diff --git a/.gitattributes b/.gitattributes
index af704cdf0c..35df6ca2fc 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -1,42 +1,58 @@
-.git*       export-ignore
-
 *           text=auto whitespace=trailing-space,space-before-tab,-indent-with-non-tab,tab-in-indent,tabwidth=4
 
-*.py        text
-*.cpp       text
-*.hpp       text
-*.cxx       text
-*.hxx       text
-*.mm        text
-*.c         text
-*.h         text
-*.i         text
-*.js        text
-*.java      text
-*.scala     text
-*.cu        text
-*.cl        text
-*.css_t     text
-*.qrc       text
-*.qss       text
-*.S         text
-*.rst       text
-*.tex       text
-*.sty       text
+.git*       text export-ignore
 
-*.aidl      text
-*.mk        text
+*.aidl         text
+*.appxmanifest text
+*.bib          text
+*.c            text
+*.cl           text
+*.conf         text
+*.cpp          text
+*.css_t        text
+*.cu           text
+*.cxx          text
+*.def          text
+*.filelist     text
+*.h            text
+*.hpp          text
+*.htm          text
+*.html         text
+*.hxx          text
+*.i            text
+*.idl          text
+*.java         text
+*.js           text
+*.mk           text
+*.mm           text
+*.plist        text
+*.properties   text
+*.py           text
+*.qrc          text
+*.qss          text
+*.S            text
+*.sbt          text
+*.scala        text
+*.sty          text
+*.tex          text
+*.txt          text
+*.xaml         text
+
+# reST underlines/overlines can look like conflict markers
+*.rst          text conflict-marker-size=80
 
 *.cmake         text whitespace=tabwidth=2
 *.cmakein       text whitespace=tabwidth=2
 *.in            text whitespace=tabwidth=2
 CMakeLists.txt  text whitespace=tabwidth=2
 
-*.png       binary
-*.jepg      binary
-*.jpg       binary
+*.avi       binary
+*.bmp       binary
 *.exr       binary
 *.ico       binary
+*.jpeg      binary
+*.jpg       binary
+*.png       binary
 
 *.a         binary
 *.so        binary
@@ -47,6 +63,7 @@ CMakeLists.txt  text whitespace=tabwidth=2
 *.pbxproj   binary
 *.vec       binary
 *.doc       binary
+*.dia       binary
 
 *.xml                      -text whitespace=cr-at-eol
 *.yml                      -text whitespace=cr-at-eol
@@ -55,9 +72,12 @@ CMakeLists.txt  text whitespace=tabwidth=2
 .cproject                  -text whitespace=cr-at-eol merge=union
 org.eclipse.jdt.core.prefs -text whitespace=cr-at-eol merge=union
 
-*.vcproj    text eol=crlf merge=union
 *.bat       text eol=crlf
 *.cmd       text eol=crlf
 *.cmd.tmpl  text eol=crlf
+*.dsp       text eol=crlf -whitespace
+*.sln       text eol=crlf -whitespace
+*.vcproj    text eol=crlf -whitespace merge=union
+*.vcxproj   text eol=crlf -whitespace merge=union
 
-*.sh        text eol=lf
\ No newline at end of file
+*.sh        text eol=lf
diff --git a/.gitignore b/.gitignore
index 0a19f3ceeb..de0707a930 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,7 +2,10 @@
 .DS_Store
 refman.rst
 OpenCV4Tegra/
+tegra/
 *.user
 .sw[a-z]
 .*.swp
 tags
+Thumbs.db
+*.autosave
diff --git a/3rdparty/.gitattributes b/3rdparty/.gitattributes
new file mode 100644
index 0000000000..562b12e16e
--- /dev/null
+++ b/3rdparty/.gitattributes
@@ -0,0 +1 @@
+* -whitespace
diff --git a/3rdparty/ffmpeg/ffmpeg_version.cmake b/3rdparty/ffmpeg/ffmpeg_version.cmake
index 3f27077d6a..7bb8bab3fb 100644
--- a/3rdparty/ffmpeg/ffmpeg_version.cmake
+++ b/3rdparty/ffmpeg/ffmpeg_version.cmake
@@ -1,4 +1,4 @@
-set(NEW_FFMPEG 1)
+set(HAVE_FFMPEG 1)
 set(HAVE_FFMPEG_CODEC 1)
 set(HAVE_FFMPEG_FORMAT 1)
 set(HAVE_FFMPEG_UTIL 1)
@@ -8,4 +8,4 @@ set(HAVE_GENTOO_FFMPEG 1)
 set(ALIASOF_libavcodec_VERSION 53.61.100)
 set(ALIASOF_libavformat_VERSION 53.32.100)
 set(ALIASOF_libavutil_VERSION 51.35.100)
-set(ALIASOF_libswscale_VERSION 2.1.100)
\ No newline at end of file
+set(ALIASOF_libswscale_VERSION 2.1.100)
diff --git a/3rdparty/ffmpeg/make.bat b/3rdparty/ffmpeg/make.bat
index 9f11b52d63..2323af9ab8 100644
--- a/3rdparty/ffmpeg/make.bat
+++ b/3rdparty/ffmpeg/make.bat
@@ -1,2 +1,2 @@
 set path=c:\dev\msys32\bin;%path% & gcc -Wall -shared -o opencv_ffmpeg.dll -O2 -x c++ -I../include -I../include/ffmpeg_ -I../../modules/highgui/src ffopencv.c -L../lib -lavformat -lavcodec -lavdevice -lswscale -lavutil -lwsock32
-set path=c:\dev\msys64\bin;%path% & gcc -m64 -Wall -shared -o opencv_ffmpeg_64.dll -O2 -x c++ -I../include -I../include/ffmpeg_ -I../../modules/highgui/src ffopencv.c -L../lib -lavformat64 -lavcodec64 -lavdevice64 -lswscale64 -lavutil64 -lavcore64 -lwsock32 -lws2_32
\ No newline at end of file
+set path=c:\dev\msys64\bin;%path% & gcc -m64 -Wall -shared -o opencv_ffmpeg_64.dll -O2 -x c++ -I../include -I../include/ffmpeg_ -I../../modules/highgui/src ffopencv.c -L../lib -lavformat64 -lavcodec64 -lavdevice64 -lswscale64 -lavutil64 -lavcore64 -lwsock32 -lws2_32
diff --git a/3rdparty/ffmpeg/readme.txt b/3rdparty/ffmpeg/readme.txt
index 1928a53039..e98b285208 100644
--- a/3rdparty/ffmpeg/readme.txt
+++ b/3rdparty/ffmpeg/readme.txt
@@ -40,5 +40,3 @@ How to update opencv_ffmpeg.dll and opencv_ffmpeg_64.dll when a new version of F
 
 8. Then, go to <opencv>\3rdparty\ffmpeg, edit make.bat
    (change paths to the actual paths to your msys32 and msys64 distributions) and then run make.bat
-
-
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r2.2.0.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r2.2.0.so
index fb34509bd5..6f28f2c554 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r2.2.0.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r2.2.0.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r2.3.3.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r2.3.3.so
index fce58145c4..010641ed15 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r2.3.3.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r2.3.3.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r3.0.1.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r3.0.1.so
index 2498763d0f..5a145b25bd 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r3.0.1.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r3.0.1.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.0.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.0.so
index 9a7047b55d..a524b743f7 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.0.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.0.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.3.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.3.so
index 98e0d9ea08..a1802f1ff3 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.3.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.0.3.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.1.1.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.1.1.so
index fe147823ec..089c7e9e2f 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.1.1.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.1.1.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.2.0.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.2.0.so
index 646ae716ec..a9ffa4b0cf 100755
Binary files a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.2.0.so and b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.2.0.so differ
diff --git a/3rdparty/lib/armeabi-v7a/libnative_camera_r4.3.0.so b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.3.0.so
new file mode 100755
index 0000000000..8ff7177ada
Binary files /dev/null and b/3rdparty/lib/armeabi-v7a/libnative_camera_r4.3.0.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r2.2.0.so b/3rdparty/lib/armeabi/libnative_camera_r2.2.0.so
index a0db0d1997..b6ce0d5a5b 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r2.2.0.so and b/3rdparty/lib/armeabi/libnative_camera_r2.2.0.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r2.3.3.so b/3rdparty/lib/armeabi/libnative_camera_r2.3.3.so
index 6f167cb0cb..635ce681a9 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r2.3.3.so and b/3rdparty/lib/armeabi/libnative_camera_r2.3.3.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r3.0.1.so b/3rdparty/lib/armeabi/libnative_camera_r3.0.1.so
index 26d286a18d..caacf39d4a 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r3.0.1.so and b/3rdparty/lib/armeabi/libnative_camera_r3.0.1.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r4.0.0.so b/3rdparty/lib/armeabi/libnative_camera_r4.0.0.so
index f29707cae3..fff4a8069b 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r4.0.0.so and b/3rdparty/lib/armeabi/libnative_camera_r4.0.0.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r4.0.3.so b/3rdparty/lib/armeabi/libnative_camera_r4.0.3.so
index 85e0320a9c..3119265558 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r4.0.3.so and b/3rdparty/lib/armeabi/libnative_camera_r4.0.3.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r4.1.1.so b/3rdparty/lib/armeabi/libnative_camera_r4.1.1.so
index 35c5141b9c..7c18baf01f 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r4.1.1.so and b/3rdparty/lib/armeabi/libnative_camera_r4.1.1.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r4.2.0.so b/3rdparty/lib/armeabi/libnative_camera_r4.2.0.so
index 816669e37c..8bb093a3d0 100755
Binary files a/3rdparty/lib/armeabi/libnative_camera_r4.2.0.so and b/3rdparty/lib/armeabi/libnative_camera_r4.2.0.so differ
diff --git a/3rdparty/lib/armeabi/libnative_camera_r4.3.0.so b/3rdparty/lib/armeabi/libnative_camera_r4.3.0.so
new file mode 100755
index 0000000000..a05f179172
Binary files /dev/null and b/3rdparty/lib/armeabi/libnative_camera_r4.3.0.so differ
diff --git a/3rdparty/lib/mips/libnative_camera_r4.0.3.so b/3rdparty/lib/mips/libnative_camera_r4.0.3.so
index 616323de96..b9500441a0 100755
Binary files a/3rdparty/lib/mips/libnative_camera_r4.0.3.so and b/3rdparty/lib/mips/libnative_camera_r4.0.3.so differ
diff --git a/3rdparty/lib/mips/libnative_camera_r4.1.1.so b/3rdparty/lib/mips/libnative_camera_r4.1.1.so
index 7ee4f25766..d11dcf036d 100755
Binary files a/3rdparty/lib/mips/libnative_camera_r4.1.1.so and b/3rdparty/lib/mips/libnative_camera_r4.1.1.so differ
diff --git a/3rdparty/lib/mips/libnative_camera_r4.2.0.so b/3rdparty/lib/mips/libnative_camera_r4.2.0.so
index 33d7745ad8..b06a6819f1 100755
Binary files a/3rdparty/lib/mips/libnative_camera_r4.2.0.so and b/3rdparty/lib/mips/libnative_camera_r4.2.0.so differ
diff --git a/3rdparty/lib/mips/libnative_camera_r4.3.0.so b/3rdparty/lib/mips/libnative_camera_r4.3.0.so
new file mode 100755
index 0000000000..844b806b9d
Binary files /dev/null and b/3rdparty/lib/mips/libnative_camera_r4.3.0.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r2.3.3.so b/3rdparty/lib/x86/libnative_camera_r2.3.3.so
index d40409f222..0dd8904ac9 100755
Binary files a/3rdparty/lib/x86/libnative_camera_r2.3.3.so and b/3rdparty/lib/x86/libnative_camera_r2.3.3.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r3.0.1.so b/3rdparty/lib/x86/libnative_camera_r3.0.1.so
index 221b833a5b..105a19d0c5 100755
Binary files a/3rdparty/lib/x86/libnative_camera_r3.0.1.so and b/3rdparty/lib/x86/libnative_camera_r3.0.1.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r4.0.3.so b/3rdparty/lib/x86/libnative_camera_r4.0.3.so
index 786d6dccbf..b01a4bd28c 100755
Binary files a/3rdparty/lib/x86/libnative_camera_r4.0.3.so and b/3rdparty/lib/x86/libnative_camera_r4.0.3.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r4.1.1.so b/3rdparty/lib/x86/libnative_camera_r4.1.1.so
index 8ec6cb74d9..a59ae39b3c 100755
Binary files a/3rdparty/lib/x86/libnative_camera_r4.1.1.so and b/3rdparty/lib/x86/libnative_camera_r4.1.1.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r4.2.0.so b/3rdparty/lib/x86/libnative_camera_r4.2.0.so
index 7fe74d21ab..b90b826448 100755
Binary files a/3rdparty/lib/x86/libnative_camera_r4.2.0.so and b/3rdparty/lib/x86/libnative_camera_r4.2.0.so differ
diff --git a/3rdparty/lib/x86/libnative_camera_r4.3.0.so b/3rdparty/lib/x86/libnative_camera_r4.3.0.so
new file mode 100755
index 0000000000..6607e5da81
Binary files /dev/null and b/3rdparty/lib/x86/libnative_camera_r4.3.0.so differ
diff --git a/3rdparty/libjasper/CMakeLists.txt b/3rdparty/libjasper/CMakeLists.txt
index 42855e2a6d..7a70a19cf0 100644
--- a/3rdparty/libjasper/CMakeLists.txt
+++ b/3rdparty/libjasper/CMakeLists.txt
@@ -23,7 +23,9 @@ if(WIN32 AND NOT MINGW)
   add_definitions(-DJAS_WIN_MSVC_BUILD)
 endif(WIN32 AND NOT MINGW)
 
-ocv_warnings_disable(CMAKE_C_FLAGS -Wno-implicit-function-declaration -Wno-uninitialized -Wmissing-prototypes -Wmissing-declarations -Wunused -Wshadow -Wsign-compare)
+ocv_warnings_disable(CMAKE_C_FLAGS -Wno-implicit-function-declaration -Wno-uninitialized -Wmissing-prototypes
+                                   -Wno-unused-but-set-parameter -Wmissing-declarations -Wunused -Wshadow
+                                   -Wsign-compare -Wstrict-overflow)
 ocv_warnings_disable(CMAKE_C_FLAGS -Wunused-parameter) # clang
 ocv_warnings_disable(CMAKE_C_FLAGS /wd4013 /wd4018 /wd4101 /wd4244 /wd4267 /wd4715) # vs2005
 
@@ -47,4 +49,3 @@ endif()
 if(NOT BUILD_SHARED_LIBS)
   install(TARGETS ${JASPER_LIBRARY} ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT main)
 endif()
-
diff --git a/3rdparty/libpng/CMakeLists.txt b/3rdparty/libpng/CMakeLists.txt
index d47dd53fcd..141c9d1518 100644
--- a/3rdparty/libpng/CMakeLists.txt
+++ b/3rdparty/libpng/CMakeLists.txt
@@ -9,7 +9,7 @@ else()
   project(${PNG_LIBRARY})
 endif()
 
-ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}" ${ZLIB_INCLUDE_DIR})
+ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}" ${ZLIB_INCLUDE_DIRS})
 
 file(GLOB lib_srcs *.c)
 file(GLOB lib_hdrs *.h)
@@ -29,8 +29,12 @@ if(MSVC)
   add_definitions(-D_CRT_SECURE_NO_DEPRECATE)
 endif(MSVC)
 
+if (HAVE_WINRT)
+  add_definitions(-DHAVE_WINRT)
+endif()
+
 add_library(${PNG_LIBRARY} STATIC ${lib_srcs} ${lib_hdrs})
-target_link_libraries(${PNG_LIBRARY} ${ZLIB_LIBRARY})
+target_link_libraries(${PNG_LIBRARY} ${ZLIB_LIBRARIES})
 
 if(UNIX)
   if(CMAKE_COMPILER_IS_GNUCXX OR CV_ICC)
diff --git a/3rdparty/libpng/opencv-libpng.patch b/3rdparty/libpng/opencv-libpng.patch
new file mode 100644
index 0000000000..6ca96392a0
--- /dev/null
+++ b/3rdparty/libpng/opencv-libpng.patch
@@ -0,0 +1,22 @@
+diff --git a/3rdparty/libpng/pngpriv.h b/3rdparty/libpng/pngpriv.h
+index 07b2b0b..e7824b8 100644
+--- a/3rdparty/libpng/pngpriv.h
++++ b/3rdparty/libpng/pngpriv.h
+@@ -360,7 +360,7 @@ typedef PNG_CONST png_uint_16p FAR * png_const_uint_16pp;
+ 
+ /* Memory model/platform independent fns */
+ #ifndef PNG_ABORT
+-#  ifdef _WINDOWS_
++#  if defined(_WINDOWS_) && !defined(HAVE_WINRT)
+ #    define PNG_ABORT() ExitProcess(0)
+ #  else
+ #    define PNG_ABORT() abort()
+@@ -378,7 +378,7 @@ typedef PNG_CONST png_uint_16p FAR * png_const_uint_16pp;
+ #  define png_memcpy  _fmemcpy
+ #  define png_memset  _fmemset
+ #else
+-#  ifdef _WINDOWS_  /* Favor Windows over C runtime fns */
++#  if defined(_WINDOWS_) && !defined(HAVE_WINRT)  /* Favor Windows over C runtime fns */
+ #    define CVT_PTR(ptr)         (ptr)
+ #    define CVT_PTR_NOCHECK(ptr) (ptr)
+ #    define png_strlen  lstrlenA
diff --git a/3rdparty/libpng/pngpriv.h b/3rdparty/libpng/pngpriv.h
index 07b2b0b05b..e7824b839e 100644
--- a/3rdparty/libpng/pngpriv.h
+++ b/3rdparty/libpng/pngpriv.h
@@ -360,7 +360,7 @@ typedef PNG_CONST png_uint_16p FAR * png_const_uint_16pp;
 
 /* Memory model/platform independent fns */
 #ifndef PNG_ABORT
-#  ifdef _WINDOWS_
+#  if defined(_WINDOWS_) && !defined(HAVE_WINRT)
 #    define PNG_ABORT() ExitProcess(0)
 #  else
 #    define PNG_ABORT() abort()
@@ -378,7 +378,7 @@ typedef PNG_CONST png_uint_16p FAR * png_const_uint_16pp;
 #  define png_memcpy  _fmemcpy
 #  define png_memset  _fmemset
 #else
-#  ifdef _WINDOWS_  /* Favor Windows over C runtime fns */
+#  if defined(_WINDOWS_) && !defined(HAVE_WINRT)  /* Favor Windows over C runtime fns */
 #    define CVT_PTR(ptr)         (ptr)
 #    define CVT_PTR_NOCHECK(ptr) (ptr)
 #    define png_strlen  lstrlenA
diff --git a/3rdparty/libtiff/CMakeLists.txt b/3rdparty/libtiff/CMakeLists.txt
index 46fef61c7d..cea2f906fd 100644
--- a/3rdparty/libtiff/CMakeLists.txt
+++ b/3rdparty/libtiff/CMakeLists.txt
@@ -17,14 +17,14 @@ check_include_file(string.h HAVE_STRING_H)
 check_include_file(sys/types.h HAVE_SYS_TYPES_H)
 check_include_file(unistd.h HAVE_UNISTD_H)
 
-if(WIN32)
+if(WIN32 AND NOT HAVE_WINRT)
   set(USE_WIN32_FILEIO 1)
 endif()
 
 configure_file("${CMAKE_CURRENT_SOURCE_DIR}/tif_config.h.cmakein"
                "${CMAKE_CURRENT_BINARY_DIR}/tif_config.h" @ONLY)
 
-ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}" "${CMAKE_CURRENT_BINARY_DIR}" ${ZLIB_INCLUDE_DIR})
+ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}" "${CMAKE_CURRENT_BINARY_DIR}" ${ZLIB_INCLUDE_DIRS})
 
 set(lib_srcs
     tif_aux.c
@@ -79,14 +79,12 @@ set(lib_srcs
     "${CMAKE_CURRENT_BINARY_DIR}/tif_config.h"
     )
 
-if(UNIX)
+if(WIN32 AND NOT HAVE_WINRT)
+  list(APPEND lib_srcs tif_win32.c)
+else()
   list(APPEND lib_srcs tif_unix.c)
 endif()
-
-if(WIN32)
-  list(APPEND lib_srcs tif_win32.c)
-endif(WIN32)
-
+  
 ocv_warnings_disable(CMAKE_C_FLAGS -Wno-unused-but-set-variable -Wmissing-prototypes -Wmissing-declarations -Wundef -Wunused -Wsign-compare
                                    -Wcast-align -Wshadow -Wno-maybe-uninitialized -Wno-pointer-to-int-cast -Wno-int-to-pointer-cast)
 ocv_warnings_disable(CMAKE_C_FLAGS -Wunused-parameter) # clang
diff --git a/3rdparty/libtiff/tif_config.h.cmakein b/3rdparty/libtiff/tif_config.h.cmakein
index 1e6bc0454c..182f2833d1 100644
--- a/3rdparty/libtiff/tif_config.h.cmakein
+++ b/3rdparty/libtiff/tif_config.h.cmakein
@@ -168,4 +168,3 @@
 
 /* Support Deflate compression */
 #define ZIP_SUPPORT 1
-
diff --git a/3rdparty/libwebp/dec/alpha.c b/3rdparty/libwebp/dec/alpha.c
index 6e65de9030..b5e68919ca 100644
--- a/3rdparty/libwebp/dec/alpha.c
+++ b/3rdparty/libwebp/dec/alpha.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Alpha-plane decompression.
@@ -13,27 +15,17 @@
 #include "./vp8i.h"
 #include "./vp8li.h"
 #include "../utils/filters.h"
-#include "../utils/quant_levels.h"
+#include "../utils/quant_levels_dec.h"
 #include "../webp/format_constants.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
 
-// TODO(skal): move to dsp/ ?
-static void CopyPlane(const uint8_t* src, int src_stride,
-                      uint8_t* dst, int dst_stride, int width, int height) {
-  while (height-- > 0) {
-    memcpy(dst, src, width);
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
 //------------------------------------------------------------------------------
 // Decodes the compressed data 'data' of size 'data_size' into the 'output'.
 // The 'output' buffer should be pre-allocated and must be of the same
-// dimension 'height'x'stride', as that of the image.
+// dimension 'height'x'width', as that of the image.
 //
 // Returns 1 on successfully decoding the compressed alpha and
 //         0 if either:
@@ -41,17 +33,16 @@ static void CopyPlane(const uint8_t* src, int src_stride,
 //           error returned by appropriate compression method.
 
 static int DecodeAlpha(const uint8_t* data, size_t data_size,
-                       int width, int height, int stride, uint8_t* output) {
-  uint8_t* decoded_data = NULL;
-  const size_t decoded_size = height * width;
-  uint8_t* unfiltered_data = NULL;
+                       int width, int height, uint8_t* output) {
   WEBP_FILTER_TYPE filter;
   int pre_processing;
   int rsrv;
   int ok = 0;
   int method;
+  const uint8_t* const alpha_data = data + ALPHA_HEADER_LEN;
+  const size_t alpha_data_size = data_size - ALPHA_HEADER_LEN;
 
-  assert(width > 0 && height > 0 && stride >= width);
+  assert(width > 0 && height > 0);
   assert(data != NULL && output != NULL);
 
   if (data_size <= ALPHA_HEADER_LEN) {
@@ -71,44 +62,26 @@ static int DecodeAlpha(const uint8_t* data, size_t data_size,
   }
 
   if (method == ALPHA_NO_COMPRESSION) {
-    ok = (data_size >= decoded_size);
-    decoded_data = (uint8_t*)data + ALPHA_HEADER_LEN;
+    const size_t alpha_decoded_size = height * width;
+    ok = (alpha_data_size >= alpha_decoded_size);
+    if (ok) memcpy(output, alpha_data, alpha_decoded_size);
   } else {
-    decoded_data = (uint8_t*)malloc(decoded_size);
-    if (decoded_data == NULL) return 0;
-    ok = VP8LDecodeAlphaImageStream(width, height,
-                                    data + ALPHA_HEADER_LEN,
-                                    data_size - ALPHA_HEADER_LEN,
-                                    decoded_data);
+    ok = VP8LDecodeAlphaImageStream(width, height, alpha_data, alpha_data_size,
+                                    output);
   }
 
   if (ok) {
-    WebPFilterFunc unfilter_func = WebPUnfilters[filter];
+    WebPUnfilterFunc unfilter_func = WebPUnfilters[filter];
     if (unfilter_func != NULL) {
-      unfiltered_data = (uint8_t*)malloc(decoded_size);
-      if (unfiltered_data == NULL) {
-        ok = 0;
-        goto Error;
-      }
       // TODO(vikas): Implement on-the-fly decoding & filter mechanism to decode
       // and apply filter per image-row.
-      unfilter_func(decoded_data, width, height, 1, width, unfiltered_data);
-      // Construct raw_data (height x stride) from alpha data (height x width).
-      CopyPlane(unfiltered_data, width, output, stride, width, height);
-      free(unfiltered_data);
-    } else {
-      // Construct raw_data (height x stride) from alpha data (height x width).
-      CopyPlane(decoded_data, width, output, stride, width, height);
+      unfilter_func(width, height, width, output);
     }
     if (pre_processing == ALPHA_PREPROCESSED_LEVELS) {
-      ok = DequantizeLevels(decoded_data, width, height);
+      ok = DequantizeLevels(output, width, height);
     }
   }
 
- Error:
-  if (method != ALPHA_NO_COMPRESSION) {
-    free(decoded_data);
-  }
   return ok;
 }
 
@@ -116,23 +89,25 @@ static int DecodeAlpha(const uint8_t* data, size_t data_size,
 
 const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec,
                                       int row, int num_rows) {
-  const int stride = dec->pic_hdr_.width_;
+  const int width = dec->pic_hdr_.width_;
+  const int height = dec->pic_hdr_.height_;
 
-  if (row < 0 || num_rows < 0 || row + num_rows > dec->pic_hdr_.height_) {
+  if (row < 0 || num_rows < 0 || row + num_rows > height) {
     return NULL;    // sanity check.
   }
 
   if (row == 0) {
     // Decode everything during the first call.
+    assert(!dec->is_alpha_decoded_);
     if (!DecodeAlpha(dec->alpha_data_, (size_t)dec->alpha_data_size_,
-                     dec->pic_hdr_.width_, dec->pic_hdr_.height_, stride,
-                     dec->alpha_plane_)) {
+                     width, height, dec->alpha_plane_)) {
       return NULL;  // Error.
     }
+    dec->is_alpha_decoded_ = 1;
   }
 
   // Return a pointer to the current decoded row.
-  return dec->alpha_plane_ + row * stride;
+  return dec->alpha_plane_ + row * width;
 }
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/dec/buffer.c b/3rdparty/libwebp/dec/buffer.c
index c159f6f248..3855715239 100644
--- a/3rdparty/libwebp/dec/buffer.c
+++ b/3rdparty/libwebp/dec/buffer.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Everything about WebPDecBuffer
diff --git a/3rdparty/libwebp/dec/decode_vp8.h b/3rdparty/libwebp/dec/decode_vp8.h
index 12c77bcbf6..acdb15aaa9 100644
--- a/3rdparty/libwebp/dec/decode_vp8.h
+++ b/3rdparty/libwebp/dec/decode_vp8.h
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  Low-level API for VP8 decoder
diff --git a/3rdparty/libwebp/dec/frame.c b/3rdparty/libwebp/dec/frame.c
index 9c91a48e17..5f6a7d9823 100644
--- a/3rdparty/libwebp/dec/frame.c
+++ b/3rdparty/libwebp/dec/frame.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Frame-reconstruction function. Memory allocation.
@@ -97,53 +99,50 @@ static void FilterRow(const VP8Decoder* const dec) {
 }
 
 //------------------------------------------------------------------------------
+// Precompute the filtering strength for each segment and each i4x4/i16x16 mode.
 
-void VP8StoreBlock(VP8Decoder* const dec) {
+static void PrecomputeFilterStrengths(VP8Decoder* const dec) {
   if (dec->filter_type_ > 0) {
-    VP8FInfo* const info = dec->f_info_ + dec->mb_x_;
-    const int skip = dec->mb_info_[dec->mb_x_].skip_;
-    int level = dec->filter_levels_[dec->segment_];
-    if (dec->filter_hdr_.use_lf_delta_) {
-      // TODO(skal): only CURRENT is handled for now.
-      level += dec->filter_hdr_.ref_lf_delta_[0];
-      if (dec->is_i4x4_) {
-        level += dec->filter_hdr_.mode_lf_delta_[0];
-      }
-    }
-    level = (level < 0) ? 0 : (level > 63) ? 63 : level;
-    info->f_level_ = level;
-
-    if (dec->filter_hdr_.sharpness_ > 0) {
-      if (dec->filter_hdr_.sharpness_ > 4) {
-        level >>= 2;
+    int s;
+    const VP8FilterHeader* const hdr = &dec->filter_hdr_;
+    for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
+      int i4x4;
+      // First, compute the initial level
+      int base_level;
+      if (dec->segment_hdr_.use_segment_) {
+        base_level = dec->segment_hdr_.filter_strength_[s];
+        if (!dec->segment_hdr_.absolute_delta_) {
+          base_level += hdr->level_;
+        }
       } else {
-        level >>= 1;
+        base_level = hdr->level_;
       }
-      if (level > 9 - dec->filter_hdr_.sharpness_) {
-        level = 9 - dec->filter_hdr_.sharpness_;
-      }
-    }
+      for (i4x4 = 0; i4x4 <= 1; ++i4x4) {
+        VP8FInfo* const info = &dec->fstrengths_[s][i4x4];
+        int level = base_level;
+        if (hdr->use_lf_delta_) {
+          // TODO(skal): only CURRENT is handled for now.
+          level += hdr->ref_lf_delta_[0];
+          if (i4x4) {
+            level += hdr->mode_lf_delta_[0];
+          }
+        }
+        level = (level < 0) ? 0 : (level > 63) ? 63 : level;
+        info->f_level_ = level;
 
-    info->f_ilevel_ = (level < 1) ? 1 : level;
-    info->f_inner_ = (!skip || dec->is_i4x4_);
-  }
-  {
-    // Transfer samples to row cache
-    int y;
-    const int y_offset = dec->cache_id_ * 16 * dec->cache_y_stride_;
-    const int uv_offset = dec->cache_id_ * 8 * dec->cache_uv_stride_;
-    uint8_t* const ydst = dec->cache_y_ + dec->mb_x_ * 16 + y_offset;
-    uint8_t* const udst = dec->cache_u_ + dec->mb_x_ * 8 + uv_offset;
-    uint8_t* const vdst = dec->cache_v_ + dec->mb_x_ * 8 + uv_offset;
-    for (y = 0; y < 16; ++y) {
-      memcpy(ydst + y * dec->cache_y_stride_,
-             dec->yuv_b_ + Y_OFF + y * BPS, 16);
-    }
-    for (y = 0; y < 8; ++y) {
-      memcpy(udst + y * dec->cache_uv_stride_,
-           dec->yuv_b_ + U_OFF + y * BPS, 8);
-      memcpy(vdst + y * dec->cache_uv_stride_,
-           dec->yuv_b_ + V_OFF + y * BPS, 8);
+        if (hdr->sharpness_ > 0) {
+          if (hdr->sharpness_ > 4) {
+            level >>= 2;
+          } else {
+            level >>= 1;
+          }
+          if (level > 9 - hdr->sharpness_) {
+            level = 9 - hdr->sharpness_;
+          }
+        }
+        info->f_ilevel_ = (level < 1) ? 1 : level;
+        info->f_inner_ = 0;
+      }
     }
   }
 }
@@ -339,6 +338,7 @@ VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io) {
       dec->br_mb_y_ = dec->mb_h_;
     }
   }
+  PrecomputeFilterStrengths(dec);
   return VP8_STATUS_OK;
 }
 
@@ -496,6 +496,7 @@ static int AllocateMemory(VP8Decoder* const dec) {
   // alpha plane
   dec->alpha_plane_ = alpha_size ? (uint8_t*)mem : NULL;
   mem += alpha_size;
+  assert(mem <= (uint8_t*)dec->mem_ + dec->mem_size_);
 
   // note: left-info is initialized once for all.
   memset(dec->mb_info_ - 1, 0, mb_info_size);
@@ -551,6 +552,7 @@ static WEBP_INLINE void Copy32b(uint8_t* dst, uint8_t* src) {
 }
 
 void VP8ReconstructBlock(VP8Decoder* const dec) {
+  int j;
   uint8_t* const y_dst = dec->yuv_b_ + Y_OFF;
   uint8_t* const u_dst = dec->yuv_b_ + U_OFF;
   uint8_t* const v_dst = dec->yuv_b_ + V_OFF;
@@ -558,7 +560,6 @@ void VP8ReconstructBlock(VP8Decoder* const dec) {
   // Rotate in the left samples from previously decoded block. We move four
   // pixels at a time for alignment reason, and because of in-loop filter.
   if (dec->mb_x_ > 0) {
-    int j;
     for (j = -1; j < 16; ++j) {
       Copy32b(&y_dst[j * BPS - 4], &y_dst[j * BPS + 12]);
     }
@@ -567,7 +568,6 @@ void VP8ReconstructBlock(VP8Decoder* const dec) {
       Copy32b(&v_dst[j * BPS - 4], &v_dst[j * BPS + 4]);
     }
   } else {
-    int j;
     for (j = 0; j < 16; ++j) {
       y_dst[j * BPS - 1] = 129;
     }
@@ -670,6 +670,21 @@ void VP8ReconstructBlock(VP8Decoder* const dec) {
       }
     }
   }
+  // Transfer reconstructed samples from yuv_b_ cache to final destination.
+  {
+    const int y_offset = dec->cache_id_ * 16 * dec->cache_y_stride_;
+    const int uv_offset = dec->cache_id_ * 8 * dec->cache_uv_stride_;
+    uint8_t* const y_out = dec->cache_y_ + dec->mb_x_ * 16 + y_offset;
+    uint8_t* const u_out = dec->cache_u_ + dec->mb_x_ * 8 + uv_offset;
+    uint8_t* const v_out = dec->cache_v_ + dec->mb_x_ * 8 + uv_offset;
+    for (j = 0; j < 16; ++j) {
+      memcpy(y_out + j * dec->cache_y_stride_, y_dst + j * BPS, 16);
+    }
+    for (j = 0; j < 8; ++j) {
+      memcpy(u_out + j * dec->cache_uv_stride_, u_dst + j * BPS, 8);
+      memcpy(v_out + j * dec->cache_uv_stride_, v_dst + j * BPS, 8);
+    }
+  }
 }
 
 //------------------------------------------------------------------------------
diff --git a/3rdparty/libwebp/dec/idec.c b/3rdparty/libwebp/dec/idec.c
index 7df790ced8..5fbf49aafe 100644
--- a/3rdparty/libwebp/dec/idec.c
+++ b/3rdparty/libwebp/dec/idec.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Incremental decoding
@@ -97,6 +99,23 @@ static WEBP_INLINE size_t MemDataSize(const MemBuffer* mem) {
   return (mem->end_ - mem->start_);
 }
 
+// Check if we need to preserve the compressed alpha data, as it may not have
+// been decoded yet.
+static int NeedCompressedAlpha(const WebPIDecoder* const idec) {
+  if (idec->state_ == STATE_PRE_VP8) {
+    // We haven't parsed the headers yet, so we don't know whether the image is
+    // lossy or lossless. This also means that we haven't parsed the ALPH chunk.
+    return 0;
+  }
+  if (idec->is_lossless_) {
+    return 0;  // ALPH chunk is not present for lossless images.
+  } else {
+    const VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
+    assert(dec != NULL);  // Must be true as idec->state_ != STATE_PRE_VP8.
+    return (dec->alpha_data_ != NULL) && !dec->is_alpha_decoded_;
+  }
+}
+
 static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) {
   MemBuffer* const mem = &idec->mem_;
   const uint8_t* const new_base = mem->buf_ + mem->start_;
@@ -122,6 +141,7 @@ static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) {
       }
       assert(last_part >= 0);
       dec->parts_[last_part].buf_end_ = mem->buf_ + mem->end_;
+      if (NeedCompressedAlpha(idec)) dec->alpha_data_ += offset;
     } else {    // Resize lossless bitreader
       VP8LDecoder* const dec = (VP8LDecoder*)idec->dec_;
       VP8LBitReaderSetBuffer(&dec->br_, new_base, MemDataSize(mem));
@@ -133,8 +153,12 @@ static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) {
 // size if required and also updates VP8BitReader's if new memory is allocated.
 static int AppendToMemBuffer(WebPIDecoder* const idec,
                              const uint8_t* const data, size_t data_size) {
+  VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
   MemBuffer* const mem = &idec->mem_;
-  const uint8_t* const old_base = mem->buf_ + mem->start_;
+  const int need_compressed_alpha = NeedCompressedAlpha(idec);
+  const uint8_t* const old_start = mem->buf_ + mem->start_;
+  const uint8_t* const old_base =
+      need_compressed_alpha ? dec->alpha_data_ : old_start;
   assert(mem->mode_ == MEM_MODE_APPEND);
   if (data_size > MAX_CHUNK_PAYLOAD) {
     // security safeguard: trying to allocate more than what the format
@@ -143,7 +167,8 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
   }
 
   if (mem->end_ + data_size > mem->buf_size_) {  // Need some free memory
-    const size_t current_size = MemDataSize(mem);
+    const size_t new_mem_start = old_start - old_base;
+    const size_t current_size = MemDataSize(mem) + new_mem_start;
     const uint64_t new_size = (uint64_t)current_size + data_size;
     const uint64_t extra_size = (new_size + CHUNK_SIZE - 1) & ~(CHUNK_SIZE - 1);
     uint8_t* const new_buf =
@@ -153,7 +178,7 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
     free(mem->buf_);
     mem->buf_ = new_buf;
     mem->buf_size_ = (size_t)extra_size;
-    mem->start_ = 0;
+    mem->start_ = new_mem_start;
     mem->end_ = current_size;
   }
 
@@ -161,14 +186,15 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
   mem->end_ += data_size;
   assert(mem->end_ <= mem->buf_size_);
 
-  DoRemap(idec, mem->buf_ + mem->start_ - old_base);
+  DoRemap(idec, mem->buf_ + mem->start_ - old_start);
   return 1;
 }
 
 static int RemapMemBuffer(WebPIDecoder* const idec,
                           const uint8_t* const data, size_t data_size) {
   MemBuffer* const mem = &idec->mem_;
-  const uint8_t* const old_base = mem->buf_ + mem->start_;
+  const uint8_t* const old_buf = mem->buf_;
+  const uint8_t* const old_start = old_buf + mem->start_;
   assert(mem->mode_ == MEM_MODE_MAP);
 
   if (data_size < mem->buf_size_) return 0;  // can't remap to a shorter buffer!
@@ -176,7 +202,7 @@ static int RemapMemBuffer(WebPIDecoder* const idec,
   mem->buf_ = (uint8_t*)data;
   mem->end_ = mem->buf_size_ = data_size;
 
-  DoRemap(idec, mem->buf_ + mem->start_ - old_base);
+  DoRemap(idec, mem->buf_ + mem->start_ - old_start);
   return 1;
 }
 
@@ -425,9 +451,8 @@ static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) {
         }
         return VP8_STATUS_SUSPENDED;
       }
+      // Reconstruct and emit samples.
       VP8ReconstructBlock(dec);
-      // Store data and save block's filtering params
-      VP8StoreBlock(dec);
 
       // Release buffer only if there is only one partition
       if (dec->num_parts_ == 1) {
@@ -596,12 +621,22 @@ void WebPIDelete(WebPIDecoder* idec) {
 
 WebPIDecoder* WebPINewRGB(WEBP_CSP_MODE mode, uint8_t* output_buffer,
                           size_t output_buffer_size, int output_stride) {
+  const int is_external_memory = (output_buffer != NULL);
   WebPIDecoder* idec;
+
   if (mode >= MODE_YUV) return NULL;
+  if (!is_external_memory) {    // Overwrite parameters to sane values.
+    output_buffer_size = 0;
+    output_stride = 0;
+  } else {  // A buffer was passed. Validate the other params.
+    if (output_stride == 0 || output_buffer_size == 0) {
+      return NULL;   // invalid parameter.
+    }
+  }
   idec = WebPINewDecoder(NULL);
   if (idec == NULL) return NULL;
   idec->output_.colorspace = mode;
-  idec->output_.is_external_memory = 1;
+  idec->output_.is_external_memory = is_external_memory;
   idec->output_.u.RGBA.rgba = output_buffer;
   idec->output_.u.RGBA.stride = output_stride;
   idec->output_.u.RGBA.size = output_buffer_size;
@@ -612,10 +647,30 @@ WebPIDecoder* WebPINewYUVA(uint8_t* luma, size_t luma_size, int luma_stride,
                            uint8_t* u, size_t u_size, int u_stride,
                            uint8_t* v, size_t v_size, int v_stride,
                            uint8_t* a, size_t a_size, int a_stride) {
-  WebPIDecoder* const idec = WebPINewDecoder(NULL);
+  const int is_external_memory = (luma != NULL);
+  WebPIDecoder* idec;
+  WEBP_CSP_MODE colorspace;
+
+  if (!is_external_memory) {    // Overwrite parameters to sane values.
+    luma_size = u_size = v_size = a_size = 0;
+    luma_stride = u_stride = v_stride = a_stride = 0;
+    u = v = a = NULL;
+    colorspace = MODE_YUVA;
+  } else {  // A luma buffer was passed. Validate the other parameters.
+    if (u == NULL || v == NULL) return NULL;
+    if (luma_size == 0 || u_size == 0 || v_size == 0) return NULL;
+    if (luma_stride == 0 || u_stride == 0 || v_stride == 0) return NULL;
+    if (a != NULL) {
+      if (a_size == 0 || a_stride == 0) return NULL;
+    }
+    colorspace = (a == NULL) ? MODE_YUV : MODE_YUVA;
+  }
+
+  idec = WebPINewDecoder(NULL);
   if (idec == NULL) return NULL;
-  idec->output_.colorspace = (a == NULL) ? MODE_YUV : MODE_YUVA;
-  idec->output_.is_external_memory = 1;
+
+  idec->output_.colorspace = colorspace;
+  idec->output_.is_external_memory = is_external_memory;
   idec->output_.u.YUVA.y = luma;
   idec->output_.u.YUVA.y_stride = luma_stride;
   idec->output_.u.YUVA.y_size = luma_size;
diff --git a/3rdparty/libwebp/dec/io.c b/3rdparty/libwebp/dec/io.c
index 594804c2e6..63810b4472 100644
--- a/3rdparty/libwebp/dec/io.c
+++ b/3rdparty/libwebp/dec/io.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // functions for sample output.
diff --git a/3rdparty/libwebp/dec/layer.c b/3rdparty/libwebp/dec/layer.c
index a3a5bdcfe8..9a4b2d9032 100644
--- a/3rdparty/libwebp/dec/layer.c
+++ b/3rdparty/libwebp/dec/layer.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Enhancement layer (for YUV444/422)
diff --git a/3rdparty/libwebp/dec/quant.c b/3rdparty/libwebp/dec/quant.c
index d54097af0d..a4cc693db7 100644
--- a/3rdparty/libwebp/dec/quant.c
+++ b/3rdparty/libwebp/dec/quant.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Quantizer initialization
diff --git a/3rdparty/libwebp/dec/tree.c b/3rdparty/libwebp/dec/tree.c
index 82484e4c55..3f02efe43f 100644
--- a/3rdparty/libwebp/dec/tree.c
+++ b/3rdparty/libwebp/dec/tree.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Coding trees and probas
diff --git a/3rdparty/libwebp/dec/vp8.c b/3rdparty/libwebp/dec/vp8.c
index b0ccfa2a06..8632e48e57 100644
--- a/3rdparty/libwebp/dec/vp8.c
+++ b/3rdparty/libwebp/dec/vp8.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // main entry for the decoder
@@ -236,20 +238,6 @@ static int ParseFilterHeader(VP8BitReader* br, VP8Decoder* const dec) {
     }
   }
   dec->filter_type_ = (hdr->level_ == 0) ? 0 : hdr->simple_ ? 1 : 2;
-  if (dec->filter_type_ > 0) {    // precompute filter levels per segment
-    if (dec->segment_hdr_.use_segment_) {
-      int s;
-      for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
-        int strength = dec->segment_hdr_.filter_strength_[s];
-        if (!dec->segment_hdr_.absolute_delta_) {
-          strength += hdr->level_;
-        }
-        dec->filter_levels_[s] = strength;
-      }
-    } else {
-      dec->filter_levels_[0] = hdr->level_;
-    }
-  }
   return !br->eof_;
 }
 
@@ -458,7 +446,7 @@ int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io) {
 //------------------------------------------------------------------------------
 // Residual decoding (Paragraph 13.2 / 13.3)
 
-static const uint8_t kBands[16 + 1] = {
+static const int kBands[16 + 1] = {
   0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7,
   0  // extra entry as sentinel
 };
@@ -474,6 +462,39 @@ static const uint8_t kZigzag[16] = {
 };
 
 typedef const uint8_t (*ProbaArray)[NUM_CTX][NUM_PROBAS];  // for const-casting
+typedef const uint8_t (*ProbaCtxArray)[NUM_PROBAS];
+
+// See section 13-2: http://tools.ietf.org/html/rfc6386#section-13.2
+static int GetLargeValue(VP8BitReader* const br, const uint8_t* const p) {
+  int v;
+  if (!VP8GetBit(br, p[3])) {
+    if (!VP8GetBit(br, p[4])) {
+      v = 2;
+    } else {
+      v = 3 + VP8GetBit(br, p[5]);
+    }
+  } else {
+    if (!VP8GetBit(br, p[6])) {
+      if (!VP8GetBit(br, p[7])) {
+        v = 5 + VP8GetBit(br, 159);
+      } else {
+        v = 7 + 2 * VP8GetBit(br, 165);
+        v += VP8GetBit(br, 145);
+      }
+    } else {
+      const uint8_t* tab;
+      const int bit1 = VP8GetBit(br, p[8]);
+      const int bit0 = VP8GetBit(br, p[9 + bit1]);
+      const int cat = 2 * bit1 + bit0;
+      v = 0;
+      for (tab = kCat3456[cat]; *tab; ++tab) {
+        v += v + VP8GetBit(br, *tab);
+      }
+      v += 3 + (8 << cat);
+    }
+  }
+  return v;
+}
 
 // Returns the position of the last non-zero coeff plus one
 // (and 0 if there's no coeff at all)
@@ -484,54 +505,26 @@ static int GetCoeffs(VP8BitReader* const br, ProbaArray prob,
   if (!VP8GetBit(br, p[0])) {   // first EOB is more a 'CBP' bit.
     return 0;
   }
-  while (1) {
-    ++n;
+  for (; n < 16; ++n) {
+    const ProbaCtxArray p_ctx = prob[kBands[n + 1]];
     if (!VP8GetBit(br, p[1])) {
-      p = prob[kBands[n]][0];
+      p = p_ctx[0];
     } else {  // non zero coeff
-      int v, j;
+      int v;
       if (!VP8GetBit(br, p[2])) {
-        p = prob[kBands[n]][1];
         v = 1;
+        p = p_ctx[1];
       } else {
-        if (!VP8GetBit(br, p[3])) {
-          if (!VP8GetBit(br, p[4])) {
-            v = 2;
-          } else {
-            v = 3 + VP8GetBit(br, p[5]);
-          }
-        } else {
-          if (!VP8GetBit(br, p[6])) {
-            if (!VP8GetBit(br, p[7])) {
-              v = 5 + VP8GetBit(br, 159);
-            } else {
-              v = 7 + 2 * VP8GetBit(br, 165);
-              v += VP8GetBit(br, 145);
-            }
-          } else {
-            const uint8_t* tab;
-            const int bit1 = VP8GetBit(br, p[8]);
-            const int bit0 = VP8GetBit(br, p[9 + bit1]);
-            const int cat = 2 * bit1 + bit0;
-            v = 0;
-            for (tab = kCat3456[cat]; *tab; ++tab) {
-              v += v + VP8GetBit(br, *tab);
-            }
-            v += 3 + (8 << cat);
-          }
-        }
-        p = prob[kBands[n]][2];
+        v = GetLargeValue(br, p);
+        p = p_ctx[2];
       }
-      j = kZigzag[n - 1];
-      out[j] = VP8GetSigned(br, v) * dq[j > 0];
-      if (n == 16 || !VP8GetBit(br, p[0])) {   // EOB
-        return n;
+      out[kZigzag[n]] = VP8GetSigned(br, v) * dq[n > 0];
+      if (n < 15 && !VP8GetBit(br, p[0])) {   // EOB
+        return n + 1;
       }
     }
-    if (n == 16) {
-      return 16;
-    }
   }
+  return 16;
 }
 
 // Alias-safe way of converting 4bytes to 32bits.
@@ -670,6 +663,12 @@ int VP8DecodeMB(VP8Decoder* const dec, VP8BitReader* const token_br) {
     dec->non_zero_ac_ = 0;
   }
 
+  if (dec->filter_type_ > 0) {  // store filter info
+    VP8FInfo* const finfo = dec->f_info_ + dec->mb_x_;
+    *finfo = dec->fstrengths_[dec->segment_][dec->is_i4x4_];
+    finfo->f_inner_ = (!info->skip_ || dec->is_i4x4_);
+  }
+
   return (!token_br->eof_);
 }
 
@@ -693,10 +692,8 @@ static int ParseFrame(VP8Decoder* const dec, VP8Io* io) {
         return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
                            "Premature end-of-file encountered.");
       }
+      // Reconstruct and emit samples.
       VP8ReconstructBlock(dec);
-
-      // Store data and save block's filtering params
-      VP8StoreBlock(dec);
     }
     if (!VP8ProcessRow(dec, io)) {
       return VP8SetError(dec, VP8_STATUS_USER_ABORT, "Output aborted.");
diff --git a/3rdparty/libwebp/dec/vp8i.h b/3rdparty/libwebp/dec/vp8i.h
index 6e27456fe1..1d0d4077eb 100644
--- a/3rdparty/libwebp/dec/vp8i.h
+++ b/3rdparty/libwebp/dec/vp8i.h
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // VP8 decoder: internal header.
@@ -27,7 +29,7 @@ extern "C" {
 
 // version numbers
 #define DEC_MAJ_VERSION 0
-#define DEC_MIN_VERSION 2
+#define DEC_MIN_VERSION 3
 #define DEC_REV_VERSION 1
 
 #define ONLY_KEYFRAME_CODE      // to remove any code related to P-Frames
@@ -157,7 +159,7 @@ typedef struct {  // filter specs
 } VP8FInfo;
 
 typedef struct {  // used for syntax-parsing
-  unsigned int nz_;          // non-zero AC/DC coeffs
+  unsigned int nz_:24;       // non-zero AC/DC coeffs (24bit)
   unsigned int dc_nz_:1;     // non-zero DC coeffs
   unsigned int skip_:1;      // block type
 } VP8MB;
@@ -269,13 +271,14 @@ struct VP8Decoder {
   uint32_t non_zero_ac_;
 
   // Filtering side-info
-  int filter_type_;                         // 0=off, 1=simple, 2=complex
-  int filter_row_;                          // per-row flag
-  uint8_t filter_levels_[NUM_MB_SEGMENTS];  // precalculated per-segment
+  int filter_type_;                          // 0=off, 1=simple, 2=complex
+  int filter_row_;                           // per-row flag
+  VP8FInfo fstrengths_[NUM_MB_SEGMENTS][2];  // precalculated per-segment/type
 
   // extensions
   const uint8_t* alpha_data_;   // compressed alpha data (if present)
   size_t alpha_data_size_;
+  int is_alpha_decoded_;  // true if alpha_data_ is decoded in alpha_plane_
   uint8_t* alpha_plane_;        // output. Persistent, contains the whole data.
 
   int layer_colorspace_;
@@ -312,8 +315,6 @@ VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io);
 int VP8ExitCritical(VP8Decoder* const dec, VP8Io* const io);
 // Process the last decoded row (filtering + output)
 int VP8ProcessRow(VP8Decoder* const dec, VP8Io* const io);
-// Store a block, along with filtering params
-void VP8StoreBlock(VP8Decoder* const dec);
 // To be called at the start of a new scanline, to initialize predictors.
 void VP8InitScanline(VP8Decoder* const dec);
 // Decode one macroblock. Returns false if there is not enough data.
diff --git a/3rdparty/libwebp/dec/vp8l.c b/3rdparty/libwebp/dec/vp8l.c
index 897e4395c7..89b5b4bf6e 100644
--- a/3rdparty/libwebp/dec/vp8l.c
+++ b/3rdparty/libwebp/dec/vp8l.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // main entry for the decoder
@@ -58,18 +60,18 @@ static const uint8_t kCodeLengthCodeOrder[NUM_CODE_LENGTH_CODES] = {
 
 #define CODE_TO_PLANE_CODES        120
 static const uint8_t code_to_plane_lut[CODE_TO_PLANE_CODES] = {
-   0x18, 0x07, 0x17, 0x19, 0x28, 0x06, 0x27, 0x29, 0x16, 0x1a,
-   0x26, 0x2a, 0x38, 0x05, 0x37, 0x39, 0x15, 0x1b, 0x36, 0x3a,
-   0x25, 0x2b, 0x48, 0x04, 0x47, 0x49, 0x14, 0x1c, 0x35, 0x3b,
-   0x46, 0x4a, 0x24, 0x2c, 0x58, 0x45, 0x4b, 0x34, 0x3c, 0x03,
-   0x57, 0x59, 0x13, 0x1d, 0x56, 0x5a, 0x23, 0x2d, 0x44, 0x4c,
-   0x55, 0x5b, 0x33, 0x3d, 0x68, 0x02, 0x67, 0x69, 0x12, 0x1e,
-   0x66, 0x6a, 0x22, 0x2e, 0x54, 0x5c, 0x43, 0x4d, 0x65, 0x6b,
-   0x32, 0x3e, 0x78, 0x01, 0x77, 0x79, 0x53, 0x5d, 0x11, 0x1f,
-   0x64, 0x6c, 0x42, 0x4e, 0x76, 0x7a, 0x21, 0x2f, 0x75, 0x7b,
-   0x31, 0x3f, 0x63, 0x6d, 0x52, 0x5e, 0x00, 0x74, 0x7c, 0x41,
-   0x4f, 0x10, 0x20, 0x62, 0x6e, 0x30, 0x73, 0x7d, 0x51, 0x5f,
-   0x40, 0x72, 0x7e, 0x61, 0x6f, 0x50, 0x71, 0x7f, 0x60, 0x70
+  0x18, 0x07, 0x17, 0x19, 0x28, 0x06, 0x27, 0x29, 0x16, 0x1a,
+  0x26, 0x2a, 0x38, 0x05, 0x37, 0x39, 0x15, 0x1b, 0x36, 0x3a,
+  0x25, 0x2b, 0x48, 0x04, 0x47, 0x49, 0x14, 0x1c, 0x35, 0x3b,
+  0x46, 0x4a, 0x24, 0x2c, 0x58, 0x45, 0x4b, 0x34, 0x3c, 0x03,
+  0x57, 0x59, 0x13, 0x1d, 0x56, 0x5a, 0x23, 0x2d, 0x44, 0x4c,
+  0x55, 0x5b, 0x33, 0x3d, 0x68, 0x02, 0x67, 0x69, 0x12, 0x1e,
+  0x66, 0x6a, 0x22, 0x2e, 0x54, 0x5c, 0x43, 0x4d, 0x65, 0x6b,
+  0x32, 0x3e, 0x78, 0x01, 0x77, 0x79, 0x53, 0x5d, 0x11, 0x1f,
+  0x64, 0x6c, 0x42, 0x4e, 0x76, 0x7a, 0x21, 0x2f, 0x75, 0x7b,
+  0x31, 0x3f, 0x63, 0x6d, 0x52, 0x5e, 0x00, 0x74, 0x7c, 0x41,
+  0x4f, 0x10, 0x20, 0x62, 0x6e, 0x30, 0x73, 0x7d, 0x51, 0x5f,
+  0x40, 0x72, 0x7e, 0x61, 0x6f, 0x50, 0x71, 0x7f, 0x60, 0x70
 };
 
 static int DecodeImageStream(int xsize, int ysize,
@@ -149,29 +151,20 @@ static WEBP_INLINE int PlaneCodeToDistance(int xsize, int plane_code) {
 //------------------------------------------------------------------------------
 // Decodes the next Huffman code from bit-stream.
 // FillBitWindow(br) needs to be called at minimum every second call
-// to ReadSymbolUnsafe.
-static int ReadSymbolUnsafe(const HuffmanTree* tree, VP8LBitReader* const br) {
-  const HuffmanTreeNode* node = tree->root_;
-  assert(node != NULL);
-  while (!HuffmanTreeNodeIsLeaf(node)) {
-    node = HuffmanTreeNextNode(node, VP8LReadOneBitUnsafe(br));
-  }
-  return node->symbol_;
-}
-
+// to ReadSymbol, in order to pre-fetch enough bits.
 static WEBP_INLINE int ReadSymbol(const HuffmanTree* tree,
                                   VP8LBitReader* const br) {
-  const int read_safe = (br->pos_ + 8 > br->len_);
-  if (!read_safe) {
-    return ReadSymbolUnsafe(tree, br);
-  } else {
-    const HuffmanTreeNode* node = tree->root_;
-    assert(node != NULL);
-    while (!HuffmanTreeNodeIsLeaf(node)) {
-      node = HuffmanTreeNextNode(node, VP8LReadOneBit(br));
-    }
-    return node->symbol_;
+  const HuffmanTreeNode* node = tree->root_;
+  int num_bits = 0;
+  uint32_t bits = VP8LPrefetchBits(br);
+  assert(node != NULL);
+  while (!HuffmanTreeNodeIsLeaf(node)) {
+    node = HuffmanTreeNextNode(node, bits & 1);
+    bits >>= 1;
+    ++num_bits;
   }
+  VP8LDiscardBits(br, num_bits);
+  return node->symbol_;
 }
 
 static int ReadHuffmanCodeLengths(
@@ -327,10 +320,10 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
     hdr->huffman_subsample_bits_ = huffman_precision;
     for (i = 0; i < huffman_pixs; ++i) {
       // The huffman data is stored in red and green bytes.
-      const int index = (huffman_image[i] >> 8) & 0xffff;
-      huffman_image[i] = index;
-      if (index >= num_htree_groups) {
-        num_htree_groups = index + 1;
+      const int group = (huffman_image[i] >> 8) & 0xffff;
+      huffman_image[i] = group;
+      if (group >= num_htree_groups) {
+        num_htree_groups = group + 1;
       }
     }
   }
@@ -634,10 +627,24 @@ static void ApplyInverseTransforms(VP8LDecoder* const dec, int num_rows,
   }
 }
 
+// Special method for paletted alpha data.
+static void ApplyInverseTransformsAlpha(VP8LDecoder* const dec, int num_rows,
+                                        const uint8_t* const rows) {
+  const int start_row = dec->last_row_;
+  const int end_row = start_row + num_rows;
+  const uint8_t* rows_in = rows;
+  uint8_t* rows_out = (uint8_t*)dec->io_->opaque + dec->io_->width * start_row;
+  VP8LTransform* const transform = &dec->transforms_[0];
+  assert(dec->next_transform_ == 1);
+  assert(transform->type_ == COLOR_INDEXING_TRANSFORM);
+  VP8LColorIndexInverseTransformAlpha(transform, start_row, end_row, rows_in,
+                                      rows_out);
+}
+
 // Processes (transforms, scales & color-converts) the rows decoded after the
 // last call.
 static void ProcessRows(VP8LDecoder* const dec, int row) {
-  const uint32_t* const rows = dec->argb_ + dec->width_ * dec->last_row_;
+  const uint32_t* const rows = dec->pixels_ + dec->width_ * dec->last_row_;
   const int num_rows = row - dec->last_row_;
 
   if (num_rows <= 0) return;  // Nothing to be done.
@@ -676,121 +683,135 @@ static void ProcessRows(VP8LDecoder* const dec, int row) {
   assert(dec->last_row_ <= dec->height_);
 }
 
-static int DecodeImageData(VP8LDecoder* const dec,
-                           uint32_t* const data, int width, int height,
-                           ProcessRowsFunc process_func) {
-  int ok = 1;
-  int col = 0, row = 0;
-  VP8LBitReader* const br = &dec->br_;
-  VP8LMetadata* const hdr = &dec->hdr_;
-  HTreeGroup* htree_group = hdr->htree_groups_;
-  uint32_t* src = data;
-  uint32_t* last_cached = data;
-  uint32_t* const src_end = data + width * height;
-  const int len_code_limit = NUM_LITERAL_CODES + NUM_LENGTH_CODES;
-  const int color_cache_limit = len_code_limit + hdr->color_cache_size_;
-  VP8LColorCache* const color_cache =
-      (hdr->color_cache_size_ > 0) ? &hdr->color_cache_ : NULL;
-  const int mask = hdr->huffman_mask_;
-
-  assert(htree_group != NULL);
-
-  while (!br->eos_ && src < src_end) {
-    int code;
-    // Only update when changing tile. Note we could use the following test:
-    //   if "((((prev_col ^ col) | prev_row ^ row)) > mask)" -> tile changed
-    // but that's actually slower and requires storing the previous col/row
-    if ((col & mask) == 0) {
-      htree_group = GetHtreeGroupForPos(hdr, col, row);
-    }
-    VP8LFillBitWindow(br);
-    code = ReadSymbol(&htree_group->htrees_[GREEN], br);
-    if (code < NUM_LITERAL_CODES) {   // Literal.
-      int red, green, blue, alpha;
-      red = ReadSymbol(&htree_group->htrees_[RED], br);
-      green = code;
-      VP8LFillBitWindow(br);
-      blue = ReadSymbol(&htree_group->htrees_[BLUE], br);
-      alpha = ReadSymbol(&htree_group->htrees_[ALPHA], br);
-      *src = (alpha << 24) + (red << 16) + (green << 8) + blue;
- AdvanceByOne:
-      ++src;
-      ++col;
-      if (col >= width) {
-        col = 0;
-        ++row;
-        if ((process_func != NULL) && (row % NUM_ARGB_CACHE_ROWS == 0)) {
-          process_func(dec, row);
-        }
-        if (color_cache != NULL) {
-          while (last_cached < src) {
-            VP8LColorCacheInsert(color_cache, *last_cached++);
-          }
-        }
-      }
-    } else if (code < len_code_limit) {           // Backward reference
-      int dist_code, dist;
-      const int length_sym = code - NUM_LITERAL_CODES;
-      const int length = GetCopyLength(length_sym, br);
-      const int dist_symbol = ReadSymbol(&htree_group->htrees_[DIST], br);
-      VP8LFillBitWindow(br);
-      dist_code = GetCopyDistance(dist_symbol, br);
-      dist = PlaneCodeToDistance(width, dist_code);
-      if (src - data < dist || src_end - src < length) {
-        ok = 0;
-        goto End;
-      }
-      {
-        int i;
-        for (i = 0; i < length; ++i) src[i] = src[i - dist];
-        src += length;
-      }
-      col += length;
-      while (col >= width) {
-        col -= width;
-        ++row;
-        if ((process_func != NULL) && (row % NUM_ARGB_CACHE_ROWS == 0)) {
-          process_func(dec, row);
-        }
-      }
-      if (src < src_end) {
-        htree_group = GetHtreeGroupForPos(hdr, col, row);
-        if (color_cache != NULL) {
-          while (last_cached < src) {
-            VP8LColorCacheInsert(color_cache, *last_cached++);
-          }
-        }
-      }
-    } else if (code < color_cache_limit) {    // Color cache.
-      const int key = code - len_code_limit;
-      assert(color_cache != NULL);
-      while (last_cached < src) {
-        VP8LColorCacheInsert(color_cache, *last_cached++);
-      }
-      *src = VP8LColorCacheLookup(color_cache, key);
-      goto AdvanceByOne;
-    } else {    // Not reached.
-      ok = 0;
-      goto End;
-    }
-    ok = !br->error_;
-    if (!ok) goto End;
-  }
-  // Process the remaining rows corresponding to last row-block.
-  if (process_func != NULL) process_func(dec, row);
-
- End:
-  if (br->error_ || !ok || (br->eos_ && src < src_end)) {
-    ok = 0;
-    dec->status_ = (!br->eos_) ?
-        VP8_STATUS_BITSTREAM_ERROR : VP8_STATUS_SUSPENDED;
-  } else if (src == src_end) {
-    dec->state_ = READ_DATA;
-  }
-
-  return ok;
+#define DECODE_DATA_FUNC(FUNC_NAME, TYPE, STORE_PIXEL)                         \
+static int FUNC_NAME(VP8LDecoder* const dec, TYPE* const data, int width,      \
+                     int height, ProcessRowsFunc process_func) {               \
+  int ok = 1;                                                                  \
+  int col = 0, row = 0;                                                        \
+  VP8LBitReader* const br = &dec->br_;                                         \
+  VP8LMetadata* const hdr = &dec->hdr_;                                        \
+  HTreeGroup* htree_group = hdr->htree_groups_;                                \
+  TYPE* src = data;                                                            \
+  TYPE* last_cached = data;                                                    \
+  TYPE* const src_end = data + width * height;                                 \
+  const int len_code_limit = NUM_LITERAL_CODES + NUM_LENGTH_CODES;             \
+  const int color_cache_limit = len_code_limit + hdr->color_cache_size_;       \
+  VP8LColorCache* const color_cache =                                          \
+      (hdr->color_cache_size_ > 0) ? &hdr->color_cache_ : NULL;                \
+  const int mask = hdr->huffman_mask_;                                         \
+  assert(htree_group != NULL);                                                 \
+  while (!br->eos_ && src < src_end) {                                         \
+    int code;                                                                  \
+    /* Only update when changing tile. Note we could use this test:        */  \
+    /* if "((((prev_col ^ col) | prev_row ^ row)) > mask)" -> tile changed */  \
+    /* but that's actually slower and needs storing the previous col/row.  */  \
+    if ((col & mask) == 0) {                                                   \
+      htree_group = GetHtreeGroupForPos(hdr, col, row);                        \
+    }                                                                          \
+    VP8LFillBitWindow(br);                                                     \
+    code = ReadSymbol(&htree_group->htrees_[GREEN], br);                       \
+    if (code < NUM_LITERAL_CODES) {  /* Literal*/                              \
+      int red, green, blue, alpha;                                             \
+      red = ReadSymbol(&htree_group->htrees_[RED], br);                        \
+      green = code;                                                            \
+      VP8LFillBitWindow(br);                                                   \
+      blue = ReadSymbol(&htree_group->htrees_[BLUE], br);                      \
+      alpha = ReadSymbol(&htree_group->htrees_[ALPHA], br);                    \
+      *src = STORE_PIXEL(alpha, red, green, blue);                             \
+    AdvanceByOne:                                                              \
+      ++src;                                                                   \
+      ++col;                                                                   \
+      if (col >= width) {                                                      \
+        col = 0;                                                               \
+        ++row;                                                                 \
+        if ((process_func != NULL) && (row % NUM_ARGB_CACHE_ROWS == 0)) {      \
+          process_func(dec, row);                                              \
+        }                                                                      \
+        if (color_cache != NULL) {                                             \
+          while (last_cached < src) {                                          \
+            VP8LColorCacheInsert(color_cache, *last_cached++);                 \
+          }                                                                    \
+        }                                                                      \
+      }                                                                        \
+    } else if (code < len_code_limit) {  /* Backward reference */              \
+      int dist_code, dist;                                                     \
+      const int length_sym = code - NUM_LITERAL_CODES;                         \
+      const int length = GetCopyLength(length_sym, br);                        \
+      const int dist_symbol = ReadSymbol(&htree_group->htrees_[DIST], br);     \
+      VP8LFillBitWindow(br);                                                   \
+      dist_code = GetCopyDistance(dist_symbol, br);                            \
+      dist = PlaneCodeToDistance(width, dist_code);                            \
+      if (src - data < dist || src_end - src < length) {                       \
+        ok = 0;                                                                \
+        goto End;                                                              \
+      }                                                                        \
+      {                                                                        \
+        int i;                                                                 \
+        for (i = 0; i < length; ++i) src[i] = src[i - dist];                   \
+        src += length;                                                         \
+      }                                                                        \
+      col += length;                                                           \
+      while (col >= width) {                                                   \
+        col -= width;                                                          \
+        ++row;                                                                 \
+        if ((process_func != NULL) && (row % NUM_ARGB_CACHE_ROWS == 0)) {      \
+          process_func(dec, row);                                              \
+        }                                                                      \
+      }                                                                        \
+      if (src < src_end) {                                                     \
+        htree_group = GetHtreeGroupForPos(hdr, col, row);                      \
+        if (color_cache != NULL) {                                             \
+          while (last_cached < src) {                                          \
+            VP8LColorCacheInsert(color_cache, *last_cached++);                 \
+          }                                                                    \
+        }                                                                      \
+      }                                                                        \
+    } else if (code < color_cache_limit) {  /* Color cache */                  \
+      const int key = code - len_code_limit;                                   \
+      assert(color_cache != NULL);                                             \
+      while (last_cached < src) {                                              \
+        VP8LColorCacheInsert(color_cache, *last_cached++);                     \
+      }                                                                        \
+      *src = VP8LColorCacheLookup(color_cache, key);                           \
+      goto AdvanceByOne;                                                       \
+    } else {  /* Not reached */                                                \
+      ok = 0;                                                                  \
+      goto End;                                                                \
+    }                                                                          \
+    ok = !br->error_;                                                          \
+    if (!ok) goto End;                                                         \
+  }                                                                            \
+  /* Process the remaining rows corresponding to last row-block. */            \
+  if (process_func != NULL) process_func(dec, row);                            \
+End:                                                                           \
+  if (br->error_ || !ok || (br->eos_ && src < src_end)) {                      \
+    ok = 0;                                                                    \
+    dec->status_ =                                                             \
+        (!br->eos_) ? VP8_STATUS_BITSTREAM_ERROR : VP8_STATUS_SUSPENDED;       \
+  } else if (src == src_end) {                                                 \
+    dec->state_ = READ_DATA;                                                   \
+  }                                                                            \
+  return ok;                                                                   \
 }
 
+static WEBP_INLINE uint32_t GetARGBPixel(int alpha, int red, int green,
+                                         int blue) {
+  return (alpha << 24) | (red << 16) | (green << 8) | blue;
+}
+
+static WEBP_INLINE uint8_t GetAlphaPixel(int alpha, int red, int green,
+                                         int blue) {
+  (void)alpha;
+  (void)red;
+  (void)blue;
+  return green;  // Alpha value is stored in green channel.
+}
+
+DECODE_DATA_FUNC(DecodeImageData, uint32_t, GetARGBPixel)
+DECODE_DATA_FUNC(DecodeAlphaData, uint8_t, GetAlphaPixel)
+
+#undef DECODE_DATA_FUNC
+
 // -----------------------------------------------------------------------------
 // VP8LTransform
 
@@ -912,8 +933,8 @@ void VP8LClear(VP8LDecoder* const dec) {
   if (dec == NULL) return;
   ClearMetadata(&dec->hdr_);
 
-  free(dec->argb_);
-  dec->argb_ = NULL;
+  free(dec->pixels_);
+  dec->pixels_ = NULL;
   for (i = 0; i < dec->next_transform_; ++i) {
     ClearTransform(&dec->transforms_[i]);
   }
@@ -1037,35 +1058,39 @@ static int DecodeImageStream(int xsize, int ysize,
 }
 
 //------------------------------------------------------------------------------
-// Allocate dec->argb_ and dec->argb_cache_ using dec->width_ and dec->height_
-
-static int AllocateARGBBuffers(VP8LDecoder* const dec, int final_width) {
+// Allocate internal buffers dec->pixels_ and dec->argb_cache_.
+static int AllocateInternalBuffers(VP8LDecoder* const dec, int final_width,
+                                   size_t bytes_per_pixel) {
+  const int argb_cache_needed = (bytes_per_pixel == sizeof(uint32_t));
   const uint64_t num_pixels = (uint64_t)dec->width_ * dec->height_;
   // Scratch buffer corresponding to top-prediction row for transforming the
-  // first row in the row-blocks.
-  const uint64_t cache_top_pixels = final_width;
-  // Scratch buffer for temporary BGRA storage.
-  const uint64_t cache_pixels = (uint64_t)final_width * NUM_ARGB_CACHE_ROWS;
+  // first row in the row-blocks. Not needed for paletted alpha.
+  const uint64_t cache_top_pixels =
+      argb_cache_needed ? (uint16_t)final_width : 0ULL;
+  // Scratch buffer for temporary BGRA storage. Not needed for paletted alpha.
+  const uint64_t cache_pixels =
+      argb_cache_needed ? (uint64_t)final_width * NUM_ARGB_CACHE_ROWS : 0ULL;
   const uint64_t total_num_pixels =
       num_pixels + cache_top_pixels + cache_pixels;
 
   assert(dec->width_ <= final_width);
-  dec->argb_ = (uint32_t*)WebPSafeMalloc(total_num_pixels, sizeof(*dec->argb_));
-  if (dec->argb_ == NULL) {
+  dec->pixels_ = (uint32_t*)WebPSafeMalloc(total_num_pixels, bytes_per_pixel);
+  if (dec->pixels_ == NULL) {
     dec->argb_cache_ = NULL;    // for sanity check
     dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
     return 0;
   }
-  dec->argb_cache_ = dec->argb_ + num_pixels + cache_top_pixels;
+  dec->argb_cache_ =
+      argb_cache_needed ? dec->pixels_ + num_pixels + cache_top_pixels : NULL;
   return 1;
 }
 
 //------------------------------------------------------------------------------
-// Special row-processing that only stores the alpha data.
 
+// Special row-processing that only stores the alpha data.
 static void ExtractAlphaRows(VP8LDecoder* const dec, int row) {
   const int num_rows = row - dec->last_row_;
-  const uint32_t* const in = dec->argb_ + dec->width_ * dec->last_row_;
+  const uint32_t* const in = dec->pixels_ + dec->width_ * dec->last_row_;
 
   if (num_rows <= 0) return;  // Nothing to be done.
   ApplyInverseTransforms(dec, num_rows, in);
@@ -1079,7 +1104,17 @@ static void ExtractAlphaRows(VP8LDecoder* const dec, int row) {
     int i;
     for (i = 0; i < cache_pixs; ++i) dst[i] = (src[i] >> 8) & 0xff;
   }
+  dec->last_row_ = dec->last_out_row_ = row;
+}
 
+// Row-processing for the special case when alpha data contains only one
+// transform: color indexing.
+static void ExtractPalettedAlphaRows(VP8LDecoder* const dec, int row) {
+  const int num_rows = row - dec->last_row_;
+  const uint8_t* const in =
+      (uint8_t*)dec->pixels_ + dec->width_ * dec->last_row_;
+  if (num_rows <= 0) return;  // Nothing to be done.
+  ApplyInverseTransformsAlpha(dec, num_rows, in);
   dec->last_row_ = dec->last_out_row_ = row;
 }
 
@@ -1088,6 +1123,7 @@ int VP8LDecodeAlphaImageStream(int width, int height, const uint8_t* const data,
   VP8Io io;
   int ok = 0;
   VP8LDecoder* const dec = VP8LNew();
+  size_t bytes_per_pixel = sizeof(uint32_t);  // Default: BGRA mode.
   if (dec == NULL) return 0;
 
   dec->width_ = width;
@@ -1106,13 +1142,25 @@ int VP8LDecodeAlphaImageStream(int width, int height, const uint8_t* const data,
   dec->action_ = READ_HDR;
   if (!DecodeImageStream(width, height, 1, dec, NULL)) goto Err;
 
-  // Allocate output (note that dec->width_ may have changed here).
-  if (!AllocateARGBBuffers(dec, width)) goto Err;
+  // Special case: if alpha data uses only the color indexing transform and
+  // doesn't use color cache (a frequent case), we will use DecodeAlphaData()
+  // method that only needs allocation of 1 byte per pixel (alpha channel).
+  if (dec->next_transform_ == 1 &&
+      dec->transforms_[0].type_ == COLOR_INDEXING_TRANSFORM &&
+      dec->hdr_.color_cache_size_ == 0) {
+    bytes_per_pixel = sizeof(uint8_t);
+  }
+
+  // Allocate internal buffers (note that dec->width_ may have changed here).
+  if (!AllocateInternalBuffers(dec, width, bytes_per_pixel)) goto Err;
 
   // Decode (with special row processing).
   dec->action_ = READ_DATA;
-  ok = DecodeImageData(dec, dec->argb_, dec->width_, dec->height_,
-                       ExtractAlphaRows);
+  ok = (bytes_per_pixel == sizeof(uint8_t)) ?
+      DecodeAlphaData(dec, (uint8_t*)dec->pixels_, dec->width_, dec->height_,
+                      ExtractPalettedAlphaRows) :
+      DecodeImageData(dec, dec->pixels_, dec->width_, dec->height_,
+                      ExtractAlphaRows);
 
  Err:
   VP8LDelete(dec);
@@ -1146,12 +1194,13 @@ int VP8LDecodeHeader(VP8LDecoder* const dec, VP8Io* const io) {
   return 1;
 
  Error:
-   VP8LClear(dec);
-   assert(dec->status_ != VP8_STATUS_OK);
-   return 0;
+  VP8LClear(dec);
+  assert(dec->status_ != VP8_STATUS_OK);
+  return 0;
 }
 
 int VP8LDecodeImage(VP8LDecoder* const dec) {
+  const size_t bytes_per_pixel = sizeof(uint32_t);
   VP8Io* io = NULL;
   WebPDecParams* params = NULL;
 
@@ -1171,13 +1220,13 @@ int VP8LDecodeImage(VP8LDecoder* const dec) {
     goto Err;
   }
 
-  if (!AllocateARGBBuffers(dec, io->width)) goto Err;
+  if (!AllocateInternalBuffers(dec, io->width, bytes_per_pixel)) goto Err;
 
   if (io->use_scaling && !AllocateAndInitRescaler(dec, io)) goto Err;
 
   // Decode.
   dec->action_ = READ_DATA;
-  if (!DecodeImageData(dec, dec->argb_, dec->width_, dec->height_,
+  if (!DecodeImageData(dec, dec->pixels_, dec->width_, dec->height_,
                        ProcessRows)) {
     goto Err;
   }
diff --git a/3rdparty/libwebp/dec/vp8li.h b/3rdparty/libwebp/dec/vp8li.h
index ee29eb5faf..543a767590 100644
--- a/3rdparty/libwebp/dec/vp8li.h
+++ b/3rdparty/libwebp/dec/vp8li.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Lossless decoder: internal header.
@@ -63,7 +65,8 @@ typedef struct {
 
   const WebPDecBuffer *output_;    // shortcut to io->opaque->output
 
-  uint32_t        *argb_;          // Internal data: always in BGRA color mode.
+  uint32_t        *pixels_;        // Internal data: either uint8_t* for alpha
+                                   // or uint32_t* for BGRA.
   uint32_t        *argb_cache_;    // Scratch buffer for temporary BGRA storage.
 
   VP8LBitReader    br_;
diff --git a/3rdparty/libwebp/dec/webp.c b/3rdparty/libwebp/dec/webp.c
index 7455da9415..97e79b64da 100644
--- a/3rdparty/libwebp/dec/webp.c
+++ b/3rdparty/libwebp/dec/webp.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Main decoding functions for WEBP images.
@@ -14,7 +16,7 @@
 #include "./vp8i.h"
 #include "./vp8li.h"
 #include "./webpi.h"
-#include "../webp/format_constants.h"
+#include "../webp/mux_types.h"  // ALPHA_FLAG
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -40,8 +42,8 @@ extern "C" {
 //   20..23  VP8X flags bit-map corresponding to the chunk-types present.
 //   24..26  Width of the Canvas Image.
 //   27..29  Height of the Canvas Image.
-// There can be extra chunks after the "VP8X" chunk (ICCP, TILE, FRM, VP8,
-// META  ...)
+// There can be extra chunks after the "VP8X" chunk (ICCP, FRGM, ANMF, VP8,
+// VP8L, XMP, EXIF  ...)
 // All sizes are in little-endian order.
 // Note: chunk data size must be padded to multiple of 2 when written.
 
@@ -192,6 +194,15 @@ static VP8StatusCode ParseOptionalChunks(const uint8_t** const data,
       return VP8_STATUS_BITSTREAM_ERROR;          // Not a valid chunk size.
     }
 
+    // Start of a (possibly incomplete) VP8/VP8L chunk implies that we have
+    // parsed all the optional chunks.
+    // Note: This check must occur before the check 'buf_size < disk_chunk_size'
+    // below to allow incomplete VP8/VP8L chunks.
+    if (!memcmp(buf, "VP8 ", TAG_SIZE) ||
+        !memcmp(buf, "VP8L", TAG_SIZE)) {
+      return VP8_STATUS_OK;
+    }
+
     if (buf_size < disk_chunk_size) {             // Insufficient data.
       return VP8_STATUS_NOT_ENOUGH_DATA;
     }
@@ -199,9 +210,6 @@ static VP8StatusCode ParseOptionalChunks(const uint8_t** const data,
     if (!memcmp(buf, "ALPH", TAG_SIZE)) {         // A valid ALPH header.
       *alpha_data = buf + CHUNK_HEADER_SIZE;
       *alpha_size = chunk_size;
-    } else if (!memcmp(buf, "VP8 ", TAG_SIZE) ||
-               !memcmp(buf, "VP8L", TAG_SIZE)) {  // A valid VP8/VP8L header.
-      return VP8_STATUS_OK;  // Found.
     }
 
     // We have a full and valid chunk; skip it.
@@ -276,6 +284,7 @@ static VP8StatusCode ParseHeadersInternal(const uint8_t* data,
                                           int* const width,
                                           int* const height,
                                           int* const has_alpha,
+                                          int* const has_animation,
                                           WebPHeaderStructure* const headers) {
   int found_riff = 0;
   int found_vp8x = 0;
@@ -308,7 +317,8 @@ static VP8StatusCode ParseHeadersInternal(const uint8_t* data,
       // necessary to send VP8X chunk to the decoder.
       return VP8_STATUS_BITSTREAM_ERROR;
     }
-    if (has_alpha != NULL) *has_alpha = !!(flags & ALPHA_FLAG_BIT);
+    if (has_alpha != NULL) *has_alpha = !!(flags & ALPHA_FLAG);
+    if (has_animation != NULL) *has_animation = !!(flags & ANIMATION_FLAG);
     if (found_vp8x && headers == NULL) {
       return VP8_STATUS_OK;  // Return features from VP8X header.
     }
@@ -370,10 +380,19 @@ static VP8StatusCode ParseHeadersInternal(const uint8_t* data,
 }
 
 VP8StatusCode WebPParseHeaders(WebPHeaderStructure* const headers) {
+  VP8StatusCode status;
+  int has_animation = 0;
   assert(headers != NULL);
   // fill out headers, ignore width/height/has_alpha.
-  return ParseHeadersInternal(headers->data, headers->data_size,
-                              NULL, NULL, NULL, headers);
+  status = ParseHeadersInternal(headers->data, headers->data_size,
+                                NULL, NULL, NULL, &has_animation, headers);
+  if (status == VP8_STATUS_OK || status == VP8_STATUS_NOT_ENOUGH_DATA) {
+    // TODO(jzern): full support of animation frames will require API additions.
+    if (has_animation) {
+      status = VP8_STATUS_UNSUPPORTED_FEATURE;
+    }
+  }
+  return status;
 }
 
 //------------------------------------------------------------------------------
@@ -625,10 +644,11 @@ static VP8StatusCode GetFeatures(const uint8_t* const data, size_t data_size,
   }
   DefaultFeatures(features);
 
-  // Only parse enough of the data to retrieve width/height/has_alpha.
+  // Only parse enough of the data to retrieve the features.
   return ParseHeadersInternal(data, data_size,
                               &features->width, &features->height,
-                              &features->has_alpha, NULL);
+                              &features->has_alpha, &features->has_animation,
+                              NULL);
 }
 
 //------------------------------------------------------------------------------
@@ -672,19 +692,13 @@ int WebPInitDecoderConfigInternal(WebPDecoderConfig* config,
 VP8StatusCode WebPGetFeaturesInternal(const uint8_t* data, size_t data_size,
                                       WebPBitstreamFeatures* features,
                                       int version) {
-  VP8StatusCode status;
   if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DECODER_ABI_VERSION)) {
     return VP8_STATUS_INVALID_PARAM;   // version mismatch
   }
   if (features == NULL) {
     return VP8_STATUS_INVALID_PARAM;
   }
-
-  status = GetFeatures(data, data_size, features);
-  if (status == VP8_STATUS_NOT_ENOUGH_DATA) {
-    return VP8_STATUS_BITSTREAM_ERROR;  // Not-enough-data treated as error.
-  }
-  return status;
+  return GetFeatures(data, data_size, features);
 }
 
 VP8StatusCode WebPDecode(const uint8_t* data, size_t data_size,
diff --git a/3rdparty/libwebp/dec/webpi.h b/3rdparty/libwebp/dec/webpi.h
index 44e5744411..4ae0bfc5a0 100644
--- a/3rdparty/libwebp/dec/webpi.h
+++ b/3rdparty/libwebp/dec/webpi.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Internal header: WebP decoding parameters and custom IO on buffer
@@ -61,10 +63,10 @@ typedef struct {
 } WebPHeaderStructure;
 
 // Skips over all valid chunks prior to the first VP8/VP8L frame header.
-// Returns VP8_STATUS_OK on success,
-//         VP8_STATUS_BITSTREAM_ERROR if an invalid header/chunk is found, and
-//         VP8_STATUS_NOT_ENOUGH_DATA if case of insufficient data.
-// In 'headers', compressed_size, offset, alpha_data, alpha_size and lossless
+// Returns: VP8_STATUS_OK, VP8_STATUS_BITSTREAM_ERROR (invalid header/chunk),
+// VP8_STATUS_NOT_ENOUGH_DATA (partial input) or VP8_STATUS_UNSUPPORTED_FEATURE
+// in the case of non-decodable features (animation for instance).
+// In 'headers', compressed_size, offset, alpha_data, alpha_size, and lossless
 // fields are updated appropriately upon success.
 VP8StatusCode WebPParseHeaders(WebPHeaderStructure* const headers);
 
diff --git a/3rdparty/libwebp/mux/demux.c b/3rdparty/libwebp/demux/demux.c
similarity index 67%
rename from 3rdparty/libwebp/mux/demux.c
rename to 3rdparty/libwebp/demux/demux.c
index 4519f7d55b..bd17ff7f69 100644
--- a/3rdparty/libwebp/mux/demux.c
+++ b/3rdparty/libwebp/demux/demux.c
@@ -1,26 +1,35 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  WebP container demux.
 //
 
-#include "../webp/mux.h"
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
 
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 
-#include "../webp/decode.h"  // WebPGetInfo
+#include "../utils/utils.h"
+#include "../webp/decode.h"     // WebPGetFeatures
+#include "../webp/demux.h"
 #include "../webp/format_constants.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
 
-#define MKFOURCC(a, b, c, d) ((uint32_t)(a) | (b) << 8 | (c) << 16 | (d) << 24)
+#define DMUX_MAJ_VERSION 0
+#define DMUX_MIN_VERSION 1
+#define DMUX_REV_VERSION 1
 
 typedef struct {
   size_t start_;        // start location of the data
@@ -39,8 +48,9 @@ typedef struct Frame {
   int x_offset_, y_offset_;
   int width_, height_;
   int duration_;
-  int is_tile_;    // this is an image fragment from a 'TILE'.
-  int frame_num_;  // the referent frame number for use in assembling tiles.
+  WebPMuxAnimDispose dispose_method_;
+  int is_fragment_;  // this is a frame fragment (and not a full frame).
+  int frame_num_;  // the referent frame number for use in assembling fragments.
   int complete_;   // img_components_ contains a full image.
   ChunkData img_components_[2];  // 0=VP8{,L} 1=ALPH
   struct Frame* next_;
@@ -58,8 +68,10 @@ struct WebPDemuxer {
   uint32_t feature_flags_;
   int canvas_width_, canvas_height_;
   int loop_count_;
+  uint32_t bgcolor_;
   int num_frames_;
   Frame* frames_;
+  Frame** frames_tail_;
   Chunk* chunks_;  // non-image chunks
 };
 
@@ -87,6 +99,12 @@ static const ChunkParser kMasterChunks[] = {
   { { '0', '0', '0', '0' }, NULL,             NULL },
 };
 
+//------------------------------------------------------------------------------
+
+int WebPGetDemuxVersion(void) {
+  return (DMUX_MAJ_VERSION << 16) | (DMUX_MIN_VERSION << 8) | DMUX_REV_VERSION;
+}
+
 // -----------------------------------------------------------------------------
 // MemBuffer
 
@@ -127,43 +145,30 @@ static WEBP_INLINE const uint8_t* GetBuffer(MemBuffer* const mem) {
   return mem->buf_ + mem->start_;
 }
 
-static WEBP_INLINE uint8_t GetByte(MemBuffer* const mem) {
+// Read from 'mem' and skip the read bytes.
+static WEBP_INLINE uint8_t ReadByte(MemBuffer* const mem) {
   const uint8_t byte = mem->buf_[mem->start_];
   Skip(mem, 1);
   return byte;
 }
 
-// Read 16, 24 or 32 bits stored in little-endian order.
-static WEBP_INLINE int ReadLE16s(const uint8_t* const data) {
-  return (int)(data[0] << 0) | (data[1] << 8);
-}
-
-static WEBP_INLINE int ReadLE24s(const uint8_t* const data) {
-  return ReadLE16s(data) | (data[2] << 16);
-}
-
-static WEBP_INLINE uint32_t ReadLE32(const uint8_t* const data) {
-  return (uint32_t)ReadLE24s(data) | (data[3] << 24);
-}
-
-// In addition to reading, skip the read bytes.
-static WEBP_INLINE int GetLE16s(MemBuffer* const mem) {
+static WEBP_INLINE int ReadLE16s(MemBuffer* const mem) {
   const uint8_t* const data = mem->buf_ + mem->start_;
-  const int val = ReadLE16s(data);
+  const int val = GetLE16(data);
   Skip(mem, 2);
   return val;
 }
 
-static WEBP_INLINE int GetLE24s(MemBuffer* const mem) {
+static WEBP_INLINE int ReadLE24s(MemBuffer* const mem) {
   const uint8_t* const data = mem->buf_ + mem->start_;
-  const int val = ReadLE24s(data);
+  const int val = GetLE24(data);
   Skip(mem, 3);
   return val;
 }
 
-static WEBP_INLINE uint32_t GetLE32(MemBuffer* const mem) {
+static WEBP_INLINE uint32_t ReadLE32(MemBuffer* const mem) {
   const uint8_t* const data = mem->buf_ + mem->start_;
-  const uint32_t val = ReadLE32(data);
+  const uint32_t val = GetLE32(data);
   Skip(mem, 4);
   return val;
 }
@@ -181,32 +186,34 @@ static void AddChunk(WebPDemuxer* const dmux, Chunk* const chunk) {
 // Add a frame to the end of the list, ensuring the last frame is complete.
 // Returns true on success, false otherwise.
 static int AddFrame(WebPDemuxer* const dmux, Frame* const frame) {
-  const Frame* last_frame = NULL;
-  Frame** f = &dmux->frames_;
-  while (*f != NULL) {
-    last_frame = *f;
-    f = &(*f)->next_;
-  }
+  const Frame* const last_frame = *dmux->frames_tail_;
   if (last_frame != NULL && !last_frame->complete_) return 0;
-  *f = frame;
+
+  *dmux->frames_tail_ = frame;
   frame->next_ = NULL;
+  dmux->frames_tail_ = &frame->next_;
   return 1;
 }
 
 // Store image bearing chunks to 'frame'.
-static ParseStatus StoreFrame(int frame_num, MemBuffer* const mem,
-                              Frame* const frame) {
+// If 'has_vp8l_alpha' is not NULL, it will be set to true if the frame is a
+// lossless image with alpha.
+static ParseStatus StoreFrame(int frame_num, uint32_t min_size,
+                              MemBuffer* const mem, Frame* const frame,
+                              int* const has_vp8l_alpha) {
   int alpha_chunks = 0;
   int image_chunks = 0;
-  int done = (MemDataSize(mem) < CHUNK_HEADER_SIZE);
+  int done = (MemDataSize(mem) < min_size);
   ParseStatus status = PARSE_OK;
 
+  if (has_vp8l_alpha != NULL) *has_vp8l_alpha = 0;  // Default.
+
   if (done) return PARSE_NEED_MORE_DATA;
 
   do {
     const size_t chunk_start_offset = mem->start_;
-    const uint32_t fourcc = GetLE32(mem);
-    const uint32_t payload_size = GetLE32(mem);
+    const uint32_t fourcc = ReadLE32(mem);
+    const uint32_t payload_size = ReadLE32(mem);
     const uint32_t payload_size_padded = payload_size + (payload_size & 1);
     const size_t payload_available = (payload_size_padded > MemDataSize(mem))
                                    ? MemDataSize(mem) : payload_size_padded;
@@ -228,23 +235,30 @@ static ParseStatus StoreFrame(int frame_num, MemBuffer* const mem,
           goto Done;
         }
         break;
-      case MKFOURCC('V', 'P', '8', ' '):
       case MKFOURCC('V', 'P', '8', 'L'):
+        if (alpha_chunks > 0) return PARSE_ERROR;  // VP8L has its own alpha
+        // fall through
+      case MKFOURCC('V', 'P', '8', ' '):
         if (image_chunks == 0) {
-          int width = 0, height = 0;
+          // Extract the bitstream features, tolerating failures when the data
+          // is incomplete.
+          WebPBitstreamFeatures features;
+          const VP8StatusCode vp8_status =
+              WebPGetFeatures(mem->buf_ + chunk_start_offset, chunk_size,
+                              &features);
+          if (status == PARSE_NEED_MORE_DATA &&
+              vp8_status == VP8_STATUS_NOT_ENOUGH_DATA) {
+            return PARSE_NEED_MORE_DATA;
+          } else if (vp8_status != VP8_STATUS_OK) {
+            // We have enough data, and yet WebPGetFeatures() failed.
+            return PARSE_ERROR;
+          }
           ++image_chunks;
           frame->img_components_[0].offset_ = chunk_start_offset;
           frame->img_components_[0].size_ = chunk_size;
-          // Extract the width and height from the bitstream, tolerating
-          // failures when the data is incomplete.
-          if (!WebPGetInfo(mem->buf_ + frame->img_components_[0].offset_,
-                           frame->img_components_[0].size_, &width, &height) &&
-              status != PARSE_NEED_MORE_DATA) {
-            return PARSE_ERROR;
-          }
-
-          frame->width_ = width;
-          frame->height_ = height;
+          frame->width_ = features.width;
+          frame->height_ = features.height;
+          if (has_vp8l_alpha != NULL) *has_vp8l_alpha = features.has_alpha;
           frame->frame_num_ = frame_num;
           frame->complete_ = (status == PARSE_OK);
           Skip(mem, payload_available);
@@ -275,42 +289,44 @@ static ParseStatus StoreFrame(int frame_num, MemBuffer* const mem,
 // Returns PARSE_OK on success with *frame pointing to the new Frame.
 // Returns PARSE_NEED_MORE_DATA with insufficient data, PARSE_ERROR otherwise.
 static ParseStatus NewFrame(const MemBuffer* const mem,
-                            uint32_t min_size, uint32_t expected_size,
-                            uint32_t actual_size, Frame** frame) {
+                            uint32_t min_size, uint32_t actual_size,
+                            Frame** frame) {
   if (SizeIsInvalid(mem, min_size)) return PARSE_ERROR;
-  if (actual_size < expected_size) return PARSE_ERROR;
+  if (actual_size < min_size) return PARSE_ERROR;
   if (MemDataSize(mem) < min_size)  return PARSE_NEED_MORE_DATA;
 
   *frame = (Frame*)calloc(1, sizeof(**frame));
   return (*frame == NULL) ? PARSE_ERROR : PARSE_OK;
 }
 
-// Parse a 'FRM ' chunk and any image bearing chunks that immediately follow.
+// Parse a 'ANMF' chunk and any image bearing chunks that immediately follow.
 // 'frame_chunk_size' is the previously validated, padded chunk size.
-static ParseStatus ParseFrame(
+static ParseStatus ParseAnimationFrame(
     WebPDemuxer* const dmux, uint32_t frame_chunk_size) {
   const int has_frames = !!(dmux->feature_flags_ & ANIMATION_FLAG);
-  const uint32_t min_size = frame_chunk_size + CHUNK_HEADER_SIZE;
+  const uint32_t anmf_payload_size = frame_chunk_size - ANMF_CHUNK_SIZE;
   int added_frame = 0;
   MemBuffer* const mem = &dmux->mem_;
   Frame* frame;
   ParseStatus status =
-      NewFrame(mem, min_size, FRAME_CHUNK_SIZE, frame_chunk_size, &frame);
+      NewFrame(mem, ANMF_CHUNK_SIZE, frame_chunk_size, &frame);
   if (status != PARSE_OK) return status;
 
-  frame->x_offset_ = 2 * GetLE24s(mem);
-  frame->y_offset_ = 2 * GetLE24s(mem);
-  frame->width_    = 1 + GetLE24s(mem);
-  frame->height_   = 1 + GetLE24s(mem);
-  frame->duration_ = 1 + GetLE24s(mem);
-  Skip(mem, frame_chunk_size - FRAME_CHUNK_SIZE);  // skip any trailing data.
+  frame->x_offset_       = 2 * ReadLE24s(mem);
+  frame->y_offset_       = 2 * ReadLE24s(mem);
+  frame->width_          = 1 + ReadLE24s(mem);
+  frame->height_         = 1 + ReadLE24s(mem);
+  frame->duration_       = ReadLE24s(mem);
+  frame->dispose_method_ = (WebPMuxAnimDispose)(ReadByte(mem) & 1);
   if (frame->width_ * (uint64_t)frame->height_ >= MAX_IMAGE_AREA) {
+    free(frame);
     return PARSE_ERROR;
   }
 
-  // Store a (potentially partial) frame only if the animation flag is set
-  // and there is some data in 'frame'.
-  status = StoreFrame(dmux->num_frames_ + 1, mem, frame);
+  // Store a frame only if the animation flag is set there is some data for
+  // this frame is available.
+  status = StoreFrame(dmux->num_frames_ + 1, anmf_payload_size, mem, frame,
+                      NULL);
   if (status != PARSE_ERROR && has_frames && frame->frame_num_ > 0) {
     added_frame = AddFrame(dmux, frame);
     if (added_frame) {
@@ -324,38 +340,43 @@ static ParseStatus ParseFrame(
   return status;
 }
 
-// Parse a 'TILE' chunk and any image bearing chunks that immediately follow.
-// 'tile_chunk_size' is the previously validated, padded chunk size.
-static ParseStatus ParseTile(WebPDemuxer* const dmux,
-                             uint32_t tile_chunk_size) {
-  const int has_tiles = !!(dmux->feature_flags_ & TILE_FLAG);
-  const uint32_t min_size = tile_chunk_size + CHUNK_HEADER_SIZE;
-  int added_tile = 0;
+#ifdef WEBP_EXPERIMENTAL_FEATURES
+// Parse a 'FRGM' chunk and any image bearing chunks that immediately follow.
+// 'fragment_chunk_size' is the previously validated, padded chunk size.
+static ParseStatus ParseFragment(WebPDemuxer* const dmux,
+                                 uint32_t fragment_chunk_size) {
+  const int frame_num = 1;  // All fragments belong to the 1st (and only) frame.
+  const int has_fragments = !!(dmux->feature_flags_ & FRAGMENTS_FLAG);
+  const uint32_t frgm_payload_size = fragment_chunk_size - FRGM_CHUNK_SIZE;
+  int added_fragment = 0;
   MemBuffer* const mem = &dmux->mem_;
   Frame* frame;
   ParseStatus status =
-      NewFrame(mem, min_size, TILE_CHUNK_SIZE, tile_chunk_size, &frame);
+      NewFrame(mem, FRGM_CHUNK_SIZE, fragment_chunk_size, &frame);
   if (status != PARSE_OK) return status;
 
-  frame->is_tile_  = 1;
-  frame->x_offset_ = 2 * GetLE24s(mem);
-  frame->y_offset_ = 2 * GetLE24s(mem);
-  Skip(mem, tile_chunk_size - TILE_CHUNK_SIZE);  // skip any trailing data.
+  frame->is_fragment_ = 1;
+  frame->x_offset_ = 2 * ReadLE24s(mem);
+  frame->y_offset_ = 2 * ReadLE24s(mem);
 
-  // Store a (potentially partial) tile only if the tile flag is set
-  // and the tile contains some data.
-  status = StoreFrame(dmux->num_frames_, mem, frame);
-  if (status != PARSE_ERROR && has_tiles && frame->frame_num_ > 0) {
-    // Note num_frames_ is incremented only when all tiles have been consumed.
-    added_tile = AddFrame(dmux, frame);
-    if (!added_tile) status = PARSE_ERROR;
+  // Store a fragment only if the fragments flag is set there is some data for
+  // this fragment is available.
+  status = StoreFrame(frame_num, frgm_payload_size, mem, frame, NULL);
+  if (status != PARSE_ERROR && has_fragments && frame->frame_num_ > 0) {
+    added_fragment = AddFrame(dmux, frame);
+    if (!added_fragment) {
+      status = PARSE_ERROR;
+    } else {
+      dmux->num_frames_ = 1;
+    }
   }
 
-  if (!added_tile) free(frame);
+  if (!added_fragment) free(frame);
   return status;
 }
+#endif  // WEBP_EXPERIMENTAL_FEATURES
 
-// General chunk storage starting with the header at 'start_offset' allowing
+// General chunk storage, starting with the header at 'start_offset', allowing
 // the user to request the payload via a fourcc string. 'size' includes the
 // header and the unpadded payload size.
 // Returns true on success, false otherwise.
@@ -384,7 +405,7 @@ static int ReadHeader(MemBuffer* const mem) {
     return 0;
   }
 
-  riff_size = ReadLE32(GetBuffer(mem) + TAG_SIZE);
+  riff_size = GetLE32(GetBuffer(mem) + TAG_SIZE);
   if (riff_size < CHUNK_HEADER_SIZE) return 0;
   if (riff_size > MAX_CHUNK_PAYLOAD) return 0;
 
@@ -403,6 +424,7 @@ static ParseStatus ParseSingleImage(WebPDemuxer* const dmux) {
   MemBuffer* const mem = &dmux->mem_;
   Frame* frame;
   ParseStatus status;
+  int has_vp8l_alpha = 0;  // Frame contains a lossless image with alpha.
 
   if (dmux->frames_ != NULL) return PARSE_ERROR;
   if (SizeIsInvalid(mem, min_size)) return PARSE_ERROR;
@@ -411,7 +433,10 @@ static ParseStatus ParseSingleImage(WebPDemuxer* const dmux) {
   frame = (Frame*)calloc(1, sizeof(*frame));
   if (frame == NULL) return PARSE_ERROR;
 
-  status = StoreFrame(1, &dmux->mem_, frame);
+  // For the single image case we allow parsing of a partial frame, but we need
+  // at least CHUNK_HEADER_SIZE for parsing.
+  status = StoreFrame(1, CHUNK_HEADER_SIZE, &dmux->mem_, frame,
+                      &has_vp8l_alpha);
   if (status != PARSE_ERROR) {
     const int has_alpha = !!(dmux->feature_flags_ & ALPHA_FLAG);
     // Clear any alpha when the alpha flag is missing.
@@ -421,10 +446,12 @@ static ParseStatus ParseSingleImage(WebPDemuxer* const dmux) {
     }
 
     // Use the frame width/height as the canvas values for non-vp8x files.
+    // Also, set ALPHA_FLAG if this is a lossless image with alpha.
     if (!dmux->is_ext_format_ && frame->width_ > 0 && frame->height_ > 0) {
       dmux->state_ = WEBP_DEMUX_PARSED_HEADER;
       dmux->canvas_width_ = frame->width_;
       dmux->canvas_height_ = frame->height_;
+      dmux->feature_flags_ |= has_vp8l_alpha ? ALPHA_FLAG : 0;
     }
     AddFrame(dmux, frame);
     dmux->num_frames_ = 1;
@@ -437,7 +464,7 @@ static ParseStatus ParseSingleImage(WebPDemuxer* const dmux) {
 
 static ParseStatus ParseVP8X(WebPDemuxer* const dmux) {
   MemBuffer* const mem = &dmux->mem_;
-  int loop_chunks = 0;
+  int anim_chunks = 0;
   uint32_t vp8x_size;
   ParseStatus status = PARSE_OK;
 
@@ -445,17 +472,17 @@ static ParseStatus ParseVP8X(WebPDemuxer* const dmux) {
 
   dmux->is_ext_format_ = 1;
   Skip(mem, TAG_SIZE);  // VP8X
-  vp8x_size = GetLE32(mem);
+  vp8x_size = ReadLE32(mem);
   if (vp8x_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
   if (vp8x_size < VP8X_CHUNK_SIZE) return PARSE_ERROR;
   vp8x_size += vp8x_size & 1;
   if (SizeIsInvalid(mem, vp8x_size)) return PARSE_ERROR;
   if (MemDataSize(mem) < vp8x_size) return PARSE_NEED_MORE_DATA;
 
-  dmux->feature_flags_ = GetByte(mem);
+  dmux->feature_flags_ = ReadByte(mem);
   Skip(mem, 3);  // Reserved.
-  dmux->canvas_width_  = 1 + GetLE24s(mem);
-  dmux->canvas_height_ = 1 + GetLE24s(mem);
+  dmux->canvas_width_  = 1 + ReadLE24s(mem);
+  dmux->canvas_height_ = 1 + ReadLE24s(mem);
   if (dmux->canvas_width_ * (uint64_t)dmux->canvas_height_ >= MAX_IMAGE_AREA) {
     return PARSE_ERROR;  // image final dimension is too large
   }
@@ -468,8 +495,8 @@ static ParseStatus ParseVP8X(WebPDemuxer* const dmux) {
   do {
     int store_chunk = 1;
     const size_t chunk_start_offset = mem->start_;
-    const uint32_t fourcc = GetLE32(mem);
-    const uint32_t chunk_size = GetLE32(mem);
+    const uint32_t fourcc = ReadLE32(mem);
+    const uint32_t chunk_size = ReadLE32(mem);
     const uint32_t chunk_size_padded = chunk_size + (chunk_size & 1);
 
     if (chunk_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
@@ -482,40 +509,50 @@ static ParseStatus ParseVP8X(WebPDemuxer* const dmux) {
       case MKFOURCC('A', 'L', 'P', 'H'):
       case MKFOURCC('V', 'P', '8', ' '):
       case MKFOURCC('V', 'P', '8', 'L'): {
+        // check that this isn't an animation (all frames should be in an ANMF).
+        if (anim_chunks > 0) return PARSE_ERROR;
+
         Rewind(mem, CHUNK_HEADER_SIZE);
         status = ParseSingleImage(dmux);
         break;
       }
-      case MKFOURCC('L', 'O', 'O', 'P'): {
-        if (chunk_size_padded < LOOP_CHUNK_SIZE) return PARSE_ERROR;
+      case MKFOURCC('A', 'N', 'I', 'M'): {
+        if (chunk_size_padded < ANIM_CHUNK_SIZE) return PARSE_ERROR;
 
         if (MemDataSize(mem) < chunk_size_padded) {
           status = PARSE_NEED_MORE_DATA;
-        } else if (loop_chunks == 0) {
-          ++loop_chunks;
-          dmux->loop_count_ = GetLE16s(mem);
-          Skip(mem, chunk_size_padded - LOOP_CHUNK_SIZE);
+        } else if (anim_chunks == 0) {
+          ++anim_chunks;
+          dmux->bgcolor_ = ReadLE32(mem);
+          dmux->loop_count_ = ReadLE16s(mem);
+          Skip(mem, chunk_size_padded - ANIM_CHUNK_SIZE);
         } else {
           store_chunk = 0;
           goto Skip;
         }
         break;
       }
-      case MKFOURCC('F', 'R', 'M', ' '): {
-        status = ParseFrame(dmux, chunk_size_padded);
+      case MKFOURCC('A', 'N', 'M', 'F'): {
+        if (anim_chunks == 0) return PARSE_ERROR;  // 'ANIM' precedes frames.
+        status = ParseAnimationFrame(dmux, chunk_size_padded);
         break;
       }
-      case MKFOURCC('T', 'I', 'L', 'E'): {
-        if (dmux->num_frames_ == 0) dmux->num_frames_ = 1;
-        status = ParseTile(dmux, chunk_size_padded);
+#ifdef WEBP_EXPERIMENTAL_FEATURES
+      case MKFOURCC('F', 'R', 'G', 'M'): {
+        status = ParseFragment(dmux, chunk_size_padded);
         break;
       }
+#endif
       case MKFOURCC('I', 'C', 'C', 'P'): {
         store_chunk = !!(dmux->feature_flags_ & ICCP_FLAG);
         goto Skip;
       }
-      case MKFOURCC('M', 'E', 'T', 'A'): {
-        store_chunk = !!(dmux->feature_flags_ & META_FLAG);
+      case MKFOURCC('X', 'M', 'P', ' '): {
+        store_chunk = !!(dmux->feature_flags_ & XMP_FLAG);
+        goto Skip;
+      }
+      case MKFOURCC('E', 'X', 'I', 'F'): {
+        store_chunk = !!(dmux->feature_flags_ & EXIF_FLAG);
         goto Skip;
       }
  Skip:
@@ -561,7 +598,7 @@ static int IsValidSimpleFormat(const WebPDemuxer* const dmux) {
 }
 
 static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
-  const int has_tiles = !!(dmux->feature_flags_ & TILE_FLAG);
+  const int has_fragments = !!(dmux->feature_flags_ & FRAGMENTS_FLAG);
   const int has_frames = !!(dmux->feature_flags_ & ANIMATION_FLAG);
   const Frame* f;
 
@@ -573,15 +610,15 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
 
   for (f = dmux->frames_; f != NULL; f = f->next_) {
     const int cur_frame_set = f->frame_num_;
-    int frame_count = 0, tile_count = 0;
+    int frame_count = 0, fragment_count = 0;
 
-    // Check frame properties and if the image is composed of tiles that each
-    // fragment came from a 'TILE'.
+    // Check frame properties and if the image is composed of fragments that
+    // each fragment came from a fragment.
     for (; f != NULL && f->frame_num_ == cur_frame_set; f = f->next_) {
       const ChunkData* const image = f->img_components_;
       const ChunkData* const alpha = f->img_components_ + 1;
 
-      if (!has_tiles && f->is_tile_) return 0;
+      if (!has_fragments && f->is_fragment_) return 0;
       if (!has_frames && f->frame_num_ > 1) return 0;
       if (f->x_offset_ < 0 || f->y_offset_ < 0) return 0;
       if (f->complete_) {
@@ -593,6 +630,9 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
 
         if (f->width_ <= 0 || f->height_ <= 0) return 0;
       } else {
+        // There shouldn't be a partial frame in a complete file.
+        if (dmux->state_ == WEBP_DEMUX_DONE) return 0;
+
         // Ensure alpha precedes image bitstream.
         if (alpha->size_ > 0 && image->size_ > 0 &&
             alpha->offset_ > image->offset_) {
@@ -602,11 +642,11 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
         if (f->next_ != NULL) return 0;
       }
 
-      tile_count += f->is_tile_;
+      fragment_count += f->is_fragment_;
       ++frame_count;
     }
-    if (!has_tiles && frame_count > 1) return 0;
-    if (tile_count > 0 && frame_count != tile_count) return 0;
+    if (!has_fragments && frame_count > 1) return 0;
+    if (fragment_count > 0 && frame_count != fragment_count) return 0;
     if (f == NULL) break;
   }
   return 1;
@@ -618,8 +658,10 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
 static void InitDemux(WebPDemuxer* const dmux, const MemBuffer* const mem) {
   dmux->state_ = WEBP_DEMUX_PARSING_HEADER;
   dmux->loop_count_ = 1;
+  dmux->bgcolor_ = 0xFFFFFFFF;  // White background by default.
   dmux->canvas_width_ = -1;
   dmux->canvas_height_ = -1;
+  dmux->frames_tail_ = &dmux->frames_;
   dmux->mem_ = *mem;
 }
 
@@ -632,9 +674,9 @@ WebPDemuxer* WebPDemuxInternal(const WebPData* data, int allow_partial,
   WebPDemuxer* dmux;
 
   if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DEMUX_ABI_VERSION)) return NULL;
-  if (data == NULL || data->bytes_ == NULL || data->size_ == 0) return NULL;
+  if (data == NULL || data->bytes == NULL || data->size == 0) return NULL;
 
-  if (!InitMemBuffer(&mem, data->bytes_, data->size_)) return NULL;
+  if (!InitMemBuffer(&mem, data->bytes, data->size)) return NULL;
   if (!ReadHeader(&mem)) return NULL;
 
   partial = (mem.buf_size_ < mem.riff_end_);
@@ -648,6 +690,7 @@ WebPDemuxer* WebPDemuxInternal(const WebPData* data, int allow_partial,
     if (!memcmp(parser->id, GetBuffer(&dmux->mem_), TAG_SIZE)) {
       status = parser->parse(dmux);
       if (status == PARSE_OK) dmux->state_ = WEBP_DEMUX_DONE;
+      if (status == PARSE_NEED_MORE_DATA && !partial) status = PARSE_ERROR;
       if (status != PARSE_ERROR && !parser->valid(dmux)) status = PARSE_ERROR;
       break;
     }
@@ -685,10 +728,12 @@ uint32_t WebPDemuxGetI(const WebPDemuxer* dmux, WebPFormatFeature feature) {
   if (dmux == NULL) return 0;
 
   switch (feature) {
-    case WEBP_FF_FORMAT_FLAGS:  return dmux->feature_flags_;
-    case WEBP_FF_CANVAS_WIDTH:  return (uint32_t)dmux->canvas_width_;
-    case WEBP_FF_CANVAS_HEIGHT: return (uint32_t)dmux->canvas_height_;
-    case WEBP_FF_LOOP_COUNT:    return (uint32_t)dmux->loop_count_;
+    case WEBP_FF_FORMAT_FLAGS:     return dmux->feature_flags_;
+    case WEBP_FF_CANVAS_WIDTH:     return (uint32_t)dmux->canvas_width_;
+    case WEBP_FF_CANVAS_HEIGHT:    return (uint32_t)dmux->canvas_height_;
+    case WEBP_FF_LOOP_COUNT:       return (uint32_t)dmux->loop_count_;
+    case WEBP_FF_BACKGROUND_COLOR: return dmux->bgcolor_;
+    case WEBP_FF_FRAME_COUNT:      return (uint32_t)dmux->num_frames_;
   }
   return 0;
 }
@@ -696,7 +741,8 @@ uint32_t WebPDemuxGetI(const WebPDemuxer* dmux, WebPFormatFeature feature) {
 // -----------------------------------------------------------------------------
 // Frame iteration
 
-// Find the first 'frame_num' frame. There may be multiple in a tiled frame.
+// Find the first 'frame_num' frame. There may be multiple such frames in a
+// fragmented frame.
 static const Frame* GetFrame(const WebPDemuxer* const dmux, int frame_num) {
   const Frame* f;
   for (f = dmux->frames_; f != NULL; f = f->next_) {
@@ -705,19 +751,19 @@ static const Frame* GetFrame(const WebPDemuxer* const dmux, int frame_num) {
   return f;
 }
 
-// Returns tile 'tile_num' and the total count.
-static const Frame* GetTile(
-    const Frame* const frame_set, int tile_num, int* const count) {
+// Returns fragment 'fragment_num' and the total count.
+static const Frame* GetFragment(
+    const Frame* const frame_set, int fragment_num, int* const count) {
   const int this_frame = frame_set->frame_num_;
   const Frame* f = frame_set;
-  const Frame* tile = NULL;
+  const Frame* fragment = NULL;
   int total;
 
   for (total = 0; f != NULL && f->frame_num_ == this_frame; f = f->next_) {
-    if (++total == tile_num) tile = f;
+    if (++total == fragment_num) fragment = f;
   }
   *count = total;
-  return tile;
+  return fragment;
 }
 
 static const uint8_t* GetFramePayload(const uint8_t* const mem_buf,
@@ -747,27 +793,31 @@ static const uint8_t* GetFramePayload(const uint8_t* const mem_buf,
 // Create a whole 'frame' from VP8 (+ alpha) or lossless.
 static int SynthesizeFrame(const WebPDemuxer* const dmux,
                            const Frame* const first_frame,
-                           int tile_num, WebPIterator* const iter) {
+                           int fragment_num, WebPIterator* const iter) {
   const uint8_t* const mem_buf = dmux->mem_.buf_;
-  int num_tiles;
+  int num_fragments;
   size_t payload_size = 0;
-  const Frame* const tile = GetTile(first_frame, tile_num, &num_tiles);
-  const uint8_t* const payload = GetFramePayload(mem_buf, tile, &payload_size);
+  const Frame* const fragment =
+      GetFragment(first_frame, fragment_num, &num_fragments);
+  const uint8_t* const payload =
+      GetFramePayload(mem_buf, fragment, &payload_size);
   if (payload == NULL) return 0;
+  assert(first_frame != NULL);
 
-  iter->frame_num_   = first_frame->frame_num_;
-  iter->num_frames_  = dmux->num_frames_;
-  iter->tile_num_    = tile_num;
-  iter->num_tiles_   = num_tiles;
-  iter->x_offset_    = tile->x_offset_;
-  iter->y_offset_    = tile->y_offset_;
-  iter->width_       = tile->width_;
-  iter->height_      = tile->height_;
-  iter->duration_    = tile->duration_;
-  iter->complete_    = tile->complete_;
-  iter->tile_.bytes_ = payload;
-  iter->tile_.size_  = payload_size;
-  // TODO(jzern): adjust offsets for 'TILE's embedded in 'FRM 's
+  iter->frame_num      = first_frame->frame_num_;
+  iter->num_frames     = dmux->num_frames_;
+  iter->fragment_num   = fragment_num;
+  iter->num_fragments  = num_fragments;
+  iter->x_offset       = fragment->x_offset_;
+  iter->y_offset       = fragment->y_offset_;
+  iter->width          = fragment->width_;
+  iter->height         = fragment->height_;
+  iter->duration       = fragment->duration_;
+  iter->dispose_method = fragment->dispose_method_;
+  iter->complete       = fragment->complete_;
+  iter->fragment.bytes = payload;
+  iter->fragment.size  = payload_size;
+  // TODO(jzern): adjust offsets for 'FRGM's embedded in 'ANMF's
   return 1;
 }
 
@@ -779,6 +829,8 @@ static int SetFrame(int frame_num, WebPIterator* const iter) {
   if (frame_num == 0) frame_num = dmux->num_frames_;
 
   frame = GetFrame(dmux, frame_num);
+  if (frame == NULL) return 0;
+
   return SynthesizeFrame(dmux, frame, 1, iter);
 }
 
@@ -792,22 +844,22 @@ int WebPDemuxGetFrame(const WebPDemuxer* dmux, int frame, WebPIterator* iter) {
 
 int WebPDemuxNextFrame(WebPIterator* iter) {
   if (iter == NULL) return 0;
-  return SetFrame(iter->frame_num_ + 1, iter);
+  return SetFrame(iter->frame_num + 1, iter);
 }
 
 int WebPDemuxPrevFrame(WebPIterator* iter) {
   if (iter == NULL) return 0;
-  if (iter->frame_num_ <= 1) return 0;
-  return SetFrame(iter->frame_num_ - 1, iter);
+  if (iter->frame_num <= 1) return 0;
+  return SetFrame(iter->frame_num - 1, iter);
 }
 
-int WebPDemuxSelectTile(WebPIterator* iter, int tile) {
-  if (iter != NULL && iter->private_ != NULL && tile > 0) {
+int WebPDemuxSelectFragment(WebPIterator* iter, int fragment_num) {
+  if (iter != NULL && iter->private_ != NULL && fragment_num > 0) {
     const WebPDemuxer* const dmux = (WebPDemuxer*)iter->private_;
-    const Frame* const frame = GetFrame(dmux, iter->frame_num_);
+    const Frame* const frame = GetFrame(dmux, iter->frame_num);
     if (frame == NULL) return 0;
 
-    return SynthesizeFrame(dmux, frame, tile, iter);
+    return SynthesizeFrame(dmux, frame, fragment_num, iter);
   }
   return 0;
 }
@@ -856,10 +908,10 @@ static int SetChunk(const char fourcc[4], int chunk_num,
   if (chunk_num <= count) {
     const uint8_t* const mem_buf = dmux->mem_.buf_;
     const Chunk* const chunk = GetChunk(dmux, fourcc, chunk_num);
-    iter->chunk_.bytes_ = mem_buf + chunk->data_.offset_ + CHUNK_HEADER_SIZE;
-    iter->chunk_.size_  = chunk->data_.size_ - CHUNK_HEADER_SIZE;
-    iter->num_chunks_   = count;
-    iter->chunk_num_    = chunk_num;
+    iter->chunk.bytes = mem_buf + chunk->data_.offset_ + CHUNK_HEADER_SIZE;
+    iter->chunk.size  = chunk->data_.size_ - CHUNK_HEADER_SIZE;
+    iter->num_chunks  = count;
+    iter->chunk_num   = chunk_num;
     return 1;
   }
   return 0;
@@ -878,17 +930,17 @@ int WebPDemuxGetChunk(const WebPDemuxer* dmux,
 int WebPDemuxNextChunk(WebPChunkIterator* iter) {
   if (iter != NULL) {
     const char* const fourcc =
-        (const char*)iter->chunk_.bytes_ - CHUNK_HEADER_SIZE;
-    return SetChunk(fourcc, iter->chunk_num_ + 1, iter);
+        (const char*)iter->chunk.bytes - CHUNK_HEADER_SIZE;
+    return SetChunk(fourcc, iter->chunk_num + 1, iter);
   }
   return 0;
 }
 
 int WebPDemuxPrevChunk(WebPChunkIterator* iter) {
-  if (iter != NULL && iter->chunk_num_ > 1) {
+  if (iter != NULL && iter->chunk_num > 1) {
     const char* const fourcc =
-        (const char*)iter->chunk_.bytes_ - CHUNK_HEADER_SIZE;
-    return SetChunk(fourcc, iter->chunk_num_ - 1, iter);
+        (const char*)iter->chunk.bytes - CHUNK_HEADER_SIZE;
+    return SetChunk(fourcc, iter->chunk_num - 1, iter);
   }
   return 0;
 }
diff --git a/3rdparty/libwebp/dsp/cpu.c b/3rdparty/libwebp/dsp/cpu.c
index 0228734457..179901e100 100644
--- a/3rdparty/libwebp/dsp/cpu.c
+++ b/3rdparty/libwebp/dsp/cpu.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // CPU detection
diff --git a/3rdparty/libwebp/dsp/dec.c b/3rdparty/libwebp/dsp/dec.c
index 9ae7b6fa76..2fbd6b1ac4 100644
--- a/3rdparty/libwebp/dsp/dec.c
+++ b/3rdparty/libwebp/dsp/dec.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Speed-critical decoding functions.
@@ -426,11 +428,16 @@ static void HE8uv(uint8_t *dst) {    // horizontal
 }
 
 // helper for chroma-DC predictions
-static WEBP_INLINE void Put8x8uv(uint64_t v, uint8_t* dst) {
+static WEBP_INLINE void Put8x8uv(uint8_t value, uint8_t* dst) {
   int j;
+#ifndef WEBP_REFERENCE_IMPLEMENTATION
+  const uint64_t v = (uint64_t)value * 0x0101010101010101ULL;
   for (j = 0; j < 8; ++j) {
     *(uint64_t*)(dst + j * BPS) = v;
   }
+#else
+  for (j = 0; j < 8; ++j) memset(dst + j * BPS, value, 8);
+#endif
 }
 
 static void DC8uv(uint8_t *dst) {     // DC
@@ -439,7 +446,7 @@ static void DC8uv(uint8_t *dst) {     // DC
   for (i = 0; i < 8; ++i) {
     dc0 += dst[i - BPS] + dst[-1 + i * BPS];
   }
-  Put8x8uv((uint64_t)((dc0 >> 4) * 0x0101010101010101ULL), dst);
+  Put8x8uv(dc0 >> 4, dst);
 }
 
 static void DC8uvNoLeft(uint8_t *dst) {   // DC with no left samples
@@ -448,7 +455,7 @@ static void DC8uvNoLeft(uint8_t *dst) {   // DC with no left samples
   for (i = 0; i < 8; ++i) {
     dc0 += dst[i - BPS];
   }
-  Put8x8uv((uint64_t)((dc0 >> 3) * 0x0101010101010101ULL), dst);
+  Put8x8uv(dc0 >> 3, dst);
 }
 
 static void DC8uvNoTop(uint8_t *dst) {  // DC with no top samples
@@ -457,11 +464,11 @@ static void DC8uvNoTop(uint8_t *dst) {  // DC with no top samples
   for (i = 0; i < 8; ++i) {
     dc0 += dst[-1 + i * BPS];
   }
-  Put8x8uv((uint64_t)((dc0 >> 3) * 0x0101010101010101ULL), dst);
+  Put8x8uv(dc0 >> 3, dst);
 }
 
 static void DC8uvNoTopLeft(uint8_t *dst) {    // DC with nothing
-  Put8x8uv(0x8080808080808080ULL, dst);
+  Put8x8uv(0x80, dst);
 }
 
 //------------------------------------------------------------------------------
diff --git a/3rdparty/libwebp/dsp/dec_neon.c b/3rdparty/libwebp/dsp/dec_neon.c
index 09cafeef1c..5dcd3b7e5e 100644
--- a/3rdparty/libwebp/dsp/dec_neon.c
+++ b/3rdparty/libwebp/dsp/dec_neon.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // ARM NEON version of dsp functions and loop filtering.
@@ -79,7 +81,7 @@ extern "C" {
   "vld4.8   {" #c1"[6], " #c2"[6], " #c3"[6], " #c4"[6]}," #b1 "," #stride"\n" \
   "vld4.8   {" #c1"[7], " #c2"[7], " #c3"[7], " #c4"[7]}," #b2 "," #stride"\n"
 
-#define STORE8x2(c1, c2, p,stride)                                             \
+#define STORE8x2(c1, c2, p, stride)                                            \
   "vst2.8   {" #c1"[0], " #c2"[0]}," #p "," #stride " \n"                      \
   "vst2.8   {" #c1"[1], " #c2"[1]}," #p "," #stride " \n"                      \
   "vst2.8   {" #c1"[2], " #c2"[2]}," #p "," #stride " \n"                      \
@@ -155,6 +157,9 @@ static void SimpleHFilter16iNEON(uint8_t* p, int stride, int thresh) {
   }
 }
 
+//-----------------------------------------------------------------------------
+// Inverse transforms (Paragraph 14.4)
+
 static void TransformOneNEON(const int16_t *in, uint8_t *dst) {
   const int kBPS = BPS;
   const int16_t constants[] = {20091, 17734, 0, 0};
@@ -311,6 +316,73 @@ static void TransformTwoNEON(const int16_t* in, uint8_t* dst, int do_two) {
   }
 }
 
+static void TransformWHT(const int16_t* in, int16_t* out) {
+  const int kStep = 32;  // The store is only incrementing the pointer as if we
+                         // had stored a single byte.
+  __asm__ volatile (
+    // part 1
+    // load data into q0, q1
+    "vld1.16         {q0, q1}, [%[in]]           \n"
+
+    "vaddl.s16       q2, d0, d3                  \n" // a0 = in[0] + in[12]
+    "vaddl.s16       q3, d1, d2                  \n" // a1 = in[4] + in[8]
+    "vsubl.s16       q4, d1, d2                  \n" // a2 = in[4] - in[8]
+    "vsubl.s16       q5, d0, d3                  \n" // a3 = in[0] - in[12]
+
+    "vadd.s32        q0, q2, q3                  \n" // tmp[0] = a0 + a1
+    "vsub.s32        q2, q2, q3                  \n" // tmp[8] = a0 - a1
+    "vadd.s32        q1, q5, q4                  \n" // tmp[4] = a3 + a2
+    "vsub.s32        q3, q5, q4                  \n" // tmp[12] = a3 - a2
+
+    // Transpose
+    // q0 = tmp[0, 4, 8, 12], q1 = tmp[2, 6, 10, 14]
+    // q2 = tmp[1, 5, 9, 13], q3 = tmp[3, 7, 11, 15]
+    "vswp            d1, d4                      \n" // vtrn.64 q0, q2
+    "vswp            d3, d6                      \n" // vtrn.64 q1, q3
+    "vtrn.32         q0, q1                      \n"
+    "vtrn.32         q2, q3                      \n"
+
+    "vmov.s32        q4, #3                      \n" // dc = 3
+    "vadd.s32        q0, q0, q4                  \n" // dc = tmp[0] + 3
+    "vadd.s32        q6, q0, q3                  \n" // a0 = dc + tmp[3]
+    "vadd.s32        q7, q1, q2                  \n" // a1 = tmp[1] + tmp[2]
+    "vsub.s32        q8, q1, q2                  \n" // a2 = tmp[1] - tmp[2]
+    "vsub.s32        q9, q0, q3                  \n" // a3 = dc - tmp[3]
+
+    "vadd.s32        q0, q6, q7                  \n"
+    "vshrn.s32       d0, q0, #3                  \n" // (a0 + a1) >> 3
+    "vadd.s32        q1, q9, q8                  \n"
+    "vshrn.s32       d1, q1, #3                  \n" // (a3 + a2) >> 3
+    "vsub.s32        q2, q6, q7                  \n"
+    "vshrn.s32       d2, q2, #3                  \n" // (a0 - a1) >> 3
+    "vsub.s32        q3, q9, q8                  \n"
+    "vshrn.s32       d3, q3, #3                  \n" // (a3 - a2) >> 3
+
+    // set the results to output
+    "vst1.16         d0[0], [%[out]], %[kStep]   \n"
+    "vst1.16         d1[0], [%[out]], %[kStep]   \n"
+    "vst1.16         d2[0], [%[out]], %[kStep]   \n"
+    "vst1.16         d3[0], [%[out]], %[kStep]   \n"
+    "vst1.16         d0[1], [%[out]], %[kStep]   \n"
+    "vst1.16         d1[1], [%[out]], %[kStep]   \n"
+    "vst1.16         d2[1], [%[out]], %[kStep]   \n"
+    "vst1.16         d3[1], [%[out]], %[kStep]   \n"
+    "vst1.16         d0[2], [%[out]], %[kStep]   \n"
+    "vst1.16         d1[2], [%[out]], %[kStep]   \n"
+    "vst1.16         d2[2], [%[out]], %[kStep]   \n"
+    "vst1.16         d3[2], [%[out]], %[kStep]   \n"
+    "vst1.16         d0[3], [%[out]], %[kStep]   \n"
+    "vst1.16         d1[3], [%[out]], %[kStep]   \n"
+    "vst1.16         d2[3], [%[out]], %[kStep]   \n"
+    "vst1.16         d3[3], [%[out]], %[kStep]   \n"
+
+    : [out] "+r"(out)  // modified registers
+    : [in] "r"(in), [kStep] "r"(kStep)  // constants
+    : "memory", "q0", "q1", "q2", "q3", "q4",
+      "q5", "q6", "q7", "q8", "q9"  // clobbered
+  );
+}
+
 #endif   // WEBP_USE_NEON
 
 //------------------------------------------------------------------------------
@@ -321,6 +393,7 @@ extern void VP8DspInitNEON(void);
 void VP8DspInitNEON(void) {
 #if defined(WEBP_USE_NEON)
   VP8Transform = TransformTwoNEON;
+  VP8TransformWHT = TransformWHT;
 
   VP8SimpleVFilter16 = SimpleVFilter16NEON;
   VP8SimpleHFilter16 = SimpleHFilter16NEON;
diff --git a/3rdparty/libwebp/dsp/dec_sse2.c b/3rdparty/libwebp/dsp/dec_sse2.c
index 2cb55d013f..6be94678ef 100644
--- a/3rdparty/libwebp/dsp/dec_sse2.c
+++ b/3rdparty/libwebp/dsp/dec_sse2.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // SSE2 version of some decoding functions (idct, loop filtering).
@@ -194,7 +196,7 @@ static void TransformSSE2(const int16_t* in, uint8_t* dst, int do_two) {
 
   // Add inverse transform to 'dst' and store.
   {
-    const __m128i zero = _mm_set1_epi16(0);
+    const __m128i zero = _mm_setzero_si128();
     // Load the reference(s).
     __m128i dst0, dst1, dst2, dst3;
     if (do_two) {
@@ -278,14 +280,14 @@ static void TransformSSE2(const int16_t* in, uint8_t* dst, int do_two) {
 
 #define GET_NOTHEV(p1, p0, q0, q1, hev_thresh, not_hev) {                      \
   const __m128i zero = _mm_setzero_si128();                                    \
-  const __m128i t1 = MM_ABS(p1, p0);                                           \
-  const __m128i t2 = MM_ABS(q1, q0);                                           \
+  const __m128i t_1 = MM_ABS(p1, p0);                                          \
+  const __m128i t_2 = MM_ABS(q1, q0);                                          \
                                                                                \
   const __m128i h = _mm_set1_epi8(hev_thresh);                                 \
-  const __m128i t3 = _mm_subs_epu8(t1, h);  /* abs(p1 - p0) - hev_tresh */     \
-  const __m128i t4 = _mm_subs_epu8(t2, h);  /* abs(q1 - q0) - hev_tresh */     \
+  const __m128i t_3 = _mm_subs_epu8(t_1, h);  /* abs(p1 - p0) - hev_tresh */   \
+  const __m128i t_4 = _mm_subs_epu8(t_2, h);  /* abs(q1 - q0) - hev_tresh */   \
                                                                                \
-  not_hev = _mm_or_si128(t3, t4);                                              \
+  not_hev = _mm_or_si128(t_3, t_4);                                            \
   not_hev = _mm_cmpeq_epi8(not_hev, zero); /* not_hev <= t1 && not_hev <= t2 */\
 }
 
@@ -314,13 +316,13 @@ static void TransformSSE2(const int16_t* in, uint8_t* dst, int do_two) {
 
 // Updates values of 2 pixels at MB edge during complex filtering.
 // Update operations:
-// q = q - a and p = p + a; where a = [(a_hi >> 7), (a_lo >> 7)]
+// q = q - delta and p = p + delta; where delta = [(a_hi >> 7), (a_lo >> 7)]
 #define UPDATE_2PIXELS(pi, qi, a_lo, a_hi) {                                   \
   const __m128i a_lo7 = _mm_srai_epi16(a_lo, 7);                               \
   const __m128i a_hi7 = _mm_srai_epi16(a_hi, 7);                               \
-  const __m128i a = _mm_packs_epi16(a_lo7, a_hi7);                             \
-  pi = _mm_adds_epi8(pi, a);                                                   \
-  qi = _mm_subs_epi8(qi, a);                                                   \
+  const __m128i delta = _mm_packs_epi16(a_lo7, a_hi7);                         \
+  pi = _mm_adds_epi8(pi, delta);                                               \
+  qi = _mm_subs_epi8(qi, delta);                                               \
 }
 
 static void NeedsFilter(const __m128i* p1, const __m128i* p0, const __m128i* q0,
diff --git a/3rdparty/libwebp/dsp/dsp.h b/3rdparty/libwebp/dsp/dsp.h
index 24d0c188c4..01a95891fe 100644
--- a/3rdparty/libwebp/dsp/dsp.h
+++ b/3rdparty/libwebp/dsp/dsp.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //   Speed-critical functions.
@@ -49,8 +51,6 @@ extern VP8CPUInfo VP8GetCPUInfo;
 //------------------------------------------------------------------------------
 // Encoding
 
-int VP8GetAlpha(const int histo[]);
-
 // Transforms
 // VP8Idct: Does one of two inverse transforms. If do_two is set, the transforms
 //          will be done for (ref, in, dst) and (ref + 4, in + 16, dst + 4).
@@ -85,10 +85,11 @@ typedef int (*VP8QuantizeBlock)(int16_t in[16], int16_t out[16],
                                 int n, const struct VP8Matrix* const mtx);
 extern VP8QuantizeBlock VP8EncQuantizeBlock;
 
-// Compute susceptibility based on DCT-coeff histograms:
-// the higher, the "easier" the macroblock is to compress.
-typedef int (*VP8CHisto)(const uint8_t* ref, const uint8_t* pred,
-                         int start_block, int end_block);
+// Collect histogram for susceptibility calculation and accumulate in histo[].
+struct VP8Histogram;
+typedef void (*VP8CHisto)(const uint8_t* ref, const uint8_t* pred,
+                          int start_block, int end_block,
+                          struct VP8Histogram* const histo);
 extern const int VP8DspScan[16 + 4 + 4];
 extern VP8CHisto VP8CollectHistogram;
 
@@ -104,7 +105,7 @@ extern VP8DecIdct2 VP8Transform;
 extern VP8DecIdct VP8TransformUV;
 extern VP8DecIdct VP8TransformDC;
 extern VP8DecIdct VP8TransformDCUV;
-extern void (*VP8TransformWHT)(const int16_t* in, int16_t* out);
+extern VP8WHT VP8TransformWHT;
 
 // *dst is the destination block, with stride BPS. Boundary samples are
 // assumed accessible when needed.
@@ -159,6 +160,9 @@ extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */];
 // Initializes SSE2 version of the fancy upsamplers.
 void WebPInitUpsamplersSSE2(void);
 
+// NEON version
+void WebPInitUpsamplersNEON(void);
+
 #endif    // FANCY_UPSAMPLING
 
 // Point-sampling methods.
@@ -200,6 +204,7 @@ extern void (*WebPApplyAlphaMultiply4444)(
 void WebPInitPremultiply(void);
 
 void WebPInitPremultiplySSE2(void);   // should not be called directly.
+void WebPInitPremultiplyNEON(void);
 
 //------------------------------------------------------------------------------
 
diff --git a/3rdparty/libwebp/dsp/enc.c b/3rdparty/libwebp/dsp/enc.c
index 02234564be..552807ad95 100644
--- a/3rdparty/libwebp/dsp/enc.c
+++ b/3rdparty/libwebp/dsp/enc.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Speed-critical encoding functions.
@@ -17,31 +19,18 @@
 extern "C" {
 #endif
 
+static WEBP_INLINE uint8_t clip_8b(int v) {
+  return (!(v & ~0xff)) ? v : (v < 0) ? 0 : 255;
+}
+
+static WEBP_INLINE int clip_max(int v, int max) {
+  return (v > max) ? max : v;
+}
+
 //------------------------------------------------------------------------------
 // Compute susceptibility based on DCT-coeff histograms:
 // the higher, the "easier" the macroblock is to compress.
 
-static int ClipAlpha(int alpha) {
-  return alpha < 0 ? 0 : alpha > 255 ? 255 : alpha;
-}
-
-int VP8GetAlpha(const int histo[MAX_COEFF_THRESH + 1]) {
-  int num = 0, den = 0, val = 0;
-  int k;
-  int alpha;
-  // note: changing this loop to avoid the numerous "k + 1" slows things down.
-  for (k = 0; k < MAX_COEFF_THRESH; ++k) {
-    if (histo[k + 1]) {
-      val += histo[k + 1];
-      num += val * (k + 1);
-      den += (k + 1) * (k + 1);
-    }
-  }
-  // we scale the value to a usable [0..255] range
-  alpha = den ? 10 * num / den - 5 : 0;
-  return ClipAlpha(alpha);
-}
-
 const int VP8DspScan[16 + 4 + 4] = {
   // Luma
   0 +  0 * BPS,  4 +  0 * BPS, 8 +  0 * BPS, 12 +  0 * BPS,
@@ -53,27 +42,23 @@ const int VP8DspScan[16 + 4 + 4] = {
   8 + 0 * BPS,  12 + 0 * BPS, 8 + 4 * BPS, 12 + 4 * BPS     // V
 };
 
-static int CollectHistogram(const uint8_t* ref, const uint8_t* pred,
-                            int start_block, int end_block) {
-  int histo[MAX_COEFF_THRESH + 1] = { 0 };
-  int16_t out[16];
-  int j, k;
+static void CollectHistogram(const uint8_t* ref, const uint8_t* pred,
+                             int start_block, int end_block,
+                             VP8Histogram* const histo) {
+  int j;
   for (j = start_block; j < end_block; ++j) {
+    int k;
+    int16_t out[16];
+
     VP8FTransform(ref + VP8DspScan[j], pred + VP8DspScan[j], out);
 
-    // Convert coefficients to bin (within out[]).
+    // Convert coefficients to bin.
     for (k = 0; k < 16; ++k) {
-      const int v = abs(out[k]) >> 2;
-      out[k] = (v > MAX_COEFF_THRESH) ? MAX_COEFF_THRESH : v;
-    }
-
-    // Use bin to update histogram.
-    for (k = 0; k < 16; ++k) {
-      histo[out[k]]++;
+      const int v = abs(out[k]) >> 3;  // TODO(skal): add rounding?
+      const int clipped_value = clip_max(v, MAX_COEFF_THRESH);
+      histo->distribution[clipped_value]++;
     }
   }
-
-  return VP8GetAlpha(histo);
 }
 
 //------------------------------------------------------------------------------
@@ -89,15 +74,12 @@ static void InitTables(void) {
   if (!tables_ok) {
     int i;
     for (i = -255; i <= 255 + 255; ++i) {
-      clip1[255 + i] = (i < 0) ? 0 : (i > 255) ? 255 : i;
+      clip1[255 + i] = clip_8b(i);
     }
     tables_ok = 1;
   }
 }
 
-static WEBP_INLINE uint8_t clip_8b(int v) {
-  return (!(v & ~0xff)) ? v : v < 0 ? 0 : 255;
-}
 
 //------------------------------------------------------------------------------
 // Transforms (Paragraph 14.4)
@@ -154,25 +136,25 @@ static void FTransform(const uint8_t* src, const uint8_t* ref, int16_t* out) {
   int i;
   int tmp[16];
   for (i = 0; i < 4; ++i, src += BPS, ref += BPS) {
-    const int d0 = src[0] - ref[0];
+    const int d0 = src[0] - ref[0];   // 9bit dynamic range ([-255,255])
     const int d1 = src[1] - ref[1];
     const int d2 = src[2] - ref[2];
     const int d3 = src[3] - ref[3];
-    const int a0 = (d0 + d3) << 3;
-    const int a1 = (d1 + d2) << 3;
-    const int a2 = (d1 - d2) << 3;
-    const int a3 = (d0 - d3) << 3;
-    tmp[0 + i * 4] = (a0 + a1);
-    tmp[1 + i * 4] = (a2 * 2217 + a3 * 5352 + 14500) >> 12;
-    tmp[2 + i * 4] = (a0 - a1);
-    tmp[3 + i * 4] = (a3 * 2217 - a2 * 5352 +  7500) >> 12;
+    const int a0 = (d0 + d3);         // 10b                      [-510,510]
+    const int a1 = (d1 + d2);
+    const int a2 = (d1 - d2);
+    const int a3 = (d0 - d3);
+    tmp[0 + i * 4] = (a0 + a1) * 8;   // 14b                      [-8160,8160]
+    tmp[1 + i * 4] = (a2 * 2217 + a3 * 5352 + 1812) >> 9;      // [-7536,7542]
+    tmp[2 + i * 4] = (a0 - a1) * 8;
+    tmp[3 + i * 4] = (a3 * 2217 - a2 * 5352 +  937) >> 9;
   }
   for (i = 0; i < 4; ++i) {
-    const int a0 = (tmp[0 + i] + tmp[12 + i]);
+    const int a0 = (tmp[0 + i] + tmp[12 + i]);  // 15b
     const int a1 = (tmp[4 + i] + tmp[ 8 + i]);
     const int a2 = (tmp[4 + i] - tmp[ 8 + i]);
     const int a3 = (tmp[0 + i] - tmp[12 + i]);
-    out[0 + i] = (a0 + a1 + 7) >> 4;
+    out[0 + i] = (a0 + a1 + 7) >> 4;            // 12b
     out[4 + i] = ((a2 * 2217 + a3 * 5352 + 12000) >> 16) + (a3 != 0);
     out[8 + i] = (a0 - a1 + 7) >> 4;
     out[12+ i] = ((a3 * 2217 - a2 * 5352 + 51000) >> 16);
@@ -207,31 +189,32 @@ static void ITransformWHT(const int16_t* in, int16_t* out) {
 }
 
 static void FTransformWHT(const int16_t* in, int16_t* out) {
-  int tmp[16];
+  // input is 12b signed
+  int16_t tmp[16];
   int i;
   for (i = 0; i < 4; ++i, in += 64) {
-    const int a0 = (in[0 * 16] + in[2 * 16]) << 2;
-    const int a1 = (in[1 * 16] + in[3 * 16]) << 2;
-    const int a2 = (in[1 * 16] - in[3 * 16]) << 2;
-    const int a3 = (in[0 * 16] - in[2 * 16]) << 2;
-    tmp[0 + i * 4] = (a0 + a1) + (a0 != 0);
+    const int a0 = (in[0 * 16] + in[2 * 16]);  // 13b
+    const int a1 = (in[1 * 16] + in[3 * 16]);
+    const int a2 = (in[1 * 16] - in[3 * 16]);
+    const int a3 = (in[0 * 16] - in[2 * 16]);
+    tmp[0 + i * 4] = a0 + a1;   // 14b
     tmp[1 + i * 4] = a3 + a2;
     tmp[2 + i * 4] = a3 - a2;
     tmp[3 + i * 4] = a0 - a1;
   }
   for (i = 0; i < 4; ++i) {
-    const int a0 = (tmp[0 + i] + tmp[8 + i]);
+    const int a0 = (tmp[0 + i] + tmp[8 + i]);  // 15b
     const int a1 = (tmp[4 + i] + tmp[12+ i]);
     const int a2 = (tmp[4 + i] - tmp[12+ i]);
     const int a3 = (tmp[0 + i] - tmp[8 + i]);
-    const int b0 = a0 + a1;
+    const int b0 = a0 + a1;    // 16b
     const int b1 = a3 + a2;
     const int b2 = a3 - a2;
     const int b3 = a0 - a1;
-    out[ 0 + i] = (b0 + (b0 > 0) + 3) >> 3;
-    out[ 4 + i] = (b1 + (b1 > 0) + 3) >> 3;
-    out[ 8 + i] = (b2 + (b2 > 0) + 3) >> 3;
-    out[12 + i] = (b3 + (b3 > 0) + 3) >> 3;
+    out[ 0 + i] = b0 >> 1;     // 15b
+    out[ 4 + i] = b1 >> 1;
+    out[ 8 + i] = b2 >> 1;
+    out[12 + i] = b3 >> 1;
   }
 }
 
@@ -589,30 +572,30 @@ static int TTransform(const uint8_t* in, const uint16_t* w) {
   int i;
   // horizontal pass
   for (i = 0; i < 4; ++i, in += BPS) {
-    const int a0 = (in[0] + in[2]) << 2;
-    const int a1 = (in[1] + in[3]) << 2;
-    const int a2 = (in[1] - in[3]) << 2;
-    const int a3 = (in[0] - in[2]) << 2;
-    tmp[0 + i * 4] = a0 + a1 + (a0 != 0);
+    const int a0 = in[0] + in[2];
+    const int a1 = in[1] + in[3];
+    const int a2 = in[1] - in[3];
+    const int a3 = in[0] - in[2];
+    tmp[0 + i * 4] = a0 + a1;
     tmp[1 + i * 4] = a3 + a2;
     tmp[2 + i * 4] = a3 - a2;
     tmp[3 + i * 4] = a0 - a1;
   }
   // vertical pass
   for (i = 0; i < 4; ++i, ++w) {
-    const int a0 = (tmp[0 + i] + tmp[8 + i]);
-    const int a1 = (tmp[4 + i] + tmp[12+ i]);
-    const int a2 = (tmp[4 + i] - tmp[12+ i]);
-    const int a3 = (tmp[0 + i] - tmp[8 + i]);
+    const int a0 = tmp[0 + i] + tmp[8 + i];
+    const int a1 = tmp[4 + i] + tmp[12+ i];
+    const int a2 = tmp[4 + i] - tmp[12+ i];
+    const int a3 = tmp[0 + i] - tmp[8 + i];
     const int b0 = a0 + a1;
     const int b1 = a3 + a2;
     const int b2 = a3 - a2;
     const int b3 = a0 - a1;
-    // abs((b + (b<0) + 3) >> 3) = (abs(b) + 3) >> 3
-    sum += w[ 0] * ((abs(b0) + 3) >> 3);
-    sum += w[ 4] * ((abs(b1) + 3) >> 3);
-    sum += w[ 8] * ((abs(b2) + 3) >> 3);
-    sum += w[12] * ((abs(b3) + 3) >> 3);
+
+    sum += w[ 0] * abs(b0);
+    sum += w[ 4] * abs(b1);
+    sum += w[ 8] * abs(b2);
+    sum += w[12] * abs(b3);
   }
   return sum;
 }
@@ -621,7 +604,7 @@ static int Disto4x4(const uint8_t* const a, const uint8_t* const b,
                     const uint16_t* const w) {
   const int sum1 = TTransform(a, w);
   const int sum2 = TTransform(b, w);
-  return (abs(sum2 - sum1) + 8) >> 4;
+  return abs(sum2 - sum1) >> 5;
 }
 
 static int Disto16x16(const uint8_t* const a, const uint8_t* const b,
@@ -651,13 +634,13 @@ static int QuantizeBlock(int16_t in[16], int16_t out[16],
   for (; n < 16; ++n) {
     const int j = kZigzag[n];
     const int sign = (in[j] < 0);
-    int coeff = (sign ? -in[j] : in[j]) + mtx->sharpen_[j];
-    if (coeff > 2047) coeff = 2047;
+    const int coeff = (sign ? -in[j] : in[j]) + mtx->sharpen_[j];
     if (coeff > mtx->zthresh_[j]) {
       const int Q = mtx->q_[j];
       const int iQ = mtx->iq_[j];
       const int B = mtx->bias_[j];
       out[n] = QUANTDIV(coeff, iQ, B);
+      if (out[n] > MAX_LEVEL) out[n] = MAX_LEVEL;
       if (sign) out[n] = -out[n];
       in[j] = out[n] * Q;
       if (out[n]) last = n;
@@ -706,6 +689,7 @@ VP8QuantizeBlock VP8EncQuantizeBlock;
 VP8BlockCopy VP8Copy4x4;
 
 extern void VP8EncDspInitSSE2(void);
+extern void VP8EncDspInitNEON(void);
 
 void VP8EncDspInit(void) {
   InitTables();
@@ -734,6 +718,10 @@ void VP8EncDspInit(void) {
     if (VP8GetCPUInfo(kSSE2)) {
       VP8EncDspInitSSE2();
     }
+#elif defined(WEBP_USE_NEON)
+    if (VP8GetCPUInfo(kNEON)) {
+      VP8EncDspInitNEON();
+    }
 #endif
   }
 }
diff --git a/3rdparty/libwebp/dsp/enc_neon.c b/3rdparty/libwebp/dsp/enc_neon.c
new file mode 100644
index 0000000000..eb256e681d
--- /dev/null
+++ b/3rdparty/libwebp/dsp/enc_neon.c
@@ -0,0 +1,639 @@
+// Copyright 2012 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// ARM NEON version of speed-critical encoding functions.
+//
+// adapted from libvpx (http://www.webmproject.org/code/)
+
+#include "./dsp.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+#if defined(WEBP_USE_NEON)
+
+#include "../enc/vp8enci.h"
+
+//------------------------------------------------------------------------------
+// Transforms (Paragraph 14.4)
+
+// Inverse transform.
+// This code is pretty much the same as TransformOneNEON in the decoder, except
+// for subtraction to *ref. See the comments there for algorithmic explanations.
+static void ITransformOne(const uint8_t* ref,
+                          const int16_t* in, uint8_t* dst) {
+  const int kBPS = BPS;
+  const int16_t kC1C2[] = { 20091, 17734, 0, 0 };  // kC1 / (kC2 >> 1) / 0 / 0
+
+  __asm__ volatile (
+    "vld1.16         {q1, q2}, [%[in]]           \n"
+    "vld1.16         {d0}, [%[kC1C2]]            \n"
+
+    // d2: in[0]
+    // d3: in[8]
+    // d4: in[4]
+    // d5: in[12]
+    "vswp            d3, d4                      \n"
+
+    // q8 = {in[4], in[12]} * kC1 * 2 >> 16
+    // q9 = {in[4], in[12]} * kC2 >> 16
+    "vqdmulh.s16     q8, q2, d0[0]               \n"
+    "vqdmulh.s16     q9, q2, d0[1]               \n"
+
+    // d22 = a = in[0] + in[8]
+    // d23 = b = in[0] - in[8]
+    "vqadd.s16       d22, d2, d3                 \n"
+    "vqsub.s16       d23, d2, d3                 \n"
+
+    //  q8 = in[4]/[12] * kC1 >> 16
+    "vshr.s16        q8, q8, #1                  \n"
+
+    // Add {in[4], in[12]} back after the multiplication.
+    "vqadd.s16       q8, q2, q8                  \n"
+
+    // d20 = c = in[4]*kC2 - in[12]*kC1
+    // d21 = d = in[4]*kC1 + in[12]*kC2
+    "vqsub.s16       d20, d18, d17               \n"
+    "vqadd.s16       d21, d19, d16               \n"
+
+    // d2 = tmp[0] = a + d
+    // d3 = tmp[1] = b + c
+    // d4 = tmp[2] = b - c
+    // d5 = tmp[3] = a - d
+    "vqadd.s16       d2, d22, d21                \n"
+    "vqadd.s16       d3, d23, d20                \n"
+    "vqsub.s16       d4, d23, d20                \n"
+    "vqsub.s16       d5, d22, d21                \n"
+
+    "vzip.16         q1, q2                      \n"
+    "vzip.16         q1, q2                      \n"
+
+    "vswp            d3, d4                      \n"
+
+    // q8 = {tmp[4], tmp[12]} * kC1 * 2 >> 16
+    // q9 = {tmp[4], tmp[12]} * kC2 >> 16
+    "vqdmulh.s16     q8, q2, d0[0]               \n"
+    "vqdmulh.s16     q9, q2, d0[1]               \n"
+
+    // d22 = a = tmp[0] + tmp[8]
+    // d23 = b = tmp[0] - tmp[8]
+    "vqadd.s16       d22, d2, d3                 \n"
+    "vqsub.s16       d23, d2, d3                 \n"
+
+    "vshr.s16        q8, q8, #1                  \n"
+    "vqadd.s16       q8, q2, q8                  \n"
+
+    // d20 = c = in[4]*kC2 - in[12]*kC1
+    // d21 = d = in[4]*kC1 + in[12]*kC2
+    "vqsub.s16       d20, d18, d17               \n"
+    "vqadd.s16       d21, d19, d16               \n"
+
+    // d2 = tmp[0] = a + d
+    // d3 = tmp[1] = b + c
+    // d4 = tmp[2] = b - c
+    // d5 = tmp[3] = a - d
+    "vqadd.s16       d2, d22, d21                \n"
+    "vqadd.s16       d3, d23, d20                \n"
+    "vqsub.s16       d4, d23, d20                \n"
+    "vqsub.s16       d5, d22, d21                \n"
+
+    "vld1.32         d6[0], [%[ref]], %[kBPS]    \n"
+    "vld1.32         d6[1], [%[ref]], %[kBPS]    \n"
+    "vld1.32         d7[0], [%[ref]], %[kBPS]    \n"
+    "vld1.32         d7[1], [%[ref]], %[kBPS]    \n"
+
+    "sub         %[ref], %[ref], %[kBPS], lsl #2 \n"
+
+    // (val) + 4 >> 3
+    "vrshr.s16       d2, d2, #3                  \n"
+    "vrshr.s16       d3, d3, #3                  \n"
+    "vrshr.s16       d4, d4, #3                  \n"
+    "vrshr.s16       d5, d5, #3                  \n"
+
+    "vzip.16         q1, q2                      \n"
+    "vzip.16         q1, q2                      \n"
+
+    // Must accumulate before saturating
+    "vmovl.u8        q8, d6                      \n"
+    "vmovl.u8        q9, d7                      \n"
+
+    "vqadd.s16       q1, q1, q8                  \n"
+    "vqadd.s16       q2, q2, q9                  \n"
+
+    "vqmovun.s16     d0, q1                      \n"
+    "vqmovun.s16     d1, q2                      \n"
+
+    "vst1.32         d0[0], [%[dst]], %[kBPS]    \n"
+    "vst1.32         d0[1], [%[dst]], %[kBPS]    \n"
+    "vst1.32         d1[0], [%[dst]], %[kBPS]    \n"
+    "vst1.32         d1[1], [%[dst]]             \n"
+
+    : [in] "+r"(in), [dst] "+r"(dst)               // modified registers
+    : [kBPS] "r"(kBPS), [kC1C2] "r"(kC1C2), [ref] "r"(ref)  // constants
+    : "memory", "q0", "q1", "q2", "q8", "q9", "q10", "q11"  // clobbered
+  );
+}
+
+static void ITransform(const uint8_t* ref,
+                       const int16_t* in, uint8_t* dst, int do_two) {
+  ITransformOne(ref, in, dst);
+  if (do_two) {
+    ITransformOne(ref + 4, in + 16, dst + 4);
+  }
+}
+
+// Same code as dec_neon.c
+static void ITransformWHT(const int16_t* in, int16_t* out) {
+  const int kStep = 32;  // The store is only incrementing the pointer as if we
+                         // had stored a single byte.
+  __asm__ volatile (
+    // part 1
+    // load data into q0, q1
+    "vld1.16         {q0, q1}, [%[in]]           \n"
+
+    "vaddl.s16       q2, d0, d3                  \n" // a0 = in[0] + in[12]
+    "vaddl.s16       q3, d1, d2                  \n" // a1 = in[4] + in[8]
+    "vsubl.s16       q4, d1, d2                  \n" // a2 = in[4] - in[8]
+    "vsubl.s16       q5, d0, d3                  \n" // a3 = in[0] - in[12]
+
+    "vadd.s32        q0, q2, q3                  \n" // tmp[0] = a0 + a1
+    "vsub.s32        q2, q2, q3                  \n" // tmp[8] = a0 - a1
+    "vadd.s32        q1, q5, q4                  \n" // tmp[4] = a3 + a2
+    "vsub.s32        q3, q5, q4                  \n" // tmp[12] = a3 - a2
+
+    // Transpose
+    // q0 = tmp[0, 4, 8, 12], q1 = tmp[2, 6, 10, 14]
+    // q2 = tmp[1, 5, 9, 13], q3 = tmp[3, 7, 11, 15]
+    "vswp            d1, d4                      \n" // vtrn.64 q0, q2
+    "vswp            d3, d6                      \n" // vtrn.64 q1, q3
+    "vtrn.32         q0, q1                      \n"
+    "vtrn.32         q2, q3                      \n"
+
+    "vmov.s32        q4, #3                      \n" // dc = 3
+    "vadd.s32        q0, q0, q4                  \n" // dc = tmp[0] + 3
+    "vadd.s32        q6, q0, q3                  \n" // a0 = dc + tmp[3]
+    "vadd.s32        q7, q1, q2                  \n" // a1 = tmp[1] + tmp[2]
+    "vsub.s32        q8, q1, q2                  \n" // a2 = tmp[1] - tmp[2]
+    "vsub.s32        q9, q0, q3                  \n" // a3 = dc - tmp[3]
+
+    "vadd.s32        q0, q6, q7                  \n"
+    "vshrn.s32       d0, q0, #3                  \n" // (a0 + a1) >> 3
+    "vadd.s32        q1, q9, q8                  \n"
+    "vshrn.s32       d1, q1, #3                  \n" // (a3 + a2) >> 3
+    "vsub.s32        q2, q6, q7                  \n"
+    "vshrn.s32       d2, q2, #3                  \n" // (a0 - a1) >> 3
+    "vsub.s32        q3, q9, q8                  \n"
+    "vshrn.s32       d3, q3, #3                  \n" // (a3 - a2) >> 3
+
+    // set the results to output
+    "vst1.16         d0[0], [%[out]], %[kStep]      \n"
+    "vst1.16         d1[0], [%[out]], %[kStep]      \n"
+    "vst1.16         d2[0], [%[out]], %[kStep]      \n"
+    "vst1.16         d3[0], [%[out]], %[kStep]      \n"
+    "vst1.16         d0[1], [%[out]], %[kStep]      \n"
+    "vst1.16         d1[1], [%[out]], %[kStep]      \n"
+    "vst1.16         d2[1], [%[out]], %[kStep]      \n"
+    "vst1.16         d3[1], [%[out]], %[kStep]      \n"
+    "vst1.16         d0[2], [%[out]], %[kStep]      \n"
+    "vst1.16         d1[2], [%[out]], %[kStep]      \n"
+    "vst1.16         d2[2], [%[out]], %[kStep]      \n"
+    "vst1.16         d3[2], [%[out]], %[kStep]      \n"
+    "vst1.16         d0[3], [%[out]], %[kStep]      \n"
+    "vst1.16         d1[3], [%[out]], %[kStep]      \n"
+    "vst1.16         d2[3], [%[out]], %[kStep]      \n"
+    "vst1.16         d3[3], [%[out]], %[kStep]      \n"
+
+    : [out] "+r"(out)  // modified registers
+    : [in] "r"(in), [kStep] "r"(kStep)  // constants
+    : "memory", "q0", "q1", "q2", "q3", "q4",
+      "q5", "q6", "q7", "q8", "q9" // clobbered
+  );
+}
+
+// Forward transform.
+
+// adapted from vp8/encoder/arm/neon/shortfdct_neon.asm
+static const int16_t kCoeff16[] = {
+  5352,  5352,  5352, 5352, 2217,  2217,  2217, 2217
+};
+static const int32_t kCoeff32[] = {
+   1812,  1812,  1812,  1812,
+    937,   937,   937,   937,
+  12000, 12000, 12000, 12000,
+  51000, 51000, 51000, 51000
+};
+
+static void FTransform(const uint8_t* src, const uint8_t* ref,
+                       int16_t* out) {
+  const int kBPS = BPS;
+  const uint8_t* src_ptr = src;
+  const uint8_t* ref_ptr = ref;
+  const int16_t* coeff16 = kCoeff16;
+  const int32_t* coeff32 = kCoeff32;
+
+  __asm__ volatile (
+    // load src into q4, q5 in high half
+    "vld1.8 {d8},  [%[src_ptr]], %[kBPS]      \n"
+    "vld1.8 {d10}, [%[src_ptr]], %[kBPS]      \n"
+    "vld1.8 {d9},  [%[src_ptr]], %[kBPS]      \n"
+    "vld1.8 {d11}, [%[src_ptr]]               \n"
+
+    // load ref into q6, q7 in high half
+    "vld1.8 {d12}, [%[ref_ptr]], %[kBPS]      \n"
+    "vld1.8 {d14}, [%[ref_ptr]], %[kBPS]      \n"
+    "vld1.8 {d13}, [%[ref_ptr]], %[kBPS]      \n"
+    "vld1.8 {d15}, [%[ref_ptr]]               \n"
+
+    // Pack the high values in to q4 and q6
+    "vtrn.32     q4, q5                       \n"
+    "vtrn.32     q6, q7                       \n"
+
+    // d[0-3] = src - ref
+    "vsubl.u8    q0, d8, d12                  \n"
+    "vsubl.u8    q1, d9, d13                  \n"
+
+    // load coeff16 into q8(d16=5352, d17=2217)
+    "vld1.16     {q8}, [%[coeff16]]           \n"
+
+    // load coeff32 high half into q9 = 1812, q10 = 937
+    "vld1.32     {q9, q10}, [%[coeff32]]!     \n"
+
+    // load coeff32 low half into q11=12000, q12=51000
+    "vld1.32     {q11,q12}, [%[coeff32]]      \n"
+
+    // part 1
+    // Transpose. Register dN is the same as dN in C
+    "vtrn.32         d0, d2                   \n"
+    "vtrn.32         d1, d3                   \n"
+    "vtrn.16         d0, d1                   \n"
+    "vtrn.16         d2, d3                   \n"
+
+    "vadd.s16        d4, d0, d3               \n" // a0 = d0 + d3
+    "vadd.s16        d5, d1, d2               \n" // a1 = d1 + d2
+    "vsub.s16        d6, d1, d2               \n" // a2 = d1 - d2
+    "vsub.s16        d7, d0, d3               \n" // a3 = d0 - d3
+
+    "vadd.s16        d0, d4, d5               \n" // a0 + a1
+    "vshl.s16        d0, d0, #3               \n" // temp[0+i*4] = (a0+a1) << 3
+    "vsub.s16        d2, d4, d5               \n" // a0 - a1
+    "vshl.s16        d2, d2, #3               \n" // (temp[2+i*4] = (a0-a1) << 3
+
+    "vmlal.s16       q9, d7, d16              \n" // a3*5352 + 1812
+    "vmlal.s16       q10, d7, d17             \n" // a3*2217 + 937
+    "vmlal.s16       q9, d6, d17              \n" // a2*2217 + a3*5352 + 1812
+    "vmlsl.s16       q10, d6, d16             \n" // a3*2217 + 937 - a2*5352
+
+    // temp[1+i*4] = (d2*2217 + d3*5352 + 1812) >> 9
+    // temp[3+i*4] = (d3*2217 + 937 - d2*5352) >> 9
+    "vshrn.s32       d1, q9, #9               \n"
+    "vshrn.s32       d3, q10, #9              \n"
+
+    // part 2
+    // transpose d0=ip[0], d1=ip[4], d2=ip[8], d3=ip[12]
+    "vtrn.32         d0, d2                   \n"
+    "vtrn.32         d1, d3                   \n"
+    "vtrn.16         d0, d1                   \n"
+    "vtrn.16         d2, d3                   \n"
+
+    "vmov.s16        d26, #7                  \n"
+
+    "vadd.s16        d4, d0, d3               \n" // a1 = ip[0] + ip[12]
+    "vadd.s16        d5, d1, d2               \n" // b1 = ip[4] + ip[8]
+    "vsub.s16        d6, d1, d2               \n" // c1 = ip[4] - ip[8]
+    "vadd.s16        d4, d4, d26              \n" // a1 + 7
+    "vsub.s16        d7, d0, d3               \n" // d1 = ip[0] - ip[12]
+
+    "vadd.s16        d0, d4, d5               \n" // op[0] = a1 + b1 + 7
+    "vsub.s16        d2, d4, d5               \n" // op[8] = a1 - b1 + 7
+
+    "vmlal.s16       q11, d7, d16             \n" // d1*5352 + 12000
+    "vmlal.s16       q12, d7, d17             \n" // d1*2217 + 51000
+
+    "vceq.s16        d4, d7, #0               \n"
+
+    "vshr.s16        d0, d0, #4               \n"
+    "vshr.s16        d2, d2, #4               \n"
+
+    "vmlal.s16       q11, d6, d17             \n" // c1*2217 + d1*5352 + 12000
+    "vmlsl.s16       q12, d6, d16             \n" // d1*2217 - c1*5352 + 51000
+
+    "vmvn            d4, d4                   \n" // !(d1 == 0)
+    // op[4] = (c1*2217 + d1*5352 + 12000)>>16
+    "vshrn.s32       d1, q11, #16             \n"
+    // op[4] += (d1!=0)
+    "vsub.s16        d1, d1, d4               \n"
+    // op[12]= (d1*2217 - c1*5352 + 51000)>>16
+    "vshrn.s32       d3, q12, #16             \n"
+
+    // set result to out array
+    "vst1.16         {q0, q1}, [%[out]]   \n"
+    : [src_ptr] "+r"(src_ptr), [ref_ptr] "+r"(ref_ptr),
+      [coeff32] "+r"(coeff32)          // modified registers
+    : [kBPS] "r"(kBPS), [coeff16] "r"(coeff16),
+      [out] "r"(out)                   // constants
+    : "memory", "q0", "q1", "q2", "q3", "q4", "q5", "q6", "q7", "q8", "q9",
+      "q10", "q11", "q12", "q13"       // clobbered
+  );
+}
+
+static void FTransformWHT(const int16_t* in, int16_t* out) {
+  const int kStep = 32;
+  __asm__ volatile (
+    // d0 = in[0 * 16] , d1 = in[1 * 16]
+    // d2 = in[2 * 16] , d3 = in[3 * 16]
+    "vld1.16         d0[0], [%[in]], %[kStep]   \n"
+    "vld1.16         d1[0], [%[in]], %[kStep]   \n"
+    "vld1.16         d2[0], [%[in]], %[kStep]   \n"
+    "vld1.16         d3[0], [%[in]], %[kStep]   \n"
+    "vld1.16         d0[1], [%[in]], %[kStep]   \n"
+    "vld1.16         d1[1], [%[in]], %[kStep]   \n"
+    "vld1.16         d2[1], [%[in]], %[kStep]   \n"
+    "vld1.16         d3[1], [%[in]], %[kStep]   \n"
+    "vld1.16         d0[2], [%[in]], %[kStep]   \n"
+    "vld1.16         d1[2], [%[in]], %[kStep]   \n"
+    "vld1.16         d2[2], [%[in]], %[kStep]   \n"
+    "vld1.16         d3[2], [%[in]], %[kStep]   \n"
+    "vld1.16         d0[3], [%[in]], %[kStep]   \n"
+    "vld1.16         d1[3], [%[in]], %[kStep]   \n"
+    "vld1.16         d2[3], [%[in]], %[kStep]   \n"
+    "vld1.16         d3[3], [%[in]], %[kStep]   \n"
+
+    "vaddl.s16       q2, d0, d2                 \n" // a0=(in[0*16]+in[2*16])
+    "vaddl.s16       q3, d1, d3                 \n" // a1=(in[1*16]+in[3*16])
+    "vsubl.s16       q4, d1, d3                 \n" // a2=(in[1*16]-in[3*16])
+    "vsubl.s16       q5, d0, d2                 \n" // a3=(in[0*16]-in[2*16])
+
+    "vqadd.s32       q6, q2, q3                 \n" // a0 + a1
+    "vqadd.s32       q7, q5, q4                 \n" // a3 + a2
+    "vqsub.s32       q8, q5, q4                 \n" // a3 - a2
+    "vqsub.s32       q9, q2, q3                 \n" // a0 - a1
+
+    // Transpose
+    // q6 = tmp[0, 1,  2,  3] ; q7 = tmp[ 4,  5,  6,  7]
+    // q8 = tmp[8, 9, 10, 11] ; q9 = tmp[12, 13, 14, 15]
+    "vswp            d13, d16                   \n" // vtrn.64 q0, q2
+    "vswp            d15, d18                   \n" // vtrn.64 q1, q3
+    "vtrn.32         q6, q7                     \n"
+    "vtrn.32         q8, q9                     \n"
+
+    "vqadd.s32       q0, q6, q8                 \n" // a0 = tmp[0] + tmp[8]
+    "vqadd.s32       q1, q7, q9                 \n" // a1 = tmp[4] + tmp[12]
+    "vqsub.s32       q2, q7, q9                 \n" // a2 = tmp[4] - tmp[12]
+    "vqsub.s32       q3, q6, q8                 \n" // a3 = tmp[0] - tmp[8]
+
+    "vqadd.s32       q4, q0, q1                 \n" // b0 = a0 + a1
+    "vqadd.s32       q5, q3, q2                 \n" // b1 = a3 + a2
+    "vqsub.s32       q6, q3, q2                 \n" // b2 = a3 - a2
+    "vqsub.s32       q7, q0, q1                 \n" // b3 = a0 - a1
+
+    "vshrn.s32       d18, q4, #1                \n" // b0 >> 1
+    "vshrn.s32       d19, q5, #1                \n" // b1 >> 1
+    "vshrn.s32       d20, q6, #1                \n" // b2 >> 1
+    "vshrn.s32       d21, q7, #1                \n" // b3 >> 1
+
+    "vst1.16         {q9, q10}, [%[out]]        \n"
+
+    : [in] "+r"(in)
+    : [kStep] "r"(kStep), [out] "r"(out)
+    : "memory", "q0", "q1", "q2", "q3", "q4", "q5",
+      "q6", "q7", "q8", "q9", "q10"       // clobbered
+  ) ;
+}
+
+//------------------------------------------------------------------------------
+// Texture distortion
+//
+// We try to match the spectral content (weighted) between source and
+// reconstructed samples.
+
+// Hadamard transform
+// Returns the weighted sum of the absolute value of transformed coefficients.
+// This uses a TTransform helper function in C
+static int Disto4x4(const uint8_t* const a, const uint8_t* const b,
+                    const uint16_t* const w) {
+  const int kBPS = BPS;
+  const uint8_t* A = a;
+  const uint8_t* B = b;
+  const uint16_t* W = w;
+  int sum;
+  __asm__ volatile (
+    "vld1.32         d0[0], [%[a]], %[kBPS]   \n"
+    "vld1.32         d0[1], [%[a]], %[kBPS]   \n"
+    "vld1.32         d2[0], [%[a]], %[kBPS]   \n"
+    "vld1.32         d2[1], [%[a]]            \n"
+
+    "vld1.32         d1[0], [%[b]], %[kBPS]   \n"
+    "vld1.32         d1[1], [%[b]], %[kBPS]   \n"
+    "vld1.32         d3[0], [%[b]], %[kBPS]   \n"
+    "vld1.32         d3[1], [%[b]]            \n"
+
+    // a d0/d2, b d1/d3
+    // d0/d1: 01 01 01 01
+    // d2/d3: 23 23 23 23
+    // But: it goes 01 45 23 67
+    // Notice the middle values are transposed
+    "vtrn.16         q0, q1                   \n"
+
+    // {a0, a1} = {in[0] + in[2], in[1] + in[3]}
+    "vaddl.u8        q2, d0, d2               \n"
+    "vaddl.u8        q10, d1, d3              \n"
+    // {a3, a2} = {in[0] - in[2], in[1] - in[3]}
+    "vsubl.u8        q3, d0, d2               \n"
+    "vsubl.u8        q11, d1, d3              \n"
+
+    // tmp[0] = a0 + a1
+    "vpaddl.s16      q0, q2                   \n"
+    "vpaddl.s16      q8, q10                  \n"
+
+    // tmp[1] = a3 + a2
+    "vpaddl.s16      q1, q3                   \n"
+    "vpaddl.s16      q9, q11                  \n"
+
+    // No pair subtract
+    // q2 = {a0, a3}
+    // q3 = {a1, a2}
+    "vtrn.16         q2, q3                   \n"
+    "vtrn.16         q10, q11                 \n"
+
+    // {tmp[3], tmp[2]} = {a0 - a1, a3 - a2}
+    "vsubl.s16       q12, d4, d6              \n"
+    "vsubl.s16       q13, d5, d7              \n"
+    "vsubl.s16       q14, d20, d22            \n"
+    "vsubl.s16       q15, d21, d23            \n"
+
+    // separate tmp[3] and tmp[2]
+    // q12 = tmp[3]
+    // q13 = tmp[2]
+    "vtrn.32         q12, q13                 \n"
+    "vtrn.32         q14, q15                 \n"
+
+    // Transpose tmp for a
+    "vswp            d1, d26                  \n" // vtrn.64
+    "vswp            d3, d24                  \n" // vtrn.64
+    "vtrn.32         q0, q1                   \n"
+    "vtrn.32         q13, q12                 \n"
+
+    // Transpose tmp for b
+    "vswp            d17, d30                 \n" // vtrn.64
+    "vswp            d19, d28                 \n" // vtrn.64
+    "vtrn.32         q8, q9                   \n"
+    "vtrn.32         q15, q14                 \n"
+
+    // The first Q register is a, the second b.
+    // q0/8 tmp[0-3]
+    // q13/15 tmp[4-7]
+    // q1/9 tmp[8-11]
+    // q12/14 tmp[12-15]
+
+    // These are still in 01 45 23 67 order. We fix it easily in the addition
+    // case but the subtraction propegates them.
+    "vswp            d3, d27                  \n"
+    "vswp            d19, d31                 \n"
+
+    // a0 = tmp[0] + tmp[8]
+    "vadd.s32        q2, q0, q1               \n"
+    "vadd.s32        q3, q8, q9               \n"
+
+    // a1 = tmp[4] + tmp[12]
+    "vadd.s32        q10, q13, q12            \n"
+    "vadd.s32        q11, q15, q14            \n"
+
+    // a2 = tmp[4] - tmp[12]
+    "vsub.s32        q13, q13, q12            \n"
+    "vsub.s32        q15, q15, q14            \n"
+
+    // a3 = tmp[0] - tmp[8]
+    "vsub.s32        q0, q0, q1               \n"
+    "vsub.s32        q8, q8, q9               \n"
+
+    // b0 = a0 + a1
+    "vadd.s32        q1, q2, q10              \n"
+    "vadd.s32        q9, q3, q11              \n"
+
+    // b1 = a3 + a2
+    "vadd.s32        q12, q0, q13             \n"
+    "vadd.s32        q14, q8, q15             \n"
+
+    // b2 = a3 - a2
+    "vsub.s32        q0, q0, q13              \n"
+    "vsub.s32        q8, q8, q15              \n"
+
+    // b3 = a0 - a1
+    "vsub.s32        q2, q2, q10              \n"
+    "vsub.s32        q3, q3, q11              \n"
+
+    "vld1.64         {q10, q11}, [%[w]]       \n"
+
+    // abs(b0)
+    "vabs.s32        q1, q1                   \n"
+    "vabs.s32        q9, q9                   \n"
+    // abs(b1)
+    "vabs.s32        q12, q12                 \n"
+    "vabs.s32        q14, q14                 \n"
+    // abs(b2)
+    "vabs.s32        q0, q0                   \n"
+    "vabs.s32        q8, q8                   \n"
+    // abs(b3)
+    "vabs.s32        q2, q2                   \n"
+    "vabs.s32        q3, q3                   \n"
+
+    // expand w before using.
+    "vmovl.u16       q13, d20                 \n"
+    "vmovl.u16       q15, d21                 \n"
+
+    // w[0] * abs(b0)
+    "vmul.u32        q1, q1, q13              \n"
+    "vmul.u32        q9, q9, q13              \n"
+
+    // w[4] * abs(b1)
+    "vmla.u32        q1, q12, q15             \n"
+    "vmla.u32        q9, q14, q15             \n"
+
+    // expand w before using.
+    "vmovl.u16       q13, d22                 \n"
+    "vmovl.u16       q15, d23                 \n"
+
+    // w[8] * abs(b1)
+    "vmla.u32        q1, q0, q13              \n"
+    "vmla.u32        q9, q8, q13              \n"
+
+    // w[12] * abs(b1)
+    "vmla.u32        q1, q2, q15              \n"
+    "vmla.u32        q9, q3, q15              \n"
+
+    // Sum the arrays
+    "vpaddl.u32      q1, q1                   \n"
+    "vpaddl.u32      q9, q9                   \n"
+    "vadd.u64        d2, d3                   \n"
+    "vadd.u64        d18, d19                 \n"
+
+    // Hadamard transform needs 4 bits of extra precision (2 bits in each
+    // direction) for dynamic raw. Weights w[] are 16bits at max, so the maximum
+    // precision for coeff is 8bit of input + 4bits of Hadamard transform +
+    // 16bits for w[] + 2 bits of abs() summation.
+    //
+    // This uses a maximum of 31 bits (signed). Discarding the top 32 bits is
+    // A-OK.
+
+    // sum2 - sum1
+    "vsub.u32        d0, d2, d18              \n"
+    // abs(sum2 - sum1)
+    "vabs.s32        d0, d0                   \n"
+    // abs(sum2 - sum1) >> 5
+    "vshr.u32        d0, #5                   \n"
+
+    // It would be better to move the value straight into r0 but I'm not
+    // entirely sure how this works with inline assembly.
+    "vmov.32         %[sum], d0[0]            \n"
+
+    : [sum] "=r"(sum), [a] "+r"(A), [b] "+r"(B), [w] "+r"(W)
+    : [kBPS] "r"(kBPS)
+    : "memory", "q0", "q1", "q2", "q3", "q4", "q5", "q6", "q7", "q8", "q9",
+      "q10", "q11", "q12", "q13", "q14", "q15"  // clobbered
+  ) ;
+
+  return sum;
+}
+
+static int Disto16x16(const uint8_t* const a, const uint8_t* const b,
+                      const uint16_t* const w) {
+  int D = 0;
+  int x, y;
+  for (y = 0; y < 16 * BPS; y += 4 * BPS) {
+    for (x = 0; x < 16; x += 4) {
+      D += Disto4x4(a + x + y, b + x + y, w);
+    }
+  }
+  return D;
+}
+
+#endif   // WEBP_USE_NEON
+
+//------------------------------------------------------------------------------
+// Entry point
+
+extern void VP8EncDspInitNEON(void);
+
+void VP8EncDspInitNEON(void) {
+#if defined(WEBP_USE_NEON)
+  VP8ITransform = ITransform;
+  VP8FTransform = FTransform;
+
+  VP8ITransformWHT = ITransformWHT;
+  VP8FTransformWHT = FTransformWHT;
+
+  VP8TDisto4x4 = Disto4x4;
+  VP8TDisto16x16 = Disto16x16;
+#endif   // WEBP_USE_NEON
+}
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
diff --git a/3rdparty/libwebp/dsp/enc_sse2.c b/3rdparty/libwebp/dsp/enc_sse2.c
index 0986e82e31..032e990762 100644
--- a/3rdparty/libwebp/dsp/enc_sse2.c
+++ b/3rdparty/libwebp/dsp/enc_sse2.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // SSE2 version of speed-critical encoding functions.
@@ -21,17 +23,48 @@ extern "C" {
 
 #include "../enc/vp8enci.h"
 
+//------------------------------------------------------------------------------
+// Quite useful macro for debugging. Left here for convenience.
+
+#if 0
+#include <stdio.h>
+static void PrintReg(const __m128i r, const char* const name, int size) {
+  int n;
+  union {
+    __m128i r;
+    uint8_t i8[16];
+    uint16_t i16[8];
+    uint32_t i32[4];
+    uint64_t i64[2];
+  } tmp;
+  tmp.r = r;
+  printf("%s\t: ", name);
+  if (size == 8) {
+    for (n = 0; n < 16; ++n) printf("%.2x ", tmp.i8[n]);
+  } else if (size == 16) {
+    for (n = 0; n < 8; ++n) printf("%.4x ", tmp.i16[n]);
+  } else if (size == 32) {
+    for (n = 0; n < 4; ++n) printf("%.8x ", tmp.i32[n]);
+  } else {
+    for (n = 0; n < 2; ++n) printf("%.16lx ", tmp.i64[n]);
+  }
+  printf("\n");
+}
+#endif
+
 //------------------------------------------------------------------------------
 // Compute susceptibility based on DCT-coeff histograms:
 // the higher, the "easier" the macroblock is to compress.
 
-static int CollectHistogramSSE2(const uint8_t* ref, const uint8_t* pred,
-                                int start_block, int end_block) {
-  int histo[MAX_COEFF_THRESH + 1] = { 0 };
-  int16_t out[16];
-  int j, k;
+static void CollectHistogramSSE2(const uint8_t* ref, const uint8_t* pred,
+                                 int start_block, int end_block,
+                                 VP8Histogram* const histo) {
   const __m128i max_coeff_thresh = _mm_set1_epi16(MAX_COEFF_THRESH);
+  int j;
   for (j = start_block; j < end_block; ++j) {
+    int16_t out[16];
+    int k;
+
     VP8FTransform(ref + VP8DspScan[j], pred + VP8DspScan[j], out);
 
     // Convert coefficients to bin (within out[]).
@@ -47,9 +80,9 @@ static int CollectHistogramSSE2(const uint8_t* ref, const uint8_t* pred,
       const __m128i xor1 = _mm_xor_si128(out1, sign1);
       const __m128i abs0 = _mm_sub_epi16(xor0, sign0);
       const __m128i abs1 = _mm_sub_epi16(xor1, sign1);
-      // v = abs(out) >> 2
-      const __m128i v0 = _mm_srai_epi16(abs0, 2);
-      const __m128i v1 = _mm_srai_epi16(abs1, 2);
+      // v = abs(out) >> 3
+      const __m128i v0 = _mm_srai_epi16(abs0, 3);
+      const __m128i v1 = _mm_srai_epi16(abs1, 3);
       // bin = min(v, MAX_COEFF_THRESH)
       const __m128i bin0 = _mm_min_epi16(v0, max_coeff_thresh);
       const __m128i bin1 = _mm_min_epi16(v1, max_coeff_thresh);
@@ -58,13 +91,11 @@ static int CollectHistogramSSE2(const uint8_t* ref, const uint8_t* pred,
       _mm_storeu_si128((__m128i*)&out[8], bin1);
     }
 
-    // Use bin to update histogram.
+    // Convert coefficients to bin.
     for (k = 0; k < 16; ++k) {
-      histo[out[k]]++;
+      histo->distribution[out[k]]++;
     }
   }
-
-  return VP8GetAlpha(histo);
 }
 
 //------------------------------------------------------------------------------
@@ -243,7 +274,7 @@ static void ITransformSSE2(const uint8_t* ref, const int16_t* in, uint8_t* dst,
 
   // Add inverse transform to 'ref' and store.
   {
-    const __m128i zero = _mm_set1_epi16(0);
+    const __m128i zero = _mm_setzero_si128();
     // Load the reference(s).
     __m128i ref0, ref1, ref2, ref3;
     if (do_two) {
@@ -295,17 +326,23 @@ static void FTransformSSE2(const uint8_t* src, const uint8_t* ref,
                            int16_t* out) {
   const __m128i zero = _mm_setzero_si128();
   const __m128i seven = _mm_set1_epi16(7);
-  const __m128i k7500 = _mm_set1_epi32(7500);
-  const __m128i k14500 = _mm_set1_epi32(14500);
+  const __m128i k937 = _mm_set1_epi32(937);
+  const __m128i k1812 = _mm_set1_epi32(1812);
   const __m128i k51000 = _mm_set1_epi32(51000);
   const __m128i k12000_plus_one = _mm_set1_epi32(12000 + (1 << 16));
   const __m128i k5352_2217 = _mm_set_epi16(5352,  2217, 5352,  2217,
                                            5352,  2217, 5352,  2217);
   const __m128i k2217_5352 = _mm_set_epi16(2217, -5352, 2217, -5352,
                                            2217, -5352, 2217, -5352);
-
+  const __m128i k88p = _mm_set_epi16(8, 8, 8, 8, 8, 8, 8, 8);
+  const __m128i k88m = _mm_set_epi16(-8, 8, -8, 8, -8, 8, -8, 8);
+  const __m128i k5352_2217p = _mm_set_epi16(2217, 5352, 2217, 5352,
+                                            2217, 5352, 2217, 5352);
+  const __m128i k5352_2217m = _mm_set_epi16(-5352, 2217, -5352, 2217,
+                                            -5352, 2217, -5352, 2217);
   __m128i v01, v32;
 
+
   // Difference between src and ref and initial transpose.
   {
     // Load src and convert to 16b.
@@ -326,73 +363,52 @@ static void FTransformSSE2(const uint8_t* src, const uint8_t* ref,
     const __m128i ref_1 = _mm_unpacklo_epi8(ref1, zero);
     const __m128i ref_2 = _mm_unpacklo_epi8(ref2, zero);
     const __m128i ref_3 = _mm_unpacklo_epi8(ref3, zero);
-    // Compute difference.
+    // Compute difference. -> 00 01 02 03 00 00 00 00
     const __m128i diff0 = _mm_sub_epi16(src_0, ref_0);
     const __m128i diff1 = _mm_sub_epi16(src_1, ref_1);
     const __m128i diff2 = _mm_sub_epi16(src_2, ref_2);
     const __m128i diff3 = _mm_sub_epi16(src_3, ref_3);
 
-    // Transpose.
+
+    // Unpack and shuffle
     // 00 01 02 03   0 0 0 0
     // 10 11 12 13   0 0 0 0
     // 20 21 22 23   0 0 0 0
     // 30 31 32 33   0 0 0 0
-    const __m128i transpose0_0 = _mm_unpacklo_epi16(diff0, diff1);
-    const __m128i transpose0_1 = _mm_unpacklo_epi16(diff2, diff3);
-    // 00 10 01 11   02 12 03 13
-    // 20 30 21 31   22 32 23 33
-    const __m128i v23 = _mm_unpackhi_epi32(transpose0_0, transpose0_1);
-    v01 = _mm_unpacklo_epi32(transpose0_0, transpose0_1);
-    v32 = _mm_shuffle_epi32(v23, _MM_SHUFFLE(1, 0, 3, 2));
-    // a02 a12 a22 a32   a03 a13 a23 a33
-    // a00 a10 a20 a30   a01 a11 a21 a31
-    // a03 a13 a23 a33   a02 a12 a22 a32
-  }
+    const __m128i shuf01 = _mm_unpacklo_epi32(diff0, diff1);
+    const __m128i shuf23 = _mm_unpacklo_epi32(diff2, diff3);
+    // 00 01 10 11 02 03 12 13
+    // 20 21 30 31 22 23 32 33
+    const __m128i shuf01_p =
+        _mm_shufflehi_epi16(shuf01, _MM_SHUFFLE(2, 3, 0, 1));
+    const __m128i shuf23_p =
+        _mm_shufflehi_epi16(shuf23, _MM_SHUFFLE(2, 3, 0, 1));
+    // 00 01 10 11 03 02 13 12
+    // 20 21 30 31 23 22 33 32
+    const __m128i s01 = _mm_unpacklo_epi64(shuf01_p, shuf23_p);
+    const __m128i s32 = _mm_unpackhi_epi64(shuf01_p, shuf23_p);
+    // 00 01 10 11 20 21 30 31
+    // 03 02 13 12 23 22 33 32
+    const __m128i a01 = _mm_add_epi16(s01, s32);
+    const __m128i a32 = _mm_sub_epi16(s01, s32);
+    // [d0 + d3 | d1 + d2 | ...] = [a0 a1 | a0' a1' | ... ]
+    // [d0 - d3 | d1 - d2 | ...] = [a3 a2 | a3' a2' | ... ]
 
-  // First pass and subsequent transpose.
-  {
-    // Same operations are done on the (0,3) and (1,2) pairs.
-    // b0 = (a0 + a3) << 3
-    // b1 = (a1 + a2) << 3
-    // b3 = (a0 - a3) << 3
-    // b2 = (a1 - a2) << 3
-    const __m128i a01 = _mm_add_epi16(v01, v32);
-    const __m128i a32 = _mm_sub_epi16(v01, v32);
-    const __m128i b01 = _mm_slli_epi16(a01, 3);
-    const __m128i b32 = _mm_slli_epi16(a32, 3);
-    const __m128i b11 = _mm_unpackhi_epi64(b01, b01);
-    const __m128i b22 = _mm_unpackhi_epi64(b32, b32);
-
-    // e0 = b0 + b1
-    // e2 = b0 - b1
-    const __m128i e0 = _mm_add_epi16(b01, b11);
-    const __m128i e2 = _mm_sub_epi16(b01, b11);
-    const __m128i e02 = _mm_unpacklo_epi64(e0, e2);
-
-    // e1 = (b3 * 5352 + b2 * 2217 + 14500) >> 12
-    // e3 = (b3 * 2217 - b2 * 5352 +  7500) >> 12
-    const __m128i b23 = _mm_unpacklo_epi16(b22, b32);
-    const __m128i c1 = _mm_madd_epi16(b23, k5352_2217);
-    const __m128i c3 = _mm_madd_epi16(b23, k2217_5352);
-    const __m128i d1 = _mm_add_epi32(c1, k14500);
-    const __m128i d3 = _mm_add_epi32(c3, k7500);
-    const __m128i e1 = _mm_srai_epi32(d1, 12);
-    const __m128i e3 = _mm_srai_epi32(d3, 12);
-    const __m128i e13 = _mm_packs_epi32(e1, e3);
-
-    // Transpose.
-    // 00 01 02 03  20 21 22 23
-    // 10 11 12 13  30 31 32 33
-    const __m128i transpose0_0 = _mm_unpacklo_epi16(e02, e13);
-    const __m128i transpose0_1 = _mm_unpackhi_epi16(e02, e13);
-    // 00 10 01 11   02 12 03 13
-    // 20 30 21 31   22 32 23 33
-    const __m128i v23 = _mm_unpackhi_epi32(transpose0_0, transpose0_1);
-    v01 = _mm_unpacklo_epi32(transpose0_0, transpose0_1);
-    v32 = _mm_shuffle_epi32(v23, _MM_SHUFFLE(1, 0, 3, 2));
-    // 02 12 22 32   03 13 23 33
-    // 00 10 20 30   01 11 21 31
-    // 03 13 23 33   02 12 22 32
+    const __m128i tmp0 = _mm_madd_epi16(a01, k88p);  // [ (a0 + a1) << 3, ... ]
+    const __m128i tmp2 = _mm_madd_epi16(a01, k88m);  // [ (a0 - a1) << 3, ... ]
+    const __m128i tmp1_1 = _mm_madd_epi16(a32, k5352_2217p);
+    const __m128i tmp3_1 = _mm_madd_epi16(a32, k5352_2217m);
+    const __m128i tmp1_2 = _mm_add_epi32(tmp1_1, k1812);
+    const __m128i tmp3_2 = _mm_add_epi32(tmp3_1, k937);
+    const __m128i tmp1   = _mm_srai_epi32(tmp1_2, 9);
+    const __m128i tmp3   = _mm_srai_epi32(tmp3_2, 9);
+    const __m128i s03 = _mm_packs_epi32(tmp0, tmp2);
+    const __m128i s12 = _mm_packs_epi32(tmp1, tmp3);
+    const __m128i s_lo = _mm_unpacklo_epi16(s03, s12);   // 0 1 0 1 0 1...
+    const __m128i s_hi = _mm_unpackhi_epi16(s03, s12);   // 2 3 2 3 2 3
+    const __m128i v23 = _mm_unpackhi_epi32(s_lo, s_hi);
+    v01 = _mm_unpacklo_epi32(s_lo, s_hi);
+    v32 = _mm_shuffle_epi32(v23, _MM_SHUFFLE(1, 0, 3, 2));  // 3 2 3 2 3 2..
   }
 
   // Second pass
@@ -406,13 +422,12 @@ static void FTransformSSE2(const uint8_t* src, const uint8_t* ref,
     const __m128i a32 = _mm_sub_epi16(v01, v32);
     const __m128i a11 = _mm_unpackhi_epi64(a01, a01);
     const __m128i a22 = _mm_unpackhi_epi64(a32, a32);
+    const __m128i a01_plus_7 = _mm_add_epi16(a01, seven);
 
     // d0 = (a0 + a1 + 7) >> 4;
     // d2 = (a0 - a1 + 7) >> 4;
-    const __m128i b0 = _mm_add_epi16(a01, a11);
-    const __m128i b2 = _mm_sub_epi16(a01, a11);
-    const __m128i c0 = _mm_add_epi16(b0, seven);
-    const __m128i c2 = _mm_add_epi16(b2, seven);
+    const __m128i c0 = _mm_add_epi16(a01_plus_7, a11);
+    const __m128i c2 = _mm_sub_epi16(a01_plus_7, a11);
     const __m128i d0 = _mm_srai_epi16(c0, 4);
     const __m128i d2 = _mm_srai_epi16(c2, 4);
 
@@ -430,6 +445,7 @@ static void FTransformSSE2(const uint8_t* src, const uint8_t* ref,
     // f1 = f1 + (a3 != 0);
     // The compare will return (0xffff, 0) for (==0, !=0). To turn that into the
     // desired (0, 1), we add one earlier through k12000_plus_one.
+    // -> f1 = f1 + 1 - (a3 == 0)
     const __m128i g1 = _mm_add_epi16(f1, _mm_cmpeq_epi16(a32, zero));
 
     _mm_storel_epi64((__m128i*)&out[ 0], d0);
@@ -439,13 +455,137 @@ static void FTransformSSE2(const uint8_t* src, const uint8_t* ref,
   }
 }
 
+static void FTransformWHTSSE2(const int16_t* in, int16_t* out) {
+  int16_t tmp[16];
+  int i;
+  for (i = 0; i < 4; ++i, in += 64) {
+    const int a0 = (in[0 * 16] + in[2 * 16]);
+    const int a1 = (in[1 * 16] + in[3 * 16]);
+    const int a2 = (in[1 * 16] - in[3 * 16]);
+    const int a3 = (in[0 * 16] - in[2 * 16]);
+    tmp[0 + i * 4] = a0 + a1;
+    tmp[1 + i * 4] = a3 + a2;
+    tmp[2 + i * 4] = a3 - a2;
+    tmp[3 + i * 4] = a0 - a1;
+  }
+  {
+    const __m128i src0 = _mm_loadl_epi64((__m128i*)&tmp[0]);
+    const __m128i src1 = _mm_loadl_epi64((__m128i*)&tmp[4]);
+    const __m128i src2 = _mm_loadl_epi64((__m128i*)&tmp[8]);
+    const __m128i src3 = _mm_loadl_epi64((__m128i*)&tmp[12]);
+    const __m128i a0 = _mm_add_epi16(src0, src2);
+    const __m128i a1 = _mm_add_epi16(src1, src3);
+    const __m128i a2 = _mm_sub_epi16(src1, src3);
+    const __m128i a3 = _mm_sub_epi16(src0, src2);
+    const __m128i b0 = _mm_srai_epi16(_mm_adds_epi16(a0, a1), 1);
+    const __m128i b1 = _mm_srai_epi16(_mm_adds_epi16(a3, a2), 1);
+    const __m128i b2 = _mm_srai_epi16(_mm_subs_epi16(a3, a2), 1);
+    const __m128i b3 = _mm_srai_epi16(_mm_subs_epi16(a0, a1), 1);
+    _mm_storel_epi64((__m128i*)&out[ 0], b0);
+    _mm_storel_epi64((__m128i*)&out[ 4], b1);
+    _mm_storel_epi64((__m128i*)&out[ 8], b2);
+    _mm_storel_epi64((__m128i*)&out[12], b3);
+  }
+}
+
 //------------------------------------------------------------------------------
 // Metric
 
-static int SSE4x4SSE2(const uint8_t* a, const uint8_t* b) {
-  const __m128i zero = _mm_set1_epi16(0);
+static int SSE_Nx4SSE2(const uint8_t* a, const uint8_t* b,
+                       int num_quads, int do_16) {
+  const __m128i zero = _mm_setzero_si128();
+  __m128i sum1 = zero;
+  __m128i sum2 = zero;
 
-  // Load values.
+  while (num_quads-- > 0) {
+    // Note: for the !do_16 case, we read 16 pixels instead of 8 but that's ok,
+    // thanks to buffer over-allocation to that effect.
+    const __m128i a0 = _mm_loadu_si128((__m128i*)&a[BPS * 0]);
+    const __m128i a1 = _mm_loadu_si128((__m128i*)&a[BPS * 1]);
+    const __m128i a2 = _mm_loadu_si128((__m128i*)&a[BPS * 2]);
+    const __m128i a3 = _mm_loadu_si128((__m128i*)&a[BPS * 3]);
+    const __m128i b0 = _mm_loadu_si128((__m128i*)&b[BPS * 0]);
+    const __m128i b1 = _mm_loadu_si128((__m128i*)&b[BPS * 1]);
+    const __m128i b2 = _mm_loadu_si128((__m128i*)&b[BPS * 2]);
+    const __m128i b3 = _mm_loadu_si128((__m128i*)&b[BPS * 3]);
+
+    // compute clip0(a-b) and clip0(b-a)
+    const __m128i a0p = _mm_subs_epu8(a0, b0);
+    const __m128i a0m = _mm_subs_epu8(b0, a0);
+    const __m128i a1p = _mm_subs_epu8(a1, b1);
+    const __m128i a1m = _mm_subs_epu8(b1, a1);
+    const __m128i a2p = _mm_subs_epu8(a2, b2);
+    const __m128i a2m = _mm_subs_epu8(b2, a2);
+    const __m128i a3p = _mm_subs_epu8(a3, b3);
+    const __m128i a3m = _mm_subs_epu8(b3, a3);
+
+    // compute |a-b| with 8b arithmetic as clip0(a-b) | clip0(b-a)
+    const __m128i diff0 = _mm_or_si128(a0p, a0m);
+    const __m128i diff1 = _mm_or_si128(a1p, a1m);
+    const __m128i diff2 = _mm_or_si128(a2p, a2m);
+    const __m128i diff3 = _mm_or_si128(a3p, a3m);
+
+    // unpack (only four operations, instead of eight)
+    const __m128i low0 = _mm_unpacklo_epi8(diff0, zero);
+    const __m128i low1 = _mm_unpacklo_epi8(diff1, zero);
+    const __m128i low2 = _mm_unpacklo_epi8(diff2, zero);
+    const __m128i low3 = _mm_unpacklo_epi8(diff3, zero);
+
+    // multiply with self
+    const __m128i low_madd0 = _mm_madd_epi16(low0, low0);
+    const __m128i low_madd1 = _mm_madd_epi16(low1, low1);
+    const __m128i low_madd2 = _mm_madd_epi16(low2, low2);
+    const __m128i low_madd3 = _mm_madd_epi16(low3, low3);
+
+    // collect in a cascading way
+    const __m128i low_sum0 = _mm_add_epi32(low_madd0, low_madd1);
+    const __m128i low_sum1 = _mm_add_epi32(low_madd2, low_madd3);
+    sum1 = _mm_add_epi32(sum1, low_sum0);
+    sum2 = _mm_add_epi32(sum2, low_sum1);
+
+    if (do_16) {  // if necessary, process the higher 8 bytes similarly
+      const __m128i hi0 = _mm_unpackhi_epi8(diff0, zero);
+      const __m128i hi1 = _mm_unpackhi_epi8(diff1, zero);
+      const __m128i hi2 = _mm_unpackhi_epi8(diff2, zero);
+      const __m128i hi3 = _mm_unpackhi_epi8(diff3, zero);
+
+      const __m128i hi_madd0 = _mm_madd_epi16(hi0, hi0);
+      const __m128i hi_madd1 = _mm_madd_epi16(hi1, hi1);
+      const __m128i hi_madd2 = _mm_madd_epi16(hi2, hi2);
+      const __m128i hi_madd3 = _mm_madd_epi16(hi3, hi3);
+      const __m128i hi_sum0 = _mm_add_epi32(hi_madd0, hi_madd1);
+      const __m128i hi_sum1 = _mm_add_epi32(hi_madd2, hi_madd3);
+      sum1 = _mm_add_epi32(sum1, hi_sum0);
+      sum2 = _mm_add_epi32(sum2, hi_sum1);
+    }
+    a += 4 * BPS;
+    b += 4 * BPS;
+  }
+  {
+    int32_t tmp[4];
+    const __m128i sum = _mm_add_epi32(sum1, sum2);
+    _mm_storeu_si128((__m128i*)tmp, sum);
+    return (tmp[3] + tmp[2] + tmp[1] + tmp[0]);
+  }
+}
+
+static int SSE16x16SSE2(const uint8_t* a, const uint8_t* b) {
+  return SSE_Nx4SSE2(a, b, 4, 1);
+}
+
+static int SSE16x8SSE2(const uint8_t* a, const uint8_t* b) {
+  return SSE_Nx4SSE2(a, b, 2, 1);
+}
+
+static int SSE8x8SSE2(const uint8_t* a, const uint8_t* b) {
+  return SSE_Nx4SSE2(a, b, 2, 0);
+}
+
+static int SSE4x4SSE2(const uint8_t* a, const uint8_t* b) {
+  const __m128i zero = _mm_setzero_si128();
+
+  // Load values. Note that we read 8 pixels instead of 4,
+  // but the a/b buffers are over-allocated to that effect.
   const __m128i a0 = _mm_loadl_epi64((__m128i*)&a[BPS * 0]);
   const __m128i a1 = _mm_loadl_epi64((__m128i*)&a[BPS * 1]);
   const __m128i a2 = _mm_loadl_epi64((__m128i*)&a[BPS * 2]);
@@ -483,6 +623,7 @@ static int SSE4x4SSE2(const uint8_t* a, const uint8_t* b) {
   const __m128i sum0 = _mm_add_epi32(madd0, madd1);
   const __m128i sum1 = _mm_add_epi32(madd2, madd3);
   const __m128i sum2 = _mm_add_epi32(sum0, sum1);
+
   int32_t tmp[4];
   _mm_storeu_si128((__m128i*)tmp, sum2);
   return (tmp[3] + tmp[2] + tmp[1] + tmp[0]);
@@ -502,8 +643,6 @@ static int TTransformSSE2(const uint8_t* inA, const uint8_t* inB,
   int32_t sum[4];
   __m128i tmp_0, tmp_1, tmp_2, tmp_3;
   const __m128i zero = _mm_setzero_si128();
-  const __m128i one = _mm_set1_epi16(1);
-  const __m128i three = _mm_set1_epi16(3);
 
   // Load, combine and tranpose inputs.
   {
@@ -550,17 +689,14 @@ static int TTransformSSE2(const uint8_t* inA, const uint8_t* inB,
   // Horizontal pass and subsequent transpose.
   {
     // Calculate a and b (two 4x4 at once).
-    const __m128i a0 = _mm_slli_epi16(_mm_add_epi16(tmp_0, tmp_2), 2);
-    const __m128i a1 = _mm_slli_epi16(_mm_add_epi16(tmp_1, tmp_3), 2);
-    const __m128i a2 = _mm_slli_epi16(_mm_sub_epi16(tmp_1, tmp_3), 2);
-    const __m128i a3 = _mm_slli_epi16(_mm_sub_epi16(tmp_0, tmp_2), 2);
-    // b0_extra = (a0 != 0);
-    const __m128i b0_extra = _mm_andnot_si128(_mm_cmpeq_epi16 (a0, zero), one);
-    const __m128i b0_base = _mm_add_epi16(a0, a1);
+    const __m128i a0 = _mm_add_epi16(tmp_0, tmp_2);
+    const __m128i a1 = _mm_add_epi16(tmp_1, tmp_3);
+    const __m128i a2 = _mm_sub_epi16(tmp_1, tmp_3);
+    const __m128i a3 = _mm_sub_epi16(tmp_0, tmp_2);
+    const __m128i b0 = _mm_add_epi16(a0, a1);
     const __m128i b1 = _mm_add_epi16(a3, a2);
     const __m128i b2 = _mm_sub_epi16(a3, a2);
     const __m128i b3 = _mm_sub_epi16(a0, a1);
-    const __m128i b0 = _mm_add_epi16(b0_base, b0_extra);
     // a00 a01 a02 a03   b00 b01 b02 b03
     // a10 a11 a12 a13   b10 b11 b12 b13
     // a20 a21 a22 a23   b20 b21 b22 b23
@@ -635,19 +771,6 @@ static int TTransformSSE2(const uint8_t* inA, const uint8_t* inB,
       B_b2 = _mm_sub_epi16(B_b2, sign_B_b2);
     }
 
-    // b = abs(b) + 3
-    A_b0 = _mm_add_epi16(A_b0, three);
-    A_b2 = _mm_add_epi16(A_b2, three);
-    B_b0 = _mm_add_epi16(B_b0, three);
-    B_b2 = _mm_add_epi16(B_b2, three);
-
-    // abs((b + (b<0) + 3) >> 3) = (abs(b) + 3) >> 3
-    // b = (abs(b) + 3) >> 3
-    A_b0 = _mm_srai_epi16(A_b0, 3);
-    A_b2 = _mm_srai_epi16(A_b2, 3);
-    B_b0 = _mm_srai_epi16(B_b0, 3);
-    B_b2 = _mm_srai_epi16(B_b2, 3);
-
     // weighted sums
     A_b0 = _mm_madd_epi16(A_b0, w_0);
     A_b2 = _mm_madd_epi16(A_b2, w_8);
@@ -666,7 +789,7 @@ static int TTransformSSE2(const uint8_t* inA, const uint8_t* inB,
 static int Disto4x4SSE2(const uint8_t* const a, const uint8_t* const b,
                         const uint16_t* const w) {
   const int diff_sum = TTransformSSE2(a, b, w);
-  return (abs(diff_sum) + 8) >> 4;
+  return abs(diff_sum) >> 5;
 }
 
 static int Disto16x16SSE2(const uint8_t* const a, const uint8_t* const b,
@@ -681,7 +804,6 @@ static int Disto16x16SSE2(const uint8_t* const a, const uint8_t* const b,
   return D;
 }
 
-
 //------------------------------------------------------------------------------
 // Quantization
 //
@@ -689,9 +811,8 @@ static int Disto16x16SSE2(const uint8_t* const a, const uint8_t* const b,
 // Simple quantization
 static int QuantizeBlockSSE2(int16_t in[16], int16_t out[16],
                              int n, const VP8Matrix* const mtx) {
-  const __m128i max_coeff_2047 = _mm_set1_epi16(2047);
-  const __m128i zero = _mm_set1_epi16(0);
-  __m128i sign0, sign8;
+  const __m128i max_coeff_2047 = _mm_set1_epi16(MAX_LEVEL);
+  const __m128i zero = _mm_setzero_si128();
   __m128i coeff0, coeff8;
   __m128i out0, out8;
   __m128i packed_out;
@@ -713,8 +834,8 @@ static int QuantizeBlockSSE2(int16_t in[16], int16_t out[16],
   const __m128i zthresh8 = _mm_loadu_si128((__m128i*)&mtx->zthresh_[8]);
 
   // sign(in) = in >> 15  (0x0000 if positive, 0xffff if negative)
-  sign0 = _mm_srai_epi16(in0, 15);
-  sign8 = _mm_srai_epi16(in8, 15);
+  const __m128i sign0 = _mm_srai_epi16(in0, 15);
+  const __m128i sign8 = _mm_srai_epi16(in8, 15);
 
   // coeff = abs(in) = (in ^ sign) - sign
   coeff0 = _mm_xor_si128(in0, sign0);
@@ -726,10 +847,6 @@ static int QuantizeBlockSSE2(int16_t in[16], int16_t out[16],
   coeff0 = _mm_add_epi16(coeff0, sharpen0);
   coeff8 = _mm_add_epi16(coeff8, sharpen8);
 
-  // if (coeff > 2047) coeff = 2047
-  coeff0 = _mm_min_epi16(coeff0, max_coeff_2047);
-  coeff8 = _mm_min_epi16(coeff8, max_coeff_2047);
-
   // out = (coeff * iQ + B) >> QFIX;
   {
     // doing calculations with 32b precision (QFIX=17)
@@ -757,9 +874,14 @@ static int QuantizeBlockSSE2(int16_t in[16], int16_t out[16],
     out_04 = _mm_srai_epi32(out_04, QFIX);
     out_08 = _mm_srai_epi32(out_08, QFIX);
     out_12 = _mm_srai_epi32(out_12, QFIX);
+
     // pack result as 16b
     out0 = _mm_packs_epi32(out_00, out_04);
     out8 = _mm_packs_epi32(out_08, out_12);
+
+    // if (coeff > 2047) coeff = 2047
+    out0 = _mm_min_epi16(out0, max_coeff_2047);
+    out8 = _mm_min_epi16(out8, max_coeff_2047);
   }
 
   // get sign back (if (sign[j]) out_n = -out_n)
@@ -832,6 +954,10 @@ void VP8EncDspInitSSE2(void) {
   VP8EncQuantizeBlock = QuantizeBlockSSE2;
   VP8ITransform = ITransformSSE2;
   VP8FTransform = FTransformSSE2;
+  VP8FTransformWHT = FTransformWHTSSE2;
+  VP8SSE16x16 = SSE16x16SSE2;
+  VP8SSE16x8 = SSE16x8SSE2;
+  VP8SSE8x8 = SSE8x8SSE2;
   VP8SSE4x4 = SSE4x4SSE2;
   VP8TDisto4x4 = Disto4x4SSE2;
   VP8TDisto16x16 = Disto16x16SSE2;
diff --git a/3rdparty/libwebp/dsp/lossless.c b/3rdparty/libwebp/dsp/lossless.c
index 62a6b7b15a..e445924ed3 100644
--- a/3rdparty/libwebp/dsp/lossless.c
+++ b/3rdparty/libwebp/dsp/lossless.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Image transforms and color space conversion methods for lossless decoder.
@@ -11,25 +13,31 @@
 //          Jyrki Alakuijala (jyrki@google.com)
 //          Urvang Joshi (urvang@google.com)
 
+#include "./dsp.h"
+
+// Define the following if target arch is sure to have SSE2
+// #define WEBP_TARGET_HAS_SSE2
+
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
 
+#if defined(WEBP_TARGET_HAS_SSE2)
+#include <emmintrin.h>
+#endif
+
 #include <math.h>
 #include <stdlib.h>
 #include "./lossless.h"
 #include "../dec/vp8li.h"
-#include "../dsp/yuv.h"
-#include "../dsp/dsp.h"
-#include "../enc/histogram.h"
+#include "./yuv.h"
 
 #define MAX_DIFF_COST (1e30f)
 
 // lookup table for small values of log2(int)
 #define APPROX_LOG_MAX  4096
 #define LOG_2_RECIPROCAL 1.44269504088896338700465094007086
-#define LOG_LOOKUP_IDX_MAX 256
-static const float kLog2Table[LOG_LOOKUP_IDX_MAX] = {
+const float kLog2Table[LOG_LOOKUP_IDX_MAX] = {
   0.0000000000000000f, 0.0000000000000000f,
   1.0000000000000000f, 1.5849625007211560f,
   2.0000000000000000f, 2.3219280948873621f,
@@ -160,16 +168,97 @@ static const float kLog2Table[LOG_LOOKUP_IDX_MAX] = {
   7.9886846867721654f, 7.9943534368588577f
 };
 
-float VP8LFastLog2(int v) {
-  if (v < LOG_LOOKUP_IDX_MAX) {
-    return kLog2Table[v];
-  } else if (v < APPROX_LOG_MAX) {
+const float kSLog2Table[LOG_LOOKUP_IDX_MAX] = {
+  0.00000000f,    0.00000000f,  2.00000000f,   4.75488750f,
+  8.00000000f,   11.60964047f,  15.50977500f,  19.65148445f,
+  24.00000000f,  28.52932501f,  33.21928095f,  38.05374781f,
+  43.01955001f,  48.10571634f,  53.30296891f,  58.60335893f,
+  64.00000000f,  69.48686830f,  75.05865003f,  80.71062276f,
+  86.43856190f,  92.23866588f,  98.10749561f,  104.04192499f,
+  110.03910002f, 116.09640474f, 122.21143267f, 128.38196256f,
+  134.60593782f, 140.88144886f, 147.20671787f, 153.58008562f,
+  160.00000000f, 166.46500594f, 172.97373660f, 179.52490559f,
+  186.11730005f, 192.74977453f, 199.42124551f, 206.13068654f,
+  212.87712380f, 219.65963219f, 226.47733176f, 233.32938445f,
+  240.21499122f, 247.13338933f, 254.08384998f, 261.06567603f,
+  268.07820003f, 275.12078236f, 282.19280949f, 289.29369244f,
+  296.42286534f, 303.57978409f, 310.76392512f, 317.97478424f,
+  325.21187564f, 332.47473081f, 339.76289772f, 347.07593991f,
+  354.41343574f, 361.77497759f, 369.16017124f, 376.56863518f,
+  384.00000000f, 391.45390785f, 398.93001188f, 406.42797576f,
+  413.94747321f, 421.48818752f, 429.04981119f, 436.63204548f,
+  444.23460010f, 451.85719280f, 459.49954906f, 467.16140179f,
+  474.84249102f, 482.54256363f, 490.26137307f, 497.99867911f,
+  505.75424759f, 513.52785023f, 521.31926438f, 529.12827280f,
+  536.95466351f, 544.79822957f, 552.65876890f, 560.53608414f,
+  568.42998244f, 576.34027536f, 584.26677867f, 592.20931226f,
+  600.16769996f, 608.14176943f, 616.13135206f, 624.13628279f,
+  632.15640007f, 640.19154569f, 648.24156472f, 656.30630539f,
+  664.38561898f, 672.47935976f, 680.58738488f, 688.70955430f,
+  696.84573069f, 704.99577935f, 713.15956818f, 721.33696754f,
+  729.52785023f, 737.73209140f, 745.94956849f, 754.18016116f,
+  762.42375127f, 770.68022275f, 778.94946161f, 787.23135586f,
+  795.52579543f, 803.83267219f, 812.15187982f, 820.48331383f,
+  828.82687147f, 837.18245171f, 845.54995518f, 853.92928416f,
+  862.32034249f, 870.72303558f, 879.13727036f, 887.56295522f,
+  896.00000000f, 904.44831595f, 912.90781569f, 921.37841320f,
+  929.86002376f, 938.35256392f, 946.85595152f, 955.37010560f,
+  963.89494641f, 972.43039537f, 980.97637504f, 989.53280911f,
+  998.09962237f, 1006.67674069f, 1015.26409097f, 1023.86160116f,
+  1032.46920021f, 1041.08681805f, 1049.71438560f, 1058.35183469f,
+  1066.99909811f, 1075.65610955f, 1084.32280357f, 1092.99911564f,
+  1101.68498204f, 1110.38033993f, 1119.08512727f, 1127.79928282f,
+  1136.52274614f, 1145.25545758f, 1153.99735821f, 1162.74838989f,
+  1171.50849518f, 1180.27761738f, 1189.05570047f, 1197.84268914f,
+  1206.63852876f, 1215.44316535f, 1224.25654560f, 1233.07861684f,
+  1241.90932703f, 1250.74862473f, 1259.59645914f, 1268.45278005f,
+  1277.31753781f, 1286.19068338f, 1295.07216828f, 1303.96194457f,
+  1312.85996488f, 1321.76618236f, 1330.68055071f, 1339.60302413f,
+  1348.53355734f, 1357.47210556f, 1366.41862452f, 1375.37307041f,
+  1384.33539991f, 1393.30557020f, 1402.28353887f, 1411.26926400f,
+  1420.26270412f, 1429.26381818f, 1438.27256558f, 1447.28890615f,
+  1456.31280014f, 1465.34420819f, 1474.38309138f, 1483.42941118f,
+  1492.48312945f, 1501.54420843f, 1510.61261078f, 1519.68829949f,
+  1528.77123795f, 1537.86138993f, 1546.95871952f, 1556.06319119f,
+  1565.17476976f, 1574.29342040f, 1583.41910860f, 1592.55180020f,
+  1601.69146137f, 1610.83805860f, 1619.99155871f, 1629.15192882f,
+  1638.31913637f, 1647.49314911f, 1656.67393509f, 1665.86146266f,
+  1675.05570047f, 1684.25661744f, 1693.46418280f, 1702.67836605f,
+  1711.89913698f, 1721.12646563f, 1730.36032233f, 1739.60067768f,
+  1748.84750254f, 1758.10076802f, 1767.36044551f, 1776.62650662f,
+  1785.89892323f, 1795.17766747f, 1804.46271172f, 1813.75402857f,
+  1823.05159087f, 1832.35537170f, 1841.66534438f, 1850.98148244f,
+  1860.30375965f, 1869.63214999f, 1878.96662767f, 1888.30716711f,
+  1897.65374295f, 1907.00633003f, 1916.36490342f, 1925.72943838f,
+  1935.09991037f, 1944.47629506f, 1953.85856831f, 1963.24670620f,
+  1972.64068498f, 1982.04048108f, 1991.44607117f, 2000.85743204f,
+  2010.27454072f, 2019.69737440f, 2029.12591044f, 2038.56012640f
+};
+
+float VP8LFastSLog2Slow(int v) {
+  assert(v >= LOG_LOOKUP_IDX_MAX);
+  if (v < APPROX_LOG_MAX) {
+    int log_cnt = 0;
+    const float v_f = (float)v;
+    while (v >= LOG_LOOKUP_IDX_MAX) {
+      ++log_cnt;
+      v = v >> 1;
+    }
+    return v_f * (kLog2Table[v] + log_cnt);
+  } else {
+    return (float)(LOG_2_RECIPROCAL * v * log((double)v));
+  }
+}
+
+float VP8LFastLog2Slow(int v) {
+  assert(v >= LOG_LOOKUP_IDX_MAX);
+  if (v < APPROX_LOG_MAX) {
     int log_cnt = 0;
     while (v >= LOG_LOOKUP_IDX_MAX) {
       ++log_cnt;
       v = v >> 1;
     }
-    return kLog2Table[v] + (float)log_cnt;
+    return kLog2Table[v] + log_cnt;
   } else {
     return (float)(LOG_2_RECIPROCAL * log((double)v));
   }
@@ -198,6 +287,61 @@ static WEBP_INLINE uint32_t Average4(uint32_t a0, uint32_t a1,
   return Average2(Average2(a0, a1), Average2(a2, a3));
 }
 
+#if defined(WEBP_TARGET_HAS_SSE2)
+static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
+                                                   uint32_t c2) {
+  const __m128i zero = _mm_setzero_si128();
+  const __m128i C0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c0), zero);
+  const __m128i C1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c1), zero);
+  const __m128i C2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c2), zero);
+  const __m128i V1 = _mm_add_epi16(C0, C1);
+  const __m128i V2 = _mm_sub_epi16(V1, C2);
+  const __m128i b = _mm_packus_epi16(V2, V2);
+  const uint32_t output = _mm_cvtsi128_si32(b);
+  return output;
+}
+
+static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
+                                                   uint32_t c2) {
+  const uint32_t ave = Average2(c0, c1);
+  const __m128i zero = _mm_setzero_si128();
+  const __m128i A0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(ave), zero);
+  const __m128i B0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c2), zero);
+  const __m128i A1 = _mm_sub_epi16(A0, B0);
+  const __m128i BgtA = _mm_cmpgt_epi16(B0, A0);
+  const __m128i A2 = _mm_sub_epi16(A1, BgtA);
+  const __m128i A3 = _mm_srai_epi16(A2, 1);
+  const __m128i A4 = _mm_add_epi16(A0, A3);
+  const __m128i A5 = _mm_packus_epi16(A4, A4);
+  const uint32_t output = _mm_cvtsi128_si32(A5);
+  return output;
+}
+
+static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
+  int pa_minus_pb;
+  const __m128i zero = _mm_setzero_si128();
+  const __m128i A0 = _mm_cvtsi32_si128(a);
+  const __m128i B0 = _mm_cvtsi32_si128(b);
+  const __m128i C0 = _mm_cvtsi32_si128(c);
+  const __m128i AC0 = _mm_subs_epu8(A0, C0);
+  const __m128i CA0 = _mm_subs_epu8(C0, A0);
+  const __m128i BC0 = _mm_subs_epu8(B0, C0);
+  const __m128i CB0 = _mm_subs_epu8(C0, B0);
+  const __m128i AC = _mm_or_si128(AC0, CA0);
+  const __m128i BC = _mm_or_si128(BC0, CB0);
+  const __m128i pa = _mm_unpacklo_epi8(AC, zero);  // |a - c|
+  const __m128i pb = _mm_unpacklo_epi8(BC, zero);  // |b - c|
+  const __m128i diff = _mm_sub_epi16(pb, pa);
+  {
+    int16_t out[8];
+    _mm_storeu_si128((__m128i*)out, diff);
+    pa_minus_pb = out[0] + out[1] + out[2] + out[3];
+  }
+  return (pa_minus_pb <= 0) ? a : b;
+}
+
+#else
+
 static WEBP_INLINE uint32_t Clip255(uint32_t a) {
   if (a < 256) {
     return a;
@@ -239,9 +383,9 @@ static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
 }
 
 static WEBP_INLINE int Sub3(int a, int b, int c) {
-  const int pa = b - c;
-  const int pb = a - c;
-  return abs(pa) - abs(pb);
+  const int pb = b - c;
+  const int pa = a - c;
+  return abs(pb) - abs(pa);
 }
 
 static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
@@ -250,9 +394,9 @@ static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
       Sub3((a >> 16) & 0xff, (b >> 16) & 0xff, (c >> 16) & 0xff) +
       Sub3((a >>  8) & 0xff, (b >>  8) & 0xff, (c >>  8) & 0xff) +
       Sub3((a      ) & 0xff, (b      ) & 0xff, (c      ) & 0xff);
-
   return (pa_minus_pb <= 0) ? a : b;
 }
+#endif
 
 //------------------------------------------------------------------------------
 // Predictors
@@ -340,35 +484,36 @@ static float PredictionCostSpatial(const int* counts,
   return (float)(-0.1 * bits);
 }
 
-// Compute the Shanon's entropy: Sum(p*log2(p))
-static float ShannonEntropy(const int* const array, int n) {
+// Compute the combined Shanon's entropy for distribution {X} and {X+Y}
+static float CombinedShannonEntropy(const int* const X,
+                                    const int* const Y, int n) {
   int i;
-  float retval = 0.f;
-  int sum = 0;
+  double retval = 0.;
+  int sumX = 0, sumXY = 0;
   for (i = 0; i < n; ++i) {
-    if (array[i] != 0) {
-      sum += array[i];
-      retval -= VP8LFastSLog2(array[i]);
+    const int x = X[i];
+    const int xy = X[i] + Y[i];
+    if (x != 0) {
+      sumX += x;
+      retval -= VP8LFastSLog2(x);
+    }
+    if (xy != 0) {
+      sumXY += xy;
+      retval -= VP8LFastSLog2(xy);
     }
   }
-  retval += VP8LFastSLog2(sum);
-  return retval;
+  retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
+  return (float)retval;
 }
 
 static float PredictionCostSpatialHistogram(int accumulated[4][256],
                                             int tile[4][256]) {
   int i;
-  int k;
-  int combo[256];
   double retval = 0;
   for (i = 0; i < 4; ++i) {
-    const double exp_val = 0.94;
-    retval += PredictionCostSpatial(&tile[i][0], 1, exp_val);
-    retval += ShannonEntropy(&tile[i][0], 256);
-    for (k = 0; k < 256; ++k) {
-      combo[k] = accumulated[i][k] + tile[i][k];
-    }
-    retval += ShannonEntropy(&combo[0], 256);
+    const double kExpValue = 0.94;
+    retval += PredictionCostSpatial(tile[i], 1, kExpValue);
+    retval += CombinedShannonEntropy(tile[i], accumulated[i], 256);
   }
   return (float)retval;
 }
@@ -572,8 +717,21 @@ static void PredictorInverseTransform(const VP8LTransform* const transform,
 }
 
 void VP8LSubtractGreenFromBlueAndRed(uint32_t* argb_data, int num_pixs) {
-  int i;
-  for (i = 0; i < num_pixs; ++i) {
+  int i = 0;
+#if defined(WEBP_TARGET_HAS_SSE2)
+  const __m128i mask = _mm_set1_epi32(0x0000ff00);
+  for (; i + 4 < num_pixs; i += 4) {
+    const __m128i in = _mm_loadu_si128((__m128i*)&argb_data[i]);
+    const __m128i in_00g0 = _mm_and_si128(in, mask);     // 00g0|00g0|...
+    const __m128i in_0g00 = _mm_slli_epi32(in_00g0, 8);  // 0g00|0g00|...
+    const __m128i in_000g = _mm_srli_epi32(in_00g0, 8);  // 000g|000g|...
+    const __m128i in_0g0g = _mm_or_si128(in_0g00, in_000g);
+    const __m128i out = _mm_sub_epi8(in, in_0g0g);
+    _mm_storeu_si128((__m128i*)&argb_data[i], out);
+  }
+  // fallthrough and finish off with plain-C
+#endif
+  for (; i < num_pixs; ++i) {
     const uint32_t argb = argb_data[i];
     const uint32_t green = (argb >> 8) & 0xff;
     const uint32_t new_r = (((argb >> 16) & 0xff) - green) & 0xff;
@@ -588,9 +746,21 @@ static void AddGreenToBlueAndRed(const VP8LTransform* const transform,
                                  int y_start, int y_end, uint32_t* data) {
   const int width = transform->xsize_;
   const uint32_t* const data_end = data + (y_end - y_start) * width;
+#if defined(WEBP_TARGET_HAS_SSE2)
+  const __m128i mask = _mm_set1_epi32(0x0000ff00);
+  for (; data + 4 < data_end; data += 4) {
+    const __m128i in = _mm_loadu_si128((__m128i*)data);
+    const __m128i in_00g0 = _mm_and_si128(in, mask);     // 00g0|00g0|...
+    const __m128i in_0g00 = _mm_slli_epi32(in_00g0, 8);  // 0g00|0g00|...
+    const __m128i in_000g = _mm_srli_epi32(in_00g0, 8);  // 000g|000g|...
+    const __m128i in_0g0g = _mm_or_si128(in_0g00, in_000g);
+    const __m128i out = _mm_add_epi8(in, in_0g0g);
+    _mm_storeu_si128((__m128i*)data, out);
+  }
+  // fallthrough and finish off with plain-C
+#endif
   while (data < data_end) {
     const uint32_t argb = *data;
-    // "* 0001001u" is equivalent to "(green << 16) + green)"
     const uint32_t green = ((argb >> 8) & 0xff);
     uint32_t red_blue = (argb & 0x00ff00ffu);
     red_blue += (green << 16) | green;
@@ -655,6 +825,25 @@ static WEBP_INLINE uint32_t TransformColor(const Multipliers* const m,
   return (argb & 0xff00ff00u) | (new_red << 16) | (new_blue);
 }
 
+static WEBP_INLINE uint8_t TransformColorRed(uint8_t green_to_red,
+                                             uint32_t argb) {
+  const uint32_t green = argb >> 8;
+  uint32_t new_red = argb >> 16;
+  new_red -= ColorTransformDelta(green_to_red, green);
+  return (new_red & 0xff);
+}
+
+static WEBP_INLINE uint8_t TransformColorBlue(uint8_t green_to_blue,
+                                              uint8_t red_to_blue,
+                                              uint32_t argb) {
+  const uint32_t green = argb >> 8;
+  const uint32_t red = argb >> 16;
+  uint8_t new_blue = argb;
+  new_blue -= ColorTransformDelta(green_to_blue, green);
+  new_blue -= ColorTransformDelta(red_to_blue, red);
+  return (new_blue & 0xff);
+}
+
 static WEBP_INLINE int SkipRepeatedPixels(const uint32_t* const argb,
                                           int ix, int xsize) {
   const uint32_t v = argb[ix];
@@ -675,14 +864,10 @@ static WEBP_INLINE int SkipRepeatedPixels(const uint32_t* const argb,
 static float PredictionCostCrossColor(const int accumulated[256],
                                       const int counts[256]) {
   // Favor low entropy, locally and globally.
-  int i;
-  int combo[256];
-  for (i = 0; i < 256; ++i) {
-    combo[i] = accumulated[i] + counts[i];
-  }
-  return ShannonEntropy(combo, 256) +
-         ShannonEntropy(counts, 256) +
-         PredictionCostSpatial(counts, 3, 2.4);  // Favor small absolute values.
+  // Favor small absolute values for PredictionCostSpatial
+  static const double kExpValue = 2.4;
+  return CombinedShannonEntropy(counts, accumulated, 256) +
+         PredictionCostSpatial(counts, 3, kExpValue);
 }
 
 static Multipliers GetBestColorTransformForTile(
@@ -712,85 +897,75 @@ static Multipliers GetBestColorTransformForTile(
   if (all_y_max > ysize) {
     all_y_max = ysize;
   }
+
   for (green_to_red = -64; green_to_red <= 64; green_to_red += halfstep) {
     int histo[256] = { 0 };
     int all_y;
-    Multipliers tx;
-    MultipliersClear(&tx);
-    tx.green_to_red_ = green_to_red & 0xff;
 
     for (all_y = tile_y_offset; all_y < all_y_max; ++all_y) {
-      uint32_t predict;
       int ix = all_y * xsize + tile_x_offset;
       int all_x;
       for (all_x = tile_x_offset; all_x < all_x_max; ++all_x, ++ix) {
         if (SkipRepeatedPixels(argb, ix, xsize)) {
           continue;
         }
-        predict = TransformColor(&tx, argb[ix], 0);
-        ++histo[(predict >> 16) & 0xff];  // red.
+        ++histo[TransformColorRed(green_to_red, argb[ix])];  // red.
       }
     }
     cur_diff = PredictionCostCrossColor(&accumulated_red_histo[0], &histo[0]);
-    if (tx.green_to_red_ == prevX.green_to_red_) {
+    if ((uint8_t)green_to_red == prevX.green_to_red_) {
       cur_diff -= 3;  // favor keeping the areas locally similar
     }
-    if (tx.green_to_red_ == prevY.green_to_red_) {
+    if ((uint8_t)green_to_red == prevY.green_to_red_) {
       cur_diff -= 3;  // favor keeping the areas locally similar
     }
-    if (tx.green_to_red_ == 0) {
+    if (green_to_red == 0) {
       cur_diff -= 3;
     }
     if (cur_diff < best_diff) {
       best_diff = cur_diff;
-      best_tx = tx;
+      best_tx.green_to_red_ = green_to_red;
     }
   }
   best_diff = MAX_DIFF_COST;
-  green_to_red = best_tx.green_to_red_;
   for (green_to_blue = -32; green_to_blue <= 32; green_to_blue += step) {
     for (red_to_blue = -32; red_to_blue <= 32; red_to_blue += step) {
       int all_y;
       int histo[256] = { 0 };
-      Multipliers tx;
-      tx.green_to_red_ = green_to_red;
-      tx.green_to_blue_ = green_to_blue;
-      tx.red_to_blue_ = red_to_blue;
       for (all_y = tile_y_offset; all_y < all_y_max; ++all_y) {
-        uint32_t predict;
         int all_x;
         int ix = all_y * xsize + tile_x_offset;
         for (all_x = tile_x_offset; all_x < all_x_max; ++all_x, ++ix) {
           if (SkipRepeatedPixels(argb, ix, xsize)) {
             continue;
           }
-          predict = TransformColor(&tx, argb[ix], 0);
-          ++histo[predict & 0xff];  // blue.
+          ++histo[TransformColorBlue(green_to_blue, red_to_blue, argb[ix])];
         }
       }
       cur_diff =
-        PredictionCostCrossColor(&accumulated_blue_histo[0], &histo[0]);
-      if (tx.green_to_blue_ == prevX.green_to_blue_) {
+          PredictionCostCrossColor(&accumulated_blue_histo[0], &histo[0]);
+      if ((uint8_t)green_to_blue == prevX.green_to_blue_) {
         cur_diff -= 3;  // favor keeping the areas locally similar
       }
-      if (tx.green_to_blue_ == prevY.green_to_blue_) {
+      if ((uint8_t)green_to_blue == prevY.green_to_blue_) {
         cur_diff -= 3;  // favor keeping the areas locally similar
       }
-      if (tx.red_to_blue_ == prevX.red_to_blue_) {
+      if ((uint8_t)red_to_blue == prevX.red_to_blue_) {
         cur_diff -= 3;  // favor keeping the areas locally similar
       }
-      if (tx.red_to_blue_ == prevY.red_to_blue_) {
+      if ((uint8_t)red_to_blue == prevY.red_to_blue_) {
         cur_diff -= 3;  // favor keeping the areas locally similar
       }
-      if (tx.green_to_blue_ == 0) {
+      if (green_to_blue == 0) {
         cur_diff -= 3;
       }
-      if (tx.red_to_blue_ == 0) {
+      if (red_to_blue == 0) {
         cur_diff -= 3;
       }
       if (cur_diff < best_diff) {
         best_diff = cur_diff;
-        best_tx = tx;
+        best_tx.green_to_blue_ = green_to_blue;
+        best_tx.red_to_blue_ = red_to_blue;
       }
     }
   }
@@ -920,39 +1095,64 @@ static void ColorSpaceInverseTransform(const VP8LTransform* const transform,
 }
 
 // Separate out pixels packed together using pixel-bundling.
-static void ColorIndexInverseTransform(
-    const VP8LTransform* const transform,
-    int y_start, int y_end, const uint32_t* src, uint32_t* dst) {
-  int y;
-  const int bits_per_pixel = 8 >> transform->bits_;
-  const int width = transform->xsize_;
-  const uint32_t* const color_map = transform->data_;
-  if (bits_per_pixel < 8) {
-    const int pixels_per_byte = 1 << transform->bits_;
-    const int count_mask = pixels_per_byte - 1;
-    const uint32_t bit_mask = (1 << bits_per_pixel) - 1;
-    for (y = y_start; y < y_end; ++y) {
-      uint32_t packed_pixels = 0;
-      int x;
-      for (x = 0; x < width; ++x) {
-        // We need to load fresh 'packed_pixels' once every 'pixels_per_byte'
-        // increments of x. Fortunately, pixels_per_byte is a power of 2, so
-        // can just use a mask for that, instead of decrementing a counter.
-        if ((x & count_mask) == 0) packed_pixels = ((*src++) >> 8) & 0xff;
-        *dst++ = color_map[packed_pixels & bit_mask];
-        packed_pixels >>= bits_per_pixel;
-      }
-    }
-  } else {
-    for (y = y_start; y < y_end; ++y) {
-      int x;
-      for (x = 0; x < width; ++x) {
-        *dst++ = color_map[((*src++) >> 8) & 0xff];
-      }
-    }
-  }
+// We define two methods for ARGB data (uint32_t) and alpha-only data (uint8_t).
+#define COLOR_INDEX_INVERSE(FUNC_NAME, TYPE, GET_INDEX, GET_VALUE)             \
+void FUNC_NAME(const VP8LTransform* const transform,                           \
+               int y_start, int y_end, const TYPE* src, TYPE* dst) {           \
+  int y;                                                                       \
+  const int bits_per_pixel = 8 >> transform->bits_;                            \
+  const int width = transform->xsize_;                                         \
+  const uint32_t* const color_map = transform->data_;                          \
+  if (bits_per_pixel < 8) {                                                    \
+    const int pixels_per_byte = 1 << transform->bits_;                         \
+    const int count_mask = pixels_per_byte - 1;                                \
+    const uint32_t bit_mask = (1 << bits_per_pixel) - 1;                       \
+    for (y = y_start; y < y_end; ++y) {                                        \
+      uint32_t packed_pixels = 0;                                              \
+      int x;                                                                   \
+      for (x = 0; x < width; ++x) {                                            \
+        /* We need to load fresh 'packed_pixels' once every                */  \
+        /* 'pixels_per_byte' increments of x. Fortunately, pixels_per_byte */  \
+        /* is a power of 2, so can just use a mask for that, instead of    */  \
+        /* decrementing a counter.                                         */  \
+        if ((x & count_mask) == 0) packed_pixels = GET_INDEX(*src++);          \
+        *dst++ = GET_VALUE(color_map[packed_pixels & bit_mask]);               \
+        packed_pixels >>= bits_per_pixel;                                      \
+      }                                                                        \
+    }                                                                          \
+  } else {                                                                     \
+    for (y = y_start; y < y_end; ++y) {                                        \
+      int x;                                                                   \
+      for (x = 0; x < width; ++x) {                                            \
+        *dst++ = GET_VALUE(color_map[GET_INDEX(*src++)]);                      \
+      }                                                                        \
+    }                                                                          \
+  }                                                                            \
 }
 
+static WEBP_INLINE uint32_t GetARGBIndex(uint32_t idx) {
+  return (idx >> 8) & 0xff;
+}
+
+static WEBP_INLINE uint8_t GetAlphaIndex(uint8_t idx) {
+  return idx;
+}
+
+static WEBP_INLINE uint32_t GetARGBValue(uint32_t val) {
+  return val;
+}
+
+static WEBP_INLINE uint8_t GetAlphaValue(uint32_t val) {
+  return (val >> 8) & 0xff;
+}
+
+static COLOR_INDEX_INVERSE(ColorIndexInverseTransform, uint32_t, GetARGBIndex,
+                           GetARGBValue)
+COLOR_INDEX_INVERSE(VP8LColorIndexInverseTransformAlpha, uint8_t, GetAlphaIndex,
+                    GetAlphaValue)
+
+#undef COLOR_INDEX_INVERSE
+
 void VP8LInverseTransform(const VP8LTransform* const transform,
                           int row_start, int row_end,
                           const uint32_t* const in, uint32_t* const out) {
@@ -1034,8 +1234,15 @@ static void ConvertBGRAToRGBA4444(const uint32_t* src,
   const uint32_t* const src_end = src + num_pixels;
   while (src < src_end) {
     const uint32_t argb = *src++;
-    *dst++ = ((argb >> 16) & 0xf0) | ((argb >> 12) & 0xf);
-    *dst++ = ((argb >>  0) & 0xf0) | ((argb >> 28) & 0xf);
+    const uint8_t rg = ((argb >> 16) & 0xf0) | ((argb >> 12) & 0xf);
+    const uint8_t ba = ((argb >>  0) & 0xf0) | ((argb >> 28) & 0xf);
+#ifdef WEBP_SWAP_16BIT_CSP
+    *dst++ = ba;
+    *dst++ = rg;
+#else
+    *dst++ = rg;
+    *dst++ = ba;
+#endif
   }
 }
 
@@ -1044,8 +1251,15 @@ static void ConvertBGRAToRGB565(const uint32_t* src,
   const uint32_t* const src_end = src + num_pixels;
   while (src < src_end) {
     const uint32_t argb = *src++;
-    *dst++ = ((argb >> 16) & 0xf8) | ((argb >> 13) & 0x7);
-    *dst++ = ((argb >>  5) & 0xe0) | ((argb >>  3) & 0x1f);
+    const uint8_t rg = ((argb >> 16) & 0xf8) | ((argb >> 13) & 0x7);
+    const uint8_t gb = ((argb >>  5) & 0xe0) | ((argb >>  3) & 0x1f);
+#ifdef WEBP_SWAP_16BIT_CSP
+    *dst++ = gb;
+    *dst++ = rg;
+#else
+    *dst++ = rg;
+    *dst++ = gb;
+#endif
   }
 }
 
@@ -1066,20 +1280,34 @@ static void CopyOrSwap(const uint32_t* src, int num_pixels, uint8_t* dst,
     const uint32_t* const src_end = src + num_pixels;
     while (src < src_end) {
       uint32_t argb = *src++;
-#if !defined(__BIG_ENDIAN__) && (defined(__i386__) || defined(__x86_64__))
+
+#if !defined(__BIG_ENDIAN__)
+#if !defined(WEBP_REFERENCE_IMPLEMENTATION)
+#if defined(__i386__) || defined(__x86_64__)
       __asm__ volatile("bswap %0" : "=r"(argb) : "0"(argb));
       *(uint32_t*)dst = argb;
-      dst += sizeof(argb);
-#elif !defined(__BIG_ENDIAN__) && defined(_MSC_VER)
+#elif defined(_MSC_VER)
       argb = _byteswap_ulong(argb);
       *(uint32_t*)dst = argb;
-      dst += sizeof(argb);
 #else
-      *dst++ = (argb >> 24) & 0xff;
-      *dst++ = (argb >> 16) & 0xff;
-      *dst++ = (argb >>  8) & 0xff;
-      *dst++ = (argb >>  0) & 0xff;
+      dst[0] = (argb >> 24) & 0xff;
+      dst[1] = (argb >> 16) & 0xff;
+      dst[2] = (argb >>  8) & 0xff;
+      dst[3] = (argb >>  0) & 0xff;
 #endif
+#else  // WEBP_REFERENCE_IMPLEMENTATION
+      dst[0] = (argb >> 24) & 0xff;
+      dst[1] = (argb >> 16) & 0xff;
+      dst[2] = (argb >>  8) & 0xff;
+      dst[3] = (argb >>  0) & 0xff;
+#endif
+#else  // __BIG_ENDIAN__
+      dst[0] = (argb >>  0) & 0xff;
+      dst[1] = (argb >>  8) & 0xff;
+      dst[2] = (argb >> 16) & 0xff;
+      dst[3] = (argb >> 24) & 0xff;
+#endif
+      dst += sizeof(argb);
     }
   } else {
     memcpy(dst, src, num_pixels * sizeof(*src));
@@ -1131,6 +1359,27 @@ void VP8LConvertFromBGRA(const uint32_t* const in_data, int num_pixels,
   }
 }
 
+// Bundles multiple (1, 2, 4 or 8) pixels into a single pixel.
+void VP8LBundleColorMap(const uint8_t* const row, int width,
+                        int xbits, uint32_t* const dst) {
+  int x;
+  if (xbits > 0) {
+    const int bit_depth = 1 << (3 - xbits);
+    const int mask = (1 << xbits) - 1;
+    uint32_t code = 0xff000000;
+    for (x = 0; x < width; ++x) {
+      const int xsub = x & mask;
+      if (xsub == 0) {
+        code = 0xff000000;
+      }
+      code |= row[x] << (8 + bit_depth * xsub);
+      dst[x >> xbits] = code;
+    }
+  } else {
+    for (x = 0; x < width; ++x) dst[x] = 0xff000000 | (row[x] << 8);
+  }
+}
+
 //------------------------------------------------------------------------------
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/dsp/lossless.h b/3rdparty/libwebp/dsp/lossless.h
index 992516fcdf..7490ec8efe 100644
--- a/3rdparty/libwebp/dsp/lossless.h
+++ b/3rdparty/libwebp/dsp/lossless.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Image transforms and color space conversion methods for lossless decoder.
@@ -33,6 +35,13 @@ void VP8LInverseTransform(const struct VP8LTransform* const transform,
                           int row_start, int row_end,
                           const uint32_t* const in, uint32_t* const out);
 
+// Similar to the static method ColorIndexInverseTransform() that is part of
+// lossless.c, but used only for alpha decoding. It takes uint8_t (rather than
+// uint32_t) arguments for 'src' and 'dst'.
+void VP8LColorIndexInverseTransformAlpha(
+    const struct VP8LTransform* const transform, int y_start, int y_end,
+    const uint8_t* src, uint8_t* dst);
+
 // Subtracts green from blue and red channels.
 void VP8LSubtractGreenFromBlueAndRed(uint32_t* argb_data, int num_pixs);
 
@@ -59,10 +68,20 @@ static WEBP_INLINE uint32_t VP8LSubSampleSize(uint32_t size,
   return (size + (1 << sampling_bits) - 1) >> sampling_bits;
 }
 
-// Faster logarithm for integers, with the property of log2(0) == 0.
-float VP8LFastLog2(int v);
+// Faster logarithm for integers. Small values use a look-up table.
+#define LOG_LOOKUP_IDX_MAX 256
+extern const float kLog2Table[LOG_LOOKUP_IDX_MAX];
+extern const float kSLog2Table[LOG_LOOKUP_IDX_MAX];
+extern float VP8LFastLog2Slow(int v);
+extern float VP8LFastSLog2Slow(int v);
+static WEBP_INLINE float VP8LFastLog2(int v) {
+  return (v < LOG_LOOKUP_IDX_MAX) ? kLog2Table[v] : VP8LFastLog2Slow(v);
+}
 // Fast calculation of v * log2(v) for integer input.
-static WEBP_INLINE float VP8LFastSLog2(int v) { return VP8LFastLog2(v) * v; }
+static WEBP_INLINE float VP8LFastSLog2(int v) {
+  return (v < LOG_LOOKUP_IDX_MAX) ? kSLog2Table[v] : VP8LFastSLog2Slow(v);
+}
+
 
 // In-place difference of each component with mod 256.
 static WEBP_INLINE uint32_t VP8LSubPixels(uint32_t a, uint32_t b) {
@@ -73,6 +92,9 @@ static WEBP_INLINE uint32_t VP8LSubPixels(uint32_t a, uint32_t b) {
   return (alpha_and_green & 0xff00ff00u) | (red_and_blue & 0x00ff00ffu);
 }
 
+void VP8LBundleColorMap(const uint8_t* const row, int width,
+                        int xbits, uint32_t* const dst);
+
 //------------------------------------------------------------------------------
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/dsp/upsampling.c b/3rdparty/libwebp/dsp/upsampling.c
index 4855eb1432..80ba4f8a5e 100644
--- a/3rdparty/libwebp/dsp/upsampling.c
+++ b/3rdparty/libwebp/dsp/upsampling.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // YUV to RGB upsampling functions.
@@ -32,7 +34,7 @@ WebPUpsampleLinePairFunc WebPUpsamplers[MODE_LAST];
 //  ([3*a +   b + 9*c + 3*d      a + 3*b + 3*c + 9*d]   [8 8]) / 16
 
 // We process u and v together stashed into 32bit (16bit each).
-#define LOAD_UV(u,v) ((u) | ((v) << 16))
+#define LOAD_UV(u, v) ((u) | ((v) << 16))
 
 #define UPSAMPLE_FUNC(FUNC_NAME, FUNC, XSTEP)                                  \
 static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
@@ -327,6 +329,11 @@ void WebPInitUpsamplers(void) {
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitUpsamplersSSE2();
     }
+#endif
+#if defined(WEBP_USE_NEON)
+    if (VP8GetCPUInfo(kNEON)) {
+      WebPInitUpsamplersNEON();
+    }
 #endif
   }
 #endif  // FANCY_UPSAMPLING
@@ -347,6 +354,11 @@ void WebPInitPremultiply(void) {
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitPremultiplySSE2();
     }
+#endif
+#if defined(WEBP_USE_NEON)
+    if (VP8GetCPUInfo(kNEON)) {
+      WebPInitPremultiplyNEON();
+    }
 #endif
   }
 #endif  // FANCY_UPSAMPLING
diff --git a/3rdparty/libwebp/dsp/upsampling_neon.c b/3rdparty/libwebp/dsp/upsampling_neon.c
new file mode 100644
index 0000000000..d118895909
--- /dev/null
+++ b/3rdparty/libwebp/dsp/upsampling_neon.c
@@ -0,0 +1,294 @@
+// Copyright 2011 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// NEON version of YUV to RGB upsampling functions.
+//
+// Author: mans@mansr.com (Mans Rullgard)
+// Based on SSE code by: somnath@google.com (Somnath Banerjee)
+
+#include "./dsp.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+#if defined(WEBP_USE_NEON)
+
+#include <assert.h>
+#include <arm_neon.h>
+#include <string.h>
+#include "./yuv.h"
+
+#ifdef FANCY_UPSAMPLING
+
+// Loads 9 pixels each from rows r1 and r2 and generates 16 pixels.
+#define UPSAMPLE_16PIXELS(r1, r2, out) {                                \
+  uint8x8_t a = vld1_u8(r1);                                            \
+  uint8x8_t b = vld1_u8(r1 + 1);                                        \
+  uint8x8_t c = vld1_u8(r2);                                            \
+  uint8x8_t d = vld1_u8(r2 + 1);                                        \
+                                                                        \
+  uint16x8_t al = vshll_n_u8(a, 1);                                     \
+  uint16x8_t bl = vshll_n_u8(b, 1);                                     \
+  uint16x8_t cl = vshll_n_u8(c, 1);                                     \
+  uint16x8_t dl = vshll_n_u8(d, 1);                                     \
+                                                                        \
+  uint8x8_t diag1, diag2;                                               \
+  uint16x8_t sl;                                                        \
+                                                                        \
+  /* a + b + c + d */                                                   \
+  sl = vaddl_u8(a,  b);                                                 \
+  sl = vaddw_u8(sl, c);                                                 \
+  sl = vaddw_u8(sl, d);                                                 \
+                                                                        \
+  al = vaddq_u16(sl, al); /* 3a +  b +  c +  d */                       \
+  bl = vaddq_u16(sl, bl); /*  a + 3b +  c +  d */                       \
+                                                                        \
+  al = vaddq_u16(al, dl); /* 3a +  b +  c + 3d */                       \
+  bl = vaddq_u16(bl, cl); /*  a + 3b + 3c +  d */                       \
+                                                                        \
+  diag2 = vshrn_n_u16(al, 3);                                           \
+  diag1 = vshrn_n_u16(bl, 3);                                           \
+                                                                        \
+  a = vrhadd_u8(a, diag1);                                              \
+  b = vrhadd_u8(b, diag2);                                              \
+  c = vrhadd_u8(c, diag2);                                              \
+  d = vrhadd_u8(d, diag1);                                              \
+                                                                        \
+  {                                                                     \
+    const uint8x8x2_t a_b = {{ a, b }};                                 \
+    const uint8x8x2_t c_d = {{ c, d }};                                 \
+    vst2_u8(out,      a_b);                                             \
+    vst2_u8(out + 32, c_d);                                             \
+  }                                                                     \
+}
+
+// Turn the macro into a function for reducing code-size when non-critical
+static void Upsample16Pixels(const uint8_t *r1, const uint8_t *r2,
+                             uint8_t *out) {
+  UPSAMPLE_16PIXELS(r1, r2, out);
+}
+
+#define UPSAMPLE_LAST_BLOCK(tb, bb, num_pixels, out) {                  \
+  uint8_t r1[9], r2[9];                                                 \
+  memcpy(r1, (tb), (num_pixels));                                       \
+  memcpy(r2, (bb), (num_pixels));                                       \
+  /* replicate last byte */                                             \
+  memset(r1 + (num_pixels), r1[(num_pixels) - 1], 9 - (num_pixels));    \
+  memset(r2 + (num_pixels), r2[(num_pixels) - 1], 9 - (num_pixels));    \
+  Upsample16Pixels(r1, r2, out);                                        \
+}
+
+#define CY  76283
+#define CVR 89858
+#define CUG 22014
+#define CVG 45773
+#define CUB 113618
+
+static const int16_t coef[4] = { CVR / 4, CUG, CVG / 2, CUB / 4 };
+
+#define CONVERT8(FMT, XSTEP, N, src_y, src_uv, out, cur_x) {            \
+  int i;                                                                \
+  for (i = 0; i < N; i += 8) {                                          \
+    int off = ((cur_x) + i) * XSTEP;                                    \
+    uint8x8_t y  = vld1_u8(src_y + (cur_x)  + i);                       \
+    uint8x8_t u  = vld1_u8((src_uv) + i);                               \
+    uint8x8_t v  = vld1_u8((src_uv) + i + 16);                          \
+    int16x8_t yy = vreinterpretq_s16_u16(vsubl_u8(y, u16));             \
+    int16x8_t uu = vreinterpretq_s16_u16(vsubl_u8(u, u128));            \
+    int16x8_t vv = vreinterpretq_s16_u16(vsubl_u8(v, u128));            \
+                                                                        \
+    int16x8_t ud = vshlq_n_s16(uu, 1);                                  \
+    int16x8_t vd = vshlq_n_s16(vv, 1);                                  \
+                                                                        \
+    int32x4_t vrl = vqdmlal_lane_s16(vshll_n_s16(vget_low_s16(vv), 1),  \
+                                     vget_low_s16(vd),  cf16, 0);       \
+    int32x4_t vrh = vqdmlal_lane_s16(vshll_n_s16(vget_high_s16(vv), 1), \
+                                     vget_high_s16(vd), cf16, 0);       \
+    int16x8_t vr = vcombine_s16(vrshrn_n_s32(vrl, 16),                  \
+                                vrshrn_n_s32(vrh, 16));                 \
+                                                                        \
+    int32x4_t vl = vmovl_s16(vget_low_s16(vv));                         \
+    int32x4_t vh = vmovl_s16(vget_high_s16(vv));                        \
+    int32x4_t ugl = vmlal_lane_s16(vl, vget_low_s16(uu),  cf16, 1);     \
+    int32x4_t ugh = vmlal_lane_s16(vh, vget_high_s16(uu), cf16, 1);     \
+    int32x4_t gcl = vqdmlal_lane_s16(ugl, vget_low_s16(vv),  cf16, 2);  \
+    int32x4_t gch = vqdmlal_lane_s16(ugh, vget_high_s16(vv), cf16, 2);  \
+    int16x8_t gc = vcombine_s16(vrshrn_n_s32(gcl, 16),                  \
+                                vrshrn_n_s32(gch, 16));                 \
+                                                                        \
+    int32x4_t ubl = vqdmlal_lane_s16(vshll_n_s16(vget_low_s16(uu), 1),  \
+                                     vget_low_s16(ud),  cf16, 3);       \
+    int32x4_t ubh = vqdmlal_lane_s16(vshll_n_s16(vget_high_s16(uu), 1), \
+                                     vget_high_s16(ud), cf16, 3);       \
+    int16x8_t ub = vcombine_s16(vrshrn_n_s32(ubl, 16),                  \
+                                vrshrn_n_s32(ubh, 16));                 \
+                                                                        \
+    int32x4_t rl = vaddl_s16(vget_low_s16(yy),  vget_low_s16(vr));      \
+    int32x4_t rh = vaddl_s16(vget_high_s16(yy), vget_high_s16(vr));     \
+    int32x4_t gl = vsubl_s16(vget_low_s16(yy),  vget_low_s16(gc));      \
+    int32x4_t gh = vsubl_s16(vget_high_s16(yy), vget_high_s16(gc));     \
+    int32x4_t bl = vaddl_s16(vget_low_s16(yy),  vget_low_s16(ub));      \
+    int32x4_t bh = vaddl_s16(vget_high_s16(yy), vget_high_s16(ub));     \
+                                                                        \
+    rl = vmulq_lane_s32(rl, cf32, 0);                                   \
+    rh = vmulq_lane_s32(rh, cf32, 0);                                   \
+    gl = vmulq_lane_s32(gl, cf32, 0);                                   \
+    gh = vmulq_lane_s32(gh, cf32, 0);                                   \
+    bl = vmulq_lane_s32(bl, cf32, 0);                                   \
+    bh = vmulq_lane_s32(bh, cf32, 0);                                   \
+                                                                        \
+    y = vqmovun_s16(vcombine_s16(vrshrn_n_s32(rl, 16),                  \
+                                 vrshrn_n_s32(rh, 16)));                \
+    u = vqmovun_s16(vcombine_s16(vrshrn_n_s32(gl, 16),                  \
+                                 vrshrn_n_s32(gh, 16)));                \
+    v = vqmovun_s16(vcombine_s16(vrshrn_n_s32(bl, 16),                  \
+                                 vrshrn_n_s32(bh, 16)));                \
+    STR_ ## FMT(out + off, y, u, v);                                    \
+  }                                                                     \
+}
+
+#define v255 vmov_n_u8(255)
+
+#define STR_Rgb(out, r, g, b) do {                                      \
+  const uint8x8x3_t r_g_b = {{ r, g, b }};                              \
+  vst3_u8(out, r_g_b);                                                  \
+} while (0)
+
+#define STR_Bgr(out, r, g, b) do {                                      \
+  const uint8x8x3_t b_g_r = {{ b, g, r }};                              \
+  vst3_u8(out, b_g_r);                                                  \
+} while (0)
+
+#define STR_Rgba(out, r, g, b) do {                                     \
+  const uint8x8x4_t r_g_b_v255 = {{ r, g, b, v255 }};                   \
+  vst4_u8(out, r_g_b_v255);                                             \
+} while (0)
+
+#define STR_Bgra(out, r, g, b) do {                                     \
+  const uint8x8x4_t b_g_r_v255 = {{ b, g, r, v255 }};                   \
+  vst4_u8(out, b_g_r_v255);                                             \
+} while (0)
+
+#define CONVERT1(FMT, XSTEP, N, src_y, src_uv, rgb, cur_x) {            \
+  int i;                                                                \
+  for (i = 0; i < N; i++) {                                             \
+    int off = ((cur_x) + i) * XSTEP;                                    \
+    int y = src_y[(cur_x) + i];                                         \
+    int u = (src_uv)[i];                                                \
+    int v = (src_uv)[i + 16];                                           \
+    VP8YuvTo ## FMT(y, u, v, rgb + off);                                \
+  }                                                                     \
+}
+
+#define CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, uv,                  \
+                      top_dst, bottom_dst, cur_x, len) {                \
+  if (top_y) {                                                          \
+    CONVERT8(FMT, XSTEP, len, top_y, uv, top_dst, cur_x)                \
+  }                                                                     \
+  if (bottom_y) {                                                       \
+    CONVERT8(FMT, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x)   \
+  }                                                                     \
+}
+
+#define CONVERT2RGB_1(FMT, XSTEP, top_y, bottom_y, uv,                  \
+                      top_dst, bottom_dst, cur_x, len) {                \
+  if (top_y) {                                                          \
+    CONVERT1(FMT, XSTEP, len, top_y, uv, top_dst, cur_x);               \
+  }                                                                     \
+  if (bottom_y) {                                                       \
+    CONVERT1(FMT, XSTEP, len, bottom_y, (uv) + 32, bottom_dst, cur_x);  \
+  }                                                                     \
+}
+
+#define NEON_UPSAMPLE_FUNC(FUNC_NAME, FMT, XSTEP)                       \
+static void FUNC_NAME(const uint8_t *top_y, const uint8_t *bottom_y,    \
+                      const uint8_t *top_u, const uint8_t *top_v,       \
+                      const uint8_t *cur_u, const uint8_t *cur_v,       \
+                      uint8_t *top_dst, uint8_t *bottom_dst, int len) { \
+  int block;                                                            \
+  /* 16 byte aligned array to cache reconstructed u and v */            \
+  uint8_t uv_buf[2 * 32 + 15];                                          \
+  uint8_t *const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);     \
+  const int uv_len = (len + 1) >> 1;                                    \
+  /* 9 pixels must be read-able for each block */                       \
+  const int num_blocks = (uv_len - 1) >> 3;                             \
+  const int leftover = uv_len - num_blocks * 8;                         \
+  const int last_pos = 1 + 16 * num_blocks;                             \
+                                                                        \
+  const int u_diag = ((top_u[0] + cur_u[0]) >> 1) + 1;                  \
+  const int v_diag = ((top_v[0] + cur_v[0]) >> 1) + 1;                  \
+                                                                        \
+  const int16x4_t cf16 = vld1_s16(coef);                                \
+  const int32x2_t cf32 = vmov_n_s32(CY);                                \
+  const uint8x8_t u16  = vmov_n_u8(16);                                 \
+  const uint8x8_t u128 = vmov_n_u8(128);                                \
+                                                                        \
+  /* Treat the first pixel in regular way */                            \
+  if (top_y) {                                                          \
+    const int u0 = (top_u[0] + u_diag) >> 1;                            \
+    const int v0 = (top_v[0] + v_diag) >> 1;                            \
+    VP8YuvTo ## FMT(top_y[0], u0, v0, top_dst);                         \
+  }                                                                     \
+  if (bottom_y) {                                                       \
+    const int u0 = (cur_u[0] + u_diag) >> 1;                            \
+    const int v0 = (cur_v[0] + v_diag) >> 1;                            \
+    VP8YuvTo ## FMT(bottom_y[0], u0, v0, bottom_dst);                   \
+  }                                                                     \
+                                                                        \
+  for (block = 0; block < num_blocks; ++block) {                        \
+    UPSAMPLE_16PIXELS(top_u, cur_u, r_uv);                              \
+    UPSAMPLE_16PIXELS(top_v, cur_v, r_uv + 16);                         \
+    CONVERT2RGB_8(FMT, XSTEP, top_y, bottom_y, r_uv,                    \
+                  top_dst, bottom_dst, 16 * block + 1, 16);             \
+    top_u += 8;                                                         \
+    cur_u += 8;                                                         \
+    top_v += 8;                                                         \
+    cur_v += 8;                                                         \
+  }                                                                     \
+                                                                        \
+  UPSAMPLE_LAST_BLOCK(top_u, cur_u, leftover, r_uv);                    \
+  UPSAMPLE_LAST_BLOCK(top_v, cur_v, leftover, r_uv + 16);               \
+  CONVERT2RGB_1(FMT, XSTEP, top_y, bottom_y, r_uv,                      \
+                top_dst, bottom_dst, last_pos, len - last_pos);         \
+}
+
+// NEON variants of the fancy upsampler.
+NEON_UPSAMPLE_FUNC(UpsampleRgbLinePairNEON,  Rgb,  3)
+NEON_UPSAMPLE_FUNC(UpsampleBgrLinePairNEON,  Bgr,  3)
+NEON_UPSAMPLE_FUNC(UpsampleRgbaLinePairNEON, Rgba, 4)
+NEON_UPSAMPLE_FUNC(UpsampleBgraLinePairNEON, Bgra, 4)
+
+#endif  // FANCY_UPSAMPLING
+
+#endif   // WEBP_USE_NEON
+
+//------------------------------------------------------------------------------
+
+extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */];
+
+void WebPInitUpsamplersNEON(void) {
+#if defined(WEBP_USE_NEON)
+  WebPUpsamplers[MODE_RGB]  = UpsampleRgbLinePairNEON;
+  WebPUpsamplers[MODE_RGBA] = UpsampleRgbaLinePairNEON;
+  WebPUpsamplers[MODE_BGR]  = UpsampleBgrLinePairNEON;
+  WebPUpsamplers[MODE_BGRA] = UpsampleBgraLinePairNEON;
+#endif   // WEBP_USE_NEON
+}
+
+void WebPInitPremultiplyNEON(void) {
+#if defined(WEBP_USE_NEON)
+  WebPUpsamplers[MODE_rgbA] = UpsampleRgbaLinePairNEON;
+  WebPUpsamplers[MODE_bgrA] = UpsampleBgraLinePairNEON;
+#endif   // WEBP_USE_NEON
+}
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
diff --git a/3rdparty/libwebp/dsp/upsampling_sse2.c b/3rdparty/libwebp/dsp/upsampling_sse2.c
index 04cc0aaec3..f31d04845e 100644
--- a/3rdparty/libwebp/dsp/upsampling_sse2.c
+++ b/3rdparty/libwebp/dsp/upsampling_sse2.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // SSE2 version of YUV to RGB upsampling functions.
@@ -51,12 +53,12 @@ extern "C" {
 
 // pack and store two alterning pixel rows
 #define PACK_AND_STORE(a, b, da, db, out) do {                                 \
-  const __m128i ta = _mm_avg_epu8(a, da);  /* (9a + 3b + 3c +  d + 8) / 16 */  \
-  const __m128i tb = _mm_avg_epu8(b, db);  /* (3a + 9b +  c + 3d + 8) / 16 */  \
-  const __m128i t1 = _mm_unpacklo_epi8(ta, tb);                                \
-  const __m128i t2 = _mm_unpackhi_epi8(ta, tb);                                \
-  _mm_store_si128(((__m128i*)(out)) + 0, t1);                                  \
-  _mm_store_si128(((__m128i*)(out)) + 1, t2);                                  \
+  const __m128i t_a = _mm_avg_epu8(a, da);  /* (9a + 3b + 3c +  d + 8) / 16 */ \
+  const __m128i t_b = _mm_avg_epu8(b, db);  /* (3a + 9b +  c + 3d + 8) / 16 */ \
+  const __m128i t_1 = _mm_unpacklo_epi8(t_a, t_b);                             \
+  const __m128i t_2 = _mm_unpackhi_epi8(t_a, t_b);                             \
+  _mm_store_si128(((__m128i*)(out)) + 0, t_1);                                 \
+  _mm_store_si128(((__m128i*)(out)) + 1, t_2);                                 \
 } while (0)
 
 // Loads 17 pixels each from rows r1 and r2 and generates 32 pixels.
@@ -128,7 +130,7 @@ static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
                       const uint8_t* top_u, const uint8_t* top_v,              \
                       const uint8_t* cur_u, const uint8_t* cur_v,              \
                       uint8_t* top_dst, uint8_t* bottom_dst, int len) {        \
-  int b;                                                                       \
+  int block;                                                                   \
   /* 16 byte aligned array to cache reconstructed u and v */                   \
   uint8_t uv_buf[4 * 32 + 15];                                                 \
   uint8_t* const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);            \
@@ -154,11 +156,11 @@ static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
     FUNC(bottom_y[0], u0, v0, bottom_dst);                                     \
   }                                                                            \
                                                                                \
-  for (b = 0; b < num_blocks; ++b) {                                           \
+  for (block = 0; block < num_blocks; ++block) {                               \
     UPSAMPLE_32PIXELS(top_u, cur_u, r_uv + 0 * 32);                            \
     UPSAMPLE_32PIXELS(top_v, cur_v, r_uv + 1 * 32);                            \
     CONVERT2RGB(FUNC, XSTEP, top_y, bottom_y, r_uv, top_dst, bottom_dst,       \
-                32 * b + 1, 32)                                                \
+                32 * block + 1, 32)                                            \
     top_u += 16;                                                               \
     cur_u += 16;                                                               \
     top_v += 16;                                                               \
@@ -211,3 +213,5 @@ void WebPInitPremultiplySSE2(void) {
 #if defined(__cplusplus) || defined(c_plusplus)
 }    // extern "C"
 #endif
+
+
diff --git a/3rdparty/libwebp/dsp/yuv.c b/3rdparty/libwebp/dsp/yuv.c
index 7f05f9a3aa..1a59f7449e 100644
--- a/3rdparty/libwebp/dsp/yuv.c
+++ b/3rdparty/libwebp/dsp/yuv.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // YUV->RGB conversion function
@@ -15,7 +17,7 @@
 extern "C" {
 #endif
 
-enum { YUV_HALF = 1 << (YUV_FIX - 1) };
+#ifdef WEBP_YUV_USE_TABLE
 
 int16_t VP8kVToR[256], VP8kUToB[256];
 int32_t VP8kVToG[256], VP8kUToG[256];
@@ -33,6 +35,7 @@ void VP8YUVInit(void) {
   if (done) {
     return;
   }
+#ifndef USE_YUVj
   for (i = 0; i < 256; ++i) {
     VP8kVToR[i] = (89858 * (i - 128) + YUV_HALF) >> YUV_FIX;
     VP8kUToG[i] = -22014 * (i - 128) + YUV_HALF;
@@ -44,9 +47,29 @@ void VP8YUVInit(void) {
     VP8kClip[i - YUV_RANGE_MIN] = clip(k, 255);
     VP8kClip4Bits[i - YUV_RANGE_MIN] = clip((k + 8) >> 4, 15);
   }
+#else
+  for (i = 0; i < 256; ++i) {
+    VP8kVToR[i] = (91881 * (i - 128) + YUV_HALF) >> YUV_FIX;
+    VP8kUToG[i] = -22554 * (i - 128) + YUV_HALF;
+    VP8kVToG[i] = -46802 * (i - 128);
+    VP8kUToB[i] = (116130 * (i - 128) + YUV_HALF) >> YUV_FIX;
+  }
+  for (i = YUV_RANGE_MIN; i < YUV_RANGE_MAX; ++i) {
+    const int k = i;
+    VP8kClip[i - YUV_RANGE_MIN] = clip(k, 255);
+    VP8kClip4Bits[i - YUV_RANGE_MIN] = clip((k + 8) >> 4, 15);
+  }
+#endif
+
   done = 1;
 }
 
+#else
+
+void VP8YUVInit(void) {}
+
+#endif  // WEBP_YUV_USE_TABLE
+
 #if defined(__cplusplus) || defined(c_plusplus)
 }    // extern "C"
 #endif
diff --git a/3rdparty/libwebp/dsp/yuv.h b/3rdparty/libwebp/dsp/yuv.h
index a569109c54..3844d8cab3 100644
--- a/3rdparty/libwebp/dsp/yuv.h
+++ b/3rdparty/libwebp/dsp/yuv.h
@@ -1,12 +1,34 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // inline YUV<->RGB conversion function
 //
+// The exact naming is Y'CbCr, following the ITU-R BT.601 standard.
+// More information at: http://en.wikipedia.org/wiki/YCbCr
+// Y = 0.2569 * R + 0.5044 * G + 0.0979 * B + 16
+// U = -0.1483 * R - 0.2911 * G + 0.4394 * B + 128
+// V = 0.4394 * R - 0.3679 * G - 0.0715 * B + 128
+// We use 16bit fixed point operations for RGB->YUV conversion.
+//
+// For the Y'CbCr to RGB conversion, the BT.601 specification reads:
+//   R = 1.164 * (Y-16) + 1.596 * (V-128)
+//   G = 1.164 * (Y-16) - 0.813 * (V-128) - 0.391 * (U-128)
+//   B = 1.164 * (Y-16)                   + 2.018 * (U-128)
+// where Y is in the [16,235] range, and U/V in the [16,240] range.
+// In the table-lookup version (WEBP_YUV_USE_TABLE), the common factor
+// "1.164 * (Y-16)" can be handled as an offset in the VP8kClip[] table.
+// So in this case the formulae should be read as:
+//   R = 1.164 * [Y + 1.371 * (V-128)                  ] - 18.624
+//   G = 1.164 * [Y - 0.698 * (V-128) - 0.336 * (U-128)] - 18.624
+//   B = 1.164 * [Y                   + 1.733 * (U-128)] - 18.624
+// once factorized. Here too, 16bit fixed precision is used.
+//
 // Author: Skal (pascal.massimino@gmail.com)
 
 #ifndef WEBP_DSP_YUV_H_
@@ -14,6 +36,19 @@
 
 #include "../dec/decode_vp8.h"
 
+// Define the following to use the LUT-based code:
+#define WEBP_YUV_USE_TABLE
+
+#if defined(WEBP_EXPERIMENTAL_FEATURES)
+// Do NOT activate this feature for real compression. This is only experimental!
+// This flag is for comparison purpose against JPEG's "YUVj" natural colorspace.
+// This colorspace is close to Rec.601's Y'CbCr model with the notable
+// difference of allowing larger range for luma/chroma.
+// See http://en.wikipedia.org/wiki/YCbCr#JPEG_conversion paragraph, and its
+// difference with http://en.wikipedia.org/wiki/YCbCr#ITU-R_BT.601_conversion
+// #define USE_YUVj
+#endif
+
 //------------------------------------------------------------------------------
 // YUV -> RGB conversion
 
@@ -22,9 +57,14 @@ extern "C" {
 #endif
 
 enum { YUV_FIX = 16,                // fixed-point precision
+       YUV_HALF = 1 << (YUV_FIX - 1),
+       YUV_MASK = (256 << YUV_FIX) - 1,
        YUV_RANGE_MIN = -227,        // min value of r/g/b output
        YUV_RANGE_MAX = 256 + 226    // max value of r/g/b output
 };
+
+#ifdef WEBP_YUV_USE_TABLE
+
 extern int16_t VP8kVToR[256], VP8kUToB[256];
 extern int32_t VP8kVToG[256], VP8kUToG[256];
 extern uint8_t VP8kClip[YUV_RANGE_MAX - YUV_RANGE_MIN];
@@ -40,34 +80,6 @@ static WEBP_INLINE void VP8YuvToRgb(uint8_t y, uint8_t u, uint8_t v,
   rgb[2] = VP8kClip[y + b_off - YUV_RANGE_MIN];
 }
 
-static WEBP_INLINE void VP8YuvToRgb565(uint8_t y, uint8_t u, uint8_t v,
-                                       uint8_t* const rgb) {
-  const int r_off = VP8kVToR[v];
-  const int g_off = (VP8kVToG[v] + VP8kUToG[u]) >> YUV_FIX;
-  const int b_off = VP8kUToB[u];
-  rgb[0] = ((VP8kClip[y + r_off - YUV_RANGE_MIN] & 0xf8) |
-            (VP8kClip[y + g_off - YUV_RANGE_MIN] >> 5));
-  rgb[1] = (((VP8kClip[y + g_off - YUV_RANGE_MIN] << 3) & 0xe0) |
-            (VP8kClip[y + b_off - YUV_RANGE_MIN] >> 3));
-}
-
-static WEBP_INLINE void VP8YuvToArgb(uint8_t y, uint8_t u, uint8_t v,
-                                     uint8_t* const argb) {
-  argb[0] = 0xff;
-  VP8YuvToRgb(y, u, v, argb + 1);
-}
-
-static WEBP_INLINE void VP8YuvToRgba4444(uint8_t y, uint8_t u, uint8_t v,
-                                         uint8_t* const argb) {
-  const int r_off = VP8kVToR[v];
-  const int g_off = (VP8kVToG[v] + VP8kUToG[u]) >> YUV_FIX;
-  const int b_off = VP8kUToB[u];
-  // Don't update alpha (last 4 bits of argb[1])
-  argb[0] = ((VP8kClip4Bits[y + r_off - YUV_RANGE_MIN] << 4) |
-             VP8kClip4Bits[y + g_off - YUV_RANGE_MIN]);
-  argb[1] = 0x0f | (VP8kClip4Bits[y + b_off - YUV_RANGE_MIN] << 4);
-}
-
 static WEBP_INLINE void VP8YuvToBgr(uint8_t y, uint8_t u, uint8_t v,
                                     uint8_t* const bgr) {
   const int r_off = VP8kVToR[v];
@@ -78,6 +90,129 @@ static WEBP_INLINE void VP8YuvToBgr(uint8_t y, uint8_t u, uint8_t v,
   bgr[2] = VP8kClip[y + r_off - YUV_RANGE_MIN];
 }
 
+static WEBP_INLINE void VP8YuvToRgb565(uint8_t y, uint8_t u, uint8_t v,
+                                       uint8_t* const rgb) {
+  const int r_off = VP8kVToR[v];
+  const int g_off = (VP8kVToG[v] + VP8kUToG[u]) >> YUV_FIX;
+  const int b_off = VP8kUToB[u];
+  const uint8_t rg = ((VP8kClip[y + r_off - YUV_RANGE_MIN] & 0xf8) |
+                      (VP8kClip[y + g_off - YUV_RANGE_MIN] >> 5));
+  const uint8_t gb = (((VP8kClip[y + g_off - YUV_RANGE_MIN] << 3) & 0xe0) |
+                      (VP8kClip[y + b_off - YUV_RANGE_MIN] >> 3));
+#ifdef WEBP_SWAP_16BIT_CSP
+  rgb[0] = gb;
+  rgb[1] = rg;
+#else
+  rgb[0] = rg;
+  rgb[1] = gb;
+#endif
+}
+
+static WEBP_INLINE void VP8YuvToRgba4444(uint8_t y, uint8_t u, uint8_t v,
+                                         uint8_t* const argb) {
+  const int r_off = VP8kVToR[v];
+  const int g_off = (VP8kVToG[v] + VP8kUToG[u]) >> YUV_FIX;
+  const int b_off = VP8kUToB[u];
+  const uint8_t rg = ((VP8kClip4Bits[y + r_off - YUV_RANGE_MIN] << 4) |
+                      VP8kClip4Bits[y + g_off - YUV_RANGE_MIN]);
+  const uint8_t ba = (VP8kClip4Bits[y + b_off - YUV_RANGE_MIN] << 4) | 0x0f;
+#ifdef WEBP_SWAP_16BIT_CSP
+  argb[0] = ba;
+  argb[1] = rg;
+#else
+  argb[0] = rg;
+  argb[1] = ba;
+#endif
+}
+
+#else   // Table-free version (slower on x86)
+
+// These constants are 16b fixed-point version of ITU-R BT.601 constants
+#define kYScale 76309      // 1.164 = 255 / 219
+#define kVToR   104597     // 1.596 = 255 / 112 * 0.701
+#define kUToG   25674      // 0.391 = 255 / 112 * 0.886 * 0.114 / 0.587
+#define kVToG   53278      // 0.813 = 255 / 112 * 0.701 * 0.299 / 0.587
+#define kUToB   132201     // 2.018 = 255 / 112 * 0.886
+#define kRCst (-kYScale * 16 - kVToR * 128 + YUV_HALF)
+#define kGCst (-kYScale * 16 + kUToG * 128 + kVToG * 128 + YUV_HALF)
+#define kBCst (-kYScale * 16 - kUToB * 128 + YUV_HALF)
+
+static WEBP_INLINE uint8_t VP8Clip8(int v) {
+  return ((v & ~YUV_MASK) == 0) ? (uint8_t)(v >> YUV_FIX)
+                                : (v < 0) ? 0u : 255u;
+}
+
+static WEBP_INLINE uint8_t VP8ClipN(int v, int N) {  // clip to N bits
+  return ((v & ~YUV_MASK) == 0) ? (uint8_t)(v >> (YUV_FIX + (8 - N)))
+                                : (v < 0) ? 0u : (255u >> (8 - N));
+}
+
+static WEBP_INLINE int VP8YUVToR(int y, int v) {
+  return kYScale * y + kVToR * v + kRCst;
+}
+
+static WEBP_INLINE int VP8YUVToG(int y, int u, int v) {
+  return kYScale * y - kUToG * u - kVToG * v + kGCst;
+}
+
+static WEBP_INLINE int VP8YUVToB(int y, int u) {
+  return kYScale * y  + kUToB * u + kBCst;
+}
+
+static WEBP_INLINE void VP8YuvToRgb(uint8_t y, uint8_t u, uint8_t v,
+                                    uint8_t* const rgb) {
+  rgb[0] = VP8Clip8(VP8YUVToR(y, v));
+  rgb[1] = VP8Clip8(VP8YUVToG(y, u, v));
+  rgb[2] = VP8Clip8(VP8YUVToB(y, u));
+}
+
+static WEBP_INLINE void VP8YuvToBgr(uint8_t y, uint8_t u, uint8_t v,
+                                    uint8_t* const bgr) {
+  bgr[0] = VP8Clip8(VP8YUVToB(y, u));
+  bgr[1] = VP8Clip8(VP8YUVToG(y, u, v));
+  bgr[2] = VP8Clip8(VP8YUVToR(y, v));
+}
+
+static WEBP_INLINE void VP8YuvToRgb565(uint8_t y, uint8_t u, uint8_t v,
+                                       uint8_t* const rgb) {
+  const int r = VP8Clip8(VP8YUVToR(y, u));
+  const int g = VP8ClipN(VP8YUVToG(y, u, v), 6);
+  const int b = VP8ClipN(VP8YUVToB(y, v), 5);
+  const uint8_t rg = (r & 0xf8) | (g >> 3);
+  const uint8_t gb = (g << 5) | b;
+#ifdef WEBP_SWAP_16BIT_CSP
+  rgb[0] = gb;
+  rgb[1] = rg;
+#else
+  rgb[0] = rg;
+  rgb[1] = gb;
+#endif
+}
+
+static WEBP_INLINE void VP8YuvToRgba4444(uint8_t y, uint8_t u, uint8_t v,
+                                         uint8_t* const argb) {
+  const int r = VP8Clip8(VP8YUVToR(y, u));
+  const int g = VP8ClipN(VP8YUVToG(y, u, v), 4);
+  const int b = VP8Clip8(VP8YUVToB(y, v));
+  const uint8_t rg = (r & 0xf0) | g;
+  const uint8_t ba = b | 0x0f;   // overwrite the lower 4 bits
+#ifdef WEBP_SWAP_16BIT_CSP
+  argb[0] = ba;
+  argb[1] = rg;
+#else
+  argb[0] = rg;
+  argb[1] = ba;
+#endif
+}
+
+#endif  // WEBP_YUV_USE_TABLE
+
+static WEBP_INLINE void VP8YuvToArgb(uint8_t y, uint8_t u, uint8_t v,
+                                     uint8_t* const argb) {
+  argb[0] = 0xff;
+  VP8YuvToRgb(y, u, v, argb + 1);
+}
+
 static WEBP_INLINE void VP8YuvToBgra(uint8_t y, uint8_t u, uint8_t v,
                                      uint8_t* const bgra) {
   VP8YuvToBgr(y, u, v, bgra);
@@ -95,18 +230,14 @@ void VP8YUVInit(void);
 
 //------------------------------------------------------------------------------
 // RGB -> YUV conversion
-// The exact naming is Y'CbCr, following the ITU-R BT.601 standard.
-// More information at: http://en.wikipedia.org/wiki/YCbCr
-// Y = 0.2569 * R + 0.5044 * G + 0.0979 * B + 16
-// U = -0.1483 * R - 0.2911 * G + 0.4394 * B + 128
-// V = 0.4394 * R - 0.3679 * G - 0.0715 * B + 128
-// We use 16bit fixed point operations.
 
 static WEBP_INLINE int VP8ClipUV(int v) {
-   v = (v + (257 << (YUV_FIX + 2 - 1))) >> (YUV_FIX + 2);
-   return ((v & ~0xff) == 0) ? v : (v < 0) ? 0 : 255;
+  v = (v + (257 << (YUV_FIX + 2 - 1))) >> (YUV_FIX + 2);
+  return ((v & ~0xff) == 0) ? v : (v < 0) ? 0 : 255;
 }
 
+#ifndef USE_YUVj
+
 static WEBP_INLINE int VP8RGBToY(int r, int g, int b) {
   const int kRound = (1 << (YUV_FIX - 1)) + (16 << YUV_FIX);
   const int luma = 16839 * r + 33059 * g + 6420 * b;
@@ -114,13 +245,38 @@ static WEBP_INLINE int VP8RGBToY(int r, int g, int b) {
 }
 
 static WEBP_INLINE int VP8RGBToU(int r, int g, int b) {
-  return VP8ClipUV(-9719 * r - 19081 * g + 28800 * b);
+  const int u = -9719 * r - 19081 * g + 28800 * b;
+  return VP8ClipUV(u);
 }
 
 static WEBP_INLINE int VP8RGBToV(int r, int g, int b) {
-  return VP8ClipUV(+28800 * r - 24116 * g - 4684 * b);
+  const int v = +28800 * r - 24116 * g - 4684 * b;
+  return VP8ClipUV(v);
 }
 
+#else
+
+// This JPEG-YUV colorspace, only for comparison!
+// These are also 16-bit precision coefficients from Rec.601, but with full
+// [0..255] output range.
+static WEBP_INLINE int VP8RGBToY(int r, int g, int b) {
+  const int kRound = (1 << (YUV_FIX - 1));
+  const int luma = 19595 * r + 38470 * g + 7471 * b;
+  return (luma + kRound) >> YUV_FIX;  // no need to clip
+}
+
+static WEBP_INLINE int VP8RGBToU(int r, int g, int b) {
+  const int u = -11058 * r - 21710 * g + 32768 * b;
+  return VP8ClipUV(u);
+}
+
+static WEBP_INLINE int VP8RGBToV(int r, int g, int b) {
+  const int v = 32768 * r - 27439 * g - 5329 * b;
+  return VP8ClipUV(v);
+}
+
+#endif    // USE_YUVj
+
 #if defined(__cplusplus) || defined(c_plusplus)
 }    // extern "C"
 #endif
diff --git a/3rdparty/libwebp/enc/alpha.c b/3rdparty/libwebp/enc/alpha.c
index 0e519b6c66..e636c96723 100644
--- a/3rdparty/libwebp/enc/alpha.c
+++ b/3rdparty/libwebp/enc/alpha.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Alpha-plane compression.
@@ -79,18 +81,17 @@ static int EncodeLossless(const uint8_t* const data, int width, int height,
   WebPConfigInit(&config);
   config.lossless = 1;
   config.method = effort_level;  // impact is very small
-  // Set moderate default quality setting for alpha. Higher qualities (80 and
-  // above) could be very slow.
-  config.quality = 10.f + 15.f * effort_level;
-  if (config.quality > 100.f) config.quality = 100.f;
+  // Set a moderate default quality setting for alpha.
+  config.quality = 10.f * effort_level;
+  assert(config.quality >= 0 && config.quality <= 100.f);
 
   ok = VP8LBitWriterInit(&tmp_bw, (width * height) >> 3);
   ok = ok && (VP8LEncodeStream(&config, &picture, &tmp_bw) == VP8_ENC_OK);
   WebPPictureFree(&picture);
   if (ok) {
-    const uint8_t* const data = VP8LBitWriterFinish(&tmp_bw);
-    const size_t data_size = VP8LBitWriterNumBytes(&tmp_bw);
-    VP8BitWriterAppend(bw, data, data_size);
+    const uint8_t* const buffer = VP8LBitWriterFinish(&tmp_bw);
+    const size_t buffer_size = VP8LBitWriterNumBytes(&tmp_bw);
+    VP8BitWriterAppend(bw, buffer, buffer_size);
   }
   VP8LBitWriterDestroy(&tmp_bw);
   return ok && !bw->error_;
@@ -128,8 +129,8 @@ static int EncodeAlphaInternal(const uint8_t* const data, int width, int height,
   VP8BitWriterAppend(bw, &header, ALPHA_HEADER_LEN);
 
   filter_func = WebPFilters[filter];
-  if (filter_func) {
-    filter_func(data, width, height, 1, width, tmp_alpha);
+  if (filter_func != NULL) {
+    filter_func(data, width, height, width, tmp_alpha);
     alpha_src = tmp_alpha;
   }  else {
     alpha_src = data;
@@ -157,6 +158,25 @@ static void CopyPlane(const uint8_t* src, int src_stride,
   }
 }
 
+static int GetNumColors(const uint8_t* data, int width, int height,
+                        int stride) {
+  int j;
+  int colors = 0;
+  uint8_t color[256] = { 0 };
+
+  for (j = 0; j < height; ++j) {
+    int i;
+    const uint8_t* const p = data + j * stride;
+    for (i = 0; i < width; ++i) {
+      color[p[i]] = 1;
+    }
+  }
+  for (j = 0; j < 256; ++j) {
+    if (color[j] > 0) ++colors;
+  }
+  return colors;
+}
+
 static int EncodeAlpha(VP8Encoder* const enc,
                        int quality, int method, int filter,
                        int effort_level,
@@ -208,18 +228,32 @@ static int EncodeAlpha(VP8Encoder* const enc,
     VP8BitWriter bw;
     int test_filter;
     uint8_t* filtered_alpha = NULL;
+    int try_filter_none = (effort_level > 3);
 
-    // We always test WEBP_FILTER_NONE first.
-    ok = EncodeAlphaInternal(quant_alpha, width, height,
-                             method, WEBP_FILTER_NONE, reduce_levels,
-                             effort_level, NULL, &bw, pic->stats);
-    if (!ok) {
-      VP8BitWriterWipeOut(&bw);
-      goto End;
+    if (filter == WEBP_FILTER_FAST) {  // Quick estimate of the best candidate.
+      const int kMinColorsForFilterNone = 16;
+      const int kMaxColorsForFilterNone = 192;
+      const int num_colors = GetNumColors(quant_alpha, width, height, width);
+      // For low number of colors, NONE yeilds better compression.
+      filter = (num_colors <= kMinColorsForFilterNone) ? WEBP_FILTER_NONE :
+               EstimateBestFilter(quant_alpha, width, height, width);
+      // For large number of colors, try FILTER_NONE in addition to the best
+      // filter as well.
+      if (num_colors > kMaxColorsForFilterNone) {
+        try_filter_none = 1;
+      }
     }
 
-    if (filter == WEBP_FILTER_FAST) {  // Quick estimate of a second candidate?
-      filter = EstimateBestFilter(quant_alpha, width, height, width);
+    // Test for WEBP_FILTER_NONE for higher effort levels.
+    if (try_filter_none || filter == WEBP_FILTER_NONE) {
+      ok = EncodeAlphaInternal(quant_alpha, width, height,
+                               method, WEBP_FILTER_NONE, reduce_levels,
+                               effort_level, NULL, &bw, pic->stats);
+
+      if (!ok) {
+        VP8BitWriterWipeOut(&bw);
+        goto End;
+      }
     }
     // Stop?
     if (filter == WEBP_FILTER_NONE) {
@@ -235,11 +269,14 @@ static int EncodeAlpha(VP8Encoder* const enc,
     // Try the other mode(s).
     {
       WebPAuxStats best_stats;
-      size_t best_score = VP8BitWriterSize(&bw);
+      size_t best_score = try_filter_none ?
+                          VP8BitWriterSize(&bw) : (size_t)~0U;
+      int wipe_tmp_bw = try_filter_none;
 
       memset(&best_stats, 0, sizeof(best_stats));  // prevent spurious warning
       if (pic->stats != NULL) best_stats = *pic->stats;
-      for (test_filter = WEBP_FILTER_HORIZONTAL;
+      for (test_filter =
+           try_filter_none ? WEBP_FILTER_HORIZONTAL : WEBP_FILTER_NONE;
            ok && (test_filter <= WEBP_FILTER_GRADIENT);
            ++test_filter) {
         VP8BitWriter tmp_bw;
@@ -263,7 +300,10 @@ static int EncodeAlpha(VP8Encoder* const enc,
         } else {
           VP8BitWriterWipeOut(&bw);
         }
-        VP8BitWriterWipeOut(&tmp_bw);
+        if (wipe_tmp_bw) {
+          VP8BitWriterWipeOut(&tmp_bw);
+        }
+        wipe_tmp_bw = 1;  // For next filter trial for WEBP_FILTER_BEST.
       }
       if (pic->stats != NULL) *pic->stats = best_stats;
     }
@@ -287,42 +327,80 @@ static int EncodeAlpha(VP8Encoder* const enc,
 //------------------------------------------------------------------------------
 // Main calls
 
+static int CompressAlphaJob(VP8Encoder* const enc, void* dummy) {
+  const WebPConfig* config = enc->config_;
+  uint8_t* alpha_data = NULL;
+  size_t alpha_size = 0;
+  const int effort_level = config->method;  // maps to [0..6]
+  const WEBP_FILTER_TYPE filter =
+      (config->alpha_filtering == 0) ? WEBP_FILTER_NONE :
+      (config->alpha_filtering == 1) ? WEBP_FILTER_FAST :
+                                       WEBP_FILTER_BEST;
+  if (!EncodeAlpha(enc, config->alpha_quality, config->alpha_compression,
+                   filter, effort_level, &alpha_data, &alpha_size)) {
+    return 0;
+  }
+  if (alpha_size != (uint32_t)alpha_size) {  // Sanity check.
+    free(alpha_data);
+    return 0;
+  }
+  enc->alpha_data_size_ = (uint32_t)alpha_size;
+  enc->alpha_data_ = alpha_data;
+  (void)dummy;
+  return 1;
+}
+
 void VP8EncInitAlpha(VP8Encoder* const enc) {
   enc->has_alpha_ = WebPPictureHasTransparency(enc->pic_);
   enc->alpha_data_ = NULL;
   enc->alpha_data_size_ = 0;
+  if (enc->thread_level_ > 0) {
+    WebPWorker* const worker = &enc->alpha_worker_;
+    WebPWorkerInit(worker);
+    worker->data1 = enc;
+    worker->data2 = NULL;
+    worker->hook = (WebPWorkerHook)CompressAlphaJob;
+  }
+}
+
+int VP8EncStartAlpha(VP8Encoder* const enc) {
+  if (enc->has_alpha_) {
+    if (enc->thread_level_ > 0) {
+      WebPWorker* const worker = &enc->alpha_worker_;
+      if (!WebPWorkerReset(worker)) {    // Makes sure worker is good to go.
+        return 0;
+      }
+      WebPWorkerLaunch(worker);
+      return 1;
+    } else {
+      return CompressAlphaJob(enc, NULL);   // just do the job right away
+    }
+  }
+  return 1;
 }
 
 int VP8EncFinishAlpha(VP8Encoder* const enc) {
   if (enc->has_alpha_) {
-    const WebPConfig* config = enc->config_;
-    uint8_t* tmp_data = NULL;
-    size_t tmp_size = 0;
-    const int effort_level = config->method;  // maps to [0..6]
-    const WEBP_FILTER_TYPE filter =
-        (config->alpha_filtering == 0) ? WEBP_FILTER_NONE :
-        (config->alpha_filtering == 1) ? WEBP_FILTER_FAST :
-                                         WEBP_FILTER_BEST;
-
-    if (!EncodeAlpha(enc, config->alpha_quality, config->alpha_compression,
-                     filter, effort_level, &tmp_data, &tmp_size)) {
-      return 0;
+    if (enc->thread_level_ > 0) {
+      WebPWorker* const worker = &enc->alpha_worker_;
+      if (!WebPWorkerSync(worker)) return 0;  // error
     }
-    if (tmp_size != (uint32_t)tmp_size) {  // Sanity check.
-      free(tmp_data);
-      return 0;
-    }
-    enc->alpha_data_size_ = (uint32_t)tmp_size;
-    enc->alpha_data_ = tmp_data;
   }
   return WebPReportProgress(enc->pic_, enc->percent_ + 20, &enc->percent_);
 }
 
-void VP8EncDeleteAlpha(VP8Encoder* const enc) {
+int VP8EncDeleteAlpha(VP8Encoder* const enc) {
+  int ok = 1;
+  if (enc->thread_level_ > 0) {
+    WebPWorker* const worker = &enc->alpha_worker_;
+    ok = WebPWorkerSync(worker);  // finish anything left in flight
+    WebPWorkerEnd(worker);  // still need to end the worker, even if !ok
+  }
   free(enc->alpha_data_);
   enc->alpha_data_ = NULL;
   enc->alpha_data_size_ = 0;
   enc->has_alpha_ = 0;
+  return ok;
 }
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/enc/analysis.c b/3rdparty/libwebp/enc/analysis.c
index 22cfb492e7..4ff3edd2a7 100644
--- a/3rdparty/libwebp/enc/analysis.c
+++ b/3rdparty/libwebp/enc/analysis.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Macroblock analysis
@@ -23,10 +25,6 @@ extern "C" {
 
 #define MAX_ITERS_K_MEANS  6
 
-static int ClipAlpha(int alpha) {
-  return alpha < 0 ? 0 : alpha > 255 ? 255 : alpha;
-}
-
 //------------------------------------------------------------------------------
 // Smooth the segment map by replacing isolated block by the majority of its
 // neighbours.
@@ -72,50 +70,10 @@ static void SmoothSegmentMap(VP8Encoder* const enc) {
 }
 
 //------------------------------------------------------------------------------
-// Finalize Segment probability based on the coding tree
-
-static int GetProba(int a, int b) {
-  int proba;
-  const int total = a + b;
-  if (total == 0) return 255;  // that's the default probability.
-  proba = (255 * a + total / 2) / total;
-  return proba;
-}
-
-static void SetSegmentProbas(VP8Encoder* const enc) {
-  int p[NUM_MB_SEGMENTS] = { 0 };
-  int n;
-
-  for (n = 0; n < enc->mb_w_ * enc->mb_h_; ++n) {
-    const VP8MBInfo* const mb = &enc->mb_info_[n];
-    p[mb->segment_]++;
-  }
-  if (enc->pic_->stats) {
-    for (n = 0; n < NUM_MB_SEGMENTS; ++n) {
-      enc->pic_->stats->segment_size[n] = p[n];
-    }
-  }
-  if (enc->segment_hdr_.num_segments_ > 1) {
-    uint8_t* const probas = enc->proba_.segments_;
-    probas[0] = GetProba(p[0] + p[1], p[2] + p[3]);
-    probas[1] = GetProba(p[0], p[1]);
-    probas[2] = GetProba(p[2], p[3]);
-
-    enc->segment_hdr_.update_map_ =
-        (probas[0] != 255) || (probas[1] != 255) || (probas[2] != 255);
-    enc->segment_hdr_.size_ =
-      p[0] * (VP8BitCost(0, probas[0]) + VP8BitCost(0, probas[1])) +
-      p[1] * (VP8BitCost(0, probas[0]) + VP8BitCost(1, probas[1])) +
-      p[2] * (VP8BitCost(1, probas[0]) + VP8BitCost(0, probas[2])) +
-      p[3] * (VP8BitCost(1, probas[0]) + VP8BitCost(1, probas[2]));
-  } else {
-    enc->segment_hdr_.update_map_ = 0;
-    enc->segment_hdr_.size_ = 0;
-  }
-}
+// set segment susceptibility alpha_ / beta_
 
 static WEBP_INLINE int clip(int v, int m, int M) {
-  return v < m ? m : v > M ? M : v;
+  return (v < m) ? m : (v > M) ? M : v;
 }
 
 static void SetSegmentAlphas(VP8Encoder* const enc,
@@ -141,23 +99,64 @@ static void SetSegmentAlphas(VP8Encoder* const enc,
   }
 }
 
+//------------------------------------------------------------------------------
+// Compute susceptibility based on DCT-coeff histograms:
+// the higher, the "easier" the macroblock is to compress.
+
+#define MAX_ALPHA 255                // 8b of precision for susceptibilities.
+#define ALPHA_SCALE (2 * MAX_ALPHA)  // scaling factor for alpha.
+#define DEFAULT_ALPHA (-1)
+#define IS_BETTER_ALPHA(alpha, best_alpha) ((alpha) > (best_alpha))
+
+static int FinalAlphaValue(int alpha) {
+  alpha = MAX_ALPHA - alpha;
+  return clip(alpha, 0, MAX_ALPHA);
+}
+
+static int GetAlpha(const VP8Histogram* const histo) {
+  int max_value = 0, last_non_zero = 1;
+  int k;
+  int alpha;
+  for (k = 0; k <= MAX_COEFF_THRESH; ++k) {
+    const int value = histo->distribution[k];
+    if (value > 0) {
+      if (value > max_value) max_value = value;
+      last_non_zero = k;
+    }
+  }
+  // 'alpha' will later be clipped to [0..MAX_ALPHA] range, clamping outer
+  // values which happen to be mostly noise. This leaves the maximum precision
+  // for handling the useful small values which contribute most.
+  alpha = (max_value > 1) ? ALPHA_SCALE * last_non_zero / max_value : 0;
+  return alpha;
+}
+
+static void MergeHistograms(const VP8Histogram* const in,
+                            VP8Histogram* const out) {
+  int i;
+  for (i = 0; i <= MAX_COEFF_THRESH; ++i) {
+    out->distribution[i] += in->distribution[i];
+  }
+}
+
 //------------------------------------------------------------------------------
 // Simplified k-Means, to assign Nb segments based on alpha-histogram
 
-static void AssignSegments(VP8Encoder* const enc, const int alphas[256]) {
+static void AssignSegments(VP8Encoder* const enc,
+                           const int alphas[MAX_ALPHA + 1]) {
   const int nb = enc->segment_hdr_.num_segments_;
   int centers[NUM_MB_SEGMENTS];
   int weighted_average = 0;
-  int map[256];
+  int map[MAX_ALPHA + 1];
   int a, n, k;
-  int min_a = 0, max_a = 255, range_a;
+  int min_a = 0, max_a = MAX_ALPHA, range_a;
   // 'int' type is ok for histo, and won't overflow
   int accum[NUM_MB_SEGMENTS], dist_accum[NUM_MB_SEGMENTS];
 
   // bracket the input
-  for (n = 0; n < 256 && alphas[n] == 0; ++n) {}
+  for (n = 0; n <= MAX_ALPHA && alphas[n] == 0; ++n) {}
   min_a = n;
-  for (n = 255; n > min_a && alphas[n] == 0; --n) {}
+  for (n = MAX_ALPHA; n > min_a && alphas[n] == 0; --n) {}
   max_a = n;
   range_a = max_a - min_a;
 
@@ -210,7 +209,7 @@ static void AssignSegments(VP8Encoder* const enc, const int alphas[256]) {
     VP8MBInfo* const mb = &enc->mb_info_[n];
     const int alpha = mb->alpha_;
     mb->segment_ = map[alpha];
-    mb->alpha_ = centers[map[alpha]];     // just for the record.
+    mb->alpha_ = centers[map[alpha]];  // for the record.
   }
 
   if (nb > 1) {
@@ -218,7 +217,6 @@ static void AssignSegments(VP8Encoder* const enc, const int alphas[256]) {
     if (smooth) SmoothSegmentMap(enc);
   }
 
-  SetSegmentProbas(enc);                             // Assign final proba
   SetSegmentAlphas(enc, centers, weighted_average);  // pick some alphas.
 }
 
@@ -227,24 +225,32 @@ static void AssignSegments(VP8Encoder* const enc, const int alphas[256]) {
 // susceptibility and set best modes for this macroblock.
 // Segment assignment is done later.
 
-// Number of modes to inspect for alpha_ evaluation. For high-quality settings,
-// we don't need to test all the possible modes during the analysis phase.
+// Number of modes to inspect for alpha_ evaluation. For high-quality settings
+// (method >= FAST_ANALYSIS_METHOD) we don't need to test all the possible modes
+// during the analysis phase.
+#define FAST_ANALYSIS_METHOD 4  // method above which we do partial analysis
 #define MAX_INTRA16_MODE 2
 #define MAX_INTRA4_MODE  2
 #define MAX_UV_MODE      2
 
 static int MBAnalyzeBestIntra16Mode(VP8EncIterator* const it) {
-  const int max_mode = (it->enc_->method_ >= 3) ? MAX_INTRA16_MODE : 4;
+  const int max_mode =
+      (it->enc_->method_ >= FAST_ANALYSIS_METHOD) ? MAX_INTRA16_MODE
+                                                  : NUM_PRED_MODES;
   int mode;
-  int best_alpha = -1;
+  int best_alpha = DEFAULT_ALPHA;
   int best_mode = 0;
 
   VP8MakeLuma16Preds(it);
   for (mode = 0; mode < max_mode; ++mode) {
-    const int alpha = VP8CollectHistogram(it->yuv_in_ + Y_OFF,
-                                          it->yuv_p_ + VP8I16ModeOffsets[mode],
-                                          0, 16);
-    if (alpha > best_alpha) {
+    VP8Histogram histo = { { 0 } };
+    int alpha;
+
+    VP8CollectHistogram(it->yuv_in_ + Y_OFF,
+                        it->yuv_p_ + VP8I16ModeOffsets[mode],
+                        0, 16, &histo);
+    alpha = GetAlpha(&histo);
+    if (IS_BETTER_ALPHA(alpha, best_alpha)) {
       best_alpha = alpha;
       best_mode = mode;
     }
@@ -256,46 +262,63 @@ static int MBAnalyzeBestIntra16Mode(VP8EncIterator* const it) {
 static int MBAnalyzeBestIntra4Mode(VP8EncIterator* const it,
                                    int best_alpha) {
   uint8_t modes[16];
-  const int max_mode = (it->enc_->method_ >= 3) ? MAX_INTRA4_MODE : NUM_BMODES;
-  int i4_alpha = 0;
+  const int max_mode =
+      (it->enc_->method_ >= FAST_ANALYSIS_METHOD) ? MAX_INTRA4_MODE
+                                                  : NUM_BMODES;
+  int i4_alpha;
+  VP8Histogram total_histo = { { 0 } };
+  int cur_histo = 0;
+
   VP8IteratorStartI4(it);
   do {
     int mode;
-    int best_mode_alpha = -1;
+    int best_mode_alpha = DEFAULT_ALPHA;
+    VP8Histogram histos[2];
     const uint8_t* const src = it->yuv_in_ + Y_OFF + VP8Scan[it->i4_];
 
     VP8MakeIntra4Preds(it);
     for (mode = 0; mode < max_mode; ++mode) {
-      const int alpha = VP8CollectHistogram(src,
-                                            it->yuv_p_ + VP8I4ModeOffsets[mode],
-                                            0, 1);
-      if (alpha > best_mode_alpha) {
+      int alpha;
+
+      memset(&histos[cur_histo], 0, sizeof(histos[cur_histo]));
+      VP8CollectHistogram(src, it->yuv_p_ + VP8I4ModeOffsets[mode],
+                          0, 1, &histos[cur_histo]);
+      alpha = GetAlpha(&histos[cur_histo]);
+      if (IS_BETTER_ALPHA(alpha, best_mode_alpha)) {
         best_mode_alpha = alpha;
         modes[it->i4_] = mode;
+        cur_histo ^= 1;   // keep track of best histo so far.
       }
     }
-    i4_alpha += best_mode_alpha;
+    // accumulate best histogram
+    MergeHistograms(&histos[cur_histo ^ 1], &total_histo);
     // Note: we reuse the original samples for predictors
   } while (VP8IteratorRotateI4(it, it->yuv_in_ + Y_OFF));
 
-  if (i4_alpha > best_alpha) {
+  i4_alpha = GetAlpha(&total_histo);
+  if (IS_BETTER_ALPHA(i4_alpha, best_alpha)) {
     VP8SetIntra4Mode(it, modes);
-    best_alpha = ClipAlpha(i4_alpha);
+    best_alpha = i4_alpha;
   }
   return best_alpha;
 }
 
 static int MBAnalyzeBestUVMode(VP8EncIterator* const it) {
-  int best_alpha = -1;
+  int best_alpha = DEFAULT_ALPHA;
   int best_mode = 0;
-  const int max_mode = (it->enc_->method_ >= 3) ? MAX_UV_MODE : 4;
+  const int max_mode =
+      (it->enc_->method_ >= FAST_ANALYSIS_METHOD) ? MAX_UV_MODE
+                                                  : NUM_PRED_MODES;
   int mode;
   VP8MakeChroma8Preds(it);
   for (mode = 0; mode < max_mode; ++mode) {
-    const int alpha = VP8CollectHistogram(it->yuv_in_ + U_OFF,
-                                          it->yuv_p_ + VP8UVModeOffsets[mode],
-                                          16, 16 + 4 + 4);
-    if (alpha > best_alpha) {
+    VP8Histogram histo = { { 0 } };
+    int alpha;
+    VP8CollectHistogram(it->yuv_in_ + U_OFF,
+                        it->yuv_p_ + VP8UVModeOffsets[mode],
+                        16, 16 + 4 + 4, &histo);
+    alpha = GetAlpha(&histo);
+    if (IS_BETTER_ALPHA(alpha, best_alpha)) {
       best_alpha = alpha;
       best_mode = mode;
     }
@@ -305,7 +328,8 @@ static int MBAnalyzeBestUVMode(VP8EncIterator* const it) {
 }
 
 static void MBAnalyze(VP8EncIterator* const it,
-                      int alphas[256], int* const uv_alpha) {
+                      int alphas[MAX_ALPHA + 1],
+                      int* const alpha, int* const uv_alpha) {
   const VP8Encoder* const enc = it->enc_;
   int best_alpha, best_uv_alpha;
 
@@ -314,7 +338,7 @@ static void MBAnalyze(VP8EncIterator* const it,
   VP8SetSegment(it, 0);      // default segment, spec-wise.
 
   best_alpha = MBAnalyzeBestIntra16Mode(it);
-  if (enc->method_ != 3) {
+  if (enc->method_ >= 5) {
     // We go and make a fast decision for intra4/intra16.
     // It's usually not a good and definitive pick, but helps seeding the stats
     // about level bit-cost.
@@ -324,10 +348,22 @@ static void MBAnalyze(VP8EncIterator* const it,
   best_uv_alpha = MBAnalyzeBestUVMode(it);
 
   // Final susceptibility mix
-  best_alpha = (best_alpha + best_uv_alpha + 1) / 2;
+  best_alpha = (3 * best_alpha + best_uv_alpha + 2) >> 2;
+  best_alpha = FinalAlphaValue(best_alpha);
   alphas[best_alpha]++;
+  it->mb_->alpha_ = best_alpha;   // for later remapping.
+
+  // Accumulate for later complexity analysis.
+  *alpha += best_alpha;   // mixed susceptibility (not just luma)
   *uv_alpha += best_uv_alpha;
-  it->mb_->alpha_ = best_alpha;   // Informative only.
+}
+
+static void DefaultMBInfo(VP8MBInfo* const mb) {
+  mb->type_ = 1;     // I16x16
+  mb->uv_mode_ = 0;
+  mb->skip_ = 0;     // not skipped
+  mb->segment_ = 0;  // default segment
+  mb->alpha_ = 0;
 }
 
 //------------------------------------------------------------------------------
@@ -340,22 +376,43 @@ static void MBAnalyze(VP8EncIterator* const it,
 // and decide intra4/intra16, but that's usually almost always a bad choice at
 // this stage.
 
+static void ResetAllMBInfo(VP8Encoder* const enc) {
+  int n;
+  for (n = 0; n < enc->mb_w_ * enc->mb_h_; ++n) {
+    DefaultMBInfo(&enc->mb_info_[n]);
+  }
+  // Default susceptibilities.
+  enc->dqm_[0].alpha_ = 0;
+  enc->dqm_[0].beta_ = 0;
+  // Note: we can't compute this alpha_ / uv_alpha_.
+  WebPReportProgress(enc->pic_, enc->percent_ + 20, &enc->percent_);
+}
+
 int VP8EncAnalyze(VP8Encoder* const enc) {
   int ok = 1;
-  int alphas[256] = { 0 };
-  VP8EncIterator it;
-
-  VP8IteratorInit(enc, &it);
+  const int do_segments =
+      enc->config_->emulate_jpeg_size ||   // We need the complexity evaluation.
+      (enc->segment_hdr_.num_segments_ > 1) ||
+      (enc->method_ == 0);  // for method 0, we need preds_[] to be filled.
+  enc->alpha_ = 0;
   enc->uv_alpha_ = 0;
-  do {
-    VP8IteratorImport(&it);
-    MBAnalyze(&it, alphas, &enc->uv_alpha_);
-    ok = VP8IteratorProgress(&it, 20);
-    // Let's pretend we have perfect lossless reconstruction.
-  } while (ok && VP8IteratorNext(&it, it.yuv_in_));
-  enc->uv_alpha_ /= enc->mb_w_ * enc->mb_h_;
-  if (ok) AssignSegments(enc, alphas);
+  if (do_segments) {
+    int alphas[MAX_ALPHA + 1] = { 0 };
+    VP8EncIterator it;
 
+    VP8IteratorInit(enc, &it);
+    do {
+      VP8IteratorImport(&it);
+      MBAnalyze(&it, alphas, &enc->alpha_, &enc->uv_alpha_);
+      ok = VP8IteratorProgress(&it, 20);
+      // Let's pretend we have perfect lossless reconstruction.
+    } while (ok && VP8IteratorNext(&it, it.yuv_in_));
+    enc->alpha_ /= enc->mb_w_ * enc->mb_h_;
+    enc->uv_alpha_ /= enc->mb_w_ * enc->mb_h_;
+    if (ok) AssignSegments(enc, alphas);
+  } else {   // Use only one default segment.
+    ResetAllMBInfo(enc);
+  }
   return ok;
 }
 
diff --git a/3rdparty/libwebp/enc/backward_references.c b/3rdparty/libwebp/enc/backward_references.c
index fd30f307df..db4f430df5 100644
--- a/3rdparty/libwebp/enc/backward_references.c
+++ b/3rdparty/libwebp/enc/backward_references.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
@@ -141,74 +143,86 @@ static void HashChainInsert(HashChain* const p,
   p->hash_to_first_index_[hash_code] = pos;
 }
 
+static void GetParamsForHashChainFindCopy(int quality, int xsize,
+                                          int cache_bits, int* window_size,
+                                          int* iter_pos, int* iter_limit) {
+  const int iter_mult = (quality < 27) ? 1 : 1 + ((quality - 27) >> 4);
+  const int iter_neg = -iter_mult * (quality >> 1);
+  // Limit the backward-ref window size for lower qualities.
+  const int max_window_size = (quality > 50) ? WINDOW_SIZE
+                            : (quality > 25) ? (xsize << 8)
+                            : (xsize << 4);
+  assert(xsize > 0);
+  *window_size = (max_window_size > WINDOW_SIZE) ? WINDOW_SIZE
+               : max_window_size;
+  *iter_pos = 8 + (quality >> 3);
+  // For lower entropy images, the rigourous search loop in HashChainFindCopy
+  // can be relaxed.
+  *iter_limit = (cache_bits > 0) ? iter_neg : iter_neg / 2;
+}
+
 static int HashChainFindCopy(const HashChain* const p,
-                             int quality, int index, int xsize,
+                             int base_position, int xsize_signed,
                              const uint32_t* const argb, int maxlen,
+                             int window_size, int iter_pos, int iter_limit,
                              int* const distance_ptr,
                              int* const length_ptr) {
-  const uint64_t hash_code = GetPixPairHash64(&argb[index]);
-  int prev_length = 0;
-  int64_t best_val = 0;
-  int best_length = 0;
-  int best_distance = 0;
-  const uint32_t* const argb_start = argb + index;
-  const int iter_min_mult = (quality < 50) ? 2 : (quality < 75) ? 4 : 8;
-  const int iter_min = -quality * iter_min_mult;
-  int iter_cnt = 10 + (quality >> 1);
-  const int min_pos = (index > WINDOW_SIZE) ? index - WINDOW_SIZE : 0;
+  const uint32_t* const argb_start = argb + base_position;
+  uint64_t best_val = 0;
+  uint32_t best_length = 1;
+  uint32_t best_distance = 0;
+  const uint32_t xsize = (uint32_t)xsize_signed;
+  const int min_pos =
+      (base_position > window_size) ? base_position - window_size : 0;
   int pos;
-
   assert(xsize > 0);
-  for (pos = p->hash_to_first_index_[hash_code];
+  for (pos = p->hash_to_first_index_[GetPixPairHash64(argb_start)];
        pos >= min_pos;
        pos = p->chain_[pos]) {
-    int64_t val;
-    int curr_length;
-    if (iter_cnt < 0) {
-      if (iter_cnt < iter_min || best_val >= 0xff0000) {
+    uint64_t val;
+    uint32_t curr_length;
+    uint32_t distance;
+    if (iter_pos < 0) {
+      if (iter_pos < iter_limit || best_val >= 0xff0000) {
         break;
       }
     }
-    --iter_cnt;
-    if (best_length != 0 &&
-        argb[pos + best_length - 1] != argb_start[best_length - 1]) {
+    --iter_pos;
+    if (argb[pos + best_length - 1] != argb_start[best_length - 1]) {
       continue;
     }
     curr_length = FindMatchLength(argb + pos, argb_start, maxlen);
-    if (curr_length < prev_length) {
+    if (curr_length < best_length) {
       continue;
     }
-    val = 65536 * curr_length;
+    distance = (uint32_t)(base_position - pos);
+    val = curr_length << 16;
     // Favoring 2d locality here gives savings for certain images.
-    if (index - pos < 9 * xsize) {
-      const int y = (index - pos) / xsize;
-      int x = (index - pos) % xsize;
-      if (x > xsize / 2) {
+    if (distance < 9 * xsize) {
+      const uint32_t y = distance / xsize;
+      uint32_t x = distance % xsize;
+      if (x > (xsize >> 1)) {
         x = xsize - x;
       }
-      if (x <= 7 && x >= -8) {
+      if (x <= 7) {
+        val += 9 * 9 + 9 * 9;
         val -= y * y + x * x;
-      } else {
-        val -= 9 * 9 + 9 * 9;
       }
-    } else {
-      val -= 9 * 9 + 9 * 9;
     }
     if (best_val < val) {
-      prev_length = curr_length;
       best_val = val;
       best_length = curr_length;
-      best_distance = index - pos;
+      best_distance = distance;
       if (curr_length >= MAX_LENGTH) {
         break;
       }
-      if ((best_distance == 1 || best_distance == xsize) &&
+      if ((best_distance == 1 || distance == xsize) &&
           best_length >= 128) {
         break;
       }
     }
   }
-  *distance_ptr = best_distance;
+  *distance_ptr = (int)best_distance;
   *length_ptr = best_length;
   return (best_length >= MIN_LENGTH);
 }
@@ -257,6 +271,9 @@ static int BackwardReferencesHashChain(int xsize, int ysize,
   const int pix_count = xsize * ysize;
   HashChain* const hash_chain = (HashChain*)malloc(sizeof(*hash_chain));
   VP8LColorCache hashers;
+  int window_size = WINDOW_SIZE;
+  int iter_pos = 1;
+  int iter_limit = -1;
 
   if (hash_chain == NULL) return 0;
   if (use_color_cache) {
@@ -267,6 +284,8 @@ static int BackwardReferencesHashChain(int xsize, int ysize,
   if (!HashChainInit(hash_chain, pix_count)) goto Error;
 
   refs->size = 0;
+  GetParamsForHashChainFindCopy(quality, xsize, cache_bits,
+                                &window_size, &iter_pos, &iter_limit);
   for (i = 0; i < pix_count; ) {
     // Alternative#1: Code the pixels starting at 'i' using backward reference.
     int offset = 0;
@@ -276,7 +295,8 @@ static int BackwardReferencesHashChain(int xsize, int ysize,
       if (maxlen > MAX_LENGTH) {
         maxlen = MAX_LENGTH;
       }
-      HashChainFindCopy(hash_chain, quality, i, xsize, argb, maxlen,
+      HashChainFindCopy(hash_chain, i, xsize, argb, maxlen,
+                        window_size, iter_pos, iter_limit,
                         &offset, &len);
     }
     if (len >= MIN_LENGTH) {
@@ -291,8 +311,9 @@ static int BackwardReferencesHashChain(int xsize, int ysize,
         if (maxlen > MAX_LENGTH) {
           maxlen = MAX_LENGTH;
         }
-        HashChainFindCopy(hash_chain, quality,
-                          i + 1, xsize, argb, maxlen, &offset2, &len2);
+        HashChainFindCopy(hash_chain, i + 1, xsize, argb, maxlen,
+                          window_size, iter_pos, iter_limit,
+                          &offset2, &len2);
         if (len2 > len + 1) {
           const uint32_t pixel = argb[i];
           // Alternative#2 is a better match. So push pixel at 'i' as literal.
@@ -362,7 +383,8 @@ typedef struct {
 
 static int BackwardReferencesTraceBackwards(
     int xsize, int ysize, int recursive_cost_model,
-    const uint32_t* const argb, int cache_bits, VP8LBackwardRefs* const refs);
+    const uint32_t* const argb, int quality, int cache_bits,
+    VP8LBackwardRefs* const refs);
 
 static void ConvertPopulationCountTableToBitEstimates(
     int num_symbols, const int population_counts[], double output[]) {
@@ -387,17 +409,16 @@ static void ConvertPopulationCountTableToBitEstimates(
 
 static int CostModelBuild(CostModel* const m, int xsize, int ysize,
                           int recursion_level, const uint32_t* const argb,
-                          int cache_bits) {
+                          int quality, int cache_bits) {
   int ok = 0;
   VP8LHistogram histo;
   VP8LBackwardRefs refs;
-  const int quality = 100;
 
   if (!VP8LBackwardRefsAlloc(&refs, xsize * ysize)) goto Error;
 
   if (recursion_level > 0) {
     if (!BackwardReferencesTraceBackwards(xsize, ysize, recursion_level - 1,
-                                          argb, cache_bits, &refs)) {
+                                          argb, quality, cache_bits, &refs)) {
       goto Error;
     }
   } else {
@@ -452,11 +473,10 @@ static WEBP_INLINE double GetDistanceCost(const CostModel* const m,
 
 static int BackwardReferencesHashChainDistanceOnly(
     int xsize, int ysize, int recursive_cost_model, const uint32_t* const argb,
-    int cache_bits, uint32_t* const dist_array) {
+    int quality, int cache_bits, uint32_t* const dist_array) {
   int i;
   int ok = 0;
   int cc_init = 0;
-  const int quality = 100;
   const int pix_count = xsize * ysize;
   const int use_color_cache = (cache_bits > 0);
   float* const cost =
@@ -466,6 +486,10 @@ static int BackwardReferencesHashChainDistanceOnly(
   VP8LColorCache hashers;
   const double mul0 = (recursive_cost_model != 0) ? 1.0 : 0.68;
   const double mul1 = (recursive_cost_model != 0) ? 1.0 : 0.82;
+  const int min_distance_code = 2;  // TODO(vikasa): tune as function of quality
+  int window_size = WINDOW_SIZE;
+  int iter_pos = 1;
+  int iter_limit = -1;
 
   if (cost == NULL || cost_model == NULL || hash_chain == NULL) goto Error;
 
@@ -477,7 +501,7 @@ static int BackwardReferencesHashChainDistanceOnly(
   }
 
   if (!CostModelBuild(cost_model, xsize, ysize, recursive_cost_model, argb,
-                      cache_bits)) {
+                      quality, cache_bits)) {
     goto Error;
   }
 
@@ -486,6 +510,8 @@ static int BackwardReferencesHashChainDistanceOnly(
   // We loop one pixel at a time, but store all currently best points to
   // non-processed locations from this point.
   dist_array[0] = 0;
+  GetParamsForHashChainFindCopy(quality, xsize, cache_bits,
+                                &window_size, &iter_pos, &iter_limit);
   for (i = 0; i < pix_count; ++i) {
     double prev_cost = 0.0;
     int shortmax;
@@ -500,7 +526,8 @@ static int BackwardReferencesHashChainDistanceOnly(
         if (maxlen > pix_count - i) {
           maxlen = pix_count - i;
         }
-        HashChainFindCopy(hash_chain, quality, i, xsize, argb, maxlen,
+        HashChainFindCopy(hash_chain, i, xsize, argb, maxlen,
+                          window_size, iter_pos, iter_limit,
                           &offset, &len);
       }
       if (len >= MIN_LENGTH) {
@@ -517,7 +544,7 @@ static int BackwardReferencesHashChainDistanceOnly(
         }
         // This if is for speedup only. It roughly doubles the speed, and
         // makes compression worse by .1 %.
-        if (len >= 128 && code < 2) {
+        if (len >= 128 && code <= min_distance_code) {
           // Long copy for short distances, let's skip the middle
           // lookups for better copies.
           // 1) insert the hashes.
@@ -528,10 +555,10 @@ static int BackwardReferencesHashChainDistanceOnly(
           }
           // 2) Add to the hash_chain (but cannot add the last pixel)
           {
-            const int last = (len < pix_count - 1 - i) ? len
-                                                       : pix_count - 1 - i;
-            for (k = 0; k < last; ++k) {
-              HashChainInsert(hash_chain, &argb[i + k], i + k);
+            const int last = (len + i < pix_count - 1) ? len + i
+                                                       : pix_count - 1;
+            for (k = i; k < last; ++k) {
+              HashChainInsert(hash_chain, &argb[k], k);
             }
           }
           // 3) jump.
@@ -571,40 +598,30 @@ Error:
   return ok;
 }
 
-static int TraceBackwards(const uint32_t* const dist_array,
-                          int dist_array_size,
-                          uint32_t** const chosen_path,
-                          int* const chosen_path_size) {
-  int i;
-  // Count how many.
-  int count = 0;
-  for (i = dist_array_size - 1; i >= 0; ) {
-    int k = dist_array[i];
-    assert(k >= 1);
-    ++count;
-    i -= k;
+// We pack the path at the end of *dist_array and return
+// a pointer to this part of the array. Example:
+// dist_array = [1x2xx3x2] => packed [1x2x1232], chosen_path = [1232]
+static void TraceBackwards(uint32_t* const dist_array,
+                           int dist_array_size,
+                           uint32_t** const chosen_path,
+                           int* const chosen_path_size) {
+  uint32_t* path = dist_array + dist_array_size;
+  uint32_t* cur = dist_array + dist_array_size - 1;
+  while (cur >= dist_array) {
+    const int k = *cur;
+    --path;
+    *path = k;
+    cur -= k;
   }
-  // Allocate.
-  *chosen_path_size = count;
-  *chosen_path =
-      (uint32_t*)WebPSafeMalloc((uint64_t)count, sizeof(**chosen_path));
-  if (*chosen_path == NULL) return 0;
-
-  // Write in reverse order.
-  for (i = dist_array_size - 1; i >= 0; ) {
-    int k = dist_array[i];
-    assert(k >= 1);
-    (*chosen_path)[--count] = k;
-    i -= k;
-  }
-  return 1;
+  *chosen_path = path;
+  *chosen_path_size = (int)(dist_array + dist_array_size - path);
 }
 
 static int BackwardReferencesHashChainFollowChosenPath(
-    int xsize, int ysize, const uint32_t* const argb, int cache_bits,
+    int xsize, int ysize, const uint32_t* const argb,
+    int quality, int cache_bits,
     const uint32_t* const chosen_path, int chosen_path_size,
     VP8LBackwardRefs* const refs) {
-  const int quality = 100;
   const int pix_count = xsize * ysize;
   const int use_color_cache = (cache_bits > 0);
   int size = 0;
@@ -613,6 +630,9 @@ static int BackwardReferencesHashChainFollowChosenPath(
   int ix;
   int ok = 0;
   int cc_init = 0;
+  int window_size = WINDOW_SIZE;
+  int iter_pos = 1;
+  int iter_limit = -1;
   HashChain* hash_chain = (HashChain*)malloc(sizeof(*hash_chain));
   VP8LColorCache hashers;
 
@@ -625,13 +645,16 @@ static int BackwardReferencesHashChainFollowChosenPath(
   }
 
   refs->size = 0;
+  GetParamsForHashChainFindCopy(quality, xsize, cache_bits,
+                                &window_size, &iter_pos, &iter_limit);
   for (ix = 0; ix < chosen_path_size; ++ix, ++size) {
     int offset = 0;
     int len = 0;
     int maxlen = chosen_path[ix];
     if (maxlen != 1) {
-      HashChainFindCopy(hash_chain, quality,
-                        i, xsize, argb, maxlen, &offset, &len);
+      HashChainFindCopy(hash_chain, i, xsize, argb, maxlen,
+                        window_size, iter_pos, iter_limit,
+                        &offset, &len);
       assert(len == maxlen);
       refs->refs[size] = PixOrCopyCreateCopy(offset, len);
       if (use_color_cache) {
@@ -674,7 +697,7 @@ Error:
 static int BackwardReferencesTraceBackwards(int xsize, int ysize,
                                             int recursive_cost_model,
                                             const uint32_t* const argb,
-                                            int cache_bits,
+                                            int quality, int cache_bits,
                                             VP8LBackwardRefs* const refs) {
   int ok = 0;
   const int dist_array_size = xsize * ysize;
@@ -686,22 +709,18 @@ static int BackwardReferencesTraceBackwards(int xsize, int ysize,
   if (dist_array == NULL) goto Error;
 
   if (!BackwardReferencesHashChainDistanceOnly(
-      xsize, ysize, recursive_cost_model, argb, cache_bits, dist_array)) {
+      xsize, ysize, recursive_cost_model, argb, quality, cache_bits,
+      dist_array)) {
     goto Error;
   }
-  if (!TraceBackwards(dist_array, dist_array_size,
-                      &chosen_path, &chosen_path_size)) {
-    goto Error;
-  }
-  free(dist_array);   // no need to retain this memory any longer
-  dist_array = NULL;
+  TraceBackwards(dist_array, dist_array_size, &chosen_path, &chosen_path_size);
   if (!BackwardReferencesHashChainFollowChosenPath(
-      xsize, ysize, argb, cache_bits, chosen_path, chosen_path_size, refs)) {
+      xsize, ysize, argb, quality, cache_bits, chosen_path, chosen_path_size,
+      refs)) {
     goto Error;
   }
   ok = 1;
  Error:
-  free(chosen_path);
   free(dist_array);
   return ok;
 }
@@ -761,18 +780,20 @@ int VP8LGetBackwardReferences(int width, int height,
 
   // Choose appropriate backward reference.
   if (lz77_is_useful) {
-    // TraceBackwards is costly. Run it for higher qualities.
-    const int try_lz77_trace_backwards = (quality >= 75);
+    // TraceBackwards is costly. Don't execute it at lower quality (q <= 10).
+    const int try_lz77_trace_backwards = (quality > 10);
     *best = refs_lz77;   // default guess: lz77 is better
     VP8LClearBackwardRefs(&refs_rle);
     if (try_lz77_trace_backwards) {
-      const int recursion_level = (num_pix < 320 * 200) ? 1 : 0;
+      // Set recursion level for large images using a color cache.
+      const int recursion_level =
+          (num_pix < 320 * 200) && (cache_bits > 0) ? 1 : 0;
       VP8LBackwardRefs refs_trace;
       if (!VP8LBackwardRefsAlloc(&refs_trace, num_pix)) {
         goto End;
       }
-      if (BackwardReferencesTraceBackwards(
-          width, height, recursion_level, argb, cache_bits, &refs_trace)) {
+      if (BackwardReferencesTraceBackwards(width, height, recursion_level, argb,
+                                           quality, cache_bits, &refs_trace)) {
         VP8LClearBackwardRefs(&refs_lz77);
         *best = refs_trace;
       }
diff --git a/3rdparty/libwebp/enc/backward_references.h b/3rdparty/libwebp/enc/backward_references.h
index 91c03361ed..b0d18135f7 100644
--- a/3rdparty/libwebp/enc/backward_references.h
+++ b/3rdparty/libwebp/enc/backward_references.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
@@ -35,7 +37,8 @@ extern "C" {
 #if defined(__GNUC__) && \
     ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
-  return n == 0 ? -1 : 31 ^ __builtin_clz(n);
+  assert(n != 0);
+  return 31 ^ __builtin_clz(n);
 }
 #elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_IX86))
 #include <intrin.h>
@@ -43,15 +46,18 @@ static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
 
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
   unsigned long first_set_bit;
-  return _BitScanReverse(&first_set_bit, n) ? first_set_bit : -1;
+  assert(n != 0);
+  _BitScanReverse(&first_set_bit, n);
+  return first_set_bit;
 }
 #else
+// Returns (int)floor(log2(n)). n must be > 0.
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
   int log = 0;
   uint32_t value = n;
   int i;
 
-  if (value == 0) return -1;
+  assert(n != 0);
   for (i = 4; i >= 0; --i) {
     const int shift = (1 << i);
     const uint32_t x = value >> shift;
@@ -65,11 +71,11 @@ static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
 #endif
 
 static WEBP_INLINE int VP8LBitsLog2Ceiling(uint32_t n) {
-  const int floor = BitsLog2Floor(n);
+  const int log_floor = BitsLog2Floor(n);
   if (n == (n & ~(n - 1)))  // zero or a power of two.
-    return floor;
+    return log_floor;
   else
-    return floor + 1;
+    return log_floor + 1;
 }
 
 // Splitting of distance and length codes into prefixes and
@@ -78,16 +84,17 @@ static WEBP_INLINE int VP8LBitsLog2Ceiling(uint32_t n) {
 static WEBP_INLINE void PrefixEncode(int distance, int* const code,
                                      int* const extra_bits_count,
                                      int* const extra_bits_value) {
-  // Collect the two most significant bits where the highest bit is 1.
-  const int highest_bit = BitsLog2Floor(--distance);
-  // & 0x3f is to make behavior well defined when highest_bit
-  // does not exist or is the least significant bit.
-  const int second_highest_bit =
-      (distance >> ((highest_bit - 1) & 0x3f)) & 1;
-  *extra_bits_count = (highest_bit > 0) ? (highest_bit - 1) : 0;
-  *extra_bits_value = distance & ((1 << *extra_bits_count) - 1);
-  *code = (highest_bit > 0) ? (2 * highest_bit + second_highest_bit)
-                            : (highest_bit == 0) ? 1 : 0;
+  if (distance > 2) {  // Collect the two most significant bits.
+    const int highest_bit = BitsLog2Floor(--distance);
+    const int second_highest_bit = (distance >> (highest_bit - 1)) & 1;
+    *extra_bits_count = highest_bit - 1;
+    *extra_bits_value = distance & ((1 << *extra_bits_count) - 1);
+    *code = 2 * highest_bit + second_highest_bit;
+  } else {
+    *extra_bits_count = 0;
+    *extra_bits_value = 0;
+    *code = (distance == 2) ? 1 : 0;
+  }
 }
 
 // -----------------------------------------------------------------------------
diff --git a/3rdparty/libwebp/enc/config.c b/3rdparty/libwebp/enc/config.c
index 1a26113554..acf96b0f30 100644
--- a/3rdparty/libwebp/enc/config.c
+++ b/3rdparty/libwebp/enc/config.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Coding tools configuration
@@ -31,9 +33,9 @@ int WebPConfigInitInternal(WebPConfig* config,
   config->target_PSNR = 0.;
   config->method = 4;
   config->sns_strength = 50;
-  config->filter_strength = 20;   // default: light filtering
+  config->filter_strength = 60;   // rather high filtering, helps w/ gradients.
   config->filter_sharpness = 0;
-  config->filter_type = 0;        // default: simple
+  config->filter_type = 1;        // default: strong (so U/V is filtered too)
   config->partitions = 0;
   config->segments = 4;
   config->pass = 1;
@@ -46,6 +48,9 @@ int WebPConfigInitInternal(WebPConfig* config,
   config->alpha_quality = 100;
   config->lossless = 0;
   config->image_hint = WEBP_HINT_DEFAULT;
+  config->emulate_jpeg_size = 0;
+  config->thread_level = 0;
+  config->low_memory = 0;
 
   // TODO(skal): tune.
   switch (preset) {
@@ -122,6 +127,12 @@ int WebPValidateConfig(const WebPConfig* config) {
     return 0;
   if (config->image_hint >= WEBP_HINT_LAST)
     return 0;
+  if (config->emulate_jpeg_size < 0 || config->emulate_jpeg_size > 1)
+    return 0;
+  if (config->thread_level < 0 || config->thread_level > 1)
+    return 0;
+  if (config->low_memory < 0 || config->low_memory > 1)
+    return 0;
   return 1;
 }
 
diff --git a/3rdparty/libwebp/enc/cost.c b/3rdparty/libwebp/enc/cost.c
index 92e0cc713c..d4916d7cd2 100644
--- a/3rdparty/libwebp/enc/cost.c
+++ b/3rdparty/libwebp/enc/cost.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Cost tables for level and modes
@@ -75,7 +77,7 @@ const uint16_t VP8LevelCodes[MAX_VARIABLE_LEVEL][2] = {
 
 // fixed costs for coding levels, deduce from the coding tree.
 // This is only the part that doesn't depend on the probability state.
-const uint16_t VP8LevelFixedCosts[2048] = {
+const uint16_t VP8LevelFixedCosts[MAX_LEVEL + 1] = {
      0,  256,  256,  256,  256,  432,  618,  630,
    731,  640,  640,  828,  901,  948, 1021, 1101,
   1174, 1221, 1294, 1042, 1085, 1115, 1158, 1202,
@@ -359,7 +361,7 @@ void VP8CalculateLevelCosts(VP8Proba* const proba) {
 
   for (ctype = 0; ctype < NUM_TYPES; ++ctype) {
     for (band = 0; band < NUM_BANDS; ++band) {
-      for(ctx = 0; ctx < NUM_CTX; ++ctx) {
+      for (ctx = 0; ctx < NUM_CTX; ++ctx) {
         const uint8_t* const p = proba->coeffs_[ctype][band][ctx];
         uint16_t* const table = proba->level_cost_[ctype][band][ctx];
         const int cost_base = VP8BitCost(1, p[1]);
diff --git a/3rdparty/libwebp/enc/cost.h b/3rdparty/libwebp/enc/cost.h
index 09b75b699d..7d7c2c79a2 100644
--- a/3rdparty/libwebp/enc/cost.h
+++ b/3rdparty/libwebp/enc/cost.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Cost tables for level and modes.
@@ -18,7 +20,8 @@
 extern "C" {
 #endif
 
-extern const uint16_t VP8LevelFixedCosts[2048];   // approximate cost per level
+// approximate cost per level:
+extern const uint16_t VP8LevelFixedCosts[MAX_LEVEL + 1];
 extern const uint16_t VP8EntropyCost[256];        // 8bit fixed-point log(p)
 
 // Cost of coding one event with probability 'proba'.
diff --git a/3rdparty/libwebp/enc/filter.c b/3rdparty/libwebp/enc/filter.c
index 7fb78a3949..aae2723d16 100644
--- a/3rdparty/libwebp/enc/filter.c
+++ b/3rdparty/libwebp/enc/filter.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Selecting filter level
diff --git a/3rdparty/libwebp/enc/frame.c b/3rdparty/libwebp/enc/frame.c
index bdd360069b..c56abed75e 100644
--- a/3rdparty/libwebp/enc/frame.c
+++ b/3rdparty/libwebp/enc/frame.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //   frame coding and analysis
@@ -45,10 +47,10 @@ const uint8_t VP8EncBands[16 + 1] = {
   0  // sentinel
 };
 
-static const uint8_t kCat3[] = { 173, 148, 140 };
-static const uint8_t kCat4[] = { 176, 155, 140, 135 };
-static const uint8_t kCat5[] = { 180, 157, 141, 134, 130 };
-static const uint8_t kCat6[] =
+const uint8_t VP8Cat3[] = { 173, 148, 140 };
+const uint8_t VP8Cat4[] = { 176, 155, 140, 135 };
+const uint8_t VP8Cat5[] = { 180, 157, 141, 134, 130 };
+const uint8_t VP8Cat6[] =
     { 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129 };
 
 //------------------------------------------------------------------------------
@@ -113,14 +115,15 @@ static int Record(int bit, proba_t* const stats) {
 // Note: no need to record the fixed probas.
 static int RecordCoeffs(int ctx, const VP8Residual* const res) {
   int n = res->first;
-  proba_t* s = res->stats[VP8EncBands[n]][ctx];
+  // should be stats[VP8EncBands[n]], but it's equivalent for n=0 or 1
+  proba_t* s = res->stats[n][ctx];
   if (res->last  < 0) {
     Record(0, s + 0);
     return 0;
   }
   while (n <= res->last) {
     int v;
-    Record(1, s + 0);
+    Record(1, s + 0);  // order of record doesn't matter
     while ((v = res->coeffs[n++]) == 0) {
       Record(0, s + 1);
       s = res->stats[VP8EncBands[n]][0];
@@ -174,8 +177,7 @@ static int BranchCost(int nb, int total, int proba) {
   return nb * VP8BitCost(1, proba) + (total - nb) * VP8BitCost(0, proba);
 }
 
-static int FinalizeTokenProbas(VP8Encoder* const enc) {
-  VP8Proba* const proba = &enc->proba_;
+static int FinalizeTokenProbas(VP8Proba* const proba) {
   int has_changed = 0;
   int size = 0;
   int t, b, c, p;
@@ -211,6 +213,47 @@ static int FinalizeTokenProbas(VP8Encoder* const enc) {
   return size;
 }
 
+//------------------------------------------------------------------------------
+// Finalize Segment probability based on the coding tree
+
+static int GetProba(int a, int b) {
+  const int total = a + b;
+  return (total == 0) ? 255     // that's the default probability.
+                      : (255 * a + total / 2) / total;  // rounded proba
+}
+
+static void SetSegmentProbas(VP8Encoder* const enc) {
+  int p[NUM_MB_SEGMENTS] = { 0 };
+  int n;
+
+  for (n = 0; n < enc->mb_w_ * enc->mb_h_; ++n) {
+    const VP8MBInfo* const mb = &enc->mb_info_[n];
+    p[mb->segment_]++;
+  }
+  if (enc->pic_->stats != NULL) {
+    for (n = 0; n < NUM_MB_SEGMENTS; ++n) {
+      enc->pic_->stats->segment_size[n] = p[n];
+    }
+  }
+  if (enc->segment_hdr_.num_segments_ > 1) {
+    uint8_t* const probas = enc->proba_.segments_;
+    probas[0] = GetProba(p[0] + p[1], p[2] + p[3]);
+    probas[1] = GetProba(p[0], p[1]);
+    probas[2] = GetProba(p[2], p[3]);
+
+    enc->segment_hdr_.update_map_ =
+        (probas[0] != 255) || (probas[1] != 255) || (probas[2] != 255);
+    enc->segment_hdr_.size_ =
+        p[0] * (VP8BitCost(0, probas[0]) + VP8BitCost(0, probas[1])) +
+        p[1] * (VP8BitCost(0, probas[0]) + VP8BitCost(1, probas[1])) +
+        p[2] * (VP8BitCost(1, probas[0]) + VP8BitCost(0, probas[2])) +
+        p[3] * (VP8BitCost(1, probas[0]) + VP8BitCost(1, probas[2]));
+  } else {
+    enc->segment_hdr_.update_map_ = 0;
+    enc->segment_hdr_.size_ = 0;
+  }
+}
+
 //------------------------------------------------------------------------------
 // helper functions for residuals struct VP8Residual.
 
@@ -239,18 +282,19 @@ static void SetResidualCoeffs(const int16_t* const coeffs,
 //------------------------------------------------------------------------------
 // Mode costs
 
-static int GetResidualCost(int ctx, const VP8Residual* const res) {
+static int GetResidualCost(int ctx0, const VP8Residual* const res) {
   int n = res->first;
-  int p0 = res->prob[VP8EncBands[n]][ctx][0];
-  const uint16_t* t = res->cost[VP8EncBands[n]][ctx];
+  // should be prob[VP8EncBands[n]], but it's equivalent for n=0 or 1
+  int p0 = res->prob[n][ctx0][0];
+  const uint16_t* t = res->cost[n][ctx0];
   int cost;
 
   if (res->last < 0) {
     return VP8BitCost(0, p0);
   }
   cost = 0;
-  while (n <= res->last) {
-    const int v = res->coeffs[n];
+  while (n < res->last) {
+    int v = res->coeffs[n];
     const int b = VP8EncBands[n + 1];
     ++n;
     if (v == 0) {
@@ -259,19 +303,28 @@ static int GetResidualCost(int ctx, const VP8Residual* const res) {
       t = res->cost[b][0];
       continue;
     }
+    v = abs(v);
     cost += VP8BitCost(1, p0);
-    if (2u >= (unsigned int)(v + 1)) {   // v = -1 or 1
-      // short-case for "VP8LevelCost(t, 1)" (256 is VP8LevelFixedCosts[1]):
-      cost += 256 + t[1];
-      p0 = res->prob[b][1][0];
-      t = res->cost[b][1];
-    } else {
-      cost += VP8LevelCost(t, abs(v));
-      p0 = res->prob[b][2][0];
-      t = res->cost[b][2];
+    cost += VP8LevelCost(t, v);
+    {
+      const int ctx = (v == 1) ? 1 : 2;
+      p0 = res->prob[b][ctx][0];
+      t = res->cost[b][ctx];
+    }
+  }
+  // Last coefficient is always non-zero
+  {
+    const int v = abs(res->coeffs[n]);
+    assert(v != 0);
+    cost += VP8BitCost(1, p0);
+    cost += VP8LevelCost(t, v);
+    if (n < 15) {
+      const int b = VP8EncBands[n + 1];
+      const int ctx = (v == 1) ? 1 : 2;
+      const int last_p0 = res->prob[b][ctx][0];
+      cost += VP8BitCost(0, last_p0);
     }
   }
-  if (n < 16) cost += VP8BitCost(0, p0);
   return cost;
 }
 
@@ -342,7 +395,8 @@ int VP8GetCostUV(VP8EncIterator* const it, const VP8ModeScore* const rd) {
 
 static int PutCoeffs(VP8BitWriter* const bw, int ctx, const VP8Residual* res) {
   int n = res->first;
-  const uint8_t* p = res->prob[VP8EncBands[n]][ctx];
+  // should be prob[VP8EncBands[n]], but it's equivalent for n=0 or 1
+  const uint8_t* p = res->prob[n][ctx];
   if (!VP8PutBit(bw, res->last >= 0, p[0])) {
     return 0;
   }
@@ -371,30 +425,30 @@ static int PutCoeffs(VP8BitWriter* const bw, int ctx, const VP8Residual* res) {
       } else {
         int mask;
         const uint8_t* tab;
-        if (v < 3 + (8 << 1)) {          // kCat3  (3b)
+        if (v < 3 + (8 << 1)) {          // VP8Cat3  (3b)
           VP8PutBit(bw, 0, p[8]);
           VP8PutBit(bw, 0, p[9]);
           v -= 3 + (8 << 0);
           mask = 1 << 2;
-          tab = kCat3;
-        } else if (v < 3 + (8 << 2)) {   // kCat4  (4b)
+          tab = VP8Cat3;
+        } else if (v < 3 + (8 << 2)) {   // VP8Cat4  (4b)
           VP8PutBit(bw, 0, p[8]);
           VP8PutBit(bw, 1, p[9]);
           v -= 3 + (8 << 1);
           mask = 1 << 3;
-          tab = kCat4;
-        } else if (v < 3 + (8 << 3)) {   // kCat5  (5b)
+          tab = VP8Cat4;
+        } else if (v < 3 + (8 << 3)) {   // VP8Cat5  (5b)
           VP8PutBit(bw, 1, p[8]);
           VP8PutBit(bw, 0, p[10]);
           v -= 3 + (8 << 2);
           mask = 1 << 4;
-          tab = kCat5;
-        } else {                         // kCat6 (11b)
+          tab = VP8Cat5;
+        } else {                         // VP8Cat6 (11b)
           VP8PutBit(bw, 1, p[8]);
           VP8PutBit(bw, 1, p[10]);
           v -= 3 + (8 << 3);
           mask = 1 << 10;
-          tab = kCat6;
+          tab = VP8Cat6;
         }
         while (mask) {
           VP8PutBit(bw, !!(v & mask), *tab++);
@@ -411,8 +465,7 @@ static int PutCoeffs(VP8BitWriter* const bw, int ctx, const VP8Residual* res) {
   return 1;
 }
 
-static void CodeResiduals(VP8BitWriter* const bw,
-                          VP8EncIterator* const it,
+static void CodeResiduals(VP8BitWriter* const bw, VP8EncIterator* const it,
                           const VP8ModeScore* const rd) {
   int x, y, ch;
   VP8Residual res;
@@ -512,146 +565,23 @@ static void RecordResiduals(VP8EncIterator* const it,
 //------------------------------------------------------------------------------
 // Token buffer
 
-#ifdef USE_TOKEN_BUFFER
+#if !defined(DISABLE_TOKEN_BUFFER)
 
-void VP8TBufferInit(VP8TBuffer* const b) {
-  b->rows_ = NULL;
-  b->tokens_ = NULL;
-  b->last_ = &b->rows_;
-  b->left_ = 0;
-  b->error_ = 0;
-}
-
-int VP8TBufferNewPage(VP8TBuffer* const b) {
-  VP8Tokens* const page = b->error_ ? NULL : (VP8Tokens*)malloc(sizeof(*page));
-  if (page == NULL) {
-    b->error_ = 1;
-    return 0;
-  }
-  *b->last_ = page;
-  b->last_ = &page->next_;
-  b->left_ = MAX_NUM_TOKEN;
-  b->tokens_ = page->tokens_;
-  return 1;
-}
-
-void VP8TBufferClear(VP8TBuffer* const b) {
-  if (b != NULL) {
-    const VP8Tokens* p = b->rows_;
-    while (p != NULL) {
-      const VP8Tokens* const next = p->next_;
-      free((void*)p);
-      p = next;
-    }
-    VP8TBufferInit(b);
-  }
-}
-
-int VP8EmitTokens(const VP8TBuffer* const b, VP8BitWriter* const bw,
-                  const uint8_t* const probas) {
-  VP8Tokens* p = b->rows_;
-  if (b->error_) return 0;
-  while (p != NULL) {
-    const int N = (p->next_ == NULL) ? b->left_ : 0;
-    int n = MAX_NUM_TOKEN;
-    while (n-- > N) {
-      VP8PutBit(bw, (p->tokens_[n] >> 15) & 1, probas[p->tokens_[n] & 0x7fff]);
-    }
-    p = p->next_;
-  }
-  return 1;
-}
-
-#define TOKEN_ID(b, ctx, p) ((p) + NUM_PROBAS * ((ctx) + (b) * NUM_CTX))
-
-static int RecordCoeffTokens(int ctx, const VP8Residual* const res,
-                             VP8TBuffer* tokens) {
-  int n = res->first;
-  int b = VP8EncBands[n];
-  if (!VP8AddToken(tokens, res->last >= 0, TOKEN_ID(b, ctx, 0))) {
-    return 0;
-  }
-
-  while (n < 16) {
-    const int c = res->coeffs[n++];
-    const int sign = c < 0;
-    int v = sign ? -c : c;
-    const int base_id = TOKEN_ID(b, ctx, 0);
-    if (!VP8AddToken(tokens, v != 0, base_id + 1)) {
-      b = VP8EncBands[n];
-      ctx = 0;
-      continue;
-    }
-    if (!VP8AddToken(tokens, v > 1, base_id + 2)) {
-      b = VP8EncBands[n];
-      ctx = 1;
-    } else {
-      if (!VP8AddToken(tokens, v > 4, base_id + 3)) {
-        if (VP8AddToken(tokens, v != 2, base_id + 4))
-          VP8AddToken(tokens, v == 4, base_id + 5);
-      } else if (!VP8AddToken(tokens, v > 10, base_id + 6)) {
-        if (!VP8AddToken(tokens, v > 6, base_id + 7)) {
-//          VP8AddToken(tokens, v == 6, 159);
-        } else {
-//          VP8AddToken(tokens, v >= 9, 165);
-//          VP8AddToken(tokens, !(v & 1), 145);
-        }
-      } else {
-        int mask;
-        const uint8_t* tab;
-        if (v < 3 + (8 << 1)) {          // kCat3  (3b)
-          VP8AddToken(tokens, 0, base_id + 8);
-          VP8AddToken(tokens, 0, base_id + 9);
-          v -= 3 + (8 << 0);
-          mask = 1 << 2;
-          tab = kCat3;
-        } else if (v < 3 + (8 << 2)) {   // kCat4  (4b)
-          VP8AddToken(tokens, 0, base_id + 8);
-          VP8AddToken(tokens, 1, base_id + 9);
-          v -= 3 + (8 << 1);
-          mask = 1 << 3;
-          tab = kCat4;
-        } else if (v < 3 + (8 << 3)) {   // kCat5  (5b)
-          VP8AddToken(tokens, 1, base_id + 8);
-          VP8AddToken(tokens, 0, base_id + 10);
-          v -= 3 + (8 << 2);
-          mask = 1 << 4;
-          tab = kCat5;
-        } else {                         // kCat6 (11b)
-          VP8AddToken(tokens, 1, base_id + 8);
-          VP8AddToken(tokens, 1, base_id + 10);
-          v -= 3 + (8 << 3);
-          mask = 1 << 10;
-          tab = kCat6;
-        }
-        while (mask) {
-          // VP8AddToken(tokens, !!(v & mask), *tab++);
-          mask >>= 1;
-        }
-      }
-      ctx = 2;
-    }
-    b = VP8EncBands[n];
-    // VP8PutBitUniform(bw, sign);
-    if (n == 16 || !VP8AddToken(tokens, n <= res->last, TOKEN_ID(b, ctx, 0))) {
-      return 1;   // EOB
-    }
-  }
-  return 1;
-}
-
-static void RecordTokens(VP8EncIterator* const it,
-                         const VP8ModeScore* const rd, VP8TBuffer tokens[2]) {
+static void RecordTokens(VP8EncIterator* const it, const VP8ModeScore* const rd,
+                         VP8TBuffer* const tokens) {
   int x, y, ch;
   VP8Residual res;
   VP8Encoder* const enc = it->enc_;
 
   VP8IteratorNzToBytes(it);
   if (it->mb_->type_ == 1) {   // i16x16
+    const int ctx = it->top_nz_[8] + it->left_nz_[8];
     InitResidual(0, 1, enc, &res);
     SetResidualCoeffs(rd->y_dc_levels, &res);
-// TODO(skal): FIX ->    it->top_nz_[8] = it->left_nz_[8] =
-      RecordCoeffTokens(it->top_nz_[8] + it->left_nz_[8], &res, &tokens[0]);
+    it->top_nz_[8] = it->left_nz_[8] =
+        VP8RecordCoeffTokens(ctx, 1,
+                             res.first, res.last, res.coeffs, tokens);
+    RecordCoeffs(ctx, &res);
     InitResidual(1, 0, enc, &res);
   } else {
     InitResidual(0, 3, enc, &res);
@@ -663,7 +593,9 @@ static void RecordTokens(VP8EncIterator* const it,
       const int ctx = it->top_nz_[x] + it->left_nz_[y];
       SetResidualCoeffs(rd->y_ac_levels[x + y * 4], &res);
       it->top_nz_[x] = it->left_nz_[y] =
-          RecordCoeffTokens(ctx, &res, &tokens[0]);
+          VP8RecordCoeffTokens(ctx, res.coeff_type,
+                               res.first, res.last, res.coeffs, tokens);
+      RecordCoeffs(ctx, &res);
     }
   }
 
@@ -675,13 +607,16 @@ static void RecordTokens(VP8EncIterator* const it,
         const int ctx = it->top_nz_[4 + ch + x] + it->left_nz_[4 + ch + y];
         SetResidualCoeffs(rd->uv_levels[ch * 2 + x + y * 2], &res);
         it->top_nz_[4 + ch + x] = it->left_nz_[4 + ch + y] =
-            RecordCoeffTokens(ctx, &res, &tokens[1]);
+            VP8RecordCoeffTokens(ctx, 2,
+                                 res.first, res.last, res.coeffs, tokens);
+        RecordCoeffs(ctx, &res);
       }
     }
   }
+  VP8IteratorBytesToNz(it);
 }
 
-#endif    // USE_TOKEN_BUFFER
+#endif    // !DISABLE_TOKEN_BUFFER
 
 //------------------------------------------------------------------------------
 // ExtraInfo map / Debug function
@@ -697,7 +632,10 @@ static void SetBlock(uint8_t* p, int value, int size) {
 #endif
 
 static void ResetSSE(VP8Encoder* const enc) {
-  memset(enc->sse_, 0, sizeof(enc->sse_));
+  enc->sse_[0] = 0;
+  enc->sse_[1] = 0;
+  enc->sse_[2] = 0;
+  // Note: enc->sse_[3] is managed by alpha.c
   enc->sse_count_ = 0;
 }
 
@@ -736,6 +674,7 @@ static void StoreSideInfo(const VP8EncIterator* const it) {
         const int b = (int)((it->luma_bits_ + it->uv_bits_ + 7) >> 3);
         *info = (b > 255) ? 255 : b; break;
       }
+      case 7: *info = mb->alpha_; break;
       default: *info = 0; break;
     };
   }
@@ -746,9 +685,173 @@ static void StoreSideInfo(const VP8EncIterator* const it) {
 #endif
 }
 
+//------------------------------------------------------------------------------
+//  StatLoop(): only collect statistics (number of skips, token usage, ...).
+//  This is used for deciding optimal probabilities. It also modifies the
+//  quantizer value if some target (size, PNSR) was specified.
+
+#define kHeaderSizeEstimate (15 + 20 + 10)      // TODO: fix better
+
+static void SetLoopParams(VP8Encoder* const enc, float q) {
+  // Make sure the quality parameter is inside valid bounds
+  if (q < 0.) {
+    q = 0;
+  } else if (q > 100.) {
+    q = 100;
+  }
+
+  VP8SetSegmentParams(enc, q);      // setup segment quantizations and filters
+  SetSegmentProbas(enc);            // compute segment probabilities
+
+  ResetStats(enc);
+  ResetTokenStats(enc);
+
+  ResetSSE(enc);
+}
+
+static int OneStatPass(VP8Encoder* const enc, float q, VP8RDLevel rd_opt,
+                       int nb_mbs, float* const PSNR, int percent_delta) {
+  VP8EncIterator it;
+  uint64_t size = 0;
+  uint64_t distortion = 0;
+  const uint64_t pixel_count = nb_mbs * 384;
+
+  SetLoopParams(enc, q);
+
+  VP8IteratorInit(enc, &it);
+  do {
+    VP8ModeScore info;
+    VP8IteratorImport(&it);
+    if (VP8Decimate(&it, &info, rd_opt)) {
+      // Just record the number of skips and act like skip_proba is not used.
+      enc->proba_.nb_skip_++;
+    }
+    RecordResiduals(&it, &info);
+    size += info.R;
+    distortion += info.D;
+    if (percent_delta && !VP8IteratorProgress(&it, percent_delta))
+      return 0;
+  } while (VP8IteratorNext(&it, it.yuv_out_) && --nb_mbs > 0);
+  size += FinalizeSkipProba(enc);
+  size += FinalizeTokenProbas(&enc->proba_);
+  size += enc->segment_hdr_.size_;
+  size = ((size + 1024) >> 11) + kHeaderSizeEstimate;
+
+  if (PSNR) {
+    *PSNR = (float)(10.* log10(255. * 255. * pixel_count / distortion));
+  }
+  return (int)size;
+}
+
+// successive refinement increments.
+static const int dqs[] = { 20, 15, 10, 8, 6, 4, 2, 1, 0 };
+
+static int StatLoop(VP8Encoder* const enc) {
+  const int method = enc->method_;
+  const int do_search = enc->do_search_;
+  const int fast_probe = ((method == 0 || method == 3) && !do_search);
+  float q = enc->config_->quality;
+  const int max_passes = enc->config_->pass;
+  const int task_percent = 20;
+  const int percent_per_pass = (task_percent + max_passes / 2) / max_passes;
+  const int final_percent = enc->percent_ + task_percent;
+  int pass;
+  int nb_mbs;
+
+  // Fast mode: quick analysis pass over few mbs. Better than nothing.
+  nb_mbs = enc->mb_w_ * enc->mb_h_;
+  if (fast_probe) {
+    if (method == 3) {  // we need more stats for method 3 to be reliable.
+      nb_mbs = (nb_mbs > 200) ? nb_mbs >> 1 : 100;
+    } else {
+      nb_mbs = (nb_mbs > 200) ? nb_mbs >> 2 : 50;
+    }
+  }
+
+  // No target size: just do several pass without changing 'q'
+  if (!do_search) {
+    for (pass = 0; pass < max_passes; ++pass) {
+      const VP8RDLevel rd_opt = (method >= 3) ? RD_OPT_BASIC : RD_OPT_NONE;
+      if (!OneStatPass(enc, q, rd_opt, nb_mbs, NULL, percent_per_pass)) {
+        return 0;
+      }
+    }
+  } else {
+    // binary search for a size close to target
+    for (pass = 0; pass < max_passes && (dqs[pass] > 0); ++pass) {
+      float PSNR;
+      int criterion;
+      const int size = OneStatPass(enc, q, RD_OPT_BASIC, nb_mbs, &PSNR,
+                                   percent_per_pass);
+#if DEBUG_SEARCH
+      printf("#%d size=%d PSNR=%.2f q=%.2f\n", pass, size, PSNR, q);
+#endif
+      if (size == 0) return 0;
+      if (enc->config_->target_PSNR > 0) {
+        criterion = (PSNR < enc->config_->target_PSNR);
+      } else {
+        criterion = (size < enc->config_->target_size);
+      }
+      // dichotomize
+      if (criterion) {
+        q += dqs[pass];
+      } else {
+        q -= dqs[pass];
+      }
+    }
+  }
+  VP8CalculateLevelCosts(&enc->proba_);  // finalize costs
+  return WebPReportProgress(enc->pic_, final_percent, &enc->percent_);
+}
+
 //------------------------------------------------------------------------------
 // Main loops
 //
+
+static const int kAverageBytesPerMB[8] = { 50, 24, 16, 9, 7, 5, 3, 2 };
+
+static int PreLoopInitialize(VP8Encoder* const enc) {
+  int p;
+  int ok = 1;
+  const int average_bytes_per_MB = kAverageBytesPerMB[enc->base_quant_ >> 4];
+  const int bytes_per_parts =
+      enc->mb_w_ * enc->mb_h_ * average_bytes_per_MB / enc->num_parts_;
+  // Initialize the bit-writers
+  for (p = 0; ok && p < enc->num_parts_; ++p) {
+    ok = VP8BitWriterInit(enc->parts_ + p, bytes_per_parts);
+  }
+  if (!ok) VP8EncFreeBitWriters(enc);  // malloc error occurred
+  return ok;
+}
+
+static int PostLoopFinalize(VP8EncIterator* const it, int ok) {
+  VP8Encoder* const enc = it->enc_;
+  if (ok) {      // Finalize the partitions, check for extra errors.
+    int p;
+    for (p = 0; p < enc->num_parts_; ++p) {
+      VP8BitWriterFinish(enc->parts_ + p);
+      ok &= !enc->parts_[p].error_;
+    }
+  }
+
+  if (ok) {      // All good. Finish up.
+    if (enc->pic_->stats) {           // finalize byte counters...
+      int i, s;
+      for (i = 0; i <= 2; ++i) {
+        for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
+          enc->residual_bytes_[i][s] = (int)((it->bit_count_[s][i] + 7) >> 3);
+        }
+      }
+    }
+    VP8AdjustFilterStrength(it);     // ...and store filter stats.
+  } else {
+    // Something bad happened -> need to do some memory cleanup.
+    VP8EncFreeBitWriters(enc);
+  }
+  return ok;
+}
+
+//------------------------------------------------------------------------------
 //  VP8EncLoop(): does the final bitstream coding.
 
 static void ResetAfterSkip(VP8EncIterator* const it) {
@@ -761,27 +864,19 @@ static void ResetAfterSkip(VP8EncIterator* const it) {
 }
 
 int VP8EncLoop(VP8Encoder* const enc) {
-  int i, s, p;
-  int ok = 1;
   VP8EncIterator it;
-  VP8ModeScore info;
-  const int dont_use_skip = !enc->proba_.use_skip_proba_;
-  const int rd_opt = enc->rd_opt_level_;
-  const int kAverageBytesPerMB = 5;     // TODO: have a kTable[quality/10]
-  const int bytes_per_parts =
-    enc->mb_w_ * enc->mb_h_ * kAverageBytesPerMB / enc->num_parts_;
+  int ok = PreLoopInitialize(enc);
+  if (!ok) return 0;
 
-  // Initialize the bit-writers
-  for (p = 0; p < enc->num_parts_; ++p) {
-    VP8BitWriterInit(enc->parts_ + p, bytes_per_parts);
-  }
-
-  ResetStats(enc);
-  ResetSSE(enc);
+  StatLoop(enc);  // stats-collection loop
 
   VP8IteratorInit(enc, &it);
   VP8InitFilter(&it);
   do {
+    VP8ModeScore info;
+    const int dont_use_skip = !enc->proba_.use_skip_proba_;
+    const VP8RDLevel rd_opt = enc->rd_opt_level_;
+
     VP8IteratorImport(&it);
     // Warning! order is important: first call VP8Decimate() and
     // *then* decide how to code the skip decision if there's one.
@@ -801,137 +896,82 @@ int VP8EncLoop(VP8Encoder* const enc) {
     ok = VP8IteratorProgress(&it, 20);
   } while (ok && VP8IteratorNext(&it, it.yuv_out_));
 
-  if (ok) {      // Finalize the partitions, check for extra errors.
-    for (p = 0; p < enc->num_parts_; ++p) {
-      VP8BitWriterFinish(enc->parts_ + p);
-      ok &= !enc->parts_[p].error_;
-    }
-  }
-
-  if (ok) {      // All good. Finish up.
-    if (enc->pic_->stats) {           // finalize byte counters...
-      for (i = 0; i <= 2; ++i) {
-        for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
-          enc->residual_bytes_[i][s] = (int)((it.bit_count_[s][i] + 7) >> 3);
-        }
-      }
-    }
-    VP8AdjustFilterStrength(&it);     // ...and store filter stats.
-  } else {
-    // Something bad happened -> need to do some memory cleanup.
-    VP8EncFreeBitWriters(enc);
-  }
-
-  return ok;
+  return PostLoopFinalize(&it, ok);
 }
 
 //------------------------------------------------------------------------------
-//  VP8StatLoop(): only collect statistics (number of skips, token usage, ...)
-//                 This is used for deciding optimal probabilities. It also
-//                 modifies the quantizer value if some target (size, PNSR)
-//                 was specified.
+// Single pass using Token Buffer.
 
-#define kHeaderSizeEstimate (15 + 20 + 10)      // TODO: fix better
+#if !defined(DISABLE_TOKEN_BUFFER)
 
-static int OneStatPass(VP8Encoder* const enc, float q, int rd_opt, int nb_mbs,
-                       float* const PSNR, int percent_delta) {
+#define MIN_COUNT 96   // minimum number of macroblocks before updating stats
+
+int VP8EncTokenLoop(VP8Encoder* const enc) {
+  int ok;
+  // Roughly refresh the proba height times per pass
+  int max_count = (enc->mb_w_ * enc->mb_h_) >> 3;
+  int cnt;
   VP8EncIterator it;
-  uint64_t size = 0;
-  uint64_t distortion = 0;
-  const uint64_t pixel_count = nb_mbs * 384;
+  VP8Proba* const proba = &enc->proba_;
+  const VP8RDLevel rd_opt = enc->rd_opt_level_;
 
-  // Make sure the quality parameter is inside valid bounds
-  if (q < 0.) {
-    q = 0;
-  } else if (q > 100.) {
-    q = 100;
-  }
+  if (max_count < MIN_COUNT) max_count = MIN_COUNT;
+  cnt = max_count;
 
-  VP8SetSegmentParams(enc, q);      // setup segment quantizations and filters
+  assert(enc->num_parts_ == 1);
+  assert(enc->use_tokens_);
+  assert(proba->use_skip_proba_ == 0);
+  assert(rd_opt >= RD_OPT_BASIC);   // otherwise, token-buffer won't be useful
+  assert(!enc->do_search_);         // TODO(skal): handle pass and dichotomy
 
-  ResetStats(enc);
-  ResetTokenStats(enc);
+  SetLoopParams(enc, enc->config_->quality);
+
+  ok = PreLoopInitialize(enc);
+  if (!ok) return 0;
 
   VP8IteratorInit(enc, &it);
+  VP8InitFilter(&it);
   do {
     VP8ModeScore info;
     VP8IteratorImport(&it);
-    if (VP8Decimate(&it, &info, rd_opt)) {
-      // Just record the number of skips and act like skip_proba is not used.
-      enc->proba_.nb_skip_++;
+    if (--cnt < 0) {
+      FinalizeTokenProbas(proba);
+      VP8CalculateLevelCosts(proba);  // refresh cost tables for rd-opt
+      cnt = max_count;
     }
-    RecordResiduals(&it, &info);
-    size += info.R;
-    distortion += info.D;
-    if (percent_delta && !VP8IteratorProgress(&it, percent_delta))
-      return 0;
-  } while (VP8IteratorNext(&it, it.yuv_out_) && --nb_mbs > 0);
-  size += FinalizeSkipProba(enc);
-  size += FinalizeTokenProbas(enc);
-  size += enc->segment_hdr_.size_;
-  size = ((size + 1024) >> 11) + kHeaderSizeEstimate;
-
-  if (PSNR) {
-    *PSNR = (float)(10.* log10(255. * 255. * pixel_count / distortion));
-  }
-  return (int)size;
-}
-
-// successive refinement increments.
-static const int dqs[] = { 20, 15, 10, 8, 6, 4, 2, 1, 0 };
-
-int VP8StatLoop(VP8Encoder* const enc) {
-  const int do_search =
-    (enc->config_->target_size > 0 || enc->config_->target_PSNR > 0);
-  const int fast_probe = (enc->method_ < 2 && !do_search);
-  float q = enc->config_->quality;
-  const int max_passes = enc->config_->pass;
-  const int task_percent = 20;
-  const int percent_per_pass = (task_percent + max_passes / 2) / max_passes;
-  const int final_percent = enc->percent_ + task_percent;
-  int pass;
-  int nb_mbs;
-
-  // Fast mode: quick analysis pass over few mbs. Better than nothing.
-  nb_mbs = enc->mb_w_ * enc->mb_h_;
-  if (fast_probe && nb_mbs > 100) nb_mbs = 100;
-
-  // No target size: just do several pass without changing 'q'
-  if (!do_search) {
-    for (pass = 0; pass < max_passes; ++pass) {
-      const int rd_opt = (enc->method_ > 2);
-      if (!OneStatPass(enc, q, rd_opt, nb_mbs, NULL, percent_per_pass)) {
-        return 0;
-      }
+    VP8Decimate(&it, &info, rd_opt);
+    RecordTokens(&it, &info, &enc->tokens_);
+#ifdef WEBP_EXPERIMENTAL_FEATURES
+    if (enc->use_layer_) {
+      VP8EncCodeLayerBlock(&it);
     }
-  } else {
-    // binary search for a size close to target
-    for (pass = 0; pass < max_passes && (dqs[pass] > 0); ++pass) {
-      const int rd_opt = 1;
-      float PSNR;
-      int criterion;
-      const int size = OneStatPass(enc, q, rd_opt, nb_mbs, &PSNR,
-                                   percent_per_pass);
-#if DEBUG_SEARCH
-      printf("#%d size=%d PSNR=%.2f q=%.2f\n", pass, size, PSNR, q);
 #endif
-      if (!size) return 0;
-      if (enc->config_->target_PSNR > 0) {
-        criterion = (PSNR < enc->config_->target_PSNR);
-      } else {
-        criterion = (size < enc->config_->target_size);
-      }
-      // dichotomize
-      if (criterion) {
-        q += dqs[pass];
-      } else {
-        q -= dqs[pass];
-      }
-    }
+    StoreSideInfo(&it);
+    VP8StoreFilterStats(&it);
+    VP8IteratorExport(&it);
+    ok = VP8IteratorProgress(&it, 20);
+  } while (ok && VP8IteratorNext(&it, it.yuv_out_));
+
+  ok = ok && WebPReportProgress(enc->pic_, enc->percent_ + 20, &enc->percent_);
+
+  if (ok) {
+    FinalizeTokenProbas(proba);
+    ok = VP8EmitTokens(&enc->tokens_, enc->parts_ + 0,
+                       (const uint8_t*)proba->coeffs_, 1);
   }
-  return WebPReportProgress(enc->pic_, final_percent, &enc->percent_);
+
+  return PostLoopFinalize(&it, ok);
 }
 
+#else
+
+int VP8EncTokenLoop(VP8Encoder* const enc) {
+  (void)enc;
+  return 0;   // we shouldn't be here.
+}
+
+#endif    // DISABLE_TOKEN_BUFFER
+
 //------------------------------------------------------------------------------
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/enc/histogram.c b/3rdparty/libwebp/enc/histogram.c
index fb4044bfd3..787ea5d183 100644
--- a/3rdparty/libwebp/enc/histogram.c
+++ b/3rdparty/libwebp/enc/histogram.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
@@ -98,8 +100,6 @@ void VP8LHistogramAddSinglePixOrCopy(VP8LHistogram* const histo,
   }
 }
 
-
-
 static double BitsEntropy(const int* const array, int n) {
   double retval = 0.;
   int sum = 0;
@@ -149,25 +149,6 @@ static double BitsEntropy(const int* const array, int n) {
   }
 }
 
-double VP8LHistogramEstimateBitsBulk(const VP8LHistogram* const p) {
-  double retval = BitsEntropy(&p->literal_[0], VP8LHistogramNumCodes(p))
-                + BitsEntropy(&p->red_[0], 256)
-                + BitsEntropy(&p->blue_[0], 256)
-                + BitsEntropy(&p->alpha_[0], 256)
-                + BitsEntropy(&p->distance_[0], NUM_DISTANCE_CODES);
-  // Compute the extra bits cost.
-  int i;
-  for (i = 2; i < NUM_LENGTH_CODES - 2; ++i) {
-    retval +=
-        (i >> 1) * p->literal_[256 + i + 2];
-  }
-  for (i = 2; i < NUM_DISTANCE_CODES - 2; ++i) {
-    retval += (i >> 1) * p->distance_[i + 2];
-  }
-  return retval;
-}
-
-
 // Returns the cost encode the rle-encoded entropy code.
 // The constants in this function are experimental.
 static double HuffmanCost(const int* const population, int length) {
@@ -207,19 +188,150 @@ static double HuffmanCost(const int* const population, int length) {
   return retval;
 }
 
-// Estimates the Huffman dictionary + other block overhead size.
-static double HistogramEstimateBitsHeader(const VP8LHistogram* const p) {
-  return HuffmanCost(&p->alpha_[0], 256) +
-         HuffmanCost(&p->red_[0], 256) +
-         HuffmanCost(&p->literal_[0], VP8LHistogramNumCodes(p)) +
-         HuffmanCost(&p->blue_[0], 256) +
-         HuffmanCost(&p->distance_[0], NUM_DISTANCE_CODES);
+static double PopulationCost(const int* const population, int length) {
+  return BitsEntropy(population, length) + HuffmanCost(population, length);
 }
 
-double VP8LHistogramEstimateBits(const VP8LHistogram* const p) {
-  return HistogramEstimateBitsHeader(p) + VP8LHistogramEstimateBitsBulk(p);
+static double ExtraCost(const int* const population, int length) {
+  int i;
+  double cost = 0.;
+  for (i = 2; i < length - 2; ++i) cost += (i >> 1) * population[i + 2];
+  return cost;
 }
 
+// Estimates the Entropy + Huffman + other block overhead size cost.
+double VP8LHistogramEstimateBits(const VP8LHistogram* const p) {
+  return PopulationCost(p->literal_, VP8LHistogramNumCodes(p))
+       + PopulationCost(p->red_, 256)
+       + PopulationCost(p->blue_, 256)
+       + PopulationCost(p->alpha_, 256)
+       + PopulationCost(p->distance_, NUM_DISTANCE_CODES)
+       + ExtraCost(p->literal_ + 256, NUM_LENGTH_CODES)
+       + ExtraCost(p->distance_, NUM_DISTANCE_CODES);
+}
+
+double VP8LHistogramEstimateBitsBulk(const VP8LHistogram* const p) {
+  return BitsEntropy(p->literal_, VP8LHistogramNumCodes(p))
+       + BitsEntropy(p->red_, 256)
+       + BitsEntropy(p->blue_, 256)
+       + BitsEntropy(p->alpha_, 256)
+       + BitsEntropy(p->distance_, NUM_DISTANCE_CODES)
+       + ExtraCost(p->literal_ + 256, NUM_LENGTH_CODES)
+       + ExtraCost(p->distance_, NUM_DISTANCE_CODES);
+}
+
+// -----------------------------------------------------------------------------
+// Various histogram combine/cost-eval functions
+
+// Adds 'in' histogram to 'out'
+static void HistogramAdd(const VP8LHistogram* const in,
+                         VP8LHistogram* const out) {
+  int i;
+  for (i = 0; i < PIX_OR_COPY_CODES_MAX; ++i) {
+    out->literal_[i] += in->literal_[i];
+  }
+  for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
+    out->distance_[i] += in->distance_[i];
+  }
+  for (i = 0; i < 256; ++i) {
+    out->red_[i] += in->red_[i];
+    out->blue_[i] += in->blue_[i];
+    out->alpha_[i] += in->alpha_[i];
+  }
+}
+
+// Performs out = a + b, computing the cost C(a+b) - C(a) - C(b) while comparing
+// to the threshold value 'cost_threshold'. The score returned is
+//  Score = C(a+b) - C(a) - C(b), where C(a) + C(b) is known and fixed.
+// Since the previous score passed is 'cost_threshold', we only need to compare
+// the partial cost against 'cost_threshold + C(a) + C(b)' to possibly bail-out
+// early.
+static double HistogramAddEval(const VP8LHistogram* const a,
+                               const VP8LHistogram* const b,
+                               VP8LHistogram* const out,
+                               double cost_threshold) {
+  double cost = 0;
+  const double sum_cost = a->bit_cost_ + b->bit_cost_;
+  int i;
+
+  cost_threshold += sum_cost;
+
+  // palette_code_bits_ is part of the cost evaluation for literal_.
+  // TODO(skal): remove/simplify this palette_code_bits_?
+  out->palette_code_bits_ =
+      (a->palette_code_bits_ > b->palette_code_bits_) ? a->palette_code_bits_ :
+                                                        b->palette_code_bits_;
+  for (i = 0; i < PIX_OR_COPY_CODES_MAX; ++i) {
+    out->literal_[i] = a->literal_[i] + b->literal_[i];
+  }
+  cost += PopulationCost(out->literal_, VP8LHistogramNumCodes(out));
+  cost += ExtraCost(out->literal_ + 256, NUM_LENGTH_CODES);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) out->red_[i] = a->red_[i] + b->red_[i];
+  cost += PopulationCost(out->red_, 256);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) out->blue_[i] = a->blue_[i] + b->blue_[i];
+  cost += PopulationCost(out->blue_, 256);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
+    out->distance_[i] = a->distance_[i] + b->distance_[i];
+  }
+  cost += PopulationCost(out->distance_, NUM_DISTANCE_CODES);
+  cost += ExtraCost(out->distance_, NUM_DISTANCE_CODES);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) out->alpha_[i] = a->alpha_[i] + b->alpha_[i];
+  cost += PopulationCost(out->alpha_, 256);
+
+  out->bit_cost_ = cost;
+  return cost - sum_cost;
+}
+
+// Same as HistogramAddEval(), except that the resulting histogram
+// is not stored. Only the cost C(a+b) - C(a) is evaluated. We omit
+// the term C(b) which is constant over all the evaluations.
+static double HistogramAddThresh(const VP8LHistogram* const a,
+                                 const VP8LHistogram* const b,
+                                 double cost_threshold) {
+  int tmp[PIX_OR_COPY_CODES_MAX];  // <= max storage we'll need
+  int i;
+  double cost = -a->bit_cost_;
+
+  for (i = 0; i < PIX_OR_COPY_CODES_MAX; ++i) {
+    tmp[i] = a->literal_[i] + b->literal_[i];
+  }
+  // note that the tests are ordered so that the usually largest
+  // cost shares come first.
+  cost += PopulationCost(tmp, VP8LHistogramNumCodes(a));
+  cost += ExtraCost(tmp + 256, NUM_LENGTH_CODES);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) tmp[i] = a->red_[i] + b->red_[i];
+  cost += PopulationCost(tmp, 256);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) tmp[i] = a->blue_[i] + b->blue_[i];
+  cost += PopulationCost(tmp, 256);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
+    tmp[i] = a->distance_[i] + b->distance_[i];
+  }
+  cost += PopulationCost(tmp, NUM_DISTANCE_CODES);
+  cost += ExtraCost(tmp, NUM_DISTANCE_CODES);
+  if (cost > cost_threshold) return cost;
+
+  for (i = 0; i < 256; ++i) tmp[i] = a->alpha_[i] + b->alpha_[i];
+  cost += PopulationCost(tmp, 256);
+
+  return cost;
+}
+
+// -----------------------------------------------------------------------------
+
 static void HistogramBuildImage(int xsize, int histo_bits,
                                 const VP8LBackwardRefs* const backward_refs,
                                 VP8LHistogramSet* const image) {
@@ -249,14 +361,15 @@ static uint32_t MyRand(uint32_t *seed) {
 }
 
 static int HistogramCombine(const VP8LHistogramSet* const in,
-                            VP8LHistogramSet* const out, int num_pairs) {
+                            VP8LHistogramSet* const out, int iter_mult,
+                            int num_pairs, int num_tries_no_success) {
   int ok = 0;
   int i, iter;
   uint32_t seed = 0;
   int tries_with_no_success = 0;
-  const int min_cluster_size = 2;
   int out_size = in->size;
-  const int outer_iters = in->size * 3;
+  const int outer_iters = in->size * iter_mult;
+  const int min_cluster_size = 2;
   VP8LHistogram* const histos = (VP8LHistogram*)malloc(2 * sizeof(*histos));
   VP8LHistogram* cur_combo = histos + 0;    // trial merged histogram
   VP8LHistogram* best_combo = histos + 1;   // best merged histogram so far
@@ -271,29 +384,26 @@ static int HistogramCombine(const VP8LHistogramSet* const in,
 
   // Collapse similar histograms in 'out'.
   for (iter = 0; iter < outer_iters && out_size >= min_cluster_size; ++iter) {
-    // We pick the best pair to be combined out of 'inner_iters' pairs.
     double best_cost_diff = 0.;
-    int best_idx1 = 0, best_idx2 = 1;
+    int best_idx1 = -1, best_idx2 = 1;
     int j;
+    const int num_tries = (num_pairs < out_size) ? num_pairs : out_size;
     seed += iter;
-    for (j = 0; j < num_pairs; ++j) {
+    for (j = 0; j < num_tries; ++j) {
       double curr_cost_diff;
       // Choose two histograms at random and try to combine them.
       const uint32_t idx1 = MyRand(&seed) % out_size;
-      const uint32_t tmp = ((j & 7) + 1) % (out_size - 1);
+      const uint32_t tmp = (j & 7) + 1;
       const uint32_t diff = (tmp < 3) ? tmp : MyRand(&seed) % (out_size - 1);
       const uint32_t idx2 = (idx1 + diff + 1) % out_size;
       if (idx1 == idx2) {
         continue;
       }
-      *cur_combo = *out->histograms[idx1];
-      VP8LHistogramAdd(cur_combo, out->histograms[idx2]);
-      cur_combo->bit_cost_ = VP8LHistogramEstimateBits(cur_combo);
       // Calculate cost reduction on combining.
-      curr_cost_diff = cur_combo->bit_cost_
-                     - out->histograms[idx1]->bit_cost_
-                     - out->histograms[idx2]->bit_cost_;
-      if (best_cost_diff > curr_cost_diff) {    // found a better pair?
+      curr_cost_diff = HistogramAddEval(out->histograms[idx1],
+                                        out->histograms[idx2],
+                                        cur_combo, best_cost_diff);
+      if (curr_cost_diff < best_cost_diff) {    // found a better pair?
         {     // swap cur/best combo histograms
           VP8LHistogram* const tmp_histo = cur_combo;
           cur_combo = best_combo;
@@ -305,7 +415,7 @@ static int HistogramCombine(const VP8LHistogramSet* const in,
       }
     }
 
-    if (best_cost_diff < 0.0) {
+    if (best_idx1 >= 0) {
       *out->histograms[best_idx1] = *best_combo;
       // swap best_idx2 slot with last one (which is now unused)
       --out_size;
@@ -315,7 +425,7 @@ static int HistogramCombine(const VP8LHistogramSet* const in,
       }
       tries_with_no_success = 0;
     }
-    if (++tries_with_no_success >= 50) {
+    if (++tries_with_no_success >= num_tries_no_success) {
       break;
     }
   }
@@ -330,20 +440,11 @@ static int HistogramCombine(const VP8LHistogramSet* const in,
 // -----------------------------------------------------------------------------
 // Histogram refinement
 
-// What is the bit cost of moving square_histogram from
-// cur_symbol to candidate_symbol.
-// TODO(skal): we don't really need to copy the histogram and Add(). Instead
-// we just need VP8LDualHistogramEstimateBits(A, B) estimation function.
+// What is the bit cost of moving square_histogram from cur_symbol to candidate.
 static double HistogramDistance(const VP8LHistogram* const square_histogram,
-                                const VP8LHistogram* const candidate) {
-  const double previous_bit_cost = candidate->bit_cost_;
-  double new_bit_cost;
-  VP8LHistogram modified_histo;
-  modified_histo = *candidate;
-  VP8LHistogramAdd(&modified_histo, square_histogram);
-  new_bit_cost = VP8LHistogramEstimateBits(&modified_histo);
-
-  return new_bit_cost - previous_bit_cost;
+                                const VP8LHistogram* const candidate,
+                                double cost_threshold) {
+  return HistogramAddThresh(candidate, square_histogram, cost_threshold);
 }
 
 // Find the best 'out' histogram for each of the 'in' histograms.
@@ -354,11 +455,12 @@ static void HistogramRemap(const VP8LHistogramSet* const in,
   int i;
   for (i = 0; i < in->size; ++i) {
     int best_out = 0;
-    double best_bits = HistogramDistance(in->histograms[i], out->histograms[0]);
+    double best_bits =
+        HistogramDistance(in->histograms[i], out->histograms[0], 1.e38);
     int k;
     for (k = 1; k < out->size; ++k) {
       const double cur_bits =
-          HistogramDistance(in->histograms[i], out->histograms[k]);
+          HistogramDistance(in->histograms[i], out->histograms[k], best_bits);
       if (cur_bits < best_bits) {
         best_bits = cur_bits;
         best_out = k;
@@ -372,7 +474,7 @@ static void HistogramRemap(const VP8LHistogramSet* const in,
     HistogramClear(out->histograms[i]);
   }
   for (i = 0; i < in->size; ++i) {
-    VP8LHistogramAdd(out->histograms[symbols[i]], in->histograms[i]);
+    HistogramAdd(in->histograms[i], out->histograms[symbols[i]]);
   }
 }
 
@@ -384,8 +486,13 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize,
   int ok = 0;
   const int histo_xsize = histo_bits ? VP8LSubSampleSize(xsize, histo_bits) : 1;
   const int histo_ysize = histo_bits ? VP8LSubSampleSize(ysize, histo_bits) : 1;
-  const int num_histo_pairs = 10 + quality / 2;  // For HistogramCombine().
   const int histo_image_raw_size = histo_xsize * histo_ysize;
+
+  // Heuristic params for HistogramCombine().
+  const int num_tries_no_success = 8 + (quality >> 1);
+  const int iter_mult = (quality < 27) ? 1 : 1 + ((quality - 27) >> 4);
+  const int num_pairs = (quality < 25) ? 10 : (5 * quality) >> 3;
+
   VP8LHistogramSet* const image_out =
       VP8LAllocateHistogramSet(histo_image_raw_size, cache_bits);
   if (image_out == NULL) return 0;
@@ -393,7 +500,8 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize,
   // Build histogram image.
   HistogramBuildImage(xsize, histo_bits, refs, image_out);
   // Collapse similar histograms.
-  if (!HistogramCombine(image_out, image_in, num_histo_pairs)) {
+  if (!HistogramCombine(image_out, image_in, iter_mult, num_pairs,
+                        num_tries_no_success)) {
     goto Error;
   }
   // Find the optimal map from original histograms to the final ones.
diff --git a/3rdparty/libwebp/enc/histogram.h b/3rdparty/libwebp/enc/histogram.h
index ec573c5c85..583b5a4f80 100644
--- a/3rdparty/libwebp/enc/histogram.h
+++ b/3rdparty/libwebp/enc/histogram.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
@@ -80,22 +82,6 @@ double VP8LHistogramEstimateBits(const VP8LHistogram* const p);
 // represent the entropy code itself.
 double VP8LHistogramEstimateBitsBulk(const VP8LHistogram* const p);
 
-static WEBP_INLINE void VP8LHistogramAdd(VP8LHistogram* const p,
-                                         const VP8LHistogram* const a) {
-  int i;
-  for (i = 0; i < PIX_OR_COPY_CODES_MAX; ++i) {
-    p->literal_[i] += a->literal_[i];
-  }
-  for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
-    p->distance_[i] += a->distance_[i];
-  }
-  for (i = 0; i < 256; ++i) {
-    p->red_[i] += a->red_[i];
-    p->blue_[i] += a->blue_[i];
-    p->alpha_[i] += a->alpha_[i];
-  }
-}
-
 static WEBP_INLINE int VP8LHistogramNumCodes(const VP8LHistogram* const p) {
   return 256 + NUM_LENGTH_CODES +
       ((p->palette_code_bits_ > 0) ? (1 << p->palette_code_bits_) : 0);
diff --git a/3rdparty/libwebp/enc/iterator.c b/3rdparty/libwebp/enc/iterator.c
index 86e473bcf0..0746659046 100644
--- a/3rdparty/libwebp/enc/iterator.c
+++ b/3rdparty/libwebp/enc/iterator.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // VP8Iterator: block iterator
diff --git a/3rdparty/libwebp/enc/layer.c b/3rdparty/libwebp/enc/layer.c
index 423127df63..fa89660994 100644
--- a/3rdparty/libwebp/enc/layer.c
+++ b/3rdparty/libwebp/enc/layer.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Enhancement layer (for YUV444/422)
diff --git a/3rdparty/libwebp/enc/picture.c b/3rdparty/libwebp/enc/picture.c
index 44eed06083..5aaa385d47 100644
--- a/3rdparty/libwebp/enc/picture.c
+++ b/3rdparty/libwebp/enc/picture.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // WebPPicture utils: colorspace conversion, crop, ...
@@ -290,8 +292,11 @@ int WebPPictureView(const WebPPicture* src,
     dst->y = src->y + top * src->y_stride + left;
     dst->u = src->u + (top >> 1) * src->uv_stride + (left >> 1);
     dst->v = src->v + (top >> 1) * src->uv_stride + (left >> 1);
+    dst->y_stride = src->y_stride;
+    dst->uv_stride = src->uv_stride;
     if (src->a != NULL) {
       dst->a = src->a + top * src->a_stride + left;
+      dst->a_stride = src->a_stride;
     }
 #ifdef WEBP_EXPERIMENTAL_FEATURES
     if (src->u0 != NULL) {
@@ -299,10 +304,12 @@ int WebPPictureView(const WebPPicture* src,
           IS_YUV_CSP(dst->colorspace, WEBP_YUV422) ? (left >> 1) : left;
       dst->u0 = src->u0 + top * src->uv0_stride + left_pos;
       dst->v0 = src->v0 + top * src->uv0_stride + left_pos;
+      dst->uv0_stride = src->uv0_stride;
     }
 #endif
   } else {
     dst->argb = src->argb + top * src->argb_stride + left;
+    dst->argb_stride = src->argb_stride;
   }
   return 1;
 }
@@ -457,7 +464,6 @@ int WebPPictureRescale(WebPPicture* pic, int width, int height) {
                  (uint8_t*)tmp.argb, width, height,
                  tmp.argb_stride * 4,
                  work, 4);
-
   }
   WebPPictureFree(pic);
   free(work);
@@ -705,7 +711,7 @@ static int Import(WebPPicture* const picture,
     for (y = 0; y < height; ++y) {
       for (x = 0; x < width; ++x) {
         const int offset = step * x + y * rgb_stride;
-        const uint32_t argb = (a_ptr[offset] << 24) |
+        const uint32_t argb = ((uint32_t)a_ptr[offset] << 24) |
                               (r_ptr[offset] << 16) |
                               (g_ptr[offset] <<  8) |
                               (b_ptr[offset]);
@@ -801,11 +807,11 @@ int WebPPictureYUVAToARGB(WebPPicture* picture) {
     // Insert alpha values if needed, in replacement for the default 0xff ones.
     if (picture->colorspace & WEBP_CSP_ALPHA_BIT) {
       for (y = 0; y < height; ++y) {
-        uint32_t* const dst = picture->argb + y * picture->argb_stride;
+        uint32_t* const argb_dst = picture->argb + y * picture->argb_stride;
         const uint8_t* const src = picture->a + y * picture->a_stride;
         int x;
         for (x = 0; x < width; ++x) {
-          dst[x] = (dst[x] & 0x00ffffffu) | (src[x] << 24);
+          argb_dst[x] = (argb_dst[x] & 0x00ffffffu) | ((uint32_t)src[x] << 24);
         }
       }
     }
@@ -906,67 +912,135 @@ void WebPCleanupTransparentArea(WebPPicture* pic) {
 #undef SIZE
 #undef SIZE2
 
+//------------------------------------------------------------------------------
+// local-min distortion
+//
+// For every pixel in the *reference* picture, we search for the local best
+// match in the compressed image. This is not a symmetrical measure.
+
+// search radius. Shouldn't be too large.
+#define RADIUS 2
+
+static float AccumulateLSIM(const uint8_t* src, int src_stride,
+                            const uint8_t* ref, int ref_stride,
+                            int w, int h) {
+  int x, y;
+  double total_sse = 0.;
+  for (y = 0; y < h; ++y) {
+    const int y_0 = (y - RADIUS < 0) ? 0 : y - RADIUS;
+    const int y_1 = (y + RADIUS + 1 >= h) ? h : y + RADIUS + 1;
+    for (x = 0; x < w; ++x) {
+      const int x_0 = (x - RADIUS < 0) ? 0 : x - RADIUS;
+      const int x_1 = (x + RADIUS + 1 >= w) ? w : x + RADIUS + 1;
+      double best_sse = 255. * 255.;
+      const double value = (double)ref[y * ref_stride + x];
+      int i, j;
+      for (j = y_0; j < y_1; ++j) {
+        const uint8_t* s = src + j * src_stride;
+        for (i = x_0; i < x_1; ++i) {
+          const double sse = (double)(s[i] - value) * (s[i] - value);
+          if (sse < best_sse) best_sse = sse;
+        }
+      }
+      total_sse += best_sse;
+    }
+  }
+  return (float)total_sse;
+}
+#undef RADIUS
 
 //------------------------------------------------------------------------------
 // Distortion
 
 // Max value returned in case of exact similarity.
 static const double kMinDistortion_dB = 99.;
+static float GetPSNR(const double v) {
+  return (float)((v > 0.) ? -4.3429448 * log(v / (255 * 255.))
+                          : kMinDistortion_dB);
+}
 
-int WebPPictureDistortion(const WebPPicture* pic1, const WebPPicture* pic2,
+int WebPPictureDistortion(const WebPPicture* src, const WebPPicture* ref,
                           int type, float result[5]) {
-  int c;
   DistoStats stats[5];
   int has_alpha;
+  int uv_w, uv_h;
 
-  if (pic1 == NULL || pic2 == NULL ||
-      pic1->width != pic2->width || pic1->height != pic2->height ||
-      pic1->y == NULL || pic2->y == NULL ||
-      pic1->u == NULL || pic2->u == NULL ||
-      pic1->v == NULL || pic2->v == NULL ||
+  if (src == NULL || ref == NULL ||
+      src->width != ref->width || src->height != ref->height ||
+      src->y == NULL || ref->y == NULL ||
+      src->u == NULL || ref->u == NULL ||
+      src->v == NULL || ref->v == NULL ||
       result == NULL) {
     return 0;
   }
   // TODO(skal): provide distortion for ARGB too.
-  if (pic1->use_argb == 1 || pic1->use_argb != pic2->use_argb) {
+  if (src->use_argb == 1 || src->use_argb != ref->use_argb) {
     return 0;
   }
 
-  has_alpha = !!(pic1->colorspace & WEBP_CSP_ALPHA_BIT);
-  if (has_alpha != !!(pic2->colorspace & WEBP_CSP_ALPHA_BIT) ||
-      (has_alpha && (pic1->a == NULL || pic2->a == NULL))) {
+  has_alpha = !!(src->colorspace & WEBP_CSP_ALPHA_BIT);
+  if (has_alpha != !!(ref->colorspace & WEBP_CSP_ALPHA_BIT) ||
+      (has_alpha && (src->a == NULL || ref->a == NULL))) {
     return 0;
   }
 
   memset(stats, 0, sizeof(stats));
-  VP8SSIMAccumulatePlane(pic1->y, pic1->y_stride,
-                         pic2->y, pic2->y_stride,
-                         pic1->width, pic1->height, &stats[0]);
-  VP8SSIMAccumulatePlane(pic1->u, pic1->uv_stride,
-                         pic2->u, pic2->uv_stride,
-                         (pic1->width + 1) >> 1, (pic1->height + 1) >> 1,
-                         &stats[1]);
-  VP8SSIMAccumulatePlane(pic1->v, pic1->uv_stride,
-                         pic2->v, pic2->uv_stride,
-                         (pic1->width + 1) >> 1, (pic1->height + 1) >> 1,
-                         &stats[2]);
-  if (has_alpha) {
-    VP8SSIMAccumulatePlane(pic1->a, pic1->a_stride,
-                           pic2->a, pic2->a_stride,
-                           pic1->width, pic1->height, &stats[3]);
-  }
-  for (c = 0; c <= 4; ++c) {
-    if (type == 1) {
-      const double v = VP8SSIMGet(&stats[c]);
-      result[c] = (float)((v < 1.) ? -10.0 * log10(1. - v)
-                                   : kMinDistortion_dB);
-    } else {
-      const double v = VP8SSIMGetSquaredError(&stats[c]);
-      result[c] = (float)((v > 0.) ? -4.3429448 * log(v / (255 * 255.))
-                                   : kMinDistortion_dB);
+
+  uv_w = HALVE(src->width);
+  uv_h = HALVE(src->height);
+  if (type >= 2) {
+    float sse[4];
+    sse[0] = AccumulateLSIM(src->y, src->y_stride,
+                            ref->y, ref->y_stride, src->width, src->height);
+    sse[1] = AccumulateLSIM(src->u, src->uv_stride,
+                            ref->u, ref->uv_stride, uv_w, uv_h);
+    sse[2] = AccumulateLSIM(src->v, src->uv_stride,
+                            ref->v, ref->uv_stride, uv_w, uv_h);
+    sse[3] = has_alpha ? AccumulateLSIM(src->a, src->a_stride,
+                                        ref->a, ref->a_stride,
+                                        src->width, src->height)
+                       : 0.f;
+    result[0] = GetPSNR(sse[0] / (src->width * src->height));
+    result[1] = GetPSNR(sse[1] / (uv_w * uv_h));
+    result[2] = GetPSNR(sse[2] / (uv_w * uv_h));
+    result[3] = GetPSNR(sse[3] / (src->width * src->height));
+    {
+      double total_sse = sse[0] + sse[1] + sse[2];
+      int total_pixels = src->width * src->height + 2 * uv_w * uv_h;
+      if (has_alpha) {
+        total_pixels += src->width * src->height;
+        total_sse += sse[3];
+      }
+      result[4] = GetPSNR(total_sse / total_pixels);
+    }
+  } else {
+    int c;
+    VP8SSIMAccumulatePlane(src->y, src->y_stride,
+                           ref->y, ref->y_stride,
+                           src->width, src->height, &stats[0]);
+    VP8SSIMAccumulatePlane(src->u, src->uv_stride,
+                           ref->u, ref->uv_stride,
+                           uv_w, uv_h, &stats[1]);
+    VP8SSIMAccumulatePlane(src->v, src->uv_stride,
+                           ref->v, ref->uv_stride,
+                           uv_w, uv_h, &stats[2]);
+    if (has_alpha) {
+      VP8SSIMAccumulatePlane(src->a, src->a_stride,
+                             ref->a, ref->a_stride,
+                             src->width, src->height, &stats[3]);
+    }
+    for (c = 0; c <= 4; ++c) {
+      if (type == 1) {
+        const double v = VP8SSIMGet(&stats[c]);
+        result[c] = (float)((v < 1.) ? -10.0 * log10(1. - v)
+                                     : kMinDistortion_dB);
+      } else {
+        const double v = VP8SSIMGetSquaredError(&stats[c]);
+        result[c] = GetPSNR(v);
+      }
+      // Accumulate forward
+      if (c < 4) VP8SSIMAddStats(&stats[c], &stats[4]);
     }
-    // Accumulate forward
-    if (c < 4) VP8SSIMAddStats(&stats[c], &stats[4]);
   }
   return 1;
 }
diff --git a/3rdparty/libwebp/enc/quant.c b/3rdparty/libwebp/enc/quant.c
index ea153849c8..462d4e9e6e 100644
--- a/3rdparty/libwebp/enc/quant.c
+++ b/3rdparty/libwebp/enc/quant.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //   Quantization
@@ -27,6 +29,8 @@
 #define SNS_TO_DQ 0.9     // Scaling constant between the sns value and the QP
                           // power-law modulation. Must be strictly less than 1.
 
+#define I4_PENALTY 4000   // Rate-penalty for quick i4/i16 decision
+
 #define MULT_8B(a, b) (((a) * (b) + 128) >> 8)
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -224,28 +228,90 @@ static void SetupFilterStrength(VP8Encoder* const enc) {
 // We want to emulate jpeg-like behaviour where the expected "good" quality
 // is around q=75. Internally, our "good" middle is around c=50. So we
 // map accordingly using linear piece-wise function
-static double QualityToCompression(double q) {
-  const double c = q / 100.;
-  return (c < 0.75) ? c * (2. / 3.) : 2. * c - 1.;
+static double QualityToCompression(double c) {
+  const double linear_c = (c < 0.75) ? c * (2. / 3.) : 2. * c - 1.;
+  // The file size roughly scales as pow(quantizer, 3.). Actually, the
+  // exponent is somewhere between 2.8 and 3.2, but we're mostly interested
+  // in the mid-quant range. So we scale the compressibility inversely to
+  // this power-law: quant ~= compression ^ 1/3. This law holds well for
+  // low quant. Finer modelling for high-quant would make use of kAcTable[]
+  // more explicitly.
+  const double v = pow(linear_c, 1 / 3.);
+  return v;
+}
+
+static double QualityToJPEGCompression(double c, double alpha) {
+  // We map the complexity 'alpha' and quality setting 'c' to a compression
+  // exponent empirically matched to the compression curve of libjpeg6b.
+  // On average, the WebP output size will be roughly similar to that of a
+  // JPEG file compressed with same quality factor.
+  const double amin = 0.30;
+  const double amax = 0.85;
+  const double exp_min = 0.4;
+  const double exp_max = 0.9;
+  const double slope = (exp_min - exp_max) / (amax - amin);
+  // Linearly interpolate 'expn' from exp_min to exp_max
+  // in the [amin, amax] range.
+  const double expn = (alpha > amax) ? exp_min
+                    : (alpha < amin) ? exp_max
+                    : exp_max + slope * (alpha - amin);
+  const double v = pow(c, expn);
+  return v;
+}
+
+static int SegmentsAreEquivalent(const VP8SegmentInfo* const S1,
+                                 const VP8SegmentInfo* const S2) {
+  return (S1->quant_ == S2->quant_) && (S1->fstrength_ == S2->fstrength_);
+}
+
+static void SimplifySegments(VP8Encoder* const enc) {
+  int map[NUM_MB_SEGMENTS] = { 0, 1, 2, 3 };
+  const int num_segments = enc->segment_hdr_.num_segments_;
+  int num_final_segments = 1;
+  int s1, s2;
+  for (s1 = 1; s1 < num_segments; ++s1) {    // find similar segments
+    const VP8SegmentInfo* const S1 = &enc->dqm_[s1];
+    int found = 0;
+    // check if we already have similar segment
+    for (s2 = 0; s2 < num_final_segments; ++s2) {
+      const VP8SegmentInfo* const S2 = &enc->dqm_[s2];
+      if (SegmentsAreEquivalent(S1, S2)) {
+        found = 1;
+        break;
+      }
+    }
+    map[s1] = s2;
+    if (!found) {
+      if (num_final_segments != s1) {
+        enc->dqm_[num_final_segments] = enc->dqm_[s1];
+      }
+      ++num_final_segments;
+    }
+  }
+  if (num_final_segments < num_segments) {  // Remap
+    int i = enc->mb_w_ * enc->mb_h_;
+    while (i-- > 0) enc->mb_info_[i].segment_ = map[enc->mb_info_[i].segment_];
+    enc->segment_hdr_.num_segments_ = num_final_segments;
+    // Replicate the trailing segment infos (it's mostly cosmetics)
+    for (i = num_final_segments; i < num_segments; ++i) {
+      enc->dqm_[i] = enc->dqm_[num_final_segments - 1];
+    }
+  }
 }
 
 void VP8SetSegmentParams(VP8Encoder* const enc, float quality) {
   int i;
   int dq_uv_ac, dq_uv_dc;
-  const int num_segments = enc->config_->segments;
+  const int num_segments = enc->segment_hdr_.num_segments_;
   const double amp = SNS_TO_DQ * enc->config_->sns_strength / 100. / 128.;
-  const double c_base = QualityToCompression(quality);
+  const double Q = quality / 100.;
+  const double c_base = enc->config_->emulate_jpeg_size ?
+      QualityToJPEGCompression(Q, enc->alpha_ / 255.) :
+      QualityToCompression(Q);
   for (i = 0; i < num_segments; ++i) {
-    // The file size roughly scales as pow(quantizer, 3.). Actually, the
-    // exponent is somewhere between 2.8 and 3.2, but we're mostly interested
-    // in the mid-quant range. So we scale the compressibility inversely to
-    // this power-law: quant ~= compression ^ 1/3. This law holds well for
-    // low quant. Finer modelling for high-quant would make use of kAcTable[]
-    // more explicitely.
-    // Additionally, we modulate the base exponent 1/3 to accommodate for the
-    // quantization susceptibility and allow denser segments to be quantized
-    // more.
-    const double expn = (1. - amp * enc->dqm_[i].alpha_) / 3.;
+    // We modulate the base coefficient to accommodate for the quantization
+    // susceptibility and allow denser segments to be quantized more.
+    const double expn = 1. - amp * enc->dqm_[i].alpha_;
     const double c = pow(c_base, expn);
     const int q = (int)(127. * (1. - c));
     assert(expn > 0.);
@@ -281,9 +347,11 @@ void VP8SetSegmentParams(VP8Encoder* const enc, float quality) {
   enc->dq_uv_dc_ = dq_uv_dc;
   enc->dq_uv_ac_ = dq_uv_ac;
 
-  SetupMatrices(enc);
-
   SetupFilterStrength(enc);   // initialize segments' filtering, eventually
+
+  if (num_segments > 1) SimplifySegments(enc);
+
+  SetupMatrices(enc);         // finalize quantization matrices
 }
 
 //------------------------------------------------------------------------------
@@ -709,7 +777,7 @@ static void PickBestIntra16(VP8EncIterator* const it, VP8ModeScore* const rd) {
   int mode;
 
   rd->mode_i16 = -1;
-  for (mode = 0; mode < 4; ++mode) {
+  for (mode = 0; mode < NUM_PRED_MODES; ++mode) {
     uint8_t* const tmp_dst = it->yuv_out2_ + Y_OFF;  // scratch buffer
     int nz;
 
@@ -838,7 +906,7 @@ static void PickBestUV(VP8EncIterator* const it, VP8ModeScore* const rd) {
 
   rd->mode_uv = -1;
   InitScore(&rd_best);
-  for (mode = 0; mode < 4; ++mode) {
+  for (mode = 0; mode < NUM_PRED_MODES; ++mode) {
     VP8ModeScore rd_uv;
 
     // Reconstruct
@@ -867,10 +935,10 @@ static void PickBestUV(VP8EncIterator* const it, VP8ModeScore* const rd) {
 
 static void SimpleQuantize(VP8EncIterator* const it, VP8ModeScore* const rd) {
   const VP8Encoder* const enc = it->enc_;
-  const int i16 = (it->mb_->type_ == 1);
+  const int is_i16 = (it->mb_->type_ == 1);
   int nz = 0;
 
-  if (i16) {
+  if (is_i16) {
     nz = ReconstructIntra16(it, rd, it->yuv_out_ + Y_OFF, it->preds_[0]);
   } else {
     VP8IteratorStartI4(it);
@@ -889,11 +957,66 @@ static void SimpleQuantize(VP8EncIterator* const it, VP8ModeScore* const rd) {
   rd->nz = nz;
 }
 
+// Refine intra16/intra4 sub-modes based on distortion only (not rate).
+static void DistoRefine(VP8EncIterator* const it, int try_both_i4_i16) {
+  const int is_i16 = (it->mb_->type_ == 1);
+  score_t best_score = MAX_COST;
+
+  if (try_both_i4_i16 || is_i16) {
+    int mode;
+    int best_mode = -1;
+    for (mode = 0; mode < NUM_PRED_MODES; ++mode) {
+      const uint8_t* const ref = it->yuv_p_ + VP8I16ModeOffsets[mode];
+      const uint8_t* const src = it->yuv_in_ + Y_OFF;
+      const score_t score = VP8SSE16x16(src, ref);
+      if (score < best_score) {
+        best_mode = mode;
+        best_score = score;
+      }
+    }
+    VP8SetIntra16Mode(it, best_mode);
+  }
+  if (try_both_i4_i16 || !is_i16) {
+    uint8_t modes_i4[16];
+    // We don't evaluate the rate here, but just account for it through a
+    // constant penalty (i4 mode usually needs more bits compared to i16).
+    score_t score_i4 = (score_t)I4_PENALTY;
+
+    VP8IteratorStartI4(it);
+    do {
+      int mode;
+      int best_sub_mode = -1;
+      score_t best_sub_score = MAX_COST;
+      const uint8_t* const src = it->yuv_in_ + Y_OFF + VP8Scan[it->i4_];
+
+      // TODO(skal): we don't really need the prediction pixels here,
+      // but just the distortion against 'src'.
+      VP8MakeIntra4Preds(it);
+      for (mode = 0; mode < NUM_BMODES; ++mode) {
+        const uint8_t* const ref = it->yuv_p_ + VP8I4ModeOffsets[mode];
+        const score_t score = VP8SSE4x4(src, ref);
+        if (score < best_sub_score) {
+          best_sub_mode = mode;
+          best_sub_score = score;
+        }
+      }
+      modes_i4[it->i4_] = best_sub_mode;
+      score_i4 += best_sub_score;
+      if (score_i4 >= best_score) break;
+    } while (VP8IteratorRotateI4(it, it->yuv_in_ + Y_OFF));
+    if (score_i4 < best_score) {
+      VP8SetIntra4Mode(it, modes_i4);
+    }
+  }
+}
+
 //------------------------------------------------------------------------------
 // Entry point
 
-int VP8Decimate(VP8EncIterator* const it, VP8ModeScore* const rd, int rd_opt) {
+int VP8Decimate(VP8EncIterator* const it, VP8ModeScore* const rd,
+                VP8RDLevel rd_opt) {
   int is_skipped;
+  const int method = it->enc_->method_;
 
   InitScore(rd);
 
@@ -902,22 +1025,21 @@ int VP8Decimate(VP8EncIterator* const it, VP8ModeScore* const rd, int rd_opt) {
   VP8MakeLuma16Preds(it);
   VP8MakeChroma8Preds(it);
 
-  // for rd_opt = 2, we perform trellis-quant on the final decision only.
-  // for rd_opt > 2, we use it for every scoring (=much slower).
-  if (rd_opt > 0) {
-    it->do_trellis_ = (rd_opt > 2);
+  if (rd_opt > RD_OPT_NONE) {
+    it->do_trellis_ = (rd_opt >= RD_OPT_TRELLIS_ALL);
     PickBestIntra16(it, rd);
-    if (it->enc_->method_ >= 2) {
+    if (method >= 2) {
       PickBestIntra4(it, rd);
     }
     PickBestUV(it, rd);
-    if (rd_opt == 2) {
+    if (rd_opt == RD_OPT_TRELLIS) {   // finish off with trellis-optim now
       it->do_trellis_ = 1;
       SimpleQuantize(it, rd);
     }
   } else {
-    // TODO: for method_ == 2, pick the best intra4/intra16 based on SSE
-    it->do_trellis_ = (it->enc_->method_ == 2);
+    // For method == 2, pick the best intra4/intra16 based on SSE (~tad slower).
+    // For method <= 1, we refine intra4 or intra16 (but don't re-examine mode).
+    DistoRefine(it, (method >= 2));
     SimpleQuantize(it, rd);
   }
   is_skipped = (rd->nz == 0);
diff --git a/3rdparty/libwebp/enc/syntax.c b/3rdparty/libwebp/enc/syntax.c
index 7c8c7b1a84..b0f7676b48 100644
--- a/3rdparty/libwebp/enc/syntax.c
+++ b/3rdparty/libwebp/enc/syntax.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Header syntax writing
@@ -11,7 +13,9 @@
 
 #include <assert.h>
 
-#include "../webp/format_constants.h"
+#include "../utils/utils.h"
+#include "../webp/format_constants.h"  // RIFF constants
+#include "../webp/mux_types.h"         // ALPHA_FLAG
 #include "./vp8enci.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -21,25 +25,12 @@ extern "C" {
 //------------------------------------------------------------------------------
 // Helper functions
 
-// TODO(later): Move to webp/format_constants.h?
-static void PutLE24(uint8_t* const data, uint32_t val) {
-  data[0] = (val >>  0) & 0xff;
-  data[1] = (val >>  8) & 0xff;
-  data[2] = (val >> 16) & 0xff;
-}
-
-static void PutLE32(uint8_t* const data, uint32_t val) {
-  PutLE24(data, val);
-  data[3] = (val >> 24) & 0xff;
-}
-
 static int IsVP8XNeeded(const VP8Encoder* const enc) {
   return !!enc->has_alpha_;  // Currently the only case when VP8X is needed.
                              // This could change in the future.
 }
 
 static int PutPaddingByte(const WebPPicture* const pic) {
-
   const uint8_t pad_byte[1] = { 0 };
   return !!pic->writer(pad_byte, 1, pic);
 }
@@ -73,14 +64,14 @@ static WebPEncodingError PutVP8XHeader(const VP8Encoder* const enc) {
   assert(pic->width <= MAX_CANVAS_SIZE && pic->height <= MAX_CANVAS_SIZE);
 
   if (enc->has_alpha_) {
-    flags |= ALPHA_FLAG_BIT;
+    flags |= ALPHA_FLAG;
   }
 
   PutLE32(vp8x + TAG_SIZE,              VP8X_CHUNK_SIZE);
   PutLE32(vp8x + CHUNK_HEADER_SIZE,     flags);
   PutLE24(vp8x + CHUNK_HEADER_SIZE + 4, pic->width - 1);
   PutLE24(vp8x + CHUNK_HEADER_SIZE + 7, pic->height - 1);
-  if(!pic->writer(vp8x, sizeof(vp8x), pic)) {
+  if (!pic->writer(vp8x, sizeof(vp8x), pic)) {
     return VP8_ENC_ERROR_BAD_WRITE;
   }
   return VP8_ENC_OK;
@@ -327,7 +318,9 @@ static size_t GeneratePartition0(VP8Encoder* const enc) {
 
   PutSegmentHeader(bw, enc);
   PutFilterHeader(bw, &enc->filter_hdr_);
-  VP8PutValue(bw, enc->config_->partitions, 2);
+  VP8PutValue(bw, enc->num_parts_ == 8 ? 3 :
+                  enc->num_parts_ == 4 ? 2 :
+                  enc->num_parts_ == 2 ? 1 : 0, 2);
   PutQuant(bw, enc);
   VP8PutBitUniform(bw, 0);   // no proba update
   VP8WriteProbas(bw, &enc->proba_);
diff --git a/3rdparty/libwebp/enc/token.c b/3rdparty/libwebp/enc/token.c
new file mode 100644
index 0000000000..6a63371f71
--- /dev/null
+++ b/3rdparty/libwebp/enc/token.c
@@ -0,0 +1,256 @@
+// Copyright 2011 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Paginated token buffer
+//
+//  A 'token' is a bit value associated with a probability, either fixed
+// or a later-to-be-determined after statistics have been collected.
+// For dynamic probability, we just record the slot id (idx) for the probability
+// value in the final probability array (uint8_t* probas in VP8EmitTokens).
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include <assert.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "./vp8enci.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+#if !defined(DISABLE_TOKEN_BUFFER)
+
+// we use pages to reduce the number of memcpy()
+#define MAX_NUM_TOKEN 8192          // max number of token per page
+#define FIXED_PROBA_BIT (1u << 14)
+
+struct VP8Tokens {
+  uint16_t tokens_[MAX_NUM_TOKEN];  // bit#15: bit
+                                    // bit #14: constant proba or idx
+                                    // bits 0..13: slot or constant proba
+  VP8Tokens* next_;
+};
+
+//------------------------------------------------------------------------------
+
+void VP8TBufferInit(VP8TBuffer* const b) {
+  b->tokens_ = NULL;
+  b->pages_ = NULL;
+  b->last_page_ = &b->pages_;
+  b->left_ = 0;
+  b->error_ = 0;
+}
+
+void VP8TBufferClear(VP8TBuffer* const b) {
+  if (b != NULL) {
+    const VP8Tokens* p = b->pages_;
+    while (p != NULL) {
+      const VP8Tokens* const next = p->next_;
+      free((void*)p);
+      p = next;
+    }
+    VP8TBufferInit(b);
+  }
+}
+
+static int TBufferNewPage(VP8TBuffer* const b) {
+  VP8Tokens* const page = b->error_ ? NULL : (VP8Tokens*)malloc(sizeof(*page));
+  if (page == NULL) {
+    b->error_ = 1;
+    return 0;
+  }
+  *b->last_page_ = page;
+  b->last_page_ = &page->next_;
+  b->left_ = MAX_NUM_TOKEN;
+  b->tokens_ = page->tokens_;
+  page->next_ = NULL;
+  return 1;
+}
+
+//------------------------------------------------------------------------------
+
+#define TOKEN_ID(t, b, ctx, p) \
+    ((p) + NUM_PROBAS * ((ctx) + NUM_CTX * ((b) + NUM_BANDS * (t))))
+
+static WEBP_INLINE int AddToken(VP8TBuffer* const b,
+                                int bit, uint32_t proba_idx) {
+  assert(proba_idx < FIXED_PROBA_BIT);
+  assert(bit == 0 || bit == 1);
+  if (b->left_ > 0 || TBufferNewPage(b)) {
+    const int slot = --b->left_;
+    b->tokens_[slot] = (bit << 15) | proba_idx;
+  }
+  return bit;
+}
+
+static WEBP_INLINE void AddConstantToken(VP8TBuffer* const b,
+                                         int bit, int proba) {
+  assert(proba < 256);
+  assert(bit == 0 || bit == 1);
+  if (b->left_ > 0 || TBufferNewPage(b)) {
+    const int slot = --b->left_;
+    b->tokens_[slot] = (bit << 15) | FIXED_PROBA_BIT | proba;
+  }
+}
+
+int VP8RecordCoeffTokens(int ctx, int coeff_type, int first, int last,
+                         const int16_t* const coeffs,
+                         VP8TBuffer* const tokens) {
+  int n = first;
+  uint32_t base_id = TOKEN_ID(coeff_type, n, ctx, 0);
+  if (!AddToken(tokens, last >= 0, base_id + 0)) {
+    return 0;
+  }
+
+  while (n < 16) {
+    const int c = coeffs[n++];
+    const int sign = c < 0;
+    int v = sign ? -c : c;
+    if (!AddToken(tokens, v != 0, base_id + 1)) {
+      ctx = 0;
+      base_id = TOKEN_ID(coeff_type, VP8EncBands[n], ctx, 0);
+      continue;
+    }
+    if (!AddToken(tokens, v > 1, base_id + 2)) {
+      ctx = 1;
+    } else {
+      if (!AddToken(tokens, v > 4, base_id + 3)) {
+        if (AddToken(tokens, v != 2, base_id + 4))
+          AddToken(tokens, v == 4, base_id + 5);
+      } else if (!AddToken(tokens, v > 10, base_id + 6)) {
+        if (!AddToken(tokens, v > 6, base_id + 7)) {
+          AddConstantToken(tokens, v == 6, 159);
+        } else {
+          AddConstantToken(tokens, v >= 9, 165);
+          AddConstantToken(tokens, !(v & 1), 145);
+        }
+      } else {
+        int mask;
+        const uint8_t* tab;
+        if (v < 3 + (8 << 1)) {          // VP8Cat3  (3b)
+          AddToken(tokens, 0, base_id + 8);
+          AddToken(tokens, 0, base_id + 9);
+          v -= 3 + (8 << 0);
+          mask = 1 << 2;
+          tab = VP8Cat3;
+        } else if (v < 3 + (8 << 2)) {   // VP8Cat4  (4b)
+          AddToken(tokens, 0, base_id + 8);
+          AddToken(tokens, 1, base_id + 9);
+          v -= 3 + (8 << 1);
+          mask = 1 << 3;
+          tab = VP8Cat4;
+        } else if (v < 3 + (8 << 3)) {   // VP8Cat5  (5b)
+          AddToken(tokens, 1, base_id + 8);
+          AddToken(tokens, 0, base_id + 10);
+          v -= 3 + (8 << 2);
+          mask = 1 << 4;
+          tab = VP8Cat5;
+        } else {                         // VP8Cat6 (11b)
+          AddToken(tokens, 1, base_id + 8);
+          AddToken(tokens, 1, base_id + 10);
+          v -= 3 + (8 << 3);
+          mask = 1 << 10;
+          tab = VP8Cat6;
+        }
+        while (mask) {
+          AddConstantToken(tokens, !!(v & mask), *tab++);
+          mask >>= 1;
+        }
+      }
+      ctx = 2;
+    }
+    AddConstantToken(tokens, sign, 128);
+    base_id = TOKEN_ID(coeff_type, VP8EncBands[n], ctx, 0);
+    if (n == 16 || !AddToken(tokens, n <= last, base_id + 0)) {
+      return 1;   // EOB
+    }
+  }
+  return 1;
+}
+
+#undef TOKEN_ID
+
+//------------------------------------------------------------------------------
+// This function works, but isn't currently used. Saved for later.
+
+#if 0
+
+static void Record(int bit, proba_t* const stats) {
+  proba_t p = *stats;
+  if (p >= 0xffff0000u) {               // an overflow is inbound.
+    p = ((p + 1u) >> 1) & 0x7fff7fffu;  // -> divide the stats by 2.
+  }
+  // record bit count (lower 16 bits) and increment total count (upper 16 bits).
+  p += 0x00010000u + bit;
+  *stats = p;
+}
+
+void VP8TokenToStats(const VP8TBuffer* const b, proba_t* const stats) {
+  const VP8Tokens* p = b->pages_;
+  while (p != NULL) {
+    const int N = (p->next_ == NULL) ? b->left_ : 0;
+    int n = MAX_NUM_TOKEN;
+    while (n-- > N) {
+      const uint16_t token = p->tokens_[n];
+      if (!(token & FIXED_PROBA_BIT)) {
+        Record((token >> 15) & 1, stats + (token & 0x3fffu));
+      }
+    }
+    p = p->next_;
+  }
+}
+
+#endif   // 0
+
+//------------------------------------------------------------------------------
+// Final coding pass, with known probabilities
+
+int VP8EmitTokens(VP8TBuffer* const b, VP8BitWriter* const bw,
+                  const uint8_t* const probas, int final_pass) {
+  const VP8Tokens* p = b->pages_;
+  (void)final_pass;
+  if (b->error_) return 0;
+  while (p != NULL) {
+    const VP8Tokens* const next = p->next_;
+    const int N = (next == NULL) ? b->left_ : 0;
+    int n = MAX_NUM_TOKEN;
+    while (n-- > N) {
+      const uint16_t token = p->tokens_[n];
+      const int bit = (token >> 15) & 1;
+      if (token & FIXED_PROBA_BIT) {
+        VP8PutBit(bw, bit, token & 0xffu);  // constant proba
+      } else {
+        VP8PutBit(bw, bit, probas[token & 0x3fffu]);
+      }
+    }
+    if (final_pass) free((void*)p);
+    p = next;
+  }
+  if (final_pass) b->pages_ = NULL;
+  return 1;
+}
+
+//------------------------------------------------------------------------------
+
+#else     // DISABLE_TOKEN_BUFFER
+
+void VP8TBufferInit(VP8TBuffer* const b) {
+  (void)b;
+}
+void VP8TBufferClear(VP8TBuffer* const b) {
+  (void)b;
+}
+
+#endif    // !DISABLE_TOKEN_BUFFER
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
diff --git a/3rdparty/libwebp/enc/tree.c b/3rdparty/libwebp/enc/tree.c
index 8b25e5e488..ecd8fb910e 100644
--- a/3rdparty/libwebp/enc/tree.c
+++ b/3rdparty/libwebp/enc/tree.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Token probabilities
diff --git a/3rdparty/libwebp/enc/vp8enci.h b/3rdparty/libwebp/enc/vp8enci.h
index f660eeec55..61d56be554 100644
--- a/3rdparty/libwebp/enc/vp8enci.h
+++ b/3rdparty/libwebp/enc/vp8enci.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //   WebP encoder: internal header.
@@ -16,6 +18,7 @@
 #include "../webp/encode.h"
 #include "../dsp/dsp.h"
 #include "../utils/bit_writer.h"
+#include "../utils/thread.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -26,12 +29,9 @@ extern "C" {
 
 // version numbers
 #define ENC_MAJ_VERSION 0
-#define ENC_MIN_VERSION 2
+#define ENC_MIN_VERSION 3
 #define ENC_REV_VERSION 1
 
-// size of histogram used by CollectHistogram.
-#define MAX_COEFF_THRESH   64
-
 // intra prediction modes
 enum { B_DC_PRED = 0,   // 4x4 modes
        B_TM_PRED = 1,
@@ -47,7 +47,8 @@ enum { B_DC_PRED = 0,   // 4x4 modes
 
        // Luma16 or UV modes
        DC_PRED = B_DC_PRED, V_PRED = B_VE_PRED,
-       H_PRED = B_HE_PRED, TM_PRED = B_TM_PRED
+       H_PRED = B_HE_PRED, TM_PRED = B_TM_PRED,
+       NUM_PRED_MODES = 4
      };
 
 enum { NUM_MB_SEGMENTS = 4,
@@ -56,10 +57,18 @@ enum { NUM_MB_SEGMENTS = 4,
        NUM_BANDS = 8,
        NUM_CTX = 3,
        NUM_PROBAS = 11,
-       MAX_LF_LEVELS = 64,      // Maximum loop filter level
-       MAX_VARIABLE_LEVEL = 67  // last (inclusive) level with variable cost
+       MAX_LF_LEVELS = 64,       // Maximum loop filter level
+       MAX_VARIABLE_LEVEL = 67,  // last (inclusive) level with variable cost
+       MAX_LEVEL = 2047          // max level (note: max codable is 2047 + 67)
      };
 
+typedef enum {   // Rate-distortion optimization levels
+  RD_OPT_NONE        = 0,  // no rd-opt
+  RD_OPT_BASIC       = 1,  // basic scoring (no trellis)
+  RD_OPT_TRELLIS     = 2,  // perform trellis-quant on the final decision only
+  RD_OPT_TRELLIS_ALL = 3   // trellis-quant for every scoring (much slower)
+} VP8RDLevel;
+
 // YUV-cache parameters. Cache is 16-pixels wide.
 // The original or reconstructed samples can be accessed using VP8Scan[]
 // The predicted blocks can be accessed using offsets to yuv_p_ and
@@ -160,7 +169,17 @@ typedef int64_t score_t;     // type used for scores, rate, distortion
 static WEBP_INLINE int QUANTDIV(int n, int iQ, int B) {
   return (n * iQ + B) >> QFIX;
 }
-extern const uint8_t VP8Zigzag[16];
+
+// size of histogram used by CollectHistogram.
+#define MAX_COEFF_THRESH   31
+typedef struct VP8Histogram VP8Histogram;
+struct VP8Histogram {
+  // TODO(skal): we only need to store the max_value and last_non_zero actually.
+  int distribution[MAX_COEFF_THRESH + 1];
+};
+
+// Uncomment the following to remove token-buffer code:
+// #define DISABLE_TOKEN_BUFFER
 
 //------------------------------------------------------------------------------
 // Headers
@@ -314,44 +333,37 @@ void VP8SetSegment(const VP8EncIterator* const it, int segment);
 //------------------------------------------------------------------------------
 // Paginated token buffer
 
-// WIP: #define USE_TOKEN_BUFFER
-
-#ifdef USE_TOKEN_BUFFER
-
-#define MAX_NUM_TOKEN 2048
-
-typedef struct VP8Tokens VP8Tokens;
-struct VP8Tokens {
-  uint16_t tokens_[MAX_NUM_TOKEN];  // bit#15: bit, bits 0..14: slot
-  int left_;
-  VP8Tokens* next_;
-};
+typedef struct VP8Tokens VP8Tokens;  // struct details in token.c
 
 typedef struct {
-  VP8Tokens* rows_;
-  uint16_t* tokens_;    // set to (*last_)->tokens_
-  VP8Tokens** last_;
-  int left_;
-  int error_;  // true in case of malloc error
+#if !defined(DISABLE_TOKEN_BUFFER)
+  VP8Tokens* pages_;        // first page
+  VP8Tokens** last_page_;   // last page
+  uint16_t* tokens_;        // set to (*last_page_)->tokens_
+  int left_;          // how many free tokens left before the page is full.
+#endif
+  int error_;         // true in case of malloc error
 } VP8TBuffer;
 
 void VP8TBufferInit(VP8TBuffer* const b);    // initialize an empty buffer
-int VP8TBufferNewPage(VP8TBuffer* const b);  // allocate a new page
-void VP8TBufferClear(VP8TBuffer* const b);   // de-allocate memory
+void VP8TBufferClear(VP8TBuffer* const b);   // de-allocate pages memory
 
-int VP8EmitTokens(const VP8TBuffer* const b, VP8BitWriter* const bw,
-                  const uint8_t* const probas);
+#if !defined(DISABLE_TOKEN_BUFFER)
 
-static WEBP_INLINE int VP8AddToken(VP8TBuffer* const b,
-                                   int bit, int proba_idx) {
-  if (b->left_ > 0 || VP8TBufferNewPage(b)) {
-    const int slot = --b->left_;
-    b->tokens_[slot] = (bit << 15) | proba_idx;
-  }
-  return bit;
-}
+// Finalizes bitstream when probabilities are known.
+// Deletes the allocated token memory if final_pass is true.
+int VP8EmitTokens(VP8TBuffer* const b, VP8BitWriter* const bw,
+                  const uint8_t* const probas, int final_pass);
 
-#endif  // USE_TOKEN_BUFFER
+// record the coding of coefficients without knowing the probabilities yet
+int VP8RecordCoeffTokens(int ctx, int coeff_type, int first, int last,
+                         const int16_t* const coeffs,
+                         VP8TBuffer* const tokens);
+
+// unused for now
+void VP8TokenToStats(const VP8TBuffer* const b, proba_t* const stats);
+
+#endif  // !DISABLE_TOKEN_BUFFER
 
 //------------------------------------------------------------------------------
 // VP8Encoder
@@ -376,6 +388,7 @@ struct VP8Encoder {
   // per-partition boolean decoders.
   VP8BitWriter bw_;                         // part0
   VP8BitWriter parts_[MAX_NUM_PARTITIONS];  // token partitions
+  VP8TBuffer tokens_;                       // token buffer
 
   int percent_;                             // for progress
 
@@ -383,6 +396,7 @@ struct VP8Encoder {
   int has_alpha_;
   uint8_t* alpha_data_;       // non-NULL if transparency is present
   uint32_t alpha_data_size_;
+  WebPWorker alpha_worker_;
 
   // enhancement layer
   int use_layer_;
@@ -394,6 +408,7 @@ struct VP8Encoder {
   VP8SegmentInfo dqm_[NUM_MB_SEGMENTS];
   int base_quant_;                 // nominal quantizer value. Only used
                                    // for relative coding of segments' quant.
+  int alpha_;                      // global susceptibility (<=> complexity)
   int uv_alpha_;                   // U/V quantization susceptibility
   // global offset of quantizers, shared by all segments
   int dq_y1_dc_;
@@ -409,9 +424,12 @@ struct VP8Encoder {
   int      block_count_[3];
 
   // quality/speed settings
-  int method_;              // 0=fastest, 6=best/slowest.
-  int rd_opt_level_;        // Deduced from method_.
-  int max_i4_header_bits_;  // partition #0 safeness factor
+  int method_;               // 0=fastest, 6=best/slowest.
+  VP8RDLevel rd_opt_level_;  // Deduced from method_.
+  int max_i4_header_bits_;   // partition #0 safeness factor
+  int thread_level_;         // derived from config->thread_level
+  int do_search_;            // derived from config->target_XXX
+  int use_tokens_;           // if true, use token buffer
 
   // Memory
   VP8MBInfo* mb_info_;   // contextual macroblock infos (mb_w_ + 1)
@@ -455,6 +473,11 @@ void VP8EncFreeBitWriters(VP8Encoder* const enc);
 
   // in frame.c
 extern const uint8_t VP8EncBands[16 + 1];
+extern const uint8_t VP8Cat3[];
+extern const uint8_t VP8Cat4[];
+extern const uint8_t VP8Cat5[];
+extern const uint8_t VP8Cat6[];
+
 // Form all the four Intra16x16 predictions in the yuv_p_ cache
 void VP8MakeLuma16Preds(const VP8EncIterator* const it);
 // Form all the four Chroma8x8 predictions in the yuv_p_ cache
@@ -466,9 +489,9 @@ void VP8MakeIntra4Preds(const VP8EncIterator* const it);
 int VP8GetCostLuma16(VP8EncIterator* const it, const VP8ModeScore* const rd);
 int VP8GetCostLuma4(VP8EncIterator* const it, const int16_t levels[16]);
 int VP8GetCostUV(VP8EncIterator* const it, const VP8ModeScore* const rd);
-// Main stat / coding passes
+// Main coding calls
 int VP8EncLoop(VP8Encoder* const enc);
-int VP8StatLoop(VP8Encoder* const enc);
+int VP8EncTokenLoop(VP8Encoder* const enc);
 
   // in webpenc.c
 // Assign an error code to a picture. Return false for convenience.
@@ -485,12 +508,14 @@ int VP8EncAnalyze(VP8Encoder* const enc);
 // Sets up segment's quantization values, base_quant_ and filter strengths.
 void VP8SetSegmentParams(VP8Encoder* const enc, float quality);
 // Pick best modes and fills the levels. Returns true if skipped.
-int VP8Decimate(VP8EncIterator* const it, VP8ModeScore* const rd, int rd_opt);
+int VP8Decimate(VP8EncIterator* const it, VP8ModeScore* const rd,
+                VP8RDLevel rd_opt);
 
   // in alpha.c
 void VP8EncInitAlpha(VP8Encoder* const enc);    // initialize alpha compression
+int VP8EncStartAlpha(VP8Encoder* const enc);    // start alpha coding process
 int VP8EncFinishAlpha(VP8Encoder* const enc);   // finalize compressed data
-void VP8EncDeleteAlpha(VP8Encoder* const enc);  // delete compressed data
+int VP8EncDeleteAlpha(VP8Encoder* const enc);   // delete compressed data
 
   // in layer.c
 void VP8EncInitLayer(VP8Encoder* const enc);     // init everything
diff --git a/3rdparty/libwebp/enc/vp8l.c b/3rdparty/libwebp/enc/vp8l.c
index 2abc30f464..945870ca46 100644
--- a/3rdparty/libwebp/enc/vp8l.c
+++ b/3rdparty/libwebp/enc/vp8l.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // main entry for the lossless encoder.
@@ -37,7 +39,8 @@ extern "C" {
 static int CompareColors(const void* p1, const void* p2) {
   const uint32_t a = *(const uint32_t*)p1;
   const uint32_t b = *(const uint32_t*)p2;
-  return (a < b) ? -1 : (a > b) ? 1 : 0;
+  assert(a != b);
+  return (a < b) ? -1 : 1;
 }
 
 // If number of colors in the image is less than or equal to MAX_PALETTE_SIZE,
@@ -85,7 +88,7 @@ static int AnalyzeAndCreatePalette(const WebPPicture* const pic,
     argb += pic->argb_stride;
   }
 
-  // TODO(skal): could we reuse in_use[] to speed up ApplyPalette()?
+  // TODO(skal): could we reuse in_use[] to speed up EncodePalette()?
   num_colors = 0;
   for (i = 0; i < (int)(sizeof(in_use) / sizeof(in_use[0])); ++i) {
     if (in_use[i]) {
@@ -220,7 +223,7 @@ static int GetHuffBitLengthsAndCodes(
   }
 
   // Create Huffman trees.
-  for (i = 0; i < histogram_image_size; ++i) {
+  for (i = 0; ok && (i < histogram_image_size); ++i) {
     HuffmanTreeCode* const codes = &huffman_codes[5 * i];
     VP8LHistogram* const histo = histogram_image->histograms[i];
     ok = ok && VP8LCreateHuffmanTree(histo->literal_, 15, codes + 0);
@@ -231,7 +234,11 @@ static int GetHuffBitLengthsAndCodes(
   }
 
  End:
-  if (!ok) free(mem_buf);
+  if (!ok) {
+    free(mem_buf);
+    // If one VP8LCreateHuffmanTree() above fails, we need to clean up behind.
+    memset(huffman_codes, 0, 5 * histogram_image_size * sizeof(*huffman_codes));
+  }
   return ok;
 }
 
@@ -406,9 +413,10 @@ static int StoreHuffmanCode(VP8LBitWriter* const bw,
 }
 
 static void WriteHuffmanCode(VP8LBitWriter* const bw,
-                             const HuffmanTreeCode* const code, int index) {
-  const int depth = code->code_lengths[index];
-  const int symbol = code->codes[index];
+                             const HuffmanTreeCode* const code,
+                             int code_index) {
+  const int depth = code->code_lengths[code_index];
+  const int symbol = code->codes[code_index];
   VP8LWriteBits(bw, depth, symbol);
 }
 
@@ -557,6 +565,9 @@ static int EncodeImageInternal(VP8LBitWriter* const bw,
       !GetHuffBitLengthsAndCodes(histogram_image, huffman_codes)) {
     goto Error;
   }
+  // Free combined histograms.
+  free(histogram_image);
+  histogram_image = NULL;
 
   // Color Cache parameters.
   VP8LWriteBits(bw, 1, use_color_cache);
@@ -576,10 +587,10 @@ static int EncodeImageInternal(VP8LBitWriter* const bw,
       uint32_t i;
       if (histogram_argb == NULL) goto Error;
       for (i = 0; i < histogram_image_xysize; ++i) {
-        const int index = histogram_symbols[i] & 0xffff;
-        histogram_argb[i] = 0xff000000 | (index << 8);
-        if (index >= max_index) {
-          max_index = index + 1;
+        const int symbol_index = histogram_symbols[i] & 0xffff;
+        histogram_argb[i] = 0xff000000 | (symbol_index << 8);
+        if (symbol_index >= max_index) {
+          max_index = symbol_index + 1;
         }
       }
       histogram_image_size = max_index;
@@ -603,9 +614,6 @@ static int EncodeImageInternal(VP8LBitWriter* const bw,
       ClearHuffmanTreeIfOnlyOneSymbol(codes);
     }
   }
-  // Free combined histograms.
-  free(histogram_image);
-  histogram_image = NULL;
 
   // Store actual literals.
   StoreImageToBitMask(bw, width, histogram_bits, &refs,
@@ -613,7 +621,7 @@ static int EncodeImageInternal(VP8LBitWriter* const bw,
   ok = 1;
 
  Error:
-  if (!ok) free(histogram_image);
+  free(histogram_image);
 
   VP8LClearBackwardRefs(&refs);
   if (huffman_codes != NULL) {
@@ -711,13 +719,6 @@ static int ApplyCrossColorFilter(const VP8LEncoder* const enc,
 
 // -----------------------------------------------------------------------------
 
-static void PutLE32(uint8_t* const data, uint32_t val) {
-  data[0] = (val >>  0) & 0xff;
-  data[1] = (val >>  8) & 0xff;
-  data[2] = (val >> 16) & 0xff;
-  data[3] = (val >> 24) & 0xff;
-}
-
 static WebPEncodingError WriteRiffHeader(const WebPPicture* const pic,
                                          size_t riff_size, size_t vp8l_size) {
   uint8_t riff[RIFF_HEADER_SIZE + CHUNK_HEADER_SIZE + VP8L_SIGNATURE_SIZE] = {
@@ -812,61 +813,94 @@ static WebPEncodingError AllocateTransformBuffer(VP8LEncoder* const enc,
   return err;
 }
 
-// Bundles multiple (2, 4 or 8) pixels into a single pixel.
-// Returns the new xsize.
-static void BundleColorMap(const WebPPicture* const pic,
-                           int xbits, uint32_t* bundled_argb, int xs) {
-  int y;
-  const int bit_depth = 1 << (3 - xbits);
-  uint32_t code = 0;
-  const uint32_t* argb = pic->argb;
-  const int width = pic->width;
-  const int height = pic->height;
-
-  for (y = 0; y < height; ++y) {
-    int x;
-    for (x = 0; x < width; ++x) {
-      const int mask = (1 << xbits) - 1;
-      const int xsub = x & mask;
-      if (xsub == 0) {
-        code = 0;
-      }
-      // TODO(vikasa): simplify the bundling logic.
-      code |= (argb[x] & 0xff00) << (bit_depth * xsub);
-      bundled_argb[y * xs + (x >> xbits)] = 0xff000000 | code;
+static void ApplyPalette(uint32_t* src, uint32_t* dst,
+                         uint32_t src_stride, uint32_t dst_stride,
+                         const uint32_t* palette, int palette_size,
+                         int width, int height, int xbits, uint8_t* row) {
+  int i, x, y;
+  int use_LUT = 1;
+  for (i = 0; i < palette_size; ++i) {
+    if ((palette[i] & 0xffff00ffu) != 0) {
+      use_LUT = 0;
+      break;
+    }
+  }
+
+  if (use_LUT) {
+    int inv_palette[MAX_PALETTE_SIZE] = { 0 };
+    for (i = 0; i < palette_size; ++i) {
+      const int color = (palette[i] >> 8) & 0xff;
+      inv_palette[color] = i;
+    }
+    for (y = 0; y < height; ++y) {
+      for (x = 0; x < width; ++x) {
+        const int color = (src[x] >> 8) & 0xff;
+        row[x] = inv_palette[color];
+      }
+      VP8LBundleColorMap(row, width, xbits, dst);
+      src += src_stride;
+      dst += dst_stride;
+    }
+  } else {
+    // Use 1 pixel cache for ARGB pixels.
+    uint32_t last_pix = palette[0];
+    int last_idx = 0;
+    for (y = 0; y < height; ++y) {
+      for (x = 0; x < width; ++x) {
+        const uint32_t pix = src[x];
+        if (pix != last_pix) {
+          for (i = 0; i < palette_size; ++i) {
+            if (pix == palette[i]) {
+              last_idx = i;
+              last_pix = pix;
+              break;
+            }
+          }
+        }
+        row[x] = last_idx;
+      }
+      VP8LBundleColorMap(row, width, xbits, dst);
+      src += src_stride;
+      dst += dst_stride;
     }
-    argb += pic->argb_stride;
   }
 }
 
 // Note: Expects "enc->palette_" to be set properly.
 // Also, "enc->palette_" will be modified after this call and should not be used
 // later.
-static WebPEncodingError ApplyPalette(VP8LBitWriter* const bw,
-                                      VP8LEncoder* const enc, int quality) {
+static WebPEncodingError EncodePalette(VP8LBitWriter* const bw,
+                                       VP8LEncoder* const enc, int quality) {
   WebPEncodingError err = VP8_ENC_OK;
-  int i, x, y;
+  int i;
   const WebPPicture* const pic = enc->pic_;
-  uint32_t* argb = pic->argb;
+  uint32_t* src = pic->argb;
+  uint32_t* dst;
   const int width = pic->width;
   const int height = pic->height;
   uint32_t* const palette = enc->palette_;
   const int palette_size = enc->palette_size_;
+  uint8_t* row = NULL;
+  int xbits;
 
   // Replace each input pixel by corresponding palette index.
-  for (y = 0; y < height; ++y) {
-    for (x = 0; x < width; ++x) {
-      const uint32_t pix = argb[x];
-      for (i = 0; i < palette_size; ++i) {
-        if (pix == palette[i]) {
-          argb[x] = 0xff000000u | (i << 8);
-          break;
-        }
-      }
-    }
-    argb += pic->argb_stride;
+  // This is done line by line.
+  if (palette_size <= 4) {
+    xbits = (palette_size <= 2) ? 3 : 2;
+  } else {
+    xbits = (palette_size <= 16) ? 1 : 0;
   }
 
+  err = AllocateTransformBuffer(enc, VP8LSubSampleSize(width, xbits), height);
+  if (err != VP8_ENC_OK) goto Error;
+  dst = enc->argb_;
+
+  row = WebPSafeMalloc((uint64_t)width, sizeof(*row));
+  if (row == NULL) return VP8_ENC_ERROR_OUT_OF_MEMORY;
+
+  ApplyPalette(src, dst, pic->argb_stride, enc->current_width_,
+               palette, palette_size, width, height, xbits, row);
+
   // Save palette to bitstream.
   VP8LWriteBits(bw, 1, TRANSFORM_PRESENT);
   VP8LWriteBits(bw, 2, COLOR_INDEXING_TRANSFORM);
@@ -880,32 +914,17 @@ static WebPEncodingError ApplyPalette(VP8LBitWriter* const bw,
     goto Error;
   }
 
-  if (palette_size <= 16) {
-    // Image can be packed (multiple pixels per uint32_t).
-    int xbits = 1;
-    if (palette_size <= 2) {
-      xbits = 3;
-    } else if (palette_size <= 4) {
-      xbits = 2;
-    }
-    err = AllocateTransformBuffer(enc, VP8LSubSampleSize(width, xbits), height);
-    if (err != VP8_ENC_OK) goto Error;
-    BundleColorMap(pic, xbits, enc->argb_, enc->current_width_);
-  }
-
  Error:
+  free(row);
   return err;
 }
 
 // -----------------------------------------------------------------------------
 
-static int GetHistoBits(const WebPConfig* const config,
-                        const WebPPicture* const pic) {
-  const int width = pic->width;
-  const int height = pic->height;
+static int GetHistoBits(int method, int use_palette, int width, int height) {
   const uint64_t hist_size = sizeof(VP8LHistogram);
   // Make tile size a function of encoding method (Range: 0 to 6).
-  int histo_bits = 7 - config->method;
+  int histo_bits = (use_palette ? 9 : 7) - method;
   while (1) {
     const uint64_t huff_image_size = VP8LSubSampleSize(width, histo_bits) *
                                      VP8LSubSampleSize(height, histo_bits) *
@@ -917,13 +936,14 @@ static int GetHistoBits(const WebPConfig* const config,
          (histo_bits > MAX_HUFFMAN_BITS) ? MAX_HUFFMAN_BITS : histo_bits;
 }
 
-static void InitEncParams(VP8LEncoder* const enc) {
+static void FinishEncParams(VP8LEncoder* const enc) {
   const WebPConfig* const config = enc->config_;
-  const WebPPicture* const picture = enc->pic_;
+  const WebPPicture* const pic = enc->pic_;
   const int method = config->method;
   const float quality = config->quality;
+  const int use_palette = enc->use_palette_;
   enc->transform_bits_ = (method < 4) ? 5 : (method > 4) ? 3 : 4;
-  enc->histo_bits_ = GetHistoBits(config, picture);
+  enc->histo_bits_ = GetHistoBits(method, use_palette, pic->width, pic->height);
   enc->cache_bits_ = (quality <= 25.f) ? 0 : 7;
 }
 
@@ -965,8 +985,6 @@ WebPEncodingError VP8LEncodeStream(const WebPConfig* const config,
     goto Error;
   }
 
-  InitEncParams(enc);
-
   // ---------------------------------------------------------------------------
   // Analyze image (entropy, num_palettes etc)
 
@@ -975,8 +993,10 @@ WebPEncodingError VP8LEncodeStream(const WebPConfig* const config,
     goto Error;
   }
 
+  FinishEncParams(enc);
+
   if (enc->use_palette_) {
-    err = ApplyPalette(bw, enc, quality);
+    err = EncodePalette(bw, enc, quality);
     if (err != VP8_ENC_OK) goto Error;
     // Color cache is disabled for palette.
     enc->cache_bits_ = 0;
diff --git a/3rdparty/libwebp/enc/vp8li.h b/3rdparty/libwebp/enc/vp8li.h
index eae90dd61f..01f01f57e5 100644
--- a/3rdparty/libwebp/enc/vp8li.h
+++ b/3rdparty/libwebp/enc/vp8li.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Lossless encoder: internal header.
diff --git a/3rdparty/libwebp/enc/webpenc.c b/3rdparty/libwebp/enc/webpenc.c
index 3c275589fc..d420d063ea 100644
--- a/3rdparty/libwebp/enc/webpenc.c
+++ b/3rdparty/libwebp/enc/webpenc.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // WebP encoder: main entry point
@@ -93,34 +95,53 @@ static void ResetBoundaryPredictions(VP8Encoder* const enc) {
   enc->nz_[-1] = 0;   // constant
 }
 
-// Map configured quality level to coding tools used.
-//-------------+---+---+---+---+---+---+
-//   Quality   | 0 | 1 | 2 | 3 | 4 | 5 +
-//-------------+---+---+---+---+---+---+
-// dynamic prob| ~ | x | x | x | x | x |
-//-------------+---+---+---+---+---+---+
-// rd-opt modes|   |   | x | x | x | x |
-//-------------+---+---+---+---+---+---+
-// fast i4/i16 | x | x |   |   |   |   |
-//-------------+---+---+---+---+---+---+
-// rd-opt i4/16|   |   | x | x | x | x |
-//-------------+---+---+---+---+---+---+
-// Trellis     |   | x |   |   | x | x |
-//-------------+---+---+---+---+---+---+
-// full-SNS    |   |   |   |   |   | x |
-//-------------+---+---+---+---+---+---+
+// Mapping from config->method_ to coding tools used.
+//-------------------+---+---+---+---+---+---+---+
+//   Method          | 0 | 1 | 2 | 3 |(4)| 5 | 6 |
+//-------------------+---+---+---+---+---+---+---+
+// fast probe        | x |   |   | x |   |   |   |
+//-------------------+---+---+---+---+---+---+---+
+// dynamic proba     | ~ | x | x | x | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
+// fast mode analysis|   |   |   |   | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
+// basic rd-opt      |   |   |   | x | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
+// disto-score i4/16 |   |   | x |   |   |   |   |
+//-------------------+---+---+---+---+---+---+---+
+// rd-opt i4/16      |   |   | ~ | x | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
+// token buffer (opt)|   |   |   | x | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
+// Trellis           |   |   |   |   |   | x |Ful|
+//-------------------+---+---+---+---+---+---+---+
+// full-SNS          |   |   |   |   | x | x | x |
+//-------------------+---+---+---+---+---+---+---+
 
 static void MapConfigToTools(VP8Encoder* const enc) {
-  const int method = enc->config_->method;
-  const int limit = 100 - enc->config_->partition_limit;
+  const WebPConfig* const config = enc->config_;
+  const int method = config->method;
+  const int limit = 100 - config->partition_limit;
   enc->method_ = method;
-  enc->rd_opt_level_ = (method >= 6) ? 3
-                     : (method >= 5) ? 2
-                     : (method >= 3) ? 1
-                     : 0;
+  enc->rd_opt_level_ = (method >= 6) ? RD_OPT_TRELLIS_ALL
+                     : (method >= 5) ? RD_OPT_TRELLIS
+                     : (method >= 3) ? RD_OPT_BASIC
+                     : RD_OPT_NONE;
   enc->max_i4_header_bits_ =
       256 * 16 * 16 *                 // upper bound: up to 16bit per 4x4 block
       (limit * limit) / (100 * 100);  // ... modulated with a quadratic curve.
+
+  enc->thread_level_ = config->thread_level;
+
+  enc->do_search_ = (config->target_size > 0 || config->target_PSNR > 0);
+  if (!config->low_memory) {
+#if !defined(DISABLE_TOKEN_BUFFER)
+    enc->use_tokens_ = (method >= 3) && !enc->do_search_;
+#endif
+    if (enc->use_tokens_) {
+      enc->num_parts_ = 1;   // doesn't work with multi-partition
+    }
+  }
 }
 
 // Memory scaling with dimensions:
@@ -259,17 +280,21 @@ static VP8Encoder* InitVP8Encoder(const WebPConfig* const config,
   VP8EncInitLayer(enc);
 #endif
 
+  VP8TBufferInit(&enc->tokens_);
   return enc;
 }
 
-static void DeleteVP8Encoder(VP8Encoder* enc) {
+static int DeleteVP8Encoder(VP8Encoder* enc) {
+  int ok = 1;
   if (enc != NULL) {
-    VP8EncDeleteAlpha(enc);
+    ok = VP8EncDeleteAlpha(enc);
 #ifdef WEBP_EXPERIMENTAL_FEATURES
     VP8EncDeleteLayer(enc);
 #endif
+    VP8TBufferClear(&enc->tokens_);
     free(enc);
   }
+  return ok;
 }
 
 //------------------------------------------------------------------------------
@@ -332,7 +357,7 @@ int WebPReportProgress(const WebPPicture* const pic,
 //------------------------------------------------------------------------------
 
 int WebPEncode(const WebPConfig* config, WebPPicture* pic) {
-  int ok;
+  int ok = 0;
 
   if (pic == NULL)
     return 0;
@@ -351,32 +376,38 @@ int WebPEncode(const WebPConfig* config, WebPPicture* pic) {
   if (!config->lossless) {
     VP8Encoder* enc = NULL;
     if (pic->y == NULL || pic->u == NULL || pic->v == NULL) {
-      if (pic->argb != NULL) {
-        if (!WebPPictureARGBToYUVA(pic, WEBP_YUV420)) return 0;
-      } else {
-        return WebPEncodingSetError(pic, VP8_ENC_ERROR_NULL_PARAMETER);
-      }
+      // Make sure we have YUVA samples.
+      if (!WebPPictureARGBToYUVA(pic, WEBP_YUV420)) return 0;
     }
 
     enc = InitVP8Encoder(config, pic);
     if (enc == NULL) return 0;  // pic->error is already set.
     // Note: each of the tasks below account for 20% in the progress report.
-    ok = VP8EncAnalyze(enc)
-      && VP8StatLoop(enc)
-      && VP8EncLoop(enc)
-      && VP8EncFinishAlpha(enc)
+    ok = VP8EncAnalyze(enc);
+
+    // Analysis is done, proceed to actual coding.
+    ok = ok && VP8EncStartAlpha(enc);   // possibly done in parallel
+    if (!enc->use_tokens_) {
+      ok = ok && VP8EncLoop(enc);
+    } else {
+      ok = ok && VP8EncTokenLoop(enc);
+    }
+    ok = ok && VP8EncFinishAlpha(enc);
 #ifdef WEBP_EXPERIMENTAL_FEATURES
-      && VP8EncFinishLayer(enc)
+    ok = ok && VP8EncFinishLayer(enc);
 #endif
-      && VP8EncWrite(enc);
+
+    ok = ok && VP8EncWrite(enc);
     StoreStats(enc);
     if (!ok) {
       VP8EncFreeBitWriters(enc);
     }
-    DeleteVP8Encoder(enc);
+    ok &= DeleteVP8Encoder(enc);  // must always be called, even if !ok
   } else {
-    if (pic->argb == NULL)
-      return WebPEncodingSetError(pic, VP8_ENC_ERROR_NULL_PARAMETER);
+    // Make sure we have ARGB samples.
+    if (pic->argb == NULL && !WebPPictureYUVAToARGB(pic)) {
+      return 0;
+    }
 
     ok = VP8LEncodeImage(config, pic);  // Sets pic->error in case of problem.
   }
diff --git a/3rdparty/libwebp/mux/muxedit.c b/3rdparty/libwebp/mux/muxedit.c
index 08629d4ae2..2d25a14d04 100644
--- a/3rdparty/libwebp/mux/muxedit.c
+++ b/3rdparty/libwebp/mux/muxedit.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Set and delete APIs for mux.
@@ -12,6 +14,7 @@
 
 #include <assert.h>
 #include "./muxi.h"
+#include "../utils/utils.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -47,8 +50,9 @@ static void MuxRelease(WebPMux* const mux) {
   MuxImageDeleteAll(&mux->images_);
   DeleteAllChunks(&mux->vp8x_);
   DeleteAllChunks(&mux->iccp_);
-  DeleteAllChunks(&mux->loop_);
-  DeleteAllChunks(&mux->meta_);
+  DeleteAllChunks(&mux->anim_);
+  DeleteAllChunks(&mux->exif_);
+  DeleteAllChunks(&mux->xmp_);
   DeleteAllChunks(&mux->unknown_);
 }
 
@@ -81,13 +85,14 @@ static WebPMuxError MuxSet(WebPMux* const mux, CHUNK_INDEX idx, uint32_t nth,
   ChunkInit(&chunk);
   SWITCH_ID_LIST(IDX_VP8X, &mux->vp8x_);
   SWITCH_ID_LIST(IDX_ICCP, &mux->iccp_);
-  SWITCH_ID_LIST(IDX_LOOP, &mux->loop_);
-  SWITCH_ID_LIST(IDX_META, &mux->meta_);
-  if (idx == IDX_UNKNOWN && data->size_ > TAG_SIZE) {
+  SWITCH_ID_LIST(IDX_ANIM, &mux->anim_);
+  SWITCH_ID_LIST(IDX_EXIF, &mux->exif_);
+  SWITCH_ID_LIST(IDX_XMP,  &mux->xmp_);
+  if (idx == IDX_UNKNOWN && data->size > TAG_SIZE) {
     // For raw-data unknown chunk, the first four bytes should be the tag to be
     // used for the chunk.
-    const WebPData tmp = { data->bytes_ + TAG_SIZE, data->size_ - TAG_SIZE };
-    err = ChunkAssignData(&chunk, &tmp, copy_data, GetLE32(data->bytes_ + 0));
+    const WebPData tmp = { data->bytes + TAG_SIZE, data->size - TAG_SIZE };
+    err = ChunkAssignData(&chunk, &tmp, copy_data, GetLE32(data->bytes + 0));
     if (err == WEBP_MUX_OK)
       err = ChunkSetNth(&chunk, &mux->unknown_, nth);
   }
@@ -106,39 +111,40 @@ static WebPMuxError MuxAddChunk(WebPMux* const mux, uint32_t nth, uint32_t tag,
   return MuxSet(mux, idx, nth, &chunk_data, copy_data);
 }
 
-// Create data for frame/tile given image data, offsets and duration.
-static WebPMuxError CreateFrameTileData(const WebPData* const image,
-                                        int x_offset, int y_offset,
-                                        int duration, int is_lossless,
-                                        int is_frame,
-                                        WebPData* const frame_tile) {
+// Create data for frame/fragment given image data, offsets and duration.
+static WebPMuxError CreateFrameFragmentData(
+    const WebPData* const image, int x_offset, int y_offset, int duration,
+    WebPMuxAnimDispose dispose_method, int is_lossless, int is_frame,
+    WebPData* const frame_frgm) {
   int width;
   int height;
-  uint8_t* frame_tile_bytes;
-  const size_t frame_tile_size = kChunks[is_frame ? IDX_FRAME : IDX_TILE].size;
+  uint8_t* frame_frgm_bytes;
+  const size_t frame_frgm_size = kChunks[is_frame ? IDX_ANMF : IDX_FRGM].size;
 
   const int ok = is_lossless ?
-      VP8LGetInfo(image->bytes_, image->size_, &width, &height, NULL) :
-      VP8GetInfo(image->bytes_, image->size_, image->size_, &width, &height);
+      VP8LGetInfo(image->bytes, image->size, &width, &height, NULL) :
+      VP8GetInfo(image->bytes, image->size, image->size, &width, &height);
   if (!ok) return WEBP_MUX_INVALID_ARGUMENT;
 
-  assert(width > 0 && height > 0 && duration > 0);
+  assert(width > 0 && height > 0 && duration >= 0);
+  assert(dispose_method == (dispose_method & 1));
   // Note: assertion on upper bounds is done in PutLE24().
 
-  frame_tile_bytes = (uint8_t*)malloc(frame_tile_size);
-  if (frame_tile_bytes == NULL) return WEBP_MUX_MEMORY_ERROR;
+  frame_frgm_bytes = (uint8_t*)malloc(frame_frgm_size);
+  if (frame_frgm_bytes == NULL) return WEBP_MUX_MEMORY_ERROR;
 
-  PutLE24(frame_tile_bytes + 0, x_offset / 2);
-  PutLE24(frame_tile_bytes + 3, y_offset / 2);
+  PutLE24(frame_frgm_bytes + 0, x_offset / 2);
+  PutLE24(frame_frgm_bytes + 3, y_offset / 2);
 
   if (is_frame) {
-    PutLE24(frame_tile_bytes + 6, width - 1);
-    PutLE24(frame_tile_bytes + 9, height - 1);
-    PutLE24(frame_tile_bytes + 12, duration - 1);
+    PutLE24(frame_frgm_bytes + 6, width - 1);
+    PutLE24(frame_frgm_bytes + 9, height - 1);
+    PutLE24(frame_frgm_bytes + 12, duration);
+    frame_frgm_bytes[15] = (dispose_method & 1);
   }
 
-  frame_tile->bytes_ = frame_tile_bytes;
-  frame_tile->size_ = frame_tile_size;
+  frame_frgm->bytes = frame_frgm_bytes;
+  frame_frgm->size = frame_frgm_size;
   return WEBP_MUX_OK;
 }
 
@@ -149,8 +155,8 @@ static WebPMuxError GetImageData(const WebPData* const bitstream,
                                  WebPData* const image, WebPData* const alpha,
                                  int* const is_lossless) {
   WebPDataInit(alpha);  // Default: no alpha.
-  if (bitstream->size_ < TAG_SIZE ||
-      memcmp(bitstream->bytes_, "RIFF", TAG_SIZE)) {
+  if (bitstream->size < TAG_SIZE ||
+      memcmp(bitstream->bytes, "RIFF", TAG_SIZE)) {
     // It is NOT webp file data. Return input data as is.
     *image = *bitstream;
   } else {
@@ -166,7 +172,7 @@ static WebPMuxError GetImageData(const WebPData* const bitstream,
     }
     WebPMuxDelete(mux);
   }
-  *is_lossless = VP8LCheckSignature(image->bytes_, image->size_);
+  *is_lossless = VP8LCheckSignature(image->bytes, image->size);
   return WEBP_MUX_OK;
 }
 
@@ -185,203 +191,96 @@ static WebPMuxError DeleteChunks(WebPChunk** chunk_list, uint32_t tag) {
   return err;
 }
 
-static WebPMuxError MuxDeleteAllNamedData(WebPMux* const mux, CHUNK_INDEX idx) {
-  const WebPChunkId id = kChunks[idx].id;
+static WebPMuxError MuxDeleteAllNamedData(WebPMux* const mux, uint32_t tag) {
+  const WebPChunkId id = ChunkGetIdFromTag(tag);
   WebPChunk** chunk_list;
 
-  if (mux == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+  assert(mux != NULL);
   if (IsWPI(id)) return WEBP_MUX_INVALID_ARGUMENT;
 
   chunk_list = MuxGetChunkListFromId(mux, id);
   if (chunk_list == NULL) return WEBP_MUX_INVALID_ARGUMENT;
 
-  return DeleteChunks(chunk_list, kChunks[idx].tag);
-}
-
-static WebPMuxError DeleteLoopCount(WebPMux* const mux) {
-  return MuxDeleteAllNamedData(mux, IDX_LOOP);
+  return DeleteChunks(chunk_list, tag);
 }
 
 //------------------------------------------------------------------------------
 // Set API(s).
 
-WebPMuxError WebPMuxSetImage(WebPMux* mux,
-                             const WebPData* bitstream, int copy_data) {
+WebPMuxError WebPMuxSetChunk(WebPMux* mux, const char fourcc[4],
+                             const WebPData* chunk_data, int copy_data) {
+  CHUNK_INDEX idx;
+  uint32_t tag;
   WebPMuxError err;
-  WebPChunk chunk;
-  WebPMuxImage wpi;
-  WebPData image;
-  WebPData alpha;
-  int is_lossless;
-  int image_tag;
-
-  if (mux == NULL || bitstream == NULL || bitstream->bytes_ == NULL ||
-      bitstream->size_ > MAX_CHUNK_PAYLOAD) {
+  if (mux == NULL || fourcc == NULL || chunk_data == NULL ||
+      chunk_data->bytes == NULL || chunk_data->size > MAX_CHUNK_PAYLOAD) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
+  idx = ChunkGetIndexFromFourCC(fourcc);
+  tag = ChunkGetTagFromFourCC(fourcc);
 
-  // If given data is for a whole webp file,
-  // extract only the VP8/VP8L data from it.
-  err = GetImageData(bitstream, &image, &alpha, &is_lossless);
-  if (err != WEBP_MUX_OK) return err;
-  image_tag = is_lossless ? kChunks[IDX_VP8L].tag : kChunks[IDX_VP8].tag;
+  // Delete existing chunk(s) with the same 'fourcc'.
+  err = MuxDeleteAllNamedData(mux, tag);
+  if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
 
-  // Delete the existing images.
-  MuxImageDeleteAll(&mux->images_);
+  // Add the given chunk.
+  return MuxSet(mux, idx, 1, chunk_data, copy_data);
+}
 
-  MuxImageInit(&wpi);
-
-  if (alpha.bytes_ != NULL) {  // Add alpha chunk.
-    ChunkInit(&chunk);
-    err = ChunkAssignData(&chunk, &alpha, copy_data, kChunks[IDX_ALPHA].tag);
-    if (err != WEBP_MUX_OK) goto Err;
-    err = ChunkSetNth(&chunk, &wpi.alpha_, 1);
-    if (err != WEBP_MUX_OK) goto Err;
-  }
-
-  // Add image chunk.
+// Creates a chunk from given 'data' and sets it as 1st chunk in 'chunk_list'.
+static WebPMuxError AddDataToChunkList(
+    const WebPData* const data, int copy_data, uint32_t tag,
+    WebPChunk** chunk_list) {
+  WebPChunk chunk;
+  WebPMuxError err;
   ChunkInit(&chunk);
-  err = ChunkAssignData(&chunk, &image, copy_data, image_tag);
+  err = ChunkAssignData(&chunk, data, copy_data, tag);
   if (err != WEBP_MUX_OK) goto Err;
-  err = ChunkSetNth(&chunk, &wpi.img_, 1);
+  err = ChunkSetNth(&chunk, chunk_list, 1);
   if (err != WEBP_MUX_OK) goto Err;
-
-  // Add this image to mux.
-  err = MuxImagePush(&wpi, &mux->images_);
-  if (err != WEBP_MUX_OK) goto Err;
-
-  // All OK.
   return WEBP_MUX_OK;
-
  Err:
-  // Something bad happened.
   ChunkRelease(&chunk);
-  MuxImageRelease(&wpi);
   return err;
 }
 
-WebPMuxError WebPMuxSetMetadata(WebPMux* mux, const WebPData* metadata,
-                                int copy_data) {
-  WebPMuxError err;
-
-  if (mux == NULL || metadata == NULL || metadata->bytes_ == NULL ||
-      metadata->size_ > MAX_CHUNK_PAYLOAD) {
-    return WEBP_MUX_INVALID_ARGUMENT;
+// Extracts image & alpha data from the given bitstream and then sets wpi.alpha_
+// and wpi.img_ appropriately.
+static WebPMuxError SetAlphaAndImageChunks(
+    const WebPData* const bitstream, int copy_data, WebPMuxImage* const wpi) {
+  int is_lossless = 0;
+  WebPData image, alpha;
+  WebPMuxError err = GetImageData(bitstream, &image, &alpha, &is_lossless);
+  const int image_tag =
+      is_lossless ? kChunks[IDX_VP8L].tag : kChunks[IDX_VP8].tag;
+  if (err != WEBP_MUX_OK) return err;
+  if (alpha.bytes != NULL) {
+    err = AddDataToChunkList(&alpha, copy_data, kChunks[IDX_ALPHA].tag,
+                             &wpi->alpha_);
+    if (err != WEBP_MUX_OK) return err;
   }
-
-  // Delete the existing metadata chunk(s).
-  err = WebPMuxDeleteMetadata(mux);
-  if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
-
-  // Add the given metadata chunk.
-  return MuxSet(mux, IDX_META, 1, metadata, copy_data);
+  return AddDataToChunkList(&image, copy_data, image_tag, &wpi->img_);
 }
 
-WebPMuxError WebPMuxSetColorProfile(WebPMux* mux, const WebPData* color_profile,
-                                    int copy_data) {
-  WebPMuxError err;
-
-  if (mux == NULL || color_profile == NULL || color_profile->bytes_ == NULL ||
-      color_profile->size_ > MAX_CHUNK_PAYLOAD) {
-    return WEBP_MUX_INVALID_ARGUMENT;
-  }
-
-  // Delete the existing ICCP chunk(s).
-  err = WebPMuxDeleteColorProfile(mux);
-  if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
-
-  // Add the given ICCP chunk.
-  return MuxSet(mux, IDX_ICCP, 1, color_profile, copy_data);
-}
-
-WebPMuxError WebPMuxSetLoopCount(WebPMux* mux, int loop_count) {
-  WebPMuxError err;
-  uint8_t* data = NULL;
-
-  if (mux == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-  if (loop_count >= MAX_LOOP_COUNT) return WEBP_MUX_INVALID_ARGUMENT;
-
-  // Delete the existing LOOP chunk(s).
-  err = DeleteLoopCount(mux);
-  if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
-
-  // Add the given loop count.
-  data = (uint8_t*)malloc(kChunks[IDX_LOOP].size);
-  if (data == NULL) return WEBP_MUX_MEMORY_ERROR;
-
-  PutLE16(data, loop_count);
-  err = MuxAddChunk(mux, 1, kChunks[IDX_LOOP].tag, data,
-                    kChunks[IDX_LOOP].size, 1);
-  free(data);
-  return err;
-}
-
-static WebPMuxError MuxPushFrameTileInternal(
-    WebPMux* const mux, const WebPData* const bitstream, int x_offset,
-    int y_offset, int duration, int copy_data, uint32_t tag) {
-  WebPChunk chunk;
-  WebPData image;
-  WebPData alpha;
+WebPMuxError WebPMuxSetImage(WebPMux* mux, const WebPData* bitstream,
+                             int copy_data) {
   WebPMuxImage wpi;
   WebPMuxError err;
-  WebPData frame_tile;
-  const int is_frame = (tag == kChunks[IDX_FRAME].tag) ? 1 : 0;
-  int is_lossless;
-  int image_tag;
 
   // Sanity checks.
-  if (mux == NULL || bitstream == NULL || bitstream->bytes_ == NULL ||
-      bitstream->size_ > MAX_CHUNK_PAYLOAD) {
-    return WEBP_MUX_INVALID_ARGUMENT;
-  }
-  if (x_offset < 0 || x_offset >= MAX_POSITION_OFFSET ||
-      y_offset < 0 || y_offset >= MAX_POSITION_OFFSET ||
-      duration <= 0 || duration > MAX_DURATION) {
+  if (mux == NULL || bitstream == NULL || bitstream->bytes == NULL ||
+      bitstream->size > MAX_CHUNK_PAYLOAD) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 
-  // Snap offsets to even positions.
-  x_offset &= ~1;
-  y_offset &= ~1;
+  if (mux->images_ != NULL) {
+    // Only one 'simple image' can be added in mux. So, remove present images.
+    MuxImageDeleteAll(&mux->images_);
+  }
 
-  // If given data is for a whole webp file,
-  // extract only the VP8/VP8L data from it.
-  err = GetImageData(bitstream, &image, &alpha, &is_lossless);
-  if (err != WEBP_MUX_OK) return err;
-  image_tag = is_lossless ? kChunks[IDX_VP8L].tag : kChunks[IDX_VP8].tag;
-
-  WebPDataInit(&frame_tile);
-  ChunkInit(&chunk);
   MuxImageInit(&wpi);
-
-  if (alpha.bytes_ != NULL) {
-    // Add alpha chunk.
-    err = ChunkAssignData(&chunk, &alpha, copy_data, kChunks[IDX_ALPHA].tag);
-    if (err != WEBP_MUX_OK) goto Err;
-    err = ChunkSetNth(&chunk, &wpi.alpha_, 1);
-    if (err != WEBP_MUX_OK) goto Err;
-    ChunkInit(&chunk);  // chunk owned by wpi.alpha_ now.
-  }
-
-  // Add image chunk.
-  err = ChunkAssignData(&chunk, &image, copy_data, image_tag);
+  err = SetAlphaAndImageChunks(bitstream, copy_data, &wpi);
   if (err != WEBP_MUX_OK) goto Err;
-  err = ChunkSetNth(&chunk, &wpi.img_, 1);
-  if (err != WEBP_MUX_OK) goto Err;
-  ChunkInit(&chunk);  // chunk owned by wpi.img_ now.
-
-  // Create frame/tile data.
-  err = CreateFrameTileData(&image, x_offset, y_offset, duration, is_lossless,
-                            is_frame, &frame_tile);
-  if (err != WEBP_MUX_OK) goto Err;
-
-  // Add frame/tile chunk (with copy_data = 1).
-  err = ChunkAssignData(&chunk, &frame_tile, 1, tag);
-  if (err != WEBP_MUX_OK) goto Err;
-  WebPDataClear(&frame_tile);
-  err = ChunkSetNth(&chunk, &wpi.header_, 1);
-  if (err != WEBP_MUX_OK) goto Err;
-  ChunkInit(&chunk);  // chunk owned by wpi.header_ now.
 
   // Add this WebPMuxImage to mux.
   err = MuxImagePush(&wpi, &mux->images_);
@@ -391,86 +290,137 @@ static WebPMuxError MuxPushFrameTileInternal(
   return WEBP_MUX_OK;
 
  Err:  // Something bad happened.
-  WebPDataClear(&frame_tile);
-  ChunkRelease(&chunk);
   MuxImageRelease(&wpi);
   return err;
 }
 
-WebPMuxError WebPMuxPushFrame(WebPMux* mux, const WebPData* bitstream,
-                              int x_offset, int y_offset,
-                              int duration, int copy_data) {
-  return MuxPushFrameTileInternal(mux, bitstream, x_offset, y_offset,
-                                  duration, copy_data, kChunks[IDX_FRAME].tag);
+WebPMuxError WebPMuxPushFrame(WebPMux* mux, const WebPMuxFrameInfo* frame,
+                              int copy_data) {
+  WebPMuxImage wpi;
+  WebPMuxError err;
+  int is_frame;
+  const WebPData* const bitstream = &frame->bitstream;
+
+  // Sanity checks.
+  if (mux == NULL || frame == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+
+  is_frame = (frame->id == WEBP_CHUNK_ANMF);
+  if (!(is_frame || (frame->id == WEBP_CHUNK_FRGM))) {
+    return WEBP_MUX_INVALID_ARGUMENT;
+  }
+#ifndef WEBP_EXPERIMENTAL_FEATURES
+  if (frame->id == WEBP_CHUNK_FRGM) {     // disabled for now.
+    return WEBP_MUX_INVALID_ARGUMENT;
+  }
+#endif
+
+  if (bitstream->bytes == NULL || bitstream->size > MAX_CHUNK_PAYLOAD) {
+    return WEBP_MUX_INVALID_ARGUMENT;
+  }
+
+  if (mux->images_ != NULL) {
+    const WebPMuxImage* const image = mux->images_;
+    const uint32_t image_id = (image->header_ != NULL) ?
+        ChunkGetIdFromTag(image->header_->tag_) : WEBP_CHUNK_IMAGE;
+    if (image_id != frame->id) {
+      return WEBP_MUX_INVALID_ARGUMENT;  // Conflicting frame types.
+    }
+  }
+
+  MuxImageInit(&wpi);
+  err = SetAlphaAndImageChunks(bitstream, copy_data, &wpi);
+  if (err != WEBP_MUX_OK) goto Err;
+  assert(wpi.img_ != NULL);  // As SetAlphaAndImageChunks() was successful.
+
+  {
+    const int is_lossless = (wpi.img_->tag_ == kChunks[IDX_VP8L].tag);
+    const int x_offset = frame->x_offset & ~1;  // Snap offsets to even.
+    const int y_offset = frame->y_offset & ~1;
+    const int duration = is_frame ? frame->duration : 1 /* unused */;
+    const WebPMuxAnimDispose dispose_method =
+        is_frame ? frame->dispose_method : 0 /* unused */;
+    const uint32_t tag = kChunks[is_frame ? IDX_ANMF : IDX_FRGM].tag;
+    WebPData frame_frgm;
+    if (x_offset < 0 || x_offset >= MAX_POSITION_OFFSET ||
+        y_offset < 0 || y_offset >= MAX_POSITION_OFFSET ||
+        (duration < 0 || duration >= MAX_DURATION) ||
+        dispose_method != (dispose_method & 1)) {
+      err = WEBP_MUX_INVALID_ARGUMENT;
+      goto Err;
+    }
+    err = CreateFrameFragmentData(&wpi.img_->data_, x_offset, y_offset,
+                                  duration, dispose_method, is_lossless,
+                                  is_frame, &frame_frgm);
+    if (err != WEBP_MUX_OK) goto Err;
+    // Add frame/fragment chunk (with copy_data = 1).
+    err = AddDataToChunkList(&frame_frgm, 1, tag, &wpi.header_);
+    WebPDataClear(&frame_frgm);  // frame_frgm owned by wpi.header_ now.
+    if (err != WEBP_MUX_OK) goto Err;
+  }
+
+  // Add this WebPMuxImage to mux.
+  err = MuxImagePush(&wpi, &mux->images_);
+  if (err != WEBP_MUX_OK) goto Err;
+
+  // All is well.
+  return WEBP_MUX_OK;
+
+ Err:  // Something bad happened.
+  MuxImageRelease(&wpi);
+  return err;
 }
 
-WebPMuxError WebPMuxPushTile(WebPMux* mux, const WebPData* bitstream,
-                             int x_offset, int y_offset,
-                             int copy_data) {
-  return MuxPushFrameTileInternal(mux, bitstream, x_offset, y_offset,
-                                  1 /* unused duration */, copy_data,
-                                  kChunks[IDX_TILE].tag);
+WebPMuxError WebPMuxSetAnimationParams(WebPMux* mux,
+                                       const WebPMuxAnimParams* params) {
+  WebPMuxError err;
+  uint8_t data[ANIM_CHUNK_SIZE];
+
+  if (mux == NULL || params == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+  if (params->loop_count < 0 || params->loop_count >= MAX_LOOP_COUNT) {
+    return WEBP_MUX_INVALID_ARGUMENT;
+  }
+
+  // Delete any existing ANIM chunk(s).
+  err = MuxDeleteAllNamedData(mux, kChunks[IDX_ANIM].tag);
+  if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
+
+  // Set the animation parameters.
+  PutLE32(data, params->bgcolor);
+  PutLE16(data + 4, params->loop_count);
+  return MuxAddChunk(mux, 1, kChunks[IDX_ANIM].tag, data, sizeof(data), 1);
 }
 
 //------------------------------------------------------------------------------
 // Delete API(s).
 
-WebPMuxError WebPMuxDeleteImage(WebPMux* mux) {
-  WebPMuxError err;
-
-  if (mux == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-
-  err = MuxValidateForImage(mux);
-  if (err != WEBP_MUX_OK) return err;
-
-  // All well, delete image.
-  MuxImageDeleteAll(&mux->images_);
-  return WEBP_MUX_OK;
-}
-
-WebPMuxError WebPMuxDeleteMetadata(WebPMux* mux) {
-  return MuxDeleteAllNamedData(mux, IDX_META);
-}
-
-WebPMuxError WebPMuxDeleteColorProfile(WebPMux* mux) {
-  return MuxDeleteAllNamedData(mux, IDX_ICCP);
-}
-
-static WebPMuxError DeleteFrameTileInternal(WebPMux* const mux, uint32_t nth,
-                                            CHUNK_INDEX idx) {
-  const WebPChunkId id = kChunks[idx].id;
-  if (mux == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-
-  assert(idx == IDX_FRAME || idx == IDX_TILE);
-  return MuxImageDeleteNth(&mux->images_, nth, id);
+WebPMuxError WebPMuxDeleteChunk(WebPMux* mux, const char fourcc[4]) {
+  if (mux == NULL || fourcc == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+  return MuxDeleteAllNamedData(mux, ChunkGetTagFromFourCC(fourcc));
 }
 
 WebPMuxError WebPMuxDeleteFrame(WebPMux* mux, uint32_t nth) {
-  return DeleteFrameTileInternal(mux, nth, IDX_FRAME);
-}
-
-WebPMuxError WebPMuxDeleteTile(WebPMux* mux, uint32_t nth) {
-  return DeleteFrameTileInternal(mux, nth, IDX_TILE);
+  if (mux == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+  return MuxImageDeleteNth(&mux->images_, nth);
 }
 
 //------------------------------------------------------------------------------
 // Assembly of the WebP RIFF file.
 
-static WebPMuxError GetFrameTileInfo(const WebPChunk* const frame_tile_chunk,
-                                     int* const x_offset, int* const y_offset,
-                                     int* const duration) {
-  const uint32_t tag = frame_tile_chunk->tag_;
-  const int is_frame = (tag == kChunks[IDX_FRAME].tag);
-  const WebPData* const data = &frame_tile_chunk->data_;
+static WebPMuxError GetFrameFragmentInfo(
+    const WebPChunk* const frame_frgm_chunk,
+    int* const x_offset, int* const y_offset, int* const duration) {
+  const uint32_t tag = frame_frgm_chunk->tag_;
+  const int is_frame = (tag == kChunks[IDX_ANMF].tag);
+  const WebPData* const data = &frame_frgm_chunk->data_;
   const size_t expected_data_size =
-      is_frame ? FRAME_CHUNK_SIZE : TILE_CHUNK_SIZE;
-  assert(frame_tile_chunk != NULL);
-  assert(tag == kChunks[IDX_FRAME].tag || tag ==  kChunks[IDX_TILE].tag);
-  if (data->size_ != expected_data_size) return WEBP_MUX_INVALID_ARGUMENT;
+      is_frame ? ANMF_CHUNK_SIZE : FRGM_CHUNK_SIZE;
+  assert(frame_frgm_chunk != NULL);
+  assert(tag == kChunks[IDX_ANMF].tag || tag ==  kChunks[IDX_FRGM].tag);
+  if (data->size != expected_data_size) return WEBP_MUX_INVALID_ARGUMENT;
 
-  *x_offset = 2 * GetLE24(data->bytes_ + 0);
-  *y_offset = 2 * GetLE24(data->bytes_ + 3);
-  if (is_frame) *duration = 1 + GetLE24(data->bytes_ + 12);
+  *x_offset = 2 * GetLE24(data->bytes + 0);
+  *y_offset = 2 * GetLE24(data->bytes + 3);
+  if (is_frame) *duration = GetLE24(data->bytes + 12);
   return WEBP_MUX_OK;
 }
 
@@ -483,8 +433,8 @@ WebPMuxError MuxGetImageWidthHeight(const WebPChunk* const image_chunk,
   assert(image_chunk != NULL);
   assert(tag == kChunks[IDX_VP8].tag || tag ==  kChunks[IDX_VP8L].tag);
   ok = (tag == kChunks[IDX_VP8].tag) ?
-      VP8GetInfo(data->bytes_, data->size_, data->size_, &w, &h) :
-      VP8LGetInfo(data->bytes_, data->size_, &w, &h, NULL);
+      VP8GetInfo(data->bytes, data->size, data->size, &w, &h) :
+      VP8LGetInfo(data->bytes, data->size, &w, &h, NULL);
   if (ok) {
     *width = w;
     *height = h;
@@ -499,11 +449,11 @@ static WebPMuxError GetImageInfo(const WebPMuxImage* const wpi,
                                  int* const duration,
                                  int* const width, int* const height) {
   const WebPChunk* const image_chunk = wpi->img_;
-  const WebPChunk* const frame_tile_chunk = wpi->header_;
+  const WebPChunk* const frame_frgm_chunk = wpi->header_;
 
-  // Get offsets and duration from FRM/TILE chunk.
+  // Get offsets and duration from ANMF/FRGM chunk.
   const WebPMuxError err =
-      GetFrameTileInfo(frame_tile_chunk, x_offset, y_offset, duration);
+      GetFrameFragmentInfo(frame_frgm_chunk, x_offset, y_offset, duration);
   if (err != WEBP_MUX_OK) return err;
 
   // Get width and height from VP8/VP8L chunk.
@@ -525,9 +475,9 @@ static WebPMuxError GetImageCanvasWidthHeight(
     int max_x = 0;
     int max_y = 0;
     int64_t image_area = 0;
-    // Aggregate the bounding box for animation frames & tiled images.
+    // Aggregate the bounding box for animation frames & fragmented images.
     for (; wpi != NULL; wpi = wpi->next_) {
-      int x_offset, y_offset, duration, w, h;
+      int x_offset = 0, y_offset = 0, duration = 0, w = 0, h = 0;
       const WebPMuxError err = GetImageInfo(wpi, &x_offset, &y_offset,
                                             &duration, &w, &h);
       const int max_x_pos = x_offset + w;
@@ -542,11 +492,11 @@ static WebPMuxError GetImageCanvasWidthHeight(
     }
     *width = max_x;
     *height = max_y;
-    // Crude check to validate that there are no image overlaps/holes for tile
-    // images. Check that the aggregated image area for individual tiles exactly
-    // matches the image area of the constructed canvas. However, the area-match
-    // is necessary but not sufficient condition.
-    if ((flags & TILE_FLAG) && (image_area != (max_x * max_y))) {
+    // Crude check to validate that there are no image overlaps/holes for
+    // fragmented images. Check that the aggregated image area for individual
+    // fragments exactly matches the image area of the constructed canvas.
+    // However, the area-match is necessary but not sufficient condition.
+    if ((flags & FRAGMENTS_FLAG) && (image_area != (max_x * max_y))) {
       *width = 0;
       *height = 0;
       return WEBP_MUX_INVALID_ARGUMENT;
@@ -580,34 +530,34 @@ static WebPMuxError CreateVP8XChunk(WebPMux* const mux) {
   assert(mux != NULL);
   images = mux->images_;  // First image.
   if (images == NULL || images->img_ == NULL ||
-      images->img_->data_.bytes_ == NULL) {
+      images->img_->data_.bytes == NULL) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 
   // If VP8X chunk(s) is(are) already present, remove them (and later add new
   // VP8X chunk with updated flags).
-  err = MuxDeleteAllNamedData(mux, IDX_VP8X);
+  err = MuxDeleteAllNamedData(mux, kChunks[IDX_VP8X].tag);
   if (err != WEBP_MUX_OK && err != WEBP_MUX_NOT_FOUND) return err;
 
   // Set flags.
-  if (mux->iccp_ != NULL && mux->iccp_->data_.bytes_ != NULL) {
+  if (mux->iccp_ != NULL && mux->iccp_->data_.bytes != NULL) {
     flags |= ICCP_FLAG;
   }
-
-  if (mux->meta_ != NULL && mux->meta_->data_.bytes_ != NULL) {
-    flags |= META_FLAG;
+  if (mux->exif_ != NULL && mux->exif_->data_.bytes != NULL) {
+    flags |= EXIF_FLAG;
+  }
+  if (mux->xmp_ != NULL && mux->xmp_->data_.bytes != NULL) {
+    flags |= XMP_FLAG;
   }
-
   if (images->header_ != NULL) {
-    if (images->header_->tag_ == kChunks[IDX_TILE].tag) {
-      // This is a tiled image.
-      flags |= TILE_FLAG;
-    } else if (images->header_->tag_ == kChunks[IDX_FRAME].tag) {
+    if (images->header_->tag_ == kChunks[IDX_FRGM].tag) {
+      // This is a fragmented image.
+      flags |= FRAGMENTS_FLAG;
+    } else if (images->header_->tag_ == kChunks[IDX_ANMF].tag) {
       // This is an image with animation.
       flags |= ANIMATION_FLAG;
     }
   }
-
   if (MuxImageCount(images, WEBP_CHUNK_ALPHA) > 0) {
     flags |= ALPHA_FLAG;  // Some images have an alpha channel.
   }
@@ -643,39 +593,63 @@ static WebPMuxError CreateVP8XChunk(WebPMux* const mux) {
   return err;
 }
 
+// Cleans up 'mux' by removing any unnecessary chunks.
+static WebPMuxError MuxCleanup(WebPMux* const mux) {
+  int num_frames;
+  int num_fragments;
+  int num_anim_chunks;
+
+  // If we have an image with single fragment or frame, convert it to a
+  // non-animated non-fragmented image (to avoid writing FRGM/ANMF chunk
+  // unnecessarily).
+  WebPMuxError err = WebPMuxNumChunks(mux, kChunks[IDX_ANMF].id, &num_frames);
+  if (err != WEBP_MUX_OK) return err;
+  err = WebPMuxNumChunks(mux, kChunks[IDX_FRGM].id, &num_fragments);
+  if (err != WEBP_MUX_OK) return err;
+  if (num_frames == 1 || num_fragments == 1) {
+    WebPMuxImage* frame_frag;
+    err = MuxImageGetNth((const WebPMuxImage**)&mux->images_, 1, &frame_frag);
+    assert(err == WEBP_MUX_OK);  // We know that one frame/fragment does exist.
+    if (frame_frag->header_ != NULL) {
+      assert(frame_frag->header_->tag_ == kChunks[IDX_ANMF].tag ||
+             frame_frag->header_->tag_ == kChunks[IDX_FRGM].tag);
+      ChunkDelete(frame_frag->header_);  // Removes ANMF/FRGM chunk.
+      frame_frag->header_ = NULL;
+    }
+    num_frames = 0;
+    num_fragments = 0;
+  }
+  // Remove ANIM chunk if this is a non-animated image.
+  err = WebPMuxNumChunks(mux, kChunks[IDX_ANIM].id, &num_anim_chunks);
+  if (err != WEBP_MUX_OK) return err;
+  if (num_anim_chunks >= 1 && num_frames == 0) {
+    err = MuxDeleteAllNamedData(mux, kChunks[IDX_ANIM].tag);
+    if (err != WEBP_MUX_OK) return err;
+  }
+  return WEBP_MUX_OK;
+}
+
 WebPMuxError WebPMuxAssemble(WebPMux* mux, WebPData* assembled_data) {
   size_t size = 0;
   uint8_t* data = NULL;
   uint8_t* dst = NULL;
-  int num_frames;
-  int num_loop_chunks;
   WebPMuxError err;
 
   if (mux == NULL || assembled_data == NULL) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 
-  // Remove LOOP chunk if unnecessary.
-  err = WebPMuxNumChunks(mux, kChunks[IDX_LOOP].id, &num_loop_chunks);
+  // Finalize mux.
+  err = MuxCleanup(mux);
   if (err != WEBP_MUX_OK) return err;
-  if (num_loop_chunks >= 1) {
-    err = WebPMuxNumChunks(mux, kChunks[IDX_FRAME].id, &num_frames);
-    if (err != WEBP_MUX_OK) return err;
-    if (num_frames == 0) {
-      err = DeleteLoopCount(mux);
-      if (err != WEBP_MUX_OK) return err;
-    }
-  }
-
-  // Create VP8X chunk.
   err = CreateVP8XChunk(mux);
   if (err != WEBP_MUX_OK) return err;
 
   // Allocate data.
   size = ChunksListDiskSize(mux->vp8x_) + ChunksListDiskSize(mux->iccp_)
-       + ChunksListDiskSize(mux->loop_) + MuxImageListDiskSize(mux->images_)
-       + ChunksListDiskSize(mux->meta_) + ChunksListDiskSize(mux->unknown_)
-       + RIFF_HEADER_SIZE;
+       + ChunksListDiskSize(mux->anim_) + MuxImageListDiskSize(mux->images_)
+       + ChunksListDiskSize(mux->exif_) + ChunksListDiskSize(mux->xmp_)
+       + ChunksListDiskSize(mux->unknown_) + RIFF_HEADER_SIZE;
 
   data = (uint8_t*)malloc(size);
   if (data == NULL) return WEBP_MUX_MEMORY_ERROR;
@@ -684,9 +658,10 @@ WebPMuxError WebPMuxAssemble(WebPMux* mux, WebPData* assembled_data) {
   dst = MuxEmitRiffHeader(data, size);
   dst = ChunkListEmit(mux->vp8x_, dst);
   dst = ChunkListEmit(mux->iccp_, dst);
-  dst = ChunkListEmit(mux->loop_, dst);
+  dst = ChunkListEmit(mux->anim_, dst);
   dst = MuxImageListEmit(mux->images_, dst);
-  dst = ChunkListEmit(mux->meta_, dst);
+  dst = ChunkListEmit(mux->exif_, dst);
+  dst = ChunkListEmit(mux->xmp_, dst);
   dst = ChunkListEmit(mux->unknown_, dst);
   assert(dst == data + size);
 
@@ -698,9 +673,9 @@ WebPMuxError WebPMuxAssemble(WebPMux* mux, WebPData* assembled_data) {
     size = 0;
   }
 
-  // Finalize.
-  assembled_data->bytes_ = data;
-  assembled_data->size_ = size;
+  // Finalize data.
+  assembled_data->bytes = data;
+  assembled_data->size = size;
 
   return err;
 }
diff --git a/3rdparty/libwebp/mux/muxi.h b/3rdparty/libwebp/mux/muxi.h
index edd8c368cd..eaed558b47 100644
--- a/3rdparty/libwebp/mux/muxi.h
+++ b/3rdparty/libwebp/mux/muxi.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Internal header for mux library.
@@ -15,7 +17,6 @@
 #include <stdlib.h>
 #include "../dec/vp8i.h"
 #include "../dec/vp8li.h"
-#include "../webp/format_constants.h"
 #include "../webp/mux.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -25,22 +26,26 @@ extern "C" {
 //------------------------------------------------------------------------------
 // Defines and constants.
 
+#define MUX_MAJ_VERSION 0
+#define MUX_MIN_VERSION 1
+#define MUX_REV_VERSION 1
+
 // Chunk object.
 typedef struct WebPChunk WebPChunk;
 struct WebPChunk {
   uint32_t        tag_;
   int             owner_;  // True if *data_ memory is owned internally.
-                           // VP8X, Loop, and other internally created chunks
-                           // like frame/tile are always owned.
+                           // VP8X, ANIM, and other internally created chunks
+                           // like ANMF/FRGM are always owned.
   WebPData        data_;
   WebPChunk*      next_;
 };
 
-// MuxImage object. Store a full webp image (including frame/tile chunk, alpha
+// MuxImage object. Store a full WebP image (including ANMF/FRGM chunk, ALPH
 // chunk and VP8/VP8L chunk),
 typedef struct WebPMuxImage WebPMuxImage;
 struct WebPMuxImage {
-  WebPChunk*  header_;      // Corresponds to WEBP_CHUNK_FRAME/WEBP_CHUNK_TILE.
+  WebPChunk*  header_;      // Corresponds to WEBP_CHUNK_ANMF/WEBP_CHUNK_FRGM.
   WebPChunk*  alpha_;       // Corresponds to WEBP_CHUNK_ALPHA.
   WebPChunk*  img_;         // Corresponds to WEBP_CHUNK_IMAGE.
   int         is_partial_;  // True if only some of the chunks are filled.
@@ -51,8 +56,9 @@ struct WebPMuxImage {
 struct WebPMux {
   WebPMuxImage*   images_;
   WebPChunk*      iccp_;
-  WebPChunk*      meta_;
-  WebPChunk*      loop_;
+  WebPChunk*      exif_;
+  WebPChunk*      xmp_;
+  WebPChunk*      anim_;
   WebPChunk*      vp8x_;
 
   WebPChunk*  unknown_;
@@ -65,13 +71,14 @@ struct WebPMux {
 typedef enum {
   IDX_VP8X = 0,
   IDX_ICCP,
-  IDX_LOOP,
-  IDX_FRAME,
-  IDX_TILE,
+  IDX_ANIM,
+  IDX_ANMF,
+  IDX_FRGM,
   IDX_ALPHA,
   IDX_VP8,
   IDX_VP8L,
-  IDX_META,
+  IDX_EXIF,
+  IDX_XMP,
   IDX_UNKNOWN,
 
   IDX_NIL,
@@ -80,8 +87,6 @@ typedef enum {
 
 #define NIL_TAG 0x00000000u  // To signal void chunk.
 
-#define MKFOURCC(a, b, c, d) ((uint32_t)(a) | (b) << 8 | (c) << 16 | (d) << 24)
-
 typedef struct {
   uint32_t      tag;
   WebPChunkId   id;
@@ -90,44 +95,6 @@ typedef struct {
 
 extern const ChunkInfo kChunks[IDX_LAST_CHUNK];
 
-//------------------------------------------------------------------------------
-// Helper functions.
-
-// Read 16, 24 or 32 bits stored in little-endian order.
-static WEBP_INLINE int GetLE16(const uint8_t* const data) {
-  return (int)(data[0] << 0) | (data[1] << 8);
-}
-
-static WEBP_INLINE int GetLE24(const uint8_t* const data) {
-  return GetLE16(data) | (data[2] << 16);
-}
-
-static WEBP_INLINE uint32_t GetLE32(const uint8_t* const data) {
-  return (uint32_t)GetLE16(data) | (GetLE16(data + 2) << 16);
-}
-
-// Store 16, 24 or 32 bits in little-endian order.
-static WEBP_INLINE void PutLE16(uint8_t* const data, int val) {
-  assert(val < (1 << 16));
-  data[0] = (val >> 0);
-  data[1] = (val >> 8);
-}
-
-static WEBP_INLINE void PutLE24(uint8_t* const data, int val) {
-  assert(val < (1 << 24));
-  PutLE16(data, val & 0xffff);
-  data[2] = (val >> 16);
-}
-
-static WEBP_INLINE void PutLE32(uint8_t* const data, uint32_t val) {
-  PutLE16(data, (int)(val & 0xffff));
-  PutLE16(data + 2, (int)(val >> 16));
-}
-
-static WEBP_INLINE size_t SizeWithPadding(size_t chunk_size) {
-  return CHUNK_HEADER_SIZE + ((chunk_size + 1) & ~1U);
-}
-
 //------------------------------------------------------------------------------
 // Chunk object management.
 
@@ -140,6 +107,12 @@ CHUNK_INDEX ChunkGetIndexFromTag(uint32_t tag);
 // Get chunk id from chunk tag. Returns WEBP_CHUNK_NIL if not found.
 WebPChunkId ChunkGetIdFromTag(uint32_t tag);
 
+// Convert a fourcc string to a tag.
+uint32_t ChunkGetTagFromFourCC(const char fourcc[4]);
+
+// Get chunk index from fourcc. Returns IDX_UNKNOWN if given fourcc is unknown.
+CHUNK_INDEX ChunkGetIndexFromFourCC(const char fourcc[4]);
+
 // Search for nth chunk with given 'tag' in the chunk list.
 // nth = 0 means "last of the list".
 WebPChunk* ChunkSearchList(WebPChunk* first, uint32_t nth, uint32_t tag);
@@ -150,7 +123,8 @@ WebPMuxError ChunkAssignData(WebPChunk* chunk, const WebPData* const data,
 
 // Sets 'chunk' at nth position in the 'chunk_list'.
 // nth = 0 has the special meaning "last of the list".
-WebPMuxError ChunkSetNth(const WebPChunk* chunk, WebPChunk** chunk_list,
+// On success ownership is transferred from 'chunk' to the 'chunk_list'.
+WebPMuxError ChunkSetNth(WebPChunk* chunk, WebPChunk** chunk_list,
                          uint32_t nth);
 
 // Releases chunk and returns chunk->next_.
@@ -159,9 +133,14 @@ WebPChunk* ChunkRelease(WebPChunk* const chunk);
 // Deletes given chunk & returns chunk->next_.
 WebPChunk* ChunkDelete(WebPChunk* const chunk);
 
+// Returns size of the chunk including chunk header and padding byte (if any).
+static WEBP_INLINE size_t SizeWithPadding(size_t chunk_size) {
+  return CHUNK_HEADER_SIZE + ((chunk_size + 1) & ~1U);
+}
+
 // Size of a chunk including header and padding.
 static WEBP_INLINE size_t ChunkDiskSize(const WebPChunk* chunk) {
-  const size_t data_size = chunk->data_.size_;
+  const size_t data_size = chunk->data_.size;
   assert(data_size < MAX_CHUNK_PAYLOAD);
   return SizeWithPadding(data_size);
 }
@@ -193,13 +172,14 @@ WebPMuxImage* MuxImageDelete(WebPMuxImage* const wpi);
 void MuxImageDeleteAll(WebPMuxImage** const wpi_list);
 
 // Count number of images matching the given tag id in the 'wpi_list'.
+// If id == WEBP_CHUNK_NIL, all images will be matched.
 int MuxImageCount(const WebPMuxImage* wpi_list, WebPChunkId id);
 
 // Check if given ID corresponds to an image related chunk.
 static WEBP_INLINE int IsWPI(WebPChunkId id) {
   switch (id) {
-    case WEBP_CHUNK_FRAME:
-    case WEBP_CHUNK_TILE:
+    case WEBP_CHUNK_ANMF:
+    case WEBP_CHUNK_FRGM:
     case WEBP_CHUNK_ALPHA:
     case WEBP_CHUNK_IMAGE:  return 1;
     default:        return 0;
@@ -211,8 +191,8 @@ static WEBP_INLINE WebPChunk** MuxImageGetListFromId(
     const WebPMuxImage* const wpi, WebPChunkId id) {
   assert(wpi != NULL);
   switch (id) {
-    case WEBP_CHUNK_FRAME:
-    case WEBP_CHUNK_TILE:  return (WebPChunk**)&wpi->header_;
+    case WEBP_CHUNK_ANMF:
+    case WEBP_CHUNK_FRGM:  return (WebPChunk**)&wpi->header_;
     case WEBP_CHUNK_ALPHA: return (WebPChunk**)&wpi->alpha_;
     case WEBP_CHUNK_IMAGE: return (WebPChunk**)&wpi->img_;
     default: return NULL;
@@ -222,13 +202,12 @@ static WEBP_INLINE WebPChunk** MuxImageGetListFromId(
 // Pushes 'wpi' at the end of 'wpi_list'.
 WebPMuxError MuxImagePush(const WebPMuxImage* wpi, WebPMuxImage** wpi_list);
 
-// Delete nth image in the image list with given tag id.
-WebPMuxError MuxImageDeleteNth(WebPMuxImage** wpi_list, uint32_t nth,
-                               WebPChunkId id);
+// Delete nth image in the image list.
+WebPMuxError MuxImageDeleteNth(WebPMuxImage** wpi_list, uint32_t nth);
 
-// Get nth image in the image list with given tag id.
+// Get nth image in the image list.
 WebPMuxError MuxImageGetNth(const WebPMuxImage** wpi_list, uint32_t nth,
-                            WebPChunkId id, WebPMuxImage** wpi);
+                            WebPMuxImage** wpi);
 
 // Total size of the given image.
 size_t MuxImageDiskSize(const WebPMuxImage* const wpi);
diff --git a/3rdparty/libwebp/mux/muxinternal.c b/3rdparty/libwebp/mux/muxinternal.c
index 6c3c4fe60a..9543c0ee5b 100644
--- a/3rdparty/libwebp/mux/muxinternal.c
+++ b/3rdparty/libwebp/mux/muxinternal.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Internal objects and utils for mux.
@@ -12,6 +14,7 @@
 
 #include <assert.h>
 #include "./muxi.h"
+#include "../utils/utils.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -22,18 +25,25 @@ extern "C" {
 const ChunkInfo kChunks[] = {
   { MKFOURCC('V', 'P', '8', 'X'),  WEBP_CHUNK_VP8X,    VP8X_CHUNK_SIZE },
   { MKFOURCC('I', 'C', 'C', 'P'),  WEBP_CHUNK_ICCP,    UNDEFINED_CHUNK_SIZE },
-  { MKFOURCC('L', 'O', 'O', 'P'),  WEBP_CHUNK_LOOP,    LOOP_CHUNK_SIZE },
-  { MKFOURCC('F', 'R', 'M', ' '),  WEBP_CHUNK_FRAME,   FRAME_CHUNK_SIZE },
-  { MKFOURCC('T', 'I', 'L', 'E'),  WEBP_CHUNK_TILE,    TILE_CHUNK_SIZE },
+  { MKFOURCC('A', 'N', 'I', 'M'),  WEBP_CHUNK_ANIM,    ANIM_CHUNK_SIZE },
+  { MKFOURCC('A', 'N', 'M', 'F'),  WEBP_CHUNK_ANMF,    ANMF_CHUNK_SIZE },
+  { MKFOURCC('F', 'R', 'G', 'M'),  WEBP_CHUNK_FRGM,    FRGM_CHUNK_SIZE },
   { MKFOURCC('A', 'L', 'P', 'H'),  WEBP_CHUNK_ALPHA,   UNDEFINED_CHUNK_SIZE },
   { MKFOURCC('V', 'P', '8', ' '),  WEBP_CHUNK_IMAGE,   UNDEFINED_CHUNK_SIZE },
   { MKFOURCC('V', 'P', '8', 'L'),  WEBP_CHUNK_IMAGE,   UNDEFINED_CHUNK_SIZE },
-  { MKFOURCC('M', 'E', 'T', 'A'),  WEBP_CHUNK_META,    UNDEFINED_CHUNK_SIZE },
+  { MKFOURCC('E', 'X', 'I', 'F'),  WEBP_CHUNK_EXIF,    UNDEFINED_CHUNK_SIZE },
+  { MKFOURCC('X', 'M', 'P', ' '),  WEBP_CHUNK_XMP,     UNDEFINED_CHUNK_SIZE },
   { MKFOURCC('U', 'N', 'K', 'N'),  WEBP_CHUNK_UNKNOWN, UNDEFINED_CHUNK_SIZE },
 
-  { NIL_TAG,                    WEBP_CHUNK_NIL,     UNDEFINED_CHUNK_SIZE }
+  { NIL_TAG,                       WEBP_CHUNK_NIL,     UNDEFINED_CHUNK_SIZE }
 };
 
+//------------------------------------------------------------------------------
+
+int WebPGetMuxVersion(void) {
+  return (MUX_MAJ_VERSION << 16) | (MUX_MIN_VERSION << 8) | MUX_REV_VERSION;
+}
+
 //------------------------------------------------------------------------------
 // Life of a chunk object.
 
@@ -73,12 +83,22 @@ WebPChunkId ChunkGetIdFromTag(uint32_t tag) {
   return WEBP_CHUNK_NIL;
 }
 
+uint32_t ChunkGetTagFromFourCC(const char fourcc[4]) {
+  return MKFOURCC(fourcc[0], fourcc[1], fourcc[2], fourcc[3]);
+}
+
+CHUNK_INDEX ChunkGetIndexFromFourCC(const char fourcc[4]) {
+  const uint32_t tag = ChunkGetTagFromFourCC(fourcc);
+  const CHUNK_INDEX idx = ChunkGetIndexFromTag(tag);
+  return (idx == IDX_NIL) ? IDX_UNKNOWN : idx;
+}
+
 //------------------------------------------------------------------------------
 // Chunk search methods.
 
 // Returns next chunk in the chunk list with the given tag.
 static WebPChunk* ChunkSearchNextInList(WebPChunk* chunk, uint32_t tag) {
-  while (chunk && chunk->tag_ != tag) {
+  while (chunk != NULL && chunk->tag_ != tag) {
     chunk = chunk->next_;
   }
   return chunk;
@@ -87,7 +107,7 @@ static WebPChunk* ChunkSearchNextInList(WebPChunk* chunk, uint32_t tag) {
 WebPChunk* ChunkSearchList(WebPChunk* first, uint32_t nth, uint32_t tag) {
   uint32_t iter = nth;
   first = ChunkSearchNextInList(first, tag);
-  if (!first) return NULL;
+  if (first == NULL) return NULL;
 
   while (--iter != 0) {
     WebPChunk* next_chunk = ChunkSearchNextInList(first->next_, tag);
@@ -99,14 +119,14 @@ WebPChunk* ChunkSearchList(WebPChunk* first, uint32_t nth, uint32_t tag) {
 
 // Outputs a pointer to 'prev_chunk->next_',
 //   where 'prev_chunk' is the pointer to the chunk at position (nth - 1).
-// Returns 1 if nth chunk was found, 0 otherwise.
+// Returns true if nth chunk was found.
 static int ChunkSearchListToSet(WebPChunk** chunk_list, uint32_t nth,
                                 WebPChunk*** const location) {
   uint32_t count = 0;
-  assert(chunk_list);
+  assert(chunk_list != NULL);
   *location = chunk_list;
 
-  while (*chunk_list) {
+  while (*chunk_list != NULL) {
     WebPChunk* const cur_chunk = *chunk_list;
     ++count;
     if (count == nth) return 1;  // Found.
@@ -124,34 +144,25 @@ static int ChunkSearchListToSet(WebPChunk** chunk_list, uint32_t nth,
 WebPMuxError ChunkAssignData(WebPChunk* chunk, const WebPData* const data,
                              int copy_data, uint32_t tag) {
   // For internally allocated chunks, always copy data & make it owner of data.
-  if (tag == kChunks[IDX_VP8X].tag || tag == kChunks[IDX_LOOP].tag) {
+  if (tag == kChunks[IDX_VP8X].tag || tag == kChunks[IDX_ANIM].tag) {
     copy_data = 1;
   }
 
   ChunkRelease(chunk);
 
   if (data != NULL) {
-    if (copy_data) {
-      // Copy data.
-      chunk->data_.bytes_ = (uint8_t*)malloc(data->size_);
-      if (chunk->data_.bytes_ == NULL) return WEBP_MUX_MEMORY_ERROR;
-      memcpy((uint8_t*)chunk->data_.bytes_, data->bytes_, data->size_);
-      chunk->data_.size_ = data->size_;
-
-      // Chunk is owner of data.
-      chunk->owner_ = 1;
-    } else {
-      // Don't copy data.
+    if (copy_data) {        // Copy data.
+      if (!WebPDataCopy(data, &chunk->data_)) return WEBP_MUX_MEMORY_ERROR;
+      chunk->owner_ = 1;    // Chunk is owner of data.
+    } else {                // Don't copy data.
       chunk->data_ = *data;
     }
   }
-
   chunk->tag_ = tag;
-
   return WEBP_MUX_OK;
 }
 
-WebPMuxError ChunkSetNth(const WebPChunk* chunk, WebPChunk** chunk_list,
+WebPMuxError ChunkSetNth(WebPChunk* chunk, WebPChunk** chunk_list,
                          uint32_t nth) {
   WebPChunk* new_chunk;
 
@@ -162,6 +173,7 @@ WebPMuxError ChunkSetNth(const WebPChunk* chunk, WebPChunk** chunk_list,
   new_chunk = (WebPChunk*)malloc(sizeof(*new_chunk));
   if (new_chunk == NULL) return WEBP_MUX_MEMORY_ERROR;
   *new_chunk = *chunk;
+  chunk->owner_ = 0;
   new_chunk->next_ = *chunk_list;
   *chunk_list = new_chunk;
   return WEBP_MUX_OK;
@@ -181,7 +193,7 @@ WebPChunk* ChunkDelete(WebPChunk* const chunk) {
 
 size_t ChunksListDiskSize(const WebPChunk* chunk_list) {
   size_t size = 0;
-  while (chunk_list) {
+  while (chunk_list != NULL) {
     size += ChunkDiskSize(chunk_list);
     chunk_list = chunk_list->next_;
   }
@@ -189,55 +201,26 @@ size_t ChunksListDiskSize(const WebPChunk* chunk_list) {
 }
 
 static uint8_t* ChunkEmit(const WebPChunk* const chunk, uint8_t* dst) {
-  const size_t chunk_size = chunk->data_.size_;
+  const size_t chunk_size = chunk->data_.size;
   assert(chunk);
   assert(chunk->tag_ != NIL_TAG);
   PutLE32(dst + 0, chunk->tag_);
   PutLE32(dst + TAG_SIZE, (uint32_t)chunk_size);
   assert(chunk_size == (uint32_t)chunk_size);
-  memcpy(dst + CHUNK_HEADER_SIZE, chunk->data_.bytes_, chunk_size);
+  memcpy(dst + CHUNK_HEADER_SIZE, chunk->data_.bytes, chunk_size);
   if (chunk_size & 1)
     dst[CHUNK_HEADER_SIZE + chunk_size] = 0;  // Add padding.
   return dst + ChunkDiskSize(chunk);
 }
 
 uint8_t* ChunkListEmit(const WebPChunk* chunk_list, uint8_t* dst) {
-  while (chunk_list) {
+  while (chunk_list != NULL) {
     dst = ChunkEmit(chunk_list, dst);
     chunk_list = chunk_list->next_;
   }
   return dst;
 }
 
-//------------------------------------------------------------------------------
-// Manipulation of a WebPData object.
-
-void WebPDataInit(WebPData* webp_data) {
-  if (webp_data != NULL) {
-    memset(webp_data, 0, sizeof(*webp_data));
-  }
-}
-
-void WebPDataClear(WebPData* webp_data) {
-  if (webp_data != NULL) {
-    free((void*)webp_data->bytes_);
-    WebPDataInit(webp_data);
-  }
-}
-
-int WebPDataCopy(const WebPData* src, WebPData* dst) {
-  if (src == NULL || dst == NULL) return 0;
-
-  WebPDataInit(dst);
-  if (src->bytes_ != NULL && src->size_ != 0) {
-    dst->bytes_ = (uint8_t*)malloc(src->size_);
-    if (dst->bytes_ == NULL) return 0;
-    memcpy((void*)dst->bytes_, src->bytes_, src->size_);
-    dst->size_ = src->size_;
-  }
-  return 1;
-}
-
 //------------------------------------------------------------------------------
 // Life of a MuxImage object.
 
@@ -265,10 +248,14 @@ int MuxImageCount(const WebPMuxImage* wpi_list, WebPChunkId id) {
   int count = 0;
   const WebPMuxImage* current;
   for (current = wpi_list; current != NULL; current = current->next_) {
-    const WebPChunk* const wpi_chunk = *MuxImageGetListFromId(current, id);
-    if (wpi_chunk != NULL) {
-      const WebPChunkId wpi_chunk_id = ChunkGetIdFromTag(wpi_chunk->tag_);
-      if (wpi_chunk_id == id) ++count;
+    if (id == WEBP_CHUNK_NIL) {
+      ++count;  // Special case: count all images.
+    } else {
+      const WebPChunk* const wpi_chunk = *MuxImageGetListFromId(current, id);
+      if (wpi_chunk != NULL) {
+        const WebPChunkId wpi_chunk_id = ChunkGetIdFromTag(wpi_chunk->tag_);
+        if (wpi_chunk_id == id) ++count;  // Count images with a matching 'id'.
+      }
     }
   }
   return count;
@@ -276,34 +263,22 @@ int MuxImageCount(const WebPMuxImage* wpi_list, WebPChunkId id) {
 
 // Outputs a pointer to 'prev_wpi->next_',
 //   where 'prev_wpi' is the pointer to the image at position (nth - 1).
-// Returns 1 if nth image with given id was found, 0 otherwise.
+// Returns true if nth image was found.
 static int SearchImageToGetOrDelete(WebPMuxImage** wpi_list, uint32_t nth,
-                                    WebPChunkId id,
                                     WebPMuxImage*** const location) {
   uint32_t count = 0;
   assert(wpi_list);
   *location = wpi_list;
 
-  // Search makes sense only for the following.
-  assert(id == WEBP_CHUNK_FRAME || id == WEBP_CHUNK_TILE ||
-         id == WEBP_CHUNK_IMAGE);
-  assert(id != WEBP_CHUNK_IMAGE || nth == 1);
-
   if (nth == 0) {
-    nth = MuxImageCount(*wpi_list, id);
+    nth = MuxImageCount(*wpi_list, WEBP_CHUNK_NIL);
     if (nth == 0) return 0;  // Not found.
   }
 
-  while (*wpi_list) {
+  while (*wpi_list != NULL) {
     WebPMuxImage* const cur_wpi = *wpi_list;
-    const WebPChunk* const wpi_chunk = *MuxImageGetListFromId(cur_wpi, id);
-    if (wpi_chunk != NULL) {
-      const WebPChunkId wpi_chunk_id = ChunkGetIdFromTag(wpi_chunk->tag_);
-      if (wpi_chunk_id == id) {
-        ++count;
-        if (count == nth) return 1;  // Found.
-      }
-    }
+    ++count;
+    if (count == nth) return 1;  // Found.
     wpi_list = &cur_wpi->next_;
     *location = wpi_list;
   }
@@ -346,15 +321,14 @@ WebPMuxImage* MuxImageDelete(WebPMuxImage* const wpi) {
 }
 
 void MuxImageDeleteAll(WebPMuxImage** const wpi_list) {
-  while (*wpi_list) {
+  while (*wpi_list != NULL) {
     *wpi_list = MuxImageDelete(*wpi_list);
   }
 }
 
-WebPMuxError MuxImageDeleteNth(WebPMuxImage** wpi_list, uint32_t nth,
-                               WebPChunkId id) {
+WebPMuxError MuxImageDeleteNth(WebPMuxImage** wpi_list, uint32_t nth) {
   assert(wpi_list);
-  if (!SearchImageToGetOrDelete(wpi_list, nth, id, &wpi_list)) {
+  if (!SearchImageToGetOrDelete(wpi_list, nth, &wpi_list)) {
     return WEBP_MUX_NOT_FOUND;
   }
   *wpi_list = MuxImageDelete(*wpi_list);
@@ -365,10 +339,10 @@ WebPMuxError MuxImageDeleteNth(WebPMuxImage** wpi_list, uint32_t nth,
 // MuxImage reader methods.
 
 WebPMuxError MuxImageGetNth(const WebPMuxImage** wpi_list, uint32_t nth,
-                            WebPChunkId id, WebPMuxImage** wpi) {
+                            WebPMuxImage** wpi) {
   assert(wpi_list);
   assert(wpi);
-  if (!SearchImageToGetOrDelete((WebPMuxImage**)wpi_list, nth, id,
+  if (!SearchImageToGetOrDelete((WebPMuxImage**)wpi_list, nth,
                                 (WebPMuxImage***)&wpi_list)) {
     return WEBP_MUX_NOT_FOUND;
   }
@@ -390,27 +364,46 @@ size_t MuxImageDiskSize(const WebPMuxImage* const wpi) {
 
 size_t MuxImageListDiskSize(const WebPMuxImage* wpi_list) {
   size_t size = 0;
-  while (wpi_list) {
+  while (wpi_list != NULL) {
     size += MuxImageDiskSize(wpi_list);
     wpi_list = wpi_list->next_;
   }
   return size;
 }
 
+// Special case as ANMF/FRGM chunk encapsulates other image chunks.
+static uint8_t* ChunkEmitSpecial(const WebPChunk* const header,
+                                 size_t total_size, uint8_t* dst) {
+  const size_t header_size = header->data_.size;
+  const size_t offset_to_next = total_size - CHUNK_HEADER_SIZE;
+  assert(header->tag_ == kChunks[IDX_ANMF].tag ||
+         header->tag_ == kChunks[IDX_FRGM].tag);
+  PutLE32(dst + 0, header->tag_);
+  PutLE32(dst + TAG_SIZE, (uint32_t)offset_to_next);
+  assert(header_size == (uint32_t)header_size);
+  memcpy(dst + CHUNK_HEADER_SIZE, header->data_.bytes, header_size);
+  if (header_size & 1) {
+    dst[CHUNK_HEADER_SIZE + header_size] = 0;  // Add padding.
+  }
+  return dst + ChunkDiskSize(header);
+}
+
 uint8_t* MuxImageEmit(const WebPMuxImage* const wpi, uint8_t* dst) {
   // Ordering of chunks to be emitted is strictly as follows:
-  // 1. Frame/Tile chunk (if present).
-  // 2. Alpha chunk (if present).
+  // 1. ANMF/FRGM chunk (if present).
+  // 2. ALPH chunk (if present).
   // 3. VP8/VP8L chunk.
   assert(wpi);
-  if (wpi->header_ != NULL) dst = ChunkEmit(wpi->header_, dst);
+  if (wpi->header_ != NULL) {
+    dst = ChunkEmitSpecial(wpi->header_, MuxImageDiskSize(wpi), dst);
+  }
   if (wpi->alpha_ != NULL) dst = ChunkEmit(wpi->alpha_, dst);
   if (wpi->img_ != NULL) dst = ChunkEmit(wpi->img_, dst);
   return dst;
 }
 
 uint8_t* MuxImageListEmit(const WebPMuxImage* wpi_list, uint8_t* dst) {
-  while (wpi_list) {
+  while (wpi_list != NULL) {
     dst = MuxImageEmit(wpi_list, dst);
     wpi_list = wpi_list->next_;
   }
@@ -441,11 +434,12 @@ uint8_t* MuxEmitRiffHeader(uint8_t* const data, size_t size) {
 
 WebPChunk** MuxGetChunkListFromId(const WebPMux* mux, WebPChunkId id) {
   assert(mux != NULL);
-  switch(id) {
+  switch (id) {
     case WEBP_CHUNK_VP8X:    return (WebPChunk**)&mux->vp8x_;
     case WEBP_CHUNK_ICCP:    return (WebPChunk**)&mux->iccp_;
-    case WEBP_CHUNK_LOOP:    return (WebPChunk**)&mux->loop_;
-    case WEBP_CHUNK_META:    return (WebPChunk**)&mux->meta_;
+    case WEBP_CHUNK_ANIM:    return (WebPChunk**)&mux->anim_;
+    case WEBP_CHUNK_EXIF:    return (WebPChunk**)&mux->exif_;
+    case WEBP_CHUNK_XMP:     return (WebPChunk**)&mux->xmp_;
     case WEBP_CHUNK_UNKNOWN: return (WebPChunk**)&mux->unknown_;
     default: return NULL;
   }
@@ -453,17 +447,17 @@ WebPChunk** MuxGetChunkListFromId(const WebPMux* mux, WebPChunkId id) {
 
 WebPMuxError MuxValidateForImage(const WebPMux* const mux) {
   const int num_images = MuxImageCount(mux->images_, WEBP_CHUNK_IMAGE);
-  const int num_frames = MuxImageCount(mux->images_, WEBP_CHUNK_FRAME);
-  const int num_tiles  = MuxImageCount(mux->images_, WEBP_CHUNK_TILE);
+  const int num_frames = MuxImageCount(mux->images_, WEBP_CHUNK_ANMF);
+  const int num_fragments = MuxImageCount(mux->images_, WEBP_CHUNK_FRGM);
 
   if (num_images == 0) {
     // No images in mux.
     return WEBP_MUX_NOT_FOUND;
-  } else if (num_images == 1 && num_frames == 0 && num_tiles == 0) {
+  } else if (num_images == 1 && num_frames == 0 && num_fragments == 0) {
     // Valid case (single image).
     return WEBP_MUX_OK;
   } else {
-    // Frame/Tile case OR an invalid mux.
+    // Frame/Fragment case OR an invalid mux.
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 }
@@ -494,10 +488,11 @@ static WebPMuxError ValidateChunk(const WebPMux* const mux, CHUNK_INDEX idx,
 
 WebPMuxError MuxValidate(const WebPMux* const mux) {
   int num_iccp;
-  int num_meta;
-  int num_loop_chunks;
+  int num_exif;
+  int num_xmp;
+  int num_anim;
   int num_frames;
-  int num_tiles;
+  int num_fragments;
   int num_vp8x;
   int num_images;
   int num_alpha;
@@ -517,29 +512,33 @@ WebPMuxError MuxValidate(const WebPMux* const mux) {
   err = ValidateChunk(mux, IDX_ICCP, ICCP_FLAG, flags, 1, &num_iccp);
   if (err != WEBP_MUX_OK) return err;
 
-  // At most one XMP metadata.
-  err = ValidateChunk(mux, IDX_META, META_FLAG, flags, 1, &num_meta);
+  // At most one EXIF metadata.
+  err = ValidateChunk(mux, IDX_EXIF, EXIF_FLAG, flags, 1, &num_exif);
   if (err != WEBP_MUX_OK) return err;
 
-  // Animation: ANIMATION_FLAG, loop chunk and frame chunk(s) are consistent.
-  // At most one loop chunk.
-  err = ValidateChunk(mux, IDX_LOOP, NO_FLAG, flags, 1, &num_loop_chunks);
+  // At most one XMP metadata.
+  err = ValidateChunk(mux, IDX_XMP, XMP_FLAG, flags, 1, &num_xmp);
   if (err != WEBP_MUX_OK) return err;
-  err = ValidateChunk(mux, IDX_FRAME, NO_FLAG, flags, -1, &num_frames);
+
+  // Animation: ANIMATION_FLAG, ANIM chunk and ANMF chunk(s) are consistent.
+  // At most one ANIM chunk.
+  err = ValidateChunk(mux, IDX_ANIM, NO_FLAG, flags, 1, &num_anim);
+  if (err != WEBP_MUX_OK) return err;
+  err = ValidateChunk(mux, IDX_ANMF, NO_FLAG, flags, -1, &num_frames);
   if (err != WEBP_MUX_OK) return err;
 
   {
     const int has_animation = !!(flags & ANIMATION_FLAG);
-    if (has_animation && (num_loop_chunks == 0 || num_frames == 0)) {
+    if (has_animation && (num_anim == 0 || num_frames == 0)) {
       return WEBP_MUX_INVALID_ARGUMENT;
     }
-    if (!has_animation && (num_loop_chunks == 1 || num_frames > 0)) {
+    if (!has_animation && (num_anim == 1 || num_frames > 0)) {
       return WEBP_MUX_INVALID_ARGUMENT;
     }
   }
 
-  // Tiling: TILE_FLAG and tile chunk(s) are consistent.
-  err = ValidateChunk(mux, IDX_TILE, TILE_FLAG, flags, -1, &num_tiles);
+  // Fragmentation: FRAGMENTS_FLAG and FRGM chunk(s) are consistent.
+  err = ValidateChunk(mux, IDX_FRGM, FRAGMENTS_FLAG, flags, -1, &num_fragments);
   if (err != WEBP_MUX_OK) return err;
 
   // Verify either VP8X chunk is present OR there is only one elem in
@@ -551,16 +550,18 @@ WebPMuxError MuxValidate(const WebPMux* const mux) {
   if (num_vp8x == 0 && num_images != 1) return WEBP_MUX_INVALID_ARGUMENT;
 
   // ALPHA_FLAG & alpha chunk(s) are consistent.
-  if (num_vp8x > 0 && MuxHasLosslessImages(mux->images_)) {
-    // Special case: we have a VP8X chunk as well as some lossless images.
-    if (!(flags & ALPHA_FLAG)) return WEBP_MUX_INVALID_ARGUMENT;
+  if (MuxHasLosslessImages(mux->images_)) {
+    if (num_vp8x > 0) {
+      // Special case: we have a VP8X chunk as well as some lossless images.
+      if (!(flags & ALPHA_FLAG)) return WEBP_MUX_INVALID_ARGUMENT;
+    }
   } else {
-    err = ValidateChunk(mux, IDX_ALPHA, ALPHA_FLAG, flags, -1, &num_alpha);
-    if (err != WEBP_MUX_OK) return err;
+      err = ValidateChunk(mux, IDX_ALPHA, ALPHA_FLAG, flags, -1, &num_alpha);
+      if (err != WEBP_MUX_OK) return err;
   }
 
-  // num_tiles & num_images are consistent.
-  if (num_tiles > 0 && num_images != num_tiles) {
+  // num_fragments & num_images are consistent.
+  if (num_fragments > 0 && num_images != num_fragments) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 
diff --git a/3rdparty/libwebp/mux/muxread.c b/3rdparty/libwebp/mux/muxread.c
index 21c3cfbaeb..2179ccb79e 100644
--- a/3rdparty/libwebp/mux/muxread.c
+++ b/3rdparty/libwebp/mux/muxread.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Read APIs for mux.
@@ -12,6 +14,7 @@
 
 #include <assert.h>
 #include "./muxi.h"
+#include "../utils/utils.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -41,8 +44,9 @@ static WebPMuxError MuxGet(const WebPMux* const mux, CHUNK_INDEX idx,
 
   SWITCH_ID_LIST(IDX_VP8X, mux->vp8x_);
   SWITCH_ID_LIST(IDX_ICCP, mux->iccp_);
-  SWITCH_ID_LIST(IDX_LOOP, mux->loop_);
-  SWITCH_ID_LIST(IDX_META, mux->meta_);
+  SWITCH_ID_LIST(IDX_ANIM, mux->anim_);
+  SWITCH_ID_LIST(IDX_EXIF, mux->exif_);
+  SWITCH_ID_LIST(IDX_XMP, mux->xmp_);
   SWITCH_ID_LIST(IDX_UNKNOWN, mux->unknown_);
   return WEBP_MUX_NOT_FOUND;
 }
@@ -50,10 +54,9 @@ static WebPMuxError MuxGet(const WebPMux* const mux, CHUNK_INDEX idx,
 
 // Fill the chunk with the given data (includes chunk header bytes), after some
 // verifications.
-static WebPMuxError ChunkVerifyAndAssignData(WebPChunk* chunk,
-                                             const uint8_t* data,
-                                             size_t data_size, size_t riff_size,
-                                             int copy_data) {
+static WebPMuxError ChunkVerifyAndAssign(WebPChunk* chunk,
+                                         const uint8_t* data, size_t data_size,
+                                         size_t riff_size, int copy_data) {
   uint32_t chunk_size;
   WebPData chunk_data;
 
@@ -68,11 +71,74 @@ static WebPMuxError ChunkVerifyAndAssignData(WebPChunk* chunk,
   }
 
   // Data assignment.
-  chunk_data.bytes_ = data + CHUNK_HEADER_SIZE;
-  chunk_data.size_ = chunk_size;
+  chunk_data.bytes = data + CHUNK_HEADER_SIZE;
+  chunk_data.size = chunk_size;
   return ChunkAssignData(chunk, &chunk_data, copy_data, GetLE32(data + 0));
 }
 
+static int MuxImageParse(const WebPChunk* const chunk, int copy_data,
+                         WebPMuxImage* const wpi) {
+  const uint8_t* bytes = chunk->data_.bytes;
+  size_t size = chunk->data_.size;
+  const uint8_t* const last = bytes + size;
+  WebPChunk subchunk;
+  size_t subchunk_size;
+  ChunkInit(&subchunk);
+
+  assert(chunk->tag_ == kChunks[IDX_ANMF].tag ||
+         chunk->tag_ == kChunks[IDX_FRGM].tag);
+  assert(!wpi->is_partial_);
+
+  // ANMF/FRGM.
+  {
+    const size_t hdr_size = (chunk->tag_ == kChunks[IDX_ANMF].tag) ?
+        ANMF_CHUNK_SIZE : FRGM_CHUNK_SIZE;
+    const WebPData temp = { bytes, hdr_size };
+    // Each of ANMF and FRGM chunk contain a header at the beginning. So, its
+    // size should at least be 'hdr_size'.
+    if (size < hdr_size) goto Fail;
+    ChunkAssignData(&subchunk, &temp, copy_data, chunk->tag_);
+  }
+  ChunkSetNth(&subchunk, &wpi->header_, 1);
+  wpi->is_partial_ = 1;  // Waiting for ALPH and/or VP8/VP8L chunks.
+
+  // Rest of the chunks.
+  subchunk_size = ChunkDiskSize(&subchunk) - CHUNK_HEADER_SIZE;
+  bytes += subchunk_size;
+  size -= subchunk_size;
+
+  while (bytes != last) {
+    ChunkInit(&subchunk);
+    if (ChunkVerifyAndAssign(&subchunk, bytes, size, size,
+                             copy_data) != WEBP_MUX_OK) {
+      goto Fail;
+    }
+    switch (ChunkGetIdFromTag(subchunk.tag_)) {
+      case WEBP_CHUNK_ALPHA:
+        if (wpi->alpha_ != NULL) goto Fail;  // Consecutive ALPH chunks.
+        if (ChunkSetNth(&subchunk, &wpi->alpha_, 1) != WEBP_MUX_OK) goto Fail;
+        wpi->is_partial_ = 1;  // Waiting for a VP8 chunk.
+        break;
+      case WEBP_CHUNK_IMAGE:
+        if (ChunkSetNth(&subchunk, &wpi->img_, 1) != WEBP_MUX_OK) goto Fail;
+        wpi->is_partial_ = 0;  // wpi is completely filled.
+        break;
+      default:
+        goto Fail;
+        break;
+    }
+    subchunk_size = ChunkDiskSize(&subchunk);
+    bytes += subchunk_size;
+    size -= subchunk_size;
+  }
+  if (wpi->is_partial_) goto Fail;
+  return 1;
+
+ Fail:
+  ChunkRelease(&subchunk);
+  return 0;
+}
+
 //------------------------------------------------------------------------------
 // Create a mux object from WebP-RIFF data.
 
@@ -94,8 +160,8 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data,
   }
   if (bitstream == NULL) return NULL;
 
-  data = bitstream->bytes_;
-  size = bitstream->size_;
+  data = bitstream->bytes;
+  size = bitstream->size;
 
   if (data == NULL) return NULL;
   if (size < RIFF_HEADER_SIZE) return NULL;
@@ -135,42 +201,48 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data,
 
   // Loop over chunks.
   while (data != end) {
+    size_t data_size;
     WebPChunkId id;
-    WebPMuxError err;
-
-    err = ChunkVerifyAndAssignData(&chunk, data, size, riff_size, copy_data);
-    if (err != WEBP_MUX_OK) goto Err;
-
+    WebPChunk** chunk_list;
+    if (ChunkVerifyAndAssign(&chunk, data, size, riff_size,
+                             copy_data) != WEBP_MUX_OK) {
+      goto Err;
+    }
+    data_size = ChunkDiskSize(&chunk);
     id = ChunkGetIdFromTag(chunk.tag_);
-
-    if (IsWPI(id)) {  // An image chunk (frame/tile/alpha/vp8).
-      WebPChunk** wpi_chunk_ptr =
-          MuxImageGetListFromId(wpi, id);  // Image chunk to set.
-      assert(wpi_chunk_ptr != NULL);
-      if (*wpi_chunk_ptr != NULL) goto Err;  // Consecutive alpha chunks or
-                                             // consecutive frame/tile chunks.
-      if (ChunkSetNth(&chunk, wpi_chunk_ptr, 1) != WEBP_MUX_OK) goto Err;
-      if (id == WEBP_CHUNK_IMAGE) {
+    switch (id) {
+      case WEBP_CHUNK_ALPHA:
+        if (wpi->alpha_ != NULL) goto Err;  // Consecutive ALPH chunks.
+        if (ChunkSetNth(&chunk, &wpi->alpha_, 1) != WEBP_MUX_OK) goto Err;
+        wpi->is_partial_ = 1;  // Waiting for a VP8 chunk.
+        break;
+      case WEBP_CHUNK_IMAGE:
+        if (ChunkSetNth(&chunk, &wpi->img_, 1) != WEBP_MUX_OK) goto Err;
         wpi->is_partial_ = 0;  // wpi is completely filled.
+ PushImage:
         // Add this to mux->images_ list.
         if (MuxImagePush(wpi, &mux->images_) != WEBP_MUX_OK) goto Err;
         MuxImageInit(wpi);  // Reset for reading next image.
-      } else {
-        wpi->is_partial_ = 1;  // wpi is only partially filled.
-      }
-    } else {  // A non-image chunk.
-      WebPChunk** chunk_list;
-      if (wpi->is_partial_) goto Err;  // Encountered a non-image chunk before
-                                       // getting all chunks of an image.
-      chunk_list = MuxGetChunkListFromId(mux, id);  // List to add this chunk.
-      if (chunk_list == NULL) chunk_list = &mux->unknown_;
-      if (ChunkSetNth(&chunk, chunk_list, 0) != WEBP_MUX_OK) goto Err;
-    }
-    {
-      const size_t data_size = ChunkDiskSize(&chunk);
-      data += data_size;
-      size -= data_size;
+        break;
+      case WEBP_CHUNK_ANMF:
+#ifdef WEBP_EXPERIMENTAL_FEATURES
+      case WEBP_CHUNK_FRGM:
+#endif
+        if (wpi->is_partial_) goto Err;  // Previous wpi is still incomplete.
+        if (!MuxImageParse(&chunk, copy_data, wpi)) goto Err;
+        ChunkRelease(&chunk);
+        goto PushImage;
+        break;
+      default:  // A non-image chunk.
+        if (wpi->is_partial_) goto Err;  // Encountered a non-image chunk before
+                                         // getting all chunks of an image.
+        chunk_list = MuxGetChunkListFromId(mux, id);  // List to add this chunk.
+        if (chunk_list == NULL) chunk_list = &mux->unknown_;
+        if (ChunkSetNth(&chunk, chunk_list, 0) != WEBP_MUX_OK) goto Err;
+        break;
     }
+    data += data_size;
+    size -= data_size;
     ChunkInit(&chunk);
   }
 
@@ -192,26 +264,30 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data,
 
 WebPMuxError WebPMuxGetFeatures(const WebPMux* mux, uint32_t* flags) {
   WebPData data;
-  WebPMuxError err;
 
   if (mux == NULL || flags == NULL) return WEBP_MUX_INVALID_ARGUMENT;
   *flags = 0;
 
   // Check if VP8X chunk is present.
-  err = MuxGet(mux, IDX_VP8X, 1, &data);
-  if (err == WEBP_MUX_NOT_FOUND) {
-    // Check if VP8/VP8L chunk is present.
-    err = WebPMuxGetImage(mux, &data);
-    WebPDataClear(&data);
-    return err;
-  } else if (err != WEBP_MUX_OK) {
-    return err;
+  if (MuxGet(mux, IDX_VP8X, 1, &data) == WEBP_MUX_OK) {
+    if (data.size < CHUNK_SIZE_BYTES) return WEBP_MUX_BAD_DATA;
+    *flags = GetLE32(data.bytes);  // All OK. Fill up flags.
+  } else {
+    WebPMuxError err = MuxValidateForImage(mux);  // Check for single image.
+    if (err != WEBP_MUX_OK) return err;
+    if (MuxHasLosslessImages(mux->images_)) {
+      const WebPData* const vp8l_data = &mux->images_->img_->data_;
+      int has_alpha = 0;
+      if (!VP8LGetInfo(vp8l_data->bytes, vp8l_data->size, NULL, NULL,
+                       &has_alpha)) {
+        return WEBP_MUX_BAD_DATA;
+      }
+      if (has_alpha) {
+        *flags = ALPHA_FLAG;
+      }
+    }
   }
 
-  if (data.size_ < CHUNK_SIZE_BYTES) return WEBP_MUX_BAD_DATA;
-
-  // All OK. Fill up flags.
-  *flags = GetLE32(data.bytes_);
   return WEBP_MUX_OK;
 }
 
@@ -230,7 +306,7 @@ static uint8_t* EmitVP8XChunk(uint8_t* const dst, int width,
 }
 
 // Assemble a single image WebP bitstream from 'wpi'.
-static WebPMuxError SynthesizeBitstream(WebPMuxImage* const wpi,
+static WebPMuxError SynthesizeBitstream(const WebPMuxImage* const wpi,
                                         WebPData* const bitstream) {
   uint8_t* dst;
 
@@ -238,7 +314,7 @@ static WebPMuxError SynthesizeBitstream(WebPMuxImage* const wpi,
   const int need_vp8x = (wpi->alpha_ != NULL);
   const size_t vp8x_size = need_vp8x ? CHUNK_HEADER_SIZE + VP8X_CHUNK_SIZE : 0;
   const size_t alpha_size = need_vp8x ? ChunkDiskSize(wpi->alpha_) : 0;
-  // Note: No need to output FRM/TILE chunk for a single image.
+  // Note: No need to output ANMF/FRGM chunk for a single image.
   const size_t size = RIFF_HEADER_SIZE + vp8x_size + alpha_size +
                       ChunkDiskSize(wpi->img_);
   uint8_t* const data = (uint8_t*)malloc(size);
@@ -265,100 +341,101 @@ static WebPMuxError SynthesizeBitstream(WebPMuxImage* const wpi,
   assert(dst == data + size);
 
   // Output.
-  bitstream->bytes_ = data;
-  bitstream->size_ = size;
+  bitstream->bytes = data;
+  bitstream->size = size;
   return WEBP_MUX_OK;
 }
 
-WebPMuxError WebPMuxGetImage(const WebPMux* mux, WebPData* bitstream) {
-  WebPMuxError err;
-  WebPMuxImage* wpi = NULL;
-
-  if (mux == NULL || bitstream == NULL) {
+WebPMuxError WebPMuxGetChunk(const WebPMux* mux, const char fourcc[4],
+                             WebPData* chunk_data) {
+  CHUNK_INDEX idx;
+  if (mux == NULL || fourcc == NULL || chunk_data == NULL) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
-
-  err = MuxValidateForImage(mux);
-  if (err != WEBP_MUX_OK) return err;
-
-  // All well. Get the image.
-  err = MuxImageGetNth((const WebPMuxImage**)&mux->images_, 1, WEBP_CHUNK_IMAGE,
-                       &wpi);
-  assert(err == WEBP_MUX_OK);  // Already tested above.
-
-  return SynthesizeBitstream(wpi, bitstream);
+  idx = ChunkGetIndexFromFourCC(fourcc);
+  if (IsWPI(kChunks[idx].id)) {     // An image chunk.
+    return WEBP_MUX_INVALID_ARGUMENT;
+  } else if (idx != IDX_UNKNOWN) {  // A known chunk type.
+    return MuxGet(mux, idx, 1, chunk_data);
+  } else {                          // An unknown chunk type.
+    const WebPChunk* const chunk =
+        ChunkSearchList(mux->unknown_, 1, ChunkGetTagFromFourCC(fourcc));
+    if (chunk == NULL) return WEBP_MUX_NOT_FOUND;
+    *chunk_data = chunk->data_;
+    return WEBP_MUX_OK;
+  }
 }
 
-WebPMuxError WebPMuxGetMetadata(const WebPMux* mux, WebPData* metadata) {
-  if (mux == NULL || metadata == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-  return MuxGet(mux, IDX_META, 1, metadata);
+static WebPMuxError MuxGetImageInternal(const WebPMuxImage* const wpi,
+                                        WebPMuxFrameInfo* const info) {
+  // Set some defaults for unrelated fields.
+  info->x_offset = 0;
+  info->y_offset = 0;
+  info->duration = 1;
+  // Extract data for related fields.
+  info->id = ChunkGetIdFromTag(wpi->img_->tag_);
+  return SynthesizeBitstream(wpi, &info->bitstream);
 }
 
-WebPMuxError WebPMuxGetColorProfile(const WebPMux* mux,
-                                    WebPData* color_profile) {
-  if (mux == NULL || color_profile == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-  return MuxGet(mux, IDX_ICCP, 1, color_profile);
+static WebPMuxError MuxGetFrameFragmentInternal(const WebPMuxImage* const wpi,
+                                                WebPMuxFrameInfo* const frame) {
+  const int is_frame = (wpi->header_->tag_ == kChunks[IDX_ANMF].tag);
+  const CHUNK_INDEX idx = is_frame ? IDX_ANMF : IDX_FRGM;
+  const WebPData* frame_frgm_data;
+#ifndef WEBP_EXPERIMENTAL_FEATURES
+  if (!is_frame) return WEBP_MUX_INVALID_ARGUMENT;
+#endif
+  assert(wpi->header_ != NULL);  // Already checked by WebPMuxGetFrame().
+  // Get frame/fragment chunk.
+  frame_frgm_data = &wpi->header_->data_;
+  if (frame_frgm_data->size < kChunks[idx].size) return WEBP_MUX_BAD_DATA;
+  // Extract info.
+  frame->x_offset = 2 * GetLE24(frame_frgm_data->bytes + 0);
+  frame->y_offset = 2 * GetLE24(frame_frgm_data->bytes + 3);
+  frame->duration = is_frame ? GetLE24(frame_frgm_data->bytes + 12) : 1;
+  frame->dispose_method =
+      is_frame ? (WebPMuxAnimDispose)(frame_frgm_data->bytes[15] & 1)
+               : WEBP_MUX_DISPOSE_NONE;
+  frame->id = ChunkGetIdFromTag(wpi->header_->tag_);
+  return SynthesizeBitstream(wpi, &frame->bitstream);
 }
 
-WebPMuxError WebPMuxGetLoopCount(const WebPMux* mux, int* loop_count) {
-  WebPData image;
-  WebPMuxError err;
-
-  if (mux == NULL || loop_count == NULL) return WEBP_MUX_INVALID_ARGUMENT;
-
-  err = MuxGet(mux, IDX_LOOP, 1, &image);
-  if (err != WEBP_MUX_OK) return err;
-  if (image.size_ < kChunks[WEBP_CHUNK_LOOP].size) return WEBP_MUX_BAD_DATA;
-  *loop_count = GetLE16(image.bytes_);
-
-  return WEBP_MUX_OK;
-}
-
-static WebPMuxError MuxGetFrameTileInternal(
-    const WebPMux* const mux, uint32_t nth, WebPData* const bitstream,
-    int* const x_offset, int* const y_offset, int* const duration,
-    uint32_t tag) {
-  const WebPData* frame_tile_data;
+WebPMuxError WebPMuxGetFrame(
+    const WebPMux* mux, uint32_t nth, WebPMuxFrameInfo* frame) {
   WebPMuxError err;
   WebPMuxImage* wpi;
 
-  const int is_frame = (tag == kChunks[WEBP_CHUNK_FRAME].tag) ? 1 : 0;
-  const CHUNK_INDEX idx = is_frame ? IDX_FRAME : IDX_TILE;
-  const WebPChunkId id = kChunks[idx].id;
-
-  if (mux == NULL || bitstream == NULL ||
-      x_offset == NULL || y_offset == NULL || (is_frame && duration == NULL)) {
+  // Sanity checks.
+  if (mux == NULL || frame == NULL) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }
 
   // Get the nth WebPMuxImage.
-  err = MuxImageGetNth((const WebPMuxImage**)&mux->images_, nth, id, &wpi);
+  err = MuxImageGetNth((const WebPMuxImage**)&mux->images_, nth, &wpi);
   if (err != WEBP_MUX_OK) return err;
 
-  // Get frame chunk.
-  assert(wpi->header_ != NULL);  // As MuxImageGetNth() already checked header_.
-  frame_tile_data = &wpi->header_->data_;
-
-  if (frame_tile_data->size_ < kChunks[idx].size) return WEBP_MUX_BAD_DATA;
-  *x_offset = 2 * GetLE24(frame_tile_data->bytes_ + 0);
-  *y_offset = 2 * GetLE24(frame_tile_data->bytes_ + 3);
-  if (is_frame) *duration = 1 + GetLE24(frame_tile_data->bytes_ + 12);
-
-  return SynthesizeBitstream(wpi, bitstream);
+  // Get frame info.
+  if (wpi->header_ == NULL) {
+    return MuxGetImageInternal(wpi, frame);
+  } else {
+    return MuxGetFrameFragmentInternal(wpi, frame);
+  }
 }
 
-WebPMuxError WebPMuxGetFrame(const WebPMux* mux, uint32_t nth,
-                             WebPData* bitstream,
-                             int* x_offset, int* y_offset, int* duration) {
-  return MuxGetFrameTileInternal(mux, nth, bitstream, x_offset, y_offset,
-                                 duration, kChunks[IDX_FRAME].tag);
-}
+WebPMuxError WebPMuxGetAnimationParams(const WebPMux* mux,
+                                       WebPMuxAnimParams* params) {
+  WebPData anim;
+  WebPMuxError err;
 
-WebPMuxError WebPMuxGetTile(const WebPMux* mux, uint32_t nth,
-                            WebPData* bitstream,
-                            int* x_offset, int* y_offset) {
-  return MuxGetFrameTileInternal(mux, nth, bitstream, x_offset, y_offset, NULL,
-                                 kChunks[IDX_TILE].tag);
+  if (mux == NULL || params == NULL) return WEBP_MUX_INVALID_ARGUMENT;
+
+  err = MuxGet(mux, IDX_ANIM, 1, &anim);
+  if (err != WEBP_MUX_OK) return err;
+  if (anim.size < kChunks[WEBP_CHUNK_ANIM].size) return WEBP_MUX_BAD_DATA;
+  params->bgcolor = GetLE32(anim.bytes);
+  params->loop_count = GetLE16(anim.bytes + 4);
+
+  return WEBP_MUX_OK;
 }
 
 // Get chunk index from chunk id. Returns IDX_NIL if not found.
diff --git a/3rdparty/libwebp/utils/bit_reader.c b/3rdparty/libwebp/utils/bit_reader.c
index 1afb1db890..ab7a8273ea 100644
--- a/3rdparty/libwebp/utils/bit_reader.c
+++ b/3rdparty/libwebp/utils/bit_reader.c
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Boolean decoder
@@ -15,7 +17,11 @@
 extern "C" {
 #endif
 
-#define MK(X) (((bit_t)(X) << (BITS)) | (MASK))
+#ifndef USE_RIGHT_JUSTIFY
+#define MK(X) (((range_t)(X) << (BITS)) | (MASK))
+#else
+#define MK(X) ((range_t)(X))
+#endif
 
 //------------------------------------------------------------------------------
 // VP8BitReader
@@ -29,7 +35,7 @@ void VP8InitBitReader(VP8BitReader* const br,
   br->buf_     = start;
   br->buf_end_ = end;
   br->value_   = 0;
-  br->missing_ = 8;   // to load the very first 8bits
+  br->bits_    = -8;   // to load the very first 8bits
   br->eof_     = 0;
 }
 
@@ -46,7 +52,7 @@ const uint8_t kVP8Log2Range[128] = {
 };
 
 // range = (range << kVP8Log2Range[range]) + trailing 1's
-const bit_t kVP8NewRange[128] = {
+const range_t kVP8NewRange[128] = {
   MK(127), MK(127), MK(191), MK(127), MK(159), MK(191), MK(223), MK(127),
   MK(143), MK(159), MK(175), MK(191), MK(207), MK(223), MK(239), MK(127),
   MK(135), MK(143), MK(151), MK(159), MK(167), MK(175), MK(183), MK(191),
@@ -71,9 +77,19 @@ void VP8LoadFinalBytes(VP8BitReader* const br) {
   assert(br != NULL && br->buf_ != NULL);
   // Only read 8bits at a time
   if (br->buf_ < br->buf_end_) {
-    br->value_ |= (bit_t)(*br->buf_++) << ((BITS) - 8 + br->missing_);
-    br->missing_ -= 8;
-  } else {
+#ifndef USE_RIGHT_JUSTIFY
+    br->value_ |= (bit_t)(*br->buf_++) << ((BITS) - 8 - br->bits_);
+#else
+    br->value_ = (bit_t)(*br->buf_++) | (br->value_ << 8);
+#endif
+    br->bits_ += 8;
+  } else if (!br->eof_) {
+#ifdef USE_RIGHT_JUSTIFY
+    // These are not strictly needed, but it makes the behaviour
+    // consistent for both USE_RIGHT_JUSTIFY and !USE_RIGHT_JUSTIFY.
+    br->value_ <<= 8;
+    br->bits_ += 8;
+#endif
     br->eof_ = 1;
   }
 }
@@ -99,6 +115,10 @@ int32_t VP8GetSignedValue(VP8BitReader* const br, int bits) {
 
 #define MAX_NUM_BIT_READ 25
 
+#define LBITS 64      // Number of bits prefetched.
+#define WBITS 32      // Minimum number of bytes needed after VP8LFillBitWindow.
+#define LOG8_WBITS 4  // Number of bytes needed to store WBITS bits.
+
 static const uint32_t kBitMask[MAX_NUM_BIT_READ] = {
   0, 1, 3, 7, 15, 31, 63, 127, 255, 511, 1023, 2047, 4095, 8191, 16383, 32767,
   65535, 131071, 262143, 524287, 1048575, 2097151, 4194303, 8388607, 16777215
@@ -120,7 +140,7 @@ void VP8LInitBitReader(VP8LBitReader* const br,
   br->eos_ = 0;
   br->error_ = 0;
   for (i = 0; i < sizeof(br->val_) && i < br->len_; ++i) {
-    br->val_ |= ((uint64_t)br->buf_[br->pos_]) << (8 * i);
+    br->val_ |= ((vp8l_val_t)br->buf_[br->pos_]) << (8 * i);
     ++br->pos_;
   }
 }
@@ -135,91 +155,56 @@ void VP8LBitReaderSetBuffer(VP8LBitReader* const br,
   br->len_ = len;
 }
 
+// If not at EOS, reload up to LBITS byte-by-byte
 static void ShiftBytes(VP8LBitReader* const br) {
   while (br->bit_pos_ >= 8 && br->pos_ < br->len_) {
     br->val_ >>= 8;
-    br->val_ |= ((uint64_t)br->buf_[br->pos_]) << 56;
+    br->val_ |= ((vp8l_val_t)br->buf_[br->pos_]) << (LBITS - 8);
     ++br->pos_;
     br->bit_pos_ -= 8;
   }
 }
 
 void VP8LFillBitWindow(VP8LBitReader* const br) {
-  if (br->bit_pos_ >= 32) {
-#if defined(__x86_64__) || defined(_M_X64)
-    if (br->pos_ + 8 < br->len_) {
-      br->val_ >>= 32;
+  if (br->bit_pos_ >= WBITS) {
+#if (defined(__x86_64__) || defined(_M_X64))
+    if (br->pos_ + sizeof(br->val_) < br->len_) {
+      br->val_ >>= WBITS;
+      br->bit_pos_ -= WBITS;
       // The expression below needs a little-endian arch to work correctly.
       // This gives a large speedup for decoding speed.
-      br->val_ |= *(const uint64_t *)(br->buf_ + br->pos_) << 32;
-      br->pos_ += 4;
-      br->bit_pos_ -= 32;
-    } else {
-      // Slow path.
-      ShiftBytes(br);
+      br->val_ |= *(const vp8l_val_t*)(br->buf_ + br->pos_) << (LBITS - WBITS);
+      br->pos_ += LOG8_WBITS;
+      return;
     }
-#else
-    // Always the slow path.
-    ShiftBytes(br);
 #endif
-  }
-  if (br->pos_ == br->len_ && br->bit_pos_ == 64) {
-    br->eos_ = 1;
-  }
-}
-
-uint32_t VP8LReadOneBit(VP8LBitReader* const br) {
-  const uint32_t val = (br->val_ >> br->bit_pos_) & 1;
-  // Flag an error at end_of_stream.
-  if (!br->eos_) {
-    ++br->bit_pos_;
-    if (br->bit_pos_ >= 32) {
-      ShiftBytes(br);
-    }
-    // After this last bit is read, check if eos needs to be flagged.
-    if (br->pos_ == br->len_ && br->bit_pos_ == 64) {
+    ShiftBytes(br);       // Slow path.
+    if (br->pos_ == br->len_ && br->bit_pos_ == LBITS) {
       br->eos_ = 1;
     }
-  } else {
-    br->error_ = 1;
   }
-  return val;
 }
 
 uint32_t VP8LReadBits(VP8LBitReader* const br, int n_bits) {
-  uint32_t val = 0;
   assert(n_bits >= 0);
   // Flag an error if end_of_stream or n_bits is more than allowed limit.
   if (!br->eos_ && n_bits < MAX_NUM_BIT_READ) {
+    const uint32_t val =
+        (uint32_t)(br->val_ >> br->bit_pos_) & kBitMask[n_bits];
+    const int new_bits = br->bit_pos_ + n_bits;
+    br->bit_pos_ = new_bits;
     // If this read is going to cross the read buffer, set the eos flag.
     if (br->pos_ == br->len_) {
-      if ((br->bit_pos_ + n_bits) >= 64) {
+      if (new_bits >= LBITS) {
         br->eos_ = 1;
-        if ((br->bit_pos_ + n_bits) > 64) return val;
-      }
-    }
-    val = (br->val_ >> br->bit_pos_) & kBitMask[n_bits];
-    br->bit_pos_ += n_bits;
-    if (br->bit_pos_ >= 40) {
-      if (br->pos_ + 5 < br->len_) {
-        br->val_ >>= 40;
-        br->val_ |=
-            (((uint64_t)br->buf_[br->pos_ + 0]) << 24) |
-            (((uint64_t)br->buf_[br->pos_ + 1]) << 32) |
-            (((uint64_t)br->buf_[br->pos_ + 2]) << 40) |
-            (((uint64_t)br->buf_[br->pos_ + 3]) << 48) |
-            (((uint64_t)br->buf_[br->pos_ + 4]) << 56);
-        br->pos_ += 5;
-        br->bit_pos_ -= 40;
-      }
-      if (br->bit_pos_ >= 8) {
-        ShiftBytes(br);
       }
     }
+    ShiftBytes(br);
+    return val;
   } else {
     br->error_ = 1;
+    return 0;
   }
-  return val;
 }
 
 //------------------------------------------------------------------------------
diff --git a/3rdparty/libwebp/utils/bit_reader.h b/3rdparty/libwebp/utils/bit_reader.h
index cc7aeece1e..588222b3be 100644
--- a/3rdparty/libwebp/utils/bit_reader.h
+++ b/3rdparty/libwebp/utils/bit_reader.h
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Boolean decoder
@@ -24,11 +26,80 @@
 extern "C" {
 #endif
 
-#define BITS 32     // can be 32, 16 or 8
-#define MASK ((((bit_t)1) << (BITS)) - 1)
-#if (BITS == 32)
-typedef uint64_t bit_t;   // natural register type
-typedef uint32_t lbit_t;  // natural type for memory I/O
+// The Boolean decoder needs to maintain infinite precision on the value_ field.
+// However, since range_ is only 8bit, we only need an active window of 8 bits
+// for value_. Left bits (MSB) gets zeroed and shifted away when value_ falls
+// below 128, range_ is updated, and fresh bits read from the bitstream are
+// brought in as LSB.
+// To avoid reading the fresh bits one by one (slow), we cache a few of them
+// ahead (actually, we cache BITS of them ahead. See below). There's two
+// strategies regarding how to shift these looked-ahead fresh bits into the
+// 8bit window of value_: either we shift them in, while keeping the position of
+// the window fixed. Or we slide the window to the right while keeping the cache
+// bits at a fixed, right-justified, position.
+//
+//  Example, for BITS=16: here is the content of value_ for both strategies:
+//
+//          !USE_RIGHT_JUSTIFY            ||        USE_RIGHT_JUSTIFY
+//                                        ||
+//   <- 8b -><- 8b -><- BITS bits  ->     ||  <- 8b+3b -><- 8b -><- 13 bits ->
+//   [unused][value_][cached bits][0]     ||  [unused...][value_][cached bits]
+//  [........00vvvvvvBBBBBBBBBBBBB000]LSB || [...........00vvvvvvBBBBBBBBBBBBB]
+//                                        ||
+// After calling VP8Shift(), where we need to shift away two zeros:
+//  [........vvvvvvvvBBBBBBBBBBB00000]LSB || [.............vvvvvvvvBBBBBBBBBBB]
+//                                        ||
+// Just before we need to call VP8LoadNewBytes(), the situation is:
+//  [........vvvvvv000000000000000000]LSB || [..........................vvvvvv]
+//                                        ||
+// And just after calling VP8LoadNewBytes():
+//  [........vvvvvvvvBBBBBBBBBBBBBBBB]LSB || [........vvvvvvvvBBBBBBBBBBBBBBBB]
+//
+// -> we're back to height active 'value_' bits (marked 'v') and BITS cached
+// bits (marked 'B')
+//
+// The right-justify strategy tends to use less shifts and is often faster.
+
+//------------------------------------------------------------------------------
+// BITS can be any multiple of 8 from 8 to 56 (inclusive).
+// Pick values that fit natural register size.
+
+#if !defined(WEBP_REFERENCE_IMPLEMENTATION)
+
+#define USE_RIGHT_JUSTIFY
+
+#if defined(__i386__) || defined(_M_IX86)      // x86 32bit
+#define BITS 16
+#elif defined(__x86_64__) || defined(_M_X64)   // x86 64bit
+#define BITS 56
+#elif defined(__arm__) || defined(_M_ARM)      // ARM
+#define BITS 24
+#else                      // reasonable default
+#define BITS 24
+#endif
+
+#else     // reference choices
+
+#define USE_RIGHT_JUSTIFY
+#define BITS 8
+
+#endif
+
+//------------------------------------------------------------------------------
+// Derived types and constants
+
+// bit_t = natural register type
+// lbit_t = natural type for memory I/O
+
+#if (BITS > 32)
+typedef uint64_t bit_t;
+typedef uint64_t lbit_t;
+#elif (BITS == 32)
+typedef uint64_t bit_t;
+typedef uint32_t lbit_t;
+#elif (BITS == 24)
+typedef uint32_t bit_t;
+typedef uint32_t lbit_t;
 #elif (BITS == 16)
 typedef uint32_t bit_t;
 typedef uint16_t lbit_t;
@@ -37,8 +108,15 @@ typedef uint32_t bit_t;
 typedef uint8_t lbit_t;
 #endif
 
+#ifndef USE_RIGHT_JUSTIFY
+typedef bit_t range_t;     // type for storing range_
+#define MASK ((((bit_t)1) << (BITS)) - 1)
+#else
+typedef uint32_t range_t;  // range_ only uses 8bits here. No need for bit_t.
+#endif
+
 //------------------------------------------------------------------------------
-// Bitreader and code-tree reader
+// Bitreader
 
 typedef struct VP8BitReader VP8BitReader;
 struct VP8BitReader {
@@ -47,9 +125,9 @@ struct VP8BitReader {
   int eof_;                   // true if input is exhausted
 
   // boolean decoder
-  bit_t range_;            // current range minus 1. In [127, 254] interval.
-  bit_t value_;            // current value
-  int missing_;            // number of missing bits in value_ (8bit)
+  range_t range_;            // current range minus 1. In [127, 254] interval.
+  bit_t value_;              // current value
+  int bits_;                 // number of valid bits left
 };
 
 // Initialize the bit reader and the boolean decoder.
@@ -67,12 +145,12 @@ int32_t VP8GetSignedValue(VP8BitReader* const br, int num_bits);
 
 // Read a bit with proba 'prob'. Speed-critical function!
 extern const uint8_t kVP8Log2Range[128];
-extern const bit_t kVP8NewRange[128];
+extern const range_t kVP8NewRange[128];
 
 void VP8LoadFinalBytes(VP8BitReader* const br);    // special case for the tail
 
 static WEBP_INLINE void VP8LoadNewBytes(VP8BitReader* const br) {
-  assert(br && br->buf_);
+  assert(br != NULL && br->buf_ != NULL);
   // Read 'BITS' bits at a time if possible.
   if (br->buf_ + sizeof(lbit_t) <= br->buf_end_) {
     // convert memory type to register type (with some zero'ing!)
@@ -80,70 +158,125 @@ static WEBP_INLINE void VP8LoadNewBytes(VP8BitReader* const br) {
     lbit_t in_bits = *(lbit_t*)br->buf_;
     br->buf_ += (BITS) >> 3;
 #if !defined(__BIG_ENDIAN__)
-#if (BITS == 32)
+#if (BITS > 32)
+// gcc 4.3 has builtin functions for swap32/swap64
+#if defined(__GNUC__) && \
+           (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3))
+    bits = (bit_t)__builtin_bswap64(in_bits);
+#elif defined(_MSC_VER)
+    bits = (bit_t)_byteswap_uint64(in_bits);
+#elif defined(__x86_64__)
+    __asm__ volatile("bswapq %0" : "=r"(bits) : "0"(in_bits));
+#else  // generic code for swapping 64-bit values (suggested by bdb@)
+    bits = (bit_t)in_bits;
+    bits = ((bits & 0xffffffff00000000ull) >> 32) |
+           ((bits & 0x00000000ffffffffull) << 32);
+    bits = ((bits & 0xffff0000ffff0000ull) >> 16) |
+           ((bits & 0x0000ffff0000ffffull) << 16);
+    bits = ((bits & 0xff00ff00ff00ff00ull) >> 8) |
+           ((bits & 0x00ff00ff00ff00ffull) << 8);
+#endif
+    bits >>= 64 - BITS;
+#elif (BITS >= 24)
 #if defined(__i386__) || defined(__x86_64__)
     __asm__ volatile("bswap %k0" : "=r"(in_bits) : "0"(in_bits));
-    bits = (bit_t)in_bits;   // 32b -> 64b zero-extension
+    bits = (bit_t)in_bits;   // 24b/32b -> 32b/64b zero-extension
 #elif defined(_MSC_VER)
-    bits = _byteswap_ulong(in_bits);
+    bits = (bit_t)_byteswap_ulong(in_bits);
 #else
     bits = (bit_t)(in_bits >> 24) | ((in_bits >> 8) & 0xff00)
          | ((in_bits << 8) & 0xff0000)  | (in_bits << 24);
 #endif  // x86
+    bits >>= (32 - BITS);
 #elif (BITS == 16)
     // gcc will recognize a 'rorw $8, ...' here:
     bits = (bit_t)(in_bits >> 8) | ((in_bits & 0xff) << 8);
 #else   // BITS == 8
     bits = (bit_t)in_bits;
 #endif
-#else    // LITTLE_ENDIAN
+#else    // BIG_ENDIAN
     bits = (bit_t)in_bits;
+    if (BITS != 8 * sizeof(bit_t)) bits >>= (8 * sizeof(bit_t) - BITS);
 #endif
-    br->value_ |= bits << br->missing_;
-    br->missing_ -= (BITS);
+#ifndef USE_RIGHT_JUSTIFY
+    br->value_ |= bits << (-br->bits_);
+#else
+    br->value_ = bits | (br->value_ << (BITS));
+#endif
+    br->bits_ += (BITS);
   } else {
     VP8LoadFinalBytes(br);    // no need to be inlined
   }
 }
 
-static WEBP_INLINE int VP8BitUpdate(VP8BitReader* const br, bit_t split) {
-  const bit_t value_split = split | (MASK);
-  if (br->missing_ > 0) {  // Make sure we have a least BITS bits in 'value_'
+static WEBP_INLINE int VP8BitUpdate(VP8BitReader* const br, range_t split) {
+  if (br->bits_ < 0) {  // Make sure we have a least BITS bits in 'value_'
     VP8LoadNewBytes(br);
   }
-  if (br->value_ > value_split) {
-    br->range_ -= value_split + 1;
-    br->value_ -= value_split + 1;
+#ifndef USE_RIGHT_JUSTIFY
+  split |= (MASK);
+  if (br->value_ > split) {
+    br->range_ -= split + 1;
+    br->value_ -= split + 1;
     return 1;
   } else {
-    br->range_ = value_split;
+    br->range_ = split;
     return 0;
   }
+#else
+  {
+    const int pos = br->bits_;
+    const range_t value = (range_t)(br->value_ >> pos);
+    if (value > split) {
+      br->range_ -= split + 1;
+      br->value_ -= (bit_t)(split + 1) << pos;
+      return 1;
+    } else {
+      br->range_ = split;
+      return 0;
+    }
+  }
+#endif
 }
 
 static WEBP_INLINE void VP8Shift(VP8BitReader* const br) {
+#ifndef USE_RIGHT_JUSTIFY
   // range_ is in [0..127] interval here.
-  const int idx = br->range_ >> (BITS);
+  const bit_t idx = br->range_ >> (BITS);
   const int shift = kVP8Log2Range[idx];
   br->range_ = kVP8NewRange[idx];
   br->value_ <<= shift;
-  br->missing_ += shift;
+  br->bits_ -= shift;
+#else
+  const int shift = kVP8Log2Range[br->range_];
+  assert(br->range_ < (range_t)128);
+  br->range_ = kVP8NewRange[br->range_];
+  br->bits_ -= shift;
+#endif
 }
-
 static WEBP_INLINE int VP8GetBit(VP8BitReader* const br, int prob) {
+#ifndef USE_RIGHT_JUSTIFY
   // It's important to avoid generating a 64bit x 64bit multiply here.
   // We just need an 8b x 8b after all.
-  const bit_t split =
-      (bit_t)((uint32_t)(br->range_ >> (BITS)) * prob) << ((BITS) - 8);
+  const range_t split =
+      (range_t)((uint32_t)(br->range_ >> (BITS)) * prob) << ((BITS) - 8);
   const int bit = VP8BitUpdate(br, split);
-  if (br->range_ <= (((bit_t)0x7e << (BITS)) | (MASK))) {
+  if (br->range_ <= (((range_t)0x7e << (BITS)) | (MASK))) {
     VP8Shift(br);
   }
   return bit;
+#else
+  const range_t split = (br->range_ * prob) >> 8;
+  const int bit = VP8BitUpdate(br, split);
+  if (br->range_ <= (range_t)0x7e) {
+    VP8Shift(br);
+  }
+  return bit;
+#endif
 }
 
 static WEBP_INLINE int VP8GetSigned(VP8BitReader* const br, int v) {
-  const bit_t split = (br->range_ >> 1);
+  const range_t split = (br->range_ >> 1);
   const int bit = VP8BitUpdate(br, split);
   VP8Shift(br);
   return bit ? -v : v;
@@ -151,16 +284,18 @@ static WEBP_INLINE int VP8GetSigned(VP8BitReader* const br, int v) {
 
 
 // -----------------------------------------------------------------------------
-// Bitreader
+// Bitreader for lossless format
+
+typedef uint64_t vp8l_val_t;  // right now, this bit-reader can only use 64bit.
 
 typedef struct {
-  uint64_t       val_;
-  const uint8_t* buf_;
-  size_t         len_;
-  size_t         pos_;
-  int            bit_pos_;
-  int            eos_;
-  int            error_;
+  vp8l_val_t     val_;        // pre-fetched bits
+  const uint8_t* buf_;        // input byte buffer
+  size_t         len_;        // buffer length
+  size_t         pos_;        // byte position in buf_
+  int            bit_pos_;    // current bit-reading position in val_
+  int            eos_;        // bitstream is finished
+  int            error_;      // an error occurred (buffer overflow attempt...)
 } VP8LBitReader;
 
 void VP8LInitBitReader(VP8LBitReader* const br,
@@ -176,17 +311,14 @@ void VP8LBitReaderSetBuffer(VP8LBitReader* const br,
 // Flags eos if this read attempt is going to cross the read buffer.
 uint32_t VP8LReadBits(VP8LBitReader* const br, int n_bits);
 
-// Reads one bit from Read Buffer. Flags an error in case end_of_stream.
-// Flags eos after reading last bit from the buffer.
-uint32_t VP8LReadOneBit(VP8LBitReader* const br);
+// Return the prefetched bits, so they can be looked up.
+static WEBP_INLINE uint32_t VP8LPrefetchBits(VP8LBitReader* const br) {
+  return (uint32_t)(br->val_ >> br->bit_pos_);
+}
 
-// VP8LReadOneBitUnsafe is faster than VP8LReadOneBit, but it can be called only
-// 32 times after the last VP8LFillBitWindow. Any subsequent calls
-// (without VP8LFillBitWindow) will return invalid data.
-static WEBP_INLINE uint32_t VP8LReadOneBitUnsafe(VP8LBitReader* const br) {
-  const uint32_t val = (br->val_ >> br->bit_pos_) & 1;
-  ++br->bit_pos_;
-  return val;
+// Discard 'num_bits' bits from the cache.
+static WEBP_INLINE void VP8LDiscardBits(VP8LBitReader* const br, int num_bits) {
+  br->bit_pos_ += num_bits;
 }
 
 // Advances the Read buffer by 4 bytes to make room for reading next 32 bits.
diff --git a/3rdparty/libwebp/utils/bit_writer.c b/3rdparty/libwebp/utils/bit_writer.c
index 671159cacd..3827a13ad1 100644
--- a/3rdparty/libwebp/utils/bit_writer.c
+++ b/3rdparty/libwebp/utils/bit_writer.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Bit writing and boolean coder
diff --git a/3rdparty/libwebp/utils/bit_writer.h b/3rdparty/libwebp/utils/bit_writer.h
index f7ca08497f..cbb095c1be 100644
--- a/3rdparty/libwebp/utils/bit_writer.h
+++ b/3rdparty/libwebp/utils/bit_writer.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Bit writing and boolean coder
diff --git a/3rdparty/libwebp/utils/color_cache.c b/3rdparty/libwebp/utils/color_cache.c
index 560f81db10..749db6128f 100644
--- a/3rdparty/libwebp/utils/color_cache.c
+++ b/3rdparty/libwebp/utils/color_cache.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Color Cache for WebP Lossless
diff --git a/3rdparty/libwebp/utils/color_cache.h b/3rdparty/libwebp/utils/color_cache.h
index 13be629f36..e5a0bd6fd4 100644
--- a/3rdparty/libwebp/utils/color_cache.h
+++ b/3rdparty/libwebp/utils/color_cache.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Color Cache for WebP Lossless
diff --git a/3rdparty/libwebp/utils/filters.c b/3rdparty/libwebp/utils/filters.c
index 08f52a3d20..eb5bb34f41 100644
--- a/3rdparty/libwebp/utils/filters.c
+++ b/3rdparty/libwebp/utils/filters.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Spatial prediction using various filters
@@ -26,8 +28,7 @@ extern "C" {
   assert(out != NULL);                                      \
   assert(width > 0);                                        \
   assert(height > 0);                                       \
-  assert(bpp > 0);                                          \
-  assert(stride >= width * bpp);
+  assert(stride >= width);
 
 static WEBP_INLINE void PredictLine(const uint8_t* src, const uint8_t* pred,
                                     uint8_t* dst, int length, int inverse) {
@@ -43,7 +44,8 @@ static WEBP_INLINE void PredictLine(const uint8_t* src, const uint8_t* pred,
 // Horizontal filter.
 
 static WEBP_INLINE void DoHorizontalFilter(const uint8_t* in,
-    int width, int height, int bpp, int stride, int inverse, uint8_t* out) {
+                                           int width, int height, int stride,
+                                           int inverse, uint8_t* out) {
   int h;
   const uint8_t* preds = (inverse ? out : in);
   SANITY_CHECK(in, out);
@@ -52,11 +54,11 @@ static WEBP_INLINE void DoHorizontalFilter(const uint8_t* in,
   for (h = 0; h < height; ++h) {
     // Leftmost pixel is predicted from above (except for topmost scanline).
     if (h == 0) {
-      memcpy((void*)out, (const void*)in, bpp);
+      out[0] = in[0];
     } else {
-      PredictLine(in, preds - stride, out, bpp, inverse);
+      PredictLine(in, preds - stride, out, 1, inverse);
     }
-    PredictLine(in + bpp, preds, out + bpp, bpp * (width - 1), inverse);
+    PredictLine(in + 1, preds, out + 1, width - 1, inverse);
     preds += stride;
     in += stride;
     out += stride;
@@ -64,46 +66,46 @@ static WEBP_INLINE void DoHorizontalFilter(const uint8_t* in,
 }
 
 static void HorizontalFilter(const uint8_t* data, int width, int height,
-                             int bpp, int stride, uint8_t* filtered_data) {
-  DoHorizontalFilter(data, width, height, bpp, stride, 0, filtered_data);
+                             int stride, uint8_t* filtered_data) {
+  DoHorizontalFilter(data, width, height, stride, 0, filtered_data);
 }
 
-static void HorizontalUnfilter(const uint8_t* data, int width, int height,
-                               int bpp, int stride, uint8_t* recon_data) {
-  DoHorizontalFilter(data, width, height, bpp, stride, 1, recon_data);
+static void HorizontalUnfilter(int width, int height, int stride,
+                               uint8_t* data) {
+  DoHorizontalFilter(data, width, height, stride, 1, data);
 }
 
 //------------------------------------------------------------------------------
 // Vertical filter.
 
 static WEBP_INLINE void DoVerticalFilter(const uint8_t* in,
-    int width, int height, int bpp, int stride, int inverse, uint8_t* out) {
+                                         int width, int height, int stride,
+                                         int inverse, uint8_t* out) {
   int h;
   const uint8_t* preds = (inverse ? out : in);
   SANITY_CHECK(in, out);
 
   // Very first top-left pixel is copied.
-  memcpy((void*)out, (const void*)in, bpp);
+  out[0] = in[0];
   // Rest of top scan-line is left-predicted.
-  PredictLine(in + bpp, preds, out + bpp, bpp * (width - 1), inverse);
+  PredictLine(in + 1, preds, out + 1, width - 1, inverse);
 
   // Filter line-by-line.
   for (h = 1; h < height; ++h) {
     in += stride;
     out += stride;
-    PredictLine(in, preds, out, bpp * width, inverse);
+    PredictLine(in, preds, out, width, inverse);
     preds += stride;
   }
 }
 
 static void VerticalFilter(const uint8_t* data, int width, int height,
-                           int bpp, int stride, uint8_t* filtered_data) {
-  DoVerticalFilter(data, width, height, bpp, stride, 0, filtered_data);
+                           int stride, uint8_t* filtered_data) {
+  DoVerticalFilter(data, width, height, stride, 0, filtered_data);
 }
 
-static void VerticalUnfilter(const uint8_t* data, int width, int height,
-                             int bpp, int stride, uint8_t* recon_data) {
-  DoVerticalFilter(data, width, height, bpp, stride, 1, recon_data);
+static void VerticalUnfilter(int width, int height, int stride, uint8_t* data) {
+  DoVerticalFilter(data, width, height, stride, 1, data);
 }
 
 //------------------------------------------------------------------------------
@@ -111,19 +113,19 @@ static void VerticalUnfilter(const uint8_t* data, int width, int height,
 
 static WEBP_INLINE int GradientPredictor(uint8_t a, uint8_t b, uint8_t c) {
   const int g = a + b - c;
-  return (g < 0) ? 0 : (g > 255) ? 255 : g;
+  return ((g & ~0xff) == 0) ? g : (g < 0) ? 0 : 255;  // clip to 8bit
 }
 
 static WEBP_INLINE
 void DoGradientFilter(const uint8_t* in, int width, int height,
-                      int bpp, int stride, int inverse, uint8_t* out) {
+                      int stride, int inverse, uint8_t* out) {
   const uint8_t* preds = (inverse ? out : in);
   int h;
   SANITY_CHECK(in, out);
 
   // left prediction for top scan-line
-  memcpy((void*)out, (const void*)in, bpp);
-  PredictLine(in + bpp, preds, out + bpp, bpp * (width - 1), inverse);
+  out[0] = in[0];
+  PredictLine(in + 1, preds, out + 1, width - 1, inverse);
 
   // Filter line-by-line.
   for (h = 1; h < height; ++h) {
@@ -132,31 +134,29 @@ void DoGradientFilter(const uint8_t* in, int width, int height,
     in += stride;
     out += stride;
     // leftmost pixel: predict from above.
-    PredictLine(in, preds - stride, out, bpp, inverse);
-    for (w = bpp; w < width * bpp; ++w) {
-      const int pred = GradientPredictor(preds[w - bpp],
+    PredictLine(in, preds - stride, out, 1, inverse);
+    for (w = 1; w < width; ++w) {
+      const int pred = GradientPredictor(preds[w - 1],
                                          preds[w - stride],
-                                         preds[w - stride - bpp]);
+                                         preds[w - stride - 1]);
       out[w] = in[w] + (inverse ? pred : -pred);
     }
   }
 }
 
 static void GradientFilter(const uint8_t* data, int width, int height,
-                           int bpp, int stride, uint8_t* filtered_data) {
-  DoGradientFilter(data, width, height, bpp, stride, 0, filtered_data);
+                           int stride, uint8_t* filtered_data) {
+  DoGradientFilter(data, width, height, stride, 0, filtered_data);
 }
 
-static void GradientUnfilter(const uint8_t* data, int width, int height,
-                             int bpp, int stride, uint8_t* recon_data) {
-  DoGradientFilter(data, width, height, bpp, stride, 1, recon_data);
+static void GradientUnfilter(int width, int height, int stride, uint8_t* data) {
+  DoGradientFilter(data, width, height, stride, 1, data);
 }
 
 #undef SANITY_CHECK
 
 // -----------------------------------------------------------------------------
-// Quick estimate of a potentially interesting filter mode to try, in addition
-// to the default NONE.
+// Quick estimate of a potentially interesting filter mode to try.
 
 #define SMAX 16
 #define SDIFF(a, b) (abs((a) - (b)) >> 4)   // Scoring diff, in [0..SMAX)
@@ -166,6 +166,7 @@ WEBP_FILTER_TYPE EstimateBestFilter(const uint8_t* data,
   int i, j;
   int bins[WEBP_FILTER_LAST][SMAX];
   memset(bins, 0, sizeof(bins));
+
   // We only sample every other pixels. That's enough.
   for (j = 2; j < height - 1; j += 2) {
     const uint8_t* const p = data + j * stride;
@@ -215,7 +216,7 @@ const WebPFilterFunc WebPFilters[WEBP_FILTER_LAST] = {
   GradientFilter     // WEBP_FILTER_GRADIENT
 };
 
-const WebPFilterFunc WebPUnfilters[WEBP_FILTER_LAST] = {
+const WebPUnfilterFunc WebPUnfilters[WEBP_FILTER_LAST] = {
   NULL,                // WEBP_FILTER_NONE
   HorizontalUnfilter,  // WEBP_FILTER_HORIZONTAL
   VerticalUnfilter,    // WEBP_FILTER_VERTICAL
diff --git a/3rdparty/libwebp/utils/filters.h b/3rdparty/libwebp/utils/filters.h
index c5cdbd6deb..1f5fa164f3 100644
--- a/3rdparty/libwebp/utils/filters.h
+++ b/3rdparty/libwebp/utils/filters.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Spatial prediction using various filters
@@ -30,18 +32,19 @@ typedef enum {
 } WEBP_FILTER_TYPE;
 
 typedef void (*WebPFilterFunc)(const uint8_t* in, int width, int height,
-                               int bpp, int stride, uint8_t* out);
+                               int stride, uint8_t* out);
+typedef void (*WebPUnfilterFunc)(int width, int height, int stride,
+                                 uint8_t* data);
 
 // Filter the given data using the given predictor.
 // 'in' corresponds to a 2-dimensional pixel array of size (stride * height)
 // in raster order.
-// 'bpp' is number of bytes per pixel, and
 // 'stride' is number of bytes per scan line (with possible padding).
 // 'out' should be pre-allocated.
 extern const WebPFilterFunc WebPFilters[WEBP_FILTER_LAST];
 
-// Reconstruct the original data from the given filtered data.
-extern const WebPFilterFunc WebPUnfilters[WEBP_FILTER_LAST];
+// In-place reconstruct the original data from the given filtered data.
+extern const WebPUnfilterFunc WebPUnfilters[WEBP_FILTER_LAST];
 
 // Fast estimate of a potentially good filter.
 extern WEBP_FILTER_TYPE EstimateBestFilter(const uint8_t* data,
diff --git a/3rdparty/libwebp/utils/huffman.c b/3rdparty/libwebp/utils/huffman.c
index 41529cc9da..0ba9d05cfc 100644
--- a/3rdparty/libwebp/utils/huffman.c
+++ b/3rdparty/libwebp/utils/huffman.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Utilities for building and looking up Huffman trees.
diff --git a/3rdparty/libwebp/utils/huffman.h b/3rdparty/libwebp/utils/huffman.h
index 70220a67fb..83a517ee60 100644
--- a/3rdparty/libwebp/utils/huffman.h
+++ b/3rdparty/libwebp/utils/huffman.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Utilities for building and looking up Huffman trees.
diff --git a/3rdparty/libwebp/utils/huffman_encode.c b/3rdparty/libwebp/utils/huffman_encode.c
index 8ccd291d22..96086663d9 100644
--- a/3rdparty/libwebp/utils/huffman_encode.c
+++ b/3rdparty/libwebp/utils/huffman_encode.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
@@ -138,13 +140,8 @@ static int CompareHuffmanTrees(const void* ptr1, const void* ptr2) {
   } else if (t1->total_count_ < t2->total_count_) {
     return 1;
   } else {
-    if (t1->value_ < t2->value_) {
-      return -1;
-    }
-    if (t1->value_ > t2->value_) {
-      return 1;
-    }
-    return 0;
+    assert(t1->value_ != t2->value_);
+    return (t1->value_ < t2->value_) ? -1 : 1;
   }
 }
 
@@ -193,6 +190,10 @@ static int GenerateOptimalTree(const int* const histogram, int histogram_size,
     }
   }
 
+  if (tree_size_orig == 0) {   // pretty optimal already!
+    return 1;
+  }
+
   // 3 * tree_size is enough to cover all the nodes representing a
   // population and all the inserted nodes combining two existing nodes.
   // The tree pool needs 2 * (tree_size_orig - 1) entities, and the
@@ -234,7 +235,7 @@ static int GenerateOptimalTree(const int* const histogram, int histogram_size,
         tree_pool[tree_pool_size++] = tree[tree_size - 1];
         tree_pool[tree_pool_size++] = tree[tree_size - 2];
         count = tree_pool[tree_pool_size - 1].total_count_ +
-            tree_pool[tree_pool_size - 2].total_count_;
+                tree_pool[tree_pool_size - 2].total_count_;
         tree_size -= 2;
         {
           // Search for the insertion point.
diff --git a/3rdparty/libwebp/utils/huffman_encode.h b/3rdparty/libwebp/utils/huffman_encode.h
index cc3b38d330..0b81f470b3 100644
--- a/3rdparty/libwebp/utils/huffman_encode.h
+++ b/3rdparty/libwebp/utils/huffman_encode.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Author: Jyrki Alakuijala (jyrki@google.com)
diff --git a/3rdparty/libwebp/utils/quant_levels.c b/3rdparty/libwebp/utils/quant_levels.c
index f6884392aa..42c7245d36 100644
--- a/3rdparty/libwebp/utils/quant_levels.c
+++ b/3rdparty/libwebp/utils/quant_levels.c
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Quantize levels for specified number of quantization-levels ([2, 256]).
@@ -140,15 +142,6 @@ int QuantizeLevels(uint8_t* const data, int width, int height,
   return 1;
 }
 
-int DequantizeLevels(uint8_t* const data, int width, int height) {
-  if (data == NULL || width <= 0 || height <= 0) return 0;
-  // TODO(skal): implement gradient smoothing.
-  (void)data;
-  (void)width;
-  (void)height;
-  return 1;
-}
-
 #if defined(__cplusplus) || defined(c_plusplus)
 }    // extern "C"
 #endif
diff --git a/3rdparty/libwebp/utils/quant_levels.h b/3rdparty/libwebp/utils/quant_levels.h
index 89ccafe40d..2d90828df9 100644
--- a/3rdparty/libwebp/utils/quant_levels.h
+++ b/3rdparty/libwebp/utils/quant_levels.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Alpha plane quantization utility
@@ -27,11 +29,6 @@ extern "C" {
 int QuantizeLevels(uint8_t* const data, int width, int height, int num_levels,
                    uint64_t* const sse);
 
-// Apply post-processing to input 'data' of size 'width'x'height' assuming
-// that the source was quantized to a reduced number of levels.
-// Returns false in case of error (data is NULL, invalid parameters, ...).
-int DequantizeLevels(uint8_t* const data, int width, int height);
-
 #if defined(__cplusplus) || defined(c_plusplus)
 }    // extern "C"
 #endif
diff --git a/3rdparty/libwebp/utils/quant_levels_dec.c b/3rdparty/libwebp/utils/quant_levels_dec.c
new file mode 100644
index 0000000000..d93594b3ba
--- /dev/null
+++ b/3rdparty/libwebp/utils/quant_levels_dec.c
@@ -0,0 +1,30 @@
+// Copyright 2013 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// TODO(skal): implement gradient smoothing.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "./quant_levels_dec.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+int DequantizeLevels(uint8_t* const data, int width, int height) {
+  if (data == NULL || width <= 0 || height <= 0) return 0;
+  (void)data;
+  (void)width;
+  (void)height;
+  return 1;
+}
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
diff --git a/3rdparty/libwebp/utils/quant_levels_dec.h b/3rdparty/libwebp/utils/quant_levels_dec.h
new file mode 100644
index 0000000000..5891067650
--- /dev/null
+++ b/3rdparty/libwebp/utils/quant_levels_dec.h
@@ -0,0 +1,32 @@
+// Copyright 2013 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Alpha plane de-quantization utility
+//
+// Author:  Vikas Arora (vikasa@google.com)
+
+#ifndef WEBP_UTILS_QUANT_LEVELS_DEC_H_
+#define WEBP_UTILS_QUANT_LEVELS_DEC_H_
+
+#include "../webp/types.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+// Apply post-processing to input 'data' of size 'width'x'height' assuming
+// that the source was quantized to a reduced number of levels.
+// Returns false in case of error (data is NULL, invalid parameters, ...).
+int DequantizeLevels(uint8_t* const data, int width, int height);
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
+
+#endif  /* WEBP_UTILS_QUANT_LEVELS_DEC_H_ */
diff --git a/3rdparty/libwebp/utils/rescaler.c b/3rdparty/libwebp/utils/rescaler.c
index 9825dcbc5f..e5ddc296ab 100644
--- a/3rdparty/libwebp/utils/rescaler.c
+++ b/3rdparty/libwebp/utils/rescaler.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Rescaling functions
@@ -20,7 +22,7 @@ extern "C" {
 #endif
 
 #define RFIX 30
-#define MULT_FIX(x,y) (((int64_t)(x) * (y) + (1 << (RFIX - 1))) >> RFIX)
+#define MULT_FIX(x, y) (((int64_t)(x) * (y) + (1 << (RFIX - 1))) >> RFIX)
 
 void WebPRescalerInit(WebPRescaler* const wrk, int src_width, int src_height,
                       uint8_t* const dst, int dst_width, int dst_height,
diff --git a/3rdparty/libwebp/utils/rescaler.h b/3rdparty/libwebp/utils/rescaler.h
index ef93d465f0..aedce46254 100644
--- a/3rdparty/libwebp/utils/rescaler.h
+++ b/3rdparty/libwebp/utils/rescaler.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Rescaling functions
diff --git a/3rdparty/libwebp/utils/thread.c b/3rdparty/libwebp/utils/thread.c
index ce89cf9dc7..b1615d0fb8 100644
--- a/3rdparty/libwebp/utils/thread.c
+++ b/3rdparty/libwebp/utils/thread.c
@@ -1,18 +1,16 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Multi-threaded worker
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
-#ifdef HAVE_CONFIG_H
-#include "config.h"
-#endif
-
 #include <assert.h>
 #include <string.h>   // for memset()
 #include "./thread.h"
diff --git a/3rdparty/libwebp/utils/thread.h b/3rdparty/libwebp/utils/thread.h
index 3191890b76..13a61a4c84 100644
--- a/3rdparty/libwebp/utils/thread.h
+++ b/3rdparty/libwebp/utils/thread.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Multi-threaded worker
@@ -12,6 +14,10 @@
 #ifndef WEBP_UTILS_THREAD_H_
 #define WEBP_UTILS_THREAD_H_
 
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
 #endif
@@ -63,13 +69,13 @@ typedef struct {
 
 // Must be called first, before any other method.
 void WebPWorkerInit(WebPWorker* const worker);
-// Must be called initialize the object and spawn the thread. Re-entrant.
+// Must be called to initialize the object and spawn the thread. Re-entrant.
 // Will potentially launch the thread. Returns false in case of error.
 int WebPWorkerReset(WebPWorker* const worker);
-// Make sure the previous work is finished. Returns true if worker->had_error
-// was not set and not error condition was triggered by the working thread.
+// Makes sure the previous work is finished. Returns true if worker->had_error
+// was not set and no error condition was triggered by the working thread.
 int WebPWorkerSync(WebPWorker* const worker);
-// Trigger the thread to call hook() with data1 and data2 argument. These
+// Triggers the thread to call hook() with data1 and data2 argument. These
 // hook/data1/data2 can be changed at any time before calling this function,
 // but not be changed afterward until the next call to WebPWorkerSync().
 void WebPWorkerLaunch(WebPWorker* const worker);
diff --git a/3rdparty/libwebp/utils/utils.c b/3rdparty/libwebp/utils/utils.c
index 673b7e284c..7eb06105b8 100644
--- a/3rdparty/libwebp/utils/utils.c
+++ b/3rdparty/libwebp/utils/utils.c
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Misc. common utility functions
@@ -19,7 +21,8 @@ extern "C" {
 //------------------------------------------------------------------------------
 // Checked memory allocation
 
-static int CheckSizeArguments(uint64_t nmemb, size_t size) {
+// Returns 0 in case of overflow of nmemb * size.
+static int CheckSizeArgumentsOverflow(uint64_t nmemb, size_t size) {
   const uint64_t total_size = nmemb * size;
   if (nmemb == 0) return 1;
   if ((uint64_t)size > WEBP_MAX_ALLOCABLE_MEMORY / nmemb) return 0;
@@ -28,12 +31,14 @@ static int CheckSizeArguments(uint64_t nmemb, size_t size) {
 }
 
 void* WebPSafeMalloc(uint64_t nmemb, size_t size) {
-  if (!CheckSizeArguments(nmemb, size)) return NULL;
+  if (!CheckSizeArgumentsOverflow(nmemb, size)) return NULL;
+  assert(nmemb * size > 0);
   return malloc((size_t)(nmemb * size));
 }
 
 void* WebPSafeCalloc(uint64_t nmemb, size_t size) {
-  if (!CheckSizeArguments(nmemb, size)) return NULL;
+  if (!CheckSizeArgumentsOverflow(nmemb, size)) return NULL;
+  assert(nmemb * size > 0);
   return calloc((size_t)nmemb, size);
 }
 
diff --git a/3rdparty/libwebp/utils/utils.h b/3rdparty/libwebp/utils/utils.h
index a034762556..e10aeeb9d7 100644
--- a/3rdparty/libwebp/utils/utils.h
+++ b/3rdparty/libwebp/utils/utils.h
@@ -1,17 +1,22 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 // Misc. common utility functions
 //
-// Author: Skal (pascal.massimino@gmail.com)
+// Authors: Skal (pascal.massimino@gmail.com)
+//          Urvang (urvang@google.com)
 
 #ifndef WEBP_UTILS_UTILS_H_
 #define WEBP_UTILS_UTILS_H_
 
+#include <assert.h>
+
 #include "../webp/types.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
@@ -35,6 +40,40 @@ void* WebPSafeMalloc(uint64_t nmemb, size_t size);
 // in order to favor the "calloc(num_foo, sizeof(foo))" pattern.
 void* WebPSafeCalloc(uint64_t nmemb, size_t size);
 
+//------------------------------------------------------------------------------
+// Reading/writing data.
+
+// Read 16, 24 or 32 bits stored in little-endian order.
+static WEBP_INLINE int GetLE16(const uint8_t* const data) {
+  return (int)(data[0] << 0) | (data[1] << 8);
+}
+
+static WEBP_INLINE int GetLE24(const uint8_t* const data) {
+  return GetLE16(data) | (data[2] << 16);
+}
+
+static WEBP_INLINE uint32_t GetLE32(const uint8_t* const data) {
+  return (uint32_t)GetLE16(data) | (GetLE16(data + 2) << 16);
+}
+
+// Store 16, 24 or 32 bits in little-endian order.
+static WEBP_INLINE void PutLE16(uint8_t* const data, int val) {
+  assert(val < (1 << 16));
+  data[0] = (val >> 0);
+  data[1] = (val >> 8);
+}
+
+static WEBP_INLINE void PutLE24(uint8_t* const data, int val) {
+  assert(val < (1 << 24));
+  PutLE16(data, val & 0xffff);
+  data[2] = (val >> 16);
+}
+
+static WEBP_INLINE void PutLE32(uint8_t* const data, uint32_t val) {
+  PutLE16(data, (int)(val & 0xffff));
+  PutLE16(data + 2, (int)(val >> 16));
+}
+
 //------------------------------------------------------------------------------
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/webp/decode.h b/3rdparty/libwebp/webp/decode.h
index 43b6c58f4f..141f8618f5 100644
--- a/3rdparty/libwebp/webp/decode.h
+++ b/3rdparty/libwebp/webp/decode.h
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  Main decoding functions for WebP images.
@@ -18,7 +20,19 @@
 extern "C" {
 #endif
 
-#define WEBP_DECODER_ABI_VERSION 0x0200    // MAJOR(8b) + MINOR(8b)
+#define WEBP_DECODER_ABI_VERSION 0x0201    // MAJOR(8b) + MINOR(8b)
+
+// Note: forward declaring enumerations is not allowed in (strict) C and C++,
+// the types are left here for reference.
+// typedef enum VP8StatusCode VP8StatusCode;
+// typedef enum WEBP_CSP_MODE WEBP_CSP_MODE;
+typedef struct WebPRGBABuffer WebPRGBABuffer;
+typedef struct WebPYUVABuffer WebPYUVABuffer;
+typedef struct WebPDecBuffer WebPDecBuffer;
+typedef struct WebPIDecoder WebPIDecoder;
+typedef struct WebPBitstreamFeatures WebPBitstreamFeatures;
+typedef struct WebPDecoderOptions WebPDecoderOptions;
+typedef struct WebPDecoderConfig WebPDecoderConfig;
 
 // Return the decoder's version number, packed in hexadecimal using 8bits for
 // each of major/minor/revision. E.g: v2.5.7 is 0x020507.
@@ -118,20 +132,28 @@ WEBP_EXTERN(uint8_t*) WebPDecodeYUVInto(
 // Note: the naming describes the byte-ordering of packed samples in memory.
 // For instance, MODE_BGRA relates to samples ordered as B,G,R,A,B,G,R,A,...
 // Non-capital names (e.g.:MODE_Argb) relates to pre-multiplied RGB channels.
-// RGB-565 and RGBA-4444 are also endian-agnostic and byte-oriented.
-typedef enum { MODE_RGB = 0, MODE_RGBA = 1,
-               MODE_BGR = 2, MODE_BGRA = 3,
-               MODE_ARGB = 4, MODE_RGBA_4444 = 5,
-               MODE_RGB_565 = 6,
-               // RGB-premultiplied transparent modes (alpha value is preserved)
-               MODE_rgbA = 7,
-               MODE_bgrA = 8,
-               MODE_Argb = 9,
-               MODE_rgbA_4444 = 10,
-               // YUV modes must come after RGB ones.
-               MODE_YUV = 11, MODE_YUVA = 12,  // yuv 4:2:0
-               MODE_LAST = 13
-             } WEBP_CSP_MODE;
+// RGBA-4444 and RGB-565 colorspaces are represented by following byte-order:
+// RGBA-4444: [r3 r2 r1 r0 g3 g2 g1 g0], [b3 b2 b1 b0 a3 a2 a1 a0], ...
+// RGB-565: [r4 r3 r2 r1 r0 g5 g4 g3], [g2 g1 g0 b4 b3 b2 b1 b0], ...
+// In the case WEBP_SWAP_16BITS_CSP is defined, the bytes are swapped for
+// these two modes:
+// RGBA-4444: [b3 b2 b1 b0 a3 a2 a1 a0], [r3 r2 r1 r0 g3 g2 g1 g0], ...
+// RGB-565: [g2 g1 g0 b4 b3 b2 b1 b0], [r4 r3 r2 r1 r0 g5 g4 g3], ...
+
+typedef enum WEBP_CSP_MODE {
+  MODE_RGB = 0, MODE_RGBA = 1,
+  MODE_BGR = 2, MODE_BGRA = 3,
+  MODE_ARGB = 4, MODE_RGBA_4444 = 5,
+  MODE_RGB_565 = 6,
+  // RGB-premultiplied transparent modes (alpha value is preserved)
+  MODE_rgbA = 7,
+  MODE_bgrA = 8,
+  MODE_Argb = 9,
+  MODE_rgbA_4444 = 10,
+  // YUV modes must come after RGB ones.
+  MODE_YUV = 11, MODE_YUVA = 12,  // yuv 4:2:0
+  MODE_LAST = 13
+} WEBP_CSP_MODE;
 
 // Some useful macros:
 static WEBP_INLINE int WebPIsPremultipliedMode(WEBP_CSP_MODE mode) {
@@ -152,13 +174,13 @@ static WEBP_INLINE int WebPIsRGBMode(WEBP_CSP_MODE mode) {
 //------------------------------------------------------------------------------
 // WebPDecBuffer: Generic structure for describing the output sample buffer.
 
-typedef struct {    // view as RGBA
+struct WebPRGBABuffer {    // view as RGBA
   uint8_t* rgba;    // pointer to RGBA samples
   int stride;       // stride in bytes from one scanline to the next.
   size_t size;      // total size of the *rgba buffer.
-} WebPRGBABuffer;
+};
 
-typedef struct {              // view as YUVA
+struct WebPYUVABuffer {              // view as YUVA
   uint8_t* y, *u, *v, *a;     // pointer to luma, chroma U/V, alpha samples
   int y_stride;               // luma stride
   int u_stride, v_stride;     // chroma strides
@@ -166,10 +188,10 @@ typedef struct {              // view as YUVA
   size_t y_size;              // luma plane size
   size_t u_size, v_size;      // chroma planes size
   size_t a_size;              // alpha-plane size
-} WebPYUVABuffer;
+};
 
 // Output buffer
-typedef struct {
+struct WebPDecBuffer {
   WEBP_CSP_MODE colorspace;  // Colorspace.
   int width, height;         // Dimensions.
   int is_external_memory;    // If true, 'internal_memory' pointer is not used.
@@ -182,7 +204,7 @@ typedef struct {
   uint8_t* private_memory;   // Internally allocated memory (only when
                              // is_external_memory is false). Should not be used
                              // externally, but accessed via the buffer union.
-} WebPDecBuffer;
+};
 
 // Internal, version-checked, entry point
 WEBP_EXTERN(int) WebPInitDecBufferInternal(WebPDecBuffer*, int);
@@ -200,7 +222,7 @@ WEBP_EXTERN(void) WebPFreeDecBuffer(WebPDecBuffer* buffer);
 //------------------------------------------------------------------------------
 // Enumeration of the status codes
 
-typedef enum {
+typedef enum VP8StatusCode {
   VP8_STATUS_OK = 0,
   VP8_STATUS_OUT_OF_MEMORY,
   VP8_STATUS_INVALID_PARAM,
@@ -237,13 +259,17 @@ typedef enum {
 //   }
 //   WebPIDelete(idec);
 
-typedef struct WebPIDecoder WebPIDecoder;
-
 // Creates a new incremental decoder with the supplied buffer parameter.
 // This output_buffer can be passed NULL, in which case a default output buffer
 // is used (with MODE_RGB). Otherwise, an internal reference to 'output_buffer'
 // is kept, which means that the lifespan of 'output_buffer' must be larger than
 // that of the returned WebPIDecoder object.
+// The supplied 'output_buffer' content MUST NOT be changed between calls to
+// WebPIAppend() or WebPIUpdate() unless 'output_buffer.is_external_memory' is
+// set to 1. In such a case, it is allowed to modify the pointers, size and
+// stride of output_buffer.u.RGBA or output_buffer.u.YUVA, provided they remain
+// within valid bounds.
+// All other fields of WebPDecBuffer MUST remain constant between calls.
 // Returns NULL if the allocation failed.
 WEBP_EXTERN(WebPIDecoder*) WebPINewDecoder(WebPDecBuffer* output_buffer);
 
@@ -251,19 +277,27 @@ WEBP_EXTERN(WebPIDecoder*) WebPINewDecoder(WebPDecBuffer* output_buffer);
 // will output the RGB/A samples specified by 'csp' into a preallocated
 // buffer 'output_buffer'. The size of this buffer is at least
 // 'output_buffer_size' and the stride (distance in bytes between two scanlines)
-// is specified by 'output_stride'. Returns NULL if the allocation failed.
+// is specified by 'output_stride'.
+// Additionally, output_buffer can be passed NULL in which case the output
+// buffer will be allocated automatically when the decoding starts. The
+// colorspace 'csp' is taken into account for allocating this buffer. All other
+// parameters are ignored.
+// Returns NULL if the allocation failed, or if some parameters are invalid.
 WEBP_EXTERN(WebPIDecoder*) WebPINewRGB(
     WEBP_CSP_MODE csp,
     uint8_t* output_buffer, size_t output_buffer_size, int output_stride);
 
 // This function allocates and initializes an incremental-decoder object, which
-// will output the raw luma/chroma samples into a preallocated planes. The luma
-// plane is specified by its pointer 'luma', its size 'luma_size' and its stride
-// 'luma_stride'. Similarly, the chroma-u plane is specified by the 'u',
-// 'u_size' and 'u_stride' parameters, and the chroma-v plane by 'v'
-// and 'v_size'. And same for the alpha-plane. The 'a' pointer can be pass
-// NULL in case one is not interested in the transparency plane.
-// Returns NULL if the allocation failed.
+// will output the raw luma/chroma samples into a preallocated planes if
+// supplied. The luma plane is specified by its pointer 'luma', its size
+// 'luma_size' and its stride 'luma_stride'. Similarly, the chroma-u plane
+// is specified by the 'u', 'u_size' and 'u_stride' parameters, and the chroma-v
+// plane by 'v' and 'v_size'. And same for the alpha-plane. The 'a' pointer
+// can be pass NULL in case one is not interested in the transparency plane.
+// Conversely, 'luma' can be passed NULL if no preallocated planes are supplied.
+// In this case, the output buffer will be automatically allocated (using
+// MODE_YUVA) when decoding starts. All parameters are then ignored.
+// Returns NULL if the allocation failed or if a parameter is invalid.
 WEBP_EXTERN(WebPIDecoder*) WebPINewYUVA(
     uint8_t* luma, size_t luma_size, int luma_stride,
     uint8_t* u, size_t u_size, int u_stride,
@@ -344,7 +378,7 @@ WEBP_EXTERN(const WebPDecBuffer*) WebPIDecodedArea(
      CHECK(WebPGetFeatures(data, data_size, &config.input) == VP8_STATUS_OK);
 
      // C) Adjust 'config', if needed
-     config.no_fancy = 1;
+     config.no_fancy_upsampling = 1;
      config.output.colorspace = MODE_BGRA;
      // etc.
 
@@ -365,10 +399,11 @@ WEBP_EXTERN(const WebPDecBuffer*) WebPIDecodedArea(
 */
 
 // Features gathered from the bitstream
-typedef struct {
-  int width;        // Width in pixels, as read from the bitstream.
-  int height;       // Height in pixels, as read from the bitstream.
-  int has_alpha;    // True if the bitstream contains an alpha channel.
+struct WebPBitstreamFeatures {
+  int width;          // Width in pixels, as read from the bitstream.
+  int height;         // Height in pixels, as read from the bitstream.
+  int has_alpha;      // True if the bitstream contains an alpha channel.
+  int has_animation;  // True if the bitstream is an animation.
 
   // Unused for now:
   int bitstream_version;        // should be 0 for now. TODO(later)
@@ -376,8 +411,8 @@ typedef struct {
                                 // recommended.
   int rotate;                   // TODO(later)
   int uv_sampling;              // should be 0 for now. TODO(later)
-  uint32_t pad[3];              // padding for later use
-} WebPBitstreamFeatures;
+  uint32_t pad[2];              // padding for later use
+};
 
 // Internal, version-checked, entry point
 WEBP_EXTERN(VP8StatusCode) WebPGetFeaturesInternal(
@@ -385,8 +420,9 @@ WEBP_EXTERN(VP8StatusCode) WebPGetFeaturesInternal(
 
 // Retrieve features from the bitstream. The *features structure is filled
 // with information gathered from the bitstream.
-// Returns false in case of error or version mismatch.
-// In case of error, features->bitstream_status will reflect the error code.
+// Returns VP8_STATUS_OK when the features are successfully retrieved. Returns
+// VP8_STATUS_NOT_ENOUGH_DATA when more data is needed to retrieve the
+// features from headers. Returns error in other cases.
 static WEBP_INLINE VP8StatusCode WebPGetFeatures(
     const uint8_t* data, size_t data_size,
     WebPBitstreamFeatures* features) {
@@ -395,7 +431,7 @@ static WEBP_INLINE VP8StatusCode WebPGetFeatures(
 }
 
 // Decoding options
-typedef struct {
+struct WebPDecoderOptions {
   int bypass_filtering;               // if true, skip the in-loop filtering
   int no_fancy_upsampling;            // if true, use faster pointwise upsampler
   int use_cropping;                   // if true, cropping is applied _first_
@@ -410,14 +446,14 @@ typedef struct {
   int force_rotation;                 // forced rotation (to be applied _last_)
   int no_enhancement;                 // if true, discard enhancement layer
   uint32_t pad[6];                    // padding for later use
-} WebPDecoderOptions;
+};
 
 // Main object storing the configuration for advanced decoding.
-typedef struct {
+struct WebPDecoderConfig {
   WebPBitstreamFeatures input;  // Immutable bitstream features (optional)
   WebPDecBuffer output;         // Output buffer (can point to external mem)
   WebPDecoderOptions options;   // Decoding options
-} WebPDecoderConfig;
+};
 
 // Internal, version-checked, entry point
 WEBP_EXTERN(int) WebPInitDecoderConfigInternal(WebPDecoderConfig*, int);
diff --git a/3rdparty/libwebp/webp/demux.h b/3rdparty/libwebp/webp/demux.h
new file mode 100644
index 0000000000..c7cd5d66db
--- /dev/null
+++ b/3rdparty/libwebp/webp/demux.h
@@ -0,0 +1,214 @@
+// Copyright 2012 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Demux API.
+// Enables extraction of image and extended format data from WebP files.
+
+// Code Example: Demuxing WebP data to extract all the frames, ICC profile
+// and EXIF/XMP metadata.
+//
+//   WebPDemuxer* demux = WebPDemux(&webp_data);
+//
+//   uint32_t width = WebPDemuxGetI(demux, WEBP_FF_CANVAS_WIDTH);
+//   uint32_t height = WebPDemuxGetI(demux, WEBP_FF_CANVAS_HEIGHT);
+//   // ... (Get information about the features present in the WebP file).
+//   uint32_t flags = WebPDemuxGetI(demux, WEBP_FF_FORMAT_FLAGS);
+//
+//   // ... (Iterate over all frames).
+//   WebPIterator iter;
+//   if (WebPDemuxGetFrame(demux, 1, &iter)) {
+//     do {
+//       // ... (Consume 'iter'; e.g. Decode 'iter.fragment' with WebPDecode(),
+//       // ... and get other frame properties like width, height, offsets etc.
+//       // ... see 'struct WebPIterator' below for more info).
+//     } while (WebPDemuxNextFrame(&iter));
+//     WebPDemuxReleaseIterator(&iter);
+//   }
+//
+//   // ... (Extract metadata).
+//   WebPChunkIterator chunk_iter;
+//   if (flags & ICCP_FLAG) WebPDemuxGetChunk(demux, "ICCP", 1, &chunk_iter);
+//   // ... (Consume the ICC profile in 'chunk_iter.chunk').
+//   WebPDemuxReleaseChunkIterator(&chunk_iter);
+//   if (flags & EXIF_FLAG) WebPDemuxGetChunk(demux, "EXIF", 1, &chunk_iter);
+//   // ... (Consume the EXIF metadata in 'chunk_iter.chunk').
+//   WebPDemuxReleaseChunkIterator(&chunk_iter);
+//   if (flags & XMP_FLAG) WebPDemuxGetChunk(demux, "XMP ", 1, &chunk_iter);
+//   // ... (Consume the XMP metadata in 'chunk_iter.chunk').
+//   WebPDemuxReleaseChunkIterator(&chunk_iter);
+//   WebPDemuxDelete(demux);
+
+#ifndef WEBP_WEBP_DEMUX_H_
+#define WEBP_WEBP_DEMUX_H_
+
+#include "./mux_types.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+#define WEBP_DEMUX_ABI_VERSION 0x0100    // MAJOR(8b) + MINOR(8b)
+
+// Note: forward declaring enumerations is not allowed in (strict) C and C++,
+// the types are left here for reference.
+// typedef enum WebPDemuxState WebPDemuxState;
+// typedef enum WebPFormatFeature WebPFormatFeature;
+typedef struct WebPDemuxer WebPDemuxer;
+typedef struct WebPIterator WebPIterator;
+typedef struct WebPChunkIterator WebPChunkIterator;
+
+//------------------------------------------------------------------------------
+
+// Returns the version number of the demux library, packed in hexadecimal using
+// 8bits for each of major/minor/revision. E.g: v2.5.7 is 0x020507.
+WEBP_EXTERN(int) WebPGetDemuxVersion(void);
+
+//------------------------------------------------------------------------------
+// Life of a Demux object
+
+typedef enum WebPDemuxState {
+  WEBP_DEMUX_PARSING_HEADER,  // Not enough data to parse full header.
+  WEBP_DEMUX_PARSED_HEADER,   // Header parsing complete, data may be available.
+  WEBP_DEMUX_DONE             // Entire file has been parsed.
+} WebPDemuxState;
+
+// Internal, version-checked, entry point
+WEBP_EXTERN(WebPDemuxer*) WebPDemuxInternal(
+    const WebPData*, int, WebPDemuxState*, int);
+
+// Parses the full WebP file given by 'data'.
+// Returns a WebPDemuxer object on successful parse, NULL otherwise.
+static WEBP_INLINE WebPDemuxer* WebPDemux(const WebPData* data) {
+  return WebPDemuxInternal(data, 0, NULL, WEBP_DEMUX_ABI_VERSION);
+}
+
+// Parses the possibly incomplete WebP file given by 'data'.
+// If 'state' is non-NULL it will be set to indicate the status of the demuxer.
+// Returns a WebPDemuxer object on successful parse, NULL otherwise.
+static WEBP_INLINE WebPDemuxer* WebPDemuxPartial(
+    const WebPData* data, WebPDemuxState* state) {
+  return WebPDemuxInternal(data, 1, state, WEBP_DEMUX_ABI_VERSION);
+}
+
+// Frees memory associated with 'dmux'.
+WEBP_EXTERN(void) WebPDemuxDelete(WebPDemuxer* dmux);
+
+//------------------------------------------------------------------------------
+// Data/information extraction.
+
+typedef enum WebPFormatFeature {
+  WEBP_FF_FORMAT_FLAGS,  // Extended format flags present in the 'VP8X' chunk.
+  WEBP_FF_CANVAS_WIDTH,
+  WEBP_FF_CANVAS_HEIGHT,
+  WEBP_FF_LOOP_COUNT,
+  WEBP_FF_BACKGROUND_COLOR,
+  WEBP_FF_FRAME_COUNT    // Number of frames present in the demux object.
+                         // In case of a partial demux, this is the number of
+                         // frames seen so far, with the last frame possibly
+                         // being partial.
+} WebPFormatFeature;
+
+// Get the 'feature' value from the 'dmux'.
+// NOTE: values are only valid if WebPDemux() was used or WebPDemuxPartial()
+// returned a state > WEBP_DEMUX_PARSING_HEADER.
+WEBP_EXTERN(uint32_t) WebPDemuxGetI(
+    const WebPDemuxer* dmux, WebPFormatFeature feature);
+
+//------------------------------------------------------------------------------
+// Frame iteration.
+
+struct WebPIterator {
+  int frame_num;
+  int num_frames;          // equivalent to WEBP_FF_FRAME_COUNT.
+  int fragment_num;
+  int num_fragments;
+  int x_offset, y_offset;  // offset relative to the canvas.
+  int width, height;       // dimensions of this frame or fragment.
+  int duration;            // display duration in milliseconds.
+  WebPMuxAnimDispose dispose_method;  // dispose method for the frame.
+  int complete;   // true if 'fragment' contains a full frame. partial images
+                  // may still be decoded with the WebP incremental decoder.
+  WebPData fragment;  // The frame or fragment given by 'frame_num' and
+                      // 'fragment_num'.
+
+  uint32_t pad[4];         // padding for later use.
+  void* private_;          // for internal use only.
+};
+
+// Retrieves frame 'frame_number' from 'dmux'.
+// 'iter->fragment' points to the first fragment on return from this function.
+// Individual fragments may be extracted using WebPDemuxSetFragment().
+// Setting 'frame_number' equal to 0 will return the last frame of the image.
+// Returns false if 'dmux' is NULL or frame 'frame_number' is not present.
+// Call WebPDemuxReleaseIterator() when use of the iterator is complete.
+// NOTE: 'dmux' must persist for the lifetime of 'iter'.
+WEBP_EXTERN(int) WebPDemuxGetFrame(
+    const WebPDemuxer* dmux, int frame_number, WebPIterator* iter);
+
+// Sets 'iter->fragment' to point to the next ('iter->frame_num' + 1) or
+// previous ('iter->frame_num' - 1) frame. These functions do not loop.
+// Returns true on success, false otherwise.
+WEBP_EXTERN(int) WebPDemuxNextFrame(WebPIterator* iter);
+WEBP_EXTERN(int) WebPDemuxPrevFrame(WebPIterator* iter);
+
+// Sets 'iter->fragment' to reflect fragment number 'fragment_num'.
+// Returns true if fragment 'fragment_num' is present, false otherwise.
+WEBP_EXTERN(int) WebPDemuxSelectFragment(WebPIterator* iter, int fragment_num);
+
+// Releases any memory associated with 'iter'.
+// Must be called before any subsequent calls to WebPDemuxGetChunk() on the same
+// iter. Also, must be called before destroying the associated WebPDemuxer with
+// WebPDemuxDelete().
+WEBP_EXTERN(void) WebPDemuxReleaseIterator(WebPIterator* iter);
+
+//------------------------------------------------------------------------------
+// Chunk iteration.
+
+struct WebPChunkIterator {
+  // The current and total number of chunks with the fourcc given to
+  // WebPDemuxGetChunk().
+  int chunk_num;
+  int num_chunks;
+  WebPData chunk;    // The payload of the chunk.
+
+  uint32_t pad[6];   // padding for later use
+  void* private_;
+};
+
+// Retrieves the 'chunk_number' instance of the chunk with id 'fourcc' from
+// 'dmux'.
+// 'fourcc' is a character array containing the fourcc of the chunk to return,
+// e.g., "ICCP", "XMP ", "EXIF", etc.
+// Setting 'chunk_number' equal to 0 will return the last chunk in a set.
+// Returns true if the chunk is found, false otherwise. Image related chunk
+// payloads are accessed through WebPDemuxGetFrame() and related functions.
+// Call WebPDemuxReleaseChunkIterator() when use of the iterator is complete.
+// NOTE: 'dmux' must persist for the lifetime of the iterator.
+WEBP_EXTERN(int) WebPDemuxGetChunk(const WebPDemuxer* dmux,
+                                   const char fourcc[4], int chunk_number,
+                                   WebPChunkIterator* iter);
+
+// Sets 'iter->chunk' to point to the next ('iter->chunk_num' + 1) or previous
+// ('iter->chunk_num' - 1) chunk. These functions do not loop.
+// Returns true on success, false otherwise.
+WEBP_EXTERN(int) WebPDemuxNextChunk(WebPChunkIterator* iter);
+WEBP_EXTERN(int) WebPDemuxPrevChunk(WebPChunkIterator* iter);
+
+// Releases any memory associated with 'iter'.
+// Must be called before destroying the associated WebPDemuxer with
+// WebPDemuxDelete().
+WEBP_EXTERN(void) WebPDemuxReleaseChunkIterator(WebPChunkIterator* iter);
+
+//------------------------------------------------------------------------------
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
+
+#endif  /* WEBP_WEBP_DEMUX_H_ */
diff --git a/3rdparty/libwebp/webp/encode.h b/3rdparty/libwebp/webp/encode.h
index 2e37cfabe7..726992f570 100644
--- a/3rdparty/libwebp/webp/encode.h
+++ b/3rdparty/libwebp/webp/encode.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //   WebP encoder: main interface
@@ -18,7 +20,18 @@
 extern "C" {
 #endif
 
-#define WEBP_ENCODER_ABI_VERSION 0x0200    // MAJOR(8b) + MINOR(8b)
+#define WEBP_ENCODER_ABI_VERSION 0x0201    // MAJOR(8b) + MINOR(8b)
+
+// Note: forward declaring enumerations is not allowed in (strict) C and C++,
+// the types are left here for reference.
+// typedef enum WebPImageHint WebPImageHint;
+// typedef enum WebPEncCSP WebPEncCSP;
+// typedef enum WebPPreset WebPPreset;
+// typedef enum WebPEncodingError WebPEncodingError;
+typedef struct WebPConfig WebPConfig;
+typedef struct WebPPicture WebPPicture;   // main structure for I/O
+typedef struct WebPAuxStats WebPAuxStats;
+typedef struct WebPMemoryWriter WebPMemoryWriter;
 
 // Return the encoder's version number, packed in hexadecimal using 8bits for
 // each of major/minor/revision. E.g: v2.5.7 is 0x020507.
@@ -66,7 +79,7 @@ WEBP_EXTERN(size_t) WebPEncodeLosslessBGRA(const uint8_t* bgra,
 // Coding parameters
 
 // Image characteristics hint for the underlying encoder.
-typedef enum {
+typedef enum WebPImageHint {
   WEBP_HINT_DEFAULT = 0,  // default preset.
   WEBP_HINT_PICTURE,      // digital picture, like portrait, inner shot
   WEBP_HINT_PHOTO,        // outdoor photograph, with natural lighting
@@ -74,7 +87,8 @@ typedef enum {
   WEBP_HINT_LAST
 } WebPImageHint;
 
-typedef struct {
+// Compression parameters.
+struct WebPConfig {
   int lossless;           // Lossless encoding (0=lossy(default), 1=lossless).
   float quality;          // between 0 (smallest file) and 100 (biggest)
   int method;             // quality/speed trade-off (0=fast, 6=slower-better)
@@ -109,13 +123,19 @@ typedef struct {
   int partition_limit;    // quality degradation allowed to fit the 512k limit
                           // on prediction modes coding (0: no degradation,
                           // 100: maximum possible degradation).
+  int emulate_jpeg_size;  // If true, compression parameters will be remapped
+                          // to better match the expected output size from
+                          // JPEG compression. Generally, the output size will
+                          // be similar but the degradation will be lower.
+  int thread_level;       // If non-zero, try and use multi-threaded encoding.
+  int low_memory;         // If set, reduce memory usage (but increase CPU use).
 
-  uint32_t pad[8];        // padding for later use
-} WebPConfig;
+  uint32_t pad[5];        // padding for later use
+};
 
 // Enumerate some predefined settings for WebPConfig, depending on the type
 // of source picture. These presets are used when calling WebPConfigPreset().
-typedef enum {
+typedef enum WebPPreset {
   WEBP_PRESET_DEFAULT = 0,  // default preset.
   WEBP_PRESET_PICTURE,      // digital picture, like portrait, inner shot
   WEBP_PRESET_PHOTO,        // outdoor photograph, with natural lighting
@@ -152,11 +172,9 @@ WEBP_EXTERN(int) WebPValidateConfig(const WebPConfig* config);
 
 //------------------------------------------------------------------------------
 // Input / Output
-
-typedef struct WebPPicture WebPPicture;   // main structure for I/O
-
 // Structure for storing auxiliary statistics (mostly for lossy encoding).
-typedef struct {
+
+struct WebPAuxStats {
   int coded_size;         // final size
 
   float PSNR[5];          // peak-signal-to-noise ratio for Y/U/V/All/Alpha
@@ -182,7 +200,7 @@ typedef struct {
   int lossless_size;           // final lossless size
 
   uint32_t pad[4];        // padding for later use
-} WebPAuxStats;
+};
 
 // Signature for output function. Should return true if writing was successful.
 // data/data_size is the segment of data to write, and 'picture' is for
@@ -192,18 +210,19 @@ typedef int (*WebPWriterFunction)(const uint8_t* data, size_t data_size,
 
 // WebPMemoryWrite: a special WebPWriterFunction that writes to memory using
 // the following WebPMemoryWriter object (to be set as a custom_ptr).
-typedef struct {
+struct WebPMemoryWriter {
   uint8_t* mem;       // final buffer (of size 'max_size', larger than 'size').
   size_t   size;      // final size
   size_t   max_size;  // total capacity
   uint32_t pad[1];    // padding for later use
-} WebPMemoryWriter;
+};
 
 // The following must be called first before any use.
 WEBP_EXTERN(void) WebPMemoryWriterInit(WebPMemoryWriter* writer);
 
 // The custom writer to be used with WebPMemoryWriter as custom_ptr. Upon
 // completion, writer.mem and writer.size will hold the coded data.
+// writer.mem must be freed using the call 'free(writer.mem)'.
 WEBP_EXTERN(int) WebPMemoryWrite(const uint8_t* data, size_t data_size,
                                  const WebPPicture* picture);
 
@@ -212,7 +231,8 @@ WEBP_EXTERN(int) WebPMemoryWrite(const uint8_t* data, size_t data_size,
 // everything is OK.
 typedef int (*WebPProgressHook)(int percent, const WebPPicture* picture);
 
-typedef enum {
+// Color spaces.
+typedef enum WebPEncCSP {
   // chroma sampling
   WEBP_YUV420 = 0,   // 4:2:0
   WEBP_YUV422 = 1,   // 4:2:2
@@ -228,7 +248,7 @@ typedef enum {
 } WebPEncCSP;
 
 // Encoding error conditions.
-typedef enum {
+typedef enum WebPEncodingError {
   VP8_ENC_OK = 0,
   VP8_ENC_ERROR_OUT_OF_MEMORY,            // memory error allocating objects
   VP8_ENC_ERROR_BITSTREAM_OUT_OF_MEMORY,  // memory error while flushing bits
@@ -248,7 +268,6 @@ typedef enum {
 
 // Main exchange structure (input samples, output bytes, statistics)
 struct WebPPicture {
-
   //   INPUT
   //////////////
   // Main flag for encoder selecting between ARGB or YUV input.
@@ -343,18 +362,19 @@ WEBP_EXTERN(int) WebPPictureAlloc(WebPPicture* picture);
 // preserved.
 WEBP_EXTERN(void) WebPPictureFree(WebPPicture* picture);
 
-// Copy the pixels of *src into *dst, using WebPPictureAlloc. Upon return,
-// *dst will fully own the copied pixels (this is not a view).
+// Copy the pixels of *src into *dst, using WebPPictureAlloc. Upon return, *dst
+// will fully own the copied pixels (this is not a view). The 'dst' picture need
+// not be initialized as its content is overwritten.
 // Returns false in case of memory allocation error.
 WEBP_EXTERN(int) WebPPictureCopy(const WebPPicture* src, WebPPicture* dst);
 
-// Compute PSNR or SSIM distortion between two pictures.
+// Compute PSNR, SSIM or LSIM distortion metric between two pictures.
 // Result is in dB, stores in result[] in the Y/U/V/Alpha/All order.
-// Returns false in case of error (pic1 and pic2 don't have same dimension, ...)
+// Returns false in case of error (src and ref don't have same dimension, ...)
 // Warning: this function is rather CPU-intensive.
 WEBP_EXTERN(int) WebPPictureDistortion(
-    const WebPPicture* pic1, const WebPPicture* pic2,
-    int metric_type,           // 0 = PSNR, 1 = SSIM
+    const WebPPicture* src, const WebPPicture* ref,
+    int metric_type,           // 0 = PSNR, 1 = SSIM, 2 = LSIM
     float result[5]);
 
 // self-crops a picture to the rectangle defined by top/left/width/height.
@@ -375,7 +395,9 @@ WEBP_EXTERN(int) WebPPictureCrop(WebPPicture* picture,
 // the top and left coordinates will be snapped to even values.
 // Picture 'src' must out-live 'dst' picture. Self-extraction of view is allowed
 // ('src' equal to 'dst') as a mean of fast-cropping (but note that doing so,
-// the original dimension will be lost).
+// the original dimension will be lost). Picture 'dst' need not be initialized
+// with WebPPictureInit() if it is different from 'src', since its content will
+// be overwritten.
 // Returns false in case of memory allocation error or invalid parameters.
 WEBP_EXTERN(int) WebPPictureView(const WebPPicture* src,
                                  int left, int top, int width, int height,
diff --git a/3rdparty/libwebp/webp/format_constants.h b/3rdparty/libwebp/webp/format_constants.h
index 7ce498f672..4c04b50c6a 100644
--- a/3rdparty/libwebp/webp/format_constants.h
+++ b/3rdparty/libwebp/webp/format_constants.h
@@ -1,8 +1,10 @@
 // Copyright 2012 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  Internal header for constants related to WebP file format.
@@ -12,6 +14,9 @@
 #ifndef WEBP_WEBP_FORMAT_CONSTANTS_H_
 #define WEBP_WEBP_FORMAT_CONSTANTS_H_
 
+// Create fourcc of the chunk from the chunk tag characters.
+#define MKFOURCC(a, b, c, d) ((uint32_t)(a) | (b) << 8 | (c) << 16 | (d) << 24)
+
 // VP8 related constants.
 #define VP8_SIGNATURE 0x9d012a              // Signature in VP8 data.
 #define VP8_MAX_PARTITION0_SIZE (1 << 19)   // max size of mode partition
@@ -65,23 +70,16 @@ typedef enum {
 #define CHUNK_SIZE_BYTES   4     // Size needed to store chunk's size.
 #define CHUNK_HEADER_SIZE  8     // Size of a chunk header.
 #define RIFF_HEADER_SIZE   12    // Size of the RIFF header ("RIFFnnnnWEBP").
-#define FRAME_CHUNK_SIZE   15    // Size of a FRM chunk.
-#define LOOP_CHUNK_SIZE    2     // Size of a LOOP chunk.
-#define TILE_CHUNK_SIZE    6     // Size of a TILE chunk.
+#define ANMF_CHUNK_SIZE    16    // Size of an ANMF chunk.
+#define ANIM_CHUNK_SIZE    6     // Size of an ANIM chunk.
+#define FRGM_CHUNK_SIZE    6     // Size of a FRGM chunk.
 #define VP8X_CHUNK_SIZE    10    // Size of a VP8X chunk.
 
-#define TILING_FLAG_BIT    0x01  // Set if tiles are possibly used.
-#define ANIMATION_FLAG_BIT 0x02  // Set if some animation is expected
-#define ICC_FLAG_BIT       0x04  // Whether ICC is present or not.
-#define METADATA_FLAG_BIT  0x08  // Set if some META chunk is possibly present.
-#define ALPHA_FLAG_BIT     0x10  // Should be same as the ALPHA_FLAG in mux.h
-#define ROTATION_FLAG_BITS 0xe0  // all 3 bits for rotation + symmetry
-
-#define MAX_CANVAS_SIZE     (1 << 24)    // 24-bit max for VP8X width/height.
-#define MAX_IMAGE_AREA      (1ULL << 32) // 32-bit max for width x height.
-#define MAX_LOOP_COUNT      (1 << 16)    // maximum value for loop-count
-#define MAX_DURATION        (1 << 24)    // maximum duration
-#define MAX_POSITION_OFFSET (1 << 24)    // maximum frame/tile x/y offset
+#define MAX_CANVAS_SIZE     (1 << 24)     // 24-bit max for VP8X width/height.
+#define MAX_IMAGE_AREA      (1ULL << 32)  // 32-bit max for width x height.
+#define MAX_LOOP_COUNT      (1 << 16)     // maximum value for loop-count
+#define MAX_DURATION        (1 << 24)     // maximum duration
+#define MAX_POSITION_OFFSET (1 << 24)     // maximum frame/fragment x/y offset
 
 // Maximum chunk payload is such that adding the header and padding won't
 // overflow a uint32_t.
diff --git a/3rdparty/libwebp/webp/mux.h b/3rdparty/libwebp/webp/mux.h
index 5139af80fa..b8c7dc62ec 100644
--- a/3rdparty/libwebp/webp/mux.h
+++ b/3rdparty/libwebp/webp/mux.h
@@ -1,8 +1,10 @@
 // Copyright 2011 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  RIFF container manipulation for WEBP images.
@@ -11,7 +13,7 @@
 //          Vikas (vikasa@google.com)
 
 // This API allows manipulation of WebP container images containing features
-// like Color profile, XMP metadata, Animation and Tiling.
+// like color profile, metadata, animation and fragmented images.
 //
 // Code Example#1: Creating a MUX with image data, color profile and XMP
 // metadata.
@@ -21,13 +23,13 @@
 //   // ... (Prepare image data).
 //   WebPMuxSetImage(mux, &image, copy_data);
 //   // ... (Prepare ICCP color profile data).
-//   WebPMuxSetColorProfile(mux, &icc_profile, copy_data);
+//   WebPMuxSetChunk(mux, "ICCP", &icc_profile, copy_data);
 //   // ... (Prepare XMP metadata).
-//   WebPMuxSetMetadata(mux, &xmp, copy_data);
+//   WebPMuxSetChunk(mux, "XMP ", &xmp, copy_data);
 //   // Get data from mux in WebP RIFF format.
 //   WebPMuxAssemble(mux, &output_data);
 //   WebPMuxDelete(mux);
-//   // ... (Consume output_data; e.g. write output_data.bytes_ to file).
+//   // ... (Consume output_data; e.g. write output_data.bytes to file).
 //   WebPDataClear(&output_data);
 //
 // Code Example#2: Get image and color profile data from a WebP file.
@@ -35,9 +37,9 @@
 //   int copy_data = 0;
 //   // ... (Read data from file).
 //   WebPMux* mux = WebPMuxCreate(&data, copy_data);
-//   WebPMuxGetImage(mux, &image);
+//   WebPMuxGetFrame(mux, 1, &image);
 //   // ... (Consume image; e.g. call WebPDecode() to decode the data).
-//   WebPMuxGetColorProfile(mux, &icc_profile);
+//   WebPMuxGetChunk(mux, "ICCP", &icc_profile);
 //   // ... (Consume icc_data).
 //   WebPMuxDelete(mux);
 //   free(data);
@@ -45,7 +47,7 @@
 #ifndef WEBP_WEBP_MUX_H_
 #define WEBP_WEBP_MUX_H_
 
-#include "./types.h"
+#include "./mux_types.h"
 
 #if defined(__cplusplus) || defined(c_plusplus)
 extern "C" {
@@ -53,8 +55,16 @@ extern "C" {
 
 #define WEBP_MUX_ABI_VERSION 0x0100        // MAJOR(8b) + MINOR(8b)
 
+// Note: forward declaring enumerations is not allowed in (strict) C and C++,
+// the types are left here for reference.
+// typedef enum WebPMuxError WebPMuxError;
+// typedef enum WebPChunkId WebPChunkId;
+typedef struct WebPMux WebPMux;   // main opaque object.
+typedef struct WebPMuxFrameInfo WebPMuxFrameInfo;
+typedef struct WebPMuxAnimParams WebPMuxAnimParams;
+
 // Error codes
-typedef enum {
+typedef enum WebPMuxError {
   WEBP_MUX_OK                 =  1,
   WEBP_MUX_NOT_FOUND          =  0,
   WEBP_MUX_INVALID_ARGUMENT   = -1,
@@ -63,51 +73,26 @@ typedef enum {
   WEBP_MUX_NOT_ENOUGH_DATA    = -4
 } WebPMuxError;
 
-// Flag values for different features used in VP8X chunk.
-typedef enum {
-  TILE_FLAG       = 0x00000001,
-  ANIMATION_FLAG  = 0x00000002,
-  ICCP_FLAG       = 0x00000004,
-  META_FLAG       = 0x00000008,
-  ALPHA_FLAG      = 0x00000010
-} WebPFeatureFlags;
-
 // IDs for different types of chunks.
-typedef enum {
+typedef enum WebPChunkId {
   WEBP_CHUNK_VP8X,     // VP8X
   WEBP_CHUNK_ICCP,     // ICCP
-  WEBP_CHUNK_LOOP,     // LOOP
-  WEBP_CHUNK_FRAME,    // FRM
-  WEBP_CHUNK_TILE,     // TILE
+  WEBP_CHUNK_ANIM,     // ANIM
+  WEBP_CHUNK_ANMF,     // ANMF
+  WEBP_CHUNK_FRGM,     // FRGM
   WEBP_CHUNK_ALPHA,    // ALPH
   WEBP_CHUNK_IMAGE,    // VP8/VP8L
-  WEBP_CHUNK_META,     // META
+  WEBP_CHUNK_EXIF,     // EXIF
+  WEBP_CHUNK_XMP,      // XMP
   WEBP_CHUNK_UNKNOWN,  // Other chunks.
   WEBP_CHUNK_NIL
 } WebPChunkId;
 
-typedef struct WebPMux WebPMux;   // main opaque object.
-
-// Data type used to describe 'raw' data, e.g., chunk data
-// (ICC profile, metadata) and WebP compressed image data.
-typedef struct {
-  const uint8_t* bytes_;
-  size_t size_;
-} WebPData;
-
 //------------------------------------------------------------------------------
-// Manipulation of a WebPData object.
 
-// Initializes the contents of the 'webp_data' object with default values.
-WEBP_EXTERN(void) WebPDataInit(WebPData* webp_data);
-
-// Clears the contents of the 'webp_data' object by calling free(). Does not
-// deallocate the object itself.
-WEBP_EXTERN(void) WebPDataClear(WebPData* webp_data);
-
-// Allocates necessary storage for 'dst' and copies the contents of 'src'.
-// Returns true on success.
-WEBP_EXTERN(int) WebPDataCopy(const WebPData* src, WebPData* dst);
+// Returns the version number of the mux library, packed in hexadecimal using
+// 8bits or each of major/minor/revision. E.g: v2.5.7 is 0x020507.
+WEBP_EXTERN(int) WebPGetMuxVersion(void);
 
 //------------------------------------------------------------------------------
 // Life of a Mux object
@@ -136,8 +121,8 @@ WEBP_EXTERN(WebPMux*) WebPMuxCreateInternal(const WebPData*, int, int);
 // Creates a mux object from raw data given in WebP RIFF format.
 // Parameters:
 //   bitstream - (in) the bitstream data in WebP RIFF format
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
+//   copy_data - (in) value 1 indicates given data WILL be copied to the mux
+//               and value 0 indicates data will NOT be copied.
 // Returns:
 //   A pointer to the mux object created from given data - on success.
 //   NULL - In case of invalid data or memory error.
@@ -147,270 +132,175 @@ static WEBP_INLINE WebPMux* WebPMuxCreate(const WebPData* bitstream,
 }
 
 //------------------------------------------------------------------------------
-// Single Image.
+// Non-image chunks.
 
-// Sets the image in the mux object. Any existing images (including frame/tile)
-// will be removed.
+// Note: Only non-image related chunks should be managed through chunk APIs.
+// (Image related chunks are: "ANMF", "FRGM", "VP8 ", "VP8L" and "ALPH").
+// To add, get and delete images, use APIs WebPMuxSetImage(),
+// WebPMuxPushFrame(), WebPMuxGetFrame() and WebPMuxDeleteFrame().
+
+// Adds a chunk with id 'fourcc' and data 'chunk_data' in the mux object.
+// Any existing chunk(s) with the same id will be removed.
+// Parameters:
+//   mux - (in/out) object to which the chunk is to be added
+//   fourcc - (in) a character array containing the fourcc of the given chunk;
+//                 e.g., "ICCP", "XMP ", "EXIF" etc.
+//   chunk_data - (in) the chunk data to be added
+//   copy_data - (in) value 1 indicates given data WILL be copied to the mux
+//               and value 0 indicates data will NOT be copied.
+// Returns:
+//   WEBP_MUX_INVALID_ARGUMENT - if mux, fourcc or chunk_data is NULL
+//                               or if fourcc corresponds to an image chunk.
+//   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
+//   WEBP_MUX_OK - on success.
+WEBP_EXTERN(WebPMuxError) WebPMuxSetChunk(
+    WebPMux* mux, const char fourcc[4], const WebPData* chunk_data,
+    int copy_data);
+
+// Gets a reference to the data of the chunk with id 'fourcc' in the mux object.
+// The caller should NOT free the returned data.
+// Parameters:
+//   mux - (in) object from which the chunk data is to be fetched
+//   fourcc - (in) a character array containing the fourcc of the chunk;
+//                 e.g., "ICCP", "XMP ", "EXIF" etc.
+//   chunk_data - (out) returned chunk data
+// Returns:
+//   WEBP_MUX_INVALID_ARGUMENT - if either mux, fourcc or chunk_data is NULL
+//                               or if fourcc corresponds to an image chunk.
+//   WEBP_MUX_NOT_FOUND - If mux does not contain a chunk with the given id.
+//   WEBP_MUX_OK - on success.
+WEBP_EXTERN(WebPMuxError) WebPMuxGetChunk(
+    const WebPMux* mux, const char fourcc[4], WebPData* chunk_data);
+
+// Deletes the chunk with the given 'fourcc' from the mux object.
+// Parameters:
+//   mux - (in/out) object from which the chunk is to be deleted
+//   fourcc - (in) a character array containing the fourcc of the chunk;
+//                 e.g., "ICCP", "XMP ", "EXIF" etc.
+// Returns:
+//   WEBP_MUX_INVALID_ARGUMENT - if mux or fourcc is NULL
+//                               or if fourcc corresponds to an image chunk.
+//   WEBP_MUX_NOT_FOUND - If mux does not contain a chunk with the given fourcc.
+//   WEBP_MUX_OK - on success.
+WEBP_EXTERN(WebPMuxError) WebPMuxDeleteChunk(
+    WebPMux* mux, const char fourcc[4]);
+
+//------------------------------------------------------------------------------
+// Images.
+
+// Encapsulates data about a single frame/fragment.
+struct WebPMuxFrameInfo {
+  WebPData    bitstream;  // image data: can either be a raw VP8/VP8L bitstream
+                          // or a single-image WebP file.
+  int         x_offset;   // x-offset of the frame.
+  int         y_offset;   // y-offset of the frame.
+  int         duration;   // duration of the frame (in milliseconds).
+
+  WebPChunkId id;         // frame type: should be one of WEBP_CHUNK_ANMF,
+                          // WEBP_CHUNK_FRGM or WEBP_CHUNK_IMAGE
+  WebPMuxAnimDispose dispose_method;  // Disposal method for the frame.
+  uint32_t    pad[2];     // padding for later use
+};
+
+// Sets the (non-animated and non-fragmented) image in the mux object.
+// Note: Any existing images (including frames/fragments) will be removed.
 // Parameters:
 //   mux - (in/out) object in which the image is to be set
 //   bitstream - (in) can either be a raw VP8/VP8L bitstream or a single-image
-//               WebP file (non-animated and non-tiled)
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
+//               WebP file (non-animated and non-fragmented)
+//   copy_data - (in) value 1 indicates given data WILL be copied to the mux
+//               and value 0 indicates data will NOT be copied.
 // Returns:
 //   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL or bitstream is NULL.
 //   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
 //   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxSetImage(WebPMux* mux,
-                                          const WebPData* bitstream,
-                                          int copy_data);
+WEBP_EXTERN(WebPMuxError) WebPMuxSetImage(
+    WebPMux* mux, const WebPData* bitstream, int copy_data);
 
-// Gets image data from the mux object.
-// The content of 'bitstream' is allocated using malloc(), and NOT
-// owned by the 'mux' object. It MUST be deallocated by the caller by calling
-// WebPDataClear().
+// Adds a frame at the end of the mux object.
+// Notes: (1) frame.id should be one of WEBP_CHUNK_ANMF or WEBP_CHUNK_FRGM
+//        (2) For setting a non-animated non-fragmented image, use
+//            WebPMuxSetImage() instead.
+//        (3) Type of frame being pushed must be same as the frames in mux.
+//        (4) As WebP only supports even offsets, any odd offset will be snapped
+//            to an even location using: offset &= ~1
 // Parameters:
-//   mux - (in) object from which the image is to be fetched
-//   bitstream - (out) the image data
+//   mux - (in/out) object to which the frame is to be added
+//   frame - (in) frame data.
+//   copy_data - (in) value 1 indicates given data WILL be copied to the mux
+//               and value 0 indicates data will NOT be copied.
 // Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either mux or bitstream is NULL
-//                               OR mux contains animation/tiling.
-//   WEBP_MUX_NOT_FOUND - if image is not present in mux object.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxGetImage(const WebPMux* mux,
-                                          WebPData* bitstream);
-
-// Deletes the image in the mux object.
-// Parameters:
-//   mux - (in/out) object from which the image is to be deleted
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
-//                               OR if mux contains animation/tiling.
-//   WEBP_MUX_NOT_FOUND - if image is not present in mux object.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxDeleteImage(WebPMux* mux);
-
-//------------------------------------------------------------------------------
-// XMP Metadata.
-
-// Sets the XMP metadata in the mux object. Any existing metadata chunk(s) will
-// be removed.
-// Parameters:
-//   mux - (in/out) object to which the XMP metadata is to be added
-//   metadata - (in) the XMP metadata data to be added
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux or metadata is NULL.
-//   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxSetMetadata(WebPMux* mux,
-                                             const WebPData* metadata,
-                                             int copy_data);
-
-// Gets a reference to the XMP metadata in the mux object.
-// The caller should NOT free the returned data.
-// Parameters:
-//   mux - (in) object from which the XMP metadata is to be fetched
-//   metadata - (out) XMP metadata
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either mux or metadata is NULL.
-//   WEBP_MUX_NOT_FOUND - if metadata is not present in mux object.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxGetMetadata(const WebPMux* mux,
-                                             WebPData* metadata);
-
-// Deletes the XMP metadata in the mux object.
-// Parameters:
-//   mux - (in/out) object from which XMP metadata is to be deleted
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
-//   WEBP_MUX_NOT_FOUND - If mux does not contain metadata.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxDeleteMetadata(WebPMux* mux);
-
-//------------------------------------------------------------------------------
-// ICC Color Profile.
-
-// Sets the color profile in the mux object. Any existing color profile chunk(s)
-// will be removed.
-// Parameters:
-//   mux - (in/out) object to which the color profile is to be added
-//   color_profile - (in) the color profile data to be added
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux or color_profile is NULL
-//   WEBP_MUX_MEMORY_ERROR - on memory allocation error
-//   WEBP_MUX_OK - on success
-WEBP_EXTERN(WebPMuxError) WebPMuxSetColorProfile(WebPMux* mux,
-                                                 const WebPData* color_profile,
-                                                 int copy_data);
-
-// Gets a reference to the color profile in the mux object.
-// The caller should NOT free the returned data.
-// Parameters:
-//   mux - (in) object from which the color profile data is to be fetched
-//   color_profile - (out) color profile data
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either mux or color_profile is NULL.
-//   WEBP_MUX_NOT_FOUND - if color profile is not present in mux object.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxGetColorProfile(const WebPMux* mux,
-                                                 WebPData* color_profile);
-
-// Deletes the color profile in the mux object.
-// Parameters:
-//   mux - (in/out) object from which color profile is to be deleted
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
-//   WEBP_MUX_NOT_FOUND - If mux does not contain color profile.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxDeleteColorProfile(WebPMux* mux);
-
-//------------------------------------------------------------------------------
-// Animation.
-
-// Adds an animation frame at the end of the mux object.
-// Note: as WebP only supports even offsets, any odd offset will be snapped to
-// an even location using: offset &= ~1
-// Parameters:
-//   mux - (in/out) object to which an animation frame is to be added
-//   bitstream - (in) the image data corresponding to the frame. It can either
-//               be a raw VP8/VP8L bitstream or a single-image WebP file
-//               (non-animated and non-tiled)
-//   x_offset - (in) x-offset of the frame to be added
-//   y_offset - (in) y-offset of the frame to be added
-//   duration - (in) duration of the frame to be added (in milliseconds)
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL or bitstream is NULL
+//   WEBP_MUX_INVALID_ARGUMENT - if mux or frame is NULL
+//                               or if content of 'frame' is invalid.
 //   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
 //   WEBP_MUX_OK - on success.
 WEBP_EXTERN(WebPMuxError) WebPMuxPushFrame(
-    WebPMux* mux, const WebPData* bitstream,
-    int x_offset, int y_offset, int duration, int copy_data);
+    WebPMux* mux, const WebPMuxFrameInfo* frame, int copy_data);
 
-// TODO(urvang): Create a struct as follows to reduce argument list size:
-// typedef struct {
-//  WebPData bitstream;
-//  int x_offset, y_offset;
-//  int duration;
-// } FrameInfo;
-
-// Gets the nth animation frame from the mux object.
-// The content of 'bitstream' is allocated using malloc(), and NOT
+// Gets the nth frame from the mux object.
+// The content of 'frame->bitstream' is allocated using malloc(), and NOT
 // owned by the 'mux' object. It MUST be deallocated by the caller by calling
 // WebPDataClear().
 // nth=0 has a special meaning - last position.
 // Parameters:
 //   mux - (in) object from which the info is to be fetched
 //   nth - (in) index of the frame in the mux object
-//   bitstream - (out) the image data
-//   x_offset - (out) x-offset of the returned frame
-//   y_offset - (out) y-offset of the returned frame
-//   duration - (out) duration of the returned frame (in milliseconds)
+//   frame - (out) data of the returned frame
 // Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either mux, bitstream, x_offset,
-//                               y_offset, or duration is NULL
+//   WEBP_MUX_INVALID_ARGUMENT - if mux or frame is NULL.
 //   WEBP_MUX_NOT_FOUND - if there are less than nth frames in the mux object.
 //   WEBP_MUX_BAD_DATA - if nth frame chunk in mux is invalid.
 //   WEBP_MUX_OK - on success.
 WEBP_EXTERN(WebPMuxError) WebPMuxGetFrame(
-    const WebPMux* mux, uint32_t nth, WebPData* bitstream,
-    int* x_offset, int* y_offset, int* duration);
+    const WebPMux* mux, uint32_t nth, WebPMuxFrameInfo* frame);
 
-// Deletes an animation frame from the mux object.
+// Deletes a frame from the mux object.
 // nth=0 has a special meaning - last position.
 // Parameters:
 //   mux - (in/out) object from which a frame is to be deleted
 //   nth - (in) The position from which the frame is to be deleted
 // Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
+//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL.
 //   WEBP_MUX_NOT_FOUND - If there are less than nth frames in the mux object
 //                        before deletion.
 //   WEBP_MUX_OK - on success.
 WEBP_EXTERN(WebPMuxError) WebPMuxDeleteFrame(WebPMux* mux, uint32_t nth);
 
-// Sets the animation loop count in the mux object. Any existing loop count
-// value(s) will be removed.
-// Parameters:
-//   mux - (in/out) object in which loop chunk is to be set/added
-//   loop_count - (in) animation loop count value.
-//                Note that loop_count of zero denotes infinite loop.
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
-//   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxSetLoopCount(WebPMux* mux, int loop_count);
-
-// Gets the animation loop count from the mux object.
-// Parameters:
-//   mux - (in) object from which the loop count is to be fetched
-//   loop_count - (out) the loop_count value present in the LOOP chunk
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either of mux or loop_count is NULL
-//   WEBP_MUX_NOT_FOUND - if loop chunk is not present in mux object.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxGetLoopCount(const WebPMux* mux,
-                                              int* loop_count);
-
 //------------------------------------------------------------------------------
-// Tiling.
+// Animation.
 
-// Adds a tile at the end of the mux object.
-// Note: as WebP only supports even offsets, any odd offset will be snapped to
-// an even location using: offset &= ~1
+// Animation parameters.
+struct WebPMuxAnimParams {
+  uint32_t bgcolor;  // Background color of the canvas stored (in MSB order) as:
+                     // Bits 00 to 07: Alpha.
+                     // Bits 08 to 15: Red.
+                     // Bits 16 to 23: Green.
+                     // Bits 24 to 31: Blue.
+  int loop_count;    // Number of times to repeat the animation [0 = infinite].
+};
+
+// Sets the animation parameters in the mux object. Any existing ANIM chunks
+// will be removed.
 // Parameters:
-//   mux - (in/out) object to which a tile is to be added.
-//   bitstream - (in) the image data corresponding to the frame. It can either
-//               be a raw VP8/VP8L bitstream or a single-image WebP file
-//               (non-animated and non-tiled)
-//   x_offset - (in) x-offset of the tile to be added
-//   y_offset - (in) y-offset of the tile to be added
-//   copy_data - (in) value 1 indicates given data WILL copied to the mux, and
-//               value 0 indicates data will NOT be copied.
+//   mux - (in/out) object in which ANIM chunk is to be set/added
+//   params - (in) animation parameters.
 // Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL or bitstream is NULL
+//   WEBP_MUX_INVALID_ARGUMENT - if either mux or params is NULL
 //   WEBP_MUX_MEMORY_ERROR - on memory allocation error.
 //   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxPushTile(
-    WebPMux* mux, const WebPData* bitstream,
-    int x_offset, int y_offset, int copy_data);
+WEBP_EXTERN(WebPMuxError) WebPMuxSetAnimationParams(
+    WebPMux* mux, const WebPMuxAnimParams* params);
 
-// Gets the nth tile from the mux object.
-// The content of 'bitstream' is allocated using malloc(), and NOT
-// owned by the 'mux' object. It MUST be deallocated by the caller by calling
-// WebPDataClear().
-// nth=0 has a special meaning - last position.
+// Gets the animation parameters from the mux object.
 // Parameters:
-//   mux - (in) object from which the info is to be fetched
-//   nth - (in) index of the tile in the mux object
-//   bitstream - (out) the image data
-//   x_offset - (out) x-offset of the returned tile
-//   y_offset - (out) y-offset of the returned tile
+//   mux - (in) object from which the animation parameters to be fetched
+//   params - (out) animation parameters extracted from the ANIM chunk
 // Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if either mux, bitstream, x_offset or
-//                               y_offset is NULL
-//   WEBP_MUX_NOT_FOUND - if there are less than nth tiles in the mux object.
-//   WEBP_MUX_BAD_DATA - if nth tile chunk in mux is invalid.
+//   WEBP_MUX_INVALID_ARGUMENT - if either of mux or params is NULL
+//   WEBP_MUX_NOT_FOUND - if ANIM chunk is not present in mux object.
 //   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxGetTile(
-    const WebPMux* mux, uint32_t nth, WebPData* bitstream,
-    int* x_offset, int* y_offset);
-
-// Deletes a tile from the mux object.
-// nth=0 has a special meaning - last position
-// Parameters:
-//   mux - (in/out) object from which a tile is to be deleted
-//   nth - (in) The position from which the tile is to be deleted
-// Returns:
-//   WEBP_MUX_INVALID_ARGUMENT - if mux is NULL
-//   WEBP_MUX_NOT_FOUND - If there are less than nth tiles in the mux object
-//                        before deletion.
-//   WEBP_MUX_OK - on success.
-WEBP_EXTERN(WebPMuxError) WebPMuxDeleteTile(WebPMux* mux, uint32_t nth);
+WEBP_EXTERN(WebPMuxError) WebPMuxGetAnimationParams(
+    const WebPMux* mux, WebPMuxAnimParams* params);
 
 //------------------------------------------------------------------------------
 // Misc Utilities.
@@ -458,143 +348,6 @@ WEBP_EXTERN(WebPMuxError) WebPMuxNumChunks(const WebPMux* mux,
 WEBP_EXTERN(WebPMuxError) WebPMuxAssemble(WebPMux* mux,
                                           WebPData* assembled_data);
 
-//------------------------------------------------------------------------------
-// Demux API.
-// Enables extraction of image and extended format data from WebP files.
-
-#define WEBP_DEMUX_ABI_VERSION 0x0100    // MAJOR(8b) + MINOR(8b)
-
-typedef struct WebPDemuxer WebPDemuxer;
-
-typedef enum {
-  WEBP_DEMUX_PARSING_HEADER,  // Not enough data to parse full header.
-  WEBP_DEMUX_PARSED_HEADER,   // Header parsing complete, data may be available.
-  WEBP_DEMUX_DONE             // Entire file has been parsed.
-} WebPDemuxState;
-
-//------------------------------------------------------------------------------
-// Life of a Demux object
-
-// Internal, version-checked, entry point
-WEBP_EXTERN(WebPDemuxer*) WebPDemuxInternal(
-    const WebPData*, int, WebPDemuxState*, int);
-
-// Parses the WebP file given by 'data'.
-// A complete WebP file must be present in 'data' for the function to succeed.
-// Returns a WebPDemuxer object on successful parse, NULL otherwise.
-static WEBP_INLINE WebPDemuxer* WebPDemux(const WebPData* data) {
-  return WebPDemuxInternal(data, 0, NULL, WEBP_DEMUX_ABI_VERSION);
-}
-
-// Parses the WebP file given by 'data'.
-// If 'state' is non-NULL it will be set to indicate the status of the demuxer.
-// Returns a WebPDemuxer object on successful parse, NULL otherwise.
-static WEBP_INLINE WebPDemuxer* WebPDemuxPartial(
-    const WebPData* data, WebPDemuxState* state) {
-  return WebPDemuxInternal(data, 1, state, WEBP_DEMUX_ABI_VERSION);
-}
-
-// Frees memory associated with 'dmux'.
-WEBP_EXTERN(void) WebPDemuxDelete(WebPDemuxer* dmux);
-
-//------------------------------------------------------------------------------
-// Data/information extraction.
-
-typedef enum {
-  WEBP_FF_FORMAT_FLAGS,  // Extended format flags present in the 'VP8X' chunk.
-  WEBP_FF_CANVAS_WIDTH,
-  WEBP_FF_CANVAS_HEIGHT,
-  WEBP_FF_LOOP_COUNT
-} WebPFormatFeature;
-
-// Get the 'feature' value from the 'dmux'.
-// NOTE: values are only valid if WebPDemux() was used or WebPDemuxPartial()
-// returned a state > WEBP_DEMUX_PARSING_HEADER.
-WEBP_EXTERN(uint32_t) WebPDemuxGetI(
-    const WebPDemuxer* dmux, WebPFormatFeature feature);
-
-//------------------------------------------------------------------------------
-// Frame iteration.
-
-typedef struct {
-  int frame_num_;
-  int num_frames_;
-  int tile_num_;
-  int num_tiles_;
-  int x_offset_, y_offset_;  // offset relative to the canvas.
-  int width_, height_;       // dimensions of this frame or tile.
-  int duration_;   // display duration in milliseconds.
-  int complete_;   // true if 'tile_' contains a full frame. partial images may
-                   // still be decoded with the WebP incremental decoder.
-  WebPData tile_;  // The frame or tile given by 'frame_num_' and 'tile_num_'.
-
-  uint32_t pad[4];           // padding for later use
-  void* private_;
-} WebPIterator;
-
-// Retrieves frame 'frame_number' from 'dmux'.
-// 'iter->tile_' points to the first tile on return from this function.
-// Individual tiles may be extracted using WebPDemuxSetTile().
-// Setting 'frame_number' equal to 0 will return the last frame of the image.
-// Returns false if 'dmux' is NULL or frame 'frame_number' is not present.
-// Call WebPDemuxReleaseIterator() when use of the iterator is complete.
-// NOTE: 'dmux' must persist for the lifetime of 'iter'.
-WEBP_EXTERN(int) WebPDemuxGetFrame(
-    const WebPDemuxer* dmux, int frame_number, WebPIterator* iter);
-
-// Sets 'iter->tile_' to point to the next ('iter->frame_num_' + 1) or previous
-// ('iter->frame_num_' - 1) frame. These functions do not loop.
-// Returns true on success, false otherwise.
-WEBP_EXTERN(int) WebPDemuxNextFrame(WebPIterator* iter);
-WEBP_EXTERN(int) WebPDemuxPrevFrame(WebPIterator* iter);
-
-// Sets 'iter->tile_' to reflect tile number 'tile_number'.
-// Returns true if tile 'tile_number' is present, false otherwise.
-WEBP_EXTERN(int) WebPDemuxSelectTile(WebPIterator* iter, int tile_number);
-
-// Releases any memory associated with 'iter'.
-// Must be called before destroying the associated WebPDemuxer with
-// WebPDemuxDelete().
-WEBP_EXTERN(void) WebPDemuxReleaseIterator(WebPIterator* iter);
-
-//------------------------------------------------------------------------------
-// Chunk iteration.
-
-typedef struct {
-  // The current and total number of chunks with the fourcc given to
-  // WebPDemuxGetChunk().
-  int chunk_num_;
-  int num_chunks_;
-  WebPData chunk_;    // The payload of the chunk.
-
-  uint32_t pad[6];    // padding for later use
-  void* private_;
-} WebPChunkIterator;
-
-// Retrieves the 'chunk_number' instance of the chunk with id 'fourcc' from
-// 'dmux'.
-// 'fourcc' is a character array containing the fourcc of the chunk to return,
-// e.g., "ICCP", "META", "EXIF", etc.
-// Setting 'chunk_number' equal to 0 will return the last chunk in a set.
-// Returns true if the chunk is found, false otherwise. Image related chunk
-// payloads are accessed through WebPDemuxGetFrame() and related functions.
-// Call WebPDemuxReleaseChunkIterator() when use of the iterator is complete.
-// NOTE: 'dmux' must persist for the lifetime of the iterator.
-WEBP_EXTERN(int) WebPDemuxGetChunk(const WebPDemuxer* dmux,
-                                   const char fourcc[4], int chunk_number,
-                                   WebPChunkIterator* iter);
-
-// Sets 'iter->chunk_' to point to the next ('iter->chunk_num_' + 1) or previous
-// ('iter->chunk_num_' - 1) chunk. These functions do not loop.
-// Returns true on success, false otherwise.
-WEBP_EXTERN(int) WebPDemuxNextChunk(WebPChunkIterator* iter);
-WEBP_EXTERN(int) WebPDemuxPrevChunk(WebPChunkIterator* iter);
-
-// Releases any memory associated with 'iter'.
-// Must be called before destroying the associated WebPDemuxer with
-// WebPDemuxDelete().
-WEBP_EXTERN(void) WebPDemuxReleaseChunkIterator(WebPChunkIterator* iter);
-
 //------------------------------------------------------------------------------
 
 #if defined(__cplusplus) || defined(c_plusplus)
diff --git a/3rdparty/libwebp/webp/mux_types.h b/3rdparty/libwebp/webp/mux_types.h
new file mode 100644
index 0000000000..b8bce363e5
--- /dev/null
+++ b/3rdparty/libwebp/webp/mux_types.h
@@ -0,0 +1,89 @@
+// Copyright 2012 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Data-types common to the mux and demux libraries.
+//
+// Author: Urvang (urvang@google.com)
+
+#ifndef WEBP_WEBP_MUX_TYPES_H_
+#define WEBP_WEBP_MUX_TYPES_H_
+
+#include <stdlib.h>  // free()
+#include <string.h>  // memset()
+#include "./types.h"
+
+#if defined(__cplusplus) || defined(c_plusplus)
+extern "C" {
+#endif
+
+// Note: forward declaring enumerations is not allowed in (strict) C and C++,
+// the types are left here for reference.
+// typedef enum WebPFeatureFlags WebPFeatureFlags;
+// typedef enum WebPMuxAnimDispose WebPMuxAnimDispose;
+typedef struct WebPData WebPData;
+
+// VP8X Feature Flags.
+typedef enum WebPFeatureFlags {
+  FRAGMENTS_FLAG  = 0x00000001,
+  ANIMATION_FLAG  = 0x00000002,
+  XMP_FLAG        = 0x00000004,
+  EXIF_FLAG       = 0x00000008,
+  ALPHA_FLAG      = 0x00000010,
+  ICCP_FLAG       = 0x00000020
+} WebPFeatureFlags;
+
+// Dispose method (animation only). Indicates how the area used by the current
+// frame is to be treated before rendering the next frame on the canvas.
+typedef enum WebPMuxAnimDispose {
+  WEBP_MUX_DISPOSE_NONE,       // Do not dispose.
+  WEBP_MUX_DISPOSE_BACKGROUND  // Dispose to background color.
+} WebPMuxAnimDispose;
+
+// Data type used to describe 'raw' data, e.g., chunk data
+// (ICC profile, metadata) and WebP compressed image data.
+struct WebPData {
+  const uint8_t* bytes;
+  size_t size;
+};
+
+// Initializes the contents of the 'webp_data' object with default values.
+static WEBP_INLINE void WebPDataInit(WebPData* webp_data) {
+  if (webp_data != NULL) {
+    memset(webp_data, 0, sizeof(*webp_data));
+  }
+}
+
+// Clears the contents of the 'webp_data' object by calling free(). Does not
+// deallocate the object itself.
+static WEBP_INLINE void WebPDataClear(WebPData* webp_data) {
+  if (webp_data != NULL) {
+    free((void*)webp_data->bytes);
+    WebPDataInit(webp_data);
+  }
+}
+
+// Allocates necessary storage for 'dst' and copies the contents of 'src'.
+// Returns true on success.
+static WEBP_INLINE int WebPDataCopy(const WebPData* src, WebPData* dst) {
+  if (src == NULL || dst == NULL) return 0;
+  WebPDataInit(dst);
+  if (src->bytes != NULL && src->size != 0) {
+    dst->bytes = (uint8_t*)malloc(src->size);
+    if (dst->bytes == NULL) return 0;
+    memcpy((void*)dst->bytes, src->bytes, src->size);
+    dst->size = src->size;
+  }
+  return 1;
+}
+
+#if defined(__cplusplus) || defined(c_plusplus)
+}    // extern "C"
+#endif
+
+#endif  /* WEBP_WEBP_MUX_TYPES_H_ */
diff --git a/3rdparty/libwebp/webp/types.h b/3rdparty/libwebp/webp/types.h
index 3e27190bef..568d1f263f 100644
--- a/3rdparty/libwebp/webp/types.h
+++ b/3rdparty/libwebp/webp/types.h
@@ -1,8 +1,10 @@
 // Copyright 2010 Google Inc. All Rights Reserved.
 //
-// This code is licensed under the same terms as WebM:
-//  Software License Agreement:  http://www.webmproject.org/license/software/
-//  Additional IP Rights Grant:  http://www.webmproject.org/license/additional/
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
 //  Common types
diff --git a/3rdparty/openexr/CMakeLists.txt b/3rdparty/openexr/CMakeLists.txt
index a77a6e1326..e10b940079 100644
--- a/3rdparty/openexr/CMakeLists.txt
+++ b/3rdparty/openexr/CMakeLists.txt
@@ -22,12 +22,14 @@ set(OPENEXR_INCLUDE_PATHS "${CMAKE_CURRENT_SOURCE_DIR}/Half"
                           "${CMAKE_CURRENT_SOURCE_DIR}/Imath"
                           "${CMAKE_CURRENT_SOURCE_DIR}/IlmImf")
 
-ocv_include_directories("${CMAKE_CURRENT_BINARY_DIR}" ${ZLIB_INCLUDE_DIR} ${OPENEXR_INCLUDE_PATHS})
+ocv_include_directories("${CMAKE_CURRENT_BINARY_DIR}" ${ZLIB_INCLUDE_DIRS} ${OPENEXR_INCLUDE_PATHS})
 
 file(GLOB lib_srcs Half/half.cpp Iex/*.cpp IlmThread/*.cpp Imath/*.cpp IlmImf/*.cpp)
 file(GLOB lib_hdrs Half/*.h Iex/Iex*.h IlmThread/IlmThread*.h Imath/Imath*.h IlmImf/*.h)
 list(APPEND lib_hdrs "${CMAKE_CURRENT_BINARY_DIR}/IlmBaseConfig.h" "${CMAKE_CURRENT_BINARY_DIR}/OpenEXRConfig.h")
 
+ocv_list_filterout(lib_srcs IlmImf/b44ExpLogTable.cpp)
+
 if(WIN32)
   ocv_list_filterout(lib_srcs Posix.*cpp)
 else()
diff --git a/3rdparty/tbb/.gitignore b/3rdparty/tbb/.gitignore
index 601e1b265e..69ce9ceb9c 100644
--- a/3rdparty/tbb/.gitignore
+++ b/3rdparty/tbb/.gitignore
@@ -1 +1 @@
-tbb*.tgz
\ No newline at end of file
+tbb*.tgz
diff --git a/3rdparty/tbb/CMakeLists.txt b/3rdparty/tbb/CMakeLists.txt
index 3c6ae98f45..4e510ffa32 100644
--- a/3rdparty/tbb/CMakeLists.txt
+++ b/3rdparty/tbb/CMakeLists.txt
@@ -1,12 +1,39 @@
 #Cross compile TBB from source
 project(tbb)
 
+if (WIN32 AND NOT ARM)
+  message(FATAL_ERROR "BUILD_TBB option supports Windows on ARM only!\nUse regular official TBB build instead of the BUILD_TBB option!")
+endif()
+
+if (WIN32 AND ARM)
+  # 4.1 update 4 - The first release that supports Windows RT. Hangs on some Android devices
+  set(tbb_ver "tbb41_20130613oss")
+  set(tbb_url "http://threadingbuildingblocks.org/sites/default/files/software_releases/source/tbb41_20130613oss_src.tgz")
+  set(tbb_md5 "108c8c1e481b0aaea61878289eb28b6a")
+  set(tbb_version_file "version_string.ver")
+  ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4702)
+else()
+  # 4.1 update 2 - works fine
+  set(tbb_ver "tbb41_20130116oss")
+  set(tbb_url "http://threadingbuildingblocks.org/sites/default/files/software_releases/source/tbb41_20130116oss_src.tgz")
+  set(tbb_md5 "3809790e1001a1b32d59c9fee590ee85")
+  set(tbb_version_file "version_string.ver")
+  ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
+endif()
+
+# 4.1 update 3 dev - Hangs on some Android devices
+#set(tbb_ver "tbb41_20130401oss")
+#set(tbb_url "http://threadingbuildingblocks.org/sites/default/files/software_releases/source/tbb41_20130401oss_src.tgz")
+#set(tbb_md5 "f2f591a0d2ca8f801e221ce7d9ea84bb")
+#set(tbb_version_file "version_string.ver")
+#ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
+
 # 4.1 update 2 - works fine
-set(tbb_ver "tbb41_20130116oss")
-set(tbb_url "http://threadingbuildingblocks.org/sites/default/files/software_releases/source/tbb41_20130116oss_src.tgz")
-set(tbb_md5 "3809790e1001a1b32d59c9fee590ee85")
-set(tbb_version_file "version_string.ver")
-ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
+#set(tbb_ver "tbb41_20130116oss")
+#set(tbb_url "http://threadingbuildingblocks.org/sites/default/files/software_releases/source/tbb41_20130116oss_src.tgz")
+#set(tbb_md5 "3809790e1001a1b32d59c9fee590ee85")
+#set(tbb_version_file "version_string.ver")
+#ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
 
 # 4.1 update 1 - works fine
 #set(tbb_ver "tbb41_20121003oss")
@@ -55,24 +82,8 @@ ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
 set(tbb_tarball "${CMAKE_CURRENT_SOURCE_DIR}/${tbb_ver}_src.tgz")
 set(tbb_src_dir "${CMAKE_CURRENT_BINARY_DIR}/${tbb_ver}")
 
-macro(getMD5 filename varname)
-  if(CMAKE_VERSION VERSION_GREATER 2.8.6)
-    file(MD5 "${filename}" ${varname})
-  else()
-    execute_process(COMMAND ${CMAKE_COMMAND} -E md5sum "${filename}"
-              RESULT_VARIABLE getMD5_RESULT
-              OUTPUT_VARIABLE getMD5_OUTPUT
-              OUTPUT_STRIP_TRAILING_WHITESPACE)
-    if(getMD5_RESULT EQUAL 0)
-      string(REGEX MATCH "^[a-z0-9]+" ${varname} "${getMD5_OUTPUT}")
-    else()
-      set(${varname} "invalid_md5")
-    endif()
-  endif()
-endmacro()
-
 if(EXISTS "${tbb_tarball}")
-  getMD5("${tbb_tarball}" tbb_local_md5)
+  file(MD5 "${tbb_tarball}" tbb_local_md5)
   if(NOT tbb_local_md5 STREQUAL tbb_md5)
     message(WARNING "Local copy of TBB source tarball has invalid MD5 hash: ${tbb_local_md5} (expected: ${tbb_md5})")
     file(REMOVE "${tbb_tarball}")
@@ -88,7 +99,7 @@ if(NOT EXISTS "${tbb_tarball}")
   if(NOT __statvar EQUAL 0)
     message(FATAL_ERROR "Failed to download TBB sources: ${tbb_url}")
   endif()
-  getMD5("${tbb_tarball}" tbb_local_md5)
+  file(MD5 "${tbb_tarball}" tbb_local_md5)
   if(NOT tbb_local_md5 STREQUAL tbb_md5)
     file(REMOVE "${tbb_tarball}")
     message(FATAL_ERROR "Downloaded TBB source tarball has invalid MD5 hash: ${tbb_local_md5} (expected: ${tbb_md5})")
@@ -107,7 +118,7 @@ if(NOT EXISTS "${tbb_src_dir}")
               RESULT_VARIABLE tbb_untar_RESULT)
 
   if(NOT tbb_untar_RESULT EQUAL 0 OR NOT EXISTS "${tbb_src_dir}")
-    message(FATAL_ERROR "Failed to unpack TBB sources")
+    message(FATAL_ERROR "Failed to unpack TBB sources from ${tbb_tarball} to ${tbb_src_dir} with error ${tbb_untar_RESULT}")
   endif()
 endif()
 
@@ -122,15 +133,41 @@ file(GLOB lib_srcs "${tbb_src_dir}/src/tbb/*.cpp")
 file(GLOB lib_hdrs "${tbb_src_dir}/src/tbb/*.h")
 list(APPEND lib_srcs "${tbb_src_dir}/src/rml/client/rml_tbb.cpp")
 
-add_definitions(-D__TBB_DYNAMIC_LOAD_ENABLED=0         #required
-                -D__TBB_BUILD=1                        #required
-                -D__TBB_SURVIVE_THREAD_SWITCH=0        #no cilk support
-                -DUSE_PTHREAD                          #required for Unix
-                -DTBB_USE_GCC_BUILTINS=1               #required for ARM GCC
-                -DTBB_USE_DEBUG=0                      #just to be sure
-                -DTBB_NO_LEGACY=1                      #don't need backward compatibility
-                -DDO_ITT_NOTIFY=0                      #it seems that we don't need these notifications
-               )
+if (WIN32)
+  add_definitions(/D__TBB_DYNAMIC_LOAD_ENABLED=0
+                  /D__TBB_BUILD=1
+                  /DTBB_NO_LEGACY=1
+                  /D_UNICODE
+                  /DUNICODE
+                  /DWINAPI_FAMILY=WINAPI_FAMILY_APP
+                  /DDO_ITT_NOTIFY=0
+                  /DUSE_WINTHREAD
+               ) # defines were copied from windows.cl.inc
+
+  if (ARM)
+    add_definitions(/D_WIN32_WINNT=0x0602
+                    /D__TBB_WIN32_USE_CL_BUILTINS
+                   )
+  endif()
+
+set(CMAKE_LINKER_FLAGS "${CMAKE_LINKER_FLAGS} /APPCONTAINER")
+else()
+  add_definitions(-D__TBB_DYNAMIC_LOAD_ENABLED=0         #required
+                  -D__TBB_BUILD=1                        #required
+                  -D__TBB_SURVIVE_THREAD_SWITCH=0        #no cilk support
+                  -DTBB_USE_DEBUG=0                      #just to be sure
+                  -DTBB_NO_LEGACY=1                      #don't need backward compatibility
+                  -DDO_ITT_NOTIFY=0                      #it seems that we don't need these notifications
+                 )
+endif()
+
+if (HAVE_LIBPTHREAD)
+  add_definitions(-DUSE_PTHREAD) #required for Unix
+endif()
+
+if (CMAKE_COMPILER_IS_GNUCXX)
+  add_definitions(-DTBB_USE_GCC_BUILTINS=1) #required for ARM GCC
+endif()
 
 if(ANDROID_COMPILER_IS_CLANG)
   add_definitions(-D__TBB_GCC_BUILTIN_ATOMICS_PRESENT=1)
@@ -145,7 +182,7 @@ endif()
 
 set(TBB_SOURCE_FILES ${lib_srcs} ${lib_hdrs})
 
-if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm")
+if (ARM AND NOT WIN32)
   if (NOT ANDROID)
     set(TBB_SOURCE_FILES ${TBB_SOURCE_FILES} "${CMAKE_CURRENT_SOURCE_DIR}/arm_linux_stub.cpp")
   endif()
@@ -156,15 +193,38 @@ endif()
 set(TBB_SOURCE_FILES ${TBB_SOURCE_FILES} "${CMAKE_CURRENT_SOURCE_DIR}/${tbb_version_file}")
 
 add_library(tbb ${TBB_SOURCE_FILES})
-target_link_libraries(tbb c m dl)
+
+if (WIN32)
+  if (ARM)
+    set(platform_macro /D_M_ARM=1)
+  endif()
+
+  add_custom_command(TARGET tbb
+                     PRE_BUILD
+                     COMMAND ${CMAKE_C_COMPILER} /nologo /TC /EP ${tbb_src_dir}\\src\\tbb\\win32-tbb-export.def /DTBB_NO_LEGACY=1 /D_CRT_SECURE_NO_DEPRECATE /D__TBB_BUILD=1 ${platform_macro} /I${tbb_src_dir}\\src /I${tbb_src_dir}\\include > "${tbb_src_dir}\\src\\tbb\\tbb.def"
+                     WORKING_DIRECTORY ${tbb_src_dir}\\src\\tbb
+                     COMMENT "Generating tbb.def file" VERBATIM
+                    )
+
+  set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /DEF:${tbb_src_dir}/src/tbb/tbb.def /DLL /MAP /fixed:no /INCREMENTAL:NO")
+else()
+  target_link_libraries(tbb c m dl)
+endif()
 
 ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wmissing-declarations)
 string(REPLACE "-Werror=non-virtual-dtor" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
 
+if (WIN32)
+  set(tbb_debug_postfix "_debug") # to fit pragmas in _windef.h inside TBB
+else()
+  set(tbb_debug_postfix ${OPENCV_DEBUG_POSTFIX})
+endif()
+
 set_target_properties(tbb
   PROPERTIES OUTPUT_NAME tbb
-  DEBUG_POSTFIX "${OPENCV_DEBUG_POSTFIX}"
+  DEBUG_POSTFIX "${tbb_debug_postfix}"
   ARCHIVE_OUTPUT_DIRECTORY ${3P_LIBRARY_OUTPUT_PATH}
+  RUNTIME_OUTPUT_DIRECTORY ${EXECUTABLE_OUTPUT_PATH}
   )
 
 if(ENABLE_SOLUTION_FOLDERS)
diff --git a/3rdparty/zlib/ChangeLog b/3rdparty/zlib/ChangeLog
index c2c643a1a9..f22aabaef5 100644
--- a/3rdparty/zlib/ChangeLog
+++ b/3rdparty/zlib/ChangeLog
@@ -1,6 +1,69 @@
 
                 ChangeLog file for zlib
 
+Changes in 1.2.8 (28 Apr 2013)
+- Update contrib/minizip/iowin32.c for Windows RT [Vollant]
+- Do not force Z_CONST for C++
+- Clean up contrib/vstudio [Ro�]
+- Correct spelling error in zlib.h
+- Fix mixed line endings in contrib/vstudio
+
+Changes in 1.2.7.3 (13 Apr 2013)
+- Fix version numbers and DLL names in contrib/vstudio/*/zlib.rc
+
+Changes in 1.2.7.2 (13 Apr 2013)
+- Change check for a four-byte type back to hexadecimal
+- Fix typo in win32/Makefile.msc
+- Add casts in gzwrite.c for pointer differences
+
+Changes in 1.2.7.1 (24 Mar 2013)
+- Replace use of unsafe string functions with snprintf if available
+- Avoid including stddef.h on Windows for Z_SOLO compile [Niessink]
+- Fix gzgetc undefine when Z_PREFIX set [Turk]
+- Eliminate use of mktemp in Makefile (not always available)
+- Fix bug in 'F' mode for gzopen()
+- Add inflateGetDictionary() function
+- Correct comment in deflate.h
+- Use _snprintf for snprintf in Microsoft C
+- On Darwin, only use /usr/bin/libtool if libtool is not Apple
+- Delete "--version" file if created by "ar --version" [Richard G.]
+- Fix configure check for veracity of compiler error return codes
+- Fix CMake compilation of static lib for MSVC2010 x64
+- Remove unused variable in infback9.c
+- Fix argument checks in gzlog_compress() and gzlog_write()
+- Clean up the usage of z_const and respect const usage within zlib
+- Clean up examples/gzlog.[ch] comparisons of different types
+- Avoid shift equal to bits in type (caused endless loop)
+- Fix unintialized value bug in gzputc() introduced by const patches
+- Fix memory allocation error in examples/zran.c [Nor]
+- Fix bug where gzopen(), gzclose() would write an empty file
+- Fix bug in gzclose() when gzwrite() runs out of memory
+- Check for input buffer malloc failure in examples/gzappend.c
+- Add note to contrib/blast to use binary mode in stdio
+- Fix comparisons of differently signed integers in contrib/blast
+- Check for invalid code length codes in contrib/puff
+- Fix serious but very rare decompression bug in inftrees.c
+- Update inflateBack() comments, since inflate() can be faster
+- Use underscored I/O function names for WINAPI_FAMILY
+- Add _tr_flush_bits to the external symbols prefixed by --zprefix
+- Add contrib/vstudio/vc10 pre-build step for static only
+- Quote --version-script argument in CMakeLists.txt
+- Don't specify --version-script on Apple platforms in CMakeLists.txt
+- Fix casting error in contrib/testzlib/testzlib.c
+- Fix types in contrib/minizip to match result of get_crc_table()
+- Simplify contrib/vstudio/vc10 with 'd' suffix
+- Add TOP support to win32/Makefile.msc
+- Suport i686 and amd64 assembler builds in CMakeLists.txt
+- Fix typos in the use of _LARGEFILE64_SOURCE in zconf.h
+- Add vc11 and vc12 build files to contrib/vstudio
+- Add gzvprintf() as an undocumented function in zlib
+- Fix configure for Sun shell
+- Remove runtime check in configure for four-byte integer type
+- Add casts and consts to ease user conversion to C++
+- Add man pages for minizip and miniunzip
+- In Makefile uninstall, don't rm if preceding cd fails
+- Do not return Z_BUF_ERROR if deflateParam() has nothing to write
+
 Changes in 1.2.7 (2 May 2012)
 - Replace use of memmove() with a simple copy for portability
 - Test for existence of strerror
diff --git a/3rdparty/zlib/README b/3rdparty/zlib/README
index 6f1255ffe4..5ca9d127ed 100644
--- a/3rdparty/zlib/README
+++ b/3rdparty/zlib/README
@@ -1,6 +1,6 @@
 ZLIB DATA COMPRESSION LIBRARY
 
-zlib 1.2.7 is a general purpose data compression library.  All the code is
+zlib 1.2.8 is a general purpose data compression library.  All the code is
 thread safe.  The data format used by the zlib library is described by RFCs
 (Request for Comments) 1950 to 1952 in the files
 http://tools.ietf.org/html/rfc1950 (zlib format), rfc1951 (deflate format) and
@@ -31,7 +31,7 @@ Mark Nelson <markn@ieee.org> wrote an article about zlib for the Jan.  1997
 issue of Dr.  Dobb's Journal; a copy of the article is available at
 http://marknelson.us/1997/01/01/zlib-engine/ .
 
-The changes made in version 1.2.7 are documented in the file ChangeLog.
+The changes made in version 1.2.8 are documented in the file ChangeLog.
 
 Unsupported third party contributions are provided in directory contrib/ .
 
@@ -84,7 +84,7 @@ Acknowledgments:
 
 Copyright notice:
 
- (C) 1995-2012 Jean-loup Gailly and Mark Adler
+ (C) 1995-2013 Jean-loup Gailly and Mark Adler
 
   This software is provided 'as-is', without any express or implied
   warranty.  In no event will the authors be held liable for any damages
diff --git a/3rdparty/zlib/compress.c b/3rdparty/zlib/compress.c
index ea4dfbe9d7..6e9762676a 100644
--- a/3rdparty/zlib/compress.c
+++ b/3rdparty/zlib/compress.c
@@ -29,7 +29,7 @@ int ZEXPORT compress2 (dest, destLen, source, sourceLen, level)
     z_stream stream;
     int err;
 
-    stream.next_in = (Bytef*)source;
+    stream.next_in = (z_const Bytef *)source;
     stream.avail_in = (uInt)sourceLen;
 #ifdef MAXSEG_64K
     /* Check for source > 64K on 16-bit machine: */
diff --git a/3rdparty/zlib/deflate.c b/3rdparty/zlib/deflate.c
index 9e4c2cbc8a..696957705b 100644
--- a/3rdparty/zlib/deflate.c
+++ b/3rdparty/zlib/deflate.c
@@ -1,5 +1,5 @@
 /* deflate.c -- compress data using the deflation algorithm
- * Copyright (C) 1995-2012 Jean-loup Gailly and Mark Adler
+ * Copyright (C) 1995-2013 Jean-loup Gailly and Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -52,7 +52,7 @@
 #include "deflate.h"
 
 const char deflate_copyright[] =
-   " deflate 1.2.7 Copyright 1995-2012 Jean-loup Gailly and Mark Adler ";
+   " deflate 1.2.8 Copyright 1995-2013 Jean-loup Gailly and Mark Adler ";
 /*
   If you use the zlib library in a product, an acknowledgment is welcome
   in the documentation of your product. If for some reason you cannot
@@ -305,7 +305,7 @@ int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
         s->pending_buf == Z_NULL) {
         s->status = FINISH_STATE;
-        strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
+        strm->msg = ERR_MSG(Z_MEM_ERROR);
         deflateEnd (strm);
         return Z_MEM_ERROR;
     }
@@ -329,7 +329,7 @@ int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
     uInt str, n;
     int wrap;
     unsigned avail;
-    unsigned char *next;
+    z_const unsigned char *next;
 
     if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL)
         return Z_STREAM_ERROR;
@@ -359,7 +359,7 @@ int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
     avail = strm->avail_in;
     next = strm->next_in;
     strm->avail_in = dictLength;
-    strm->next_in = (Bytef *)dictionary;
+    strm->next_in = (z_const Bytef *)dictionary;
     fill_window(s);
     while (s->lookahead >= MIN_MATCH) {
         str = s->strstart;
@@ -513,6 +513,8 @@ int ZEXPORT deflateParams(strm, level, strategy)
         strm->total_in != 0) {
         /* Flush the last buffer: */
         err = deflate(strm, Z_BLOCK);
+        if (err == Z_BUF_ERROR && s->pending == 0)
+            err = Z_OK;
     }
     if (s->level != level) {
         s->level = level;
diff --git a/3rdparty/zlib/deflate.h b/3rdparty/zlib/deflate.h
index fbac44d908..ce0299edd1 100644
--- a/3rdparty/zlib/deflate.h
+++ b/3rdparty/zlib/deflate.h
@@ -104,7 +104,7 @@ typedef struct internal_state {
     int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
     gz_headerp  gzhead;  /* gzip header information to write */
     uInt   gzindex;      /* where in extra, name, or comment */
-    Byte  method;        /* STORED (for zip only) or DEFLATED */
+    Byte  method;        /* can only be DEFLATED */
     int   last_flush;    /* value of flush param for previous deflate call */
 
                 /* used by deflate.c: */
diff --git a/3rdparty/zlib/gzguts.h b/3rdparty/zlib/gzguts.h
index ee3f281aa5..d87659d031 100644
--- a/3rdparty/zlib/gzguts.h
+++ b/3rdparty/zlib/gzguts.h
@@ -1,5 +1,5 @@
 /* gzguts.h -- zlib internal header definitions for gz* operations
- * Copyright (C) 2004, 2005, 2010, 2011, 2012 Mark Adler
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -35,6 +35,13 @@
 #  include <io.h>
 #endif
 
+#ifdef WINAPI_FAMILY
+#  define open _open
+#  define read _read
+#  define write _write
+#  define close _close
+#endif
+
 #ifdef NO_DEFLATE       /* for compatibility with old definition */
 #  define NO_GZCOMPRESS
 #endif
@@ -60,7 +67,7 @@
 #ifndef HAVE_VSNPRINTF
 #  ifdef MSDOS
 /* vsnprintf may exist on some MS-DOS compilers (DJGPP?),
- but for now we just assume it doesn't. */
+   but for now we just assume it doesn't. */
 #    define NO_vsnprintf
 #  endif
 #  ifdef __TURBOC__
@@ -88,6 +95,14 @@
 #  endif
 #endif
 
+/* unlike snprintf (which is required in C99, yet still not supported by
+   Microsoft more than a decade later!), _snprintf does not guarantee null
+   termination of the result -- however this is only used in gzlib.c where
+   the result is assured to fit in the space provided */
+#ifdef _MSC_VER
+#  define snprintf _snprintf
+#endif
+
 #ifndef local
 #  define local static
 #endif
@@ -127,7 +142,8 @@
 #  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
 #endif
 
-/* default i/o buffer size -- double this for output when reading */
+/* default i/o buffer size -- double this for output when reading (this and
+   twice this must be able to fit in an unsigned type) */
 #define GZBUFSIZE 8192
 
 /* gzip modes, also provide a little integrity check on the passed structure */
diff --git a/3rdparty/zlib/gzlib.c b/3rdparty/zlib/gzlib.c
index ca55c6ea92..fae202ef89 100644
--- a/3rdparty/zlib/gzlib.c
+++ b/3rdparty/zlib/gzlib.c
@@ -1,5 +1,5 @@
 /* gzlib.c -- zlib functions common to reading and writing gzip files
- * Copyright (C) 2004, 2010, 2011, 2012 Mark Adler
+ * Copyright (C) 2004, 2010, 2011, 2012, 2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -108,7 +108,7 @@ local gzFile gz_open(path, fd, mode)
         return NULL;
 
     /* allocate gzFile structure to return */
-    state = malloc(sizeof(gz_state));
+    state = (gz_statep)malloc(sizeof(gz_state));
     if (state == NULL)
         return NULL;
     state->size = 0;            /* no buffers allocated yet */
@@ -162,8 +162,10 @@ local gzFile gz_open(path, fd, mode)
                 break;
             case 'F':
                 state->strategy = Z_FIXED;
+                break;
             case 'T':
                 state->direct = 1;
+                break;
             default:        /* could consider as an error, but just ignore */
                 ;
             }
@@ -194,8 +196,8 @@ local gzFile gz_open(path, fd, mode)
     }
     else
 #endif
-        len = strlen(path);
-    state->path = malloc(len + 1);
+        len = strlen((const char *)path);
+    state->path = (char *)malloc(len + 1);
     if (state->path == NULL) {
         free(state);
         return NULL;
@@ -208,7 +210,11 @@ local gzFile gz_open(path, fd, mode)
             *(state->path) = 0;
     else
 #endif
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+        snprintf(state->path, len + 1, "%s", (const char *)path);
+#else
         strcpy(state->path, path);
+#endif
 
     /* compute the flags for open() */
     oflag =
@@ -236,7 +242,7 @@ local gzFile gz_open(path, fd, mode)
 #ifdef _WIN32
         fd == -2 ? _wopen(path, oflag, 0666) :
 #endif
-        open(path, oflag, 0666));
+        open((const char *)path, oflag, 0666));
     if (state->fd == -1) {
         free(state->path);
         free(state);
@@ -282,9 +288,13 @@ gzFile ZEXPORT gzdopen(fd, mode)
     char *path;         /* identifier for error messages */
     gzFile gz;
 
-    if (fd == -1 || (path = malloc(7 + 3 * sizeof(int))) == NULL)
+    if (fd == -1 || (path = (char *)malloc(7 + 3 * sizeof(int))) == NULL)
         return NULL;
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    snprintf(path, 7 + 3 * sizeof(int), "<fd:%d>", fd); /* for debugging */
+#else
     sprintf(path, "<fd:%d>", fd);   /* for debugging */
+#endif
     gz = gz_open(path, fd, mode);
     free(path);
     return gz;
@@ -531,7 +541,8 @@ const char * ZEXPORT gzerror(file, errnum)
     /* return error information */
     if (errnum != NULL)
         *errnum = state->err;
-    return state->msg == NULL ? "" : state->msg;
+    return state->err == Z_MEM_ERROR ? "out of memory" :
+                                       (state->msg == NULL ? "" : state->msg);
 }
 
 /* -- see zlib.h -- */
@@ -582,21 +593,24 @@ void ZLIB_INTERNAL gz_error(state, err, msg)
     if (msg == NULL)
         return;
 
-    /* for an out of memory error, save as static string */
-    if (err == Z_MEM_ERROR) {
-        state->msg = (char *)msg;
+    /* for an out of memory error, return literal string when requested */
+    if (err == Z_MEM_ERROR)
         return;
-    }
 
     /* construct error message with path */
-    if ((state->msg = malloc(strlen(state->path) + strlen(msg) + 3)) == NULL) {
+    if ((state->msg = (char *)malloc(strlen(state->path) + strlen(msg) + 3)) ==
+            NULL) {
         state->err = Z_MEM_ERROR;
-        state->msg = (char *)"out of memory";
         return;
     }
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    snprintf(state->msg, strlen(state->path) + strlen(msg) + 3,
+             "%s%s%s", state->path, ": ", msg);
+#else
     strcpy(state->msg, state->path);
     strcat(state->msg, ": ");
     strcat(state->msg, msg);
+#endif
     return;
 }
 
diff --git a/3rdparty/zlib/gzread.c b/3rdparty/zlib/gzread.c
index 3493d34d4e..bf4538eb27 100644
--- a/3rdparty/zlib/gzread.c
+++ b/3rdparty/zlib/gzread.c
@@ -1,5 +1,5 @@
 /* gzread.c -- zlib functions for reading gzip files
- * Copyright (C) 2004, 2005, 2010, 2011, 2012 Mark Adler
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -58,7 +58,8 @@ local int gz_avail(state)
         return -1;
     if (state->eof == 0) {
         if (strm->avail_in) {       /* copy what's there to the start */
-            unsigned char *p = state->in, *q = strm->next_in;
+            unsigned char *p = state->in;
+            unsigned const char *q = strm->next_in;
             unsigned n = strm->avail_in;
             do {
                 *p++ = *q++;
@@ -90,8 +91,8 @@ local int gz_look(state)
     /* allocate read buffers and inflate memory */
     if (state->size == 0) {
         /* allocate buffers */
-        state->in = malloc(state->want);
-        state->out = malloc(state->want << 1);
+        state->in = (unsigned char *)malloc(state->want);
+        state->out = (unsigned char *)malloc(state->want << 1);
         if (state->in == NULL || state->out == NULL) {
             if (state->out != NULL)
                 free(state->out);
@@ -352,14 +353,14 @@ int ZEXPORT gzread(file, buf, len)
 
         /* large len -- read directly into user buffer */
         else if (state->how == COPY) {      /* read directly */
-            if (gz_load(state, buf, len, &n) == -1)
+            if (gz_load(state, (unsigned char *)buf, len, &n) == -1)
                 return -1;
         }
 
         /* large len -- decompress directly into user buffer */
         else {  /* state->how == GZIP */
             strm->avail_out = len;
-            strm->next_out = buf;
+            strm->next_out = (unsigned char *)buf;
             if (gz_decomp(state) == -1)
                 return -1;
             n = state->x.have;
@@ -378,7 +379,11 @@ int ZEXPORT gzread(file, buf, len)
 }
 
 /* -- see zlib.h -- */
-#undef gzgetc
+#ifdef Z_PREFIX_SET
+#  undef z_gzgetc
+#else
+#  undef gzgetc
+#endif
 int ZEXPORT gzgetc(file)
     gzFile file;
 {
@@ -518,7 +523,7 @@ char * ZEXPORT gzgets(file, buf, len)
 
         /* look for end-of-line in current output buffer */
         n = state->x.have > left ? left : state->x.have;
-        eol = memchr(state->x.next, '\n', n);
+        eol = (unsigned char *)memchr(state->x.next, '\n', n);
         if (eol != NULL)
             n = (unsigned)(eol - state->x.next) + 1;
 
diff --git a/3rdparty/zlib/gzwrite.c b/3rdparty/zlib/gzwrite.c
index 27cb3428e3..aa767fbf63 100644
--- a/3rdparty/zlib/gzwrite.c
+++ b/3rdparty/zlib/gzwrite.c
@@ -1,5 +1,5 @@
 /* gzwrite.c -- zlib functions for writing gzip files
- * Copyright (C) 2004, 2005, 2010, 2011, 2012 Mark Adler
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -19,7 +19,7 @@ local int gz_init(state)
     z_streamp strm = &(state->strm);
 
     /* allocate input buffer */
-    state->in = malloc(state->want);
+    state->in = (unsigned char *)malloc(state->want);
     if (state->in == NULL) {
         gz_error(state, Z_MEM_ERROR, "out of memory");
         return -1;
@@ -28,7 +28,7 @@ local int gz_init(state)
     /* only need output buffer and deflate state if compressing */
     if (!state->direct) {
         /* allocate output buffer */
-        state->out = malloc(state->want);
+        state->out = (unsigned char *)malloc(state->want);
         if (state->out == NULL) {
             free(state->in);
             gz_error(state, Z_MEM_ERROR, "out of memory");
@@ -168,7 +168,6 @@ int ZEXPORT gzwrite(file, buf, len)
     unsigned len;
 {
     unsigned put = len;
-    unsigned n;
     gz_statep state;
     z_streamp strm;
 
@@ -208,16 +207,19 @@ int ZEXPORT gzwrite(file, buf, len)
     if (len < state->size) {
         /* copy to input buffer, compress when full */
         do {
+            unsigned have, copy;
+
             if (strm->avail_in == 0)
                 strm->next_in = state->in;
-            n = state->size - strm->avail_in;
-            if (n > len)
-                n = len;
-            memcpy(strm->next_in + strm->avail_in, buf, n);
-            strm->avail_in += n;
-            state->x.pos += n;
-            buf = (char *)buf + n;
-            len -= n;
+            have = (unsigned)((strm->next_in + strm->avail_in) - state->in);
+            copy = state->size - have;
+            if (copy > len)
+                copy = len;
+            memcpy(state->in + have, buf, copy);
+            strm->avail_in += copy;
+            state->x.pos += copy;
+            buf = (const char *)buf + copy;
+            len -= copy;
             if (len && gz_comp(state, Z_NO_FLUSH) == -1)
                 return 0;
         } while (len);
@@ -229,7 +231,7 @@ int ZEXPORT gzwrite(file, buf, len)
 
         /* directly compress user buffer to file */
         strm->avail_in = len;
-        strm->next_in = (voidp)buf;
+        strm->next_in = (z_const Bytef *)buf;
         state->x.pos += len;
         if (gz_comp(state, Z_NO_FLUSH) == -1)
             return 0;
@@ -244,6 +246,7 @@ int ZEXPORT gzputc(file, c)
     gzFile file;
     int c;
 {
+    unsigned have;
     unsigned char buf[1];
     gz_statep state;
     z_streamp strm;
@@ -267,12 +270,16 @@ int ZEXPORT gzputc(file, c)
 
     /* try writing to input buffer for speed (state->size == 0 if buffer not
        initialized) */
-    if (strm->avail_in < state->size) {
+    if (state->size) {
         if (strm->avail_in == 0)
             strm->next_in = state->in;
-        strm->next_in[strm->avail_in++] = c;
-        state->x.pos++;
-        return c & 0xff;
+        have = (unsigned)((strm->next_in + strm->avail_in) - state->in);
+        if (have < state->size) {
+            state->in[have] = c;
+            strm->avail_in++;
+            state->x.pos++;
+            return c & 0xff;
+        }
     }
 
     /* no room in buffer or not initialized, use gz_write() */
@@ -300,12 +307,11 @@ int ZEXPORT gzputs(file, str)
 #include <stdarg.h>
 
 /* -- see zlib.h -- */
-int ZEXPORTVA gzprintf (gzFile file, const char *format, ...)
+int ZEXPORTVA gzvprintf(gzFile file, const char *format, va_list va)
 {
     int size, len;
     gz_statep state;
     z_streamp strm;
-    va_list va;
 
     /* get internal structure */
     if (file == NULL)
@@ -335,25 +341,20 @@ int ZEXPORTVA gzprintf (gzFile file, const char *format, ...)
     /* do the printf() into the input buffer, put length in len */
     size = (int)(state->size);
     state->in[size - 1] = 0;
-    va_start(va, format);
 #ifdef NO_vsnprintf
 #  ifdef HAS_vsprintf_void
     (void)vsprintf((char *)(state->in), format, va);
-    va_end(va);
     for (len = 0; len < size; len++)
         if (state->in[len] == 0) break;
 #  else
     len = vsprintf((char *)(state->in), format, va);
-    va_end(va);
 #  endif
 #else
 #  ifdef HAS_vsnprintf_void
     (void)vsnprintf((char *)(state->in), size, format, va);
-    va_end(va);
     len = strlen((char *)(state->in));
 #  else
     len = vsnprintf((char *)(state->in), size, format, va);
-    va_end(va);
 #  endif
 #endif
 
@@ -368,6 +369,17 @@ int ZEXPORTVA gzprintf (gzFile file, const char *format, ...)
     return len;
 }
 
+int ZEXPORTVA gzprintf(gzFile file, const char *format, ...)
+{
+    va_list va;
+    int ret;
+
+    va_start(va, format);
+    ret = gzvprintf(file, format, va);
+    va_end(va);
+    return ret;
+}
+
 #else /* !STDC && !Z_HAVE_STDARG_H */
 
 /* -- see zlib.h -- */
@@ -547,9 +559,9 @@ int ZEXPORT gzclose_w(file)
     }
 
     /* flush, free memory, and close file */
+    if (gz_comp(state, Z_FINISH) == -1)
+        ret = state->err;
     if (state->size) {
-        if (gz_comp(state, Z_FINISH) == -1)
-            ret = state->err;
         if (!state->direct) {
             (void)deflateEnd(&(state->strm));
             free(state->out);
diff --git a/3rdparty/zlib/infback.c b/3rdparty/zlib/infback.c
index 981aff17c2..f3833c2e43 100644
--- a/3rdparty/zlib/infback.c
+++ b/3rdparty/zlib/infback.c
@@ -255,7 +255,7 @@ out_func out;
 void FAR *out_desc;
 {
     struct inflate_state FAR *state;
-    unsigned char FAR *next;    /* next input */
+    z_const unsigned char FAR *next;    /* next input */
     unsigned char FAR *put;     /* next output */
     unsigned have, left;        /* available input and output */
     unsigned long hold;         /* bit buffer */
diff --git a/3rdparty/zlib/inffast.c b/3rdparty/zlib/inffast.c
index 2f1d60b43b..bda59ceb6a 100644
--- a/3rdparty/zlib/inffast.c
+++ b/3rdparty/zlib/inffast.c
@@ -1,5 +1,5 @@
 /* inffast.c -- fast decoding
- * Copyright (C) 1995-2008, 2010 Mark Adler
+ * Copyright (C) 1995-2008, 2010, 2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -69,8 +69,8 @@ z_streamp strm;
 unsigned start;         /* inflate()'s starting value for strm->avail_out */
 {
     struct inflate_state FAR *state;
-    unsigned char FAR *in;      /* local strm->next_in */
-    unsigned char FAR *last;    /* while in < last, enough input available */
+    z_const unsigned char FAR *in;      /* local strm->next_in */
+    z_const unsigned char FAR *last;    /* have enough input while in < last */
     unsigned char FAR *out;     /* local strm->next_out */
     unsigned char FAR *beg;     /* inflate()'s initial strm->next_out */
     unsigned char FAR *end;     /* while out < end, enough space available */
diff --git a/3rdparty/zlib/inflate.c b/3rdparty/zlib/inflate.c
index 47418a1e1e..870f89bb4d 100644
--- a/3rdparty/zlib/inflate.c
+++ b/3rdparty/zlib/inflate.c
@@ -93,11 +93,12 @@
 
 /* function prototypes */
 local void fixedtables OF((struct inflate_state FAR *state));
-local int updatewindow OF((z_streamp strm, unsigned out));
+local int updatewindow OF((z_streamp strm, const unsigned char FAR *end,
+                           unsigned copy));
 #ifdef BUILDFIXED
    void makefixed OF((void));
 #endif
-local unsigned syncsearch OF((unsigned FAR *have, unsigned char FAR *buf,
+local unsigned syncsearch OF((unsigned FAR *have, const unsigned char FAR *buf,
                               unsigned len));
 
 int ZEXPORT inflateResetKeep(strm)
@@ -375,12 +376,13 @@ void makefixed()
    output will fall in the output data, making match copies simpler and faster.
    The advantage may be dependent on the size of the processor's data caches.
  */
-local int updatewindow(strm, out)
+local int updatewindow(strm, end, copy)
 z_streamp strm;
-unsigned out;
+const Bytef *end;
+unsigned copy;
 {
     struct inflate_state FAR *state;
-    unsigned copy, dist;
+    unsigned dist;
 
     state = (struct inflate_state FAR *)strm->state;
 
@@ -400,19 +402,18 @@ unsigned out;
     }
 
     /* copy state->wsize or less output bytes into the circular window */
-    copy = out - strm->avail_out;
     if (copy >= state->wsize) {
-        zmemcpy(state->window, strm->next_out - state->wsize, state->wsize);
+        zmemcpy(state->window, end - state->wsize, state->wsize);
         state->wnext = 0;
         state->whave = state->wsize;
     }
     else {
         dist = state->wsize - state->wnext;
         if (dist > copy) dist = copy;
-        zmemcpy(state->window + state->wnext, strm->next_out - copy, dist);
+        zmemcpy(state->window + state->wnext, end - copy, dist);
         copy -= dist;
         if (copy) {
-            zmemcpy(state->window, strm->next_out - copy, copy);
+            zmemcpy(state->window, end - copy, copy);
             state->wnext = copy;
             state->whave = state->wsize;
         }
@@ -606,7 +607,7 @@ z_streamp strm;
 int flush;
 {
     struct inflate_state FAR *state;
-    unsigned char FAR *next;    /* next input */
+    z_const unsigned char FAR *next;    /* next input */
     unsigned char FAR *put;     /* next output */
     unsigned have, left;        /* available input and output */
     unsigned long hold;         /* bit buffer */
@@ -920,7 +921,7 @@ int flush;
             while (state->have < 19)
                 state->lens[order[state->have++]] = 0;
             state->next = state->codes;
-            state->lencode = (code const FAR *)(state->next);
+            state->lencode = (const code FAR *)(state->next);
             state->lenbits = 7;
             ret = inflate_table(CODES, state->lens, 19, &(state->next),
                                 &(state->lenbits), state->work);
@@ -994,7 +995,7 @@ int flush;
                values here (9 and 6) without reading the comments in inftrees.h
                concerning the ENOUGH constants, which depend on those values */
             state->next = state->codes;
-            state->lencode = (code const FAR *)(state->next);
+            state->lencode = (const code FAR *)(state->next);
             state->lenbits = 9;
             ret = inflate_table(LENS, state->lens, state->nlen, &(state->next),
                                 &(state->lenbits), state->work);
@@ -1003,7 +1004,7 @@ int flush;
                 state->mode = BAD;
                 break;
             }
-            state->distcode = (code const FAR *)(state->next);
+            state->distcode = (const code FAR *)(state->next);
             state->distbits = 6;
             ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist,
                             &(state->next), &(state->distbits), state->work);
@@ -1230,7 +1231,7 @@ int flush;
     RESTORE();
     if (state->wsize || (out != strm->avail_out && state->mode < BAD &&
             (state->mode < CHECK || flush != Z_FINISH)))
-        if (updatewindow(strm, out)) {
+        if (updatewindow(strm, strm->next_out, out - strm->avail_out)) {
             state->mode = MEM;
             return Z_MEM_ERROR;
         }
@@ -1264,6 +1265,29 @@ z_streamp strm;
     return Z_OK;
 }
 
+int ZEXPORT inflateGetDictionary(strm, dictionary, dictLength)
+z_streamp strm;
+Bytef *dictionary;
+uInt *dictLength;
+{
+    struct inflate_state FAR *state;
+
+    /* check state */
+    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* copy dictionary */
+    if (state->whave && dictionary != Z_NULL) {
+        zmemcpy(dictionary, state->window + state->wnext,
+                state->whave - state->wnext);
+        zmemcpy(dictionary + state->whave - state->wnext,
+                state->window, state->wnext);
+    }
+    if (dictLength != Z_NULL)
+        *dictLength = state->whave;
+    return Z_OK;
+}
+
 int ZEXPORT inflateSetDictionary(strm, dictionary, dictLength)
 z_streamp strm;
 const Bytef *dictionary;
@@ -1271,8 +1295,6 @@ uInt dictLength;
 {
     struct inflate_state FAR *state;
     unsigned long dictid;
-    unsigned char *next;
-    unsigned avail;
     int ret;
 
     /* check state */
@@ -1291,13 +1313,7 @@ uInt dictLength;
 
     /* copy dictionary to window using updatewindow(), which will amend the
        existing dictionary if appropriate */
-    next = strm->next_out;
-    avail = strm->avail_out;
-    strm->next_out = (Bytef *)dictionary + dictLength;
-    strm->avail_out = 0;
-    ret = updatewindow(strm, dictLength);
-    strm->avail_out = avail;
-    strm->next_out = next;
+    ret = updatewindow(strm, dictionary + dictLength, dictLength);
     if (ret) {
         state->mode = MEM;
         return Z_MEM_ERROR;
@@ -1337,7 +1353,7 @@ gz_headerp head;
  */
 local unsigned syncsearch(have, buf, len)
 unsigned FAR *have;
-unsigned char FAR *buf;
+const unsigned char FAR *buf;
 unsigned len;
 {
     unsigned got;
diff --git a/3rdparty/zlib/inftrees.c b/3rdparty/zlib/inftrees.c
index abcd7c45ed..44d89cf24e 100644
--- a/3rdparty/zlib/inftrees.c
+++ b/3rdparty/zlib/inftrees.c
@@ -1,5 +1,5 @@
 /* inftrees.c -- generate Huffman trees for efficient decoding
- * Copyright (C) 1995-2012 Mark Adler
+ * Copyright (C) 1995-2013 Mark Adler
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -9,7 +9,7 @@
 #define MAXBITS 15
 
 const char inflate_copyright[] =
-   " inflate 1.2.7 Copyright 1995-2012 Mark Adler ";
+   " inflate 1.2.8 Copyright 1995-2013 Mark Adler ";
 /*
   If you use the zlib library in a product, an acknowledgment is welcome
   in the documentation of your product. If for some reason you cannot
@@ -62,7 +62,7 @@ unsigned short FAR *work;
         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
     static const unsigned short lext[31] = { /* Length codes 257..285 extra */
         16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
-        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 78, 68};
+        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 72, 78};
     static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
@@ -208,8 +208,8 @@ unsigned short FAR *work;
     mask = used - 1;            /* mask for comparing low */
 
     /* check available table space */
-    if ((type == LENS && used >= ENOUGH_LENS) ||
-        (type == DISTS && used >= ENOUGH_DISTS))
+    if ((type == LENS && used > ENOUGH_LENS) ||
+        (type == DISTS && used > ENOUGH_DISTS))
         return 1;
 
     /* process all codes and make table entries */
@@ -277,8 +277,8 @@ unsigned short FAR *work;
 
             /* check for enough space */
             used += 1U << curr;
-            if ((type == LENS && used >= ENOUGH_LENS) ||
-                (type == DISTS && used >= ENOUGH_DISTS))
+            if ((type == LENS && used > ENOUGH_LENS) ||
+                (type == DISTS && used > ENOUGH_DISTS))
                 return 1;
 
             /* point entry in root table to sub-table */
diff --git a/3rdparty/zlib/trees.c b/3rdparty/zlib/trees.c
index 8c32b214b1..1fd7759ef0 100644
--- a/3rdparty/zlib/trees.c
+++ b/3rdparty/zlib/trees.c
@@ -146,8 +146,8 @@ local void send_tree      OF((deflate_state *s, ct_data *tree, int max_code));
 local int  build_bl_tree  OF((deflate_state *s));
 local void send_all_trees OF((deflate_state *s, int lcodes, int dcodes,
                               int blcodes));
-local void compress_block OF((deflate_state *s, ct_data *ltree,
-                              ct_data *dtree));
+local void compress_block OF((deflate_state *s, const ct_data *ltree,
+                              const ct_data *dtree));
 local int  detect_data_type OF((deflate_state *s));
 local unsigned bi_reverse OF((unsigned value, int length));
 local void bi_windup      OF((deflate_state *s));
@@ -972,7 +972,8 @@ void ZLIB_INTERNAL _tr_flush_block(s, buf, stored_len, last)
     } else if (s->strategy == Z_FIXED || static_lenb == opt_lenb) {
 #endif
         send_bits(s, (STATIC_TREES<<1)+last, 3);
-        compress_block(s, (ct_data *)static_ltree, (ct_data *)static_dtree);
+        compress_block(s, (const ct_data *)static_ltree,
+                       (const ct_data *)static_dtree);
 #ifdef DEBUG
         s->compressed_len += 3 + s->static_len;
 #endif
@@ -980,7 +981,8 @@ void ZLIB_INTERNAL _tr_flush_block(s, buf, stored_len, last)
         send_bits(s, (DYN_TREES<<1)+last, 3);
         send_all_trees(s, s->l_desc.max_code+1, s->d_desc.max_code+1,
                        max_blindex+1);
-        compress_block(s, (ct_data *)s->dyn_ltree, (ct_data *)s->dyn_dtree);
+        compress_block(s, (const ct_data *)s->dyn_ltree,
+                       (const ct_data *)s->dyn_dtree);
 #ifdef DEBUG
         s->compressed_len += 3 + s->opt_len;
 #endif
@@ -1057,8 +1059,8 @@ int ZLIB_INTERNAL _tr_tally (s, dist, lc)
  */
 local void compress_block(s, ltree, dtree)
     deflate_state *s;
-    ct_data *ltree; /* literal tree */
-    ct_data *dtree; /* distance tree */
+    const ct_data *ltree; /* literal tree */
+    const ct_data *dtree; /* distance tree */
 {
     unsigned dist;      /* distance of matched string */
     int lc;             /* match length or unmatched char (if dist == 0) */
diff --git a/3rdparty/zlib/uncompr.c b/3rdparty/zlib/uncompr.c
index ad98be3a5d..242e9493df 100644
--- a/3rdparty/zlib/uncompr.c
+++ b/3rdparty/zlib/uncompr.c
@@ -30,7 +30,7 @@ int ZEXPORT uncompress (dest, destLen, source, sourceLen)
     z_stream stream;
     int err;
 
-    stream.next_in = (Bytef*)source;
+    stream.next_in = (z_const Bytef *)source;
     stream.avail_in = (uInt)sourceLen;
     /* Check for source > 64K on 16-bit machine: */
     if ((uLong)stream.avail_in != sourceLen) return Z_BUF_ERROR;
diff --git a/3rdparty/zlib/zconf.h.cmakein b/3rdparty/zlib/zconf.h.cmakein
index 1e4e14045f..6528fb4df8 100644
--- a/3rdparty/zlib/zconf.h.cmakein
+++ b/3rdparty/zlib/zconf.h.cmakein
@@ -1,5 +1,5 @@
 /* zconf.h -- configuration of the zlib compression library
- * Copyright (C) 1995-2012 Jean-loup Gailly.
+ * Copyright (C) 1995-2013 Jean-loup Gailly.
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -23,6 +23,7 @@
 #  define _dist_code            z__dist_code
 #  define _length_code          z__length_code
 #  define _tr_align             z__tr_align
+#  define _tr_flush_bits        z__tr_flush_bits
 #  define _tr_flush_block       z__tr_flush_block
 #  define _tr_init              z__tr_init
 #  define _tr_stored_block      z__tr_stored_block
@@ -79,6 +80,7 @@
 #      define gzopen_w              z_gzopen_w
 #    endif
 #    define gzprintf              z_gzprintf
+#    define gzvprintf             z_gzvprintf
 #    define gzputc                z_gzputc
 #    define gzputs                z_gzputs
 #    define gzread                z_gzread
@@ -105,6 +107,7 @@
 #  define inflateReset          z_inflateReset
 #  define inflateReset2         z_inflateReset2
 #  define inflateSetDictionary  z_inflateSetDictionary
+#  define inflateGetDictionary  z_inflateGetDictionary
 #  define inflateSync           z_inflateSync
 #  define inflateSyncPoint      z_inflateSyncPoint
 #  define inflateUndermine      z_inflateUndermine
@@ -390,20 +393,14 @@ typedef uLong FAR uLongf;
    typedef Byte       *voidp;
 #endif
 
-/* ./configure may #define Z_U4 here */
-
 #if !defined(Z_U4) && !defined(Z_SOLO) && defined(STDC)
 #  include <limits.h>
 #  if (UINT_MAX == 0xffffffffUL)
 #    define Z_U4 unsigned
-#  else
-#    if (ULONG_MAX == 0xffffffffUL)
-#      define Z_U4 unsigned long
-#    else
-#      if (USHRT_MAX == 0xffffffffUL)
-#        define Z_U4 unsigned short
-#      endif
-#    endif
+#  elif (ULONG_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned long
+#  elif (USHRT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned short
 #  endif
 #endif
 
@@ -427,8 +424,16 @@ typedef uLong FAR uLongf;
 #  endif
 #endif
 
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+#    include <stdarg.h>         /* for va_list */
+#  endif
+#endif
+
 #ifdef _WIN32
-#  include <stddef.h>           /* for wchar_t */
+#  ifndef Z_SOLO
+#    include <stddef.h>         /* for wchar_t */
+#  endif
 #endif
 
 /* a little trick to accommodate both "#define _LARGEFILE64_SOURCE" and
@@ -437,7 +442,7 @@ typedef uLong FAR uLongf;
  * both "#undef _LARGEFILE64_SOURCE" and "#define _LARGEFILE64_SOURCE 0" as
  * equivalently requesting no 64-bit operations
  */
-#if defined(LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1
+#if defined(_LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1
 #  undef _LARGEFILE64_SOURCE
 #endif
 
@@ -445,7 +450,7 @@ typedef uLong FAR uLongf;
 #  define Z_HAVE_UNISTD_H
 #endif
 #ifndef Z_SOLO
-#  if defined(Z_HAVE_UNISTD_H) || defined(LARGEFILE64_SOURCE)
+#  if defined(Z_HAVE_UNISTD_H) || defined(_LARGEFILE64_SOURCE)
 #    include <unistd.h>         /* for SEEK_*, off_t, and _LFS64_LARGEFILE */
 #    ifdef VMS
 #      include <unixio.h>       /* for off_t */
diff --git a/3rdparty/zlib/zlib.h b/3rdparty/zlib/zlib.h
index 3edf3acdb5..3e0c7672ac 100644
--- a/3rdparty/zlib/zlib.h
+++ b/3rdparty/zlib/zlib.h
@@ -1,7 +1,7 @@
 /* zlib.h -- interface of the 'zlib' general purpose compression library
-  version 1.2.7, May 2nd, 2012
+  version 1.2.8, April 28th, 2013
 
-  Copyright (C) 1995-2012 Jean-loup Gailly and Mark Adler
+  Copyright (C) 1995-2013 Jean-loup Gailly and Mark Adler
 
   This software is provided 'as-is', without any express or implied
   warranty.  In no event will the authors be held liable for any damages
@@ -37,11 +37,11 @@
 extern "C" {
 #endif
 
-#define ZLIB_VERSION "1.2.7"
-#define ZLIB_VERNUM 0x1270
+#define ZLIB_VERSION "1.2.8"
+#define ZLIB_VERNUM 0x1280
 #define ZLIB_VER_MAJOR 1
 #define ZLIB_VER_MINOR 2
-#define ZLIB_VER_REVISION 7
+#define ZLIB_VER_REVISION 8
 #define ZLIB_VER_SUBREVISION 0
 
 /*
@@ -839,6 +839,21 @@ ZEXTERN int ZEXPORT inflateSetDictionary OF((z_streamp strm,
    inflate().
 */
 
+ZEXTERN int ZEXPORT inflateGetDictionary OF((z_streamp strm,
+                                             Bytef *dictionary,
+                                             uInt  *dictLength));
+/*
+     Returns the sliding dictionary being maintained by inflate.  dictLength is
+   set to the number of bytes in the dictionary, and that many bytes are copied
+   to dictionary.  dictionary must have enough space, where 32768 bytes is
+   always enough.  If inflateGetDictionary() is called with dictionary equal to
+   Z_NULL, then only the dictionary length is returned, and nothing is copied.
+   Similary, if dictLength is Z_NULL, then it is not set.
+
+     inflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the
+   stream state is inconsistent.
+*/
+
 ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm));
 /*
      Skips invalid compressed data until a possible full flush point (see above
@@ -846,7 +861,7 @@ ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm));
    available input is skipped.  No output is provided.
 
      inflateSync searches for a 00 00 FF FF pattern in the compressed data.
-   All full flush points have this pattern, but not all occurences of this
+   All full flush points have this pattern, but not all occurrences of this
    pattern are full flush points.
 
      inflateSync returns Z_OK if a possible full flush point has been found,
@@ -1007,7 +1022,8 @@ ZEXTERN int ZEXPORT inflateBackInit OF((z_streamp strm, int windowBits,
    the version of the header file.
 */
 
-typedef unsigned (*in_func) OF((void FAR *, unsigned char FAR * FAR *));
+typedef unsigned (*in_func) OF((void FAR *,
+                                z_const unsigned char FAR * FAR *));
 typedef int (*out_func) OF((void FAR *, unsigned char FAR *, unsigned));
 
 ZEXTERN int ZEXPORT inflateBack OF((z_streamp strm,
@@ -1015,11 +1031,12 @@ ZEXTERN int ZEXPORT inflateBack OF((z_streamp strm,
                                     out_func out, void FAR *out_desc));
 /*
      inflateBack() does a raw inflate with a single call using a call-back
-   interface for input and output.  This is more efficient than inflate() for
-   file i/o applications in that it avoids copying between the output and the
-   sliding window by simply making the window itself the output buffer.  This
-   function trusts the application to not change the output buffer passed by
-   the output function, at least until inflateBack() returns.
+   interface for input and output.  This is potentially more efficient than
+   inflate() for file i/o applications, in that it avoids copying between the
+   output and the sliding window by simply making the window itself the output
+   buffer.  inflate() can be faster on modern CPUs when used with large
+   buffers.  inflateBack() trusts the application to not change the output
+   buffer passed by the output function, at least until inflateBack() returns.
 
      inflateBackInit() must be called first to allocate the internal state
    and to initialize the state with the user-provided window buffer.
@@ -1736,6 +1753,13 @@ ZEXTERN int            ZEXPORT deflateResetKeep OF((z_streamp));
 ZEXTERN gzFile         ZEXPORT gzopen_w OF((const wchar_t *path,
                                             const char *mode));
 #endif
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+ZEXTERN int            ZEXPORTVA gzvprintf Z_ARG((gzFile file,
+                                                  const char *format,
+                                                  va_list va));
+#  endif
+#endif
 
 #ifdef __cplusplus
 }
diff --git a/3rdparty/zlib/zutil.c b/3rdparty/zlib/zutil.c
index 65e0d3b72b..23d2ebef00 100644
--- a/3rdparty/zlib/zutil.c
+++ b/3rdparty/zlib/zutil.c
@@ -14,7 +14,7 @@
 struct internal_state      {int dummy;}; /* for buggy compilers */
 #endif
 
-const char * const z_errmsg[10] = {
+z_const char * const z_errmsg[10] = {
 "need dictionary",     /* Z_NEED_DICT       2  */
 "stream end",          /* Z_STREAM_END      1  */
 "",                    /* Z_OK              0  */
diff --git a/3rdparty/zlib/zutil.h b/3rdparty/zlib/zutil.h
index 4e3dcc6ae9..24ab06b1cf 100644
--- a/3rdparty/zlib/zutil.h
+++ b/3rdparty/zlib/zutil.h
@@ -1,5 +1,5 @@
 /* zutil.h -- internal interface and configuration of the compression library
- * Copyright (C) 1995-2012 Jean-loup Gailly.
+ * Copyright (C) 1995-2013 Jean-loup Gailly.
  * For conditions of distribution and use, see copyright notice in zlib.h
  */
 
@@ -44,13 +44,13 @@ typedef unsigned short ush;
 typedef ush FAR ushf;
 typedef unsigned long  ulg;
 
-extern const char * const z_errmsg[10]; /* indexed by 2-zlib_error */
+extern z_const char * const z_errmsg[10]; /* indexed by 2-zlib_error */
 /* (size given to avoid silly warnings with Visual C++) */
 
 #define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)]
 
 #define ERR_RETURN(strm,err) \
-  return (strm->msg = (char*)ERR_MSG(err), (err))
+  return (strm->msg = ERR_MSG(err), (err))
 /* To be used only when the state is known to be valid */
 
         /* common constants */
@@ -168,7 +168,8 @@ extern const char * const z_errmsg[10]; /* indexed by 2-zlib_error */
 #endif
 
 /* provide prototypes for these when building zlib without LFS */
-#if !defined(_WIN32) && (!defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0)
+#if !defined(_WIN32) && \
+    (!defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0)
     ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t));
     ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t));
 #endif
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 470c5520dd..c795ad3ff3 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -4,20 +4,14 @@
 #    From the off-tree build directory, invoke:
 #      $ cmake <PATH_TO_OPENCV_ROOT>
 #
-#
-#   - OCT-2008: Initial version <joseluisblancoc@gmail.com>
-#
 # ----------------------------------------------------------------------------
 
-set(CMAKE_ALLOW_LOOSE_LOOP_CONSTRUCTS true)
+include(cmake/OpenCVMinDepVersions.cmake)
 
-# --------------------------------------------------------------
-# Indicate CMake 2.7 and above that we don't want to mix relative
-#  and absolute paths in linker lib lists.
-# Run "cmake --help-policy CMP0003" for more information.
-# --------------------------------------------------------------
-if(COMMAND cmake_policy)
-  cmake_policy(SET CMP0003 NEW)
+if(CMAKE_GENERATOR MATCHES Xcode AND XCODE_VERSION VERSION_GREATER 4.3)
+  cmake_minimum_required(VERSION 2.8.8 FATAL_ERROR)
+else()
+  cmake_minimum_required(VERSION "${MIN_VER_CMAKE}" FATAL_ERROR)
 endif()
 
 # Following block can broke build in case of cross-compilng
@@ -41,20 +35,10 @@ else(NOT CMAKE_TOOLCHAIN_FILE)
   set(CMAKE_INSTALL_PREFIX "${CMAKE_BINARY_DIR}/install" CACHE PATH "Installation Directory")
 endif(NOT CMAKE_TOOLCHAIN_FILE)
 
-# --------------------------------------------------------------
-# Top level OpenCV project
-# --------------------------------------------------------------
-if(CMAKE_GENERATOR MATCHES Xcode AND XCODE_VERSION VERSION_GREATER 4.3)
-  cmake_minimum_required(VERSION 2.8.8)
-elseif(IOS)
-  cmake_minimum_required(VERSION 2.8.0)
-else()
-  cmake_minimum_required(VERSION 2.6.3)
-endif()
 
 # must go before the project command
 set(CMAKE_CONFIGURATION_TYPES "Debug;Release" CACHE STRING "Configs" FORCE)
-if(DEFINED CMAKE_BUILD_TYPE AND CMAKE_VERSION VERSION_GREATER "2.8")
+if(DEFINED CMAKE_BUILD_TYPE)
   set_property( CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS ${CMAKE_CONFIGURATION_TYPES} )
 endif()
 
@@ -103,6 +87,19 @@ if(UNIX AND NOT ANDROID)
   endif()
 endif()
 
+# Add these standard paths to the search paths for FIND_PATH
+# to find include files from these locations first
+if(MINGW)
+  if(EXISTS /mingw)
+      list(APPEND CMAKE_INCLUDE_PATH /mingw)
+  endif()
+  if(EXISTS /mingw32)
+      list(APPEND CMAKE_INCLUDE_PATH /mingw32)
+  endif()
+  if(EXISTS /mingw64)
+      list(APPEND CMAKE_INCLUDE_PATH /mingw64)
+  endif()
+endif()
 
 # ----------------------------------------------------------------------------
 # OpenCV cmake options
@@ -110,41 +107,45 @@ endif()
 
 # Optional 3rd party components
 # ===================================================
-OCV_OPTION(WITH_1394           "Include IEEE1394 support"                    ON   IF (UNIX AND NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_1394           "Include IEEE1394 support"                    ON   IF (NOT ANDROID AND NOT IOS) )
 OCV_OPTION(WITH_AVFOUNDATION   "Use AVFoundation for Video I/O"              ON   IF IOS)
 OCV_OPTION(WITH_CARBON         "Use Carbon for UI instead of Cocoa"          OFF  IF APPLE )
-OCV_OPTION(WITH_CUDA           "Include NVidia Cuda Runtime support"         ON   IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_CUFFT          "Include NVidia Cuda Fast Fourier Transform (FFT) library support"            ON  IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_CUBLAS         "Include NVidia Cuda Basic Linear Algebra Subprograms (BLAS) library support" OFF IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_NVCUVID        "Include NVidia Video Decoding library support"                               OFF IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS AND NOT APPLE) )
+OCV_OPTION(WITH_CUDA           "Include NVidia Cuda Runtime support"                                         ON  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_CUFFT          "Include NVidia Cuda Fast Fourier Transform (FFT) library support"            ON  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_CUBLAS         "Include NVidia Cuda Basic Linear Algebra Subprograms (BLAS) library support" OFF IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_NVCUVID        "Include NVidia Video Decoding library support"                               OFF IF (NOT ANDROID AND NOT IOS AND NOT APPLE) )
 OCV_OPTION(WITH_EIGEN          "Include Eigen2/Eigen3 support"               ON)
+OCV_OPTION(WITH_VFW            "Include Video for Windows support"           ON   IF WIN32 )
 OCV_OPTION(WITH_FFMPEG         "Include FFMPEG support"                      ON   IF (NOT ANDROID AND NOT IOS))
 OCV_OPTION(WITH_GSTREAMER      "Include Gstreamer support"                   ON   IF (UNIX AND NOT APPLE AND NOT ANDROID) )
+OCV_OPTION(WITH_GSTREAMER_1_X  "Include Gstreamer 1.x support"               OFF)
 OCV_OPTION(WITH_GTK            "Include GTK support"                         ON   IF (UNIX AND NOT APPLE AND NOT ANDROID) )
 OCV_OPTION(WITH_IPP            "Include Intel IPP support"                   OFF  IF (MSVC OR X86 OR X86_64) )
 OCV_OPTION(WITH_JASPER         "Include JPEG2K support"                      ON   IF (NOT IOS) )
-OCV_OPTION(WITH_JPEG           "Include JPEG support"                        ON   IF (NOT IOS) )
+OCV_OPTION(WITH_JPEG           "Include JPEG support"                        ON)
 OCV_OPTION(WITH_WEBP           "Include WebP support"                        ON   IF (NOT IOS) )
 OCV_OPTION(WITH_OPENEXR        "Include ILM support via OpenEXR"             ON   IF (NOT IOS) )
 OCV_OPTION(WITH_OPENGL         "Include OpenGL support"                      OFF  IF (NOT ANDROID AND NOT APPLE) )
 OCV_OPTION(WITH_OPENNI         "Include OpenNI support"                      OFF  IF (NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_PNG            "Include PNG support"                         ON   IF (NOT IOS) )
+OCV_OPTION(WITH_PNG            "Include PNG support"                         ON)
 OCV_OPTION(WITH_PVAPI          "Include Prosilica GigE support"              ON   IF (NOT ANDROID AND NOT IOS) )
 OCV_OPTION(WITH_GIGEAPI        "Include Smartek GigE support"                ON   IF (NOT ANDROID AND NOT IOS) )
 OCV_OPTION(WITH_QT             "Build with Qt Backend support"               OFF  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_WIN32UI        "Build with Win32 UI Backend support"         ON   IF WIN32 )
 OCV_OPTION(WITH_QUICKTIME      "Use QuickTime for Video I/O insted of QTKit" OFF  IF APPLE )
 OCV_OPTION(WITH_TBB            "Include Intel TBB support"                   OFF  IF (NOT IOS) )
 OCV_OPTION(WITH_CSTRIPES       "Include C= support"                          OFF  IF WIN32 )
 OCV_OPTION(WITH_TIFF           "Include TIFF support"                        ON   IF (NOT IOS) )
 OCV_OPTION(WITH_UNICAP         "Include Unicap support (GPL)"                OFF  IF (UNIX AND NOT APPLE AND NOT ANDROID) )
 OCV_OPTION(WITH_V4L            "Include Video 4 Linux support"               ON   IF (UNIX AND NOT ANDROID) )
-OCV_OPTION(WITH_VIDEOINPUT     "Build HighGUI with DirectShow support"       ON   IF WIN32 )
+OCV_OPTION(WITH_DSHOW          "Build HighGUI with DirectShow support"       ON   IF (WIN32 AND NOT ARM) )
+OCV_OPTION(WITH_MSMF           "Build HighGUI with Media Foundation support" OFF  IF WIN32 )
 OCV_OPTION(WITH_XIMEA          "Include XIMEA cameras support"               OFF  IF (NOT ANDROID AND NOT APPLE) )
 OCV_OPTION(WITH_XINE           "Include Xine support (GPL)"                  OFF  IF (UNIX AND NOT APPLE AND NOT ANDROID) )
 OCV_OPTION(WITH_CLP            "Include Clp support (EPL)"                   OFF)
-OCV_OPTION(WITH_OPENCL         "Include OpenCL Runtime support"              OFF  IF (NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_OPENCLAMDFFT   "Include AMD OpenCL FFT library support"      OFF  IF (NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_OPENCLAMDBLAS  "Include AMD OpenCL BLAS library support"     OFF  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_OPENCL         "Include OpenCL Runtime support"              ON  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_OPENCLAMDFFT   "Include AMD OpenCL FFT library support"      ON  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_OPENCLAMDBLAS  "Include AMD OpenCL BLAS library support"     ON  IF (NOT ANDROID AND NOT IOS) )
 
 
 # OpenCV build components
@@ -170,7 +171,7 @@ OCV_OPTION(BUILD_JASPER             "Build libjasper from source"        WIN32 O
 OCV_OPTION(BUILD_JPEG               "Build libjpeg from source"          WIN32 OR ANDROID OR APPLE )
 OCV_OPTION(BUILD_PNG                "Build libpng from source"           WIN32 OR ANDROID OR APPLE )
 OCV_OPTION(BUILD_OPENEXR            "Build openexr from source"          WIN32 OR ANDROID OR APPLE )
-OCV_OPTION(BUILD_TBB                "Download and build TBB from source" ANDROID IF CMAKE_COMPILER_IS_GNUCXX )
+OCV_OPTION(BUILD_TBB                "Download and build TBB from source" ANDROID )
 
 # OpenCV installation options
 # ===================================================
@@ -183,7 +184,7 @@ OCV_OPTION(INSTALL_TO_MANGLED_PATHS "Enables mangled install paths, that help wi
 # OpenCV build options
 # ===================================================
 OCV_OPTION(ENABLE_PRECOMPILED_HEADERS "Use precompiled headers"                                  ON   IF (NOT IOS) )
-OCV_OPTION(ENABLE_SOLUTION_FOLDERS    "Solution folder in Visual Studio or in other IDEs"        (MSVC_IDE OR CMAKE_GENERATOR MATCHES Xcode) IF (CMAKE_VERSION VERSION_GREATER "2.8.0") )
+OCV_OPTION(ENABLE_SOLUTION_FOLDERS    "Solution folder in Visual Studio or in other IDEs"        (MSVC_IDE OR CMAKE_GENERATOR MATCHES Xcode) )
 OCV_OPTION(ENABLE_PROFILING           "Enable profiling in the GCC compiler (Add flags: -g -pg)" OFF  IF CMAKE_COMPILER_IS_GNUCXX )
 OCV_OPTION(ENABLE_OMIT_FRAME_POINTER  "Enable -fomit-frame-pointer for GCC"                      ON   IF CMAKE_COMPILER_IS_GNUCXX AND NOT (APPLE AND CMAKE_COMPILER_IS_CLANGCXX) )
 OCV_OPTION(ENABLE_POWERPC             "Enable PowerPC for GCC"                                   ON   IF (CMAKE_COMPILER_IS_GNUCXX AND CMAKE_SYSTEM_PROCESSOR MATCHES powerpc.*) )
@@ -197,15 +198,7 @@ OCV_OPTION(ENABLE_SSE42               "Enable SSE4.2 instructions"
 OCV_OPTION(ENABLE_AVX                 "Enable AVX instructions"                                  OFF  IF ((MSVC OR CMAKE_COMPILER_IS_GNUCXX) AND (X86 OR X86_64)) )
 OCV_OPTION(ENABLE_NOISY_WARNINGS      "Show all warnings even if they are too noisy"             OFF )
 OCV_OPTION(OPENCV_WARNINGS_ARE_ERRORS "Treat warnings as errors"                                 OFF )
-
-
-# uncategorized options
-# ===================================================
-OCV_OPTION(CMAKE_VERBOSE "Verbose mode" OFF )
-
-# backward compatibility
-# ===================================================
-include(cmake/OpenCVLegacyOptions.cmake OPTIONAL)
+OCV_OPTION(ENABLE_WINRT_MODE          "Build with Windows Runtime support"                       OFF  IF WIN32 )
 
 
 # ----------------------------------------------------------------------------
@@ -272,37 +265,25 @@ if(DEFINED CMAKE_DEBUG_POSTFIX)
   set(OPENCV_DEBUG_POSTFIX "${CMAKE_DEBUG_POSTFIX}")
 endif()
 
-if(CMAKE_VERBOSE)
-  set(CMAKE_VERBOSE_MAKEFILE 1)
-endif()
-
 
 # ----------------------------------------------------------------------------
 #  Path for build/platform -specific headers
 # ----------------------------------------------------------------------------
 set(OPENCV_CONFIG_FILE_INCLUDE_DIR "${CMAKE_BINARY_DIR}/" CACHE PATH "Where to create the platform-dependant cvconfig.h")
-add_definitions(-DHAVE_CVCONFIG_H)
 ocv_include_directories(${OPENCV_CONFIG_FILE_INCLUDE_DIR})
 
+# ----------------------------------------------------------------------------
+#  Path for additional modules
+# ----------------------------------------------------------------------------
+set(OPENCV_EXTRA_MODULES_PATH "" CACHE PATH "Where to look for additional OpenCV modules")
 
 # ----------------------------------------------------------------------------
 #  Autodetect if we are in a GIT repository
 # ----------------------------------------------------------------------------
+find_host_package(Git QUIET)
 
-# don't use FindGit because it requires CMake 2.8.2
-set(git_names git eg) # eg = easy git
-# Prefer .cmd variants on Windows unless running in a Makefile in the MSYS shell
-if(CMAKE_HOST_WIN32)
-  if(NOT CMAKE_GENERATOR MATCHES "MSYS")
-    set(git_names git.cmd git eg.cmd eg)
-  endif()
-endif()
-
-find_host_program(GIT_EXECUTABLE NAMES ${git_names} PATH_SUFFIXES Git/cmd Git/bin DOC "git command line client")
-mark_as_advanced(GIT_EXECUTABLE)
-
-if(GIT_EXECUTABLE)
-  execute_process(COMMAND ${GIT_EXECUTABLE} describe --tags --always --dirty --match "2.[0-9].[0-9]*"
+if(GIT_FOUND)
+  execute_process(COMMAND "${GIT_EXECUTABLE}" describe --tags --always --dirty --match "2.[0-9].[0-9]*"
     WORKING_DIRECTORY "${OpenCV_SOURCE_DIR}"
     OUTPUT_VARIABLE OPENCV_VCSVERSION
     RESULT_VARIABLE GIT_RESULT
@@ -346,14 +327,11 @@ endif(WIN32 AND NOT MINGW)
 #       CHECK FOR SYSTEM LIBRARIES, OPTIONS, ETC..
 # ----------------------------------------------------------------------------
 if(UNIX)
-  include(cmake/OpenCVFindPkgConfig.cmake OPTIONAL)
+  find_package(PkgConfig QUIET)
   include(CheckFunctionExists)
   include(CheckIncludeFile)
 
   if(NOT APPLE)
-    CHECK_INCLUDE_FILE(alloca.h HAVE_ALLOCA_H)
-    CHECK_FUNCTION_EXISTS(alloca HAVE_ALLOCA)
-    CHECK_INCLUDE_FILE(unistd.h HAVE_UNISTD_H)
     CHECK_INCLUDE_FILE(pthread.h HAVE_LIBPTHREAD)
     if(ANDROID)
       set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} dl m log)
@@ -363,7 +341,7 @@ if(UNIX)
       set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} dl m pthread rt)
     endif()
   else()
-    add_definitions(-DHAVE_ALLOCA -DHAVE_ALLOCA_H -DHAVE_LIBPTHREAD -DHAVE_UNISTD_H)
+    set(HAVE_LIBPTHREAD YES)
   endif()
 endif()
 
@@ -400,7 +378,7 @@ if(ANDROID)
   if(NOT ANDROID_TOOLS_Pkg_Revision GREATER 13)
     message(WARNING "OpenCV requires Android SDK tools revision 14 or newer. Otherwise tests and samples will no be compiled.")
   endif()
-elseif(ANT_EXECUTABLE)
+else()
   find_package(JNI)
 endif()
 
@@ -413,15 +391,6 @@ endif()
 # --- OpenCL ---
 if(WITH_OPENCL)
   include(cmake/OpenCVDetectOpenCL.cmake)
-  if(OPENCL_FOUND)
-    set(HAVE_OPENCL 1)
-  endif()
-  if(WITH_OPENCLAMDFFT AND CLAMDFFT_INCLUDE_DIR)
-    set(HAVE_CLAMDFFT 1)
-  endif()
-  if(WITH_OPENCLAMDBLAS AND CLAMDBLAS_INCLUDE_DIR)
-    set(HAVE_CLAMDBLAS 1)
-  endif()
 endif()
 
 # ----------------------------------------------------------------------------
@@ -463,15 +432,15 @@ if(BUILD_EXAMPLES OR BUILD_ANDROID_EXAMPLES OR INSTALL_PYTHON_EXAMPLES)
 endif()
 
 if(ANDROID)
-  add_subdirectory(android/service)
+  add_subdirectory(platforms/android/service)
 endif()
 
 if(BUILD_ANDROID_PACKAGE)
-  add_subdirectory(android/package)
+  add_subdirectory(platforms/android/package)
 endif()
 
 if (ANDROID)
-  add_subdirectory(android/libinfo)
+  add_subdirectory(platforms/android/libinfo)
 endif()
 
 # ----------------------------------------------------------------------------
@@ -491,6 +460,8 @@ include(cmake/OpenCVGenAndroidMK.cmake)
 # Generate OpenCVСonfig.cmake and OpenCVConfig-version.cmake for cmake projects
 include(cmake/OpenCVGenConfig.cmake)
 
+# Generate Info.plist for the IOS framework
+include(cmake/OpenCVGenInfoPlist.cmake)
 
 # ----------------------------------------------------------------------------
 # Summary:
@@ -597,19 +568,32 @@ if(ANDROID)
   status("    Android examples:"    BUILD_ANDROID_EXAMPLES AND CAN_BUILD_ANDROID_PROJECTS         THEN YES ELSE NO)
 endif()
 
+# ================== Windows RT features ==================
+if(WIN32)
+status("")
+    status("  Windows RT support:" HAVE_WINRT THEN YES ELSE NO)
+    if (ENABLE_WINRT_MODE)
+      status("    Windows SDK v8.0:" ${WINDOWS_SDK_PATH})
+      status("    Visual Studio 2012:" ${VISUAL_STUDIO_PATH})
+    endif()
+endif(WIN32)
+
 # ========================== GUI ==========================
 status("")
 status("  GUI: ")
 
-if(HAVE_QT)
+if(HAVE_QT5)
+  status("    QT 5.x:"            HAVE_QT        THEN "YES (ver ${Qt5Core_VERSION_STRING})" ELSE NO)
+  status("    QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${Qt5OpenGL_LIBRARIES} ${Qt5OpenGL_VERSION_STRING})" ELSE NO)
+elseif(HAVE_QT)
   status("    QT 4.x:"            HAVE_QT        THEN "YES (ver ${QT_VERSION_MAJOR}.${QT_VERSION_MINOR}.${QT_VERSION_PATCH} ${QT_EDITION})" ELSE NO)
   status("    QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${QT_QTOPENGL_LIBRARY})" ELSE NO)
 else()
   if(DEFINED WITH_QT)
-    status("    QT 4.x:" NO)
+    status("    QT:" NO)
   endif()
-  if(WIN32)
-    status("    Win32 UI:" YES)
+  if(DEFINED WITH_WIN32UI)
+    status("    Win32 UI:" HAVE_WIN32UI THEN YES ELSE NO)
   else()
     if(APPLE)
       if(WITH_CARBON)
@@ -630,7 +614,7 @@ status("    OpenGL support:" HAVE_OPENGL THEN "YES (${OPENGL_LIBRARIES})" ELSE N
 # ========================== MEDIA IO ==========================
 status("")
 status("  Media I/O: ")
-status("    ZLib:"         BUILD_ZLIB    THEN "build (ver ${ZLIB_VERSION_STRING})"               ELSE "${ZLIB_LIBRARY} (ver ${ZLIB_VERSION_STRING})")
+status("    ZLib:"         BUILD_ZLIB    THEN "build (ver ${ZLIB_VERSION_STRING})"               ELSE "${ZLIB_LIBRARIES} (ver ${ZLIB_VERSION_STRING})")
 
 if(WITH_JPEG)
   status("    JPEG:"       JPEG_FOUND    THEN "${JPEG_LIBRARY} (ver ${JPEG_LIB_VERSION})"        ELSE "build (ver ${JPEG_LIB_VERSION})")
@@ -673,6 +657,10 @@ endif()
 status("")
 status("  Video I/O:")
 
+if (DEFINED WITH_VFW)
+  status("    Video for Windows:" HAVE_VFW         THEN YES                                        ELSE NO)
+endif(DEFINED WITH_VFW)
+
 if(DEFINED WITH_1394)
   status("    DC1394 1.x:"     HAVE_DC1394         THEN "YES (ver ${ALIASOF_libdc1394_VERSION})"   ELSE NO)
   status("    DC1394 2.x:"     HAVE_DC1394_2       THEN "YES (ver ${ALIASOF_libdc1394-2_VERSION})" ELSE NO)
@@ -707,10 +695,12 @@ endif(DEFINED WITH_FFMPEG)
 if(DEFINED WITH_GSTREAMER)
   status("    GStreamer:"      HAVE_GSTREAMER      THEN ""                                         ELSE NO)
   if(HAVE_GSTREAMER)
-    status("      base:"       "YES (ver ${ALIASOF_gstreamer-base-0.10_VERSION})")
-    status("      app:"        "YES (ver ${ALIASOF_gstreamer-app-0.10_VERSION})")
-    status("      video:"      "YES (ver ${ALIASOF_gstreamer-video-0.10_VERSION})")
-  endif()
+    status("      base:"       "YES (ver ${GSTREAMER_BASE_VERSION})")
+    status("      video:"      "YES (ver ${GSTREAMER_VIDEO_VERSION})")
+    status("      app:"        "YES (ver ${GSTREAMER_APP_VERSION})")
+    status("      riff:"       "YES (ver ${GSTREAMER_RIFF_VERSION})")
+    status("      pbutils:"    "YES (ver ${GSTREAMER_PBUTILS_VERSION})")
+  endif(HAVE_GSTREAMER)
 endif(DEFINED WITH_GSTREAMER)
 
 if(DEFINED WITH_OPENNI)
@@ -729,8 +719,8 @@ if(DEFINED WITH_GIGEAPI)
 endif(DEFINED WITH_GIGEAPI)
 
 if(DEFINED WITH_QUICKTIME)
-  status("    QuickTime:"      WITH_QUICKTIME      THEN YES                                        ELSE NO)
-  status("    QTKit:"          WITH_QUICKTIME      THEN NO                                         ELSE YES)
+  status("    QuickTime:"      HAVE_QUICKTIME      THEN YES                                        ELSE NO)
+  status("    QTKit:"          HAVE_QTKIT          THEN YES                                        ELSE NO)
 endif(DEFINED WITH_QUICKTIME)
 
 if(DEFINED WITH_UNICAP)
@@ -755,9 +745,13 @@ if(DEFINED WITH_V4L)
                                                    ELSE "${HAVE_CAMV4L_STR}/${HAVE_CAMV4L2_STR}")
 endif(DEFINED WITH_V4L)
 
-if(DEFINED WITH_VIDEOINPUT)
-  status("    DirectShow:"     HAVE_VIDEOINPUT     THEN YES                                        ELSE NO)
-endif(DEFINED WITH_VIDEOINPUT)
+if(DEFINED WITH_DSHOW)
+  status("    DirectShow:"     HAVE_DSHOW     THEN YES                                        ELSE NO)
+endif(DEFINED WITH_DSHOW)
+
+if(DEFINED WITH_MSMF)
+  status("    Media Foundation:" HAVE_MSMF    THEN YES                                        ELSE NO)
+endif(DEFINED WITH_MSMF)
 
 if(DEFINED WITH_XIMEA)
   status("    XIMEA:"          HAVE_XIMEA          THEN YES                                        ELSE NO)
@@ -799,11 +793,11 @@ if(HAVE_CUDA)
   status("    Use fast math:"        CUDA_FAST_MATH THEN YES ELSE NO)
 endif()
 
-if(HAVE_OPENCL AND BUILD_opencv_ocl)
+if(HAVE_OPENCL)
   status("")
   status("  OpenCL")
   if(OPENCL_INCLUDE_DIR)
-    status("    Include:"            ${OPENCL_INCLUDE_DIR})
+    status("    Include path:"       ${OPENCL_INCLUDE_DIRS})
   endif()
   if(OPENCL_LIBRARIES)
     status("    libraries:"          ${OPENCL_LIBRARIES})
@@ -815,15 +809,15 @@ endif()
 # ========================== python ==========================
 status("")
 status("  Python:")
-status("    Interpreter:"     PYTHON_EXECUTABLE   THEN "${PYTHON_EXECUTABLE} (ver ${PYTHON_VERSION_FULL})"         ELSE NO)
+status("    Interpreter:"     PYTHONINTERP_FOUND  THEN "${PYTHON_EXECUTABLE} (ver ${PYTHON_VERSION_STRING})"       ELSE NO)
 if(BUILD_opencv_python)
   if(PYTHONLIBS_VERSION_STRING)
     status("    Libraries:"   HAVE_opencv_python  THEN  "${PYTHON_LIBRARIES} (ver ${PYTHONLIBS_VERSION_STRING})"   ELSE NO)
   else()
-    status("    Libraries:"   HAVE_opencv_python  THEN  ${PYTHON_LIBRARIES}                                        ELSE NO)
+    status("    Libraries:"   HAVE_opencv_python  THEN  "${PYTHON_LIBRARIES}"                                      ELSE NO)
   endif()
-  status("    numpy:"         PYTHON_USE_NUMPY    THEN "${PYTHON_NUMPY_INCLUDE_DIR} (ver ${PYTHON_NUMPY_VERSION})" ELSE "NO (Python wrappers can not be generated)")
-  status("    packages path:" PYTHON_EXECUTABLE   THEN "${PYTHON_PACKAGES_PATH}"                                   ELSE "-")
+  status("    numpy:"         PYTHON_NUMPY_INCLUDE_DIRS THEN "${PYTHON_NUMPY_INCLUDE_DIRS} (ver ${PYTHON_NUMPY_VERSION})" ELSE "NO (Python wrappers can not be generated)")
+  status("    packages path:" PYTHON_EXECUTABLE         THEN "${PYTHON_PACKAGES_PATH}"                                    ELSE "-")
 endif()
 
 # ========================== java ==========================
@@ -833,7 +827,7 @@ status("    ant:"           ANT_EXECUTABLE      THEN "${ANT_EXECUTABLE} (ver ${A
 if(NOT ANDROID)
   status("    JNI:"         JNI_INCLUDE_DIRS    THEN "${JNI_INCLUDE_DIRS}"                                       ELSE NO)
 endif()
-status("    Java tests:"    BUILD_TESTS AND (NOT ANDROID OR CAN_BUILD_ANDROID_PROJECTS)                 THEN YES ELSE NO)
+status("    Java tests:"    BUILD_TESTS AND (CAN_BUILD_ANDROID_PROJECTS OR HAVE_opencv_java)            THEN YES ELSE NO)
 
 # ========================== documentation ==========================
 if(BUILD_DOCS)
@@ -846,6 +840,7 @@ if(BUILD_DOCS)
   endif()
   status("    Sphinx:"              HAVE_SPHINX              THEN "${SPHINX_BUILD} (ver ${SPHINX_VERSION})" ELSE NO)
   status("    PdfLaTeX compiler:"   PDFLATEX_COMPILER        THEN "${PDFLATEX_COMPILER}" ELSE NO)
+  status("    PlantUML:"            PLANTUML                 THEN "${PLANTUML}" ELSE NO)
 endif()
 
 # ========================== samples and tests ==========================
@@ -871,4 +866,3 @@ ocv_finalize_status()
 if("${CMAKE_CURRENT_SOURCE_DIR}" STREQUAL "${CMAKE_CURRENT_BINARY_DIR}")
   message(WARNING "The source directory is the same as binary directory. \"make clean\" may damage the source tree")
 endif()
-
diff --git a/README b/README
deleted file mode 100644
index 9dd45a230b..0000000000
--- a/README
+++ /dev/null
@@ -1,6 +0,0 @@
-OpenCV: open source computer vision library
-
-Homepage:    http://opencv.org
-Online docs: http://docs.opencv.org
-Q&A forum:   http://answers.opencv.org
-Dev zone:    http://code.opencv.org
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000..403f118eed
--- /dev/null
+++ b/README.md
@@ -0,0 +1,23 @@
+### OpenCV: Open Source Computer Vision Library
+
+#### Resources
+
+* Homepage: <http://opencv.org>
+* Docs: <http://docs.opencv.org>
+* Q&A forum: <http://answers.opencv.org>
+* Issue tracking: <http://code.opencv.org>
+
+#### Contributing
+
+Please read before starting work on a pull request: <http://code.opencv.org/projects/opencv/wiki/How_to_contribute>
+
+Summary of guidelines:
+
+* One pull request per issue;
+* Choose the right base branch;
+* Include tests and documentation;
+* Clean up "oops" commits before submitting;
+* Follow the coding style guide.
+
+[![Donate OpenCV project](http://opencv.org/wp-content/uploads/2013/07/gittip1.png)](https://www.gittip.com/OpenCV/)
+[![Donate OpenCV project](http://opencv.org/wp-content/uploads/2013/07/paypal-donate-button.png)](https://www.paypal.com/cgi-bin/webscr?item_name=Donation+to+OpenCV&cmd=_donations&business=accountant%40opencv.org)
\ No newline at end of file
diff --git a/android/scripts/build.cmd b/android/scripts/build.cmd
deleted file mode 100644
index 3e0f1666b6..0000000000
--- a/android/scripts/build.cmd
+++ /dev/null
@@ -1,90 +0,0 @@
-@ECHO OFF
-
-:: enable command extensions
-VERIFY BADVALUE 2>NUL
-SETLOCAL ENABLEEXTENSIONS || (ECHO Unable to enable command extensions. & EXIT \B)
-
-:: build environment
-SET SOURCE_DIR=%cd%
-IF EXIST .\android.toolchain.cmake (SET BUILD_OPENCV=1) ELSE (SET BUILD_OPENCV=0)
-IF EXIST .\jni\nul (SET BUILD_JAVA_PART=1) ELSE (SET BUILD_JAVA_PART=0)
-
-:: load configuration
-PUSHD %~dp0
-SET SCRIPTS_DIR=%cd%
-IF EXIST .\wincfg.cmd CALL .\wincfg.cmd
-POPD
-
-:: inherit old names
-IF NOT DEFINED CMAKE SET CMAKE=%CMAKE_EXE%
-IF NOT DEFINED MAKE SET MAKE=%MAKE_EXE%
-
-:: defaults
-IF NOT DEFINED BUILD_DIR SET BUILD_DIR=build
-IF NOT DEFINED ANDROID_ABI SET ANDROID_ABI=armeabi-v7a
-SET OPENCV_BUILD_DIR=%SCRIPTS_DIR%\..\%BUILD_DIR%
-
-:: check that all required variables defined
-PUSHD .
-IF NOT DEFINED ANDROID_NDK (ECHO. & ECHO You should set an environment variable ANDROID_NDK to the full path to your copy of Android NDK & GOTO end)
-(CD "%ANDROID_NDK%") || (ECHO. & ECHO Directory "%ANDROID_NDK%" specified by ANDROID_NDK variable does not exist & GOTO end)
-
-IF NOT EXIST "%CMAKE%" (ECHO. & ECHO You should set an environment variable CMAKE to the full path to cmake executable & GOTO end)
-IF NOT EXIST "%MAKE%" (ECHO. & ECHO You should set an environment variable MAKE to the full path to native port of make executable & GOTO end)
-
-IF NOT %BUILD_JAVA_PART%==1 GOTO required_variables_checked
-
-IF NOT DEFINED ANDROID_SDK (ECHO. & ECHO You should set an environment variable ANDROID_SDK to the full path to your copy of Android SDK & GOTO end)
-(CD "%ANDROID_SDK%" 2>NUL) || (ECHO. & ECHO Directory "%ANDROID_SDK%" specified by ANDROID_SDK variable does not exist & GOTO end)
-
-IF NOT DEFINED ANT_DIR (ECHO. & ECHO You should set an environment variable ANT_DIR to the full path to Apache Ant root & GOTO end)
-(CD "%ANT_DIR%" 2>NUL) || (ECHO. & ECHO Directory "%ANT_DIR%" specified by ANT_DIR variable does not exist & GOTO end)
-
-IF NOT DEFINED JAVA_HOME (ECHO. & ECHO You should set an environment variable JAVA_HOME to the full path to JDK & GOTO end)
-(CD "%JAVA_HOME%" 2>NUL) || (ECHO. & ECHO Directory "%JAVA_HOME%" specified by JAVA_HOME variable does not exist & GOTO end)
-
-:required_variables_checked
-POPD
-
-:: check for ninja
-echo "%MAKE%"|findstr /i ninja >nul:
-IF %errorlevel%==1 (SET BUILD_WITH_NINJA=0) ELSE (SET BUILD_WITH_NINJA=1)
-IF %BUILD_WITH_NINJA%==1 (SET CMAKE_GENERATOR=Ninja) ELSE (SET CMAKE_GENERATOR=MinGW Makefiles)
-
-:: create build dir
-IF DEFINED REBUILD rmdir /S /Q "%BUILD_DIR%" 2>NUL
-MKDIR "%BUILD_DIR%" 2>NUL
-PUSHD "%BUILD_DIR%" || (ECHO. & ECHO Directory "%BUILD_DIR%" is not found & GOTO end)
-
-:: run cmake
-ECHO. & ECHO Runnning cmake...
-ECHO ANDROID_ABI=%ANDROID_ABI%
-ECHO.
-IF NOT %BUILD_OPENCV%==1 GOTO other-cmake
-:opencv-cmake
-("%CMAKE%" -G"%CMAKE_GENERATOR%" -DANDROID_ABI="%ANDROID_ABI%" -DCMAKE_TOOLCHAIN_FILE="%SOURCE_DIR%"\android.toolchain.cmake -DCMAKE_MAKE_PROGRAM="%MAKE%" %* "%SOURCE_DIR%\..") && GOTO cmakefin
-ECHO. & ECHO cmake failed &	GOTO end
-:other-cmake
-("%CMAKE%" -G"%CMAKE_GENERATOR%" -DANDROID_ABI="%ANDROID_ABI%" -DOpenCV_DIR="%OPENCV_BUILD_DIR%" -DCMAKE_TOOLCHAIN_FILE="%OPENCV_BUILD_DIR%\..\android.toolchain.cmake" -DCMAKE_MAKE_PROGRAM="%MAKE%" %* "%SOURCE_DIR%") && GOTO cmakefin
-ECHO. & ECHO cmake failed &	GOTO end
-:cmakefin
-
-:: run make
-ECHO. & ECHO Building native libs...
-IF %BUILD_WITH_NINJA%==0 ("%MAKE%" -j %NUMBER_OF_PROCESSORS% VERBOSE=%VERBOSE%) || (ECHO. & ECHO make failed & GOTO end)
-IF %BUILD_WITH_NINJA%==1 ("%MAKE%") || (ECHO. & ECHO ninja failed & GOTO end)
-
-IF NOT %BUILD_JAVA_PART%==1 GOTO end
-POPD && PUSHD %SOURCE_DIR%
-
-:: configure java part
-ECHO. & ECHO Updating Android project...
-(CALL "%ANDROID_SDK%\tools\android" update project --name %PROJECT_NAME% --path .) || (ECHO. & ECHO failed to update android project & GOTO end)
-
-:: compile java part
-ECHO. & ECHO Compiling Android project...
-(CALL "%ANT_DIR%\bin\ant" debug) || (ECHO. & ECHO failed to compile android project & GOTO end)
-
-:end
-POPD
-ENDLOCAL
diff --git a/android/scripts/cmake_android.cmd b/android/scripts/cmake_android.cmd
deleted file mode 100644
index 212c04b47e..0000000000
--- a/android/scripts/cmake_android.cmd
+++ /dev/null
@@ -1,5 +0,0 @@
-@ECHO OFF
-
-PUSHD %~dp0..
-CALL .\scripts\build.cmd %* -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON
-POPD
\ No newline at end of file
diff --git a/android/scripts/cmake_android_armeabi.sh b/android/scripts/cmake_android_armeabi.sh
deleted file mode 100755
index 9c711d8855..0000000000
--- a/android/scripts/cmake_android_armeabi.sh
+++ /dev/null
@@ -1,8 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_armeabi
-cd build_armeabi
-
-cmake -DANDROID_ABI=armeabi -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake $@ ../..
-
diff --git a/android/scripts/cmake_android_mips.sh b/android/scripts/cmake_android_mips.sh
deleted file mode 100755
index 17d2ff937e..0000000000
--- a/android/scripts/cmake_android_mips.sh
+++ /dev/null
@@ -1,8 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_mips
-cd build_mips
-
-cmake -DANDROID_ABI=mips -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake $@ ../..
-
diff --git a/android/scripts/cmake_android_neon.sh b/android/scripts/cmake_android_neon.sh
deleted file mode 100755
index 5e85605b56..0000000000
--- a/android/scripts/cmake_android_neon.sh
+++ /dev/null
@@ -1,8 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_neon
-cd build_neon
-
-cmake -DANDROID_ABI="armeabi-v7a with NEON" -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake $@ ../..
-
diff --git a/android/scripts/cmake_android_service.sh b/android/scripts/cmake_android_service.sh
deleted file mode 100755
index 0dbd482520..0000000000
--- a/android/scripts/cmake_android_service.sh
+++ /dev/null
@@ -1,7 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_service
-cd build_service
-
-cmake -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake -DANDROID_TOOLCHAIN_NAME="arm-linux-androideabi-4.4.3" -DANDROID_STL=stlport_static -DANDROID_STL_FORCE_FEATURES=OFF -DBUILD_ANDROID_SERVICE=ON -DANDROID_SOURCE_TREE=~/Projects/AndroidSource/ServiceStub/ $@ ../..
diff --git a/android/scripts/cmake_android_x86.sh b/android/scripts/cmake_android_x86.sh
deleted file mode 100755
index a01df2e668..0000000000
--- a/android/scripts/cmake_android_x86.sh
+++ /dev/null
@@ -1,9 +0,0 @@
-#!/bin/sh
-
-cd `dirname $0`/..
-
-mkdir -p build_x86
-cd build_x86
-
-cmake -DANDROID_ABI=x86 -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake $@ ../..
-
diff --git a/android/scripts/wincfg.cmd.tmpl b/android/scripts/wincfg.cmd.tmpl
deleted file mode 100644
index 166a5e7b02..0000000000
--- a/android/scripts/wincfg.cmd.tmpl
+++ /dev/null
@@ -1,30 +0,0 @@
-:: variables required for OpenCV build ::
-:: Note: all pathes should be specified without tailing slashes!
-SET ANDROID_NDK=C:\full\path\to\your\copy\of\android\NDK\android-ndk-r7b
-SET CMAKE_EXE=C:\full\path\to\cmake\utility\cmake.exe
-SET MAKE_EXE=%ANDROID_NDK%\prebuilt\windows\bin\make.exe
-
-:: variables required for android-opencv build ::
-SET ANDROID_SDK=C:\full\path\to\your\copy\of\android\SDK\android-sdk-windows
-SET ANT_DIR=C:\full\path\to\ant\directory\apache-ant-1.8.2
-SET JAVA_HOME=C:\full\path\to\JDK\jdk1.6.0_25
-
-:: configuration options ::
-:::: general ARM-V7 settings
-SET ANDROID_ABI=armeabi-v7a
-SET BUILD_DIR=build
-
-:::: uncomment following lines to compile for old emulator or old device
-::SET ANDROID_ABI=armeabi
-::SET BUILD_DIR=build_armeabi
-
-:::: uncomment following lines to compile for ARM-V7 with NEON support
-::SET ANDROID_ABI=armeabi-v7a with NEON
-::SET BUILD_DIR=build_neon
-
-:::: uncomment following lines to compile for x86
-::SET ANDROID_ABI=x86
-::SET BUILD_DIR=build_x86
-
-:::: other options
-::SET ANDROID_NATIVE_API_LEVEL=8   &:: android-3 is enough for native part of OpenCV but android-8 is required for Java API
diff --git a/android/service/doc/Makefile b/android/service/doc/Makefile
deleted file mode 100644
index b8e7bba113..0000000000
--- a/android/service/doc/Makefile
+++ /dev/null
@@ -1,89 +0,0 @@
-# Makefile for Sphinx documentation
-#
-
-# You can set these variables from the command line.
-SPHINXOPTS    =
-SPHINXBUILD   = sphinx-build
-PAPER         =
-BUILDDIR      = _build
-
-# Internal variables.
-PAPEROPT_a4     = -D latex_paper_size=a4
-PAPEROPT_letter = -D latex_paper_size=letter
-ALLSPHINXOPTS   = -d $(BUILDDIR)/doctrees $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
-
-.PHONY: help clean html dirhtml pickle json htmlhelp qthelp latex changes linkcheck doctest
-
-help:
-	@echo "Please use \`make <target>' where <target> is one of"
-	@echo "  html      to make standalone HTML files"
-	@echo "  dirhtml   to make HTML files named index.html in directories"
-	@echo "  pickle    to make pickle files"
-	@echo "  json      to make JSON files"
-	@echo "  htmlhelp  to make HTML files and a HTML help project"
-	@echo "  qthelp    to make HTML files and a qthelp project"
-	@echo "  latex     to make LaTeX files, you can set PAPER=a4 or PAPER=letter"
-	@echo "  changes   to make an overview of all changed/added/deprecated items"
-	@echo "  linkcheck to check all external links for integrity"
-	@echo "  doctest   to run all doctests embedded in the documentation (if enabled)"
-
-clean:
-	-rm -rf $(BUILDDIR)/*
-
-html:
-	$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)/html
-	@echo
-	@echo "Build finished. The HTML pages are in $(BUILDDIR)/html."
-
-dirhtml:
-	$(SPHINXBUILD) -b dirhtml $(ALLSPHINXOPTS) $(BUILDDIR)/dirhtml
-	@echo
-	@echo "Build finished. The HTML pages are in $(BUILDDIR)/dirhtml."
-
-pickle:
-	$(SPHINXBUILD) -b pickle $(ALLSPHINXOPTS) $(BUILDDIR)/pickle
-	@echo
-	@echo "Build finished; now you can process the pickle files."
-
-json:
-	$(SPHINXBUILD) -b json $(ALLSPHINXOPTS) $(BUILDDIR)/json
-	@echo
-	@echo "Build finished; now you can process the JSON files."
-
-htmlhelp:
-	$(SPHINXBUILD) -b htmlhelp $(ALLSPHINXOPTS) $(BUILDDIR)/htmlhelp
-	@echo
-	@echo "Build finished; now you can run HTML Help Workshop with the" \
-	      ".hhp project file in $(BUILDDIR)/htmlhelp."
-
-qthelp:
-	$(SPHINXBUILD) -b qthelp $(ALLSPHINXOPTS) $(BUILDDIR)/qthelp
-	@echo
-	@echo "Build finished; now you can run "qcollectiongenerator" with the" \
-	      ".qhcp project file in $(BUILDDIR)/qthelp, like this:"
-	@echo "# qcollectiongenerator $(BUILDDIR)/qthelp/OpenCVEngine.qhcp"
-	@echo "To view the help file:"
-	@echo "# assistant -collectionFile $(BUILDDIR)/qthelp/OpenCVEngine.qhc"
-
-latex:
-	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
-	@echo
-	@echo "Build finished; the LaTeX files are in $(BUILDDIR)/latex."
-	@echo "Run \`make all-pdf' or \`make all-ps' in that directory to" \
-	      "run these through (pdf)latex."
-
-changes:
-	$(SPHINXBUILD) -b changes $(ALLSPHINXOPTS) $(BUILDDIR)/changes
-	@echo
-	@echo "The overview file is in $(BUILDDIR)/changes."
-
-linkcheck:
-	$(SPHINXBUILD) -b linkcheck $(ALLSPHINXOPTS) $(BUILDDIR)/linkcheck
-	@echo
-	@echo "Link check complete; look for any errors in the above output " \
-	      "or in $(BUILDDIR)/linkcheck/output.txt."
-
-doctest:
-	$(SPHINXBUILD) -b doctest $(ALLSPHINXOPTS) $(BUILDDIR)/doctest
-	@echo "Testing of doctests in the sources finished, look at the " \
-	      "results in $(BUILDDIR)/doctest/output.txt."
diff --git a/android/service/engine/jni/BinderComponent/TegraDetector.cpp b/android/service/engine/jni/BinderComponent/TegraDetector.cpp
deleted file mode 100644
index f7db1fa85d..0000000000
--- a/android/service/engine/jni/BinderComponent/TegraDetector.cpp
+++ /dev/null
@@ -1,53 +0,0 @@
-#include "TegraDetector.h"
-#include <zlib.h>
-#include <string.h>
-
-#define KERNEL_CONFIG "/proc/config.gz"
-#define KERNEL_CONFIG_MAX_LINE_WIDTH 512
-#define KERNEL_CONFIG_TEGRA_MAGIC "CONFIG_ARCH_TEGRA=y"
-#define KERNEL_CONFIG_TEGRA2_MAGIC "CONFIG_ARCH_TEGRA_2x_SOC=y"
-#define KERNEL_CONFIG_TEGRA3_MAGIC "CONFIG_ARCH_TEGRA_3x_SOC=y"
-#define MAX_DATA_LEN    4096
-
-int DetectTegra()
-{
-    int result = TEGRA_NOT_TEGRA;
-    gzFile kernelConfig = gzopen(KERNEL_CONFIG, "r");
-    if (kernelConfig != 0)
-    {
-        char tmpbuf[KERNEL_CONFIG_MAX_LINE_WIDTH];
-        const char *tegra_config = KERNEL_CONFIG_TEGRA_MAGIC;
-        const char *tegra2_config = KERNEL_CONFIG_TEGRA2_MAGIC;
-        const char *tegra3_config = KERNEL_CONFIG_TEGRA3_MAGIC;
-        int len = strlen(tegra_config);
-        int len2 = strlen(tegra2_config);
-        int len3 = strlen(tegra3_config);
-        while (0 != gzgets(kernelConfig, tmpbuf, KERNEL_CONFIG_MAX_LINE_WIDTH))
-        {
-            if (0 == strncmp(tmpbuf, tegra_config, len))
-            {
-                result = 1;
-            }
-
-            if (0 == strncmp(tmpbuf, tegra2_config, len2))
-            {
-                result = 2;
-                break;
-            }
-
-            if (0 == strncmp(tmpbuf, tegra3_config, len3))
-            {
-                result = 3;
-                break;
-            }
-
-        }
-        gzclose(kernelConfig);
-    }
-    else
-    {
-        result = TEGRA_DETECTOR_ERROR;
-    }
-
-    return result;
-}
\ No newline at end of file
diff --git a/android/service/engine/jni/BinderComponent/TegraDetector.h b/android/service/engine/jni/BinderComponent/TegraDetector.h
deleted file mode 100644
index 8fbdb607a1..0000000000
--- a/android/service/engine/jni/BinderComponent/TegraDetector.h
+++ /dev/null
@@ -1,9 +0,0 @@
-#ifndef __TEGRA_DETECTOR_H__
-#define __TEGRA_DETECTOR_H__
-
-#define TEGRA_DETECTOR_ERROR -2
-#define TEGRA_NOT_TEGRA -1
-
-int DetectTegra();
-
-#endif
\ No newline at end of file
diff --git a/android/service/engine/src/org/opencv/engine/BinderConnector.java b/android/service/engine/src/org/opencv/engine/BinderConnector.java
deleted file mode 100644
index fd23fbfe49..0000000000
--- a/android/service/engine/src/org/opencv/engine/BinderConnector.java
+++ /dev/null
@@ -1,26 +0,0 @@
-package org.opencv.engine;
-
-import android.os.IBinder;
-
-public class BinderConnector
-{
-    public BinderConnector(MarketConnector Market)
-    {
-        Init(Market);
-    }
-    public native IBinder Connect();
-    public boolean Disconnect()
-    {
-        Final();
-        return true;
-    }
-
-    static
-    {
-        System.loadLibrary("OpenCVEngine");
-        System.loadLibrary("OpenCVEngine_jni");
-    }
-
-    private native boolean Init(MarketConnector Market);
-    public native void Final();
-}
diff --git a/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java b/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java
deleted file mode 100644
index df31c7fe8e..0000000000
--- a/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java
+++ /dev/null
@@ -1,42 +0,0 @@
-package org.opencv.engine;
-
-import android.app.Service;
-import android.content.Intent;
-import android.os.IBinder;
-import android.util.Log;
-
-
-public class OpenCVEngineService extends Service
-{
-    private static final String TAG = "OpenCVEngine/Service";
-    private IBinder mEngineInterface;
-    private MarketConnector mMarket;
-    private BinderConnector mNativeBinder;
-    public void onCreate()
-    {
-        Log.i(TAG, "Service starting");
-        super.onCreate();
-        Log.i(TAG, "Engine binder component creating");
-        mMarket = new MarketConnector(getBaseContext());
-        mNativeBinder = new BinderConnector(mMarket);
-        mEngineInterface = mNativeBinder.Connect();
-        Log.i(TAG, "Service started successfully");
-    }
-
-    public IBinder onBind(Intent intent)
-    {
-        Log.i(TAG, "Service onBind called for intent " + intent.toString());
-        return mEngineInterface;
-    }
-    public boolean onUnbind(Intent intent)
-    {
-        Log.i(TAG, "Service onUnbind called for intent " + intent.toString());
-        return true;
-    }
-    public void OnDestroy()
-    {
-        Log.i(TAG, "OpenCV Engine service destruction");
-        mNativeBinder.Disconnect();
-    }
-
-}
diff --git a/apps/CMakeLists.txt b/apps/CMakeLists.txt
index ad0f0daade..f5ac42a018 100644
--- a/apps/CMakeLists.txt
+++ b/apps/CMakeLists.txt
@@ -3,4 +3,3 @@ add_definitions(-D__OPENCV_BUILD=1)
 add_subdirectory(haartraining)
 add_subdirectory(traincascade)
 add_subdirectory(sft)
-
diff --git a/apps/haartraining/CMakeLists.txt b/apps/haartraining/CMakeLists.txt
index 953be3b7e5..7a197db837 100644
--- a/apps/haartraining/CMakeLists.txt
+++ b/apps/haartraining/CMakeLists.txt
@@ -1,4 +1,4 @@
-SET(OPENCV_HAARTRAINING_DEPS opencv_core opencv_imgproc opencv_highgui opencv_objdetect opencv_calib3d opencv_video opencv_features2d opencv_flann opencv_legacy)
+SET(OPENCV_HAARTRAINING_DEPS opencv_core opencv_imgproc opencv_photo opencv_ml opencv_highgui opencv_objdetect opencv_calib3d opencv_video opencv_features2d opencv_flann opencv_legacy)
 ocv_check_dependencies(${OPENCV_HAARTRAINING_DEPS})
 
 if(NOT OCV_DEPENDENCIES_FOUND)
@@ -79,4 +79,3 @@ if(ENABLE_SOLUTION_FOLDERS)
   set_target_properties(opencv_haartraining PROPERTIES FOLDER "applications")
   set_target_properties(opencv_haartraining_engine PROPERTIES FOLDER "applications")
 endif()
-
diff --git a/apps/haartraining/_cvcommon.h b/apps/haartraining/_cvcommon.h
index 5b363f1fce..92fee8e84a 100644
--- a/apps/haartraining/_cvcommon.h
+++ b/apps/haartraining/_cvcommon.h
@@ -43,7 +43,6 @@
 #define __CVCOMMON_H_
 
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "cxcore.h"
 #include "cv.h"
@@ -53,15 +52,6 @@
 #define __END__  __CV_END__
 #define EXIT __CV_EXIT__
 
-#define CV_DECLARE_QSORT( func_name, T, less_than )                     \
-void func_name( T* array, size_t length, int aux );
-
-#define less_than( a, b ) ((a) < (b))
-
-CV_DECLARE_QSORT( icvSort_32f, float, less_than )
-
-CV_DECLARE_QSORT( icvSort_32s, int, less_than )
-
 #ifndef PATH_MAX
 #define PATH_MAX 512
 #endif /* PATH_MAX */
@@ -100,4 +90,3 @@ int icvGetIdxAt( CvMat* idx, int pos )
 void icvSave( const CvArr* ptr, const char* filename, int line );
 
 #endif /* __CVCOMMON_H_ */
-
diff --git a/apps/haartraining/cvboost.cpp b/apps/haartraining/cvboost.cpp
index 8dfd3dd316..4b9f24f1bc 100644
--- a/apps/haartraining/cvboost.cpp
+++ b/apps/haartraining/cvboost.cpp
@@ -39,9 +39,7 @@
 //
 //M*/
 
-#ifdef HAVE_CVCONFIG_H
-  #include "cvconfig.h"
-#endif
+#include "cvconfig.h"
 
 #ifdef HAVE_MALLOC_H
   #include <malloc.h>
@@ -76,15 +74,18 @@ typedef struct CvValArray
     size_t step;
 } CvValArray;
 
-#define CMP_VALUES( idx1, idx2 )                                 \
-    ( *( (float*) (aux->data + ((int) (idx1)) * aux->step ) ) <  \
-      *( (float*) (aux->data + ((int) (idx2)) * aux->step ) ) )
-
-static CV_IMPLEMENT_QSORT_EX( icvSortIndexedValArray_16s, short, CMP_VALUES, CvValArray* )
-
-static CV_IMPLEMENT_QSORT_EX( icvSortIndexedValArray_32s, int,   CMP_VALUES, CvValArray* )
-
-static CV_IMPLEMENT_QSORT_EX( icvSortIndexedValArray_32f, float, CMP_VALUES, CvValArray* )
+template<typename T, typename Idx>
+class LessThanValArray
+{
+public:
+    LessThanValArray( const T* _aux ) : aux(_aux) {}
+    bool operator()(Idx a, Idx b) const
+    {
+        return *( (float*) (aux->data + ((int) (a)) * aux->step ) ) <
+               *( (float*) (aux->data + ((int) (b)) * aux->step ) );
+    }
+    const T* aux;
+};
 
 CV_BOOST_IMPL
 void cvGetSortedIndices( CvMat* val, CvMat* idx, int sortcols )
@@ -130,8 +131,9 @@ void cvGetSortedIndices( CvMat* val, CvMat* idx, int sortcols )
                 {
                     CV_MAT_ELEM( *idx, short, i, j ) = (short) j;
                 }
-                icvSortIndexedValArray_16s( (short*) (idx->data.ptr + (size_t)i * idx->step),
-                                            idx->cols, &va );
+                std::sort((short*) (idx->data.ptr + (size_t)i * idx->step),
+                          (short*) (idx->data.ptr + (size_t)i * idx->step) + idx->cols,
+                          LessThanValArray<CvValArray, short>(&va));
                 va.data += istep;
             }
             break;
@@ -143,8 +145,9 @@ void cvGetSortedIndices( CvMat* val, CvMat* idx, int sortcols )
                 {
                     CV_MAT_ELEM( *idx, int, i, j ) = j;
                 }
-                icvSortIndexedValArray_32s( (int*) (idx->data.ptr + (size_t)i * idx->step),
-                                            idx->cols, &va );
+                std::sort((int*) (idx->data.ptr + (size_t)i * idx->step),
+                          (int*) (idx->data.ptr + (size_t)i * idx->step) + idx->cols,
+                          LessThanValArray<CvValArray, int>(&va));
                 va.data += istep;
             }
             break;
@@ -156,8 +159,9 @@ void cvGetSortedIndices( CvMat* val, CvMat* idx, int sortcols )
                 {
                     CV_MAT_ELEM( *idx, float, i, j ) = (float) j;
                 }
-                icvSortIndexedValArray_32f( (float*) (idx->data.ptr + (size_t)i * idx->step),
-                                            idx->cols, &va );
+                std::sort((float*) (idx->data.ptr + (size_t)i * idx->step),
+                          (float*) (idx->data.ptr + (size_t)i * idx->step) + idx->cols,
+                          LessThanValArray<CvValArray, float>(&va));
                 va.data += istep;
             }
             break;
@@ -545,7 +549,7 @@ CvClassifier* cvCreateStumpClassifier( CvMat* trainData,
 
         va.data = data + i * ((size_t) cstep);
         va.step = sstep;
-        icvSortIndexedValArray_32s( idx, l, &va );
+        std::sort(idx, idx + l, LessThanValArray<CvValArray, int>(&va));
         if( findStumpThreshold_32s[(int) ((CvStumpTrainParams*) trainParams)->error]
               ( data + i * ((size_t) cstep), sstep,
                 wdata, wstep, ydata, ystep, (uchar*) idx, sizeof( int ), l,
@@ -1028,7 +1032,7 @@ CvClassifier* cvCreateMTStumpClassifier( CvMat* trainData,
             {
                 va.data = t_data + ti * t_cstep;
                 va.step = t_sstep;
-                icvSortIndexedValArray_32s( t_idx, l, &va );
+                std::sort(t_idx, t_idx + l, LessThanValArray<CvValArray, int>(&va));
                 if( findStumpThreshold_32s[stumperror](
                         t_data + ti * t_cstep, t_sstep,
                         wdata, wstep, ydata, ystep,
@@ -2096,7 +2100,7 @@ static void icvZeroApproxMed( float* approx, CvBtTrainer* trainer )
         trainer->f[i] = *((float*) (trainer->ydata + idx * trainer->ystep));
     }
 
-    icvSort_32f( trainer->f, trainer->numsamples, 0 );
+    std::sort(trainer->f, trainer->f + trainer->numsamples);
     approx[0] = trainer->f[trainer->numsamples / 2];
 }
 
@@ -2341,7 +2345,7 @@ static void icvBtNext_LADREG( CvCARTClassifier** trees, CvBtTrainer* trainer )
         }
         if( respnum > 0 )
         {
-            icvSort_32f( resp, respnum, 0 );
+            std::sort(resp, resp + respnum);
             val = resp[respnum / 2];
         }
         else
@@ -2394,7 +2398,7 @@ static void icvBtNext_MREG( CvCARTClassifier** trees, CvBtTrainer* trainer )
     }
 
     /* delta = quantile_alpha{abs(resid_i)} */
-    icvSort_32f( resp, trainer->numsamples, 0 );
+    std::sort(resp, resp + trainer->numsamples);
     delta = resp[(int)(trainer->param[1] * (trainer->numsamples - 1))];
 
     /* yhat_i */
@@ -2434,7 +2438,7 @@ static void icvBtNext_MREG( CvCARTClassifier** trees, CvBtTrainer* trainer )
         if( respnum > 0 )
         {
             /* rhat = median(y_i - F_(m-1)(x_i)) */
-            icvSort_32f( resp, respnum, 0 );
+            std::sort(resp, resp + respnum);
             rhat = resp[respnum / 2];
 
             /* val = sum{sign(r_i - rhat_i) * min(delta, abs(r_i - rhat_i)}
@@ -2531,7 +2535,7 @@ static void icvBtNext_L2CLASS( CvCARTClassifier** trees, CvBtTrainer* trainer )
         float threshold;
         int count;
 
-        icvSort_32f( sorted_weights, trainer->numsamples, 0 );
+        std::sort(sorted_weights, sorted_weights + trainer->numsamples);
 
         sum_weights *= (1.0F - trainer->param[1]);
 
@@ -2693,7 +2697,7 @@ static void icvBtNext_LKCLASS( CvCARTClassifier** trees, CvBtTrainer* trainer )
             float threshold;
             int count;
 
-            icvSort_32f( sorted_weights, trainer->numsamples, 0 );
+            std::sort(sorted_weights, sorted_weights + trainer->numsamples);
 
             sum_weights *= (1.0F - trainer->param[1]);
 
@@ -3504,7 +3508,7 @@ CvMat* cvTrimWeights( CvMat* weights, CvMat* idx, float factor )
             sum_weights += sorted_weights[i];
         }
 
-        icvSort_32f( sorted_weights, num, 0 );
+        std::sort(sorted_weights, sorted_weights + num);
 
         sum_weights *= (1.0F - factor);
 
diff --git a/apps/haartraining/cvclassifier.h b/apps/haartraining/cvclassifier.h
index df644ed173..9a43441187 100644
--- a/apps/haartraining/cvclassifier.h
+++ b/apps/haartraining/cvclassifier.h
@@ -112,7 +112,9 @@ CV_INLINE float cvLogRatio( float val )
 /* each trainData matrix row is a sample */
 #define CV_ROW_SAMPLE 1
 
-#define CV_IS_ROW_SAMPLE( flags ) ( ( flags ) & CV_ROW_SAMPLE )
+#ifndef CV_IS_ROW_SAMPLE
+#  define CV_IS_ROW_SAMPLE( flags ) ( ( flags ) & CV_ROW_SAMPLE )
+#endif
 
 /* Classifier supports tune function */
 #define CV_TUNABLE    (1 << 1)
diff --git a/apps/haartraining/cvcommon.cpp b/apps/haartraining/cvcommon.cpp
index 013256b6d8..e23fab263b 100644
--- a/apps/haartraining/cvcommon.cpp
+++ b/apps/haartraining/cvcommon.cpp
@@ -50,11 +50,6 @@
 #include <direct.h>
 #endif /* _WIN32 */
 
-
-CV_IMPLEMENT_QSORT( icvSort_32f, float, less_than )
-
-CV_IMPLEMENT_QSORT( icvSort_32s, int, less_than )
-
 int icvMkDir( const char* filename )
 {
     char path[PATH_MAX];
diff --git a/apps/haartraining/cvhaartraining.cpp b/apps/haartraining/cvhaartraining.cpp
index 661bc959be..2fcf76a4b5 100644
--- a/apps/haartraining/cvhaartraining.cpp
+++ b/apps/haartraining/cvhaartraining.cpp
@@ -1088,7 +1088,7 @@ CvIntHaarClassifier* icvCreateCARTStageClassifier( CvHaarTrainingData* data,
                 numpos++;
             }
         }
-        icvSort_32f( eval.data.fl, numpos, 0 );
+        std::sort(eval.data.fl, eval.data.fl + numpos);
         threshold = eval.data.fl[(int) ((1.0F - minhitrate) * numpos)];
 
         numneg = 0;
@@ -2291,7 +2291,7 @@ static CvMat* icvGetUsedValues( CvHaarTrainingData* training_data,
             feature_idx->data.i[total++] = cart->compidx[j];
         }
     }
-    icvSort_32s( feature_idx->data.i, total, 0 );
+    std::sort(feature_idx->data.i, feature_idx->data.i + total);
 
     last = 0;
     for( i = 1; i < total; i++ )
@@ -2676,7 +2676,7 @@ void cvCreateTreeCascadeClassifier( const char* dirname,
 
                     CvSplit* curSplit;
                     CV_CALL( curSplit = (CvSplit*) cvAlloc( sizeof( *curSplit ) ) );
-                    CV_ZERO_OBJ( curSplit );
+                    memset(curSplit, 0, sizeof(*curSplit));
 
                     if( last_split ) last_split->next = curSplit;
                     else first_split = curSplit;
diff --git a/apps/haartraining/performance.cpp b/apps/haartraining/performance.cpp
index 0620f21a1b..cb8dda1c83 100644
--- a/apps/haartraining/performance.cpp
+++ b/apps/haartraining/performance.cpp
@@ -45,7 +45,6 @@
  * Measure performance of classifier
  */
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "cv.h"
 #include "highgui.h"
@@ -376,4 +375,3 @@ int main( int argc, char* argv[] )
 
     return 0;
 }
-
diff --git a/apps/sft/CMakeLists.txt b/apps/sft/CMakeLists.txt
index 8b950225c8..05bd337c34 100644
--- a/apps/sft/CMakeLists.txt
+++ b/apps/sft/CMakeLists.txt
@@ -30,4 +30,4 @@ if(ENABLE_SOLUTION_FOLDERS)
   set_target_properties(${the_target} PROPERTIES FOLDER "applications")
 endif()
 
-install(TARGETS ${the_target} RUNTIME DESTINATION bin COMPONENT main)
\ No newline at end of file
+install(TARGETS ${the_target} RUNTIME DESTINATION bin COMPONENT main)
diff --git a/apps/sft/config.cpp b/apps/sft/config.cpp
index 3cc64c7fec..9157575a1e 100644
--- a/apps/sft/config.cpp
+++ b/apps/sft/config.cpp
@@ -159,4 +159,4 @@ std::ostream& sft::operator<<(std::ostream& out, const Config& m)
         << std::setw(14) << std::left  << "featureType"  << m.featureType  << std::endl;
 
     return out;
-}
\ No newline at end of file
+}
diff --git a/apps/sft/include/sft/common.hpp b/apps/sft/include/sft/common.hpp
index 93439a48ed..5c142a749d 100644
--- a/apps/sft/include/sft/common.hpp
+++ b/apps/sft/include/sft/common.hpp
@@ -44,6 +44,7 @@
 #define __SFT_COMMON_HPP__
 
 #include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
 #include <opencv2/softcascade.hpp>
 
 namespace cv {using namespace softcascade;}
@@ -53,7 +54,7 @@ namespace sft
     using cv::Mat;
     struct ICF;
 
-    typedef std::string   string;
+    typedef cv::String   string;
 
     typedef std::vector<ICF>           Icfvector;
     typedef std::vector<sft::string>   svector;
@@ -70,4 +71,4 @@ namespace sft
 # define dprintf(format, ...)
 #endif
 
-#endif
\ No newline at end of file
+#endif
diff --git a/apps/sft/include/sft/config.hpp b/apps/sft/include/sft/config.hpp
index 3d39d32722..c6e85b264c 100644
--- a/apps/sft/include/sft/config.hpp
+++ b/apps/sft/include/sft/config.hpp
@@ -135,4 +135,4 @@ std::ostream& operator<<(std::ostream& out, const Config& m);
 
 }
 
-#endif
\ No newline at end of file
+#endif
diff --git a/apps/sft/include/sft/dataset.hpp b/apps/sft/include/sft/dataset.hpp
index 98d6192568..7504f4033d 100644
--- a/apps/sft/include/sft/dataset.hpp
+++ b/apps/sft/include/sft/dataset.hpp
@@ -64,4 +64,4 @@ private:
 };
 }
 
-#endif
\ No newline at end of file
+#endif
diff --git a/apps/sft/sft.cpp b/apps/sft/sft.cpp
index 7b51387638..79d41032eb 100644
--- a/apps/sft/sft.cpp
+++ b/apps/sft/sft.cpp
@@ -165,4 +165,4 @@ int main(int argc, char** argv)
     fso.release();
     std::cout << "Training complete..." << std::endl;
     return 0;
-}
\ No newline at end of file
+}
diff --git a/apps/traincascade/CMakeLists.txt b/apps/traincascade/CMakeLists.txt
index f6fa679a9a..eb0c83df5a 100644
--- a/apps/traincascade/CMakeLists.txt
+++ b/apps/traincascade/CMakeLists.txt
@@ -1,4 +1,4 @@
-set(OPENCV_TRAINCASCADE_DEPS opencv_core opencv_ml opencv_imgproc opencv_objdetect opencv_highgui opencv_calib3d opencv_video opencv_features2d opencv_flann opencv_legacy)
+set(OPENCV_TRAINCASCADE_DEPS opencv_core opencv_ml opencv_imgproc opencv_photo opencv_objdetect opencv_highgui opencv_calib3d opencv_video opencv_features2d opencv_flann opencv_legacy)
 ocv_check_dependencies(${OPENCV_TRAINCASCADE_DEPS})
 
 if(NOT OCV_DEPENDENCIES_FOUND)
@@ -34,4 +34,3 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()
 
 install(TARGETS ${the_target} RUNTIME DESTINATION bin COMPONENT main)
-
diff --git a/apps/traincascade/HOGfeatures.cpp b/apps/traincascade/HOGfeatures.cpp
index eaf4bd7960..132e2e62e9 100644
--- a/apps/traincascade/HOGfeatures.cpp
+++ b/apps/traincascade/HOGfeatures.cpp
@@ -1,5 +1,5 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include "HOGfeatures.h"
 #include "cascadeclassifier.h"
diff --git a/apps/traincascade/boost.cpp b/apps/traincascade/boost.cpp
index 1b0f390a5b..732704a4a2 100644
--- a/apps/traincascade/boost.cpp
+++ b/apps/traincascade/boost.cpp
@@ -1,11 +1,52 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "boost.h"
 #include "cascadeclassifier.h"
 #include <queue>
 #include "cxmisc.h"
 
+#include "cvconfig.h"
+#ifdef HAVE_TBB
+#  include "tbb/tbb_stddef.h"
+#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#    include "tbb/tbb.h"
+#    include "tbb/task.h"
+#    undef min
+#    undef max
+#  else
+#    undef HAVE_TBB
+#  endif
+#endif
+
+#ifdef HAVE_TBB
+    typedef tbb::blocked_range<int> BlockedRange;
+
+    template<typename Body> static inline
+    void parallel_for( const BlockedRange& range, const Body& body )
+    {
+        tbb::parallel_for(range, body);
+    }
+#else
+    class BlockedRange
+    {
+    public:
+        BlockedRange() : _begin(0), _end(0), _grainsize(0) {}
+        BlockedRange(int b, int e, int g=1) : _begin(b), _end(e), _grainsize(g) {}
+        int begin() const { return _begin; }
+        int end() const { return _end; }
+        int grainsize() const { return _grainsize; }
+
+    protected:
+        int _begin, _end, _grainsize;
+    };
+
+    template<typename Body> static inline
+    void parallel_for( const BlockedRange& range, const Body& body )
+    {
+        body(range);
+    }
+#endif
+
 using namespace std;
 
 static inline double
@@ -18,12 +59,20 @@ logRatio( double val )
     return log( val/(1. - val) );
 }
 
-#define CV_CMP_FLT(i,j) (i < j)
-static CV_IMPLEMENT_QSORT_EX( icvSortFlt, float, CV_CMP_FLT, const float* )
+template<typename T, typename Idx>
+class LessThanIdx
+{
+public:
+    LessThanIdx( const T* _arr ) : arr(_arr) {}
+    bool operator()(Idx a, Idx b) const { return arr[a] < arr[b]; }
+    const T* arr;
+};
 
-#define CV_CMP_NUM_IDX(i,j) (aux[i] < aux[j])
-static CV_IMPLEMENT_QSORT_EX( icvSortIntAux, int, CV_CMP_NUM_IDX, const float* )
-static CV_IMPLEMENT_QSORT_EX( icvSortUShAux, unsigned short, CV_CMP_NUM_IDX, const float* )
+static inline int cvAlign( int size, int align )
+{
+    CV_DbgAssert( (align & (align-1)) == 0 && size < INT_MAX );
+    return (size + align - 1) & -align;
+}
 
 #define CV_THRESHOLD_EPS (0.00001F)
 
@@ -43,7 +92,7 @@ static CvMat* cvPreprocessIndexArray( const CvMat* idx_arr, int data_arr_size, b
 
     CV_FUNCNAME( "cvPreprocessIndexArray" );
 
-    __BEGIN__;
+    __CV_BEGIN__;
 
     int i, idx_total, idx_selected = 0, step, type, prev = INT_MIN, is_sorted = 1;
     uchar* srcb = 0;
@@ -131,7 +180,7 @@ static CvMat* cvPreprocessIndexArray( const CvMat* idx_arr, int data_arr_size, b
         }
     }
 
-    __END__;
+    __CV_END__;
 
     if( cvGetErrStatus() < 0 )
         cvReleaseMat( &idx );
@@ -722,7 +771,7 @@ void CvCascadeBoostTrainData::get_ord_var_data( CvDTreeNode* n, int vi, float* o
                 sampleValues[i] = (*featureEvaluator)( vi, sampleIndices[i]);
             }
         }
-        icvSortIntAux( sortedIndicesBuf, nodeSampleCount, &sampleValues[0] );
+        std::sort(sortedIndicesBuf, sortedIndicesBuf + nodeSampleCount, LessThanIdx<float, int>(&sampleValues[0]) );
         for( int i = 0; i < nodeSampleCount; i++ )
             ordValuesBuf[i] = (&sampleValues[0])[sortedIndicesBuf[i]];
         *sortedIndices = sortedIndicesBuf;
@@ -766,7 +815,7 @@ float CvCascadeBoostTrainData::getVarValue( int vi, int si )
 }
 
 
-struct FeatureIdxOnlyPrecalc
+struct FeatureIdxOnlyPrecalc : ParallelLoopBody
 {
     FeatureIdxOnlyPrecalc( const CvFeatureEvaluator* _featureEvaluator, CvMat* _buf, int _sample_count, bool _is_buf_16u )
     {
@@ -776,11 +825,11 @@ struct FeatureIdxOnlyPrecalc
         idst = _buf->data.i;
         is_buf_16u = _is_buf_16u;
     }
-    void operator()( const BlockedRange& range ) const
+    void operator()( const Range& range ) const
     {
         cv::AutoBuffer<float> valCache(sample_count);
         float* valCachePtr = (float*)valCache;
-        for ( int fi = range.begin(); fi < range.end(); fi++)
+        for ( int fi = range.start; fi < range.end; fi++)
         {
             for( int si = 0; si < sample_count; si++ )
             {
@@ -791,9 +840,9 @@ struct FeatureIdxOnlyPrecalc
                     *(idst + fi*sample_count + si) = si;
             }
             if ( is_buf_16u )
-                icvSortUShAux( udst + fi*sample_count, sample_count, valCachePtr );
+                std::sort(udst + fi*sample_count, udst + (fi + 1)*sample_count, LessThanIdx<float, unsigned short>(valCachePtr) );
             else
-                icvSortIntAux( idst + fi*sample_count, sample_count, valCachePtr );
+                std::sort(idst + fi*sample_count, idst + (fi + 1)*sample_count, LessThanIdx<float, int>(valCachePtr) );
         }
     }
     const CvFeatureEvaluator* featureEvaluator;
@@ -803,7 +852,7 @@ struct FeatureIdxOnlyPrecalc
     bool is_buf_16u;
 };
 
-struct FeatureValAndIdxPrecalc
+struct FeatureValAndIdxPrecalc : ParallelLoopBody
 {
     FeatureValAndIdxPrecalc( const CvFeatureEvaluator* _featureEvaluator, CvMat* _buf, Mat* _valCache, int _sample_count, bool _is_buf_16u )
     {
@@ -814,9 +863,9 @@ struct FeatureValAndIdxPrecalc
         idst = _buf->data.i;
         is_buf_16u = _is_buf_16u;
     }
-    void operator()( const BlockedRange& range ) const
+    void operator()( const Range& range ) const
     {
-        for ( int fi = range.begin(); fi < range.end(); fi++)
+        for ( int fi = range.start; fi < range.end; fi++)
         {
             for( int si = 0; si < sample_count; si++ )
             {
@@ -827,9 +876,9 @@ struct FeatureValAndIdxPrecalc
                     *(idst + fi*sample_count + si) = si;
             }
             if ( is_buf_16u )
-                icvSortUShAux( udst + fi*sample_count, sample_count, valCache->ptr<float>(fi) );
+                std::sort(idst + fi*sample_count, idst + (fi + 1)*sample_count, LessThanIdx<float, unsigned short>(valCache->ptr<float>(fi)) );
             else
-                icvSortIntAux( idst + fi*sample_count, sample_count, valCache->ptr<float>(fi) );
+                std::sort(idst + fi*sample_count, idst + (fi + 1)*sample_count, LessThanIdx<float, int>(valCache->ptr<float>(fi)) );
         }
     }
     const CvFeatureEvaluator* featureEvaluator;
@@ -840,7 +889,7 @@ struct FeatureValAndIdxPrecalc
     bool is_buf_16u;
 };
 
-struct FeatureValOnlyPrecalc
+struct FeatureValOnlyPrecalc : ParallelLoopBody
 {
     FeatureValOnlyPrecalc( const CvFeatureEvaluator* _featureEvaluator, Mat* _valCache, int _sample_count )
     {
@@ -848,9 +897,9 @@ struct FeatureValOnlyPrecalc
         valCache = _valCache;
         sample_count = _sample_count;
     }
-    void operator()( const BlockedRange& range ) const
+    void operator()( const Range& range ) const
     {
-        for ( int fi = range.begin(); fi < range.end(); fi++)
+        for ( int fi = range.start; fi < range.end; fi++)
             for( int si = 0; si < sample_count; si++ )
                 valCache->at<float>(fi,si) = (*featureEvaluator)( fi, si );
     }
@@ -864,12 +913,12 @@ void CvCascadeBoostTrainData::precalculate()
     int minNum = MIN( numPrecalcVal, numPrecalcIdx);
 
     double proctime = -TIME( 0 );
-    parallel_for( BlockedRange(numPrecalcVal, numPrecalcIdx),
-                  FeatureIdxOnlyPrecalc(featureEvaluator, buf, sample_count, is_buf_16u!=0) );
-    parallel_for( BlockedRange(0, minNum),
-                  FeatureValAndIdxPrecalc(featureEvaluator, buf, &valCache, sample_count, is_buf_16u!=0) );
-    parallel_for( BlockedRange(minNum, numPrecalcVal),
-                  FeatureValOnlyPrecalc(featureEvaluator, &valCache, sample_count) );
+    parallel_for_( Range(numPrecalcVal, numPrecalcIdx),
+                   FeatureIdxOnlyPrecalc(featureEvaluator, buf, sample_count, is_buf_16u!=0) );
+    parallel_for_( Range(0, minNum),
+                   FeatureValAndIdxPrecalc(featureEvaluator, buf, &valCache, sample_count, is_buf_16u!=0) );
+    parallel_for_( Range(minNum, numPrecalcVal),
+                   FeatureValOnlyPrecalc(featureEvaluator, &valCache, sample_count) );
     cout << "Precalculation time: " << (proctime + TIME( 0 )) << endl;
 }
 
@@ -908,7 +957,7 @@ void CvCascadeBoostTree::write( FileStorage &fs, const Mat& featureMap )
     int subsetN = (maxCatCount + 31)/32;
     queue<CvDTreeNode*> internalNodesQueue;
     int size = (int)pow( 2.f, (float)ensemble->get_params().max_depth);
-    Ptr<float> leafVals = new float[size];
+    std::vector<float> leafVals(size);
     int leafValIdx = 0;
     int internalNodeIdx = 1;
     CvDTreeNode* tempNode;
@@ -1602,7 +1651,7 @@ bool CvCascadeBoost::isErrDesired()
         if( ((CvCascadeBoostTrainData*)data)->featureEvaluator->getCls( i ) == 1.0F )
             eval[numPos++] = predict( i, true );
 
-    icvSortFlt( &eval[0], numPos, 0 );
+    std::sort(&eval[0], &eval[0] + numPos);
 
     int thresholdIdx = (int)((1.0F - minHitRate) * numPos);
 
diff --git a/apps/traincascade/cascadeclassifier.cpp b/apps/traincascade/cascadeclassifier.cpp
index f6752da0c9..5c96b45f76 100644
--- a/apps/traincascade/cascadeclassifier.cpp
+++ b/apps/traincascade/cascadeclassifier.cpp
@@ -1,5 +1,4 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "cascadeclassifier.h"
 #include <queue>
@@ -160,10 +159,10 @@ bool CvCascadeClassifier::train( const string _cascadeDirName,
         cascadeParams = _cascadeParams;
         featureParams = CvFeatureParams::create(cascadeParams.featureType);
         featureParams->init(_featureParams);
-        stageParams = new CvCascadeBoostParams;
+        stageParams = makePtr<CvCascadeBoostParams>();
         *stageParams = _stageParams;
         featureEvaluator = CvFeatureEvaluator::create(cascadeParams.featureType);
-        featureEvaluator->init( (CvFeatureParams*)featureParams, numPos + numNeg, cascadeParams.winSize );
+        featureEvaluator->init( featureParams, numPos + numNeg, cascadeParams.winSize );
         stageClassifiers.reserve( numStages );
     }
     cout << "PARAMETERS:" << endl;
@@ -207,10 +206,10 @@ bool CvCascadeClassifier::train( const string _cascadeDirName,
             break;
         }
 
-        CvCascadeBoost* tempStage = new CvCascadeBoost;
-        bool isStageTrained = tempStage->train( (CvFeatureEvaluator*)featureEvaluator,
+        Ptr<CvCascadeBoost> tempStage = makePtr<CvCascadeBoost>();
+        bool isStageTrained = tempStage->train( featureEvaluator,
                                                 curNumSamples, _precalcValBufSize, _precalcIdxBufSize,
-                                                *((CvCascadeBoostParams*)stageParams) );
+                                                *stageParams );
         cout << "END>" << endl;
 
         if(!isStageTrained)
@@ -326,7 +325,7 @@ void CvCascadeClassifier::writeParams( FileStorage &fs ) const
 
 void CvCascadeClassifier::writeFeatures( FileStorage &fs, const Mat& featureMap ) const
 {
-    ((CvFeatureEvaluator*)((Ptr<CvFeatureEvaluator>)featureEvaluator))->writeFeatures( fs, featureMap );
+    featureEvaluator->writeFeatures( fs, featureMap );
 }
 
 void CvCascadeClassifier::writeStages( FileStorage &fs, const Mat& featureMap ) const
@@ -340,7 +339,7 @@ void CvCascadeClassifier::writeStages( FileStorage &fs, const Mat& featureMap )
         sprintf( cmnt, "stage %d", i );
         cvWriteComment( fs.fs, cmnt, 0 );
         fs << "{";
-        ((CvCascadeBoost*)((Ptr<CvCascadeBoost>)*it))->write( fs, featureMap );
+        (*it)->write( fs, featureMap );
         fs << "}";
     }
     fs << "]";
@@ -351,7 +350,7 @@ bool CvCascadeClassifier::readParams( const FileNode &node )
     if ( !node.isMap() || !cascadeParams.read( node ) )
         return false;
 
-    stageParams = new CvCascadeBoostParams;
+    stageParams = makePtr<CvCascadeBoostParams>();
     FileNode rnode = node[CC_STAGE_PARAMS];
     if ( !stageParams->read( rnode ) )
         return false;
@@ -372,12 +371,9 @@ bool CvCascadeClassifier::readStages( const FileNode &node)
     FileNodeIterator it = rnode.begin();
     for( int i = 0; i < min( (int)rnode.size(), numStages ); i++, it++ )
     {
-        CvCascadeBoost* tempStage = new CvCascadeBoost;
-        if ( !tempStage->read( *it, (CvFeatureEvaluator *)featureEvaluator, *((CvCascadeBoostParams*)stageParams) ) )
-        {
-            delete tempStage;
+        Ptr<CvCascadeBoost> tempStage = makePtr<CvCascadeBoost>();
+        if ( !tempStage->read( *it, featureEvaluator, *stageParams) )
             return false;
-        }
         stageClassifiers.push_back(tempStage);
     }
     return true;
@@ -454,7 +450,7 @@ void CvCascadeClassifier::save( const string filename, bool baseFormat )
 
                     fs << "{";
                     fs << ICV_HAAR_FEATURE_NAME << "{";
-                    ((CvHaarEvaluator*)((CvFeatureEvaluator*)featureEvaluator))->writeFeature( fs, tempNode->split->var_idx );
+                    ((CvHaarEvaluator*)featureEvaluator.get())->writeFeature( fs, tempNode->split->var_idx );
                     fs << "}";
 
                     fs << ICV_HAAR_THRESHOLD_NAME << tempNode->split->ord.c;
@@ -500,7 +496,7 @@ bool CvCascadeClassifier::load( const string cascadeDirName )
     if ( !readParams( node ) )
         return false;
     featureEvaluator = CvFeatureEvaluator::create(cascadeParams.featureType);
-    featureEvaluator->init( ((CvFeatureParams*)featureParams), numPos + numNeg, cascadeParams.winSize );
+    featureEvaluator->init( featureParams, numPos + numNeg, cascadeParams.winSize );
     fs.release();
 
     char buf[10];
@@ -511,11 +507,10 @@ bool CvCascadeClassifier::load( const string cascadeDirName )
         node = fs.getFirstTopLevelNode();
         if ( !fs.isOpened() )
             break;
-        CvCascadeBoost *tempStage = new CvCascadeBoost;
+        Ptr<CvCascadeBoost> tempStage = makePtr<CvCascadeBoost>();
 
-        if ( !tempStage->read( node, (CvFeatureEvaluator*)featureEvaluator, *((CvCascadeBoostParams*)stageParams )) )
+        if ( !tempStage->read( node, featureEvaluator, *stageParams ))
         {
-            delete tempStage;
             fs.release();
             break;
         }
@@ -532,7 +527,7 @@ void CvCascadeClassifier::getUsedFeaturesIdxMap( Mat& featureMap )
 
     for( vector< Ptr<CvCascadeBoost> >::const_iterator it = stageClassifiers.begin();
         it != stageClassifiers.end(); it++ )
-        ((CvCascadeBoost*)((Ptr<CvCascadeBoost>)(*it)))->markUsedFeaturesInMap( featureMap );
+        (*it)->markUsedFeaturesInMap( featureMap );
 
     for( int fi = 0, idx = 0; fi < varCount; fi++ )
         if ( featureMap.at<int>(0, fi) >= 0 )
diff --git a/apps/traincascade/features.cpp b/apps/traincascade/features.cpp
index 4aebc80940..7f0c0a7079 100644
--- a/apps/traincascade/features.cpp
+++ b/apps/traincascade/features.cpp
@@ -1,5 +1,4 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "traincascade_features.h"
 #include "cascadeclassifier.h"
diff --git a/apps/traincascade/haarfeatures.cpp b/apps/traincascade/haarfeatures.cpp
index 59f1fd4ca7..ceb730aafb 100644
--- a/apps/traincascade/haarfeatures.cpp
+++ b/apps/traincascade/haarfeatures.cpp
@@ -1,5 +1,5 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include "haarfeatures.h"
 #include "cascadeclassifier.h"
diff --git a/apps/traincascade/imagestorage.cpp b/apps/traincascade/imagestorage.cpp
index e7922a9afc..a8426e074d 100644
--- a/apps/traincascade/imagestorage.cpp
+++ b/apps/traincascade/imagestorage.cpp
@@ -1,7 +1,7 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
-#include "cv.h"
 #include "imagestorage.h"
 #include <stdio.h>
 #include <iostream>
diff --git a/apps/traincascade/lbpfeatures.cpp b/apps/traincascade/lbpfeatures.cpp
index 5e23d9e461..dd1659f9de 100644
--- a/apps/traincascade/lbpfeatures.cpp
+++ b/apps/traincascade/lbpfeatures.cpp
@@ -1,5 +1,5 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include "lbpfeatures.h"
 #include "cascadeclassifier.h"
diff --git a/apps/traincascade/traincascade.cpp b/apps/traincascade/traincascade.cpp
index c85046e6b7..a896c216c9 100644
--- a/apps/traincascade/traincascade.cpp
+++ b/apps/traincascade/traincascade.cpp
@@ -1,5 +1,4 @@
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
 
 #include "cv.h"
 #include "cascadeclassifier.h"
@@ -19,9 +18,9 @@ int main( int argc, char* argv[] )
 
     CvCascadeParams cascadeParams;
     CvCascadeBoostParams stageParams;
-    Ptr<CvFeatureParams> featureParams[] = { Ptr<CvFeatureParams>(new CvHaarFeatureParams),
-                                             Ptr<CvFeatureParams>(new CvLBPFeatureParams),
-                                             Ptr<CvFeatureParams>(new CvHOGFeatureParams)
+    Ptr<CvFeatureParams> featureParams[] = { makePtr<CvHaarFeatureParams>(),
+                                             makePtr<CvLBPFeatureParams>(),
+                                             makePtr<CvHOGFeatureParams>()
                                            };
     int fc = sizeof(featureParams)/sizeof(featureParams[0]);
     if( argc == 1 )
diff --git a/cmake/CMakeParseArguments.cmake b/cmake/CMakeParseArguments.cmake
deleted file mode 100644
index 7ce4c49ae5..0000000000
--- a/cmake/CMakeParseArguments.cmake
+++ /dev/null
@@ -1,138 +0,0 @@
-# CMAKE_PARSE_ARGUMENTS(<prefix> <options> <one_value_keywords> <multi_value_keywords> args...)
-#
-# CMAKE_PARSE_ARGUMENTS() is intended to be used in macros or functions for
-# parsing the arguments given to that macro or function.
-# It processes the arguments and defines a set of variables which hold the
-# values of the respective options.
-#
-# The <options> argument contains all options for the respective macro,
-# i.e. keywords which can be used when calling the macro without any value
-# following, like e.g. the OPTIONAL keyword of the install() command.
-#
-# The <one_value_keywords> argument contains all keywords for this macro
-# which are followed by one value, like e.g. DESTINATION keyword of the
-# install() command.
-#
-# The <multi_value_keywords> argument contains all keywords for this macro
-# which can be followed by more than one value, like e.g. the TARGETS or
-# FILES keywords of the install() command.
-#
-# When done, CMAKE_PARSE_ARGUMENTS() will have defined for each of the
-# keywords listed in <options>, <one_value_keywords> and
-# <multi_value_keywords> a variable composed of the given <prefix>
-# followed by "_" and the name of the respective keyword.
-# These variables will then hold the respective value from the argument list.
-# For the <options> keywords this will be TRUE or FALSE.
-#
-# All remaining arguments are collected in a variable
-# <prefix>_UNPARSED_ARGUMENTS, this can be checked afterwards to see whether
-# your macro was called with unrecognized parameters.
-#
-# As an example here a my_install() macro, which takes similar arguments as the
-# real install() command:
-#
-#   function(MY_INSTALL)
-#     set(options OPTIONAL FAST)
-#     set(oneValueArgs DESTINATION RENAME)
-#     set(multiValueArgs TARGETS CONFIGURATIONS)
-#     cmake_parse_arguments(MY_INSTALL "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN} )
-#     ...
-#
-# Assume my_install() has been called like this:
-#   my_install(TARGETS foo bar DESTINATION bin OPTIONAL blub)
-#
-# After the cmake_parse_arguments() call the macro will have set the following
-# variables:
-#   MY_INSTALL_OPTIONAL = TRUE
-#   MY_INSTALL_FAST = FALSE (this option was not used when calling my_install()
-#   MY_INSTALL_DESTINATION = "bin"
-#   MY_INSTALL_RENAME = "" (was not used)
-#   MY_INSTALL_TARGETS = "foo;bar"
-#   MY_INSTALL_CONFIGURATIONS = "" (was not used)
-#   MY_INSTALL_UNPARSED_ARGUMENTS = "blub" (no value expected after "OPTIONAL"
-#
-# You can the continue and process these variables.
-#
-# Keywords terminate lists of values, e.g. if directly after a one_value_keyword
-# another recognized keyword follows, this is interpreted as the beginning of
-# the new option.
-# E.g. my_install(TARGETS foo DESTINATION OPTIONAL) would result in
-# MY_INSTALL_DESTINATION set to "OPTIONAL", but MY_INSTALL_DESTINATION would
-# be empty and MY_INSTALL_OPTIONAL would be set to TRUE therefor.
-
-#=============================================================================
-# Copyright 2010 Alexander Neundorf <neundorf@kde.org>
-#
-# Distributed under the OSI-approved BSD License (the "License");
-# see accompanying file Copyright.txt for details.
-#
-# This software is distributed WITHOUT ANY WARRANTY; without even the
-# implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
-# See the License for more information.
-#=============================================================================
-# (To distribute this file outside of CMake, substitute the full
-#  License text for the above reference.)
-
-
-if(__CMAKE_PARSE_ARGUMENTS_INCLUDED)
-  return()
-endif()
-set(__CMAKE_PARSE_ARGUMENTS_INCLUDED TRUE)
-
-
-function(CMAKE_PARSE_ARGUMENTS prefix _optionNames _singleArgNames _multiArgNames)
-  # first set all result variables to empty/FALSE
-  foreach(arg_name ${_singleArgNames} ${_multiArgNames})
-    set(${prefix}_${arg_name})
-  endforeach(arg_name)
-
-  foreach(option ${_optionNames})
-    set(${prefix}_${option} FALSE)
-  endforeach(option)
-
-  set(${prefix}_UNPARSED_ARGUMENTS)
-
-  set(insideValues FALSE)
-  set(currentArgName)
-
-  # now iterate over all arguments and fill the result variables
-  foreach(currentArg ${ARGN})
-    list(FIND _optionNames "${currentArg}" optionIndex)  # ... then this marks the end of the arguments belonging to this keyword
-    list(FIND _singleArgNames "${currentArg}" singleArgIndex)  # ... then this marks the end of the arguments belonging to this keyword
-    list(FIND _multiArgNames "${currentArg}" multiArgIndex)  # ... then this marks the end of the arguments belonging to this keyword
-
-    if(${optionIndex} EQUAL -1  AND  ${singleArgIndex} EQUAL -1  AND  ${multiArgIndex} EQUAL -1)
-      if(insideValues)
-        if("${insideValues}" STREQUAL "SINGLE")
-          set(${prefix}_${currentArgName} ${currentArg})
-          set(insideValues FALSE)
-        elseif("${insideValues}" STREQUAL "MULTI")
-          list(APPEND ${prefix}_${currentArgName} ${currentArg})
-        endif()
-      else(insideValues)
-        list(APPEND ${prefix}_UNPARSED_ARGUMENTS ${currentArg})
-      endif(insideValues)
-    else()
-      if(NOT ${optionIndex} EQUAL -1)
-        set(${prefix}_${currentArg} TRUE)
-        set(insideValues FALSE)
-      elseif(NOT ${singleArgIndex} EQUAL -1)
-        set(currentArgName ${currentArg})
-        set(${prefix}_${currentArgName})
-        set(insideValues "SINGLE")
-      elseif(NOT ${multiArgIndex} EQUAL -1)
-        set(currentArgName ${currentArg})
-        set(${prefix}_${currentArgName})
-        set(insideValues "MULTI")
-      endif()
-    endif()
-
-  endforeach(currentArg)
-
-  # propagate the result variables to the caller:
-  foreach(arg_name ${_singleArgNames} ${_multiArgNames} ${_optionNames})
-    set(${prefix}_${arg_name}  ${${prefix}_${arg_name}} PARENT_SCOPE)
-  endforeach(arg_name)
-  set(${prefix}_UNPARSED_ARGUMENTS ${${prefix}_UNPARSED_ARGUMENTS} PARENT_SCOPE)
-
-endfunction(CMAKE_PARSE_ARGUMENTS _options _singleArgs _multiArgs)
diff --git a/cmake/OpenCVCRTLinkage.cmake b/cmake/OpenCVCRTLinkage.cmake
index 7514285d9f..8a297c6857 100644
--- a/cmake/OpenCVCRTLinkage.cmake
+++ b/cmake/OpenCVCRTLinkage.cmake
@@ -2,6 +2,45 @@ if(NOT MSVC)
   message(FATAL_ERROR "CRT options are available only for MSVC")
 endif()
 
+#INCLUDE (CheckIncludeFiles)
+
+set(HAVE_WINRT FALSE)
+
+# search Windows Platform SDK
+message(STATUS "Checking for Windows Platform SDK")
+GET_FILENAME_COMPONENT(WINDOWS_SDK_PATH  "[HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Microsoft SDKs\\Windows\\v8.0;InstallationFolder]" ABSOLUTE CACHE)
+if (WINDOWS_SDK_PATH STREQUAL "")
+  set(HAVE_MSPDK FALSE)
+  message(STATUS "Windows Platform SDK 8.0 was not found")
+else()
+  set(HAVE_MSPDK TRUE)
+endif()
+
+#search for Visual Studio 11.0 install directory
+message(STATUS "Checking for Visual Studio 2012")
+GET_FILENAME_COMPONENT(VISUAL_STUDIO_PATH [HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\VisualStudio\\11.0\\Setup\\VS;ProductDir] REALPATH CACHE)
+if (VISUAL_STUDIO_PATH STREQUAL "")
+  set(HAVE_MSVC2012 FALSE)
+  message(STATUS "Visual Studio 2012 was not found")
+else()
+  set(HAVE_MSVC2012 TRUE)
+endif()
+
+try_compile(HAVE_WINRT_SDK
+  "${OpenCV_BINARY_DIR}"
+  "${OpenCV_SOURCE_DIR}/cmake/checks/winrttest.cpp")
+
+if (ENABLE_WINRT_MODE AND HAVE_WINRT_SDK AND HAVE_MSVC2012 AND HAVE_MSPDK)
+  set(HAVE_WINRT TRUE)
+endif()
+
+if (HAVE_WINRT)
+  add_definitions(/DWINVER=0x0602 /DNTDDI_VERSION=NTDDI_WIN8 /D_WIN32_WINNT=0x0602)
+  set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /appcontainer")
+  set(CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS} /appcontainer")
+  set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} /appcontainer")
+endif()
+
 if(NOT BUILD_SHARED_LIBS AND BUILD_WITH_STATIC_CRT)
   foreach(flag_var
           CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE
@@ -62,4 +101,3 @@ if(NOT BUILD_WITH_DEBUG_INFO AND NOT MSVC)
   string(REPLACE "/Zi" "" CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG}")
   string(REPLACE "/Zi" "" CMAKE_CXX_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG}")
 endif()
-
diff --git a/cmake/OpenCVCompilerOptions.cmake b/cmake/OpenCVCompilerOptions.cmake
index a9baa9780d..bfca6a0bc9 100644
--- a/cmake/OpenCVCompilerOptions.cmake
+++ b/cmake/OpenCVCompilerOptions.cmake
@@ -47,6 +47,9 @@ macro(add_extra_compiler_option option)
   endif()
 endmacro()
 
+# OpenCV fails some tests when 'char' is 'unsigned' by default
+add_extra_compiler_option(-fsigned-char)
+
 if(MINGW)
   # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838
   # here we are trying to workaround the problem
@@ -230,6 +233,10 @@ if(MSVC)
       set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /fp:fast") # !! important - be on the same wave with x64 compilers
     endif()
   endif()
+
+  if(OPENCV_WARNINGS_ARE_ERRORS)
+    set(OPENCV_EXTRA_FLAGS "${OPENCV_EXTRA_FLAGS} /WX")
+  endif()
 endif()
 
 # Extra link libs if the user selects building static libs:
@@ -250,6 +257,12 @@ set(OPENCV_EXTRA_EXE_LINKER_FLAGS         "${OPENCV_EXTRA_EXE_LINKER_FLAGS}"
 set(OPENCV_EXTRA_EXE_LINKER_FLAGS_RELEASE "${OPENCV_EXTRA_EXE_LINKER_FLAGS_RELEASE}" CACHE INTERNAL "Extra linker flags for Release build")
 set(OPENCV_EXTRA_EXE_LINKER_FLAGS_DEBUG   "${OPENCV_EXTRA_EXE_LINKER_FLAGS_DEBUG}"   CACHE INTERNAL "Extra linker flags for Debug build")
 
+# set default visibility to hidden
+if(CMAKE_COMPILER_IS_GNUCXX AND CMAKE_OPENCV_GCC_VERSION_NUM GREATER 399)
+  add_extra_compiler_option(-fvisibility=hidden)
+  add_extra_compiler_option(-fvisibility-inlines-hidden)
+endif()
+
 #combine all "extra" options
 set(CMAKE_C_FLAGS           "${CMAKE_C_FLAGS} ${OPENCV_EXTRA_FLAGS} ${OPENCV_EXTRA_C_FLAGS}")
 set(CMAKE_CXX_FLAGS         "${CMAKE_CXX_FLAGS} ${OPENCV_EXTRA_FLAGS} ${OPENCV_EXTRA_CXX_FLAGS}")
@@ -285,4 +298,4 @@ if(MSVC)
   if(NOT ENABLE_NOISY_WARNINGS)
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /wd4251") #class 'std::XXX' needs to have dll-interface to be used by clients of YYY
   endif()
-endif()
\ No newline at end of file
+endif()
diff --git a/cmake/OpenCVConfig.cmake b/cmake/OpenCVConfig.cmake
deleted file mode 100644
index 300fcd4193..0000000000
--- a/cmake/OpenCVConfig.cmake
+++ /dev/null
@@ -1,159 +0,0 @@
-# ===================================================================================
-#  The OpenCV CMake configuration file
-#
-#             ** File generated automatically, do not modify **
-#
-#  Usage from an external project:
-#    In your CMakeLists.txt, add these lines:
-#
-#    FIND_PACKAGE(OpenCV REQUIRED)
-#    TARGET_LINK_LIBRARIES(MY_TARGET_NAME ${OpenCV_LIBS})
-#
-#    Or you can search for specific OpenCV modules:
-#
-#    FIND_PACKAGE(OpenCV REQUIRED core highgui)
-#
-#    If the module is found then OPENCV_<MODULE>_FOUND is set to TRUE.
-#
-#    This file will define the following variables:
-#      - OpenCV_LIBS                     : The list of libraries to links against.
-#      - OpenCV_LIB_DIR                  : The directory(es) where lib files are. Calling LINK_DIRECTORIES
-#                                          with this path is NOT needed.
-#      - OpenCV_INCLUDE_DIRS             : The OpenCV include directories.
-#      - OpenCV_COMPUTE_CAPABILITIES     : The version of compute capability
-#      - OpenCV_ANDROID_NATIVE_API_LEVEL : Minimum required level of Android API
-#      - OpenCV_VERSION                  : The version of this OpenCV build. Example: "2.4.0"
-#      - OpenCV_VERSION_MAJOR            : Major version part of OpenCV_VERSION. Example: "2"
-#      - OpenCV_VERSION_MINOR            : Minor version part of OpenCV_VERSION. Example: "4"
-#      - OpenCV_VERSION_PATCH            : Patch version part of OpenCV_VERSION. Example: "0"
-#
-#    Advanced variables:
-#      - OpenCV_SHARED
-#      - OpenCV_CONFIG_PATH
-#      - OpenCV_LIB_COMPONENTS
-#
-# ===================================================================================
-#
-#    Windows pack specific options:
-#      - OpenCV_STATIC
-#      - OpenCV_CUDA
-
-if(CMAKE_VERSION VERSION_GREATER 2.6)
-  get_property(OpenCV_LANGUAGES GLOBAL PROPERTY ENABLED_LANGUAGES)
-  if(NOT ";${OpenCV_LANGUAGES};" MATCHES ";CXX;")
-    enable_language(CXX)
-  endif()
-endif()
-
-if(NOT DEFINED OpenCV_STATIC)
-  # look for global setting
-  if(NOT DEFINED BUILD_SHARED_LIBS OR BUILD_SHARED_LIBS)
-    set(OpenCV_STATIC OFF)
-  else()
-    set(OpenCV_STATIC ON)
-  endif()
-endif()
-
-if(NOT DEFINED OpenCV_CUDA)
-  # if user' app uses CUDA, then it probably wants CUDA-enabled OpenCV binaries
-  if(CUDA_FOUND)
-    set(OpenCV_CUDA ON)
-  endif()
-endif()
-
-if(MSVC)
-  if(CMAKE_CL_64)
-    set(OpenCV_ARCH x64)
-    set(OpenCV_TBB_ARCH intel64)
-  else()
-    set(OpenCV_ARCH x86)
-    set(OpenCV_TBB_ARCH ia32)
-  endif()
-  if(MSVC_VERSION EQUAL 1400)
-    set(OpenCV_RUNTIME vc8)
-  elseif(MSVC_VERSION EQUAL 1500)
-    set(OpenCV_RUNTIME vc9)
-  elseif(MSVC_VERSION EQUAL 1600)
-    set(OpenCV_RUNTIME vc10)
-  elseif(MSVC_VERSION EQUAL 1700)
-    set(OpenCV_RUNTIME vc11)
-  endif()
-elseif(MINGW)
-  set(OpenCV_RUNTIME mingw)
-
-  execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpmachine
-                  OUTPUT_VARIABLE OPENCV_GCC_TARGET_MACHINE
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
-  if(CMAKE_OPENCV_GCC_TARGET_MACHINE MATCHES "64")
-    set(MINGW64 1)
-    set(OpenCV_ARCH x64)
-  else()
-    set(OpenCV_ARCH x86)
-  endif()
-endif()
-
-if(CMAKE_VERSION VERSION_GREATER 2.6.2)
-  unset(OpenCV_CONFIG_PATH CACHE)
-endif()
-
-get_filename_component(OpenCV_CONFIG_PATH "${CMAKE_CURRENT_LIST_FILE}" PATH CACHE)
-if(OpenCV_RUNTIME AND OpenCV_ARCH)
-  if(OpenCV_STATIC AND EXISTS "${OpenCV_CONFIG_PATH}/${OpenCV_ARCH}/${OpenCV_RUNTIME}/staticlib/OpenCVConfig.cmake")
-    if(OpenCV_CUDA AND EXISTS "${OpenCV_CONFIG_PATH}/gpu/${OpenCV_ARCH}/${OpenCV_RUNTIME}/staticlib/OpenCVConfig.cmake")
-      set(OpenCV_LIB_PATH "${OpenCV_CONFIG_PATH}/gpu/${OpenCV_ARCH}/${OpenCV_RUNTIME}/staticlib")
-    else()
-      set(OpenCV_LIB_PATH "${OpenCV_CONFIG_PATH}/${OpenCV_ARCH}/${OpenCV_RUNTIME}/staticlib")
-    endif()
-  elseif(EXISTS "${OpenCV_CONFIG_PATH}/${OpenCV_ARCH}/${OpenCV_RUNTIME}/lib/OpenCVConfig.cmake")
-    if(OpenCV_CUDA AND EXISTS "${OpenCV_CONFIG_PATH}/gpu/${OpenCV_ARCH}/${OpenCV_RUNTIME}/lib/OpenCVConfig.cmake")
-      set(OpenCV_LIB_PATH "${OpenCV_CONFIG_PATH}/gpu/${OpenCV_ARCH}/${OpenCV_RUNTIME}/lib")
-    else()
-      set(OpenCV_LIB_PATH "${OpenCV_CONFIG_PATH}/${OpenCV_ARCH}/${OpenCV_RUNTIME}/lib")
-    endif()
-  endif()
-endif()
-
-if(OpenCV_LIB_PATH AND EXISTS "${OpenCV_LIB_PATH}/OpenCVConfig.cmake")
-  set(OpenCV_LIB_DIR_OPT "${OpenCV_LIB_PATH}" CACHE PATH "Path where release OpenCV libraries are located" FORCE)
-  set(OpenCV_LIB_DIR_DBG "${OpenCV_LIB_PATH}" CACHE PATH "Path where debug OpenCV libraries are located" FORCE)
-  set(OpenCV_3RDPARTY_LIB_DIR_OPT "${OpenCV_LIB_PATH}" CACHE PATH "Path where release 3rdpaty OpenCV dependencies are located" FORCE)
-  set(OpenCV_3RDPARTY_LIB_DIR_DBG "${OpenCV_LIB_PATH}" CACHE PATH "Path where debug 3rdpaty OpenCV dependencies are located" FORCE)
-
-  include("${OpenCV_LIB_PATH}/OpenCVConfig.cmake")
-
-  if(OpenCV_CUDA)
-    set(_OpenCV_LIBS "")
-    foreach(_lib ${OpenCV_LIBS})
-      string(REPLACE "${OpenCV_CONFIG_PATH}/gpu/${OpenCV_ARCH}/${OpenCV_RUNTIME}" "${OpenCV_CONFIG_PATH}/${OpenCV_ARCH}/${OpenCV_RUNTIME}" _lib2 "${_lib}")
-      if(NOT EXISTS "${_lib}" AND EXISTS "${_lib2}")
-        list(APPEND _OpenCV_LIBS "${_lib2}")
-      else()
-        list(APPEND _OpenCV_LIBS "${_lib}")
-      endif()
-    endforeach()
-    set(OpenCV_LIBS ${_OpenCV_LIBS})
-  endif()
-  set(OpenCV_FOUND TRUE CACHE BOOL "" FORCE)
-  set(OPENCV_FOUND TRUE CACHE BOOL "" FORCE)
-
-  if(NOT OpenCV_FIND_QUIETLY)
-    message(STATUS "Found OpenCV ${OpenCV_VERSION} in ${OpenCV_LIB_PATH}")
-    if(NOT OpenCV_LIB_PATH MATCHES "/staticlib")
-      get_filename_component(_OpenCV_LIB_PATH "${OpenCV_LIB_PATH}/../bin" ABSOLUTE)
-      file(TO_NATIVE_PATH "${_OpenCV_LIB_PATH}" _OpenCV_LIB_PATH)
-      message(STATUS "You might need to add ${_OpenCV_LIB_PATH} to your PATH to be able to run your applications.")
-      if(OpenCV_LIB_PATH MATCHES "/gpu/")
-        string(REPLACE "\\gpu" "" _OpenCV_LIB_PATH2 "${_OpenCV_LIB_PATH}")
-        message(STATUS "GPU support is enabled so you might also need ${_OpenCV_LIB_PATH2} in your PATH (it must go after the ${_OpenCV_LIB_PATH}).")
-      endif()
-    endif()
-  endif()
-else()
-  if(NOT OpenCV_FIND_QUIETLY)
-    message(WARNING "Found OpenCV 2.4.3 Windows Super Pack but it has not binaries compatible with your configuration.
-    You should manually point CMake variable OpenCV_DIR to your build of OpenCV library.")
-  endif()
-  set(OpenCV_FOUND FALSE CACHE BOOL "" FORCE)
-  set(OPENCV_FOUND FALSE CACHE BOOL "" FORCE)
-endif()
-
diff --git a/cmake/OpenCVDetectAndroidSDK.cmake b/cmake/OpenCVDetectAndroidSDK.cmake
index 0e0240ca86..0173223d4e 100644
--- a/cmake/OpenCVDetectAndroidSDK.cmake
+++ b/cmake/OpenCVDetectAndroidSDK.cmake
@@ -124,7 +124,7 @@ if(ANDROID_EXECUTABLE)
   if(NOT ANDROID_SDK_TARGET)
     set(ANDROID_SDK_TARGET "" CACHE STRING "Android SDK target for the OpenCV Java API and samples")
   endif()
-  if(ANDROID_SDK_TARGETS AND CMAKE_VERSION VERSION_GREATER "2.8")
+  if(ANDROID_SDK_TARGETS)
     set_property( CACHE ANDROID_SDK_TARGET PROPERTY STRINGS ${ANDROID_SDK_TARGETS} )
   endif()
 endif(ANDROID_EXECUTABLE)
@@ -176,7 +176,8 @@ macro(android_get_compatible_target VAR)
 endmacro()
 
 unset(__android_project_chain CACHE)
-#add_android_project(target_name ${path} NATIVE_DEPS opencv_core LIBRARY_DEPS ${OpenCV_BINARY_DIR} SDK_TARGET 11)
+
+# add_android_project(target_name ${path} NATIVE_DEPS opencv_core LIBRARY_DEPS ${OpenCV_BINARY_DIR} SDK_TARGET 11)
 macro(add_android_project target path)
   # parse arguments
   set(android_proj_arglist NATIVE_DEPS LIBRARY_DEPS SDK_TARGET IGNORE_JAVA IGNORE_MANIFEST)
@@ -212,6 +213,16 @@ macro(add_android_project target path)
     ocv_check_dependencies(${android_proj_NATIVE_DEPS} opencv_java)
   endif()
 
+  if(EXISTS "${path}/jni/Android.mk" )
+    # find if native_app_glue is used
+    file(STRINGS "${path}/jni/Android.mk" NATIVE_APP_GLUE REGEX ".*(call import-module,android/native_app_glue)" )
+    if(NATIVE_APP_GLUE)
+      if(ANDROID_NATIVE_API_LEVEL LESS 9 OR NOT EXISTS "${ANDROID_NDK}/sources/android/native_app_glue")
+        set(OCV_DEPENDENCIES_FOUND FALSE)
+      endif()
+    endif()
+  endif()
+
   if(OCV_DEPENDENCIES_FOUND AND android_proj_sdk_target AND ANDROID_EXECUTABLE AND ANT_EXECUTABLE AND ANDROID_TOOLS_Pkg_Revision GREATER 13 AND EXISTS "${path}/${ANDROID_MANIFEST_FILE}")
 
     project(${target})
@@ -264,6 +275,7 @@ macro(add_android_project target path)
     ocv_list_filterout(android_proj_jni_files "\\\\.svn")
 
     if(android_proj_jni_files AND EXISTS ${path}/jni/Android.mk AND NOT DEFINED JNI_LIB_NAME)
+      # find local module name in Android.mk file to build native lib
       file(STRINGS "${path}/jni/Android.mk" JNI_LIB_NAME REGEX "LOCAL_MODULE[ ]*:=[ ]*.*" )
       string(REGEX REPLACE "LOCAL_MODULE[ ]*:=[ ]*([a-zA-Z_][a-zA-Z_0-9]*)[ ]*" "\\1" JNI_LIB_NAME "${JNI_LIB_NAME}")
 
@@ -271,6 +283,13 @@ macro(add_android_project target path)
         ocv_include_modules_recurse(${android_proj_NATIVE_DEPS})
         ocv_include_directories("${path}/jni")
 
+        if(NATIVE_APP_GLUE)
+          include_directories(${ANDROID_NDK}/sources/android/native_app_glue)
+          list(APPEND android_proj_jni_files ${ANDROID_NDK}/sources/android/native_app_glue/android_native_app_glue.c)
+          ocv_warnings_disable(CMAKE_C_FLAGS -Wstrict-prototypes -Wunused-parameter -Wmissing-prototypes)
+          set(android_proj_NATIVE_DEPS ${android_proj_NATIVE_DEPS} android)
+        endif()
+
         add_library(${JNI_LIB_NAME} MODULE ${android_proj_jni_files})
         target_link_libraries(${JNI_LIB_NAME} ${OPENCV_LINKER_LIBS} ${android_proj_NATIVE_DEPS})
 
@@ -280,9 +299,9 @@ macro(add_android_project target path)
             )
 
         get_target_property(android_proj_jni_location "${JNI_LIB_NAME}" LOCATION)
-    if (NOT (CMAKE_BUILD_TYPE MATCHES "debug"))
-        add_custom_command(TARGET ${JNI_LIB_NAME} POST_BUILD COMMAND ${CMAKE_STRIP} --strip-unneeded "${android_proj_jni_location}")
-    endif()
+        if (NOT (CMAKE_BUILD_TYPE MATCHES "debug"))
+            add_custom_command(TARGET ${JNI_LIB_NAME} POST_BUILD COMMAND ${CMAKE_STRIP} --strip-unneeded "${android_proj_jni_location}")
+        endif()
       endif()
     endif()
 
diff --git a/cmake/OpenCVDetectApacheAnt.cmake b/cmake/OpenCVDetectApacheAnt.cmake
index c98fc48b3e..7b7e9a6da9 100644
--- a/cmake/OpenCVDetectApacheAnt.cmake
+++ b/cmake/OpenCVDetectApacheAnt.cmake
@@ -16,10 +16,16 @@ find_host_program(ANT_EXECUTABLE NAMES ${ANT_NAME})
 
 if(ANT_EXECUTABLE)
   execute_process(COMMAND ${ANT_EXECUTABLE} -version
+    RESULT_VARIABLE ANT_ERROR_LEVEL
     OUTPUT_VARIABLE ANT_VERSION_FULL
     OUTPUT_STRIP_TRAILING_WHITESPACE)
-  string(REGEX MATCH "[0-9]+.[0-9]+.[0-9]+" ANT_VERSION "${ANT_VERSION_FULL}")
-  set(ANT_VERSION "${ANT_VERSION}" CACHE INTERNAL "Detected ant vesion")
+  if (ANT_ERROR_LEVEL)
+    unset(ANT_EXECUTABLE)
+    unset(ANT_EXECUTABLE CACHE)
+  else()
+    string(REGEX MATCH "[0-9]+.[0-9]+.[0-9]+" ANT_VERSION "${ANT_VERSION_FULL}")
+    set(ANT_VERSION "${ANT_VERSION}" CACHE INTERNAL "Detected ant vesion")
 
-  message(STATUS "Found apache ant ${ANT_VERSION}: ${ANT_EXECUTABLE}")
+    message(STATUS "Found apache ant ${ANT_VERSION}: ${ANT_EXECUTABLE}")
+  endif()
 endif()
diff --git a/cmake/OpenCVDetectCUDA.cmake b/cmake/OpenCVDetectCUDA.cmake
index e853a8d0a0..076da108f7 100644
--- a/cmake/OpenCVDetectCUDA.cmake
+++ b/cmake/OpenCVDetectCUDA.cmake
@@ -1,8 +1,3 @@
-if(${CMAKE_VERSION} VERSION_LESS "2.8.3")
-  message(STATUS "WITH_CUDA flag requires CMake 2.8.3 or newer. CUDA support is disabled.")
-  return()
-endif()
-
 if(WIN32 AND NOT MSVC)
   message(STATUS "CUDA compilation is disabled (due to only Visual Studio compiler supported on your platform).")
   return()
@@ -26,8 +21,52 @@ if(CUDA_FOUND)
     set(HAVE_CUBLAS 1)
   endif()
 
+  if(${CUDA_VERSION} VERSION_LESS "5.5")
+    find_cuda_helper_libs(npp)
+  else()
+    # hack for CUDA 5.5
+    if(${CMAKE_SYSTEM_PROCESSOR} STREQUAL "arm")
+      unset(CUDA_TOOLKIT_INCLUDE CACHE)
+      unset(CUDA_CUDART_LIBRARY CACHE)
+      unset(CUDA_cublas_LIBRARY CACHE)
+      unset(CUDA_cufft_LIBRARY CACHE)
+      unset(CUDA_npp_LIBRARY CACHE)
+
+      if(SOFTFP)
+        set(cuda_arm_path "${CUDA_TOOLKIT_ROOT_DIR}/targets/armv7-linux-gnueabi")
+      else()
+        set(cuda_arm_path "${CUDA_TOOLKIT_ROOT_DIR}/targets/armv7-linux-gnueabihf")
+      endif()
+
+      set(CUDA_TOOLKIT_INCLUDE "${cuda_arm_path}/include" CACHE PATH "include path")
+      set(CUDA_INCLUDE_DIRS ${CUDA_TOOLKIT_INCLUDE})
+
+      set(cuda_arm_library_path "${cuda_arm_path}/lib")
+
+      set(CUDA_CUDART_LIBRARY "${cuda_arm_library_path}/libcudart.so" CACHE FILEPATH "cudart library")
+      set(CUDA_LIBRARIES ${CUDA_CUDART_LIBRARY})
+      set(CUDA_cublas_LIBRARY "${cuda_arm_library_path}/libcublas.so" CACHE FILEPATH "cublas library")
+      set(CUDA_cufft_LIBRARY "${cuda_arm_library_path}/libcufft.so" CACHE FILEPATH "cufft library")
+      set(CUDA_nppc_LIBRARY "${cuda_arm_library_path}/libnppc.so" CACHE FILEPATH "nppc library")
+      set(CUDA_nppi_LIBRARY "${cuda_arm_library_path}/libnppi.so" CACHE FILEPATH "nppi library")
+      set(CUDA_npps_LIBRARY "${cuda_arm_library_path}/libnpps.so" CACHE FILEPATH "npps library")
+      set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppi_LIBRARY};${CUDA_npps_LIBRARY}" CACHE STRING "npp library")
+    else()
+      unset(CUDA_npp_LIBRARY CACHE)
+
+      find_cuda_helper_libs(nppc)
+      find_cuda_helper_libs(nppi)
+      find_cuda_helper_libs(npps)
+
+      set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppi_LIBRARY};${CUDA_npps_LIBRARY}" CACHE STRING "npp library")
+    endif()
+  endif()
+
   if(WITH_NVCUVID)
     find_cuda_helper_libs(nvcuvid)
+    if(WIN32)
+      find_cuda_helper_libs(nvcuvenc)
+    endif()
     set(HAVE_NVCUVID 1)
   endif()
 
@@ -55,9 +94,13 @@ if(CUDA_FOUND)
   if(CUDA_GENERATION STREQUAL "Fermi")
     set(__cuda_arch_bin "2.0 2.1(2.0)")
   elseif(CUDA_GENERATION STREQUAL "Kepler")
-    set(__cuda_arch_bin "3.0")
+    if(${CUDA_VERSION} VERSION_LESS "5.0")
+      set(__cuda_arch_bin "3.0")
+    else()
+      set(__cuda_arch_bin "3.0 3.5")
+    endif()
   elseif(CUDA_GENERATION STREQUAL "Auto")
-    execute_process( COMMAND "${CUDA_NVCC_EXECUTABLE}" "${OpenCV_SOURCE_DIR}/cmake/OpenCVDetectCudaArch.cu" "--run"
+    execute_process( COMMAND "${CUDA_NVCC_EXECUTABLE}" "${OpenCV_SOURCE_DIR}/cmake/checks/OpenCVDetectCudaArch.cu" "--run"
                      WORKING_DIRECTORY "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/"
                      RESULT_VARIABLE _nvcc_res OUTPUT_VARIABLE _nvcc_out
                      ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
@@ -69,8 +112,12 @@ if(CUDA_FOUND)
   endif()
 
   if(NOT DEFINED __cuda_arch_bin)
-    set(__cuda_arch_bin "1.1 1.2 1.3 2.0 2.1(2.0) 3.0")
-    set(__cuda_arch_ptx "2.0 3.0")
+    if(${CUDA_VERSION} VERSION_LESS "5.0")
+      set(__cuda_arch_bin "1.1 1.2 1.3 2.0 2.1(2.0) 3.0")
+    else()
+      set(__cuda_arch_bin "1.1 1.2 1.3 2.0 2.1(2.0) 3.0 3.5")
+    endif()
+    set(__cuda_arch_ptx "3.0")
   endif()
 
   set(CUDA_ARCH_BIN ${__cuda_arch_bin} CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
@@ -136,17 +183,22 @@ if(CUDA_FOUND)
 
   mark_as_advanced(CUDA_BUILD_CUBIN CUDA_BUILD_EMULATION CUDA_VERBOSE_BUILD CUDA_SDK_ROOT_DIR)
 
-  find_cuda_helper_libs(npp)
-
   macro(ocv_cuda_compile VAR)
     foreach(var CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_RELEASE CMAKE_CXX_FLAGS_DEBUG)
       set(${var}_backup_in_cuda_compile_ "${${var}}")
 
-      # we reomove /EHa as it leasd warnings under windows
+      # we remove /EHa as it generates warnings under windows
       string(REPLACE "/EHa" "" ${var} "${${var}}")
 
       # we remove -ggdb3 flag as it leads to preprocessor errors when compiling CUDA files (CUDA 4.1)
       string(REPLACE "-ggdb3" "" ${var} "${${var}}")
+
+      # we remove -Wsign-promo as it generates warnings under linux
+      string(REPLACE "-Wsign-promo" "" ${var} "${${var}}")
+
+      # we remove -fvisibility-inlines-hidden because it's used for C++ compiler
+      # but NVCC uses C compiler by default
+      string(REPLACE "-fvisibility-inlines-hidden" "" ${var} "${${var}}")
     endforeach()
 
     if(BUILD_SHARED_LIBS)
diff --git a/cmake/OpenCVDetectCXXCompiler.cmake b/cmake/OpenCVDetectCXXCompiler.cmake
index 6e02780009..7efcba7c1e 100644
--- a/cmake/OpenCVDetectCXXCompiler.cmake
+++ b/cmake/OpenCVDetectCXXCompiler.cmake
@@ -27,23 +27,23 @@ endif()
 #  the -fPIC flag should be used.
 # ----------------------------------------------------------------------------
 if(UNIX)
-    if  (__ICL)
-        set(CV_ICC   __ICL)
-    elseif(__ICC)
-        set(CV_ICC   __ICC)
-    elseif(__ECL)
-        set(CV_ICC   __ECL)
-    elseif(__ECC)
-        set(CV_ICC   __ECC)
-    elseif(__INTEL_COMPILER)
-        set(CV_ICC   __INTEL_COMPILER)
-    elseif(CMAKE_C_COMPILER MATCHES "icc")
-        set(CV_ICC   icc_matches_c_compiler)
-    endif()
+  if  (__ICL)
+    set(CV_ICC   __ICL)
+  elseif(__ICC)
+    set(CV_ICC   __ICC)
+  elseif(__ECL)
+    set(CV_ICC   __ECL)
+  elseif(__ECC)
+    set(CV_ICC   __ECC)
+  elseif(__INTEL_COMPILER)
+    set(CV_ICC   __INTEL_COMPILER)
+  elseif(CMAKE_C_COMPILER MATCHES "icc")
+    set(CV_ICC   icc_matches_c_compiler)
+  endif()
 endif()
 
 if(MSVC AND CMAKE_C_COMPILER MATCHES "icc")
-    set(CV_ICC   __INTEL_COMPILER_FOR_WINDOWS)
+  set(CV_ICC   __INTEL_COMPILER_FOR_WINDOWS)
 endif()
 
 # ----------------------------------------------------------------------------
@@ -64,43 +64,49 @@ if(CMAKE_COMPILER_IS_CLANGCXX)
   string(REGEX MATCH "[0-9]+\\.[0-9]+" CMAKE_CLANG_REGEX_VERSION "${CMAKE_OPENCV_CLANG_VERSION_FULL}")
 
 elseif(CMAKE_COMPILER_IS_GNUCXX)
-    execute_process(COMMAND ${CMAKE_CXX_COMPILER} ${CMAKE_CXX_COMPILER_ARG1} -dumpversion
-                  OUTPUT_VARIABLE CMAKE_OPENCV_GCC_VERSION_FULL
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
+  execute_process(COMMAND ${CMAKE_CXX_COMPILER} ${CMAKE_CXX_COMPILER_ARG1} -dumpversion
+                OUTPUT_VARIABLE CMAKE_OPENCV_GCC_VERSION_FULL
+                OUTPUT_STRIP_TRAILING_WHITESPACE)
 
-    execute_process(COMMAND ${CMAKE_CXX_COMPILER} ${CMAKE_CXX_COMPILER_ARG1} -v
-                  ERROR_VARIABLE CMAKE_OPENCV_GCC_INFO_FULL
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
+  execute_process(COMMAND ${CMAKE_CXX_COMPILER} ${CMAKE_CXX_COMPILER_ARG1} -v
+                ERROR_VARIABLE CMAKE_OPENCV_GCC_INFO_FULL
+                OUTPUT_STRIP_TRAILING_WHITESPACE)
 
-    # Typical output in CMAKE_OPENCV_GCC_VERSION_FULL: "c+//0 (whatever) 4.2.3 (...)"
-    # Look for the version number
-    string(REGEX MATCH "[0-9]+\\.[0-9]+\\.[0-9]+" CMAKE_GCC_REGEX_VERSION "${CMAKE_OPENCV_GCC_VERSION_FULL}")
-    if(NOT CMAKE_GCC_REGEX_VERSION)
-      string(REGEX MATCH "[0-9]+\\.[0-9]+" CMAKE_GCC_REGEX_VERSION "${CMAKE_OPENCV_GCC_VERSION_FULL}")
-    endif()
-
-    # Split the three parts:
-    string(REGEX MATCHALL "[0-9]+" CMAKE_OPENCV_GCC_VERSIONS "${CMAKE_GCC_REGEX_VERSION}")
-
-    list(GET CMAKE_OPENCV_GCC_VERSIONS 0 CMAKE_OPENCV_GCC_VERSION_MAJOR)
-    list(GET CMAKE_OPENCV_GCC_VERSIONS 1 CMAKE_OPENCV_GCC_VERSION_MINOR)
-
-    set(CMAKE_OPENCV_GCC_VERSION ${CMAKE_OPENCV_GCC_VERSION_MAJOR}${CMAKE_OPENCV_GCC_VERSION_MINOR})
-    math(EXPR CMAKE_OPENCV_GCC_VERSION_NUM "${CMAKE_OPENCV_GCC_VERSION_MAJOR}*100 + ${CMAKE_OPENCV_GCC_VERSION_MINOR}")
-    message(STATUS "Detected version of GNU GCC: ${CMAKE_OPENCV_GCC_VERSION} (${CMAKE_OPENCV_GCC_VERSION_NUM})")
-
-    if(WIN32)
-        execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpmachine
-                  OUTPUT_VARIABLE CMAKE_OPENCV_GCC_TARGET_MACHINE
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
-        if(CMAKE_OPENCV_GCC_TARGET_MACHINE MATCHES "64")
-            set(MINGW64 1)
-        endif()
+  # Typical output in CMAKE_OPENCV_GCC_VERSION_FULL: "c+//0 (whatever) 4.2.3 (...)"
+  # Look for the version number
+  string(REGEX MATCH "[0-9]+\\.[0-9]+\\.[0-9]+" CMAKE_GCC_REGEX_VERSION "${CMAKE_OPENCV_GCC_VERSION_FULL}")
+  if(NOT CMAKE_GCC_REGEX_VERSION)
+    string(REGEX MATCH "[0-9]+\\.[0-9]+" CMAKE_GCC_REGEX_VERSION "${CMAKE_OPENCV_GCC_VERSION_FULL}")
+  endif()
+
+  # Split the three parts:
+  string(REGEX MATCHALL "[0-9]+" CMAKE_OPENCV_GCC_VERSIONS "${CMAKE_GCC_REGEX_VERSION}")
+
+  list(GET CMAKE_OPENCV_GCC_VERSIONS 0 CMAKE_OPENCV_GCC_VERSION_MAJOR)
+  list(GET CMAKE_OPENCV_GCC_VERSIONS 1 CMAKE_OPENCV_GCC_VERSION_MINOR)
+
+  set(CMAKE_OPENCV_GCC_VERSION ${CMAKE_OPENCV_GCC_VERSION_MAJOR}${CMAKE_OPENCV_GCC_VERSION_MINOR})
+  math(EXPR CMAKE_OPENCV_GCC_VERSION_NUM "${CMAKE_OPENCV_GCC_VERSION_MAJOR}*100 + ${CMAKE_OPENCV_GCC_VERSION_MINOR}")
+  message(STATUS "Detected version of GNU GCC: ${CMAKE_OPENCV_GCC_VERSION} (${CMAKE_OPENCV_GCC_VERSION_NUM})")
+
+  if(WIN32)
+    execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpmachine
+              OUTPUT_VARIABLE CMAKE_OPENCV_GCC_TARGET_MACHINE
+              OUTPUT_STRIP_TRAILING_WHITESPACE)
+    if(CMAKE_OPENCV_GCC_TARGET_MACHINE MATCHES "amd64|x86_64|AMD64")
+      set(MINGW64 1)
     endif()
+  endif()
 endif()
 
-if(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*" OR CMAKE_GENERATOR MATCHES "Visual Studio.*Win64")
-    set(X86_64 1)
+if(MSVC64 OR MINGW64)
+  set(X86_64 1)
+elseif(MINGW OR (MSVC AND NOT CMAKE_CROSSCOMPILING))
+  set(X86 1)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
+  set(X86_64 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*|amd64.*|AMD64.*")
-    set(X86 1)
+  set(X86 1)
+elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "arm.*|ARM.*")
+  set(ARM 1)
 endif()
diff --git a/cmake/OpenCVDetectOpenCL.cmake b/cmake/OpenCVDetectOpenCL.cmake
index 12ab9d3eae..c96df82e2b 100644
--- a/cmake/OpenCVDetectOpenCL.cmake
+++ b/cmake/OpenCVDetectOpenCL.cmake
@@ -1,154 +1,126 @@
 if(APPLE)
   set(OPENCL_FOUND YES)
-  set(OPENCL_LIBRARIES "-framework OpenCL")
-else()
-  find_package(OpenCL QUIET)
-  if(WITH_OPENCLAMDFFT)
-    set(CLAMDFFT_SEARCH_PATH $ENV{CLAMDFFT_PATH})
-    if(NOT CLAMDFFT_SEARCH_PATH)
-      if(WIN32)
-        set( CLAMDFFT_SEARCH_PATH "C:\\Program Files (x86)\\AMD\\clAmdFft" )
-      endif()
-    endif()
-    set( CLAMDFFT_INCLUDE_SEARCH_PATH ${CLAMDFFT_SEARCH_PATH}/include )
-    if(UNIX)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(CLAMDFFT_LIB_SEARCH_PATH /usr/lib)
-      else()
-        set(CLAMDFFT_LIB_SEARCH_PATH /usr/lib64)
-      endif()
-    else()
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(CLAMDFFT_LIB_SEARCH_PATH ${CLAMDFFT_SEARCH_PATH}\\lib32\\import)
-      else()
-        set(CLAMDFFT_LIB_SEARCH_PATH ${CLAMDFFT_SEARCH_PATH}\\lib64\\import)
-      endif()
-    endif()
-    find_path(CLAMDFFT_INCLUDE_DIR
-      NAMES clAmdFft.h
-      PATHS ${CLAMDFFT_INCLUDE_SEARCH_PATH}
-      PATH_SUFFIXES clAmdFft
-      NO_DEFAULT_PATH)
-    find_library(CLAMDFFT_LIBRARY
-      NAMES clAmdFft.Runtime
-      PATHS ${CLAMDFFT_LIB_SEARCH_PATH}
-      NO_DEFAULT_PATH)
-    if(CLAMDFFT_LIBRARY)
-      set(CLAMDFFT_LIBRARIES ${CLAMDFFT_LIBRARY})
-    else()
-      set(CLAMDFFT_LIBRARIES "")
-    endif()
-  endif()
-  if(WITH_OPENCLAMDBLAS)
-    set(CLAMDBLAS_SEARCH_PATH $ENV{CLAMDBLAS_PATH})
-    if(NOT CLAMDBLAS_SEARCH_PATH)
-      if(WIN32)
-        set( CLAMDBLAS_SEARCH_PATH "C:\\Program Files (x86)\\AMD\\clAmdBlas" )
-      endif()
-    endif()
-    set( CLAMDBLAS_INCLUDE_SEARCH_PATH ${CLAMDBLAS_SEARCH_PATH}/include )
-    if(UNIX)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(CLAMDBLAS_LIB_SEARCH_PATH /usr/lib)
-      else()
-        set(CLAMDBLAS_LIB_SEARCH_PATH /usr/lib64)
-      endif()
-    else()
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(CLAMDBLAS_LIB_SEARCH_PATH ${CLAMDBLAS_SEARCH_PATH}\\lib32\\import)
-      else()
-        set(CLAMDBLAS_LIB_SEARCH_PATH ${CLAMDBLAS_SEARCH_PATH}\\lib64\\import)
-      endif()
-    endif()
-    find_path(CLAMDBLAS_INCLUDE_DIR
-      NAMES clAmdBlas.h
-      PATHS ${CLAMDBLAS_INCLUDE_SEARCH_PATH}
-      PATH_SUFFIXES clAmdBlas
-      NO_DEFAULT_PATH)
-    find_library(CLAMDBLAS_LIBRARY
-      NAMES clAmdBlas
-      PATHS ${CLAMDBLAS_LIB_SEARCH_PATH}
-      NO_DEFAULT_PATH)
-    if(CLAMDBLAS_LIBRARY)
-      set(CLAMDBLAS_LIBRARIES ${CLAMDBLAS_LIBRARY})
-    else()
-      set(CLAMDBLAS_LIBRARIES "")
-    endif()
-  endif()
-  # Try AMD/ATI Stream SDK
+  set(OPENCL_LIBRARY "-framework OpenCL" CACHE STRING "OpenCL library")
+  set(OPENCL_INCLUDE_DIR "" CACHE STRING "OpenCL include directory")
+  mark_as_advanced(OPENCL_INCLUDE_DIR OPENCL_LIBRARY)
+else(APPLE)
+  #find_package(OpenCL QUIET)
+
   if (NOT OPENCL_FOUND)
-    set(ENV_AMDSTREAMSDKROOT $ENV{AMDAPPSDKROOT})
-    set(ENV_AMDAPPSDKROOT $ENV{AMDAPPSDKROOT})
-    set(ENV_OPENCLROOT $ENV{OPENCLROOT})
-    set(ENV_CUDA_PATH $ENV{CUDA_PATH})
-    set(ENV_INTELOCLSDKROOT $ENV{INTELOCLSDKROOT})
-    if(ENV_AMDSTREAMSDKROOT)
-      set(OPENCL_INCLUDE_SEARCH_PATH ${ENV_AMDAPPSDKROOT}/include)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_AMDAPPSDKROOT}/lib/x86)
-      else()
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_AMDAPPSDKROOT}/lib/x86_64)
-      endif()
-    elseif(ENV_AMDSTREAMSDKROOT)
-      set(OPENCL_INCLUDE_SEARCH_PATH ${ENV_AMDSTREAMSDKROOT}/include)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_AMDSTREAMSDKROOT}/lib/x86)
-      else()
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_AMDSTREAMSDKROOT}/lib/x86_64)
-      endif()
-    elseif(ENV_CUDA_PATH AND WIN32)
-      set(OPENCL_INCLUDE_SEARCH_PATH ${ENV_CUDA_PATH}/include)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_CUDA_PATH}/lib/Win32)
-      else()
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_CUDA_PATH}/lib/x64)
-      endif()
-    elseif(ENV_OPENCLROOT AND UNIX)
-      set(OPENCL_INCLUDE_SEARCH_PATH ${ENV_OPENCLROOT}/inc)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} /usr/lib)
-      else()
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} /usr/lib64)
-      endif()
-    elseif(ENV_INTELOCLSDKROOT)
-      set(OPENCL_INCLUDE_SEARCH_PATH ${ENV_INTELOCLSDKROOT}/include)
-      if(CMAKE_SIZEOF_VOID_P EQUAL 4)
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_INTELOCLSDKROOT}/lib/x86)
-      else()
-        set(OPENCL_LIB_SEARCH_PATH ${OPENCL_LIB_SEARCH_PATH} ${ENV_INTELOCLSDKROOT}/lib/x64)
-      endif()
-    endif()
+    find_path(OPENCL_ROOT_DIR
+              NAMES OpenCL/cl.h CL/cl.h include/CL/cl.h include/nvidia-current/CL/cl.h
+              PATHS ENV OCLROOT ENV AMDAPPSDKROOT ENV CUDA_PATH ENV INTELOCLSDKROOT
+              DOC "OpenCL root directory"
+              NO_DEFAULT_PATH)
 
-    if(OPENCL_INCLUDE_SEARCH_PATH)
-      find_path(OPENCL_INCLUDE_DIR
-        NAMES CL/cl.h OpenCL/cl.h
-        PATHS ${OPENCL_INCLUDE_SEARCH_PATH}
-        NO_DEFAULT_PATH)
+    find_path(OPENCL_INCLUDE_DIR
+              NAMES OpenCL/cl.h CL/cl.h
+              HINTS ${OPENCL_ROOT_DIR}
+              PATH_SUFFIXES include include/nvidia-current
+              DOC "OpenCL include directory"
+              NO_DEFAULT_PATH)
+
+    if(WIN32)
+      if(X86_64)
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib/Win64 lib/x86_64 lib/x64)
+      elseif(X86)
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib/Win32 lib/x86)
+      else()
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib)
+      endif()
+    elseif(UNIX)
+      if(X86_64)
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib64 lib)
+      elseif(X86)
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib32 lib)
+      else()
+        set(OPENCL_POSSIBLE_LIB_SUFFIXES lib)
+      endif()
     else()
-      find_path(OPENCL_INCLUDE_DIR
-        NAMES CL/cl.h OpenCL/cl.h)
+      set(OPENCL_POSSIBLE_LIB_SUFFIXES lib)
     endif()
 
-    if(OPENCL_LIB_SEARCH_PATH)
-      find_library(OPENCL_LIBRARY NAMES OpenCL PATHS ${OPENCL_LIB_SEARCH_PATH} NO_DEFAULT_PATH)
-    else()
-      find_library(OPENCL_LIBRARY NAMES OpenCL)
-    endif()
+    find_library(OPENCL_LIBRARY
+              NAMES OpenCL
+              HINTS ${OPENCL_ROOT_DIR}
+              PATH_SUFFIXES ${OPENCL_POSSIBLE_LIB_SUFFIXES}
+              DOC "OpenCL library"
+              NO_DEFAULT_PATH)
 
+    mark_as_advanced(OPENCL_INCLUDE_DIR OPENCL_LIBRARY)
     include(FindPackageHandleStandardArgs)
-    find_package_handle_standard_args(
-      OPENCL
-      DEFAULT_MSG
-      OPENCL_LIBRARY OPENCL_INCLUDE_DIR
-      )
+    FIND_PACKAGE_HANDLE_STANDARD_ARGS(OPENCL DEFAULT_MSG OPENCL_LIBRARY OPENCL_INCLUDE_DIR )
+  endif()
+endif(APPLE)
 
-    if(OPENCL_FOUND)
-      set(OPENCL_LIBRARIES ${OPENCL_LIBRARY})
-      set(HAVE_OPENCL 1)
-    else()
-      set(OPENCL_LIBRARIES)
+if(OPENCL_FOUND)
+  set(HAVE_OPENCL 1)
+  set(OPENCL_INCLUDE_DIRS ${OPENCL_INCLUDE_DIR})
+  set(OPENCL_LIBRARIES    ${OPENCL_LIBRARY})
+
+  if(WIN32 AND X86_64)
+    set(CLAMD_POSSIBLE_LIB_SUFFIXES lib64/import)
+  elseif(WIN32)
+    set(CLAMD_POSSIBLE_LIB_SUFFIXES lib32/import)
+  endif()
+
+  if(X86_64 AND UNIX)
+    set(CLAMD_POSSIBLE_LIB_SUFFIXES lib64)
+  elseif(X86 AND UNIX)
+    set(CLAMD_POSSIBLE_LIB_SUFFIXES lib32)
+  endif()
+
+  if(WITH_OPENCLAMDFFT)
+    find_path(CLAMDFFT_ROOT_DIR
+              NAMES include/clAmdFft.h
+              PATHS ENV CLAMDFFT_PATH ENV ProgramFiles
+              PATH_SUFFIXES clAmdFft AMD/clAmdFft
+              DOC "AMD FFT root directory"
+              NO_DEFAULT_PATH)
+
+    find_path(CLAMDFFT_INCLUDE_DIR
+              NAMES clAmdFft.h
+              HINTS ${CLAMDFFT_ROOT_DIR}
+              PATH_SUFFIXES include
+              DOC "clAmdFft include directory")
+
+    find_library(CLAMDFFT_LIBRARY
+              NAMES clAmdFft.Runtime
+              HINTS ${CLAMDFFT_ROOT_DIR}
+              PATH_SUFFIXES ${CLAMD_POSSIBLE_LIB_SUFFIXES}
+              DOC "clAmdFft library")
+
+    if(CLAMDFFT_LIBRARY AND CLAMDFFT_INCLUDE_DIR)
+      set(HAVE_CLAMDFFT 1)
+      list(APPEND OPENCL_INCLUDE_DIRS "${CLAMDFFT_INCLUDE_DIR}")
+      list(APPEND OPENCL_LIBRARIES    "${CLAMDFFT_LIBRARY}")
+    endif()
+  endif()
+
+  if(WITH_OPENCLAMDBLAS)
+    find_path(CLAMDBLAS_ROOT_DIR
+              NAMES include/clAmdBlas.h
+              PATHS ENV CLAMDBLAS_PATH ENV ProgramFiles
+              PATH_SUFFIXES clAmdBlas AMD/clAmdBlas
+              DOC "AMD FFT root directory"
+              NO_DEFAULT_PATH)
+
+    find_path(CLAMDBLAS_INCLUDE_DIR
+              NAMES clAmdBlas.h
+              HINTS ${CLAMDBLAS_ROOT_DIR}
+              PATH_SUFFIXES include
+              DOC "clAmdFft include directory")
+
+    find_library(CLAMDBLAS_LIBRARY
+              NAMES clAmdBlas
+              HINTS ${CLAMDBLAS_ROOT_DIR}
+              PATH_SUFFIXES ${CLAMD_POSSIBLE_LIB_SUFFIXES}
+              DOC "clAmdBlas library")
+
+    if(CLAMDBLAS_LIBRARY AND CLAMDBLAS_INCLUDE_DIR)
+      set(HAVE_CLAMDBLAS 1)
+      list(APPEND OPENCL_INCLUDE_DIRS "${CLAMDBLAS_INCLUDE_DIR}")
+      list(APPEND OPENCL_LIBRARIES    "${CLAMDBLAS_LIBRARY}")
     endif()
-  else()
-    set(HAVE_OPENCL 1)
   endif()
 endif()
diff --git a/cmake/OpenCVDetectPython.cmake b/cmake/OpenCVDetectPython.cmake
index 73143b7341..0ef0be9c99 100644
--- a/cmake/OpenCVDetectPython.cmake
+++ b/cmake/OpenCVDetectPython.cmake
@@ -2,7 +2,7 @@ if(WIN32 AND NOT PYTHON_EXECUTABLE)
   # search for executable with the same bitness as resulting binaries
   # standard FindPythonInterp always prefers executable from system path
   # this is really important because we are using the interpreter for numpy search and for choosing the install location
-  foreach(_CURRENT_VERSION ${Python_ADDITIONAL_VERSIONS} 2.7 2.6 2.5 2.4 2.3 2.2 2.1 2.0)
+  foreach(_CURRENT_VERSION ${Python_ADDITIONAL_VERSIONS} 2.7 "${MIN_VER_PYTHON}")
     find_host_program(PYTHON_EXECUTABLE
       NAMES python${_CURRENT_VERSION} python
       PATHS
@@ -12,45 +12,20 @@ if(WIN32 AND NOT PYTHON_EXECUTABLE)
     )
   endforeach()
 endif()
-find_host_package(PythonInterp 2.0)
+find_host_package(PythonInterp "${MIN_VER_PYTHON}")
 
-unset(PYTHON_USE_NUMPY CACHE)
 unset(HAVE_SPHINX CACHE)
-if(PYTHON_EXECUTABLE)
-  if(PYTHON_VERSION_STRING)
-    set(PYTHON_VERSION_MAJOR_MINOR "${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}")
-    set(PYTHON_VERSION_FULL "${PYTHON_VERSION_STRING}")
-  else()
-    execute_process(COMMAND ${PYTHON_EXECUTABLE} --version
-      ERROR_VARIABLE PYTHON_VERSION_FULL
-      ERROR_STRIP_TRAILING_WHITESPACE)
 
-    string(REGEX MATCH "[0-9]+.[0-9]+" PYTHON_VERSION_MAJOR_MINOR "${PYTHON_VERSION_FULL}")
-  endif()
-
-  if("${PYTHON_VERSION_FULL}" MATCHES "[0-9]+.[0-9]+.[0-9]+")
-    set(PYTHON_VERSION_FULL "${CMAKE_MATCH_0}")
-  elseif("${PYTHON_VERSION_FULL}" MATCHES "[0-9]+.[0-9]+")
-    set(PYTHON_VERSION_FULL "${CMAKE_MATCH_0}")
-  else()
-    unset(PYTHON_VERSION_FULL)
-  endif()
+if(PYTHONINTERP_FOUND)
+  set(PYTHON_VERSION_MAJOR_MINOR "${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}")
 
   if(NOT ANDROID AND NOT IOS)
-    if(CMAKE_VERSION VERSION_GREATER 2.8.8 AND PYTHON_VERSION_FULL)
-      find_host_package(PythonLibs ${PYTHON_VERSION_FULL} EXACT)
-    else()
-      find_host_package(PythonLibs ${PYTHON_VERSION_FULL})
-    endif()
-    # cmake 2.4 (at least on Ubuntu 8.04 (hardy)) don't define PYTHONLIBS_FOUND
-    if(NOT PYTHONLIBS_FOUND AND PYTHON_INCLUDE_PATH)
-      set(PYTHONLIBS_FOUND ON)
-    endif()
+    find_host_package(PythonLibs "${PYTHON_VERSION_STRING}" EXACT)
   endif()
 
   if(NOT ANDROID AND NOT IOS)
     if(CMAKE_HOST_UNIX)
-      execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "from distutils.sysconfig import *; print get_python_lib()"
+      execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "from distutils.sysconfig import *; print(get_python_lib())"
                       RESULT_VARIABLE PYTHON_CVPY_PROCESS
                       OUTPUT_VARIABLE PYTHON_STD_PACKAGES_PATH
                       OUTPUT_STRIP_TRAILING_WHITESPACE)
@@ -79,46 +54,40 @@ if(PYTHON_EXECUTABLE)
     endif()
     SET(PYTHON_PACKAGES_PATH "${_PYTHON_PACKAGES_PATH}" CACHE PATH "Where to install the python packages.")
 
-    if(NOT PYTHON_NUMPY_INCLUDE_DIR)
+    if(NOT PYTHON_NUMPY_INCLUDE_DIRS)
       # Attempt to discover the NumPy include directory. If this succeeds, then build python API with NumPy
-      execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "import os; os.environ['DISTUTILS_USE_SDK']='1'; import numpy.distutils; print numpy.distutils.misc_util.get_numpy_include_dirs()[0]"
+      execute_process(COMMAND "${PYTHON_EXECUTABLE}" -c
+                        "import os; os.environ['DISTUTILS_USE_SDK']='1'; import numpy.distutils; print(os.pathsep.join(numpy.distutils.misc_util.get_numpy_include_dirs()))"
                       RESULT_VARIABLE PYTHON_NUMPY_PROCESS
-                      OUTPUT_VARIABLE PYTHON_NUMPY_INCLUDE_DIR
+                      OUTPUT_VARIABLE PYTHON_NUMPY_INCLUDE_DIRS
                       OUTPUT_STRIP_TRAILING_WHITESPACE)
 
       if(PYTHON_NUMPY_PROCESS EQUAL 0)
-        file(TO_CMAKE_PATH "${PYTHON_NUMPY_INCLUDE_DIR}" _PYTHON_NUMPY_INCLUDE_DIR)
-        set(PYTHON_NUMPY_INCLUDE_DIR ${_PYTHON_NUMPY_INCLUDE_DIR} CACHE PATH "Path to numpy headers")
+        file(TO_CMAKE_PATH "${PYTHON_NUMPY_INCLUDE_DIRS}" _PYTHON_NUMPY_INCLUDE_DIRS)
+        set(PYTHON_NUMPY_INCLUDE_DIRS "${_PYTHON_NUMPY_INCLUDE_DIRS}" CACHE PATH "Path to numpy headers")
       endif()
     endif()
 
-    if(PYTHON_NUMPY_INCLUDE_DIR)
-      set(PYTHON_USE_NUMPY TRUE)
-      execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "import numpy; print numpy.version.version"
-                        RESULT_VARIABLE PYTHON_NUMPY_PROCESS
-                        OUTPUT_VARIABLE PYTHON_NUMPY_VERSION
-                        OUTPUT_STRIP_TRAILING_WHITESPACE)
+    if(PYTHON_NUMPY_INCLUDE_DIRS)
+      execute_process(COMMAND "${PYTHON_EXECUTABLE}" -c "import numpy; print(numpy.version.version)"
+                      OUTPUT_VARIABLE PYTHON_NUMPY_VERSION
+                      OUTPUT_STRIP_TRAILING_WHITESPACE)
     endif()
   endif(NOT ANDROID AND NOT IOS)
+endif()
 
-  if(BUILD_DOCS)
-    find_host_program(SPHINX_BUILD sphinx-build)
-    if(SPHINX_BUILD)
-        if(UNIX)
-            execute_process(COMMAND sh -c "${SPHINX_BUILD} -_ 2>&1 | sed -ne 1p"
-                             RESULT_VARIABLE SPHINX_PROCESS
-                             OUTPUT_VARIABLE SPHINX_VERSION
-                             OUTPUT_STRIP_TRAILING_WHITESPACE)
-        else()
-            execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "import sphinx; print sphinx.__version__"
-                            RESULT_VARIABLE SPHINX_PROCESS
-                            OUTPUT_VARIABLE SPHINX_VERSION
-                            OUTPUT_STRIP_TRAILING_WHITESPACE)
-        endif()
-        if(SPHINX_PROCESS EQUAL 0)
-          set(HAVE_SPHINX 1)
-          message(STATUS "Found Sphinx ${SPHINX_VERSION}: ${SPHINX_BUILD}")
-        endif()
-    endif()
-  endif(BUILD_DOCS)
-endif(PYTHON_EXECUTABLE)
+if(BUILD_DOCS)
+  find_host_program(SPHINX_BUILD sphinx-build)
+  find_host_program(PLANTUML plantuml)
+  if(SPHINX_BUILD)
+      execute_process(COMMAND "${SPHINX_BUILD}"
+                      OUTPUT_QUIET
+                      ERROR_VARIABLE SPHINX_OUTPUT
+                      OUTPUT_STRIP_TRAILING_WHITESPACE)
+      if(SPHINX_OUTPUT MATCHES "Sphinx v([0-9][^ \n]*)")
+        set(SPHINX_VERSION "${CMAKE_MATCH_1}")
+        set(HAVE_SPHINX 1)
+        message(STATUS "Found Sphinx ${SPHINX_VERSION}: ${SPHINX_BUILD}")
+      endif()
+  endif()
+endif(BUILD_DOCS)
diff --git a/cmake/OpenCVExtraTargets.cmake b/cmake/OpenCVExtraTargets.cmake
index 936e3a2635..b4d339155a 100644
--- a/cmake/OpenCVExtraTargets.cmake
+++ b/cmake/OpenCVExtraTargets.cmake
@@ -13,57 +13,7 @@ endif()
 
 
 # ----------------------------------------------------------------------------
-# Source package, for "make package_source"
-# ----------------------------------------------------------------------------
-if(BUILD_PACKAGE)
-  set(TARBALL_NAME "${CMAKE_PROJECT_NAME}-${OPENCV_VERSION}")
-  if (NOT WIN32)
-    if(APPLE)
-      set(TAR_CMD gnutar)
-    else()
-      set(TAR_CMD tar)
-    endif()
-    set(TAR_TRANSFORM "\"s,^,${TARBALL_NAME}/,\"")
-    add_custom_target(package_source
-      #TODO: maybe we should not remove dll's
-      COMMAND ${TAR_CMD} --transform ${TAR_TRANSFORM} -cjpf ${CMAKE_CURRENT_BINARY_DIR}/${TARBALL_NAME}.tar.bz2 --exclude=".svn" --exclude=".git" --exclude="*.pyc" --exclude="*.vcproj" --exclude="*/lib/*" --exclude="*.dll" ./
-      WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
-  else()
-    add_custom_target(package_source
-      COMMAND zip -9 -r ${CMAKE_CURRENT_BINARY_DIR}/${TARBALL_NAME}.zip . -x '*/.svn/*' '*/.git/*' '*.vcproj' '*.pyc'
-      WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
-  endif()
-  if(ENABLE_SOLUTION_FOLDERS)
-    set_target_properties(package_source PROPERTIES FOLDER "extra")
-  endif()
-endif()
-
-
-# ----------------------------------------------------------------------------
-# performance tests, for "make perf"
-# ----------------------------------------------------------------------------
-if(BUILD_PERF_TESTS AND PYTHON_EXECUTABLE)
-  if(CMAKE_VERSION VERSION_GREATER "2.8.2")
-    add_custom_target(perf
-      ${PYTHON_EXECUTABLE} "${OpenCV_SOURCE_DIR}/modules/ts/misc/run.py" --configuration $<CONFIGURATION> "${CMAKE_BINARY_DIR}"
-      WORKING_DIRECTORY "${CMAKE_BINARY_DIR}"
-      DEPENDS "${OpenCV_SOURCE_DIR}/modules/ts/misc/run.py"
-     )
-  else()
-    add_custom_target(perf
-      ${PYTHON_EXECUTABLE} "${OpenCV_SOURCE_DIR}/modules/ts/misc/run.py" "${CMAKE_BINARY_DIR}"
-      WORKING_DIRECTORY "${CMAKE_BINARY_DIR}"
-      DEPENDS "${OpenCV_SOURCE_DIR}/modules/ts/misc/run.py"
-      )
-  endif()
-  if(ENABLE_SOLUTION_FOLDERS)
-    set_target_properties(perf PROPERTIES FOLDER "extra")
-  endif()
-endif()
-
-
-# ----------------------------------------------------------------------------
-# spefial targets to build all OpenCV modules
+# target building all OpenCV modules
 # ----------------------------------------------------------------------------
 add_custom_target(opencv_modules)
 if(ENABLE_SOLUTION_FOLDERS)
@@ -72,7 +22,7 @@ endif()
 
 
 # ----------------------------------------------------------------------------
-# spefial targets to build all tests
+# targets building all tests
 # ----------------------------------------------------------------------------
 if(BUILD_TESTS)
   add_custom_target(opencv_tests)
diff --git a/cmake/OpenCVFindIPP.cmake b/cmake/OpenCVFindIPP.cmake
index d1af605653..772cae886f 100644
--- a/cmake/OpenCVFindIPP.cmake
+++ b/cmake/OpenCVFindIPP.cmake
@@ -136,12 +136,20 @@ endfunction()
 
 # ------------------------------------------------------------------------
 # This is auxiliary function called from set_ipp_variables()
-# to set IPP_LIBRARIES variable in IPP 7.x style
+# to set IPP_LIBRARIES variable in IPP 7.x and 8.x style
 # ------------------------------------------------------------------------
-function(set_ipp_new_libraries)
+function(set_ipp_new_libraries _LATEST_VERSION)
     set(IPP_PREFIX "ipp")
-    set(IPP_SUFFIX "_l")       # static not threaded libs suffix
-    set(IPP_THRD   "_t")       # static threaded libs suffix
+
+    if(${_LATEST_VERSION} VERSION_LESS "8.0")
+        set(IPP_SUFFIX "_l")        # static not threaded libs suffix IPP 7.x
+    else()
+        if(WIN32)
+            set(IPP_SUFFIX "mt")    # static not threaded libs suffix IPP 8.x for Windows
+        else()
+            set(IPP_SUFFIX "")      # static not threaded libs suffix IPP 8.x for Linux/OS X
+        endif()
+    endif()
     set(IPPCORE    "core")     # core functionality
     set(IPPSP      "s")        # signal processing
     set(IPPIP      "i")        # image processing
@@ -199,7 +207,9 @@ function(set_ipp_variables _LATEST_VERSION)
         # set INCLUDE and LIB folders
         set(IPP_INCLUDE_DIRS ${IPP_ROOT_DIR}/include PARENT_SCOPE)
 
-        if (IPP_X64)
+        if (APPLE)
+            set(IPP_LIBRARY_DIRS ${IPP_ROOT_DIR}/lib PARENT_SCOPE)
+        elseif (IPP_X64)
             if(NOT EXISTS ${IPP_ROOT_DIR}/lib/intel64)
                 message(SEND_ERROR "IPP EM64T libraries not found")
             endif()
@@ -211,8 +221,8 @@ function(set_ipp_variables _LATEST_VERSION)
             set(IPP_LIBRARY_DIRS ${IPP_ROOT_DIR}/lib/ia32 PARENT_SCOPE)
         endif()
 
-        # set IPP_LIBRARIES variable (7.x lib names)
-        set_ipp_new_libraries()
+        # set IPP_LIBRARIES variable (7.x or 8.x lib names)
+        set_ipp_new_libraries(${_LATEST_VERSION})
         set(IPP_LIBRARIES ${IPP_LIBRARIES} PARENT_SCOPE)
         message(STATUS "IPP libs: ${IPP_LIBRARIES}")
 
@@ -336,4 +346,4 @@ if(WIN32 AND MINGW AND NOT IPP_LATEST_VERSION_MAJOR LESS 7)
     #     See http://code.opencv.org/issues/1906 for additional details
     set(MSV_NTDLL    "ntdll")
     set(IPP_LIBRARIES ${IPP_LIBRARIES} ${MSV_NTDLL}${IPP_LIB_SUFFIX})
-endif()
\ No newline at end of file
+endif()
diff --git a/cmake/OpenCVFindLibsGUI.cmake b/cmake/OpenCVFindLibsGUI.cmake
index c883a80ced..c80beca055 100644
--- a/cmake/OpenCVFindLibsGUI.cmake
+++ b/cmake/OpenCVFindLibsGUI.cmake
@@ -2,13 +2,39 @@
 #  Detect 3rd-party GUI libraries
 # ----------------------------------------------------------------------------
 
+#--- Win32 UI ---
+ocv_clear_vars(HAVE_WIN32UI)
+if(WITH_WIN32UI)
+  try_compile(HAVE_WIN32UI
+    "${OpenCV_BINARY_DIR}"
+    "${OpenCV_SOURCE_DIR}/cmake/checks/win32uitest.cpp"
+    CMAKE_FLAGS "-DLINK_LIBRARIES:STRING=user32;gdi32")
+endif()
+
 # --- QT4 ---
-ocv_clear_vars(HAVE_QT)
+ocv_clear_vars(HAVE_QT HAVE_QT5)
 if(WITH_QT)
-  find_package(Qt4)
-  if(QT4_FOUND)
-    set(HAVE_QT TRUE)
-    add_definitions(-DHAVE_QT) # We need to define the macro this way, using cvconfig.h does not work
+  if(NOT WITH_QT EQUAL 4)
+    find_package(Qt5Core)
+    find_package(Qt5Gui)
+    find_package(Qt5Widgets)
+    find_package(Qt5Test)
+    find_package(Qt5Concurrent)
+    if(Qt5Core_FOUND AND Qt5Gui_FOUND AND Qt5Widgets_FOUND AND Qt5Test_FOUND AND Qt5Concurrent_FOUND)
+      set(HAVE_QT5 ON)
+      set(HAVE_QT  ON)
+      find_package(Qt5OpenGL)
+      if(Qt5OpenGL_FOUND)
+        set(QT_QTOPENGL_FOUND ON)
+      endif()
+    endif()
+  endif()
+
+  if(NOT HAVE_QT)
+    find_package(Qt4 REQUIRED QtCore QtGui QtTest)
+    if(QT4_FOUND)
+      set(HAVE_QT TRUE)
+    endif()
   endif()
 endif()
 
@@ -25,17 +51,25 @@ endif()
 # --- OpenGl ---
 ocv_clear_vars(HAVE_OPENGL HAVE_QT_OPENGL)
 if(WITH_OPENGL)
-  if(WIN32 OR QT_QTOPENGL_FOUND OR HAVE_GTKGLEXT)
+  if(WITH_WIN32UI OR (HAVE_QT AND QT_QTOPENGL_FOUND) OR HAVE_GTKGLEXT)
     find_package (OpenGL QUIET)
     if(OPENGL_FOUND)
       set(HAVE_OPENGL TRUE)
       list(APPEND OPENCV_LINKER_LIBS ${OPENGL_LIBRARIES})
       if(QT_QTOPENGL_FOUND)
         set(HAVE_QT_OPENGL TRUE)
-        add_definitions(-DHAVE_QT_OPENGL)
       else()
         ocv_include_directories(${OPENGL_INCLUDE_DIR})
       endif()
     endif()
   endif()
 endif(WITH_OPENGL)
+
+# --- Carbon & Cocoa ---
+if(APPLE)
+  if(WITH_CARBON)
+    set(HAVE_CARBON YES)
+  elseif(NOT IOS)
+    set(HAVE_COCOA YES)
+  endif()
+endif()
diff --git a/cmake/OpenCVFindLibsGrfmt.cmake b/cmake/OpenCVFindLibsGrfmt.cmake
index 501f8ff3e8..d8ddcfeb71 100644
--- a/cmake/OpenCVFindLibsGrfmt.cmake
+++ b/cmake/OpenCVFindLibsGrfmt.cmake
@@ -6,22 +6,21 @@
 if(BUILD_ZLIB)
   ocv_clear_vars(ZLIB_FOUND)
 else()
-  include(FindZLIB)
+  find_package(ZLIB "${MIN_VER_ZLIB}")
   if(ZLIB_FOUND AND ANDROID)
-    if(ZLIB_LIBRARY STREQUAL "${ANDROID_SYSROOT}/usr/lib/libz.so")
-      set(ZLIB_LIBRARY z)
+    if(ZLIB_LIBRARIES STREQUAL "${ANDROID_SYSROOT}/usr/lib/libz.so")
       set(ZLIB_LIBRARIES z)
     endif()
   endif()
 endif()
 
 if(NOT ZLIB_FOUND)
-  ocv_clear_vars(ZLIB_LIBRARY ZLIB_LIBRARIES ZLIB_INCLUDE_DIR)
+  ocv_clear_vars(ZLIB_LIBRARY ZLIB_LIBRARIES ZLIB_INCLUDE_DIRS)
 
   set(ZLIB_LIBRARY zlib)
-  set(ZLIB_LIBRARIES ${ZLIB_LIBRARY})
   add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/zlib")
-  set(ZLIB_INCLUDE_DIR "${${ZLIB_LIBRARY}_SOURCE_DIR}" "${${ZLIB_LIBRARY}_BINARY_DIR}")
+  set(ZLIB_INCLUDE_DIRS "${${ZLIB_LIBRARY}_SOURCE_DIR}" "${${ZLIB_LIBRARY}_BINARY_DIR}")
+  set(ZLIB_LIBRARIES ${ZLIB_LIBRARY})
 
   ocv_parse_header2(ZLIB "${${ZLIB_LIBRARY}_SOURCE_DIR}/zlib.h" ZLIB_VERSION)
 endif()
@@ -36,34 +35,36 @@ if(WITH_TIFF)
       ocv_parse_header("${TIFF_INCLUDE_DIR}/tiff.h" TIFF_VERSION_LINES TIFF_VERSION_CLASSIC TIFF_VERSION_BIG TIFF_VERSION TIFF_BIGTIFF_VERSION)
     endif()
   endif()
-endif()
 
-if(WITH_TIFF AND NOT TIFF_FOUND)
-  ocv_clear_vars(TIFF_LIBRARY TIFF_LIBRARIES TIFF_INCLUDE_DIR)
+  if(NOT TIFF_FOUND)
+    ocv_clear_vars(TIFF_LIBRARY TIFF_LIBRARIES TIFF_INCLUDE_DIR)
 
-  set(TIFF_LIBRARY libtiff)
-  set(TIFF_LIBRARIES ${TIFF_LIBRARY})
-  add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libtiff")
-  set(TIFF_INCLUDE_DIR "${${TIFF_LIBRARY}_SOURCE_DIR}" "${${TIFF_LIBRARY}_BINARY_DIR}")
-  ocv_parse_header("${${TIFF_LIBRARY}_SOURCE_DIR}/tiff.h" TIFF_VERSION_LINES TIFF_VERSION_CLASSIC TIFF_VERSION_BIG TIFF_VERSION TIFF_BIGTIFF_VERSION)
-endif()
-
-if(TIFF_VERSION_CLASSIC AND NOT TIFF_VERSION)
-  set(TIFF_VERSION ${TIFF_VERSION_CLASSIC})
-endif()
-
-if(TIFF_BIGTIFF_VERSION AND NOT TIFF_VERSION_BIG)
-  set(TIFF_VERSION_BIG ${TIFF_BIGTIFF_VERSION})
-endif()
-
-if(NOT TIFF_VERSION_STRING AND TIFF_INCLUDE_DIR)
-  list(GET TIFF_INCLUDE_DIR 0 _TIFF_INCLUDE_DIR)
-  if(EXISTS "${_TIFF_INCLUDE_DIR}/tiffvers.h")
-    file(STRINGS "${_TIFF_INCLUDE_DIR}/tiffvers.h" tiff_version_str REGEX "^#define[\t ]+TIFFLIB_VERSION_STR[\t ]+\"LIBTIFF, Version .*")
-    string(REGEX REPLACE "^#define[\t ]+TIFFLIB_VERSION_STR[\t ]+\"LIBTIFF, Version +([^ \\n]*).*" "\\1" TIFF_VERSION_STRING "${tiff_version_str}")
-    unset(tiff_version_str)
+    set(TIFF_LIBRARY libtiff)
+    set(TIFF_LIBRARIES ${TIFF_LIBRARY})
+    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libtiff")
+    set(TIFF_INCLUDE_DIR "${${TIFF_LIBRARY}_SOURCE_DIR}" "${${TIFF_LIBRARY}_BINARY_DIR}")
+    ocv_parse_header("${${TIFF_LIBRARY}_SOURCE_DIR}/tiff.h" TIFF_VERSION_LINES TIFF_VERSION_CLASSIC TIFF_VERSION_BIG TIFF_VERSION TIFF_BIGTIFF_VERSION)
   endif()
-  unset(_TIFF_INCLUDE_DIR)
+
+  if(TIFF_VERSION_CLASSIC AND NOT TIFF_VERSION)
+    set(TIFF_VERSION ${TIFF_VERSION_CLASSIC})
+  endif()
+
+  if(TIFF_BIGTIFF_VERSION AND NOT TIFF_VERSION_BIG)
+    set(TIFF_VERSION_BIG ${TIFF_BIGTIFF_VERSION})
+  endif()
+
+  if(NOT TIFF_VERSION_STRING AND TIFF_INCLUDE_DIR)
+    list(GET TIFF_INCLUDE_DIR 0 _TIFF_INCLUDE_DIR)
+    if(EXISTS "${_TIFF_INCLUDE_DIR}/tiffvers.h")
+      file(STRINGS "${_TIFF_INCLUDE_DIR}/tiffvers.h" tiff_version_str REGEX "^#define[\t ]+TIFFLIB_VERSION_STR[\t ]+\"LIBTIFF, Version .*")
+      string(REGEX REPLACE "^#define[\t ]+TIFFLIB_VERSION_STR[\t ]+\"LIBTIFF, Version +([^ \\n]*).*" "\\1" TIFF_VERSION_STRING "${tiff_version_str}")
+      unset(tiff_version_str)
+    endif()
+    unset(_TIFF_INCLUDE_DIR)
+  endif()
+
+  set(HAVE_TIFF YES)
 endif()
 
 # --- libjpeg (optional) ---
@@ -73,19 +74,20 @@ if(WITH_JPEG)
   else()
     include(FindJPEG)
   endif()
+
+  if(NOT JPEG_FOUND)
+    ocv_clear_vars(JPEG_LIBRARY JPEG_LIBRARIES JPEG_INCLUDE_DIR)
+
+    set(JPEG_LIBRARY libjpeg)
+    set(JPEG_LIBRARIES ${JPEG_LIBRARY})
+    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libjpeg")
+    set(JPEG_INCLUDE_DIR "${${JPEG_LIBRARY}_SOURCE_DIR}")
+  endif()
+
+  ocv_parse_header("${JPEG_INCLUDE_DIR}/jpeglib.h" JPEG_VERSION_LINES JPEG_LIB_VERSION)
+  set(HAVE_JPEG YES)
 endif()
 
-if(WITH_JPEG AND NOT JPEG_FOUND)
-  ocv_clear_vars(JPEG_LIBRARY JPEG_LIBRARIES JPEG_INCLUDE_DIR)
-
-  set(JPEG_LIBRARY libjpeg)
-  set(JPEG_LIBRARIES ${JPEG_LIBRARY})
-  add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libjpeg")
-  set(JPEG_INCLUDE_DIR "${${JPEG_LIBRARY}_SOURCE_DIR}")
-endif()
-
-ocv_parse_header("${JPEG_INCLUDE_DIR}/jpeglib.h" JPEG_VERSION_LINES JPEG_LIB_VERSION)
-
 # --- libwebp (optional) ---
 
 if(WITH_WEBP)
@@ -129,19 +131,21 @@ if(WITH_JASPER)
   else()
     include(FindJasper)
   endif()
-endif()
 
-if(WITH_JASPER AND NOT JASPER_FOUND)
-  ocv_clear_vars(JASPER_LIBRARY JASPER_LIBRARIES JASPER_INCLUDE_DIR)
+  if(NOT JASPER_FOUND)
+    ocv_clear_vars(JASPER_LIBRARY JASPER_LIBRARIES JASPER_INCLUDE_DIR)
 
-  set(JASPER_LIBRARY libjasper)
-  set(JASPER_LIBRARIES ${JASPER_LIBRARY})
-  add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libjasper")
-  set(JASPER_INCLUDE_DIR "${${JASPER_LIBRARY}_SOURCE_DIR}")
-endif()
+    set(JASPER_LIBRARY libjasper)
+    set(JASPER_LIBRARIES ${JASPER_LIBRARY})
+    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libjasper")
+    set(JASPER_INCLUDE_DIR "${${JASPER_LIBRARY}_SOURCE_DIR}")
+  endif()
 
-if(NOT JASPER_VERSION_STRING)
-  ocv_parse_header2(JASPER "${JASPER_INCLUDE_DIR}/jasper/jas_config.h" JAS_VERSION "")
+  set(HAVE_JASPER YES)
+
+  if(NOT JASPER_VERSION_STRING)
+    ocv_parse_header2(JASPER "${JASPER_INCLUDE_DIR}/jasper/jas_config.h" JAS_VERSION "")
+  endif()
 endif()
 
 # --- libpng (optional, should be searched after zlib) ---
@@ -151,30 +155,31 @@ if(WITH_PNG)
   else()
     include(FindPNG)
     if(PNG_FOUND)
-      check_include_file("${PNG_PNG_INCLUDE_DIR}/png.h"        HAVE_PNG_H)
+      include(CheckIncludeFile)
       check_include_file("${PNG_PNG_INCLUDE_DIR}/libpng/png.h" HAVE_LIBPNG_PNG_H)
-      if(HAVE_PNG_H)
-        ocv_parse_header("${PNG_PNG_INCLUDE_DIR}/png.h" PNG_VERSION_LINES PNG_LIBPNG_VER_MAJOR PNG_LIBPNG_VER_MINOR PNG_LIBPNG_VER_RELEASE)
-      elseif(HAVE_LIBPNG_PNG_H)
+      if(HAVE_LIBPNG_PNG_H)
         ocv_parse_header("${PNG_PNG_INCLUDE_DIR}/libpng/png.h" PNG_VERSION_LINES PNG_LIBPNG_VER_MAJOR PNG_LIBPNG_VER_MINOR PNG_LIBPNG_VER_RELEASE)
+      else()
+        ocv_parse_header("${PNG_PNG_INCLUDE_DIR}/png.h" PNG_VERSION_LINES PNG_LIBPNG_VER_MAJOR PNG_LIBPNG_VER_MINOR PNG_LIBPNG_VER_RELEASE)
       endif()
     endif()
   endif()
+
+  if(NOT PNG_FOUND)
+    ocv_clear_vars(PNG_LIBRARY PNG_LIBRARIES PNG_INCLUDE_DIR PNG_PNG_INCLUDE_DIR HAVE_LIBPNG_PNG_H PNG_DEFINITIONS)
+
+    set(PNG_LIBRARY libpng)
+    set(PNG_LIBRARIES ${PNG_LIBRARY})
+    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libpng")
+    set(PNG_INCLUDE_DIR "${${PNG_LIBRARY}_SOURCE_DIR}")
+    set(PNG_DEFINITIONS "")
+    ocv_parse_header("${PNG_INCLUDE_DIR}/png.h" PNG_VERSION_LINES PNG_LIBPNG_VER_MAJOR PNG_LIBPNG_VER_MINOR PNG_LIBPNG_VER_RELEASE)
+  endif()
+
+  set(HAVE_PNG YES)
+  set(PNG_VERSION "${PNG_LIBPNG_VER_MAJOR}.${PNG_LIBPNG_VER_MINOR}.${PNG_LIBPNG_VER_RELEASE}")
 endif()
 
-if(WITH_PNG AND NOT PNG_FOUND)
-  ocv_clear_vars(PNG_LIBRARY PNG_LIBRARIES PNG_INCLUDE_DIR PNG_PNG_INCLUDE_DIR HAVE_PNG_H HAVE_LIBPNG_PNG_H PNG_DEFINITIONS)
-
-  set(PNG_LIBRARY libpng)
-  set(PNG_LIBRARIES ${PNG_LIBRARY})
-  add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/libpng")
-  set(PNG_INCLUDE_DIR "${${PNG_LIBRARY}_SOURCE_DIR}")
-  set(PNG_DEFINITIONS "")
-  ocv_parse_header("${PNG_INCLUDE_DIR}/png.h" PNG_VERSION_LINES PNG_LIBPNG_VER_MAJOR PNG_LIBPNG_VER_MINOR PNG_LIBPNG_VER_RELEASE)
-endif()
-
-set(PNG_VERSION "${PNG_LIBPNG_VER_MAJOR}.${PNG_LIBPNG_VER_MINOR}.${PNG_LIBPNG_VER_RELEASE}")
-
 # --- OpenEXR (optional) ---
 if(WITH_OPENEXR)
   if(BUILD_OPENEXR)
@@ -182,17 +187,14 @@ if(WITH_OPENEXR)
   else()
     include("${OpenCV_SOURCE_DIR}/cmake/OpenCVFindOpenEXR.cmake")
   endif()
-endif()
 
-if(WITH_OPENEXR AND NOT OPENEXR_FOUND)
-  ocv_clear_vars(OPENEXR_INCLUDE_PATHS OPENEXR_LIBRARIES OPENEXR_ILMIMF_LIBRARY OPENEXR_VERSION)
+  if(NOT OPENEXR_FOUND)
+    ocv_clear_vars(OPENEXR_INCLUDE_PATHS OPENEXR_LIBRARIES OPENEXR_ILMIMF_LIBRARY OPENEXR_VERSION)
 
-  set(OPENEXR_LIBRARIES IlmImf)
-  set(OPENEXR_ILMIMF_LIBRARY IlmImf)
-  add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/openexr")
-endif()
+    set(OPENEXR_LIBRARIES IlmImf)
+    set(OPENEXR_ILMIMF_LIBRARY IlmImf)
+    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/openexr")
+  endif()
 
-#cmake 2.8.2 bug - it fails to determine zlib version
-if(ZLIB_FOUND)
-  ocv_parse_header2(ZLIB "${ZLIB_INCLUDE_DIR}/zlib.h" ZLIB_VERSION)
+  set(HAVE_OPENEXR YES)
 endif()
diff --git a/cmake/OpenCVFindLibsPerf.cmake b/cmake/OpenCVFindLibsPerf.cmake
index 6e497accd3..b8945c257b 100644
--- a/cmake/OpenCVFindLibsPerf.cmake
+++ b/cmake/OpenCVFindLibsPerf.cmake
@@ -27,7 +27,7 @@ endif(WITH_CUDA)
 # --- Eigen ---
 if(WITH_EIGEN)
   find_path(EIGEN_INCLUDE_PATH "Eigen/Core"
-            PATHS /usr/local /opt /usr $ENV{EIGEN_ROOT}/include ENV ProgramFiles ENV ProgramW6432 
+            PATHS /usr/local /opt /usr $ENV{EIGEN_ROOT}/include ENV ProgramFiles ENV ProgramW6432
             PATH_SUFFIXES include/eigen3 include/eigen2 Eigen/include/eigen3 Eigen/include/eigen2
             DOC "The path to Eigen3/Eigen2 headers"
             CMAKE_FIND_ROOT_PATH_BOTH)
@@ -88,8 +88,9 @@ endif()
 # --- OpenMP ---
 if(NOT HAVE_TBB AND NOT HAVE_CSTRIPES)
   set(_fname "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/omptest.cpp")
-  FILE(WRITE "${_fname}" "#ifndef _OPENMP\n#error\n#endif\nint main() { return 0; }\n")
-  TRY_COMPILE(HAVE_OPENMP "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp" "${_fname}")
+  file(WRITE "${_fname}" "#ifndef _OPENMP\n#error\n#endif\nint main() { return 0; }\n")
+  try_compile(HAVE_OPENMP "${CMAKE_BINARY_DIR}" "${_fname}")
+  file(REMOVE "${_fname}")
 else()
   set(HAVE_OPENMP 0)
 endif()
@@ -104,8 +105,9 @@ endif()
 # --- Concurrency ---
 if(MSVC AND NOT HAVE_TBB AND NOT HAVE_CSTRIPES AND NOT HAVE_OPENMP)
   set(_fname "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/concurrencytest.cpp")
-  FILE(WRITE "${_fname}" "#if _MSC_VER < 1600\n#error\n#endif\nint main() { return 0; }\n")
-  TRY_COMPILE(HAVE_CONCURRENCY "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp" "${_fname}")
+  file(WRITE "${_fname}" "#if _MSC_VER < 1600\n#error\n#endif\nint main() { return 0; }\n")
+  try_compile(HAVE_CONCURRENCY "${CMAKE_BINARY_DIR}" "${_fname}")
+  file(REMOVE "${_fname}")
 else()
   set(HAVE_CONCURRENCY 0)
 endif()
diff --git a/cmake/OpenCVFindLibsVideo.cmake b/cmake/OpenCVFindLibsVideo.cmake
index 414918527b..eba6fa988c 100644
--- a/cmake/OpenCVFindLibsVideo.cmake
+++ b/cmake/OpenCVFindLibsVideo.cmake
@@ -2,17 +2,54 @@
 #  Detect 3rd-party video IO libraries
 # ----------------------------------------------------------------------------
 
+ocv_clear_vars(HAVE_VFW)
+if(WITH_VFW)
+  try_compile(HAVE_VFW
+    "${OpenCV_BINARY_DIR}"
+    "${OpenCV_SOURCE_DIR}/cmake/checks/vfwtest.cpp"
+    CMAKE_FLAGS "-DLINK_LIBRARIES:STRING=vfw32")
+endif(WITH_VFW)
+
 # --- GStreamer ---
 ocv_clear_vars(HAVE_GSTREAMER)
-if(WITH_GSTREAMER)
-  CHECK_MODULE(gstreamer-base-0.10 HAVE_GSTREAMER)
-  if(HAVE_GSTREAMER)
-    CHECK_MODULE(gstreamer-app-0.10 HAVE_GSTREAMER)
+# try to find gstreamer 0.10 first
+if(WITH_GSTREAMER AND NOT WITH_GSTREAMER_1_X)
+  CHECK_MODULE(gstreamer-base-0.10 HAVE_GSTREAMER_BASE)
+  CHECK_MODULE(gstreamer-video-0.10 HAVE_GSTREAMER_VIDEO)
+  CHECK_MODULE(gstreamer-app-0.10 HAVE_GSTREAMER_APP)
+  CHECK_MODULE(gstreamer-riff-0.10 HAVE_GSTREAMER_RIFF)
+  CHECK_MODULE(gstreamer-pbutils-0.10 HAVE_GSTREAMER_PBUTILS)
+
+  if(HAVE_GSTREAMER_BASE AND HAVE_GSTREAMER_VIDEO AND HAVE_GSTREAMER_APP AND HAVE_GSTREAMER_RIFF AND HAVE_GSTREAMER_PBUTILS)
+      set(HAVE_GSTREAMER TRUE)
+      set(GSTREAMER_BASE_VERSION ${ALIASOF_gstreamer-base-0.10_VERSION})
+      set(GSTREAMER_VIDEO_VERSION ${ALIASOF_gstreamer-video-0.10_VERSION})
+      set(GSTREAMER_APP_VERSION ${ALIASOF_gstreamer-app-0.10_VERSION})
+      set(GSTREAMER_RIFF_VERSION ${ALIASOF_gstreamer-riff-0.10_VERSION})
+      set(GSTREAMER_PBUTILS_VERSION ${ALIASOF_gstreamer-pbutils-0.10_VERSION})
   endif()
-  if(HAVE_GSTREAMER)
-    CHECK_MODULE(gstreamer-video-0.10 HAVE_GSTREAMER)
+
+endif(WITH_GSTREAMER AND NOT WITH_GSTREAMER_1_X)
+
+# if gstreamer 0.10 was not found, or we specified we wanted 1.x, try to find it
+if(WITH_GSTREAMER_1_X OR NOT HAVE_GSTREAMER)
+  #check for 1.x
+  CHECK_MODULE(gstreamer-base-1.0 HAVE_GSTREAMER_BASE)
+  CHECK_MODULE(gstreamer-video-1.0 HAVE_GSTREAMER_VIDEO)
+  CHECK_MODULE(gstreamer-app-1.0 HAVE_GSTREAMER_APP)
+  CHECK_MODULE(gstreamer-riff-1.0 HAVE_GSTREAMER_RIFF)
+  CHECK_MODULE(gstreamer-pbutils-1.0 HAVE_GSTREAMER_PBUTILS)
+
+  if(HAVE_GSTREAMER_BASE AND HAVE_GSTREAMER_VIDEO AND HAVE_GSTREAMER_APP AND HAVE_GSTREAMER_RIFF AND HAVE_GSTREAMER_PBUTILS)
+      set(HAVE_GSTREAMER TRUE)
+      set(GSTREAMER_BASE_VERSION ${ALIASOF_gstreamer-base-1.0_VERSION})
+      set(GSTREAMER_VIDEO_VERSION ${ALIASOF_gstreamer-video-1.0_VERSION})
+      set(GSTREAMER_APP_VERSION ${ALIASOF_gstreamer-app-1.0_VERSION})
+      set(GSTREAMER_RIFF_VERSION ${ALIASOF_gstreamer-riff-1.0_VERSION})
+      set(GSTREAMER_PBUTILS_VERSION ${ALIASOF_gstreamer-pbutils-1.0_VERSION})
   endif()
-endif(WITH_GSTREAMER)
+
+endif(WITH_GSTREAMER_1_X OR NOT HAVE_GSTREAMER)
 
 # --- unicap ---
 ocv_clear_vars(HAVE_UNICAP)
@@ -37,7 +74,7 @@ if(WITH_PVAPI)
       set(PVAPI_SDK_SUBDIR x86)
     elseif(X86_64)
       set(PVAPI_SDK_SUBDIR x64)
-    elseif(CMAKE_SYSTEM_PROCESSOR MATCHES arm)
+    elseif(ARM)
       set(PVAPI_SDK_SUBDIR arm)
     endif()
 
@@ -49,7 +86,14 @@ if(WITH_PVAPI)
       set(_PVAPI_LIBRARY "${_PVAPI_LIBRARY}/${CMAKE_OPENCV_GCC_VERSION_MAJOR}.${CMAKE_OPENCV_GCC_VERSION_MINOR}")
     endif()
 
-    set(PVAPI_LIBRARY "${_PVAPI_LIBRARY}/${CMAKE_STATIC_LIBRARY_PREFIX}PvAPI${CMAKE_STATIC_LIBRARY_SUFFIX}" CACHE PATH "The PvAPI library")
+    if(WIN32)
+      if(MINGW)
+        set(PVAPI_DEFINITIONS "-DPVDECL=__stdcall")
+      endif(MINGW)
+      set(PVAPI_LIBRARY "${_PVAPI_LIBRARY}/PvAPI.lib" CACHE PATH "The PvAPI library")
+    else(WIN32)
+      set(PVAPI_LIBRARY "${_PVAPI_LIBRARY}/${CMAKE_STATIC_LIBRARY_PREFIX}PvAPI${CMAKE_STATIC_LIBRARY_SUFFIX}" CACHE PATH "The PvAPI library")
+    endif(WIN32)
     if(EXISTS "${PVAPI_LIBRARY}")
       set(HAVE_PVAPI TRUE)
     endif()
@@ -72,10 +116,33 @@ endif(WITH_GIGEAPI)
 # --- Dc1394 ---
 ocv_clear_vars(HAVE_DC1394 HAVE_DC1394_2)
 if(WITH_1394)
-  CHECK_MODULE(libdc1394-2 HAVE_DC1394_2)
-  if(NOT HAVE_DC1394_2)
-    CHECK_MODULE(libdc1394 HAVE_DC1394)
-  endif()
+  if(WIN32 AND MINGW)
+      find_path(CMU1394_INCLUDE_PATH "/1394common.h"
+                PATH_SUFFIXES include
+                DOC "The path to cmu1394 headers")
+      find_path(DC1394_2_INCLUDE_PATH "/dc1394/dc1394.h"
+                PATH_SUFFIXES include
+                DOC "The path to DC1394 2.x headers")
+      if(CMU1394_INCLUDE_PATH AND DC1394_2_INCLUDE_PATH)
+        set(CMU1394_LIB_DIR  "${CMU1394_INCLUDE_PATH}/../lib"  CACHE PATH "Full path of CMU1394 library directory")
+        set(DC1394_2_LIB_DIR "${DC1394_2_INCLUDE_PATH}/../lib" CACHE PATH "Full path of DC1394 2.x library directory")
+        if(EXISTS "${CMU1394_LIB_DIR}/lib1394camera.a" AND EXISTS "${DC1394_2_LIB_DIR}/libdc1394.a")
+          set(HAVE_DC1394_2 TRUE)
+        endif()
+      endif()
+      if(HAVE_DC1394_2)
+        ocv_parse_pkg("libdc1394-2" "${DC1394_2_LIB_DIR}/pkgconfig" "")
+        ocv_include_directories(${DC1394_2_INCLUDE_PATH})
+        set(HIGHGUI_LIBRARIES ${HIGHGUI_LIBRARIES}
+            "${DC1394_2_LIB_DIR}/libdc1394.a"
+            "${CMU1394_LIB_DIR}/lib1394camera.a")
+      endif(HAVE_DC1394_2)
+  else(WIN32 AND MINGW)
+    CHECK_MODULE(libdc1394-2 HAVE_DC1394_2)
+    if(NOT HAVE_DC1394_2)
+      CHECK_MODULE(libdc1394 HAVE_DC1394)
+    endif()
+  endif(WIN32 AND MINGW)
 endif(WITH_1394)
 
 # --- xine ---
@@ -111,7 +178,7 @@ endif(WITH_XIMEA)
 # --- FFMPEG ---
 ocv_clear_vars(HAVE_FFMPEG HAVE_FFMPEG_CODEC HAVE_FFMPEG_FORMAT HAVE_FFMPEG_UTIL HAVE_FFMPEG_SWSCALE HAVE_GENTOO_FFMPEG HAVE_FFMPEG_FFMPEG)
 if(WITH_FFMPEG)
-  if(WIN32)
+  if(WIN32 AND NOT ARM)
     include("${OpenCV_SOURCE_DIR}/3rdparty/ffmpeg/ffmpeg_version.cmake")
   elseif(UNIX)
     CHECK_MODULE(libavcodec HAVE_FFMPEG_CODEC)
@@ -175,15 +242,20 @@ if(WITH_FFMPEG)
   endif(APPLE)
 endif(WITH_FFMPEG)
 
-# --- VideoInput ---
-if(WITH_VIDEOINPUT)
+# --- VideoInput/DirectShow ---
+if(WITH_DSHOW)
   # always have VideoInput on Windows
-  set(HAVE_VIDEOINPUT 1)
-endif(WITH_VIDEOINPUT)
+  set(HAVE_DSHOW 1)
+endif(WITH_DSHOW)
+
+# --- VideoInput/Microsoft Media Foundation ---
+if(WITH_MSMF)
+  check_include_file(Mfapi.h HAVE_MSMF)
+endif(WITH_MSMF)
 
 # --- Extra HighGUI libs on Windows ---
 if(WIN32)
-  list(APPEND HIGHGUI_LIBRARIES comctl32 gdi32 ole32 vfw32)
+  list(APPEND HIGHGUI_LIBRARIES comctl32 gdi32 ole32 setupapi ws2_32 vfw32)
   if(MINGW64)
     list(APPEND HIGHGUI_LIBRARIES avifil32 avicap32 winmm msvfw32)
     list(REMOVE_ITEM HIGHGUI_LIBRARIES vfw32)
@@ -191,3 +263,17 @@ if(WIN32)
     list(APPEND HIGHGUI_LIBRARIES winmm)
   endif()
 endif(WIN32)
+
+# --- Apple AV Foundation ---
+if(WITH_AVFOUNDATION)
+  set(HAVE_AVFOUNDATION YES)
+endif()
+
+# --- QuickTime ---
+if (NOT IOS)
+  if(WITH_QUICKTIME)
+    set(HAVE_QUICKTIME YES)
+  elseif(APPLE)
+    set(HAVE_QTKIT YES)
+  endif()
+endif()
diff --git a/cmake/OpenCVFindPkgConfig.cmake b/cmake/OpenCVFindPkgConfig.cmake
deleted file mode 100644
index 7e439fd8dc..0000000000
--- a/cmake/OpenCVFindPkgConfig.cmake
+++ /dev/null
@@ -1,365 +0,0 @@
-#
-# OpenCV note: the file has been extracted from CMake 2.6.2 distribution.
-# It is used to build OpenCV with CMake 2.4.x
-#
-
-# - a pkg-config module for CMake
-#
-# Usage:
-#   pkg_check_modules(<PREFIX> [REQUIRED] <MODULE> [<MODULE>]*)
-#     checks for all the given modules
-#
-#   pkg_search_module(<PREFIX> [REQUIRED] <MODULE> [<MODULE>]*)
-#     checks for given modules and uses the first working one
-#
-# When the 'REQUIRED' argument was set, macros will fail with an error
-# when module(s) could not be found
-#
-# It sets the following variables:
-#   PKG_CONFIG_FOUND         ... true if pkg-config works on the system
-#   PKG_CONFIG_EXECUTABLE    ... pathname of the pkg-config program
-#   <PREFIX>_FOUND           ... set to 1 if module(s) exist
-#
-# For the following variables two sets of values exist; first one is the
-# common one and has the given PREFIX. The second set contains flags
-# which are given out when pkgconfig was called with the '--static'
-# option.
-#   <XPREFIX>_LIBRARIES      ... only the libraries (w/o the '-l')
-#   <XPREFIX>_LIBRARY_DIRS   ... the paths of the libraries (w/o the '-L')
-#   <XPREFIX>_LDFLAGS        ... all required linker flags
-#   <XPREFIX>_LDFLAGS_OTHER  ... all other linker flags
-#   <XPREFIX>_INCLUDE_DIRS   ... the '-I' preprocessor flags (w/o the '-I')
-#   <XPREFIX>_CFLAGS         ... all required cflags
-#   <XPREFIX>_CFLAGS_OTHER   ... the other compiler flags
-#
-#   <XPREFIX> = <PREFIX>        for common case
-#   <XPREFIX> = <PREFIX>_STATIC for static linking
-#
-# There are some special variables whose prefix depends on the count
-# of given modules. When there is only one module, <PREFIX> stays
-# unchanged. When there are multiple modules, the prefix will be
-# changed to <PREFIX>_<MODNAME>:
-#   <XPREFIX>_VERSION    ... version of the module
-#   <XPREFIX>_PREFIX     ... prefix-directory of the module
-#   <XPREFIX>_INCLUDEDIR ... include-dir of the module
-#   <XPREFIX>_LIBDIR     ... lib-dir of the module
-#
-#   <XPREFIX> = <PREFIX>  when |MODULES| == 1, else
-#   <XPREFIX> = <PREFIX>_<MODNAME>
-#
-# A <MODULE> parameter can have the following formats:
-#   {MODNAME}            ... matches any version
-#   {MODNAME}>={VERSION} ... at least version <VERSION> is required
-#   {MODNAME}={VERSION}  ... exactly version <VERSION> is required
-#   {MODNAME}<={VERSION} ... modules must not be newer than <VERSION>
-#
-# Examples
-#   pkg_check_modules (GLIB2   glib-2.0)
-#
-#   pkg_check_modules (GLIB2   glib-2.0>=2.10)
-#     requires at least version 2.10 of glib2 and defines e.g.
-#       GLIB2_VERSION=2.10.3
-#
-#   pkg_check_modules (FOO     glib-2.0>=2.10 gtk+-2.0)
-#     requires both glib2 and gtk2, and defines e.g.
-#       FOO_glib-2.0_VERSION=2.10.3
-#       FOO_gtk+-2.0_VERSION=2.8.20
-#
-#   pkg_check_modules (XRENDER REQUIRED xrender)
-#     defines e.g.:
-#       XRENDER_LIBRARIES=Xrender;X11
-#       XRENDER_STATIC_LIBRARIES=Xrender;X11;pthread;Xau;Xdmcp
-#
-#   pkg_search_module (BAR     libxml-2.0 libxml2 libxml>=2)
-
-
-# Copyright (C) 2006 Enrico Scholz <enrico.scholz@informatik.tu-chemnitz.de>
-#
-# Redistribution and use, with or without modification, are permitted
-# provided that the following conditions are met:
-#
-#    1. Redistributions must retain the above copyright notice, this
-#       list of conditions and the following disclaimer.
-#    2. The name of the author may not be used to endorse or promote
-#       products derived from this software without specific prior
-#       written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
-# IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
-# ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
-# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
-# GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
-# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
-# IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
-# OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
-# IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-
-### Common stuff ####
-set(PKG_CONFIG_VERSION 1)
-set(PKG_CONFIG_FOUND   0)
-
-find_program(PKG_CONFIG_EXECUTABLE NAMES pkg-config DOC "pkg-config executable")
-mark_as_advanced(PKG_CONFIG_EXECUTABLE)
-
-if(PKG_CONFIG_EXECUTABLE)
-  set(PKG_CONFIG_FOUND 1)
-endif(PKG_CONFIG_EXECUTABLE)
-
-
-# Unsets the given variables
-macro(_pkgconfig_unset var)
-  set(${var} "" CACHE INTERNAL "")
-endmacro(_pkgconfig_unset)
-
-macro(_pkgconfig_set var value)
-  set(${var} ${value} CACHE INTERNAL "")
-endmacro(_pkgconfig_set)
-
-# Invokes pkgconfig, cleans up the result and sets variables
-macro(_pkgconfig_invoke _pkglist _prefix _varname _regexp)
-  set(_pkgconfig_invoke_result)
-
-  execute_process(
-    COMMAND ${PKG_CONFIG_EXECUTABLE} ${ARGN} ${_pkglist}
-    OUTPUT_VARIABLE _pkgconfig_invoke_result
-    RESULT_VARIABLE _pkgconfig_failed)
-
-  if (_pkgconfig_failed)
-    set(_pkgconfig_${_varname} "")
-    _pkgconfig_unset(${_prefix}_${_varname})
-  else(_pkgconfig_failed)
-    string(REGEX REPLACE "[\r\n]"                  " " _pkgconfig_invoke_result "${_pkgconfig_invoke_result}")
-    string(REGEX REPLACE " +$"                     ""  _pkgconfig_invoke_result "${_pkgconfig_invoke_result}")
-
-    if (NOT ${_regexp} STREQUAL "")
-      string(REGEX REPLACE "${_regexp}" " " _pkgconfig_invoke_result "${_pkgconfig_invoke_result}")
-    endif(NOT ${_regexp} STREQUAL "")
-
-    separate_arguments(_pkgconfig_invoke_result)
-
-    #message(STATUS "  ${_varname} ... ${_pkgconfig_invoke_result}")
-    set(_pkgconfig_${_varname} ${_pkgconfig_invoke_result})
-    _pkgconfig_set(${_prefix}_${_varname} "${_pkgconfig_invoke_result}")
-  endif(_pkgconfig_failed)
-endmacro(_pkgconfig_invoke)
-
-# Invokes pkgconfig two times; once without '--static' and once with
-# '--static'
-macro(_pkgconfig_invoke_dyn _pkglist _prefix _varname cleanup_regexp)
-  _pkgconfig_invoke("${_pkglist}" ${_prefix}        ${_varname} "${cleanup_regexp}" ${ARGN})
-  _pkgconfig_invoke("${_pkglist}" ${_prefix} STATIC_${_varname} "${cleanup_regexp}" --static  ${ARGN})
-endmacro(_pkgconfig_invoke_dyn)
-
-# Splits given arguments into options and a package list
-macro(_pkgconfig_parse_options _result _is_req)
-  set(${_is_req} 0)
-
-  foreach(_pkg ${ARGN})
-    if (_pkg STREQUAL "REQUIRED")
-      set(${_is_req} 1)
-    endif (_pkg STREQUAL "REQUIRED")
-  endforeach(_pkg ${ARGN})
-
-  set(${_result} ${ARGN})
-  list(REMOVE_ITEM ${_result} "REQUIRED")
-endmacro(_pkgconfig_parse_options)
-
-###
-macro(_pkg_check_modules_internal _is_required _is_silent _prefix)
-  _pkgconfig_unset(${_prefix}_FOUND)
-  _pkgconfig_unset(${_prefix}_VERSION)
-  _pkgconfig_unset(${_prefix}_PREFIX)
-  _pkgconfig_unset(${_prefix}_INCLUDEDIR)
-  _pkgconfig_unset(${_prefix}_LIBDIR)
-  _pkgconfig_unset(${_prefix}_LIBS)
-  _pkgconfig_unset(${_prefix}_LIBS_L)
-  _pkgconfig_unset(${_prefix}_LIBS_PATHS)
-  _pkgconfig_unset(${_prefix}_LIBS_OTHER)
-  _pkgconfig_unset(${_prefix}_CFLAGS)
-  _pkgconfig_unset(${_prefix}_CFLAGS_I)
-  _pkgconfig_unset(${_prefix}_CFLAGS_OTHER)
-  _pkgconfig_unset(${_prefix}_STATIC_LIBDIR)
-  _pkgconfig_unset(${_prefix}_STATIC_LIBS)
-  _pkgconfig_unset(${_prefix}_STATIC_LIBS_L)
-  _pkgconfig_unset(${_prefix}_STATIC_LIBS_PATHS)
-  _pkgconfig_unset(${_prefix}_STATIC_LIBS_OTHER)
-  _pkgconfig_unset(${_prefix}_STATIC_CFLAGS)
-  _pkgconfig_unset(${_prefix}_STATIC_CFLAGS_I)
-  _pkgconfig_unset(${_prefix}_STATIC_CFLAGS_OTHER)
-
-  # create a better addressable variable of the modules and calculate its size
-  set(_pkg_check_modules_list ${ARGN})
-  list(LENGTH _pkg_check_modules_list _pkg_check_modules_cnt)
-
-  if(PKG_CONFIG_EXECUTABLE)
-    # give out status message telling checked module
-    if (NOT ${_is_silent})
-      if (_pkg_check_modules_cnt EQUAL 1)
-        message(STATUS "checking for module '${_pkg_check_modules_list}'")
-      else(_pkg_check_modules_cnt EQUAL 1)
-        message(STATUS "checking for modules '${_pkg_check_modules_list}'")
-      endif(_pkg_check_modules_cnt EQUAL 1)
-    endif(NOT ${_is_silent})
-
-    set(_pkg_check_modules_packages)
-    set(_pkg_check_modules_failed)
-
-    # iterate through module list and check whether they exist and match the required version
-    foreach (_pkg_check_modules_pkg ${_pkg_check_modules_list})
-      set(_pkg_check_modules_exist_query)
-
-      # check whether version is given
-      if (_pkg_check_modules_pkg MATCHES ".*(>=|=|<=).*")
-        string(REGEX REPLACE "(.*[^><])(>=|=|<=)(.*)" "\\1" _pkg_check_modules_pkg_name "${_pkg_check_modules_pkg}")
-        string(REGEX REPLACE "(.*[^><])(>=|=|<=)(.*)" "\\2" _pkg_check_modules_pkg_op   "${_pkg_check_modules_pkg}")
-        string(REGEX REPLACE "(.*[^><])(>=|=|<=)(.*)" "\\3" _pkg_check_modules_pkg_ver  "${_pkg_check_modules_pkg}")
-      else(_pkg_check_modules_pkg MATCHES ".*(>=|=|<=).*")
-        set(_pkg_check_modules_pkg_name "${_pkg_check_modules_pkg}")
-        set(_pkg_check_modules_pkg_op)
-        set(_pkg_check_modules_pkg_ver)
-      endif(_pkg_check_modules_pkg MATCHES ".*(>=|=|<=).*")
-
-      # handle the operands
-      if (_pkg_check_modules_pkg_op STREQUAL ">=")
-        list(APPEND _pkg_check_modules_exist_query --atleast-version)
-      endif(_pkg_check_modules_pkg_op STREQUAL ">=")
-
-      if (_pkg_check_modules_pkg_op STREQUAL "=")
-        list(APPEND _pkg_check_modules_exist_query --exact-version)
-      endif(_pkg_check_modules_pkg_op STREQUAL "=")
-
-      if (_pkg_check_modules_pkg_op STREQUAL "<=")
-        list(APPEND _pkg_check_modules_exist_query --max-version)
-      endif(_pkg_check_modules_pkg_op STREQUAL "<=")
-
-      # create the final query which is of the format:
-      # * --atleast-version <version> <pkg-name>
-      # * --exact-version <version> <pkg-name>
-      # * --max-version <version> <pkg-name>
-      # * --exists <pkg-name>
-      if (_pkg_check_modules_pkg_op)
-        list(APPEND _pkg_check_modules_exist_query "${_pkg_check_modules_pkg_ver}")
-      else(_pkg_check_modules_pkg_op)
-        list(APPEND _pkg_check_modules_exist_query --exists)
-      endif(_pkg_check_modules_pkg_op)
-
-      _pkgconfig_unset(${_prefix}_${_pkg_check_modules_pkg_name}_VERSION)
-      _pkgconfig_unset(${_prefix}_${_pkg_check_modules_pkg_name}_PREFIX)
-      _pkgconfig_unset(${_prefix}_${_pkg_check_modules_pkg_name}_INCLUDEDIR)
-      _pkgconfig_unset(${_prefix}_${_pkg_check_modules_pkg_name}_LIBDIR)
-
-      list(APPEND _pkg_check_modules_exist_query "${_pkg_check_modules_pkg_name}")
-      list(APPEND _pkg_check_modules_packages    "${_pkg_check_modules_pkg_name}")
-
-      # execute the query
-      execute_process(
-        COMMAND ${PKG_CONFIG_EXECUTABLE} ${_pkg_check_modules_exist_query}
-        RESULT_VARIABLE _pkgconfig_retval)
-
-      # evaluate result and tell failures
-      if (_pkgconfig_retval)
-        if(NOT ${_is_silent})
-          message(STATUS "  package '${_pkg_check_modules_pkg}' not found")
-        endif(NOT ${_is_silent})
-
-        set(_pkg_check_modules_failed 1)
-      endif(_pkgconfig_retval)
-    endforeach(_pkg_check_modules_pkg)
-
-    if(_pkg_check_modules_failed)
-      # fail when requested
-      if (${_is_required})
-        message(SEND_ERROR "A required package was not found")
-      endif (${_is_required})
-    else(_pkg_check_modules_failed)
-      # when we are here, we checked whether requested modules
-      # exist. Now, go through them and set variables
-
-      _pkgconfig_set(${_prefix}_FOUND 1)
-      list(LENGTH _pkg_check_modules_packages pkg_count)
-
-      # iterate through all modules again and set individual variables
-      foreach (_pkg_check_modules_pkg ${_pkg_check_modules_packages})
-        # handle case when there is only one package required
-        if (pkg_count EQUAL 1)
-          set(_pkg_check_prefix "${_prefix}")
-        else(pkg_count EQUAL 1)
-          set(_pkg_check_prefix "${_prefix}_${_pkg_check_modules_pkg}")
-        endif(pkg_count EQUAL 1)
-
-        _pkgconfig_invoke(${_pkg_check_modules_pkg} "${_pkg_check_prefix}" VERSION    ""   --modversion )
-        _pkgconfig_invoke(${_pkg_check_modules_pkg} "${_pkg_check_prefix}" PREFIX     ""   --variable=prefix )
-        _pkgconfig_invoke(${_pkg_check_modules_pkg} "${_pkg_check_prefix}" INCLUDEDIR ""   --variable=includedir )
-        _pkgconfig_invoke(${_pkg_check_modules_pkg} "${_pkg_check_prefix}" LIBDIR     ""   --variable=libdir )
-
-        message(STATUS "  found ${_pkg_check_modules_pkg}, version ${_pkgconfig_VERSION}")
-      endforeach(_pkg_check_modules_pkg)
-
-      # set variables which are combined for multiple modules
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" LIBRARIES           "(^| )-l" --libs-only-l )
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" LIBRARY_DIRS        "(^| )-L" --libs-only-L )
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" LDFLAGS             ""        --libs )
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" LDFLAGS_OTHER       ""        --libs-only-other )
-
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" INCLUDE_DIRS        "(^| )-I" --cflags-only-I )
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" CFLAGS              ""        --cflags )
-      _pkgconfig_invoke_dyn("${_pkg_check_modules_packages}" "${_prefix}" CFLAGS_OTHER        ""        --cflags-only-other )
-    endif(_pkg_check_modules_failed)
-  else(PKG_CONFIG_EXECUTABLE)
-    if (${_is_required})
-      message(SEND_ERROR "pkg-config tool not found")
-    endif (${_is_required})
-  endif(PKG_CONFIG_EXECUTABLE)
-endmacro(_pkg_check_modules_internal)
-
-###
-### User visible macros start here
-###
-
-###
-macro(pkg_check_modules _prefix _module0)
-  # check cached value
-  if (NOT DEFINED __pkg_config_checked_${_prefix} OR __pkg_config_checked_${_prefix} LESS ${PKG_CONFIG_VERSION} OR NOT ${_prefix}_FOUND)
-    _pkgconfig_parse_options   (_pkg_modules _pkg_is_required "${_module0}" ${ARGN})
-    _pkg_check_modules_internal("${_pkg_is_required}" 0 "${_prefix}" ${_pkg_modules})
-
-    _pkgconfig_set(__pkg_config_checked_${_prefix} ${PKG_CONFIG_VERSION})
-  endif(NOT DEFINED __pkg_config_checked_${_prefix} OR __pkg_config_checked_${_prefix} LESS ${PKG_CONFIG_VERSION} OR NOT ${_prefix}_FOUND)
-endmacro(pkg_check_modules)
-
-###
-macro(pkg_search_module _prefix _module0)
-  # check cached value
-  if (NOT DEFINED __pkg_config_checked_${_prefix} OR __pkg_config_checked_${_prefix} LESS ${PKG_CONFIG_VERSION} OR NOT ${_prefix}_FOUND)
-    set(_pkg_modules_found 0)
-    _pkgconfig_parse_options(_pkg_modules_alt _pkg_is_required "${_module0}" ${ARGN})
-
-    message(STATUS "checking for one of the modules '${_pkg_modules_alt}'")
-
-    # iterate through all modules and stop at the first working one.
-    foreach(_pkg_alt ${_pkg_modules_alt})
-      if(NOT _pkg_modules_found)
-        _pkg_check_modules_internal(0 1 "${_prefix}" "${_pkg_alt}")
-      endif(NOT _pkg_modules_found)
-
-      if (${_prefix}_FOUND)
-        set(_pkg_modules_found 1)
-      endif(${_prefix}_FOUND)
-    endforeach(_pkg_alt)
-
-    if (NOT ${_prefix}_FOUND)
-      if(${_pkg_is_required})
-        message(SEND_ERROR "None of the required '${_pkg_modules_alt}' found")
-      endif(${_pkg_is_required})
-    endif(NOT ${_prefix}_FOUND)
-
-    _pkgconfig_set(__pkg_config_checked_${_prefix} ${PKG_CONFIG_VERSION})
-  endif(NOT DEFINED __pkg_config_checked_${_prefix} OR __pkg_config_checked_${_prefix} LESS ${PKG_CONFIG_VERSION} OR NOT ${_prefix}_FOUND)
-endmacro(pkg_search_module)
-
-### Local Variables:
-### mode: cmake
-### End:
diff --git a/cmake/OpenCVFindXimea.cmake b/cmake/OpenCVFindXimea.cmake
index 5600275f47..6b86b609e9 100644
--- a/cmake/OpenCVFindXimea.cmake
+++ b/cmake/OpenCVFindXimea.cmake
@@ -9,6 +9,7 @@
 #
 # Created: 5 Aug 2011 by Marian Zajko (marian.zajko@ximea.com)
 # Updated: 25 June 2012 by Igor Kuzmin (parafin@ximea.com)
+# Updated: 22 October 2012 by Marian Zajko (marian.zajko@ximea.com)
 #
 
 set(XIMEA_FOUND)
@@ -19,10 +20,14 @@ if(WIN32)
   # Try to find the XIMEA API path in registry.
   GET_FILENAME_COMPONENT(XIMEA_PATH "[HKEY_CURRENT_USER\\Software\\XIMEA\\CamSupport\\API;Path]" ABSOLUTE)
 
-  if(EXISTS XIMEA_PATH)
+  if(EXISTS ${XIMEA_PATH})
     set(XIMEA_FOUND 1)
     # set LIB folders
-    set(XIMEA_LIBRARY_DIR "${XIMEA_PATH}/x86")
+    if(CMAKE_CL_64)
+      set(XIMEA_LIBRARY_DIR "${XIMEA_PATH}/x64")
+    else()
+      set(XIMEA_LIBRARY_DIR "${XIMEA_PATH}/x86")
+    endif()
   else()
     set(XIMEA_FOUND 0)
   endif()
@@ -39,4 +44,3 @@ endif()
 mark_as_advanced(FORCE XIMEA_FOUND)
 mark_as_advanced(FORCE XIMEA_PATH)
 mark_as_advanced(FORCE XIMEA_LIBRARY_DIR)
-
diff --git a/cmake/OpenCVGenConfig.cmake b/cmake/OpenCVGenConfig.cmake
index 705ccc8df1..7ccf3ef42e 100644
--- a/cmake/OpenCVGenConfig.cmake
+++ b/cmake/OpenCVGenConfig.cmake
@@ -84,7 +84,7 @@ macro(ocv_generate_dependencies_map_configcmake suffix configuration)
 
     set(OPENCV_DEPENDENCIES_MAP_${suffix} "${OPENCV_DEPENDENCIES_MAP_${suffix}}set(OpenCV_${__ocv_lib}_LIBNAME_${suffix} \"${__libname}\")\n")
     set(OPENCV_DEPENDENCIES_MAP_${suffix} "${OPENCV_DEPENDENCIES_MAP_${suffix}}set(OpenCV_${__ocv_lib}_DEPS_${suffix} ${__mod_deps})\n")
-    set(OPENCV_DEPENDENCIES_MAP_${suffix} "${OPENCV_DEPENDENCIES_MAP_${suffix}}set(OpenCV_${__ocv_lib}_EXTRA_DEPS_${suffix} ${__ext_deps})\n")
+    set(OPENCV_DEPENDENCIES_MAP_${suffix} "${OPENCV_DEPENDENCIES_MAP_${suffix}}set(OpenCV_${__ocv_lib}_EXTRA_DEPS_${suffix} \"${__ext_deps}\")\n")
 
     list(APPEND OPENCV_PROCESSED_LIBS ${__ocv_lib})
     list(APPEND OPENCV_LIBS_TO_PROCESS ${${__ocv_lib}_MODULE_DEPS_${suffix}})
@@ -162,7 +162,7 @@ if(UNIX)
 endif()
 
 if(ANDROID)
-  install(FILES "${OpenCV_SOURCE_DIR}/android/android.toolchain.cmake" DESTINATION ${OPENCV_CONFIG_INSTALL_PATH}/)
+  install(FILES "${OpenCV_SOURCE_DIR}/platforms/android/android.toolchain.cmake" DESTINATION ${OPENCV_CONFIG_INSTALL_PATH}/)
 endif()
 
 # --------------------------------------------------------------------------------------------
diff --git a/cmake/OpenCVGenHeaders.cmake b/cmake/OpenCVGenHeaders.cmake
index d2d38c90a7..c7129fefa3 100644
--- a/cmake/OpenCVGenHeaders.cmake
+++ b/cmake/OpenCVGenHeaders.cmake
@@ -1,13 +1,3 @@
-# ----------------------------------------------------------------------------
-#  Variables for cvconfig.h.cmake
-# ----------------------------------------------------------------------------
-set(PACKAGE "opencv")
-set(PACKAGE_BUGREPORT "opencvlibrary-devel@lists.sourceforge.net")
-set(PACKAGE_NAME "opencv")
-set(PACKAGE_STRING "${PACKAGE} ${OPENCV_VERSION}")
-set(PACKAGE_TARNAME "${PACKAGE}")
-set(PACKAGE_VERSION "${OPENCV_VERSION}")
-
 # platform-specific config file
 configure_file("${OpenCV_SOURCE_DIR}/cmake/templates/cvconfig.h.cmake" "${OPENCV_CONFIG_FILE_INCLUDE_DIR}/cvconfig.h")
 
diff --git a/cmake/OpenCVGenInfoPlist.cmake b/cmake/OpenCVGenInfoPlist.cmake
new file mode 100644
index 0000000000..97c674ceb7
--- /dev/null
+++ b/cmake/OpenCVGenInfoPlist.cmake
@@ -0,0 +1,4 @@
+if(IOS)
+  configure_file("${OpenCV_SOURCE_DIR}/platforms/ios/Info.plist.in"
+                 "${CMAKE_BINARY_DIR}/ios/Info.plist")
+endif()
diff --git a/cmake/OpenCVGenPkgconfig.cmake b/cmake/OpenCVGenPkgconfig.cmake
index 49d6707345..7bfc7bc5af 100644
--- a/cmake/OpenCVGenPkgconfig.cmake
+++ b/cmake/OpenCVGenPkgconfig.cmake
@@ -12,7 +12,6 @@ set(prefix      "${CMAKE_INSTALL_PREFIX}")
 set(exec_prefix "\${prefix}")
 set(libdir      "") #TODO: need link paths for OpenCV_EXTRA_COMPONENTS
 set(includedir  "\${prefix}/${OPENCV_INCLUDE_INSTALL_PATH}")
-set(VERSION     ${OPENCV_VERSION})
 
 if(CMAKE_BUILD_TYPE MATCHES "Release")
   set(ocv_optkind OPT)
diff --git a/cmake/OpenCVLegacyOptions.cmake b/cmake/OpenCVLegacyOptions.cmake
deleted file mode 100644
index a34c9e5abc..0000000000
--- a/cmake/OpenCVLegacyOptions.cmake
+++ /dev/null
@@ -1,24 +0,0 @@
-macro(ocv_legacy_option old superseded_by)
-  if(DEFINED ${old})
-    if(ARGV2)
-      set(${superseded_by} ${${old}} CACHE ${ARGV2} "Set via depricated ${old}" FORCE)
-    else()
-      set(${superseded_by} ${${old}} CACHE BOOL     "Set via depricated ${old}" FORCE)
-    endif()
-    unset(${old} CACHE)
-  endif()
-endmacro()
-
-ocv_legacy_option(BUILD_NEW_PYTHON_SUPPORT BUILD_opencv_python)
-ocv_legacy_option(BUILD_JAVA_SUPPORT       BUILD_opencv_java)
-ocv_legacy_option(WITH_ANDROID_CAMERA      BUILD_opencv_androidcamera)
-
-if(DEFINED OPENCV_BUILD_3RDPARTY_LIBS)
-  set(BUILD_ZLIB   ${OPENCV_BUILD_3RDPARTY_LIBS} CACHE BOOL "Set via depricated OPENCV_BUILD_3RDPARTY_LIBS" FORCE)
-  set(BUILD_TIFF   ${OPENCV_BUILD_3RDPARTY_LIBS} CACHE BOOL "Set via depricated OPENCV_BUILD_3RDPARTY_LIBS" FORCE)
-  set(BUILD_JASPER ${OPENCV_BUILD_3RDPARTY_LIBS} CACHE BOOL "Set via depricated OPENCV_BUILD_3RDPARTY_LIBS" FORCE)
-  set(BUILD_JPEG   ${OPENCV_BUILD_3RDPARTY_LIBS} CACHE BOOL "Set via depricated OPENCV_BUILD_3RDPARTY_LIBS" FORCE)
-  set(BUILD_PNG    ${OPENCV_BUILD_3RDPARTY_LIBS} CACHE BOOL "Set via depricated OPENCV_BUILD_3RDPARTY_LIBS" FORCE)
-  unset(OPENCV_BUILD_3RDPARTY_LIBS CACHE)
-endif()
-
diff --git a/cmake/OpenCVMinDepVersions.cmake b/cmake/OpenCVMinDepVersions.cmake
new file mode 100644
index 0000000000..b659a83794
--- /dev/null
+++ b/cmake/OpenCVMinDepVersions.cmake
@@ -0,0 +1,3 @@
+set(MIN_VER_CMAKE 2.8.7)
+set(MIN_VER_PYTHON 2.6)
+set(MIN_VER_ZLIB 1.2.3)
diff --git a/cmake/OpenCVModule.cmake b/cmake/OpenCVModule.cmake
index 48aa7139a2..4956edf911 100644
--- a/cmake/OpenCVModule.cmake
+++ b/cmake/OpenCVModule.cmake
@@ -33,6 +33,7 @@
 #   <add extra installation rules>
 #   ocv_add_accuracy_tests(<extra dependencies>)
 #   ocv_add_perf_tests(<extra dependencies>)
+#   ocv_add_samples(<extra dependencies>)
 #
 #
 # If module have no "extra" then you can define it in one line:
@@ -303,7 +304,7 @@ macro(ocv_glob_modules)
   # collect modules
   set(OPENCV_INITIAL_PASS ON)
   foreach(__path ${ARGN})
-    ocv_get_real_path(__path "${__path}")
+    get_filename_component(__path "${__path}" ABSOLUTE)
 
     list(FIND __directories_observed "${__path}" __pathIdx)
     if(__pathIdx GREATER -1)
@@ -315,7 +316,7 @@ macro(ocv_glob_modules)
     if(__ocvmodules)
       list(SORT __ocvmodules)
       foreach(mod ${__ocvmodules})
-        ocv_get_real_path(__modpath "${__path}/${mod}")
+        get_filename_component(__modpath "${__path}/${mod}" ABSOLUTE)
         if(EXISTS "${__modpath}/CMakeLists.txt")
 
           list(FIND __directories_observed "${__modpath}" __pathIdx)
@@ -432,20 +433,32 @@ macro(ocv_glob_module_sources)
   file(GLOB lib_hdrs     "include/opencv2/*.hpp" "include/opencv2/${name}/*.hpp" "include/opencv2/${name}/*.h")
   file(GLOB lib_hdrs_detail "include/opencv2/${name}/detail/*.hpp" "include/opencv2/${name}/detail/*.h")
 
-  file(GLOB lib_device_srcs "src/cuda/*.cu")
-  set(device_objs "")
-  set(lib_device_hdrs "")
+  file(GLOB lib_cuda_srcs "src/cuda/*.cu")
+  set(cuda_objs "")
+  set(lib_cuda_hdrs "")
 
-  if (HAVE_CUDA AND lib_device_srcs)
+  if(HAVE_CUDA AND lib_cuda_srcs)
     ocv_include_directories(${CUDA_INCLUDE_DIRS})
-    file(GLOB lib_device_hdrs "src/cuda/*.hpp")
+    file(GLOB lib_cuda_hdrs "src/cuda/*.hpp")
 
-    ocv_cuda_compile(device_objs ${lib_device_srcs} ${lib_device_hdrs})
-    source_group("Src\\Cuda"      FILES ${lib_device_srcs} ${lib_device_hdrs})
+    ocv_cuda_compile(cuda_objs ${lib_cuda_srcs} ${lib_cuda_hdrs})
+    source_group("Src\\Cuda"      FILES ${lib_cuda_srcs} ${lib_cuda_hdrs})
+  endif()
+
+  file(GLOB cl_kernels "src/opencl/*.cl")
+
+  if(HAVE_OPENCL AND cl_kernels)
+    ocv_include_directories(${OPENCL_INCLUDE_DIRS})
+    add_custom_command(
+      OUTPUT "${CMAKE_CURRENT_BINARY_DIR}/kernels.cpp"
+      COMMAND ${CMAKE_COMMAND} -DCL_DIR="${CMAKE_CURRENT_SOURCE_DIR}/src/opencl" -DOUTPUT="${CMAKE_CURRENT_BINARY_DIR}/kernels.cpp" -P "${OpenCV_SOURCE_DIR}/cmake/cl2cpp.cmake"
+      DEPENDS ${cl_kernels} "${OpenCV_SOURCE_DIR}/cmake/cl2cpp.cmake")
+    source_group("Src\\OpenCL" FILES ${cl_kernels} "${CMAKE_CURRENT_BINARY_DIR}/kernels.cpp")
+    list(APPEND lib_srcs ${cl_kernels} "${CMAKE_CURRENT_BINARY_DIR}/kernels.cpp")
   endif()
 
   ocv_set_module_sources(${ARGN} HEADERS ${lib_hdrs} ${lib_hdrs_detail}
-                                 SOURCES ${lib_srcs} ${lib_int_hdrs} ${device_objs} ${lib_device_srcs} ${lib_device_hdrs})
+                                 SOURCES ${lib_srcs} ${lib_int_hdrs} ${cuda_objs} ${lib_cuda_srcs} ${lib_cuda_hdrs})
 
   source_group("Src" FILES ${lib_srcs} ${lib_int_hdrs})
   source_group("Include" FILES ${lib_hdrs})
@@ -458,13 +471,28 @@ endmacro()
 #   ocv_create_module(<extra link dependencies>)
 #   ocv_create_module(SKIP_LINK)
 macro(ocv_create_module)
-  add_library(${the_module} ${OPENCV_MODULE_TYPE} ${OPENCV_MODULE_${the_module}_HEADERS} ${OPENCV_MODULE_${the_module}_SOURCES})
+  # The condition we ought to be testing here is whether ocv_add_precompiled_headers will
+  # be called at some point in the future. We can't look into the future, though,
+  # so this will have to do.
+  if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/src/precomp.hpp")
+    get_native_precompiled_header(${the_module} precomp.hpp)
+  endif()
+
+  add_library(${the_module} ${OPENCV_MODULE_TYPE} ${OPENCV_MODULE_${the_module}_HEADERS} ${OPENCV_MODULE_${the_module}_SOURCES}
+    "${OPENCV_CONFIG_FILE_INCLUDE_DIR}/cvconfig.h" "${OPENCV_CONFIG_FILE_INCLUDE_DIR}/opencv2/opencv_modules.hpp"
+    ${${the_module}_pch})
+  if(NOT the_module STREQUAL opencv_ts)
+    set_target_properties(${the_module} PROPERTIES COMPILE_DEFINITIONS OPENCV_NOSTL)
+  endif()
 
   if(NOT "${ARGN}" STREQUAL "SKIP_LINK")
     target_link_libraries(${the_module} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_MODULE_${the_module}_DEPS_EXT} ${OPENCV_LINKER_LIBS} ${IPP_LIBS} ${ARGN})
     if (HAVE_CUDA)
       target_link_libraries(${the_module} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
     endif()
+    if(HAVE_OPENCL AND OPENCL_LIBRARIES)
+      target_link_libraries(${the_module} ${OPENCL_LIBRARIES})
+    endif()
   endif()
 
   add_dependencies(opencv_modules ${the_module})
@@ -492,7 +520,8 @@ macro(ocv_create_module)
     )
   endif()
 
-  if(BUILD_SHARED_LIBS)
+  if((NOT DEFINED OPENCV_MODULE_TYPE AND BUILD_SHARED_LIBS)
+      OR (DEFINED OPENCV_MODULE_TYPE AND OPENCV_MODULE_TYPE STREQUAL SHARED))
     if(MSVC)
       set_target_properties(${the_module} PROPERTIES DEFINE_SYMBOL CVAPI_EXPORTS)
     else()
@@ -517,7 +546,7 @@ macro(ocv_create_module)
   if(OPENCV_MODULE_${the_module}_HEADERS AND ";${OPENCV_MODULES_PUBLIC};" MATCHES ";${the_module};")
     foreach(hdr ${OPENCV_MODULE_${the_module}_HEADERS})
       string(REGEX REPLACE "^.*opencv2/" "opencv2/" hdr2 "${hdr}")
-      if(hdr2 MATCHES "^(opencv2/.*)/[^/]+.h(..)?$")
+      if(NOT hdr2 MATCHES "opencv2/${the_module}/private.*" AND hdr2 MATCHES "^(opencv2/?.*)/[^/]+.h(..)?$" )
         install(FILES ${hdr} DESTINATION "${OPENCV_INCLUDE_INSTALL_PATH}/${CMAKE_MATCH_1}" COMPONENT main)
       endif()
     endforeach()
@@ -553,6 +582,7 @@ macro(ocv_define_module module_name)
 
   ocv_add_accuracy_tests()
   ocv_add_perf_tests()
+  ocv_add_samples()
 endmacro()
 
 # ensures that all passed modules are available
@@ -620,7 +650,9 @@ function(ocv_add_perf_tests)
         set(OPENCV_PERF_${the_module}_SOURCES ${perf_srcs} ${perf_hdrs})
       endif()
 
-      add_executable(${the_target} ${OPENCV_PERF_${the_module}_SOURCES})
+      get_native_precompiled_header(${the_target} perf_precomp.hpp)
+
+      add_executable(${the_target} ${OPENCV_PERF_${the_module}_SOURCES} ${${the_target}_pch})
       target_link_libraries(${the_target} ${OPENCV_MODULE_${the_module}_DEPS} ${perf_deps} ${OPENCV_LINKER_LIBS})
       add_dependencies(opencv_perf_tests ${the_target})
 
@@ -636,9 +668,6 @@ function(ocv_add_perf_tests)
 
       ocv_add_precompiled_headers(${the_target})
 
-      if (PYTHON_EXECUTABLE)
-        add_dependencies(perf ${the_target})
-      endif()
     else(OCV_DEPENDENCIES_FOUND)
       # TODO: warn about unsatisfied dependencies
     endif(OCV_DEPENDENCIES_FOUND)
@@ -671,7 +700,9 @@ function(ocv_add_accuracy_tests)
         set(OPENCV_TEST_${the_module}_SOURCES ${test_srcs} ${test_hdrs})
       endif()
 
-      add_executable(${the_target} ${OPENCV_TEST_${the_module}_SOURCES})
+      get_native_precompiled_header(${the_target} test_precomp.hpp)
+
+      add_executable(${the_target} ${OPENCV_TEST_${the_module}_SOURCES} ${${the_target}_pch})
       target_link_libraries(${the_target} ${OPENCV_MODULE_${the_module}_DEPS} ${test_deps} ${OPENCV_LINKER_LIBS})
       add_dependencies(opencv_tests ${the_target})
 
@@ -696,6 +727,48 @@ function(ocv_add_accuracy_tests)
   endif()
 endfunction()
 
+function(ocv_add_samples)
+  set(samples_path "${CMAKE_CURRENT_SOURCE_DIR}/samples")
+  string(REGEX REPLACE "^opencv_" "" module_id ${the_module})
+
+  if(BUILD_EXAMPLES AND EXISTS "${samples_path}")
+    set(samples_deps ${the_module} ${OPENCV_MODULE_${the_module}_DEPS} opencv_highgui ${ARGN})
+    ocv_check_dependencies(${samples_deps})
+
+    if(OCV_DEPENDENCIES_FOUND)
+      file(GLOB sample_sources "${samples_path}/*.cpp")
+      ocv_include_modules(${OPENCV_MODULE_${the_module}_DEPS})
+
+      foreach(source ${sample_sources})
+        get_filename_component(name "${source}" NAME_WE)
+        set(the_target "example_${module_id}_${name}")
+
+        add_executable(${the_target} "${source}")
+        target_link_libraries(${the_target} ${samples_deps})
+
+        set_target_properties(${the_target} PROPERTIES PROJECT_LABEL "(sample) ${name}")
+
+        if(ENABLE_SOLUTION_FOLDERS)
+          set_target_properties(${the_target} PROPERTIES
+            OUTPUT_NAME "${module_id}-example-${name}"
+            FOLDER "samples/${module_id}")
+        endif()
+
+        if(WIN32)
+          install(TARGETS ${the_target} RUNTIME DESTINATION "samples/${module_id}" COMPONENT main)
+        endif()
+      endforeach()
+    endif()
+  endif()
+
+  if(INSTALL_C_EXAMPLES AND NOT WIN32 AND EXISTS "${samples_path}")
+    file(GLOB sample_files "${samples_path}/*")
+    install(FILES ${sample_files}
+            DESTINATION share/OpenCV/samples/${module_id}
+            PERMISSIONS OWNER_READ GROUP_READ WORLD_READ)
+  endif()
+endfunction()
+
 # internal macro; finds all link dependencies of the module
 # should be used at the end of CMake processing
 macro(__ocv_track_module_link_dependencies the_module optkind)
diff --git a/cmake/OpenCVPCHSupport.cmake b/cmake/OpenCVPCHSupport.cmake
index cfc4bfa724..060965346a 100644
--- a/cmake/OpenCVPCHSupport.cmake
+++ b/cmake/OpenCVPCHSupport.cmake
@@ -46,6 +46,13 @@ MACRO(_PCH_GET_COMPILE_FLAGS _out_compile_flags)
             LIST(APPEND ${_out_compile_flags} "-fPIC")
         ENDIF()
 
+        GET_TARGET_PROPERTY(_target_definitions ${_PCH_current_target} COMPILE_DEFINITIONS)
+        if(_target_definitions)
+          foreach(_def ${_target_definitions})
+            LIST(APPEND ${_out_compile_flags} "-D${_def}")
+          endforeach()
+        endif()
+
     ELSE()
         ## TODO ... ? or does it work out of the box
     ENDIF()
@@ -272,12 +279,9 @@ ENDMACRO(ADD_PRECOMPILED_HEADER)
 MACRO(GET_NATIVE_PRECOMPILED_HEADER _targetName _input)
 
     if(CMAKE_GENERATOR MATCHES "^Visual.*$")
-        SET(_dummy_str "#include \"${_input}\"\n"
-"// This is required to suppress LNK4221.  Very annoying.\n"
-"void *g_${_targetName}Dummy = 0\;\n")
+        set(_dummy_str "#include \"${_input}\"\n")
 
-        # Use of cxx extension for generated files (as Qt does)
-        SET(${_targetName}_pch ${CMAKE_CURRENT_BINARY_DIR}/${_targetName}_pch.cxx)
+        set(${_targetName}_pch ${CMAKE_CURRENT_BINARY_DIR}/${_targetName}_pch.cpp)
         if(EXISTS ${${_targetName}_pch})
             # Check if contents is the same, if not rewrite
             # todo
@@ -337,11 +341,7 @@ ENDMACRO(ADD_NATIVE_PRECOMPILED_HEADER)
 
 macro(ocv_add_precompiled_header_to_target the_target pch_header)
   if(PCHSupport_FOUND AND ENABLE_PRECOMPILED_HEADERS AND EXISTS "${pch_header}")
-    if(CMAKE_GENERATOR MATCHES Visual)
-      string(REGEX REPLACE "hpp$" "cpp" ${the_target}_pch "${pch_header}")
-      add_native_precompiled_header(${the_target} ${pch_header})
-      unset(${the_target}_pch)
-    elseif(CMAKE_GENERATOR MATCHES Xcode)
+    if(CMAKE_GENERATOR MATCHES "^Visual" OR CMAKE_GENERATOR MATCHES Xcode)
       add_native_precompiled_header(${the_target} ${pch_header})
     elseif(CMAKE_COMPILER_IS_GNUCXX AND CMAKE_GENERATOR MATCHES "Makefiles|Ninja")
       add_precompiled_header(${the_target} ${pch_header})
diff --git a/cmake/OpenCVUtils.cmake b/cmake/OpenCVUtils.cmake
index db24c99708..ddf0290673 100644
--- a/cmake/OpenCVUtils.cmake
+++ b/cmake/OpenCVUtils.cmake
@@ -77,7 +77,7 @@ MACRO(ocv_check_compiler_flag LANG FLAG RESULT)
     if(_fname)
       MESSAGE(STATUS "Performing Test ${RESULT}")
       TRY_COMPILE(${RESULT}
-        ${CMAKE_BINARY_DIR}
+        "${CMAKE_BINARY_DIR}"
         "${_fname}"
         COMPILE_DEFINITIONS "${FLAG}"
         OUTPUT_VARIABLE OUTPUT)
@@ -411,16 +411,6 @@ macro(ocv_regex_escape var regex)
 endmacro()
 
 
-# get absolute path with symlinks resolved
-macro(ocv_get_real_path VAR PATHSTR)
-  if(CMAKE_VERSION VERSION_LESS 2.8)
-    get_filename_component(${VAR} "${PATHSTR}" ABSOLUTE)
-  else()
-    get_filename_component(${VAR} "${PATHSTR}" REALPATH)
-  endif()
-endmacro()
-
-
 # convert list of paths to full paths
 macro(ocv_convert_to_full_paths VAR)
   if(${VAR})
@@ -511,6 +501,13 @@ macro(ocv_parse_header2 LIBNAME HDR_PATH VARNAME)
   endif()
 endmacro()
 
+# read single version info from the pkg file
+macro(ocv_parse_pkg LIBNAME PKG_PATH SCOPE)
+  if(EXISTS "${PKG_PATH}/${LIBNAME}.pc")
+    file(STRINGS "${PKG_PATH}/${LIBNAME}.pc" line_to_parse REGEX "^Version:[ \t]+[0-9.]*.*$" LIMIT_COUNT 1)
+    STRING(REGEX REPLACE ".*Version: ([^ ]+).*" "\\1" ALIASOF_${LIBNAME}_VERSION "${line_to_parse}" )
+  endif()
+endmacro()
 
 ################################################################################################
 # short command to setup source group
@@ -518,4 +515,4 @@ function(ocv_source_group group)
   cmake_parse_arguments(OCV_SOURCE_GROUP "" "" "GLOB" ${ARGN})
   file(GLOB srcs ${OCV_SOURCE_GROUP_GLOB})
   source_group(${group} FILES ${srcs})
-endfunction()
\ No newline at end of file
+endfunction()
diff --git a/cmake/OpenCVVersion.cmake b/cmake/OpenCVVersion.cmake
index 03de98068b..c8ce40308f 100644
--- a/cmake/OpenCVVersion.cmake
+++ b/cmake/OpenCVVersion.cmake
@@ -1,16 +1,19 @@
 SET(OPENCV_VERSION_FILE "${CMAKE_CURRENT_SOURCE_DIR}/modules/core/include/opencv2/core/version.hpp")
-FILE(STRINGS "${OPENCV_VERSION_FILE}" OPENCV_VERSION_PARTS REGEX "#define CV_VERSION_[A-Z]+[ ]+[0-9]+" )
+file(STRINGS "${OPENCV_VERSION_FILE}" OPENCV_VERSION_PARTS REGEX "#define CV_VERSION_[A-Z]+[ ]+" )
 
 string(REGEX REPLACE ".+CV_VERSION_EPOCH[ ]+([0-9]+).*" "\\1" OPENCV_VERSION_MAJOR "${OPENCV_VERSION_PARTS}")
 string(REGEX REPLACE ".+CV_VERSION_MAJOR[ ]+([0-9]+).*" "\\1" OPENCV_VERSION_MINOR "${OPENCV_VERSION_PARTS}")
 string(REGEX REPLACE ".+CV_VERSION_MINOR[ ]+([0-9]+).*" "\\1" OPENCV_VERSION_PATCH "${OPENCV_VERSION_PARTS}")
 string(REGEX REPLACE ".+CV_VERSION_REVISION[ ]+([0-9]+).*" "\\1" OPENCV_VERSION_TWEAK "${OPENCV_VERSION_PARTS}")
+string(REGEX REPLACE ".+CV_VERSION_STATUS[ ]+\"([^\"]*)\".*" "\\1" OPENCV_VERSION_STATUS "${OPENCV_VERSION_PARTS}")
 
-set(OPENCV_VERSION "${OPENCV_VERSION_MAJOR}.${OPENCV_VERSION_MINOR}.${OPENCV_VERSION_PATCH}")
+set(OPENCV_VERSION_PLAIN "${OPENCV_VERSION_MAJOR}.${OPENCV_VERSION_MINOR}.${OPENCV_VERSION_PATCH}")
 if(OPENCV_VERSION_TWEAK GREATER 0)
-  set(OPENCV_VERSION "${OPENCV_VERSION}.${OPENCV_VERSION_TWEAK}")
+  set(OPENCV_VERSION_PLAIN "${OPENCV_VERSION_PLAIN}.${OPENCV_VERSION_TWEAK}")
 endif()
 
+set(OPENCV_VERSION "${OPENCV_VERSION_PLAIN}${OPENCV_VERSION_STATUS}")
+
 set(OPENCV_SOVERSION "${OPENCV_VERSION_MAJOR}.${OPENCV_VERSION_MINOR}")
 set(OPENCV_LIBVERSION "${OPENCV_VERSION_MAJOR}.${OPENCV_VERSION_MINOR}.${OPENCV_VERSION_PATCH}")
 
diff --git a/cmake/OpenCVDetectCudaArch.cu b/cmake/checks/OpenCVDetectCudaArch.cu
similarity index 99%
rename from cmake/OpenCVDetectCudaArch.cu
rename to cmake/checks/OpenCVDetectCudaArch.cu
index 008f8ba8d0..9d7086cf24 100644
--- a/cmake/OpenCVDetectCudaArch.cu
+++ b/cmake/checks/OpenCVDetectCudaArch.cu
@@ -11,4 +11,4 @@ int main()
         printf("%d.%d ", prop.major, prop.minor);
     }
     return 0;
-}
\ No newline at end of file
+}
diff --git a/cmake/checks/vfwtest.cpp b/cmake/checks/vfwtest.cpp
new file mode 100644
index 0000000000..8d8ecb2719
--- /dev/null
+++ b/cmake/checks/vfwtest.cpp
@@ -0,0 +1,10 @@
+
+#include <windows.h>
+#include <vfw.h>
+
+int main()
+{
+  AVIFileInit();
+  AVIFileExit();
+  return 0;
+}
diff --git a/cmake/checks/win32uitest.cpp b/cmake/checks/win32uitest.cpp
new file mode 100644
index 0000000000..f475e1c963
--- /dev/null
+++ b/cmake/checks/win32uitest.cpp
@@ -0,0 +1,11 @@
+#include <windows.h>
+
+int main(int argc, char** argv)
+{
+    CreateWindow(NULL /*lpClassName*/, NULL /*lpWindowName*/, 0 /*dwStyle*/, 0 /*x*/,
+                 0 /*y*/, 0 /*nWidth*/, 0 /*nHeight*/, NULL /*hWndParent*/, NULL /*hMenu*/,
+                NULL /*hInstance*/,  NULL /*lpParam*/);
+    DeleteDC(NULL);
+
+    return 0;
+}
diff --git a/cmake/checks/winrttest.cpp b/cmake/checks/winrttest.cpp
new file mode 100644
index 0000000000..9ec0c9ac1f
--- /dev/null
+++ b/cmake/checks/winrttest.cpp
@@ -0,0 +1,6 @@
+#include <wrl/client.h>
+
+int main(int, char**)
+{
+    return 0;
+}
diff --git a/modules/ocl/cl2cpp.cmake b/cmake/cl2cpp.cmake
similarity index 89%
rename from modules/ocl/cl2cpp.cmake
rename to cmake/cl2cpp.cmake
index ca17c61b43..1e932eabdb 100644
--- a/modules/ocl/cl2cpp.cmake
+++ b/cmake/cl2cpp.cmake
@@ -18,6 +18,7 @@ foreach(cl ${cl_list})
   string(REPLACE "\t" "  " lines "${lines}")
 
   string(REGEX REPLACE "/\\*([^*]/|\\*[^/]|[^*/])*\\*/" ""   lines "${lines}") # multiline comments
+  string(REGEX REPLACE "/\\*([^\n])*\\*/"               ""   lines "${lines}") # single-line comments
   string(REGEX REPLACE "[ ]*//[^\n]*\n"                 "\n" lines "${lines}") # single-line comments
   string(REGEX REPLACE "\n[ ]*(\n[ ]*)*"                "\n" lines "${lines}") # empty lines & leading whitespace
   string(REGEX REPLACE "^\n"                            ""   lines "${lines}") # leading new line
@@ -31,4 +32,4 @@ foreach(cl ${cl_list})
   file(APPEND ${OUTPUT} "const char* ${cl_filename}=\"${lines};\n")
 endforeach()
 
-file(APPEND ${OUTPUT} "}\n}\n")
\ No newline at end of file
+file(APPEND ${OUTPUT} "}\n}\n")
diff --git a/cmake/templates/OpenCVConfig-version.cmake.in b/cmake/templates/OpenCVConfig-version.cmake.in
index 0b432a074c..b5ac5f8e29 100644
--- a/cmake/templates/OpenCVConfig-version.cmake.in
+++ b/cmake/templates/OpenCVConfig-version.cmake.in
@@ -1,4 +1,4 @@
-set(OpenCV_VERSION @OPENCV_VERSION@)
+set(OpenCV_VERSION @OPENCV_VERSION_PLAIN@)
 set(PACKAGE_VERSION ${OpenCV_VERSION})
 
 set(PACKAGE_VERSION_EXACT False)
diff --git a/cmake/templates/OpenCVConfig.cmake.in b/cmake/templates/OpenCVConfig.cmake.in
index 1dcd840307..90749488a7 100644
--- a/cmake/templates/OpenCVConfig.cmake.in
+++ b/cmake/templates/OpenCVConfig.cmake.in
@@ -22,11 +22,12 @@
 #      - OpenCV_INCLUDE_DIRS             : The OpenCV include directories.
 #      - OpenCV_COMPUTE_CAPABILITIES     : The version of compute capability
 #      - OpenCV_ANDROID_NATIVE_API_LEVEL : Minimum required level of Android API
-#      - OpenCV_VERSION                  : The version of this OpenCV build: "@OPENCV_VERSION@"
+#      - OpenCV_VERSION                  : The version of this OpenCV build: "@OPENCV_VERSION_PLAIN@"
 #      - OpenCV_VERSION_MAJOR            : Major version part of OpenCV_VERSION: "@OPENCV_VERSION_MAJOR@"
 #      - OpenCV_VERSION_MINOR            : Minor version part of OpenCV_VERSION: "@OPENCV_VERSION_MINOR@"
 #      - OpenCV_VERSION_PATCH            : Patch version part of OpenCV_VERSION: "@OPENCV_VERSION_PATCH@"
 #      - OpenCV_VERSION_TWEAK            : Tweak version part of OpenCV_VERSION: "@OPENCV_VERSION_TWEAK@"
+#      - OpenCV_VERSION_STATUS           : Development status of this build: "@OPENCV_VERSION_STATUS@"
 #
 #    Advanced variables:
 #      - OpenCV_SHARED
@@ -96,11 +97,12 @@ mark_as_advanced(FORCE OpenCV_LIB_DIR_OPT OpenCV_LIB_DIR_DBG OpenCV_3RDPARTY_LIB
 # ======================================================
 #  Version variables:
 # ======================================================
-SET(OpenCV_VERSION @OPENCV_VERSION@)
+SET(OpenCV_VERSION @OPENCV_VERSION_PLAIN@)
 SET(OpenCV_VERSION_MAJOR  @OPENCV_VERSION_MAJOR@)
 SET(OpenCV_VERSION_MINOR  @OPENCV_VERSION_MINOR@)
 SET(OpenCV_VERSION_PATCH  @OPENCV_VERSION_PATCH@)
 SET(OpenCV_VERSION_TWEAK  @OPENCV_VERSION_TWEAK@)
+SET(OpenCV_VERSION_STATUS "@OPENCV_VERSION_STATUS@")
 
 # ====================================================================
 # Link libraries: e.g.   libopencv_core.so, opencv_imgproc220d.lib, etc...
diff --git a/cmake/templates/cmake_uninstall.cmake.in b/cmake/templates/cmake_uninstall.cmake.in
index 14e601019d..0e63d705cd 100644
--- a/cmake/templates/cmake_uninstall.cmake.in
+++ b/cmake/templates/cmake_uninstall.cmake.in
@@ -23,5 +23,3 @@ FOREACH(file ${files})
     MESSAGE(STATUS "File \"$ENV{DESTDIR}${file}\" does not exist.")
   ENDIF(EXISTS "$ENV{DESTDIR}${file}")
 ENDFOREACH(file)
-
-
diff --git a/cmake/templates/cvconfig.h.cmake b/cmake/templates/cvconfig.h.cmake
index 62c8b68b38..56c5d5aadc 100644
--- a/cmake/templates/cvconfig.h.cmake
+++ b/cmake/templates/cvconfig.h.cmake
@@ -1,17 +1,20 @@
-/* Define to one of `_getb67', `GETB67', `getb67' for Cray-2 and Cray-YMP
-   systems. This function is required for `alloca.c' support on those systems.
-   */
-#cmakedefine  CRAY_STACKSEG_END
+/* OpenCV compiled as static or dynamic libs */
+#cmakedefine BUILD_SHARED_LIBS
 
-/* Define to 1 if using `alloca.c'. */
-#cmakedefine C_ALLOCA
+/* Compile for 'real' NVIDIA GPU architectures */
+#define CUDA_ARCH_BIN "${OPENCV_CUDA_ARCH_BIN}"
 
-/* Define to 1 if you have `alloca', as a function or macro. */
-#cmakedefine HAVE_ALLOCA 1
+/* Create PTX or BIN for 1.0 compute capability */
+#cmakedefine CUDA_ARCH_BIN_OR_PTX_10
 
-/* Define to 1 if you have <alloca.h> and it should be used (not on Ultrix).
-   */
-#cmakedefine HAVE_ALLOCA_H 1
+/* NVIDIA GPU features are used */
+#define CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES}"
+
+/* Compile for 'virtual' NVIDIA PTX architectures */
+#define CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX}"
+
+/* AVFoundation video libraries */
+#cmakedefine HAVE_AVFOUNDATION
 
 /* V4L capturing support */
 #cmakedefine HAVE_CAMV4L
@@ -19,15 +22,33 @@
 /* V4L2 capturing support */
 #cmakedefine HAVE_CAMV4L2
 
-/* V4L2 capturing support in videoio.h */
-#cmakedefine HAVE_VIDEOIO
-
-/* V4L/V4L2 capturing support via libv4l */
-#cmakedefine HAVE_LIBV4L
-
 /* Carbon windowing environment */
 #cmakedefine HAVE_CARBON
 
+/* AMD's Basic Linear Algebra Subprograms Library*/
+#cmakedefine HAVE_CLAMDBLAS
+
+/* AMD's OpenCL Fast Fourier Transform Library*/
+#cmakedefine HAVE_CLAMDFFT
+
+/* Clp support */
+#cmakedefine HAVE_CLP
+
+/* Cocoa API */
+#cmakedefine HAVE_COCOA
+
+/* C= */
+#cmakedefine HAVE_CSTRIPES
+
+/* NVidia Cuda Basic Linear Algebra Subprograms (BLAS) API*/
+#cmakedefine HAVE_CUBLAS
+
+/* NVidia Cuda Runtime API*/
+#cmakedefine HAVE_CUDA
+
+/* NVidia Cuda Fast Fourier Transform (FFT) API*/
+#cmakedefine HAVE_CUFFT
+
 /* IEEE1394 capturing support */
 #cmakedefine HAVE_DC1394
 
@@ -37,182 +58,108 @@
 /* IEEE1394 capturing support - libdc1394 v2.x */
 #cmakedefine HAVE_DC1394_2
 
+/* DirectShow Video Capture library */
+#cmakedefine HAVE_DSHOW
+
+/* Eigen Matrix & Linear Algebra Library */
+#cmakedefine HAVE_EIGEN
+
+/* FFMpeg video library */
+#cmakedefine HAVE_FFMPEG
+
+/* ffmpeg's libswscale */
+#cmakedefine HAVE_FFMPEG_SWSCALE
+
 /* ffmpeg in Gentoo */
 #cmakedefine HAVE_GENTOO_FFMPEG
 
-/* FFMpeg video library */
-#cmakedefine  HAVE_FFMPEG
-
-/* FFMpeg version flag */
-#cmakedefine  NEW_FFMPEG
-
-/* ffmpeg's libswscale */
-#cmakedefine  HAVE_FFMPEG_SWSCALE
-
 /* GStreamer multimedia framework */
-#cmakedefine  HAVE_GSTREAMER
+#cmakedefine HAVE_GSTREAMER
 
 /* GTK+ 2.0 Thread support */
-#cmakedefine  HAVE_GTHREAD
+#cmakedefine HAVE_GTHREAD
 
 /* GTK+ 2.x toolkit */
-#cmakedefine  HAVE_GTK
-
-/* OpenEXR codec */
-#cmakedefine  HAVE_ILMIMF
+#cmakedefine HAVE_GTK
 
 /* Define to 1 if you have the <inttypes.h> header file. */
-#cmakedefine  HAVE_INTTYPES_H 1
-
-/* JPEG-2000 codec */
-#cmakedefine  HAVE_JASPER
-
-/* IJG JPEG codec */
-#cmakedefine  HAVE_JPEG
-
-/* Define to 1 if you have the `dl' library (-ldl). */
-#cmakedefine  HAVE_LIBDL 1
-
-/* Define to 1 if you have the `gomp' library (-lgomp). */
-#cmakedefine  HAVE_LIBGOMP 1
-
-/* Define to 1 if you have the `m' library (-lm). */
-#cmakedefine  HAVE_LIBM 1
-
-/* libpng/png.h needs to be included */
-#cmakedefine  HAVE_LIBPNG_PNG_H
-
-/* Define to 1 if you have the `pthread' library (-lpthread). */
-#cmakedefine  HAVE_LIBPTHREAD 1
-
-/* Define to 1 if you have the `lrint' function. */
-#cmakedefine  HAVE_LRINT 1
-
-/* PNG codec */
-#cmakedefine  HAVE_PNG
-
-/* Define to 1 if you have the `png_get_valid' function. */
-#cmakedefine  HAVE_PNG_GET_VALID 1
-
-/* png.h needs to be included */
-#cmakedefine  HAVE_PNG_H
-
-/* Define to 1 if you have the `png_set_tRNS_to_alpha' function. */
-#cmakedefine  HAVE_PNG_SET_TRNS_TO_ALPHA 1
-
-/* QuickTime video libraries */
-#cmakedefine  HAVE_QUICKTIME
-
-/* AVFoundation video libraries */
-#cmakedefine  HAVE_AVFOUNDATION
-
-/* TIFF codec */
-#cmakedefine  HAVE_TIFF
-
-/* Unicap video capture library */
-#cmakedefine  HAVE_UNICAP
-
-/* Define to 1 if you have the <unistd.h> header file. */
-#cmakedefine  HAVE_UNISTD_H 1
-
-/* Xine video library */
-#cmakedefine  HAVE_XINE
-
-/* OpenNI library */
-#cmakedefine  HAVE_OPENNI
-
-/* LZ77 compression/decompression library (used for PNG) */
-#cmakedefine  HAVE_ZLIB
+#cmakedefine HAVE_INTTYPES_H 1
 
 /* Intel Integrated Performance Primitives */
-#cmakedefine  HAVE_IPP
+#cmakedefine HAVE_IPP
 
-/* OpenCV compiled as static or dynamic libs */
-#cmakedefine  BUILD_SHARED_LIBS
+/* JPEG-2000 codec */
+#cmakedefine HAVE_JASPER
 
-/* Name of package */
-#define  PACKAGE "${PACKAGE}"
+/* IJG JPEG codec */
+#cmakedefine HAVE_JPEG
 
-/* Define to the address where bug reports for this package should be sent. */
-#define  PACKAGE_BUGREPORT "${PACKAGE_BUGREPORT}"
+/* libpng/png.h needs to be included */
+#cmakedefine HAVE_LIBPNG_PNG_H
 
-/* Define to the full name of this package. */
-#define  PACKAGE_NAME "${PACKAGE_NAME}"
+/* V4L/V4L2 capturing support via libv4l */
+#cmakedefine HAVE_LIBV4L
 
-/* Define to the full name and version of this package. */
-#define  PACKAGE_STRING "${PACKAGE_STRING}"
-
-/* Define to the one symbol short name of this package. */
-#define  PACKAGE_TARNAME "${PACKAGE_TARNAME}"
-
-/* Define to the version of this package. */
-#define  PACKAGE_VERSION "${PACKAGE_VERSION}"
-
-/* If using the C implementation of alloca, define if you know the
-   direction of stack growth for your system; otherwise it will be
-   automatically deduced at runtime.
-    STACK_DIRECTION > 0 => grows toward higher addresses
-    STACK_DIRECTION < 0 => grows toward lower addresses
-    STACK_DIRECTION = 0 => direction of growth unknown */
-#cmakedefine  STACK_DIRECTION
-
-/* Version number of package */
-#define  VERSION "${PACKAGE_VERSION}"
-
-/* Define to 1 if your processor stores words with the most significant byte
-   first (like Motorola and SPARC, unlike Intel and VAX). */
-#cmakedefine  WORDS_BIGENDIAN
-
-/* Intel Threading Building Blocks */
-#cmakedefine  HAVE_TBB
-
-/* C= */
-#cmakedefine  HAVE_CSTRIPES
-
-/* Eigen Matrix & Linear Algebra Library */
-#cmakedefine  HAVE_EIGEN
-
-/* NVidia Cuda Runtime API*/
-#cmakedefine HAVE_CUDA
-
-/* NVidia Cuda Fast Fourier Transform (FFT) API*/
-#cmakedefine HAVE_CUFFT
-
-/* NVidia Cuda Basic Linear Algebra Subprograms (BLAS) API*/
-#cmakedefine HAVE_CUBLAS
+/* Microsoft Media Foundation Capture library */
+#cmakedefine HAVE_MSMF
 
 /* NVidia Video Decoding API*/
 #cmakedefine HAVE_NVCUVID
 
-/* Compile for 'real' NVIDIA GPU architectures */
-#define CUDA_ARCH_BIN "${OPENCV_CUDA_ARCH_BIN}"
-
-/* Compile for 'virtual' NVIDIA PTX architectures */
-#define CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX}"
-
-/* NVIDIA GPU features are used */
-#define CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES}"
-
-/* Create PTX or BIN for 1.0 compute capability */
-#cmakedefine CUDA_ARCH_BIN_OR_PTX_10
-
 /* OpenCL Support */
 #cmakedefine HAVE_OPENCL
 
-/* AMD's OpenCL Fast Fourier Transform Library*/
-#cmakedefine HAVE_CLAMDFFT
-
-/* AMD's Basic Linear Algebra Subprograms Library*/
-#cmakedefine HAVE_CLAMDBLAS
-
-/* VideoInput library */
-#cmakedefine HAVE_VIDEOINPUT
-
-/* XIMEA camera support */
-#cmakedefine HAVE_XIMEA
+/* OpenEXR codec */
+#cmakedefine HAVE_OPENEXR
 
 /* OpenGL support*/
 #cmakedefine HAVE_OPENGL
 
-/* Clp support */
-#cmakedefine HAVE_CLP
+/* OpenNI library */
+#cmakedefine HAVE_OPENNI
+
+/* PNG codec */
+#cmakedefine HAVE_PNG
+
+/* Qt support */
+#cmakedefine HAVE_QT
+
+/* Qt OpenGL support */
+#cmakedefine HAVE_QT_OPENGL
+
+/* QuickTime video libraries */
+#cmakedefine HAVE_QUICKTIME
+
+/* QTKit video libraries */
+#cmakedefine HAVE_QTKIT
+
+/* Intel Threading Building Blocks */
+#cmakedefine HAVE_TBB
+
+/* TIFF codec */
+#cmakedefine HAVE_TIFF
+
+/* Unicap video capture library */
+#cmakedefine HAVE_UNICAP
+
+/* Video for Windows support */
+#cmakedefine HAVE_VFW
+
+/* V4L2 capturing support in videoio.h */
+#cmakedefine HAVE_VIDEOIO
+
+/* Win32 UI */
+#cmakedefine HAVE_WIN32UI
+
+/* Windows Runtime support */
+#cmakedefine HAVE_WINRT
+
+/* XIMEA camera support */
+#cmakedefine HAVE_XIMEA
+
+/* Xine video library */
+#cmakedefine HAVE_XINE
+
+/* Define to 1 if your processor stores words with the most significant byte
+   first (like Motorola and SPARC, unlike Intel and VAX). */
+#cmakedefine WORDS_BIGENDIAN
diff --git a/cmake/templates/opencv-XXX.pc.cmake.in b/cmake/templates/opencv-XXX.pc.cmake.in
index 4f6415cab4..5b94e4d6c3 100644
--- a/cmake/templates/opencv-XXX.pc.cmake.in
+++ b/cmake/templates/opencv-XXX.pc.cmake.in
@@ -8,6 +8,6 @@ includedir_new=@includedir@
 
 Name: OpenCV
 Description: Open Source Computer Vision Library
-Version: @VERSION@
+Version: @OPENCV_VERSION_PLAIN@
 Libs: @OpenCV_LIB_COMPONENTS@
 Cflags: -I${includedir_old} -I${includedir_new}
diff --git a/cmake/templates/opencv_modules.hpp.in b/cmake/templates/opencv_modules.hpp.in
index 5d58e40e4c..1498715028 100644
--- a/cmake/templates/opencv_modules.hpp.in
+++ b/cmake/templates/opencv_modules.hpp.in
@@ -6,4 +6,4 @@
  *
 */
 
-@OPENCV_MODULE_DEFINITIONS_CONFIGMAKE@
\ No newline at end of file
+@OPENCV_MODULE_DEFINITIONS_CONFIGMAKE@
diff --git a/doc/CMakeLists.txt b/doc/CMakeLists.txt
index 0f2695fc9a..888740bd4f 100644
--- a/doc/CMakeLists.txt
+++ b/doc/CMakeLists.txt
@@ -17,7 +17,7 @@ if(BUILD_DOCS AND HAVE_SPHINX)
   set(OPTIONAL_DOC_LIST "")
 
 
-  set(OPENCV2_BASE_MODULES core imgproc highgui video calib3d features2d objdetect ml flann gpu photo stitching nonfree contrib legacy)
+  set(OPENCV2_BASE_MODULES core imgproc highgui video calib3d features2d objdetect ml flann gpu photo stitching nonfree contrib legacy bioinspired)
 
   # build lists of modules to be documented
   set(OPENCV2_MODULES "")
@@ -49,12 +49,12 @@ if(BUILD_DOCS AND HAVE_SPHINX)
     set(toc_file "${OPENCV_MODULE_opencv_${mod}_LOCATION}/doc/${mod}.rst")
     if(EXISTS "${toc_file}")
       file(RELATIVE_PATH toc_file "${OpenCV_SOURCE_DIR}/modules" "${toc_file}")
-      set(OPENCV_REFMAN_TOC "${OPENCV_REFMAN_TOC}   ${toc_file}\r\n")
+      set(OPENCV_REFMAN_TOC "${OPENCV_REFMAN_TOC}   ${toc_file}\n")
     endif()
   endforeach()
 
-  file(GLOB_RECURSE _OPENCV_FILES_REF "${OpenCV_SOURCE_DIR}/android/service/doc/*.rst")
-  file(GLOB_RECURSE _OPENCV_FILES_REF_PICT "${OpenCV_SOURCE_DIR}/android/service/doc/*.png" "${OpenCV_SOURCE_DIR}/android/service/doc/*.jpg")
+  file(GLOB_RECURSE _OPENCV_FILES_REF "${OpenCV_SOURCE_DIR}/platforms/android/service/doc/*.rst")
+  file(GLOB_RECURSE _OPENCV_FILES_REF_PICT "${OpenCV_SOURCE_DIR}/platforms/android/service/doc/*.png" "${OpenCV_SOURCE_DIR}/platforms/android/service/doc/*.jpg")
   list(APPEND OPENCV_FILES_REF ${_OPENCV_FILES_REF})
   list(APPEND OPENCV_FILES_REF_PICT ${_OPENCV_FILES_REF_PICT})
 
@@ -67,9 +67,14 @@ if(BUILD_DOCS AND HAVE_SPHINX)
   set(OPENCV_DOC_DEPS conf.py ${OPENCV_FILES_REF} ${OPENCV_FILES_REF_PICT}
            ${OPENCV_FILES_UG} ${OPENCV_FILES_TUT} ${OPENCV_FILES_TUT_PICT})
 
+  set(BUILD_PLANTUML "")
+  if(PLANTUML)
+    set(BUILD_PLANTUML "-tplantuml")
+  endif()
+
   if(PDFLATEX_COMPILER)
     add_custom_target(docs
-      COMMAND ${SPHINX_BUILD} -b latex -c ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/.. .
+      COMMAND ${SPHINX_BUILD} ${BUILD_PLANTUML} -b latex -c ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/.. .
       COMMAND ${CMAKE_COMMAND} -E copy_directory ${CMAKE_CURRENT_SOURCE_DIR}/pics ${CMAKE_CURRENT_BINARY_DIR}/doc/opencv1/pics
       COMMAND ${CMAKE_COMMAND} -E copy_if_different ${CMAKE_CURRENT_SOURCE_DIR}/mymath.sty ${CMAKE_CURRENT_BINARY_DIR}
       COMMAND ${PYTHON_EXECUTABLE} "${CMAKE_CURRENT_SOURCE_DIR}/patch_refman_latex.py" opencv2refman.tex
@@ -103,7 +108,7 @@ if(BUILD_DOCS AND HAVE_SPHINX)
   endif()
 
   add_custom_target(html_docs
-    COMMAND ${SPHINX_BUILD} -b html -c ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/.. ./_html
+    COMMAND ${SPHINX_BUILD} ${BUILD_PLANTUML} -b html -c ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_SOURCE_DIR}/.. ./_html
     COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/mymath.sty ${CMAKE_CURRENT_BINARY_DIR}
     DEPENDS ${OPENCV_DOC_DEPS}
     WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
@@ -122,4 +127,4 @@ if(BUILD_DOCS AND HAVE_SPHINX)
     install(FILES "${f}" DESTINATION "${OPENCV_DOC_INSTALL_PATH}" OPTIONAL)
   endforeach()
 
-endif()
\ No newline at end of file
+endif()
diff --git a/doc/_static/insertIframe.js b/doc/_static/insertIframe.js
index 904ff361de..0a3099a8c2 100644
--- a/doc/_static/insertIframe.js
+++ b/doc/_static/insertIframe.js
@@ -1,4 +1,4 @@
-function insertIframe (elementId, iframeSrc) 
+function insertIframe (elementId, iframeSrc)
 {
   var iframe;
   if (document.createElement && (iframe = document.createElement('iframe')))
@@ -10,4 +10,3 @@ function insertIframe (elementId, iframeSrc)
     element.parentNode.replaceChild(iframe, element);
   }
 }
-
diff --git a/doc/_themes/blue/layout.html b/doc/_themes/blue/layout.html
index 8bba49b17a..d0d43031c8 100644
--- a/doc/_themes/blue/layout.html
+++ b/doc/_themes/blue/layout.html
@@ -183,7 +183,7 @@
                   {% if theme_lang == 'c' %}
                   {% endif %}
                   {% if theme_lang == 'cpp' %}
-                    <li>Try the <a href="http://opencv.willowgarage.com/wiki/Welcome?action=AttachFile&do=get&target=opencv_cheatsheet.pdf">Cheatsheet</a>.</li>
+                    <li>Try the <a href="http://docs.opencv.org/opencv_cheatsheet.pdf">Cheatsheet</a>.</li>
                   {% endif %}
                   {% if theme_lang == 'py' %}
                     <li>Try the <a href="cookbook.html">Cookbook</a>.</li>
diff --git a/doc/_themes/blue/static/default.css_t b/doc/_themes/blue/static/default.css_t
index e2127db1da..4d410be8aa 100644
--- a/doc/_themes/blue/static/default.css_t
+++ b/doc/_themes/blue/static/default.css_t
@@ -310,21 +310,6 @@ dl.pyfunction > dt:first-child {
     margin-bottom: 7px;
 }
 
-dl.pyoldfunction > dt:first-child {
-    margin-bottom: 7px;
-    color: #8080B0;
-}
-
-dl.pyoldfunction > dt:first-child tt.descname
-{
-    color: #8080B0;
-}
-
-dl.pyoldfunction > dt:first-child tt.descclassname
-{
-    color: #8080B0;
-}
-
 dl.jfunction > dt:first-child {
     margin-bottom: 7px;
 }
@@ -402,4 +387,4 @@ div.sphinxsidebar #searchbox input[type="text"] {
 
 div.sphinxsidebar #searchbox input[type="submit"] {
     width:auto;
-}
\ No newline at end of file
+}
diff --git a/doc/_themes/blue/theme.conf b/doc/_themes/blue/theme.conf
index 206193f6f2..bc0b99b2a3 100644
--- a/doc/_themes/blue/theme.conf
+++ b/doc/_themes/blue/theme.conf
@@ -28,4 +28,4 @@ feedbacklinkcolor = #ffffff
 bodyfont = sans-serif
 headfont = 'Trebuchet MS', sans-serif
 guifont = "Lucida Sans","Lucida Sans Unicode","Lucida Grande",Verdana,Arial,Helvetica,sans-serif
-lang = none
\ No newline at end of file
+lang = none
diff --git a/doc/check_docs.py b/doc/check_docs.py
index 487b390e8a..0290fc70f4 100755
--- a/doc/check_docs.py
+++ b/doc/check_docs.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys, glob
 
@@ -184,5 +184,3 @@ p = RSTParser()
 for m in opencv_module_list:
     print "\n\n*************************** " + m + " *************************\n"
     p.check_module_docs(m)
-
-
diff --git a/doc/check_docs2.py b/doc/check_docs2.py
index 963c7ff1da..36d7068d86 100755
--- a/doc/check_docs2.py
+++ b/doc/check_docs2.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import os, sys, fnmatch, re
 
@@ -83,11 +83,15 @@ def get_cv2_object(name):
     elif name == "DescriptorExtractor":
         return cv2.DescriptorExtractor_create("ORB"), name
     elif name == "BackgroundSubtractor":
-        return cv2.BackgroundSubtractorMOG(), name
+        return cv2.createBackgroundSubtractorMOG(), name
     elif name == "StatModel":
         return cv2.KNearest(), name
     else:
-        return getattr(cv2, name)(), name
+        try:
+            obj = getattr(cv2, name)()
+        except AttributeError:
+            obj = getattr(cv2, "create" + name)()
+        return obj, name
 
 def compareSignatures(f, s):
     # function names
@@ -197,9 +201,9 @@ def process_module(module, path):
             hdrlist.append(os.path.join(root, filename))
 
     if module == "gpu":
-        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "cuda_devptrs.hpp"))
-        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "gpumat.hpp"))
-        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "stream_accessor.hpp"))
+        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "gpu_types.hpp"))
+        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "gpu.hpp"))
+        hdrlist.append(os.path.join(path, "..", "core", "include", "opencv2", "core", "gpu_stream_accessor.hpp"))
 
     decls = []
     for hname in hdrlist:
diff --git a/doc/conf.py b/doc/conf.py
index 7b9b02ecf4..2cbcdf8867 100755
--- a/doc/conf.py
+++ b/doc/conf.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 # -*- coding: utf-8 -*-
 #
@@ -28,6 +28,16 @@ sys.path.insert(0, os.path.abspath('.'))
 # Add any Sphinx extension module names here, as strings. They can be extensions
 # coming with Sphinx (named 'sphinx.ext.*') or your custom ones.
 extensions = ['sphinx.ext.pngmath', 'sphinx.ext.ifconfig', 'sphinx.ext.todo', 'sphinx.ext.extlinks', 'ocv', 'sphinx.ext.doctest']
+
+have_plantuml_ext = False
+if tags.has('plantuml'):
+  try:
+    import sphinxcontrib.plantuml
+    extensions.append("sphinxcontrib.plantuml")
+    have_plantuml_ext = True
+  except ImportError:
+    print "No module sphinxcontrib.plantuml found, sphinx will not render UML diagrams"
+
 doctest_test_doctest_blocks = 'block'
 
 # Add any paths that contain templates here, relative to this directory.
@@ -55,6 +65,7 @@ version_epoch = re.search("^W*#\W*define\W+CV_VERSION_EPOCH\W+(\d+)\W*$", versio
 version_major = re.search("^W*#\W*define\W+CV_VERSION_MAJOR\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
 version_minor = re.search("^W*#\W*define\W+CV_VERSION_MINOR\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
 version_patch = re.search("^W*#\W*define\W+CV_VERSION_REVISION\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
+version_status = re.search("^W*#\W*define\W+CV_VERSION_STATUS\W+\"(.*?)\"\W*$", version_file, re.MULTILINE).group(1)
 
 # The short X.Y version.
 version = version_epoch + '.' + version_major
@@ -62,6 +73,7 @@ version = version_epoch + '.' + version_major
 release = version_epoch + '.' + version_major + '.' + version_minor
 if version_patch:
     release = release + '.' + version_patch
+release += version_status
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -77,6 +89,9 @@ if version_patch:
 # directories to ignore when looking for source files.
 exclude_patterns = ['doc/tutorials/definitions']
 
+if not have_plantuml_ext:
+  exclude_patterns.append('**/uml/*')
+
 # The reST default role (used for this markup: `text`) to use for all documents.
 #default_role = None
 
@@ -239,7 +254,7 @@ latex_documents = [
    u'', 'manual'),
   ('doc/tutorials/tutorials', 'opencv_tutorials.tex', u'The OpenCV Tutorials',
    u'', 'manual'),
-  ('android/refman', 'opencv2manager.tex', u'The OpenCV Manager Manual',
+  ('platforms/android/refman', 'opencv2manager.tex', u'The OpenCV Manager Manual',
    u'', 'manual'),
 ]
 
@@ -284,120 +299,122 @@ latex_domain_indices = True
 # (source start file, name, description, authors, manual section).
 man_pages = [
     ('index', 'opencv', u'The OpenCV Reference Manual',
-     [u'opencv-dev@itseez.com'], 1)
+     [u'admin@opencv.org'], 1)
 ]
 
 # ---- External links for tutorials -----------------
 extlinks = {
-            'basicstructures' : ('http://opencv.itseez.com/modules/core/doc/basic_structures.html#%s', None),
-            'oldbasicstructures' : ('http://opencv.itseez.com/modules/core/doc/old_basic_structures.html#%s', None),
-            'readwriteimagevideo' : ('http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None),
-            'operationsonarrays' : ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html#%s', None),
-            'utilitysystemfunctions':('http://opencv.itseez.com/modules/core/doc/utility_and_system_functions_and_macros.html#%s', None),
-            'imgprocfilter':('http://opencv.itseez.com/modules/imgproc/doc/filtering.html#%s', None),
-            'svms':('http://opencv.itseez.com/modules/ml/doc/support_vector_machines.html#%s', None),
-            'drawingfunc':('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#%s', None),
-            'xmlymlpers':('http://opencv.itseez.com/modules/core/doc/xml_yaml_persistence.html#%s', None),
-            'hgvideo' : ('http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None),
-            'gpuinit' : ('http://opencv.itseez.com/modules/gpu/doc/initalization_and_information.html#%s', None),
-            'gpudatastructure' : ('http://opencv.itseez.com/modules/gpu/doc/data_structures.html#%s', None),
-            'gpuopmatrices' : ('http://opencv.itseez.com/modules/gpu/doc/operations_on_matrices.html#%s', None),
-            'gpuperelement' : ('http://opencv.itseez.com/modules/gpu/doc/per_element_operations.html#%s', None),
-            'gpuimgproc' : ('http://opencv.itseez.com/modules/gpu/doc/image_processing.html#%s', None),
-            'gpumatrixreduct' : ('http://opencv.itseez.com/modules/gpu/doc/matrix_reductions.html#%s', None),
-            'filtering':('http://opencv.itseez.com/modules/imgproc/doc/filtering.html#%s', None),
-            'flann' : ('http://opencv.itseez.com/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.html#%s', None ),
-            'calib3d' : ('http://opencv.itseez.com/trunk/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html#%s', None ),
-            'feature2d' : ('http://opencv.itseez.com/trunk/modules/imgproc/doc/feature_detection.html#%s', None ),
-            'imgproc_geometric' : ('http://opencv.itseez.com/trunk/modules/imgproc/doc/geometric_transformations.html#%s', None ),
+            'basicstructures' : ('http://docs.opencv.org/modules/core/doc/basic_structures.html#%s', None),
+            'oldbasicstructures' : ('http://docs.opencv.org/modules/core/doc/old_basic_structures.html#%s', None),
+            'readwriteimagevideo' : ('http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None),
+            'operationsonarrays' : ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html#%s', None),
+            'utilitysystemfunctions':('http://docs.opencv.org/modules/core/doc/utility_and_system_functions_and_macros.html#%s', None),
+            'imgprocfilter':('http://docs.opencv.org/modules/imgproc/doc/filtering.html#%s', None),
+            'svms':('http://docs.opencv.org/modules/ml/doc/support_vector_machines.html#%s', None),
+            'drawingfunc':('http://docs.opencv.org/modules/core/doc/drawing_functions.html#%s', None),
+            'xmlymlpers':('http://docs.opencv.org/modules/core/doc/xml_yaml_persistence.html#%s', None),
+            'hgvideo' : ('http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None),
+            'gpuinit' : ('http://docs.opencv.org/modules/gpu/doc/initalization_and_information.html#%s', None),
+            'gpudatastructure' : ('http://docs.opencv.org/modules/gpu/doc/data_structures.html#%s', None),
+            'gpuopmatrices' : ('http://docs.opencv.org/modules/gpu/doc/operations_on_matrices.html#%s', None),
+            'gpuperelement' : ('http://docs.opencv.org/modules/gpu/doc/per_element_operations.html#%s', None),
+            'gpuimgproc' : ('http://docs.opencv.org/modules/gpu/doc/image_processing.html#%s', None),
+            'gpumatrixreduct' : ('http://docs.opencv.org/modules/gpu/doc/matrix_reductions.html#%s', None),
+            'filtering':('http://docs.opencv.org/modules/imgproc/doc/filtering.html#%s', None),
+            'flann' : ('http://docs.opencv.org/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.html#%s', None ),
+            'calib3d' : ('http://docs.opencv.org/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html#%s', None ),
+            'feature2d' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html#%s', None ),
+            'imgproc_geometric' : ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html#%s', None ),
 
-            'opencv_group' : ('http://tech.groups.yahoo.com/group/OpenCV/%s', None),
+            # 'opencv_group' : ('http://answers.opencv.org/%s', None),
+            'opencv_qa' : ('http://answers.opencv.org/%s', None),
+            'how_to_contribute' : ('http://code.opencv.org/projects/opencv/wiki/How_to_contribute/%s', None),
 
-            'cvt_color': ('http://opencv.itseez.com/modules/imgproc/doc/miscellaneous_transformations.html?highlight=cvtcolor#cvtcolor%s', None),
-            'imread':    ('http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imread#imread%s', None),
-            'imwrite':   ('http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imwrite#imwrite%s', None),
-            'imshow':    ('http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=imshow#imshow%s', None),
-            'named_window': ('http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=namedwindow#namedwindow%s', None),
-            'wait_key': ('http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=waitkey#waitkey%s', None),
-            'add_weighted': ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=addweighted#addweighted%s', None),
-            'saturate_cast': ('http://opencv.itseez.com/modules/core/doc/utility_and_system_functions_and_macros.html?highlight=saturate_cast#saturate-cast%s', None),
-            'mat_zeros': ('http://opencv.itseez.com/modules/core/doc/basic_structures.html?highlight=zeros#mat-zeros%s', None),
-            'convert_to': ('http://opencv.itseez.com/modules/core/doc/basic_structures.html#mat-convertto%s', None),
-            'create_trackbar': ('http://opencv.itseez.com/modules/highgui/doc/user_interface.html?highlight=createtrackbar#createtrackbar%s', None),
-            'point': ('http://opencv.itseez.com/modules/core/doc/basic_structures.html#point%s', None),
-            'scalar': ('http://opencv.itseez.com/modules/core/doc/basic_structures.html#scalar%s', None),
-            'line': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#line%s', None),
-            'ellipse': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#ellipse%s', None),
-            'rectangle': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#rectangle%s', None),
-            'circle': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#circle%s', None),
-            'fill_poly': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#fillpoly%s', None),
-            'rng': ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=rng#rng%s', None),
-            'put_text': ('http://opencv.itseez.com/modules/core/doc/drawing_functions.html#puttext%s', None),
-            'gaussian_blur': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=gaussianblur#gaussianblur%s', None),
-            'blur': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=blur#blur%s', None),
-            'median_blur': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=medianblur#medianblur%s', None),
-            'bilateral_filter': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=bilateralfilter#bilateralfilter%s', None),
-            'erode': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=erode#erode%s', None),
-            'dilate': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=dilate#dilate%s', None),
-            'get_structuring_element': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=getstructuringelement#getstructuringelement%s', None),
-            'flood_fill': ( 'http://opencv.itseez.com/modules/imgproc/doc/miscellaneous_transformations.html?highlight=floodfill#floodfill%s', None),
-            'morphology_ex': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=morphologyex#morphologyex%s', None),
-            'pyr_down': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=pyrdown#pyrdown%s', None),
-            'pyr_up': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=pyrup#pyrup%s', None),
-            'resize': ('http://opencv.itseez.com/modules/imgproc/doc/geometric_transformations.html?highlight=resize#resize%s', None),
-            'threshold': ('http://opencv.itseez.com/modules/imgproc/doc/miscellaneous_transformations.html?highlight=threshold#threshold%s', None),
-            'filter2d': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=filter2d#filter2d%s', None),
-            'copy_make_border': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=copymakeborder#copymakeborder%s', None),
-            'sobel': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=sobel#sobel%s', None),
-            'scharr': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=scharr#scharr%s', None),
-            'laplacian': ('http://opencv.itseez.com/modules/imgproc/doc/filtering.html?highlight=laplacian#laplacian%s', None),
-            'canny': ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=canny#canny%s', None),
-            'copy_to': ('http://opencv.itseez.com/modules/core/doc/basic_structures.html?highlight=copyto#mat-copyto%s', None),
-            'hough_lines' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=houghlines#houghlines%s', None),
-            'hough_lines_p' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=houghlinesp#houghlinesp%s', None),
-            'hough_circles' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=houghcircles#houghcircles%s', None),
-            'remap' : ('http://opencv.itseez.com/modules/imgproc/doc/geometric_transformations.html?highlight=remap#remap%s', None),
-            'warp_affine' : ('http://opencv.itseez.com/modules/imgproc/doc/geometric_transformations.html?highlight=warpaffine#warpaffine%s' , None),
-            'get_rotation_matrix_2d' : ('http://opencv.itseez.com/modules/imgproc/doc/geometric_transformations.html?highlight=getrotationmatrix2d#getrotationmatrix2d%s', None),
-            'get_affine_transform' : ('http://opencv.itseez.com/modules/imgproc/doc/geometric_transformations.html?highlight=getaffinetransform#getaffinetransform%s', None),
-            'equalize_hist' : ('http://opencv.itseez.com/modules/imgproc/doc/histograms.html?highlight=equalizehist#equalizehist%s', None),
-            'split' : ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=split#split%s', None),
-            'calc_hist' : ('http://opencv.itseez.com/modules/imgproc/doc/histograms.html?highlight=calchist#calchist%s', None),
-            'normalize' : ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=normalize#normalize%s', None),
-            'match_template' : ('http://opencv.itseez.com/modules/imgproc/doc/object_detection.html?highlight=matchtemplate#matchtemplate%s', None),
-            'min_max_loc' : ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=minmaxloc#minmaxloc%s', None),
-            'mix_channels' : ( 'http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=mixchannels#mixchannels%s', None),
-            'calc_back_project' : ('http://opencv.itseez.com/modules/imgproc/doc/histograms.html?highlight=calcbackproject#calcbackproject%s', None),
-            'compare_hist' : ('http://opencv.itseez.com/modules/imgproc/doc/histograms.html?highlight=comparehist#comparehist%s', None),
-            'corner_harris' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=cornerharris#cornerharris%s', None),
-            'good_features_to_track' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=goodfeaturestotrack#goodfeaturestotrack%s', None),
-            'corner_min_eigenval' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=cornermineigenval#cornermineigenval%s', None),
-            'corner_eigenvals_and_vecs' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=cornereigenvalsandvecs#cornereigenvalsandvecs%s', None),
-            'corner_sub_pix' : ('http://opencv.itseez.com/modules/imgproc/doc/feature_detection.html?highlight=cornersubpix#cornersubpix%s', None),
-            'find_contours' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=findcontours#findcontours%s', None),
-            'convex_hull' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=convexhull#convexhull%s', None),
-            'draw_contours' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=drawcontours#drawcontours%s', None),
-            'bounding_rect' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=boundingrect#boundingrect%s', None),
-            'min_enclosing_circle' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=minenclosingcircle#minenclosingcircle%s', None),
-            'min_area_rect' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=minarearect#minarearect%s', None),
-            'fit_ellipse' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=fitellipse#fitellipse%s', None),
-            'moments' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=moments#moments%s', None),
-            'contour_area' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=contourarea#contourarea%s', None),
-            'arc_length' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=arclength#arclength%s', None),
-            'point_polygon_test' : ('http://opencv.itseez.com/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=pointpolygontest#pointpolygontest%s', None),
-            'feature_detector' : ( 'http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=featuredetector#FeatureDetector%s', None),
-            'feature_detector_detect' : ('http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=detect#featuredetector-detect%s', None ),
-            'surf_feature_detector' : ('http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=surffeaturedetector#surffeaturedetector%s', None ),
-            'draw_keypoints' : ('http://opencv.itseez.com/modules/features2d/doc/drawing_function_of_keypoints_and_matches.html?highlight=drawkeypoints#drawkeypoints%s', None ),
-            'descriptor_extractor': ( 'http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=descriptorextractor#descriptorextractor%s', None ),
-            'descriptor_extractor_compute' : ( 'http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=compute#descriptorextractor-compute%s', None ),
-            'surf_descriptor_extractor' : ( 'http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=surfdescriptorextractor#surfdescriptorextractor%s', None ),
-            'draw_matches' : ( 'http://opencv.itseez.com/modules/features2d/doc/drawing_function_of_keypoints_and_matches.html?highlight=drawmatches#drawmatches%s', None ),
-            'find_homography' : ('http://opencv.itseez.com/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html?highlight=findhomography#findhomography%s', None),
-            'perspective_transform' : ('http://opencv.itseez.com/modules/core/doc/operations_on_arrays.html?highlight=perspectivetransform#perspectivetransform%s', None ),
-            'flann_based_matcher' : ('http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html?highlight=flannbasedmatcher#flannbasedmatcher%s', None),
-            'brute_force_matcher' : ('http://opencv.itseez.com/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html?highlight=bruteforcematcher#bruteforcematcher%s', None ),
-            'cascade_classifier' : ('http://opencv.itseez.com/modules/objdetect/doc/cascade_classification.html?highlight=cascadeclassifier#cascadeclassifier%s', None ),
-            'cascade_classifier_load' : ('http://opencv.itseez.com/modules/objdetect/doc/cascade_classification.html?highlight=load#cascadeclassifier-load%s', None ),
-            'cascade_classifier_detect_multiscale' : ('http://opencv.itseez.com/modules/objdetect/doc/cascade_classification.html?highlight=detectmultiscale#cascadeclassifier-detectmultiscale%s', None )
+            'cvt_color': ('http://docs.opencv.org/modules/imgproc/doc/miscellaneous_transformations.html?highlight=cvtcolor#cvtcolor%s', None),
+            'imread':    ('http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imread#imread%s', None),
+            'imwrite':   ('http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html?highlight=imwrite#imwrite%s', None),
+            'imshow':    ('http://docs.opencv.org/modules/highgui/doc/user_interface.html?highlight=imshow#imshow%s', None),
+            'named_window': ('http://docs.opencv.org/modules/highgui/doc/user_interface.html?highlight=namedwindow#namedwindow%s', None),
+            'wait_key': ('http://docs.opencv.org/modules/highgui/doc/user_interface.html?highlight=waitkey#waitkey%s', None),
+            'add_weighted': ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=addweighted#addweighted%s', None),
+            'saturate_cast': ('http://docs.opencv.org/modules/core/doc/utility_and_system_functions_and_macros.html?highlight=saturate_cast#saturate-cast%s', None),
+            'mat_zeros': ('http://docs.opencv.org/modules/core/doc/basic_structures.html?highlight=zeros#mat-zeros%s', None),
+            'convert_to': ('http://docs.opencv.org/modules/core/doc/basic_structures.html#mat-convertto%s', None),
+            'create_trackbar': ('http://docs.opencv.org/modules/highgui/doc/user_interface.html?highlight=createtrackbar#createtrackbar%s', None),
+            'point': ('http://docs.opencv.org/modules/core/doc/basic_structures.html#point%s', None),
+            'scalar': ('http://docs.opencv.org/modules/core/doc/basic_structures.html#scalar%s', None),
+            'line': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#line%s', None),
+            'ellipse': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#ellipse%s', None),
+            'rectangle': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#rectangle%s', None),
+            'circle': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#circle%s', None),
+            'fill_poly': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#fillpoly%s', None),
+            'rng': ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=rng#rng%s', None),
+            'put_text': ('http://docs.opencv.org/modules/core/doc/drawing_functions.html#puttext%s', None),
+            'gaussian_blur': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=gaussianblur#gaussianblur%s', None),
+            'blur': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=blur#blur%s', None),
+            'median_blur': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=medianblur#medianblur%s', None),
+            'bilateral_filter': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=bilateralfilter#bilateralfilter%s', None),
+            'erode': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=erode#erode%s', None),
+            'dilate': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=dilate#dilate%s', None),
+            'get_structuring_element': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=getstructuringelement#getstructuringelement%s', None),
+            'flood_fill': ( 'http://docs.opencv.org/modules/imgproc/doc/miscellaneous_transformations.html?highlight=floodfill#floodfill%s', None),
+            'morphology_ex': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=morphologyex#morphologyex%s', None),
+            'pyr_down': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=pyrdown#pyrdown%s', None),
+            'pyr_up': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=pyrup#pyrup%s', None),
+            'resize': ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html?highlight=resize#resize%s', None),
+            'threshold': ('http://docs.opencv.org/modules/imgproc/doc/miscellaneous_transformations.html?highlight=threshold#threshold%s', None),
+            'filter2d': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=filter2d#filter2d%s', None),
+            'copy_make_border': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=copymakeborder#copymakeborder%s', None),
+            'sobel': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=sobel#sobel%s', None),
+            'scharr': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=scharr#scharr%s', None),
+            'laplacian': ('http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=laplacian#laplacian%s', None),
+            'canny': ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=canny#canny%s', None),
+            'copy_to': ('http://docs.opencv.org/modules/core/doc/basic_structures.html?highlight=copyto#mat-copyto%s', None),
+            'hough_lines' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=houghlines#houghlines%s', None),
+            'hough_lines_p' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=houghlinesp#houghlinesp%s', None),
+            'hough_circles' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=houghcircles#houghcircles%s', None),
+            'remap' : ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html?highlight=remap#remap%s', None),
+            'warp_affine' : ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html?highlight=warpaffine#warpaffine%s' , None),
+            'get_rotation_matrix_2d' : ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html?highlight=getrotationmatrix2d#getrotationmatrix2d%s', None),
+            'get_affine_transform' : ('http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html?highlight=getaffinetransform#getaffinetransform%s', None),
+            'equalize_hist' : ('http://docs.opencv.org/modules/imgproc/doc/histograms.html?highlight=equalizehist#equalizehist%s', None),
+            'split' : ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=split#split%s', None),
+            'calc_hist' : ('http://docs.opencv.org/modules/imgproc/doc/histograms.html?highlight=calchist#calchist%s', None),
+            'normalize' : ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=normalize#normalize%s', None),
+            'match_template' : ('http://docs.opencv.org/modules/imgproc/doc/object_detection.html?highlight=matchtemplate#matchtemplate%s', None),
+            'min_max_loc' : ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=minmaxloc#minmaxloc%s', None),
+            'mix_channels' : ( 'http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=mixchannels#mixchannels%s', None),
+            'calc_back_project' : ('http://docs.opencv.org/modules/imgproc/doc/histograms.html?highlight=calcbackproject#calcbackproject%s', None),
+            'compare_hist' : ('http://docs.opencv.org/modules/imgproc/doc/histograms.html?highlight=comparehist#comparehist%s', None),
+            'corner_harris' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=cornerharris#cornerharris%s', None),
+            'good_features_to_track' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=goodfeaturestotrack#goodfeaturestotrack%s', None),
+            'corner_min_eigenval' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=cornermineigenval#cornermineigenval%s', None),
+            'corner_eigenvals_and_vecs' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=cornereigenvalsandvecs#cornereigenvalsandvecs%s', None),
+            'corner_sub_pix' : ('http://docs.opencv.org/modules/imgproc/doc/feature_detection.html?highlight=cornersubpix#cornersubpix%s', None),
+            'find_contours' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=findcontours#findcontours%s', None),
+            'convex_hull' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=convexhull#convexhull%s', None),
+            'draw_contours' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=drawcontours#drawcontours%s', None),
+            'bounding_rect' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=boundingrect#boundingrect%s', None),
+            'min_enclosing_circle' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=minenclosingcircle#minenclosingcircle%s', None),
+            'min_area_rect' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=minarearect#minarearect%s', None),
+            'fit_ellipse' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=fitellipse#fitellipse%s', None),
+            'moments' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=moments#moments%s', None),
+            'contour_area' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=contourarea#contourarea%s', None),
+            'arc_length' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=arclength#arclength%s', None),
+            'point_polygon_test' : ('http://docs.opencv.org/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=pointpolygontest#pointpolygontest%s', None),
+            'feature_detector' : ( 'http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=featuredetector#FeatureDetector%s', None),
+            'feature_detector_detect' : ('http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=detect#featuredetector-detect%s', None ),
+            'surf_feature_detector' : ('http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_feature_detectors.html?highlight=surffeaturedetector#surffeaturedetector%s', None ),
+            'draw_keypoints' : ('http://docs.opencv.org/modules/features2d/doc/drawing_function_of_keypoints_and_matches.html?highlight=drawkeypoints#drawkeypoints%s', None ),
+            'descriptor_extractor': ( 'http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=descriptorextractor#descriptorextractor%s', None ),
+            'descriptor_extractor_compute' : ( 'http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=compute#descriptorextractor-compute%s', None ),
+            'surf_descriptor_extractor' : ( 'http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_extractors.html?highlight=surfdescriptorextractor#surfdescriptorextractor%s', None ),
+            'draw_matches' : ( 'http://docs.opencv.org/modules/features2d/doc/drawing_function_of_keypoints_and_matches.html?highlight=drawmatches#drawmatches%s', None ),
+            'find_homography' : ('http://docs.opencv.org/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html?highlight=findhomography#findhomography%s', None),
+            'perspective_transform' : ('http://docs.opencv.org/modules/core/doc/operations_on_arrays.html?highlight=perspectivetransform#perspectivetransform%s', None ),
+            'flann_based_matcher' : ('http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html?highlight=flannbasedmatcher#flannbasedmatcher%s', None),
+            'brute_force_matcher' : ('http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html?highlight=bruteforcematcher#bruteforcematcher%s', None ),
+            'cascade_classifier' : ('http://docs.opencv.org/modules/objdetect/doc/cascade_classification.html?highlight=cascadeclassifier#cascadeclassifier%s', None ),
+            'cascade_classifier_load' : ('http://docs.opencv.org/modules/objdetect/doc/cascade_classification.html?highlight=load#cascadeclassifier-load%s', None ),
+            'cascade_classifier_detect_multiscale' : ('http://docs.opencv.org/modules/objdetect/doc/cascade_classification.html?highlight=detectmultiscale#cascadeclassifier-detectmultiscale%s', None )
            }
diff --git a/doc/mymath.sty b/doc/mymath.sty
index 491c160cc3..4da8282121 100644
--- a/doc/mymath.sty
+++ b/doc/mymath.sty
@@ -39,4 +39,3 @@
  #7 & #8 & #9
 \end{bmatrix}
 }
-
diff --git a/doc/ocv.py b/doc/ocv.py
index 4ff8a6deb5..a22d3daa56 100755
--- a/doc/ocv.py
+++ b/doc/ocv.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 # -*- coding: utf-8 -*-
 """
     ocv domain, a modified copy of sphinx.domains.cpp + shpinx.domains.python.
@@ -267,10 +267,6 @@ class OCVPyModulelevel(OCVPyObject):
         else:
             return ''
 
-class OCVPyOldModulelevel(OCVPyModulelevel):
-    directive_prefix = 'pyold'
-    pass
-
 class OCVPyXRefRole(XRefRole):
     def process_link(self, env, refnode, has_explicit_title, title, target):
         refnode['ocv:module'] = env.temp_data.get('ocv:module')
@@ -1491,7 +1487,6 @@ class OCVDomain(Domain):
         'cfunction': ObjType(l_('cfunction'), 'cfunc', 'cfuncx'),
         'jfunction': ObjType(l_('jfunction'), 'jfunc', 'jfuncx'),
         'pyfunction': ObjType(l_('pyfunction'), 'pyfunc'),
-        'pyoldfunction': ObjType(l_('pyoldfunction'), 'pyoldfunc'),
         'member':   ObjType(l_('member'),   'member'),
         'emember':   ObjType(l_('emember'),   'emember'),
         'type':     ObjType(l_('type'),     'type'),
@@ -1505,7 +1500,6 @@ class OCVDomain(Domain):
         'cfunction':    OCVCFunctionObject,
         'jfunction':    OCVJavaFunctionObject,
         'pyfunction':   OCVPyModulelevel,
-        'pyoldfunction':   OCVPyOldModulelevel,
         'member':       OCVMemberObject,
         'emember':      OCVEnumMemberObject,
         'type':         OCVTypeObject,
@@ -1522,7 +1516,6 @@ class OCVDomain(Domain):
         'jfunc' :  OCVXRefRole(fix_parens=True),
         'jfuncx' :  OCVXRefRole(),
         'pyfunc' :  OCVPyXRefRole(),
-        'pyoldfunc' :  OCVPyXRefRole(),
         'member': OCVXRefRole(),
         'emember': OCVXRefRole(),
         'type':   OCVXRefRole(),
@@ -1612,7 +1605,6 @@ class OCVDomain(Domain):
             'cfunction':     _('C function'),
             'jfunction':     _('Java method'),
             'pyfunction':    _('Python function'),
-            'pyoldfunction': _('Legacy Python function'),
             'member':        _('C++ member'),
             'emember':       _('enum member'),
             'type':          _('C/C++ type'),
diff --git a/doc/opencv-logo2.png b/doc/opencv-logo2.png
index 615fd2add4..bc71a2ae50 100644
Binary files a/doc/opencv-logo2.png and b/doc/opencv-logo2.png differ
diff --git a/doc/opencv_cheatsheet.tex b/doc/opencv_cheatsheet.tex
index e76bd016e9..fc0c0fa41b 100644
--- a/doc/opencv_cheatsheet.tex
+++ b/doc/opencv_cheatsheet.tex
@@ -75,11 +75,11 @@
 % if using A4 paper. (This probably isn't strictly necessary.)
 % If using another size paper, use default 1cm margins.
 \ifthenelse{\lengthtest { \paperwidth = 11in}}
-	{ \geometry{top=.5in,left=.5in,right=.5in,bottom=.5in} }
-	{\ifthenelse{ \lengthtest{ \paperwidth = 297mm}}
-		{\geometry{top=1cm,left=1cm,right=1cm,bottom=1cm} }
-		{\geometry{top=1cm,left=1cm,right=1cm,bottom=1cm} }
-	}
+    { \geometry{top=.5in,left=.5in,right=.5in,bottom=.5in} }
+    {\ifthenelse{ \lengthtest{ \paperwidth = 297mm}}
+        {\geometry{top=1cm,left=1cm,right=1cm,bottom=1cm} }
+        {\geometry{top=1cm,left=1cm,right=1cm,bottom=1cm} }
+    }
 
 % Turn off header and footer
 % \pagestyle{empty}
diff --git a/doc/patch_refman_latex.py b/doc/patch_refman_latex.py
index 352c46cb56..ff762fc8f3 100755
--- a/doc/patch_refman_latex.py
+++ b/doc/patch_refman_latex.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys
 
diff --git a/doc/pattern_tools/gen_pattern.py b/doc/pattern_tools/gen_pattern.py
index 45b45af2db..3643b6d3b2 100755
--- a/doc/pattern_tools/gen_pattern.py
+++ b/doc/pattern_tools/gen_pattern.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 """gen_pattern.py
 To run:
diff --git a/doc/pattern_tools/svgfig.py b/doc/pattern_tools/svgfig.py
index bf182a8b09..c690c9ff08 100755
--- a/doc/pattern_tools/svgfig.py
+++ b/doc/pattern_tools/svgfig.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 # svgfig.py copyright (C) 2008 Jim Pivarski <jpivarski@gmail.com>
 #
@@ -3667,4 +3667,3 @@ class YErrorBars:
             output.append(LineAxis(x, start, x, end, start, end, bars, False, False, **self.attr).SVG(trans))
 
         return output
-
diff --git a/doc/reformat.py b/doc/reformat.py
index 00e4aae9e0..017efebb38 100755
--- a/doc/reformat.py
+++ b/doc/reformat.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import os, sys, re
 
diff --git a/doc/tutorials/contrib/retina_model/images/retina_TreeHdr_retina.jpg b/doc/tutorials/bioinspired/retina_model/images/retina_TreeHdr_retina.jpg
similarity index 100%
rename from doc/tutorials/contrib/retina_model/images/retina_TreeHdr_retina.jpg
rename to doc/tutorials/bioinspired/retina_model/images/retina_TreeHdr_retina.jpg
diff --git a/doc/tutorials/contrib/retina_model/images/retina_TreeHdr_small.jpg b/doc/tutorials/bioinspired/retina_model/images/retina_TreeHdr_small.jpg
similarity index 100%
rename from doc/tutorials/contrib/retina_model/images/retina_TreeHdr_small.jpg
rename to doc/tutorials/bioinspired/retina_model/images/retina_TreeHdr_small.jpg
diff --git a/doc/tutorials/contrib/retina_model/images/studentsSample_input.jpg b/doc/tutorials/bioinspired/retina_model/images/studentsSample_input.jpg
similarity index 100%
rename from doc/tutorials/contrib/retina_model/images/studentsSample_input.jpg
rename to doc/tutorials/bioinspired/retina_model/images/studentsSample_input.jpg
diff --git a/doc/tutorials/contrib/retina_model/images/studentsSample_magno.jpg b/doc/tutorials/bioinspired/retina_model/images/studentsSample_magno.jpg
similarity index 100%
rename from doc/tutorials/contrib/retina_model/images/studentsSample_magno.jpg
rename to doc/tutorials/bioinspired/retina_model/images/studentsSample_magno.jpg
diff --git a/doc/tutorials/contrib/retina_model/images/studentsSample_parvo.jpg b/doc/tutorials/bioinspired/retina_model/images/studentsSample_parvo.jpg
similarity index 100%
rename from doc/tutorials/contrib/retina_model/images/studentsSample_parvo.jpg
rename to doc/tutorials/bioinspired/retina_model/images/studentsSample_parvo.jpg
diff --git a/doc/tutorials/contrib/retina_model/retina_model.rst b/doc/tutorials/bioinspired/retina_model/retina_model.rst
similarity index 79%
rename from doc/tutorials/contrib/retina_model/retina_model.rst
rename to doc/tutorials/bioinspired/retina_model/retina_model.rst
index 184bb343e1..e8527ee8b6 100644
--- a/doc/tutorials/contrib/retina_model/retina_model.rst
+++ b/doc/tutorials/bioinspired/retina_model/retina_model.rst
@@ -21,7 +21,7 @@ In this tutorial you will learn how to:
 General overview
 ================
 
-The proposed model originates from Jeanny Herault's research at `Gipsa <http://www.gipsa-lab.inpg.fr>`_. It is involved in image processing applications with `Listic <http://www.listic.univ-savoie.fr>`_ (code maintainer) lab. This is not a complete model but it already present interesting properties that can be involved for enhanced image processing experience. The model allows the following human retina properties to be used :
+The proposed model originates from Jeanny Herault's research [herault2010]_ at `Gipsa <http://www.gipsa-lab.inpg.fr>`_. It is involved in image processing applications with `Listic <http://www.listic.univ-savoie.fr>`_ (code maintainer and user) lab. This is not a complete model but it already present interesting properties that can be involved for enhanced image processing experience. The model allows the following human retina properties to be used :
 
 * spectral whitening that has 3 important effects: high spatio-temporal frequency signals canceling (noise), mid-frequencies details enhancement and low frequencies luminance energy reduction. This *all in one* property directly allows visual signals cleaning of classical undesired distortions introduced by image sensors and input luminance range.
 
@@ -37,7 +37,7 @@ In the figure below, the OpenEXR image sample *CrissyField.exr*, a High Dynamic
    :alt: A High dynamic range image linearly rescaled within range [0-255].
    :align: center
 
-In the following image, as your retina does, local luminance adaptation, spatial noise removal and spectral whitening work together and transmit accurate information on lower range 8bit data channels. On this picture, noise in significantly removed, local details hidden by strong luminance contrasts are enhanced. Output image keeps its naturalness and visual content is enhanced.
+In the following image, applying the ideas proposed in [benoit2010]_, as your retina does, local luminance adaptation, spatial noise removal and spectral whitening work together and transmit accurate information on lower range 8bit data channels. On this picture, noise in significantly removed, local details hidden by strong luminance contrasts are enhanced. Output image keeps its naturalness and visual content is enhanced. Color processing is based on the color multiplexing/demultiplexing method proposed in [chaix2007]_.
 
 .. image:: images/retina_TreeHdr_retina.jpg
    :alt: A High dynamic range image compressed within range [0-255] using the retina.
@@ -86,31 +86,37 @@ This model can be used basically for spatio-temporal video effects but also in t
 
 * performing motion analysis also taking benefit of the previously cited properties.
 
+Literature
+==========
 For more information, refer to the following papers :
 
-* Benoit A., Caplier A., Durette B., Herault, J., "Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
+.. [benoit2010] Benoit A., Caplier A., Durette B., Herault, J., "Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
 
 * Please have a look at the reference work of Jeanny Herault that you can read in his book :
 
-Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+.. [herault2010] Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
 
 This retina filter code includes the research contributions of phd/research collegues from which code has been redrawn by the author :
 
-* take a look at the *retinacolor.hpp* module to discover Brice Chaix de Lavarene phD color mosaicing/demosaicing and his reference paper: B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+* take a look at the *retinacolor.hpp* module to discover Brice Chaix de Lavarene phD color mosaicing/demosaicing and his reference paper:
 
-* take a look at *imagelogpolprojection.hpp* to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions. ====> more information in the above cited Jeanny Heraults's book.
+.. [chaix2007] B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+
+* take a look at *imagelogpolprojection.hpp* to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions. More informations in the above cited Jeanny Heraults's book.
 
 Code tutorial
 =============
 
-Please refer to the original tutorial source code in file *opencv_folder/samples/cpp/tutorial_code/contrib/retina_tutorial.cpp*.
+Please refer to the original tutorial source code in file *opencv_folder/samples/cpp/tutorial_code/bioinspired/retina_tutorial.cpp*.
 
-To compile it, assuming OpenCV is correctly installed, use the following command. It requires the opencv_core *(cv::Mat and friends objects management)*, opencv_highgui *(display and image/video read)* and opencv_contrib *(Retina description)* libraries to compile.
+**Note :** do not forget that the retina model is included in the following namespace : *cv::bioinspired*.
+
+To compile it, assuming OpenCV is correctly installed, use the following command. It requires the opencv_core *(cv::Mat and friends objects management)*, opencv_highgui *(display and image/video read)* and opencv_bioinspired *(Retina description)* libraries to compile.
 
 .. code-block:: cpp
 
    // compile
-   gcc retina_tutorial.cpp -o Retina_tuto -lopencv_core -lopencv_highgui -lopencv_contrib
+   gcc retina_tutorial.cpp -o Retina_tuto -lopencv_core -lopencv_highgui -lopencv_bioinspired
 
    // Run commands : add 'log' as a last parameter to apply a spatial log sampling (simulates retina sampling)
    // run on webcam
@@ -124,7 +130,7 @@ To compile it, assuming OpenCV is correctly installed, use the following command
 
 Here is a code explanation :
 
-Retina definition is present in the contrib package and a simple include allows to use it
+Retina definition is present in the bioinspired package and a simple include allows to use it. You can rather use the specific header : *opencv2/bioinspired.hpp* if you prefer but then include the other required openv modules : *opencv2/core.hpp* and *opencv2/highgui.hpp*
 
 .. code-block:: cpp
 
@@ -225,72 +231,71 @@ Once all input parameters are processed, a first image should have been loaded,
         return -1;
     }
 
-Now, everything is ready to run the retina model. I propose here to allocate a retina instance and to manage the eventual log sampling option. The Retina constructor expects at least a cv::Size object that shows the input data size that will have to be managed. One can activate other options such as color and its related color multiplexing strategy (here Bayer multiplexing is chosen using enum cv::RETINA_COLOR_BAYER). If using log sampling, the image reduction factor (smaller output images) and log sampling strengh can be adjusted.
+Now, everything is ready to run the retina model. I propose here to allocate a retina instance and to manage the eventual log sampling option. The Retina constructor expects at least a cv::Size object that shows the input data size that will have to be managed. One can activate other options such as color and its related color multiplexing strategy (here Bayer multiplexing is chosen using *enum cv::bioinspired::RETINA_COLOR_BAYER*). If using log sampling, the image reduction factor (smaller output images) and log sampling strengh can be adjusted.
 
 .. code-block:: cpp
 
-        // pointer to a retina object
-        cv::Ptr<cv::Retina> myRetina;
-
-        // if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
-        if (useLogSampling)
-        {
-            myRetina = new cv::Retina(inputFrame.size(), true, cv::RETINA_COLOR_BAYER, true, 2.0, 10.0);
-        }
-        else// -> else allocate "classical" retina :
-            myRetina = new cv::Retina(inputFrame.size());
+    // pointer to a retina object
+    cv::Ptr<cv::bioinspired::Retina> myRetina;
 
+    // if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
+    if (useLogSampling)
+    {
+        myRetina = cv::bioinspired::createRetina(inputFrame.size(), true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+    }
+    else// -> else allocate "classical" retina :
+        myRetina = cv::bioinspired::createRetina(inputFrame.size());
 
 Once done, the proposed code writes a default xml file that contains the default parameters of the retina. This is useful to make your own config using this template. Here generated template xml file is called *RetinaDefaultParameters.xml*.
 
 .. code-block:: cpp
 
-        // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
-        myRetina->write("RetinaDefaultParameters.xml");
+    // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
+    myRetina->write("RetinaDefaultParameters.xml");
 
 In the following line, the retina attempts to load another xml file called *RetinaSpecificParameters.xml*. If you created it and introduced your own setup, it will be loaded, in the other case, default retina parameters are used.
 
 .. code-block:: cpp
 
-        // load parameters if file exists
-        myRetina->setup("RetinaSpecificParameters.xml");
+    // load parameters if file exists
+    myRetina->setup("RetinaSpecificParameters.xml");
 
 It is not required here but just to show it is possible, you can reset the retina buffers to zero to force it to forget past events.
 
 .. code-block:: cpp
 
-        // reset all retina buffers (imagine you close your eyes for a long time)
-        myRetina->clearBuffers();
+    // reset all retina buffers (imagine you close your eyes for a long time)
+    myRetina->clearBuffers();
 
 Now, it is time to run the retina ! First create some output buffers ready to receive the two retina channels outputs
 
 .. code-block:: cpp
 
-        // declare retina output buffers
-        cv::Mat retinaOutput_parvo;
-        cv::Mat retinaOutput_magno;
+    // declare retina output buffers
+    cv::Mat retinaOutput_parvo;
+    cv::Mat retinaOutput_magno;
 
 Then, run retina in a loop, load new frames from video sequence if necessary and get retina outputs back to dedicated buffers.
 
 .. code-block:: cpp
 
-        // processing loop with no stop condition
-        while(true)
-        {
-            // if using video stream, then, grabbing a new frame, else, input remains the same
-            if (videoCapture.isOpened())
-                videoCapture>>inputFrame;
+    // processing loop with no stop condition
+    while(true)
+    {
+        // if using video stream, then, grabbing a new frame, else, input remains the same
+        if (videoCapture.isOpened())
+            videoCapture>>inputFrame;
 
-            // run retina filter on the loaded input frame
-            myRetina->run(inputFrame);
-            // Retrieve and display retina output
-            myRetina->getParvo(retinaOutput_parvo);
-            myRetina->getMagno(retinaOutput_magno);
-            cv::imshow("retina input", inputFrame);
-            cv::imshow("Retina Parvo", retinaOutput_parvo);
-            cv::imshow("Retina Magno", retinaOutput_magno);
-            cv::waitKey(10);
-        }
+        // run retina filter on the loaded input frame
+        myRetina->run(inputFrame);
+        // Retrieve and display retina output
+        myRetina->getParvo(retinaOutput_parvo);
+        myRetina->getMagno(retinaOutput_magno);
+        cv::imshow("retina input", inputFrame);
+        cv::imshow("Retina Parvo", retinaOutput_parvo);
+        cv::imshow("Retina Magno", retinaOutput_magno);
+        cv::waitKey(10);
+    }
 
 That's done ! But if you want to secure the system, take care and manage Exceptions. The retina can throw some when it sees irrelevant data (no input frame, wrong setup, etc.).
 Then, i recommend to surround all the retina code by a try/catch system like this :
@@ -323,29 +328,28 @@ Once done open the configuration file *RetinaDefaultParameters.xml* generated by
 
 .. code-block:: cpp
 
-        <?xml version="1.0"?>
-        <opencv_storage>
-        <OPLandIPLparvo>
-          <colorMode>1</colorMode>
-          <normaliseOutput>1</normaliseOutput>
-          <photoreceptorsLocalAdaptationSensitivity>7.0e-01</photoreceptorsLocalAdaptationSensitivity>
-          <photoreceptorsTemporalConstant>5.0e-01</photoreceptorsTemporalConstant>
-          <photoreceptorsSpatialConstant>5.3e-01</photoreceptorsSpatialConstant>
-          <horizontalCellsGain>0.</horizontalCellsGain>
-          <hcellsTemporalConstant>1.</hcellsTemporalConstant>
-          <hcellsSpatialConstant>7.</hcellsSpatialConstant>
-          <ganglionCellsSensitivity>7.0e-01</ganglionCellsSensitivity></OPLandIPLparvo>
-        <IPLmagno>
-          <normaliseOutput>1</normaliseOutput>
-          <parasolCells_beta>0.</parasolCells_beta>
-          <parasolCells_tau>0.</parasolCells_tau>
-          <parasolCells_k>7.</parasolCells_k>
-          <amacrinCellsTemporalCutFrequency>1.2e+00</amacrinCellsTemporalCutFrequency>
-          <V0CompressionParameter>9.5e-01</V0CompressionParameter>
-          <localAdaptintegration_tau>0.</localAdaptintegration_tau>
-          <localAdaptintegration_k>7.</localAdaptintegration_k></IPLmagno>
-        </opencv_storage>
-
+    <?xml version="1.0"?>
+    <opencv_storage>
+    <OPLandIPLparvo>
+        <colorMode>1</colorMode>
+        <normaliseOutput>1</normaliseOutput>
+        <photoreceptorsLocalAdaptationSensitivity>7.5e-01</photoreceptorsLocalAdaptationSensitivity>
+        <photoreceptorsTemporalConstant>9.0e-01</photoreceptorsTemporalConstant>
+        <photoreceptorsSpatialConstant>5.7e-01</photoreceptorsSpatialConstant>
+        <horizontalCellsGain>0.01</horizontalCellsGain>
+        <hcellsTemporalConstant>0.5</hcellsTemporalConstant>
+        <hcellsSpatialConstant>7.</hcellsSpatialConstant>
+        <ganglionCellsSensitivity>7.5e-01</ganglionCellsSensitivity></OPLandIPLparvo>
+    <IPLmagno>
+        <normaliseOutput>1</normaliseOutput>
+        <parasolCells_beta>0.</parasolCells_beta>
+        <parasolCells_tau>0.</parasolCells_tau>
+        <parasolCells_k>7.</parasolCells_k>
+        <amacrinCellsTemporalCutFrequency>2.0e+00</amacrinCellsTemporalCutFrequency>
+        <V0CompressionParameter>9.5e-01</V0CompressionParameter>
+        <localAdaptintegration_tau>0.</localAdaptintegration_tau>
+        <localAdaptintegration_k>7.</localAdaptintegration_k></IPLmagno>
+    </opencv_storage>
 
 Here are some hints but actually, the best parameter setup depends more on what you want to do with the retina rather than the images input that you give to retina. Apart from the more specific case of High Dynamic Range images (HDR) that require more specific setup for specific luminance compression objective, the retina behaviors should be rather stable from content to content. Note that OpenCV is able to manage such HDR format thanks to the OpenEXR images compatibility.
 
@@ -381,7 +385,7 @@ This parameter set tunes the neural network connected to the photo-receptors, th
 
 * **horizontalCellsGain** here is a critical parameter ! If you are not interested by the mean luminance and focus on details enhancement, then, set to zero. But if you want to keep some environment luminance data, let some low spatial frequencies pass into the system and set a higher value (<1).
 
-* **hcellsTemporalConstant** similar to photo-receptors, this acts on the temporal constant of a low pass temporal filter that smooths input data. Here, a high value generates a high retina after effect while a lower value makes the retina more reactive.
+* **hcellsTemporalConstant** similar to photo-receptors, this acts on the temporal constant of a low pass temporal filter that smooths input data. Here, a high value generates a high retina after effect while a lower value makes the retina more reactive. This value should be lower than **photoreceptorsTemporalConstant** to limit strong retina after effects.
 
 * **hcellsSpatialConstant** is the spatial constant of the low pass filter of these cells filter. It specifies the lowest spatial frequency allowed in the following. Visually, a high value leads to very low spatial frequencies processing and leads to salient halo effects. Lower values reduce this effect but the limit is : do not go lower than the value of **photoreceptorsSpatialConstant**. Those 2 parameters actually specify the spatial band-pass of the retina.
 
diff --git a/doc/tutorials/bioinspired/retina_model/retina_model.rst~ b/doc/tutorials/bioinspired/retina_model/retina_model.rst~
new file mode 100644
index 0000000000..50bdcbb629
--- /dev/null
+++ b/doc/tutorials/bioinspired/retina_model/retina_model.rst~
@@ -0,0 +1,418 @@
+.. _Retina_Model:
+
+Discovering the human retina and its use for image processing
+*************************************************************
+
+Goal
+=====
+
+I present here a model of human retina that shows some interesting properties for image preprocessing and enhancement.
+In this tutorial you will learn how to:
+
+.. container:: enumeratevisibleitemswithsquare
+
+   + discover the main two channels outing from your retina
+
+   + see the basics to use the retina model
+
+   + discover some parameters tweaks
+
+
+General overview
+================
+
+The proposed model originates from Jeanny Herault's research [herault2010]_ at `Gipsa <http://www.gipsa-lab.inpg.fr>`_. It is involved in image processing applications with `Listic <http://www.listic.univ-savoie.fr>`_ (code maintainer and user) lab. This is not a complete model but it already present interesting properties that can be involved for enhanced image processing experience. The model allows the following human retina properties to be used :
+
+* spectral whitening that has 3 important effects: high spatio-temporal frequency signals canceling (noise), mid-frequencies details enhancement and low frequencies luminance energy reduction. This *all in one* property directly allows visual signals cleaning of classical undesired distortions introduced by image sensors and input luminance range.
+
+* local logarithmic luminance compression allows details to be enhanced even in low light conditions.
+
+* decorrelation of the details information (Parvocellular output channel) and transient information (events, motion made available at the Magnocellular output channel).
+
+The first two points are illustrated below :
+
+In the figure below, the OpenEXR image sample *CrissyField.exr*, a High Dynamic Range image is shown. In order to make it visible on this web-page, the original input image is linearly rescaled to the classical image luminance range [0-255] and is converted to 8bit/channel format. Such strong conversion hides many details because of too strong local contrasts. Furthermore, noise energy is also strong and pollutes visual information.
+
+.. image:: images/retina_TreeHdr_small.jpg
+   :alt: A High dynamic range image linearly rescaled within range [0-255].
+   :align: center
+
+In the following image, applying the ideas proposed in [benoit2010]_, as your retina does, local luminance adaptation, spatial noise removal and spectral whitening work together and transmit accurate information on lower range 8bit data channels. On this picture, noise in significantly removed, local details hidden by strong luminance contrasts are enhanced. Output image keeps its naturalness and visual content is enhanced. Color processing is based on the color multiplexing/demultiplexing method proposed in [chaix2007]_.
+
+.. image:: images/retina_TreeHdr_retina.jpg
+   :alt: A High dynamic range image compressed within range [0-255] using the retina.
+   :align: center
+
+
+*Note :* image sample can be downloaded from the `OpenEXR website <http://www.openexr.com>`_. Regarding this demonstration, before retina processing, input image has been linearly rescaled within 0-255 keeping its channels float format. 5% of its histogram ends has been cut (mostly removes wrong HDR pixels). Check out the sample *opencv/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp* for similar processing. The following demonstration will only consider classical 8bit/channel images.
+
+The retina model output channels
+================================
+
+The retina model presents two outputs that benefit from the above cited behaviors.
+
+* The first one is called the Parvocellular channel. It is mainly active in the foveal retina area (high resolution central vision with color sensitive photo-receptors), its aim is to provide accurate color vision for visual details remaining static on the retina. On the other hand objects moving on the retina projection are blurred.
+
+* The second well known channel is the Magnocellular channel. It is mainly active in the retina peripheral vision and send signals related to change events (motion, transient events, etc.). These outing signals also help visual system to focus/center retina on 'transient'/moving areas for more detailed analysis thus improving visual scene context and object classification.
+
+**NOTE :** regarding the proposed model, contrary to the real retina, we apply these two channels on the entire input images using the same resolution. This allows enhanced visual details and motion information to be extracted on all the considered images... but remember, that these two channels are complementary. For example, if Magnocellular channel gives strong energy in an area, then, the Parvocellular channel is certainly blurred there since there is a transient event.
+
+As an illustration, we apply in the following the retina model on a webcam video stream of a dark visual scene. In this visual scene, captured in an amphitheater of the university, some students are moving while talking to the teacher.
+
+In this video sequence, because of the dark ambiance, signal to noise ratio is low and color artifacts are present on visual features edges because of the low quality image capture tool-chain.
+
+.. image:: images/studentsSample_input.jpg
+   :alt: an input video stream extract sample
+   :align: center
+
+Below is shown the retina foveal vision applied on the entire image. In the used retina configuration, global luminance is preserved and local contrasts are enhanced. Also, signal to noise ratio is improved : since high frequency spatio-temporal noise is reduced, enhanced details are not corrupted by any enhanced noise.
+
+.. image:: images/studentsSample_parvo.jpg
+   :alt: the retina Parvocellular output. Enhanced details, luminance adaptation and noise removal. A processing tool for image analysis.
+   :align: center
+
+Below is the output of the Magnocellular output of the retina model. Its signals are strong where transient events occur. Here, a student is moving at the bottom of the image thus generating high energy. The remaining of the image is static however, it is corrupted by a strong noise. Here, the retina filters out most of the noise thus generating low false motion area 'alarms'. This channel can be used as a transient/moving areas detector : it would provide relevant information for a low cost segmentation tool that would highlight areas in which an event is occurring.
+
+.. image:: images/studentsSample_magno.jpg
+   :alt: the retina Magnocellular output. Enhanced transient signals (motion, etc.). A preprocessing tool for event detection.
+   :align: center
+
+Retina use case
+===============
+
+This model can be used basically for spatio-temporal video effects but also in the aim of :
+
+* performing texture analysis with enhanced signal to noise ratio and enhanced details robust against input images luminance ranges (check out the Parvocellular retina channel output)
+
+* performing motion analysis also taking benefit of the previously cited properties.
+
+Literature
+==========
+For more information, refer to the following papers :
+
+.. [benoit2010] Benoit A., Caplier A., Durette B., Herault, J., "Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
+
+* Please have a look at the reference work of Jeanny Herault that you can read in his book :
+
+.. [herault2010] Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+
+This retina filter code includes the research contributions of phd/research collegues from which code has been redrawn by the author :
+
+* take a look at the *retinacolor.hpp* module to discover Brice Chaix de Lavarene phD color mosaicing/demosaicing and his reference paper:
+
+.. [chaix2007] B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+
+* take a look at *imagelogpolprojection.hpp* to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions. More informations in the above cited Jeanny Heraults's book.
+
+Code tutorial
+=============
+
+Please refer to the original tutorial source code in file *opencv_folder/samples/cpp/tutorial_code/bioinspired/retina_tutorial.cpp*.
+
+**Note :** do not forget that the retina model is included in the following namespace : *cv::bioinspired*.
+
+To compile it, assuming OpenCV is correctly installed, use the following command. It requires the opencv_core *(cv::Mat and friends objects management)*, opencv_highgui *(display and image/video read)* and opencv_bioinspired *(Retina description)* libraries to compile.
+
+.. code-block:: cpp
+
+   // compile
+   gcc retina_tutorial.cpp -o Retina_tuto -lopencv_core -lopencv_highgui -lopencv_bioinspired
+
+   // Run commands : add 'log' as a last parameter to apply a spatial log sampling (simulates retina sampling)
+   // run on webcam
+   ./Retina_tuto -video
+   // run on video file
+   ./Retina_tuto -video myVideo.avi
+   // run on an image
+   ./Retina_tuto -image myPicture.jpg
+   // run on an image with log sampling
+   ./Retina_tuto -image myPicture.jpg log
+
+Here is a code explanation :
+
+Retina definition is present in the bioinspired package and a simple include allows to use it. You can rather use the specific header : *opencv2/bioinspired.hpp* if you prefer but then include the other required openv modules : *opencv2/core.hpp* and *opencv2/highgui.hpp* 
+
+.. code-block:: cpp
+
+   #include "opencv2/opencv.hpp"
+
+Provide user some hints to run the program with a help function
+
+.. code-block:: cpp
+
+   // the help procedure
+   static void help(std::string errorMessage)
+   {
+    std::cout<<"Program init error : "<<errorMessage<<std::endl;
+    std::cout<<"\nProgram call procedure : retinaDemo [processing mode] [Optional : media target] [Optional LAST parameter: \"log\" to activate retina log sampling]"<<std::endl;
+    std::cout<<"\t[processing mode] :"<<std::endl;
+    std::cout<<"\t -image : for still image processing"<<std::endl;
+    std::cout<<"\t -video : for video stream processing"<<std::endl;
+    std::cout<<"\t[Optional : media target] :"<<std::endl;
+    std::cout<<"\t if processing an image or video file, then, specify the path and filename of the target to process"<<std::endl;
+    std::cout<<"\t leave empty if processing video stream coming from a connected video device"<<std::endl;
+    std::cout<<"\t[Optional : activate retina log sampling] : an optional last parameter can be specified for retina spatial log sampling"<<std::endl;
+    std::cout<<"\t set \"log\" without quotes to activate this sampling, output frame size will be divided by 4"<<std::endl;
+    std::cout<<"\nExamples:"<<std::endl;
+    std::cout<<"\t-Image processing : ./retinaDemo -image lena.jpg"<<std::endl;
+    std::cout<<"\t-Image processing with log sampling : ./retinaDemo -image lena.jpg log"<<std::endl;
+    std::cout<<"\t-Video processing : ./retinaDemo -video myMovie.mp4"<<std::endl;
+    std::cout<<"\t-Live video processing : ./retinaDemo -video"<<std::endl;
+    std::cout<<"\nPlease start again with new parameters"<<std::endl;
+    std::cout<<"****************************************************"<<std::endl;
+    std::cout<<" NOTE : this program generates the default retina parameters file 'RetinaDefaultParameters.xml'"<<std::endl;
+    std::cout<<" => you can use this to fine tune parameters and load them if you save to file 'RetinaSpecificParameters.xml'"<<std::endl;
+   }
+
+Then, start the main program and first declare a *cv::Mat* matrix in which input images will be loaded. Also allocate a *cv::VideoCapture* object ready to load video streams (if necessary)
+
+.. code-block:: cpp
+
+  int main(int argc, char* argv[]) {
+    // declare the retina input buffer... that will be fed differently in regard of the input media
+    cv::Mat inputFrame;
+    cv::VideoCapture videoCapture; // in case a video media is used, its manager is declared here
+
+
+In the main program, before processing, first check input command parameters. Here it loads a first input image coming from a single loaded image (if user chose command *-image*) or from a video stream (if user chose command *-video*). Also, if the user added *log* command at the end of its program call, the spatial logarithmic image sampling performed by the retina is taken into account by the Boolean flag *useLogSampling*.
+
+.. code-block:: cpp
+
+  // welcome message
+    std::cout<<"****************************************************"<<std::endl;
+    std::cout<<"* Retina demonstration : demonstrates the use of is a wrapper class of the Gipsa/Listic Labs retina model."<<std::endl;
+    std::cout<<"* This demo will try to load the file 'RetinaSpecificParameters.xml' (if exists).\nTo create it, copy the autogenerated template 'RetinaDefaultParameters.xml'.\nThen twaek it with your own retina parameters."<<std::endl;
+    // basic input arguments checking
+    if (argc<2)
+    {
+        help("bad number of parameter");
+        return -1;
+    }
+
+    bool useLogSampling = !strcmp(argv[argc-1], "log"); // check if user wants retina log sampling processing
+
+    std::string inputMediaType=argv[1];
+
+    //////////////////////////////////////////////////////////////////////////////
+    // checking input media type (still image, video file, live video acquisition)
+    if (!strcmp(inputMediaType.c_str(), "-image") && argc >= 3)
+    {
+        std::cout<<"RetinaDemo: processing image "<<argv[2]<<std::endl;
+        // image processing case
+        inputFrame = cv::imread(std::string(argv[2]), 1); // load image in RGB mode
+    }else
+        if (!strcmp(inputMediaType.c_str(), "-video"))
+        {
+            if (argc == 2 || (argc == 3 && useLogSampling)) // attempt to grab images from a video capture device
+            {
+                videoCapture.open(0);
+            }else// attempt to grab images from a video filestream
+            {
+                std::cout<<"RetinaDemo: processing video stream "<<argv[2]<<std::endl;
+                videoCapture.open(argv[2]);
+            }
+
+            // grab a first frame to check if everything is ok
+            videoCapture>>inputFrame;
+        }else
+        {
+            // bad command parameter
+            help("bad command parameter");
+            return -1;
+        }
+
+Once all input parameters are processed, a first image should have been loaded, if not, display error and stop program :
+
+.. code-block:: cpp
+
+    if (inputFrame.empty())
+    {
+        help("Input media could not be loaded, aborting");
+        return -1;
+    }
+
+Now, everything is ready to run the retina model. I propose here to allocate a retina instance and to manage the eventual log sampling option. The Retina constructor expects at least a cv::Size object that shows the input data size that will have to be managed. One can activate other options such as color and its related color multiplexing strategy (here Bayer multiplexing is chosen using *enum cv::bioinspired::RETINA_COLOR_BAYER*). If using log sampling, the image reduction factor (smaller output images) and log sampling strengh can be adjusted.
+
+.. code-block:: cpp
+
+    // pointer to a retina object
+    cv::Ptr<Retina> myRetina;
+
+    // if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
+    if (useLogSampling)
+    {
+        myRetina = cv::bioinspired::createRetina(inputFrame.size(), true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+    }
+    else// -> else allocate "classical" retina :
+        myRetina = cv::bioinspired::createRetina(inputFrame.size());
+
+Once done, the proposed code writes a default xml file that contains the default parameters of the retina. This is useful to make your own config using this template. Here generated template xml file is called *RetinaDefaultParameters.xml*.
+
+.. code-block:: cpp
+
+    // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
+    myRetina->write("RetinaDefaultParameters.xml");
+
+In the following line, the retina attempts to load another xml file called *RetinaSpecificParameters.xml*. If you created it and introduced your own setup, it will be loaded, in the other case, default retina parameters are used.
+
+.. code-block:: cpp
+
+    // load parameters if file exists
+    myRetina->setup("RetinaSpecificParameters.xml");
+
+It is not required here but just to show it is possible, you can reset the retina buffers to zero to force it to forget past events.
+
+.. code-block:: cpp
+
+    // reset all retina buffers (imagine you close your eyes for a long time)
+    myRetina->clearBuffers();
+
+Now, it is time to run the retina ! First create some output buffers ready to receive the two retina channels outputs
+
+.. code-block:: cpp
+
+    // declare retina output buffers
+    cv::Mat retinaOutput_parvo;
+    cv::Mat retinaOutput_magno;
+
+Then, run retina in a loop, load new frames from video sequence if necessary and get retina outputs back to dedicated buffers.
+
+.. code-block:: cpp
+
+    // processing loop with no stop condition
+    while(true)
+    {
+        // if using video stream, then, grabbing a new frame, else, input remains the same
+        if (videoCapture.isOpened())
+            videoCapture>>inputFrame;
+
+        // run retina filter on the loaded input frame
+        myRetina->run(inputFrame);
+        // Retrieve and display retina output
+        myRetina->getParvo(retinaOutput_parvo);
+        myRetina->getMagno(retinaOutput_magno);
+        cv::imshow("retina input", inputFrame);
+        cv::imshow("Retina Parvo", retinaOutput_parvo);
+        cv::imshow("Retina Magno", retinaOutput_magno);
+        cv::waitKey(10);
+    }
+
+That's done ! But if you want to secure the system, take care and manage Exceptions. The retina can throw some when it sees irrelevant data (no input frame, wrong setup, etc.).
+Then, i recommend to surround all the retina code by a try/catch system like this :
+
+.. code-block:: cpp
+
+    try{
+         // pointer to a retina object
+         cv::Ptr<cv::Retina> myRetina;
+         [---]
+         // processing loop with no stop condition
+         while(true)
+         {
+             [---]
+         }
+
+    }catch(cv::Exception e)
+    {
+        std::cerr<<"Error using Retina : "<<e.what()<<std::endl;
+    }
+
+Retina parameters, what to do ?
+===============================
+
+First, it is recommended to read the reference paper :
+
+* Benoit A., Caplier A., Durette B., Herault, J., *"Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing"*, Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
+
+Once done open the configuration file *RetinaDefaultParameters.xml* generated by the demo and let's have a look at it.
+
+.. code-block:: cpp
+
+    <?xml version="1.0"?>
+    <opencv_storage>
+    <OPLandIPLparvo>
+        <colorMode>1</colorMode>
+        <normaliseOutput>1</normaliseOutput>
+        <photoreceptorsLocalAdaptationSensitivity>7.5e-01</photoreceptorsLocalAdaptationSensitivity>
+        <photoreceptorsTemporalConstant>9.0e-01</photoreceptorsTemporalConstant>
+        <photoreceptorsSpatialConstant>5.7e-01</photoreceptorsSpatialConstant>
+        <horizontalCellsGain>0.01</horizontalCellsGain>
+        <hcellsTemporalConstant>0.5</hcellsTemporalConstant>
+        <hcellsSpatialConstant>7.</hcellsSpatialConstant>
+        <ganglionCellsSensitivity>7.5e-01</ganglionCellsSensitivity></OPLandIPLparvo>
+    <IPLmagno>
+        <normaliseOutput>1</normaliseOutput>
+        <parasolCells_beta>0.</parasolCells_beta>
+        <parasolCells_tau>0.</parasolCells_tau>
+        <parasolCells_k>7.</parasolCells_k>
+        <amacrinCellsTemporalCutFrequency>2.0e+00</amacrinCellsTemporalCutFrequency>
+        <V0CompressionParameter>9.5e-01</V0CompressionParameter>
+        <localAdaptintegration_tau>0.</localAdaptintegration_tau>
+        <localAdaptintegration_k>7.</localAdaptintegration_k></IPLmagno>
+    </opencv_storage>
+
+Here are some hints but actually, the best parameter setup depends more on what you want to do with the retina rather than the images input that you give to retina. Apart from the more specific case of High Dynamic Range images (HDR) that require more specific setup for specific luminance compression objective, the retina behaviors should be rather stable from content to content. Note that OpenCV is able to manage such HDR format thanks to the OpenEXR images compatibility.
+
+Then, if the application target requires details enhancement prior to specific image processing, you need to know if mean luminance information is required or not. If not, the the retina can cancel or significantly reduce its energy thus giving more visibility to higher spatial frequency details.
+
+
+Basic parameters
+----------------
+
+The most simple parameters are the following :
+
+* **colorMode** : let the retina process color information (if 1) or gray scale images (if 0). In this last case, only the first channel of the input will be processed.
+
+* **normaliseOutput** : each channel has this parameter, if value is 1, then the considered channel output is rescaled between 0 and 255. Take care in this case at the Magnocellular output level (motion/transient channel detection). Residual noise will also be rescaled !
+
+**Note :** using color requires color channels multiplexing/demultipexing which requires more processing. You can expect much faster processing using gray levels : it would require around 30 product per pixel for all the retina processes and it has recently been parallelized for multicore architectures.
+
+Photo-receptors parameters
+--------------------------
+
+The following parameters act on the entry point of the retina - photo-receptors - and impact all the following processes. These sensors are low pass spatio-temporal filters that smooth temporal and spatial data and also adjust there sensitivity to local luminance thus improving details extraction and high frequency noise canceling.
+
+* **photoreceptorsLocalAdaptationSensitivity** between 0 and 1. Values close to 1 allow high luminance log compression effect at the photo-receptors level. Values closer to 0 give a more linear sensitivity. Increased alone, it can burn the *Parvo (details channel)* output image. If adjusted in collaboration with **ganglionCellsSensitivity** images can be very contrasted whatever the local luminance there is... at the price of a naturalness decrease.
+
+* **photoreceptorsTemporalConstant** this setups the temporal constant of the low pass filter effect at the entry of the retina. High value lead to strong temporal smoothing effect : moving objects are blurred and can disappear while static object are favored. But when starting the retina processing, stable state is reached lately.
+
+* **photoreceptorsSpatialConstant** specifies the spatial constant related to photo-receptors low pass filter effect. This parameters specify the minimum allowed spatial signal period allowed in the following. Typically, this filter should cut high frequency noise. Then a 0 value doesn't cut anything noise while higher values start to cut high spatial frequencies and more and more lower frequencies... Then, do not go to high if you wanna see some details of the input images ! A good compromise for color images is 0.53 since this won't affect too much the color spectrum. Higher values would lead to gray and blurred output images.
+
+Horizontal cells parameters
+---------------------------
+
+This parameter set tunes the neural network connected to the photo-receptors, the horizontal cells. It modulates photo-receptors sensitivity and completes the processing for final spectral whitening (part of the spatial band pass effect thus favoring visual details enhancement).
+
+* **horizontalCellsGain** here is a critical parameter ! If you are not interested by the mean luminance and focus on details enhancement, then, set to zero. But if you want to keep some environment luminance data, let some low spatial frequencies pass into the system and set a higher value (<1).
+
+* **hcellsTemporalConstant** similar to photo-receptors, this acts on the temporal constant of a low pass temporal filter that smooths input data. Here, a high value generates a high retina after effect while a lower value makes the retina more reactive. This value should be lower than **photoreceptorsTemporalConstant** to limit strong retina after effects.
+
+* **hcellsSpatialConstant** is the spatial constant of the low pass filter of these cells filter. It specifies the lowest spatial frequency allowed in the following. Visually, a high value leads to very low spatial frequencies processing and leads to salient halo effects. Lower values reduce this effect but the limit is : do not go lower than the value of **photoreceptorsSpatialConstant**. Those 2 parameters actually specify the spatial band-pass of the retina.
+
+**NOTE** after the processing managed by the previous parameters, input data is cleaned from noise and luminance in already partly enhanced. The following parameters act on the last processing stages of the two outing retina signals.
+
+Parvo (details channel) dedicated parameter
+-------------------------------------------
+
+* **ganglionCellsSensitivity** specifies the strength of the final local adaptation occurring at the output of this details dedicated channel. Parameter values remain between 0 and 1. Low value tend to give a linear response while higher values enforces the remaining low contrasted areas.
+
+**Note :** this parameter can correct eventual burned images by favoring low energetic details of the visual scene, even in bright areas.
+
+IPL Magno (motion/transient channel) parameters
+-----------------------------------------------
+
+Once image information is cleaned, this channel acts as a high pass temporal filter that only selects signals related to transient signals (events, motion, etc.). A low pass spatial filter smooths extracted transient data and a final logarithmic compression enhances low transient events thus enhancing event sensitivity.
+
+* **parasolCells_beta** generally set to zero, can be considered as an amplifier gain at the entry point of this processing stage. Generally set to 0.
+
+* **parasolCells_tau** the temporal smoothing effect that can be added
+
+* **parasolCells_k** the spatial constant of the spatial filtering effect, set it at a high value to favor low spatial frequency signals that are lower subject to residual noise.
+
+* **amacrinCellsTemporalCutFrequency** specifies the temporal constant of the high pass filter. High values let slow transient events to be selected.
+
+* **V0CompressionParameter** specifies the strength of the log compression. Similar behaviors to previous description but here it enforces sensitivity of transient events.
+
+* **localAdaptintegration_tau** generally set to 0, no real use here actually
+
+* **localAdaptintegration_k** specifies the size of the area on which local adaptation is performed. Low values lead to short range local adaptation (higher sensitivity to noise), high values secure log compression.
diff --git a/doc/tutorials/contrib/table_of_content_contrib/images/retina_TreeHdr_small.jpg b/doc/tutorials/bioinspired/table_of_content_bioinspired/images/retina_TreeHdr_small.jpg
similarity index 100%
rename from doc/tutorials/contrib/table_of_content_contrib/images/retina_TreeHdr_small.jpg
rename to doc/tutorials/bioinspired/table_of_content_bioinspired/images/retina_TreeHdr_small.jpg
diff --git a/doc/tutorials/bioinspired/table_of_content_bioinspired/table_of_content_bioinspired.rst b/doc/tutorials/bioinspired/table_of_content_bioinspired/table_of_content_bioinspired.rst
new file mode 100644
index 0000000000..88869e98f9
--- /dev/null
+++ b/doc/tutorials/bioinspired/table_of_content_bioinspired/table_of_content_bioinspired.rst
@@ -0,0 +1,36 @@
+.. _Table-Of-Content-Bioinspired:
+
+*bioinspired* module. Algorithms inspired from biological models
+----------------------------------------------------------------
+
+Here you will learn how to use additional modules of OpenCV defined in the "bioinspired" module.
+
+ .. include:: ../../definitions/tocDefinitions.rst
+
++
+  .. tabularcolumns:: m{100pt} m{300pt}
+  .. cssclass:: toctableopencv
+
+  =============== ======================================================
+  |RetinaDemoImg| **Title:** :ref:`Retina_Model`
+
+                  *Compatibility:* > OpenCV 2.4
+
+                  *Author:* |Author_AlexB|
+
+                  You will learn how to process images and video streams with a model of retina filter for details enhancement, spatio-temporal noise removal, luminance correction and spatio-temporal events detection.
+
+  =============== ======================================================
+
+  .. |RetinaDemoImg| image:: images/retina_TreeHdr_small.jpg
+                   :height: 90pt
+                   :width:  90pt
+
+ .. raw:: latex
+
+    \pagebreak
+
+.. toctree::
+   :hidden:
+
+   ../retina_model/retina_model
diff --git a/doc/tutorials/contrib/table_of_content_contrib/table_of_content_contrib.rst b/doc/tutorials/bioinspired/table_of_content_bioinspired/table_of_content_bioinspired.rst~
similarity index 100%
rename from doc/tutorials/contrib/table_of_content_contrib/table_of_content_contrib.rst
rename to doc/tutorials/bioinspired/table_of_content_bioinspired/table_of_content_bioinspired.rst~
diff --git a/doc/tutorials/calib3d/camera_calibration/camera_calibration.rst b/doc/tutorials/calib3d/camera_calibration/camera_calibration.rst
index 77723b2b8b..2cf00f42a9 100644
--- a/doc/tutorials/calib3d/camera_calibration/camera_calibration.rst
+++ b/doc/tutorials/calib3d/camera_calibration/camera_calibration.rst
@@ -3,42 +3,42 @@
 Camera calibration With OpenCV
 ******************************
 
-Cameras have been around for a long-long time. However, with the introduction of the cheap *pinhole* cameras in the late 20th century, they became a common occurrence in our everyday life. Unfortunately, this cheapness comes with its price: significant distortion. Luckily, these are constants and with a calibration and some remapping we can correct this. Furthermore, with calibration you may also determinate the relation between the camera's natural units (pixels) and the real world units (for example millimeters).
+Cameras have been around for a long-long time. However, with the introduction of the cheap *pinhole* cameras in the late 20th century, they became a common occurrence in our everyday life. Unfortunately, this cheapness comes with its price: significant distortion. Luckily, these are constants and with a calibration and some remapping we can correct this. Furthermore, with calibration you may also determine the relation between the camera's natural units (pixels) and the real world units (for example millimeters).
 
 Theory
 ======
 
-For the distortion OpenCV takes into account the radial and tangential factors. For the radial one uses the following formula:
+For the distortion OpenCV takes into account the radial and tangential factors. For the radial factor one uses the following formula:
 
 .. math::
 
    x_{corrected} = x( 1 + k_1 r^2 + k_2 r^4 + k_3 r^6) \\
    y_{corrected} = y( 1 + k_1 r^2 + k_2 r^4 + k_3 r^6)
 
-So for an old pixel point at :math:`(x,y)` coordinate in the input image, for a corrected output image its position will be :math:`(x_{corrected} y_{corrected})` . The presence of the radial distortion manifests in form of the "barrel" or "fish-eye" effect.
+So for an old pixel point at :math:`(x,y)` coordinates in the input image, its position on the corrected output image will be :math:`(x_{corrected} y_{corrected})`. The presence of the radial distortion manifests in form of the "barrel" or "fish-eye" effect.
 
-Tangential distortion occurs because the image taking lenses are not perfectly parallel to the imaging plane. Correcting this is made via the formulas:
+Tangential distortion occurs because the image taking lenses are not perfectly parallel to the imaging plane. It can be corrected via the formulas:
 
 .. math::
 
    x_{corrected} = x + [ 2p_1xy + p_2(r^2+2x^2)] \\
    y_{corrected} = y + [ p_1(r^2+ 2y^2)+ 2p_2xy]
 
-So we have five distortion parameters, which in OpenCV are organized in a 5 column one row matrix:
+So we have five distortion parameters which in OpenCV are presented as one row matrix with 5 columns:
 
 .. math::
 
   Distortion_{coefficients}=(k_1 \hspace{10pt} k_2 \hspace{10pt} p_1 \hspace{10pt} p_2 \hspace{10pt} k_3)
 
-Now for the unit conversion, we use the following formula:
+Now for the unit conversion we use the following formula:
 
 .. math::
 
    \left [  \begin{matrix}   x \\   y \\  w \end{matrix} \right ] = \left [ \begin{matrix}   f_x & 0 & c_x \\  0 & f_y & c_y \\   0 & 0 & 1 \end{matrix} \right ] \left [ \begin{matrix}  X \\  Y \\   Z \end{matrix} \right ]
 
-Here the presence of the :math:`w` is cause we use a homography coordinate system (and :math:`w=Z`). The unknown parameters are :math:`f_x` and :math:`f_y` (camera focal lengths) and :math:`(c_x, c_y)` what are the optical centers expressed in pixels coordinates. If for both axes a common focal length is used with a given :math:`a` aspect ratio (usually 1), then :math:`f_y=f_x*a` and in the upper formula we will have a single :math:`f` focal length. The matrix containing these four parameters is referred to as the *camera matrix*. While the distortion coefficients are the same regardless of the camera resolutions used, these should be scaled along with the current resolution from the calibrated resolution.
+Here the presence of :math:`w` is explained by the use of homography coordinate system (and :math:`w=Z`). The unknown parameters are :math:`f_x` and :math:`f_y` (camera focal lengths) and :math:`(c_x, c_y)` which are the optical centers expressed in pixels coordinates. If for both axes a common focal length is used with a given :math:`a` aspect ratio (usually 1), then :math:`f_y=f_x*a` and in the upper formula we will have a single focal length :math:`f`. The matrix containing these four parameters is referred to as the *camera matrix*. While the distortion coefficients are the same regardless of the camera resolutions used, these should be scaled along with the current resolution from the calibrated resolution.
 
-The process of determining these two matrices is the calibration. Calculating these parameters is done by some basic geometrical equations. The equations used depend on the calibrating objects used. Currently OpenCV supports three types of object for calibration:
+The process of determining these two matrices is the calibration. Calculation of these parameters is done through basic geometrical equations. The equations used depend on the chosen calibrating objects. Currently OpenCV supports three types of objects for calibration:
 
 .. container:: enumeratevisibleitemswithsquare
 
@@ -46,7 +46,7 @@ The process of determining these two matrices is the calibration. Calculating th
    + Symmetrical circle pattern
    + Asymmetrical circle pattern
 
-Basically, you need to take snapshots of these patterns with your camera and let OpenCV find them. Each found pattern equals in a new equation. To solve the equation you need at least a predetermined number of pattern snapshots to form a well-posed equation system. This number is higher for the chessboard pattern and less for the circle ones. For example, in theory the chessboard one requires at least two. However, in practice we have a good amount of noise present in our input images, so for good results you will probably want at least 10 good snapshots of the input pattern in different position.
+Basically, you need to take snapshots of these patterns with your camera and let OpenCV find them. Each found pattern results in a new equation. To solve the equation you need at least a predetermined number of pattern snapshots to form a well-posed equation system. This number is higher for the chessboard pattern and less for the circle ones. For example, in theory the chessboard pattern requires at least two snapshots. However, in practice we have a good amount of noise present in our input images, so for good results you will probably need at least 10 good snapshots of the input pattern in different positions.
 
 Goal
 ====
@@ -55,19 +55,19 @@ The sample application will:
 
 .. container:: enumeratevisibleitemswithsquare
 
-   + Determinate the distortion matrix
-   + Determinate the camera matrix
-   + Input from Camera, Video and Image file list
-   + Configuration from XML/YAML file
+   + Determine the distortion matrix
+   + Determine the camera matrix
+   + Take input from Camera, Video and Image file list
+   + Read configuration from XML/YAML file
    + Save the results into XML/YAML file
    + Calculate re-projection error
 
 Source code
 ===========
 
-You may also find the source code in the :file:`samples/cpp/tutorial_code/calib3d/camera_calibration/` folder of the OpenCV source library or :download:`download it from here <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp>`. The program has a single argument. The name of its configuration file. If none given it will try to open the one named "default.xml". :download:`Here's a sample configuration file <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/in_VID5.xml>` in XML format. In the configuration file you may choose to use as input a camera, a video file or an image list. If you opt for the later one, you need to create a configuration file where you enumerate the images to use. Here's :download:`an example of this <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/VID5.xml>`. The important part to remember is that the images needs to be specified using the absolute path or the relative one from your applications working directory. You may find all this in the beforehand mentioned directory.
+You may also find the source code in the :file:`samples/cpp/tutorial_code/calib3d/camera_calibration/` folder of the OpenCV source library or :download:`download it from here <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp>`. The program has a single argument: the name of its configuration file. If none is given then it will try to open the one named "default.xml". :download:`Here's a sample configuration file <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/in_VID5.xml>` in XML format. In the configuration file you may choose to use camera as an input, a video file or an image list. If you opt for the last one, you will need to create a configuration file where you enumerate the images to use. Here's :download:`an example of this <../../../../samples/cpp/tutorial_code/calib3d/camera_calibration/VID5.xml>`. The important part to remember is that the images need to be specified using the absolute path or the relative one from your application's working directory. You may find all this in the samples directory mentioned above.
 
-The application starts up with reading the settings from the configuration file. Although, this is an important part of it, it has nothing to do with the subject of this tutorial: *camera calibration*. Therefore, I've chosen to do not post here the code part for that. The technical background on how to do this you can find in the :ref:`fileInputOutputXMLYAML` tutorial.
+The application starts up with reading the settings from the configuration file. Although, this is an important part of it, it has nothing to do with the subject of this tutorial: *camera calibration*. Therefore, I've chosen not to post the code for that part here. Technical background on how to do this you can find in the :ref:`fileInputOutputXMLYAML` tutorial.
 
 Explanation
 ===========
@@ -93,9 +93,9 @@ Explanation
             return -1;
       }
 
-   For this I've used simple OpenCV class input operation. After reading the file I've an additional post-process function that checks for the validity of the input. Only if all of them are good will be the *goodInput* variable true.
+   For this I've used simple OpenCV class input operation. After reading the file I've an additional post-processing function that checks validity of the input. Only if all inputs are good then *goodInput* variable will be true.
 
-#. **Get next input, if it fails or we have enough of them calibrate**. After this we have a big loop where we do the following operations: get the next image from the image list, camera or video file. If this fails or we have enough images we run the calibration process. In case of image we step out of the loop and otherwise the remaining frames will be undistorted (if the option is set) via changing from *DETECTION* mode to *CALIBRATED* one.
+#. **Get next input, if it fails or we have enough of them - calibrate**. After this we have a big loop where we do the following operations: get the next image from the image list, camera or video file. If this fails or we have enough images then we run the calibration process. In case of image we step out of the loop and otherwise the remaining frames will be undistorted (if the option is set) via changing from *DETECTION* mode to the *CALIBRATED* one.
 
    .. code-block:: cpp
 
@@ -125,7 +125,7 @@ Explanation
 
    For some cameras we may need to flip the input image. Here we do this too.
 
-#. **Find the pattern in the current input**. The formation of the equations I mentioned above consists of finding the major patterns in the input: in case of the chessboard this is their corners of the squares and for the circles, well, the circles itself. The position of these will form the result and is collected into the *pointBuf* vector.
+#. **Find the pattern in the current input**. The formation of the equations I mentioned above aims to finding major patterns in the input: in case of the chessboard this are corners of the squares and for the circles, well, the circles themselves. The position of these will form the result which will be written into the *pointBuf* vector.
 
    .. code-block:: cpp
 
@@ -146,9 +146,9 @@ Explanation
         break;
       }
 
-   Depending on the type of the input pattern you use either the :calib3d:`findChessboardCorners <findchessboardcorners>` or the :calib3d:`findCirclesGrid <findcirclesgrid>` function. For both of them you pass on the current image, the size of the board and you'll get back the positions of the patterns. Furthermore, they return a boolean variable that states if in the input we could find or not the pattern (we only need to take into account images where this is true!).
+   Depending on the type of the input pattern you use either the :calib3d:`findChessboardCorners <findchessboardcorners>` or the :calib3d:`findCirclesGrid <findcirclesgrid>` function. For both of them you pass the current image and the size of the board and you'll get the positions of the patterns. Furthermore, they return a boolean variable which states if the pattern was found in the input (we only need to take into account those images where this is true!).
 
-   Then again in case of cameras we only take camera images after an input delay time passed. This is in order to allow for the user to move the chessboard around and as getting different images. Same images mean same equations, and same equations at the calibration will form an ill-posed problem, so the calibration will fail. For square images the position of the corners are only approximate. We may improve this by calling the :feature2d:`cornerSubPix <cornersubpix>` function. This way will get a better calibration result. After this we add a valid inputs result to the *imagePoints* vector to collect all of the equations into a single container. Finally, for visualization feedback purposes we will draw the found points on the input image with the :calib3d:`findChessboardCorners <drawchessboardcorners>` function.
+   Then again in case of cameras we only take camera images when an input delay time is passed. This is done in order to allow user moving the chessboard around and getting different images. Similar images result in similar equations, and similar equations at the calibration step will form an ill-posed problem, so the calibration will fail. For square images the positions of the corners are only approximate. We may improve this by calling the :feature2d:`cornerSubPix <cornersubpix>` function. It will produce better calibration result. After this we add a valid inputs result to the *imagePoints* vector to collect all of the equations into a single container. Finally, for visualization feedback purposes we will draw the found points on the input image using :calib3d:`findChessboardCorners <drawchessboardcorners>` function.
 
    .. code-block:: cpp
 
@@ -175,7 +175,7 @@ Explanation
               drawChessboardCorners( view, s.boardSize, Mat(pointBuf), found );
         }
 
-#. **Show state and result for the user, plus command line control of the application**. The showing part consists of a text output on the live feed, and for video or camera input to show the "capturing" frame we simply bitwise negate the input image.
+#. **Show state and result to the user, plus command line control of the application**. This  part shows text output on the image.
 
    .. code-block:: cpp
 
@@ -199,7 +199,7 @@ Explanation
       if( blinkOutput )
          bitwise_not(view, view);
 
-   If we only ran the calibration and got the camera matrix plus the distortion coefficients we may just as correct the image with the :imgproc_geometric:`undistort <undistort>` function:
+   If we ran calibration and got camera's matrix with the distortion coefficients we may want to correct the image using :imgproc_geometric:`undistort <undistort>` function:
 
    .. code-block:: cpp
 
@@ -212,7 +212,7 @@ Explanation
       //------------------------------ Show image and check for input commands -------------------
       imshow("Image View", view);
 
-   Then we wait for an input key and if this is *u* we toggle the distortion removal, if it is *g* we start all over the detection process (or simply start it), and finally for the *ESC* key quit the application:
+   Then we wait for an input key and if this is *u* we toggle the distortion removal, if it is *g* we start again the detection process, and finally for the *ESC* key we quit the application:
 
    .. code-block:: cpp
 
@@ -229,7 +229,7 @@ Explanation
         imagePoints.clear();
       }
 
-#. **Show the distortion removal for the images too**. When you work with an image list it is not possible to remove the distortion inside the loop. Therefore, you must append this after the loop. Taking advantage of this now I'll expand the :imgproc_geometric:`undistort <undistort>` function, which is in fact first a call of the :imgproc_geometric:`initUndistortRectifyMap <initundistortrectifymap>` to find out the transformation matrices and then doing the transformation with the :imgproc_geometric:`remap <remap>` function. Because, after a successful calibration the map calculation needs to be done only once, by using this expanded form you may speed up your application:
+#. **Show the distortion removal for the images too**. When you work with an image list it is not possible to remove the distortion inside the loop. Therefore, you must do this after the loop. Taking advantage of this now I'll expand the :imgproc_geometric:`undistort <undistort>` function, which is in fact first calls :imgproc_geometric:`initUndistortRectifyMap <initundistortrectifymap>` to find transformation matrices and then performs transformation using :imgproc_geometric:`remap <remap>` function. Because, after successful calibration map calculation needs to be done only once, by using this expanded form you may speed up your application:
 
    .. code-block:: cpp
 
@@ -256,7 +256,7 @@ Explanation
 The calibration and save
 ========================
 
-Because the calibration needs to be only once per camera it makes sense to save them after a successful calibration. This way later on you can just load these values into your program. Due to this we first make the calibration, and if it succeeds we save the result into an OpenCV style XML or YAML file, depending on the extension you give in the configuration file.
+Because the calibration needs to be done only once per camera, it makes sense to save it after a successful calibration. This way later on you can just load these values into your program. Due to this we first make the calibration, and if it succeeds we save the result into an OpenCV style XML or YAML file, depending on the extension you give in the configuration file.
 
 Therefore in the first function we just split up these two processes. Because we want to save many of the calibration variables we'll create these variables here and pass on both of them to the calibration and saving function. Again, I'll not show the saving part as that has little in common with the calibration. Explore the source file in order to find out how and what:
 
@@ -280,7 +280,7 @@ Therefore in the first function we just split up these two processes. Because we
     return ok;
    }
 
-We do the calibration with the help of the :calib3d:`calibrateCamera <calibratecamera>` function. This has the following parameters:
+We do the calibration with the help of the :calib3d:`calibrateCamera <calibratecamera>` function. It has the following parameters:
 
 .. container:: enumeratevisibleitemswithsquare
 
@@ -318,11 +318,11 @@ We do the calibration with the help of the :calib3d:`calibrateCamera <calibratec
         calcBoardCornerPositions(s.boardSize, s.squareSize, objectPoints[0], s.calibrationPattern);
         objectPoints.resize(imagePoints.size(),objectPoints[0]);
 
-   + The image points. This is a vector of *Point2f* vector that for each input image contains where the important points (corners for chessboard, and center of circles for the circle patterns) were found. We already collected this from what the :calib3d:`findChessboardCorners <findchessboardcorners>` or the :calib3d:`findCirclesGrid <findcirclesgrid>` function returned. We just need to pass it on.
+   + The image points. This is a vector of *Point2f* vector which for each input image contains coordinates of the important points (corners for chessboard and centers of the circles for the circle pattern). We have already collected this from :calib3d:`findChessboardCorners <findchessboardcorners>` or :calib3d:`findCirclesGrid <findcirclesgrid>` function. We just need to pass it on.
 
    + The size of the image acquired from the camera, video file or the images.
 
-   + The camera matrix. If we used the fix aspect ratio option we need to set the :math:`f_x` to zero:
+   + The camera matrix. If we used the fixed aspect ratio option we need to set the :math:`f_x` to zero:
 
      .. code-block:: cpp
 
@@ -336,16 +336,16 @@ We do the calibration with the help of the :calib3d:`calibrateCamera <calibratec
 
         distCoeffs = Mat::zeros(8, 1, CV_64F);
 
-   + The function will calculate for all the views the rotation and translation vector that transform the object points (given in the model coordinate space) to the image points (given in the world coordinate space). The 7th and 8th parameters are an output vector of matrices containing in the ith position the rotation and translation vector for the ith object point to the ith image point.
+   + For all the views the function will calculate rotation and translation vectors which transform the object points (given in the model coordinate space) to the image points (given in the world coordinate space). The 7-th and 8-th parameters are the output vector of matrices containing in the i-th position the rotation and translation vector for the i-th object point to the i-th image point.
 
-   + The final argument is a flag. You need to specify here options like fix the aspect ratio for the focal length, assume zero tangential distortion or to fix the principal point.
+   + The final argument is the flag. You need to specify here options like fix the aspect ratio for the focal length, assume zero tangential distortion or to fix the principal point.
 
    .. code-block:: cpp
 
      double rms = calibrateCamera(objectPoints, imagePoints, imageSize, cameraMatrix,
                                  distCoeffs, rvecs, tvecs, s.flag|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5);
 
-   + The function returns the average re-projection error. This number gives a good estimation of just how exact is the found parameters. This should be as close to zero as possible. Given the intrinsic, distortion, rotation and translation matrices we may calculate the error for one view by using the :calib3d:`projectPoints <projectpoints>` to first transform the object point to image point. Then we calculate the absolute norm between what we got with our transformation and the corner/circle finding algorithm. To find the average error we calculate the arithmetical mean of the errors calculate for all the calibration images.
+   + The function returns the average re-projection error. This number gives a good estimation of precision of the found parameters. This should be as close to zero as possible. Given the intrinsic, distortion, rotation and translation matrices we may calculate the error for one view by using the :calib3d:`projectPoints <projectpoints>` to first transform the object point to image point. Then we calculate the absolute norm between what we got with our transformation and the corner/circle finding algorithm. To find the average error we calculate the arithmetical mean of the errors calculated for all the calibration images.
 
      .. code-block:: cpp
 
@@ -378,25 +378,25 @@ We do the calibration with the help of the :calib3d:`calibrateCamera <calibratec
 Results
 =======
 
-Let there be :download:`this input chessboard pattern <../../../pattern.png>` that has a size of 9 X 6. I've used an AXIS IP camera to create a couple of snapshots of the board and saved it into a VID5 directory. I've put this inside the :file:`images/CameraCalibraation` folder of my working directory and created the following :file:`VID5.XML` file that describes which images to use:
+Let there be :download:`this input chessboard pattern <../../../pattern.png>` which has a size of 9 X 6. I've used an AXIS IP camera to create a couple of snapshots of the board and saved it into VID5 directory. I've put this inside the :file:`images/CameraCalibration` folder of my working directory and created the following :file:`VID5.XML` file that describes which images to use:
 
 .. code-block:: xml
 
    <?xml version="1.0"?>
    <opencv_storage>
    <images>
-   images/CameraCalibraation/VID5/xx1.jpg
-   images/CameraCalibraation/VID5/xx2.jpg
-   images/CameraCalibraation/VID5/xx3.jpg
-   images/CameraCalibraation/VID5/xx4.jpg
-   images/CameraCalibraation/VID5/xx5.jpg
-   images/CameraCalibraation/VID5/xx6.jpg
-   images/CameraCalibraation/VID5/xx7.jpg
-   images/CameraCalibraation/VID5/xx8.jpg
+   images/CameraCalibration/VID5/xx1.jpg
+   images/CameraCalibration/VID5/xx2.jpg
+   images/CameraCalibration/VID5/xx3.jpg
+   images/CameraCalibration/VID5/xx4.jpg
+   images/CameraCalibration/VID5/xx5.jpg
+   images/CameraCalibration/VID5/xx6.jpg
+   images/CameraCalibration/VID5/xx7.jpg
+   images/CameraCalibration/VID5/xx8.jpg
    </images>
    </opencv_storage>
 
-Then specified the :file:`images/CameraCalibraation/VID5/VID5.XML` as input in the configuration file. Here's a chessboard pattern found during the runtime of the application:
+Then passed :file:`images/CameraCalibration/VID5/VID5.XML` as an input in the configuration file. Here's a chessboard pattern found during the runtime of the application:
 
 .. image:: images/fileListImage.jpg
    :alt: A found chessboard
@@ -433,7 +433,7 @@ In both cases in the specified output XML/YAML file you'll find the camera and d
     -4.1802327176423804e-001 5.0715244063187526e-001 0. 0.
     -5.7843597214487474e-001</data></Distortion_Coefficients>
 
-Add these values as constants to your program, call the :imgproc_geometric:`initUndistortRectifyMap <initundistortrectifymap>` and the :imgproc_geometric:`remap <remap>` function to remove distortion and enjoy distortion free inputs with cheap and low quality cameras.
+Add these values as constants to your program, call the :imgproc_geometric:`initUndistortRectifyMap <initundistortrectifymap>` and the :imgproc_geometric:`remap <remap>` function to remove distortion and enjoy distortion free inputs for cheap and low quality cameras.
 
 You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=ViPN810E0SU>`_.
 
diff --git a/doc/tutorials/calib3d/camera_calibration_square_chess/camera_calibration_square_chess.rst b/doc/tutorials/calib3d/camera_calibration_square_chess/camera_calibration_square_chess.rst
index 4eff2640ad..e08550750d 100644
--- a/doc/tutorials/calib3d/camera_calibration_square_chess/camera_calibration_square_chess.rst
+++ b/doc/tutorials/calib3d/camera_calibration_square_chess/camera_calibration_square_chess.rst
@@ -59,4 +59,4 @@ Now, let us write a code that detects a chessboard in a new image and finds its
 #.
     Calculate reprojection error like it is done in ``calibration`` sample (see ``opencv/samples/cpp/calibration.cpp``, function ``computeReprojectionErrors``).
 
-Question: how to calculate the distance from the camera origin to any of the corners?
\ No newline at end of file
+Question: how to calculate the distance from the camera origin to any of the corners?
diff --git a/doc/tutorials/core/adding_images/adding_images.rst b/doc/tutorials/core/adding_images/adding_images.rst
index e3135693de..3559132dc2 100644
--- a/doc/tutorials/core/adding_images/adding_images.rst
+++ b/doc/tutorials/core/adding_images/adding_images.rst
@@ -35,8 +35,7 @@ As usual, after the not-so-lengthy explanation, let's go to the code:
 
 .. code-block:: cpp
 
-   #include <cv.h>
-   #include <highgui.h>
+   #include <opencv2/opencv.hpp>
    #include <iostream>
 
    using namespace cv;
diff --git a/doc/tutorials/core/basic_linear_transform/basic_linear_transform.rst b/doc/tutorials/core/basic_linear_transform/basic_linear_transform.rst
index 4c2fb70358..75c2628683 100644
--- a/doc/tutorials/core/basic_linear_transform/basic_linear_transform.rst
+++ b/doc/tutorials/core/basic_linear_transform/basic_linear_transform.rst
@@ -72,8 +72,7 @@ Code
 
 .. code-block:: cpp
 
-   #include <cv.h>
-   #include <highgui.h>
+   #include <opencv2/opencv.hpp>
    #include <iostream>
 
    using namespace cv;
diff --git a/doc/tutorials/core/file_input_output_with_xml_yml/file_input_output_with_xml_yml.rst b/doc/tutorials/core/file_input_output_with_xml_yml/file_input_output_with_xml_yml.rst
index 87166b7cc3..42f6a60915 100644
--- a/doc/tutorials/core/file_input_output_with_xml_yml/file_input_output_with_xml_yml.rst
+++ b/doc/tutorials/core/file_input_output_with_xml_yml/file_input_output_with_xml_yml.rst
@@ -277,4 +277,3 @@ You may observe a runtime instance of this on the `YouTube here <https://www.you
   <div align="center">
   <iframe title="File Input and Output using XML and YAML files in OpenCV" width="560" height="349" src="http://www.youtube.com/embed/A4yqVnByMMM?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
   </div>
-
diff --git a/doc/tutorials/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.rst b/doc/tutorials/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.rst
index 99d6692743..9285509b07 100644
--- a/doc/tutorials/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.rst
+++ b/doc/tutorials/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.rst
@@ -127,6 +127,3 @@ You may observe a runtime instance of this on the `YouTube here <https://www.you
   <div align="center">
   <iframe title="Interoperability with OpenCV 1" width="560" height="349" src="http://www.youtube.com/embed/qckm-zvo31w?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
   </div>
-
-
-
diff --git a/doc/tutorials/core/mat_the_basic_image_container/mat_the_basic_image_container.rst b/doc/tutorials/core/mat_the_basic_image_container/mat_the_basic_image_container.rst
index 03d82bbd46..76c5a4541c 100644
--- a/doc/tutorials/core/mat_the_basic_image_container/mat_the_basic_image_container.rst
+++ b/doc/tutorials/core/mat_the_basic_image_container/mat_the_basic_image_container.rst
@@ -143,7 +143,7 @@ Although *Mat* works really well as an image container, it is also a general mat
 
     You cannot initialize the matrix values with this construction. It will only reallocate its matrix data memory if the new size will not fit into the old one.
 
-   + MATLAB style initializer: :basicstructures:`zeros() <mat-zeros>`, :basicstructures:`ones() <mat-ones>`, ::basicstructures:`eyes() <mat-eye>`. Specify size and data type to use:
+   + MATLAB style initializer: :basicstructures:`zeros() <mat-zeros>`, :basicstructures:`ones() <mat-ones>`, :basicstructures:`eye() <mat-eye>`. Specify size and data type to use:
 
      .. literalinclude:: ../../../../samples/cpp/tutorial_code/core/mat_the_basic_image_container/mat_the_basic_image_container.cpp
         :language: cpp
diff --git a/doc/tutorials/core/table_of_content_core/table_of_content_core.rst b/doc/tutorials/core/table_of_content_core/table_of_content_core.rst
index b50d97635f..4150265d10 100644
--- a/doc/tutorials/core/table_of_content_core/table_of_content_core.rst
+++ b/doc/tutorials/core/table_of_content_core/table_of_content_core.rst
@@ -218,4 +218,4 @@ Here you will learn the about the basic building blocks of the library. A must r
    ../random_generator_and_text/random_generator_and_text
    ../discrete_fourier_transform/discrete_fourier_transform
    ../file_input_output_with_xml_yml/file_input_output_with_xml_yml
-   ../interoperability_with_OpenCV_1/interoperability_with_OpenCV_1
\ No newline at end of file
+   ../interoperability_with_OpenCV_1/interoperability_with_OpenCV_1
diff --git a/doc/tutorials/definitions/README.txt b/doc/tutorials/definitions/README.txt
index 829e27cacb..a598a95dd7 100644
--- a/doc/tutorials/definitions/README.txt
+++ b/doc/tutorials/definitions/README.txt
@@ -1 +1 @@
-Include in this directory only defintion files. None of the reST files entered here will be parsed by the Sphinx Builder.
\ No newline at end of file
+Include in this directory only defintion files. None of the reST files entered here will be parsed by the Sphinx Builder.
diff --git a/doc/tutorials/definitions/noContent.rst b/doc/tutorials/definitions/noContent.rst
index ded9e78fa1..c2780c266d 100644
--- a/doc/tutorials/definitions/noContent.rst
+++ b/doc/tutorials/definitions/noContent.rst
@@ -1,3 +1,3 @@
 
 .. note::
-   Unfortunetly we have no tutorials into this section. Nevertheless, our tutorial writting team is working on it. If you have a tutorial suggestion or you have writen yourself a tutorial (or coded a sample code) that you would like to see here please contact us via our :opencv_group:`user group <>`.
\ No newline at end of file
+   Unfortunetly we have no tutorials into this section. And you can help us with that, since OpenCV is a community effort. If you have a tutorial suggestion or you have written a tutorial yourself (or coded a sample code) that you would like to see here, please contact follow these instructions: :ref:`howToWriteTutorial` and :how_to_contribute:`How to contribute <>`.
diff --git a/doc/tutorials/features2d/feature_description/feature_description.rst b/doc/tutorials/features2d/feature_description/feature_description.rst
index aa1a4a88be..b6e2dfa2bd 100644
--- a/doc/tutorials/features2d/feature_description/feature_description.rst
+++ b/doc/tutorials/features2d/feature_description/feature_description.rst
@@ -13,7 +13,7 @@ In this tutorial you will learn how to:
    * Use the :descriptor_extractor:`DescriptorExtractor<>` interface in order to find the feature vector correspondent to the keypoints. Specifically:
 
      * Use :surf_descriptor_extractor:`SurfDescriptorExtractor<>` and its function :descriptor_extractor:`compute<>` to perform the required calculations.
-     * Use a :brute_force_matcher:`BruteForceMatcher<>`	to match the features vector
+     * Use a :brute_force_matcher:`BFMatcher<>`	to match the features vector
      * Use the function :draw_matches:`drawMatches<>` to draw the detected matches.
 
 
@@ -69,7 +69,7 @@ This tutorial code's is shown lines below. You can also download it from `here <
      extractor.compute( img_2, keypoints_2, descriptors_2 );
 
      //-- Step 3: Matching descriptor vectors with a brute force matcher
-     BruteForceMatcher< L2<float> > matcher;
+     BFMatcher matcher(NORM_L2);
      std::vector< DMatch > matches;
      matcher.match( descriptors_1, descriptors_2, matches );
 
@@ -100,6 +100,3 @@ Result
    .. image:: images/Feature_Description_BruteForce_Result.jpg
       :align: center
       :height: 200pt
-
-
-
diff --git a/doc/tutorials/features2d/feature_detection/feature_detection.rst b/doc/tutorials/features2d/feature_detection/feature_detection.rst
index 1c9ca7cf82..02da6d0806 100644
--- a/doc/tutorials/features2d/feature_detection/feature_detection.rst
+++ b/doc/tutorials/features2d/feature_detection/feature_detection.rst
@@ -31,6 +31,7 @@ This tutorial code's is shown lines below. You can also download it from `here <
    #include "opencv2/core.hpp"
    #include "opencv2/features2d.hpp"
    #include "opencv2/highgui.hpp"
+   #include "opencv2/nonfree.hpp"
 
    using namespace cv;
 
@@ -94,4 +95,3 @@ Result
    .. image:: images/Feature_Detection_Result_b.jpg
       :align: center
       :height: 200pt
-
diff --git a/doc/tutorials/features2d/feature_flann_matcher/feature_flann_matcher.rst b/doc/tutorials/features2d/feature_flann_matcher/feature_flann_matcher.rst
index 9309b05c1d..9839c89841 100644
--- a/doc/tutorials/features2d/feature_flann_matcher/feature_flann_matcher.rst
+++ b/doc/tutorials/features2d/feature_flann_matcher/feature_flann_matcher.rst
@@ -28,6 +28,7 @@ This tutorial code's is shown lines below. You can also download it from `here <
    #include "opencv2/core.hpp"
    #include "opencv2/features2d.hpp"
    #include "opencv2/highgui.hpp"
+   #include "opencv2/nonfree.hpp"
 
    using namespace cv;
 
@@ -85,7 +86,7 @@ This tutorial code's is shown lines below. You can also download it from `here <
      std::vector< DMatch > good_matches;
 
      for( int i = 0; i < descriptors_1.rows; i++ )
-     { if( matches[i].distance < 2*min_dist )
+     { if( matches[i].distance <= 2*min_dist )
        { good_matches.push_back( matches[i]); }
      }
 
@@ -127,6 +128,3 @@ Result
    .. image:: images/Feature_FlannMatcher_Keypoints_Result.jpg
       :align: center
       :height: 250pt
-
-
-
diff --git a/doc/tutorials/features2d/feature_homography/feature_homography.rst b/doc/tutorials/features2d/feature_homography/feature_homography.rst
index 0d78229599..3040ed7d75 100644
--- a/doc/tutorials/features2d/feature_homography/feature_homography.rst
+++ b/doc/tutorials/features2d/feature_homography/feature_homography.rst
@@ -30,6 +30,7 @@ This tutorial code's is shown lines below. You can also download it from `here <
    #include "opencv2/features2d.hpp"
    #include "opencv2/highgui.hpp"
    #include "opencv2/calib3d.hpp"
+   #include "opencv2/nonfree.hpp"
 
    using namespace cv;
 
@@ -145,4 +146,3 @@ Result
    .. image:: images/Feature_Homography_Result.jpg
       :align: center
       :height: 200pt
-
diff --git a/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.rst b/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.rst
index cc90082b85..f4107804bb 100644
--- a/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.rst
+++ b/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.rst
@@ -201,4 +201,3 @@ Learn about how to use the feature points  detectors, descriptors and matching f
    ../feature_flann_matcher/feature_flann_matcher
    ../feature_homography/feature_homography
    ../detection_of_planar_objects/detection_of_planar_objects
-
diff --git a/doc/tutorials/features2d/trackingmotion/corner_subpixeles/corner_subpixeles.rst b/doc/tutorials/features2d/trackingmotion/corner_subpixeles/corner_subpixeles.rst
index a267b13803..4e9015cb68 100644
--- a/doc/tutorials/features2d/trackingmotion/corner_subpixeles/corner_subpixeles.rst
+++ b/doc/tutorials/features2d/trackingmotion/corner_subpixeles/corner_subpixeles.rst
@@ -135,4 +135,3 @@ Here is the result:
 
 .. image:: images/Corner_Subpixeles_Result.jpg
               :align: center
-
diff --git a/doc/tutorials/features2d/trackingmotion/generic_corner_detector/generic_corner_detector.rst b/doc/tutorials/features2d/trackingmotion/generic_corner_detector/generic_corner_detector.rst
index 465ff216cb..d33bf3df6c 100644
--- a/doc/tutorials/features2d/trackingmotion/generic_corner_detector/generic_corner_detector.rst
+++ b/doc/tutorials/features2d/trackingmotion/generic_corner_detector/generic_corner_detector.rst
@@ -37,4 +37,3 @@ Result
 
 .. image:: images/My_Shi_Tomasi_corner_detector_Result.jpg
               :align: center
-
diff --git a/doc/tutorials/features2d/trackingmotion/good_features_to_track/good_features_to_track.rst b/doc/tutorials/features2d/trackingmotion/good_features_to_track/good_features_to_track.rst
index 9f71e987f2..1495befc75 100644
--- a/doc/tutorials/features2d/trackingmotion/good_features_to_track/good_features_to_track.rst
+++ b/doc/tutorials/features2d/trackingmotion/good_features_to_track/good_features_to_track.rst
@@ -118,5 +118,3 @@ Result
 
 .. image:: images/Feature_Detection_Result_a.jpg
               :align: center
-
-
diff --git a/doc/tutorials/features2d/trackingmotion/harris_detector/harris_detector.rst b/doc/tutorials/features2d/trackingmotion/harris_detector/harris_detector.rst
index 0f6bb33fa8..e0e4ea47e4 100644
--- a/doc/tutorials/features2d/trackingmotion/harris_detector/harris_detector.rst
+++ b/doc/tutorials/features2d/trackingmotion/harris_detector/harris_detector.rst
@@ -243,5 +243,3 @@ The detected corners are surrounded by a small black circle
 
 .. image:: images/Harris_Detector_Result.jpg
               :align: center
-
-
diff --git a/doc/tutorials/general/table_of_content_general/table_of_content_general.rst b/doc/tutorials/general/table_of_content_general/table_of_content_general.rst
index 2a6d57ffb9..6e127a98f6 100644
--- a/doc/tutorials/general/table_of_content_general/table_of_content_general.rst
+++ b/doc/tutorials/general/table_of_content_general/table_of_content_general.rst
@@ -10,4 +10,3 @@ These tutorials are the bottom of the iceberg as they link together multiple of
 .. raw:: latex
 
    \pagebreak
-
diff --git a/doc/tutorials/gpu/gpu-basics-similarity/gpu-basics-similarity.rst b/doc/tutorials/gpu/gpu-basics-similarity/gpu-basics-similarity.rst
index 86400bfae4..2e8fe9559b 100644
--- a/doc/tutorials/gpu/gpu-basics-similarity/gpu-basics-similarity.rst
+++ b/doc/tutorials/gpu/gpu-basics-similarity/gpu-basics-similarity.rst
@@ -1 +1,234 @@
-.. _gpuBasicsSimilarity:

Similarity check (PNSR and SSIM) on the GPU
*******************************************

Goal
====

In the :ref:`videoInputPSNRMSSIM` tutorial I already presented the PSNR and SSIM methods for
checking the similarity between the two images. And as you could see there performing these takes
quite some time, especially in the case of the SSIM. However, if the performance numbers of an
OpenCV implementation for the CPU do not satisfy you and you happen to have an NVidia CUDA GPU
device in your system all is not lost. You may try to port or write your algorithm for the video
card.

This tutorial will give a good grasp on how to approach coding by using the GPU module of OpenCV. As
a prerequisite you should already know how to handle the core, highgui and imgproc modules. So, our
goals are:

.. container:: enumeratevisibleitemswithsquare

   + What's different compared to the CPU?
   + Create the GPU code for the PSNR and SSIM
   + Optimize the code for maximal performance

The source code
===============

You may also find the source code and these video file in the
:file:`samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity` folder of the
OpenCV source library or :download:`download it from here
<../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp>`. The
full source code is quite long (due to the controlling of the application via the command line
arguments and performance measurement). Therefore, to avoid cluttering up these sections with those
you'll find here only the functions itself.

The PSNR returns a float number, that if the two inputs are similar between 30 and 50 (higher is
better).

.. literalinclude:: ../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
   :language: cpp
   :linenos:
   :tab-width: 4
   :lines: 165-210, 18-23, 210-235

The SSIM returns the MSSIM of the images. This is too a float number between zero and one (higher is
better), however we have one for each channel. Therefore, we return a *Scalar* OpenCV data
structure:

.. literalinclude:: ../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
   :language: cpp
   :linenos:
   :tab-width: 4
   :lines: 235-355, 26-42, 357-

How to do it? - The GPU
=======================

Now as you can see we have three types of functions for each operation. One for the CPU and two for
the GPU. The reason I made two for the GPU is too illustrate that often simple porting your CPU to
GPU will actually make it slower. If you want some performance gain you will need to remember a few
rules, whose I'm going to detail later on.

The development of the GPU module was made so that it resembles as much as possible its CPU
counterpart. This is to make porting easy. The first thing you need to do before writing any code is
to link the GPU module to your project, and include the header file for the module. All the
functions and data structures of the GPU are in a *gpu* sub namespace of the *cv* namespace. You may
add this to the default one via the *use namespace* keyword, or mark it everywhere explicitly via
the cv:: to avoid confusion. I'll do the later.

.. code-block:: cpp

   #include <opencv2/gpu.hpp>        // GPU structures and methods

GPU stands for **g**\ raphics **p**\ rocessing **u**\ nit. It was originally build to render
graphical scenes. These scenes somehow build on a lot of data. Nevertheless, these aren't all
dependent one from another in a sequential way and as it is possible a parallel processing of them.
Due to this a GPU will contain multiple smaller processing units. These aren't the state of the art
processors and on a one on one test with a CPU it will fall behind. However, its strength lies in
its numbers. In the last years there has been an increasing trend to harvest these massive parallel
powers of the GPU in non-graphical scene rendering too. This gave birth to the general-purpose
computation on graphics processing units (GPGPU).

The GPU has its own memory. When you read data from the hard drive with OpenCV into a *Mat* object
that takes place in your systems memory. The CPU works somehow directly on this (via its cache),
however the GPU cannot. He has too transferred the information he will use for calculations from the
system memory to its own. This is done via an upload process and takes time. In the end the result
will have to be downloaded back to your system memory for your CPU to see it and use it. Porting
small functions to GPU is not recommended as the upload/download time will be larger than the amount
you gain by a parallel execution.

Mat objects are stored only in the system memory (or the CPU cache). For getting an OpenCV matrix
to the GPU you'll need to use its GPU counterpart :gpudatastructure:`GpuMat <gpu-gpumat>`. It works
similar to the Mat with a 2D only limitation and no reference returning for its functions (cannot
mix GPU references with CPU ones). To upload a Mat object to the GPU you need to call the upload
function after creating an instance of the class. To download you may use simple assignment to a
Mat object or use the download function.

.. code-block:: cpp

   Mat I1;         // Main memory item - read image into with imread for example
   gpu::GpuMat gI; // GPU matrix - for now empty
   gI1.upload(I1); // Upload a data from the system memory to the GPU memory

   I1 = gI1;       // Download, gI1.download(I1) will work too

Once you have your data up in the GPU memory you may call GPU enabled functions of OpenCV. Most of
the functions keep the same name just as on the CPU, with the difference that they only accept
*GpuMat* inputs. A full list of these you will find in the documentation: `online here
<http://opencv.itseez.com/modules/gpu/doc/gpu.html>`_ or the OpenCV reference manual that comes with
the source code.

Another thing to keep in mind is that not for all channel numbers you can make efficient algorithms
on the GPU. Generally, I found that the input images for the GPU images need to be either one or
four channel ones and one of the char or float type for the item sizes. No double support on the
GPU, sorry. Passing other types of objects for some functions will result in an exception thrown,
and an error message on the error output. The documentation details in most of the places the types
accepted for the inputs. If you have three channel images as an input you can do two things: either
adds a new channel (and use char elements) or split up the image and call the function for each
image. The first one isn't really recommended as you waste memory.

For some functions, where the position of the elements (neighbor items) doesn't matter quick
solution is to just reshape it into a single channel image. This is the case for the PSNR
implementation where for the *absdiff* method the value of the neighbors is not important. However,
for the *GaussianBlur* this isn't an option and such need to use the split method for the SSIM. With
this knowledge you can already make a GPU viable code (like mine GPU one) and run it. You'll be
surprised to see that it might turn out slower than your CPU implementation.

Optimization
============

The reason for this is that you're throwing out on the window the price for memory allocation and
data transfer. And on the GPU this is damn high. Another possibility for optimization is to
introduce asynchronous OpenCV GPU calls too with the help of the
:gpudatastructure:`gpu::Stream<gpu-stream>`.

1. Memory allocation on the GPU is considerable. Therefore, if it’s possible allocate new memory as
   few times as possible. If you create a function what you intend to call multiple times it is a
   good idea to allocate any local parameters for the function only once, during the first call.
   To do this you create a data structure containing all the local variables you will use. For
   instance in case of the PSNR these are:

   .. code-block:: cpp

      struct BufferPSNR                                     // Optimized GPU versions
        {   // Data allocations are very expensive on GPU. Use a buffer to solve: allocate once reuse later.
        gpu::GpuMat gI1, gI2, gs, t1,t2;

        gpu::GpuMat buf;
      };

   Then create an instance of this in the main program:

   .. code-block:: cpp

      BufferPSNR bufferPSNR;

   And finally pass this to the function each time you call it:

   .. code-block:: cpp

      double getPSNR_GPU_optimized(const Mat& I1, const Mat& I2, BufferPSNR& b)

   Now you access these local parameters as: *b.gI1*, *b.buf* and so on. The GpuMat will only
   reallocate itself on a new call if the new matrix size is different from the previous one.

#. Avoid unnecessary function data transfers. Any small data transfer will be significant one once
   you go to the GPU. Therefore, if possible make all calculations in-place (in other words do not
   create new memory objects - for reasons explained at the previous point). For example, although
   expressing arithmetical operations may be easier to express in one line formulas, it will be
   slower. In case of the SSIM at one point I need to calculate:

   .. code-block:: cpp

      b.t1 = 2 * b.mu1_mu2 + C1;

   Although the upper call will succeed observe that there is a hidden data transfer present. Before
   it makes the addition it needs to store somewhere the multiplication. Therefore, it will create a
   local matrix in the background, add to that the *C1* value and finally assign that to *t1*. To
   avoid this we use the gpu functions, instead of the arithmetic operators:

   .. code-block:: cpp

      gpu::multiply(b.mu1_mu2, 2, b.t1); //b.t1 = 2 * b.mu1_mu2 + C1;
      gpu::add(b.t1, C1, b.t1);

#. Use asynchronous calls (the :gpudatastructure:`gpu::Stream <gpu-stream>`). By default whenever
   you call a gpu function it will wait for the call to finish and return with the result
   afterwards. However, it is possible to make asynchronous calls, meaning it will call for the
   operation execution, make the costly data allocations for the algorithm and return back right
   away. Now you can call another function if you wish to do so. For the MSSIM this is a small
   optimization point. In our default implementation we split up the image into channels and call
   then for each channel the gpu functions. A small degree of parallelization is possible with the
   stream. By using a stream we can make the data allocation, upload operations while the GPU is
   already executing a given method. For example we need to upload two images. We queue these one
   after another and call already the function that processes it. The functions will wait for the
   upload to finish, however while that happens makes the output buffer allocations for the function
   to be executed next.

   .. code-block:: cpp

      gpu::Stream stream;

      stream.enqueueConvert(b.gI1, b.t1, CV_32F);    // Upload

      gpu::split(b.t1, b.vI1, stream);              // Methods (pass the stream as final parameter).
      gpu::multiply(b.vI1[i], b.vI1[i], b.I1_2, stream);        // I1^2

Result and conclusion
=====================

On an Intel P8700 laptop CPU paired with a low end NVidia GT220M here are the performance numbers:

.. code-block:: cpp

   Time of PSNR CPU (averaged for 10 runs): 41.4122 milliseconds. With result of: 19.2506
   Time of PSNR GPU (averaged for 10 runs): 158.977 milliseconds. With result of: 19.2506
   Initial call GPU optimized:              31.3418 milliseconds. With result of: 19.2506
   Time of PSNR GPU OPTIMIZED ( / 10 runs): 24.8171 milliseconds. With result of: 19.2506

   Time of MSSIM CPU (averaged for 10 runs): 484.343 milliseconds. With result of B0.890964 G0.903845 R0.936934
   Time of MSSIM GPU (averaged for 10 runs): 745.105 milliseconds. With result of B0.89922 G0.909051 R0.968223
   Time of MSSIM GPU Initial Call            357.746 milliseconds. With result of B0.890964 G0.903845 R0.936934
   Time of MSSIM GPU OPTIMIZED ( / 10 runs): 203.091 milliseconds. With result of B0.890964 G0.903845 R0.936934

In both cases we managed a performance increase of almost 100% compared to the CPU implementation.
It may be just the improvement needed for your application to work. You may observe a runtime
instance of this on the `YouTube here <https://www.youtube.com/watch?v=3_ESXmFlnvY>`_.

.. raw:: html

  <div align="center">
  <iframe title="Similarity check (PNSR and SSIM) on the GPU" width="560" height="349" src="http://www.youtube.com/embed/3_ESXmFlnvY?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
  </div>
\ No newline at end of file
+.. _gpuBasicsSimilarity:
+
+Similarity check (PNSR and SSIM) on the GPU
+*******************************************
+
+Goal
+====
+
+In the :ref:`videoInputPSNRMSSIM` tutorial I already presented the PSNR and SSIM methods for
+checking the similarity between the two images. And as you could see there performing these takes
+quite some time, especially in the case of the SSIM. However, if the performance numbers of an
+OpenCV implementation for the CPU do not satisfy you and you happen to have an NVidia CUDA GPU
+device in your system all is not lost. You may try to port or write your algorithm for the video
+card.
+
+This tutorial will give a good grasp on how to approach coding by using the GPU module of OpenCV. As
+a prerequisite you should already know how to handle the core, highgui and imgproc modules. So, our
+goals are:
+
+.. container:: enumeratevisibleitemswithsquare
+
+   + What's different compared to the CPU?
+   + Create the GPU code for the PSNR and SSIM
+   + Optimize the code for maximal performance
+
+The source code
+===============
+
+You may also find the source code and these video file in the
+:file:`samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity` folder of the
+OpenCV source library or :download:`download it from here
+<../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp>`. The
+full source code is quite long (due to the controlling of the application via the command line
+arguments and performance measurement). Therefore, to avoid cluttering up these sections with those
+you'll find here only the functions itself.
+
+The PSNR returns a float number, that if the two inputs are similar between 30 and 50 (higher is
+better).
+
+.. literalinclude:: ../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
+   :language: cpp
+   :linenos:
+   :tab-width: 4
+   :lines: 165-210, 18-23, 210-235
+
+The SSIM returns the MSSIM of the images. This is too a float number between zero and one (higher is
+better), however we have one for each channel. Therefore, we return a *Scalar* OpenCV data
+structure:
+
+.. literalinclude:: ../../../../samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
+   :language: cpp
+   :linenos:
+   :tab-width: 4
+   :lines: 235-355, 26-42, 357-
+
+How to do it? - The GPU
+=======================
+
+Now as you can see we have three types of functions for each operation. One for the CPU and two for
+the GPU. The reason I made two for the GPU is too illustrate that often simple porting your CPU to
+GPU will actually make it slower. If you want some performance gain you will need to remember a few
+rules, whose I'm going to detail later on.
+
+The development of the GPU module was made so that it resembles as much as possible its CPU
+counterpart. This is to make porting easy. The first thing you need to do before writing any code is
+to link the GPU module to your project, and include the header file for the module. All the
+functions and data structures of the GPU are in a *gpu* sub namespace of the *cv* namespace. You may
+add this to the default one via the *use namespace* keyword, or mark it everywhere explicitly via
+the cv:: to avoid confusion. I'll do the later.
+
+.. code-block:: cpp
+
+   #include <opencv2/gpu.hpp>        // GPU structures and methods
+
+GPU stands for **g**\ raphics **p**\ rocessing **u**\ nit. It was originally build to render
+graphical scenes. These scenes somehow build on a lot of data. Nevertheless, these aren't all
+dependent one from another in a sequential way and as it is possible a parallel processing of them.
+Due to this a GPU will contain multiple smaller processing units. These aren't the state of the art
+processors and on a one on one test with a CPU it will fall behind. However, its strength lies in
+its numbers. In the last years there has been an increasing trend to harvest these massive parallel
+powers of the GPU in non-graphical scene rendering too. This gave birth to the general-purpose
+computation on graphics processing units (GPGPU).
+
+The GPU has its own memory. When you read data from the hard drive with OpenCV into a *Mat* object
+that takes place in your systems memory. The CPU works somehow directly on this (via its cache),
+however the GPU cannot. He has too transferred the information he will use for calculations from the
+system memory to its own. This is done via an upload process and takes time. In the end the result
+will have to be downloaded back to your system memory for your CPU to see it and use it. Porting
+small functions to GPU is not recommended as the upload/download time will be larger than the amount
+you gain by a parallel execution.
+
+Mat objects are stored only in the system memory (or the CPU cache). For getting an OpenCV matrix
+to the GPU you'll need to use its GPU counterpart :gpudatastructure:`GpuMat <gpu-gpumat>`. It works
+similar to the Mat with a 2D only limitation and no reference returning for its functions (cannot
+mix GPU references with CPU ones). To upload a Mat object to the GPU you need to call the upload
+function after creating an instance of the class. To download you may use simple assignment to a
+Mat object or use the download function.
+
+.. code-block:: cpp
+
+   Mat I1;         // Main memory item - read image into with imread for example
+   gpu::GpuMat gI; // GPU matrix - for now empty
+   gI1.upload(I1); // Upload a data from the system memory to the GPU memory
+
+   I1 = gI1;       // Download, gI1.download(I1) will work too
+
+Once you have your data up in the GPU memory you may call GPU enabled functions of OpenCV. Most of
+the functions keep the same name just as on the CPU, with the difference that they only accept
+*GpuMat* inputs. A full list of these you will find in the documentation: `online here
+<http://docs.opencv.org/modules/gpu/doc/gpu.html>`_ or the OpenCV reference manual that comes with
+the source code.
+
+Another thing to keep in mind is that not for all channel numbers you can make efficient algorithms
+on the GPU. Generally, I found that the input images for the GPU images need to be either one or
+four channel ones and one of the char or float type for the item sizes. No double support on the
+GPU, sorry. Passing other types of objects for some functions will result in an exception thrown,
+and an error message on the error output. The documentation details in most of the places the types
+accepted for the inputs. If you have three channel images as an input you can do two things: either
+adds a new channel (and use char elements) or split up the image and call the function for each
+image. The first one isn't really recommended as you waste memory.
+
+For some functions, where the position of the elements (neighbor items) doesn't matter quick
+solution is to just reshape it into a single channel image. This is the case for the PSNR
+implementation where for the *absdiff* method the value of the neighbors is not important. However,
+for the *GaussianBlur* this isn't an option and such need to use the split method for the SSIM. With
+this knowledge you can already make a GPU viable code (like mine GPU one) and run it. You'll be
+surprised to see that it might turn out slower than your CPU implementation.
+
+Optimization
+============
+
+The reason for this is that you're throwing out on the window the price for memory allocation and
+data transfer. And on the GPU this is damn high. Another possibility for optimization is to
+introduce asynchronous OpenCV GPU calls too with the help of the
+:gpudatastructure:`gpu::Stream<gpu-stream>`.
+
+1. Memory allocation on the GPU is considerable. Therefore, if it’s possible allocate new memory as
+   few times as possible. If you create a function what you intend to call multiple times it is a
+   good idea to allocate any local parameters for the function only once, during the first call.
+   To do this you create a data structure containing all the local variables you will use. For
+   instance in case of the PSNR these are:
+
+   .. code-block:: cpp
+
+      struct BufferPSNR                                     // Optimized GPU versions
+        {   // Data allocations are very expensive on GPU. Use a buffer to solve: allocate once reuse later.
+        gpu::GpuMat gI1, gI2, gs, t1,t2;
+
+        gpu::GpuMat buf;
+      };
+
+   Then create an instance of this in the main program:
+
+   .. code-block:: cpp
+
+      BufferPSNR bufferPSNR;
+
+   And finally pass this to the function each time you call it:
+
+   .. code-block:: cpp
+
+      double getPSNR_GPU_optimized(const Mat& I1, const Mat& I2, BufferPSNR& b)
+
+   Now you access these local parameters as: *b.gI1*, *b.buf* and so on. The GpuMat will only
+   reallocate itself on a new call if the new matrix size is different from the previous one.
+
+#. Avoid unnecessary function data transfers. Any small data transfer will be significant one once
+   you go to the GPU. Therefore, if possible make all calculations in-place (in other words do not
+   create new memory objects - for reasons explained at the previous point). For example, although
+   expressing arithmetical operations may be easier to express in one line formulas, it will be
+   slower. In case of the SSIM at one point I need to calculate:
+
+   .. code-block:: cpp
+
+      b.t1 = 2 * b.mu1_mu2 + C1;
+
+   Although the upper call will succeed observe that there is a hidden data transfer present. Before
+   it makes the addition it needs to store somewhere the multiplication. Therefore, it will create a
+   local matrix in the background, add to that the *C1* value and finally assign that to *t1*. To
+   avoid this we use the gpu functions, instead of the arithmetic operators:
+
+   .. code-block:: cpp
+
+      gpu::multiply(b.mu1_mu2, 2, b.t1); //b.t1 = 2 * b.mu1_mu2 + C1;
+      gpu::add(b.t1, C1, b.t1);
+
+#. Use asynchronous calls (the :gpudatastructure:`gpu::Stream <gpu-stream>`). By default whenever
+   you call a gpu function it will wait for the call to finish and return with the result
+   afterwards. However, it is possible to make asynchronous calls, meaning it will call for the
+   operation execution, make the costly data allocations for the algorithm and return back right
+   away. Now you can call another function if you wish to do so. For the MSSIM this is a small
+   optimization point. In our default implementation we split up the image into channels and call
+   then for each channel the gpu functions. A small degree of parallelization is possible with the
+   stream. By using a stream we can make the data allocation, upload operations while the GPU is
+   already executing a given method. For example we need to upload two images. We queue these one
+   after another and call already the function that processes it. The functions will wait for the
+   upload to finish, however while that happens makes the output buffer allocations for the function
+   to be executed next.
+
+   .. code-block:: cpp
+
+      gpu::Stream stream;
+
+      stream.enqueueConvert(b.gI1, b.t1, CV_32F);    // Upload
+
+      gpu::split(b.t1, b.vI1, stream);              // Methods (pass the stream as final parameter).
+      gpu::multiply(b.vI1[i], b.vI1[i], b.I1_2, stream);        // I1^2
+
+Result and conclusion
+=====================
+
+On an Intel P8700 laptop CPU paired with a low end NVidia GT220M here are the performance numbers:
+
+.. code-block:: cpp
+
+   Time of PSNR CPU (averaged for 10 runs): 41.4122 milliseconds. With result of: 19.2506
+   Time of PSNR GPU (averaged for 10 runs): 158.977 milliseconds. With result of: 19.2506
+   Initial call GPU optimized:              31.3418 milliseconds. With result of: 19.2506
+   Time of PSNR GPU OPTIMIZED ( / 10 runs): 24.8171 milliseconds. With result of: 19.2506
+
+   Time of MSSIM CPU (averaged for 10 runs): 484.343 milliseconds. With result of B0.890964 G0.903845 R0.936934
+   Time of MSSIM GPU (averaged for 10 runs): 745.105 milliseconds. With result of B0.89922 G0.909051 R0.968223
+   Time of MSSIM GPU Initial Call            357.746 milliseconds. With result of B0.890964 G0.903845 R0.936934
+   Time of MSSIM GPU OPTIMIZED ( / 10 runs): 203.091 milliseconds. With result of B0.890964 G0.903845 R0.936934
+
+In both cases we managed a performance increase of almost 100% compared to the CPU implementation.
+It may be just the improvement needed for your application to work. You may observe a runtime
+instance of this on the `YouTube here <https://www.youtube.com/watch?v=3_ESXmFlnvY>`_.
+
+.. raw:: html
+
+  <div align="center">
+  <iframe title="Similarity check (PNSR and SSIM) on the GPU" width="560" height="349" src="http://www.youtube.com/embed/3_ESXmFlnvY?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
+  </div>
diff --git a/doc/tutorials/highgui/table_of_content_highgui/table_of_content_highgui.rst b/doc/tutorials/highgui/table_of_content_highgui/table_of_content_highgui.rst
index 0ba7c323cb..ef6eacce27 100644
--- a/doc/tutorials/highgui/table_of_content_highgui/table_of_content_highgui.rst
+++ b/doc/tutorials/highgui/table_of_content_highgui/table_of_content_highgui.rst
@@ -74,4 +74,4 @@ This section contains valuable tutorials about how to read/save your image/video
 
    ../trackbar/trackbar
    ../video-input-psnr-ssim/video-input-psnr-ssim
-   ../video-write/video-write
\ No newline at end of file
+   ../video-write/video-write
diff --git a/doc/tutorials/highgui/trackbar/trackbar.rst b/doc/tutorials/highgui/trackbar/trackbar.rst
index d6f7202ce1..5749123c10 100644
--- a/doc/tutorials/highgui/trackbar/trackbar.rst
+++ b/doc/tutorials/highgui/trackbar/trackbar.rst
@@ -28,9 +28,7 @@ Let's modify the program made in the tutorial :ref:`Adding_Images`. We will let
 
 .. code-block:: cpp
 
-   #include <cv.h>
-   #include <highgui.h>
-
+   #include <opencv2/opencv.hpp>
    using namespace cv;
 
    /// Global Variables
@@ -50,41 +48,41 @@ Let's modify the program made in the tutorial :ref:`Adding_Images`. We will let
     */
    void on_trackbar( int, void* )
    {
-    alpha = (double) alpha_slider/alpha_slider_max ;
-    beta = ( 1.0 - alpha );
+     alpha = (double) alpha_slider/alpha_slider_max ;
+     beta = ( 1.0 - alpha );
 
-    addWeighted( src1, alpha, src2, beta, 0.0, dst);
+     addWeighted( src1, alpha, src2, beta, 0.0, dst);
 
-    imshow( "Linear Blend", dst );
+     imshow( "Linear Blend", dst );
    }
 
    int main( int argc, char** argv )
    {
-    /// Read image ( same size, same type )
-    src1 = imread("../../images/LinuxLogo.jpg");
-    src2 = imread("../../images/WindowsLogo.jpg");
+     /// Read image ( same size, same type )
+     src1 = imread("../../images/LinuxLogo.jpg");
+     src2 = imread("../../images/WindowsLogo.jpg");
 
-    if( !src1.data ) { printf("Error loading src1 \n"); return -1; }
-    if( !src2.data ) { printf("Error loading src2 \n"); return -1; }
+     if( !src1.data ) { printf("Error loading src1 \n"); return -1; }
+     if( !src2.data ) { printf("Error loading src2 \n"); return -1; }
 
-    /// Initialize values
-    alpha_slider = 0;
+     /// Initialize values
+     alpha_slider = 0;
 
-    /// Create Windows
-    namedWindow("Linear Blend", 1);
+     /// Create Windows
+     namedWindow("Linear Blend", 1);
 
-    /// Create Trackbars
-    char TrackbarName[50];
-    sprintf( TrackbarName, "Alpha x %d", alpha_slider_max );
+     /// Create Trackbars
+     char TrackbarName[50];
+     sprintf( TrackbarName, "Alpha x %d", alpha_slider_max );
 
-    createTrackbar( TrackbarName, "Linear Blend", &alpha_slider, alpha_slider_max, on_trackbar );
+     createTrackbar( TrackbarName, "Linear Blend", &alpha_slider, alpha_slider_max, on_trackbar );
 
-    /// Show some stuff
-    on_trackbar( alpha_slider, 0 );
+     /// Show some stuff
+     on_trackbar( alpha_slider, 0 );
 
-    /// Wait until user press some key
-    waitKey(0);
-    return 0;
+     /// Wait until user press some key
+     waitKey(0);
+     return 0;
    }
 
 
@@ -154,8 +152,3 @@ Result
   .. image:: images/Adding_Trackbars_Tutorial_Result_1.jpg
      :alt: Adding Trackbars - Lena
      :align: center
-
-
-
-
-
diff --git a/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.rst b/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.rst
index e163348f03..f313cb9086 100644
--- a/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.rst
+++ b/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.rst
@@ -60,7 +60,7 @@ Erosion
 
 * As the kernel :math:`B` is scanned over the image, we compute the minimal pixel value overlapped by :math:`B` and replace the image pixel under the anchor point with that minimal value.
 
-* Analagously to the example for dilation, we can apply the erosion operator to the original image (shown above). You can see in the result below that the bright areas of the image (the background, apparently), get thinner, whereas the dark zones (the "writing"( gets bigger.
+* Analagously to the example for dilation, we can apply the erosion operator to the original image (shown above). You can see in the result below that the bright areas of the image (the background, apparently), get thinner, whereas the dark zones (the "writing") gets bigger.
 
   .. image:: images/Morphology_1_Tutorial_Theory_Erosion.png
      :alt: Erosion result - Theory example
diff --git a/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.rst b/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.rst
index 133a613add..9277a101d6 100644
--- a/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.rst
+++ b/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.rst
@@ -329,4 +329,3 @@ Result
 
      .. image:: images/Histogram_Calculation_Result.jpg
               :align: center
-
diff --git a/doc/tutorials/imgproc/histograms/template_matching/template_matching.rst b/doc/tutorials/imgproc/histograms/template_matching/template_matching.rst
index d8a91560e9..afe15ee0dd 100644
--- a/doc/tutorials/imgproc/histograms/template_matching/template_matching.rst
+++ b/doc/tutorials/imgproc/histograms/template_matching/template_matching.rst
@@ -369,4 +369,3 @@ Results
 
    .. image:: images/Template_Matching_Image_Result.jpg
             :align: center
-
diff --git a/doc/tutorials/imgproc/imgtrans/canny_detector/canny_detector.rst b/doc/tutorials/imgproc/imgtrans/canny_detector/canny_detector.rst
index 01c9050f26..ae2d0c8ace 100644
--- a/doc/tutorials/imgproc/imgtrans/canny_detector/canny_detector.rst
+++ b/doc/tutorials/imgproc/imgtrans/canny_detector/canny_detector.rst
@@ -282,6 +282,3 @@ Result
            :align: center
 
 * Notice how the image is superposed to the black background on the edge regions.
-
-
-
diff --git a/doc/tutorials/imgproc/imgtrans/hough_circle/hough_circle.rst b/doc/tutorials/imgproc/imgtrans/hough_circle/hough_circle.rst
index ecd4ba2192..96257683ef 100644
--- a/doc/tutorials/imgproc/imgtrans/hough_circle/hough_circle.rst
+++ b/doc/tutorials/imgproc/imgtrans/hough_circle/hough_circle.rst
@@ -40,7 +40,7 @@ Code
    * Display the detected circle in a window.
 
    .. |TutorialHoughCirclesSimpleDownload| replace:: here
-   .. _TutorialHoughCirclesSimpleDownload: http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/cpp/houghlines.cpp
+   .. _TutorialHoughCirclesSimpleDownload: http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/cpp/houghcircles.cpp
    .. |TutorialHoughCirclesFancyDownload| replace:: here
    .. _TutorialHoughCirclesFancyDownload: http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/cpp/tutorial_code/ImgTrans/HoughCircle_Demo.cpp
 
diff --git a/doc/tutorials/imgproc/imgtrans/hough_lines/hough_lines.rst b/doc/tutorials/imgproc/imgtrans/hough_lines/hough_lines.rst
index 6b872bff51..d716c08323 100644
--- a/doc/tutorials/imgproc/imgtrans/hough_lines/hough_lines.rst
+++ b/doc/tutorials/imgproc/imgtrans/hough_lines/hough_lines.rst
@@ -290,4 +290,3 @@ We get the following result by using the Probabilistic Hough Line Transform:
    :align: center
 
 You may observe that the number of lines detected vary while you change the *threshold*. The explanation is sort of evident: If you establish a higher threshold, fewer lines will be detected (since you will need more points to declare a line detected).
-
diff --git a/doc/tutorials/imgproc/imgtrans/remap/remap.rst b/doc/tutorials/imgproc/imgtrans/remap/remap.rst
index a8b9fdf744..a060aa779e 100644
--- a/doc/tutorials/imgproc/imgtrans/remap/remap.rst
+++ b/doc/tutorials/imgproc/imgtrans/remap/remap.rst
@@ -311,4 +311,3 @@ Result
          :alt: Result 0 for remapping
          :width: 250pt
          :align: center
-
diff --git a/doc/tutorials/imgproc/imgtrans/warp_affine/warp_affine.rst b/doc/tutorials/imgproc/imgtrans/warp_affine/warp_affine.rst
index b601929781..d61b493371 100644
--- a/doc/tutorials/imgproc/imgtrans/warp_affine/warp_affine.rst
+++ b/doc/tutorials/imgproc/imgtrans/warp_affine/warp_affine.rst
@@ -306,4 +306,3 @@ Result
             :alt: Original image
             :width: 250pt
             :align: center
-
diff --git a/doc/tutorials/imgproc/opening_closing_hats/opening_closing_hats.rst b/doc/tutorials/imgproc/opening_closing_hats/opening_closing_hats.rst
index 6baa3a3d71..ca9d445464 100644
--- a/doc/tutorials/imgproc/opening_closing_hats/opening_closing_hats.rst
+++ b/doc/tutorials/imgproc/opening_closing_hats/opening_closing_hats.rst
@@ -279,4 +279,3 @@ Results
   .. image:: images/Morphology_2_Tutorial_Cover.jpg
      :alt: Morphology 2: Result sample
      :align: center
-
diff --git a/doc/tutorials/imgproc/pyramids/pyramids.rst b/doc/tutorials/imgproc/pyramids/pyramids.rst
index 00baae2ac1..dacc5b93a1 100644
--- a/doc/tutorials/imgproc/pyramids/pyramids.rst
+++ b/doc/tutorials/imgproc/pyramids/pyramids.rst
@@ -259,5 +259,3 @@ Results
   .. image:: images/Pyramids_Tutorial_PyrUp_Result.jpg
      :alt: Pyramids: PyrUp Result
      :align: center
-
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/bounding_rects_circles/bounding_rects_circles.rst b/doc/tutorials/imgproc/shapedescriptors/bounding_rects_circles/bounding_rects_circles.rst
index f91730ec88..24c276dc31 100644
--- a/doc/tutorials/imgproc/shapedescriptors/bounding_rects_circles/bounding_rects_circles.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/bounding_rects_circles/bounding_rects_circles.rst
@@ -121,4 +121,3 @@ Result
 
    .. |BRC_1|  image:: images/Bounding_Rects_Circles_Result.jpg
                     :align: middle
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/bounding_rotated_ellipses/bounding_rotated_ellipses.rst b/doc/tutorials/imgproc/shapedescriptors/bounding_rotated_ellipses/bounding_rotated_ellipses.rst
index 0986e1edcd..9c7fbb85f6 100644
--- a/doc/tutorials/imgproc/shapedescriptors/bounding_rotated_ellipses/bounding_rotated_ellipses.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/bounding_rotated_ellipses/bounding_rotated_ellipses.rst
@@ -123,4 +123,3 @@ Result
 
    .. |BRE_1|  image:: images/Bounding_Rotated_Ellipses_Result.jpg
                     :align: middle
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/find_contours/find_contours.rst b/doc/tutorials/imgproc/shapedescriptors/find_contours/find_contours.rst
index 1fca7df93a..68cf80dc7c 100644
--- a/doc/tutorials/imgproc/shapedescriptors/find_contours/find_contours.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/find_contours/find_contours.rst
@@ -104,4 +104,3 @@ Result
 
    .. |contour_1|  image:: images/Find_Contours_Result.jpg
                      :align: middle
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/hull/hull.rst b/doc/tutorials/imgproc/shapedescriptors/hull/hull.rst
index c1ed79cea1..eb5d19e3b5 100644
--- a/doc/tutorials/imgproc/shapedescriptors/hull/hull.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/hull/hull.rst
@@ -113,4 +113,3 @@ Result
 
    .. |Hull_1|  image:: images/Hull_Result.jpg
                      :align: middle
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/moments/moments.rst b/doc/tutorials/imgproc/shapedescriptors/moments/moments.rst
index 15ac2f51f7..350ca38d46 100644
--- a/doc/tutorials/imgproc/shapedescriptors/moments/moments.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/moments/moments.rst
@@ -133,4 +133,3 @@ Result
    .. |MU_2|  image:: images/Moments_Result2.jpg
                     :width: 250pt
                     :align: middle
-
diff --git a/doc/tutorials/imgproc/shapedescriptors/point_polygon_test/point_polygon_test.rst b/doc/tutorials/imgproc/shapedescriptors/point_polygon_test/point_polygon_test.rst
index b7f72c8150..2d68cc2470 100644
--- a/doc/tutorials/imgproc/shapedescriptors/point_polygon_test/point_polygon_test.rst
+++ b/doc/tutorials/imgproc/shapedescriptors/point_polygon_test/point_polygon_test.rst
@@ -114,4 +114,3 @@ Result
 
    .. |PPT_1|  image:: images/Point_Polygon_Test_Result.jpg
                     :align: middle
-
diff --git a/doc/tutorials/imgproc/table_of_content_imgproc/table_of_content_imgproc.rst b/doc/tutorials/imgproc/table_of_content_imgproc/table_of_content_imgproc.rst
index 1be239cfb1..a371c02c87 100644
--- a/doc/tutorials/imgproc/table_of_content_imgproc/table_of_content_imgproc.rst
+++ b/doc/tutorials/imgproc/table_of_content_imgproc/table_of_content_imgproc.rst
@@ -539,6 +539,3 @@ In this section you will learn about the image processing (manipulation) functio
    ../shapedescriptors/bounding_rotated_ellipses/bounding_rotated_ellipses
    ../shapedescriptors/moments/moments
    ../shapedescriptors/point_polygon_test/point_polygon_test
-
-
-
diff --git a/doc/tutorials/introduction/android_binary_package/O4A_SDK.rst b/doc/tutorials/introduction/android_binary_package/O4A_SDK.rst
index b37126f308..78566e7d28 100644
--- a/doc/tutorials/introduction/android_binary_package/O4A_SDK.rst
+++ b/doc/tutorials/introduction/android_binary_package/O4A_SDK.rst
@@ -48,10 +48,10 @@ The structure of package contents looks as follows:
 
 ::
 
-    OpenCV-2.4.4-android-sdk
+    OpenCV-2.4.6-android-sdk
     |_ apk
-    |   |_ OpenCV_2.4.4_binary_pack_armv7a.apk
-    |   |_ OpenCV_2.4.4_Manager_2.6_XXX.apk
+    |   |_ OpenCV_2.4.6_binary_pack_armv7a.apk
+    |   |_ OpenCV_2.4.6_Manager_2.9_XXX.apk
     |
     |_ doc
     |_ samples
@@ -98,7 +98,7 @@ The structure of package contents looks as follows:
 * :file:`doc` folder contains various OpenCV documentation in PDF format.
   It's also available online at http://docs.opencv.org.
 
-  .. note:: The most recent docs (nightly build) are at http://docs.opencv.org/trunk/.
+  .. note:: The most recent docs (nightly build) are at http://docs.opencv.org/2.4.
             Generally, it's more up-to-date, but can refer to not-yet-released functionality.
 
 .. TODO: I'm not sure that this is the best place to talk about OpenCV Manager
@@ -157,10 +157,10 @@ Get the OpenCV4Android SDK
 
    .. code-block:: bash
 
-      unzip ~/Downloads/OpenCV-2.4.4-android-sdk.zip
+      unzip ~/Downloads/OpenCV-2.4.6-android-sdk.zip
 
-.. |opencv_android_bin_pack| replace:: :file:`OpenCV-2.4.4-android-sdk.zip`
-.. _opencv_android_bin_pack_url: http://sourceforge.net/projects/opencvlibrary/files/opencv-android/2.4.4/OpenCV-2.4.4-android-sdk.zip/download
+.. |opencv_android_bin_pack| replace:: :file:`OpenCV-2.4.6-android-sdk.zip`
+.. _opencv_android_bin_pack_url: http://sourceforge.net/projects/opencvlibrary/files/opencv-android/2.4.6/OpenCV-2.4.6-android-sdk.zip/download
 .. |opencv_android_bin_pack_url| replace:: |opencv_android_bin_pack|
 .. |seven_zip| replace:: 7-Zip
 .. _seven_zip: http://www.7-zip.org/
@@ -295,7 +295,7 @@ Well, running samples from Eclipse is very simple:
   .. code-block:: sh
     :linenos:
 
-    <Android SDK path>/platform-tools/adb install <OpenCV4Android SDK path>/apk/OpenCV_2.4.4_Manager_2.6_armv7a-neon.apk
+    <Android SDK path>/platform-tools/adb install <OpenCV4Android SDK path>/apk/OpenCV_2.4.6_Manager_2.9_armv7a-neon.apk
 
   .. note:: ``armeabi``, ``armv7a-neon``, ``arm7a-neon-android8``, ``mips`` and ``x86`` stand for
             platform targets:
diff --git a/doc/tutorials/introduction/android_binary_package/dev_with_OCV_on_Android.rst b/doc/tutorials/introduction/android_binary_package/dev_with_OCV_on_Android.rst
index 41f5166292..243dc35dd8 100644
--- a/doc/tutorials/introduction/android_binary_package/dev_with_OCV_on_Android.rst
+++ b/doc/tutorials/introduction/android_binary_package/dev_with_OCV_on_Android.rst
@@ -55,14 +55,14 @@ Manager to access OpenCV libraries externally installed in the target system.
    :guilabel:`File -> Import -> Existing project in your workspace`.
 
    Press :guilabel:`Browse`  button and locate OpenCV4Android SDK
-   (:file:`OpenCV-2.4.4-android-sdk/sdk`).
+   (:file:`OpenCV-2.4.6-android-sdk/sdk`).
 
    .. image:: images/eclipse_opencv_dependency0.png
         :alt: Add dependency from OpenCV library
         :align: center
 
 #. In application project add a reference to the OpenCV Java SDK in
-   :guilabel:`Project -> Properties -> Android -> Library -> Add` select ``OpenCV Library - 2.4.4``.
+   :guilabel:`Project -> Properties -> Android -> Library -> Add` select ``OpenCV Library - 2.4.6``.
 
    .. image:: images/eclipse_opencv_dependency1.png
         :alt: Add dependency from OpenCV library
@@ -101,7 +101,7 @@ See the "15-puzzle" OpenCV sample for details.
         public void onResume()
         {
             super.onResume();
-            OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_3, this, mLoaderCallback);
+            OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_6, this, mLoaderCallback);
         }
 
         ...
@@ -128,27 +128,27 @@ described above.
 #. Add the OpenCV library project to your workspace the same way as for the async initialization
    above. Use menu :guilabel:`File -> Import -> Existing project in your workspace`,
    press :guilabel:`Browse` button and select OpenCV SDK path
-   (:file:`OpenCV-2.4.4-android-sdk/sdk`).
+   (:file:`OpenCV-2.4.6-android-sdk/sdk`).
 
    .. image:: images/eclipse_opencv_dependency0.png
         :alt: Add dependency from OpenCV library
         :align: center
 
 #. In the application project add a reference to the OpenCV4Android SDK in
-   :guilabel:`Project -> Properties -> Android -> Library -> Add` select ``OpenCV Library - 2.4.4``;
+   :guilabel:`Project -> Properties -> Android -> Library -> Add` select ``OpenCV Library - 2.4.6``;
 
    .. image:: images/eclipse_opencv_dependency1.png
        :alt: Add dependency from OpenCV library
        :align: center
 
 #. If your application project **doesn't have a JNI part**, just copy the corresponding OpenCV
-   native libs from :file:`<OpenCV-2.4.4-android-sdk>/sdk/native/libs/<target_arch>` to your
+   native libs from :file:`<OpenCV-2.4.6-android-sdk>/sdk/native/libs/<target_arch>` to your
    project directory to folder :file:`libs/<target_arch>`.
 
    In case of the application project **with a JNI part**, instead of manual libraries copying you
    need to modify your ``Android.mk`` file:
    add the following two code lines after the ``"include $(CLEAR_VARS)"`` and before
-   ``"include path_to_OpenCV-2.4.4-android-sdk/sdk/native/jni/OpenCV.mk"``
+   ``"include path_to_OpenCV-2.4.6-android-sdk/sdk/native/jni/OpenCV.mk"``
 
    .. code-block:: make
       :linenos:
@@ -221,7 +221,7 @@ taken:
 
    .. code-block:: make
 
-      include C:\Work\OpenCV4Android\OpenCV-2.4.4-android-sdk\sdk\native\jni\OpenCV.mk
+      include C:\Work\OpenCV4Android\OpenCV-2.4.6-android-sdk\sdk\native\jni\OpenCV.mk
 
    Should be inserted into the :file:`jni/Android.mk` file **after** this line:
 
@@ -379,7 +379,7 @@ result.
        public void onResume()
        {
            super.onResume();
-           OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_3, this, mLoaderCallback);
+           OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_6, this, mLoaderCallback);
        }
 
 #. Defines that your activity implements ``CvViewFrameListener2`` interface and fix activity related
diff --git a/doc/tutorials/introduction/how_to_write_a_tutorial/how_to_write_a_tutorial.rst b/doc/tutorials/introduction/how_to_write_a_tutorial/how_to_write_a_tutorial.rst
index 26ec076f50..5ae5062492 100644
--- a/doc/tutorials/introduction/how_to_write_a_tutorial/how_to_write_a_tutorial.rst
+++ b/doc/tutorials/introduction/how_to_write_a_tutorial/how_to_write_a_tutorial.rst
@@ -1 +1,440 @@
-.. _howToWriteTutorial:

How to write a tutorial for OpenCV?
***********************************

Okay, so assume you have just finished a project of yours implementing something based on OpenCV and you want to present/share it with the community. Luckily, OpenCV is an *open source project*. This means that in theory anyone has access to the full source code and may extend it. While making a robust and practical library (like OpenCV) is great, the success of a library also depends on how user friendly it is. To improve on this aspect, the OpenCV team has already been listening to user feedback from its :opencv_group:`Yahoo user group <>` and by making samples you can find in the source directories sample folder. The addition of the tutorials (in both online and PDF format) is an extension of these efforts.

Goal
====

.. _reST: http://docutils.sourceforge.net/rst.html
.. |reST| replace:: reStructuredText
.. |Sphinx| replace:: Sphinx
.. _Sphinx: http://sphinx.pocoo.org/

The tutorials are just as an important part of the library as  the implementation of those crafty data structures and algorithms you can find in OpenCV. Therefore, the source codes for the tutorials are part of the library. And yes, I meant source codes. The reason for this formulation is that the tutorials are written by using the |Sphinx|_ documentation generation system. This is based on the popular python documentation system called |reST|_ (reST). ReStructuredText is a really neat language that by using a few simple conventions (indentation, directives) and emulating old school e-mail writing techniques (text only) tries to offer a simple way to create and edit documents. Sphinx extends this with some new features and creates the resulting document in both HTML (for web) and PDF (for offline usage) format.

Usually, an OpenCV tutorial has the following parts:

1. A source code demonstration of an OpenCV feature:

   a. One or more CPP, Python, Java or other type of files depending for what OpenCV offers support and for what language you make the tutorial.
   #. Occasionaly, input resource files required for running your tutorials application.


#. A table of content entry (so people may easily find the tutorial):

   a. Adding your stuff to the tutorials table of content (**reST** file).
   #. Add an image file near the TOC entry.


#. The content of the tutorial itself:

   a. The **reST** text of the tutorial
   #. Images following the idea that "*A picture is worth a thousand words*".
   #. For more complex demonstrations you may create a video.

As you can see you will need at least some basic knowledge of the *reST* system in order to complete the task at hand with success. However, don't worry *reST* (and *Sphinx*) was made with simplicity in mind. It is easy to grasp its basics. I found that the `OpenAlea documentations introduction on this subject <http://openalea.gforge.inria.fr/doc/openalea/doc/_build/html/source/tutorial/rest_syntax.html>`_ (or the `Thomas Cokelaer one <http://thomas-cokelaer.info/tutorials/sphinx/rest_syntax.html>`_ ) should enough for this. If for some directive or feature you need a more in-depth description look it up in the official |reST|_ help files or at the |Sphinx|_ documentation.

In our world achieving some tasks is possible in multiple ways. However, some of the roads to take may have obvious or hidden advantages over others. Then again, in some other cases it may come down to just simple user preference. Here, I'll present how I decided to write the tutorials, based on my personal experience. If for some of them you know a better solution and you can back it up feel free to use that. I've nothing against it, as long as it gets the job done in an elegant fashion.

Now the best would be if you could make the integration yourself. For this you need first to have the source code. I recommend following the guides for your operating system on acquiring OpenCV sources. For Linux users look :ref:`here <Linux-Installation>` and for :ref:`Windows here <Windows_Installation>`. You must also install python and sphinx with its dependencies in order to be able to build the documentation.

Once you have downloaded the repository to your hard drive you can take a look in the OpenCV directory to make sure you have both the samples and doc folder present. Anyone may download the trunk source files from  :file:`git://code.opencv.org/opencv.git` . Nevertheless, not everyone has upload (commit/submit) rights. This is to protect the integrity of the library. If you plan doing more than one tutorial, and would like to have an account with commit user rights you should first register an account at http://code.opencv.org/ and then contact dr. Gary Bradski at -delete-bradski@-delete-willowgarage.com. Otherwise, you can just send the resulting files to us via the :opencv_group:`Yahoo user group <>` or to me at -delete-bernat@-delete-primeranks.net and I'll add it. If you have questions, suggestions or constructive critics I will gladly listen to them. If you send it to the OpenCV group please tag its subject with a **[Tutorial]** entry.

Format the Source Code
======================

Before I start this let it be clear: the main goal is to have a working sample code. However, for your tutorial to be of a top notch quality you should follow a few guide lines I am going to present here.

In case you have an application by using the older interface (with *IplImage*, *CVMat*, *cvLoadImage* and such) consider migrating it to the new C++ interface. The tutorials are intended to be an up to date help for our users. And as of OpenCV 2 the OpenCV emphasis on using the less error prone and clearer C++ interface. Therefore, if possible please convert your code to the C++ interface. For this it may help to read the :ref:`InteroperabilityWithOpenCV1` tutorial. However, once you have an OpenCV 2 working code, then you should make your source code snippet as easy to read as possible. Here're a couple of advices for this:

.. container:: enumeratevisibleitemswithsquare

   + Add a standard output with the description of what your program does. Keep it short and yet, descriptive. This output is at the start of the program. In my example files this usually takes the form of a *help* function containing the output. This way both the source file viewer and application runner can see what all is about in your sample. Here's an instance of this:

     .. code-block:: cpp

        void help()
        {
        cout
        << "--------------------------------------------------------------------------"   << endl
        << "This program shows how to write video files. You can extract the R or G or B color channel "
        << " of the input video. You can choose to use the source codec (Y) or select a custom one. (N)"<< endl
        << "Usage:"                                                                       << endl
        << "./video-write inputvideoName [ R | G | B] [Y | N]"                            << endl
        << "--------------------------------------------------------------------------"   << endl
        << endl;
        }
        // ...
        int main(int argc, char *argv[], char *window_name)
        {
        help();
        // here comes the actual source code
        }

     Additionally, finalize the description with a short usage guide. This way the user will know how to call your programs, what leads us to the next point.

   + Prefer command line argument controlling instead of hard coded one. If your program has some variables that may be changed use command line arguments for this. The tutorials, can be a simple try-out ground for the user. If you offer command line controlling for the input image (for example), then you offer the possibility for the user to try it out with his/her own images, without the need to mess in the source code. In the upper example you can see that the input image, channel and codec selection may all be changed from the command line. Just compile the program and run it with your own input arguments.

   + Be as verbose as possible. There is no shame in filling the source code with comments. This way the more advanced user may figure out what's happening right from the sample code. This advice goes for the output console too. Specify to the user what's happening. Never leave the user hanging there and thinking on: "Is this program now crashing or just doing some computationally intensive task?." So, if you do a training task that may take some time, make sure you print out a message about this before starting and after finishing it.

   + Throw out unnecessary stuff from your source code. This is a warning to not take the previous point too seriously. Balance is the key. If it's something that can be done in a fewer lines or simpler than that's the way you should do it. Nevertheless, if for some reason you have such sections notify the user why you have chosen to do so. Keep the amount of information as low as possible, while still getting the job done in an elegant way.

   + Put your sample file into the :file:`opencv/samples/cpp/tutorial_code/sectionName` folder. If you write a tutorial for other languages than cpp, then change that part of the path. Before completing this you need to decide that to what section (module) does your tutorial goes. Think about on what module relies most heavily your code and that is the one to use. If the answer to this question is more than one modules then the *general* section is the one to use. For finding the *opencv* directory open up your file system and navigate where you downloaded our repository.

   + If the input resources are hard to acquire for the end user consider adding a few of them to the :file:`opencv/samples/cpp/tutorial_code/images`. Make sure that who reads your code can try it out!

Add the TOC entry
=================

For this you will need to know some |reST|_. There is no going around this. |reST|_ files have **rst** extensions. However, these are simple text files. Use any text editor you like. Finding a text editor that offers syntax highlighting for |reST|_ was quite a challenge at the time of writing this tutorial. In my experience, `Intype <http://intype.info/>`_ is a solid option on Windows, although there is still place for improvement.

Adding your source code to a table of content is important for multiple reasons. First and foremost this will allow for the user base to find your tutorial from our websites tutorial table of content. Secondly, if you omit this *Sphinx* will throw a warning that your tutorial file isn't part of any TOC tree entry. And there is nothing more than the developer team hates than an ever increasing warning/error list for their builds. *Sphinx* also uses this to build up the previous-back-up buttons on the website. Finally, omitting this step will lead to that your tutorial will **not** be added to the PDF version of the tutorials.

Navigate to the :file:`opencv/doc/tutorials/section/table_of_content_section` folder (where the section is the module to which you're adding the tutorial). Open the *table_of_content_section* file. Now this may have two forms. If no prior tutorials are present in this section that there is a template message about this and has the following form:

.. code-block:: rst

  .. _Table-Of-Content-Section:

   Section title
   -----------------------------------------------------------

   Description about the section.

   .. include:: ../../definitions/noContent.rst

   .. raw:: latex

      \pagebreak

The first line is a reference to the section title in the reST system. The section title will be a link and you may refer to it via the ``:ref:`` directive. The *include* directive imports the template text from the definitions directories *noContent.rst* file. *Sphinx* does not creates the PDF from scratch. It does this by first creating a latex file. Then creates the PDF from the latex file. With the *raw* directive you can directly add to this output commands. Its unique argument is for what kind of output to add the content of the directive. For the PDFs it may happen that multiple sections will overlap on a single page. To avoid this at the end of the TOC we add a *pagebreak* latex command, that hints to the LATEX system that the next line should be on a new page.

If you have one of this, try to transform it to the following form:

.. include:: ../../definitions/tocDefinitions.rst

.. code-block:: rst

   .. _Table-Of-Content-Section:

   Section title
   -----------------------------------------------------------

   .. include:: ../../definitions/tocDefinitions.rst

   +
     .. tabularcolumns:: m{100pt} m{300pt}
     .. cssclass:: toctableopencv

     =============== ======================================================
      |MatBasicIma|  **Title:** :ref:`matTheBasicImageContainer`

                     *Compatibility:* > OpenCV 2.0

                     *Author:* |Author_BernatG|

                     You will learn how to store images in the memory and how to print out their content to the console.

     =============== =====================================================

     .. |MatBasicIma| image:: images/matTheBasicImageStructure.jpg
                      :height: 90pt
                      :width:  90pt

   .. raw:: latex

      \pagebreak

   .. toctree::
      :hidden:

      ../mat - the basic image container/mat - the basic image container

If this is already present just add a new section of the content between the include and the raw directives (excluding those lines). Here you'll see a new include directive. This should be present only once in a TOC tree and the reST file contains the definitions of all the authors contributing to the OpenCV tutorials. We are a multicultural community and some of our name may contain some funky characters. However, reST **only supports** ANSI characters. Luckily we can specify Unicode characters with the *unicode* directive. Doing this for all of your tutorials is a troublesome procedure. Therefore, the tocDefinitions file contains the definition of your author name. Add it here once and afterwards just use the replace construction. For example here's the definition for my name:

.. code-block:: rst

   .. |Author_BernatG| unicode:: Bern U+00E1 t U+0020 G U+00E1 bor

The ``|Author_BernatG|`` is the text definitions alias. I can use later this to add the definition, like I've done in the TOCs *Author* part. After the ``::`` and a space you start the definition. If you want to add an UNICODE character (non-ASCI) leave an empty space and specify it in the format U+(UNICODE code). To find the UNICODE code of a character I recommend using the `FileFormat <http://www.fileformat.info>`_ websites service. Spaces are trimmed from the definition, therefore we add a space by its UNICODE character (U+0020).

Until the *raw* directive what you can see is a TOC tree entry. Here's how a TOC entry will look like:

+
  .. tabularcolumns:: m{100pt} m{300pt}
  .. cssclass:: toctableopencv

  =============== ======================================================
   |MatBasicIma|  **Title:** :ref:`matTheBasicImageContainer`

                  *Compatibility:* > OpenCV 2.0

                  *Author:* |Author_BernatG|

                  You will learn how to store images in the memory and how to print out their content to the console.

  =============== ======================================================

  .. |MatBasicIma| image:: images/matTheBasicImageStructure.jpg
                   :height: 90pt
                   :width:  90pt

As you can see we have an image to the left and a description box to the right. To create two boxes we use a table with two columns and a single row. In the left column is the image and in the right one the description. However, the image directive is way too long to fit in a column. Therefore, we need to use the substitution definition system. We add this definition after the TOC tree. All images for the TOC tree are to be put in the images folder near its |reST|_ file. We use the point measurement system because we are also creating PDFs. PDFs are printable documents, where there is no such thing that pixels (px), just points (pt). And while generally space is no problem for web pages (we have monitors with **huge** resolutions) the size of the paper (A4 or letter) is constant and will be for a long time in the future. Therefore, size constrains come in play more like for the PDF, than the generated HTML code.

Now your images should be as small as possible, while still offering the intended information for the user. Remember that the tutorial will become part of the OpenCV source code. If you add large images (that manifest in form of large image size) it will just increase the size of the repository pointlessly. If someone wants to download it later, its download time will be that much longer. Not to mention the larger PDF size for the tutorials and the longer load time for the web pages. In terms of pixels a TOC image should not be larger than 120 X 120 pixels. Resize your images if they are larger!

.. note::

   If you add a larger image and specify a smaller image size, *Sphinx* will not resize that. At build time will add the full size image and the resize will be done by your browser after the image is loaded. A 120 X 120 image is somewhere below 10KB. If you add a 110KB image, you have just pointlessly added a 100KB extra data to transfer over the internet for every user!

Generally speaking you shouldn't need to specify your images size (excluding the TOC entries). If no such is found *Sphinx* will use the size of the image itself (so no resize occurs). Then again if for some reason you decide to specify a size that should be the **width** of the image rather than its height. The reason for this again goes back to the PDFs. On a PDF page the height is larger than the width. In the PDF the images will not be resized. If you specify a size that does not fit in the page, then what does not fits in **will be cut off**. When creating your images for your tutorial you should try to keep the image widths below 500 pixels, and calculate with around 400 point page width when specifying image widths.

The image format depends on the content of the image. If you have some complex scene (many random like colors) then use *jpg*. Otherwise, prefer using *png*. They are even some tools out there that optimize the size of *PNG* images, such as `PNGGauntlet <http://pnggauntlet.com/>`_. Use them to make your images as small as possible in size.

Now on the right side column of the table we add the information about the tutorial:

.. container:: enumeratevisibleitemswithsquare

   + In the first line it is the title of the tutorial. However, there is no need to specify it explicitly. We use the reference system. We'll start up our tutorial with a reference specification, just like in case of this TOC entry with its  `` .. _Table-Of-Content-Section:`` . If after this you have a title (pointed out by the following line of -), then Sphinx will replace the ``:ref:`Table-Of-Content-Section``` directive with the tile of the section in reference form (creates a link in web page). Here's how the definition looks in my case:

     .. code-block:: rst

        .. _matTheBasicImageContainer:

           Mat - The Basic Image Container
           *******************************

     Note, that according to the |reST|_ rules the * should be as long as your title.

   + Compatibility. What version of OpenCV is required to run your sample code.

   + Author. Use the substitution markup of |reST|_.

   + A short sentence describing the essence of your tutorial.

Now before each TOC entry you need to add the three lines of:

.. code-block:: cpp

   +
     .. tabularcolumns:: m{100pt} m{300pt}
     .. cssclass:: toctableopencv

The plus sign (+) is to enumerate tutorials by using bullet points. So for every TOC entry we have a corresponding bullet point represented by the +. Sphinx is highly indenting sensitive. Indentation is used to express from which point until to which point does a construction last. Un-indentation means end of that construction. So to keep all the bullet points to the same group the following TOC entries (until the next +) should be indented by two spaces.

Here, I should also mention that **always** prefer using spaces instead of tabs. Working with only spaces makes possible that if we both use monotype fonts we will see the same thing. Tab size is text editor dependent and as should be avoided. *Sphinx* translates all tabs into 8 spaces before interpreting it.

It turns out that the automatic formatting of both the HTML and PDF(LATEX) system messes up our tables. Therefore, we need to help them out a little. For the PDF generation we add the ``.. tabularcolumns:: m{100pt} m{300pt}`` directive. This means that the first column should be 100 points wide and middle aligned. For the HTML look we simply name the following table of a *toctableopencv* class type. Then, we can modify the look of the table by modifying the CSS of our web page. The CSS definitions go into the :file:`opencv/doc/_themes/blue/static/default.css_t` file.

.. code-block:: css

   .toctableopencv
   {
    width: 100% ;
    table-layout: fixed;
   }


   .toctableopencv colgroup col:first-child
   {
    width: 100pt !important;
    max-width: 100pt !important;
    min-width: 100pt !important;
   }

   .toctableopencv colgroup col:nth-child(2)
   {
    width: 100% !important;
   }

However, you should not need to modify this. Just add these three lines (plus keep the two space indentation) for all TOC entries you add. At the end of the TOC file you'll find:

.. code-block:: rst

   .. raw:: latex

      \pagebreak

   .. toctree::
      :hidden:

      ../mat - the basic image container/mat - the basic image container

The page break entry comes for separating sections and should be only one in a TOC tree |reST|_ file. Finally, at the end of the TOC tree we need to add our tutorial to the *Sphinx* TOC tree system. *Sphinx* will generate from this the previous-next-up information for the HTML file and add items to the PDF according to the order here. By default this TOC tree directive generates a simple table of contents. However, we already created a fancy looking one so we no longer need this basic one. Therefore, we add the *hidden* option to do not show it.

The path is of a relative type. We step back in the file system and then go into the :file:`mat - the basic image container` directory for the :file:`mat - the basic image container.rst` file. Putting out the *rst* extension for the file is optional.

Write the tutorial
==================

Create a folder with the name of your tutorial. Preferably, use small letters only. Then create a text file in this folder with *rst* extension and the same name. If you have images for the tutorial create an :file:`images` folder and add your images there. When creating your images follow the guidelines described in the previous part!

Now here's our recommendation for the structure of the tutorial (although, remember that this is not carved in the stone; if you have a better idea, use it!):


.. container:: enumeratevisibleitemswithsquare

   + Create the reference point and the title.

     .. code-block:: rst

        .. _matTheBasicImageContainer:

        Mat - The Basic Image Container
        *******************************

     You start the tutorial by specifying a reference point by the ``.. _matTheBasicImageContainer:`` and then its title. The name of the reference point should be a unique one over the whole documentation. Therefore, do not use general names like *tutorial1*. Use the * character to underline the title for its full width. The subtitles of the tutorial should be underlined with = charachter.

   + Goals. You start your tutorial by specifying what you will present. You can also enumerate the sub jobs to be done. For this you can use a bullet point construction. There is a single configuration file for both the reference manual and the tutorial documentation. In the reference manuals at the argument enumeration we do not want any kind of bullet point style enumeration. Therefore, by default all the bullet points at this level are set to do not show the dot before the entries in the HTML. You can override this by putting the bullet point in a container. I've defined a square type bullet point view under the name *enumeratevisibleitemswithsquare*. The CSS style definition for this is again in the  :file:`opencv\doc\_themes\blue\static\default.css_t` file. Here's a quick example of using it:

     .. code-block:: rst

        .. container:: enumeratevisibleitemswithsquare

           + Create the reference point and the title.
           + Second entry
           + Third entry

     Note that you need the keep the indentation of the container directive. Directive indentations are always three (3) spaces. Here you may even give usage tips for your sample code.

   + Source code. Present your samples code to the user. It's a good idea to offer a quick download link for the HTML page by using the *download* directive and pointing out where the user may find your source code in the file system by using the *file* directive:

     .. code-block:: rst

        Text :file:`samples/cpp/tutorial_code/highgui/video-write/` folder of the OpenCV source library
        or :download:`text to appear in the webpage
        <../../../../samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp>`.

     For the download link the path is a relative one, hence the multiple back stepping operations (..). Then you can add the source code either by using the *code block* directive or the *literal include* one. In case of the code block you will need to actually add all the source code text into your |reST|_ text and also apply the required indentation:

     .. code-block:: rst

        .. code-block:: cpp

           int i = 0;
           l = ++j;

     The only argument of the directive is the language used (here CPP). Then you add the source code into its content (meaning one empty line after the directive) by keeping the indentation of the directive (3 spaces). With the *literal include* directive you do not need to add the source code of the sample. You just specify the sample and *Sphinx* will load it for you, during build time. Here's an example usage:

     .. code-block:: rst

        .. literalinclude:: ../../../../samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp
           :language: cpp
           :linenos:
           :tab-width: 4
           :lines: 1-8, 21-22, 24-

     After the directive you specify a relative path to the file from what to import. It has four options: the language to use, if you add the ``:linenos:`` the line numbers will be shown, you can specify the tab size with the ``:tab-width:`` and you do not need to load the whole file, you can show just the important lines. Use the *lines* option to do not show redundant information (such as the *help* function). Here basically you specify ranges, if the second range line number is missing than that means that until the end of the file. The ranges specified here do no need to be in an ascending order, you may even reorganize the structure of how you want to show your sample inside the tutorial.

   + The tutorial. Well here goes the explanation for why and what have you used. Try to be short, clear, concise and yet a thorough one. There's no magic formula. Look into a few already made tutorials and start out from there. Try to mix sample OpenCV code with your explanations. If with words is hard to describe something do not hesitate to add in a reasonable size image, to overcome this issue.

     When you present OpenCV functionality it's a good idea to give a link to the used OpenCV data structure or function. Because the OpenCV tutorials and reference manual are in separate PDF files it is not possible to make this link work for the PDF format. Therefore, we use here only web page links to the **opencv.itseez.com** website. The OpenCV functions and data structures may be used for multiple tasks. Nevertheless, we want to avoid that every users creates its own reference to a commonly used function. So for this we use the global link collection of *Sphinx*. This is defined in the file:`opencv/doc/conf.py` configuration file. Open it and go all the way down to the last entry:

     .. code-block:: py

       # ---- External links for tutorials -----------------
       extlinks = {
           'hgvideo' : ('http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None)
           }

     In short here we defined a new **hgvideo** directive that refers to an external webpage link. Its usage is:

     .. code-block:: rst

       A sample function of the highgui modules image write and read page is the :hgvideo:`imread() function <imread>`.

     Which turns to: A sample function of the highgui modules image write and read page is the :hgvideo:`imread() function <imread>`. The argument you give between the <> will be put in place of the ``%s`` in the upper definition, and as the link will anchor to the correct function. To find out the anchor of a given function just open up a web page, search for the function and click on it. In the address bar it should appear like: ``http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#imread`` .  Look here for the name of the directives for each page of the OpenCV reference manual. If none present for one of them feel free to add one for it.

     For formulas you can add LATEX code that will translate in the web pages into images. You do this by using the *math* directive. A usage tip:

     .. code-block:: latex

        .. math::

           MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}

     That after build turns into:

     .. math::

        MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}

     You can even use it inline as ``:math:` MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}``` that turns into :math:`MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}`.

     If you use some crazy LATEX library extension you need to add those to the ones to use at build time. Look into the file:`opencv/doc/conf.py` configuration file for more information on this.

   + Results. Well, here depending on your program show one of more of the following:

     - Console outputs by using the code block directive.
     - Output images.
     - Runtime videos, visualization. For this use your favorite screens capture software. `Camtasia Studio <http://www.techsmith.com/camtasia/>`_ certainly is one of the better choices, however their prices are out of this world. `CamStudio <http://camstudio.org/>`_ is a free alternative, but less powerful. If you do a video you can upload it to YouTube and then use the raw directive with HTML option to embed it into the generated web page:

       .. code-block:: rst

          You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=jpBwHxsl1_0>`_.

          .. raw:: html

             <div align="center">
             <iframe title="Creating a video with OpenCV" width="560" height="349" src="http://www.youtube.com/embed/jpBwHxsl1_0?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
             </div>

       This results in the text and video: You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=jpBwHxsl1_0>`_.

       .. raw:: html

          <div align="center">
          <iframe title="Creating a video with OpenCV" width="560" height="349" src="http://www.youtube.com/embed/jpBwHxsl1_0?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
          </div>

     When these aren't self-explanatory make sure to throw in a few guiding lines about what and why we can see.

   + Build the documentation and check for errors or warnings. In the CMake make sure you check or pass the option for building documentation. Then simply build the **docs** project for the PDF file and the **docs_html** project for the web page. Read the output of the build and check for errors/warnings for what you have added. This is also the time to observe and correct any kind of *not so good looking* parts. Remember to keep clean our build logs.

   + Read again your tutorial and check for both programming and spelling errors. If found any, please correct them.


Take home the pride and joy of a job well done!
===============================================

Once you are done contact me or dr. Gary Bradski with the tutorial. We may submit the tutorial ourselves to the trunk branch of our repository or ask you to do so.

Now, to see your work **live** you may need to wait some time. The PDFs are updated usually at the launch of a new OpenCV version. The web pages are a little more diverse. They are automatically rebuilt in each evening. However, the **opencv.itseez.com** website contains only the most recent **stable branch** of OpenCV. Currently this is 2.3. When we add something new (like a tutorial) that first goes to the **trunk branch** of our repository. A build of this you may find on the **opencv.itseez.com/trunk** website. Although, we try to make a build every night occasionally we might freeze any of the branches to fix upcoming issues. During this it may take a little longer to see your work *live*, however if you submited it, be sure that eventually it will show up.

If you have any questions or advices relating to this tutorial you can contact me at -delete-bernat@-delete-primeranks.net. Of course, delete the -delete- parts of that e-mail address.
\ No newline at end of file
+.. _howToWriteTutorial:
+
+How to write a tutorial for OpenCV
+**********************************
+
+Okay, so assume you have just finished a project of yours implementing something
+based on OpenCV and you want to present/share it with the community. Luckily, OpenCV
+is an *open source project*. This means that anyone has access to the full source
+code and may propose extensions. And a good tutorial is a valuable addition to the
+library! Please read instructions on contribution process here:
+http://opencv.org/contribute.html. You may also find this page helpful:
+:how_to_contribute:`How to contribute <>`.
+
+While making a robust and practical library (like OpenCV) is great, the success of a
+library also depends on how user friendly it is. To improve on this aspect, the
+OpenCV team has already been listening to user feedback at :opencv_qa:`OpenCV Q&A
+forum <>` and by making samples you can find in the source directories
+:file:`samples` folder. The addition of the tutorials (in both online and PDF format)
+is an extension of these efforts.
+
+Goal
+====
+
+.. _reST: http://docutils.sourceforge.net/rst.html
+.. |reST| replace:: reStructuredText
+.. |Sphinx| replace:: Sphinx
+.. _Sphinx: http://sphinx.pocoo.org/
+
+The tutorials are just as an important part of the library as the implementation of
+those crafty data structures and algorithms you can find in OpenCV. Therefore, the
+source codes for the tutorials are part of the library. And yes, I meant source
+codes. The reason for this formulation is that the tutorials are written by using the
+|Sphinx|_ documentation generation system. This is based on the popular Python
+documentation system called |reST|_ (reST). ReStructuredText is a really neat
+language that by using a few simple conventions (indentation, directives) and
+emulating old school email writing techniques (text only) tries to offer a simple
+way to create and edit documents. Sphinx extends this with some new features and
+creates the resulting document in both HTML (for web) and PDF (for offline usage)
+format.
+
+
+Usually, an OpenCV tutorial has the following parts:
+
+1. A source code demonstration of an OpenCV feature:
+
+   a. One or more CPP, Python, Java or other type of files depending for what OpenCV offers support and for what language you make the tutorial.
+   #. Occasionaly, input resource files required for running your tutorials application.
+
+
+#. A table of content entry (so people may easily find the tutorial):
+
+   a. Adding your stuff to the tutorials table of content (**reST** file).
+   #. Add an image file near the TOC entry.
+
+
+#. The content of the tutorial itself:
+
+   a. The **reST** text of the tutorial
+   #. Images following the idea that "*A picture is worth a thousand words*".
+   #. For more complex demonstrations you may create a video.
+
+As you can see you will need at least some basic knowledge of the *reST* system in order to complete the task at hand with success. However, don't worry *reST* (and *Sphinx*) was made with simplicity in mind. It is easy to grasp its basics. I found that the `OpenAlea documentations introduction on this subject <http://openalea.gforge.inria.fr/doc/openalea/doc/_build/html/source/tutorial/rest_syntax.html>`_ (or the `Thomas Cokelaer one <http://thomas-cokelaer.info/tutorials/sphinx/rest_syntax.html>`_ ) should enough for this. If for some directive or feature you need a more in-depth description look it up in the official |reST|_ help files or at the |Sphinx|_ documentation.
+
+In our world achieving some tasks is possible in multiple ways. However, some of the roads to take may have obvious or hidden advantages over others. Then again, in some other cases it may come down to just simple user preference. Here, I'll present how I decided to write the tutorials, based on my personal experience. If for some of them you know a better solution and you can back it up feel free to use that. I've nothing against it, as long as it gets the job done in an elegant fashion.
+
+Now the best would be if you could make the integration yourself. For this you need first to have the source code. I recommend following the guides for your operating system on acquiring OpenCV sources. For Linux users look :ref:`here <Linux-Installation>` and for :ref:`Windows here <Windows_Installation>`. You must also install python and sphinx with its dependencies in order to be able to build the documentation.
+
+Once you have downloaded the repository to your hard drive you can take a look in the OpenCV directory to make sure you have both the samples and doc folder present. Anyone may download the latest source files from :file:`git://github.com/Itseez/opencv.git` . Nevertheless, not everyone has upload (commit/submit) rights. This is to protect the integrity of the library. If you plan doing more than one tutorial, and would like to have an account with commit user rights you should first register an account at http://code.opencv.org/ and then contact OpenCV administrator -delete-admin@-delete-opencv.org. Otherwise, you can just send the resulting files to us at -delete-admin@-delete-opencv.org and we'll add it.
+
+
+Format the Source Code
+======================
+
+Before I start this let it be clear: the main goal is to have a working sample code. However, for your tutorial to be of a top notch quality you should follow a few guide lines I am going to present here. In case you have an application by using the older interface (with *IplImage*, *cvMat*, *cvLoadImage* and such) consider migrating it to the new C++ interface. The tutorials are intended to be an up to date help for our users. And as of OpenCV 2 the OpenCV emphasis on using the less error prone and clearer C++ interface. Therefore, if possible please convert your code to the C++ interface. For this it may help to read the :ref:`InteroperabilityWithOpenCV1` tutorial. However, once you have an OpenCV 2 working code, then you should make your source code snippet as easy to read as possible. Here're a couple of advices for this:
+
+
+.. container:: enumeratevisibleitemswithsquare
+
+   + Add a standard output with the description of what your program does. Keep it short and yet, descriptive. This output is at the start of the program. In my example files this usually takes the form of a *help* function containing the output. This way both the source file viewer and application runner can see what all is about in your sample. Here's an instance of this:
+
+     .. code-block:: cpp
+
+        void help()
+        {
+        cout
+        << "--------------------------------------------------------------------------"   << endl
+        << "This program shows how to write video files. You can extract the R or G or B color channel "
+        << " of the input video. You can choose to use the source codec (Y) or select a custom one. (N)"<< endl
+        << "Usage:"                                                                       << endl
+        << "./video-write inputvideoName [ R | G | B] [Y | N]"                            << endl
+        << "--------------------------------------------------------------------------"   << endl
+        << endl;
+        }
+        // ...
+        int main(int argc, char *argv[], char *window_name)
+        {
+        help();
+        // here comes the actual source code
+        }
+
+     Additionally, finalize the description with a short usage guide. This way the user will know how to call your programs, what leads us to the next point.
+
+   + Prefer command line argument controlling instead of hard coded one. If your program has some variables that may be changed use command line arguments for this. The tutorials, can be a simple try-out ground for the user. If you offer command line controlling for the input image (for example), then you offer the possibility for the user to try it out with his/her own images, without the need to mess in the source code. In the upper example you can see that the input image, channel and codec selection may all be changed from the command line. Just compile the program and run it with your own input arguments.
+
+   + Be as verbose as possible. There is no shame in filling the source code with comments. This way the more advanced user may figure out what's happening right from the sample code. This advice goes for the output console too. Specify to the user what's happening. Never leave the user hanging there and thinking on: "Is this program now crashing or just doing some computationally intensive task?." So, if you do a training task that may take some time, make sure you print out a message about this before starting and after finishing it.
+
+   + Throw out unnecessary stuff from your source code. This is a warning to not take the previous point too seriously. Balance is the key. If it's something that can be done in a fewer lines or simpler than that's the way you should do it. Nevertheless, if for some reason you have such sections notify the user why you have chosen to do so. Keep the amount of information as low as possible, while still getting the job done in an elegant way.
+
+   + Put your sample file into the :file:`opencv/samples/cpp/tutorial_code/sectionName` folder. If you write a tutorial for other languages than cpp, then change that part of the path. Before completing this you need to decide that to what section (module) does your tutorial goes. Think about on what module relies most heavily your code and that is the one to use. If the answer to this question is more than one modules then the *general* section is the one to use. For finding the *opencv* directory open up your file system and navigate where you downloaded our repository.
+
+   + If the input resources are hard to acquire for the end user consider adding a few of them to the :file:`opencv/samples/cpp/tutorial_code/images`. Make sure that who reads your code can try it out!
+
+Add the TOC entry
+=================
+
+For this you will need to know some |reST|_. There is no going around this. |reST|_ files have **rst** extensions. However, these are simple text files. Use any text editor you like. Finding a text editor that offers syntax highlighting for |reST|_ was quite a challenge at the time of writing this tutorial. In my experience, `Intype <http://intype.info/>`_ is a solid option on Windows, although there is still place for improvement.
+
+Adding your source code to a table of content is important for multiple reasons. First and foremost this will allow for the user base to find your tutorial from our websites tutorial table of content. Secondly, if you omit this *Sphinx* will throw a warning that your tutorial file isn't part of any TOC tree entry. And there is nothing more than the developer team hates than an ever increasing warning/error list for their builds. *Sphinx* also uses this to build up the previous-back-up buttons on the website. Finally, omitting this step will lead to that your tutorial will **not** be added to the PDF version of the tutorials.
+
+Navigate to the :file:`opencv/doc/tutorials/section/table_of_content_section` folder (where the section is the module to which you're adding the tutorial). Open the *table_of_content_section* file. Now this may have two forms. If no prior tutorials are present in this section that there is a template message about this and has the following form:
+
+.. code-block:: rst
+
+  .. _Table-Of-Content-Section:
+
+   Section title
+   -----------------------------------------------------------
+
+   Description about the section.
+
+   .. include:: ../../definitions/noContent.rst
+
+   .. raw:: latex
+
+      \pagebreak
+
+The first line is a reference to the section title in the reST system. The section title will be a link and you may refer to it via the ``:ref:`` directive. The *include* directive imports the template text from the definitions directories *noContent.rst* file. *Sphinx* does not creates the PDF from scratch. It does this by first creating a latex file. Then creates the PDF from the latex file. With the *raw* directive you can directly add to this output commands. Its unique argument is for what kind of output to add the content of the directive. For the PDFs it may happen that multiple sections will overlap on a single page. To avoid this at the end of the TOC we add a *pagebreak* latex command, that hints to the LATEX system that the next line should be on a new page.
+
+If you have one of this, try to transform it to the following form:
+
+.. include:: ../../definitions/tocDefinitions.rst
+
+.. code-block:: rst
+
+   .. _Table-Of-Content-Section:
+
+   Section title
+   -----------------------------------------------------------
+
+   .. include:: ../../definitions/tocDefinitions.rst
+
+   +
+     .. tabularcolumns:: m{100pt} m{300pt}
+     .. cssclass:: toctableopencv
+
+     =============== ======================================================
+      |MatBasicIma|  **Title:** :ref:`matTheBasicImageContainer`
+
+                     *Compatibility:* > OpenCV 2.0
+
+                     *Author:* |Author_BernatG|
+
+                     You will learn how to store images in the memory and how to print out their content to the console.
+
+     =============== =====================================================
+
+     .. |MatBasicIma| image:: images/matTheBasicImageStructure.jpg
+                      :height: 90pt
+                      :width:  90pt
+
+   .. raw:: latex
+
+      \pagebreak
+
+   .. toctree::
+      :hidden:
+
+      ../mat - the basic image container/mat - the basic image container
+
+If this is already present just add a new section of the content between the include and the raw directives (excluding those lines). Here you'll see a new include directive. This should be present only once in a TOC tree and the reST file contains the definitions of all the authors contributing to the OpenCV tutorials. We are a multicultural community and some of our name may contain some funky characters. However, reST **only supports** ANSI characters. Luckily we can specify Unicode characters with the *unicode* directive. Doing this for all of your tutorials is a troublesome procedure. Therefore, the tocDefinitions file contains the definition of your author name. Add it here once and afterwards just use the replace construction. For example here's the definition for my name:
+
+.. code-block:: rst
+
+   .. |Author_BernatG| unicode:: Bern U+00E1 t U+0020 G U+00E1 bor
+
+The ``|Author_BernatG|`` is the text definitions alias. I can use later this to add the definition, like I've done in the TOCs *Author* part. After the ``::`` and a space you start the definition. If you want to add an UNICODE character (non-ASCI) leave an empty space and specify it in the format U+(UNICODE code). To find the UNICODE code of a character I recommend using the `FileFormat <http://www.fileformat.info>`_ websites service. Spaces are trimmed from the definition, therefore we add a space by its UNICODE character (U+0020).
+
+Until the *raw* directive what you can see is a TOC tree entry. Here's how a TOC entry will look like:
+
++
+  .. tabularcolumns:: m{100pt} m{300pt}
+  .. cssclass:: toctableopencv
+
+  =============== ======================================================
+   |MatBasicIma|  **Title:** :ref:`matTheBasicImageContainer`
+
+                  *Compatibility:* > OpenCV 2.0
+
+                  *Author:* |Author_BernatG|
+
+                  You will learn how to store images in the memory and how to print out their content to the console.
+
+  =============== ======================================================
+
+  .. |MatBasicIma| image:: images/matTheBasicImageStructure.jpg
+                   :height: 90pt
+                   :width:  90pt
+
+As you can see we have an image to the left and a description box to the right. To create two boxes we use a table with two columns and a single row. In the left column is the image and in the right one the description. However, the image directive is way too long to fit in a column. Therefore, we need to use the substitution definition system. We add this definition after the TOC tree. All images for the TOC tree are to be put in the images folder near its |reST|_ file. We use the point measurement system because we are also creating PDFs. PDFs are printable documents, where there is no such thing that pixels (px), just points (pt). And while generally space is no problem for web pages (we have monitors with **huge** resolutions) the size of the paper (A4 or letter) is constant and will be for a long time in the future. Therefore, size constrains come in play more like for the PDF, than the generated HTML code.
+
+Now your images should be as small as possible, while still offering the intended information for the user. Remember that the tutorial will become part of the OpenCV source code. If you add large images (that manifest in form of large image size) it will just increase the size of the repository pointlessly. If someone wants to download it later, its download time will be that much longer. Not to mention the larger PDF size for the tutorials and the longer load time for the web pages. In terms of pixels a TOC image should not be larger than 120 X 120 pixels. Resize your images if they are larger!
+
+.. note::
+
+   If you add a larger image and specify a smaller image size, *Sphinx* will not resize that. At build time will add the full size image and the resize will be done by your browser after the image is loaded. A 120 X 120 image is somewhere below 10KB. If you add a 110KB image, you have just pointlessly added a 100KB extra data to transfer over the internet for every user!
+
+Generally speaking you shouldn't need to specify your images size (excluding the TOC entries). If no such is found *Sphinx* will use the size of the image itself (so no resize occurs). Then again if for some reason you decide to specify a size that should be the **width** of the image rather than its height. The reason for this again goes back to the PDFs. On a PDF page the height is larger than the width. In the PDF the images will not be resized. If you specify a size that does not fit in the page, then what does not fits in **will be cut off**. When creating your images for your tutorial you should try to keep the image widths below 500 pixels, and calculate with around 400 point page width when specifying image widths.
+
+The image format depends on the content of the image. If you have some complex scene (many random like colors) then use *jpg*. Otherwise, prefer using *png*. They are even some tools out there that optimize the size of *PNG* images, such as `PNGGauntlet <http://pnggauntlet.com/>`_. Use them to make your images as small as possible in size.
+
+Now on the right side column of the table we add the information about the tutorial:
+
+.. container:: enumeratevisibleitemswithsquare
+
+   + In the first line it is the title of the tutorial. However, there is no need to specify it explicitly. We use the reference system. We'll start up our tutorial with a reference specification, just like in case of this TOC entry with its  `` .. _Table-Of-Content-Section:`` . If after this you have a title (pointed out by the following line of -), then Sphinx will replace the ``:ref:`Table-Of-Content-Section``` directive with the tile of the section in reference form (creates a link in web page). Here's how the definition looks in my case:
+
+     .. code-block:: rst
+
+        .. _matTheBasicImageContainer:
+
+           Mat - The Basic Image Container
+           *******************************
+
+     Note, that according to the |reST|_ rules the * should be as long as your title.
+
+   + Compatibility. What version of OpenCV is required to run your sample code.
+
+   + Author. Use the substitution markup of |reST|_.
+
+   + A short sentence describing the essence of your tutorial.
+
+Now before each TOC entry you need to add the three lines of:
+
+.. code-block:: cpp
+
+   +
+     .. tabularcolumns:: m{100pt} m{300pt}
+     .. cssclass:: toctableopencv
+
+The plus sign (+) is to enumerate tutorials by using bullet points. So for every TOC entry we have a corresponding bullet point represented by the +. Sphinx is highly indenting sensitive. Indentation is used to express from which point until to which point does a construction last. Un-indentation means end of that construction. So to keep all the bullet points to the same group the following TOC entries (until the next +) should be indented by two spaces.
+
+Here, I should also mention that **always** prefer using spaces instead of tabs. Working with only spaces makes possible that if we both use monotype fonts we will see the same thing. Tab size is text editor dependent and as should be avoided. *Sphinx* translates all tabs into 8 spaces before interpreting it.
+
+It turns out that the automatic formatting of both the HTML and PDF(LATEX) system messes up our tables. Therefore, we need to help them out a little. For the PDF generation we add the ``.. tabularcolumns:: m{100pt} m{300pt}`` directive. This means that the first column should be 100 points wide and middle aligned. For the HTML look we simply name the following table of a *toctableopencv* class type. Then, we can modify the look of the table by modifying the CSS of our web page. The CSS definitions go into the :file:`opencv/doc/_themes/blue/static/default.css_t` file.
+
+.. code-block:: css
+
+   .toctableopencv
+   {
+    width: 100% ;
+    table-layout: fixed;
+   }
+
+
+   .toctableopencv colgroup col:first-child
+   {
+    width: 100pt !important;
+    max-width: 100pt !important;
+    min-width: 100pt !important;
+   }
+
+   .toctableopencv colgroup col:nth-child(2)
+   {
+    width: 100% !important;
+   }
+
+However, you should not need to modify this. Just add these three lines (plus keep the two space indentation) for all TOC entries you add. At the end of the TOC file you'll find:
+
+.. code-block:: rst
+
+   .. raw:: latex
+
+      \pagebreak
+
+   .. toctree::
+      :hidden:
+
+      ../mat - the basic image container/mat - the basic image container
+
+The page break entry comes for separating sections and should be only one in a TOC tree |reST|_ file. Finally, at the end of the TOC tree we need to add our tutorial to the *Sphinx* TOC tree system. *Sphinx* will generate from this the previous-next-up information for the HTML file and add items to the PDF according to the order here. By default this TOC tree directive generates a simple table of contents. However, we already created a fancy looking one so we no longer need this basic one. Therefore, we add the *hidden* option to do not show it.
+
+The path is of a relative type. We step back in the file system and then go into the :file:`mat - the basic image container` directory for the :file:`mat - the basic image container.rst` file. Putting out the *rst* extension for the file is optional.
+
+Write the tutorial
+==================
+
+Create a folder with the name of your tutorial. Preferably, use small letters only. Then create a text file in this folder with *rst* extension and the same name. If you have images for the tutorial create an :file:`images` folder and add your images there. When creating your images follow the guidelines described in the previous part!
+
+Now here's our recommendation for the structure of the tutorial (although, remember that this is not carved in the stone; if you have a better idea, use it!):
+
+
+.. container:: enumeratevisibleitemswithsquare
+
+   + Create the reference point and the title.
+
+     .. code-block:: rst
+
+        .. _matTheBasicImageContainer:
+
+        Mat - The Basic Image Container
+        *******************************
+
+     You start the tutorial by specifying a reference point by the ``.. _matTheBasicImageContainer:`` and then its title. The name of the reference point should be a unique one over the whole documentation. Therefore, do not use general names like *tutorial1*. Use the * character to underline the title for its full width. The subtitles of the tutorial should be underlined with = charachter.
+
+   + Goals. You start your tutorial by specifying what you will present. You can also enumerate the sub jobs to be done. For this you can use a bullet point construction. There is a single configuration file for both the reference manual and the tutorial documentation. In the reference manuals at the argument enumeration we do not want any kind of bullet point style enumeration. Therefore, by default all the bullet points at this level are set to do not show the dot before the entries in the HTML. You can override this by putting the bullet point in a container. I've defined a square type bullet point view under the name *enumeratevisibleitemswithsquare*. The CSS style definition for this is again in the  :file:`opencv\doc\_themes\blue\static\default.css_t` file. Here's a quick example of using it:
+
+     .. code-block:: rst
+
+        .. container:: enumeratevisibleitemswithsquare
+
+           + Create the reference point and the title.
+           + Second entry
+           + Third entry
+
+     Note that you need the keep the indentation of the container directive. Directive indentations are always three (3) spaces. Here you may even give usage tips for your sample code.
+
+   + Source code. Present your samples code to the user. It's a good idea to offer a quick download link for the HTML page by using the *download* directive and pointing out where the user may find your source code in the file system by using the *file* directive:
+
+     .. code-block:: rst
+
+        Text :file:`samples/cpp/tutorial_code/highgui/video-write/` folder of the OpenCV source library
+        or :download:`text to appear in the webpage
+        <../../../../samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp>`.
+
+     For the download link the path is a relative one, hence the multiple back stepping operations (..). Then you can add the source code either by using the *code block* directive or the *literal include* one. In case of the code block you will need to actually add all the source code text into your |reST|_ text and also apply the required indentation:
+
+     .. code-block:: rst
+
+        .. code-block:: cpp
+
+           int i = 0;
+           l = ++j;
+
+     The only argument of the directive is the language used (here CPP). Then you add the source code into its content (meaning one empty line after the directive) by keeping the indentation of the directive (3 spaces). With the *literal include* directive you do not need to add the source code of the sample. You just specify the sample and *Sphinx* will load it for you, during build time. Here's an example usage:
+
+     .. code-block:: rst
+
+        .. literalinclude:: ../../../../samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp
+           :language: cpp
+           :linenos:
+           :tab-width: 4
+           :lines: 1-8, 21-22, 24-
+
+     After the directive you specify a relative path to the file from what to import. It has four options: the language to use, if you add the ``:linenos:`` the line numbers will be shown, you can specify the tab size with the ``:tab-width:`` and you do not need to load the whole file, you can show just the important lines. Use the *lines* option to do not show redundant information (such as the *help* function). Here basically you specify ranges, if the second range line number is missing than that means that until the end of the file. The ranges specified here do no need to be in an ascending order, you may even reorganize the structure of how you want to show your sample inside the tutorial.
+
+   + The tutorial. Well here goes the explanation for why and what have you used. Try to be short, clear, concise and yet a thorough one. There's no magic formula. Look into a few already made tutorials and start out from there. Try to mix sample OpenCV code with your explanations. If with words is hard to describe something do not hesitate to add in a reasonable size image, to overcome this issue.
+
+     When you present OpenCV functionality it's a good idea to give a link to the used OpenCV data structure or function. Because the OpenCV tutorials and reference manual are in separate PDF files it is not possible to make this link work for the PDF format. Therefore, we use here only web page links to the http://docs.opencv.org website. The OpenCV functions and data structures may be used for multiple tasks. Nevertheless, we want to avoid that every users creates its own reference to a commonly used function. So for this we use the global link collection of *Sphinx*. This is defined in the file:`opencv/doc/conf.py` configuration file. Open it and go all the way down to the last entry:
+
+     .. code-block:: py
+
+       # ---- External links for tutorials -----------------
+       extlinks = {
+           'hgvideo' : ('http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html#%s', None)
+           }
+
+     In short here we defined a new **hgvideo** directive that refers to an external webpage link. Its usage is:
+
+     .. code-block:: rst
+
+       A sample function of the highgui modules image write and read page is the :hgvideo:`imread() function <imread>`.
+
+     Which turns to: A sample function of the highgui modules image write and read page is the :hgvideo:`imread() function <imread>`. The argument you give between the <> will be put in place of the ``%s`` in the upper definition, and as the link will anchor to the correct function. To find out the anchor of a given function just open up a web page, search for the function and click on it. In the address bar it should appear like: ``http://docs.opencv.org/modules/highgui/doc/reading_and_writing_images_and_video.html#imread`` .  Look here for the name of the directives for each page of the OpenCV reference manual. If none present for one of them feel free to add one for it.
+
+     For formulas you can add LATEX code that will translate in the web pages into images. You do this by using the *math* directive. A usage tip:
+
+     .. code-block:: latex
+
+        .. math::
+
+           MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}
+
+     That after build turns into:
+
+     .. math::
+
+        MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}
+
+     You can even use it inline as ``:math:` MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}``` that turns into :math:`MSE = \frac{1}{c*i*j} \sum{(I_1-I_2)^2}`.
+
+     If you use some crazy LATEX library extension you need to add those to the ones to use at build time. Look into the file:`opencv/doc/conf.py` configuration file for more information on this.
+
+   + Results. Well, here depending on your program show one of more of the following:
+
+     - Console outputs by using the code block directive.
+     - Output images.
+     - Runtime videos, visualization. For this use your favorite screens capture software. `Camtasia Studio <http://www.techsmith.com/camtasia/>`_ certainly is one of the better choices, however their prices are out of this world. `CamStudio <http://camstudio.org/>`_ is a free alternative, but less powerful. If you do a video you can upload it to YouTube and then use the raw directive with HTML option to embed it into the generated web page:
+
+       .. code-block:: rst
+
+          You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=jpBwHxsl1_0>`_.
+
+          .. raw:: html
+
+             <div align="center">
+             <iframe title="Creating a video with OpenCV" width="560" height="349" src="http://www.youtube.com/embed/jpBwHxsl1_0?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
+             </div>
+
+       This results in the text and video: You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=jpBwHxsl1_0>`_.
+
+       .. raw:: html
+
+          <div align="center">
+          <iframe title="Creating a video with OpenCV" width="560" height="349" src="http://www.youtube.com/embed/jpBwHxsl1_0?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
+          </div>
+
+     When these aren't self-explanatory make sure to throw in a few guiding lines about what and why we can see.
+
+   + Build the documentation and check for errors or warnings. In the CMake make sure you check or pass the option for building documentation. Then simply build the **docs** project for the PDF file and the **docs_html** project for the web page. Read the output of the build and check for errors/warnings for what you have added. This is also the time to observe and correct any kind of *not so good looking* parts. Remember to keep clean our build logs.
+
+   + Read again your tutorial and check for both programming and spelling errors. If found any, please correct them.
+
+
+Take home the pride and joy of a job well done!
+===============================================
+
+Once you are done please make a GitHub pull request with the tutorial. Now, to see
+your work **live** you may need to wait some time. The PDFs are updated usually at
+the launch of a new OpenCV version. The web pages are a little more diverse. They are
+automatically rebuilt nightly. Currently we use ``2.4`` and ``master`` branches for
+daily builds. So, if your pull request was merged to any of these branches, your
+material will be published at `docs.opencv.org/2.4 <http:/docs.opencv.org/2.4>`_ or
+`docs.opencv.org/master <http:/docs.opencv.org/master>`_ correspondingly. Everything
+that was added to ``2.4`` is merged to ``master`` branch every week. Although, we try
+to make a build every night, occasionally we might freeze any of the branches to fix
+upcoming issues. During this it may take a little longer to see your work online,
+however if you submitted it, be sure that eventually it will show up.
+
+If you have any questions or advices relating to this tutorial you can contact us at
+-delete-admin@-delete-opencv.org (delete the -delete- parts of that email address).
diff --git a/doc/tutorials/introduction/ios_install/ios_install.rst b/doc/tutorials/introduction/ios_install/ios_install.rst
index cb9e0650b2..2973b7ec2b 100644
--- a/doc/tutorials/introduction/ios_install/ios_install.rst
+++ b/doc/tutorials/introduction/ios_install/ios_install.rst
@@ -37,7 +37,7 @@ Building OpenCV from Source, using CMake and Command Line
     .. code-block:: bash
 
        cd ~/<my_working_directory>
-       python opencv/ios/build_framework.py ios
+       python opencv/platforms/ios/build_framework.py ios
 
 If everything's fine, a few minutes later you will get ~/<my_working_directory>/ios/opencv2.framework. You can add this framework to your Xcode projects.
 
diff --git a/doc/tutorials/introduction/linux_eclipse/linux_eclipse.rst b/doc/tutorials/introduction/linux_eclipse/linux_eclipse.rst
index c1a6fac000..74e0ff9e09 100644
--- a/doc/tutorials/introduction/linux_eclipse/linux_eclipse.rst
+++ b/doc/tutorials/introduction/linux_eclipse/linux_eclipse.rst
@@ -65,8 +65,7 @@ Making a project
 
    .. code-block:: cpp
 
-      #include <cv.h>
-      #include <highgui.h>
+      #include <opencv2/opencv.hpp>
 
       using namespace cv;
 
@@ -81,7 +80,7 @@ Making a project
             return -1;
           }
 
-        namedWindow( "Display Image", CV_WINDOW_AUTOSIZE );
+        namedWindow( "Display Image", WINDOW_AUTOSIZE );
         imshow( "Display Image", image );
 
         waitKey(0);
@@ -201,29 +200,20 @@ Assuming that the image to use as the argument would be located in <DisplayImage
 V2: Using CMake+OpenCV with Eclipse (plugin CDT)
 ==================================================
 
-(See the `getting started <http://opencv.willowgarage.com/wiki/Getting_started>` section of the OpenCV Wiki)
-
 Say you have or create a new file, *helloworld.cpp* in a directory called *foo*:
 
 .. code-block:: cpp
 
 
-   #include <cv.h>
-   #include <highgui.h>
+   #include <opencv2/opencv.hpp>
+   using namespace cv;
+
    int main ( int argc, char **argv )
    {
-     cvNamedWindow( "My Window", 1 );
-     IplImage *img = cvCreateImage( cvSize( 640, 480 ), IPL_DEPTH_8U, 1 );
-     CvFont font;
-     double hScale = 1.0;
-     double vScale = 1.0;
-     int lineWidth = 1;
-     cvInitFont( &font, CV_FONT_HERSHEY_SIMPLEX | CV_FONT_ITALIC,
-                 hScale, vScale, 0, lineWidth );
-     cvPutText( img, "Hello World!", cvPoint( 200, 400 ), &font,
-                cvScalar( 255, 255, 0 ) );
-     cvShowImage( "My Window", img );
-     cvWaitKey();
+     Mat img(480, 640, CV_8U);
+     putText(img, "Hello World!", Point( 200, 400 ), FONT_HERSHEY_SIMPLEX | FONT_ITALIC, 1.0, Scalar( 255, 255, 0 ));
+     imshow("My Window", img);
+     waitKey();
      return 0;
    }
 
@@ -255,6 +245,3 @@ Say you have or create a new file, *helloworld.cpp* in a directory called *foo*:
  a. You can also optionally modify the ``Build command:`` from ``make`` to something like ``make VERBOSE=1 -j4`` which tells the compiler to produce detailed symbol files for debugging and also to compile in 4 parallel threads.
 
 #. Done!
-
-
-
diff --git a/doc/tutorials/introduction/linux_install/linux_install.rst b/doc/tutorials/introduction/linux_install/linux_install.rst
index e8b96dab73..1e02b64c9d 100644
--- a/doc/tutorials/introduction/linux_install/linux_install.rst
+++ b/doc/tutorials/introduction/linux_install/linux_install.rst
@@ -80,4 +80,3 @@ Building OpenCV from Source Using CMake, Using the Command Line
 .. note::
 
    If the size of the created library is a critical issue (like in case of an Android build) you can use the ``install/strip`` command to get the smallest size as possible. The *stripped* version appears to be twice as small. However, we do not recommend using this unless those extra megabytes do really matter.
-
diff --git a/doc/tutorials/introduction/load_save_image/load_save_image.rst b/doc/tutorials/introduction/load_save_image/load_save_image.rst
index 50fb9ea37f..ac0ee02e5d 100644
--- a/doc/tutorials/introduction/load_save_image/load_save_image.rst
+++ b/doc/tutorials/introduction/load_save_image/load_save_image.rst
@@ -26,8 +26,7 @@ Here it is:
 .. code-block:: cpp
    :linenos:
 
-   #include <cv.h>
-   #include <highgui.h>
+   #include <opencv2/opencv.hpp>
 
    using namespace cv;
 
@@ -45,12 +44,12 @@ Here it is:
     }
 
     Mat gray_image;
-    cvtColor( image, gray_image, CV_BGR2GRAY );
+    cvtColor( image, gray_image, COLOR_BGR2GRAY );
 
     imwrite( "../../images/Gray_Image.jpg", gray_image );
 
-    namedWindow( imageName, CV_WINDOW_AUTOSIZE );
-    namedWindow( "Gray image", CV_WINDOW_AUTOSIZE );
+    namedWindow( imageName, WINDOW_AUTOSIZE );
+    namedWindow( "Gray image", WINDOW_AUTOSIZE );
 
     imshow( imageName, image );
     imshow( "Gray image", gray_image );
diff --git a/doc/tutorials/introduction/table_of_content_introduction/images/visual_studio_image_watch.png b/doc/tutorials/introduction/table_of_content_introduction/images/visual_studio_image_watch.png
new file mode 100755
index 0000000000..e693344df8
Binary files /dev/null and b/doc/tutorials/introduction/table_of_content_introduction/images/visual_studio_image_watch.png differ
diff --git a/doc/tutorials/introduction/table_of_content_introduction/table_of_content_introduction.rst b/doc/tutorials/introduction/table_of_content_introduction/table_of_content_introduction.rst
index 5e91a5392c..2238b87878 100644
--- a/doc/tutorials/introduction/table_of_content_introduction/table_of_content_introduction.rst
+++ b/doc/tutorials/introduction/table_of_content_introduction/table_of_content_introduction.rst
@@ -103,6 +103,21 @@ world of the OpenCV.
                       :height: 90pt
                       :width:  90pt
 
+  =========== ======================================================
+  |WinVSVis|  **Title:** :ref:`Windows_Visual_Studio_Image_Watch`
+
+              *Compatibility:* >= OpenCV 2.4
+
+              *Author:* Wolf Kienzle
+
+              You will learn how to visualize OpenCV matrices and images within Visual Studio 2012.
+
+  =========== ======================================================
+
+     .. |WinVSVis| image:: images/visual_studio_image_watch.png
+                        :height: 90pt
+                        :width:  90pt
+
 * **Desktop Java**
 
   .. tabularcolumns:: m{100pt} m{300pt}
@@ -278,6 +293,7 @@ world of the OpenCV.
    ../linux_eclipse/linux_eclipse
    ../windows_install/windows_install
    ../windows_visual_studio_Opencv/windows_visual_studio_Opencv
+   ../windows_visual_studio_image_watch/windows_visual_studio_image_watch
    ../desktop_java/java_dev_intro
    ../android_binary_package/android_dev_intro
    ../android_binary_package/O4A_SDK
diff --git a/doc/tutorials/introduction/windows_install/windows_install.rst b/doc/tutorials/introduction/windows_install/windows_install.rst
index eebda7b066..c29c13aede 100644
--- a/doc/tutorials/introduction/windows_install/windows_install.rst
+++ b/doc/tutorials/introduction/windows_install/windows_install.rst
@@ -292,7 +292,7 @@ Building the library
 
    This will create an *Install* directory inside the *Build* one collecting all the built binaries into a single place. Use this only after you built both the *Release* and *Debug* versions.
 
-   To test your build just go into the :file:`Build/bin/Debug` or :file:`Build/bin/Release` directory and start a couple of applications like the *contours.exe*. If they run, you are done. Otherwise, something definitely went awfully wrong. In this case you should contact us via our :opencv_group:`user group <>`.
+   To test your build just go into the :file:`Build/bin/Debug` or :file:`Build/bin/Release` directory and start a couple of applications like the *contours.exe*. If they run, you are done. Otherwise, something definitely went awfully wrong. In this case you should contact us at our :opencv_qa:`Q&A forum <>`.
    If everything is okay the *contours.exe* output should resemble the following image (if built with Qt support):
 
    .. image:: images/WindowsQtContoursOutput.png
@@ -312,9 +312,13 @@ First we set an enviroment variable to make easier our work. This will hold the
 
 ::
 
-   setx -m OPENCV_DIR D:\OpenCV\Build\x86\vc10
+   setx -m OPENCV_DIR D:\OpenCV\Build\x86\vc10     (suggested for Visual Studio 2010 - 32 bit Windows)
+   setx -m OPENCV_DIR D:\OpenCV\Build\x64\vc10     (suggested for Visual Studio 2010 - 64 bit Windows)
 
-Here the directory is where you have your OpenCV binaries (*extracted* or *built*). You can have different platform (e.g. x64 instead of x86) or compiler type, so substitute appropriate value. Inside this you should have folders like *bin* and *include*. The -m should be added if you wish to make the settings computer wise, instead of user wise.
+   setx -m OPENCV_DIR D:\OpenCV\Build\x86\vc11     (suggested for Visual Studio 2012 - 32 bit Windows)
+   setx -m OPENCV_DIR D:\OpenCV\Build\x64\vc11     (suggested for Visual Studio 2012 - 64 bit Windows)
+
+Here the directory is where you have your OpenCV binaries (*extracted* or *built*). You can have different platform (e.g. x64 instead of x86) or compiler type, so substitute appropriate value. Inside this you should have two folders called *lib* and *bin*. The -m should be added if you wish to make the settings computer wise, instead of user wise.
 
 If you built static libraries then you are done. Otherwise, you need to add the *bin* folders path to the systems path. This is cause you will use the OpenCV library in form of *\"Dynamic-link libraries\"* (also known as **DLL**). Inside these are stored all the algorithms and information the OpenCV library contains. The operating system will load them only on demand, during runtime. However, to do this he needs to know where they are. The systems **PATH** contains a list of folders where DLLs can be found. Add the OpenCV library path to this and the OS will know where to look if he ever needs the OpenCV binaries. Otherwise, you will need to copy the used DLLs right beside the applications executable file (*exe*) for the OS to find it, which is highly unpleasent if you work on many projects. To do this start up again the |PathEditor|_ and add the following new entry (right click in the application to bring up the menu):
 
diff --git a/doc/tutorials/introduction/windows_visual_studio_Opencv/images/PropertySheetOpenCVInclude.jpg b/doc/tutorials/introduction/windows_visual_studio_Opencv/images/PropertySheetOpenCVInclude.jpg
index 5826f4dd6c..bc61d33862 100644
Binary files a/doc/tutorials/introduction/windows_visual_studio_Opencv/images/PropertySheetOpenCVInclude.jpg and b/doc/tutorials/introduction/windows_visual_studio_Opencv/images/PropertySheetOpenCVInclude.jpg differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_Opencv/windows_visual_studio_Opencv.rst b/doc/tutorials/introduction/windows_visual_studio_Opencv/windows_visual_studio_Opencv.rst
index ec227e7241..f3058a74d2 100644
--- a/doc/tutorials/introduction/windows_visual_studio_Opencv/windows_visual_studio_Opencv.rst
+++ b/doc/tutorials/introduction/windows_visual_studio_Opencv/windows_visual_studio_Opencv.rst
@@ -52,7 +52,7 @@ Use for example the *OpenCV_Debug* name. Then by selecting the sheet :menuselect
 
 .. code-block:: bash
 
-   $(OPENCV_DIR)\include
+   $(OPENCV_DIR)\..\..\include
 
 .. image:: images/PropertySheetOpenCVInclude.jpg
    :alt: Add the include dir like this.
@@ -64,7 +64,7 @@ Next go to the :menuselection:`Linker --> General` and under the *"Additional Li
 
 .. code-block:: bash
 
-   $(OPENCV_DIR)\libs
+   $(OPENCV_DIR)\lib
 
 .. image:: images/PropertySheetOpenCVLib.jpg
    :alt: Add the library folder like this.
@@ -86,7 +86,7 @@ The names of the libraries are as follow:
 
    opencv_(The Name of the module)(The version Number of the library you use)d.lib
 
-A full list, for the currently latest trunk version would contain:
+A full list, for the latest version would contain:
 
 .. code-block:: bash
 
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/breakpoint.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/breakpoint.png
new file mode 100755
index 0000000000..2597a8deb8
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/breakpoint.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/edges_zoom.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/edges_zoom.png
new file mode 100755
index 0000000000..8c76959988
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/edges_zoom.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/help_button.jpg b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/help_button.jpg
new file mode 100644
index 0000000000..38aa809d5b
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/help_button.jpg differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/input_zoom.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/input_zoom.png
new file mode 100755
index 0000000000..d4e2df7fb0
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/input_zoom.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/toolwindow.jpg b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/toolwindow.jpg
new file mode 100644
index 0000000000..cd60c00b2d
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/toolwindow.jpg differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer.jpg b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer.jpg
new file mode 100644
index 0000000000..95f4b81a58
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer.jpg differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer_context_menu.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer_context_menu.png
new file mode 100755
index 0000000000..902f23a3dc
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/viewer_context_menu.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/visual_studio_image_watch.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/visual_studio_image_watch.png
new file mode 100755
index 0000000000..e693344df8
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/visual_studio_image_watch.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/images/vs_locals.png b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/vs_locals.png
new file mode 100755
index 0000000000..0df403467a
Binary files /dev/null and b/doc/tutorials/introduction/windows_visual_studio_image_watch/images/vs_locals.png differ
diff --git a/doc/tutorials/introduction/windows_visual_studio_image_watch/windows_visual_studio_image_watch.rst b/doc/tutorials/introduction/windows_visual_studio_image_watch/windows_visual_studio_image_watch.rst
new file mode 100644
index 0000000000..91b411682d
--- /dev/null
+++ b/doc/tutorials/introduction/windows_visual_studio_image_watch/windows_visual_studio_image_watch.rst
@@ -0,0 +1,144 @@
+.. _Windows_Visual_Studio_Image_Watch:
+
+Image Watch: viewing in-memory images in the Visual Studio debugger
+*******************************************************************
+
+Image Watch is a plug-in for Microsoft Visual Studio that lets you to visualize in-memory images (*cv::Mat* or *IplImage\_* objects, for example) while debugging an application. This can be helpful for tracking down bugs, or for simply understanding what a given piece of code is doing.
+
+Prerequisites
+=============
+
+This tutorial assumes that you have the following available:
+
+#. Visual Studio 2012 Professional (or better) with Update 1 installed. Update 1 can be downloaded `here <http://www.microsoft.com/en-us/download/details.aspx?id=35774>`_.
+
+#. An OpenCV installation on your Windows machine (Tutorial: :ref:`Windows_Installation`).
+
+#. Ability to create and build OpenCV projects in Visual Studio (Tutorial: :ref:`Windows_Visual_Studio_How_To`).
+
+Installation
+============
+
+`Download <http://go.microsoft.com/fwlink/?LinkId=285460>`_ the Image Watch installer. The installer comes in a single file with extension .vsix (*Visual Studio Extension*). To launch it, simply double-click on the .vsix file in Windows Explorer. When the installer has finished, make sure to restart Visual Studio to complete the installation.
+
+Example
+========
+
+Image Watch works with any existing project that uses OpenCV image objects (for example, *cv::Mat*). In this example, we use a minimal test program that loads an image from a file and runs an edge detector. To build the program, create a console application project in Visual Studio, name it "image-watch-demo", and insert the source code below.
+
+.. code-block:: c++
+
+    // Test application for the Visual Studio Image Watch Debugger extension
+
+    #include <iostream>                        // std::cout
+    #include <opencv2/core/core.hpp>           // cv::Mat
+    #include <opencv2/highgui/highgui.hpp>     // cv::imread()
+    #include <opencv2/imgproc/imgproc.hpp>     // cv::Canny()
+
+    using namespace std;
+    using namespace cv;
+
+    void help()
+    {
+        cout
+            << "----------------------------------------------------" << endl
+            << "This is a test program for the Image Watch Debugger " << endl
+            << "plug-in for Visual Studio. The program loads an     " << endl
+            << "image from a file and runs the Canny edge detector. " << endl
+            << "No output is displayed or written to disk."
+            << endl
+            << "Usage:"                                               << endl
+            << "image-watch-demo inputimage"                          << endl
+            << "----------------------------------------------------" << endl
+            << endl;
+    }
+
+    int main(int argc, char *argv[])
+    {
+        help();
+
+        if (argc != 2)
+        {
+            cout << "Wrong number of parameters" << endl;
+            return -1;
+        }
+
+        cout << "Loading input image: " << argv[1] << endl;
+        Mat input;
+        input = imread(argv[1], CV_LOAD_IMAGE_COLOR);
+
+        cout << "Detecting edges in input image" << endl;
+        Mat edges;
+        Canny(input, edges, 10, 100);
+
+        return 0;
+    }
+
+Make sure your active solution configuration (:menuselection:`Build --> Configuration Manager`) is set to a debug build (usually called "Debug"). This should disable compiler optimizations so that viewing variables in the debugger can work reliably.
+
+Build your solution (:menuselection:`Build --> Build Solution`, or press *F7*).
+
+Now set a breakpoint on the source line that says
+
+.. code-block:: c++
+
+    Mat edges;
+
+To set the breakpoint, right-click on the source line and select :menuselection:`Breakpoints --> Insert Breakpoint` from the context menu.
+
+Launch the program in the debugger (:menuselection:`Debug --> Start Debugging`, or hit *F5*). When the breakpoint is hit, the program is paused and Visual Studio displays a yellow instruction pointer at the breakpoint:
+
+.. image:: images/breakpoint.png
+
+Now you can inspect the state of you program. For example, you can bring up the *Locals* window (:menuselection:`Debug --> Windows --> Locals`), which will show the names and values of the variables in the current scope:
+
+.. image:: images/vs_locals.png
+
+Note that the built-in *Locals* window will display text only. This is where the Image Watch plug-in comes in. Image Watch is like another *Locals* window, but with an image viewer built into it. To bring up Image Watch, select :menuselection:`View --> Other Windows --> Image Watch`. Like Visual Studio's *Locals* window, Image Watch can dock to the Visual Studio IDE. Also, Visual Studio will remember whether you had Image Watch open, and where it was located between debugging sessions. This means you only have to do this once--the next time you start debugging, Image Watch will be back where you left it. Here's what the docked Image Watch window looks like at our breakpoint:
+
+.. image:: images/toolwindow.jpg
+    :height: 320pt
+
+The radio button at the top left (*Locals/Watch*) selects what is shown in the *Image List* below: *Locals* lists all OpenCV image objects in the current scope (this list is automatically populated). *Watch* shows image expressions that have been pinned for continuous inspection (not described here, see `Image Watch documentation <http://go.microsoft.com/fwlink/?LinkId=285461>`_ for details). The image list shows basic information such as width, height, number of channels, and, if available, a thumbnail. In our example, the image list contains our two local image variables, *input* and *edges*.
+
+If an image has a thumbnail, left-clicking on that image will select it for detailed viewing in the *Image Viewer* on the right. The viewer lets you pan (drag mouse) and zoom (mouse wheel). It also displays the pixel coordinate and value at the current mouse position.
+
+.. image:: images/viewer.jpg
+    :height: 160pt
+
+Note that the second image in the list, *edges*, is shown as "invalid". This indicates that some data members of this image object have corrupt or invalid values (for example, a negative image width). This is expected at this point in the program, since the C++ constructor for *edges* has not run yet, and so its members have undefined values (in debug mode they are usually filled with "0xCD" bytes).
+
+From here you can single-step through your code (:menuselection:`Debug->Step Over`, or press *F10*) and watch the pixels change: if you step once, over the *Mat edges;* statement, the *edges* image will change from "invalid" to "empty", which means that it is now in a valid state (default constructed), even though it has not been initialized yet (using *cv::Mat::create()*, for example). If you make one more step over the *cv::Canny()* call, you will see a thumbnail of the edge image appear in the image list.
+
+Now assume you want to do a visual sanity check of the *cv::Canny()* implementation. Bring the *edges* image into the viewer by selecting it in the *Image List* and zoom into a region with a clearly defined edge:
+
+.. image:: images/edges_zoom.png
+    :height: 160pt
+
+Right-click on the *Image Viewer* to bring up the view context menu and enable :menuselection:`Link Views` (a check box next to the menu item indicates whether the option is enabled).
+
+.. image:: images/viewer_context_menu.png
+    :height: 120pt
+
+The :menuselection:`Link Views` feature keeps the view region fixed when flipping between images of the same size. To see how this works, select the input image from the image list--you should now see the corresponding zoomed-in region in the input image:
+
+.. image:: images/input_zoom.png
+    :height: 160pt
+
+You may also switch back and forth between viewing input and edges with your up/down cursor keys. That way you can easily  verify that the detected edges line up nicely with the data in the input image.
+
+More ...
+====================
+
+Image watch has a number of more advanced features, such as
+
+#. pinning images to a *Watch* list for inspection across scopes or between debugging sessions
+
+#. clamping, thresholding, or diff'ing images directly inside the Watch window
+
+#. comparing an in-memory image against a reference image from a file
+
+Please refer to the online `Image Watch Documentation <http://go.microsoft.com/fwlink/?LinkId=285461>`_ for details--you also can get to the documentation page by clicking on the *Help* link in the Image Watch window:
+
+.. image:: images/help_button.jpg
+    :height: 80pt
diff --git a/doc/tutorials/ios/hello/hello.rst b/doc/tutorials/ios/hello/hello.rst
index b2a37aa703..8435be54dc 100644
--- a/doc/tutorials/ios/hello/hello.rst
+++ b/doc/tutorials/ios/hello/hello.rst
@@ -73,4 +73,3 @@ Now we will learn how to write a simple Hello World Application in Xcode using O
 .. image:: images/output.png
      :alt: output
      :align: center
-
diff --git a/doc/tutorials/ios/image_manipulation/image_manipulation.rst b/doc/tutorials/ios/image_manipulation/image_manipulation.rst
index fd2d9c6e3f..c4cde19907 100644
--- a/doc/tutorials/ios/image_manipulation/image_manipulation.rst
+++ b/doc/tutorials/ios/image_manipulation/image_manipulation.rst
@@ -127,4 +127,4 @@ Check out an instance of running code with more Image Effects on `YouTube <http:
 
   <div align="center">
  <iframe width="560" height="350" src="http://www.youtube.com/embed/Ko3K_xdhJ1I" frameborder="0" allowfullscreen></iframe>
-  </div>
\ No newline at end of file
+  </div>
diff --git a/doc/tutorials/ml/introduction_to_svm/introduction_to_svm.rst b/doc/tutorials/ml/introduction_to_svm/introduction_to_svm.rst
index 051765def1..50f734803d 100644
--- a/doc/tutorials/ml/introduction_to_svm/introduction_to_svm.rst
+++ b/doc/tutorials/ml/introduction_to_svm/introduction_to_svm.rst
@@ -129,7 +129,7 @@ Explanation
 
 3. **Train the SVM**
 
-   We call the method `CvSVM::train <http://opencv.itseez.com/modules/ml/doc/support_vector_machines.html#cvsvm-train>`_ to build the SVM model.
+   We call the method `CvSVM::train <http://docs.opencv.org/modules/ml/doc/support_vector_machines.html#cvsvm-train>`_ to build the SVM model.
 
    .. code-block:: cpp
 
@@ -185,4 +185,3 @@ Results
 .. image:: images/result.png
   :alt: The seperated planes
   :align: center
-
diff --git a/doc/tutorials/ml/non_linear_svms/non_linear_svms.rst b/doc/tutorials/ml/non_linear_svms/non_linear_svms.rst
index 8fbcc563a7..57e0b1b6ea 100644
--- a/doc/tutorials/ml/non_linear_svms/non_linear_svms.rst
+++ b/doc/tutorials/ml/non_linear_svms/non_linear_svms.rst
@@ -1 +1,232 @@
-.. _nonLinearSvmS:
 
Support Vector Machines for Non-Linearly Separable Data
*******************************************************

Goal
====

In this tutorial you will learn how to:

.. container:: enumeratevisibleitemswithsquare

  + Define the optimization problem for SVMs when it is not possible to separate linearly the training data.

  + How to configure the parameters in :svms:`CvSVMParams <cvsvmparams>` to adapt your SVM for this class of problems.

Motivation
==========

Why is it interesting to extend the SVM optimation problem in order to handle non-linearly separable training data? Most of the applications in which SVMs are used in computer vision require a more powerful tool than a simple linear classifier. This stems from the fact that in these tasks **the training data can be rarely separated using an hyperplane**.

Consider one of these tasks, for example, face detection. The training data in this case is composed by a set of images that are faces and another set of images that are non-faces (*every other thing in the world except from faces*). This training data is too complex so as to find a representation of each sample (*feature vector*) that could make the whole set of faces linearly separable from the whole set of non-faces.

Extension of the Optimization Problem
=====================================

Remember that using SVMs we obtain a separating hyperplane. Therefore, since the training data is now non-linearly separable, we must admit that the hyperplane found will misclassify some of the samples. This *misclassification* is a new variable in the optimization that must be taken into account. The new model has to include both the old requirement of finding the hyperplane that gives the biggest margin and the new one of generalizing the training data correctly by not allowing too many classification errors.  

We start here from the formulation of the optimization problem of finding the hyperplane which maximizes the **margin** (this is explained in the :ref:`previous tutorial <introductiontosvms>`):

.. math::
  \min_{\beta, \beta_{0}} L(\beta) = \frac{1}{2}||\beta||^{2} \text{ subject to } y_{i}(\beta^{T} x_{i} + \beta_{0}) \geq 1 \text{ } \forall i

There are multiple ways in which this model can be modified so it takes into account the misclassification errors. For example, one could think of minimizing the same quantity plus a constant times the number of misclassification errors in the training data, i.e.:

.. math::
  \min ||\beta||^{2} + C \text{(\# misclassication errors)}

However, this one is not a very good solution since, among some other reasons, we do not distinguish between samples that are misclassified with a small distance to their appropriate decision region or samples that are not. Therefore, a better solution will take into account the *distance of the misclassified samples to their correct decision regions*, i.e.:

.. math::
  \min ||\beta||^{2} + C \text{(distance of misclassified samples to their correct regions)}

For each sample of the training data a new parameter :math:`\xi_{i}` is defined. Each one of these parameters contains the distance from its corresponding training sample to their correct decision region. The following picture shows non-linearly separable training data from two classes, a separating hyperplane and the distances to their correct regions of the samples that are misclassified.

.. image:: images/sample-errors-dist.png
   :alt: Samples misclassified and their distances to their correct regions
   :align: center 

.. note:: Only the distances of the samples that are misclassified are shown in the picture. The distances of the rest of the samples are zero since they lay already in their correct decision region.

The red and blue lines that appear on the picture are the margins to each one of the decision regions. It is very **important** to realize that each of the :math:`\xi_{i}` goes from a misclassified training sample to the margin of its appropriate region.

Finally, the new formulation for the optimization problem is:

.. math::
  \min_{\beta, \beta_{0}} L(\beta) = ||\beta||^{2} + C \sum_{i} {\xi_{i}} \text{ subject to } y_{i}(\beta^{T} x_{i} + \beta_{0}) \geq 1 - \xi_{i} \text{ and } \xi_{i} \geq 0 \text{ } \forall i 

How should the parameter C be chosen? It is obvious that the answer to this question depends on how the training data is distributed. Although there is no general answer, it is useful to take into account these rules:

.. container:: enumeratevisibleitemswithsquare

   * Large values of C give solutions with *less misclassification errors* but a *smaller margin*. Consider that in this case it is expensive to make misclassification errors. Since the aim of the optimization is to minimize the argument, few misclassifications errors are allowed.

   * Small values of C give solutions with *bigger margin* and *more classification errors*. In this case the minimization does not consider that much the term of the sum so it focuses more on finding a hyperplane with big margin.

Source Code
===========

You may also find the source code and these video file in the :file:`samples/cpp/tutorial_code/gpu/non_linear_svms/non_linear_svms` folder of the OpenCV source library or :download:`download it from here <../../../../samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp>`.

.. literalinclude:: ../../../../samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp
   :language: cpp
   :linenos:
   :tab-width: 4
   :lines: 1-11, 22-23, 26-

Explanation
===========

1. **Set up the training data**

  The training data of this exercise is formed by a set of labeled 2D-points that belong to one of two different classes. To make the exercise more appealing, the training data is generated randomly using a uniform probability density functions (PDFs).

  We have divided the generation of the training data into two main parts.

  In the first part we generate data for both classes that is linearly separable.

  .. code-block:: cpp

     // Generate random points for the class 1
     Mat trainClass = trainData.rowRange(0, nLinearSamples);
     // The x coordinate of the points is in [0, 0.4)
     Mat c = trainClass.colRange(0, 1);
     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(0.4 * WIDTH));
     // The y coordinate of the points is in [0, 1)
     c = trainClass.colRange(1,2);
     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));

     // Generate random points for the class 2
     trainClass = trainData.rowRange(2*NTRAINING_SAMPLES-nLinearSamples, 2*NTRAINING_SAMPLES);
     // The x coordinate of the points is in [0.6, 1]
     c = trainClass.colRange(0 , 1); 
     rng.fill(c, RNG::UNIFORM, Scalar(0.6*WIDTH), Scalar(WIDTH));
     // The y coordinate of the points is in [0, 1)
     c = trainClass.colRange(1,2);
     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));

  In the second part we create data for both classes that is non-linearly separable, data that overlaps.

  .. code-block:: cpp

     // Generate random points for the classes 1 and 2
     trainClass = trainData.rowRange(  nLinearSamples, 2*NTRAINING_SAMPLES-nLinearSamples);
     // The x coordinate of the points is in [0.4, 0.6)
     c = trainClass.colRange(0,1);
     rng.fill(c, RNG::UNIFORM, Scalar(0.4*WIDTH), Scalar(0.6*WIDTH)); 
     // The y coordinate of the points is in [0, 1)
     c = trainClass.colRange(1,2);
     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT)); 

2. **Set up SVM's parameters**

  .. seealso::

      In the previous tutorial :ref:`introductiontosvms` there is an explanation of the atributes of the class :svms:`CvSVMParams <cvsvmparams>` that we configure here before training the SVM.

  .. code-block:: cpp

     CvSVMParams params;
     params.svm_type    = SVM::C_SVC;
     params.C              = 0.1;
     params.kernel_type = SVM::LINEAR;
     params.term_crit   = TermCriteria(CV_TERMCRIT_ITER, (int)1e7, 1e-6);

  There are just two differences between the configuration we do here and the one that was done in the :ref:`previous tutorial <introductiontosvms>` that we use as reference.

  * *CvSVM::C_SVC*. We chose here a small value of this parameter in order not to punish too much the misclassification errors in the optimization. The idea of doing this stems from the will of obtaining a solution close to the one intuitively expected. However, we recommend to get a better insight of the problem by making adjustments to this parameter.

      .. note:: Here there are just very few points in the overlapping region between classes, giving a smaller value to **FRAC_LINEAR_SEP** the density of points can be incremented and the impact of the parameter **CvSVM::C_SVC** explored deeply.

  * *Termination Criteria of the algorithm*. The maximum number of iterations has to be increased considerably in order to solve correctly a problem with non-linearly separable training data. In particular, we have increased in five orders of magnitude this value.   

3. **Train the SVM**

  We call the method :svms:`CvSVM::train <cvsvm-train>` to build the SVM model. Watch out that the training process may take a quite long time. Have patiance when your run the program.

  .. code-block:: cpp

     CvSVM svm;
     svm.train(trainData, labels, Mat(), Mat(), params);

4. **Show the Decision Regions**

  The method :svms:`CvSVM::predict <cvsvm-predict>` is used to classify an input sample using a trained SVM. In this example we have used this method in order to color the space depending on the prediction done by the SVM. In other words, an image is traversed interpreting its pixels as points of the Cartesian plane. Each of the points is colored depending on the class predicted by the SVM; in dark green if it is the class with label 1 and in dark blue if it is the class with label 2.

  .. code-block:: cpp

     Vec3b green(0,100,0), blue (100,0,0);
     for (int i = 0; i < I.rows; ++i)
          for (int j = 0; j < I.cols; ++j)
          {
               Mat sampleMat = (Mat_<float>(1,2) << i, j);
               float response = svm.predict(sampleMat);

               if      (response == 1)    I.at<Vec3b>(j, i)  = green;
               else if (response == 2)    I.at<Vec3b>(j, i)  = blue;
          }

5. **Show the training data**

  The method :drawingFunc:`circle <circle>` is used to show the samples that compose the training data. The samples of the class labeled with 1 are shown in light green and in light blue the samples of the class labeled with 2.

  .. code-block:: cpp

     int thick = -1;
     int lineType = 8;
     float px, py;
     // Class 1
     for (int i = 0; i < NTRAINING_SAMPLES; ++i)
     {
          px = trainData.at<float>(i,0);
          py = trainData.at<float>(i,1);
          circle(I, Point( (int) px,  (int) py ), 3, Scalar(0, 255, 0), thick, lineType);
     }
     // Class 2
     for (int i = NTRAINING_SAMPLES; i <2*NTRAINING_SAMPLES; ++i)
     {
          px = trainData.at<float>(i,0);
          py = trainData.at<float>(i,1);
          circle(I, Point( (int) px, (int) py ), 3, Scalar(255, 0, 0), thick, lineType);
     }

6. **Support vectors**

  We use here a couple of methods to obtain information about the support vectors. The method :svms:`CvSVM::get_support_vector_count <cvsvm-get-support-vector>` outputs the total number of support vectors used in the problem and with the method :svms:`CvSVM::get_support_vector <cvsvm-get-support-vector>` we obtain each of the support vectors using an index. We have used this methods here to find the training examples that are support vectors and highlight them.

  .. code-block:: cpp

     thick = 2;
     lineType  = 8;
     int x     = svm.get_support_vector_count();

     for (int i = 0; i < x; ++i)
     {
          const float* v = svm.get_support_vector(i);
          circle(     I,  Point( (int) v[0], (int) v[1]), 6, Scalar(128, 128, 128), thick, lineType);
     }

Results
=======

.. container:: enumeratevisibleitemswithsquare

   * The code opens an image and shows the training examples of both classes. The points of one class are represented with light green and light blue ones are used for the other class.

   * The SVM is trained and used to classify all the pixels of the image. This results in a division of the image in a blue region and a green region. The boundary between both regions is the separating hyperplane. Since the training data is non-linearly separable, it can be seen that some of the examples of both classes are misclassified; some green points lay on the blue region and some blue points lay on the green one.

   * Finally the support vectors are shown using gray rings around the training examples.

.. image:: images/result.png
  :alt: Training data and decision regions given by the SVM
  :width: 300pt
  :align: center 

You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=vFv2yPcSo-Q>`_. 

.. raw:: html

  <div align="center">
  <iframe title="Support Vector Machines for Non-Linearly Separable Data" width="560" height="349" src="http://www.youtube.com/embed/vFv2yPcSo-Q?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
  </div>
\ No newline at end of file
+.. _nonLinearSvmS:
+
+Support Vector Machines for Non-Linearly Separable Data
+*******************************************************
+
+Goal
+====
+
+In this tutorial you will learn how to:
+
+.. container:: enumeratevisibleitemswithsquare
+
+  + Define the optimization problem for SVMs when it is not possible to separate linearly the training data.
+
+  + How to configure the parameters in :svms:`CvSVMParams <cvsvmparams>` to adapt your SVM for this class of problems.
+
+Motivation
+==========
+
+Why is it interesting to extend the SVM optimation problem in order to handle non-linearly separable training data? Most of the applications in which SVMs are used in computer vision require a more powerful tool than a simple linear classifier. This stems from the fact that in these tasks **the training data can be rarely separated using an hyperplane**.
+
+Consider one of these tasks, for example, face detection. The training data in this case is composed by a set of images that are faces and another set of images that are non-faces (*every other thing in the world except from faces*). This training data is too complex so as to find a representation of each sample (*feature vector*) that could make the whole set of faces linearly separable from the whole set of non-faces.
+
+Extension of the Optimization Problem
+=====================================
+
+Remember that using SVMs we obtain a separating hyperplane. Therefore, since the training data is now non-linearly separable, we must admit that the hyperplane found will misclassify some of the samples. This *misclassification* is a new variable in the optimization that must be taken into account. The new model has to include both the old requirement of finding the hyperplane that gives the biggest margin and the new one of generalizing the training data correctly by not allowing too many classification errors.
+
+We start here from the formulation of the optimization problem of finding the hyperplane which maximizes the **margin** (this is explained in the :ref:`previous tutorial <introductiontosvms>`):
+
+.. math::
+  \min_{\beta, \beta_{0}} L(\beta) = \frac{1}{2}||\beta||^{2} \text{ subject to } y_{i}(\beta^{T} x_{i} + \beta_{0}) \geq 1 \text{ } \forall i
+
+There are multiple ways in which this model can be modified so it takes into account the misclassification errors. For example, one could think of minimizing the same quantity plus a constant times the number of misclassification errors in the training data, i.e.:
+
+.. math::
+  \min ||\beta||^{2} + C \text{(\# misclassication errors)}
+
+However, this one is not a very good solution since, among some other reasons, we do not distinguish between samples that are misclassified with a small distance to their appropriate decision region or samples that are not. Therefore, a better solution will take into account the *distance of the misclassified samples to their correct decision regions*, i.e.:
+
+.. math::
+  \min ||\beta||^{2} + C \text{(distance of misclassified samples to their correct regions)}
+
+For each sample of the training data a new parameter :math:`\xi_{i}` is defined. Each one of these parameters contains the distance from its corresponding training sample to their correct decision region. The following picture shows non-linearly separable training data from two classes, a separating hyperplane and the distances to their correct regions of the samples that are misclassified.
+
+.. image:: images/sample-errors-dist.png
+   :alt: Samples misclassified and their distances to their correct regions
+   :align: center
+
+.. note:: Only the distances of the samples that are misclassified are shown in the picture. The distances of the rest of the samples are zero since they lay already in their correct decision region.
+
+The red and blue lines that appear on the picture are the margins to each one of the decision regions. It is very **important** to realize that each of the :math:`\xi_{i}` goes from a misclassified training sample to the margin of its appropriate region.
+
+Finally, the new formulation for the optimization problem is:
+
+.. math::
+  \min_{\beta, \beta_{0}} L(\beta) = ||\beta||^{2} + C \sum_{i} {\xi_{i}} \text{ subject to } y_{i}(\beta^{T} x_{i} + \beta_{0}) \geq 1 - \xi_{i} \text{ and } \xi_{i} \geq 0 \text{ } \forall i
+
+How should the parameter C be chosen? It is obvious that the answer to this question depends on how the training data is distributed. Although there is no general answer, it is useful to take into account these rules:
+
+.. container:: enumeratevisibleitemswithsquare
+
+   * Large values of C give solutions with *less misclassification errors* but a *smaller margin*. Consider that in this case it is expensive to make misclassification errors. Since the aim of the optimization is to minimize the argument, few misclassifications errors are allowed.
+
+   * Small values of C give solutions with *bigger margin* and *more classification errors*. In this case the minimization does not consider that much the term of the sum so it focuses more on finding a hyperplane with big margin.
+
+Source Code
+===========
+
+You may also find the source code and these video file in the :file:`samples/cpp/tutorial_code/gpu/non_linear_svms/non_linear_svms` folder of the OpenCV source library or :download:`download it from here <../../../../samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp>`.
+
+.. literalinclude:: ../../../../samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp
+   :language: cpp
+   :linenos:
+   :tab-width: 4
+   :lines: 1-11, 22-23, 26-
+
+Explanation
+===========
+
+1. **Set up the training data**
+
+  The training data of this exercise is formed by a set of labeled 2D-points that belong to one of two different classes. To make the exercise more appealing, the training data is generated randomly using a uniform probability density functions (PDFs).
+
+  We have divided the generation of the training data into two main parts.
+
+  In the first part we generate data for both classes that is linearly separable.
+
+  .. code-block:: cpp
+
+     // Generate random points for the class 1
+     Mat trainClass = trainData.rowRange(0, nLinearSamples);
+     // The x coordinate of the points is in [0, 0.4)
+     Mat c = trainClass.colRange(0, 1);
+     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(0.4 * WIDTH));
+     // The y coordinate of the points is in [0, 1)
+     c = trainClass.colRange(1,2);
+     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));
+
+     // Generate random points for the class 2
+     trainClass = trainData.rowRange(2*NTRAINING_SAMPLES-nLinearSamples, 2*NTRAINING_SAMPLES);
+     // The x coordinate of the points is in [0.6, 1]
+     c = trainClass.colRange(0 , 1);
+     rng.fill(c, RNG::UNIFORM, Scalar(0.6*WIDTH), Scalar(WIDTH));
+     // The y coordinate of the points is in [0, 1)
+     c = trainClass.colRange(1,2);
+     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));
+
+  In the second part we create data for both classes that is non-linearly separable, data that overlaps.
+
+  .. code-block:: cpp
+
+     // Generate random points for the classes 1 and 2
+     trainClass = trainData.rowRange(  nLinearSamples, 2*NTRAINING_SAMPLES-nLinearSamples);
+     // The x coordinate of the points is in [0.4, 0.6)
+     c = trainClass.colRange(0,1);
+     rng.fill(c, RNG::UNIFORM, Scalar(0.4*WIDTH), Scalar(0.6*WIDTH));
+     // The y coordinate of the points is in [0, 1)
+     c = trainClass.colRange(1,2);
+     rng.fill(c, RNG::UNIFORM, Scalar(1), Scalar(HEIGHT));
+
+2. **Set up SVM's parameters**
+
+  .. seealso::
+
+      In the previous tutorial :ref:`introductiontosvms` there is an explanation of the atributes of the class :svms:`CvSVMParams <cvsvmparams>` that we configure here before training the SVM.
+
+  .. code-block:: cpp
+
+     CvSVMParams params;
+     params.svm_type    = SVM::C_SVC;
+     params.C              = 0.1;
+     params.kernel_type = SVM::LINEAR;
+     params.term_crit   = TermCriteria(CV_TERMCRIT_ITER, (int)1e7, 1e-6);
+
+  There are just two differences between the configuration we do here and the one that was done in the :ref:`previous tutorial <introductiontosvms>` that we use as reference.
+
+  * *CvSVM::C_SVC*. We chose here a small value of this parameter in order not to punish too much the misclassification errors in the optimization. The idea of doing this stems from the will of obtaining a solution close to the one intuitively expected. However, we recommend to get a better insight of the problem by making adjustments to this parameter.
+
+      .. note:: Here there are just very few points in the overlapping region between classes, giving a smaller value to **FRAC_LINEAR_SEP** the density of points can be incremented and the impact of the parameter **CvSVM::C_SVC** explored deeply.
+
+  * *Termination Criteria of the algorithm*. The maximum number of iterations has to be increased considerably in order to solve correctly a problem with non-linearly separable training data. In particular, we have increased in five orders of magnitude this value.
+
+3. **Train the SVM**
+
+  We call the method :svms:`CvSVM::train <cvsvm-train>` to build the SVM model. Watch out that the training process may take a quite long time. Have patiance when your run the program.
+
+  .. code-block:: cpp
+
+     CvSVM svm;
+     svm.train(trainData, labels, Mat(), Mat(), params);
+
+4. **Show the Decision Regions**
+
+  The method :svms:`CvSVM::predict <cvsvm-predict>` is used to classify an input sample using a trained SVM. In this example we have used this method in order to color the space depending on the prediction done by the SVM. In other words, an image is traversed interpreting its pixels as points of the Cartesian plane. Each of the points is colored depending on the class predicted by the SVM; in dark green if it is the class with label 1 and in dark blue if it is the class with label 2.
+
+  .. code-block:: cpp
+
+     Vec3b green(0,100,0), blue (100,0,0);
+     for (int i = 0; i < I.rows; ++i)
+          for (int j = 0; j < I.cols; ++j)
+          {
+               Mat sampleMat = (Mat_<float>(1,2) << i, j);
+               float response = svm.predict(sampleMat);
+
+               if      (response == 1)    I.at<Vec3b>(j, i)  = green;
+               else if (response == 2)    I.at<Vec3b>(j, i)  = blue;
+          }
+
+5. **Show the training data**
+
+  The method :drawingFunc:`circle <circle>` is used to show the samples that compose the training data. The samples of the class labeled with 1 are shown in light green and in light blue the samples of the class labeled with 2.
+
+  .. code-block:: cpp
+
+     int thick = -1;
+     int lineType = 8;
+     float px, py;
+     // Class 1
+     for (int i = 0; i < NTRAINING_SAMPLES; ++i)
+     {
+          px = trainData.at<float>(i,0);
+          py = trainData.at<float>(i,1);
+          circle(I, Point( (int) px,  (int) py ), 3, Scalar(0, 255, 0), thick, lineType);
+     }
+     // Class 2
+     for (int i = NTRAINING_SAMPLES; i <2*NTRAINING_SAMPLES; ++i)
+     {
+          px = trainData.at<float>(i,0);
+          py = trainData.at<float>(i,1);
+          circle(I, Point( (int) px, (int) py ), 3, Scalar(255, 0, 0), thick, lineType);
+     }
+
+6. **Support vectors**
+
+  We use here a couple of methods to obtain information about the support vectors. The method :svms:`CvSVM::get_support_vector_count <cvsvm-get-support-vector>` outputs the total number of support vectors used in the problem and with the method :svms:`CvSVM::get_support_vector <cvsvm-get-support-vector>` we obtain each of the support vectors using an index. We have used this methods here to find the training examples that are support vectors and highlight them.
+
+  .. code-block:: cpp
+
+     thick = 2;
+     lineType  = 8;
+     int x     = svm.get_support_vector_count();
+
+     for (int i = 0; i < x; ++i)
+     {
+          const float* v = svm.get_support_vector(i);
+          circle(     I,  Point( (int) v[0], (int) v[1]), 6, Scalar(128, 128, 128), thick, lineType);
+     }
+
+Results
+========
+
+.. container:: enumeratevisibleitemswithsquare
+
+   * The code opens an image and shows the training examples of both classes. The points of one class are represented with light green and light blue ones are used for the other class.
+
+   * The SVM is trained and used to classify all the pixels of the image. This results in a division of the image in a blue region and a green region. The boundary between both regions is the separating hyperplane. Since the training data is non-linearly separable, it can be seen that some of the examples of both classes are misclassified; some green points lay on the blue region and some blue points lay on the green one.
+
+   * Finally the support vectors are shown using gray rings around the training examples.
+
+.. image:: images/result.png
+  :alt: Training data and decision regions given by the SVM
+  :width: 300pt
+  :align: center
+
+You may observe a runtime instance of this on the `YouTube here <https://www.youtube.com/watch?v=vFv2yPcSo-Q>`_.
+
+.. raw:: html
+
+  <div align="center">
+  <iframe title="Support Vector Machines for Non-Linearly Separable Data" width="560" height="349" src="http://www.youtube.com/embed/vFv2yPcSo-Q?rel=0&loop=1" frameborder="0" allowfullscreen align="middle"></iframe>
+  </div>
diff --git a/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.rst b/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.rst
index 146a0ec0be..ba9c2740e2 100644
--- a/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.rst
+++ b/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.rst
@@ -26,91 +26,90 @@ This tutorial code's is shown lines below. You can also download it from `here <
 
 .. code-block:: cpp
 
-   #include "opencv2/objdetect.hpp"
-   #include "opencv2/highgui.hpp"
-   #include "opencv2/imgproc.hpp"
+    #include "opencv2/objdetect.hpp"
+    #include "opencv2/highgui.hpp"
+    #include "opencv2/imgproc.hpp"
 
-   #include <iostream>
-   #include <stdio.h>
+    #include <iostream>
+    #include <stdio.h>
 
-   using namespace std;
-   using namespace cv;
+    using namespace std;
+    using namespace cv;
 
-   /** Function Headers */
-   void detectAndDisplay( Mat frame );
+    /** Function Headers */
+    void detectAndDisplay( Mat frame );
 
-   /** Global variables */
-   String face_cascade_name = "haarcascade_frontalface_alt.xml";
-   String eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
-   CascadeClassifier face_cascade;
-   CascadeClassifier eyes_cascade;
-   string window_name = "Capture - Face detection";
-   RNG rng(12345);
+    /** Global variables */
+    String face_cascade_name = "haarcascade_frontalface_alt.xml";
+    String eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
+    CascadeClassifier face_cascade;
+    CascadeClassifier eyes_cascade;
+    String window_name = "Capture - Face detection";
 
-   /** @function main */
-   int main( int argc, const char** argv )
-   {
-     CvCapture* capture;
-     Mat frame;
-
-     //-- 1. Load the cascades
-     if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
-     if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
-
-     //-- 2. Read the video stream
-     capture = cvCaptureFromCAM( -1 );
-     if( capture )
-     {
-       while( true )
-       {
-     frame = cvQueryFrame( capture );
-
-     //-- 3. Apply the classifier to the frame
-         if( !frame.empty() )
-         { detectAndDisplay( frame ); }
-         else
-         { printf(" --(!) No captured frame -- Break!"); break; }
-
-         int c = waitKey(10);
-         if( (char)c == 'c' ) { break; }
-        }
-     }
-     return 0;
-   }
-
-  /** @function detectAndDisplay */
-  void detectAndDisplay( Mat frame )
-  {
-    std::vector<Rect> faces;
-    Mat frame_gray;
-
-    cvtColor( frame, frame_gray, CV_BGR2GRAY );
-    equalizeHist( frame_gray, frame_gray );
-
-    //-- Detect faces
-    face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
-
-    for( int i = 0; i < faces.size(); i++ )
+    /** @function main */
+    int main( void )
     {
-      Point center( faces[i].x + faces[i].width*0.5, faces[i].y + faces[i].height*0.5 );
-      ellipse( frame, center, Size( faces[i].width*0.5, faces[i].height*0.5), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
+        VideoCapture capture;
+        Mat frame;
 
-      Mat faceROI = frame_gray( faces[i] );
-      std::vector<Rect> eyes;
+        //-- 1. Load the cascades
+        if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading face cascade\n"); return -1; };
+        if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading eyes cascade\n"); return -1; };
 
-      //-- In each face, detect eyes
-      eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CV_HAAR_SCALE_IMAGE, Size(30, 30) );
+        //-- 2. Read the video stream
+        capture.open( -1 );
+        if ( ! capture.isOpened() ) { printf("--(!)Error opening video capture\n"); return -1; }
 
-      for( int j = 0; j < eyes.size(); j++ )
-       {
-         Point center( faces[i].x + eyes[j].x + eyes[j].width*0.5, faces[i].y + eyes[j].y + eyes[j].height*0.5 );
-         int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
-         circle( frame, center, radius, Scalar( 255, 0, 0 ), 4, 8, 0 );
-       }
+        while (  capture.read(frame) )
+        {
+            if( frame.empty() )
+            {
+                printf(" --(!) No captured frame -- Break!");
+                break;
+            }
+
+            //-- 3. Apply the classifier to the frame
+            detectAndDisplay( frame );
+
+            int c = waitKey(10);
+            if( (char)c == 27 ) { break; } // escape
+        }
+        return 0;
+    }
+
+    /** @function detectAndDisplay */
+    void detectAndDisplay( Mat frame )
+    {
+        std::vector<Rect> faces;
+        Mat frame_gray;
+
+        cvtColor( frame, frame_gray, COLOR_BGR2GRAY );
+        equalizeHist( frame_gray, frame_gray );
+
+        //-- Detect faces
+        face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CASCADE_SCALE_IMAGE, Size(30, 30) );
+
+        for( size_t i = 0; i < faces.size(); i++ )
+        {
+            Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
+            ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
+
+            Mat faceROI = frame_gray( faces[i] );
+            std::vector<Rect> eyes;
+
+            //-- In each face, detect eyes
+            eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CASCADE_SCALE_IMAGE, Size(30, 30) );
+
+            for( size_t j = 0; j < eyes.size(); j++ )
+            {
+                Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
+                int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
+                circle( frame, eye_center, radius, Scalar( 255, 0, 0 ), 4, 8, 0 );
+            }
+        }
+        //-- Show what you got
+        imshow( window_name, frame );
     }
-    //-- Show what you got
-    imshow( window_name, frame );
-   }
 
 Explanation
 ============
@@ -131,4 +130,3 @@ Result
    .. image:: images/Cascade_Classifier_Tutorial_Result_LBP.jpg
       :align: center
       :height: 300pt
-
diff --git a/doc/tutorials/tutorials.rst b/doc/tutorials/tutorials.rst
index cbc51c1956..07bc515d53 100644
--- a/doc/tutorials/tutorials.rst
+++ b/doc/tutorials/tutorials.rst
@@ -156,17 +156,17 @@ As always, we would be happy to hear your comments and receive your contribution
                  :width:  80pt
                  :alt: gpu icon
 
-*  :ref:`Table-Of-Content-Contrib`
+*  :ref:`Table-Of-Content-Bioinspired`
 
    .. tabularcolumns:: m{100pt} m{300pt}
    .. cssclass:: toctableopencv
 
-   =========== =======================================================
-   |Contrib|       Discover additional contribution to OpenCV.
+   ============= =======================================================
+   |Bioinspired|       Algorithms inspired from biological models.
 
-   =========== =======================================================
+   ============= =======================================================
 
-   .. |Contrib| image:: images/retina.jpg
+   .. |Bioinspired| image:: images/retina.jpg
                  :height: 80pt
                  :width:  80pt
                  :alt: gpu icon
@@ -219,6 +219,6 @@ As always, we would be happy to hear your comments and receive your contribution
    objdetect/table_of_content_objdetect/table_of_content_objdetect
    ml/table_of_content_ml/table_of_content_ml
    gpu/table_of_content_gpu/table_of_content_gpu
-   contrib/table_of_content_contrib/table_of_content_contrib
+   bioinspired/table_of_content_bioinspired/table_of_content_bioinspired
    ios/table_of_content_ios/table_of_content_ios
    general/table_of_content_general/table_of_content_general
diff --git a/include/CMakeLists.txt b/include/CMakeLists.txt
index 1d5909616a..ed3b85a8fc 100644
--- a/include/CMakeLists.txt
+++ b/include/CMakeLists.txt
@@ -5,5 +5,3 @@ install(FILES ${old_hdrs}
 install(FILES "opencv2/opencv.hpp"
     DESTINATION ${OPENCV_INCLUDE_INSTALL_PATH}/opencv2
     COMPONENT main)
-
-
diff --git a/include/opencv/cv.h b/include/opencv/cv.h
index c878e68818..1ed020a356 100644
--- a/include/opencv/cv.h
+++ b/include/opencv/cv.h
@@ -61,23 +61,15 @@
 //CV_WARNING("This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module")
 
 #include "opencv2/core/core_c.h"
-#include "opencv2/core.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/video.hpp"
-#include "opencv2/features2d.hpp"
-#include "opencv2/flann.hpp"
-#include "opencv2/calib3d.hpp"
-#include "opencv2/objdetect.hpp"
+#include "opencv2/photo/photo_c.h"
+#include "opencv2/video/tracking_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
+#include "opencv2/legacy.hpp"
 #include "opencv2/legacy/compat.hpp"
 
 #if !defined(CV_IMPL)
 #define CV_IMPL extern "C"
 #endif //CV_IMPL
 
-#if defined(__cplusplus)
-#include "opencv2/core/internal.hpp"
-#endif //__cplusplus
-
 #endif // __OPENCV_OLD_CV_H_
-
diff --git a/include/opencv/cv.hpp b/include/opencv/cv.hpp
index 37b523b31c..e498d7ac1a 100644
--- a/include/opencv/cv.hpp
+++ b/include/opencv/cv.hpp
@@ -47,6 +47,14 @@
 //#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
 //#endif
 
-#include <cv.h>
+#include "cv.h"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/photo.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/objdetect.hpp"
 
 #endif
diff --git a/include/opencv/cvaux.h b/include/opencv/cvaux.h
index 52d4f0b648..cb49c086ba 100644
--- a/include/opencv/cvaux.h
+++ b/include/opencv/cvaux.h
@@ -47,17 +47,16 @@
 //#endif
 
 #include "opencv2/core/core_c.h"
-#include "opencv2/core.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/video.hpp"
-#include "opencv2/features2d.hpp"
-#include "opencv2/calib3d.hpp"
-#include "opencv2/objdetect.hpp"
+#include "opencv2/photo/photo_c.h"
+#include "opencv2/video/tracking_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
+#include "opencv2/contrib/compat.hpp"
+
 #include "opencv2/legacy.hpp"
 #include "opencv2/legacy/compat.hpp"
 #include "opencv2/legacy/blobtrack.hpp"
-#include "opencv2/contrib.hpp"
+
 
 #endif
 
diff --git a/include/opencv/cvaux.hpp b/include/opencv/cvaux.hpp
index 952210b0be..b0e60a3038 100644
--- a/include/opencv/cvaux.hpp
+++ b/include/opencv/cvaux.hpp
@@ -46,6 +46,7 @@
 //#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
 //#endif
 
-#include <cvaux.h>
+#include "cvaux.h"
+#include "opencv2/core/utility.hpp"
 
 #endif
diff --git a/include/opencv/cxcore.h b/include/opencv/cxcore.h
index 9be80ae96f..0982bd7508 100644
--- a/include/opencv/cxcore.h
+++ b/include/opencv/cxcore.h
@@ -48,6 +48,5 @@
 //#endif
 
 #include "opencv2/core/core_c.h"
-#include "opencv2/core.hpp"
 
 #endif
diff --git a/include/opencv/cxcore.hpp b/include/opencv/cxcore.hpp
index 033b36556c..9af4ac7460 100644
--- a/include/opencv/cxcore.hpp
+++ b/include/opencv/cxcore.hpp
@@ -47,6 +47,7 @@
 //#warning "This is a deprecated opencv header provided for compatibility. Please include a header from a corresponding opencv module"
 //#endif
 
-#include <cxcore.h>
+#include "cxcore.h"
+#include "opencv2/core.hpp"
 
 #endif
diff --git a/include/opencv/cxeigen.hpp b/include/opencv/cxeigen.hpp
index c503f713d6..1f04d1a3a2 100644
--- a/include/opencv/cxeigen.hpp
+++ b/include/opencv/cxeigen.hpp
@@ -46,4 +46,3 @@
 #include "opencv2/core/eigen.hpp"
 
 #endif
-
diff --git a/include/opencv/cxmisc.h b/include/opencv/cxmisc.h
index 6446944643..6c93a0cce4 100644
--- a/include/opencv/cxmisc.h
+++ b/include/opencv/cxmisc.h
@@ -1,6 +1,8 @@
 #ifndef __OPENCV_OLD_CXMISC_H__
 #define __OPENCV_OLD_CXMISC_H__
 
-#include "opencv2/core/internal.hpp"
+#ifdef __cplusplus
+#  include "opencv2/core/utility.hpp"
+#endif
 
 #endif
diff --git a/include/opencv/highgui.h b/include/opencv/highgui.h
index 56874e7383..0261029c09 100644
--- a/include/opencv/highgui.h
+++ b/include/opencv/highgui.h
@@ -43,8 +43,6 @@
 #define __OPENCV_OLD_HIGHGUI_H__
 
 #include "opencv2/core/core_c.h"
-#include "opencv2/core.hpp"
 #include "opencv2/highgui/highgui_c.h"
-#include "opencv2/highgui.hpp"
 
 #endif
diff --git a/include/opencv/ml.h b/include/opencv/ml.h
index ac21f3c219..d8e967f81b 100644
--- a/include/opencv/ml.h
+++ b/include/opencv/ml.h
@@ -42,7 +42,6 @@
 #define __OPENCV_OLD_ML_H__
 
 #include "opencv2/core/core_c.h"
-#include "opencv2/core.hpp"
 #include "opencv2/ml.hpp"
 
 #endif
diff --git a/include/opencv2/opencv.hpp b/include/opencv2/opencv.hpp
index f9c15df3d4..3b96bdd368 100644
--- a/include/opencv2/opencv.hpp
+++ b/include/opencv2/opencv.hpp
@@ -43,19 +43,16 @@
 #ifndef __OPENCV_ALL_HPP__
 #define __OPENCV_ALL_HPP__
 
-#include "opencv2/core/core_c.h"
 #include "opencv2/core.hpp"
-#include "opencv2/flann/miniflann.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/photo.hpp"
 #include "opencv2/video.hpp"
 #include "opencv2/features2d.hpp"
 #include "opencv2/objdetect.hpp"
 #include "opencv2/calib3d.hpp"
-#include "opencv2/ml.hpp"
-#include "opencv2/highgui/highgui_c.h"
 #include "opencv2/highgui.hpp"
 #include "opencv2/contrib.hpp"
+#include "opencv2/bioinspired.hpp"
+#include "opencv2/ml.hpp"
 
 #endif
diff --git a/index.rst b/index.rst
index 909bf908b8..5f50b66d0f 100644
--- a/index.rst
+++ b/index.rst
@@ -10,7 +10,7 @@ Welcome to opencv documentation!
    :maxdepth: 2
 
    modules/refman.rst
-   android/refman.rst
+   platforms/android/refman.rst
    doc/user_guide/user_guide.rst
    doc/tutorials/tutorials.rst
 
diff --git a/ios/configure-device_xcode.sh b/ios/configure-device_xcode.sh
deleted file mode 100755
index 8c28a3e909..0000000000
--- a/ios/configure-device_xcode.sh
+++ /dev/null
@@ -1 +0,0 @@
-cmake -GXcode -DCMAKE_TOOLCHAIN_FILE=../opencv/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake -DCMAKE_INSTALL_PREFIX=../OpenCV_iPhoneOS ../opencv 
diff --git a/ios/configure-simulator_xcode.sh b/ios/configure-simulator_xcode.sh
deleted file mode 100755
index 50e00261db..0000000000
--- a/ios/configure-simulator_xcode.sh
+++ /dev/null
@@ -1 +0,0 @@
-cmake -GXcode -DCMAKE_TOOLCHAIN_FILE=../opencv/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake -DCMAKE_INSTALL_PREFIX=../OpenCV_iPhoneSimulator ../opencv 
diff --git a/ios/readme.txt b/ios/readme.txt
deleted file mode 100644
index 1441b241b7..0000000000
--- a/ios/readme.txt
+++ /dev/null
@@ -1,15 +0,0 @@
-Assuming that your build directory is on the same level that opencv source,
-From the build directory run
-  ../opencv/ios/configure-device_xcode.sh
-or
-  ../opencv/ios/configure-simulator_xcode.sh
-
-Then from the same folder invoke
-
-xcodebuild -sdk iphoneos -configuration Release -target ALL_BUILD
-xcodebuild -sdk iphoneos -configuration Release -target install install
-
-or
-
-xcodebuild -sdk iphonesimulator -configuration Release -target ALL_BUILD
-xcodebuild -sdk iphonesimulator -configuration Release -target install install
\ No newline at end of file
diff --git a/modules/CMakeLists.txt b/modules/CMakeLists.txt
index c865eb01ba..400b2a8381 100644
--- a/modules/CMakeLists.txt
+++ b/modules/CMakeLists.txt
@@ -4,4 +4,4 @@ if(NOT OPENCV_MODULES_PATH)
   set(OPENCV_MODULES_PATH "${CMAKE_CURRENT_SOURCE_DIR}")
 endif()
 
-ocv_glob_modules(${OPENCV_MODULES_PATH})
+ocv_glob_modules(${OPENCV_MODULES_PATH} ${OPENCV_EXTRA_MODULES_PATH})
diff --git a/modules/androidcamera/CMakeLists.txt b/modules/androidcamera/CMakeLists.txt
index d54dd5d208..8ac8ced88e 100644
--- a/modules/androidcamera/CMakeLists.txt
+++ b/modules/androidcamera/CMakeLists.txt
@@ -6,7 +6,7 @@ set(the_description "Auxiliary module for Android native camera support")
 set(OPENCV_MODULE_TYPE STATIC)
 
 ocv_define_module(androidcamera INTERNAL opencv_core log dl)
-ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}/camera_wrapper" "${OpenCV_SOURCE_DIR}/android/service/engine/jni/include")
+ocv_include_directories("${CMAKE_CURRENT_SOURCE_DIR}/camera_wrapper" "${OpenCV_SOURCE_DIR}/platforms/android/service/engine/jni/include")
 
 # Android source tree for native camera
 SET (ANDROID_SOURCE_TREE "ANDROID_SOURCE_TREE-NOTFOUND" CACHE PATH
diff --git a/modules/androidcamera/camera_wrapper/camera_wrapper.cpp b/modules/androidcamera/camera_wrapper/camera_wrapper.cpp
index f6ec2f09cf..ca631fc21e 100644
--- a/modules/androidcamera/camera_wrapper/camera_wrapper.cpp
+++ b/modules/androidcamera/camera_wrapper/camera_wrapper.cpp
@@ -1,5 +1,8 @@
-#if !defined(ANDROID_r2_2_0) && !defined(ANDROID_r2_3_3) && !defined(ANDROID_r3_0_1) && !defined(ANDROID_r4_0_0) && !defined(ANDROID_r4_0_3) && !defined(ANDROID_r4_1_1) && !defined(ANDROID_r4_2_0)
-# error Building camera wrapper for your version of Android is not supported by OpenCV. You need to modify OpenCV sources in order to compile camera wrapper for your version of Android.
+#if !defined(ANDROID_r2_2_0) && !defined(ANDROID_r2_3_3) && !defined(ANDROID_r3_0_1) && \
+ !defined(ANDROID_r4_0_0) && !defined(ANDROID_r4_0_3) && !defined(ANDROID_r4_1_1) && \
+ !defined(ANDROID_r4_2_0) && !defined(ANDROID_r4_3_0)
+# error Building camera wrapper for your version of Android is not supported by OpenCV.\
+ You need to modify OpenCV sources in order to compile camera wrapper for your version of Android.
 #endif
 
 #include <camera/Camera.h>
@@ -16,17 +19,18 @@
 //Include SurfaceTexture.h file with the SurfaceTexture class
 # include <gui/SurfaceTexture.h>
 # define MAGIC_OPENCV_TEXTURE_ID (0x10)
-#else // defined(ANDROID_r3_0_1) || defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3)
-//TODO: This is either 2.2 or 2.3. Include the headers for ISurface.h access
-#if defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0)
-#include <gui/ISurface.h>
-#include <gui/BufferQueue.h>
+#elif defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0)
+# include <gui/ISurface.h>
+# include <gui/BufferQueue.h>
+#elif defined(ANDROID_r4_3_0)
+# include <gui/IGraphicBufferProducer.h>
+# include <gui/BufferQueue.h>
 #else
 # include <surfaceflinger/ISurface.h>
-#endif  // defined(ANDROID_r4_1_1)
-#endif  // defined(ANDROID_r3_0_1) || defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3)
+#endif
 
 #include <string>
+#include <fstream>
 
 //undef logging macro from /system/core/libcutils/loghack.h
 #ifdef LOGD
@@ -45,7 +49,6 @@
 # undef LOGE
 #endif
 
-
 // LOGGING
 #include <android/log.h>
 #define CAMERA_LOG_TAG "OpenCV_NativeCamera"
@@ -60,7 +63,7 @@ using namespace android;
 
 void debugShowFPS();
 
-#if defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0)
+#if defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
 class ConsumerListenerStub: public BufferQueue::ConsumerListener
 {
 public:
@@ -73,6 +76,29 @@ public:
 };
 #endif
 
+std::string getProcessName()
+{
+    std::string result;
+    std::ifstream f;
+
+    f.open("/proc/self/cmdline");
+    if (f.is_open())
+    {
+        std::string fullPath;
+        std::getline(f, fullPath, '\0');
+        if (!fullPath.empty())
+        {
+            int i = fullPath.size()-1;
+            while ((i >= 0) && (fullPath[i] != '/')) i--;
+            result = fullPath.substr(i+1, std::string::npos);
+        }
+    }
+
+    f.close();
+
+    return result;
+}
+
 void debugShowFPS()
 {
     static int mFrameCount = 0;
@@ -280,7 +306,7 @@ public:
     }
 
     virtual void postData(int32_t msgType, const sp<IMemory>& dataPtr
-    #if defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3) || defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0)
+    #if defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3) || defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
                           ,camera_frame_metadata_t*
 #endif
                           )
@@ -361,6 +387,11 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
     typedef sp<Camera> (*Android22ConnectFuncType)();
     typedef sp<Camera> (*Android23ConnectFuncType)(int);
     typedef sp<Camera> (*Android3DConnectFuncType)(int, int);
+    typedef sp<Camera> (*Android43ConnectFuncType)(int, const String16&, int);
+
+    const int ANY_CAMERA_INDEX = -1;
+    const int BACK_CAMERA_INDEX = 99;
+    const int FRONT_CAMERA_INDEX = 98;
 
     enum {
     CAMERA_SUPPORT_MODE_2D = 0x01, /* Camera Sensor supports 2D mode. */
@@ -369,11 +400,65 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
     CAMERA_SUPPORT_MODE_ZSL = 0x08 /* Camera Sensor supports ZSL mode. */
     };
 
+    // used for Android 4.3
+    enum {
+        USE_CALLING_UID = -1
+    };
+
     const char Android22ConnectName[] = "_ZN7android6Camera7connectEv";
     const char Android23ConnectName[] = "_ZN7android6Camera7connectEi";
     const char Android3DConnectName[] = "_ZN7android6Camera7connectEii";
+    const char Android43ConnectName[] = "_ZN7android6Camera7connectEiRKNS_8String16Ei";
 
-    LOGD("CameraHandler::initCameraConnect(%p, %d, %p, %p)", callback, cameraId, userData, prevCameraParameters);
+    int localCameraIndex = cameraId;
+
+    if (cameraId == ANY_CAMERA_INDEX)
+    {
+        localCameraIndex = 0;
+    }
+#if !defined(ANDROID_r2_2_0)
+    else if (cameraId == BACK_CAMERA_INDEX)
+    {
+        LOGD("Back camera selected");
+        for (int i = 0; i < Camera::getNumberOfCameras(); i++)
+        {
+            CameraInfo info;
+            Camera::getCameraInfo(i, &info);
+            if (info.facing == CAMERA_FACING_BACK)
+            {
+                localCameraIndex = i;
+                break;
+            }
+        }
+    }
+    else if (cameraId == FRONT_CAMERA_INDEX)
+    {
+        LOGD("Front camera selected");
+        for (int i = 0; i < Camera::getNumberOfCameras(); i++)
+        {
+            CameraInfo info;
+            Camera::getCameraInfo(i, &info);
+            if (info.facing == CAMERA_FACING_FRONT)
+            {
+                localCameraIndex = i;
+                break;
+            }
+        }
+    }
+
+    if (localCameraIndex == BACK_CAMERA_INDEX)
+    {
+        LOGE("Back camera not found!");
+        return NULL;
+    }
+    else if (localCameraIndex == FRONT_CAMERA_INDEX)
+    {
+        LOGE("Front camera not found!");
+        return NULL;
+    }
+#endif
+
+    LOGD("CameraHandler::initCameraConnect(%p, %d, %p, %p)", callback, localCameraIndex, userData, prevCameraParameters);
 
     sp<Camera> camera = 0;
 
@@ -381,8 +466,8 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
 
     if (!CameraHALHandle)
     {
-    LOGE("Cannot link to \"libcamera_client.so\"");
-    return NULL;
+        LOGE("Cannot link to \"libcamera_client.so\"");
+        return NULL;
     }
 
     // reset errors
@@ -390,24 +475,30 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
 
     if (Android22ConnectFuncType Android22Connect = (Android22ConnectFuncType)dlsym(CameraHALHandle, Android22ConnectName))
     {
-    LOGD("Connecting to CameraService v 2.2");
-    camera = Android22Connect();
+        LOGD("Connecting to CameraService v 2.2");
+        camera = Android22Connect();
     }
     else if (Android23ConnectFuncType Android23Connect = (Android23ConnectFuncType)dlsym(CameraHALHandle, Android23ConnectName))
     {
-    LOGD("Connecting to CameraService v 2.3");
-    camera = Android23Connect(cameraId);
+        LOGD("Connecting to CameraService v 2.3");
+        camera = Android23Connect(localCameraIndex);
     }
     else if (Android3DConnectFuncType Android3DConnect = (Android3DConnectFuncType)dlsym(CameraHALHandle, Android3DConnectName))
     {
-    LOGD("Connecting to CameraService v 3D");
-    camera = Android3DConnect(cameraId, CAMERA_SUPPORT_MODE_2D);
+        LOGD("Connecting to CameraService v 3D");
+        camera = Android3DConnect(localCameraIndex, CAMERA_SUPPORT_MODE_2D);
+    }
+    else if (Android43ConnectFuncType Android43Connect = (Android43ConnectFuncType)dlsym(CameraHALHandle, Android43ConnectName))
+    {
+        std::string currentProcName = getProcessName();
+        LOGD("Current process name for camera init: %s", currentProcName.c_str());
+        camera = Android43Connect(localCameraIndex, String16(currentProcName.c_str()), USE_CALLING_UID);
     }
     else
     {
-    dlclose(CameraHALHandle);
-    LOGE("Cannot connect to CameraService. Connect method was not found!");
-    return NULL;
+        dlclose(CameraHALHandle);
+        LOGE("Cannot connect to CameraService. Connect method was not found!");
+        return NULL;
     }
 
     dlclose(CameraHALHandle);
@@ -422,9 +513,9 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
     camera->setListener(handler);
 
     handler->camera = camera;
-    handler->cameraId = cameraId;
+    handler->cameraId = localCameraIndex;
 
-    if (prevCameraParameters != 0)
+    if (prevCameraParameters != NULL)
     {
         LOGI("initCameraConnect: Setting paramers from previous camera handler");
         camera->setParameters(prevCameraParameters->flatten());
@@ -456,11 +547,11 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
         const char* available_focus_modes = handler->params.get(CameraParameters::KEY_SUPPORTED_FOCUS_MODES);
         if (available_focus_modes != 0)
         {
-        if (strstr(available_focus_modes, "continuous-video") != NULL)
-        {
-        handler->params.set(CameraParameters::KEY_FOCUS_MODE, CameraParameters::FOCUS_MODE_CONTINUOUS_VIDEO);
+            if (strstr(available_focus_modes, "continuous-video") != NULL)
+            {
+                handler->params.set(CameraParameters::KEY_FOCUS_MODE, CameraParameters::FOCUS_MODE_CONTINUOUS_VIDEO);
 
-        status_t resParams = handler->camera->setParameters(handler->params.flatten());
+                status_t resParams = handler->camera->setParameters(handler->params.flatten());
 
                 if (resParams != 0)
                 {
@@ -470,8 +561,8 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
                 {
                     LOGD("initCameraConnect: autofocus is set to mode \"continuous-video\"");
                 }
+            }
         }
-    }
 #endif
 
         //check if yuv420sp format available. Set this format as preview format.
@@ -513,26 +604,25 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
         }
     }
 
-    status_t pdstatus;
+    status_t bufferStatus;
 #if defined(ANDROID_r2_2_0)
-    pdstatus = camera->setPreviewDisplay(sp<ISurface>(0 /*new DummySurface*/));
-    if (pdstatus != 0)
-        LOGE("initCameraConnect: failed setPreviewDisplay(0) call; camera migth not work correctly on some devices");
+    bufferStatus = camera->setPreviewDisplay(sp<ISurface>(0 /*new DummySurface*/));
+    if (bufferStatus != 0)
+        LOGE("initCameraConnect: failed setPreviewDisplay(0) call (status %d); camera might not work correctly on some devices", bufferStatus);
 #elif defined(ANDROID_r2_3_3)
     /* Do nothing in case of 2.3 for now */
-
 #elif defined(ANDROID_r3_0_1) || defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3)
     sp<SurfaceTexture> surfaceTexture = new SurfaceTexture(MAGIC_OPENCV_TEXTURE_ID);
-    pdstatus = camera->setPreviewTexture(surfaceTexture);
-    if (pdstatus != 0)
-        LOGE("initCameraConnect: failed setPreviewTexture call; camera migth not work correctly");
-#elif defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0)
+    bufferStatus = camera->setPreviewTexture(surfaceTexture);
+    if (bufferStatus != 0)
+        LOGE("initCameraConnect: failed setPreviewTexture call (status %d); camera might not work correctly", bufferStatus);
+#elif defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
     sp<BufferQueue> bufferQueue = new BufferQueue();
     sp<BufferQueue::ConsumerListener> queueListener = new ConsumerListenerStub();
     bufferQueue->consumerConnect(queueListener);
-    pdstatus = camera->setPreviewTexture(bufferQueue);
-    if (pdstatus != 0)
-    LOGE("initCameraConnect: failed setPreviewTexture call; camera migth not work correctly");
+    bufferStatus = camera->setPreviewTexture(bufferQueue);
+    if (bufferStatus != 0)
+        LOGE("initCameraConnect: failed setPreviewTexture call; camera might not work correctly");
 #endif
 
 #if (defined(ANDROID_r2_2_0) || defined(ANDROID_r2_3_3) || defined(ANDROID_r3_0_1))
@@ -548,9 +638,9 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
 #endif //!(defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3))
 
     LOGD("Starting preview");
-    status_t resStart = camera->startPreview();
+    status_t previewStatus = camera->startPreview();
 
-    if (resStart != 0)
+    if (previewStatus != 0)
     {
         LOGE("initCameraConnect: startPreview() fails. Closing camera connection...");
         handler->closeCameraConnect();
@@ -558,7 +648,7 @@ CameraHandler* CameraHandler::initCameraConnect(const CameraCallback& callback,
     }
     else
     {
-    LOGD("Preview started successfully");
+        LOGD("Preview started successfully");
     }
 
     return handler;
@@ -573,9 +663,11 @@ void CameraHandler::closeCameraConnect()
     }
 
     camera->stopPreview();
+#if defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3) || defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
+    camera->setPreviewCallbackFlags(CAMERA_FRAME_CALLBACK_FLAG_NOOP);
+#endif
     camera->disconnect();
     camera.clear();
-
     camera=NULL;
     // ATTENTION!!!!!!!!!!!!!!!!!!!!!!!!!!
     // When we set
@@ -816,14 +908,60 @@ void CameraHandler::applyProperties(CameraHandler** ppcameraHandler)
 
     if (*ppcameraHandler == 0)
     {
-        LOGE("applyProperties: Passed null *ppcameraHandler");
+        LOGE("applyProperties: Passed NULL *ppcameraHandler");
         return;
     }
 
-    LOGD("CameraHandler::applyProperties()");
-    CameraHandler* previousCameraHandler=*ppcameraHandler;
-    CameraParameters curCameraParameters(previousCameraHandler->params.flatten());
+    CameraParameters curCameraParameters((*ppcameraHandler)->params.flatten());
 
+#if defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3) || defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
+    CameraHandler* handler=*ppcameraHandler;
+
+    handler->camera->stopPreview();
+    handler->camera->setPreviewCallbackFlags(CAMERA_FRAME_CALLBACK_FLAG_NOOP);
+
+    status_t reconnectStatus = handler->camera->reconnect();
+    if (reconnectStatus != 0)
+    {
+        LOGE("applyProperties: failed to reconnect camera (status %d)", reconnectStatus);
+        return;
+    }
+
+    handler->camera->setParameters(curCameraParameters.flatten());
+    handler->params.unflatten(curCameraParameters.flatten());
+
+    status_t bufferStatus;
+# if defined(ANDROID_r4_0_0) || defined(ANDROID_r4_0_3)
+    sp<SurfaceTexture> surfaceTexture = new SurfaceTexture(MAGIC_OPENCV_TEXTURE_ID);
+    bufferStatus = handler->camera->setPreviewTexture(surfaceTexture);
+    if (bufferStatus != 0)
+        LOGE("applyProperties: failed setPreviewTexture call (status %d); camera might not work correctly", bufferStatus);
+# elif defined(ANDROID_r4_1_1) || defined(ANDROID_r4_2_0) || defined(ANDROID_r4_3_0)
+    sp<BufferQueue> bufferQueue = new BufferQueue();
+    sp<BufferQueue::ConsumerListener> queueListener = new ConsumerListenerStub();
+    bufferQueue->consumerConnect(queueListener);
+    bufferStatus = handler->camera->setPreviewTexture(bufferQueue);
+    if (bufferStatus != 0)
+        LOGE("applyProperties: failed setPreviewTexture call; camera might not work correctly");
+# endif
+
+    handler->camera->setPreviewCallbackFlags( CAMERA_FRAME_CALLBACK_FLAG_ENABLE_MASK | CAMERA_FRAME_CALLBACK_FLAG_COPY_OUT_MASK);//with copy
+
+    LOGD("Starting preview");
+    status_t previewStatus = handler->camera->startPreview();
+
+    if (previewStatus != 0)
+    {
+        LOGE("initCameraConnect: startPreview() fails. Closing camera connection...");
+        handler->closeCameraConnect();
+        handler = NULL;
+    }
+    else
+    {
+        LOGD("Preview started successfully");
+    }
+#else
+    CameraHandler* previousCameraHandler=*ppcameraHandler;
     CameraCallback cameraCallback=previousCameraHandler->cameraCallback;
     void* userData=previousCameraHandler->userData;
     int cameraId=previousCameraHandler->cameraId;
@@ -832,7 +970,6 @@ void CameraHandler::applyProperties(CameraHandler** ppcameraHandler)
     previousCameraHandler->closeCameraConnect();
     LOGD("CameraHandler::applyProperties(): after previousCameraHandler->closeCameraConnect");
 
-
     LOGD("CameraHandler::applyProperties(): before initCameraConnect");
     CameraHandler* handler=initCameraConnect(cameraCallback, cameraId, userData, &curCameraParameters);
     LOGD("CameraHandler::applyProperties(): after initCameraConnect, handler=0x%x", (int)handler);
@@ -845,6 +982,7 @@ void CameraHandler::applyProperties(CameraHandler** ppcameraHandler)
         }
     }
     (*ppcameraHandler)=handler;
+#endif
 }
 
 
diff --git a/modules/androidcamera/camera_wrapper/camera_wrapper.h b/modules/androidcamera/camera_wrapper/camera_wrapper.h
index 88c9c4100b..430dea2dd5 100644
--- a/modules/androidcamera/camera_wrapper/camera_wrapper.h
+++ b/modules/androidcamera/camera_wrapper/camera_wrapper.h
@@ -14,4 +14,3 @@ double getCameraPropertyC(void* camera, int propIdx);
 void setCameraPropertyC(void* camera, int propIdx, double value);
 void applyCameraPropertiesC(void** camera);
 }
-
diff --git a/modules/androidcamera/include/camera_activity.hpp b/modules/androidcamera/include/camera_activity.hpp
index 8aa25b620e..2af7befe30 100644
--- a/modules/androidcamera/include/camera_activity.hpp
+++ b/modules/androidcamera/include/camera_activity.hpp
@@ -44,4 +44,4 @@ private:
     int frameHeight;
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/androidcamera/src/camera_activity.cpp b/modules/androidcamera/src/camera_activity.cpp
index 9ea54d78be..a265d2e8d1 100644
--- a/modules/androidcamera/src/camera_activity.cpp
+++ b/modules/androidcamera/src/camera_activity.cpp
@@ -4,7 +4,6 @@
 #include <dirent.h>
 #include <android/log.h>
 #include <cctype>
-#include <string>
 #include <vector>
 #include <algorithm>
 #include <opencv2/core/version.hpp>
@@ -12,6 +11,8 @@
 #include "camera_wrapper.h"
 #include "EngineCommon.h"
 
+#include "opencv2/core.hpp"
+
 #undef LOG_TAG
 #undef LOGE
 #undef LOGD
@@ -28,6 +29,11 @@
 #include <sys/types.h>
 #include <dirent.h>
 
+struct str_greater
+{
+    bool operator() (const cv::String& a, const cv::String& b) { return a > b; }
+};
+
 class CameraWrapperConnector
 {
 public:
@@ -37,17 +43,17 @@ public:
     static CameraActivity::ErrorCode getProperty(void* camera, int propIdx, double* value);
     static CameraActivity::ErrorCode applyProperties(void** ppcamera);
 
-    static void setPathLibFolder(const std::string& path);
+    static void setPathLibFolder(const cv::String& path);
 
 private:
-    static std::string pathLibFolder;
+    static cv::String pathLibFolder;
     static bool isConnectedToLib;
 
-    static std::string getPathLibFolder();
-    static std::string getDefaultPathLibFolder();
+    static cv::String getPathLibFolder();
+    static cv::String getDefaultPathLibFolder();
     static CameraActivity::ErrorCode connectToLib();
     static CameraActivity::ErrorCode getSymbolFromLib(void * libHandle, const char* symbolName, void** ppSymbol);
-    static void fillListWrapperLibs(const std::string& folderPath, std::vector<std::string>& listLibs);
+    static void fillListWrapperLibs(const cv::String& folderPath, std::vector<cv::String>& listLibs);
 
     static InitCameraConnectC pInitCameraC;
     static CloseCameraConnectC pCloseCameraC;
@@ -58,7 +64,7 @@ private:
     friend bool nextFrame(void* buffer, size_t bufferSize, void* userData);
 };
 
-std::string CameraWrapperConnector::pathLibFolder;
+cv::String CameraWrapperConnector::pathLibFolder;
 
 bool CameraWrapperConnector::isConnectedToLib = false;
 InitCameraConnectC  CameraWrapperConnector::pInitCameraC = 0;
@@ -165,7 +171,7 @@ CameraActivity::ErrorCode CameraWrapperConnector::connectToLib()
     }
 
     dlerror();
-    std::string folderPath = getPathLibFolder();
+    cv::String folderPath = getPathLibFolder();
     if (folderPath.empty())
     {
         LOGD("Trying to find native camera in default OpenCV packages");
@@ -174,12 +180,12 @@ CameraActivity::ErrorCode CameraWrapperConnector::connectToLib()
 
     LOGD("CameraWrapperConnector::connectToLib: folderPath=%s", folderPath.c_str());
 
-    std::vector<std::string> listLibs;
+    std::vector<cv::String> listLibs;
     fillListWrapperLibs(folderPath, listLibs);
-    std::sort(listLibs.begin(), listLibs.end(), std::greater<std::string>());
+    std::sort(listLibs.begin(), listLibs.end(), str_greater());
 
     void * libHandle=0;
-    std::string cur_path;
+    cv::String cur_path;
     for(size_t i = 0; i < listLibs.size(); i++) {
         cur_path=folderPath + listLibs[i];
         LOGD("try to load library '%s'", listLibs[i].c_str());
@@ -245,7 +251,7 @@ CameraActivity::ErrorCode CameraWrapperConnector::getSymbolFromLib(void* libHand
     return CameraActivity::NO_ERROR;
 }
 
-void CameraWrapperConnector::fillListWrapperLibs(const std::string& folderPath, std::vector<std::string>& listLibs)
+void CameraWrapperConnector::fillListWrapperLibs(const cv::String& folderPath, std::vector<cv::String>& listLibs)
 {
     DIR *dp;
     struct dirent *ep;
@@ -264,7 +270,7 @@ void CameraWrapperConnector::fillListWrapperLibs(const std::string& folderPath,
     }
 }
 
-std::string CameraWrapperConnector::getDefaultPathLibFolder()
+cv::String CameraWrapperConnector::getDefaultPathLibFolder()
 {
     #define BIN_PACKAGE_NAME(x) "org.opencv.lib_v" CVAUX_STR(CV_VERSION_EPOCH) CVAUX_STR(CV_VERSION_MAJOR) "_" x
     const char* const packageList[] = {BIN_PACKAGE_NAME("armv7a"), OPENCV_ENGINE_PACKAGE};
@@ -287,10 +293,10 @@ std::string CameraWrapperConnector::getDefaultPathLibFolder()
         }
     }
 
-    return std::string();
+    return cv::String();
 }
 
-std::string CameraWrapperConnector::getPathLibFolder()
+cv::String CameraWrapperConnector::getPathLibFolder()
 {
     if (!pathLibFolder.empty())
         return pathLibFolder;
@@ -358,10 +364,10 @@ std::string CameraWrapperConnector::getPathLibFolder()
         LOGE("Could not get library name and base address");
     }
 
-    return std::string();
+    return cv::String();
 }
 
-void CameraWrapperConnector::setPathLibFolder(const std::string& path)
+void CameraWrapperConnector::setPathLibFolder(const cv::String& path)
 {
     pathLibFolder=path;
 }
@@ -428,14 +434,14 @@ void CameraActivity::applyProperties()
 int CameraActivity::getFrameWidth()
 {
     if (frameWidth <= 0)
-    frameWidth = getProperty(ANDROID_CAMERA_PROPERTY_FRAMEWIDTH);
+        frameWidth = getProperty(ANDROID_CAMERA_PROPERTY_FRAMEWIDTH);
     return frameWidth;
 }
 
 int CameraActivity::getFrameHeight()
 {
     if (frameHeight <= 0)
-    frameHeight = getProperty(ANDROID_CAMERA_PROPERTY_FRAMEHEIGHT);
+        frameHeight = getProperty(ANDROID_CAMERA_PROPERTY_FRAMEHEIGHT);
     return frameHeight;
 }
 
diff --git a/modules/bioinspired/CMakeLists.txt b/modules/bioinspired/CMakeLists.txt
new file mode 100644
index 0000000000..b0f152cfcf
--- /dev/null
+++ b/modules/bioinspired/CMakeLists.txt
@@ -0,0 +1,2 @@
+set(the_description "Biologically inspired algorithms")
+ocv_define_module(bioinspired opencv_core OPTIONAL opencv_highgui opencv_ocl)
diff --git a/modules/bioinspired/doc/bioinspired.rst b/modules/bioinspired/doc/bioinspired.rst
new file mode 100644
index 0000000000..6bffcdcf28
--- /dev/null
+++ b/modules/bioinspired/doc/bioinspired.rst
@@ -0,0 +1,10 @@
+********************************************************************
+bioinspired. Biologically inspired vision models and derivated tools
+********************************************************************
+
+The module provides biological visual systems models (human visual system and others). It also provides derivated objects that take advantage of those bio-inspired models.
+
+.. toctree::
+    :maxdepth: 2
+
+    Human retina documentation <retina/index>
diff --git a/modules/bioinspired/doc/retina/images/retinaInput.jpg b/modules/bioinspired/doc/retina/images/retinaInput.jpg
new file mode 100644
index 0000000000..d3cdeeecbd
Binary files /dev/null and b/modules/bioinspired/doc/retina/images/retinaInput.jpg differ
diff --git a/modules/bioinspired/doc/retina/images/retinaOutput_default.jpg b/modules/bioinspired/doc/retina/images/retinaOutput_default.jpg
new file mode 100644
index 0000000000..0b14a5308f
Binary files /dev/null and b/modules/bioinspired/doc/retina/images/retinaOutput_default.jpg differ
diff --git a/modules/bioinspired/doc/retina/images/retinaOutput_realistic.jpg b/modules/bioinspired/doc/retina/images/retinaOutput_realistic.jpg
new file mode 100644
index 0000000000..1bd60f80cc
Binary files /dev/null and b/modules/bioinspired/doc/retina/images/retinaOutput_realistic.jpg differ
diff --git a/modules/contrib/doc/retina/index.rst b/modules/bioinspired/doc/retina/index.rst
similarity index 57%
rename from modules/contrib/doc/retina/index.rst
rename to modules/bioinspired/doc/retina/index.rst
index a3a5bc82fc..242416baa7 100644
--- a/modules/contrib/doc/retina/index.rst
+++ b/modules/bioinspired/doc/retina/index.rst
@@ -5,56 +5,96 @@ Retina : a Bio mimetic human retina model
 
 Retina
 ======
+.. ocv:class:: Retina : public Algorithm
 
-.. ocv:class:: Retina
+**Note** : do not forget that the retina model is included in the following namespace : *cv::bioinspired*.
 
-Class which provides the main controls to the Gipsa/Listic labs human  retina model. Spatio-temporal filtering modelling the two main retina information channels :
+Introduction
+++++++++++++
 
-* foveal vision for detailled color vision : the parvocellular pathway).
+Class which provides the main controls to the Gipsa/Listic labs human  retina model. This is a non separable spatio-temporal filter modelling the two main retina information channels :
 
-* periphearal vision for sensitive transient signals detection (motion and events) : the magnocellular pathway.
+* foveal vision for detailled color vision : the parvocellular pathway.
 
-**NOTE : See the Retina tutorial in the tutorial/contrib section for complementary explanations.**
+* peripheral vision for sensitive transient signals detection (motion and events) : the magnocellular pathway.
 
-The retina can be settled up with various parameters, by default, the retina cancels mean luminance and enforces all details of the visual scene. In order to use your own parameters, you can use at least one time the *write(std::string fs)* method which will write a proper XML file with all default parameters. Then, tweak it on your own and reload them at any time using method *setup(std::string fs)*. These methods update a *Retina::RetinaParameters* member structure that is described hereafter. ::
+From a general point of view, this filter whitens the image spectrum and corrects luminance thanks to local adaptation. An other important property is its hability to filter out spatio-temporal noise while enhancing details.
+This model originates from Jeanny Herault work [Herault2010]_. It has been involved in Alexandre Benoit phd and his current research [Benoit2010]_, [Strat2013]_ (he currently maintains this module within OpenCV). It includes the work of other Jeanny's phd student such as [Chaix2007]_ and the log polar transformations of Barthelemy Durette described in Jeanny's book.
 
-  class Retina
+**NOTES :**
+
+* For ease of use in computer vision applications, the two retina channels are applied homogeneously on all the input images. This does not follow the real retina topology but this can still be done using the log sampling capabilities proposed within the class.
+
+* Extend the retina description and code use in the tutorial/contrib section for complementary explanations.
+
+Preliminary illustration
+++++++++++++++++++++++++
+
+As a preliminary presentation, let's start with a visual example. We propose to apply the filter on a low quality color jpeg image with backlight problems. Here is the considered input... *"Well, my eyes were able to see more that this strange black shadow..."*
+
+.. image:: images/retinaInput.jpg
+   :alt: a low quality color jpeg image with backlight problems.
+   :align: center
+
+Below, the retina foveal model applied on the entire image with default parameters. Here contours are enforced, halo effects are voluntary visible with this configuration. See parameters discussion below and increase horizontalCellsGain near 1 to remove them.
+
+.. image:: images/retinaOutput_default.jpg
+   :alt: the retina foveal model applied on the entire image with default parameters. Here contours are enforced, luminance is corrected and halo effects are voluntary visible with this configuration, increase horizontalCellsGain near 1 to remove them.
+   :align: center
+
+Below, a second retina foveal model output applied on the entire image with a parameters setup focused on naturalness perception. *"Hey, i now recognize my cat, looking at the mountains at the end of the day !"*. Here contours are enforced, luminance is corrected but halos are avoided with this configuration. The backlight effect is corrected and highlight details are still preserved. Then, even on a low quality jpeg image, if some luminance information remains, the retina is able to reconstruct a proper visual signal. Such configuration is also usefull for High Dynamic Range (*HDR*) images compression to 8bit images as discussed in [benoit2010]_ and in the demonstration codes discussed below.
+As shown at the end of the page, parameters change from defaults are :
+
+* horizontalCellsGain=0.3
+
+* photoreceptorsLocalAdaptationSensitivity=ganglioncellsSensitivity=0.89.
+
+.. image:: images/retinaOutput_realistic.jpg
+   :alt: the retina foveal model applied on the entire image with 'naturalness' parameters. Here contours are enforced but are avoided with this configuration, horizontalCellsGain is 0.3 and photoreceptorsLocalAdaptationSensitivity=ganglioncellsSensitivity=0.89.
+   :align: center
+
+As observed in this preliminary demo, the retina can be settled up with various parameters, by default, as shown on the figure above, the retina strongly reduces mean luminance energy and enforces all details of the visual scene. Luminance energy and halo effects can be modulated (exagerated to cancelled as shown on the two examples). In order to use your own parameters, you can use at least one time the *write(String fs)* method which will write a proper XML file with all default parameters. Then, tweak it on your own and reload them at any time using method *setup(String fs)*. These methods update a *Retina::RetinaParameters* member structure that is described hereafter. XML parameters file samples are shown at the end of the page.
+
+Here is an overview of the abstract Retina interface, allocate one instance with the *createRetina* functions.::
+
+  namespace cv{namespace bioinspired{
+
+  class Retina : public Algorithm
   {
   public:
     // parameters setup instance
     struct RetinaParameters; // this class is detailled later
 
-    // constructors
-    Retina (Size inputSize);
-    Retina (Size inputSize, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
+    // main method for input frame processing (all use method, can also perform High Dynamic Range tone mapping)
+    void run (InputArray inputImage);
 
-    // main method for input frame processing
-    void run (const Mat &inputImage);
+    // specific method aiming at correcting luminance only (faster High Dynamic Range tone mapping)
+    void applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)
 
     // output buffers retreival methods
     // -> foveal color vision details channel with luminance and noise correction
-    void getParvo (Mat &retinaOutput_parvo);
-    void getParvo (std::valarray< float > &retinaOutput_parvo);
-    const std::valarray< float > & getParvo () const;
+    void getParvo (OutputArray retinaOutput_parvo);
+    void getParvoRAW (OutputArray retinaOutput_parvo);// retreive original output buffers without any normalisation
+    const Mat getParvoRAW () const;// retreive original output buffers without any normalisation
     // -> peripheral monochrome motion and events (transient information) channel
-    void getMagno (Mat &retinaOutput_magno);
-    void getMagno (std::valarray< float > &retinaOutput_magno);
-    const std::valarray< float > & getMagno () const;
+    void getMagno (OutputArray retinaOutput_magno);
+    void getMagnoRAW (OutputArray retinaOutput_magno); // retreive original output buffers without any normalisation
+    const Mat getMagnoRAW () const;// retreive original output buffers without any normalisation
 
     // reset retina buffers... equivalent to closing your eyes for some seconds
     void clearBuffers ();
 
     // retreive input and output buffers sizes
-    Size inputSize ();
-    Size outputSize ();
+    Size getInputSize ();
+    Size getOutputSize ();
 
     // setup methods with specific parameters specification of global xml config file loading/write
-    void setup (std::string retinaParameterFile="", const bool applyDefaultSetupOnFailure=true);
+    void setup (String retinaParameterFile="", const bool applyDefaultSetupOnFailure=true);
     void setup (FileStorage &fs, const bool applyDefaultSetupOnFailure=true);
     void setup (RetinaParameters newParameters);
     struct Retina::RetinaParameters getParameters ();
-    const std::string printSetup ();
-    virtual void write (std::string fs) const;
+    const String printSetup ();
+    virtual void write (String fs) const;
     virtual void write (FileStorage &fs) const;
     void setupOPLandIPLParvoChannel (const bool colorMode=true, const bool normaliseOutput=true, const float photoreceptorsLocalAdaptationSensitivity=0.7, const float photoreceptorsTemporalConstant=0.5, const float photoreceptorsSpatialConstant=0.53, const float horizontalCellsGain=0, const float HcellsTemporalConstant=1, const float HcellsSpatialConstant=7, const float ganglionCellsSensitivity=0.7);
     void setupIPLMagnoChannel (const bool normaliseOutput=true, const float parasolCells_beta=0, const float parasolCells_tau=0, const float parasolCells_k=7, const float amacrinCellsTemporalCutFrequency=1.2, const float V0CompressionParameter=0.95, const float localAdaptintegration_tau=0, const float localAdaptintegration_k=7);
@@ -63,11 +103,21 @@ The retina can be settled up with various parameters, by default, the retina can
     void activateContoursProcessing (const bool activate);
   };
 
+    // Allocators
+    cv::Ptr<Retina> createRetina (Size inputSize);
+    cv::Ptr<Retina> createRetina (Size inputSize, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
+    }} // cv and bioinspired namespaces end
+
+.. Sample code::
+
+   * An example on retina tone mapping can be found at opencv_source_code/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp
+   * An example on retina tone mapping on video input can be found at opencv_source_code/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp
+   * A complete example illustrating the retina interface can be found at opencv_source_code/samples/cpp/retinaDemo.cpp
 
 Description
 +++++++++++
 
-Class which allows the `Gipsa <http://www.gipsa-lab.inpg.fr>`_ (preliminary work) / `Listic <http://www.listic.univ-savoie.fr>`_ (code maintainer) labs retina model to be used. This class allows human retina spatio-temporal image processing to be applied on still images, images sequences and video sequences. Briefly, here are the main human retina model properties:
+Class which allows the `Gipsa <http://www.gipsa-lab.inpg.fr>`_ (preliminary work) / `Listic <http://www.listic.univ-savoie.fr>`_ (code maintainer and user) labs retina model to be used. This class allows human retina spatio-temporal image processing to be applied on still images, images sequences and video sequences. Briefly, here are the main human retina model properties:
 
 * spectral whithening (mid-frequency details enhancement)
 
@@ -83,19 +133,35 @@ Use : this model can be used basically for spatio-temporal video effects but als
 
 * performing motion analysis also taking benefit of the previously cited properties  (check out the magnocellular retina channel output, by using the provided **getMagno** methods)
 
+* general image/video sequence description using either one or both channels. An example of the use of Retina in a Bag of Words approach is given in [Strat2013]_.
+
+Literature
+==========
 For more information, refer to the following papers :
 
-* Benoit A., Caplier A., Durette B., Herault, J., "Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
+* Model description :
+
+.. [Benoit2010] Benoit A., Caplier A., Durette B., Herault, J., "Using Human Visual System Modeling For Bio-Inspired Low Level Image Processing", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773. DOI <http://dx.doi.org/10.1016/j.cviu.2010.01.011>
+
+* Model use in a Bag of Words approach :
+
+.. [Strat2013] Strat S., Benoit A., Lambert P., "Retina enhanced SIFT descriptors for video indexing", CBMI2013, Veszprém, Hungary, 2013.
 
 * Please have a look at the reference work of Jeanny Herault that you can read in his book :
 
-Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+.. [Herault2010] Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
 
 This retina filter code includes the research contributions of phd/research collegues from which code has been redrawn by the author :
 
-* take a look at the *retinacolor.hpp* module to discover Brice Chaix de Lavarene phD color mosaicing/demosaicing and his reference paper: B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+* take a look at the *retinacolor.hpp* module to discover Brice Chaix de Lavarene phD color mosaicing/demosaicing and his reference paper:
 
-* take a look at *imagelogpolprojection.hpp* to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions. ====> more informations in the above cited Jeanny Heraults's book.
+.. [Chaix2007] B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+
+* take a look at *imagelogpolprojection.hpp* to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions. More informations in the above cited Jeanny Heraults's book.
+
+* Meylan&al work on HDR tone mapping that is implemented as a specific method within the model :
+
+.. [Meylan2007] L. Meylan , D. Alleysson, S. Susstrunk, "A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images", Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
 
 Demos and experiments !
 =======================
@@ -120,32 +186,38 @@ Take a look at the provided C++ examples provided with OpenCV :
 
    Then, take a HDR image using bracketing with your camera and generate an OpenEXR image and then process it using the demo.
 
-   Typical use, supposing that you have the OpenEXR image *memorial.exr* (present in the samples/cpp/ folder)
+   Typical use, supposing that you have the OpenEXR image such as *memorial.exr* (present in the samples/cpp/ folder)
 
-   **OpenCVReleaseFolder/bin/OpenEXRimages_HighDynamicRange_Retina_toneMapping memorial.exr**
+   **OpenCVReleaseFolder/bin/OpenEXRimages_HighDynamicRange_Retina_toneMapping memorial.exr [optionnal: 'fast']**
 
       Note that some sliders are made available to allow you to play with luminance compression.
 
+      If not using the 'fast' option, then, tone mapping is performed using the full retina model [Benoit2010]_. It includes spectral whitening that allows luminance energy to be reduced. When using the 'fast' option, then, a simpler method is used, it is an adaptation of the algorithm presented in [Meylan2007]_. This method gives also good results and is faster to process but it sometimes requires some more parameters adjustement.
+
 
 Methods description
 ===================
 
 Here are detailled the main methods to control the retina model
 
-Retina::Retina
-++++++++++++++
+Ptr<Retina>::createRetina
++++++++++++++++++++++++++
 
-.. ocv:function:: Retina::Retina(Size inputSize)
-.. ocv:function:: Retina::Retina(Size inputSize, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod = RETINA_COLOR_BAYER, const bool useRetinaLogSampling = false, const double reductionFactor = 1.0, const double samplingStrenght = 10.0 )
+.. ocv:function:: Ptr<cv::bioinspired::Retina> createRetina(Size inputSize)
+.. ocv:function:: Ptr<cv::bioinspired::Retina> createRetina(Size inputSize, const bool colorMode, cv::bioinspired::RETINA_COLORSAMPLINGMETHOD colorSamplingMethod = cv::bioinspired::RETINA_COLOR_BAYER, const bool useRetinaLogSampling = false, const double reductionFactor = 1.0, const double samplingStrenght = 10.0 )
 
-    Constructors
+    Constructors from standardized interfaces : retreive a smart pointer to a Retina instance
 
     :param inputSize: the input frame size
     :param colorMode: the chosen processing mode : with or without color processing
-    :param colorSamplingMethod: specifies which kind of color sampling will be used
-        * RETINA_COLOR_RANDOM: each pixel position is either R, G or B in a random choice
-        * RETINA_COLOR_DIAGONAL: color sampling is RGBRGBRGB..., line 2 BRGBRGBRG..., line 3, GBRGBRGBR...
-        * RETINA_COLOR_BAYER: standard bayer sampling
+    :param colorSamplingMethod: specifies which kind of color sampling will be used :
+
+        * cv::bioinspired::RETINA_COLOR_RANDOM: each pixel position is either R, G or B in a random choice
+
+        * cv::bioinspired::RETINA_COLOR_DIAGONAL: color sampling is RGBRGBRGB..., line 2 BRGBRGBRG..., line 3, GBRGBRGBR...
+
+        * cv::bioinspired::RETINA_COLOR_BAYER: standard bayer sampling
+
     :param useRetinaLogSampling: activate retina log sampling, if true, the 2 following parameters can be used
     :param reductionFactor: only usefull if param useRetinaLogSampling=true, specifies the reduction factor of the output frame (as the center (fovea) is high resolution and corners can be underscaled, then a reduction of the output is allowed without precision leak
     :param samplingStrenght: only usefull if param useRetinaLogSampling=true, specifies the strenght of the log scale that is applied
@@ -178,55 +250,46 @@ Retina::clearBuffers
 Retina::getParvo
 ++++++++++++++++
 
-.. ocv:function:: void Retina::getParvo( Mat & retinaOutput_parvo )
-.. ocv:function:: void Retina::getParvo( std::valarray<float> & retinaOutput_parvo )
-.. ocv:function:: const std::valarray<float> & Retina::getParvo() const
+.. ocv:function:: void Retina::getParvo( OutputArray retinaOutput_parvo )
+.. ocv:function:: void Retina::getParvoRAW( OutputArray retinaOutput_parvo )
+.. ocv:function:: const Mat Retina::getParvoRAW() const
 
-    Accessor of the details channel of the retina (models foveal vision)
+    Accessor of the details channel of the retina (models foveal vision). Warning, getParvoRAW methods return buffers that are not rescaled within range [0;255] while the non RAW method allows a normalized matrix to be retrieved.
 
     :param retinaOutput_parvo: the output buffer (reallocated if necessary), format can be :
 
         * a Mat, this output is rescaled for standard 8bits image processing use in OpenCV
 
-        * a 1D std::valarray Buffer (encoding is R1, R2, ... Rn), this output is the original retina filter model output, without any quantification or rescaling
+        * RAW methods actually return a 1D matrix (encoding is R1, R2, ... Rn, G1, G2, ..., Gn, B1, B2, ...Bn), this output is the original retina filter model output, without any quantification or rescaling.
 
 Retina::getMagno
 ++++++++++++++++
 
-.. ocv:function:: void Retina::getMagno( Mat & retinaOutput_magno )
-.. ocv:function:: void Retina::getMagno( std::valarray<float> & retinaOutput_magno )
-.. ocv:function:: const std::valarray<float> & Retina::getMagno() const
+.. ocv:function:: void Retina::getMagno( OutputArray retinaOutput_magno )
+.. ocv:function:: void Retina::getMagnoRAW( OutputArray retinaOutput_magno )
+.. ocv:function:: const Mat Retina::getMagnoRAW() const
 
-    Accessor of the motion channel of the retina (models peripheral vision)
+    Accessor of the motion channel of the retina (models peripheral vision). Warning, getMagnoRAW methods return buffers that are not rescaled within range [0;255] while the non RAW method allows a normalized matrix to be retrieved.
 
     :param retinaOutput_magno: the output buffer (reallocated if necessary), format can be :
 
         * a Mat, this output is rescaled for standard 8bits image processing use in OpenCV
 
-        * a 1D std::valarray Buffer (encoding is R1, R2, ... Rn), this output is the original retina filter model output, without any quantification or rescaling
+        * RAW methods actually return a 1D matrix (encoding is M1, M2,... Mn), this output is the original retina filter model output, without any quantification or rescaling.
 
-Retina::getParameters
-+++++++++++++++++++++
+Retina::getInputSize
+++++++++++++++++++++
 
-.. ocv:function:: Retina::RetinaParameters Retina::getParameters()
-
-    Retrieve the current parameters values in a *Retina::RetinaParameters* structure
-
-    :return: the current parameters setup
-
-Retina::inputSize
-+++++++++++++++++
-
-.. ocv:function:: Size Retina::inputSize()
+.. ocv:function:: Size Retina::getInputSize()
 
     Retreive retina input buffer size
 
     :return: the retina input buffer size
 
-Retina::outputSize
-++++++++++++++++++
+Retina::getOutputSize
++++++++++++++++++++++
 
-.. ocv:function:: Size Retina::outputSize()
+.. ocv:function:: Size Retina::getOutputSize()
 
     Retreive retina output buffer size that can be different from the input if a spatial log transformation is applied
 
@@ -235,21 +298,33 @@ Retina::outputSize
 Retina::printSetup
 ++++++++++++++++++
 
-.. ocv:function:: const std::string Retina::printSetup()
+.. ocv:function:: const String Retina::printSetup()
 
     Outputs a string showing the used parameters setup
 
-    :return: a string which contains formatted parameters information
+    :return: a string which contains formated parameters information
 
 Retina::run
 +++++++++++
 
-.. ocv:function:: void Retina::run(const Mat & inputImage)
+.. ocv:function:: void Retina::run(InputArray inputImage)
 
     Method which allows retina to be applied on an input image, after run, encapsulated retina module is ready to deliver its outputs using dedicated acccessors, see getParvo and getMagno methods
 
     :param inputImage: the input Mat image to be processed, can be gray level or BGR coded in any format (from 8bit to 16bits)
 
+Retina::applyFastToneMapping
+++++++++++++++++++++++++++++
+
+.. ocv:function:: void Retina::applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)
+
+    Method which processes an image in the aim to correct its luminance : correct backlight problems, enhance details in shadows. This method is designed to perform High Dynamic Range image tone mapping (compress >8bit/pixel images to 8bit/pixel). This is a simplified version of the Retina Parvocellular model (simplified version of the run/getParvo methods call) since it does not include the spatio-temporal filter modelling the Outer Plexiform Layer of the retina that performs spectral whitening and many other stuff. However, it works great for tone mapping and in a faster way.
+
+    Check the demos and experiments section to see examples and the way to perform tone mapping using the original retina model and the method.
+
+    :param inputImage: the input image to process (should be coded in float format : CV_32F, CV_32FC1, CV_32F_C3, CV_32F_C4, the 4th channel won't be considered).
+    :param outputToneMappedImage: the output 8bit/channel tone mapped image (CV_8U or CV_8UC3 format).
+
 Retina::setColorSaturation
 ++++++++++++++++++++++++++
 
@@ -264,7 +339,7 @@ Retina::setColorSaturation
 Retina::setup
 +++++++++++++
 
-.. ocv:function:: void Retina::setup(std::string retinaParameterFile = "", const bool applyDefaultSetupOnFailure = true )
+.. ocv:function:: void Retina::setup(String retinaParameterFile = "", const bool applyDefaultSetupOnFailure = true )
 .. ocv:function:: void Retina::setup(FileStorage & fs, const bool applyDefaultSetupOnFailure = true )
 .. ocv:function:: void Retina::setup(RetinaParameters newParameters)
 
@@ -273,12 +348,12 @@ Retina::setup
     :param retinaParameterFile: the parameters filename
     :param applyDefaultSetupOnFailure: set to true if an error must be thrown on error
     :param fs: the open Filestorage which contains retina parameters
-    :param newParameters: a parameters structures updated with the new target configuration
+    :param newParameters: a parameters structures updated with the new target configuration. You can retreive the current parameers structure using method *Retina::RetinaParameters Retina::getParameters()* and update it before running method *setup*.
 
 Retina::write
 +++++++++++++
 
-.. ocv:function:: void Retina::write( std::string fs ) const
+.. ocv:function:: void Retina::write( String fs ) const
 .. ocv:function:: void Retina::write( FileStorage& fs ) const
 
     Write xml/yml formated parameters information
@@ -335,7 +410,7 @@ Retina::RetinaParameters
                   photoreceptorsTemporalConstant(0.5f),// the time constant of the first order low pass filter of the photoreceptors, use it to cut high temporal frequencies (noise or fast motion), unit is frames, typical value is 1 frame
                   photoreceptorsSpatialConstant(0.53f),// the spatial constant of the first order low pass filter of the photoreceptors, use it to cut high spatial frequencies (noise or thick contours), unit is pixels, typical value is 1 pixel
                   horizontalCellsGain(0.0f),//gain of the horizontal cells network, if 0, then the mean value of the output is zero, if the parameter is near 1, then, the luminance is not filtered and is still reachable at the output, typicall value is 0
-                  hcellsTemporalConstant(1.f),// the time constant of the first order low pass filter of the horizontal cells, use it to cut low temporal frequencies (local luminance variations), unit is frames, typical value is 1 frame, as the photoreceptors
+                  hcellsTemporalConstant(1.f),// the time constant of the first order low pass filter of the horizontal cells, use it to cut low temporal frequencies (local luminance variations), unit is frames, typical value is 1 frame, as the photoreceptors. Reduce to 0.5 to limit retina after effects.
                   hcellsSpatialConstant(7.f),//the spatial constant of the first order low pass filter of the horizontal cells, use it to cut low spatial frequencies (local luminance), unit is pixels, typical value is 5 pixel, this value is also used for local contrast computing when computing the local contrast adaptation at the ganglion cells level (Inner Plexiform Layer parvocellular channel model)
                   ganglionCellsSensitivity(0.7f)//the compression strengh of the ganglion cells local adaptation output, set a value between 0.6 and 1 for best results, a high value increases more the low value sensitivity... and the output saturates faster, recommended value: 0.7
                   {};// default setup
@@ -359,3 +434,60 @@ Retina::RetinaParameters
             struct OPLandIplParvoParameters OPLandIplParvo;
             struct IplMagnoParameters IplMagno;
     };
+
+Retina parameters files examples
+++++++++++++++++++++++++++++++++
+
+Here is the default configuration file of the retina module. It gives results such as the first retina output shown on the top of this page.
+
+.. code-block:: cpp
+
+    <?xml version="1.0"?>
+    <opencv_storage>
+    <OPLandIPLparvo>
+        <colorMode>1</colorMode>
+        <normaliseOutput>1</normaliseOutput>
+        <photoreceptorsLocalAdaptationSensitivity>7.5e-01</photoreceptorsLocalAdaptationSensitivity>
+        <photoreceptorsTemporalConstant>9.0e-01</photoreceptorsTemporalConstant>
+        <photoreceptorsSpatialConstant>5.3e-01</photoreceptorsSpatialConstant>
+        <horizontalCellsGain>0.01</horizontalCellsGain>
+        <hcellsTemporalConstant>0.5</hcellsTemporalConstant>
+        <hcellsSpatialConstant>7.</hcellsSpatialConstant>
+        <ganglionCellsSensitivity>7.5e-01</ganglionCellsSensitivity></OPLandIPLparvo>
+    <IPLmagno>
+        <normaliseOutput>1</normaliseOutput>
+        <parasolCells_beta>0.</parasolCells_beta>
+        <parasolCells_tau>0.</parasolCells_tau>
+        <parasolCells_k>7.</parasolCells_k>
+        <amacrinCellsTemporalCutFrequency>2.0e+00</amacrinCellsTemporalCutFrequency>
+        <V0CompressionParameter>9.5e-01</V0CompressionParameter>
+        <localAdaptintegration_tau>0.</localAdaptintegration_tau>
+        <localAdaptintegration_k>7.</localAdaptintegration_k></IPLmagno>
+    </opencv_storage>
+
+Here is the 'realistic" setup used to obtain the second retina output shown on the top of this page.
+
+.. code-block:: cpp
+
+    <?xml version="1.0"?>
+    <opencv_storage>
+    <OPLandIPLparvo>
+      <colorMode>1</colorMode>
+      <normaliseOutput>1</normaliseOutput>
+      <photoreceptorsLocalAdaptationSensitivity>8.9e-01</photoreceptorsLocalAdaptationSensitivity>
+      <photoreceptorsTemporalConstant>9.0e-01</photoreceptorsTemporalConstant>
+      <photoreceptorsSpatialConstant>5.3e-01</photoreceptorsSpatialConstant>
+      <horizontalCellsGain>0.3</horizontalCellsGain>
+      <hcellsTemporalConstant>0.5</hcellsTemporalConstant>
+      <hcellsSpatialConstant>7.</hcellsSpatialConstant>
+      <ganglionCellsSensitivity>8.9e-01</ganglionCellsSensitivity></OPLandIPLparvo>
+    <IPLmagno>
+      <normaliseOutput>1</normaliseOutput>
+      <parasolCells_beta>0.</parasolCells_beta>
+      <parasolCells_tau>0.</parasolCells_tau>
+      <parasolCells_k>7.</parasolCells_k>
+      <amacrinCellsTemporalCutFrequency>2.0e+00</amacrinCellsTemporalCutFrequency>
+      <V0CompressionParameter>9.5e-01</V0CompressionParameter>
+      <localAdaptintegration_tau>0.</localAdaptintegration_tau>
+      <localAdaptintegration_k>7.</localAdaptintegration_k></IPLmagno>
+    </opencv_storage>
diff --git a/modules/bioinspired/include/opencv2/bioinspired.hpp b/modules/bioinspired/include/opencv2/bioinspired.hpp
new file mode 100644
index 0000000000..5f2f8644d2
--- /dev/null
+++ b/modules/bioinspired/include/opencv2/bioinspired.hpp
@@ -0,0 +1,50 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_BIOINSPIRED_HPP__
+#define __OPENCV_BIOINSPIRED_HPP__
+
+#include "opencv2/core.hpp"
+#include "opencv2/bioinspired/retina.hpp"
+#include "opencv2/bioinspired/retinafasttonemapping.hpp"
+
+#endif
diff --git a/modules/bioinspired/include/opencv2/bioinspired/bioinspired.hpp b/modules/bioinspired/include/opencv2/bioinspired/bioinspired.hpp
new file mode 100644
index 0000000000..40be2854eb
--- /dev/null
+++ b/modules/bioinspired/include/opencv2/bioinspired/bioinspired.hpp
@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/bioinspired.hpp"
diff --git a/modules/contrib/include/opencv2/contrib/retina.hpp b/modules/bioinspired/include/opencv2/bioinspired/retina.hpp
similarity index 73%
rename from modules/contrib/include/opencv2/contrib/retina.hpp
rename to modules/bioinspired/include/opencv2/bioinspired/retina.hpp
index 456daab443..b4fda7038e 100644
--- a/modules/contrib/include/opencv2/contrib/retina.hpp
+++ b/modules/bioinspired/include/opencv2/bioinspired/retina.hpp
@@ -6,12 +6,12 @@
  ** copy or use the software.
  **
  **
- ** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+ ** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
  ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
  **
  ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
  **
- **  Creation - enhancement process 2007-2011
+ **  Creation - enhancement process 2007-2013
  **      Author: Alexandre Benoit (benoit.alexandre.vision@gmail.com), LISTIC lab, Annecy le vieux, France
  **
  ** Theses algorithm have been developped by Alexandre BENOIT since his thesis with Alice Caplier at Gipsa-Lab (www.gipsa-lab.inpg.fr) and the research he pursues at LISTIC Lab (www.listic.univ-savoie.fr).
@@ -32,7 +32,7 @@
  ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
  ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
  **
- **               For Human Visual System tools (hvstools)
+ **               For Human Visual System tools (bioinspired)
  ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
  **
  ** Third party copyrights are property of their respective owners.
@@ -62,8 +62,8 @@
  ** the use of this software, even if advised of the possibility of such damage.
  *******************************************************************************/
 
-#ifndef __OPENCV_CONTRIB_RETINA_HPP__
-#define __OPENCV_CONTRIB_RETINA_HPP__
+#ifndef __OPENCV_BIOINSPIRED_RETINA_HPP__
+#define __OPENCV_BIOINSPIRED_RETINA_HPP__
 
 /*
  * Retina.hpp
@@ -73,22 +73,19 @@
  */
 
 #include "opencv2/core.hpp" // for all OpenCV core functionalities access, including cv::Exception support
-#include <valarray>
 
-namespace cv
-{
 
-enum RETINA_COLORSAMPLINGMETHOD
-{
+namespace cv{
+namespace bioinspired{
+
+enum {
     RETINA_COLOR_RANDOM, //!< each pixel position is either R, G or B in a random choice
     RETINA_COLOR_DIAGONAL,//!< color sampling is RGBRGBRGB..., line 2 BRGBRGBRG..., line 3, GBRGBRGBR...
     RETINA_COLOR_BAYER//!< standard bayer sampling
 };
 
-class RetinaFilter;
-
 /**
- * @class Retina a wrapper class which allows the Gipsa/Listic Labs model to be used.
+ * @class Retina a wrapper class which allows the Gipsa/Listic Labs model to be used with OpenCV.
  * This retina model allows spatio-temporal image processing (applied on still images, video sequences).
  * As a summary, these are the retina model properties:
  * => It applies a spectral whithening (mid-frequency details enhancement)
@@ -110,7 +107,7 @@ class RetinaFilter;
  * _take a look at imagelogpolprojection.hpp to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions.
  * ====> more informations in the above cited Jeanny Heraults's book.
  */
-class CV_EXPORTS Retina {
+class CV_EXPORTS Retina : public Algorithm {
 
 public:
 
@@ -119,13 +116,13 @@ public:
         struct OPLandIplParvoParameters{ // Outer Plexiform Layer (OPL) and Inner Plexiform Layer Parvocellular (IplParvo) parameters
                OPLandIplParvoParameters():colorMode(true),
                                  normaliseOutput(true),
-                                 photoreceptorsLocalAdaptationSensitivity(0.7f),
-                                 photoreceptorsTemporalConstant(0.5f),
+                                 photoreceptorsLocalAdaptationSensitivity(0.75f),
+                                 photoreceptorsTemporalConstant(0.9f),
                                  photoreceptorsSpatialConstant(0.53f),
-                                 horizontalCellsGain(0.0f),
-                                 hcellsTemporalConstant(1.f),
+                                 horizontalCellsGain(0.01f),
+                                 hcellsTemporalConstant(0.5f),
                                  hcellsSpatialConstant(7.f),
-                                 ganglionCellsSensitivity(0.7f){};// default setup
+                                 ganglionCellsSensitivity(0.75f){};// default setup
                bool colorMode, normaliseOutput;
                float photoreceptorsLocalAdaptationSensitivity, photoreceptorsTemporalConstant, photoreceptorsSpatialConstant, horizontalCellsGain, hcellsTemporalConstant, hcellsSpatialConstant, ganglionCellsSensitivity;
            };
@@ -135,7 +132,7 @@ public:
                           parasolCells_beta(0.f),
                           parasolCells_tau(0.f),
                           parasolCells_k(7.f),
-                          amacrinCellsTemporalCutFrequency(1.2f),
+                          amacrinCellsTemporalCutFrequency(2.0f),
                           V0CompressionParameter(0.95f),
                           localAdaptintegration_tau(0.f),
                           localAdaptintegration_k(7.f){};// default setup
@@ -146,34 +143,15 @@ public:
             struct IplMagnoParameters IplMagno;
     };
 
-    /**
-     * Main constructor with most commun use setup : create an instance of color ready retina model
-     * @param inputSize : the input frame size
-     */
-    Retina(Size inputSize);
-
-    /**
-     * Complete Retina filter constructor which allows all basic structural parameters definition
-         * @param inputSize : the input frame size
-     * @param colorMode : the chosen processing mode : with or without color processing
-     * @param colorSamplingMethod: specifies which kind of color sampling will be used
-     * @param useRetinaLogSampling: activate retina log sampling, if true, the 2 following parameters can be used
-     * @param reductionFactor: only usefull if param useRetinaLogSampling=true, specifies the reduction factor of the output frame (as the center (fovea) is high resolution and corners can be underscaled, then a reduction of the output is allowed without precision leak
-     * @param samplingStrenght: only usefull if param useRetinaLogSampling=true, specifies the strenght of the log scale that is applied
-     */
-    Retina(Size inputSize, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
-
-    virtual ~Retina();
-
     /**
     * retreive retina input buffer size
     */
-    Size inputSize();
+    virtual Size getInputSize()=0;
 
     /**
     * retreive retina output buffer size
     */
-    Size outputSize();
+    virtual Size getOutputSize()=0;
 
     /**
      * try to open an XML retina parameters file to adjust current retina instance setup
@@ -182,8 +160,7 @@ public:
      * @param retinaParameterFile : the parameters filename
          * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
      */
-    void setup(std::string retinaParameterFile="", const bool applyDefaultSetupOnFailure=true);
-
+    virtual void setup(String retinaParameterFile="", const bool applyDefaultSetupOnFailure=true)=0;
 
     /**
      * try to open an XML retina parameters file to adjust current retina instance setup
@@ -192,7 +169,7 @@ public:
      * @param fs : the open Filestorage which contains retina parameters
      * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
      */
-    void setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure=true);
+    virtual void setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure=true)=0;
 
     /**
      * try to open an XML retina parameters file to adjust current retina instance setup
@@ -201,31 +178,30 @@ public:
      * @param newParameters : a parameters structures updated with the new target configuration
          * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
      */
-    void setup(RetinaParameters newParameters);
+    virtual void setup(RetinaParameters newParameters)=0;
 
     /**
-     * @return the current parameters setup
-     */
-    Retina::RetinaParameters getParameters();
+    * @return the current parameters setup
+    */
+    virtual struct Retina::RetinaParameters getParameters()=0;
 
     /**
      * parameters setup display method
      * @return a string which contains formatted parameters information
      */
-    const std::string printSetup();
+    virtual const String printSetup()=0;
 
     /**
      * write xml/yml formated parameters information
      * @rparam fs : the filename of the xml file that will be open and writen with formatted parameters information
      */
-    virtual void write( std::string fs ) const;
-
+    virtual void write( String fs ) const=0;
 
     /**
      * write xml/yml formated parameters information
      * @param fs : a cv::Filestorage object ready to be filled
          */
-    virtual void write( FileStorage& fs ) const;
+    virtual void write( FileStorage& fs ) const=0;
 
     /**
      * setup the OPL and IPL parvo channels (see biologocal model)
@@ -242,7 +218,7 @@ public:
      * @param HcellsSpatialConstant: the spatial constant of the first order low pass filter of the horizontal cells, use it to cut low spatial frequencies (local luminance), unit is pixels, typical value is 5 pixel, this value is also used for local contrast computing when computing the local contrast adaptation at the ganglion cells level (Inner Plexiform Layer parvocellular channel model)
      * @param ganglionCellsSensitivity: the compression strengh of the ganglion cells local adaptation output, set a value between 160 and 250 for best results, a high value increases more the low value sensitivity... and the output saturates faster, recommended value: 230
      */
-    void setupOPLandIPLParvoChannel(const bool colorMode=true, const bool normaliseOutput = true, const float photoreceptorsLocalAdaptationSensitivity=0.7, const float photoreceptorsTemporalConstant=0.5, const float photoreceptorsSpatialConstant=0.53, const float horizontalCellsGain=0, const float HcellsTemporalConstant=1, const float HcellsSpatialConstant=7, const float ganglionCellsSensitivity=0.7);
+    virtual void setupOPLandIPLParvoChannel(const bool colorMode=true, const bool normaliseOutput = true, const float photoreceptorsLocalAdaptationSensitivity=0.7, const float photoreceptorsTemporalConstant=0.5, const float photoreceptorsSpatialConstant=0.53, const float horizontalCellsGain=0, const float HcellsTemporalConstant=1, const float HcellsSpatialConstant=7, const float ganglionCellsSensitivity=0.7)=0;
 
     /**
      * set parameters values for the Inner Plexiform Layer (IPL) magnocellular channel
@@ -256,41 +232,49 @@ public:
      * @param localAdaptintegration_tau: specifies the temporal constant of the low pas filter involved in the computation of the local "motion mean" for the local adaptation computation
      * @param localAdaptintegration_k: specifies the spatial constant of the low pas filter involved in the computation of the local "motion mean" for the local adaptation computation
      */
-    void setupIPLMagnoChannel(const bool normaliseOutput = true, const float parasolCells_beta=0, const float parasolCells_tau=0, const float parasolCells_k=7, const float amacrinCellsTemporalCutFrequency=1.2, const float V0CompressionParameter=0.95, const float localAdaptintegration_tau=0, const float localAdaptintegration_k=7);
+    virtual void setupIPLMagnoChannel(const bool normaliseOutput = true, const float parasolCells_beta=0, const float parasolCells_tau=0, const float parasolCells_k=7, const float amacrinCellsTemporalCutFrequency=1.2, const float V0CompressionParameter=0.95, const float localAdaptintegration_tau=0, const float localAdaptintegration_k=7)=0;
 
     /**
      * method which allows retina to be applied on an input image, after run, encapsulated retina module is ready to deliver its outputs using dedicated acccessors, see getParvo and getMagno methods
      * @param inputImage : the input cv::Mat image to be processed, can be gray level or BGR coded in any format (from 8bit to 16bits)
      */
-    void run(const Mat &inputImage);
+    virtual void run(InputArray inputImage)=0;
+
+    /**
+     * method that applies a luminance correction (initially High Dynamic Range (HDR) tone mapping) using only the 2 local adaptation stages of the retina parvo channel : photoreceptors level and ganlion cells level. Spatio temporal filtering is applied but limited to temporal smoothing and eventually high frequencies attenuation. This is a lighter method than the one available using the regular run method. It is then faster but it does not include complete temporal filtering nor retina spectral whitening. Then, it can have a more limited effect on images with a very high dynamic range. This is an adptation of the original still image HDR tone mapping algorithm of David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+    * -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+     @param inputImage the input image to process RGB or gray levels
+     @param outputToneMappedImage the output tone mapped image
+     */
+    virtual void applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)=0;
 
     /**
      * accessor of the details channel of the retina (models foveal vision)
      * @param retinaOutput_parvo : the output buffer (reallocated if necessary), this output is rescaled for standard 8bits image processing use in OpenCV
      */
-    void getParvo(Mat &retinaOutput_parvo);
+    virtual void getParvo(OutputArray retinaOutput_parvo)=0;
 
     /**
      * accessor of the details channel of the retina (models foveal vision)
-     * @param retinaOutput_parvo : the output buffer (reallocated if necessary), this output is the original retina filter model output, without any quantification or rescaling
+     * @param retinaOutput_parvo : a cv::Mat header filled with the internal parvo buffer of the retina module. This output is the original retina filter model output, without any quantification or rescaling
      */
-    void getParvo(std::valarray<float> &retinaOutput_parvo);
+    virtual void getParvoRAW(OutputArray retinaOutput_parvo)=0;
 
     /**
      * accessor of the motion channel of the retina (models peripheral vision)
      * @param retinaOutput_magno : the output buffer (reallocated if necessary), this output is rescaled for standard 8bits image processing use in OpenCV
      */
-    void getMagno(Mat &retinaOutput_magno);
+    virtual void getMagno(OutputArray retinaOutput_magno)=0;
 
     /**
      * accessor of the motion channel of the retina (models peripheral vision)
-     * @param retinaOutput_magno : the output buffer (reallocated if necessary), this output is the original retina filter model output, without any quantification or rescaling
+     * @param retinaOutput_magno : a cv::Mat header filled with the internal retina magno buffer of the retina module. This output is the original retina filter model output, without any quantification or rescaling
      */
-    void getMagno(std::valarray<float> &retinaOutput_magno);
+    virtual void getMagnoRAW(OutputArray retinaOutput_magno)=0;
 
-    // original API level data accessors : get buffers addresses...
-    const std::valarray<float> & getMagno() const;
-    const std::valarray<float> & getParvo() const;
+    // original API level data accessors : get buffers addresses from a Mat header, similar to getParvoRAW and getMagnoRAW...
+    virtual const Mat getMagnoRAW() const=0;
+    virtual const Mat getParvoRAW() const=0;
 
     /**
      * activate color saturation as the final step of the color demultiplexing process
@@ -298,59 +282,30 @@ public:
      * @param saturateColors: boolean that activates color saturation (if true) or desactivate (if false)
      * @param colorSaturationValue: the saturation factor
      */
-    void setColorSaturation(const bool saturateColors=true, const float colorSaturationValue=4.0);
+    virtual void setColorSaturation(const bool saturateColors=true, const float colorSaturationValue=4.0)=0;
 
     /**
      * clear all retina buffers (equivalent to opening the eyes after a long period of eye close ;o)
      */
-    void clearBuffers();
+    virtual void clearBuffers()=0;
 
     /**
     * Activate/desactivate the Magnocellular pathway processing (motion information extraction), by default, it is activated
     * @param activate: true if Magnocellular output should be activated, false if not
     */
-    void activateMovingContoursProcessing(const bool activate);
+    virtual void activateMovingContoursProcessing(const bool activate)=0;
 
     /**
     * Activate/desactivate the Parvocellular pathway processing (contours information extraction), by default, it is activated
     * @param activate: true if Parvocellular (contours information extraction) output should be activated, false if not
     */
-    void activateContoursProcessing(const bool activate);
-
-protected:
-    // Parameteres setup members
-    RetinaParameters _retinaParameters; // structure of parameters
-
-        // Retina model related modules
-    std::valarray<float> _inputBuffer; //!< buffer used to convert input cv::Mat to internal retina buffers format (valarrays)
-
-    // pointer to retina model
-    RetinaFilter* _retinaFilter; //!< the pointer to the retina module, allocated with instance construction
-
-    /**
-     * exports a valarray buffer outing from HVStools objects to a cv::Mat in CV_8UC1 (gray level picture) or CV_8UC3 (color) format
-     * @param grayMatrixToConvert the valarray to export to OpenCV
-     * @param nbRows : the number of rows of the valarray flatten matrix
-     * @param nbColumns : the number of rows of the valarray flatten matrix
-     * @param colorMode : a flag which mentions if matrix is color (true) or graylevel (false)
-     * @param outBuffer : the output matrix which is reallocated to satisfy Retina output buffer dimensions
-     */
-    void _convertValarrayBuffer2cvMat(const std::valarray<float> &grayMatrixToConvert, const unsigned int nbRows, const unsigned int nbColumns, const bool colorMode, Mat &outBuffer);
-
-    /**
-     *
-     * @param inputMatToConvert : the OpenCV cv::Mat that has to be converted to gray or RGB valarray buffer that will be processed by the retina model
-     * @param outputValarrayMatrix : the output valarray
-     * @return the input image color mode (color=true, gray levels=false)
-     */
-        bool _convertCvMat2ValarrayBuffer(const cv::Mat inputMatToConvert, std::valarray<float> &outputValarrayMatrix);
-
-    //! private method called by constructors, gathers their parameters and use them in a unified way
-    void _init(const Size inputSize, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
-
-
+    virtual void activateContoursProcessing(const bool activate)=0;
 };
+CV_EXPORTS Ptr<Retina> createRetina(Size inputSize);
+CV_EXPORTS Ptr<Retina> createRetina(Size inputSize, const bool colorMode, int colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
 
+CV_EXPORTS Ptr<Retina> createRetina_OCL(Size inputSize);
+CV_EXPORTS Ptr<Retina> createRetina_OCL(Size inputSize, const bool colorMode, int colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
 }
-#endif /* __OPENCV_CONTRIB_RETINA_HPP__ */
-
+}
+#endif /* __OPENCV_BIOINSPIRED_RETINA_HPP__ */
diff --git a/modules/bioinspired/include/opencv2/bioinspired/retinafasttonemapping.hpp b/modules/bioinspired/include/opencv2/bioinspired/retinafasttonemapping.hpp
new file mode 100644
index 0000000000..6c83f885c8
--- /dev/null
+++ b/modules/bioinspired/include/opencv2/bioinspired/retinafasttonemapping.hpp
@@ -0,0 +1,121 @@
+
+/*#******************************************************************************
+ ** IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+ **
+ ** By downloading, copying, installing or using the software you agree to this license.
+ ** If you do not agree to this license, do not download, install,
+ ** copy or use the software.
+ **
+ **
+ ** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+ **
+ ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
+ **
+ **  Creation - enhancement process 2007-2013
+ **      Author: Alexandre Benoit (benoit.alexandre.vision@gmail.com), LISTIC lab, Annecy le vieux, France
+ **
+ ** Theses algorithm have been developped by Alexandre BENOIT since his thesis with Alice Caplier at Gipsa-Lab (www.gipsa-lab.inpg.fr) and the research he pursues at LISTIC Lab (www.listic.univ-savoie.fr).
+ ** Refer to the following research paper for more information:
+ ** Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+ ** This work have been carried out thanks to Jeanny Herault who's research and great discussions are the basis of all this work, please take a look at his book:
+ ** Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+ **
+ **
+ **
+ **
+ **
+ ** This class is based on image processing tools of the author and already used within the Retina class (this is the same code as method retina::applyFastToneMapping, but in an independent class, it is ligth from a memory requirement point of view). It implements an adaptation of the efficient tone mapping algorithm propose by David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+ ** -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+ **
+ **
+ **                          License Agreement
+ **               For Open Source Computer Vision Library
+ **
+ ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+ ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+ **
+ **               For Human Visual System tools (bioinspired)
+ ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
+ **
+ ** Third party copyrights are property of their respective owners.
+ **
+ ** Redistribution and use in source and binary forms, with or without modification,
+ ** are permitted provided that the following conditions are met:
+ **
+ ** * Redistributions of source code must retain the above copyright notice,
+ **    this list of conditions and the following disclaimer.
+ **
+ ** * Redistributions in binary form must reproduce the above copyright notice,
+ **    this list of conditions and the following disclaimer in the documentation
+ **    and/or other materials provided with the distribution.
+ **
+ ** * The name of the copyright holders may not be used to endorse or promote products
+ **    derived from this software without specific prior written permission.
+ **
+ ** This software is provided by the copyright holders and contributors "as is" and
+ ** any express or implied warranties, including, but not limited to, the implied
+ ** warranties of merchantability and fitness for a particular purpose are disclaimed.
+ ** In no event shall the Intel Corporation or contributors be liable for any direct,
+ ** indirect, incidental, special, exemplary, or consequential damages
+ ** (including, but not limited to, procurement of substitute goods or services;
+ ** loss of use, data, or profits; or business interruption) however caused
+ ** and on any theory of liability, whether in contract, strict liability,
+ ** or tort (including negligence or otherwise) arising in any way out of
+ ** the use of this software, even if advised of the possibility of such damage.
+ *******************************************************************************/
+
+#ifndef __OPENCV_BIOINSPIRED_RETINAFASTTONEMAPPING_HPP__
+#define __OPENCV_BIOINSPIRED_RETINAFASTTONEMAPPING_HPP__
+
+/*
+ * retinafasttonemapping.hpp
+ *
+ *  Created on: May 26, 2013
+ *      Author: Alexandre Benoit
+ */
+
+#include "opencv2/core.hpp" // for all OpenCV core functionalities access, including cv::Exception support
+
+namespace cv{
+namespace bioinspired{
+
+/**
+ * @class RetinaFastToneMappingImpl a wrapper class which allows the tone mapping algorithm of Meylan&al(2007) to be used with OpenCV.
+ * This algorithm is already implemented in thre Retina class (retina::applyFastToneMapping) but used it does not require all the retina model to be allocated. This allows a light memory use for low memory devices (smartphones, etc.
+ * As a summary, these are the model properties:
+ * => 2 stages of local luminance adaptation with a different local neighborhood for each.
+ * => first stage models the retina photorecetors local luminance adaptation
+ * => second stage models th ganglion cells local information adaptation
+ * => compared to the initial publication, this class uses spatio-temporal low pass filters instead of spatial only filters.
+ * ====> this can help noise robustness and temporal stability for video sequence use cases.
+ * for more information, read to the following papers :
+ *  Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+ * regarding spatio-temporal filter and the bigger retina model :
+ * Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+ */
+class CV_EXPORTS RetinaFastToneMapping : public Algorithm
+{
+public:
+
+    /**
+     * method that applies a luminance correction (initially High Dynamic Range (HDR) tone mapping) using only the 2 local adaptation stages of the retina parvocellular channel : photoreceptors level and ganlion cells level. Spatio temporal filtering is applied but limited to temporal smoothing and eventually high frequencies attenuation. This is a lighter method than the one available using the regular retina::run method. It is then faster but it does not include complete temporal filtering nor retina spectral whitening. Then, it can have a more limited effect on images with a very high dynamic range. This is an adptation of the original still image HDR tone mapping algorithm of David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+    * -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+     @param inputImage the input image to process RGB or gray levels
+     @param outputToneMappedImage the output tone mapped image
+     */
+    virtual void applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)=0;
+
+    /**
+     * setup method that updates tone mapping behaviors by adjusing the local luminance computation area
+     * @param photoreceptorsNeighborhoodRadius the first stage local adaptation area
+     * @param ganglioncellsNeighborhoodRadius the second stage local adaptation area
+     * @param meanLuminanceModulatorK the factor applied to modulate the meanLuminance information (default is 1, see reference paper)
+     */
+    virtual void setup(const float photoreceptorsNeighborhoodRadius=3.f, const float ganglioncellsNeighborhoodRadius=1.f, const float meanLuminanceModulatorK=1.f)=0;
+};
+
+CV_EXPORTS Ptr<RetinaFastToneMapping> createRetinaFastToneMapping(Size inputSize);
+
+}
+}
+#endif /* __OPENCV_BIOINSPIRED_RETINAFASTTONEMAPPING_HPP__ */
diff --git a/modules/contrib/src/basicretinafilter.cpp b/modules/bioinspired/src/basicretinafilter.cpp
similarity index 98%
rename from modules/contrib/src/basicretinafilter.cpp
rename to modules/bioinspired/src/basicretinafilter.cpp
index 020b8f04e1..7e7b467faa 100644
--- a/modules/contrib/src/basicretinafilter.cpp
+++ b/modules/bioinspired/src/basicretinafilter.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -72,7 +72,8 @@
 
 namespace cv
 {
-
+namespace bioinspired
+{
 // @author Alexandre BENOIT, benoit.alexandre.vision@gmail.com, LISTIC : www.listic.univ-savoie.fr Gipsa-Lab, France: www.gipsa-lab.inpg.fr/
 
 //////////////////////////////////////////////////////////
@@ -883,4 +884,5 @@ void BasicRetinaFilter::_verticalAnticausalFilter_Irregular_multGain(float *outp
     }
 
 }
-}
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/basicretinafilter.hpp b/modules/bioinspired/src/basicretinafilter.hpp
similarity index 98%
rename from modules/contrib/src/basicretinafilter.hpp
rename to modules/bioinspired/src/basicretinafilter.hpp
index a2ece415db..323bff9408 100644
--- a/modules/contrib/src/basicretinafilter.hpp
+++ b/modules/bioinspired/src/basicretinafilter.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -113,6 +113,8 @@
 //#define __BASIC_RETINA_ELEMENT_DEBUG
 
 namespace cv
+{
+namespace bioinspired
 {
     class BasicRetinaFilter
     {
@@ -287,7 +289,7 @@ namespace cv
         * @param maxInputValue: the maximum amplitude value measured after local adaptation processing (c.f. function runFilter_LocalAdapdation & runFilter_LocalAdapdation_autonomous)
         * @param meanLuminance: the a priori meann luminance of the input data (should be 128 for 8bits images but can vary greatly in case of High Dynamic Range Images (HDRI)
         */
-        void setV0CompressionParameterToneMapping(const float v0, const float maxInputValue, const float meanLuminance=128.0f){ _v0=v0*maxInputValue; _localLuminanceFactor=1.0f; _localLuminanceAddon=meanLuminance*_v0; _maxInputValue=maxInputValue;};
+        void setV0CompressionParameterToneMapping(const float v0, const float maxInputValue, const float meanLuminance=128.0f){ _v0=v0*maxInputValue; _localLuminanceFactor=1.0f; _localLuminanceAddon=meanLuminance*v0; _maxInputValue=maxInputValue;};
 
         /**
         * update compression parameters while keeping v0 parameter value
@@ -650,7 +652,6 @@ namespace cv
 
     };
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 #endif
-
-
diff --git a/modules/contrib/src/imagelogpolprojection.cpp b/modules/bioinspired/src/imagelogpolprojection.cpp
similarity index 98%
rename from modules/contrib/src/imagelogpolprojection.cpp
rename to modules/bioinspired/src/imagelogpolprojection.cpp
index 22f5214e7c..0a4c1ed0de 100644
--- a/modules/contrib/src/imagelogpolprojection.cpp
+++ b/modules/bioinspired/src/imagelogpolprojection.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -72,7 +72,8 @@
 
 namespace cv
 {
-
+namespace bioinspired
+{
 // constructor
 ImageLogPolProjection::ImageLogPolProjection(const unsigned int nbRows, const unsigned int nbColumns, const PROJECTIONTYPE projection, const bool colorModeCapable)
 :BasicRetinaFilter(nbRows, nbColumns),
@@ -446,4 +447,5 @@ std::valarray<float> &ImageLogPolProjection::runProjection(const std::valarray<f
     return _sampledFrame;
 }
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/imagelogpolprojection.hpp b/modules/bioinspired/src/imagelogpolprojection.hpp
similarity index 96%
rename from modules/contrib/src/imagelogpolprojection.hpp
rename to modules/bioinspired/src/imagelogpolprojection.hpp
index 1572ee83e5..41ecd5eaf3 100644
--- a/modules/contrib/src/imagelogpolprojection.hpp
+++ b/modules/bioinspired/src/imagelogpolprojection.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -103,6 +103,8 @@
 
 namespace cv
 {
+namespace bioinspired
+{
 
 class ImageLogPolProjection:public BasicRetinaFilter
 {
@@ -236,5 +238,6 @@ private:
 
 };
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 #endif /*IMAGELOGPOLPROJECTION_H_*/
diff --git a/modules/contrib/src/magnoretinafilter.cpp b/modules/bioinspired/src/magnoretinafilter.cpp
similarity index 95%
rename from modules/contrib/src/magnoretinafilter.cpp
rename to modules/bioinspired/src/magnoretinafilter.cpp
index 48e10cf539..81fdb1df54 100644
--- a/modules/contrib/src/magnoretinafilter.cpp
+++ b/modules/bioinspired/src/magnoretinafilter.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -72,6 +72,8 @@
 
 namespace cv
 {
+namespace bioinspired
+{
 // Constructor and Desctructor of the OPL retina filter
 MagnoRetinaFilter::MagnoRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns)
 :BasicRetinaFilter(NBrows, NBcolumns, 2),
@@ -206,6 +208,5 @@ const std::valarray<float> &MagnoRetinaFilter::runFilter(const std::valarray<flo
 
     return (*_magnoYOutput);
 }
-}
-
-
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/magnoretinafilter.hpp b/modules/bioinspired/src/magnoretinafilter.hpp
similarity index 96%
rename from modules/contrib/src/magnoretinafilter.hpp
rename to modules/bioinspired/src/magnoretinafilter.hpp
index 991e75daf5..e06d14ddc2 100644
--- a/modules/contrib/src/magnoretinafilter.hpp
+++ b/modules/bioinspired/src/magnoretinafilter.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -99,7 +99,8 @@
 
 namespace cv
 {
-
+namespace bioinspired
+{
     class MagnoRetinaFilter: public BasicRetinaFilter
     {
     public:
@@ -238,8 +239,7 @@ namespace cv
 #endif
     };
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 
 #endif /*MagnoRetinaFilter_H_*/
-
-
diff --git a/modules/bioinspired/src/opencl/retina_kernel.cl b/modules/bioinspired/src/opencl/retina_kernel.cl
new file mode 100644
index 0000000000..169be4d270
--- /dev/null
+++ b/modules/bioinspired/src/opencl/retina_kernel.cl
@@ -0,0 +1,779 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+//data (which is float) is aligend in 32 bytes
+#define WIDTH_MULTIPLE (32 >> 2)
+
+/////////////////////////////////////////////////////////
+//*******************************************************
+// basicretinafilter
+//////////////// _spatiotemporalLPfilter ////////////////
+//_horizontalCausalFilter_addInput
+kernel void horizontalCausalFilter_addInput(
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int in_offset,
+    const int out_offset,
+    const float _tau,
+    const float _a
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= rows)
+    {
+        return;
+    }
+
+    global const float * iptr =
+        input  + mad24(gid, elements_per_row, in_offset / 4);
+    global float * optr =
+        output + mad24(gid, elements_per_row, out_offset / 4);
+
+    float res;
+    float4 in_v4, out_v4, res_v4 = (float4)(0);
+    //vectorize to increase throughput
+    for(int i = 0; i < cols / 4; ++i, iptr += 4, optr += 4)
+    {
+        in_v4  = vload4(0, iptr);
+        out_v4 = vload4(0, optr);
+
+        res_v4.x = in_v4.x + _tau * out_v4.x + _a * res_v4.w;
+        res_v4.y = in_v4.y + _tau * out_v4.y + _a * res_v4.x;
+        res_v4.z = in_v4.z + _tau * out_v4.z + _a * res_v4.y;
+        res_v4.w = in_v4.w + _tau * out_v4.w + _a * res_v4.z;
+
+        vstore4(res_v4, 0, optr);
+    }
+    res = res_v4.w;
+    // there may be left some
+    for(int i = 0; i < cols % 4;  ++i, ++iptr, ++optr)
+    {
+        res = *iptr + _tau * *optr + _a * res;
+        *optr = res;
+    }
+}
+
+//_horizontalAnticausalFilter
+kernel void horizontalAnticausalFilter(
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int out_offset,
+    const float _a
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= rows)
+    {
+        return;
+    }
+
+    global float * optr = output +
+                          mad24(gid + 1, elements_per_row, - 1 + out_offset / 4);
+
+    float4 result_v4 = (float4)(0), out_v4;
+    float result = 0;
+    // we assume elements_per_row is multple of WIDTH_MULTIPLE
+    for(int i = 0; i < WIDTH_MULTIPLE; ++ i, -- optr)
+    {
+        if(i >= elements_per_row - cols)
+        {
+            result = *optr + _a * result;
+        }
+        *optr = result;
+    }
+    result_v4.x = result;
+    optr -= 3;
+    for(int i = WIDTH_MULTIPLE / 4; i < elements_per_row / 4; ++i, optr -= 4)
+    {
+        // shift left, `offset` is type `size_t` so it cannot be negative
+        out_v4 = vload4(0, optr);
+
+        result_v4.w = out_v4.w + _a * result_v4.x;
+        result_v4.z = out_v4.z + _a * result_v4.w;
+        result_v4.y = out_v4.y + _a * result_v4.z;
+        result_v4.x = out_v4.x + _a * result_v4.y;
+
+        vstore4(result_v4, 0, optr);
+    }
+}
+
+//_verticalCausalFilter
+kernel void verticalCausalFilter(
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int out_offset,
+    const float _a
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= cols)
+    {
+        return;
+    }
+
+    global float * optr = output + gid + out_offset / 4;
+    float result = 0;
+    for(int i = 0; i < rows; ++i, optr += elements_per_row)
+    {
+        result = *optr + _a * result;
+        *optr = result;
+    }
+}
+
+//_verticalCausalFilter
+kernel void verticalAnticausalFilter_multGain(
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int out_offset,
+    const float _a,
+    const float _gain
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= cols)
+    {
+        return;
+    }
+
+    global float * optr = output + (rows - 1) * elements_per_row + gid + out_offset / 4;
+    float result = 0;
+    for(int i = 0; i < rows; ++i, optr -= elements_per_row)
+    {
+        result = *optr + _a * result;
+        *optr = _gain * result;
+    }
+}
+//
+// end of _spatiotemporalLPfilter
+/////////////////////////////////////////////////////////////////////
+
+//////////////// horizontalAnticausalFilter_Irregular ////////////////
+kernel void horizontalAnticausalFilter_Irregular(
+    global float * output,
+    global float * buffer,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int out_offset,
+    const int buffer_offset
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= rows)
+    {
+        return;
+    }
+
+    global float * optr =
+        output + mad24(rows - gid, elements_per_row, -1 + out_offset / 4);
+    global float * bptr =
+        buffer + mad24(rows - gid, elements_per_row, -1 + buffer_offset / 4);
+
+    float4 buf_v4, out_v4, res_v4 = (float4)(0);
+    float result = 0;
+    // we assume elements_per_row is multple of WIDTH_MULTIPLE
+    for(int i = 0; i < WIDTH_MULTIPLE; ++ i, -- optr, -- bptr)
+    {
+        if(i >= elements_per_row - cols)
+        {
+            result = *optr + *bptr * result;
+        }
+        *optr = result;
+    }
+    res_v4.x = result;
+    optr -= 3;
+    bptr -= 3;
+    for(int i = WIDTH_MULTIPLE / 4; i < elements_per_row / 4; ++i, optr -= 4, bptr -= 4)
+    {
+        buf_v4 = vload4(0, bptr);
+        out_v4 = vload4(0, optr);
+
+        res_v4.w = out_v4.w + buf_v4.w * res_v4.x;
+        res_v4.z = out_v4.z + buf_v4.z * res_v4.w;
+        res_v4.y = out_v4.y + buf_v4.y * res_v4.z;
+        res_v4.x = out_v4.x + buf_v4.x * res_v4.y;
+
+        vstore4(res_v4, 0, optr);
+    }
+}
+
+//////////////// verticalCausalFilter_Irregular ////////////////
+kernel void verticalCausalFilter_Irregular(
+    global float * output,
+    global float * buffer,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int out_offset,
+    const int buffer_offset
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= cols)
+    {
+        return;
+    }
+
+    global float * optr = output + gid + out_offset / 4;
+    global float * bptr = buffer + gid + buffer_offset / 4;
+    float result = 0;
+    for(int i = 0; i < rows; ++i, optr += elements_per_row, bptr += elements_per_row)
+    {
+        result = *optr + *bptr * result;
+        *optr = result;
+    }
+}
+
+//////////////// _adaptiveHorizontalCausalFilter_addInput ////////////////
+kernel void adaptiveHorizontalCausalFilter_addInput(
+    global const float * input,
+    global const float * gradient,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int in_offset,
+    const int grad_offset,
+    const int out_offset
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= rows)
+    {
+        return;
+    }
+
+    global const float * iptr =
+        input + mad24(gid, elements_per_row, in_offset / 4);
+    global const float * gptr =
+        gradient + mad24(gid, elements_per_row, grad_offset / 4);
+    global float * optr =
+        output + mad24(gid, elements_per_row, out_offset / 4);
+
+    float4 in_v4, grad_v4, out_v4, res_v4 = (float4)(0);
+    for(int i = 0; i < cols / 4; ++i, iptr += 4, gptr += 4, optr += 4)
+    {
+        in_v4   = vload4(0, iptr);
+        grad_v4 = vload4(0, gptr);
+
+        res_v4.x = in_v4.x + grad_v4.x * res_v4.w;
+        res_v4.y = in_v4.y + grad_v4.y * res_v4.x;
+        res_v4.z = in_v4.z + grad_v4.z * res_v4.y;
+        res_v4.w = in_v4.w + grad_v4.w * res_v4.z;
+
+        vstore4(res_v4, 0, optr);
+    }
+    for(int i = 0; i < cols % 4; ++i, ++iptr, ++gptr, ++optr)
+    {
+        res_v4.w = *iptr + *gptr * res_v4.w;
+        *optr = res_v4.w;
+    }
+}
+
+//////////////// _adaptiveVerticalAnticausalFilter_multGain ////////////////
+kernel void adaptiveVerticalAnticausalFilter_multGain(
+    global const float * gradient,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const int grad_offset,
+    const int out_offset,
+    const float gain
+)
+{
+    int gid = get_global_id(0);
+    if(gid >= cols)
+    {
+        return;
+    }
+
+    int start_idx = mad24(rows - 1, elements_per_row, gid);
+
+    global const float * gptr = gradient + start_idx + grad_offset / 4;
+    global float * optr = output + start_idx + out_offset / 4;
+
+    float result = 0;
+    for(int i = 0; i < rows; ++i, gptr -= elements_per_row, optr -= elements_per_row)
+    {
+        result = *optr + *gptr * result;
+        *optr = gain * result;
+    }
+}
+
+//////////////// _localLuminanceAdaptation ////////////////
+// FIXME:
+//  This kernel seems to have precision problem on GPU
+kernel void localLuminanceAdaptation(
+    global const float * luma,
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float _localLuminanceAddon,
+    const float _localLuminanceFactor,
+    const float _maxInputValue
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    int offset = mad24(gidy, elements_per_row, gidx);
+
+    float X0 = luma[offset] * _localLuminanceFactor + _localLuminanceAddon;
+    float input_val = input[offset];
+    // output of the following line may be different between GPU and CPU
+    output[offset] = (_maxInputValue + X0) * input_val / (input_val + X0 + 0.00000000001f);
+}
+// end of basicretinafilter
+//*******************************************************
+/////////////////////////////////////////////////////////
+
+
+
+/////////////////////////////////////////////////////////
+//******************************************************
+// magno
+// TODO: this kernel has too many buffer accesses, better to make it
+//   vector read/write for fetch efficiency
+kernel void amacrineCellsComputing(
+    global const float * opl_on,
+    global const float * opl_off,
+    global float * prev_in_on,
+    global float * prev_in_off,
+    global float * out_on,
+    global float * out_off,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float coeff
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+
+    int offset = mad24(gidy, elements_per_row, gidx);
+    opl_on      += offset;
+    opl_off     += offset;
+    prev_in_on  += offset;
+    prev_in_off += offset;
+    out_on      += offset;
+    out_off     += offset;
+
+    float magnoXonPixelResult = coeff * (*out_on + *opl_on - *prev_in_on);
+    *out_on = fmax(magnoXonPixelResult, 0);
+    float magnoXoffPixelResult = coeff * (*out_off + *opl_off - *prev_in_off);
+    *out_off = fmax(magnoXoffPixelResult, 0);
+
+    *prev_in_on = *opl_on;
+    *prev_in_off = *opl_off;
+}
+
+/////////////////////////////////////////////////////////
+//******************************************************
+// parvo
+// TODO: this kernel has too many buffer accesses, needs optimization
+kernel void OPL_OnOffWaysComputing(
+    global float4 * photo_out,
+    global float4 * horiz_out,
+    global float4 * bipol_on,
+    global float4 * bipol_off,
+    global float4 * parvo_on,
+    global float4 * parvo_off,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx * 4 >= cols || gidy >= rows)
+    {
+        return;
+    }
+    // we assume elements_per_row must be multiples of 4
+    int offset = mad24(gidy, elements_per_row >> 2, gidx);
+    photo_out += offset;
+    horiz_out += offset;
+    bipol_on  += offset;
+    bipol_off += offset;
+    parvo_on  += offset;
+    parvo_off += offset;
+
+    float4 diff = *photo_out - *horiz_out;
+    float4 isPositive;// = convert_float4(diff > (float4)(0.0f, 0.0f, 0.0f, 0.0f));
+    isPositive.x = diff.x > 0.0f;
+    isPositive.y = diff.y > 0.0f;
+    isPositive.z = diff.z > 0.0f;
+    isPositive.w = diff.w > 0.0f;
+    float4 res_on  = isPositive * diff;
+    float4 res_off = (isPositive - (float4)(1.0f)) * diff;
+
+    *bipol_on = res_on;
+    *parvo_on = res_on;
+
+    *bipol_off = res_off;
+    *parvo_off = res_off;
+}
+
+/////////////////////////////////////////////////////////
+//******************************************************
+// retinacolor
+inline int bayerSampleOffset(int step, int rows, int x, int y)
+{
+    return mad24(y, step, x) +
+           ((y % 2) + (x % 2)) * rows * step;
+}
+
+
+/////// colorMultiplexing //////
+kernel void runColorMultiplexingBayer(
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+
+    int offset = mad24(gidy, elements_per_row, gidx);
+    output[offset] = input[bayerSampleOffset(elements_per_row, rows, gidx, gidy)];
+}
+
+kernel void runColorDemultiplexingBayer(
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+
+    int offset = mad24(gidy, elements_per_row, gidx);
+    output[bayerSampleOffset(elements_per_row, rows, gidx, gidy)] = input[offset];
+}
+
+kernel void demultiplexAssign(
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+
+    int offset = bayerSampleOffset(elements_per_row, rows, gidx, gidy);
+    output[offset] = input[offset];
+}
+
+
+//// normalizeGrayOutputCentredSigmoide
+kernel void normalizeGrayOutputCentredSigmoide(
+    global const float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float meanval,
+    const float X0
+)
+
+{
+    int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    int offset = mad24(gidy, elements_per_row, gidx);
+
+    float input_val = input[offset];
+    output[offset] = meanval +
+                     (meanval + X0) * (input_val - meanval) / (fabs(input_val - meanval) + X0);
+}
+
+//// normalize by photoreceptors density
+kernel void normalizePhotoDensity(
+    global const float * chroma,
+    global const float * colorDensity,
+    global const float * multiplex,
+    global float * luma,
+    global float * demultiplex,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float pG
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+    int index = offset;
+
+    float Cr = chroma[index] * colorDensity[index];
+    index += elements_per_row * rows;
+    float Cg = chroma[index] * colorDensity[index];
+    index += elements_per_row * rows;
+    float Cb = chroma[index] * colorDensity[index];
+
+    const float luma_res = (Cr + Cg + Cb) * pG;
+    luma[offset] = luma_res;
+    demultiplex[bayerSampleOffset(elements_per_row, rows, gidx, gidy)] =
+        multiplex[offset] - luma_res;
+}
+
+
+
+//////// computeGradient ///////
+// TODO:
+// this function maybe accelerated by image2d_t or lds
+kernel void computeGradient(
+    global const float * luma,
+    global float * gradient,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    int gidx = get_global_id(0) + 2, gidy = get_global_id(1) + 2;
+    if(gidx >= cols - 2 || gidy >= rows - 2)
+    {
+        return;
+    }
+    int offset = mad24(gidy, elements_per_row, gidx);
+    luma += offset;
+
+    // horizontal and vertical local gradients
+    const float v_grad = fabs(luma[elements_per_row] - luma[- elements_per_row]);
+    const float h_grad = fabs(luma[1] - luma[-1]);
+
+    // neighborhood horizontal and vertical gradients
+    const float cur_val  = luma[0];
+    const float v_grad_p = fabs(cur_val - luma[- 2 * elements_per_row]);
+    const float h_grad_p = fabs(cur_val - luma[- 2]);
+    const float v_grad_n = fabs(cur_val - luma[2 * elements_per_row]);
+    const float h_grad_n = fabs(cur_val - luma[2]);
+
+    const float horiz_grad = 0.5f * h_grad + 0.25f * (h_grad_p + h_grad_n);
+    const float verti_grad = 0.5f * v_grad + 0.25f * (v_grad_p + v_grad_n);
+    const bool is_vertical_greater = horiz_grad < verti_grad;
+
+    gradient[offset + elements_per_row * rows] = is_vertical_greater ? 0.06f : 0.57f;
+    gradient[offset                          ] = is_vertical_greater ? 0.57f : 0.06f;
+}
+
+
+/////// substractResidual ///////
+kernel void substractResidual(
+    global float * input,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float pR,
+    const float pG,
+    const float pB
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    int indices [3] =
+    {
+        mad24(gidy, elements_per_row, gidx),
+        mad24(gidy + rows, elements_per_row, gidx),
+        mad24(gidy + 2 * rows, elements_per_row, gidx)
+    };
+    float vals[3] = {input[indices[0]], input[indices[1]], input[indices[2]]};
+    float residu = pR * vals[0] + pG * vals[1] + pB * vals[2];
+
+    input[indices[0]] = vals[0] - residu;
+    input[indices[1]] = vals[1] - residu;
+    input[indices[2]] = vals[2] - residu;
+}
+
+///// clipRGBOutput_0_maxInputValue /////
+kernel void clipRGBOutput_0_maxInputValue(
+    global float * input,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float maxVal
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+    float val = input[offset];
+    val = clamp(val, 0.0f, maxVal);
+    input[offset] = val;
+}
+
+//// normalizeGrayOutputNearZeroCentreredSigmoide ////
+kernel void normalizeGrayOutputNearZeroCentreredSigmoide(
+    global float * input,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float maxVal,
+    const float X0cube
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+    float currentCubeLuminance = input[offset];
+    currentCubeLuminance = currentCubeLuminance * currentCubeLuminance * currentCubeLuminance;
+    output[offset] = currentCubeLuminance * X0cube / (X0cube + currentCubeLuminance);
+}
+
+//// centerReductImageLuminance ////
+kernel void centerReductImageLuminance(
+    global float * input,
+    const int cols,
+    const int rows,
+    const int elements_per_row,
+    const float mean,
+    const float std_dev
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+
+    float val = input[offset];
+    input[offset] = (val - mean) / std_dev;
+}
+
+//// inverseValue ////
+kernel void inverseValue(
+    global float * input,
+    const int cols,
+    const int rows,
+    const int elements_per_row
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+    input[offset] = 1.f / input[offset];
+}
+
+#define CV_PI 3.1415926535897932384626433832795
+
+//// _processRetinaParvoMagnoMapping ////
+kernel void processRetinaParvoMagnoMapping(
+    global float * parvo,
+    global float * magno,
+    global float * output,
+    const int cols,
+    const int rows,
+    const int halfCols,
+    const int halfRows,
+    const int elements_per_row,
+    const float minDistance
+)
+{
+    const int gidx = get_global_id(0), gidy = get_global_id(1);
+    if(gidx >= cols || gidy >= rows)
+    {
+        return;
+    }
+    const int offset = mad24(gidy, elements_per_row, gidx);
+
+    float distanceToCenter =
+        sqrt(((float)(gidy - halfRows) * (gidy - halfRows) + (gidx - halfCols) * (gidx - halfCols)));
+
+    float a = distanceToCenter < minDistance ?
+              (0.5f + 0.5f * (float)cos(CV_PI * distanceToCenter / minDistance)) : 0;
+    float b = 1.f - a;
+
+    output[offset] = parvo[offset] * a + magno[offset] * b;
+}
diff --git a/modules/contrib/src/parvoretinafilter.cpp b/modules/bioinspired/src/parvoretinafilter.cpp
similarity index 96%
rename from modules/contrib/src/parvoretinafilter.cpp
rename to modules/bioinspired/src/parvoretinafilter.cpp
index 6043b72328..a276d97a4c 100644
--- a/modules/contrib/src/parvoretinafilter.cpp
+++ b/modules/bioinspired/src/parvoretinafilter.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -73,6 +73,8 @@
 
 namespace cv
 {
+namespace bioinspired
+{
 //////////////////////////////////////////////////////////
 //                 OPL RETINA FILTER
 //////////////////////////////////////////////////////////
@@ -227,5 +229,5 @@ void ParvoRetinaFilter::_OPL_OnOffWaysComputing() // WARNING : this method requi
     }
 #endif
 }
-}
-
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/parvoretinafilter.hpp b/modules/bioinspired/src/parvoretinafilter.hpp
similarity index 97%
rename from modules/contrib/src/parvoretinafilter.hpp
rename to modules/bioinspired/src/parvoretinafilter.hpp
index 40f71ebe9f..f5ffa1a06d 100644
--- a/modules/contrib/src/parvoretinafilter.hpp
+++ b/modules/bioinspired/src/parvoretinafilter.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -103,6 +103,8 @@
 
 namespace cv
 {
+namespace bioinspired
+{
 //retina classes that derivate from the Basic Retrina class
 class ParvoRetinaFilter: public BasicRetinaFilter
 {
@@ -256,6 +258,6 @@ private:
 #endif
 
 };
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 #endif
-
diff --git a/modules/bioinspired/src/precomp.hpp b/modules/bioinspired/src/precomp.hpp
new file mode 100644
index 0000000000..541b970325
--- /dev/null
+++ b/modules/bioinspired/src/precomp.hpp
@@ -0,0 +1,67 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include "opencv2/opencv_modules.hpp"
+#include "opencv2/bioinspired.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
+
+#include <valarray>
+
+#ifdef HAVE_OPENCV_OCL
+    #include "opencv2/ocl/private/util.hpp"
+#endif
+
+namespace cv
+{
+
+// special function to get pointer to constant valarray elements, since
+// simple &arr[0] does not compile on VS2005/VS2008.
+template<typename T> inline const T* get_data(const std::valarray<T>& arr)
+{ return &((std::valarray<T>&)arr)[0]; }
+
+}
+
+#endif
diff --git a/modules/bioinspired/src/retina.cpp b/modules/bioinspired/src/retina.cpp
new file mode 100644
index 0000000000..4604331076
--- /dev/null
+++ b/modules/bioinspired/src/retina.cpp
@@ -0,0 +1,743 @@
+/*#******************************************************************************
+ ** IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+ **
+ ** By downloading, copying, installing or using the software you agree to this license.
+ ** If you do not agree to this license, do not download, install,
+ ** copy or use the software.
+ **
+ **
+ ** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+ ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
+ **
+ ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
+ **
+ **  Creation - enhancement process 2007-2011
+ **      Author: Alexandre Benoit (benoit.alexandre.vision@gmail.com), LISTIC lab, Annecy le vieux, France
+ **
+ ** Theses algorithm have been developped by Alexandre BENOIT since his thesis with Alice Caplier at Gipsa-Lab (www.gipsa-lab.inpg.fr) and the research he pursues at LISTIC Lab (www.listic.univ-savoie.fr).
+ ** Refer to the following research paper for more information:
+ ** Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+ ** This work have been carried out thanks to Jeanny Herault who's research and great discussions are the basis of all this work, please take a look at his book:
+ ** Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+ **
+ ** The retina filter includes the research contributions of phd/research collegues from which code has been redrawn by the author :
+ ** _take a look at the retinacolor.hpp module to discover Brice Chaix de Lavarene color mosaicing/demosaicing and the reference paper:
+ ** ====> B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
+ ** _take a look at imagelogpolprojection.hpp to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions.
+ ** ====> more informations in the above cited Jeanny Heraults's book.
+ **
+ **                          License Agreement
+ **               For Open Source Computer Vision Library
+ **
+ ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+ ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+ **
+ **               For Human Visual System tools (bioinspired)
+ ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
+ **
+ ** Third party copyrights are property of their respective owners.
+ **
+ ** Redistribution and use in source and binary forms, with or without modification,
+ ** are permitted provided that the following conditions are met:
+ **
+ ** * Redistributions of source code must retain the above copyright notice,
+ **    this list of conditions and the following disclaimer.
+ **
+ ** * Redistributions in binary form must reproduce the above copyright notice,
+ **    this list of conditions and the following disclaimer in the documentation
+ **    and/or other materials provided with the distribution.
+ **
+ ** * The name of the copyright holders may not be used to endorse or promote products
+ **    derived from this software without specific prior written permission.
+ **
+ ** This software is provided by the copyright holders and contributors "as is" and
+ ** any express or implied warranties, including, but not limited to, the implied
+ ** warranties of merchantability and fitness for a particular purpose are disclaimed.
+ ** In no event shall the Intel Corporation or contributors be liable for any direct,
+ ** indirect, incidental, special, exemplary, or consequential damages
+ ** (including, but not limited to, procurement of substitute goods or services;
+ ** loss of use, data, or profits; or business interruption) however caused
+ ** and on any theory of liability, whether in contract, strict liability,
+ ** or tort (including negligence or otherwise) arising in any way out of
+ ** the use of this software, even if advised of the possibility of such damage.
+ *******************************************************************************/
+
+/*
+ * Retina.cpp
+ *
+ *  Created on: Jul 19, 2011
+ *      Author: Alexandre Benoit
+ */
+#include "precomp.hpp"
+#include "retinafilter.hpp"
+#include <cstdio>
+#include <sstream>
+#include <valarray>
+
+namespace cv
+{
+namespace bioinspired
+{
+
+class RetinaImpl : public Retina
+{
+public:
+    /**
+     * Main constructor with most commun use setup : create an instance of color ready retina model
+     * @param inputSize : the input frame size
+     */
+    RetinaImpl(Size inputSize);
+
+    /**
+     * Complete Retina filter constructor which allows all basic structural parameters definition
+         * @param inputSize : the input frame size
+     * @param colorMode : the chosen processing mode : with or without color processing
+     * @param colorSamplingMethod: specifies which kind of color sampling will be used
+     * @param useRetinaLogSampling: activate retina log sampling, if true, the 2 following parameters can be used
+     * @param reductionFactor: only usefull if param useRetinaLogSampling=true, specifies the reduction factor of the output frame (as the center (fovea) is high resolution and corners can be underscaled, then a reduction of the output is allowed without precision leak
+     * @param samplingStrenght: only usefull if param useRetinaLogSampling=true, specifies the strenght of the log scale that is applied
+     */
+    RetinaImpl(Size inputSize, const bool colorMode, int colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
+
+    virtual ~RetinaImpl();
+    /**
+        * retreive retina input buffer size
+        */
+        Size getInputSize();
+
+    /**
+        * retreive retina output buffer size
+        */
+        Size getOutputSize();
+
+    /**
+     * try to open an XML retina parameters file to adjust current retina instance setup
+     * => if the xml file does not exist, then default setup is applied
+     * => warning, Exceptions are thrown if read XML file is not valid
+     * @param retinaParameterFile : the parameters filename
+         * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
+     */
+    void setup(String retinaParameterFile="", const bool applyDefaultSetupOnFailure=true);
+
+
+    /**
+     * try to open an XML retina parameters file to adjust current retina instance setup
+     * => if the xml file does not exist, then default setup is applied
+     * => warning, Exceptions are thrown if read XML file is not valid
+     * @param fs : the open Filestorage which contains retina parameters
+         * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
+     */
+        void setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure=true);
+
+    /**
+     * try to open an XML retina parameters file to adjust current retina instance setup
+     * => if the xml file does not exist, then default setup is applied
+     * => warning, Exceptions are thrown if read XML file is not valid
+     * @param newParameters : a parameters structures updated with the new target configuration
+         * @param applyDefaultSetupOnFailure : set to true if an error must be thrown on error
+     */
+    void setup(Retina::RetinaParameters newParameters);
+
+    /**
+    * @return the current parameters setup
+    */
+    struct Retina::RetinaParameters getParameters();
+
+    /**
+     * parameters setup display method
+     * @return a string which contains formatted parameters information
+     */
+    const String printSetup();
+
+    /**
+     * write xml/yml formated parameters information
+     * @rparam fs : the filename of the xml file that will be open and writen with formatted parameters information
+     */
+    virtual void write( String fs ) const;
+
+
+    /**
+     * write xml/yml formated parameters information
+     * @param fs : a cv::Filestorage object ready to be filled
+         */
+    virtual void write( FileStorage& fs ) const;
+
+    /**
+     * setup the OPL and IPL parvo channels (see biologocal model)
+     * OPL is referred as Outer Plexiform Layer of the retina, it allows the spatio-temporal filtering which withens the spectrum and reduces spatio-temporal noise while attenuating global luminance (low frequency energy)
+     * IPL parvo is the OPL next processing stage, it refers to Inner Plexiform layer of the retina, it allows high contours sensitivity in foveal vision.
+     * for more informations, please have a look at the paper Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+     * @param colorMode : specifies if (true) color is processed of not (false) to then processing gray level image
+     * @param normaliseOutput : specifies if (true) output is rescaled between 0 and 255 of not (false)
+     * @param photoreceptorsLocalAdaptationSensitivity: the photoreceptors sensitivity renage is 0-1 (more log compression effect when value increases)
+     * @param photoreceptorsTemporalConstant: the time constant of the first order low pass filter of the photoreceptors, use it to cut high temporal frequencies (noise or fast motion), unit is frames, typical value is 1 frame
+     * @param photoreceptorsSpatialConstant: the spatial constant of the first order low pass filter of the photoreceptors, use it to cut high spatial frequencies (noise or thick contours), unit is pixels, typical value is 1 pixel
+     * @param horizontalCellsGain: gain of the horizontal cells network, if 0, then the mean value of the output is zero, if the parameter is near 1, then, the luminance is not filtered and is still reachable at the output, typicall value is 0
+     * @param HcellsTemporalConstant: the time constant of the first order low pass filter of the horizontal cells, use it to cut low temporal frequencies (local luminance variations), unit is frames, typical value is 1 frame, as the photoreceptors
+     * @param HcellsSpatialConstant: the spatial constant of the first order low pass filter of the horizontal cells, use it to cut low spatial frequencies (local luminance), unit is pixels, typical value is 5 pixel, this value is also used for local contrast computing when computing the local contrast adaptation at the ganglion cells level (Inner Plexiform Layer parvocellular channel model)
+     * @param ganglionCellsSensitivity: the compression strengh of the ganglion cells local adaptation output, set a value between 160 and 250 for best results, a high value increases more the low value sensitivity... and the output saturates faster, recommended value: 230
+     */
+    void setupOPLandIPLParvoChannel(const bool colorMode=true, const bool normaliseOutput = true, const float photoreceptorsLocalAdaptationSensitivity=0.7, const float photoreceptorsTemporalConstant=0.5, const float photoreceptorsSpatialConstant=0.53, const float horizontalCellsGain=0, const float HcellsTemporalConstant=1, const float HcellsSpatialConstant=7, const float ganglionCellsSensitivity=0.7);
+
+    /**
+     * set parameters values for the Inner Plexiform Layer (IPL) magnocellular channel
+     * this channel processes signals outpint from OPL processing stage in peripheral vision, it allows motion information enhancement. It is decorrelated from the details channel. See reference paper for more details.
+     * @param normaliseOutput : specifies if (true) output is rescaled between 0 and 255 of not (false)
+     * @param parasolCells_beta: the low pass filter gain used for local contrast adaptation at the IPL level of the retina (for ganglion cells local adaptation), typical value is 0
+     * @param parasolCells_tau: the low pass filter time constant used for local contrast adaptation at the IPL level of the retina (for ganglion cells local adaptation), unit is frame, typical value is 0 (immediate response)
+     * @param parasolCells_k: the low pass filter spatial constant used for local contrast adaptation at the IPL level of the retina (for ganglion cells local adaptation), unit is pixels, typical value is 5
+     * @param amacrinCellsTemporalCutFrequency: the time constant of the first order high pass fiter of the magnocellular way (motion information channel), unit is frames, tipicall value is 5
+     * @param V0CompressionParameter: the compression strengh of the ganglion cells local adaptation output, set a value between 160 and 250 for best results, a high value increases more the low value sensitivity... and the output saturates faster, recommended value: 200
+     * @param localAdaptintegration_tau: specifies the temporal constant of the low pas filter involved in the computation of the local "motion mean" for the local adaptation computation
+     * @param localAdaptintegration_k: specifies the spatial constant of the low pas filter involved in the computation of the local "motion mean" for the local adaptation computation
+     */
+    void setupIPLMagnoChannel(const bool normaliseOutput = true, const float parasolCells_beta=0, const float parasolCells_tau=0, const float parasolCells_k=7, const float amacrinCellsTemporalCutFrequency=1.2, const float V0CompressionParameter=0.95, const float localAdaptintegration_tau=0, const float localAdaptintegration_k=7);
+
+    /**
+     * method which allows retina to be applied on an input image, after run, encapsulated retina module is ready to deliver its outputs using dedicated acccessors, see getParvo and getMagno methods
+     * @param inputImage : the input cv::Mat image to be processed, can be gray level or BGR coded in any format (from 8bit to 16bits)
+     */
+    void run(InputArray inputImage);
+
+    /**
+     * method that applies a luminance correction (initially High Dynamic Range (HDR) tone mapping) using only the 2 local adaptation stages of the retina parvo channel : photoreceptors level and ganlion cells level. Spatio temporal filtering is applied but limited to temporal smoothing and eventually high frequencies attenuation. This is a lighter method than the one available using the regular run method. It is then faster but it does not include complete temporal filtering nor retina spectral whitening. This is an adptation of the original still image HDR tone mapping algorithm of David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+    * -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+     @param inputImage the input image to process RGB or gray levels
+     @param outputToneMappedImage the output tone mapped image
+     */
+    void applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage);
+
+    /**
+     * accessor of the details channel of the retina (models foveal vision)
+     * @param retinaOutput_parvo : the output buffer (reallocated if necessary), this output is rescaled for standard 8bits image processing use in OpenCV
+     */
+    void getParvo(OutputArray retinaOutput_parvo);
+
+    /**
+     * accessor of the details channel of the retina (models foveal vision)
+     * @param retinaOutput_parvo : a cv::Mat header filled with the internal parvo buffer of the retina module. This output is the original retina filter model output, without any quantification or rescaling
+     */
+    void getParvoRAW(OutputArray retinaOutput_parvo);
+
+    /**
+     * accessor of the motion channel of the retina (models peripheral vision)
+     * @param retinaOutput_magno : the output buffer (reallocated if necessary), this output is rescaled for standard 8bits image processing use in OpenCV
+     */
+    void getMagno(OutputArray retinaOutput_magno);
+
+    /**
+     * accessor of the motion channel of the retina (models peripheral vision)
+     * @param retinaOutput_magno : a cv::Mat header filled with the internal retina magno buffer of the retina module. This output is the original retina filter model output, without any quantification or rescaling
+     */
+    void getMagnoRAW(OutputArray retinaOutput_magno);
+
+    // original API level data accessors : get buffers addresses from a Mat header, similar to getParvoRAW and getMagnoRAW...
+    const Mat getMagnoRAW() const;
+    const Mat getParvoRAW() const;
+
+    /**
+     * activate color saturation as the final step of the color demultiplexing process
+     * -> this saturation is a sigmoide function applied to each channel of the demultiplexed image.
+     * @param saturateColors: boolean that activates color saturation (if true) or desactivate (if false)
+     * @param colorSaturationValue: the saturation factor
+     */
+    void setColorSaturation(const bool saturateColors=true, const float colorSaturationValue=4.0);
+
+    /**
+     * clear all retina buffers (equivalent to opening the eyes after a long period of eye close ;o)
+     */
+    void clearBuffers();
+
+        /**
+        * Activate/desactivate the Magnocellular pathway processing (motion information extraction), by default, it is activated
+        * @param activate: true if Magnocellular output should be activated, false if not
+        */
+        void activateMovingContoursProcessing(const bool activate);
+
+        /**
+        * Activate/desactivate the Parvocellular pathway processing (contours information extraction), by default, it is activated
+        * @param activate: true if Parvocellular (contours information extraction) output should be activated, false if not
+        */
+        void activateContoursProcessing(const bool activate);
+private:
+
+    // Parameteres setup members
+    RetinaParameters _retinaParameters; // structure of parameters
+
+    // Retina model related modules
+    std::valarray<float> _inputBuffer; //!< buffer used to convert input cv::Mat to internal retina buffers format (valarrays)
+
+    // pointer to retina model
+    RetinaFilter* _retinaFilter; //!< the pointer to the retina module, allocated with instance construction
+
+    //! private method called by constructors, gathers their parameters and use them in a unified way
+    void _init(const Size inputSize, const bool colorMode, int colorSamplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
+
+    /**
+     * exports a valarray buffer outing from bioinspired objects to a cv::Mat in CV_8UC1 (gray level picture) or CV_8UC3 (color) format
+     * @param grayMatrixToConvert the valarray to export to OpenCV
+     * @param nbRows : the number of rows of the valarray flatten matrix
+     * @param nbColumns : the number of rows of the valarray flatten matrix
+     * @param colorMode : a flag which mentions if matrix is color (true) or graylevel (false)
+     * @param outBuffer : the output matrix which is reallocated to satisfy Retina output buffer dimensions
+     */
+    void _convertValarrayBuffer2cvMat(const std::valarray<float> &grayMatrixToConvert, const unsigned int nbRows, const unsigned int nbColumns, const bool colorMode, OutputArray outBuffer);
+
+    /**
+     * convert a cv::Mat to a valarray buffer in float format
+     * @param inputMatToConvert : the OpenCV cv::Mat that has to be converted to gray or RGB valarray buffer that will be processed by the retina model
+     * @param outputValarrayMatrix : the output valarray
+     * @return the input image color mode (color=true, gray levels=false)
+     */
+    bool _convertCvMat2ValarrayBuffer(InputArray inputMatToConvert, std::valarray<float> &outputValarrayMatrix);
+
+
+};
+
+// smart pointers allocation :
+Ptr<Retina> createRetina(Size inputSize){ return makePtr<RetinaImpl>(inputSize); }
+Ptr<Retina> createRetina(Size inputSize, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght){
+    return makePtr<RetinaImpl>(inputSize, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+}
+
+
+// RetinaImpl code
+RetinaImpl::RetinaImpl(const cv::Size inputSz)
+{
+    _retinaFilter = 0;
+    _init(inputSz, true, RETINA_COLOR_BAYER, false);
+}
+
+RetinaImpl::RetinaImpl(const cv::Size inputSz, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+{
+    _retinaFilter = 0;
+    _init(inputSz, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+};
+
+RetinaImpl::~RetinaImpl()
+{
+    if (_retinaFilter)
+        delete _retinaFilter;
+}
+
+/**
+* retreive retina input buffer size
+*/
+Size RetinaImpl::getInputSize(){return cv::Size(_retinaFilter->getInputNBcolumns(), _retinaFilter->getInputNBrows());}
+
+/**
+* retreive retina output buffer size
+*/
+Size RetinaImpl::getOutputSize(){return cv::Size(_retinaFilter->getOutputNBcolumns(), _retinaFilter->getOutputNBrows());}
+
+
+void RetinaImpl::setColorSaturation(const bool saturateColors, const float colorSaturationValue)
+{
+    _retinaFilter->setColorSaturation(saturateColors, colorSaturationValue);
+}
+
+struct Retina::RetinaParameters RetinaImpl::getParameters(){return _retinaParameters;}
+
+void RetinaImpl::setup(String retinaParameterFile, const bool applyDefaultSetupOnFailure)
+{
+    try
+    {
+        // opening retinaParameterFile in read mode
+        cv::FileStorage fs(retinaParameterFile, cv::FileStorage::READ);
+        setup(fs, applyDefaultSetupOnFailure);
+    }
+    catch(Exception &e)
+    {
+        printf("Retina::setup: wrong/unappropriate xml parameter file : error report :`n=>%s\n", e.what());
+        if (applyDefaultSetupOnFailure)
+        {
+            printf("Retina::setup: resetting retina with default parameters\n");
+            setupOPLandIPLParvoChannel();
+            setupIPLMagnoChannel();
+        }
+        else
+        {
+            printf("=> keeping current parameters\n");
+        }
+    }
+}
+
+void RetinaImpl::setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure)
+{
+    try
+    {
+        // read parameters file if it exists or apply default setup if asked for
+        if (!fs.isOpened())
+        {
+            printf("Retina::setup: provided parameters file could not be open... skeeping configuration\n");
+            return;
+            // implicit else case : retinaParameterFile could be open (it exists at least)
+        }
+                // OPL and Parvo init first... update at the same time the parameters structure and the retina core
+        cv::FileNode rootFn = fs.root(), currFn=rootFn["OPLandIPLparvo"];
+        currFn["colorMode"]>>_retinaParameters.OPLandIplParvo.colorMode;
+        currFn["normaliseOutput"]>>_retinaParameters.OPLandIplParvo.normaliseOutput;
+        currFn["photoreceptorsLocalAdaptationSensitivity"]>>_retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
+        currFn["photoreceptorsTemporalConstant"]>>_retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
+        currFn["photoreceptorsSpatialConstant"]>>_retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
+        currFn["horizontalCellsGain"]>>_retinaParameters.OPLandIplParvo.horizontalCellsGain;
+        currFn["hcellsTemporalConstant"]>>_retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
+        currFn["hcellsSpatialConstant"]>>_retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
+        currFn["ganglionCellsSensitivity"]>>_retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
+        setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+
+        // init retina IPL magno setup... update at the same time the parameters structure and the retina core
+        currFn=rootFn["IPLmagno"];
+        currFn["normaliseOutput"]>>_retinaParameters.IplMagno.normaliseOutput;
+        currFn["parasolCells_beta"]>>_retinaParameters.IplMagno.parasolCells_beta;
+        currFn["parasolCells_tau"]>>_retinaParameters.IplMagno.parasolCells_tau;
+        currFn["parasolCells_k"]>>_retinaParameters.IplMagno.parasolCells_k;
+        currFn["amacrinCellsTemporalCutFrequency"]>>_retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
+        currFn["V0CompressionParameter"]>>_retinaParameters.IplMagno.V0CompressionParameter;
+        currFn["localAdaptintegration_tau"]>>_retinaParameters.IplMagno.localAdaptintegration_tau;
+        currFn["localAdaptintegration_k"]>>_retinaParameters.IplMagno.localAdaptintegration_k;
+
+        setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency,_retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
+
+    }catch(Exception &e)
+    {
+        printf("RetinaImpl::setup: resetting retina with default parameters\n");
+        if (applyDefaultSetupOnFailure)
+        {
+            setupOPLandIPLParvoChannel();
+            setupIPLMagnoChannel();
+        }
+        printf("Retina::setup: wrong/unappropriate xml parameter file : error report :`n=>%s\n", e.what());
+        printf("=> keeping current parameters\n");
+    }
+
+    // report current configuration
+    printf("%s\n", printSetup().c_str());
+}
+
+void RetinaImpl::setup(Retina::RetinaParameters newConfiguration)
+{
+    // simply copy structures
+    memcpy(&_retinaParameters, &newConfiguration, sizeof(Retina::RetinaParameters));
+    // apply setup
+    setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+    setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency,_retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
+
+}
+
+const String RetinaImpl::printSetup()
+{
+    std::stringstream outmessage;
+
+    // displaying OPL and IPL parvo setup
+    outmessage<<"Current Retina instance setup :"
+            <<"\nOPLandIPLparvo"<<"{"
+            << "\n\t colorMode : " << _retinaParameters.OPLandIplParvo.colorMode
+            << "\n\t normalizeParvoOutput :" << _retinaParameters.OPLandIplParvo.normaliseOutput
+            << "\n\t photoreceptorsLocalAdaptationSensitivity : " << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity
+            << "\n\t photoreceptorsTemporalConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant
+            << "\n\t photoreceptorsSpatialConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant
+            << "\n\t horizontalCellsGain : " << _retinaParameters.OPLandIplParvo.horizontalCellsGain
+            << "\n\t hcellsTemporalConstant : " << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant
+            << "\n\t hcellsSpatialConstant : " << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant
+            << "\n\t parvoGanglionCellsSensitivity : " << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity
+            <<"}\n";
+
+    // displaying IPL magno setup
+    outmessage<<"Current Retina instance setup :"
+            <<"\nIPLmagno"<<"{"
+            << "\n\t normaliseOutput : " << _retinaParameters.IplMagno.normaliseOutput
+            << "\n\t parasolCells_beta : " << _retinaParameters.IplMagno.parasolCells_beta
+            << "\n\t parasolCells_tau : " << _retinaParameters.IplMagno.parasolCells_tau
+            << "\n\t parasolCells_k : " << _retinaParameters.IplMagno.parasolCells_k
+            << "\n\t amacrinCellsTemporalCutFrequency : " << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency
+            << "\n\t V0CompressionParameter : " << _retinaParameters.IplMagno.V0CompressionParameter
+            << "\n\t localAdaptintegration_tau : " << _retinaParameters.IplMagno.localAdaptintegration_tau
+            << "\n\t localAdaptintegration_k : " << _retinaParameters.IplMagno.localAdaptintegration_k
+            <<"}";
+    return outmessage.str().c_str();
+}
+
+void RetinaImpl::write( String fs ) const
+{
+    FileStorage parametersSaveFile(fs, cv::FileStorage::WRITE );
+    write(parametersSaveFile);
+}
+
+void RetinaImpl::write( FileStorage& fs ) const
+{
+    if (!fs.isOpened())
+        return; // basic error case
+    fs<<"OPLandIPLparvo"<<"{";
+    fs << "colorMode" << _retinaParameters.OPLandIplParvo.colorMode;
+    fs << "normaliseOutput" << _retinaParameters.OPLandIplParvo.normaliseOutput;
+    fs << "photoreceptorsLocalAdaptationSensitivity" << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
+    fs << "photoreceptorsTemporalConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
+    fs << "photoreceptorsSpatialConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
+    fs << "horizontalCellsGain" << _retinaParameters.OPLandIplParvo.horizontalCellsGain;
+    fs << "hcellsTemporalConstant" << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
+    fs << "hcellsSpatialConstant" << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
+    fs << "ganglionCellsSensitivity" << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
+    fs << "}";
+    fs<<"IPLmagno"<<"{";
+    fs << "normaliseOutput" << _retinaParameters.IplMagno.normaliseOutput;
+    fs << "parasolCells_beta" << _retinaParameters.IplMagno.parasolCells_beta;
+    fs << "parasolCells_tau" << _retinaParameters.IplMagno.parasolCells_tau;
+    fs << "parasolCells_k" << _retinaParameters.IplMagno.parasolCells_k;
+    fs << "amacrinCellsTemporalCutFrequency" << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
+    fs << "V0CompressionParameter" << _retinaParameters.IplMagno.V0CompressionParameter;
+    fs << "localAdaptintegration_tau" << _retinaParameters.IplMagno.localAdaptintegration_tau;
+    fs << "localAdaptintegration_k" << _retinaParameters.IplMagno.localAdaptintegration_k;
+    fs<<"}";
+}
+
+void RetinaImpl::setupOPLandIPLParvoChannel(const bool colorMode, const bool normaliseOutput, const float photoreceptorsLocalAdaptationSensitivity, const float photoreceptorsTemporalConstant, const float photoreceptorsSpatialConstant, const float horizontalCellsGain, const float HcellsTemporalConstant, const float HcellsSpatialConstant, const float ganglionCellsSensitivity)
+{
+    // retina core parameters setup
+    _retinaFilter->setColorMode(colorMode);
+    _retinaFilter->setPhotoreceptorsLocalAdaptationSensitivity(photoreceptorsLocalAdaptationSensitivity);
+    _retinaFilter->setOPLandParvoParameters(0, photoreceptorsTemporalConstant, photoreceptorsSpatialConstant, horizontalCellsGain, HcellsTemporalConstant, HcellsSpatialConstant, ganglionCellsSensitivity);
+    _retinaFilter->setParvoGanglionCellsLocalAdaptationSensitivity(ganglionCellsSensitivity);
+    _retinaFilter->activateNormalizeParvoOutput_0_maxOutputValue(normaliseOutput);
+
+        // update parameters struture
+
+    _retinaParameters.OPLandIplParvo.colorMode = colorMode;
+    _retinaParameters.OPLandIplParvo.normaliseOutput = normaliseOutput;
+    _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity = photoreceptorsLocalAdaptationSensitivity;
+    _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant = photoreceptorsTemporalConstant;
+    _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant = photoreceptorsSpatialConstant;
+    _retinaParameters.OPLandIplParvo.horizontalCellsGain = horizontalCellsGain;
+    _retinaParameters.OPLandIplParvo.hcellsTemporalConstant = HcellsTemporalConstant;
+    _retinaParameters.OPLandIplParvo.hcellsSpatialConstant = HcellsSpatialConstant;
+    _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity = ganglionCellsSensitivity;
+
+}
+
+void RetinaImpl::setupIPLMagnoChannel(const bool normaliseOutput, const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float V0CompressionParameter, const float localAdaptintegration_tau, const float localAdaptintegration_k)
+{
+
+    _retinaFilter->setMagnoCoefficientsTable(parasolCells_beta, parasolCells_tau, parasolCells_k, amacrinCellsTemporalCutFrequency, V0CompressionParameter, localAdaptintegration_tau, localAdaptintegration_k);
+    _retinaFilter->activateNormalizeMagnoOutput_0_maxOutputValue(normaliseOutput);
+
+        // update parameters struture
+    _retinaParameters.IplMagno.normaliseOutput = normaliseOutput;
+    _retinaParameters.IplMagno.parasolCells_beta = parasolCells_beta;
+    _retinaParameters.IplMagno.parasolCells_tau = parasolCells_tau;
+    _retinaParameters.IplMagno.parasolCells_k = parasolCells_k;
+    _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency = amacrinCellsTemporalCutFrequency;
+    _retinaParameters.IplMagno.V0CompressionParameter = V0CompressionParameter;
+    _retinaParameters.IplMagno.localAdaptintegration_tau = localAdaptintegration_tau;
+    _retinaParameters.IplMagno.localAdaptintegration_k = localAdaptintegration_k;
+}
+
+void RetinaImpl::run(InputArray inputMatToConvert)
+{
+    // first convert input image to the compatible format : std::valarray<float>
+    const bool colorMode = _convertCvMat2ValarrayBuffer(inputMatToConvert.getMat(), _inputBuffer);
+    // process the retina
+    if (!_retinaFilter->runFilter(_inputBuffer, colorMode, false, _retinaParameters.OPLandIplParvo.colorMode && colorMode, false))
+        throw cv::Exception(-1, "RetinaImpl cannot be applied, wrong input buffer size", "RetinaImpl::run", "RetinaImpl.h", 0);
+}
+
+void RetinaImpl::applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)
+{
+    // first convert input image to the compatible format :
+    const bool colorMode = _convertCvMat2ValarrayBuffer(inputImage.getMat(), _inputBuffer);
+    const unsigned int nbPixels=_retinaFilter->getOutputNBrows()*_retinaFilter->getOutputNBcolumns();
+
+    // process tone mapping
+    if (colorMode)
+    {
+        std::valarray<float> imageOutput(nbPixels*3);
+        _retinaFilter->runRGBToneMapping(_inputBuffer, imageOutput, true, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+        _convertValarrayBuffer2cvMat(imageOutput, _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), true, outputToneMappedImage);
+    }else
+    {
+        std::valarray<float> imageOutput(nbPixels);
+        _retinaFilter->runGrayToneMapping(_inputBuffer, imageOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+        _convertValarrayBuffer2cvMat(imageOutput, _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), false, outputToneMappedImage);
+    }
+
+}
+
+void RetinaImpl::getParvo(OutputArray retinaOutput_parvo)
+{
+    if (_retinaFilter->getColorMode())
+    {
+        // reallocate output buffer (if necessary)
+        _convertValarrayBuffer2cvMat(_retinaFilter->getColorOutput(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), true, retinaOutput_parvo);
+    }else
+    {
+        // reallocate output buffer (if necessary)
+        _convertValarrayBuffer2cvMat(_retinaFilter->getContours(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), false, retinaOutput_parvo);
+    }
+    //retinaOutput_parvo/=255.0;
+}
+void RetinaImpl::getMagno(OutputArray retinaOutput_magno)
+{
+    // reallocate output buffer (if necessary)
+    _convertValarrayBuffer2cvMat(_retinaFilter->getMovingContours(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), false, retinaOutput_magno);
+    //retinaOutput_magno/=255.0;
+}
+
+// original API level data accessors : copy buffers if size matches, reallocate if required
+void RetinaImpl::getMagnoRAW(OutputArray magnoOutputBufferCopy){
+    // get magno channel header
+    const cv::Mat magnoChannel=cv::Mat(getMagnoRAW());
+    // copy data
+    magnoChannel.copyTo(magnoOutputBufferCopy);
+}
+
+void RetinaImpl::getParvoRAW(OutputArray parvoOutputBufferCopy){
+    // get parvo channel header
+    const cv::Mat parvoChannel=cv::Mat(getMagnoRAW());
+    // copy data
+    parvoChannel.copyTo(parvoOutputBufferCopy);
+}
+
+// original API level data accessors : get buffers addresses...
+const Mat RetinaImpl::getMagnoRAW() const {
+    // create a cv::Mat header for the valarray
+    return Mat(_retinaFilter->getMovingContours().size(),1, CV_32F, (void*)get_data(_retinaFilter->getMovingContours()));
+
+}
+
+const Mat RetinaImpl::getParvoRAW() const {
+    if (_retinaFilter->getColorMode()) // check if color mode is enabled
+    {
+        // create a cv::Mat table (for RGB planes as a single vector)
+        return Mat(_retinaFilter->getColorOutput().size(), 1, CV_32F, (void*)get_data(_retinaFilter->getColorOutput()));
+    }
+    // otherwise, output is gray level
+    // create a cv::Mat header for the valarray
+    return Mat( _retinaFilter->getContours().size(), 1, CV_32F, (void*)get_data(_retinaFilter->getContours()));
+}
+
+// private method called by constructirs
+void RetinaImpl::_init(const cv::Size inputSz, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+{
+    // basic error check
+    if (inputSz.height*inputSz.width <= 0)
+        throw cv::Exception(-1, "Bad retina size setup : size height and with must be superior to zero", "RetinaImpl::setup", "Retina.cpp", 0);
+
+    unsigned int nbPixels=inputSz.height*inputSz.width;
+    // resize buffers if size does not match
+    _inputBuffer.resize(nbPixels*3); // buffer supports gray images but also 3 channels color buffers... (larger is better...)
+
+    // allocate the retina model
+        if (_retinaFilter)
+           delete _retinaFilter;
+    _retinaFilter = new RetinaFilter(inputSz.height, inputSz.width, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+
+    _retinaParameters.OPLandIplParvo.colorMode = colorMode;
+    // prepare the default parameter XML file with default setup
+    setup(_retinaParameters);
+
+    // init retina
+    _retinaFilter->clearAllBuffers();
+
+    // report current configuration
+    printf("%s\n", printSetup().c_str());
+}
+
+void RetinaImpl::_convertValarrayBuffer2cvMat(const std::valarray<float> &grayMatrixToConvert, const unsigned int nbRows, const unsigned int nbColumns, const bool colorMode, OutputArray outBuffer)
+{
+    // fill output buffer with the valarray buffer
+    const float *valarrayPTR=get_data(grayMatrixToConvert);
+    if (!colorMode)
+    {
+        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8U);
+        Mat outMat = outBuffer.getMat();
+        for (unsigned int i=0;i<nbRows;++i)
+        {
+            for (unsigned int j=0;j<nbColumns;++j)
+            {
+                cv::Point2d pixel(j,i);
+                outMat.at<unsigned char>(pixel)=(unsigned char)*(valarrayPTR++);
+            }
+        }
+    }else
+    {
+        const unsigned int nbPixels=nbColumns*nbRows;
+        const unsigned int doubleNBpixels=nbColumns*nbRows*2;
+        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8UC3);
+        Mat outMat = outBuffer.getMat();
+        for (unsigned int i=0;i<nbRows;++i)
+        {
+            for (unsigned int j=0;j<nbColumns;++j,++valarrayPTR)
+            {
+                cv::Point2d pixel(j,i);
+                cv::Vec3b pixelValues;
+                pixelValues[2]=(unsigned char)*(valarrayPTR);
+                pixelValues[1]=(unsigned char)*(valarrayPTR+nbPixels);
+                pixelValues[0]=(unsigned char)*(valarrayPTR+doubleNBpixels);
+
+                outMat.at<cv::Vec3b>(pixel)=pixelValues;
+            }
+        }
+    }
+}
+
+bool RetinaImpl::_convertCvMat2ValarrayBuffer(InputArray inputMat, std::valarray<float> &outputValarrayMatrix)
+{
+    const Mat inputMatToConvert=inputMat.getMat();
+    // first check input consistency
+    if (inputMatToConvert.empty())
+        throw cv::Exception(-1, "RetinaImpl cannot be applied, input buffer is empty", "RetinaImpl::run", "RetinaImpl.h", 0);
+
+    // retreive color mode from image input
+    int imageNumberOfChannels = inputMatToConvert.channels();
+
+        // convert to float AND fill the valarray buffer
+    typedef float T; // define here the target pixel format, here, float
+    const int dsttype = DataType<T>::depth; // output buffer is float format
+
+    const unsigned int nbPixels=inputMat.getMat().rows*inputMat.getMat().cols;
+    const unsigned int doubleNBpixels=inputMat.getMat().rows*inputMat.getMat().cols*2;
+
+    if(imageNumberOfChannels==4)
+    {
+    // create a cv::Mat table (for RGBA planes)
+        cv::Mat planes[4] =
+        {
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[doubleNBpixels]),
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[nbPixels]),
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
+        };
+        planes[3] = cv::Mat(inputMatToConvert.size(), dsttype);     // last channel (alpha) does not point on the valarray (not usefull in our case)
+        // split color cv::Mat in 4 planes... it fills valarray directely
+        cv::split(Mat_<Vec<T, 4> >(inputMatToConvert), planes);
+    }
+    else if (imageNumberOfChannels==3)
+    {
+        // create a cv::Mat table (for RGB planes)
+        cv::Mat planes[] =
+        {
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[doubleNBpixels]),
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[nbPixels]),
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
+        };
+        // split color cv::Mat in 3 planes... it fills valarray directely
+        cv::split(cv::Mat_<Vec<T, 3> >(inputMatToConvert), planes);
+    }
+    else if(imageNumberOfChannels==1)
+    {
+        // create a cv::Mat header for the valarray
+        cv::Mat dst(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0]);
+        inputMatToConvert.convertTo(dst, dsttype);
+    }
+        else
+            CV_Error(Error::StsUnsupportedFormat, "input image must be single channel (gray levels), bgr format (color) or bgra (color with transparency which won't be considered");
+
+    return imageNumberOfChannels>1; // return bool : false for gray level image processing, true for color mode
+}
+
+void RetinaImpl::clearBuffers() {_retinaFilter->clearAllBuffers();}
+
+void RetinaImpl::activateMovingContoursProcessing(const bool activate){_retinaFilter->activateMovingContoursProcessing(activate);}
+
+void RetinaImpl::activateContoursProcessing(const bool activate){_retinaFilter->activateContoursProcessing(activate);}
+
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/bioinspired/src/retina_ocl.cpp b/modules/bioinspired/src/retina_ocl.cpp
new file mode 100644
index 0000000000..c294ebfebe
--- /dev/null
+++ b/modules/bioinspired/src/retina_ocl.cpp
@@ -0,0 +1,1648 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include "retina_ocl.hpp"
+#include <iostream>
+#include <sstream>
+
+#ifdef HAVE_OPENCV_OCL
+
+#define NOT_IMPLEMENTED CV_Error(cv::Error::StsNotImplemented, "Not implemented")
+
+namespace cv
+{
+namespace ocl
+{
+//OpenCL kernel file string pointer
+extern const char * retina_kernel;
+}
+}
+
+namespace cv
+{
+namespace bioinspired
+{
+namespace ocl
+{
+using namespace cv::ocl;
+
+class RetinaOCLImpl : public Retina
+{
+public:
+    RetinaOCLImpl(Size getInputSize);
+    RetinaOCLImpl(Size getInputSize, const bool colorMode, int colorSamplingMethod = RETINA_COLOR_BAYER, const bool useRetinaLogSampling = false, const double reductionFactor = 1.0, const double samplingStrenght = 10.0);
+    virtual ~RetinaOCLImpl();
+
+    Size getInputSize();
+    Size getOutputSize();
+
+    void setup(String retinaParameterFile = "", const bool applyDefaultSetupOnFailure = true);
+    void setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure = true);
+    void setup(RetinaParameters newParameters);
+
+    RetinaOCLImpl::RetinaParameters getParameters();
+
+    const String printSetup();
+    virtual void write( String fs ) const;
+    virtual void write( FileStorage& fs ) const;
+
+    void setupOPLandIPLParvoChannel(const bool colorMode = true, const bool normaliseOutput = true, const float photoreceptorsLocalAdaptationSensitivity = 0.7, const float photoreceptorsTemporalConstant = 0.5, const float photoreceptorsSpatialConstant = 0.53, const float horizontalCellsGain = 0, const float HcellsTemporalConstant = 1, const float HcellsSpatialConstant = 7, const float ganglionCellsSensitivity = 0.7);
+    void setupIPLMagnoChannel(const bool normaliseOutput = true, const float parasolCells_beta = 0, const float parasolCells_tau = 0, const float parasolCells_k = 7, const float amacrinCellsTemporalCutFrequency = 1.2, const float V0CompressionParameter = 0.95, const float localAdaptintegration_tau = 0, const float localAdaptintegration_k = 7);
+
+    void run(InputArray inputImage);
+    void getParvo(OutputArray retinaOutput_parvo);
+    void getMagno(OutputArray retinaOutput_magno);
+
+    void setColorSaturation(const bool saturateColors = true, const float colorSaturationValue = 4.0);
+    void clearBuffers();
+    void activateMovingContoursProcessing(const bool activate);
+    void activateContoursProcessing(const bool activate);
+
+    // unimplemented interfaces:
+    void applyFastToneMapping(InputArray /*inputImage*/, OutputArray /*outputToneMappedImage*/) { NOT_IMPLEMENTED; }
+    void getParvoRAW(OutputArray /*retinaOutput_parvo*/) { NOT_IMPLEMENTED; }
+    void getMagnoRAW(OutputArray /*retinaOutput_magno*/) { NOT_IMPLEMENTED; }
+    const Mat getMagnoRAW() const { NOT_IMPLEMENTED; return Mat(); }
+    const Mat getParvoRAW() const { NOT_IMPLEMENTED; return Mat(); }
+
+protected:
+    RetinaParameters _retinaParameters;
+    cv::ocl::oclMat _inputBuffer;
+    RetinaFilter* _retinaFilter;
+    bool convertToColorPlanes(const cv::ocl::oclMat& input, cv::ocl::oclMat &output);
+    void convertToInterleaved(const cv::ocl::oclMat& input, bool colorMode, cv::ocl::oclMat &output);
+    void _init(const Size getInputSize, const bool colorMode, int colorSamplingMethod = RETINA_COLOR_BAYER, const bool useRetinaLogSampling = false, const double reductionFactor = 1.0, const double samplingStrenght = 10.0);
+};
+
+RetinaOCLImpl::RetinaOCLImpl(const cv::Size inputSz)
+{
+    _retinaFilter = 0;
+    _init(inputSz, true, RETINA_COLOR_BAYER, false);
+}
+
+RetinaOCLImpl::RetinaOCLImpl(const cv::Size inputSz, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+{
+    _retinaFilter = 0;
+    _init(inputSz, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+};
+
+RetinaOCLImpl::~RetinaOCLImpl()
+{
+    if (_retinaFilter)
+    {
+        delete _retinaFilter;
+    }
+}
+
+/**
+* retreive retina input buffer size
+*/
+Size RetinaOCLImpl::getInputSize()
+{
+    return cv::Size(_retinaFilter->getInputNBcolumns(), _retinaFilter->getInputNBrows());
+}
+
+/**
+* retreive retina output buffer size
+*/
+Size RetinaOCLImpl::getOutputSize()
+{
+    return cv::Size(_retinaFilter->getOutputNBcolumns(), _retinaFilter->getOutputNBrows());
+}
+
+
+void RetinaOCLImpl::setColorSaturation(const bool saturateColors, const float colorSaturationValue)
+{
+    _retinaFilter->setColorSaturation(saturateColors, colorSaturationValue);
+}
+
+struct RetinaOCLImpl::RetinaParameters RetinaOCLImpl::getParameters()
+{
+    return _retinaParameters;
+}
+
+
+void RetinaOCLImpl::setup(String retinaParameterFile, const bool applyDefaultSetupOnFailure)
+{
+    try
+    {
+        // opening retinaParameterFile in read mode
+        cv::FileStorage fs(retinaParameterFile, cv::FileStorage::READ);
+        setup(fs, applyDefaultSetupOnFailure);
+    }
+    catch(Exception &e)
+    {
+        std::cout << "RetinaOCLImpl::setup: wrong/unappropriate xml parameter file : error report :`n=>" << e.what() << std::endl;
+        if (applyDefaultSetupOnFailure)
+        {
+            std::cout << "RetinaOCLImpl::setup: resetting retina with default parameters" << std::endl;
+            setupOPLandIPLParvoChannel();
+            setupIPLMagnoChannel();
+        }
+        else
+        {
+            std::cout << "=> keeping current parameters" << std::endl;
+        }
+    }
+}
+
+void RetinaOCLImpl::setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure)
+{
+    try
+    {
+        // read parameters file if it exists or apply default setup if asked for
+        if (!fs.isOpened())
+        {
+            std::cout << "RetinaOCLImpl::setup: provided parameters file could not be open... skeeping configuration" << std::endl;
+            return;
+            // implicit else case : retinaParameterFile could be open (it exists at least)
+        }
+        // OPL and Parvo init first... update at the same time the parameters structure and the retina core
+        cv::FileNode rootFn = fs.root(), currFn = rootFn["OPLandIPLparvo"];
+        currFn["colorMode"] >> _retinaParameters.OPLandIplParvo.colorMode;
+        currFn["normaliseOutput"] >> _retinaParameters.OPLandIplParvo.normaliseOutput;
+        currFn["photoreceptorsLocalAdaptationSensitivity"] >> _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
+        currFn["photoreceptorsTemporalConstant"] >> _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
+        currFn["photoreceptorsSpatialConstant"] >> _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
+        currFn["horizontalCellsGain"] >> _retinaParameters.OPLandIplParvo.horizontalCellsGain;
+        currFn["hcellsTemporalConstant"] >> _retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
+        currFn["hcellsSpatialConstant"] >> _retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
+        currFn["ganglionCellsSensitivity"] >> _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
+        setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+
+        // init retina IPL magno setup... update at the same time the parameters structure and the retina core
+        currFn = rootFn["IPLmagno"];
+        currFn["normaliseOutput"] >> _retinaParameters.IplMagno.normaliseOutput;
+        currFn["parasolCells_beta"] >> _retinaParameters.IplMagno.parasolCells_beta;
+        currFn["parasolCells_tau"] >> _retinaParameters.IplMagno.parasolCells_tau;
+        currFn["parasolCells_k"] >> _retinaParameters.IplMagno.parasolCells_k;
+        currFn["amacrinCellsTemporalCutFrequency"] >> _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
+        currFn["V0CompressionParameter"] >> _retinaParameters.IplMagno.V0CompressionParameter;
+        currFn["localAdaptintegration_tau"] >> _retinaParameters.IplMagno.localAdaptintegration_tau;
+        currFn["localAdaptintegration_k"] >> _retinaParameters.IplMagno.localAdaptintegration_k;
+
+        setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency, _retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
+
+    }
+    catch(Exception &e)
+    {
+        std::cout << "RetinaOCLImpl::setup: resetting retina with default parameters" << std::endl;
+        if (applyDefaultSetupOnFailure)
+        {
+            setupOPLandIPLParvoChannel();
+            setupIPLMagnoChannel();
+        }
+        std::cout << "RetinaOCLImpl::setup: wrong/unappropriate xml parameter file : error report :`n=>" << e.what() << std::endl;
+        std::cout << "=> keeping current parameters" << std::endl;
+    }
+}
+
+void RetinaOCLImpl::setup(cv::bioinspired::Retina::RetinaParameters newConfiguration)
+{
+    // simply copy structures
+    memcpy(&_retinaParameters, &newConfiguration, sizeof(cv::bioinspired::Retina::RetinaParameters));
+    // apply setup
+    setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
+    setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency, _retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
+}
+
+const String RetinaOCLImpl::printSetup()
+{
+    std::stringstream outmessage;
+
+    // displaying OPL and IPL parvo setup
+    outmessage << "Current Retina instance setup :"
+               << "\nOPLandIPLparvo" << "{"
+               << "\n==> colorMode : " << _retinaParameters.OPLandIplParvo.colorMode
+               << "\n==> normalizeParvoOutput :" << _retinaParameters.OPLandIplParvo.normaliseOutput
+               << "\n==> photoreceptorsLocalAdaptationSensitivity : " << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity
+               << "\n==> photoreceptorsTemporalConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant
+               << "\n==> photoreceptorsSpatialConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant
+               << "\n==> horizontalCellsGain : " << _retinaParameters.OPLandIplParvo.horizontalCellsGain
+               << "\n==> hcellsTemporalConstant : " << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant
+               << "\n==> hcellsSpatialConstant : " << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant
+               << "\n==> parvoGanglionCellsSensitivity : " << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity
+               << "}\n";
+
+    // displaying IPL magno setup
+    outmessage << "Current Retina instance setup :"
+               << "\nIPLmagno" << "{"
+               << "\n==> normaliseOutput : " << _retinaParameters.IplMagno.normaliseOutput
+               << "\n==> parasolCells_beta : " << _retinaParameters.IplMagno.parasolCells_beta
+               << "\n==> parasolCells_tau : " << _retinaParameters.IplMagno.parasolCells_tau
+               << "\n==> parasolCells_k : " << _retinaParameters.IplMagno.parasolCells_k
+               << "\n==> amacrinCellsTemporalCutFrequency : " << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency
+               << "\n==> V0CompressionParameter : " << _retinaParameters.IplMagno.V0CompressionParameter
+               << "\n==> localAdaptintegration_tau : " << _retinaParameters.IplMagno.localAdaptintegration_tau
+               << "\n==> localAdaptintegration_k : " << _retinaParameters.IplMagno.localAdaptintegration_k
+               << "}";
+    return outmessage.str().c_str();
+}
+
+void RetinaOCLImpl::write( String fs ) const
+{
+    FileStorage parametersSaveFile(fs, cv::FileStorage::WRITE );
+    write(parametersSaveFile);
+}
+
+void RetinaOCLImpl::write( FileStorage& fs ) const
+{
+    if (!fs.isOpened())
+    {
+        return;    // basic error case
+    }
+    fs << "OPLandIPLparvo" << "{";
+    fs << "colorMode" << _retinaParameters.OPLandIplParvo.colorMode;
+    fs << "normaliseOutput" << _retinaParameters.OPLandIplParvo.normaliseOutput;
+    fs << "photoreceptorsLocalAdaptationSensitivity" << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
+    fs << "photoreceptorsTemporalConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
+    fs << "photoreceptorsSpatialConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
+    fs << "horizontalCellsGain" << _retinaParameters.OPLandIplParvo.horizontalCellsGain;
+    fs << "hcellsTemporalConstant" << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
+    fs << "hcellsSpatialConstant" << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
+    fs << "ganglionCellsSensitivity" << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
+    fs << "}";
+    fs << "IPLmagno" << "{";
+    fs << "normaliseOutput" << _retinaParameters.IplMagno.normaliseOutput;
+    fs << "parasolCells_beta" << _retinaParameters.IplMagno.parasolCells_beta;
+    fs << "parasolCells_tau" << _retinaParameters.IplMagno.parasolCells_tau;
+    fs << "parasolCells_k" << _retinaParameters.IplMagno.parasolCells_k;
+    fs << "amacrinCellsTemporalCutFrequency" << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
+    fs << "V0CompressionParameter" << _retinaParameters.IplMagno.V0CompressionParameter;
+    fs << "localAdaptintegration_tau" << _retinaParameters.IplMagno.localAdaptintegration_tau;
+    fs << "localAdaptintegration_k" << _retinaParameters.IplMagno.localAdaptintegration_k;
+    fs << "}";
+}
+
+void RetinaOCLImpl::setupOPLandIPLParvoChannel(const bool colorMode, const bool normaliseOutput, const float photoreceptorsLocalAdaptationSensitivity, const float photoreceptorsTemporalConstant, const float photoreceptorsSpatialConstant, const float horizontalCellsGain, const float HcellsTemporalConstant, const float HcellsSpatialConstant, const float ganglionCellsSensitivity)
+{
+    // retina core parameters setup
+    _retinaFilter->setColorMode(colorMode);
+    _retinaFilter->setPhotoreceptorsLocalAdaptationSensitivity(photoreceptorsLocalAdaptationSensitivity);
+    _retinaFilter->setOPLandParvoParameters(0, photoreceptorsTemporalConstant, photoreceptorsSpatialConstant, horizontalCellsGain, HcellsTemporalConstant, HcellsSpatialConstant, ganglionCellsSensitivity);
+    _retinaFilter->setParvoGanglionCellsLocalAdaptationSensitivity(ganglionCellsSensitivity);
+    _retinaFilter->activateNormalizeParvoOutput_0_maxOutputValue(normaliseOutput);
+
+    // update parameters struture
+
+    _retinaParameters.OPLandIplParvo.colorMode = colorMode;
+    _retinaParameters.OPLandIplParvo.normaliseOutput = normaliseOutput;
+    _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity = photoreceptorsLocalAdaptationSensitivity;
+    _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant = photoreceptorsTemporalConstant;
+    _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant = photoreceptorsSpatialConstant;
+    _retinaParameters.OPLandIplParvo.horizontalCellsGain = horizontalCellsGain;
+    _retinaParameters.OPLandIplParvo.hcellsTemporalConstant = HcellsTemporalConstant;
+    _retinaParameters.OPLandIplParvo.hcellsSpatialConstant = HcellsSpatialConstant;
+    _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity = ganglionCellsSensitivity;
+}
+
+void RetinaOCLImpl::setupIPLMagnoChannel(const bool normaliseOutput, const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float V0CompressionParameter, const float localAdaptintegration_tau, const float localAdaptintegration_k)
+{
+
+    _retinaFilter->setMagnoCoefficientsTable(parasolCells_beta, parasolCells_tau, parasolCells_k, amacrinCellsTemporalCutFrequency, V0CompressionParameter, localAdaptintegration_tau, localAdaptintegration_k);
+    _retinaFilter->activateNormalizeMagnoOutput_0_maxOutputValue(normaliseOutput);
+
+    // update parameters struture
+    _retinaParameters.IplMagno.normaliseOutput = normaliseOutput;
+    _retinaParameters.IplMagno.parasolCells_beta = parasolCells_beta;
+    _retinaParameters.IplMagno.parasolCells_tau = parasolCells_tau;
+    _retinaParameters.IplMagno.parasolCells_k = parasolCells_k;
+    _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency = amacrinCellsTemporalCutFrequency;
+    _retinaParameters.IplMagno.V0CompressionParameter = V0CompressionParameter;
+    _retinaParameters.IplMagno.localAdaptintegration_tau = localAdaptintegration_tau;
+    _retinaParameters.IplMagno.localAdaptintegration_k = localAdaptintegration_k;
+}
+
+void RetinaOCLImpl::run(const InputArray input)
+{
+    oclMat &inputMatToConvert = getOclMatRef(input);
+    bool colorMode = convertToColorPlanes(inputMatToConvert, _inputBuffer);
+    // first convert input image to the compatible format : std::valarray<float>
+    // process the retina
+    if (!_retinaFilter->runFilter(_inputBuffer, colorMode, false, _retinaParameters.OPLandIplParvo.colorMode && colorMode, false))
+    {
+        throw cv::Exception(-1, "Retina cannot be applied, wrong input buffer size", "RetinaOCLImpl::run", "Retina.h", 0);
+    }
+}
+
+void RetinaOCLImpl::getParvo(OutputArray output)
+{
+    oclMat &retinaOutput_parvo = getOclMatRef(output);
+    if (_retinaFilter->getColorMode())
+    {
+        // reallocate output buffer (if necessary)
+        convertToInterleaved(_retinaFilter->getColorOutput(), true, retinaOutput_parvo);
+    }
+    else
+    {
+        // reallocate output buffer (if necessary)
+        convertToInterleaved(_retinaFilter->getContours(), false, retinaOutput_parvo);
+    }
+    //retinaOutput_parvo/=255.0;
+}
+void RetinaOCLImpl::getMagno(OutputArray output)
+{
+    oclMat &retinaOutput_magno = getOclMatRef(output);
+    // reallocate output buffer (if necessary)
+    convertToInterleaved(_retinaFilter->getMovingContours(), false, retinaOutput_magno);
+    //retinaOutput_magno/=255.0;
+}
+// private method called by constructirs
+void RetinaOCLImpl::_init(const cv::Size inputSz, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+{
+    // basic error check
+    if (inputSz.height*inputSz.width <= 0)
+    {
+        throw cv::Exception(-1, "Bad retina size setup : size height and with must be superior to zero", "RetinaOCLImpl::setup", "Retina.h", 0);
+    }
+
+    // allocate the retina model
+    if (_retinaFilter)
+    {
+        delete _retinaFilter;
+    }
+    _retinaFilter = new RetinaFilter(inputSz.height, inputSz.width, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+
+    // prepare the default parameter XML file with default setup
+    setup(_retinaParameters);
+
+    // init retina
+    _retinaFilter->clearAllBuffers();
+}
+
+bool RetinaOCLImpl::convertToColorPlanes(const oclMat& input, oclMat &output)
+{
+    oclMat convert_input;
+    input.convertTo(convert_input, CV_32F);
+    if(convert_input.channels() == 3 || convert_input.channels() == 4)
+    {
+        ocl::ensureSizeIsEnough(int(_retinaFilter->getInputNBrows() * 4),
+                                int(_retinaFilter->getInputNBcolumns()), CV_32FC1, output);
+        oclMat channel_splits[4] =
+        {
+            output(Rect(Point(0, _retinaFilter->getInputNBrows() * 2), getInputSize())),
+            output(Rect(Point(0, _retinaFilter->getInputNBrows()), getInputSize())),
+            output(Rect(Point(0, 0), getInputSize())),
+            output(Rect(Point(0, _retinaFilter->getInputNBrows() * 3), getInputSize()))
+        };
+        ocl::split(convert_input, channel_splits);
+        return true;
+    }
+    else if(convert_input.channels() == 1)
+    {
+        convert_input.copyTo(output);
+        return false;
+    }
+    else
+    {
+        CV_Error(-1, "Retina ocl only support 1, 3, 4 channel input");
+        return false;
+    }
+}
+void RetinaOCLImpl::convertToInterleaved(const oclMat& input, bool colorMode, oclMat &output)
+{
+    input.convertTo(output, CV_8U);
+    if(colorMode)
+    {
+        int numOfSplits = input.rows / getInputSize().height;
+        std::vector<oclMat> channel_splits(numOfSplits);
+        for(int i = 0; i < static_cast<int>(channel_splits.size()); i ++)
+        {
+            channel_splits[i] =
+                output(Rect(Point(0, _retinaFilter->getInputNBrows() * (numOfSplits - i - 1)), getInputSize()));
+        }
+        merge(channel_splits, output);
+    }
+    else
+    {
+        //...
+    }
+}
+
+void RetinaOCLImpl::clearBuffers()
+{
+    _retinaFilter->clearAllBuffers();
+}
+
+void RetinaOCLImpl::activateMovingContoursProcessing(const bool activate)
+{
+    _retinaFilter->activateMovingContoursProcessing(activate);
+}
+
+void RetinaOCLImpl::activateContoursProcessing(const bool activate)
+{
+    _retinaFilter->activateContoursProcessing(activate);
+}
+
+///////////////////////////////////////
+///////// BasicRetinaFilter ///////////
+///////////////////////////////////////
+BasicRetinaFilter::BasicRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns, const unsigned int parametersListSize, const bool)
+    : _NBrows(NBrows), _NBcols(NBcolumns),
+      _filterOutput(NBrows, NBcolumns, CV_32FC1),
+      _localBuffer(NBrows, NBcolumns, CV_32FC1),
+      _filteringCoeficientsTable(3 * parametersListSize)
+{
+    _halfNBrows = _filterOutput.rows / 2;
+    _halfNBcolumns = _filterOutput.cols / 2;
+
+    // set default values
+    _maxInputValue = 256.0;
+
+    // reset all buffers
+    clearAllBuffers();
+}
+
+BasicRetinaFilter::~BasicRetinaFilter()
+{
+}
+
+void BasicRetinaFilter::resize(const unsigned int NBrows, const unsigned int NBcolumns)
+{
+    // resizing buffers
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _filterOutput);
+
+    // updating variables
+    _halfNBrows = _filterOutput.rows / 2;
+    _halfNBcolumns = _filterOutput.cols / 2;
+
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _localBuffer);
+    // reset buffers
+    clearAllBuffers();
+}
+
+void BasicRetinaFilter::setLPfilterParameters(const float beta, const float tau, const float desired_k, const unsigned int filterIndex)
+{
+    float _beta = beta + tau;
+    float k = desired_k;
+    // check if the spatial constant is correct (avoid 0 value to avoid division by 0)
+    if (desired_k <= 0)
+    {
+        k = 0.001f;
+        std::cerr << "BasicRetinaFilter::spatial constant of the low pass filter must be superior to zero !!! correcting parameter setting to 0,001" << std::endl;
+    }
+
+    float _alpha = k * k;
+    float _mu = 0.8f;
+    unsigned int tableOffset = filterIndex * 3;
+    if (k <= 0)
+    {
+        std::cerr << "BasicRetinaFilter::spatial filtering coefficient must be superior to zero, correcting value to 0.01" << std::endl;
+        _alpha = 0.0001f;
+    }
+
+    float _temp =  (1.0f + _beta) / (2.0f * _mu * _alpha);
+    float a = _filteringCoeficientsTable[tableOffset] = 1.0f + _temp - (float)sqrt( (1.0f + _temp) * (1.0f + _temp) - 1.0f);
+    _filteringCoeficientsTable[1 + tableOffset] = (1.0f - a) * (1.0f - a) * (1.0f - a) * (1.0f - a) / (1.0f + _beta);
+    _filteringCoeficientsTable[2 + tableOffset] = tau;
+}
+const oclMat &BasicRetinaFilter::runFilter_LocalAdapdation(const oclMat &inputFrame, const oclMat &localLuminance)
+{
+    _localLuminanceAdaptation(inputFrame, localLuminance, _filterOutput);
+    return _filterOutput;
+}
+
+
+void BasicRetinaFilter::runFilter_LocalAdapdation(const oclMat &inputFrame, const oclMat &localLuminance, oclMat &outputFrame)
+{
+    _localLuminanceAdaptation(inputFrame, localLuminance, outputFrame);
+}
+
+const oclMat &BasicRetinaFilter::runFilter_LocalAdapdation_autonomous(const oclMat &inputFrame)
+{
+    _spatiotemporalLPfilter(inputFrame, _filterOutput);
+    _localLuminanceAdaptation(inputFrame, _filterOutput, _filterOutput);
+    return _filterOutput;
+}
+void BasicRetinaFilter::runFilter_LocalAdapdation_autonomous(const oclMat &inputFrame, oclMat &outputFrame)
+{
+    _spatiotemporalLPfilter(inputFrame, _filterOutput);
+    _localLuminanceAdaptation(inputFrame, _filterOutput, outputFrame);
+}
+
+void BasicRetinaFilter::_localLuminanceAdaptation(oclMat &inputOutputFrame, const oclMat &localLuminance)
+{
+    _localLuminanceAdaptation(inputOutputFrame, localLuminance, inputOutputFrame, false);
+}
+
+void BasicRetinaFilter::_localLuminanceAdaptation(const oclMat &inputFrame, const oclMat &localLuminance, oclMat &outputFrame, const bool updateLuminanceMean)
+{
+    if (updateLuminanceMean)
+    {
+        float meanLuminance = saturate_cast<float>(ocl::sum(inputFrame)[0]) / getNBpixels();
+        updateCompressionParameter(meanLuminance);
+    }
+    int elements_per_row = static_cast<int>(inputFrame.step / inputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, _NBrows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &localLuminance.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &_localLuminanceAddon));
+    args.push_back(std::make_pair(sizeof(cl_float), &_localLuminanceFactor));
+    args.push_back(std::make_pair(sizeof(cl_float), &_maxInputValue));
+    openCLExecuteKernel(ctx, &retina_kernel, "localLuminanceAdaptation", globalSize, localSize, args, -1, -1);
+}
+
+const oclMat &BasicRetinaFilter::runFilter_LPfilter(const oclMat &inputFrame, const unsigned int filterIndex)
+{
+    _spatiotemporalLPfilter(inputFrame, _filterOutput, filterIndex);
+    return _filterOutput;
+}
+void BasicRetinaFilter::runFilter_LPfilter(const oclMat &inputFrame, oclMat &outputFrame, const unsigned int filterIndex)
+{
+    _spatiotemporalLPfilter(inputFrame, outputFrame, filterIndex);
+}
+
+void BasicRetinaFilter::_spatiotemporalLPfilter(const oclMat &inputFrame, oclMat &LPfilterOutput, const unsigned int filterIndex)
+{
+    unsigned int coefTableOffset = filterIndex * 3;
+
+    _a = _filteringCoeficientsTable[coefTableOffset];
+    _gain = _filteringCoeficientsTable[1 + coefTableOffset];
+    _tau = _filteringCoeficientsTable[2 + coefTableOffset];
+
+    _horizontalCausalFilter_addInput(inputFrame, LPfilterOutput);
+    _horizontalAnticausalFilter(LPfilterOutput);
+    _verticalCausalFilter(LPfilterOutput);
+    _verticalAnticausalFilter_multGain(LPfilterOutput);
+}
+
+void BasicRetinaFilter::_horizontalCausalFilter_addInput(const oclMat &inputFrame, oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(inputFrame.step / inputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBrows, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_float), &_tau));
+    args.push_back(std::make_pair(sizeof(cl_float), &_a));
+    openCLExecuteKernel(ctx, &retina_kernel, "horizontalCausalFilter_addInput", globalSize, localSize, args, -1, -1);
+}
+
+void BasicRetinaFilter::_horizontalAnticausalFilter(oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBrows, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_float), &_a));
+    openCLExecuteKernel(ctx, &retina_kernel, "horizontalAnticausalFilter", globalSize, localSize, args, -1, -1);
+}
+
+void BasicRetinaFilter::_verticalCausalFilter(oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_float), &_a));
+    openCLExecuteKernel(ctx, &retina_kernel, "verticalCausalFilter", globalSize, localSize, args, -1, -1);
+}
+
+void BasicRetinaFilter::_verticalAnticausalFilter_multGain(oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_float), &_a));
+    args.push_back(std::make_pair(sizeof(cl_float), &_gain));
+    openCLExecuteKernel(ctx, &retina_kernel, "verticalAnticausalFilter_multGain", globalSize, localSize, args, -1, -1);
+}
+
+void BasicRetinaFilter::_horizontalAnticausalFilter_Irregular(oclMat &outputFrame, const oclMat &spatialConstantBuffer)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {outputFrame.rows, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &spatialConstantBuffer.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &spatialConstantBuffer.offset));
+    openCLExecuteKernel(ctx, &retina_kernel, "horizontalAnticausalFilter_Irregular", globalSize, localSize, args, -1, -1);
+}
+
+//  vertical anticausal filter
+void BasicRetinaFilter::_verticalCausalFilter_Irregular(oclMat &outputFrame, const oclMat &spatialConstantBuffer)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {outputFrame.cols, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &spatialConstantBuffer.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &spatialConstantBuffer.offset));
+    openCLExecuteKernel(ctx, &retina_kernel, "verticalCausalFilter_Irregular", globalSize, localSize, args, -1, -1);
+}
+
+void normalizeGrayOutput_0_maxOutputValue(oclMat &inputOutputBuffer, const float maxOutputValue)
+{
+    double min_val, max_val;
+    ocl::minMax(inputOutputBuffer, &min_val, &max_val);
+    float factor = maxOutputValue / static_cast<float>(max_val - min_val);
+    float offset = - static_cast<float>(min_val) * factor;
+    ocl::multiply(factor, inputOutputBuffer, inputOutputBuffer);
+    ocl::add(inputOutputBuffer, offset, inputOutputBuffer);
+}
+
+void normalizeGrayOutputCentredSigmoide(const float meanValue, const float sensitivity, oclMat &in, oclMat &out, const float maxValue)
+{
+    if (sensitivity == 1.0f)
+    {
+        std::cerr << "TemplateBuffer::TemplateBuffer<type>::normalizeGrayOutputCentredSigmoide error: 2nd parameter (sensitivity) must not equal 0, copying original data..." << std::endl;
+        in.copyTo(out);
+        return;
+    }
+
+    float X0 = maxValue / (sensitivity - 1.0f);
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {in.cols, out.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    int elements_per_row = static_cast<int>(out.step / out.elemSize());
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &in.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &out.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &in.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &in.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &meanValue));
+    args.push_back(std::make_pair(sizeof(cl_float), &X0));
+    openCLExecuteKernel(ctx, &retina_kernel, "normalizeGrayOutputCentredSigmoide", globalSize, localSize, args, -1, -1);
+}
+
+void normalizeGrayOutputNearZeroCentreredSigmoide(oclMat &inputPicture, oclMat &outputBuffer, const float sensitivity, const float maxOutputValue)
+{
+    float X0cube = sensitivity * sensitivity * sensitivity;
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {inputPicture.cols, inputPicture.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    int elements_per_row = static_cast<int>(inputPicture.step / inputPicture.elemSize());
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputPicture.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputBuffer.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputPicture.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputPicture.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &maxOutputValue));
+    args.push_back(std::make_pair(sizeof(cl_float), &X0cube));
+    openCLExecuteKernel(ctx, &retina_kernel, "normalizeGrayOutputNearZeroCentreredSigmoide", globalSize, localSize, args, -1, -1);
+}
+
+void centerReductImageLuminance(oclMat &inputoutput)
+{
+    Scalar mean, stddev;
+    cv::meanStdDev((Mat)inputoutput, mean, stddev);
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {inputoutput.cols, inputoutput.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    float f_mean = static_cast<float>(mean[0]);
+    float f_stddev = static_cast<float>(stddev[0]);
+    int elements_per_row = static_cast<int>(inputoutput.step / inputoutput.elemSize());
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputoutput.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputoutput.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputoutput.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &f_mean));
+    args.push_back(std::make_pair(sizeof(cl_float), &f_stddev));
+    openCLExecuteKernel(ctx, &retina_kernel, "centerReductImageLuminance", globalSize, localSize, args, -1, -1);
+}
+
+///////////////////////////////////////
+///////// ParvoRetinaFilter ///////////
+///////////////////////////////////////
+ParvoRetinaFilter::ParvoRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns)
+    : BasicRetinaFilter(NBrows, NBcolumns, 3),
+      _photoreceptorsOutput(NBrows, NBcolumns, CV_32FC1),
+      _horizontalCellsOutput(NBrows, NBcolumns, CV_32FC1),
+      _parvocellularOutputON(NBrows, NBcolumns, CV_32FC1),
+      _parvocellularOutputOFF(NBrows, NBcolumns, CV_32FC1),
+      _bipolarCellsOutputON(NBrows, NBcolumns, CV_32FC1),
+      _bipolarCellsOutputOFF(NBrows, NBcolumns, CV_32FC1),
+      _localAdaptationOFF(NBrows, NBcolumns, CV_32FC1)
+{
+    // link to the required local parent adaptation buffers
+    _localAdaptationON = _localBuffer;
+    _parvocellularOutputONminusOFF = _filterOutput;
+
+    // init: set all the values to 0
+    clearAllBuffers();
+}
+
+ParvoRetinaFilter::~ParvoRetinaFilter()
+{
+}
+
+void ParvoRetinaFilter::clearAllBuffers()
+{
+    BasicRetinaFilter::clearAllBuffers();
+    _photoreceptorsOutput = 0;
+    _horizontalCellsOutput = 0;
+    _parvocellularOutputON = 0;
+    _parvocellularOutputOFF = 0;
+    _bipolarCellsOutputON = 0;
+    _bipolarCellsOutputOFF = 0;
+    _localAdaptationOFF = 0;
+}
+void ParvoRetinaFilter::resize(const unsigned int NBrows, const unsigned int NBcolumns)
+{
+    BasicRetinaFilter::resize(NBrows, NBcolumns);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _photoreceptorsOutput);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _horizontalCellsOutput);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _parvocellularOutputON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _parvocellularOutputOFF);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _bipolarCellsOutputON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _bipolarCellsOutputOFF);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _localAdaptationOFF);
+
+    // link to the required local parent adaptation buffers
+    _localAdaptationON = _localBuffer;
+    _parvocellularOutputONminusOFF = _filterOutput;
+
+    // clean buffers
+    clearAllBuffers();
+}
+
+void ParvoRetinaFilter::setOPLandParvoFiltersParameters(const float beta1, const float tau1, const float k1, const float beta2, const float tau2, const float k2)
+{
+    // init photoreceptors low pass filter
+    setLPfilterParameters(beta1, tau1, k1);
+    // init horizontal cells low pass filter
+    setLPfilterParameters(beta2, tau2, k2, 1);
+    // init parasol ganglion cells low pass filter (default parameters)
+    setLPfilterParameters(0, tau1, k1, 2);
+
+}
+const oclMat &ParvoRetinaFilter::runFilter(const oclMat &inputFrame, const bool useParvoOutput)
+{
+    _spatiotemporalLPfilter(inputFrame, _photoreceptorsOutput);
+    _spatiotemporalLPfilter(_photoreceptorsOutput, _horizontalCellsOutput, 1);
+    _OPL_OnOffWaysComputing();
+
+    if (useParvoOutput)
+    {
+        // local adaptation processes on ON and OFF ways
+        _spatiotemporalLPfilter(_bipolarCellsOutputON, _localAdaptationON, 2);
+        _localLuminanceAdaptation(_parvocellularOutputON, _localAdaptationON);
+        _spatiotemporalLPfilter(_bipolarCellsOutputOFF, _localAdaptationOFF, 2);
+        _localLuminanceAdaptation(_parvocellularOutputOFF, _localAdaptationOFF);
+        ocl::subtract(_parvocellularOutputON, _parvocellularOutputOFF, _parvocellularOutputONminusOFF);
+    }
+
+    return _parvocellularOutputONminusOFF;
+}
+void ParvoRetinaFilter::_OPL_OnOffWaysComputing()
+{
+    int elements_per_row = static_cast<int>(_photoreceptorsOutput.step / _photoreceptorsOutput.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {(_photoreceptorsOutput.cols + 3) / 4, _photoreceptorsOutput.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_photoreceptorsOutput.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_horizontalCellsOutput.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_bipolarCellsOutputON.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_bipolarCellsOutputOFF.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_parvocellularOutputON.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_parvocellularOutputOFF.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_photoreceptorsOutput.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_photoreceptorsOutput.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "OPL_OnOffWaysComputing", globalSize, localSize, args, -1, -1);
+}
+
+///////////////////////////////////////
+//////////// MagnoFilter //////////////
+///////////////////////////////////////
+MagnoRetinaFilter::MagnoRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns)
+    : BasicRetinaFilter(NBrows, NBcolumns, 2),
+      _previousInput_ON(NBrows, NBcolumns, CV_32FC1),
+      _previousInput_OFF(NBrows, NBcolumns, CV_32FC1),
+      _amacrinCellsTempOutput_ON(NBrows, NBcolumns, CV_32FC1),
+      _amacrinCellsTempOutput_OFF(NBrows, NBcolumns, CV_32FC1),
+      _magnoXOutputON(NBrows, NBcolumns, CV_32FC1),
+      _magnoXOutputOFF(NBrows, NBcolumns, CV_32FC1),
+      _localProcessBufferON(NBrows, NBcolumns, CV_32FC1),
+      _localProcessBufferOFF(NBrows, NBcolumns, CV_32FC1)
+{
+    _magnoYOutput = _filterOutput;
+    _magnoYsaturated = _localBuffer;
+
+    clearAllBuffers();
+}
+
+MagnoRetinaFilter::~MagnoRetinaFilter()
+{
+}
+void MagnoRetinaFilter::clearAllBuffers()
+{
+    BasicRetinaFilter::clearAllBuffers();
+    _previousInput_ON = 0;
+    _previousInput_OFF = 0;
+    _amacrinCellsTempOutput_ON = 0;
+    _amacrinCellsTempOutput_OFF = 0;
+    _magnoXOutputON = 0;
+    _magnoXOutputOFF = 0;
+    _localProcessBufferON = 0;
+    _localProcessBufferOFF = 0;
+
+}
+void MagnoRetinaFilter::resize(const unsigned int NBrows, const unsigned int NBcolumns)
+{
+    BasicRetinaFilter::resize(NBrows, NBcolumns);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _previousInput_ON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _previousInput_OFF);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _amacrinCellsTempOutput_ON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _amacrinCellsTempOutput_OFF);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _magnoXOutputON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _magnoXOutputOFF);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _localProcessBufferON);
+    ensureSizeIsEnough(NBrows, NBcolumns, CV_32FC1, _localProcessBufferOFF);
+
+    // to be sure, relink buffers
+    _magnoYOutput = _filterOutput;
+    _magnoYsaturated = _localBuffer;
+
+    // reset all buffers
+    clearAllBuffers();
+}
+
+void MagnoRetinaFilter::setCoefficientsTable(const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float localAdaptIntegration_tau, const float localAdaptIntegration_k )
+{
+    _temporalCoefficient = (float)std::exp(-1.0f / amacrinCellsTemporalCutFrequency);
+    // the first set of parameters is dedicated to the low pass filtering property of the ganglion cells
+    BasicRetinaFilter::setLPfilterParameters(parasolCells_beta, parasolCells_tau, parasolCells_k, 0);
+    // the second set of parameters is dedicated to the ganglion cells output intergartion for their local adaptation property
+    BasicRetinaFilter::setLPfilterParameters(0, localAdaptIntegration_tau, localAdaptIntegration_k, 1);
+}
+
+void MagnoRetinaFilter::_amacrineCellsComputing(
+    const oclMat &OPL_ON,
+    const oclMat &OPL_OFF
+)
+{
+    int elements_per_row = static_cast<int>(OPL_ON.step / OPL_ON.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {OPL_ON.cols, OPL_ON.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &OPL_ON.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &OPL_OFF.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_previousInput_ON.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_previousInput_OFF.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_amacrinCellsTempOutput_ON.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &_amacrinCellsTempOutput_OFF.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &OPL_ON.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &OPL_ON.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &_temporalCoefficient));
+    openCLExecuteKernel(ctx, &retina_kernel, "amacrineCellsComputing", globalSize, localSize, args, -1, -1);
+}
+
+const oclMat &MagnoRetinaFilter::runFilter(const oclMat &OPL_ON, const oclMat &OPL_OFF)
+{
+    // Compute the high pass temporal filter
+    _amacrineCellsComputing(OPL_ON, OPL_OFF);
+
+    // apply low pass filtering on ON and OFF ways after temporal high pass filtering
+    _spatiotemporalLPfilter(_amacrinCellsTempOutput_ON, _magnoXOutputON, 0);
+    _spatiotemporalLPfilter(_amacrinCellsTempOutput_OFF, _magnoXOutputOFF, 0);
+
+    // local adaptation of the ganglion cells to the local contrast of the moving contours
+    _spatiotemporalLPfilter(_magnoXOutputON, _localProcessBufferON, 1);
+    _localLuminanceAdaptation(_magnoXOutputON, _localProcessBufferON);
+
+    _spatiotemporalLPfilter(_magnoXOutputOFF, _localProcessBufferOFF, 1);
+    _localLuminanceAdaptation(_magnoXOutputOFF, _localProcessBufferOFF);
+
+    _magnoYOutput = _magnoXOutputON + _magnoXOutputOFF;
+
+    return _magnoYOutput;
+}
+
+///////////////////////////////////////
+//////////// RetinaColor //////////////
+///////////////////////////////////////
+
+// define an array of ROI headers of input x
+#define MAKE_OCLMAT_SLICES(x, n) \
+    oclMat x##_slices[n];\
+    for(int _SLICE_INDEX_ = 0; _SLICE_INDEX_ < n; _SLICE_INDEX_ ++)\
+    {\
+        x##_slices[_SLICE_INDEX_] = x(getROI(_SLICE_INDEX_));\
+    }
+
+RetinaColor::RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const int samplingMethod)
+    : BasicRetinaFilter(NBrows, NBcolumns, 3),
+      _RGBmosaic(NBrows * 3, NBcolumns, CV_32FC1),
+      _tempMultiplexedFrame(NBrows, NBcolumns, CV_32FC1),
+      _demultiplexedTempBuffer(NBrows * 3, NBcolumns, CV_32FC1),
+      _demultiplexedColorFrame(NBrows * 3, NBcolumns, CV_32FC1),
+      _chrominance(NBrows * 3, NBcolumns, CV_32FC1),
+      _colorLocalDensity(NBrows * 3, NBcolumns, CV_32FC1),
+      _imageGradient(NBrows * 3, NBcolumns, CV_32FC1)
+{
+    // link to parent buffers (let's recycle !)
+    _luminance = _filterOutput;
+    _multiplexedFrame = _localBuffer;
+
+    _objectInit = false;
+    _samplingMethod = samplingMethod;
+    _saturateColors = false;
+    _colorSaturationValue = 4.0;
+
+    // set default spatio-temporal filter parameters
+    setLPfilterParameters(0.0, 0.0, 1.5);
+    setLPfilterParameters(0.0, 0.0, 10.5, 1);// for the low pass filter dedicated to contours energy extraction (demultiplexing process)
+    setLPfilterParameters(0.f, 0.f, 0.9f, 2);
+
+    // init default value on image Gradient
+    _imageGradient = 0.57f;
+
+    // init color sampling map
+    _initColorSampling();
+
+    // flush all buffers
+    clearAllBuffers();
+}
+
+RetinaColor::~RetinaColor()
+{
+
+}
+
+void RetinaColor::clearAllBuffers()
+{
+    BasicRetinaFilter::clearAllBuffers();
+    _tempMultiplexedFrame = 0.f;
+    _demultiplexedTempBuffer = 0.f;
+
+    _demultiplexedColorFrame = 0.f;
+    _chrominance = 0.f;
+    _imageGradient = 0.57f;
+}
+
+void RetinaColor::resize(const unsigned int NBrows, const unsigned int NBcolumns)
+{
+    BasicRetinaFilter::clearAllBuffers();
+    ensureSizeIsEnough(NBrows,     NBcolumns, CV_32FC1, _tempMultiplexedFrame);
+    ensureSizeIsEnough(NBrows * 2, NBcolumns, CV_32FC1, _imageGradient);
+    ensureSizeIsEnough(NBrows * 3, NBcolumns, CV_32FC1, _RGBmosaic);
+    ensureSizeIsEnough(NBrows * 3, NBcolumns, CV_32FC1, _demultiplexedTempBuffer);
+    ensureSizeIsEnough(NBrows * 3, NBcolumns, CV_32FC1, _demultiplexedColorFrame);
+    ensureSizeIsEnough(NBrows * 3, NBcolumns, CV_32FC1, _chrominance);
+    ensureSizeIsEnough(NBrows * 3, NBcolumns, CV_32FC1, _colorLocalDensity);
+
+    // link to parent buffers (let's recycle !)
+    _luminance = _filterOutput;
+    _multiplexedFrame = _localBuffer;
+
+    // init color sampling map
+    _initColorSampling();
+
+    // clean buffers
+    clearAllBuffers();
+}
+
+static void inverseValue(oclMat &input)
+{
+    int elements_per_row = static_cast<int>(input.step / input.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {input.cols, input.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &input.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &input.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &input.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "inverseValue", globalSize, localSize, args, -1, -1);
+}
+
+void RetinaColor::_initColorSampling()
+{
+    CV_Assert(_samplingMethod == RETINA_COLOR_BAYER);
+    _pR = _pB = 0.25;
+    _pG = 0.5;
+    // filling the mosaic buffer:
+    _RGBmosaic = 0;
+    Mat tmp_mat(_NBrows * 3, _NBcols, CV_32FC1);
+    float * tmp_mat_ptr = tmp_mat.ptr<float>();
+    tmp_mat.setTo(0);
+    for (unsigned int index = 0 ; index < getNBpixels(); ++index)
+    {
+        tmp_mat_ptr[bayerSampleOffset(index)] = 1.0;
+    }
+    _RGBmosaic.upload(tmp_mat);
+    // computing photoreceptors local density
+    MAKE_OCLMAT_SLICES(_RGBmosaic, 3);
+    MAKE_OCLMAT_SLICES(_colorLocalDensity, 3);
+    _colorLocalDensity.setTo(0);
+    _spatiotemporalLPfilter(_RGBmosaic_slices[0], _colorLocalDensity_slices[0]);
+    _spatiotemporalLPfilter(_RGBmosaic_slices[1], _colorLocalDensity_slices[1]);
+    _spatiotemporalLPfilter(_RGBmosaic_slices[2], _colorLocalDensity_slices[2]);
+
+    //_colorLocalDensity = oclMat(_colorLocalDensity.size(), _colorLocalDensity.type(), 1.f) / _colorLocalDensity;
+    inverseValue(_colorLocalDensity);
+
+    _objectInit = true;
+}
+
+static void demultiplex(const oclMat &input, oclMat &ouput)
+{
+    int elements_per_row = static_cast<int>(input.step / input.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {input.cols, input.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &input.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &ouput.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &input.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &input.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "runColorDemultiplexingBayer", globalSize, localSize, args, -1, -1);
+}
+
+static void normalizePhotoDensity(
+    const oclMat &chroma,
+    const oclMat &colorDensity,
+    const oclMat &multiplex,
+    oclMat &ocl_luma,
+    oclMat &demultiplex,
+    const float pG
+)
+{
+    int elements_per_row = static_cast<int>(ocl_luma.step / ocl_luma.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {ocl_luma.cols, ocl_luma.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &chroma.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &colorDensity.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &multiplex.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &ocl_luma.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &demultiplex.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &ocl_luma.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &ocl_luma.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &pG));
+    openCLExecuteKernel(ctx, &retina_kernel, "normalizePhotoDensity", globalSize, localSize, args, -1, -1);
+}
+
+static void substractResidual(
+    oclMat &colorDemultiplex,
+    float pR,
+    float pG,
+    float pB
+)
+{
+    int elements_per_row = static_cast<int>(colorDemultiplex.step / colorDemultiplex.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    int rows = colorDemultiplex.rows / 3, cols = colorDemultiplex.cols;
+    size_t globalSize[] = {cols, rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &colorDemultiplex.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &pR));
+    args.push_back(std::make_pair(sizeof(cl_float), &pG));
+    args.push_back(std::make_pair(sizeof(cl_float), &pB));
+    openCLExecuteKernel(ctx, &retina_kernel, "substractResidual", globalSize, localSize, args, -1, -1);
+}
+
+static void demultiplexAssign(const oclMat& input, const oclMat& output)
+{
+    // only supports bayer
+    int elements_per_row = static_cast<int>(input.step / input.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    int rows = input.rows / 3, cols = input.cols;
+    size_t globalSize[] = {cols, rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &input.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &output.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "demultiplexAssign", globalSize, localSize, args, -1, -1);
+}
+
+void RetinaColor::runColorDemultiplexing(
+    const oclMat &ocl_multiplexed_input,
+    const bool adaptiveFiltering,
+    const float maxInputValue
+)
+{
+    MAKE_OCLMAT_SLICES(_demultiplexedTempBuffer, 3);
+    MAKE_OCLMAT_SLICES(_chrominance, 3);
+    MAKE_OCLMAT_SLICES(_RGBmosaic, 3);
+    MAKE_OCLMAT_SLICES(_demultiplexedColorFrame, 3);
+    MAKE_OCLMAT_SLICES(_colorLocalDensity, 3);
+
+    _demultiplexedTempBuffer.setTo(0);
+    demultiplex(ocl_multiplexed_input, _demultiplexedTempBuffer);
+
+    // interpolate the demultiplexed frame depending on the color sampling method
+    if (!adaptiveFiltering)
+    {
+        CV_Assert(adaptiveFiltering == false);
+    }
+
+    _spatiotemporalLPfilter(_demultiplexedTempBuffer_slices[0], _chrominance_slices[0]);
+    _spatiotemporalLPfilter(_demultiplexedTempBuffer_slices[1], _chrominance_slices[1]);
+    _spatiotemporalLPfilter(_demultiplexedTempBuffer_slices[2], _chrominance_slices[2]);
+
+    if (!adaptiveFiltering)// compute the gradient on the luminance
+    {
+        // TODO: implement me!
+        CV_Assert(adaptiveFiltering == false);
+    }
+    else
+    {
+        normalizePhotoDensity(_chrominance, _colorLocalDensity, ocl_multiplexed_input, _luminance, _demultiplexedTempBuffer, _pG);
+        // compute the gradient of the luminance
+        _computeGradient(_luminance, _imageGradient);
+
+        _adaptiveSpatialLPfilter(_RGBmosaic_slices[0], _imageGradient, _chrominance_slices[0]);
+        _adaptiveSpatialLPfilter(_RGBmosaic_slices[1], _imageGradient, _chrominance_slices[1]);
+        _adaptiveSpatialLPfilter(_RGBmosaic_slices[2], _imageGradient, _chrominance_slices[2]);
+
+        _adaptiveSpatialLPfilter(_demultiplexedTempBuffer_slices[0], _imageGradient, _demultiplexedColorFrame_slices[0]);
+        _adaptiveSpatialLPfilter(_demultiplexedTempBuffer_slices[1], _imageGradient, _demultiplexedColorFrame_slices[1]);
+        _adaptiveSpatialLPfilter(_demultiplexedTempBuffer_slices[2], _imageGradient, _demultiplexedColorFrame_slices[2]);
+
+        _demultiplexedColorFrame /= _chrominance; // per element division
+        substractResidual(_demultiplexedColorFrame, _pR, _pG, _pB);
+        runColorMultiplexing(_demultiplexedColorFrame, _tempMultiplexedFrame);
+
+        _demultiplexedTempBuffer.setTo(0);
+        _luminance = ocl_multiplexed_input - _tempMultiplexedFrame;
+        demultiplexAssign(_demultiplexedColorFrame, _demultiplexedTempBuffer);
+
+        for(int i = 0; i < 3; i ++)
+        {
+            _spatiotemporalLPfilter(_demultiplexedTempBuffer_slices[i], _demultiplexedTempBuffer_slices[i]);
+            _demultiplexedColorFrame_slices[i] = _demultiplexedTempBuffer_slices[i] * _colorLocalDensity_slices[i] + _luminance;
+        }
+    }
+    // eliminate saturated colors by simple clipping values to the input range
+    clipRGBOutput_0_maxInputValue(_demultiplexedColorFrame, maxInputValue);
+
+    if (_saturateColors)
+    {
+        ocl::normalizeGrayOutputCentredSigmoide(128, maxInputValue, _demultiplexedColorFrame, _demultiplexedColorFrame);
+    }
+}
+void RetinaColor::runColorMultiplexing(const oclMat &demultiplexedInputFrame, oclMat &multiplexedFrame)
+{
+    int elements_per_row = static_cast<int>(multiplexedFrame.step / multiplexedFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {multiplexedFrame.cols, multiplexedFrame.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &demultiplexedInputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &multiplexedFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &multiplexedFrame.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &multiplexedFrame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "runColorMultiplexingBayer", globalSize, localSize, args, -1, -1);
+}
+
+void RetinaColor::clipRGBOutput_0_maxInputValue(oclMat &inputOutputBuffer, const float maxInputValue)
+{
+    // the kernel is equivalent to:
+    //ocl::threshold(inputOutputBuffer, inputOutputBuffer, maxInputValue, maxInputValue, THRESH_TRUNC);
+    //ocl::threshold(inputOutputBuffer, inputOutputBuffer, 0, 0, THRESH_TOZERO);
+    int elements_per_row = static_cast<int>(inputOutputBuffer.step / inputOutputBuffer.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, inputOutputBuffer.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputOutputBuffer.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputOutputBuffer.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &maxInputValue));
+    openCLExecuteKernel(ctx, &retina_kernel, "clipRGBOutput_0_maxInputValue", globalSize, localSize, args, -1, -1);
+}
+
+void RetinaColor::_adaptiveSpatialLPfilter(const oclMat &inputFrame, const oclMat &gradient, oclMat &outputFrame)
+{
+    /**********/
+    _gain = (1 - 0.57f) * (1 - 0.57f) * (1 - 0.06f) * (1 - 0.06f);
+
+    // launch the serie of 1D directional filters in order to compute the 2D low pass filter
+    // -> horizontal filters work with the first layer of imageGradient
+    _adaptiveHorizontalCausalFilter_addInput(inputFrame, gradient, outputFrame);
+    _horizontalAnticausalFilter_Irregular(outputFrame, gradient);
+    // -> horizontal filters work with the second layer of imageGradient
+    _verticalCausalFilter_Irregular(outputFrame, gradient(getROI(1)));
+    _adaptiveVerticalAnticausalFilter_multGain(gradient, outputFrame);
+}
+
+void RetinaColor::_adaptiveHorizontalCausalFilter_addInput(const oclMat &inputFrame, const oclMat &gradient, oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(inputFrame.step / inputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBrows, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &inputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &gradient.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &inputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &gradient.offset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    openCLExecuteKernel(ctx, &retina_kernel, "adaptiveHorizontalCausalFilter_addInput", globalSize, localSize, args, -1, -1);
+}
+
+void RetinaColor::_adaptiveVerticalAnticausalFilter_multGain(const oclMat &gradient, oclMat &outputFrame)
+{
+    int elements_per_row = static_cast<int>(outputFrame.step / outputFrame.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, 1, 1};
+    size_t localSize[]  = {256, 1, 1};
+
+    int gradOffset = gradient.offset + static_cast<int>(gradient.step * _NBrows);
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &gradient.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &outputFrame.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_int),   &gradOffset));
+    args.push_back(std::make_pair(sizeof(cl_int),   &outputFrame.offset));
+    args.push_back(std::make_pair(sizeof(cl_float), &_gain));
+    openCLExecuteKernel(ctx, &retina_kernel, "adaptiveVerticalAnticausalFilter_multGain", globalSize, localSize, args, -1, -1);
+}
+void RetinaColor::_computeGradient(const oclMat &luminance, oclMat &gradient)
+{
+    int elements_per_row = static_cast<int>(luminance.step / luminance.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {_NBcols, _NBrows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &luminance.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &gradient.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBcols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &_NBrows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    openCLExecuteKernel(ctx, &retina_kernel, "computeGradient", globalSize, localSize, args, -1, -1);
+}
+
+///////////////////////////////////////
+//////////// RetinaFilter /////////////
+///////////////////////////////////////
+RetinaFilter::RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode, const int samplingMethod, const bool useRetinaLogSampling, const double, const double)
+    :
+    _photoreceptorsPrefilter(sizeRows, sizeColumns, 4),
+    _ParvoRetinaFilter(sizeRows, sizeColumns),
+    _MagnoRetinaFilter(sizeRows, sizeColumns),
+    _colorEngine(sizeRows, sizeColumns, samplingMethod)
+{
+    CV_Assert(!useRetinaLogSampling);
+
+    // set default processing activities
+    _useParvoOutput = true;
+    _useMagnoOutput = true;
+
+    _useColorMode = colorMode;
+
+    // set default parameters
+    setGlobalParameters();
+
+    // stability controls values init
+    _setInitPeriodCount();
+    _globalTemporalConstant = 25;
+
+    // reset all buffers
+    clearAllBuffers();
+}
+
+RetinaFilter::~RetinaFilter()
+{
+}
+
+void RetinaFilter::clearAllBuffers()
+{
+    _photoreceptorsPrefilter.clearAllBuffers();
+    _ParvoRetinaFilter.clearAllBuffers();
+    _MagnoRetinaFilter.clearAllBuffers();
+    _colorEngine.clearAllBuffers();
+    // stability controls value init
+    _setInitPeriodCount();
+}
+
+void RetinaFilter::resize(const unsigned int NBrows, const unsigned int NBcolumns)
+{
+    unsigned int rows = NBrows, cols = NBcolumns;
+
+    // resize optionnal member and adjust other modules size if required
+    _photoreceptorsPrefilter.resize(rows, cols);
+    _ParvoRetinaFilter.resize(rows, cols);
+    _MagnoRetinaFilter.resize(rows, cols);
+    _colorEngine.resize(rows, cols);
+
+    // clean buffers
+    clearAllBuffers();
+
+}
+
+void RetinaFilter::_setInitPeriodCount()
+{
+    // find out the maximum temporal constant value and apply a security factor
+    // false value (obviously too long) but appropriate for simple use
+    _globalTemporalConstant = (unsigned int)(_ParvoRetinaFilter.getPhotoreceptorsTemporalConstant() + _ParvoRetinaFilter.getHcellsTemporalConstant() + _MagnoRetinaFilter.getTemporalConstant());
+    // reset frame counter
+    _ellapsedFramesSinceLastReset = 0;
+}
+
+void RetinaFilter::setGlobalParameters(const float OPLspatialResponse1, const float OPLtemporalresponse1, const float OPLassymetryGain, const float OPLspatialResponse2, const float OPLtemporalresponse2, const float LPfilterSpatialResponse, const float LPfilterGain, const float LPfilterTemporalresponse, const float MovingContoursExtractorCoefficient, const bool normalizeParvoOutput_0_maxOutputValue, const bool normalizeMagnoOutput_0_maxOutputValue, const float maxOutputValue, const float maxInputValue, const float meanValue)
+{
+    _normalizeParvoOutput_0_maxOutputValue = normalizeParvoOutput_0_maxOutputValue;
+    _normalizeMagnoOutput_0_maxOutputValue = normalizeMagnoOutput_0_maxOutputValue;
+    _maxOutputValue = maxOutputValue;
+    _photoreceptorsPrefilter.setV0CompressionParameter(0.9f, maxInputValue, meanValue);
+    _photoreceptorsPrefilter.setLPfilterParameters(0, 0, 10, 3); // keeps low pass filter with low cut frequency in memory (usefull for the tone mapping function)
+    _ParvoRetinaFilter.setOPLandParvoFiltersParameters(0, OPLtemporalresponse1, OPLspatialResponse1, OPLassymetryGain, OPLtemporalresponse2, OPLspatialResponse2);
+    _ParvoRetinaFilter.setV0CompressionParameter(0.9f, maxInputValue, meanValue);
+    _MagnoRetinaFilter.setCoefficientsTable(LPfilterGain, LPfilterTemporalresponse, LPfilterSpatialResponse, MovingContoursExtractorCoefficient, 0, 2.0f * LPfilterSpatialResponse);
+    _MagnoRetinaFilter.setV0CompressionParameter(0.7f, maxInputValue, meanValue);
+
+    // stability controls value init
+    _setInitPeriodCount();
+}
+
+bool RetinaFilter::checkInput(const oclMat &input, const bool)
+{
+    BasicRetinaFilter *inputTarget = &_photoreceptorsPrefilter;
+
+    bool test = (input.rows == static_cast<int>(inputTarget->getNBrows())
+                 || input.rows == static_cast<int>(inputTarget->getNBrows()) * 3
+                 || input.rows == static_cast<int>(inputTarget->getNBrows()) * 4)
+                && input.cols == static_cast<int>(inputTarget->getNBcolumns());
+    if (!test)
+    {
+        std::cerr << "RetinaFilter::checkInput: input buffer does not match retina buffer size, conversion aborted" << std::endl;
+        return false;
+    }
+
+    return true;
+}
+
+// main function that runs the filter for a given input frame
+bool RetinaFilter::runFilter(const oclMat &imageInput, const bool useAdaptiveFiltering, const bool processRetinaParvoMagnoMapping, const bool useColorMode, const bool inputIsColorMultiplexed)
+{
+    // preliminary check
+    bool processSuccess = true;
+    if (!checkInput(imageInput, useColorMode))
+    {
+        return false;
+    }
+
+    // run the color multiplexing if needed and compute each suub filter of the retina:
+    // -> local adaptation
+    // -> contours OPL extraction
+    // -> moving contours extraction
+
+    // stability controls value update
+    ++_ellapsedFramesSinceLastReset;
+
+    _useColorMode = useColorMode;
+
+    oclMat selectedPhotoreceptorsLocalAdaptationInput = imageInput;
+    oclMat selectedPhotoreceptorsColorInput = imageInput;
+
+    //********** Following is input data specific photoreceptors processing
+    if (useColorMode && (!inputIsColorMultiplexed)) // not multiplexed color input case
+    {
+        _colorEngine.runColorMultiplexing(selectedPhotoreceptorsColorInput);
+        selectedPhotoreceptorsLocalAdaptationInput = _colorEngine.getMultiplexedFrame();
+    }
+    //********** Following is generic Retina processing
+
+    // photoreceptors local adaptation
+    _photoreceptorsPrefilter.runFilter_LocalAdapdation(selectedPhotoreceptorsLocalAdaptationInput, _ParvoRetinaFilter.getHorizontalCellsOutput());
+
+    // run parvo filter
+    _ParvoRetinaFilter.runFilter(_photoreceptorsPrefilter.getOutput(), _useParvoOutput);
+
+    if (_useParvoOutput)
+    {
+        _ParvoRetinaFilter.normalizeGrayOutputCentredSigmoide(); // models the saturation of the cells, usefull for visualisation of the ON-OFF Parvo Output, Bipolar cells outputs do not change !!!
+        _ParvoRetinaFilter.centerReductImageLuminance(); // best for further spectrum analysis
+
+        if (_normalizeParvoOutput_0_maxOutputValue)
+        {
+            _ParvoRetinaFilter.normalizeGrayOutput_0_maxOutputValue(_maxOutputValue);
+        }
+    }
+
+    if (_useParvoOutput && _useMagnoOutput)
+    {
+        _MagnoRetinaFilter.runFilter(_ParvoRetinaFilter.getBipolarCellsON(), _ParvoRetinaFilter.getBipolarCellsOFF());
+        if (_normalizeMagnoOutput_0_maxOutputValue)
+        {
+            _MagnoRetinaFilter.normalizeGrayOutput_0_maxOutputValue(_maxOutputValue);
+        }
+        _MagnoRetinaFilter.normalizeGrayOutputNearZeroCentreredSigmoide();
+    }
+
+    if (_useParvoOutput && _useMagnoOutput && processRetinaParvoMagnoMapping)
+    {
+        _processRetinaParvoMagnoMapping();
+        if (_useColorMode)
+        {
+            _colorEngine.runColorDemultiplexing(_retinaParvoMagnoMappedFrame, useAdaptiveFiltering, _maxOutputValue);
+        }
+        return processSuccess;
+    }
+
+    if (_useParvoOutput && _useColorMode)
+    {
+        _colorEngine.runColorDemultiplexing(_ParvoRetinaFilter.getOutput(), useAdaptiveFiltering, _maxOutputValue);
+    }
+    return processSuccess;
+}
+
+const oclMat &RetinaFilter::getContours()
+{
+    if (_useColorMode)
+    {
+        return _colorEngine.getLuminance();
+    }
+    else
+    {
+        return _ParvoRetinaFilter.getOutput();
+    }
+}
+void RetinaFilter::_processRetinaParvoMagnoMapping()
+{
+    oclMat parvo = _ParvoRetinaFilter.getOutput();
+    oclMat magno = _MagnoRetinaFilter.getOutput();
+
+    int halfRows = parvo.rows / 2;
+    int halfCols = parvo.cols / 2;
+    float minDistance = MIN(halfRows, halfCols) * 0.7f;
+
+    int elements_per_row = static_cast<int>(parvo.step / parvo.elemSize());
+
+    Context * ctx = Context::getContext();
+    std::vector<std::pair<size_t, const void *> > args;
+    size_t globalSize[] = {parvo.cols, parvo.rows, 1};
+    size_t localSize[]  = {16, 16, 1};
+
+    args.push_back(std::make_pair(sizeof(cl_mem),   &parvo.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),   &magno.data));
+    args.push_back(std::make_pair(sizeof(cl_int),   &parvo.cols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &parvo.rows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &halfCols));
+    args.push_back(std::make_pair(sizeof(cl_int),   &halfRows));
+    args.push_back(std::make_pair(sizeof(cl_int),   &elements_per_row));
+    args.push_back(std::make_pair(sizeof(cl_float), &minDistance));
+    openCLExecuteKernel(ctx, &retina_kernel, "processRetinaParvoMagnoMapping", globalSize, localSize, args, -1, -1);
+}
+}  /* namespace ocl */
+
+Ptr<Retina> createRetina_OCL(Size getInputSize){ return makePtr<ocl::RetinaOCLImpl>(getInputSize); }
+Ptr<Retina> createRetina_OCL(Size getInputSize, const bool colorMode, int colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+{
+    return makePtr<ocl::RetinaOCLImpl>(getInputSize, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
+}
+
+}  /* namespace bioinspired */
+}  /* namespace cv */
+
+#endif /* #ifdef HAVE_OPENCV_OCL */
diff --git a/modules/bioinspired/src/retina_ocl.hpp b/modules/bioinspired/src/retina_ocl.hpp
new file mode 100644
index 0000000000..90df0601c4
--- /dev/null
+++ b/modules/bioinspired/src/retina_ocl.hpp
@@ -0,0 +1,634 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OCL_RETINA_HPP__
+#define __OCL_RETINA_HPP__
+
+#include "precomp.hpp"
+
+#ifdef HAVE_OPENCV_OCL
+
+// please refer to c++ headers for API comments
+namespace cv
+{
+namespace bioinspired
+{
+namespace ocl
+{
+void normalizeGrayOutputCentredSigmoide(const float meanValue, const float sensitivity, cv::ocl::oclMat &in, cv::ocl::oclMat &out, const float maxValue = 255.f);
+void normalizeGrayOutput_0_maxOutputValue(cv::ocl::oclMat &inputOutputBuffer, const float maxOutputValue = 255.0);
+void normalizeGrayOutputNearZeroCentreredSigmoide(cv::ocl::oclMat &inputPicture, cv::ocl::oclMat &outputBuffer, const float sensitivity = 40, const float maxOutputValue = 255.0f);
+void centerReductImageLuminance(cv::ocl::oclMat &inputOutputBuffer);
+
+class BasicRetinaFilter
+{
+public:
+    BasicRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns, const unsigned int parametersListSize = 1, const bool useProgressiveFilter = false);
+    ~BasicRetinaFilter();
+    inline void clearOutputBuffer()
+    {
+        _filterOutput = 0;
+    };
+    inline void clearSecondaryBuffer()
+    {
+        _localBuffer = 0;
+    };
+    inline void clearAllBuffers()
+    {
+        clearOutputBuffer();
+        clearSecondaryBuffer();
+    };
+    void  resize(const unsigned int NBrows, const unsigned int NBcolumns);
+    const cv::ocl::oclMat &runFilter_LPfilter(const cv::ocl::oclMat &inputFrame, const unsigned int filterIndex = 0);
+    void  runFilter_LPfilter(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame, const unsigned int filterIndex = 0);
+    void  runFilter_LPfilter_Autonomous(cv::ocl::oclMat &inputOutputFrame, const unsigned int filterIndex = 0);
+    const cv::ocl::oclMat &runFilter_LocalAdapdation(const cv::ocl::oclMat &inputOutputFrame, const cv::ocl::oclMat &localLuminance);
+    void  runFilter_LocalAdapdation(const cv::ocl::oclMat &inputFrame, const cv::ocl::oclMat &localLuminance, cv::ocl::oclMat &outputFrame);
+    const cv::ocl::oclMat &runFilter_LocalAdapdation_autonomous(const cv::ocl::oclMat &inputFrame);
+    void  runFilter_LocalAdapdation_autonomous(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame);
+    void  setLPfilterParameters(const float beta, const float tau, const float k, const unsigned int filterIndex = 0);
+    inline void setV0CompressionParameter(const float v0, const float maxInputValue, const float)
+    {
+        _v0 = v0 * maxInputValue;
+        _localLuminanceFactor = v0;
+        _localLuminanceAddon = maxInputValue * (1.0f - v0);
+        _maxInputValue = maxInputValue;
+    };
+    inline void setV0CompressionParameter(const float v0, const float meanLuminance)
+    {
+        this->setV0CompressionParameter(v0, _maxInputValue, meanLuminance);
+    };
+    inline void setV0CompressionParameter(const float v0)
+    {
+        _v0 = v0 * _maxInputValue;
+        _localLuminanceFactor = v0;
+        _localLuminanceAddon = _maxInputValue * (1.0f - v0);
+    };
+    inline void setV0CompressionParameterToneMapping(const float v0, const float maxInputValue, const float meanLuminance = 128.0f)
+    {
+        _v0 = v0 * maxInputValue;
+        _localLuminanceFactor = 1.0f;
+        _localLuminanceAddon = meanLuminance * _v0;
+        _maxInputValue = maxInputValue;
+    };
+    inline void updateCompressionParameter(const float meanLuminance)
+    {
+        _localLuminanceFactor = 1;
+        _localLuminanceAddon = meanLuminance * _v0;
+    };
+    inline float getV0CompressionParameter()
+    {
+        return _v0 / _maxInputValue;
+    };
+    inline const cv::ocl::oclMat &getOutput() const
+    {
+        return _filterOutput;
+    };
+    inline unsigned int getNBrows()
+    {
+        return _filterOutput.rows;
+    };
+    inline unsigned int getNBcolumns()
+    {
+        return _filterOutput.cols;
+    };
+    inline unsigned int getNBpixels()
+    {
+        return _filterOutput.size().area();
+    };
+    inline void normalizeGrayOutput_0_maxOutputValue(const float maxValue)
+    {
+        ocl::normalizeGrayOutput_0_maxOutputValue(_filterOutput, maxValue);
+    };
+    inline void normalizeGrayOutputCentredSigmoide()
+    {
+        ocl::normalizeGrayOutputCentredSigmoide(0.0, 2.0, _filterOutput, _filterOutput);
+    };
+    inline void centerReductImageLuminance()
+    {
+        ocl::centerReductImageLuminance(_filterOutput);
+    };
+    inline float getMaxInputValue()
+    {
+        return this->_maxInputValue;
+    };
+    inline void setMaxInputValue(const float newMaxInputValue)
+    {
+        this->_maxInputValue = newMaxInputValue;
+    };
+
+protected:
+
+    int _NBrows;
+    int _NBcols;
+    unsigned int _halfNBrows;
+    unsigned int _halfNBcolumns;
+
+    cv::ocl::oclMat _filterOutput;
+    cv::ocl::oclMat _localBuffer;
+
+    std::valarray <float>_filteringCoeficientsTable;
+    float _v0;
+    float _maxInputValue;
+    float _meanInputValue;
+    float _localLuminanceFactor;
+    float _localLuminanceAddon;
+
+    float _a;
+    float _tau;
+    float _gain;
+
+    void _spatiotemporalLPfilter(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &LPfilterOutput, const unsigned int coefTableOffset = 0);
+    float _squaringSpatiotemporalLPfilter(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame, const unsigned int filterIndex = 0);
+    void _spatiotemporalLPfilter_Irregular(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame, const unsigned int filterIndex = 0);
+    void _localSquaringSpatioTemporalLPfilter(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &LPfilterOutput, const unsigned int *integrationAreas, const unsigned int filterIndex = 0);
+    void _localLuminanceAdaptation(const cv::ocl::oclMat &inputFrame, const cv::ocl::oclMat &localLuminance, cv::ocl::oclMat &outputFrame, const bool updateLuminanceMean = true);
+    void _localLuminanceAdaptation(cv::ocl::oclMat &inputOutputFrame, const cv::ocl::oclMat &localLuminance);
+    void _localLuminanceAdaptationPosNegValues(const cv::ocl::oclMat &inputFrame, const cv::ocl::oclMat &localLuminance, float *outputFrame);
+    void _horizontalCausalFilter_addInput(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame);
+    void _horizontalAnticausalFilter(cv::ocl::oclMat &outputFrame);
+    void _verticalCausalFilter(cv::ocl::oclMat &outputFrame);
+    void _horizontalAnticausalFilter_Irregular(cv::ocl::oclMat &outputFrame, const cv::ocl::oclMat &spatialConstantBuffer);
+    void _verticalCausalFilter_Irregular(cv::ocl::oclMat &outputFrame, const cv::ocl::oclMat &spatialConstantBuffer);
+    void _verticalAnticausalFilter_multGain(cv::ocl::oclMat &outputFrame);
+};
+
+class MagnoRetinaFilter: public BasicRetinaFilter
+{
+public:
+    MagnoRetinaFilter(const unsigned int NBrows, const unsigned int NBcolumns);
+    virtual ~MagnoRetinaFilter();
+    void clearAllBuffers();
+    void resize(const unsigned int NBrows, const unsigned int NBcolumns);
+    void setCoefficientsTable(const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float localAdaptIntegration_tau, const float localAdaptIntegration_k);
+
+    const cv::ocl::oclMat &runFilter(const cv::ocl::oclMat &OPL_ON, const cv::ocl::oclMat &OPL_OFF);
+
+    inline const cv::ocl::oclMat &getMagnoON() const
+    {
+        return _magnoXOutputON;
+    };
+    inline const cv::ocl::oclMat &getMagnoOFF() const
+    {
+        return _magnoXOutputOFF;
+    };
+    inline const cv::ocl::oclMat &getMagnoYsaturated() const
+    {
+        return _magnoYsaturated;
+    };
+    inline void normalizeGrayOutputNearZeroCentreredSigmoide()
+    {
+        ocl::normalizeGrayOutputNearZeroCentreredSigmoide(_magnoYOutput, _magnoYsaturated);
+    };
+    inline float getTemporalConstant()
+    {
+        return this->_filteringCoeficientsTable[2];
+    };
+private:
+    cv::ocl::oclMat _previousInput_ON;
+    cv::ocl::oclMat _previousInput_OFF;
+    cv::ocl::oclMat _amacrinCellsTempOutput_ON;
+    cv::ocl::oclMat _amacrinCellsTempOutput_OFF;
+    cv::ocl::oclMat _magnoXOutputON;
+    cv::ocl::oclMat _magnoXOutputOFF;
+    cv::ocl::oclMat _localProcessBufferON;
+    cv::ocl::oclMat _localProcessBufferOFF;
+    cv::ocl::oclMat _magnoYOutput;
+    cv::ocl::oclMat _magnoYsaturated;
+
+    float _temporalCoefficient;
+    void _amacrineCellsComputing(const cv::ocl::oclMat &OPL_ON,  const cv::ocl::oclMat &OPL_OFF);
+};
+
+class ParvoRetinaFilter: public BasicRetinaFilter
+{
+public:
+    ParvoRetinaFilter(const unsigned int NBrows = 480, const unsigned int NBcolumns = 640);
+    virtual ~ParvoRetinaFilter();
+    void resize(const unsigned int NBrows, const unsigned int NBcolumns);
+    void clearAllBuffers();
+    void setOPLandParvoFiltersParameters(const float beta1, const float tau1, const float k1, const float beta2, const float tau2, const float k2);
+
+    inline void setGanglionCellsLocalAdaptationLPfilterParameters(const float tau, const float k)
+    {
+        BasicRetinaFilter::setLPfilterParameters(0, tau, k, 2);
+    };
+    const cv::ocl::oclMat &runFilter(const cv::ocl::oclMat &inputFrame, const bool useParvoOutput = true);
+
+    inline const cv::ocl::oclMat &getPhotoreceptorsLPfilteringOutput() const
+    {
+        return _photoreceptorsOutput;
+    };
+
+    inline const cv::ocl::oclMat &getHorizontalCellsOutput() const
+    {
+        return _horizontalCellsOutput;
+    };
+
+    inline const cv::ocl::oclMat &getParvoON() const
+    {
+        return _parvocellularOutputON;
+    };
+
+    inline const cv::ocl::oclMat &getParvoOFF() const
+    {
+        return _parvocellularOutputOFF;
+    };
+
+    inline const cv::ocl::oclMat &getBipolarCellsON() const
+    {
+        return _bipolarCellsOutputON;
+    };
+
+    inline const cv::ocl::oclMat &getBipolarCellsOFF() const
+    {
+        return _bipolarCellsOutputOFF;
+    };
+
+    inline float getPhotoreceptorsTemporalConstant()
+    {
+        return this->_filteringCoeficientsTable[2];
+    };
+
+    inline float getHcellsTemporalConstant()
+    {
+        return this->_filteringCoeficientsTable[5];
+    };
+private:
+    cv::ocl::oclMat _photoreceptorsOutput;
+    cv::ocl::oclMat _horizontalCellsOutput;
+    cv::ocl::oclMat _parvocellularOutputON;
+    cv::ocl::oclMat _parvocellularOutputOFF;
+    cv::ocl::oclMat _bipolarCellsOutputON;
+    cv::ocl::oclMat _bipolarCellsOutputOFF;
+    cv::ocl::oclMat _localAdaptationOFF;
+    cv::ocl::oclMat _localAdaptationON;
+    cv::ocl::oclMat _parvocellularOutputONminusOFF;
+    void _OPL_OnOffWaysComputing();
+};
+class RetinaColor: public BasicRetinaFilter
+{
+public:
+    RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const int samplingMethod = RETINA_COLOR_DIAGONAL);
+    virtual ~RetinaColor();
+
+    void clearAllBuffers();
+    void resize(const unsigned int NBrows, const unsigned int NBcolumns);
+    inline void runColorMultiplexing(const cv::ocl::oclMat &inputRGBFrame)
+    {
+        runColorMultiplexing(inputRGBFrame, _multiplexedFrame);
+    };
+    void runColorMultiplexing(const cv::ocl::oclMat &demultiplexedInputFrame, cv::ocl::oclMat &multiplexedFrame);
+    void runColorDemultiplexing(const cv::ocl::oclMat &multiplexedColorFrame, const bool adaptiveFiltering = false, const float maxInputValue = 255.0);
+
+    void setColorSaturation(const bool saturateColors = true, const float colorSaturationValue = 4.0)
+    {
+        _saturateColors = saturateColors;
+        _colorSaturationValue = colorSaturationValue;
+    };
+
+    void setChrominanceLPfilterParameters(const float beta, const float tau, const float k)
+    {
+        setLPfilterParameters(beta, tau, k);
+    };
+
+    bool applyKrauskopfLMS2Acr1cr2Transform(cv::ocl::oclMat &result);
+    bool applyLMS2LabTransform(cv::ocl::oclMat &result);
+    inline const cv::ocl::oclMat &getMultiplexedFrame() const
+    {
+        return _multiplexedFrame;
+    };
+
+    inline const cv::ocl::oclMat &getDemultiplexedColorFrame() const
+    {
+        return _demultiplexedColorFrame;
+    };
+
+    inline const cv::ocl::oclMat &getLuminance() const
+    {
+        return _luminance;
+    };
+    inline const cv::ocl::oclMat &getChrominance() const
+    {
+        return _chrominance;
+    };
+    void clipRGBOutput_0_maxInputValue(cv::ocl::oclMat &inputOutputBuffer, const float maxOutputValue = 255.0);
+    void normalizeRGBOutput_0_maxOutputValue(const float maxOutputValue = 255.0);
+    inline void setDemultiplexedColorFrame(const cv::ocl::oclMat &demultiplexedImage)
+    {
+        _demultiplexedColorFrame = demultiplexedImage;
+    };
+protected:
+    inline unsigned int bayerSampleOffset(unsigned int index)
+    {
+        return index + ((index / getNBcolumns()) % 2) * getNBpixels() + ((index % getNBcolumns()) % 2) * getNBpixels();
+    }
+    inline Rect getROI(int idx)
+    {
+        return Rect(0, idx * _NBrows, _NBcols, _NBrows);
+    }
+    int _samplingMethod;
+    bool _saturateColors;
+    float _colorSaturationValue;
+    cv::ocl::oclMat _luminance;
+    cv::ocl::oclMat _multiplexedFrame;
+    cv::ocl::oclMat _RGBmosaic;
+    cv::ocl::oclMat _tempMultiplexedFrame;
+    cv::ocl::oclMat _demultiplexedTempBuffer;
+    cv::ocl::oclMat _demultiplexedColorFrame;
+    cv::ocl::oclMat _chrominance;
+    cv::ocl::oclMat _colorLocalDensity;
+    cv::ocl::oclMat _imageGradient;
+
+    float _pR, _pG, _pB;
+    bool _objectInit;
+
+    void _initColorSampling();
+    void _adaptiveSpatialLPfilter(const cv::ocl::oclMat &inputFrame, const cv::ocl::oclMat &gradient, cv::ocl::oclMat &outputFrame);
+    void _adaptiveHorizontalCausalFilter_addInput(const cv::ocl::oclMat &inputFrame, const cv::ocl::oclMat &gradient, cv::ocl::oclMat &outputFrame);
+    void _adaptiveVerticalAnticausalFilter_multGain(const cv::ocl::oclMat &gradient, cv::ocl::oclMat &outputFrame);
+    void _computeGradient(const cv::ocl::oclMat &luminance, cv::ocl::oclMat &gradient);
+    void _normalizeOutputs_0_maxOutputValue(void);
+    void _applyImageColorSpaceConversion(const cv::ocl::oclMat &inputFrame, cv::ocl::oclMat &outputFrame, const float *transformTable);
+};
+class RetinaFilter
+{
+public:
+    RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode = false, const int samplingMethod = RETINA_COLOR_BAYER, const bool useRetinaLogSampling = false, const double reductionFactor = 1.0, const double samplingStrenght = 10.0);
+    ~RetinaFilter();
+
+    void clearAllBuffers();
+    void resize(const unsigned int NBrows, const unsigned int NBcolumns);
+    bool checkInput(const cv::ocl::oclMat &input, const bool colorMode);
+    bool runFilter(const cv::ocl::oclMat &imageInput, const bool useAdaptiveFiltering = true, const bool processRetinaParvoMagnoMapping = false, const bool useColorMode = false, const bool inputIsColorMultiplexed = false);
+
+    void setGlobalParameters(const float OPLspatialResponse1 = 0.7, const float OPLtemporalresponse1 = 1, const float OPLassymetryGain = 0, const float OPLspatialResponse2 = 5, const float OPLtemporalresponse2 = 1, const float LPfilterSpatialResponse = 5, const float LPfilterGain = 0, const float LPfilterTemporalresponse = 0, const float MovingContoursExtractorCoefficient = 5, const bool normalizeParvoOutput_0_maxOutputValue = false, const bool normalizeMagnoOutput_0_maxOutputValue = false, const float maxOutputValue = 255.0, const float maxInputValue = 255.0, const float meanValue = 128.0);
+
+    inline void setPhotoreceptorsLocalAdaptationSensitivity(const float V0CompressionParameter)
+    {
+        _photoreceptorsPrefilter.setV0CompressionParameter(1 - V0CompressionParameter);
+        _setInitPeriodCount();
+    };
+
+    inline void setParvoGanglionCellsLocalAdaptationSensitivity(const float V0CompressionParameter)
+    {
+        _ParvoRetinaFilter.setV0CompressionParameter(V0CompressionParameter);
+        _setInitPeriodCount();
+    };
+
+    inline void setGanglionCellsLocalAdaptationLPfilterParameters(const float spatialResponse, const float temporalResponse)
+    {
+        _ParvoRetinaFilter.setGanglionCellsLocalAdaptationLPfilterParameters(temporalResponse, spatialResponse);
+        _setInitPeriodCount();
+    };
+
+    inline void setMagnoGanglionCellsLocalAdaptationSensitivity(const float V0CompressionParameter)
+    {
+        _MagnoRetinaFilter.setV0CompressionParameter(V0CompressionParameter);
+        _setInitPeriodCount();
+    };
+
+    void setOPLandParvoParameters(const float beta1, const float tau1, const float k1, const float beta2, const float tau2, const float k2, const float V0CompressionParameter)
+    {
+        _ParvoRetinaFilter.setOPLandParvoFiltersParameters(beta1, tau1, k1, beta2, tau2, k2);
+        _ParvoRetinaFilter.setV0CompressionParameter(V0CompressionParameter);
+        _setInitPeriodCount();
+    };
+
+    void setMagnoCoefficientsTable(const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float V0CompressionParameter, const float localAdaptintegration_tau, const float localAdaptintegration_k)
+    {
+        _MagnoRetinaFilter.setCoefficientsTable(parasolCells_beta, parasolCells_tau, parasolCells_k, amacrinCellsTemporalCutFrequency, localAdaptintegration_tau, localAdaptintegration_k);
+        _MagnoRetinaFilter.setV0CompressionParameter(V0CompressionParameter);
+        _setInitPeriodCount();
+    };
+
+    inline void activateNormalizeParvoOutput_0_maxOutputValue(const bool normalizeParvoOutput_0_maxOutputValue)
+    {
+        _normalizeParvoOutput_0_maxOutputValue = normalizeParvoOutput_0_maxOutputValue;
+    };
+
+    inline void activateNormalizeMagnoOutput_0_maxOutputValue(const bool normalizeMagnoOutput_0_maxOutputValue)
+    {
+        _normalizeMagnoOutput_0_maxOutputValue = normalizeMagnoOutput_0_maxOutputValue;
+    };
+
+    inline void setMaxOutputValue(const float maxOutputValue)
+    {
+        _maxOutputValue = maxOutputValue;
+    };
+
+    void setColorMode(const bool desiredColorMode)
+    {
+        _useColorMode = desiredColorMode;
+    };
+    inline void setColorSaturation(const bool saturateColors = true, const float colorSaturationValue = 4.0)
+    {
+        _colorEngine.setColorSaturation(saturateColors, colorSaturationValue);
+    };
+    inline const cv::ocl::oclMat &getLocalAdaptation() const
+    {
+        return _photoreceptorsPrefilter.getOutput();
+    };
+    inline const cv::ocl::oclMat &getPhotoreceptors() const
+    {
+        return _ParvoRetinaFilter.getPhotoreceptorsLPfilteringOutput();
+    };
+
+    inline const cv::ocl::oclMat &getHorizontalCells() const
+    {
+        return _ParvoRetinaFilter.getHorizontalCellsOutput();
+    };
+    inline bool areContoursProcessed()
+    {
+        return _useParvoOutput;
+    };
+    bool getParvoFoveaResponse(cv::ocl::oclMat &parvoFovealResponse);
+    inline void activateContoursProcessing(const bool useParvoOutput)
+    {
+        _useParvoOutput = useParvoOutput;
+    };
+
+    const cv::ocl::oclMat &getContours();
+
+    inline const cv::ocl::oclMat &getContoursON() const
+    {
+        return _ParvoRetinaFilter.getParvoON();
+    };
+
+    inline const cv::ocl::oclMat &getContoursOFF() const
+    {
+        return _ParvoRetinaFilter.getParvoOFF();
+    };
+
+    inline bool areMovingContoursProcessed()
+    {
+        return _useMagnoOutput;
+    };
+
+    inline void activateMovingContoursProcessing(const bool useMagnoOutput)
+    {
+        _useMagnoOutput = useMagnoOutput;
+    };
+
+    inline const cv::ocl::oclMat &getMovingContours() const
+    {
+        return _MagnoRetinaFilter.getOutput();
+    };
+
+    inline const cv::ocl::oclMat &getMovingContoursSaturated() const
+    {
+        return _MagnoRetinaFilter.getMagnoYsaturated();
+    };
+
+    inline const cv::ocl::oclMat &getMovingContoursON() const
+    {
+        return _MagnoRetinaFilter.getMagnoON();
+    };
+
+    inline const cv::ocl::oclMat &getMovingContoursOFF() const
+    {
+        return _MagnoRetinaFilter.getMagnoOFF();
+    };
+
+    inline const cv::ocl::oclMat &getRetinaParvoMagnoMappedOutput() const
+    {
+        return _retinaParvoMagnoMappedFrame;
+    };
+
+    inline const cv::ocl::oclMat &getParvoContoursChannel() const
+    {
+        return _colorEngine.getLuminance();
+    };
+
+    inline const cv::ocl::oclMat &getParvoChrominance() const
+    {
+        return _colorEngine.getChrominance();
+    };
+    inline const cv::ocl::oclMat &getColorOutput() const
+    {
+        return _colorEngine.getDemultiplexedColorFrame();
+    };
+
+    inline bool isColorMode()
+    {
+        return _useColorMode;
+    };
+    bool getColorMode()
+    {
+        return _useColorMode;
+    };
+
+    inline bool isInitTransitionDone()
+    {
+        if (_ellapsedFramesSinceLastReset < _globalTemporalConstant)
+        {
+            return false;
+        }
+        return true;
+    };
+    inline float getRetinaSamplingBackProjection(const float projectedRadiusLength)
+    {
+        return projectedRadiusLength;
+    };
+
+    inline unsigned int getInputNBrows()
+    {
+        return _photoreceptorsPrefilter.getNBrows();
+    };
+
+    inline unsigned int getInputNBcolumns()
+    {
+        return _photoreceptorsPrefilter.getNBcolumns();
+    };
+
+    inline unsigned int getInputNBpixels()
+    {
+        return _photoreceptorsPrefilter.getNBpixels();
+    };
+
+    inline unsigned int getOutputNBrows()
+    {
+        return _photoreceptorsPrefilter.getNBrows();
+    };
+
+    inline unsigned int getOutputNBcolumns()
+    {
+        return _photoreceptorsPrefilter.getNBcolumns();
+    };
+
+    inline unsigned int getOutputNBpixels()
+    {
+        return _photoreceptorsPrefilter.getNBpixels();
+    };
+private:
+    bool _useParvoOutput;
+    bool _useMagnoOutput;
+
+    unsigned int _ellapsedFramesSinceLastReset;
+    unsigned int _globalTemporalConstant;
+
+    cv::ocl::oclMat _retinaParvoMagnoMappedFrame;
+    BasicRetinaFilter _photoreceptorsPrefilter;
+    ParvoRetinaFilter _ParvoRetinaFilter;
+    MagnoRetinaFilter _MagnoRetinaFilter;
+    RetinaColor       _colorEngine;
+
+    bool _useMinimalMemoryForToneMappingONLY;
+    bool _normalizeParvoOutput_0_maxOutputValue;
+    bool _normalizeMagnoOutput_0_maxOutputValue;
+    float _maxOutputValue;
+    bool _useColorMode;
+
+    void _setInitPeriodCount();
+    void _processRetinaParvoMagnoMapping();
+    void _runGrayToneMapping(const cv::ocl::oclMat &grayImageInput, cv::ocl::oclMat &grayImageOutput , const float PhotoreceptorsCompression = 0.6, const float ganglionCellsCompression = 0.6);
+};
+
+}  /* namespace ocl */
+}  /* namespace bioinspired */
+}  /* namespace cv */
+
+#endif  /* HAVE_OPENCV_OCL */
+#endif  /* __OCL_RETINA_HPP__ */
diff --git a/modules/contrib/src/retinacolor.cpp b/modules/bioinspired/src/retinacolor.cpp
similarity index 97%
rename from modules/contrib/src/retinacolor.cpp
rename to modules/bioinspired/src/retinacolor.cpp
index d138582ce4..3fbc553852 100644
--- a/modules/contrib/src/retinacolor.cpp
+++ b/modules/bioinspired/src/retinacolor.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -73,14 +73,15 @@
 
 namespace cv
 {
-
+namespace bioinspired
+{
 // init static values
 static float _LMStoACr1Cr2[]={1.0,  1.0, 0.0,  1.0, -1.0, 0.0,  -0.5, -0.5, 1.0};
 //static double _ACr1Cr2toLMS[]={0.5,  0.5, 0.0,   0.5, -0.5, 0.0,  0.5,  0.0, 1.0};
 static float _LMStoLab[]={0.5774f, 0.5774f, 0.5774f, 0.4082f, 0.4082f, -0.8165f, 0.7071f, -0.7071f, 0.f};
 
 // constructor/desctructor
-RetinaColor::RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const RETINA_COLORSAMPLINGMETHOD samplingMethod)
+RetinaColor::RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const int samplingMethod)
 :BasicRetinaFilter(NBrows, NBcolumns, 3),
  _colorSampling(NBrows*NBcolumns),
  _RGBmosaic(NBrows*NBcolumns*3),
@@ -338,8 +339,11 @@ void RetinaColor::runColorDemultiplexing(const std::valarray<float> &multiplexed
         }
 
         // compute the gradient of the luminance
+#ifdef MAKE_PARALLEL // call the TemplateBuffer TBB clipping method
+        cv::parallel_for_(cv::Range(2,_filterOutput.getNBrows()-2), Parallel_computeGradient(_filterOutput.getNBcolumns(), _filterOutput.getNBrows(), &(*_luminance)[0], &_imageGradient[0]));
+#else
         _computeGradient(&(*_luminance)[0]);
-
+#endif
         // adaptively filter the submosaics to get the adaptive densities, here the buffer _chrominance is used as a temp buffer
         _adaptiveSpatialLPfilter(&_RGBmosaic[0], &_chrominance[0]);
         _adaptiveSpatialLPfilter(&_RGBmosaic[0]+_filterOutput.getNBpixels(), &_chrominance[0]+_filterOutput.getNBpixels());
@@ -717,4 +721,5 @@ void RetinaColor::_applyImageColorSpaceConversion(const std::valarray<float> &in
     }
 }
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/retinacolor.hpp b/modules/bioinspired/src/retinacolor.hpp
similarity index 86%
rename from modules/contrib/src/retinacolor.hpp
rename to modules/bioinspired/src/retinacolor.hpp
index 056d006e4b..3fb6be8f9e 100644
--- a/modules/contrib/src/retinacolor.hpp
+++ b/modules/bioinspired/src/retinacolor.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -85,7 +85,8 @@
 
 namespace cv
 {
-
+namespace bioinspired
+{
     class RetinaColor: public BasicRetinaFilter
     {
     public:
@@ -99,7 +100,7 @@ namespace cv
         * @param NBcolumns: number of columns of the input image
         * @param samplingMethod: the chosen color sampling method
         */
-        RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const RETINA_COLORSAMPLINGMETHOD samplingMethod=RETINA_COLOR_DIAGONAL);
+        RetinaColor(const unsigned int NBrows, const unsigned int NBcolumns, const int samplingMethod=RETINA_COLOR_BAYER);
 
         /**
         * standard destructor
@@ -219,7 +220,7 @@ namespace cv
     protected:
 
         // private functions
-        RETINA_COLORSAMPLINGMETHOD _samplingMethod;
+        int _samplingMethod;
         bool _saturateColors;
         float _colorSaturationValue;
         // links to parent buffers (more convienient names
@@ -333,10 +334,56 @@ namespace cv
                 }
             }
         };
+
+        class Parallel_computeGradient: public cv::ParallelLoopBody
+        {
+        private:
+            float *imageGradient;
+            const float *luminance;
+            unsigned int nbColumns, doubleNbColumns, nbRows, nbPixels;
+        public:
+            Parallel_computeGradient(const unsigned int nbCols, const unsigned int nbRws, const float *lum, float *imageGrad)
+            :imageGradient(imageGrad), luminance(lum), nbColumns(nbCols), doubleNbColumns(2*nbCols), nbRows(nbRws), nbPixels(nbRws*nbCols){};
+
+            virtual void operator()( const Range& r ) const {
+                for (int idLine=r.start;idLine!=r.end;++idLine)
+                {
+                    for (unsigned int idColumn=2;idColumn<nbColumns-2;++idColumn)
+                    {
+                        const unsigned int pixelIndex=idColumn+nbColumns*idLine;
+
+                        // horizontal and vertical local gradients
+                        const float verticalGrad=fabs(luminance[pixelIndex+nbColumns]-luminance[pixelIndex-nbColumns]);
+                        const float horizontalGrad=fabs(luminance[pixelIndex+1]-luminance[pixelIndex-1]);
+
+                        // neighborhood horizontal and vertical gradients
+                        const float verticalGrad_p=fabs(luminance[pixelIndex]-luminance[pixelIndex-doubleNbColumns]);
+                        const float horizontalGrad_p=fabs(luminance[pixelIndex]-luminance[pixelIndex-2]);
+                        const float verticalGrad_n=fabs(luminance[pixelIndex+doubleNbColumns]-luminance[pixelIndex]);
+                        const float horizontalGrad_n=fabs(luminance[pixelIndex+2]-luminance[pixelIndex]);
+
+                        const float horizontalGradient=0.5f*horizontalGrad+0.25f*(horizontalGrad_p+horizontalGrad_n);
+                        const float verticalGradient=0.5f*verticalGrad+0.25f*(verticalGrad_p+verticalGrad_n);
+
+                        // compare local gradient means and fill the appropriate filtering coefficient value that will be used in adaptative filters
+                        if (horizontalGradient<verticalGradient)
+                        {
+                            imageGradient[pixelIndex+nbPixels]=0.06f;
+                            imageGradient[pixelIndex]=0.57f;
+                        }
+                        else
+                        {
+                            imageGradient[pixelIndex+nbPixels]=0.57f;
+                            imageGradient[pixelIndex]=0.06f;
+                        }
+                    }
+                }
+            }
+        };
+
 #endif
     };
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 
 #endif /*RETINACOLOR_HPP_*/
-
-
diff --git a/modules/bioinspired/src/retinafasttonemapping.cpp b/modules/bioinspired/src/retinafasttonemapping.cpp
new file mode 100644
index 0000000000..2713d7449c
--- /dev/null
+++ b/modules/bioinspired/src/retinafasttonemapping.cpp
@@ -0,0 +1,316 @@
+
+/*#******************************************************************************
+ ** IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+ **
+ ** By downloading, copying, installing or using the software you agree to this license.
+ ** If you do not agree to this license, do not download, install,
+ ** copy or use the software.
+ **
+ **
+ ** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+ **
+ ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
+ **
+ **  Creation - enhancement process 2007-2013
+ **      Author: Alexandre Benoit (benoit.alexandre.vision@gmail.com), LISTIC lab, Annecy le vieux, France
+ **
+ ** Theses algorithm have been developped by Alexandre BENOIT since his thesis with Alice Caplier at Gipsa-Lab (www.gipsa-lab.inpg.fr) and the research he pursues at LISTIC Lab (www.listic.univ-savoie.fr).
+ ** Refer to the following research paper for more information:
+ ** Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+ ** This work have been carried out thanks to Jeanny Herault who's research and great discussions are the basis of all this work, please take a look at his book:
+ ** Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+ **
+ **
+ ** This class is based on image processing tools of the author and already used within the Retina class (this is the same code as method retina::applyFastToneMapping, but in an independent class, it is ligth from a memory requirement point of view). It implements an adaptation of the efficient tone mapping algorithm propose by David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+ ** -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+ **
+ **
+ **                          License Agreement
+ **               For Open Source Computer Vision Library
+ **
+ ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+ ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+ **
+ **               For Human Visual System tools (bioinspired)
+ ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
+ **
+ ** Third party copyrights are property of their respective owners.
+ **
+ ** Redistribution and use in source and binary forms, with or without modification,
+ ** are permitted provided that the following conditions are met:
+ **
+ ** * Redistributions of source code must retain the above copyright notice,
+ **    this list of conditions and the following disclaimer.
+ **
+ ** * Redistributions in binary form must reproduce the above copyright notice,
+ **    this list of conditions and the following disclaimer in the documentation
+ **    and/or other materials provided with the distribution.
+ **
+ ** * The name of the copyright holders may not be used to endorse or promote products
+ **    derived from this software without specific prior written permission.
+ **
+ ** This software is provided by the copyright holders and contributors "as is" and
+ ** any express or implied warranties, including, but not limited to, the implied
+ ** warranties of merchantability and fitness for a particular purpose are disclaimed.
+ ** In no event shall the Intel Corporation or contributors be liable for any direct,
+ ** indirect, incidental, special, exemplary, or consequential damages
+ ** (including, but not limited to, procurement of substitute goods or services;
+ ** loss of use, data, or profits; or business interruption) however caused
+ ** and on any theory of liability, whether in contract, strict liability,
+ ** or tort (including negligence or otherwise) arising in any way out of
+ ** the use of this software, even if advised of the possibility of such damage.
+ *******************************************************************************/
+
+/*
+ * retinafasttonemapping.cpp
+ *
+ *  Created on: May 26, 2013
+ *      Author: Alexandre Benoit
+ */
+
+#include "precomp.hpp"
+#include "basicretinafilter.hpp"
+#include "retinacolor.hpp"
+#include <cstdio>
+#include <sstream>
+#include <valarray>
+
+namespace cv
+{
+namespace bioinspired
+{
+/**
+ * @class RetinaFastToneMappingImpl a wrapper class which allows the tone mapping algorithm of Meylan&al(2007) to be used with OpenCV.
+ * This algorithm is already implemented in thre Retina class (retina::applyFastToneMapping) but used it does not require all the retina model to be allocated. This allows a light memory use for low memory devices (smartphones, etc.
+ * As a summary, these are the model properties:
+ * => 2 stages of local luminance adaptation with a different local neighborhood for each.
+ * => first stage models the retina photorecetors local luminance adaptation
+ * => second stage models th ganglion cells local information adaptation
+ * => compared to the initial publication, this class uses spatio-temporal low pass filters instead of spatial only filters.
+ * ====> this can help noise robustness and temporal stability for video sequence use cases.
+ * for more information, read to the following papers :
+ *  Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
+ * regarding spatio-temporal filter and the bigger retina model :
+ * Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
+ */
+
+class RetinaFastToneMappingImpl : public RetinaFastToneMapping
+{
+public:
+    /**
+     * constructor
+     * @param imageInput: the size of the images to process
+     */
+    RetinaFastToneMappingImpl(Size imageInput)
+    {
+        unsigned int nbPixels=imageInput.height*imageInput.width;
+
+        // basic error check
+        if (nbPixels <= 0)
+        throw cv::Exception(-1, "Bad retina size setup : size height and with must be superior to zero", "RetinaImpl::setup", "retinafasttonemapping.cpp", 0);
+
+        // resize buffers
+        _inputBuffer.resize(nbPixels*3); // buffer supports gray images but also 3 channels color buffers... (larger is better...)
+        _imageOutput.resize(nbPixels*3);
+        _temp2.resize(nbPixels);
+        // allocate the main filter with 2 setup sets properties (one for each low pass filter
+        _multiuseFilter = makePtr<BasicRetinaFilter>(imageInput.height, imageInput.width, 2);
+        // allocate the color manager (multiplexer/demultiplexer
+        _colorEngine = makePtr<RetinaColor>(imageInput.height, imageInput.width);
+        // setup filter behaviors with default values
+        setup();
+    }
+
+    /**
+     * basic destructor
+     */
+    virtual ~RetinaFastToneMappingImpl(){};
+
+    /**
+     * method that applies a luminance correction (initially High Dynamic Range (HDR) tone mapping) using only the 2 local adaptation stages of the retina parvocellular channel : photoreceptors level and ganlion cells level. Spatio temporal filtering is applied but limited to temporal smoothing and eventually high frequencies attenuation. This is a lighter method than the one available using the regular retina::run method. It is then faster but it does not include complete temporal filtering nor retina spectral whitening. Then, it can have a more limited effect on images with a very high dynamic range. This is an adptation of the original still image HDR tone mapping algorithm of David Alleyson, Sabine Susstruck and Laurence Meylan's work, please cite:
+    * -> Meylan L., Alleysson D., and Susstrunk S., A Model of Retinal Local Adaptation for the Tone Mapping of Color Filter Array Images, Journal of Optical Society of America, A, Vol. 24, N 9, September, 1st, 2007, pp. 2807-2816
+     @param inputImage the input image to process RGB or gray levels
+     @param outputToneMappedImage the output tone mapped image
+     */
+    virtual void applyFastToneMapping(InputArray inputImage, OutputArray outputToneMappedImage)
+    {
+        // first convert input image to the compatible format :
+        const bool colorMode = _convertCvMat2ValarrayBuffer(inputImage.getMat(), _inputBuffer);
+
+        // process tone mapping
+        if (colorMode)
+        {
+            _runRGBToneMapping(_inputBuffer, _imageOutput, true);
+            _convertValarrayBuffer2cvMat(_imageOutput, _multiuseFilter->getNBrows(), _multiuseFilter->getNBcolumns(), true, outputToneMappedImage);
+        }else
+        {
+            _runGrayToneMapping(_inputBuffer, _imageOutput);
+            _convertValarrayBuffer2cvMat(_imageOutput, _multiuseFilter->getNBrows(), _multiuseFilter->getNBcolumns(), false, outputToneMappedImage);
+        }
+
+    }
+
+    /**
+     * setup method that updates tone mapping behaviors by adjusing the local luminance computation area
+     * @param photoreceptorsNeighborhoodRadius the first stage local adaptation area
+     * @param ganglioncellsNeighborhoodRadius the second stage local adaptation area
+     * @param meanLuminanceModulatorK the factor applied to modulate the meanLuminance information (default is 1, see reference paper)
+     */
+    virtual void setup(const float photoreceptorsNeighborhoodRadius=3.f, const float ganglioncellsNeighborhoodRadius=1.f, const float meanLuminanceModulatorK=1.f)
+    {
+        // setup the spatio-temporal properties of each filter
+        _meanLuminanceModulatorK = meanLuminanceModulatorK;
+        _multiuseFilter->setV0CompressionParameter(1.f, 255.f, 128.f);
+        _multiuseFilter->setLPfilterParameters(0.f, 0.f, photoreceptorsNeighborhoodRadius, 1);
+        _multiuseFilter->setLPfilterParameters(0.f, 0.f, ganglioncellsNeighborhoodRadius, 2);
+    }
+
+private:
+    // a filter able to perform local adaptation and low pass spatio-temporal filtering
+    cv::Ptr <BasicRetinaFilter> _multiuseFilter;
+    cv::Ptr <RetinaColor> _colorEngine;
+
+    //!< buffer used to convert input cv::Mat to internal retina buffers format (valarrays)
+    std::valarray<float> _inputBuffer;
+    std::valarray<float> _imageOutput;
+    std::valarray<float> _temp2;
+    float _meanLuminanceModulatorK;
+
+
+void _convertValarrayBuffer2cvMat(const std::valarray<float> &grayMatrixToConvert, const unsigned int nbRows, const unsigned int nbColumns, const bool colorMode, OutputArray outBuffer)
+{
+    // fill output buffer with the valarray buffer
+    const float *valarrayPTR=get_data(grayMatrixToConvert);
+    if (!colorMode)
+    {
+        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8U);
+        Mat outMat = outBuffer.getMat();
+        for (unsigned int i=0;i<nbRows;++i)
+        {
+            for (unsigned int j=0;j<nbColumns;++j)
+            {
+                cv::Point2d pixel(j,i);
+                outMat.at<unsigned char>(pixel)=(unsigned char)*(valarrayPTR++);
+            }
+        }
+    }else
+    {
+        const unsigned int nbPixels=nbColumns*nbRows;
+        const unsigned int doubleNBpixels=nbColumns*nbRows*2;
+        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8UC3);
+        Mat outMat = outBuffer.getMat();
+        for (unsigned int i=0;i<nbRows;++i)
+        {
+            for (unsigned int j=0;j<nbColumns;++j,++valarrayPTR)
+            {
+                cv::Point2d pixel(j,i);
+                cv::Vec3b pixelValues;
+                pixelValues[2]=(unsigned char)*(valarrayPTR);
+                pixelValues[1]=(unsigned char)*(valarrayPTR+nbPixels);
+                pixelValues[0]=(unsigned char)*(valarrayPTR+doubleNBpixels);
+
+                outMat.at<cv::Vec3b>(pixel)=pixelValues;
+            }
+        }
+    }
+}
+
+bool _convertCvMat2ValarrayBuffer(InputArray inputMat, std::valarray<float> &outputValarrayMatrix)
+{
+    const Mat inputMatToConvert=inputMat.getMat();
+    // first check input consistency
+    if (inputMatToConvert.empty())
+        throw cv::Exception(-1, "RetinaImpl cannot be applied, input buffer is empty", "RetinaImpl::run", "RetinaImpl.h", 0);
+
+    // retreive color mode from image input
+    int imageNumberOfChannels = inputMatToConvert.channels();
+
+        // convert to float AND fill the valarray buffer
+    typedef float T; // define here the target pixel format, here, float
+    const int dsttype = DataType<T>::depth; // output buffer is float format
+
+    const unsigned int nbPixels=inputMat.getMat().rows*inputMat.getMat().cols;
+    const unsigned int doubleNBpixels=inputMat.getMat().rows*inputMat.getMat().cols*2;
+
+    if(imageNumberOfChannels==4)
+    {
+    // create a cv::Mat table (for RGBA planes)
+        cv::Mat planes[4] =
+        {
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[doubleNBpixels]),
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[nbPixels]),
+            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
+        };
+        planes[3] = cv::Mat(inputMatToConvert.size(), dsttype);     // last channel (alpha) does not point on the valarray (not usefull in our case)
+        // split color cv::Mat in 4 planes... it fills valarray directely
+        cv::split(Mat_<Vec<T, 4> >(inputMatToConvert), planes);
+    }
+    else if (imageNumberOfChannels==3)
+    {
+        // create a cv::Mat table (for RGB planes)
+        cv::Mat planes[] =
+        {
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[doubleNBpixels]),
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[nbPixels]),
+        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
+        };
+        // split color cv::Mat in 3 planes... it fills valarray directely
+        cv::split(cv::Mat_<Vec<T, 3> >(inputMatToConvert), planes);
+    }
+    else if(imageNumberOfChannels==1)
+    {
+        // create a cv::Mat header for the valarray
+        cv::Mat dst(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0]);
+        inputMatToConvert.convertTo(dst, dsttype);
+    }
+        else
+            CV_Error(Error::StsUnsupportedFormat, "input image must be single channel (gray levels), bgr format (color) or bgra (color with transparency which won't be considered");
+
+    return imageNumberOfChannels>1; // return bool : false for gray level image processing, true for color mode
+}
+
+
+    // run the initilized retina filter in order to perform gray image tone mapping, after this call all retina outputs are updated
+    void _runGrayToneMapping(const std::valarray<float> &grayImageInput, std::valarray<float> &grayImageOutput)
+    {
+         // apply tone mapping on the multiplexed image
+        // -> photoreceptors local adaptation (large area adaptation)
+        _multiuseFilter->runFilter_LPfilter(grayImageInput, grayImageOutput, 0); // compute low pass filtering modeling the horizontal cells filtering to acess local luminance
+        _multiuseFilter->setV0CompressionParameterToneMapping(1.f, grayImageOutput.max(), _meanLuminanceModulatorK*grayImageOutput.sum()/(float)_multiuseFilter->getNBpixels());
+        _multiuseFilter->runFilter_LocalAdapdation(grayImageInput, grayImageOutput, _temp2); // adapt contrast to local luminance
+
+        // -> ganglion cells local adaptation (short area adaptation)
+        _multiuseFilter->runFilter_LPfilter(_temp2, grayImageOutput, 1); // compute low pass filtering (high cut frequency (remove spatio-temporal noise)
+        _multiuseFilter->setV0CompressionParameterToneMapping(1.f, _temp2.max(), _meanLuminanceModulatorK*grayImageOutput.sum()/(float)_multiuseFilter->getNBpixels());
+        _multiuseFilter->runFilter_LocalAdapdation(_temp2, grayImageOutput, grayImageOutput); // adapt contrast to local luminance
+
+    }
+
+ // run the initilized retina filter in order to perform color tone mapping, after this call all retina outputs are updated
+    void _runRGBToneMapping(const std::valarray<float> &RGBimageInput, std::valarray<float> &RGBimageOutput, const bool useAdaptiveFiltering)
+    {
+        // multiplex the image with the color sampling method specified in the constructor
+        _colorEngine->runColorMultiplexing(RGBimageInput);
+
+        // apply tone mapping on the multiplexed image
+        _runGrayToneMapping(_colorEngine->getMultiplexedFrame(), RGBimageOutput);
+
+        // demultiplex tone maped image
+        _colorEngine->runColorDemultiplexing(RGBimageOutput, useAdaptiveFiltering, _multiuseFilter->getMaxInputValue());//_ColorEngine->getMultiplexedFrame());//_ParvoRetinaFilter->getPhotoreceptorsLPfilteringOutput());
+
+        // rescaling result between 0 and 255
+        _colorEngine->normalizeRGBOutput_0_maxOutputValue(255.0);
+
+        // return the result
+        RGBimageOutput=_colorEngine->getDemultiplexedColorFrame();
+    }
+
+};
+
+CV_EXPORTS Ptr<RetinaFastToneMapping> createRetinaFastToneMapping(Size inputSize)
+{
+    return makePtr<RetinaFastToneMappingImpl>(inputSize);
+}
+
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/retinafilter.cpp b/modules/bioinspired/src/retinafilter.cpp
similarity index 96%
rename from modules/contrib/src/retinafilter.cpp
rename to modules/bioinspired/src/retinafilter.cpp
index 4cf6019f57..e1e24c89bf 100644
--- a/modules/contrib/src/retinafilter.cpp
+++ b/modules/bioinspired/src/retinafilter.cpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -72,9 +72,11 @@
 #include <cmath>
 
 namespace cv
+{
+namespace bioinspired
 {
     // standard constructor without any log sampling of the input frame
-    RetinaFilter::RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode, const RETINA_COLORSAMPLINGMETHOD samplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
+    RetinaFilter::RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode, const int samplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
         :
     _retinaParvoMagnoMappedFrame(0),
         _retinaParvoMagnoMapCoefTable(0),
@@ -375,21 +377,15 @@ namespace cv
         // apply tone mapping on the multiplexed image
         // -> photoreceptors local adaptation (large area adaptation)
         _photoreceptorsPrefilter.runFilter_LPfilter(grayImageInput, grayImageOutput, 2); // compute low pass filtering modeling the horizontal cells filtering to acess local luminance
-        _photoreceptorsPrefilter.setV0CompressionParameterToneMapping(PhotoreceptorsCompression, grayImageOutput.sum()/(float)_photoreceptorsPrefilter.getNBpixels());
+        _photoreceptorsPrefilter.setV0CompressionParameterToneMapping(1.f-PhotoreceptorsCompression, grayImageOutput.max(), 1.f*grayImageOutput.sum()/(float)_photoreceptorsPrefilter.getNBpixels());
         _photoreceptorsPrefilter.runFilter_LocalAdapdation(grayImageInput, grayImageOutput, temp2); // adapt contrast to local luminance
 
-        // high pass filter
-        //_spatiotemporalLPfilter(_localBuffer, _filterOutput, 2); // compute low pass filtering (high cut frequency (remove spatio-temporal noise)
-
-        //for (unsigned int i=0;i<_NBpixels;++i)
-        //  _localBuffer[i]-= _filterOutput[i]/2.0;
-
         // -> ganglion cells local adaptation (short area adaptation)
         _photoreceptorsPrefilter.runFilter_LPfilter(temp2, grayImageOutput, 1); // compute low pass filtering (high cut frequency (remove spatio-temporal noise)
-        _photoreceptorsPrefilter.setV0CompressionParameterToneMapping(ganglionCellsCompression, temp2.max(), temp2.sum()/(float)_photoreceptorsPrefilter.getNBpixels());
+        _photoreceptorsPrefilter.setV0CompressionParameterToneMapping(1.f-ganglionCellsCompression, temp2.max(), 1.f*temp2.sum()/(float)_photoreceptorsPrefilter.getNBpixels());
         _photoreceptorsPrefilter.runFilter_LocalAdapdation(temp2, grayImageOutput, grayImageOutput); // adapt contrast to local luminance
-
     }
+
     // run the initilized retina filter in order to perform color tone mapping, after this call all retina outputs are updated
     void RetinaFilter::runRGBToneMapping(const std::valarray<float> &RGBimageInput, std::valarray<float> &RGBimageOutput, const bool useAdaptiveFiltering, const float PhotoreceptorsCompression, const float ganglionCellsCompression)
     {
@@ -526,4 +522,5 @@ namespace cv
 
         return true;
     }
-}
+}// end of namespace bioinspired
+}// end of namespace cv
diff --git a/modules/contrib/src/retinafilter.hpp b/modules/bioinspired/src/retinafilter.hpp
similarity index 98%
rename from modules/contrib/src/retinafilter.hpp
rename to modules/bioinspired/src/retinafilter.hpp
index 7fa2a078c8..3e204885fc 100644
--- a/modules/contrib/src/retinafilter.hpp
+++ b/modules/bioinspired/src/retinafilter.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -110,7 +110,8 @@
 //#define __RETINADEBUG // define RETINADEBUG to display debug data
 namespace cv
 {
-
+namespace bioinspired
+{
 // retina class that process the 3 outputs of the retina filtering stages
 class RetinaFilter//: public BasicRetinaFilter
 {
@@ -126,7 +127,7 @@ public:
     * @param reductionFactor: only usefull if param useRetinaLogSampling=true, specifies the reduction factor of the output frame (as the center (fovea) is high resolution and corners can be underscaled, then a reduction of the output is allowed without precision leak
     * @param samplingStrenght: only usefull if param useRetinaLogSampling=true, specifies the strenght of the log scale that is applied
     */
-    RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode=false, const RETINA_COLORSAMPLINGMETHOD samplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
+    RetinaFilter(const unsigned int sizeRows, const unsigned int sizeColumns, const bool colorMode=false, const int samplingMethod=RETINA_COLOR_BAYER, const bool useRetinaLogSampling=false, const double reductionFactor=1.0, const double samplingStrenght=10.0);
 
     /**
     * standard destructor
@@ -541,9 +542,7 @@ private:
 
 };
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
+
 #endif /*RETINACLASSES_H_*/
-
-
-
-
diff --git a/modules/contrib/src/templatebuffer.hpp b/modules/bioinspired/src/templatebuffer.hpp
similarity index 98%
rename from modules/contrib/src/templatebuffer.hpp
rename to modules/bioinspired/src/templatebuffer.hpp
index 21414b4da7..827eb709fa 100644
--- a/modules/contrib/src/templatebuffer.hpp
+++ b/modules/bioinspired/src/templatebuffer.hpp
@@ -6,7 +6,7 @@
 ** copy or use the software.
 **
 **
-** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
+** bioinspired : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
 ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
 **
 ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
@@ -32,7 +32,7 @@
 ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
 **
-**               For Human Visual System tools (hvstools)
+**               For Human Visual System tools (bioinspired)
 ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
 **
 ** Third party copyrights are property of their respective owners.
@@ -71,6 +71,12 @@
 #include <cmath>
 
 
+//#define __TEMPLATEBUFFERDEBUG //define TEMPLATEBUFFERDEBUG in order to display debug information
+
+namespace cv
+{
+namespace bioinspired
+{
 //// If a parallelization method is available then, you should define MAKE_PARALLEL, in the other case, the classical serial code will be used
 #define MAKE_PARALLEL
 // ==> then include required includes
@@ -101,10 +107,6 @@ public:
 };
 #endif
 
-//#define __TEMPLATEBUFFERDEBUG //define TEMPLATEBUFFERDEBUG in order to display debug information
-
-namespace cv
-{
     /**
     * @class TemplateBuffer
     * @brief this class is a simple template memory buffer which contains basic functions to get information on or normalize the buffer content
@@ -548,8 +550,6 @@ namespace cv
         return std::fabs(x);
     }
 
-}
+}// end of namespace bioinspired
+}// end of namespace cv
 #endif
-
-
-
diff --git a/modules/features2d/test/test_precomp.cpp b/modules/bioinspired/test/test_main.cpp
similarity index 58%
rename from modules/features2d/test/test_precomp.cpp
rename to modules/bioinspired/test/test_main.cpp
index 5956e13e3e..6b24993447 100644
--- a/modules/features2d/test/test_precomp.cpp
+++ b/modules/bioinspired/test/test_main.cpp
@@ -1 +1,3 @@
 #include "test_precomp.hpp"
+
+CV_TEST_MAIN("cv")
diff --git a/modules/bioinspired/test/test_precomp.hpp b/modules/bioinspired/test/test_precomp.hpp
new file mode 100644
index 0000000000..b1672149ab
--- /dev/null
+++ b/modules/bioinspired/test/test_precomp.hpp
@@ -0,0 +1,16 @@
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/bioinspired.hpp"
+#include <iostream>
+
+#endif
diff --git a/modules/bioinspired/test/test_retina_ocl.cpp b/modules/bioinspired/test/test_retina_ocl.cpp
new file mode 100644
index 0000000000..b09ce50366
--- /dev/null
+++ b/modules/bioinspired/test/test_retina_ocl.cpp
@@ -0,0 +1,144 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+#include "opencv2/opencv_modules.hpp"
+#include "opencv2/bioinspired.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+
+#if defined(HAVE_OPENCV_OCL)
+
+#include "opencv2/ocl.hpp"
+#define RETINA_ITERATIONS 5
+
+static double checkNear(const cv::Mat &m1, const cv::Mat &m2)
+{
+    return cv::norm(m1, m2, cv::NORM_INF);
+}
+
+#define PARAM_TEST_CASE(name, ...) struct name : testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
+#define GET_PARAM(k) std::tr1::get< k >(GetParam())
+
+static int oclInit = false;
+
+PARAM_TEST_CASE(Retina_OCL, bool, int, bool, double, double)
+{
+    bool colorMode;
+    int colorSamplingMethod;
+    bool useLogSampling;
+    double reductionFactor;
+    double samplingStrength;
+
+    std::vector<cv::ocl::Info> infos;
+
+    virtual void SetUp()
+    {
+        colorMode           = GET_PARAM(0);
+        colorSamplingMethod = GET_PARAM(1);
+        useLogSampling      = GET_PARAM(2);
+        reductionFactor     = GET_PARAM(3);
+        samplingStrength    = GET_PARAM(4);
+
+        if(!oclInit)
+        {
+            cv::ocl::getDevice(infos);
+            std::cout << "Device name:" << infos[0].DeviceName[0] << std::endl;
+            oclInit = true;
+        }
+    }
+};
+
+TEST_P(Retina_OCL, Accuracy)
+{
+    using namespace cv;
+    Mat input = imread(cvtest::TS::ptr()->get_data_path() + "shared/lena.png", colorMode);
+    CV_Assert(!input.empty());
+    ocl::oclMat ocl_input(input);
+
+    Ptr<bioinspired::Retina> ocl_retina = bioinspired::createRetina_OCL(
+        input.size(),
+        colorMode,
+        colorSamplingMethod,
+        useLogSampling,
+        reductionFactor,
+        samplingStrength);
+
+    Ptr<bioinspired::Retina> gold_retina = bioinspired::createRetina(
+        input.size(),
+        colorMode,
+        colorSamplingMethod,
+        useLogSampling,
+        reductionFactor,
+        samplingStrength);
+
+    Mat gold_parvo;
+    Mat gold_magno;
+    ocl::oclMat ocl_parvo;
+    ocl::oclMat ocl_magno;
+
+    for(int i = 0; i < RETINA_ITERATIONS; i ++)
+    {
+        ocl_retina->run(ocl_input);
+        gold_retina->run(input);
+
+        gold_retina->getParvo(gold_parvo);
+        gold_retina->getMagno(gold_magno);
+
+        ocl_retina->getParvo(ocl_parvo);
+        ocl_retina->getMagno(ocl_magno);
+
+        EXPECT_LE(checkNear(gold_parvo, (Mat)ocl_parvo), 1.0);
+        EXPECT_LE(checkNear(gold_magno, (Mat)ocl_magno), 1.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(Contrib, Retina_OCL, testing::Combine(
+                            testing::Values(false, true),
+                            testing::Values((int)cv::bioinspired::RETINA_COLOR_BAYER),
+                            testing::Values(false/*,true*/),
+                            testing::Values(1.0, 0.5),
+                            testing::Values(10.0, 5.0)));
+#endif
diff --git a/modules/calib3d/doc/calib3d.rst b/modules/calib3d/doc/calib3d.rst
index 0ac8d51960..0c056843d2 100644
--- a/modules/calib3d/doc/calib3d.rst
+++ b/modules/calib3d/doc/calib3d.rst
@@ -6,4 +6,3 @@ calib3d. Camera Calibration and 3D Reconstruction
     :maxdepth: 2
 
     camera_calibration_and_3d_reconstruction
-
diff --git a/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.rst b/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.rst
index 26c73f1b69..cb30dc36d5 100644
--- a/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.rst
+++ b/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.rst
@@ -81,7 +81,7 @@ is extended as:
 :math:`s_1`,
 :math:`s_2`,
 :math:`s_3`, and
-:math:`s_4`, are the thin prism distortion coefficients. 
+:math:`s_4`, are the thin prism distortion coefficients.
 Higher-order coefficients are not considered in OpenCV. In the functions below the coefficients are passed or returned as
 
 .. math::
@@ -109,20 +109,27 @@ The functions below use the above model to do the following:
 
  * Estimate the relative position and orientation of the stereo camera "heads" and compute the *rectification* transformation that makes the camera optical axes parallel.
 
+.. note::
 
+   * A calibration sample for 3 cameras in horizontal position can be found at opencv_source_code/samples/cpp/3calibration.cpp
+   * A calibration sample based on a sequence of images can be found at opencv_source_code/samples/cpp/calibration.cpp
+   * A calibration sample in order to do 3D reconstruction can be found at opencv_source_code/samples/cpp/build3dmodel.cpp
+   * A calibration sample of an artificially generated camera and chessboard patterns can be found at opencv_source_code/samples/cpp/calibration_artificial.cpp
+   * A calibration example on stereo calibration can be found at opencv_source_code/samples/cpp/stereo_calib.cpp
+   * A calibration example on stereo matching can be found at opencv_source_code/samples/cpp/stereo_match.cpp
+
+   * (Python) A camera calibration sample can be found at opencv_source_code/samples/python2/calibrate.py
 
 calibrateCamera
 ---------------
 Finds the camera intrinsic and extrinsic parameters from several views of a calibration pattern.
 
-.. ocv:function:: double calibrateCamera( InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, Size imageSize, InputOutputArray cameraMatrix, InputOutputArray distCoeffs, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs, int flags=0, TermCriteria criteria=TermCriteria( TermCriteria::COUNT+TermCriteria::EPS, 30, DBL_EPSILON) )
+.. ocv:function:: double calibrateCamera( InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, Size imageSize, InputOutputArray cameraMatrix, InputOutputArray distCoeffs, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs, int flags=0, TermCriteria criteria=TermCriteria( TermCriteria::COUNT + TermCriteria::EPS, 30, DBL_EPSILON) )
 
 .. ocv:pyfunction:: cv2.calibrateCamera(objectPoints, imagePoints, imageSize, cameraMatrix, distCoeffs[, rvecs[, tvecs[, flags[, criteria]]]]) -> retval, cameraMatrix, distCoeffs, rvecs, tvecs
 
 .. ocv:cfunction:: double cvCalibrateCamera2( const CvMat* object_points, const CvMat* image_points, const CvMat* point_counts, CvSize image_size, CvMat* camera_matrix, CvMat* distortion_coeffs, CvMat* rotation_vectors=NULL, CvMat* translation_vectors=NULL, int flags=0, CvTermCriteria term_crit=cvTermCriteria( CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,DBL_EPSILON) )
 
-.. ocv:pyoldfunction:: cv.CalibrateCamera2(objectPoints, imagePoints, pointCounts, imageSize, cameraMatrix, distCoeffs, rvecs, tvecs, flags=0)-> None
-
     :param objectPoints: In the new interface it is a vector of vectors of calibration pattern points in the calibration pattern coordinate space. The outer vector contains as many elements as the number of the pattern views. If the same calibration pattern is shown in each view and it is fully visible, all the vectors will be the same. Although, it is possible to use partially occluded patterns, or even different patterns in different views. Then, the vectors will be different. The points are 3D, but since they are in a pattern coordinate system, then, if the rig is planar, it may make sense to put the model to a XY coordinate plane so that Z-coordinate of each input object point is 0.
 
         In the old interface all the vectors of object points from different views are concatenated together.
@@ -156,9 +163,9 @@ Finds the camera intrinsic and extrinsic parameters from several views of a cali
         * **CV_CALIB_FIX_K1,...,CV_CALIB_FIX_K6** The corresponding radial distortion coefficient is not changed during the optimization. If  ``CV_CALIB_USE_INTRINSIC_GUESS``  is set, the coefficient from the supplied  ``distCoeffs``  matrix is used. Otherwise, it is set to 0.
 
         * **CV_CALIB_RATIONAL_MODEL** Coefficients k4, k5, and k6 are enabled. To provide the backward compatibility, this extra flag should be explicitly specified to make the calibration function use the rational model and return 8 coefficients. If the flag is not set, the function computes  and returns  only 5 distortion coefficients.
-        
+
         * **CALIB_THIN_PRISM_MODEL** Coefficients s1, s2, s3 and s4 are enabled. To provide the backward compatibility, this extra flag should be explicitly specified to make the calibration function use the thin prism model and return 12 coefficients. If the flag is not set, the function computes  and returns  only 5 distortion coefficients.
-        
+
         * **CALIB_FIX_S1_S2_S3_S4** The thin prism distortion coefficients are not changed during the optimization. If  ``CV_CALIB_USE_INTRINSIC_GUESS``  is set, the coefficient from the supplied  ``distCoeffs``  matrix is used. Otherwise, it is set to 0.
 
 
@@ -279,7 +286,7 @@ For points in an image of a stereo pair, computes the corresponding epilines in
 
 .. ocv:cfunction:: void cvComputeCorrespondEpilines( const CvMat* points, int which_image, const CvMat* fundamental_matrix, CvMat* correspondent_lines )
 
-.. ocv:pyoldfunction:: cv.ComputeCorrespondEpilines(points, whichImage, F, lines) -> None
+.. ocv:pyfunction:: cv2.computeCorrespondEpilines(points, whichImage, F[, lines]) -> lines
 
     :param points: Input points.  :math:`N \times 1`  or  :math:`1 \times N`  matrix of type  ``CV_32FC2``  or  ``vector<Point2f>`` .
 
@@ -354,7 +361,6 @@ Converts points to/from homogeneous coordinates.
 .. ocv:function:: void convertPointsHomogeneous( InputArray src, OutputArray dst )
 
 .. ocv:cfunction:: void cvConvertPointsHomogeneous( const CvMat* src, CvMat* dst )
-.. ocv:pyoldfunction:: cv.ConvertPointsHomogeneous(src, dst) -> None
 
     :param src: Input array or vector of 2D, 3D, or 4D points.
 
@@ -400,8 +406,6 @@ Decomposes a projection matrix into a rotation matrix and a camera matrix.
 
 .. ocv:cfunction:: void cvDecomposeProjectionMatrix( const CvMat * projMatr, CvMat * calibMatr, CvMat * rotMatr, CvMat * posVect, CvMat * rotMatrX=NULL, CvMat * rotMatrY=NULL, CvMat * rotMatrZ=NULL, CvPoint3D64f * eulerAngles=NULL )
 
-.. ocv:pyoldfunction:: cv.DecomposeProjectionMatrix(projMatrix, cameraMatrix, rotMatrix, transVect, rotMatrX=None, rotMatrY=None, rotMatrZ=None) -> eulerAngles
-
     :param projMatrix: 3x4 input projection matrix P.
 
     :param cameraMatrix: Output 3x3 camera matrix K.
@@ -436,7 +440,6 @@ Renders the detected chessboard corners.
 .. ocv:pyfunction:: cv2.drawChessboardCorners(image, patternSize, corners, patternWasFound) -> image
 
 .. ocv:cfunction:: void cvDrawChessboardCorners( CvArr* image, CvSize pattern_size, CvPoint2D32f* corners, int count, int pattern_was_found )
-.. ocv:pyoldfunction:: cv.DrawChessboardCorners(image, patternSize, corners, patternWasFound)-> None
 
     :param image: Destination image. It must be an 8-bit color image.
 
@@ -454,12 +457,11 @@ findChessboardCorners
 -------------------------
 Finds the positions of internal corners of the chessboard.
 
-.. ocv:function:: bool findChessboardCorners( InputArray image, Size patternSize, OutputArray corners, int flags=CALIB_CB_ADAPTIVE_THRESH+CALIB_CB_NORMALIZE_IMAGE )
+.. ocv:function:: bool findChessboardCorners( InputArray image, Size patternSize, OutputArray corners, int flags=CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE )
 
 .. ocv:pyfunction:: cv2.findChessboardCorners(image, patternSize[, corners[, flags]]) -> retval, corners
 
 .. ocv:cfunction:: int cvFindChessboardCorners( const void* image, CvSize pattern_size, CvPoint2D32f* corners, int* corner_count=NULL, int flags=CV_CALIB_CB_ADAPTIVE_THRESH+CV_CALIB_CB_NORMALIZE_IMAGE )
-.. ocv:pyoldfunction:: cv.FindChessboardCorners(image, patternSize, flags=CV_CALIB_CB_ADAPTIVE_THRESH) -> corners
 
     :param image: Source chessboard view. It must be an 8-bit grayscale or color image.
 
@@ -513,9 +515,9 @@ findCirclesGrid
 -------------------
 Finds centers in the grid of circles.
 
-.. ocv:function:: bool findCirclesGrid( InputArray image, Size patternSize, OutputArray centers, int flags=CALIB_CB_SYMMETRIC_GRID, const Ptr<FeatureDetector> &blobDetector = new SimpleBlobDetector() )
+.. ocv:function:: bool findCirclesGrid( InputArray image, Size patternSize, OutputArray centers, int flags=CALIB_CB_SYMMETRIC_GRID, const Ptr<FeatureDetector> &blobDetector = makePtr<SimpleBlobDetector>() )
 
-.. ocv:pyfunction:: cv2.findCirclesGridDefault(image, patternSize[, centers[, flags]]) -> retval, centers
+.. ocv:pyfunction:: cv2.findCirclesGrid(image, patternSize[, centers[, flags[, blobDetector]]]) -> retval, centers
 
     :param image: grid view of input circles; it must be an 8-bit grayscale or color image.
 
@@ -564,8 +566,6 @@ Finds an object pose from 3D-2D point correspondences.
 
 .. ocv:cfunction:: void cvFindExtrinsicCameraParams2( const CvMat* object_points, const CvMat* image_points, const CvMat* camera_matrix, const CvMat* distortion_coeffs, CvMat* rotation_vector, CvMat* translation_vector, int use_extrinsic_guess=0 )
 
-.. ocv:pyoldfunction:: cv.FindExtrinsicCameraParams2(objectPoints, imagePoints, cameraMatrix, distCoeffs, rvec, tvec, useExtrinsicGuess=0 ) -> None
-
     :param objectPoints: Array of object points in the object coordinate space, 3xN/Nx3 1-channel or 1xN/Nx1 3-channel, where N is the number of points.  ``vector<Point3f>``  can be also passed here.
 
     :param imagePoints: Array of corresponding image points, 2xN/Nx2 1-channel or 1xN/Nx1 2-channel, where N is the number of points.  ``vector<Point2f>``  can be also passed here.
@@ -588,7 +588,9 @@ Finds an object pose from 3D-2D point correspondences.
 
 The function estimates the object pose given a set of object points, their corresponding image projections, as well as the camera matrix and the distortion coefficients.
 
+.. note::
 
+   * An example of how to use solvePNP for planar augmented reality can be found at opencv_source_code/samples/python2/plane_ar.py
 
 solvePnPRansac
 ------------------
@@ -636,7 +638,6 @@ Calculates a fundamental matrix from the corresponding points in two images.
 .. ocv:pyfunction:: cv2.findFundamentalMat(points1, points2[, method[, param1[, param2[, mask]]]]) -> retval, mask
 
 .. ocv:cfunction:: int cvFindFundamentalMat( const CvMat* points1, const CvMat* points2, CvMat* fundamental_matrix, int method=CV_FM_RANSAC, double param1=3., double param2=0.99, CvMat* status=NULL )
-.. ocv:pyoldfunction:: cv.FindFundamentalMat(points1, points2, fundamentalMatrix, method=CV_FM_RANSAC, param1=1., param2=0.99, status=None) -> retval
 
     :param points1: Array of  ``N``  points from the first image. The point coordinates should be floating-point (single or double precision).
 
@@ -694,7 +695,7 @@ findEssentialMat
 ------------------
 Calculates an essential matrix from the corresponding points in two images.
 
-.. ocv:function:: Mat findEssentialMat( InputArray points1, InputArray points2, double focal=1.0, Point2d pp=Point2d(0, 0), int method=CV_RANSAC, double prob=0.999, double threshold=1.0, OutputArray mask=noArray() )
+.. ocv:function:: Mat findEssentialMat( InputArray points1, InputArray points2, double focal=1.0, Point2d pp=Point2d(0, 0), int method=RANSAC, double prob=0.999, double threshold=1.0, OutputArray mask=noArray() )
 
     :param points1: Array of  ``N`` ``(N >= 5)`` 2D points from the first image. The point coordinates should be floating-point (single or double precision).
 
@@ -715,7 +716,7 @@ Calculates an essential matrix from the corresponding points in two images.
 
     :param mask: Output array of N elements, every element of which is set to 0 for outliers and to 1 for the other points. The array is computed only in the RANSAC and LMedS methods.
 
-This function estimates essential matrix based on an implementation of five-point algorithm [Nister03]_ [SteweniusCFS]_.
+This function estimates essential matrix based on the five-point algorithm solver in [Nister03]_. [SteweniusCFS]_ is also a related.
 The epipolar geometry is described by the following equation:
 
 .. math::
@@ -820,8 +821,6 @@ Finds a perspective transformation between two planes.
 
 .. ocv:cfunction:: int cvFindHomography( const CvMat* src_points, const CvMat* dst_points, CvMat* homography, int method=0, double ransacReprojThreshold=3, CvMat* mask=0 )
 
-.. ocv:pyoldfunction:: cv.FindHomography(srcPoints, dstPoints, H, method=0, ransacReprojThreshold=3.0, status=None) -> None
-
     :param srcPoints: Coordinates of the points in the original plane, a matrix of the type  ``CV_32FC2``  or ``vector<Point2f>`` .
 
     :param dstPoints: Coordinates of the points in the target plane, a matrix of the type  ``CV_32FC2``  or a  ``vector<Point2f>`` .
@@ -893,6 +892,9 @@ Homography matrix is determined up to a scale. Thus, it is normalized so that
     :ocv:func:`warpPerspective`,
     :ocv:func:`perspectiveTransform`
 
+.. note::
+
+   * A example on calculating a homography for image matching can be found at opencv_source_code/samples/cpp/video_homography.cpp
 
 estimateAffine3D
 --------------------
@@ -946,8 +948,6 @@ Returns the new camera matrix based on the free scaling parameter.
 
 .. ocv:cfunction:: void cvGetOptimalNewCameraMatrix( const CvMat* camera_matrix, const CvMat* dist_coeffs, CvSize image_size, double alpha, CvMat* new_camera_matrix, CvSize new_imag_size=cvSize(0,0), CvRect* valid_pixel_ROI=0, int center_principal_point=0 )
 
-.. ocv:pyoldfunction:: cv.GetOptimalNewCameraMatrix(cameraMatrix, distCoeffs, imageSize, alpha, newCameraMatrix, newImageSize=(0, 0), validPixROI=0, centerPrincipalPoint=0) -> None
-
     :param cameraMatrix: Input camera matrix.
 
     :param distCoeffs: Input vector of distortion coefficients  :math:`(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6],[s_1, s_2, s_3, s_4]])`  of 4, 5, 8 or 12 elements. If the vector is NULL/empty, the zero distortion coefficients are assumed.
@@ -975,14 +975,12 @@ initCameraMatrix2D
 ----------------------
 Finds an initial camera matrix from 3D-2D point correspondences.
 
-.. ocv:function:: Mat initCameraMatrix2D( InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, Size imageSize, double aspectRatio=1.)
+.. ocv:function:: Mat initCameraMatrix2D( InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, Size imageSize, double aspectRatio=1.0 )
 
 .. ocv:pyfunction:: cv2.initCameraMatrix2D(objectPoints, imagePoints, imageSize[, aspectRatio]) -> retval
 
 .. ocv:cfunction:: void cvInitIntrinsicParams2D( const CvMat* object_points, const CvMat* image_points, const CvMat* npoints, CvSize image_size, CvMat* camera_matrix, double aspect_ratio=1. )
 
-.. ocv:pyoldfunction:: cv.InitIntrinsicParams2D(objectPoints, imagePoints, npoints, imageSize, cameraMatrix, aspectRatio=1.) -> None
-
     :param objectPoints: Vector of vectors of the calibration pattern points in the calibration pattern coordinate space. In the old interface all the per-view vectors are concatenated. See :ocv:func:`calibrateCamera` for details.
 
     :param imagePoints: Vector of vectors of the projections of the calibration pattern points. In the old interface all the per-view vectors are concatenated.
@@ -1030,8 +1028,6 @@ Projects 3D points to an image plane.
 
 .. ocv:cfunction:: void cvProjectPoints2( const CvMat* object_points, const CvMat* rotation_vector, const CvMat* translation_vector, const CvMat* camera_matrix, const CvMat* distortion_coeffs, CvMat* image_points, CvMat* dpdrot=NULL, CvMat* dpdt=NULL, CvMat* dpdf=NULL, CvMat* dpdc=NULL, CvMat* dpddist=NULL, double aspect_ratio=0 )
 
-.. ocv:pyoldfunction:: cv.ProjectPoints2(objectPoints, rvec, tvec, cameraMatrix, distCoeffs, imagePoints, dpdrot=None, dpdt=None, dpdf=None, dpdc=None, dpddist=None)-> None
-
     :param objectPoints: Array of object points, 3xN/Nx3 1-channel or 1xN/Nx1 3-channel  (or  ``vector<Point3f>`` ), where N is the number of points in the view.
 
     :param rvec: Rotation vector. See  :ocv:func:`Rodrigues` for details.
@@ -1075,15 +1071,13 @@ Reprojects a disparity image to 3D space.
 
 .. ocv:cfunction:: void cvReprojectImageTo3D( const CvArr* disparityImage, CvArr* _3dImage, const CvMat* Q, int handleMissingValues=0 )
 
-.. ocv:pyoldfunction:: cv.ReprojectImageTo3D(disparity, _3dImage, Q, handleMissingValues=0) -> None
-
     :param disparity: Input single-channel 8-bit unsigned, 16-bit signed, 32-bit signed or 32-bit floating-point disparity image.
 
     :param _3dImage: Output 3-channel floating-point image of the same size as  ``disparity`` . Each element of  ``_3dImage(x,y)``  contains 3D coordinates of the point  ``(x,y)``  computed from the disparity map.
 
     :param Q: :math:`4 \times 4`  perspective transformation matrix that can be obtained with  :ocv:func:`stereoRectify`.
 
-    :param handleMissingValues: Indicates, whether the function should handle missing values (i.e. points where the disparity was not computed). If ``handleMissingValues=true``, then pixels with the minimal disparity that corresponds to the outliers (see  :ocv:funcx:`StereoBM::operator()` ) are transformed to 3D points with a very large Z value (currently set to 10000).
+    :param handleMissingValues: Indicates, whether the function should handle missing values (i.e. points where the disparity was not computed). If ``handleMissingValues=true``, then pixels with the minimal disparity that corresponds to the outliers (see  :ocv:funcx:`StereoMatcher::compute` ) are transformed to 3D points with a very large Z value (currently set to 10000).
 
     :param ddepth: The optional output array depth. If it is ``-1``, the output image will have ``CV_32F`` depth. ``ddepth`` can also be set to ``CV_16S``, ``CV_32S`` or ``CV_32F``.
 
@@ -1109,7 +1103,6 @@ Computes an RQ decomposition of 3x3 matrices.
 .. ocv:pyfunction:: cv2.RQDecomp3x3(src[, mtxR[, mtxQ[, Qx[, Qy[, Qz]]]]]) -> retval, mtxR, mtxQ, Qx, Qy, Qz
 
 .. ocv:cfunction:: void cvRQDecomp3x3( const CvMat * matrixM, CvMat * matrixR, CvMat * matrixQ, CvMat * matrixQx=NULL, CvMat * matrixQy=NULL, CvMat * matrixQz=NULL, CvPoint3D64f * eulerAngles=NULL )
-.. ocv:pyoldfunction:: cv.RQDecomp3x3(M, R, Q, Qx=None, Qy=None, Qz=None) -> eulerAngles
 
     :param src: 3x3 input matrix.
 
@@ -1140,8 +1133,6 @@ Converts a rotation matrix to a rotation vector or vice versa.
 
 .. ocv:cfunction:: int cvRodrigues2( const CvMat* src, CvMat* dst, CvMat* jacobian=0 )
 
-.. ocv:pyoldfunction:: cv.Rodrigues2(src, dst, jacobian=0)-> None
-
     :param src: Input rotation vector (3x1 or 1x3) or rotation matrix (3x3).
 
     :param dst: Output rotation matrix (3x3) or rotation vector (3x1 or 1x3), respectively.
@@ -1166,155 +1157,84 @@ used in the global 3D geometry optimization procedures like
 :ocv:func:`solvePnP` .
 
 
+StereoMatcher
+-------------
+.. ocv:class:: StereoMatcher : public Algorithm
 
-StereoBM
---------
-.. ocv:class:: StereoBM
+The base class for stereo correspondence algorithms.
 
-Class for computing stereo correspondence using the block matching algorithm. ::
-
-    // Block matching stereo correspondence algorithm class StereoBM
-    {
-        enum { NORMALIZED_RESPONSE = CV_STEREO_BM_NORMALIZED_RESPONSE,
-            BASIC_PRESET=CV_STEREO_BM_BASIC,
-            FISH_EYE_PRESET=CV_STEREO_BM_FISH_EYE,
-            NARROW_PRESET=CV_STEREO_BM_NARROW };
-
-        StereoBM();
-        // the preset is one of ..._PRESET above.
-        // ndisparities is the size of disparity range,
-        // in which the optimal disparity at each pixel is searched for.
-        // SADWindowSize is the size of averaging window used to match pixel blocks
-        //    (larger values mean better robustness to noise, but yield blurry disparity maps)
-        StereoBM(int preset, int ndisparities=0, int SADWindowSize=21);
-        // separate initialization function
-        void init(int preset, int ndisparities=0, int SADWindowSize=21);
-        // computes the disparity for the two rectified 8-bit single-channel images.
-        // the disparity will be 16-bit signed (fixed-point) or 32-bit floating-point image of the same size as left.
-        void operator()( InputArray left, InputArray right, OutputArray disparity, int disptype=CV_16S );
-
-        Ptr<CvStereoBMState> state;
-    };
-
-The class is a C++ wrapper for the associated functions. In particular, :ocv:funcx:`StereoBM::operator()` is the wrapper for
-:ocv:cfunc:`cvFindStereoCorrespondenceBM`.
-
-
-StereoBM::StereoBM
-------------------
-The constructors.
-
-.. ocv:function:: StereoBM::StereoBM()
-.. ocv:function:: StereoBM::StereoBM(int preset, int ndisparities=0, int SADWindowSize=21)
-
-.. ocv:pyfunction:: cv2.StereoBM([preset[, ndisparities[, SADWindowSize]]]) -> <StereoBM object>
-
-.. ocv:cfunction:: CvStereoBMState* cvCreateStereoBMState( int preset=CV_STEREO_BM_BASIC, int numberOfDisparities=0 )
-
-.. ocv:pyoldfunction:: cv.CreateStereoBMState(preset=CV_STEREO_BM_BASIC, numberOfDisparities=0)-> CvStereoBMState
-
-    :param preset: specifies the whole set of algorithm parameters, one of:
-
-            * BASIC_PRESET - parameters suitable for general cameras
-            * FISH_EYE_PRESET - parameters suitable for wide-angle cameras
-            * NARROW_PRESET - parameters suitable for narrow-angle cameras
-
-        After constructing the class, you can override any parameters set by the preset.
-
-    :param ndisparities: the disparity search range. For each pixel algorithm will find the best disparity from 0 (default minimum disparity) to ``ndisparities``. The search range can then be shifted by changing the minimum disparity.
-
-    :param SADWindowSize: the linear size of the blocks compared by the algorithm. The size should be odd (as the block is centered at the current pixel). Larger block size implies smoother, though less accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher chance for algorithm to find a wrong correspondence.
-
-The constructors initialize ``StereoBM`` state. You can then call ``StereoBM::operator()`` to compute disparity for a specific stereo pair.
-
-.. note:: In the C API you need to deallocate ``CvStereoBM`` state when it is not needed anymore using ``cvReleaseStereoBMState(&stereobm)``.
-
-StereoBM::operator()
+StereoMatcher::compute
 -----------------------
-Computes disparity using the BM algorithm for a rectified stereo pair.
+Computes disparity map for the specified stereo pair
 
-.. ocv:function:: void StereoBM::operator()( InputArray left, InputArray right, OutputArray disparity, int disptype=CV_16S )
+.. ocv:function:: void StereoMatcher::compute( InputArray left, InputArray right, OutputArray disparity )
 
-.. ocv:pyfunction:: cv2.StereoBM.compute(left, right[, disparity[, disptype]]) -> disparity
-
-.. ocv:cfunction:: void cvFindStereoCorrespondenceBM( const CvArr* left, const CvArr* right, CvArr* disparity, CvStereoBMState* state )
-
-.. ocv:pyoldfunction:: cv.FindStereoCorrespondenceBM(left, right, disparity, state)-> None
+.. ocv:pyfunction:: cv2.StereoBM.compute(left, right[, disparity]) -> disparity
 
     :param left: Left 8-bit single-channel image.
 
     :param right: Right image of the same size and the same type as the left one.
 
-    :param disparity: Output disparity map. It has the same size as the input images. When ``disptype==CV_16S``, the map is a 16-bit signed single-channel image, containing disparity values scaled by 16. To get the true disparity values from such fixed-point representation, you will need to divide each  ``disp`` element by 16. If ``disptype==CV_32F``, the disparity map will already contain the real disparity values on output.
+    :param disparity: Output disparity map. It has the same size as the input images. Some algorithms, like StereoBM or StereoSGBM compute 16-bit fixed-point disparity map (where each disparity value has 4 fractional bits), whereas other algorithms output 32-bit floating-point disparity map.
 
-    :param disptype: Type of the output disparity map, ``CV_16S`` (default) or ``CV_32F``.
 
-    :param state: The pre-initialized ``CvStereoBMState`` structure in the case of the old API.
+StereoBM
+--------
+.. ocv:class:: StereoBM : public StereoMatcher
 
-The method executes the BM algorithm on a rectified stereo pair. See the ``stereo_match.cpp`` OpenCV sample on how to prepare images and call the method. Note that the method is not constant, thus you should not use the same ``StereoBM`` instance from within different threads simultaneously. The function is parallelized with the TBB library.
+Class for computing stereo correspondence using the block matching algorithm, introduced and contributed to OpenCV by K. Konolige.
 
+.. Sample code:
+
+   (Ocl) An example for using the stereoBM matching algorithm can be found at opencv_source_code/samples/ocl/stereo_match.cpp
+
+createStereoBM
+------------------
+Creates StereoBM object
+
+.. ocv:function:: Ptr<StereoBM> createStereoBM(int numDisparities=0, int blockSize=21)
+
+.. ocv:pyfunction:: cv2.createStereoBM([numDisparities[, blockSize]]) -> retval
+
+    :param numDisparities: the disparity search range. For each pixel algorithm will find the best disparity from 0 (default minimum disparity) to ``numDisparities``. The search range can then be shifted by changing the minimum disparity.
+
+    :param blockSize: the linear size of the blocks compared by the algorithm. The size should be odd (as the block is centered at the current pixel). Larger block size implies smoother, though less accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher chance for algorithm to find a wrong correspondence.
+
+The function create ``StereoBM`` object. You can then call ``StereoBM::compute()`` to compute disparity for a specific stereo pair.
 
 
 StereoSGBM
 ----------
 
-.. ocv:class:: StereoSGBM
-
-Class for computing stereo correspondence using the semi-global block matching algorithm. ::
-
-    class StereoSGBM
-    {
-        StereoSGBM();
-        StereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
-                   int P1=0, int P2=0, int disp12MaxDiff=0,
-                   int preFilterCap=0, int uniquenessRatio=0,
-                   int speckleWindowSize=0, int speckleRange=0,
-                   bool fullDP=false);
-        virtual ~StereoSGBM();
-
-        virtual void operator()(InputArray left, InputArray right, OutputArray disp);
-
-        int minDisparity;
-        int numberOfDisparities;
-        int SADWindowSize;
-        int preFilterCap;
-        int uniquenessRatio;
-        int P1, P2;
-        int speckleWindowSize;
-        int speckleRange;
-        int disp12MaxDiff;
-        bool fullDP;
-
-        ...
-    };
+.. ocv:class:: StereoSGBM : public StereoMatcher
 
 The class implements the modified H. Hirschmuller algorithm [HH08]_ that differs from the original one as follows:
 
- * By default, the algorithm is single-pass, which means that you consider only 5 directions instead of 8. Set ``fullDP=true`` to run the full variant of the algorithm but beware that it may consume a lot of memory.
+ * By default, the algorithm is single-pass, which means that you consider only 5 directions instead of 8. Set ``mode=StereoSGBM::MODE_HH`` in ``createStereoSGBM`` to run the full variant of the algorithm but beware that it may consume a lot of memory.
 
- * The algorithm matches blocks, not individual pixels. Though, setting ``SADWindowSize=1`` reduces the blocks to single pixels.
+ * The algorithm matches blocks, not individual pixels. Though, setting ``blockSize=1`` reduces the blocks to single pixels.
 
  * Mutual information cost function is not implemented. Instead, a simpler Birchfield-Tomasi sub-pixel metric from [BT98]_ is used. Though, the color images are supported as well.
 
- * Some pre- and post- processing steps from K. Konolige algorithm :ocv:funcx:`StereoBM::operator()`  are included, for example: pre-filtering (``CV_STEREO_BM_XSOBEL`` type) and post-filtering (uniqueness check, quadratic interpolation and speckle filtering).
+ * Some pre- and post- processing steps from K. Konolige algorithm ``StereoBM``  are included, for example: pre-filtering (``StereoBM::PREFILTER_XSOBEL`` type) and post-filtering (uniqueness check, quadratic interpolation and speckle filtering).
 
+.. note::
 
+   * (Python) An example illustrating the use of the StereoSGBM matching algorithm can be found at opencv_source_code/samples/python2/stereo_match.py
 
-StereoSGBM::StereoSGBM
+createStereoSGBM
 --------------------------
-.. ocv:function:: StereoSGBM::StereoSGBM()
+Creates StereoSGBM object
 
-.. ocv:function:: StereoSGBM::StereoSGBM( int minDisparity, int numDisparities, int SADWindowSize, int P1=0, int P2=0, int disp12MaxDiff=0, int preFilterCap=0, int uniquenessRatio=0, int speckleWindowSize=0, int speckleRange=0, bool fullDP=false)
+.. ocv:function:: Ptr<StereoSGBM> createStereoSGBM( int minDisparity, int numDisparities, int blockSize, int P1=0, int P2=0, int disp12MaxDiff=0, int preFilterCap=0, int uniquenessRatio=0, int speckleWindowSize=0, int speckleRange=0, int mode=StereoSGBM::MODE_SGBM)
 
-.. ocv:pyfunction:: cv2.StereoSGBM([minDisparity, numDisparities, SADWindowSize[, P1[, P2[, disp12MaxDiff[, preFilterCap[, uniquenessRatio[, speckleWindowSize[, speckleRange[, fullDP]]]]]]]]]) -> <StereoSGBM object>
-
-    Initializes ``StereoSGBM`` and sets parameters to custom values.??
+.. ocv:pyfunction:: cv2.createStereoSGBM(minDisparity, numDisparities, blockSize[, P1[, P2[, disp12MaxDiff[, preFilterCap[, uniquenessRatio[, speckleWindowSize[, speckleRange[, mode]]]]]]]]) -> retval
 
     :param minDisparity: Minimum possible disparity value. Normally, it is zero but sometimes rectification algorithms can shift images, so this parameter needs to be adjusted accordingly.
 
     :param numDisparities: Maximum disparity minus minimum disparity. The value is always greater than zero. In the current implementation, this parameter must be divisible by 16.
 
-    :param SADWindowSize: Matched block size. It must be an odd number  ``>=1`` . Normally, it should be somewhere in  the ``3..11``  range.
+    :param blockSize: Matched block size. It must be an odd number  ``>=1`` . Normally, it should be somewhere in  the ``3..11``  range.
 
     :param P1: The first parameter controlling the disparity smoothness. See below.
 
@@ -1330,32 +1250,12 @@ StereoSGBM::StereoSGBM
 
     :param speckleRange: Maximum disparity variation within each connected component. If you do speckle filtering, set the parameter to a positive value, it will be implicitly multiplied by 16. Normally, 1 or 2 is good enough.
 
-    :param fullDP: Set it to  ``true``  to run the full-scale two-pass dynamic programming algorithm. It will consume O(W*H*numDisparities) bytes, which is large for 640x480 stereo and huge for HD-size pictures. By default, it is set to ``false`` .
+    :param mode: Set it to  ``StereoSGBM::MODE_HH``  to run the full-scale two-pass dynamic programming algorithm. It will consume O(W*H*numDisparities) bytes, which is large for 640x480 stereo and huge for HD-size pictures. By default, it is set to ``false`` .
 
-The first constructor initializes ``StereoSGBM`` with all the default parameters. So, you only have to set ``StereoSGBM::numberOfDisparities`` at minimum. The second constructor enables you to set each parameter to a custom value.
+The first constructor initializes ``StereoSGBM`` with all the default parameters. So, you only have to set ``StereoSGBM::numDisparities`` at minimum. The second constructor enables you to set each parameter to a custom value.
 
 
 
-StereoSGBM::operator ()
------------------------
-
-.. ocv:function:: void StereoSGBM::operator()(InputArray left, InputArray right, OutputArray disp)
-
-.. ocv:pyfunction:: cv2.StereoSGBM.compute(left, right[, disp]) -> disp
-
-    Computes disparity using the SGBM algorithm for a rectified stereo pair.
-
-    :param left: Left 8-bit single-channel or 3-channel image.
-
-    :param right: Right image of the same size and the same type as the left one.
-
-    :param disp: Output disparity map. It is a 16-bit signed single-channel image of the same size as the input image. It contains disparity values  scaled by 16. So, to get the floating-point disparity map, you need to divide each  ``disp``  element by 16.
-
-The method executes the SGBM algorithm on a rectified stereo pair. See ``stereo_match.cpp`` OpenCV sample on how to prepare images and call the method.
-
-.. note:: The method is not constant, so you should not use the same ``StereoSGBM`` instance from different threads simultaneously.
-
-
 stereoCalibrate
 -------------------
 Calibrates the stereo camera.
@@ -1366,8 +1266,6 @@ Calibrates the stereo camera.
 
 .. ocv:cfunction:: double cvStereoCalibrate( const CvMat* object_points, const CvMat* image_points1, const CvMat* image_points2, const CvMat* npoints, CvMat* camera_matrix1, CvMat* dist_coeffs1, CvMat* camera_matrix2, CvMat* dist_coeffs2, CvSize image_size, CvMat* R, CvMat* T, CvMat* E=0, CvMat* F=0, CvTermCriteria term_crit=cvTermCriteria( CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,1e-6), int flags=CV_CALIB_FIX_INTRINSIC )
 
-.. ocv:pyoldfunction:: cv.StereoCalibrate(objectPoints, imagePoints1, imagePoints2, pointCounts, cameraMatrix1, distCoeffs1, cameraMatrix2, distCoeffs2, imageSize, R, T, E=None, F=None, term_crit=(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 30, 1e-6), flags=CV_CALIB_FIX_INTRINSIC)-> None
-
     :param objectPoints: Vector of vectors of the calibration pattern points.
 
     :param imagePoints1: Vector of vectors of the projections of the calibration pattern points, observed by the first camera.
@@ -1413,9 +1311,9 @@ Calibrates the stereo camera.
             * **CV_CALIB_FIX_K1,...,CV_CALIB_FIX_K6** Do not change the corresponding radial distortion coefficient during the optimization. If  ``CV_CALIB_USE_INTRINSIC_GUESS``  is set, the coefficient from the supplied  ``distCoeffs``  matrix is used. Otherwise, it is set to 0.
 
             * **CV_CALIB_RATIONAL_MODEL** Enable coefficients k4, k5, and k6. To provide the backward compatibility, this extra flag should be explicitly specified to make the calibration function use the rational model and return 8 coefficients. If the flag is not set, the function computes  and returns only 5 distortion coefficients.
-            
+
             * **CALIB_THIN_PRISM_MODEL** Coefficients s1, s2, s3 and s4 are enabled. To provide the backward compatibility, this extra flag should be explicitly specified to make the calibration function use the thin prism model and return 12 coefficients. If the flag is not set, the function computes  and returns  only 5 distortion coefficients.
-        
+
             * **CALIB_FIX_S1_S2_S3_S4** The thin prism distortion coefficients are not changed during the optimization. If  ``CV_CALIB_USE_INTRINSIC_GUESS``  is set, the coefficient from the supplied  ``distCoeffs``  matrix is used. Otherwise, it is set to 0.
 
 The function estimates transformation between two cameras making a stereo pair. If you have a stereo camera where the relative position and orientation of two cameras is fixed, and if you computed poses of an object relative to the first camera and to the second camera, (R1, T1) and (R2, T2), respectively (this can be done with
@@ -1459,8 +1357,6 @@ Computes rectification transforms for each head of a calibrated stereo camera.
 
 .. ocv:cfunction:: void cvStereoRectify( const CvMat* camera_matrix1, const CvMat* camera_matrix2, const CvMat* dist_coeffs1, const CvMat* dist_coeffs2, CvSize image_size, const CvMat* R, const CvMat* T, CvMat* R1, CvMat* R2, CvMat* P1, CvMat* P2, CvMat* Q=0, int flags=CV_CALIB_ZERO_DISPARITY, double alpha=-1, CvSize new_image_size=cvSize(0,0), CvRect* valid_pix_ROI1=0, CvRect* valid_pix_ROI2=0 )
 
-.. ocv:pyoldfunction:: cv.StereoRectify(cameraMatrix1, cameraMatrix2, distCoeffs1, distCoeffs2, imageSize, R, T, R1, R2, P1, P2, Q=None, flags=CV_CALIB_ZERO_DISPARITY, alpha=-1, newImageSize=(0, 0)) -> (roi1, roi2)
-
     :param cameraMatrix1: First camera matrix.
 
     :param cameraMatrix2: Second camera matrix.
@@ -1548,8 +1444,6 @@ Computes a rectification transform for an uncalibrated stereo camera.
 
 .. ocv:cfunction:: int cvStereoRectifyUncalibrated( const CvMat* points1, const CvMat* points2, const CvMat* F, CvSize img_size, CvMat* H1, CvMat* H2, double threshold=5 )
 
-.. ocv:pyoldfunction:: cv.StereoRectifyUncalibrated(points1, points2, F, imageSize, H1, H2, threshold=5)-> None
-
     :param points1: Array of feature points in the first image.
 
     :param points2: The corresponding points in the second image. The same formats as in  :ocv:func:`findFundamentalMat` are supported.
@@ -1614,6 +1508,6 @@ The function reconstructs 3-dimensional points (in homogeneous coordinates) by u
 
 .. [SteweniusCFS] Stewénius, H., Calibrated Fivepoint solver. http://www.vis.uky.edu/~stewe/FIVEPOINT/
 
-.. [Slabaugh] Slabaugh, G.G. Computing Euler angles from a rotation matrix. http://gregslabaugh.name/publications/euler.pdf
+.. [Slabaugh] Slabaugh, G.G. Computing Euler angles from a rotation matrix. http://www.soi.city.ac.uk/~sbbh653/publications/euler.pdf (verified: 2013-04-15)
 
 .. [Zhang2000] Z. Zhang. A Flexible New Technique for Camera Calibration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 22(11):1330-1334, 2000.
diff --git a/modules/calib3d/include/opencv2/calib3d.hpp b/modules/calib3d/include/opencv2/calib3d.hpp
index 989912cda8..1a8777b93c 100644
--- a/modules/calib3d/include/opencv2/calib3d.hpp
+++ b/modules/calib3d/include/opencv2/calib3d.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -46,556 +47,181 @@
 #include "opencv2/core.hpp"
 #include "opencv2/features2d.hpp"
 
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/****************************************************************************************\
-*                      Camera Calibration, Pose Estimation and Stereo                    *
-\****************************************************************************************/
-
-typedef struct CvPOSITObject CvPOSITObject;
-
-/* Allocates and initializes CvPOSITObject structure before doing cvPOSIT */
-CVAPI(CvPOSITObject*)  cvCreatePOSITObject( CvPoint3D32f* points, int point_count );
-
-
-/* Runs POSIT (POSe from ITeration) algorithm for determining 3d position of
-   an object given its model and projection in a weak-perspective case */
-CVAPI(void)  cvPOSIT(  CvPOSITObject* posit_object, CvPoint2D32f* image_points,
-                       double focal_length, CvTermCriteria criteria,
-                       float* rotation_matrix, float* translation_vector);
-
-/* Releases CvPOSITObject structure */
-CVAPI(void)  cvReleasePOSITObject( CvPOSITObject**  posit_object );
-
-/* updates the number of RANSAC iterations */
-CVAPI(int) cvRANSACUpdateNumIters( double p, double err_prob,
-                                   int model_points, int max_iters );
-
-CVAPI(void) cvConvertPointsHomogeneous( const CvMat* src, CvMat* dst );
-
-/* Calculates fundamental matrix given a set of corresponding points */
-#define CV_FM_7POINT 1
-#define CV_FM_8POINT 2
-
-#define CV_LMEDS 4
-#define CV_RANSAC 8
-
-#define CV_FM_LMEDS_ONLY  CV_LMEDS
-#define CV_FM_RANSAC_ONLY CV_RANSAC
-#define CV_FM_LMEDS CV_LMEDS
-#define CV_FM_RANSAC CV_RANSAC
-
-enum
-{
-    CV_ITERATIVE = 0,
-    CV_EPNP = 1, // F.Moreno-Noguer, V.Lepetit and P.Fua "EPnP: Efficient Perspective-n-Point Camera Pose Estimation"
-    CV_P3P = 2 // X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang; "Complete Solution Classification for the Perspective-Three-Point Problem"
-};
-
-CVAPI(int) cvFindFundamentalMat( const CvMat* points1, const CvMat* points2,
-                                 CvMat* fundamental_matrix,
-                                 int method CV_DEFAULT(CV_FM_RANSAC),
-                                 double param1 CV_DEFAULT(3.), double param2 CV_DEFAULT(0.99),
-                                 CvMat* status CV_DEFAULT(NULL) );
-
-/* For each input point on one of images
-   computes parameters of the corresponding
-   epipolar line on the other image */
-CVAPI(void) cvComputeCorrespondEpilines( const CvMat* points,
-                                         int which_image,
-                                         const CvMat* fundamental_matrix,
-                                         CvMat* correspondent_lines );
-
-/* Triangulation functions */
-
-CVAPI(void) cvTriangulatePoints(CvMat* projMatr1, CvMat* projMatr2,
-                                CvMat* projPoints1, CvMat* projPoints2,
-                                CvMat* points4D);
-
-CVAPI(void) cvCorrectMatches(CvMat* F, CvMat* points1, CvMat* points2,
-                             CvMat* new_points1, CvMat* new_points2);
-
-
-/* Computes the optimal new camera matrix according to the free scaling parameter alpha:
-   alpha=0 - only valid pixels will be retained in the undistorted image
-   alpha=1 - all the source image pixels will be retained in the undistorted image
-*/
-CVAPI(void) cvGetOptimalNewCameraMatrix( const CvMat* camera_matrix,
-                                         const CvMat* dist_coeffs,
-                                         CvSize image_size, double alpha,
-                                         CvMat* new_camera_matrix,
-                                         CvSize new_imag_size CV_DEFAULT(cvSize(0,0)),
-                                         CvRect* valid_pixel_ROI CV_DEFAULT(0),
-                                         int center_principal_point CV_DEFAULT(0));
-
-/* Converts rotation vector to rotation matrix or vice versa */
-CVAPI(int) cvRodrigues2( const CvMat* src, CvMat* dst,
-                         CvMat* jacobian CV_DEFAULT(0) );
-
-/* Finds perspective transformation between the object plane and image (view) plane */
-CVAPI(int) cvFindHomography( const CvMat* src_points,
-                             const CvMat* dst_points,
-                             CvMat* homography,
-                             int method CV_DEFAULT(0),
-                             double ransacReprojThreshold CV_DEFAULT(3),
-                             CvMat* mask CV_DEFAULT(0));
-
-/* Computes RQ decomposition for 3x3 matrices */
-CVAPI(void) cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
-                           CvMat *matrixQx CV_DEFAULT(NULL),
-                           CvMat *matrixQy CV_DEFAULT(NULL),
-                           CvMat *matrixQz CV_DEFAULT(NULL),
-                           CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
-
-/* Computes projection matrix decomposition */
-CVAPI(void) cvDecomposeProjectionMatrix( const CvMat *projMatr, CvMat *calibMatr,
-                                         CvMat *rotMatr, CvMat *posVect,
-                                         CvMat *rotMatrX CV_DEFAULT(NULL),
-                                         CvMat *rotMatrY CV_DEFAULT(NULL),
-                                         CvMat *rotMatrZ CV_DEFAULT(NULL),
-                                         CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
-
-/* Computes d(AB)/dA and d(AB)/dB */
-CVAPI(void) cvCalcMatMulDeriv( const CvMat* A, const CvMat* B, CvMat* dABdA, CvMat* dABdB );
-
-/* Computes r3 = rodrigues(rodrigues(r2)*rodrigues(r1)),
-   t3 = rodrigues(r2)*t1 + t2 and the respective derivatives */
-CVAPI(void) cvComposeRT( const CvMat* _rvec1, const CvMat* _tvec1,
-                         const CvMat* _rvec2, const CvMat* _tvec2,
-                         CvMat* _rvec3, CvMat* _tvec3,
-                         CvMat* dr3dr1 CV_DEFAULT(0), CvMat* dr3dt1 CV_DEFAULT(0),
-                         CvMat* dr3dr2 CV_DEFAULT(0), CvMat* dr3dt2 CV_DEFAULT(0),
-                         CvMat* dt3dr1 CV_DEFAULT(0), CvMat* dt3dt1 CV_DEFAULT(0),
-                         CvMat* dt3dr2 CV_DEFAULT(0), CvMat* dt3dt2 CV_DEFAULT(0) );
-
-/* Projects object points to the view plane using
-   the specified extrinsic and intrinsic camera parameters */
-CVAPI(void) cvProjectPoints2( const CvMat* object_points, const CvMat* rotation_vector,
-                              const CvMat* translation_vector, const CvMat* camera_matrix,
-                              const CvMat* distortion_coeffs, CvMat* image_points,
-                              CvMat* dpdrot CV_DEFAULT(NULL), CvMat* dpdt CV_DEFAULT(NULL),
-                              CvMat* dpdf CV_DEFAULT(NULL), CvMat* dpdc CV_DEFAULT(NULL),
-                              CvMat* dpddist CV_DEFAULT(NULL),
-                              double aspect_ratio CV_DEFAULT(0));
-
-/* Finds extrinsic camera parameters from
-   a few known corresponding point pairs and intrinsic parameters */
-CVAPI(void) cvFindExtrinsicCameraParams2( const CvMat* object_points,
-                                          const CvMat* image_points,
-                                          const CvMat* camera_matrix,
-                                          const CvMat* distortion_coeffs,
-                                          CvMat* rotation_vector,
-                                          CvMat* translation_vector,
-                                          int use_extrinsic_guess CV_DEFAULT(0) );
-
-/* Computes initial estimate of the intrinsic camera parameters
-   in case of planar calibration target (e.g. chessboard) */
-CVAPI(void) cvInitIntrinsicParams2D( const CvMat* object_points,
-                                     const CvMat* image_points,
-                                     const CvMat* npoints, CvSize image_size,
-                                     CvMat* camera_matrix,
-                                     double aspect_ratio CV_DEFAULT(1.) );
-
-#define CV_CALIB_CB_ADAPTIVE_THRESH  1
-#define CV_CALIB_CB_NORMALIZE_IMAGE  2
-#define CV_CALIB_CB_FILTER_QUADS     4
-#define CV_CALIB_CB_FAST_CHECK       8
-
-// Performs a fast check if a chessboard is in the input image. This is a workaround to
-// a problem of cvFindChessboardCorners being slow on images with no chessboard
-// - src: input image
-// - size: chessboard size
-// Returns 1 if a chessboard can be in this image and findChessboardCorners should be called,
-// 0 if there is no chessboard, -1 in case of error
-CVAPI(int) cvCheckChessboard(IplImage* src, CvSize size);
-
-    /* Detects corners on a chessboard calibration pattern */
-CVAPI(int) cvFindChessboardCorners( const void* image, CvSize pattern_size,
-                                    CvPoint2D32f* corners,
-                                    int* corner_count CV_DEFAULT(NULL),
-                                    int flags CV_DEFAULT(CV_CALIB_CB_ADAPTIVE_THRESH+CV_CALIB_CB_NORMALIZE_IMAGE) );
-
-/* Draws individual chessboard corners or the whole chessboard detected */
-CVAPI(void) cvDrawChessboardCorners( CvArr* image, CvSize pattern_size,
-                                     CvPoint2D32f* corners,
-                                     int count, int pattern_was_found );
-
-#define CV_CALIB_USE_INTRINSIC_GUESS  1
-#define CV_CALIB_FIX_ASPECT_RATIO     2
-#define CV_CALIB_FIX_PRINCIPAL_POINT  4
-#define CV_CALIB_ZERO_TANGENT_DIST    8
-#define CV_CALIB_FIX_FOCAL_LENGTH 16
-#define CV_CALIB_FIX_K1  32
-#define CV_CALIB_FIX_K2  64
-#define CV_CALIB_FIX_K3  128
-#define CV_CALIB_FIX_K4  2048
-#define CV_CALIB_FIX_K5  4096
-#define CV_CALIB_FIX_K6  8192
-#define CV_CALIB_RATIONAL_MODEL 16384
-#define CV_CALIB_THIN_PRISM_MODEL 32768
-#define CV_CALIB_FIX_S1_S2_S3_S4  65536
-
-
-/* Finds intrinsic and extrinsic camera parameters
-   from a few views of known calibration pattern */
-CVAPI(double) cvCalibrateCamera2( const CvMat* object_points,
-                                const CvMat* image_points,
-                                const CvMat* point_counts,
-                                CvSize image_size,
-                                CvMat* camera_matrix,
-                                CvMat* distortion_coeffs,
-                                CvMat* rotation_vectors CV_DEFAULT(NULL),
-                                CvMat* translation_vectors CV_DEFAULT(NULL),
-                                int flags CV_DEFAULT(0),
-                                CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
-                                    CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,DBL_EPSILON)) );
-
-/* Computes various useful characteristics of the camera from the data computed by
-   cvCalibrateCamera2 */
-CVAPI(void) cvCalibrationMatrixValues( const CvMat *camera_matrix,
-                                CvSize image_size,
-                                double aperture_width CV_DEFAULT(0),
-                                double aperture_height CV_DEFAULT(0),
-                                double *fovx CV_DEFAULT(NULL),
-                                double *fovy CV_DEFAULT(NULL),
-                                double *focal_length CV_DEFAULT(NULL),
-                                CvPoint2D64f *principal_point CV_DEFAULT(NULL),
-                                double *pixel_aspect_ratio CV_DEFAULT(NULL));
-
-#define CV_CALIB_FIX_INTRINSIC  256
-#define CV_CALIB_SAME_FOCAL_LENGTH 512
-
-/* Computes the transformation from one camera coordinate system to another one
-   from a few correspondent views of the same calibration target. Optionally, calibrates
-   both cameras */
-CVAPI(double) cvStereoCalibrate( const CvMat* object_points, const CvMat* image_points1,
-                               const CvMat* image_points2, const CvMat* npoints,
-                               CvMat* camera_matrix1, CvMat* dist_coeffs1,
-                               CvMat* camera_matrix2, CvMat* dist_coeffs2,
-                               CvSize image_size, CvMat* R, CvMat* T,
-                               CvMat* E CV_DEFAULT(0), CvMat* F CV_DEFAULT(0),
-                               CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
-                                   CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,1e-6)),
-                               int flags CV_DEFAULT(CV_CALIB_FIX_INTRINSIC));
-
-#define CV_CALIB_ZERO_DISPARITY 1024
-
-/* Computes 3D rotations (+ optional shift) for each camera coordinate system to make both
-   views parallel (=> to make all the epipolar lines horizontal or vertical) */
-CVAPI(void) cvStereoRectify( const CvMat* camera_matrix1, const CvMat* camera_matrix2,
-                             const CvMat* dist_coeffs1, const CvMat* dist_coeffs2,
-                             CvSize image_size, const CvMat* R, const CvMat* T,
-                             CvMat* R1, CvMat* R2, CvMat* P1, CvMat* P2,
-                             CvMat* Q CV_DEFAULT(0),
-                             int flags CV_DEFAULT(CV_CALIB_ZERO_DISPARITY),
-                             double alpha CV_DEFAULT(-1),
-                             CvSize new_image_size CV_DEFAULT(cvSize(0,0)),
-                             CvRect* valid_pix_ROI1 CV_DEFAULT(0),
-                             CvRect* valid_pix_ROI2 CV_DEFAULT(0));
-
-/* Computes rectification transformations for uncalibrated pair of images using a set
-   of point correspondences */
-CVAPI(int) cvStereoRectifyUncalibrated( const CvMat* points1, const CvMat* points2,
-                                        const CvMat* F, CvSize img_size,
-                                        CvMat* H1, CvMat* H2,
-                                        double threshold CV_DEFAULT(5));
-
-
-
-/* stereo correspondence parameters and functions */
-
-#define CV_STEREO_BM_NORMALIZED_RESPONSE  0
-#define CV_STEREO_BM_XSOBEL               1
-
-/* Block matching algorithm structure */
-typedef struct CvStereoBMState
-{
-    // pre-filtering (normalization of input images)
-    int preFilterType; // =CV_STEREO_BM_NORMALIZED_RESPONSE now
-    int preFilterSize; // averaging window size: ~5x5..21x21
-    int preFilterCap; // the output of pre-filtering is clipped by [-preFilterCap,preFilterCap]
-
-    // correspondence using Sum of Absolute Difference (SAD)
-    int SADWindowSize; // ~5x5..21x21
-    int minDisparity;  // minimum disparity (can be negative)
-    int numberOfDisparities; // maximum disparity - minimum disparity (> 0)
-
-    // post-filtering
-    int textureThreshold;  // the disparity is only computed for pixels
-                           // with textured enough neighborhood
-    int uniquenessRatio;   // accept the computed disparity d* only if
-                           // SAD(d) >= SAD(d*)*(1 + uniquenessRatio/100.)
-                           // for any d != d*+/-1 within the search range.
-    int speckleWindowSize; // disparity variation window
-    int speckleRange; // acceptable range of variation in window
-
-    int trySmallerWindows; // if 1, the results may be more accurate,
-                           // at the expense of slower processing
-    CvRect roi1, roi2;
-    int disp12MaxDiff;
-
-    // temporary buffers
-    CvMat* preFilteredImg0;
-    CvMat* preFilteredImg1;
-    CvMat* slidingSumBuf;
-    CvMat* cost;
-    CvMat* disp;
-} CvStereoBMState;
-
-#define CV_STEREO_BM_BASIC 0
-#define CV_STEREO_BM_FISH_EYE 1
-#define CV_STEREO_BM_NARROW 2
-
-CVAPI(CvStereoBMState*) cvCreateStereoBMState(int preset CV_DEFAULT(CV_STEREO_BM_BASIC),
-                                              int numberOfDisparities CV_DEFAULT(0));
-
-CVAPI(void) cvReleaseStereoBMState( CvStereoBMState** state );
-
-CVAPI(void) cvFindStereoCorrespondenceBM( const CvArr* left, const CvArr* right,
-                                          CvArr* disparity, CvStereoBMState* state );
-
-CVAPI(CvRect) cvGetValidDisparityROI( CvRect roi1, CvRect roi2, int minDisparity,
-                                      int numberOfDisparities, int SADWindowSize );
-
-CVAPI(void) cvValidateDisparity( CvArr* disparity, const CvArr* cost,
-                                 int minDisparity, int numberOfDisparities,
-                                 int disp12MaxDiff CV_DEFAULT(1) );
-
-/* Reprojects the computed disparity image to the 3D space using the specified 4x4 matrix */
-CVAPI(void)  cvReprojectImageTo3D( const CvArr* disparityImage,
-                                   CvArr* _3dImage, const CvMat* Q,
-                                   int handleMissingValues CV_DEFAULT(0) );
-
-
-#ifdef __cplusplus
-}
-
-//////////////////////////////////////////////////////////////////////////////////////////
-class CV_EXPORTS CvLevMarq
-{
-public:
-    CvLevMarq();
-    CvLevMarq( int nparams, int nerrs, CvTermCriteria criteria=
-              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
-              bool completeSymmFlag=false );
-    ~CvLevMarq();
-    void init( int nparams, int nerrs, CvTermCriteria criteria=
-              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
-              bool completeSymmFlag=false );
-    bool update( const CvMat*& param, CvMat*& J, CvMat*& err );
-    bool updateAlt( const CvMat*& param, CvMat*& JtJ, CvMat*& JtErr, double*& errNorm );
-
-    void clear();
-    void step();
-    enum { DONE=0, STARTED=1, CALC_J=2, CHECK_ERR=3 };
-
-    cv::Ptr<CvMat> mask;
-    cv::Ptr<CvMat> prevParam;
-    cv::Ptr<CvMat> param;
-    cv::Ptr<CvMat> J;
-    cv::Ptr<CvMat> err;
-    cv::Ptr<CvMat> JtJ;
-    cv::Ptr<CvMat> JtJN;
-    cv::Ptr<CvMat> JtErr;
-    cv::Ptr<CvMat> JtJV;
-    cv::Ptr<CvMat> JtJW;
-    double prevErrNorm, errNorm;
-    int lambdaLg10;
-    CvTermCriteria criteria;
-    int state;
-    int iters;
-    bool completeSymmFlag;
-};
-
 namespace cv
 {
-//! converts rotation vector to rotation matrix or vice versa using Rodrigues transformation
-CV_EXPORTS_W void Rodrigues(InputArray src, OutputArray dst, OutputArray jacobian=noArray());
 
 //! type of the robust estimation algorithm
-enum
-{
-    LMEDS=CV_LMEDS, //!< least-median algorithm
-    RANSAC=CV_RANSAC //!< RANSAC algorithm
-};
+enum { LMEDS  = 4, //!< least-median algorithm
+       RANSAC = 8  //!< RANSAC algorithm
+     };
+
+enum { ITERATIVE = 0,
+       EPNP      = 1, // F.Moreno-Noguer, V.Lepetit and P.Fua "EPnP: Efficient Perspective-n-Point Camera Pose Estimation"
+       P3P       = 2  // X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang; "Complete Solution Classification for the Perspective-Three-Point Problem"
+     };
+
+enum { CALIB_CB_ADAPTIVE_THRESH = 1,
+       CALIB_CB_NORMALIZE_IMAGE = 2,
+       CALIB_CB_FILTER_QUADS    = 4,
+       CALIB_CB_FAST_CHECK      = 8
+     };
+
+enum { CALIB_CB_SYMMETRIC_GRID  = 1,
+       CALIB_CB_ASYMMETRIC_GRID = 2,
+       CALIB_CB_CLUSTERING      = 4
+     };
+
+enum { CALIB_USE_INTRINSIC_GUESS = 0x00001,
+       CALIB_FIX_ASPECT_RATIO    = 0x00002,
+       CALIB_FIX_PRINCIPAL_POINT = 0x00004,
+       CALIB_ZERO_TANGENT_DIST   = 0x00008,
+       CALIB_FIX_FOCAL_LENGTH    = 0x00010,
+       CALIB_FIX_K1              = 0x00020,
+       CALIB_FIX_K2              = 0x00040,
+       CALIB_FIX_K3              = 0x00080,
+       CALIB_FIX_K4              = 0x00800,
+       CALIB_FIX_K5              = 0x01000,
+       CALIB_FIX_K6              = 0x02000,
+       CALIB_RATIONAL_MODEL      = 0x04000,
+       CALIB_THIN_PRISM_MODEL    = 0x08000,
+       CALIB_FIX_S1_S2_S3_S4     = 0x10000,
+       // only for stereo
+       CALIB_FIX_INTRINSIC       = 0x00100,
+       CALIB_SAME_FOCAL_LENGTH   = 0x00200,
+       // for stereo rectification
+       CALIB_ZERO_DISPARITY      = 0x00400
+     };
+
+//! the algorithm for finding fundamental matrix
+enum { FM_7POINT = 1, //!< 7-point algorithm
+       FM_8POINT = 2, //!< 8-point algorithm
+       FM_LMEDS  = 4, //!< least-median algorithm
+       FM_RANSAC = 8  //!< RANSAC algorithm
+     };
+
+
+
+//! converts rotation vector to rotation matrix or vice versa using Rodrigues transformation
+CV_EXPORTS_W void Rodrigues( InputArray src, OutputArray dst, OutputArray jacobian = noArray() );
 
 //! computes the best-fit perspective transformation mapping srcPoints to dstPoints.
 CV_EXPORTS_W Mat findHomography( InputArray srcPoints, InputArray dstPoints,
-                                 int method=0, double ransacReprojThreshold=3,
+                                 int method = 0, double ransacReprojThreshold = 3,
                                  OutputArray mask=noArray());
 
 //! variant of findHomography for backward compatibility
 CV_EXPORTS Mat findHomography( InputArray srcPoints, InputArray dstPoints,
-                               OutputArray mask, int method=0, double ransacReprojThreshold=3);
+                               OutputArray mask, int method = 0, double ransacReprojThreshold = 3 );
 
 //! Computes RQ decomposition of 3x3 matrix
 CV_EXPORTS_W Vec3d RQDecomp3x3( InputArray src, OutputArray mtxR, OutputArray mtxQ,
-                                OutputArray Qx=noArray(),
-                                OutputArray Qy=noArray(),
-                                OutputArray Qz=noArray());
+                                OutputArray Qx = noArray(),
+                                OutputArray Qy = noArray(),
+                                OutputArray Qz = noArray());
 
 //! Decomposes the projection matrix into camera matrix and the rotation martix and the translation vector
 CV_EXPORTS_W void decomposeProjectionMatrix( InputArray projMatrix, OutputArray cameraMatrix,
                                              OutputArray rotMatrix, OutputArray transVect,
-                                             OutputArray rotMatrixX=noArray(),
-                                             OutputArray rotMatrixY=noArray(),
-                                             OutputArray rotMatrixZ=noArray(),
-                                             OutputArray eulerAngles=noArray() );
+                                             OutputArray rotMatrixX = noArray(),
+                                             OutputArray rotMatrixY = noArray(),
+                                             OutputArray rotMatrixZ = noArray(),
+                                             OutputArray eulerAngles =noArray() );
 
 //! computes derivatives of the matrix product w.r.t each of the multiplied matrix coefficients
-CV_EXPORTS_W void matMulDeriv( InputArray A, InputArray B,
-                               OutputArray dABdA,
-                               OutputArray dABdB );
+CV_EXPORTS_W void matMulDeriv( InputArray A, InputArray B, OutputArray dABdA, OutputArray dABdB );
 
 //! composes 2 [R|t] transformations together. Also computes the derivatives of the result w.r.t the arguments
 CV_EXPORTS_W void composeRT( InputArray rvec1, InputArray tvec1,
                              InputArray rvec2, InputArray tvec2,
                              OutputArray rvec3, OutputArray tvec3,
-                             OutputArray dr3dr1=noArray(), OutputArray dr3dt1=noArray(),
-                             OutputArray dr3dr2=noArray(), OutputArray dr3dt2=noArray(),
-                             OutputArray dt3dr1=noArray(), OutputArray dt3dt1=noArray(),
-                             OutputArray dt3dr2=noArray(), OutputArray dt3dt2=noArray() );
+                             OutputArray dr3dr1 = noArray(), OutputArray dr3dt1 = noArray(),
+                             OutputArray dr3dr2 = noArray(), OutputArray dr3dt2 = noArray(),
+                             OutputArray dt3dr1 = noArray(), OutputArray dt3dt1 = noArray(),
+                             OutputArray dt3dr2 = noArray(), OutputArray dt3dt2 = noArray() );
 
 //! projects points from the model coordinate space to the image coordinates. Also computes derivatives of the image coordinates w.r.t the intrinsic and extrinsic camera parameters
 CV_EXPORTS_W void projectPoints( InputArray objectPoints,
                                  InputArray rvec, InputArray tvec,
                                  InputArray cameraMatrix, InputArray distCoeffs,
                                  OutputArray imagePoints,
-                                 OutputArray jacobian=noArray(),
-                                 double aspectRatio=0 );
+                                 OutputArray jacobian = noArray(),
+                                 double aspectRatio = 0 );
 
 //! computes the camera pose from a few 3D points and the corresponding projections. The outliers are not handled.
-enum
-{
-    ITERATIVE=CV_ITERATIVE,
-    EPNP=CV_EPNP,
-    P3P=CV_P3P
-};
 CV_EXPORTS_W bool solvePnP( InputArray objectPoints, InputArray imagePoints,
                             InputArray cameraMatrix, InputArray distCoeffs,
                             OutputArray rvec, OutputArray tvec,
-                            bool useExtrinsicGuess=false, int flags=ITERATIVE);
+                            bool useExtrinsicGuess = false, int flags = ITERATIVE );
 
 //! computes the camera pose from a few 3D points and the corresponding projections. The outliers are possible.
-CV_EXPORTS_W void solvePnPRansac( InputArray objectPoints,
-                                  InputArray imagePoints,
-                                  InputArray cameraMatrix,
-                                  InputArray distCoeffs,
-                                  OutputArray rvec,
-                                  OutputArray tvec,
-                                  bool useExtrinsicGuess = false,
-                                  int iterationsCount = 100,
-                                  float reprojectionError = 8.0,
-                                  int minInliersCount = 100,
-                                  OutputArray inliers = noArray(),
-                                  int flags = ITERATIVE);
+CV_EXPORTS_W void solvePnPRansac( InputArray objectPoints, InputArray imagePoints,
+                                  InputArray cameraMatrix, InputArray distCoeffs,
+                                  OutputArray rvec, OutputArray tvec,
+                                  bool useExtrinsicGuess = false, int iterationsCount = 100,
+                                  float reprojectionError = 8.0, int minInliersCount = 100,
+                                  OutputArray inliers = noArray(), int flags = ITERATIVE );
 
 //! initializes camera matrix from a few 3D points and the corresponding projections.
 CV_EXPORTS_W Mat initCameraMatrix2D( InputArrayOfArrays objectPoints,
                                      InputArrayOfArrays imagePoints,
-                                     Size imageSize, double aspectRatio=1. );
-
-enum { CALIB_CB_ADAPTIVE_THRESH = 1, CALIB_CB_NORMALIZE_IMAGE = 2,
-       CALIB_CB_FILTER_QUADS = 4, CALIB_CB_FAST_CHECK = 8 };
+                                     Size imageSize, double aspectRatio = 1.0 );
 
 //! finds checkerboard pattern of the specified size in the image
-CV_EXPORTS_W bool findChessboardCorners( InputArray image, Size patternSize,
-                                         OutputArray corners,
-                                         int flags=CALIB_CB_ADAPTIVE_THRESH+CALIB_CB_NORMALIZE_IMAGE );
+CV_EXPORTS_W bool findChessboardCorners( InputArray image, Size patternSize, OutputArray corners,
+                                         int flags = CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE );
 
 //! finds subpixel-accurate positions of the chessboard corners
-CV_EXPORTS bool find4QuadCornerSubpix(InputArray img, InputOutputArray corners, Size region_size);
+CV_EXPORTS bool find4QuadCornerSubpix( InputArray img, InputOutputArray corners, Size region_size );
 
 //! draws the checkerboard pattern (found or partly found) in the image
 CV_EXPORTS_W void drawChessboardCorners( InputOutputArray image, Size patternSize,
                                          InputArray corners, bool patternWasFound );
 
-enum { CALIB_CB_SYMMETRIC_GRID = 1, CALIB_CB_ASYMMETRIC_GRID = 2,
-       CALIB_CB_CLUSTERING = 4 };
-
 //! finds circles' grid pattern of the specified size in the image
 CV_EXPORTS_W bool findCirclesGrid( InputArray image, Size patternSize,
-                                 OutputArray centers, int flags=CALIB_CB_SYMMETRIC_GRID,
-                                 const Ptr<FeatureDetector> &blobDetector = new SimpleBlobDetector());
-
-//! the deprecated function. Use findCirclesGrid() instead of it.
-CV_EXPORTS_W bool findCirclesGridDefault( InputArray image, Size patternSize,
-                                          OutputArray centers, int flags=CALIB_CB_SYMMETRIC_GRID );
-enum
-{
-    CALIB_USE_INTRINSIC_GUESS = CV_CALIB_USE_INTRINSIC_GUESS,
-    CALIB_FIX_ASPECT_RATIO = CV_CALIB_FIX_ASPECT_RATIO,
-    CALIB_FIX_PRINCIPAL_POINT = CV_CALIB_FIX_PRINCIPAL_POINT,
-    CALIB_ZERO_TANGENT_DIST = CV_CALIB_ZERO_TANGENT_DIST,
-    CALIB_FIX_FOCAL_LENGTH = CV_CALIB_FIX_FOCAL_LENGTH,
-    CALIB_FIX_K1 = CV_CALIB_FIX_K1,
-    CALIB_FIX_K2 = CV_CALIB_FIX_K2,
-    CALIB_FIX_K3 = CV_CALIB_FIX_K3,
-    CALIB_FIX_K4 = CV_CALIB_FIX_K4,
-    CALIB_FIX_K5 = CV_CALIB_FIX_K5,
-    CALIB_FIX_K6 = CV_CALIB_FIX_K6,
-    CALIB_RATIONAL_MODEL = CV_CALIB_RATIONAL_MODEL,
-    CALIB_THIN_PRISM_MODEL = CV_CALIB_THIN_PRISM_MODEL,
-    CALIB_FIX_S1_S2_S3_S4=CV_CALIB_FIX_S1_S2_S3_S4,
-    // only for stereo
-    CALIB_FIX_INTRINSIC = CV_CALIB_FIX_INTRINSIC,
-    CALIB_SAME_FOCAL_LENGTH = CV_CALIB_SAME_FOCAL_LENGTH,
-    // for stereo rectification
-    CALIB_ZERO_DISPARITY = CV_CALIB_ZERO_DISPARITY
-};
+                                   OutputArray centers, int flags = CALIB_CB_SYMMETRIC_GRID,
+                                   const Ptr<FeatureDetector> &blobDetector = makePtr<SimpleBlobDetector>());
 
 //! finds intrinsic and extrinsic camera parameters from several fews of a known calibration pattern.
 CV_EXPORTS_W double calibrateCamera( InputArrayOfArrays objectPoints,
-                                     InputArrayOfArrays imagePoints,
-                                     Size imageSize,
-                                     InputOutputArray cameraMatrix,
-                                     InputOutputArray distCoeffs,
+                                     InputArrayOfArrays imagePoints, Size imageSize,
+                                     InputOutputArray cameraMatrix, InputOutputArray distCoeffs,
                                      OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs,
-                                     int flags=0, TermCriteria criteria = TermCriteria(
-                                         TermCriteria::COUNT+TermCriteria::EPS, 30, DBL_EPSILON) );
+                                     int flags = 0, TermCriteria criteria = TermCriteria(
+                                        TermCriteria::COUNT + TermCriteria::EPS, 30, DBL_EPSILON) );
 
 //! computes several useful camera characteristics from the camera matrix, camera frame resolution and the physical sensor size.
-CV_EXPORTS_W void calibrationMatrixValues( InputArray cameraMatrix,
-                                Size imageSize,
-                                double apertureWidth,
-                                double apertureHeight,
-                                CV_OUT double& fovx,
-                                CV_OUT double& fovy,
-                                CV_OUT double& focalLength,
-                                CV_OUT Point2d& principalPoint,
-                                CV_OUT double& aspectRatio );
+CV_EXPORTS_W void calibrationMatrixValues( InputArray cameraMatrix, Size imageSize,
+                                           double apertureWidth, double apertureHeight,
+                                           CV_OUT double& fovx, CV_OUT double& fovy,
+                                           CV_OUT double& focalLength, CV_OUT Point2d& principalPoint,
+                                           CV_OUT double& aspectRatio );
 
 //! finds intrinsic and extrinsic parameters of a stereo camera
 CV_EXPORTS_W double stereoCalibrate( InputArrayOfArrays objectPoints,
-                                     InputArrayOfArrays imagePoints1,
-                                     InputArrayOfArrays imagePoints2,
-                                     InputOutputArray cameraMatrix1,
-                                     InputOutputArray distCoeffs1,
-                                     InputOutputArray cameraMatrix2,
-                                     InputOutputArray distCoeffs2,
-                                     Size imageSize, OutputArray R,
-                                     OutputArray T, OutputArray E, OutputArray F,
+                                     InputArrayOfArrays imagePoints1, InputArrayOfArrays imagePoints2,
+                                     InputOutputArray cameraMatrix1, InputOutputArray distCoeffs1,
+                                     InputOutputArray cameraMatrix2, InputOutputArray distCoeffs2,
+                                     Size imageSize, OutputArray R,OutputArray T, OutputArray E, OutputArray F,
                                      TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 1e-6),
-                                     int flags=CALIB_FIX_INTRINSIC );
+                                     int flags = CALIB_FIX_INTRINSIC );
 
 
 //! computes the rectification transformation for a stereo camera from its intrinsic and extrinsic parameters
 CV_EXPORTS_W void stereoRectify( InputArray cameraMatrix1, InputArray distCoeffs1,
-                               InputArray cameraMatrix2, InputArray distCoeffs2,
-                               Size imageSize, InputArray R, InputArray T,
-                               OutputArray R1, OutputArray R2,
-                               OutputArray P1, OutputArray P2,
-                               OutputArray Q, int flags=CALIB_ZERO_DISPARITY,
-                               double alpha=-1, Size newImageSize=Size(),
-                               CV_OUT Rect* validPixROI1=0, CV_OUT Rect* validPixROI2=0 );
+                                 InputArray cameraMatrix2, InputArray distCoeffs2,
+                                 Size imageSize, InputArray R, InputArray T,
+                                 OutputArray R1, OutputArray R2,
+                                 OutputArray P1, OutputArray P2,
+                                 OutputArray Q, int flags = CALIB_ZERO_DISPARITY,
+                                 double alpha = -1, Size newImageSize = Size(),
+                                 CV_OUT Rect* validPixROI1 = 0, CV_OUT Rect* validPixROI2 = 0 );
 
 //! computes the rectification transformation for an uncalibrated stereo camera (zero distortion is assumed)
 CV_EXPORTS_W bool stereoRectifyUncalibrated( InputArray points1, InputArray points2,
                                              InputArray F, Size imgSize,
                                              OutputArray H1, OutputArray H2,
-                                             double threshold=5 );
+                                             double threshold = 5 );
 
 //! computes the rectification transformations for 3-head camera, where all the heads are on the same line.
 CV_EXPORTS_W float rectify3Collinear( InputArray cameraMatrix1, InputArray distCoeffs1,
@@ -611,8 +237,9 @@ CV_EXPORTS_W float rectify3Collinear( InputArray cameraMatrix1, InputArray distC
 
 //! returns the optimal new camera matrix
 CV_EXPORTS_W Mat getOptimalNewCameraMatrix( InputArray cameraMatrix, InputArray distCoeffs,
-                                            Size imageSize, double alpha, Size newImgSize=Size(),
-                                            CV_OUT Rect* validPixROI=0, bool centerPrincipalPoint=false);
+                                            Size imageSize, double alpha, Size newImgSize = Size(),
+                                            CV_OUT Rect* validPixROI = 0,
+                                            bool centerPrincipalPoint = false);
 
 //! converts point coordinates from normal pixel coordinates to homogeneous coordinates ((x,y)->(x,y,1))
 CV_EXPORTS_W void convertPointsToHomogeneous( InputArray src, OutputArray dst );
@@ -623,44 +250,36 @@ CV_EXPORTS_W void convertPointsFromHomogeneous( InputArray src, OutputArray dst
 //! for backward compatibility
 CV_EXPORTS void convertPointsHomogeneous( InputArray src, OutputArray dst );
 
-//! the algorithm for finding fundamental matrix
-enum
-{
-    FM_7POINT = CV_FM_7POINT, //!< 7-point algorithm
-    FM_8POINT = CV_FM_8POINT, //!< 8-point algorithm
-    FM_LMEDS = CV_FM_LMEDS,  //!< least-median algorithm
-    FM_RANSAC = CV_FM_RANSAC  //!< RANSAC algorithm
-};
-
 //! finds fundamental matrix from a set of corresponding 2D points
 CV_EXPORTS_W Mat findFundamentalMat( InputArray points1, InputArray points2,
-                                     int method=FM_RANSAC,
-                                     double param1=3., double param2=0.99,
-                                     OutputArray mask=noArray());
+                                     int method = FM_RANSAC,
+                                     double param1 = 3., double param2 = 0.99,
+                                     OutputArray mask = noArray() );
 
 //! variant of findFundamentalMat for backward compatibility
 CV_EXPORTS Mat findFundamentalMat( InputArray points1, InputArray points2,
-                                   OutputArray mask, int method=FM_RANSAC,
-                                   double param1=3., double param2=0.99);
+                                   OutputArray mask, int method = FM_RANSAC,
+                                   double param1 = 3., double param2 = 0.99 );
 
 //! finds essential matrix from a set of corresponding 2D points using five-point algorithm
-CV_EXPORTS Mat findEssentialMat( InputArray points1, InputArray points2, double focal = 1.0, Point2d pp = Point2d(0, 0),
-                                 int method = CV_RANSAC,
-                                 double prob = 0.999, double threshold = 1.0, OutputArray mask = noArray() );
+CV_EXPORTS_W Mat findEssentialMat( InputArray points1, InputArray points2,
+                                 double focal = 1.0, Point2d pp = Point2d(0, 0),
+                                 int method = RANSAC, double prob = 0.999,
+                                 double threshold = 1.0, OutputArray mask = noArray() );
 
 //! decompose essential matrix to possible rotation matrix and one translation vector
-CV_EXPORTS void decomposeEssentialMat( InputArray E, OutputArray R1, OutputArray R2, OutputArray t );
+CV_EXPORTS_W void decomposeEssentialMat( InputArray E, OutputArray R1, OutputArray R2, OutputArray t );
 
 //! recover relative camera pose from a set of corresponding 2D points
-CV_EXPORTS int recoverPose( InputArray E, InputArray points1, InputArray points2, OutputArray R, OutputArray t,
+CV_EXPORTS_W int recoverPose( InputArray E, InputArray points1, InputArray points2,
+                            OutputArray R, OutputArray t,
                             double focal = 1.0, Point2d pp = Point2d(0, 0),
-                            InputOutputArray mask = noArray());
+                            InputOutputArray mask = noArray() );
 
 
 //! finds coordinates of epipolar lines corresponding the specified points
-CV_EXPORTS void computeCorrespondEpilines( InputArray points,
-                                           int whichImage, InputArray F,
-                                           OutputArray lines );
+CV_EXPORTS_W void computeCorrespondEpilines( InputArray points, int whichImage,
+                                             InputArray F, OutputArray lines );
 
 CV_EXPORTS_W void triangulatePoints( InputArray projMatr1, InputArray projMatr2,
                                      InputArray projPoints1, InputArray projPoints2,
@@ -669,89 +288,10 @@ CV_EXPORTS_W void triangulatePoints( InputArray projMatr1, InputArray projMatr2,
 CV_EXPORTS_W void correctMatches( InputArray F, InputArray points1, InputArray points2,
                                   OutputArray newPoints1, OutputArray newPoints2 );
 
-
-class CV_EXPORTS_W StereoMatcher : public Algorithm
-{
-public:
-    CV_WRAP virtual void compute( InputArray left, InputArray right,
-                                  OutputArray disparity ) = 0;
-};
-
-enum { STEREO_DISP_SCALE=16, STEREO_PREFILTER_NORMALIZED_RESPONSE = 0, STEREO_PREFILTER_XSOBEL = 1 };
-
-CV_EXPORTS Ptr<StereoMatcher> createStereoBM(int numDisparities=0, int SADWindowSize=21);
-
-CV_EXPORTS Ptr<StereoMatcher> createStereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
-                                               int P1=0, int P2=0, int disp12MaxDiff=0,
-                                               int preFilterCap=0, int uniquenessRatio=0,
-                                               int speckleWindowSize=0, int speckleRange=0,
-                                               bool fullDP=false);
-
-template<> CV_EXPORTS void Ptr<CvStereoBMState>::delete_obj();
-
-// to be moved to "compat" module
-class CV_EXPORTS_W StereoBM
-{
-public:
-    enum { PREFILTER_NORMALIZED_RESPONSE = 0, PREFILTER_XSOBEL = 1,
-           BASIC_PRESET=0, FISH_EYE_PRESET=1, NARROW_PRESET=2 };
-
-    //! the default constructor
-    CV_WRAP StereoBM();
-    //! the full constructor taking the camera-specific preset, number of disparities and the SAD window size
-    CV_WRAP StereoBM(int preset, int ndisparities=0, int SADWindowSize=21);
-    //! the method that reinitializes the state. The previous content is destroyed
-    void init(int preset, int ndisparities=0, int SADWindowSize=21);
-    //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair
-    CV_WRAP_AS(compute) void operator()( InputArray left, InputArray right,
-                                         OutputArray disparity, int disptype=CV_16S );
-
-    //! pointer to the underlying CvStereoBMState
-    Ptr<CvStereoBMState> state;
-};
-
-
-// to be moved to "compat" module
-class CV_EXPORTS_W StereoSGBM
-{
-public:
-    enum { DISP_SHIFT=4, DISP_SCALE = (1<<DISP_SHIFT) };
-
-    //! the default constructor
-    CV_WRAP StereoSGBM();
-
-    //! the full constructor taking all the necessary algorithm parameters
-    CV_WRAP StereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
-               int P1=0, int P2=0, int disp12MaxDiff=0,
-               int preFilterCap=0, int uniquenessRatio=0,
-               int speckleWindowSize=0, int speckleRange=0,
-               bool fullDP=false);
-    //! the destructor
-    virtual ~StereoSGBM();
-
-    //! the stereo correspondence operator that computes disparity map for the specified rectified stereo pair
-    CV_WRAP_AS(compute) virtual void operator()(InputArray left, InputArray right,
-                                                OutputArray disp);
-
-    CV_PROP_RW int minDisparity;
-    CV_PROP_RW int numberOfDisparities;
-    CV_PROP_RW int SADWindowSize;
-    CV_PROP_RW int preFilterCap;
-    CV_PROP_RW int uniquenessRatio;
-    CV_PROP_RW int P1;
-    CV_PROP_RW int P2;
-    CV_PROP_RW int speckleWindowSize;
-    CV_PROP_RW int speckleRange;
-    CV_PROP_RW int disp12MaxDiff;
-    CV_PROP_RW bool fullDP;
-
-protected:
-    Ptr<StereoMatcher> sm;
-};
-
 //! filters off speckles (small regions of incorrectly computed disparity)
-CV_EXPORTS_W void filterSpeckles( InputOutputArray img, double newVal, int maxSpeckleSize, double maxDiff,
-                                  InputOutputArray buf=noArray() );
+CV_EXPORTS_W void filterSpeckles( InputOutputArray img, double newVal,
+                                  int maxSpeckleSize, double maxDiff,
+                                  InputOutputArray buf = noArray() );
 
 //! computes valid disparity ROI from the valid ROIs of the rectified images (that are returned by cv::stereoRectify())
 CV_EXPORTS_W Rect getValidDisparityROI( Rect roi1, Rect roi2,
@@ -761,20 +301,116 @@ CV_EXPORTS_W Rect getValidDisparityROI( Rect roi1, Rect roi2,
 //! validates disparity using the left-right check. The matrix "cost" should be computed by the stereo correspondence algorithm
 CV_EXPORTS_W void validateDisparity( InputOutputArray disparity, InputArray cost,
                                      int minDisparity, int numberOfDisparities,
-                                     int disp12MaxDisp=1 );
+                                     int disp12MaxDisp = 1 );
 
 //! reprojects disparity image to 3D: (x,y,d)->(X,Y,Z) using the matrix Q returned by cv::stereoRectify
 CV_EXPORTS_W void reprojectImageTo3D( InputArray disparity,
                                       OutputArray _3dImage, InputArray Q,
-                                      bool handleMissingValues=false,
-                                      int ddepth=-1 );
+                                      bool handleMissingValues = false,
+                                      int ddepth = -1 );
 
 CV_EXPORTS_W  int estimateAffine3D(InputArray src, InputArray dst,
                                    OutputArray out, OutputArray inliers,
-                                   double ransacThreshold=3, double confidence=0.99);
+                                   double ransacThreshold = 3, double confidence = 0.99);
 
-}
-
-#endif
+
+
+class CV_EXPORTS_W StereoMatcher : public Algorithm
+{
+public:
+    enum { DISP_SHIFT = 4,
+           DISP_SCALE = (1 << DISP_SHIFT)
+         };
+
+    CV_WRAP virtual void compute( InputArray left, InputArray right,
+                                  OutputArray disparity ) = 0;
+
+    CV_WRAP virtual int getMinDisparity() const = 0;
+    CV_WRAP virtual void setMinDisparity(int minDisparity) = 0;
+
+    CV_WRAP virtual int getNumDisparities() const = 0;
+    CV_WRAP virtual void setNumDisparities(int numDisparities) = 0;
+
+    CV_WRAP virtual int getBlockSize() const = 0;
+    CV_WRAP virtual void setBlockSize(int blockSize) = 0;
+
+    CV_WRAP virtual int getSpeckleWindowSize() const = 0;
+    CV_WRAP virtual void setSpeckleWindowSize(int speckleWindowSize) = 0;
+
+    CV_WRAP virtual int getSpeckleRange() const = 0;
+    CV_WRAP virtual void setSpeckleRange(int speckleRange) = 0;
+
+    CV_WRAP virtual int getDisp12MaxDiff() const = 0;
+    CV_WRAP virtual void setDisp12MaxDiff(int disp12MaxDiff) = 0;
+};
+
+
+
+class CV_EXPORTS_W StereoBM : public StereoMatcher
+{
+public:
+    enum { PREFILTER_NORMALIZED_RESPONSE = 0,
+           PREFILTER_XSOBEL              = 1
+         };
+
+    CV_WRAP virtual int getPreFilterType() const = 0;
+    CV_WRAP virtual void setPreFilterType(int preFilterType) = 0;
+
+    CV_WRAP virtual int getPreFilterSize() const = 0;
+    CV_WRAP virtual void setPreFilterSize(int preFilterSize) = 0;
+
+    CV_WRAP virtual int getPreFilterCap() const = 0;
+    CV_WRAP virtual void setPreFilterCap(int preFilterCap) = 0;
+
+    CV_WRAP virtual int getTextureThreshold() const = 0;
+    CV_WRAP virtual void setTextureThreshold(int textureThreshold) = 0;
+
+    CV_WRAP virtual int getUniquenessRatio() const = 0;
+    CV_WRAP virtual void setUniquenessRatio(int uniquenessRatio) = 0;
+
+    CV_WRAP virtual int getSmallerBlockSize() const = 0;
+    CV_WRAP virtual void setSmallerBlockSize(int blockSize) = 0;
+
+    CV_WRAP virtual Rect getROI1() const = 0;
+    CV_WRAP virtual void setROI1(Rect roi1) = 0;
+
+    CV_WRAP virtual Rect getROI2() const = 0;
+    CV_WRAP virtual void setROI2(Rect roi2) = 0;
+};
+
+CV_EXPORTS_W Ptr<StereoBM> createStereoBM(int numDisparities = 0, int blockSize = 21);
+
+
+class CV_EXPORTS_W StereoSGBM : public StereoMatcher
+{
+public:
+    enum { MODE_SGBM = 0,
+           MODE_HH   = 1
+         };
+
+    CV_WRAP virtual int getPreFilterCap() const = 0;
+    CV_WRAP virtual void setPreFilterCap(int preFilterCap) = 0;
+
+    CV_WRAP virtual int getUniquenessRatio() const = 0;
+    CV_WRAP virtual void setUniquenessRatio(int uniquenessRatio) = 0;
+
+    CV_WRAP virtual int getP1() const = 0;
+    CV_WRAP virtual void setP1(int P1) = 0;
+
+    CV_WRAP virtual int getP2() const = 0;
+    CV_WRAP virtual void setP2(int P2) = 0;
+
+    CV_WRAP virtual int getMode() const = 0;
+    CV_WRAP virtual void setMode(int mode) = 0;
+};
+
+
+CV_EXPORTS_W Ptr<StereoSGBM> createStereoSGBM(int minDisparity, int numDisparities, int blockSize,
+                                            int P1 = 0, int P2 = 0, int disp12MaxDiff = 0,
+                                            int preFilterCap = 0, int uniquenessRatio = 0,
+                                            int speckleWindowSize = 0, int speckleRange = 0,
+                                            int mode = StereoSGBM::MODE_SGBM);
+
+} // cv
 
 #endif
diff --git a/modules/calib3d/include/opencv2/calib3d/calib3d.hpp b/modules/calib3d/include/opencv2/calib3d/calib3d.hpp
index 302de229bd..b3da45edd5 100644
--- a/modules/calib3d/include/opencv2/calib3d/calib3d.hpp
+++ b/modules/calib3d/include/opencv2/calib3d/calib3d.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/calib3d.hpp"
\ No newline at end of file
+#include "opencv2/calib3d.hpp"
diff --git a/modules/calib3d/include/opencv2/calib3d/calib3d_c.h b/modules/calib3d/include/opencv2/calib3d/calib3d_c.h
new file mode 100644
index 0000000000..a505d526db
--- /dev/null
+++ b/modules/calib3d/include/opencv2/calib3d/calib3d_c.h
@@ -0,0 +1,413 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CALIB3D_C_H__
+#define __OPENCV_CALIB3D_C_H__
+
+#include "opencv2/core/core_c.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/****************************************************************************************\
+*                      Camera Calibration, Pose Estimation and Stereo                    *
+\****************************************************************************************/
+
+typedef struct CvPOSITObject CvPOSITObject;
+
+/* Allocates and initializes CvPOSITObject structure before doing cvPOSIT */
+CVAPI(CvPOSITObject*)  cvCreatePOSITObject( CvPoint3D32f* points, int point_count );
+
+
+/* Runs POSIT (POSe from ITeration) algorithm for determining 3d position of
+   an object given its model and projection in a weak-perspective case */
+CVAPI(void)  cvPOSIT(  CvPOSITObject* posit_object, CvPoint2D32f* image_points,
+                       double focal_length, CvTermCriteria criteria,
+                       float* rotation_matrix, float* translation_vector);
+
+/* Releases CvPOSITObject structure */
+CVAPI(void)  cvReleasePOSITObject( CvPOSITObject**  posit_object );
+
+/* updates the number of RANSAC iterations */
+CVAPI(int) cvRANSACUpdateNumIters( double p, double err_prob,
+                                   int model_points, int max_iters );
+
+CVAPI(void) cvConvertPointsHomogeneous( const CvMat* src, CvMat* dst );
+
+/* Calculates fundamental matrix given a set of corresponding points */
+#define CV_FM_7POINT 1
+#define CV_FM_8POINT 2
+
+#define CV_LMEDS 4
+#define CV_RANSAC 8
+
+#define CV_FM_LMEDS_ONLY  CV_LMEDS
+#define CV_FM_RANSAC_ONLY CV_RANSAC
+#define CV_FM_LMEDS CV_LMEDS
+#define CV_FM_RANSAC CV_RANSAC
+
+enum
+{
+    CV_ITERATIVE = 0,
+    CV_EPNP = 1, // F.Moreno-Noguer, V.Lepetit and P.Fua "EPnP: Efficient Perspective-n-Point Camera Pose Estimation"
+    CV_P3P = 2 // X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang; "Complete Solution Classification for the Perspective-Three-Point Problem"
+};
+
+CVAPI(int) cvFindFundamentalMat( const CvMat* points1, const CvMat* points2,
+                                 CvMat* fundamental_matrix,
+                                 int method CV_DEFAULT(CV_FM_RANSAC),
+                                 double param1 CV_DEFAULT(3.), double param2 CV_DEFAULT(0.99),
+                                 CvMat* status CV_DEFAULT(NULL) );
+
+/* For each input point on one of images
+   computes parameters of the corresponding
+   epipolar line on the other image */
+CVAPI(void) cvComputeCorrespondEpilines( const CvMat* points,
+                                         int which_image,
+                                         const CvMat* fundamental_matrix,
+                                         CvMat* correspondent_lines );
+
+/* Triangulation functions */
+
+CVAPI(void) cvTriangulatePoints(CvMat* projMatr1, CvMat* projMatr2,
+                                CvMat* projPoints1, CvMat* projPoints2,
+                                CvMat* points4D);
+
+CVAPI(void) cvCorrectMatches(CvMat* F, CvMat* points1, CvMat* points2,
+                             CvMat* new_points1, CvMat* new_points2);
+
+
+/* Computes the optimal new camera matrix according to the free scaling parameter alpha:
+   alpha=0 - only valid pixels will be retained in the undistorted image
+   alpha=1 - all the source image pixels will be retained in the undistorted image
+*/
+CVAPI(void) cvGetOptimalNewCameraMatrix( const CvMat* camera_matrix,
+                                         const CvMat* dist_coeffs,
+                                         CvSize image_size, double alpha,
+                                         CvMat* new_camera_matrix,
+                                         CvSize new_imag_size CV_DEFAULT(cvSize(0,0)),
+                                         CvRect* valid_pixel_ROI CV_DEFAULT(0),
+                                         int center_principal_point CV_DEFAULT(0));
+
+/* Converts rotation vector to rotation matrix or vice versa */
+CVAPI(int) cvRodrigues2( const CvMat* src, CvMat* dst,
+                         CvMat* jacobian CV_DEFAULT(0) );
+
+/* Finds perspective transformation between the object plane and image (view) plane */
+CVAPI(int) cvFindHomography( const CvMat* src_points,
+                             const CvMat* dst_points,
+                             CvMat* homography,
+                             int method CV_DEFAULT(0),
+                             double ransacReprojThreshold CV_DEFAULT(3),
+                             CvMat* mask CV_DEFAULT(0));
+
+/* Computes RQ decomposition for 3x3 matrices */
+CVAPI(void) cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
+                           CvMat *matrixQx CV_DEFAULT(NULL),
+                           CvMat *matrixQy CV_DEFAULT(NULL),
+                           CvMat *matrixQz CV_DEFAULT(NULL),
+                           CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
+
+/* Computes projection matrix decomposition */
+CVAPI(void) cvDecomposeProjectionMatrix( const CvMat *projMatr, CvMat *calibMatr,
+                                         CvMat *rotMatr, CvMat *posVect,
+                                         CvMat *rotMatrX CV_DEFAULT(NULL),
+                                         CvMat *rotMatrY CV_DEFAULT(NULL),
+                                         CvMat *rotMatrZ CV_DEFAULT(NULL),
+                                         CvPoint3D64f *eulerAngles CV_DEFAULT(NULL));
+
+/* Computes d(AB)/dA and d(AB)/dB */
+CVAPI(void) cvCalcMatMulDeriv( const CvMat* A, const CvMat* B, CvMat* dABdA, CvMat* dABdB );
+
+/* Computes r3 = rodrigues(rodrigues(r2)*rodrigues(r1)),
+   t3 = rodrigues(r2)*t1 + t2 and the respective derivatives */
+CVAPI(void) cvComposeRT( const CvMat* _rvec1, const CvMat* _tvec1,
+                         const CvMat* _rvec2, const CvMat* _tvec2,
+                         CvMat* _rvec3, CvMat* _tvec3,
+                         CvMat* dr3dr1 CV_DEFAULT(0), CvMat* dr3dt1 CV_DEFAULT(0),
+                         CvMat* dr3dr2 CV_DEFAULT(0), CvMat* dr3dt2 CV_DEFAULT(0),
+                         CvMat* dt3dr1 CV_DEFAULT(0), CvMat* dt3dt1 CV_DEFAULT(0),
+                         CvMat* dt3dr2 CV_DEFAULT(0), CvMat* dt3dt2 CV_DEFAULT(0) );
+
+/* Projects object points to the view plane using
+   the specified extrinsic and intrinsic camera parameters */
+CVAPI(void) cvProjectPoints2( const CvMat* object_points, const CvMat* rotation_vector,
+                              const CvMat* translation_vector, const CvMat* camera_matrix,
+                              const CvMat* distortion_coeffs, CvMat* image_points,
+                              CvMat* dpdrot CV_DEFAULT(NULL), CvMat* dpdt CV_DEFAULT(NULL),
+                              CvMat* dpdf CV_DEFAULT(NULL), CvMat* dpdc CV_DEFAULT(NULL),
+                              CvMat* dpddist CV_DEFAULT(NULL),
+                              double aspect_ratio CV_DEFAULT(0));
+
+/* Finds extrinsic camera parameters from
+   a few known corresponding point pairs and intrinsic parameters */
+CVAPI(void) cvFindExtrinsicCameraParams2( const CvMat* object_points,
+                                          const CvMat* image_points,
+                                          const CvMat* camera_matrix,
+                                          const CvMat* distortion_coeffs,
+                                          CvMat* rotation_vector,
+                                          CvMat* translation_vector,
+                                          int use_extrinsic_guess CV_DEFAULT(0) );
+
+/* Computes initial estimate of the intrinsic camera parameters
+   in case of planar calibration target (e.g. chessboard) */
+CVAPI(void) cvInitIntrinsicParams2D( const CvMat* object_points,
+                                     const CvMat* image_points,
+                                     const CvMat* npoints, CvSize image_size,
+                                     CvMat* camera_matrix,
+                                     double aspect_ratio CV_DEFAULT(1.) );
+
+#define CV_CALIB_CB_ADAPTIVE_THRESH  1
+#define CV_CALIB_CB_NORMALIZE_IMAGE  2
+#define CV_CALIB_CB_FILTER_QUADS     4
+#define CV_CALIB_CB_FAST_CHECK       8
+
+// Performs a fast check if a chessboard is in the input image. This is a workaround to
+// a problem of cvFindChessboardCorners being slow on images with no chessboard
+// - src: input image
+// - size: chessboard size
+// Returns 1 if a chessboard can be in this image and findChessboardCorners should be called,
+// 0 if there is no chessboard, -1 in case of error
+CVAPI(int) cvCheckChessboard(IplImage* src, CvSize size);
+
+    /* Detects corners on a chessboard calibration pattern */
+CVAPI(int) cvFindChessboardCorners( const void* image, CvSize pattern_size,
+                                    CvPoint2D32f* corners,
+                                    int* corner_count CV_DEFAULT(NULL),
+                                    int flags CV_DEFAULT(CV_CALIB_CB_ADAPTIVE_THRESH+CV_CALIB_CB_NORMALIZE_IMAGE) );
+
+/* Draws individual chessboard corners or the whole chessboard detected */
+CVAPI(void) cvDrawChessboardCorners( CvArr* image, CvSize pattern_size,
+                                     CvPoint2D32f* corners,
+                                     int count, int pattern_was_found );
+
+#define CV_CALIB_USE_INTRINSIC_GUESS  1
+#define CV_CALIB_FIX_ASPECT_RATIO     2
+#define CV_CALIB_FIX_PRINCIPAL_POINT  4
+#define CV_CALIB_ZERO_TANGENT_DIST    8
+#define CV_CALIB_FIX_FOCAL_LENGTH 16
+#define CV_CALIB_FIX_K1  32
+#define CV_CALIB_FIX_K2  64
+#define CV_CALIB_FIX_K3  128
+#define CV_CALIB_FIX_K4  2048
+#define CV_CALIB_FIX_K5  4096
+#define CV_CALIB_FIX_K6  8192
+#define CV_CALIB_RATIONAL_MODEL 16384
+#define CV_CALIB_THIN_PRISM_MODEL 32768
+#define CV_CALIB_FIX_S1_S2_S3_S4  65536
+
+
+/* Finds intrinsic and extrinsic camera parameters
+   from a few views of known calibration pattern */
+CVAPI(double) cvCalibrateCamera2( const CvMat* object_points,
+                                const CvMat* image_points,
+                                const CvMat* point_counts,
+                                CvSize image_size,
+                                CvMat* camera_matrix,
+                                CvMat* distortion_coeffs,
+                                CvMat* rotation_vectors CV_DEFAULT(NULL),
+                                CvMat* translation_vectors CV_DEFAULT(NULL),
+                                int flags CV_DEFAULT(0),
+                                CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
+                                    CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,DBL_EPSILON)) );
+
+/* Computes various useful characteristics of the camera from the data computed by
+   cvCalibrateCamera2 */
+CVAPI(void) cvCalibrationMatrixValues( const CvMat *camera_matrix,
+                                CvSize image_size,
+                                double aperture_width CV_DEFAULT(0),
+                                double aperture_height CV_DEFAULT(0),
+                                double *fovx CV_DEFAULT(NULL),
+                                double *fovy CV_DEFAULT(NULL),
+                                double *focal_length CV_DEFAULT(NULL),
+                                CvPoint2D64f *principal_point CV_DEFAULT(NULL),
+                                double *pixel_aspect_ratio CV_DEFAULT(NULL));
+
+#define CV_CALIB_FIX_INTRINSIC  256
+#define CV_CALIB_SAME_FOCAL_LENGTH 512
+
+/* Computes the transformation from one camera coordinate system to another one
+   from a few correspondent views of the same calibration target. Optionally, calibrates
+   both cameras */
+CVAPI(double) cvStereoCalibrate( const CvMat* object_points, const CvMat* image_points1,
+                               const CvMat* image_points2, const CvMat* npoints,
+                               CvMat* camera_matrix1, CvMat* dist_coeffs1,
+                               CvMat* camera_matrix2, CvMat* dist_coeffs2,
+                               CvSize image_size, CvMat* R, CvMat* T,
+                               CvMat* E CV_DEFAULT(0), CvMat* F CV_DEFAULT(0),
+                               CvTermCriteria term_crit CV_DEFAULT(cvTermCriteria(
+                                   CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,1e-6)),
+                               int flags CV_DEFAULT(CV_CALIB_FIX_INTRINSIC));
+
+#define CV_CALIB_ZERO_DISPARITY 1024
+
+/* Computes 3D rotations (+ optional shift) for each camera coordinate system to make both
+   views parallel (=> to make all the epipolar lines horizontal or vertical) */
+CVAPI(void) cvStereoRectify( const CvMat* camera_matrix1, const CvMat* camera_matrix2,
+                             const CvMat* dist_coeffs1, const CvMat* dist_coeffs2,
+                             CvSize image_size, const CvMat* R, const CvMat* T,
+                             CvMat* R1, CvMat* R2, CvMat* P1, CvMat* P2,
+                             CvMat* Q CV_DEFAULT(0),
+                             int flags CV_DEFAULT(CV_CALIB_ZERO_DISPARITY),
+                             double alpha CV_DEFAULT(-1),
+                             CvSize new_image_size CV_DEFAULT(cvSize(0,0)),
+                             CvRect* valid_pix_ROI1 CV_DEFAULT(0),
+                             CvRect* valid_pix_ROI2 CV_DEFAULT(0));
+
+/* Computes rectification transformations for uncalibrated pair of images using a set
+   of point correspondences */
+CVAPI(int) cvStereoRectifyUncalibrated( const CvMat* points1, const CvMat* points2,
+                                        const CvMat* F, CvSize img_size,
+                                        CvMat* H1, CvMat* H2,
+                                        double threshold CV_DEFAULT(5));
+
+
+
+/* stereo correspondence parameters and functions */
+
+#define CV_STEREO_BM_NORMALIZED_RESPONSE  0
+#define CV_STEREO_BM_XSOBEL               1
+
+/* Block matching algorithm structure */
+typedef struct CvStereoBMState
+{
+    // pre-filtering (normalization of input images)
+    int preFilterType; // =CV_STEREO_BM_NORMALIZED_RESPONSE now
+    int preFilterSize; // averaging window size: ~5x5..21x21
+    int preFilterCap; // the output of pre-filtering is clipped by [-preFilterCap,preFilterCap]
+
+    // correspondence using Sum of Absolute Difference (SAD)
+    int SADWindowSize; // ~5x5..21x21
+    int minDisparity;  // minimum disparity (can be negative)
+    int numberOfDisparities; // maximum disparity - minimum disparity (> 0)
+
+    // post-filtering
+    int textureThreshold;  // the disparity is only computed for pixels
+                           // with textured enough neighborhood
+    int uniquenessRatio;   // accept the computed disparity d* only if
+                           // SAD(d) >= SAD(d*)*(1 + uniquenessRatio/100.)
+                           // for any d != d*+/-1 within the search range.
+    int speckleWindowSize; // disparity variation window
+    int speckleRange; // acceptable range of variation in window
+
+    int trySmallerWindows; // if 1, the results may be more accurate,
+                           // at the expense of slower processing
+    CvRect roi1, roi2;
+    int disp12MaxDiff;
+
+    // temporary buffers
+    CvMat* preFilteredImg0;
+    CvMat* preFilteredImg1;
+    CvMat* slidingSumBuf;
+    CvMat* cost;
+    CvMat* disp;
+} CvStereoBMState;
+
+#define CV_STEREO_BM_BASIC 0
+#define CV_STEREO_BM_FISH_EYE 1
+#define CV_STEREO_BM_NARROW 2
+
+CVAPI(CvStereoBMState*) cvCreateStereoBMState(int preset CV_DEFAULT(CV_STEREO_BM_BASIC),
+                                              int numberOfDisparities CV_DEFAULT(0));
+
+CVAPI(void) cvReleaseStereoBMState( CvStereoBMState** state );
+
+CVAPI(void) cvFindStereoCorrespondenceBM( const CvArr* left, const CvArr* right,
+                                          CvArr* disparity, CvStereoBMState* state );
+
+CVAPI(CvRect) cvGetValidDisparityROI( CvRect roi1, CvRect roi2, int minDisparity,
+                                      int numberOfDisparities, int SADWindowSize );
+
+CVAPI(void) cvValidateDisparity( CvArr* disparity, const CvArr* cost,
+                                 int minDisparity, int numberOfDisparities,
+                                 int disp12MaxDiff CV_DEFAULT(1) );
+
+/* Reprojects the computed disparity image to the 3D space using the specified 4x4 matrix */
+CVAPI(void)  cvReprojectImageTo3D( const CvArr* disparityImage,
+                                   CvArr* _3dImage, const CvMat* Q,
+                                   int handleMissingValues CV_DEFAULT(0) );
+
+#ifdef __cplusplus
+} // extern "C"
+
+//////////////////////////////////////////////////////////////////////////////////////////
+class CV_EXPORTS CvLevMarq
+{
+public:
+    CvLevMarq();
+    CvLevMarq( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    ~CvLevMarq();
+    void init( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    bool update( const CvMat*& param, CvMat*& J, CvMat*& err );
+    bool updateAlt( const CvMat*& param, CvMat*& JtJ, CvMat*& JtErr, double*& errNorm );
+
+    void clear();
+    void step();
+    enum { DONE=0, STARTED=1, CALC_J=2, CHECK_ERR=3 };
+
+    cv::Ptr<CvMat> mask;
+    cv::Ptr<CvMat> prevParam;
+    cv::Ptr<CvMat> param;
+    cv::Ptr<CvMat> J;
+    cv::Ptr<CvMat> err;
+    cv::Ptr<CvMat> JtJ;
+    cv::Ptr<CvMat> JtJN;
+    cv::Ptr<CvMat> JtErr;
+    cv::Ptr<CvMat> JtJV;
+    cv::Ptr<CvMat> JtJW;
+    double prevErrNorm, errNorm;
+    int lambdaLg10;
+    CvTermCriteria criteria;
+    int state;
+    int iters;
+    bool completeSymmFlag;
+};
+
+#endif
+
+#endif /* __OPENCV_CALIB3D_C_H__ */
diff --git a/modules/calib3d/perf/perf_pnp.cpp b/modules/calib3d/perf/perf_pnp.cpp
index e0ffd70cfb..7a7acb052d 100644
--- a/modules/calib3d/perf/perf_pnp.cpp
+++ b/modules/calib3d/perf/perf_pnp.cpp
@@ -1,5 +1,8 @@
 #include "perf_precomp.hpp"
-#include "opencv2/core/internal.hpp"
+
+#ifdef HAVE_TBB
+#include "tbb/task_scheduler_init.h"
+#endif
 
 using namespace std;
 using namespace cv;
@@ -7,7 +10,7 @@ using namespace perf;
 using std::tr1::make_tuple;
 using std::tr1::get;
 
-CV_ENUM(pnpAlgo, CV_ITERATIVE, CV_EPNP /*, CV_P3P*/)
+CV_ENUM(pnpAlgo, ITERATIVE, EPNP /*, P3P*/)
 
 typedef std::tr1::tuple<int, pnpAlgo> PointsNum_Algo_t;
 typedef perf::TestBaseWithParam<PointsNum_Algo_t> PointsNum_Algo;
@@ -17,7 +20,7 @@ typedef perf::TestBaseWithParam<int> PointsNum;
 PERF_TEST_P(PointsNum_Algo, solvePnP,
             testing::Combine(
                 testing::Values(/*4,*/ 3*9, 7*13), //TODO: find why results on 4 points are too unstable
-                testing::Values((int)CV_ITERATIVE, (int)CV_EPNP)
+                testing::Values((int)ITERATIVE, (int)EPNP)
                 )
             )
 {
@@ -90,7 +93,7 @@ PERF_TEST(PointsNum_Algo, solveP3P)
 
     TEST_CYCLE_N(1000)
     {
-        solvePnP(points3d, points2d, intrinsics, distortion, rvec, tvec, false, CV_P3P);
+        solvePnP(points3d, points2d, intrinsics, distortion, rvec, tvec, false, P3P);
     }
 
     SANITY_CHECK(rvec, 1e-6);
@@ -127,7 +130,7 @@ PERF_TEST_P(PointsNum, DISABLED_SolvePnPRansac, testing::Values(4, 3*9, 7*13))
 
 #ifdef HAVE_TBB
     // limit concurrency to get determenistic result
-    cv::Ptr<tbb::task_scheduler_init> one_thread = new tbb::task_scheduler_init(1);
+    tbb::task_scheduler_init one_thread(1);
 #endif
 
     TEST_CYCLE()
diff --git a/modules/calib3d/perf/perf_precomp.cpp b/modules/calib3d/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/calib3d/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/calib3d/src/calib3d_init.cpp b/modules/calib3d/src/calib3d_init.cpp
index 06303bd985..6192c7998a 100644
--- a/modules/calib3d/src/calib3d_init.cpp
+++ b/modules/calib3d/src/calib3d_init.cpp
@@ -47,7 +47,7 @@ using namespace cv;
 //////////////////////////////////////////////////////////////////////////////////////////////////////////
 
 
-//////////////////////////////////////////////////////////////////////////////////////////////////////////                  
+//////////////////////////////////////////////////////////////////////////////////////////////////////////
 
 
 
diff --git a/modules/calib3d/src/calibinit.cpp b/modules/calib3d/src/calibinit.cpp
index 0a349584e7..844fde4999 100644
--- a/modules/calib3d/src/calibinit.cpp
+++ b/modules/calib3d/src/calibinit.cpp
@@ -60,6 +60,8 @@
 \************************************************************************************/
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/calib3d/calib3d_c.h"
 #include "circlesgrid.hpp"
 #include <stdarg.h>
 
@@ -269,8 +271,8 @@ int cvFindChessboardCorners( const void* arr, CvSize pattern_size,
     if( !out_corners )
         CV_Error( CV_StsNullPtr, "Null pointer to corners" );
 
-    storage = cvCreateMemStorage(0);
-    thresh_img = cvCreateMat( img->rows, img->cols, CV_8UC1 );
+    storage.reset(cvCreateMemStorage(0));
+    thresh_img.reset(cvCreateMat( img->rows, img->cols, CV_8UC1 ));
 
 #ifdef DEBUG_CHESSBOARD
     dbg_img = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 3 );
@@ -282,7 +284,7 @@ int cvFindChessboardCorners( const void* arr, CvSize pattern_size,
     {
         // equalize the input image histogram -
         // that should make the contrast between "black" and "white" areas big enough
-        norm_img = cvCreateMat( img->rows, img->cols, CV_8UC1 );
+        norm_img.reset(cvCreateMat( img->rows, img->cols, CV_8UC1 ));
 
         if( CV_MAT_CN(img->type) != 1 )
         {
@@ -539,12 +541,12 @@ int cvFindChessboardCorners( const void* arr, CvSize pattern_size,
         cv::Ptr<CvMat> gray;
         if( CV_MAT_CN(img->type) != 1 )
         {
-            gray = cvCreateMat(img->rows, img->cols, CV_8UC1);
+            gray.reset(cvCreateMat(img->rows, img->cols, CV_8UC1));
             cvCvtColor(img, gray, CV_BGR2GRAY);
         }
         else
         {
-            gray = cvCloneMat(img);
+            gray.reset(cvCloneMat(img));
         }
         int wsize = 2;
         cvFindCornerSubPix( gray, out_corners, pattern_size.width*pattern_size.height,
@@ -625,7 +627,7 @@ icvOrderFoundConnectedQuads( int quad_count, CvCBQuad **quads,
         int *all_count, CvCBQuad **all_quads, CvCBCorner **corners,
         CvSize pattern_size, CvMemStorage* storage )
 {
-    cv::Ptr<CvMemStorage> temp_storage = cvCreateChildMemStorage( storage );
+    cv::Ptr<CvMemStorage> temp_storage(cvCreateChildMemStorage( storage ));
     CvSeq* stack = cvCreateSeq( 0, sizeof(*stack), sizeof(void*), temp_storage );
 
     // first find an interior quad
@@ -1095,7 +1097,7 @@ icvOrderQuad(CvCBQuad *quad, CvCBCorner *corner, int common)
 static int
 icvCleanFoundConnectedQuads( int quad_count, CvCBQuad **quad_group, CvSize pattern_size )
 {
-    CvPoint2D32f center = {0,0};
+    CvPoint2D32f center;
     int i, j, k;
     // number of quads this pattern should contain
     int count = ((pattern_size.width + 1)*(pattern_size.height + 1) + 1)/2;
@@ -1107,11 +1109,11 @@ icvCleanFoundConnectedQuads( int quad_count, CvCBQuad **quad_group, CvSize patte
 
     // create an array of quadrangle centers
     cv::AutoBuffer<CvPoint2D32f> centers( quad_count );
-    cv::Ptr<CvMemStorage> temp_storage = cvCreateMemStorage(0);
+    cv::Ptr<CvMemStorage> temp_storage(cvCreateMemStorage(0));
 
     for( i = 0; i < quad_count; i++ )
     {
-        CvPoint2D32f ci = {0,0};
+        CvPoint2D32f ci;
         CvCBQuad* q = quad_group[i];
 
         for( j = 0; j < 4; j++ )
@@ -1203,7 +1205,7 @@ static int
 icvFindConnectedQuads( CvCBQuad *quad, int quad_count, CvCBQuad **out_group,
                        int group_idx, CvMemStorage* storage )
 {
-    cv::Ptr<CvMemStorage> temp_storage = cvCreateChildMemStorage( storage );
+    cv::Ptr<CvMemStorage> temp_storage(cvCreateChildMemStorage( storage ));
     CvSeq* stack = cvCreateSeq( 0, sizeof(*stack), sizeof(void*), temp_storage );
     int i, count = 0;
 
@@ -1672,7 +1674,7 @@ icvGenerateQuads( CvCBQuad **out_quads, CvCBCorner **out_corners,
     min_size = 25; //cvRound( image->cols * image->rows * .03 * 0.01 * 0.92 );
 
     // create temporary storage for contours and the sequence of pointers to found quadrangles
-    temp_storage = cvCreateChildMemStorage( storage );
+    temp_storage.reset(cvCreateChildMemStorage( storage ));
     root = cvCreateSeq( 0, sizeof(CvSeq), sizeof(CvSeq*), temp_storage );
 
     // initialize contour retrieving routine
@@ -1833,7 +1835,7 @@ cvDrawChessboardCorners( CvArr* _image, CvSize pattern_size,
 
     if( !found )
     {
-        CvScalar color = {{0,0,255}};
+        CvScalar color(0,0,255,0);
         if( cn == 1 )
             color = cvScalarAll(200);
         color.val[0] *= scale;
@@ -1856,17 +1858,17 @@ cvDrawChessboardCorners( CvArr* _image, CvSize pattern_size,
     else
     {
         int x, y;
-        CvPoint prev_pt = {0, 0};
+        CvPoint prev_pt;
         const int line_max = 7;
         static const CvScalar line_colors[line_max] =
         {
-            {{0,0,255}},
-            {{0,128,255}},
-            {{0,200,200}},
-            {{0,255,0}},
-            {{200,200,0}},
-            {{255,0,0}},
-            {{255,0,255}}
+            CvScalar(0,0,255),
+            CvScalar(0,128,255),
+            CvScalar(0,200,200),
+            CvScalar(0,255,0),
+            CvScalar(200,200,0),
+            CvScalar(255,0,0),
+            CvScalar(255,0,255)
         };
 
         for( y = 0, i = 0; y < pattern_size.height; y++ )
diff --git a/modules/calib3d/src/calibration.cpp b/modules/calib3d/src/calibration.cpp
index e36c1e5ede..ff5e425b1b 100644
--- a/modules/calib3d/src/calibration.cpp
+++ b/modules/calib3d/src/calibration.cpp
@@ -41,6 +41,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/calib3d/calib3d_c.h"
 #include <stdio.h>
 #include <iterator>
 
@@ -566,7 +568,7 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
         (objectPoints->rows == count && CV_MAT_CN(objectPoints->type)*objectPoints->cols == 3) ||
         (objectPoints->rows == 3 && CV_MAT_CN(objectPoints->type) == 1 && objectPoints->cols == count)))
     {
-        matM = cvCreateMat( objectPoints->rows, objectPoints->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(objectPoints->type)) );
+        matM.reset(cvCreateMat( objectPoints->rows, objectPoints->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(objectPoints->type)) ));
         cvConvert(objectPoints, matM);
     }
     else
@@ -582,7 +584,7 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
         (imagePoints->rows == count && CV_MAT_CN(imagePoints->type)*imagePoints->cols == 2) ||
         (imagePoints->rows == 2 && CV_MAT_CN(imagePoints->type) == 1 && imagePoints->cols == count)))
     {
-        _m = cvCreateMat( imagePoints->rows, imagePoints->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(imagePoints->type)) );
+        _m.reset(cvCreateMat( imagePoints->rows, imagePoints->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(imagePoints->type)) ));
         cvConvert(imagePoints, _m);
     }
     else
@@ -662,10 +664,10 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
 
         if( CV_MAT_TYPE(dpdr->type) == CV_64FC1 )
         {
-            _dpdr = cvCloneMat(dpdr);
+            _dpdr.reset(cvCloneMat(dpdr));
         }
         else
-            _dpdr = cvCreateMat( 2*count, 3, CV_64FC1 );
+            _dpdr.reset(cvCreateMat( 2*count, 3, CV_64FC1 ));
         dpdr_p = _dpdr->data.db;
         dpdr_step = _dpdr->step/sizeof(dpdr_p[0]);
     }
@@ -680,10 +682,10 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
 
         if( CV_MAT_TYPE(dpdt->type) == CV_64FC1 )
         {
-            _dpdt = cvCloneMat(dpdt);
+            _dpdt.reset(cvCloneMat(dpdt));
         }
         else
-            _dpdt = cvCreateMat( 2*count, 3, CV_64FC1 );
+            _dpdt.reset(cvCreateMat( 2*count, 3, CV_64FC1 ));
         dpdt_p = _dpdt->data.db;
         dpdt_step = _dpdt->step/sizeof(dpdt_p[0]);
     }
@@ -697,10 +699,10 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
 
         if( CV_MAT_TYPE(dpdf->type) == CV_64FC1 )
         {
-            _dpdf = cvCloneMat(dpdf);
+            _dpdf.reset(cvCloneMat(dpdf));
         }
         else
-            _dpdf = cvCreateMat( 2*count, 2, CV_64FC1 );
+            _dpdf.reset(cvCreateMat( 2*count, 2, CV_64FC1 ));
         dpdf_p = _dpdf->data.db;
         dpdf_step = _dpdf->step/sizeof(dpdf_p[0]);
     }
@@ -714,10 +716,10 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
 
         if( CV_MAT_TYPE(dpdc->type) == CV_64FC1 )
         {
-            _dpdc = cvCloneMat(dpdc);
+            _dpdc.reset(cvCloneMat(dpdc));
         }
         else
-            _dpdc = cvCreateMat( 2*count, 2, CV_64FC1 );
+            _dpdc.reset(cvCreateMat( 2*count, 2, CV_64FC1 ));
         dpdc_p = _dpdc->data.db;
         dpdc_step = _dpdc->step/sizeof(dpdc_p[0]);
     }
@@ -734,10 +736,10 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
 
         if( CV_MAT_TYPE(dpdk->type) == CV_64FC1 )
         {
-            _dpdk = cvCloneMat(dpdk);
+            _dpdk.reset(cvCloneMat(dpdk));
         }
         else
-            _dpdk = cvCreateMat( dpdk->rows, dpdk->cols, CV_64FC1 );
+            _dpdk.reset(cvCreateMat( dpdk->rows, dpdk->cols, CV_64FC1 ));
         dpdk_p = _dpdk->data.db;
         dpdk_step = _dpdk->step/sizeof(dpdk_p[0]);
     }
@@ -824,7 +826,7 @@ CV_IMPL void cvProjectPoints2( const CvMat* objectPoints,
                             dpdk_p[dpdk_step+7] = fy*y*cdist*(-icdist2)*icdist2*r6;
                             if( _dpdk->cols > 8 )
                             {
-                                dpdk_p[8] = fx*r2; //s1 
+                                dpdk_p[8] = fx*r2; //s1
                                 dpdk_p[9] = fx*r4; //s2
                                 dpdk_p[10] = 0;//s3
                                 dpdk_p[11] = 0;//s4
@@ -948,8 +950,8 @@ CV_IMPL void cvFindExtrinsicCameraParams2( const CvMat* objectPoints,
         CV_IS_MAT(A) && CV_IS_MAT(rvec) && CV_IS_MAT(tvec) );
 
     count = MAX(objectPoints->cols, objectPoints->rows);
-    matM = cvCreateMat( 1, count, CV_64FC3 );
-    _m = cvCreateMat( 1, count, CV_64FC2 );
+    matM.reset(cvCreateMat( 1, count, CV_64FC3 ));
+    _m.reset(cvCreateMat( 1, count, CV_64FC2 ));
 
     cvConvertPointsHomogeneous( objectPoints, matM );
     cvConvertPointsHomogeneous( imagePoints, _m );
@@ -961,8 +963,8 @@ CV_IMPL void cvFindExtrinsicCameraParams2( const CvMat* objectPoints,
     CV_Assert( (CV_MAT_DEPTH(tvec->type) == CV_64F || CV_MAT_DEPTH(tvec->type) == CV_32F) &&
         (tvec->rows == 1 || tvec->cols == 1) && tvec->rows*tvec->cols*CV_MAT_CN(tvec->type) == 3 );
 
-    _mn = cvCreateMat( 1, count, CV_64FC2 );
-    _Mxy = cvCreateMat( 1, count, CV_64FC2 );
+    _mn.reset(cvCreateMat( 1, count, CV_64FC2 ));
+    _Mxy.reset(cvCreateMat( 1, count, CV_64FC2 ));
 
     // normalize image points
     // (unapply the intrinsic matrix transformation and distortion)
@@ -1053,7 +1055,7 @@ CV_IMPL void cvFindExtrinsicCameraParams2( const CvMat* objectPoints,
             CvPoint3D64f* M = (CvPoint3D64f*)matM->data.db;
             CvPoint2D64f* mn = (CvPoint2D64f*)_mn->data.db;
 
-            matL = cvCreateMat( 2*count, 12, CV_64F );
+            matL.reset(cvCreateMat( 2*count, 12, CV_64F ));
             L = matL->data.db;
 
             for( i = 0; i < count; i++, L += 24 )
@@ -1160,11 +1162,11 @@ CV_IMPL void cvInitIntrinsicParams2D( const CvMat* objectPoints,
     if( objectPoints->rows != 1 || imagePoints->rows != 1 )
         CV_Error( CV_StsBadSize, "object points and image points must be a single-row matrices" );
 
-    matA = cvCreateMat( 2*nimages, 2, CV_64F );
-    _b = cvCreateMat( 2*nimages, 1, CV_64F );
+    matA.reset(cvCreateMat( 2*nimages, 2, CV_64F ));
+    _b.reset(cvCreateMat( 2*nimages, 1, CV_64F ));
     a[2] = (imageSize.width - 1)*0.5;
     a[5] = (imageSize.height - 1)*0.5;
-    _allH = cvCreateMat( nimages, 9, CV_64F );
+    _allH.reset(cvCreateMat( nimages, 9, CV_64F ));
 
     // extract vanishing points in order to obtain initial value for the focal length
     for( i = 0, pos = 0; i < nimages; i++, pos += ni )
@@ -1254,7 +1256,7 @@ CV_IMPL double cvCalibrateCamera2( const CvMat* objectPoints,
     //when the thin prism model is used the distortion coefficients matrix must have 12 parameters
     if((flags & CV_CALIB_THIN_PRISM_MODEL) && (distCoeffs->cols*distCoeffs->rows != 12))
         CV_Error( CV_StsBadArg, "Thin prism model must have 12 parameters in the distortion matrix" );
-    
+
     nimages = npoints->rows*npoints->cols;
     npstep = npoints->rows == 1 ? 1 : npoints->step/CV_ELEM_SIZE(npoints->type);
 
@@ -1308,16 +1310,16 @@ CV_IMPL double cvCalibrateCamera2( const CvMat* objectPoints,
         total += ni;
     }
 
-    matM = cvCreateMat( 1, total, CV_64FC3 );
-    _m = cvCreateMat( 1, total, CV_64FC2 );
+    matM.reset(cvCreateMat( 1, total, CV_64FC3 ));
+    _m.reset(cvCreateMat( 1, total, CV_64FC2 ));
 
     cvConvertPointsHomogeneous( objectPoints, matM );
     cvConvertPointsHomogeneous( imagePoints, _m );
 
     nparams = NINTRINSIC + nimages*6;
-    _Ji = cvCreateMat( maxPoints*2, NINTRINSIC, CV_64FC1 );
-    _Je = cvCreateMat( maxPoints*2, 6, CV_64FC1 );
-    _err = cvCreateMat( maxPoints*2, 1, CV_64FC1 );
+    _Ji.reset(cvCreateMat( maxPoints*2, NINTRINSIC, CV_64FC1 ));
+    _Je.reset(cvCreateMat( maxPoints*2, 6, CV_64FC1 ));
+    _err.reset(cvCreateMat( maxPoints*2, 1, CV_64FC1 ));
     cvZero( _Ji );
 
     _k = cvMat( distCoeffs->rows, distCoeffs->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(distCoeffs->type)), k);
@@ -1636,12 +1638,12 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
                         CvTermCriteria termCrit,
                         int flags )
 {
-    const int NINTRINSIC = 12;
+    const int NINTRINSIC = 16;
     Ptr<CvMat> npoints, err, J_LR, Je, Ji, imagePoints[2], objectPoints, RT0;
     CvLevMarq solver;
     double reprojErr = 0;
 
-    double A[2][9], dk[2][8]={{0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0}}, rlr[9];
+    double A[2][9], dk[2][12]={{0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0}}, rlr[9];
     CvMat K[2], Dist[2], om_LR, T_LR;
     CvMat R_LR = cvMat(3, 3, CV_64F, rlr);
     int i, k, p, ni = 0, ofs, nimages, pointsTotal, maxPoints = 0;
@@ -1660,7 +1662,7 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
                CV_MAT_TYPE(_npoints->type) == CV_32SC1 );
 
     nimages = _npoints->cols + _npoints->rows - 1;
-    npoints = cvCreateMat( _npoints->rows, _npoints->cols, _npoints->type );
+    npoints.reset(cvCreateMat( _npoints->rows, _npoints->cols, _npoints->type ));
     cvCopy( _npoints, npoints );
 
     for( i = 0, pointsTotal = 0; i < nimages; i++ )
@@ -1669,8 +1671,8 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
         pointsTotal += npoints->data.i[i];
     }
 
-    objectPoints = cvCreateMat( _objectPoints->rows, _objectPoints->cols,
-                                CV_64FC(CV_MAT_CN(_objectPoints->type)));
+    objectPoints.reset(cvCreateMat( _objectPoints->rows, _objectPoints->cols,
+                                    CV_64FC(CV_MAT_CN(_objectPoints->type))));
     cvConvert( _objectPoints, objectPoints );
     cvReshape( objectPoints, objectPoints, 3, 1 );
 
@@ -1687,9 +1689,9 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
                 (_imagePoints1->rows == 1 && _imagePoints1->cols == pointsTotal && cn == 2)) );
 
         K[k] = cvMat(3,3,CV_64F,A[k]);
-        Dist[k] = cvMat(1,8,CV_64F,dk[k]);
+        Dist[k] = cvMat(1,12,CV_64F,dk[k]);
 
-        imagePoints[k] = cvCreateMat( points->rows, points->cols, CV_64FC(CV_MAT_CN(points->type)));
+        imagePoints[k].reset(cvCreateMat( points->rows, points->cols, CV_64FC(CV_MAT_CN(points->type))));
         cvConvert( points, imagePoints[k] );
         cvReshape( imagePoints[k], imagePoints[k], 2, 1 );
 
@@ -1727,10 +1729,10 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
 
     recomputeIntrinsics = (flags & CV_CALIB_FIX_INTRINSIC) == 0;
 
-    err = cvCreateMat( maxPoints*2, 1, CV_64F );
-    Je = cvCreateMat( maxPoints*2, 6, CV_64F );
-    J_LR = cvCreateMat( maxPoints*2, 6, CV_64F );
-    Ji = cvCreateMat( maxPoints*2, NINTRINSIC, CV_64F );
+    err.reset(cvCreateMat( maxPoints*2, 1, CV_64F ));
+    Je.reset(cvCreateMat( maxPoints*2, 6, CV_64F ));
+    J_LR.reset(cvCreateMat( maxPoints*2, 6, CV_64F ));
+    Ji.reset(cvCreateMat( maxPoints*2, NINTRINSIC, CV_64F ));
     cvZero( Ji );
 
     // we optimize for the inter-camera R(3),t(3), then, optionally,
@@ -1738,7 +1740,7 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
     nparams = 6*(nimages+1) + (recomputeIntrinsics ? NINTRINSIC*2 : 0);
 
     // storage for initial [om(R){i}|t{i}] (in order to compute the median for each component)
-    RT0 = cvCreateMat( 6, nimages, CV_64F );
+    RT0.reset(cvCreateMat( 6, nimages, CV_64F ));
 
     solver.init( nparams, 0, termCrit );
     if( recomputeIntrinsics )
@@ -1746,6 +1748,8 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
         uchar* imask = solver.mask->data.ptr + nparams - NINTRINSIC*2;
         if( !(flags & CV_CALIB_RATIONAL_MODEL) )
             flags |= CV_CALIB_FIX_K4 | CV_CALIB_FIX_K5 | CV_CALIB_FIX_K6;
+        if( !(flags & CV_CALIB_THIN_PRISM_MODEL) )
+            flags |= CV_CALIB_FIX_S1_S2_S3_S4;
         if( flags & CV_CALIB_FIX_ASPECT_RATIO )
             imask[0] = imask[NINTRINSIC] = 0;
         if( flags & CV_CALIB_FIX_FOCAL_LENGTH )
@@ -1766,6 +1770,13 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
             imask[10] = imask[NINTRINSIC+10] = 0;
         if( flags & CV_CALIB_FIX_K6 )
             imask[11] = imask[NINTRINSIC+11] = 0;
+        if( flags & CV_CALIB_FIX_S1_S2_S3_S4 )
+        {
+            imask[12] = imask[NINTRINSIC+12] = 0;
+            imask[13] = imask[NINTRINSIC+13] = 0;
+            imask[14] = imask[NINTRINSIC+14] = 0;
+            imask[15] = imask[NINTRINSIC+15] = 0;
+        }
     }
 
     /*
@@ -1840,6 +1851,10 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
             iparam[4] = dk[k][0]; iparam[5] = dk[k][1]; iparam[6] = dk[k][2];
             iparam[7] = dk[k][3]; iparam[8] = dk[k][4]; iparam[9] = dk[k][5];
             iparam[10] = dk[k][6]; iparam[11] = dk[k][7];
+            iparam[12] = dk[k][8];
+            iparam[13] = dk[k][9];
+            iparam[14] = dk[k][10];
+            iparam[15] = dk[k][11];
         }
 
     om_LR = cvMat(3, 1, CV_64F, solver.param->data.db);
@@ -1906,6 +1921,10 @@ double cvStereoCalibrate( const CvMat* _objectPoints, const CvMat* _imagePoints1
                 dk[k][5] = iparam[k*NINTRINSIC+9];
                 dk[k][6] = iparam[k*NINTRINSIC+10];
                 dk[k][7] = iparam[k*NINTRINSIC+11];
+                dk[k][8] = iparam[k*NINTRINSIC+12];
+                dk[k][9] = iparam[k*NINTRINSIC+13];
+                dk[k][10] = iparam[k*NINTRINSIC+14];
+                dk[k][11] = iparam[k*NINTRINSIC+15];
             }
         }
 
@@ -2078,7 +2097,7 @@ icvGetRectangles( const CvMat* cameraMatrix, const CvMat* distCoeffs,
 {
     const int N = 9;
     int x, y, k;
-    cv::Ptr<CvMat> _pts = cvCreateMat(1, N*N, CV_32FC2);
+    cv::Ptr<CvMat> _pts(cvCreateMat(1, N*N, CV_32FC2));
     CvPoint2D32f* pts = (CvPoint2D32f*)(_pts->data.ptr);
 
     for( y = k = 0; y < N; y++ )
@@ -2437,10 +2456,10 @@ CV_IMPL int cvStereoRectifyUncalibrated(
 
     npoints = _points1->rows * _points1->cols * CV_MAT_CN(_points1->type) / 2;
 
-    _m1 = cvCreateMat( _points1->rows, _points1->cols, CV_64FC(CV_MAT_CN(_points1->type)) );
-    _m2 = cvCreateMat( _points2->rows, _points2->cols, CV_64FC(CV_MAT_CN(_points2->type)) );
-    _lines1 = cvCreateMat( 1, npoints, CV_64FC3 );
-    _lines2 = cvCreateMat( 1, npoints, CV_64FC3 );
+    _m1.reset(cvCreateMat( _points1->rows, _points1->cols, CV_64FC(CV_MAT_CN(_points1->type)) ));
+    _m2.reset(cvCreateMat( _points2->rows, _points2->cols, CV_64FC(CV_MAT_CN(_points2->type)) ));
+    _lines1.reset(cvCreateMat( 1, npoints, CV_64FC3 ));
+    _lines2.reset(cvCreateMat( 1, npoints, CV_64FC3 ));
 
     cvConvert( F0, &F );
 
@@ -2688,7 +2707,7 @@ void cv::reprojectImageTo3D( InputArray _disparity,
             for( x = 0; x < cols*3; x++ )
             {
                 int ival = cvRound(dptr[x]);
-                dptr0[x] = CV_CAST_16S(ival);
+                dptr0[x] = cv::saturate_cast<short>(ival);
             }
         }
         else if( dtype == CV_32SC3 )
diff --git a/modules/calib3d/src/checkchessboard.cpp b/modules/calib3d/src/checkchessboard.cpp
index 9191fc8c84..715fe73ef8 100644
--- a/modules/calib3d/src/checkchessboard.cpp
+++ b/modules/calib3d/src/checkchessboard.cpp
@@ -40,6 +40,8 @@
  //M*/
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/calib3d/calib3d_c.h"
 
 #include <vector>
 #include <algorithm>
diff --git a/modules/calib3d/src/circlesgrid.cpp b/modules/calib3d/src/circlesgrid.cpp
index 9d893ffe97..969611fbc6 100644
--- a/modules/calib3d/src/circlesgrid.cpp
+++ b/modules/calib3d/src/circlesgrid.cpp
@@ -41,6 +41,7 @@
  //M*/
 
 #include "circlesgrid.hpp"
+#include <limits>
 //#define DEBUG_CIRCLES
 
 #ifdef DEBUG_CIRCLES
@@ -202,12 +203,12 @@ void CirclesGridClusterFinder::findCorners(const std::vector<cv::Point2f> &hull2
   //corners are the most sharp angles (6)
   Mat anglesMat = Mat(angles);
   Mat sortedIndices;
-  sortIdx(anglesMat, sortedIndices, CV_SORT_EVERY_COLUMN + CV_SORT_DESCENDING);
+  sortIdx(anglesMat, sortedIndices, SORT_EVERY_COLUMN + SORT_DESCENDING);
   CV_Assert(sortedIndices.type() == CV_32SC1);
   CV_Assert(sortedIndices.cols == 1);
   const int cornersCount = isAsymmetricGrid ? 6 : 4;
   Mat cornersIndices;
-  cv::sort(sortedIndices.rowRange(0, cornersCount), cornersIndices, CV_SORT_EVERY_COLUMN + CV_SORT_ASCENDING);
+  cv::sort(sortedIndices.rowRange(0, cornersCount), cornersIndices, SORT_EVERY_COLUMN + SORT_ASCENDING);
   corners.clear();
   for(int i=0; i<cornersCount; i++)
   {
@@ -438,15 +439,15 @@ bool Graph::doesVertexExist(size_t id) const
 
 void Graph::addVertex(size_t id)
 {
-  assert( !doesVertexExist( id ) );
+  CV_Assert( !doesVertexExist( id ) );
 
   vertices.insert(std::pair<size_t, Vertex> (id, Vertex()));
 }
 
 void Graph::addEdge(size_t id1, size_t id2)
 {
-  assert( doesVertexExist( id1 ) );
-  assert( doesVertexExist( id2 ) );
+  CV_Assert( doesVertexExist( id1 ) );
+  CV_Assert( doesVertexExist( id2 ) );
 
   vertices[id1].neighbors.insert(id2);
   vertices[id2].neighbors.insert(id1);
@@ -454,8 +455,8 @@ void Graph::addEdge(size_t id1, size_t id2)
 
 void Graph::removeEdge(size_t id1, size_t id2)
 {
-  assert( doesVertexExist( id1 ) );
-  assert( doesVertexExist( id2 ) );
+  CV_Assert( doesVertexExist( id1 ) );
+  CV_Assert( doesVertexExist( id2 ) );
 
   vertices[id1].neighbors.erase(id2);
   vertices[id2].neighbors.erase(id1);
@@ -463,8 +464,8 @@ void Graph::removeEdge(size_t id1, size_t id2)
 
 bool Graph::areVerticesAdjacent(size_t id1, size_t id2) const
 {
-  assert( doesVertexExist( id1 ) );
-  assert( doesVertexExist( id2 ) );
+  CV_Assert( doesVertexExist( id1 ) );
+  CV_Assert( doesVertexExist( id2 ) );
 
   Vertices::const_iterator it = vertices.find(id1);
   return it->second.neighbors.find(id2) != it->second.neighbors.end();
@@ -477,7 +478,7 @@ size_t Graph::getVerticesCount() const
 
 size_t Graph::getDegree(size_t id) const
 {
-  assert( doesVertexExist(id) );
+  CV_Assert( doesVertexExist(id) );
 
   Vertices::const_iterator it = vertices.find(id);
   return it->second.neighbors.size();
@@ -495,7 +496,7 @@ void Graph::floydWarshall(cv::Mat &distanceMatrix, int infinity) const
     distanceMatrix.at<int> ((int)it1->first, (int)it1->first) = 0;
     for (Neighbors::const_iterator it2 = it1->second.neighbors.begin(); it2 != it1->second.neighbors.end(); it2++)
     {
-      assert( it1->first != *it2 );
+      CV_Assert( it1->first != *it2 );
       distanceMatrix.at<int> ((int)it1->first, (int)*it2) = edgeWeight;
     }
   }
@@ -524,7 +525,7 @@ void Graph::floydWarshall(cv::Mat &distanceMatrix, int infinity) const
 
 const Graph::Neighbors& Graph::getNeighbors(size_t id) const
 {
-  assert( doesVertexExist(id) );
+  CV_Assert( doesVertexExist(id) );
 
   Vertices::const_iterator it = vertices.find(id);
   return it->second.neighbors;
@@ -604,7 +605,7 @@ bool CirclesGridFinder::findHoles()
     }
 
     default:
-      CV_Error(CV_StsBadArg, "Unkown pattern type");
+      CV_Error(Error::StsBadArg, "Unkown pattern type");
   }
   return (isDetectionCorrect());
   //CV_Error( 0, "Detection is not correct" );
@@ -813,7 +814,7 @@ void CirclesGridFinder::findMCS(const std::vector<Point2f> &basis, std::vector<G
 Mat CirclesGridFinder::rectifyGrid(Size detectedGridSize, const std::vector<Point2f>& centers,
                                    const std::vector<Point2f> &keypoints, std::vector<Point2f> &warpedKeypoints)
 {
-  assert( !centers.empty() );
+  CV_Assert( !centers.empty() );
   const float edgeLength = 30;
   const Point2f offset(150, 150);
 
@@ -832,7 +833,7 @@ Mat CirclesGridFinder::rectifyGrid(Size detectedGridSize, const std::vector<Poin
     }
   }
 
-  Mat H = findHomography(Mat(centers), Mat(dstPoints), CV_RANSAC);
+  Mat H = findHomography(Mat(centers), Mat(dstPoints), RANSAC);
   //Mat H = findHomography( Mat( corners ), Mat( dstPoints ) );
 
   std::vector<Point2f> srcKeypoints;
@@ -912,7 +913,7 @@ void CirclesGridFinder::findCandidateLine(std::vector<size_t> &line, size_t seed
     }
   }
 
-  assert( line.size() == seeds.size() );
+  CV_Assert( line.size() == seeds.size() );
 }
 
 void CirclesGridFinder::findCandidateHoles(std::vector<size_t> &above, std::vector<size_t> &below, bool addRow, Point2f basisVec,
@@ -927,9 +928,9 @@ void CirclesGridFinder::findCandidateHoles(std::vector<size_t> &above, std::vect
   size_t lastIdx = addRow ? holes.size() - 1 : holes[0].size() - 1;
   findCandidateLine(below, lastIdx, addRow, basisVec, belowSeeds);
 
-  assert( below.size() == above.size() );
-  assert( belowSeeds.size() == aboveSeeds.size() );
-  assert( below.size() == belowSeeds.size() );
+  CV_Assert( below.size() == above.size() );
+  CV_Assert( belowSeeds.size() == aboveSeeds.size() );
+  CV_Assert( below.size() == belowSeeds.size() );
 }
 
 bool CirclesGridFinder::areCentersNew(const std::vector<size_t> &newCenters, const std::vector<std::vector<size_t> > &holes)
@@ -1000,10 +1001,10 @@ void CirclesGridFinder::insertWinner(float aboveConfidence, float belowConfidenc
 float CirclesGridFinder::computeGraphConfidence(const std::vector<Graph> &basisGraphs, bool addRow,
                                                 const std::vector<size_t> &points, const std::vector<size_t> &seeds)
 {
-  assert( points.size() == seeds.size() );
+  CV_Assert( points.size() == seeds.size() );
   float confidence = 0;
   const size_t vCount = basisGraphs[0].getVerticesCount();
-  assert( basisGraphs[0].getVerticesCount() == basisGraphs[1].getVerticesCount() );
+  CV_Assert( basisGraphs[0].getVerticesCount() == basisGraphs[1].getVerticesCount() );
 
   for (size_t i = 0; i < seeds.size(); i++)
   {
@@ -1087,7 +1088,7 @@ void CirclesGridFinder::findBasis(const std::vector<Point2f> &samples, std::vect
   const int clustersCount = 4;
   kmeans(Mat(samples).reshape(1, 0), clustersCount, bestLabels, termCriteria, parameters.kmeansAttempts,
          KMEANS_RANDOM_CENTERS, centers);
-  assert( centers.type() == CV_32FC1 );
+  CV_Assert( centers.type() == CV_32FC1 );
 
   std::vector<int> basisIndices;
   //TODO: only remove duplicate
@@ -1204,7 +1205,7 @@ void CirclesGridFinder::computeRNG(Graph &rng, std::vector<cv::Point2f> &vectors
 
 void computePredecessorMatrix(const Mat &dm, int verticesCount, Mat &predecessorMatrix)
 {
-  assert( dm.type() == CV_32SC1 );
+  CV_Assert( dm.type() == CV_32SC1 );
   predecessorMatrix.create(verticesCount, verticesCount, CV_32SC1);
   predecessorMatrix = -1;
   for (int i = 0; i < predecessorMatrix.rows; i++)
@@ -1253,7 +1254,6 @@ size_t CirclesGridFinder::findLongestPath(std::vector<Graph> &basisGraphs, Path
 
     double maxVal;
     Point maxLoc;
-    assert (infinity < 0);
     minMaxLoc(distanceMatrix, 0, &maxVal, 0, &maxLoc);
 
     if (maxVal > longestPaths[0].length)
@@ -1364,7 +1364,7 @@ void CirclesGridFinder::drawHoles(const Mat &srcImage, Mat &drawImage) const
   const Scalar holeColor = Scalar(0, 255, 0);
 
   if (srcImage.channels() == 1)
-    cvtColor(srcImage, drawImage, CV_GRAY2RGB);
+    cvtColor(srcImage, drawImage, COLOR_GRAY2RGB);
   else
     srcImage.copyTo(drawImage);
 
@@ -1594,9 +1594,3 @@ size_t CirclesGridFinder::getFirstCorner(std::vector<Point> &largeCornerIndices,
 
   return cornerIdx;
 }
-
-bool cv::findCirclesGridDefault( InputArray image, Size patternSize,
-                                 OutputArray centers, int flags )
-{
-  return findCirclesGrid(image, patternSize, centers, flags);
-}
diff --git a/modules/calib3d/src/circlesgrid.hpp b/modules/calib3d/src/circlesgrid.hpp
index 3028c58b56..fae6b3cac7 100644
--- a/modules/calib3d/src/circlesgrid.hpp
+++ b/modules/calib3d/src/circlesgrid.hpp
@@ -47,6 +47,7 @@
 #include <set>
 #include <list>
 #include <numeric>
+#include <map>
 
 #include "precomp.hpp"
 
diff --git a/modules/calib3d/src/compat_ptsetreg.cpp b/modules/calib3d/src/compat_ptsetreg.cpp
index dca0d0dabe..e8f4108580 100644
--- a/modules/calib3d/src/compat_ptsetreg.cpp
+++ b/modules/calib3d/src/compat_ptsetreg.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 /************************************************************************************\
        Some backward compatibility stuff, to be moved to legacy or compat module
@@ -52,7 +53,6 @@ using cv::Ptr;
 
 CvLevMarq::CvLevMarq()
 {
-    mask = prevParam = param = J = err = JtJ = JtJN = JtErr = JtJV = JtJW = Ptr<CvMat>();
     lambdaLg10 = 0; state = DONE;
     criteria = cvTermCriteria(0,0,0);
     iters = 0;
@@ -61,7 +61,6 @@ CvLevMarq::CvLevMarq()
 
 CvLevMarq::CvLevMarq( int nparams, int nerrs, CvTermCriteria criteria0, bool _completeSymmFlag )
 {
-    mask = prevParam = param = J = err = JtJ = JtJN = JtErr = JtJV = JtJW = Ptr<CvMat>();
     init(nparams, nerrs, criteria0, _completeSymmFlag);
 }
 
@@ -88,19 +87,19 @@ void CvLevMarq::init( int nparams, int nerrs, CvTermCriteria criteria0, bool _co
 {
     if( !param || param->rows != nparams || nerrs != (err ? err->rows : 0) )
         clear();
-    mask = cvCreateMat( nparams, 1, CV_8U );
+    mask.reset(cvCreateMat( nparams, 1, CV_8U ));
     cvSet(mask, cvScalarAll(1));
-    prevParam = cvCreateMat( nparams, 1, CV_64F );
-    param = cvCreateMat( nparams, 1, CV_64F );
-    JtJ = cvCreateMat( nparams, nparams, CV_64F );
-    JtJN = cvCreateMat( nparams, nparams, CV_64F );
-    JtJV = cvCreateMat( nparams, nparams, CV_64F );
-    JtJW = cvCreateMat( nparams, 1, CV_64F );
-    JtErr = cvCreateMat( nparams, 1, CV_64F );
+    prevParam.reset(cvCreateMat( nparams, 1, CV_64F ));
+    param.reset(cvCreateMat( nparams, 1, CV_64F ));
+    JtJ.reset(cvCreateMat( nparams, nparams, CV_64F ));
+    JtJN.reset(cvCreateMat( nparams, nparams, CV_64F ));
+    JtJV.reset(cvCreateMat( nparams, nparams, CV_64F ));
+    JtJW.reset(cvCreateMat( nparams, 1, CV_64F ));
+    JtErr.reset(cvCreateMat( nparams, 1, CV_64F ));
     if( nerrs > 0 )
     {
-        J = cvCreateMat( nerrs, nparams, CV_64F );
-        err = cvCreateMat( nerrs, 1, CV_64F );
+        J.reset(cvCreateMat( nerrs, nparams, CV_64F ));
+        err.reset(cvCreateMat( nerrs, 1, CV_64F ));
     }
     prevErrNorm = DBL_MAX;
     lambdaLg10 = -3;
@@ -195,7 +194,7 @@ bool CvLevMarq::updateAlt( const CvMat*& _param, CvMat*& _JtJ, CvMat*& _JtErr, d
 {
     double change;
 
-    CV_Assert( err.empty() );
+    CV_Assert( !err );
     if( state == DONE )
     {
         _param = param;
@@ -427,4 +426,3 @@ CV_IMPL void cvConvertPointsHomogeneous( const CvMat* _src, CvMat* _dst )
             dst.convertTo(dst0, dst0.type());
     }
 }
-
diff --git a/modules/calib3d/src/compat_stereo.cpp b/modules/calib3d/src/compat_stereo.cpp
index 7b9ed33e41..16eefc6847 100644
--- a/modules/calib3d/src/compat_stereo.cpp
+++ b/modules/calib3d/src/compat_stereo.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 CvStereoBMState* cvCreateStereoBMState( int /*preset*/, int numberOfDisparities )
 {
@@ -83,10 +84,6 @@ void cvReleaseStereoBMState( CvStereoBMState** state )
     cvFree( state );
 }
 
-template<> void cv::Ptr<CvStereoBMState>::delete_obj()
-{ cvReleaseStereoBMState(&obj); }
-
-
 void cvFindStereoCorrespondenceBM( const CvArr* leftarr, const CvArr* rightarr,
                                    CvArr* disparr, CvStereoBMState* state )
 {
@@ -124,93 +121,3 @@ void cvValidateDisparity( CvArr* _disp, const CvArr* _cost, int minDisparity,
     cv::Mat disp = cv::cvarrToMat(_disp), cost = cv::cvarrToMat(_cost);
     cv::validateDisparity( disp, cost, minDisparity, numberOfDisparities, disp12MaxDiff );
 }
-
-namespace cv
-{
-
-StereoBM::StereoBM()
-{ init(BASIC_PRESET); }
-
-StereoBM::StereoBM(int _preset, int _ndisparities, int _SADWindowSize)
-{ init(_preset, _ndisparities, _SADWindowSize); }
-
-void StereoBM::init(int _preset, int _ndisparities, int _SADWindowSize)
-{
-    state = cvCreateStereoBMState(_preset, _ndisparities);
-    state->SADWindowSize = _SADWindowSize;
-}
-
-void StereoBM::operator()( InputArray _left, InputArray _right,
-                          OutputArray _disparity, int disptype )
-{
-    Mat left = _left.getMat(), right = _right.getMat();
-    CV_Assert( disptype == CV_16S || disptype == CV_32F );
-    _disparity.create(left.size(), disptype);
-    Mat disp = _disparity.getMat();
-
-    CvMat left_c = left, right_c = right, disp_c = disp;
-    cvFindStereoCorrespondenceBM(&left_c, &right_c, &disp_c, state);
-}
-
-
-StereoSGBM::StereoSGBM()
-{
-    minDisparity = numberOfDisparities = 0;
-    SADWindowSize = 0;
-    P1 = P2 = 0;
-    disp12MaxDiff = 0;
-    preFilterCap = 0;
-    uniquenessRatio = 0;
-    speckleWindowSize = 0;
-    speckleRange = 0;
-    fullDP = false;
-
-    sm = createStereoSGBM(0, 0, 0);
-}
-
-StereoSGBM::StereoSGBM( int _minDisparity, int _numDisparities, int _SADWindowSize,
-                       int _P1, int _P2, int _disp12MaxDiff, int _preFilterCap,
-                       int _uniquenessRatio, int _speckleWindowSize, int _speckleRange,
-                       bool _fullDP )
-{
-    minDisparity = _minDisparity;
-    numberOfDisparities = _numDisparities;
-    SADWindowSize = _SADWindowSize;
-    P1 = _P1;
-    P2 = _P2;
-    disp12MaxDiff = _disp12MaxDiff;
-    preFilterCap = _preFilterCap;
-    uniquenessRatio = _uniquenessRatio;
-    speckleWindowSize = _speckleWindowSize;
-    speckleRange = _speckleRange;
-    fullDP = _fullDP;
-
-    sm = createStereoSGBM(0, 0, 0);
-}
-
-StereoSGBM::~StereoSGBM()
-{
-}
-
-void StereoSGBM::operator ()( InputArray _left, InputArray _right,
-                             OutputArray _disp )
-{
-    sm->set("minDisparity", minDisparity);
-    sm->set("numDisparities", numberOfDisparities);
-    sm->set("SADWindowSize", SADWindowSize);
-    sm->set("P1", P1);
-    sm->set("P2", P2);
-    sm->set("disp12MaxDiff", disp12MaxDiff);
-    sm->set("preFilterCap", preFilterCap);
-    sm->set("uniquenessRatio", uniquenessRatio);
-    sm->set("speckleWindowSize", speckleWindowSize);
-    sm->set("speckleRange", speckleRange);
-    sm->set("fullDP", fullDP);
-
-    sm->compute(_left, _right, _disp);
-}
-
-}
-
-
-
diff --git a/modules/calib3d/src/epnp.cpp b/modules/calib3d/src/epnp.cpp
index 7fb63254dc..edbcaffd39 100644
--- a/modules/calib3d/src/epnp.cpp
+++ b/modules/calib3d/src/epnp.cpp
@@ -621,4 +621,3 @@ void epnp::qr_solve(CvMat * A, CvMat * b, CvMat * X)
     pX[i] = (pb[i] - sum) / A2[i];
   }
 }
-
diff --git a/modules/calib3d/src/epnp.h b/modules/calib3d/src/epnp.h
index 203830a6d6..fe01606305 100644
--- a/modules/calib3d/src/epnp.h
+++ b/modules/calib3d/src/epnp.h
@@ -2,6 +2,7 @@
 #define epnp_h
 
 #include "precomp.hpp"
+#include "opencv2/core/core_c.h"
 
 class epnp {
  public:
diff --git a/modules/calib3d/src/five-point.cpp b/modules/calib3d/src/five-point.cpp
index 57f941e381..9922247008 100644
--- a/modules/calib3d/src/five-point.cpp
+++ b/modules/calib3d/src/five-point.cpp
@@ -1,47 +1,35 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
+/* This is a 5-point algorithm contributed to OpenCV by the author, Bo Li.
+   It implements the 5-point algorithm solver from Nister's paper:
+   Nister, An efficient solution to the five-point relative pose problem, PAMI, 2004.
+*/
+
+/*  Copyright (c) 2013, Bo Li (prclibo@gmail.com), ETH Zurich
+    All rights reserved.
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are met:
+        * Redistributions of source code must retain the above copyright
+          notice, this list of conditions and the following disclaimer.
+        * Redistributions in binary form must reproduce the above copyright
+          notice, this list of conditions and the following disclaimer in the
+          documentation and/or other materials provided with the distribution.
+        * Neither the name of the copyright holder nor the
+          names of its contributors may be used to endorse or promote products
+          derived from this software without specific prior written permission.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+    ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+    WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+    DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE FOR ANY
+    DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+    (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+    LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+    ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*/
 
 #include "precomp.hpp"
-#include <complex>
 
 namespace cv
 {
@@ -71,16 +59,10 @@ public:
         SVD::compute(Q, W, U, Vt, SVD::MODIFY_A | SVD::FULL_UV);
 
         Mat EE = Mat(Vt.t()).colRange(5, 9) * 1.0;
-        Mat AA(20, 10, CV_64F);
-        EE = EE.t();
-        calibrated_fivepoint_helper((double*)EE.data, (double*)AA.data);
-        AA = AA.t();
-        EE = EE.t();
-
         Mat A(10, 20, CV_64F);
-        int perm[20] = {0, 3, 1, 2, 4, 10, 6, 12, 5, 11, 7, 13, 16, 8, 14, 17, 9, 15, 18, 19};
-        for (int i = 0; i < 20; i++)
-            A.col(i) = AA.col(perm[i]) * 1.0;
+        EE = EE.t();
+        getCoeffMat((double*)EE.data, (double*)A.data);
+        EE = EE.t();
 
         A = A.colRange(0, 10).inv() * A.colRange(10, 20);
 
@@ -118,7 +100,7 @@ public:
         c[1] = (b[29]*b[7]*b[24]-b[29]*b[20]*b[11]+b[2]*b[20]*b[38]-b[2]*b[25]*b[33]-b[28]*b[20]*b[12]+b[28]*b[7]*b[25]-b[29]*b[19]*b[12]-b[3]*b[24]*b[33]+b[15]*b[33]*b[12]+b[3]*b[19]*b[38]-b[16]*b[6]*b[38]+b[3]*b[20]*b[37]+b[16]*b[32]*b[12]+b[29]*b[6]*b[25]-b[16]*b[7]*b[37]-b[3]*b[25]*b[32]-b[15]*b[7]*b[38]+b[16]*b[33]*b[11]);
         c[0] = -b[29]*b[20]*b[12]+b[29]*b[7]*b[25]+b[16]*b[33]*b[12]-b[16]*b[7]*b[38]+b[3]*b[20]*b[38]-b[3]*b[25]*b[33];
 
-        std::vector<std::complex<double> > roots;
+        std::vector<Complex<double> > roots;
         solvePoly(coeffs, roots);
 
         std::vector<double> xs, ys, zs;
@@ -128,8 +110,8 @@ public:
         double* e = ematrix.ptr<double>();
         for (size_t i = 0; i < roots.size(); i++)
         {
-            if (fabs(roots[i].imag()) > 1e-10) continue;
-            double z1 = roots[i].real();
+            if (fabs(roots[i].im) > 1e-10) continue;
+            double z1 = roots[i].re;
             double z2 = z1 * z1;
             double z3 = z2 * z1;
             double z4 = z3 * z1;
@@ -146,355 +128,248 @@ public:
             Mat Bz(3, 3, CV_64F, bz);
             cv::Mat xy1;
             SVD::solveZ(Bz, xy1);
-            
-            if (fabs(xy1.at<double>(2)) < 1e-10) continue; 
-            xs.push_back(xy1.at<double>(0) / xy1.at<double>(2)); 
-            ys.push_back(xy1.at<double>(1) / xy1.at<double>(2)); 
-            zs.push_back(z1); 
-            
-            cv::Mat Evec = EE.col(0) * xs.back() + EE.col(1) * ys.back() + EE.col(2) * zs.back() + EE.col(3); 
-            Evec /= norm(Evec); 
-            
-            memcpy(e + count * 9, Evec.data, 9 * sizeof(double)); 
-            count++; 
+
+            if (fabs(xy1.at<double>(2)) < 1e-10) continue;
+            xs.push_back(xy1.at<double>(0) / xy1.at<double>(2));
+            ys.push_back(xy1.at<double>(1) / xy1.at<double>(2));
+            zs.push_back(z1);
+
+            cv::Mat Evec = EE.col(0) * xs.back() + EE.col(1) * ys.back() + EE.col(2) * zs.back() + EE.col(3);
+            Evec /= norm(Evec);
+
+            memcpy(e + count * 9, Evec.data, 9 * sizeof(double));
+            count++;
         }
 
         ematrix.rowRange(0, count*3).copyTo(_model);
         return count;
     }
 
-protected: 
-    void calibrated_fivepoint_helper( double *EE, double* A ) const
+protected:
+    void getCoeffMat(double *e, double *A) const
     {
-        double e00,e01,e02,e03,e04,e05,e06,e07,e08;
-        double e10,e11,e12,e13,e14,e15,e16,e17,e18;
-        double e20,e21,e22,e23,e24,e25,e26,e27,e28;
-        double e30,e31,e32,e33,e34,e35,e36,e37,e38;
+        double ep2[36], ep3[36];
+        for (int i = 0; i < 36; i++)
+        {
+            ep2[i] = e[i] * e[i];
+            ep3[i] = ep2[i] * e[i];
+        }
 
-        double e002,e012,e022,e032,e042,e052,e062,e072,e082;
-        double e102,e112,e122,e132,e142,e152,e162,e172,e182;
-        double e202,e212,e222,e232,e242,e252,e262,e272,e282;
-        double e302,e312,e322,e332,e342,e352,e362,e372,e382;
+        A[0]=e[33]*e[28]*e[32]-e[33]*e[31]*e[29]+e[30]*e[34]*e[29]-e[30]*e[28]*e[35]-e[27]*e[32]*e[34]+e[27]*e[31]*e[35];
+        A[146]=.5000000000*e[6]*ep2[8]-.5000000000*e[6]*ep2[5]+.5000000000*ep3[6]+.5000000000*e[6]*ep2[7]-.5000000000*e[6]*ep2[4]+e[0]*e[2]*e[8]+e[3]*e[4]*e[7]+e[3]*e[5]*e[8]+e[0]*e[1]*e[7]-.5000000000*e[6]*ep2[1]-.5000000000*e[6]*ep2[2]+.5000000000*ep2[0]*e[6]+.5000000000*ep2[3]*e[6];
+        A[1]=e[30]*e[34]*e[2]+e[33]*e[1]*e[32]-e[3]*e[28]*e[35]+e[0]*e[31]*e[35]+e[3]*e[34]*e[29]-e[30]*e[1]*e[35]+e[27]*e[31]*e[8]-e[27]*e[32]*e[7]-e[30]*e[28]*e[8]-e[33]*e[31]*e[2]-e[0]*e[32]*e[34]+e[6]*e[28]*e[32]-e[33]*e[4]*e[29]+e[33]*e[28]*e[5]+e[30]*e[7]*e[29]+e[27]*e[4]*e[35]-e[27]*e[5]*e[34]-e[6]*e[31]*e[29];
+        A[147]=e[9]*e[27]*e[15]+e[9]*e[29]*e[17]+e[9]*e[11]*e[35]+e[9]*e[28]*e[16]+e[9]*e[10]*e[34]+e[27]*e[11]*e[17]+e[27]*e[10]*e[16]+e[12]*e[30]*e[15]+e[12]*e[32]*e[17]+e[12]*e[14]*e[35]+e[12]*e[31]*e[16]+e[12]*e[13]*e[34]+e[30]*e[14]*e[17]+e[30]*e[13]*e[16]+e[15]*e[35]*e[17]+e[15]*e[34]*e[16]-1.*e[15]*e[28]*e[10]-1.*e[15]*e[31]*e[13]-1.*e[15]*e[32]*e[14]-1.*e[15]*e[29]*e[11]+.5000000000*ep2[9]*e[33]+.5000000000*e[33]*ep2[16]-.5000000000*e[33]*ep2[11]+.5000000000*e[33]*ep2[12]+1.500000000*e[33]*ep2[15]+.5000000000*e[33]*ep2[17]-.5000000000*e[33]*ep2[10]-.5000000000*e[33]*ep2[14]-.5000000000*e[33]*ep2[13];
+        A[2]=-e[33]*e[22]*e[29]-e[33]*e[31]*e[20]-e[27]*e[32]*e[25]+e[27]*e[22]*e[35]-e[27]*e[23]*e[34]+e[27]*e[31]*e[26]+e[33]*e[28]*e[23]-e[21]*e[28]*e[35]+e[30]*e[25]*e[29]+e[24]*e[28]*e[32]-e[24]*e[31]*e[29]+e[18]*e[31]*e[35]-e[30]*e[28]*e[26]-e[30]*e[19]*e[35]+e[21]*e[34]*e[29]+e[33]*e[19]*e[32]-e[18]*e[32]*e[34]+e[30]*e[34]*e[20];
+        A[144]=e[18]*e[2]*e[17]+e[3]*e[21]*e[15]+e[3]*e[12]*e[24]+e[3]*e[23]*e[17]+e[3]*e[14]*e[26]+e[3]*e[22]*e[16]+e[3]*e[13]*e[25]+3.*e[6]*e[24]*e[15]+e[6]*e[26]*e[17]+e[6]*e[25]*e[16]+e[0]*e[20]*e[17]+e[0]*e[11]*e[26]+e[0]*e[19]*e[16]+e[0]*e[10]*e[25]+e[15]*e[26]*e[8]-1.*e[15]*e[20]*e[2]-1.*e[15]*e[19]*e[1]-1.*e[15]*e[22]*e[4]+e[15]*e[25]*e[7]-1.*e[15]*e[23]*e[5]+e[12]*e[21]*e[6]+e[12]*e[22]*e[7]+e[12]*e[4]*e[25]+e[12]*e[23]*e[8]+e[12]*e[5]*e[26]-1.*e[24]*e[11]*e[2]-1.*e[24]*e[10]*e[1]-1.*e[24]*e[13]*e[4]+e[24]*e[16]*e[7]-1.*e[24]*e[14]*e[5]+e[24]*e[17]*e[8]+e[21]*e[13]*e[7]+e[21]*e[4]*e[16]+e[21]*e[14]*e[8]+e[21]*e[5]*e[17]-1.*e[6]*e[23]*e[14]-1.*e[6]*e[20]*e[11]-1.*e[6]*e[19]*e[10]-1.*e[6]*e[22]*e[13]+e[9]*e[18]*e[6]+e[9]*e[0]*e[24]+e[9]*e[19]*e[7]+e[9]*e[1]*e[25]+e[9]*e[20]*e[8]+e[9]*e[2]*e[26]+e[18]*e[0]*e[15]+e[18]*e[10]*e[7]+e[18]*e[1]*e[16]+e[18]*e[11]*e[8];
+        A[3]=e[33]*e[10]*e[32]+e[33]*e[28]*e[14]-e[33]*e[13]*e[29]-e[33]*e[31]*e[11]+e[9]*e[31]*e[35]-e[9]*e[32]*e[34]+e[27]*e[13]*e[35]-e[27]*e[32]*e[16]+e[27]*e[31]*e[17]-e[27]*e[14]*e[34]+e[12]*e[34]*e[29]-e[12]*e[28]*e[35]+e[30]*e[34]*e[11]+e[30]*e[16]*e[29]-e[30]*e[10]*e[35]-e[30]*e[28]*e[17]+e[15]*e[28]*e[32]-e[15]*e[31]*e[29];
+        A[145]=e[0]*e[27]*e[6]+e[0]*e[28]*e[7]+e[0]*e[1]*e[34]+e[0]*e[29]*e[8]+e[0]*e[2]*e[35]+e[6]*e[34]*e[7]-1.*e[6]*e[32]*e[5]+e[6]*e[30]*e[3]+e[6]*e[35]*e[8]-1.*e[6]*e[29]*e[2]-1.*e[6]*e[28]*e[1]-1.*e[6]*e[31]*e[4]+e[27]*e[1]*e[7]+e[27]*e[2]*e[8]+e[3]*e[31]*e[7]+e[3]*e[4]*e[34]+e[3]*e[32]*e[8]+e[3]*e[5]*e[35]+e[30]*e[4]*e[7]+e[30]*e[5]*e[8]+.5000000000*ep2[0]*e[33]+1.500000000*e[33]*ep2[6]-.5000000000*e[33]*ep2[4]-.5000000000*e[33]*ep2[5]-.5000000000*e[33]*ep2[1]+.5000000000*e[33]*ep2[7]+.5000000000*e[33]*ep2[3]-.5000000000*e[33]*ep2[2]+.5000000000*e[33]*ep2[8];
+        A[4]=-e[0]*e[23]*e[16]+e[9]*e[4]*e[26]+e[9]*e[22]*e[8]-e[9]*e[5]*e[25]-e[9]*e[23]*e[7]+e[18]*e[4]*e[17]+e[18]*e[13]*e[8]-e[18]*e[5]*e[16]-e[18]*e[14]*e[7]+e[3]*e[16]*e[20]+e[3]*e[25]*e[11]-e[3]*e[10]*e[26]-e[3]*e[19]*e[17]+e[12]*e[7]*e[20]+e[12]*e[25]*e[2]-e[12]*e[1]*e[26]-e[12]*e[19]*e[8]+e[21]*e[7]*e[11]+e[21]*e[16]*e[2]-e[21]*e[1]*e[17]-e[21]*e[10]*e[8]+e[6]*e[10]*e[23]+e[6]*e[19]*e[14]-e[6]*e[13]*e[20]-e[6]*e[22]*e[11]+e[15]*e[1]*e[23]+e[15]*e[19]*e[5]-e[15]*e[4]*e[20]-e[15]*e[22]*e[2]+e[24]*e[1]*e[14]+e[24]*e[10]*e[5]-e[24]*e[4]*e[11]-e[24]*e[13]*e[2]+e[0]*e[13]*e[26]+e[0]*e[22]*e[17]-e[0]*e[14]*e[25];
+        A[150]=e[18]*e[19]*e[25]+.5000000000*ep3[24]-.5000000000*e[24]*ep2[23]+e[18]*e[20]*e[26]+e[21]*e[22]*e[25]+e[21]*e[23]*e[26]-.5000000000*e[24]*ep2[19]+.5000000000*ep2[21]*e[24]+.5000000000*e[24]*ep2[26]-.5000000000*e[24]*ep2[20]+.5000000000*ep2[18]*e[24]-.5000000000*e[24]*ep2[22]+.5000000000*e[24]*ep2[25];
+        A[5]=-e[3]*e[1]*e[35]-e[0]*e[32]*e[7]+e[27]*e[4]*e[8]+e[33]*e[1]*e[5]-e[33]*e[4]*e[2]+e[0]*e[4]*e[35]+e[3]*e[34]*e[2]-e[30]*e[1]*e[8]+e[30]*e[7]*e[2]-e[6]*e[4]*e[29]+e[3]*e[7]*e[29]+e[6]*e[1]*e[32]-e[0]*e[5]*e[34]-e[3]*e[28]*e[8]+e[0]*e[31]*e[8]+e[6]*e[28]*e[5]-e[6]*e[31]*e[2]-e[27]*e[5]*e[7];
+        A[151]=e[33]*e[16]*e[7]-1.*e[33]*e[14]*e[5]+e[33]*e[17]*e[8]+e[30]*e[13]*e[7]+e[30]*e[4]*e[16]+e[30]*e[14]*e[8]+e[30]*e[5]*e[17]+e[6]*e[27]*e[9]-1.*e[6]*e[28]*e[10]-1.*e[6]*e[31]*e[13]-1.*e[6]*e[32]*e[14]-1.*e[6]*e[29]*e[11]+e[9]*e[28]*e[7]+e[9]*e[1]*e[34]+e[9]*e[29]*e[8]+e[9]*e[2]*e[35]+e[27]*e[10]*e[7]+e[27]*e[1]*e[16]+e[27]*e[11]*e[8]+e[27]*e[2]*e[17]+e[3]*e[30]*e[15]+e[3]*e[12]*e[33]+e[3]*e[32]*e[17]+e[3]*e[14]*e[35]+e[3]*e[31]*e[16]+e[3]*e[13]*e[34]+3.*e[6]*e[33]*e[15]+e[6]*e[35]*e[17]+e[6]*e[34]*e[16]+e[0]*e[27]*e[15]+e[0]*e[9]*e[33]+e[0]*e[29]*e[17]+e[0]*e[11]*e[35]+e[0]*e[28]*e[16]+e[0]*e[10]*e[34]+e[15]*e[34]*e[7]-1.*e[15]*e[32]*e[5]+e[15]*e[35]*e[8]-1.*e[15]*e[29]*e[2]-1.*e[15]*e[28]*e[1]-1.*e[15]*e[31]*e[4]+e[12]*e[30]*e[6]+e[12]*e[31]*e[7]+e[12]*e[4]*e[34]+e[12]*e[32]*e[8]+e[12]*e[5]*e[35]-1.*e[33]*e[11]*e[2]-1.*e[33]*e[10]*e[1]-1.*e[33]*e[13]*e[4];
+        A[6]=e[6]*e[1]*e[5]-e[6]*e[4]*e[2]+e[3]*e[7]*e[2]+e[0]*e[4]*e[8]-e[0]*e[5]*e[7]-e[3]*e[1]*e[8];
+        A[148]=.5000000000*ep3[15]+e[9]*e[10]*e[16]-.5000000000*e[15]*ep2[11]+e[9]*e[11]*e[17]+.5000000000*ep2[12]*e[15]+.5000000000*e[15]*ep2[16]+.5000000000*e[15]*ep2[17]-.5000000000*e[15]*ep2[13]+.5000000000*ep2[9]*e[15]+e[12]*e[14]*e[17]-.5000000000*e[15]*ep2[10]-.5000000000*e[15]*ep2[14]+e[12]*e[13]*e[16];
+        A[7]=e[15]*e[28]*e[14]-e[15]*e[13]*e[29]-e[15]*e[31]*e[11]+e[33]*e[10]*e[14]-e[33]*e[13]*e[11]+e[9]*e[13]*e[35]-e[9]*e[32]*e[16]+e[9]*e[31]*e[17]-e[9]*e[14]*e[34]+e[27]*e[13]*e[17]-e[27]*e[14]*e[16]+e[12]*e[34]*e[11]+e[12]*e[16]*e[29]-e[12]*e[10]*e[35]-e[12]*e[28]*e[17]+e[30]*e[16]*e[11]-e[30]*e[10]*e[17]+e[15]*e[10]*e[32];
+        A[149]=e[18]*e[27]*e[24]+e[18]*e[28]*e[25]+e[18]*e[19]*e[34]+e[18]*e[29]*e[26]+e[18]*e[20]*e[35]+e[27]*e[19]*e[25]+e[27]*e[20]*e[26]+e[21]*e[30]*e[24]+e[21]*e[31]*e[25]+e[21]*e[22]*e[34]+e[21]*e[32]*e[26]+e[21]*e[23]*e[35]+e[30]*e[22]*e[25]+e[30]*e[23]*e[26]+e[24]*e[34]*e[25]+e[24]*e[35]*e[26]-1.*e[24]*e[29]*e[20]-1.*e[24]*e[31]*e[22]-1.*e[24]*e[32]*e[23]-1.*e[24]*e[28]*e[19]+1.500000000*e[33]*ep2[24]+.5000000000*e[33]*ep2[25]+.5000000000*e[33]*ep2[26]-.5000000000*e[33]*ep2[23]-.5000000000*e[33]*ep2[19]-.5000000000*e[33]*ep2[20]-.5000000000*e[33]*ep2[22]+.5000000000*ep2[18]*e[33]+.5000000000*ep2[21]*e[33];
+        A[9]=e[21]*e[25]*e[29]-e[27]*e[23]*e[25]+e[24]*e[19]*e[32]-e[21]*e[28]*e[26]-e[21]*e[19]*e[35]+e[18]*e[31]*e[26]-e[30]*e[19]*e[26]-e[24]*e[31]*e[20]+e[24]*e[28]*e[23]+e[27]*e[22]*e[26]+e[30]*e[25]*e[20]-e[33]*e[22]*e[20]+e[33]*e[19]*e[23]+e[21]*e[34]*e[20]-e[18]*e[23]*e[34]-e[24]*e[22]*e[29]-e[18]*e[32]*e[25]+e[18]*e[22]*e[35];
+        A[155]=e[12]*e[14]*e[8]+e[12]*e[5]*e[17]+e[15]*e[16]*e[7]+e[15]*e[17]*e[8]+e[0]*e[11]*e[17]+e[0]*e[9]*e[15]+e[0]*e[10]*e[16]+e[3]*e[14]*e[17]+e[3]*e[13]*e[16]+e[9]*e[10]*e[7]+e[9]*e[1]*e[16]+e[9]*e[11]*e[8]+e[9]*e[2]*e[17]-1.*e[15]*e[11]*e[2]-1.*e[15]*e[10]*e[1]-1.*e[15]*e[13]*e[4]-1.*e[15]*e[14]*e[5]+e[12]*e[3]*e[15]+e[12]*e[13]*e[7]+e[12]*e[4]*e[16]+.5000000000*ep2[12]*e[6]+1.500000000*ep2[15]*e[6]+.5000000000*e[6]*ep2[17]+.5000000000*e[6]*ep2[16]+.5000000000*e[6]*ep2[9]-.5000000000*e[6]*ep2[11]-.5000000000*e[6]*ep2[10]-.5000000000*e[6]*ep2[14]-.5000000000*e[6]*ep2[13];
+        A[8]=-e[9]*e[14]*e[16]-e[12]*e[10]*e[17]+e[9]*e[13]*e[17]-e[15]*e[13]*e[11]+e[15]*e[10]*e[14]+e[12]*e[16]*e[11];
+        A[154]=e[21]*e[14]*e[17]+e[21]*e[13]*e[16]+e[15]*e[26]*e[17]+e[15]*e[25]*e[16]-1.*e[15]*e[23]*e[14]-1.*e[15]*e[20]*e[11]-1.*e[15]*e[19]*e[10]-1.*e[15]*e[22]*e[13]+e[9]*e[20]*e[17]+e[9]*e[11]*e[26]+e[9]*e[19]*e[16]+e[9]*e[10]*e[25]+.5000000000*ep2[12]*e[24]+1.500000000*e[24]*ep2[15]+.5000000000*e[24]*ep2[17]+.5000000000*e[24]*ep2[16]+.5000000000*ep2[9]*e[24]-.5000000000*e[24]*ep2[11]-.5000000000*e[24]*ep2[10]-.5000000000*e[24]*ep2[14]-.5000000000*e[24]*ep2[13]+e[18]*e[11]*e[17]+e[18]*e[9]*e[15]+e[18]*e[10]*e[16]+e[12]*e[21]*e[15]+e[12]*e[23]*e[17]+e[12]*e[14]*e[26]+e[12]*e[22]*e[16]+e[12]*e[13]*e[25];
+        A[11]=-e[9]*e[5]*e[34]+e[9]*e[31]*e[8]-e[9]*e[32]*e[7]+e[27]*e[4]*e[17]+e[27]*e[13]*e[8]-e[27]*e[5]*e[16]-e[27]*e[14]*e[7]+e[0]*e[13]*e[35]-e[0]*e[32]*e[16]+e[0]*e[31]*e[17]-e[0]*e[14]*e[34]+e[9]*e[4]*e[35]+e[6]*e[10]*e[32]+e[6]*e[28]*e[14]-e[6]*e[13]*e[29]-e[6]*e[31]*e[11]+e[15]*e[1]*e[32]+e[3]*e[34]*e[11]+e[3]*e[16]*e[29]-e[3]*e[10]*e[35]-e[3]*e[28]*e[17]-e[12]*e[1]*e[35]+e[12]*e[7]*e[29]+e[12]*e[34]*e[2]-e[12]*e[28]*e[8]+e[15]*e[28]*e[5]-e[15]*e[4]*e[29]-e[15]*e[31]*e[2]+e[33]*e[1]*e[14]+e[33]*e[10]*e[5]-e[33]*e[4]*e[11]-e[33]*e[13]*e[2]+e[30]*e[7]*e[11]+e[30]*e[16]*e[2]-e[30]*e[1]*e[17]-e[30]*e[10]*e[8];
+        A[153]=e[21]*e[31]*e[7]+e[21]*e[4]*e[34]+e[21]*e[32]*e[8]+e[21]*e[5]*e[35]+e[30]*e[22]*e[7]+e[30]*e[4]*e[25]+e[30]*e[23]*e[8]+e[30]*e[5]*e[26]+3.*e[24]*e[33]*e[6]+e[24]*e[34]*e[7]+e[24]*e[35]*e[8]+e[33]*e[25]*e[7]+e[33]*e[26]*e[8]+e[0]*e[27]*e[24]+e[0]*e[18]*e[33]+e[0]*e[28]*e[25]+e[0]*e[19]*e[34]+e[0]*e[29]*e[26]+e[0]*e[20]*e[35]+e[18]*e[27]*e[6]+e[18]*e[28]*e[7]+e[18]*e[1]*e[34]+e[18]*e[29]*e[8]+e[18]*e[2]*e[35]+e[27]*e[19]*e[7]+e[27]*e[1]*e[25]+e[27]*e[20]*e[8]+e[27]*e[2]*e[26]+e[3]*e[30]*e[24]+e[3]*e[21]*e[33]+e[3]*e[31]*e[25]+e[3]*e[22]*e[34]+e[3]*e[32]*e[26]+e[3]*e[23]*e[35]+e[6]*e[30]*e[21]-1.*e[6]*e[29]*e[20]+e[6]*e[35]*e[26]-1.*e[6]*e[31]*e[22]-1.*e[6]*e[32]*e[23]-1.*e[6]*e[28]*e[19]+e[6]*e[34]*e[25]-1.*e[24]*e[32]*e[5]-1.*e[24]*e[29]*e[2]-1.*e[24]*e[28]*e[1]-1.*e[24]*e[31]*e[4]-1.*e[33]*e[20]*e[2]-1.*e[33]*e[19]*e[1]-1.*e[33]*e[22]*e[4]-1.*e[33]*e[23]*e[5];
+        A[10]=e[21]*e[25]*e[20]-e[21]*e[19]*e[26]+e[18]*e[22]*e[26]-e[18]*e[23]*e[25]-e[24]*e[22]*e[20]+e[24]*e[19]*e[23];
+        A[152]=e[3]*e[4]*e[25]+e[3]*e[23]*e[8]+e[3]*e[5]*e[26]+e[21]*e[4]*e[7]+e[21]*e[5]*e[8]+e[6]*e[25]*e[7]+e[6]*e[26]*e[8]+e[0]*e[19]*e[7]+e[0]*e[1]*e[25]+e[0]*e[20]*e[8]+e[0]*e[2]*e[26]-1.*e[6]*e[20]*e[2]-1.*e[6]*e[19]*e[1]-1.*e[6]*e[22]*e[4]-1.*e[6]*e[23]*e[5]+e[18]*e[1]*e[7]+e[18]*e[0]*e[6]+e[18]*e[2]*e[8]+e[3]*e[21]*e[6]+e[3]*e[22]*e[7]-.5000000000*e[24]*ep2[4]+.5000000000*e[24]*ep2[0]+1.500000000*e[24]*ep2[6]-.5000000000*e[24]*ep2[5]-.5000000000*e[24]*ep2[1]+.5000000000*e[24]*ep2[7]+.5000000000*e[24]*ep2[3]-.5000000000*e[24]*ep2[2]+.5000000000*e[24]*ep2[8];
+        A[13]=e[6]*e[28]*e[23]-e[6]*e[22]*e[29]-e[6]*e[31]*e[20]-e[3]*e[19]*e[35]+e[3]*e[34]*e[20]+e[3]*e[25]*e[29]-e[21]*e[1]*e[35]+e[21]*e[7]*e[29]+e[21]*e[34]*e[2]+e[24]*e[1]*e[32]+e[24]*e[28]*e[5]-e[24]*e[4]*e[29]-e[24]*e[31]*e[2]+e[33]*e[1]*e[23]+e[33]*e[19]*e[5]-e[33]*e[4]*e[20]-e[33]*e[22]*e[2]-e[21]*e[28]*e[8]+e[30]*e[7]*e[20]+e[30]*e[25]*e[2]-e[30]*e[1]*e[26]+e[18]*e[4]*e[35]-e[18]*e[5]*e[34]+e[18]*e[31]*e[8]-e[18]*e[32]*e[7]+e[27]*e[4]*e[26]+e[27]*e[22]*e[8]-e[27]*e[5]*e[25]-e[27]*e[23]*e[7]-e[3]*e[28]*e[26]-e[0]*e[32]*e[25]+e[0]*e[22]*e[35]-e[0]*e[23]*e[34]+e[0]*e[31]*e[26]-e[30]*e[19]*e[8]+e[6]*e[19]*e[32];
+        A[159]=.5000000000*ep2[18]*e[6]+.5000000000*ep2[21]*e[6]+1.500000000*ep2[24]*e[6]+.5000000000*e[6]*ep2[26]-.5000000000*e[6]*ep2[23]-.5000000000*e[6]*ep2[19]-.5000000000*e[6]*ep2[20]-.5000000000*e[6]*ep2[22]+.5000000000*e[6]*ep2[25]+e[21]*e[3]*e[24]+e[18]*e[20]*e[8]+e[21]*e[4]*e[25]+e[18]*e[19]*e[7]+e[18]*e[1]*e[25]+e[21]*e[22]*e[7]+e[21]*e[23]*e[8]+e[18]*e[0]*e[24]+e[18]*e[2]*e[26]+e[21]*e[5]*e[26]+e[24]*e[26]*e[8]-1.*e[24]*e[20]*e[2]-1.*e[24]*e[19]*e[1]-1.*e[24]*e[22]*e[4]+e[24]*e[25]*e[7]-1.*e[24]*e[23]*e[5]+e[0]*e[19]*e[25]+e[0]*e[20]*e[26]+e[3]*e[22]*e[25]+e[3]*e[23]*e[26];
+        A[12]=e[18]*e[4]*e[8]+e[3]*e[7]*e[20]+e[3]*e[25]*e[2]-e[3]*e[1]*e[26]-e[18]*e[5]*e[7]+e[6]*e[1]*e[23]+e[6]*e[19]*e[5]-e[6]*e[4]*e[20]-e[6]*e[22]*e[2]+e[21]*e[7]*e[2]-e[21]*e[1]*e[8]+e[24]*e[1]*e[5]-e[24]*e[4]*e[2]-e[3]*e[19]*e[8]+e[0]*e[4]*e[26]+e[0]*e[22]*e[8]-e[0]*e[5]*e[25]-e[0]*e[23]*e[7];
+        A[158]=e[9]*e[1]*e[7]+e[9]*e[0]*e[6]+e[9]*e[2]*e[8]+e[3]*e[12]*e[6]+e[3]*e[13]*e[7]+e[3]*e[4]*e[16]+e[3]*e[14]*e[8]+e[3]*e[5]*e[17]+e[12]*e[4]*e[7]+e[12]*e[5]*e[8]+e[6]*e[16]*e[7]+e[6]*e[17]*e[8]-1.*e[6]*e[11]*e[2]-1.*e[6]*e[10]*e[1]-1.*e[6]*e[13]*e[4]-1.*e[6]*e[14]*e[5]+e[0]*e[10]*e[7]+e[0]*e[1]*e[16]+e[0]*e[11]*e[8]+e[0]*e[2]*e[17]+.5000000000*ep2[3]*e[15]+1.500000000*e[15]*ep2[6]+.5000000000*e[15]*ep2[7]+.5000000000*e[15]*ep2[8]+.5000000000*ep2[0]*e[15]-.5000000000*e[15]*ep2[4]-.5000000000*e[15]*ep2[5]-.5000000000*e[15]*ep2[1]-.5000000000*e[15]*ep2[2];
+        A[15]=-e[15]*e[13]*e[2]-e[6]*e[13]*e[11]-e[15]*e[4]*e[11]+e[12]*e[16]*e[2]-e[3]*e[10]*e[17]+e[3]*e[16]*e[11]+e[0]*e[13]*e[17]-e[0]*e[14]*e[16]+e[15]*e[1]*e[14]-e[12]*e[10]*e[8]+e[9]*e[4]*e[17]+e[9]*e[13]*e[8]-e[9]*e[5]*e[16]-e[9]*e[14]*e[7]+e[15]*e[10]*e[5]+e[12]*e[7]*e[11]+e[6]*e[10]*e[14]-e[12]*e[1]*e[17];
+        A[157]=e[12]*e[30]*e[24]+e[12]*e[21]*e[33]+e[12]*e[31]*e[25]+e[12]*e[22]*e[34]+e[12]*e[32]*e[26]+e[12]*e[23]*e[35]+e[9]*e[27]*e[24]+e[9]*e[18]*e[33]+e[9]*e[28]*e[25]+e[9]*e[19]*e[34]+e[9]*e[29]*e[26]+e[9]*e[20]*e[35]+e[21]*e[30]*e[15]+e[21]*e[32]*e[17]+e[21]*e[14]*e[35]+e[21]*e[31]*e[16]+e[21]*e[13]*e[34]+e[30]*e[23]*e[17]+e[30]*e[14]*e[26]+e[30]*e[22]*e[16]+e[30]*e[13]*e[25]+e[15]*e[27]*e[18]+3.*e[15]*e[33]*e[24]-1.*e[15]*e[29]*e[20]+e[15]*e[35]*e[26]-1.*e[15]*e[31]*e[22]-1.*e[15]*e[32]*e[23]-1.*e[15]*e[28]*e[19]+e[15]*e[34]*e[25]+e[18]*e[29]*e[17]+e[18]*e[11]*e[35]+e[18]*e[28]*e[16]+e[18]*e[10]*e[34]+e[27]*e[20]*e[17]+e[27]*e[11]*e[26]+e[27]*e[19]*e[16]+e[27]*e[10]*e[25]-1.*e[24]*e[28]*e[10]-1.*e[24]*e[31]*e[13]-1.*e[24]*e[32]*e[14]+e[24]*e[34]*e[16]+e[24]*e[35]*e[17]-1.*e[24]*e[29]*e[11]-1.*e[33]*e[23]*e[14]+e[33]*e[25]*e[16]+e[33]*e[26]*e[17]-1.*e[33]*e[20]*e[11]-1.*e[33]*e[19]*e[10]-1.*e[33]*e[22]*e[13];
+        A[14]=e[18]*e[13]*e[17]+e[9]*e[13]*e[26]+e[9]*e[22]*e[17]-e[9]*e[14]*e[25]-e[18]*e[14]*e[16]-e[15]*e[13]*e[20]-e[15]*e[22]*e[11]+e[12]*e[16]*e[20]+e[12]*e[25]*e[11]-e[12]*e[10]*e[26]-e[12]*e[19]*e[17]+e[21]*e[16]*e[11]-e[21]*e[10]*e[17]-e[9]*e[23]*e[16]+e[24]*e[10]*e[14]-e[24]*e[13]*e[11]+e[15]*e[10]*e[23]+e[15]*e[19]*e[14];
+        A[156]=e[21]*e[12]*e[24]+e[21]*e[23]*e[17]+e[21]*e[14]*e[26]+e[21]*e[22]*e[16]+e[21]*e[13]*e[25]+e[24]*e[26]*e[17]+e[24]*e[25]*e[16]+e[9]*e[19]*e[25]+e[9]*e[18]*e[24]+e[9]*e[20]*e[26]+e[12]*e[22]*e[25]+e[12]*e[23]*e[26]+e[18]*e[20]*e[17]+e[18]*e[11]*e[26]+e[18]*e[19]*e[16]+e[18]*e[10]*e[25]-1.*e[24]*e[23]*e[14]-1.*e[24]*e[20]*e[11]-1.*e[24]*e[19]*e[10]-1.*e[24]*e[22]*e[13]+.5000000000*ep2[21]*e[15]+1.500000000*ep2[24]*e[15]+.5000000000*e[15]*ep2[25]+.5000000000*e[15]*ep2[26]+.5000000000*e[15]*ep2[18]-.5000000000*e[15]*ep2[23]-.5000000000*e[15]*ep2[19]-.5000000000*e[15]*ep2[20]-.5000000000*e[15]*ep2[22];
+        A[18]=e[6]*e[1]*e[14]+e[15]*e[1]*e[5]-e[0]*e[5]*e[16]-e[0]*e[14]*e[7]+e[0]*e[13]*e[8]-e[15]*e[4]*e[2]+e[12]*e[7]*e[2]+e[6]*e[10]*e[5]+e[3]*e[7]*e[11]-e[6]*e[4]*e[11]+e[3]*e[16]*e[2]-e[6]*e[13]*e[2]-e[3]*e[1]*e[17]-e[9]*e[5]*e[7]-e[3]*e[10]*e[8]-e[12]*e[1]*e[8]+e[0]*e[4]*e[17]+e[9]*e[4]*e[8];
+        A[128]=-.5000000000*e[14]*ep2[16]-.5000000000*e[14]*ep2[10]-.5000000000*e[14]*ep2[9]+e[11]*e[9]*e[12]+.5000000000*ep3[14]+e[17]*e[13]*e[16]+.5000000000*e[14]*ep2[12]+e[11]*e[10]*e[13]-.5000000000*e[14]*ep2[15]+.5000000000*e[14]*ep2[17]+e[17]*e[12]*e[15]+.5000000000*ep2[11]*e[14]+.5000000000*e[14]*ep2[13];
+        A[19]=-e[21]*e[19]*e[8]+e[18]*e[4]*e[26]-e[18]*e[5]*e[25]-e[18]*e[23]*e[7]+e[21]*e[25]*e[2]-e[21]*e[1]*e[26]+e[6]*e[19]*e[23]+e[18]*e[22]*e[8]-e[0]*e[23]*e[25]-e[6]*e[22]*e[20]+e[24]*e[1]*e[23]+e[24]*e[19]*e[5]-e[24]*e[4]*e[20]-e[24]*e[22]*e[2]+e[3]*e[25]*e[20]-e[3]*e[19]*e[26]+e[0]*e[22]*e[26]+e[21]*e[7]*e[20];
+        A[129]=.5000000000*ep2[20]*e[32]+1.500000000*e[32]*ep2[23]+.5000000000*e[32]*ep2[22]+.5000000000*e[32]*ep2[21]+.5000000000*e[32]*ep2[26]-.5000000000*e[32]*ep2[18]-.5000000000*e[32]*ep2[19]-.5000000000*e[32]*ep2[24]-.5000000000*e[32]*ep2[25]+e[20]*e[27]*e[21]+e[20]*e[18]*e[30]+e[20]*e[28]*e[22]+e[20]*e[19]*e[31]+e[20]*e[29]*e[23]+e[29]*e[19]*e[22]+e[29]*e[18]*e[21]+e[23]*e[30]*e[21]+e[23]*e[31]*e[22]+e[26]*e[30]*e[24]+e[26]*e[21]*e[33]+e[26]*e[31]*e[25]+e[26]*e[22]*e[34]+e[26]*e[23]*e[35]+e[35]*e[22]*e[25]+e[35]*e[21]*e[24]-1.*e[23]*e[27]*e[18]-1.*e[23]*e[33]*e[24]-1.*e[23]*e[28]*e[19]-1.*e[23]*e[34]*e[25];
+        A[16]=-e[9]*e[23]*e[25]-e[21]*e[10]*e[26]-e[21]*e[19]*e[17]-e[18]*e[23]*e[16]+e[18]*e[13]*e[26]+e[12]*e[25]*e[20]-e[12]*e[19]*e[26]-e[15]*e[22]*e[20]+e[21]*e[16]*e[20]+e[21]*e[25]*e[11]+e[24]*e[10]*e[23]+e[24]*e[19]*e[14]-e[24]*e[13]*e[20]-e[24]*e[22]*e[11]+e[18]*e[22]*e[17]-e[18]*e[14]*e[25]+e[9]*e[22]*e[26]+e[15]*e[19]*e[23];
+        A[130]=.5000000000*e[23]*ep2[21]+e[20]*e[19]*e[22]+e[20]*e[18]*e[21]+.5000000000*ep3[23]+e[26]*e[22]*e[25]+.5000000000*e[23]*ep2[26]-.5000000000*e[23]*ep2[18]+.5000000000*e[23]*ep2[22]-.5000000000*e[23]*ep2[19]+e[26]*e[21]*e[24]+.5000000000*ep2[20]*e[23]-.5000000000*e[23]*ep2[24]-.5000000000*e[23]*ep2[25];
+        A[17]=e[18]*e[13]*e[35]-e[18]*e[32]*e[16]+e[18]*e[31]*e[17]-e[18]*e[14]*e[34]+e[27]*e[13]*e[26]+e[27]*e[22]*e[17]-e[27]*e[14]*e[25]-e[27]*e[23]*e[16]-e[9]*e[32]*e[25]+e[9]*e[22]*e[35]-e[9]*e[23]*e[34]+e[9]*e[31]*e[26]+e[15]*e[19]*e[32]+e[15]*e[28]*e[23]-e[15]*e[22]*e[29]-e[15]*e[31]*e[20]+e[24]*e[10]*e[32]+e[24]*e[28]*e[14]-e[24]*e[13]*e[29]-e[24]*e[31]*e[11]+e[33]*e[10]*e[23]+e[33]*e[19]*e[14]-e[33]*e[13]*e[20]-e[33]*e[22]*e[11]+e[21]*e[16]*e[29]-e[21]*e[10]*e[35]-e[21]*e[28]*e[17]+e[30]*e[16]*e[20]+e[30]*e[25]*e[11]-e[30]*e[10]*e[26]-e[30]*e[19]*e[17]-e[12]*e[28]*e[26]-e[12]*e[19]*e[35]+e[12]*e[34]*e[20]+e[12]*e[25]*e[29]+e[21]*e[34]*e[11];
+        A[131]=-1.*e[32]*e[10]*e[1]+e[32]*e[13]*e[4]-1.*e[32]*e[16]*e[7]-1.*e[32]*e[15]*e[6]-1.*e[32]*e[9]*e[0]+e[32]*e[12]*e[3]+e[17]*e[30]*e[6]+e[17]*e[3]*e[33]+e[17]*e[31]*e[7]+e[17]*e[4]*e[34]+e[17]*e[5]*e[35]-1.*e[5]*e[27]*e[9]-1.*e[5]*e[28]*e[10]-1.*e[5]*e[33]*e[15]-1.*e[5]*e[34]*e[16]+e[5]*e[29]*e[11]+e[35]*e[12]*e[6]+e[35]*e[3]*e[15]+e[35]*e[13]*e[7]+e[35]*e[4]*e[16]+e[11]*e[27]*e[3]+e[11]*e[0]*e[30]+e[11]*e[28]*e[4]+e[11]*e[1]*e[31]+e[29]*e[9]*e[3]+e[29]*e[0]*e[12]+e[29]*e[10]*e[4]+e[29]*e[1]*e[13]+e[5]*e[30]*e[12]+3.*e[5]*e[32]*e[14]+e[5]*e[31]*e[13]+e[8]*e[30]*e[15]+e[8]*e[12]*e[33]+e[8]*e[32]*e[17]+e[8]*e[14]*e[35]+e[8]*e[31]*e[16]+e[8]*e[13]*e[34]+e[2]*e[27]*e[12]+e[2]*e[9]*e[30]+e[2]*e[29]*e[14]+e[2]*e[11]*e[32]+e[2]*e[28]*e[13]+e[2]*e[10]*e[31]-1.*e[14]*e[27]*e[0]-1.*e[14]*e[34]*e[7]-1.*e[14]*e[33]*e[6]+e[14]*e[30]*e[3]-1.*e[14]*e[28]*e[1]+e[14]*e[31]*e[4];
+        A[22]=.5000000000*e[18]*ep2[29]+.5000000000*e[18]*ep2[28]+.5000000000*e[18]*ep2[30]+.5000000000*e[18]*ep2[33]-.5000000000*e[18]*ep2[32]-.5000000000*e[18]*ep2[31]-.5000000000*e[18]*ep2[34]-.5000000000*e[18]*ep2[35]+1.500000000*e[18]*ep2[27]+e[27]*e[28]*e[19]+e[27]*e[29]*e[20]+e[21]*e[27]*e[30]+e[21]*e[29]*e[32]+e[21]*e[28]*e[31]+e[30]*e[28]*e[22]+e[30]*e[19]*e[31]+e[30]*e[29]*e[23]+e[30]*e[20]*e[32]+e[24]*e[27]*e[33]+e[24]*e[29]*e[35]+e[24]*e[28]*e[34]+e[33]*e[28]*e[25]+e[33]*e[19]*e[34]+e[33]*e[29]*e[26]+e[33]*e[20]*e[35]-1.*e[27]*e[35]*e[26]-1.*e[27]*e[31]*e[22]-1.*e[27]*e[32]*e[23]-1.*e[27]*e[34]*e[25];
+        A[132]=e[20]*e[1]*e[4]+e[20]*e[0]*e[3]+e[20]*e[2]*e[5]+e[5]*e[21]*e[3]+e[5]*e[22]*e[4]+e[8]*e[21]*e[6]+e[8]*e[3]*e[24]+e[8]*e[22]*e[7]+e[8]*e[4]*e[25]+e[8]*e[5]*e[26]+e[26]*e[4]*e[7]+e[26]*e[3]*e[6]+e[2]*e[18]*e[3]+e[2]*e[0]*e[21]+e[2]*e[19]*e[4]+e[2]*e[1]*e[22]-1.*e[5]*e[19]*e[1]-1.*e[5]*e[18]*e[0]-1.*e[5]*e[25]*e[7]-1.*e[5]*e[24]*e[6]+.5000000000*e[23]*ep2[4]-.5000000000*e[23]*ep2[0]-.5000000000*e[23]*ep2[6]+1.500000000*e[23]*ep2[5]-.5000000000*e[23]*ep2[1]-.5000000000*e[23]*ep2[7]+.5000000000*e[23]*ep2[3]+.5000000000*e[23]*ep2[2]+.5000000000*e[23]*ep2[8];
+        A[23]=1.500000000*e[9]*ep2[27]+.5000000000*e[9]*ep2[29]+.5000000000*e[9]*ep2[28]-.5000000000*e[9]*ep2[32]-.5000000000*e[9]*ep2[31]+.5000000000*e[9]*ep2[33]+.5000000000*e[9]*ep2[30]-.5000000000*e[9]*ep2[34]-.5000000000*e[9]*ep2[35]+e[33]*e[27]*e[15]+e[33]*e[29]*e[17]+e[33]*e[11]*e[35]+e[33]*e[28]*e[16]+e[33]*e[10]*e[34]+e[27]*e[29]*e[11]+e[27]*e[28]*e[10]+e[27]*e[30]*e[12]-1.*e[27]*e[31]*e[13]-1.*e[27]*e[32]*e[14]-1.*e[27]*e[34]*e[16]-1.*e[27]*e[35]*e[17]+e[30]*e[29]*e[14]+e[30]*e[11]*e[32]+e[30]*e[28]*e[13]+e[30]*e[10]*e[31]+e[12]*e[29]*e[32]+e[12]*e[28]*e[31]+e[15]*e[29]*e[35]+e[15]*e[28]*e[34];
+        A[133]=-1.*e[32]*e[24]*e[6]+e[8]*e[30]*e[24]+e[8]*e[21]*e[33]+e[8]*e[31]*e[25]+e[8]*e[22]*e[34]+e[26]*e[30]*e[6]+e[26]*e[3]*e[33]+e[26]*e[31]*e[7]+e[26]*e[4]*e[34]+e[26]*e[32]*e[8]+e[26]*e[5]*e[35]+e[35]*e[21]*e[6]+e[35]*e[3]*e[24]+e[35]*e[22]*e[7]+e[35]*e[4]*e[25]+e[35]*e[23]*e[8]+e[2]*e[27]*e[21]+e[2]*e[18]*e[30]+e[2]*e[28]*e[22]+e[2]*e[19]*e[31]+e[2]*e[29]*e[23]+e[2]*e[20]*e[32]+e[20]*e[27]*e[3]+e[20]*e[0]*e[30]+e[20]*e[28]*e[4]+e[20]*e[1]*e[31]+e[20]*e[29]*e[5]+e[29]*e[18]*e[3]+e[29]*e[0]*e[21]+e[29]*e[19]*e[4]+e[29]*e[1]*e[22]+e[5]*e[30]*e[21]+e[5]*e[31]*e[22]+3.*e[5]*e[32]*e[23]-1.*e[5]*e[27]*e[18]-1.*e[5]*e[33]*e[24]-1.*e[5]*e[28]*e[19]-1.*e[5]*e[34]*e[25]-1.*e[23]*e[27]*e[0]-1.*e[23]*e[34]*e[7]-1.*e[23]*e[33]*e[6]+e[23]*e[30]*e[3]-1.*e[23]*e[28]*e[1]+e[23]*e[31]*e[4]+e[32]*e[21]*e[3]-1.*e[32]*e[19]*e[1]+e[32]*e[22]*e[4]-1.*e[32]*e[18]*e[0]-1.*e[32]*e[25]*e[7];
+        A[20]=.5000000000*e[27]*ep2[33]-.5000000000*e[27]*ep2[32]-.5000000000*e[27]*ep2[31]-.5000000000*e[27]*ep2[34]-.5000000000*e[27]*ep2[35]+e[33]*e[29]*e[35]+.5000000000*e[27]*ep2[29]+e[30]*e[29]*e[32]+e[30]*e[28]*e[31]+e[33]*e[28]*e[34]+.5000000000*e[27]*ep2[28]+.5000000000*e[27]*ep2[30]+.5000000000*ep3[27];
+        A[134]=e[14]*e[21]*e[12]+e[14]*e[22]*e[13]+e[17]*e[21]*e[15]+e[17]*e[12]*e[24]+e[17]*e[14]*e[26]+e[17]*e[22]*e[16]+e[17]*e[13]*e[25]+e[26]*e[12]*e[15]+e[26]*e[13]*e[16]-1.*e[14]*e[24]*e[15]-1.*e[14]*e[25]*e[16]-1.*e[14]*e[18]*e[9]-1.*e[14]*e[19]*e[10]+e[11]*e[18]*e[12]+e[11]*e[9]*e[21]+e[11]*e[19]*e[13]+e[11]*e[10]*e[22]+e[20]*e[11]*e[14]+e[20]*e[9]*e[12]+e[20]*e[10]*e[13]+1.500000000*e[23]*ep2[14]+.5000000000*e[23]*ep2[12]+.5000000000*e[23]*ep2[13]+.5000000000*e[23]*ep2[17]+.5000000000*ep2[11]*e[23]-.5000000000*e[23]*ep2[16]-.5000000000*e[23]*ep2[9]-.5000000000*e[23]*ep2[15]-.5000000000*e[23]*ep2[10];
+        A[21]=1.500000000*e[0]*ep2[27]+.5000000000*e[0]*ep2[29]+.5000000000*e[0]*ep2[28]+.5000000000*e[0]*ep2[30]-.5000000000*e[0]*ep2[32]-.5000000000*e[0]*ep2[31]+.5000000000*e[0]*ep2[33]-.5000000000*e[0]*ep2[34]-.5000000000*e[0]*ep2[35]-1.*e[27]*e[31]*e[4]+e[3]*e[27]*e[30]+e[3]*e[29]*e[32]+e[3]*e[28]*e[31]+e[30]*e[28]*e[4]+e[30]*e[1]*e[31]+e[30]*e[29]*e[5]+e[30]*e[2]*e[32]+e[6]*e[27]*e[33]+e[6]*e[29]*e[35]+e[6]*e[28]*e[34]+e[27]*e[28]*e[1]+e[27]*e[29]*e[2]+e[33]*e[28]*e[7]+e[33]*e[1]*e[34]+e[33]*e[29]*e[8]+e[33]*e[2]*e[35]-1.*e[27]*e[34]*e[7]-1.*e[27]*e[32]*e[5]-1.*e[27]*e[35]*e[8];
+        A[135]=e[14]*e[12]*e[3]+e[14]*e[13]*e[4]+e[17]*e[12]*e[6]+e[17]*e[3]*e[15]+e[17]*e[13]*e[7]+e[17]*e[4]*e[16]+e[17]*e[14]*e[8]+e[8]*e[12]*e[15]+e[8]*e[13]*e[16]+e[2]*e[11]*e[14]+e[2]*e[9]*e[12]+e[2]*e[10]*e[13]+e[11]*e[9]*e[3]+e[11]*e[0]*e[12]+e[11]*e[10]*e[4]+e[11]*e[1]*e[13]-1.*e[14]*e[10]*e[1]-1.*e[14]*e[16]*e[7]-1.*e[14]*e[15]*e[6]-1.*e[14]*e[9]*e[0]-.5000000000*e[5]*ep2[16]-.5000000000*e[5]*ep2[9]+.5000000000*e[5]*ep2[11]+.5000000000*e[5]*ep2[12]-.5000000000*e[5]*ep2[15]-.5000000000*e[5]*ep2[10]+.5000000000*e[5]*ep2[13]+1.500000000*ep2[14]*e[5]+.5000000000*e[5]*ep2[17];
+        A[27]=1.500000000*e[27]*ep2[9]-.5000000000*e[27]*ep2[16]+.5000000000*e[27]*ep2[11]+.5000000000*e[27]*ep2[12]+.5000000000*e[27]*ep2[15]-.5000000000*e[27]*ep2[17]+.5000000000*e[27]*ep2[10]-.5000000000*e[27]*ep2[14]-.5000000000*e[27]*ep2[13]+e[12]*e[10]*e[31]+e[30]*e[11]*e[14]+e[30]*e[10]*e[13]+e[15]*e[9]*e[33]+e[15]*e[29]*e[17]+e[15]*e[11]*e[35]+e[15]*e[28]*e[16]+e[15]*e[10]*e[34]+e[33]*e[11]*e[17]+e[33]*e[10]*e[16]-1.*e[9]*e[31]*e[13]-1.*e[9]*e[32]*e[14]-1.*e[9]*e[34]*e[16]-1.*e[9]*e[35]*e[17]+e[9]*e[29]*e[11]+e[9]*e[28]*e[10]+e[12]*e[9]*e[30]+e[12]*e[29]*e[14]+e[12]*e[11]*e[32]+e[12]*e[28]*e[13];
+        A[137]=e[29]*e[18]*e[12]+e[29]*e[9]*e[21]+e[29]*e[19]*e[13]+e[29]*e[10]*e[22]+e[17]*e[30]*e[24]+e[17]*e[21]*e[33]+e[17]*e[31]*e[25]+e[17]*e[22]*e[34]+e[17]*e[32]*e[26]+e[17]*e[23]*e[35]-1.*e[23]*e[27]*e[9]-1.*e[23]*e[28]*e[10]-1.*e[23]*e[33]*e[15]-1.*e[23]*e[34]*e[16]-1.*e[32]*e[24]*e[15]-1.*e[32]*e[25]*e[16]-1.*e[32]*e[18]*e[9]-1.*e[32]*e[19]*e[10]+e[26]*e[30]*e[15]+e[26]*e[12]*e[33]+e[26]*e[31]*e[16]+e[26]*e[13]*e[34]+e[35]*e[21]*e[15]+e[35]*e[12]*e[24]+e[35]*e[22]*e[16]+e[35]*e[13]*e[25]+e[14]*e[30]*e[21]+e[14]*e[31]*e[22]+3.*e[14]*e[32]*e[23]+e[11]*e[27]*e[21]+e[11]*e[18]*e[30]+e[11]*e[28]*e[22]+e[11]*e[19]*e[31]+e[11]*e[29]*e[23]+e[11]*e[20]*e[32]+e[23]*e[30]*e[12]+e[23]*e[31]*e[13]+e[32]*e[21]*e[12]+e[32]*e[22]*e[13]-1.*e[14]*e[27]*e[18]-1.*e[14]*e[33]*e[24]+e[14]*e[29]*e[20]+e[14]*e[35]*e[26]-1.*e[14]*e[28]*e[19]-1.*e[14]*e[34]*e[25]+e[20]*e[27]*e[12]+e[20]*e[9]*e[30]+e[20]*e[28]*e[13]+e[20]*e[10]*e[31];
+        A[26]=.5000000000*e[0]*ep2[1]+.5000000000*e[0]*ep2[2]+e[6]*e[2]*e[8]+e[6]*e[1]*e[7]+.5000000000*e[0]*ep2[3]+e[3]*e[1]*e[4]+.5000000000*e[0]*ep2[6]+e[3]*e[2]*e[5]-.5000000000*e[0]*ep2[5]-.5000000000*e[0]*ep2[8]+.5000000000*ep3[0]-.5000000000*e[0]*ep2[7]-.5000000000*e[0]*ep2[4];
+        A[136]=1.500000000*ep2[23]*e[14]+.5000000000*e[14]*ep2[26]-.5000000000*e[14]*ep2[18]-.5000000000*e[14]*ep2[19]+.5000000000*e[14]*ep2[20]+.5000000000*e[14]*ep2[22]-.5000000000*e[14]*ep2[24]+.5000000000*e[14]*ep2[21]-.5000000000*e[14]*ep2[25]+e[23]*e[21]*e[12]+e[23]*e[22]*e[13]+e[26]*e[21]*e[15]+e[26]*e[12]*e[24]+e[26]*e[23]*e[17]+e[26]*e[22]*e[16]+e[26]*e[13]*e[25]+e[17]*e[22]*e[25]+e[17]*e[21]*e[24]+e[11]*e[19]*e[22]+e[11]*e[18]*e[21]+e[11]*e[20]*e[23]+e[20]*e[18]*e[12]+e[20]*e[9]*e[21]+e[20]*e[19]*e[13]+e[20]*e[10]*e[22]-1.*e[23]*e[24]*e[15]-1.*e[23]*e[25]*e[16]-1.*e[23]*e[18]*e[9]-1.*e[23]*e[19]*e[10];
+        A[25]=1.500000000*e[27]*ep2[0]-.5000000000*e[27]*ep2[4]+.5000000000*e[27]*ep2[6]-.5000000000*e[27]*ep2[5]+.5000000000*e[27]*ep2[1]-.5000000000*e[27]*ep2[7]+.5000000000*e[27]*ep2[3]+.5000000000*e[27]*ep2[2]-.5000000000*e[27]*ep2[8]+e[0]*e[33]*e[6]+e[0]*e[30]*e[3]-1.*e[0]*e[35]*e[8]-1.*e[0]*e[31]*e[4]+e[3]*e[28]*e[4]+e[3]*e[1]*e[31]+e[3]*e[29]*e[5]+e[3]*e[2]*e[32]+e[30]*e[1]*e[4]+e[30]*e[2]*e[5]+e[6]*e[28]*e[7]+e[6]*e[1]*e[34]+e[6]*e[29]*e[8]+e[6]*e[2]*e[35]+e[33]*e[1]*e[7]+e[33]*e[2]*e[8]+e[0]*e[28]*e[1]+e[0]*e[29]*e[2]-1.*e[0]*e[34]*e[7]-1.*e[0]*e[32]*e[5];
+        A[139]=e[8]*e[22]*e[25]+e[8]*e[21]*e[24]+e[20]*e[18]*e[3]+e[20]*e[0]*e[21]+e[20]*e[19]*e[4]+e[20]*e[1]*e[22]+e[20]*e[2]*e[23]+e[23]*e[21]*e[3]+e[23]*e[22]*e[4]+e[23]*e[26]*e[8]-1.*e[23]*e[19]*e[1]-1.*e[23]*e[18]*e[0]-1.*e[23]*e[25]*e[7]-1.*e[23]*e[24]*e[6]+e[2]*e[19]*e[22]+e[2]*e[18]*e[21]+e[26]*e[21]*e[6]+e[26]*e[3]*e[24]+e[26]*e[22]*e[7]+e[26]*e[4]*e[25]+.5000000000*ep2[20]*e[5]+1.500000000*ep2[23]*e[5]+.5000000000*e[5]*ep2[22]+.5000000000*e[5]*ep2[21]+.5000000000*e[5]*ep2[26]-.5000000000*e[5]*ep2[18]-.5000000000*e[5]*ep2[19]-.5000000000*e[5]*ep2[24]-.5000000000*e[5]*ep2[25];
+        A[24]=e[24]*e[11]*e[8]+e[24]*e[2]*e[17]+3.*e[9]*e[18]*e[0]+e[9]*e[19]*e[1]+e[9]*e[20]*e[2]+e[18]*e[10]*e[1]+e[18]*e[11]*e[2]+e[3]*e[18]*e[12]+e[3]*e[9]*e[21]+e[3]*e[20]*e[14]+e[3]*e[11]*e[23]+e[3]*e[19]*e[13]+e[3]*e[10]*e[22]+e[6]*e[18]*e[15]+e[6]*e[9]*e[24]+e[6]*e[20]*e[17]+e[6]*e[11]*e[26]+e[6]*e[19]*e[16]+e[6]*e[10]*e[25]+e[0]*e[20]*e[11]+e[0]*e[19]*e[10]-1.*e[9]*e[26]*e[8]-1.*e[9]*e[22]*e[4]-1.*e[9]*e[25]*e[7]-1.*e[9]*e[23]*e[5]+e[12]*e[0]*e[21]+e[12]*e[19]*e[4]+e[12]*e[1]*e[22]+e[12]*e[20]*e[5]+e[12]*e[2]*e[23]-1.*e[18]*e[13]*e[4]-1.*e[18]*e[16]*e[7]-1.*e[18]*e[14]*e[5]-1.*e[18]*e[17]*e[8]+e[21]*e[10]*e[4]+e[21]*e[1]*e[13]+e[21]*e[11]*e[5]+e[21]*e[2]*e[14]+e[15]*e[0]*e[24]+e[15]*e[19]*e[7]+e[15]*e[1]*e[25]+e[15]*e[20]*e[8]+e[15]*e[2]*e[26]-1.*e[0]*e[23]*e[14]-1.*e[0]*e[25]*e[16]-1.*e[0]*e[26]*e[17]-1.*e[0]*e[22]*e[13]+e[24]*e[10]*e[7]+e[24]*e[1]*e[16];
+        A[138]=e[11]*e[1]*e[4]+e[11]*e[0]*e[3]+e[11]*e[2]*e[5]+e[5]*e[12]*e[3]+e[5]*e[13]*e[4]+e[8]*e[12]*e[6]+e[8]*e[3]*e[15]+e[8]*e[13]*e[7]+e[8]*e[4]*e[16]+e[8]*e[5]*e[17]+e[17]*e[4]*e[7]+e[17]*e[3]*e[6]-1.*e[5]*e[10]*e[1]-1.*e[5]*e[16]*e[7]-1.*e[5]*e[15]*e[6]-1.*e[5]*e[9]*e[0]+e[2]*e[9]*e[3]+e[2]*e[0]*e[12]+e[2]*e[10]*e[4]+e[2]*e[1]*e[13]+.5000000000*ep2[2]*e[14]-.5000000000*e[14]*ep2[0]-.5000000000*e[14]*ep2[6]-.5000000000*e[14]*ep2[1]-.5000000000*e[14]*ep2[7]+1.500000000*e[14]*ep2[5]+.5000000000*e[14]*ep2[4]+.5000000000*e[14]*ep2[3]+.5000000000*e[14]*ep2[8];
+        A[31]=e[3]*e[27]*e[12]+e[3]*e[9]*e[30]+e[3]*e[29]*e[14]+e[3]*e[11]*e[32]+e[3]*e[28]*e[13]+e[3]*e[10]*e[31]+e[6]*e[27]*e[15]+e[6]*e[9]*e[33]+e[6]*e[29]*e[17]+e[6]*e[11]*e[35]+e[6]*e[28]*e[16]+e[6]*e[10]*e[34]+3.*e[0]*e[27]*e[9]+e[0]*e[29]*e[11]+e[0]*e[28]*e[10]-1.*e[9]*e[34]*e[7]-1.*e[9]*e[32]*e[5]-1.*e[9]*e[35]*e[8]+e[9]*e[29]*e[2]+e[9]*e[28]*e[1]-1.*e[9]*e[31]*e[4]+e[12]*e[0]*e[30]+e[12]*e[28]*e[4]+e[12]*e[1]*e[31]+e[12]*e[29]*e[5]+e[12]*e[2]*e[32]+e[27]*e[11]*e[2]+e[27]*e[10]*e[1]-1.*e[27]*e[13]*e[4]-1.*e[27]*e[16]*e[7]-1.*e[27]*e[14]*e[5]-1.*e[27]*e[17]*e[8]+e[30]*e[10]*e[4]+e[30]*e[1]*e[13]+e[30]*e[11]*e[5]+e[30]*e[2]*e[14]+e[15]*e[0]*e[33]+e[15]*e[28]*e[7]+e[15]*e[1]*e[34]+e[15]*e[29]*e[8]+e[15]*e[2]*e[35]-1.*e[0]*e[31]*e[13]-1.*e[0]*e[32]*e[14]-1.*e[0]*e[34]*e[16]-1.*e[0]*e[35]*e[17]+e[33]*e[10]*e[7]+e[33]*e[1]*e[16]+e[33]*e[11]*e[8]+e[33]*e[2]*e[17];
+        A[141]=.5000000000*ep2[30]*e[6]+.5000000000*e[6]*ep2[27]-.5000000000*e[6]*ep2[32]-.5000000000*e[6]*ep2[28]-.5000000000*e[6]*ep2[29]-.5000000000*e[6]*ep2[31]+1.500000000*e[6]*ep2[33]+.5000000000*e[6]*ep2[34]+.5000000000*e[6]*ep2[35]+e[0]*e[27]*e[33]+e[0]*e[29]*e[35]+e[0]*e[28]*e[34]+e[3]*e[30]*e[33]+e[3]*e[32]*e[35]+e[3]*e[31]*e[34]+e[30]*e[31]*e[7]+e[30]*e[4]*e[34]+e[30]*e[32]*e[8]+e[30]*e[5]*e[35]+e[27]*e[28]*e[7]+e[27]*e[1]*e[34]+e[27]*e[29]*e[8]+e[27]*e[2]*e[35]+e[33]*e[34]*e[7]+e[33]*e[35]*e[8]-1.*e[33]*e[32]*e[5]-1.*e[33]*e[29]*e[2]-1.*e[33]*e[28]*e[1]-1.*e[33]*e[31]*e[4];
+        A[30]=e[24]*e[20]*e[26]+e[21]*e[19]*e[22]-.5000000000*e[18]*ep2[22]-.5000000000*e[18]*ep2[25]+.5000000000*ep3[18]+.5000000000*e[18]*ep2[21]+e[21]*e[20]*e[23]+.5000000000*e[18]*ep2[20]+.5000000000*e[18]*ep2[19]+.5000000000*e[18]*ep2[24]+e[24]*e[19]*e[25]-.5000000000*e[18]*ep2[23]-.5000000000*e[18]*ep2[26];
+        A[140]=.5000000000*e[33]*ep2[35]+.5000000000*ep3[33]+.5000000000*ep2[27]*e[33]+.5000000000*ep2[30]*e[33]-.5000000000*e[33]*ep2[29]+.5000000000*e[33]*ep2[34]-.5000000000*e[33]*ep2[32]-.5000000000*e[33]*ep2[28]+e[30]*e[32]*e[35]-.5000000000*e[33]*ep2[31]+e[27]*e[29]*e[35]+e[27]*e[28]*e[34]+e[30]*e[31]*e[34];
+        A[29]=1.500000000*e[27]*ep2[18]+.5000000000*e[27]*ep2[19]+.5000000000*e[27]*ep2[20]+.5000000000*e[27]*ep2[21]+.5000000000*e[27]*ep2[24]-.5000000000*e[27]*ep2[26]-.5000000000*e[27]*ep2[23]-.5000000000*e[27]*ep2[22]-.5000000000*e[27]*ep2[25]+e[33]*e[20]*e[26]-1.*e[18]*e[35]*e[26]-1.*e[18]*e[31]*e[22]-1.*e[18]*e[32]*e[23]-1.*e[18]*e[34]*e[25]+e[18]*e[28]*e[19]+e[18]*e[29]*e[20]+e[21]*e[18]*e[30]+e[21]*e[28]*e[22]+e[21]*e[19]*e[31]+e[21]*e[29]*e[23]+e[21]*e[20]*e[32]+e[30]*e[19]*e[22]+e[30]*e[20]*e[23]+e[24]*e[18]*e[33]+e[24]*e[28]*e[25]+e[24]*e[19]*e[34]+e[24]*e[29]*e[26]+e[24]*e[20]*e[35]+e[33]*e[19]*e[25];
+        A[143]=e[9]*e[27]*e[33]+e[9]*e[29]*e[35]+e[9]*e[28]*e[34]+e[33]*e[35]*e[17]+e[33]*e[34]*e[16]+e[27]*e[29]*e[17]+e[27]*e[11]*e[35]+e[27]*e[28]*e[16]+e[27]*e[10]*e[34]+e[33]*e[30]*e[12]-1.*e[33]*e[28]*e[10]-1.*e[33]*e[31]*e[13]-1.*e[33]*e[32]*e[14]-1.*e[33]*e[29]*e[11]+e[30]*e[32]*e[17]+e[30]*e[14]*e[35]+e[30]*e[31]*e[16]+e[30]*e[13]*e[34]+e[12]*e[32]*e[35]+e[12]*e[31]*e[34]+.5000000000*e[15]*ep2[27]-.5000000000*e[15]*ep2[32]-.5000000000*e[15]*ep2[28]-.5000000000*e[15]*ep2[29]-.5000000000*e[15]*ep2[31]+1.500000000*e[15]*ep2[33]+.5000000000*e[15]*ep2[30]+.5000000000*e[15]*ep2[34]+.5000000000*e[15]*ep2[35];
+        A[28]=.5000000000*e[9]*ep2[12]-.5000000000*e[9]*ep2[16]+.5000000000*e[9]*ep2[10]-.5000000000*e[9]*ep2[17]-.5000000000*e[9]*ep2[13]+e[15]*e[10]*e[16]+e[12]*e[11]*e[14]+.5000000000*e[9]*ep2[11]+.5000000000*e[9]*ep2[15]-.5000000000*e[9]*ep2[14]+e[15]*e[11]*e[17]+.5000000000*ep3[9]+e[12]*e[10]*e[13];
+        A[142]=e[18]*e[27]*e[33]+e[18]*e[29]*e[35]+e[18]*e[28]*e[34]+e[27]*e[28]*e[25]+e[27]*e[19]*e[34]+e[27]*e[29]*e[26]+e[27]*e[20]*e[35]+e[21]*e[30]*e[33]+e[21]*e[32]*e[35]+e[21]*e[31]*e[34]+e[30]*e[31]*e[25]+e[30]*e[22]*e[34]+e[30]*e[32]*e[26]+e[30]*e[23]*e[35]+e[33]*e[34]*e[25]+e[33]*e[35]*e[26]-1.*e[33]*e[29]*e[20]-1.*e[33]*e[31]*e[22]-1.*e[33]*e[32]*e[23]-1.*e[33]*e[28]*e[19]+.5000000000*ep2[27]*e[24]+.5000000000*ep2[30]*e[24]+1.500000000*e[24]*ep2[33]+.5000000000*e[24]*ep2[35]+.5000000000*e[24]*ep2[34]-.5000000000*e[24]*ep2[32]-.5000000000*e[24]*ep2[28]-.5000000000*e[24]*ep2[29]-.5000000000*e[24]*ep2[31];
+        A[36]=.5000000000*e[9]*ep2[21]+.5000000000*e[9]*ep2[24]+.5000000000*e[9]*ep2[19]+1.500000000*e[9]*ep2[18]+.5000000000*e[9]*ep2[20]-.5000000000*e[9]*ep2[26]-.5000000000*e[9]*ep2[23]-.5000000000*e[9]*ep2[22]-.5000000000*e[9]*ep2[25]+e[21]*e[18]*e[12]+e[21]*e[20]*e[14]+e[21]*e[11]*e[23]+e[21]*e[19]*e[13]+e[21]*e[10]*e[22]+e[24]*e[18]*e[15]+e[24]*e[20]*e[17]+e[24]*e[11]*e[26]+e[24]*e[19]*e[16]+e[24]*e[10]*e[25]+e[15]*e[19]*e[25]+e[15]*e[20]*e[26]+e[12]*e[19]*e[22]+e[12]*e[20]*e[23]+e[18]*e[20]*e[11]+e[18]*e[19]*e[10]-1.*e[18]*e[23]*e[14]-1.*e[18]*e[25]*e[16]-1.*e[18]*e[26]*e[17]-1.*e[18]*e[22]*e[13];
+        A[182]=.5000000000*ep2[29]*e[26]+.5000000000*ep2[32]*e[26]+.5000000000*e[26]*ep2[33]+1.500000000*e[26]*ep2[35]+.5000000000*e[26]*ep2[34]-.5000000000*e[26]*ep2[27]-.5000000000*e[26]*ep2[28]-.5000000000*e[26]*ep2[31]-.5000000000*e[26]*ep2[30]+e[20]*e[27]*e[33]+e[20]*e[29]*e[35]+e[20]*e[28]*e[34]+e[29]*e[27]*e[24]+e[29]*e[18]*e[33]+e[29]*e[28]*e[25]+e[29]*e[19]*e[34]+e[23]*e[30]*e[33]+e[23]*e[32]*e[35]+e[23]*e[31]*e[34]+e[32]*e[30]*e[24]+e[32]*e[21]*e[33]+e[32]*e[31]*e[25]+e[32]*e[22]*e[34]+e[35]*e[33]*e[24]+e[35]*e[34]*e[25]-1.*e[35]*e[27]*e[18]-1.*e[35]*e[30]*e[21]-1.*e[35]*e[31]*e[22]-1.*e[35]*e[28]*e[19];
+        A[37]=e[12]*e[19]*e[31]+e[12]*e[29]*e[23]+e[12]*e[20]*e[32]+3.*e[9]*e[27]*e[18]+e[9]*e[28]*e[19]+e[9]*e[29]*e[20]+e[21]*e[9]*e[30]+e[21]*e[29]*e[14]+e[21]*e[11]*e[32]+e[21]*e[28]*e[13]+e[21]*e[10]*e[31]+e[30]*e[20]*e[14]+e[30]*e[11]*e[23]+e[30]*e[19]*e[13]+e[30]*e[10]*e[22]+e[9]*e[33]*e[24]-1.*e[9]*e[35]*e[26]-1.*e[9]*e[31]*e[22]-1.*e[9]*e[32]*e[23]-1.*e[9]*e[34]*e[25]+e[18]*e[29]*e[11]+e[18]*e[28]*e[10]+e[27]*e[20]*e[11]+e[27]*e[19]*e[10]+e[15]*e[27]*e[24]+e[15]*e[18]*e[33]+e[15]*e[28]*e[25]+e[15]*e[19]*e[34]+e[15]*e[29]*e[26]+e[15]*e[20]*e[35]-1.*e[18]*e[31]*e[13]-1.*e[18]*e[32]*e[14]-1.*e[18]*e[34]*e[16]-1.*e[18]*e[35]*e[17]-1.*e[27]*e[23]*e[14]-1.*e[27]*e[25]*e[16]-1.*e[27]*e[26]*e[17]-1.*e[27]*e[22]*e[13]+e[24]*e[29]*e[17]+e[24]*e[11]*e[35]+e[24]*e[28]*e[16]+e[24]*e[10]*e[34]+e[33]*e[20]*e[17]+e[33]*e[11]*e[26]+e[33]*e[19]*e[16]+e[33]*e[10]*e[25]+e[12]*e[27]*e[21]+e[12]*e[18]*e[30]+e[12]*e[28]*e[22];
+        A[183]=-.5000000000*e[17]*ep2[27]+.5000000000*e[17]*ep2[32]-.5000000000*e[17]*ep2[28]+.5000000000*e[17]*ep2[29]-.5000000000*e[17]*ep2[31]+.5000000000*e[17]*ep2[33]-.5000000000*e[17]*ep2[30]+.5000000000*e[17]*ep2[34]+1.500000000*e[17]*ep2[35]+e[32]*e[30]*e[15]+e[32]*e[12]*e[33]+e[32]*e[31]*e[16]+e[32]*e[13]*e[34]+e[14]*e[30]*e[33]+e[14]*e[31]*e[34]+e[11]*e[27]*e[33]+e[11]*e[29]*e[35]+e[11]*e[28]*e[34]+e[35]*e[33]*e[15]+e[35]*e[34]*e[16]+e[29]*e[27]*e[15]+e[29]*e[9]*e[33]+e[29]*e[28]*e[16]+e[29]*e[10]*e[34]-1.*e[35]*e[27]*e[9]-1.*e[35]*e[30]*e[12]-1.*e[35]*e[28]*e[10]-1.*e[35]*e[31]*e[13]+e[35]*e[32]*e[14];
+        A[38]=.5000000000*e[9]*ep2[1]+1.500000000*e[9]*ep2[0]+.5000000000*e[9]*ep2[2]+.5000000000*e[9]*ep2[3]+.5000000000*e[9]*ep2[6]-.5000000000*e[9]*ep2[4]-.5000000000*e[9]*ep2[5]-.5000000000*e[9]*ep2[7]-.5000000000*e[9]*ep2[8]+e[6]*e[0]*e[15]+e[6]*e[10]*e[7]+e[6]*e[1]*e[16]+e[6]*e[11]*e[8]+e[6]*e[2]*e[17]+e[15]*e[1]*e[7]+e[15]*e[2]*e[8]+e[0]*e[11]*e[2]+e[0]*e[10]*e[1]-1.*e[0]*e[13]*e[4]-1.*e[0]*e[16]*e[7]-1.*e[0]*e[14]*e[5]-1.*e[0]*e[17]*e[8]+e[3]*e[0]*e[12]+e[3]*e[10]*e[4]+e[3]*e[1]*e[13]+e[3]*e[11]*e[5]+e[3]*e[2]*e[14]+e[12]*e[1]*e[4]+e[12]*e[2]*e[5];
+        A[180]=.5000000000*e[35]*ep2[33]+.5000000000*e[35]*ep2[34]-.5000000000*e[35]*ep2[27]-.5000000000*e[35]*ep2[28]-.5000000000*e[35]*ep2[31]-.5000000000*e[35]*ep2[30]+e[32]*e[31]*e[34]+.5000000000*ep2[29]*e[35]+.5000000000*ep2[32]*e[35]+e[29]*e[28]*e[34]+e[32]*e[30]*e[33]+.5000000000*ep3[35]+e[29]*e[27]*e[33];
+        A[39]=.5000000000*e[0]*ep2[19]+.5000000000*e[0]*ep2[20]+.5000000000*e[0]*ep2[24]-.5000000000*e[0]*ep2[26]-.5000000000*e[0]*ep2[23]-.5000000000*e[0]*ep2[22]-.5000000000*e[0]*ep2[25]+1.500000000*ep2[18]*e[0]+.5000000000*e[0]*ep2[21]+e[18]*e[19]*e[1]+e[18]*e[20]*e[2]+e[21]*e[18]*e[3]+e[21]*e[19]*e[4]+e[21]*e[1]*e[22]+e[21]*e[20]*e[5]+e[21]*e[2]*e[23]-1.*e[18]*e[26]*e[8]-1.*e[18]*e[22]*e[4]-1.*e[18]*e[25]*e[7]-1.*e[18]*e[23]*e[5]+e[18]*e[24]*e[6]+e[3]*e[19]*e[22]+e[3]*e[20]*e[23]+e[24]*e[19]*e[7]+e[24]*e[1]*e[25]+e[24]*e[20]*e[8]+e[24]*e[2]*e[26]+e[6]*e[19]*e[25]+e[6]*e[20]*e[26];
+        A[181]=.5000000000*ep2[32]*e[8]-.5000000000*e[8]*ep2[27]-.5000000000*e[8]*ep2[28]+.5000000000*e[8]*ep2[29]-.5000000000*e[8]*ep2[31]+.5000000000*e[8]*ep2[33]-.5000000000*e[8]*ep2[30]+.5000000000*e[8]*ep2[34]+1.500000000*e[8]*ep2[35]+e[2]*e[27]*e[33]+e[2]*e[29]*e[35]+e[2]*e[28]*e[34]+e[5]*e[30]*e[33]+e[5]*e[32]*e[35]+e[5]*e[31]*e[34]+e[32]*e[30]*e[6]+e[32]*e[3]*e[33]+e[32]*e[31]*e[7]+e[32]*e[4]*e[34]+e[29]*e[27]*e[6]+e[29]*e[0]*e[33]+e[29]*e[28]*e[7]+e[29]*e[1]*e[34]+e[35]*e[33]*e[6]+e[35]*e[34]*e[7]-1.*e[35]*e[27]*e[0]-1.*e[35]*e[30]*e[3]-1.*e[35]*e[28]*e[1]-1.*e[35]*e[31]*e[4];
+        A[32]=-.5000000000*e[18]*ep2[4]+1.500000000*e[18]*ep2[0]+.5000000000*e[18]*ep2[6]-.5000000000*e[18]*ep2[5]+.5000000000*e[18]*ep2[1]-.5000000000*e[18]*ep2[7]+.5000000000*e[18]*ep2[3]+.5000000000*e[18]*ep2[2]-.5000000000*e[18]*ep2[8]+e[3]*e[0]*e[21]+e[3]*e[19]*e[4]+e[3]*e[1]*e[22]+e[3]*e[20]*e[5]+e[3]*e[2]*e[23]+e[21]*e[1]*e[4]+e[21]*e[2]*e[5]+e[6]*e[0]*e[24]+e[6]*e[19]*e[7]+e[6]*e[1]*e[25]+e[6]*e[20]*e[8]+e[6]*e[2]*e[26]+e[24]*e[1]*e[7]+e[24]*e[2]*e[8]+e[0]*e[19]*e[1]+e[0]*e[20]*e[2]-1.*e[0]*e[26]*e[8]-1.*e[0]*e[22]*e[4]-1.*e[0]*e[25]*e[7]-1.*e[0]*e[23]*e[5];
+        A[178]=e[10]*e[1]*e[7]+e[10]*e[0]*e[6]+e[10]*e[2]*e[8]+e[4]*e[12]*e[6]+e[4]*e[3]*e[15]+e[4]*e[13]*e[7]+e[4]*e[14]*e[8]+e[4]*e[5]*e[17]+e[13]*e[3]*e[6]+e[13]*e[5]*e[8]+e[7]*e[15]*e[6]+e[7]*e[17]*e[8]-1.*e[7]*e[11]*e[2]-1.*e[7]*e[9]*e[0]-1.*e[7]*e[14]*e[5]-1.*e[7]*e[12]*e[3]+e[1]*e[9]*e[6]+e[1]*e[0]*e[15]+e[1]*e[11]*e[8]+e[1]*e[2]*e[17]+1.500000000*e[16]*ep2[7]+.5000000000*e[16]*ep2[6]+.5000000000*e[16]*ep2[8]+.5000000000*ep2[1]*e[16]-.5000000000*e[16]*ep2[0]-.5000000000*e[16]*ep2[5]-.5000000000*e[16]*ep2[3]-.5000000000*e[16]*ep2[2]+.5000000000*ep2[4]*e[16];
+        A[33]=e[0]*e[30]*e[21]-1.*e[0]*e[35]*e[26]-1.*e[0]*e[31]*e[22]-1.*e[0]*e[32]*e[23]-1.*e[0]*e[34]*e[25]-1.*e[18]*e[34]*e[7]-1.*e[18]*e[32]*e[5]-1.*e[18]*e[35]*e[8]-1.*e[18]*e[31]*e[4]-1.*e[27]*e[26]*e[8]-1.*e[27]*e[22]*e[4]-1.*e[27]*e[25]*e[7]-1.*e[27]*e[23]*e[5]+e[6]*e[28]*e[25]+e[6]*e[19]*e[34]+e[6]*e[29]*e[26]+e[6]*e[20]*e[35]+e[21]*e[28]*e[4]+e[21]*e[1]*e[31]+e[21]*e[29]*e[5]+e[21]*e[2]*e[32]+e[30]*e[19]*e[4]+e[30]*e[1]*e[22]+e[30]*e[20]*e[5]+e[30]*e[2]*e[23]+e[24]*e[27]*e[6]+e[24]*e[0]*e[33]+e[24]*e[28]*e[7]+e[24]*e[1]*e[34]+e[24]*e[29]*e[8]+e[24]*e[2]*e[35]+e[33]*e[18]*e[6]+e[33]*e[19]*e[7]+e[33]*e[1]*e[25]+e[33]*e[20]*e[8]+e[33]*e[2]*e[26]+3.*e[0]*e[27]*e[18]+e[0]*e[28]*e[19]+e[0]*e[29]*e[20]+e[18]*e[28]*e[1]+e[18]*e[29]*e[2]+e[27]*e[19]*e[1]+e[27]*e[20]*e[2]+e[3]*e[27]*e[21]+e[3]*e[18]*e[30]+e[3]*e[28]*e[22]+e[3]*e[19]*e[31]+e[3]*e[29]*e[23]+e[3]*e[20]*e[32];
+        A[179]=e[19]*e[18]*e[6]+e[19]*e[0]*e[24]+e[19]*e[1]*e[25]+e[19]*e[20]*e[8]+e[19]*e[2]*e[26]+e[22]*e[21]*e[6]+e[22]*e[3]*e[24]+e[22]*e[4]*e[25]+e[22]*e[23]*e[8]+e[22]*e[5]*e[26]-1.*e[25]*e[21]*e[3]+e[25]*e[26]*e[8]-1.*e[25]*e[20]*e[2]-1.*e[25]*e[18]*e[0]-1.*e[25]*e[23]*e[5]+e[25]*e[24]*e[6]+e[1]*e[18]*e[24]+e[1]*e[20]*e[26]+e[4]*e[21]*e[24]+e[4]*e[23]*e[26]+.5000000000*ep2[19]*e[7]+.5000000000*ep2[22]*e[7]+1.500000000*ep2[25]*e[7]+.5000000000*e[7]*ep2[26]-.5000000000*e[7]*ep2[18]-.5000000000*e[7]*ep2[23]-.5000000000*e[7]*ep2[20]+.5000000000*e[7]*ep2[24]-.5000000000*e[7]*ep2[21];
+        A[34]=.5000000000*e[18]*ep2[11]+1.500000000*e[18]*ep2[9]+.5000000000*e[18]*ep2[10]+.5000000000*e[18]*ep2[12]+.5000000000*e[18]*ep2[15]-.5000000000*e[18]*ep2[16]-.5000000000*e[18]*ep2[17]-.5000000000*e[18]*ep2[14]-.5000000000*e[18]*ep2[13]+e[12]*e[9]*e[21]+e[12]*e[20]*e[14]+e[12]*e[11]*e[23]+e[12]*e[19]*e[13]+e[12]*e[10]*e[22]+e[21]*e[11]*e[14]+e[21]*e[10]*e[13]+e[15]*e[9]*e[24]+e[15]*e[20]*e[17]+e[15]*e[11]*e[26]+e[15]*e[19]*e[16]+e[15]*e[10]*e[25]+e[24]*e[11]*e[17]+e[24]*e[10]*e[16]-1.*e[9]*e[23]*e[14]-1.*e[9]*e[25]*e[16]-1.*e[9]*e[26]*e[17]+e[9]*e[20]*e[11]+e[9]*e[19]*e[10]-1.*e[9]*e[22]*e[13];
+        A[176]=e[13]*e[21]*e[24]+e[13]*e[23]*e[26]+e[19]*e[18]*e[15]+e[19]*e[9]*e[24]+e[19]*e[20]*e[17]+e[19]*e[11]*e[26]-1.*e[25]*e[23]*e[14]-1.*e[25]*e[20]*e[11]-1.*e[25]*e[18]*e[9]-1.*e[25]*e[21]*e[12]+e[22]*e[21]*e[15]+e[22]*e[12]*e[24]+e[22]*e[23]*e[17]+e[22]*e[14]*e[26]+e[22]*e[13]*e[25]+e[25]*e[24]*e[15]+e[25]*e[26]*e[17]+e[10]*e[19]*e[25]+e[10]*e[18]*e[24]+e[10]*e[20]*e[26]-.5000000000*e[16]*ep2[18]-.5000000000*e[16]*ep2[23]+.5000000000*e[16]*ep2[19]-.5000000000*e[16]*ep2[20]-.5000000000*e[16]*ep2[21]+.5000000000*ep2[22]*e[16]+1.500000000*ep2[25]*e[16]+.5000000000*e[16]*ep2[24]+.5000000000*e[16]*ep2[26];
+        A[35]=.5000000000*e[0]*ep2[12]+.5000000000*e[0]*ep2[15]+.5000000000*e[0]*ep2[11]+1.500000000*e[0]*ep2[9]+.5000000000*e[0]*ep2[10]-.5000000000*e[0]*ep2[16]-.5000000000*e[0]*ep2[17]-.5000000000*e[0]*ep2[14]-.5000000000*e[0]*ep2[13]+e[12]*e[9]*e[3]+e[12]*e[10]*e[4]+e[12]*e[1]*e[13]+e[12]*e[11]*e[5]+e[12]*e[2]*e[14]+e[15]*e[9]*e[6]+e[15]*e[10]*e[7]+e[15]*e[1]*e[16]+e[15]*e[11]*e[8]+e[15]*e[2]*e[17]+e[6]*e[11]*e[17]+e[6]*e[10]*e[16]+e[3]*e[11]*e[14]+e[3]*e[10]*e[13]+e[9]*e[10]*e[1]+e[9]*e[11]*e[2]-1.*e[9]*e[13]*e[4]-1.*e[9]*e[16]*e[7]-1.*e[9]*e[14]*e[5]-1.*e[9]*e[17]*e[8];
+        A[177]=e[19]*e[11]*e[35]+e[28]*e[18]*e[15]+e[28]*e[9]*e[24]+e[28]*e[20]*e[17]+e[28]*e[11]*e[26]-1.*e[25]*e[27]*e[9]-1.*e[25]*e[30]*e[12]-1.*e[25]*e[32]*e[14]+e[25]*e[33]*e[15]+e[25]*e[35]*e[17]-1.*e[25]*e[29]*e[11]-1.*e[34]*e[23]*e[14]+e[34]*e[24]*e[15]+e[34]*e[26]*e[17]-1.*e[34]*e[20]*e[11]-1.*e[34]*e[18]*e[9]-1.*e[34]*e[21]*e[12]+e[13]*e[30]*e[24]+e[13]*e[21]*e[33]+e[13]*e[31]*e[25]+e[13]*e[22]*e[34]+e[13]*e[32]*e[26]+e[13]*e[23]*e[35]+e[10]*e[27]*e[24]+e[10]*e[18]*e[33]+e[10]*e[28]*e[25]+e[10]*e[19]*e[34]+e[10]*e[29]*e[26]+e[10]*e[20]*e[35]+e[22]*e[30]*e[15]+e[22]*e[12]*e[33]+e[22]*e[32]*e[17]+e[22]*e[14]*e[35]+e[22]*e[31]*e[16]+e[31]*e[21]*e[15]+e[31]*e[12]*e[24]+e[31]*e[23]*e[17]+e[31]*e[14]*e[26]-1.*e[16]*e[27]*e[18]+e[16]*e[33]*e[24]-1.*e[16]*e[30]*e[21]-1.*e[16]*e[29]*e[20]+e[16]*e[35]*e[26]-1.*e[16]*e[32]*e[23]+e[16]*e[28]*e[19]+3.*e[16]*e[34]*e[25]+e[19]*e[27]*e[15]+e[19]*e[9]*e[33]+e[19]*e[29]*e[17];
+        A[45]=e[4]*e[27]*e[3]+e[4]*e[0]*e[30]+e[4]*e[29]*e[5]+e[4]*e[2]*e[32]+e[31]*e[0]*e[3]+e[31]*e[2]*e[5]+e[7]*e[27]*e[6]+e[7]*e[0]*e[33]+e[7]*e[29]*e[8]+e[7]*e[2]*e[35]+e[34]*e[0]*e[6]+e[34]*e[2]*e[8]+e[1]*e[27]*e[0]+e[1]*e[29]*e[2]+e[1]*e[34]*e[7]-1.*e[1]*e[32]*e[5]-1.*e[1]*e[33]*e[6]-1.*e[1]*e[30]*e[3]-1.*e[1]*e[35]*e[8]+e[1]*e[31]*e[4]+1.500000000*e[28]*ep2[1]+.5000000000*e[28]*ep2[4]+.5000000000*e[28]*ep2[0]-.5000000000*e[28]*ep2[6]-.5000000000*e[28]*ep2[5]+.5000000000*e[28]*ep2[7]-.5000000000*e[28]*ep2[3]+.5000000000*e[28]*ep2[2]-.5000000000*e[28]*ep2[8];
+        A[191]=-1.*e[35]*e[10]*e[1]-1.*e[35]*e[13]*e[4]+e[35]*e[16]*e[7]+e[35]*e[15]*e[6]-1.*e[35]*e[9]*e[0]-1.*e[35]*e[12]*e[3]+e[32]*e[12]*e[6]+e[32]*e[3]*e[15]+e[32]*e[13]*e[7]+e[32]*e[4]*e[16]-1.*e[8]*e[27]*e[9]-1.*e[8]*e[30]*e[12]-1.*e[8]*e[28]*e[10]-1.*e[8]*e[31]*e[13]+e[8]*e[29]*e[11]+e[11]*e[27]*e[6]+e[11]*e[0]*e[33]+e[11]*e[28]*e[7]+e[11]*e[1]*e[34]+e[29]*e[9]*e[6]+e[29]*e[0]*e[15]+e[29]*e[10]*e[7]+e[29]*e[1]*e[16]+e[5]*e[30]*e[15]+e[5]*e[12]*e[33]+e[5]*e[32]*e[17]+e[5]*e[14]*e[35]+e[5]*e[31]*e[16]+e[5]*e[13]*e[34]+e[8]*e[33]*e[15]+3.*e[8]*e[35]*e[17]+e[8]*e[34]*e[16]+e[2]*e[27]*e[15]+e[2]*e[9]*e[33]+e[2]*e[29]*e[17]+e[2]*e[11]*e[35]+e[2]*e[28]*e[16]+e[2]*e[10]*e[34]-1.*e[17]*e[27]*e[0]+e[17]*e[34]*e[7]+e[17]*e[33]*e[6]-1.*e[17]*e[30]*e[3]-1.*e[17]*e[28]*e[1]-1.*e[17]*e[31]*e[4]+e[14]*e[30]*e[6]+e[14]*e[3]*e[33]+e[14]*e[31]*e[7]+e[14]*e[4]*e[34]+e[14]*e[32]*e[8];
+        A[44]=e[19]*e[11]*e[2]+e[4]*e[18]*e[12]+e[4]*e[9]*e[21]+e[4]*e[20]*e[14]+e[4]*e[11]*e[23]+e[4]*e[19]*e[13]+e[4]*e[10]*e[22]+e[7]*e[18]*e[15]+e[7]*e[9]*e[24]+e[7]*e[20]*e[17]+e[7]*e[11]*e[26]+e[7]*e[19]*e[16]+e[7]*e[10]*e[25]+e[1]*e[18]*e[9]+e[1]*e[20]*e[11]-1.*e[10]*e[21]*e[3]-1.*e[10]*e[26]*e[8]-1.*e[10]*e[23]*e[5]-1.*e[10]*e[24]*e[6]+e[13]*e[18]*e[3]+e[13]*e[0]*e[21]+e[13]*e[1]*e[22]+e[13]*e[20]*e[5]+e[13]*e[2]*e[23]-1.*e[19]*e[15]*e[6]-1.*e[19]*e[14]*e[5]-1.*e[19]*e[12]*e[3]-1.*e[19]*e[17]*e[8]+e[22]*e[9]*e[3]+e[22]*e[0]*e[12]+e[22]*e[11]*e[5]+e[22]*e[2]*e[14]+e[16]*e[18]*e[6]+e[16]*e[0]*e[24]+e[16]*e[1]*e[25]+e[16]*e[20]*e[8]+e[16]*e[2]*e[26]-1.*e[1]*e[23]*e[14]-1.*e[1]*e[24]*e[15]-1.*e[1]*e[26]*e[17]-1.*e[1]*e[21]*e[12]+e[25]*e[9]*e[6]+e[25]*e[0]*e[15]+e[25]*e[11]*e[8]+e[25]*e[2]*e[17]+e[10]*e[18]*e[0]+3.*e[10]*e[19]*e[1]+e[10]*e[20]*e[2]+e[19]*e[9]*e[0];
+        A[190]=.5000000000*ep2[23]*e[26]+.5000000000*e[26]*ep2[25]+.5000000000*ep2[20]*e[26]-.5000000000*e[26]*ep2[18]+.5000000000*ep3[26]+.5000000000*e[26]*ep2[24]+e[20]*e[19]*e[25]-.5000000000*e[26]*ep2[19]-.5000000000*e[26]*ep2[21]+e[20]*e[18]*e[24]-.5000000000*e[26]*ep2[22]+e[23]*e[21]*e[24]+e[23]*e[22]*e[25];
+        A[47]=e[16]*e[9]*e[33]+e[16]*e[29]*e[17]+e[16]*e[11]*e[35]+e[16]*e[10]*e[34]+e[34]*e[11]*e[17]+e[34]*e[9]*e[15]-1.*e[10]*e[30]*e[12]-1.*e[10]*e[32]*e[14]-1.*e[10]*e[33]*e[15]-1.*e[10]*e[35]*e[17]+e[10]*e[27]*e[9]+e[10]*e[29]*e[11]+e[13]*e[27]*e[12]+e[13]*e[9]*e[30]+e[13]*e[29]*e[14]+e[13]*e[11]*e[32]+e[13]*e[10]*e[31]+e[31]*e[11]*e[14]+e[31]*e[9]*e[12]+e[16]*e[27]*e[15]+1.500000000*e[28]*ep2[10]+.5000000000*e[28]*ep2[16]+.5000000000*e[28]*ep2[9]+.5000000000*e[28]*ep2[11]-.5000000000*e[28]*ep2[12]-.5000000000*e[28]*ep2[15]-.5000000000*e[28]*ep2[17]-.5000000000*e[28]*ep2[14]+.5000000000*e[28]*ep2[13];
+        A[189]=.5000000000*ep2[20]*e[35]+.5000000000*ep2[23]*e[35]+1.500000000*e[35]*ep2[26]+.5000000000*e[35]*ep2[25]+.5000000000*e[35]*ep2[24]-.5000000000*e[35]*ep2[18]-.5000000000*e[35]*ep2[19]-.5000000000*e[35]*ep2[22]-.5000000000*e[35]*ep2[21]+e[20]*e[27]*e[24]+e[20]*e[18]*e[33]+e[20]*e[28]*e[25]+e[20]*e[19]*e[34]+e[20]*e[29]*e[26]+e[29]*e[19]*e[25]+e[29]*e[18]*e[24]+e[23]*e[30]*e[24]+e[23]*e[21]*e[33]+e[23]*e[31]*e[25]+e[23]*e[22]*e[34]+e[23]*e[32]*e[26]+e[32]*e[22]*e[25]+e[32]*e[21]*e[24]+e[26]*e[33]*e[24]+e[26]*e[34]*e[25]-1.*e[26]*e[27]*e[18]-1.*e[26]*e[30]*e[21]-1.*e[26]*e[31]*e[22]-1.*e[26]*e[28]*e[19];
+        A[46]=e[4]*e[2]*e[5]+.5000000000*e[1]*ep2[0]-.5000000000*e[1]*ep2[6]+e[7]*e[0]*e[6]+.5000000000*e[1]*ep2[7]+.5000000000*e[1]*ep2[4]-.5000000000*e[1]*ep2[8]+.5000000000*e[1]*ep2[2]-.5000000000*e[1]*ep2[3]+.5000000000*ep3[1]+e[7]*e[2]*e[8]-.5000000000*e[1]*ep2[5]+e[4]*e[0]*e[3];
+        A[188]=-.5000000000*e[17]*ep2[13]-.5000000000*e[17]*ep2[9]+.5000000000*e[17]*ep2[16]+.5000000000*e[17]*ep2[15]+.5000000000*ep3[17]-.5000000000*e[17]*ep2[10]+e[14]*e[13]*e[16]+e[14]*e[12]*e[15]+.5000000000*ep2[14]*e[17]+e[11]*e[10]*e[16]-.5000000000*e[17]*ep2[12]+.5000000000*ep2[11]*e[17]+e[11]*e[9]*e[15];
+        A[41]=e[4]*e[27]*e[30]+e[4]*e[29]*e[32]+e[4]*e[28]*e[31]+e[31]*e[27]*e[3]+e[31]*e[0]*e[30]+e[31]*e[29]*e[5]+e[31]*e[2]*e[32]+e[7]*e[27]*e[33]+e[7]*e[29]*e[35]+e[7]*e[28]*e[34]+e[28]*e[27]*e[0]+e[28]*e[29]*e[2]+e[34]*e[27]*e[6]+e[34]*e[0]*e[33]+e[34]*e[29]*e[8]+e[34]*e[2]*e[35]-1.*e[28]*e[32]*e[5]-1.*e[28]*e[33]*e[6]-1.*e[28]*e[30]*e[3]-1.*e[28]*e[35]*e[8]+.5000000000*e[1]*ep2[27]+.5000000000*e[1]*ep2[29]+1.500000000*e[1]*ep2[28]+.5000000000*e[1]*ep2[31]-.5000000000*e[1]*ep2[32]-.5000000000*e[1]*ep2[33]-.5000000000*e[1]*ep2[30]+.5000000000*e[1]*ep2[34]-.5000000000*e[1]*ep2[35];
+        A[187]=.5000000000*ep2[11]*e[35]+.5000000000*e[35]*ep2[16]-.5000000000*e[35]*ep2[9]-.5000000000*e[35]*ep2[12]+.5000000000*e[35]*ep2[15]+1.500000000*e[35]*ep2[17]-.5000000000*e[35]*ep2[10]+.5000000000*e[35]*ep2[14]-.5000000000*e[35]*ep2[13]+e[11]*e[27]*e[15]+e[11]*e[9]*e[33]+e[11]*e[29]*e[17]+e[11]*e[28]*e[16]+e[11]*e[10]*e[34]+e[29]*e[9]*e[15]+e[29]*e[10]*e[16]+e[14]*e[30]*e[15]+e[14]*e[12]*e[33]+e[14]*e[32]*e[17]+e[14]*e[31]*e[16]+e[14]*e[13]*e[34]+e[32]*e[12]*e[15]+e[32]*e[13]*e[16]+e[17]*e[33]*e[15]+e[17]*e[34]*e[16]-1.*e[17]*e[27]*e[9]-1.*e[17]*e[30]*e[12]-1.*e[17]*e[28]*e[10]-1.*e[17]*e[31]*e[13];
+        A[40]=e[34]*e[27]*e[33]+e[34]*e[29]*e[35]-.5000000000*e[28]*ep2[30]-.5000000000*e[28]*ep2[35]+.5000000000*ep3[28]+.5000000000*e[28]*ep2[27]+.5000000000*e[28]*ep2[29]+e[31]*e[27]*e[30]+e[31]*e[29]*e[32]-.5000000000*e[28]*ep2[32]-.5000000000*e[28]*ep2[33]+.5000000000*e[28]*ep2[31]+.5000000000*e[28]*ep2[34];
+        A[186]=.5000000000*ep2[5]*e[8]+e[2]*e[0]*e[6]+.5000000000*ep2[2]*e[8]+.5000000000*ep3[8]-.5000000000*e[8]*ep2[0]+e[5]*e[4]*e[7]+e[5]*e[3]*e[6]+.5000000000*e[8]*ep2[7]+e[2]*e[1]*e[7]-.5000000000*e[8]*ep2[1]-.5000000000*e[8]*ep2[4]-.5000000000*e[8]*ep2[3]+.5000000000*e[8]*ep2[6];
+        A[43]=e[28]*e[27]*e[9]+e[28]*e[29]*e[11]-1.*e[28]*e[30]*e[12]+e[28]*e[31]*e[13]-1.*e[28]*e[32]*e[14]-1.*e[28]*e[33]*e[15]-1.*e[28]*e[35]*e[17]+e[31]*e[27]*e[12]+e[31]*e[9]*e[30]+e[31]*e[29]*e[14]+e[31]*e[11]*e[32]+e[13]*e[27]*e[30]+e[13]*e[29]*e[32]+e[16]*e[27]*e[33]+e[16]*e[29]*e[35]+e[34]*e[27]*e[15]+e[34]*e[9]*e[33]+e[34]*e[29]*e[17]+e[34]*e[11]*e[35]+e[34]*e[28]*e[16]+.5000000000*e[10]*ep2[27]+.5000000000*e[10]*ep2[29]+1.500000000*e[10]*ep2[28]-.5000000000*e[10]*ep2[32]+.5000000000*e[10]*ep2[31]-.5000000000*e[10]*ep2[33]-.5000000000*e[10]*ep2[30]+.5000000000*e[10]*ep2[34]-.5000000000*e[10]*ep2[35];
+        A[185]=-.5000000000*e[35]*ep2[1]+.5000000000*e[35]*ep2[7]-.5000000000*e[35]*ep2[3]+.5000000000*ep2[2]*e[35]+1.500000000*e[35]*ep2[8]-.5000000000*e[35]*ep2[4]-.5000000000*e[35]*ep2[0]+.5000000000*e[35]*ep2[6]+.5000000000*e[35]*ep2[5]+e[2]*e[27]*e[6]+e[2]*e[0]*e[33]+e[2]*e[28]*e[7]+e[2]*e[1]*e[34]+e[2]*e[29]*e[8]-1.*e[8]*e[27]*e[0]+e[8]*e[34]*e[7]+e[8]*e[32]*e[5]+e[8]*e[33]*e[6]-1.*e[8]*e[30]*e[3]-1.*e[8]*e[28]*e[1]-1.*e[8]*e[31]*e[4]+e[29]*e[1]*e[7]+e[29]*e[0]*e[6]+e[5]*e[30]*e[6]+e[5]*e[3]*e[33]+e[5]*e[31]*e[7]+e[5]*e[4]*e[34]+e[32]*e[4]*e[7]+e[32]*e[3]*e[6];
+        A[42]=e[28]*e[27]*e[18]+e[28]*e[29]*e[20]+e[22]*e[27]*e[30]+e[22]*e[29]*e[32]+e[22]*e[28]*e[31]+e[31]*e[27]*e[21]+e[31]*e[18]*e[30]+e[31]*e[29]*e[23]+e[31]*e[20]*e[32]+e[25]*e[27]*e[33]+e[25]*e[29]*e[35]+e[25]*e[28]*e[34]+e[34]*e[27]*e[24]+e[34]*e[18]*e[33]+e[34]*e[29]*e[26]+e[34]*e[20]*e[35]-1.*e[28]*e[33]*e[24]-1.*e[28]*e[30]*e[21]-1.*e[28]*e[35]*e[26]-1.*e[28]*e[32]*e[23]-.5000000000*e[19]*ep2[33]-.5000000000*e[19]*ep2[30]-.5000000000*e[19]*ep2[35]+.5000000000*e[19]*ep2[27]+.5000000000*e[19]*ep2[29]+1.500000000*e[19]*ep2[28]+.5000000000*e[19]*ep2[31]+.5000000000*e[19]*ep2[34]-.5000000000*e[19]*ep2[32];
+        A[184]=e[23]*e[3]*e[15]-1.*e[17]*e[19]*e[1]-1.*e[17]*e[22]*e[4]-1.*e[17]*e[18]*e[0]+e[17]*e[25]*e[7]+e[17]*e[24]*e[6]+e[14]*e[21]*e[6]+e[14]*e[3]*e[24]+e[14]*e[22]*e[7]+e[14]*e[4]*e[25]+e[14]*e[23]*e[8]-1.*e[26]*e[10]*e[1]-1.*e[26]*e[13]*e[4]+e[26]*e[16]*e[7]+e[26]*e[15]*e[6]-1.*e[26]*e[9]*e[0]-1.*e[26]*e[12]*e[3]+e[23]*e[12]*e[6]+e[11]*e[18]*e[6]+e[11]*e[0]*e[24]+e[11]*e[19]*e[7]+e[11]*e[1]*e[25]+e[11]*e[20]*e[8]+e[11]*e[2]*e[26]+e[20]*e[9]*e[6]+e[20]*e[0]*e[15]+e[20]*e[10]*e[7]+e[20]*e[1]*e[16]+e[20]*e[2]*e[17]+e[5]*e[21]*e[15]+e[5]*e[12]*e[24]+e[5]*e[23]*e[17]+e[5]*e[14]*e[26]+e[5]*e[22]*e[16]+e[5]*e[13]*e[25]+e[8]*e[24]*e[15]+3.*e[8]*e[26]*e[17]+e[8]*e[25]*e[16]+e[2]*e[18]*e[15]+e[2]*e[9]*e[24]+e[2]*e[19]*e[16]+e[2]*e[10]*e[25]-1.*e[17]*e[21]*e[3]+e[23]*e[4]*e[16]+e[23]*e[13]*e[7]-1.*e[8]*e[18]*e[9]-1.*e[8]*e[21]*e[12]-1.*e[8]*e[19]*e[10]-1.*e[8]*e[22]*e[13];
+        A[54]=e[13]*e[18]*e[12]+e[13]*e[9]*e[21]+e[13]*e[20]*e[14]+e[13]*e[11]*e[23]+e[13]*e[10]*e[22]+e[22]*e[11]*e[14]+e[22]*e[9]*e[12]+e[16]*e[18]*e[15]+e[16]*e[9]*e[24]+e[16]*e[20]*e[17]+e[16]*e[11]*e[26]+e[16]*e[10]*e[25]+e[25]*e[11]*e[17]+e[25]*e[9]*e[15]-1.*e[10]*e[23]*e[14]-1.*e[10]*e[24]*e[15]-1.*e[10]*e[26]*e[17]+e[10]*e[20]*e[11]+e[10]*e[18]*e[9]-1.*e[10]*e[21]*e[12]+.5000000000*e[19]*ep2[11]+.5000000000*e[19]*ep2[9]+1.500000000*e[19]*ep2[10]+.5000000000*e[19]*ep2[13]+.5000000000*e[19]*ep2[16]-.5000000000*e[19]*ep2[12]-.5000000000*e[19]*ep2[15]-.5000000000*e[19]*ep2[17]-.5000000000*e[19]*ep2[14];
+        A[164]=e[10]*e[18]*e[6]+e[10]*e[0]*e[24]+e[10]*e[19]*e[7]+e[10]*e[1]*e[25]+e[10]*e[20]*e[8]+e[10]*e[2]*e[26]+e[19]*e[9]*e[6]+e[19]*e[0]*e[15]+e[19]*e[1]*e[16]+e[19]*e[11]*e[8]+e[19]*e[2]*e[17]+e[4]*e[21]*e[15]+e[4]*e[12]*e[24]+e[4]*e[23]*e[17]+e[4]*e[14]*e[26]+e[4]*e[22]*e[16]+e[4]*e[13]*e[25]+e[7]*e[24]*e[15]+e[7]*e[26]*e[17]+3.*e[7]*e[25]*e[16]+e[1]*e[18]*e[15]+e[1]*e[9]*e[24]+e[1]*e[20]*e[17]+e[1]*e[11]*e[26]-1.*e[16]*e[21]*e[3]+e[16]*e[26]*e[8]-1.*e[16]*e[20]*e[2]-1.*e[16]*e[18]*e[0]-1.*e[16]*e[23]*e[5]+e[16]*e[24]*e[6]+e[13]*e[21]*e[6]+e[13]*e[3]*e[24]+e[13]*e[22]*e[7]+e[13]*e[23]*e[8]+e[13]*e[5]*e[26]-1.*e[25]*e[11]*e[2]+e[25]*e[15]*e[6]-1.*e[25]*e[9]*e[0]-1.*e[25]*e[14]*e[5]-1.*e[25]*e[12]*e[3]+e[25]*e[17]*e[8]+e[22]*e[12]*e[6]+e[22]*e[3]*e[15]+e[22]*e[14]*e[8]+e[22]*e[5]*e[17]-1.*e[7]*e[23]*e[14]-1.*e[7]*e[20]*e[11]-1.*e[7]*e[18]*e[9]-1.*e[7]*e[21]*e[12];
+        A[55]=e[13]*e[9]*e[3]+e[13]*e[0]*e[12]+e[13]*e[10]*e[4]+e[13]*e[11]*e[5]+e[13]*e[2]*e[14]+e[16]*e[9]*e[6]+e[16]*e[0]*e[15]+e[16]*e[10]*e[7]+e[16]*e[11]*e[8]+e[16]*e[2]*e[17]+e[7]*e[11]*e[17]+e[7]*e[9]*e[15]+e[4]*e[11]*e[14]+e[4]*e[9]*e[12]+e[10]*e[9]*e[0]+e[10]*e[11]*e[2]-1.*e[10]*e[15]*e[6]-1.*e[10]*e[14]*e[5]-1.*e[10]*e[12]*e[3]-1.*e[10]*e[17]*e[8]+.5000000000*e[1]*ep2[11]+.5000000000*e[1]*ep2[9]+1.500000000*e[1]*ep2[10]-.5000000000*e[1]*ep2[12]-.5000000000*e[1]*ep2[15]-.5000000000*e[1]*ep2[17]-.5000000000*e[1]*ep2[14]+.5000000000*e[1]*ep2[13]+.5000000000*e[1]*ep2[16];
+        A[165]=e[1]*e[27]*e[6]+e[1]*e[0]*e[33]+e[1]*e[28]*e[7]+e[1]*e[29]*e[8]+e[1]*e[2]*e[35]-1.*e[7]*e[27]*e[0]-1.*e[7]*e[32]*e[5]+e[7]*e[33]*e[6]-1.*e[7]*e[30]*e[3]+e[7]*e[35]*e[8]-1.*e[7]*e[29]*e[2]+e[7]*e[31]*e[4]+e[28]*e[0]*e[6]+e[28]*e[2]*e[8]+e[4]*e[30]*e[6]+e[4]*e[3]*e[33]+e[4]*e[32]*e[8]+e[4]*e[5]*e[35]+e[31]*e[3]*e[6]+e[31]*e[5]*e[8]+.5000000000*ep2[1]*e[34]+1.500000000*e[34]*ep2[7]+.5000000000*e[34]*ep2[4]-.5000000000*e[34]*ep2[0]+.5000000000*e[34]*ep2[6]-.5000000000*e[34]*ep2[5]-.5000000000*e[34]*ep2[3]-.5000000000*e[34]*ep2[2]+.5000000000*e[34]*ep2[8];
+        A[52]=e[4]*e[18]*e[3]+e[4]*e[0]*e[21]+e[4]*e[1]*e[22]+e[4]*e[20]*e[5]+e[4]*e[2]*e[23]+e[22]*e[0]*e[3]+e[22]*e[2]*e[5]+e[7]*e[18]*e[6]+e[7]*e[0]*e[24]+e[7]*e[1]*e[25]+e[7]*e[20]*e[8]+e[7]*e[2]*e[26]+e[25]*e[0]*e[6]+e[25]*e[2]*e[8]+e[1]*e[18]*e[0]+e[1]*e[20]*e[2]-1.*e[1]*e[21]*e[3]-1.*e[1]*e[26]*e[8]-1.*e[1]*e[23]*e[5]-1.*e[1]*e[24]*e[6]+.5000000000*e[19]*ep2[4]+.5000000000*e[19]*ep2[0]-.5000000000*e[19]*ep2[6]-.5000000000*e[19]*ep2[5]+1.500000000*e[19]*ep2[1]+.5000000000*e[19]*ep2[7]-.5000000000*e[19]*ep2[3]+.5000000000*e[19]*ep2[2]-.5000000000*e[19]*ep2[8];
+        A[166]=-.5000000000*e[7]*ep2[0]+e[4]*e[5]*e[8]+.5000000000*ep2[4]*e[7]-.5000000000*e[7]*ep2[2]+.5000000000*e[7]*ep2[8]-.5000000000*e[7]*ep2[5]+.5000000000*e[7]*ep2[6]+e[1]*e[0]*e[6]+.5000000000*ep3[7]+e[4]*e[3]*e[6]+e[1]*e[2]*e[8]-.5000000000*e[7]*ep2[3]+.5000000000*ep2[1]*e[7];
+        A[53]=-1.*e[1]*e[32]*e[23]-1.*e[19]*e[32]*e[5]-1.*e[19]*e[33]*e[6]-1.*e[19]*e[30]*e[3]-1.*e[19]*e[35]*e[8]-1.*e[28]*e[21]*e[3]-1.*e[28]*e[26]*e[8]-1.*e[28]*e[23]*e[5]-1.*e[28]*e[24]*e[6]+e[7]*e[27]*e[24]+e[7]*e[18]*e[33]+e[7]*e[29]*e[26]+e[7]*e[20]*e[35]+e[22]*e[27]*e[3]+e[22]*e[0]*e[30]+e[22]*e[29]*e[5]+e[22]*e[2]*e[32]+e[31]*e[18]*e[3]+e[31]*e[0]*e[21]+e[31]*e[20]*e[5]+e[31]*e[2]*e[23]+e[25]*e[27]*e[6]+e[25]*e[0]*e[33]+e[25]*e[28]*e[7]+e[25]*e[1]*e[34]+e[25]*e[29]*e[8]+e[25]*e[2]*e[35]+e[34]*e[18]*e[6]+e[34]*e[0]*e[24]+e[34]*e[19]*e[7]+e[34]*e[20]*e[8]+e[34]*e[2]*e[26]+e[1]*e[27]*e[18]+3.*e[1]*e[28]*e[19]+e[1]*e[29]*e[20]+e[19]*e[27]*e[0]+e[19]*e[29]*e[2]+e[28]*e[18]*e[0]+e[28]*e[20]*e[2]+e[4]*e[27]*e[21]+e[4]*e[18]*e[30]+e[4]*e[28]*e[22]+e[4]*e[19]*e[31]+e[4]*e[29]*e[23]+e[4]*e[20]*e[32]-1.*e[1]*e[33]*e[24]-1.*e[1]*e[30]*e[21]-1.*e[1]*e[35]*e[26]+e[1]*e[31]*e[22];
+        A[167]=e[10]*e[27]*e[15]+e[10]*e[9]*e[33]+e[10]*e[29]*e[17]+e[10]*e[11]*e[35]+e[10]*e[28]*e[16]+e[28]*e[11]*e[17]+e[28]*e[9]*e[15]+e[13]*e[30]*e[15]+e[13]*e[12]*e[33]+e[13]*e[32]*e[17]+e[13]*e[14]*e[35]+e[13]*e[31]*e[16]+e[31]*e[14]*e[17]+e[31]*e[12]*e[15]+e[16]*e[33]*e[15]+e[16]*e[35]*e[17]-1.*e[16]*e[27]*e[9]-1.*e[16]*e[30]*e[12]-1.*e[16]*e[32]*e[14]-1.*e[16]*e[29]*e[11]+.5000000000*ep2[10]*e[34]+1.500000000*e[34]*ep2[16]-.5000000000*e[34]*ep2[9]-.5000000000*e[34]*ep2[11]-.5000000000*e[34]*ep2[12]+.5000000000*e[34]*ep2[15]+.5000000000*e[34]*ep2[17]-.5000000000*e[34]*ep2[14]+.5000000000*e[34]*ep2[13];
+        A[50]=.5000000000*e[19]*ep2[18]+.5000000000*e[19]*ep2[25]+.5000000000*e[19]*ep2[22]+e[25]*e[20]*e[26]-.5000000000*e[19]*ep2[21]+.5000000000*e[19]*ep2[20]-.5000000000*e[19]*ep2[26]-.5000000000*e[19]*ep2[23]-.5000000000*e[19]*ep2[24]+.5000000000*ep3[19]+e[22]*e[20]*e[23]+e[25]*e[18]*e[24]+e[22]*e[18]*e[21];
+        A[160]=.5000000000*e[34]*ep2[33]+.5000000000*e[34]*ep2[35]-.5000000000*e[34]*ep2[27]-.5000000000*e[34]*ep2[32]-.5000000000*e[34]*ep2[29]-.5000000000*e[34]*ep2[30]+.5000000000*ep2[28]*e[34]+e[31]*e[30]*e[33]+e[31]*e[32]*e[35]+e[28]*e[27]*e[33]+.5000000000*ep3[34]+e[28]*e[29]*e[35]+.5000000000*ep2[31]*e[34];
+        A[51]=e[4]*e[28]*e[13]+e[4]*e[10]*e[31]+e[7]*e[27]*e[15]+e[7]*e[9]*e[33]+e[7]*e[29]*e[17]+e[7]*e[11]*e[35]+e[7]*e[28]*e[16]+e[7]*e[10]*e[34]+e[1]*e[27]*e[9]+e[1]*e[29]*e[11]+3.*e[1]*e[28]*e[10]+e[10]*e[27]*e[0]-1.*e[10]*e[32]*e[5]-1.*e[10]*e[33]*e[6]-1.*e[10]*e[30]*e[3]-1.*e[10]*e[35]*e[8]+e[10]*e[29]*e[2]+e[13]*e[27]*e[3]+e[13]*e[0]*e[30]+e[13]*e[1]*e[31]+e[13]*e[29]*e[5]+e[13]*e[2]*e[32]+e[28]*e[11]*e[2]-1.*e[28]*e[15]*e[6]+e[28]*e[9]*e[0]-1.*e[28]*e[14]*e[5]-1.*e[28]*e[12]*e[3]-1.*e[28]*e[17]*e[8]+e[31]*e[9]*e[3]+e[31]*e[0]*e[12]+e[31]*e[11]*e[5]+e[31]*e[2]*e[14]+e[16]*e[27]*e[6]+e[16]*e[0]*e[33]+e[16]*e[1]*e[34]+e[16]*e[29]*e[8]+e[16]*e[2]*e[35]-1.*e[1]*e[30]*e[12]-1.*e[1]*e[32]*e[14]-1.*e[1]*e[33]*e[15]-1.*e[1]*e[35]*e[17]+e[34]*e[9]*e[6]+e[34]*e[0]*e[15]+e[34]*e[11]*e[8]+e[34]*e[2]*e[17]+e[4]*e[27]*e[12]+e[4]*e[9]*e[30]+e[4]*e[29]*e[14]+e[4]*e[11]*e[32];
+        A[161]=e[4]*e[30]*e[33]+e[4]*e[32]*e[35]+e[4]*e[31]*e[34]+e[31]*e[30]*e[6]+e[31]*e[3]*e[33]+e[31]*e[32]*e[8]+e[31]*e[5]*e[35]+e[28]*e[27]*e[6]+e[28]*e[0]*e[33]+e[28]*e[29]*e[8]+e[28]*e[2]*e[35]+e[34]*e[33]*e[6]+e[34]*e[35]*e[8]-1.*e[34]*e[27]*e[0]-1.*e[34]*e[32]*e[5]-1.*e[34]*e[30]*e[3]-1.*e[34]*e[29]*e[2]+e[1]*e[27]*e[33]+e[1]*e[29]*e[35]+e[1]*e[28]*e[34]+.5000000000*ep2[31]*e[7]-.5000000000*e[7]*ep2[27]-.5000000000*e[7]*ep2[32]+.5000000000*e[7]*ep2[28]-.5000000000*e[7]*ep2[29]+.5000000000*e[7]*ep2[33]-.5000000000*e[7]*ep2[30]+1.500000000*e[7]*ep2[34]+.5000000000*e[7]*ep2[35];
+        A[48]=-.5000000000*e[10]*ep2[14]-.5000000000*e[10]*ep2[17]-.5000000000*e[10]*ep2[15]+e[13]*e[11]*e[14]+e[16]*e[11]*e[17]+.5000000000*e[10]*ep2[13]+e[13]*e[9]*e[12]-.5000000000*e[10]*ep2[12]+.5000000000*ep3[10]+e[16]*e[9]*e[15]+.5000000000*e[10]*ep2[16]+.5000000000*e[10]*ep2[11]+.5000000000*e[10]*ep2[9];
+        A[162]=e[22]*e[32]*e[35]+e[22]*e[31]*e[34]+e[31]*e[30]*e[24]+e[31]*e[21]*e[33]+e[31]*e[32]*e[26]+e[31]*e[23]*e[35]+e[34]*e[33]*e[24]+e[34]*e[35]*e[26]-1.*e[34]*e[27]*e[18]-1.*e[34]*e[30]*e[21]-1.*e[34]*e[29]*e[20]-1.*e[34]*e[32]*e[23]+e[19]*e[27]*e[33]+e[19]*e[29]*e[35]+e[19]*e[28]*e[34]+e[28]*e[27]*e[24]+e[28]*e[18]*e[33]+e[28]*e[29]*e[26]+e[28]*e[20]*e[35]+e[22]*e[30]*e[33]+.5000000000*ep2[28]*e[25]+.5000000000*ep2[31]*e[25]+.5000000000*e[25]*ep2[33]+.5000000000*e[25]*ep2[35]+1.500000000*e[25]*ep2[34]-.5000000000*e[25]*ep2[27]-.5000000000*e[25]*ep2[32]-.5000000000*e[25]*ep2[29]-.5000000000*e[25]*ep2[30];
+        A[49]=-1.*e[19]*e[35]*e[26]-1.*e[19]*e[32]*e[23]+e[19]*e[27]*e[18]+e[19]*e[29]*e[20]+e[22]*e[27]*e[21]+e[22]*e[18]*e[30]+e[22]*e[19]*e[31]+e[22]*e[29]*e[23]+e[22]*e[20]*e[32]+e[31]*e[18]*e[21]+e[31]*e[20]*e[23]+e[25]*e[27]*e[24]+e[25]*e[18]*e[33]+e[25]*e[19]*e[34]+e[25]*e[29]*e[26]+e[25]*e[20]*e[35]+e[34]*e[18]*e[24]+e[34]*e[20]*e[26]-1.*e[19]*e[33]*e[24]-1.*e[19]*e[30]*e[21]+1.500000000*e[28]*ep2[19]+.5000000000*e[28]*ep2[18]+.5000000000*e[28]*ep2[20]+.5000000000*e[28]*ep2[22]+.5000000000*e[28]*ep2[25]-.5000000000*e[28]*ep2[26]-.5000000000*e[28]*ep2[23]-.5000000000*e[28]*ep2[24]-.5000000000*e[28]*ep2[21];
+        A[163]=e[10]*e[27]*e[33]+e[10]*e[29]*e[35]+e[10]*e[28]*e[34]+e[34]*e[33]*e[15]+e[34]*e[35]*e[17]+e[28]*e[27]*e[15]+e[28]*e[9]*e[33]+e[28]*e[29]*e[17]+e[28]*e[11]*e[35]-1.*e[34]*e[27]*e[9]-1.*e[34]*e[30]*e[12]+e[34]*e[31]*e[13]-1.*e[34]*e[32]*e[14]-1.*e[34]*e[29]*e[11]+e[31]*e[30]*e[15]+e[31]*e[12]*e[33]+e[31]*e[32]*e[17]+e[31]*e[14]*e[35]+e[13]*e[30]*e[33]+e[13]*e[32]*e[35]-.5000000000*e[16]*ep2[27]-.5000000000*e[16]*ep2[32]+.5000000000*e[16]*ep2[28]-.5000000000*e[16]*ep2[29]+.5000000000*e[16]*ep2[31]+.5000000000*e[16]*ep2[33]-.5000000000*e[16]*ep2[30]+1.500000000*e[16]*ep2[34]+.5000000000*e[16]*ep2[35];
+        A[63]=e[29]*e[32]*e[14]-1.*e[29]*e[33]*e[15]-1.*e[29]*e[34]*e[16]+e[32]*e[27]*e[12]+e[32]*e[9]*e[30]+e[32]*e[28]*e[13]+e[32]*e[10]*e[31]+e[14]*e[27]*e[30]+e[14]*e[28]*e[31]+e[17]*e[27]*e[33]+e[17]*e[28]*e[34]+e[35]*e[27]*e[15]+e[35]*e[9]*e[33]+e[35]*e[29]*e[17]+e[35]*e[28]*e[16]+e[35]*e[10]*e[34]+e[29]*e[27]*e[9]+e[29]*e[28]*e[10]-1.*e[29]*e[30]*e[12]-1.*e[29]*e[31]*e[13]+.5000000000*e[11]*ep2[27]+1.500000000*e[11]*ep2[29]+.5000000000*e[11]*ep2[28]+.5000000000*e[11]*ep2[32]-.5000000000*e[11]*ep2[31]-.5000000000*e[11]*ep2[33]-.5000000000*e[11]*ep2[30]-.5000000000*e[11]*ep2[34]+.5000000000*e[11]*ep2[35];
+        A[173]=e[1]*e[20]*e[35]+e[19]*e[27]*e[6]+e[19]*e[0]*e[33]+e[19]*e[28]*e[7]+e[19]*e[29]*e[8]+e[19]*e[2]*e[35]+e[28]*e[18]*e[6]+e[28]*e[0]*e[24]+e[28]*e[20]*e[8]+e[28]*e[2]*e[26]+e[4]*e[30]*e[24]+e[4]*e[21]*e[33]+e[4]*e[31]*e[25]+e[4]*e[22]*e[34]+e[4]*e[32]*e[26]+e[4]*e[23]*e[35]-1.*e[7]*e[27]*e[18]+e[7]*e[33]*e[24]-1.*e[7]*e[30]*e[21]-1.*e[7]*e[29]*e[20]+e[7]*e[35]*e[26]+e[7]*e[31]*e[22]-1.*e[7]*e[32]*e[23]-1.*e[25]*e[27]*e[0]-1.*e[25]*e[32]*e[5]-1.*e[25]*e[30]*e[3]-1.*e[25]*e[29]*e[2]-1.*e[34]*e[21]*e[3]-1.*e[34]*e[20]*e[2]-1.*e[34]*e[18]*e[0]-1.*e[34]*e[23]*e[5]+e[22]*e[30]*e[6]+e[22]*e[3]*e[33]+e[22]*e[32]*e[8]+e[22]*e[5]*e[35]+e[31]*e[21]*e[6]+e[31]*e[3]*e[24]+e[31]*e[23]*e[8]+e[31]*e[5]*e[26]+e[34]*e[26]*e[8]+e[1]*e[27]*e[24]+e[1]*e[18]*e[33]+e[1]*e[28]*e[25]+e[1]*e[19]*e[34]+e[1]*e[29]*e[26]+e[34]*e[24]*e[6]+e[25]*e[33]*e[6]+3.*e[25]*e[34]*e[7]+e[25]*e[35]*e[8];
+        A[62]=.5000000000*e[20]*ep2[27]+1.500000000*e[20]*ep2[29]+.5000000000*e[20]*ep2[28]+.5000000000*e[20]*ep2[32]+.5000000000*e[20]*ep2[35]-.5000000000*e[20]*ep2[31]-.5000000000*e[20]*ep2[33]-.5000000000*e[20]*ep2[30]-.5000000000*e[20]*ep2[34]+e[29]*e[27]*e[18]+e[29]*e[28]*e[19]+e[23]*e[27]*e[30]+e[23]*e[29]*e[32]+e[23]*e[28]*e[31]+e[32]*e[27]*e[21]+e[32]*e[18]*e[30]+e[32]*e[28]*e[22]+e[32]*e[19]*e[31]+e[26]*e[27]*e[33]+e[26]*e[29]*e[35]+e[26]*e[28]*e[34]+e[35]*e[27]*e[24]+e[35]*e[18]*e[33]+e[35]*e[28]*e[25]+e[35]*e[19]*e[34]-1.*e[29]*e[33]*e[24]-1.*e[29]*e[30]*e[21]-1.*e[29]*e[31]*e[22]-1.*e[29]*e[34]*e[25];
+        A[172]=e[19]*e[1]*e[7]+e[19]*e[0]*e[6]+e[19]*e[2]*e[8]+e[4]*e[21]*e[6]+e[4]*e[3]*e[24]+e[4]*e[22]*e[7]+e[4]*e[23]*e[8]+e[4]*e[5]*e[26]+e[22]*e[3]*e[6]+e[22]*e[5]*e[8]+e[7]*e[24]*e[6]+e[7]*e[26]*e[8]+e[1]*e[18]*e[6]+e[1]*e[0]*e[24]+e[1]*e[20]*e[8]+e[1]*e[2]*e[26]-1.*e[7]*e[21]*e[3]-1.*e[7]*e[20]*e[2]-1.*e[7]*e[18]*e[0]-1.*e[7]*e[23]*e[5]+.5000000000*e[25]*ep2[4]-.5000000000*e[25]*ep2[0]+.5000000000*e[25]*ep2[6]-.5000000000*e[25]*ep2[5]+.5000000000*e[25]*ep2[1]+1.500000000*e[25]*ep2[7]-.5000000000*e[25]*ep2[3]-.5000000000*e[25]*ep2[2]+.5000000000*e[25]*ep2[8];
+        A[61]=e[5]*e[27]*e[30]+e[5]*e[29]*e[32]+e[5]*e[28]*e[31]+e[32]*e[27]*e[3]+e[32]*e[0]*e[30]+e[32]*e[28]*e[4]+e[32]*e[1]*e[31]+e[8]*e[27]*e[33]+e[8]*e[29]*e[35]+e[8]*e[28]*e[34]+e[29]*e[27]*e[0]+e[29]*e[28]*e[1]+e[35]*e[27]*e[6]+e[35]*e[0]*e[33]+e[35]*e[28]*e[7]+e[35]*e[1]*e[34]-1.*e[29]*e[34]*e[7]-1.*e[29]*e[33]*e[6]-1.*e[29]*e[30]*e[3]-1.*e[29]*e[31]*e[4]+.5000000000*e[2]*ep2[27]+1.500000000*e[2]*ep2[29]+.5000000000*e[2]*ep2[28]+.5000000000*e[2]*ep2[32]-.5000000000*e[2]*ep2[31]-.5000000000*e[2]*ep2[33]-.5000000000*e[2]*ep2[30]-.5000000000*e[2]*ep2[34]+.5000000000*e[2]*ep2[35];
+        A[175]=e[13]*e[12]*e[6]+e[13]*e[3]*e[15]+e[13]*e[4]*e[16]+e[13]*e[14]*e[8]+e[13]*e[5]*e[17]+e[16]*e[15]*e[6]+e[16]*e[17]*e[8]+e[1]*e[11]*e[17]+e[1]*e[9]*e[15]+e[1]*e[10]*e[16]+e[4]*e[14]*e[17]+e[4]*e[12]*e[15]+e[10]*e[9]*e[6]+e[10]*e[0]*e[15]+e[10]*e[11]*e[8]+e[10]*e[2]*e[17]-1.*e[16]*e[11]*e[2]-1.*e[16]*e[9]*e[0]-1.*e[16]*e[14]*e[5]-1.*e[16]*e[12]*e[3]+.5000000000*ep2[13]*e[7]+1.500000000*ep2[16]*e[7]+.5000000000*e[7]*ep2[17]+.5000000000*e[7]*ep2[15]-.5000000000*e[7]*ep2[9]-.5000000000*e[7]*ep2[11]-.5000000000*e[7]*ep2[12]+.5000000000*e[7]*ep2[10]-.5000000000*e[7]*ep2[14];
+        A[60]=.5000000000*e[29]*ep2[32]+.5000000000*e[29]*ep2[35]-.5000000000*e[29]*ep2[31]-.5000000000*e[29]*ep2[33]-.5000000000*e[29]*ep2[30]-.5000000000*e[29]*ep2[34]+e[32]*e[27]*e[30]+.5000000000*ep3[29]+.5000000000*e[29]*ep2[28]+e[35]*e[28]*e[34]+.5000000000*e[29]*ep2[27]+e[35]*e[27]*e[33]+e[32]*e[28]*e[31];
+        A[174]=-1.*e[16]*e[21]*e[12]+e[10]*e[18]*e[15]+e[10]*e[9]*e[24]+e[10]*e[20]*e[17]+e[10]*e[11]*e[26]+e[19]*e[11]*e[17]+e[19]*e[9]*e[15]+e[19]*e[10]*e[16]+e[13]*e[21]*e[15]+e[13]*e[12]*e[24]+e[13]*e[23]*e[17]+e[13]*e[14]*e[26]+e[13]*e[22]*e[16]+e[22]*e[14]*e[17]+e[22]*e[12]*e[15]+e[16]*e[24]*e[15]+e[16]*e[26]*e[17]-1.*e[16]*e[23]*e[14]-1.*e[16]*e[20]*e[11]-1.*e[16]*e[18]*e[9]+.5000000000*ep2[13]*e[25]+1.500000000*e[25]*ep2[16]+.5000000000*e[25]*ep2[17]+.5000000000*e[25]*ep2[15]+.5000000000*ep2[10]*e[25]-.5000000000*e[25]*ep2[9]-.5000000000*e[25]*ep2[11]-.5000000000*e[25]*ep2[12]-.5000000000*e[25]*ep2[14];
+        A[59]=e[19]*e[20]*e[2]+e[22]*e[18]*e[3]+e[22]*e[0]*e[21]+e[22]*e[19]*e[4]+e[22]*e[20]*e[5]+e[22]*e[2]*e[23]-1.*e[19]*e[21]*e[3]-1.*e[19]*e[26]*e[8]+e[19]*e[25]*e[7]-1.*e[19]*e[23]*e[5]-1.*e[19]*e[24]*e[6]+e[4]*e[18]*e[21]+e[4]*e[20]*e[23]+e[25]*e[18]*e[6]+e[25]*e[0]*e[24]+e[25]*e[20]*e[8]+e[25]*e[2]*e[26]+e[7]*e[18]*e[24]+e[7]*e[20]*e[26]+e[19]*e[18]*e[0]+1.500000000*ep2[19]*e[1]+.5000000000*e[1]*ep2[22]+.5000000000*e[1]*ep2[18]+.5000000000*e[1]*ep2[20]+.5000000000*e[1]*ep2[25]-.5000000000*e[1]*ep2[26]-.5000000000*e[1]*ep2[23]-.5000000000*e[1]*ep2[24]-.5000000000*e[1]*ep2[21];
+        A[169]=e[19]*e[27]*e[24]+e[19]*e[18]*e[33]+e[19]*e[28]*e[25]+e[19]*e[29]*e[26]+e[19]*e[20]*e[35]+e[28]*e[18]*e[24]+e[28]*e[20]*e[26]+e[22]*e[30]*e[24]+e[22]*e[21]*e[33]+e[22]*e[31]*e[25]+e[22]*e[32]*e[26]+e[22]*e[23]*e[35]+e[31]*e[21]*e[24]+e[31]*e[23]*e[26]+e[25]*e[33]*e[24]+e[25]*e[35]*e[26]-1.*e[25]*e[27]*e[18]-1.*e[25]*e[30]*e[21]-1.*e[25]*e[29]*e[20]-1.*e[25]*e[32]*e[23]-.5000000000*e[34]*ep2[18]-.5000000000*e[34]*ep2[23]-.5000000000*e[34]*ep2[20]-.5000000000*e[34]*ep2[21]+.5000000000*ep2[19]*e[34]+.5000000000*ep2[22]*e[34]+1.500000000*e[34]*ep2[25]+.5000000000*e[34]*ep2[24]+.5000000000*e[34]*ep2[26];
+        A[58]=e[16]*e[0]*e[6]+e[16]*e[2]*e[8]+e[1]*e[11]*e[2]-1.*e[1]*e[15]*e[6]+e[1]*e[9]*e[0]-1.*e[1]*e[14]*e[5]-1.*e[1]*e[12]*e[3]-1.*e[1]*e[17]*e[8]+e[4]*e[9]*e[3]+e[4]*e[0]*e[12]+e[4]*e[1]*e[13]+e[4]*e[11]*e[5]+e[4]*e[2]*e[14]+e[13]*e[0]*e[3]+e[13]*e[2]*e[5]+e[7]*e[9]*e[6]+e[7]*e[0]*e[15]+e[7]*e[1]*e[16]+e[7]*e[11]*e[8]+e[7]*e[2]*e[17]-.5000000000*e[10]*ep2[6]-.5000000000*e[10]*ep2[5]-.5000000000*e[10]*ep2[3]-.5000000000*e[10]*ep2[8]+1.500000000*e[10]*ep2[1]+.5000000000*e[10]*ep2[0]+.5000000000*e[10]*ep2[2]+.5000000000*e[10]*ep2[4]+.5000000000*e[10]*ep2[7];
+        A[168]=e[13]*e[14]*e[17]+e[13]*e[12]*e[15]+e[10]*e[9]*e[15]+.5000000000*e[16]*ep2[15]-.5000000000*e[16]*ep2[11]-.5000000000*e[16]*ep2[12]-.5000000000*e[16]*ep2[14]+e[10]*e[11]*e[17]+.5000000000*ep2[10]*e[16]+.5000000000*ep3[16]-.5000000000*e[16]*ep2[9]+.5000000000*e[16]*ep2[17]+.5000000000*ep2[13]*e[16];
+        A[57]=e[10]*e[29]*e[20]+e[22]*e[27]*e[12]+e[22]*e[9]*e[30]+e[22]*e[29]*e[14]+e[22]*e[11]*e[32]+e[22]*e[10]*e[31]+e[31]*e[18]*e[12]+e[31]*e[9]*e[21]+e[31]*e[20]*e[14]+e[31]*e[11]*e[23]-1.*e[10]*e[33]*e[24]-1.*e[10]*e[30]*e[21]-1.*e[10]*e[35]*e[26]-1.*e[10]*e[32]*e[23]+e[10]*e[34]*e[25]+e[19]*e[27]*e[9]+e[19]*e[29]*e[11]+e[28]*e[18]*e[9]+e[28]*e[20]*e[11]+e[16]*e[27]*e[24]+e[16]*e[18]*e[33]+e[16]*e[28]*e[25]+e[16]*e[19]*e[34]+e[16]*e[29]*e[26]+e[16]*e[20]*e[35]-1.*e[19]*e[30]*e[12]-1.*e[19]*e[32]*e[14]-1.*e[19]*e[33]*e[15]-1.*e[19]*e[35]*e[17]-1.*e[28]*e[23]*e[14]-1.*e[28]*e[24]*e[15]-1.*e[28]*e[26]*e[17]-1.*e[28]*e[21]*e[12]+e[25]*e[27]*e[15]+e[25]*e[9]*e[33]+e[25]*e[29]*e[17]+e[25]*e[11]*e[35]+e[34]*e[18]*e[15]+e[34]*e[9]*e[24]+e[34]*e[20]*e[17]+e[34]*e[11]*e[26]+e[13]*e[27]*e[21]+e[13]*e[18]*e[30]+e[13]*e[28]*e[22]+e[13]*e[19]*e[31]+e[13]*e[29]*e[23]+e[13]*e[20]*e[32]+e[10]*e[27]*e[18]+3.*e[10]*e[28]*e[19];
+        A[171]=e[4]*e[30]*e[15]+e[4]*e[12]*e[33]+e[4]*e[32]*e[17]+e[4]*e[14]*e[35]+e[4]*e[31]*e[16]+e[4]*e[13]*e[34]+e[7]*e[33]*e[15]+e[7]*e[35]*e[17]+3.*e[7]*e[34]*e[16]+e[1]*e[27]*e[15]+e[1]*e[9]*e[33]+e[1]*e[29]*e[17]+e[1]*e[11]*e[35]+e[1]*e[28]*e[16]+e[1]*e[10]*e[34]-1.*e[16]*e[27]*e[0]-1.*e[16]*e[32]*e[5]+e[16]*e[33]*e[6]-1.*e[16]*e[30]*e[3]+e[16]*e[35]*e[8]-1.*e[16]*e[29]*e[2]+e[13]*e[30]*e[6]+e[13]*e[3]*e[33]+e[13]*e[31]*e[7]+e[13]*e[32]*e[8]+e[13]*e[5]*e[35]-1.*e[34]*e[11]*e[2]+e[34]*e[15]*e[6]-1.*e[34]*e[9]*e[0]-1.*e[34]*e[14]*e[5]-1.*e[34]*e[12]*e[3]+e[34]*e[17]*e[8]+e[31]*e[12]*e[6]+e[31]*e[3]*e[15]+e[31]*e[14]*e[8]+e[31]*e[5]*e[17]-1.*e[7]*e[27]*e[9]-1.*e[7]*e[30]*e[12]+e[7]*e[28]*e[10]-1.*e[7]*e[32]*e[14]+e[10]*e[27]*e[6]+e[10]*e[0]*e[33]+e[10]*e[29]*e[8]+e[10]*e[2]*e[35]+e[28]*e[9]*e[6]+e[28]*e[0]*e[15]+e[28]*e[11]*e[8]+e[28]*e[2]*e[17]-1.*e[7]*e[29]*e[11];
+        A[56]=e[22]*e[18]*e[12]+e[22]*e[9]*e[21]+e[22]*e[20]*e[14]+e[22]*e[11]*e[23]+e[22]*e[19]*e[13]+e[25]*e[18]*e[15]+e[25]*e[9]*e[24]+e[25]*e[20]*e[17]+e[25]*e[11]*e[26]+e[25]*e[19]*e[16]+e[16]*e[18]*e[24]+e[16]*e[20]*e[26]+e[13]*e[18]*e[21]+e[13]*e[20]*e[23]+e[19]*e[18]*e[9]+e[19]*e[20]*e[11]-1.*e[19]*e[23]*e[14]-1.*e[19]*e[24]*e[15]-1.*e[19]*e[26]*e[17]-1.*e[19]*e[21]*e[12]+.5000000000*e[10]*ep2[22]+.5000000000*e[10]*ep2[25]+1.500000000*e[10]*ep2[19]+.5000000000*e[10]*ep2[18]+.5000000000*e[10]*ep2[20]-.5000000000*e[10]*ep2[26]-.5000000000*e[10]*ep2[23]-.5000000000*e[10]*ep2[24]-.5000000000*e[10]*ep2[21];
+        A[170]=e[19]*e[20]*e[26]-.5000000000*e[25]*ep2[20]+e[22]*e[21]*e[24]+e[19]*e[18]*e[24]+.5000000000*ep2[22]*e[25]-.5000000000*e[25]*ep2[21]-.5000000000*e[25]*ep2[23]+.5000000000*ep2[19]*e[25]-.5000000000*e[25]*ep2[18]+.5000000000*e[25]*ep2[24]+.5000000000*e[25]*ep2[26]+.5000000000*ep3[25]+e[22]*e[23]*e[26];
+        A[73]=-1.*e[20]*e[33]*e[6]-1.*e[20]*e[30]*e[3]-1.*e[20]*e[31]*e[4]-1.*e[29]*e[21]*e[3]-1.*e[29]*e[22]*e[4]-1.*e[29]*e[25]*e[7]-1.*e[29]*e[24]*e[6]+e[8]*e[27]*e[24]+e[8]*e[18]*e[33]+e[8]*e[28]*e[25]+e[8]*e[19]*e[34]+e[23]*e[27]*e[3]+e[23]*e[0]*e[30]+e[23]*e[28]*e[4]+e[23]*e[1]*e[31]+e[32]*e[18]*e[3]+e[32]*e[0]*e[21]+e[32]*e[19]*e[4]+e[32]*e[1]*e[22]+e[26]*e[27]*e[6]+e[26]*e[0]*e[33]+e[26]*e[28]*e[7]+e[26]*e[1]*e[34]+e[26]*e[29]*e[8]+e[26]*e[2]*e[35]+e[35]*e[18]*e[6]+e[35]*e[0]*e[24]+e[35]*e[19]*e[7]+e[35]*e[1]*e[25]+e[35]*e[20]*e[8]+e[2]*e[27]*e[18]+e[2]*e[28]*e[19]+3.*e[2]*e[29]*e[20]+e[20]*e[27]*e[0]+e[20]*e[28]*e[1]+e[29]*e[18]*e[0]+e[29]*e[19]*e[1]+e[5]*e[27]*e[21]+e[5]*e[18]*e[30]+e[5]*e[28]*e[22]+e[5]*e[19]*e[31]+e[5]*e[29]*e[23]+e[5]*e[20]*e[32]-1.*e[2]*e[33]*e[24]-1.*e[2]*e[30]*e[21]-1.*e[2]*e[31]*e[22]+e[2]*e[32]*e[23]-1.*e[2]*e[34]*e[25]-1.*e[20]*e[34]*e[7];
+        A[72]=e[5]*e[18]*e[3]+e[5]*e[0]*e[21]+e[5]*e[19]*e[4]+e[5]*e[1]*e[22]+e[5]*e[2]*e[23]+e[23]*e[1]*e[4]+e[23]*e[0]*e[3]+e[8]*e[18]*e[6]+e[8]*e[0]*e[24]+e[8]*e[19]*e[7]+e[8]*e[1]*e[25]+e[8]*e[2]*e[26]+e[26]*e[1]*e[7]+e[26]*e[0]*e[6]+e[2]*e[18]*e[0]+e[2]*e[19]*e[1]-1.*e[2]*e[21]*e[3]-1.*e[2]*e[22]*e[4]-1.*e[2]*e[25]*e[7]-1.*e[2]*e[24]*e[6]-.5000000000*e[20]*ep2[4]+.5000000000*e[20]*ep2[0]-.5000000000*e[20]*ep2[6]+.5000000000*e[20]*ep2[5]+.5000000000*e[20]*ep2[1]-.5000000000*e[20]*ep2[7]-.5000000000*e[20]*ep2[3]+1.500000000*e[20]*ep2[2]+.5000000000*e[20]*ep2[8];
+        A[75]=e[14]*e[9]*e[3]+e[14]*e[0]*e[12]+e[14]*e[10]*e[4]+e[14]*e[1]*e[13]+e[14]*e[11]*e[5]+e[17]*e[9]*e[6]+e[17]*e[0]*e[15]+e[17]*e[10]*e[7]+e[17]*e[1]*e[16]+e[17]*e[11]*e[8]+e[8]*e[9]*e[15]+e[8]*e[10]*e[16]+e[5]*e[9]*e[12]+e[5]*e[10]*e[13]+e[11]*e[9]*e[0]+e[11]*e[10]*e[1]-1.*e[11]*e[13]*e[4]-1.*e[11]*e[16]*e[7]-1.*e[11]*e[15]*e[6]-1.*e[11]*e[12]*e[3]+.5000000000*e[2]*ep2[14]+.5000000000*e[2]*ep2[17]+1.500000000*e[2]*ep2[11]+.5000000000*e[2]*ep2[9]+.5000000000*e[2]*ep2[10]-.5000000000*e[2]*ep2[16]-.5000000000*e[2]*ep2[12]-.5000000000*e[2]*ep2[15]-.5000000000*e[2]*ep2[13];
+        A[74]=e[14]*e[18]*e[12]+e[14]*e[9]*e[21]+e[14]*e[11]*e[23]+e[14]*e[19]*e[13]+e[14]*e[10]*e[22]+e[23]*e[9]*e[12]+e[23]*e[10]*e[13]+e[17]*e[18]*e[15]+e[17]*e[9]*e[24]+e[17]*e[11]*e[26]+e[17]*e[19]*e[16]+e[17]*e[10]*e[25]+e[26]*e[9]*e[15]+e[26]*e[10]*e[16]-1.*e[11]*e[24]*e[15]-1.*e[11]*e[25]*e[16]+e[11]*e[18]*e[9]-1.*e[11]*e[21]*e[12]+e[11]*e[19]*e[10]-1.*e[11]*e[22]*e[13]+1.500000000*e[20]*ep2[11]+.5000000000*e[20]*ep2[9]+.5000000000*e[20]*ep2[10]+.5000000000*e[20]*ep2[14]+.5000000000*e[20]*ep2[17]-.5000000000*e[20]*ep2[16]-.5000000000*e[20]*ep2[12]-.5000000000*e[20]*ep2[15]-.5000000000*e[20]*ep2[13];
+        A[77]=e[23]*e[10]*e[31]+e[32]*e[18]*e[12]+e[32]*e[9]*e[21]+e[32]*e[19]*e[13]+e[32]*e[10]*e[22]-1.*e[11]*e[33]*e[24]-1.*e[11]*e[30]*e[21]+e[11]*e[35]*e[26]-1.*e[11]*e[31]*e[22]-1.*e[11]*e[34]*e[25]+e[20]*e[27]*e[9]+e[20]*e[28]*e[10]+e[29]*e[18]*e[9]+e[29]*e[19]*e[10]+e[17]*e[27]*e[24]+e[17]*e[18]*e[33]+e[17]*e[28]*e[25]+e[17]*e[19]*e[34]+e[17]*e[29]*e[26]+e[17]*e[20]*e[35]-1.*e[20]*e[30]*e[12]-1.*e[20]*e[31]*e[13]-1.*e[20]*e[33]*e[15]-1.*e[20]*e[34]*e[16]-1.*e[29]*e[24]*e[15]-1.*e[29]*e[25]*e[16]-1.*e[29]*e[21]*e[12]-1.*e[29]*e[22]*e[13]+e[26]*e[27]*e[15]+e[26]*e[9]*e[33]+e[26]*e[28]*e[16]+e[26]*e[10]*e[34]+e[35]*e[18]*e[15]+e[35]*e[9]*e[24]+e[35]*e[19]*e[16]+e[35]*e[10]*e[25]+e[14]*e[27]*e[21]+e[14]*e[18]*e[30]+e[14]*e[28]*e[22]+e[14]*e[19]*e[31]+e[14]*e[29]*e[23]+e[14]*e[20]*e[32]+e[11]*e[27]*e[18]+e[11]*e[28]*e[19]+3.*e[11]*e[29]*e[20]+e[23]*e[27]*e[12]+e[23]*e[9]*e[30]+e[23]*e[11]*e[32]+e[23]*e[28]*e[13];
+        A[76]=e[23]*e[18]*e[12]+e[23]*e[9]*e[21]+e[23]*e[20]*e[14]+e[23]*e[19]*e[13]+e[23]*e[10]*e[22]+e[26]*e[18]*e[15]+e[26]*e[9]*e[24]+e[26]*e[20]*e[17]+e[26]*e[19]*e[16]+e[26]*e[10]*e[25]+e[17]*e[19]*e[25]+e[17]*e[18]*e[24]+e[14]*e[19]*e[22]+e[14]*e[18]*e[21]+e[20]*e[18]*e[9]+e[20]*e[19]*e[10]-1.*e[20]*e[24]*e[15]-1.*e[20]*e[25]*e[16]-1.*e[20]*e[21]*e[12]-1.*e[20]*e[22]*e[13]+.5000000000*e[11]*ep2[23]+.5000000000*e[11]*ep2[26]+.5000000000*e[11]*ep2[19]+.5000000000*e[11]*ep2[18]+1.500000000*e[11]*ep2[20]-.5000000000*e[11]*ep2[22]-.5000000000*e[11]*ep2[24]-.5000000000*e[11]*ep2[21]-.5000000000*e[11]*ep2[25];
+        A[79]=-1.*e[20]*e[21]*e[3]+e[20]*e[26]*e[8]-1.*e[20]*e[22]*e[4]-1.*e[20]*e[25]*e[7]-1.*e[20]*e[24]*e[6]+e[5]*e[19]*e[22]+e[5]*e[18]*e[21]+e[26]*e[18]*e[6]+e[26]*e[0]*e[24]+e[26]*e[19]*e[7]+e[26]*e[1]*e[25]+e[8]*e[19]*e[25]+e[8]*e[18]*e[24]+e[20]*e[18]*e[0]+e[20]*e[19]*e[1]+e[23]*e[18]*e[3]+e[23]*e[0]*e[21]+e[23]*e[19]*e[4]+e[23]*e[1]*e[22]+e[23]*e[20]*e[5]+1.500000000*ep2[20]*e[2]+.5000000000*e[2]*ep2[23]+.5000000000*e[2]*ep2[19]+.5000000000*e[2]*ep2[18]+.5000000000*e[2]*ep2[26]-.5000000000*e[2]*ep2[22]-.5000000000*e[2]*ep2[24]-.5000000000*e[2]*ep2[21]-.5000000000*e[2]*ep2[25];
+        A[78]=-1.*e[2]*e[15]*e[6]+e[2]*e[9]*e[0]-1.*e[2]*e[12]*e[3]+e[5]*e[9]*e[3]+e[5]*e[0]*e[12]+e[5]*e[10]*e[4]+e[5]*e[1]*e[13]+e[5]*e[2]*e[14]+e[14]*e[1]*e[4]+e[14]*e[0]*e[3]+e[8]*e[9]*e[6]+e[8]*e[0]*e[15]+e[8]*e[10]*e[7]+e[8]*e[1]*e[16]+e[8]*e[2]*e[17]+e[17]*e[1]*e[7]+e[17]*e[0]*e[6]+e[2]*e[10]*e[1]-1.*e[2]*e[13]*e[4]-1.*e[2]*e[16]*e[7]+.5000000000*e[11]*ep2[1]+.5000000000*e[11]*ep2[0]+1.500000000*e[11]*ep2[2]+.5000000000*e[11]*ep2[5]+.5000000000*e[11]*ep2[8]-.5000000000*e[11]*ep2[4]-.5000000000*e[11]*ep2[6]-.5000000000*e[11]*ep2[7]-.5000000000*e[11]*ep2[3];
+        A[64]=e[5]*e[19]*e[13]+e[5]*e[10]*e[22]+e[8]*e[18]*e[15]+e[8]*e[9]*e[24]+e[8]*e[20]*e[17]+e[8]*e[11]*e[26]+e[8]*e[19]*e[16]+e[8]*e[10]*e[25]+e[2]*e[18]*e[9]+e[2]*e[19]*e[10]-1.*e[11]*e[21]*e[3]-1.*e[11]*e[22]*e[4]-1.*e[11]*e[25]*e[7]-1.*e[11]*e[24]*e[6]+e[14]*e[18]*e[3]+e[14]*e[0]*e[21]+e[14]*e[19]*e[4]+e[14]*e[1]*e[22]+e[14]*e[2]*e[23]-1.*e[20]*e[13]*e[4]-1.*e[20]*e[16]*e[7]-1.*e[20]*e[15]*e[6]-1.*e[20]*e[12]*e[3]+e[23]*e[9]*e[3]+e[23]*e[0]*e[12]+e[23]*e[10]*e[4]+e[23]*e[1]*e[13]+e[17]*e[18]*e[6]+e[17]*e[0]*e[24]+e[17]*e[19]*e[7]+e[17]*e[1]*e[25]+e[17]*e[2]*e[26]-1.*e[2]*e[24]*e[15]-1.*e[2]*e[25]*e[16]-1.*e[2]*e[21]*e[12]-1.*e[2]*e[22]*e[13]+e[26]*e[9]*e[6]+e[26]*e[0]*e[15]+e[26]*e[10]*e[7]+e[26]*e[1]*e[16]+e[11]*e[18]*e[0]+e[11]*e[19]*e[1]+3.*e[11]*e[20]*e[2]+e[20]*e[9]*e[0]+e[20]*e[10]*e[1]+e[5]*e[18]*e[12]+e[5]*e[9]*e[21]+e[5]*e[20]*e[14]+e[5]*e[11]*e[23];
+        A[65]=e[32]*e[1]*e[4]+e[32]*e[0]*e[3]+e[8]*e[27]*e[6]+e[8]*e[0]*e[33]+e[8]*e[28]*e[7]+e[8]*e[1]*e[34]+e[35]*e[1]*e[7]+e[35]*e[0]*e[6]+e[2]*e[27]*e[0]+e[2]*e[28]*e[1]-1.*e[2]*e[34]*e[7]+e[2]*e[32]*e[5]-1.*e[2]*e[33]*e[6]-1.*e[2]*e[30]*e[3]+e[2]*e[35]*e[8]-1.*e[2]*e[31]*e[4]+e[5]*e[27]*e[3]+e[5]*e[0]*e[30]+e[5]*e[28]*e[4]+e[5]*e[1]*e[31]+1.500000000*e[29]*ep2[2]-.5000000000*e[29]*ep2[4]+.5000000000*e[29]*ep2[0]-.5000000000*e[29]*ep2[6]+.5000000000*e[29]*ep2[5]+.5000000000*e[29]*ep2[1]-.5000000000*e[29]*ep2[7]-.5000000000*e[29]*ep2[3]+.5000000000*e[29]*ep2[8];
+        A[66]=e[5]*e[0]*e[3]+e[8]*e[1]*e[7]+e[8]*e[0]*e[6]+e[5]*e[1]*e[4]-.5000000000*e[2]*ep2[4]+.5000000000*ep3[2]+.5000000000*e[2]*ep2[1]-.5000000000*e[2]*ep2[3]+.5000000000*e[2]*ep2[0]+.5000000000*e[2]*ep2[8]+.5000000000*e[2]*ep2[5]-.5000000000*e[2]*ep2[6]-.5000000000*e[2]*ep2[7];
+        A[67]=e[35]*e[9]*e[15]+e[35]*e[10]*e[16]-1.*e[11]*e[30]*e[12]-1.*e[11]*e[31]*e[13]-1.*e[11]*e[33]*e[15]-1.*e[11]*e[34]*e[16]+e[11]*e[27]*e[9]+e[11]*e[28]*e[10]+e[14]*e[27]*e[12]+e[14]*e[9]*e[30]+e[14]*e[11]*e[32]+e[14]*e[28]*e[13]+e[14]*e[10]*e[31]+e[32]*e[9]*e[12]+e[32]*e[10]*e[13]+e[17]*e[27]*e[15]+e[17]*e[9]*e[33]+e[17]*e[11]*e[35]+e[17]*e[28]*e[16]+e[17]*e[10]*e[34]+1.500000000*e[29]*ep2[11]-.5000000000*e[29]*ep2[16]+.5000000000*e[29]*ep2[9]-.5000000000*e[29]*ep2[12]-.5000000000*e[29]*ep2[15]+.5000000000*e[29]*ep2[17]+.5000000000*e[29]*ep2[10]+.5000000000*e[29]*ep2[14]-.5000000000*e[29]*ep2[13];
+        A[68]=e[14]*e[9]*e[12]+e[17]*e[10]*e[16]+e[17]*e[9]*e[15]+.5000000000*ep3[11]+e[14]*e[10]*e[13]+.5000000000*e[11]*ep2[10]-.5000000000*e[11]*ep2[15]+.5000000000*e[11]*ep2[14]-.5000000000*e[11]*ep2[13]-.5000000000*e[11]*ep2[12]+.5000000000*e[11]*ep2[9]-.5000000000*e[11]*ep2[16]+.5000000000*e[11]*ep2[17];
+        A[69]=e[20]*e[27]*e[18]+e[20]*e[28]*e[19]+e[23]*e[27]*e[21]+e[23]*e[18]*e[30]+e[23]*e[28]*e[22]+e[23]*e[19]*e[31]+e[23]*e[20]*e[32]+e[32]*e[19]*e[22]+e[32]*e[18]*e[21]+e[26]*e[27]*e[24]+e[26]*e[18]*e[33]+e[26]*e[28]*e[25]+e[26]*e[19]*e[34]+e[26]*e[20]*e[35]+e[35]*e[19]*e[25]+e[35]*e[18]*e[24]-1.*e[20]*e[33]*e[24]-1.*e[20]*e[30]*e[21]-1.*e[20]*e[31]*e[22]-1.*e[20]*e[34]*e[25]+.5000000000*e[29]*ep2[23]+.5000000000*e[29]*ep2[26]-.5000000000*e[29]*ep2[22]-.5000000000*e[29]*ep2[24]-.5000000000*e[29]*ep2[21]-.5000000000*e[29]*ep2[25]+1.500000000*e[29]*ep2[20]+.5000000000*e[29]*ep2[19]+.5000000000*e[29]*ep2[18];
+        A[70]=.5000000000*e[20]*ep2[26]+.5000000000*e[20]*ep2[18]+.5000000000*ep3[20]+.5000000000*e[20]*ep2[19]+e[26]*e[18]*e[24]+.5000000000*e[20]*ep2[23]-.5000000000*e[20]*ep2[25]+e[23]*e[19]*e[22]-.5000000000*e[20]*ep2[24]-.5000000000*e[20]*ep2[21]-.5000000000*e[20]*ep2[22]+e[23]*e[18]*e[21]+e[26]*e[19]*e[25];
+        A[71]=e[8]*e[28]*e[16]+e[8]*e[10]*e[34]+e[2]*e[27]*e[9]+3.*e[2]*e[29]*e[11]+e[2]*e[28]*e[10]+e[11]*e[27]*e[0]-1.*e[11]*e[34]*e[7]-1.*e[11]*e[33]*e[6]-1.*e[11]*e[30]*e[3]+e[11]*e[28]*e[1]-1.*e[11]*e[31]*e[4]+e[14]*e[27]*e[3]+e[14]*e[0]*e[30]+e[14]*e[28]*e[4]+e[14]*e[1]*e[31]+e[14]*e[2]*e[32]+e[29]*e[10]*e[1]-1.*e[29]*e[13]*e[4]-1.*e[29]*e[16]*e[7]-1.*e[29]*e[15]*e[6]+e[29]*e[9]*e[0]-1.*e[29]*e[12]*e[3]+e[32]*e[9]*e[3]+e[32]*e[0]*e[12]+e[32]*e[10]*e[4]+e[32]*e[1]*e[13]+e[17]*e[27]*e[6]+e[17]*e[0]*e[33]+e[17]*e[28]*e[7]+e[17]*e[1]*e[34]+e[17]*e[2]*e[35]-1.*e[2]*e[30]*e[12]-1.*e[2]*e[31]*e[13]-1.*e[2]*e[33]*e[15]-1.*e[2]*e[34]*e[16]+e[35]*e[9]*e[6]+e[35]*e[0]*e[15]+e[35]*e[10]*e[7]+e[35]*e[1]*e[16]+e[5]*e[27]*e[12]+e[5]*e[9]*e[30]+e[5]*e[29]*e[14]+e[5]*e[11]*e[32]+e[5]*e[28]*e[13]+e[5]*e[10]*e[31]+e[8]*e[27]*e[15]+e[8]*e[9]*e[33]+e[8]*e[29]*e[17]+e[8]*e[11]*e[35];
+        A[91]=-1.*e[12]*e[34]*e[7]+e[12]*e[32]*e[5]-1.*e[12]*e[35]*e[8]-1.*e[12]*e[29]*e[2]-1.*e[12]*e[28]*e[1]+e[12]*e[31]*e[4]-1.*e[30]*e[11]*e[2]-1.*e[30]*e[10]*e[1]+e[30]*e[13]*e[4]-1.*e[30]*e[16]*e[7]+e[30]*e[14]*e[5]-1.*e[30]*e[17]*e[8]+e[15]*e[3]*e[33]+e[15]*e[31]*e[7]+e[15]*e[4]*e[34]+e[15]*e[32]*e[8]+e[15]*e[5]*e[35]+e[3]*e[27]*e[9]-1.*e[3]*e[28]*e[10]-1.*e[3]*e[34]*e[16]-1.*e[3]*e[35]*e[17]-1.*e[3]*e[29]*e[11]+e[33]*e[13]*e[7]+e[33]*e[4]*e[16]+e[33]*e[14]*e[8]+e[33]*e[5]*e[17]+e[9]*e[28]*e[4]+e[9]*e[1]*e[31]+e[9]*e[29]*e[5]+e[9]*e[2]*e[32]+e[27]*e[10]*e[4]+e[27]*e[1]*e[13]+e[27]*e[11]*e[5]+e[27]*e[2]*e[14]+3.*e[3]*e[30]*e[12]+e[3]*e[32]*e[14]+e[3]*e[31]*e[13]+e[6]*e[30]*e[15]+e[6]*e[12]*e[33]+e[6]*e[32]*e[17]+e[6]*e[14]*e[35]+e[6]*e[31]*e[16]+e[6]*e[13]*e[34]+e[0]*e[27]*e[12]+e[0]*e[9]*e[30]+e[0]*e[29]*e[14]+e[0]*e[11]*e[32]+e[0]*e[28]*e[13]+e[0]*e[10]*e[31];
+        A[90]=.5000000000*e[21]*ep2[24]-.5000000000*e[21]*ep2[25]+.5000000000*e[21]*ep2[23]-.5000000000*e[21]*ep2[26]+.5000000000*ep2[18]*e[21]+.5000000000*e[21]*ep2[22]-.5000000000*e[21]*ep2[20]+e[24]*e[22]*e[25]+e[24]*e[23]*e[26]-.5000000000*e[21]*ep2[19]+e[18]*e[19]*e[22]+e[18]*e[20]*e[23]+.5000000000*ep3[21];
+        A[89]=-.5000000000*e[30]*ep2[26]-.5000000000*e[30]*ep2[19]-.5000000000*e[30]*ep2[20]-.5000000000*e[30]*ep2[25]+.5000000000*ep2[18]*e[30]+1.500000000*e[30]*ep2[21]+.5000000000*e[30]*ep2[22]+.5000000000*e[30]*ep2[23]+.5000000000*e[30]*ep2[24]+e[18]*e[27]*e[21]+e[18]*e[28]*e[22]+e[18]*e[19]*e[31]+e[18]*e[29]*e[23]+e[18]*e[20]*e[32]+e[27]*e[19]*e[22]+e[27]*e[20]*e[23]+e[21]*e[31]*e[22]+e[21]*e[32]*e[23]+e[24]*e[21]*e[33]+e[24]*e[31]*e[25]+e[24]*e[22]*e[34]+e[24]*e[32]*e[26]+e[24]*e[23]*e[35]+e[33]*e[22]*e[25]+e[33]*e[23]*e[26]-1.*e[21]*e[29]*e[20]-1.*e[21]*e[35]*e[26]-1.*e[21]*e[28]*e[19]-1.*e[21]*e[34]*e[25];
+        A[88]=.5000000000*e[12]*ep2[15]-.5000000000*e[12]*ep2[17]+e[15]*e[13]*e[16]-.5000000000*e[12]*ep2[10]+e[15]*e[14]*e[17]-.5000000000*e[12]*ep2[16]-.5000000000*e[12]*ep2[11]+e[9]*e[10]*e[13]+.5000000000*e[12]*ep2[13]+.5000000000*ep2[9]*e[12]+.5000000000*ep3[12]+e[9]*e[11]*e[14]+.5000000000*e[12]*ep2[14];
+        A[95]=e[12]*e[13]*e[4]+e[12]*e[14]*e[5]+e[15]*e[12]*e[6]+e[15]*e[13]*e[7]+e[15]*e[4]*e[16]+e[15]*e[14]*e[8]+e[15]*e[5]*e[17]+e[6]*e[14]*e[17]+e[6]*e[13]*e[16]+e[0]*e[11]*e[14]+e[0]*e[9]*e[12]+e[0]*e[10]*e[13]+e[9]*e[10]*e[4]+e[9]*e[1]*e[13]+e[9]*e[11]*e[5]+e[9]*e[2]*e[14]-1.*e[12]*e[11]*e[2]-1.*e[12]*e[10]*e[1]-1.*e[12]*e[16]*e[7]-1.*e[12]*e[17]*e[8]+1.500000000*ep2[12]*e[3]+.5000000000*e[3]*ep2[15]-.5000000000*e[3]*ep2[16]+.5000000000*e[3]*ep2[9]-.5000000000*e[3]*ep2[11]-.5000000000*e[3]*ep2[17]-.5000000000*e[3]*ep2[10]+.5000000000*e[3]*ep2[14]+.5000000000*e[3]*ep2[13];
+        A[94]=e[18]*e[11]*e[14]+e[18]*e[9]*e[12]+e[18]*e[10]*e[13]+e[12]*e[23]*e[14]+e[12]*e[22]*e[13]+e[15]*e[12]*e[24]+e[15]*e[23]*e[17]+e[15]*e[14]*e[26]+e[15]*e[22]*e[16]+e[15]*e[13]*e[25]+e[24]*e[14]*e[17]+e[24]*e[13]*e[16]-1.*e[12]*e[25]*e[16]-1.*e[12]*e[26]*e[17]-1.*e[12]*e[20]*e[11]-1.*e[12]*e[19]*e[10]+e[9]*e[20]*e[14]+e[9]*e[11]*e[23]+e[9]*e[19]*e[13]+e[9]*e[10]*e[22]+.5000000000*ep2[9]*e[21]-.5000000000*e[21]*ep2[16]-.5000000000*e[21]*ep2[11]-.5000000000*e[21]*ep2[17]-.5000000000*e[21]*ep2[10]+1.500000000*e[21]*ep2[12]+.5000000000*e[21]*ep2[14]+.5000000000*e[21]*ep2[13]+.5000000000*e[21]*ep2[15];
+        A[93]=-1.*e[21]*e[35]*e[8]-1.*e[21]*e[29]*e[2]-1.*e[21]*e[28]*e[1]+e[21]*e[31]*e[4]-1.*e[30]*e[26]*e[8]-1.*e[30]*e[20]*e[2]-1.*e[30]*e[19]*e[1]+e[30]*e[22]*e[4]-1.*e[30]*e[25]*e[7]+e[30]*e[23]*e[5]+e[6]*e[31]*e[25]+e[6]*e[22]*e[34]+e[6]*e[32]*e[26]+e[6]*e[23]*e[35]+e[24]*e[30]*e[6]+e[24]*e[3]*e[33]+e[24]*e[31]*e[7]+e[24]*e[4]*e[34]+e[24]*e[32]*e[8]+e[24]*e[5]*e[35]+e[33]*e[21]*e[6]+e[33]*e[22]*e[7]+e[33]*e[4]*e[25]+e[33]*e[23]*e[8]+e[33]*e[5]*e[26]+e[0]*e[27]*e[21]+e[0]*e[18]*e[30]+e[0]*e[28]*e[22]+e[0]*e[19]*e[31]+e[0]*e[29]*e[23]+e[0]*e[20]*e[32]+e[18]*e[27]*e[3]+e[18]*e[28]*e[4]+e[18]*e[1]*e[31]+e[18]*e[29]*e[5]+e[18]*e[2]*e[32]+e[27]*e[19]*e[4]+e[27]*e[1]*e[22]+e[27]*e[20]*e[5]+e[27]*e[2]*e[23]+3.*e[3]*e[30]*e[21]+e[3]*e[31]*e[22]+e[3]*e[32]*e[23]-1.*e[3]*e[29]*e[20]-1.*e[3]*e[35]*e[26]-1.*e[3]*e[28]*e[19]-1.*e[3]*e[34]*e[25]-1.*e[21]*e[34]*e[7]+e[21]*e[32]*e[5];
+        A[92]=e[18]*e[1]*e[4]+e[18]*e[0]*e[3]+e[18]*e[2]*e[5]+e[3]*e[22]*e[4]+e[3]*e[23]*e[5]+e[6]*e[3]*e[24]+e[6]*e[22]*e[7]+e[6]*e[4]*e[25]+e[6]*e[23]*e[8]+e[6]*e[5]*e[26]+e[24]*e[4]*e[7]+e[24]*e[5]*e[8]+e[0]*e[19]*e[4]+e[0]*e[1]*e[22]+e[0]*e[20]*e[5]+e[0]*e[2]*e[23]-1.*e[3]*e[26]*e[8]-1.*e[3]*e[20]*e[2]-1.*e[3]*e[19]*e[1]-1.*e[3]*e[25]*e[7]+.5000000000*e[21]*ep2[4]+.5000000000*e[21]*ep2[0]+.5000000000*e[21]*ep2[6]+.5000000000*e[21]*ep2[5]-.5000000000*e[21]*ep2[1]-.5000000000*e[21]*ep2[7]+1.500000000*e[21]*ep2[3]-.5000000000*e[21]*ep2[2]-.5000000000*e[21]*ep2[8];
+        A[82]=.5000000000*ep2[27]*e[21]+1.500000000*e[21]*ep2[30]+.5000000000*e[21]*ep2[32]+.5000000000*e[21]*ep2[31]+.5000000000*e[21]*ep2[33]-.5000000000*e[21]*ep2[28]-.5000000000*e[21]*ep2[29]-.5000000000*e[21]*ep2[34]-.5000000000*e[21]*ep2[35]+e[18]*e[27]*e[30]+e[18]*e[29]*e[32]+e[18]*e[28]*e[31]+e[27]*e[28]*e[22]+e[27]*e[19]*e[31]+e[27]*e[29]*e[23]+e[27]*e[20]*e[32]+e[30]*e[31]*e[22]+e[30]*e[32]*e[23]+e[24]*e[30]*e[33]+e[24]*e[32]*e[35]+e[24]*e[31]*e[34]+e[33]*e[31]*e[25]+e[33]*e[22]*e[34]+e[33]*e[32]*e[26]+e[33]*e[23]*e[35]-1.*e[30]*e[29]*e[20]-1.*e[30]*e[35]*e[26]-1.*e[30]*e[28]*e[19]-1.*e[30]*e[34]*e[25];
+        A[192]=-.5000000000*e[26]*ep2[4]-.5000000000*e[26]*ep2[0]+.5000000000*e[26]*ep2[6]+.5000000000*e[26]*ep2[5]-.5000000000*e[26]*ep2[1]+.5000000000*e[26]*ep2[7]-.5000000000*e[26]*ep2[3]+.5000000000*e[26]*ep2[2]+1.500000000*e[26]*ep2[8]+e[20]*e[0]*e[6]+e[20]*e[2]*e[8]+e[5]*e[21]*e[6]+e[5]*e[3]*e[24]+e[5]*e[22]*e[7]+e[5]*e[4]*e[25]+e[5]*e[23]*e[8]+e[23]*e[4]*e[7]+e[23]*e[3]*e[6]+e[8]*e[24]*e[6]+e[8]*e[25]*e[7]+e[2]*e[18]*e[6]+e[2]*e[0]*e[24]+e[2]*e[19]*e[7]+e[2]*e[1]*e[25]-1.*e[8]*e[21]*e[3]-1.*e[8]*e[19]*e[1]-1.*e[8]*e[22]*e[4]-1.*e[8]*e[18]*e[0]+e[20]*e[1]*e[7];
+        A[83]=e[9]*e[27]*e[30]+e[9]*e[29]*e[32]+e[9]*e[28]*e[31]+e[33]*e[30]*e[15]+e[33]*e[32]*e[17]+e[33]*e[14]*e[35]+e[33]*e[31]*e[16]+e[33]*e[13]*e[34]+e[27]*e[29]*e[14]+e[27]*e[11]*e[32]+e[27]*e[28]*e[13]+e[27]*e[10]*e[31]-1.*e[30]*e[28]*e[10]+e[30]*e[31]*e[13]+e[30]*e[32]*e[14]-1.*e[30]*e[34]*e[16]-1.*e[30]*e[35]*e[17]-1.*e[30]*e[29]*e[11]+e[15]*e[32]*e[35]+e[15]*e[31]*e[34]-.5000000000*e[12]*ep2[34]-.5000000000*e[12]*ep2[35]+.5000000000*e[12]*ep2[27]+.5000000000*e[12]*ep2[32]-.5000000000*e[12]*ep2[28]-.5000000000*e[12]*ep2[29]+.5000000000*e[12]*ep2[31]+.5000000000*e[12]*ep2[33]+1.500000000*e[12]*ep2[30];
+        A[193]=e[23]*e[30]*e[6]+e[23]*e[3]*e[33]+e[23]*e[31]*e[7]+e[23]*e[4]*e[34]+e[32]*e[21]*e[6]+e[32]*e[3]*e[24]+e[32]*e[22]*e[7]+e[32]*e[4]*e[25]+e[26]*e[33]*e[6]+e[26]*e[34]*e[7]+3.*e[26]*e[35]*e[8]+e[35]*e[24]*e[6]+e[35]*e[25]*e[7]+e[2]*e[27]*e[24]+e[2]*e[18]*e[33]+e[2]*e[28]*e[25]+e[2]*e[19]*e[34]+e[2]*e[29]*e[26]+e[2]*e[20]*e[35]+e[20]*e[27]*e[6]+e[20]*e[0]*e[33]+e[20]*e[28]*e[7]+e[20]*e[1]*e[34]+e[20]*e[29]*e[8]+e[29]*e[18]*e[6]+e[29]*e[0]*e[24]+e[29]*e[19]*e[7]+e[29]*e[1]*e[25]+e[5]*e[30]*e[24]+e[5]*e[21]*e[33]+e[5]*e[31]*e[25]+e[5]*e[22]*e[34]+e[5]*e[32]*e[26]+e[5]*e[23]*e[35]-1.*e[8]*e[27]*e[18]+e[8]*e[33]*e[24]-1.*e[8]*e[30]*e[21]-1.*e[8]*e[31]*e[22]+e[8]*e[32]*e[23]-1.*e[8]*e[28]*e[19]+e[8]*e[34]*e[25]-1.*e[26]*e[27]*e[0]-1.*e[26]*e[30]*e[3]-1.*e[26]*e[28]*e[1]-1.*e[26]*e[31]*e[4]-1.*e[35]*e[21]*e[3]-1.*e[35]*e[19]*e[1]-1.*e[35]*e[22]*e[4]-1.*e[35]*e[18]*e[0];
+        A[80]=e[27]*e[29]*e[32]+e[27]*e[28]*e[31]+e[33]*e[32]*e[35]+e[33]*e[31]*e[34]+.5000000000*ep3[30]-.5000000000*e[30]*ep2[28]-.5000000000*e[30]*ep2[29]-.5000000000*e[30]*ep2[34]+.5000000000*e[30]*ep2[33]+.5000000000*ep2[27]*e[30]+.5000000000*e[30]*ep2[32]+.5000000000*e[30]*ep2[31]-.5000000000*e[30]*ep2[35];
+        A[194]=.5000000000*ep2[14]*e[26]+1.500000000*e[26]*ep2[17]+.5000000000*e[26]*ep2[15]+.5000000000*e[26]*ep2[16]+.5000000000*ep2[11]*e[26]-.5000000000*e[26]*ep2[9]-.5000000000*e[26]*ep2[12]-.5000000000*e[26]*ep2[10]-.5000000000*e[26]*ep2[13]+e[20]*e[11]*e[17]+e[20]*e[9]*e[15]+e[20]*e[10]*e[16]+e[14]*e[21]*e[15]+e[14]*e[12]*e[24]+e[14]*e[23]*e[17]+e[14]*e[22]*e[16]+e[14]*e[13]*e[25]+e[23]*e[12]*e[15]+e[23]*e[13]*e[16]+e[17]*e[24]*e[15]+e[17]*e[25]*e[16]-1.*e[17]*e[18]*e[9]-1.*e[17]*e[21]*e[12]-1.*e[17]*e[19]*e[10]-1.*e[17]*e[22]*e[13]+e[11]*e[18]*e[15]+e[11]*e[9]*e[24]+e[11]*e[19]*e[16]+e[11]*e[10]*e[25];
+        A[81]=e[0]*e[27]*e[30]+e[0]*e[29]*e[32]+e[0]*e[28]*e[31]+e[30]*e[31]*e[4]+e[30]*e[32]*e[5]+e[6]*e[30]*e[33]+e[6]*e[32]*e[35]+e[6]*e[31]*e[34]+e[27]*e[28]*e[4]+e[27]*e[1]*e[31]+e[27]*e[29]*e[5]+e[27]*e[2]*e[32]+e[33]*e[31]*e[7]+e[33]*e[4]*e[34]+e[33]*e[32]*e[8]+e[33]*e[5]*e[35]-1.*e[30]*e[34]*e[7]-1.*e[30]*e[35]*e[8]-1.*e[30]*e[29]*e[2]-1.*e[30]*e[28]*e[1]+1.500000000*e[3]*ep2[30]+.5000000000*e[3]*ep2[32]+.5000000000*e[3]*ep2[31]+.5000000000*e[3]*ep2[27]-.5000000000*e[3]*ep2[28]-.5000000000*e[3]*ep2[29]+.5000000000*e[3]*ep2[33]-.5000000000*e[3]*ep2[34]-.5000000000*e[3]*ep2[35];
+        A[195]=.5000000000*ep2[14]*e[8]+1.500000000*ep2[17]*e[8]+.5000000000*e[8]*ep2[15]+.5000000000*e[8]*ep2[16]-.5000000000*e[8]*ep2[9]+.5000000000*e[8]*ep2[11]-.5000000000*e[8]*ep2[12]-.5000000000*e[8]*ep2[10]-.5000000000*e[8]*ep2[13]+e[14]*e[12]*e[6]+e[14]*e[3]*e[15]+e[14]*e[13]*e[7]+e[14]*e[4]*e[16]+e[14]*e[5]*e[17]+e[17]*e[15]*e[6]+e[17]*e[16]*e[7]+e[2]*e[11]*e[17]+e[2]*e[9]*e[15]+e[2]*e[10]*e[16]+e[5]*e[12]*e[15]+e[5]*e[13]*e[16]+e[11]*e[9]*e[6]+e[11]*e[0]*e[15]+e[11]*e[10]*e[7]+e[11]*e[1]*e[16]-1.*e[17]*e[10]*e[1]-1.*e[17]*e[13]*e[4]-1.*e[17]*e[9]*e[0]-1.*e[17]*e[12]*e[3];
+        A[86]=-.5000000000*e[3]*ep2[1]-.5000000000*e[3]*ep2[7]+.5000000000*ep3[3]-.5000000000*e[3]*ep2[8]+e[0]*e[2]*e[5]+.5000000000*e[3]*ep2[6]+.5000000000*e[3]*ep2[4]-.5000000000*e[3]*ep2[2]+e[0]*e[1]*e[4]+e[6]*e[4]*e[7]+.5000000000*ep2[0]*e[3]+.5000000000*e[3]*ep2[5]+e[6]*e[5]*e[8];
+        A[196]=.5000000000*ep2[23]*e[17]+1.500000000*ep2[26]*e[17]+.5000000000*e[17]*ep2[25]+.5000000000*e[17]*ep2[24]-.5000000000*e[17]*ep2[18]-.5000000000*e[17]*ep2[19]+.5000000000*e[17]*ep2[20]-.5000000000*e[17]*ep2[22]-.5000000000*e[17]*ep2[21]+e[23]*e[21]*e[15]+e[23]*e[12]*e[24]+e[23]*e[14]*e[26]+e[23]*e[22]*e[16]+e[23]*e[13]*e[25]+e[26]*e[24]*e[15]+e[26]*e[25]*e[16]+e[11]*e[19]*e[25]+e[11]*e[18]*e[24]+e[11]*e[20]*e[26]+e[14]*e[22]*e[25]+e[14]*e[21]*e[24]+e[20]*e[18]*e[15]+e[20]*e[9]*e[24]+e[20]*e[19]*e[16]+e[20]*e[10]*e[25]-1.*e[26]*e[18]*e[9]-1.*e[26]*e[21]*e[12]-1.*e[26]*e[19]*e[10]-1.*e[26]*e[22]*e[13];
+        A[87]=-1.*e[12]*e[34]*e[16]-1.*e[12]*e[35]*e[17]-1.*e[12]*e[29]*e[11]+e[9]*e[27]*e[12]+e[9]*e[29]*e[14]+e[9]*e[11]*e[32]+e[9]*e[28]*e[13]+e[9]*e[10]*e[31]+e[27]*e[11]*e[14]+e[27]*e[10]*e[13]+e[12]*e[32]*e[14]+e[12]*e[31]*e[13]+e[15]*e[12]*e[33]+e[15]*e[32]*e[17]+e[15]*e[14]*e[35]+e[15]*e[31]*e[16]+e[15]*e[13]*e[34]+e[33]*e[14]*e[17]+e[33]*e[13]*e[16]-1.*e[12]*e[28]*e[10]+.5000000000*ep2[9]*e[30]-.5000000000*e[30]*ep2[16]-.5000000000*e[30]*ep2[11]+1.500000000*e[30]*ep2[12]+.5000000000*e[30]*ep2[15]-.5000000000*e[30]*ep2[17]-.5000000000*e[30]*ep2[10]+.5000000000*e[30]*ep2[14]+.5000000000*e[30]*ep2[13];
+        A[197]=e[32]*e[22]*e[16]+e[32]*e[13]*e[25]-1.*e[17]*e[27]*e[18]+e[17]*e[33]*e[24]-1.*e[17]*e[30]*e[21]+e[17]*e[29]*e[20]+3.*e[17]*e[35]*e[26]-1.*e[17]*e[31]*e[22]-1.*e[17]*e[28]*e[19]+e[17]*e[34]*e[25]+e[20]*e[27]*e[15]+e[20]*e[9]*e[33]+e[20]*e[28]*e[16]+e[20]*e[10]*e[34]+e[29]*e[18]*e[15]+e[29]*e[9]*e[24]+e[29]*e[19]*e[16]+e[29]*e[10]*e[25]-1.*e[26]*e[27]*e[9]-1.*e[26]*e[30]*e[12]-1.*e[26]*e[28]*e[10]-1.*e[26]*e[31]*e[13]+e[26]*e[33]*e[15]+e[26]*e[34]*e[16]+e[35]*e[24]*e[15]+e[35]*e[25]*e[16]-1.*e[35]*e[18]*e[9]-1.*e[35]*e[21]*e[12]-1.*e[35]*e[19]*e[10]-1.*e[35]*e[22]*e[13]+e[14]*e[30]*e[24]+e[14]*e[21]*e[33]+e[14]*e[31]*e[25]+e[14]*e[22]*e[34]+e[14]*e[32]*e[26]+e[14]*e[23]*e[35]+e[11]*e[27]*e[24]+e[11]*e[18]*e[33]+e[11]*e[28]*e[25]+e[11]*e[19]*e[34]+e[11]*e[29]*e[26]+e[11]*e[20]*e[35]+e[23]*e[30]*e[15]+e[23]*e[12]*e[33]+e[23]*e[32]*e[17]+e[23]*e[31]*e[16]+e[23]*e[13]*e[34]+e[32]*e[21]*e[15]+e[32]*e[12]*e[24];
+        A[84]=e[6]*e[23]*e[17]+e[6]*e[14]*e[26]+e[6]*e[22]*e[16]+e[6]*e[13]*e[25]+e[0]*e[20]*e[14]+e[0]*e[11]*e[23]+e[0]*e[19]*e[13]+e[0]*e[10]*e[22]-1.*e[12]*e[26]*e[8]-1.*e[12]*e[20]*e[2]-1.*e[12]*e[19]*e[1]+e[12]*e[22]*e[4]-1.*e[12]*e[25]*e[7]+e[12]*e[23]*e[5]-1.*e[21]*e[11]*e[2]-1.*e[21]*e[10]*e[1]+e[21]*e[13]*e[4]-1.*e[21]*e[16]*e[7]+e[21]*e[14]*e[5]-1.*e[21]*e[17]*e[8]+e[15]*e[3]*e[24]+e[15]*e[22]*e[7]+e[15]*e[4]*e[25]+e[15]*e[23]*e[8]+e[15]*e[5]*e[26]-1.*e[3]*e[25]*e[16]-1.*e[3]*e[26]*e[17]-1.*e[3]*e[20]*e[11]-1.*e[3]*e[19]*e[10]+e[24]*e[13]*e[7]+e[24]*e[4]*e[16]+e[24]*e[14]*e[8]+e[24]*e[5]*e[17]+e[9]*e[18]*e[3]+e[9]*e[0]*e[21]+e[9]*e[19]*e[4]+e[9]*e[1]*e[22]+e[9]*e[20]*e[5]+e[9]*e[2]*e[23]+e[18]*e[0]*e[12]+e[18]*e[10]*e[4]+e[18]*e[1]*e[13]+e[18]*e[11]*e[5]+e[18]*e[2]*e[14]+3.*e[3]*e[21]*e[12]+e[3]*e[23]*e[14]+e[3]*e[22]*e[13]+e[6]*e[21]*e[15]+e[6]*e[12]*e[24];
+        A[198]=.5000000000*ep2[5]*e[17]+1.500000000*e[17]*ep2[8]+.5000000000*e[17]*ep2[7]+.5000000000*e[17]*ep2[6]+.5000000000*ep2[2]*e[17]-.5000000000*e[17]*ep2[4]-.5000000000*e[17]*ep2[0]-.5000000000*e[17]*ep2[1]-.5000000000*e[17]*ep2[3]+e[11]*e[1]*e[7]+e[11]*e[0]*e[6]+e[11]*e[2]*e[8]+e[5]*e[12]*e[6]+e[5]*e[3]*e[15]+e[5]*e[13]*e[7]+e[5]*e[4]*e[16]+e[5]*e[14]*e[8]+e[14]*e[4]*e[7]+e[14]*e[3]*e[6]+e[8]*e[15]*e[6]+e[8]*e[16]*e[7]-1.*e[8]*e[10]*e[1]-1.*e[8]*e[13]*e[4]-1.*e[8]*e[9]*e[0]-1.*e[8]*e[12]*e[3]+e[2]*e[9]*e[6]+e[2]*e[0]*e[15]+e[2]*e[10]*e[7]+e[2]*e[1]*e[16];
+        A[85]=e[6]*e[4]*e[34]+e[6]*e[32]*e[8]+e[6]*e[5]*e[35]+e[33]*e[4]*e[7]+e[33]*e[5]*e[8]+e[0]*e[27]*e[3]+e[0]*e[28]*e[4]+e[0]*e[1]*e[31]+e[0]*e[29]*e[5]+e[0]*e[2]*e[32]-1.*e[3]*e[34]*e[7]+e[3]*e[32]*e[5]+e[3]*e[33]*e[6]-1.*e[3]*e[35]*e[8]-1.*e[3]*e[29]*e[2]-1.*e[3]*e[28]*e[1]+e[3]*e[31]*e[4]+e[27]*e[1]*e[4]+e[27]*e[2]*e[5]+e[6]*e[31]*e[7]+.5000000000*e[30]*ep2[4]+.5000000000*e[30]*ep2[6]+.5000000000*e[30]*ep2[5]-.5000000000*e[30]*ep2[1]-.5000000000*e[30]*ep2[7]-.5000000000*e[30]*ep2[2]-.5000000000*e[30]*ep2[8]+.5000000000*ep2[0]*e[30]+1.500000000*e[30]*ep2[3];
+        A[199]=.5000000000*ep2[23]*e[8]+1.500000000*ep2[26]*e[8]-.5000000000*e[8]*ep2[18]-.5000000000*e[8]*ep2[19]-.5000000000*e[8]*ep2[22]+.5000000000*e[8]*ep2[24]-.5000000000*e[8]*ep2[21]+.5000000000*e[8]*ep2[25]+.5000000000*ep2[20]*e[8]+e[20]*e[18]*e[6]+e[20]*e[0]*e[24]+e[20]*e[19]*e[7]+e[20]*e[1]*e[25]+e[20]*e[2]*e[26]+e[23]*e[21]*e[6]+e[23]*e[3]*e[24]+e[23]*e[22]*e[7]+e[23]*e[4]*e[25]+e[23]*e[5]*e[26]-1.*e[26]*e[21]*e[3]-1.*e[26]*e[19]*e[1]-1.*e[26]*e[22]*e[4]-1.*e[26]*e[18]*e[0]+e[26]*e[25]*e[7]+e[26]*e[24]*e[6]+e[2]*e[19]*e[25]+e[2]*e[18]*e[24]+e[5]*e[22]*e[25]+e[5]*e[21]*e[24];
+        A[109]=e[19]*e[27]*e[21]+e[19]*e[18]*e[30]+e[19]*e[28]*e[22]+e[19]*e[29]*e[23]+e[19]*e[20]*e[32]+e[28]*e[18]*e[21]+e[28]*e[20]*e[23]+e[22]*e[30]*e[21]+e[22]*e[32]*e[23]+e[25]*e[30]*e[24]+e[25]*e[21]*e[33]+e[25]*e[22]*e[34]+e[25]*e[32]*e[26]+e[25]*e[23]*e[35]+e[34]*e[21]*e[24]+e[34]*e[23]*e[26]-1.*e[22]*e[27]*e[18]-1.*e[22]*e[33]*e[24]-1.*e[22]*e[29]*e[20]-1.*e[22]*e[35]*e[26]+.5000000000*ep2[19]*e[31]+1.500000000*e[31]*ep2[22]+.5000000000*e[31]*ep2[21]+.5000000000*e[31]*ep2[23]+.5000000000*e[31]*ep2[25]-.5000000000*e[31]*ep2[26]-.5000000000*e[31]*ep2[18]-.5000000000*e[31]*ep2[20]-.5000000000*e[31]*ep2[24];
+        A[108]=-.5000000000*e[13]*ep2[15]+.5000000000*e[13]*ep2[16]+.5000000000*e[13]*ep2[12]+e[16]*e[12]*e[15]+.5000000000*ep3[13]+e[10]*e[11]*e[14]+.5000000000*e[13]*ep2[14]-.5000000000*e[13]*ep2[17]-.5000000000*e[13]*ep2[11]-.5000000000*e[13]*ep2[9]+.5000000000*ep2[10]*e[13]+e[10]*e[9]*e[12]+e[16]*e[14]*e[17];
+        A[111]=-1.*e[13]*e[29]*e[2]-1.*e[31]*e[11]*e[2]-1.*e[31]*e[15]*e[6]-1.*e[31]*e[9]*e[0]+e[31]*e[14]*e[5]+e[31]*e[12]*e[3]-1.*e[31]*e[17]*e[8]+e[16]*e[30]*e[6]+e[16]*e[3]*e[33]+e[16]*e[4]*e[34]+e[16]*e[32]*e[8]+e[16]*e[5]*e[35]-1.*e[4]*e[27]*e[9]+e[4]*e[28]*e[10]-1.*e[4]*e[33]*e[15]-1.*e[4]*e[35]*e[17]-1.*e[4]*e[29]*e[11]+e[34]*e[12]*e[6]+e[34]*e[3]*e[15]+e[34]*e[14]*e[8]+e[34]*e[5]*e[17]+e[10]*e[27]*e[3]+e[10]*e[0]*e[30]+e[10]*e[29]*e[5]+e[10]*e[2]*e[32]+e[28]*e[9]*e[3]+e[28]*e[0]*e[12]+e[28]*e[11]*e[5]+e[28]*e[2]*e[14]+e[4]*e[30]*e[12]+e[4]*e[32]*e[14]+3.*e[4]*e[31]*e[13]+e[7]*e[30]*e[15]+e[7]*e[12]*e[33]+e[7]*e[32]*e[17]+e[7]*e[14]*e[35]+e[7]*e[31]*e[16]+e[7]*e[13]*e[34]+e[1]*e[27]*e[12]+e[1]*e[9]*e[30]+e[1]*e[29]*e[14]+e[1]*e[11]*e[32]+e[1]*e[28]*e[13]+e[1]*e[10]*e[31]-1.*e[13]*e[27]*e[0]+e[13]*e[32]*e[5]-1.*e[13]*e[33]*e[6]+e[13]*e[30]*e[3]-1.*e[13]*e[35]*e[8];
+        A[110]=e[25]*e[23]*e[26]+e[19]*e[20]*e[23]+e[19]*e[18]*e[21]+e[25]*e[21]*e[24]+.5000000000*ep3[22]+.5000000000*e[22]*ep2[23]+.5000000000*ep2[19]*e[22]-.5000000000*e[22]*ep2[18]-.5000000000*e[22]*ep2[24]+.5000000000*e[22]*ep2[21]+.5000000000*e[22]*ep2[25]-.5000000000*e[22]*ep2[20]-.5000000000*e[22]*ep2[26];
+        A[105]=e[34]*e[5]*e[8]+e[1]*e[27]*e[3]+e[1]*e[0]*e[30]+e[1]*e[28]*e[4]+e[1]*e[29]*e[5]+e[1]*e[2]*e[32]-1.*e[4]*e[27]*e[0]+e[4]*e[34]*e[7]+e[4]*e[32]*e[5]-1.*e[4]*e[33]*e[6]+e[4]*e[30]*e[3]-1.*e[4]*e[35]*e[8]-1.*e[4]*e[29]*e[2]+e[28]*e[0]*e[3]+e[28]*e[2]*e[5]+e[7]*e[30]*e[6]+e[7]*e[3]*e[33]+e[7]*e[32]*e[8]+e[7]*e[5]*e[35]+e[34]*e[3]*e[6]+.5000000000*ep2[1]*e[31]+1.500000000*e[31]*ep2[4]-.5000000000*e[31]*ep2[0]-.5000000000*e[31]*ep2[6]+.5000000000*e[31]*ep2[5]+.5000000000*e[31]*ep2[7]+.5000000000*e[31]*ep2[3]-.5000000000*e[31]*ep2[2]-.5000000000*e[31]*ep2[8];
+        A[104]=e[1]*e[20]*e[14]+e[1]*e[11]*e[23]+e[13]*e[21]*e[3]-1.*e[13]*e[26]*e[8]-1.*e[13]*e[20]*e[2]-1.*e[13]*e[18]*e[0]+e[13]*e[23]*e[5]-1.*e[13]*e[24]*e[6]-1.*e[22]*e[11]*e[2]-1.*e[22]*e[15]*e[6]-1.*e[22]*e[9]*e[0]+e[22]*e[14]*e[5]+e[22]*e[12]*e[3]-1.*e[22]*e[17]*e[8]+e[16]*e[21]*e[6]+e[16]*e[3]*e[24]+e[16]*e[4]*e[25]+e[16]*e[23]*e[8]+e[16]*e[5]*e[26]-1.*e[4]*e[24]*e[15]-1.*e[4]*e[26]*e[17]-1.*e[4]*e[20]*e[11]-1.*e[4]*e[18]*e[9]+e[25]*e[12]*e[6]+e[25]*e[3]*e[15]+e[25]*e[14]*e[8]+e[25]*e[5]*e[17]+e[10]*e[18]*e[3]+e[10]*e[0]*e[21]+e[10]*e[19]*e[4]+e[10]*e[1]*e[22]+e[10]*e[20]*e[5]+e[10]*e[2]*e[23]+e[19]*e[9]*e[3]+e[19]*e[0]*e[12]+e[19]*e[1]*e[13]+e[19]*e[11]*e[5]+e[19]*e[2]*e[14]+e[4]*e[21]*e[12]+e[4]*e[23]*e[14]+3.*e[4]*e[22]*e[13]+e[7]*e[21]*e[15]+e[7]*e[12]*e[24]+e[7]*e[23]*e[17]+e[7]*e[14]*e[26]+e[7]*e[22]*e[16]+e[7]*e[13]*e[25]+e[1]*e[18]*e[12]+e[1]*e[9]*e[21];
+        A[107]=e[10]*e[27]*e[12]+e[10]*e[9]*e[30]+e[10]*e[29]*e[14]+e[10]*e[11]*e[32]+e[10]*e[28]*e[13]+e[28]*e[11]*e[14]+e[28]*e[9]*e[12]+e[13]*e[30]*e[12]+e[13]*e[32]*e[14]+e[16]*e[30]*e[15]+e[16]*e[12]*e[33]+e[16]*e[32]*e[17]+e[16]*e[14]*e[35]+e[16]*e[13]*e[34]+e[34]*e[14]*e[17]+e[34]*e[12]*e[15]-1.*e[13]*e[27]*e[9]-1.*e[13]*e[33]*e[15]-1.*e[13]*e[35]*e[17]-1.*e[13]*e[29]*e[11]+.5000000000*ep2[10]*e[31]+.5000000000*e[31]*ep2[16]-.5000000000*e[31]*ep2[9]-.5000000000*e[31]*ep2[11]+.5000000000*e[31]*ep2[12]-.5000000000*e[31]*ep2[15]-.5000000000*e[31]*ep2[17]+.5000000000*e[31]*ep2[14]+1.500000000*e[31]*ep2[13];
+        A[106]=-.5000000000*e[4]*ep2[6]-.5000000000*e[4]*ep2[0]+e[1]*e[2]*e[5]+.5000000000*e[4]*ep2[7]+e[1]*e[0]*e[3]+e[7]*e[5]*e[8]-.5000000000*e[4]*ep2[8]+.5000000000*e[4]*ep2[3]+.5000000000*e[4]*ep2[5]+e[7]*e[3]*e[6]-.5000000000*e[4]*ep2[2]+.5000000000*ep3[4]+.5000000000*ep2[1]*e[4];
+        A[100]=e[34]*e[32]*e[35]-.5000000000*e[31]*ep2[35]+.5000000000*e[31]*ep2[34]+.5000000000*ep2[28]*e[31]+.5000000000*ep3[31]+.5000000000*e[31]*ep2[32]+e[34]*e[30]*e[33]-.5000000000*e[31]*ep2[27]+.5000000000*e[31]*ep2[30]-.5000000000*e[31]*ep2[33]-.5000000000*e[31]*ep2[29]+e[28]*e[29]*e[32]+e[28]*e[27]*e[30];
+        A[101]=e[1]*e[27]*e[30]+e[1]*e[29]*e[32]+e[1]*e[28]*e[31]+e[31]*e[30]*e[3]+e[31]*e[32]*e[5]+e[7]*e[30]*e[33]+e[7]*e[32]*e[35]+e[7]*e[31]*e[34]+e[28]*e[27]*e[3]+e[28]*e[0]*e[30]+e[28]*e[29]*e[5]+e[28]*e[2]*e[32]+e[34]*e[30]*e[6]+e[34]*e[3]*e[33]+e[34]*e[32]*e[8]+e[34]*e[5]*e[35]-1.*e[31]*e[27]*e[0]-1.*e[31]*e[33]*e[6]-1.*e[31]*e[35]*e[8]-1.*e[31]*e[29]*e[2]+.5000000000*e[4]*ep2[30]+.5000000000*e[4]*ep2[32]+1.500000000*e[4]*ep2[31]-.5000000000*e[4]*ep2[27]+.5000000000*e[4]*ep2[28]-.5000000000*e[4]*ep2[29]-.5000000000*e[4]*ep2[33]+.5000000000*e[4]*ep2[34]-.5000000000*e[4]*ep2[35];
+        A[102]=.5000000000*e[22]*ep2[30]+.5000000000*e[22]*ep2[32]+1.500000000*e[22]*ep2[31]+.5000000000*e[22]*ep2[34]-.5000000000*e[22]*ep2[27]-.5000000000*e[22]*ep2[29]-.5000000000*e[22]*ep2[33]-.5000000000*e[22]*ep2[35]+e[28]*e[18]*e[30]+e[28]*e[29]*e[23]+e[28]*e[20]*e[32]+e[31]*e[30]*e[21]+e[31]*e[32]*e[23]+e[25]*e[30]*e[33]+e[25]*e[32]*e[35]+e[25]*e[31]*e[34]+e[34]*e[30]*e[24]+e[34]*e[21]*e[33]+e[34]*e[32]*e[26]+e[34]*e[23]*e[35]-1.*e[31]*e[27]*e[18]-1.*e[31]*e[33]*e[24]-1.*e[31]*e[29]*e[20]-1.*e[31]*e[35]*e[26]+e[19]*e[27]*e[30]+e[19]*e[29]*e[32]+e[19]*e[28]*e[31]+e[28]*e[27]*e[21]+.5000000000*ep2[28]*e[22];
+        A[103]=e[16]*e[30]*e[33]+e[16]*e[32]*e[35]+e[10]*e[27]*e[30]+e[10]*e[29]*e[32]+e[10]*e[28]*e[31]+e[34]*e[30]*e[15]+e[34]*e[12]*e[33]+e[34]*e[32]*e[17]+e[34]*e[14]*e[35]+e[34]*e[31]*e[16]+e[28]*e[27]*e[12]+e[28]*e[9]*e[30]+e[28]*e[29]*e[14]+e[28]*e[11]*e[32]-1.*e[31]*e[27]*e[9]+e[31]*e[30]*e[12]+e[31]*e[32]*e[14]-1.*e[31]*e[33]*e[15]-1.*e[31]*e[35]*e[17]-1.*e[31]*e[29]*e[11]-.5000000000*e[13]*ep2[27]+.5000000000*e[13]*ep2[32]+.5000000000*e[13]*ep2[28]-.5000000000*e[13]*ep2[29]+1.500000000*e[13]*ep2[31]-.5000000000*e[13]*ep2[33]+.5000000000*e[13]*ep2[30]+.5000000000*e[13]*ep2[34]-.5000000000*e[13]*ep2[35];
+        A[96]=e[21]*e[23]*e[14]+e[21]*e[22]*e[13]+e[24]*e[21]*e[15]+e[24]*e[23]*e[17]+e[24]*e[14]*e[26]+e[24]*e[22]*e[16]+e[24]*e[13]*e[25]+e[15]*e[22]*e[25]+e[15]*e[23]*e[26]+e[9]*e[19]*e[22]+e[9]*e[18]*e[21]+e[9]*e[20]*e[23]+e[18]*e[20]*e[14]+e[18]*e[11]*e[23]+e[18]*e[19]*e[13]+e[18]*e[10]*e[22]-1.*e[21]*e[25]*e[16]-1.*e[21]*e[26]*e[17]-1.*e[21]*e[20]*e[11]-1.*e[21]*e[19]*e[10]+1.500000000*ep2[21]*e[12]+.5000000000*e[12]*ep2[24]-.5000000000*e[12]*ep2[26]+.5000000000*e[12]*ep2[18]+.5000000000*e[12]*ep2[23]-.5000000000*e[12]*ep2[19]-.5000000000*e[12]*ep2[20]+.5000000000*e[12]*ep2[22]-.5000000000*e[12]*ep2[25];
+        A[97]=-1.*e[12]*e[29]*e[20]-1.*e[12]*e[35]*e[26]-1.*e[12]*e[28]*e[19]-1.*e[12]*e[34]*e[25]+e[18]*e[29]*e[14]+e[18]*e[11]*e[32]+e[18]*e[28]*e[13]+e[18]*e[10]*e[31]+e[27]*e[20]*e[14]+e[27]*e[11]*e[23]+e[27]*e[19]*e[13]+e[27]*e[10]*e[22]+e[15]*e[30]*e[24]+e[15]*e[21]*e[33]+e[15]*e[31]*e[25]+e[15]*e[22]*e[34]+e[15]*e[32]*e[26]+e[15]*e[23]*e[35]-1.*e[21]*e[28]*e[10]-1.*e[21]*e[34]*e[16]-1.*e[21]*e[35]*e[17]-1.*e[21]*e[29]*e[11]-1.*e[30]*e[25]*e[16]-1.*e[30]*e[26]*e[17]-1.*e[30]*e[20]*e[11]-1.*e[30]*e[19]*e[10]+e[24]*e[32]*e[17]+e[24]*e[14]*e[35]+e[24]*e[31]*e[16]+e[24]*e[13]*e[34]+e[33]*e[23]*e[17]+e[33]*e[14]*e[26]+e[33]*e[22]*e[16]+e[33]*e[13]*e[25]+3.*e[12]*e[30]*e[21]+e[12]*e[31]*e[22]+e[12]*e[32]*e[23]+e[9]*e[27]*e[21]+e[9]*e[18]*e[30]+e[9]*e[28]*e[22]+e[9]*e[19]*e[31]+e[9]*e[29]*e[23]+e[9]*e[20]*e[32]+e[21]*e[32]*e[14]+e[21]*e[31]*e[13]+e[30]*e[23]*e[14]+e[30]*e[22]*e[13]+e[12]*e[27]*e[18]+e[12]*e[33]*e[24];
+        A[98]=e[0]*e[11]*e[5]+e[0]*e[2]*e[14]+e[9]*e[1]*e[4]+e[9]*e[0]*e[3]+e[9]*e[2]*e[5]+e[3]*e[13]*e[4]+e[3]*e[14]*e[5]+e[6]*e[3]*e[15]+e[6]*e[13]*e[7]+e[6]*e[4]*e[16]+e[6]*e[14]*e[8]+e[6]*e[5]*e[17]+e[15]*e[4]*e[7]+e[15]*e[5]*e[8]-1.*e[3]*e[11]*e[2]-1.*e[3]*e[10]*e[1]-1.*e[3]*e[16]*e[7]-1.*e[3]*e[17]*e[8]+e[0]*e[10]*e[4]+e[0]*e[1]*e[13]+1.500000000*e[12]*ep2[3]+.5000000000*e[12]*ep2[4]+.5000000000*e[12]*ep2[5]+.5000000000*e[12]*ep2[6]+.5000000000*ep2[0]*e[12]-.5000000000*e[12]*ep2[1]-.5000000000*e[12]*ep2[7]-.5000000000*e[12]*ep2[2]-.5000000000*e[12]*ep2[8];
+        A[99]=e[21]*e[24]*e[6]+e[0]*e[19]*e[22]+e[0]*e[20]*e[23]+e[24]*e[22]*e[7]+e[24]*e[4]*e[25]+e[24]*e[23]*e[8]+e[24]*e[5]*e[26]+e[6]*e[22]*e[25]+e[6]*e[23]*e[26]+e[18]*e[0]*e[21]+e[18]*e[19]*e[4]+e[18]*e[1]*e[22]+e[18]*e[20]*e[5]+e[18]*e[2]*e[23]+e[21]*e[22]*e[4]+e[21]*e[23]*e[5]-1.*e[21]*e[26]*e[8]-1.*e[21]*e[20]*e[2]-1.*e[21]*e[19]*e[1]-1.*e[21]*e[25]*e[7]+1.500000000*ep2[21]*e[3]+.5000000000*e[3]*ep2[22]+.5000000000*e[3]*ep2[23]+.5000000000*e[3]*ep2[24]-.5000000000*e[3]*ep2[26]-.5000000000*e[3]*ep2[19]-.5000000000*e[3]*ep2[20]-.5000000000*e[3]*ep2[25]+.5000000000*ep2[18]*e[3];
+        A[127]=e[11]*e[27]*e[12]+e[11]*e[9]*e[30]+e[11]*e[29]*e[14]+e[11]*e[28]*e[13]+e[11]*e[10]*e[31]+e[29]*e[9]*e[12]+e[29]*e[10]*e[13]+e[14]*e[30]*e[12]+e[14]*e[31]*e[13]+e[17]*e[30]*e[15]+e[17]*e[12]*e[33]+e[17]*e[14]*e[35]+e[17]*e[31]*e[16]+e[17]*e[13]*e[34]+e[35]*e[12]*e[15]+e[35]*e[13]*e[16]-1.*e[14]*e[27]*e[9]-1.*e[14]*e[28]*e[10]-1.*e[14]*e[33]*e[15]-1.*e[14]*e[34]*e[16]+.5000000000*ep2[11]*e[32]-.5000000000*e[32]*ep2[16]-.5000000000*e[32]*ep2[9]+.5000000000*e[32]*ep2[12]-.5000000000*e[32]*ep2[15]+.5000000000*e[32]*ep2[17]-.5000000000*e[32]*ep2[10]+1.500000000*e[32]*ep2[14]+.5000000000*e[32]*ep2[13];
+        A[126]=e[8]*e[3]*e[6]+.5000000000*ep2[2]*e[5]-.5000000000*e[5]*ep2[0]+.5000000000*e[5]*ep2[4]-.5000000000*e[5]*ep2[6]+.5000000000*e[5]*ep2[8]+e[8]*e[4]*e[7]+.5000000000*ep3[5]+e[2]*e[0]*e[3]+.5000000000*e[5]*ep2[3]-.5000000000*e[5]*ep2[7]+e[2]*e[1]*e[4]-.5000000000*e[5]*ep2[1];
+        A[125]=e[2]*e[27]*e[3]+e[2]*e[0]*e[30]+e[2]*e[28]*e[4]+e[2]*e[1]*e[31]+e[2]*e[29]*e[5]-1.*e[5]*e[27]*e[0]-1.*e[5]*e[34]*e[7]-1.*e[5]*e[33]*e[6]+e[5]*e[30]*e[3]+e[5]*e[35]*e[8]-1.*e[5]*e[28]*e[1]+e[5]*e[31]*e[4]+e[29]*e[1]*e[4]+e[29]*e[0]*e[3]+e[8]*e[30]*e[6]+e[8]*e[3]*e[33]+e[8]*e[31]*e[7]+e[8]*e[4]*e[34]+e[35]*e[4]*e[7]+e[35]*e[3]*e[6]+.5000000000*ep2[2]*e[32]+1.500000000*e[32]*ep2[5]+.5000000000*e[32]*ep2[4]-.5000000000*e[32]*ep2[0]-.5000000000*e[32]*ep2[6]-.5000000000*e[32]*ep2[1]-.5000000000*e[32]*ep2[7]+.5000000000*e[32]*ep2[3]+.5000000000*e[32]*ep2[8];
+        A[124]=-1.*e[14]*e[19]*e[1]+e[14]*e[22]*e[4]-1.*e[14]*e[18]*e[0]-1.*e[14]*e[25]*e[7]-1.*e[14]*e[24]*e[6]-1.*e[23]*e[10]*e[1]+e[23]*e[13]*e[4]-1.*e[23]*e[16]*e[7]-1.*e[23]*e[15]*e[6]-1.*e[23]*e[9]*e[0]+e[23]*e[12]*e[3]+e[17]*e[21]*e[6]+e[17]*e[3]*e[24]+e[17]*e[22]*e[7]+e[17]*e[4]*e[25]+e[17]*e[5]*e[26]-1.*e[5]*e[24]*e[15]-1.*e[5]*e[25]*e[16]-1.*e[5]*e[18]*e[9]-1.*e[5]*e[19]*e[10]+e[26]*e[12]*e[6]+e[26]*e[3]*e[15]+e[26]*e[13]*e[7]+e[26]*e[4]*e[16]+e[11]*e[18]*e[3]+e[11]*e[0]*e[21]+e[11]*e[19]*e[4]+e[11]*e[1]*e[22]+e[11]*e[20]*e[5]+e[11]*e[2]*e[23]+e[20]*e[9]*e[3]+e[20]*e[0]*e[12]+e[20]*e[10]*e[4]+e[20]*e[1]*e[13]+e[20]*e[2]*e[14]+e[5]*e[21]*e[12]+3.*e[5]*e[23]*e[14]+e[5]*e[22]*e[13]+e[8]*e[21]*e[15]+e[8]*e[12]*e[24]+e[8]*e[23]*e[17]+e[8]*e[14]*e[26]+e[8]*e[22]*e[16]+e[8]*e[13]*e[25]+e[2]*e[18]*e[12]+e[2]*e[9]*e[21]+e[2]*e[19]*e[13]+e[2]*e[10]*e[22]+e[14]*e[21]*e[3];
+        A[123]=-.5000000000*e[14]*ep2[27]+1.500000000*e[14]*ep2[32]-.5000000000*e[14]*ep2[28]+.5000000000*e[14]*ep2[29]+.5000000000*e[14]*ep2[31]-.5000000000*e[14]*ep2[33]+.5000000000*e[14]*ep2[30]-.5000000000*e[14]*ep2[34]+.5000000000*e[14]*ep2[35]+e[11]*e[27]*e[30]+e[11]*e[29]*e[32]+e[11]*e[28]*e[31]+e[35]*e[30]*e[15]+e[35]*e[12]*e[33]+e[35]*e[32]*e[17]+e[35]*e[31]*e[16]+e[35]*e[13]*e[34]+e[29]*e[27]*e[12]+e[29]*e[9]*e[30]+e[29]*e[28]*e[13]+e[29]*e[10]*e[31]-1.*e[32]*e[27]*e[9]+e[32]*e[30]*e[12]-1.*e[32]*e[28]*e[10]+e[32]*e[31]*e[13]-1.*e[32]*e[33]*e[15]-1.*e[32]*e[34]*e[16]+e[17]*e[30]*e[33]+e[17]*e[31]*e[34];
+        A[122]=-.5000000000*e[23]*ep2[33]-.5000000000*e[23]*ep2[34]+.5000000000*ep2[29]*e[23]+.5000000000*e[23]*ep2[30]+1.500000000*e[23]*ep2[32]+.5000000000*e[23]*ep2[31]+.5000000000*e[23]*ep2[35]-.5000000000*e[23]*ep2[27]-.5000000000*e[23]*ep2[28]+e[32]*e[30]*e[21]+e[32]*e[31]*e[22]+e[26]*e[30]*e[33]+e[26]*e[32]*e[35]+e[26]*e[31]*e[34]+e[35]*e[30]*e[24]+e[35]*e[21]*e[33]+e[35]*e[31]*e[25]+e[35]*e[22]*e[34]-1.*e[32]*e[27]*e[18]-1.*e[32]*e[33]*e[24]-1.*e[32]*e[28]*e[19]-1.*e[32]*e[34]*e[25]+e[20]*e[27]*e[30]+e[20]*e[29]*e[32]+e[20]*e[28]*e[31]+e[29]*e[27]*e[21]+e[29]*e[18]*e[30]+e[29]*e[28]*e[22]+e[29]*e[19]*e[31];
+        A[121]=e[2]*e[27]*e[30]+e[2]*e[29]*e[32]+e[2]*e[28]*e[31]+e[32]*e[30]*e[3]+e[32]*e[31]*e[4]+e[8]*e[30]*e[33]+e[8]*e[32]*e[35]+e[8]*e[31]*e[34]+e[29]*e[27]*e[3]+e[29]*e[0]*e[30]+e[29]*e[28]*e[4]+e[29]*e[1]*e[31]+e[35]*e[30]*e[6]+e[35]*e[3]*e[33]+e[35]*e[31]*e[7]+e[35]*e[4]*e[34]-1.*e[32]*e[27]*e[0]-1.*e[32]*e[34]*e[7]-1.*e[32]*e[33]*e[6]-1.*e[32]*e[28]*e[1]+.5000000000*e[5]*ep2[30]+1.500000000*e[5]*ep2[32]+.5000000000*e[5]*ep2[31]-.5000000000*e[5]*ep2[27]-.5000000000*e[5]*ep2[28]+.5000000000*e[5]*ep2[29]-.5000000000*e[5]*ep2[33]-.5000000000*e[5]*ep2[34]+.5000000000*e[5]*ep2[35];
+        A[120]=.5000000000*e[32]*ep2[31]+.5000000000*e[32]*ep2[35]-.5000000000*e[32]*ep2[27]+e[29]*e[27]*e[30]+e[29]*e[28]*e[31]+e[35]*e[30]*e[33]+e[35]*e[31]*e[34]+.5000000000*ep2[29]*e[32]+.5000000000*ep3[32]-.5000000000*e[32]*ep2[33]-.5000000000*e[32]*ep2[34]+.5000000000*e[32]*ep2[30]-.5000000000*e[32]*ep2[28];
+        A[118]=e[10]*e[1]*e[4]+e[10]*e[0]*e[3]+e[10]*e[2]*e[5]+e[4]*e[12]*e[3]+e[4]*e[14]*e[5]+e[7]*e[12]*e[6]+e[7]*e[3]*e[15]+e[7]*e[4]*e[16]+e[7]*e[14]*e[8]+e[7]*e[5]*e[17]+e[16]*e[3]*e[6]+e[16]*e[5]*e[8]-1.*e[4]*e[11]*e[2]-1.*e[4]*e[15]*e[6]-1.*e[4]*e[9]*e[0]-1.*e[4]*e[17]*e[8]+e[1]*e[9]*e[3]+e[1]*e[0]*e[12]+e[1]*e[11]*e[5]+e[1]*e[2]*e[14]+1.500000000*e[13]*ep2[4]+.5000000000*e[13]*ep2[3]+.5000000000*e[13]*ep2[5]+.5000000000*e[13]*ep2[7]+.5000000000*ep2[1]*e[13]-.5000000000*e[13]*ep2[0]-.5000000000*e[13]*ep2[6]-.5000000000*e[13]*ep2[2]-.5000000000*e[13]*ep2[8];
+        A[119]=e[25]*e[21]*e[6]+e[25]*e[3]*e[24]+e[25]*e[23]*e[8]+e[25]*e[5]*e[26]+e[7]*e[21]*e[24]+e[7]*e[23]*e[26]+e[19]*e[18]*e[3]+e[19]*e[0]*e[21]+e[19]*e[1]*e[22]+e[19]*e[20]*e[5]+e[19]*e[2]*e[23]+e[22]*e[21]*e[3]+e[22]*e[23]*e[5]-1.*e[22]*e[26]*e[8]-1.*e[22]*e[20]*e[2]-1.*e[22]*e[18]*e[0]+e[22]*e[25]*e[7]-1.*e[22]*e[24]*e[6]+e[1]*e[18]*e[21]+e[1]*e[20]*e[23]+.5000000000*e[4]*ep2[25]-.5000000000*e[4]*ep2[26]-.5000000000*e[4]*ep2[18]-.5000000000*e[4]*ep2[20]-.5000000000*e[4]*ep2[24]+.5000000000*ep2[19]*e[4]+1.500000000*ep2[22]*e[4]+.5000000000*e[4]*ep2[21]+.5000000000*e[4]*ep2[23];
+        A[116]=e[22]*e[21]*e[12]+e[22]*e[23]*e[14]+e[25]*e[21]*e[15]+e[25]*e[12]*e[24]+e[25]*e[23]*e[17]+e[25]*e[14]*e[26]+e[25]*e[22]*e[16]+e[16]*e[21]*e[24]+e[16]*e[23]*e[26]+e[10]*e[19]*e[22]+e[10]*e[18]*e[21]+e[10]*e[20]*e[23]+e[19]*e[18]*e[12]+e[19]*e[9]*e[21]+e[19]*e[20]*e[14]+e[19]*e[11]*e[23]-1.*e[22]*e[24]*e[15]-1.*e[22]*e[26]*e[17]-1.*e[22]*e[20]*e[11]-1.*e[22]*e[18]*e[9]-.5000000000*e[13]*ep2[26]-.5000000000*e[13]*ep2[18]+.5000000000*e[13]*ep2[23]+.5000000000*e[13]*ep2[19]-.5000000000*e[13]*ep2[20]-.5000000000*e[13]*ep2[24]+.5000000000*e[13]*ep2[21]+1.500000000*ep2[22]*e[13]+.5000000000*e[13]*ep2[25];
+        A[117]=e[13]*e[30]*e[21]+3.*e[13]*e[31]*e[22]+e[13]*e[32]*e[23]+e[10]*e[27]*e[21]+e[10]*e[18]*e[30]+e[10]*e[28]*e[22]+e[10]*e[19]*e[31]+e[10]*e[29]*e[23]+e[10]*e[20]*e[32]+e[22]*e[30]*e[12]+e[22]*e[32]*e[14]+e[31]*e[21]*e[12]+e[31]*e[23]*e[14]-1.*e[13]*e[27]*e[18]-1.*e[13]*e[33]*e[24]-1.*e[13]*e[29]*e[20]-1.*e[13]*e[35]*e[26]+e[13]*e[28]*e[19]+e[13]*e[34]*e[25]+e[19]*e[27]*e[12]+e[19]*e[9]*e[30]+e[19]*e[29]*e[14]+e[19]*e[11]*e[32]+e[28]*e[18]*e[12]+e[28]*e[9]*e[21]+e[28]*e[20]*e[14]+e[28]*e[11]*e[23]+e[16]*e[30]*e[24]+e[16]*e[21]*e[33]+e[16]*e[31]*e[25]+e[16]*e[22]*e[34]+e[16]*e[32]*e[26]+e[16]*e[23]*e[35]-1.*e[22]*e[27]*e[9]-1.*e[22]*e[33]*e[15]-1.*e[22]*e[35]*e[17]-1.*e[22]*e[29]*e[11]-1.*e[31]*e[24]*e[15]-1.*e[31]*e[26]*e[17]-1.*e[31]*e[20]*e[11]-1.*e[31]*e[18]*e[9]+e[25]*e[30]*e[15]+e[25]*e[12]*e[33]+e[25]*e[32]*e[17]+e[25]*e[14]*e[35]+e[34]*e[21]*e[15]+e[34]*e[12]*e[24]+e[34]*e[23]*e[17]+e[34]*e[14]*e[26];
+        A[114]=e[19]*e[11]*e[14]+e[19]*e[9]*e[12]+e[19]*e[10]*e[13]+e[13]*e[21]*e[12]+e[13]*e[23]*e[14]+e[16]*e[21]*e[15]+e[16]*e[12]*e[24]+e[16]*e[23]*e[17]+e[16]*e[14]*e[26]+e[16]*e[13]*e[25]+e[25]*e[14]*e[17]+e[25]*e[12]*e[15]-1.*e[13]*e[24]*e[15]-1.*e[13]*e[26]*e[17]-1.*e[13]*e[20]*e[11]-1.*e[13]*e[18]*e[9]+e[10]*e[18]*e[12]+e[10]*e[9]*e[21]+e[10]*e[20]*e[14]+e[10]*e[11]*e[23]+1.500000000*e[22]*ep2[13]+.5000000000*e[22]*ep2[14]+.5000000000*e[22]*ep2[12]+.5000000000*e[22]*ep2[16]+.5000000000*ep2[10]*e[22]-.5000000000*e[22]*ep2[9]-.5000000000*e[22]*ep2[11]-.5000000000*e[22]*ep2[15]-.5000000000*e[22]*ep2[17];
+        A[115]=e[13]*e[12]*e[3]+e[13]*e[14]*e[5]+e[16]*e[12]*e[6]+e[16]*e[3]*e[15]+e[16]*e[13]*e[7]+e[16]*e[14]*e[8]+e[16]*e[5]*e[17]+e[7]*e[14]*e[17]+e[7]*e[12]*e[15]+e[1]*e[11]*e[14]+e[1]*e[9]*e[12]+e[1]*e[10]*e[13]+e[10]*e[9]*e[3]+e[10]*e[0]*e[12]+e[10]*e[11]*e[5]+e[10]*e[2]*e[14]-1.*e[13]*e[11]*e[2]-1.*e[13]*e[15]*e[6]-1.*e[13]*e[9]*e[0]-1.*e[13]*e[17]*e[8]+1.500000000*ep2[13]*e[4]+.5000000000*e[4]*ep2[16]-.5000000000*e[4]*ep2[9]-.5000000000*e[4]*ep2[11]+.5000000000*e[4]*ep2[12]-.5000000000*e[4]*ep2[15]-.5000000000*e[4]*ep2[17]+.5000000000*e[4]*ep2[10]+.5000000000*e[4]*ep2[14];
+        A[112]=e[19]*e[1]*e[4]+e[19]*e[0]*e[3]+e[19]*e[2]*e[5]+e[4]*e[21]*e[3]+e[4]*e[23]*e[5]+e[7]*e[21]*e[6]+e[7]*e[3]*e[24]+e[7]*e[4]*e[25]+e[7]*e[23]*e[8]+e[7]*e[5]*e[26]+e[25]*e[3]*e[6]+e[25]*e[5]*e[8]+e[1]*e[18]*e[3]+e[1]*e[0]*e[21]+e[1]*e[20]*e[5]+e[1]*e[2]*e[23]-1.*e[4]*e[26]*e[8]-1.*e[4]*e[20]*e[2]-1.*e[4]*e[18]*e[0]-1.*e[4]*e[24]*e[6]+1.500000000*e[22]*ep2[4]-.5000000000*e[22]*ep2[0]-.5000000000*e[22]*ep2[6]+.5000000000*e[22]*ep2[5]+.5000000000*e[22]*ep2[1]+.5000000000*e[22]*ep2[7]+.5000000000*e[22]*ep2[3]-.5000000000*e[22]*ep2[2]-.5000000000*e[22]*ep2[8];
+        A[113]=-1.*e[31]*e[20]*e[2]-1.*e[31]*e[18]*e[0]+e[31]*e[23]*e[5]-1.*e[31]*e[24]*e[6]+e[7]*e[30]*e[24]+e[7]*e[21]*e[33]+e[7]*e[32]*e[26]+e[7]*e[23]*e[35]+e[25]*e[30]*e[6]+e[25]*e[3]*e[33]+e[25]*e[31]*e[7]+e[25]*e[4]*e[34]+e[25]*e[32]*e[8]+e[25]*e[5]*e[35]+e[34]*e[21]*e[6]+e[34]*e[3]*e[24]+e[34]*e[22]*e[7]+e[34]*e[23]*e[8]+e[34]*e[5]*e[26]+e[1]*e[27]*e[21]+e[1]*e[18]*e[30]+e[1]*e[28]*e[22]+e[1]*e[19]*e[31]+e[1]*e[29]*e[23]+e[1]*e[20]*e[32]+e[19]*e[27]*e[3]+e[19]*e[0]*e[30]+e[19]*e[28]*e[4]+e[19]*e[29]*e[5]+e[19]*e[2]*e[32]+e[28]*e[18]*e[3]+e[28]*e[0]*e[21]+e[28]*e[20]*e[5]+e[28]*e[2]*e[23]+e[4]*e[30]*e[21]+3.*e[4]*e[31]*e[22]+e[4]*e[32]*e[23]-1.*e[4]*e[27]*e[18]-1.*e[4]*e[33]*e[24]-1.*e[4]*e[29]*e[20]-1.*e[4]*e[35]*e[26]-1.*e[22]*e[27]*e[0]+e[22]*e[32]*e[5]-1.*e[22]*e[33]*e[6]+e[22]*e[30]*e[3]-1.*e[22]*e[35]*e[8]-1.*e[22]*e[29]*e[2]+e[31]*e[21]*e[3]-1.*e[31]*e[26]*e[8];
 
-        double e003,e013,e023,e033,e043,e053,e063,e073,e083;
-        double e103,e113,e123,e133,e143,e153,e163,e173,e183;
-        double e203,e213,e223,e233,e243,e253,e263,e273,e283;
-        double e303,e313,e323,e333,e343,e353,e363,e373,e383;
-        e00 = EE[0*9 + 0 ];
-        e10 = EE[1*9 + 0 ];
-        e20 = EE[2*9 + 0 ];
-        e30 = EE[3*9 + 0 ];
-        e01 = EE[0*9 + 1 ];
-        e11 = EE[1*9 + 1 ];
-        e21 = EE[2*9 + 1 ];
-        e31 = EE[3*9 + 1 ];
-        e02 = EE[0*9 + 2 ];
-        e12 = EE[1*9 + 2 ];
-        e22 = EE[2*9 + 2 ];
-        e32 = EE[3*9 + 2 ];
-        e03 = EE[0*9 + 3 ];
-        e13 = EE[1*9 + 3 ];
-        e23 = EE[2*9 + 3 ];
-        e33 = EE[3*9 + 3 ];
-        e04 = EE[0*9 + 4 ];
-        e14 = EE[1*9 + 4 ];
-        e24 = EE[2*9 + 4 ];
-        e34 = EE[3*9 + 4 ];
-        e05 = EE[0*9 + 5 ];
-        e15 = EE[1*9 + 5 ];
-        e25 = EE[2*9 + 5 ];
-        e35 = EE[3*9 + 5 ];
-        e06 = EE[0*9 + 6 ];
-        e16 = EE[1*9 + 6 ];
-        e26 = EE[2*9 + 6 ];
-        e36 = EE[3*9 + 6 ];
-        e07 = EE[0*9 + 7 ];
-        e17 = EE[1*9 + 7 ];
-        e27 = EE[2*9 + 7 ];
-        e37 = EE[3*9 + 7 ];
-        e08 = EE[0*9 + 8 ];
-        e18 = EE[1*9 + 8 ];
-        e28 = EE[2*9 + 8 ];
-        e38 = EE[3*9 + 8 ];
+        int perm[20] = {6, 8, 18, 15, 12, 5, 14, 7, 4, 11, 19, 13, 1, 16, 17, 3, 10, 9, 2, 0};
+        double AA[200];
+        for (int i = 0; i < 20; i++)
+        {
+            for (int j = 0; j < 10; j++) AA[i + j * 20] = A[perm[i] + j * 20];
+        }
 
-
-        e002 =e00*e00;
-        e102 =e10*e10;
-        e202 =e20*e20;
-        e302 =e30*e30;
-        e012 =e01*e01;
-        e112 =e11*e11;
-        e212 =e21*e21;
-        e312 =e31*e31;
-        e022 =e02*e02;
-        e122 =e12*e12;
-        e222 =e22*e22;
-        e322 =e32*e32;
-        e032 =e03*e03;
-        e132 =e13*e13;
-        e232 =e23*e23;
-        e332 =e33*e33;
-        e042 =e04*e04;
-        e142 =e14*e14;
-        e242 =e24*e24;
-        e342 =e34*e34;
-        e052 =e05*e05;
-        e152 =e15*e15;
-        e252 =e25*e25;
-        e352 =e35*e35;
-        e062 =e06*e06;
-        e162 =e16*e16;
-        e262 =e26*e26;
-        e362 =e36*e36;
-        e072 =e07*e07;
-        e172 =e17*e17;
-        e272 =e27*e27;
-        e372 =e37*e37;
-        e082 =e08*e08;
-        e182 =e18*e18;
-        e282 =e28*e28;
-        e382 =e38*e38;
-
-        e003 =e00*e00*e00;
-        e103 =e10*e10*e10;
-        e203 =e20*e20*e20;
-        e303 =e30*e30*e30;
-        e013 =e01*e01*e01;
-        e113 =e11*e11*e11;
-        e213 =e21*e21*e21;
-        e313 =e31*e31*e31;
-        e023 =e02*e02*e02;
-        e123 =e12*e12*e12;
-        e223 =e22*e22*e22;
-        e323 =e32*e32*e32;
-        e033 =e03*e03*e03;
-        e133 =e13*e13*e13;
-        e233 =e23*e23*e23;
-        e333 =e33*e33*e33;
-        e043 =e04*e04*e04;
-        e143 =e14*e14*e14;
-        e243 =e24*e24*e24;
-        e343 =e34*e34*e34;
-        e053 =e05*e05*e05;
-        e153 =e15*e15*e15;
-        e253 =e25*e25*e25;
-        e353 =e35*e35*e35;
-        e063 =e06*e06*e06;
-        e163 =e16*e16*e16;
-        e263 =e26*e26*e26;
-        e363 =e36*e36*e36;
-        e073 =e07*e07*e07;
-        e173 =e17*e17*e17;
-        e273 =e27*e27*e27;
-        e373 =e37*e37*e37;
-        e083 =e08*e08*e08;
-        e183 =e18*e18*e18;
-        e283 =e28*e28*e28;
-        e383 =e38*e38*e38;
-
-
-        A[0 + 10*0]=0.5*e003+0.5*e00*e012+0.5*e00*e022+0.5*e00*e032+e03*e01*e04+e03*e02*e05+0.5*e00*e062+e06*e01*e07+e06*e02*e08-0.5*e00*e042-0.5*e00*e052-0.5*e00*e072-0.5*e00*e082;
-        A[0 + 10*1]=e00*e11*e01+e00*e12*e02+e03*e00*e13+e03*e11*e04+e03*e01*e14+e03*e12*e05+e03*e02*e15+e13*e01*e04+e13*e02*e05+e06*e00*e16+1.5*e10*e002+0.5*e10*e012+0.5*e10*e022+0.5*e10*e062-0.5*e10*e042-0.5*e10*e052-0.5*e10*e072+0.5*e10*e032+e06*e11*e07+e06*e01*e17+e06*e12*e08+e06*e02*e18+e16*e01*e07+e16*e02*e08-e00*e14*e04-e00*e17*e07-e00*e15*e05-e00*e18*e08-0.5*e10*e082;
-        A[0 + 10*2]=e16*e02*e18+e03*e12*e15+e10*e11*e01+e10*e12*e02+e03*e10*e13+e03*e11*e14+e13*e11*e04+e13*e01*e14+e13*e12*e05+e13*e02*e15+e06*e10*e16+e06*e12*e18+e06*e11*e17+e16*e11*e07+e16*e01*e17+e16*e12*e08-e10*e14*e04-e10*e17*e07-e10*e15*e05-e10*e18*e08+1.5*e00*e102+0.5*e00*e122+0.5*e00*e112+0.5*e00*e132+0.5*e00*e162-0.5*e00*e152-0.5*e00*e172-0.5*e00*e182-0.5*e00*e142;
-        A[0 + 10*3]=0.5*e103+0.5*e10*e122+0.5*e10*e112+0.5*e10*e132+e13*e12*e15+e13*e11*e14+0.5*e10*e162+e16*e12*e18+e16*e11*e17-0.5*e10*e152-0.5*e10*e172-0.5*e10*e182-0.5*e10*e142;
-        A[0 + 10*4]=-e00*e28*e08-e00*e25*e05-e00*e27*e07-e00*e24*e04+e26*e02*e08+e26*e01*e07+e06*e02*e28+e06*e22*e08+e06*e01*e27+e06*e21*e07+e23*e02*e05+e23*e01*e04+e03*e02*e25+e03*e22*e05+e03*e01*e24+e03*e21*e04+e00*e22*e02+e00*e21*e01-0.5*e20*e082-0.5*e20*e052-0.5*e20*e072-0.5*e20*e042+e06*e00*e26+0.5*e20*e062+e03*e00*e23+0.5*e20*e022+1.5*e20*e002+0.5*e20*e032+0.5*e20*e012;
-        A[0 + 10*5]=-e10*e24*e04-e10*e27*e07-e10*e25*e05-e10*e28*e08-e20*e14*e04-e20*e17*e07-e20*e15*e05-e20*e18*e08-e00*e24*e14-e00*e25*e15-e00*e27*e17-e00*e28*e18+e06*e21*e17+e06*e22*e18+e06*e12*e28+e16*e00*e26+e16*e21*e07+e16*e01*e27+e16*e22*e08+e16*e02*e28+e26*e11*e07+e26*e01*e17+e26*e12*e08+e26*e02*e18+e06*e11*e27+e23*e11*e04+e23*e01*e14+e23*e12*e05+e23*e02*e15+e06*e20*e16+e06*e10*e26+e03*e21*e14+e03*e22*e15+e03*e12*e25+e13*e00*e23+e13*e21*e04+e13*e01*e24+e13*e22*e05+e13*e02*e25+e03*e11*e24+e03*e20*e13+e03*e10*e23+e00*e21*e11+3*e00*e20*e10+e00*e22*e12+e20*e12*e02+e20*e11*e01+e10*e22*e02+e10*e21*e01;
-        A[0 + 10*6]=-0.5*e20*e152+e26*e11*e17-e10*e24*e14-e10*e25*e15-e10*e27*e17-e10*e28*e18+0.5*e20*e162+e13*e10*e23+e13*e22*e15+e23*e12*e15+e23*e11*e14+e16*e10*e26+e16*e21*e17+e16*e11*e27+e16*e22*e18+e16*e12*e28+e26*e12*e18+e13*e12*e25+0.5*e20*e132+1.5*e20*e102+0.5*e20*e122+0.5*e20*e112+e10*e21*e11+e10*e22*e12+e13*e11*e24-0.5*e20*e172-0.5*e20*e182-0.5*e20*e142+e13*e21*e14;
-        A[0 + 10*7]=-e20*e25*e05-e20*e28*e08-0.5*e00*e272-0.5*e00*e282-0.5*e00*e242+0.5*e00*e262-0.5*e00*e252+e06*e20*e26+0.5*e00*e232+e06*e22*e28+e06*e21*e27+e26*e21*e07+e26*e01*e27+e26*e22*e08+e26*e02*e28-e20*e24*e04-e20*e27*e07+e03*e20*e23+e03*e22*e25+e03*e21*e24+e23*e21*e04+e23*e01*e24+e23*e22*e05+e23*e02*e25+e20*e21*e01+e20*e22*e02+1.5*e00*e202+0.5*e00*e222+0.5*e00*e212;
-        A[0 + 10*8]=e23*e21*e14+e23*e11*e24+e23*e22*e15+e23*e12*e25+e16*e20*e26+e16*e22*e28+e16*e21*e27+e26*e21*e17+e26*e11*e27+e26*e22*e18+e26*e12*e28+1.5*e10*e202+0.5*e10*e222+0.5*e10*e212+0.5*e10*e232+e20*e21*e11+e20*e22*e12+e13*e20*e23+e13*e22*e25+e13*e21*e24-e20*e24*e14-e20*e25*e15-e20*e27*e17-e20*e28*e18-0.5*e10*e272-0.5*e10*e282-0.5*e10*e242-0.5*e10*e252+0.5*e10*e262;
-        A[0 + 10*9]=0.5*e203+0.5*e20*e222+0.5*e20*e212+0.5*e20*e232+e23*e22*e25+e23*e21*e24+0.5*e20*e262+e26*e22*e28+e26*e21*e27-0.5*e20*e252-0.5*e20*e272-0.5*e20*e282-0.5*e20*e242;
-        A[0 + 10*10]=e06*e32*e08-0.5*e30*e082-0.5*e30*e042-0.5*e30*e052-0.5*e30*e072+0.5*e30*e012+0.5*e30*e022+0.5*e30*e032+0.5*e30*e062+1.5*e30*e002+e00*e31*e01+e00*e32*e02+e03*e31*e04+e03*e01*e34+e03*e32*e05+e03*e02*e35+e33*e01*e04+e33*e02*e05+e06*e00*e36+e06*e31*e07+e06*e01*e37+e06*e02*e38+e36*e01*e07+e36*e02*e08-e00*e34*e04-e00*e37*e07-e00*e35*e05-e00*e38*e08+e03*e00*e33;
-        A[0 + 10*11]=e06*e30*e16+e03*e30*e13+e16*e31*e07+e06*e10*e36-e10*e37*e07+3*e00*e30*e10+e00*e32*e12-e00*e38*e18-e10*e34*e04-e10*e35*e05-e10*e38*e08-e30*e14*e04-e30*e17*e07-e30*e15*e05-e30*e18*e08+e00*e31*e11+e10*e31*e01+e10*e32*e02+e30*e11*e01+e30*e12*e02+e03*e10*e33-e00*e34*e14-e00*e35*e15-e00*e37*e17+e03*e31*e14+e03*e11*e34+e03*e32*e15+e03*e12*e35+e13*e00*e33+e13*e31*e04+e13*e01*e34+e13*e32*e05+e13*e02*e35+e33*e11*e04+e33*e01*e14+e33*e12*e05+e33*e02*e15+e06*e31*e17+e06*e11*e37+e06*e32*e18+e06*e12*e38+e16*e00*e36+e16*e01*e37+e16*e32*e08+e16*e02*e38+e36*e11*e07+e36*e01*e17+e36*e12*e08+e36*e02*e18;
-        A[0 + 10*12]=e13*e10*e33+e33*e11*e14+e16*e10*e36+e16*e31*e17+e16*e11*e37+e16*e32*e18+e16*e12*e38+e36*e12*e18+e36*e11*e17-e10*e34*e14-e10*e35*e15-e10*e37*e17-e10*e38*e18+e10*e31*e11+e10*e32*e12+e13*e31*e14+e13*e11*e34+e13*e32*e15+e13*e12*e35+e33*e12*e15+1.5*e30*e102+0.5*e30*e122+0.5*e30*e112+0.5*e30*e132+0.5*e30*e162-0.5*e30*e152-0.5*e30*e172-0.5*e30*e182-0.5*e30*e142;
-        A[0 + 10*13]=e00*e32*e22+3*e00*e30*e20+e00*e31*e21+e20*e31*e01+e20*e32*e02+e30*e21*e01+e30*e22*e02+e03*e20*e33+e03*e32*e25+e03*e22*e35+e03*e31*e24+e03*e21*e34+e23*e00*e33+e23*e31*e04+e23*e01*e34+e23*e32*e05+e23*e02*e35+e33*e21*e04+e33*e01*e24+e33*e22*e05+e33*e02*e25+e06*e30*e26+e06*e20*e36+e06*e32*e28+e06*e22*e38+e06*e31*e27+e06*e21*e37+e26*e00*e36+e26*e31*e07+e03*e30*e23+e26*e01*e37+e26*e32*e08+e26*e02*e38+e36*e21*e07+e36*e01*e27+e36*e22*e08+e36*e02*e28-e00*e35*e25-e00*e37*e27-e00*e38*e28-e00*e34*e24-e20*e34*e04-e20*e37*e07-e20*e35*e05-e20*e38*e08-e30*e24*e04-e30*e27*e07-e30*e25*e05-e30*e28*e08;
-        A[0 + 10*14]=e16*e30*e26+e13*e21*e34+3*e10*e30*e20+e10*e32*e22+e10*e31*e21+e20*e31*e11+e20*e32*e12+e30*e21*e11+e30*e22*e12+e13*e30*e23+e13*e20*e33+e13*e32*e25+e13*e22*e35+e13*e31*e24+e23*e10*e33+e23*e31*e14+e23*e11*e34+e23*e32*e15+e23*e12*e35+e33*e21*e14+e33*e11*e24+e33*e22*e15+e33*e12*e25+e16*e20*e36+e16*e32*e28+e16*e22*e38+e16*e31*e27+e16*e21*e37+e26*e10*e36+e26*e31*e17+e26*e11*e37+e26*e32*e18+e26*e12*e38+e36*e21*e17+e36*e11*e27+e36*e22*e18+e36*e12*e28-e10*e35*e25-e10*e37*e27-e10*e38*e28-e10*e34*e24-e20*e34*e14-e20*e35*e15-e20*e37*e17-e20*e38*e18-e30*e24*e14-e30*e25*e15-e30*e27*e17-e30*e28*e18;
-        A[0 + 10*15]=-e20*e34*e24+0.5*e30*e262-0.5*e30*e252-0.5*e30*e272-0.5*e30*e282-0.5*e30*e242+1.5*e30*e202+0.5*e30*e222+0.5*e30*e212+0.5*e30*e232+e20*e32*e22+e20*e31*e21+e23*e20*e33+e23*e32*e25+e23*e22*e35+e23*e31*e24+e23*e21*e34+e33*e22*e25+e33*e21*e24+e26*e20*e36+e26*e32*e28+e26*e22*e38+e26*e31*e27+e26*e21*e37+e36*e22*e28+e36*e21*e27-e20*e35*e25-e20*e37*e27-e20*e38*e28;
-        A[0 + 10*16]=0.5*e00*e322+e30*e32*e02+e30*e31*e01+1.5*e00*e302+0.5*e00*e312+e03*e32*e35+e33*e31*e04+e33*e01*e34+e33*e32*e05+e33*e02*e35+e06*e30*e36+e06*e31*e37+e06*e32*e38+e36*e31*e07+e36*e01*e37+e36*e32*e08+e36*e02*e38-e30*e34*e04-e30*e37*e07-e30*e35*e05-e30*e38*e08+0.5*e00*e332+0.5*e00*e362-0.5*e00*e382-0.5*e00*e352-0.5*e00*e342-0.5*e00*e372+e03*e30*e33+e03*e31*e34;
-        A[0 + 10*17]=0.5*e10*e362-0.5*e10*e382-0.5*e10*e352-0.5*e10*e342-0.5*e10*e372+e36*e31*e17+e36*e11*e37+e36*e32*e18+e36*e12*e38-e30*e34*e14-e30*e35*e15-e30*e37*e17-e30*e38*e18+1.5*e10*e302+0.5*e10*e312+0.5*e10*e322+0.5*e10*e332+e30*e31*e11+e30*e32*e12+e13*e30*e33+e13*e31*e34+e13*e32*e35+e33*e31*e14+e33*e11*e34+e33*e32*e15+e33*e12*e35+e16*e30*e36+e16*e31*e37+e16*e32*e38;
-        A[0 + 10*18]=e33*e31*e24+e33*e21*e34+e26*e30*e36+e26*e31*e37+e26*e32*e38+e36*e32*e28+e36*e22*e38+e36*e31*e27+e36*e21*e37-e30*e35*e25-e30*e37*e27-e30*e38*e28-e30*e34*e24+e33*e22*e35+1.5*e20*e302+0.5*e20*e312+0.5*e20*e322+0.5*e20*e332+0.5*e20*e362-0.5*e20*e382-0.5*e20*e352-0.5*e20*e342-0.5*e20*e372+e30*e32*e22+e30*e31*e21+e23*e30*e33+e23*e31*e34+e23*e32*e35+e33*e32*e25;
-        A[0 + 10*19]=0.5*e303+0.5*e30*e312+0.5*e30*e322+0.5*e30*e332+e33*e31*e34+e33*e32*e35+0.5*e30*e362+e36*e31*e37+e36*e32*e38-0.5*e30*e382-0.5*e30*e352-0.5*e30*e342-0.5*e30*e372;
-        A[1 + 10*0]=e00*e01*e04+0.5*e002*e03+e00*e02*e05+0.5*e033+0.5*e03*e042+0.5*e03*e052+0.5*e03*e062+e06*e04*e07+e06*e05*e08-0.5*e03*e012-0.5*e03*e072-0.5*e03*e022-0.5*e03*e082;
-        A[1 + 10*1]=e03*e14*e04+e10*e01*e04+e16*e05*e08+e00*e10*e03+e00*e11*e04+e00*e01*e14+e00*e12*e05+e00*e02*e15+e10*e02*e05+e03*e15*e05+e06*e03*e16+e06*e14*e07+e06*e04*e17+e06*e15*e08+e06*e05*e18+0.5*e002*e13+1.5*e13*e032+0.5*e13*e042+0.5*e13*e052+0.5*e13*e062-0.5*e13*e012-0.5*e13*e072-0.5*e13*e022-0.5*e13*e082+e16*e04*e07-e03*e12*e02-e03*e11*e01-e03*e17*e07-e03*e18*e08;
-        A[1 + 10*2]=-e13*e11*e01+e00*e10*e13+e00*e12*e15+e00*e11*e14+e10*e11*e04+e10*e01*e14+e10*e12*e05+e10*e02*e15+e13*e14*e04+e13*e15*e05+e06*e13*e16+e06*e15*e18+e06*e14*e17+e16*e14*e07+e16*e04*e17+e16*e15*e08+e16*e05*e18-e13*e12*e02-e13*e17*e07-e13*e18*e08+0.5*e102*e03+1.5*e03*e132+0.5*e03*e152+0.5*e03*e142+0.5*e03*e162-0.5*e03*e112-0.5*e03*e172-0.5*e03*e122-0.5*e03*e182;
-        A[1 + 10*3]=0.5*e102*e13+e10*e11*e14+e10*e12*e15+0.5*e133+0.5*e13*e152+0.5*e13*e142+0.5*e13*e162+e16*e15*e18+e16*e14*e17-0.5*e13*e112-0.5*e13*e122-0.5*e13*e172-0.5*e13*e182;
-        A[1 + 10*4]=-e03*e28*e08-e03*e27*e07-e03*e21*e01-e03*e22*e02+e26*e05*e08+e26*e04*e07+e06*e05*e28+e06*e25*e08+e06*e04*e27+e06*e24*e07+e03*e25*e05+e03*e24*e04+e20*e02*e05+e20*e01*e04+e00*e02*e25+e00*e22*e05+e00*e01*e24+e00*e21*e04+e00*e20*e03-0.5*e23*e072-0.5*e23*e082-0.5*e23*e022-0.5*e23*e012+e06*e03*e26+0.5*e23*e052+0.5*e23*e062+1.5*e23*e032+0.5*e23*e042+0.5*e002*e23;
-        A[1 + 10*5]=e00*e21*e14+e00*e11*e24+e00*e10*e23+e00*e22*e15+e00*e12*e25+e20*e12*e05+e20*e01*e14+e20*e11*e04+e00*e20*e13+e10*e02*e25+e10*e22*e05+e10*e01*e24+e10*e21*e04+e10*e20*e03+e23*e15*e05+e23*e14*e04+e13*e25*e05+e13*e24*e04+e03*e24*e14+e03*e25*e15+3*e03*e23*e13+e20*e02*e15+e16*e03*e26+e06*e14*e27-e23*e18*e08+e06*e24*e17+e06*e15*e28+e06*e25*e18+e06*e13*e26+e06*e23*e16+e26*e04*e17+e26*e14*e07+e16*e05*e28+e16*e25*e08+e16*e04*e27+e16*e24*e07-e03*e22*e12-e03*e21*e11+e26*e05*e18+e26*e15*e08-e03*e27*e17-e03*e28*e18-e13*e22*e02-e13*e28*e08-e13*e27*e07-e13*e21*e01-e23*e17*e07-e23*e11*e01-e23*e12*e02;
-        A[1 + 10*6]=-0.5*e23*e182-0.5*e23*e172-0.5*e23*e112-0.5*e23*e122-e13*e22*e12-e13*e27*e17-e13*e28*e18+e26*e15*e18+e26*e14*e17-e13*e21*e11+e20*e12*e15+e13*e25*e15+e13*e24*e14+e16*e13*e26+e16*e25*e18+e16*e15*e28+e16*e24*e17+e16*e14*e27+1.5*e23*e132+0.5*e23*e152+0.5*e23*e142+0.5*e23*e162+e10*e20*e13+e10*e21*e14+e10*e11*e24+e10*e22*e15+e10*e12*e25+e20*e11*e14+0.5*e102*e23;
-        A[1 + 10*7]=e26*e04*e27+e00*e22*e25-e23*e28*e08+0.5*e03*e262-0.5*e03*e212-0.5*e03*e272-0.5*e03*e222-0.5*e03*e282+e23*e24*e04+e23*e25*e05+0.5*e202*e03+e06*e23*e26+e06*e24*e27+e06*e25*e28+e26*e24*e07+e26*e25*e08+e26*e05*e28-e23*e22*e02-e23*e21*e01-e23*e27*e07+e00*e20*e23+e00*e21*e24+e20*e21*e04+e20*e01*e24+e20*e22*e05+e20*e02*e25+1.5*e03*e232+0.5*e03*e242+0.5*e03*e252;
-        A[1 + 10*8]=e20*e11*e24-0.5*e13*e212-0.5*e13*e272-0.5*e13*e222-0.5*e13*e282-e23*e27*e17-e23*e28*e18+e26*e25*e18+e26*e24*e17+e26*e14*e27-e23*e21*e11-e23*e22*e12+e26*e15*e28+e23*e25*e15+e23*e24*e14+e16*e23*e26+e16*e24*e27+e16*e25*e28+0.5*e13*e262+e20*e21*e14+e20*e22*e15+e20*e12*e25+0.5*e13*e242+0.5*e13*e252+0.5*e202*e13+1.5*e13*e232+e10*e20*e23+e10*e22*e25+e10*e21*e24;
-        A[1 + 10*9]=0.5*e202*e23+e20*e22*e25+e20*e21*e24+0.5*e233+0.5*e23*e242+0.5*e23*e252+0.5*e23*e262+e26*e24*e27+e26*e25*e28-0.5*e23*e212-0.5*e23*e272-0.5*e23*e222-0.5*e23*e282;
-        A[1 + 10*10]=e00*e30*e03+0.5*e33*e062-0.5*e33*e012-0.5*e33*e022-0.5*e33*e072+e03*e35*e05+e06*e03*e36+e06*e34*e07+e06*e04*e37+e06*e35*e08+e06*e05*e38+e36*e04*e07+e36*e05*e08-e03*e32*e02-e03*e31*e01-e03*e37*e07+e00*e31*e04+e00*e01*e34+e00*e32*e05+e00*e02*e35+e30*e01*e04+e30*e02*e05+e03*e34*e04-e03*e38*e08+0.5*e002*e33+1.5*e33*e032+0.5*e33*e042+0.5*e33*e052-0.5*e33*e082;
-        A[1 + 10*11]=e06*e35*e18+e06*e33*e16+e00*e30*e13+e00*e10*e33+e00*e31*e14+e00*e11*e34+e00*e32*e15+e00*e12*e35+e10*e30*e03-e33*e17*e07-e33*e18*e08+e10*e31*e04+e10*e01*e34+e10*e32*e05+e10*e02*e35+e30*e11*e04+e30*e01*e14+e30*e12*e05+e30*e02*e15+3*e03*e33*e13+e03*e35*e15+e03*e34*e14+e13*e34*e04+e13*e35*e05+e33*e14*e04+e33*e15*e05+e06*e13*e36+e06*e15*e38+e06*e34*e17+e06*e14*e37+e16*e03*e36+e16*e34*e07+e16*e04*e37+e16*e35*e08+e16*e05*e38+e36*e14*e07+e36*e04*e17+e36*e15*e08+e36*e05*e18-e03*e31*e11-e03*e32*e12-e03*e37*e17-e03*e38*e18-e13*e32*e02-e13*e31*e01-e13*e37*e07-e13*e38*e08-e33*e12*e02-e33*e11*e01;
-        A[1 + 10*12]=e16*e13*e36+e10*e11*e34+0.5*e33*e152+0.5*e33*e142+0.5*e33*e162-0.5*e33*e112-0.5*e33*e122-0.5*e33*e172-0.5*e33*e182+0.5*e102*e33+1.5*e33*e132+e10*e30*e13+e10*e31*e14+e10*e32*e15+e10*e12*e35+e30*e11*e14+e30*e12*e15+e13*e35*e15+e13*e34*e14+e16*e35*e18+e16*e15*e38+e16*e34*e17+e16*e14*e37+e36*e15*e18+e36*e14*e17-e13*e31*e11-e13*e32*e12-e13*e37*e17-e13*e38*e18;
-        A[1 + 10*13]=e06*e35*e28+e36*e04*e27+e00*e20*e33+e00*e30*e23+3*e03*e33*e23+e03*e34*e24+e03*e35*e25+e23*e34*e04+e23*e35*e05+e33*e24*e04+e33*e25*e05+e06*e33*e26+e06*e23*e36+e06*e34*e27+e06*e24*e37+e06*e25*e38+e26*e03*e36+e26*e34*e07+e26*e04*e37+e26*e35*e08+e26*e05*e38+e36*e24*e07+e36*e25*e08+e36*e05*e28-e03*e31*e21-e03*e37*e27-e03*e32*e22-e03*e38*e28-e23*e32*e02-e23*e31*e01-e23*e37*e07-e23*e38*e08-e33*e22*e02-e33*e21*e01-e33*e27*e07-e33*e28*e08+e00*e32*e25+e00*e22*e35+e00*e31*e24+e00*e21*e34+e20*e30*e03+e20*e31*e04+e20*e01*e34+e20*e32*e05+e20*e02*e35+e30*e21*e04+e30*e01*e24+e30*e22*e05+e30*e02*e25;
-        A[1 + 10*14]=e10*e30*e23+e10*e20*e33+e10*e22*e35+e10*e32*e25+e10*e31*e24+e10*e21*e34+e20*e30*e13+e20*e31*e14+e20*e11*e34+e20*e32*e15+e20*e12*e35+e30*e21*e14+e30*e11*e24+e30*e22*e15+e30*e12*e25+3*e13*e33*e23+e13*e34*e24+e13*e35*e25+e23*e35*e15+e23*e34*e14+e33*e25*e15+e33*e24*e14+e16*e33*e26+e16*e23*e36+e16*e34*e27+e16*e24*e37+e16*e35*e28+e16*e25*e38+e26*e13*e36+e26*e35*e18+e26*e15*e38+e26*e34*e17+e26*e14*e37+e36*e25*e18+e36*e15*e28+e36*e24*e17+e36*e14*e27-e13*e31*e21-e13*e37*e27-e13*e32*e22-e13*e38*e28-e23*e31*e11-e23*e32*e12-e23*e37*e17-e23*e38*e18-e33*e21*e11-e33*e22*e12-e33*e27*e17-e33*e28*e18;
-        A[1 + 10*15]=-0.5*e33*e212-0.5*e33*e272-0.5*e33*e222-0.5*e33*e282+e26*e23*e36+e20*e30*e23+e20*e32*e25+e20*e22*e35+e20*e31*e24+e20*e21*e34+e30*e22*e25+e30*e21*e24+e23*e34*e24+e23*e35*e25+e26*e34*e27+e26*e24*e37+e26*e35*e28+e26*e25*e38+e36*e24*e27+e36*e25*e28-e23*e31*e21-e23*e37*e27-e23*e32*e22-e23*e38*e28+0.5*e202*e33+1.5*e33*e232+0.5*e33*e242+0.5*e33*e252+0.5*e33*e262;
-        A[1 + 10*16]=e33*e35*e05+e30*e32*e05+0.5*e03*e362+0.5*e302*e03+1.5*e03*e332+0.5*e03*e352+0.5*e03*e342+e00*e30*e33+e00*e31*e34+e00*e32*e35+e30*e31*e04+e30*e01*e34+e30*e02*e35+e33*e34*e04+e06*e33*e36+e06*e35*e38+e06*e34*e37+e36*e34*e07+e36*e04*e37+e36*e35*e08+e36*e05*e38-e33*e32*e02-e33*e31*e01-e33*e37*e07-e33*e38*e08-0.5*e03*e322-0.5*e03*e382-0.5*e03*e312-0.5*e03*e372;
-        A[1 + 10*17]=-e33*e31*e11-e33*e32*e12-e33*e38*e18+e30*e11*e34+e30*e32*e15+e30*e12*e35+e33*e35*e15+e33*e34*e14+e16*e33*e36+e16*e35*e38+e16*e34*e37+e36*e35*e18+e36*e15*e38+e36*e34*e17+e36*e14*e37-e33*e37*e17+0.5*e302*e13+1.5*e13*e332+0.5*e13*e352+0.5*e13*e342+0.5*e13*e362-0.5*e13*e322-0.5*e13*e382-0.5*e13*e312-0.5*e13*e372+e10*e30*e33+e10*e31*e34+e10*e32*e35+e30*e31*e14;
-        A[1 + 10*18]=e36*e25*e38+0.5*e302*e23+1.5*e23*e332+0.5*e23*e352+0.5*e23*e342+0.5*e23*e362-0.5*e23*e322-0.5*e23*e382-0.5*e23*e312-0.5*e23*e372+e20*e30*e33+e20*e31*e34+e20*e32*e35+e30*e32*e25+e30*e22*e35+e30*e31*e24+e30*e21*e34+e33*e34*e24+e33*e35*e25+e26*e33*e36+e26*e35*e38+e26*e34*e37+e36*e34*e27+e36*e24*e37+e36*e35*e28-e33*e31*e21-e33*e37*e27-e33*e32*e22-e33*e38*e28;
-        A[1 + 10*19]=0.5*e302*e33+e30*e31*e34+e30*e32*e35+0.5*e333+0.5*e33*e352+0.5*e33*e342+0.5*e33*e362+e36*e35*e38+e36*e34*e37-0.5*e33*e322-0.5*e33*e382-0.5*e33*e312-0.5*e33*e372;
-        A[2 + 10*0]=0.5*e002*e06+e00*e01*e07+e00*e02*e08+0.5*e032*e06+e03*e04*e07+e03*e05*e08+0.5*e063+0.5*e06*e072+0.5*e06*e082-0.5*e06*e012-0.5*e06*e022-0.5*e06*e042-0.5*e06*e052;
-        A[2 + 10*1]=e00*e10*e06+0.5*e002*e16+0.5*e032*e16+1.5*e16*e062+0.5*e16*e072+0.5*e16*e082-0.5*e16*e012-0.5*e16*e022-0.5*e16*e042-0.5*e16*e052+e00*e11*e07+e00*e01*e17+e00*e12*e08+e00*e02*e18+e10*e01*e07+e10*e02*e08+e03*e13*e06+e03*e14*e07+e03*e04*e17+e03*e15*e08+e03*e05*e18+e13*e04*e07+e13*e05*e08+e06*e17*e07+e06*e18*e08-e06*e12*e02-e06*e11*e01-e06*e14*e04-e06*e15*e05;
-        A[2 + 10*2]=e13*e14*e07+0.5*e102*e06+e00*e10*e16+e00*e12*e18+e00*e11*e17+e10*e11*e07+e10*e01*e17+e10*e12*e08+e10*e02*e18+e03*e13*e16+e03*e15*e18+e03*e14*e17+e13*e04*e17+e13*e15*e08+e13*e05*e18+e16*e17*e07+e16*e18*e08-e16*e12*e02-e16*e11*e01-e16*e14*e04-e16*e15*e05+0.5*e132*e06+1.5*e06*e162+0.5*e06*e182+0.5*e06*e172-0.5*e06*e112-0.5*e06*e122-0.5*e06*e142-0.5*e06*e152;
-        A[2 + 10*3]=0.5*e102*e16+e10*e12*e18+e10*e11*e17+0.5*e132*e16+e13*e15*e18+e13*e14*e17+0.5*e163+0.5*e16*e182+0.5*e16*e172-0.5*e16*e112-0.5*e16*e122-0.5*e16*e142-0.5*e16*e152;
-        A[2 + 10*4]=e06*e27*e07+e23*e05*e08+e23*e04*e07+e03*e05*e28+e03*e25*e08+e03*e04*e27+e03*e24*e07+e20*e02*e08+e20*e01*e07+e00*e02*e28+e00*e22*e08+e00*e01*e27+e00*e21*e07+e00*e20*e06-e06*e25*e05-e06*e24*e04-e06*e21*e01-e06*e22*e02+e06*e28*e08-0.5*e26*e042-0.5*e26*e052-0.5*e26*e012-0.5*e26*e022+0.5*e26*e082+0.5*e26*e072+1.5*e26*e062+0.5*e002*e26+e03*e23*e06+0.5*e032*e26;
-        A[2 + 10*5]=e13*e05*e28+e00*e12*e28+e13*e25*e08+e13*e04*e27+e13*e24*e07+e13*e23*e06+e03*e14*e27+e03*e24*e17+e03*e15*e28+e03*e25*e18+e03*e13*e26+e03*e23*e16+e20*e02*e18+e20*e12*e08+e20*e01*e17+e20*e11*e07+e00*e21*e17+e10*e02*e28+e10*e22*e08+e10*e01*e27+e10*e21*e07+e10*e20*e06+e00*e11*e27-e26*e15*e05-e26*e14*e04-e26*e11*e01-e26*e12*e02-e16*e25*e05-e16*e24*e04-e16*e21*e01-e16*e22*e02-e06*e24*e14-e06*e22*e12-e06*e21*e11-e06*e25*e15+e00*e20*e16+e00*e22*e18+e00*e10*e26+e26*e18*e08+e26*e17*e07+e16*e28*e08+e16*e27*e07+e06*e27*e17+e06*e28*e18+3*e06*e26*e16+e23*e05*e18+e23*e15*e08+e23*e04*e17+e23*e14*e07;
-        A[2 + 10*6]=e10*e22*e18+0.5*e26*e182+0.5*e26*e172+e16*e28*e18+e16*e27*e17-e16*e25*e15-e16*e21*e11-e16*e22*e12+1.5*e26*e162+e13*e15*e28+e13*e24*e17+e13*e14*e27+e23*e15*e18+e23*e14*e17+e10*e12*e28+e10*e21*e17+e10*e11*e27+e20*e12*e18+e20*e11*e17+e13*e23*e16+e13*e25*e18+e10*e20*e16+0.5*e102*e26-0.5*e26*e122-0.5*e26*e142-0.5*e26*e152-e16*e24*e14-0.5*e26*e112+0.5*e132*e26;
-        A[2 + 10*7]=-0.5*e06*e212-0.5*e06*e252-0.5*e06*e242+0.5*e06*e272+0.5*e06*e282-0.5*e06*e222+e20*e02*e28+e03*e23*e26+e03*e24*e27+e03*e25*e28+e23*e24*e07+e23*e04*e27+e23*e25*e08+e23*e05*e28+e26*e28*e08-e26*e22*e02-e26*e21*e01-e26*e24*e04-e26*e25*e05+e26*e27*e07+e00*e20*e26+e00*e21*e27+e00*e22*e28+e20*e21*e07+e20*e01*e27+e20*e22*e08+0.5*e202*e06+0.5*e232*e06+1.5*e06*e262;
-        A[2 + 10*8]=-e26*e24*e14-0.5*e16*e212-0.5*e16*e252-0.5*e16*e242-e26*e25*e15-0.5*e16*e222-e26*e21*e11+e26*e28*e18+e26*e27*e17-e26*e22*e12+e23*e15*e28+e23*e24*e17+e23*e14*e27+0.5*e232*e16+1.5*e16*e262+0.5*e16*e272+0.5*e16*e282+e10*e20*e26+e10*e21*e27+e10*e22*e28+e20*e22*e18+e20*e12*e28+e20*e21*e17+e20*e11*e27+e13*e23*e26+e13*e24*e27+e13*e25*e28+e23*e25*e18+0.5*e202*e16;
-        A[2 + 10*9]=0.5*e202*e26+e20*e21*e27+e20*e22*e28+0.5*e232*e26+e23*e24*e27+e23*e25*e28+0.5*e263+0.5*e26*e272+0.5*e26*e282-0.5*e26*e222-0.5*e26*e212-0.5*e26*e252-0.5*e26*e242;
-        A[2 + 10*10]=e03*e34*e07+0.5*e032*e36+1.5*e36*e062+e03*e33*e06+e00*e31*e07+e00*e01*e37+e00*e32*e08+e00*e02*e38+e30*e01*e07+e30*e02*e08+e03*e04*e37+e03*e35*e08+e03*e05*e38+0.5*e002*e36-0.5*e36*e022-0.5*e36*e042-0.5*e36*e052+0.5*e36*e072+0.5*e36*e082-0.5*e36*e012+e33*e04*e07+e33*e05*e08+e06*e37*e07+e06*e38*e08-e06*e32*e02-e06*e31*e01-e06*e34*e04-e06*e35*e05+e00*e30*e06;
-        A[2 + 10*11]=e13*e33*e06+e13*e34*e07+e13*e04*e37+e13*e35*e08+e13*e05*e38+e33*e14*e07+e33*e04*e17+e33*e15*e08+e33*e05*e18+3*e06*e36*e16+e06*e38*e18+e06*e37*e17+e16*e37*e07+e16*e38*e08+e36*e17*e07+e36*e18*e08-e06*e35*e15-e06*e31*e11-e06*e32*e12+e00*e31*e17+e00*e11*e37+e10*e30*e06+e10*e31*e07+e10*e01*e37+e10*e32*e08+e10*e02*e38+e30*e11*e07+e30*e01*e17+e30*e12*e08+e30*e02*e18+e03*e33*e16+e03*e13*e36+e03*e35*e18+e03*e15*e38+e03*e34*e17+e03*e14*e37+e00*e30*e16+e00*e12*e38-e06*e34*e14-e16*e32*e02-e16*e31*e01-e16*e34*e04-e16*e35*e05-e36*e12*e02-e36*e11*e01-e36*e14*e04-e36*e15*e05+e00*e10*e36+e00*e32*e18;
-        A[2 + 10*12]=0.5*e36*e182+0.5*e36*e172-0.5*e36*e112-0.5*e36*e122-0.5*e36*e142-0.5*e36*e152+0.5*e102*e36+0.5*e132*e36+1.5*e36*e162+e10*e30*e16+e10*e32*e18+e10*e12*e38+e10*e31*e17+e10*e11*e37+e30*e12*e18+e30*e11*e17+e13*e33*e16+e13*e35*e18+e13*e15*e38+e13*e34*e17+e13*e14*e37+e33*e15*e18+e33*e14*e17+e16*e38*e18+e16*e37*e17-e16*e35*e15-e16*e31*e11-e16*e32*e12-e16*e34*e14;
-        A[2 + 10*13]=e00*e20*e36+e00*e31*e27+e00*e21*e37+e00*e32*e28+e00*e22*e38+e20*e30*e06+e20*e31*e07+e20*e01*e37+e20*e32*e08+e20*e02*e38+e30*e21*e07+e30*e01*e27+e30*e22*e08+e30*e02*e28+e03*e33*e26+e03*e23*e36+e03*e34*e27+e03*e24*e37+e03*e35*e28-e26*e31*e01-e26*e35*e05-e36*e22*e02-e36*e21*e01-e36*e24*e04-e36*e25*e05-e26*e34*e04+e03*e25*e38+e23*e34*e07+e23*e04*e37+e23*e35*e08+e23*e05*e38+e33*e24*e07+e33*e04*e27+e33*e25*e08+e33*e05*e28+3*e06*e36*e26+e06*e37*e27+e06*e38*e28+e26*e37*e07+e26*e38*e08+e36*e27*e07+e36*e28*e08-e06*e32*e22-e06*e31*e21-e06*e35*e25-e06*e34*e24-e26*e32*e02+e00*e30*e26+e23*e33*e06;
-        A[2 + 10*14]=e10*e30*e26+e10*e20*e36+e10*e31*e27+e10*e21*e37+e10*e32*e28+e10*e22*e38+e20*e30*e16+e20*e32*e18+e20*e12*e38+e20*e31*e17+e20*e11*e37+e30*e22*e18+e30*e12*e28+e30*e21*e17+e30*e11*e27+e13*e33*e26+e13*e23*e36+e13*e34*e27+e13*e24*e37+e13*e35*e28+e13*e25*e38+e23*e33*e16+e23*e35*e18+e23*e15*e38+e23*e34*e17+e23*e14*e37+e33*e25*e18+e33*e15*e28+e33*e24*e17+e33*e14*e27+3*e16*e36*e26+e16*e37*e27+e16*e38*e28+e26*e38*e18+e26*e37*e17+e36*e28*e18+e36*e27*e17-e16*e32*e22-e16*e31*e21-e16*e35*e25-e16*e34*e24-e26*e35*e15-e26*e31*e11-e26*e32*e12-e26*e34*e14-e36*e25*e15-e36*e21*e11-e36*e22*e12-e36*e24*e14;
-        A[2 + 10*15]=e33*e25*e28+e20*e30*e26+e20*e32*e28+e20*e31*e27+e20*e21*e37+e20*e22*e38+e30*e21*e27+e30*e22*e28+e23*e33*e26+e23*e34*e27+e23*e24*e37+e23*e35*e28+e23*e25*e38+e33*e24*e27+e26*e37*e27+e26*e38*e28-e26*e32*e22-e26*e31*e21-e26*e35*e25-e26*e34*e24+0.5*e202*e36+0.5*e232*e36+1.5*e36*e262+0.5*e36*e272+0.5*e36*e282-0.5*e36*e222-0.5*e36*e212-0.5*e36*e252-0.5*e36*e242;
-        A[2 + 10*16]=e00*e30*e36+e00*e32*e38+e00*e31*e37+e30*e31*e07+e30*e01*e37+e30*e32*e08+e30*e02*e38+e03*e33*e36-0.5*e06*e342+e03*e35*e38+e33*e34*e07+e33*e04*e37+e33*e35*e08+e33*e05*e38+e36*e37*e07+e36*e38*e08-e36*e32*e02-e36*e31*e01-e36*e34*e04-e36*e35*e05+e03*e34*e37+0.5*e302*e06+0.5*e332*e06+1.5*e06*e362+0.5*e06*e382+0.5*e06*e372-0.5*e06*e352-0.5*e06*e312-0.5*e06*e322;
-        A[2 + 10*17]=-e36*e35*e15+e10*e30*e36+0.5*e302*e16+0.5*e332*e16+1.5*e16*e362+0.5*e16*e382+0.5*e16*e372-0.5*e16*e352-0.5*e16*e312-0.5*e16*e322-0.5*e16*e342+e10*e32*e38+e10*e31*e37+e30*e32*e18+e30*e12*e38+e30*e31*e17+e30*e11*e37+e13*e33*e36+e13*e35*e38+e13*e34*e37+e33*e35*e18+e33*e15*e38+e33*e34*e17+e33*e14*e37+e36*e38*e18+e36*e37*e17-e36*e31*e11-e36*e32*e12-e36*e34*e14;
-        A[2 + 10*18]=-e36*e35*e25+e30*e32*e28+0.5*e302*e26+0.5*e332*e26+1.5*e26*e362+0.5*e26*e382+0.5*e26*e372-0.5*e26*e352-0.5*e26*e312-0.5*e26*e322-0.5*e26*e342+e20*e30*e36+e20*e32*e38+e20*e31*e37+e30*e31*e27+e30*e21*e37+e30*e22*e38+e23*e33*e36+e23*e35*e38+e23*e34*e37+e33*e34*e27+e33*e24*e37+e33*e35*e28+e33*e25*e38+e36*e37*e27+e36*e38*e28-e36*e32*e22-e36*e31*e21-e36*e34*e24;
-        A[2 + 10*19]=0.5*e302*e36+e30*e32*e38+e30*e31*e37+0.5*e332*e36+e33*e35*e38+e33*e34*e37+0.5*e363+0.5*e36*e382+0.5*e36*e372-0.5*e36*e352-0.5*e36*e312-0.5*e36*e322-0.5*e36*e342;
-        A[3 + 10*0]=0.5*e01*e002+0.5*e013+0.5*e01*e022+e04*e00*e03+0.5*e01*e042+e04*e02*e05+e07*e00*e06+0.5*e01*e072+e07*e02*e08-0.5*e01*e032-0.5*e01*e052-0.5*e01*e062-0.5*e01*e082;
-        A[3 + 10*1]=1.5*e11*e012+0.5*e11*e002+0.5*e11*e022+0.5*e11*e042+0.5*e11*e072-0.5*e11*e032-0.5*e11*e052-0.5*e11*e062-0.5*e11*e082+e01*e10*e00+e01*e12*e02+e04*e10*e03+e04*e00*e13+e04*e01*e14+e04*e12*e05+e04*e02*e15+e14*e00*e03+e14*e02*e05+e07*e10*e06+e07*e00*e16+e07*e01*e17+e07*e12*e08+e07*e02*e18+e17*e00*e06+e17*e02*e08-e01*e13*e03-e01*e16*e06-e01*e15*e05-e01*e18*e08;
-        A[3 + 10*2]=e17*e02*e18+e14*e10*e03+e11*e12*e02-e11*e18*e08+0.5*e01*e102+0.5*e01*e122+1.5*e01*e112+0.5*e01*e142+0.5*e01*e172-0.5*e01*e132-0.5*e01*e152-0.5*e01*e162-0.5*e01*e182+e11*e10*e00+e04*e10*e13+e04*e12*e15+e04*e11*e14+e14*e00*e13+e14*e12*e05+e14*e02*e15+e07*e10*e16+e07*e12*e18+e07*e11*e17+e17*e10*e06+e17*e00*e16+e17*e12*e08-e11*e13*e03-e11*e16*e06-e11*e15*e05;
-        A[3 + 10*3]=0.5*e11*e102+0.5*e11*e122+0.5*e113+e14*e10*e13+e14*e12*e15+0.5*e11*e142+e17*e10*e16+e17*e12*e18+0.5*e11*e172-0.5*e11*e132-0.5*e11*e152-0.5*e11*e162-0.5*e11*e182;
-        A[3 + 10*4]=-e01*e25*e05-e01*e26*e06-e01*e23*e03+e27*e02*e08+e27*e00*e06+e07*e02*e28+e07*e22*e08+e07*e01*e27+e07*e00*e26+e24*e02*e05+e24*e00*e03+e04*e02*e25+e04*e22*e05+e04*e01*e24+e04*e00*e23+e04*e20*e03+e01*e22*e02+e01*e20*e00-e01*e28*e08+e07*e20*e06+0.5*e21*e072+0.5*e21*e042+0.5*e21*e022+0.5*e21*e002+1.5*e21*e012-0.5*e21*e082-0.5*e21*e052-0.5*e21*e062-0.5*e21*e032;
-        A[3 + 10*5]=e11*e20*e00+e07*e20*e16+3*e01*e21*e11+e01*e22*e12-e21*e18*e08-e21*e15*e05-e21*e16*e06-e21*e13*e03-e11*e28*e08-e11*e25*e05-e11*e26*e06-e11*e23*e03-e01*e28*e18-e01*e23*e13-e01*e25*e15-e01*e26*e16+e27*e02*e18+e27*e12*e08+e27*e00*e16+e27*e10*e06+e17*e02*e28+e17*e22*e08+e17*e01*e27+e17*e00*e26+e17*e20*e06+e07*e11*e27+e07*e21*e17+e07*e12*e28+e07*e22*e18+e07*e10*e26+e24*e02*e15+e24*e12*e05+e24*e00*e13+e24*e10*e03+e14*e02*e25+e14*e22*e05+e14*e01*e24+e14*e00*e23+e14*e20*e03+e04*e11*e24+e04*e21*e14+e04*e12*e25+e04*e22*e15+e21*e12*e02+e04*e20*e13+e01*e20*e10+e11*e22*e02+e21*e10*e00+e04*e10*e23;
-        A[3 + 10*6]=1.5*e21*e112+0.5*e21*e102+0.5*e21*e122+e11*e20*e10+e11*e22*e12+e14*e10*e23+e14*e22*e15+e14*e12*e25-0.5*e21*e162-0.5*e21*e152-0.5*e21*e132-0.5*e21*e182+e27*e12*e18-e11*e26*e16-e11*e25*e15-e11*e23*e13-e11*e28*e18+e17*e20*e16+e17*e10*e26+e17*e22*e18+e17*e12*e28+e17*e11*e27+e27*e10*e16+0.5*e21*e172+e14*e11*e24+e24*e10*e13+e24*e12*e15+0.5*e21*e142+e14*e20*e13;
-        A[3 + 10*7]=-0.5*e01*e262-0.5*e01*e282-0.5*e01*e252-0.5*e01*e232+0.5*e01*e272+e27*e22*e08+e27*e02*e28-e21*e23*e03-e21*e26*e06-e21*e25*e05-e21*e28*e08+e04*e22*e25+e24*e20*e03+e24*e00*e23+e24*e22*e05+e24*e02*e25+e07*e20*e26+e07*e21*e27+e07*e22*e28+e27*e20*e06+e27*e00*e26+e21*e20*e00+e21*e22*e02+e04*e20*e23+e04*e21*e24+0.5*e01*e222+0.5*e01*e242+1.5*e01*e212+0.5*e01*e202;
-        A[3 + 10*8]=-0.5*e11*e282-0.5*e11*e252-e21*e26*e16+e27*e12*e28-e21*e25*e15-e21*e23*e13-e21*e28*e18+e17*e20*e26+e17*e21*e27+e17*e22*e28+e27*e20*e16+e27*e10*e26+e27*e22*e18+0.5*e11*e242+0.5*e11*e272-0.5*e11*e232-0.5*e11*e262+0.5*e11*e202+1.5*e11*e212+0.5*e11*e222+e21*e20*e10+e14*e20*e23+e14*e21*e24+e14*e22*e25+e24*e20*e13+e24*e10*e23+e24*e22*e15+e24*e12*e25+e21*e22*e12;
-        A[3 + 10*9]=0.5*e21*e202+0.5*e213+0.5*e21*e222+e24*e20*e23+0.5*e21*e242+e24*e22*e25+e27*e20*e26+0.5*e21*e272+e27*e22*e28-0.5*e21*e232-0.5*e21*e262-0.5*e21*e282-0.5*e21*e252;
-        A[3 + 10*10]=-0.5*e31*e032-0.5*e31*e052-0.5*e31*e062-0.5*e31*e082+e07*e30*e06+e07*e00*e36+e07*e01*e37+e07*e32*e08+e07*e02*e38+e37*e00*e06+e37*e02*e08-e01*e33*e03-e01*e36*e06-e01*e35*e05-e01*e38*e08+0.5*e31*e072+e04*e30*e03+e04*e00*e33+e04*e01*e34+e04*e32*e05+e04*e02*e35+e34*e00*e03+e34*e02*e05+0.5*e31*e002+0.5*e31*e022+0.5*e31*e042+e01*e30*e00+e01*e32*e02+1.5*e31*e012;
-        A[3 + 10*11]=e34*e12*e05+e34*e02*e15+e07*e10*e36+e07*e32*e18+e07*e12*e38+e07*e31*e17+e07*e11*e37+e17*e30*e06+e17*e00*e36+e17*e01*e37+e17*e32*e08+e17*e02*e38+e37*e10*e06+e37*e00*e16+e37*e12*e08+e37*e02*e18-e01*e36*e16-e01*e35*e15-e01*e33*e13-e01*e38*e18-e11*e33*e03-e11*e36*e06-e11*e35*e05+e01*e30*e10+e01*e32*e12+3*e01*e31*e11+e11*e30*e00+e11*e32*e02+e31*e10*e00+e31*e12*e02+e04*e30*e13+e04*e10*e33+e04*e32*e15+e04*e12*e35+e04*e31*e14+e04*e11*e34+e14*e30*e03+e14*e00*e33+e14*e01*e34+e14*e32*e05+e14*e02*e35+e34*e10*e03+e34*e00*e13+e07*e30*e16-e11*e38*e08-e31*e13*e03-e31*e16*e06-e31*e15*e05-e31*e18*e08;
-        A[3 + 10*12]=-e11*e33*e13-e11*e38*e18+0.5*e31*e142+0.5*e31*e172-0.5*e31*e162-0.5*e31*e152-0.5*e31*e132-0.5*e31*e182+0.5*e31*e122+0.5*e31*e102+e11*e30*e10+e11*e32*e12+e14*e30*e13+e14*e10*e33+e14*e32*e15+e14*e12*e35+e14*e11*e34+e34*e10*e13+e34*e12*e15+e17*e30*e16+e17*e10*e36+e17*e32*e18+e17*e12*e38+e17*e11*e37+e37*e10*e16+e37*e12*e18-e11*e36*e16-e11*e35*e15+1.5*e31*e112;
-        A[3 + 10*13]=-e21*e35*e05+e07*e32*e28+e01*e30*e20-e21*e33*e03-e21*e36*e06-e21*e38*e08-e31*e23*e03-e31*e26*e06-e31*e25*e05-e31*e28*e08+3*e01*e31*e21+e01*e32*e22+e21*e30*e00+e21*e32*e02+e31*e20*e00+e31*e22*e02+e04*e30*e23+e04*e20*e33+e04*e31*e24+e04*e21*e34+e04*e32*e25+e04*e22*e35+e24*e30*e03+e24*e00*e33+e24*e01*e34+e24*e32*e05+e24*e02*e35+e34*e20*e03+e34*e00*e23+e34*e22*e05+e34*e02*e25+e07*e30*e26+e07*e20*e36+e07*e31*e27+e07*e21*e37+e07*e22*e38+e27*e30*e06+e27*e00*e36+e27*e01*e37+e27*e32*e08+e27*e02*e38+e37*e00*e26+e37*e22*e08+e37*e02*e28-e01*e33*e23-e01*e36*e26-e01*e38*e28-e01*e35*e25+e37*e20*e06;
-        A[3 + 10*14]=e11*e32*e22+e34*e12*e25+e11*e30*e20+3*e11*e31*e21+e21*e30*e10+e21*e32*e12+e34*e10*e23+e34*e22*e15+e17*e30*e26+e17*e20*e36+e17*e31*e27+e17*e21*e37+e17*e32*e28+e17*e22*e38+e27*e30*e16+e27*e10*e36+e27*e32*e18+e27*e12*e38+e27*e11*e37+e37*e20*e16+e37*e10*e26+e37*e22*e18+e37*e12*e28-e11*e33*e23-e11*e36*e26-e11*e38*e28-e11*e35*e25-e21*e36*e16-e21*e35*e15-e21*e33*e13-e21*e38*e18-e31*e26*e16-e31*e25*e15-e31*e23*e13-e31*e28*e18+e31*e20*e10+e31*e22*e12+e14*e30*e23+e14*e20*e33+e14*e31*e24+e14*e21*e34+e14*e32*e25+e14*e22*e35+e24*e30*e13+e24*e10*e33+e24*e32*e15+e24*e12*e35+e24*e11*e34+e34*e20*e13;
-        A[3 + 10*15]=-e21*e36*e26+e37*e22*e28-e21*e33*e23-e21*e38*e28-e21*e35*e25+0.5*e31*e222+0.5*e31*e242+0.5*e31*e272-0.5*e31*e232-0.5*e31*e262-0.5*e31*e282-0.5*e31*e252+e21*e30*e20+e21*e32*e22+e24*e30*e23+e24*e20*e33+e24*e21*e34+e24*e32*e25+e24*e22*e35+e34*e20*e23+e34*e22*e25+e27*e30*e26+e27*e20*e36+e27*e21*e37+e27*e32*e28+e27*e22*e38+e37*e20*e26+1.5*e31*e212+0.5*e31*e202;
-        A[3 + 10*16]=e04*e32*e35+0.5*e01*e372-0.5*e01*e352-0.5*e01*e362-0.5*e01*e332-0.5*e01*e382+e04*e31*e34+e34*e30*e03+e34*e00*e33+e34*e32*e05+e34*e02*e35+e07*e30*e36+e07*e32*e38+e07*e31*e37+e37*e30*e06+e37*e00*e36+e37*e32*e08+e04*e30*e33+e37*e02*e38-e31*e33*e03-e31*e36*e06-e31*e35*e05-e31*e38*e08+0.5*e01*e302+0.5*e01*e322+1.5*e01*e312+0.5*e01*e342+e31*e30*e00+e31*e32*e02;
-        A[3 + 10*17]=e31*e32*e12+e14*e30*e33+e14*e32*e35+e14*e31*e34+e34*e30*e13+e34*e10*e33+e34*e32*e15+e34*e12*e35+e17*e30*e36+e17*e32*e38+e17*e31*e37+e37*e30*e16+e37*e10*e36+e37*e32*e18+e37*e12*e38-e31*e36*e16-e31*e35*e15+0.5*e11*e302+0.5*e11*e322+1.5*e11*e312-e31*e33*e13-e31*e38*e18+0.5*e11*e342+0.5*e11*e372+e31*e30*e10-0.5*e11*e352-0.5*e11*e362-0.5*e11*e332-0.5*e11*e382;
-        A[3 + 10*18]=e34*e32*e25+0.5*e21*e342+0.5*e21*e372-0.5*e21*e352-0.5*e21*e362-0.5*e21*e332-0.5*e21*e382+0.5*e21*e302+0.5*e21*e322+1.5*e21*e312+e31*e30*e20+e31*e32*e22+e24*e30*e33+e24*e32*e35+e24*e31*e34+e34*e30*e23+e34*e20*e33+e34*e22*e35+e27*e30*e36+e27*e32*e38+e27*e31*e37+e37*e30*e26+e37*e20*e36+e37*e32*e28+e37*e22*e38-e31*e33*e23-e31*e36*e26-e31*e38*e28-e31*e35*e25;
-        A[3 + 10*19]=0.5*e31*e302+0.5*e31*e322+0.5*e313+e34*e30*e33+e34*e32*e35+0.5*e31*e342+e37*e30*e36+e37*e32*e38+0.5*e31*e372-0.5*e31*e352-0.5*e31*e362-0.5*e31*e332-0.5*e31*e382;
-        A[4 + 10*0]=e01*e00*e03+0.5*e012*e04+e01*e02*e05+0.5*e04*e032+0.5*e043+0.5*e04*e052+e07*e03*e06+0.5*e04*e072+e07*e05*e08-0.5*e04*e002-0.5*e04*e022-0.5*e04*e062-0.5*e04*e082;
-        A[4 + 10*1]=e07*e13*e06+e01*e10*e03-0.5*e14*e002-0.5*e14*e022-0.5*e14*e062-0.5*e14*e082+e01*e00*e13+e01*e11*e04+e01*e12*e05+e01*e02*e15+e11*e00*e03+e11*e02*e05+e04*e13*e03+e04*e15*e05+e07*e03*e16+e07*e04*e17+e07*e15*e08+e07*e05*e18+e17*e03*e06+e17*e05*e08-e04*e10*e00-e04*e12*e02-e04*e16*e06-e04*e18*e08+0.5*e012*e14+1.5*e14*e042+0.5*e14*e032+0.5*e14*e052+0.5*e14*e072;
-        A[4 + 10*2]=e11*e10*e03+0.5*e112*e04+0.5*e04*e132+0.5*e04*e152+1.5*e04*e142+0.5*e04*e172-0.5*e04*e102-0.5*e04*e162-0.5*e04*e122-0.5*e04*e182+e01*e10*e13+e01*e12*e15+e01*e11*e14+e11*e00*e13+e11*e12*e05+e11*e02*e15+e14*e13*e03+e14*e15*e05+e07*e13*e16+e07*e15*e18+e07*e14*e17+e17*e13*e06+e17*e03*e16+e17*e15*e08+e17*e05*e18-e14*e10*e00-e14*e12*e02-e14*e16*e06-e14*e18*e08;
-        A[4 + 10*3]=e11*e10*e13+e11*e12*e15+0.5*e112*e14+0.5*e14*e132+0.5*e14*e152+0.5*e143+e17*e13*e16+e17*e15*e18+0.5*e14*e172-0.5*e14*e102-0.5*e14*e162-0.5*e14*e122-0.5*e14*e182;
-        A[4 + 10*4]=-e04*e28*e08-e04*e26*e06-e04*e22*e02-e04*e20*e00+e27*e05*e08+e27*e03*e06+e07*e05*e28+e07*e25*e08+e07*e04*e27+e07*e03*e26+e07*e23*e06+e04*e25*e05+e04*e23*e03+e21*e02*e05+e21*e00*e03+e01*e02*e25+e01*e22*e05+e01*e21*e04+e01*e00*e23+e01*e20*e03+0.5*e012*e24+0.5*e24*e072+0.5*e24*e052+0.5*e24*e032+1.5*e24*e042-0.5*e24*e022-0.5*e24*e002-0.5*e24*e082-0.5*e24*e062;
-        A[4 + 10*5]=e11*e02*e25+e11*e22*e05+e11*e21*e04-e24*e18*e08-e24*e16*e06-e24*e12*e02-e14*e28*e08-e14*e26*e06-e14*e22*e02-e14*e20*e00-e04*e28*e18-e04*e22*e12-e04*e26*e16-e04*e20*e10+e27*e05*e18+e27*e15*e08+e27*e03*e16+e27*e13*e06+e17*e05*e28+e17*e25*e08+e17*e04*e27+e17*e03*e26+e17*e23*e06+e07*e14*e27+e07*e24*e17+e07*e15*e28+e07*e25*e18+e07*e13*e26+e07*e23*e16+e24*e15*e05+e24*e13*e03+e14*e25*e05+e14*e23*e03+3*e04*e24*e14+e04*e25*e15+e04*e23*e13+e21*e02*e15+e21*e12*e05+e21*e00*e13+e21*e10*e03-e24*e10*e00+e01*e20*e13+e01*e10*e23+e01*e22*e15+e01*e12*e25+e01*e11*e24+e11*e20*e03+e11*e00*e23+e01*e21*e14;
-        A[4 + 10*6]=e11*e12*e25-0.5*e24*e182-0.5*e24*e102-0.5*e24*e162-0.5*e24*e122+e27*e13*e16+e27*e15*e18-e14*e20*e10-e14*e26*e16-e14*e22*e12-e14*e28*e18+e17*e15*e28+e17*e14*e27+1.5*e24*e142+0.5*e24*e132+0.5*e24*e152+0.5*e112*e24+0.5*e24*e172+e11*e21*e14+e11*e22*e15+e11*e10*e23+e11*e20*e13+e21*e10*e13+e21*e12*e15+e17*e13*e26+e17*e23*e16+e14*e25*e15+e14*e23*e13+e17*e25*e18;
-        A[4 + 10*7]=e27*e03*e26+e27*e25*e08+e27*e05*e28-e24*e20*e00-e24*e22*e02-e24*e26*e06-e24*e28*e08+0.5*e04*e232+1.5*e04*e242+0.5*e04*e252+0.5*e04*e272-0.5*e04*e202-0.5*e04*e222-0.5*e04*e262-0.5*e04*e282+e24*e23*e03+e24*e25*e05+e07*e23*e26+e07*e24*e27+e07*e25*e28+e27*e23*e06+e21*e20*e03+e21*e00*e23+e21*e22*e05+e21*e02*e25+0.5*e212*e04+e01*e20*e23+e01*e21*e24+e01*e22*e25;
-        A[4 + 10*8]=-e24*e22*e12-e24*e28*e18+0.5*e14*e272-0.5*e14*e202-0.5*e14*e222-0.5*e14*e262-0.5*e14*e282+e17*e23*e26+e17*e24*e27+e17*e25*e28+e27*e23*e16+e27*e13*e26+e27*e25*e18+e27*e15*e28-e24*e20*e10-e24*e26*e16+0.5*e14*e232+1.5*e14*e242+0.5*e14*e252+e21*e10*e23+e21*e22*e15+e21*e12*e25+e24*e23*e13+e24*e25*e15+e21*e20*e13+0.5*e212*e14+e11*e20*e23+e11*e22*e25+e11*e21*e24;
-        A[4 + 10*9]=e21*e20*e23+0.5*e212*e24+e21*e22*e25+0.5*e24*e232+0.5*e243+0.5*e24*e252+e27*e23*e26+0.5*e24*e272+e27*e25*e28-0.5*e24*e202-0.5*e24*e222-0.5*e24*e262-0.5*e24*e282;
-        A[4 + 10*10]=-e04*e38*e08-e04*e32*e02-e04*e36*e06-0.5*e34*e002-0.5*e34*e022-0.5*e34*e062-0.5*e34*e082+e37*e03*e06+e37*e05*e08-e04*e30*e00+0.5*e34*e032+0.5*e34*e052+0.5*e34*e072+1.5*e34*e042+e01*e30*e03+e01*e00*e33+e01*e31*e04+e01*e32*e05+e01*e02*e35+e31*e00*e03+e31*e02*e05+e04*e33*e03+e04*e35*e05+e07*e03*e36+e07*e04*e37+e07*e35*e08+e07*e05*e38+0.5*e012*e34+e07*e33*e06;
-        A[4 + 10*11]=e07*e13*e36+e01*e12*e35-e04*e30*e10+e17*e04*e37+e17*e35*e08+e17*e05*e38+e37*e13*e06+e37*e03*e16+e37*e15*e08+e37*e05*e18-e04*e36*e16+e17*e33*e06+e04*e33*e13+e04*e35*e15+3*e04*e34*e14+e14*e33*e03+e14*e35*e05+e34*e13*e03+e34*e15*e05+e07*e33*e16+e07*e35*e18+e07*e15*e38+e07*e34*e17+e07*e14*e37+e17*e03*e36+e31*e10*e03+e01*e30*e13+e01*e10*e33+e01*e32*e15+e01*e31*e14+e01*e11*e34+e11*e30*e03+e11*e00*e33+e11*e31*e04+e11*e32*e05+e11*e02*e35+e31*e00*e13+e31*e12*e05+e31*e02*e15-e34*e12*e02-e34*e16*e06-e34*e18*e08-e14*e32*e02-e14*e36*e06-e14*e38*e08-e34*e10*e00-e04*e32*e12-e04*e38*e18-e14*e30*e00;
-        A[4 + 10*12]=e11*e32*e15-0.5*e34*e102-0.5*e34*e162-0.5*e34*e122-0.5*e34*e182+e37*e13*e16+0.5*e112*e34+1.5*e34*e142+0.5*e34*e132+0.5*e34*e152+0.5*e34*e172+e11*e30*e13+e11*e10*e33+e11*e12*e35+e11*e31*e14+e31*e10*e13+e31*e12*e15+e14*e33*e13+e14*e35*e15+e17*e33*e16+e17*e13*e36+e17*e35*e18+e17*e15*e38+e17*e14*e37+e37*e15*e18-e14*e30*e10-e14*e36*e16-e14*e32*e12-e14*e38*e18;
-        A[4 + 10*13]=e01*e22*e35-e04*e30*e20-e04*e32*e22+e01*e31*e24+e01*e21*e34+e01*e32*e25+e21*e30*e03+e21*e00*e33+e21*e31*e04+e21*e32*e05+e21*e02*e35+e31*e20*e03+e31*e00*e23+e31*e22*e05+e31*e02*e25+e04*e33*e23+3*e04*e34*e24+e04*e35*e25+e24*e33*e03+e37*e05*e28-e04*e36*e26-e04*e38*e28-e24*e30*e00-e24*e32*e02-e24*e36*e06-e24*e38*e08+e24*e35*e05+e34*e23*e03+e34*e25*e05+e07*e33*e26+e07*e23*e36+e07*e34*e27+e07*e24*e37+e07*e35*e28+e07*e25*e38+e27*e33*e06+e27*e03*e36+e27*e04*e37+e27*e35*e08+e27*e05*e38+e37*e23*e06+e37*e03*e26+e37*e25*e08-e34*e20*e00-e34*e22*e02-e34*e26*e06-e34*e28*e08+e01*e30*e23+e01*e20*e33;
-        A[4 + 10*14]=e21*e10*e33+e11*e30*e23+e11*e20*e33+e11*e31*e24+e11*e21*e34+e11*e32*e25+e11*e22*e35+e21*e30*e13+e21*e32*e15+e21*e12*e35+e21*e31*e14+e31*e20*e13+e31*e10*e23+e31*e22*e15+e31*e12*e25+e14*e33*e23+3*e14*e34*e24+e14*e35*e25+e24*e33*e13+e24*e35*e15+e34*e23*e13+e34*e25*e15+e17*e33*e26+e17*e23*e36+e17*e34*e27+e17*e24*e37+e17*e35*e28+e17*e25*e38+e27*e33*e16+e27*e13*e36+e27*e35*e18+e27*e15*e38+e27*e14*e37+e37*e23*e16+e37*e13*e26+e37*e25*e18+e37*e15*e28-e34*e28*e18-e34*e22*e12-e14*e32*e22-e14*e36*e26-e14*e38*e28-e24*e30*e10-e24*e36*e16-e24*e32*e12-e24*e38*e18-e34*e20*e10-e34*e26*e16-e14*e30*e20;
-        A[4 + 10*15]=-0.5*e34*e202-0.5*e34*e222-0.5*e34*e262-0.5*e34*e282+e37*e25*e28-e24*e32*e22-e24*e36*e26-e24*e38*e28-e24*e30*e20+0.5*e212*e34+1.5*e34*e242+0.5*e34*e232+0.5*e34*e252+0.5*e34*e272+e21*e30*e23+e21*e20*e33+e21*e31*e24+e21*e32*e25+e21*e22*e35+e31*e20*e23+e31*e22*e25+e24*e33*e23+e24*e35*e25+e27*e33*e26+e27*e23*e36+e27*e24*e37+e27*e35*e28+e27*e25*e38+e37*e23*e26;
-        A[4 + 10*16]=e37*e33*e06+e01*e30*e33+e01*e31*e34+e31*e30*e03+e31*e02*e35+e34*e33*e03+e34*e35*e05+e07*e33*e36+e07*e35*e38+e07*e34*e37+e37*e03*e36+e37*e35*e08+e37*e05*e38-e34*e32*e02-e34*e36*e06-e34*e38*e08+e31*e32*e05+e31*e00*e33+0.5*e312*e04+0.5*e04*e332+0.5*e04*e352+1.5*e04*e342+0.5*e04*e372+e01*e32*e35-0.5*e04*e302-0.5*e04*e322-0.5*e04*e362-0.5*e04*e382-e34*e30*e00;
-        A[4 + 10*17]=0.5*e14*e372-0.5*e14*e302-0.5*e14*e322-0.5*e14*e362-0.5*e14*e382+0.5*e312*e14+0.5*e14*e332+0.5*e14*e352+1.5*e14*e342+e11*e30*e33+e11*e32*e35+e11*e31*e34+e31*e30*e13+e31*e10*e33+e31*e32*e15+e31*e12*e35+e34*e33*e13+e34*e35*e15+e17*e33*e36+e17*e35*e38+e17*e34*e37+e37*e33*e16+e37*e13*e36+e37*e35*e18+e37*e15*e38-e34*e30*e10-e34*e36*e16-e34*e32*e12-e34*e38*e18;
-        A[4 + 10*18]=-e34*e32*e22-e34*e36*e26-e34*e38*e28+0.5*e24*e332+0.5*e24*e352+1.5*e24*e342+0.5*e24*e372-0.5*e24*e302-0.5*e24*e322-0.5*e24*e362-0.5*e24*e382+e21*e30*e33+0.5*e312*e24+e21*e32*e35+e21*e31*e34+e31*e30*e23+e31*e20*e33+e31*e32*e25+e31*e22*e35+e34*e33*e23+e34*e35*e25+e27*e33*e36+e27*e35*e38+e27*e34*e37+e37*e33*e26+e37*e23*e36+e37*e35*e28+e37*e25*e38-e34*e30*e20;
-        A[4 + 10*19]=e31*e30*e33+e31*e32*e35+0.5*e312*e34+0.5*e34*e332+0.5*e34*e352+0.5*e343+e37*e33*e36+e37*e35*e38+0.5*e34*e372-0.5*e34*e302-0.5*e34*e322-0.5*e34*e362-0.5*e34*e382;
-        A[5 + 10*0]=e01*e00*e06+0.5*e012*e07+e01*e02*e08+e04*e03*e06+0.5*e042*e07+e04*e05*e08+0.5*e07*e062+0.5*e073+0.5*e07*e082-0.5*e07*e002-0.5*e07*e022-0.5*e07*e032-0.5*e07*e052;
-        A[5 + 10*1]=e04*e13*e06+0.5*e042*e17+1.5*e17*e072+0.5*e17*e062+0.5*e17*e082-0.5*e17*e002-0.5*e17*e022-0.5*e17*e032-0.5*e17*e052+e01*e10*e06+e07*e16*e06+e07*e18*e08-e07*e10*e00-e07*e12*e02-e07*e13*e03-e07*e15*e05+e01*e00*e16+e01*e11*e07+e01*e12*e08+e01*e02*e18+e11*e00*e06+e11*e02*e08+e04*e03*e16+e04*e14*e07+e04*e15*e08+e04*e05*e18+e14*e03*e06+e14*e05*e08+0.5*e012*e17;
-        A[5 + 10*2]=-e17*e10*e00+0.5*e112*e07+0.5*e142*e07+0.5*e07*e162+0.5*e07*e182+1.5*e07*e172-0.5*e07*e102-0.5*e07*e152-0.5*e07*e132-0.5*e07*e122+e01*e10*e16+e01*e12*e18+e01*e11*e17+e11*e10*e06+e11*e00*e16+e11*e12*e08+e11*e02*e18+e04*e13*e16+e04*e15*e18+e04*e14*e17+e14*e13*e06+e14*e03*e16+e14*e15*e08+e14*e05*e18+e17*e16*e06+e17*e18*e08-e17*e12*e02-e17*e13*e03-e17*e15*e05;
-        A[5 + 10*3]=e11*e10*e16+e11*e12*e18+0.5*e112*e17+e14*e13*e16+e14*e15*e18+0.5*e142*e17+0.5*e17*e162+0.5*e17*e182+0.5*e173-0.5*e17*e102-0.5*e17*e152-0.5*e17*e132-0.5*e17*e122;
-        A[5 + 10*4]=e01*e22*e08+e07*e28*e08-e07*e20*e00-e07*e23*e03-e07*e22*e02-e07*e25*e05+0.5*e012*e27+0.5*e042*e27+1.5*e27*e072+0.5*e27*e062+0.5*e27*e082-0.5*e27*e002-0.5*e27*e022-0.5*e27*e032-0.5*e27*e052+e07*e26*e06+e01*e20*e06+e01*e00*e26+e01*e21*e07+e01*e02*e28+e21*e00*e06+e21*e02*e08+e04*e23*e06+e04*e03*e26+e04*e24*e07+e04*e25*e08+e04*e05*e28+e24*e03*e06+e24*e05*e08;
-        A[5 + 10*5]=e14*e24*e07+e14*e03*e26+e14*e23*e06+e04*e14*e27+e04*e24*e17+e04*e15*e28+e04*e25*e18+e04*e13*e26+e04*e23*e16+e21*e02*e18+e21*e12*e08-e27*e15*e05-e27*e13*e03-e27*e12*e02-e27*e10*e00-e17*e25*e05-e17*e23*e03-e17*e22*e02-e17*e20*e00-e07*e22*e12-e07*e23*e13-e07*e25*e15-e07*e20*e10+e27*e18*e08+e27*e16*e06+e17*e28*e08+e17*e26*e06+3*e07*e27*e17+e07*e28*e18+e07*e26*e16+e24*e05*e18+e24*e15*e08+e24*e03*e16+e24*e13*e06+e14*e05*e28+e14*e25*e08+e01*e12*e28+e01*e20*e16+e01*e10*e26+e01*e22*e18+e01*e21*e17+e11*e20*e06+e01*e11*e27+e21*e00*e16+e21*e10*e06+e11*e21*e07+e11*e22*e08+e11*e02*e28+e11*e00*e26;
-        A[5 + 10*6]=-0.5*e27*e102-0.5*e27*e152-0.5*e27*e132-0.5*e27*e122+0.5*e142*e27+1.5*e27*e172+0.5*e27*e162+0.5*e27*e182+0.5*e112*e27+e11*e22*e18+e11*e10*e26+e11*e20*e16-e17*e22*e12-e17*e23*e13-e17*e25*e15-e17*e20*e10+e17*e28*e18+e17*e26*e16+e24*e15*e18+e24*e13*e16+e14*e24*e17+e14*e15*e28+e14*e25*e18+e14*e13*e26+e14*e23*e16+e21*e12*e18+e21*e10*e16+e11*e21*e17+e11*e12*e28;
-        A[5 + 10*7]=-0.5*e07*e252+e27*e26*e06+e27*e28*e08-e27*e20*e00-e27*e22*e02-e27*e23*e03-e27*e25*e05+1.5*e07*e272+0.5*e07*e282+e01*e22*e28+e21*e20*e06+e21*e00*e26+e21*e22*e08+e21*e02*e28+e04*e23*e26+e04*e24*e27+e04*e25*e28+e24*e23*e06+e24*e03*e26+e24*e25*e08+e24*e05*e28+0.5*e212*e07+0.5*e242*e07+0.5*e07*e262+e01*e20*e26+e01*e21*e27-0.5*e07*e202-0.5*e07*e232-0.5*e07*e222;
-        A[5 + 10*8]=-e27*e25*e15-e27*e23*e13-e27*e22*e12-0.5*e17*e252-0.5*e17*e202-0.5*e17*e222-0.5*e17*e232+0.5*e17*e262+1.5*e17*e272+0.5*e17*e282+e24*e23*e16+e24*e13*e26+e24*e25*e18+e24*e15*e28+e27*e26*e16+e27*e28*e18-e27*e20*e10+e14*e24*e27+e14*e25*e28+0.5*e212*e17+0.5*e242*e17+e11*e20*e26+e11*e21*e27+e11*e22*e28+e21*e20*e16+e21*e10*e26+e21*e22*e18+e21*e12*e28+e14*e23*e26;
-        A[5 + 10*9]=e21*e20*e26+0.5*e212*e27+e21*e22*e28+e24*e23*e26+0.5*e242*e27+e24*e25*e28+0.5*e27*e262+0.5*e273+0.5*e27*e282-0.5*e27*e202-0.5*e27*e222-0.5*e27*e232-0.5*e27*e252;
-        A[5 + 10*10]=e04*e05*e38+e01*e30*e06-0.5*e37*e002-0.5*e37*e022-0.5*e37*e032-0.5*e37*e052-e07*e32*e02-e07*e35*e05-e07*e33*e03+e07*e36*e06+e07*e38*e08-e07*e30*e00+1.5*e37*e072+0.5*e37*e062+0.5*e37*e082+e01*e02*e38+e31*e00*e06+e31*e02*e08+e04*e33*e06+e04*e03*e36+e04*e34*e07+e04*e35*e08+e34*e03*e06+e34*e05*e08+0.5*e012*e37+0.5*e042*e37+e01*e00*e36+e01*e31*e07+e01*e32*e08;
-        A[5 + 10*11]=e14*e33*e06+e11*e30*e06+e11*e00*e36+e11*e31*e07+e31*e10*e06+e11*e32*e08+e11*e02*e38+e31*e00*e16+e31*e12*e08+e31*e02*e18+e04*e33*e16+e04*e13*e36+e04*e35*e18+e04*e15*e38+e01*e10*e36+e01*e32*e18+e01*e12*e38+e01*e31*e17+e01*e11*e37+e01*e30*e16-e17*e35*e05-e37*e10*e00-e37*e12*e02-e37*e13*e03-e37*e15*e05+e37*e18*e08-e07*e30*e10-e07*e35*e15-e07*e33*e13-e07*e32*e12-e17*e30*e00-e17*e32*e02-e17*e33*e03+e07*e38*e18+3*e07*e37*e17+e17*e36*e06+e17*e38*e08+e37*e16*e06+e04*e34*e17+e04*e14*e37+e14*e03*e36+e14*e34*e07+e14*e35*e08+e14*e05*e38+e34*e13*e06+e34*e03*e16+e34*e15*e08+e34*e05*e18+e07*e36*e16;
-        A[5 + 10*12]=e11*e32*e18-0.5*e37*e102-0.5*e37*e152-0.5*e37*e132-0.5*e37*e122+0.5*e112*e37+0.5*e142*e37+1.5*e37*e172+0.5*e37*e162+0.5*e37*e182+e11*e10*e36+e11*e12*e38+e11*e31*e17+e31*e10*e16+e31*e12*e18+e14*e33*e16+e14*e13*e36+e14*e35*e18+e14*e15*e38+e14*e34*e17+e34*e13*e16+e34*e15*e18+e17*e36*e16+e17*e38*e18-e17*e30*e10-e17*e35*e15-e17*e33*e13-e17*e32*e12+e11*e30*e16;
-        A[5 + 10*13]=e01*e20*e36+e01*e31*e27+e01*e21*e37+e01*e32*e28+e01*e22*e38+e21*e30*e06+e21*e00*e36+e21*e31*e07+e21*e32*e08+e21*e02*e38+e01*e30*e26+e31*e20*e06+e31*e00*e26+e31*e22*e08+e31*e02*e28+e04*e33*e26+e04*e23*e36+e04*e34*e27+e04*e24*e37+e04*e35*e28+e04*e25*e38+e24*e33*e06+e24*e03*e36+e24*e34*e07+e24*e35*e08+e24*e05*e38+e34*e23*e06+e34*e03*e26+e34*e25*e08+e34*e05*e28+e07*e36*e26+3*e07*e37*e27+e07*e38*e28+e27*e36*e06+e27*e38*e08+e37*e26*e06+e37*e28*e08-e07*e30*e20-e07*e32*e22-e07*e33*e23-e07*e35*e25-e27*e30*e00-e27*e32*e02-e27*e33*e03-e27*e35*e05-e37*e20*e00-e37*e22*e02-e37*e23*e03-e37*e25*e05;
-        A[5 + 10*14]=e11*e30*e26+e11*e20*e36+e11*e31*e27+e11*e21*e37+e11*e32*e28+e11*e22*e38+e21*e10*e36+e21*e32*e18+e21*e12*e38+e21*e31*e17+e31*e20*e16+e31*e10*e26+e31*e22*e18+e31*e12*e28+e14*e33*e26+e14*e23*e36+e14*e34*e27+e14*e24*e37+e14*e35*e28+e14*e25*e38+e24*e33*e16+e24*e13*e36+e24*e35*e18+e24*e15*e38+e24*e34*e17+e34*e23*e16+e34*e13*e26+e34*e25*e18+e34*e15*e28+e17*e36*e26+3*e17*e37*e27+e17*e38*e28+e27*e36*e16+e27*e38*e18+e37*e26*e16+e37*e28*e18-e17*e30*e20-e17*e32*e22-e17*e33*e23-e17*e35*e25-e27*e30*e10-e27*e35*e15-e27*e33*e13-e27*e32*e12-e37*e20*e10-e37*e25*e15-e37*e23*e13-e37*e22*e12+e21*e30*e16;
-        A[5 + 10*15]=e21*e20*e36+e21*e31*e27+e21*e32*e28+e21*e22*e38+e31*e22*e28+e24*e33*e26+e24*e23*e36+e24*e34*e27+e24*e35*e28+e24*e25*e38+e34*e23*e26+e34*e25*e28+e27*e36*e26+e27*e38*e28-e27*e30*e20-e27*e32*e22-e27*e33*e23-e27*e35*e25+0.5*e242*e37+1.5*e37*e272+0.5*e37*e262+0.5*e37*e282+e31*e20*e26+e21*e30*e26+0.5*e212*e37-0.5*e37*e202-0.5*e37*e222-0.5*e37*e232-0.5*e37*e252;
-        A[5 + 10*16]=e01*e30*e36+e01*e32*e38+e01*e31*e37+e31*e30*e06+e31*e00*e36+e31*e32*e08+e31*e02*e38+e04*e33*e36+e04*e35*e38+e04*e34*e37+e34*e33*e06+e34*e03*e36+e34*e35*e08+e34*e05*e38+e37*e36*e06+e37*e38*e08-e37*e30*e00-e37*e32*e02-e37*e33*e03-e37*e35*e05+0.5*e312*e07+0.5*e342*e07+0.5*e07*e362+0.5*e07*e382+1.5*e07*e372-0.5*e07*e302-0.5*e07*e352-0.5*e07*e322-0.5*e07*e332;
-        A[5 + 10*17]=0.5*e312*e17+0.5*e342*e17+0.5*e17*e362+0.5*e17*e382+1.5*e17*e372-0.5*e17*e302-0.5*e17*e352-0.5*e17*e322-0.5*e17*e332-e37*e32*e12-e37*e33*e13+e11*e30*e36+e11*e32*e38+e11*e31*e37+e31*e30*e16+e31*e10*e36+e31*e32*e18+e31*e12*e38+e14*e33*e36+e14*e35*e38+e14*e34*e37+e34*e33*e16+e34*e13*e36+e34*e35*e18+e34*e15*e38+e37*e36*e16+e37*e38*e18-e37*e30*e10-e37*e35*e15;
-        A[5 + 10*18]=e21*e31*e37-0.5*e27*e332+e21*e30*e36+e21*e32*e38+e31*e30*e26+e31*e20*e36+e31*e32*e28+e31*e22*e38+e24*e33*e36+e24*e35*e38+e24*e34*e37+e34*e33*e26+e34*e23*e36+e34*e35*e28+e34*e25*e38+e37*e36*e26+e37*e38*e28-e37*e30*e20-e37*e32*e22-e37*e33*e23-e37*e35*e25+0.5*e312*e27+0.5*e342*e27+0.5*e27*e362+0.5*e27*e382+1.5*e27*e372-0.5*e27*e302-0.5*e27*e352-0.5*e27*e322;
-        A[5 + 10*19]=e31*e30*e36+e31*e32*e38+0.5*e312*e37+e34*e33*e36+e34*e35*e38+0.5*e342*e37+0.5*e37*e362+0.5*e37*e382+0.5*e373-0.5*e37*e302-0.5*e37*e352-0.5*e37*e322-0.5*e37*e332;
-        A[6 + 10*0]=0.5*e02*e002+0.5*e02*e012+0.5*e023+e05*e00*e03+e05*e01*e04+0.5*e02*e052+e08*e00*e06+e08*e01*e07+0.5*e02*e082-0.5*e02*e032-0.5*e02*e042-0.5*e02*e062-0.5*e02*e072;
-        A[6 + 10*1]=-0.5*e12*e042-0.5*e12*e062-0.5*e12*e072+0.5*e12*e082-0.5*e12*e032+1.5*e12*e022+0.5*e12*e002+0.5*e12*e012+0.5*e12*e052+e02*e10*e00+e02*e11*e01+e05*e10*e03+e05*e00*e13+e05*e11*e04+e05*e01*e14+e05*e02*e15+e15*e00*e03+e15*e01*e04+e08*e10*e06+e08*e00*e16+e08*e11*e07+e08*e01*e17+e08*e02*e18+e18*e00*e06+e18*e01*e07-e02*e13*e03-e02*e14*e04-e02*e16*e06-e02*e17*e07;
-        A[6 + 10*2]=0.5*e02*e102+1.5*e02*e122+0.5*e02*e112+0.5*e02*e152+0.5*e02*e182-0.5*e02*e162-0.5*e02*e172-0.5*e02*e132-0.5*e02*e142+e12*e10*e00+e12*e11*e01+e05*e10*e13+e05*e12*e15+e05*e11*e14+e15*e10*e03+e15*e00*e13+e15*e11*e04+e15*e01*e14+e08*e10*e16+e08*e12*e18+e08*e11*e17+e18*e10*e06+e18*e00*e16+e18*e11*e07+e18*e01*e17-e12*e13*e03-e12*e14*e04-e12*e16*e06-e12*e17*e07;
-        A[6 + 10*3]=0.5*e12*e102+0.5*e123+0.5*e12*e112+e15*e10*e13+0.5*e12*e152+e15*e11*e14+e18*e10*e16+0.5*e12*e182+e18*e11*e17-0.5*e12*e162-0.5*e12*e172-0.5*e12*e132-0.5*e12*e142;
-        A[6 + 10*4]=-0.5*e22*e032-0.5*e22*e042-0.5*e22*e062-0.5*e22*e072+0.5*e22*e082+1.5*e22*e022+0.5*e22*e002+0.5*e22*e012+0.5*e22*e052+e02*e20*e00+e02*e21*e01+e05*e20*e03+e05*e00*e23+e05*e21*e04+e05*e01*e24+e05*e02*e25+e25*e00*e03+e25*e01*e04+e08*e20*e06+e08*e00*e26+e08*e21*e07+e08*e01*e27+e08*e02*e28+e28*e00*e06+e28*e01*e07-e02*e27*e07-e02*e23*e03-e02*e24*e04-e02*e26*e06;
-        A[6 + 10*5]=-e22*e17*e07-e22*e16*e06-e22*e14*e04-e22*e13*e03-e12*e26*e06-e12*e24*e04-e12*e23*e03-e12*e27*e07-e02*e24*e14-e02*e23*e13-e02*e27*e17-e02*e26*e16+e28*e01*e17+e28*e11*e07+e28*e00*e16+e28*e10*e06+e18*e02*e28+e18*e01*e27+e18*e21*e07+e18*e00*e26+e18*e20*e06+e08*e11*e27+e08*e21*e17+e08*e12*e28+e08*e22*e18+e08*e10*e26+e25*e01*e14+e25*e11*e04+e25*e00*e13+e25*e10*e03+e15*e01*e24+e02*e21*e11+e12*e21*e01+e15*e02*e25+e15*e21*e04+e05*e22*e15+e05*e11*e24+e15*e20*e03+e15*e00*e23+e05*e10*e23+e05*e12*e25+e05*e21*e14+e22*e10*e00+e22*e11*e01+e02*e20*e10+3*e02*e22*e12+e12*e20*e00+e08*e20*e16+e05*e20*e13;
-        A[6 + 10*6]=-e12*e24*e14-e12*e23*e13-e12*e27*e17-e12*e26*e16+e28*e11*e17+e28*e10*e16+e18*e11*e27+e18*e21*e17+e18*e12*e28+e18*e10*e26+e18*e20*e16+e25*e11*e14+e25*e10*e13+e15*e11*e24+e15*e21*e14+e15*e12*e25+e15*e10*e23+e15*e20*e13+e12*e21*e11+0.5*e22*e182+0.5*e22*e152+1.5*e22*e122+0.5*e22*e102+e12*e20*e10+0.5*e22*e112-0.5*e22*e172-0.5*e22*e132-0.5*e22*e142-0.5*e22*e162;
-        A[6 + 10*7]=0.5*e02*e282+e28*e01*e27-e22*e27*e07-e22*e23*e03-e22*e24*e04-e22*e26*e06+0.5*e02*e252+e05*e20*e23+e05*e22*e25+e25*e20*e03+e25*e00*e23+e25*e21*e04+e25*e01*e24+e08*e20*e26+e08*e21*e27+e08*e22*e28+e28*e20*e06+e28*e00*e26+e28*e21*e07+e05*e21*e24+0.5*e02*e202+0.5*e02*e212+1.5*e02*e222+e22*e20*e00+e22*e21*e01-0.5*e02*e272-0.5*e02*e242-0.5*e02*e232-0.5*e02*e262;
-        A[6 + 10*8]=-e22*e27*e17-e22*e23*e13-e22*e24*e14-0.5*e12*e232-0.5*e12*e262-0.5*e12*e242-0.5*e12*e272+0.5*e12*e282+e18*e21*e27+e28*e20*e16+e28*e10*e26+e28*e21*e17+e28*e11*e27-e22*e26*e16+e18*e22*e28+0.5*e12*e252+0.5*e12*e202+0.5*e12*e212+1.5*e12*e222+e22*e20*e10+e15*e20*e23+e15*e21*e24+e15*e22*e25+e25*e20*e13+e25*e10*e23+e25*e21*e14+e25*e11*e24+e18*e20*e26+e22*e21*e11;
-        A[6 + 10*9]=0.5*e22*e202+0.5*e22*e212+0.5*e223+e25*e20*e23+e25*e21*e24+0.5*e22*e252+e28*e20*e26+e28*e21*e27+0.5*e22*e282-0.5*e22*e232-0.5*e22*e262-0.5*e22*e242-0.5*e22*e272;
-        A[6 + 10*10]=e08*e31*e07-0.5*e32*e032-e02*e33*e03-e02*e34*e04-e02*e36*e06-0.5*e32*e042-0.5*e32*e062-0.5*e32*e072+e38*e01*e07+e38*e00*e06-e02*e37*e07+e05*e31*e04+e05*e01*e34+e05*e02*e35+e35*e01*e04+e35*e00*e03+e08*e30*e06+e08*e00*e36+e08*e01*e37+e08*e02*e38+0.5*e32*e052+e02*e30*e00+e02*e31*e01+e05*e30*e03+e05*e00*e33+1.5*e32*e022+0.5*e32*e012+0.5*e32*e002+0.5*e32*e082;
-        A[6 + 10*11]=e05*e32*e15+e32*e11*e01+e38*e10*e06+e08*e12*e38-e32*e14*e04-e32*e16*e06-e32*e17*e07-e12*e36*e06-e32*e13*e03-e02*e34*e14-e12*e37*e07-e12*e33*e03-e12*e34*e04-e02*e37*e17-e02*e33*e13+e38*e01*e17-e02*e36*e16+e18*e01*e37+e18*e02*e38+e38*e00*e16+e38*e11*e07+e08*e30*e16+e08*e10*e36+e08*e32*e18+e08*e31*e17+e08*e11*e37+e18*e30*e06+e18*e00*e36+e18*e31*e07+e35*e10*e03+e35*e00*e13+e35*e11*e04+e35*e01*e14+e15*e02*e35+e05*e10*e33+e05*e12*e35+e05*e31*e14+e05*e11*e34+e15*e30*e03+e15*e00*e33+e15*e31*e04+e15*e01*e34+e05*e30*e13+e02*e30*e10+e02*e31*e11+3*e02*e32*e12+e12*e30*e00+e12*e31*e01+e32*e10*e00;
-        A[6 + 10*12]=0.5*e32*e102+0.5*e32*e112+e12*e30*e10+1.5*e32*e122+e12*e31*e11+e15*e30*e13+e15*e10*e33+e15*e12*e35+e15*e31*e14+e15*e11*e34+e35*e10*e13-0.5*e32*e162-0.5*e32*e172-0.5*e32*e132-0.5*e32*e142-e12*e37*e17-e12*e33*e13-e12*e34*e14+0.5*e32*e182+0.5*e32*e152+e35*e11*e14+e18*e30*e16+e18*e10*e36+e18*e12*e38+e18*e31*e17+e18*e11*e37+e38*e10*e16+e38*e11*e17-e12*e36*e16;
-        A[6 + 10*13]=3*e02*e32*e22+e05*e31*e24+e08*e22*e38+e02*e31*e21+e22*e30*e00+e22*e31*e01+e32*e20*e00+e32*e21*e01+e05*e30*e23+e05*e20*e33+e05*e21*e34-e22*e37*e07-e22*e33*e03-e22*e34*e04-e22*e36*e06-e32*e27*e07-e32*e23*e03-e32*e24*e04-e32*e26*e06+e05*e32*e25+e25*e30*e03+e25*e00*e33+e25*e31*e04+e25*e01*e34+e25*e02*e35+e35*e20*e03+e35*e00*e23+e35*e21*e04+e35*e01*e24+e08*e30*e26+e08*e20*e36+e08*e31*e27+e08*e21*e37+e08*e32*e28+e28*e30*e06+e28*e00*e36+e28*e31*e07+e28*e01*e37+e28*e02*e38+e38*e20*e06+e38*e00*e26+e38*e21*e07+e38*e01*e27-e02*e33*e23-e02*e36*e26-e02*e34*e24-e02*e37*e27+e05*e22*e35+e02*e30*e20;
-        A[6 + 10*14]=e18*e22*e38+e12*e31*e21+3*e12*e32*e22+e22*e30*e10+e22*e31*e11+e32*e20*e10+e32*e21*e11+e15*e30*e23+e15*e20*e33+e15*e31*e24+e15*e21*e34+e15*e32*e25+e15*e22*e35+e25*e30*e13+e25*e10*e33+e25*e12*e35+e25*e31*e14+e25*e11*e34+e35*e20*e13+e35*e10*e23+e35*e21*e14+e35*e11*e24+e18*e30*e26+e18*e20*e36+e18*e31*e27+e18*e21*e37+e18*e32*e28+e28*e30*e16+e28*e10*e36+e28*e12*e38+e28*e31*e17+e28*e11*e37+e38*e20*e16+e38*e10*e26+e12*e30*e20-e22*e37*e17-e22*e33*e13-e22*e34*e14-e32*e26*e16-e32*e27*e17-e32*e23*e13-e32*e24*e14-e22*e36*e16+e38*e21*e17+e38*e11*e27-e12*e33*e23-e12*e36*e26-e12*e34*e24-e12*e37*e27;
-        A[6 + 10*15]=e25*e30*e23+e22*e30*e20+e22*e31*e21+e25*e20*e33+e25*e31*e24+e25*e21*e34+e25*e22*e35+e35*e20*e23+e35*e21*e24+e28*e30*e26+e28*e20*e36+e28*e31*e27+e28*e21*e37+e28*e22*e38+e38*e20*e26+e38*e21*e27-e22*e33*e23-e22*e36*e26-e22*e34*e24-e22*e37*e27+0.5*e32*e212+0.5*e32*e252+1.5*e32*e222+0.5*e32*e202+0.5*e32*e282-0.5*e32*e232-0.5*e32*e262-0.5*e32*e242-0.5*e32*e272;
-        A[6 + 10*16]=0.5*e02*e302+1.5*e02*e322+0.5*e02*e312+0.5*e02*e352+0.5*e02*e382-0.5*e02*e342-0.5*e02*e362-0.5*e02*e332-0.5*e02*e372+e38*e30*e06+e32*e30*e00+e32*e31*e01+e05*e30*e33+e05*e32*e35+e05*e31*e34+e35*e30*e03+e35*e00*e33+e35*e31*e04+e35*e01*e34+e08*e30*e36+e08*e32*e38+e08*e31*e37+e38*e00*e36+e38*e31*e07+e38*e01*e37-e32*e37*e07-e32*e33*e03-e32*e34*e04-e32*e36*e06;
-        A[6 + 10*17]=e32*e30*e10+e32*e31*e11+e15*e30*e33+e15*e32*e35+e15*e31*e34+e35*e30*e13+e35*e10*e33+e35*e31*e14+e35*e11*e34+e18*e30*e36+e18*e32*e38+e18*e31*e37+e38*e30*e16+e38*e10*e36+e38*e31*e17+e38*e11*e37-e32*e36*e16-e32*e37*e17-e32*e33*e13-e32*e34*e14+0.5*e12*e382-0.5*e12*e342-0.5*e12*e362-0.5*e12*e332-0.5*e12*e372+0.5*e12*e352+1.5*e12*e322+0.5*e12*e312+0.5*e12*e302;
-        A[6 + 10*18]=0.5*e22*e302+0.5*e22*e312+0.5*e22*e352+0.5*e22*e382-0.5*e22*e342-0.5*e22*e362-0.5*e22*e332-0.5*e22*e372+1.5*e22*e322+e32*e30*e20+e32*e31*e21+e25*e30*e33+e25*e32*e35+e25*e31*e34+e35*e30*e23+e35*e20*e33+e35*e31*e24+e35*e21*e34+e28*e30*e36+e28*e32*e38+e28*e31*e37+e38*e30*e26+e38*e20*e36+e38*e31*e27+e38*e21*e37-e32*e33*e23-e32*e36*e26-e32*e34*e24-e32*e37*e27;
-        A[6 + 10*19]=0.5*e32*e302+0.5*e323+0.5*e32*e312+e35*e30*e33+0.5*e32*e352+e35*e31*e34+e38*e30*e36+0.5*e32*e382+e38*e31*e37-0.5*e32*e342-0.5*e32*e362-0.5*e32*e332-0.5*e32*e372;
-        A[7 + 10*0]=e02*e01*e04+e02*e00*e03+0.5*e022*e05+0.5*e05*e032+0.5*e05*e042+0.5*e053+e08*e03*e06+e08*e04*e07+0.5*e05*e082-0.5*e05*e002-0.5*e05*e062-0.5*e05*e012-0.5*e05*e072;
-        A[7 + 10*1]=e08*e13*e06+e02*e10*e03+e02*e00*e13+e02*e11*e04+e02*e01*e14+e02*e12*e05+e12*e01*e04+e12*e00*e03+e05*e13*e03+e05*e14*e04+e08*e03*e16+e08*e14*e07+e08*e04*e17+e08*e05*e18+e18*e03*e06+e18*e04*e07-e05*e10*e00-e05*e11*e01-e05*e16*e06-e05*e17*e07+0.5*e022*e15+1.5*e15*e052+0.5*e15*e032+0.5*e15*e042+0.5*e15*e082-0.5*e15*e002-0.5*e15*e062-0.5*e15*e012-0.5*e15*e072;
-        A[7 + 10*2]=0.5*e122*e05+0.5*e05*e132+1.5*e05*e152+0.5*e05*e142+0.5*e05*e182-0.5*e05*e102-0.5*e05*e162-0.5*e05*e112-0.5*e05*e172+e02*e10*e13+e02*e12*e15+e02*e11*e14+e12*e10*e03+e12*e00*e13+e12*e11*e04+e12*e01*e14+e15*e13*e03+e15*e14*e04+e08*e13*e16+e08*e15*e18+e08*e14*e17+e18*e13*e06+e18*e03*e16+e18*e14*e07+e18*e04*e17-e15*e11*e01-e15*e16*e06-e15*e17*e07-e15*e10*e00;
-        A[7 + 10*3]=e12*e10*e13+0.5*e122*e15+e12*e11*e14+0.5*e15*e132+0.5*e153+0.5*e15*e142+e18*e13*e16+0.5*e15*e182+e18*e14*e17-0.5*e15*e102-0.5*e15*e162-0.5*e15*e112-0.5*e15*e172;
-        A[7 + 10*4]=0.5*e25*e082-0.5*e25*e002-0.5*e25*e062-0.5*e25*e012-0.5*e25*e072+e02*e20*e03+e02*e00*e23+e02*e21*e04+e02*e01*e24+e02*e22*e05+e22*e01*e04+e22*e00*e03+e05*e23*e03+e05*e24*e04+e08*e23*e06+e08*e03*e26+e08*e24*e07+e08*e04*e27+e08*e05*e28+e28*e03*e06+e28*e04*e07-e05*e20*e00-e05*e27*e07-e05*e21*e01-e05*e26*e06+0.5*e022*e25+1.5*e25*e052+0.5*e25*e032+0.5*e25*e042;
-        A[7 + 10*5]=-e25*e17*e07-e25*e16*e06-e25*e11*e01-e25*e10*e00-e15*e26*e06-e15*e21*e01-e15*e27*e07-e15*e20*e00-e05*e27*e17-e05*e21*e11-e05*e26*e16-e05*e20*e10+e28*e04*e17+e28*e14*e07+e28*e03*e16+e28*e13*e06+e18*e05*e28+e18*e04*e27+e18*e24*e07+e18*e03*e26+e18*e23*e06+e08*e14*e27+e08*e24*e17+e08*e15*e28+e08*e25*e18+e08*e13*e26+e08*e23*e16+e25*e14*e04+e25*e13*e03+e15*e24*e04+e15*e23*e03+e05*e24*e14+3*e05*e25*e15+e05*e23*e13+e22*e01*e14+e22*e11*e04+e22*e00*e13+e22*e10*e03+e12*e22*e05+e12*e01*e24+e12*e21*e04+e12*e00*e23+e12*e20*e03+e02*e11*e24+e02*e21*e14+e02*e12*e25+e02*e22*e15+e02*e10*e23+e02*e20*e13;
-        A[7 + 10*6]=-e15*e27*e17-e15*e21*e11-e15*e26*e16+e28*e14*e17+e28*e13*e16+e18*e14*e27+e18*e24*e17+e18*e15*e28+e18*e13*e26+e15*e24*e14+e15*e23*e13+e22*e11*e14+e22*e10*e13+e12*e11*e24+e12*e21*e14+e12*e22*e15+e12*e10*e23+e18*e23*e16+0.5*e25*e142+0.5*e25*e182+1.5*e25*e152+0.5*e25*e132+0.5*e122*e25+e12*e20*e13-0.5*e25*e172-0.5*e25*e162-0.5*e25*e112-0.5*e25*e102-e15*e20*e10;
-        A[7 + 10*7]=e28*e24*e07-0.5*e05*e272-0.5*e05*e262-0.5*e05*e212+0.5*e05*e282-0.5*e05*e202+e28*e23*e06+e08*e23*e26+e08*e25*e28+e08*e24*e27+e28*e03*e26+e28*e04*e27-e25*e27*e07-e25*e21*e01-e25*e26*e06+e02*e20*e23+e02*e22*e25+e02*e21*e24+e22*e20*e03+e22*e00*e23+e22*e21*e04+e22*e01*e24+e25*e23*e03+e25*e24*e04+0.5*e222*e05+0.5*e05*e232+1.5*e05*e252+0.5*e05*e242-e25*e20*e00;
-        A[7 + 10*8]=-0.5*e15*e202-0.5*e15*e262-0.5*e15*e212-0.5*e15*e272+e18*e23*e26+e18*e25*e28+e18*e24*e27+e28*e23*e16+e28*e13*e26+e28*e24*e17+e28*e14*e27-e25*e20*e10-e25*e26*e16-e25*e21*e11-e25*e27*e17+0.5*e15*e282+0.5*e15*e232+1.5*e15*e252+0.5*e15*e242+0.5*e222*e15+e12*e21*e24+e22*e20*e13+e22*e10*e23+e22*e21*e14+e22*e11*e24+e25*e23*e13+e25*e24*e14+e12*e20*e23+e12*e22*e25;
-        A[7 + 10*9]=e22*e20*e23+0.5*e222*e25+e22*e21*e24+0.5*e25*e232+0.5*e253+0.5*e25*e242+e28*e23*e26+0.5*e25*e282+e28*e24*e27-0.5*e25*e202-0.5*e25*e262-0.5*e25*e212-0.5*e25*e272;
-        A[7 + 10*10]=-0.5*e35*e062-0.5*e35*e012-0.5*e35*e072-e05*e30*e00-e05*e31*e01-e05*e36*e06-e05*e37*e07-0.5*e35*e002+0.5*e35*e082+e05*e34*e04+e08*e33*e06+e08*e03*e36+e08*e34*e07+e08*e04*e37+e08*e05*e38+e38*e04*e07+e38*e03*e06+0.5*e022*e35+1.5*e35*e052+0.5*e35*e042+0.5*e35*e032+e02*e30*e03+e02*e00*e33+e02*e31*e04+e02*e01*e34+e02*e32*e05+e32*e01*e04+e32*e00*e03+e05*e33*e03;
-        A[7 + 10*11]=e08*e33*e16-e35*e16*e06-e35*e17*e07-e15*e30*e00-e15*e37*e07-e15*e31*e01-e15*e36*e06-e35*e10*e00-e35*e11*e01-e05*e37*e17-e05*e31*e11+e38*e04*e17-e05*e30*e10-e05*e36*e16+e18*e33*e06+e18*e03*e36+e18*e34*e07+e18*e04*e37+e18*e05*e38+e38*e13*e06+e38*e03*e16+e38*e14*e07+e35*e14*e04+e08*e13*e36+e08*e35*e18+e08*e15*e38+e08*e34*e17+e08*e14*e37+e35*e13*e03+e05*e33*e13+3*e05*e35*e15+e05*e34*e14+e15*e33*e03+e15*e34*e04+e12*e01*e34+e12*e32*e05+e32*e10*e03+e32*e00*e13+e32*e11*e04+e32*e01*e14+e12*e30*e03+e02*e30*e13+e02*e32*e15+e02*e10*e33+e02*e12*e35+e12*e00*e33+e02*e31*e14+e02*e11*e34+e12*e31*e04;
-        A[7 + 10*12]=-0.5*e35*e162-0.5*e35*e172-e15*e36*e16-e15*e31*e11-e15*e37*e17-0.5*e35*e102-0.5*e35*e112-e15*e30*e10+e18*e13*e36+e18*e15*e38+e18*e34*e17+e18*e14*e37+e38*e13*e16+e38*e14*e17+e18*e33*e16+1.5*e35*e152+0.5*e35*e132+0.5*e35*e142+0.5*e35*e182+0.5*e122*e35+e32*e10*e13+e32*e11*e14+e15*e33*e13+e15*e34*e14+e12*e10*e33+e12*e32*e15+e12*e31*e14+e12*e11*e34+e12*e30*e13;
-        A[7 + 10*13]=e05*e33*e23+3*e05*e35*e25+e05*e34*e24+e25*e33*e03+e25*e34*e04+e35*e23*e03+e35*e24*e04+e08*e33*e26+e08*e23*e36+e08*e35*e28+e02*e20*e33+e02*e32*e25+e02*e22*e35+e02*e31*e24+e02*e21*e34+e22*e30*e03+e22*e00*e33+e22*e31*e04+e22*e01*e34+e22*e32*e05+e32*e20*e03+e32*e00*e23+e32*e21*e04+e32*e01*e24+e02*e30*e23-e35*e27*e07-e35*e21*e01-e35*e26*e06+e08*e25*e38+e08*e34*e27+e08*e24*e37+e28*e33*e06+e28*e03*e36+e28*e34*e07+e28*e04*e37+e28*e05*e38+e38*e23*e06+e38*e03*e26+e38*e24*e07+e38*e04*e27-e05*e30*e20-e05*e36*e26-e05*e31*e21-e05*e37*e27-e25*e30*e00-e25*e37*e07-e25*e31*e01-e25*e36*e06-e35*e20*e00;
-        A[7 + 10*14]=e12*e21*e34+e18*e25*e38+e12*e30*e23+e12*e20*e33+e12*e32*e25+e12*e22*e35+e12*e31*e24+e22*e30*e13+e22*e10*e33+e22*e32*e15+e22*e31*e14+e22*e11*e34+e32*e20*e13+e32*e10*e23+e32*e21*e14-e25*e30*e10-e25*e36*e16-e25*e31*e11-e25*e37*e17-e35*e20*e10-e35*e26*e16-e35*e21*e11-e35*e27*e17+e15*e33*e23+3*e15*e35*e25+e15*e34*e24+e25*e33*e13+e25*e34*e14+e35*e23*e13+e35*e24*e14+e18*e33*e26+e18*e23*e36+e18*e35*e28+e18*e34*e27+e18*e24*e37+e28*e33*e16+e28*e13*e36+e28*e15*e38+e28*e34*e17+e28*e14*e37+e38*e23*e16+e38*e13*e26+e38*e24*e17+e38*e14*e27-e15*e30*e20-e15*e36*e26-e15*e31*e21-e15*e37*e27+e32*e11*e24;
-        A[7 + 10*15]=-0.5*e35*e202-0.5*e35*e262-0.5*e35*e212-0.5*e35*e272+e25*e34*e24+e28*e23*e36+e28*e25*e38+e28*e34*e27+e28*e24*e37+e38*e23*e26+e38*e24*e27-e25*e30*e20-e25*e36*e26-e25*e31*e21-e25*e37*e27+e25*e33*e23+0.5*e222*e35+1.5*e35*e252+0.5*e35*e232+0.5*e35*e242+0.5*e35*e282+e22*e30*e23+e22*e20*e33+e22*e32*e25+e22*e31*e24+e22*e21*e34+e32*e20*e23+e32*e21*e24+e28*e33*e26;
-        A[7 + 10*16]=-e35*e30*e00-e35*e31*e01-e35*e36*e06-e35*e37*e07+0.5*e322*e05+0.5*e05*e332+0.5*e05*e342+1.5*e05*e352+0.5*e05*e382-0.5*e05*e302-0.5*e05*e362-0.5*e05*e312-0.5*e05*e372+e02*e30*e33+e02*e31*e34+e02*e32*e35+e32*e30*e03+e32*e00*e33+e32*e31*e04+e32*e01*e34+e35*e33*e03+e35*e34*e04+e08*e33*e36+e08*e34*e37+e08*e35*e38+e38*e33*e06+e38*e03*e36+e38*e34*e07+e38*e04*e37;
-        A[7 + 10*17]=-e35*e30*e10+e12*e32*e35-0.5*e15*e362-0.5*e15*e312-0.5*e15*e372-e35*e36*e16+0.5*e322*e15+0.5*e15*e332+0.5*e15*e342+1.5*e15*e352+0.5*e15*e382-0.5*e15*e302+e12*e30*e33+e12*e31*e34+e32*e30*e13+e32*e10*e33+e32*e31*e14+e32*e11*e34+e35*e33*e13+e35*e34*e14+e18*e33*e36+e18*e34*e37+e18*e35*e38+e38*e33*e16+e38*e13*e36+e38*e34*e17+e38*e14*e37-e35*e31*e11-e35*e37*e17;
-        A[7 + 10*18]=-0.5*e25*e302-0.5*e25*e362-0.5*e25*e312-0.5*e25*e372+0.5*e322*e25+0.5*e25*e332+0.5*e25*e342+1.5*e25*e352+0.5*e25*e382+e22*e30*e33+e22*e31*e34+e22*e32*e35+e32*e30*e23+e32*e20*e33+e32*e31*e24+e32*e21*e34+e35*e33*e23+e35*e34*e24+e28*e33*e36+e28*e34*e37+e28*e35*e38+e38*e33*e26+e38*e23*e36+e38*e34*e27+e38*e24*e37-e35*e30*e20-e35*e36*e26-e35*e31*e21-e35*e37*e27;
-        A[7 + 10*19]=e32*e30*e33+e32*e31*e34+0.5*e322*e35+0.5*e35*e332+0.5*e35*e342+0.5*e353+e38*e33*e36+e38*e34*e37+0.5*e35*e382-0.5*e35*e302-0.5*e35*e362-0.5*e35*e312-0.5*e35*e372;
-        A[8 + 10*0]=e02*e00*e06+e02*e01*e07+0.5*e022*e08+e05*e04*e07+e05*e03*e06+0.5*e052*e08+0.5*e08*e062+0.5*e08*e072+0.5*e083-0.5*e08*e042-0.5*e08*e002-0.5*e08*e012-0.5*e08*e032;
-        A[8 + 10*1]=e02*e10*e06+e02*e00*e16+e02*e11*e07+e02*e01*e17+e02*e12*e08+e12*e00*e06+e12*e01*e07+e05*e13*e06+e05*e03*e16+e05*e14*e07+e05*e04*e17+e05*e15*e08+e15*e04*e07+e15*e03*e06+e08*e16*e06+e08*e17*e07-e08*e10*e00-e08*e11*e01-e08*e13*e03-e08*e14*e04+0.5*e022*e18+0.5*e052*e18+1.5*e18*e082+0.5*e18*e062+0.5*e18*e072-0.5*e18*e042-0.5*e18*e002-0.5*e18*e012-0.5*e18*e032;
-        A[8 + 10*2]=e12*e01*e17+0.5*e152*e08+0.5*e08*e162+1.5*e08*e182+0.5*e08*e172-0.5*e08*e102-0.5*e08*e112-0.5*e08*e132-0.5*e08*e142+e05*e13*e16+e05*e14*e17+e05*e15*e18+e15*e13*e06+e15*e03*e16+e15*e14*e07+e15*e04*e17+e18*e16*e06+e18*e17*e07-e18*e10*e00-e18*e11*e01-e18*e13*e03-e18*e14*e04+0.5*e122*e08+e02*e10*e16+e02*e12*e18+e02*e11*e17+e12*e10*e06+e12*e00*e16+e12*e11*e07;
-        A[8 + 10*3]=e12*e10*e16+0.5*e122*e18+e12*e11*e17+e15*e13*e16+e15*e14*e17+0.5*e152*e18+0.5*e18*e162+0.5*e183+0.5*e18*e172-0.5*e18*e102-0.5*e18*e112-0.5*e18*e132-0.5*e18*e142;
-        A[8 + 10*4]=-e08*e20*e00+e08*e27*e07-e08*e21*e01-e08*e23*e03-e08*e24*e04+e02*e20*e06+e02*e00*e26+e02*e21*e07+e02*e01*e27+e02*e22*e08+e22*e00*e06+e22*e01*e07+e05*e23*e06+e05*e03*e26+e05*e24*e07+e05*e04*e27+e05*e25*e08+e25*e04*e07+e25*e03*e06+e08*e26*e06+0.5*e022*e28+0.5*e052*e28+1.5*e28*e082+0.5*e28*e062+0.5*e28*e072-0.5*e28*e042-0.5*e28*e002-0.5*e28*e012-0.5*e28*e032;
-        A[8 + 10*5]=e22*e10*e06+e22*e11*e07+e22*e01*e17+e05*e23*e16+e05*e13*e26+e05*e25*e18+e05*e15*e28+e05*e24*e17+e05*e14*e27+e15*e23*e06+e15*e03*e26+e15*e24*e07+e15*e04*e27+e15*e25*e08+e25*e13*e06+e25*e03*e16+e25*e14*e07+e25*e04*e17+e08*e26*e16+3*e08*e28*e18+e08*e27*e17+e18*e26*e06+e18*e27*e07+e22*e00*e16+e28*e16*e06+e28*e17*e07-e08*e20*e10-e08*e21*e11-e08*e23*e13-e08*e24*e14-e18*e20*e00-e18*e21*e01-e18*e23*e03-e18*e24*e04-e28*e10*e00-e28*e11*e01-e28*e13*e03-e28*e14*e04+e02*e20*e16+e02*e10*e26+e02*e22*e18+e02*e12*e28+e02*e21*e17+e02*e11*e27+e12*e20*e06+e12*e00*e26+e12*e21*e07+e12*e01*e27+e12*e22*e08;
-        A[8 + 10*6]=-e18*e24*e14-e18*e21*e11-e18*e23*e13-e18*e20*e10+e18*e27*e17+e18*e26*e16+e25*e14*e17+e25*e13*e16+e15*e25*e18+e15*e14*e27+e15*e24*e17+e15*e13*e26+e15*e23*e16+e22*e11*e17+e22*e10*e16+e12*e11*e27+e12*e21*e17+e12*e22*e18+e12*e10*e26+e12*e20*e16+0.5*e28*e162+0.5*e28*e172+1.5*e28*e182+0.5*e152*e28-0.5*e28*e142-0.5*e28*e112-0.5*e28*e132-0.5*e28*e102+0.5*e122*e28;
-        A[8 + 10*7]=-e28*e24*e04-e28*e21*e01-e28*e23*e03-e28*e20*e00+e28*e27*e07+e28*e26*e06+e25*e04*e27+e25*e24*e07+e25*e03*e26+e05*e24*e27+e05*e25*e28+e05*e23*e26+e22*e01*e27+e22*e21*e07+e22*e00*e26+e22*e20*e06+e02*e22*e28+e02*e20*e26+e02*e21*e27+0.5*e222*e08-0.5*e08*e242-0.5*e08*e212-0.5*e08*e232-0.5*e08*e202+0.5*e08*e262+0.5*e08*e272+1.5*e08*e282+0.5*e252*e08+e25*e23*e06;
-        A[8 + 10*8]=e25*e24*e17+e25*e14*e27+e28*e26*e16+e28*e27*e17-e28*e21*e11-e28*e24*e14+e12*e22*e28+e22*e10*e26+e22*e21*e17+e22*e11*e27+e15*e23*e26+e15*e25*e28+e15*e24*e27+e25*e23*e16+e25*e13*e26+e22*e20*e16+0.5*e222*e18+0.5*e252*e18+0.5*e18*e262+0.5*e18*e272+e12*e20*e26+e12*e21*e27-e28*e20*e10-0.5*e18*e232-0.5*e18*e242-e28*e23*e13-0.5*e18*e212+1.5*e18*e282-0.5*e18*e202;
-        A[8 + 10*9]=e22*e20*e26+e22*e21*e27+0.5*e222*e28+e25*e23*e26+0.5*e252*e28+e25*e24*e27+0.5*e28*e262+0.5*e28*e272+0.5*e283-0.5*e28*e202-0.5*e28*e212-0.5*e28*e232-0.5*e28*e242;
-        A[8 + 10*10]=-e08*e30*e00-0.5*e38*e042-0.5*e38*e002-0.5*e38*e012-0.5*e38*e032+1.5*e38*e082+0.5*e38*e062+0.5*e38*e072+e32*e01*e07+e05*e33*e06+e05*e03*e36+e05*e34*e07+e05*e04*e37+e05*e35*e08+e35*e04*e07+e35*e03*e06+e08*e36*e06+e08*e37*e07+0.5*e052*e38+e32*e00*e06+e02*e30*e06+e02*e00*e36+e02*e31*e07+e02*e01*e37+e02*e32*e08+0.5*e022*e38-e08*e33*e03-e08*e31*e01-e08*e34*e04;
-        A[8 + 10*11]=-e38*e11*e01-e38*e14*e04-e38*e10*e00-e38*e13*e03-e18*e30*e00-e18*e33*e03-e18*e31*e01-e18*e34*e04-e08*e30*e10-e08*e33*e13-e08*e31*e11-e08*e34*e14+3*e08*e38*e18+e08*e37*e17+e18*e36*e06+e18*e37*e07+e38*e16*e06+e38*e17*e07+e15*e35*e08+e35*e13*e06+e35*e03*e16+e35*e14*e07+e35*e04*e17+e08*e36*e16+e05*e35*e18+e05*e15*e38+e15*e33*e06+e15*e03*e36+e15*e34*e07+e15*e04*e37+e05*e14*e37+e12*e30*e06+e12*e31*e07+e12*e01*e37+e12*e00*e36+e12*e32*e08+e32*e10*e06+e32*e00*e16+e32*e11*e07+e32*e01*e17+e05*e33*e16+e05*e13*e36+e05*e34*e17+e02*e30*e16+e02*e10*e36+e02*e32*e18+e02*e12*e38+e02*e31*e17+e02*e11*e37;
-        A[8 + 10*12]=e12*e30*e16+e12*e10*e36+e12*e32*e18+e12*e31*e17+e12*e11*e37+e32*e10*e16+e32*e11*e17+e15*e33*e16+e15*e13*e36-0.5*e38*e102-0.5*e38*e112-0.5*e38*e132-0.5*e38*e142+0.5*e38*e162+0.5*e38*e172+e15*e34*e17+e15*e14*e37+e15*e35*e18+e35*e13*e16+e35*e14*e17+e18*e36*e16+e18*e37*e17-e18*e30*e10-e18*e33*e13-e18*e31*e11-e18*e34*e14+0.5*e122*e38+0.5*e152*e38+1.5*e38*e182;
-        A[8 + 10*13]=e22*e30*e06-e28*e34*e04+e05*e35*e28+e02*e22*e38+e22*e00*e36+e22*e31*e07+e22*e01*e37+e02*e32*e28+e02*e21*e37-e38*e20*e00-e28*e31*e01-e38*e23*e03-e38*e21*e01-e38*e24*e04-e28*e30*e00-e08*e30*e20-e08*e31*e21-e08*e33*e23-e08*e34*e24-e28*e33*e03+e35*e24*e07+e35*e04*e27+e08*e36*e26+e08*e37*e27+3*e08*e38*e28+e28*e36*e06+e28*e37*e07+e38*e26*e06+e38*e27*e07+e25*e04*e37+e25*e35*e08+e35*e23*e06+e35*e03*e26+e05*e23*e36+e05*e25*e38+e05*e34*e27+e05*e24*e37+e25*e33*e06+e25*e03*e36+e25*e34*e07+e05*e33*e26+e32*e21*e07+e32*e01*e27+e22*e32*e08+e32*e20*e06+e32*e00*e26+e02*e30*e26+e02*e20*e36+e02*e31*e27;
-        A[8 + 10*14]=e35*e13*e26-e38*e21*e11-e38*e24*e14+e35*e24*e17+e35*e14*e27+e18*e36*e26+e18*e37*e27+3*e18*e38*e28+e28*e36*e16+e28*e37*e17+e38*e26*e16+e38*e27*e17-e18*e30*e20-e18*e31*e21-e18*e33*e23-e18*e34*e24-e28*e30*e10-e28*e33*e13-e28*e31*e11-e28*e34*e14-e38*e20*e10-e38*e23*e13+e35*e23*e16+e12*e20*e36+e12*e30*e26+e12*e31*e27+e12*e21*e37+e12*e32*e28+e12*e22*e38+e22*e30*e16+e22*e10*e36+e22*e32*e18+e22*e31*e17+e22*e11*e37+e32*e20*e16+e32*e10*e26+e32*e21*e17+e32*e11*e27+e15*e33*e26+e15*e23*e36+e15*e35*e28+e15*e25*e38+e15*e34*e27+e15*e24*e37+e25*e33*e16+e25*e13*e36+e25*e34*e17+e25*e14*e37+e25*e35*e18;
-        A[8 + 10*15]=-e28*e30*e20+e22*e30*e26+e22*e20*e36+e22*e31*e27+e22*e21*e37+e22*e32*e28+e32*e20*e26+e32*e21*e27+e25*e33*e26+e25*e23*e36+e25*e35*e28+e25*e34*e27+e25*e24*e37+e35*e23*e26+e35*e24*e27+e28*e36*e26+e28*e37*e27-e28*e31*e21-e28*e33*e23-e28*e34*e24-0.5*e38*e242+0.5*e252*e38+1.5*e38*e282+0.5*e38*e262+0.5*e38*e272-0.5*e38*e202-0.5*e38*e212-0.5*e38*e232+0.5*e222*e38;
-        A[8 + 10*16]=-0.5*e08*e312-0.5*e08*e342+0.5*e352*e08+0.5*e08*e362+1.5*e08*e382+0.5*e08*e372-0.5*e08*e302-0.5*e08*e332+e02*e30*e36+e02*e32*e38+e02*e31*e37+e32*e30*e06+e32*e00*e36+e32*e31*e07+e32*e01*e37+e05*e33*e36+e05*e34*e37+e05*e35*e38+e35*e33*e06+e35*e03*e36+e35*e34*e07+e35*e04*e37+0.5*e322*e08+e38*e36*e06+e38*e37*e07-e38*e30*e00-e38*e33*e03-e38*e31*e01-e38*e34*e04;
-        A[8 + 10*17]=-e38*e30*e10+e38*e36*e16+e38*e37*e17-e38*e33*e13-e38*e31*e11-e38*e34*e14+0.5*e18*e362+e12*e30*e36+e12*e32*e38+e12*e31*e37+e32*e30*e16+e32*e10*e36+e32*e31*e17+e32*e11*e37+e15*e33*e36+e15*e34*e37+e15*e35*e38+e35*e33*e16+e35*e13*e36+e35*e34*e17+e35*e14*e37+0.5*e322*e18+0.5*e352*e18+1.5*e18*e382+0.5*e18*e372-0.5*e18*e302-0.5*e18*e332-0.5*e18*e312-0.5*e18*e342;
-        A[8 + 10*18]=-e38*e30*e20+e25*e35*e38+e22*e30*e36+e22*e32*e38+e22*e31*e37+e32*e30*e26+e32*e20*e36+e32*e31*e27+e32*e21*e37+e25*e33*e36+e25*e34*e37+e35*e33*e26+e35*e23*e36+e35*e34*e27+e35*e24*e37+e38*e36*e26+e38*e37*e27-e38*e31*e21-e38*e33*e23-e38*e34*e24-0.5*e28*e332-0.5*e28*e312-0.5*e28*e342+0.5*e322*e28+0.5*e352*e28+0.5*e28*e362+1.5*e28*e382+0.5*e28*e372-0.5*e28*e302;
-        A[8 + 10*19]=e32*e30*e36+0.5*e322*e38+e32*e31*e37+e35*e33*e36+e35*e34*e37+0.5*e352*e38+0.5*e38*e362+0.5*e383+0.5*e38*e372-0.5*e38*e302-0.5*e38*e332-0.5*e38*e312-0.5*e38*e342;
-        A[9 + 10*0]=e00*e04*e08-e00*e05*e07+e03*e02*e07-e03*e01*e08-e06*e02*e04+e06*e01*e05;
-        A[9 + 10*1]=e06*e01*e15-e16*e02*e04+e16*e01*e05+e03*e02*e17-e13*e01*e08+e06*e11*e05+e13*e02*e07+e00*e04*e18+e00*e14*e08-e00*e05*e17-e10*e05*e07-e00*e15*e07-e06*e12*e04-e06*e02*e14-e03*e01*e18-e03*e11*e08+e10*e04*e08+e03*e12*e07;
-        A[9 + 10*2]=-e13*e01*e18-e13*e11*e08+e13*e12*e07+e13*e02*e17+e03*e12*e17-e10*e15*e07+e10*e04*e18+e10*e14*e08-e10*e05*e17-e00*e15*e17+e00*e14*e18+e16*e01*e15+e06*e11*e15-e06*e12*e14-e16*e12*e04-e16*e02*e14+e16*e11*e05-e03*e11*e18;
-        A[9 + 10*3]=e10*e14*e18-e10*e15*e17-e13*e11*e18+e13*e12*e17+e16*e11*e15-e16*e12*e14;
-        A[9 + 10*4]=-e20*e05*e07+e03*e22*e07+e06*e21*e05+e06*e01*e25-e23*e01*e08+e23*e02*e07+e00*e24*e08-e00*e25*e07-e00*e05*e27+e00*e04*e28-e06*e22*e04-e06*e02*e24-e03*e21*e08-e03*e01*e28-e26*e02*e04+e26*e01*e05+e03*e02*e27+e20*e04*e08;
-        A[9 + 10*5]=e23*e12*e07-e26*e02*e14+e16*e21*e05-e23*e11*e08+e10*e24*e08-e20*e05*e17+e26*e11*e05+e26*e01*e15+e10*e04*e28+e00*e24*e18-e00*e15*e27+e03*e22*e17-e13*e01*e28+e23*e02*e17+e16*e01*e25+e20*e04*e18+e06*e11*e25+e13*e02*e27-e23*e01*e18-e20*e15*e07-e10*e25*e07+e13*e22*e07-e06*e22*e14-e26*e12*e04-e03*e11*e28-e03*e21*e18-e16*e22*e04-e16*e02*e24-e06*e12*e24+e06*e21*e15+e00*e14*e28-e00*e25*e17+e20*e14*e08-e13*e21*e08-e10*e05*e27+e03*e12*e27;
-        A[9 + 10*6]=-e13*e11*e28+e13*e12*e27+e13*e22*e17+e16*e11*e25+e10*e14*e28-e13*e21*e18-e23*e11*e18+e23*e12*e17+e20*e14*e18-e20*e15*e17+e26*e11*e15-e10*e15*e27-e10*e25*e17-e16*e22*e14-e16*e12*e24+e16*e21*e15-e26*e12*e14+e10*e24*e18;
-        A[9 + 10*7]=e26*e21*e05+e26*e01*e25+e20*e04*e28+e20*e24*e08-e20*e25*e07+e23*e22*e07+e03*e22*e27-e03*e21*e28-e26*e22*e04-e20*e05*e27-e00*e25*e27+e06*e21*e25-e06*e22*e24+e00*e24*e28-e26*e02*e24-e23*e21*e08-e23*e01*e28+e23*e02*e27;
-        A[9 + 10*8]=-e10*e25*e27+e10*e24*e28-e20*e15*e27-e20*e25*e17+e20*e14*e28+e20*e24*e18+e26*e11*e25+e23*e22*e17-e23*e11*e28+e23*e12*e27-e23*e21*e18-e13*e21*e28+e13*e22*e27-e26*e12*e24+e26*e21*e15-e16*e22*e24+e16*e21*e25-e26*e22*e14;
-        A[9 + 10*9]=-e20*e25*e27+e20*e24*e28-e23*e21*e28+e23*e22*e27-e26*e22*e24+e26*e21*e25;
-        A[9 + 10*10]=e03*e02*e37-e03*e31*e08-e03*e01*e38+e03*e32*e07-e00*e35*e07+e30*e04*e08+e06*e31*e05-e36*e02*e04+e36*e01*e05-e06*e32*e04-e06*e02*e34+e06*e01*e35+e00*e04*e38-e00*e05*e37+e33*e02*e07-e33*e01*e08-e30*e05*e07+e00*e34*e08;
-        A[9 + 10*11]=-e36*e12*e04+e30*e04*e18-e30*e15*e07-e36*e02*e14-e30*e05*e17+e30*e14*e08-e00*e35*e17-e00*e15*e37+e33*e02*e17-e06*e32*e14-e06*e12*e34-e16*e32*e04+e06*e31*e15+e06*e11*e35+e00*e34*e18-e10*e35*e07-e33*e11*e08-e33*e01*e18+e16*e01*e35-e16*e02*e34+e16*e31*e05-e03*e31*e18-e03*e11*e38+e03*e32*e17+e13*e02*e37-e13*e31*e08-e13*e01*e38+e10*e34*e08+e00*e14*e38+e36*e11*e05+e36*e01*e15+e03*e12*e37-e10*e05*e37+e10*e04*e38+e33*e12*e07+e13*e32*e07;
-        A[9 + 10*12]=-e36*e12*e14-e30*e15*e17+e13*e32*e17-e13*e31*e18-e33*e11*e18+e33*e12*e17+e10*e14*e38+e30*e14*e18-e13*e11*e38+e13*e12*e37-e10*e35*e17+e10*e34*e18-e16*e12*e34-e16*e32*e14+e16*e11*e35+e16*e31*e15+e36*e11*e15-e10*e15*e37;
-        A[9 + 10*13]=-e06*e22*e34-e06*e32*e24-e00*e25*e37-e00*e35*e27+e23*e02*e37+e00*e24*e38-e23*e01*e38-e03*e31*e28-e33*e01*e28+e03*e22*e37+e03*e32*e27+e33*e02*e27-e03*e21*e38-e26*e32*e04-e33*e21*e08+e36*e01*e25+e36*e21*e05-e20*e05*e37+e20*e04*e38+e30*e04*e28-e20*e35*e07+e33*e22*e07+e30*e24*e08-e30*e25*e07-e23*e31*e08+e23*e32*e07+e00*e34*e28+e06*e21*e35+e06*e31*e25-e36*e02*e24+e26*e01*e35-e36*e22*e04+e26*e31*e05-e26*e02*e34+e20*e34*e08-e30*e05*e27;
-        A[9 + 10*14]=e33*e22*e17+e33*e12*e27+e16*e21*e35-e16*e22*e34-e16*e32*e24+e23*e32*e17-e23*e11*e38-e23*e31*e18+e23*e12*e37-e13*e21*e38-e13*e31*e28+e13*e22*e37+e36*e21*e15-e36*e12*e24+e36*e11*e25-e26*e12*e34-e20*e35*e17+e20*e14*e38+e20*e34*e18+e30*e24*e18-e30*e15*e27-e30*e25*e17+e30*e14*e28-e33*e21*e18+e10*e34*e28+e10*e24*e38-e10*e35*e27-e10*e25*e37-e20*e15*e37-e26*e32*e14+e26*e11*e35+e26*e31*e15-e36*e22*e14+e13*e32*e27+e16*e31*e25-e33*e11*e28;
-        A[9 + 10*15]=-e20*e35*e27-e20*e25*e37+e20*e34*e28+e20*e24*e38+e30*e24*e28-e30*e25*e27+e23*e32*e27+e23*e22*e37-e23*e31*e28-e23*e21*e38+e33*e22*e27-e26*e22*e34-e26*e32*e24+e26*e21*e35+e26*e31*e25-e36*e22*e24+e36*e21*e25-e33*e21*e28;
-        A[9 + 10*16]=-e33*e01*e38-e03*e31*e38+e00*e34*e38+e33*e32*e07+e03*e32*e37+e06*e31*e35-e00*e35*e37-e36*e32*e04-e06*e32*e34-e36*e02*e34+e36*e01*e35+e36*e31*e05+e30*e04*e38+e30*e34*e08-e33*e31*e08+e33*e02*e37-e30*e05*e37-e30*e35*e07;
-        A[9 + 10*17]=-e33*e31*e18-e33*e11*e38+e10*e34*e38+e30*e14*e38-e10*e35*e37-e30*e15*e37-e13*e31*e38+e13*e32*e37-e30*e35*e17+e33*e12*e37+e30*e34*e18+e33*e32*e17+e16*e31*e35-e16*e32*e34-e36*e12*e34-e36*e32*e14+e36*e11*e35+e36*e31*e15;
-        A[9 + 10*18]=-e20*e35*e37+e20*e34*e38+e30*e24*e38-e30*e35*e27-e30*e25*e37+e30*e34*e28+e23*e32*e37-e23*e31*e38-e33*e21*e38-e33*e31*e28+e33*e22*e37+e33*e32*e27+e26*e31*e35-e26*e32*e34-e36*e22*e34-e36*e32*e24+e36*e21*e35+e36*e31*e25;
-        A[9 + 10*19]=-e33*e31*e38-e30*e35*e37+e36*e31*e35+e33*e32*e37+e30*e34*e38-e36*e32*e34;
+        for (int i = 0; i < 200; i++)
+        {
+            A[i] = AA[i];
+        }
     }
 
+
     void computeError( InputArray _m1, InputArray _m2, InputArray _model, OutputArray _err ) const
     {
         Mat X1 = _m1.getMat(), X2 = _m2.getMat(), model = _model.getMat();
@@ -518,31 +393,31 @@ protected:
             double b = Ex1[1] * Ex1[1];
             double c = Etx2[0] * Etx2[0];
             double d = Etx2[1] * Etx2[1];
-            
+
             err.at<float>(i) = (float)(x2tEx1 * x2tEx1 / (a + b + c + d));
         }
     }
-}; 
+};
 
 }
 
 // Input should be a vector of n 2D points or a Nx2 matrix
 cv::Mat cv::findEssentialMat( InputArray _points1, InputArray _points2, double focal, Point2d pp,
-					          int method, double prob, double threshold, OutputArray _mask) 
+                              int method, double prob, double threshold, OutputArray _mask)
 {
-	Mat points1, points2; 
-	_points1.getMat().convertTo(points1, CV_64F);
-	_points2.getMat().convertTo(points2, CV_64F);
+    Mat points1, points2;
+    _points1.getMat().convertTo(points1, CV_64F);
+    _points2.getMat().convertTo(points2, CV_64F);
 
-	int npoints = points1.checkVector(2);
+    int npoints = points1.checkVector(2);
     CV_Assert( npoints >= 5 && points2.checkVector(2) == npoints &&
-				              points1.type() == points2.type());
+                              points1.type() == points2.type());
 
-	if( points1.channels() > 1 )
-	{
-		points1 = points1.reshape(1, npoints); 
-		points2 = points2.reshape(1, npoints); 
-	}
+    if( points1.channels() > 1 )
+    {
+        points1 = points1.reshape(1, npoints);
+        points2 = points2.reshape(1, npoints);
+    }
 
     double ifocal = focal != 0 ? 1./focal : 1.;
     for( int i = 0; i < npoints; i++ )
@@ -553,177 +428,187 @@ cv::Mat cv::findEssentialMat( InputArray _points1, InputArray _points2, double f
         points2.at<double>(i, 1) = (points2.at<double>(i, 1) - pp.y)*ifocal;
     }
 
-	// Reshape data to fit opencv ransac function
-	points1 = points1.reshape(2, npoints);
-	points2 = points2.reshape(2, npoints);
+    // Reshape data to fit opencv ransac function
+    points1 = points1.reshape(2, npoints);
+    points2 = points2.reshape(2, npoints);
 
-	threshold /= focal;
+    threshold /= focal;
 
     Mat E;
-    if( method == CV_RANSAC )
-        createRANSACPointSetRegistrator(new EMEstimatorCallback, 5, threshold, prob)->run(points1, points2, E, _mask);
+    if( method == RANSAC )
+        createRANSACPointSetRegistrator(makePtr<EMEstimatorCallback>(), 5, threshold, prob)->run(points1, points2, E, _mask);
     else
-        createLMeDSPointSetRegistrator(new EMEstimatorCallback, 5, prob)->run(points1, points2, E, _mask);
+        createLMeDSPointSetRegistrator(makePtr<EMEstimatorCallback>(), 5, prob)->run(points1, points2, E, _mask);
 
-	return E; 
+    return E;
 }
 
 int cv::recoverPose( InputArray E, InputArray _points1, InputArray _points2, OutputArray _R,
-                     OutputArray _t, double focal, Point2d pp, InputOutputArray _mask) 
+                     OutputArray _t, double focal, Point2d pp, InputOutputArray _mask)
 {
-	Mat points1, points2; 
-	_points1.getMat().copyTo(points1); 
-	_points2.getMat().copyTo(points2); 
-    
-	int npoints = points1.checkVector(2);
+    Mat points1, points2;
+    _points1.getMat().copyTo(points1);
+    _points2.getMat().copyTo(points2);
+
+    int npoints = points1.checkVector(2);
     CV_Assert( npoints >= 0 && points2.checkVector(2) == npoints &&
-				              points1.type() == points2.type());
+                              points1.type() == points2.type());
 
-	if (points1.channels() > 1)
-	{
-		points1 = points1.reshape(1, npoints); 
-		points2 = points2.reshape(1, npoints); 
-	}
-	points1.convertTo(points1, CV_64F); 
-	points2.convertTo(points2, CV_64F); 
+    if (points1.channels() > 1)
+    {
+        points1 = points1.reshape(1, npoints);
+        points2 = points2.reshape(1, npoints);
+    }
+    points1.convertTo(points1, CV_64F);
+    points2.convertTo(points2, CV_64F);
 
-	points1.col(0) = (points1.col(0) - pp.x) / focal; 
-	points2.col(0) = (points2.col(0) - pp.x) / focal; 
-	points1.col(1) = (points1.col(1) - pp.y) / focal; 
-	points2.col(1) = (points2.col(1) - pp.y) / focal; 
+    points1.col(0) = (points1.col(0) - pp.x) / focal;
+    points2.col(0) = (points2.col(0) - pp.x) / focal;
+    points1.col(1) = (points1.col(1) - pp.y) / focal;
+    points2.col(1) = (points2.col(1) - pp.y) / focal;
 
-	points1 = points1.t(); 
-	points2 = points2.t(); 
-	
-	Mat R1, R2, t; 
-	decomposeEssentialMat(E, R1, R2, t); 
-	Mat P0 = Mat::eye(3, 4, R1.type()); 
-	Mat P1(3, 4, R1.type()), P2(3, 4, R1.type()), P3(3, 4, R1.type()), P4(3, 4, R1.type()); 
-	P1(Range::all(), Range(0, 3)) = R1 * 1.0; P1.col(3) = t * 1.0; 
-	P2(Range::all(), Range(0, 3)) = R2 * 1.0; P2.col(3) = t * 1.0; 
-	P3(Range::all(), Range(0, 3)) = R1 * 1.0; P3.col(3) = -t * 1.0; 
-	P4(Range::all(), Range(0, 3)) = R2 * 1.0; P4.col(3) = -t * 1.0; 
+    points1 = points1.t();
+    points2 = points2.t();
 
-	// Do the cheirality check. 
-	// Notice here a threshold dist is used to filter
-	// out far away points (i.e. infinite points) since 
-	// there depth may vary between postive and negtive. 
-	double dist = 50.0; 
-	Mat Q; 
-	triangulatePoints(P0, P1, points1, points2, Q); 
-	Mat mask1 = Q.row(2).mul(Q.row(3)) > 0; 
-	Q.row(0) /= Q.row(3); 
-	Q.row(1) /= Q.row(3); 
-	Q.row(2) /= Q.row(3); 
-	Q.row(3) /= Q.row(3); 
-	mask1 = (Q.row(2) < dist) & mask1; 
-	Q = P1 * Q; 
-	mask1 = (Q.row(2) > 0) & mask1; 
-	mask1 = (Q.row(2) < dist) & mask1; 
+    Mat R1, R2, t;
+    decomposeEssentialMat(E, R1, R2, t);
+    Mat P0 = Mat::eye(3, 4, R1.type());
+    Mat P1(3, 4, R1.type()), P2(3, 4, R1.type()), P3(3, 4, R1.type()), P4(3, 4, R1.type());
+    P1(Range::all(), Range(0, 3)) = R1 * 1.0; P1.col(3) = t * 1.0;
+    P2(Range::all(), Range(0, 3)) = R2 * 1.0; P2.col(3) = t * 1.0;
+    P3(Range::all(), Range(0, 3)) = R1 * 1.0; P3.col(3) = -t * 1.0;
+    P4(Range::all(), Range(0, 3)) = R2 * 1.0; P4.col(3) = -t * 1.0;
 
-	triangulatePoints(P0, P2, points1, points2, Q); 
-	Mat mask2 = Q.row(2).mul(Q.row(3)) > 0; 
-	Q.row(0) /= Q.row(3); 
-	Q.row(1) /= Q.row(3); 
-	Q.row(2) /= Q.row(3); 
-	Q.row(3) /= Q.row(3); 
-	mask2 = (Q.row(2) < dist) & mask2; 
-	Q = P2 * Q; 
-	mask2 = (Q.row(2) > 0) & mask2; 
-	mask2 = (Q.row(2) < dist) & mask2; 
+    // Do the cheirality check.
+    // Notice here a threshold dist is used to filter
+    // out far away points (i.e. infinite points) since
+    // there depth may vary between postive and negtive.
+    double dist = 50.0;
+    Mat Q;
+    triangulatePoints(P0, P1, points1, points2, Q);
+    Mat mask1 = Q.row(2).mul(Q.row(3)) > 0;
+    Q.row(0) /= Q.row(3);
+    Q.row(1) /= Q.row(3);
+    Q.row(2) /= Q.row(3);
+    Q.row(3) /= Q.row(3);
+    mask1 = (Q.row(2) < dist) & mask1;
+    Q = P1 * Q;
+    mask1 = (Q.row(2) > 0) & mask1;
+    mask1 = (Q.row(2) < dist) & mask1;
 
-	triangulatePoints(P0, P3, points1, points2, Q); 
-	Mat mask3 = Q.row(2).mul(Q.row(3)) > 0; 
-	Q.row(0) /= Q.row(3); 
-	Q.row(1) /= Q.row(3); 
-	Q.row(2) /= Q.row(3); 
-	Q.row(3) /= Q.row(3); 
-	mask3 = (Q.row(2) < dist) & mask3; 
-	Q = P3 * Q; 
-	mask3 = (Q.row(2) > 0) & mask3; 
-	mask3 = (Q.row(2) < dist) & mask3; 
+    triangulatePoints(P0, P2, points1, points2, Q);
+    Mat mask2 = Q.row(2).mul(Q.row(3)) > 0;
+    Q.row(0) /= Q.row(3);
+    Q.row(1) /= Q.row(3);
+    Q.row(2) /= Q.row(3);
+    Q.row(3) /= Q.row(3);
+    mask2 = (Q.row(2) < dist) & mask2;
+    Q = P2 * Q;
+    mask2 = (Q.row(2) > 0) & mask2;
+    mask2 = (Q.row(2) < dist) & mask2;
 
-	triangulatePoints(P0, P4, points1, points2, Q); 
-	Mat mask4 = Q.row(2).mul(Q.row(3)) > 0; 
-	Q.row(0) /= Q.row(3); 
-	Q.row(1) /= Q.row(3); 
-	Q.row(2) /= Q.row(3); 
-	Q.row(3) /= Q.row(3); 
-	mask4 = (Q.row(2) < dist) & mask4; 
-	Q = P4 * Q; 
-	mask4 = (Q.row(2) > 0) & mask4; 
-	mask4 = (Q.row(2) < dist) & mask4; 
+    triangulatePoints(P0, P3, points1, points2, Q);
+    Mat mask3 = Q.row(2).mul(Q.row(3)) > 0;
+    Q.row(0) /= Q.row(3);
+    Q.row(1) /= Q.row(3);
+    Q.row(2) /= Q.row(3);
+    Q.row(3) /= Q.row(3);
+    mask3 = (Q.row(2) < dist) & mask3;
+    Q = P3 * Q;
+    mask3 = (Q.row(2) > 0) & mask3;
+    mask3 = (Q.row(2) < dist) & mask3;
 
-	// If _mask is given, then use it to filter outliers. 
-	if (_mask.needed())
-	{
-		_mask.create(1, npoints, CV_8U, -1, true); 
-		Mat mask = _mask.getMat(); 
-		bitwise_and(mask, mask1, mask1); 
-		bitwise_and(mask, mask2, mask2); 
-		bitwise_and(mask, mask3, mask3); 
-		bitwise_and(mask, mask4, mask4); 
-	}
+    triangulatePoints(P0, P4, points1, points2, Q);
+    Mat mask4 = Q.row(2).mul(Q.row(3)) > 0;
+    Q.row(0) /= Q.row(3);
+    Q.row(1) /= Q.row(3);
+    Q.row(2) /= Q.row(3);
+    Q.row(3) /= Q.row(3);
+    mask4 = (Q.row(2) < dist) & mask4;
+    Q = P4 * Q;
+    mask4 = (Q.row(2) > 0) & mask4;
+    mask4 = (Q.row(2) < dist) & mask4;
 
-    CV_Assert(_R.needed() && _t.needed()); 
-    _R.create(3, 3, R1.type()); 
-    _t.create(3, 1, t.type()); 
+    mask1 = mask1.t();
+    mask2 = mask2.t();
+    mask3 = mask3.t();
+    mask4 = mask4.t();
 
-	int good1 = countNonZero(mask1); 
-	int good2 = countNonZero(mask2); 
-	int good3 = countNonZero(mask3); 
-	int good4 = countNonZero(mask4); 
-	if (good1 >= good2 && good1 >= good3 && good1 >= good4)
-	{
-        R1.copyTo(_R); 
-        t.copyTo(_t);
-		if (_mask.needed()) mask1.copyTo(_mask);
-		return good1; 
-	}
-	else if (good2 >= good1 && good2 >= good3 && good2 >= good4)
-	{
-        R2.copyTo(_R);
-        t.copyTo(_t);
-		if (_mask.needed()) mask2.copyTo(_mask); 
-		return good2; 
-	}
-	else if (good3 >= good1 && good3 >= good2 && good3 >= good4)
-	{
-        t = -t; 
+    // If _mask is given, then use it to filter outliers.
+    if (!_mask.empty())
+    {
+        Mat mask = _mask.getMat();
+        CV_Assert(mask.size() == mask1.size());
+        bitwise_and(mask, mask1, mask1);
+        bitwise_and(mask, mask2, mask2);
+        bitwise_and(mask, mask3, mask3);
+        bitwise_and(mask, mask4, mask4);
+    }
+    if (_mask.empty() && _mask.needed())
+    {
+        _mask.create(mask1.size(), CV_8U);
+    }
+
+    CV_Assert(_R.needed() && _t.needed());
+    _R.create(3, 3, R1.type());
+    _t.create(3, 1, t.type());
+
+    int good1 = countNonZero(mask1);
+    int good2 = countNonZero(mask2);
+    int good3 = countNonZero(mask3);
+    int good4 = countNonZero(mask4);
+
+    if (good1 >= good2 && good1 >= good3 && good1 >= good4)
+    {
         R1.copyTo(_R);
-        t.copyTo(_t); 
-		if (_mask.needed()) mask3.copyTo(_mask); 
-		return good3; 
-	}
-	else  
-	{
-        t = -t; 
+        t.copyTo(_t);
+        if (_mask.needed()) mask1.copyTo(_mask);
+        return good1;
+    }
+    else if (good2 >= good1 && good2 >= good3 && good2 >= good4)
+    {
         R2.copyTo(_R);
         t.copyTo(_t);
-		if (_mask.needed()) mask4.copyTo(_mask);
-		return good4; 
-	}
+        if (_mask.needed()) mask2.copyTo(_mask);
+        return good2;
+    }
+    else if (good3 >= good1 && good3 >= good2 && good3 >= good4)
+    {
+        t = -t;
+        R1.copyTo(_R);
+        t.copyTo(_t);
+        if (_mask.needed()) mask3.copyTo(_mask);
+        return good3;
+    }
+    else
+    {
+        t = -t;
+        R2.copyTo(_R);
+        t.copyTo(_t);
+        if (_mask.needed()) mask4.copyTo(_mask);
+        return good4;
+    }
 }
 
 
-void cv::decomposeEssentialMat( InputArray _E, OutputArray _R1, OutputArray _R2, OutputArray _t ) 
+void cv::decomposeEssentialMat( InputArray _E, OutputArray _R1, OutputArray _R2, OutputArray _t )
 {
     Mat E = _E.getMat().reshape(1, 3);
-	CV_Assert(E.cols == 3 && E.rows == 3);
-	
-    Mat D, U, Vt; 
-	SVD::compute(E, D, U, Vt); 
-	
-    if (determinant(U) < 0) U *= -1.;
-	if (determinant(Vt) < 0) Vt *= -1.;
+    CV_Assert(E.cols == 3 && E.rows == 3);
 
-    Mat W = (Mat_<double>(3, 3) << 0, 1, 0, -1, 0, 0, 0, 0, 1); 
-	W.convertTo(W, E.type()); 
-	
-    Mat R1, R2, t; 
-    R1 = U * W * Vt; 
-	R2 = U * W.t() * Vt; 
-	t = U.col(2) * 1.0; 
+    Mat D, U, Vt;
+    SVD::compute(E, D, U, Vt);
+
+    if (determinant(U) < 0) U *= -1.;
+    if (determinant(Vt) < 0) Vt *= -1.;
+
+    Mat W = (Mat_<double>(3, 3) << 0, 1, 0, -1, 0, 0, 0, 0, 1);
+    W.convertTo(W, E.type());
+
+    Mat R1, R2, t;
+    R1 = U * W * Vt;
+    R2 = U * W.t() * Vt;
+    t = U.col(2) * 1.0;
 
     R1.copyTo(_R1);
     R2.copyTo(_R2);
diff --git a/modules/calib3d/src/fundam.cpp b/modules/calib3d/src/fundam.cpp
index 0db8fcab9b..d1c6e8cd0b 100644
--- a/modules/calib3d/src/fundam.cpp
+++ b/modules/calib3d/src/fundam.cpp
@@ -181,12 +181,12 @@ public:
                     LtL[j][k] += Lx[j]*Lx[k] + Ly[j]*Ly[k];
         }
         completeSymm( _LtL );
-        
+
         eigen( _LtL, matW, matV );
         _Htemp = _invHnorm*_H0;
         _H0 = _Htemp*_Hnorm2;
         _H0.convertTo(_model, _H0.type(), 1./_H0.at<double>(2,2) );
-        
+
         return 1;
     }
 
@@ -259,6 +259,8 @@ public:
                 Jptr[8] = Jptr[9] = Jptr[10] = 0.;
                 Jptr[11] = Mx*ww; Jptr[12] = My*ww; Jptr[13] = ww;
                 Jptr[14] = -Mx*ww*yi; Jptr[15] = -My*ww*yi;
+
+                Jptr += 16;
             }
         }
 
@@ -292,7 +294,7 @@ cv::Mat cv::findHomography( InputArray _points1, InputArray _points2,
         {
             npoints = p.checkVector(3, -1, false);
             if( npoints < 0 )
-                CV_Error(CV_StsBadArg, "The input arrays should be 2D or 3D point sets");
+                CV_Error(Error::StsBadArg, "The input arrays should be 2D or 3D point sets");
             if( npoints == 0 )
                 return Mat();
             convertPointsFromHomogeneous(p, p);
@@ -305,7 +307,7 @@ cv::Mat cv::findHomography( InputArray _points1, InputArray _points2,
     if( ransacReprojThreshold <= 0 )
         ransacReprojThreshold = defaultRANSACReprojThreshold;
 
-    Ptr<PointSetRegistrator::Callback> cb = new HomographyEstimatorCallback;
+    Ptr<PointSetRegistrator::Callback> cb = makePtr<HomographyEstimatorCallback>();
 
     if( method == 0 || npoints == 4 )
     {
@@ -317,7 +319,7 @@ cv::Mat cv::findHomography( InputArray _points1, InputArray _points2,
     else if( method == LMEDS )
         result = createLMeDSPointSetRegistrator(cb, 4, confidence, maxIters)->run(src, dst, H, tempMask);
     else
-        CV_Error(CV_StsBadArg, "Unknown estimation method");
+        CV_Error(Error::StsBadArg, "Unknown estimation method");
 
     if( result && npoints > 4 )
     {
@@ -332,7 +334,7 @@ cv::Mat cv::findHomography( InputArray _points1, InputArray _points2,
             if( method == RANSAC || method == LMEDS )
                 cb->runKernel( src, dst, H );
             Mat H8(8, 1, CV_64F, H.ptr<double>());
-            createLMSolver(new HomographyRefineCallback(src, dst), 10)->run(H8);
+            createLMSolver(makePtr<HomographyRefineCallback>(src, dst), 10)->run(H8);
         }
     }
 
@@ -475,7 +477,7 @@ static int run7Point( const Mat& _m1, const Mat& _m2, Mat& _fmatrix )
     return n;
 }
 
-    
+
 static int run8Point( const Mat& _m1, const Mat& _m2, Mat& _fmatrix )
 {
     double a[9*9], w[9], v[9*9];
@@ -585,11 +587,11 @@ static int run8Point( const Mat& _m1, const Mat& _m2, Mat& _fmatrix )
     gemm( T2, F0, 1., 0, 0., TF, GEMM_1_T );
     F0 = Mat(3, 3, CV_64F, fmatrix);
     gemm( TF, T1, 1., 0, 0., F0, 0 );
-    
+
     // make F(3,3) = 1
     if( fabs(F0.at<double>(2,2)) > FLT_EPSILON )
         F0 *= 1./F0.at<double>(2,2);
-    
+
     return 1;
 }
 
@@ -671,7 +673,7 @@ cv::Mat cv::findFundamentalMat( InputArray _points1, InputArray _points2,
         {
             npoints = p.checkVector(3, -1, false);
             if( npoints < 0 )
-                CV_Error(CV_StsBadArg, "The input arrays should be 2D or 3D point sets");
+                CV_Error(Error::StsBadArg, "The input arrays should be 2D or 3D point sets");
             if( npoints == 0 )
                 return Mat();
             convertPointsFromHomogeneous(p, p);
@@ -684,7 +686,7 @@ cv::Mat cv::findFundamentalMat( InputArray _points1, InputArray _points2,
     if( npoints < 7 )
         return Mat();
 
-    Ptr<PointSetRegistrator::Callback> cb = new FMEstimatorCallback;
+    Ptr<PointSetRegistrator::Callback> cb = makePtr<FMEstimatorCallback>();
     int result;
 
     if( npoints == 7 || method == FM_8POINT )
@@ -739,7 +741,7 @@ void cv::computeCorrespondEpilines( InputArray _points, int whichImage,
     {
         npoints = points.checkVector(3);
         if( npoints < 0 )
-            CV_Error( CV_StsBadArg, "The input should be a 2D or 3D point set");
+            CV_Error( Error::StsBadArg, "The input should be a 2D or 3D point set");
         Mat temp;
         convertPointsFromHomogeneous(points, temp);
         points = temp;
@@ -893,7 +895,7 @@ void cv::convertPointsFromHomogeneous( InputArray _src, OutputArray _dst )
         }
     }
     else
-        CV_Error(CV_StsUnsupportedFormat, "");
+        CV_Error(Error::StsUnsupportedFormat, "");
 }
 
 
@@ -974,7 +976,7 @@ void cv::convertPointsToHomogeneous( InputArray _src, OutputArray _dst )
         }
     }
     else
-        CV_Error(CV_StsUnsupportedFormat, "");
+        CV_Error(Error::StsUnsupportedFormat, "");
 }
 
 
diff --git a/modules/calib3d/src/levmarq.cpp b/modules/calib3d/src/levmarq.cpp
index 539c804e2b..55704132cf 100644
--- a/modules/calib3d/src/levmarq.cpp
+++ b/modules/calib3d/src/levmarq.cpp
@@ -47,30 +47,30 @@
    This is translation to C++ of the Matlab's LMSolve package by Miroslav Balda.
    Here is the original copyright:
    ============================================================================
-   
+
    Copyright (c) 2007, Miroslav Balda
    All rights reserved.
 
-   Redistribution and use in source and binary forms, with or without 
-   modification, are permitted provided that the following conditions are 
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
    met:
 
-       * Redistributions of source code must retain the above copyright 
+       * Redistributions of source code must retain the above copyright
          notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above copyright 
-         notice, this list of conditions and the following disclaimer in 
+       * Redistributions in binary form must reproduce the above copyright
+         notice, this list of conditions and the following disclaimer in
          the documentation and/or other materials provided with the distribution
 
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
-   AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
-   IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
-   ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
-   LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
-   CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
-   SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 
-   INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 
-   CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 
-   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+   AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+   IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+   ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+   LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+   CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+   SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+   INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+   CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
    POSSIBILITY OF SUCH DAMAGE.
 */
 
@@ -95,7 +95,7 @@ public:
         int ptype = param0.type();
 
         CV_Assert( (param0.cols == 1 || param0.rows == 1) && (ptype == CV_32F || ptype == CV_64F));
-        CV_Assert( !cb.empty() );
+        CV_Assert( cb );
 
         int lx = param0.rows + param0.cols - 1;
         param0.convertTo(x, CV_64F);
@@ -112,7 +112,7 @@ public:
         gemm(J, r, 1, noArray(), 0, v, GEMM_1_T);
 
         Mat D = A.diag().clone();
-        
+
         const double Rlo = 0.25, Rhi = 0.75;
         double lambda = 1, lc = 0.75;
         int i, iter = 0;
@@ -220,7 +220,7 @@ CV_INIT_ALGORITHM(LMSolverImpl, "LMSolver",
 Ptr<LMSolver> createLMSolver(const Ptr<LMSolver::Callback>& cb, int maxIters)
 {
     CV_Assert( !LMSolverImpl_info_auto.name().empty() );
-    return new LMSolverImpl(cb, maxIters);
+    return makePtr<LMSolverImpl>(cb, maxIters);
 }
-    
+
 }
diff --git a/modules/calib3d/src/p3p.cpp b/modules/calib3d/src/p3p.cpp
index 92e7954721..882868d6b8 100644
--- a/modules/calib3d/src/p3p.cpp
+++ b/modules/calib3d/src/p3p.cpp
@@ -411,4 +411,3 @@ bool p3p::jacobi_4x4(double * A, double * D, double * U)
 
     return false;
 }
-
diff --git a/modules/calib3d/src/p3p.h b/modules/calib3d/src/p3p.h
index a58b0fa149..57f8d7df88 100644
--- a/modules/calib3d/src/p3p.h
+++ b/modules/calib3d/src/p3p.h
@@ -59,4 +59,3 @@ class p3p
 };
 
 #endif // P3P_H
-
diff --git a/modules/calib3d/src/posit.cpp b/modules/calib3d/src/posit.cpp
index c8c1c2c2f1..14c33e1e7b 100644
--- a/modules/calib3d/src/posit.cpp
+++ b/modules/calib3d/src/posit.cpp
@@ -39,6 +39,7 @@
 //
 //M*/
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 /* POSIT structure */
 struct CvPOSITObject
diff --git a/modules/calib3d/src/precomp.cpp b/modules/calib3d/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/calib3d/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/calib3d/src/precomp.hpp b/modules/calib3d/src/precomp.hpp
index 5951ea1fd2..12885f37a7 100644
--- a/modules/calib3d/src/precomp.hpp
+++ b/modules/calib3d/src/precomp.hpp
@@ -42,16 +42,12 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/calib3d.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
 #include "opencv2/features2d.hpp"
-#include <vector>
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/core/private.hpp"
 
 #ifdef HAVE_TEGRA_OPTIMIZATION
 #include "opencv2/calib3d/calib3d_tegra.hpp"
diff --git a/modules/calib3d/src/ptsetreg.cpp b/modules/calib3d/src/ptsetreg.cpp
index 58cad5cf38..aa361a9118 100644
--- a/modules/calib3d/src/ptsetreg.cpp
+++ b/modules/calib3d/src/ptsetreg.cpp
@@ -53,7 +53,7 @@ namespace cv
 int RANSACUpdateNumIters( double p, double ep, int modelPoints, int maxIters )
 {
     if( modelPoints <= 0 )
-        CV_Error( CV_StsOutOfRange, "the number of model points should be positive" );
+        CV_Error( Error::StsOutOfRange, "the number of model points should be positive" );
 
     p = MAX(p, 0.);
     p = MIN(p, 1.);
@@ -154,7 +154,7 @@ public:
                 continue;
             break;
         }
-        
+
         return i == modelPoints && iters < maxAttempts;
     }
 
@@ -171,7 +171,7 @@ public:
 
         RNG rng((uint64)-1);
 
-        CV_Assert( !cb.empty() );
+        CV_Assert( cb );
         CV_Assert( confidence > 0 && confidence < 1 );
 
         CV_Assert( count >= 0 && count2 == count );
@@ -235,7 +235,7 @@ public:
                 }
             }
         }
-        
+
         if( maxGoodCount > 0 )
         {
             if( bestMask.data != bestMask0.data )
@@ -250,7 +250,7 @@ public:
         }
         else
             _model.release();
-        
+
         return result;
     }
 
@@ -267,9 +267,6 @@ public:
     int maxIters;
 };
 
-
-static CV_IMPLEMENT_QSORT( sortDistances, int, CV_LT )
-
 class LMeDSPointSetRegistrator : public RANSACPointSetRegistrator
 {
 public:
@@ -291,7 +288,7 @@ public:
 
         RNG rng((uint64)-1);
 
-        CV_Assert( !cb.empty() );
+        CV_Assert( cb );
         CV_Assert( confidence > 0 && confidence < 1 );
 
         CV_Assert( count >= 0 && count2 == count );
@@ -347,7 +344,7 @@ public:
                 else
                     errf = err;
                 CV_Assert( errf.isContinuous() && errf.type() == CV_32F && (int)errf.total() == count );
-                sortDistances( (int*)errf.data, count, 0 );
+                std::sort((int*)errf.data, (int*)errf.data + count);
 
                 double median = count % 2 != 0 ?
                 errf.at<float>(count/2) : (errf.at<float>(count/2-1) + errf.at<float>(count/2))*0.5;
@@ -359,7 +356,7 @@ public:
                 }
             }
         }
-        
+
         if( minMedian < DBL_MAX )
         {
             sigma = 2.5*1.4826*(1 + 5./(count - modelPoints))*std::sqrt(minMedian);
@@ -378,7 +375,7 @@ public:
         }
         else
             _model.release();
-        
+
         return result;
     }
 
@@ -400,7 +397,8 @@ Ptr<PointSetRegistrator> createRANSACPointSetRegistrator(const Ptr<PointSetRegis
                                                          double _confidence, int _maxIters)
 {
     CV_Assert( !RANSACPointSetRegistrator_info_auto.name().empty() );
-    return new RANSACPointSetRegistrator(_cb, _modelPoints, _threshold, _confidence, _maxIters);
+    return Ptr<PointSetRegistrator>(
+        new RANSACPointSetRegistrator(_cb, _modelPoints, _threshold, _confidence, _maxIters));
 }
 
 
@@ -408,7 +406,8 @@ Ptr<PointSetRegistrator> createLMeDSPointSetRegistrator(const Ptr<PointSetRegist
                              int _modelPoints, double _confidence, int _maxIters)
 {
     CV_Assert( !LMeDSPointSetRegistrator_info_auto.name().empty() );
-    return new LMeDSPointSetRegistrator(_cb, _modelPoints, _confidence, _maxIters);
+    return Ptr<PointSetRegistrator>(
+        new LMeDSPointSetRegistrator(_cb, _modelPoints, _confidence, _maxIters));
 }
 
 class Affine3DEstimatorCallback : public PointSetRegistrator::Callback
@@ -534,7 +533,6 @@ int cv::estimateAffine3D(InputArray _from, InputArray _to,
     const double epsilon = DBL_EPSILON;
     param1 = param1 <= 0 ? 3 : param1;
     param2 = (param2 < epsilon) ? 0.99 : (param2 > 1 - epsilon) ? 0.99 : param2;
-    
-    return createRANSACPointSetRegistrator(new Affine3DEstimatorCallback, 4, param1, param2)->run(dFrom, dTo, _out, _inliers);
-}
 
+    return createRANSACPointSetRegistrator(makePtr<Affine3DEstimatorCallback>(), 4, param1, param2)->run(dFrom, dTo, _out, _inliers);
+}
diff --git a/modules/calib3d/src/quadsubpix.cpp b/modules/calib3d/src/quadsubpix.cpp
index 26f26b5807..2e98a462b5 100644
--- a/modules/calib3d/src/quadsubpix.cpp
+++ b/modules/calib3d/src/quadsubpix.cpp
@@ -47,40 +47,8 @@
 
 #include <math.h>
 
-//#define _SUBPIX_VERBOSE
-
-#undef max
-
 namespace cv {
 
-
-// static void drawCircles(Mat& img, const std::vector<Point2f>& corners, const std::vector<float>& radius)
-// {
-//     for(size_t i = 0; i < corners.size(); i++)
-//     {
-//         circle(img, corners[i], cvRound(radius[i]), CV_RGB(255, 0, 0));
-//     }
-// }
-
-// static int histQuantile(const Mat& hist, float quantile)
-// {
-//     if(hist.dims > 1) return -1; // works for 1D histograms only
-
-//     float cur_sum = 0;
-//     float total_sum = (float)sum(hist).val[0];
-//     float quantile_sum = total_sum*quantile;
-//     for(int j = 0; j < hist.size[0]; j++)
-//     {
-//         cur_sum += (float)hist.at<float>(j);
-//         if(cur_sum > quantile_sum)
-//         {
-//             return j;
-//         }
-//     }
-
-//     return hist.size[0] - 1;
-// }
-
 inline bool is_smaller(const std::pair<int, float>& p1, const std::pair<int, float>& p2)
 {
     return p1.second < p2.second;
@@ -124,29 +92,6 @@ static void findLinesCrossPoint(Point2f origin1, Point2f dir1, Point2f origin2,
     cross_point = origin1 + dir1*alpha;
 }
 
-// static void findCorner(const std::vector<Point>& contour, Point2f point, Point2f& corner)
-// {
-//     // find the nearest point
-//     double min_dist = std::numeric_limits<double>::max();
-//     int min_idx = -1;
-
-//     // find corner idx
-//     for(size_t i = 0; i < contour.size(); i++)
-//     {
-//         double dist = norm(Point2f((float)contour[i].x, (float)contour[i].y) - point);
-//         if(dist < min_dist)
-//         {
-//             min_dist = dist;
-//             min_idx = (int)i;
-//         }
-//     }
-//     assert(min_idx >= 0);
-
-//     // temporary solution, have to make something more precise
-//     corner = contour[min_idx];
-//     return;
-// }
-
 static void findCorner(const std::vector<Point2f>& contour, Point2f point, Point2f& corner)
 {
     // find the nearest point
@@ -163,7 +108,7 @@ static void findCorner(const std::vector<Point2f>& contour, Point2f point, Point
             min_idx = (int)i;
         }
     }
-    assert(min_idx >= 0);
+    CV_Assert(min_idx >= 0);
 
     // temporary solution, have to make something more precise
     corner = contour[min_idx];
@@ -173,13 +118,7 @@ static void findCorner(const std::vector<Point2f>& contour, Point2f point, Point
 static int segment_hist_max(const Mat& hist, int& low_thresh, int& high_thresh)
 {
     Mat bw;
-    //const double max_bell_width = 20; // we expect two bells with width bounded above
-    //const double min_bell_width = 5; // and below
-
     double total_sum = sum(hist).val[0];
-    //double thresh = total_sum/(2*max_bell_width)*0.25f; // quarter of a bar inside a bell
-
-//    threshold(hist, bw, thresh, 255.0, CV_THRESH_BINARY);
 
     double quantile_sum = 0.0;
     //double min_quantile = 0.2;
@@ -233,12 +172,6 @@ bool cv::find4QuadCornerSubpix(InputArray _img, InputOutputArray _corners, Size
     const float* _ranges = ranges;
     Mat hist;
 
-#if defined(_SUBPIX_VERBOSE)
-    std::vector<float> radius;
-    radius.assign(corners.size(), 0.0f);
-#endif //_SUBPIX_VERBOSE
-
-
     Mat black_comp, white_comp;
     for(int i = 0; i < ncorners; i++)
     {
@@ -248,39 +181,20 @@ bool cv::find4QuadCornerSubpix(InputArray _img, InputOutputArray _corners, Size
         Mat img_roi = img(roi);
         calcHist(&img_roi, 1, &channels, Mat(), hist, 1, &nbins, &_ranges);
 
-#if 0
-        int black_thresh = histQuantile(hist, 0.45f);
-        int white_thresh = histQuantile(hist, 0.55f);
-#else
         int black_thresh = 0, white_thresh = 0;
         segment_hist_max(hist, black_thresh, white_thresh);
-#endif
 
-        threshold(img, black_comp, black_thresh, 255.0, CV_THRESH_BINARY_INV);
-        threshold(img, white_comp, white_thresh, 255.0, CV_THRESH_BINARY);
+        threshold(img, black_comp, black_thresh, 255.0, THRESH_BINARY_INV);
+        threshold(img, white_comp, white_thresh, 255.0, THRESH_BINARY);
 
         const int erode_count = 1;
         erode(black_comp, black_comp, Mat(), Point(-1, -1), erode_count);
         erode(white_comp, white_comp, Mat(), Point(-1, -1), erode_count);
 
-#if defined(_SUBPIX_VERBOSE)
-        namedWindow("roi", 1);
-        imshow("roi", img_roi);
-        imwrite("test.jpg", img);
-        namedWindow("black", 1);
-        imshow("black", black_comp);
-        namedWindow("white", 1);
-        imshow("white", white_comp);
-        cvWaitKey(0);
-        imwrite("black.jpg", black_comp);
-        imwrite("white.jpg", white_comp);
-#endif
-
-
         std::vector<std::vector<Point> > white_contours, black_contours;
         std::vector<Vec4i> white_hierarchy, black_hierarchy;
-        findContours(black_comp, black_contours, black_hierarchy, CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE);
-        findContours(white_comp, white_contours, white_hierarchy, CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE);
+        findContours(black_comp, black_contours, black_hierarchy, RETR_LIST, CHAIN_APPROX_SIMPLE);
+        findContours(white_comp, white_contours, white_hierarchy, RETR_LIST, CHAIN_APPROX_SIMPLE);
 
         if(black_contours.size() < 5 || white_contours.size() < 5) continue;
 
@@ -302,15 +216,11 @@ bool cv::find4QuadCornerSubpix(InputArray _img, InputOutputArray _corners, Size
         Point2f quad_corners[4];
         for(int k = 0; k < 4; k++)
         {
-#if 1
             std::vector<Point2f> temp;
             for(size_t j = 0; j < quads[k]->size(); j++) temp.push_back((*quads[k])[j]);
             approxPolyDP(Mat(temp), quads_approx[k], 0.5, true);
 
             findCorner(quads_approx[k], corners[i], quad_corners[k]);
-#else
-            findCorner(*quads[k], corners[i], quad_corners[k]);
-#endif
             quad_corners[k] += Point2f(0.5f, 0.5f);
         }
 
@@ -323,44 +233,7 @@ bool cv::find4QuadCornerSubpix(InputArray _img, InputOutputArray _corners, Size
         if(cvIsNaN(angle) || cvIsInf(angle) || angle < 0.5 || angle > CV_PI - 0.5) continue;
 
         findLinesCrossPoint(origin1, dir1, origin2, dir2, corners[i]);
-
-#if defined(_SUBPIX_VERBOSE)
-        radius[i] = norm(corners[i] - ground_truth_corners[ground_truth_idx])*6;
-
-#if 1
-        Mat test(img.size(), CV_32FC3);
-        cvtColor(img, test, CV_GRAY2RGB);
-//        line(test, quad_corners[0] - corners[i] + Point2f(30, 30), quad_corners[1] - corners[i] + Point2f(30, 30), cvScalar(0, 255, 0));
-//        line(test, quad_corners[2] - corners[i] + Point2f(30, 30), quad_corners[3] - corners[i] + Point2f(30, 30), cvScalar(0, 255, 0));
-        std::vector<std::vector<Point> > contrs;
-        contrs.resize(1);
-        for(int k = 0; k < 4; k++)
-        {
-            //contrs[0] = quads_approx[k];
-            contrs[0].clear();
-            for(size_t j = 0; j < quads_approx[k].size(); j++) contrs[0].push_back(quads_approx[k][j]);
-            drawContours(test, contrs, 0, CV_RGB(0, 0, 255), 1, 1, std::vector<Vec4i>(), 2);
-            circle(test, quad_corners[k], 0.5, CV_RGB(255, 0, 0));
-        }
-        Mat test1 = test(Rect(corners[i].x - 30, corners[i].y - 30, 60, 60));
-        namedWindow("1", 1);
-        imshow("1", test1);
-        imwrite("test.jpg", test);
-        waitKey(0);
-#endif
-#endif //_SUBPIX_VERBOSE
-
     }
 
-#if defined(_SUBPIX_VERBOSE)
-    Mat test(img.size(), CV_32FC3);
-    cvtColor(img, test, CV_GRAY2RGB);
-    drawCircles(test, corners, radius);
-
-    namedWindow("corners", 1);
-    imshow("corners", test);
-    waitKey();
-#endif //_SUBPIX_VERBOSE
-
     return true;
 }
diff --git a/modules/calib3d/src/solvepnp.cpp b/modules/calib3d/src/solvepnp.cpp
index 5bcffb3166..698302b942 100644
--- a/modules/calib3d/src/solvepnp.cpp
+++ b/modules/calib3d/src/solvepnp.cpp
@@ -43,6 +43,8 @@
 #include "precomp.hpp"
 #include "epnp.h"
 #include "p3p.h"
+#include "opencv2/calib3d/calib3d_c.h"
+
 #include <iostream>
 using namespace cv;
 
@@ -57,7 +59,7 @@ bool cv::solvePnP( InputArray _opoints, InputArray _ipoints,
     _tvec.create(3, 1, CV_64F);
     Mat cameraMatrix = _cameraMatrix.getMat(), distCoeffs = _distCoeffs.getMat();
 
-    if (flags == CV_EPNP)
+    if (flags == EPNP)
     {
         cv::Mat undistortedPoints;
         cv::undistortPoints(ipoints, undistortedPoints, cameraMatrix, distCoeffs);
@@ -68,7 +70,7 @@ bool cv::solvePnP( InputArray _opoints, InputArray _ipoints,
         cv::Rodrigues(R, rvec);
         return true;
     }
-    else if (flags == CV_P3P)
+    else if (flags == P3P)
     {
         CV_Assert( npoints == 4);
         cv::Mat undistortedPoints;
@@ -81,7 +83,7 @@ bool cv::solvePnP( InputArray _opoints, InputArray _ipoints,
             cv::Rodrigues(R, rvec);
         return result;
     }
-    else if (flags == CV_ITERATIVE)
+    else if (flags == ITERATIVE)
     {
         CvMat c_objectPoints = opoints, c_imagePoints = ipoints;
         CvMat c_cameraMatrix = cameraMatrix, c_distCoeffs = distCoeffs;
@@ -115,31 +117,6 @@ namespace cv
             transform(points, modif_points, transformation);
         }
 
-        class Mutex
-        {
-        public:
-            Mutex() {
-            }
-            void lock()
-            {
-#ifdef HAVE_TBB
-                resultsMutex.lock();
-#endif
-            }
-
-            void unlock()
-            {
-#ifdef HAVE_TBB
-                resultsMutex.unlock();
-#endif
-            }
-
-        private:
-#ifdef HAVE_TBB
-            tbb::mutex resultsMutex;
-#endif
-        };
-
         struct CameraParameters
         {
             void init(Mat _intrinsics, Mat _distCoeffs)
@@ -342,7 +319,7 @@ void cv::solvePnPRansac(InputArray _opoints, InputArray _ipoints,
 
     if (localInliers.size() >= (size_t)pnpransac::MIN_POINTS_COUNT)
     {
-        if (flags != CV_P3P)
+        if (flags != P3P)
         {
             int i, pointsCount = (int)localInliers.size();
             Mat inlierObjectPoints(1, pointsCount, CV_32FC3), inlierImagePoints(1, pointsCount, CV_32FC2);
@@ -371,4 +348,3 @@ void cv::solvePnPRansac(InputArray _opoints, InputArray _ipoints,
     }
     return;
 }
-
diff --git a/modules/calib3d/src/stereobm.cpp b/modules/calib3d/src/stereobm.cpp
index 158881bd94..ee131db76d 100644
--- a/modules/calib3d/src/stereobm.cpp
+++ b/modules/calib3d/src/stereobm.cpp
@@ -56,7 +56,7 @@ struct StereoBMParams
 {
     StereoBMParams(int _numDisparities=64, int _SADWindowSize=21)
     {
-        preFilterType = STEREO_PREFILTER_XSOBEL;
+        preFilterType = StereoBM::PREFILTER_XSOBEL;
         preFilterSize = 9;
         preFilterCap = 31;
         SADWindowSize = _SADWindowSize;
@@ -84,7 +84,7 @@ struct StereoBMParams
     int disp12MaxDiff;
     int dispType;
 };
-    
+
 
 static void prefilterNorm( const Mat& src, Mat& dst, int winsize, int ftzero, uchar* buf )
 {
@@ -183,7 +183,7 @@ prefilterXSobel( const Mat& src, Mat& dst, int ftzero )
         if( useSIMD )
         {
             __m128i z = _mm_setzero_si128(), ftz = _mm_set1_epi16((short)ftzero),
-            ftz2 = _mm_set1_epi8(CV_CAST_8U(ftzero*2));
+            ftz2 = _mm_set1_epi8(cv::saturate_cast<uchar>(ftzero*2));
             for( ; x <= size.width-9; x += 8 )
             {
                 __m128i c0 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow0 + x - 1)), z);
@@ -195,9 +195,9 @@ prefilterXSobel( const Mat& src, Mat& dst, int ftzero )
                 d1 = _mm_sub_epi16(d1, c1);
 
                 __m128i c2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow2 + x - 1)), z);
-                __m128i c3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow2 + x - 1)), z);
+                __m128i c3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow3 + x - 1)), z);
                 __m128i d2 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow2 + x + 1)), z);
-                __m128i d3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow2 + x + 1)), z);
+                __m128i d3 = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)(srow3 + x + 1)), z);
 
                 d2 = _mm_sub_epi16(d2, c2);
                 d3 = _mm_sub_epi16(d3, c3);
@@ -676,7 +676,7 @@ struct PrefilterInvoker : public ParallelLoopBody
     {
         for( int i = range.start; i < range.end; i++ )
         {
-            if( state->preFilterType == STEREO_PREFILTER_NORMALIZED_RESPONSE )
+            if( state->preFilterType == StereoBM::PREFILTER_NORMALIZED_RESPONSE )
                 prefilterNorm( *imgs0[i], *imgs[i], state->preFilterSize, state->preFilterCap, buf[i] );
             else
                 prefilterXSobel( *imgs0[i], *imgs[i], state->preFilterCap );
@@ -771,8 +771,7 @@ protected:
     Rect validDisparityRect;
 };
 
-    
-class StereoBMImpl : public StereoMatcher
+class StereoBMImpl : public StereoBM
 {
 public:
     StereoBMImpl()
@@ -784,46 +783,46 @@ public:
     {
         params = StereoBMParams(_numDisparities, _SADWindowSize);
     }
-    
+
     void compute( InputArray leftarr, InputArray rightarr, OutputArray disparr )
     {
         Mat left0 = leftarr.getMat(), right0 = rightarr.getMat();
         int dtype = disparr.fixedType() ? disparr.type() : params.dispType;
 
         if (left0.size() != right0.size())
-            CV_Error( CV_StsUnmatchedSizes, "All the images must have the same size" );
+            CV_Error( Error::StsUnmatchedSizes, "All the images must have the same size" );
 
         if (left0.type() != CV_8UC1 || right0.type() != CV_8UC1)
-            CV_Error( CV_StsUnsupportedFormat, "Both input images must have CV_8UC1" );
+            CV_Error( Error::StsUnsupportedFormat, "Both input images must have CV_8UC1" );
 
         if (dtype != CV_16SC1 && dtype != CV_32FC1)
-            CV_Error( CV_StsUnsupportedFormat, "Disparity image must have CV_16SC1 or CV_32FC1 format" );
+            CV_Error( Error::StsUnsupportedFormat, "Disparity image must have CV_16SC1 or CV_32FC1 format" );
 
         disparr.create(left0.size(), dtype);
         Mat disp0 = disparr.getMat();
 
-        if( params.preFilterType != STEREO_PREFILTER_NORMALIZED_RESPONSE &&
-            params.preFilterType != STEREO_PREFILTER_XSOBEL )
-            CV_Error( CV_StsOutOfRange, "preFilterType must be = CV_STEREO_BM_NORMALIZED_RESPONSE" );
+        if( params.preFilterType != PREFILTER_NORMALIZED_RESPONSE &&
+            params.preFilterType != PREFILTER_XSOBEL )
+            CV_Error( Error::StsOutOfRange, "preFilterType must be = CV_STEREO_BM_NORMALIZED_RESPONSE" );
 
         if( params.preFilterSize < 5 || params.preFilterSize > 255 || params.preFilterSize % 2 == 0 )
-            CV_Error( CV_StsOutOfRange, "preFilterSize must be odd and be within 5..255" );
+            CV_Error( Error::StsOutOfRange, "preFilterSize must be odd and be within 5..255" );
 
         if( params.preFilterCap < 1 || params.preFilterCap > 63 )
-            CV_Error( CV_StsOutOfRange, "preFilterCap must be within 1..63" );
+            CV_Error( Error::StsOutOfRange, "preFilterCap must be within 1..63" );
 
         if( params.SADWindowSize < 5 || params.SADWindowSize > 255 || params.SADWindowSize % 2 == 0 ||
             params.SADWindowSize >= std::min(left0.cols, left0.rows) )
-            CV_Error( CV_StsOutOfRange, "SADWindowSize must be odd, be within 5..255 and be not larger than image width or height" );
+            CV_Error( Error::StsOutOfRange, "SADWindowSize must be odd, be within 5..255 and be not larger than image width or height" );
 
         if( params.numDisparities <= 0 || params.numDisparities % 16 != 0 )
-            CV_Error( CV_StsOutOfRange, "numDisparities must be positive and divisble by 16" );
+            CV_Error( Error::StsOutOfRange, "numDisparities must be positive and divisble by 16" );
 
         if( params.textureThreshold < 0 )
-            CV_Error( CV_StsOutOfRange, "texture threshold must be non-negative" );
+            CV_Error( Error::StsOutOfRange, "texture threshold must be non-negative" );
 
         if( params.uniquenessRatio < 0 )
-            CV_Error( CV_StsOutOfRange, "uniqueness ratio must be non-negative" );
+            CV_Error( Error::StsOutOfRange, "uniqueness ratio must be non-negative" );
 
         preFilteredImg0.create( left0.size(), CV_8U );
         preFilteredImg1.create( left0.size(), CV_8U );
@@ -888,48 +887,111 @@ public:
                                                   R2.area() > 0 ? Rect(0, 0, width, height) : validDisparityRect,
                                                   params.minDisparity, params.numDisparities,
                                                   params.SADWindowSize);
-        
+
         parallel_for_(Range(0, nstripes),
                       FindStereoCorrespInvoker(left, right, disp, &params, nstripes,
                                                bufSize0, useShorts, validDisparityRect,
                                                slidingSumBuf, cost));
-        
+
         if( params.speckleRange >= 0 && params.speckleWindowSize > 0 )
             filterSpeckles(disp, FILTERED, params.speckleWindowSize, params.speckleRange, slidingSumBuf);
-        
+
         if (disp0.data != disp.data)
             disp.convertTo(disp0, disp0.type(), 1./(1 << DISPARITY_SHIFT), 0);
     }
 
-    AlgorithmInfo* info() const;
+    AlgorithmInfo* info() const { return 0; }
+
+    int getMinDisparity() const { return params.minDisparity; }
+    void setMinDisparity(int minDisparity) { params.minDisparity = minDisparity; }
+
+    int getNumDisparities() const { return params.numDisparities; }
+    void setNumDisparities(int numDisparities) { params.numDisparities = numDisparities; }
+
+    int getBlockSize() const { return params.SADWindowSize; }
+    void setBlockSize(int blockSize) { params.SADWindowSize = blockSize; }
+
+    int getSpeckleWindowSize() const { return params.speckleWindowSize; }
+    void setSpeckleWindowSize(int speckleWindowSize) { params.speckleWindowSize = speckleWindowSize; }
+
+    int getSpeckleRange() const { return params.speckleRange; }
+    void setSpeckleRange(int speckleRange) { params.speckleRange = speckleRange; }
+
+    int getDisp12MaxDiff() const { return params.disp12MaxDiff; }
+    void setDisp12MaxDiff(int disp12MaxDiff) { params.disp12MaxDiff = disp12MaxDiff; }
+
+    int getPreFilterType() const { return params.preFilterType; }
+    void setPreFilterType(int preFilterType) { params.preFilterType = preFilterType; }
+
+    int getPreFilterSize() const { return params.preFilterSize; }
+    void setPreFilterSize(int preFilterSize) { params.preFilterSize = preFilterSize; }
+
+    int getPreFilterCap() const { return params.preFilterCap; }
+    void setPreFilterCap(int preFilterCap) { params.preFilterCap = preFilterCap; }
+
+    int getTextureThreshold() const { return params.textureThreshold; }
+    void setTextureThreshold(int textureThreshold) { params.textureThreshold = textureThreshold; }
+
+    int getUniquenessRatio() const { return params.uniquenessRatio; }
+    void setUniquenessRatio(int uniquenessRatio) { params.uniquenessRatio = uniquenessRatio; }
+
+    int getSmallerBlockSize() const { return 0; }
+    void setSmallerBlockSize(int) {}
+
+    Rect getROI1() const { return params.roi1; }
+    void setROI1(Rect roi1) { params.roi1 = roi1; }
+
+    Rect getROI2() const { return params.roi2; }
+    void setROI2(Rect roi2) { params.roi2 = roi2; }
+
+    void write(FileStorage& fs) const
+    {
+        fs << "name" << name_
+        << "minDisparity" << params.minDisparity
+        << "numDisparities" << params.numDisparities
+        << "blockSize" << params.SADWindowSize
+        << "speckleWindowSize" << params.speckleWindowSize
+        << "speckleRange" << params.speckleRange
+        << "disp12MaxDiff" << params.disp12MaxDiff
+        << "preFilterType" << params.preFilterType
+        << "preFilterSize" << params.preFilterSize
+        << "preFilterCap" << params.preFilterCap
+        << "textureThreshold" << params.textureThreshold
+        << "uniquenessRatio" << params.uniquenessRatio;
+    }
+
+    void read(const FileNode& fn)
+    {
+        FileNode n = fn["name"];
+        CV_Assert( n.isString() && String(n) == name_ );
+        params.minDisparity = (int)fn["minDisparity"];
+        params.numDisparities = (int)fn["numDisparities"];
+        params.SADWindowSize = (int)fn["blockSize"];
+        params.speckleWindowSize = (int)fn["speckleWindowSize"];
+        params.speckleRange = (int)fn["speckleRange"];
+        params.disp12MaxDiff = (int)fn["disp12MaxDiff"];
+        params.preFilterType = (int)fn["preFilterType"];
+        params.preFilterSize = (int)fn["preFilterSize"];
+        params.preFilterCap = (int)fn["preFilterCap"];
+        params.textureThreshold = (int)fn["textureThreshold"];
+        params.uniquenessRatio = (int)fn["uniquenessRatio"];
+        params.roi1 = params.roi2 = Rect();
+    }
 
     StereoBMParams params;
     Mat preFilteredImg0, preFilteredImg1, cost, dispbuf;
     Mat slidingSumBuf;
+
+    static const char* name_;
 };
 
-#define add_param(n) \
-    obj.info()->addParam(obj, #n, obj.params.n)
-
-CV_INIT_ALGORITHM(StereoBMImpl, "StereoMatcher.BM",
-                  add_param(preFilterType);
-                  add_param(preFilterSize);
-                  add_param(preFilterCap);
-                  add_param(SADWindowSize);
-                  add_param(minDisparity);
-                  add_param(numDisparities);
-                  add_param(textureThreshold);
-                  add_param(uniquenessRatio);
-                  add_param(speckleRange);
-                  add_param(speckleWindowSize);
-                  add_param(disp12MaxDiff);
-                  add_param(dispType));
+const char* StereoBMImpl::name_ = "StereoMatcher.BM";
 
 }
 
-cv::Ptr<cv::StereoMatcher> cv::createStereoBM(int _numDisparities, int _SADWindowSize)
+cv::Ptr<cv::StereoBM> cv::createStereoBM(int _numDisparities, int _SADWindowSize)
 {
-    return new StereoBMImpl(_numDisparities, _SADWindowSize);
+    return makePtr<StereoBMImpl>(_numDisparities, _SADWindowSize);
 }
 
 /* End of file. */
diff --git a/modules/calib3d/src/stereosgbm.cpp b/modules/calib3d/src/stereosgbm.cpp
index 026201fc03..6d75d8f53d 100644
--- a/modules/calib3d/src/stereosgbm.cpp
+++ b/modules/calib3d/src/stereosgbm.cpp
@@ -75,13 +75,13 @@ struct StereoSGBMParams
         uniquenessRatio = 0;
         speckleWindowSize = 0;
         speckleRange = 0;
-        fullDP = false;
+        mode = StereoSGBM::MODE_SGBM;
     }
 
     StereoSGBMParams( int _minDisparity, int _numDisparities, int _SADWindowSize,
                       int _P1, int _P2, int _disp12MaxDiff, int _preFilterCap,
                       int _uniquenessRatio, int _speckleWindowSize, int _speckleRange,
-                      bool _fullDP )
+                      int _mode )
     {
         minDisparity = _minDisparity;
         numDisparities = _numDisparities;
@@ -93,7 +93,7 @@ struct StereoSGBMParams
         uniquenessRatio = _uniquenessRatio;
         speckleWindowSize = _speckleWindowSize;
         speckleRange = _speckleRange;
-        fullDP = _fullDP;
+        mode = _mode;
     }
 
     int minDisparity;
@@ -106,7 +106,7 @@ struct StereoSGBMParams
     int speckleWindowSize;
     int speckleRange;
     int disp12MaxDiff;
-    bool fullDP;
+    int mode;
 };
 
 /*
@@ -328,8 +328,8 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
 #endif
 
     const int ALIGN = 16;
-    const int DISP_SHIFT = StereoSGBM::DISP_SHIFT;
-    const int DISP_SCALE = StereoSGBM::DISP_SCALE;
+    const int DISP_SHIFT = StereoMatcher::DISP_SHIFT;
+    const int DISP_SCALE = (1 << DISP_SHIFT);
     const CostType MAX_COST = SHRT_MAX;
 
     int minD = params.minDisparity, maxD = minD + params.numDisparities;
@@ -344,7 +344,8 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
     int D = maxD - minD, width1 = maxX1 - minX1;
     int INVALID_DISP = minD - 1, INVALID_DISP_SCALED = INVALID_DISP*DISP_SCALE;
     int SW2 = SADWindowSize.width/2, SH2 = SADWindowSize.height/2;
-    int npasses = params.fullDP ? 2 : 1;
+    bool fullDP = params.mode == StereoSGBM::MODE_HH;
+    int npasses = fullDP ? 2 : 1;
     const int TAB_OFS = 256*4, TAB_SIZE = 256 + TAB_OFS*2;
     PixType clipTab[TAB_SIZE];
 
@@ -373,7 +374,7 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
     // we keep pixel difference cost (C) and the summary cost over NR directions (S).
     // we also keep all the partial costs for the previous line L_r(x,d) and also min_k L_r(x, k)
     size_t costBufSize = width1*D;
-    size_t CSBufSize = costBufSize*(params.fullDP ? height : 1);
+    size_t CSBufSize = costBufSize*(fullDP ? height : 1);
     size_t minLrSize = (width1 + LrBorder*2)*NR2, LrSize = minLrSize*D2;
     int hsumBufNRows = SH2*2 + 2;
     size_t totalBufSize = (LrSize + minLrSize)*NLR*sizeof(CostType) + // minLr[] and Lr[]
@@ -434,8 +435,8 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
         {
             int x, d;
             DispType* disp1ptr = disp1.ptr<DispType>(y);
-            CostType* C = Cbuf + (!params.fullDP ? 0 : y*costBufSize);
-            CostType* S = Sbuf + (!params.fullDP ? 0 : y*costBufSize);
+            CostType* C = Cbuf + (!fullDP ? 0 : y*costBufSize);
+            CostType* S = Sbuf + (!fullDP ? 0 : y*costBufSize);
 
             if( pass == 1 ) // compute C on the first pass, and reuse it on the second pass, if any.
             {
@@ -460,7 +461,7 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
                         if( y > 0 )
                         {
                             const CostType* hsumSub = hsumBuf + (std::max(y - SH2 - 1, 0) % hsumBufNRows)*costBufSize;
-                            const CostType* Cprev = !params.fullDP || y == 0 ? C : C - costBufSize;
+                            const CostType* Cprev = !fullDP || y == 0 ? C : C - costBufSize;
 
                             for( x = D; x < width1*D; x += D )
                             {
@@ -828,8 +829,7 @@ static void computeDisparitySGBM( const Mat& img1, const Mat& img2,
     }
 }
 
-
-class StereoSGBMImpl : public StereoMatcher
+class StereoSGBMImpl : public StereoSGBM
 {
 public:
     StereoSGBMImpl()
@@ -840,12 +840,12 @@ public:
     StereoSGBMImpl( int _minDisparity, int _numDisparities, int _SADWindowSize,
                     int _P1, int _P2, int _disp12MaxDiff, int _preFilterCap,
                     int _uniquenessRatio, int _speckleWindowSize, int _speckleRange,
-                    bool _fullDP )
+                    int _mode )
     {
         params = StereoSGBMParams( _minDisparity, _numDisparities, _SADWindowSize,
                                    _P1, _P2, _disp12MaxDiff, _preFilterCap,
                                    _uniquenessRatio, _speckleWindowSize, _speckleRange,
-                                   _fullDP );
+                                   _mode );
     }
 
     void compute( InputArray leftarr, InputArray rightarr, OutputArray disparr )
@@ -861,47 +861,100 @@ public:
         medianBlur(disp, disp, 3);
 
         if( params.speckleWindowSize > 0 )
-            filterSpeckles(disp, (params.minDisparity - 1)*STEREO_DISP_SCALE, params.speckleWindowSize,
-                           STEREO_DISP_SCALE*params.speckleRange, buffer);
+            filterSpeckles(disp, (params.minDisparity - 1)*StereoMatcher::DISP_SCALE, params.speckleWindowSize,
+                           StereoMatcher::DISP_SCALE*params.speckleRange, buffer);
     }
 
-    AlgorithmInfo* info() const;
+    AlgorithmInfo* info() const { return 0; }
+
+    int getMinDisparity() const { return params.minDisparity; }
+    void setMinDisparity(int minDisparity) { params.minDisparity = minDisparity; }
+
+    int getNumDisparities() const { return params.numDisparities; }
+    void setNumDisparities(int numDisparities) { params.numDisparities = numDisparities; }
+
+    int getBlockSize() const { return params.SADWindowSize; }
+    void setBlockSize(int blockSize) { params.SADWindowSize = blockSize; }
+
+    int getSpeckleWindowSize() const { return params.speckleWindowSize; }
+    void setSpeckleWindowSize(int speckleWindowSize) { params.speckleWindowSize = speckleWindowSize; }
+
+    int getSpeckleRange() const { return params.speckleRange; }
+    void setSpeckleRange(int speckleRange) { params.speckleRange = speckleRange; }
+
+    int getDisp12MaxDiff() const { return params.disp12MaxDiff; }
+    void setDisp12MaxDiff(int disp12MaxDiff) { params.disp12MaxDiff = disp12MaxDiff; }
+
+    int getPreFilterCap() const { return params.preFilterCap; }
+    void setPreFilterCap(int preFilterCap) { params.preFilterCap = preFilterCap; }
+
+    int getUniquenessRatio() const { return params.uniquenessRatio; }
+    void setUniquenessRatio(int uniquenessRatio) { params.uniquenessRatio = uniquenessRatio; }
+
+    int getP1() const { return params.P1; }
+    void setP1(int P1) { params.P1 = P1; }
+
+    int getP2() const { return params.P2; }
+    void setP2(int P2) { params.P2 = P2; }
+
+    int getMode() const { return params.mode; }
+    void setMode(int mode) { params.mode = mode; }
+
+    void write(FileStorage& fs) const
+    {
+        fs << "name" << name_
+        << "minDisparity" << params.minDisparity
+        << "numDisparities" << params.numDisparities
+        << "blockSize" << params.SADWindowSize
+        << "speckleWindowSize" << params.speckleWindowSize
+        << "speckleRange" << params.speckleRange
+        << "disp12MaxDiff" << params.disp12MaxDiff
+        << "preFilterCap" << params.preFilterCap
+        << "uniquenessRatio" << params.uniquenessRatio
+        << "P1" << params.P1
+        << "P2" << params.P2
+        << "mode" << params.mode;
+    }
+
+    void read(const FileNode& fn)
+    {
+        FileNode n = fn["name"];
+        CV_Assert( n.isString() && String(n) == name_ );
+        params.minDisparity = (int)fn["minDisparity"];
+        params.numDisparities = (int)fn["numDisparities"];
+        params.SADWindowSize = (int)fn["blockSize"];
+        params.speckleWindowSize = (int)fn["speckleWindowSize"];
+        params.speckleRange = (int)fn["speckleRange"];
+        params.disp12MaxDiff = (int)fn["disp12MaxDiff"];
+        params.preFilterCap = (int)fn["preFilterCap"];
+        params.uniquenessRatio = (int)fn["uniquenessRatio"];
+        params.P1 = (int)fn["P1"];
+        params.P2 = (int)fn["P2"];
+        params.mode = (int)fn["mode"];
+    }
 
     StereoSGBMParams params;
     Mat buffer;
+    static const char* name_;
 };
 
+const char* StereoSGBMImpl::name_ = "StereoMatcher.SGBM";
 
-Ptr<StereoMatcher> createStereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
-                                    int P1, int P2, int disp12MaxDiff,
-                                    int preFilterCap, int uniquenessRatio,
-                                    int speckleWindowSize, int speckleRange,
-                                    bool fullDP)
+
+Ptr<StereoSGBM> createStereoSGBM(int minDisparity, int numDisparities, int SADWindowSize,
+                                 int P1, int P2, int disp12MaxDiff,
+                                 int preFilterCap, int uniquenessRatio,
+                                 int speckleWindowSize, int speckleRange,
+                                 int mode)
 {
-    return new StereoSGBMImpl(minDisparity, numDisparities, SADWindowSize,
-                              P1, P2, disp12MaxDiff,
-                              preFilterCap, uniquenessRatio,
-                              speckleWindowSize, speckleRange,
-                              fullDP);
+    return Ptr<StereoSGBM>(
+        new StereoSGBMImpl(minDisparity, numDisparities, SADWindowSize,
+                           P1, P2, disp12MaxDiff,
+                           preFilterCap, uniquenessRatio,
+                           speckleWindowSize, speckleRange,
+                           mode));
 }
 
-
-#define add_param(n) \
-    obj.info()->addParam(obj, #n, obj.params.n)
-
-CV_INIT_ALGORITHM(StereoSGBMImpl, "StereoMatcher.SGBM",
-                  add_param(minDisparity);
-                  add_param(numDisparities);
-                  add_param(SADWindowSize);
-                  add_param(preFilterCap);
-                  add_param(uniquenessRatio);
-                  add_param(P1);
-                  add_param(P2);
-                  add_param(speckleWindowSize);
-                  add_param(speckleRange);
-                  add_param(disp12MaxDiff);
-                  add_param(fullDP));
-
 Rect getValidDisparityROI( Rect roi1, Rect roi2,
                           int minDisparity,
                           int numberOfDisparities,
@@ -1117,4 +1170,3 @@ void cv::validateDisparity( InputOutputArray _disp, InputArray _cost, int minDis
         }
     }
 }
-
diff --git a/modules/calib3d/src/triangulate.cpp b/modules/calib3d/src/triangulate.cpp
index 9f52b37163..b0af3dc466 100644
--- a/modules/calib3d/src/triangulate.cpp
+++ b/modules/calib3d/src/triangulate.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 // cvCorrectMatches function is Copyright (C) 2009, Jostein Austvik Jacobsen.
 // cvTriangulatePoints function is derived from icvReconstructPointsFor3View, originally by Valery Mosyagin.
@@ -239,32 +240,32 @@ cvCorrectMatches(CvMat *F_, CvMat *points1_, CvMat *points2_, CvMat *new_points1
     }
 
     // Make sure F uses double precision
-    F = cvCreateMat(3,3,CV_64FC1);
+    F.reset(cvCreateMat(3,3,CV_64FC1));
     cvConvert(F_, F);
 
     // Make sure points1 uses double precision
-    points1 = cvCreateMat(points1_->rows,points1_->cols,CV_64FC2);
+    points1.reset(cvCreateMat(points1_->rows,points1_->cols,CV_64FC2));
     cvConvert(points1_, points1);
 
     // Make sure points2 uses double precision
-    points2 = cvCreateMat(points2_->rows,points2_->cols,CV_64FC2);
+    points2.reset(cvCreateMat(points2_->rows,points2_->cols,CV_64FC2));
     cvConvert(points2_, points2);
 
-    tmp33 = cvCreateMat(3,3,CV_64FC1);
-    tmp31 = cvCreateMat(3,1,CV_64FC1), tmp31_2 = cvCreateMat(3,1,CV_64FC1);
-    T1i = cvCreateMat(3,3,CV_64FC1), T2i = cvCreateMat(3,3,CV_64FC1);
-    R1 = cvCreateMat(3,3,CV_64FC1), R2 = cvCreateMat(3,3,CV_64FC1);
-    TFT = cvCreateMat(3,3,CV_64FC1), TFTt = cvCreateMat(3,3,CV_64FC1), RTFTR = cvCreateMat(3,3,CV_64FC1);
-    U = cvCreateMat(3,3,CV_64FC1);
-    S = cvCreateMat(3,3,CV_64FC1);
-    V = cvCreateMat(3,3,CV_64FC1);
-    e1 = cvCreateMat(3,1,CV_64FC1), e2 = cvCreateMat(3,1,CV_64FC1);
+    tmp33.reset(cvCreateMat(3,3,CV_64FC1));
+    tmp31.reset(cvCreateMat(3,1,CV_64FC1)), tmp31_2.reset(cvCreateMat(3,1,CV_64FC1));
+    T1i.reset(cvCreateMat(3,3,CV_64FC1)), T2i.reset(cvCreateMat(3,3,CV_64FC1));
+    R1.reset(cvCreateMat(3,3,CV_64FC1)), R2.reset(cvCreateMat(3,3,CV_64FC1));
+    TFT.reset(cvCreateMat(3,3,CV_64FC1)), TFTt.reset(cvCreateMat(3,3,CV_64FC1)), RTFTR.reset(cvCreateMat(3,3,CV_64FC1));
+    U.reset(cvCreateMat(3,3,CV_64FC1));
+    S.reset(cvCreateMat(3,3,CV_64FC1));
+    V.reset(cvCreateMat(3,3,CV_64FC1));
+    e1.reset(cvCreateMat(3,1,CV_64FC1)), e2.reset(cvCreateMat(3,1,CV_64FC1));
 
     double x1, y1, x2, y2;
     double scale;
     double f1, f2, a, b, c, d;
-    polynomial = cvCreateMat(1,7,CV_64FC1);
-    result = cvCreateMat(1,6,CV_64FC2);
+    polynomial.reset(cvCreateMat(1,7,CV_64FC1));
+    result.reset(cvCreateMat(1,6,CV_64FC2));
     double t_min, s_val, t, s;
     for (int p = 0; p < points1->cols; ++p) {
         // Replace F by T2-t * F * T1-t
diff --git a/modules/calib3d/test/test_affine3.cpp b/modules/calib3d/test/test_affine3.cpp
index 196d428e45..62326e9203 100644
--- a/modules/calib3d/test/test_affine3.cpp
+++ b/modules/calib3d/test/test_affine3.cpp
@@ -52,30 +52,30 @@ TEST(Calib3d_Affine3f, accuracy)
 
     cv::Mat expected;
     cv::Rodrigues(rvec, expected);
-    
-    
+
+
     ASSERT_EQ(0, norm(cv::Mat(affine.matrix, false).colRange(0, 3).rowRange(0, 3) != expected));
     ASSERT_EQ(0, norm(cv::Mat(affine.linear()) != expected));
-    
-    
+
+
     cv::Matx33d R = cv::Matx33d::eye();
-    
+
     double angle = 50;
     R.val[0] = R.val[4] = std::cos(CV_PI*angle/180.0);
     R.val[3] = std::sin(CV_PI*angle/180.0);
     R.val[1] = -R.val[3];
-    
-    
+
+
     cv::Affine3d affine1(cv::Mat(cv::Vec3d(0.2, 0.5, 0.3)).reshape(1, 1), cv::Vec3d(4, 5, 6));
     cv::Affine3d affine2(R, cv::Vec3d(1, 1, 0.4));
-    
+
     cv::Affine3d result = affine1.inv() * affine2;
-    
+
     expected = cv::Mat(affine1.matrix.inv(cv::DECOMP_SVD)) * cv::Mat(affine2.matrix, false);
-    
+
 
     cv::Mat diff;
     cv::absdiff(expected, result.matrix, diff);
-    
+
     ASSERT_LT(cv::norm(diff, cv::NORM_INF), 1e-15);
 }
diff --git a/modules/calib3d/test/test_affine3d_estimator.cpp b/modules/calib3d/test/test_affine3d_estimator.cpp
index ff061aacfa..f31e2e7324 100644
--- a/modules/calib3d/test/test_affine3d_estimator.cpp
+++ b/modules/calib3d/test/test_affine3d_estimator.cpp
@@ -195,4 +195,3 @@ void CV_Affine3D_EstTest::run( int /* start_from */)
 }
 
 TEST(Calib3d_EstimateAffineTransform, accuracy) { CV_Affine3D_EstTest test; test.safe_run(); }
-
diff --git a/modules/calib3d/test/test_cameracalibration.cpp b/modules/calib3d/test/test_cameracalibration.cpp
index e8b5570949..fb82382377 100644
--- a/modules/calib3d/test/test_cameracalibration.cpp
+++ b/modules/calib3d/test/test_cameracalibration.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 #include <limits>
 
diff --git a/modules/calib3d/test/test_cameracalibration_artificial.cpp b/modules/calib3d/test/test_cameracalibration_artificial.cpp
index 733f0069c7..1ff13c9e87 100644
--- a/modules/calib3d/test/test_cameracalibration_artificial.cpp
+++ b/modules/calib3d/test/test_cameracalibration_artificial.cpp
@@ -304,7 +304,7 @@ protected:
             for(size_t i = 0; i < brdsNum; ++i)
             {
                 Mat gray;
-                cvtColor(boards[i], gray, CV_BGR2GRAY);
+                cvtColor(boards[i], gray, COLOR_BGR2GRAY);
                 vector<Point2f> tmp = imagePoints_findCb[i];
                 cornerSubPix(gray, tmp, Size(5, 5), Size(-1,-1), tc);
                 imagePoints.push_back(tmp);
@@ -314,7 +314,7 @@ protected:
             for(size_t i = 0; i < brdsNum; ++i)
             {
                 Mat gray;
-                cvtColor(boards[i], gray, CV_BGR2GRAY);
+                cvtColor(boards[i], gray, COLOR_BGR2GRAY);
                 vector<Point2f> tmp = imagePoints_findCb[i];
                 find4QuadCornerSubpix(gray, tmp, Size(5, 5));
                 imagePoints.push_back(tmp);
@@ -327,7 +327,7 @@ protected:
         Mat camMat_est = Mat::eye(3, 3, CV_64F), distCoeffs_est = Mat::zeros(1, 5, CV_64F);
         vector<Mat> rvecs_est, tvecs_est;
 
-        int flags = /*CV_CALIB_FIX_K3|*/CV_CALIB_FIX_K4|CV_CALIB_FIX_K5|CV_CALIB_FIX_K6; //CALIB_FIX_K3; //CALIB_FIX_ASPECT_RATIO |  | CALIB_ZERO_TANGENT_DIST;
+        int flags = /*CALIB_FIX_K3|*/CALIB_FIX_K4|CALIB_FIX_K5|CALIB_FIX_K6; //CALIB_FIX_K3; //CALIB_FIX_ASPECT_RATIO |  | CALIB_ZERO_TANGENT_DIST;
         TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 100, DBL_EPSILON);
         double rep_error = calibrateCamera(objectPoints, imagePoints, imgSize, camMat_est, distCoeffs_est, rvecs_est, tvecs_est, flags, criteria);
         rep_error /= brdsNum * cornersSize.area();
diff --git a/modules/calib3d/test/test_cameracalibration_badarg.cpp b/modules/calib3d/test/test_cameracalibration_badarg.cpp
index b805e71a49..3edab8becf 100644
--- a/modules/calib3d/test/test_cameracalibration_badarg.cpp
+++ b/modules/calib3d/test/test_cameracalibration_badarg.cpp
@@ -41,6 +41,7 @@
 
 #include "test_precomp.hpp"
 #include "test_chessboardgenerator.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 #include <iostream>
 
@@ -734,5 +735,3 @@ protected:
 TEST(Calib3d_CalibrateCamera_C, badarg) { CV_CameraCalibrationBadArgTest test; test.safe_run(); }
 TEST(Calib3d_Rodrigues_C, badarg) { CV_Rodrigues2BadArgTest test; test.safe_run(); }
 TEST(Calib3d_ProjectPoints_C, badarg) { CV_ProjectPoints2BadArgTest test; test.safe_run(); }
-
-
diff --git a/modules/calib3d/test/test_chessboardgenerator.cpp b/modules/calib3d/test/test_chessboardgenerator.cpp
index 7c761fc572..7c0bd34dc7 100644
--- a/modules/calib3d/test/test_chessboardgenerator.cpp
+++ b/modules/calib3d/test/test_chessboardgenerator.cpp
@@ -161,15 +161,15 @@ Mat cv::ChessBoardGenerator::generateChessBoard(const Mat& bg, const Mat& camMat
     if (rendererResolutionMultiplier == 1)
     {
         result = bg.clone();
-        drawContours(result, whole_contour, -1, Scalar::all(255), CV_FILLED, CV_AA);
-        drawContours(result, squares_black, -1, Scalar::all(0), CV_FILLED, CV_AA);
+        drawContours(result, whole_contour, -1, Scalar::all(255), FILLED, LINE_AA);
+        drawContours(result, squares_black, -1, Scalar::all(0), FILLED, LINE_AA);
     }
     else
     {
         Mat tmp;
         resize(bg, tmp, bg.size() * rendererResolutionMultiplier);
-        drawContours(tmp, whole_contour, -1, Scalar::all(255), CV_FILLED, CV_AA);
-        drawContours(tmp, squares_black, -1, Scalar::all(0), CV_FILLED, CV_AA);
+        drawContours(tmp, whole_contour, -1, Scalar::all(255), FILLED, LINE_AA);
+        drawContours(tmp, squares_black, -1, Scalar::all(0), FILLED, LINE_AA);
         resize(tmp, result, bg.size(), 0, 0, INTER_AREA);
     }
 
@@ -329,4 +329,3 @@ Mat cv::ChessBoardGenerator::operator ()(const Mat& bg, const Mat& camMat, const
     return generateChessBoard(bg, camMat, distCoeffs, zero, pb1, pb2,
         squareSize.width, squareSize.height,  pts3d, corners);
 }
-
diff --git a/modules/calib3d/test/test_chesscorners.cpp b/modules/calib3d/test/test_chesscorners.cpp
index 42c25b86ad..f9625d540a 100644
--- a/modules/calib3d/test/test_chesscorners.cpp
+++ b/modules/calib3d/test/test_chesscorners.cpp
@@ -57,14 +57,14 @@ void show_points( const Mat& gray, const Mat& u, const vector<Point2f>& v, Size
     merge(vector<Mat>(3, gray), rgb);
 
     for(size_t i = 0; i < v.size(); i++ )
-        circle( rgb, v[i], 3, CV_RGB(255, 0, 0), CV_FILLED);
+        circle( rgb, v[i], 3, Scalar(255, 0, 0), FILLED);
 
     if( !u.empty() )
     {
         const Point2f* u_data = u.ptr<Point2f>();
         size_t count = u.cols * u.rows;
         for(size_t i = 0; i < count; i++ )
-            circle( rgb, u_data[i], 3, CV_RGB(0, 255, 0), CV_FILLED);
+            circle( rgb, u_data[i], 3, Scalar(0, 255, 0), FILLED);
     }
     if (!v.empty())
     {
@@ -208,7 +208,7 @@ void CV_ChessboardDetectorTest::run_batch( const string& filename )
     }
 
     int progress = 0;
-    int max_idx = board_list.node->data.seq->total/2;
+    int max_idx = board_list.size()/2;
     double sum_error = 0.0;
     int count = 0;
 
@@ -217,7 +217,7 @@ void CV_ChessboardDetectorTest::run_batch( const string& filename )
         ts->update_context( this, idx, true );
 
         /* read the image */
-        string img_file = board_list[idx * 2];
+        String img_file = board_list[idx * 2];
         Mat gray = imread( folder + img_file, 0);
 
         if( gray.empty() )
@@ -227,7 +227,7 @@ void CV_ChessboardDetectorTest::run_batch( const string& filename )
             return;
         }
 
-        string _filename = folder + (string)board_list[idx * 2 + 1];
+        String _filename = folder + (String)board_list[idx * 2 + 1];
         bool doesContatinChessboard;
         Mat expected;
         {
@@ -244,7 +244,7 @@ void CV_ChessboardDetectorTest::run_batch( const string& filename )
         switch( pattern )
         {
             case CHESSBOARD:
-                result = findChessboardCorners(gray, pattern_size, v, CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_NORMALIZE_IMAGE);
+                result = findChessboardCorners(gray, pattern_size, v, CALIB_CB_ADAPTIVE_THRESH | CALIB_CB_NORMALIZE_IMAGE);
                 break;
             case CIRCLES_GRID:
                 result = findCirclesGrid(gray, pattern_size, v);
@@ -459,7 +459,7 @@ bool CV_ChessboardDetectorTest::checkByGenerator()
         vector<Point>& cnt = cnts[0];
         cnt.push_back(cg[  0]); cnt.push_back(cg[0+2]);
         cnt.push_back(cg[7+0]); cnt.push_back(cg[7+2]);
-        cv::drawContours(cb, cnts, -1, Scalar::all(128), CV_FILLED);
+        cv::drawContours(cb, cnts, -1, Scalar::all(128), FILLED);
 
         found = findChessboardCorners(cb, cbg.cornersSize(), corners_found);
         if (found)
diff --git a/modules/calib3d/test/test_chesscorners_badarg.cpp b/modules/calib3d/test/test_chesscorners_badarg.cpp
index d7c4c4f4ce..318912eeb4 100644
--- a/modules/calib3d/test/test_chesscorners_badarg.cpp
+++ b/modules/calib3d/test/test_chesscorners_badarg.cpp
@@ -41,6 +41,7 @@
 
 #include "test_precomp.hpp"
 #include "test_chessboardgenerator.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 #include <limits>
 
diff --git a/modules/calib3d/test/test_chesscorners_timing.cpp b/modules/calib3d/test/test_chesscorners_timing.cpp
index 6195c04b07..47653f88d8 100644
--- a/modules/calib3d/test/test_chesscorners_timing.cpp
+++ b/modules/calib3d/test/test_chesscorners_timing.cpp
@@ -40,6 +40,8 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/calib3d/calib3d_c.h"
 
 class CV_ChessboardDetectorTimingTest : public cvtest::BaseTest
 {
@@ -66,7 +68,7 @@ void CV_ChessboardDetectorTimingTest::run( int start_from )
     CvMat*  _v = 0;
     CvPoint2D32f* v;
 
-    IplImage* img = 0;
+    IplImage img;
     IplImage* gray = 0;
     IplImage* thresh = 0;
 
@@ -105,9 +107,10 @@ void CV_ChessboardDetectorTimingTest::run( int start_from )
         /* read the image */
         sprintf( filename, "%s%s", filepath, imgname );
 
-        img = cvLoadImage( filename );
+        cv::Mat img2 = cv::imread( filename );
+        img = img2;
 
-        if( !img )
+        if( img2.empty() )
         {
             ts->printf( cvtest::TS::LOG, "one of chessboard images can't be read: %s\n", filename );
             if( max_idx == 1 )
@@ -120,9 +123,9 @@ void CV_ChessboardDetectorTimingTest::run( int start_from )
 
         ts->printf(cvtest::TS::LOG, "%s: chessboard %d:\n", imgname, is_chessboard);
 
-        gray = cvCreateImage( cvSize( img->width, img->height ), IPL_DEPTH_8U, 1 );
-        thresh = cvCreateImage( cvSize( img->width, img->height ), IPL_DEPTH_8U, 1 );
-        cvCvtColor( img, gray, CV_BGR2GRAY );
+        gray = cvCreateImage( cvSize( img.width, img.height ), IPL_DEPTH_8U, 1 );
+        thresh = cvCreateImage( cvSize( img.width, img.height ), IPL_DEPTH_8U, 1 );
+        cvCvtColor( &img, gray, CV_BGR2GRAY );
 
 
         count0 = pattern_size.width*pattern_size.height;
@@ -164,7 +167,6 @@ void CV_ChessboardDetectorTimingTest::run( int start_from )
                    find_chessboard_time*1e-6, find_chessboard_time/num_pixels);
 
         cvReleaseMat( &_v );
-        cvReleaseImage( &img );
         cvReleaseImage( &gray );
         cvReleaseImage( &thresh );
         progress = update_progress( progress, idx-1, max_idx, 0 );
@@ -175,7 +177,6 @@ _exit_:
     /* release occupied memory */
     cvReleaseMat( &_v );
     cvReleaseFileStorage( &fs );
-    cvReleaseImage( &img );
     cvReleaseImage( &gray );
     cvReleaseImage( &thresh );
 
diff --git a/modules/calib3d/test/test_compose_rt.cpp b/modules/calib3d/test/test_compose_rt.cpp
index b71288e6d4..577cc06270 100644
--- a/modules/calib3d/test/test_compose_rt.cpp
+++ b/modules/calib3d/test/test_compose_rt.cpp
@@ -212,4 +212,3 @@ protected:
 };
 
 TEST(Calib3d_ComposeRT, accuracy) { CV_composeRT_Test test; test.safe_run(); }
-
diff --git a/modules/calib3d/test/test_cornerssubpix.cpp b/modules/calib3d/test/test_cornerssubpix.cpp
index 88f7deee06..4426d5ea5e 100644
--- a/modules/calib3d/test/test_cornerssubpix.cpp
+++ b/modules/calib3d/test/test_cornerssubpix.cpp
@@ -40,6 +40,7 @@
  //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 #include <limits>
 #include "test_chessboardgenerator.hpp"
 
diff --git a/modules/calib3d/test/test_fundam.cpp b/modules/calib3d/test/test_fundam.cpp
index 19ee70fa1b..7e6f9a8e71 100644
--- a/modules/calib3d/test/test_fundam.cpp
+++ b/modules/calib3d/test/test_fundam.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 using namespace cv;
 using namespace std;
@@ -1079,7 +1080,7 @@ protected:
     void run_func();
     void prepare_to_validation( int );
 
-    double sampson_error(const double* f, double x1, double y1, double x2, double y2); 
+    double sampson_error(const double* f, double x1, double y1, double x2, double y2);
 
     int method;
     int img_size;
@@ -1145,9 +1146,8 @@ void CV_EssentialMatTest::get_test_array_types_and_sizes( int /*test_case_idx*/,
     int pt_count = MAX(5, cvRound(exp(pt_count_exp)));
 
     dims = cvtest::randInt(rng) % 2 + 2;
-    dims = 2; 
+    dims = 2;
     method = CV_LMEDS << (cvtest::randInt(rng) % 2);
-    
 
     types[INPUT][0] = CV_MAKETYPE(pt_depth, 1);
 
@@ -1192,11 +1192,11 @@ void CV_EssentialMatTest::get_test_array_types_and_sizes( int /*test_case_idx*/,
     sizes[OUTPUT][0] = sizes[REF_OUTPUT][0] = cvSize(3,1);
     types[OUTPUT][0] = types[REF_OUTPUT][0] = CV_64FC1;
     sizes[OUTPUT][1] = sizes[REF_OUTPUT][1] = cvSize(pt_count,1);
-    types[OUTPUT][1] = types[REF_OUTPUT][1] = CV_8UC1;       
+    types[OUTPUT][1] = types[REF_OUTPUT][1] = CV_8UC1;
     sizes[OUTPUT][2] = sizes[REF_OUTPUT][2] = cvSize(1,1);
     types[OUTPUT][2] = types[REF_OUTPUT][2] = CV_64FC1;
     sizes[OUTPUT][3] = sizes[REF_OUTPUT][3] = cvSize(1,1);
-    types[OUTPUT][3] = types[REF_OUTPUT][3] = CV_8UC1;       
+    types[OUTPUT][3] = types[REF_OUTPUT][3] = CV_8UC1;
 
 }
 
@@ -1289,46 +1289,46 @@ int CV_EssentialMatTest::prepare_test_case( int test_case_idx )
 void CV_EssentialMatTest::run_func()
 {
     Mat _input0(test_mat[INPUT][0]), _input1(test_mat[INPUT][1]);
-    Mat K(test_mat[INPUT][4]); 
-    double focal(K.at<double>(0, 0)); 
-    cv::Point2d pp(K.at<double>(0, 2), K.at<double>(1, 2)); 
+    Mat K(test_mat[INPUT][4]);
+    double focal(K.at<double>(0, 0));
+    cv::Point2d pp(K.at<double>(0, 2), K.at<double>(1, 2));
 
     RNG& rng = ts->get_rng();
     Mat E, mask1(test_mat[TEMP][1]);
-    E = cv::findEssentialMat( _input0, _input1, focal, pp, method, 0.99, MAX(sigma*3, 0.0001), mask1 ); 
-    if (E.rows > 3) 
+    E = cv::findEssentialMat( _input0, _input1, focal, pp, method, 0.99, MAX(sigma*3, 0.0001), mask1 );
+    if (E.rows > 3)
     {
-        int count = E.rows / 3; 
-        int row = (cvtest::randInt(rng) % count) * 3; 
-        E = E.rowRange(row, row + 3) * 1.0; 
+        int count = E.rows / 3;
+        int row = (cvtest::randInt(rng) % count) * 3;
+        E = E.rowRange(row, row + 3) * 1.0;
     }
 
-    E.copyTo(test_mat[TEMP][0]); 
+    E.copyTo(test_mat[TEMP][0]);
 
-    Mat R, t, mask2; 
-    recoverPose( E, _input0, _input1, R, t, focal, pp, mask2 ); 
-    R.copyTo(test_mat[TEMP][2]); 
-    t.copyTo(test_mat[TEMP][3]); 
-    mask2.copyTo(test_mat[TEMP][4]); 
+    Mat R, t, mask2;
+    recoverPose( E, _input0, _input1, R, t, focal, pp, mask2 );
+    R.copyTo(test_mat[TEMP][2]);
+    t.copyTo(test_mat[TEMP][3]);
+    mask2.copyTo(test_mat[TEMP][4]);
 }
 
 double CV_EssentialMatTest::sampson_error(const double * f, double x1, double y1, double x2, double y2)
 {
     double Fx1[3] = {
-        f[0] * x1 + f[1] * y1 + f[2], 
-        f[3] * x1 + f[4] * y1 + f[5], 
+        f[0] * x1 + f[1] * y1 + f[2],
+        f[3] * x1 + f[4] * y1 + f[5],
         f[6] * x1 + f[7] * y1 + f[8]
-    }; 
+    };
     double Ftx2[3] = {
-        f[0] * x2 + f[3] * y2 + f[6], 
-        f[1] * x2 + f[4] * y2 + f[7], 
+        f[0] * x2 + f[3] * y2 + f[6],
+        f[1] * x2 + f[4] * y2 + f[7],
         f[2] * x2 + f[5] * y2 + f[8]
-    }; 
-    double x2tFx1 = Fx1[0] * x2 + Fx1[1] * y2 + Fx1[2]; 
+    };
+    double x2tFx1 = Fx1[0] * x2 + Fx1[1] * y2 + Fx1[2];
 
-    double error = x2tFx1 * x2tFx1 / (Fx1[0] * Fx1[0] + Fx1[1] * Fx1[1] + Ftx2[0] * Ftx2[0] + Ftx2[1] * Ftx2[1]); 
-    error = sqrt(error); 
-    return error; 
+    double error = x2tFx1 * x2tFx1 / (Fx1[0] * Fx1[0] + Fx1[1] * Fx1[1] + Ftx2[0] * Ftx2[0] + Ftx2[1] * Ftx2[1]);
+    error = sqrt(error);
+    return error;
 
 }
 
@@ -1338,7 +1338,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     const Mat& A = test_mat[INPUT][4];
     double f0[9], f[9], e[9];
     Mat F0(3, 3, CV_64FC1, f0), F(3, 3, CV_64F, f);
-    Mat E(3, 3, CV_64F, e); 
+    Mat E(3, 3, CV_64F, e);
 
     Mat invA, R=Rt0.colRange(0, 3), T1, T2;
 
@@ -1362,7 +1362,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     uchar* mtfm2 = test_mat[OUTPUT][1].data;
     double* e_prop1 = (double*)test_mat[REF_OUTPUT][0].data;
     double* e_prop2 = (double*)test_mat[OUTPUT][0].data;
-    Mat E_prop2 = Mat(3, 1, CV_64F, e_prop2); 
+    Mat E_prop2 = Mat(3, 1, CV_64F, e_prop2);
 
     int i, pt_count = test_mat[INPUT][2].cols;
     Mat p1( 1, pt_count, CV_64FC2 );
@@ -1381,8 +1381,8 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
         double y1 = p1.at<Point2d>(i).y;
         double x2 = p2.at<Point2d>(i).x;
         double y2 = p2.at<Point2d>(i).y;
-//        double t0 = sampson_error(f0, x1, y1, x2, y2); 
-//        double t = sampson_error(f, x1, y1, x2, y2); 
+//        double t0 = sampson_error(f0, x1, y1, x2, y2);
+//        double t = sampson_error(f, x1, y1, x2, y2);
         double n1 = 1./sqrt(x1*x1 + y1*y1 + 1);
         double n2 = 1./sqrt(x2*x2 + y2*y2 + 1);
         double t0 = fabs(f0[0]*x2*x1 + f0[1]*x2*y1 + f0[2]*x2 +
@@ -1394,7 +1394,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
         mtfm1[i] = 1;
         mtfm2[i] = !status[i] || t0 > err_level || t < err_level;
     }
-    
+
     e_prop1[0] = sqrt(0.5);
     e_prop1[1] = sqrt(0.5);
     e_prop1[2] = 0;
@@ -1402,26 +1402,26 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     e_prop2[0] = 0;
     e_prop2[1] = 0;
     e_prop2[2] = 0;
-    SVD::compute(E, E_prop2); 
+    SVD::compute(E, E_prop2);
 
 
 
-    double* pose_prop1 = (double*)test_mat[REF_OUTPUT][2].data; 
-    double* pose_prop2 = (double*)test_mat[OUTPUT][2].data; 
-    double terr1 = norm(Rt0.col(3) / norm(Rt0.col(3)) + test_mat[TEMP][3]); 
-    double terr2 = norm(Rt0.col(3) / norm(Rt0.col(3)) - test_mat[TEMP][3]); 
-    Mat rvec; 
-    Rodrigues(Rt0.colRange(0, 3), rvec); 
-    pose_prop1[0] = 0; 
-    // No check for CV_LMeDS on translation. Since it 
-    // involves with some degraded problem, when data is exact inliers. 
-    pose_prop2[0] = method == CV_LMEDS || pt_count == 5 ? 0 : MIN(terr1, terr2); 
+    double* pose_prop1 = (double*)test_mat[REF_OUTPUT][2].data;
+    double* pose_prop2 = (double*)test_mat[OUTPUT][2].data;
+    double terr1 = norm(Rt0.col(3) / norm(Rt0.col(3)) + test_mat[TEMP][3]);
+    double terr2 = norm(Rt0.col(3) / norm(Rt0.col(3)) - test_mat[TEMP][3]);
+    Mat rvec;
+    Rodrigues(Rt0.colRange(0, 3), rvec);
+    pose_prop1[0] = 0;
+    // No check for CV_LMeDS on translation. Since it
+    // involves with some degraded problem, when data is exact inliers.
+    pose_prop2[0] = method == CV_LMEDS || pt_count == 5 ? 0 : MIN(terr1, terr2);
 
 
-//    int inliers_count = countNonZero(test_mat[TEMP][1]); 
-//    int good_count = countNonZero(test_mat[TEMP][4]); 
-    test_mat[OUTPUT][3] = true; //good_count >= inliers_count / 2; 
-    test_mat[REF_OUTPUT][3] = true; 
+//    int inliers_count = countNonZero(test_mat[TEMP][1]);
+//    int good_count = countNonZero(test_mat[TEMP][4]);
+    test_mat[OUTPUT][3] = true; //good_count >= inliers_count / 2;
+    test_mat[REF_OUTPUT][3] = true;
 
 
 }
diff --git a/modules/calib3d/test/test_homography.cpp b/modules/calib3d/test/test_homography.cpp
index f68af1d7c1..5bb50bb265 100644
--- a/modules/calib3d/test/test_homography.cpp
+++ b/modules/calib3d/test/test_homography.cpp
@@ -65,7 +65,7 @@
 #define METHODS_COUNT 3
 
 int NORM_TYPE[COUNT_NORM_TYPES] = {cv::NORM_L1, cv::NORM_L2, cv::NORM_INF};
-int METHOD[METHODS_COUNT] = {0, CV_RANSAC, CV_LMEDS};
+int METHOD[METHODS_COUNT] = {0, cv::RANSAC, cv::LMEDS};
 
 using namespace cv;
 using namespace std;
@@ -309,7 +309,7 @@ void CV_HomographyTest::run(int)
             switch (method)
             {
             case 0:
-            case CV_LMEDS:
+            case LMEDS:
                 {
                     Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f, method),
                                          cv::findHomography(src_mat_2f, dst_vec, method),
@@ -339,14 +339,14 @@ void CV_HomographyTest::run(int)
 
                     continue;
                 }
-            case CV_RANSAC:
+            case RANSAC:
                 {
                     cv::Mat mask [4]; double diff;
 
-                    Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f, CV_RANSAC, reproj_threshold, mask[0]),
-                                         cv::findHomography(src_mat_2f, dst_vec, CV_RANSAC, reproj_threshold, mask[1]),
-                                         cv::findHomography(src_vec, dst_mat_2f, CV_RANSAC, reproj_threshold, mask[2]),
-                                         cv::findHomography(src_vec, dst_vec, CV_RANSAC, reproj_threshold, mask[3]) };
+                    Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f, RANSAC, reproj_threshold, mask[0]),
+                                         cv::findHomography(src_mat_2f, dst_vec, RANSAC, reproj_threshold, mask[1]),
+                                         cv::findHomography(src_vec, dst_mat_2f, RANSAC, reproj_threshold, mask[2]),
+                                         cv::findHomography(src_vec, dst_vec, RANSAC, reproj_threshold, mask[3]) };
 
                     for (int j = 0; j < 4; ++j)
                     {
@@ -411,7 +411,7 @@ void CV_HomographyTest::run(int)
             switch (method)
             {
             case 0:
-            case CV_LMEDS:
+            case LMEDS:
                 {
                     Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f),
                                          cv::findHomography(src_mat_2f, dst_vec),
@@ -466,14 +466,14 @@ void CV_HomographyTest::run(int)
 
                     continue;
                 }
-            case CV_RANSAC:
+            case RANSAC:
                 {
                     cv::Mat mask_res [4];
 
-                    Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f, CV_RANSAC, reproj_threshold, mask_res[0]),
-                                         cv::findHomography(src_mat_2f, dst_vec, CV_RANSAC, reproj_threshold, mask_res[1]),
-                                         cv::findHomography(src_vec, dst_mat_2f, CV_RANSAC, reproj_threshold, mask_res[2]),
-                                         cv::findHomography(src_vec, dst_vec, CV_RANSAC, reproj_threshold, mask_res[3]) };
+                    Mat H_res_64 [4] = { cv::findHomography(src_mat_2f, dst_mat_2f, RANSAC, reproj_threshold, mask_res[0]),
+                                         cv::findHomography(src_mat_2f, dst_vec, RANSAC, reproj_threshold, mask_res[1]),
+                                         cv::findHomography(src_vec, dst_mat_2f, RANSAC, reproj_threshold, mask_res[2]),
+                                         cv::findHomography(src_vec, dst_vec, RANSAC, reproj_threshold, mask_res[3]) };
 
                     for (int j = 0; j < 4; ++j)
                     {
diff --git a/modules/calib3d/test/test_modelest.cpp b/modules/calib3d/test/test_modelest.cpp
index 5b0a860169..55cb79ebaf 100644
--- a/modules/calib3d/test/test_modelest.cpp
+++ b/modules/calib3d/test/test_modelest.cpp
@@ -229,4 +229,3 @@ void CV_ModelEstimator2_Test::run_func()
 TEST(Calib3d_ModelEstimator2, accuracy) { CV_ModelEstimator2_Test test; test.safe_run(); }
 
 #endif
-
diff --git a/modules/calib3d/test/test_posit.cpp b/modules/calib3d/test/test_posit.cpp
index 34f3cc4d53..8a77d19ad8 100644
--- a/modules/calib3d/test/test_posit.cpp
+++ b/modules/calib3d/test/test_posit.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 using namespace cv;
 using namespace std;
@@ -189,13 +190,13 @@ void CV_POSITTest::run( int start_from )
                  rotation->data.fl, translation->data.fl );
         cvReleasePOSITObject( &object );
 
-        //Mat _rotation = cvarrToMat(rotation), _true_rotation = cvarrToMat(true_rotation);
-        //Mat _translation = cvarrToMat(translation), _true_translation = cvarrToMat(true_translation);
-        code = cvtest::cmpEps2( ts, rotation, true_rotation, flEpsilon, false, "rotation matrix" );
+        Mat _rotation = cvarrToMat(rotation), _true_rotation = cvarrToMat(true_rotation);
+        Mat _translation = cvarrToMat(translation), _true_translation = cvarrToMat(true_translation);
+        code = cvtest::cmpEps2( ts, _rotation, _true_rotation, flEpsilon, false, "rotation matrix" );
         if( code < 0 )
             break;
 
-        code = cvtest::cmpEps2( ts, translation, true_translation, flEpsilon, false, "translation vector" );
+        code = cvtest::cmpEps2( ts, _translation, _true_translation, flEpsilon, false, "translation vector" );
         if( code < 0 )
             break;
     }
diff --git a/modules/calib3d/test/test_precomp.hpp b/modules/calib3d/test/test_precomp.hpp
index fee6551f0c..d670a4c220 100644
--- a/modules/calib3d/test/test_precomp.hpp
+++ b/modules/calib3d/test/test_precomp.hpp
@@ -9,12 +9,11 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/calib3d.hpp"
 #include "opencv2/highgui.hpp"
-#include <iostream>
 
 namespace cvtest
 {
@@ -22,4 +21,3 @@ namespace cvtest
 }
 
 #endif
-
diff --git a/modules/calib3d/test/test_reproject_image_to_3d.cpp b/modules/calib3d/test/test_reproject_image_to_3d.cpp
index a93804f74e..daf89e0f0d 100644
--- a/modules/calib3d/test/test_reproject_image_to_3d.cpp
+++ b/modules/calib3d/test/test_reproject_image_to_3d.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 #include <string>
 #include <limits>
 
@@ -172,4 +173,3 @@ protected:
 };
 
 TEST(Calib3d_ReprojectImageTo3D, accuracy) { CV_ReprojectImageTo3DTest test; test.safe_run(); }
-
diff --git a/modules/calib3d/test/test_solvepnp_ransac.cpp b/modules/calib3d/test/test_solvepnp_ransac.cpp
index dc66c1dc70..ae744a4d7e 100644
--- a/modules/calib3d/test/test_solvepnp_ransac.cpp
+++ b/modules/calib3d/test/test_solvepnp_ransac.cpp
@@ -41,7 +41,10 @@
 //M*/
 
 #include "test_precomp.hpp"
-#include "opencv2/core/internal.hpp"
+
+#ifdef HAVE_TBB
+#include "tbb/task_scheduler_init.h"
+#endif
 
 using namespace cv;
 using namespace std;
@@ -51,9 +54,9 @@ class CV_solvePnPRansac_Test : public cvtest::BaseTest
 public:
     CV_solvePnPRansac_Test()
     {
-        eps[CV_ITERATIVE] = 1.0e-2;
-        eps[CV_EPNP] = 1.0e-2;
-        eps[CV_P3P] = 1.0e-2;
+        eps[ITERATIVE] = 1.0e-2;
+        eps[EPNP] = 1.0e-2;
+        eps[P3P] = 1.0e-2;
         totalTestsCount = 10;
     }
     ~CV_solvePnPRansac_Test() {}
@@ -190,9 +193,9 @@ class CV_solvePnP_Test : public CV_solvePnPRansac_Test
 public:
     CV_solvePnP_Test()
     {
-        eps[CV_ITERATIVE] = 1.0e-6;
-        eps[CV_EPNP] = 1.0e-6;
-        eps[CV_P3P] = 1.0e-4;
+        eps[ITERATIVE] = 1.0e-6;
+        eps[EPNP] = 1.0e-6;
+        eps[P3P] = 1.0e-4;
         totalTestsCount = 1000;
     }
 
@@ -236,7 +239,7 @@ protected:
     }
 };
 
-TEST(Calib3d_SolvePnPRansac, accuracy) { CV_solvePnPRansac_Test test; test.safe_run(); }
+TEST(DISABLED_Calib3d_SolvePnPRansac, accuracy) { CV_solvePnPRansac_Test test; test.safe_run(); }
 TEST(Calib3d_SolvePnP, accuracy) { CV_solvePnP_Test test; test.safe_run(); }
 
 
@@ -273,7 +276,7 @@ TEST(DISABLED_Calib3d_SolvePnPRansac, concurrency)
     {
         // limit concurrency to get determenistic result
         cv::theRNG().state = 20121010;
-        cv::Ptr<tbb::task_scheduler_init> one_thread = new tbb::task_scheduler_init(1);
+        tbb::task_scheduler_init one_thread(1);
         solvePnPRansac(object, image, camera_mat, dist_coef, rvec1, tvec1);
     }
 
@@ -292,7 +295,7 @@ TEST(DISABLED_Calib3d_SolvePnPRansac, concurrency)
     {
         // single thread again
         cv::theRNG().state = 20121010;
-        cv::Ptr<tbb::task_scheduler_init> one_thread = new tbb::task_scheduler_init(1);
+        tbb::task_scheduler_init one_thread(1);
         solvePnPRansac(object, image, camera_mat, dist_coef, rvec2, tvec2);
     }
 
@@ -303,4 +306,4 @@ TEST(DISABLED_Calib3d_SolvePnPRansac, concurrency)
     EXPECT_LT(tnorm, 1e-6);
 
 }
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/calib3d/test/test_stereomatching.cpp b/modules/calib3d/test/test_stereomatching.cpp
index 4b35dad990..8e1120e475 100644
--- a/modules/calib3d/test/test_stereomatching.cpp
+++ b/modules/calib3d/test/test_stereomatching.cpp
@@ -48,6 +48,7 @@
 #include "test_precomp.hpp"
 #include <limits>
 #include <cstdio>
+#include <map>
 
 using namespace std;
 using namespace cv;
@@ -74,12 +75,12 @@ void computeTextureBasedMasks( const Mat& _img, Mat* texturelessMask, Mat* textu
     if( !texturelessMask && !texturedMask )
         return;
     if( _img.empty() )
-        CV_Error( CV_StsBadArg, "img is empty" );
+        CV_Error( Error::StsBadArg, "img is empty" );
 
     Mat img = _img;
     if( _img.channels() > 1)
     {
-        Mat tmp; cvtColor( _img, tmp, CV_BGR2GRAY ); img = tmp;
+        Mat tmp; cvtColor( _img, tmp, COLOR_BGR2GRAY ); img = tmp;
     }
     Mat dxI; Sobel( img, dxI, CV_32FC1, 1, 0, 3 );
     Mat dxI2; pow( dxI / 8.f/*normalize*/, 2, dxI2 );
@@ -94,21 +95,21 @@ void computeTextureBasedMasks( const Mat& _img, Mat* texturelessMask, Mat* textu
 void checkTypeAndSizeOfDisp( const Mat& dispMap, const Size* sz )
 {
     if( dispMap.empty() )
-        CV_Error( CV_StsBadArg, "dispMap is empty" );
+        CV_Error( Error::StsBadArg, "dispMap is empty" );
     if( dispMap.type() != CV_32FC1 )
-        CV_Error( CV_StsBadArg, "dispMap must have CV_32FC1 type" );
+        CV_Error( Error::StsBadArg, "dispMap must have CV_32FC1 type" );
     if( sz && (dispMap.rows != sz->height || dispMap.cols != sz->width) )
-        CV_Error( CV_StsBadArg, "dispMap has incorrect size" );
+        CV_Error( Error::StsBadArg, "dispMap has incorrect size" );
 }
 
 void checkTypeAndSizeOfMask( const Mat& mask, Size sz )
 {
     if( mask.empty() )
-        CV_Error( CV_StsBadArg, "mask is empty" );
+        CV_Error( Error::StsBadArg, "mask is empty" );
     if( mask.type() != CV_8UC1 )
-        CV_Error( CV_StsBadArg, "mask must have CV_8UC1 type" );
+        CV_Error( Error::StsBadArg, "mask must have CV_8UC1 type" );
     if( mask.rows != sz.height || mask.cols != sz.width )
-        CV_Error( CV_StsBadArg, "mask has incorrect size" );
+        CV_Error( Error::StsBadArg, "mask has incorrect size" );
 }
 
 void checkDispMapsAndUnknDispMasks( const Mat& leftDispMap, const Mat& rightDispMap,
@@ -142,7 +143,7 @@ void checkDispMapsAndUnknDispMasks( const Mat& leftDispMap, const Mat& rightDisp
             minMaxLoc( rightDispMap, &rightMinVal, 0, 0, 0, ~rightUnknDispMask );
     }
     if( leftMinVal < 0 || rightMinVal < 0)
-        CV_Error( CV_StsBadArg, "known disparity values must be positive" );
+        CV_Error( Error::StsBadArg, "known disparity values must be positive" );
 }
 
 /*
@@ -162,7 +163,7 @@ void computeOcclusionBasedMasks( const Mat& leftDisp, const Mat& _rightDisp,
     if( _rightDisp.empty() )
     {
         if( !rightUnknDispMask.empty() )
-           CV_Error( CV_StsBadArg, "rightUnknDispMask must be empty if _rightDisp is empty" );
+           CV_Error( Error::StsBadArg, "rightUnknDispMask must be empty if _rightDisp is empty" );
         rightDisp.create(leftDisp.size(), CV_32FC1);
         rightDisp.setTo(Scalar::all(0) );
         for( int leftY = 0; leftY < leftDisp.rows; leftY++ )
@@ -229,9 +230,9 @@ void computeDepthDiscontMask( const Mat& disp, Mat& depthDiscontMask, const Mat&
                                  float dispGap = EVAL_DISP_GAP, int discontWidth = EVAL_DISCONT_WIDTH )
 {
     if( disp.empty() )
-        CV_Error( CV_StsBadArg, "disp is empty" );
+        CV_Error( Error::StsBadArg, "disp is empty" );
     if( disp.type() != CV_32FC1 )
-        CV_Error( CV_StsBadArg, "disp must have CV_32FC1 type" );
+        CV_Error( Error::StsBadArg, "disp must have CV_32FC1 type" );
     if( !unknDispMask.empty() )
         checkTypeAndSizeOfMask( unknDispMask, disp.size() );
 
@@ -459,14 +460,29 @@ void CV_StereoMatchingTest::run(int)
             continue;
         }
         int dispScaleFactor = datasetsParams[datasetName].dispScaleFactor;
-        Mat tmp; trueLeftDisp.convertTo( tmp, CV_32FC1, 1.f/dispScaleFactor ); trueLeftDisp = tmp; tmp.release();
+        Mat tmp;
+
+        trueLeftDisp.convertTo( tmp, CV_32FC1, 1.f/dispScaleFactor );
+        trueLeftDisp = tmp;
+        tmp.release();
+
         if( !trueRightDisp.empty() )
-            trueRightDisp.convertTo( tmp, CV_32FC1, 1.f/dispScaleFactor ); trueRightDisp = tmp; tmp.release();
+        {
+            trueRightDisp.convertTo( tmp, CV_32FC1, 1.f/dispScaleFactor );
+            trueRightDisp = tmp;
+            tmp.release();
+        }
 
         Mat leftDisp, rightDisp;
         int ignBorder = max(runStereoMatchingAlgorithm(leftImg, rightImg, leftDisp, rightDisp, ci), EVAL_IGNORE_BORDER);
-        leftDisp.convertTo( tmp, CV_32FC1 ); leftDisp = tmp; tmp.release();
-        rightDisp.convertTo( tmp, CV_32FC1 ); rightDisp = tmp; tmp.release();
+
+        leftDisp.convertTo( tmp, CV_32FC1 );
+        leftDisp = tmp;
+        tmp.release();
+
+        rightDisp.convertTo( tmp, CV_32FC1 );
+        rightDisp = tmp;
+        tmp.release();
 
         int tempCode = processStereoMatchingResults( resFS, ci, isWrite,
                    leftImg, rightImg, trueLeftDisp, trueRightDisp, leftDisp, rightDisp, QualityEvalParams(ignBorder));
@@ -530,7 +546,8 @@ int CV_StereoMatchingTest::processStereoMatchingResults( FileStorage& fs, int ca
     // rightDisp is not used in current test virsion
     int code = cvtest::TS::OK;
     assert( fs.isOpened() );
-    assert( trueLeftDisp.type() == CV_32FC1 && trueRightDisp.type() == CV_32FC1 );
+    assert( trueLeftDisp.type() == CV_32FC1 );
+    assert( trueRightDisp.empty() || trueRightDisp.type() == CV_32FC1 );
     assert( leftDisp.type() == CV_32FC1 && rightDisp.type() == CV_32FC1 );
 
     // get masks for unknown ground truth disparity values
@@ -554,9 +571,9 @@ int CV_StereoMatchingTest::processStereoMatchingResults( FileStorage& fs, int ca
     if( isWrite )
     {
         fs << caseNames[caseIdx] << "{";
-        cvWriteComment( fs.fs, RMS_STR.c_str(), 0 );
+        //cvWriteComment( fs.fs, RMS_STR.c_str(), 0 );
         writeErrors( RMS_STR, rmss, &fs );
-        cvWriteComment( fs.fs, BAD_PXLS_FRACTION_STR.c_str(), 0 );
+        //cvWriteComment( fs.fs, BAD_PXLS_FRACTION_STR.c_str(), 0 );
         writeErrors( BAD_PXLS_FRACTION_STR, badPxlsFractions, &fs );
         fs << "}"; // datasetName
     }
@@ -593,10 +610,10 @@ int CV_StereoMatchingTest::readDatasetsParams( FileStorage& fs )
     assert(fn.isSeq());
     for( int i = 0; i < (int)fn.size(); i+=3 )
     {
-        string _name = fn[i];
+        String _name = fn[i];
         DatasetParams params;
-        string sf = fn[i+1]; params.dispScaleFactor = atoi(sf.c_str());
-        string uv = fn[i+2]; params.dispUnknVal = atoi(uv.c_str());
+        String sf = fn[i+1]; params.dispScaleFactor = atoi(sf.c_str());
+        String uv = fn[i+2]; params.dispUnknVal = atoi(uv.c_str());
         datasetsParams[_name] = params;
     }
     return cvtest::TS::OK;
@@ -680,10 +697,10 @@ protected:
         assert(fn.isSeq());
         for( int i = 0; i < (int)fn.size(); i+=4 )
         {
-            string caseName = fn[i], datasetName = fn[i+1];
+            String caseName = fn[i], datasetName = fn[i+1];
             RunParams params;
-            string ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
-            string winSize = fn[i+3]; params.winSize = atoi(winSize.c_str());
+            String ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
+            String winSize = fn[i+3]; params.winSize = atoi(winSize.c_str());
             caseNames.push_back( caseName );
             caseDatasets.push_back( datasetName );
             caseRunParams.push_back( params );
@@ -697,11 +714,13 @@ protected:
         RunParams params = caseRunParams[caseIdx];
         assert( params.ndisp%16 == 0 );
         assert( _leftImg.type() == CV_8UC3 && _rightImg.type() == CV_8UC3 );
-        Mat leftImg; cvtColor( _leftImg, leftImg, CV_BGR2GRAY );
-        Mat rightImg; cvtColor( _rightImg, rightImg, CV_BGR2GRAY );
+        Mat leftImg; cvtColor( _leftImg, leftImg, COLOR_BGR2GRAY );
+        Mat rightImg; cvtColor( _rightImg, rightImg, COLOR_BGR2GRAY );
 
-        StereoBM bm( StereoBM::BASIC_PRESET, params.ndisp, params.winSize );
-        bm( leftImg, rightImg, leftDisp, CV_32F );
+        Ptr<StereoBM> bm = createStereoBM( params.ndisp, params.winSize );
+        Mat tempDisp;
+        bm->compute( leftImg, rightImg, tempDisp );
+        tempDisp.convertTo(leftDisp, CV_32F, 1./StereoMatcher::DISP_SCALE);
         return params.winSize/2;
     }
 };
@@ -734,11 +753,11 @@ protected:
         assert(fn.isSeq());
         for( int i = 0; i < (int)fn.size(); i+=5 )
         {
-            string caseName = fn[i], datasetName = fn[i+1];
+            String caseName = fn[i], datasetName = fn[i+1];
             RunParams params;
-            string ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
-            string winSize = fn[i+3]; params.winSize = atoi(winSize.c_str());
-            string fullDP = fn[i+4]; params.fullDP = atoi(fullDP.c_str()) == 0 ? false : true;
+            String ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
+            String winSize = fn[i+3]; params.winSize = atoi(winSize.c_str());
+            String fullDP = fn[i+4]; params.fullDP = atoi(fullDP.c_str()) == 0 ? false : true;
             caseNames.push_back( caseName );
             caseDatasets.push_back( datasetName );
             caseRunParams.push_back( params );
@@ -751,10 +770,13 @@ protected:
     {
         RunParams params = caseRunParams[caseIdx];
         assert( params.ndisp%16 == 0 );
-        StereoSGBM sgbm( 0, params.ndisp, params.winSize, 10*params.winSize*params.winSize, 40*params.winSize*params.winSize,
-                         1, 63, 10, 100, 32, params.fullDP );
-        sgbm( leftImg, rightImg, leftDisp );
-        assert( leftDisp.type() == CV_16SC1 );
+        Ptr<StereoSGBM> sgbm = createStereoSGBM( 0, params.ndisp, params.winSize,
+                                                 10*params.winSize*params.winSize,
+                                                 40*params.winSize*params.winSize,
+                                                 1, 63, 10, 100, 32, params.fullDP ?
+                                                 StereoSGBM::MODE_HH : StereoSGBM::MODE_SGBM );
+        sgbm->compute( leftImg, rightImg, leftDisp );
+        CV_Assert( leftDisp.type() == CV_16SC1 );
         leftDisp/=16;
         return 0;
     }
diff --git a/modules/calib3d/test/test_undistort.cpp b/modules/calib3d/test/test_undistort.cpp
index 745daeaf6d..959d8bec71 100644
--- a/modules/calib3d/test/test_undistort.cpp
+++ b/modules/calib3d/test/test_undistort.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 
 using namespace cv;
 using namespace std;
@@ -491,7 +492,7 @@ void CV_UndistortPointsTest::distortPoints(const CvMat* _src, CvMat* _dst, const
         __P = cvCreateMat(3,4,CV_64F);
     if (matP)
     {
-        cvTsConvert(matP,__P);
+        cvtest::convert(cvarrToMat(matP), cvarrToMat(__P), -1);
     }
     else
     {
@@ -500,7 +501,7 @@ void CV_UndistortPointsTest::distortPoints(const CvMat* _src, CvMat* _dst, const
         __P->data.db[4] = 1;
         __P->data.db[8] = 1;
     }
-    CvMat* __R = cvCreateMat(3,3,CV_64F);;
+    CvMat* __R = cvCreateMat(3,3,CV_64F);
     if (matR)
     {
         cvCopy(matR,__R);
@@ -847,16 +848,16 @@ void CV_InitUndistortRectifyMapTest::prepare_to_validation(int/* test_case_idx*/
     CvMat _input3 = test_mat[INPUT][3];
     CvMat _input4 = test_mat[INPUT][4];
 
-    cvTsConvert(&_input1,&_camera);
-    cvTsConvert(&_input2,&_distort);
-    cvTsConvert(&_input3,&_rot);
-    cvTsConvert(&_input4,&_new_cam);
+    cvtest::convert(cvarrToMat(&_input1), cvarrToMat(&_camera), -1);
+    cvtest::convert(cvarrToMat(&_input2), cvarrToMat(&_distort), -1);
+    cvtest::convert(cvarrToMat(&_input3), cvarrToMat(&_rot), -1);
+    cvtest::convert(cvarrToMat(&_input4), cvarrToMat(&_new_cam), -1);
 
     //Applying precalculated undistort rectify map
     if (!useCPlus)
     {
-        mapx = cv::Mat(_mapx);
-        mapy = cv::Mat(_mapy);
+        mapx = cv::cvarrToMat(_mapx);
+        mapy = cv::cvarrToMat(_mapy);
     }
     cv::Mat map1,map2;
     cv::convertMaps(mapx,mapy,map1,map2,CV_32FC1);
@@ -876,7 +877,7 @@ void CV_InitUndistortRectifyMapTest::prepare_to_validation(int/* test_case_idx*/
                       zero_distortion ? 0 : &_distort, zero_R ? 0 : &_rot, zero_new_cam ? &_camera : &_new_cam);
     //cvTsDistortPoints(&_points,&ref_points,&_camera,&_distort,&_rot,&_new_cam);
     CvMat dst = test_mat[REF_OUTPUT][0];
-    cvTsConvert(&ref_points,&dst);
+    cvtest::convert(cvarrToMat(&ref_points), cvarrToMat(&dst), -1);
 
     cvtest::copy(test_mat[INPUT][0],test_mat[OUTPUT][0]);
 
diff --git a/modules/calib3d/test/test_undistort_badarg.cpp b/modules/calib3d/test/test_undistort_badarg.cpp
index 074d193731..60460a55a0 100644
--- a/modules/calib3d/test/test_undistort_badarg.cpp
+++ b/modules/calib3d/test/test_undistort_badarg.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 
 class CV_UndistortPointsBadArgTest : public cvtest::BadArgTest
 {
@@ -243,27 +244,27 @@ void CV_UndistortPointsBadArgTest::run(int)
     //C++ tests
     useCPlus = true;
 
-    camera_mat = cv::Mat(&_camera_mat_orig);
-    distortion_coeffs = cv::Mat(&_distortion_coeffs_orig);
-    P = cv::Mat(&_P_orig);
-    R = cv::Mat(&_R_orig);
-    src_points = cv::Mat(&_src_points_orig);
+    camera_mat = cv::cvarrToMat(&_camera_mat_orig);
+    distortion_coeffs = cv::cvarrToMat(&_distortion_coeffs_orig);
+    P = cv::cvarrToMat(&_P_orig);
+    R = cv::cvarrToMat(&_R_orig);
+    src_points = cv::cvarrToMat(&_src_points_orig);
 
     temp = cvCreateMat(2,2,CV_32FC2);
-    src_points = cv::Mat(temp);
+    src_points = cv::cvarrToMat(temp);
     errcount += run_test_case( CV_StsAssert, "Invalid input data matrix size" );
-    src_points = cv::Mat(&_src_points_orig);
+    src_points = cv::cvarrToMat(&_src_points_orig);
     cvReleaseMat(&temp);
 
     temp = cvCreateMat(1,4,CV_64FC2);
-    src_points = cv::Mat(temp);
+    src_points = cv::cvarrToMat(temp);
     errcount += run_test_case( CV_StsAssert, "Invalid input data matrix type" );
-    src_points = cv::Mat(&_src_points_orig);
+    src_points = cv::cvarrToMat(&_src_points_orig);
     cvReleaseMat(&temp);
 
     src_points = cv::Mat();
     errcount += run_test_case( CV_StsAssert, "Input data matrix is not continuous" );
-    src_points = cv::Mat(&_src_points_orig);
+    src_points = cv::cvarrToMat(&_src_points_orig);
     cvReleaseMat(&temp);
 
 
@@ -360,12 +361,12 @@ void CV_InitUndistortRectifyMapBadArgTest::run(int)
     //C++ tests
     useCPlus = true;
 
-    camera_mat = cv::Mat(&_camera_mat_orig);
-    distortion_coeffs = cv::Mat(&_distortion_coeffs_orig);
-    new_camera_mat = cv::Mat(&_new_camera_mat_orig);
-    R = cv::Mat(&_R_orig);
-    mapx = cv::Mat(&_mapx_orig);
-    mapy = cv::Mat(&_mapy_orig);
+    camera_mat = cv::cvarrToMat(&_camera_mat_orig);
+    distortion_coeffs = cv::cvarrToMat(&_distortion_coeffs_orig);
+    new_camera_mat = cv::cvarrToMat(&_new_camera_mat_orig);
+    R = cv::cvarrToMat(&_R_orig);
+    mapx = cv::cvarrToMat(&_mapx_orig);
+    mapy = cv::cvarrToMat(&_mapy_orig);
 
 
     mat_type = CV_64F;
@@ -373,21 +374,21 @@ void CV_InitUndistortRectifyMapBadArgTest::run(int)
     mat_type = mat_type_orig;
 
     temp = cvCreateMat(3,2,CV_32FC1);
-    camera_mat = cv::Mat(temp);
+    camera_mat = cv::cvarrToMat(temp);
     errcount += run_test_case( CV_StsAssert, "Invalid camera data matrix size" );
-    camera_mat = cv::Mat(&_camera_mat_orig);
+    camera_mat = cv::cvarrToMat(&_camera_mat_orig);
     cvReleaseMat(&temp);
 
     temp = cvCreateMat(4,3,CV_32FC1);
-    R = cv::Mat(temp);
+    R = cv::cvarrToMat(temp);
     errcount += run_test_case( CV_StsAssert, "Invalid R data matrix size" );
-    R = cv::Mat(&_R_orig);
+    R = cv::cvarrToMat(&_R_orig);
     cvReleaseMat(&temp);
 
     temp = cvCreateMat(6,1,CV_32FC1);
-    distortion_coeffs = cv::Mat(temp);
+    distortion_coeffs = cv::cvarrToMat(temp);
     errcount += run_test_case( CV_StsAssert, "Invalid distortion coefficients data matrix size" );
-    distortion_coeffs = cv::Mat(&_distortion_coeffs_orig);
+    distortion_coeffs = cv::cvarrToMat(&_distortion_coeffs_orig);
     cvReleaseMat(&temp);
 
 //------------
@@ -499,11 +500,11 @@ void CV_UndistortBadArgTest::run(int)
     //C++ tests
     useCPlus = true;
 
-    camera_mat = cv::Mat(&_camera_mat_orig);
-    distortion_coeffs = cv::Mat(&_distortion_coeffs_orig);
-    new_camera_mat = cv::Mat(&_new_camera_mat_orig);
-    src = cv::Mat(&_src_orig);
-    dst = cv::Mat(&_dst_orig);
+    camera_mat = cv::cvarrToMat(&_camera_mat_orig);
+    distortion_coeffs = cv::cvarrToMat(&_distortion_coeffs_orig);
+    new_camera_mat = cv::cvarrToMat(&_new_camera_mat_orig);
+    src = cv::cvarrToMat(&_src_orig);
+    dst = cv::cvarrToMat(&_dst_orig);
 
 //------------
     delete[] arr_src;
diff --git a/modules/calib3d/test/test_undistort_points.cpp b/modules/calib3d/test/test_undistort_points.cpp
index c1ace3d4df..5dabd213db 100644
--- a/modules/calib3d/test/test_undistort_points.cpp
+++ b/modules/calib3d/test/test_undistort_points.cpp
@@ -94,4 +94,4 @@ void CV_UndistortTest::run(int /* start_from */)
     }
 }
 
-TEST(Calib3d_Undistort, accuracy) { CV_UndistortTest test; test.safe_run(); }
\ No newline at end of file
+TEST(Calib3d_Undistort, accuracy) { CV_UndistortTest test; test.safe_run(); }
diff --git a/modules/contrib/CMakeLists.txt b/modules/contrib/CMakeLists.txt
index 81c9ea4aa5..1d3432b598 100644
--- a/modules/contrib/CMakeLists.txt
+++ b/modules/contrib/CMakeLists.txt
@@ -1 +1 @@
-ocv_define_module(contrib opencv_imgproc opencv_calib3d opencv_features2d opencv_ml opencv_video opencv_objdetect OPTIONAL opencv_highgui)
+ocv_define_module(contrib opencv_imgproc opencv_calib3d opencv_ml opencv_video opencv_objdetect OPTIONAL opencv_highgui)
diff --git a/modules/contrib/doc/contrib.rst b/modules/contrib/doc/contrib.rst
index 798d38d5bb..de14d33efe 100644
--- a/modules/contrib/doc/contrib.rst
+++ b/modules/contrib/doc/contrib.rst
@@ -9,5 +9,4 @@ The module contains some recently added functionality that has not been stabiliz
 
     stereo
     FaceRecognizer Documentation <facerec/index>
-    Retina Documentation <retina/index>
     openfabmap
diff --git a/modules/contrib/doc/facerec/colormaps.rst b/modules/contrib/doc/facerec/colormaps.rst
index beba413fdb..95750ab5c0 100644
--- a/modules/contrib/doc/facerec/colormaps.rst
+++ b/modules/contrib/doc/facerec/colormaps.rst
@@ -51,9 +51,9 @@ In OpenCV 2.4 you only need :ocv:func:`applyColorMap` to apply a colormap on a g
     int main(int argc, const char *argv[]) {
         // Get the path to the image, if it was given
         // if no arguments were given.
-        string filename;
+        String filename;
         if (argc > 1) {
-            filename = string(argv[1]);
+            filename = String(argv[1]);
         }
         // The following lines show how to apply a colormap on a given image
         // and show it with cv::imshow example with an image. An exception is
diff --git a/modules/contrib/doc/facerec/facerec_api.rst b/modules/contrib/doc/facerec/facerec_api.rst
index 8bea7070ae..c2c7f56d0c 100644
--- a/modules/contrib/doc/facerec/facerec_api.rst
+++ b/modules/contrib/doc/facerec/facerec_api.rst
@@ -3,6 +3,12 @@ FaceRecognizer
 
 .. highlight:: cpp
 
+.. Sample code::
+
+   * An example using the FaceRecognizer class can be found at opencv_source_code/samples/cpp/facerec_demo.cpp
+
+   * (Python)  An example using the FaceRecognizer class can be found at opencv_source_code/samples/python2/facerec_demo.py
+
 FaceRecognizer
 --------------
 
@@ -30,10 +36,10 @@ a unified access to all face recongition algorithms in OpenCV. ::
       virtual void predict(InputArray src, int &label, double &confidence) const = 0;
 
       // Serializes this object to a given filename.
-      virtual void save(const string& filename) const;
+      virtual void save(const String& filename) const;
 
       // Deserializes this object from a given filename.
-      virtual void load(const string& filename);
+      virtual void load(const String& filename);
 
       // Serializes this object to a given cv::FileStorage.
       virtual void save(FileStorage& fs) const = 0;
@@ -52,7 +58,7 @@ I'll go a bit more into detail explaining :ocv:class:`FaceRecognizer`, because i
 
 * So called “virtual constructor”. That is, each Algorithm derivative is registered at program start and you can get the list of registered algorithms and create instance of a particular algorithm by its name (see :ocv:func:`Algorithm::create`). If you plan to add your own algorithms, it is good practice to add a unique prefix to your algorithms to distinguish them from other algorithms.
 
-* Setting/Retrieving algorithm parameters by name. If you used video capturing functionality from OpenCV highgui module, you are probably familar with :ocv:cfunc:`cvSetCaptureProperty`, :ocv:cfunc:`cvGetCaptureProperty`, :ocv:func:`VideoCapture::set` and :ocv:func:`VideoCapture::get`. :ocv:class:`Algorithm` provides similar method where instead of integer id's you specify the parameter names as text strings. See :ocv:func:`Algorithm::set` and :ocv:func:`Algorithm::get` for details.
+* Setting/Retrieving algorithm parameters by name. If you used video capturing functionality from OpenCV highgui module, you are probably familar with :ocv:cfunc:`cvSetCaptureProperty`, :ocv:cfunc:`cvGetCaptureProperty`, :ocv:func:`VideoCapture::set` and :ocv:func:`VideoCapture::get`. :ocv:class:`Algorithm` provides similar method where instead of integer id's you specify the parameter names as text Strings. See :ocv:func:`Algorithm::set` and :ocv:func:`Algorithm::get` for details.
 
 * Reading and writing parameters from/to XML or YAML files. Every Algorithm derivative can store all its parameters and then read them back. There is no need to re-implement it each time.
 
@@ -64,6 +70,8 @@ Moreover every :ocv:class:`FaceRecognizer` supports the:
 
 * **Loading/Saving** the model state from/to a given XML or YAML.
 
+.. note:: When using the FaceRecognizer interface in combination with Python, please stick to Python 2. Some underlying scripts like create_csv will not work in other versions, like Python 3.
+
 Setting the Thresholds
 +++++++++++++++++++++++
 
@@ -113,7 +121,7 @@ Since every :ocv:class:`FaceRecognizer` is a :ocv:class:`Algorithm`, you can use
     // Create a FaceRecognizer:
     Ptr<FaceRecognizer> model = createEigenFaceRecognizer();
     // And here's how to get its name:
-    std::string name = model->name();
+    String name = model->name();
 
 
 FaceRecognizer::train
@@ -251,7 +259,7 @@ FaceRecognizer::save
 
 Saves a :ocv:class:`FaceRecognizer` and its model state.
 
-.. ocv:function:: void FaceRecognizer::save(const string& filename) const
+.. ocv:function:: void FaceRecognizer::save(const String& filename) const
 
     Saves this model to a given filename, either as XML or YAML.
 
@@ -265,7 +273,7 @@ Saves a :ocv:class:`FaceRecognizer` and its model state.
 
 
 Every :ocv:class:`FaceRecognizer` overwrites ``FaceRecognizer::save(FileStorage& fs)``
-to save the internal model state. ``FaceRecognizer::save(const string& filename)`` saves
+to save the internal model state. ``FaceRecognizer::save(const String& filename)`` saves
 the state of a model to the given filename.
 
 The suffix ``const`` means that prediction does not affect the internal model
@@ -276,13 +284,13 @@ FaceRecognizer::load
 
 Loads a :ocv:class:`FaceRecognizer` and its model state.
 
-.. ocv:function:: void FaceRecognizer::load( const string& filename )
+.. ocv:function:: void FaceRecognizer::load( const String& filename )
 .. ocv:function:: void FaceRecognizer::load( const FileStorage& fs ) = 0
 
 Loads a persisted model and state from a given XML or YAML file . Every
 :ocv:class:`FaceRecognizer` has to overwrite ``FaceRecognizer::load(FileStorage& fs)``
 to enable loading the model state. ``FaceRecognizer::load(FileStorage& fs)`` in
-turn gets called by ``FaceRecognizer::load(const string& filename)``, to ease
+turn gets called by ``FaceRecognizer::load(const String& filename)``, to ease
 saving a model.
 
 createEigenFaceRecognizer
diff --git a/modules/contrib/doc/facerec/facerec_tutorial.rst b/modules/contrib/doc/facerec/facerec_tutorial.rst
index 170da8ff24..cbfb417971 100644
--- a/modules/contrib/doc/facerec/facerec_tutorial.rst
+++ b/modules/contrib/doc/facerec/facerec_tutorial.rst
@@ -7,7 +7,7 @@ Face Recognition with OpenCV
 Introduction
 ============
 
-`OpenCV (Open Source Computer Vision) <http://opencv.willowgarage.com>`_ is a popular computer vision library started by `Intel <http://www.intel.com>`_ in 1999. The cross-platform library sets its focus on real-time image processing and includes patent-free implementations of the latest computer vision algorithms. In 2008 `Willow Garage <http://www.willowgarage.com>`_ took over support and OpenCV 2.3.1 now comes with a programming interface to C, C++, `Python <http://www.python.org>`_ and `Android <http://www.android.com>`_. OpenCV is released under a BSD license so it is used in academic projects and commercial products alike.
+`OpenCV (Open Source Computer Vision) <http://opencv.org>`_ is a popular computer vision library started by `Intel <http://www.intel.com>`_ in 1999. The cross-platform library sets its focus on real-time image processing and includes patent-free implementations of the latest computer vision algorithms. In 2008 `Willow Garage <http://www.willowgarage.com>`_ took over support and OpenCV 2.3.1 now comes with a programming interface to C, C++, `Python <http://www.python.org>`_ and `Android <http://www.android.com>`_. OpenCV is released under a BSD license so it is used in academic projects and commercial products alike.
 
 OpenCV 2.4 now comes with the very new :ocv:class:`FaceRecognizer` class for face recognition, so you can start experimenting with face recognition right away. This document is the guide I've wished for, when I was working myself into face recognition. It shows you how to perform face recognition with :ocv:class:`FaceRecognizer` in OpenCV (with full source code listings) and gives you an introduction into the algorithms behind. I'll also show how to create the visualizations you can find in many publications, because a lot of people asked for.
 
@@ -626,5 +626,3 @@ CSV for the AT&T Facedatabase
 .. literalinclude:: etc/at.txt
    :language: none
    :linenos:
-
-
diff --git a/modules/contrib/doc/facerec/index.rst b/modules/contrib/doc/facerec/index.rst
index ce8a15e677..b871448c55 100644
--- a/modules/contrib/doc/facerec/index.rst
+++ b/modules/contrib/doc/facerec/index.rst
@@ -30,4 +30,3 @@ Indices and tables
 * :ref:`genindex`
 * :ref:`modindex`
 * :ref:`search`
-
diff --git a/modules/contrib/doc/facerec/src/CMakeLists.txt b/modules/contrib/doc/facerec/src/CMakeLists.txt
index 10720048c0..94aa36fbe0 100644
--- a/modules/contrib/doc/facerec/src/CMakeLists.txt
+++ b/modules/contrib/doc/facerec/src/CMakeLists.txt
@@ -6,7 +6,7 @@ project(facerec_cpp_samples)
 #SET(OpenCV_DIR /path/to/your/opencv/installation)
 
 # packages
-find_package(OpenCV REQUIRED) # http://opencv.willowgarage.com
+find_package(OpenCV REQUIRED) # http://opencv.org
 
 # probably you should loop through the sample files here
 add_executable(facerec_demo facerec_demo.cpp)
@@ -23,4 +23,3 @@ target_link_libraries(facerec_fisherfaces opencv_contrib opencv_core opencv_imgp
 
 add_executable(facerec_lbph facerec_lbph.cpp)
 target_link_libraries(facerec_lbph opencv_contrib opencv_core opencv_imgproc opencv_highgui)
-
diff --git a/modules/contrib/doc/facerec/src/create_csv.py b/modules/contrib/doc/facerec/src/create_csv.py
index 71d773c017..c4de778f98 100755
--- a/modules/contrib/doc/facerec/src/create_csv.py
+++ b/modules/contrib/doc/facerec/src/create_csv.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys
 import os.path
diff --git a/modules/contrib/doc/facerec/tutorial/facerec_gender_classification.rst b/modules/contrib/doc/facerec/tutorial/facerec_gender_classification.rst
index 0c7398d2fb..770083170d 100644
--- a/modules/contrib/doc/facerec/tutorial/facerec_gender_classification.rst
+++ b/modules/contrib/doc/facerec/tutorial/facerec_gender_classification.rst
@@ -231,5 +231,3 @@ Here are some examples:
 +---------------------------------+----------------------------------------------------------------------------+
 | 0.2 (20%), 0.2 (20%), (70,70)   | .. image:: ../img/tutorial/gender_classification/arnie_20_20_70_70.jpg     |
 +---------------------------------+----------------------------------------------------------------------------+
-
-
diff --git a/modules/contrib/doc/facerec/tutorial/facerec_save_load.rst b/modules/contrib/doc/facerec/tutorial/facerec_save_load.rst
index 7948bcd73a..2d0b65dffd 100644
--- a/modules/contrib/doc/facerec/tutorial/facerec_save_load.rst
+++ b/modules/contrib/doc/facerec/tutorial/facerec_save_load.rst
@@ -44,4 +44,3 @@ And here is the Reconstruction, which is the same as the original:
 
 .. image:: ../img/eigenface_reconstruction_opencv.png
     :align: center
-
diff --git a/modules/contrib/doc/facerec/tutorial/facerec_video_recognition.rst b/modules/contrib/doc/facerec/tutorial/facerec_video_recognition.rst
index b692fe5137..76e76eebe2 100644
--- a/modules/contrib/doc/facerec/tutorial/facerec_video_recognition.rst
+++ b/modules/contrib/doc/facerec/tutorial/facerec_video_recognition.rst
@@ -205,5 +205,3 @@ Here are some examples:
 +---------------------------------+----------------------------------------------------------------------------+
 | 0.2 (20%), 0.2 (20%), (70,70)   | .. image:: ../img/tutorial/gender_classification/arnie_20_20_70_70.jpg     |
 +---------------------------------+----------------------------------------------------------------------------+
-
-
diff --git a/modules/contrib/doc/openfabmap.rst b/modules/contrib/doc/openfabmap.rst
index 2f2ad40745..e2f157a075 100644
--- a/modules/contrib/doc/openfabmap.rst
+++ b/modules/contrib/doc/openfabmap.rst
@@ -9,6 +9,10 @@ FAB-MAP is an approach to appearance-based place recognition. FAB-MAP compares i
 
 openFABMAP requires training data (e.g. a collection of images from a similar but not identical environment) to construct a visual vocabulary for the visual bag-of-words model, along with a Chow-Liu tree representation of feature likelihood and for use in the Sampled new place method (see below).
 
+.. note::
+
+   * An example using the openFABMAP package can be found at opencv_source_code/samples/cpp/fabmap_sample.cpp
+
 of2::FabMap
 --------------------
 
diff --git a/modules/contrib/doc/stereo.rst b/modules/contrib/doc/stereo.rst
index d5f2d00801..103bd0f3fd 100644
--- a/modules/contrib/doc/stereo.rst
+++ b/modules/contrib/doc/stereo.rst
@@ -113,5 +113,3 @@ The method executes the variational algorithm on a rectified stereo pair. See ``
 **Note**:
 
 The method is not constant, so you should not use the same ``StereoVar`` instance from different threads simultaneously.
-
-
diff --git a/modules/contrib/include/opencv2/contrib.hpp b/modules/contrib/include/opencv2/contrib.hpp
index 44bf3313ed..75c6f3db0e 100644
--- a/modules/contrib/include/opencv2/contrib.hpp
+++ b/modules/contrib/include/opencv2/contrib.hpp
@@ -48,927 +48,591 @@
 #include "opencv2/features2d.hpp"
 #include "opencv2/objdetect.hpp"
 
-#ifdef __cplusplus
-
-/****************************************************************************************\
-*                                   Adaptive Skin Detector                               *
-\****************************************************************************************/
-
-class CV_EXPORTS CvAdaptiveSkinDetector
-{
-private:
-    enum {
-        GSD_HUE_LT = 3,
-        GSD_HUE_UT = 33,
-        GSD_INTENSITY_LT = 15,
-        GSD_INTENSITY_UT = 250
-    };
-
-    class CV_EXPORTS Histogram
-    {
-    private:
-        enum {
-            HistogramSize = (GSD_HUE_UT - GSD_HUE_LT + 1)
-        };
-
-    protected:
-        int findCoverageIndex(double surfaceToCover, int defaultValue = 0);
-
-    public:
-        CvHistogram *fHistogram;
-        Histogram();
-        virtual ~Histogram();
-
-        void findCurveThresholds(int &x1, int &x2, double percent = 0.05);
-        void mergeWith(Histogram *source, double weight);
-    };
-
-    int nStartCounter, nFrameCount, nSkinHueLowerBound, nSkinHueUpperBound, nMorphingMethod, nSamplingDivider;
-    double fHistogramMergeFactor, fHuePercentCovered;
-    Histogram histogramHueMotion, skinHueHistogram;
-    IplImage *imgHueFrame, *imgSaturationFrame, *imgLastGrayFrame, *imgMotionFrame, *imgFilteredFrame;
-    IplImage *imgShrinked, *imgTemp, *imgGrayFrame, *imgHSVFrame;
-
-protected:
-    void initData(IplImage *src, int widthDivider, int heightDivider);
-    void adaptiveFilter();
-
-public:
-
-    enum {
-        MORPHING_METHOD_NONE = 0,
-        MORPHING_METHOD_ERODE = 1,
-        MORPHING_METHOD_ERODE_ERODE = 2,
-        MORPHING_METHOD_ERODE_DILATE = 3
-    };
-
-    CvAdaptiveSkinDetector(int samplingDivider = 1, int morphingMethod = MORPHING_METHOD_NONE);
-    virtual ~CvAdaptiveSkinDetector();
-
-    virtual void process(IplImage *inputBGRImage, IplImage *outputHueMask);
-};
-
-
-/****************************************************************************************\
- *                                  Fuzzy MeanShift Tracker                               *
- \****************************************************************************************/
-
-class CV_EXPORTS CvFuzzyPoint {
-public:
-    double x, y, value;
-
-    CvFuzzyPoint(double _x, double _y);
-};
-
-class CV_EXPORTS CvFuzzyCurve {
-private:
-    std::vector<CvFuzzyPoint> points;
-    double value, centre;
-
-    bool between(double x, double x1, double x2);
-
-public:
-    CvFuzzyCurve();
-    ~CvFuzzyCurve();
-
-    void setCentre(double _centre);
-    double getCentre();
-    void clear();
-    void addPoint(double x, double y);
-    double calcValue(double param);
-    double getValue();
-    void setValue(double _value);
-};
-
-class CV_EXPORTS CvFuzzyFunction {
-public:
-    std::vector<CvFuzzyCurve> curves;
-
-    CvFuzzyFunction();
-    ~CvFuzzyFunction();
-    void addCurve(CvFuzzyCurve *curve, double value = 0);
-    void resetValues();
-    double calcValue();
-    CvFuzzyCurve *newCurve();
-};
-
-class CV_EXPORTS CvFuzzyRule {
-private:
-    CvFuzzyCurve *fuzzyInput1, *fuzzyInput2;
-    CvFuzzyCurve *fuzzyOutput;
-public:
-    CvFuzzyRule();
-    ~CvFuzzyRule();
-    void setRule(CvFuzzyCurve *c1, CvFuzzyCurve *c2, CvFuzzyCurve *o1);
-    double calcValue(double param1, double param2);
-    CvFuzzyCurve *getOutputCurve();
-};
-
-class CV_EXPORTS CvFuzzyController {
-private:
-    std::vector<CvFuzzyRule*> rules;
-public:
-    CvFuzzyController();
-    ~CvFuzzyController();
-    void addRule(CvFuzzyCurve *c1, CvFuzzyCurve *c2, CvFuzzyCurve *o1);
-    double calcOutput(double param1, double param2);
-};
-
-class CV_EXPORTS CvFuzzyMeanShiftTracker
-{
-private:
-    class FuzzyResizer
-    {
-    private:
-        CvFuzzyFunction iInput, iOutput;
-        CvFuzzyController fuzzyController;
-    public:
-        FuzzyResizer();
-        int calcOutput(double edgeDensity, double density);
-    };
-
-    class SearchWindow
-    {
-    public:
-        FuzzyResizer *fuzzyResizer;
-        int x, y;
-        int width, height, maxWidth, maxHeight, ellipseHeight, ellipseWidth;
-        int ldx, ldy, ldw, ldh, numShifts, numIters;
-        int xGc, yGc;
-        long m00, m01, m10, m11, m02, m20;
-        double ellipseAngle;
-        double density;
-        unsigned int depthLow, depthHigh;
-        int verticalEdgeLeft, verticalEdgeRight, horizontalEdgeTop, horizontalEdgeBottom;
-
-        SearchWindow();
-        ~SearchWindow();
-        void setSize(int _x, int _y, int _width, int _height);
-        void initDepthValues(IplImage *maskImage, IplImage *depthMap);
-        bool shift();
-        void extractInfo(IplImage *maskImage, IplImage *depthMap, bool initDepth);
-        void getResizeAttribsEdgeDensityLinear(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
-        void getResizeAttribsInnerDensity(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
-        void getResizeAttribsEdgeDensityFuzzy(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
-        bool meanShift(IplImage *maskImage, IplImage *depthMap, int maxIteration, bool initDepth);
-    };
-
-public:
-    enum TrackingState
-    {
-        tsNone          = 0,
-        tsSearching     = 1,
-        tsTracking      = 2,
-        tsSetWindow     = 3,
-        tsDisabled      = 10
-    };
-
-    enum ResizeMethod {
-        rmEdgeDensityLinear     = 0,
-        rmEdgeDensityFuzzy      = 1,
-        rmInnerDensity          = 2
-    };
-
-    enum {
-        MinKernelMass           = 1000
-    };
-
-    SearchWindow kernel;
-    int searchMode;
-
-private:
-    enum
-    {
-        MaxMeanShiftIteration   = 5,
-        MaxSetSizeIteration     = 5
-    };
-
-    void findOptimumSearchWindow(SearchWindow &searchWindow, IplImage *maskImage, IplImage *depthMap, int maxIteration, int resizeMethod, bool initDepth);
-
-public:
-    CvFuzzyMeanShiftTracker();
-    ~CvFuzzyMeanShiftTracker();
-
-    void track(IplImage *maskImage, IplImage *depthMap, int resizeMethod, bool resetSearch, int minKernelMass = MinKernelMass);
-};
-
-
 namespace cv
 {
-
-    class CV_EXPORTS Octree
+class CV_EXPORTS Octree
+{
+public:
+    struct Node
     {
-    public:
-        struct Node
-        {
-            Node() {}
-            int begin, end;
-            float x_min, x_max, y_min, y_max, z_min, z_max;
-            int maxLevels;
-            bool isLeaf;
-            int children[8];
-        };
-
-        Octree();
-        Octree( const std::vector<Point3f>& points, int maxLevels = 10, int minPoints = 20 );
-        virtual ~Octree();
-
-        virtual void buildTree( const std::vector<Point3f>& points, int maxLevels = 10, int minPoints = 20 );
-        virtual void getPointsWithinSphere( const Point3f& center, float radius,
-                                           std::vector<Point3f>& points ) const;
-        const std::vector<Node>& getNodes() const { return nodes; }
-    private:
-        int minPoints;
-        std::vector<Point3f> points;
-        std::vector<Node> nodes;
-
-        virtual void buildNext(size_t node_ind);
+        Node() {}
+        int begin, end;
+        float x_min, x_max, y_min, y_max, z_min, z_max;
+        int maxLevels;
+        bool isLeaf;
+        int children[8];
     };
 
+    Octree();
+    Octree( const std::vector<Point3f>& points, int maxLevels = 10, int minPoints = 20 );
+    virtual ~Octree();
 
-    class CV_EXPORTS Mesh3D
-    {
-    public:
-        struct EmptyMeshException {};
+    virtual void buildTree( const std::vector<Point3f>& points, int maxLevels = 10, int minPoints = 20 );
+    virtual void getPointsWithinSphere( const Point3f& center, float radius,
+                                       std::vector<Point3f>& points ) const;
+    const std::vector<Node>& getNodes() const { return nodes; }
+private:
+    int minPoints;
+    std::vector<Point3f> points;
+    std::vector<Node> nodes;
 
-        Mesh3D();
-        Mesh3D(const std::vector<Point3f>& vtx);
-        ~Mesh3D();
+    virtual void buildNext(size_t node_ind);
+};
 
-        void buildOctree();
-        void clearOctree();
-        float estimateResolution(float tryRatio = 0.1f);
-        void computeNormals(float normalRadius, int minNeighbors = 20);
-        void computeNormals(const std::vector<int>& subset, float normalRadius, int minNeighbors = 20);
 
-        void writeAsVrml(const std::string& file, const std::vector<Scalar>& colors = std::vector<Scalar>()) const;
+class CV_EXPORTS Mesh3D
+{
+public:
+    struct EmptyMeshException {};
 
-        std::vector<Point3f> vtx;
-        std::vector<Point3f> normals;
-        float resolution;
-        Octree octree;
+    Mesh3D();
+    Mesh3D(const std::vector<Point3f>& vtx);
+    ~Mesh3D();
 
-        const static Point3f allzero;
-    };
+    void buildOctree();
+    void clearOctree();
+    float estimateResolution(float tryRatio = 0.1f);
+    void computeNormals(float normalRadius, int minNeighbors = 20);
+    void computeNormals(const std::vector<int>& subset, float normalRadius, int minNeighbors = 20);
 
-    class CV_EXPORTS SpinImageModel
-    {
-    public:
+    void writeAsVrml(const String& file, const std::vector<Scalar>& colors = std::vector<Scalar>()) const;
 
-        /* model parameters, leave unset for default or auto estimate */
-        float normalRadius;
-        int minNeighbors;
+    std::vector<Point3f> vtx;
+    std::vector<Point3f> normals;
+    float resolution;
+    Octree octree;
 
-        float binSize;
-        int imageWidth;
+    const static Point3f allzero;
+};
 
-        float lambda;
-        float gamma;
+class CV_EXPORTS SpinImageModel
+{
+public:
 
-        float T_GeometriccConsistency;
-        float T_GroupingCorespondances;
+    /* model parameters, leave unset for default or auto estimate */
+    float normalRadius;
+    int minNeighbors;
 
-        /* public interface */
-        SpinImageModel();
-        explicit SpinImageModel(const Mesh3D& mesh);
-        ~SpinImageModel();
+    float binSize;
+    int imageWidth;
 
-        void setLogger(std::ostream* log);
-        void selectRandomSubset(float ratio);
-        void setSubset(const std::vector<int>& subset);
-        void compute();
+    float lambda;
+    float gamma;
 
-        void match(const SpinImageModel& scene, std::vector< std::vector<Vec2i> >& result);
+    float T_GeometriccConsistency;
+    float T_GroupingCorespondances;
 
-        Mat packRandomScaledSpins(bool separateScale = false, size_t xCount = 10, size_t yCount = 10) const;
+    /* public interface */
+    SpinImageModel();
+    explicit SpinImageModel(const Mesh3D& mesh);
+    ~SpinImageModel();
 
-        size_t getSpinCount() const { return spinImages.rows; }
-        Mat getSpinImage(size_t index) const { return spinImages.row((int)index); }
-        const Point3f& getSpinVertex(size_t index) const { return mesh.vtx[subset[index]]; }
-        const Point3f& getSpinNormal(size_t index) const { return mesh.normals[subset[index]]; }
+    void selectRandomSubset(float ratio);
+    void setSubset(const std::vector<int>& subset);
+    void compute();
 
-        const Mesh3D& getMesh() const { return mesh; }
-        Mesh3D& getMesh() { return mesh; }
+    void match(const SpinImageModel& scene, std::vector< std::vector<Vec2i> >& result);
 
-        /* static utility functions */
-        static bool spinCorrelation(const Mat& spin1, const Mat& spin2, float lambda, float& result);
+    Mat packRandomScaledSpins(bool separateScale = false, size_t xCount = 10, size_t yCount = 10) const;
 
-        static Point2f calcSpinMapCoo(const Point3f& point, const Point3f& vertex, const Point3f& normal);
+    size_t getSpinCount() const { return spinImages.rows; }
+    Mat getSpinImage(size_t index) const { return spinImages.row((int)index); }
+    const Point3f& getSpinVertex(size_t index) const { return mesh.vtx[subset[index]]; }
+    const Point3f& getSpinNormal(size_t index) const { return mesh.normals[subset[index]]; }
 
-        static float geometricConsistency(const Point3f& pointScene1, const Point3f& normalScene1,
-                                          const Point3f& pointModel1, const Point3f& normalModel1,
-                                          const Point3f& pointScene2, const Point3f& normalScene2,
-                                          const Point3f& pointModel2, const Point3f& normalModel2);
+    const Mesh3D& getMesh() const { return mesh; }
+    Mesh3D& getMesh() { return mesh; }
 
-        static float groupingCreteria(const Point3f& pointScene1, const Point3f& normalScene1,
+    /* static utility functions */
+    static bool spinCorrelation(const Mat& spin1, const Mat& spin2, float lambda, float& result);
+
+    static Point2f calcSpinMapCoo(const Point3f& point, const Point3f& vertex, const Point3f& normal);
+
+    static float geometricConsistency(const Point3f& pointScene1, const Point3f& normalScene1,
                                       const Point3f& pointModel1, const Point3f& normalModel1,
                                       const Point3f& pointScene2, const Point3f& normalScene2,
-                                      const Point3f& pointModel2, const Point3f& normalModel2,
-                                      float gamma);
-    protected:
-        void defaultParams();
+                                      const Point3f& pointModel2, const Point3f& normalModel2);
 
-        void matchSpinToModel(const Mat& spin, std::vector<int>& indeces,
-                              std::vector<float>& corrCoeffs, bool useExtremeOutliers = true) const;
+    static float groupingCreteria(const Point3f& pointScene1, const Point3f& normalScene1,
+                                  const Point3f& pointModel1, const Point3f& normalModel1,
+                                  const Point3f& pointScene2, const Point3f& normalScene2,
+                                  const Point3f& pointModel2, const Point3f& normalModel2,
+                                  float gamma);
+protected:
+    void defaultParams();
 
-        void repackSpinImages(const std::vector<uchar>& mask, Mat& spinImages, bool reAlloc = true) const;
+    void matchSpinToModel(const Mat& spin, std::vector<int>& indeces,
+                          std::vector<float>& corrCoeffs, bool useExtremeOutliers = true) const;
 
-        std::vector<int> subset;
-        Mesh3D mesh;
-        Mat spinImages;
-        std::ostream* out;
-    };
+    void repackSpinImages(const std::vector<uchar>& mask, Mat& spinImages, bool reAlloc = true) const;
 
-    class CV_EXPORTS TickMeter
+    std::vector<int> subset;
+    Mesh3D mesh;
+    Mat spinImages;
+};
+
+class CV_EXPORTS TickMeter
+{
+public:
+    TickMeter();
+    void start();
+    void stop();
+
+    int64 getTimeTicks() const;
+    double getTimeMicro() const;
+    double getTimeMilli() const;
+    double getTimeSec()   const;
+    int64 getCounter() const;
+
+    void reset();
+private:
+    int64 counter;
+    int64 sumTime;
+    int64 startTime;
+};
+
+//CV_EXPORTS std::ostream& operator<<(std::ostream& out, const TickMeter& tm);
+
+class CV_EXPORTS SelfSimDescriptor
+{
+public:
+    SelfSimDescriptor();
+    SelfSimDescriptor(int _ssize, int _lsize,
+                      int _startDistanceBucket=DEFAULT_START_DISTANCE_BUCKET,
+                      int _numberOfDistanceBuckets=DEFAULT_NUM_DISTANCE_BUCKETS,
+                      int _nangles=DEFAULT_NUM_ANGLES);
+    SelfSimDescriptor(const SelfSimDescriptor& ss);
+    virtual ~SelfSimDescriptor();
+    SelfSimDescriptor& operator = (const SelfSimDescriptor& ss);
+
+    size_t getDescriptorSize() const;
+    Size getGridSize( Size imgsize, Size winStride ) const;
+
+    virtual void compute(const Mat& img, std::vector<float>& descriptors, Size winStride=Size(),
+                         const std::vector<Point>& locations=std::vector<Point>()) const;
+    virtual void computeLogPolarMapping(Mat& mappingMask) const;
+    virtual void SSD(const Mat& img, Point pt, Mat& ssd) const;
+
+    int smallSize;
+    int largeSize;
+    int startDistanceBucket;
+    int numberOfDistanceBuckets;
+    int numberOfAngles;
+
+    enum { DEFAULT_SMALL_SIZE = 5, DEFAULT_LARGE_SIZE = 41,
+        DEFAULT_NUM_ANGLES = 20, DEFAULT_START_DISTANCE_BUCKET = 3,
+        DEFAULT_NUM_DISTANCE_BUCKETS = 7 };
+};
+
+
+CV_EXPORTS_W int chamerMatching( Mat& img, Mat& templ,
+                              CV_OUT std::vector<std::vector<Point> >& results, CV_OUT std::vector<float>& cost,
+                              double templScale=1, int maxMatches = 20,
+                              double minMatchDistance = 1.0, int padX = 3,
+                              int padY = 3, int scales = 5, double minScale = 0.6, double maxScale = 1.6,
+                              double orientationWeight = 0.5, double truncate = 20);
+
+
+class CV_EXPORTS_W StereoVar
+{
+public:
+    // Flags
+    enum {USE_INITIAL_DISPARITY = 1, USE_EQUALIZE_HIST = 2, USE_SMART_ID = 4, USE_AUTO_PARAMS = 8, USE_MEDIAN_FILTERING = 16};
+    enum {CYCLE_O, CYCLE_V};
+    enum {PENALIZATION_TICHONOV, PENALIZATION_CHARBONNIER, PENALIZATION_PERONA_MALIK};
+
+    //! the default constructor
+    CV_WRAP StereoVar();
+
+    //! the full constructor taking all the necessary algorithm parameters
+    CV_WRAP StereoVar(int levels, double pyrScale, int nIt, int minDisp, int maxDisp, int poly_n, double poly_sigma, float fi, float lambda, int penalization, int cycle, int flags);
+
+    //! the destructor
+    virtual ~StereoVar();
+
+    //! the stereo correspondence operator that computes disparity map for the specified rectified stereo pair
+    CV_WRAP_AS(compute) virtual void operator()(const Mat& left, const Mat& right, CV_OUT Mat& disp);
+
+    CV_PROP_RW int      levels;
+    CV_PROP_RW double   pyrScale;
+    CV_PROP_RW int      nIt;
+    CV_PROP_RW int      minDisp;
+    CV_PROP_RW int      maxDisp;
+    CV_PROP_RW int      poly_n;
+    CV_PROP_RW double   poly_sigma;
+    CV_PROP_RW float    fi;
+    CV_PROP_RW float    lambda;
+    CV_PROP_RW int      penalization;
+    CV_PROP_RW int      cycle;
+    CV_PROP_RW int      flags;
+
+private:
+    void autoParams();
+    void FMG(Mat &I1, Mat &I2, Mat &I2x, Mat &u, int level);
+    void VCycle_MyFAS(Mat &I1_h, Mat &I2_h, Mat &I2x_h, Mat &u_h, int level);
+    void VariationalSolver(Mat &I1_h, Mat &I2_h, Mat &I2x_h, Mat &u_h, int level);
+};
+
+CV_EXPORTS void polyfit(const Mat& srcx, const Mat& srcy, Mat& dst, int order);
+
+class CV_EXPORTS Directory
+{
+    public:
+        static std::vector<String> GetListFiles  ( const String& path, const String & exten = "*", bool addPath = true );
+        static std::vector<String> GetListFilesR ( const String& path, const String & exten = "*", bool addPath = true );
+        static std::vector<String> GetListFolders( const String& path, const String & exten = "*", bool addPath = true );
+};
+
+/*
+ * Generation of a set of different colors by the following way:
+ * 1) generate more then need colors (in "factor" times) in RGB,
+ * 2) convert them to Lab,
+ * 3) choose the needed count of colors from the set that are more different from
+ *    each other,
+ * 4) convert the colors back to RGB
+ */
+CV_EXPORTS void generateColors( std::vector<Scalar>& colors, size_t count, size_t factor=100 );
+
+
+/*
+ *  Estimate the rigid body motion from frame0 to frame1. The method is based on the paper
+ *  "Real-Time Visual Odometry from Dense RGB-D Images", F. Steinbucker, J. Strum, D. Cremers, ICCV, 2011.
+ */
+enum { ROTATION          = 1,
+       TRANSLATION       = 2,
+       RIGID_BODY_MOTION = 4
+     };
+CV_EXPORTS bool RGBDOdometry( Mat& Rt, const Mat& initRt,
+                              const Mat& image0, const Mat& depth0, const Mat& mask0,
+                              const Mat& image1, const Mat& depth1, const Mat& mask1,
+                              const Mat& cameraMatrix, float minDepth=0.f, float maxDepth=4.f, float maxDepthDiff=0.07f,
+                              const std::vector<int>& iterCounts=std::vector<int>(),
+                              const std::vector<float>& minGradientMagnitudes=std::vector<float>(),
+                              int transformType=RIGID_BODY_MOTION );
+
+/**
+*Bilinear interpolation technique.
+*
+*The value of a desired cortical pixel is obtained through a bilinear interpolation of the values
+*of the four nearest neighbouring Cartesian pixels to the center of the RF.
+*The same principle is applied to the inverse transformation.
+*
+*More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
+*/
+class CV_EXPORTS LogPolar_Interp
+{
+public:
+
+    LogPolar_Interp() {}
+
+    /**
+    *Constructor
+    *\param w the width of the input image
+    *\param h the height of the input image
+    *\param center the transformation center: where the output precision is maximal
+    *\param R the number of rings of the cortical image (default value 70 pixel)
+    *\param ro0 the radius of the blind spot (default value 3 pixel)
+    *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
+    *            \a 0 means that the retinal image is computed within the inscribed circle.
+    *\param S the number of sectors of the cortical image (default value 70 pixel).
+    *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
+    *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
+    *          \a 0 means that the parameter \a S is provided by the user.
+    */
+    LogPolar_Interp(int w, int h, Point2i center, int R=70, double ro0=3.0,
+                    int interp=INTER_LINEAR, int full=1, int S=117, int sp=1);
+    /**
+    *Transformation from Cartesian image to cortical (log-polar) image.
+    *\param source the Cartesian image
+    *\return the transformed image (cortical image)
+    */
+    const Mat to_cortical(const Mat &source);
+    /**
+    *Transformation from cortical image to retinal (inverse log-polar) image.
+    *\param source the cortical image
+    *\return the transformed image (retinal image)
+    */
+    const Mat to_cartesian(const Mat &source);
+    /**
+    *Destructor
+    */
+    ~LogPolar_Interp();
+
+protected:
+
+    Mat Rsri;
+    Mat Csri;
+
+    int S, R, M, N;
+    int top, bottom,left,right;
+    double ro0, romax, a, q;
+    int interp;
+
+    Mat ETAyx;
+    Mat CSIyx;
+
+    void create_map(int M, int N, int R, int S, double ro0);
+};
+
+/**
+*Overlapping circular receptive fields technique
+*
+*The Cartesian plane is divided in two regions: the fovea and the periphery.
+*The fovea (oversampling) is handled by using the bilinear interpolation technique described above, whereas in
+*the periphery we use the overlapping Gaussian circular RFs.
+*
+*More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
+*/
+class CV_EXPORTS LogPolar_Overlapping
+{
+public:
+    LogPolar_Overlapping() {}
+
+    /**
+    *Constructor
+    *\param w the width of the input image
+    *\param h the height of the input image
+    *\param center the transformation center: where the output precision is maximal
+    *\param R the number of rings of the cortical image (default value 70 pixel)
+    *\param ro0 the radius of the blind spot (default value 3 pixel)
+    *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
+    *            \a 0 means that the retinal image is computed within the inscribed circle.
+    *\param S the number of sectors of the cortical image (default value 70 pixel).
+    *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
+    *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
+    *          \a 0 means that the parameter \a S is provided by the user.
+    */
+    LogPolar_Overlapping(int w, int h, Point2i center, int R=70,
+                         double ro0=3.0, int full=1, int S=117, int sp=1);
+    /**
+    *Transformation from Cartesian image to cortical (log-polar) image.
+    *\param source the Cartesian image
+    *\return the transformed image (cortical image)
+    */
+    const Mat to_cortical(const Mat &source);
+    /**
+    *Transformation from cortical image to retinal (inverse log-polar) image.
+    *\param source the cortical image
+    *\return the transformed image (retinal image)
+    */
+    const Mat to_cartesian(const Mat &source);
+    /**
+    *Destructor
+    */
+    ~LogPolar_Overlapping();
+
+protected:
+
+    Mat Rsri;
+    Mat Csri;
+    std::vector<int> Rsr;
+    std::vector<int> Csr;
+    std::vector<double> Wsr;
+
+    int S, R, M, N, ind1;
+    int top, bottom,left,right;
+    double ro0, romax, a, q;
+
+    struct kernel
     {
-    public:
-        TickMeter();
-        void start();
-        void stop();
-
-        int64 getTimeTicks() const;
-        double getTimeMicro() const;
-        double getTimeMilli() const;
-        double getTimeSec()   const;
-        int64 getCounter() const;
-
-        void reset();
-    private:
-        int64 counter;
-        int64 sumTime;
-        int64 startTime;
+        kernel() { w = 0; }
+        std::vector<double> weights;
+        int w;
     };
 
-    CV_EXPORTS std::ostream& operator<<(std::ostream& out, const TickMeter& tm);
+    Mat ETAyx;
+    Mat CSIyx;
+    std::vector<kernel> w_ker_2D;
 
-    class CV_EXPORTS SelfSimDescriptor
-    {
-    public:
-        SelfSimDescriptor();
-        SelfSimDescriptor(int _ssize, int _lsize,
-                          int _startDistanceBucket=DEFAULT_START_DISTANCE_BUCKET,
-                          int _numberOfDistanceBuckets=DEFAULT_NUM_DISTANCE_BUCKETS,
-                          int _nangles=DEFAULT_NUM_ANGLES);
-        SelfSimDescriptor(const SelfSimDescriptor& ss);
-        virtual ~SelfSimDescriptor();
-        SelfSimDescriptor& operator = (const SelfSimDescriptor& ss);
+    void create_map(int M, int N, int R, int S, double ro0);
+};
 
-        size_t getDescriptorSize() const;
-        Size getGridSize( Size imgsize, Size winStride ) const;
+/**
+* Adjacent receptive fields technique
+*
+*All the Cartesian pixels, whose coordinates in the cortical domain share the same integer part, are assigned to the same RF.
+*The precision of the boundaries of the RF can be improved by breaking each pixel into subpixels and assigning each of them to the correct RF.
+*This technique is implemented from: Traver, V., Pla, F.: Log-polar mapping template design: From task-level requirements
+*to geometry parameters. Image Vision Comput. 26(10) (2008) 1354-1370
+*
+*More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
+*/
+class CV_EXPORTS LogPolar_Adjacent
+{
+public:
+    LogPolar_Adjacent() {}
 
-        virtual void compute(const Mat& img, std::vector<float>& descriptors, Size winStride=Size(),
-                             const std::vector<Point>& locations=std::vector<Point>()) const;
-        virtual void computeLogPolarMapping(Mat& mappingMask) const;
-        virtual void SSD(const Mat& img, Point pt, Mat& ssd) const;
-
-        int smallSize;
-        int largeSize;
-        int startDistanceBucket;
-        int numberOfDistanceBuckets;
-        int numberOfAngles;
-
-        enum { DEFAULT_SMALL_SIZE = 5, DEFAULT_LARGE_SIZE = 41,
-            DEFAULT_NUM_ANGLES = 20, DEFAULT_START_DISTANCE_BUCKET = 3,
-            DEFAULT_NUM_DISTANCE_BUCKETS = 7 };
-    };
-
-
-    typedef bool (*BundleAdjustCallback)(int iteration, double norm_error, void* user_data);
-
-    class CV_EXPORTS LevMarqSparse {
-    public:
-        LevMarqSparse();
-        LevMarqSparse(int npoints, // number of points
-                      int ncameras, // number of cameras
-                      int nPointParams, // number of params per one point  (3 in case of 3D points)
-                      int nCameraParams, // number of parameters per one camera
-                      int nErrParams, // number of parameters in measurement vector
-                      // for 1 point at one camera (2 in case of 2D projections)
-                      Mat& visibility, // visibility matrix. rows correspond to points, columns correspond to cameras
-                      // 1 - point is visible for the camera, 0 - invisible
-                      Mat& P0, // starting vector of parameters, first cameras then points
-                      Mat& X, // measurements, in order of visibility. non visible cases are skipped
-                      TermCriteria criteria, // termination criteria
-
-                      // callback for estimation of Jacobian matrices
-                      void (CV_CDECL * fjac)(int i, int j, Mat& point_params,
-                                             Mat& cam_params, Mat& A, Mat& B, void* data),
-                      // callback for estimation of backprojection errors
-                      void (CV_CDECL * func)(int i, int j, Mat& point_params,
-                                             Mat& cam_params, Mat& estim, void* data),
-                      void* data, // user-specific data passed to the callbacks
-                      BundleAdjustCallback cb, void* user_data
-                      );
-
-        virtual ~LevMarqSparse();
-
-        virtual void run( int npoints, // number of points
-                         int ncameras, // number of cameras
-                         int nPointParams, // number of params per one point  (3 in case of 3D points)
-                         int nCameraParams, // number of parameters per one camera
-                         int nErrParams, // number of parameters in measurement vector
-                         // for 1 point at one camera (2 in case of 2D projections)
-                         Mat& visibility, // visibility matrix. rows correspond to points, columns correspond to cameras
-                         // 1 - point is visible for the camera, 0 - invisible
-                         Mat& P0, // starting vector of parameters, first cameras then points
-                         Mat& X, // measurements, in order of visibility. non visible cases are skipped
-                         TermCriteria criteria, // termination criteria
-
-                         // callback for estimation of Jacobian matrices
-                         void (CV_CDECL * fjac)(int i, int j, Mat& point_params,
-                                                Mat& cam_params, Mat& A, Mat& B, void* data),
-                         // callback for estimation of backprojection errors
-                         void (CV_CDECL * func)(int i, int j, Mat& point_params,
-                                                Mat& cam_params, Mat& estim, void* data),
-                         void* data // user-specific data passed to the callbacks
-                         );
-
-        virtual void clear();
-
-        // useful function to do simple bundle adjustment tasks
-        static void bundleAdjust(std::vector<Point3d>& points, // positions of points in global coordinate system (input and output)
-                                 const std::vector<std::vector<Point2d> >& imagePoints, // projections of 3d points for every camera
-                                 const std::vector<std::vector<int> >& visibility, // visibility of 3d points for every camera
-                                 std::vector<Mat>& cameraMatrix, // intrinsic matrices of all cameras (input and output)
-                                 std::vector<Mat>& R, // rotation matrices of all cameras (input and output)
-                                 std::vector<Mat>& T, // translation vector of all cameras (input and output)
-                                 std::vector<Mat>& distCoeffs, // distortion coefficients of all cameras (input and output)
-                                 const TermCriteria& criteria=
-                                 TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, DBL_EPSILON),
-                                 BundleAdjustCallback cb = 0, void* user_data = 0);
-
-    public:
-        virtual void optimize(CvMat &_vis); //main function that runs minimization
-
-        //iteratively asks for measurement for visible camera-point pairs
-        void ask_for_proj(CvMat &_vis,bool once=false);
-        //iteratively asks for Jacobians for every camera_point pair
-        void ask_for_projac(CvMat &_vis);
-
-        CvMat* err; //error X-hX
-        double prevErrNorm, errNorm;
-        double lambda;
-        CvTermCriteria criteria;
-        int iters;
-
-        CvMat** U; //size of array is equal to number of cameras
-        CvMat** V; //size of array is equal to number of points
-        CvMat** inv_V_star; //inverse of V*
-
-        CvMat** A;
-        CvMat** B;
-        CvMat** W;
-
-        CvMat* X; //measurement
-        CvMat* hX; //current measurement extimation given new parameter vector
-
-        CvMat* prevP; //current already accepted parameter.
-        CvMat* P; // parameters used to evaluate function with new params
-        // this parameters may be rejected
-
-        CvMat* deltaP; //computed increase of parameters (result of normal system solution )
-
-        CvMat** ea; // sum_i  AijT * e_ij , used as right part of normal equation
-        // length of array is j = number of cameras
-        CvMat** eb; // sum_j  BijT * e_ij , used as right part of normal equation
-        // length of array is i = number of points
-
-        CvMat** Yj; //length of array is i = num_points
-
-        CvMat* S; //big matrix of block Sjk  , each block has size num_cam_params x num_cam_params
-
-        CvMat* JtJ_diag; //diagonal of JtJ,  used to backup diagonal elements before augmentation
-
-        CvMat* Vis_index; // matrix which element is index of measurement for point i and camera j
-
-        int num_cams;
-        int num_points;
-        int num_err_param;
-        int num_cam_param;
-        int num_point_param;
-
-        //target function and jacobian pointers, which needs to be initialized
-        void (*fjac)(int i, int j, Mat& point_params, Mat& cam_params, Mat& A, Mat& B, void* data);
-        void (*func)(int i, int j, Mat& point_params, Mat& cam_params, Mat& estim, void* data);
-
-        void* data;
-
-        BundleAdjustCallback cb;
-        void* user_data;
-    };
-
-    CV_EXPORTS_W int chamerMatching( Mat& img, Mat& templ,
-                                  CV_OUT std::vector<std::vector<Point> >& results, CV_OUT std::vector<float>& cost,
-                                  double templScale=1, int maxMatches = 20,
-                                  double minMatchDistance = 1.0, int padX = 3,
-                                  int padY = 3, int scales = 5, double minScale = 0.6, double maxScale = 1.6,
-                                  double orientationWeight = 0.5, double truncate = 20);
-
-
-    class CV_EXPORTS_W StereoVar
-    {
-    public:
-        // Flags
-        enum {USE_INITIAL_DISPARITY = 1, USE_EQUALIZE_HIST = 2, USE_SMART_ID = 4, USE_AUTO_PARAMS = 8, USE_MEDIAN_FILTERING = 16};
-        enum {CYCLE_O, CYCLE_V};
-        enum {PENALIZATION_TICHONOV, PENALIZATION_CHARBONNIER, PENALIZATION_PERONA_MALIK};
-
-        //! the default constructor
-        CV_WRAP StereoVar();
-
-        //! the full constructor taking all the necessary algorithm parameters
-        CV_WRAP StereoVar(int levels, double pyrScale, int nIt, int minDisp, int maxDisp, int poly_n, double poly_sigma, float fi, float lambda, int penalization, int cycle, int flags);
-
-        //! the destructor
-        virtual ~StereoVar();
-
-        //! the stereo correspondence operator that computes disparity map for the specified rectified stereo pair
-        CV_WRAP_AS(compute) virtual void operator()(const Mat& left, const Mat& right, CV_OUT Mat& disp);
-
-        CV_PROP_RW int      levels;
-        CV_PROP_RW double   pyrScale;
-        CV_PROP_RW int      nIt;
-        CV_PROP_RW int      minDisp;
-        CV_PROP_RW int      maxDisp;
-        CV_PROP_RW int      poly_n;
-        CV_PROP_RW double   poly_sigma;
-        CV_PROP_RW float    fi;
-        CV_PROP_RW float    lambda;
-        CV_PROP_RW int      penalization;
-        CV_PROP_RW int      cycle;
-        CV_PROP_RW int      flags;
-
-    private:
-        void autoParams();
-        void FMG(Mat &I1, Mat &I2, Mat &I2x, Mat &u, int level);
-        void VCycle_MyFAS(Mat &I1_h, Mat &I2_h, Mat &I2x_h, Mat &u_h, int level);
-        void VariationalSolver(Mat &I1_h, Mat &I2_h, Mat &I2x_h, Mat &u_h, int level);
-    };
-
-    CV_EXPORTS void polyfit(const Mat& srcx, const Mat& srcy, Mat& dst, int order);
-
-    class CV_EXPORTS Directory
-    {
-        public:
-            static std::vector<std::string> GetListFiles  ( const std::string& path, const std::string & exten = "*", bool addPath = true );
-            static std::vector<std::string> GetListFilesR ( const std::string& path, const std::string & exten = "*", bool addPath = true );
-            static std::vector<std::string> GetListFolders( const std::string& path, const std::string & exten = "*", bool addPath = true );
-    };
-
-    /*
-     * Generation of a set of different colors by the following way:
-     * 1) generate more then need colors (in "factor" times) in RGB,
-     * 2) convert them to Lab,
-     * 3) choose the needed count of colors from the set that are more different from
-     *    each other,
-     * 4) convert the colors back to RGB
+    /**
+     *Constructor
+     *\param w the width of the input image
+     *\param h the height of the input image
+     *\param center the transformation center: where the output precision is maximal
+     *\param R the number of rings of the cortical image (default value 70 pixel)
+     *\param ro0 the radius of the blind spot (default value 3 pixel)
+     *\param smin the size of the subpixel (default value 0.25 pixel)
+     *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
+     *            \a 0 means that the retinal image is computed within the inscribed circle.
+     *\param S the number of sectors of the cortical image (default value 70 pixel).
+     *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
+     *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
+     *          \a 0 means that the parameter \a S is provided by the user.
      */
-    CV_EXPORTS void generateColors( std::vector<Scalar>& colors, size_t count, size_t factor=100 );
-
-
-    /*
-     *  Estimate the rigid body motion from frame0 to frame1. The method is based on the paper
-     *  "Real-Time Visual Odometry from Dense RGB-D Images", F. Steinbucker, J. Strum, D. Cremers, ICCV, 2011.
+    LogPolar_Adjacent(int w, int h, Point2i center, int R=70, double ro0=3.0, double smin=0.25, int full=1, int S=117, int sp=1);
+    /**
+     *Transformation from Cartesian image to cortical (log-polar) image.
+     *\param source the Cartesian image
+     *\return the transformed image (cortical image)
      */
-    enum { ROTATION          = 1,
-           TRANSLATION       = 2,
-           RIGID_BODY_MOTION = 4
-         };
-    CV_EXPORTS bool RGBDOdometry( Mat& Rt, const Mat& initRt,
-                                  const Mat& image0, const Mat& depth0, const Mat& mask0,
-                                  const Mat& image1, const Mat& depth1, const Mat& mask1,
-                                  const Mat& cameraMatrix, float minDepth=0.f, float maxDepth=4.f, float maxDepthDiff=0.07f,
-                                  const std::vector<int>& iterCounts=std::vector<int>(),
-                                  const std::vector<float>& minGradientMagnitudes=std::vector<float>(),
-                                  int transformType=RIGID_BODY_MOTION );
-
+    const Mat to_cortical(const Mat &source);
     /**
-    *Bilinear interpolation technique.
-    *
-    *The value of a desired cortical pixel is obtained through a bilinear interpolation of the values
-    *of the four nearest neighbouring Cartesian pixels to the center of the RF.
-    *The same principle is applied to the inverse transformation.
-    *
-    *More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
-    */
-    class CV_EXPORTS LogPolar_Interp
-    {
-    public:
-
-        LogPolar_Interp() {}
-
-        /**
-        *Constructor
-        *\param w the width of the input image
-        *\param h the height of the input image
-        *\param center the transformation center: where the output precision is maximal
-        *\param R the number of rings of the cortical image (default value 70 pixel)
-        *\param ro0 the radius of the blind spot (default value 3 pixel)
-        *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
-        *            \a 0 means that the retinal image is computed within the inscribed circle.
-        *\param S the number of sectors of the cortical image (default value 70 pixel).
-        *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
-        *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
-        *          \a 0 means that the parameter \a S is provided by the user.
-        */
-        LogPolar_Interp(int w, int h, Point2i center, int R=70, double ro0=3.0,
-                        int interp=INTER_LINEAR, int full=1, int S=117, int sp=1);
-        /**
-        *Transformation from Cartesian image to cortical (log-polar) image.
-        *\param source the Cartesian image
-        *\return the transformed image (cortical image)
-        */
-        const Mat to_cortical(const Mat &source);
-        /**
-        *Transformation from cortical image to retinal (inverse log-polar) image.
-        *\param source the cortical image
-        *\return the transformed image (retinal image)
-        */
-        const Mat to_cartesian(const Mat &source);
-        /**
-        *Destructor
-        */
-        ~LogPolar_Interp();
-
-    protected:
-
-        Mat Rsri;
-        Mat Csri;
-
-        int S, R, M, N;
-        int top, bottom,left,right;
-        double ro0, romax, a, q;
-        int interp;
-
-        Mat ETAyx;
-        Mat CSIyx;
-
-        void create_map(int M, int N, int R, int S, double ro0);
-    };
-
+     *Transformation from cortical image to retinal (inverse log-polar) image.
+     *\param source the cortical image
+     *\return the transformed image (retinal image)
+     */
+    const Mat to_cartesian(const Mat &source);
     /**
-    *Overlapping circular receptive fields technique
-    *
-    *The Cartesian plane is divided in two regions: the fovea and the periphery.
-    *The fovea (oversampling) is handled by using the bilinear interpolation technique described above, whereas in
-    *the periphery we use the overlapping Gaussian circular RFs.
-    *
-    *More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
-    */
-    class CV_EXPORTS LogPolar_Overlapping
+     *Destructor
+     */
+    ~LogPolar_Adjacent();
+
+protected:
+    struct pixel
     {
-    public:
-        LogPolar_Overlapping() {}
-
-        /**
-        *Constructor
-        *\param w the width of the input image
-        *\param h the height of the input image
-        *\param center the transformation center: where the output precision is maximal
-        *\param R the number of rings of the cortical image (default value 70 pixel)
-        *\param ro0 the radius of the blind spot (default value 3 pixel)
-        *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
-        *            \a 0 means that the retinal image is computed within the inscribed circle.
-        *\param S the number of sectors of the cortical image (default value 70 pixel).
-        *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
-        *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
-        *          \a 0 means that the parameter \a S is provided by the user.
-        */
-        LogPolar_Overlapping(int w, int h, Point2i center, int R=70,
-                             double ro0=3.0, int full=1, int S=117, int sp=1);
-        /**
-        *Transformation from Cartesian image to cortical (log-polar) image.
-        *\param source the Cartesian image
-        *\return the transformed image (cortical image)
-        */
-        const Mat to_cortical(const Mat &source);
-        /**
-        *Transformation from cortical image to retinal (inverse log-polar) image.
-        *\param source the cortical image
-        *\return the transformed image (retinal image)
-        */
-        const Mat to_cartesian(const Mat &source);
-        /**
-        *Destructor
-        */
-        ~LogPolar_Overlapping();
-
-    protected:
-
-        Mat Rsri;
-        Mat Csri;
-        std::vector<int> Rsr;
-        std::vector<int> Csr;
-        std::vector<double> Wsr;
-
-        int S, R, M, N, ind1;
-        int top, bottom,left,right;
-        double ro0, romax, a, q;
-
-        struct kernel
-        {
-            kernel() { w = 0; }
-            std::vector<double> weights;
-            int w;
-        };
-
-        Mat ETAyx;
-        Mat CSIyx;
-        std::vector<kernel> w_ker_2D;
-
-        void create_map(int M, int N, int R, int S, double ro0);
+        pixel() { u = v = 0; a = 0.; }
+        int u;
+        int v;
+        double a;
     };
+    int S, R, M, N;
+    int top, bottom,left,right;
+    double ro0, romax, a, q;
+    std::vector<std::vector<pixel> > L;
+    std::vector<double> A;
 
-    /**
-    * Adjacent receptive fields technique
-    *
-    *All the Cartesian pixels, whose coordinates in the cortical domain share the same integer part, are assigned to the same RF.
-    *The precision of the boundaries of the RF can be improved by breaking each pixel into subpixels and assigning each of them to the correct RF.
-    *This technique is implemented from: Traver, V., Pla, F.: Log-polar mapping template design: From task-level requirements
-    *to geometry parameters. Image Vision Comput. 26(10) (2008) 1354-1370
-    *
-    *More details can be found in http://dx.doi.org/10.1007/978-3-642-23968-7_5
-    */
-    class CV_EXPORTS LogPolar_Adjacent
+    void subdivide_recursively(double x, double y, int i, int j, double length, double smin);
+    bool get_uv(double x, double y, int&u, int&v);
+    void create_map(int M, int N, int R, int S, double ro0, double smin);
+};
+
+CV_EXPORTS Mat subspaceProject(InputArray W, InputArray mean, InputArray src);
+CV_EXPORTS Mat subspaceReconstruct(InputArray W, InputArray mean, InputArray src);
+
+class CV_EXPORTS LDA
+{
+public:
+    // Initializes a LDA with num_components (default 0) and specifies how
+    // samples are aligned (default dataAsRow=true).
+    LDA(int num_components = 0) :
+        _num_components(num_components) {};
+
+    // Initializes and performs a Discriminant Analysis with Fisher's
+    // Optimization Criterion on given data in src and corresponding labels
+    // in labels. If 0 (or less) number of components are given, they are
+    // automatically determined for given data in computation.
+    LDA(InputArrayOfArrays src, InputArray labels,
+            int num_components = 0) :
+                _num_components(num_components)
     {
-    public:
-        LogPolar_Adjacent() {}
+        this->compute(src, labels); //! compute eigenvectors and eigenvalues
+    }
 
-        /**
-         *Constructor
-         *\param w the width of the input image
-         *\param h the height of the input image
-         *\param center the transformation center: where the output precision is maximal
-         *\param R the number of rings of the cortical image (default value 70 pixel)
-         *\param ro0 the radius of the blind spot (default value 3 pixel)
-         *\param smin the size of the subpixel (default value 0.25 pixel)
-         *\param full \a 1 (default value) means that the retinal image (the inverse transform) is computed within the circumscribing circle.
-         *            \a 0 means that the retinal image is computed within the inscribed circle.
-         *\param S the number of sectors of the cortical image (default value 70 pixel).
-         *         Its value is usually internally computed to obtain a pixel aspect ratio equals to 1.
-         *\param sp \a 1 (default value) means that the parameter \a S is internally computed.
-         *          \a 0 means that the parameter \a S is provided by the user.
-         */
-        LogPolar_Adjacent(int w, int h, Point2i center, int R=70, double ro0=3.0, double smin=0.25, int full=1, int S=117, int sp=1);
-        /**
-         *Transformation from Cartesian image to cortical (log-polar) image.
-         *\param source the Cartesian image
-         *\return the transformed image (cortical image)
-         */
-        const Mat to_cortical(const Mat &source);
-        /**
-         *Transformation from cortical image to retinal (inverse log-polar) image.
-         *\param source the cortical image
-         *\return the transformed image (retinal image)
-         */
-        const Mat to_cartesian(const Mat &source);
-        /**
-         *Destructor
-         */
-        ~LogPolar_Adjacent();
+    // Serializes this object to a given filename.
+    void save(const String& filename) const;
 
-    protected:
-        struct pixel
-        {
-            pixel() { u = v = 0; a = 0.; }
-            int u;
-            int v;
-            double a;
-        };
-        int S, R, M, N;
-        int top, bottom,left,right;
-        double ro0, romax, a, q;
-        std::vector<std::vector<pixel> > L;
-        std::vector<double> A;
+    // Deserializes this object from a given filename.
+    void load(const String& filename);
 
-        void subdivide_recursively(double x, double y, int i, int j, double length, double smin);
-        bool get_uv(double x, double y, int&u, int&v);
-        void create_map(int M, int N, int R, int S, double ro0, double smin);
-    };
-
-    CV_EXPORTS Mat subspaceProject(InputArray W, InputArray mean, InputArray src);
-    CV_EXPORTS Mat subspaceReconstruct(InputArray W, InputArray mean, InputArray src);
-
-    class CV_EXPORTS LDA
-    {
-    public:
-        // Initializes a LDA with num_components (default 0) and specifies how
-        // samples are aligned (default dataAsRow=true).
-        LDA(int num_components = 0) :
-            _num_components(num_components) {};
-
-        // Initializes and performs a Discriminant Analysis with Fisher's
-        // Optimization Criterion on given data in src and corresponding labels
-        // in labels. If 0 (or less) number of components are given, they are
-        // automatically determined for given data in computation.
-        LDA(InputArrayOfArrays src, InputArray labels,
-                int num_components = 0) :
-                    _num_components(num_components)
-        {
-            this->compute(src, labels); //! compute eigenvectors and eigenvalues
-        }
-
-        // Serializes this object to a given filename.
-        void save(const std::string& filename) const;
-
-        // Deserializes this object from a given filename.
-        void load(const std::string& filename);
-
-        // Serializes this object to a given cv::FileStorage.
-        void save(FileStorage& fs) const;
-
-            // Deserializes this object from a given cv::FileStorage.
-        void load(const FileStorage& node);
-
-        // Destructor.
-        ~LDA() {}
-
-        //! Compute the discriminants for data in src and labels.
-        void compute(InputArrayOfArrays src, InputArray labels);
-
-        // Projects samples into the LDA subspace.
-        Mat project(InputArray src);
-
-        // Reconstructs projections from the LDA subspace.
-        Mat reconstruct(InputArray src);
-
-        // Returns the eigenvectors of this LDA.
-        Mat eigenvectors() const { return _eigenvectors; };
-
-        // Returns the eigenvalues of this LDA.
-        Mat eigenvalues() const { return _eigenvalues; }
-
-    protected:
-        bool _dataAsRow;
-        int _num_components;
-        Mat _eigenvectors;
-        Mat _eigenvalues;
-
-        void lda(InputArrayOfArrays src, InputArray labels);
-    };
-
-    class CV_EXPORTS_W FaceRecognizer : public Algorithm
-    {
-    public:
-        //! virtual destructor
-        virtual ~FaceRecognizer() {}
-
-        // Trains a FaceRecognizer.
-        CV_WRAP virtual void train(InputArrayOfArrays src, InputArray labels) = 0;
-
-        // Updates a FaceRecognizer.
-        CV_WRAP virtual void update(InputArrayOfArrays src, InputArray labels);
-
-        // Gets a prediction from a FaceRecognizer.
-        virtual int predict(InputArray src) const = 0;
-
-        // Predicts the label and confidence for a given sample.
-        CV_WRAP virtual void predict(InputArray src, CV_OUT int &label, CV_OUT double &confidence) const = 0;
-
-        // Serializes this object to a given filename.
-        CV_WRAP virtual void save(const std::string& filename) const;
-
-        // Deserializes this object from a given filename.
-        CV_WRAP virtual void load(const std::string& filename);
-
-        // Serializes this object to a given cv::FileStorage.
-        virtual void save(FileStorage& fs) const = 0;
+    // Serializes this object to a given cv::FileStorage.
+    void save(FileStorage& fs) const;
 
         // Deserializes this object from a given cv::FileStorage.
-        virtual void load(const FileStorage& fs) = 0;
+    void load(const FileStorage& node);
 
-    };
+    // Destructor.
+    ~LDA() {}
 
-    CV_EXPORTS_W Ptr<FaceRecognizer> createEigenFaceRecognizer(int num_components = 0, double threshold = DBL_MAX);
-    CV_EXPORTS_W Ptr<FaceRecognizer> createFisherFaceRecognizer(int num_components = 0, double threshold = DBL_MAX);
-    CV_EXPORTS_W Ptr<FaceRecognizer> createLBPHFaceRecognizer(int radius=1, int neighbors=8,
-                                                            int grid_x=8, int grid_y=8, double threshold = DBL_MAX);
+    //! Compute the discriminants for data in src and labels.
+    void compute(InputArrayOfArrays src, InputArray labels);
 
-    enum
-    {
-        COLORMAP_AUTUMN = 0,
-        COLORMAP_BONE = 1,
-        COLORMAP_JET = 2,
-        COLORMAP_WINTER = 3,
-        COLORMAP_RAINBOW = 4,
-        COLORMAP_OCEAN = 5,
-        COLORMAP_SUMMER = 6,
-        COLORMAP_SPRING = 7,
-        COLORMAP_COOL = 8,
-        COLORMAP_HSV = 9,
-        COLORMAP_PINK = 10,
-        COLORMAP_HOT = 11
-    };
+    // Projects samples into the LDA subspace.
+    Mat project(InputArray src);
 
-    CV_EXPORTS_W void applyColorMap(InputArray src, OutputArray dst, int colormap);
+    // Reconstructs projections from the LDA subspace.
+    Mat reconstruct(InputArray src);
 
-    CV_EXPORTS bool initModule_contrib();
+    // Returns the eigenvectors of this LDA.
+    Mat eigenvectors() const { return _eigenvectors; };
+
+    // Returns the eigenvalues of this LDA.
+    Mat eigenvalues() const { return _eigenvalues; }
+
+protected:
+    bool _dataAsRow;
+    int _num_components;
+    Mat _eigenvectors;
+    Mat _eigenvalues;
+
+    void lda(InputArrayOfArrays src, InputArray labels);
+};
+
+class CV_EXPORTS_W FaceRecognizer : public Algorithm
+{
+public:
+    //! virtual destructor
+    virtual ~FaceRecognizer() {}
+
+    // Trains a FaceRecognizer.
+    CV_WRAP virtual void train(InputArrayOfArrays src, InputArray labels) = 0;
+
+    // Updates a FaceRecognizer.
+    CV_WRAP virtual void update(InputArrayOfArrays src, InputArray labels);
+
+    // Gets a prediction from a FaceRecognizer.
+    virtual int predict(InputArray src) const = 0;
+
+    // Predicts the label and confidence for a given sample.
+    CV_WRAP virtual void predict(InputArray src, CV_OUT int &label, CV_OUT double &confidence) const = 0;
+
+    // Serializes this object to a given filename.
+    CV_WRAP virtual void save(const String& filename) const;
+
+    // Deserializes this object from a given filename.
+    CV_WRAP virtual void load(const String& filename);
+
+    // Serializes this object to a given cv::FileStorage.
+    virtual void save(FileStorage& fs) const = 0;
+
+    // Deserializes this object from a given cv::FileStorage.
+    virtual void load(const FileStorage& fs) = 0;
+
+};
+
+CV_EXPORTS_W Ptr<FaceRecognizer> createEigenFaceRecognizer(int num_components = 0, double threshold = DBL_MAX);
+CV_EXPORTS_W Ptr<FaceRecognizer> createFisherFaceRecognizer(int num_components = 0, double threshold = DBL_MAX);
+CV_EXPORTS_W Ptr<FaceRecognizer> createLBPHFaceRecognizer(int radius=1, int neighbors=8,
+                                                        int grid_x=8, int grid_y=8, double threshold = DBL_MAX);
+
+enum
+{
+    COLORMAP_AUTUMN = 0,
+    COLORMAP_BONE = 1,
+    COLORMAP_JET = 2,
+    COLORMAP_WINTER = 3,
+    COLORMAP_RAINBOW = 4,
+    COLORMAP_OCEAN = 5,
+    COLORMAP_SUMMER = 6,
+    COLORMAP_SPRING = 7,
+    COLORMAP_COOL = 8,
+    COLORMAP_HSV = 9,
+    COLORMAP_PINK = 10,
+    COLORMAP_HOT = 11
+};
+
+CV_EXPORTS_W void applyColorMap(InputArray src, OutputArray dst, int colormap);
+
+CV_EXPORTS bool initModule_contrib();
 }
 
-#include "opencv2/contrib/retina.hpp"
-
 #include "opencv2/contrib/openfabmap.hpp"
 
 #endif
-
-#endif
diff --git a/modules/contrib/include/opencv2/contrib/compat.hpp b/modules/contrib/include/opencv2/contrib/compat.hpp
new file mode 100644
index 0000000000..ba758c235f
--- /dev/null
+++ b/modules/contrib/include/opencv2/contrib/compat.hpp
@@ -0,0 +1,384 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CONTRIB_COMPAT_HPP__
+#define __OPENCV_CONTRIB_COMPAT_HPP__
+
+#include "opencv2/core/core_c.h"
+
+#ifdef __cplusplus
+
+/****************************************************************************************\
+*                                   Adaptive Skin Detector                               *
+\****************************************************************************************/
+
+class CV_EXPORTS CvAdaptiveSkinDetector
+{
+private:
+    enum {
+        GSD_HUE_LT = 3,
+        GSD_HUE_UT = 33,
+        GSD_INTENSITY_LT = 15,
+        GSD_INTENSITY_UT = 250
+    };
+
+    class CV_EXPORTS Histogram
+    {
+    private:
+        enum {
+            HistogramSize = (GSD_HUE_UT - GSD_HUE_LT + 1)
+        };
+
+    protected:
+        int findCoverageIndex(double surfaceToCover, int defaultValue = 0);
+
+    public:
+        CvHistogram *fHistogram;
+        Histogram();
+        virtual ~Histogram();
+
+        void findCurveThresholds(int &x1, int &x2, double percent = 0.05);
+        void mergeWith(Histogram *source, double weight);
+    };
+
+    int nStartCounter, nFrameCount, nSkinHueLowerBound, nSkinHueUpperBound, nMorphingMethod, nSamplingDivider;
+    double fHistogramMergeFactor, fHuePercentCovered;
+    Histogram histogramHueMotion, skinHueHistogram;
+    IplImage *imgHueFrame, *imgSaturationFrame, *imgLastGrayFrame, *imgMotionFrame, *imgFilteredFrame;
+    IplImage *imgShrinked, *imgTemp, *imgGrayFrame, *imgHSVFrame;
+
+protected:
+    void initData(IplImage *src, int widthDivider, int heightDivider);
+    void adaptiveFilter();
+
+public:
+
+    enum {
+        MORPHING_METHOD_NONE = 0,
+        MORPHING_METHOD_ERODE = 1,
+        MORPHING_METHOD_ERODE_ERODE = 2,
+        MORPHING_METHOD_ERODE_DILATE = 3
+    };
+
+    CvAdaptiveSkinDetector(int samplingDivider = 1, int morphingMethod = MORPHING_METHOD_NONE);
+    virtual ~CvAdaptiveSkinDetector();
+
+    virtual void process(IplImage *inputBGRImage, IplImage *outputHueMask);
+};
+
+
+/****************************************************************************************\
+ *                                  Fuzzy MeanShift Tracker                               *
+ \****************************************************************************************/
+
+class CV_EXPORTS CvFuzzyPoint {
+public:
+    double x, y, value;
+
+    CvFuzzyPoint(double _x, double _y);
+};
+
+class CV_EXPORTS CvFuzzyCurve {
+private:
+    std::vector<CvFuzzyPoint> points;
+    double value, centre;
+
+    bool between(double x, double x1, double x2);
+
+public:
+    CvFuzzyCurve();
+    ~CvFuzzyCurve();
+
+    void setCentre(double _centre);
+    double getCentre();
+    void clear();
+    void addPoint(double x, double y);
+    double calcValue(double param);
+    double getValue();
+    void setValue(double _value);
+};
+
+class CV_EXPORTS CvFuzzyFunction {
+public:
+    std::vector<CvFuzzyCurve> curves;
+
+    CvFuzzyFunction();
+    ~CvFuzzyFunction();
+    void addCurve(CvFuzzyCurve *curve, double value = 0);
+    void resetValues();
+    double calcValue();
+    CvFuzzyCurve *newCurve();
+};
+
+class CV_EXPORTS CvFuzzyRule {
+private:
+    CvFuzzyCurve *fuzzyInput1, *fuzzyInput2;
+    CvFuzzyCurve *fuzzyOutput;
+public:
+    CvFuzzyRule();
+    ~CvFuzzyRule();
+    void setRule(CvFuzzyCurve *c1, CvFuzzyCurve *c2, CvFuzzyCurve *o1);
+    double calcValue(double param1, double param2);
+    CvFuzzyCurve *getOutputCurve();
+};
+
+class CV_EXPORTS CvFuzzyController {
+private:
+    std::vector<CvFuzzyRule*> rules;
+public:
+    CvFuzzyController();
+    ~CvFuzzyController();
+    void addRule(CvFuzzyCurve *c1, CvFuzzyCurve *c2, CvFuzzyCurve *o1);
+    double calcOutput(double param1, double param2);
+};
+
+class CV_EXPORTS CvFuzzyMeanShiftTracker
+{
+private:
+    class FuzzyResizer
+    {
+    private:
+        CvFuzzyFunction iInput, iOutput;
+        CvFuzzyController fuzzyController;
+    public:
+        FuzzyResizer();
+        int calcOutput(double edgeDensity, double density);
+    };
+
+    class SearchWindow
+    {
+    public:
+        FuzzyResizer *fuzzyResizer;
+        int x, y;
+        int width, height, maxWidth, maxHeight, ellipseHeight, ellipseWidth;
+        int ldx, ldy, ldw, ldh, numShifts, numIters;
+        int xGc, yGc;
+        long m00, m01, m10, m11, m02, m20;
+        double ellipseAngle;
+        double density;
+        unsigned int depthLow, depthHigh;
+        int verticalEdgeLeft, verticalEdgeRight, horizontalEdgeTop, horizontalEdgeBottom;
+
+        SearchWindow();
+        ~SearchWindow();
+        void setSize(int _x, int _y, int _width, int _height);
+        void initDepthValues(IplImage *maskImage, IplImage *depthMap);
+        bool shift();
+        void extractInfo(IplImage *maskImage, IplImage *depthMap, bool initDepth);
+        void getResizeAttribsEdgeDensityLinear(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
+        void getResizeAttribsInnerDensity(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
+        void getResizeAttribsEdgeDensityFuzzy(int &resizeDx, int &resizeDy, int &resizeDw, int &resizeDh);
+        bool meanShift(IplImage *maskImage, IplImage *depthMap, int maxIteration, bool initDepth);
+    };
+
+public:
+    enum TrackingState
+    {
+        tsNone          = 0,
+        tsSearching     = 1,
+        tsTracking      = 2,
+        tsSetWindow     = 3,
+        tsDisabled      = 10
+    };
+
+    enum ResizeMethod {
+        rmEdgeDensityLinear     = 0,
+        rmEdgeDensityFuzzy      = 1,
+        rmInnerDensity          = 2
+    };
+
+    enum {
+        MinKernelMass           = 1000
+    };
+
+    SearchWindow kernel;
+    int searchMode;
+
+private:
+    enum
+    {
+        MaxMeanShiftIteration   = 5,
+        MaxSetSizeIteration     = 5
+    };
+
+    void findOptimumSearchWindow(SearchWindow &searchWindow, IplImage *maskImage, IplImage *depthMap, int maxIteration, int resizeMethod, bool initDepth);
+
+public:
+    CvFuzzyMeanShiftTracker();
+    ~CvFuzzyMeanShiftTracker();
+
+    void track(IplImage *maskImage, IplImage *depthMap, int resizeMethod, bool resetSearch, int minKernelMass = MinKernelMass);
+};
+
+
+namespace cv
+{
+
+typedef bool (*BundleAdjustCallback)(int iteration, double norm_error, void* user_data);
+
+class CV_EXPORTS LevMarqSparse {
+public:
+    LevMarqSparse();
+    LevMarqSparse(int npoints, // number of points
+                  int ncameras, // number of cameras
+                  int nPointParams, // number of params per one point  (3 in case of 3D points)
+                  int nCameraParams, // number of parameters per one camera
+                  int nErrParams, // number of parameters in measurement vector
+                  // for 1 point at one camera (2 in case of 2D projections)
+                  Mat& visibility, // visibility matrix. rows correspond to points, columns correspond to cameras
+                  // 1 - point is visible for the camera, 0 - invisible
+                  Mat& P0, // starting vector of parameters, first cameras then points
+                  Mat& X, // measurements, in order of visibility. non visible cases are skipped
+                  TermCriteria criteria, // termination criteria
+
+                  // callback for estimation of Jacobian matrices
+                  void (*fjac)(int i, int j, Mat& point_params,
+                                         Mat& cam_params, Mat& A, Mat& B, void* data),
+                  // callback for estimation of backprojection errors
+                  void (*func)(int i, int j, Mat& point_params,
+                                         Mat& cam_params, Mat& estim, void* data),
+                  void* data, // user-specific data passed to the callbacks
+                  BundleAdjustCallback cb, void* user_data
+                  );
+
+    virtual ~LevMarqSparse();
+
+    virtual void run( int npoints, // number of points
+                     int ncameras, // number of cameras
+                     int nPointParams, // number of params per one point  (3 in case of 3D points)
+                     int nCameraParams, // number of parameters per one camera
+                     int nErrParams, // number of parameters in measurement vector
+                     // for 1 point at one camera (2 in case of 2D projections)
+                     Mat& visibility, // visibility matrix. rows correspond to points, columns correspond to cameras
+                     // 1 - point is visible for the camera, 0 - invisible
+                     Mat& P0, // starting vector of parameters, first cameras then points
+                     Mat& X, // measurements, in order of visibility. non visible cases are skipped
+                     TermCriteria criteria, // termination criteria
+
+                     // callback for estimation of Jacobian matrices
+                     void (CV_CDECL * fjac)(int i, int j, Mat& point_params,
+                                            Mat& cam_params, Mat& A, Mat& B, void* data),
+                     // callback for estimation of backprojection errors
+                     void (CV_CDECL * func)(int i, int j, Mat& point_params,
+                                            Mat& cam_params, Mat& estim, void* data),
+                     void* data // user-specific data passed to the callbacks
+                     );
+
+    virtual void clear();
+
+    // useful function to do simple bundle adjustment tasks
+    static void bundleAdjust(std::vector<Point3d>& points, // positions of points in global coordinate system (input and output)
+                             const std::vector<std::vector<Point2d> >& imagePoints, // projections of 3d points for every camera
+                             const std::vector<std::vector<int> >& visibility, // visibility of 3d points for every camera
+                             std::vector<Mat>& cameraMatrix, // intrinsic matrices of all cameras (input and output)
+                             std::vector<Mat>& R, // rotation matrices of all cameras (input and output)
+                             std::vector<Mat>& T, // translation vector of all cameras (input and output)
+                             std::vector<Mat>& distCoeffs, // distortion coefficients of all cameras (input and output)
+                             const TermCriteria& criteria=
+                             TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, DBL_EPSILON),
+                             BundleAdjustCallback cb = 0, void* user_data = 0);
+
+public:
+    virtual void optimize(CvMat &_vis); //main function that runs minimization
+
+    //iteratively asks for measurement for visible camera-point pairs
+    void ask_for_proj(CvMat &_vis,bool once=false);
+    //iteratively asks for Jacobians for every camera_point pair
+    void ask_for_projac(CvMat &_vis);
+
+    CvMat* err; //error X-hX
+    double prevErrNorm, errNorm;
+    double lambda;
+    CvTermCriteria criteria;
+    int iters;
+
+    CvMat** U; //size of array is equal to number of cameras
+    CvMat** V; //size of array is equal to number of points
+    CvMat** inv_V_star; //inverse of V*
+
+    CvMat** A;
+    CvMat** B;
+    CvMat** W;
+
+    CvMat* X; //measurement
+    CvMat* hX; //current measurement extimation given new parameter vector
+
+    CvMat* prevP; //current already accepted parameter.
+    CvMat* P; // parameters used to evaluate function with new params
+    // this parameters may be rejected
+
+    CvMat* deltaP; //computed increase of parameters (result of normal system solution )
+
+    CvMat** ea; // sum_i  AijT * e_ij , used as right part of normal equation
+    // length of array is j = number of cameras
+    CvMat** eb; // sum_j  BijT * e_ij , used as right part of normal equation
+    // length of array is i = number of points
+
+    CvMat** Yj; //length of array is i = num_points
+
+    CvMat* S; //big matrix of block Sjk  , each block has size num_cam_params x num_cam_params
+
+    CvMat* JtJ_diag; //diagonal of JtJ,  used to backup diagonal elements before augmentation
+
+    CvMat* Vis_index; // matrix which element is index of measurement for point i and camera j
+
+    int num_cams;
+    int num_points;
+    int num_err_param;
+    int num_cam_param;
+    int num_point_param;
+
+    //target function and jacobian pointers, which needs to be initialized
+    void (*fjac)(int i, int j, Mat& point_params, Mat& cam_params, Mat& A, Mat& B, void* data);
+    void (*func)(int i, int j, Mat& point_params, Mat& cam_params, Mat& estim, void* data);
+
+    void* data;
+
+    BundleAdjustCallback cb;
+    void* user_data;
+};
+
+} // cv
+
+#endif /* __cplusplus */
+
+#endif /* __OPENCV_CONTRIB_COMPAT_HPP__ */
diff --git a/modules/contrib/include/opencv2/contrib/contrib.hpp b/modules/contrib/include/opencv2/contrib/contrib.hpp
index 7a43631fce..b9edf94f59 100644
--- a/modules/contrib/include/opencv2/contrib/contrib.hpp
+++ b/modules/contrib/include/opencv2/contrib/contrib.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/contrib.hpp"
\ No newline at end of file
+#include "opencv2/contrib.hpp"
diff --git a/modules/contrib/include/opencv2/contrib/detection_based_tracker.hpp b/modules/contrib/include/opencv2/contrib/detection_based_tracker.hpp
index c3db03ab9f..c11904ea19 100644
--- a/modules/contrib/include/opencv2/contrib/detection_based_tracker.hpp
+++ b/modules/contrib/include/opencv2/contrib/detection_based_tracker.hpp
@@ -1,6 +1,6 @@
 #pragma once
 
-#if defined(__linux__) || defined(LINUX) || defined(__APPLE__) || defined(ANDROID)
+#if defined(__linux__) || defined(LINUX) || defined(__APPLE__) || defined(__ANDROID__)
 
 #include <opencv2/core.hpp>
 #include <opencv2/objdetect.hpp>
@@ -9,7 +9,7 @@
 
 namespace cv
 {
-class DetectionBasedTracker
+class CV_EXPORTS DetectionBasedTracker
 {
     public:
         struct Parameters
diff --git a/modules/contrib/src/adaptiveskindetector.cpp b/modules/contrib/src/adaptiveskindetector.cpp
index a44c42077f..092e48cb47 100644
--- a/modules/contrib/src/adaptiveskindetector.cpp
+++ b/modules/contrib/src/adaptiveskindetector.cpp
@@ -35,6 +35,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/contrib/compat.hpp"
 
 #define ASD_INTENSITY_SET_PIXEL(pointer, qq) {(*pointer) = (unsigned char)qq;}
 
@@ -284,5 +286,3 @@ void CvAdaptiveSkinDetector::Histogram::mergeWith(CvAdaptiveSkinDetector::Histog
         }
     }
 };
-
-
diff --git a/modules/contrib/src/ba.cpp b/modules/contrib/src/ba.cpp
index 39acd45798..9a8002e96f 100644
--- a/modules/contrib/src/ba.cpp
+++ b/modules/contrib/src/ba.cpp
@@ -41,6 +41,8 @@
 
 #include "precomp.hpp"
 #include "opencv2/calib3d.hpp"
+#include "opencv2/contrib/compat.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 #include <iostream>
 
 using namespace cv;
@@ -359,7 +361,7 @@ void LevMarqSparse::ask_for_proj(CvMat &/*_vis*/,bool once) {
                 cvGetSubRect( P, &cam_mat, cvRect( 0, j * num_cam_param, 1, num_cam_param ));
                 CvMat measur_mat;
                 cvGetSubRect( hX, &measur_mat, cvRect( 0, ind * num_err_param, 1, num_err_param ));
-                Mat _point_mat(&point_mat), _cam_mat(&cam_mat), _measur_mat(&measur_mat);
+                Mat _point_mat = cv::cvarrToMat(&point_mat), _cam_mat = cv::cvarrToMat(&cam_mat), _measur_mat = cv::cvarrToMat(&measur_mat);
                 func( i, j, _point_mat, _cam_mat, _measur_mat, data);
                 assert( ind*num_err_param == ((int*)(Vis_index->data.ptr + i * Vis_index->step))[j]);
                 ind+=1;
@@ -398,7 +400,7 @@ void LevMarqSparse::ask_for_projac(CvMat &/*_vis*/)   //should be evaluated at p
 
                 //CvMat* Bij = B_line[j];
                 //CvMat* Bij = ((CvMat**)(B->data.ptr + B->step * i))[j];
-                Mat _point_mat(&point_mat), _cam_mat(&cam_mat), _Aij(Aij), _Bij(Bij);
+                Mat _point_mat = cv::cvarrToMat(&point_mat), _cam_mat = cv::cvarrToMat(&cam_mat), _Aij = cv::cvarrToMat(Aij), _Bij = cv::cvarrToMat(Bij);
                 (*fjac)(i, j, _point_mat, _cam_mat, _Aij, _Bij, data);
             }
         }
@@ -744,7 +746,7 @@ static void fjac(int /*i*/, int /*j*/, CvMat *point_params, CvMat* cam_params, C
   CvMat* _mp = cvCreateMat(1, 1, CV_64FC2 ); //projection of the point
 
   //split camera params into different matrices
-  CvMat _ri, _ti, _k;
+  CvMat _ri, _ti, _k = cvMat(0, 0, CV_64F, NULL); // dummy initialization to fix warning of cl.exe
   cvGetRows( cam_params, &_ri, 0, 3 );
   cvGetRows( cam_params, &_ti, 3, 6 );
 
@@ -1100,16 +1102,17 @@ void LevMarqSparse::bundleAdjust( std::vector<Point3d>& points, //positions of p
   }
   //fill camera params
   //R.clear();T.clear();cameraMatrix.clear();
+  Mat levmarP = cv::cvarrToMat(levmar.P);
   for( int i = 0; i < num_cameras; i++ ) {
     //rotation
-    Mat rot_vec = Mat(levmar.P).rowRange(i*num_cam_param, i*num_cam_param+3);
+    Mat rot_vec = levmarP.rowRange(i*num_cam_param, i*num_cam_param+3);
     Rodrigues( rot_vec, R[i] );
     //translation
-    T[i] = Mat(levmar.P).rowRange(i*num_cam_param + 3, i*num_cam_param+6);
+    levmarP.rowRange(i*num_cam_param + 3, i*num_cam_param+6).copyTo(T[i]);
 
     //intrinsic camera matrix
     double* intr_data = (double*)cameraMatrix[i].data;
-    double* intr = (double*)(Mat(levmar.P).data + Mat(levmar.P).step * (i*num_cam_param+6));
+    double* intr = (double*)(levmarP.data +levmarP.step * (i*num_cam_param+6));
     //focals
     intr_data[0] = intr[0];  //fx
     intr_data[4] = intr[1];  //fy
@@ -1119,7 +1122,7 @@ void LevMarqSparse::bundleAdjust( std::vector<Point3d>& points, //positions of p
 
     //add distortion if exists
     if( distCoeffs.size() ) {
-      Mat(levmar.P).rowRange(i*num_cam_param + 10, i*num_cam_param+10+numdist).copyTo(distCoeffs[i]);
+      levmarP.rowRange(i*num_cam_param + 10, i*num_cam_param+10+numdist).copyTo(distCoeffs[i]);
     }
   }
 }
diff --git a/modules/contrib/src/bowmsctrainer.cpp b/modules/contrib/src/bowmsctrainer.cpp
index c1e945f9ba..3057691204 100644
--- a/modules/contrib/src/bowmsctrainer.cpp
+++ b/modules/contrib/src/bowmsctrainer.cpp
@@ -136,4 +136,3 @@ Mat BOWMSCTrainer::cluster(const Mat& _descriptors) const {
 }
 
 }
-
diff --git a/modules/contrib/src/chamfermatching.cpp b/modules/contrib/src/chamfermatching.cpp
index bd30ad9916..fd2899c5f0 100644
--- a/modules/contrib/src/chamfermatching.cpp
+++ b/modules/contrib/src/chamfermatching.cpp
@@ -142,7 +142,7 @@ private:
         LocationScaleImageRange(const std::vector<Point>& locations, const std::vector<float>& _scales) :
         locations_(locations), scales_(_scales)
         {
-            assert(locations.size()==_scales.size());
+            CV_Assert(locations.size()==_scales.size());
         }
 
         ImageIterator* iterator() const
@@ -393,7 +393,7 @@ private:
         LocationScaleImageIterator(const std::vector<Point>& locations, const std::vector<float>& _scales) :
         locations_(locations), scales_(_scales)
         {
-            assert(locations.size()==_scales.size());
+            CV_Assert(locations.size()==_scales.size());
             reset();
         }
 
@@ -622,7 +622,7 @@ void ChamferMatcher::Matching::followContour(Mat& templ_img, template_coords_t&
     coordinate_t next;
     unsigned char ptr;
 
-    assert (direction==-1 || !coords.empty());
+    CV_Assert (direction==-1 || !coords.empty());
 
     coordinate_t crt = coords.back();
 
@@ -903,19 +903,18 @@ void ChamferMatcher::Template::show() const
             p2.x = x + pad*(int)(sin(orientations[i])*100)/100;
             p2.y = y + pad*(int)(cos(orientations[i])*100)/100;
 
-            line(templ_color, p1,p2, CV_RGB(255,0,0));
+            line(templ_color, p1,p2, Scalar(255,0,0));
         }
     }
 
-    circle(templ_color,Point(center.x + pad, center.y + pad),1,CV_RGB(0,255,0));
+    circle(templ_color,Point(center.x + pad, center.y + pad),1,Scalar(0,255,0));
 
 #ifdef HAVE_OPENCV_HIGHGUI
     namedWindow("templ",1);
     imshow("templ",templ_color);
-
-    cvWaitKey(0);
+    waitKey();
 #else
-    CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without GUI support");
+    CV_Error(Error::StsNotImplemented, "OpenCV has been compiled without GUI support");
 #endif
 
     templ_color.release();
@@ -1060,7 +1059,7 @@ void ChamferMatcher::Matching::fillNonContourOrientations(Mat& annotated_img, Ma
     int cols = annotated_img.cols;
     int rows = annotated_img.rows;
 
-    assert(orientation_img.cols==cols && orientation_img.rows==rows);
+    CV_Assert(orientation_img.cols==cols && orientation_img.rows==rows);
 
     for (int y=0;y<rows;++y) {
         for (int x=0;x<cols;++x) {
@@ -1280,7 +1279,7 @@ void ChamferMatcher::showMatch(Mat& img, int index)
         std::cout << "Index too big.\n" << std::endl;
     }
 
-    assert(img.channels()==3);
+    CV_Assert(img.channels()==3);
 
     Match match = matches[index];
 
@@ -1299,7 +1298,7 @@ void ChamferMatcher::showMatch(Mat& img, int index)
 
 void ChamferMatcher::showMatch(Mat& img, Match match)
 {
-    assert(img.channels()==3);
+    CV_Assert(img.channels()==3);
 
     const template_coords_t& templ_coords = match.tpl->coords;
     for (size_t i=0;i<templ_coords.size();++i) {
diff --git a/modules/contrib/src/chowliutree.cpp b/modules/contrib/src/chowliutree.cpp
index 8c6acabdcd..d4aed5383a 100644
--- a/modules/contrib/src/chowliutree.cpp
+++ b/modules/contrib/src/chowliutree.cpp
@@ -287,4 +287,3 @@ bool ChowLiuTree::reduceEdgesToMinSpan(std::list<info>& edges) {
 }
 
 }
-
diff --git a/modules/contrib/src/colormap.cpp b/modules/contrib/src/colormap.cpp
index ca98b3a918..08ff44a5c6 100644
--- a/modules/contrib/src/colormap.cpp
+++ b/modules/contrib/src/colormap.cpp
@@ -40,7 +40,7 @@ static Mat linspace(float x0, float x1, int n)
 static void sortMatrixRowsByIndices(InputArray _src, InputArray _indices, OutputArray _dst)
 {
     if(_indices.getMat().type() != CV_32SC1)
-        CV_Error(CV_StsUnsupportedFormat, "cv::sortRowsByIndices only works on integer indices!");
+        CV_Error(Error::StsUnsupportedFormat, "cv::sortRowsByIndices only works on integer indices!");
     Mat src = _src.getMat();
     std::vector<int> indices = _indices.getMat();
     _dst.create(src.rows, src.cols, src.type());
@@ -64,8 +64,8 @@ static Mat argsort(InputArray _src, bool ascending=true)
 {
     Mat src = _src.getMat();
     if (src.rows != 1 && src.cols != 1)
-        CV_Error(CV_StsBadArg, "cv::argsort only sorts 1D matrices.");
-    int flags = CV_SORT_EVERY_ROW+(ascending ? CV_SORT_ASCENDING : CV_SORT_DESCENDING);
+        CV_Error(Error::StsBadArg, "cv::argsort only sorts 1D matrices.");
+    int flags = SORT_EVERY_ROW | (ascending ? SORT_ASCENDING : SORT_DESCENDING);
     Mat sorted_indices;
     sortIdx(src.reshape(1,1),sorted_indices,flags);
     return sorted_indices;
@@ -116,8 +116,8 @@ static Mat interp1(InputArray _x, InputArray _Y, InputArray _xi)
     Mat Y = _Y.getMat();
     Mat xi = _xi.getMat();
     // check types & alignment
-    assert((x.type() == Y.type()) && (Y.type() == xi.type()));
-    assert((x.cols == 1) && (x.rows == Y.rows) && (x.cols == Y.cols));
+    CV_Assert((x.type() == Y.type()) && (Y.type() == xi.type()));
+    CV_Assert((x.cols == 1) && (x.rows == Y.rows) && (x.cols == Y.cols));
     // call templated interp1
     switch(x.type()) {
         case CV_8SC1: return interp1_<char>(x,Y,xi); break;
@@ -127,7 +127,7 @@ static Mat interp1(InputArray _x, InputArray _Y, InputArray _xi)
         case CV_32SC1: return interp1_<int>(x,Y,xi); break;
         case CV_32FC1: return interp1_<float>(x,Y,xi); break;
         case CV_64FC1: return interp1_<double>(x,Y,xi); break;
-        default: CV_Error(CV_StsUnsupportedFormat, ""); break;
+        default: CV_Error(Error::StsUnsupportedFormat, ""); break;
     }
     return Mat();
 }
@@ -473,7 +473,7 @@ namespace colormap
     void ColorMap::operator()(InputArray _src, OutputArray _dst) const
     {
         if(_lut.total() != 256)
-            CV_Error(CV_StsAssert, "cv::LUT only supports tables of size 256.");
+            CV_Error(Error::StsAssert, "cv::LUT only supports tables of size 256.");
         Mat src = _src.getMat();
         // Return original matrix if wrong type is given (is fail loud better here?)
         if(src.type() != CV_8UC1 && src.type() != CV_8UC3)
@@ -483,8 +483,8 @@ namespace colormap
         }
         // Turn into a BGR matrix into its grayscale representation.
         if(src.type() == CV_8UC3)
-            cvtColor(src.clone(), src, CV_BGR2GRAY);
-        cvtColor(src.clone(), src, CV_GRAY2BGR);
+            cvtColor(src.clone(), src, COLOR_BGR2GRAY);
+        cvtColor(src.clone(), src, COLOR_GRAY2BGR);
         // Apply the ColorMap.
         LUT(src, _lut, _dst);
     }
@@ -521,7 +521,7 @@ namespace colormap
             colormap == COLORMAP_WINTER ? (colormap::ColorMap*)(new colormap::Winter) : 0;
 
         if( !cm )
-            CV_Error( CV_StsBadArg, "Unknown colormap id; use one of COLORMAP_*");
+            CV_Error( Error::StsBadArg, "Unknown colormap id; use one of COLORMAP_*");
 
         (*cm)(src, dst);
 
diff --git a/modules/contrib/src/colortracker.cpp b/modules/contrib/src/colortracker.cpp
index 5a9d9b7c43..a3eeb052b2 100644
--- a/modules/contrib/src/colortracker.cpp
+++ b/modules/contrib/src/colortracker.cpp
@@ -60,7 +60,7 @@ void CvMeanShiftTracker::newTrackingWindow(Mat image, Rect selection)
     float srange[] = { 0, 1 };
     const float* ranges[] = {hrange, srange};
 
-    cvtColor(image, hsv, CV_BGR2HSV);
+    cvtColor(image, hsv, COLOR_BGR2HSV);
     inRange(hsv, Scalar(0, 30, MIN(10, 256)), Scalar(180, 256, MAX(10, 256)), mask);
 
     hue.create(hsv.size(), CV_8UC2);
@@ -83,7 +83,7 @@ RotatedRect CvMeanShiftTracker::updateTrackingWindow(Mat image)
     float srange[] = { 0, 1 };
     const float* ranges[] = {hrange, srange};
 
-    cvtColor(image, hsv, CV_BGR2HSV);
+    cvtColor(image, hsv, COLOR_BGR2HSV);
     inRange(hsv, Scalar(0, 30, MIN(10, 256)), Scalar(180, 256, MAX(10, 256)), mask);
     hue.create(hsv.size(), CV_8UC2);
     mixChannels(&hsv, 1, &hue, 1, channels, 2);
@@ -132,5 +132,3 @@ Point2f CvMeanShiftTracker::getTrackingCenter()
 {
     return prev_center;
 }
-
-
diff --git a/modules/contrib/src/contrib_init.cpp b/modules/contrib/src/contrib_init.cpp
index a80f6f5e11..317867a59f 100644
--- a/modules/contrib/src/contrib_init.cpp
+++ b/modules/contrib/src/contrib_init.cpp
@@ -41,4 +41,3 @@
 //M*/
 
 #include "precomp.hpp"
-
diff --git a/modules/contrib/src/detection_based_tracker.cpp b/modules/contrib/src/detection_based_tracker.cpp
index 117a6e916a..27807290af 100644
--- a/modules/contrib/src/detection_based_tracker.cpp
+++ b/modules/contrib/src/detection_based_tracker.cpp
@@ -1,5 +1,6 @@
 #if defined(__linux__) || defined(LINUX) || defined(__APPLE__) || defined(ANDROID)
 #include "opencv2/contrib/detection_based_tracker.hpp"
+#include "opencv2/core/utility.hpp"
 
 #include <pthread.h>
 
@@ -127,7 +128,7 @@ cv::DetectionBasedTracker::SeparateDetectionWork::SeparateDetectionWork(Detectio
     stateThread(STATE_THREAD_STOPPED),
     timeWhenDetectingThreadStartedWork(-1)
 {
-    CV_Assert(!_detector.empty());
+    CV_Assert(_detector);
 
     cascadeInThread = _detector;
 
@@ -461,11 +462,11 @@ cv::DetectionBasedTracker::DetectionBasedTracker(cv::Ptr<IDetector> mainDetector
     cascadeForTracking(trackingDetector)
 {
     CV_Assert( (params.maxTrackLifetime >= 0)
-//            && (!mainDetector.empty())
-            && (!trackingDetector.empty()) );
+//            && mainDetector
+            && trackingDetector );
 
-    if (!mainDetector.empty()) {
-        separateDetectionWork = new SeparateDetectionWork(*this, mainDetector);
+    if (mainDetector) {
+        separateDetectionWork.reset(new SeparateDetectionWork(*this, mainDetector));
     }
 
     weightsPositionsSmoothing.push_back(1);
@@ -482,7 +483,7 @@ void DetectionBasedTracker::process(const Mat& imageGray)
 {
     CV_Assert(imageGray.type()==CV_8UC1);
 
-    if ( (!separateDetectionWork.empty()) && (!separateDetectionWork->isWorking()) ) {
+    if ( separateDetectionWork && !separateDetectionWork->isWorking() ) {
         separateDetectionWork->run();
     }
 
@@ -500,7 +501,7 @@ void DetectionBasedTracker::process(const Mat& imageGray)
 
     std::vector<Rect> rectsWhereRegions;
     bool shouldHandleResult=false;
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         shouldHandleResult = separateDetectionWork->communicateWithDetectingThread(imageGray, rectsWhereRegions);
     }
 
@@ -588,7 +589,7 @@ void cv::DetectionBasedTracker::getObjects(std::vector<ExtObject>& result) const
 
 bool cv::DetectionBasedTracker::run()
 {
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         return separateDetectionWork->run();
     }
     return false;
@@ -596,14 +597,14 @@ bool cv::DetectionBasedTracker::run()
 
 void cv::DetectionBasedTracker::stop()
 {
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         separateDetectionWork->stop();
     }
 }
 
 void cv::DetectionBasedTracker::resetTracking()
 {
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         separateDetectionWork->resetTracking();
     }
     trackedObjects.clear();
@@ -875,11 +876,11 @@ bool cv::DetectionBasedTracker::setParameters(const Parameters& params)
         return false;
     }
 
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         separateDetectionWork->lock();
     }
     parameters=params;
-    if (!separateDetectionWork.empty()) {
+    if (separateDetectionWork) {
         separateDetectionWork->unlock();
     }
     return true;
@@ -891,4 +892,3 @@ const cv::DetectionBasedTracker::Parameters& DetectionBasedTracker::getParameter
 }
 
 #endif
-
diff --git a/modules/contrib/src/facerec.cpp b/modules/contrib/src/facerec.cpp
index 9ce082819e..1bea74e89e 100644
--- a/modules/contrib/src/facerec.cpp
+++ b/modules/contrib/src/facerec.cpp
@@ -17,6 +17,7 @@
  */
 #include "precomp.hpp"
 #include <set>
+#include <limits>
 
 namespace cv
 {
@@ -35,7 +36,7 @@ inline void readFileNodeList(const FileNode& fn, std::vector<_Tp>& result) {
 
 // Writes the a list of given items to a cv::FileStorage.
 template<typename _Tp>
-inline void writeFileNodeList(FileStorage& fs, const std::string& name,
+inline void writeFileNodeList(FileStorage& fs, const String& name,
                               const std::vector<_Tp>& items) {
     // typedefs
     typedef typename std::vector<_Tp>::const_iterator constVecIterator;
@@ -50,8 +51,8 @@ inline void writeFileNodeList(FileStorage& fs, const std::string& name,
 static Mat asRowMatrix(InputArrayOfArrays src, int rtype, double alpha=1, double beta=0) {
     // make sure the input data is a vector of matrices or vector of vector
     if(src.kind() != _InputArray::STD_VECTOR_MAT && src.kind() != _InputArray::STD_VECTOR_VECTOR) {
-        std::string error_message = "The data is expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "The data is expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
+        CV_Error(Error::StsBadArg, error_message);
     }
     // number of samples
     size_t n = src.total();
@@ -66,8 +67,8 @@ static Mat asRowMatrix(InputArrayOfArrays src, int rtype, double alpha=1, double
     for(unsigned int i = 0; i < n; i++) {
         // make sure data can be reshaped, throw exception if not!
         if(src.getMat(i).total() != d) {
-            std::string error_message = format("Wrong number of elements in matrix #%d! Expected %d was %d.", i, d, src.getMat(i).total());
-            CV_Error(CV_StsBadArg, error_message);
+            String error_message = format("Wrong number of elements in matrix #%d! Expected %d was %d.", i, d, src.getMat(i).total());
+            CV_Error(Error::StsBadArg, error_message);
         }
         // get a hold of the current row
         Mat xi = data.row(i);
@@ -305,22 +306,22 @@ void FaceRecognizer::update(InputArrayOfArrays src, InputArray labels ) {
         return;
     }
 
-    std::string error_msg = format("This FaceRecognizer (%s) does not support updating, you have to use FaceRecognizer::train to update it.", this->name().c_str());
-    CV_Error(CV_StsNotImplemented, error_msg);
+    String error_msg = format("This FaceRecognizer (%s) does not support updating, you have to use FaceRecognizer::train to update it.", this->name().c_str());
+    CV_Error(Error::StsNotImplemented, error_msg);
 }
 
-void FaceRecognizer::save(const std::string& filename) const {
+void FaceRecognizer::save(const String& filename) const {
     FileStorage fs(filename, FileStorage::WRITE);
     if (!fs.isOpened())
-        CV_Error(CV_StsError, "File can't be opened for writing!");
+        CV_Error(Error::StsError, "File can't be opened for writing!");
     this->save(fs);
     fs.release();
 }
 
-void FaceRecognizer::load(const std::string& filename) {
+void FaceRecognizer::load(const String& filename) {
     FileStorage fs(filename, FileStorage::READ);
     if (!fs.isOpened())
-        CV_Error(CV_StsError, "File can't be opened for writing!");
+        CV_Error(Error::StsError, "File can't be opened for writing!");
     this->load(fs);
     fs.release();
 }
@@ -330,18 +331,18 @@ void FaceRecognizer::load(const std::string& filename) {
 //------------------------------------------------------------------------------
 void Eigenfaces::train(InputArrayOfArrays _src, InputArray _local_labels) {
     if(_src.total() == 0) {
-        std::string error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
+        CV_Error(Error::StsBadArg, error_message);
     } else if(_local_labels.getMat().type() != CV_32SC1) {
-        std::string error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _local_labels.type());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _local_labels.type());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // make sure data has correct size
     if(_src.total() > 1) {
         for(int i = 1; i < static_cast<int>(_src.total()); i++) {
             if(_src.getMat(i-1).total() != _src.getMat(i).total()) {
-                std::string error_message = format("In the Eigenfaces method all input samples (training images) must be of equal size! Expected %d pixels, but was %d pixels.", _src.getMat(i-1).total(), _src.getMat(i).total());
-                CV_Error(CV_StsUnsupportedFormat, error_message);
+                String error_message = format("In the Eigenfaces method all input samples (training images) must be of equal size! Expected %d pixels, but was %d pixels.", _src.getMat(i-1).total(), _src.getMat(i).total());
+                CV_Error(Error::StsUnsupportedFormat, error_message);
             }
         }
     }
@@ -354,8 +355,8 @@ void Eigenfaces::train(InputArrayOfArrays _src, InputArray _local_labels) {
    int n = data.rows;
     // assert there are as much samples as labels
     if(static_cast<int>(labels.total()) != n) {
-        std::string error_message = format("The number of samples (src) must equal the number of labels (labels)! len(src)=%d, len(labels)=%d.", n, labels.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("The number of samples (src) must equal the number of labels (labels)! len(src)=%d, len(labels)=%d.", n, labels.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // clear existing model data
     _labels.release();
@@ -365,7 +366,7 @@ void Eigenfaces::train(InputArrayOfArrays _src, InputArray _local_labels) {
         _num_components = n;
 
     // perform the PCA
-    PCA pca(data, Mat(), CV_PCA_DATA_AS_ROW, _num_components);
+    PCA pca(data, Mat(), PCA::DATA_AS_ROW, _num_components);
     // copy the PCA results
     _mean = pca.mean.reshape(1,1); // store the mean vector
     _eigenvalues = pca.eigenvalues.clone(); // eigenvalues by row
@@ -385,12 +386,12 @@ void Eigenfaces::predict(InputArray _src, int &minClass, double &minDist) const
     // make sure the user is passing correct data
     if(_projections.empty()) {
         // throw error if no data (or simply return -1?)
-        std::string error_message = "This Eigenfaces model is not computed yet. Did you call Eigenfaces::train?";
-        CV_Error(CV_StsError, error_message);
+        String error_message = "This Eigenfaces model is not computed yet. Did you call Eigenfaces::train?";
+        CV_Error(Error::StsError, error_message);
     } else if(_eigenvectors.rows != static_cast<int>(src.total())) {
         // check data alignment just for clearer exception messages
-        std::string error_message = format("Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with %d elements, but got %d.", _eigenvectors.rows, src.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with %d elements, but got %d.", _eigenvectors.rows, src.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // project into PCA subspace
     Mat q = subspaceProject(_eigenvectors, _mean, src.reshape(1,1));
@@ -439,18 +440,18 @@ void Eigenfaces::save(FileStorage& fs) const {
 //------------------------------------------------------------------------------
 void Fisherfaces::train(InputArrayOfArrays src, InputArray _lbls) {
     if(src.total() == 0) {
-        std::string error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
+        CV_Error(Error::StsBadArg, error_message);
     } else if(_lbls.getMat().type() != CV_32SC1) {
-        std::string error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _lbls.type());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _lbls.type());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // make sure data has correct size
     if(src.total() > 1) {
         for(int i = 1; i < static_cast<int>(src.total()); i++) {
             if(src.getMat(i-1).total() != src.getMat(i).total()) {
-                std::string error_message = format("In the Fisherfaces method all input samples (training images) must be of equal size! Expected %d pixels, but was %d pixels.", src.getMat(i-1).total(), src.getMat(i).total());
-                CV_Error(CV_StsUnsupportedFormat, error_message);
+                String error_message = format("In the Fisherfaces method all input samples (training images) must be of equal size! Expected %d pixels, but was %d pixels.", src.getMat(i-1).total(), src.getMat(i).total());
+                CV_Error(Error::StsUnsupportedFormat, error_message);
             }
         }
     }
@@ -461,11 +462,11 @@ void Fisherfaces::train(InputArrayOfArrays src, InputArray _lbls) {
     int N = data.rows;
     // make sure labels are passed in correct shape
     if(labels.total() != (size_t) N) {
-        std::string error_message = format("The number of samples (src) must equal the number of labels (labels)! len(src)=%d, len(labels)=%d.", N, labels.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("The number of samples (src) must equal the number of labels (labels)! len(src)=%d, len(labels)=%d.", N, labels.total());
+        CV_Error(Error::StsBadArg, error_message);
     } else if(labels.rows != 1 && labels.cols != 1) {
-        std::string error_message = format("Expected the labels in a matrix with one row or column! Given dimensions are rows=%s, cols=%d.", labels.rows, labels.cols);
-       CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Expected the labels in a matrix with one row or column! Given dimensions are rows=%s, cols=%d.", labels.rows, labels.cols);
+       CV_Error(Error::StsBadArg, error_message);
     }
     // clear existing model data
     _labels.release();
@@ -481,7 +482,7 @@ void Fisherfaces::train(InputArrayOfArrays src, InputArray _lbls) {
     if((_num_components <= 0) || (_num_components > (C-1)))
         _num_components = (C-1);
     // perform a PCA and keep (N-C) components
-    PCA pca(data, Mat(), CV_PCA_DATA_AS_ROW, (N-C));
+    PCA pca(data, Mat(), PCA::DATA_AS_ROW, (N-C));
     // project the data and perform a LDA on it
     LDA lda(pca.project(data),labels, _num_components);
     // store the total mean vector
@@ -505,11 +506,11 @@ void Fisherfaces::predict(InputArray _src, int &minClass, double &minDist) const
     // check data alignment just for clearer exception messages
     if(_projections.empty()) {
         // throw error if no data (or simply return -1?)
-        std::string error_message = "This Fisherfaces model is not computed yet. Did you call Fisherfaces::train?";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "This Fisherfaces model is not computed yet. Did you call Fisherfaces::train?";
+        CV_Error(Error::StsBadArg, error_message);
     } else if(src.total() != (size_t) _eigenvectors.rows) {
-        std::string error_message = format("Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with %d elements, but got %d.", _eigenvectors.rows, src.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with %d elements, but got %d.", _eigenvectors.rows, src.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // project into LDA subspace
     Mat q = subspaceProject(_eigenvectors, _mean, src.reshape(1,1));
@@ -640,8 +641,8 @@ static void elbp(InputArray src, OutputArray dst, int radius, int neighbors)
     case CV_32FC1:  elbp_<float>(src,dst, radius, neighbors); break;
     case CV_64FC1:  elbp_<double>(src,dst, radius, neighbors); break;
     default:
-        std::string error_msg = format("Using Original Local Binary Patterns for feature extraction only works on single-channel images (given %d). Please pass the image data as a grayscale image!", type);
-        CV_Error(CV_StsNotImplemented, error_msg);
+        String error_msg = format("Using Original Local Binary Patterns for feature extraction only works on single-channel images (given %d). Please pass the image data as a grayscale image!", type);
+        CV_Error(Error::StsNotImplemented, error_msg);
         break;
     }
 }
@@ -687,7 +688,7 @@ static Mat histc(InputArray _src, int minVal, int maxVal, bool normed)
             return histc_(src, minVal, maxVal, normed);
             break;
         default:
-            CV_Error(CV_StsUnmatchedFormats, "This type is not implemented yet."); break;
+            CV_Error(Error::StsUnmatchedFormats, "This type is not implemented yet."); break;
     }
     return Mat();
 }
@@ -768,15 +769,15 @@ void LBPH::update(InputArrayOfArrays _in_src, InputArray _in_labels) {
 
 void LBPH::train(InputArrayOfArrays _in_src, InputArray _in_labels, bool preserveData) {
     if(_in_src.kind() != _InputArray::STD_VECTOR_MAT && _in_src.kind() != _InputArray::STD_VECTOR_VECTOR) {
-        std::string error_message = "The images are expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "The images are expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
+        CV_Error(Error::StsBadArg, error_message);
     }
     if(_in_src.total() == 0) {
-        std::string error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
-        CV_Error(CV_StsUnsupportedFormat, error_message);
+        String error_message = format("Empty training data was given. You'll need more than one sample to learn a model.");
+        CV_Error(Error::StsUnsupportedFormat, error_message);
     } else if(_in_labels.getMat().type() != CV_32SC1) {
-        std::string error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _in_labels.type());
-        CV_Error(CV_StsUnsupportedFormat, error_message);
+        String error_message = format("Labels must be given as integer (CV_32SC1). Expected %d, but was %d.", CV_32SC1, _in_labels.type());
+        CV_Error(Error::StsUnsupportedFormat, error_message);
     }
     // get the vector of matrices
     std::vector<Mat> src;
@@ -785,8 +786,8 @@ void LBPH::train(InputArrayOfArrays _in_src, InputArray _in_labels, bool preserv
     Mat labels = _in_labels.getMat();
     // check if data is well- aligned
     if(labels.total() != src.size()) {
-        std::string error_message = format("The number of samples (src) must equal the number of labels (labels). Was len(samples)=%d, len(labels)=%d.", src.size(), _labels.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("The number of samples (src) must equal the number of labels (labels). Was len(samples)=%d, len(labels)=%d.", src.size(), _labels.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // if this model should be trained without preserving old data, delete old model data
     if(!preserveData) {
@@ -816,8 +817,8 @@ void LBPH::train(InputArrayOfArrays _in_src, InputArray _in_labels, bool preserv
 void LBPH::predict(InputArray _src, int &minClass, double &minDist) const {
     if(_histograms.empty()) {
         // throw error if no data (or simply return -1?)
-        std::string error_message = "This LBPH model is not computed yet. Did you call the train method?";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "This LBPH model is not computed yet. Did you call the train method?";
+        CV_Error(Error::StsBadArg, error_message);
     }
     Mat src = _src.getMat();
     // get the spatial histogram from input image
@@ -832,7 +833,7 @@ void LBPH::predict(InputArray _src, int &minClass, double &minDist) const {
     minDist = DBL_MAX;
     minClass = -1;
     for(size_t sampleIdx = 0; sampleIdx < _histograms.size(); sampleIdx++) {
-        double dist = compareHist(_histograms[sampleIdx], query, CV_COMP_CHISQR);
+        double dist = compareHist(_histograms[sampleIdx], query, HISTCMP_CHISQR);
         if((dist < minDist) && (dist < _threshold)) {
             minDist = dist;
             minClass = _labels.at<int>((int) sampleIdx);
@@ -850,18 +851,18 @@ int LBPH::predict(InputArray _src) const {
 
 Ptr<FaceRecognizer> createEigenFaceRecognizer(int num_components, double threshold)
 {
-    return new Eigenfaces(num_components, threshold);
+    return makePtr<Eigenfaces>(num_components, threshold);
 }
 
 Ptr<FaceRecognizer> createFisherFaceRecognizer(int num_components, double threshold)
 {
-    return new Fisherfaces(num_components, threshold);
+    return makePtr<Fisherfaces>(num_components, threshold);
 }
 
 Ptr<FaceRecognizer> createLBPHFaceRecognizer(int radius, int neighbors,
                                              int grid_x, int grid_y, double threshold)
 {
-    return new LBPH(radius, neighbors, grid_x, grid_y, threshold);
+    return makePtr<LBPH>(radius, neighbors, grid_x, grid_y, threshold);
 }
 
 CV_INIT_ALGORITHM(Eigenfaces, "FaceRecognizer.Eigenfaces",
@@ -893,7 +894,7 @@ CV_INIT_ALGORITHM(LBPH, "FaceRecognizer.LBPH",
 
 bool initModule_contrib()
 {
-    Ptr<Algorithm> efaces = createEigenfaces(), ffaces = createFisherfaces(), lbph = createLBPH();
+    Ptr<Algorithm> efaces = createEigenfaces_ptr_hidden(), ffaces = createFisherfaces_ptr_hidden(), lbph = createLBPH_ptr_hidden();
     return efaces->info() != 0 && ffaces->info() != 0 && lbph->info() != 0;
 }
 
diff --git a/modules/contrib/src/featuretracker.cpp b/modules/contrib/src/featuretracker.cpp
index 575f69abf9..795c1a08fa 100644
--- a/modules/contrib/src/featuretracker.cpp
+++ b/modules/contrib/src/featuretracker.cpp
@@ -54,23 +54,26 @@ CvFeatureTracker::CvFeatureTracker(CvFeatureTrackerParams _params) :
     {
     case CvFeatureTrackerParams::SIFT:
         dd = Algorithm::create<Feature2D>("Feature2D.SIFT");
-        if( dd.empty() )
+        if( !dd )
             CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without SIFT support");
         dd->set("nOctaveLayers", 5);
         dd->set("contrastThreshold", 0.04);
         dd->set("edgeThreshold", 10.7);
+        break;
     case CvFeatureTrackerParams::SURF:
         dd = Algorithm::create<Feature2D>("Feature2D.SURF");
-        if( dd.empty() )
+        if( !dd )
             CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without SURF support");
         dd->set("hessianThreshold", 400);
         dd->set("nOctaves", 3);
         dd->set("nOctaveLayers", 4);
+        break;
     default:
         CV_Error(CV_StsBadArg, "Unknown feature type");
+        break;
     }
 
-    matcher = new BFMatcher(NORM_L2);
+    matcher = makePtr<BFMatcher>(int(NORM_L2));
 }
 
 CvFeatureTracker::~CvFeatureTracker()
@@ -80,7 +83,7 @@ CvFeatureTracker::~CvFeatureTracker()
 void CvFeatureTracker::newTrackingWindow(Mat image, Rect selection)
 {
     image.copyTo(prev_image);
-    cvtColor(prev_image, prev_image_bw, CV_BGR2GRAY);
+    cvtColor(prev_image, prev_image_bw, COLOR_BGR2GRAY);
     prev_trackwindow = selection;
     prev_center.x = selection.x;
     prev_center.y = selection.y;
@@ -131,7 +134,7 @@ Rect CvFeatureTracker::updateTrackingWindowWithSIFT(Mat image)
             curr_keys.push_back(curr_keypoints[matches[i].trainIdx].pt);
         }
 
-        Mat T = findHomography(prev_keys, curr_keys, CV_LMEDS);
+        Mat T = findHomography(prev_keys, curr_keys, LMEDS);
 
         prev_trackwindow.x += cvRound(T.at<double> (0, 2));
         prev_trackwindow.y += cvRound(T.at<double> (1, 2));
@@ -148,12 +151,12 @@ Rect CvFeatureTracker::updateTrackingWindowWithFlow(Mat image)
     ittr++;
     Size subPixWinSize(10,10), winSize(31,31);
     Mat image_bw;
-    TermCriteria termcrit(CV_TERMCRIT_ITER | CV_TERMCRIT_EPS, 20, 0.03);
+    TermCriteria termcrit(TermCriteria::COUNT | TermCriteria::EPS, 20, 0.03);
     std::vector<uchar> status;
     std::vector<float> err;
 
-    cvtColor(image, image_bw, CV_BGR2GRAY);
-    cvtColor(prev_image, prev_image_bw, CV_BGR2GRAY);
+    cvtColor(image, image_bw, COLOR_BGR2GRAY);
+    cvtColor(prev_image, prev_image_bw, COLOR_BGR2GRAY);
 
     if (ittr == 1)
     {
@@ -218,4 +221,3 @@ Point2f CvFeatureTracker::getTrackingCenter()
     center.y = (float)(prev_center.y + prev_trackwindow.height/2.0);
     return center;
 }
-
diff --git a/modules/contrib/src/fuzzymeanshifttracker.cpp b/modules/contrib/src/fuzzymeanshifttracker.cpp
index 443b961ed1..5e5ebc7e99 100644
--- a/modules/contrib/src/fuzzymeanshifttracker.cpp
+++ b/modules/contrib/src/fuzzymeanshifttracker.cpp
@@ -35,6 +35,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/contrib/compat.hpp"
 
 CvFuzzyPoint::CvFuzzyPoint(double _x, double _y)
 {
@@ -380,6 +381,7 @@ void CvFuzzyMeanShiftTracker::SearchWindow::initDepthValues(IplImage *maskImage,
                 {
                     if (*depthData)
                     {
+                        d = *depthData;
                         m1 += d;
                         if (d < mind)
                             mind = d;
@@ -719,4 +721,3 @@ void CvFuzzyMeanShiftTracker::track(IplImage *maskImage, IplImage *depthMap, int
                 searchMode = tsTracking;
     }
 };
-
diff --git a/modules/contrib/src/gencolors.cpp b/modules/contrib/src/gencolors.cpp
index 504015308c..24796ec03d 100644
--- a/modules/contrib/src/gencolors.cpp
+++ b/modules/contrib/src/gencolors.cpp
@@ -85,7 +85,7 @@ static void downsamplePoints( const Mat& src, Mat& dst, size_t count )
         candidatePointsMask.at<uchar>(0, maxLoc.x) = 0;
 
         Mat minDists;
-        reduce( activedDists, minDists, 0, CV_REDUCE_MIN );
+        reduce( activedDists, minDists, 0, REDUCE_MIN );
         minMaxLoc( minDists, 0, &maxVal, 0, &maxLoc, candidatePointsMask );
         dst.at<Point3_<uchar> >((int)i) = src.at<Point3_<uchar> >(maxLoc.x);
     }
@@ -118,7 +118,7 @@ void cv::generateColors( std::vector<Scalar>& colors, size_t count, size_t facto
     // Convert the colors set to Lab space.
     // Distances between colors in this space correspond a human perception.
     Mat lab;
-    cvtColor( bgr, lab, CV_BGR2Lab);
+    cvtColor( bgr, lab, COLOR_BGR2Lab);
 
     // Subsample colors from the generated set so that
     // to maximize the minimum distances between each other.
@@ -128,7 +128,7 @@ void cv::generateColors( std::vector<Scalar>& colors, size_t count, size_t facto
 
     // Convert subsampled colors back to RGB
     Mat bgr_subset;
-    cvtColor( lab_subset, bgr_subset, CV_Lab2BGR );
+    cvtColor( lab_subset, bgr_subset, COLOR_Lab2BGR );
 
     CV_Assert( bgr_subset.total() == count );
     for( size_t i = 0; i < count; i++ )
diff --git a/modules/contrib/src/hybridtracker.cpp b/modules/contrib/src/hybridtracker.cpp
index 23a6ecbbad..d93f76d0d3 100644
--- a/modules/contrib/src/hybridtracker.cpp
+++ b/modules/contrib/src/hybridtracker.cpp
@@ -233,4 +233,3 @@ void CvHybridTracker::updateTrackerWithLowPassFilter(Mat) {
 Rect CvHybridTracker::getTrackingWindow() {
     return prev_window;
 }
-
diff --git a/modules/contrib/src/inputoutput.cpp b/modules/contrib/src/inputoutput.cpp
index ee97e1287b..7f583558e8 100644
--- a/modules/contrib/src/inputoutput.cpp
+++ b/modules/contrib/src/inputoutput.cpp
@@ -1,7 +1,7 @@
-
 #include "opencv2/contrib.hpp"
+#include "cvconfig.h"
 
-#ifdef WIN32
+#if defined(WIN32) || defined(_WIN32)
     #include <windows.h>
     #include <tchar.h>
 #else
@@ -10,16 +10,27 @@
 
 namespace cv
 {
-    std::vector<std::string> Directory::GetListFiles(  const std::string& path, const std::string & exten, bool addPath )
+    std::vector<String> Directory::GetListFiles( const String& path, const String & exten, bool addPath )
     {
-        std::vector<std::string> list;
+        std::vector<String> list;
         list.clear();
-        std::string path_f = path + "/" + exten;
+        String path_f = path + "/" + exten;
         #ifdef WIN32
-            WIN32_FIND_DATA FindFileData;
-            HANDLE hFind;
+        #ifdef HAVE_WINRT
+            WIN32_FIND_DATAW FindFileData;
+        #else
+            WIN32_FIND_DATAA FindFileData;
+        #endif
+        HANDLE hFind;
 
-            hFind = FindFirstFile((LPCSTR)path_f.c_str(), &FindFileData);
+        #ifdef HAVE_WINRT
+            wchar_t wpath[MAX_PATH];
+            size_t copied = mbstowcs(wpath, path_f.c_str(), MAX_PATH);
+            CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+            hFind = FindFirstFileExW(wpath, FindExInfoStandard, &FindFileData, FindExSearchNameMatch, NULL, 0);
+        #else
+            hFind = FindFirstFileA((LPCSTR)path_f.c_str(), &FindFileData);
+        #endif
             if (hFind == INVALID_HANDLE_VALUE)
             {
                 return list;
@@ -34,13 +45,26 @@ namespace cv
                         FindFileData.dwFileAttributes == FILE_ATTRIBUTE_SYSTEM  ||
                         FindFileData.dwFileAttributes == FILE_ATTRIBUTE_READONLY)
                     {
+                        char* fname;
+                    #ifdef HAVE_WINRT
+                        char fname_tmp[MAX_PATH] = {0};
+                        size_t copied = wcstombs(fname_tmp, FindFileData.cFileName, MAX_PATH);
+                        CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+                        fname = fname_tmp;
+                    #else
+                        fname = FindFileData.cFileName;
+                    #endif
                         if (addPath)
-                            list.push_back(path + "/" + FindFileData.cFileName);
+                            list.push_back(path + "/" + String(fname));
                         else
-                            list.push_back(FindFileData.cFileName);
+                            list.push_back(String(fname));
                     }
                 }
-                while(FindNextFile(hFind, &FindFileData));
+            #ifdef HAVE_WINRT
+                while(FindNextFileW(hFind, &FindFileData));
+            #else
+                while(FindNextFileA(hFind, &FindFileData));
+            #endif
                 FindClose(hFind);
             }
         #else
@@ -57,10 +81,10 @@ namespace cv
                 if (dirp->d_type == DT_REG)
                 {
                     if (exten.compare("*") == 0)
-                        list.push_back(static_cast<std::string>(dirp->d_name));
+                        list.push_back(static_cast<String>(dirp->d_name));
                     else
-                        if (std::string(dirp->d_name).find(exten) != std::string::npos)
-                            list.push_back(static_cast<std::string>(dirp->d_name));
+                        if (String(dirp->d_name).find(exten) != String::npos)
+                            list.push_back(static_cast<String>(dirp->d_name));
                 }
             }
             closedir(dp);
@@ -69,16 +93,28 @@ namespace cv
         return list;
     }
 
-    std::vector<std::string> Directory::GetListFolders( const std::string& path, const std::string & exten, bool addPath )
+    std::vector<String> Directory::GetListFolders( const String& path, const String & exten, bool addPath )
     {
-        std::vector<std::string> list;
-        std::string path_f = path + "/" + exten;
+        std::vector<String> list;
+        String path_f = path + "/" + exten;
         list.clear();
         #ifdef WIN32
-            WIN32_FIND_DATA FindFileData;
+        #ifdef HAVE_WINRT
+            WIN32_FIND_DATAW FindFileData;
+        #else
+            WIN32_FIND_DATAA FindFileData;
+        #endif
             HANDLE hFind;
 
-            hFind = FindFirstFile((LPCSTR)path_f.c_str(), &FindFileData);
+        #ifdef HAVE_WINRT
+            wchar_t wpath [MAX_PATH];
+            size_t copied = mbstowcs(wpath, path_f.c_str(), path_f.size());
+            CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+
+            hFind = FindFirstFileExW(wpath, FindExInfoStandard, &FindFileData, FindExSearchNameMatch, NULL, 0);
+        #else
+            hFind = FindFirstFileA((LPCSTR)path_f.c_str(), &FindFileData);
+        #endif
             if (hFind == INVALID_HANDLE_VALUE)
             {
                 return list;
@@ -87,17 +123,37 @@ namespace cv
             {
                 do
                 {
+#ifdef HAVE_WINRT
+                    if (FindFileData.dwFileAttributes == FILE_ATTRIBUTE_DIRECTORY &&
+                        wcscmp(FindFileData.cFileName, L".") != 0 &&
+                        wcscmp(FindFileData.cFileName, L"..") != 0)
+#else
                     if (FindFileData.dwFileAttributes == FILE_ATTRIBUTE_DIRECTORY &&
                         strcmp(FindFileData.cFileName, ".") != 0 &&
                         strcmp(FindFileData.cFileName, "..") != 0)
+#endif
                     {
+                        char* fname;
+                    #ifdef HAVE_WINRT
+                        char fname_tmp[MAX_PATH];
+                        size_t copied = wcstombs(fname_tmp, FindFileData.cFileName, MAX_PATH);
+                        CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+                        fname = fname_tmp;
+                    #else
+                        fname = FindFileData.cFileName;
+                    #endif
+
                         if (addPath)
-                            list.push_back(path + "/" + FindFileData.cFileName);
+                            list.push_back(path + "/" + String(fname));
                         else
-                            list.push_back(FindFileData.cFileName);
+                            list.push_back(String(fname));
                     }
                 }
-                while(FindNextFile(hFind, &FindFileData));
+            #ifdef HAVE_WINRT
+                while(FindNextFileW(hFind, &FindFileData));
+            #else
+                while(FindNextFileA(hFind, &FindFileData));
+            #endif
                 FindClose(hFind);
             }
 
@@ -117,10 +173,10 @@ namespace cv
                     strcmp(dirp->d_name, "..") != 0 )
                 {
                     if (exten.compare("*") == 0)
-                        list.push_back(static_cast<std::string>(dirp->d_name));
+                        list.push_back(static_cast<String>(dirp->d_name));
                     else
-                        if (std::string(dirp->d_name).find(exten) != std::string::npos)
-                            list.push_back(static_cast<std::string>(dirp->d_name));
+                        if (String(dirp->d_name).find(exten) != String::npos)
+                            list.push_back(static_cast<String>(dirp->d_name));
                 }
             }
             closedir(dp);
@@ -129,16 +185,16 @@ namespace cv
         return list;
     }
 
-    std::vector<std::string> Directory::GetListFilesR ( const std::string& path, const std::string & exten, bool addPath )
+    std::vector<String> Directory::GetListFilesR ( const String& path, const String & exten, bool addPath )
     {
-        std::vector<std::string> list = Directory::GetListFiles(path, exten, addPath);
+        std::vector<String> list = Directory::GetListFiles(path, exten, addPath);
 
-        std::vector<std::string> dirs = Directory::GetListFolders(path, exten, addPath);
+        std::vector<String> dirs = Directory::GetListFolders(path, exten, addPath);
 
-        std::vector<std::string>::const_iterator it;
+        std::vector<String>::const_iterator it;
         for (it = dirs.begin(); it != dirs.end(); ++it)
         {
-            std::vector<std::string> cl = Directory::GetListFiles(*it, exten, addPath);
+            std::vector<String> cl = Directory::GetListFiles(*it, exten, addPath);
             list.insert(list.end(), cl.begin(), cl.end());
         }
 
diff --git a/modules/contrib/src/lda.cpp b/modules/contrib/src/lda.cpp
index 442d567508..fe60e89237 100644
--- a/modules/contrib/src/lda.cpp
+++ b/modules/contrib/src/lda.cpp
@@ -42,10 +42,10 @@ static Mat argsort(InputArray _src, bool ascending=true)
 {
     Mat src = _src.getMat();
     if (src.rows != 1 && src.cols != 1) {
-        std::string error_message = "Wrong shape of input matrix! Expected a matrix with one row or column.";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "Wrong shape of input matrix! Expected a matrix with one row or column.";
+        CV_Error(Error::StsBadArg, error_message);
     }
-    int flags = CV_SORT_EVERY_ROW+(ascending ? CV_SORT_ASCENDING : CV_SORT_DESCENDING);
+    int flags = SORT_EVERY_ROW | (ascending ? SORT_ASCENDING : SORT_DESCENDING);
     Mat sorted_indices;
     sortIdx(src.reshape(1,1),sorted_indices,flags);
     return sorted_indices;
@@ -54,8 +54,8 @@ static Mat argsort(InputArray _src, bool ascending=true)
 static Mat asRowMatrix(InputArrayOfArrays src, int rtype, double alpha=1, double beta=0) {
     // make sure the input data is a vector of matrices or vector of vector
     if(src.kind() != _InputArray::STD_VECTOR_MAT && src.kind() != _InputArray::STD_VECTOR_VECTOR) {
-        std::string error_message = "The data is expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "The data is expected as InputArray::STD_VECTOR_MAT (a std::vector<Mat>) or _InputArray::STD_VECTOR_VECTOR (a std::vector< std::vector<...> >).";
+        CV_Error(Error::StsBadArg, error_message);
     }
     // number of samples
     size_t n = src.total();
@@ -70,8 +70,8 @@ static Mat asRowMatrix(InputArrayOfArrays src, int rtype, double alpha=1, double
     for(int i = 0; i < (int)n; i++) {
         // make sure data can be reshaped, throw exception if not!
         if(src.getMat(i).total() != d) {
-            std::string error_message = format("Wrong number of elements in matrix #%d! Expected %d was %d.", i, (int)d, (int)src.getMat(i).total());
-            CV_Error(CV_StsBadArg, error_message);
+            String error_message = format("Wrong number of elements in matrix #%d! Expected %d was %d.", i, (int)d, (int)src.getMat(i).total());
+            CV_Error(Error::StsBadArg, error_message);
         }
         // get a hold of the current row
         Mat xi = data.row(i);
@@ -87,7 +87,7 @@ static Mat asRowMatrix(InputArrayOfArrays src, int rtype, double alpha=1, double
 
 static void sortMatrixColumnsByIndices(InputArray _src, InputArray _indices, OutputArray _dst) {
     if(_indices.getMat().type() != CV_32SC1) {
-        CV_Error(CV_StsUnsupportedFormat, "cv::sortColumnsByIndices only works on integer indices!");
+        CV_Error(Error::StsUnsupportedFormat, "cv::sortColumnsByIndices only works on integer indices!");
     }
     Mat src = _src.getMat();
     std::vector<int> indices = _indices.getMat();
@@ -178,13 +178,13 @@ Mat subspaceProject(InputArray _W, InputArray _mean, InputArray _src) {
     int d = src.cols;
     // make sure the data has the correct shape
     if(W.rows != d) {
-        std::string error_message = format("Wrong shapes for given matrices. Was size(src) = (%d,%d), size(W) = (%d,%d).", src.rows, src.cols, W.rows, W.cols);
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong shapes for given matrices. Was size(src) = (%d,%d), size(W) = (%d,%d).", src.rows, src.cols, W.rows, W.cols);
+        CV_Error(Error::StsBadArg, error_message);
     }
     // make sure mean is correct if not empty
     if(!mean.empty() && (mean.total() != (size_t) d)) {
-        std::string error_message = format("Wrong mean shape for the given data matrix. Expected %d, but was %d.", d, mean.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong mean shape for the given data matrix. Expected %d, but was %d.", d, mean.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // create temporary matrices
     Mat X, Y;
@@ -216,13 +216,13 @@ Mat subspaceReconstruct(InputArray _W, InputArray _mean, InputArray _src)
     int d = src.cols;
     // make sure the data has the correct shape
     if(W.cols != d) {
-        std::string error_message = format("Wrong shapes for given matrices. Was size(src) = (%d,%d), size(W) = (%d,%d).", src.rows, src.cols, W.rows, W.cols);
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong shapes for given matrices. Was size(src) = (%d,%d), size(W) = (%d,%d).", src.rows, src.cols, W.rows, W.cols);
+        CV_Error(Error::StsBadArg, error_message);
     }
     // make sure mean is correct if not empty
     if(!mean.empty() && (mean.total() != (size_t) W.rows)) {
-        std::string error_message = format("Wrong mean shape for the given eigenvector matrix. Expected %d, but was %d.", W.cols, mean.total());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("Wrong mean shape for the given eigenvector matrix. Expected %d, but was %d.", W.cols, mean.total());
+        CV_Error(Error::StsBadArg, error_message);
     }
     // initalize temporary matrices
     Mat X, Y;
@@ -936,20 +936,20 @@ public:
 //------------------------------------------------------------------------------
 // Linear Discriminant Analysis implementation
 //------------------------------------------------------------------------------
-void LDA::save(const std::string& filename) const {
+void LDA::save(const String& filename) const {
     FileStorage fs(filename, FileStorage::WRITE);
     if (!fs.isOpened()) {
-        CV_Error(CV_StsError, "File can't be opened for writing!");
+        CV_Error(Error::StsError, "File can't be opened for writing!");
     }
     this->save(fs);
     fs.release();
 }
 
 // Deserializes this object from a given filename.
-void LDA::load(const std::string& filename) {
+void LDA::load(const String& filename) {
     FileStorage fs(filename, FileStorage::READ);
     if (!fs.isOpened())
-       CV_Error(CV_StsError, "File can't be opened for writing!");
+       CV_Error(Error::StsError, "File can't be opened for writing!");
     this->load(fs);
     fs.release();
 }
@@ -1001,13 +1001,13 @@ void LDA::lda(InputArrayOfArrays _src, InputArray _lbls) {
     // we can't do a LDA on one class, what do you
     // want to separate from each other then?
     if(C == 1) {
-        std::string error_message = "At least two classes are needed to perform a LDA. Reason: Only one class was given!";
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = "At least two classes are needed to perform a LDA. Reason: Only one class was given!";
+        CV_Error(Error::StsBadArg, error_message);
     }
     // throw error if less labels, than samples
     if (labels.size() != static_cast<size_t>(N)) {
-        std::string error_message = format("The number of samples must equal the number of labels. Given %d labels, %d samples. ", labels.size(), N);
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message = format("The number of samples must equal the number of labels. Given %d labels, %d samples. ", labels.size(), N);
+        CV_Error(Error::StsBadArg, error_message);
     }
     // warn if within-classes scatter matrix becomes singular
     if (N < D) {
@@ -1089,8 +1089,8 @@ void LDA::compute(InputArrayOfArrays _src, InputArray _lbls) {
         lda(_src.getMat(), _lbls);
         break;
     default:
-        std::string error_message= format("InputArray Datatype %d is not supported.", _src.kind());
-        CV_Error(CV_StsBadArg, error_message);
+        String error_message= format("InputArray Datatype %d is not supported.", _src.kind());
+        CV_Error(Error::StsBadArg, error_message);
         break;
     }
 }
@@ -1106,4 +1106,3 @@ Mat LDA::reconstruct(InputArray src) {
 }
 
 }
-
diff --git a/modules/contrib/src/logpolar_bsm.cpp b/modules/contrib/src/logpolar_bsm.cpp
index 70c7437bbe..8327b0a2da 100644
--- a/modules/contrib/src/logpolar_bsm.cpp
+++ b/modules/contrib/src/logpolar_bsm.cpp
@@ -649,4 +649,3 @@ LogPolar_Adjacent::~LogPolar_Adjacent()
 }
 
 }
-
diff --git a/modules/contrib/src/octree.cpp b/modules/contrib/src/octree.cpp
index 80d2564945..b612372753 100644
--- a/modules/contrib/src/octree.cpp
+++ b/modules/contrib/src/octree.cpp
@@ -258,7 +258,7 @@ namespace cv
 
     void Octree::buildTree(const std::vector<Point3f>& points3d, int maxLevels, int _minPoints)
     {
-        assert((size_t)maxLevels * 8 < MAX_STACK_SIZE);
+        CV_Assert((size_t)maxLevels * 8 < MAX_STACK_SIZE);
         points.resize(points3d.size());
         std::copy(points3d.begin(), points3d.end(), points.begin());
         minPoints = _minPoints;
diff --git a/modules/contrib/src/precomp.cpp b/modules/contrib/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/contrib/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/contrib/src/precomp.hpp b/modules/contrib/src/precomp.hpp
index 7819148ddc..09ffd086ce 100644
--- a/modules/contrib/src/precomp.hpp
+++ b/modules/contrib/src/precomp.hpp
@@ -43,16 +43,13 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/contrib.hpp"
 #include "opencv2/features2d.hpp"
 #include "opencv2/objdetect.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/core/private.hpp"
 
 namespace cv
 {
diff --git a/modules/contrib/src/retina.cpp b/modules/contrib/src/retina.cpp
deleted file mode 100644
index 1464896871..0000000000
--- a/modules/contrib/src/retina.cpp
+++ /dev/null
@@ -1,463 +0,0 @@
-/*#******************************************************************************
- ** IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
- **
- ** By downloading, copying, installing or using the software you agree to this license.
- ** If you do not agree to this license, do not download, install,
- ** copy or use the software.
- **
- **
- ** HVStools : interfaces allowing OpenCV users to integrate Human Vision System models. Presented models originate from Jeanny Herault's original research and have been reused and adapted by the author&collaborators for computed vision applications since his thesis with Alice Caplier at Gipsa-Lab.
- ** Use: extract still images & image sequences features, from contours details to motion spatio-temporal features, etc. for high level visual scene analysis. Also contribute to image enhancement/compression such as tone mapping.
- **
- ** Maintainers : Listic lab (code author current affiliation & applications) and Gipsa Lab (original research origins & applications)
- **
- **  Creation - enhancement process 2007-2011
- **      Author: Alexandre Benoit (benoit.alexandre.vision@gmail.com), LISTIC lab, Annecy le vieux, France
- **
- ** Theses algorithm have been developped by Alexandre BENOIT since his thesis with Alice Caplier at Gipsa-Lab (www.gipsa-lab.inpg.fr) and the research he pursues at LISTIC Lab (www.listic.univ-savoie.fr).
- ** Refer to the following research paper for more information:
- ** Benoit A., Caplier A., Durette B., Herault, J., "USING HUMAN VISUAL SYSTEM MODELING FOR BIO-INSPIRED LOW LEVEL IMAGE PROCESSING", Elsevier, Computer Vision and Image Understanding 114 (2010), pp. 758-773, DOI: http://dx.doi.org/10.1016/j.cviu.2010.01.011
- ** This work have been carried out thanks to Jeanny Herault who's research and great discussions are the basis of all this work, please take a look at his book:
- ** Vision: Images, Signals and Neural Networks: Models of Neural Processing in Visual Perception (Progress in Neural Processing),By: Jeanny Herault, ISBN: 9814273686. WAPI (Tower ID): 113266891.
- **
- ** The retina filter includes the research contributions of phd/research collegues from which code has been redrawn by the author :
- ** _take a look at the retinacolor.hpp module to discover Brice Chaix de Lavarene color mosaicing/demosaicing and the reference paper:
- ** ====> B. Chaix de Lavarene, D. Alleysson, B. Durette, J. Herault (2007). "Efficient demosaicing through recursive filtering", IEEE International Conference on Image Processing ICIP 2007
- ** _take a look at imagelogpolprojection.hpp to discover retina spatial log sampling which originates from Barthelemy Durette phd with Jeanny Herault. A Retina / V1 cortex projection is also proposed and originates from Jeanny's discussions.
- ** ====> more informations in the above cited Jeanny Heraults's book.
- **
- **                          License Agreement
- **               For Open Source Computer Vision Library
- **
- ** Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
- ** Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
- **
- **               For Human Visual System tools (hvstools)
- ** Copyright (C) 2007-2011, LISTIC Lab, Annecy le Vieux and GIPSA Lab, Grenoble, France, all rights reserved.
- **
- ** Third party copyrights are property of their respective owners.
- **
- ** Redistribution and use in source and binary forms, with or without modification,
- ** are permitted provided that the following conditions are met:
- **
- ** * Redistributions of source code must retain the above copyright notice,
- **    this list of conditions and the following disclaimer.
- **
- ** * Redistributions in binary form must reproduce the above copyright notice,
- **    this list of conditions and the following disclaimer in the documentation
- **    and/or other materials provided with the distribution.
- **
- ** * The name of the copyright holders may not be used to endorse or promote products
- **    derived from this software without specific prior written permission.
- **
- ** This software is provided by the copyright holders and contributors "as is" and
- ** any express or implied warranties, including, but not limited to, the implied
- ** warranties of merchantability and fitness for a particular purpose are disclaimed.
- ** In no event shall the Intel Corporation or contributors be liable for any direct,
- ** indirect, incidental, special, exemplary, or consequential damages
- ** (including, but not limited to, procurement of substitute goods or services;
- ** loss of use, data, or profits; or business interruption) however caused
- ** and on any theory of liability, whether in contract, strict liability,
- ** or tort (including negligence or otherwise) arising in any way out of
- ** the use of this software, even if advised of the possibility of such damage.
- *******************************************************************************/
-
-/*
- * Retina.cpp
- *
- *  Created on: Jul 19, 2011
- *      Author: Alexandre Benoit
- */
-#include "precomp.hpp"
-#include "retinafilter.hpp"
-#include <iostream>
-
-namespace cv
-{
-
-Retina::Retina(const cv::Size inputSz)
-{
-    _retinaFilter = 0;
-    _init(inputSz, true, RETINA_COLOR_BAYER, false);
-}
-
-Retina::Retina(const cv::Size inputSz, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
-{
-    _retinaFilter = 0;
-    _init(inputSz, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
-};
-
-Retina::~Retina()
-{
-    if (_retinaFilter)
-        delete _retinaFilter;
-}
-
-/**
-* retreive retina input buffer size
-*/
-Size Retina::inputSize(){return cv::Size(_retinaFilter->getInputNBcolumns(), _retinaFilter->getInputNBrows());}
-
-/**
-* retreive retina output buffer size
-*/
-Size Retina::outputSize(){return cv::Size(_retinaFilter->getOutputNBcolumns(), _retinaFilter->getOutputNBrows());}
-
-
-void Retina::setColorSaturation(const bool saturateColors, const float colorSaturationValue)
-{
-    _retinaFilter->setColorSaturation(saturateColors, colorSaturationValue);
-}
-
-struct Retina::RetinaParameters Retina::getParameters(){return _retinaParameters;}
-
-
-void Retina::setup(std::string retinaParameterFile, const bool applyDefaultSetupOnFailure)
-{
-    try
-    {
-        // opening retinaParameterFile in read mode
-        cv::FileStorage fs(retinaParameterFile, cv::FileStorage::READ);
-        setup(fs, applyDefaultSetupOnFailure);
-    }catch(Exception &e)
-    {
-    std::cout<<"Retina::setup: wrong/unappropriate xml parameter file : error report :`n=>"<<e.what()<<std::endl;
-    if (applyDefaultSetupOnFailure)
-    {
-            std::cout<<"Retina::setup: resetting retina with default parameters"<<std::endl;
-        setupOPLandIPLParvoChannel();
-        setupIPLMagnoChannel();
-    }
-        else
-        {
-        std::cout<<"=> keeping current parameters"<<std::endl;
-        }
-    }
-}
-
-void Retina::setup(cv::FileStorage &fs, const bool applyDefaultSetupOnFailure)
-{
-    try
-    {
-        // read parameters file if it exists or apply default setup if asked for
-        if (!fs.isOpened())
-        {
-            std::cout<<"Retina::setup: provided parameters file could not be open... skeeping configuration"<<std::endl;
-            return;
-            // implicit else case : retinaParameterFile could be open (it exists at least)
-        }
-                // OPL and Parvo init first... update at the same time the parameters structure and the retina core
-        cv::FileNode rootFn = fs.root(), currFn=rootFn["OPLandIPLparvo"];
-        currFn["colorMode"]>>_retinaParameters.OPLandIplParvo.colorMode;
-        currFn["normaliseOutput"]>>_retinaParameters.OPLandIplParvo.normaliseOutput;
-        currFn["photoreceptorsLocalAdaptationSensitivity"]>>_retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
-        currFn["photoreceptorsTemporalConstant"]>>_retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
-        currFn["photoreceptorsSpatialConstant"]>>_retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
-        currFn["horizontalCellsGain"]>>_retinaParameters.OPLandIplParvo.horizontalCellsGain;
-        currFn["hcellsTemporalConstant"]>>_retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
-        currFn["hcellsSpatialConstant"]>>_retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
-        currFn["ganglionCellsSensitivity"]>>_retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
-        setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
-
-        // init retina IPL magno setup... update at the same time the parameters structure and the retina core
-        currFn=rootFn["IPLmagno"];
-        currFn["normaliseOutput"]>>_retinaParameters.IplMagno.normaliseOutput;
-        currFn["parasolCells_beta"]>>_retinaParameters.IplMagno.parasolCells_beta;
-        currFn["parasolCells_tau"]>>_retinaParameters.IplMagno.parasolCells_tau;
-        currFn["parasolCells_k"]>>_retinaParameters.IplMagno.parasolCells_k;
-        currFn["amacrinCellsTemporalCutFrequency"]>>_retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
-        currFn["V0CompressionParameter"]>>_retinaParameters.IplMagno.V0CompressionParameter;
-        currFn["localAdaptintegration_tau"]>>_retinaParameters.IplMagno.localAdaptintegration_tau;
-        currFn["localAdaptintegration_k"]>>_retinaParameters.IplMagno.localAdaptintegration_k;
-
-        setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency,_retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
-
-    }catch(Exception &e)
-    {
-        std::cout<<"Retina::setup: resetting retina with default parameters"<<std::endl;
-        if (applyDefaultSetupOnFailure)
-        {
-            setupOPLandIPLParvoChannel();
-            setupIPLMagnoChannel();
-        }
-        std::cout<<"Retina::setup: wrong/unappropriate xml parameter file : error report :`n=>"<<e.what()<<std::endl;
-        std::cout<<"=> keeping current parameters"<<std::endl;
-    }
-
-    // report current configuration
-    std::cout<<printSetup()<<std::endl;
-}
-
-void Retina::setup(cv::Retina::RetinaParameters newConfiguration)
-{
-    // simply copy structures
-    memcpy(&_retinaParameters, &newConfiguration, sizeof(cv::Retina::RetinaParameters));
-    // apply setup
-    setupOPLandIPLParvoChannel(_retinaParameters.OPLandIplParvo.colorMode, _retinaParameters.OPLandIplParvo.normaliseOutput, _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity, _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant, _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant, _retinaParameters.OPLandIplParvo.horizontalCellsGain, _retinaParameters.OPLandIplParvo.hcellsTemporalConstant, _retinaParameters.OPLandIplParvo.hcellsSpatialConstant, _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity);
-    setupIPLMagnoChannel(_retinaParameters.IplMagno.normaliseOutput, _retinaParameters.IplMagno.parasolCells_beta, _retinaParameters.IplMagno.parasolCells_tau, _retinaParameters.IplMagno.parasolCells_k, _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency,_retinaParameters.IplMagno.V0CompressionParameter, _retinaParameters.IplMagno.localAdaptintegration_tau, _retinaParameters.IplMagno.localAdaptintegration_k);
-
-
-}
-
-const std::string Retina::printSetup()
-{
-    std::stringstream outmessage;
-
-    // displaying OPL and IPL parvo setup
-    outmessage<<"Current Retina instance setup :"
-            <<"\nOPLandIPLparvo"<<"{"
-            << "\n==> colorMode : " << _retinaParameters.OPLandIplParvo.colorMode
-            << "\n==> normalizeParvoOutput :" << _retinaParameters.OPLandIplParvo.normaliseOutput
-            << "\n==> photoreceptorsLocalAdaptationSensitivity : " << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity
-            << "\n==> photoreceptorsTemporalConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant
-            << "\n==> photoreceptorsSpatialConstant : " << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant
-            << "\n==> horizontalCellsGain : " << _retinaParameters.OPLandIplParvo.horizontalCellsGain
-            << "\n==> hcellsTemporalConstant : " << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant
-            << "\n==> hcellsSpatialConstant : " << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant
-            << "\n==> parvoGanglionCellsSensitivity : " << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity
-            <<"}\n";
-
-    // displaying IPL magno setup
-    outmessage<<"Current Retina instance setup :"
-            <<"\nIPLmagno"<<"{"
-            << "\n==> normaliseOutput : " << _retinaParameters.IplMagno.normaliseOutput
-            << "\n==> parasolCells_beta : " << _retinaParameters.IplMagno.parasolCells_beta
-            << "\n==> parasolCells_tau : " << _retinaParameters.IplMagno.parasolCells_tau
-            << "\n==> parasolCells_k : " << _retinaParameters.IplMagno.parasolCells_k
-            << "\n==> amacrinCellsTemporalCutFrequency : " << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency
-            << "\n==> V0CompressionParameter : " << _retinaParameters.IplMagno.V0CompressionParameter
-            << "\n==> localAdaptintegration_tau : " << _retinaParameters.IplMagno.localAdaptintegration_tau
-            << "\n==> localAdaptintegration_k : " << _retinaParameters.IplMagno.localAdaptintegration_k
-            <<"}";
-    return outmessage.str();
-}
-
-void Retina::write( std::string fs ) const
-{
-    FileStorage parametersSaveFile(fs, cv::FileStorage::WRITE );
-    write(parametersSaveFile);
-}
-
-void Retina::write( FileStorage& fs ) const
-{
-    if (!fs.isOpened())
-        return; // basic error case
-    fs<<"OPLandIPLparvo"<<"{";
-    fs << "colorMode" << _retinaParameters.OPLandIplParvo.colorMode;
-    fs << "normaliseOutput" << _retinaParameters.OPLandIplParvo.normaliseOutput;
-    fs << "photoreceptorsLocalAdaptationSensitivity" << _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity;
-    fs << "photoreceptorsTemporalConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant;
-    fs << "photoreceptorsSpatialConstant" << _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant;
-    fs << "horizontalCellsGain" << _retinaParameters.OPLandIplParvo.horizontalCellsGain;
-    fs << "hcellsTemporalConstant" << _retinaParameters.OPLandIplParvo.hcellsTemporalConstant;
-    fs << "hcellsSpatialConstant" << _retinaParameters.OPLandIplParvo.hcellsSpatialConstant;
-    fs << "ganglionCellsSensitivity" << _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity;
-    fs << "}";
-    fs<<"IPLmagno"<<"{";
-    fs << "normaliseOutput" << _retinaParameters.IplMagno.normaliseOutput;
-    fs << "parasolCells_beta" << _retinaParameters.IplMagno.parasolCells_beta;
-    fs << "parasolCells_tau" << _retinaParameters.IplMagno.parasolCells_tau;
-    fs << "parasolCells_k" << _retinaParameters.IplMagno.parasolCells_k;
-    fs << "amacrinCellsTemporalCutFrequency" << _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency;
-    fs << "V0CompressionParameter" << _retinaParameters.IplMagno.V0CompressionParameter;
-    fs << "localAdaptintegration_tau" << _retinaParameters.IplMagno.localAdaptintegration_tau;
-    fs << "localAdaptintegration_k" << _retinaParameters.IplMagno.localAdaptintegration_k;
-    fs<<"}";
-}
-
-void Retina::setupOPLandIPLParvoChannel(const bool colorMode, const bool normaliseOutput, const float photoreceptorsLocalAdaptationSensitivity, const float photoreceptorsTemporalConstant, const float photoreceptorsSpatialConstant, const float horizontalCellsGain, const float HcellsTemporalConstant, const float HcellsSpatialConstant, const float ganglionCellsSensitivity)
-{
-    // retina core parameters setup
-    _retinaFilter->setColorMode(colorMode);
-    _retinaFilter->setPhotoreceptorsLocalAdaptationSensitivity(photoreceptorsLocalAdaptationSensitivity);
-    _retinaFilter->setOPLandParvoParameters(0, photoreceptorsTemporalConstant, photoreceptorsSpatialConstant, horizontalCellsGain, HcellsTemporalConstant, HcellsSpatialConstant, ganglionCellsSensitivity);
-    _retinaFilter->setParvoGanglionCellsLocalAdaptationSensitivity(ganglionCellsSensitivity);
-    _retinaFilter->activateNormalizeParvoOutput_0_maxOutputValue(normaliseOutput);
-
-        // update parameters struture
-
-    _retinaParameters.OPLandIplParvo.colorMode = colorMode;
-    _retinaParameters.OPLandIplParvo.normaliseOutput = normaliseOutput;
-    _retinaParameters.OPLandIplParvo.photoreceptorsLocalAdaptationSensitivity = photoreceptorsLocalAdaptationSensitivity;
-    _retinaParameters.OPLandIplParvo.photoreceptorsTemporalConstant = photoreceptorsTemporalConstant;
-    _retinaParameters.OPLandIplParvo.photoreceptorsSpatialConstant = photoreceptorsSpatialConstant;
-    _retinaParameters.OPLandIplParvo.horizontalCellsGain = horizontalCellsGain;
-    _retinaParameters.OPLandIplParvo.hcellsTemporalConstant = HcellsTemporalConstant;
-    _retinaParameters.OPLandIplParvo.hcellsSpatialConstant = HcellsSpatialConstant;
-    _retinaParameters.OPLandIplParvo.ganglionCellsSensitivity = ganglionCellsSensitivity;
-
-}
-
-void Retina::setupIPLMagnoChannel(const bool normaliseOutput, const float parasolCells_beta, const float parasolCells_tau, const float parasolCells_k, const float amacrinCellsTemporalCutFrequency, const float V0CompressionParameter, const float localAdaptintegration_tau, const float localAdaptintegration_k)
-{
-
-    _retinaFilter->setMagnoCoefficientsTable(parasolCells_beta, parasolCells_tau, parasolCells_k, amacrinCellsTemporalCutFrequency, V0CompressionParameter, localAdaptintegration_tau, localAdaptintegration_k);
-    _retinaFilter->activateNormalizeMagnoOutput_0_maxOutputValue(normaliseOutput);
-
-        // update parameters struture
-    _retinaParameters.IplMagno.normaliseOutput = normaliseOutput;
-    _retinaParameters.IplMagno.parasolCells_beta = parasolCells_beta;
-    _retinaParameters.IplMagno.parasolCells_tau = parasolCells_tau;
-    _retinaParameters.IplMagno.parasolCells_k = parasolCells_k;
-    _retinaParameters.IplMagno.amacrinCellsTemporalCutFrequency = amacrinCellsTemporalCutFrequency;
-    _retinaParameters.IplMagno.V0CompressionParameter = V0CompressionParameter;
-    _retinaParameters.IplMagno.localAdaptintegration_tau = localAdaptintegration_tau;
-    _retinaParameters.IplMagno.localAdaptintegration_k = localAdaptintegration_k;
-}
-
-void Retina::run(const cv::Mat &inputMatToConvert)
-{
-    // first convert input image to the compatible format : std::valarray<float>
-    const bool colorMode = _convertCvMat2ValarrayBuffer(inputMatToConvert, _inputBuffer);
-    // process the retina
-    if (!_retinaFilter->runFilter(_inputBuffer, colorMode, false, _retinaParameters.OPLandIplParvo.colorMode && colorMode, false))
-        throw cv::Exception(-1, "Retina cannot be applied, wrong input buffer size", "Retina::run", "Retina.h", 0);
-}
-
-void Retina::getParvo(cv::Mat &retinaOutput_parvo)
-{
-    if (_retinaFilter->getColorMode())
-    {
-        // reallocate output buffer (if necessary)
-        _convertValarrayBuffer2cvMat(_retinaFilter->getColorOutput(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), true, retinaOutput_parvo);
-    }else
-    {
-        // reallocate output buffer (if necessary)
-        _convertValarrayBuffer2cvMat(_retinaFilter->getContours(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), false, retinaOutput_parvo);
-    }
-    //retinaOutput_parvo/=255.0;
-}
-void Retina::getMagno(cv::Mat &retinaOutput_magno)
-{
-    // reallocate output buffer (if necessary)
-    _convertValarrayBuffer2cvMat(_retinaFilter->getMovingContours(), _retinaFilter->getOutputNBrows(), _retinaFilter->getOutputNBcolumns(), false, retinaOutput_magno);
-    //retinaOutput_magno/=255.0;
-}
-
-// original API level data accessors : copy buffers if size matches
-void Retina::getMagno(std::valarray<float> &magnoOutputBufferCopy){if (magnoOutputBufferCopy.size()==_retinaFilter->getMovingContours().size()) magnoOutputBufferCopy = _retinaFilter->getMovingContours();}
-void Retina::getParvo(std::valarray<float> &parvoOutputBufferCopy){if (parvoOutputBufferCopy.size()==_retinaFilter->getContours().size()) parvoOutputBufferCopy = _retinaFilter->getContours();}
-// original API level data accessors : get buffers addresses...
-const std::valarray<float> & Retina::getMagno() const {return _retinaFilter->getMovingContours();}
-const std::valarray<float> & Retina::getParvo() const {if (_retinaFilter->getColorMode())return _retinaFilter->getColorOutput(); /* implicite else */return _retinaFilter->getContours();}
-
-// private method called by constructirs
-void Retina::_init(const cv::Size inputSz, const bool colorMode, RETINA_COLORSAMPLINGMETHOD colorSamplingMethod, const bool useRetinaLogSampling, const double reductionFactor, const double samplingStrenght)
-{
-    // basic error check
-    if (inputSz.height*inputSz.width <= 0)
-        throw cv::Exception(-1, "Bad retina size setup : size height and with must be superior to zero", "Retina::setup", "Retina.h", 0);
-
-    unsigned int nbPixels=inputSz.height*inputSz.width;
-    // resize buffers if size does not match
-    _inputBuffer.resize(nbPixels*3); // buffer supports gray images but also 3 channels color buffers... (larger is better...)
-
-    // allocate the retina model
-        if (_retinaFilter)
-           delete _retinaFilter;
-    _retinaFilter = new RetinaFilter(inputSz.height, inputSz.width, colorMode, colorSamplingMethod, useRetinaLogSampling, reductionFactor, samplingStrenght);
-
-    // prepare the default parameter XML file with default setup
-        setup(_retinaParameters);
-
-    // init retina
-    _retinaFilter->clearAllBuffers();
-
-    // report current configuration
-    std::cout<<printSetup()<<std::endl;
-}
-
-void Retina::_convertValarrayBuffer2cvMat(const std::valarray<float> &grayMatrixToConvert, const unsigned int nbRows, const unsigned int nbColumns, const bool colorMode, cv::Mat &outBuffer)
-{
-    // fill output buffer with the valarray buffer
-    const float *valarrayPTR=get_data(grayMatrixToConvert);
-    if (!colorMode)
-    {
-        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8U);
-        for (unsigned int i=0;i<nbRows;++i)
-        {
-            for (unsigned int j=0;j<nbColumns;++j)
-            {
-                cv::Point2d pixel(j,i);
-                outBuffer.at<unsigned char>(pixel)=(unsigned char)*(valarrayPTR++);
-            }
-        }
-    }else
-    {
-        const unsigned int doubleNBpixels=_retinaFilter->getOutputNBpixels()*2;
-        outBuffer.create(cv::Size(nbColumns, nbRows), CV_8UC3);
-        for (unsigned int i=0;i<nbRows;++i)
-        {
-            for (unsigned int j=0;j<nbColumns;++j,++valarrayPTR)
-            {
-                cv::Point2d pixel(j,i);
-                cv::Vec3b pixelValues;
-                pixelValues[2]=(unsigned char)*(valarrayPTR);
-                pixelValues[1]=(unsigned char)*(valarrayPTR+_retinaFilter->getOutputNBpixels());
-                pixelValues[0]=(unsigned char)*(valarrayPTR+doubleNBpixels);
-
-                outBuffer.at<cv::Vec3b>(pixel)=pixelValues;
-            }
-        }
-    }
-}
-
-bool Retina::_convertCvMat2ValarrayBuffer(const cv::Mat inputMatToConvert, std::valarray<float> &outputValarrayMatrix)
-{
-    // first check input consistency
-    if (inputMatToConvert.empty())
-        throw cv::Exception(-1, "Retina cannot be applied, input buffer is empty", "Retina::run", "Retina.h", 0);
-
-    // retreive color mode from image input
-    int imageNumberOfChannels = inputMatToConvert.channels();
-
-        // convert to float AND fill the valarray buffer
-    typedef float T; // define here the target pixel format, here, float
-        const int dsttype = DataType<T>::depth; // output buffer is float format
-
-
-    if(imageNumberOfChannels==4)
-    {
-    // create a cv::Mat table (for RGBA planes)
-        cv::Mat planes[4] =
-        {
-            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[_retinaFilter->getInputNBpixels()*2]),
-            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[_retinaFilter->getInputNBpixels()]),
-            cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
-        };
-        planes[3] = cv::Mat(inputMatToConvert.size(), dsttype);     // last channel (alpha) does not point on the valarray (not usefull in our case)
-        // split color cv::Mat in 4 planes... it fills valarray directely
-        cv::split(cv::Mat_<Vec<T, 4> >(inputMatToConvert), planes);
-    }
-    else if (imageNumberOfChannels==3)
-    {
-        // create a cv::Mat table (for RGB planes)
-        cv::Mat planes[] =
-        {
-        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[_retinaFilter->getInputNBpixels()*2]),
-        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[_retinaFilter->getInputNBpixels()]),
-        cv::Mat(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0])
-        };
-        // split color cv::Mat in 3 planes... it fills valarray directely
-        cv::split(cv::Mat_<Vec<T, 3> >(inputMatToConvert), planes);
-    }
-    else if(imageNumberOfChannels==1)
-    {
-        // create a cv::Mat header for the valarray
-        cv::Mat dst(inputMatToConvert.size(), dsttype, &outputValarrayMatrix[0]);
-        inputMatToConvert.convertTo(dst, dsttype);
-    }
-        else
-            CV_Error(CV_StsUnsupportedFormat, "input image must be single channel (gray levels), bgr format (color) or bgra (color with transparency which won't be considered");
-
-    return imageNumberOfChannels>1; // return bool : false for gray level image processing, true for color mode
-}
-
-void Retina::clearBuffers() {_retinaFilter->clearAllBuffers();}
-
-void Retina::activateMovingContoursProcessing(const bool activate){_retinaFilter->activateMovingContoursProcessing(activate);}
-
-void Retina::activateContoursProcessing(const bool activate){_retinaFilter->activateContoursProcessing(activate);}
-
-} // end of namespace cv
-
diff --git a/modules/contrib/src/rgbdodometry.cpp b/modules/contrib/src/rgbdodometry.cpp
index 1fa800b0ae..6f86f17ec4 100644
--- a/modules/contrib/src/rgbdodometry.cpp
+++ b/modules/contrib/src/rgbdodometry.cpp
@@ -54,7 +54,6 @@
 #include <iostream>
 #include <limits>
 
-#include "opencv2/core/internal.hpp"
 #if defined(HAVE_EIGEN) && EIGEN_WORLD_VERSION == 3
 #  ifdef ANDROID
      template <typename Scalar> Scalar log2(Scalar v) { return std::log(v)/std::log(Scalar(2)); }
@@ -423,7 +422,7 @@ bool computeKsi( int transformType,
         computeCFuncPtr = computeC_Translation;
     }
     else
-        CV_Error( CV_StsBadFlag, "Unsupported value of transformation type flag.");
+        CV_Error(Error::StsBadFlag, "Unsupported value of transformation type flag.");
 
     Mat C( correspsCount, Cwidth, CV_64FC1 );
     Mat dI_dt( correspsCount, 1, CV_64FC1 );
diff --git a/modules/contrib/src/spinimages.cpp b/modules/contrib/src/spinimages.cpp
index 7f7a8ad63f..4e58472bfb 100644
--- a/modules/contrib/src/spinimages.cpp
+++ b/modules/contrib/src/spinimages.cpp
@@ -56,24 +56,24 @@ namespace
 {
     const static Scalar colors[] =
     {
-        CV_RGB(255,   0,   0),
-        CV_RGB(  0, 255,   0),
-        CV_RGB(  0,   0, 255),
-        CV_RGB(255, 255,   0),
-        CV_RGB(255,   0, 255),
-        CV_RGB(  0, 255, 255),
-        CV_RGB(255, 127, 127),
-        CV_RGB(127, 127, 255),
-        CV_RGB(127, 255, 127),
-        CV_RGB(255, 255, 127),
-        CV_RGB(127, 255, 255),
-        CV_RGB(255, 127, 255),
-        CV_RGB(127,   0,   0),
-        CV_RGB(  0, 127,   0),
-        CV_RGB(  0,   0, 127),
-        CV_RGB(127, 127,   0),
-        CV_RGB(127,   0, 127),
-        CV_RGB(  0, 127, 127)
+        Scalar(255,   0,   0),
+        Scalar(  0, 255,   0),
+        Scalar(  0,   0, 255),
+        Scalar(255, 255,   0),
+        Scalar(255,   0, 255),
+        Scalar(  0, 255, 255),
+        Scalar(255, 127, 127),
+        Scalar(127, 127, 255),
+        Scalar(127, 255, 127),
+        Scalar(255, 255, 127),
+        Scalar(127, 255, 255),
+        Scalar(255, 127, 255),
+        Scalar(127,   0,   0),
+        Scalar(  0, 127,   0),
+        Scalar(  0,   0, 127),
+        Scalar(127, 127,   0),
+        Scalar(127,   0, 127),
+        Scalar(  0, 127, 127)
     };
     size_t colors_mum = sizeof(colors)/sizeof(colors[0]);
 
@@ -199,7 +199,7 @@ void convertTransformMatrix(const float* matrix, float* sseMatrix)
 
 inline __m128 transformSSE(const __m128* matrix, const __m128& in)
 {
-    assert(((size_t)matrix & 15) == 0);
+    CV_DbgAssert(((size_t)matrix & 15) == 0);
     __m128 a0 = _mm_mul_ps(_mm_load_ps((float*)(matrix+0)), _mm_shuffle_ps(in,in,_MM_SHUFFLE(0,0,0,0)));
     __m128 a1 = _mm_mul_ps(_mm_load_ps((float*)(matrix+1)), _mm_shuffle_ps(in,in,_MM_SHUFFLE(1,1,1,1)));
     __m128 a2 = _mm_mul_ps(_mm_load_ps((float*)(matrix+2)), _mm_shuffle_ps(in,in,_MM_SHUFFLE(2,2,2,2)));
@@ -221,8 +221,8 @@ void computeSpinImages( const Octree& Octree, const std::vector<Point3f>& points
     float pixelsPerMeter = 1.f / binSize;
     float support = imageWidth * binSize;
 
-    assert(normals.size() == points.size());
-    assert(mask.size() == points.size());
+    CV_Assert(normals.size() == points.size());
+    CV_Assert(mask.size() == points.size());
 
     size_t points_size = points.size();
     mask.resize(points_size);
@@ -250,7 +250,7 @@ void computeSpinImages( const Octree& Octree, const std::vector<Point3f>& points
         if (mask[i] == 0)
             continue;
 
-        int t = cvGetThreadNum();
+        int t = getThreadNum();
         std::vector<Point3f>& pointsInSphere = pointsInSpherePool[t];
 
         const Point3f& center = points[i];
@@ -289,7 +289,7 @@ void computeSpinImages( const Octree& Octree, const std::vector<Point3f>& points
             __m128 ppm4 = _mm_set1_ps(pixelsPerMeter);
             __m128i height4m1 = _mm_set1_epi32(spinImage.rows-1);
             __m128i width4m1 = _mm_set1_epi32(spinImage.cols-1);
-            assert( spinImage.step <= 0xffff );
+            CV_Assert( spinImage.step <= 0xffff );
             __m128i step4 = _mm_set1_epi16((short)step);
             __m128i zero4 = _mm_setzero_si128();
             __m128i one4i = _mm_set1_epi32(1);
@@ -472,7 +472,7 @@ float cv::Mesh3D::estimateResolution(float /*tryRatio*/)
 
     return resolution = (float)dist[ dist.size() / 2 ];
 #else
-    CV_Error(CV_StsNotImplemented, "");
+    CV_Error(Error::StsNotImplemented, "");
     return 1.f;
 #endif
 }
@@ -494,7 +494,7 @@ void cv::Mesh3D::computeNormals(const std::vector<int>& subset, float normalRadi
     ::computeNormals(octree, vtx, normals, mask, normalRadius, minNeighbors);
 }
 
-void cv::Mesh3D::writeAsVrml(const std::string& file, const std::vector<Scalar>& _colors) const
+void cv::Mesh3D::writeAsVrml(const String& file, const std::vector<Scalar>& _colors) const
 {
     std::ofstream ofs(file.c_str());
 
@@ -686,16 +686,15 @@ inline float cv::SpinImageModel::groupingCreteria(const Point3f& pointScene1, co
 }
 
 
-cv::SpinImageModel::SpinImageModel(const Mesh3D& _mesh) : mesh(_mesh) , out(0)
+cv::SpinImageModel::SpinImageModel(const Mesh3D& _mesh) : mesh(_mesh)
 {
      if (mesh.vtx.empty())
          throw Mesh3D::EmptyMeshException();
     defaultParams();
 }
-cv::SpinImageModel::SpinImageModel() : out(0) { defaultParams(); }
-cv::SpinImageModel::~SpinImageModel() {}
 
-void cv::SpinImageModel::setLogger(std::ostream* log) { out = log; }
+cv::SpinImageModel::SpinImageModel() { defaultParams(); }
+cv::SpinImageModel::~SpinImageModel() {}
 
 void cv::SpinImageModel::defaultParams()
 {
@@ -756,7 +755,7 @@ Mat cv::SpinImageModel::packRandomScaledSpins(bool separateScale, size_t xCount,
     int sz = spins.front().cols;
 
     Mat result((int)(yCount * sz + (yCount - 1)), (int)(xCount * sz + (xCount - 1)), CV_8UC3);
-    result = colors[(static_cast<int64>(cvGetTickCount()/cvGetTickFrequency())/1000) % colors_mum];
+    result = colors[(static_cast<int64>(getTickCount()/getTickFrequency())/1000) % colors_mum];
 
     int pos = 0;
     for(int y = 0; y < (int)yCount; ++y)
@@ -770,7 +769,7 @@ Mat cv::SpinImageModel::packRandomScaledSpins(bool separateScale, size_t xCount,
                 int endx   = (x + 1) * sz + x;
 
                 Mat color;
-                cvtColor(spins[pos++], color, CV_GRAY2BGR);
+                cvtColor(spins[pos++], color, COLOR_GRAY2BGR);
                 Mat roi = result(Range(starty, endy), Range(startx, endx));
                 color.copyTo(roi);
             }
@@ -809,7 +808,7 @@ void cv::SpinImageModel::selectRandomSubset(float ratio)
             left[pos] = left.back();
             left.resize(left.size() - 1);
         }
-        sort(subset, std::less<int>());
+        std::sort(subset.begin(), subset.end());
     }
 }
 
@@ -928,7 +927,7 @@ void cv::SpinImageModel::matchSpinToModel(const Mat& spin, std::vector<int>& ind
     if(total < 5)
         return;
 
-    sort(cleanCorrs, std::less<float>());
+    std::sort(cleanCorrs.begin(), cleanCorrs.end());
 
     float lower_fourth = cleanCorrs[(1 * total) / 4 - 1];
     float upper_fourth = cleanCorrs[(3 * total) / 4 - 0];
@@ -1016,7 +1015,7 @@ private:
         std::vector<int> nonzero(model.spinImages.rows);
         for(int i = 0; i < model.spinImages.rows; ++i)
             nonzero[i] = countNonZero(model.spinImages.row(i));
-        sort(nonzero, std::less<int>());
+        std::sort(nonzero.begin(), nonzero.end());
         model.lambda = static_cast<float>( nonzero[ nonzero.size()/2 ] ) / 2;
     }
 
@@ -1030,12 +1029,8 @@ private:
         matchSpinToModel(scene.spinImages.row(i), indeces, coeffs);
         for(size_t t = 0; t < indeces.size(); ++t)
             allMatches.push_back(Match(i, indeces[t], coeffs[t]));
-
-        if (out) if (i % 100 == 0) *out << "Comparing scene spinimage " << i << " of " << scene.spinImages.rows << std::endl;
     }
     corr_timer.stop();
-    if (out) *out << "Spin correlation time  = " << corr_timer << std::endl;
-    if (out) *out << "Matches number = " << allMatches.size() << std::endl;
 
     if(allMatches.empty())
         return;
@@ -1046,7 +1041,6 @@ private:
     allMatches.erase(
         remove_if(allMatches.begin(), allMatches.end(), bind2nd(std::less<float>(), maxMeasure * fraction)),
         allMatches.end());
-    if (out) *out << "Matches number [filtered by similarity measure] = " << allMatches.size() << std::endl;
 
     int matchesSize = (int)allMatches.size();
     if(matchesSize == 0)
@@ -1095,15 +1089,12 @@ private:
     allMatches.erase(
       std::remove_if(allMatches.begin(), allMatches.end(), std::bind2nd(std::equal_to<float>(), infinity)),
       allMatches.end());
-    if (out) *out << "Matches number [filtered by geometric consistency] = " << allMatches.size() << std::endl;
 
 
     matchesSize = (int)allMatches.size();
     if(matchesSize == 0)
         return;
 
-    if (out) *out << "grouping ..." << std::endl;
-
     Mat groupingMat((int)matchesSize, (int)matchesSize, CV_32F);
     groupingMat = Scalar(0);
 
@@ -1151,8 +1142,6 @@ private:
 
     for(int g = 0; g < matchesSize; ++g)
     {
-        if (out) if (g % 100 == 0) *out << "G = " << g << std::endl;
-
         group_t left = allMatchesInds;
         group_t group;
 
@@ -1201,16 +1190,16 @@ private:
 
 cv::TickMeter::TickMeter() { reset(); }
 int64 cv::TickMeter::getTimeTicks() const { return sumTime; }
-double cv::TickMeter::getTimeMicro() const { return (double)getTimeTicks()/cvGetTickFrequency(); }
+double cv::TickMeter::getTimeMicro() const { return (double)getTimeTicks()/getTickFrequency(); }
 double cv::TickMeter::getTimeMilli() const { return getTimeMicro()*1e-3; }
 double cv::TickMeter::getTimeSec()   const { return getTimeMilli()*1e-3; }
 int64 cv::TickMeter::getCounter() const { return counter; }
 void  cv::TickMeter::reset() {startTime = 0; sumTime = 0; counter = 0; }
 
-void cv::TickMeter::start(){ startTime = cvGetTickCount(); }
+void cv::TickMeter::start(){ startTime = getTickCount(); }
 void cv::TickMeter::stop()
 {
-    int64 time = cvGetTickCount();
+    int64 time = getTickCount();
     if ( startTime == 0 )
         return;
 
@@ -1220,4 +1209,4 @@ void cv::TickMeter::stop()
     startTime = 0;
 }
 
-std::ostream& cv::operator<<(std::ostream& out, const TickMeter& tm){ return out << tm.getTimeSec() << "sec"; }
+//std::ostream& cv::operator<<(std::ostream& out, const TickMeter& tm){ return out << tm.getTimeSec() << "sec"; }
diff --git a/modules/contrib/src/stereovar.cpp b/modules/contrib/src/stereovar.cpp
index a825ad8d82..b2141b1f15 100644
--- a/modules/contrib/src/stereovar.cpp
+++ b/modules/contrib/src/stereovar.cpp
@@ -239,8 +239,8 @@ void StereoVar::VariationalSolver(Mat &I1, Mat &I2, Mat &I2x, Mat &u, int level)
 
 void StereoVar::VCycle_MyFAS(Mat &I1, Mat &I2, Mat &I2x, Mat &_u, int level)
 {
-    CvSize imgSize = _u.size();
-    CvSize frmSize = cvSize((int) (imgSize.width * pyrScale + 0.5), (int) (imgSize.height * pyrScale + 0.5));
+    Size imgSize = _u.size();
+    Size frmSize = Size((int) (imgSize.width * pyrScale + 0.5), (int) (imgSize.height * pyrScale + 0.5));
     Mat I1_h, I2_h, I2x_h, u_h, U, U_h;
 
     //PRE relaxation
@@ -285,7 +285,7 @@ void StereoVar::VCycle_MyFAS(Mat &I1, Mat &I2, Mat &I2x, Mat &_u, int level)
 void StereoVar::FMG(Mat &I1, Mat &I2, Mat &I2x, Mat &u, int level)
 {
     double  scale = std::pow(pyrScale, (double) level);
-    CvSize  frmSize = cvSize((int) (u.cols * scale + 0.5), (int) (u.rows * scale + 0.5));
+    Size  frmSize = Size((int) (u.cols * scale + 0.5), (int) (u.rows * scale + 0.5));
     Mat I1_h, I2_h, I2x_h, u_h;
 
     //scaling DOWN
@@ -350,7 +350,7 @@ void StereoVar::autoParams()
 void StereoVar::operator ()( const Mat& left, const Mat& right, Mat& disp )
 {
     CV_Assert(left.size() == right.size() && left.type() == right.type());
-    CvSize imgSize = left.size();
+    Size imgSize = left.size();
     int MaxD = MAX(labs(minDisp), labs(maxDisp));
     int SignD = 1; if (MIN(minDisp, maxDisp) < 0) SignD = -1;
     if (minDisp >= maxDisp) {MaxD = 256; SignD = 1;}
@@ -367,8 +367,8 @@ void StereoVar::operator ()( const Mat& left, const Mat& right, Mat& disp )
     // Preprocessing
     Mat leftgray, rightgray;
     if (left.type() != CV_8UC1) {
-        cvtColor(left, leftgray, CV_BGR2GRAY);
-        cvtColor(right, rightgray, CV_BGR2GRAY);
+        cvtColor(left, leftgray, COLOR_BGR2GRAY);
+        cvtColor(right, rightgray, COLOR_BGR2GRAY);
     } else {
         left.copyTo(leftgray);
         right.copyTo(rightgray);
@@ -378,8 +378,8 @@ void StereoVar::operator ()( const Mat& left, const Mat& right, Mat& disp )
         equalizeHist(rightgray, rightgray);
     }
     if (poly_sigma > 0.0001) {
-        GaussianBlur(leftgray, leftgray, cvSize(poly_n, poly_n), poly_sigma);
-        GaussianBlur(rightgray, rightgray, cvSize(poly_n, poly_n), poly_sigma);
+        GaussianBlur(leftgray, leftgray, Size(poly_n, poly_n), poly_sigma);
+        GaussianBlur(rightgray, rightgray, Size(poly_n, poly_n), poly_sigma);
     }
 
     if (flags & USE_AUTO_PARAMS) {
@@ -408,4 +408,4 @@ void StereoVar::operator ()( const Mat& left, const Mat& right, Mat& disp )
 
     u.release();
 }
-} // namespace
\ No newline at end of file
+} // namespace
diff --git a/modules/contrib/test/test_precomp.cpp b/modules/contrib/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/contrib/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/contrib/test/test_precomp.hpp b/modules/contrib/test/test_precomp.hpp
index de9e283ee2..992ff28dfa 100644
--- a/modules/contrib/test/test_precomp.hpp
+++ b/modules/contrib/test/test_precomp.hpp
@@ -14,4 +14,3 @@
 #include <iostream>
 
 #endif
-
diff --git a/modules/core/CMakeLists.txt b/modules/core/CMakeLists.txt
index e538d1acae..b83a591490 100644
--- a/modules/core/CMakeLists.txt
+++ b/modules/core/CMakeLists.txt
@@ -1,12 +1,24 @@
 set(the_description "The Core Functionality")
-ocv_add_module(core ${ZLIB_LIBRARIES})
-ocv_module_include_directories(${ZLIB_INCLUDE_DIR} "${OpenCV_SOURCE_DIR}/modules/gpu/include")
+ocv_add_module(core ${ZLIB_LIBRARIES} OPTIONAL opencv_cudev)
+ocv_module_include_directories(${ZLIB_INCLUDE_DIRS})
 
-if(HAVE_CUDA)
-  ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef)
+if (HAVE_WINRT)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /ZW /GS /Gm- /AI\"${WINDOWS_SDK_PATH}/References/CommonConfiguration/Neutral\" /AI\"${VISUAL_STUDIO_PATH}/vcpackages\"")
 endif()
 
-ocv_glob_module_sources(SOURCES "${opencv_core_BINARY_DIR}/version_string.inc")
+if(HAVE_CUDA)
+  ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wenum-compare -Wunused-function)
+endif()
+
+file(GLOB lib_cuda_hdrs        "include/opencv2/${name}/cuda/*.hpp"        "include/opencv2/${name}/cuda/*.h")
+file(GLOB lib_cuda_hdrs_detail "include/opencv2/${name}/cuda/detail/*.hpp" "include/opencv2/${name}/cuda/detail/*.h")
+
+source_group("Cuda Headers"         FILES ${lib_cuda_hdrs})
+source_group("Cuda Headers\\Detail" FILES ${lib_cuda_hdrs_detail})
+
+ocv_glob_module_sources(SOURCES "${opencv_core_BINARY_DIR}/version_string.inc"
+                        HEADERS ${lib_cuda_hdrs} ${lib_cuda_hdrs_detail})
+
 ocv_create_module()
 ocv_add_precompiled_headers(${the_module})
 
diff --git a/modules/core/doc/basic_structures.rst b/modules/core/doc/basic_structures.rst
index d9cc303315..93e7ca479c 100644
--- a/modules/core/doc/basic_structures.rst
+++ b/modules/core/doc/basic_structures.rst
@@ -49,6 +49,43 @@ Point\_
 -------
 .. ocv:class:: Point_
 
+::
+
+    template<typename _Tp> class CV_EXPORTS Point_
+    {
+    public:
+        typedef _Tp value_type;
+
+        // various constructors
+        Point_();
+        Point_(_Tp _x, _Tp _y);
+        Point_(const Point_& pt);
+        Point_(const CvPoint& pt);
+        Point_(const CvPoint2D32f& pt);
+        Point_(const Size_<_Tp>& sz);
+        Point_(const Vec<_Tp, 2>& v);
+
+        Point_& operator = (const Point_& pt);
+        //! conversion to another data type
+        template<typename _Tp2> operator Point_<_Tp2>() const;
+
+        //! conversion to the old-style C structures
+        operator CvPoint() const;
+        operator CvPoint2D32f() const;
+        operator Vec<_Tp, 2>() const;
+
+        //! dot product
+        _Tp dot(const Point_& pt) const;
+        //! dot product computed in double-precision arithmetics
+        double ddot(const Point_& pt) const;
+        //! cross-product
+        double cross(const Point_& pt) const;
+        //! checks whether the point is inside the specified rectangle
+        bool inside(const Rect_<_Tp>& r) const;
+
+        _Tp x, y; //< the point coordinates
+    };
+
 Template class for 2D points specified by its coordinates
 :math:`x` and
 :math:`y` .
@@ -84,6 +121,39 @@ Point3\_
 --------
 .. ocv:class:: Point3_
 
+::
+
+    template<typename _Tp> class CV_EXPORTS Point3_
+    {
+    public:
+        typedef _Tp value_type;
+
+        // various constructors
+        Point3_();
+        Point3_(_Tp _x, _Tp _y, _Tp _z);
+        Point3_(const Point3_& pt);
+        explicit Point3_(const Point_<_Tp>& pt);
+        Point3_(const CvPoint3D32f& pt);
+        Point3_(const Vec<_Tp, 3>& v);
+
+        Point3_& operator = (const Point3_& pt);
+        //! conversion to another data type
+        template<typename _Tp2> operator Point3_<_Tp2>() const;
+        //! conversion to the old-style CvPoint...
+        operator CvPoint3D32f() const;
+        //! conversion to cv::Vec<>
+        operator Vec<_Tp, 3>() const;
+
+        //! dot product
+        _Tp dot(const Point3_& pt) const;
+        //! dot product computed in double-precision arithmetics
+        double ddot(const Point3_& pt) const;
+        //! cross product of the 2 3D points
+        Point3_ cross(const Point3_& pt) const;
+
+        _Tp x, y, z; //< the point coordinates
+    };
+
 Template class for 3D points specified by its coordinates
 :math:`x`,
 :math:`y` and
@@ -100,6 +170,35 @@ Size\_
 ------
 .. ocv:class:: Size_
 
+::
+
+    template<typename _Tp> class CV_EXPORTS Size_
+    {
+    public:
+        typedef _Tp value_type;
+
+        //! various constructors
+        Size_();
+        Size_(_Tp _width, _Tp _height);
+        Size_(const Size_& sz);
+        Size_(const CvSize& sz);
+        Size_(const CvSize2D32f& sz);
+        Size_(const Point_<_Tp>& pt);
+
+        Size_& operator = (const Size_& sz);
+        //! the area (width*height)
+        _Tp area() const;
+
+        //! conversion of another data type.
+        template<typename _Tp2> operator Size_<_Tp2>() const;
+
+        //! conversion to the old-style OpenCV types
+        operator CvSize() const;
+        operator CvSize2D32f() const;
+
+        _Tp width, height; // the width and the height
+    };
+
 Template class for specifying the size of an image or rectangle. The class includes two members called ``width`` and ``height``. The structure can be converted to and from the old OpenCV structures
 ``CvSize`` and ``CvSize2D32f`` . The same set of arithmetic and comparison operations as for ``Point_`` is available.
 
@@ -113,6 +212,43 @@ Rect\_
 ------
 .. ocv:class:: Rect_
 
+::
+
+    template<typename _Tp> class CV_EXPORTS Rect_
+    {
+    public:
+        typedef _Tp value_type;
+
+        //! various constructors
+        Rect_();
+        Rect_(_Tp _x, _Tp _y, _Tp _width, _Tp _height);
+        Rect_(const Rect_& r);
+        Rect_(const CvRect& r);
+        Rect_(const Point_<_Tp>& org, const Size_<_Tp>& sz);
+        Rect_(const Point_<_Tp>& pt1, const Point_<_Tp>& pt2);
+
+        Rect_& operator = ( const Rect_& r );
+        //! the top-left corner
+        Point_<_Tp> tl() const;
+        //! the bottom-right corner
+        Point_<_Tp> br() const;
+
+        //! size (width, height) of the rectangle
+        Size_<_Tp> size() const;
+        //! area (width*height) of the rectangle
+        _Tp area() const;
+
+        //! conversion to another data type
+        template<typename _Tp2> operator Rect_<_Tp2>() const;
+        //! conversion to the old-style CvRect
+        operator CvRect() const;
+
+        //! checks whether the rectangle contains the point
+        bool contains(const Point_<_Tp>& pt) const;
+
+        _Tp x, y, width, height; //< the top-left corner, as well as width and height of the rectangle
+    };
+
 Template class for 2D rectangles, described by the following parameters:
 
 * Coordinates of the top-left corner. This is a default interpretation of ``Rect_::x`` and ``Rect_::y`` in OpenCV. Though, in your algorithms you may count ``x`` and ``y`` from the bottom-left corner.
@@ -171,11 +307,32 @@ RotatedRect
 -----------
 .. ocv:class:: RotatedRect
 
+::
+
+    class CV_EXPORTS RotatedRect
+    {
+    public:
+        //! various constructors
+        RotatedRect();
+        RotatedRect(const Point2f& center, const Size2f& size, float angle);
+        RotatedRect(const CvBox2D& box);
+
+        //! returns 4 vertices of the rectangle
+        void points(Point2f pts[]) const;
+        //! returns the minimal up-right rectangle containing the rotated rectangle
+        Rect boundingRect() const;
+        //! conversion to the old-style CvBox2D structure
+        operator CvBox2D() const;
+
+        Point2f center; //< the rectangle mass center
+        Size2f size;    //< width and height of the rectangle
+        float angle;    //< the rotation angle. When the angle is 0, 90, 180, 270 etc., the rectangle becomes an up-right rectangle.
+    };
+
 The class represents rotated (i.e. not up-right) rectangles on a plane. Each rectangle is specified by the center point (mass center), length of each side (represented by cv::Size2f structure) and the rotation angle in degrees.
 
     .. ocv:function:: RotatedRect::RotatedRect()
     .. ocv:function:: RotatedRect::RotatedRect(const Point2f& center, const Size2f& size, float angle)
-    .. ocv:function:: RotatedRect::RotatedRect(const CvBox2D& box)
 
         :param center: The rectangle mass center.
         :param size: Width and height of the rectangle.
@@ -184,7 +341,6 @@ The class represents rotated (i.e. not up-right) rectangles on a plane. Each rec
 
     .. ocv:function:: void RotatedRect::points( Point2f pts[] ) const
     .. ocv:function:: Rect RotatedRect::boundingRect() const
-    .. ocv:function:: RotatedRect::operator CvBox2D() const
 
         :param pts: The points array for storing rectangle vertices.
 
@@ -219,7 +375,33 @@ TermCriteria
 ------------
 .. ocv:class:: TermCriteria
 
-  The class defining termination criteria for iterative algorithms. You can initialize it by default constructor and then override any parameters, or the structure may be fully initialized using the advanced variant of the constructor.
+::
+
+    class CV_EXPORTS TermCriteria
+    {
+    public:
+        enum
+        {
+            COUNT=1, //!< the maximum number of iterations or elements to compute
+            MAX_ITER=COUNT, //!< ditto
+            EPS=2 //!< the desired accuracy or change in parameters at which the iterative algorithm stops
+        };
+
+        //! default constructor
+        TermCriteria();
+        //! full constructor
+        TermCriteria(int type, int maxCount, double epsilon);
+        //! conversion from CvTermCriteria
+        TermCriteria(const CvTermCriteria& criteria);
+        //! conversion to CvTermCriteria
+        operator CvTermCriteria() const;
+
+        int type; //!< the type of termination criteria: COUNT, EPS or COUNT + EPS
+        int maxCount; // the maximum number of iterations/elements
+        double epsilon; // the desired accuracy
+    };
+
+The class defining termination criteria for iterative algorithms. You can initialize it by default constructor and then override any parameters, or the structure may be fully initialized using the advanced variant of the constructor.
 
 TermCriteria::TermCriteria
 --------------------------
@@ -229,8 +411,6 @@ The constructors.
 
 .. ocv:function:: TermCriteria::TermCriteria(int type, int maxCount, double epsilon)
 
-.. ocv:function:: TermCriteria::TermCriteria(const CvTermCriteria& criteria)
-
     :param type: The type of termination criteria: ``TermCriteria::COUNT``, ``TermCriteria::EPS`` or ``TermCriteria::COUNT`` + ``TermCriteria::EPS``.
 
     :param maxCount: The maximum number of iterations or elements to compute.
@@ -239,11 +419,6 @@ The constructors.
 
     :param criteria: Termination criteria in the deprecated ``CvTermCriteria`` format.
 
-TermCriteria::operator CvTermCriteria
--------------------------------------
-Converts to the deprecated ``CvTermCriteria`` format.
-
-.. ocv:function:: TermCriteria::operator CvTermCriteria() const
 
 Matx
 ----
@@ -330,9 +505,36 @@ Scalar\_
 --------
 .. ocv:class:: Scalar_
 
-Template class for a 4-element vector derived from Vec. ::
+Template class for a 4-element vector derived from Vec.
 
-    template<typename _Tp> class Scalar_ : public Vec<_Tp, 4> { ... };
+::
+
+    template<typename _Tp> class CV_EXPORTS Scalar_ : public Vec<_Tp, 4>
+    {
+    public:
+        //! various constructors
+        Scalar_();
+        Scalar_(_Tp v0, _Tp v1, _Tp v2=0, _Tp v3=0);
+        Scalar_(const CvScalar& s);
+        Scalar_(_Tp v0);
+
+        //! returns a scalar with all elements set to v0
+        static Scalar_<_Tp> all(_Tp v0);
+        //! conversion to the old-style CvScalar
+        operator CvScalar() const;
+
+        //! conversion to another data type
+        template<typename T2> operator Scalar_<T2>() const;
+
+        //! per-element product
+        Scalar_<_Tp> mul(const Scalar_<_Tp>& t, double scale=1 ) const;
+
+        // returns (v0, -v1, -v2, -v3)
+        Scalar_<_Tp> conj() const;
+
+        // returns true iff v1 == v2 == v3 == 0
+        bool isReal() const;
+    };
 
     typedef Scalar_<double> Scalar;
 
@@ -342,12 +544,21 @@ Range
 -----
 .. ocv:class:: Range
 
-Template class specifying a continuous subsequence (slice) of a sequence. ::
+Template class specifying a continuous subsequence (slice) of a sequence.
 
-    class Range
+::
+
+    class CV_EXPORTS Range
     {
     public:
-        ...
+        Range();
+        Range(int _start, int _end);
+        Range(const CvSlice& slice);
+        int size() const;
+        bool empty() const;
+        static Range all();
+        operator CvSlice() const;
+
         int start, end;
     };
 
@@ -368,187 +579,466 @@ The static method ``Range::all()`` returns a special variable that means "the wh
     }
 
 
-.. _Ptr:
+KeyPoint
+--------
+.. ocv:class:: KeyPoint
+
+  Data structure for salient point detectors.
+
+  .. ocv:member:: Point2f pt
+
+     coordinates of the keypoint
+
+  .. ocv:member:: float size
+
+     diameter of the meaningful keypoint neighborhood
+
+  .. ocv:member:: float angle
+
+     computed orientation of the keypoint (-1 if not applicable). Its possible values are in a range [0,360) degrees. It is measured relative to image coordinate system (y-axis is directed downward), ie in clockwise.
+
+  .. ocv:member:: float response
+
+     the response by which the most strong keypoints have been selected. Can be used for further sorting or subsampling
+
+  .. ocv:member:: int octave
+
+     octave (pyramid layer) from which the keypoint has been extracted
+
+  .. ocv:member:: int class_id
+
+     object id that can be used to clustered keypoints by an object they belong to
+
+KeyPoint::KeyPoint
+------------------
+The keypoint constructors
+
+.. ocv:function:: KeyPoint::KeyPoint()
+
+.. ocv:function:: KeyPoint::KeyPoint(Point2f _pt, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1)
+
+.. ocv:function:: KeyPoint::KeyPoint(float x, float y, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1)
+
+.. ocv:pyfunction:: cv2.KeyPoint([x, y, _size[, _angle[, _response[, _octave[, _class_id]]]]]) -> <KeyPoint object>
+
+    :param x: x-coordinate of the keypoint
+
+    :param y: y-coordinate of the keypoint
+
+    :param _pt: x & y coordinates of the keypoint
+
+    :param _size: keypoint diameter
+
+    :param _angle: keypoint orientation
+
+    :param _response: keypoint detector response on the keypoint (that is, strength of the keypoint)
+
+    :param _octave: pyramid octave in which the keypoint has been detected
+
+    :param _class_id: object id
+
+
+KeyPoint::convert
+--------------------
+
+This method converts vector of keypoints to vector of points or the reverse, where each keypoint is assigned the same size and the same orientation.
+
+.. ocv:function:: void KeyPoint::convert(const std::vector<KeyPoint>& keypoints, std::vector<Point2f>& points2f, const std::vector<int>& keypointIndexes=std::vector<int>())
+
+.. ocv:function:: void KeyPoint::convert(const std::vector<Point2f>& points2f, std::vector<KeyPoint>& keypoints, float size=1, float response=1, int octave=0, int class_id=-1)
+
+.. ocv:pyfunction:: cv2.KeyPoint_convert(keypoints[, keypointIndexes]) -> points2f
+
+.. ocv:pyfunction:: cv2.KeyPoint_convert(points2f[, size[, response[, octave[, class_id]]]]) -> keypoints
+
+    :param keypoints: Keypoints obtained from any feature detection algorithm like SIFT/SURF/ORB
+
+    :param points2f: Array of (x,y) coordinates of each keypoint
+
+    :param keypointIndexes: Array of indexes of keypoints to be converted to points. (Acts like a mask to convert only specified keypoints)
+
+    :param _size: keypoint diameter
+
+    :param _response: keypoint detector response on the keypoint (that is, strength of the keypoint)
+
+    :param _octave: pyramid octave in which the keypoint has been detected
+
+    :param _class_id: object id
+
+
+KeyPoint::overlap
+--------------------
+
+This method computes overlap for pair of keypoints. Overlap is the ratio between area of keypoint regions' intersection and area of keypoint regions' union (considering keypoint region as circle). If they don't overlap, we get zero. If they coincide at same location with same size, we get 1.
+
+.. ocv:function:: float KeyPoint::overlap(const KeyPoint& kp1, const KeyPoint& kp2)
+
+.. ocv:pyfunction:: cv2.KeyPoint_overlap(kp1, kp2) -> retval
+
+    :param kp1: First keypoint
+
+    :param kp2: Second keypoint
+
+
+DMatch
+------
+.. ocv:class:: DMatch
+
+Class for matching keypoint descriptors: query descriptor index,
+train descriptor index, train image index, and distance between descriptors. ::
+
+    class DMatch
+    {
+    public:
+        DMatch() : queryIdx(-1), trainIdx(-1), imgIdx(-1),
+                   distance(std::numeric_limits<float>::max()) {}
+        DMatch( int _queryIdx, int _trainIdx, float _distance ) :
+                queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(-1),
+                distance(_distance) {}
+        DMatch( int _queryIdx, int _trainIdx, int _imgIdx, float _distance ) :
+                queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(_imgIdx),
+                distance(_distance) {}
+
+        int queryIdx; // query descriptor index
+        int trainIdx; // train descriptor index
+        int imgIdx;   // train image index
+
+        float distance;
+
+        // less is better
+        bool operator<( const DMatch &m ) const;
+    };
+
 
 Ptr
 ---
 .. ocv:class:: Ptr
 
-Template class for smart reference-counting pointers ::
+Template class for smart pointers with shared ownership. ::
 
-    template<typename _Tp> class Ptr
+    template<typename T>
+    struct Ptr
     {
-    public:
-        // default constructor
+        typedef T element_type;
+
         Ptr();
-        // constructor that wraps the object pointer
-        Ptr(_Tp* _obj);
-        // destructor: calls release()
+
+        template<typename Y>
+        explicit Ptr(Y* p);
+        template<typename Y, typename D>
+        Ptr(Y* p, D d);
+
+        Ptr(const Ptr& o);
+        template<typename Y>
+        Ptr(const Ptr<Y>& o);
+        template<typename Y>
+        Ptr(const Ptr<Y>& o, T* p);
+
         ~Ptr();
-        // copy constructor; increments ptr's reference counter
-        Ptr(const Ptr& ptr);
-        // assignment operator; decrements own reference counter
-        // (with release()) and increments ptr's reference counter
-        Ptr& operator = (const Ptr& ptr);
-        // increments reference counter
-        void addref();
-        // decrements reference counter; when it becomes 0,
-        // delete_obj() is called
+
+        Ptr& operator = (const Ptr& o);
+        template<typename Y>
+        Ptr& operator = (const Ptr<Y>& o);
+
         void release();
-        // user-specified custom object deletion operation.
-        // by default, "delete obj;" is called
-        void delete_obj();
-        // returns true if obj == 0;
+
+        template<typename Y>
+        void reset(Y* p);
+        template<typename Y, typename D>
+        void reset(Y* p, D d);
+
+        void swap(Ptr& o);
+
+        T* get() const;
+
+        T& operator * () const;
+        T* operator -> () const;
+        operator T* () const;
+
         bool empty() const;
 
-        // provide access to the object fields and methods
-        _Tp* operator -> ();
-        const _Tp* operator -> () const;
-
-        // return the underlying object pointer;
-        // thanks to the methods, the Ptr<_Tp> can be
-        // used instead of _Tp*
-        operator _Tp* ();
-        operator const _Tp*() const;
-    protected:
-        // the encapsulated object pointer
-        _Tp* obj;
-        // the associated reference counter
-        int* refcount;
+        template<typename Y>
+        Ptr<Y> staticCast() const;
+        template<typename Y>
+        Ptr<Y> constCast() const;
+        template<typename Y>
+        Ptr<Y> dynamicCast() const;
     };
 
 
-The ``Ptr<_Tp>`` class is a template class that wraps pointers of the corresponding type. It is
-similar to ``shared_ptr`` that is part of the Boost library
-(http://www.boost.org/doc/libs/1_40_0/libs/smart_ptr/shared_ptr.htm) and also part of the
-`C++0x <http://en.wikipedia.org/wiki/C++0x>`_ standard.
+A ``Ptr<T>`` pretends to be a pointer to an object of type T.
+Unlike an ordinary pointer, however, the object will be automatically
+cleaned up once all ``Ptr`` instances pointing to it are destroyed.
 
-This class provides the following options:
+``Ptr`` is similar to ``boost::shared_ptr`` that is part of the Boost library
+(http://www.boost.org/doc/libs/release/libs/smart_ptr/shared_ptr.htm)
+and ``std::shared_ptr`` from the `C++11 <http://en.wikipedia.org/wiki/C++11>`_ standard.
+
+This class provides the following advantages:
 
 *
     Default constructor, copy constructor, and assignment operator for an arbitrary C++ class
-    or a C structure. For some objects, like files, windows, mutexes, sockets, and others, a copy
+    or C structure. For some objects, like files, windows, mutexes, sockets, and others, a copy
     constructor or an assignment operator are difficult to define. For some other objects, like
     complex classifiers in OpenCV, copy constructors are absent and not easy to implement. Finally,
     some of complex OpenCV and your own data structures may be written in C.
-    However, copy constructors and default constructors can simplify programming a lot.Besides,
-    they are often required (for example, by STL containers). By wrapping a pointer to such a
-    complex object ``TObj`` to ``Ptr<TObj>``, you automatically get all of the necessary
+    However, copy constructors and default constructors can simplify programming a lot. Besides,
+    they are often required (for example, by STL containers). By using a ``Ptr`` to such an
+    object instead of the object itself, you automatically get all of the necessary
     constructors and the assignment operator.
 
 *
     *O(1)* complexity of the above-mentioned operations. While some structures, like ``std::vector``,
     provide a copy constructor and an assignment operator, the operations may take a considerable
-    amount of time if the data structures are large. But if the structures are put into ``Ptr<>``,
+    amount of time if the data structures are large. But if the structures are put into a ``Ptr``,
     the overhead is small and independent of the data size.
 
 *
-    Automatic destruction, even for C structures. See the example below with ``FILE*``.
+    Automatic and customizable cleanup, even for C structures. See the example below with ``FILE*``.
 
 *
     Heterogeneous collections of objects. The standard STL and most other C++ and OpenCV containers
     can store only objects of the same type and the same size. The classical solution to store objects
-    of different types in the same container is to store pointers to the base class ``base_class_t*``
-    instead but then you loose the automatic memory management. Again, by using ``Ptr<base_class_t>()``
-    instead of the raw pointers, you can solve the problem.
+    of different types in the same container is to store pointers to the base class (``Base*``)
+    instead but then you lose the automatic memory management. Again, by using ``Ptr<Base>``
+    instead of raw pointers, you can solve the problem.
 
-The ``Ptr`` class treats the wrapped object as a black box. The reference counter is allocated and
-managed separately. The only thing the pointer class needs to know about the object is how to
-deallocate it. This knowledge is encapsulated in the ``Ptr::delete_obj()`` method that is called when
-the reference counter becomes 0. If the object is a C++ class instance, no additional coding is
-needed, because the default implementation of this method calls ``delete obj;``. However, if the
-object is deallocated in a different way, the specialized method should be created. For example,
-if you want to wrap ``FILE``, the ``delete_obj`` may be implemented as follows: ::
+A ``Ptr`` is said to *own* a pointer - that is, for each ``Ptr`` there is a pointer that will be deleted
+once all ``Ptr`` instances that own it are destroyed. The owned pointer may be null, in which case nothing is deleted.
+Each ``Ptr`` also *stores* a pointer. The stored pointer is the pointer the ``Ptr`` pretends to be;
+that is, the one you get when you use :ocv:func:`Ptr::get` or the conversion to ``T*``. It's usually
+the same as the owned pointer, but if you use casts or the general shared-ownership constructor, the two may diverge:
+the ``Ptr`` will still own the original pointer, but will itself point to something else.
 
-    template<> inline void Ptr<FILE>::delete_obj()
-    {
-        fclose(obj); // no need to clear the pointer afterwards,
-                     // it is done externally.
-    }
-    ...
+The owned pointer is treated as a black box. The only thing ``Ptr`` needs to know about it is how to
+delete it. This knowledge is encapsulated in the *deleter* - an auxiliary object that is associated
+with the owned pointer and shared between all ``Ptr`` instances that own it. The default deleter is
+an instance of ``DefaultDeleter``, which uses the standard C++ ``delete`` operator; as such it
+will work with any pointer allocated with the standard ``new`` operator.
 
-    // now use it:
-    Ptr<FILE> f(fopen("myfile.txt", "r"));
-    if(f.empty())
-        throw ...;
+However, if the pointer must be deleted in a different way, you must specify a custom deleter upon
+``Ptr`` construction. A deleter is simply a callable object that accepts the pointer as its sole argument.
+For example, if you want to wrap ``FILE``, you may do so as follows::
+
+    Ptr<FILE> f(fopen("myfile.txt", "w"), fclose);
+    if(!f) throw ...;
     fprintf(f, ....);
     ...
-    // the file will be closed automatically by the Ptr<FILE> destructor.
+    // the file will be closed automatically by f's destructor.
 
+Alternatively, if you want all pointers of a particular type to be deleted the same way,
+you can specialize ``DefaultDeleter<T>::operator()`` for that type, like this::
 
-.. note:: The reference increment/decrement operations are implemented as atomic operations,
-          and therefore it is normally safe to use the classes in multi-threaded applications.
-          The same is true for :ocv:class:`Mat` and other C++ OpenCV classes that operate on
-          the reference counters.
+    namespace cv {
+    template<> void DefaultDeleter<FILE>::operator ()(FILE * obj) const
+    {
+        fclose(obj);
+    }
+    }
 
-Ptr::Ptr
---------
-Various Ptr constructors.
+For convenience, the following types from the OpenCV C API already have such a specialization
+that calls the appropriate release function:
+
+* ``CvCapture``
+* :ocv:struct:`CvDTreeSplit`
+* :ocv:struct:`CvFileStorage`
+* ``CvHaarClassifierCascade``
+* :ocv:struct:`CvMat`
+* :ocv:struct:`CvMatND`
+* :ocv:struct:`CvMemStorage`
+* :ocv:struct:`CvSparseMat`
+* ``CvVideoWriter``
+* :ocv:struct:`IplImage`
+
+.. note:: The shared ownership mechanism is implemented with reference counting. As such,
+          cyclic ownership (e.g. when object ``a`` contains a ``Ptr`` to object ``b``, which
+          contains a ``Ptr`` to object ``a``) will lead to all involved objects never being
+          cleaned up. Avoid such situations.
+
+.. note:: It is safe to concurrently read (but not write) a ``Ptr`` instance from multiple threads
+          and therefore it is normally safe to use it in multi-threaded applications.
+          The same is true for :ocv:class:`Mat` and other C++ OpenCV classes that use internal
+          reference counts.
+
+Ptr::Ptr (null)
+------------------
 
 .. ocv:function:: Ptr::Ptr()
-.. ocv:function:: Ptr::Ptr(_Tp* _obj)
-.. ocv:function:: Ptr::Ptr(const Ptr& ptr)
+
+    The default constructor creates a null ``Ptr`` - one that owns and stores a null pointer.
+
+Ptr::Ptr (assuming ownership)
+-----------------------------
+
+.. ocv:function:: template<typename Y> Ptr::Ptr(Y* p)
+.. ocv:function:: template<typename Y, typename D> Ptr::Ptr(Y* p, D d)
+
+    :param d: Deleter to use for the owned pointer.
+    :param p: Pointer to own.
+
+    If ``p`` is null, these are equivalent to the default constructor.
+
+    Otherwise, these constructors assume ownership of ``p`` - that is, the created ``Ptr`` owns
+    and stores ``p`` and assumes it is the sole owner of it. Don't use them if ``p`` is already
+    owned by another ``Ptr``, or else ``p`` will get deleted twice.
+
+    With the first constructor, ``DefaultDeleter<Y>()`` becomes the associated deleter (so ``p``
+    will eventually be deleted with the standard ``delete`` operator). ``Y`` must be a complete
+    type at the point of invocation.
+
+    With the second constructor, ``d`` becomes the associated deleter.
+
+    ``Y*`` must be convertible to ``T*``.
+
+    .. note:: It is often easier to use :ocv:func:`makePtr` instead.
+
+Ptr::Ptr (sharing ownership)
+----------------------------
+
+.. ocv:function:: Ptr::Ptr(const Ptr& o)
+.. ocv:function:: template<typename Y> Ptr::Ptr(const Ptr<Y>& o)
+.. ocv:function:: template<typename Y> Ptr::Ptr(const Ptr<Y>& o, T* p)
+
+    :param o: ``Ptr`` to share ownership with.
+    :param p: Pointer to store.
+
+    These constructors create a ``Ptr`` that shares ownership with another ``Ptr`` - that is,
+    own the same pointer as ``o``.
+
+    With the first two, the same pointer is stored, as well; for the second, ``Y*`` must be convertible to ``T*``.
+
+    With the third, ``p`` is stored, and ``Y`` may be any type. This constructor allows to have completely
+    unrelated owned and stored pointers, and should be used with care to avoid confusion. A relatively
+    benign use is to create a non-owning ``Ptr``, like this::
+
+        ptr = Ptr<T>(Ptr<T>(), dont_delete_me); // owns nothing; will not delete the pointer.
 
 Ptr::~Ptr
 ---------
-The Ptr destructor.
 
 .. ocv:function:: Ptr::~Ptr()
 
+    The destructor is equivalent to calling :ocv:func:`Ptr::release`.
+
 Ptr::operator =
 ----------------
-Assignment operator.
 
-.. ocv:function:: Ptr& Ptr::operator = (const Ptr& ptr)
+.. ocv:function:: Ptr& Ptr::operator = (const Ptr& o)
+.. ocv:function:: template<typename Y> Ptr& Ptr::operator = (const Ptr<Y>& o)
 
-Decrements own reference counter (with ``release()``) and increments ptr's reference counter.
+    :param o: ``Ptr`` to share ownership with.
 
-Ptr::addref
------------
-Increments reference counter.
+    Assignment replaces the current ``Ptr`` instance with one that owns and stores same
+    pointers as ``o`` and then destroys the old instance.
 
-.. ocv:function:: void Ptr::addref()
 
 Ptr::release
 ------------
-Decrements reference counter; when it becomes 0, ``delete_obj()`` is called.
 
 .. ocv:function:: void Ptr::release()
 
-Ptr::delete_obj
----------------
-User-specified custom object deletion operation. By default, ``delete obj;`` is called.
+    If no other ``Ptr`` instance owns the owned pointer, deletes it with the associated deleter.
+    Then sets both the owned and the stored pointers to ``NULL``.
 
-.. ocv:function:: void Ptr::delete_obj()
+
+Ptr::reset
+----------
+
+.. ocv:function:: template<typename Y> void Ptr::reset(Y* p)
+.. ocv:function:: template<typename Y, typename D> void Ptr::reset(Y* p, D d)
+
+    :param d: Deleter to use for the owned pointer.
+    :param p: Pointer to own.
+
+    ``ptr.reset(...)`` is equivalent to ``ptr = Ptr<T>(...)``.
+
+Ptr::swap
+---------
+
+.. ocv:function:: void Ptr::swap(Ptr& o)
+
+    :param o: ``Ptr`` to swap with.
+
+    Swaps the owned and stored pointers (and deleters, if any) of this and ``o``.
+
+Ptr::get
+--------
+
+.. ocv:function:: T* Ptr::get() const
+
+    Returns the stored pointer.
+
+Ptr pointer emulation
+---------------------
+
+.. ocv:function:: T& Ptr::operator * () const
+.. ocv:function:: T* Ptr::operator -> () const
+.. ocv:function:: Ptr::operator T* () const
+
+    These operators are what allows ``Ptr`` to pretend to be a pointer.
+
+    If ``ptr`` is a ``Ptr<T>``, then ``*ptr`` is equivalent to ``*ptr.get()``
+    and ``ptr->foo`` is equivalent to ``ptr.get()->foo``. In addition, ``ptr``
+    is implicitly convertible to ``T*``, and such conversion is equivalent to
+    ``ptr.get()``. As a corollary, ``if (ptr)`` is equivalent to ``if (ptr.get())``.
+    In other words, a ``Ptr`` behaves as if it was its own stored pointer.
 
 Ptr::empty
 ----------
-Returns true if obj == 0;
 
-bool empty() const;
+.. ocv:function:: bool Ptr::empty() const
 
-Ptr::operator ->
-----------------
-Provide access to the object fields and methods.
+    ``ptr.empty()`` is equivalent to ``!ptr.get()``.
 
- .. ocv:function:: template<typename _Tp> _Tp* Ptr::operator -> ()
- .. ocv:function:: template<typename _Tp> const _Tp* Ptr::operator -> () const
+Ptr casts
+---------
 
+.. ocv:function:: template<typename Y> Ptr<Y> Ptr::staticCast() const
+.. ocv:function:: template<typename Y> Ptr<Y> Ptr::constCast() const
+.. ocv:function:: template<typename Y> Ptr<Y> Ptr::dynamicCast() const
 
-Ptr::operator _Tp*
-------------------
-Returns the underlying object pointer. Thanks to the methods, the ``Ptr<_Tp>`` can be used instead
-of ``_Tp*``.
+    If ``ptr`` is a ``Ptr``, then ``ptr.fooCast<Y>()`` is equivalent to
+    ``Ptr<Y>(ptr, foo_cast<Y>(ptr.get()))``. That is, these functions create
+    a new ``Ptr`` with the same owned pointer and a cast stored pointer.
 
- .. ocv:function:: template<typename _Tp> Ptr::operator _Tp* ()
- .. ocv:function:: template<typename _Tp> Ptr::operator const _Tp*() const
+Ptr global swap
+---------------
 
+.. ocv:function:: template<typename T> void swap(Ptr<T>& ptr1, Ptr<T>& ptr2)
+
+    Equivalent to ``ptr1.swap(ptr2)``. Provided to help write generic algorithms.
+
+Ptr comparisons
+---------------
+
+.. ocv:function:: template<typename T> bool operator == (const Ptr<T>& ptr1, const Ptr<T>& ptr2)
+.. ocv:function:: template<typename T> bool operator != (const Ptr<T>& ptr1, const Ptr<T>& ptr2)
+
+    Return whether ``ptr1.get()`` and ``ptr2.get()`` are equal and not equal, respectively.
+
+makePtr
+-------
+
+.. ocv:function:: template<typename T> Ptr<T> makePtr()
+.. ocv:function:: template<typename T, typename A1> Ptr<T> makePtr(const A1& a1)
+.. ocv:function:: template<typename T, typename A1, typename A2> Ptr<T> makePtr(const A1& a1, const A2& a2)
+.. ocv:function:: template<typename T, typename A1, typename A2, typename A3> Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3)
+
+    (and so on...)
+
+    ``makePtr<T>(...)`` is equivalent to ``Ptr<T>(new T(...))``. It is shorter than the latter, and
+    it's marginally safer than using a constructor or :ocv:func:`Ptr::reset`, since it ensures that
+    the owned pointer is new and thus not owned by any other ``Ptr`` instance.
+
+    Unfortunately, perfect forwarding is impossible to implement in C++03, and so ``makePtr`` is limited
+    to constructors of ``T`` that have up to 10 arguments, none of which are non-const references.
 
 Mat
 ---
 .. ocv:class:: Mat
 
-OpenCV C++ n-dimensional dense array class ::
+OpenCV C++ n-dimensional dense array class
+::
 
     class CV_EXPORTS Mat
     {
@@ -578,7 +1068,6 @@ OpenCV C++ n-dimensional dense array class ::
         ...
     };
 
-
 The class ``Mat`` represents an n-dimensional dense numerical single-channel or multi-channel array. It can be used to store real or complex-valued vectors and matrices, grayscale or color images, voxel volumes, vector fields, point clouds, tensors, histograms (though, very high-dimensional histograms may be better stored in a ``SparseMat`` ). The data layout of the array
 :math:`M` is defined by the array ``M.step[]``, so that the address of element
 :math:`(i_0,...,i_{M.dims-1})`, where
@@ -798,6 +1287,9 @@ Finally, there are STL-style iterators that are smart enough to skip gaps betwee
 
 The matrix iterators are random-access iterators, so they can be passed to any STL algorithm, including ``std::sort()`` .
 
+.. note::
+
+   * An example demonstrating the serial out capabilities of cv::Mat can be found at opencv_source_code/samples/cpp/cout_mat.cpp
 
 .. _MatrixExpressions:
 
@@ -911,10 +1403,6 @@ Various Mat constructors
 
 .. ocv:function:: Mat::Mat(const Mat& m, const Rect& roi)
 
-.. ocv:function:: Mat::Mat(const CvMat* m, bool copyData=false)
-
-.. ocv:function:: Mat::Mat(const IplImage* img, bool copyData=false)
-
 .. ocv:function:: template<typename T, int n> explicit Mat::Mat(const Vec<T, n>& vec, bool copyData=true)
 
 .. ocv:function:: template<typename T, int m, int n> explicit Mat::Mat(const Matx<T, m, n>& vec, bool copyData=true)
@@ -1064,7 +1552,7 @@ The method makes a new header for the specified row span of the matrix. Similarl
 
 Mat::colRange
 -------------
-Creates a matrix header for the specified row span.
+Creates a matrix header for the specified column span.
 
 .. ocv:function:: Mat Mat::colRange(int startcol, int endcol) const
 
@@ -1398,7 +1886,7 @@ Such a scheme makes the memory management robust and efficient at the same time
     Mat color;
     ...
     Mat gray(color.rows, color.cols, color.depth());
-    cvtColor(color, gray, CV_BGR2GRAY);
+    cvtColor(color, gray, COLOR_BGR2GRAY);
 
 
 you can simply write: ::
@@ -1406,7 +1894,7 @@ you can simply write: ::
     Mat color;
     ...
     Mat gray;
-    cvtColor(color, gray, CV_BGR2GRAY);
+    cvtColor(color, gray, COLOR_BGR2GRAY);
 
 
 because ``cvtColor`` , as well as the most of OpenCV functions, calls ``Mat::create()`` for the output array internally.
@@ -1465,6 +1953,7 @@ Adds elements to the bottom of the matrix.
 .. ocv:function:: void Mat::push_back( const Mat& m )
 
     :param elem: Added element(s).
+    :param m: Added line(s).
 
 The methods add one or more elements to the bottom of the matrix. They emulate the corresponding method of the STL vector class. When ``elem`` is ``Mat`` , its type and the number of columns must be the same as in the container matrix.
 
@@ -1552,33 +2041,6 @@ The operators make a new header for the specified sub-array of ``*this`` . They
 :ocv:func:`Mat::colRange` . For example, ``A(Range(0, 10), Range::all())`` is equivalent to ``A.rowRange(0, 10)`` . Similarly to all of the above, the operators are O(1) operations, that is, no matrix data is copied.
 
 
-Mat::operator CvMat
--------------------
-Creates the ``CvMat`` header for the matrix.
-
-.. ocv:function:: Mat::operator CvMat() const
-
-
-The operator creates the ``CvMat`` header for the matrix without copying the underlying data. The reference counter is not taken into account by this operation. Thus, you should make sure than the original matrix is not deallocated while the ``CvMat`` header is used. The operator is useful for intermixing the new and the old OpenCV API's, for example: ::
-
-    Mat img(Size(320, 240), CV_8UC3);
-    ...
-
-    CvMat cvimg = img;
-    mycvOldFunc( &cvimg, ...);
-
-
-where ``mycvOldFunc`` is a function written to work with OpenCV 1.x data structures.
-
-
-Mat::operator IplImage
-----------------------
-Creates the ``IplImage`` header for the matrix.
-
-.. ocv:function:: Mat::operator IplImage() const
-
-The operator creates the ``IplImage`` header for the matrix without copying the underlying data. You should make sure than the original matrix is not deallocated while the ``IplImage`` header is used. Similarly to ``Mat::operator CvMat`` , the operator is useful for intermixing the new and the old OpenCV API's.
-
 Mat::total
 ----------
 Returns the total number of array elements.
@@ -1691,7 +2153,7 @@ Returns the depth of a matrix element.
 
 .. ocv:function:: int Mat::depth() const
 
-The method returns the identifier of the matrix element depth (the type of each individual channel). For example, for a 16-bit signed 3-channel array, the method returns ``CV_16S`` . A complete list of matrix types contains the following values:
+The method returns the identifier of the matrix element depth (the type of each individual channel). For example, for a 16-bit signed element array, the method returns ``CV_16S`` . A complete list of matrix types contains the following values:
 
 * ``CV_8U``     - 8-bit unsigned integers ( ``0..255``     )
 
@@ -2153,14 +2615,12 @@ Various SparseMat constructors.
 .. ocv:function:: SparseMat::SparseMat( int dims, const int* _sizes, int _type )
 .. ocv:function:: SparseMat::SparseMat( const SparseMat& m )
 .. ocv:function:: SparseMat::SparseMat( const Mat& m )
-.. ocv:function:: SparseMat::SparseMat( const CvSparseMat* m )
 
 
     :param m: Source matrix for copy constructor. If m is dense matrix (ocv:class:`Mat`) then it will be converted to sparse representation.
     :param dims: Array dimensionality.
     :param _sizes: Sparce matrix size on all dementions.
     :param _type: Sparse matrix data type.
-    :param try1d: if try1d is true and matrix is a single-column matrix (Nx1), then the sparse matrix will be 1-dimensional.
 
 SparseMat::~SparseMat
 ---------------------
@@ -2175,6 +2635,8 @@ Provides sparse matrix assignment operators.
 .. ocv:function:: SparseMat& SparseMat::operator = (const SparseMat& m)
 .. ocv:function:: SparseMat& SparseMat::operator = (const Mat& m)
 
+    :param m: Matrix for assignment.
+
 The last variant is equivalent to the corresponding constructor with try1d=false.
 
 
@@ -2202,6 +2664,10 @@ Convert sparse matrix with possible type change and scaling.
 .. ocv:function:: void SparseMat::convertTo( SparseMat& m, int rtype, double alpha=1 ) const
 .. ocv:function:: void SparseMat::convertTo( Mat& m, int rtype, double alpha=1, double beta=0 ) const
 
+    :param m: Destination matrix.
+    :param rtype: Destination matrix type.
+    :param alpha: Conversion multiplier.
+
 The first version converts arbitrary sparse matrix to dense matrix and multiplies all the matrix elements by the specified scalar.
 The second versiob converts sparse matrix to dense matrix with optional type conversion and scaling.
 When rtype=-1, the destination element type will be the same as the sparse matrix element type.
@@ -2294,7 +2760,7 @@ The method returns the number of matrix channels.
 
 SparseMat::size
 ---------------
-Returns the array of sizes or matrix size by i dimention and 0 if the matrix is not allocated.
+Returns the array of sizes or matrix size by i dimension and 0 if the matrix is not allocated.
 
 .. ocv:function:: const int* SparseMat::size() const
 .. ocv:function:: int SparseMat::size(int i) const
@@ -2322,6 +2788,11 @@ Compute element hash value from the element indices.
 .. ocv:function:: size_t SparseMat::hash(int i0, int i1, int i2) const
 .. ocv:function:: size_t SparseMat::hash(const int* idx) const
 
+    :param i0: The first dimension index.
+    :param i1: The second dimension index.
+    :param i2: The third dimension index.
+    :param idx: Array of element indices for multidimensional matices.
+
 SparseMat::ptr
 --------------
 Low-level element-access functions, special variants for 1D, 2D, 3D cases, and the generic one for n-D case.
@@ -2331,6 +2802,12 @@ Low-level element-access functions, special variants for 1D, 2D, 3D cases, and t
 .. ocv:function:: uchar* SparseMat::ptr(int i0, int i1, int i2, bool createMissing, size_t* hashval=0)
 .. ocv:function:: uchar* SparseMat::ptr(const int* idx, bool createMissing, size_t* hashval=0)
 
+    :param i0: The first dimension index.
+    :param i1: The second dimension index.
+    :param i2: The third dimension index.
+    :param idx: Array of element indices for multidimensional matices.
+    :param createMissing: Create new element with 0 value if it does not exist in SparseMat.
+
 Return pointer to the matrix element. If the element is there (it is non-zero), the pointer to it is returned.
 If it is not there and ``createMissing=false``, NULL pointer is returned. If it is not there and ``createMissing=true``,
 the new elementis created and initialized with 0. Pointer to it is returned. If the optional hashval pointer is not ``NULL``,
@@ -2344,6 +2821,11 @@ Erase the specified matrix element. When there is no such an element, the method
 .. ocv:function:: void SparseMat::erase(int i0, int i1, int i2, size_t* hashval=0)
 .. ocv:function:: void SparseMat::erase(const int* idx, size_t* hashval=0)
 
+    :param i0: The first dimension index.
+    :param i1: The second dimension index.
+    :param i2: The third dimension index.
+    :param idx: Array of element indices for multidimensional matices.
+
 SparseMat\_
 -----------
 .. ocv:class:: SparseMat_
@@ -2412,6 +2894,82 @@ Algorithm
 ---------
 .. ocv:class:: Algorithm
 
+::
+
+    class CV_EXPORTS_W Algorithm
+    {
+    public:
+        Algorithm();
+        virtual ~Algorithm();
+        string name() const;
+
+        template<typename _Tp> typename ParamType<_Tp>::member_type get(const string& name) const;
+        template<typename _Tp> typename ParamType<_Tp>::member_type get(const char* name) const;
+
+        CV_WRAP int getInt(const string& name) const;
+        CV_WRAP double getDouble(const string& name) const;
+        CV_WRAP bool getBool(const string& name) const;
+        CV_WRAP string getString(const string& name) const;
+        CV_WRAP Mat getMat(const string& name) const;
+        CV_WRAP vector<Mat> getMatVector(const string& name) const;
+        CV_WRAP Ptr<Algorithm> getAlgorithm(const string& name) const;
+
+        void set(const string& name, int value);
+        void set(const string& name, double value);
+        void set(const string& name, bool value);
+        void set(const string& name, const string& value);
+        void set(const string& name, const Mat& value);
+        void set(const string& name, const vector<Mat>& value);
+        void set(const string& name, const Ptr<Algorithm>& value);
+        template<typename _Tp> void set(const string& name, const Ptr<_Tp>& value);
+
+        CV_WRAP void setInt(const string& name, int value);
+        CV_WRAP void setDouble(const string& name, double value);
+        CV_WRAP void setBool(const string& name, bool value);
+        CV_WRAP void setString(const string& name, const string& value);
+        CV_WRAP void setMat(const string& name, const Mat& value);
+        CV_WRAP void setMatVector(const string& name, const vector<Mat>& value);
+        CV_WRAP void setAlgorithm(const string& name, const Ptr<Algorithm>& value);
+        template<typename _Tp> void setAlgorithm(const string& name, const Ptr<_Tp>& value);
+
+        void set(const char* name, int value);
+        void set(const char* name, double value);
+        void set(const char* name, bool value);
+        void set(const char* name, const string& value);
+        void set(const char* name, const Mat& value);
+        void set(const char* name, const vector<Mat>& value);
+        void set(const char* name, const Ptr<Algorithm>& value);
+        template<typename _Tp> void set(const char* name, const Ptr<_Tp>& value);
+
+        void setInt(const char* name, int value);
+        void setDouble(const char* name, double value);
+        void setBool(const char* name, bool value);
+        void setString(const char* name, const string& value);
+        void setMat(const char* name, const Mat& value);
+        void setMatVector(const char* name, const vector<Mat>& value);
+        void setAlgorithm(const char* name, const Ptr<Algorithm>& value);
+        template<typename _Tp> void setAlgorithm(const char* name, const Ptr<_Tp>& value);
+
+        CV_WRAP string paramHelp(const string& name) const;
+        int paramType(const char* name) const;
+        CV_WRAP int paramType(const string& name) const;
+        CV_WRAP void getParams(CV_OUT vector<string>& names) const;
+
+
+        virtual void write(FileStorage& fs) const;
+        virtual void read(const FileNode& fn);
+
+        typedef Algorithm* (*Constructor)(void);
+        typedef int (Algorithm::*Getter)() const;
+        typedef void (Algorithm::*Setter)(int);
+
+        CV_WRAP static void getList(CV_OUT vector<string>& algorithms);
+        CV_WRAP static Ptr<Algorithm> _create(const string& name);
+        template<typename _Tp> static Ptr<_Tp> create(const string& name);
+
+        virtual AlgorithmInfo* info() const /* TODO: make it = 0;*/ { return 0; }
+    };
+
 This is a base class for all more or less complex algorithms in OpenCV, especially for classes of algorithms, for which there can be multiple implementations. The examples are stereo correspondence (for which there are algorithms like block matching, semi-global block matching, graph-cut etc.), background subtraction (which can be done using mixture-of-gaussians models, codebook-based algorithm etc.), optical flow (block matching, Lucas-Kanade, Horn-Schunck etc.).
 
 The class provides the following features for all derived classes:
@@ -2457,13 +3015,13 @@ Algorithm::name
 ---------------
 Returns the algorithm name
 
-.. ocv:function:: string Algorithm::name() const
+.. ocv:function:: String Algorithm::name() const
 
 Algorithm::get
 --------------
 Returns the algorithm parameter
 
-.. ocv:function:: template<typename _Tp> typename ParamType<_Tp>::member_type Algorithm::get(const string& name) const
+.. ocv:function:: template<typename _Tp> typename ParamType<_Tp>::member_type Algorithm::get(const String& name) const
 
     :param name: The parameter name.
 
@@ -2472,7 +3030,7 @@ The method returns value of the particular parameter. Since the compiler can not
     * myalgo.get<int>("param_name")
     * myalgo.get<double>("param_name")
     * myalgo.get<bool>("param_name")
-    * myalgo.get<string>("param_name")
+    * myalgo.get<String>("param_name")
     * myalgo.get<Mat>("param_name")
     * myalgo.get<vector<Mat> >("param_name")
     * myalgo.get<Algorithm>("param_name") (it returns Ptr<Algorithm>).
@@ -2484,13 +3042,13 @@ Algorithm::set
 --------------
 Sets the algorithm parameter
 
-.. ocv:function:: void Algorithm::set(const string& name, int value)
-.. ocv:function:: void Algorithm::set(const string& name, double value)
-.. ocv:function:: void Algorithm::set(const string& name, bool value)
-.. ocv:function:: void Algorithm::set(const string& name, const string& value)
-.. ocv:function:: void Algorithm::set(const string& name, const Mat& value)
-.. ocv:function:: void Algorithm::set(const string& name, const vector<Mat>& value)
-.. ocv:function:: void Algorithm::set(const string& name, const Ptr<Algorithm>& value)
+.. ocv:function:: void Algorithm::set(const String& name, int value)
+.. ocv:function:: void Algorithm::set(const String& name, double value)
+.. ocv:function:: void Algorithm::set(const String& name, bool value)
+.. ocv:function:: void Algorithm::set(const String& name, const String& value)
+.. ocv:function:: void Algorithm::set(const String& name, const Mat& value)
+.. ocv:function:: void Algorithm::set(const String& name, const vector<Mat>& value)
+.. ocv:function:: void Algorithm::set(const String& name, const Ptr<Algorithm>& value)
 
     :param name: The parameter name.
     :param value: The parameter value.
@@ -2529,13 +3087,13 @@ Algorithm::getList
 ------------------
 Returns the list of registered algorithms
 
-.. ocv:function:: void Algorithm::getList(vector<string>& algorithms)
+.. ocv:function:: void Algorithm::getList(vector<String>& algorithms)
 
     :param algorithms: The output vector of algorithm names.
 
 This static method returns the list of registered algorithms in alphabetical order. Here is how to use it ::
 
-    vector<string> algorithms;
+    vector<String> algorithms;
     Algorithm::getList(algorithms);
     cout << "Algorithms: " << algorithms.size() << endl;
     for (size_t i=0; i < algorithms.size(); i++)
@@ -2546,11 +3104,11 @@ Algorithm::create
 -----------------
 Creates algorithm instance by name
 
-.. ocv:function:: template<typename _Tp> Ptr<_Tp> Algorithm::create(const string& name)
+.. ocv:function:: template<typename _Tp> Ptr<_Tp> Algorithm::create(const String& name)
 
     :param name: The algorithm name, one of the names returned by ``Algorithm::getList()``.
 
-This static method creates a new instance of the specified algorithm. If there is no such algorithm, the method will silently return null pointer (that can be checked by ``Ptr::empty()`` method). Also, you should specify the particular ``Algorithm`` subclass as ``_Tp`` (or simply ``Algorithm`` if you do not know it at that point). ::
+This static method creates a new instance of the specified algorithm. If there is no such algorithm, the method will silently return a null pointer. Also, you should specify the particular ``Algorithm`` subclass as ``_Tp`` (or simply ``Algorithm`` if you do not know it at that point). ::
 
     Ptr<BackgroundSubtractor> bgfg = Algorithm::create<BackgroundSubtractor>("BackgroundSubtractor.MOG2");
 
@@ -2566,4 +3124,3 @@ The above methods are usually enough for users. If you want to make your own alg
  * Add public virtual method ``AlgorithmInfo* info() const;`` to your class.
  * Add constructor function, ``AlgorithmInfo`` instance and implement the ``info()`` method. The simplest way is to take  http://code.opencv.org/projects/opencv/repository/revisions/master/entry/modules/ml/src/ml_init.cpp as the reference and modify it according to the list of your parameters.
  * Add some public function (e.g. ``initModule_<mymodule>()``) that calls info() of your algorithm and put it into the same source file as ``info()`` implementation. This is to force C++ linker to include this object file into the target application. See ``Algorithm::create()`` for details.
-
diff --git a/modules/core/doc/clustering.rst b/modules/core/doc/clustering.rst
index 60fb189285..dc9b28799d 100644
--- a/modules/core/doc/clustering.rst
+++ b/modules/core/doc/clustering.rst
@@ -13,16 +13,20 @@ Finds centers of clusters and groups input samples around the clusters.
 
 .. ocv:cfunction:: int cvKMeans2( const CvArr* samples, int cluster_count, CvArr* labels, CvTermCriteria termcrit, int attempts=1, CvRNG* rng=0, int flags=0, CvArr* _centers=0, double* compactness=0 )
 
-.. ocv:pyoldfunction:: cv.KMeans2(samples, nclusters, labels, termcrit, attempts=1, flags=0, centers=None) -> float
-
     :param samples: Floating-point matrix of input samples, one row per sample.
 
+    :param data: Data for clustering.
+
     :param cluster_count: Number of clusters to split the set by.
 
+    :param K: Number of clusters to split the set by.
+
     :param labels: Input/output integer array that stores the cluster indices for every sample.
 
     :param criteria: The algorithm termination criteria, that is, the maximum number of iterations and/or the desired accuracy. The accuracy is specified as ``criteria.epsilon``. As soon as each of the cluster centers moves by less than ``criteria.epsilon`` on some iteration, the algorithm stops.
 
+    :param termcrit: The algorithm termination criteria, that is, the maximum number of iterations and/or the desired accuracy.
+
     :param attempts: Flag to specify the number of times the algorithm is executed using different initial labellings. The algorithm returns the labels that yield the best compactness (see the last function parameter).
 
     :param rng: CvRNG state initialized by RNG().
@@ -37,6 +41,8 @@ Finds centers of clusters and groups input samples around the clusters.
 
     :param centers: Output matrix of the cluster centers, one row per each cluster center.
 
+    :param _centers: Output matrix of the cluster centers, one row per each cluster center.
+
     :param compactness: The returned value that is described below.
 
 The function ``kmeans`` implements a k-means algorithm that finds the
@@ -58,6 +64,12 @@ Basically, you can use only the core of the function, set the number of
 attempts to 1, initialize labels each time using a custom algorithm, pass them with the
 ( ``flags`` = ``KMEANS_USE_INITIAL_LABELS`` ) flag, and then choose the best (most-compact) clustering.
 
+.. note::
+
+   * An example on K-means clustering can be found at opencv_source_code/samples/cpp/kmeans.cpp
+
+   * (Python) An example on K-means clustering can be found at opencv_source_code/samples/python2/kmeans.py
+
 partition
 -------------
 Splits an element set into equivalency classes.
diff --git a/modules/core/doc/command_line_parser.rst b/modules/core/doc/command_line_parser.rst
index 53396b23bc..c1f69f4e50 100644
--- a/modules/core/doc/command_line_parser.rst
+++ b/modules/core/doc/command_line_parser.rst
@@ -10,13 +10,13 @@ CommandLineParser
 The CommandLineParser class is designed for command line arguments parsing
 
 
-    .. ocv:function:: CommandLineParser::CommandLineParser( int argc, const char* const argv[], const string& keys )
+    .. ocv:function:: CommandLineParser::CommandLineParser( int argc, const char* const argv[], const String& keys )
 
         :param argc:
         :param argv:
         :param keys:
 
-    .. ocv:function:: template<typename T> T CommandLineParser::get<T>(const std::string& name, bool space_delete = true)
+    .. ocv:function:: template<typename T> T CommandLineParser::get<T>(const String& name, bool space_delete = true)
 
         :param name:
         :param space_delete:
@@ -26,14 +26,14 @@ The CommandLineParser class is designed for command line arguments parsing
         :param index:
         :param space_delete:
 
-    .. ocv:function:: bool CommandLineParser::has(const std::string& name)
+    .. ocv:function:: bool CommandLineParser::has(const String& name)
 
         :param name:
 
     .. ocv:function:: bool CommandLineParser::check()
 
 
-    .. ocv:function:: void CommandLineParser::about( const string& message )
+    .. ocv:function:: void CommandLineParser::about( const String& message )
 
         :param message:
 
@@ -41,7 +41,7 @@ The CommandLineParser class is designed for command line arguments parsing
 
     .. ocv:function:: void CommandLineParser::printErrors()
 
-    .. ocv:function:: std::string CommandLineParser::getPathToApplication()
+    .. ocv:function:: String CommandLineParser::getPathToApplication()
 
 
 The sample below demonstrates how to use CommandLineParser:
@@ -59,12 +59,12 @@ The sample below demonstrates how to use CommandLineParser:
 
     int N = parser.get<int>("N");
     double fps = parser.get<double>("fps");
-    std::string path = parser.get<std::string>("path");
+    String path = parser.get<String>("path");
 
     use_time_stamp = parser.has("timestamp");
 
-    std::string img1 = parser.get<string>(0);
-    std::string img2 = parser.get<string>(1);
+    String img1 = parser.get<String>(0);
+    String img2 = parser.get<String>(1);
 
     int repeat = parser.get<int>(2);
 
@@ -78,7 +78,7 @@ Syntax:
 
 ::
 
-    const std::string keys =
+    const String keys =
         "{help h usage ? |      | print this message   }"
         "{@image1        |      | image1 for compare   }"
         "{@image2        |      | image2 for compare   }"
@@ -98,4 +98,3 @@ Use:
     # ./app -fps=aaa
     ERRORS:
     Exception: can not convert: [aaa] to [double]
-
diff --git a/modules/core/doc/core.rst b/modules/core/doc/core.rst
index 90caeebc79..73bfc35367 100644
--- a/modules/core/doc/core.rst
+++ b/modules/core/doc/core.rst
@@ -15,4 +15,4 @@ core. The Core Functionality
     old_xml_yaml_persistence
     clustering
     utility_and_system_functions_and_macros
-
+    opengl_interop
diff --git a/modules/core/doc/drawing_functions.rst b/modules/core/doc/drawing_functions.rst
index 6c0ac4b35f..258cfe6ef3 100644
--- a/modules/core/doc/drawing_functions.rst
+++ b/modules/core/doc/drawing_functions.rst
@@ -6,7 +6,7 @@ Drawing Functions
 Drawing functions work with matrices/images of arbitrary depth.
 The boundaries of the shapes can be rendered with antialiasing (implemented only for 8-bit images for now).
 All the functions include the parameter ``color`` that uses an RGB value (that may be constructed
-with ``CV_RGB`` or the  :ocv:class:`Scalar_`  constructor
+with the ``Scalar``  constructor
 ) for color
 images and brightness for grayscale images. For color images, the channel ordering
 is normally *Blue, Green, Red*.
@@ -26,18 +26,20 @@ If a drawn figure is partially or completely outside the image, the drawing func
 
 .. note:: The functions do not support alpha-transparency when the target image is 4-channel. In this case, the ``color[3]`` is simply copied to the repainted pixels. Thus, if you want to paint semi-transparent shapes, you can paint them in a separate buffer and then blend it with the main image.
 
+.. note::
+
+   * An example on using variate drawing functions like line, rectangle, ... can be found at opencv_source_code/samples/cpp/drawing.cpp
+
 circle
 ----------
 Draws a circle.
 
-.. ocv:function:: void circle(Mat& img, Point center, int radius, const Scalar& color, int thickness=1, int lineType=8, int shift=0)
+.. ocv:function:: void circle( Mat& img, Point center, int radius, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
 .. ocv:pyfunction:: cv2.circle(img, center, radius, color[, thickness[, lineType[, shift]]]) -> img
 
 .. ocv:cfunction:: void cvCircle( CvArr* img, CvPoint center, int radius, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.Circle(img, center, radius, color, thickness=1, lineType=8, shift=0)-> None
-
     :param img: Image where the circle is drawn.
 
     :param center: Center of the circle.
@@ -66,8 +68,6 @@ Clips the line against the image rectangle.
 
 .. ocv:cfunction:: int cvClipLine( CvSize img_size, CvPoint* pt1, CvPoint* pt2 )
 
-.. ocv:pyoldfunction:: cv.ClipLine(imgSize, pt1, pt2) -> (point1, point2)
-
     :param imgSize: Image size. The image rectangle is  ``Rect(0, 0, imgSize.width, imgSize.height)`` .
 
     :param imgRect: Image rectangle.
@@ -83,9 +83,9 @@ ellipse
 -----------
 Draws a simple or thick elliptic arc or fills an ellipse sector.
 
-.. ocv:function:: void ellipse(Mat& img, Point center, Size axes, double angle, double startAngle, double endAngle, const Scalar& color, int thickness=1, int lineType=8, int shift=0)
+.. ocv:function:: void ellipse( Mat& img, Point center, Size axes, double angle, double startAngle, double endAngle, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
-.. ocv:function:: void ellipse(Mat& img, const RotatedRect& box, const Scalar& color, int thickness=1, int lineType=8)
+.. ocv:function:: void ellipse( Mat& img, const RotatedRect& box, const Scalar& color, int thickness=1, int lineType=LINE_8 )
 
 .. ocv:pyfunction:: cv2.ellipse(img, center, axes, angle, startAngle, endAngle, color[, thickness[, lineType[, shift]]]) -> img
 
@@ -93,12 +93,8 @@ Draws a simple or thick elliptic arc or fills an ellipse sector.
 
 .. ocv:cfunction:: void cvEllipse( CvArr* img, CvPoint center, CvSize axes, double angle, double start_angle, double end_angle, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.Ellipse(img, center, axes, angle, start_angle, end_angle, color, thickness=1, lineType=8, shift=0)-> None
-
 .. ocv:cfunction:: void cvEllipseBox( CvArr* img, CvBox2D box, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.EllipseBox(img, box, color, thickness=1, lineType=8, shift=0)-> None
-
     :param img: Image.
 
     :param center: Center of the ellipse.
@@ -162,14 +158,14 @@ fillConvexPoly
 ------------------
 Fills a convex polygon.
 
-.. ocv:function:: void fillConvexPoly(Mat& img, const Point* pts, int npts, const Scalar& color, int lineType=8, int shift=0)
+.. ocv:function:: void fillConvexPoly( Mat& img, const Point* pts, int npts, const Scalar& color, int lineType=LINE_8, int shift=0 )
+
+.. ocv:function:: void fillConvexPoly( InputOutputArray img, InputArray points, const Scalar& color, int lineType=LINE_8, int shift=0 )
 
 .. ocv:pyfunction:: cv2.fillConvexPoly(img, points, color[, lineType[, shift]]) -> img
 
 .. ocv:cfunction:: void cvFillConvexPoly( CvArr* img, const CvPoint* pts, int npts, CvScalar color, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.FillConvexPoly(img, pn, color, lineType=8, shift=0)-> None
-
     :param img: Image.
 
     :param pts: Polygon vertices.
@@ -192,14 +188,14 @@ fillPoly
 ------------
 Fills the area bounded by one or more polygons.
 
-.. ocv:function:: void fillPoly(Mat& img, const Point** pts,               const int* npts, int ncontours,              const Scalar& color, int lineType=8,              int shift=0, Point offset=Point() )
+.. ocv:function:: void fillPoly( Mat& img, const Point** pts, const int* npts, int ncontours, const Scalar& color, int lineType=LINE_8, int shift=0, Point offset=Point() )
+
+.. ocv:function:: void fillPoly( InputOutputArray img, InputArrayOfArrays pts, const Scalar& color, int lineType=LINE_8, int shift=0, Point offset=Point() )
 
 .. ocv:pyfunction:: cv2.fillPoly(img, pts, color[, lineType[, shift[, offset]]]) -> img
 
 .. ocv:cfunction:: void cvFillPoly( CvArr* img, CvPoint** pts, const int* npts, int contours, CvScalar color, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.FillPoly(img, polys, color, lineType=8, shift=0)-> None
-
     :param img: Image.
 
     :param pts: Array of polygons where each polygon is represented as an array of points.
@@ -225,16 +221,16 @@ getTextSize
 ---------------
 Calculates the width and height of a text string.
 
-.. ocv:function:: Size getTextSize(const string& text, int fontFace, double fontScale, int thickness, int* baseLine)
+.. ocv:function:: Size getTextSize(const String& text, int fontFace, double fontScale, int thickness, int* baseLine)
 
 .. ocv:pyfunction:: cv2.getTextSize(text, fontFace, fontScale, thickness) -> retval, baseLine
 
 .. ocv:cfunction:: void cvGetTextSize( const char* text_string, const CvFont* font, CvSize* text_size, int* baseline )
 
-.. ocv:pyoldfunction:: cv.GetTextSize(textString, font)-> (textSize, baseline)
-
     :param text: Input text string.
 
+    :param text_string: Input text string in C format.
+
     :param fontFace: Font to use. See the  :ocv:func:`putText` for details.
 
     :param fontScale: Font scale. See the  :ocv:func:`putText`  for details.
@@ -243,10 +239,16 @@ Calculates the width and height of a text string.
 
     :param baseLine: Output parameter - y-coordinate of the baseline relative to the bottom-most text point.
 
+    :param baseline: Output parameter - y-coordinate of the baseline relative to the bottom-most text point.
+
+    :param font: Font description in terms of old C API.
+
+    :param text_size: Output parameter - The size of a box that contains the specified text.
+
 The function ``getTextSize`` calculates and returns the size of a box that contains the specified text.
 That is, the following code renders some text, the tight box surrounding it, and the baseline: ::
 
-    string text = "Funny text inside the box";
+    String text = "Funny text inside the box";
     int fontFace = FONT_HERSHEY_SCRIPT_SIMPLEX;
     double fontScale = 2;
     int thickness = 3;
@@ -330,14 +332,12 @@ line
 --------
 Draws a line segment connecting two points.
 
-.. ocv:function:: void line(Mat& img, Point pt1, Point pt2, const Scalar& color,          int thickness=1, int lineType=8, int shift=0)
+.. ocv:function:: void line( Mat& img, Point pt1, Point pt2, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
 .. ocv:pyfunction:: cv2.line(img, pt1, pt2, color[, thickness[, lineType[, shift]]]) -> img
 
 .. ocv:cfunction:: void cvLine( CvArr* img, CvPoint pt1, CvPoint pt2, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.Line(img, pt1, pt2, color, thickness=1, lineType=8, shift=0)-> None
-
     :param img: Image.
 
     :param pt1: First point of the line segment.
@@ -350,16 +350,16 @@ Draws a line segment connecting two points.
 
     :param lineType: Type of the line:
 
-            * **8** (or omitted) - 8-connected line.
+            * **LINE_8** (or omitted) - 8-connected line.
 
-            * **4** - 4-connected line.
+            * **LINE_4** - 4-connected line.
 
-            * **CV_AA** - antialiased line.
+            * **LINE_AA** - antialiased line.
 
     :param shift: Number of fractional bits in the point coordinates.
 
 The function ``line`` draws the line segment between ``pt1`` and ``pt2`` points in the image. The line is clipped by the image boundaries. For non-antialiased lines with integer coordinates, the 8-connected or 4-connected Bresenham algorithm is used. Thick lines are drawn with rounding endings.
-Antialiased lines are drawn using Gaussian filtering. To specify the line color, you may use the macro ``CV_RGB(r, g, b)`` .
+Antialiased lines are drawn using Gaussian filtering.
 
 
 LineIterator
@@ -418,16 +418,14 @@ rectangle
 -------------
 Draws a simple, thick, or filled up-right rectangle.
 
-.. ocv:function:: void rectangle(Mat& img, Point pt1, Point pt2, const Scalar& color, int thickness=1, int lineType=8, int shift=0)
+.. ocv:function:: void rectangle( Mat& img, Point pt1, Point pt2, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
-.. ocv:function:: void rectangle( Mat& img, Rect rec, const Scalar& color, int thickness=1, int lineType=8, int shift=0 )
+.. ocv:function:: void rectangle( Mat& img, Rect rec, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
 .. ocv:pyfunction:: cv2.rectangle(img, pt1, pt2, color[, thickness[, lineType[, shift]]]) -> img
 
 .. ocv:cfunction:: void cvRectangle( CvArr* img, CvPoint pt1, CvPoint pt2, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.Rectangle(img, pt1, pt2, color, thickness=1, lineType=8, shift=0)-> None
-
     :param img: Image.
 
     :param pt1: Vertex of the rectangle.
@@ -452,16 +450,14 @@ polylines
 -------------
 Draws several polygonal curves.
 
-.. ocv:function:: void polylines( Mat& img, const Point* const* pts, const int* npts, int ncontours, bool isClosed, const Scalar& color, int thickness=1, int lineType=8, int shift=0 )
+.. ocv:function:: void polylines( Mat& img, const Point* const* pts, const int* npts, int ncontours, bool isClosed, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
-.. ocv:function:: void polylines( InputOutputArray img, InputArrayOfArrays pts, bool isClosed, const Scalar& color, int thickness=1, int lineType=8, int shift=0 )
+.. ocv:function:: void polylines( InputOutputArray img, InputArrayOfArrays pts, bool isClosed, const Scalar& color, int thickness=1, int lineType=LINE_8, int shift=0 )
 
 .. ocv:pyfunction:: cv2.polylines(img, pts, isClosed, color[, thickness[, lineType[, shift]]]) -> img
 
 .. ocv:cfunction:: void cvPolyLine( CvArr* img, CvPoint** pts, const int* npts, int contours, int is_closed, CvScalar color, int thickness=1, int line_type=8, int shift=0 )
 
-.. ocv:pyoldfunction:: cv.PolyLine(img, polys, is_closed, color, thickness=1, lineType=8, shift=0) -> None
-
     :param img: Image.
 
     :param pts: Array of polygonal curves.
@@ -487,14 +483,12 @@ drawContours
 ----------------
 Draws contours outlines or filled contours.
 
-.. ocv:function:: void drawContours( InputOutputArray image, InputArrayOfArrays contours, int contourIdx, const Scalar& color, int thickness=1, int lineType=8, InputArray hierarchy=noArray(), int maxLevel=INT_MAX, Point offset=Point() )
+.. ocv:function:: void drawContours( InputOutputArray image, InputArrayOfArrays contours, int contourIdx, const Scalar& color, int thickness=1, int lineType=LINE_8, InputArray hierarchy=noArray(), int maxLevel=INT_MAX, Point offset=Point() )
 
 .. ocv:pyfunction:: cv2.drawContours(image, contours, contourIdx, color[, thickness[, lineType[, hierarchy[, maxLevel[, offset]]]]]) -> image
 
 .. ocv:cfunction:: void cvDrawContours( CvArr * img, CvSeq* contour, CvScalar external_color, CvScalar hole_color, int max_level, int thickness=1, int line_type=8, CvPoint offset=cvPoint(0,0) )
 
-.. ocv:pyoldfunction:: cv.DrawContours(img, contour, external_color, hole_color, max_level, thickness=1, lineType=8, offset=(0, 0))-> None
-
     :param image: Destination image.
 
     :param contours: All the input contours. Each contour is stored as a point vector.
@@ -525,10 +519,11 @@ The function draws contour outlines in the image if
 :math:`\texttt{thickness} \ge 0` or fills the area bounded by the contours if
 :math:`\texttt{thickness}<0` . The example below shows how to retrieve connected components from the binary image and label them: ::
 
-    #include "cv.h"
-    #include "highgui.h"
+    #include "opencv2/imgproc.hpp"
+    #include "opencv2/highgui.hpp"
 
     using namespace cv;
+    using namespace std;
 
     int main( int argc, char** argv )
     {
@@ -548,7 +543,7 @@ The function draws contour outlines in the image if
         vector<Vec4i> hierarchy;
 
         findContours( src, contours, hierarchy,
-            CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
+            RETR_CCOMP, CHAIN_APPROX_SIMPLE );
 
         // iterate through all the top-level contours,
         // draw each connected component with its own random color
@@ -556,7 +551,7 @@ The function draws contour outlines in the image if
         for( ; idx >= 0; idx = hierarchy[idx][0] )
         {
             Scalar color( rand()&255, rand()&255, rand()&255 );
-            drawContours( dst, contours, idx, color, CV_FILLED, 8, hierarchy );
+            drawContours( dst, contours, idx, color, FILLED, 8, hierarchy );
         }
 
         namedWindow( "Components", 1 );
@@ -564,18 +559,23 @@ The function draws contour outlines in the image if
         waitKey(0);
     }
 
+.. note::
+
+   * An example using the drawContour functionality can be found at opencv_source_code/samples/cpp/contours2.cpp
+   * An example using drawContours to clean up a background segmentation result at opencv_source_code/samples/cpp/segment_objects.cpp
+
+   * (Python) An example using the drawContour functionality can be found at opencv_source/samples/python2/contours.py
 
 
 putText
 -----------
 Draws a text string.
 
-.. ocv:function:: void putText( Mat& img, const string& text, Point org,              int fontFace, double fontScale, Scalar color,              int thickness=1, int lineType=8,              bool bottomLeftOrigin=false )
+.. ocv:function:: void putText( Mat& img, const String& text, Point org, int fontFace, double fontScale, Scalar color, int thickness=1, int lineType=LINE_8, bool bottomLeftOrigin=false )
 
 .. ocv:pyfunction:: cv2.putText(img, text, org, fontFace, fontScale, color[, thickness[, lineType[, bottomLeftOrigin]]]) -> None
 
 .. ocv:cfunction:: void cvPutText( CvArr* img, const char* text, CvPoint org, const CvFont* font, CvScalar color )
-.. ocv:pyoldfunction:: cv.PutText(img, text, org, font, color)-> None
 
     :param img: Image.
 
@@ -602,4 +602,3 @@ The function ``putText`` renders the specified text string in the image.
 Symbols that cannot be rendered using the specified font are
 replaced by question marks. See
 :ocv:func:`getTextSize` for a text rendering code example.
-
diff --git a/modules/core/doc/dynamic_structures.rst b/modules/core/doc/dynamic_structures.rst
index e333971c06..0f3df5d12e 100644
--- a/modules/core/doc/dynamic_structures.rst
+++ b/modules/core/doc/dynamic_structures.rst
@@ -279,7 +279,6 @@ CloneSeq
 Creates a copy of a sequence.
 
 .. ocv:cfunction:: CvSeq* cvCloneSeq(  const CvSeq* seq, CvMemStorage* storage=NULL )
-.. ocv:pyoldfunction:: cv.CloneSeq(seq, storage)-> None
 
     :param seq: Sequence
 
@@ -388,9 +387,6 @@ Creates memory storage.
 
 .. ocv:cfunction:: CvMemStorage* cvCreateMemStorage( int block_size=0 )
 
-.. ocv:pyoldfunction:: cv.CreateMemStorage(blockSize=0) -> memstorage
-
-
     :param block_size: Size of the storage blocks in bytes. If it is 0, the block size is set to a default value - currently it is  about 64K.
 
 The function creates an empty memory storage. See
@@ -1584,4 +1580,3 @@ Gathers all node pointers to a single sequence.
     :param storage: Container for the sequence
 
 The function puts pointers of all nodes reachable from  ``first`` into a single sequence. The pointers are written sequentially in the depth-first order.
-
diff --git a/modules/core/doc/intro.rst b/modules/core/doc/intro.rst
index 806f434b26..6d9fdfca5e 100644
--- a/modules/core/doc/intro.rst
+++ b/modules/core/doc/intro.rst
@@ -4,7 +4,7 @@ Introduction
 
 .. highlight:: cpp
 
-OpenCV (Open Source Computer Vision Library: http://opencv.willowgarage.com/wiki/) is an open-source BSD-licensed library that includes several hundreds of computer vision algorithms. The document describes the so-called OpenCV 2.x API, which is essentially a C++ API, as opposite to the C-based OpenCV 1.x API. The latter is described in opencv1x.pdf.
+OpenCV (Open Source Computer Vision Library: http://opencv.org) is an open-source BSD-licensed library that includes several hundreds of computer vision algorithms. The document describes the so-called OpenCV 2.x API, which is essentially a C++ API, as opposite to the C-based OpenCV 1.x API. The latter is described in opencv1x.pdf.
 
 OpenCV has a modular structure, which means that the package includes several shared or static libraries. The following modules are available:
 
@@ -83,17 +83,22 @@ First of all, ``std::vector``, ``Mat``, and other data structures used by the fu
     // matrix will be deallocated, since it is not referenced by anyone
     C = C.clone();
 
-You see that the use of ``Mat`` and other basic structures is simple. But what about high-level classes or even user data types created without taking automatic memory management into account? For them, OpenCV offers the ``Ptr<>`` template class that is similar to ``std::shared_ptr`` from C++ TR1. So, instead of using plain pointers::
+You see that the use of ``Mat`` and other basic structures is simple. But what about high-level classes or even user
+data types created without taking automatic memory management into account? For them, OpenCV offers the :ocv:class:`Ptr`
+template class that is similar to ``std::shared_ptr`` from C++11. So, instead of using plain pointers::
 
    T* ptr = new T(...);
 
 you can use::
 
-   Ptr<T> ptr = new T(...);
+   Ptr<T> ptr(new T(...));
 
-That is, ``Ptr<T> ptr`` encapsulates a pointer to a ``T`` instance and a reference counter associated with the pointer. See the
-:ocv:class:`Ptr`
-description for details.
+or::
+
+   Ptr<T> ptr = makePtr<T>(...);
+
+``Ptr<T>`` encapsulates a pointer to a ``T`` instance and a reference counter associated with the pointer. See the
+:ocv:class:`Ptr` description for details.
 
 .. _AutomaticAllocation:
 
@@ -104,8 +109,8 @@ OpenCV deallocates the memory automatically, as well as automatically allocates
 
 Example: ::
 
-    #include "cv.h"
-    #include "highgui.h"
+    #include "opencv2/imgproc.hpp"
+    #include "opencv2/highgui.hpp"
 
     using namespace cv;
 
@@ -119,7 +124,7 @@ Example: ::
         for(;;)
         {
             cap >> frame;
-            cvtColor(frame, edges, CV_BGR2GRAY);
+            cvtColor(frame, edges, COLOR_BGR2GRAY);
             GaussianBlur(edges, edges, Size(7,7), 1.5, 1.5);
             Canny(edges, edges, 0, 30, 3);
             imshow("edges", edges);
@@ -128,7 +133,7 @@ Example: ::
         return 0;
     }
 
-The array ``frame`` is automatically allocated by the ``>>`` operator since the video frame resolution and the bit-depth is known to the video capturing module. The array ``edges`` is automatically allocated by the ``cvtColor`` function. It has the same size and the bit-depth as the input array. The number of channels is 1 because the color conversion code ``CV_BGR2GRAY`` is passed, which means a color to grayscale conversion. Note that ``frame`` and ``edges`` are allocated only once during the first execution of the loop body since all the next video frames have the same resolution. If you somehow change the video resolution, the arrays are automatically reallocated.
+The array ``frame`` is automatically allocated by the ``>>`` operator since the video frame resolution and the bit-depth is known to the video capturing module. The array ``edges`` is automatically allocated by the ``cvtColor`` function. It has the same size and the bit-depth as the input array. The number of channels is 1 because the color conversion code ``COLOR_BGR2GRAY`` is passed, which means a color to grayscale conversion. Note that ``frame`` and ``edges`` are allocated only once during the first execution of the loop body since all the next video frames have the same resolution. If you somehow change the video resolution, the arrays are automatically reallocated.
 
 The key component of this technology is the ``Mat::create`` method. It takes the desired array size and type. If the array already has the specified size and type, the method does nothing. Otherwise, it releases the previously allocated data, if any (this part involves decrementing the reference counter and comparing it with zero), and then allocates a new buffer of the required size. Most functions call the ``Mat::create`` method for each output array, and so the automatic output data allocation is implemented.
 
diff --git a/modules/core/doc/old_basic_structures.rst b/modules/core/doc/old_basic_structures.rst
index 0596e04bff..906306dcb2 100644
--- a/modules/core/doc/old_basic_structures.rst
+++ b/modules/core/doc/old_basic_structures.rst
@@ -478,8 +478,6 @@ Clears a specific array element.
 
 .. ocv:cfunction:: void cvClearND( CvArr* arr, const int* idx )
 
-.. ocv:pyoldfunction:: cv.ClearND(arr, idx)-> None
-
     :param arr: Input array
     :param idx: Array of the element indices
 
@@ -490,7 +488,6 @@ CloneImage
 Makes a full copy of an image, including the header, data, and ROI.
 
 .. ocv:cfunction:: IplImage* cvCloneImage(const IplImage* image)
-.. ocv:pyoldfunction:: cv.CloneImage(image) -> image
 
     :param image: The original image
 
@@ -499,7 +496,6 @@ CloneMat
 Creates a full matrix copy.
 
 .. ocv:cfunction:: CvMat* cvCloneMat(const CvMat* mat)
-.. ocv:pyoldfunction:: cv.CloneMat(mat) -> mat
 
     :param mat: Matrix to be copied
 
@@ -510,7 +506,6 @@ CloneMatND
 Creates full copy of a multi-dimensional array and returns a pointer to the copy.
 
 .. ocv:cfunction:: CvMatND* cvCloneMatND(const CvMatND* mat)
-.. ocv:pyoldfunction:: cv.CloneMatND(mat) -> matND
 
     :param mat: Input array
 
@@ -530,8 +525,6 @@ ConvertScale
 Converts one array to another with optional linear transformation.
 
 .. ocv:cfunction:: void cvConvertScale(const CvArr* src, CvArr* dst, double scale=1, double shift=0)
-.. ocv:pyoldfunction:: cv.ConvertScale(src, dst, scale=1.0, shift=0.0)-> None
-.. ocv:pyoldfunction:: cv.Convert(src, dst)-> None
 
     ::
 
@@ -569,7 +562,6 @@ Copy
 Copies one array to another.
 
 .. ocv:cfunction:: void cvCopy(const CvArr* src, CvArr* dst, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.Copy(src, dst, mask=None)-> None
 
     :param src: The source array
 
@@ -591,7 +583,6 @@ CreateData
 Allocates array data
 
 .. ocv:cfunction:: void cvCreateData(CvArr* arr)
-.. ocv:pyoldfunction:: cv.CreateData(arr) -> None
 
     :param arr: Array header
 
@@ -603,7 +594,6 @@ CreateImage
 Creates an image header and allocates the image data.
 
 .. ocv:cfunction:: IplImage* cvCreateImage(CvSize size, int depth, int channels)
-.. ocv:pyoldfunction:: cv.CreateImage(size, depth, channels)->image
 
     :param size: Image width and height
 
@@ -621,7 +611,6 @@ CreateImageHeader
 Creates an image header but does not allocate the image data.
 
 .. ocv:cfunction:: IplImage* cvCreateImageHeader(CvSize size, int depth, int channels)
-.. ocv:pyoldfunction:: cv.CreateImageHeader(size, depth, channels) -> image
 
     :param size: Image width and height
 
@@ -634,7 +623,6 @@ CreateMat
 Creates a matrix header and allocates the matrix data.
 
 .. ocv:cfunction:: CvMat* cvCreateMat( int rows, int cols, int type)
-.. ocv:pyoldfunction:: cv.CreateMat(rows, cols, type) -> mat
 
     :param rows: Number of rows in the matrix
 
@@ -652,7 +640,6 @@ CreateMatHeader
 Creates a matrix header but does not allocate the matrix data.
 
 .. ocv:cfunction:: CvMat* cvCreateMatHeader( int rows, int cols, int type)
-.. ocv:pyoldfunction:: cv.CreateMatHeader(rows, cols, type) -> mat
 
     :param rows: Number of rows in the matrix
 
@@ -667,7 +654,6 @@ CreateMatND
 Creates the header and allocates the data for a multi-dimensional dense array.
 
 .. ocv:cfunction:: CvMatND* cvCreateMatND( int dims, const int* sizes, int type)
-.. ocv:pyoldfunction:: cv.CreateMatND(dims, type) -> matND
 
     :param dims: Number of array dimensions. This must not exceed CV_MAX_DIM (32 by default, but can be changed at build time).
 
@@ -685,7 +671,6 @@ CreateMatNDHeader
 Creates a new matrix header but does not allocate the matrix data.
 
 .. ocv:cfunction:: CvMatND* cvCreateMatNDHeader( int dims, const int* sizes, int type)
-.. ocv:pyoldfunction:: cv.CreateMatNDHeader(dims, type) -> matND
 
     :param dims: Number of array dimensions
 
@@ -716,7 +701,6 @@ CrossProduct
 Calculates the cross product of two 3D vectors.
 
 .. ocv:cfunction:: void cvCrossProduct(const CvArr* src1, const CvArr* src2, CvArr* dst)
-.. ocv:pyoldfunction:: cv.CrossProduct(src1, src2, dst)-> None
 
     :param src1: The first source vector
 
@@ -742,7 +726,6 @@ DotProduct
 Calculates the dot product of two arrays in Euclidean metrics.
 
 .. ocv:cfunction:: double cvDotProduct(const CvArr* src1, const CvArr* src2)
-.. ocv:pyoldfunction:: cv.DotProduct(src1, src2) -> float
 
     :param src1: The first source array
 
@@ -767,11 +750,6 @@ Get?D
 .. ocv:cfunction:: CvScalar cvGet3D(const CvArr* arr, int idx0, int idx1, int idx2)
 .. ocv:cfunction:: CvScalar cvGetND( const CvArr* arr, const int* idx )
 
-.. ocv:pyoldfunction:: cv.Get1D(arr, idx) -> scalar
-.. ocv:pyoldfunction:: cv.Get2D(arr, idx0, idx1) -> scalar
-.. ocv:pyoldfunction:: cv.Get3D(arr, idx0, idx1, idx2) -> scalar
-.. ocv:pyoldfunction:: cv.GetND(arr, indices) -> scalar
-
     Return a specific array element.
 
     :param arr: Input array
@@ -794,10 +772,6 @@ Returns one of more array columns.
 
 .. ocv:cfunction:: CvMat* cvGetCols( const CvArr* arr, CvMat* submat, int start_col, int end_col )
 
-.. ocv:pyoldfunction:: cv.GetCol(arr, col)-> submat
-
-.. ocv:pyoldfunction:: cv.GetCols(arr, startCol, endCol)-> submat
-
     :param arr: Input array
 
     :param submat: Pointer to the resulting sub-array header
@@ -815,7 +789,6 @@ GetDiag
 Returns one of array diagonals.
 
 .. ocv:cfunction:: CvMat* cvGetDiag(const CvArr* arr, CvMat* submat, int diag=0)
-.. ocv:pyoldfunction:: cv.GetDiag(arr, diag=0)-> submat
 
     :param arr: Input array
 
@@ -830,7 +803,6 @@ GetDims
 Return number of array dimensions
 
 .. ocv:cfunction:: int cvGetDims(const CvArr* arr, int* sizes=NULL)
-.. ocv:pyoldfunction:: cv.GetDims(arr) -> (dim1, dim2, ...)
 
     :param arr: Input array
 
@@ -861,7 +833,6 @@ GetElemType
 Returns type of array elements.
 
 .. ocv:cfunction:: int cvGetElemType(const CvArr* arr)
-.. ocv:pyoldfunction:: cv.GetElemType(arr)-> int
 
     :param arr: Input array
 
@@ -877,8 +848,6 @@ Returns image header for arbitrary array.
 
 .. ocv:cfunction:: IplImage* cvGetImage( const CvArr* arr, IplImage* image_header )
 
-.. ocv:pyoldfunction:: cv.GetImage(arr) -> iplimage
-
     :param arr: Input array
 
     :param image_header: Pointer to  ``IplImage``  structure used as a temporary buffer
@@ -890,7 +859,6 @@ GetImageCOI
 Returns the index of the channel of interest.
 
 .. ocv:cfunction:: int cvGetImageCOI(const IplImage* image)
-.. ocv:pyoldfunction:: cv.GetImageCOI(image) -> int
 
     :param image: A pointer to the image header
 
@@ -902,7 +870,6 @@ GetImageROI
 Returns the image ROI.
 
 .. ocv:cfunction:: CvRect cvGetImageROI(const IplImage* image)
-.. ocv:pyoldfunction:: cv.GetImageROI(image)-> CvRect
 
     :param image: A pointer to the image header
 
@@ -913,7 +880,6 @@ GetMat
 Returns matrix header for arbitrary array.
 
 .. ocv:cfunction:: CvMat* cvGetMat(const CvArr* arr, CvMat* header, int* coi=NULL, int allowND=0)
-.. ocv:pyoldfunction:: cv.GetMat(arr, allowND=0) -> mat
 
     :param arr: Input array
 
@@ -1002,11 +968,6 @@ Return a specific element of single-channel 1D, 2D, 3D or nD array.
 .. ocv:cfunction:: double cvGetReal3D(const CvArr* arr, int idx0, int idx1, int idx2)
 .. ocv:cfunction:: double cvGetRealND( const CvArr* arr, const int* idx )
 
-.. ocv:pyoldfunction:: cv.GetReal1D(arr, idx0)->float
-.. ocv:pyoldfunction:: cv.GetReal2D(arr, idx0, idx1)->float
-.. ocv:pyoldfunction:: cv.GetReal3D(arr, idx0, idx1, idx2)->float
-.. ocv:pyoldfunction:: cv.GetRealND(arr, idx)->float
-
     :param arr: Input array. Must have a single channel.
 
     :param idx0: The first zero-based component of the element index
@@ -1030,9 +991,6 @@ Returns array row or row span.
 
 .. ocv:cfunction:: CvMat* cvGetRows( const CvArr* arr, CvMat* submat, int start_row, int end_row, int delta_row=1 )
 
-.. ocv:pyoldfunction:: cv.GetRow(arr, row)-> submat
-.. ocv:pyoldfunction:: cv.GetRows(arr, startRow, endRow, deltaRow=1)-> submat
-
     :param arr: Input array
 
     :param submat: Pointer to the resulting sub-array header
@@ -1053,7 +1011,6 @@ GetSize
 Returns size of matrix or image ROI.
 
 .. ocv:cfunction:: CvSize cvGetSize(const CvArr* arr)
-.. ocv:pyoldfunction:: cv.GetSize(arr)-> (width, height)
 
     :param arr: array header
 
@@ -1064,7 +1021,6 @@ GetSubRect
 Returns matrix header corresponding to the rectangular sub-array of input image or matrix.
 
 .. ocv:cfunction:: CvMat* cvGetSubRect(const CvArr* arr, CvMat* submat, CvRect rect)
-.. ocv:pyoldfunction:: cv.GetSubRect(arr, rect) -> submat
 
     :param arr: Input array
 
@@ -1357,7 +1313,6 @@ ResetImageROI
 Resets the image ROI to include the entire image and releases the ROI structure.
 
 .. ocv:cfunction:: void cvResetImageROI(IplImage* image)
-.. ocv:pyoldfunction:: cv.ResetImageROI(image)-> None
 
     :param image: A pointer to the image header
 
@@ -1374,8 +1329,6 @@ Changes shape of matrix/image without copying data.
 
 .. ocv:cfunction:: CvMat* cvReshape( const CvArr* arr, CvMat* header, int new_cn, int new_rows=0 )
 
-.. ocv:pyoldfunction:: cv.Reshape(arr, newCn, newRows=0) -> mat
-
     :param arr: Input array
 
     :param header: Output header to be filled
@@ -1412,8 +1365,6 @@ Changes the shape of a multi-dimensional array without copying the data.
 
 .. ocv:cfunction:: CvArr* cvReshapeMatND( const CvArr* arr, int sizeof_header, CvArr* header, int new_cn, int new_dims, int* new_sizes )
 
-.. ocv:pyoldfunction:: cv.ReshapeMatND(arr, newCn, newDims) -> mat
-
     :param arr: Input array
 
     :param sizeof_header: Size of output header to distinguish between IplImage, CvMat and CvMatND output headers
@@ -1453,7 +1404,6 @@ Set
 Sets every element of an array to a given value.
 
 .. ocv:cfunction:: void cvSet(CvArr* arr, CvScalar value, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.Set(arr, value, mask=None)-> None
 
     :param arr: The destination array
 
@@ -1481,12 +1431,6 @@ Change the particular array element.
 
 .. ocv:cfunction:: void cvSetND( CvArr* arr, const int* idx, CvScalar value )
 
-.. ocv:pyoldfunction:: cv.Set1D(arr, idx, value) -> None
-.. ocv:pyoldfunction:: cv.Set2D(arr, idx0, idx1, value) -> None
-.. ocv:pyoldfunction:: cv.Set3D(arr, idx0, idx1, idx2, value) -> None
-.. ocv:pyoldfunction:: cv.SetND(arr, indices, value) -> None
-
-
     :param arr: Input array
 
     :param idx0: The first zero-based component of the element index
@@ -1506,7 +1450,6 @@ SetData
 Assigns user data to the array header.
 
 .. ocv:cfunction:: void cvSetData(CvArr* arr, void* data, int step)
-.. ocv:pyoldfunction:: cv.SetData(arr, data, step)-> None
 
     :param arr: Array header
 
@@ -1525,7 +1468,6 @@ SetImageCOI
 Sets the channel of interest in an IplImage.
 
 .. ocv:cfunction:: void cvSetImageCOI( IplImage* image, int coi)
-.. ocv:pyoldfunction:: cv.SetImageCOI(image, coi)-> None
 
     :param image: A pointer to the image header
 
@@ -1539,7 +1481,6 @@ SetImageROI
 Sets an image Region Of Interest (ROI) for a given rectangle.
 
 .. ocv:cfunction:: void cvSetImageROI( IplImage* image, CvRect rect)
-.. ocv:pyoldfunction:: cv.SetImageROI(image, rect)-> None
 
     :param image: A pointer to the image header
 
@@ -1562,11 +1503,6 @@ Change a specific array element.
 
 .. ocv:cfunction:: void cvSetRealND( CvArr* arr, const int* idx, double value )
 
-.. ocv:pyoldfunction:: cv.SetReal1D(arr, idx, value) -> None
-.. ocv:pyoldfunction:: cv.SetReal2D(arr, idx0, idx1, value) -> None
-.. ocv:pyoldfunction:: cv.SetReal3D(arr, idx0, idx1, idx2, value) -> None
-.. ocv:pyoldfunction:: cv.SetRealND(arr, indices, value) -> None
-
     :param arr: Input array
 
     :param idx0: The first zero-based component of the element index
@@ -1588,7 +1524,6 @@ SetZero
 Clears the array.
 
 .. ocv:cfunction:: void cvSetZero(CvArr* arr)
-.. ocv:pyoldfunction:: cv.SetZero(arr) -> None
 
     :param arr: Array to be cleared
 
@@ -1599,7 +1534,6 @@ mGet
 Returns the particular element of single-channel floating-point matrix.
 
 .. ocv:cfunction:: double cvmGet(const CvMat* mat, int row, int col)
-.. ocv:pyoldfunction:: cv.mGet(mat, row, col) -> float
 
     :param mat: Input matrix
 
@@ -1614,7 +1548,6 @@ mSet
 Sets a specific element of a single-channel floating-point matrix.
 
 .. ocv:cfunction:: void cvmSet(CvMat* mat, int row, int col, double value)
-.. ocv:pyoldfunction:: cv.mSet(mat, row, col, value)-> None
 
     :param mat: The matrix
 
@@ -1655,7 +1588,6 @@ RNG
 Initializes a random number generator state.
 
 .. ocv:cfunction:: CvRNG cvRNG(int64 seed=-1)
-.. ocv:pyoldfunction:: cv.RNG(seed=-1LL)-> CvRNG
 
     :param seed: 64-bit value used to initiate a random sequence
 
@@ -1670,8 +1602,6 @@ Fills an array with random numbers and updates the RNG state.
 
 .. ocv:cfunction:: void cvRandArr( CvRNG* rng, CvArr* arr, int dist_type, CvScalar param1, CvScalar param2 )
 
-.. ocv:pyoldfunction:: cv.RandArr(rng, arr, distType, param1, param2)-> None
-
     :param rng: CvRNG state initialized by :ocv:cfunc:`RNG`
 
     :param arr: The destination array
@@ -1695,7 +1625,6 @@ RandInt
 Returns a 32-bit unsigned integer and updates RNG.
 
 .. ocv:cfunction:: unsigned cvRandInt(CvRNG* rng)
-.. ocv:pyoldfunction:: cv.RandInt(rng)-> unsigned
 
     :param rng: CvRNG state initialized by  :ocv:cfunc:`RNG`.
 
@@ -1707,7 +1636,6 @@ RandReal
 Returns a floating-point random number and updates RNG.
 
 .. ocv:cfunction:: double cvRandReal(CvRNG* rng)
-.. ocv:pyoldfunction:: cv.RandReal(rng) -> float
 
     :param rng: RNG state initialized by  :ocv:cfunc:`RNG`
 
@@ -1718,8 +1646,6 @@ fromarray
 ---------
 Create a CvMat from an object that supports the array interface.
 
-.. ocv:pyoldfunction:: cv.fromarray(array, allowND=False) -> mat
-
     :param object: Any object that supports the array interface
 
     :param allowND: If true, will return a CvMatND
@@ -1751,4 +1677,3 @@ For example, `NumPy <http://numpy.scipy.org/>`_ arrays support the array interfa
     (480, 640, 3) 1
 
 .. note:: In the new Python wrappers (**cv2** module) the function is not needed, since cv2 can process  Numpy arrays (and this is the only supported array type).
-
diff --git a/modules/core/doc/old_xml_yaml_persistence.rst b/modules/core/doc/old_xml_yaml_persistence.rst
index bdd4c2ded4..7492213b6c 100644
--- a/modules/core/doc/old_xml_yaml_persistence.rst
+++ b/modules/core/doc/old_xml_yaml_persistence.rst
@@ -327,8 +327,6 @@ Loads an object from a file.
 
 .. ocv:cfunction:: void* cvLoad( const char* filename, CvMemStorage* memstorage=NULL, const char* name=NULL, const char** real_name=NULL )
 
-.. ocv:pyoldfunction:: cv.Load(filename, storage=None, name=None)-> generic
-
     :param filename: File name
 
     :param memstorage: Memory storage for dynamic structures, such as  :ocv:struct:`CvSeq`  or  :ocv:struct:`CvGraph`  . It is not used for matrices or images.
@@ -596,8 +594,6 @@ Saves an object to a file.
 
 .. ocv:cfunction:: void cvSave( const char* filename, const void* struct_ptr, const char* name=NULL, const char* comment=NULL, CvAttrList attributes=cvAttrList() )
 
-.. ocv:pyoldfunction:: cv.Save(filename, structPtr, name=None, comment=None)-> None
-
     :param filename: File name
 
     :param struct_ptr: Object to save
diff --git a/modules/core/doc/opengl_interop.rst b/modules/core/doc/opengl_interop.rst
new file mode 100644
index 0000000000..5267c4d451
--- /dev/null
+++ b/modules/core/doc/opengl_interop.rst
@@ -0,0 +1,543 @@
+OpenGL interoperability
+=======================
+
+.. highlight:: cpp
+
+
+
+General Information
+-------------------
+This section describes OpenGL interoperability.
+
+To enable OpenGL support, configure OpenCV using ``CMake`` with ``WITH_OPENGL=ON`` .
+Currently OpenGL is supported only with WIN32, GTK and Qt backends on Windows and Linux (MacOS and Android are not supported).
+For GTK backend ``gtkglext-1.0`` library is required.
+
+To use OpenGL functionality you should first create OpenGL context (window or frame buffer).
+You can do this with :ocv:func:`namedWindow` function or with other OpenGL toolkit (GLUT, for example).
+
+
+
+ogl::Buffer
+-----------
+Smart pointer for OpenGL buffer object with reference counting.
+
+.. ocv:class:: ogl::Buffer
+
+Buffer Objects are OpenGL objects that store an array of unformatted memory allocated by the OpenGL context.
+These can be used to store vertex data, pixel data retrieved from images or the framebuffer, and a variety of other things.
+
+``ogl::Buffer`` has interface similar with :ocv:class:`Mat` interface and represents 2D array memory.
+
+``ogl::Buffer`` supports memory transfers between host and device and also can be mapped to CUDA memory.
+
+
+
+ogl::Buffer::Target
+-------------------
+The target defines how you intend to use the buffer object.
+
+.. ocv:enum:: ogl::Buffer::Target
+
+    .. ocv:emember:: ARRAY_BUFFER
+
+        The buffer will be used as a source for vertex data.
+
+    .. ocv:emember:: ELEMENT_ARRAY_BUFFER
+
+        The buffer will be used for indices (in ``glDrawElements`` or :ocv:func:`ogl::render`, for example).
+
+    .. ocv:emember:: PIXEL_PACK_BUFFER
+
+        The buffer will be used for reading from OpenGL textures.
+
+    .. ocv:emember:: PIXEL_UNPACK_BUFFER
+
+        The buffer will be used for writing to OpenGL textures.
+
+
+
+ogl::Buffer::Buffer
+-------------------
+The constructors.
+
+.. ocv:function:: ogl::Buffer::Buffer()
+
+.. ocv:function:: ogl::Buffer::Buffer(int arows, int acols, int atype, unsigned int abufId, bool autoRelease = false)
+
+.. ocv:function:: ogl::Buffer::Buffer(Size asize, int atype, unsigned int abufId, bool autoRelease = false)
+
+.. ocv:function:: ogl::Buffer::Buffer(int arows, int acols, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+.. ocv:function:: ogl::Buffer::Buffer(Size asize, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+.. ocv:function:: ogl::Buffer::Buffer(InputArray arr, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+    :param arows: Number of rows in a 2D array.
+
+    :param acols: Number of columns in a 2D array.
+
+    :param asize: 2D array size.
+
+    :param atype: Array type ( ``CV_8UC1, ..., CV_64FC4`` ). See :ocv:class:`Mat` for details.
+
+    :param abufId: Buffer object name.
+
+    :param arr: Input array (host or device memory, it can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` or ``std::vector`` ).
+
+    :param target: Buffer usage. See :ocv:enum:`ogl::Buffer::Target` .
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+Creates empty ``ogl::Buffer`` object, creates ``ogl::Buffer`` object from existed buffer ( ``abufId`` parameter),
+allocates memory for ``ogl::Buffer`` object or copies from host/device memory.
+
+
+
+ogl::Buffer::create
+-------------------
+Allocates memory for ``ogl::Buffer`` object.
+
+.. ocv:function:: void ogl::Buffer::create(int arows, int acols, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+.. ocv:function:: void ogl::Buffer::create(Size asize, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+    :param arows: Number of rows in a 2D array.
+
+    :param acols: Number of columns in a 2D array.
+
+    :param asize: 2D array size.
+
+    :param atype: Array type ( ``CV_8UC1, ..., CV_64FC4`` ). See :ocv:class:`Mat` for details.
+
+    :param target: Buffer usage. See :ocv:enum:`ogl::Buffer::Target` .
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+
+
+ogl::Buffer::release
+--------------------
+Decrements the reference counter and destroys the buffer object if needed.
+
+.. ocv:function:: void ogl::Buffer::release()
+
+The function will call `setAutoRelease(true)` .
+
+
+
+ogl::Buffer::setAutoRelease
+---------------------------
+Sets auto release mode.
+
+.. ocv:function:: void ogl::Buffer::setAutoRelease(bool flag)
+
+    :param flag: Auto release mode (if true, release will be called in object's destructor).
+
+The lifetime of the OpenGL object is tied to the lifetime of the context.
+If OpenGL context was bound to a window it could be released at any time (user can close a window).
+If object's destructor is called after destruction of the context it will cause an error.
+Thus ``ogl::Buffer`` doesn't destroy OpenGL object in destructor by default (all OpenGL resources will be released with OpenGL context).
+This function can force ``ogl::Buffer`` destructor to destroy OpenGL object.
+
+
+
+ogl::Buffer::copyFrom
+---------------------
+Copies from host/device memory to OpenGL buffer.
+
+.. ocv:function:: void ogl::Buffer::copyFrom(InputArray arr, Target target = ARRAY_BUFFER, bool autoRelease = false)
+
+    :param arr: Input array (host or device memory, it can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` or ``std::vector`` ).
+
+    :param target: Buffer usage. See :ocv:enum:`ogl::Buffer::Target` .
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+
+
+ogl::Buffer::copyTo
+-------------------
+Copies from OpenGL buffer to host/device memory or another OpenGL buffer object.
+
+.. ocv:function:: void ogl::Buffer::copyTo(OutputArray arr) const
+
+    :param arr: Destination array (host or device memory, can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` , ``std::vector`` or ``ogl::Buffer`` ).
+
+
+
+ogl::Buffer::clone
+------------------
+Creates a full copy of the buffer object and the underlying data.
+
+.. ocv:function:: Buffer ogl::Buffer::clone(Target target = ARRAY_BUFFER, bool autoRelease = false) const
+
+    :param target: Buffer usage for destination buffer.
+
+    :param autoRelease: Auto release mode for destination buffer.
+
+
+
+ogl::Buffer::bind
+-----------------
+Binds OpenGL buffer to the specified buffer binding point.
+
+.. ocv:function:: void ogl::Buffer::bind(Target target) const
+
+    :param target: Binding point. See :ocv:enum:`ogl::Buffer::Target` .
+
+
+
+ogl::Buffer::unbind
+-------------------
+Unbind any buffers from the specified binding point.
+
+.. ocv:function:: static void ogl::Buffer::unbind(Target target)
+
+    :param target: Binding point. See :ocv:enum:`ogl::Buffer::Target` .
+
+
+
+ogl::Buffer::mapHost
+--------------------
+Maps OpenGL buffer to host memory.
+
+.. ocv:function:: Mat ogl::Buffer::mapHost(Access access)
+
+    :param access: Access policy, indicating whether it will be possible to read from, write to, or both read from and write to the buffer object's mapped data store. The symbolic constant must be ``ogl::Buffer::READ_ONLY`` , ``ogl::Buffer::WRITE_ONLY`` or ``ogl::Buffer::READ_WRITE`` .
+
+``mapHost`` maps to the client's address space the entire data store of the buffer object.
+The data can then be directly read and/or written relative to the returned pointer, depending on the specified ``access`` policy.
+
+A mapped data store must be unmapped with :ocv:func:`ogl::Buffer::unmapHost` before its buffer object is used.
+
+This operation can lead to memory transfers between host and device.
+
+Only one buffer object can be mapped at a time.
+
+
+
+ogl::Buffer::unmapHost
+----------------------
+Unmaps OpenGL buffer.
+
+.. ocv:function:: void ogl::Buffer::unmapHost()
+
+
+
+ogl::Buffer::mapDevice
+----------------------
+Maps OpenGL buffer to CUDA device memory.
+
+.. ocv:function:: gpu::GpuMat ogl::Buffer::mapDevice()
+
+This operatation doesn't copy data.
+Several buffer objects can be mapped to CUDA memory at a time.
+
+A mapped data store must be unmapped with :ocv:func:`ogl::Buffer::unmapDevice` before its buffer object is used.
+
+
+
+ogl::Buffer::unmapDevice
+------------------------
+Unmaps OpenGL buffer.
+
+.. ocv:function:: void ogl::Buffer::unmapDevice()
+
+
+
+ogl::Texture2D
+--------------
+Smart pointer for OpenGL 2D texture memory with reference counting.
+
+.. ocv:class:: ogl::Texture2D
+
+
+
+ogl::Texture2D::Format
+----------------------
+An Image Format describes the way that the images in Textures store their data.
+
+.. ocv:enum:: ogl::Texture2D::Format
+
+    .. ocv:emember:: NONE
+    .. ocv:emember:: DEPTH_COMPONENT
+    .. ocv:emember:: RGB
+    .. ocv:emember:: RGBA
+
+
+
+ogl::Texture2D::Texture2D
+-------------------------
+The constructors.
+
+.. ocv:function:: ogl::Texture2D::Texture2D()
+
+.. ocv:function:: ogl::Texture2D::Texture2D(int arows, int acols, Format aformat, unsigned int atexId, bool autoRelease = false)
+
+.. ocv:function:: ogl::Texture2D::Texture2D(Size asize, Format aformat, unsigned int atexId, bool autoRelease = false)
+
+.. ocv:function:: ogl::Texture2D::Texture2D(int arows, int acols, Format aformat, bool autoRelease = false)
+
+.. ocv:function:: ogl::Texture2D::Texture2D(Size asize, Format aformat, bool autoRelease = false)
+
+.. ocv:function:: ogl::Texture2D::Texture2D(InputArray arr, bool autoRelease = false)
+
+    :param arows: Number of rows.
+
+    :param acols: Number of columns.
+
+    :param asize: 2D array size.
+
+    :param aformat: Image format. See :ocv:enum:`ogl::Texture2D::Format` .
+
+    :param arr: Input array (host or device memory, it can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` or :ocv:class:`ogl::Buffer` ).
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+Creates empty ``ogl::Texture2D`` object, allocates memory for ``ogl::Texture2D`` object or copies from host/device memory.
+
+
+
+ogl::Texture2D::create
+----------------------
+Allocates memory for ``ogl::Texture2D`` object.
+
+.. ocv:function:: void ogl::Texture2D::create(int arows, int acols, Format aformat, bool autoRelease = false)
+
+.. ocv:function:: void ogl::Texture2D::create(Size asize, Format aformat, bool autoRelease = false)
+
+    :param arows: Number of rows.
+
+    :param acols: Number of columns.
+
+    :param asize: 2D array size.
+
+    :param aformat: Image format. See :ocv:enum:`ogl::Texture2D::Format` .
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+
+
+ogl::Texture2D::release
+-----------------------
+Decrements the reference counter and destroys the texture object if needed.
+
+.. ocv:function:: void ogl::Texture2D::release()
+
+The function will call `setAutoRelease(true)` .
+
+
+
+ogl::Texture2D::setAutoRelease
+------------------------------
+Sets auto release mode.
+
+.. ocv:function:: void ogl::Texture2D::setAutoRelease(bool flag)
+
+    :param flag: Auto release mode (if true, release will be called in object's destructor).
+
+The lifetime of the OpenGL object is tied to the lifetime of the context.
+If OpenGL context was bound to a window it could be released at any time (user can close a window).
+If object's destructor is called after destruction of the context it will cause an error.
+Thus ``ogl::Texture2D`` doesn't destroy OpenGL object in destructor by default (all OpenGL resources will be released with OpenGL context).
+This function can force ``ogl::Texture2D`` destructor to destroy OpenGL object.
+
+
+
+ogl::Texture2D::copyFrom
+------------------------
+Copies from host/device memory to OpenGL texture.
+
+.. ocv:function:: void ogl::Texture2D::copyFrom(InputArray arr, bool autoRelease = false)
+
+    :param arr: Input array (host or device memory, it can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` or :ocv:class:`ogl::Buffer` ).
+
+    :param autoRelease: Auto release mode (if true, release will be called in object's destructor).
+
+
+
+ogl::Texture2D::copyTo
+----------------------
+Copies from OpenGL texture to host/device memory or another OpenGL texture object.
+
+.. ocv:function:: void ogl::Texture2D::copyTo(OutputArray arr, int ddepth = CV_32F, bool autoRelease = false) const
+
+    :param arr: Destination array (host or device memory, can be :ocv:class:`Mat` , :ocv:class:`gpu::GpuMat` , :ocv:class:`ogl::Buffer` or ``ogl::Texture2D`` ).
+
+    :param ddepth: Destination depth.
+
+    :param autoRelease: Auto release mode for destination buffer (if ``arr`` is OpenGL buffer or texture).
+
+
+
+ogl::Texture2D::bind
+--------------------
+Binds texture to current active texture unit for ``GL_TEXTURE_2D`` target.
+
+.. ocv:function:: void ogl::Texture2D::bind() const
+
+
+
+ogl::Arrays
+-----------
+Wrapper for OpenGL Client-Side Vertex arrays.
+
+.. ocv:class:: ogl::Arrays
+
+``ogl::Arrays`` stores vertex data in :ocv:class:`ogl::Buffer` objects.
+
+
+
+ogl::Arrays::setVertexArray
+---------------------------
+Sets an array of vertex coordinates.
+
+.. ocv:function:: void ogl::Arrays::setVertexArray(InputArray vertex)
+
+    :param vertex: array with vertex coordinates, can be both host and device memory.
+
+
+
+ogl::Arrays::resetVertexArray
+-----------------------------
+Resets vertex coordinates.
+
+.. ocv:function:: void ogl::Arrays::resetVertexArray()
+
+
+
+ogl::Arrays::setColorArray
+--------------------------
+Sets an array of vertex colors.
+
+.. ocv:function:: void ogl::Arrays::setColorArray(InputArray color)
+
+    :param color: array with vertex colors, can be both host and device memory.
+
+
+
+ogl::Arrays::resetColorArray
+----------------------------
+Resets vertex colors.
+
+.. ocv:function:: void ogl::Arrays::resetColorArray()
+
+
+
+ogl::Arrays::setNormalArray
+---------------------------
+Sets an array of vertex normals.
+
+.. ocv:function:: void ogl::Arrays::setNormalArray(InputArray normal)
+
+    :param normal: array with vertex normals, can be both host and device memory.
+
+
+
+ogl::Arrays::resetNormalArray
+-----------------------------
+Resets vertex normals.
+
+.. ocv:function:: void ogl::Arrays::resetNormalArray()
+
+
+
+ogl::Arrays::setTexCoordArray
+-----------------------------
+Sets an array of vertex texture coordinates.
+
+.. ocv:function:: void ogl::Arrays::setTexCoordArray(InputArray texCoord)
+
+    :param texCoord: array with vertex texture coordinates, can be both host and device memory.
+
+
+
+ogl::Arrays::resetTexCoordArray
+-------------------------------
+Resets vertex texture coordinates.
+
+.. ocv:function:: void ogl::Arrays::resetTexCoordArray()
+
+
+
+ogl::Arrays::release
+--------------------
+Releases all inner buffers.
+
+.. ocv:function:: void ogl::Arrays::release()
+
+
+
+ogl::Arrays::setAutoRelease
+---------------------------
+Sets auto release mode all inner buffers.
+
+.. ocv:function:: void ogl::Arrays::setAutoRelease(bool flag)
+
+    :param flag: Auto release mode.
+
+
+
+ogl::Arrays::bind
+-----------------
+Binds all vertex arrays.
+
+.. ocv:function:: void ogl::Arrays::bind() const
+
+
+
+ogl::Arrays::size
+-----------------
+Returns the vertex count.
+
+.. ocv:function:: int ogl::Arrays::size() const
+
+
+
+ogl::render
+-----------
+Render OpenGL texture or primitives.
+
+.. ocv:function:: void ogl::render(const Texture2D& tex, Rect_<double> wndRect = Rect_<double>(0.0, 0.0, 1.0, 1.0), Rect_<double> texRect = Rect_<double>(0.0, 0.0, 1.0, 1.0))
+
+.. ocv:function:: void ogl::render(const Arrays& arr, int mode = POINTS, Scalar color = Scalar::all(255))
+
+.. ocv:function:: void ogl::render(const Arrays& arr, InputArray indices, int mode = POINTS, Scalar color = Scalar::all(255))
+
+    :param tex: Texture to draw.
+
+    :param wndRect: Region of window, where to draw a texture (normalized coordinates).
+
+    :param texRect: Region of texture to draw (normalized coordinates).
+
+    :param arr: Array of privitives vertices.
+
+    :param indices: Array of vertices indices (host or device memory).
+
+    :param mode: Render mode. Available options:
+
+        * **POINTS**
+        * **LINES**
+        * **LINE_LOOP**
+        * **LINE_STRIP**
+        * **TRIANGLES**
+        * **TRIANGLE_STRIP**
+        * **TRIANGLE_FAN**
+        * **QUADS**
+        * **QUAD_STRIP**
+        * **POLYGON**
+
+    :param color: Color for all vertices. Will be used if ``arr`` doesn't contain color array.
+
+
+
+gpu::setGlDevice
+----------------
+Sets a CUDA device and initializes it for the current thread with OpenGL interoperability.
+
+.. ocv:function:: void gpu::setGlDevice( int device = 0 )
+
+    :param device: System index of a GPU device starting with 0.
+
+This function should be explicitly called after OpenGL context creation and before any CUDA calls.
diff --git a/modules/core/doc/operations_on_arrays.rst b/modules/core/doc/operations_on_arrays.rst
index 64ca686184..d38f57ac33 100644
--- a/modules/core/doc/operations_on_arrays.rst
+++ b/modules/core/doc/operations_on_arrays.rst
@@ -36,8 +36,6 @@ Calculates the per-element absolute difference between two arrays or between an
 
 .. ocv:cfunction:: void cvAbsDiff(const CvArr* src1, const CvArr* src2, CvArr* dst)
 .. ocv:cfunction:: void cvAbsDiffS(const CvArr* src, CvArr* dst, CvScalar value)
-.. ocv:pyoldfunction:: cv.AbsDiff(src1, src2, dst)-> None
-.. ocv:pyoldfunction:: cv.AbsDiffS(src, dst, value)-> None
 
     :param src1: first input array or a scalar.
 
@@ -90,8 +88,6 @@ Calculates the per-element sum of two arrays or an array and a scalar.
 
 .. ocv:cfunction:: void cvAdd(const CvArr* src1, const CvArr* src2, CvArr* dst, const CvArr* mask=NULL)
 .. ocv:cfunction:: void cvAddS(const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.Add(src1, src2, dst, mask=None)-> None
-.. ocv:pyoldfunction:: cv.AddS(src, value, dst, mask=None)-> None
 
     :param src1: first input array or a scalar.
 
@@ -103,7 +99,7 @@ Calculates the per-element sum of two arrays or an array and a scalar.
 
     :param dst: output array that has the same size and number of channels as the input array(s); the depth is defined by ``dtype`` or ``src1``/``src2``.
 
-    :param mask: optional operation mask – 8-bit single channel array, that specifies elements of the output array to be changed.
+    :param mask: optional operation mask - 8-bit single channel array, that specifies elements of the output array to be changed.
 
     :param dtype: optional depth of the output array (see the discussion below).
 
@@ -160,7 +156,6 @@ Calculates the weighted sum of two arrays.
 .. ocv:pyfunction:: cv2.addWeighted(src1, alpha, src2, beta, gamma[, dst[, dtype]]) -> dst
 
 .. ocv:cfunction:: void cvAddWeighted(const CvArr* src1, double alpha, const CvArr* src2, double beta, double gamma, CvArr* dst)
-.. ocv:pyoldfunction:: cv.AddWeighted(src1, alpha, src2, beta, gamma, dst)-> None
 
     :param src1: first input array.
 
@@ -210,8 +205,6 @@ Calculates the per-element bit-wise conjunction of two arrays or an array and a
 
 .. ocv:cfunction:: void cvAnd(const CvArr* src1, const CvArr* src2, CvArr* dst, const CvArr* mask=NULL)
 .. ocv:cfunction:: void cvAndS(const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.And(src1, src2, dst, mask=None)-> None
-.. ocv:pyoldfunction:: cv.AndS(src, value, dst, mask=None)-> None
 
     :param src1: first input array or a scalar.
 
@@ -262,7 +255,6 @@ Inverts every bit of an array.
 .. ocv:pyfunction:: cv2.bitwise_not(src[, dst[, mask]]) -> dst
 
 .. ocv:cfunction:: void cvNot(const CvArr* src, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Not(src, dst)-> None
 
     :param src: input array.
 
@@ -290,8 +282,6 @@ Calculates the per-element bit-wise disjunction of two arrays or an array and a
 
 .. ocv:cfunction:: void cvOr(const CvArr* src1, const CvArr* src2, CvArr* dst, const CvArr* mask=NULL)
 .. ocv:cfunction:: void cvOrS(const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.Or(src1, src2, dst, mask=None)-> None
-.. ocv:pyoldfunction:: cv.OrS(src, value, dst, mask=None)-> None
 
     :param src1: first input array or a scalar.
 
@@ -342,8 +332,6 @@ Calculates the per-element bit-wise "exclusive or" operation on two arrays or an
 
 .. ocv:cfunction:: void cvXor(const CvArr* src1, const CvArr* src2, CvArr* dst, const CvArr* mask=NULL)
 .. ocv:cfunction:: void cvXorS(const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL)
-.. ocv:pyoldfunction:: cv.Xor(src1, src2, dst, mask=None)-> None
-.. ocv:pyoldfunction:: cv.XorS(src, value, dst, mask=None)-> None
 
     :param src1: first input array or a scalar.
 
@@ -396,8 +384,6 @@ Calculates the covariance matrix of a set of vectors.
 
 .. ocv:cfunction:: void cvCalcCovarMatrix( const CvArr** vects, int count, CvArr* cov_mat, CvArr* avg, int flags )
 
-.. ocv:pyoldfunction:: cv.CalcCovarMatrix(vects, covMat, avg, flags)-> None
-
     :param samples: samples stored either as separate matrices or as rows/columns of a single matrix.
 
     :param nsamples: number of samples when they are stored separately.
@@ -456,8 +442,6 @@ Calculates the magnitude and angle of 2D vectors.
 
 .. ocv:cfunction:: void cvCartToPolar( const CvArr* x, const CvArr* y, CvArr* magnitude, CvArr* angle=NULL, int angle_in_degrees=0 )
 
-.. ocv:pyoldfunction:: cv.CartToPolar(x, y, magnitude, angle=None, angleInDegrees=0)-> None
-
     :param x: array of x-coordinates; this must be a single-precision or double-precision floating-point array.
 
     :param y: array of y-coordinates, that must have the same size and same type as ``x``.
@@ -518,12 +502,8 @@ Performs the per-element comparison of two arrays or an array and scalar value.
 
 .. ocv:cfunction:: void cvCmp( const CvArr* src1, const CvArr* src2, CvArr* dst, int cmp_op )
 
-.. ocv:pyoldfunction:: cv.Cmp(src1, src2, dst, cmpOp)-> None
-
 .. ocv:cfunction:: void cvCmpS( const CvArr* src, double value, CvArr* dst, int cmp_op )
 
-.. ocv:pyoldfunction:: cv.CmpS(src, value, dst, cmpOp)-> None
-
     :param src1: first input array or a scalar (in the case of ``cvCmp``, ``cv.Cmp``, ``cvCmpS``, ``cv.CmpS`` it is always an array); when it is an array, it must have a single channel.
 
     :param src2: second input array or a scalar (in the case of ``cvCmp`` and ``cv.Cmp`` it is always an array; in the case of ``cvCmpS``, ``cv.CmpS`` it is always a scalar); when it is an array, it must have a single channel.
@@ -532,7 +512,7 @@ Performs the per-element comparison of two arrays or an array and scalar value.
 
     :param value: scalar value.
 
-    :param dst: output array that has the same size as the input arrays and type= ``CV_8UC1`` .
+    :param dst: output array that has the same size and type as the input arrays.
 
     :param cmpop: a flag, that specifies correspondence between the arrays:
 
@@ -624,7 +604,6 @@ Scales, calculates absolute values, and converts the result to 8-bit.
 .. ocv:pyfunction:: cv2.convertScaleAbs(src[, dst[, alpha[, beta]]]) -> dst
 
 .. ocv:cfunction:: void cvConvertScaleAbs(const CvArr* src, CvArr* dst, double scale=1, double shift=0)
-.. ocv:pyoldfunction:: cv.ConvertScaleAbs(src, dst, scale=1.0, shift=0.0)-> None
 
     :param src: input array.
 
@@ -668,8 +647,6 @@ Counts non-zero array elements.
 
 .. ocv:cfunction:: int cvCountNonZero(const CvArr* arr)
 
-.. ocv:pyoldfunction:: cv.CountNonZero(arr)-> int
-
     :param src: single-channel array.
 
 The function returns the number of non-zero elements in ``src`` :
@@ -760,7 +737,6 @@ Performs a forward or inverse discrete Cosine transform of 1D or 2D array.
 .. ocv:pyfunction:: cv2.dct(src[, dst[, flags]]) -> dst
 
 .. ocv:cfunction:: void cvDCT(const CvArr* src, CvArr* dst, int flags)
-.. ocv:pyoldfunction:: cv.DCT(src, dst, flags)-> None
 
     :param src: input floating-point array.
 
@@ -855,8 +831,6 @@ Performs a forward or inverse Discrete Fourier transform of a 1D or 2D floating-
 
 .. ocv:cfunction:: void cvDFT( const CvArr* src, CvArr* dst, int flags, int nonzero_rows=0 )
 
-.. ocv:pyoldfunction:: cv.DFT(src, dst, flags, nonzeroRows=0)-> None
-
     :param src: input array that could be real or complex.
 
     :param dst: output array whose size and type depends on the  ``flags`` .
@@ -868,7 +842,7 @@ Performs a forward or inverse Discrete Fourier transform of a 1D or 2D floating-
             * **DFT_SCALE** scales the result: divide it by the number of array elements. Normally, it is combined with  ``DFT_INVERSE``.
             * **DFT_ROWS** performs a forward or inverse transform of every individual row of the input matrix; this flag enables you to transform multiple vectors simultaneously and can be used to decrease the overhead (which is sometimes several times larger than the processing itself) to perform 3D and higher-dimensional transformations and so forth.
 
-            * **DFT_COMPLEX_OUTPUT** performs a forward transformation of 1D or 2D real array; the result, though being a complex array, has complex-conjugate symmetry (*CCS*, see the function description below for details), and such an array can be packed into a real array of the same size as input, which is the fastest option and which is what the function does by default; however, you may wish to get a full complex array (for simpler spectrum analysis, and so on) – pass the flag to enable the function to produce a full-size complex output array.
+            * **DFT_COMPLEX_OUTPUT** performs a forward transformation of 1D or 2D real array; the result, though being a complex array, has complex-conjugate symmetry (*CCS*, see the function description below for details), and such an array can be packed into a real array of the same size as input, which is the fastest option and which is what the function does by default; however, you may wish to get a full complex array (for simpler spectrum analysis, and so on) - pass the flag to enable the function to produce a full-size complex output array.
 
             * **DFT_REAL_OUTPUT** performs an inverse transformation of a 1D or 2D complex array; the result is normally a complex array of the same size, however, if the input array has conjugate-complex symmetry (for example, it is a result of forward transformation with  ``DFT_COMPLEX_OUTPUT``  flag), the output is a real array; while the function itself does not check whether the input is symmetrical or not, you can pass the flag and then the function will assume the symmetry and produce the real output array (note that when the input is packed into a real array and inverse transformation is executed, the function treats the input as a packed complex-conjugate symmetrical array, and the output will also be a real array).
 
@@ -997,6 +971,12 @@ All of the above improvements have been implemented in :ocv:func:`matchTemplate`
 
 .. seealso:: :ocv:func:`dct` , :ocv:func:`getOptimalDFTSize` , :ocv:func:`mulSpectrums`, :ocv:func:`filter2D` , :ocv:func:`matchTemplate` , :ocv:func:`flip` , :ocv:func:`cartToPolar` , :ocv:func:`magnitude` , :ocv:func:`phase`
 
+.. note::
+
+   * An example using the discrete fourier transform can be found at opencv_source_code/samples/cpp/dft.cpp
+
+   * (Python) An example using the dft functionality to perform Wiener deconvolution can be found at opencv_source/samples/python2/deconvolution.py
+   * (Python) An example rearranging the quadrants of a Fourier image can be found at opencv_source/samples/python2/dft.py
 
 
 divide
@@ -1011,7 +991,6 @@ Performs per-element division of two arrays or a scalar by an array.
 .. ocv:pyfunction:: cv2.divide(scale, src2[, dst[, dtype]]) -> dst
 
 .. ocv:cfunction:: void cvDiv(const CvArr* src1, const CvArr* src2, CvArr* dst, double scale=1)
-.. ocv:pyoldfunction:: cv.Div(src1, src2, dst, scale=1) -> None
 
     :param src1: first input array.
 
@@ -1058,10 +1037,10 @@ Returns the determinant of a square floating-point matrix.
 
 .. ocv:cfunction:: double cvDet( const CvArr* mat )
 
-.. ocv:pyoldfunction:: cv.Det(mat) -> float
-
     :param mtx: input matrix that must have ``CV_32FC1`` or ``CV_64FC1`` type and square size.
 
+    :param mat: input matrix that must have ``CV_32FC1`` or ``CV_64FC1`` type and square size.
+
 The function ``determinant`` calculates and returns the determinant of the specified matrix. For small matrices ( ``mtx.cols=mtx.rows<=3`` ),
 the direct method is used. For larger matrices, the function uses LU factorization with partial pivoting.
 
@@ -1081,16 +1060,12 @@ eigen
 -----
 Calculates eigenvalues and eigenvectors of a symmetric matrix.
 
-.. ocv:function:: bool eigen(InputArray src, OutputArray eigenvalues, int lowindex=-1, int highindex=-1)
+.. ocv:function:: bool eigen( InputArray src, OutputArray eigenvalues, OutputArray eigenvectors=noArray() )
 
-.. ocv:function:: bool eigen(InputArray src, OutputArray eigenvalues, OutputArray eigenvectors, int lowindex=-1,int highindex=-1)
-
-.. ocv:pyfunction:: cv2.eigen(src, computeEigenvectors[, eigenvalues[, eigenvectors]]) -> retval, eigenvalues, eigenvectors
+.. ocv:pyfunction:: cv2.eigen(src[, eigenvalues[, eigenvectors]]) -> retval, eigenvalues, eigenvectors
 
 .. ocv:cfunction:: void cvEigenVV( CvArr* mat, CvArr* evects, CvArr* evals, double eps=0, int lowindex=-1, int highindex=-1 )
 
-.. ocv:pyoldfunction:: cv.EigenVV(mat, evects, evals, eps, lowindex=-1, highindex=-1)-> None
-
     :param src: input matrix that must have ``CV_32FC1`` or ``CV_64FC1`` type, square size and be symmetrical (``src`` :sup:`T` == ``src``).
 
     :param eigenvalues: output vector of eigenvalues of the same type as ``src``; the eigenvalues are stored in the descending order.
@@ -1120,7 +1095,6 @@ Calculates the exponent of every array element.
 .. ocv:pyfunction:: cv2.exp(src[, dst]) -> dst
 
 .. ocv:cfunction:: void cvExp(const CvArr* src, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Exp(src, dst)-> None
 
     :param src: input array.
 
@@ -1201,8 +1175,6 @@ Flips a 2D array around vertical, horizontal, or both axes.
 
 .. ocv:cfunction:: void cvFlip( const CvArr* src, CvArr* dst=NULL, int flip_mode=0 )
 
-.. ocv:pyoldfunction:: cv.Flip(src, dst=None, flipMode=0)-> None
-
     :param src: input array.
 
     :param dst: output array of the same size and type as ``src``.
@@ -1249,7 +1221,6 @@ Performs generalized matrix multiplication.
 .. ocv:pyfunction:: cv2.gemm(src1, src2, alpha, src3, gamma[, dst[, flags]]) -> dst
 
 .. ocv:cfunction:: void cvGEMM( const CvArr* src1, const CvArr* src2, double alpha, const CvArr* src3, double beta, CvArr* dst, int tABC=0)
-.. ocv:pyoldfunction:: cv.GEMM(src1, src2, alpha, src3, beta, dst, tABC=0)-> None
 
     :param src1: first multiplied input matrix that should have ``CV_32FC1``, ``CV_64FC1``, ``CV_32FC2``, or ``CV_64FC2`` type.
 
@@ -1284,40 +1255,6 @@ The function can be replaced with a matrix expression. For example, the above ca
 
 
 
-getConvertElem
---------------
-Returns a conversion function for a single pixel.
-
-.. ocv:function:: ConvertData getConvertElem(int fromType, int toType)
-
-.. ocv:function:: ConvertScaleData getConvertScaleElem(int fromType, int toType)
-
-    :param fromType: input pixel type.
-
-    :param toType: output pixel type.
-
-    :param from: callback parameter: pointer to the input pixel.
-
-    :param to: callback parameter: pointer to the output pixel
-
-    :param cn: callback parameter: the number of channels; it can be arbitrary, 1, 100, 100000, etc.
-
-    :param alpha: ``ConvertScaleData`` callback optional parameter: the scale factor.
-
-    :param beta: ``ConvertScaleData`` callback optional parameter: the delta or offset.
-
-The functions ``getConvertElem`` and ``getConvertScaleElem`` return pointers to the functions for converting individual pixels from one type to another. While the main function purpose is to convert single pixels (actually, for converting sparse matrices from one type to another), you can use them to convert the whole row of a dense matrix or the whole matrix at once, by setting ``cn = matrix.cols*matrix.rows*matrix.channels()`` if the matrix data is continuous.
-
-``ConvertData`` and ``ConvertScaleData`` are defined as: ::
-
-    typedef void (*ConvertData)(const void* from, void* to, int cn)
-    typedef void (*ConvertScaleData)(const void* from, void* to,
-                                     int cn, double alpha, double beta)
-
-.. seealso:: :ocv:func:`Mat::convertTo` , :ocv:func:`SparseMat::convertTo`
-
-
-
 getOptimalDFTSize
 -----------------
 Returns the optimal DFT size for a given vector size.
@@ -1327,7 +1264,6 @@ Returns the optimal DFT size for a given vector size.
 .. ocv:pyfunction:: cv2.getOptimalDFTSize(vecsize) -> retval
 
 .. ocv:cfunction:: int cvGetOptimalDFTSize(int size0)
-.. ocv:pyoldfunction:: cv.GetOptimalDFTSize(size0)-> int
 
     :param vecsize: vector size.
 
@@ -1411,8 +1347,6 @@ Checks if array elements lie between the elements of two other arrays.
 
 .. ocv:cfunction:: void cvInRange(const CvArr* src, const CvArr* lower, const CvArr* upper, CvArr* dst)
 .. ocv:cfunction:: void cvInRangeS(const CvArr* src, CvScalar lower, CvScalar upper, CvArr* dst)
-.. ocv:pyoldfunction:: cv.InRange(src, lower, upper, dst)-> None
-.. ocv:pyoldfunction:: cv.InRangeS(src, lower, upper, dst)-> None
 
     :param src: first input array.
 
@@ -1453,8 +1387,6 @@ Finds the inverse or pseudo-inverse of a matrix.
 
 .. ocv:cfunction:: double cvInvert( const CvArr* src, CvArr* dst, int method=CV_LU )
 
-.. ocv:pyoldfunction:: cv.Invert(src, dst, method=CV_LU) -> float
-
     :param src: input floating-point ``M x N`` matrix.
 
     :param dst: output matrix of ``N x M`` size and the same type as ``src``.
@@ -1492,7 +1424,6 @@ Calculates the natural logarithm of every array element.
 .. ocv:pyfunction:: cv2.log(src[, dst]) -> dst
 
 .. ocv:cfunction:: void cvLog(const CvArr* src, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Log(src, dst)-> None
 
     :param src: input array.
 
@@ -1523,12 +1454,11 @@ LUT
 ---
 Performs a look-up table transform of an array.
 
-.. ocv:function:: void LUT( InputArray src, InputArray lut, OutputArray dst, int interpolation=0 )
+.. ocv:function:: void LUT( InputArray src, InputArray lut, OutputArray dst )
 
-.. ocv:pyfunction:: cv2.LUT(src, lut[, dst[, interpolation]]) -> dst
+.. ocv:pyfunction:: cv2.LUT(src, lut[, dst]) -> dst
 
 .. ocv:cfunction:: void cvLUT(const CvArr* src, CvArr* dst, const CvArr* lut)
-.. ocv:pyoldfunction:: cv.LUT(src, dst, lut)-> None
 
     :param src: input array of 8-bit elements.
 
@@ -1594,8 +1524,6 @@ Calculates the Mahalanobis distance between two vectors.
 
 .. ocv:cfunction:: double cvMahalanobis( const CvArr* vec1, const CvArr* vec2, const CvArr* mat )
 
-.. ocv:pyoldfunction:: cv.Mahalonobis(vec1, vec2, mat) -> None
-
     :param vec1: first 1D input vector.
 
     :param vec2: second 1D input vector.
@@ -1628,14 +1556,10 @@ Calculates per-element maximum of two arrays or an array and a scalar.
 
 .. ocv:function:: void max(const Mat& src1, const Mat& src2, Mat& dst)
 
-.. ocv:function:: void max( const Mat& src1, double src2, Mat& dst )
-
 .. ocv:pyfunction:: cv2.max(src1, src2[, dst]) -> dst
 
 .. ocv:cfunction:: void cvMax(const CvArr* src1, const CvArr* src2, CvArr* dst)
 .. ocv:cfunction:: void cvMaxS(const CvArr* src, double value, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Max(src1, src2, dst)-> None
-.. ocv:pyoldfunction:: cv.MaxS(src, value, dst)-> None
 
     :param src1: first input array.
 
@@ -1681,8 +1605,6 @@ Calculates an average (mean) of array elements.
 
 .. ocv:cfunction:: CvScalar cvAvg( const CvArr* arr, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.Avg(arr, mask=None) -> scalar
-
     :param src: input array that should have from 1 to 4 channels so that the result can be stored in :ocv:class:`Scalar_` .
 
     :param mask: optional operation mask.
@@ -1714,8 +1636,6 @@ Calculates a mean and standard deviation of array elements.
 
 .. ocv:cfunction:: void cvAvgSdv( const CvArr* arr, CvScalar* mean, CvScalar* std_dev, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.AvgSdv(arr, mask=None) -> (mean, stdDev)
-
     :param src: input array that should have from 1 to 4 channels so that the results can be stored in  :ocv:class:`Scalar_` 's.
 
     :param mean: output parameter: calculated mean value.
@@ -1755,7 +1675,6 @@ Creates one multichannel array out of several single-channel ones.
 .. ocv:pyfunction:: cv2.merge(mv[, dst]) -> dst
 
 .. ocv:cfunction:: void cvMerge(const CvArr* src0, const CvArr* src1, const CvArr* src2, const CvArr* src3, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Merge(src0, src1, src2, src3, dst)-> None
 
     :param mv: input array or vector of matrices to be merged; all the matrices in ``mv`` must have the same size and the same depth.
 
@@ -1791,14 +1710,10 @@ Calculates per-element minimum of two arrays or an array and a scalar.
 
 .. ocv:function:: void min(const Mat& src1, const Mat& src2, Mat& dst)
 
-.. ocv:function:: void min( const Mat& src1, double src2, Mat& dst )
-
 .. ocv:pyfunction:: cv2.min(src1, src2[, dst]) -> dst
 
 .. ocv:cfunction:: void cvMin(const CvArr* src1, const CvArr* src2, CvArr* dst)
 .. ocv:cfunction:: void cvMinS(const CvArr* src, double value, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Min(src1, src2, dst)-> None
-.. ocv:pyoldfunction:: cv.MinS(src, value, dst)-> None
 
     :param src1: first input array.
 
@@ -1838,7 +1753,7 @@ minMaxIdx
 ---------
 Finds the global minimum and maximum in an array
 
-.. ocv:function:: void minMaxIdx(InputArray src, double* minVal, double* maxVal, int* minIdx=0, int* maxIdx=0, InputArray mask=noArray())
+.. ocv:function:: void minMaxIdx(InputArray src, double* minVal, double* maxVal = 0, int* minIdx=0, int* maxIdx=0, InputArray mask=noArray())
 
     :param src: input single-channel array.
 
@@ -1878,8 +1793,6 @@ Finds the global minimum and maximum in an array.
 
 .. ocv:cfunction:: void cvMinMaxLoc( const CvArr* arr, double* min_val, double* max_val, CvPoint* min_loc=NULL, CvPoint* max_loc=NULL, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.MinMaxLoc(arr, mask=None)-> (minVal, maxVal, minLoc, maxLoc)
-
     :param src: input single-channel array.
 
     :param minVal: pointer to the returned minimum value;  ``NULL`` is used if not required.
@@ -1920,14 +1833,14 @@ Copies specified channels from input arrays to the specified channels of output
 
 .. ocv:function:: void mixChannels( const Mat* src, size_t nsrcs, Mat* dst, size_t ndsts, const int* fromTo, size_t npairs )
 
-.. ocv:function:: void mixChannels( const vector<Mat>& src, vector<Mat>& dst, const int* fromTo, size_t npairs )
+.. ocv:function:: void mixChannels( InputArrayOfArrays src, InputOutputArrayOfArrays dst, const int* fromTo, size_t npairs )
 
-.. ocv:pyfunction:: cv2.mixChannels(src, dst, fromTo) -> None
+.. ocv:function:: void mixChannels( InputArrayOfArrays src, InputOutputArrayOfArrays dst, const std::vector<int>& fromTo )
+
+.. ocv:pyfunction:: cv2.mixChannels(src, dst, fromTo) -> dst
 
 .. ocv:cfunction:: void cvMixChannels( const CvArr** src, int src_count, CvArr** dst, int dst_count, const int* from_to, int pair_count )
 
-.. ocv:pyoldfunction:: cv.MixChannels(src, dst, fromTo) -> None
-
     :param src: input array or vector of matricesl; all of the matrices must have the same size and the same depth.
 
     :param nsrcs: number of matrices in ``src``.
@@ -1980,7 +1893,6 @@ Performs the per-element multiplication of two Fourier spectrums.
 .. ocv:pyfunction:: cv2.mulSpectrums(a, b, flags[, c[, conjB]]) -> c
 
 .. ocv:cfunction:: void cvMulSpectrums( const CvArr* src1, const CvArr* src2, CvArr* dst, int flags)
-.. ocv:pyoldfunction:: cv.MulSpectrums(src1, src2, dst, flags)-> None
 
     :param src1: first input array.
 
@@ -2010,7 +1922,6 @@ Calculates the per-element scaled product of two arrays.
 .. ocv:pyfunction:: cv2.multiply(src1, src2[, dst[, scale[, dtype]]]) -> dst
 
 .. ocv:cfunction:: void cvMul(const CvArr* src1, const CvArr* src2, CvArr* dst, double scale=1)
-.. ocv:pyoldfunction:: cv.Mul(src1, src2, dst, scale=1) -> None
 
     :param src1: first input array.
 
@@ -2060,8 +1971,6 @@ Calculates the product of a matrix and its transposition.
 
 .. ocv:cfunction:: void cvMulTransposed( const CvArr* src, CvArr* dst, int order, const CvArr* delta=NULL, double scale=1. )
 
-.. ocv:pyoldfunction:: cv.MulTransposed(src, dst, order, delta=None, scale=1.0) -> None
-
     :param src: input single-channel matrix. Note that unlike :ocv:func:`gemm`, the function can multiply not only floating-point matrices.
 
     :param dst: output square matrix.
@@ -2112,8 +2021,6 @@ Calculates an absolute array norm, an absolute difference norm, or a relative di
 
 .. ocv:cfunction:: double cvNorm( const CvArr* arr1, const CvArr* arr2=NULL, int norm_type=CV_L2, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.Norm(arr1, arr2, normType=CV_L2, mask=None) -> float
-
     :param src1: first input array.
 
     :param src2: second input array of the same size and the same type as ``src1``.
@@ -2260,7 +2167,9 @@ The sample below is the function that takes two matrices. The first function sto
     :ocv:func:`dft`,
     :ocv:func:`dct`
 
+.. note::
 
+   * An example using PCA for dimensionality reduction while maintaining an amount of variance can be found at opencv_source_code/samples/cpp/pca.cpp
 
 PCA::PCA
 --------
@@ -2369,7 +2278,6 @@ Performs the perspective matrix transformation of vectors.
 .. ocv:pyfunction:: cv2.perspectiveTransform(src, m[, dst]) -> dst
 
 .. ocv:cfunction:: void cvPerspectiveTransform(const CvArr* src, CvArr* dst, const CvMat* mat)
-.. ocv:pyoldfunction:: cv.PerspectiveTransform(src, dst, mat)-> None
 
     :param src: input two-channel or three-channel floating-point array; each element is a 2D/3D vector to be transformed.
 
@@ -2443,8 +2351,6 @@ Calculates x and y coordinates of 2D vectors from their magnitude and angle.
 
 .. ocv:cfunction:: void cvPolarToCart( const CvArr* magnitude, const CvArr* angle, CvArr* x, CvArr* y, int angle_in_degrees=0 )
 
-.. ocv:pyoldfunction:: cv.PolarToCart(magnitude, angle, x, y, angleInDegrees=0)-> None
-
     :param magnitude: input floating-point array of magnitudes of 2D vectors; it can be an empty matrix (``=Mat()``), in this case, the function assumes that all the magnitudes are =1; if it is not empty, it must have the same size and type as ``angle``.
 
     :param angle: input floating-point array of angles of 2D vectors.
@@ -2484,7 +2390,6 @@ Raises every array element to a power.
 .. ocv:pyfunction:: cv2.pow(src, power[, dst]) -> dst
 
 .. ocv:cfunction:: void cvPow( const CvArr* src, CvArr* dst, double power)
-.. ocv:pyoldfunction:: cv.Pow(src, dst, power)-> None
 
     :param src: input array.
 
@@ -2752,7 +2657,6 @@ Reduces a matrix to a vector.
 .. ocv:pyfunction:: cv2.reduce(src, dim, rtype[, dst[, dtype]]) -> dst
 
 .. ocv:cfunction:: void cvReduce(const CvArr* src, CvArr* dst, int dim=-1, int op=CV_REDUCE_SUM)
-.. ocv:pyoldfunction:: cv.Reduce(src, dst, dim=-1, op=CV_REDUCE_SUM)-> None
 
     :param src: input 2D matrix.
 
@@ -2790,8 +2694,6 @@ Fills the output array with repeated copies of the input array.
 
 .. ocv:cfunction:: void cvRepeat(const CvArr* src, CvArr* dst)
 
-.. ocv:pyoldfunction:: cv.Repeat(src, dst)-> None
-
     :param src: input array to replicate.
 
     :param dst: output array of the same type as ``src``.
@@ -2826,7 +2728,6 @@ Calculates the sum of a scaled array and another array.
 .. ocv:pyfunction:: cv2.scaleAdd(src1, alpha, src2[, dst]) -> dst
 
 .. ocv:cfunction:: void cvScaleAdd(const CvArr* src1, CvScalar scale, const CvArr* src2, CvArr* dst)
-.. ocv:pyoldfunction:: cv.ScaleAdd(src1, scale, src2, dst)-> None
 
     :param src1: first input array.
 
@@ -2870,8 +2771,6 @@ Initializes a scaled identity matrix.
 
 .. ocv:cfunction:: void cvSetIdentity(CvArr* mat, CvScalar value=cvRealScalar(1))
 
-.. ocv:pyoldfunction:: cv.SetIdentity(mat, value=1)-> None
-
     :param mtx: matrix to initialize (not necessarily square).
 
     :param value: value to assign to diagonal elements.
@@ -2908,7 +2807,6 @@ Solves one or more linear systems or least-squares problems.
 .. ocv:pyfunction:: cv2.solve(src1, src2[, dst[, flags]]) -> retval, dst
 
 .. ocv:cfunction:: int cvSolve(const CvArr* src1, const CvArr* src2, CvArr* dst, int method=CV_LU)
-.. ocv:pyoldfunction:: cv.Solve(A, B, X, method=CV_LU)-> None
 
     :param src1: input matrix on the left-hand side of the system.
 
@@ -2959,8 +2857,6 @@ Finds the real roots of a cubic equation.
 
 .. ocv:cfunction:: int cvSolveCubic( const CvMat* coeffs, CvMat* roots )
 
-.. ocv:pyoldfunction:: cv.SolveCubic(coeffs, roots)-> None
-
     :param coeffs: equation coefficients, an array of 3 or 4 elements.
 
     :param roots: output array of real roots that has 1 or 3 elements.
@@ -3086,8 +2982,6 @@ Divides a multi-channel array into several single-channel arrays.
 
 .. ocv:cfunction:: void cvSplit(const CvArr* src, CvArr* dst0, CvArr* dst1, CvArr* dst2, CvArr* dst3)
 
-.. ocv:pyoldfunction:: cv.Split(src, dst0, dst1, dst2, dst3)-> None
-
     :param src: input multi-channel array.
 
     :param mv: output array or vector of arrays; in the first variant of the function the number of arrays must match ``src.channels()``; the arrays themselves are reallocated, if needed.
@@ -3118,7 +3012,6 @@ Calculates a square root of array elements.
 .. ocv:pyfunction:: cv2.sqrt(src[, dst]) -> dst
 
 .. ocv:cfunction:: float cvSqrt(float value)
-.. ocv:pyoldfunction:: cv.Sqrt(value)-> float
 
     :param src: input floating-point array.
 
@@ -3145,10 +3038,6 @@ Calculates the per-element difference between two arrays or array and a scalar.
 .. ocv:cfunction:: void cvSubRS( const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL )
 .. ocv:cfunction:: void cvSubS( const CvArr* src, CvScalar value, CvArr* dst, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.Sub(src1, src2, dst, mask=None) -> None
-.. ocv:pyoldfunction:: cv.SubRS(src, value, dst, mask=None) -> None
-.. ocv:pyoldfunction:: cv.SubS(src, value, dst, mask=None) -> None
-
     :param src1: first input array or a scalar.
 
     :param src2: second input array or a scalar.
@@ -3281,8 +3170,6 @@ Performs SVD of a matrix
 
 .. ocv:cfunction:: void cvSVD( CvArr* A, CvArr* W, CvArr* U=NULL, CvArr* V=NULL, int flags=0 )
 
-.. ocv:pyoldfunction:: cv.SVD(A, W, U=None, V=None, flags=0) -> None
-
     :param src: decomposed matrix
 
     :param w: calculated singular values
@@ -3331,8 +3218,6 @@ Performs a singular value back substitution.
 
 .. ocv:cfunction:: void cvSVBkSb( const CvArr* W, const CvArr* U, const CvArr* V, const CvArr* B, CvArr* X, int flags )
 
-.. ocv:pyoldfunction:: cv.SVBkSb(W, U, V, B, X, flags) -> None
-
     :param w: singular values
 
     :param u: left singular vectors
@@ -3367,8 +3252,6 @@ Calculates the sum of array elements.
 
 .. ocv:cfunction:: CvScalar cvSum(const CvArr* arr)
 
-.. ocv:pyoldfunction:: cv.Sum(arr) -> scalar
-
     :param arr: input array that must have from 1 to 4 channels.
 
 The functions ``sum`` calculate and return the sum of array elements, independently for each channel.
@@ -3412,8 +3295,6 @@ Returns the trace of a matrix.
 
 .. ocv:cfunction:: CvScalar cvTrace(const CvArr* mat)
 
-.. ocv:pyoldfunction:: cv.Trace(mat) -> scalar
-
     :param mat: input matrix.
 
 The function ``trace`` returns the sum of the diagonal elements of the matrix ``mtx`` .
@@ -3434,8 +3315,6 @@ Performs the matrix transformation of every array element.
 
 .. ocv:cfunction:: void cvTransform( const CvArr* src, CvArr* dst, const CvMat* transmat, const CvMat* shiftvec=NULL )
 
-.. ocv:pyoldfunction:: cv.Transform(src, dst, transmat, shiftvec=None)-> None
-
     :param src: input array that must have as many channels (1 to 4) as ``m.cols`` or ``m.cols-1``.
 
     :param dst: output array of the same size and depth as ``src``; it has as many channels as ``m.rows``.
@@ -3485,7 +3364,6 @@ Transposes a matrix.
 .. ocv:pyfunction:: cv2.transpose(src[, dst]) -> dst
 
 .. ocv:cfunction:: void cvTranspose(const CvArr* src, CvArr* dst)
-.. ocv:pyoldfunction:: cv.Transpose(src, dst)-> None
 
     :param src: input array.
 
@@ -3498,3 +3376,101 @@ The function :ocv:func:`transpose` transposes the matrix ``src`` :
     \texttt{dst} (i,j) =  \texttt{src} (j,i)
 
 .. note:: No complex conjugation is done in case of a complex matrix. It it should be done separately if needed.
+
+
+borderInterpolate
+-----------------
+Computes the source location of an extrapolated pixel.
+
+.. ocv:function:: int borderInterpolate( int p, int len, int borderType )
+
+.. ocv:pyfunction:: cv2.borderInterpolate(p, len, borderType) -> retval
+
+    :param p: 0-based coordinate of the extrapolated pixel along one of the axes,
+              likely <0 or >= ``len`` .
+
+    :param len: Length of the array along the corresponding axis.
+
+    :param borderType: Border type, one of the  ``BORDER_*`` , except for  ``BORDER_TRANSPARENT``
+                       and  ``BORDER_ISOLATED`` . When  ``borderType==BORDER_CONSTANT`` , the
+                       function always returns -1, regardless of  ``p``  and  ``len`` .
+
+The function computes and returns the coordinate of a donor pixel corresponding to the specified
+extrapolated pixel when using the specified extrapolation border mode. For example, if you use
+``BORDER_WRAP`` mode in the horizontal direction, ``BORDER_REFLECT_101`` in the vertical direction
+and want to compute value of the "virtual" pixel ``Point(-5, 100)`` in a floating-point image
+``img`` , it looks like: ::
+
+    float val = img.at<float>(borderInterpolate(100, img.rows, BORDER_REFLECT_101),
+                              borderInterpolate(-5, img.cols, BORDER_WRAP));
+
+
+Normally, the function is not called directly. It is used inside :ocv:class:`FilterEngine`
+and :ocv:func:`copyMakeBorder` to compute tables for quick extrapolation.
+
+.. seealso::
+
+    :ocv:class:`FilterEngine`,
+    :ocv:func:`copyMakeBorder`
+
+
+copyMakeBorder
+--------------
+Forms a border around an image.
+
+.. ocv:function:: void copyMakeBorder( InputArray src, OutputArray dst, int top, int bottom, int left, int right, int borderType, const Scalar& value=Scalar() )
+
+.. ocv:pyfunction:: cv2.copyMakeBorder(src, top, bottom, left, right, borderType[, dst[, value]]) -> dst
+
+    :param src: Source image.
+
+    :param dst: Destination image of the same type as  ``src``  and the
+                size ``Size(src.cols+left+right, src.rows+top+bottom)`` .
+
+    :param top:
+
+    :param bottom:
+
+    :param left:
+
+    :param right: Parameter specifying how many pixels in each direction from the source image
+                  rectangle to extrapolate. For example,  ``top=1, bottom=1, left=1, right=1``
+                  mean that 1 pixel-wide border needs to be built.
+
+    :param borderType: Border type. See  :ocv:func:`borderInterpolate` for details.
+
+    :param value: Border value if  ``borderType==BORDER_CONSTANT`` .
+
+The function copies the source image into the middle of the destination image. The areas to the
+left, to the right, above and below the copied source image will be filled with extrapolated pixels.
+This is not what :ocv:class:`FilterEngine` or filtering functions based on it do (they extrapolate
+pixels on-fly), but what other more complex functions, including your own, may do to simplify image
+boundary handling.
+
+The function supports the mode when ``src`` is already in the middle of ``dst`` . In this case, the
+function does not copy ``src`` itself but simply constructs the border, for example: ::
+
+    // let border be the same in all directions
+    int border=2;
+    // constructs a larger image to fit both the image and the border
+    Mat gray_buf(rgb.rows + border*2, rgb.cols + border*2, rgb.depth());
+    // select the middle part of it w/o copying data
+    Mat gray(gray_canvas, Rect(border, border, rgb.cols, rgb.rows));
+    // convert image from RGB to grayscale
+    cvtColor(rgb, gray, COLOR_RGB2GRAY);
+    // form a border in-place
+    copyMakeBorder(gray, gray_buf, border, border,
+                   border, border, BORDER_REPLICATE);
+    // now do some custom filtering ...
+    ...
+
+
+.. note::
+
+    When the source image is a part (ROI) of a bigger image, the function will try to use the pixels
+    outside of the ROI to form a border. To disable this feature and always do extrapolation, as if
+    ``src`` was not a ROI, use ``borderType | BORDER_ISOLATED``.
+
+.. seealso::
+
+    :ocv:func:`borderInterpolate`
diff --git a/modules/core/doc/utility_and_system_functions_and_macros.rst b/modules/core/doc/utility_and_system_functions_and_macros.rst
index 54198b058a..d5052fa5ac 100644
--- a/modules/core/doc/utility_and_system_functions_and_macros.rst
+++ b/modules/core/doc/utility_and_system_functions_and_macros.rst
@@ -76,7 +76,6 @@ Calculates the angle of a 2D vector in degrees.
 .. ocv:pyfunction:: cv2.fastAtan2(y, x) -> retval
 
 .. ocv:cfunction:: float cvFastArctan(float y, float x)
-.. ocv:pyoldfunction:: cv.FastArctan(y, x)-> float
 
     :param x: x-coordinate of the vector.
 
@@ -95,8 +94,6 @@ Computes the cube root of an argument.
 
 .. ocv:cfunction:: float cvCbrt( float value )
 
-.. ocv:pyoldfunction:: cv.Cbrt(value)-> float
-
     :param val: A function argument.
 
 The function ``cubeRoot`` computes :math:`\sqrt[3]{\texttt{val}}`. Negative arguments are handled correctly. NaN and Inf are not handled. The accuracy approaches the maximum possible accuracy for single-precision data.
@@ -107,7 +104,6 @@ Ceil
 Rounds floating-point number to the nearest integer not smaller than the original.
 
 .. ocv:cfunction:: int cvCeil(double value)
-.. ocv:pyoldfunction:: cv.Ceil(value) -> int
 
     :param value: floating-point number. If the value is outside of ``INT_MIN`` ... ``INT_MAX`` range, the result is not defined.
 
@@ -123,7 +119,6 @@ Floor
 Rounds floating-point number to the nearest integer not larger than the original.
 
 .. ocv:cfunction:: int cvFloor(double value)
-.. ocv:pyoldfunction:: cv.Floor(value) -> int
 
     :param value: floating-point number. If the value is outside of ``INT_MIN`` ... ``INT_MAX`` range, the result is not defined.
 
@@ -139,7 +134,6 @@ Round
 Rounds floating-point number to the nearest integer
 
 .. ocv:cfunction:: int cvRound(double value)
-.. ocv:pyoldfunction:: cv.Round(value) -> int
 
     :param value: floating-point number. If the value is outside of ``INT_MIN`` ... ``INT_MAX`` range, the result is not defined.
 
@@ -149,7 +143,6 @@ IsInf
 Determines if the argument is Infinity.
 
 .. ocv:cfunction:: int cvIsInf(double value)
-.. ocv:pyoldfunction:: cv.IsInf(value)-> int
 
         :param value: The input floating-point value
 
@@ -160,7 +153,6 @@ IsNaN
 Determines if the argument is Not A Number.
 
 .. ocv:cfunction:: int cvIsNaN(double value)
-.. ocv:pyoldfunction:: cv.IsNaN(value)-> int
 
         :param value: The input floating-point value
 
@@ -173,6 +165,8 @@ Checks a condition at runtime and throws exception if it fails
 
 .. ocv:function:: CV_Assert(expr)
 
+    :param expr: Expression for check.
+
 The macros ``CV_Assert`` (and ``CV_DbgAssert``) evaluate the specified expression. If it is 0, the macros raise an error (see :ocv:func:`error` ). The macro ``CV_Assert`` checks the condition in both Debug and Release configurations while ``CV_DbgAssert`` is only retained in the Debug configuration.
 
 
@@ -188,8 +182,14 @@ Signals an error and raises an exception.
 
     :param status: Error code. Normally, it is a negative value. The list of pre-defined error codes can be found in  ``cxerror.h`` .
 
+    :param func_name: The function name where error occurs.
+
     :param err_msg: Text of the error message.
 
+    :param file_name: The file name where error occurs.
+
+    :param line: The line number where error occurs.
+
     :param args: ``printf`` -like formatted error message in parentheses.
 
 The function and the helper macros ``CV_Error`` and ``CV_Error_``: ::
@@ -218,19 +218,19 @@ Exception class passed to an error. ::
     public:
         // various constructors and the copy operation
         Exception() { code = 0; line = 0; }
-        Exception(int _code, const string& _err,
-                  const string& _func, const string& _file, int _line);
+        Exception(int _code, const String& _err,
+                  const String& _func, const String& _file, int _line);
         Exception(const Exception& exc);
         Exception& operator = (const Exception& exc);
 
         // the error code
         int code;
         // the error text message
-        string err;
+        String err;
         // function name where the error happened
-        string func;
+        String func;
         // the source file name where the error happened
-        string file;
+        String file;
         // the source file line where the error happened
         int line;
     };
@@ -249,6 +249,7 @@ Allocates an aligned memory buffer.
 .. ocv:cfunction:: void* cvAlloc( size_t size )
 
     :param size: Allocated buffer size.
+    :param bufSize: Allocated buffer size.
 
 The function allocates the buffer of the specified size and returns it. When the buffer size is 16 bytes or more, the returned buffer is aligned to 16 bytes.
 
@@ -272,7 +273,7 @@ format
 ------
 Returns a text string formatted using the ``printf``\ -like expression.
 
-.. ocv:function:: string format( const char* fmt, ... )
+.. ocv:function:: String format( const char* fmt, ... )
 
     :param fmt: ``printf`` -compatible formatting specifiers.
 
@@ -284,7 +285,7 @@ getBuildInformation
 -------------------
 Returns full configuration time cmake output.
 
-.. ocv:function:: const std::string& getBuildInformation()
+.. ocv:function:: const String& getBuildInformation()
 
 Returned value is raw cmake output including version control system revision, compiler version, compiler flags, enabled modules and third party libraries, etc. Output format depends on target architecture.
 
diff --git a/modules/core/doc/xml_yaml_persistence.rst b/modules/core/doc/xml_yaml_persistence.rst
index 102dafc704..cf3cfb4476 100644
--- a/modules/core/doc/xml_yaml_persistence.rst
+++ b/modules/core/doc/xml_yaml_persistence.rst
@@ -11,7 +11,7 @@ You can store and then restore various OpenCV data structures to/from XML (http:
 
 Use the following procedure to write something to XML or YAML:
  #. Create new :ocv:class:`FileStorage` and open it for writing. It can be done with a single call to :ocv:func:`FileStorage::FileStorage` constructor that takes a filename, or you can use the default constructor and then call :ocv:func:`FileStorage::open`. Format of the file (XML or YAML) is determined from the filename extension (".xml" and ".yml"/".yaml", respectively)
- #. Write all the data you want using the streaming operator ``>>``, just like in the case of STL streams.
+ #. Write all the data you want using the streaming operator ``<<``, just like in the case of STL streams.
  #. Close the file using :ocv:func:`FileStorage::release`. ``FileStorage`` destructor also closes the file.
 
 Here is an example: ::
@@ -91,6 +91,10 @@ Several things can be noted by looking at the sample code and the output:
  *
    In YAML (but not XML), mappings and sequences can be written in a compact Python-like inline form. In the sample above matrix elements, as well as each feature, including its lbp value, is stored in such inline form. To store a mapping/sequence in a compact form, put ":" after the opening character, e.g. use **"{:"** instead of **"{"** and **"[:"** instead of **"["**. When the data is written to XML, those extra ":" are ignored.
 
+.. note::
+
+   * A complete example using the FileStorage interface can be found at opencv_source_code/samples/cpp/filestorage.cpp
+
 
 Reading data from a file storage.
 ---------------------------------
@@ -113,7 +117,7 @@ Here is how to read the file created by the code sample above: ::
     // first method: use (type) operator on FileNode.
     int frameCount = (int)fs2["frameCount"];
 
-    std::string date;
+    String date;
     // second method: use FileNode::operator >>
     fs2["calibrationDate"] >> date;
 
@@ -156,7 +160,7 @@ The constructors.
 
 .. ocv:function:: FileStorage::FileStorage()
 
-.. ocv:function:: FileStorage::FileStorage(const string& source, int flags, const string& encoding=string())
+.. ocv:function:: FileStorage::FileStorage(const String& source, int flags, const String& encoding=String())
 
     :param source: Name of the file to open or the text string to read the data from. Extension of the file (``.xml`` or ``.yml``/``.yaml``) determines its format (XML or YAML respectively). Also you can append ``.gz`` to work with compressed files, for example ``myHugeMatrix.xml.gz``. If both ``FileStorage::WRITE`` and ``FileStorage::MEMORY`` flags are specified, ``source`` is used just to specify the output file format (e.g. ``mydata.xml``, ``.yml`` etc.).
 
@@ -179,7 +183,18 @@ FileStorage::open
 -----------------
 Opens a file.
 
-.. ocv:function:: bool FileStorage::open(const string& filename, int flags, const string& encoding=string())
+.. ocv:function:: bool FileStorage::open(const String& filename, int flags, const String& encoding=String())
+
+    :param filename: Name of the file to open or the text string to read the data from.
+                     Extension of the file (``.xml`` or ``.yml``/``.yaml``) determines its format (XML or YAML respectively).
+                     Also you can append ``.gz`` to work with compressed files, for example ``myHugeMatrix.xml.gz``.
+                     If both ``FileStorage::WRITE`` and ``FileStorage::MEMORY`` flags are specified, ``source``
+                     is used just to specify the output file format (e.g. ``mydata.xml``, ``.yml`` etc.).
+
+    :param flags: Mode of operation. See FileStorage constructor for more details.
+
+    :param encoding: Encoding of the file. Note that UTF-16 XML encoding is not supported currently and you should use 8-bit encoding instead of it.
+
 
 See description of parameters in :ocv:func:`FileStorage::FileStorage`. The method calls :ocv:func:`FileStorage::release` before opening the file.
 
@@ -208,7 +223,7 @@ FileStorage::releaseAndGetString
 --------------------------------
 Closes the file and releases all the memory buffers.
 
-.. ocv:function:: string FileStorage::releaseAndGetString()
+.. ocv:function:: String FileStorage::releaseAndGetString()
 
 Call this method after all I/O operations with the storage are finished. If the storage was opened for writing data and ``FileStorage::WRITE`` was specified
 
@@ -237,7 +252,7 @@ FileStorage::operator[]
 -----------------------
 Returns the specified element of the top-level mapping.
 
-.. ocv:function:: FileNode FileStorage::operator[](const string& nodename) const
+.. ocv:function:: FileNode FileStorage::operator[](const String& nodename) const
 
 .. ocv:function:: FileNode FileStorage::operator[](const char* nodename) const
 
@@ -261,7 +276,7 @@ FileStorage::writeRaw
 ---------------------
 Writes multiple numbers.
 
-.. ocv:function:: void FileStorage::writeRaw( const string& fmt, const uchar* vec, size_t len )
+.. ocv:function:: void FileStorage::writeRaw( const String& fmt, const uchar* vec, size_t len )
 
      :param fmt: Specification of each array element that has the following format  ``([count]{'u'|'c'|'w'|'s'|'i'|'f'|'d'})...`` where the characters correspond to fundamental C++ types:
 
@@ -293,7 +308,7 @@ FileStorage::writeObj
 ---------------------
 Writes the registered C structure (CvMat, CvMatND, CvSeq).
 
-.. ocv:function:: void FileStorage::writeObj( const string& name, const void* obj )
+.. ocv:function:: void FileStorage::writeObj( const String& name, const void* obj )
 
     :param name: Name of the written object.
 
@@ -306,7 +321,7 @@ FileStorage::getDefaultObjectName
 ---------------------------------
 Returns the normalized object name for the specified name of a file.
 
-.. ocv:function:: static string FileStorage::getDefaultObjectName(const string& filename)
+.. ocv:function:: static String FileStorage::getDefaultObjectName(const String& filename)
 
    :param filename: Name of a file
 
@@ -383,7 +398,7 @@ FileNode::operator[]
 --------------------
 Returns element of a mapping node or a sequence node.
 
-.. ocv:function:: FileNode FileNode::operator[](const string& nodename) const
+.. ocv:function:: FileNode FileNode::operator[](const String& nodename) const
 
 .. ocv:function:: FileNode FileNode::operator[](const char* nodename) const
 
@@ -507,7 +522,7 @@ FileNode::name
 --------------
 Returns the node name.
 
-.. ocv:function:: string FileNode::name() const
+.. ocv:function:: String FileNode::name() const
 
     :returns: The node name or an empty string if the node is nameless.
 
@@ -548,11 +563,11 @@ Returns the node content as double.
     :returns: The node content as double.
 
 
-FileNode::operator std::string
+FileNode::operator String
 ------------------------------
 Returns the node content as text string.
 
-.. ocv:function:: FileNode::operator std::string() const
+.. ocv:function:: FileNode::operator String() const
 
     :returns: The node content as a text string.
 
@@ -588,7 +603,7 @@ FileNode::readRaw
 -----------------
 Reads node elements to the buffer with the specified format.
 
-.. ocv:function:: void FileNode::readRaw( const string& fmt, uchar* vec, size_t len ) const
+.. ocv:function:: void FileNode::readRaw( const String& fmt, uchar* vec, size_t len ) const
 
     :param fmt: Specification of each array element. It has the same format as in :ocv:func:`FileStorage::writeRaw`.
 
@@ -692,7 +707,7 @@ FileNodeIterator::readRaw
 -------------------------
 Reads node elements to the buffer with the specified format.
 
-.. ocv:function:: FileNodeIterator& FileNodeIterator::readRaw( const string& fmt, uchar* vec, size_t maxCount=(size_t)INT_MAX )
+.. ocv:function:: FileNodeIterator& FileNodeIterator::readRaw( const String& fmt, uchar* vec, size_t maxCount=(size_t)INT_MAX )
 
     :param fmt: Specification of each array element. It has the same format as in :ocv:func:`FileStorage::writeRaw`.
 
diff --git a/modules/core/include/opencv2/core.hpp b/modules/core/include/opencv2/core.hpp
index 5e204e8411..c7f07ed459 100644
--- a/modules/core/include/opencv2/core.hpp
+++ b/modules/core/include/opencv2/core.hpp
@@ -1,6 +1,3 @@
-/*! \file core.hpp
-    \brief The Core Functionality
- */
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
 //  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
@@ -46,77 +43,25 @@
 #ifndef __OPENCV_CORE_HPP__
 #define __OPENCV_CORE_HPP__
 
+#ifndef __cplusplus
+#  error core.hpp header must be compiled as C++
+#endif
+
 #include "opencv2/core/cvdef.h"
-
-#include "opencv2/core/types_c.h"
 #include "opencv2/core/version.hpp"
-
-#ifdef __cplusplus
-
-#ifndef SKIP_INCLUDES
-#include <limits.h>
-#include <algorithm>
-#include <cmath>
-#include <cstddef>
-#include <complex>
-#include <map>
-#include <new>
-#include <string>
-#include <vector>
-#include <sstream>
-#endif // SKIP_INCLUDES
+#include "opencv2/core/base.hpp"
+#include "opencv2/core/cvstd.hpp"
+#include "opencv2/core/traits.hpp"
+#include "opencv2/core/matx.hpp"
+#include "opencv2/core/types.hpp"
+#include "opencv2/core/mat.hpp"
+#include "opencv2/core/persistence.hpp"
 
 /*! \namespace cv
     Namespace where all the C++ OpenCV functionality resides
 */
 namespace cv {
 
-template<typename _Tp> class CV_EXPORTS Size_;
-template<typename _Tp> class CV_EXPORTS Point_;
-template<typename _Tp> class CV_EXPORTS Rect_;
-template<typename _Tp, int cn> class CV_EXPORTS Vec;
-template<typename _Tp, int m, int n> class CV_EXPORTS Matx;
-
-class Mat;
-class SparseMat;
-typedef Mat MatND;
-
-namespace ogl {
-    class Buffer;
-    class Texture2D;
-    class Arrays;
-}
-
-namespace gpu {
-    class GpuMat;
-}
-
-class CV_EXPORTS MatExpr;
-class CV_EXPORTS MatOp_Base;
-class CV_EXPORTS MatArg;
-class CV_EXPORTS MatConstIterator;
-
-template<typename _Tp> class CV_EXPORTS Mat_;
-template<typename _Tp> class CV_EXPORTS MatIterator_;
-template<typename _Tp> class CV_EXPORTS MatConstIterator_;
-template<typename _Tp> class CV_EXPORTS MatCommaInitializer_;
-
-template<typename _Tp, size_t fixed_size = 1024/sizeof(_Tp)+8> class CV_EXPORTS AutoBuffer;
-
-CV_EXPORTS std::string format( const char* fmt, ... );
-CV_EXPORTS std::string tempfile( const char* suffix CV_DEFAULT(0));
-
-CV_EXPORTS void glob(std::string pattern, std::vector<std::string>& result, bool recursive = false);
-
-// matrix decomposition types
-enum { DECOMP_LU=0, DECOMP_SVD=1, DECOMP_EIG=2, DECOMP_CHOLESKY=3, DECOMP_QR=4, DECOMP_NORMAL=16 };
-enum { NORM_INF=1, NORM_L1=2, NORM_L2=4, NORM_L2SQR=5, NORM_HAMMING=6, NORM_HAMMING2=7, NORM_TYPE_MASK=7, NORM_RELATIVE=8, NORM_MINMAX=32 };
-enum { CMP_EQ=0, CMP_GT=1, CMP_GE=2, CMP_LT=3, CMP_LE=4, CMP_NE=5 };
-enum { GEMM_1_T=1, GEMM_2_T=2, GEMM_3_T=4 };
-enum { DFT_INVERSE=1, DFT_SCALE=2, DFT_ROWS=4, DFT_COMPLEX_OUTPUT=16, DFT_REAL_OUTPUT=32,
-    DCT_INVERSE = DFT_INVERSE, DCT_ROWS=DFT_ROWS };
-
-
 /*!
  The standard OpenCV exception class.
  Instances of the class are thrown by various functions and methods in the case of critical errors.
@@ -132,7 +77,7 @@ public:
      Full constructor. Normally the constuctor is not called explicitly.
      Instead, the macros CV_Error(), CV_Error_() and CV_Assert() are used.
     */
-    Exception(int _code, const std::string& _err, const std::string& _func, const std::string& _file, int _line);
+    Exception(int _code, const String& _err, const String& _func, const String& _file, int _line);
     virtual ~Exception() throw();
 
     /*!
@@ -141,12 +86,12 @@ public:
     virtual const char *what() const throw();
     void formatMessage();
 
-    std::string msg; ///< the formatted error message
+    String msg; ///< the formatted error message
 
     int code; ///< error code @see CVStatus
-    std::string err; ///< error description
-    std::string func; ///< function name. Available only when the compiler supports __func__ macro
-    std::string file; ///< source file name where the error has occured
+    String err; ///< error description
+    String func; ///< function name. Available only when the compiler supports __func__ macro
+    String file; ///< source file name where the error has occured
     int line; ///< line number in the source file where the error has occured
 };
 
@@ -160,2027 +105,179 @@ public:
 
   \param exc the exception raisen.
  */
+//TODO: drop this version
 CV_EXPORTS void error( const Exception& exc );
 
-//! Sets/resets the break-on-error mode.
+
+enum { SORT_EVERY_ROW    = 0,
+       SORT_EVERY_COLUMN = 1,
+       SORT_ASCENDING    = 0,
+       SORT_DESCENDING   = 16
+     };
+
+enum { COVAR_SCRAMBLED = 0,
+       COVAR_NORMAL    = 1,
+       COVAR_USE_AVG   = 2,
+       COVAR_SCALE     = 4,
+       COVAR_ROWS      = 8,
+       COVAR_COLS      = 16
+     };
 
 /*!
-  When the break-on-error mode is set, the default error handler
-  issues a hardware exception, which can make debugging more convenient.
-
-  \return the previous state
- */
-CV_EXPORTS bool setBreakOnError(bool flag);
-
-typedef int (CV_CDECL *ErrorCallback)( int status, const char* func_name,
-                                       const char* err_msg, const char* file_name,
-                                       int line, void* userdata );
-
-//! Sets the new error handler and the optional user data.
-
-/*!
-  The function sets the new error handler, called from cv::error().
-
-  \param errCallback the new error handler. If NULL, the default error handler is used.
-  \param userdata the optional user data pointer, passed to the callback.
-  \param prevUserdata the optional output parameter where the previous user data pointer is stored
-
-  \return the previous error handler
+ k-Means flags
 */
-CV_EXPORTS ErrorCallback redirectError( ErrorCallback errCallback,
-                                        void* userdata=0, void** prevUserdata=0);
+enum { KMEANS_RANDOM_CENTERS     = 0, // Chooses random centers for k-Means initialization
+       KMEANS_PP_CENTERS         = 2, // Uses k-Means++ algorithm for initialization
+       KMEANS_USE_INITIAL_LABELS = 1  // Uses the user-provided labels for K-Means initialization
+     };
+
+enum { FILLED  = -1,
+       LINE_4  = 4,
+       LINE_8  = 8,
+       LINE_AA = 16
+     };
+
+enum { FONT_HERSHEY_SIMPLEX        = 0,
+       FONT_HERSHEY_PLAIN          = 1,
+       FONT_HERSHEY_DUPLEX         = 2,
+       FONT_HERSHEY_COMPLEX        = 3,
+       FONT_HERSHEY_TRIPLEX        = 4,
+       FONT_HERSHEY_COMPLEX_SMALL  = 5,
+       FONT_HERSHEY_SCRIPT_SIMPLEX = 6,
+       FONT_HERSHEY_SCRIPT_COMPLEX = 7,
+       FONT_ITALIC                 = 16
+     };
+
+enum { REDUCE_SUM = 0,
+       REDUCE_AVG = 1,
+       REDUCE_MAX = 2,
+       REDUCE_MIN = 3
+     };
 
-#ifdef __GNUC__
-#define CV_Error( code, msg ) cv::error( cv::Exception(code, msg, __func__, __FILE__, __LINE__) )
-#define CV_Error_( code, args ) cv::error( cv::Exception(code, cv::format args, __func__, __FILE__, __LINE__) )
-#define CV_Assert( expr ) if(!!(expr)) ; else cv::error( cv::Exception(CV_StsAssert, #expr, __func__, __FILE__, __LINE__) )
-#else
-#define CV_Error( code, msg ) cv::error( cv::Exception(code, msg, "", __FILE__, __LINE__) )
-#define CV_Error_( code, args ) cv::error( cv::Exception(code, cv::format args, "", __FILE__, __LINE__) )
-#define CV_Assert( expr ) if(!!(expr)) ; else cv::error( cv::Exception(CV_StsAssert, #expr, "", __FILE__, __LINE__) )
-#endif
-
-#ifdef _DEBUG
-#define CV_DbgAssert(expr) CV_Assert(expr)
-#else
-#define CV_DbgAssert(expr)
-#endif
-
-CV_EXPORTS void setNumThreads(int nthreads);
-CV_EXPORTS int getNumThreads();
-CV_EXPORTS int getThreadNum();
-
-CV_EXPORTS_W const std::string& getBuildInformation();
-
-//! Returns the number of ticks.
-
-/*!
-  The function returns the number of ticks since the certain event (e.g. when the machine was turned on).
-  It can be used to initialize cv::RNG or to measure a function execution time by reading the tick count
-  before and after the function call. The granularity of ticks depends on the hardware and OS used. Use
-  cv::getTickFrequency() to convert ticks to seconds.
-*/
-CV_EXPORTS_W int64 getTickCount();
-
-/*!
-  Returns the number of ticks per seconds.
-
-  The function returns the number of ticks (as returned by cv::getTickCount()) per second.
-  The following code computes the execution time in milliseconds:
-
-  \code
-  double exec_time = (double)getTickCount();
-  // do something ...
-  exec_time = ((double)getTickCount() - exec_time)*1000./getTickFrequency();
-  \endcode
-*/
-CV_EXPORTS_W double getTickFrequency();
-
-/*!
-  Returns the number of CPU ticks.
-
-  On platforms where the feature is available, the function returns the number of CPU ticks
-  since the certain event (normally, the system power-on moment). Using this function
-  one can accurately measure the execution time of very small code fragments,
-  for which cv::getTickCount() granularity is not enough.
-*/
-CV_EXPORTS_W int64 getCPUTickCount();
-
-/*!
-  Returns SSE etc. support status
-
-  The function returns true if certain hardware features are available.
-  Currently, the following features are recognized:
-  - CV_CPU_MMX - MMX
-  - CV_CPU_SSE - SSE
-  - CV_CPU_SSE2 - SSE 2
-  - CV_CPU_SSE3 - SSE 3
-  - CV_CPU_SSSE3 - SSSE 3
-  - CV_CPU_SSE4_1 - SSE 4.1
-  - CV_CPU_SSE4_2 - SSE 4.2
-  - CV_CPU_POPCNT - POPCOUNT
-  - CV_CPU_AVX - AVX
-
-  \note {Note that the function output is not static. Once you called cv::useOptimized(false),
-  most of the hardware acceleration is disabled and thus the function will returns false,
-  until you call cv::useOptimized(true)}
-*/
-CV_EXPORTS_W bool checkHardwareSupport(int feature);
-
-//! returns the number of CPUs (including hyper-threading)
-CV_EXPORTS_W int getNumberOfCPUs();
-
-/*!
-  Allocates memory buffer
-
-  This is specialized OpenCV memory allocation function that returns properly aligned memory buffers.
-  The usage is identical to malloc(). The allocated buffers must be freed with cv::fastFree().
-  If there is not enough memory, the function calls cv::error(), which raises an exception.
-
-  \param bufSize buffer size in bytes
-  \return the allocated memory buffer.
-*/
-CV_EXPORTS void* fastMalloc(size_t bufSize);
-
-/*!
-  Frees the memory allocated with cv::fastMalloc
-
-  This is the corresponding deallocation function for cv::fastMalloc().
-  When ptr==NULL, the function has no effect.
-*/
-CV_EXPORTS void fastFree(void* ptr);
-
-template<typename _Tp> static inline _Tp* allocate(size_t n)
-{
-    return new _Tp[n];
-}
-
-template<typename _Tp> static inline void deallocate(_Tp* ptr, size_t)
-{
-    delete[] ptr;
-}
-
-/*!
-  Aligns pointer by the certain number of bytes
-
-  This small inline function aligns the pointer by the certian number of bytes by shifting
-  it forward by 0 or a positive offset.
-*/
-template<typename _Tp> static inline _Tp* alignPtr(_Tp* ptr, int n=(int)sizeof(_Tp))
-{
-    return (_Tp*)(((size_t)ptr + n-1) & -n);
-}
-
-/*!
-  Aligns buffer size by the certain number of bytes
-
-  This small inline function aligns a buffer size by the certian number of bytes by enlarging it.
-*/
-static inline size_t alignSize(size_t sz, int n)
-{
-    return (sz + n-1) & -n;
-}
-
-/*!
-  Turns on/off available optimization
-
-  The function turns on or off the optimized code in OpenCV. Some optimization can not be enabled
-  or disabled, but, for example, most of SSE code in OpenCV can be temporarily turned on or off this way.
-
-  \note{Since optimization may imply using special data structures, it may be unsafe
-  to call this function anywhere in the code. Instead, call it somewhere at the top level.}
-*/
-CV_EXPORTS_W void setUseOptimized(bool onoff);
-
-/*!
-  Returns the current optimization status
-
-  The function returns the current optimization status, which is controlled by cv::setUseOptimized().
-*/
-CV_EXPORTS_W bool useOptimized();
-
-/*!
-  The STL-compilant memory Allocator based on cv::fastMalloc() and cv::fastFree()
-*/
-template<typename _Tp> class CV_EXPORTS Allocator
-{
-public:
-    typedef _Tp value_type;
-    typedef value_type* pointer;
-    typedef const value_type* const_pointer;
-    typedef value_type& reference;
-    typedef const value_type& const_reference;
-    typedef size_t size_type;
-    typedef ptrdiff_t difference_type;
-    template<typename U> class rebind { typedef Allocator<U> other; };
-
-    explicit Allocator() {}
-    ~Allocator() {}
-    explicit Allocator(Allocator const&) {}
-    template<typename U>
-    explicit Allocator(Allocator<U> const&) {}
-
-    // address
-    pointer address(reference r) { return &r; }
-    const_pointer address(const_reference r) { return &r; }
-
-    pointer allocate(size_type count, const void* =0)
-    { return reinterpret_cast<pointer>(fastMalloc(count * sizeof (_Tp))); }
-
-    void deallocate(pointer p, size_type) {fastFree(p); }
-
-    size_type max_size() const
-    { return max(static_cast<_Tp>(-1)/sizeof(_Tp), 1); }
-
-    void construct(pointer p, const _Tp& v) { new(static_cast<void*>(p)) _Tp(v); }
-    void destroy(pointer p) { p->~_Tp(); }
-};
-
-/////////////////////// Vec (used as element of multi-channel images /////////////////////
-
-/*!
-  A helper class for cv::DataType
-
-  The class is specialized for each fundamental numerical data type supported by OpenCV.
-  It provides DataDepth<T>::value constant.
-*/
-template<typename _Tp> class CV_EXPORTS DataDepth {};
-
-template<> class DataDepth<bool> { public: enum { value = CV_8U, fmt=(int)'u' }; };
-template<> class DataDepth<uchar> { public: enum { value = CV_8U, fmt=(int)'u' }; };
-template<> class DataDepth<schar> { public: enum { value = CV_8S, fmt=(int)'c' }; };
-template<> class DataDepth<char> { public: enum { value = CV_8S, fmt=(int)'c' }; };
-template<> class DataDepth<ushort> { public: enum { value = CV_16U, fmt=(int)'w' }; };
-template<> class DataDepth<short> { public: enum { value = CV_16S, fmt=(int)'s' }; };
-template<> class DataDepth<int> { public: enum { value = CV_32S, fmt=(int)'i' }; };
-// this is temporary solution to support 32-bit unsigned integers
-template<> class DataDepth<unsigned> { public: enum { value = CV_32S, fmt=(int)'i' }; };
-template<> class DataDepth<float> { public: enum { value = CV_32F, fmt=(int)'f' }; };
-template<> class DataDepth<double> { public: enum { value = CV_64F, fmt=(int)'d' }; };
-template<typename _Tp> class DataDepth<_Tp*> { public: enum { value = CV_USRTYPE1, fmt=(int)'r' }; };
-
-
-////////////////////////////// Small Matrix ///////////////////////////
-
-/*!
- A short numerical vector.
-
- This template class represents short numerical vectors (of 1, 2, 3, 4 ... elements)
- on which you can perform basic arithmetical operations, access individual elements using [] operator etc.
- The vectors are allocated on stack, as opposite to std::valarray, std::vector, cv::Mat etc.,
- which elements are dynamically allocated in the heap.
-
- The template takes 2 parameters:
- -# _Tp element type
- -# cn the number of elements
-
- In addition to the universal notation like Vec<float, 3>, you can use shorter aliases
- for the most popular specialized variants of Vec, e.g. Vec3f ~ Vec<float, 3>.
- */
-
-struct CV_EXPORTS Matx_AddOp {};
-struct CV_EXPORTS Matx_SubOp {};
-struct CV_EXPORTS Matx_ScaleOp {};
-struct CV_EXPORTS Matx_MulOp {};
-struct CV_EXPORTS Matx_MatMulOp {};
-struct CV_EXPORTS Matx_TOp {};
-
-template<typename _Tp, int m, int n> class CV_EXPORTS Matx
-{
-public:
-    typedef _Tp value_type;
-    typedef Matx<_Tp, (m < n ? m : n), 1> diag_type;
-    typedef Matx<_Tp, m, n> mat_type;
-    enum { depth = DataDepth<_Tp>::value, rows = m, cols = n, channels = rows*cols,
-           type = CV_MAKETYPE(depth, channels) };
-
-    //! default constructor
-    Matx();
-
-    Matx(_Tp v0); //!< 1x1 matrix
-    Matx(_Tp v0, _Tp v1); //!< 1x2 or 2x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2); //!< 1x3 or 3x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3); //!< 1x4, 2x2 or 4x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4); //!< 1x5 or 5x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5); //!< 1x6, 2x3, 3x2 or 6x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6); //!< 1x7 or 7x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7); //!< 1x8, 2x4, 4x2 or 8x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8); //!< 1x9, 3x3 or 9x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9); //!< 1x10, 2x5 or 5x2 or 10x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-         _Tp v8, _Tp v9, _Tp v10, _Tp v11); //!< 1x12, 2x6, 3x4, 4x3, 6x2 or 12x1 matrix
-    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-         _Tp v8, _Tp v9, _Tp v10, _Tp v11,
-         _Tp v12, _Tp v13, _Tp v14, _Tp v15); //!< 1x16, 4x4 or 16x1 matrix
-    explicit Matx(const _Tp* vals); //!< initialize from a plain array
-
-    static Matx all(_Tp alpha);
-    static Matx zeros();
-    static Matx ones();
-    static Matx eye();
-    static Matx diag(const diag_type& d);
-    static Matx randu(_Tp a, _Tp b);
-    static Matx randn(_Tp a, _Tp b);
-
-    //! dot product computed with the default precision
-    _Tp dot(const Matx<_Tp, m, n>& v) const;
-
-    //! dot product computed in double-precision arithmetics
-    double ddot(const Matx<_Tp, m, n>& v) const;
-
-    //! convertion to another data type
-    template<typename T2> operator Matx<T2, m, n>() const;
-
-    //! change the matrix shape
-    template<int m1, int n1> Matx<_Tp, m1, n1> reshape() const;
-
-    //! extract part of the matrix
-    template<int m1, int n1> Matx<_Tp, m1, n1> get_minor(int i, int j) const;
-
-    //! extract the matrix row
-    Matx<_Tp, 1, n> row(int i) const;
-
-    //! extract the matrix column
-    Matx<_Tp, m, 1> col(int i) const;
-
-    //! extract the matrix diagonal
-    diag_type diag() const;
-
-    //! transpose the matrix
-    Matx<_Tp, n, m> t() const;
-
-    //! invert matrix the matrix
-    Matx<_Tp, n, m> inv(int method=DECOMP_LU) const;
-
-    //! solve linear system
-    template<int l> Matx<_Tp, n, l> solve(const Matx<_Tp, m, l>& rhs, int flags=DECOMP_LU) const;
-    Vec<_Tp, n> solve(const Vec<_Tp, m>& rhs, int method) const;
-
-    //! multiply two matrices element-wise
-    Matx<_Tp, m, n> mul(const Matx<_Tp, m, n>& a) const;
-
-    //! element access
-    const _Tp& operator ()(int i, int j) const;
-    _Tp& operator ()(int i, int j);
-
-    //! 1D element access
-    const _Tp& operator ()(int i) const;
-    _Tp& operator ()(int i);
-
-    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_AddOp);
-    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_SubOp);
-    template<typename _T2> Matx(const Matx<_Tp, m, n>& a, _T2 alpha, Matx_ScaleOp);
-    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_MulOp);
-    template<int l> Matx(const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b, Matx_MatMulOp);
-    Matx(const Matx<_Tp, n, m>& a, Matx_TOp);
-
-    _Tp val[m*n]; //< matrix elements
-};
-
-
-typedef Matx<float, 1, 2> Matx12f;
-typedef Matx<double, 1, 2> Matx12d;
-typedef Matx<float, 1, 3> Matx13f;
-typedef Matx<double, 1, 3> Matx13d;
-typedef Matx<float, 1, 4> Matx14f;
-typedef Matx<double, 1, 4> Matx14d;
-typedef Matx<float, 1, 6> Matx16f;
-typedef Matx<double, 1, 6> Matx16d;
-
-typedef Matx<float, 2, 1> Matx21f;
-typedef Matx<double, 2, 1> Matx21d;
-typedef Matx<float, 3, 1> Matx31f;
-typedef Matx<double, 3, 1> Matx31d;
-typedef Matx<float, 4, 1> Matx41f;
-typedef Matx<double, 4, 1> Matx41d;
-typedef Matx<float, 6, 1> Matx61f;
-typedef Matx<double, 6, 1> Matx61d;
-
-typedef Matx<float, 2, 2> Matx22f;
-typedef Matx<double, 2, 2> Matx22d;
-typedef Matx<float, 2, 3> Matx23f;
-typedef Matx<double, 2, 3> Matx23d;
-typedef Matx<float, 3, 2> Matx32f;
-typedef Matx<double, 3, 2> Matx32d;
-
-typedef Matx<float, 3, 3> Matx33f;
-typedef Matx<double, 3, 3> Matx33d;
-
-typedef Matx<float, 3, 4> Matx34f;
-typedef Matx<double, 3, 4> Matx34d;
-typedef Matx<float, 4, 3> Matx43f;
-typedef Matx<double, 4, 3> Matx43d;
-
-typedef Matx<float, 4, 4> Matx44f;
-typedef Matx<double, 4, 4> Matx44d;
-typedef Matx<float, 6, 6> Matx66f;
-typedef Matx<double, 6, 6> Matx66d;
-
-
-/*!
-  A short numerical vector.
-
-  This template class represents short numerical vectors (of 1, 2, 3, 4 ... elements)
-  on which you can perform basic arithmetical operations, access individual elements using [] operator etc.
-  The vectors are allocated on stack, as opposite to std::valarray, std::vector, cv::Mat etc.,
-  which elements are dynamically allocated in the heap.
-
-  The template takes 2 parameters:
-  -# _Tp element type
-  -# cn the number of elements
-
-  In addition to the universal notation like Vec<float, 3>, you can use shorter aliases
-  for the most popular specialized variants of Vec, e.g. Vec3f ~ Vec<float, 3>.
-*/
-template<typename _Tp, int cn> class CV_EXPORTS Vec : public Matx<_Tp, cn, 1>
-{
-public:
-    typedef _Tp value_type;
-    enum { depth = DataDepth<_Tp>::value, channels = cn, type = CV_MAKETYPE(depth, channels) };
-
-    //! default constructor
-    Vec();
-
-    Vec(_Tp v0); //!< 1-element vector constructor
-    Vec(_Tp v0, _Tp v1); //!< 2-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2); //!< 3-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3); //!< 4-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4); //!< 5-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5); //!< 6-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6); //!< 7-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7); //!< 8-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8); //!< 9-element vector constructor
-    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9); //!< 10-element vector constructor
-    explicit Vec(const _Tp* values);
-
-    Vec(const Vec<_Tp, cn>& v);
-
-    static Vec all(_Tp alpha);
-
-    //! per-element multiplication
-    Vec mul(const Vec<_Tp, cn>& v) const;
-
-    //! conjugation (makes sense for complex numbers and quaternions)
-    Vec conj() const;
-
-    /*!
-      cross product of the two 3D vectors.
-
-      For other dimensionalities the exception is raised
-    */
-    Vec cross(const Vec& v) const;
-    //! convertion to another data type
-    template<typename T2> operator Vec<T2, cn>() const;
-    //! conversion to 4-element CvScalar.
-    operator CvScalar() const;
-
-    /*! element access */
-    const _Tp& operator [](int i) const;
-    _Tp& operator[](int i);
-    const _Tp& operator ()(int i) const;
-    _Tp& operator ()(int i);
-
-    Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_AddOp);
-    Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_SubOp);
-    template<typename _T2> Vec(const Matx<_Tp, cn, 1>& a, _T2 alpha, Matx_ScaleOp);
-};
-
-
-/* \typedef
-
-   Shorter aliases for the most popular specializations of Vec<T,n>
-*/
-typedef Vec<uchar, 2> Vec2b;
-typedef Vec<uchar, 3> Vec3b;
-typedef Vec<uchar, 4> Vec4b;
-
-typedef Vec<short, 2> Vec2s;
-typedef Vec<short, 3> Vec3s;
-typedef Vec<short, 4> Vec4s;
-
-typedef Vec<ushort, 2> Vec2w;
-typedef Vec<ushort, 3> Vec3w;
-typedef Vec<ushort, 4> Vec4w;
-
-typedef Vec<int, 2> Vec2i;
-typedef Vec<int, 3> Vec3i;
-typedef Vec<int, 4> Vec4i;
-typedef Vec<int, 6> Vec6i;
-typedef Vec<int, 8> Vec8i;
-
-typedef Vec<float, 2> Vec2f;
-typedef Vec<float, 3> Vec3f;
-typedef Vec<float, 4> Vec4f;
-typedef Vec<float, 6> Vec6f;
-
-typedef Vec<double, 2> Vec2d;
-typedef Vec<double, 3> Vec3d;
-typedef Vec<double, 4> Vec4d;
-typedef Vec<double, 6> Vec6d;
-
-
-//////////////////////////////// Complex //////////////////////////////
-
-/*!
-  A complex number class.
-
-  The template class is similar and compatible with std::complex, however it provides slightly
-  more convenient access to the real and imaginary parts using through the simple field access, as opposite
-  to std::complex::real() and std::complex::imag().
-*/
-template<typename _Tp> class CV_EXPORTS Complex
-{
-public:
-
-    //! constructors
-    Complex();
-    Complex( _Tp _re, _Tp _im=0 );
-    Complex( const std::complex<_Tp>& c );
-
-    //! conversion to another data type
-    template<typename T2> operator Complex<T2>() const;
-    //! conjugation
-    Complex conj() const;
-    //! conversion to std::complex
-    operator std::complex<_Tp>() const;
-
-    _Tp re, im; //< the real and the imaginary parts
-};
-
-
-/*!
-  \typedef
-*/
-typedef Complex<float> Complexf;
-typedef Complex<double> Complexd;
-
-
-//////////////////////////////// Point_ ////////////////////////////////
-
-/*!
-  template 2D point class.
-
-  The class defines a point in 2D space. Data type of the point coordinates is specified
-  as a template parameter. There are a few shorter aliases available for user convenience.
-  See cv::Point, cv::Point2i, cv::Point2f and cv::Point2d.
-*/
-template<typename _Tp> class CV_EXPORTS Point_
-{
-public:
-    typedef _Tp value_type;
-
-    // various constructors
-    Point_();
-    Point_(_Tp _x, _Tp _y);
-    Point_(const Point_& pt);
-    Point_(const CvPoint& pt);
-    Point_(const CvPoint2D32f& pt);
-    Point_(const Size_<_Tp>& sz);
-    Point_(const Vec<_Tp, 2>& v);
-
-    Point_& operator = (const Point_& pt);
-    //! conversion to another data type
-    template<typename _Tp2> operator Point_<_Tp2>() const;
-
-    //! conversion to the old-style C structures
-    operator CvPoint() const;
-    operator CvPoint2D32f() const;
-    operator Vec<_Tp, 2>() const;
-
-    //! dot product
-    _Tp dot(const Point_& pt) const;
-    //! dot product computed in double-precision arithmetics
-    double ddot(const Point_& pt) const;
-    //! cross-product
-    double cross(const Point_& pt) const;
-    //! checks whether the point is inside the specified rectangle
-    bool inside(const Rect_<_Tp>& r) const;
-
-    _Tp x, y; //< the point coordinates
-};
-
-/*!
-  template 3D point class.
-
-  The class defines a point in 3D space. Data type of the point coordinates is specified
-  as a template parameter.
-
-  \see cv::Point3i, cv::Point3f and cv::Point3d
-*/
-template<typename _Tp> class CV_EXPORTS Point3_
-{
-public:
-    typedef _Tp value_type;
-
-    // various constructors
-    Point3_();
-    Point3_(_Tp _x, _Tp _y, _Tp _z);
-    Point3_(const Point3_& pt);
-    explicit Point3_(const Point_<_Tp>& pt);
-    Point3_(const CvPoint3D32f& pt);
-    Point3_(const Vec<_Tp, 3>& v);
-
-    Point3_& operator = (const Point3_& pt);
-    //! conversion to another data type
-    template<typename _Tp2> operator Point3_<_Tp2>() const;
-    //! conversion to the old-style CvPoint...
-    operator CvPoint3D32f() const;
-    //! conversion to cv::Vec<>
-    operator Vec<_Tp, 3>() const;
-
-    //! dot product
-    _Tp dot(const Point3_& pt) const;
-    //! dot product computed in double-precision arithmetics
-    double ddot(const Point3_& pt) const;
-    //! cross product of the 2 3D points
-    Point3_ cross(const Point3_& pt) const;
-
-    _Tp x, y, z; //< the point coordinates
-};
-
-//////////////////////////////// Size_ ////////////////////////////////
-
-/*!
-  The 2D size class
-
-  The class represents the size of a 2D rectangle, image size, matrix size etc.
-  Normally, cv::Size ~ cv::Size_<int> is used.
-*/
-template<typename _Tp> class CV_EXPORTS Size_
-{
-public:
-    typedef _Tp value_type;
-
-    //! various constructors
-    Size_();
-    Size_(_Tp _width, _Tp _height);
-    Size_(const Size_& sz);
-    Size_(const CvSize& sz);
-    Size_(const CvSize2D32f& sz);
-    Size_(const Point_<_Tp>& pt);
-
-    Size_& operator = (const Size_& sz);
-    //! the area (width*height)
-    _Tp area() const;
-
-    //! conversion of another data type.
-    template<typename _Tp2> operator Size_<_Tp2>() const;
-
-    //! conversion to the old-style OpenCV types
-    operator CvSize() const;
-    operator CvSize2D32f() const;
-
-    _Tp width, height; // the width and the height
-};
-
-//////////////////////////////// Rect_ ////////////////////////////////
-
-/*!
-  The 2D up-right rectangle class
-
-  The class represents a 2D rectangle with coordinates of the specified data type.
-  Normally, cv::Rect ~ cv::Rect_<int> is used.
-*/
-template<typename _Tp> class CV_EXPORTS Rect_
-{
-public:
-    typedef _Tp value_type;
-
-    //! various constructors
-    Rect_();
-    Rect_(_Tp _x, _Tp _y, _Tp _width, _Tp _height);
-    Rect_(const Rect_& r);
-    Rect_(const CvRect& r);
-    Rect_(const Point_<_Tp>& org, const Size_<_Tp>& sz);
-    Rect_(const Point_<_Tp>& pt1, const Point_<_Tp>& pt2);
-
-    Rect_& operator = ( const Rect_& r );
-    //! the top-left corner
-    Point_<_Tp> tl() const;
-    //! the bottom-right corner
-    Point_<_Tp> br() const;
-
-    //! size (width, height) of the rectangle
-    Size_<_Tp> size() const;
-    //! area (width*height) of the rectangle
-    _Tp area() const;
-
-    //! conversion to another data type
-    template<typename _Tp2> operator Rect_<_Tp2>() const;
-    //! conversion to the old-style CvRect
-    operator CvRect() const;
-
-    //! checks whether the rectangle contains the point
-    bool contains(const Point_<_Tp>& pt) const;
-
-    _Tp x, y, width, height; //< the top-left corner, as well as width and height of the rectangle
-};
-
-
-/*!
-  \typedef
-
-  shorter aliases for the most popular cv::Point_<>, cv::Size_<> and cv::Rect_<> specializations
-*/
-typedef Point_<int> Point2i;
-typedef Point2i Point;
-typedef Size_<int> Size2i;
-typedef Size2i Size;
-typedef Rect_<int> Rect;
-typedef Point_<float> Point2f;
-typedef Point_<double> Point2d;
-typedef Size_<float> Size2f;
-typedef Point3_<int> Point3i;
-typedef Point3_<float> Point3f;
-typedef Point3_<double> Point3d;
-
-
-/*!
-  The rotated 2D rectangle.
-
-  The class represents rotated (i.e. not up-right) rectangles on a plane.
-  Each rectangle is described by the center point (mass center), length of each side
-  (represented by cv::Size2f structure) and the rotation angle in degrees.
-*/
-class CV_EXPORTS RotatedRect
-{
-public:
-    //! various constructors
-    RotatedRect();
-    RotatedRect(const Point2f& center, const Size2f& size, float angle);
-    RotatedRect(const CvBox2D& box);
-
-    //! returns 4 vertices of the rectangle
-    void points(Point2f pts[]) const;
-    //! returns the minimal up-right rectangle containing the rotated rectangle
-    Rect boundingRect() const;
-    //! conversion to the old-style CvBox2D structure
-    operator CvBox2D() const;
-
-    Point2f center; //< the rectangle mass center
-    Size2f size;    //< width and height of the rectangle
-    float angle;    //< the rotation angle. When the angle is 0, 90, 180, 270 etc., the rectangle becomes an up-right rectangle.
-};
-
-//////////////////////////////// Scalar_ ///////////////////////////////
-
-/*!
-   The template scalar class.
-
-   This is partially specialized cv::Vec class with the number of elements = 4, i.e. a short vector of four elements.
-   Normally, cv::Scalar ~ cv::Scalar_<double> is used.
-*/
-template<typename _Tp> class CV_EXPORTS Scalar_ : public Vec<_Tp, 4>
-{
-public:
-    //! various constructors
-    Scalar_();
-    Scalar_(_Tp v0, _Tp v1, _Tp v2=0, _Tp v3=0);
-    Scalar_(const CvScalar& s);
-    Scalar_(_Tp v0);
-
-    //! returns a scalar with all elements set to v0
-    static Scalar_<_Tp> all(_Tp v0);
-    //! conversion to the old-style CvScalar
-    operator CvScalar() const;
-
-    //! conversion to another data type
-    template<typename T2> operator Scalar_<T2>() const;
-
-    //! per-element product
-    Scalar_<_Tp> mul(const Scalar_<_Tp>& t, double scale=1 ) const;
-
-    // returns (v0, -v1, -v2, -v3)
-    Scalar_<_Tp> conj() const;
-
-    // returns true iff v1 == v2 == v3 == 0
-    bool isReal() const;
-};
-
-typedef Scalar_<double> Scalar;
-
-CV_EXPORTS void scalarToRawData(const Scalar& s, void* buf, int type, int unroll_to=0);
-
-//////////////////////////////// Range /////////////////////////////////
-
-/*!
-   The 2D range class
-
-   This is the class used to specify a continuous subsequence, i.e. part of a contour, or a column span in a matrix.
-*/
-class CV_EXPORTS Range
-{
-public:
-    Range();
-    Range(int _start, int _end);
-    Range(const CvSlice& slice);
-    int size() const;
-    bool empty() const;
-    static Range all();
-    operator CvSlice() const;
-
-    int start, end;
-};
-
-/////////////////////////////// DataType ////////////////////////////////
-
-/*!
-   Informative template class for OpenCV "scalars".
-
-   The class is specialized for each primitive numerical type supported by OpenCV (such as unsigned char or float),
-   as well as for more complex types, like cv::Complex<>, std::complex<>, cv::Vec<> etc.
-   The common property of all such types (called "scalars", do not confuse it with cv::Scalar_)
-   is that each of them is basically a tuple of numbers of the same type. Each "scalar" can be represented
-   by the depth id (CV_8U ... CV_64F) and the number of channels.
-   OpenCV matrices, 2D or nD, dense or sparse, can store "scalars",
-   as long as the number of channels does not exceed CV_CN_MAX.
-*/
-template<typename _Tp> class DataType
-{
-public:
-    typedef _Tp value_type;
-    typedef value_type work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 1, depth = -1, channels = 1, fmt=0,
-        type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<bool>
-{
-public:
-    typedef bool value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<uchar>
-{
-public:
-    typedef uchar value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<schar>
-{
-public:
-    typedef schar value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<char>
-{
-public:
-    typedef schar value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<ushort>
-{
-public:
-    typedef ushort value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<short>
-{
-public:
-    typedef short value_type;
-    typedef int work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<int>
-{
-public:
-    typedef int value_type;
-    typedef value_type work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<float>
-{
-public:
-    typedef float value_type;
-    typedef value_type work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<> class DataType<double>
-{
-public:
-    typedef double value_type;
-    typedef value_type work_type;
-    typedef value_type channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 1,
-           fmt=DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<typename _Tp, int m, int n> class DataType<Matx<_Tp, m, n> >
-{
-public:
-    typedef Matx<_Tp, m, n> value_type;
-    typedef Matx<typename DataType<_Tp>::work_type, m, n> work_type;
-    typedef _Tp channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = m*n,
-        fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-        type = CV_MAKETYPE(depth, channels) };
-};
-
-template<typename _Tp, int cn> class DataType<Vec<_Tp, cn> >
-{
-public:
-    typedef Vec<_Tp, cn> value_type;
-    typedef Vec<typename DataType<_Tp>::work_type, cn> work_type;
-    typedef _Tp channel_type;
-    typedef value_type vec_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = cn,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-};
-
-template<typename _Tp> class DataType<std::complex<_Tp> >
-{
-public:
-    typedef std::complex<_Tp> value_type;
-    typedef value_type work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 2,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Complex<_Tp> >
-{
-public:
-    typedef Complex<_Tp> value_type;
-    typedef value_type work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 2,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Point_<_Tp> >
-{
-public:
-    typedef Point_<_Tp> value_type;
-    typedef Point_<typename DataType<_Tp>::work_type> work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 2,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Point3_<_Tp> >
-{
-public:
-    typedef Point3_<_Tp> value_type;
-    typedef Point3_<typename DataType<_Tp>::work_type> work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 3,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Size_<_Tp> >
-{
-public:
-    typedef Size_<_Tp> value_type;
-    typedef Size_<typename DataType<_Tp>::work_type> work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 2,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Rect_<_Tp> >
-{
-public:
-    typedef Rect_<_Tp> value_type;
-    typedef Rect_<typename DataType<_Tp>::work_type> work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 4,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<typename _Tp> class DataType<Scalar_<_Tp> >
-{
-public:
-    typedef Scalar_<_Tp> value_type;
-    typedef Scalar_<typename DataType<_Tp>::work_type> work_type;
-    typedef _Tp channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 4,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-template<> class DataType<Range>
-{
-public:
-    typedef Range value_type;
-    typedef value_type work_type;
-    typedef int channel_type;
-    enum { generic_type = 0, depth = DataDepth<channel_type>::value, channels = 2,
-           fmt = ((channels-1)<<8) + DataDepth<channel_type>::fmt,
-           type = CV_MAKETYPE(depth, channels) };
-    typedef Vec<channel_type, channels> vec_type;
-};
-
-//////////////////// generic_type ref-counting pointer class for C/C++ objects ////////////////////////
-
-/*!
-  Smart pointer to dynamically allocated objects.
-
-  This is template pointer-wrapping class that stores the associated reference counter along with the
-  object pointer. The class is similar to std::smart_ptr<> from the recent addons to the C++ standard,
-  but is shorter to write :) and self-contained (i.e. does add any dependency on the compiler or an external library).
-
-  Basically, you can use "Ptr<MyObjectType> ptr" (or faster "const Ptr<MyObjectType>& ptr" for read-only access)
-  everywhere instead of "MyObjectType* ptr", where MyObjectType is some C structure or a C++ class.
-  To make it all work, you need to specialize Ptr<>::delete_obj(), like:
-
-  \code
-  template<> void Ptr<MyObjectType>::delete_obj() { call_destructor_func(obj); }
-  \endcode
-
-  \note{if MyObjectType is a C++ class with a destructor, you do not need to specialize delete_obj(),
-  since the default implementation calls "delete obj;"}
-
-  \note{Another good property of the class is that the operations on the reference counter are atomic,
-  i.e. it is safe to use the class in multi-threaded applications}
-*/
-template<typename _Tp> class CV_EXPORTS Ptr
-{
-public:
-    //! empty constructor
-    Ptr();
-    //! take ownership of the pointer. The associated reference counter is allocated and set to 1
-    Ptr(_Tp* _obj);
-    //! calls release()
-    ~Ptr();
-    //! copy constructor. Copies the members and calls addref()
-    Ptr(const Ptr& ptr);
-    template<typename _Tp2> Ptr(const Ptr<_Tp2>& ptr);
-    //! copy operator. Calls ptr.addref() and release() before copying the members
-    Ptr& operator = (const Ptr& ptr);
-    //! increments the reference counter
-    void addref();
-    //! decrements the reference counter. If it reaches 0, delete_obj() is called
-    void release();
-    //! deletes the object. Override if needed
-    void delete_obj();
-    //! returns true iff obj==NULL
-    bool empty() const;
-
-    //! cast pointer to another type
-    template<typename _Tp2> Ptr<_Tp2> ptr();
-    template<typename _Tp2> const Ptr<_Tp2> ptr() const;
-
-    //! helper operators making "Ptr<T> ptr" use very similar to "T* ptr".
-    _Tp* operator -> ();
-    const _Tp* operator -> () const;
-
-    operator _Tp* ();
-    operator const _Tp*() const;
-
-    _Tp* obj; //< the object pointer.
-    int* refcount; //< the associated reference counter
-};
-
-template<class T, class U> bool operator==(Ptr<T> const & a, Ptr<U> const & b);
-template<class T, class U> bool operator!=(Ptr<T> const & a, Ptr<U> const & b);
-
-
-//////////////////////// Input/Output Array Arguments /////////////////////////////////
-
-/*!
- Proxy datatype for passing Mat's and vector<>'s as input parameters
- */
-class CV_EXPORTS _InputArray
-{
-public:
-    enum {
-        KIND_SHIFT = 16,
-        FIXED_TYPE = 0x8000 << KIND_SHIFT,
-        FIXED_SIZE = 0x4000 << KIND_SHIFT,
-        KIND_MASK = ~(FIXED_TYPE|FIXED_SIZE) - (1 << KIND_SHIFT) + 1,
-
-        NONE              = 0 << KIND_SHIFT,
-        MAT               = 1 << KIND_SHIFT,
-        MATX              = 2 << KIND_SHIFT,
-        STD_VECTOR        = 3 << KIND_SHIFT,
-        STD_VECTOR_VECTOR = 4 << KIND_SHIFT,
-        STD_VECTOR_MAT    = 5 << KIND_SHIFT,
-        EXPR              = 6 << KIND_SHIFT,
-        OPENGL_BUFFER     = 7 << KIND_SHIFT,
-        OPENGL_TEXTURE    = 8 << KIND_SHIFT,
-        GPU_MAT           = 9 << KIND_SHIFT
-    };
-    _InputArray();
-
-    _InputArray(const Mat& m);
-    _InputArray(const MatExpr& expr);
-    template<typename _Tp> _InputArray(const _Tp* vec, int n);
-    template<typename _Tp> _InputArray(const std::vector<_Tp>& vec);
-    template<typename _Tp> _InputArray(const std::vector<std::vector<_Tp> >& vec);
-    _InputArray(const std::vector<Mat>& vec);
-    template<typename _Tp> _InputArray(const std::vector<Mat_<_Tp> >& vec);
-    template<typename _Tp> _InputArray(const Mat_<_Tp>& m);
-    template<typename _Tp, int m, int n> _InputArray(const Matx<_Tp, m, n>& matx);
-    _InputArray(const Scalar& s);
-    _InputArray(const double& val);
-    _InputArray(const gpu::GpuMat& d_mat);
-    _InputArray(const ogl::Buffer& buf);
-    _InputArray(const ogl::Texture2D& tex);
-
-    virtual Mat getMat(int i=-1) const;
-    virtual void getMatVector(std::vector<Mat>& mv) const;
-    virtual gpu::GpuMat getGpuMat() const;
-    virtual ogl::Buffer getOGlBuffer() const;
-    virtual ogl::Texture2D getOGlTexture2D() const;
-
-    virtual int kind() const;
-    virtual Size size(int i=-1) const;
-    virtual size_t total(int i=-1) const;
-    virtual int type(int i=-1) const;
-    virtual int depth(int i=-1) const;
-    virtual int channels(int i=-1) const;
-    virtual bool empty() const;
-
-    virtual ~_InputArray();
-
-    int flags;
-    void* obj;
-    Size sz;
-};
-
-
-enum
-{
-    DEPTH_MASK_8U = 1 << CV_8U,
-    DEPTH_MASK_8S = 1 << CV_8S,
-    DEPTH_MASK_16U = 1 << CV_16U,
-    DEPTH_MASK_16S = 1 << CV_16S,
-    DEPTH_MASK_32S = 1 << CV_32S,
-    DEPTH_MASK_32F = 1 << CV_32F,
-    DEPTH_MASK_64F = 1 << CV_64F,
-    DEPTH_MASK_ALL = (DEPTH_MASK_64F<<1)-1,
-    DEPTH_MASK_ALL_BUT_8S = DEPTH_MASK_ALL & ~DEPTH_MASK_8S,
-    DEPTH_MASK_FLT = DEPTH_MASK_32F + DEPTH_MASK_64F
-};
-
-
-/*!
- Proxy datatype for passing Mat's and vector<>'s as input parameters
- */
-class CV_EXPORTS _OutputArray : public _InputArray
-{
-public:
-    _OutputArray();
-
-    _OutputArray(Mat& m);
-    template<typename _Tp> _OutputArray(std::vector<_Tp>& vec);
-    template<typename _Tp> _OutputArray(std::vector<std::vector<_Tp> >& vec);
-    _OutputArray(std::vector<Mat>& vec);
-    template<typename _Tp> _OutputArray(std::vector<Mat_<_Tp> >& vec);
-    template<typename _Tp> _OutputArray(Mat_<_Tp>& m);
-    template<typename _Tp, int m, int n> _OutputArray(Matx<_Tp, m, n>& matx);
-    template<typename _Tp> _OutputArray(_Tp* vec, int n);
-    _OutputArray(gpu::GpuMat& d_mat);
-    _OutputArray(ogl::Buffer& buf);
-    _OutputArray(ogl::Texture2D& tex);
-
-    _OutputArray(const Mat& m);
-    template<typename _Tp> _OutputArray(const std::vector<_Tp>& vec);
-    template<typename _Tp> _OutputArray(const std::vector<std::vector<_Tp> >& vec);
-    _OutputArray(const std::vector<Mat>& vec);
-    template<typename _Tp> _OutputArray(const std::vector<Mat_<_Tp> >& vec);
-    template<typename _Tp> _OutputArray(const Mat_<_Tp>& m);
-    template<typename _Tp, int m, int n> _OutputArray(const Matx<_Tp, m, n>& matx);
-    template<typename _Tp> _OutputArray(const _Tp* vec, int n);
-    _OutputArray(const gpu::GpuMat& d_mat);
-    _OutputArray(const ogl::Buffer& buf);
-    _OutputArray(const ogl::Texture2D& tex);
-
-    virtual bool fixedSize() const;
-    virtual bool fixedType() const;
-    virtual bool needed() const;
-    virtual Mat& getMatRef(int i=-1) const;
-    virtual gpu::GpuMat& getGpuMatRef() const;
-    virtual ogl::Buffer& getOGlBufferRef() const;
-    virtual ogl::Texture2D& getOGlTexture2DRef() const;
-    virtual void create(Size sz, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
-    virtual void create(int rows, int cols, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
-    virtual void create(int dims, const int* size, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
-    virtual void release() const;
-    virtual void clear() const;
-
-    virtual ~_OutputArray();
-};
-
-typedef const _InputArray& InputArray;
-typedef InputArray InputArrayOfArrays;
-typedef const _OutputArray& OutputArray;
-typedef OutputArray OutputArrayOfArrays;
-typedef OutputArray InputOutputArray;
-typedef OutputArray InputOutputArrayOfArrays;
-
-CV_EXPORTS OutputArray noArray();
-
-/////////////////////////////////////// Mat ///////////////////////////////////////////
-
-enum { MAGIC_MASK=0xFFFF0000, TYPE_MASK=0x00000FFF, DEPTH_MASK=7 };
-
-static inline size_t getElemSize(int type) { return CV_ELEM_SIZE(type); }
-
-/*!
-   Custom array allocator
-
-*/
-class CV_EXPORTS MatAllocator
-{
-public:
-    MatAllocator() {}
-    virtual ~MatAllocator() {}
-    virtual void allocate(int dims, const int* sizes, int type, int*& refcount,
-                          uchar*& datastart, uchar*& data, size_t* step) = 0;
-    virtual void deallocate(int* refcount, uchar* datastart, uchar* data) = 0;
-};
-
-/*!
-   The n-dimensional matrix class.
-
-   The class represents an n-dimensional dense numerical array that can act as
-   a matrix, image, optical flow map, 3-focal tensor etc.
-   It is very similar to CvMat and CvMatND types from earlier versions of OpenCV,
-   and similarly to those types, the matrix can be multi-channel. It also fully supports ROI mechanism.
-
-   There are many different ways to create cv::Mat object. Here are the some popular ones:
-   <ul>
-   <li> using cv::Mat::create(nrows, ncols, type) method or
-     the similar constructor cv::Mat::Mat(nrows, ncols, type[, fill_value]) constructor.
-     A new matrix of the specified size and specifed type will be allocated.
-     "type" has the same meaning as in cvCreateMat function,
-     e.g. CV_8UC1 means 8-bit single-channel matrix, CV_32FC2 means 2-channel (i.e. complex)
-     floating-point matrix etc:
-
-     \code
-     // make 7x7 complex matrix filled with 1+3j.
-     cv::Mat M(7,7,CV_32FC2,Scalar(1,3));
-     // and now turn M to 100x60 15-channel 8-bit matrix.
-     // The old content will be deallocated
-     M.create(100,60,CV_8UC(15));
-     \endcode
-
-     As noted in the introduction of this chapter, Mat::create()
-     will only allocate a new matrix when the current matrix dimensionality
-     or type are different from the specified.
-
-   <li> by using a copy constructor or assignment operator, where on the right side it can
-     be a matrix or expression, see below. Again, as noted in the introduction,
-     matrix assignment is O(1) operation because it only copies the header
-     and increases the reference counter. cv::Mat::clone() method can be used to get a full
-     (a.k.a. deep) copy of the matrix when you need it.
-
-   <li> by constructing a header for a part of another matrix. It can be a single row, single column,
-     several rows, several columns, rectangular region in the matrix (called a minor in algebra) or
-     a diagonal. Such operations are also O(1), because the new header will reference the same data.
-     You can actually modify a part of the matrix using this feature, e.g.
-
-     \code
-     // add 5-th row, multiplied by 3 to the 3rd row
-     M.row(3) = M.row(3) + M.row(5)*3;
-
-     // now copy 7-th column to the 1-st column
-     // M.col(1) = M.col(7); // this will not work
-     Mat M1 = M.col(1);
-     M.col(7).copyTo(M1);
-
-     // create new 320x240 image
-     cv::Mat img(Size(320,240),CV_8UC3);
-     // select a roi
-     cv::Mat roi(img, Rect(10,10,100,100));
-     // fill the ROI with (0,255,0) (which is green in RGB space);
-     // the original 320x240 image will be modified
-     roi = Scalar(0,255,0);
-     \endcode
-
-     Thanks to the additional cv::Mat::datastart and cv::Mat::dataend members, it is possible to
-     compute the relative sub-matrix position in the main "container" matrix using cv::Mat::locateROI():
-
-     \code
-     Mat A = Mat::eye(10, 10, CV_32S);
-     // extracts A columns, 1 (inclusive) to 3 (exclusive).
-     Mat B = A(Range::all(), Range(1, 3));
-     // extracts B rows, 5 (inclusive) to 9 (exclusive).
-     // that is, C ~ A(Range(5, 9), Range(1, 3))
-     Mat C = B(Range(5, 9), Range::all());
-     Size size; Point ofs;
-     C.locateROI(size, ofs);
-     // size will be (width=10,height=10) and the ofs will be (x=1, y=5)
-     \endcode
-
-     As in the case of whole matrices, if you need a deep copy, use cv::Mat::clone() method
-     of the extracted sub-matrices.
-
-   <li> by making a header for user-allocated-data. It can be useful for
-      <ol>
-      <li> processing "foreign" data using OpenCV (e.g. when you implement
-         a DirectShow filter or a processing module for gstreamer etc.), e.g.
-
-         \code
-         void process_video_frame(const unsigned char* pixels,
-                                  int width, int height, int step)
-         {
-            cv::Mat img(height, width, CV_8UC3, pixels, step);
-            cv::GaussianBlur(img, img, cv::Size(7,7), 1.5, 1.5);
-         }
-         \endcode
-
-      <li> for quick initialization of small matrices and/or super-fast element access
-
-         \code
-         double m[3][3] = {{a, b, c}, {d, e, f}, {g, h, i}};
-         cv::Mat M = cv::Mat(3, 3, CV_64F, m).inv();
-         \endcode
-      </ol>
-
-       partial yet very common cases of this "user-allocated data" case are conversions
-       from CvMat and IplImage to cv::Mat. For this purpose there are special constructors
-       taking pointers to CvMat or IplImage and the optional
-       flag indicating whether to copy the data or not.
-
-       Backward conversion from cv::Mat to CvMat or IplImage is provided via cast operators
-       cv::Mat::operator CvMat() an cv::Mat::operator IplImage().
-       The operators do not copy the data.
-
-
-       \code
-       IplImage* img = cvLoadImage("greatwave.jpg", 1);
-       Mat mtx(img); // convert IplImage* -> cv::Mat
-       CvMat oldmat = mtx; // convert cv::Mat -> CvMat
-       CV_Assert(oldmat.cols == img->width && oldmat.rows == img->height &&
-           oldmat.data.ptr == (uchar*)img->imageData && oldmat.step == img->widthStep);
-       \endcode
-
-   <li> by using MATLAB-style matrix initializers, cv::Mat::zeros(), cv::Mat::ones(), cv::Mat::eye(), e.g.:
-
-   \code
-   // create a double-precision identity martix and add it to M.
-   M += Mat::eye(M.rows, M.cols, CV_64F);
-   \endcode
-
-   <li> by using comma-separated initializer:
-
-   \code
-   // create 3x3 double-precision identity matrix
-   Mat M = (Mat_<double>(3,3) << 1, 0, 0, 0, 1, 0, 0, 0, 1);
-   \endcode
-
-   here we first call constructor of cv::Mat_ class (that we describe further) with the proper matrix,
-   and then we just put "<<" operator followed by comma-separated values that can be constants,
-   variables, expressions etc. Also, note the extra parentheses that are needed to avoid compiler errors.
-
-   </ul>
-
-   Once matrix is created, it will be automatically managed by using reference-counting mechanism
-   (unless the matrix header is built on top of user-allocated data,
-   in which case you should handle the data by yourself).
-   The matrix data will be deallocated when no one points to it;
-   if you want to release the data pointed by a matrix header before the matrix destructor is called,
-   use cv::Mat::release().
-
-   The next important thing to learn about the matrix class is element access. Here is how the matrix is stored.
-   The elements are stored in row-major order (row by row). The cv::Mat::data member points to the first element of the first row,
-   cv::Mat::rows contains the number of matrix rows and cv::Mat::cols - the number of matrix columns. There is yet another member,
-   cv::Mat::step that is used to actually compute address of a matrix element. cv::Mat::step is needed because the matrix can be
-   a part of another matrix or because there can some padding space in the end of each row for a proper alignment.
-
-   \image html roi.png
-
-   Given these parameters, address of the matrix element M_{ij} is computed as following:
-
-   addr(M_{ij})=M.data + M.step*i + j*M.elemSize()
-
-   if you know the matrix element type, e.g. it is float, then you can use cv::Mat::at() method:
-
-   addr(M_{ij})=&M.at<float>(i,j)
-
-   (where & is used to convert the reference returned by cv::Mat::at() to a pointer).
-   if you need to process a whole row of matrix, the most efficient way is to get
-   the pointer to the row first, and then just use plain C operator []:
-
-   \code
-   // compute sum of positive matrix elements
-   // (assuming that M is double-precision matrix)
-   double sum=0;
-   for(int i = 0; i < M.rows; i++)
-   {
-       const double* Mi = M.ptr<double>(i);
-       for(int j = 0; j < M.cols; j++)
-           sum += std::max(Mi[j], 0.);
-   }
-   \endcode
-
-   Some operations, like the above one, do not actually depend on the matrix shape,
-   they just process elements of a matrix one by one (or elements from multiple matrices
-   that are sitting in the same place, e.g. matrix addition). Such operations are called
-   element-wise and it makes sense to check whether all the input/output matrices are continuous,
-   i.e. have no gaps in the end of each row, and if yes, process them as a single long row:
-
-   \code
-   // compute sum of positive matrix elements, optimized variant
-   double sum=0;
-   int cols = M.cols, rows = M.rows;
-   if(M.isContinuous())
-   {
-       cols *= rows;
-       rows = 1;
-   }
-   for(int i = 0; i < rows; i++)
-   {
-       const double* Mi = M.ptr<double>(i);
-       for(int j = 0; j < cols; j++)
-           sum += std::max(Mi[j], 0.);
-   }
-   \endcode
-   in the case of continuous matrix the outer loop body will be executed just once,
-   so the overhead will be smaller, which will be especially noticeable in the case of small matrices.
-
-   Finally, there are STL-style iterators that are smart enough to skip gaps between successive rows:
-   \code
-   // compute sum of positive matrix elements, iterator-based variant
-   double sum=0;
-   MatConstIterator_<double> it = M.begin<double>(), it_end = M.end<double>();
-   for(; it != it_end; ++it)
-       sum += std::max(*it, 0.);
-   \endcode
-
-   The matrix iterators are random-access iterators, so they can be passed
-   to any STL algorithm, including std::sort().
-*/
-class CV_EXPORTS Mat
-{
-public:
-    //! default constructor
-    Mat();
-    //! constructs 2D matrix of the specified size and type
-    // (_type is CV_8UC1, CV_64FC3, CV_32SC(12) etc.)
-    Mat(int rows, int cols, int type);
-    Mat(Size size, int type);
-    //! constucts 2D matrix and fills it with the specified value _s.
-    Mat(int rows, int cols, int type, const Scalar& s);
-    Mat(Size size, int type, const Scalar& s);
-
-    //! constructs n-dimensional matrix
-    Mat(int ndims, const int* sizes, int type);
-    Mat(int ndims, const int* sizes, int type, const Scalar& s);
-
-    //! copy constructor
-    Mat(const Mat& m);
-    //! constructor for matrix headers pointing to user-allocated data
-    Mat(int rows, int cols, int type, void* data, size_t step=AUTO_STEP);
-    Mat(Size size, int type, void* data, size_t step=AUTO_STEP);
-    Mat(int ndims, const int* sizes, int type, void* data, const size_t* steps=0);
-
-    //! creates a matrix header for a part of the bigger matrix
-    Mat(const Mat& m, const Range& rowRange, const Range& colRange=Range::all());
-    Mat(const Mat& m, const Rect& roi);
-    Mat(const Mat& m, const Range* ranges);
-    //! converts old-style CvMat to the new matrix; the data is not copied by default
-    Mat(const CvMat* m, bool copyData=false);
-    //! converts old-style CvMatND to the new matrix; the data is not copied by default
-    Mat(const CvMatND* m, bool copyData=false);
-    //! converts old-style IplImage to the new matrix; the data is not copied by default
-    Mat(const IplImage* img, bool copyData=false);
-    //! builds matrix from std::vector with or without copying the data
-    template<typename _Tp> explicit Mat(const std::vector<_Tp>& vec, bool copyData=false);
-    //! builds matrix from cv::Vec; the data is copied by default
-    template<typename _Tp, int n> explicit Mat(const Vec<_Tp, n>& vec, bool copyData=true);
-    //! builds matrix from cv::Matx; the data is copied by default
-    template<typename _Tp, int m, int n> explicit Mat(const Matx<_Tp, m, n>& mtx, bool copyData=true);
-    //! builds matrix from a 2D point
-    template<typename _Tp> explicit Mat(const Point_<_Tp>& pt, bool copyData=true);
-    //! builds matrix from a 3D point
-    template<typename _Tp> explicit Mat(const Point3_<_Tp>& pt, bool copyData=true);
-    //! builds matrix from comma initializer
-    template<typename _Tp> explicit Mat(const MatCommaInitializer_<_Tp>& commaInitializer);
-
-    //! download data from GpuMat
-    explicit Mat(const gpu::GpuMat& m);
-
-    //! destructor - calls release()
-    ~Mat();
-    //! assignment operators
-    Mat& operator = (const Mat& m);
-    Mat& operator = (const MatExpr& expr);
-
-    //! returns a new matrix header for the specified row
-    Mat row(int y) const;
-    //! returns a new matrix header for the specified column
-    Mat col(int x) const;
-    //! ... for the specified row span
-    Mat rowRange(int startrow, int endrow) const;
-    Mat rowRange(const Range& r) const;
-    //! ... for the specified column span
-    Mat colRange(int startcol, int endcol) const;
-    Mat colRange(const Range& r) const;
-    //! ... for the specified diagonal
-    // (d=0 - the main diagonal,
-    //  >0 - a diagonal from the lower half,
-    //  <0 - a diagonal from the upper half)
-    Mat diag(int d=0) const;
-    //! constructs a square diagonal matrix which main diagonal is vector "d"
-    static Mat diag(const Mat& d);
-
-    //! returns deep copy of the matrix, i.e. the data is copied
-    Mat clone() const;
-    //! copies the matrix content to "m".
-    // It calls m.create(this->size(), this->type()).
-    void copyTo( OutputArray m ) const;
-    //! copies those matrix elements to "m" that are marked with non-zero mask elements.
-    void copyTo( OutputArray m, InputArray mask ) const;
-    //! converts matrix to another datatype with optional scalng. See cvConvertScale.
-    void convertTo( OutputArray m, int rtype, double alpha=1, double beta=0 ) const;
-
-    void assignTo( Mat& m, int type=-1 ) const;
-
-    //! sets every matrix element to s
-    Mat& operator = (const Scalar& s);
-    //! sets some of the matrix elements to s, according to the mask
-    Mat& setTo(InputArray value, InputArray mask=noArray());
-    //! creates alternative matrix header for the same data, with different
-    // number of channels and/or different number of rows. see cvReshape.
-    Mat reshape(int cn, int rows=0) const;
-    Mat reshape(int cn, int newndims, const int* newsz) const;
-
-    //! matrix transposition by means of matrix expressions
-    MatExpr t() const;
-    //! matrix inversion by means of matrix expressions
-    MatExpr inv(int method=DECOMP_LU) const;
-    //! per-element matrix multiplication by means of matrix expressions
-    MatExpr mul(InputArray m, double scale=1) const;
-
-    //! computes cross-product of 2 3D vectors
-    Mat cross(InputArray m) const;
-    //! computes dot-product
-    double dot(InputArray m) const;
-
-    //! Matlab-style matrix initialization
-    static MatExpr zeros(int rows, int cols, int type);
-    static MatExpr zeros(Size size, int type);
-    static MatExpr zeros(int ndims, const int* sz, int type);
-    static MatExpr ones(int rows, int cols, int type);
-    static MatExpr ones(Size size, int type);
-    static MatExpr ones(int ndims, const int* sz, int type);
-    static MatExpr eye(int rows, int cols, int type);
-    static MatExpr eye(Size size, int type);
-
-    //! allocates new matrix data unless the matrix already has specified size and type.
-    // previous data is unreferenced if needed.
-    void create(int rows, int cols, int type);
-    void create(Size size, int type);
-    void create(int ndims, const int* sizes, int type);
-
-    //! increases the reference counter; use with care to avoid memleaks
-    void addref();
-    //! decreases reference counter;
-    // deallocates the data when reference counter reaches 0.
-    void release();
-
-    //! deallocates the matrix data
-    void deallocate();
-    //! internal use function; properly re-allocates _size, _step arrays
-    void copySize(const Mat& m);
-
-    //! reserves enough space to fit sz hyper-planes
-    void reserve(size_t sz);
-    //! resizes matrix to the specified number of hyper-planes
-    void resize(size_t sz);
-    //! resizes matrix to the specified number of hyper-planes; initializes the newly added elements
-    void resize(size_t sz, const Scalar& s);
-    //! internal function
-    void push_back_(const void* elem);
-    //! adds element to the end of 1d matrix (or possibly multiple elements when _Tp=Mat)
-    template<typename _Tp> void push_back(const _Tp& elem);
-    template<typename _Tp> void push_back(const Mat_<_Tp>& elem);
-    void push_back(const Mat& m);
-    //! removes several hyper-planes from bottom of the matrix
-    void pop_back(size_t nelems=1);
-
-    //! locates matrix header within a parent matrix. See below
-    void locateROI( Size& wholeSize, Point& ofs ) const;
-    //! moves/resizes the current matrix ROI inside the parent matrix.
-    Mat& adjustROI( int dtop, int dbottom, int dleft, int dright );
-    //! extracts a rectangular sub-matrix
-    // (this is a generalized form of row, rowRange etc.)
-    Mat operator()( Range rowRange, Range colRange ) const;
-    Mat operator()( const Rect& roi ) const;
-    Mat operator()( const Range* ranges ) const;
-
-    //! converts header to CvMat; no data is copied
-    operator CvMat() const;
-    //! converts header to CvMatND; no data is copied
-    operator CvMatND() const;
-    //! converts header to IplImage; no data is copied
-    operator IplImage() const;
-
-    template<typename _Tp> operator std::vector<_Tp>() const;
-    template<typename _Tp, int n> operator Vec<_Tp, n>() const;
-    template<typename _Tp, int m, int n> operator Matx<_Tp, m, n>() const;
-
-    //! returns true iff the matrix data is continuous
-    // (i.e. when there are no gaps between successive rows).
-    // similar to CV_IS_MAT_CONT(cvmat->type)
-    bool isContinuous() const;
-
-    //! returns true if the matrix is a submatrix of another matrix
-    bool isSubmatrix() const;
-
-    //! returns element size in bytes,
-    // similar to CV_ELEM_SIZE(cvmat->type)
-    size_t elemSize() const;
-    //! returns the size of element channel in bytes.
-    size_t elemSize1() const;
-    //! returns element type, similar to CV_MAT_TYPE(cvmat->type)
-    int type() const;
-    //! returns element type, similar to CV_MAT_DEPTH(cvmat->type)
-    int depth() const;
-    //! returns element type, similar to CV_MAT_CN(cvmat->type)
-    int channels() const;
-    //! returns step/elemSize1()
-    size_t step1(int i=0) const;
-    //! returns true if matrix data is NULL
-    bool empty() const;
-    //! returns the total number of matrix elements
-    size_t total() const;
-
-    //! returns N if the matrix is 1-channel (N x ptdim) or ptdim-channel (1 x N) or (N x 1); negative number otherwise
-    int checkVector(int elemChannels, int depth=-1, bool requireContinuous=true) const;
-
-    //! returns pointer to i0-th submatrix along the dimension #0
-    uchar* ptr(int i0=0);
-    const uchar* ptr(int i0=0) const;
-
-    //! returns pointer to (i0,i1) submatrix along the dimensions #0 and #1
-    uchar* ptr(int i0, int i1);
-    const uchar* ptr(int i0, int i1) const;
-
-    //! returns pointer to (i0,i1,i3) submatrix along the dimensions #0, #1, #2
-    uchar* ptr(int i0, int i1, int i2);
-    const uchar* ptr(int i0, int i1, int i2) const;
-
-    //! returns pointer to the matrix element
-    uchar* ptr(const int* idx);
-    //! returns read-only pointer to the matrix element
-    const uchar* ptr(const int* idx) const;
-
-    template<int n> uchar* ptr(const Vec<int, n>& idx);
-    template<int n> const uchar* ptr(const Vec<int, n>& idx) const;
-
-    //! template version of the above method
-    template<typename _Tp> _Tp* ptr(int i0=0);
-    template<typename _Tp> const _Tp* ptr(int i0=0) const;
-
-    template<typename _Tp> _Tp* ptr(int i0, int i1);
-    template<typename _Tp> const _Tp* ptr(int i0, int i1) const;
-
-    template<typename _Tp> _Tp* ptr(int i0, int i1, int i2);
-    template<typename _Tp> const _Tp* ptr(int i0, int i1, int i2) const;
-
-    template<typename _Tp> _Tp* ptr(const int* idx);
-    template<typename _Tp> const _Tp* ptr(const int* idx) const;
-
-    template<typename _Tp, int n> _Tp* ptr(const Vec<int, n>& idx);
-    template<typename _Tp, int n> const _Tp* ptr(const Vec<int, n>& idx) const;
-
-    //! the same as above, with the pointer dereferencing
-    template<typename _Tp> _Tp& at(int i0=0);
-    template<typename _Tp> const _Tp& at(int i0=0) const;
-
-    template<typename _Tp> _Tp& at(int i0, int i1);
-    template<typename _Tp> const _Tp& at(int i0, int i1) const;
-
-    template<typename _Tp> _Tp& at(int i0, int i1, int i2);
-    template<typename _Tp> const _Tp& at(int i0, int i1, int i2) const;
-
-    template<typename _Tp> _Tp& at(const int* idx);
-    template<typename _Tp> const _Tp& at(const int* idx) const;
-
-    template<typename _Tp, int n> _Tp& at(const Vec<int, n>& idx);
-    template<typename _Tp, int n> const _Tp& at(const Vec<int, n>& idx) const;
-
-    //! special versions for 2D arrays (especially convenient for referencing image pixels)
-    template<typename _Tp> _Tp& at(Point pt);
-    template<typename _Tp> const _Tp& at(Point pt) const;
-
-    //! template methods for iteration over matrix elements.
-    // the iterators take care of skipping gaps in the end of rows (if any)
-    template<typename _Tp> MatIterator_<_Tp> begin();
-    template<typename _Tp> MatIterator_<_Tp> end();
-    template<typename _Tp> MatConstIterator_<_Tp> begin() const;
-    template<typename _Tp> MatConstIterator_<_Tp> end() const;
-
-    enum { MAGIC_VAL=0x42FF0000, AUTO_STEP=0, CONTINUOUS_FLAG=CV_MAT_CONT_FLAG, SUBMATRIX_FLAG=CV_SUBMAT_FLAG };
-
-    /*! includes several bit-fields:
-         - the magic signature
-         - continuity flag
-         - depth
-         - number of channels
-     */
-    int flags;
-    //! the matrix dimensionality, >= 2
-    int dims;
-    //! the number of rows and columns or (-1, -1) when the matrix has more than 2 dimensions
-    int rows, cols;
-    //! pointer to the data
-    uchar* data;
-
-    //! pointer to the reference counter;
-    // when matrix points to user-allocated data, the pointer is NULL
-    int* refcount;
-
-    //! helper fields used in locateROI and adjustROI
-    uchar* datastart;
-    uchar* dataend;
-    uchar* datalimit;
-
-    //! custom allocator
-    MatAllocator* allocator;
-
-    struct CV_EXPORTS MSize
-    {
-        MSize(int* _p);
-        Size operator()() const;
-        const int& operator[](int i) const;
-        int& operator[](int i);
-        operator const int*() const;
-        bool operator == (const MSize& sz) const;
-        bool operator != (const MSize& sz) const;
-
-        int* p;
-    };
-
-    struct CV_EXPORTS MStep
-    {
-        MStep();
-        MStep(size_t s);
-        const size_t& operator[](int i) const;
-        size_t& operator[](int i);
-        operator size_t() const;
-        MStep& operator = (size_t s);
-
-        size_t* p;
-        size_t buf[2];
-    protected:
-        MStep& operator = (const MStep&);
-    };
-
-    MSize size;
-    MStep step;
-
-protected:
-    void initEmpty();
-};
-
-
-/*!
-   Random Number Generator
-
-   The class implements RNG using Multiply-with-Carry algorithm
-*/
-class CV_EXPORTS RNG
-{
-public:
-    enum { UNIFORM=0, NORMAL=1 };
-
-    RNG();
-    RNG(uint64 state);
-    //! updates the state and returns the next 32-bit unsigned integer random number
-    unsigned next();
-
-    operator uchar();
-    operator schar();
-    operator ushort();
-    operator short();
-    operator unsigned();
-    //! returns a random integer sampled uniformly from [0, N).
-    unsigned operator ()(unsigned N);
-    unsigned operator ()();
-    operator int();
-    operator float();
-    operator double();
-    //! returns uniformly distributed integer random number from [a,b) range
-    int uniform(int a, int b);
-    //! returns uniformly distributed floating-point random number from [a,b) range
-    float uniform(float a, float b);
-    //! returns uniformly distributed double-precision floating-point random number from [a,b) range
-    double uniform(double a, double b);
-    void fill( InputOutputArray mat, int distType, InputArray a, InputArray b, bool saturateRange=false );
-    //! returns Gaussian random variate with mean zero.
-    double gaussian(double sigma);
-
-    uint64 state;
-};
-
-class CV_EXPORTS RNG_MT19937
-{
-public:
-    RNG_MT19937();
-    RNG_MT19937(unsigned s);
-    void seed(unsigned s);
-
-    unsigned next();
-
-    operator int();
-    operator unsigned();
-    operator float();
-    operator double();
-
-    unsigned operator ()(unsigned N);
-    unsigned operator ()();
-
-    // returns uniformly distributed integer random number from [a,b) range
-    int uniform(int a, int b);
-    // returns uniformly distributed floating-point random number from [a,b) range
-    float uniform(float a, float b);
-    // returns uniformly distributed double-precision floating-point random number from [a,b) range
-    double uniform(double a, double b);
-
-private:
-    enum PeriodParameters {N = 624, M = 397};
-    unsigned state[N];
-    int mti;
-};
-
-/*!
- Termination criteria in iterative algorithms
- */
-class CV_EXPORTS TermCriteria
-{
-public:
-    enum
-    {
-        COUNT=1, //!< the maximum number of iterations or elements to compute
-        MAX_ITER=COUNT, //!< ditto
-        EPS=2 //!< the desired accuracy or change in parameters at which the iterative algorithm stops
-    };
-
-    //! default constructor
-    TermCriteria();
-    //! full constructor
-    TermCriteria(int type, int maxCount, double epsilon);
-    //! conversion from CvTermCriteria
-    TermCriteria(const CvTermCriteria& criteria);
-    //! conversion to CvTermCriteria
-    operator CvTermCriteria() const;
-
-    int type; //!< the type of termination criteria: COUNT, EPS or COUNT + EPS
-    int maxCount; // the maximum number of iterations/elements
-    double epsilon; // the desired accuracy
-};
-
-
-typedef void (*BinaryFunc)(const uchar* src1, size_t step1,
-                           const uchar* src2, size_t step2,
-                           uchar* dst, size_t step, Size sz,
-                           void*);
-
-CV_EXPORTS BinaryFunc getConvertFunc(int sdepth, int ddepth);
-CV_EXPORTS BinaryFunc getConvertScaleFunc(int sdepth, int ddepth);
-CV_EXPORTS BinaryFunc getCopyMaskFunc(size_t esz);
 
 //! swaps two matrices
 CV_EXPORTS void swap(Mat& a, Mat& b);
 
-//! converts array (CvMat or IplImage) to cv::Mat
-CV_EXPORTS Mat cvarrToMat(const CvArr* arr, bool copyData=false,
-                          bool allowND=true, int coiMode=0,
-                          AutoBuffer<double>* buf=0);
-//! extracts Channel of Interest from CvMat or IplImage and makes cv::Mat out of it.
-CV_EXPORTS void extractImageCOI(const CvArr* arr, OutputArray coiimg, int coi=-1);
-//! inserts single-channel cv::Mat into a multi-channel CvMat or IplImage
-CV_EXPORTS void insertImageCOI(InputArray coiimg, CvArr* arr, int coi=-1);
+//! 1D interpolation function: returns coordinate of the "donor" pixel for the specified location p.
+CV_EXPORTS_W int borderInterpolate(int p, int len, int borderType);
+
+//! copies 2D array to a larger destination array with extrapolation of the outer part of src using the specified border mode
+CV_EXPORTS_W void copyMakeBorder(InputArray src, OutputArray dst,
+                                 int top, int bottom, int left, int right,
+                                 int borderType, const Scalar& value = Scalar() );
 
 //! adds one matrix to another (dst = src1 + src2)
 CV_EXPORTS_W void add(InputArray src1, InputArray src2, OutputArray dst,
-                      InputArray mask=noArray(), int dtype=-1);
+                      InputArray mask = noArray(), int dtype = -1);
+
 //! subtracts one matrix from another (dst = src1 - src2)
 CV_EXPORTS_W void subtract(InputArray src1, InputArray src2, OutputArray dst,
-                           InputArray mask=noArray(), int dtype=-1);
+                           InputArray mask = noArray(), int dtype = -1);
 
 //! computes element-wise weighted product of the two arrays (dst = scale*src1*src2)
 CV_EXPORTS_W void multiply(InputArray src1, InputArray src2,
-                           OutputArray dst, double scale=1, int dtype=-1);
+                           OutputArray dst, double scale = 1, int dtype = -1);
 
-//! computes element-wise weighted quotient of the two arrays (dst = scale*src1/src2)
+//! computes element-wise weighted quotient of the two arrays (dst = scale * src1 / src2)
 CV_EXPORTS_W void divide(InputArray src1, InputArray src2, OutputArray dst,
-                         double scale=1, int dtype=-1);
+                         double scale = 1, int dtype = -1);
 
 //! computes element-wise weighted reciprocal of an array (dst = scale/src2)
 CV_EXPORTS_W void divide(double scale, InputArray src2,
-                         OutputArray dst, int dtype=-1);
+                         OutputArray dst, int dtype = -1);
 
 //! adds scaled array to another one (dst = alpha*src1 + src2)
 CV_EXPORTS_W void scaleAdd(InputArray src1, double alpha, InputArray src2, OutputArray dst);
 
 //! computes weighted sum of two arrays (dst = alpha*src1 + beta*src2 + gamma)
 CV_EXPORTS_W void addWeighted(InputArray src1, double alpha, InputArray src2,
-                              double beta, double gamma, OutputArray dst, int dtype=-1);
+                              double beta, double gamma, OutputArray dst, int dtype = -1);
 
 //! scales array elements, computes absolute values and converts the results to 8-bit unsigned integers: dst(i)=saturate_cast<uchar>abs(src(i)*alpha+beta)
 CV_EXPORTS_W void convertScaleAbs(InputArray src, OutputArray dst,
-                                  double alpha=1, double beta=0);
+                                  double alpha = 1, double beta = 0);
+
 //! transforms array of numbers using a lookup table: dst(i)=lut(src(i))
-CV_EXPORTS_W void LUT(InputArray src, InputArray lut, OutputArray dst,
-                      int interpolation=0);
+CV_EXPORTS_W void LUT(InputArray src, InputArray lut, OutputArray dst);
 
 //! computes sum of array elements
 CV_EXPORTS_AS(sumElems) Scalar sum(InputArray src);
+
 //! computes the number of nonzero array elements
 CV_EXPORTS_W int countNonZero( InputArray src );
+
 //! returns the list of locations of non-zero pixels
 CV_EXPORTS_W void findNonZero( InputArray src, OutputArray idx );
 
 //! computes mean value of selected array elements
-CV_EXPORTS_W Scalar mean(InputArray src, InputArray mask=noArray());
+CV_EXPORTS_W Scalar mean(InputArray src, InputArray mask = noArray());
+
 //! computes mean value and standard deviation of all or selected array elements
 CV_EXPORTS_W void meanStdDev(InputArray src, OutputArray mean, OutputArray stddev,
                              InputArray mask=noArray());
+
 //! computes norm of the selected array part
-CV_EXPORTS_W double norm(InputArray src1, int normType=NORM_L2, InputArray mask=noArray());
+CV_EXPORTS_W double norm(InputArray src1, int normType = NORM_L2, InputArray mask = noArray());
+
 //! computes norm of selected part of the difference between two arrays
 CV_EXPORTS_W double norm(InputArray src1, InputArray src2,
-                         int normType=NORM_L2, InputArray mask=noArray());
+                         int normType = NORM_L2, InputArray mask = noArray());
+
+//! computes PSNR image/video quality metric
+CV_EXPORTS_W double PSNR(InputArray src1, InputArray src2);
+
+//! computes norm of a sparse matrix
+CV_EXPORTS double norm( const SparseMat& src, int normType );
 
 //! naive nearest neighbor finder
 CV_EXPORTS_W void batchDistance(InputArray src1, InputArray src2,
                                 OutputArray dist, int dtype, OutputArray nidx,
-                                int normType=NORM_L2, int K=0,
-                                InputArray mask=noArray(), int update=0,
-                                bool crosscheck=false);
+                                int normType = NORM_L2, int K = 0,
+                                InputArray mask = noArray(), int update = 0,
+                                bool crosscheck = false);
 
 //! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
-CV_EXPORTS_W void normalize( InputArray src, OutputArray dst, double alpha=1, double beta=0,
-                             int norm_type=NORM_L2, int dtype=-1, InputArray mask=noArray());
+CV_EXPORTS_W void normalize( InputArray src, OutputArray dst, double alpha = 1, double beta = 0,
+                             int norm_type = NORM_L2, int dtype = -1, InputArray mask = noArray());
+
+//! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
+CV_EXPORTS void normalize( const SparseMat& src, SparseMat& dst, double alpha, int normType );
 
 //! finds global minimum and maximum array elements and returns their values and their locations
 CV_EXPORTS_W void minMaxLoc(InputArray src, CV_OUT double* minVal,
-                           CV_OUT double* maxVal=0, CV_OUT Point* minLoc=0,
-                           CV_OUT Point* maxLoc=0, InputArray mask=noArray());
-CV_EXPORTS void minMaxIdx(InputArray src, double* minVal, double* maxVal,
-                          int* minIdx=0, int* maxIdx=0, InputArray mask=noArray());
+                            CV_OUT double* maxVal = 0, CV_OUT Point* minLoc = 0,
+                            CV_OUT Point* maxLoc = 0, InputArray mask = noArray());
+
+CV_EXPORTS void minMaxIdx(InputArray src, double* minVal, double* maxVal = 0,
+                          int* minIdx = 0, int* maxIdx = 0, InputArray mask = noArray());
+
+//! finds global minimum and maximum sparse array elements and returns their values and their locations
+CV_EXPORTS void minMaxLoc(const SparseMat& a, double* minVal,
+                          double* maxVal, int* minIdx = 0, int* maxIdx = 0);
 
 //! transforms 2D matrix to 1D row or column vector by taking sum, minimum, maximum or mean value over all the rows
-CV_EXPORTS_W void reduce(InputArray src, OutputArray dst, int dim, int rtype, int dtype=-1);
+CV_EXPORTS_W void reduce(InputArray src, OutputArray dst, int dim, int rtype, int dtype = -1);
 
 //! makes multi-channel array out of several single-channel arrays
 CV_EXPORTS void merge(const Mat* mv, size_t count, OutputArray dst);
+
 //! makes multi-channel array out of several single-channel arrays
 CV_EXPORTS_W void merge(InputArrayOfArrays mv, OutputArray dst);
 
 //! copies each plane of a multi-channel array to a dedicated array
 CV_EXPORTS void split(const Mat& src, Mat* mvbegin);
+
 //! copies each plane of a multi-channel array to a dedicated array
 CV_EXPORTS_W void split(InputArray m, OutputArrayOfArrays mv);
 
 //! copies selected channels from the input arrays to the selected channels of the output arrays
 CV_EXPORTS void mixChannels(const Mat* src, size_t nsrcs, Mat* dst, size_t ndsts,
                             const int* fromTo, size_t npairs);
-CV_EXPORTS void mixChannels(const std::vector<Mat>& src, std::vector<Mat>& dst,
+
+CV_EXPORTS void mixChannels(InputArrayOfArrays src, InputOutputArrayOfArrays dst,
                             const int* fromTo, size_t npairs);
-CV_EXPORTS_W void mixChannels(InputArrayOfArrays src, InputArrayOfArrays dst,
+
+CV_EXPORTS_W void mixChannels(InputArrayOfArrays src, InputOutputArrayOfArrays dst,
                               const std::vector<int>& fromTo);
 
 //! extracts a single channel from src (coi is 0-based index)
@@ -2194,155 +291,299 @@ CV_EXPORTS_W void flip(InputArray src, OutputArray dst, int flipCode);
 
 //! replicates the input matrix the specified number of times in the horizontal and/or vertical direction
 CV_EXPORTS_W void repeat(InputArray src, int ny, int nx, OutputArray dst);
+
 CV_EXPORTS Mat repeat(const Mat& src, int ny, int nx);
 
 CV_EXPORTS void hconcat(const Mat* src, size_t nsrc, OutputArray dst);
+
 CV_EXPORTS void hconcat(InputArray src1, InputArray src2, OutputArray dst);
+
 CV_EXPORTS_W void hconcat(InputArrayOfArrays src, OutputArray dst);
 
 CV_EXPORTS void vconcat(const Mat* src, size_t nsrc, OutputArray dst);
+
 CV_EXPORTS void vconcat(InputArray src1, InputArray src2, OutputArray dst);
+
 CV_EXPORTS_W void vconcat(InputArrayOfArrays src, OutputArray dst);
 
 //! computes bitwise conjunction of the two arrays (dst = src1 & src2)
 CV_EXPORTS_W void bitwise_and(InputArray src1, InputArray src2,
-                              OutputArray dst, InputArray mask=noArray());
+                              OutputArray dst, InputArray mask = noArray());
+
 //! computes bitwise disjunction of the two arrays (dst = src1 | src2)
 CV_EXPORTS_W void bitwise_or(InputArray src1, InputArray src2,
-                             OutputArray dst, InputArray mask=noArray());
+                             OutputArray dst, InputArray mask = noArray());
+
 //! computes bitwise exclusive-or of the two arrays (dst = src1 ^ src2)
 CV_EXPORTS_W void bitwise_xor(InputArray src1, InputArray src2,
-                              OutputArray dst, InputArray mask=noArray());
+                              OutputArray dst, InputArray mask = noArray());
+
 //! inverts each bit of array (dst = ~src)
 CV_EXPORTS_W void bitwise_not(InputArray src, OutputArray dst,
-                              InputArray mask=noArray());
+                              InputArray mask = noArray());
+
 //! computes element-wise absolute difference of two arrays (dst = abs(src1 - src2))
 CV_EXPORTS_W void absdiff(InputArray src1, InputArray src2, OutputArray dst);
+
 //! set mask elements for those array elements which are within the element-specific bounding box (dst = lowerb <= src && src < upperb)
 CV_EXPORTS_W void inRange(InputArray src, InputArray lowerb,
                           InputArray upperb, OutputArray dst);
+
 //! compares elements of two arrays (dst = src1 <cmpop> src2)
 CV_EXPORTS_W void compare(InputArray src1, InputArray src2, OutputArray dst, int cmpop);
+
 //! computes per-element minimum of two arrays (dst = min(src1, src2))
 CV_EXPORTS_W void min(InputArray src1, InputArray src2, OutputArray dst);
+
 //! computes per-element maximum of two arrays (dst = max(src1, src2))
 CV_EXPORTS_W void max(InputArray src1, InputArray src2, OutputArray dst);
 
+// the following overloads are needed to avoid conflicts with
+//     const _Tp& std::min(const _Tp&, const _Tp&, _Compare)
 //! computes per-element minimum of two arrays (dst = min(src1, src2))
 CV_EXPORTS void min(const Mat& src1, const Mat& src2, Mat& dst);
-//! computes per-element minimum of array and scalar (dst = min(src1, src2))
-CV_EXPORTS void min(const Mat& src1, double src2, Mat& dst);
 //! computes per-element maximum of two arrays (dst = max(src1, src2))
 CV_EXPORTS void max(const Mat& src1, const Mat& src2, Mat& dst);
-//! computes per-element maximum of array and scalar (dst = max(src1, src2))
-CV_EXPORTS void max(const Mat& src1, double src2, Mat& dst);
 
 //! computes square root of each matrix element (dst = src**0.5)
 CV_EXPORTS_W void sqrt(InputArray src, OutputArray dst);
+
 //! raises the input matrix elements to the specified power (b = a**power)
 CV_EXPORTS_W void pow(InputArray src, double power, OutputArray dst);
+
 //! computes exponent of each matrix element (dst = e**src)
 CV_EXPORTS_W void exp(InputArray src, OutputArray dst);
+
 //! computes natural logarithm of absolute value of each matrix element: dst = log(abs(src))
 CV_EXPORTS_W void log(InputArray src, OutputArray dst);
-//! computes cube root of the argument
-CV_EXPORTS_W float cubeRoot(float val);
-//! computes the angle in degrees (0..360) of the vector (x,y)
-CV_EXPORTS_W float fastAtan2(float y, float x);
-
-CV_EXPORTS void exp(const float* src, float* dst, int n);
-CV_EXPORTS void log(const float* src, float* dst, int n);
-CV_EXPORTS void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
-CV_EXPORTS void magnitude(const float* x, const float* y, float* dst, int n);
 
 //! converts polar coordinates to Cartesian
 CV_EXPORTS_W void polarToCart(InputArray magnitude, InputArray angle,
-                              OutputArray x, OutputArray y, bool angleInDegrees=false);
+                              OutputArray x, OutputArray y, bool angleInDegrees = false);
+
 //! converts Cartesian coordinates to polar
 CV_EXPORTS_W void cartToPolar(InputArray x, InputArray y,
                               OutputArray magnitude, OutputArray angle,
-                              bool angleInDegrees=false);
+                              bool angleInDegrees = false);
+
 //! computes angle (angle(i)) of each (x(i), y(i)) vector
 CV_EXPORTS_W void phase(InputArray x, InputArray y, OutputArray angle,
-                        bool angleInDegrees=false);
+                        bool angleInDegrees = false);
+
 //! computes magnitude (magnitude(i)) of each (x(i), y(i)) vector
 CV_EXPORTS_W void magnitude(InputArray x, InputArray y, OutputArray magnitude);
+
 //! checks that each matrix element is within the specified range.
-CV_EXPORTS_W bool checkRange(InputArray a, bool quiet=true, CV_OUT Point* pos=0,
-                            double minVal=-DBL_MAX, double maxVal=DBL_MAX);
+CV_EXPORTS_W bool checkRange(InputArray a, bool quiet = true, CV_OUT Point* pos = 0,
+                            double minVal = -DBL_MAX, double maxVal = DBL_MAX);
+
 //! converts NaN's to the given number
-CV_EXPORTS_W void patchNaNs(InputOutputArray a, double val=0);
+CV_EXPORTS_W void patchNaNs(InputOutputArray a, double val = 0);
 
 //! implements generalized matrix product algorithm GEMM from BLAS
 CV_EXPORTS_W void gemm(InputArray src1, InputArray src2, double alpha,
-                       InputArray src3, double gamma, OutputArray dst, int flags=0);
+                       InputArray src3, double gamma, OutputArray dst, int flags = 0);
+
 //! multiplies matrix by its transposition from the left or from the right
 CV_EXPORTS_W void mulTransposed( InputArray src, OutputArray dst, bool aTa,
-                                 InputArray delta=noArray(),
-                                 double scale=1, int dtype=-1 );
+                                 InputArray delta = noArray(),
+                                 double scale = 1, int dtype = -1 );
+
 //! transposes the matrix
 CV_EXPORTS_W void transpose(InputArray src, OutputArray dst);
+
 //! performs affine transformation of each element of multi-channel input matrix
 CV_EXPORTS_W void transform(InputArray src, OutputArray dst, InputArray m );
+
 //! performs perspective transformation of each element of multi-channel input matrix
 CV_EXPORTS_W void perspectiveTransform(InputArray src, OutputArray dst, InputArray m );
 
 //! extends the symmetrical matrix from the lower half or from the upper half
-CV_EXPORTS_W void completeSymm(InputOutputArray mtx, bool lowerToUpper=false);
+CV_EXPORTS_W void completeSymm(InputOutputArray mtx, bool lowerToUpper = false);
+
 //! initializes scaled identity matrix
-CV_EXPORTS_W void setIdentity(InputOutputArray mtx, const Scalar& s=Scalar(1));
+CV_EXPORTS_W void setIdentity(InputOutputArray mtx, const Scalar& s = Scalar(1));
+
 //! computes determinant of a square matrix
 CV_EXPORTS_W double determinant(InputArray mtx);
+
 //! computes trace of a matrix
 CV_EXPORTS_W Scalar trace(InputArray mtx);
+
 //! computes inverse or pseudo-inverse matrix
-CV_EXPORTS_W double invert(InputArray src, OutputArray dst, int flags=DECOMP_LU);
+CV_EXPORTS_W double invert(InputArray src, OutputArray dst, int flags = DECOMP_LU);
+
 //! solves linear system or a least-square problem
 CV_EXPORTS_W bool solve(InputArray src1, InputArray src2,
-                        OutputArray dst, int flags=DECOMP_LU);
-
-enum
-{
-    SORT_EVERY_ROW=0,
-    SORT_EVERY_COLUMN=1,
-    SORT_ASCENDING=0,
-    SORT_DESCENDING=16
-};
+                        OutputArray dst, int flags = DECOMP_LU);
 
 //! sorts independently each matrix row or each matrix column
 CV_EXPORTS_W void sort(InputArray src, OutputArray dst, int flags);
+
 //! sorts independently each matrix row or each matrix column
 CV_EXPORTS_W void sortIdx(InputArray src, OutputArray dst, int flags);
+
 //! finds real roots of a cubic polynomial
 CV_EXPORTS_W int solveCubic(InputArray coeffs, OutputArray roots);
-//! finds real and complex roots of a polynomial
-CV_EXPORTS_W double solvePoly(InputArray coeffs, OutputArray roots, int maxIters=300);
-//! finds eigenvalues of a symmetric matrix
-CV_EXPORTS bool eigen(InputArray src, OutputArray eigenvalues, int lowindex=-1,
-                      int highindex=-1);
-//! finds eigenvalues and eigenvectors of a symmetric matrix
-CV_EXPORTS bool eigen(InputArray src, OutputArray eigenvalues,
-                      OutputArray eigenvectors,
-                      int lowindex=-1, int highindex=-1);
-CV_EXPORTS_W bool eigen(InputArray src, bool computeEigenvectors,
-                        OutputArray eigenvalues, OutputArray eigenvectors);
 
-enum
-{
-    COVAR_SCRAMBLED=0,
-    COVAR_NORMAL=1,
-    COVAR_USE_AVG=2,
-    COVAR_SCALE=4,
-    COVAR_ROWS=8,
-    COVAR_COLS=16
-};
+//! finds real and complex roots of a polynomial
+CV_EXPORTS_W double solvePoly(InputArray coeffs, OutputArray roots, int maxIters = 300);
+
+//! finds eigenvalues and eigenvectors of a symmetric matrix
+CV_EXPORTS_W bool eigen(InputArray src, OutputArray eigenvalues,
+                        OutputArray eigenvectors = noArray());
 
 //! computes covariation matrix of a set of samples
 CV_EXPORTS void calcCovarMatrix( const Mat* samples, int nsamples, Mat& covar, Mat& mean,
-                                 int flags, int ctype=CV_64F);
+                                 int flags, int ctype = CV_64F); //TODO: InputArrayOfArrays
+
 //! computes covariation matrix of a set of samples
 CV_EXPORTS_W void calcCovarMatrix( InputArray samples, OutputArray covar,
-                                   OutputArray mean, int flags, int ctype=CV_64F);
+                                   OutputArray mean, int flags, int ctype = CV_64F);
+
+CV_EXPORTS_W void PCACompute(InputArray data, InputOutputArray mean,
+                             OutputArray eigenvectors, int maxComponents = 0);
+
+CV_EXPORTS_W void PCACompute(InputArray data, InputOutputArray mean,
+                             OutputArray eigenvectors, double retainedVariance);
+
+CV_EXPORTS_W void PCAProject(InputArray data, InputArray mean,
+                             InputArray eigenvectors, OutputArray result);
+
+CV_EXPORTS_W void PCABackProject(InputArray data, InputArray mean,
+                                 InputArray eigenvectors, OutputArray result);
+
+//! computes SVD of src
+CV_EXPORTS_W void SVDecomp( InputArray src, OutputArray w, OutputArray u, OutputArray vt, int flags = 0 );
+
+//! performs back substitution for the previously computed SVD
+CV_EXPORTS_W void SVBackSubst( InputArray w, InputArray u, InputArray vt,
+                               InputArray rhs, OutputArray dst );
+
+//! computes Mahalanobis distance between two vectors: sqrt((v1-v2)'*icovar*(v1-v2)), where icovar is the inverse covariation matrix
+CV_EXPORTS_W double Mahalanobis(InputArray v1, InputArray v2, InputArray icovar);
+
+//! performs forward or inverse 1D or 2D Discrete Fourier Transformation
+CV_EXPORTS_W void dft(InputArray src, OutputArray dst, int flags = 0, int nonzeroRows = 0);
+
+//! performs inverse 1D or 2D Discrete Fourier Transformation
+CV_EXPORTS_W void idft(InputArray src, OutputArray dst, int flags = 0, int nonzeroRows = 0);
+
+//! performs forward or inverse 1D or 2D Discrete Cosine Transformation
+CV_EXPORTS_W void dct(InputArray src, OutputArray dst, int flags = 0);
+
+//! performs inverse 1D or 2D Discrete Cosine Transformation
+CV_EXPORTS_W void idct(InputArray src, OutputArray dst, int flags = 0);
+
+//! computes element-wise product of the two Fourier spectrums. The second spectrum can optionally be conjugated before the multiplication
+CV_EXPORTS_W void mulSpectrums(InputArray a, InputArray b, OutputArray c,
+                               int flags, bool conjB = false);
+
+//! computes the minimal vector size vecsize1 >= vecsize so that the dft() of the vector of length vecsize1 can be computed efficiently
+CV_EXPORTS_W int getOptimalDFTSize(int vecsize);
+
+//! clusters the input data using k-Means algorithm
+CV_EXPORTS_W double kmeans( InputArray data, int K, InputOutputArray bestLabels,
+                            TermCriteria criteria, int attempts,
+                            int flags, OutputArray centers = noArray() );
+
+//! returns the thread-local Random number generator
+CV_EXPORTS RNG& theRNG();
+
+//! fills array with uniformly-distributed random numbers from the range [low, high)
+CV_EXPORTS_W void randu(InputOutputArray dst, InputArray low, InputArray high);
+
+//! fills array with normally-distributed random numbers with the specified mean and the standard deviation
+CV_EXPORTS_W void randn(InputOutputArray dst, InputArray mean, InputArray stddev);
+
+//! shuffles the input array elements
+CV_EXPORTS_W void randShuffle(InputOutputArray dst, double iterFactor = 1., RNG* rng = 0);
+
+//! draws the line segment (pt1, pt2) in the image
+CV_EXPORTS_W void line(CV_IN_OUT Mat& img, Point pt1, Point pt2, const Scalar& color,
+                     int thickness = 1, int lineType = LINE_8, int shift = 0);
+
+//! draws the rectangle outline or a solid rectangle with the opposite corners pt1 and pt2 in the image
+CV_EXPORTS_W void rectangle(CV_IN_OUT Mat& img, Point pt1, Point pt2,
+                          const Scalar& color, int thickness = 1,
+                          int lineType = LINE_8, int shift = 0);
+
+//! draws the rectangle outline or a solid rectangle covering rec in the image
+CV_EXPORTS void rectangle(CV_IN_OUT Mat& img, Rect rec,
+                          const Scalar& color, int thickness = 1,
+                          int lineType = LINE_8, int shift = 0);
+
+//! draws the circle outline or a solid circle in the image
+CV_EXPORTS_W void circle(CV_IN_OUT Mat& img, Point center, int radius,
+                       const Scalar& color, int thickness = 1,
+                       int lineType = LINE_8, int shift = 0);
+
+//! draws an elliptic arc, ellipse sector or a rotated ellipse in the image
+CV_EXPORTS_W void ellipse(CV_IN_OUT Mat& img, Point center, Size axes,
+                        double angle, double startAngle, double endAngle,
+                        const Scalar& color, int thickness = 1,
+                        int lineType = LINE_8, int shift = 0);
+
+//! draws a rotated ellipse in the image
+CV_EXPORTS_W void ellipse(CV_IN_OUT Mat& img, const RotatedRect& box, const Scalar& color,
+                        int thickness = 1, int lineType = LINE_8);
+
+//! draws a filled convex polygon in the image
+CV_EXPORTS void fillConvexPoly(Mat& img, const Point* pts, int npts,
+                               const Scalar& color, int lineType = LINE_8,
+                               int shift = 0);
+
+CV_EXPORTS_W void fillConvexPoly(InputOutputArray img, InputArray points,
+                                 const Scalar& color, int lineType = LINE_8,
+                                 int shift = 0);
+
+//! fills an area bounded by one or more polygons
+CV_EXPORTS void fillPoly(Mat& img, const Point** pts,
+                         const int* npts, int ncontours,
+                         const Scalar& color, int lineType = LINE_8, int shift = 0,
+                         Point offset = Point() );
+
+CV_EXPORTS_W void fillPoly(InputOutputArray img, InputArrayOfArrays pts,
+                           const Scalar& color, int lineType = LINE_8, int shift = 0,
+                           Point offset = Point() );
+
+//! draws one or more polygonal curves
+CV_EXPORTS void polylines(Mat& img, const Point* const* pts, const int* npts,
+                          int ncontours, bool isClosed, const Scalar& color,
+                          int thickness = 1, int lineType = LINE_8, int shift = 0 );
+
+CV_EXPORTS_W void polylines(InputOutputArray img, InputArrayOfArrays pts,
+                            bool isClosed, const Scalar& color,
+                            int thickness = 1, int lineType = LINE_8, int shift = 0 );
+
+//! draws contours in the image
+CV_EXPORTS_W void drawContours( InputOutputArray image, InputArrayOfArrays contours,
+                              int contourIdx, const Scalar& color,
+                              int thickness = 1, int lineType = LINE_8,
+                              InputArray hierarchy = noArray(),
+                              int maxLevel = INT_MAX, Point offset = Point() );
+
+//! clips the line segment by the rectangle Rect(0, 0, imgSize.width, imgSize.height)
+CV_EXPORTS bool clipLine(Size imgSize, CV_IN_OUT Point& pt1, CV_IN_OUT Point& pt2);
+
+//! clips the line segment by the rectangle imgRect
+CV_EXPORTS_W bool clipLine(Rect imgRect, CV_OUT CV_IN_OUT Point& pt1, CV_OUT CV_IN_OUT Point& pt2);
+
+//! converts elliptic arc to a polygonal curve
+CV_EXPORTS_W void ellipse2Poly( Point center, Size axes, int angle,
+                                int arcStart, int arcEnd, int delta,
+                                CV_OUT std::vector<Point>& pts );
+
+//! renders text string in the image
+CV_EXPORTS_W void putText( Mat& img, const String& text, Point org,
+                         int fontFace, double fontScale, Scalar color,
+                         int thickness = 1, int lineType = LINE_8,
+                         bool bottomLeftOrigin = false );
+
+//! returns bounding box of the text string
+CV_EXPORTS_W Size getTextSize(const String& text, int fontFace,
+                            double fontScale, int thickness,
+                            CV_OUT int* baseLine);
 
 /*!
     Principal Component Analysis
@@ -2369,9 +610,9 @@ CV_EXPORTS_W void calcCovarMatrix( InputArray samples, OutputArray covar,
         PCA pca(pcaset, // pass the data
                 Mat(), // we do not have a pre-computed mean vector,
                        // so let the PCA engine to compute it
-                CV_PCA_DATA_AS_ROW, // indicate that the vectors
+                PCA::DATA_AS_ROW, // indicate that the vectors
                                     // are stored as matrix rows
-                                    // (use CV_PCA_DATA_AS_COL if the vectors are
+                                    // (use PCA::DATA_AS_COL if the vectors are
                                     // the matrix columns)
                 maxComponents // specify, how many principal components to retain
                 );
@@ -2401,39 +642,43 @@ CV_EXPORTS_W void calcCovarMatrix( InputArray samples, OutputArray covar,
 class CV_EXPORTS PCA
 {
 public:
+    enum { DATA_AS_ROW = 0,
+           DATA_AS_COL = 1,
+           USE_AVG     = 2
+         };
+
     //! default constructor
     PCA();
+
     //! the constructor that performs PCA
-    PCA(InputArray data, InputArray mean, int flags, int maxComponents=0);
+    PCA(InputArray data, InputArray mean, int flags, int maxComponents = 0);
     PCA(InputArray data, InputArray mean, int flags, double retainedVariance);
+
     //! operator that performs PCA. The previously stored data, if any, is released
-    PCA& operator()(InputArray data, InputArray mean, int flags, int maxComponents=0);
+    PCA& operator()(InputArray data, InputArray mean, int flags, int maxComponents = 0);
     PCA& operator()(InputArray data, InputArray mean, int flags, double retainedVariance);
+
     //! projects vector from the original space to the principal components subspace
     Mat project(InputArray vec) const;
+
     //! projects vector from the original space to the principal components subspace
     void project(InputArray vec, OutputArray result) const;
+
     //! reconstructs the original vector from the projection
     Mat backProject(InputArray vec) const;
+
     //! reconstructs the original vector from the projection
     void backProject(InputArray vec, OutputArray result) const;
 
+    //! write and load PCA matrix
+    void write(FileStorage& fs ) const;
+    void read(const FileNode& fs);
+
     Mat eigenvectors; //!< eigenvectors of the covariation matrix
     Mat eigenvalues; //!< eigenvalues of the covariation matrix
     Mat mean; //!< mean value subtracted before the projection and added after the back projection
 };
 
-CV_EXPORTS_W void PCACompute(InputArray data, InputOutputArray mean,
-                             OutputArray eigenvectors, int maxComponents=0);
-
-CV_EXPORTS_W void PCACompute(InputArray data, InputOutputArray mean,
-                             OutputArray eigenvectors, double retainedVariance);
-
-CV_EXPORTS_W void PCAProject(InputArray data, InputArray mean,
-                             InputArray eigenvectors, OutputArray result);
-
-CV_EXPORTS_W void PCABackProject(InputArray data, InputArray mean,
-                                 InputArray eigenvectors, OutputArray result);
 
 
 /*!
@@ -2452,163 +697,51 @@ CV_EXPORTS_W void PCABackProject(InputArray data, InputArray mean,
 class CV_EXPORTS SVD
 {
 public:
-    enum { MODIFY_A=1, NO_UV=2, FULL_UV=4 };
+    enum { MODIFY_A = 1,
+           NO_UV    = 2,
+           FULL_UV  = 4
+         };
+
     //! the default constructor
     SVD();
+
     //! the constructor that performs SVD
-    SVD( InputArray src, int flags=0 );
+    SVD( InputArray src, int flags = 0 );
+
     //! the operator that performs SVD. The previously allocated SVD::u, SVD::w are SVD::vt are released.
-    SVD& operator ()( InputArray src, int flags=0 );
+    SVD& operator ()( InputArray src, int flags = 0 );
 
     //! decomposes matrix and stores the results to user-provided matrices
     static void compute( InputArray src, OutputArray w,
-                         OutputArray u, OutputArray vt, int flags=0 );
+                         OutputArray u, OutputArray vt, int flags = 0 );
+
     //! computes singular values of a matrix
-    static void compute( InputArray src, OutputArray w, int flags=0 );
+    static void compute( InputArray src, OutputArray w, int flags = 0 );
+
     //! performs back substitution
     static void backSubst( InputArray w, InputArray u,
                            InputArray vt, InputArray rhs,
                            OutputArray dst );
 
-    template<typename _Tp, int m, int n, int nm> static void compute( const Matx<_Tp, m, n>& a,
-        Matx<_Tp, nm, 1>& w, Matx<_Tp, m, nm>& u, Matx<_Tp, n, nm>& vt );
-    template<typename _Tp, int m, int n, int nm> static void compute( const Matx<_Tp, m, n>& a,
-        Matx<_Tp, nm, 1>& w );
-    template<typename _Tp, int m, int n, int nm, int nb> static void backSubst( const Matx<_Tp, nm, 1>& w,
-        const Matx<_Tp, m, nm>& u, const Matx<_Tp, n, nm>& vt, const Matx<_Tp, m, nb>& rhs, Matx<_Tp, n, nb>& dst );
-
     //! finds dst = arg min_{|dst|=1} |m*dst|
     static void solveZ( InputArray src, OutputArray dst );
+
     //! performs back substitution, so that dst is the solution or pseudo-solution of m*dst = rhs, where m is the decomposed matrix
     void backSubst( InputArray rhs, OutputArray dst ) const;
 
+    template<typename _Tp, int m, int n, int nm> static
+    void compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w, Matx<_Tp, m, nm>& u, Matx<_Tp, n, nm>& vt );
+
+    template<typename _Tp, int m, int n, int nm> static
+    void compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w );
+
+    template<typename _Tp, int m, int n, int nm, int nb> static
+    void backSubst( const Matx<_Tp, nm, 1>& w, const Matx<_Tp, m, nm>& u, const Matx<_Tp, n, nm>& vt, const Matx<_Tp, m, nb>& rhs, Matx<_Tp, n, nb>& dst );
+
     Mat u, w, vt;
 };
 
-//! computes SVD of src
-CV_EXPORTS_W void SVDecomp( InputArray src, OutputArray w, OutputArray u, OutputArray vt, int flags=0 );
 
-//! performs back substitution for the previously computed SVD
-CV_EXPORTS_W void SVBackSubst( InputArray w, InputArray u, InputArray vt,
-                               InputArray rhs, OutputArray dst );
-
-//! computes Mahalanobis distance between two vectors: sqrt((v1-v2)'*icovar*(v1-v2)), where icovar is the inverse covariation matrix
-CV_EXPORTS_W double Mahalanobis(InputArray v1, InputArray v2, InputArray icovar);
-//! a synonym for Mahalanobis
-CV_EXPORTS double Mahalonobis(InputArray v1, InputArray v2, InputArray icovar);
-
-//! performs forward or inverse 1D or 2D Discrete Fourier Transformation
-CV_EXPORTS_W void dft(InputArray src, OutputArray dst, int flags=0, int nonzeroRows=0);
-//! performs inverse 1D or 2D Discrete Fourier Transformation
-CV_EXPORTS_W void idft(InputArray src, OutputArray dst, int flags=0, int nonzeroRows=0);
-//! performs forward or inverse 1D or 2D Discrete Cosine Transformation
-CV_EXPORTS_W void dct(InputArray src, OutputArray dst, int flags=0);
-//! performs inverse 1D or 2D Discrete Cosine Transformation
-CV_EXPORTS_W void idct(InputArray src, OutputArray dst, int flags=0);
-//! computes element-wise product of the two Fourier spectrums. The second spectrum can optionally be conjugated before the multiplication
-CV_EXPORTS_W void mulSpectrums(InputArray a, InputArray b, OutputArray c,
-                               int flags, bool conjB=false);
-//! computes the minimal vector size vecsize1 >= vecsize so that the dft() of the vector of length vecsize1 can be computed efficiently
-CV_EXPORTS_W int getOptimalDFTSize(int vecsize);
-
-/*!
- Various k-Means flags
-*/
-enum
-{
-    KMEANS_RANDOM_CENTERS=0, // Chooses random centers for k-Means initialization
-    KMEANS_PP_CENTERS=2,     // Uses k-Means++ algorithm for initialization
-    KMEANS_USE_INITIAL_LABELS=1 // Uses the user-provided labels for K-Means initialization
-};
-//! clusters the input data using k-Means algorithm
-CV_EXPORTS_W double kmeans( InputArray data, int K, InputOutputArray bestLabels,
-                            TermCriteria criteria, int attempts,
-                            int flags, OutputArray centers=noArray() );
-
-//! returns the thread-local Random number generator
-CV_EXPORTS RNG& theRNG();
-
-//! returns the next unifomly-distributed random number of the specified type
-template<typename _Tp> static inline _Tp randu() { return (_Tp)theRNG(); }
-
-//! fills array with uniformly-distributed random numbers from the range [low, high)
-CV_EXPORTS_W void randu(InputOutputArray dst, InputArray low, InputArray high);
-
-//! fills array with normally-distributed random numbers with the specified mean and the standard deviation
-CV_EXPORTS_W void randn(InputOutputArray dst, InputArray mean, InputArray stddev);
-
-//! shuffles the input array elements
-CV_EXPORTS void randShuffle(InputOutputArray dst, double iterFactor=1., RNG* rng=0);
-CV_EXPORTS_AS(randShuffle) void randShuffle_(InputOutputArray dst, double iterFactor=1.);
-
-//! draws the line segment (pt1, pt2) in the image
-CV_EXPORTS_W void line(CV_IN_OUT Mat& img, Point pt1, Point pt2, const Scalar& color,
-                     int thickness=1, int lineType=8, int shift=0);
-
-//! draws the rectangle outline or a solid rectangle with the opposite corners pt1 and pt2 in the image
-CV_EXPORTS_W void rectangle(CV_IN_OUT Mat& img, Point pt1, Point pt2,
-                          const Scalar& color, int thickness=1,
-                          int lineType=8, int shift=0);
-
-//! draws the rectangle outline or a solid rectangle covering rec in the image
-CV_EXPORTS void rectangle(CV_IN_OUT Mat& img, Rect rec,
-                          const Scalar& color, int thickness=1,
-                          int lineType=8, int shift=0);
-
-//! draws the circle outline or a solid circle in the image
-CV_EXPORTS_W void circle(CV_IN_OUT Mat& img, Point center, int radius,
-                       const Scalar& color, int thickness=1,
-                       int lineType=8, int shift=0);
-
-//! draws an elliptic arc, ellipse sector or a rotated ellipse in the image
-CV_EXPORTS_W void ellipse(CV_IN_OUT Mat& img, Point center, Size axes,
-                        double angle, double startAngle, double endAngle,
-                        const Scalar& color, int thickness=1,
-                        int lineType=8, int shift=0);
-
-//! draws a rotated ellipse in the image
-CV_EXPORTS_W void ellipse(CV_IN_OUT Mat& img, const RotatedRect& box, const Scalar& color,
-                        int thickness=1, int lineType=8);
-
-//! draws a filled convex polygon in the image
-CV_EXPORTS void fillConvexPoly(Mat& img, const Point* pts, int npts,
-                               const Scalar& color, int lineType=8,
-                               int shift=0);
-CV_EXPORTS_W void fillConvexPoly(InputOutputArray img, InputArray points,
-                                 const Scalar& color, int lineType=8,
-                                 int shift=0);
-
-//! fills an area bounded by one or more polygons
-CV_EXPORTS void fillPoly(Mat& img, const Point** pts,
-                         const int* npts, int ncontours,
-                         const Scalar& color, int lineType=8, int shift=0,
-                         Point offset=Point() );
-
-CV_EXPORTS_W void fillPoly(InputOutputArray img, InputArrayOfArrays pts,
-                           const Scalar& color, int lineType=8, int shift=0,
-                           Point offset=Point() );
-
-//! draws one or more polygonal curves
-CV_EXPORTS void polylines(Mat& img, const Point* const* pts, const int* npts,
-                          int ncontours, bool isClosed, const Scalar& color,
-                          int thickness=1, int lineType=8, int shift=0 );
-
-CV_EXPORTS_W void polylines(InputOutputArray img, InputArrayOfArrays pts,
-                            bool isClosed, const Scalar& color,
-                            int thickness=1, int lineType=8, int shift=0 );
-
-//! draws contours in the image
-CV_EXPORTS_W void drawContours( InputOutputArray image, InputArrayOfArrays contours,
-                              int contourIdx, const Scalar& color,
-                              int thickness=1, int lineType=8,
-                              InputArray hierarchy=noArray(),
-                              int maxLevel=INT_MAX, Point offset=Point() );
-
-//! clips the line segment by the rectangle Rect(0, 0, imgSize.width, imgSize.height)
-CV_EXPORTS bool clipLine(Size imgSize, CV_IN_OUT Point& pt1, CV_IN_OUT Point& pt2);
-
-//! clips the line segment by the rectangle imgRect
-CV_EXPORTS_W bool clipLine(Rect imgRect, CV_OUT CV_IN_OUT Point& pt1, CV_OUT CV_IN_OUT Point& pt2);
 
 /*!
    Line iterator class
@@ -2621,7 +754,7 @@ class CV_EXPORTS LineIterator
 public:
     //! intializes the iterator
     LineIterator( const Mat& img, Point pt1, Point pt2,
-                  int connectivity=8, bool leftToRight=false );
+                  int connectivity = 8, bool leftToRight = false );
     //! returns pointer to the current pixel
     uchar* operator *();
     //! prefix increment operator (++it). shifts iterator to the next pixel
@@ -2639,1191 +772,7 @@ public:
     int minusStep, plusStep;
 };
 
-//! converts elliptic arc to a polygonal curve
-CV_EXPORTS_W void ellipse2Poly( Point center, Size axes, int angle,
-                                int arcStart, int arcEnd, int delta,
-                                CV_OUT std::vector<Point>& pts );
 
-enum
-{
-    FONT_HERSHEY_SIMPLEX = 0,
-    FONT_HERSHEY_PLAIN = 1,
-    FONT_HERSHEY_DUPLEX = 2,
-    FONT_HERSHEY_COMPLEX = 3,
-    FONT_HERSHEY_TRIPLEX = 4,
-    FONT_HERSHEY_COMPLEX_SMALL = 5,
-    FONT_HERSHEY_SCRIPT_SIMPLEX = 6,
-    FONT_HERSHEY_SCRIPT_COMPLEX = 7,
-    FONT_ITALIC = 16
-};
-
-//! renders text string in the image
-CV_EXPORTS_W void putText( Mat& img, const std::string& text, Point org,
-                         int fontFace, double fontScale, Scalar color,
-                         int thickness=1, int lineType=8,
-                         bool bottomLeftOrigin=false );
-
-//! returns bounding box of the text string
-CV_EXPORTS_W Size getTextSize(const std::string& text, int fontFace,
-                            double fontScale, int thickness,
-                            CV_OUT int* baseLine);
-
-///////////////////////////////// Mat_<_Tp> ////////////////////////////////////
-
-/*!
- Template matrix class derived from Mat
-
- The class Mat_ is a "thin" template wrapper on top of cv::Mat. It does not have any extra data fields,
- nor it or cv::Mat have any virtual methods and thus references or pointers to these two classes
- can be safely converted one to another. But do it with care, for example:
-
- \code
- // create 100x100 8-bit matrix
- Mat M(100,100,CV_8U);
- // this will compile fine. no any data conversion will be done.
- Mat_<float>& M1 = (Mat_<float>&)M;
- // the program will likely crash at the statement below
- M1(99,99) = 1.f;
- \endcode
-
- While cv::Mat is sufficient in most cases, cv::Mat_ can be more convenient if you use a lot of element
- access operations and if you know matrix type at compile time.
- Note that cv::Mat::at<_Tp>(int y, int x) and cv::Mat_<_Tp>::operator ()(int y, int x) do absolutely the
- same thing and run at the same speed, but the latter is certainly shorter:
-
- \code
- Mat_<double> M(20,20);
- for(int i = 0; i < M.rows; i++)
-    for(int j = 0; j < M.cols; j++)
-       M(i,j) = 1./(i+j+1);
- Mat E, V;
- eigen(M,E,V);
- cout << E.at<double>(0,0)/E.at<double>(M.rows-1,0);
- \endcode
-
- It is easy to use Mat_ for multi-channel images/matrices - just pass cv::Vec as cv::Mat_ template parameter:
-
- \code
- // allocate 320x240 color image and fill it with green (in RGB space)
- Mat_<Vec3b> img(240, 320, Vec3b(0,255,0));
- // now draw a diagonal white line
- for(int i = 0; i < 100; i++)
-     img(i,i)=Vec3b(255,255,255);
- // and now modify the 2nd (red) channel of each pixel
- for(int i = 0; i < img.rows; i++)
-    for(int j = 0; j < img.cols; j++)
-       img(i,j)[2] ^= (uchar)(i ^ j); // img(y,x)[c] accesses c-th channel of the pixel (x,y)
- \endcode
-*/
-template<typename _Tp> class CV_EXPORTS Mat_ : public Mat
-{
-public:
-    typedef _Tp value_type;
-    typedef typename DataType<_Tp>::channel_type channel_type;
-    typedef MatIterator_<_Tp> iterator;
-    typedef MatConstIterator_<_Tp> const_iterator;
-
-    //! default constructor
-    Mat_();
-    //! equivalent to Mat(_rows, _cols, DataType<_Tp>::type)
-    Mat_(int _rows, int _cols);
-    //! constructor that sets each matrix element to specified value
-    Mat_(int _rows, int _cols, const _Tp& value);
-    //! equivalent to Mat(_size, DataType<_Tp>::type)
-    explicit Mat_(Size _size);
-    //! constructor that sets each matrix element to specified value
-    Mat_(Size _size, const _Tp& value);
-    //! n-dim array constructor
-    Mat_(int _ndims, const int* _sizes);
-    //! n-dim array constructor that sets each matrix element to specified value
-    Mat_(int _ndims, const int* _sizes, const _Tp& value);
-    //! copy/conversion contructor. If m is of different type, it's converted
-    Mat_(const Mat& m);
-    //! copy constructor
-    Mat_(const Mat_& m);
-    //! constructs a matrix on top of user-allocated data. step is in bytes(!!!), regardless of the type
-    Mat_(int _rows, int _cols, _Tp* _data, size_t _step=AUTO_STEP);
-    //! constructs n-dim matrix on top of user-allocated data. steps are in bytes(!!!), regardless of the type
-    Mat_(int _ndims, const int* _sizes, _Tp* _data, const size_t* _steps=0);
-    //! selects a submatrix
-    Mat_(const Mat_& m, const Range& rowRange, const Range& colRange=Range::all());
-    //! selects a submatrix
-    Mat_(const Mat_& m, const Rect& roi);
-    //! selects a submatrix, n-dim version
-    Mat_(const Mat_& m, const Range* ranges);
-    //! from a matrix expression
-    explicit Mat_(const MatExpr& e);
-    //! makes a matrix out of Vec, std::vector, Point_ or Point3_. The matrix will have a single column
-    explicit Mat_(const std::vector<_Tp>& vec, bool copyData=false);
-    template<int n> explicit Mat_(const Vec<typename DataType<_Tp>::channel_type, n>& vec, bool copyData=true);
-    template<int m, int n> explicit Mat_(const Matx<typename DataType<_Tp>::channel_type, m, n>& mtx, bool copyData=true);
-    explicit Mat_(const Point_<typename DataType<_Tp>::channel_type>& pt, bool copyData=true);
-    explicit Mat_(const Point3_<typename DataType<_Tp>::channel_type>& pt, bool copyData=true);
-    explicit Mat_(const MatCommaInitializer_<_Tp>& commaInitializer);
-
-    Mat_& operator = (const Mat& m);
-    Mat_& operator = (const Mat_& m);
-    //! set all the elements to s.
-    Mat_& operator = (const _Tp& s);
-    //! assign a matrix expression
-    Mat_& operator = (const MatExpr& e);
-
-    //! iterators; they are smart enough to skip gaps in the end of rows
-    iterator begin();
-    iterator end();
-    const_iterator begin() const;
-    const_iterator end() const;
-
-    //! equivalent to Mat::create(_rows, _cols, DataType<_Tp>::type)
-    void create(int _rows, int _cols);
-    //! equivalent to Mat::create(_size, DataType<_Tp>::type)
-    void create(Size _size);
-    //! equivalent to Mat::create(_ndims, _sizes, DatType<_Tp>::type)
-    void create(int _ndims, const int* _sizes);
-    //! cross-product
-    Mat_ cross(const Mat_& m) const;
-    //! data type conversion
-    template<typename T2> operator Mat_<T2>() const;
-    //! overridden forms of Mat::row() etc.
-    Mat_ row(int y) const;
-    Mat_ col(int x) const;
-    Mat_ diag(int d=0) const;
-    Mat_ clone() const;
-
-    //! overridden forms of Mat::elemSize() etc.
-    size_t elemSize() const;
-    size_t elemSize1() const;
-    int type() const;
-    int depth() const;
-    int channels() const;
-    size_t step1(int i=0) const;
-    //! returns step()/sizeof(_Tp)
-    size_t stepT(int i=0) const;
-
-    //! overridden forms of Mat::zeros() etc. Data type is omitted, of course
-    static MatExpr zeros(int rows, int cols);
-    static MatExpr zeros(Size size);
-    static MatExpr zeros(int _ndims, const int* _sizes);
-    static MatExpr ones(int rows, int cols);
-    static MatExpr ones(Size size);
-    static MatExpr ones(int _ndims, const int* _sizes);
-    static MatExpr eye(int rows, int cols);
-    static MatExpr eye(Size size);
-
-    //! some more overriden methods
-    Mat_& adjustROI( int dtop, int dbottom, int dleft, int dright );
-    Mat_ operator()( const Range& rowRange, const Range& colRange ) const;
-    Mat_ operator()( const Rect& roi ) const;
-    Mat_ operator()( const Range* ranges ) const;
-
-    //! more convenient forms of row and element access operators
-    _Tp* operator [](int y);
-    const _Tp* operator [](int y) const;
-
-    //! returns reference to the specified element
-    _Tp& operator ()(const int* idx);
-    //! returns read-only reference to the specified element
-    const _Tp& operator ()(const int* idx) const;
-
-    //! returns reference to the specified element
-    template<int n> _Tp& operator ()(const Vec<int, n>& idx);
-    //! returns read-only reference to the specified element
-    template<int n> const _Tp& operator ()(const Vec<int, n>& idx) const;
-
-    //! returns reference to the specified element (1D case)
-    _Tp& operator ()(int idx0);
-    //! returns read-only reference to the specified element (1D case)
-    const _Tp& operator ()(int idx0) const;
-    //! returns reference to the specified element (2D case)
-    _Tp& operator ()(int idx0, int idx1);
-    //! returns read-only reference to the specified element (2D case)
-    const _Tp& operator ()(int idx0, int idx1) const;
-    //! returns reference to the specified element (3D case)
-    _Tp& operator ()(int idx0, int idx1, int idx2);
-    //! returns read-only reference to the specified element (3D case)
-    const _Tp& operator ()(int idx0, int idx1, int idx2) const;
-
-    _Tp& operator ()(Point pt);
-    const _Tp& operator ()(Point pt) const;
-
-    //! conversion to vector.
-    operator std::vector<_Tp>() const;
-    //! conversion to Vec
-    template<int n> operator Vec<typename DataType<_Tp>::channel_type, n>() const;
-    //! conversion to Matx
-    template<int m, int n> operator Matx<typename DataType<_Tp>::channel_type, m, n>() const;
-};
-
-typedef Mat_<uchar> Mat1b;
-typedef Mat_<Vec2b> Mat2b;
-typedef Mat_<Vec3b> Mat3b;
-typedef Mat_<Vec4b> Mat4b;
-
-typedef Mat_<short> Mat1s;
-typedef Mat_<Vec2s> Mat2s;
-typedef Mat_<Vec3s> Mat3s;
-typedef Mat_<Vec4s> Mat4s;
-
-typedef Mat_<ushort> Mat1w;
-typedef Mat_<Vec2w> Mat2w;
-typedef Mat_<Vec3w> Mat3w;
-typedef Mat_<Vec4w> Mat4w;
-
-typedef Mat_<int>   Mat1i;
-typedef Mat_<Vec2i> Mat2i;
-typedef Mat_<Vec3i> Mat3i;
-typedef Mat_<Vec4i> Mat4i;
-
-typedef Mat_<float> Mat1f;
-typedef Mat_<Vec2f> Mat2f;
-typedef Mat_<Vec3f> Mat3f;
-typedef Mat_<Vec4f> Mat4f;
-
-typedef Mat_<double> Mat1d;
-typedef Mat_<Vec2d> Mat2d;
-typedef Mat_<Vec3d> Mat3d;
-typedef Mat_<Vec4d> Mat4d;
-
-//////////// Iterators & Comma initializers //////////////////
-
-class CV_EXPORTS MatConstIterator
-{
-public:
-    typedef uchar* value_type;
-    typedef ptrdiff_t difference_type;
-    typedef const uchar** pointer;
-    typedef uchar* reference;
-    typedef std::random_access_iterator_tag iterator_category;
-
-    //! default constructor
-    MatConstIterator();
-    //! constructor that sets the iterator to the beginning of the matrix
-    MatConstIterator(const Mat* _m);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator(const Mat* _m, int _row, int _col=0);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator(const Mat* _m, Point _pt);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator(const Mat* _m, const int* _idx);
-    //! copy constructor
-    MatConstIterator(const MatConstIterator& it);
-
-    //! copy operator
-    MatConstIterator& operator = (const MatConstIterator& it);
-    //! returns the current matrix element
-    uchar* operator *() const;
-    //! returns the i-th matrix element, relative to the current
-    uchar* operator [](ptrdiff_t i) const;
-
-    //! shifts the iterator forward by the specified number of elements
-    MatConstIterator& operator += (ptrdiff_t ofs);
-    //! shifts the iterator backward by the specified number of elements
-    MatConstIterator& operator -= (ptrdiff_t ofs);
-    //! decrements the iterator
-    MatConstIterator& operator --();
-    //! decrements the iterator
-    MatConstIterator operator --(int);
-    //! increments the iterator
-    MatConstIterator& operator ++();
-    //! increments the iterator
-    MatConstIterator operator ++(int);
-    //! returns the current iterator position
-    Point pos() const;
-    //! returns the current iterator position
-    void pos(int* _idx) const;
-    ptrdiff_t lpos() const;
-    void seek(ptrdiff_t ofs, bool relative=false);
-    void seek(const int* _idx, bool relative=false);
-
-    const Mat* m;
-    size_t elemSize;
-    uchar* ptr;
-    uchar* sliceStart;
-    uchar* sliceEnd;
-};
-
-/*!
- Matrix read-only iterator
-
- */
-template<typename _Tp>
-class CV_EXPORTS MatConstIterator_ : public MatConstIterator
-{
-public:
-    typedef _Tp value_type;
-    typedef ptrdiff_t difference_type;
-    typedef const _Tp* pointer;
-    typedef const _Tp& reference;
-    typedef std::random_access_iterator_tag iterator_category;
-
-    //! default constructor
-    MatConstIterator_();
-    //! constructor that sets the iterator to the beginning of the matrix
-    MatConstIterator_(const Mat_<_Tp>* _m);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator_(const Mat_<_Tp>* _m, int _row, int _col=0);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator_(const Mat_<_Tp>* _m, Point _pt);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatConstIterator_(const Mat_<_Tp>* _m, const int* _idx);
-    //! copy constructor
-    MatConstIterator_(const MatConstIterator_& it);
-
-    //! copy operator
-    MatConstIterator_& operator = (const MatConstIterator_& it);
-    //! returns the current matrix element
-    _Tp operator *() const;
-    //! returns the i-th matrix element, relative to the current
-    _Tp operator [](ptrdiff_t i) const;
-
-    //! shifts the iterator forward by the specified number of elements
-    MatConstIterator_& operator += (ptrdiff_t ofs);
-    //! shifts the iterator backward by the specified number of elements
-    MatConstIterator_& operator -= (ptrdiff_t ofs);
-    //! decrements the iterator
-    MatConstIterator_& operator --();
-    //! decrements the iterator
-    MatConstIterator_ operator --(int);
-    //! increments the iterator
-    MatConstIterator_& operator ++();
-    //! increments the iterator
-    MatConstIterator_ operator ++(int);
-    //! returns the current iterator position
-    Point pos() const;
-};
-
-
-/*!
- Matrix read-write iterator
-
-*/
-template<typename _Tp>
-class CV_EXPORTS MatIterator_ : public MatConstIterator_<_Tp>
-{
-public:
-    typedef _Tp* pointer;
-    typedef _Tp& reference;
-    typedef std::random_access_iterator_tag iterator_category;
-
-    //! the default constructor
-    MatIterator_();
-    //! constructor that sets the iterator to the beginning of the matrix
-    MatIterator_(Mat_<_Tp>* _m);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatIterator_(Mat_<_Tp>* _m, int _row, int _col=0);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatIterator_(const Mat_<_Tp>* _m, Point _pt);
-    //! constructor that sets the iterator to the specified element of the matrix
-    MatIterator_(const Mat_<_Tp>* _m, const int* _idx);
-    //! copy constructor
-    MatIterator_(const MatIterator_& it);
-    //! copy operator
-    MatIterator_& operator = (const MatIterator_<_Tp>& it );
-
-    //! returns the current matrix element
-    _Tp& operator *() const;
-    //! returns the i-th matrix element, relative to the current
-    _Tp& operator [](ptrdiff_t i) const;
-
-    //! shifts the iterator forward by the specified number of elements
-    MatIterator_& operator += (ptrdiff_t ofs);
-    //! shifts the iterator backward by the specified number of elements
-    MatIterator_& operator -= (ptrdiff_t ofs);
-    //! decrements the iterator
-    MatIterator_& operator --();
-    //! decrements the iterator
-    MatIterator_ operator --(int);
-    //! increments the iterator
-    MatIterator_& operator ++();
-    //! increments the iterator
-    MatIterator_ operator ++(int);
-};
-
-template<typename _Tp> class CV_EXPORTS MatOp_Iter_;
-
-/*!
- Comma-separated Matrix Initializer
-
- The class instances are usually not created explicitly.
- Instead, they are created on "matrix << firstValue" operator.
-
- The sample below initializes 2x2 rotation matrix:
-
- \code
- double angle = 30, a = cos(angle*CV_PI/180), b = sin(angle*CV_PI/180);
- Mat R = (Mat_<double>(2,2) << a, -b, b, a);
- \endcode
-*/
-template<typename _Tp> class CV_EXPORTS MatCommaInitializer_
-{
-public:
-    //! the constructor, created by "matrix << firstValue" operator, where matrix is cv::Mat
-    MatCommaInitializer_(Mat_<_Tp>* _m);
-    //! the operator that takes the next value and put it to the matrix
-    template<typename T2> MatCommaInitializer_<_Tp>& operator , (T2 v);
-    //! another form of conversion operator
-    Mat_<_Tp> operator *() const;
-    operator Mat_<_Tp>() const;
-protected:
-    MatIterator_<_Tp> it;
-};
-
-
-template<typename _Tp, int m, int n> class CV_EXPORTS MatxCommaInitializer
-{
-public:
-    MatxCommaInitializer(Matx<_Tp, m, n>* _mtx);
-    template<typename T2> MatxCommaInitializer<_Tp, m, n>& operator , (T2 val);
-    Matx<_Tp, m, n> operator *() const;
-
-    Matx<_Tp, m, n>* dst;
-    int idx;
-};
-
-template<typename _Tp, int m> class CV_EXPORTS VecCommaInitializer : public MatxCommaInitializer<_Tp, m, 1>
-{
-public:
-    VecCommaInitializer(Vec<_Tp, m>* _vec);
-    template<typename T2> VecCommaInitializer<_Tp, m>& operator , (T2 val);
-    Vec<_Tp, m> operator *() const;
-};
-
-/*!
- Automatically Allocated Buffer Class
-
- The class is used for temporary buffers in functions and methods.
- If a temporary buffer is usually small (a few K's of memory),
- but its size depends on the parameters, it makes sense to create a small
- fixed-size array on stack and use it if it's large enough. If the required buffer size
- is larger than the fixed size, another buffer of sufficient size is allocated dynamically
- and released after the processing. Therefore, in typical cases, when the buffer size is small,
- there is no overhead associated with malloc()/free().
- At the same time, there is no limit on the size of processed data.
-
- This is what AutoBuffer does. The template takes 2 parameters - type of the buffer elements and
- the number of stack-allocated elements. Here is how the class is used:
-
- \code
- void my_func(const cv::Mat& m)
- {
-    cv::AutoBuffer<float> buf; // create automatic buffer containing 1000 floats
-
-    buf.allocate(m.rows); // if m.rows <= 1000, the pre-allocated buffer is used,
-                          // otherwise the buffer of "m.rows" floats will be allocated
-                          // dynamically and deallocated in cv::AutoBuffer destructor
-    ...
- }
- \endcode
-*/
-template<typename _Tp, size_t fixed_size> class CV_EXPORTS AutoBuffer
-{
-public:
-    typedef _Tp value_type;
-
-    //! the default contructor
-    AutoBuffer();
-    //! constructor taking the real buffer size
-    AutoBuffer(size_t _size);
-
-    //! the copy constructor
-    AutoBuffer(const AutoBuffer<_Tp, fixed_size>& buf);
-    //! the assignment operator
-    AutoBuffer<_Tp, fixed_size>& operator = (const AutoBuffer<_Tp, fixed_size>& buf);
-
-    //! destructor. calls deallocate()
-    ~AutoBuffer();
-
-    //! allocates the new buffer of size _size. if the _size is small enough, stack-allocated buffer is used
-    void allocate(size_t _size);
-    //! deallocates the buffer if it was dynamically allocated
-    void deallocate();
-    //! resizes the buffer and preserves the content
-    void resize(size_t _size);
-    //! returns the current buffer size
-    size_t size() const;
-    //! returns pointer to the real buffer, stack-allocated or head-allocated
-    operator _Tp* ();
-    //! returns read-only pointer to the real buffer, stack-allocated or head-allocated
-    operator const _Tp* () const;
-
-protected:
-    //! pointer to the real buffer, can point to buf if the buffer is small enough
-    _Tp* ptr;
-    //! size of the real buffer
-    size_t sz;
-    //! pre-allocated buffer
-    _Tp buf[fixed_size];
-};
-
-/////////////////////////// multi-dimensional dense matrix //////////////////////////
-
-/*!
- n-Dimensional Dense Matrix Iterator Class.
-
- The class cv::NAryMatIterator is used for iterating over one or more n-dimensional dense arrays (cv::Mat's).
-
- The iterator is completely different from cv::Mat_ and cv::SparseMat_ iterators.
- It iterates through the slices (or planes), not the elements, where "slice" is a continuous part of the arrays.
-
- Here is the example on how the iterator can be used to normalize 3D histogram:
-
- \code
- void normalizeColorHist(Mat& hist)
- {
- #if 1
-     // intialize iterator (the style is different from STL).
-     // after initialization the iterator will contain
-     // the number of slices or planes
-     // the iterator will go through
-     Mat* arrays[] = { &hist, 0 };
-     Mat planes[1];
-     NAryMatIterator it(arrays, planes);
-     double s = 0;
-     // iterate through the matrix. on each iteration
-     // it.planes[i] (of type Mat) will be set to the current plane of
-     // i-th n-dim matrix passed to the iterator constructor.
-     for(int p = 0; p < it.nplanes; p++, ++it)
-        s += sum(it.planes[0])[0];
-     it = NAryMatIterator(hist);
-     s = 1./s;
-     for(int p = 0; p < it.nplanes; p++, ++it)
-        it.planes[0] *= s;
- #elif 1
-     // this is a shorter implementation of the above
-     // using built-in operations on Mat
-     double s = sum(hist)[0];
-     hist.convertTo(hist, hist.type(), 1./s, 0);
- #else
-     // and this is even shorter one
-     // (assuming that the histogram elements are non-negative)
-     normalize(hist, hist, 1, 0, NORM_L1);
- #endif
- }
- \endcode
-
- You can iterate through several matrices simultaneously as long as they have the same geometry
- (dimensionality and all the dimension sizes are the same), which is useful for binary
- and n-ary operations on such matrices. Just pass those matrices to cv::MatNDIterator.
- Then, during the iteration it.planes[0], it.planes[1], ... will
- be the slices of the corresponding matrices
-*/
-class CV_EXPORTS NAryMatIterator
-{
-public:
-    //! the default constructor
-    NAryMatIterator();
-    //! the full constructor taking arbitrary number of n-dim matrices
-    NAryMatIterator(const Mat** arrays, uchar** ptrs, int narrays=-1);
-    //! the full constructor taking arbitrary number of n-dim matrices
-    NAryMatIterator(const Mat** arrays, Mat* planes, int narrays=-1);
-    //! the separate iterator initialization method
-    void init(const Mat** arrays, Mat* planes, uchar** ptrs, int narrays=-1);
-
-    //! proceeds to the next plane of every iterated matrix
-    NAryMatIterator& operator ++();
-    //! proceeds to the next plane of every iterated matrix (postfix increment operator)
-    NAryMatIterator operator ++(int);
-
-    //! the iterated arrays
-    const Mat** arrays;
-    //! the current planes
-    Mat* planes;
-    //! data pointers
-    uchar** ptrs;
-    //! the number of arrays
-    int narrays;
-    //! the number of hyper-planes that the iterator steps through
-    size_t nplanes;
-    //! the size of each segment (in elements)
-    size_t size;
-protected:
-    int iterdepth;
-    size_t idx;
-};
-
-//typedef NAryMatIterator NAryMatNDIterator;
-
-typedef void (*ConvertData)(const void* from, void* to, int cn);
-typedef void (*ConvertScaleData)(const void* from, void* to, int cn, double alpha, double beta);
-
-//! returns the function for converting pixels from one data type to another
-CV_EXPORTS ConvertData getConvertElem(int fromType, int toType);
-//! returns the function for converting pixels from one data type to another with the optional scaling
-CV_EXPORTS ConvertScaleData getConvertScaleElem(int fromType, int toType);
-
-
-/////////////////////////// multi-dimensional sparse matrix //////////////////////////
-
-class SparseMatIterator;
-class SparseMatConstIterator;
-template<typename _Tp> class SparseMatIterator_;
-template<typename _Tp> class SparseMatConstIterator_;
-
-/*!
- Sparse matrix class.
-
- The class represents multi-dimensional sparse numerical arrays. Such a sparse array can store elements
- of any type that cv::Mat is able to store. "Sparse" means that only non-zero elements
- are stored (though, as a result of some operations on a sparse matrix, some of its stored elements
- can actually become 0. It's user responsibility to detect such elements and delete them using cv::SparseMat::erase().
- The non-zero elements are stored in a hash table that grows when it's filled enough,
- so that the search time remains O(1) in average. Elements can be accessed using the following methods:
-
- <ol>
- <li>Query operations: cv::SparseMat::ptr() and the higher-level cv::SparseMat::ref(),
-      cv::SparseMat::value() and cv::SparseMat::find, for example:
- \code
- const int dims = 5;
- int size[] = {10, 10, 10, 10, 10};
- SparseMat sparse_mat(dims, size, CV_32F);
- for(int i = 0; i < 1000; i++)
- {
-     int idx[dims];
-     for(int k = 0; k < dims; k++)
-        idx[k] = rand()%sparse_mat.size(k);
-     sparse_mat.ref<float>(idx) += 1.f;
- }
- \endcode
-
- <li>Sparse matrix iterators. Like cv::Mat iterators and unlike cv::Mat iterators, the sparse matrix iterators are STL-style,
- that is, the iteration is done as following:
- \code
- // prints elements of a sparse floating-point matrix and the sum of elements.
- SparseMatConstIterator_<float>
-        it = sparse_mat.begin<float>(),
-        it_end = sparse_mat.end<float>();
- double s = 0;
- int dims = sparse_mat.dims();
- for(; it != it_end; ++it)
- {
-     // print element indices and the element value
-     const Node* n = it.node();
-     printf("(")
-     for(int i = 0; i < dims; i++)
-        printf("%3d%c", n->idx[i], i < dims-1 ? ',' : ')');
-     printf(": %f\n", *it);
-     s += *it;
- }
- printf("Element sum is %g\n", s);
- \endcode
- If you run this loop, you will notice that elements are enumerated
- in no any logical order (lexicographical etc.),
- they come in the same order as they stored in the hash table, i.e. semi-randomly.
-
- You may collect pointers to the nodes and sort them to get the proper ordering.
- Note, however, that pointers to the nodes may become invalid when you add more
- elements to the matrix; this is because of possible buffer reallocation.
-
- <li>A combination of the above 2 methods when you need to process 2 or more sparse
- matrices simultaneously, e.g. this is how you can compute unnormalized
- cross-correlation of the 2 floating-point sparse matrices:
- \code
- double crossCorr(const SparseMat& a, const SparseMat& b)
- {
-     const SparseMat *_a = &a, *_b = &b;
-     // if b contains less elements than a,
-     // it's faster to iterate through b
-     if(_a->nzcount() > _b->nzcount())
-        std::swap(_a, _b);
-     SparseMatConstIterator_<float> it = _a->begin<float>(),
-                                    it_end = _a->end<float>();
-     double ccorr = 0;
-     for(; it != it_end; ++it)
-     {
-         // take the next element from the first matrix
-         float avalue = *it;
-         const Node* anode = it.node();
-         // and try to find element with the same index in the second matrix.
-         // since the hash value depends only on the element index,
-         // we reuse hashvalue stored in the node
-         float bvalue = _b->value<float>(anode->idx,&anode->hashval);
-         ccorr += avalue*bvalue;
-     }
-     return ccorr;
- }
- \endcode
- </ol>
-*/
-class CV_EXPORTS SparseMat
-{
-public:
-    typedef SparseMatIterator iterator;
-    typedef SparseMatConstIterator const_iterator;
-
-    //! the sparse matrix header
-    struct CV_EXPORTS Hdr
-    {
-        Hdr(int _dims, const int* _sizes, int _type);
-        void clear();
-        int refcount;
-        int dims;
-        int valueOffset;
-        size_t nodeSize;
-        size_t nodeCount;
-        size_t freeList;
-        std::vector<uchar> pool;
-        std::vector<size_t> hashtab;
-        int size[CV_MAX_DIM];
-    };
-
-    //! sparse matrix node - element of a hash table
-    struct CV_EXPORTS Node
-    {
-        //! hash value
-        size_t hashval;
-        //! index of the next node in the same hash table entry
-        size_t next;
-        //! index of the matrix element
-        int idx[CV_MAX_DIM];
-    };
-
-    //! default constructor
-    SparseMat();
-    //! creates matrix of the specified size and type
-    SparseMat(int dims, const int* _sizes, int _type);
-    //! copy constructor
-    SparseMat(const SparseMat& m);
-    //! converts dense 2d matrix to the sparse form
-    /*!
-     \param m the input matrix
-     \param try1d if true and m is a single-column matrix (Nx1),
-            then the sparse matrix will be 1-dimensional.
-    */
-    explicit SparseMat(const Mat& m);
-    //! converts old-style sparse matrix to the new-style. All the data is copied
-    SparseMat(const CvSparseMat* m);
-    //! the destructor
-    ~SparseMat();
-
-    //! assignment operator. This is O(1) operation, i.e. no data is copied
-    SparseMat& operator = (const SparseMat& m);
-    //! equivalent to the corresponding constructor
-    SparseMat& operator = (const Mat& m);
-
-    //! creates full copy of the matrix
-    SparseMat clone() const;
-
-    //! copies all the data to the destination matrix. All the previous content of m is erased
-    void copyTo( SparseMat& m ) const;
-    //! converts sparse matrix to dense matrix.
-    void copyTo( Mat& m ) const;
-    //! multiplies all the matrix elements by the specified scale factor alpha and converts the results to the specified data type
-    void convertTo( SparseMat& m, int rtype, double alpha=1 ) const;
-    //! converts sparse matrix to dense n-dim matrix with optional type conversion and scaling.
-    /*!
-      \param rtype The output matrix data type. When it is =-1, the output array will have the same data type as (*this)
-      \param alpha The scale factor
-      \param beta The optional delta added to the scaled values before the conversion
-    */
-    void convertTo( Mat& m, int rtype, double alpha=1, double beta=0 ) const;
-
-    // not used now
-    void assignTo( SparseMat& m, int type=-1 ) const;
-
-    //! reallocates sparse matrix.
-    /*!
-        If the matrix already had the proper size and type,
-        it is simply cleared with clear(), otherwise,
-        the old matrix is released (using release()) and the new one is allocated.
-    */
-    void create(int dims, const int* _sizes, int _type);
-    //! sets all the sparse matrix elements to 0, which means clearing the hash table.
-    void clear();
-    //! manually increments the reference counter to the header.
-    void addref();
-    // decrements the header reference counter. When the counter reaches 0, the header and all the underlying data are deallocated.
-    void release();
-
-    //! converts sparse matrix to the old-style representation; all the elements are copied.
-    operator CvSparseMat*() const;
-    //! returns the size of each element in bytes (not including the overhead - the space occupied by SparseMat::Node elements)
-    size_t elemSize() const;
-    //! returns elemSize()/channels()
-    size_t elemSize1() const;
-
-    //! returns type of sparse matrix elements
-    int type() const;
-    //! returns the depth of sparse matrix elements
-    int depth() const;
-    //! returns the number of channels
-    int channels() const;
-
-    //! returns the array of sizes, or NULL if the matrix is not allocated
-    const int* size() const;
-    //! returns the size of i-th matrix dimension (or 0)
-    int size(int i) const;
-    //! returns the matrix dimensionality
-    int dims() const;
-    //! returns the number of non-zero elements (=the number of hash table nodes)
-    size_t nzcount() const;
-
-    //! computes the element hash value (1D case)
-    size_t hash(int i0) const;
-    //! computes the element hash value (2D case)
-    size_t hash(int i0, int i1) const;
-    //! computes the element hash value (3D case)
-    size_t hash(int i0, int i1, int i2) const;
-    //! computes the element hash value (nD case)
-    size_t hash(const int* idx) const;
-
-    //@{
-    /*!
-     specialized variants for 1D, 2D, 3D cases and the generic_type one for n-D case.
-
-     return pointer to the matrix element.
-     <ul>
-      <li>if the element is there (it's non-zero), the pointer to it is returned
-      <li>if it's not there and createMissing=false, NULL pointer is returned
-      <li>if it's not there and createMissing=true, then the new element
-        is created and initialized with 0. Pointer to it is returned
-      <li>if the optional hashval pointer is not NULL, the element hash value is
-      not computed, but *hashval is taken instead.
-     </ul>
-    */
-    //! returns pointer to the specified element (1D case)
-    uchar* ptr(int i0, bool createMissing, size_t* hashval=0);
-    //! returns pointer to the specified element (2D case)
-    uchar* ptr(int i0, int i1, bool createMissing, size_t* hashval=0);
-    //! returns pointer to the specified element (3D case)
-    uchar* ptr(int i0, int i1, int i2, bool createMissing, size_t* hashval=0);
-    //! returns pointer to the specified element (nD case)
-    uchar* ptr(const int* idx, bool createMissing, size_t* hashval=0);
-    //@}
-
-    //@{
-    /*!
-     return read-write reference to the specified sparse matrix element.
-
-     ref<_Tp>(i0,...[,hashval]) is equivalent to *(_Tp*)ptr(i0,...,true[,hashval]).
-     The methods always return a valid reference.
-     If the element did not exist, it is created and initialiazed with 0.
-    */
-    //! returns reference to the specified element (1D case)
-    template<typename _Tp> _Tp& ref(int i0, size_t* hashval=0);
-    //! returns reference to the specified element (2D case)
-    template<typename _Tp> _Tp& ref(int i0, int i1, size_t* hashval=0);
-    //! returns reference to the specified element (3D case)
-    template<typename _Tp> _Tp& ref(int i0, int i1, int i2, size_t* hashval=0);
-    //! returns reference to the specified element (nD case)
-    template<typename _Tp> _Tp& ref(const int* idx, size_t* hashval=0);
-    //@}
-
-    //@{
-    /*!
-     return value of the specified sparse matrix element.
-
-     value<_Tp>(i0,...[,hashval]) is equivalent
-
-     \code
-     { const _Tp* p = find<_Tp>(i0,...[,hashval]); return p ? *p : _Tp(); }
-     \endcode
-
-     That is, if the element did not exist, the methods return 0.
-     */
-    //! returns value of the specified element (1D case)
-    template<typename _Tp> _Tp value(int i0, size_t* hashval=0) const;
-    //! returns value of the specified element (2D case)
-    template<typename _Tp> _Tp value(int i0, int i1, size_t* hashval=0) const;
-    //! returns value of the specified element (3D case)
-    template<typename _Tp> _Tp value(int i0, int i1, int i2, size_t* hashval=0) const;
-    //! returns value of the specified element (nD case)
-    template<typename _Tp> _Tp value(const int* idx, size_t* hashval=0) const;
-    //@}
-
-    //@{
-    /*!
-     Return pointer to the specified sparse matrix element if it exists
-
-     find<_Tp>(i0,...[,hashval]) is equivalent to (_const Tp*)ptr(i0,...false[,hashval]).
-
-     If the specified element does not exist, the methods return NULL.
-    */
-    //! returns pointer to the specified element (1D case)
-    template<typename _Tp> const _Tp* find(int i0, size_t* hashval=0) const;
-    //! returns pointer to the specified element (2D case)
-    template<typename _Tp> const _Tp* find(int i0, int i1, size_t* hashval=0) const;
-    //! returns pointer to the specified element (3D case)
-    template<typename _Tp> const _Tp* find(int i0, int i1, int i2, size_t* hashval=0) const;
-    //! returns pointer to the specified element (nD case)
-    template<typename _Tp> const _Tp* find(const int* idx, size_t* hashval=0) const;
-
-    //! erases the specified element (2D case)
-    void erase(int i0, int i1, size_t* hashval=0);
-    //! erases the specified element (3D case)
-    void erase(int i0, int i1, int i2, size_t* hashval=0);
-    //! erases the specified element (nD case)
-    void erase(const int* idx, size_t* hashval=0);
-
-    //@{
-    /*!
-       return the sparse matrix iterator pointing to the first sparse matrix element
-    */
-    //! returns the sparse matrix iterator at the matrix beginning
-    SparseMatIterator begin();
-    //! returns the sparse matrix iterator at the matrix beginning
-    template<typename _Tp> SparseMatIterator_<_Tp> begin();
-    //! returns the read-only sparse matrix iterator at the matrix beginning
-    SparseMatConstIterator begin() const;
-    //! returns the read-only sparse matrix iterator at the matrix beginning
-    template<typename _Tp> SparseMatConstIterator_<_Tp> begin() const;
-    //@}
-    /*!
-       return the sparse matrix iterator pointing to the element following the last sparse matrix element
-    */
-    //! returns the sparse matrix iterator at the matrix end
-    SparseMatIterator end();
-    //! returns the read-only sparse matrix iterator at the matrix end
-    SparseMatConstIterator end() const;
-    //! returns the typed sparse matrix iterator at the matrix end
-    template<typename _Tp> SparseMatIterator_<_Tp> end();
-    //! returns the typed read-only sparse matrix iterator at the matrix end
-    template<typename _Tp> SparseMatConstIterator_<_Tp> end() const;
-
-    //! returns the value stored in the sparse martix node
-    template<typename _Tp> _Tp& value(Node* n);
-    //! returns the value stored in the sparse martix node
-    template<typename _Tp> const _Tp& value(const Node* n) const;
-
-    ////////////// some internal-use methods ///////////////
-    Node* node(size_t nidx);
-    const Node* node(size_t nidx) const;
-
-    uchar* newNode(const int* idx, size_t hashval);
-    void removeNode(size_t hidx, size_t nidx, size_t previdx);
-    void resizeHashTab(size_t newsize);
-
-    enum { MAGIC_VAL=0x42FD0000, MAX_DIM=CV_MAX_DIM, HASH_SCALE=0x5bd1e995, HASH_BIT=0x80000000 };
-
-    int flags;
-    Hdr* hdr;
-};
-
-//! finds global minimum and maximum sparse array elements and returns their values and their locations
-CV_EXPORTS void minMaxLoc(const SparseMat& a, double* minVal,
-                          double* maxVal, int* minIdx=0, int* maxIdx=0);
-//! computes norm of a sparse matrix
-CV_EXPORTS double norm( const SparseMat& src, int normType );
-//! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
-CV_EXPORTS void normalize( const SparseMat& src, SparseMat& dst, double alpha, int normType );
-
-/*!
- Read-Only Sparse Matrix Iterator.
- Here is how to use the iterator to compute the sum of floating-point sparse matrix elements:
-
- \code
- SparseMatConstIterator it = m.begin(), it_end = m.end();
- double s = 0;
- CV_Assert( m.type() == CV_32F );
- for( ; it != it_end; ++it )
-    s += it.value<float>();
- \endcode
-*/
-class CV_EXPORTS SparseMatConstIterator
-{
-public:
-    //! the default constructor
-    SparseMatConstIterator();
-    //! the full constructor setting the iterator to the first sparse matrix element
-    SparseMatConstIterator(const SparseMat* _m);
-    //! the copy constructor
-    SparseMatConstIterator(const SparseMatConstIterator& it);
-
-    //! the assignment operator
-    SparseMatConstIterator& operator = (const SparseMatConstIterator& it);
-
-    //! template method returning the current matrix element
-    template<typename _Tp> const _Tp& value() const;
-    //! returns the current node of the sparse matrix. it.node->idx is the current element index
-    const SparseMat::Node* node() const;
-
-    //! moves iterator to the previous element
-    SparseMatConstIterator& operator --();
-    //! moves iterator to the previous element
-    SparseMatConstIterator operator --(int);
-    //! moves iterator to the next element
-    SparseMatConstIterator& operator ++();
-    //! moves iterator to the next element
-    SparseMatConstIterator operator ++(int);
-
-    //! moves iterator to the element after the last element
-    void seekEnd();
-
-    const SparseMat* m;
-    size_t hashidx;
-    uchar* ptr;
-};
-
-/*!
- Read-write Sparse Matrix Iterator
-
- The class is similar to cv::SparseMatConstIterator,
- but can be used for in-place modification of the matrix elements.
-*/
-class CV_EXPORTS SparseMatIterator : public SparseMatConstIterator
-{
-public:
-    //! the default constructor
-    SparseMatIterator();
-    //! the full constructor setting the iterator to the first sparse matrix element
-    SparseMatIterator(SparseMat* _m);
-    //! the full constructor setting the iterator to the specified sparse matrix element
-    SparseMatIterator(SparseMat* _m, const int* idx);
-    //! the copy constructor
-    SparseMatIterator(const SparseMatIterator& it);
-
-    //! the assignment operator
-    SparseMatIterator& operator = (const SparseMatIterator& it);
-    //! returns read-write reference to the current sparse matrix element
-    template<typename _Tp> _Tp& value() const;
-    //! returns pointer to the current sparse matrix node. it.node->idx is the index of the current element (do not modify it!)
-    SparseMat::Node* node() const;
-
-    //! moves iterator to the next element
-    SparseMatIterator& operator ++();
-    //! moves iterator to the next element
-    SparseMatIterator operator ++(int);
-};
-
-/*!
- The Template Sparse Matrix class derived from cv::SparseMat
-
- The class provides slightly more convenient operations for accessing elements.
-
- \code
- SparseMat m;
- ...
- SparseMat_<int> m_ = (SparseMat_<int>&)m;
- m_.ref(1)++; // equivalent to m.ref<int>(1)++;
- m_.ref(2) += m_(3); // equivalent to m.ref<int>(2) += m.value<int>(3);
- \endcode
-*/
-template<typename _Tp> class CV_EXPORTS SparseMat_ : public SparseMat
-{
-public:
-    typedef SparseMatIterator_<_Tp> iterator;
-    typedef SparseMatConstIterator_<_Tp> const_iterator;
-
-    //! the default constructor
-    SparseMat_();
-    //! the full constructor equivelent to SparseMat(dims, _sizes, DataType<_Tp>::type)
-    SparseMat_(int dims, const int* _sizes);
-    //! the copy constructor. If DataType<_Tp>.type != m.type(), the m elements are converted
-    SparseMat_(const SparseMat& m);
-    //! the copy constructor. This is O(1) operation - no data is copied
-    SparseMat_(const SparseMat_& m);
-    //! converts dense matrix to the sparse form
-    SparseMat_(const Mat& m);
-    //! converts the old-style sparse matrix to the C++ class. All the elements are copied
-    SparseMat_(const CvSparseMat* m);
-    //! the assignment operator. If DataType<_Tp>.type != m.type(), the m elements are converted
-    SparseMat_& operator = (const SparseMat& m);
-    //! the assignment operator. This is O(1) operation - no data is copied
-    SparseMat_& operator = (const SparseMat_& m);
-    //! converts dense matrix to the sparse form
-    SparseMat_& operator = (const Mat& m);
-
-    //! makes full copy of the matrix. All the elements are duplicated
-    SparseMat_ clone() const;
-    //! equivalent to cv::SparseMat::create(dims, _sizes, DataType<_Tp>::type)
-    void create(int dims, const int* _sizes);
-    //! converts sparse matrix to the old-style CvSparseMat. All the elements are copied
-    operator CvSparseMat*() const;
-
-    //! returns type of the matrix elements
-    int type() const;
-    //! returns depth of the matrix elements
-    int depth() const;
-    //! returns the number of channels in each matrix element
-    int channels() const;
-
-    //! equivalent to SparseMat::ref<_Tp>(i0, hashval)
-    _Tp& ref(int i0, size_t* hashval=0);
-    //! equivalent to SparseMat::ref<_Tp>(i0, i1, hashval)
-    _Tp& ref(int i0, int i1, size_t* hashval=0);
-    //! equivalent to SparseMat::ref<_Tp>(i0, i1, i2, hashval)
-    _Tp& ref(int i0, int i1, int i2, size_t* hashval=0);
-    //! equivalent to SparseMat::ref<_Tp>(idx, hashval)
-    _Tp& ref(const int* idx, size_t* hashval=0);
-
-    //! equivalent to SparseMat::value<_Tp>(i0, hashval)
-    _Tp operator()(int i0, size_t* hashval=0) const;
-    //! equivalent to SparseMat::value<_Tp>(i0, i1, hashval)
-    _Tp operator()(int i0, int i1, size_t* hashval=0) const;
-    //! equivalent to SparseMat::value<_Tp>(i0, i1, i2, hashval)
-    _Tp operator()(int i0, int i1, int i2, size_t* hashval=0) const;
-    //! equivalent to SparseMat::value<_Tp>(idx, hashval)
-    _Tp operator()(const int* idx, size_t* hashval=0) const;
-
-    //! returns sparse matrix iterator pointing to the first sparse matrix element
-    SparseMatIterator_<_Tp> begin();
-    //! returns read-only sparse matrix iterator pointing to the first sparse matrix element
-    SparseMatConstIterator_<_Tp> begin() const;
-    //! returns sparse matrix iterator pointing to the element following the last sparse matrix element
-    SparseMatIterator_<_Tp> end();
-    //! returns read-only sparse matrix iterator pointing to the element following the last sparse matrix element
-    SparseMatConstIterator_<_Tp> end() const;
-};
-
-
-/*!
- Template Read-Only Sparse Matrix Iterator Class.
-
- This is the derived from SparseMatConstIterator class that
- introduces more convenient operator *() for accessing the current element.
-*/
-template<typename _Tp> class CV_EXPORTS SparseMatConstIterator_ : public SparseMatConstIterator
-{
-public:
-    typedef std::forward_iterator_tag iterator_category;
-
-    //! the default constructor
-    SparseMatConstIterator_();
-    //! the full constructor setting the iterator to the first sparse matrix element
-    SparseMatConstIterator_(const SparseMat_<_Tp>* _m);
-    //! the copy constructor
-    SparseMatConstIterator_(const SparseMatConstIterator_& it);
-
-    //! the assignment operator
-    SparseMatConstIterator_& operator = (const SparseMatConstIterator_& it);
-    //! the element access operator
-    const _Tp& operator *() const;
-
-    //! moves iterator to the next element
-    SparseMatConstIterator_& operator ++();
-    //! moves iterator to the next element
-    SparseMatConstIterator_ operator ++(int);
-};
-
-/*!
- Template Read-Write Sparse Matrix Iterator Class.
-
- This is the derived from cv::SparseMatConstIterator_ class that
- introduces more convenient operator *() for accessing the current element.
-*/
-template<typename _Tp> class CV_EXPORTS SparseMatIterator_ : public SparseMatConstIterator_<_Tp>
-{
-public:
-    typedef std::forward_iterator_tag iterator_category;
-
-    //! the default constructor
-    SparseMatIterator_();
-    //! the full constructor setting the iterator to the first sparse matrix element
-    SparseMatIterator_(SparseMat_<_Tp>* _m);
-    //! the copy constructor
-    SparseMatIterator_(const SparseMatIterator_& it);
-
-    //! the assignment operator
-    SparseMatIterator_& operator = (const SparseMatIterator_& it);
-    //! returns the reference to the current element
-    _Tp& operator *() const;
-
-    //! moves the iterator to the next element
-    SparseMatIterator_& operator ++();
-    //! moves the iterator to the next element
-    SparseMatIterator_ operator ++(int);
-};
-
-//////////////////// Fast Nearest-Neighbor Search Structure ////////////////////
 
 /*!
  Fast Nearest Neighbor Search Class.
@@ -3860,6 +809,7 @@ public:
         Node() : idx(-1), left(-1), right(-1), boundary(0.f) {}
         Node(int _idx, int _left, int _right, float _boundary)
             : idx(_idx), left(_left), right(_right), boundary(_boundary) {}
+
         //! split dimension; >=0 for nodes (dim), < 0 for leaves (index of the point)
         int idx;
         //! node indices of the left and the right branches
@@ -3871,32 +821,32 @@ public:
     //! the default constructor
     CV_WRAP KDTree();
     //! the full constructor that builds the search tree
-    CV_WRAP KDTree(InputArray points, bool copyAndReorderPoints=false);
+    CV_WRAP KDTree(InputArray points, bool copyAndReorderPoints = false);
     //! the full constructor that builds the search tree
     CV_WRAP KDTree(InputArray points, InputArray _labels,
-                   bool copyAndReorderPoints=false);
+                   bool copyAndReorderPoints = false);
     //! builds the search tree
-    CV_WRAP void build(InputArray points, bool copyAndReorderPoints=false);
+    CV_WRAP void build(InputArray points, bool copyAndReorderPoints = false);
     //! builds the search tree
     CV_WRAP void build(InputArray points, InputArray labels,
-                       bool copyAndReorderPoints=false);
+                       bool copyAndReorderPoints = false);
     //! finds the K nearest neighbors of "vec" while looking at Emax (at most) leaves
     CV_WRAP int findNearest(InputArray vec, int K, int Emax,
                             OutputArray neighborsIdx,
-                            OutputArray neighbors=noArray(),
-                            OutputArray dist=noArray(),
-                            OutputArray labels=noArray()) const;
+                            OutputArray neighbors = noArray(),
+                            OutputArray dist = noArray(),
+                            OutputArray labels = noArray()) const;
     //! finds all the points from the initial set that belong to the specified box
     CV_WRAP void findOrthoRange(InputArray minBounds,
                                 InputArray maxBounds,
                                 OutputArray neighborsIdx,
-                                OutputArray neighbors=noArray(),
-                                OutputArray labels=noArray()) const;
+                                OutputArray neighbors = noArray(),
+                                OutputArray labels = noArray()) const;
     //! returns vectors with the specified indices
     CV_WRAP void getPoints(InputArray idx, OutputArray pts,
-                           OutputArray labels=noArray()) const;
+                           OutputArray labels = noArray()) const;
     //! return a vector with the specified index
-    const float* getPoint(int ptidx, int* label=0) const;
+    const float* getPoint(int ptidx, int* label = 0) const;
     //! returns the search space dimensionality
     CV_WRAP int dims() const;
 
@@ -3907,447 +857,119 @@ public:
     CV_PROP_RW int normType; //!< type of the distance (cv::NORM_L1 or cv::NORM_L2) used for search. Initially set to cv::NORM_L2, but you can modify it
 };
 
-//////////////////////////////////////// XML & YAML I/O ////////////////////////////////////
 
-class CV_EXPORTS FileNode;
 
 /*!
- XML/YAML File Storage Class.
+   Random Number Generator
 
- The class describes an object associated with XML or YAML file.
- It can be used to store data to such a file or read and decode the data.
-
- The storage is organized as a tree of nested sequences (or lists) and mappings.
- Sequence is a heterogenious array, which elements are accessed by indices or sequentially using an iterator.
- Mapping is analogue of std::map or C structure, which elements are accessed by names.
- The most top level structure is a mapping.
- Leaves of the file storage tree are integers, floating-point numbers and text strings.
-
- For example, the following code:
-
- \code
- // open file storage for writing. Type of the file is determined from the extension
- FileStorage fs("test.yml", FileStorage::WRITE);
- fs << "test_int" << 5 << "test_real" << 3.1 << "test_string" << "ABCDEFGH";
- fs << "test_mat" << Mat::eye(3,3,CV_32F);
-
- fs << "test_list" << "[" << 0.0000000000001 << 2 << CV_PI << -3435345 << "2-502 2-029 3egegeg" <<
- "{:" << "month" << 12 << "day" << 31 << "year" << 1969 << "}" << "]";
- fs << "test_map" << "{" << "x" << 1 << "y" << 2 << "width" << 100 << "height" << 200 << "lbp" << "[:";
-
- const uchar arr[] = {0, 1, 1, 0, 1, 1, 0, 1};
- fs.writeRaw("u", arr, (int)(sizeof(arr)/sizeof(arr[0])));
-
- fs << "]" << "}";
- \endcode
-
- will produce the following file:
-
- \verbatim
- %YAML:1.0
- test_int: 5
- test_real: 3.1000000000000001e+00
- test_string: ABCDEFGH
- test_mat: !!opencv-matrix
-     rows: 3
-     cols: 3
-     dt: f
-     data: [ 1., 0., 0., 0., 1., 0., 0., 0., 1. ]
- test_list:
-     - 1.0000000000000000e-13
-     - 2
-     - 3.1415926535897931e+00
-     - -3435345
-     - "2-502 2-029 3egegeg"
-     - { month:12, day:31, year:1969 }
- test_map:
-     x: 1
-     y: 2
-     width: 100
-     height: 200
-     lbp: [ 0, 1, 1, 0, 1, 1, 0, 1 ]
- \endverbatim
-
- and to read the file above, the following code can be used:
-
- \code
- // open file storage for reading.
- // Type of the file is determined from the content, not the extension
- FileStorage fs("test.yml", FileStorage::READ);
- int test_int = (int)fs["test_int"];
- double test_real = (double)fs["test_real"];
- std::string test_string = (std::string)fs["test_string"];
-
- Mat M;
- fs["test_mat"] >> M;
-
- FileNode tl = fs["test_list"];
- CV_Assert(tl.type() == FileNode::SEQ && tl.size() == 6);
- double tl0 = (double)tl[0];
- int tl1 = (int)tl[1];
- double tl2 = (double)tl[2];
- int tl3 = (int)tl[3];
- std::string tl4 = (std::string)tl[4];
- CV_Assert(tl[5].type() == FileNode::MAP && tl[5].size() == 3);
-
- int month = (int)tl[5]["month"];
- int day = (int)tl[5]["day"];
- int year = (int)tl[5]["year"];
-
- FileNode tm = fs["test_map"];
-
- int x = (int)tm["x"];
- int y = (int)tm["y"];
- int width = (int)tm["width"];
- int height = (int)tm["height"];
-
- int lbp_val = 0;
- FileNodeIterator it = tm["lbp"].begin();
-
- for(int k = 0; k < 8; k++, ++it)
-    lbp_val |= ((int)*it) << k;
- \endcode
+   The class implements RNG using Multiply-with-Carry algorithm
 */
-class CV_EXPORTS_W FileStorage
+class CV_EXPORTS RNG
 {
 public:
-    //! file storage mode
-    enum
-    {
-        READ=0, //! read mode
-        WRITE=1, //! write mode
-        APPEND=2, //! append mode
-        MEMORY=4,
-        FORMAT_MASK=(7<<3),
-        FORMAT_AUTO=0,
-        FORMAT_XML=(1<<3),
-        FORMAT_YAML=(2<<3)
-    };
-    enum
-    {
-        UNDEFINED=0,
-        VALUE_EXPECTED=1,
-        NAME_EXPECTED=2,
-        INSIDE_MAP=4
-    };
-    //! the default constructor
-    CV_WRAP FileStorage();
-    //! the full constructor that opens file storage for reading or writing
-    CV_WRAP FileStorage(const std::string& source, int flags, const std::string& encoding=std::string());
-    //! the constructor that takes pointer to the C FileStorage structure
-    FileStorage(CvFileStorage* fs);
-    //! the destructor. calls release()
-    virtual ~FileStorage();
+    enum { UNIFORM = 0,
+           NORMAL  = 1
+         };
 
-    //! opens file storage for reading or writing. The previous storage is closed with release()
-    CV_WRAP virtual bool open(const std::string& filename, int flags, const std::string& encoding=std::string());
-    //! returns true if the object is associated with currently opened file.
-    CV_WRAP virtual bool isOpened() const;
-    //! closes the file and releases all the memory buffers
-    CV_WRAP virtual void release();
-    //! closes the file, releases all the memory buffers and returns the text string
-    CV_WRAP virtual std::string releaseAndGetString();
+    RNG();
+    RNG(uint64 state);
+    //! updates the state and returns the next 32-bit unsigned integer random number
+    unsigned next();
 
-    //! returns the first element of the top-level mapping
-    CV_WRAP FileNode getFirstTopLevelNode() const;
-    //! returns the top-level mapping. YAML supports multiple streams
-    CV_WRAP FileNode root(int streamidx=0) const;
-    //! returns the specified element of the top-level mapping
-    FileNode operator[](const std::string& nodename) const;
-    //! returns the specified element of the top-level mapping
-    CV_WRAP FileNode operator[](const char* nodename) const;
+    operator uchar();
+    operator schar();
+    operator ushort();
+    operator short();
+    operator unsigned();
+    //! returns a random integer sampled uniformly from [0, N).
+    unsigned operator ()(unsigned N);
+    unsigned operator ()();
+    operator int();
+    operator float();
+    operator double();
+    //! returns uniformly distributed integer random number from [a,b) range
+    int uniform(int a, int b);
+    //! returns uniformly distributed floating-point random number from [a,b) range
+    float uniform(float a, float b);
+    //! returns uniformly distributed double-precision floating-point random number from [a,b) range
+    double uniform(double a, double b);
+    void fill( InputOutputArray mat, int distType, InputArray a, InputArray b, bool saturateRange = false );
+    //! returns Gaussian random variate with mean zero.
+    double gaussian(double sigma);
 
-    //! returns pointer to the underlying C FileStorage structure
-    CvFileStorage* operator *() { return fs; }
-    //! returns pointer to the underlying C FileStorage structure
-    const CvFileStorage* operator *() const { return fs; }
-    //! writes one or more numbers of the specified format to the currently written structure
-    void writeRaw( const std::string& fmt, const uchar* vec, size_t len );
-    //! writes the registered C structure (CvMat, CvMatND, CvSeq). See cvWrite()
-    void writeObj( const std::string& name, const void* obj );
-
-    //! returns the normalized object name for the specified file name
-    static std::string getDefaultObjectName(const std::string& filename);
-
-    Ptr<CvFileStorage> fs; //!< the underlying C FileStorage structure
-    std::string elname; //!< the currently written element
-    std::vector<char> structs; //!< the stack of written structures
-    int state; //!< the writer state
+    uint64 state;
 };
 
-class CV_EXPORTS FileNodeIterator;
-
-/*!
- File Storage Node class
-
- The node is used to store each and every element of the file storage opened for reading -
- from the primitive objects, such as numbers and text strings, to the complex nodes:
- sequences, mappings and the registered objects.
-
- Note that file nodes are only used for navigating file storages opened for reading.
- When a file storage is opened for writing, no data is stored in memory after it is written.
-*/
-class CV_EXPORTS_W_SIMPLE FileNode
+class CV_EXPORTS RNG_MT19937
 {
 public:
-    //! type of the file storage node
-    enum
-    {
-        NONE=0, //!< empty node
-        INT=1, //!< an integer
-        REAL=2, //!< floating-point number
-        FLOAT=REAL, //!< synonym or REAL
-        STR=3, //!< text string in UTF-8 encoding
-        STRING=STR, //!< synonym for STR
-        REF=4, //!< integer of size size_t. Typically used for storing complex dynamic structures where some elements reference the others
-        SEQ=5, //!< sequence
-        MAP=6, //!< mapping
-        TYPE_MASK=7,
-        FLOW=8, //!< compact representation of a sequence or mapping. Used only by YAML writer
-        USER=16, //!< a registered object (e.g. a matrix)
-        EMPTY=32, //!< empty structure (sequence or mapping)
-        NAMED=64 //!< the node has a name (i.e. it is element of a mapping)
-    };
-    //! the default constructor
-    CV_WRAP FileNode();
-    //! the full constructor wrapping CvFileNode structure.
-    FileNode(const CvFileStorage* fs, const CvFileNode* node);
-    //! the copy constructor
-    FileNode(const FileNode& node);
-    //! returns element of a mapping node
-    FileNode operator[](const std::string& nodename) const;
-    //! returns element of a mapping node
-    CV_WRAP FileNode operator[](const char* nodename) const;
-    //! returns element of a sequence node
-    CV_WRAP FileNode operator[](int i) const;
-    //! returns type of the node
-    CV_WRAP int type() const;
+    RNG_MT19937();
+    RNG_MT19937(unsigned s);
+    void seed(unsigned s);
 
-    //! returns true if the node is empty
-    CV_WRAP bool empty() const;
-    //! returns true if the node is a "none" object
-    CV_WRAP bool isNone() const;
-    //! returns true if the node is a sequence
-    CV_WRAP bool isSeq() const;
-    //! returns true if the node is a mapping
-    CV_WRAP bool isMap() const;
-    //! returns true if the node is an integer
-    CV_WRAP bool isInt() const;
-    //! returns true if the node is a floating-point number
-    CV_WRAP bool isReal() const;
-    //! returns true if the node is a text string
-    CV_WRAP bool isString() const;
-    //! returns true if the node has a name
-    CV_WRAP bool isNamed() const;
-    //! returns the node name or an empty string if the node is nameless
-    CV_WRAP std::string name() const;
-    //! returns the number of elements in the node, if it is a sequence or mapping, or 1 otherwise.
-    CV_WRAP size_t size() const;
-    //! returns the node content as an integer. If the node stores floating-point number, it is rounded.
-    operator int() const;
-    //! returns the node content as float
-    operator float() const;
-    //! returns the node content as double
-    operator double() const;
-    //! returns the node content as text string
-    operator std::string() const;
+    unsigned next();
 
-    //! returns pointer to the underlying file node
-    CvFileNode* operator *();
-    //! returns pointer to the underlying file node
-    const CvFileNode* operator* () const;
+    operator int();
+    operator unsigned();
+    operator float();
+    operator double();
 
-    //! returns iterator pointing to the first node element
-    FileNodeIterator begin() const;
-    //! returns iterator pointing to the element following the last node element
-    FileNodeIterator end() const;
+    unsigned operator ()(unsigned N);
+    unsigned operator ()();
 
-    //! reads node elements to the buffer with the specified format
-    void readRaw( const std::string& fmt, uchar* vec, size_t len ) const;
-    //! reads the registered object and returns pointer to it
-    void* readObj() const;
+    // returns uniformly distributed integer random number from [a,b) range
+    int uniform(int a, int b);
+    // returns uniformly distributed floating-point random number from [a,b) range
+    float uniform(float a, float b);
+    // returns uniformly distributed double-precision floating-point random number from [a,b) range
+    double uniform(double a, double b);
 
-    // do not use wrapper pointer classes for better efficiency
-    const CvFileStorage* fs;
-    const CvFileNode* node;
+private:
+    enum PeriodParameters {N = 624, M = 397};
+    unsigned state[N];
+    int mti;
 };
 
 
-/*!
- File Node Iterator
 
- The class is used for iterating sequences (usually) and mappings.
- */
-class CV_EXPORTS FileNodeIterator
+/////////////////////////////// Formatted output of cv::Mat ///////////////////////////
+
+class CV_EXPORTS Formatted
 {
 public:
-    //! the default constructor
-    FileNodeIterator();
-    //! the full constructor set to the ofs-th element of the node
-    FileNodeIterator(const CvFileStorage* fs, const CvFileNode* node, size_t ofs=0);
-    //! the copy constructor
-    FileNodeIterator(const FileNodeIterator& it);
-    //! returns the currently observed element
-    FileNode operator *() const;
-    //! accesses the currently observed element methods
-    FileNode operator ->() const;
-
-    //! moves iterator to the next node
-    FileNodeIterator& operator ++ ();
-    //! moves iterator to the next node
-    FileNodeIterator operator ++ (int);
-    //! moves iterator to the previous node
-    FileNodeIterator& operator -- ();
-    //! moves iterator to the previous node
-    FileNodeIterator operator -- (int);
-    //! moves iterator forward by the specified offset (possibly negative)
-    FileNodeIterator& operator += (int ofs);
-    //! moves iterator backward by the specified offset (possibly negative)
-    FileNodeIterator& operator -= (int ofs);
-
-    //! reads the next maxCount elements (or less, if the sequence/mapping last element occurs earlier) to the buffer with the specified format
-    FileNodeIterator& readRaw( const std::string& fmt, uchar* vec,
-                               size_t maxCount=(size_t)INT_MAX );
-
-    const CvFileStorage* fs;
-    const CvFileNode* container;
-    CvSeqReader reader;
-    size_t remaining;
-};
-
-////////////// convenient wrappers for operating old-style dynamic structures //////////////
-
-template<typename _Tp> class SeqIterator;
-
-typedef Ptr<CvMemStorage> MemStorage;
-
-/*!
- Template Sequence Class derived from CvSeq
-
- The class provides more convenient access to sequence elements,
- STL-style operations and iterators.
-
- \note The class is targeted for simple data types,
-    i.e. no constructors or destructors
-    are called for the sequence elements.
-*/
-template<typename _Tp> class CV_EXPORTS Seq
-{
-public:
-    typedef SeqIterator<_Tp> iterator;
-    typedef SeqIterator<_Tp> const_iterator;
-
-    //! the default constructor
-    Seq();
-    //! the constructor for wrapping CvSeq structure. The real element type in CvSeq should match _Tp.
-    Seq(const CvSeq* seq);
-    //! creates the empty sequence that resides in the specified storage
-    Seq(MemStorage& storage, int headerSize = sizeof(CvSeq));
-    //! returns read-write reference to the specified element
-    _Tp& operator [](int idx);
-    //! returns read-only reference to the specified element
-    const _Tp& operator[](int idx) const;
-    //! returns iterator pointing to the beginning of the sequence
-    SeqIterator<_Tp> begin() const;
-    //! returns iterator pointing to the element following the last sequence element
-    SeqIterator<_Tp> end() const;
-    //! returns the number of elements in the sequence
-    size_t size() const;
-    //! returns the type of sequence elements (CV_8UC1 ... CV_64FC(CV_CN_MAX) ...)
-    int type() const;
-    //! returns the depth of sequence elements (CV_8U ... CV_64F)
-    int depth() const;
-    //! returns the number of channels in each sequence element
-    int channels() const;
-    //! returns the size of each sequence element
-    size_t elemSize() const;
-    //! returns index of the specified sequence element
-    size_t index(const _Tp& elem) const;
-    //! appends the specified element to the end of the sequence
-    void push_back(const _Tp& elem);
-    //! appends the specified element to the front of the sequence
-    void push_front(const _Tp& elem);
-    //! appends zero or more elements to the end of the sequence
-    void push_back(const _Tp* elems, size_t count);
-    //! appends zero or more elements to the front of the sequence
-    void push_front(const _Tp* elems, size_t count);
-    //! inserts the specified element to the specified position
-    void insert(int idx, const _Tp& elem);
-    //! inserts zero or more elements to the specified position
-    void insert(int idx, const _Tp* elems, size_t count);
-    //! removes element at the specified position
-    void remove(int idx);
-    //! removes the specified subsequence
-    void remove(const Range& r);
-
-    //! returns reference to the first sequence element
-    _Tp& front();
-    //! returns read-only reference to the first sequence element
-    const _Tp& front() const;
-    //! returns reference to the last sequence element
-    _Tp& back();
-    //! returns read-only reference to the last sequence element
-    const _Tp& back() const;
-    //! returns true iff the sequence contains no elements
-    bool empty() const;
-
-    //! removes all the elements from the sequence
-    void clear();
-    //! removes the first element from the sequence
-    void pop_front();
-    //! removes the last element from the sequence
-    void pop_back();
-    //! removes zero or more elements from the beginning of the sequence
-    void pop_front(_Tp* elems, size_t count);
-    //! removes zero or more elements from the end of the sequence
-    void pop_back(_Tp* elems, size_t count);
-
-    //! copies the whole sequence or the sequence slice to the specified vector
-    void copyTo(std::vector<_Tp>& vec, const Range& range=Range::all()) const;
-    //! returns the vector containing all the sequence elements
-    operator std::vector<_Tp>() const;
-
-    CvSeq* seq;
+    virtual const char* next() = 0;
+    virtual void reset() = 0;
+    virtual ~Formatted();
 };
 
 
-/*!
- STL-style Sequence Iterator inherited from the CvSeqReader structure
-*/
-template<typename _Tp> class CV_EXPORTS SeqIterator : public CvSeqReader
+class CV_EXPORTS Formatter
 {
 public:
-    //! the default constructor
-    SeqIterator();
-    //! the constructor setting the iterator to the beginning or to the end of the sequence
-    SeqIterator(const Seq<_Tp>& seq, bool seekEnd=false);
-    //! positions the iterator within the sequence
-    void seek(size_t pos);
-    //! reports the current iterator position
-    size_t tell() const;
-    //! returns reference to the current sequence element
-    _Tp& operator *();
-    //! returns read-only reference to the current sequence element
-    const _Tp& operator *() const;
-    //! moves iterator to the next sequence element
-    SeqIterator& operator ++();
-    //! moves iterator to the next sequence element
-    SeqIterator operator ++(int) const;
-    //! moves iterator to the previous sequence element
-    SeqIterator& operator --();
-    //! moves iterator to the previous sequence element
-    SeqIterator operator --(int) const;
+    enum { FMT_MATLAB  = 0,
+           FMT_CSV     = 1,
+           FMT_PYTHON  = 2,
+           FMT_NUMPY   = 3,
+           FMT_C       = 4,
+           FMT_DEFAULT = FMT_MATLAB
+         };
 
-    //! moves iterator forward by the specified offset (possibly negative)
-    SeqIterator& operator +=(int);
-    //! moves iterator backward by the specified offset (possibly negative)
-    SeqIterator& operator -=(int);
+    virtual ~Formatter();
+
+    virtual Ptr<Formatted> format(const Mat& mtx) const = 0;
+
+    virtual void set32fPrecision(int p = 8) = 0;
+    virtual void set64fPrecision(int p = 16) = 0;
+    virtual void setMultiline(bool ml = true) = 0;
+
+    static Ptr<Formatter> get(int fmt = FMT_DEFAULT);
 
-    // this is index of the current element module seq->total*2
-    // (to distinguish between 0 and seq->total)
-    int index;
 };
 
+
+
+//////////////////////////////////////// Algorithm ////////////////////////////////////
+
 class CV_EXPORTS Algorithm;
 class CV_EXPORTS AlgorithmInfo;
 struct CV_EXPORTS AlgorithmInfoData;
@@ -4362,41 +984,41 @@ class CV_EXPORTS_W Algorithm
 public:
     Algorithm();
     virtual ~Algorithm();
-    std::string name() const;
+    String name() const;
 
-    template<typename _Tp> typename ParamType<_Tp>::member_type get(const std::string& name) const;
+    template<typename _Tp> typename ParamType<_Tp>::member_type get(const String& name) const;
     template<typename _Tp> typename ParamType<_Tp>::member_type get(const char* name) const;
 
-    CV_WRAP int getInt(const std::string& name) const;
-    CV_WRAP double getDouble(const std::string& name) const;
-    CV_WRAP bool getBool(const std::string& name) const;
-    CV_WRAP std::string getString(const std::string& name) const;
-    CV_WRAP Mat getMat(const std::string& name) const;
-    CV_WRAP std::vector<Mat> getMatVector(const std::string& name) const;
-    CV_WRAP Ptr<Algorithm> getAlgorithm(const std::string& name) const;
+    CV_WRAP int getInt(const String& name) const;
+    CV_WRAP double getDouble(const String& name) const;
+    CV_WRAP bool getBool(const String& name) const;
+    CV_WRAP String getString(const String& name) const;
+    CV_WRAP Mat getMat(const String& name) const;
+    CV_WRAP std::vector<Mat> getMatVector(const String& name) const;
+    CV_WRAP Ptr<Algorithm> getAlgorithm(const String& name) const;
 
-    void set(const std::string& name, int value);
-    void set(const std::string& name, double value);
-    void set(const std::string& name, bool value);
-    void set(const std::string& name, const std::string& value);
-    void set(const std::string& name, const Mat& value);
-    void set(const std::string& name, const std::vector<Mat>& value);
-    void set(const std::string& name, const Ptr<Algorithm>& value);
-    template<typename _Tp> void set(const std::string& name, const Ptr<_Tp>& value);
+    void set(const String& name, int value);
+    void set(const String& name, double value);
+    void set(const String& name, bool value);
+    void set(const String& name, const String& value);
+    void set(const String& name, const Mat& value);
+    void set(const String& name, const std::vector<Mat>& value);
+    void set(const String& name, const Ptr<Algorithm>& value);
+    template<typename _Tp> void set(const String& name, const Ptr<_Tp>& value);
 
-    CV_WRAP void setInt(const std::string& name, int value);
-    CV_WRAP void setDouble(const std::string& name, double value);
-    CV_WRAP void setBool(const std::string& name, bool value);
-    CV_WRAP void setString(const std::string& name, const std::string& value);
-    CV_WRAP void setMat(const std::string& name, const Mat& value);
-    CV_WRAP void setMatVector(const std::string& name, const std::vector<Mat>& value);
-    CV_WRAP void setAlgorithm(const std::string& name, const Ptr<Algorithm>& value);
-    template<typename _Tp> void setAlgorithm(const std::string& name, const Ptr<_Tp>& value);
+    CV_WRAP void setInt(const String& name, int value);
+    CV_WRAP void setDouble(const String& name, double value);
+    CV_WRAP void setBool(const String& name, bool value);
+    CV_WRAP void setString(const String& name, const String& value);
+    CV_WRAP void setMat(const String& name, const Mat& value);
+    CV_WRAP void setMatVector(const String& name, const std::vector<Mat>& value);
+    CV_WRAP void setAlgorithm(const String& name, const Ptr<Algorithm>& value);
+    template<typename _Tp> void setAlgorithm(const String& name, const Ptr<_Tp>& value);
 
     void set(const char* name, int value);
     void set(const char* name, double value);
     void set(const char* name, bool value);
-    void set(const char* name, const std::string& value);
+    void set(const char* name, const String& value);
     void set(const char* name, const Mat& value);
     void set(const char* name, const std::vector<Mat>& value);
     void set(const char* name, const Ptr<Algorithm>& value);
@@ -4405,16 +1027,16 @@ public:
     void setInt(const char* name, int value);
     void setDouble(const char* name, double value);
     void setBool(const char* name, bool value);
-    void setString(const char* name, const std::string& value);
+    void setString(const char* name, const String& value);
     void setMat(const char* name, const Mat& value);
     void setMatVector(const char* name, const std::vector<Mat>& value);
     void setAlgorithm(const char* name, const Ptr<Algorithm>& value);
     template<typename _Tp> void setAlgorithm(const char* name, const Ptr<_Tp>& value);
 
-    CV_WRAP std::string paramHelp(const std::string& name) const;
+    CV_WRAP String paramHelp(const String& name) const;
     int paramType(const char* name) const;
-    CV_WRAP int paramType(const std::string& name) const;
-    CV_WRAP void getParams(CV_OUT std::vector<std::string>& names) const;
+    CV_WRAP int paramType(const String& name) const;
+    CV_WRAP void getParams(CV_OUT std::vector<String>& names) const;
 
 
     virtual void write(FileStorage& fs) const;
@@ -4424,9 +1046,9 @@ public:
     typedef int (Algorithm::*Getter)() const;
     typedef void (Algorithm::*Setter)(int);
 
-    CV_WRAP static void getList(CV_OUT std::vector<std::string>& algorithms);
-    CV_WRAP static Ptr<Algorithm> _create(const std::string& name);
-    template<typename _Tp> static Ptr<_Tp> create(const std::string& name);
+    CV_WRAP static void getList(CV_OUT std::vector<String>& algorithms);
+    CV_WRAP static Ptr<Algorithm> _create(const String& name);
+    template<typename _Tp> static Ptr<_Tp> create(const String& name);
 
     virtual AlgorithmInfo* info() const /* TODO: make it = 0;*/ { return 0; }
 };
@@ -4436,86 +1058,86 @@ class CV_EXPORTS AlgorithmInfo
 {
 public:
     friend class Algorithm;
-    AlgorithmInfo(const std::string& name, Algorithm::Constructor create);
+    AlgorithmInfo(const String& name, Algorithm::Constructor create);
     ~AlgorithmInfo();
     void get(const Algorithm* algo, const char* name, int argType, void* value) const;
     void addParam_(Algorithm& algo, const char* name, int argType,
                    void* value, bool readOnly,
                    Algorithm::Getter getter, Algorithm::Setter setter,
-                   const std::string& help=std::string());
-    std::string paramHelp(const char* name) const;
+                   const String& help=String());
+    String paramHelp(const char* name) const;
     int paramType(const char* name) const;
-    void getParams(std::vector<std::string>& names) const;
+    void getParams(std::vector<String>& names) const;
 
     void write(const Algorithm* algo, FileStorage& fs) const;
     void read(Algorithm* algo, const FileNode& fn) const;
-    std::string name() const;
+    String name() const;
 
     void addParam(Algorithm& algo, const char* name,
                   int& value, bool readOnly=false,
                   int (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(int)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   bool& value, bool readOnly=false,
                   int (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(int)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   double& value, bool readOnly=false,
                   double (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(double)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
-                  std::string& value, bool readOnly=false,
-                  std::string (Algorithm::*getter)()=0,
-                  void (Algorithm::*setter)(const std::string&)=0,
-                  const std::string& help=std::string());
+                  String& value, bool readOnly=false,
+                  String (Algorithm::*getter)()=0,
+                  void (Algorithm::*setter)(const String&)=0,
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   Mat& value, bool readOnly=false,
                   Mat (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(const Mat&)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   std::vector<Mat>& value, bool readOnly=false,
                   std::vector<Mat> (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(const std::vector<Mat>&)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   Ptr<Algorithm>& value, bool readOnly=false,
                   Ptr<Algorithm> (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(const Ptr<Algorithm>&)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   float& value, bool readOnly=false,
                   float (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(float)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   unsigned int& value, bool readOnly=false,
                   unsigned int (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(unsigned int)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   uint64& value, bool readOnly=false,
                   uint64 (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(uint64)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     void addParam(Algorithm& algo, const char* name,
                   uchar& value, bool readOnly=false,
                   uchar (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(uchar)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     template<typename _Tp, typename _Base> void addParam(Algorithm& algo, const char* name,
                   Ptr<_Tp>& value, bool readOnly=false,
                   Ptr<_Tp> (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(const Ptr<_Tp>&)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
     template<typename _Tp> void addParam(Algorithm& algo, const char* name,
                   Ptr<_Tp>& value, bool readOnly=false,
                   Ptr<_Tp> (Algorithm::*getter)()=0,
                   void (Algorithm::*setter)(const Ptr<_Tp>&)=0,
-                  const std::string& help=std::string());
+                  const String& help=String());
 protected:
     AlgorithmInfoData* data;
     void set(Algorithm* algo, const char* name, int argType,
@@ -4531,13 +1153,13 @@ struct CV_EXPORTS Param
     Param(int _type, bool _readonly, int _offset,
           Algorithm::Getter _getter=0,
           Algorithm::Setter _setter=0,
-          const std::string& _help=std::string());
+          const String& _help=String());
     int type;
     int offset;
     bool readonly;
     Algorithm::Getter getter;
     Algorithm::Setter setter;
-    std::string help;
+    String help;
 };
 
 template<> struct ParamType<bool>
@@ -4564,10 +1186,10 @@ template<> struct ParamType<double>
     enum { type = Param::REAL };
 };
 
-template<> struct ParamType<std::string>
+template<> struct ParamType<String>
 {
-    typedef const std::string& const_param_type;
-    typedef std::string member_type;
+    typedef const String& const_param_type;
+    typedef String member_type;
 
     enum { type = Param::STRING };
 };
@@ -4628,95 +1250,10 @@ template<> struct ParamType<uchar>
     enum { type = Param::UCHAR };
 };
 
-// The CommandLineParser class is designed for command line arguments parsing
-
-class CV_EXPORTS CommandLineParser
-{
-    public:
-    CommandLineParser(int argc, const char* const argv[], const std::string& keys);
-    CommandLineParser(const CommandLineParser& parser);
-    CommandLineParser& operator = (const CommandLineParser& parser);
-
-    std::string getPathToApplication() const;
-
-    template <typename T>
-    T get(const std::string& name, bool space_delete = true) const
-        {
-        T val = T();
-        getByName(name, space_delete, ParamType<T>::type, (void*)&val);
-        return val;
-        }
-
-    template <typename T>
-    T get(int index, bool space_delete = true) const
-    {
-        T val = T();
-        getByIndex(index, space_delete, ParamType<T>::type, (void*)&val);
-        return val;
-    }
-
-    bool has(const std::string& name) const;
-
-    bool check() const;
-
-    void about(const std::string& message);
-
-    void printMessage() const;
-    void printErrors() const;
-
-protected:
-    void getByName(const std::string& name, bool space_delete, int type, void* dst) const;
-    void getByIndex(int index, bool space_delete, int type, void* dst) const;
-
-    struct Impl;
-    Impl* impl;
-};
-
-/////////////////////////////// Parallel Primitives //////////////////////////////////
-
-// a base body class
-class CV_EXPORTS ParallelLoopBody
-{
-public:
-    virtual ~ParallelLoopBody();
-    virtual void operator() (const Range& range) const = 0;
-};
-
-CV_EXPORTS void parallel_for_(const Range& range, const ParallelLoopBody& body, double nstripes=-1.);
-
-/////////////////////////// Synchronization Primitives ///////////////////////////////
-
-class CV_EXPORTS Mutex
-{
-public:
-    Mutex();
-    ~Mutex();
-    Mutex(const Mutex& m);
-    Mutex& operator = (const Mutex& m);
-
-    void lock();
-    bool trylock();
-    void unlock();
-
-    struct Impl;
-protected:
-    Impl* impl;
-};
-
-class CV_EXPORTS AutoLock
-{
-public:
-    AutoLock(Mutex& m) : mutex(&m) { mutex->lock(); }
-    ~AutoLock() { mutex->unlock(); }
-protected:
-    Mutex* mutex;
-};
-
-}
-
-#endif // __cplusplus
+} //namespace cv
 
 #include "opencv2/core/operations.hpp"
-#include "opencv2/core/mat.hpp"
+#include "opencv2/core/cvstd.inl.hpp"
+
 
 #endif /*__OPENCV_CORE_HPP__*/
diff --git a/modules/core/include/opencv2/core/affine.hpp b/modules/core/include/opencv2/core/affine.hpp
index c9329c9b64..ed8bd29a18 100644
--- a/modules/core/include/opencv2/core/affine.hpp
+++ b/modules/core/include/opencv2/core/affine.hpp
@@ -1,44 +1,45 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
- //
- //  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
- //
- //  By downloading, copying, installing or using the software you agree to this license.
- //  If you do not agree to this license, do not download, install,
- //  copy or use the software.
- //
- //
- //                           License Agreement
- //                For Open Source Computer Vision Library
- //
- // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
- // Copyright (C) 2008-2013, Willow Garage Inc., all rights reserved.
- // Third party copyrights are property of their respective owners.
- //
- // Redistribution and use in source and binary forms, with or without modification,
- // are permitted provided that the following conditions are met:
- //
- //   * Redistribution's of source code must retain the above copyright notice,
- //     this list of conditions and the following disclaimer.
- //
- //   * Redistribution's in binary form must reproduce the above copyright notice,
- //     this list of conditions and the following disclaimer in the documentation
- //     and / or other materials provided with the distribution.
- //
- //   * The name of the copyright holders may not be used to endorse or promote products
- //     derived from this software without specific prior written permission.
- //
- // This software is provided by the copyright holders and contributors "as is" and
- // any express or implied warranties, including, but not limited to, the implied
- // warranties of merchantability and fitness for a particular purpose are disclaimed.
- // In no event shall the Intel Corporation or contributors be liable for any direct,
- // indirect, incidental, special, exemplary, or consequential damages
- // (including, but not limited to, procurement of substitute goods or services;
- // loss of use, data, or profits; or business interruption) however caused
- // and on any theory of liability, whether in contract, strict liability,
- // or tort (including negligence or otherwise) arising in any way out of
- // the use of this software, even if advised of the possibility of such damage.
- //
- //M*/
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
 #ifndef __OPENCV_CORE_AFFINE3_HPP__
 #define __OPENCV_CORE_AFFINE3_HPP__
@@ -68,7 +69,7 @@ namespace cv
 
         //Rodrigues vector
         Affine3(const Vec3& rvec, const Vec3& t = Vec3::all(0));
-        
+
         //Combines all contructors above. Supports 4x4, 3x3, 1x3, 3x1 sizes of data matrix
         explicit Affine3(const cv::Mat& data, const Vec3& t = Vec3::all(0));
 
@@ -79,13 +80,13 @@ namespace cv
 
         //Rotation matrix
         void rotation(const Mat3& R);
-        
+
         //Rodrigues vector
         void rotation(const Vec3& rvec);
-        
+
         //Combines rotation methods above. Suports 3x3, 1x3, 3x1 sizes of data matrix;
         void rotation(const Mat& data);
-        
+
         //Euler angles
         void rotation(float_type alpha, float_type beta, float_type gamma);
 
@@ -109,8 +110,6 @@ namespace cv
 
         template <typename Y> operator Affine3<Y>() const;
 
-        operator cv::Mat();
-
         Mat4 matrix;
 
 #if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
@@ -121,24 +120,36 @@ namespace cv
 #endif
     };
 
-    template<typename T> Affine3<T> operator*(const Affine3<T>& affine1, const Affine3<T>& affine2);
-    template<typename T, typename V> V operator*(const Affine3<T>& affine, const V& vector);
+    template<typename T> static
+    Affine3<T> operator*(const Affine3<T>& affine1, const Affine3<T>& affine2);
+
+    template<typename T, typename V> static
+    V operator*(const Affine3<T>& affine, const V& vector);
 
     typedef Affine3<float> Affine3f;
     typedef Affine3<double> Affine3d;
 
-    cv::Vec3f operator*(const cv::Affine3f& affine, const cv::Vec3f& vector);
-    cv::Vec3d operator*(const cv::Affine3d& affine, const cv::Vec3d& vector);
+    static cv::Vec3f operator*(const cv::Affine3f& affine, const cv::Vec3f& vector);
+    static cv::Vec3d operator*(const cv::Affine3d& affine, const cv::Vec3d& vector);
 }
 
 
+
 ///////////////////////////////////////////////////////////////////////////////////
 /// Implementaiton
 
-template<typename T> inline cv::Affine3<T>::Affine3() : matrix(Mat4::eye()) {}
-template<typename T> inline cv::Affine3<T>::Affine3(const Mat4& affine) : matrix(affine) {}
+template<typename T> inline
+cv::Affine3<T>::Affine3()
+    : matrix(Mat4::eye())
+{}
 
-template<typename T> inline cv::Affine3<T>::Affine3(const Mat3& R, const Vec3& t)
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat4& affine)
+    : matrix(affine)
+{}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat3& R, const Vec3& t)
 {
     rotation(R);
     translation(t);
@@ -146,7 +157,8 @@ template<typename T> inline cv::Affine3<T>::Affine3(const Mat3& R, const Vec3& t
     matrix.val[15] = 1;
 }
 
-template<typename T> inline cv::Affine3<T>::Affine3(const Vec3& rvec, const Vec3& t)
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Vec3& rvec, const Vec3& t)
 {
     rotation(rvec);
     translation(t);
@@ -154,7 +166,8 @@ template<typename T> inline cv::Affine3<T>::Affine3(const Vec3& rvec, const Vec3
     matrix.val[15] = 1;
 }
 
-template<typename T> inline cv::Affine3<T>::Affine3(const cv::Mat& data, const Vec3& t)
+template<typename T> inline
+cv::Affine3<T>::Affine3(const cv::Mat& data, const Vec3& t)
 {
     CV_Assert(data.type() == cv::DataType<T>::type);
 
@@ -170,7 +183,8 @@ template<typename T> inline cv::Affine3<T>::Affine3(const cv::Mat& data, const V
     matrix.val[15] = 1;
 }
 
-template<typename T> inline cv::Affine3<T>::Affine3(float_type alpha, float_type beta, float_type gamma, const Vec3& t)
+template<typename T> inline
+cv::Affine3<T>::Affine3(float_type alpha, float_type beta, float_type gamma, const Vec3& t)
 {
     rotation(alpha, beta, gamma);
     translation(t);
@@ -178,14 +192,20 @@ template<typename T> inline cv::Affine3<T>::Affine3(float_type alpha, float_type
     matrix.val[15] = 1;
 }
 
-template<typename T> inline cv::Affine3<T> cv::Affine3<T>::Identity()
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::Identity()
 {
     return Affine3<T>(cv::Affine3<T>::Mat4::eye());
 }
 
-template<typename T> inline void cv::Affine3<T>::rotation(const Mat3& R) { linear(R); }
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Mat3& R)
+{
+    linear(R);
+}
 
-template<typename T> inline void cv::Affine3<T>::rotation(const Vec3& rvec)
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Vec3& rvec)
 {
     double rx = rvec[0], ry = rvec[1], rz = rvec[2];
     double theta = std::sqrt(rx*rx + ry*ry + rz*rz);
@@ -217,10 +237,11 @@ template<typename T> inline void cv::Affine3<T>::rotation(const Vec3& rvec)
 }
 
 //Combines rotation methods above. Suports 3x3, 1x3, 3x1 sizes of data matrix;
-template<typename T> inline void cv::Affine3<T>::rotation(const cv::Mat& data)
+template<typename T> inline
+void cv::Affine3<T>::rotation(const cv::Mat& data)
 {
     CV_Assert(data.type() == cv::DataType<T>::type);
-    
+
     if (data.cols == 3 && data.rows == 3)
     {
         Mat3 R;
@@ -237,44 +258,56 @@ template<typename T> inline void cv::Affine3<T>::rotation(const cv::Mat& data)
         CV_Assert(!"Input marix can be 3x3, 1x3 or 3x1");
 }
 
-template<typename T> inline void cv::Affine3<T>::rotation(float_type alpha, float_type beta, float_type gamma)
+template<typename T> inline
+void cv::Affine3<T>::rotation(float_type alpha, float_type beta, float_type gamma)
 {
     rotation(Vec3(alpha, beta, gamma));
 }
 
-template<typename T> inline void cv::Affine3<T>::linear(const Mat3& L)
+template<typename T> inline
+void cv::Affine3<T>::linear(const Mat3& L)
 {
     matrix.val[0] = L.val[0]; matrix.val[1] = L.val[1];  matrix.val[ 2] = L.val[2];
     matrix.val[4] = L.val[3]; matrix.val[5] = L.val[4];  matrix.val[ 6] = L.val[5];
     matrix.val[8] = L.val[6]; matrix.val[9] = L.val[7];  matrix.val[10] = L.val[8];
 }
 
-template<typename T> inline void cv::Affine3<T>::translation(const Vec3& t)
+template<typename T> inline
+void cv::Affine3<T>::translation(const Vec3& t)
 {
     matrix.val[3] = t[0]; matrix.val[7] = t[1]; matrix.val[11] = t[2];
 }
 
-template<typename T> inline typename cv::Affine3<T>::Mat3 cv::Affine3<T>::rotation() const { return linear(); }
-template<typename T> inline typename cv::Affine3<T>::Mat3 cv::Affine3<T>::linear() const
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::rotation() const
 {
-    cv::Affine3<T>::Mat3 R;
+    return linear();
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::linear() const
+{
+    typename cv::Affine3<T>::Mat3 R;
     R.val[0] = matrix.val[0];  R.val[1] = matrix.val[1];  R.val[2] = matrix.val[ 2];
     R.val[3] = matrix.val[4];  R.val[4] = matrix.val[5];  R.val[5] = matrix.val[ 6];
     R.val[6] = matrix.val[8];  R.val[7] = matrix.val[9];  R.val[8] = matrix.val[10];
     return R;
 }
 
-template<typename T> inline typename cv::Affine3<T>::Vec3 cv::Affine3<T>::translation() const
+template<typename T> inline
+typename cv::Affine3<T>::Vec3 cv::Affine3<T>::translation() const
 {
     return Vec3(matrix.val[3], matrix.val[7], matrix.val[11]);
 }
 
-template<typename T> inline cv::Affine3<T> cv::Affine3<T>::inv(int method) const
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::inv(int method) const
 {
     return matrix.inv(method);
 }
 
-template<typename T> inline cv::Affine3<T> cv::Affine3<T>::rotate(const Mat3& R) const
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::rotate(const Mat3& R) const
 {
     Mat3 Lc = linear();
     Vec3 tc = translation();
@@ -297,7 +330,8 @@ template<typename T> inline cv::Affine3<T> cv::Affine3<T>::rotate(const Mat3& R)
     return result;
 }
 
-template<typename T> inline cv::Affine3<T> cv::Affine3<T>::translate(const Vec3& t) const
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::translate(const Vec3& t) const
 {
     Mat4 m = matrix;
     m.val[ 3] += t[0];
@@ -306,24 +340,26 @@ template<typename T> inline cv::Affine3<T> cv::Affine3<T>::translate(const Vec3&
     return m;
 }
 
-template<typename T> inline cv::Affine3<T> cv::Affine3<T>::concatenate(const Affine3<T>& affine) const
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::concatenate(const Affine3<T>& affine) const
 {
     return (*this).rotate(affine.rotation()).translate(affine.translation());
 }
 
-template<typename T> template <typename Y> inline cv::Affine3<T>::operator Affine3<Y>() const
+template<typename T> template <typename Y> inline
+cv::Affine3<T>::operator Affine3<Y>() const
 {
     return Affine3<Y>(matrix);
 }
 
-template<typename T> inline cv::Affine3<T>::operator cv::Mat() { return cv::Mat(matrix, false); }
-
-template<typename T> inline cv::Affine3<T> cv::operator*(const cv::Affine3<T>& affine1, const cv::Affine3<T>& affine2)
+template<typename T> inline
+cv::Affine3<T> cv::operator*(const cv::Affine3<T>& affine1, const cv::Affine3<T>& affine2)
 {
     return affine2.concatenate(affine1);
 }
 
-template<typename T, typename V> inline V cv::operator*(const cv::Affine3<T>& affine, const V& v)
+template<typename T, typename V> inline
+V cv::operator*(const cv::Affine3<T>& affine, const V& v)
 {
     const typename Affine3<T>::Mat4& m = affine.matrix;
 
@@ -334,7 +370,8 @@ template<typename T, typename V> inline V cv::operator*(const cv::Affine3<T>& af
     return r;
 }
 
-inline cv::Vec3f cv::operator*(const cv::Affine3f& affine, const cv::Vec3f& v)
+static inline
+cv::Vec3f cv::operator*(const cv::Affine3f& affine, const cv::Vec3f& v)
 {
     const cv::Matx44f& m = affine.matrix;
     cv::Vec3f r;
@@ -344,7 +381,8 @@ inline cv::Vec3f cv::operator*(const cv::Affine3f& affine, const cv::Vec3f& v)
     return r;
 }
 
-inline cv::Vec3d cv::operator*(const cv::Affine3d& affine, const cv::Vec3d& v)
+static inline
+cv::Vec3d cv::operator*(const cv::Affine3d& affine, const cv::Vec3d& v)
 {
     const cv::Matx44d& m = affine.matrix;
     cv::Vec3d r;
@@ -354,20 +392,25 @@ inline cv::Vec3d cv::operator*(const cv::Affine3d& affine, const cv::Vec3d& v)
     return r;
 }
 
-#if (defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H) || defined CV_AFFINE_FORCE_EIGEN_PLUGIN
 
-template<typename T> inline cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine)
+
+#if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine)
 {
     cv::Mat(4, 4, cv::DataType<T>::type, affine.matrix().data()).copyTo(matrix);
 }
 
-template<typename T> inline cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine)
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine)
 {
     Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> a = affine;
     cv::Mat(4, 4, cv::DataType<T>::type, a.matrix().data()).copyTo(matrix);
 }
 
-template<typename T> inline cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const
 {
     Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> r;
     cv::Mat hdr(4, 4, cv::DataType<T>::type, r.matrix().data());
@@ -375,7 +418,8 @@ template<typename T> inline cv::Affine3<T>::operator Eigen::Transform<T, 3, Eige
     return r;
 }
 
-template<typename T> inline cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine>() const
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine>() const
 {
     return this->operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>();
 }
@@ -386,5 +430,3 @@ template<typename T> inline cv::Affine3<T>::operator Eigen::Transform<T, 3, Eige
 #endif /* __cplusplus */
 
 #endif /* __OPENCV_CORE_AFFINE3_HPP__ */
-
-
diff --git a/modules/core/include/opencv2/core/base.hpp b/modules/core/include/opencv2/core/base.hpp
new file mode 100644
index 0000000000..013f573a64
--- /dev/null
+++ b/modules/core/include/opencv2/core/base.hpp
@@ -0,0 +1,508 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_BASE_HPP__
+#define __OPENCV_CORE_BASE_HPP__
+
+#ifndef __cplusplus
+#  error base.hpp header must be compiled as C++
+#endif
+
+#include <climits>
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+
+namespace cv
+{
+
+// error codes
+namespace Error {
+enum {
+    StsOk=                       0,  /* everithing is ok                */
+    StsBackTrace=               -1,  /* pseudo error for back trace     */
+    StsError=                   -2,  /* unknown /unspecified error      */
+    StsInternal=                -3,  /* internal error (bad state)      */
+    StsNoMem=                   -4,  /* insufficient memory             */
+    StsBadArg=                  -5,  /* function arg/param is bad       */
+    StsBadFunc=                 -6,  /* unsupported function            */
+    StsNoConv=                  -7,  /* iter. didn't converge           */
+    StsAutoTrace=               -8,  /* tracing                         */
+    HeaderIsNull=               -9,  /* image header is NULL            */
+    BadImageSize=              -10,  /* image size is invalid           */
+    BadOffset=                 -11,  /* offset is invalid               */
+    BadDataPtr=                -12,  /**/
+    BadStep=                   -13,  /**/
+    BadModelOrChSeq=           -14,  /**/
+    BadNumChannels=            -15,  /**/
+    BadNumChannel1U=           -16,  /**/
+    BadDepth=                  -17,  /**/
+    BadAlphaChannel=           -18,  /**/
+    BadOrder=                  -19,  /**/
+    BadOrigin=                 -20,  /**/
+    BadAlign=                  -21,  /**/
+    BadCallBack=               -22,  /**/
+    BadTileSize=               -23,  /**/
+    BadCOI=                    -24,  /**/
+    BadROISize=                -25,  /**/
+    MaskIsTiled=               -26,  /**/
+    StsNullPtr=                -27,  /* null pointer */
+    StsVecLengthErr=           -28,  /* incorrect vector length */
+    StsFilterStructContentErr= -29,  /* incorr. filter structure content */
+    StsKernelStructContentErr= -30,  /* incorr. transform kernel content */
+    StsFilterOffsetErr=        -31,  /* incorrect filter ofset value */
+    StsBadSize=                -201, /* the input/output structure size is incorrect  */
+    StsDivByZero=              -202, /* division by zero */
+    StsInplaceNotSupported=    -203, /* in-place operation is not supported */
+    StsObjectNotFound=         -204, /* request can't be completed */
+    StsUnmatchedFormats=       -205, /* formats of input/output arrays differ */
+    StsBadFlag=                -206, /* flag is wrong or not supported */
+    StsBadPoint=               -207, /* bad CvPoint */
+    StsBadMask=                -208, /* bad format of mask (neither 8uC1 nor 8sC1)*/
+    StsUnmatchedSizes=         -209, /* sizes of input/output structures do not match */
+    StsUnsupportedFormat=      -210, /* the data format/type is not supported by the function*/
+    StsOutOfRange=             -211, /* some of parameters are out of range */
+    StsParseError=             -212, /* invalid syntax/structure of the parsed file */
+    StsNotImplemented=         -213, /* the requested function/feature is not implemented */
+    StsBadMemBlock=            -214, /* an allocated block has been corrupted */
+    StsAssert=                 -215, /* assertion failed */
+    GpuNotSupported=           -216,
+    GpuApiCallError=           -217,
+    OpenGlNotSupported=        -218,
+    OpenGlApiCallError=        -219,
+    OpenCLApiCallError=        -220
+};
+} //Error
+
+// matrix decomposition types
+enum { DECOMP_LU       = 0,
+       DECOMP_SVD      = 1,
+       DECOMP_EIG      = 2,
+       DECOMP_CHOLESKY = 3,
+       DECOMP_QR       = 4,
+       DECOMP_NORMAL   = 16
+     };
+
+// norm types
+enum { NORM_INF       = 1,
+       NORM_L1        = 2,
+       NORM_L2        = 4,
+       NORM_L2SQR     = 5,
+       NORM_HAMMING   = 6,
+       NORM_HAMMING2  = 7,
+       NORM_TYPE_MASK = 7,
+       NORM_RELATIVE  = 8,
+       NORM_MINMAX    = 32
+     };
+
+// comparison types
+enum { CMP_EQ = 0,
+       CMP_GT = 1,
+       CMP_GE = 2,
+       CMP_LT = 3,
+       CMP_LE = 4,
+       CMP_NE = 5
+     };
+
+enum { GEMM_1_T = 1,
+       GEMM_2_T = 2,
+       GEMM_3_T = 4
+     };
+
+enum { DFT_INVERSE        = 1,
+       DFT_SCALE          = 2,
+       DFT_ROWS           = 4,
+       DFT_COMPLEX_OUTPUT = 16,
+       DFT_REAL_OUTPUT    = 32,
+       DCT_INVERSE        = DFT_INVERSE,
+       DCT_ROWS           = DFT_ROWS
+     };
+
+//! Various border types, image boundaries are denoted with '|'
+enum {
+       BORDER_CONSTANT    = 0, // iiiiii|abcdefgh|iiiiiii  with some specified 'i'
+       BORDER_REPLICATE   = 1, // aaaaaa|abcdefgh|hhhhhhh
+       BORDER_REFLECT     = 2, // fedcba|abcdefgh|hgfedcb
+       BORDER_WRAP        = 3, // cdefgh|abcdefgh|abcdefg
+       BORDER_REFLECT_101 = 4, // gfedcb|abcdefgh|gfedcba
+       BORDER_TRANSPARENT = 5, // uvwxyz|absdefgh|ijklmno
+
+       BORDER_REFLECT101  = BORDER_REFLECT_101,
+       BORDER_DEFAULT     = BORDER_REFLECT_101,
+       BORDER_ISOLATED    = 16 // do not look outside of ROI
+     };
+
+
+
+//////////////// static assert /////////////////
+
+#define CVAUX_CONCAT_EXP(a, b) a##b
+#define CVAUX_CONCAT(a, b) CVAUX_CONCAT_EXP(a,b)
+
+#if defined(__clang__)
+#  ifndef __has_extension
+#    define __has_extension __has_feature /* compatibility, for older versions of clang */
+#  endif
+#  if __has_extension(cxx_static_assert)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(__GNUC__)
+#  if (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(_MSC_VER)
+#  if _MSC_VER >= 1600 /* MSVC 10 */
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#endif
+#ifndef CV_StaticAssert
+#  if defined(__GNUC__) && (__GNUC__ > 3) && (__GNUC_MINOR__ > 2)
+#    define CV_StaticAssert(condition, reason) ({ extern int __attribute__((error("CV_StaticAssert: " reason " " #condition))) CV_StaticAssert(); ((condition) ? 0 : CV_StaticAssert()); })
+#  else
+     template <bool x> struct CV_StaticAssert_failed;
+     template <> struct CV_StaticAssert_failed<true> { enum { val = 1 }; };
+     template<int x> struct CV_StaticAssert_test {};
+#    define CV_StaticAssert(condition, reason)\
+       typedef cv::CV_StaticAssert_test< sizeof(cv::CV_StaticAssert_failed< static_cast<bool>(condition) >) > CVAUX_CONCAT(CV_StaticAssert_failed_at_, __LINE__)
+#  endif
+#endif
+
+
+
+//! Signals an error and raises the exception.
+/*!
+  By default the function prints information about the error to stderr,
+  then it either stops if setBreakOnError() had been called before or raises the exception.
+  It is possible to alternate error processing by using redirectError().
+
+  \param exc the exception raisen.
+ */
+CV_EXPORTS void error(int _code, const String& _err, const char* _func, const char* _file, int _line);
+
+#ifdef __GNUC__
+#  define CV_Error( code, msg ) cv::error( code, msg, __func__, __FILE__, __LINE__ )
+#  define CV_Error_( code, args ) cv::error( code, cv::format args, __func__, __FILE__, __LINE__ )
+#  define CV_Assert( expr ) if(!!(expr)) ; else cv::error( cv::Error::StsAssert, #expr, __func__, __FILE__, __LINE__ )
+#else
+#  define CV_Error( code, msg ) cv::error( code, msg, "", __FILE__, __LINE__ )
+#  define CV_Error_( code, args ) cv::error( code, cv::format args, "", __FILE__, __LINE__ )
+#  define CV_Assert( expr ) if(!!(expr)) ; else cv::error( cv::Error::StsAssert, #expr, "", __FILE__, __LINE__ )
+#endif
+
+#ifdef _DEBUG
+#  define CV_DbgAssert(expr) CV_Assert(expr)
+#else
+#  define CV_DbgAssert(expr)
+#endif
+
+
+
+/////////////// saturate_cast (used in image & signal processing) ///////////////////
+
+template<typename _Tp> static inline _Tp saturate_cast(uchar v)    { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(schar v)    { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(ushort v)   { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(short v)    { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(unsigned v) { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(int v)      { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(float v)    { return _Tp(v); }
+template<typename _Tp> static inline _Tp saturate_cast(double v)   { return _Tp(v); }
+
+template<> inline uchar saturate_cast<uchar>(schar v)        { return (uchar)std::max((int)v, 0); }
+template<> inline uchar saturate_cast<uchar>(ushort v)       { return (uchar)std::min((unsigned)v, (unsigned)UCHAR_MAX); }
+template<> inline uchar saturate_cast<uchar>(int v)          { return (uchar)((unsigned)v <= UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
+template<> inline uchar saturate_cast<uchar>(short v)        { return saturate_cast<uchar>((int)v); }
+template<> inline uchar saturate_cast<uchar>(unsigned v)     { return (uchar)std::min(v, (unsigned)UCHAR_MAX); }
+template<> inline uchar saturate_cast<uchar>(float v)        { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
+template<> inline uchar saturate_cast<uchar>(double v)       { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
+
+template<> inline schar saturate_cast<schar>(uchar v)        { return (schar)std::min((int)v, SCHAR_MAX); }
+template<> inline schar saturate_cast<schar>(ushort v)       { return (schar)std::min((unsigned)v, (unsigned)SCHAR_MAX); }
+template<> inline schar saturate_cast<schar>(int v)          { return (schar)((unsigned)(v-SCHAR_MIN) <= (unsigned)UCHAR_MAX ? v : v > 0 ? SCHAR_MAX : SCHAR_MIN); }
+template<> inline schar saturate_cast<schar>(short v)        { return saturate_cast<schar>((int)v); }
+template<> inline schar saturate_cast<schar>(unsigned v)     { return (schar)std::min(v, (unsigned)SCHAR_MAX); }
+template<> inline schar saturate_cast<schar>(float v)        { int iv = cvRound(v); return saturate_cast<schar>(iv); }
+template<> inline schar saturate_cast<schar>(double v)       { int iv = cvRound(v); return saturate_cast<schar>(iv); }
+
+template<> inline ushort saturate_cast<ushort>(schar v)      { return (ushort)std::max((int)v, 0); }
+template<> inline ushort saturate_cast<ushort>(short v)      { return (ushort)std::max((int)v, 0); }
+template<> inline ushort saturate_cast<ushort>(int v)        { return (ushort)((unsigned)v <= (unsigned)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
+template<> inline ushort saturate_cast<ushort>(unsigned v)   { return (ushort)std::min(v, (unsigned)USHRT_MAX); }
+template<> inline ushort saturate_cast<ushort>(float v)      { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
+template<> inline ushort saturate_cast<ushort>(double v)     { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
+
+template<> inline short saturate_cast<short>(ushort v)       { return (short)std::min((int)v, SHRT_MAX); }
+template<> inline short saturate_cast<short>(int v)          { return (short)((unsigned)(v - SHRT_MIN) <= (unsigned)USHRT_MAX ? v : v > 0 ? SHRT_MAX : SHRT_MIN); }
+template<> inline short saturate_cast<short>(unsigned v)     { return (short)std::min(v, (unsigned)SHRT_MAX); }
+template<> inline short saturate_cast<short>(float v)        { int iv = cvRound(v); return saturate_cast<short>(iv); }
+template<> inline short saturate_cast<short>(double v)       { int iv = cvRound(v); return saturate_cast<short>(iv); }
+
+template<> inline int saturate_cast<int>(float v)            { return cvRound(v); }
+template<> inline int saturate_cast<int>(double v)           { return cvRound(v); }
+
+// we intentionally do not clip negative numbers, to make -1 become 0xffffffff etc.
+template<> inline unsigned saturate_cast<unsigned>(float v)  { return cvRound(v); }
+template<> inline unsigned saturate_cast<unsigned>(double v) { return cvRound(v); }
+
+
+
+//////////////////////////////// low-level functions ////////////////////////////////
+
+CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+
+CV_EXPORTS int normL1_(const uchar* a, const uchar* b, int n);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n, int cellSize);
+CV_EXPORTS float normL1_(const float* a, const float* b, int n);
+CV_EXPORTS float normL2Sqr_(const float* a, const float* b, int n);
+
+CV_EXPORTS void exp(const float* src, float* dst, int n);
+CV_EXPORTS void log(const float* src, float* dst, int n);
+CV_EXPORTS void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
+CV_EXPORTS void magnitude(const float* x, const float* y, float* dst, int n);
+
+//! computes cube root of the argument
+CV_EXPORTS_W float cubeRoot(float val);
+//! computes the angle in degrees (0..360) of the vector (x,y)
+CV_EXPORTS_W float fastAtan2(float y, float x);
+
+
+
+/////////////////////////////////// inline norms ////////////////////////////////////
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i=0;
+#if CV_ENABLE_UNROLLED
+    for( ; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = a[i], v1 = a[i+1], v2 = a[i+2], v3 = a[i+3];
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = a[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i = 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        s += (_AccTp)std::abs(a[i]) + (_AccTp)std::abs(a[i+1]) +
+            (_AccTp)std::abs(a[i+2]) + (_AccTp)std::abs(a[i+3]);
+    }
+#endif
+    for( ; i < n; i++ )
+        s += std::abs(a[i]);
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+        s = std::max(s, (_AccTp)std::abs(a[i]));
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += v*v;
+    }
+    return s;
+}
+
+template<> inline
+float normL2Sqr(const float* a, const float* b, int n)
+{
+    if( n >= 8 )
+        return normL2Sqr_(a, b, n);
+    float s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        float v = a[i] - b[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += std::abs(v0) + std::abs(v1) + std::abs(v2) + std::abs(v3);
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += std::abs(v);
+    }
+    return s;
+}
+
+template<> inline
+float normL1(const float* a, const float* b, int n)
+{
+    if( n >= 8 )
+        return normL1_(a, b, n);
+    float s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        float v = a[i] - b[i];
+        s += std::abs(v);
+    }
+    return s;
+}
+
+template<> inline
+int normL1(const uchar* a, const uchar* b, int n)
+{
+    return normL1_(a, b, n);
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        _AccTp v0 = a[i] - b[i];
+        s = std::max(s, std::abs(v0));
+    }
+    return s;
+}
+
+
+
+////////////////// forward declarations for important OpenCV types //////////////////
+
+template<typename _Tp, int cn> class Vec;
+template<typename _Tp, int m, int n> class Matx;
+
+template<typename _Tp> class Complex;
+template<typename _Tp> class Point_;
+template<typename _Tp> class Point3_;
+template<typename _Tp> class Size_;
+template<typename _Tp> class Rect_;
+template<typename _Tp> class Scalar_;
+
+class CV_EXPORTS RotatedRect;
+class CV_EXPORTS Range;
+class CV_EXPORTS TermCriteria;
+class CV_EXPORTS KeyPoint;
+class CV_EXPORTS DMatch;
+class CV_EXPORTS RNG;
+
+class CV_EXPORTS Mat;
+class CV_EXPORTS MatExpr;
+
+class CV_EXPORTS SparseMat;
+typedef Mat MatND;
+
+template<typename _Tp> class Mat_;
+template<typename _Tp> class SparseMat_;
+
+class CV_EXPORTS MatConstIterator;
+class CV_EXPORTS SparseMatIterator;
+class CV_EXPORTS SparseMatConstIterator;
+template<typename _Tp> class MatIterator_;
+template<typename _Tp> class MatConstIterator_;
+template<typename _Tp> class SparseMatIterator_;
+template<typename _Tp> class SparseMatConstIterator_;
+
+namespace ogl
+{
+    class CV_EXPORTS Buffer;
+    class CV_EXPORTS Texture2D;
+    class CV_EXPORTS Arrays;
+}
+
+namespace gpu
+{
+    class CV_EXPORTS GpuMat;
+    class CV_EXPORTS CudaMem;
+    class CV_EXPORTS Stream;
+    class CV_EXPORTS Event;
+}
+
+namespace cudev
+{
+    template <typename _Tp> class GpuMat_;
+}
+
+} // cv
+
+#endif //__OPENCV_CORE_BASE_HPP__
diff --git a/modules/core/include/opencv2/core/core.hpp b/modules/core/include/opencv2/core/core.hpp
index 5ceb8ff96c..438918359b 100644
--- a/modules/core/include/opencv2/core/core.hpp
+++ b/modules/core/include/opencv2/core/core.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/core.hpp"
\ No newline at end of file
+#include "opencv2/core.hpp"
diff --git a/modules/core/include/opencv2/core/core_c.h b/modules/core/include/opencv2/core/core_c.h
index df763ab9a4..ca8413ee23 100644
--- a/modules/core/include/opencv2/core/core_c.h
+++ b/modules/core/include/opencv2/core/core_c.h
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -46,6 +47,20 @@
 
 #include "opencv2/core/types_c.h"
 
+#ifdef __cplusplus
+#  ifdef _MSC_VER
+/* disable warning C4190: 'function' has C-linkage specified, but returns UDT 'typename'
+                          which is incompatible with C
+
+   It is OK to disable it because we only extend few plain structures with
+   C++ construrtors for simpler interoperability with C++ API of the library
+*/
+#    pragma warning(disable:4190)
+#  elif defined __clang__ && __clang_major__ >= 3
+#    pragma GCC diagnostic ignored "-Wreturn-type-c-linkage"
+#  endif
+#endif
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -1478,27 +1493,9 @@ CVAPI(int) cvKMeans2( const CvArr* samples, int cluster_count, CvArr* labels,
 *                                    System functions                                    *
 \****************************************************************************************/
 
-/* Add the function pointers table with associated information to the IPP primitives list */
-CVAPI(int)  cvRegisterModule( const CvModuleInfo* module_info );
-
 /* Loads optimized functions from IPP, MKL etc. or switches back to pure C code */
 CVAPI(int)  cvUseOptimized( int on_off );
 
-/* Retrieves information about the registered modules and loaded optimized plugins */
-CVAPI(void)  cvGetModuleInfo( const char* module_name,
-                              const char** version,
-                              const char** loaded_addon_plugins );
-
-typedef void* (CV_CDECL *CvAllocFunc)(size_t size, void* userdata);
-typedef int (CV_CDECL *CvFreeFunc)(void* pptr, void* userdata);
-
-/* Set user-defined memory managment functions (substitutors for malloc and free) that
-   will be called by cvAlloc, cvFree and higher-level functions (e.g. cvCreateImage) */
-CVAPI(void) cvSetMemoryManager( CvAllocFunc alloc_func CV_DEFAULT(NULL),
-                               CvFreeFunc free_func CV_DEFAULT(NULL),
-                               void* userdata CV_DEFAULT(NULL));
-
-
 typedef IplImage* (CV_STDCALL* Cv_iplCreateImageHeader)
                             (int,int,int,char*,char*,int,int,int,int,int,
                             IplROI*,IplImage*,void*,IplTileInfo*);
@@ -1696,18 +1693,6 @@ CVAPI(double) cvGetTickFrequency( void );
 
 /*********************************** CPU capabilities ***********************************/
 
-#define CV_CPU_NONE    0
-#define CV_CPU_MMX     1
-#define CV_CPU_SSE     2
-#define CV_CPU_SSE2    3
-#define CV_CPU_SSE3    4
-#define CV_CPU_SSSE3   5
-#define CV_CPU_SSE4_1  6
-#define CV_CPU_SSE4_2  7
-#define CV_CPU_POPCNT  8
-#define CV_CPU_AVX    10
-#define CV_HARDWARE_MAX_FEATURE 255
-
 CVAPI(int) cvCheckHardwareSupport(int feature);
 
 /*********************************** Multi-Threading ************************************/
@@ -1855,19 +1840,11 @@ static char cvFuncName[] = Name
 #define __CV_EXIT__        goto exit
 
 #ifdef __cplusplus
-}
-
-// classes for automatic module/RTTI data registration/unregistration
-struct CV_EXPORTS CvModule
-{
-    CvModule( CvModuleInfo* _info );
-    ~CvModule();
-    CvModuleInfo* info;
-
-    static CvModuleInfo* first;
-    static CvModuleInfo* last;
-};
+} // extern "C"
+#endif
 
+#ifdef __cplusplus
+// class for automatic module/RTTI data registration/unregistration
 struct CV_EXPORTS CvType
 {
     CvType( const char* type_name,
@@ -1880,6 +1857,396 @@ struct CV_EXPORTS CvType
     static CvTypeInfo* last;
 };
 
+#include "opencv2/core/utility.hpp"
+
+namespace cv
+{
+
+/////////////////////////////////////////// glue ///////////////////////////////////////////
+
+//! converts array (CvMat or IplImage) to cv::Mat
+CV_EXPORTS Mat cvarrToMat(const CvArr* arr, bool copyData=false,
+                          bool allowND=true, int coiMode=0,
+                          AutoBuffer<double>* buf=0);
+
+static inline Mat cvarrToMatND(const CvArr* arr, bool copyData=false, int coiMode=0)
+{
+    return cvarrToMat(arr, copyData, true, coiMode);
+}
+
+
+//! extracts Channel of Interest from CvMat or IplImage and makes cv::Mat out of it.
+CV_EXPORTS void extractImageCOI(const CvArr* arr, OutputArray coiimg, int coi=-1);
+//! inserts single-channel cv::Mat into a multi-channel CvMat or IplImage
+CV_EXPORTS void insertImageCOI(InputArray coiimg, CvArr* arr, int coi=-1);
+
+
+
+////// specialized implementations of DefaultDeleter::operator() for classic OpenCV types //////
+
+template<> CV_EXPORTS void DefaultDeleter<CvMat>::operator ()(CvMat* obj) const;
+template<> CV_EXPORTS void DefaultDeleter<IplImage>::operator ()(IplImage* obj) const;
+template<> CV_EXPORTS void DefaultDeleter<CvMatND>::operator ()(CvMatND* obj) const;
+template<> CV_EXPORTS void DefaultDeleter<CvSparseMat>::operator ()(CvSparseMat* obj) const;
+template<> CV_EXPORTS void DefaultDeleter<CvMemStorage>::operator ()(CvMemStorage* obj) const;
+
+////////////// convenient wrappers for operating old-style dynamic structures //////////////
+
+template<typename _Tp> class SeqIterator;
+
+typedef Ptr<CvMemStorage> MemStorage;
+
+/*!
+ Template Sequence Class derived from CvSeq
+
+ The class provides more convenient access to sequence elements,
+ STL-style operations and iterators.
+
+ \note The class is targeted for simple data types,
+    i.e. no constructors or destructors
+    are called for the sequence elements.
+*/
+template<typename _Tp> class Seq
+{
+public:
+    typedef SeqIterator<_Tp> iterator;
+    typedef SeqIterator<_Tp> const_iterator;
+
+    //! the default constructor
+    Seq();
+    //! the constructor for wrapping CvSeq structure. The real element type in CvSeq should match _Tp.
+    Seq(const CvSeq* seq);
+    //! creates the empty sequence that resides in the specified storage
+    Seq(MemStorage& storage, int headerSize = sizeof(CvSeq));
+    //! returns read-write reference to the specified element
+    _Tp& operator [](int idx);
+    //! returns read-only reference to the specified element
+    const _Tp& operator[](int idx) const;
+    //! returns iterator pointing to the beginning of the sequence
+    SeqIterator<_Tp> begin() const;
+    //! returns iterator pointing to the element following the last sequence element
+    SeqIterator<_Tp> end() const;
+    //! returns the number of elements in the sequence
+    size_t size() const;
+    //! returns the type of sequence elements (CV_8UC1 ... CV_64FC(CV_CN_MAX) ...)
+    int type() const;
+    //! returns the depth of sequence elements (CV_8U ... CV_64F)
+    int depth() const;
+    //! returns the number of channels in each sequence element
+    int channels() const;
+    //! returns the size of each sequence element
+    size_t elemSize() const;
+    //! returns index of the specified sequence element
+    size_t index(const _Tp& elem) const;
+    //! appends the specified element to the end of the sequence
+    void push_back(const _Tp& elem);
+    //! appends the specified element to the front of the sequence
+    void push_front(const _Tp& elem);
+    //! appends zero or more elements to the end of the sequence
+    void push_back(const _Tp* elems, size_t count);
+    //! appends zero or more elements to the front of the sequence
+    void push_front(const _Tp* elems, size_t count);
+    //! inserts the specified element to the specified position
+    void insert(int idx, const _Tp& elem);
+    //! inserts zero or more elements to the specified position
+    void insert(int idx, const _Tp* elems, size_t count);
+    //! removes element at the specified position
+    void remove(int idx);
+    //! removes the specified subsequence
+    void remove(const Range& r);
+
+    //! returns reference to the first sequence element
+    _Tp& front();
+    //! returns read-only reference to the first sequence element
+    const _Tp& front() const;
+    //! returns reference to the last sequence element
+    _Tp& back();
+    //! returns read-only reference to the last sequence element
+    const _Tp& back() const;
+    //! returns true iff the sequence contains no elements
+    bool empty() const;
+
+    //! removes all the elements from the sequence
+    void clear();
+    //! removes the first element from the sequence
+    void pop_front();
+    //! removes the last element from the sequence
+    void pop_back();
+    //! removes zero or more elements from the beginning of the sequence
+    void pop_front(_Tp* elems, size_t count);
+    //! removes zero or more elements from the end of the sequence
+    void pop_back(_Tp* elems, size_t count);
+
+    //! copies the whole sequence or the sequence slice to the specified vector
+    void copyTo(std::vector<_Tp>& vec, const Range& range=Range::all()) const;
+    //! returns the vector containing all the sequence elements
+    operator std::vector<_Tp>() const;
+
+    CvSeq* seq;
+};
+
+
+/*!
+ STL-style Sequence Iterator inherited from the CvSeqReader structure
+*/
+template<typename _Tp> class SeqIterator : public CvSeqReader
+{
+public:
+    //! the default constructor
+    SeqIterator();
+    //! the constructor setting the iterator to the beginning or to the end of the sequence
+    SeqIterator(const Seq<_Tp>& seq, bool seekEnd=false);
+    //! positions the iterator within the sequence
+    void seek(size_t pos);
+    //! reports the current iterator position
+    size_t tell() const;
+    //! returns reference to the current sequence element
+    _Tp& operator *();
+    //! returns read-only reference to the current sequence element
+    const _Tp& operator *() const;
+    //! moves iterator to the next sequence element
+    SeqIterator& operator ++();
+    //! moves iterator to the next sequence element
+    SeqIterator operator ++(int) const;
+    //! moves iterator to the previous sequence element
+    SeqIterator& operator --();
+    //! moves iterator to the previous sequence element
+    SeqIterator operator --(int) const;
+
+    //! moves iterator forward by the specified offset (possibly negative)
+    SeqIterator& operator +=(int);
+    //! moves iterator backward by the specified offset (possibly negative)
+    SeqIterator& operator -=(int);
+
+    // this is index of the current element module seq->total*2
+    // (to distinguish between 0 and seq->total)
+    int index;
+};
+
+
+
+// bridge C++ => C Seq API
+CV_EXPORTS schar*  seqPush( CvSeq* seq, const void* element=0);
+CV_EXPORTS schar*  seqPushFront( CvSeq* seq, const void* element=0);
+CV_EXPORTS void  seqPop( CvSeq* seq, void* element=0);
+CV_EXPORTS void  seqPopFront( CvSeq* seq, void* element=0);
+CV_EXPORTS void  seqPopMulti( CvSeq* seq, void* elements,
+                              int count, int in_front=0 );
+CV_EXPORTS void  seqRemove( CvSeq* seq, int index );
+CV_EXPORTS void  clearSeq( CvSeq* seq );
+CV_EXPORTS schar*  getSeqElem( const CvSeq* seq, int index );
+CV_EXPORTS void  seqRemoveSlice( CvSeq* seq, CvSlice slice );
+CV_EXPORTS void  seqInsertSlice( CvSeq* seq, int before_index, const CvArr* from_arr );
+
+template<typename _Tp> inline Seq<_Tp>::Seq() : seq(0) {}
+template<typename _Tp> inline Seq<_Tp>::Seq( const CvSeq* _seq ) : seq((CvSeq*)_seq)
+{
+    CV_Assert(!_seq || _seq->elem_size == sizeof(_Tp));
+}
+
+template<typename _Tp> inline Seq<_Tp>::Seq( MemStorage& storage,
+                                             int headerSize )
+{
+    CV_Assert(headerSize >= (int)sizeof(CvSeq));
+    seq = cvCreateSeq(DataType<_Tp>::type, headerSize, sizeof(_Tp), storage);
+}
+
+template<typename _Tp> inline _Tp& Seq<_Tp>::operator [](int idx)
+{ return *(_Tp*)getSeqElem(seq, idx); }
+
+template<typename _Tp> inline const _Tp& Seq<_Tp>::operator [](int idx) const
+{ return *(_Tp*)getSeqElem(seq, idx); }
+
+template<typename _Tp> inline SeqIterator<_Tp> Seq<_Tp>::begin() const
+{ return SeqIterator<_Tp>(*this); }
+
+template<typename _Tp> inline SeqIterator<_Tp> Seq<_Tp>::end() const
+{ return SeqIterator<_Tp>(*this, true); }
+
+template<typename _Tp> inline size_t Seq<_Tp>::size() const
+{ return seq ? seq->total : 0; }
+
+template<typename _Tp> inline int Seq<_Tp>::type() const
+{ return seq ? CV_MAT_TYPE(seq->flags) : 0; }
+
+template<typename _Tp> inline int Seq<_Tp>::depth() const
+{ return seq ? CV_MAT_DEPTH(seq->flags) : 0; }
+
+template<typename _Tp> inline int Seq<_Tp>::channels() const
+{ return seq ? CV_MAT_CN(seq->flags) : 0; }
+
+template<typename _Tp> inline size_t Seq<_Tp>::elemSize() const
+{ return seq ? seq->elem_size : 0; }
+
+template<typename _Tp> inline size_t Seq<_Tp>::index(const _Tp& elem) const
+{ return cvSeqElemIdx(seq, &elem); }
+
+template<typename _Tp> inline void Seq<_Tp>::push_back(const _Tp& elem)
+{ cvSeqPush(seq, &elem); }
+
+template<typename _Tp> inline void Seq<_Tp>::push_front(const _Tp& elem)
+{ cvSeqPushFront(seq, &elem); }
+
+template<typename _Tp> inline void Seq<_Tp>::push_back(const _Tp* elem, size_t count)
+{ cvSeqPushMulti(seq, elem, (int)count, 0); }
+
+template<typename _Tp> inline void Seq<_Tp>::push_front(const _Tp* elem, size_t count)
+{ cvSeqPushMulti(seq, elem, (int)count, 1); }
+
+template<typename _Tp> inline _Tp& Seq<_Tp>::back()
+{ return *(_Tp*)getSeqElem(seq, -1); }
+
+template<typename _Tp> inline const _Tp& Seq<_Tp>::back() const
+{ return *(const _Tp*)getSeqElem(seq, -1); }
+
+template<typename _Tp> inline _Tp& Seq<_Tp>::front()
+{ return *(_Tp*)getSeqElem(seq, 0); }
+
+template<typename _Tp> inline const _Tp& Seq<_Tp>::front() const
+{ return *(const _Tp*)getSeqElem(seq, 0); }
+
+template<typename _Tp> inline bool Seq<_Tp>::empty() const
+{ return !seq || seq->total == 0; }
+
+template<typename _Tp> inline void Seq<_Tp>::clear()
+{ if(seq) clearSeq(seq); }
+
+template<typename _Tp> inline void Seq<_Tp>::pop_back()
+{ seqPop(seq); }
+
+template<typename _Tp> inline void Seq<_Tp>::pop_front()
+{ seqPopFront(seq); }
+
+template<typename _Tp> inline void Seq<_Tp>::pop_back(_Tp* elem, size_t count)
+{ seqPopMulti(seq, elem, (int)count, 0); }
+
+template<typename _Tp> inline void Seq<_Tp>::pop_front(_Tp* elem, size_t count)
+{ seqPopMulti(seq, elem, (int)count, 1); }
+
+template<typename _Tp> inline void Seq<_Tp>::insert(int idx, const _Tp& elem)
+{ seqInsert(seq, idx, &elem); }
+
+template<typename _Tp> inline void Seq<_Tp>::insert(int idx, const _Tp* elems, size_t count)
+{
+    CvMat m = cvMat(1, count, DataType<_Tp>::type, elems);
+    seqInsertSlice(seq, idx, &m);
+}
+
+template<typename _Tp> inline void Seq<_Tp>::remove(int idx)
+{ seqRemove(seq, idx); }
+
+template<typename _Tp> inline void Seq<_Tp>::remove(const Range& r)
+{ seqRemoveSlice(seq, cvSlice(r.start, r.end)); }
+
+template<typename _Tp> inline void Seq<_Tp>::copyTo(std::vector<_Tp>& vec, const Range& range) const
+{
+    size_t len = !seq ? 0 : range == Range::all() ? seq->total : range.end - range.start;
+    vec.resize(len);
+    if( seq && len )
+        cvCvtSeqToArray(seq, &vec[0], range);
+}
+
+template<typename _Tp> inline Seq<_Tp>::operator std::vector<_Tp>() const
+{
+    std::vector<_Tp> vec;
+    copyTo(vec);
+    return vec;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp>::SeqIterator()
+{ memset(this, 0, sizeof(*this)); }
+
+template<typename _Tp> inline SeqIterator<_Tp>::SeqIterator(const Seq<_Tp>& _seq, bool seekEnd)
+{
+    cvStartReadSeq(_seq.seq, this);
+    index = seekEnd ? _seq.seq->total : 0;
+}
+
+template<typename _Tp> inline void SeqIterator<_Tp>::seek(size_t pos)
+{
+    cvSetSeqReaderPos(this, (int)pos, false);
+    index = pos;
+}
+
+template<typename _Tp> inline size_t SeqIterator<_Tp>::tell() const
+{ return index; }
+
+template<typename _Tp> inline _Tp& SeqIterator<_Tp>::operator *()
+{ return *(_Tp*)ptr; }
+
+template<typename _Tp> inline const _Tp& SeqIterator<_Tp>::operator *() const
+{ return *(const _Tp*)ptr; }
+
+template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator ++()
+{
+    CV_NEXT_SEQ_ELEM(sizeof(_Tp), *this);
+    if( ++index >= seq->total*2 )
+        index = 0;
+    return *this;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp> SeqIterator<_Tp>::operator ++(int) const
+{
+    SeqIterator<_Tp> it = *this;
+    ++*this;
+    return it;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator --()
+{
+    CV_PREV_SEQ_ELEM(sizeof(_Tp), *this);
+    if( --index < 0 )
+        index = seq->total*2-1;
+    return *this;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp> SeqIterator<_Tp>::operator --(int) const
+{
+    SeqIterator<_Tp> it = *this;
+    --*this;
+    return it;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator +=(int delta)
+{
+    cvSetSeqReaderPos(this, delta, 1);
+    index += delta;
+    int n = seq->total*2;
+    if( index < 0 )
+        index += n;
+    if( index >= n )
+        index -= n;
+    return *this;
+}
+
+template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator -=(int delta)
+{
+    return (*this += -delta);
+}
+
+template<typename _Tp> inline ptrdiff_t operator - (const SeqIterator<_Tp>& a,
+                                                    const SeqIterator<_Tp>& b)
+{
+    ptrdiff_t delta = a.index - b.index, n = a.seq->total;
+    if( delta > n || delta < -n )
+        delta += delta < 0 ? n : -n;
+    return delta;
+}
+
+template<typename _Tp> inline bool operator == (const SeqIterator<_Tp>& a,
+                                                const SeqIterator<_Tp>& b)
+{
+    return a.seq == b.seq && a.index == b.index;
+}
+
+template<typename _Tp> inline bool operator != (const SeqIterator<_Tp>& a,
+                                                const SeqIterator<_Tp>& b)
+{
+    return !(a == b);
+}
+
+} // cv
+
 #endif
 
 #endif
diff --git a/modules/gpu/include/opencv2/gpu/device/block.hpp b/modules/core/include/opencv2/core/cuda/block.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/block.hpp
rename to modules/core/include/opencv2/core/cuda/block.hpp
index 86ce205bc7..04bfdba714 100644
--- a/modules/gpu/include/opencv2/gpu/device/block.hpp
+++ b/modules/core/include/opencv2/core/cuda/block.hpp
@@ -43,7 +43,7 @@
 #ifndef __OPENCV_GPU_DEVICE_BLOCK_HPP__
 #define __OPENCV_GPU_DEVICE_BLOCK_HPP__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     struct Block
     {
@@ -201,5 +201,3 @@ namespace cv { namespace gpu { namespace device
 }}}
 
 #endif /* __OPENCV_GPU_DEVICE_BLOCK_HPP__ */
-
-
diff --git a/modules/gpu/include/opencv2/gpu/device/border_interpolate.hpp b/modules/core/include/opencv2/core/cuda/border_interpolate.hpp
similarity index 96%
rename from modules/gpu/include/opencv2/gpu/device/border_interpolate.hpp
rename to modules/core/include/opencv2/core/cuda/border_interpolate.hpp
index 2343ccab2a..6c53f09eee 100644
--- a/modules/gpu/include/opencv2/gpu/device/border_interpolate.hpp
+++ b/modules/core/include/opencv2/core/cuda/border_interpolate.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,7 +47,7 @@
 #include "vec_traits.hpp"
 #include "vec_math.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     //////////////////////////////////////////////////////////////
     // BrdConstant
@@ -73,8 +73,8 @@ namespace cv { namespace gpu { namespace device
             return (x >= 0 && x < width) ? saturate_cast<D>(data[x]) : val;
         }
 
-        const int width;
-        const D val;
+        int width;
+        D val;
     };
 
     template <typename D> struct BrdColConstant
@@ -98,8 +98,8 @@ namespace cv { namespace gpu { namespace device
             return (y >= 0 && y < height) ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
         }
 
-        const int height;
-        const D val;
+        int height;
+        D val;
     };
 
     template <typename D> struct BrdConstant
@@ -120,9 +120,9 @@ namespace cv { namespace gpu { namespace device
             return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
         }
 
-        const int height;
-        const int width;
-        const D val;
+        int height;
+        int width;
+        D val;
     };
 
     //////////////////////////////////////////////////////////////
@@ -165,7 +165,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(data[idx_col(x)]);
         }
 
-        const int last_col;
+        int last_col;
     };
 
     template <typename D> struct BrdColReplicate
@@ -205,7 +205,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(*(const T*)((const char*)data + idx_row(y) * step));
         }
 
-        const int last_row;
+        int last_row;
     };
 
     template <typename D> struct BrdReplicate
@@ -255,8 +255,8 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(src(idx_row(y), idx_col(x)));
         }
 
-        const int last_row;
-        const int last_col;
+        int last_row;
+        int last_col;
     };
 
     //////////////////////////////////////////////////////////////
@@ -299,7 +299,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(data[idx_col(x)]);
         }
 
-        const int last_col;
+        int last_col;
     };
 
     template <typename D> struct BrdColReflect101
@@ -339,7 +339,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
         }
 
-        const int last_row;
+        int last_row;
     };
 
     template <typename D> struct BrdReflect101
@@ -389,8 +389,8 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(src(idx_row(y), idx_col(x)));
         }
 
-        const int last_row;
-        const int last_col;
+        int last_row;
+        int last_col;
     };
 
     //////////////////////////////////////////////////////////////
@@ -433,7 +433,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(data[idx_col(x)]);
         }
 
-        const int last_col;
+        int last_col;
     };
 
     template <typename D> struct BrdColReflect
@@ -473,7 +473,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
         }
 
-        const int last_row;
+        int last_row;
     };
 
     template <typename D> struct BrdReflect
@@ -523,8 +523,8 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(src(idx_row(y), idx_col(x)));
         }
 
-        const int last_row;
-        const int last_col;
+        int last_row;
+        int last_col;
     };
 
     //////////////////////////////////////////////////////////////
@@ -567,7 +567,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(data[idx_col(x)]);
         }
 
-        const int width;
+        int width;
     };
 
     template <typename D> struct BrdColWrap
@@ -607,7 +607,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
         }
 
-        const int height;
+        int height;
     };
 
     template <typename D> struct BrdWrap
@@ -664,8 +664,8 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<D>(src(idx_row(y), idx_col(x)));
         }
 
-        const int height;
-        const int width;
+        int height;
+        int width;
     };
 
     //////////////////////////////////////////////////////////////
@@ -683,8 +683,8 @@ namespace cv { namespace gpu { namespace device
             return b.at(y, x, ptr);
         }
 
-        const Ptr2D ptr;
-        const B b;
+        Ptr2D ptr;
+        B b;
     };
 
     // under win32 there is some bug with templated types that passed as kernel parameters
@@ -704,11 +704,11 @@ namespace cv { namespace gpu { namespace device
             return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
         }
 
-        const Ptr2D src;
-        const int height;
-        const int width;
-        const D val;
+        Ptr2D src;
+        int height;
+        int width;
+        D val;
     };
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_BORDER_INTERPOLATE_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/color.hpp b/modules/core/include/opencv2/core/cuda/color.hpp
similarity index 92%
rename from modules/gpu/include/opencv2/gpu/device/color.hpp
rename to modules/core/include/opencv2/core/cuda/color.hpp
index 53d3d6b086..d5f94c92a4 100644
--- a/modules/gpu/include/opencv2/gpu/device/color.hpp
+++ b/modules/core/include/opencv2/core/cuda/color.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -45,7 +45,7 @@
 
 #include "detail/color_detail.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     // All OPENCV_GPU_IMPLEMENT_*_TRAITS(ColorSpace1_to_ColorSpace2, ...) macros implements
     // template <typename T> class ColorSpace1_to_ColorSpace2_traits
@@ -107,25 +107,25 @@ namespace cv { namespace gpu { namespace device
 
     #undef OPENCV_GPU_IMPLEMENT_RGB2GRAY_TRAITS
 
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv, 3, 3, 0)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv, 4, 3, 0)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv4, 3, 4, 0)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv4, 4, 4, 0)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv, 3, 3, 2)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv, 4, 3, 2)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv4, 3, 4, 2)
-    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv4, 4, 4, 2)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv, 3, 3, 2)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv, 4, 3, 2)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv4, 3, 4, 2)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv4, 4, 4, 2)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv, 3, 3, 0)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv, 4, 3, 0)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv4, 3, 4, 0)
+    OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv4, 4, 4, 0)
 
     #undef OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS
 
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgb, 3, 3, 0)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgba, 3, 4, 0)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgb, 4, 3, 0)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgba, 4, 4, 0)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgr, 3, 3, 2)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgra, 3, 4, 2)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgr, 4, 3, 2)
-    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgra, 4, 4, 2)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgb, 3, 3, 2)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgba, 3, 4, 2)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgb, 4, 3, 2)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgba, 4, 4, 2)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgr, 3, 3, 0)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgra, 3, 4, 0)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgr, 4, 3, 0)
+    OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgra, 4, 4, 0)
 
     #undef OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS
 
@@ -296,6 +296,6 @@ namespace cv { namespace gpu { namespace device
     OPENCV_GPU_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lbgra, 4, 4, false, 0)
 
     #undef OPENCV_GPU_IMPLEMENT_Luv2RGB_TRAITS
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_BORDER_INTERPOLATE_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/common.hpp b/modules/core/include/opencv2/core/cuda/common.hpp
similarity index 74%
rename from modules/gpu/include/opencv2/gpu/device/common.hpp
rename to modules/core/include/opencv2/core/cuda/common.hpp
index 64d82c83b0..6b3fef19a4 100644
--- a/modules/gpu/include/opencv2/gpu/device/common.hpp
+++ b/modules/core/include/opencv2/core/cuda/common.hpp
@@ -44,11 +44,9 @@
 #define __OPENCV_GPU_COMMON_HPP__
 
 #include <cuda_runtime.h>
-#include "opencv2/core/cuda_devptrs.hpp"
-
-#ifndef CV_PI
-    #define CV_PI   3.1415926535897932384626433832795
-#endif
+#include "opencv2/core/gpu_types.hpp"
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/base.hpp"
 
 #ifndef CV_PI_F
     #ifndef CV_PI
@@ -58,16 +56,24 @@
     #endif
 #endif
 
-#if defined(__GNUC__)
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__)
+namespace cv { namespace gpu {
+    static inline void checkCudaError(cudaError_t err, const char* file, const int line, const char* func)
+    {
+        if (cudaSuccess != err)
+            cv::error(cv::Error::GpuApiCallError, cudaGetErrorString(err), func, file, line);
+    }
+}}
+
+#ifndef cudaSafeCall
+    #if defined(__GNUC__)
+        #define cudaSafeCall(expr)  cv::gpu::checkCudaError(expr, __FILE__, __LINE__, __func__)
+    #else /* defined(__CUDACC__) || defined(__MSVC__) */
+        #define cudaSafeCall(expr)  cv::gpu::checkCudaError(expr, __FILE__, __LINE__, "")
+    #endif
 #endif
 
 namespace cv { namespace gpu
 {
-    void error(const char *error_string, const char *file, const int line, const char *func);
-
     template <typename T> static inline bool isAligned(const T* ptr, size_t size)
     {
         return reinterpret_cast<size_t>(ptr) % size == 0;
@@ -79,37 +85,20 @@ namespace cv { namespace gpu
     }
 }}
 
-static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
-{
-    if (cudaSuccess != err)
-        cv::gpu::error(cudaGetErrorString(err), file, line, func);
-}
-
 namespace cv { namespace gpu
 {
-    __host__ __device__ __forceinline__ int divUp(int total, int grain)
+    namespace cudev
     {
-        return (total + grain - 1) / grain;
-    }
-
-    namespace device
-    {
-        using cv::gpu::divUp;
-
-#ifdef __CUDACC__
-        typedef unsigned char uchar;
-        typedef unsigned short ushort;
-        typedef signed char schar;
-        #if defined (_WIN32) || defined (__APPLE__)
-            typedef unsigned int uint;
-        #endif
+        __host__ __device__ __forceinline__ int divUp(int total, int grain)
+        {
+            return (total + grain - 1) / grain;
+        }
 
         template<class T> inline void bindTexture(const textureReference* tex, const PtrStepSz<T>& img)
         {
             cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
             cudaSafeCall( cudaBindTexture2D(0, tex, img.ptr(), &desc, img.cols, img.rows, img.step) );
         }
-#endif // __CUDACC__
     }
 }}
 
diff --git a/modules/gpu/include/opencv2/gpu/device/datamov_utils.hpp b/modules/core/include/opencv2/core/cuda/datamov_utils.hpp
similarity index 96%
rename from modules/gpu/include/opencv2/gpu/device/datamov_utils.hpp
rename to modules/core/include/opencv2/core/cuda/datamov_utils.hpp
index e05a22477c..10df540936 100644
--- a/modules/gpu/include/opencv2/gpu/device/datamov_utils.hpp
+++ b/modules/core/include/opencv2/core/cuda/datamov_utils.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -45,7 +45,7 @@
 
 #include "common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 200
 
@@ -100,6 +100,6 @@ namespace cv { namespace gpu { namespace device
         #undef OPENCV_GPU_ASM_PTR
 
     #endif // __CUDA_ARCH__ >= 200
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_DATAMOV_UTILS_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/color_detail.hpp b/modules/core/include/opencv2/core/cuda/detail/color_detail.hpp
similarity index 94%
rename from modules/gpu/include/opencv2/gpu/device/detail/color_detail.hpp
rename to modules/core/include/opencv2/core/cuda/detail/color_detail.hpp
index 0d9c08620d..a0a673410a 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/color_detail.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/color_detail.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -49,7 +49,7 @@
 #include "../limits.hpp"
 #include "../functional.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     #ifndef CV_DESCALE
         #define CV_DESCALE(x, n) (((x) + (1 << ((n)-1))) >> (n))
@@ -120,11 +120,8 @@ namespace cv { namespace gpu { namespace device
                 return dst;
             }
 
-            __device__ __forceinline__ RGB2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-
-            __device__ __forceinline__ RGB2RGB(const RGB2RGB& other_)
-                :unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2RGB() {}
+            __host__ __device__ __forceinline__ RGB2RGB(const RGB2RGB&) {}
         };
 
         template <> struct RGB2RGB<uchar, 4, 4, 2> : unary_function<uint, uint>
@@ -141,15 +138,15 @@ namespace cv { namespace gpu { namespace device
                 return dst;
             }
 
-            __device__ __forceinline__ RGB2RGB():unary_function<uint, uint>(){}
-            __device__ __forceinline__ RGB2RGB(const RGB2RGB& other_):unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ RGB2RGB() {}
+            __host__ __device__ __forceinline__ RGB2RGB(const RGB2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2RGB<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2RGB<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -203,8 +200,8 @@ namespace cv { namespace gpu { namespace device
                 return RGB2RGB5x5Converter<green_bits, bidx>::cvt(src);
             }
 
-            __device__ __forceinline__ RGB2RGB5x5():unary_function<uchar3, ushort>(){}
-            __device__ __forceinline__ RGB2RGB5x5(const RGB2RGB5x5& other_):unary_function<uchar3, ushort>(){}
+            __host__ __device__ __forceinline__ RGB2RGB5x5() {}
+            __host__ __device__ __forceinline__ RGB2RGB5x5(const RGB2RGB5x5&) {}
         };
 
         template<int bidx, int green_bits> struct RGB2RGB5x5<4, bidx,green_bits> : unary_function<uint, ushort>
@@ -214,15 +211,15 @@ namespace cv { namespace gpu { namespace device
                 return RGB2RGB5x5Converter<green_bits, bidx>::cvt(src);
             }
 
-            __device__ __forceinline__ RGB2RGB5x5():unary_function<uint, ushort>(){}
-            __device__ __forceinline__ RGB2RGB5x5(const RGB2RGB5x5& other_):unary_function<uint, ushort>(){}
+            __host__ __device__ __forceinline__ RGB2RGB5x5() {}
+            __host__ __device__ __forceinline__ RGB2RGB5x5(const RGB2RGB5x5&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2RGB5x5_TRAITS(name, scn, bidx, green_bits) \
     struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2RGB5x5<scn, bidx, green_bits> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2RGB5x5<scn, bidx, green_bits> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -282,8 +279,8 @@ namespace cv { namespace gpu { namespace device
                 RGB5x52RGBConverter<green_bits, bidx>::cvt(src, dst);
                 return dst;
             }
-            __device__ __forceinline__ RGB5x52RGB():unary_function<ushort, uchar3>(){}
-            __device__ __forceinline__ RGB5x52RGB(const RGB5x52RGB& other_):unary_function<ushort, uchar3>(){}
+            __host__ __device__ __forceinline__ RGB5x52RGB() {}
+            __host__ __device__ __forceinline__ RGB5x52RGB(const RGB5x52RGB&) {}
 
         };
 
@@ -295,15 +292,15 @@ namespace cv { namespace gpu { namespace device
                 RGB5x52RGBConverter<green_bits, bidx>::cvt(src, dst);
                 return dst;
             }
-            __device__ __forceinline__ RGB5x52RGB():unary_function<ushort, uint>(){}
-            __device__ __forceinline__ RGB5x52RGB(const RGB5x52RGB& other_):unary_function<ushort, uint>(){}
+            __host__ __device__ __forceinline__ RGB5x52RGB() {}
+            __host__ __device__ __forceinline__ RGB5x52RGB(const RGB5x52RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB5x52RGB_TRAITS(name, dcn, bidx, green_bits) \
     struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB5x52RGB<dcn, bidx, green_bits> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB5x52RGB<dcn, bidx, green_bits> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -325,9 +322,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Gray2RGB():unary_function<T, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ Gray2RGB(const Gray2RGB& other_)
-                : unary_function<T, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ Gray2RGB() {}
+            __host__ __device__ __forceinline__ Gray2RGB(const Gray2RGB&) {}
         };
 
         template <> struct Gray2RGB<uchar, 4> : unary_function<uchar, uint>
@@ -342,15 +338,15 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Gray2RGB():unary_function<uchar, uint>(){}
-            __device__ __forceinline__ Gray2RGB(const Gray2RGB& other_):unary_function<uchar, uint>(){}
+            __host__ __device__ __forceinline__ Gray2RGB() {}
+            __host__ __device__ __forceinline__ Gray2RGB(const Gray2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_GRAY2RGB_TRAITS(name, dcn) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::Gray2RGB<T, dcn> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::Gray2RGB<T, dcn> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -384,15 +380,15 @@ namespace cv { namespace gpu { namespace device
                 return Gray2RGB5x5Converter<green_bits>::cvt(src);
             }
 
-            __device__ __forceinline__ Gray2RGB5x5():unary_function<uchar, ushort>(){}
-            __device__ __forceinline__ Gray2RGB5x5(const Gray2RGB5x5& other_):unary_function<uchar, ushort>(){}
+            __host__ __device__ __forceinline__ Gray2RGB5x5() {}
+            __host__ __device__ __forceinline__ Gray2RGB5x5(const Gray2RGB5x5&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_GRAY2RGB5x5_TRAITS(name, green_bits) \
     struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::Gray2RGB5x5<green_bits> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::Gray2RGB5x5<green_bits> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -426,15 +422,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB5x52GrayConverter<green_bits>::cvt(src);
             }
-            __device__ __forceinline__ RGB5x52Gray() : unary_function<ushort, uchar>(){}
-            __device__ __forceinline__ RGB5x52Gray(const RGB5x52Gray& other_) : unary_function<ushort, uchar>(){}
+            __host__ __device__ __forceinline__ RGB5x52Gray() {}
+            __host__ __device__ __forceinline__ RGB5x52Gray(const RGB5x52Gray&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB5x52GRAY_TRAITS(name, green_bits) \
     struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB5x52Gray<green_bits> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB5x52Gray<green_bits> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -467,9 +463,8 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB2GrayConvert<bidx>(&src.x);
             }
-            __device__ __forceinline__ RGB2Gray() : unary_function<typename TypeVec<T, scn>::vec_type, T>(){}
-            __device__ __forceinline__ RGB2Gray(const RGB2Gray& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, T>(){}
+            __host__ __device__ __forceinline__ RGB2Gray() {}
+            __host__ __device__ __forceinline__ RGB2Gray(const RGB2Gray&) {}
         };
 
         template <int bidx> struct RGB2Gray<uchar, 4, bidx> : unary_function<uint, uchar>
@@ -478,15 +473,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB2GrayConvert<bidx>(src);
             }
-            __device__ __forceinline__ RGB2Gray() : unary_function<uint, uchar>(){}
-            __device__ __forceinline__ RGB2Gray(const RGB2Gray& other_) : unary_function<uint, uchar>(){}
+            __host__ __device__ __forceinline__ RGB2Gray() {}
+            __host__ __device__ __forceinline__ RGB2Gray(const RGB2Gray&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2GRAY_TRAITS(name, scn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2Gray<T, scn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2Gray<T, scn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -529,17 +524,15 @@ namespace cv { namespace gpu { namespace device
                 RGB2YUVConvert<bidx>(&src.x, dst);
                 return dst;
             }
-            __device__ __forceinline__ RGB2YUV()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ RGB2YUV(const RGB2YUV& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2YUV() {}
+            __host__ __device__ __forceinline__ RGB2YUV(const RGB2YUV&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2YUV_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2YUV<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2YUV<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -609,10 +602,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ YUV2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ YUV2RGB(const YUV2RGB& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ YUV2RGB() {}
+            __host__ __device__ __forceinline__ YUV2RGB(const YUV2RGB&) {}
         };
 
         template <int bidx> struct YUV2RGB<uchar, 4, 4, bidx> : unary_function<uint, uint>
@@ -621,15 +612,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return YUV2RGBConvert<bidx>(src);
             }
-            __device__ __forceinline__ YUV2RGB() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ YUV2RGB(const YUV2RGB& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ YUV2RGB() {}
+            __host__ __device__ __forceinline__ YUV2RGB(const YUV2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_YUV2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::YUV2RGB<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::YUV2RGB<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -689,10 +680,8 @@ namespace cv { namespace gpu { namespace device
                 RGB2YCrCbConvert<bidx>(&src.x, dst);
                 return dst;
             }
-            __device__ __forceinline__ RGB2YCrCb()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ RGB2YCrCb(const RGB2YCrCb& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2YCrCb() {}
+            __host__ __device__ __forceinline__ RGB2YCrCb(const RGB2YCrCb&) {}
         };
 
         template <int bidx> struct RGB2YCrCb<uchar, 4, 4, bidx> : unary_function<uint, uint>
@@ -702,15 +691,15 @@ namespace cv { namespace gpu { namespace device
                 return RGB2YCrCbConvert<bidx>(src);
             }
 
-            __device__ __forceinline__ RGB2YCrCb() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ RGB2YCrCb(const RGB2YCrCb& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ RGB2YCrCb() {}
+            __host__ __device__ __forceinline__ RGB2YCrCb(const RGB2YCrCb&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2YCrCb_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2YCrCb<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2YCrCb<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -771,10 +760,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ YCrCb2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ YCrCb2RGB(const YCrCb2RGB& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ YCrCb2RGB() {}
+            __host__ __device__ __forceinline__ YCrCb2RGB(const YCrCb2RGB&) {}
         };
 
         template <int bidx> struct YCrCb2RGB<uchar, 4, 4, bidx> : unary_function<uint, uint>
@@ -783,15 +770,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return YCrCb2RGBConvert<bidx>(src);
             }
-            __device__ __forceinline__ YCrCb2RGB() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ YCrCb2RGB(const YCrCb2RGB& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ YCrCb2RGB() {}
+            __host__ __device__ __forceinline__ YCrCb2RGB(const YCrCb2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_YCrCb2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::YCrCb2RGB<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::YCrCb2RGB<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -849,10 +836,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2XYZ()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ RGB2XYZ(const RGB2XYZ& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2XYZ() {}
+            __host__ __device__ __forceinline__ RGB2XYZ(const RGB2XYZ&) {}
         };
 
         template <int bidx> struct RGB2XYZ<uchar, 4, 4, bidx> : unary_function<uint, uint>
@@ -861,15 +846,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB2XYZConvert<bidx>(src);
             }
-            __device__ __forceinline__ RGB2XYZ() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ RGB2XYZ(const RGB2XYZ& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ RGB2XYZ() {}
+            __host__ __device__ __forceinline__ RGB2XYZ(const RGB2XYZ&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2XYZ_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2XYZ<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2XYZ<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -926,10 +911,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ XYZ2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ XYZ2RGB(const XYZ2RGB& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ XYZ2RGB() {}
+            __host__ __device__ __forceinline__ XYZ2RGB(const XYZ2RGB&) {}
         };
 
         template <int bidx> struct XYZ2RGB<uchar, 4, 4, bidx> : unary_function<uint, uint>
@@ -938,15 +921,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return XYZ2RGBConvert<bidx>(src);
             }
-            __device__ __forceinline__ XYZ2RGB() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ XYZ2RGB(const XYZ2RGB& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ XYZ2RGB() {}
+            __host__ __device__ __forceinline__ XYZ2RGB(const XYZ2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_XYZ2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::XYZ2RGB<T, scn, dcn, bidx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::XYZ2RGB<T, scn, dcn, bidx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1066,10 +1049,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2HSV()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ RGB2HSV(const RGB2HSV& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2HSV() {}
+            __host__ __device__ __forceinline__ RGB2HSV(const RGB2HSV&) {}
         };
 
         template <int bidx, int hr> struct RGB2HSV<uchar, 4, 4, bidx, hr> : unary_function<uint, uint>
@@ -1078,15 +1059,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB2HSVConvert<bidx, hr>(src);
             }
-            __device__ __forceinline__ RGB2HSV():unary_function<uint, uint>(){}
-            __device__ __forceinline__ RGB2HSV(const RGB2HSV& other_):unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ RGB2HSV() {}
+            __host__ __device__ __forceinline__ RGB2HSV(const RGB2HSV&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2HSV_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HSV<T, scn, dcn, bidx, 180> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HSV<T, scn, dcn, bidx, 180> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1094,7 +1075,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <typename T> struct name ## _full_traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HSV<T, scn, dcn, bidx, 256> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HSV<T, scn, dcn, bidx, 256> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1102,7 +1083,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HSV<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HSV<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1110,7 +1091,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _full_traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HSV<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HSV<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1208,10 +1189,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ HSV2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ HSV2RGB(const HSV2RGB& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ HSV2RGB() {}
+            __host__ __device__ __forceinline__ HSV2RGB(const HSV2RGB&) {}
         };
 
         template <int bidx, int hr> struct HSV2RGB<uchar, 4, 4, bidx, hr> : unary_function<uint, uint>
@@ -1220,15 +1199,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return HSV2RGBConvert<bidx, hr>(src);
             }
-            __device__ __forceinline__ HSV2RGB():unary_function<uint, uint>(){}
-            __device__ __forceinline__ HSV2RGB(const HSV2RGB& other_):unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ HSV2RGB() {}
+            __host__ __device__ __forceinline__ HSV2RGB(const HSV2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_HSV2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::HSV2RGB<T, scn, dcn, bidx, 180> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HSV2RGB<T, scn, dcn, bidx, 180> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1236,7 +1215,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <typename T> struct name ## _full_traits \
     { \
-        typedef ::cv::gpu::device::color_detail::HSV2RGB<T, scn, dcn, bidx, 255> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HSV2RGB<T, scn, dcn, bidx, 255> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1244,7 +1223,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::HSV2RGB<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HSV2RGB<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1252,7 +1231,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _full_traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::HSV2RGB<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HSV2RGB<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1343,10 +1322,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2HLS()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ RGB2HLS(const RGB2HLS& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ RGB2HLS() {}
+            __host__ __device__ __forceinline__ RGB2HLS(const RGB2HLS&) {}
         };
 
         template <int bidx, int hr> struct RGB2HLS<uchar, 4, 4, bidx, hr> : unary_function<uint, uint>
@@ -1355,15 +1332,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return RGB2HLSConvert<bidx, hr>(src);
             }
-            __device__ __forceinline__ RGB2HLS() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ RGB2HLS(const RGB2HLS& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ RGB2HLS() {}
+            __host__ __device__ __forceinline__ RGB2HLS(const RGB2HLS&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2HLS_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HLS<T, scn, dcn, bidx, 180> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HLS<T, scn, dcn, bidx, 180> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1371,7 +1348,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <typename T> struct name ## _full_traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HLS<T, scn, dcn, bidx, 256> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HLS<T, scn, dcn, bidx, 256> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1379,7 +1356,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HLS<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HLS<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1387,7 +1364,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _full_traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2HLS<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2HLS<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1485,10 +1462,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ HLS2RGB()
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
-            __device__ __forceinline__ HLS2RGB(const HLS2RGB& other_)
-                : unary_function<typename TypeVec<T, scn>::vec_type, typename TypeVec<T, dcn>::vec_type>(){}
+            __host__ __device__ __forceinline__ HLS2RGB() {}
+            __host__ __device__ __forceinline__ HLS2RGB(const HLS2RGB&) {}
         };
 
         template <int bidx, int hr> struct HLS2RGB<uchar, 4, 4, bidx, hr> : unary_function<uint, uint>
@@ -1497,15 +1472,15 @@ namespace cv { namespace gpu { namespace device
             {
                 return HLS2RGBConvert<bidx, hr>(src);
             }
-            __device__ __forceinline__ HLS2RGB() : unary_function<uint, uint>(){}
-            __device__ __forceinline__ HLS2RGB(const HLS2RGB& other_) : unary_function<uint, uint>(){}
+            __host__ __device__ __forceinline__ HLS2RGB() {}
+            __host__ __device__ __forceinline__ HLS2RGB(const HLS2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_HLS2RGB_TRAITS(name, scn, dcn, bidx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::HLS2RGB<T, scn, dcn, bidx, 180> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HLS2RGB<T, scn, dcn, bidx, 180> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1513,7 +1488,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <typename T> struct name ## _full_traits \
     { \
-        typedef ::cv::gpu::device::color_detail::HLS2RGB<T, scn, dcn, bidx, 255> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HLS2RGB<T, scn, dcn, bidx, 255> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1521,7 +1496,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::HLS2RGB<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HLS2RGB<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1529,7 +1504,7 @@ namespace cv { namespace gpu { namespace device
     }; \
     template <> struct name ## _full_traits<float> \
     { \
-        typedef ::cv::gpu::device::color_detail::HLS2RGB<float, scn, dcn, bidx, 360> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::HLS2RGB<float, scn, dcn, bidx, 360> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1651,8 +1626,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2Lab() {}
-            __device__ __forceinline__ RGB2Lab(const RGB2Lab& other_) {}
+            __host__ __device__ __forceinline__ RGB2Lab() {}
+            __host__ __device__ __forceinline__ RGB2Lab(const RGB2Lab&) {}
         };
         template <int scn, int dcn, bool srgb, int blueIdx>
         struct RGB2Lab<float, scn, dcn, srgb, blueIdx>
@@ -1666,15 +1641,15 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2Lab() {}
-            __device__ __forceinline__ RGB2Lab(const RGB2Lab& other_) {}
+            __host__ __device__ __forceinline__ RGB2Lab() {}
+            __host__ __device__ __forceinline__ RGB2Lab(const RGB2Lab&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2Lab_TRAITS(name, scn, dcn, srgb, blueIdx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2Lab<T, scn, dcn, srgb, blueIdx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2Lab<T, scn, dcn, srgb, blueIdx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1764,8 +1739,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Lab2RGB() {}
-            __device__ __forceinline__ Lab2RGB(const Lab2RGB& other_) {}
+            __host__ __device__ __forceinline__ Lab2RGB() {}
+            __host__ __device__ __forceinline__ Lab2RGB(const Lab2RGB&) {}
         };
         template <int scn, int dcn, bool srgb, int blueIdx>
         struct Lab2RGB<float, scn, dcn, srgb, blueIdx>
@@ -1779,15 +1754,15 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Lab2RGB() {}
-            __device__ __forceinline__ Lab2RGB(const Lab2RGB& other_) {}
+            __host__ __device__ __forceinline__ Lab2RGB() {}
+            __host__ __device__ __forceinline__ Lab2RGB(const Lab2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_Lab2RGB_TRAITS(name, scn, dcn, srgb, blueIdx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::Lab2RGB<T, scn, dcn, srgb, blueIdx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::Lab2RGB<T, scn, dcn, srgb, blueIdx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1863,8 +1838,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2Luv() {}
-            __device__ __forceinline__ RGB2Luv(const RGB2Luv& other_) {}
+            __host__ __device__ __forceinline__ RGB2Luv() {}
+            __host__ __device__ __forceinline__ RGB2Luv(const RGB2Luv&) {}
         };
         template <int scn, int dcn, bool srgb, int blueIdx>
         struct RGB2Luv<float, scn, dcn, srgb, blueIdx>
@@ -1878,15 +1853,15 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ RGB2Luv() {}
-            __device__ __forceinline__ RGB2Luv(const RGB2Luv& other_) {}
+            __host__ __device__ __forceinline__ RGB2Luv() {}
+            __host__ __device__ __forceinline__ RGB2Luv(const RGB2Luv&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_RGB2Luv_TRAITS(name, scn, dcn, srgb, blueIdx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::RGB2Luv<T, scn, dcn, srgb, blueIdx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::RGB2Luv<T, scn, dcn, srgb, blueIdx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1964,8 +1939,8 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Luv2RGB() {}
-            __device__ __forceinline__ Luv2RGB(const Luv2RGB& other_) {}
+            __host__ __device__ __forceinline__ Luv2RGB() {}
+            __host__ __device__ __forceinline__ Luv2RGB(const Luv2RGB&) {}
         };
         template <int scn, int dcn, bool srgb, int blueIdx>
         struct Luv2RGB<float, scn, dcn, srgb, blueIdx>
@@ -1979,15 +1954,15 @@ namespace cv { namespace gpu { namespace device
 
                 return dst;
             }
-            __device__ __forceinline__ Luv2RGB() {}
-            __device__ __forceinline__ Luv2RGB(const Luv2RGB& other_) {}
+            __host__ __device__ __forceinline__ Luv2RGB() {}
+            __host__ __device__ __forceinline__ Luv2RGB(const Luv2RGB&) {}
         };
     }
 
 #define OPENCV_GPU_IMPLEMENT_Luv2RGB_TRAITS(name, scn, dcn, srgb, blueIdx) \
     template <typename T> struct name ## _traits \
     { \
-        typedef ::cv::gpu::device::color_detail::Luv2RGB<T, scn, dcn, srgb, blueIdx> functor_type; \
+        typedef ::cv::gpu::cudev::color_detail::Luv2RGB<T, scn, dcn, srgb, blueIdx> functor_type; \
         static __host__ __device__ __forceinline__ functor_type create_functor() \
         { \
             return functor_type(); \
@@ -1996,6 +1971,6 @@ namespace cv { namespace gpu { namespace device
 
     #undef CV_DESCALE
 
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_COLOR_DETAIL_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/reduce.hpp b/modules/core/include/opencv2/core/cuda/detail/reduce.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/detail/reduce.hpp
rename to modules/core/include/opencv2/core/cuda/detail/reduce.hpp
index 091a160e31..eba9b41a78 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/reduce.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/reduce.hpp
@@ -47,7 +47,7 @@
 #include "../warp.hpp"
 #include "../warp_shuffle.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace reduce_detail
     {
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/reduce_key_val.hpp b/modules/core/include/opencv2/core/cuda/detail/reduce_key_val.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/detail/reduce_key_val.hpp
rename to modules/core/include/opencv2/core/cuda/detail/reduce_key_val.hpp
index a84e0c2fd0..1049e6714c 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/reduce_key_val.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/reduce_key_val.hpp
@@ -47,7 +47,7 @@
 #include "../warp.hpp"
 #include "../warp_shuffle.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace reduce_key_val_detail
     {
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/transform_detail.hpp b/modules/core/include/opencv2/core/cuda/detail/transform_detail.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/detail/transform_detail.hpp
rename to modules/core/include/opencv2/core/cuda/detail/transform_detail.hpp
index 10da5938c5..2ac309b0c6 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/transform_detail.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/transform_detail.hpp
@@ -47,7 +47,7 @@
 #include "../vec_traits.hpp"
 #include "../functional.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace transform_detail
     {
@@ -345,7 +345,7 @@ namespace cv { namespace gpu { namespace device
             {
                 typedef TransformFunctorTraits<UnOp> ft;
 
-                StaticAssert<ft::smart_shift != 1>::check();
+                CV_StaticAssert(ft::smart_shift != 1, "");
 
                 if (!isAligned(src.data, ft::smart_shift * sizeof(T)) || !isAligned(src.step, ft::smart_shift * sizeof(T)) ||
                     !isAligned(dst.data, ft::smart_shift * sizeof(D)) || !isAligned(dst.step, ft::smart_shift * sizeof(D)))
@@ -369,7 +369,7 @@ namespace cv { namespace gpu { namespace device
             {
                 typedef TransformFunctorTraits<BinOp> ft;
 
-                StaticAssert<ft::smart_shift != 1>::check();
+                CV_StaticAssert(ft::smart_shift != 1, "");
 
                 if (!isAligned(src1.data, ft::smart_shift * sizeof(T1)) || !isAligned(src1.step, ft::smart_shift * sizeof(T1)) ||
                     !isAligned(src2.data, ft::smart_shift * sizeof(T2)) || !isAligned(src2.step, ft::smart_shift * sizeof(T2)) ||
@@ -390,6 +390,6 @@ namespace cv { namespace gpu { namespace device
             }
         };
     } // namespace transform_detail
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_TRANSFORM_DETAIL_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/type_traits_detail.hpp b/modules/core/include/opencv2/core/cuda/detail/type_traits_detail.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/detail/type_traits_detail.hpp
rename to modules/core/include/opencv2/core/cuda/detail/type_traits_detail.hpp
index 97ff00d8f6..4292d88003 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/type_traits_detail.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/type_traits_detail.hpp
@@ -46,7 +46,7 @@
 #include "../common.hpp"
 #include "../vec_traits.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace type_traits_detail
     {
@@ -182,6 +182,6 @@ namespace cv { namespace gpu { namespace device
             enum { value = 1 };
         };
     } // namespace type_traits_detail
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_TYPE_TRAITS_DETAIL_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/detail/vec_distance_detail.hpp b/modules/core/include/opencv2/core/cuda/detail/vec_distance_detail.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/detail/vec_distance_detail.hpp
rename to modules/core/include/opencv2/core/cuda/detail/vec_distance_detail.hpp
index 78ab5565cb..a2a31a703c 100644
--- a/modules/gpu/include/opencv2/gpu/device/detail/vec_distance_detail.hpp
+++ b/modules/core/include/opencv2/core/cuda/detail/vec_distance_detail.hpp
@@ -45,7 +45,7 @@
 
 #include "../datamov_utils.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace vec_distance_detail
     {
@@ -112,6 +112,6 @@ namespace cv { namespace gpu { namespace device
             }
         };
     } // namespace vec_distance_detail
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_VEC_DISTANCE_DETAIL_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/dynamic_smem.hpp b/modules/core/include/opencv2/core/cuda/dynamic_smem.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/dynamic_smem.hpp
rename to modules/core/include/opencv2/core/cuda/dynamic_smem.hpp
index cf431d9524..aa20e53b82 100644
--- a/modules/gpu/include/opencv2/gpu/device/dynamic_smem.hpp
+++ b/modules/core/include/opencv2/core/cuda/dynamic_smem.hpp
@@ -43,7 +43,7 @@
 #ifndef __OPENCV_GPU_DYNAMIC_SMEM_HPP__
 #define __OPENCV_GPU_DYNAMIC_SMEM_HPP__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template<class T> struct DynamicSharedMem
     {
diff --git a/modules/core/include/opencv2/core/cuda/emulation.hpp b/modules/core/include/opencv2/core/cuda/emulation.hpp
new file mode 100644
index 0000000000..b484f2378e
--- /dev/null
+++ b/modules/core/include/opencv2/core/cuda/emulation.hpp
@@ -0,0 +1,261 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_GPU_EMULATION_HPP_
+#define OPENCV_GPU_EMULATION_HPP_
+
+#include "common.hpp"
+#include "warp_reduce.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    struct Emulation
+    {
+
+        static __device__ __forceinline__ int syncthreadsOr(int pred)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 200)
+                // just campilation stab
+                return 0;
+#else
+                return __syncthreads_or(pred);
+#endif
+        }
+
+        template<int CTA_SIZE>
+        static __forceinline__ __device__ int Ballot(int predicate)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ >= 200)
+            return __ballot(predicate);
+#else
+            __shared__ volatile int cta_buffer[CTA_SIZE];
+
+            int tid = threadIdx.x;
+            cta_buffer[tid] = predicate ? (1 << (tid & 31)) : 0;
+            return warp_reduce(cta_buffer);
+#endif
+        }
+
+        struct smem
+        {
+            enum { TAG_MASK = (1U << ( (sizeof(unsigned int) << 3) - 5U)) - 1U };
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicInc(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + 1);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - 1;
+#else
+                return ::atomicInc(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicAdd(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + val);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - val;
+#else
+                return ::atomicAdd(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicMin(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count = ::min(*address, val);
+                do
+                {
+                    *address = count;
+                } while (*address > count);
+
+                return count;
+#else
+                return ::atomicMin(address, val);
+#endif
+            }
+        }; // struct cmem
+
+        struct glob
+        {
+            static __device__ __forceinline__ int atomicAdd(int* address, int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ unsigned int atomicAdd(unsigned int* address, unsigned int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ float atomicAdd(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 200
+                return ::atomicAdd(address, val);
+            #else
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(val + __int_as_float(assumed)));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #endif
+            }
+            static __device__ __forceinline__ double atomicAdd(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(val + __longlong_as_double(assumed)));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                (void) address;
+                (void) val;
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMin(int* address, int val)
+            {
+                return ::atomicMin(address, val);
+            }
+            static __device__ __forceinline__ float atomicMin(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fminf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                (void) address;
+                (void) val;
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMin(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmin(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                (void) address;
+                (void) val;
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMax(int* address, int val)
+            {
+                return ::atomicMax(address, val);
+            }
+            static __device__ __forceinline__ float atomicMax(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fmaxf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                (void) address;
+                (void) val;
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMax(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmax(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                (void) address;
+                (void) val;
+                return 0.0;
+            #endif
+            }
+        };
+    }; //struct Emulation
+}}} // namespace cv { namespace gpu { namespace cudev
+
+#endif /* OPENCV_GPU_EMULATION_HPP_ */
diff --git a/modules/gpu/include/opencv2/gpu/device/filters.hpp b/modules/core/include/opencv2/core/cuda/filters.hpp
similarity index 97%
rename from modules/gpu/include/opencv2/gpu/device/filters.hpp
rename to modules/core/include/opencv2/core/cuda/filters.hpp
index d193969a79..f35f662e8b 100644
--- a/modules/gpu/include/opencv2/gpu/device/filters.hpp
+++ b/modules/core/include/opencv2/core/cuda/filters.hpp
@@ -48,7 +48,7 @@
 #include "vec_math.hpp"
 #include "type_traits.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <typename Ptr2D> struct PointFilter
     {
@@ -67,7 +67,7 @@ namespace cv { namespace gpu { namespace device
             return src(__float2int_rz(y), __float2int_rz(x));
         }
 
-        const Ptr2D src;
+        Ptr2D src;
     };
 
     template <typename Ptr2D> struct LinearFilter
@@ -107,7 +107,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<elem_type>(out);
         }
 
-        const Ptr2D src;
+        Ptr2D src;
     };
 
     template <typename Ptr2D> struct CubicFilter
@@ -166,7 +166,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<elem_type>(res);
         }
 
-        const Ptr2D src;
+        Ptr2D src;
     };
     // for integer scaling
     template <typename Ptr2D> struct IntegerAreaFilter
@@ -203,7 +203,7 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<elem_type>(out);
         }
 
-        const Ptr2D src;
+        Ptr2D src;
         float scale_x, scale_y ,scale;
     };
 
@@ -269,10 +269,10 @@ namespace cv { namespace gpu { namespace device
             return saturate_cast<elem_type>(out);
         }
 
-        const Ptr2D src;
+        Ptr2D src;
         float scale_x, scale_y;
         int width, haight;
     };
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_FILTERS_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/funcattrib.hpp b/modules/core/include/opencv2/core/cuda/funcattrib.hpp
similarity index 95%
rename from modules/gpu/include/opencv2/gpu/device/funcattrib.hpp
rename to modules/core/include/opencv2/core/cuda/funcattrib.hpp
index 05df4d10b4..46ef81926c 100644
--- a/modules/gpu/include/opencv2/gpu/device/funcattrib.hpp
+++ b/modules/core/include/opencv2/core/cuda/funcattrib.hpp
@@ -40,13 +40,12 @@
 //
 //M*/
 
-
 #ifndef __OPENCV_GPU_DEVICE_FUNCATTRIB_HPP_
 #define __OPENCV_GPU_DEVICE_FUNCATTRIB_HPP_
 
 #include <cstdio>
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template<class Func>
     void printFuncAttrib(Func& func)
@@ -67,6 +66,6 @@ namespace cv { namespace gpu { namespace device
         printf("\n");
         fflush(stdout);
     }
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-#endif  /* __OPENCV_GPU_DEVICE_FUNCATTRIB_HPP_ */
\ No newline at end of file
+#endif  /* __OPENCV_GPU_DEVICE_FUNCATTRIB_HPP_ */
diff --git a/modules/gpu/include/opencv2/gpu/device/functional.hpp b/modules/core/include/opencv2/core/cuda/functional.hpp
similarity index 73%
rename from modules/gpu/include/opencv2/gpu/device/functional.hpp
rename to modules/core/include/opencv2/core/cuda/functional.hpp
index 6064e8e99c..d6c019236a 100644
--- a/modules/gpu/include/opencv2/gpu/device/functional.hpp
+++ b/modules/core/include/opencv2/core/cuda/functional.hpp
@@ -49,7 +49,7 @@
 #include "type_traits.hpp"
 #include "device_functions.h"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     // Function Objects
     template<typename Argument, typename Result> struct unary_function : public std::unary_function<Argument, Result> {};
@@ -63,8 +63,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a + b;
         }
-        __device__ __forceinline__ plus(const plus& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ plus():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ plus() {}
+        __host__ __device__ __forceinline__ plus(const plus&) {}
     };
 
     template <typename T> struct minus : binary_function<T, T, T>
@@ -74,8 +74,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a - b;
         }
-        __device__ __forceinline__ minus(const minus& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ minus():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ minus() {}
+        __host__ __device__ __forceinline__ minus(const minus&) {}
     };
 
     template <typename T> struct multiplies : binary_function<T, T, T>
@@ -85,8 +85,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a * b;
         }
-        __device__ __forceinline__ multiplies(const multiplies& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ multiplies():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ multiplies() {}
+        __host__ __device__ __forceinline__ multiplies(const multiplies&) {}
     };
 
     template <typename T> struct divides : binary_function<T, T, T>
@@ -96,8 +96,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a / b;
         }
-        __device__ __forceinline__ divides(const divides& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ divides():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ divides() {}
+        __host__ __device__ __forceinline__ divides(const divides&) {}
     };
 
     template <typename T> struct modulus : binary_function<T, T, T>
@@ -107,8 +107,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a % b;
         }
-        __device__ __forceinline__ modulus(const modulus& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ modulus():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ modulus() {}
+        __host__ __device__ __forceinline__ modulus(const modulus&) {}
     };
 
     template <typename T> struct negate : unary_function<T, T>
@@ -117,8 +117,8 @@ namespace cv { namespace gpu { namespace device
         {
             return -a;
         }
-        __device__ __forceinline__ negate(const negate& other):unary_function<T,T>(){}
-        __device__ __forceinline__ negate():unary_function<T,T>(){}
+        __host__ __device__ __forceinline__ negate() {}
+        __host__ __device__ __forceinline__ negate(const negate&) {}
     };
 
     // Comparison Operations
@@ -129,8 +129,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a == b;
         }
-        __device__ __forceinline__ equal_to(const equal_to& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ equal_to():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ equal_to() {}
+        __host__ __device__ __forceinline__ equal_to(const equal_to&) {}
     };
 
     template <typename T> struct not_equal_to : binary_function<T, T, bool>
@@ -140,8 +140,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a != b;
         }
-        __device__ __forceinline__ not_equal_to(const not_equal_to& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ not_equal_to():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ not_equal_to() {}
+        __host__ __device__ __forceinline__ not_equal_to(const not_equal_to&) {}
     };
 
     template <typename T> struct greater : binary_function<T, T, bool>
@@ -151,8 +151,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a > b;
         }
-        __device__ __forceinline__ greater(const greater& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ greater():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ greater() {}
+        __host__ __device__ __forceinline__ greater(const greater&) {}
     };
 
     template <typename T> struct less : binary_function<T, T, bool>
@@ -162,8 +162,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a < b;
         }
-        __device__ __forceinline__ less(const less& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ less():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ less() {}
+        __host__ __device__ __forceinline__ less(const less&) {}
     };
 
     template <typename T> struct greater_equal : binary_function<T, T, bool>
@@ -173,8 +173,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a >= b;
         }
-        __device__ __forceinline__ greater_equal(const greater_equal& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ greater_equal():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ greater_equal() {}
+        __host__ __device__ __forceinline__ greater_equal(const greater_equal&) {}
     };
 
     template <typename T> struct less_equal : binary_function<T, T, bool>
@@ -184,8 +184,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a <= b;
         }
-        __device__ __forceinline__ less_equal(const less_equal& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ less_equal():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ less_equal() {}
+        __host__ __device__ __forceinline__ less_equal(const less_equal&) {}
     };
 
     // Logical Operations
@@ -196,8 +196,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a && b;
         }
-        __device__ __forceinline__ logical_and(const logical_and& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ logical_and():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ logical_and() {}
+        __host__ __device__ __forceinline__ logical_and(const logical_and&) {}
     };
 
     template <typename T> struct logical_or : binary_function<T, T, bool>
@@ -207,8 +207,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a || b;
         }
-        __device__ __forceinline__ logical_or(const logical_or& other):binary_function<T,T,bool>(){}
-        __device__ __forceinline__ logical_or():binary_function<T,T,bool>(){}
+        __host__ __device__ __forceinline__ logical_or() {}
+        __host__ __device__ __forceinline__ logical_or(const logical_or&) {}
     };
 
     template <typename T> struct logical_not : unary_function<T, bool>
@@ -217,8 +217,8 @@ namespace cv { namespace gpu { namespace device
         {
             return !a;
         }
-        __device__ __forceinline__ logical_not(const logical_not& other):unary_function<T,bool>(){}
-        __device__ __forceinline__ logical_not():unary_function<T,bool>(){}
+        __host__ __device__ __forceinline__ logical_not() {}
+        __host__ __device__ __forceinline__ logical_not(const logical_not&) {}
     };
 
     // Bitwise Operations
@@ -229,8 +229,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a & b;
         }
-        __device__ __forceinline__ bit_and(const bit_and& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ bit_and():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ bit_and() {}
+        __host__ __device__ __forceinline__ bit_and(const bit_and&) {}
     };
 
     template <typename T> struct bit_or : binary_function<T, T, T>
@@ -240,8 +240,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a | b;
         }
-        __device__ __forceinline__ bit_or(const bit_or& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ bit_or():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ bit_or() {}
+        __host__ __device__ __forceinline__ bit_or(const bit_or&) {}
     };
 
     template <typename T> struct bit_xor : binary_function<T, T, T>
@@ -251,8 +251,8 @@ namespace cv { namespace gpu { namespace device
         {
             return a ^ b;
         }
-        __device__ __forceinline__ bit_xor(const bit_xor& other):binary_function<T,T,T>(){}
-        __device__ __forceinline__ bit_xor():binary_function<T,T,T>(){}
+        __host__ __device__ __forceinline__ bit_xor() {}
+        __host__ __device__ __forceinline__ bit_xor(const bit_xor&) {}
     };
 
     template <typename T> struct bit_not : unary_function<T, T>
@@ -261,8 +261,8 @@ namespace cv { namespace gpu { namespace device
         {
             return ~v;
         }
-        __device__ __forceinline__ bit_not(const bit_not& other):unary_function<T,T>(){}
-        __device__ __forceinline__ bit_not():unary_function<T,T>(){}
+        __host__ __device__ __forceinline__ bit_not() {}
+        __host__ __device__ __forceinline__ bit_not(const bit_not&) {}
     };
 
     // Generalized Identity Operations
@@ -272,8 +272,8 @@ namespace cv { namespace gpu { namespace device
         {
             return x;
         }
-        __device__ __forceinline__ identity(const identity& other):unary_function<T,T>(){}
-        __device__ __forceinline__ identity():unary_function<T,T>(){}
+        __host__ __device__ __forceinline__ identity() {}
+        __host__ __device__ __forceinline__ identity(const identity&) {}
     };
 
     template <typename T1, typename T2> struct project1st : binary_function<T1, T2, T1>
@@ -282,8 +282,8 @@ namespace cv { namespace gpu { namespace device
         {
             return lhs;
         }
-        __device__ __forceinline__ project1st(const project1st& other):binary_function<T1,T2,T1>(){}
-        __device__ __forceinline__ project1st():binary_function<T1,T2,T1>(){}
+        __host__ __device__ __forceinline__ project1st() {}
+        __host__ __device__ __forceinline__ project1st(const project1st&) {}
     };
 
     template <typename T1, typename T2> struct project2nd : binary_function<T1, T2, T2>
@@ -292,8 +292,8 @@ namespace cv { namespace gpu { namespace device
         {
             return rhs;
         }
-        __device__ __forceinline__ project2nd(const project2nd& other):binary_function<T1,T2,T2>(){}
-        __device__ __forceinline__ project2nd():binary_function<T1,T2,T2>(){}
+        __host__ __device__ __forceinline__ project2nd() {}
+        __host__ __device__ __forceinline__ project2nd(const project2nd&) {}
     };
 
     // Min/Max Operations
@@ -302,8 +302,8 @@ namespace cv { namespace gpu { namespace device
     template <> struct name<type> : binary_function<type, type, type> \
     { \
         __device__ __forceinline__ type operator()(type lhs, type rhs) const {return op(lhs, rhs);} \
-        __device__ __forceinline__ name() {}\
-        __device__ __forceinline__ name(const name&) {}\
+        __host__ __device__ __forceinline__ name() {}\
+        __host__ __device__ __forceinline__ name(const name&) {}\
     };
 
     template <typename T> struct maximum : binary_function<T, T, T>
@@ -312,8 +312,8 @@ namespace cv { namespace gpu { namespace device
         {
             return max(lhs, rhs);
         }
-        __device__ __forceinline__ maximum() {}
-        __device__ __forceinline__ maximum(const maximum&) {}
+        __host__ __device__ __forceinline__ maximum() {}
+        __host__ __device__ __forceinline__ maximum(const maximum&) {}
     };
 
     OPENCV_GPU_IMPLEMENT_MINMAX(maximum, uchar, ::max)
@@ -332,8 +332,8 @@ namespace cv { namespace gpu { namespace device
         {
             return min(lhs, rhs);
         }
-        __device__ __forceinline__ minimum() {}
-        __device__ __forceinline__ minimum(const minimum&) {}
+        __host__ __device__ __forceinline__ minimum() {}
+        __host__ __device__ __forceinline__ minimum(const minimum&) {}
     };
 
     OPENCV_GPU_IMPLEMENT_MINMAX(minimum, uchar, ::min)
@@ -349,7 +349,6 @@ namespace cv { namespace gpu { namespace device
 #undef OPENCV_GPU_IMPLEMENT_MINMAX
 
     // Math functions
-///bound=========================================
 
     template <typename T> struct abs_func : unary_function<T, T>
     {
@@ -358,8 +357,8 @@ namespace cv { namespace gpu { namespace device
             return abs(x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<unsigned char> : unary_function<unsigned char, unsigned char>
     {
@@ -368,8 +367,8 @@ namespace cv { namespace gpu { namespace device
             return x;
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<signed char> : unary_function<signed char, signed char>
     {
@@ -378,8 +377,8 @@ namespace cv { namespace gpu { namespace device
             return ::abs((int)x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<char> : unary_function<char, char>
     {
@@ -388,8 +387,8 @@ namespace cv { namespace gpu { namespace device
             return ::abs((int)x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<unsigned short> : unary_function<unsigned short, unsigned short>
     {
@@ -398,8 +397,8 @@ namespace cv { namespace gpu { namespace device
             return x;
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<short> : unary_function<short, short>
     {
@@ -408,8 +407,8 @@ namespace cv { namespace gpu { namespace device
             return ::abs((int)x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<unsigned int> : unary_function<unsigned int, unsigned int>
     {
@@ -418,8 +417,8 @@ namespace cv { namespace gpu { namespace device
             return x;
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<int> : unary_function<int, int>
     {
@@ -428,8 +427,8 @@ namespace cv { namespace gpu { namespace device
             return ::abs(x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<float> : unary_function<float, float>
     {
@@ -438,8 +437,8 @@ namespace cv { namespace gpu { namespace device
             return ::fabsf(x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
     template <> struct abs_func<double> : unary_function<double, double>
     {
@@ -448,8 +447,8 @@ namespace cv { namespace gpu { namespace device
             return ::fabs(x);
         }
 
-        __device__ __forceinline__ abs_func() {}
-        __device__ __forceinline__ abs_func(const abs_func&) {}
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
     };
 
 #define OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(name, func) \
@@ -459,8 +458,8 @@ namespace cv { namespace gpu { namespace device
         { \
             return func ## f(v); \
         } \
-        __device__ __forceinline__ name ## _func() {} \
-        __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
     }; \
     template <> struct name ## _func<double> : unary_function<double, double> \
     { \
@@ -468,8 +467,8 @@ namespace cv { namespace gpu { namespace device
         { \
             return func(v); \
         } \
-        __device__ __forceinline__ name ## _func() {} \
-        __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
     };
 
 #define OPENCV_GPU_IMPLEMENT_BIN_FUNCTOR(name, func) \
@@ -479,6 +478,8 @@ namespace cv { namespace gpu { namespace device
         { \
             return func ## f(v1, v2); \
         } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
     }; \
     template <> struct name ## _func<double> : binary_function<double, double, double> \
     { \
@@ -486,6 +487,8 @@ namespace cv { namespace gpu { namespace device
         { \
             return func(v1, v2); \
         } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
     };
 
     OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(sqrt, ::sqrt)
@@ -522,8 +525,8 @@ namespace cv { namespace gpu { namespace device
         {
             return src1 * src1 + src2 * src2;
         }
-        __device__ __forceinline__ hypot_sqr_func(const hypot_sqr_func& other) : binary_function<T, T, float>(){}
-        __device__ __forceinline__ hypot_sqr_func() : binary_function<T, T, float>(){}
+        __host__ __device__ __forceinline__ hypot_sqr_func() {}
+        __host__ __device__ __forceinline__ hypot_sqr_func(const hypot_sqr_func&) {}
     };
 
     // Saturate Cast Functor
@@ -533,8 +536,8 @@ namespace cv { namespace gpu { namespace device
         {
             return saturate_cast<D>(v);
         }
-        __device__ __forceinline__ saturate_cast_func(const saturate_cast_func& other):unary_function<T, D>(){}
-        __device__ __forceinline__ saturate_cast_func():unary_function<T, D>(){}
+        __host__ __device__ __forceinline__ saturate_cast_func() {}
+        __host__ __device__ __forceinline__ saturate_cast_func(const saturate_cast_func&) {}
     };
 
     // Threshold Functors
@@ -547,13 +550,12 @@ namespace cv { namespace gpu { namespace device
             return (src > thresh) * maxVal;
         }
 
-        __device__ __forceinline__ thresh_binary_func(const thresh_binary_func& other)
-            : unary_function<T, T>(), thresh(other.thresh), maxVal(other.maxVal){}
+        __host__ __device__ __forceinline__ thresh_binary_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_func(const thresh_binary_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
 
-        __device__ __forceinline__ thresh_binary_func():unary_function<T, T>(){}
-
-        const T thresh;
-        const T maxVal;
+        T thresh;
+        T maxVal;
     };
 
     template <typename T> struct thresh_binary_inv_func : unary_function<T, T>
@@ -565,13 +567,12 @@ namespace cv { namespace gpu { namespace device
             return (src <= thresh) * maxVal;
         }
 
-        __device__ __forceinline__ thresh_binary_inv_func(const thresh_binary_inv_func& other)
-            : unary_function<T, T>(), thresh(other.thresh), maxVal(other.maxVal){}
+        __host__ __device__ __forceinline__ thresh_binary_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_inv_func(const thresh_binary_inv_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
 
-        __device__ __forceinline__ thresh_binary_inv_func():unary_function<T, T>(){}
-
-        const T thresh;
-        const T maxVal;
+        T thresh;
+        T maxVal;
     };
 
     template <typename T> struct thresh_trunc_func : unary_function<T, T>
@@ -583,12 +584,11 @@ namespace cv { namespace gpu { namespace device
             return minimum<T>()(src, thresh);
         }
 
-        __device__ __forceinline__ thresh_trunc_func(const thresh_trunc_func& other)
-            : unary_function<T, T>(), thresh(other.thresh){}
+        __host__ __device__ __forceinline__ thresh_trunc_func() {}
+        __host__ __device__ __forceinline__ thresh_trunc_func(const thresh_trunc_func& other)
+            : thresh(other.thresh) {}
 
-        __device__ __forceinline__ thresh_trunc_func():unary_function<T, T>(){}
-
-        const T thresh;
+        T thresh;
     };
 
     template <typename T> struct thresh_to_zero_func : unary_function<T, T>
@@ -599,12 +599,12 @@ namespace cv { namespace gpu { namespace device
         {
             return (src > thresh) * src;
         }
-        __device__ __forceinline__ thresh_to_zero_func(const thresh_to_zero_func& other)
-            : unary_function<T, T>(), thresh(other.thresh){}
 
-        __device__ __forceinline__ thresh_to_zero_func():unary_function<T, T>(){}
+        __host__ __device__ __forceinline__ thresh_to_zero_func() {}
+       __host__  __device__ __forceinline__ thresh_to_zero_func(const thresh_to_zero_func& other)
+            : thresh(other.thresh) {}
 
-        const T thresh;
+        T thresh;
     };
 
     template <typename T> struct thresh_to_zero_inv_func : unary_function<T, T>
@@ -615,14 +615,14 @@ namespace cv { namespace gpu { namespace device
         {
             return (src <= thresh) * src;
         }
-        __device__ __forceinline__ thresh_to_zero_inv_func(const thresh_to_zero_inv_func& other)
-            : unary_function<T, T>(), thresh(other.thresh){}
 
-        __device__ __forceinline__ thresh_to_zero_inv_func():unary_function<T, T>(){}
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func(const thresh_to_zero_inv_func& other)
+            : thresh(other.thresh) {}
 
-        const T thresh;
+        T thresh;
     };
-//bound!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ============>
+
     // Function Object Adaptors
     template <typename Predicate> struct unary_negate : unary_function<typename Predicate::argument_type, bool>
     {
@@ -633,10 +633,10 @@ namespace cv { namespace gpu { namespace device
           return !pred(x);
       }
 
-        __device__ __forceinline__ unary_negate(const unary_negate& other) : unary_function<typename Predicate::argument_type, bool>(){}
-        __device__ __forceinline__ unary_negate() : unary_function<typename Predicate::argument_type, bool>(){}
+      __host__ __device__ __forceinline__ unary_negate() {}
+      __host__ __device__ __forceinline__ unary_negate(const unary_negate& other) : pred(other.pred) {}
 
-      const Predicate pred;
+      Predicate pred;
     };
 
     template <typename Predicate> __host__ __device__ __forceinline__ unary_negate<Predicate> not1(const Predicate& pred)
@@ -653,13 +653,11 @@ namespace cv { namespace gpu { namespace device
         {
             return !pred(x,y);
         }
-        __device__ __forceinline__ binary_negate(const binary_negate& other)
-        : binary_function<typename Predicate::first_argument_type, typename Predicate::second_argument_type, bool>(){}
 
-        __device__ __forceinline__ binary_negate() :
-        binary_function<typename Predicate::first_argument_type, typename Predicate::second_argument_type, bool>(){}
+        __host__ __device__ __forceinline__ binary_negate() {}
+        __host__ __device__ __forceinline__ binary_negate(const binary_negate& other) : pred(other.pred) {}
 
-        const Predicate pred;
+        Predicate pred;
     };
 
     template <typename BinaryPredicate> __host__ __device__ __forceinline__ binary_negate<BinaryPredicate> not2(const BinaryPredicate& pred)
@@ -676,11 +674,11 @@ namespace cv { namespace gpu { namespace device
             return op(arg1, a);
         }
 
-        __device__ __forceinline__ binder1st(const binder1st& other) :
-        unary_function<typename Op::second_argument_type, typename Op::result_type>(){}
+        __host__ __device__ __forceinline__ binder1st() {}
+        __host__ __device__ __forceinline__ binder1st(const binder1st& other) : op(other.op), arg1(other.arg1) {}
 
-        const Op op;
-        const typename Op::first_argument_type arg1;
+        Op op;
+        typename Op::first_argument_type arg1;
     };
 
     template <typename Op, typename T> __host__ __device__ __forceinline__ binder1st<Op> bind1st(const Op& op, const T& x)
@@ -697,11 +695,11 @@ namespace cv { namespace gpu { namespace device
             return op(a, arg2);
         }
 
-         __device__ __forceinline__ binder2nd(const binder2nd& other) :
-        unary_function<typename Op::first_argument_type, typename Op::result_type>(), op(other.op), arg2(other.arg2){}
+        __host__ __device__ __forceinline__ binder2nd() {}
+        __host__ __device__ __forceinline__ binder2nd(const binder2nd& other) : op(other.op), arg2(other.arg2) {}
 
-        const Op op;
-        const typename Op::second_argument_type arg2;
+        Op op;
+        typename Op::second_argument_type arg2;
     };
 
     template <typename Op, typename T> __host__ __device__ __forceinline__ binder2nd<Op> bind2nd(const Op& op, const T& x)
@@ -786,6 +784,6 @@ namespace cv { namespace gpu { namespace device
 
 #define OPENCV_GPU_TRANSFORM_FUNCTOR_TRAITS(type) \
     template <> struct TransformFunctorTraits< type > : DefaultTransformFunctorTraits< type >
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_FUNCTIONAL_HPP__
diff --git a/modules/core/include/opencv2/core/cuda/limits.hpp b/modules/core/include/opencv2/core/cuda/limits.hpp
new file mode 100644
index 0000000000..0439de795c
--- /dev/null
+++ b/modules/core/include/opencv2/core/cuda/limits.hpp
@@ -0,0 +1,122 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPU_LIMITS_GPU_HPP__
+#define __OPENCV_GPU_LIMITS_GPU_HPP__
+
+#include <limits.h>
+#include <float.h>
+#include "common.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+
+template <class T> struct numeric_limits;
+
+template <> struct numeric_limits<bool>
+{
+    __device__ __forceinline__ static bool min() { return false; }
+    __device__ __forceinline__ static bool max() { return true;  }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<signed char>
+{
+    __device__ __forceinline__ static signed char min() { return SCHAR_MIN; }
+    __device__ __forceinline__ static signed char max() { return SCHAR_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned char>
+{
+    __device__ __forceinline__ static unsigned char min() { return 0; }
+    __device__ __forceinline__ static unsigned char max() { return UCHAR_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<short>
+{
+    __device__ __forceinline__ static short min() { return SHRT_MIN; }
+    __device__ __forceinline__ static short max() { return SHRT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned short>
+{
+    __device__ __forceinline__ static unsigned short min() { return 0; }
+    __device__ __forceinline__ static unsigned short max() { return USHRT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<int>
+{
+    __device__ __forceinline__ static int min() { return INT_MIN; }
+    __device__ __forceinline__ static int max() { return INT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned int>
+{
+    __device__ __forceinline__ static unsigned int min() { return 0; }
+    __device__ __forceinline__ static unsigned int max() { return UINT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<float>
+{
+    __device__ __forceinline__ static float min() { return FLT_MIN; }
+    __device__ __forceinline__ static float max() { return FLT_MAX; }
+    __device__ __forceinline__ static float epsilon() { return FLT_EPSILON; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<double>
+{
+    __device__ __forceinline__ static double min() { return DBL_MIN; }
+    __device__ __forceinline__ static double max() { return DBL_MAX; }
+    __device__ __forceinline__ static double epsilon() { return DBL_EPSILON; }
+    static const bool is_signed = true;
+};
+
+}}} // namespace cv { namespace gpu { namespace cudev {
+
+#endif // __OPENCV_GPU_LIMITS_GPU_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/reduce.hpp b/modules/core/include/opencv2/core/cuda/reduce.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/reduce.hpp
rename to modules/core/include/opencv2/core/cuda/reduce.hpp
index 2161b06495..722e2bbeb6 100644
--- a/modules/gpu/include/opencv2/gpu/device/reduce.hpp
+++ b/modules/core/include/opencv2/core/cuda/reduce.hpp
@@ -47,7 +47,7 @@
 #include "detail/reduce.hpp"
 #include "detail/reduce_key_val.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <int N, typename T, class Op>
     __device__ __forceinline__ void reduce(volatile T* smem, T& val, unsigned int tid, const Op& op)
diff --git a/modules/gpu/include/opencv2/gpu/device/saturate_cast.hpp b/modules/core/include/opencv2/core/cuda/saturate_cast.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/saturate_cast.hpp
rename to modules/core/include/opencv2/core/cuda/saturate_cast.hpp
index 7a2799fa37..b30f5e7ce0 100644
--- a/modules/gpu/include/opencv2/gpu/device/saturate_cast.hpp
+++ b/modules/core/include/opencv2/core/cuda/saturate_cast.hpp
@@ -45,7 +45,7 @@
 
 #include "common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(uchar v) { return _Tp(v); }
     template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(schar v) { return _Tp(v); }
diff --git a/modules/gpu/include/opencv2/gpu/device/scan.hpp b/modules/core/include/opencv2/core/cuda/scan.hpp
similarity index 69%
rename from modules/gpu/include/opencv2/gpu/device/scan.hpp
rename to modules/core/include/opencv2/core/cuda/scan.hpp
index f6dc6937fb..ecde123bb3 100644
--- a/modules/gpu/include/opencv2/gpu/device/scan.hpp
+++ b/modules/core/include/opencv2/core/cuda/scan.hpp
@@ -43,9 +43,12 @@
 #ifndef __OPENCV_GPU_SCAN_HPP__
 #define __OPENCV_GPU_SCAN_HPP__
 
-#include "common.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/warp.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     enum ScanKind { EXCLUSIVE = 0,  INCLUSIVE = 1 };
 
@@ -166,6 +169,82 @@ namespace cv { namespace gpu { namespace device
         static const int warp_log  = 5;
         static const int warp_mask = 31;
     };
+
+    template <typename T>
+    __device__ T warpScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+    #if __CUDA_ARCH__ >= 300
+        const unsigned int laneId = cv::gpu::cudev::Warp::laneId();
+
+        // scan on shuffl functions
+        #pragma unroll
+        for (int i = 1; i <= (OPENCV_GPU_WARP_SIZE / 2); i *= 2)
+        {
+            const T n = cv::gpu::cudev::shfl_up(idata, i);
+            if (laneId >= i)
+                  idata += n;
+        }
+
+        return idata;
+    #else
+        unsigned int pos = 2 * tid - (tid & (OPENCV_GPU_WARP_SIZE - 1));
+        s_Data[pos] = 0;
+        pos += OPENCV_GPU_WARP_SIZE;
+        s_Data[pos] = idata;
+
+        s_Data[pos] += s_Data[pos - 1];
+        s_Data[pos] += s_Data[pos - 2];
+        s_Data[pos] += s_Data[pos - 4];
+        s_Data[pos] += s_Data[pos - 8];
+        s_Data[pos] += s_Data[pos - 16];
+
+        return s_Data[pos];
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T warpScanExclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        return warpScanInclusive(idata, s_Data, tid) - idata;
+    }
+
+    template <int tiNumScanThreads, typename T>
+    __device__ T blockScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        if (tiNumScanThreads > OPENCV_GPU_WARP_SIZE)
+        {
+            //Bottom-level inclusive warp scan
+            T warpResult = warpScanInclusive(idata, s_Data, tid);
+
+            //Save top elements of each warp for exclusive warp scan
+            //sync to wait for warp scans to complete (because s_Data is being overwritten)
+            __syncthreads();
+            if ((tid & (OPENCV_GPU_WARP_SIZE - 1)) == (OPENCV_GPU_WARP_SIZE - 1))
+            {
+                s_Data[tid >> OPENCV_GPU_LOG_WARP_SIZE] = warpResult;
+            }
+
+            //wait for warp scans to complete
+            __syncthreads();
+
+            if (tid < (tiNumScanThreads / OPENCV_GPU_WARP_SIZE) )
+            {
+                //grab top warp elements
+                T val = s_Data[tid];
+                //calculate exclusive scan and write back to shared memory
+                s_Data[tid] = warpScanExclusive(val, s_Data, tid);
+            }
+
+            //return updated warp scans with exclusive scan results
+            __syncthreads();
+
+            return warpResult + s_Data[tid >> OPENCV_GPU_LOG_WARP_SIZE];
+        }
+        else
+        {
+            return warpScanInclusive(idata, s_Data, tid);
+        }
+    }
 }}}
 
 #endif // __OPENCV_GPU_SCAN_HPP__
diff --git a/modules/core/include/opencv2/core/cuda/simd_functions.hpp b/modules/core/include/opencv2/core/cuda/simd_functions.hpp
new file mode 100644
index 0000000000..aedd5632f2
--- /dev/null
+++ b/modules/core/include/opencv2/core/cuda/simd_functions.hpp
@@ -0,0 +1,909 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/*
+ * Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ *
+ *   Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ *
+ *   Neither the name of NVIDIA Corporation nor the names of its contributors
+ *   may be used to endorse or promote products derived from this software
+ *   without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __OPENCV_GPU_SIMD_FUNCTIONS_HPP__
+#define __OPENCV_GPU_SIMD_FUNCTIONS_HPP__
+
+#include "common.hpp"
+
+/*
+  This header file contains inline functions that implement intra-word SIMD
+  operations, that are hardware accelerated on sm_3x (Kepler) GPUs. Efficient
+  emulation code paths are provided for earlier architectures (sm_1x, sm_2x)
+  to make the code portable across all GPUs supported by CUDA. The following
+  functions are currently implemented:
+
+  vadd2(a,b)      per-halfword unsigned addition, with wrap-around: a + b
+  vsub2(a,b)      per-halfword unsigned subtraction, with wrap-around: a - b
+  vabsdiff2(a,b)  per-halfword unsigned absolute difference: |a - b|
+  vavg2(a,b)      per-halfword unsigned average: (a + b) / 2
+  vavrg2(a,b)     per-halfword unsigned rounded average: (a + b + 1) / 2
+  vseteq2(a,b)    per-halfword unsigned comparison: a == b ? 1 : 0
+  vcmpeq2(a,b)    per-halfword unsigned comparison: a == b ? 0xffff : 0
+  vsetge2(a,b)    per-halfword unsigned comparison: a >= b ? 1 : 0
+  vcmpge2(a,b)    per-halfword unsigned comparison: a >= b ? 0xffff : 0
+  vsetgt2(a,b)    per-halfword unsigned comparison: a > b ? 1 : 0
+  vcmpgt2(a,b)    per-halfword unsigned comparison: a > b ? 0xffff : 0
+  vsetle2(a,b)    per-halfword unsigned comparison: a <= b ? 1 : 0
+  vcmple2(a,b)    per-halfword unsigned comparison: a <= b ? 0xffff : 0
+  vsetlt2(a,b)    per-halfword unsigned comparison: a < b ? 1 : 0
+  vcmplt2(a,b)    per-halfword unsigned comparison: a < b ? 0xffff : 0
+  vsetne2(a,b)    per-halfword unsigned comparison: a != b ? 1 : 0
+  vcmpne2(a,b)    per-halfword unsigned comparison: a != b ? 0xffff : 0
+  vmax2(a,b)      per-halfword unsigned maximum: max(a, b)
+  vmin2(a,b)      per-halfword unsigned minimum: min(a, b)
+
+  vadd4(a,b)      per-byte unsigned addition, with wrap-around: a + b
+  vsub4(a,b)      per-byte unsigned subtraction, with wrap-around: a - b
+  vabsdiff4(a,b)  per-byte unsigned absolute difference: |a - b|
+  vavg4(a,b)      per-byte unsigned average: (a + b) / 2
+  vavrg4(a,b)     per-byte unsigned rounded average: (a + b + 1) / 2
+  vseteq4(a,b)    per-byte unsigned comparison: a == b ? 1 : 0
+  vcmpeq4(a,b)    per-byte unsigned comparison: a == b ? 0xff : 0
+  vsetge4(a,b)    per-byte unsigned comparison: a >= b ? 1 : 0
+  vcmpge4(a,b)    per-byte unsigned comparison: a >= b ? 0xff : 0
+  vsetgt4(a,b)    per-byte unsigned comparison: a > b ? 1 : 0
+  vcmpgt4(a,b)    per-byte unsigned comparison: a > b ? 0xff : 0
+  vsetle4(a,b)    per-byte unsigned comparison: a <= b ? 1 : 0
+  vcmple4(a,b)    per-byte unsigned comparison: a <= b ? 0xff : 0
+  vsetlt4(a,b)    per-byte unsigned comparison: a < b ? 1 : 0
+  vcmplt4(a,b)    per-byte unsigned comparison: a < b ? 0xff : 0
+  vsetne4(a,b)    per-byte unsigned comparison: a != b ? 1: 0
+  vcmpne4(a,b)    per-byte unsigned comparison: a != b ? 0xff: 0
+  vmax4(a,b)      per-byte unsigned maximum: max(a, b)
+  vmin4(a,b)      per-byte unsigned minimum: min(a, b)
+*/
+
+namespace cv { namespace gpu { namespace cudev
+{
+    // 2
+
+    static __device__ __forceinline__ unsigned int vadd2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a + b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // carry-in to high word (= carry-out from low word)
+        r = r - s;          // subtract out carry-out from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a - b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // borrow to high word
+        r = r + s;          // compensate for borrow from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u, v;
+        s = a & 0x0000ffff; // extract low halfword
+        r = b & 0x0000ffff; // extract low halfword
+        u = ::max(r, s);    // maximum of low halfwords
+        v = ::min(r, s);    // minimum of low halfwords
+        s = a & 0xffff0000; // extract high halfword
+        r = b & 0xffff0000; // extract high halfword
+        t = ::max(r, s);    // maximum of high halfwords
+        s = ::min(r, s);    // minimum of high halfwords
+        r = u | t;          // maximum of both halfwords
+        s = v | s;          // minimum of both halfwords
+        r = r - s;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross halfword boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int s;
+        s = a ^ b;
+        r = a | b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross half-word boundaries
+        s = s >> 1;
+        r = r - s;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm ("vset2.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::max(r, s);    // maximum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::max(r, s);    // maximum of high halfwords
+        r = t | u;          // combine halfword maximums
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmin2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::min(r, s);    // minimum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::min(r, s);    // minimum of high halfwords
+        r = t | u;          // combine halfword minimums
+    #endif
+
+        return r;
+    }
+
+    // 4
+
+    static __device__ __forceinline__ unsigned int vadd4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ b;          // sum bits
+        r = a & 0x7f7f7f7f; // clear msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // msb sum bits
+        r = r + t;          // add without msbs, record carry-out in msbs
+        r = r ^ s;          // sum of msb sum and carry-in bits, w/o carry-out
+    #endif /* __CUDA_ARCH__ >= 300 */
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ ~b;         // inverted sum bits
+        r = a | 0x80808080; // set msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // inverted msb sum bits
+        r = r - t;          // subtract w/o msbs, record inverted borrows in msb
+        r = r ^ s;          // combine inverted msb sum bits and borrows
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int c;
+        c = a ^ b;
+        r = a | b;
+        c = c & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        c = c >> 1;
+        r = r - c;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x80
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r & ~c;         // msb = 1, if r was 0x00
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, t;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq4(a, b);
+        t = r << 8;         // convert bool
+        r = t - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        t = a ^ b;          // 0x00 if a == b
+        r = t | 0x80808080; // set msbs, to catch carry out
+        t = t ^ r;          // extract msbs, msb = 1 if t < 0x80
+        r = r - 0x01010101; // msb = 0, if t was 0x00 or 0x80
+        r = t & ~r;         // msb = 1, if t was 0x00
+        t = r >> 7;         // build mask
+        t = r - t;          //  from
+        r = t | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg4(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge4(a, b);
+        s = r << 8;         // convert bool
+        r = s - r;          //  to mask
+    #else
+        asm ("not.b32 %0,%0;" : "+r"(b));
+        r = vavrg4 (a, b);  // (a + ~b + 1) / 2 = (a - b) / 2
+        r = r & 0x80808080; // msb = carry-outs
+        s = r >> 7;         // build mask
+        s = r - s;          //  from
+        r = s | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a ^ b;          //
+        s = (r &  s) ^ b;   // select a when a >= b, else select b => max(a,b)
+        r = s ^ r;          // select a when b >= a, else select b => min(a,b)
+        r = s - r;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;           // byte-wise unsigned maximum
+    }
+
+    static __device__ __forceinline__ unsigned int vmin4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(b, a);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;
+    }
+}}}
+
+#endif // __OPENCV_GPU_SIMD_FUNCTIONS_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/transform.hpp b/modules/core/include/opencv2/core/cuda/transform.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/transform.hpp
rename to modules/core/include/opencv2/core/cuda/transform.hpp
index 636caac63f..7c82e36469 100644
--- a/modules/gpu/include/opencv2/gpu/device/transform.hpp
+++ b/modules/core/include/opencv2/core/cuda/transform.hpp
@@ -47,7 +47,7 @@
 #include "utility.hpp"
 #include "detail/transform_detail.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <typename T, typename D, typename UnOp, typename Mask>
     static inline void transform(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, const Mask& mask, cudaStream_t stream)
diff --git a/modules/gpu/include/opencv2/gpu/device/type_traits.hpp b/modules/core/include/opencv2/core/cuda/type_traits.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/type_traits.hpp
rename to modules/core/include/opencv2/core/cuda/type_traits.hpp
index 1b36acca5d..8a58264bfc 100644
--- a/modules/gpu/include/opencv2/gpu/device/type_traits.hpp
+++ b/modules/core/include/opencv2/core/cuda/type_traits.hpp
@@ -45,7 +45,7 @@
 
 #include "detail/type_traits_detail.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <typename T> struct IsSimpleParameter
     {
diff --git a/modules/gpu/include/opencv2/gpu/device/utility.hpp b/modules/core/include/opencv2/core/cuda/utility.hpp
similarity index 96%
rename from modules/gpu/include/opencv2/gpu/device/utility.hpp
rename to modules/core/include/opencv2/core/cuda/utility.hpp
index 83eaaa21ce..82c61a2014 100644
--- a/modules/gpu/include/opencv2/gpu/device/utility.hpp
+++ b/modules/core/include/opencv2/core/cuda/utility.hpp
@@ -46,7 +46,7 @@
 #include "saturate_cast.hpp"
 #include "datamov_utils.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     #define OPENCV_GPU_LOG_WARP_SIZE        (5)
     #define OPENCV_GPU_WARP_SIZE            (1 << OPENCV_GPU_LOG_WARP_SIZE)
@@ -124,8 +124,8 @@ namespace cv { namespace gpu { namespace device
 
     struct WithOutMask
     {
-        __device__ __forceinline__ WithOutMask(){}
-        __device__ __forceinline__ WithOutMask(const WithOutMask& mask){}
+        __host__ __device__ __forceinline__ WithOutMask(){}
+        __host__ __device__ __forceinline__ WithOutMask(const WithOutMask&){}
 
         __device__ __forceinline__ void next() const
         {
@@ -208,6 +208,6 @@ namespace cv { namespace gpu { namespace device
 
         return false;
     }
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_UTILITY_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/vec_distance.hpp b/modules/core/include/opencv2/core/cuda/vec_distance.hpp
similarity index 98%
rename from modules/gpu/include/opencv2/gpu/device/vec_distance.hpp
rename to modules/core/include/opencv2/core/cuda/vec_distance.hpp
index d5b4bb202c..4b88410207 100644
--- a/modules/gpu/include/opencv2/gpu/device/vec_distance.hpp
+++ b/modules/core/include/opencv2/core/cuda/vec_distance.hpp
@@ -47,7 +47,7 @@
 #include "functional.hpp"
 #include "detail/vec_distance_detail.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <typename T> struct L1Dist
     {
@@ -219,6 +219,6 @@ namespace cv { namespace gpu { namespace device
 
         U vec1Vals[MAX_LEN / THREAD_DIM];
     };
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_VEC_DISTANCE_HPP__
diff --git a/modules/core/include/opencv2/core/cuda/vec_math.hpp b/modules/core/include/opencv2/core/cuda/vec_math.hpp
new file mode 100644
index 0000000000..0dc92c3ef6
--- /dev/null
+++ b/modules/core/include/opencv2/core/cuda/vec_math.hpp
@@ -0,0 +1,922 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPU_VECMATH_HPP__
+#define __OPENCV_GPU_VECMATH_HPP__
+
+#include "vec_traits.hpp"
+#include "saturate_cast.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+
+// saturate_cast
+
+namespace vec_math_detail
+{
+    template <int cn, typename VecD> struct SatCastHelper;
+    template <typename VecD> struct SatCastHelper<1, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<2, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<3, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<4, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z), saturate_cast<D>(v.w));
+        }
+    };
+
+    template <typename VecD, typename VecS> static __device__ __forceinline__ VecD saturate_cast_helper(const VecS& v)
+    {
+        return SatCastHelper<VecTraits<VecD>::cn, VecD>::cast(v);
+    }
+}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+// unary operators
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(op (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(op (a.x), op (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(op (a.x), op (a.y), op (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(op (a.x), op (a.y), op (a.z), op (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_OP
+
+// unary functions
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x), func (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x), func (a.y), func (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x), func (a.y), func (a.z), func (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabsf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabs, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrt, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::exp, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::log, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::cosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanh, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC
+
+// binary operators (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op b.x, a.y op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op b.x, a.y op b.y, a.z op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op b.x, a.y op b.y, a.z op b.z, a.w op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_OP
+
+// binary operators (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(op, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op s); \
+    } \
+    __device__ __forceinline__ output_type ## 1 operator op(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(s op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op s, a.y op s); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(s op b.x, s op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op s, a.y op s, a.z op s); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(s op b.x, s op b.y, s op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op s, a.y op s, a.z op s, a.w op s); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(s op b.x, s op b.y, s op b.z, s op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uint, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP
+
+// binary function (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x, b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x, b.x), func (a.y, b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z), func (a.w, b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmaxf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmax, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fminf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fmin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypot, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC
+
+// binary function (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(func_name, func, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) a.x, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 1 func_name(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) s, (output_type) b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s), func ((output_type) a.w, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z), func ((output_type) s, (output_type) b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, double, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC
+
+}}} // namespace cv { namespace gpu { namespace device
+
+#endif // __OPENCV_GPU_VECMATH_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/vec_traits.hpp b/modules/core/include/opencv2/core/cuda/vec_traits.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/vec_traits.hpp
rename to modules/core/include/opencv2/core/cuda/vec_traits.hpp
index 8d179c83f5..304b05c919 100644
--- a/modules/gpu/include/opencv2/gpu/device/vec_traits.hpp
+++ b/modules/core/include/opencv2/core/cuda/vec_traits.hpp
@@ -45,7 +45,7 @@
 
 #include "common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template<typename T, int N> struct TypeVec;
 
@@ -275,6 +275,6 @@ namespace cv { namespace gpu { namespace device
         static __device__ __host__ __forceinline__ char8 make(schar a0, schar a1, schar a2, schar a3, schar a4, schar a5, schar a6, schar a7) {return make_char8(a0, a1, a2, a3, a4, a5, a6, a7);}
         static __device__ __host__ __forceinline__ char8 make(const schar* v) {return make_char8(v[0], v[1], v[2], v[3], v[4], v[5], v[6], v[7]);}
     };
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif // __OPENCV_GPU_VEC_TRAITS_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/warp.hpp b/modules/core/include/opencv2/core/cuda/warp.hpp
similarity index 97%
rename from modules/gpu/include/opencv2/gpu/device/warp.hpp
rename to modules/core/include/opencv2/core/cuda/warp.hpp
index 0f1dc794ab..6d2b7745f1 100644
--- a/modules/gpu/include/opencv2/gpu/device/warp.hpp
+++ b/modules/core/include/opencv2/core/cuda/warp.hpp
@@ -43,7 +43,7 @@
 #ifndef __OPENCV_GPU_DEVICE_WARP_HPP__
 #define __OPENCV_GPU_DEVICE_WARP_HPP__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     struct Warp
     {
@@ -126,6 +126,6 @@ namespace cv { namespace gpu { namespace device
                 *t = value;
         }
     };
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif /* __OPENCV_GPU_DEVICE_WARP_HPP__ */
diff --git a/modules/gpu/include/opencv2/gpu/device/warp_reduce.hpp b/modules/core/include/opencv2/core/cuda/warp_reduce.hpp
similarity index 92%
rename from modules/gpu/include/opencv2/gpu/device/warp_reduce.hpp
rename to modules/core/include/opencv2/core/cuda/warp_reduce.hpp
index 7ac85b095d..82185e8c0c 100644
--- a/modules/gpu/include/opencv2/gpu/device/warp_reduce.hpp
+++ b/modules/core/include/opencv2/core/cuda/warp_reduce.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -40,11 +40,10 @@
 //
 //M*/
 
-
 #ifndef OPENCV_GPU_WARP_REDUCE_HPP__
 #define OPENCV_GPU_WARP_REDUCE_HPP__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <class T>
     __device__ __forceinline__ T warp_reduce(volatile T *ptr , const unsigned int tid = threadIdx.x)
@@ -64,6 +63,6 @@ namespace cv { namespace gpu { namespace device
 
         return ptr[tid - lane];
     }
-}}} // namespace cv { namespace gpu { namespace device {
+}}} // namespace cv { namespace gpu { namespace cudev {
 
-#endif /* OPENCV_GPU_WARP_REDUCE_HPP__ */
\ No newline at end of file
+#endif /* OPENCV_GPU_WARP_REDUCE_HPP__ */
diff --git a/modules/gpu/include/opencv2/gpu/device/warp_shuffle.hpp b/modules/core/include/opencv2/core/cuda/warp_shuffle.hpp
similarity index 99%
rename from modules/gpu/include/opencv2/gpu/device/warp_shuffle.hpp
rename to modules/core/include/opencv2/core/cuda/warp_shuffle.hpp
index 8b4479a79b..aabcacfa4b 100644
--- a/modules/gpu/include/opencv2/gpu/device/warp_shuffle.hpp
+++ b/modules/core/include/opencv2/core/cuda/warp_shuffle.hpp
@@ -43,7 +43,7 @@
 #ifndef __OPENCV_GPU_WARP_SHUFFLE_HPP__
 #define __OPENCV_GPU_WARP_SHUFFLE_HPP__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <typename T>
     __device__ __forceinline__ T shfl(T val, int srcLane, int width = warpSize)
diff --git a/modules/core/include/opencv2/core/cvdef.h b/modules/core/include/opencv2/core/cvdef.h
index 06b2b5f255..46f294962d 100644
--- a/modules/core/include/opencv2/core/cvdef.h
+++ b/modules/core/include/opencv2/core/cvdef.h
@@ -73,8 +73,10 @@
 #  define CV_ENABLE_UNROLLED 1
 #endif
 
-#if (defined WIN32 || defined _WIN32 || defined WINCE) && defined CVAPI_EXPORTS
+#if (defined WIN32 || defined _WIN32 || defined WINCE || defined __CYGWIN__) && defined CVAPI_EXPORTS
 #  define CV_EXPORTS __declspec(dllexport)
+#elif defined __GNUC__ && __GNUC__ >= 4
+#  define CV_EXPORTS __attribute__ ((visibility ("default")))
 #else
 #  define CV_EXPORTS
 #endif
@@ -82,7 +84,7 @@
 #ifndef CV_INLINE
 #  if defined __cplusplus
 #    define CV_INLINE static inline
-#  elif (defined WIN32 || defined _WIN32 || defined WINCE) && !defined __GNUC__
+#  elif defined _MSC_VER
 #    define CV_INLINE __inline
 #  else
 #    define CV_INLINE static
@@ -97,17 +99,33 @@
 #  endif
 #endif
 
-/* intrinsics support */
+/* CPU features and intrinsics support */
+#define CV_CPU_NONE    0
+#define CV_CPU_MMX     1
+#define CV_CPU_SSE     2
+#define CV_CPU_SSE2    3
+#define CV_CPU_SSE3    4
+#define CV_CPU_SSSE3   5
+#define CV_CPU_SSE4_1  6
+#define CV_CPU_SSE4_2  7
+#define CV_CPU_POPCNT  8
+#define CV_CPU_AVX    10
+#define CV_CPU_NEON   11
+#define CV_HARDWARE_MAX_FEATURE 255
+
+// do not include SSE/AVX/NEON headers for NVCC compiler
+#ifndef __CUDACC__
+
 #if defined __SSE2__ || defined _M_X64  || (defined _M_IX86_FP && _M_IX86_FP >= 2)
-#  include "emmintrin.h"
+#  include <emmintrin.h>
 #  define CV_SSE 1
 #  define CV_SSE2 1
 #  if defined __SSE3__ || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include "pmmintrin.h"
+#    include <pmmintrin.h>
 #    define CV_SSE3 1
 #  endif
 #  if defined __SSSE3__  || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include "tmmintrin.h"
+#    include <tmmintrin.h>
 #    define CV_SSSE3 1
 #  endif
 #  if defined __SSE4_1__ || (defined _MSC_VER && _MSC_VER >= 1500)
@@ -131,11 +149,18 @@
 #  endif
 #endif
 
-#ifdef __ARM_NEON__
+#if (defined WIN32 || defined _WIN32) && defined(_M_ARM)
+# include <Intrin.h>
+# include "arm_neon.h"
+# define CV_NEON 1
+# define CPU_HAS_NEON_FEATURE (true)
+#elif defined(__ARM_NEON__)
 #  include <arm_neon.h>
 #  define CV_NEON 1
 #endif
 
+#endif // __CUDACC__
+
 #ifndef CV_SSE
 #  define CV_SSE 0
 #endif
@@ -174,7 +199,15 @@
 */
 
 #if !defined _MSC_VER && !defined __BORLANDC__
-#  include <stdint.h>
+#  if defined __cplusplus && __cplusplus >= 201103L
+#    include <cstdint>
+     typedef std::uint32_t uint;
+#  else
+#    include <stdint.h>
+     typedef uint32_t uint;
+#  endif
+#else
+   typedef unsigned uint;
 #endif
 
 typedef signed char schar;
@@ -305,17 +338,27 @@ typedef signed char schar;
 
 #if defined __BORLANDC__
 #  include <fastmath.h>
+#elif defined __cplusplus
+#  include <cmath>
 #else
 #  include <math.h>
 #endif
 
+#ifndef MIN
+#  define MIN(a,b)  ((a) > (b) ? (b) : (a))
+#endif
+
+#ifndef MAX
+#  define MAX(a,b)  ((a) < (b) ? (b) : (a))
+#endif
+
 #ifdef HAVE_TEGRA_OPTIMIZATION
 #  include "tegra_round.hpp"
 #endif
 
 CV_INLINE int cvRound( double value )
 {
-#if (defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ && defined __SSE2__ && !defined __APPLE__)
+#if ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
     __m128d t = _mm_set_sd( value );
     return _mm_cvtsd_si32(t);
 #elif defined _MSC_VER && defined _M_IX86
@@ -326,21 +369,27 @@ CV_INLINE int cvRound( double value )
         fistp t;
     }
     return t;
-#elif defined HAVE_LRINT || defined CV_ICC || defined __GNUC__
+#elif defined _MSC_VER && defined _M_ARM && defined HAVE_TEGRA_OPTIMIZATION
+    TEGRA_ROUND(value);
+#elif defined CV_ICC || defined __GNUC__
 #  ifdef HAVE_TEGRA_OPTIMIZATION
     TEGRA_ROUND(value);
 #  else
     return (int)lrint(value);
 #  endif
 #else
-    // while this is not IEEE754-compliant rounding, it's usually a good enough approximation
-    return (int)(value + (value >= 0 ? 0.5 : -0.5));
+    double intpart, fractpart;
+    fractpart = modf(value, &intpart);
+    if ((fabs(fractpart) != 0.5) || ((((int)intpart) % 2) != 0))
+        return (int)(value + (value >= 0 ? 0.5 : -0.5));
+    else
+        return (int)intpart;
 #endif
 }
 
 CV_INLINE int cvFloor( double value )
 {
-#if defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__ && !defined __APPLE__)
+#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
     __m128d t = _mm_set_sd( value );
     int i = _mm_cvtsd_si32(t);
     return i - _mm_movemask_pd(_mm_cmplt_sd(t, _mm_cvtsi32_sd(t,i)));
@@ -356,7 +405,7 @@ CV_INLINE int cvFloor( double value )
 
 CV_INLINE int cvCeil( double value )
 {
-#if defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__&& !defined __APPLE__)
+#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__&& !defined __APPLE__)) && !defined(__CUDACC__)
     __m128d t = _mm_set_sd( value );
     int i = _mm_cvtsd_si32(t);
     return i + _mm_movemask_pd(_mm_cmplt_sd(_mm_cvtsi32_sd(t,i), t));
@@ -401,11 +450,11 @@ CV_INLINE int cvIsInf( double value )
 #      define CV_XADD(addr, delta) __atomic_fetch_add((_Atomic(int)*)(addr), delta, 4)
 #    endif
 #  else
-#    ifdef __ATOMIC_ACQ_REL
+#    if defined __ATOMIC_ACQ_REL && !defined __clang__
        // version for gcc >= 4.7
-#      define CV_XADD(addr, delta) __atomic_fetch_add(addr, delta, __ATOMIC_ACQ_REL)
+#      define CV_XADD(addr, delta) (int)__atomic_fetch_add((unsigned*)(addr), (unsigned)(delta), __ATOMIC_ACQ_REL)
 #    else
-#      define CV_XADD(addr, delta) __sync_fetch_and_add(addr, delta)
+#      define CV_XADD(addr, delta) (int)__sync_fetch_and_add((unsigned*)(addr), (unsigned)(delta))
 #    endif
 #  endif
 #elif (defined WIN32 || defined _WIN32 || defined WINCE) && (!defined RC_INVOKED)
@@ -421,4 +470,4 @@ CV_INLINE int cvIsInf( double value )
    CV_INLINE CV_XADD(int* addr, int delta) { int tmp = *addr; *addr += delta; return tmp; }
 #endif
 
-#endif // __OPENCV_CORE_CVDEF_H__
\ No newline at end of file
+#endif // __OPENCV_CORE_CVDEF_H__
diff --git a/modules/core/include/opencv2/core/cvstd.hpp b/modules/core/include/opencv2/core/cvstd.hpp
new file mode 100644
index 0000000000..afdeb2549a
--- /dev/null
+++ b/modules/core/include/opencv2/core/cvstd.hpp
@@ -0,0 +1,882 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_CVSTD_HPP__
+#define __OPENCV_CORE_CVSTD_HPP__
+
+#ifndef __cplusplus
+#  error cvstd.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/cvdef.h"
+
+#include <cstddef>
+#include <cstring>
+#include <cctype>
+
+#ifndef OPENCV_NOSTL
+#  include <string>
+#endif
+
+// import useful primitives from stl
+#ifndef OPENCV_NOSTL_TRANSITIONAL
+#  include <algorithm>
+#  include <utility>
+#  include <cstdlib> //for abs(int)
+#  include <cmath>
+
+namespace cv
+{
+    using std::min;
+    using std::max;
+    using std::abs;
+    using std::swap;
+    using std::sqrt;
+    using std::exp;
+    using std::pow;
+    using std::log;
+}
+
+namespace std
+{
+    static inline uchar abs(uchar a) { return a; }
+    static inline ushort abs(ushort a) { return a; }
+    static inline unsigned abs(unsigned a) { return a; }
+    static inline uint64 abs(uint64 a) { return a; }
+}
+
+#else
+namespace cv
+{
+    template<typename T> static inline T min(T a, T b) { return a < b ? a : b; }
+    template<typename T> static inline T max(T a, T b) { return a > b ? a : b; }
+    template<typename T> static inline T abs(T a) { return a < 0 ? -a : a; }
+    template<typename T> static inline void swap(T& a, T& b) { T tmp = a; a = b; b = tmp; }
+
+    template<> inline uchar abs(uchar a) { return a; }
+    template<> inline ushort abs(ushort a) { return a; }
+    template<> inline unsigned abs(unsigned a) { return a; }
+    template<> inline uint64 abs(uint64 a) { return a; }
+}
+#endif
+
+namespace cv {
+
+//////////////////////////// memory management functions ////////////////////////////
+
+/*!
+  Allocates memory buffer
+
+  This is specialized OpenCV memory allocation function that returns properly aligned memory buffers.
+  The usage is identical to malloc(). The allocated buffers must be freed with cv::fastFree().
+  If there is not enough memory, the function calls cv::error(), which raises an exception.
+
+  \param bufSize buffer size in bytes
+  \return the allocated memory buffer.
+*/
+CV_EXPORTS void* fastMalloc(size_t bufSize);
+
+/*!
+  Frees the memory allocated with cv::fastMalloc
+
+  This is the corresponding deallocation function for cv::fastMalloc().
+  When ptr==NULL, the function has no effect.
+*/
+CV_EXPORTS void fastFree(void* ptr);
+
+/*!
+  The STL-compilant memory Allocator based on cv::fastMalloc() and cv::fastFree()
+*/
+template<typename _Tp> class Allocator
+{
+public:
+    typedef _Tp value_type;
+    typedef value_type* pointer;
+    typedef const value_type* const_pointer;
+    typedef value_type& reference;
+    typedef const value_type& const_reference;
+    typedef size_t size_type;
+    typedef ptrdiff_t difference_type;
+    template<typename U> class rebind { typedef Allocator<U> other; };
+
+    explicit Allocator() {}
+    ~Allocator() {}
+    explicit Allocator(Allocator const&) {}
+    template<typename U>
+    explicit Allocator(Allocator<U> const&) {}
+
+    // address
+    pointer address(reference r) { return &r; }
+    const_pointer address(const_reference r) { return &r; }
+
+    pointer allocate(size_type count, const void* =0) { return reinterpret_cast<pointer>(fastMalloc(count * sizeof (_Tp))); }
+    void deallocate(pointer p, size_type) { fastFree(p); }
+
+    void construct(pointer p, const _Tp& v) { new(static_cast<void*>(p)) _Tp(v); }
+    void destroy(pointer p) { p->~_Tp(); }
+
+    size_type max_size() const { return cv::max(static_cast<_Tp>(-1)/sizeof(_Tp), 1); }
+};
+
+namespace detail
+{
+
+// Metafunction to avoid taking a reference to void.
+template<typename T>
+struct RefOrVoid { typedef T& type; };
+
+template<>
+struct RefOrVoid<void>{ typedef void type; };
+
+template<>
+struct RefOrVoid<const void>{ typedef const void type; };
+
+template<>
+struct RefOrVoid<volatile void>{ typedef volatile void type; };
+
+template<>
+struct RefOrVoid<const volatile void>{ typedef const volatile void type; };
+
+// This class would be private to Ptr, if it didn't have to be a non-template.
+struct PtrOwner;
+
+}
+
+template<typename Y>
+struct DefaultDeleter
+{
+    void operator () (Y* p) const;
+};
+
+/*
+  A smart shared pointer class with reference counting.
+
+  A Ptr<T> stores a pointer and owns a (potentially different) pointer.
+  The stored pointer has type T and is the one returned by get() et al,
+  while the owned pointer can have any type and is the one deleted
+  when there are no more Ptrs that own it. You can't directly obtain the
+  owned pointer.
+
+  The interface of this class is mostly a subset of that of C++11's
+  std::shared_ptr.
+*/
+template<typename T>
+struct Ptr
+{
+    /* Generic programming support. */
+    typedef T element_type;
+
+    /* Ptr that owns NULL and stores NULL. */
+    Ptr();
+
+    /* Ptr that owns p and stores p. The owned pointer will be deleted with
+       DefaultDeleter<Y>. Y must be a complete type and Y* must be
+       convertible to T*. */
+    template<typename Y>
+    explicit Ptr(Y* p);
+
+    /* Ptr that owns p and stores p. The owned pointer will be deleted by
+       calling d(p). Y* must be convertible to T*. */
+    template<typename Y, typename D>
+    Ptr(Y* p, D d);
+
+    /* Same as the constructor below; it exists to suppress the generation
+       of the implicit copy constructor. */
+    Ptr(const Ptr& o);
+
+    /* Ptr that owns the same pointer as o and stores the same pointer as o,
+       converted to T*. Naturally, Y* must be convertible to T*. */
+    template<typename Y>
+    Ptr(const Ptr<Y>& o);
+
+    /* Ptr that owns same pointer as o, and stores p. Useful for casts and
+       creating non-owning Ptrs. */
+    template<typename Y>
+    Ptr(const Ptr<Y>& o, T* p);
+
+    /* Equivalent to release(). */
+    ~Ptr();
+
+    /* Same as assignment below; exists to suppress the generation of the
+       implicit assignment operator. */
+    Ptr& operator = (const Ptr& o);
+
+    template<typename Y>
+    Ptr& operator = (const Ptr<Y>& o);
+
+    /* Resets both the owned and stored pointers to NULL. Deletes the owned
+       pointer with the associated deleter if it's not owned by any other
+       Ptr and is non-zero. It's called reset() in std::shared_ptr; here
+       it is release() for compatibility with old OpenCV versions. */
+    void release();
+
+    /* Equivalent to assigning from Ptr<T>(p). */
+    template<typename Y>
+    void reset(Y* p);
+
+    /* Equivalent to assigning from Ptr<T>(p, d). */
+    template<typename Y, typename D>
+    void reset(Y* p, D d);
+
+    /* Swaps the stored and owned pointers of this and o. */
+    void swap(Ptr& o);
+
+    /* Returns the stored pointer. */
+    T* get() const;
+
+    /* Ordinary pointer emulation. */
+    typename detail::RefOrVoid<T>::type operator * () const;
+    T* operator -> () const;
+
+    /* Equivalent to get(). */
+    operator T* () const;
+
+    /* Equivalent to !*this. */
+    bool empty() const;
+
+    /* Returns a Ptr that owns the same pointer as this, and stores the same
+       pointer as this, except converted via static_cast to Y*. */
+    template<typename Y>
+    Ptr<Y> staticCast() const;
+
+    /* Ditto for const_cast. */
+    template<typename Y>
+    Ptr<Y> constCast() const;
+
+    /* Ditto for dynamic_cast. */
+    template<typename Y>
+    Ptr<Y> dynamicCast() const;
+
+private:
+    detail::PtrOwner* owner;
+    T* stored;
+
+    template<typename Y>
+    friend struct Ptr; // have to do this for the cross-type copy constructor
+};
+
+/* Overload of the generic swap. */
+template<typename T>
+void swap(Ptr<T>& ptr1, Ptr<T>& ptr2);
+
+/* Obvious comparisons. */
+template<typename T>
+bool operator == (const Ptr<T>& ptr1, const Ptr<T>& ptr2);
+template<typename T>
+bool operator != (const Ptr<T>& ptr1, const Ptr<T>& ptr2);
+
+/* Convenience creation functions. In the far future, there may be variadic templates here. */
+template<typename T>
+Ptr<T> makePtr();
+template<typename T, typename A1>
+Ptr<T> makePtr(const A1& a1);
+template<typename T, typename A1, typename A2>
+Ptr<T> makePtr(const A1& a1, const A2& a2);
+template<typename T, typename A1, typename A2, typename A3>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3);
+template<typename T, typename A1, typename A2, typename A3, typename A4>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8, typename A9>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9);
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8, typename A9, typename A10>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9, const A10& a10);
+
+
+//////////////////////////////// string class ////////////////////////////////
+
+class CV_EXPORTS FileNode; //for string constructor from FileNode
+
+class CV_EXPORTS String
+{
+public:
+    typedef char value_type;
+    typedef char& reference;
+    typedef const char& const_reference;
+    typedef char* pointer;
+    typedef const char* const_pointer;
+    typedef ptrdiff_t difference_type;
+    typedef size_t size_type;
+    typedef char* iterator;
+    typedef const char* const_iterator;
+
+    static const size_t npos = size_t(-1);
+
+    explicit String();
+    String(const String& str);
+    String(const String& str, size_t pos, size_t len = npos);
+    String(const char* s);
+    String(const char* s, size_t n);
+    String(size_t n, char c);
+    String(const char* first, const char* last);
+    template<typename Iterator> String(Iterator first, Iterator last);
+    explicit String(const FileNode& fn);
+    ~String();
+
+    String& operator=(const String& str);
+    String& operator=(const char* s);
+    String& operator=(char c);
+
+    size_t size() const;
+    size_t length() const;
+
+    char operator[](size_t idx) const;
+    char operator[](int idx) const;
+
+    const char* begin() const;
+    const char* end() const;
+
+    const char* c_str() const;
+
+    bool empty() const;
+    void clear();
+
+    int compare(const char* s) const;
+    int compare(const String& str) const;
+
+    void swap(String& str);
+    String substr(size_t pos = 0, size_t len = npos) const;
+
+    size_t find(const char* s, size_t pos, size_t n) const;
+    size_t find(char c, size_t pos = 0) const;
+    size_t find(const String& str, size_t pos = 0) const;
+    size_t find(const char* s, size_t pos = 0) const;
+
+    size_t rfind(const char* s, size_t pos, size_t n) const;
+    size_t rfind(char c, size_t pos = npos) const;
+    size_t rfind(const String& str, size_t pos = npos) const;
+    size_t rfind(const char* s, size_t pos = npos) const;
+
+    size_t find_first_of(const char* s, size_t pos, size_t n) const;
+    size_t find_first_of(char c, size_t pos = 0) const;
+    size_t find_first_of(const String& str, size_t pos = 0) const;
+    size_t find_first_of(const char* s, size_t pos = 0) const;
+
+    size_t find_last_of(const char* s, size_t pos, size_t n) const;
+    size_t find_last_of(char c, size_t pos = npos) const;
+    size_t find_last_of(const String& str, size_t pos = npos) const;
+    size_t find_last_of(const char* s, size_t pos = npos) const;
+
+    friend String operator+ (const String& lhs, const String& rhs);
+    friend String operator+ (const String& lhs, const char*   rhs);
+    friend String operator+ (const char*   lhs, const String& rhs);
+    friend String operator+ (const String& lhs, char          rhs);
+    friend String operator+ (char          lhs, const String& rhs);
+
+    String toLowerCase() const;
+
+#ifndef OPENCV_NOSTL
+    String(const std::string& str);
+    String(const std::string& str, size_t pos, size_t len = npos);
+    String& operator=(const std::string& str);
+    operator std::string() const;
+
+    friend String operator+ (const String& lhs, const std::string& rhs);
+    friend String operator+ (const std::string& lhs, const String& rhs);
+#endif
+
+private:
+    char*  cstr_;
+    size_t len_;
+
+    char* allocate(size_t len); // len without trailing 0
+    void deallocate();
+};
+
+
+////////////////////////// cv::String implementation /////////////////////////
+
+inline
+String::String()
+    : cstr_(0), len_(0)
+{}
+
+inline
+String::String(const String& str)
+    : cstr_(str.cstr_), len_(str.len_)
+{
+    if (cstr_)
+        CV_XADD(((int*)cstr_)-1, 1);
+}
+
+inline
+String::String(const String& str, size_t pos, size_t len)
+    : cstr_(0), len_(0)
+{
+    pos = min(pos, str.len_);
+    len = min(str.len_ - pos, len);
+    if (!len) return;
+    if (len == str.len_)
+    {
+        CV_XADD(((int*)str.cstr_)-1, 1);
+        cstr_ = str.cstr_;
+        len_ = str.len_;
+        return;
+    }
+    memcpy(allocate(len), str.cstr_ + pos, len);
+}
+
+inline
+String::String(const char* s)
+    : cstr_(0), len_(0)
+{
+    if (!s) return;
+    size_t len = strlen(s);
+    memcpy(allocate(len), s, len);
+}
+
+inline
+String::String(const char* s, size_t n)
+    : cstr_(0), len_(0)
+{
+    if (!n) return;
+    memcpy(allocate(n), s, n);
+}
+
+inline
+String::String(size_t n, char c)
+    : cstr_(0), len_(0)
+{
+    memset(allocate(n), c, n);
+}
+
+inline
+String::String(const char* first, const char* last)
+    : cstr_(0), len_(0)
+{
+    size_t len = (size_t)(last - first);
+    memcpy(allocate(len), first, len);
+}
+
+template<typename Iterator> inline
+String::String(Iterator first, Iterator last)
+    : cstr_(0), len_(0)
+{
+    size_t len = (size_t)(last - first);
+    char* str = allocate(len);
+    while (first != last)
+    {
+        *str++ = *first;
+        ++first;
+    }
+}
+
+inline
+String::~String()
+{
+    deallocate();
+}
+
+inline
+String& String::operator=(const String& str)
+{
+    if (&str == this) return *this;
+
+    deallocate();
+    if (str.cstr_) CV_XADD(((int*)str.cstr_)-1, 1);
+    cstr_ = str.cstr_;
+    len_ = str.len_;
+    return *this;
+}
+
+inline
+String& String::operator=(const char* s)
+{
+    deallocate();
+    if (!s) return *this;
+    size_t len = strlen(s);
+    memcpy(allocate(len), s, len);
+    return *this;
+}
+
+inline
+String& String::operator=(char c)
+{
+    deallocate();
+    allocate(1)[0] = c;
+    return *this;
+}
+
+inline
+size_t String::size() const
+{
+    return len_;
+}
+
+inline
+size_t String::length() const
+{
+    return len_;
+}
+
+inline
+char String::operator[](size_t idx) const
+{
+    return cstr_[idx];
+}
+
+inline
+char String::operator[](int idx) const
+{
+    return cstr_[idx];
+}
+
+inline
+const char* String::begin() const
+{
+    return cstr_;
+}
+
+inline
+const char* String::end() const
+{
+    return len_ ? cstr_ + 1 : 0;
+}
+
+inline
+bool String::empty() const
+{
+    return len_ == 0;
+}
+
+inline
+const char* String::c_str() const
+{
+    return cstr_ ? cstr_ : "";
+}
+
+inline
+void String::swap(String& str)
+{
+    cv::swap(cstr_, str.cstr_);
+    cv::swap(len_, str.len_);
+}
+
+inline
+void String::clear()
+{
+    deallocate();
+}
+
+inline
+int String::compare(const char* s) const
+{
+    if (cstr_ == s) return 0;
+    return strcmp(c_str(), s);
+}
+
+inline
+int String::compare(const String& str) const
+{
+    if (cstr_ == str.cstr_) return 0;
+    return strcmp(c_str(), str.c_str());
+}
+
+inline
+String String::substr(size_t pos, size_t len) const
+{
+    return String(*this, pos, len);
+}
+
+inline
+size_t String::find(const char* s, size_t pos, size_t n) const
+{
+    if (n == 0 || pos + n > len_) return npos;
+    const char* lmax = cstr_ + len_ - n;
+    for (const char* i = cstr_ + pos; i <= lmax; ++i)
+    {
+        size_t j = 0;
+        while (j < n && s[j] == i[j]) ++j;
+        if (j == n) return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::find(char c, size_t pos) const
+{
+    return find(&c, pos, 1);
+}
+
+inline
+size_t String::find(const String& str, size_t pos) const
+{
+    return find(str.c_str(), pos, str.len_);
+}
+
+inline
+size_t String::find(const char* s, size_t pos) const
+{
+    if (pos >= len_ || !s[0]) return npos;
+    const char* lmax = cstr_ + len_;
+    for (const char* i = cstr_ + pos; i < lmax; ++i)
+    {
+        size_t j = 0;
+        while (s[j] && s[j] == i[j])
+        {   if(i + j >= lmax) return npos;
+            ++j;
+        }
+        if (!s[j]) return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::rfind(const char* s, size_t pos, size_t n) const
+{
+    if (n > len_) return npos;
+    if (pos > len_ - n) pos = len_ - n;
+    for (const char* i = cstr_ + pos; i >= cstr_; --i)
+    {
+        size_t j = 0;
+        while (j < n && s[j] == i[j]) ++j;
+        if (j == n) return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::rfind(char c, size_t pos) const
+{
+    return rfind(&c, pos, 1);
+}
+
+inline
+size_t String::rfind(const String& str, size_t pos) const
+{
+    return rfind(str.c_str(), pos, str.len_);
+}
+
+inline
+size_t String::rfind(const char* s, size_t pos) const
+{
+    return rfind(s, pos, strlen(s));
+}
+
+inline
+size_t String::find_first_of(const char* s, size_t pos, size_t n) const
+{
+    if (n == 0 || pos + n > len_) return npos;
+    const char* lmax = cstr_ + len_;
+    for (const char* i = cstr_ + pos; i < lmax; ++i)
+    {
+        for (size_t j = 0; j < n; ++j)
+            if (s[j] == *i)
+                return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::find_first_of(char c, size_t pos) const
+{
+    return find_first_of(&c, pos, 1);
+}
+
+inline
+size_t String::find_first_of(const String& str, size_t pos) const
+{
+    return find_first_of(str.c_str(), pos, str.len_);
+}
+
+inline
+size_t String::find_first_of(const char* s, size_t pos) const
+{
+    if (pos >= len_ || !s[0]) return npos;
+    const char* lmax = cstr_ + len_;
+    for (const char* i = cstr_ + pos; i < lmax; ++i)
+    {
+        for (size_t j = 0; s[j]; ++j)
+            if (s[j] == *i)
+                return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::find_last_of(const char* s, size_t pos, size_t n) const
+{
+    if (pos >= len_) pos = len_ - 1;
+    for (const char* i = cstr_ + pos; i >= cstr_; --i)
+    {
+        for (size_t j = 0; j < n; ++j)
+            if (s[j] == *i)
+                return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+size_t String::find_last_of(char c, size_t pos) const
+{
+    return find_last_of(&c, pos, 1);
+}
+
+inline
+size_t String::find_last_of(const String& str, size_t pos) const
+{
+    return find_last_of(str.c_str(), pos, str.len_);
+}
+
+inline
+size_t String::find_last_of(const char* s, size_t pos) const
+{
+    if (pos >= len_) pos = len_ - 1;
+    for (const char* i = cstr_ + pos; i >= cstr_; --i)
+    {
+        for (size_t j = 0; s[j]; ++j)
+            if (s[j] == *i)
+                return (size_t)(i - cstr_);
+    }
+    return npos;
+}
+
+inline
+String String::toLowerCase() const
+{
+    String res(cstr_, len_);
+
+    for (size_t i = 0; i < len_; ++i)
+        res.cstr_[i] = (char) ::tolower(cstr_[i]);
+
+    return res;
+}
+
+// ************************* cv::String non-member functions *************************
+
+inline
+String operator + (const String& lhs, const String& rhs)
+{
+    String s;
+    s.allocate(lhs.len_ + rhs.len_);
+    memcpy(s.cstr_, lhs.cstr_, lhs.len_);
+    memcpy(s.cstr_ + lhs.len_, rhs.cstr_, rhs.len_);
+    return s;
+}
+
+inline
+String operator + (const String& lhs, const char* rhs)
+{
+    String s;
+    size_t rhslen = strlen(rhs);
+    s.allocate(lhs.len_ + rhslen);
+    memcpy(s.cstr_, lhs.cstr_, lhs.len_);
+    memcpy(s.cstr_ + lhs.len_, rhs, rhslen);
+    return s;
+}
+
+inline
+String operator + (const char* lhs, const String& rhs)
+{
+    String s;
+    size_t lhslen = strlen(lhs);
+    s.allocate(lhslen + rhs.len_);
+    memcpy(s.cstr_, lhs, lhslen);
+    memcpy(s.cstr_ + lhslen, rhs.cstr_, rhs.len_);
+    return s;
+}
+
+inline
+String operator + (const String& lhs, char rhs)
+{
+    String s;
+    s.allocate(lhs.len_ + 1);
+    memcpy(s.cstr_, lhs.cstr_, lhs.len_);
+    s.cstr_[lhs.len_] = rhs;
+    return s;
+}
+
+inline
+String operator + (char lhs, const String& rhs)
+{
+    String s;
+    s.allocate(rhs.len_ + 1);
+    s.cstr_[0] = lhs;
+    memcpy(s.cstr_ + 1, rhs.cstr_, rhs.len_);
+    return s;
+}
+
+static inline bool operator== (const String& lhs, const String& rhs) { return 0 == lhs.compare(rhs); }
+static inline bool operator== (const char*   lhs, const String& rhs) { return 0 == rhs.compare(lhs); }
+static inline bool operator== (const String& lhs, const char*   rhs) { return 0 == lhs.compare(rhs); }
+static inline bool operator!= (const String& lhs, const String& rhs) { return 0 != lhs.compare(rhs); }
+static inline bool operator!= (const char*   lhs, const String& rhs) { return 0 != rhs.compare(lhs); }
+static inline bool operator!= (const String& lhs, const char*   rhs) { return 0 != lhs.compare(rhs); }
+static inline bool operator<  (const String& lhs, const String& rhs) { return lhs.compare(rhs) <  0; }
+static inline bool operator<  (const char*   lhs, const String& rhs) { return rhs.compare(lhs) >  0; }
+static inline bool operator<  (const String& lhs, const char*   rhs) { return lhs.compare(rhs) <  0; }
+static inline bool operator<= (const String& lhs, const String& rhs) { return lhs.compare(rhs) <= 0; }
+static inline bool operator<= (const char*   lhs, const String& rhs) { return rhs.compare(lhs) >= 0; }
+static inline bool operator<= (const String& lhs, const char*   rhs) { return lhs.compare(rhs) <= 0; }
+static inline bool operator>  (const String& lhs, const String& rhs) { return lhs.compare(rhs) >  0; }
+static inline bool operator>  (const char*   lhs, const String& rhs) { return rhs.compare(lhs) <  0; }
+static inline bool operator>  (const String& lhs, const char*   rhs) { return lhs.compare(rhs) >  0; }
+static inline bool operator>= (const String& lhs, const String& rhs) { return lhs.compare(rhs) >= 0; }
+static inline bool operator>= (const char*   lhs, const String& rhs) { return rhs.compare(lhs) <= 0; }
+static inline bool operator>= (const String& lhs, const char*   rhs) { return lhs.compare(rhs) >= 0; }
+
+} // cv
+
+#ifndef OPENCV_NOSTL_TRANSITIONAL
+namespace std
+#else
+namespace cv
+#endif
+{
+    template<> inline
+    void swap<cv::String>(cv::String& a, cv::String& b)
+    {
+        a.swap(b);
+    }
+}
+
+#include "opencv2/core/ptr.inl.hpp"
+
+#endif //__OPENCV_CORE_CVSTD_HPP__
diff --git a/modules/core/include/opencv2/core/cvstd.inl.hpp b/modules/core/include/opencv2/core/cvstd.inl.hpp
new file mode 100644
index 0000000000..8642b7442e
--- /dev/null
+++ b/modules/core/include/opencv2/core/cvstd.inl.hpp
@@ -0,0 +1,250 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_CVSTDINL_HPP__
+#define __OPENCV_CORE_CVSTDINL_HPP__
+
+#ifndef OPENCV_NOSTL
+#  include <complex>
+#  include <ostream>
+#endif
+
+namespace cv
+{
+#ifndef OPENCV_NOSTL
+
+template<typename _Tp> class DataType< std::complex<_Tp> >
+{
+public:
+    typedef std::complex<_Tp>  value_type;
+    typedef value_type         work_type;
+    typedef _Tp                channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels) };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+inline
+String::String(const std::string& str)
+    : cstr_(0), len_(0)
+{
+    if (!str.empty())
+    {
+        size_t len = str.size();
+        memcpy(allocate(len), str.c_str(), len);
+    }
+}
+
+inline
+String::String(const std::string& str, size_t pos, size_t len)
+    : cstr_(0), len_(0)
+{
+    size_t strlen = str.size();
+    pos = max(pos, strlen);
+    len = min(strlen - pos, len);
+    if (!len) return;
+    memcpy(allocate(len), str.c_str() + pos, len);
+}
+
+inline
+String& String::operator = (const std::string& str)
+{
+    deallocate();
+    if (!str.empty())
+    {
+        size_t len = str.size();
+        memcpy(allocate(len), str.c_str(), len);
+    }
+    return *this;
+}
+
+inline
+String::operator std::string() const
+{
+    return std::string(cstr_, len_);
+}
+
+inline
+String operator + (const String& lhs, const std::string& rhs)
+{
+    String s;
+    size_t rhslen = rhs.size();
+    s.allocate(lhs.len_ + rhslen);
+    memcpy(s.cstr_, lhs.cstr_, lhs.len_);
+    memcpy(s.cstr_ + lhs.len_, rhs.c_str(), rhslen);
+    return s;
+}
+
+inline
+String operator + (const std::string& lhs, const String& rhs)
+{
+    String s;
+    size_t lhslen = lhs.size();
+    s.allocate(lhslen + rhs.len_);
+    memcpy(s.cstr_, lhs.c_str(), lhslen);
+    memcpy(s.cstr_ + lhslen, rhs.cstr_, rhs.len_);
+    return s;
+}
+
+inline
+FileNode::operator std::string() const
+{
+    String value;
+    read(*this, value, value);
+    return value;
+}
+
+template<> inline
+void operator >> (const FileNode& n, std::string& value)
+{
+    String val;
+    read(n, val, val);
+    value = val;
+}
+
+template<> inline
+FileStorage& operator << (FileStorage& fs, const std::string& value)
+{
+    return fs << cv::String(value);
+}
+
+static inline
+std::ostream& operator << (std::ostream& os, const String& str)
+{
+    return os << str.c_str();
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, Ptr<Formatted> fmtd)
+{
+    fmtd->reset();
+    for(const char* str = fmtd->next(); str; str = fmtd->next())
+        out << str;
+    return out;
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, const Mat& mtx)
+{
+    return out << Formatter::get()->format(mtx);
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point3_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp, int m, int n> static inline
+std::ostream& operator << (std::ostream& out, const Matx<_Tp, m, n>& matx)
+{
+    return out << Formatter::get()->format(matx);
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << "]";
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point3_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << ", " << p.z << "]";
+    return out;
+}
+
+template<typename _Tp, int n> static inline
+std::ostream& operator << (std::ostream& out, const Vec<_Tp, n>& vec)
+{
+    out << "[";
+
+    if(Vec<_Tp, n>::depth < CV_32F)
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << (int)vec[i] << ", ";
+        }
+        out << (int)vec[n-1] << "]";
+    }
+    else
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << vec[i] << ", ";
+        }
+        out << vec[n-1] << "]";
+    }
+
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Size_<_Tp>& size)
+{
+    return out << "[" << size.width << " x " << size.height << "]";
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Rect_<_Tp>& rect)
+{
+    return out << "[" << rect.width << " x " << rect.height << " from (" << rect.x << ", " << rect.y << ")]";
+}
+
+
+#endif // OPENCV_NOSTL
+} // cv
+
+#endif // __OPENCV_CORE_CVSTDINL_HPP__
diff --git a/modules/core/include/opencv2/core/eigen.hpp b/modules/core/include/opencv2/core/eigen.hpp
index 6e47206c54..3005bfbfd4 100644
--- a/modules/core/include/opencv2/core/eigen.hpp
+++ b/modules/core/include/opencv2/core/eigen.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,12 +41,10 @@
 //
 //M*/
 
+
 #ifndef __OPENCV_CORE_EIGEN_HPP__
 #define __OPENCV_CORE_EIGEN_HPP__
 
-#ifdef __cplusplus
-
-#include "opencv2/core/core_c.h"
 #include "opencv2/core.hpp"
 
 #if defined _MSC_VER && _MSC_VER >= 1200
@@ -57,7 +56,7 @@
 namespace cv
 {
 
-template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols>
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
 void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src, Mat& dst )
 {
     if( !(src.Flags & Eigen::RowMajorBit) )
@@ -75,7 +74,7 @@ void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCo
 }
 
 // Matx case
-template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols>
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
 void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src,
                Matx<_Tp, _rows, _cols>& dst )
 {
@@ -89,14 +88,14 @@ void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCo
     }
 }
 
-template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols>
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
 void cv2eigen( const Mat& src,
                Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
 {
     CV_DbgAssert(src.rows == _rows && src.cols == _cols);
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
+        const Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         if( src.type() == _dst.type() )
             transpose(src, _dst);
@@ -107,46 +106,42 @@ void cv2eigen( const Mat& src,
         }
         else
             Mat(src.t()).convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
+        const Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         src.convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
 // Matx case
-template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols>
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
 void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
                Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
 {
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(_cols, _rows, DataType<_Tp>::type,
+        const Mat _dst(_cols, _rows, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         transpose(src, _dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(_rows, _cols, DataType<_Tp>::type,
+        const Mat _dst(_rows, _cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         Mat(src).copyTo(_dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
-template<typename _Tp>
+template<typename _Tp>  static inline
 void cv2eigen( const Mat& src,
                Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
 {
     dst.resize(src.rows, src.cols);
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
+        const Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
              dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         if( src.type() == _dst.type() )
             transpose(src, _dst);
@@ -157,40 +152,36 @@ void cv2eigen( const Mat& src,
         }
         else
             Mat(src.t()).convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
+        const Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         src.convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
 // Matx case
-template<typename _Tp, int _rows, int _cols>
+template<typename _Tp, int _rows, int _cols> static inline
 void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
                Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
 {
     dst.resize(_rows, _cols);
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(_cols, _rows, DataType<_Tp>::type,
+        const Mat _dst(_cols, _rows, DataType<_Tp>::type,
              dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         transpose(src, _dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(_rows, _cols, DataType<_Tp>::type,
+        const Mat _dst(_rows, _cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         Mat(src).copyTo(_dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
-template<typename _Tp>
+template<typename _Tp> static inline
 void cv2eigen( const Mat& src,
                Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
 {
@@ -199,25 +190,23 @@ void cv2eigen( const Mat& src,
 
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
+        const Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         if( src.type() == _dst.type() )
             transpose(src, _dst);
         else
             Mat(src.t()).convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
+        const Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         src.convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
 // Matx case
-template<typename _Tp, int _rows>
+template<typename _Tp, int _rows> static inline
 void cv2eigen( const Matx<_Tp, _rows, 1>& src,
                Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
 {
@@ -225,22 +214,20 @@ void cv2eigen( const Matx<_Tp, _rows, 1>& src,
 
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(1, _rows, DataType<_Tp>::type,
+        const Mat _dst(1, _rows, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         transpose(src, _dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(_rows, 1, DataType<_Tp>::type,
+        const Mat _dst(_rows, 1, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         src.copyTo(_dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
 
-template<typename _Tp>
+template<typename _Tp> static inline
 void cv2eigen( const Mat& src,
                Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
 {
@@ -248,49 +235,41 @@ void cv2eigen( const Mat& src,
     dst.resize(src.cols);
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
+        const Mat _dst(src.cols, src.rows, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         if( src.type() == _dst.type() )
             transpose(src, _dst);
         else
             Mat(src.t()).convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
+        const Mat _dst(src.rows, src.cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         src.convertTo(_dst, _dst.type());
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
 //Matx
-template<typename _Tp, int _cols>
+template<typename _Tp, int _cols> static inline
 void cv2eigen( const Matx<_Tp, 1, _cols>& src,
                Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
 {
     dst.resize(_cols);
     if( !(dst.Flags & Eigen::RowMajorBit) )
     {
-        Mat _dst(_cols, 1, DataType<_Tp>::type,
+        const Mat _dst(_cols, 1, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         transpose(src, _dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
     else
     {
-        Mat _dst(1, _cols, DataType<_Tp>::type,
+        const Mat _dst(1, _cols, DataType<_Tp>::type,
                  dst.data(), (size_t)(dst.stride()*sizeof(_Tp)));
         Mat(src).copyTo(_dst);
-        CV_DbgAssert(_dst.data == (uchar*)dst.data());
     }
 }
 
-
-}
+} // cv
 
 #endif
-
-#endif
-
diff --git a/modules/core/include/opencv2/core/gpu.hpp b/modules/core/include/opencv2/core/gpu.hpp
new file mode 100644
index 0000000000..dfd4dd92ac
--- /dev/null
+++ b/modules/core/include/opencv2/core/gpu.hpp
@@ -0,0 +1,672 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_GPU_HPP__
+#define __OPENCV_CORE_GPU_HPP__
+
+#ifndef __cplusplus
+#  error gpu.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core.hpp"
+#include "opencv2/core/gpu_types.hpp"
+
+namespace cv { namespace gpu {
+
+//////////////////////////////// GpuMat ///////////////////////////////
+
+// Smart pointer for GPU memory with reference counting.
+// Its interface is mostly similar with cv::Mat.
+
+class CV_EXPORTS GpuMat
+{
+public:
+    //! default constructor
+    GpuMat();
+
+    //! constructs GpuMat of the specified size and type
+    GpuMat(int rows, int cols, int type);
+    GpuMat(Size size, int type);
+
+    //! constucts GpuMat and fills it with the specified value _s
+    GpuMat(int rows, int cols, int type, Scalar s);
+    GpuMat(Size size, int type, Scalar s);
+
+    //! copy constructor
+    GpuMat(const GpuMat& m);
+
+    //! constructor for GpuMat headers pointing to user-allocated data
+    GpuMat(int rows, int cols, int type, void* data, size_t step = Mat::AUTO_STEP);
+    GpuMat(Size size, int type, void* data, size_t step = Mat::AUTO_STEP);
+
+    //! creates a GpuMat header for a part of the bigger matrix
+    GpuMat(const GpuMat& m, Range rowRange, Range colRange);
+    GpuMat(const GpuMat& m, Rect roi);
+
+    //! builds GpuMat from host memory (Blocking call)
+    explicit GpuMat(InputArray arr);
+
+    //! destructor - calls release()
+    ~GpuMat();
+
+    //! assignment operators
+    GpuMat& operator =(const GpuMat& m);
+
+    //! allocates new GpuMat data unless the GpuMat already has specified size and type
+    void create(int rows, int cols, int type);
+    void create(Size size, int type);
+
+    //! decreases reference counter, deallocate the data when reference counter reaches 0
+    void release();
+
+    //! swaps with other smart pointer
+    void swap(GpuMat& mat);
+
+    //! pefroms upload data to GpuMat (Blocking call)
+    void upload(InputArray arr);
+
+    //! pefroms upload data to GpuMat (Non-Blocking call)
+    void upload(InputArray arr, Stream& stream);
+
+    //! pefroms download data from device to host memory (Blocking call)
+    void download(OutputArray dst) const;
+
+    //! pefroms download data from device to host memory (Non-Blocking call)
+    void download(OutputArray dst, Stream& stream) const;
+
+    //! returns deep copy of the GpuMat, i.e. the data is copied
+    GpuMat clone() const;
+
+    //! copies the GpuMat content to device memory (Blocking call)
+    void copyTo(OutputArray dst) const;
+
+    //! copies the GpuMat content to device memory (Non-Blocking call)
+    void copyTo(OutputArray dst, Stream& stream) const;
+
+    //! copies those GpuMat elements to "m" that are marked with non-zero mask elements (Blocking call)
+    void copyTo(OutputArray dst, InputArray mask) const;
+
+    //! copies those GpuMat elements to "m" that are marked with non-zero mask elements (Non-Blocking call)
+    void copyTo(OutputArray dst, InputArray mask, Stream& stream) const;
+
+    //! sets some of the GpuMat elements to s (Blocking call)
+    GpuMat& setTo(Scalar s);
+
+    //! sets some of the GpuMat elements to s (Non-Blocking call)
+    GpuMat& setTo(Scalar s, Stream& stream);
+
+    //! sets some of the GpuMat elements to s, according to the mask (Blocking call)
+    GpuMat& setTo(Scalar s, InputArray mask);
+
+    //! sets some of the GpuMat elements to s, according to the mask (Non-Blocking call)
+    GpuMat& setTo(Scalar s, InputArray mask, Stream& stream);
+
+    //! converts GpuMat to another datatype (Blocking call)
+    void convertTo(OutputArray dst, int rtype) const;
+
+    //! converts GpuMat to another datatype (Non-Blocking call)
+    void convertTo(OutputArray dst, int rtype, Stream& stream) const;
+
+    //! converts GpuMat to another datatype with scaling (Blocking call)
+    void convertTo(OutputArray dst, int rtype, double alpha, double beta = 0.0) const;
+
+    //! converts GpuMat to another datatype with scaling (Non-Blocking call)
+    void convertTo(OutputArray dst, int rtype, double alpha, Stream& stream) const;
+
+    //! converts GpuMat to another datatype with scaling (Non-Blocking call)
+    void convertTo(OutputArray dst, int rtype, double alpha, double beta, Stream& stream) const;
+
+    void assignTo(GpuMat& m, int type=-1) const;
+
+    //! returns pointer to y-th row
+    uchar* ptr(int y = 0);
+    const uchar* ptr(int y = 0) const;
+
+    //! template version of the above method
+    template<typename _Tp> _Tp* ptr(int y = 0);
+    template<typename _Tp> const _Tp* ptr(int y = 0) const;
+
+    template <typename _Tp> operator PtrStepSz<_Tp>() const;
+    template <typename _Tp> operator PtrStep<_Tp>() const;
+
+    //! returns a new GpuMat header for the specified row
+    GpuMat row(int y) const;
+
+    //! returns a new GpuMat header for the specified column
+    GpuMat col(int x) const;
+
+    //! ... for the specified row span
+    GpuMat rowRange(int startrow, int endrow) const;
+    GpuMat rowRange(Range r) const;
+
+    //! ... for the specified column span
+    GpuMat colRange(int startcol, int endcol) const;
+    GpuMat colRange(Range r) const;
+
+    //! extracts a rectangular sub-GpuMat (this is a generalized form of row, rowRange etc.)
+    GpuMat operator ()(Range rowRange, Range colRange) const;
+    GpuMat operator ()(Rect roi) const;
+
+    //! creates alternative GpuMat header for the same data, with different
+    //! number of channels and/or different number of rows
+    GpuMat reshape(int cn, int rows = 0) const;
+
+    //! locates GpuMat header within a parent GpuMat
+    void locateROI(Size& wholeSize, Point& ofs) const;
+
+    //! moves/resizes the current GpuMat ROI inside the parent GpuMat
+    GpuMat& adjustROI(int dtop, int dbottom, int dleft, int dright);
+
+    //! returns true iff the GpuMat data is continuous
+    //! (i.e. when there are no gaps between successive rows)
+    bool isContinuous() const;
+
+    //! returns element size in bytes
+    size_t elemSize() const;
+
+    //! returns the size of element channel in bytes
+    size_t elemSize1() const;
+
+    //! returns element type
+    int type() const;
+
+    //! returns element type
+    int depth() const;
+
+    //! returns number of channels
+    int channels() const;
+
+    //! returns step/elemSize1()
+    size_t step1() const;
+
+    //! returns GpuMat size : width == number of columns, height == number of rows
+    Size size() const;
+
+    //! returns true if GpuMat data is NULL
+    bool empty() const;
+
+    /*! includes several bit-fields:
+    - the magic signature
+    - continuity flag
+    - depth
+    - number of channels
+    */
+    int flags;
+
+    //! the number of rows and columns
+    int rows, cols;
+
+    //! a distance between successive rows in bytes; includes the gap if any
+    size_t step;
+
+    //! pointer to the data
+    uchar* data;
+
+    //! pointer to the reference counter;
+    //! when GpuMat points to user-allocated data, the pointer is NULL
+    int* refcount;
+
+    //! helper fields used in locateROI and adjustROI
+    uchar* datastart;
+    uchar* dataend;
+};
+
+//! creates continuous matrix
+CV_EXPORTS void createContinuous(int rows, int cols, int type, OutputArray arr);
+
+//! ensures that size of the given matrix is not less than (rows, cols) size
+//! and matrix type is match specified one too
+CV_EXPORTS void ensureSizeIsEnough(int rows, int cols, int type, OutputArray arr);
+
+CV_EXPORTS GpuMat allocMatFromBuf(int rows, int cols, int type, GpuMat& mat);
+
+//////////////////////////////// CudaMem ////////////////////////////////
+
+// CudaMem is limited cv::Mat with page locked memory allocation.
+// Page locked memory is only needed for async and faster coping to GPU.
+// It is convertable to cv::Mat header without reference counting
+// so you can use it with other opencv functions.
+
+class CV_EXPORTS CudaMem
+{
+public:
+    enum AllocType { PAGE_LOCKED = 1, SHARED = 2, WRITE_COMBINED = 4 };
+
+    explicit CudaMem(AllocType alloc_type = PAGE_LOCKED);
+
+    CudaMem(const CudaMem& m);
+
+    CudaMem(int rows, int cols, int type, AllocType alloc_type = PAGE_LOCKED);
+    CudaMem(Size size, int type, AllocType alloc_type = PAGE_LOCKED);
+
+    //! creates from host memory with coping data
+    explicit CudaMem(InputArray arr, AllocType alloc_type = PAGE_LOCKED);
+
+    ~CudaMem();
+
+    CudaMem& operator =(const CudaMem& m);
+
+    //! swaps with other smart pointer
+    void swap(CudaMem& b);
+
+    //! returns deep copy of the matrix, i.e. the data is copied
+    CudaMem clone() const;
+
+    //! allocates new matrix data unless the matrix already has specified size and type.
+    void create(int rows, int cols, int type);
+    void create(Size size, int type);
+
+    //! creates alternative CudaMem header for the same data, with different
+    //! number of channels and/or different number of rows
+    CudaMem reshape(int cn, int rows = 0) const;
+
+    //! decrements reference counter and released memory if needed.
+    void release();
+
+    //! returns matrix header with disabled reference counting for CudaMem data.
+    Mat createMatHeader() const;
+
+    //! maps host memory into device address space and returns GpuMat header for it. Throws exception if not supported by hardware.
+    GpuMat createGpuMatHeader() const;
+
+    // Please see cv::Mat for descriptions
+    bool isContinuous() const;
+    size_t elemSize() const;
+    size_t elemSize1() const;
+    int type() const;
+    int depth() const;
+    int channels() const;
+    size_t step1() const;
+    Size size() const;
+    bool empty() const;
+
+    // Please see cv::Mat for descriptions
+    int flags;
+    int rows, cols;
+    size_t step;
+
+    uchar* data;
+    int* refcount;
+
+    uchar* datastart;
+    uchar* dataend;
+
+    AllocType alloc_type;
+};
+
+//! page-locks the matrix m memory and maps it for the device(s)
+CV_EXPORTS void registerPageLocked(Mat& m);
+
+//! unmaps the memory of matrix m, and makes it pageable again
+CV_EXPORTS void unregisterPageLocked(Mat& m);
+
+///////////////////////////////// Stream //////////////////////////////////
+
+// Encapculates Cuda Stream. Provides interface for async coping.
+// Passed to each function that supports async kernel execution.
+// Reference counting is enabled.
+
+class CV_EXPORTS Stream
+{
+    typedef void (Stream::*bool_type)() const;
+    void this_type_does_not_support_comparisons() const {}
+
+public:
+    typedef void (*StreamCallback)(int status, void* userData);
+
+    //! creates a new asynchronous stream
+    Stream();
+
+    //! queries an asynchronous stream for completion status
+    bool queryIfComplete() const;
+
+    //! waits for stream tasks to complete
+    void waitForCompletion();
+
+    //! makes a compute stream wait on an event
+    void waitEvent(const Event& event);
+
+    //! adds a callback to be called on the host after all currently enqueued items in the stream have completed
+    void enqueueHostCallback(StreamCallback callback, void* userData);
+
+    //! return Stream object for default CUDA stream
+    static Stream& Null();
+
+    //! returns true if stream object is not default (!= 0)
+    operator bool_type() const;
+
+    class Impl;
+
+private:
+    Ptr<Impl> impl_;
+    Stream(const Ptr<Impl>& impl);
+
+    friend struct StreamAccessor;
+};
+
+class CV_EXPORTS Event
+{
+public:
+    enum CreateFlags
+    {
+        DEFAULT        = 0x00,  /**< Default event flag */
+        BLOCKING_SYNC  = 0x01,  /**< Event uses blocking synchronization */
+        DISABLE_TIMING = 0x02,  /**< Event will not record timing data */
+        INTERPROCESS   = 0x04   /**< Event is suitable for interprocess use. DisableTiming must be set */
+    };
+
+    explicit Event(CreateFlags flags = DEFAULT);
+
+    //! records an event
+    void record(Stream& stream = Stream::Null());
+
+    //! queries an event's status
+    bool queryIfComplete() const;
+
+    //! waits for an event to complete
+    void waitForCompletion();
+
+    //! computes the elapsed time between events
+    static float elapsedTime(const Event& start, const Event& end);
+
+    class Impl;
+
+private:
+    Ptr<Impl> impl_;
+
+    friend struct EventAccessor;
+};
+
+//////////////////////////////// Initialization & Info ////////////////////////
+
+//! this is the only function that do not throw exceptions if the library is compiled without CUDA
+CV_EXPORTS int getCudaEnabledDeviceCount();
+
+//! set device to be used for GPU executions for the calling host thread
+CV_EXPORTS void setDevice(int device);
+
+//! returns which device is currently being used for the calling host thread
+CV_EXPORTS int getDevice();
+
+//! explicitly destroys and cleans up all resources associated with the current device in the current process
+//! any subsequent API call to this device will reinitialize the device
+CV_EXPORTS void resetDevice();
+
+enum FeatureSet
+{
+    FEATURE_SET_COMPUTE_10 = 10,
+    FEATURE_SET_COMPUTE_11 = 11,
+    FEATURE_SET_COMPUTE_12 = 12,
+    FEATURE_SET_COMPUTE_13 = 13,
+    FEATURE_SET_COMPUTE_20 = 20,
+    FEATURE_SET_COMPUTE_21 = 21,
+    FEATURE_SET_COMPUTE_30 = 30,
+    FEATURE_SET_COMPUTE_35 = 35,
+
+    GLOBAL_ATOMICS = FEATURE_SET_COMPUTE_11,
+    SHARED_ATOMICS = FEATURE_SET_COMPUTE_12,
+    NATIVE_DOUBLE = FEATURE_SET_COMPUTE_13,
+    WARP_SHUFFLE_FUNCTIONS = FEATURE_SET_COMPUTE_30,
+    DYNAMIC_PARALLELISM = FEATURE_SET_COMPUTE_35
+};
+
+//! checks whether current device supports the given feature
+CV_EXPORTS bool deviceSupports(FeatureSet feature_set);
+
+//! information about what GPU archs this OpenCV GPU module was compiled for
+class CV_EXPORTS TargetArchs
+{
+public:
+    static bool builtWith(FeatureSet feature_set);
+
+    static bool has(int major, int minor);
+    static bool hasPtx(int major, int minor);
+    static bool hasBin(int major, int minor);
+
+    static bool hasEqualOrLessPtx(int major, int minor);
+    static bool hasEqualOrGreater(int major, int minor);
+    static bool hasEqualOrGreaterPtx(int major, int minor);
+    static bool hasEqualOrGreaterBin(int major, int minor);
+};
+
+//! information about the given GPU.
+class CV_EXPORTS DeviceInfo
+{
+public:
+    //! creates DeviceInfo object for the current GPU
+    DeviceInfo();
+
+    //! creates DeviceInfo object for the given GPU
+    DeviceInfo(int device_id);
+
+    //! device number.
+    int deviceID() const;
+
+    //! ASCII string identifying device
+    const char* name() const;
+
+    //! global memory available on device in bytes
+    size_t totalGlobalMem() const;
+
+    //! shared memory available per block in bytes
+    size_t sharedMemPerBlock() const;
+
+    //! 32-bit registers available per block
+    int regsPerBlock() const;
+
+    //! warp size in threads
+    int warpSize() const;
+
+    //! maximum pitch in bytes allowed by memory copies
+    size_t memPitch() const;
+
+    //! maximum number of threads per block
+    int maxThreadsPerBlock() const;
+
+    //! maximum size of each dimension of a block
+    Vec3i maxThreadsDim() const;
+
+    //! maximum size of each dimension of a grid
+    Vec3i maxGridSize() const;
+
+    //! clock frequency in kilohertz
+    int clockRate() const;
+
+    //! constant memory available on device in bytes
+    size_t totalConstMem() const;
+
+    //! major compute capability
+    int majorVersion() const;
+
+    //! minor compute capability
+    int minorVersion() const;
+
+    //! alignment requirement for textures
+    size_t textureAlignment() const;
+
+    //! pitch alignment requirement for texture references bound to pitched memory
+    size_t texturePitchAlignment() const;
+
+    //! number of multiprocessors on device
+    int multiProcessorCount() const;
+
+    //! specified whether there is a run time limit on kernels
+    bool kernelExecTimeoutEnabled() const;
+
+    //! device is integrated as opposed to discrete
+    bool integrated() const;
+
+    //! device can map host memory with cudaHostAlloc/cudaHostGetDevicePointer
+    bool canMapHostMemory() const;
+
+    enum ComputeMode
+    {
+        ComputeModeDefault,         /**< default compute mode (Multiple threads can use ::cudaSetDevice() with this device) */
+        ComputeModeExclusive,       /**< compute-exclusive-thread mode (Only one thread in one process will be able to use ::cudaSetDevice() with this device) */
+        ComputeModeProhibited,      /**< compute-prohibited mode (No threads can use ::cudaSetDevice() with this device) */
+        ComputeModeExclusiveProcess /**< compute-exclusive-process mode (Many threads in one process will be able to use ::cudaSetDevice() with this device) */
+    };
+
+    //! compute mode
+    ComputeMode computeMode() const;
+
+    //! maximum 1D texture size
+    int maxTexture1D() const;
+
+    //! maximum 1D mipmapped texture size
+    int maxTexture1DMipmap() const;
+
+    //! maximum size for 1D textures bound to linear memory
+    int maxTexture1DLinear() const;
+
+    //! maximum 2D texture dimensions
+    Vec2i maxTexture2D() const;
+
+    //! maximum 2D mipmapped texture dimensions
+    Vec2i maxTexture2DMipmap() const;
+
+    //! maximum dimensions (width, height, pitch) for 2D textures bound to pitched memory
+    Vec3i maxTexture2DLinear() const;
+
+    //! maximum 2D texture dimensions if texture gather operations have to be performed
+    Vec2i maxTexture2DGather() const;
+
+    //! maximum 3D texture dimensions
+    Vec3i maxTexture3D() const;
+
+    //! maximum Cubemap texture dimensions
+    int maxTextureCubemap() const;
+
+    //! maximum 1D layered texture dimensions
+    Vec2i maxTexture1DLayered() const;
+
+    //! maximum 2D layered texture dimensions
+    Vec3i maxTexture2DLayered() const;
+
+    //! maximum Cubemap layered texture dimensions
+    Vec2i maxTextureCubemapLayered() const;
+
+    //! maximum 1D surface size
+    int maxSurface1D() const;
+
+    //! maximum 2D surface dimensions
+    Vec2i maxSurface2D() const;
+
+    //! maximum 3D surface dimensions
+    Vec3i maxSurface3D() const;
+
+    //! maximum 1D layered surface dimensions
+    Vec2i maxSurface1DLayered() const;
+
+    //! maximum 2D layered surface dimensions
+    Vec3i maxSurface2DLayered() const;
+
+    //! maximum Cubemap surface dimensions
+    int maxSurfaceCubemap() const;
+
+    //! maximum Cubemap layered surface dimensions
+    Vec2i maxSurfaceCubemapLayered() const;
+
+    //! alignment requirements for surfaces
+    size_t surfaceAlignment() const;
+
+    //! device can possibly execute multiple kernels concurrently
+    bool concurrentKernels() const;
+
+    //! device has ECC support enabled
+    bool ECCEnabled() const;
+
+    //! PCI bus ID of the device
+    int pciBusID() const;
+
+    //! PCI device ID of the device
+    int pciDeviceID() const;
+
+    //! PCI domain ID of the device
+    int pciDomainID() const;
+
+    //! true if device is a Tesla device using TCC driver, false otherwise
+    bool tccDriver() const;
+
+    //! number of asynchronous engines
+    int asyncEngineCount() const;
+
+    //! device shares a unified address space with the host
+    bool unifiedAddressing() const;
+
+    //! peak memory clock frequency in kilohertz
+    int memoryClockRate() const;
+
+    //! global memory bus width in bits
+    int memoryBusWidth() const;
+
+    //! size of L2 cache in bytes
+    int l2CacheSize() const;
+
+    //! maximum resident threads per multiprocessor
+    int maxThreadsPerMultiProcessor() const;
+
+    //! gets free and total device memory
+    void queryMemory(size_t& totalMemory, size_t& freeMemory) const;
+    size_t freeMemory() const;
+    size_t totalMemory() const;
+
+    //! checks whether device supports the given feature
+    bool supports(FeatureSet feature_set) const;
+
+    //! checks whether the GPU module can be run on the given device
+    bool isCompatible() const;
+
+private:
+    int device_id_;
+};
+
+CV_EXPORTS void printCudaDeviceInfo(int device);
+CV_EXPORTS void printShortCudaDeviceInfo(int device);
+
+}} // namespace cv { namespace gpu {
+
+
+#include "opencv2/core/gpu.inl.hpp"
+
+#endif /* __OPENCV_CORE_GPU_HPP__ */
diff --git a/modules/core/include/opencv2/core/gpu.inl.hpp b/modules/core/include/opencv2/core/gpu.inl.hpp
new file mode 100644
index 0000000000..13861170cb
--- /dev/null
+++ b/modules/core/include/opencv2/core/gpu.inl.hpp
@@ -0,0 +1,605 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_GPUINL_HPP__
+#define __OPENCV_CORE_GPUINL_HPP__
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpu {
+
+//////////////////////////////// GpuMat ///////////////////////////////
+
+inline
+GpuMat::GpuMat()
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_, Scalar s_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{
+    if (rows_ > 0 && cols_ > 0)
+    {
+        create(rows_, cols_, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_, Scalar s_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{
+    if (size_.height > 0 && size_.width > 0)
+    {
+        create(size_.height, size_.width, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(const GpuMat& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend)
+{
+    if (refcount)
+        CV_XADD(refcount, 1);
+}
+
+inline
+GpuMat::GpuMat(InputArray arr) :
+    flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
+{
+    upload(arr);
+}
+
+inline
+GpuMat::~GpuMat()
+{
+    release();
+}
+
+inline
+GpuMat& GpuMat::operator =(const GpuMat& m)
+{
+    if (this != &m)
+    {
+        GpuMat temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void GpuMat::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+void GpuMat::swap(GpuMat& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+}
+
+inline
+GpuMat GpuMat::clone() const
+{
+    GpuMat m;
+    copyTo(m);
+    return m;
+}
+
+inline
+void GpuMat::copyTo(OutputArray dst, InputArray mask) const
+{
+    copyTo(dst, mask, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s)
+{
+    return setTo(s, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s, InputArray mask)
+{
+    return setTo(s, mask, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype) const
+{
+    convertTo(dst, rtype, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, double beta) const
+{
+    convertTo(dst, rtype, alpha, beta, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, Stream& stream) const
+{
+    convertTo(dst, rtype, alpha, 0.0, stream);
+}
+
+inline
+void GpuMat::assignTo(GpuMat& m, int _type) const
+{
+    if (_type < 0)
+        m = *this;
+    else
+        convertTo(m, _type);
+}
+
+inline
+uchar* GpuMat::ptr(int y)
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+inline
+const uchar* GpuMat::ptr(int y) const
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+template<typename _Tp> inline
+_Tp* GpuMat::ptr(int y)
+{
+    return (_Tp*)ptr(y);
+}
+
+template<typename _Tp> inline
+const _Tp* GpuMat::ptr(int y) const
+{
+    return (const _Tp*)ptr(y);
+}
+
+template <class T> inline
+GpuMat::operator PtrStepSz<T>() const
+{
+    return PtrStepSz<T>(rows, cols, (T*)data, step);
+}
+
+template <class T> inline
+GpuMat::operator PtrStep<T>() const
+{
+    return PtrStep<T>((T*)data, step);
+}
+
+inline
+GpuMat GpuMat::row(int y) const
+{
+    return GpuMat(*this, Range(y, y+1), Range::all());
+}
+
+inline
+GpuMat GpuMat::col(int x) const
+{
+    return GpuMat(*this, Range::all(), Range(x, x+1));
+}
+
+inline
+GpuMat GpuMat::rowRange(int startrow, int endrow) const
+{
+    return GpuMat(*this, Range(startrow, endrow), Range::all());
+}
+
+inline
+GpuMat GpuMat::rowRange(Range r) const
+{
+    return GpuMat(*this, r, Range::all());
+}
+
+inline
+GpuMat GpuMat::colRange(int startcol, int endcol) const
+{
+    return GpuMat(*this, Range::all(), Range(startcol, endcol));
+}
+
+inline
+GpuMat GpuMat::colRange(Range r) const
+{
+    return GpuMat(*this, Range::all(), r);
+}
+
+inline
+GpuMat GpuMat::operator ()(Range rowRange_, Range colRange_) const
+{
+    return GpuMat(*this, rowRange_, colRange_);
+}
+
+inline
+GpuMat GpuMat::operator ()(Rect roi) const
+{
+    return GpuMat(*this, roi);
+}
+
+inline
+bool GpuMat::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t GpuMat::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t GpuMat::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int GpuMat::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int GpuMat::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int GpuMat::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t GpuMat::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size GpuMat::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool GpuMat::empty() const
+{
+    return data == 0;
+}
+
+static inline
+GpuMat createContinuous(int rows, int cols, int type)
+{
+    GpuMat m;
+    createContinuous(rows, cols, type, m);
+    return m;
+}
+
+static inline
+void createContinuous(Size size, int type, OutputArray arr)
+{
+    createContinuous(size.height, size.width, type, arr);
+}
+
+static inline
+GpuMat createContinuous(Size size, int type)
+{
+    GpuMat m;
+    createContinuous(size, type, m);
+    return m;
+}
+
+static inline
+void ensureSizeIsEnough(Size size, int type, OutputArray arr)
+{
+    ensureSizeIsEnough(size.height, size.width, type, arr);
+}
+
+static inline
+void swap(GpuMat& a, GpuMat& b)
+{
+    a.swap(b);
+}
+
+//////////////////////////////// CudaMem ////////////////////////////////
+
+inline
+CudaMem::CudaMem(AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+}
+
+inline
+CudaMem::CudaMem(const CudaMem& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), alloc_type(m.alloc_type)
+{
+    if( refcount )
+        CV_XADD(refcount, 1);
+}
+
+inline
+CudaMem::CudaMem(int rows_, int cols_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+CudaMem::CudaMem(Size size_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+CudaMem::CudaMem(InputArray arr, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    arr.getMat().copyTo(*this);
+}
+
+inline
+CudaMem::~CudaMem()
+{
+    release();
+}
+
+inline
+CudaMem& CudaMem::operator =(const CudaMem& m)
+{
+    if (this != &m)
+    {
+        CudaMem temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void CudaMem::swap(CudaMem& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+    std::swap(alloc_type, b.alloc_type);
+}
+
+inline
+CudaMem CudaMem::clone() const
+{
+    CudaMem m(size(), type(), alloc_type);
+    createMatHeader().copyTo(m);
+    return m;
+}
+
+inline
+void CudaMem::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+Mat CudaMem::createMatHeader() const
+{
+    return Mat(size(), type(), data, step);
+}
+
+inline
+bool CudaMem::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t CudaMem::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t CudaMem::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int CudaMem::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int CudaMem::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int CudaMem::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t CudaMem::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size CudaMem::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool CudaMem::empty() const
+{
+    return data == 0;
+}
+
+static inline
+void swap(CudaMem& a, CudaMem& b)
+{
+    a.swap(b);
+}
+
+//////////////////////////////// Stream ///////////////////////////////
+
+inline
+Stream::Stream(const Ptr<Impl>& impl)
+    : impl_(impl)
+{
+}
+
+//////////////////////////////// Initialization & Info ////////////////////////
+
+inline
+bool TargetArchs::has(int major, int minor)
+{
+    return hasPtx(major, minor) || hasBin(major, minor);
+}
+
+inline
+bool TargetArchs::hasEqualOrGreater(int major, int minor)
+{
+    return hasEqualOrGreaterPtx(major, minor) || hasEqualOrGreaterBin(major, minor);
+}
+
+inline
+DeviceInfo::DeviceInfo()
+{
+    device_id_ = getDevice();
+}
+
+inline
+DeviceInfo::DeviceInfo(int device_id)
+{
+    CV_Assert( device_id >= 0 && device_id < getCudaEnabledDeviceCount() );
+    device_id_ = device_id;
+}
+
+inline
+int DeviceInfo::deviceID() const
+{
+    return device_id_;
+}
+
+inline
+size_t DeviceInfo::freeMemory() const
+{
+    size_t _totalMemory, _freeMemory;
+    queryMemory(_totalMemory, _freeMemory);
+    return _freeMemory;
+}
+
+inline
+size_t DeviceInfo::totalMemory() const
+{
+    size_t _totalMemory, _freeMemory;
+    queryMemory(_totalMemory, _freeMemory);
+    return _totalMemory;
+}
+
+inline
+bool DeviceInfo::supports(FeatureSet feature_set) const
+{
+    int version = majorVersion() * 10 + minorVersion();
+    return version >= feature_set;
+}
+
+}} // namespace cv { namespace gpu {
+
+//////////////////////////////// Mat ////////////////////////////////
+
+namespace cv {
+
+inline
+Mat::Mat(const gpu::GpuMat& m)
+    : flags(0), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0), datalimit(0), allocator(0), size(&rows)
+{
+    m.download(*this);
+}
+
+}
+
+#endif // __OPENCV_CORE_GPUINL_HPP__
diff --git a/modules/core/include/opencv2/core/stream_accessor.hpp b/modules/core/include/opencv2/core/gpu_stream_accessor.hpp
similarity index 73%
rename from modules/core/include/opencv2/core/stream_accessor.hpp
rename to modules/core/include/opencv2/core/gpu_stream_accessor.hpp
index 9748bef80c..cf7d3c4316 100644
--- a/modules/core/include/opencv2/core/stream_accessor.hpp
+++ b/modules/core/include/opencv2/core/gpu_stream_accessor.hpp
@@ -22,7 +22,7 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
@@ -40,25 +40,38 @@
 //
 //M*/
 
-#ifndef __OPENCV_CUDA_STREAM_ACCESSOR_HPP__
-#define __OPENCV_CUDA_STREAM_ACCESSOR_HPP__
+#ifndef __OPENCV_CORE_GPU_STREAM_ACCESSOR_HPP__
+#define __OPENCV_CORE_GPU_STREAM_ACCESSOR_HPP__
 
-#include "opencv2/core/gpumat.hpp"
-#include "cuda_runtime_api.h"
+#ifndef __cplusplus
+#  error gpu_stream_accessor.hpp header must be compiled as C++
+#endif
+
+// This is only header file that depends on Cuda. All other headers are independent.
+// So if you use OpenCV binaries you do noot need to install Cuda Toolkit.
+// But of you wanna use GPU by yourself, may get cuda stream instance using the class below.
+// In this case you have to install Cuda Toolkit.
+
+#include <cuda_runtime.h>
+#include "opencv2/core/cvdef.h"
 
 namespace cv
 {
     namespace gpu
     {
-        // This is only header file that depends on Cuda. All other headers are independent.
-        // So if you use OpenCV binaries you do noot need to install Cuda Toolkit.
-        // But of you wanna use GPU by yourself, may get cuda stream instance using the class below.
-        // In this case you have to install Cuda Toolkit.
+        class Stream;
+        class Event;
+
         struct StreamAccessor
         {
             CV_EXPORTS static cudaStream_t getStream(const Stream& stream);
         };
+
+        struct EventAccessor
+        {
+            CV_EXPORTS static cudaEvent_t getEvent(const Event& event);
+        };
     }
 }
 
-#endif /* __OPENCV_CUDA_STREAM_ACCESSOR_HPP__ */
\ No newline at end of file
+#endif /* __OPENCV_CORE_GPU_STREAM_ACCESSOR_HPP__ */
diff --git a/modules/core/include/opencv2/core/cuda_devptrs.hpp b/modules/core/include/opencv2/core/gpu_types.hpp
similarity index 69%
rename from modules/core/include/opencv2/core/cuda_devptrs.hpp
rename to modules/core/include/opencv2/core/gpu_types.hpp
index 26fc2403f9..4997034cca 100644
--- a/modules/core/include/opencv2/core/cuda_devptrs.hpp
+++ b/modules/core/include/opencv2/core/gpu_types.hpp
@@ -22,7 +22,7 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
@@ -40,10 +40,12 @@
 //
 //M*/
 
-#ifndef __OPENCV_CORE_DEVPTRS_HPP__
-#define __OPENCV_CORE_DEVPTRS_HPP__
+#ifndef __OPENCV_CORE_GPU_TYPES_HPP__
+#define __OPENCV_CORE_GPU_TYPES_HPP__
 
-#ifdef __cplusplus
+#ifndef __cplusplus
+#  error gpu_types.hpp header must be compiled as C++
+#endif
 
 #ifdef __CUDACC__
     #define __CV_GPU_HOST_DEVICE__ __host__ __device__ __forceinline__
@@ -58,10 +60,7 @@ namespace cv
         // Simple lightweight structures that encapsulates information about an image on device.
         // It is intended to pass to nvcc-compiled code. GpuMat depends on headers that nvcc can't compile
 
-        template <bool expr> struct StaticAssert;
-        template <> struct StaticAssert<true> {static __CV_GPU_HOST_DEVICE__ void check(){}};
-
-        template<typename T> struct DevPtr
+        template <typename T> struct DevPtr
         {
             typedef T elem_type;
             typedef int index_type;
@@ -78,7 +77,7 @@ namespace cv
             __CV_GPU_HOST_DEVICE__ operator const T*() const { return data; }
         };
 
-        template<typename T> struct PtrSz : public DevPtr<T>
+        template <typename T> struct PtrSz : public DevPtr<T>
         {
             __CV_GPU_HOST_DEVICE__ PtrSz() : size(0) {}
             __CV_GPU_HOST_DEVICE__ PtrSz(T* data_, size_t size_) : DevPtr<T>(data_), size(size_) {}
@@ -86,12 +85,12 @@ namespace cv
             size_t size;
         };
 
-        template<typename T> struct PtrStep : public DevPtr<T>
+        template <typename T> struct PtrStep : public DevPtr<T>
         {
             __CV_GPU_HOST_DEVICE__ PtrStep() : step(0) {}
             __CV_GPU_HOST_DEVICE__ PtrStep(T* data_, size_t step_) : DevPtr<T>(data_), step(step_) {}
 
-            /** \brief stride between two consecutive rows in bytes. Step is stored always and everywhere in bytes!!! */
+            //! stride between two consecutive rows in bytes. Step is stored always and everywhere in bytes!!!
             size_t step;
 
             __CV_GPU_HOST_DEVICE__       T* ptr(int y = 0)       { return (      T*)( (      char*)DevPtr<T>::data + y * step); }
@@ -121,54 +120,7 @@ namespace cv
         typedef PtrStep<unsigned char> PtrStepb;
         typedef PtrStep<float> PtrStepf;
         typedef PtrStep<int> PtrStepi;
-
-
-#if defined __GNUC__
-    #define __CV_GPU_DEPR_BEFORE__
-    #define __CV_GPU_DEPR_AFTER__ __attribute__ ((deprecated))
-#elif defined(__MSVC__) //|| defined(__CUDACC__)
-    #pragma deprecated(DevMem2D_)
-    #define __CV_GPU_DEPR_BEFORE__ __declspec(deprecated)
-    #define __CV_GPU_DEPR_AFTER__
-#else
-    #define __CV_GPU_DEPR_BEFORE__
-    #define __CV_GPU_DEPR_AFTER__
-#endif
-
-        template <typename T> struct __CV_GPU_DEPR_BEFORE__ DevMem2D_ : public PtrStepSz<T>
-        {
-            DevMem2D_() {}
-            DevMem2D_(int rows_, int cols_, T* data_, size_t step_) : PtrStepSz<T>(rows_, cols_, data_, step_) {}
-
-            template <typename U>
-            explicit __CV_GPU_DEPR_BEFORE__ DevMem2D_(const DevMem2D_<U>& d) : PtrStepSz<T>(d.rows, d.cols, (T*)d.data, d.step) {}
-        } __CV_GPU_DEPR_AFTER__ ;
-
-        typedef DevMem2D_<unsigned char> DevMem2Db;
-        typedef DevMem2Db DevMem2D;
-        typedef DevMem2D_<float> DevMem2Df;
-        typedef DevMem2D_<int> DevMem2Di;
-
-//#undef __CV_GPU_DEPR_BEFORE__
-//#undef __CV_GPU_DEPR_AFTER__
-
-        namespace device
-        {
-            using cv::gpu::PtrSz;
-            using cv::gpu::PtrStep;
-            using cv::gpu::PtrStepSz;
-
-            using cv::gpu::PtrStepSzb;
-            using cv::gpu::PtrStepSzf;
-            using cv::gpu::PtrStepSzi;
-
-            using cv::gpu::PtrStepb;
-            using cv::gpu::PtrStepf;
-            using cv::gpu::PtrStepi;
-        }
     }
 }
 
-#endif // __cplusplus
-
-#endif /* __OPENCV_CORE_DEVPTRS_HPP__ */
+#endif /* __OPENCV_CORE_GPU_TYPES_HPP__ */
diff --git a/modules/core/include/opencv2/core/gpumat.hpp b/modules/core/include/opencv2/core/gpumat.hpp
deleted file mode 100644
index bb436ec553..0000000000
--- a/modules/core/include/opencv2/core/gpumat.hpp
+++ /dev/null
@@ -1,691 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_GPUMAT_HPP__
-#define __OPENCV_GPUMAT_HPP__
-
-#ifdef __cplusplus
-
-#include "opencv2/core.hpp"
-#include "opencv2/core/cuda_devptrs.hpp"
-
-namespace cv { namespace gpu
-{
-    //////////////////////////////// CudaMem ////////////////////////////////
-    // CudaMem is limited cv::Mat with page locked memory allocation.
-    // Page locked memory is only needed for async and faster coping to GPU.
-    // It is convertable to cv::Mat header without reference counting
-    // so you can use it with other opencv functions.
-
-    // Page-locks the matrix m memory and maps it for the device(s)
-    CV_EXPORTS void registerPageLocked(Mat& m);
-    // Unmaps the memory of matrix m, and makes it pageable again.
-    CV_EXPORTS void unregisterPageLocked(Mat& m);
-
-    class CV_EXPORTS CudaMem
-    {
-    public:
-        enum  { ALLOC_PAGE_LOCKED = 1, ALLOC_ZEROCOPY = 2, ALLOC_WRITE_COMBINED = 4 };
-
-        CudaMem();
-        CudaMem(const CudaMem& m);
-
-        CudaMem(int rows, int cols, int type, int _alloc_type = ALLOC_PAGE_LOCKED);
-        CudaMem(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
-
-
-        //! creates from cv::Mat with coping data
-        explicit CudaMem(const Mat& m, int alloc_type = ALLOC_PAGE_LOCKED);
-
-        ~CudaMem();
-
-        CudaMem& operator = (const CudaMem& m);
-
-        //! returns deep copy of the matrix, i.e. the data is copied
-        CudaMem clone() const;
-
-        //! allocates new matrix data unless the matrix already has specified size and type.
-        void create(int rows, int cols, int type, int alloc_type = ALLOC_PAGE_LOCKED);
-        void create(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
-
-        //! decrements reference counter and released memory if needed.
-        void release();
-
-        //! returns matrix header with disabled reference counting for CudaMem data.
-        Mat createMatHeader() const;
-        operator Mat() const;
-
-        //! maps host memory into device address space and returns GpuMat header for it. Throws exception if not supported by hardware.
-        GpuMat createGpuMatHeader() const;
-        operator GpuMat() const;
-
-        //returns if host memory can be mapperd to gpu address space;
-        static bool canMapHostMemory();
-
-        // Please see cv::Mat for descriptions
-        bool isContinuous() const;
-        size_t elemSize() const;
-        size_t elemSize1() const;
-        int type() const;
-        int depth() const;
-        int channels() const;
-        size_t step1() const;
-        Size size() const;
-        bool empty() const;
-
-
-        // Please see cv::Mat for descriptions
-        int flags;
-        int rows, cols;
-        size_t step;
-
-        uchar* data;
-        int* refcount;
-
-        uchar* datastart;
-        uchar* dataend;
-
-        int alloc_type;
-    };
-
-
-    //////////////////////////////// CudaStream ////////////////////////////////
-    // Encapculates Cuda Stream. Provides interface for async coping.
-    // Passed to each function that supports async kernel execution.
-    // Reference counting is enabled
-
-    class CV_EXPORTS Stream
-    {
-    public:
-        Stream();
-        ~Stream();
-
-        Stream(const Stream&);
-        Stream& operator =(const Stream&);
-
-        bool queryIfComplete();
-        void waitForCompletion();
-
-        //! downloads asynchronously
-        // Warning! cv::Mat must point to page locked memory (i.e. to CudaMem data or to its subMat)
-        void enqueueDownload(const GpuMat& src, CudaMem& dst);
-        void enqueueDownload(const GpuMat& src, Mat& dst);
-
-        //! uploads asynchronously
-        // Warning! cv::Mat must point to page locked memory (i.e. to CudaMem data or to its ROI)
-        void enqueueUpload(const CudaMem& src, GpuMat& dst);
-        void enqueueUpload(const Mat& src, GpuMat& dst);
-
-        //! copy asynchronously
-        void enqueueCopy(const GpuMat& src, GpuMat& dst);
-
-        //! memory set asynchronously
-        void enqueueMemSet(GpuMat& src, Scalar val);
-        void enqueueMemSet(GpuMat& src, Scalar val, const GpuMat& mask);
-
-        //! converts matrix type, ex from float to uchar depending on type
-        void enqueueConvert(const GpuMat& src, GpuMat& dst, int dtype, double a = 1, double b = 0);
-
-        //! adds a callback to be called on the host after all currently enqueued items in the stream have completed
-        typedef void (*StreamCallback)(Stream& stream, int status, void* userData);
-        void enqueueHostCallback(StreamCallback callback, void* userData);
-
-        static Stream& Null();
-
-        operator bool() const;
-
-    private:
-        struct Impl;
-
-        explicit Stream(Impl* impl);
-        void create();
-        void release();
-
-        Impl *impl;
-
-        friend struct StreamAccessor;
-    };
-
-    //////////////////////////////// Initialization & Info ////////////////////////
-
-    //! This is the only function that do not throw exceptions if the library is compiled without Cuda.
-    CV_EXPORTS int getCudaEnabledDeviceCount();
-
-    //! Functions below throw cv::Expception if the library is compiled without Cuda.
-
-    CV_EXPORTS void setDevice(int device);
-    CV_EXPORTS int getDevice();
-
-    //! Explicitly destroys and cleans up all resources associated with the current device in the current process.
-    //! Any subsequent API call to this device will reinitialize the device.
-    CV_EXPORTS void resetDevice();
-
-    enum FeatureSet
-    {
-        FEATURE_SET_COMPUTE_10 = 10,
-        FEATURE_SET_COMPUTE_11 = 11,
-        FEATURE_SET_COMPUTE_12 = 12,
-        FEATURE_SET_COMPUTE_13 = 13,
-        FEATURE_SET_COMPUTE_20 = 20,
-        FEATURE_SET_COMPUTE_21 = 21,
-        FEATURE_SET_COMPUTE_30 = 30,
-        FEATURE_SET_COMPUTE_35 = 35,
-
-        GLOBAL_ATOMICS = FEATURE_SET_COMPUTE_11,
-        SHARED_ATOMICS = FEATURE_SET_COMPUTE_12,
-        NATIVE_DOUBLE = FEATURE_SET_COMPUTE_13,
-        WARP_SHUFFLE_FUNCTIONS = FEATURE_SET_COMPUTE_30,
-        DYNAMIC_PARALLELISM = FEATURE_SET_COMPUTE_35
-    };
-
-    // Checks whether current device supports the given feature
-    CV_EXPORTS bool deviceSupports(FeatureSet feature_set);
-
-    // Gives information about what GPU archs this OpenCV GPU module was
-    // compiled for
-    class CV_EXPORTS TargetArchs
-    {
-    public:
-        static bool builtWith(FeatureSet feature_set);
-        static bool has(int major, int minor);
-        static bool hasPtx(int major, int minor);
-        static bool hasBin(int major, int minor);
-        static bool hasEqualOrLessPtx(int major, int minor);
-        static bool hasEqualOrGreater(int major, int minor);
-        static bool hasEqualOrGreaterPtx(int major, int minor);
-        static bool hasEqualOrGreaterBin(int major, int minor);
-    private:
-        TargetArchs();
-    };
-
-    // Gives information about the given GPU
-    class CV_EXPORTS DeviceInfo
-    {
-    public:
-        // Creates DeviceInfo object for the current GPU
-        DeviceInfo() : device_id_(getDevice()) { query(); }
-
-        // Creates DeviceInfo object for the given GPU
-        DeviceInfo(int device_id) : device_id_(device_id) { query(); }
-
-        std::string name() const { return name_; }
-
-        // Return compute capability versions
-        int majorVersion() const { return majorVersion_; }
-        int minorVersion() const { return minorVersion_; }
-
-        int multiProcessorCount() const { return multi_processor_count_; }
-
-        size_t sharedMemPerBlock() const;
-
-        void queryMemory(size_t& totalMemory, size_t& freeMemory) const;
-        size_t freeMemory() const;
-        size_t totalMemory() const;
-
-        // Checks whether device supports the given feature
-        bool supports(FeatureSet feature_set) const;
-
-        // Checks whether the GPU module can be run on the given device
-        bool isCompatible() const;
-
-        int deviceID() const { return device_id_; }
-
-    private:
-        void query();
-
-        int device_id_;
-
-        std::string name_;
-        int multi_processor_count_;
-        int majorVersion_;
-        int minorVersion_;
-    };
-
-    CV_EXPORTS void printCudaDeviceInfo(int device);
-    CV_EXPORTS void printShortCudaDeviceInfo(int device);
-
-    //////////////////////////////// GpuMat ///////////////////////////////
-
-    //! Smart pointer for GPU memory with reference counting. Its interface is mostly similar with cv::Mat.
-    class CV_EXPORTS GpuMat
-    {
-    public:
-        //! default constructor
-        GpuMat();
-
-        //! constructs GpuMatrix of the specified size and type (_type is CV_8UC1, CV_64FC3, CV_32SC(12) etc.)
-        GpuMat(int rows, int cols, int type);
-        GpuMat(Size size, int type);
-
-        //! constucts GpuMatrix and fills it with the specified value _s.
-        GpuMat(int rows, int cols, int type, Scalar s);
-        GpuMat(Size size, int type, Scalar s);
-
-        //! copy constructor
-        GpuMat(const GpuMat& m);
-
-        //! constructor for GpuMatrix headers pointing to user-allocated data
-        GpuMat(int rows, int cols, int type, void* data, size_t step = Mat::AUTO_STEP);
-        GpuMat(Size size, int type, void* data, size_t step = Mat::AUTO_STEP);
-
-        //! creates a matrix header for a part of the bigger matrix
-        GpuMat(const GpuMat& m, Range rowRange, Range colRange);
-        GpuMat(const GpuMat& m, Rect roi);
-
-        //! builds GpuMat from Mat. Perfom blocking upload to device.
-        explicit GpuMat(const Mat& m);
-
-        //! destructor - calls release()
-        ~GpuMat();
-
-        //! assignment operators
-        GpuMat& operator = (const GpuMat& m);
-
-        //! pefroms blocking upload data to GpuMat.
-        void upload(const Mat& m);
-
-        //! downloads data from device to host memory. Blocking calls.
-        void download(Mat& m) const;
-
-        //! returns a new GpuMatrix header for the specified row
-        GpuMat row(int y) const;
-        //! returns a new GpuMatrix header for the specified column
-        GpuMat col(int x) const;
-        //! ... for the specified row span
-        GpuMat rowRange(int startrow, int endrow) const;
-        GpuMat rowRange(Range r) const;
-        //! ... for the specified column span
-        GpuMat colRange(int startcol, int endcol) const;
-        GpuMat colRange(Range r) const;
-
-        //! returns deep copy of the GpuMatrix, i.e. the data is copied
-        GpuMat clone() const;
-        //! copies the GpuMatrix content to "m".
-        // It calls m.create(this->size(), this->type()).
-        void copyTo(GpuMat& m) const;
-        //! copies those GpuMatrix elements to "m" that are marked with non-zero mask elements.
-        void copyTo(GpuMat& m, const GpuMat& mask) const;
-        //! converts GpuMatrix to another datatype with optional scalng. See cvConvertScale.
-        void convertTo(GpuMat& m, int rtype, double alpha = 1, double beta = 0) const;
-
-        void assignTo(GpuMat& m, int type=-1) const;
-
-        //! sets every GpuMatrix element to s
-        GpuMat& operator = (Scalar s);
-        //! sets some of the GpuMatrix elements to s, according to the mask
-        GpuMat& setTo(Scalar s, const GpuMat& mask = GpuMat());
-        //! creates alternative GpuMatrix header for the same data, with different
-        // number of channels and/or different number of rows. see cvReshape.
-        GpuMat reshape(int cn, int rows = 0) const;
-
-        //! allocates new GpuMatrix data unless the GpuMatrix already has specified size and type.
-        // previous data is unreferenced if needed.
-        void create(int rows, int cols, int type);
-        void create(Size size, int type);
-        //! decreases reference counter;
-        // deallocate the data when reference counter reaches 0.
-        void release();
-
-        //! swaps with other smart pointer
-        void swap(GpuMat& mat);
-
-        //! locates GpuMatrix header within a parent GpuMatrix. See below
-        void locateROI(Size& wholeSize, Point& ofs) const;
-        //! moves/resizes the current GpuMatrix ROI inside the parent GpuMatrix.
-        GpuMat& adjustROI(int dtop, int dbottom, int dleft, int dright);
-        //! extracts a rectangular sub-GpuMatrix
-        // (this is a generalized form of row, rowRange etc.)
-        GpuMat operator()(Range rowRange, Range colRange) const;
-        GpuMat operator()(Rect roi) const;
-
-        //! returns true iff the GpuMatrix data is continuous
-        // (i.e. when there are no gaps between successive rows).
-        // similar to CV_IS_GpuMat_CONT(cvGpuMat->type)
-        bool isContinuous() const;
-        //! returns element size in bytes,
-        // similar to CV_ELEM_SIZE(cvMat->type)
-        size_t elemSize() const;
-        //! returns the size of element channel in bytes.
-        size_t elemSize1() const;
-        //! returns element type, similar to CV_MAT_TYPE(cvMat->type)
-        int type() const;
-        //! returns element type, similar to CV_MAT_DEPTH(cvMat->type)
-        int depth() const;
-        //! returns element type, similar to CV_MAT_CN(cvMat->type)
-        int channels() const;
-        //! returns step/elemSize1()
-        size_t step1() const;
-        //! returns GpuMatrix size:
-        // width == number of columns, height == number of rows
-        Size size() const;
-        //! returns true if GpuMatrix data is NULL
-        bool empty() const;
-
-        //! returns pointer to y-th row
-        uchar* ptr(int y = 0);
-        const uchar* ptr(int y = 0) const;
-
-        //! template version of the above method
-        template<typename _Tp> _Tp* ptr(int y = 0);
-        template<typename _Tp> const _Tp* ptr(int y = 0) const;
-
-        template <typename _Tp> operator PtrStepSz<_Tp>() const;
-        template <typename _Tp> operator PtrStep<_Tp>() const;
-
-        // Deprecated function
-        __CV_GPU_DEPR_BEFORE__ template <typename _Tp> operator DevMem2D_<_Tp>() const __CV_GPU_DEPR_AFTER__;
-        #undef __CV_GPU_DEPR_BEFORE__
-        #undef __CV_GPU_DEPR_AFTER__
-
-        /*! includes several bit-fields:
-        - the magic signature
-        - continuity flag
-        - depth
-        - number of channels
-        */
-        int flags;
-
-        //! the number of rows and columns
-        int rows, cols;
-
-        //! a distance between successive rows in bytes; includes the gap if any
-        size_t step;
-
-        //! pointer to the data
-        uchar* data;
-
-        //! pointer to the reference counter;
-        // when GpuMatrix points to user-allocated data, the pointer is NULL
-        int* refcount;
-
-        //! helper fields used in locateROI and adjustROI
-        uchar* datastart;
-        uchar* dataend;
-    };
-
-    //! Creates continuous GPU matrix
-    CV_EXPORTS void createContinuous(int rows, int cols, int type, GpuMat& m);
-    CV_EXPORTS GpuMat createContinuous(int rows, int cols, int type);
-    CV_EXPORTS void createContinuous(Size size, int type, GpuMat& m);
-    CV_EXPORTS GpuMat createContinuous(Size size, int type);
-
-    //! Ensures that size of the given matrix is not less than (rows, cols) size
-    //! and matrix type is match specified one too
-    CV_EXPORTS void ensureSizeIsEnough(int rows, int cols, int type, GpuMat& m);
-    CV_EXPORTS void ensureSizeIsEnough(Size size, int type, GpuMat& m);
-
-    CV_EXPORTS GpuMat allocMatFromBuf(int rows, int cols, int type, GpuMat &mat);
-
-    ////////////////////////////////////////////////////////////////////////
-    // Error handling
-
-    CV_EXPORTS void error(const char* error_string, const char* file, const int line, const char* func = "");
-
-    ////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////////////////////////////////////////////
-
-    inline GpuMat::GpuMat()
-        : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-    {
-    }
-
-    inline GpuMat::GpuMat(int rows_, int cols_, int type_)
-        : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-    {
-        if (rows_ > 0 && cols_ > 0)
-            create(rows_, cols_, type_);
-    }
-
-    inline GpuMat::GpuMat(Size size_, int type_)
-        : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-    {
-        if (size_.height > 0 && size_.width > 0)
-            create(size_.height, size_.width, type_);
-    }
-
-    inline GpuMat::GpuMat(int rows_, int cols_, int type_, Scalar s_)
-        : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-    {
-        if (rows_ > 0 && cols_ > 0)
-        {
-            create(rows_, cols_, type_);
-            setTo(s_);
-        }
-    }
-
-    inline GpuMat::GpuMat(Size size_, int type_, Scalar s_)
-        : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-    {
-        if (size_.height > 0 && size_.width > 0)
-        {
-            create(size_.height, size_.width, type_);
-            setTo(s_);
-        }
-    }
-
-    inline GpuMat::~GpuMat()
-    {
-        release();
-    }
-
-    inline GpuMat GpuMat::clone() const
-    {
-        GpuMat m;
-        copyTo(m);
-        return m;
-    }
-
-    inline void GpuMat::assignTo(GpuMat& m, int _type) const
-    {
-        if (_type < 0)
-            m = *this;
-        else
-            convertTo(m, _type);
-    }
-
-    inline size_t GpuMat::step1() const
-    {
-        return step / elemSize1();
-    }
-
-    inline bool GpuMat::empty() const
-    {
-        return data == 0;
-    }
-
-    template<typename _Tp> inline _Tp* GpuMat::ptr(int y)
-    {
-        return (_Tp*)ptr(y);
-    }
-
-    template<typename _Tp> inline const _Tp* GpuMat::ptr(int y) const
-    {
-        return (const _Tp*)ptr(y);
-    }
-
-    inline void swap(GpuMat& a, GpuMat& b)
-    {
-        a.swap(b);
-    }
-
-    inline GpuMat GpuMat::row(int y) const
-    {
-        return GpuMat(*this, Range(y, y+1), Range::all());
-    }
-
-    inline GpuMat GpuMat::col(int x) const
-    {
-        return GpuMat(*this, Range::all(), Range(x, x+1));
-    }
-
-    inline GpuMat GpuMat::rowRange(int startrow, int endrow) const
-    {
-        return GpuMat(*this, Range(startrow, endrow), Range::all());
-    }
-
-    inline GpuMat GpuMat::rowRange(Range r) const
-    {
-        return GpuMat(*this, r, Range::all());
-    }
-
-    inline GpuMat GpuMat::colRange(int startcol, int endcol) const
-    {
-        return GpuMat(*this, Range::all(), Range(startcol, endcol));
-    }
-
-    inline GpuMat GpuMat::colRange(Range r) const
-    {
-        return GpuMat(*this, Range::all(), r);
-    }
-
-    inline void GpuMat::create(Size size_, int type_)
-    {
-        create(size_.height, size_.width, type_);
-    }
-
-    inline GpuMat GpuMat::operator()(Range _rowRange, Range _colRange) const
-    {
-        return GpuMat(*this, _rowRange, _colRange);
-    }
-
-    inline GpuMat GpuMat::operator()(Rect roi) const
-    {
-        return GpuMat(*this, roi);
-    }
-
-    inline bool GpuMat::isContinuous() const
-    {
-        return (flags & Mat::CONTINUOUS_FLAG) != 0;
-    }
-
-    inline size_t GpuMat::elemSize() const
-    {
-        return CV_ELEM_SIZE(flags);
-    }
-
-    inline size_t GpuMat::elemSize1() const
-    {
-        return CV_ELEM_SIZE1(flags);
-    }
-
-    inline int GpuMat::type() const
-    {
-        return CV_MAT_TYPE(flags);
-    }
-
-    inline int GpuMat::depth() const
-    {
-        return CV_MAT_DEPTH(flags);
-    }
-
-    inline int GpuMat::channels() const
-    {
-        return CV_MAT_CN(flags);
-    }
-
-    inline Size GpuMat::size() const
-    {
-        return Size(cols, rows);
-    }
-
-    inline uchar* GpuMat::ptr(int y)
-    {
-        CV_DbgAssert((unsigned)y < (unsigned)rows);
-        return data + step * y;
-    }
-
-    inline const uchar* GpuMat::ptr(int y) const
-    {
-        CV_DbgAssert((unsigned)y < (unsigned)rows);
-        return data + step * y;
-    }
-
-    inline GpuMat& GpuMat::operator = (Scalar s)
-    {
-        setTo(s);
-        return *this;
-    }
-
-    template <class T> inline GpuMat::operator PtrStepSz<T>() const
-    {
-        return PtrStepSz<T>(rows, cols, (T*)data, step);
-    }
-
-    template <class T> inline GpuMat::operator PtrStep<T>() const
-    {
-        return PtrStep<T>((T*)data, step);
-    }
-
-    template <class T> inline GpuMat::operator DevMem2D_<T>() const
-    {
-        return DevMem2D_<T>(rows, cols, (T*)data, step);
-    }
-
-    inline GpuMat createContinuous(int rows, int cols, int type)
-    {
-        GpuMat m;
-        createContinuous(rows, cols, type, m);
-        return m;
-    }
-
-    inline void createContinuous(Size size, int type, GpuMat& m)
-    {
-        createContinuous(size.height, size.width, type, m);
-    }
-
-    inline GpuMat createContinuous(Size size, int type)
-    {
-        GpuMat m;
-        createContinuous(size, type, m);
-        return m;
-    }
-
-    inline void ensureSizeIsEnough(Size size, int type, GpuMat& m)
-    {
-        ensureSizeIsEnough(size.height, size.width, type, m);
-    }
-}}
-
-#endif // __cplusplus
-
-#endif // __OPENCV_GPUMAT_HPP__
diff --git a/modules/core/include/opencv2/core/internal.hpp b/modules/core/include/opencv2/core/internal.hpp
deleted file mode 100644
index 1faa0944a5..0000000000
--- a/modules/core/include/opencv2/core/internal.hpp
+++ /dev/null
@@ -1,763 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-/* The header is for internal use and it is likely to change.
-   It contains some macro definitions that are used in cxcore, cv, cvaux
-   and, probably, other libraries. If you need some of this functionality,
-   the safe way is to copy it into your code and rename the macros.
-*/
-#ifndef __OPENCV_CORE_INTERNAL_HPP__
-#define __OPENCV_CORE_INTERNAL_HPP__
-
-#include <vector>
-
-#if defined WIN32 || defined _WIN32
-#  ifndef WIN32
-#    define WIN32
-#  endif
-#  ifndef _WIN32
-#    define _WIN32
-#  endif
-#endif
-
-#if !defined WIN32 && !defined WINCE
-#  include <pthread.h>
-#endif
-
-#ifdef __BORLANDC__
-#  ifndef WIN32
-#    define WIN32
-#  endif
-#  ifndef _WIN32
-#    define _WIN32
-#  endif
-#  define CV_DLL
-#  undef _CV_ALWAYS_PROFILE_
-#  define _CV_ALWAYS_NO_PROFILE_
-#endif
-
-#ifndef FALSE
-#  define FALSE 0
-#endif
-#ifndef TRUE
-#  define TRUE 1
-#endif
-
-#define __BEGIN__ __CV_BEGIN__
-#define __END__ __CV_END__
-#define EXIT __CV_EXIT__
-
-#ifdef HAVE_IPP
-#  include "ipp.h"
-
-CV_INLINE IppiSize ippiSize(int width, int height)
-{
-    IppiSize size = { width, height };
-    return size;
-}
-#endif
-
-#ifndef IPPI_CALL
-#  define IPPI_CALL(func) CV_Assert((func) >= 0)
-#endif
-
-#if defined __SSE2__ || defined _M_X64  || (defined _M_IX86_FP && _M_IX86_FP >= 2)
-#  include "emmintrin.h"
-#  define CV_SSE 1
-#  define CV_SSE2 1
-#  if defined __SSE3__ || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include "pmmintrin.h"
-#    define CV_SSE3 1
-#  endif
-#  if defined __SSSE3__  || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include "tmmintrin.h"
-#    define CV_SSSE3 1
-#  endif
-#  if defined __SSE4_1__ || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include <smmintrin.h>
-#    define CV_SSE4_1 1
-#  endif
-#  if defined __SSE4_2__ || (defined _MSC_VER && _MSC_VER >= 1500)
-#    include <nmmintrin.h>
-#    define CV_SSE4_2 1
-#  endif
-#  if defined __AVX__ || (defined _MSC_FULL_VER && _MSC_FULL_VER >= 160040219)
-// MS Visual Studio 2010 (2012?) has no macro pre-defined to identify the use of /arch:AVX
-// See: http://connect.microsoft.com/VisualStudio/feedback/details/605858/arch-avx-should-define-a-predefined-macro-in-x64-and-set-a-unique-value-for-m-ix86-fp-in-win32
-#    include <immintrin.h>
-#    define CV_AVX 1
-#    if defined(_XCR_XFEATURE_ENABLED_MASK)
-#      define __xgetbv() _xgetbv(_XCR_XFEATURE_ENABLED_MASK)
-#    else
-#      define __xgetbv() 0
-#    endif
-#  endif
-#endif
-
-#ifdef __ARM_NEON__
-#  include <arm_neon.h>
-#  define CV_NEON 1
-#  define CPU_HAS_NEON_FEATURE (true)
-#endif
-
-#ifndef CV_SSE
-#  define CV_SSE 0
-#endif
-#ifndef CV_SSE2
-#  define CV_SSE2 0
-#endif
-#ifndef CV_SSE3
-#  define CV_SSE3 0
-#endif
-#ifndef CV_SSSE3
-#  define CV_SSSE3 0
-#endif
-#ifndef CV_SSE4_1
-#  define CV_SSE4_1 0
-#endif
-#ifndef CV_SSE4_2
-#  define CV_SSE4_2 0
-#endif
-#ifndef CV_AVX
-#  define CV_AVX 0
-#endif
-#ifndef CV_NEON
-#  define CV_NEON 0
-#endif
-
-#ifdef HAVE_TBB
-#  include "tbb/tbb_stddef.h"
-#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
-#    include "tbb/tbb.h"
-#    include "tbb/task.h"
-#    undef min
-#    undef max
-#  else
-#    undef HAVE_TBB
-#  endif
-#endif
-
-#ifdef HAVE_EIGEN
-#  if defined __GNUC__ && defined __APPLE__
-#    pragma GCC diagnostic ignored "-Wshadow"
-#  endif
-#  include <Eigen/Core>
-#  include "opencv2/core/eigen.hpp"
-#endif
-
-#ifdef __cplusplus
-
-namespace cv
-{
-#ifdef HAVE_TBB
-
-    typedef tbb::blocked_range<int> BlockedRange;
-
-    template<typename Body> static inline
-    void parallel_for( const BlockedRange& range, const Body& body )
-    {
-        tbb::parallel_for(range, body);
-    }
-
-    template<typename Iterator, typename Body> static inline
-    void parallel_do( Iterator first, Iterator last, const Body& body )
-    {
-        tbb::parallel_do(first, last, body);
-    }
-
-    typedef tbb::split Split;
-
-    template<typename Body> static inline
-    void parallel_reduce( const BlockedRange& range, Body& body )
-    {
-        tbb::parallel_reduce(range, body);
-    }
-
-    typedef tbb::concurrent_vector<Rect> ConcurrentRectVector;
-    typedef tbb::concurrent_vector<double> ConcurrentDoubleVector;
-#else
-    class BlockedRange
-    {
-    public:
-        BlockedRange() : _begin(0), _end(0), _grainsize(0) {}
-        BlockedRange(int b, int e, int g=1) : _begin(b), _end(e), _grainsize(g) {}
-        int begin() const { return _begin; }
-        int end() const { return _end; }
-        int grainsize() const { return _grainsize; }
-
-    protected:
-        int _begin, _end, _grainsize;
-    };
-
-    template<typename Body> static inline
-    void parallel_for( const BlockedRange& range, const Body& body )
-    {
-        body(range);
-    }
-    typedef std::vector<Rect> ConcurrentRectVector;
-    typedef std::vector<double> ConcurrentDoubleVector;
-
-    template<typename Iterator, typename Body> static inline
-    void parallel_do( Iterator first, Iterator last, const Body& body )
-    {
-        for( ; first != last; ++first )
-            body(*first);
-    }
-
-    class Split {};
-
-    template<typename Body> static inline
-    void parallel_reduce( const BlockedRange& range, Body& body )
-    {
-        body(range);
-    }
-#endif
-} //namespace cv
-
-#define CV_INIT_ALGORITHM(classname, algname, memberinit) \
-    static ::cv::Algorithm* create##classname() \
-    { \
-        return new classname; \
-    } \
-    \
-    static ::cv::AlgorithmInfo& classname##_info() \
-    { \
-        static ::cv::AlgorithmInfo classname##_info_var(algname, create##classname); \
-        return classname##_info_var; \
-    } \
-    \
-    static ::cv::AlgorithmInfo& classname##_info_auto = classname##_info(); \
-    \
-    ::cv::AlgorithmInfo* classname::info() const \
-    { \
-        static volatile bool initialized = false; \
-        \
-        if( !initialized ) \
-        { \
-            initialized = true; \
-            classname obj; \
-            memberinit; \
-        } \
-        return &classname##_info(); \
-    }
-
-#endif //__cplusplus
-
-/* maximal size of vector to run matrix operations on it inline (i.e. w/o ipp calls) */
-#define  CV_MAX_INLINE_MAT_OP_SIZE  10
-
-/* maximal linear size of matrix to allocate it on stack. */
-#define  CV_MAX_LOCAL_MAT_SIZE  32
-
-/* maximal size of local memory storage */
-#define  CV_MAX_LOCAL_SIZE  \
-    (CV_MAX_LOCAL_MAT_SIZE*CV_MAX_LOCAL_MAT_SIZE*(int)sizeof(double))
-
-/* default image row align (in bytes) */
-#define  CV_DEFAULT_IMAGE_ROW_ALIGN  4
-
-/* matrices are continuous by default */
-#define  CV_DEFAULT_MAT_ROW_ALIGN  1
-
-/* maximum size of dynamic memory buffer.
-   cvAlloc reports an error if a larger block is requested. */
-#define  CV_MAX_ALLOC_SIZE    (((size_t)1 << (sizeof(size_t)*8-2)))
-
-/* the alignment of all the allocated buffers */
-#define  CV_MALLOC_ALIGN    16
-
-/* default alignment for dynamic data strucutures, resided in storages. */
-#define  CV_STRUCT_ALIGN    ((int)sizeof(double))
-
-/* default storage block size */
-#define  CV_STORAGE_BLOCK_SIZE   ((1<<16) - 128)
-
-/* default memory block for sparse array elements */
-#define  CV_SPARSE_MAT_BLOCK    (1<<12)
-
-/* initial hash table size */
-#define  CV_SPARSE_HASH_SIZE0    (1<<10)
-
-/* maximal average node_count/hash_size ratio beyond which hash table is resized */
-#define  CV_SPARSE_HASH_RATIO    3
-
-/* max length of strings */
-#define  CV_MAX_STRLEN  1024
-
-#if 0 /*def  CV_CHECK_FOR_NANS*/
-#  define CV_CHECK_NANS( arr ) cvCheckArray((arr))
-#else
-#  define CV_CHECK_NANS( arr )
-#endif
-
-/****************************************************************************************\
-*                                  Common declarations                                   *
-\****************************************************************************************/
-
-/* get alloca declaration */
-#ifdef __GNUC__
-#  undef alloca
-#  define alloca __builtin_alloca
-#  define CV_HAVE_ALLOCA 1
-#elif defined WIN32 || defined _WIN32 || \
-      defined WINCE || defined _MSC_VER || defined __BORLANDC__
-#  include <malloc.h>
-#  define CV_HAVE_ALLOCA 1
-#elif defined HAVE_ALLOCA_H
-#  include <alloca.h>
-#  define CV_HAVE_ALLOCA 1
-#elif defined HAVE_ALLOCA
-#  include <stdlib.h>
-#  define CV_HAVE_ALLOCA 1
-#else
-#  undef CV_HAVE_ALLOCA
-#endif
-
-#ifdef __GNUC__
-#  define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
-#elif defined _MSC_VER
-#  define CV_DECL_ALIGNED(x) __declspec(align(x))
-#else
-#  define CV_DECL_ALIGNED(x)
-#endif
-
-#if CV_HAVE_ALLOCA
-/* ! DO NOT make it an inline function */
-#  define cvStackAlloc(size) cvAlignPtr( alloca((size) + CV_MALLOC_ALIGN), CV_MALLOC_ALIGN )
-#endif
-
-#ifndef CV_IMPL
-#  define CV_IMPL CV_EXTERN_C
-#endif
-
-#define CV_DBG_BREAK() { volatile int* crashMe = 0; *crashMe = 0; }
-
-/* default step, set in case of continuous data
-   to work around checks for valid step in some ipp functions */
-#define  CV_STUB_STEP     (1 << 30)
-
-#define  CV_SIZEOF_FLOAT ((int)sizeof(float))
-#define  CV_SIZEOF_SHORT ((int)sizeof(short))
-
-#define  CV_ORIGIN_TL  0
-#define  CV_ORIGIN_BL  1
-
-/* IEEE754 constants and macros */
-#define  CV_POS_INF       0x7f800000
-#define  CV_NEG_INF       0x807fffff /* CV_TOGGLE_FLT(0xff800000) */
-#define  CV_1F            0x3f800000
-#define  CV_TOGGLE_FLT(x) ((x)^((int)(x) < 0 ? 0x7fffffff : 0))
-#define  CV_TOGGLE_DBL(x) \
-    ((x)^((int64)(x) < 0 ? CV_BIG_INT(0x7fffffffffffffff) : 0))
-
-#define  CV_NOP(a)      (a)
-#define  CV_ADD(a, b)   ((a) + (b))
-#define  CV_SUB(a, b)   ((a) - (b))
-#define  CV_MUL(a, b)   ((a) * (b))
-#define  CV_AND(a, b)   ((a) & (b))
-#define  CV_OR(a, b)    ((a) | (b))
-#define  CV_XOR(a, b)   ((a) ^ (b))
-#define  CV_ANDN(a, b)  (~(a) & (b))
-#define  CV_ORN(a, b)   (~(a) | (b))
-#define  CV_SQR(a)      ((a) * (a))
-
-#define  CV_LT(a, b)    ((a) < (b))
-#define  CV_LE(a, b)    ((a) <= (b))
-#define  CV_EQ(a, b)    ((a) == (b))
-#define  CV_NE(a, b)    ((a) != (b))
-#define  CV_GT(a, b)    ((a) > (b))
-#define  CV_GE(a, b)    ((a) >= (b))
-
-#define  CV_NONZERO(a)      ((a) != 0)
-#define  CV_NONZERO_FLT(a)  (((a)+(a)) != 0)
-
-/* general-purpose saturation macros */
-#define  CV_CAST_8U(t)  (uchar)(!((t) & ~255) ? (t) : (t) > 0 ? 255 : 0)
-#define  CV_CAST_8S(t)  (schar)(!(((t)+128) & ~255) ? (t) : (t) > 0 ? 127 : -128)
-#define  CV_CAST_16U(t) (ushort)(!((t) & ~65535) ? (t) : (t) > 0 ? 65535 : 0)
-#define  CV_CAST_16S(t) (short)(!(((t)+32768) & ~65535) ? (t) : (t) > 0 ? 32767 : -32768)
-#define  CV_CAST_32S(t) (int)(t)
-#define  CV_CAST_64S(t) (int64)(t)
-#define  CV_CAST_32F(t) (float)(t)
-#define  CV_CAST_64F(t) (double)(t)
-
-#define  CV_PASTE2(a,b) a##b
-#define  CV_PASTE(a,b)  CV_PASTE2(a,b)
-
-#define  CV_EMPTY
-#define  CV_MAKE_STR(a) #a
-
-#define  CV_ZERO_OBJ(x) memset((x), 0, sizeof(*(x)))
-
-#define  CV_DIM(static_array) ((int)(sizeof(static_array)/sizeof((static_array)[0])))
-
-#define  cvUnsupportedFormat "Unsupported format"
-
-CV_INLINE void* cvAlignPtr( const void* ptr, int align CV_DEFAULT(32) )
-{
-    assert( (align & (align-1)) == 0 );
-    return (void*)( ((size_t)ptr + align - 1) & ~(size_t)(align-1) );
-}
-
-CV_INLINE int cvAlign( int size, int align )
-{
-    assert( (align & (align-1)) == 0 && size < INT_MAX );
-    return (size + align - 1) & -align;
-}
-
-CV_INLINE  CvSize  cvGetMatSize( const CvMat* mat )
-{
-    CvSize size;
-    size.width = mat->cols;
-    size.height = mat->rows;
-    return size;
-}
-
-#define  CV_DESCALE(x,n)     (((x) + (1 << ((n)-1))) >> (n))
-#define  CV_FLT_TO_FIX(x,n)  cvRound((x)*(1<<(n)))
-
-/****************************************************************************************\
-
-  Generic implementation of QuickSort algorithm.
-  ----------------------------------------------
-  Using this macro user can declare customized sort function that can be much faster
-  than built-in qsort function because of lower overhead on elements
-  comparison and exchange. The macro takes less_than (or LT) argument - a macro or function
-  that takes 2 arguments returns non-zero if the first argument should be before the second
-  one in the sorted sequence and zero otherwise.
-
-  Example:
-
-    Suppose that the task is to sort points by ascending of y coordinates and if
-    y's are equal x's should ascend.
-
-    The code is:
-    ------------------------------------------------------------------------------
-           #define cmp_pts( pt1, pt2 ) \
-               ((pt1).y < (pt2).y || ((pt1).y < (pt2).y && (pt1).x < (pt2).x))
-
-           [static] CV_IMPLEMENT_QSORT( icvSortPoints, CvPoint, cmp_pts )
-    ------------------------------------------------------------------------------
-
-    After that the function "void icvSortPoints( CvPoint* array, size_t total, int aux );"
-    is available to user.
-
-  aux is an additional parameter, which can be used when comparing elements.
-  The current implementation was derived from *BSD system qsort():
-
-    * Copyright (c) 1992, 1993
-    *  The Regents of the University of California.  All rights reserved.
-    *
-    * Redistribution and use in source and binary forms, with or without
-    * modification, are permitted provided that the following conditions
-    * are met:
-    * 1. Redistributions of source code must retain the above copyright
-    *    notice, this list of conditions and the following disclaimer.
-    * 2. Redistributions in binary form must reproduce the above copyright
-    *    notice, this list of conditions and the following disclaimer in the
-    *    documentation and/or other materials provided with the distribution.
-    * 3. All advertising materials mentioning features or use of this software
-    *    must display the following acknowledgement:
-    *  This product includes software developed by the University of
-    *  California, Berkeley and its contributors.
-    * 4. Neither the name of the University nor the names of its contributors
-    *    may be used to endorse or promote products derived from this software
-    *    without specific prior written permission.
-    *
-    * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
-    * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-    * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
-    * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
-    * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-    * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
-    * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
-    * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
-    * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
-    * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
-    * SUCH DAMAGE.
-
-\****************************************************************************************/
-
-#define CV_IMPLEMENT_QSORT_EX( func_name, T, LT, user_data_type )                   \
-void func_name( T *array, size_t total, user_data_type aux )                        \
-{                                                                                   \
-    int isort_thresh = 7;                                                           \
-    T t;                                                                            \
-    int sp = 0;                                                                     \
-                                                                                    \
-    struct                                                                          \
-    {                                                                               \
-        T *lb;                                                                      \
-        T *ub;                                                                      \
-    }                                                                               \
-    stack[48];                                                                      \
-                                                                                    \
-    aux = aux;                                                                      \
-                                                                                    \
-    if( total <= 1 )                                                                \
-        return;                                                                     \
-                                                                                    \
-    stack[0].lb = array;                                                            \
-    stack[0].ub = array + (total - 1);                                              \
-                                                                                    \
-    while( sp >= 0 )                                                                \
-    {                                                                               \
-        T* left = stack[sp].lb;                                                     \
-        T* right = stack[sp--].ub;                                                  \
-                                                                                    \
-        for(;;)                                                                     \
-        {                                                                           \
-            int i, n = (int)(right - left) + 1, m;                                  \
-            T* ptr;                                                                 \
-            T* ptr2;                                                                \
-                                                                                    \
-            if( n <= isort_thresh )                                                 \
-            {                                                                       \
-            insert_sort:                                                            \
-                for( ptr = left + 1; ptr <= right; ptr++ )                          \
-                {                                                                   \
-                    for( ptr2 = ptr; ptr2 > left && LT(ptr2[0],ptr2[-1]); ptr2--)   \
-                        CV_SWAP( ptr2[0], ptr2[-1], t );                            \
-                }                                                                   \
-                break;                                                              \
-            }                                                                       \
-            else                                                                    \
-            {                                                                       \
-                T* left0;                                                           \
-                T* left1;                                                           \
-                T* right0;                                                          \
-                T* right1;                                                          \
-                T* pivot;                                                           \
-                T* a;                                                               \
-                T* b;                                                               \
-                T* c;                                                               \
-                int swap_cnt = 0;                                                   \
-                                                                                    \
-                left0 = left;                                                       \
-                right0 = right;                                                     \
-                pivot = left + (n/2);                                               \
-                                                                                    \
-                if( n > 40 )                                                        \
-                {                                                                   \
-                    int d = n / 8;                                                  \
-                    a = left, b = left + d, c = left + 2*d;                         \
-                    left = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))     \
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));    \
-                                                                                    \
-                    a = pivot - d, b = pivot, c = pivot + d;                        \
-                    pivot = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))    \
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));    \
-                                                                                    \
-                    a = right - 2*d, b = right - d, c = right;                      \
-                    right = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))    \
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));    \
-                }                                                                   \
-                                                                                    \
-                a = left, b = pivot, c = right;                                     \
-                pivot = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))        \
-                                   : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));       \
-                if( pivot != left0 )                                                \
-                {                                                                   \
-                    CV_SWAP( *pivot, *left0, t );                                   \
-                    pivot = left0;                                                  \
-                }                                                                   \
-                left = left1 = left0 + 1;                                           \
-                right = right1 = right0;                                            \
-                                                                                    \
-                for(;;)                                                             \
-                {                                                                   \
-                    while( left <= right && !LT(*pivot, *left) )                    \
-                    {                                                               \
-                        if( !LT(*left, *pivot) )                                    \
-                        {                                                           \
-                            if( left > left1 )                                      \
-                                CV_SWAP( *left1, *left, t );                        \
-                            swap_cnt = 1;                                           \
-                            left1++;                                                \
-                        }                                                           \
-                        left++;                                                     \
-                    }                                                               \
-                                                                                    \
-                    while( left <= right && !LT(*right, *pivot) )                   \
-                    {                                                               \
-                        if( !LT(*pivot, *right) )                                   \
-                        {                                                           \
-                            if( right < right1 )                                    \
-                                CV_SWAP( *right1, *right, t );                      \
-                            swap_cnt = 1;                                           \
-                            right1--;                                               \
-                        }                                                           \
-                        right--;                                                    \
-                    }                                                               \
-                                                                                    \
-                    if( left > right )                                              \
-                        break;                                                      \
-                    CV_SWAP( *left, *right, t );                                    \
-                    swap_cnt = 1;                                                   \
-                    left++;                                                         \
-                    right--;                                                        \
-                }                                                                   \
-                                                                                    \
-                if( swap_cnt == 0 )                                                 \
-                {                                                                   \
-                    left = left0, right = right0;                                   \
-                    goto insert_sort;                                               \
-                }                                                                   \
-                                                                                    \
-                n = MIN( (int)(left1 - left0), (int)(left - left1) );               \
-                for( i = 0; i < n; i++ )                                            \
-                    CV_SWAP( left0[i], left[i-n], t );                              \
-                                                                                    \
-                n = MIN( (int)(right0 - right1), (int)(right1 - right) );           \
-                for( i = 0; i < n; i++ )                                            \
-                    CV_SWAP( left[i], right0[i-n+1], t );                           \
-                n = (int)(left - left1);                                            \
-                m = (int)(right1 - right);                                          \
-                if( n > 1 )                                                         \
-                {                                                                   \
-                    if( m > 1 )                                                     \
-                    {                                                               \
-                        if( n > m )                                                 \
-                        {                                                           \
-                            stack[++sp].lb = left0;                                 \
-                            stack[sp].ub = left0 + n - 1;                           \
-                            left = right0 - m + 1, right = right0;                  \
-                        }                                                           \
-                        else                                                        \
-                        {                                                           \
-                            stack[++sp].lb = right0 - m + 1;                        \
-                            stack[sp].ub = right0;                                  \
-                            left = left0, right = left0 + n - 1;                    \
-                        }                                                           \
-                    }                                                               \
-                    else                                                            \
-                        left = left0, right = left0 + n - 1;                        \
-                }                                                                   \
-                else if( m > 1 )                                                    \
-                    left = right0 - m + 1, right = right0;                          \
-                else                                                                \
-                    break;                                                          \
-            }                                                                       \
-        }                                                                           \
-    }                                                                               \
-}
-
-#define CV_IMPLEMENT_QSORT( func_name, T, cmp )  \
-    CV_IMPLEMENT_QSORT_EX( func_name, T, cmp, int )
-
-/****************************************************************************************\
-*                     Structures and macros for integration with IPP                     *
-\****************************************************************************************/
-
-/* IPP-compatible return codes */
-typedef enum CvStatus
-{
-    CV_BADMEMBLOCK_ERR          = -113,
-    CV_INPLACE_NOT_SUPPORTED_ERR= -112,
-    CV_UNMATCHED_ROI_ERR        = -111,
-    CV_NOTFOUND_ERR             = -110,
-    CV_BADCONVERGENCE_ERR       = -109,
-
-    CV_BADDEPTH_ERR             = -107,
-    CV_BADROI_ERR               = -106,
-    CV_BADHEADER_ERR            = -105,
-    CV_UNMATCHED_FORMATS_ERR    = -104,
-    CV_UNSUPPORTED_COI_ERR      = -103,
-    CV_UNSUPPORTED_CHANNELS_ERR = -102,
-    CV_UNSUPPORTED_DEPTH_ERR    = -101,
-    CV_UNSUPPORTED_FORMAT_ERR   = -100,
-
-    CV_BADARG_ERR               = -49,  //ipp comp
-    CV_NOTDEFINED_ERR           = -48,  //ipp comp
-
-    CV_BADCHANNELS_ERR          = -47,  //ipp comp
-    CV_BADRANGE_ERR             = -44,  //ipp comp
-    CV_BADSTEP_ERR              = -29,  //ipp comp
-
-    CV_BADFLAG_ERR              =  -12,
-    CV_DIV_BY_ZERO_ERR          =  -11, //ipp comp
-    CV_BADCOEF_ERR              =  -10,
-
-    CV_BADFACTOR_ERR            =  -7,
-    CV_BADPOINT_ERR             =  -6,
-    CV_BADSCALE_ERR             =  -4,
-    CV_OUTOFMEM_ERR             =  -3,
-    CV_NULLPTR_ERR              =  -2,
-    CV_BADSIZE_ERR              =  -1,
-    CV_NO_ERR                   =   0,
-    CV_OK                       =   CV_NO_ERR
-}
-CvStatus;
-
-#define CV_NOTHROW throw()
-
-typedef struct CvFuncTable
-{
-    void*   fn_2d[CV_DEPTH_MAX];
-}
-CvFuncTable;
-
-typedef struct CvBigFuncTable
-{
-    void*   fn_2d[CV_DEPTH_MAX*4];
-} CvBigFuncTable;
-
-#define CV_INIT_FUNC_TAB( tab, FUNCNAME, FLAG )         \
-    (tab).fn_2d[CV_8U] = (void*)FUNCNAME##_8u##FLAG;    \
-    (tab).fn_2d[CV_8S] = 0;                             \
-    (tab).fn_2d[CV_16U] = (void*)FUNCNAME##_16u##FLAG;  \
-    (tab).fn_2d[CV_16S] = (void*)FUNCNAME##_16s##FLAG;  \
-    (tab).fn_2d[CV_32S] = (void*)FUNCNAME##_32s##FLAG;  \
-    (tab).fn_2d[CV_32F] = (void*)FUNCNAME##_32f##FLAG;  \
-    (tab).fn_2d[CV_64F] = (void*)FUNCNAME##_64f##FLAG
-
-namespace cv { namespace ogl {
-CV_EXPORTS bool checkError(const char* file, const int line, const char* func = "");
-}}
-
-#if defined(__GNUC__)
-    #define CV_CheckGlError() CV_DbgAssert( (cv::ogl::checkError(__FILE__, __LINE__, __func__)) )
-#else
-    #define CV_CheckGlError() CV_DbgAssert( (cv::ogl::checkError(__FILE__, __LINE__)) )
-#endif
-
-#endif // __OPENCV_CORE_INTERNAL_HPP__
diff --git a/modules/core/include/opencv2/core/mat.hpp b/modules/core/include/opencv2/core/mat.hpp
index 8e7890568c..c132be9456 100644
--- a/modules/core/include/opencv2/core/mat.hpp
+++ b/modules/core/include/opencv2/core/mat.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,1137 +41,1819 @@
 //
 //M*/
 
-#ifndef __OPENCV_CORE_MATRIX_OPERATIONS_HPP__
-#define __OPENCV_CORE_MATRIX_OPERATIONS_HPP__
+#ifndef __OPENCV_CORE_MAT_HPP__
+#define __OPENCV_CORE_MAT_HPP__
 
-#ifndef SKIP_INCLUDES
-#include <limits.h>
-#include <string.h>
-#endif // SKIP_INCLUDES
+#ifndef __cplusplus
+#  error mat.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/matx.hpp"
+#include "opencv2/core/types.hpp"
 
-#ifdef __cplusplus
 
 namespace cv
 {
 
-//////////////////////////////// Mat ////////////////////////////////
+//////////////////////// Input/Output Array Arguments /////////////////////////////////
 
-inline void Mat::initEmpty()
+/*!
+ Proxy datatype for passing Mat's and vector<>'s as input parameters
+ */
+class CV_EXPORTS _InputArray
 {
-    flags = MAGIC_VAL;
-    dims = rows = cols = 0;
-    data = datastart = dataend = datalimit = 0;
-    refcount = 0;
-    allocator = 0;
-}
+public:
+    enum {
+        KIND_SHIFT = 16,
+        FIXED_TYPE = 0x8000 << KIND_SHIFT,
+        FIXED_SIZE = 0x4000 << KIND_SHIFT,
+        KIND_MASK = ~(FIXED_TYPE|FIXED_SIZE) - (1 << KIND_SHIFT) + 1,
 
-inline Mat::Mat() : size(&rows)
-{
-    initEmpty();
-}
+        NONE              = 0 << KIND_SHIFT,
+        MAT               = 1 << KIND_SHIFT,
+        MATX              = 2 << KIND_SHIFT,
+        STD_VECTOR        = 3 << KIND_SHIFT,
+        STD_VECTOR_VECTOR = 4 << KIND_SHIFT,
+        STD_VECTOR_MAT    = 5 << KIND_SHIFT,
+        EXPR              = 6 << KIND_SHIFT,
+        OPENGL_BUFFER     = 7 << KIND_SHIFT,
+        CUDA_MEM          = 8 << KIND_SHIFT,
+        GPU_MAT           = 9 << KIND_SHIFT,
+        OCL_MAT           =10 << KIND_SHIFT
+    };
 
-inline Mat::Mat(int _rows, int _cols, int _type) : size(&rows)
-{
-    initEmpty();
-    create(_rows, _cols, _type);
-}
+    _InputArray();
+    _InputArray(const Mat& m);
+    _InputArray(const MatExpr& expr);
+    _InputArray(const std::vector<Mat>& vec);
+    template<typename _Tp> _InputArray(const Mat_<_Tp>& m);
+    template<typename _Tp> _InputArray(const std::vector<_Tp>& vec);
+    template<typename _Tp> _InputArray(const std::vector<std::vector<_Tp> >& vec);
+    template<typename _Tp> _InputArray(const std::vector<Mat_<_Tp> >& vec);
+    template<typename _Tp> _InputArray(const _Tp* vec, int n);
+    template<typename _Tp, int m, int n> _InputArray(const Matx<_Tp, m, n>& matx);
+    _InputArray(const double& val);
+    _InputArray(const gpu::GpuMat& d_mat);
+    _InputArray(const ogl::Buffer& buf);
+    _InputArray(const gpu::CudaMem& cuda_mem);
+    template<typename _Tp> _InputArray(const cudev::GpuMat_<_Tp>& m);
 
-inline Mat::Mat(int _rows, int _cols, int _type, const Scalar& _s) : size(&rows)
-{
-    initEmpty();
-    create(_rows, _cols, _type);
-    *this = _s;
-}
+    virtual Mat getMat(int i=-1) const;
+    virtual void getMatVector(std::vector<Mat>& mv) const;
+    virtual gpu::GpuMat getGpuMat() const;
+    virtual ogl::Buffer getOGlBuffer() const;
 
-inline Mat::Mat(Size _sz, int _type) : size(&rows)
-{
-    initEmpty();
-    create( _sz.height, _sz.width, _type );
-}
+    virtual int kind() const;
+    virtual Size size(int i=-1) const;
+    virtual size_t total(int i=-1) const;
+    virtual int type(int i=-1) const;
+    virtual int depth(int i=-1) const;
+    virtual int channels(int i=-1) const;
+    virtual bool empty() const;
 
-inline Mat::Mat(Size _sz, int _type, const Scalar& _s) : size(&rows)
-{
-    initEmpty();
-    create(_sz.height, _sz.width, _type);
-    *this = _s;
-}
+    virtual ~_InputArray();
 
-inline Mat::Mat(int _dims, const int* _sz, int _type) : size(&rows)
-{
-    initEmpty();
-    create(_dims, _sz, _type);
-}
+    int flags;
+    void* obj;
+    Size sz;
+};
 
-inline Mat::Mat(int _dims, const int* _sz, int _type, const Scalar& _s) : size(&rows)
-{
-    initEmpty();
-    create(_dims, _sz, _type);
-    *this = _s;
-}
 
-inline Mat::Mat(const Mat& m)
-    : flags(m.flags), dims(m.dims), rows(m.rows), cols(m.cols), data(m.data),
-    refcount(m.refcount), datastart(m.datastart), dataend(m.dataend),
-    datalimit(m.datalimit), allocator(m.allocator), size(&rows)
+/*!
+ Proxy datatype for passing Mat's and vector<>'s as input parameters
+ */
+class CV_EXPORTS _OutputArray : public _InputArray
 {
-    if( refcount )
-        CV_XADD(refcount, 1);
-    if( m.dims <= 2 )
+public:
+    enum
     {
-        step[0] = m.step[0]; step[1] = m.step[1];
-    }
-    else
+        DEPTH_MASK_8U = 1 << CV_8U,
+        DEPTH_MASK_8S = 1 << CV_8S,
+        DEPTH_MASK_16U = 1 << CV_16U,
+        DEPTH_MASK_16S = 1 << CV_16S,
+        DEPTH_MASK_32S = 1 << CV_32S,
+        DEPTH_MASK_32F = 1 << CV_32F,
+        DEPTH_MASK_64F = 1 << CV_64F,
+        DEPTH_MASK_ALL = (DEPTH_MASK_64F<<1)-1,
+        DEPTH_MASK_ALL_BUT_8S = DEPTH_MASK_ALL & ~DEPTH_MASK_8S,
+        DEPTH_MASK_FLT = DEPTH_MASK_32F + DEPTH_MASK_64F
+    };
+
+    _OutputArray();
+    _OutputArray(Mat& m);
+    _OutputArray(std::vector<Mat>& vec);
+    _OutputArray(gpu::GpuMat& d_mat);
+    _OutputArray(ogl::Buffer& buf);
+    _OutputArray(gpu::CudaMem& cuda_mem);
+    template<typename _Tp> _OutputArray(cudev::GpuMat_<_Tp>& m);
+    template<typename _Tp> _OutputArray(std::vector<_Tp>& vec);
+    template<typename _Tp> _OutputArray(std::vector<std::vector<_Tp> >& vec);
+    template<typename _Tp> _OutputArray(std::vector<Mat_<_Tp> >& vec);
+    template<typename _Tp> _OutputArray(Mat_<_Tp>& m);
+    template<typename _Tp> _OutputArray(_Tp* vec, int n);
+    template<typename _Tp, int m, int n> _OutputArray(Matx<_Tp, m, n>& matx);
+
+    _OutputArray(const Mat& m);
+    _OutputArray(const std::vector<Mat>& vec);
+    _OutputArray(const gpu::GpuMat& d_mat);
+    _OutputArray(const ogl::Buffer& buf);
+    _OutputArray(const gpu::CudaMem& cuda_mem);
+    template<typename _Tp> _OutputArray(const cudev::GpuMat_<_Tp>& m);
+    template<typename _Tp> _OutputArray(const std::vector<_Tp>& vec);
+    template<typename _Tp> _OutputArray(const std::vector<std::vector<_Tp> >& vec);
+    template<typename _Tp> _OutputArray(const std::vector<Mat_<_Tp> >& vec);
+    template<typename _Tp> _OutputArray(const Mat_<_Tp>& m);
+    template<typename _Tp> _OutputArray(const _Tp* vec, int n);
+    template<typename _Tp, int m, int n> _OutputArray(const Matx<_Tp, m, n>& matx);
+
+    virtual bool fixedSize() const;
+    virtual bool fixedType() const;
+    virtual bool needed() const;
+    virtual Mat& getMatRef(int i=-1) const;
+    virtual gpu::GpuMat& getGpuMatRef() const;
+    virtual ogl::Buffer& getOGlBufferRef() const;
+    virtual gpu::CudaMem& getCudaMemRef() const;
+    virtual void create(Size sz, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
+    virtual void create(int rows, int cols, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
+    virtual void create(int dims, const int* size, int type, int i=-1, bool allowTransposed=false, int fixedDepthMask=0) const;
+    virtual void release() const;
+    virtual void clear() const;
+
+    virtual ~_OutputArray();
+};
+
+typedef const _InputArray& InputArray;
+typedef InputArray InputArrayOfArrays;
+typedef const _OutputArray& OutputArray;
+typedef OutputArray OutputArrayOfArrays;
+typedef OutputArray InputOutputArray;
+typedef OutputArray InputOutputArrayOfArrays;
+
+CV_EXPORTS OutputArray noArray();
+
+
+
+/////////////////////////////////// MatAllocator //////////////////////////////////////
+
+/*!
+   Custom array allocator
+
+*/
+class CV_EXPORTS MatAllocator
+{
+public:
+    MatAllocator() {}
+    virtual ~MatAllocator() {}
+    virtual void allocate(int dims, const int* sizes, int type, int*& refcount,
+                          uchar*& datastart, uchar*& data, size_t* step) = 0;
+    virtual void deallocate(int* refcount, uchar* datastart, uchar* data) = 0;
+};
+
+
+
+//////////////////////////////// MatCommaInitializer //////////////////////////////////
+
+/*!
+ Comma-separated Matrix Initializer
+
+ The class instances are usually not created explicitly.
+ Instead, they are created on "matrix << firstValue" operator.
+
+ The sample below initializes 2x2 rotation matrix:
+
+ \code
+ double angle = 30, a = cos(angle*CV_PI/180), b = sin(angle*CV_PI/180);
+ Mat R = (Mat_<double>(2,2) << a, -b, b, a);
+ \endcode
+*/
+template<typename _Tp> class MatCommaInitializer_
+{
+public:
+    //! the constructor, created by "matrix << firstValue" operator, where matrix is cv::Mat
+    MatCommaInitializer_(Mat_<_Tp>* _m);
+    //! the operator that takes the next value and put it to the matrix
+    template<typename T2> MatCommaInitializer_<_Tp>& operator , (T2 v);
+    //! another form of conversion operator
+    operator Mat_<_Tp>() const;
+protected:
+    MatIterator_<_Tp> it;
+};
+
+
+
+
+/////////////////////////////////////// Mat ///////////////////////////////////////////
+
+/*!
+   The n-dimensional matrix class.
+
+   The class represents an n-dimensional dense numerical array that can act as
+   a matrix, image, optical flow map, 3-focal tensor etc.
+   It is very similar to CvMat and CvMatND types from earlier versions of OpenCV,
+   and similarly to those types, the matrix can be multi-channel. It also fully supports ROI mechanism.
+
+   There are many different ways to create cv::Mat object. Here are the some popular ones:
+   <ul>
+   <li> using cv::Mat::create(nrows, ncols, type) method or
+     the similar constructor cv::Mat::Mat(nrows, ncols, type[, fill_value]) constructor.
+     A new matrix of the specified size and specifed type will be allocated.
+     "type" has the same meaning as in cvCreateMat function,
+     e.g. CV_8UC1 means 8-bit single-channel matrix, CV_32FC2 means 2-channel (i.e. complex)
+     floating-point matrix etc:
+
+     \code
+     // make 7x7 complex matrix filled with 1+3j.
+     cv::Mat M(7,7,CV_32FC2,Scalar(1,3));
+     // and now turn M to 100x60 15-channel 8-bit matrix.
+     // The old content will be deallocated
+     M.create(100,60,CV_8UC(15));
+     \endcode
+
+     As noted in the introduction of this chapter, Mat::create()
+     will only allocate a new matrix when the current matrix dimensionality
+     or type are different from the specified.
+
+   <li> by using a copy constructor or assignment operator, where on the right side it can
+     be a matrix or expression, see below. Again, as noted in the introduction,
+     matrix assignment is O(1) operation because it only copies the header
+     and increases the reference counter. cv::Mat::clone() method can be used to get a full
+     (a.k.a. deep) copy of the matrix when you need it.
+
+   <li> by constructing a header for a part of another matrix. It can be a single row, single column,
+     several rows, several columns, rectangular region in the matrix (called a minor in algebra) or
+     a diagonal. Such operations are also O(1), because the new header will reference the same data.
+     You can actually modify a part of the matrix using this feature, e.g.
+
+     \code
+     // add 5-th row, multiplied by 3 to the 3rd row
+     M.row(3) = M.row(3) + M.row(5)*3;
+
+     // now copy 7-th column to the 1-st column
+     // M.col(1) = M.col(7); // this will not work
+     Mat M1 = M.col(1);
+     M.col(7).copyTo(M1);
+
+     // create new 320x240 image
+     cv::Mat img(Size(320,240),CV_8UC3);
+     // select a roi
+     cv::Mat roi(img, Rect(10,10,100,100));
+     // fill the ROI with (0,255,0) (which is green in RGB space);
+     // the original 320x240 image will be modified
+     roi = Scalar(0,255,0);
+     \endcode
+
+     Thanks to the additional cv::Mat::datastart and cv::Mat::dataend members, it is possible to
+     compute the relative sub-matrix position in the main "container" matrix using cv::Mat::locateROI():
+
+     \code
+     Mat A = Mat::eye(10, 10, CV_32S);
+     // extracts A columns, 1 (inclusive) to 3 (exclusive).
+     Mat B = A(Range::all(), Range(1, 3));
+     // extracts B rows, 5 (inclusive) to 9 (exclusive).
+     // that is, C ~ A(Range(5, 9), Range(1, 3))
+     Mat C = B(Range(5, 9), Range::all());
+     Size size; Point ofs;
+     C.locateROI(size, ofs);
+     // size will be (width=10,height=10) and the ofs will be (x=1, y=5)
+     \endcode
+
+     As in the case of whole matrices, if you need a deep copy, use cv::Mat::clone() method
+     of the extracted sub-matrices.
+
+   <li> by making a header for user-allocated-data. It can be useful for
+      <ol>
+      <li> processing "foreign" data using OpenCV (e.g. when you implement
+         a DirectShow filter or a processing module for gstreamer etc.), e.g.
+
+         \code
+         void process_video_frame(const unsigned char* pixels,
+                                  int width, int height, int step)
+         {
+            cv::Mat img(height, width, CV_8UC3, pixels, step);
+            cv::GaussianBlur(img, img, cv::Size(7,7), 1.5, 1.5);
+         }
+         \endcode
+
+      <li> for quick initialization of small matrices and/or super-fast element access
+
+         \code
+         double m[3][3] = {{a, b, c}, {d, e, f}, {g, h, i}};
+         cv::Mat M = cv::Mat(3, 3, CV_64F, m).inv();
+         \endcode
+      </ol>
+
+       partial yet very common cases of this "user-allocated data" case are conversions
+       from CvMat and IplImage to cv::Mat. For this purpose there are special constructors
+       taking pointers to CvMat or IplImage and the optional
+       flag indicating whether to copy the data or not.
+
+       Backward conversion from cv::Mat to CvMat or IplImage is provided via cast operators
+       cv::Mat::operator CvMat() an cv::Mat::operator IplImage().
+       The operators do not copy the data.
+
+
+       \code
+       IplImage* img = cvLoadImage("greatwave.jpg", 1);
+       Mat mtx(img); // convert IplImage* -> cv::Mat
+       CvMat oldmat = mtx; // convert cv::Mat -> CvMat
+       CV_Assert(oldmat.cols == img->width && oldmat.rows == img->height &&
+           oldmat.data.ptr == (uchar*)img->imageData && oldmat.step == img->widthStep);
+       \endcode
+
+   <li> by using MATLAB-style matrix initializers, cv::Mat::zeros(), cv::Mat::ones(), cv::Mat::eye(), e.g.:
+
+   \code
+   // create a double-precision identity martix and add it to M.
+   M += Mat::eye(M.rows, M.cols, CV_64F);
+   \endcode
+
+   <li> by using comma-separated initializer:
+
+   \code
+   // create 3x3 double-precision identity matrix
+   Mat M = (Mat_<double>(3,3) << 1, 0, 0, 0, 1, 0, 0, 0, 1);
+   \endcode
+
+   here we first call constructor of cv::Mat_ class (that we describe further) with the proper matrix,
+   and then we just put "<<" operator followed by comma-separated values that can be constants,
+   variables, expressions etc. Also, note the extra parentheses that are needed to avoid compiler errors.
+
+   </ul>
+
+   Once matrix is created, it will be automatically managed by using reference-counting mechanism
+   (unless the matrix header is built on top of user-allocated data,
+   in which case you should handle the data by yourself).
+   The matrix data will be deallocated when no one points to it;
+   if you want to release the data pointed by a matrix header before the matrix destructor is called,
+   use cv::Mat::release().
+
+   The next important thing to learn about the matrix class is element access. Here is how the matrix is stored.
+   The elements are stored in row-major order (row by row). The cv::Mat::data member points to the first element of the first row,
+   cv::Mat::rows contains the number of matrix rows and cv::Mat::cols - the number of matrix columns. There is yet another member,
+   cv::Mat::step that is used to actually compute address of a matrix element. cv::Mat::step is needed because the matrix can be
+   a part of another matrix or because there can some padding space in the end of each row for a proper alignment.
+
+   \image html roi.png
+
+   Given these parameters, address of the matrix element M_{ij} is computed as following:
+
+   addr(M_{ij})=M.data + M.step*i + j*M.elemSize()
+
+   if you know the matrix element type, e.g. it is float, then you can use cv::Mat::at() method:
+
+   addr(M_{ij})=&M.at<float>(i,j)
+
+   (where & is used to convert the reference returned by cv::Mat::at() to a pointer).
+   if you need to process a whole row of matrix, the most efficient way is to get
+   the pointer to the row first, and then just use plain C operator []:
+
+   \code
+   // compute sum of positive matrix elements
+   // (assuming that M is double-precision matrix)
+   double sum=0;
+   for(int i = 0; i < M.rows; i++)
+   {
+       const double* Mi = M.ptr<double>(i);
+       for(int j = 0; j < M.cols; j++)
+           sum += std::max(Mi[j], 0.);
+   }
+   \endcode
+
+   Some operations, like the above one, do not actually depend on the matrix shape,
+   they just process elements of a matrix one by one (or elements from multiple matrices
+   that are sitting in the same place, e.g. matrix addition). Such operations are called
+   element-wise and it makes sense to check whether all the input/output matrices are continuous,
+   i.e. have no gaps in the end of each row, and if yes, process them as a single long row:
+
+   \code
+   // compute sum of positive matrix elements, optimized variant
+   double sum=0;
+   int cols = M.cols, rows = M.rows;
+   if(M.isContinuous())
+   {
+       cols *= rows;
+       rows = 1;
+   }
+   for(int i = 0; i < rows; i++)
+   {
+       const double* Mi = M.ptr<double>(i);
+       for(int j = 0; j < cols; j++)
+           sum += std::max(Mi[j], 0.);
+   }
+   \endcode
+   in the case of continuous matrix the outer loop body will be executed just once,
+   so the overhead will be smaller, which will be especially noticeable in the case of small matrices.
+
+   Finally, there are STL-style iterators that are smart enough to skip gaps between successive rows:
+   \code
+   // compute sum of positive matrix elements, iterator-based variant
+   double sum=0;
+   MatConstIterator_<double> it = M.begin<double>(), it_end = M.end<double>();
+   for(; it != it_end; ++it)
+       sum += std::max(*it, 0.);
+   \endcode
+
+   The matrix iterators are random-access iterators, so they can be passed
+   to any STL algorithm, including std::sort().
+*/
+class CV_EXPORTS Mat
+{
+public:
+    //! default constructor
+    Mat();
+    //! constructs 2D matrix of the specified size and type
+    // (_type is CV_8UC1, CV_64FC3, CV_32SC(12) etc.)
+    Mat(int rows, int cols, int type);
+    Mat(Size size, int type);
+    //! constucts 2D matrix and fills it with the specified value _s.
+    Mat(int rows, int cols, int type, const Scalar& s);
+    Mat(Size size, int type, const Scalar& s);
+
+    //! constructs n-dimensional matrix
+    Mat(int ndims, const int* sizes, int type);
+    Mat(int ndims, const int* sizes, int type, const Scalar& s);
+
+    //! copy constructor
+    Mat(const Mat& m);
+    //! constructor for matrix headers pointing to user-allocated data
+    Mat(int rows, int cols, int type, void* data, size_t step=AUTO_STEP);
+    Mat(Size size, int type, void* data, size_t step=AUTO_STEP);
+    Mat(int ndims, const int* sizes, int type, void* data, const size_t* steps=0);
+
+    //! creates a matrix header for a part of the bigger matrix
+    Mat(const Mat& m, const Range& rowRange, const Range& colRange=Range::all());
+    Mat(const Mat& m, const Rect& roi);
+    Mat(const Mat& m, const Range* ranges);
+    //! builds matrix from std::vector with or without copying the data
+    template<typename _Tp> explicit Mat(const std::vector<_Tp>& vec, bool copyData=false);
+    //! builds matrix from cv::Vec; the data is copied by default
+    template<typename _Tp, int n> explicit Mat(const Vec<_Tp, n>& vec, bool copyData=true);
+    //! builds matrix from cv::Matx; the data is copied by default
+    template<typename _Tp, int m, int n> explicit Mat(const Matx<_Tp, m, n>& mtx, bool copyData=true);
+    //! builds matrix from a 2D point
+    template<typename _Tp> explicit Mat(const Point_<_Tp>& pt, bool copyData=true);
+    //! builds matrix from a 3D point
+    template<typename _Tp> explicit Mat(const Point3_<_Tp>& pt, bool copyData=true);
+    //! builds matrix from comma initializer
+    template<typename _Tp> explicit Mat(const MatCommaInitializer_<_Tp>& commaInitializer);
+
+    // //! converts old-style CvMat to the new matrix; the data is not copied by default
+    // Mat(const CvMat* m, bool copyData=false);
+    // //! converts old-style CvMatND to the new matrix; the data is not copied by default
+    // Mat(const CvMatND* m, bool copyData=false);
+    // //! converts old-style IplImage to the new matrix; the data is not copied by default
+    // Mat(const IplImage* img, bool copyData=false);
+    //Mat(const void* img, bool copyData=false);
+
+    //! download data from GpuMat
+    explicit Mat(const gpu::GpuMat& m);
+
+    //! destructor - calls release()
+    ~Mat();
+    //! assignment operators
+    Mat& operator = (const Mat& m);
+    Mat& operator = (const MatExpr& expr);
+
+    //! returns a new matrix header for the specified row
+    Mat row(int y) const;
+    //! returns a new matrix header for the specified column
+    Mat col(int x) const;
+    //! ... for the specified row span
+    Mat rowRange(int startrow, int endrow) const;
+    Mat rowRange(const Range& r) const;
+    //! ... for the specified column span
+    Mat colRange(int startcol, int endcol) const;
+    Mat colRange(const Range& r) const;
+    //! ... for the specified diagonal
+    // (d=0 - the main diagonal,
+    //  >0 - a diagonal from the lower half,
+    //  <0 - a diagonal from the upper half)
+    Mat diag(int d=0) const;
+    //! constructs a square diagonal matrix which main diagonal is vector "d"
+    static Mat diag(const Mat& d);
+
+    //! returns deep copy of the matrix, i.e. the data is copied
+    Mat clone() const;
+    //! copies the matrix content to "m".
+    // It calls m.create(this->size(), this->type()).
+    void copyTo( OutputArray m ) const;
+    //! copies those matrix elements to "m" that are marked with non-zero mask elements.
+    void copyTo( OutputArray m, InputArray mask ) const;
+    //! converts matrix to another datatype with optional scalng. See cvConvertScale.
+    void convertTo( OutputArray m, int rtype, double alpha=1, double beta=0 ) const;
+
+    void assignTo( Mat& m, int type=-1 ) const;
+
+    //! sets every matrix element to s
+    Mat& operator = (const Scalar& s);
+    //! sets some of the matrix elements to s, according to the mask
+    Mat& setTo(InputArray value, InputArray mask=noArray());
+    //! creates alternative matrix header for the same data, with different
+    // number of channels and/or different number of rows. see cvReshape.
+    Mat reshape(int cn, int rows=0) const;
+    Mat reshape(int cn, int newndims, const int* newsz) const;
+
+    //! matrix transposition by means of matrix expressions
+    MatExpr t() const;
+    //! matrix inversion by means of matrix expressions
+    MatExpr inv(int method=DECOMP_LU) const;
+    //! per-element matrix multiplication by means of matrix expressions
+    MatExpr mul(InputArray m, double scale=1) const;
+
+    //! computes cross-product of 2 3D vectors
+    Mat cross(InputArray m) const;
+    //! computes dot-product
+    double dot(InputArray m) const;
+
+    //! Matlab-style matrix initialization
+    static MatExpr zeros(int rows, int cols, int type);
+    static MatExpr zeros(Size size, int type);
+    static MatExpr zeros(int ndims, const int* sz, int type);
+    static MatExpr ones(int rows, int cols, int type);
+    static MatExpr ones(Size size, int type);
+    static MatExpr ones(int ndims, const int* sz, int type);
+    static MatExpr eye(int rows, int cols, int type);
+    static MatExpr eye(Size size, int type);
+
+    //! allocates new matrix data unless the matrix already has specified size and type.
+    // previous data is unreferenced if needed.
+    void create(int rows, int cols, int type);
+    void create(Size size, int type);
+    void create(int ndims, const int* sizes, int type);
+
+    //! increases the reference counter; use with care to avoid memleaks
+    void addref();
+    //! decreases reference counter;
+    // deallocates the data when reference counter reaches 0.
+    void release();
+
+    //! deallocates the matrix data
+    void deallocate();
+    //! internal use function; properly re-allocates _size, _step arrays
+    void copySize(const Mat& m);
+
+    //! reserves enough space to fit sz hyper-planes
+    void reserve(size_t sz);
+    //! resizes matrix to the specified number of hyper-planes
+    void resize(size_t sz);
+    //! resizes matrix to the specified number of hyper-planes; initializes the newly added elements
+    void resize(size_t sz, const Scalar& s);
+    //! internal function
+    void push_back_(const void* elem);
+    //! adds element to the end of 1d matrix (or possibly multiple elements when _Tp=Mat)
+    template<typename _Tp> void push_back(const _Tp& elem);
+    template<typename _Tp> void push_back(const Mat_<_Tp>& elem);
+    void push_back(const Mat& m);
+    //! removes several hyper-planes from bottom of the matrix
+    void pop_back(size_t nelems=1);
+
+    //! locates matrix header within a parent matrix. See below
+    void locateROI( Size& wholeSize, Point& ofs ) const;
+    //! moves/resizes the current matrix ROI inside the parent matrix.
+    Mat& adjustROI( int dtop, int dbottom, int dleft, int dright );
+    //! extracts a rectangular sub-matrix
+    // (this is a generalized form of row, rowRange etc.)
+    Mat operator()( Range rowRange, Range colRange ) const;
+    Mat operator()( const Rect& roi ) const;
+    Mat operator()( const Range* ranges ) const;
+
+    // //! converts header to CvMat; no data is copied
+    // operator CvMat() const;
+    // //! converts header to CvMatND; no data is copied
+    // operator CvMatND() const;
+    // //! converts header to IplImage; no data is copied
+    // operator IplImage() const;
+
+    template<typename _Tp> operator std::vector<_Tp>() const;
+    template<typename _Tp, int n> operator Vec<_Tp, n>() const;
+    template<typename _Tp, int m, int n> operator Matx<_Tp, m, n>() const;
+
+    //! returns true iff the matrix data is continuous
+    // (i.e. when there are no gaps between successive rows).
+    // similar to CV_IS_MAT_CONT(cvmat->type)
+    bool isContinuous() const;
+
+    //! returns true if the matrix is a submatrix of another matrix
+    bool isSubmatrix() const;
+
+    //! returns element size in bytes,
+    // similar to CV_ELEM_SIZE(cvmat->type)
+    size_t elemSize() const;
+    //! returns the size of element channel in bytes.
+    size_t elemSize1() const;
+    //! returns element type, similar to CV_MAT_TYPE(cvmat->type)
+    int type() const;
+    //! returns element type, similar to CV_MAT_DEPTH(cvmat->type)
+    int depth() const;
+    //! returns element type, similar to CV_MAT_CN(cvmat->type)
+    int channels() const;
+    //! returns step/elemSize1()
+    size_t step1(int i=0) const;
+    //! returns true if matrix data is NULL
+    bool empty() const;
+    //! returns the total number of matrix elements
+    size_t total() const;
+
+    //! returns N if the matrix is 1-channel (N x ptdim) or ptdim-channel (1 x N) or (N x 1); negative number otherwise
+    int checkVector(int elemChannels, int depth=-1, bool requireContinuous=true) const;
+
+    //! returns pointer to i0-th submatrix along the dimension #0
+    uchar* ptr(int i0=0);
+    const uchar* ptr(int i0=0) const;
+
+    //! returns pointer to (i0,i1) submatrix along the dimensions #0 and #1
+    uchar* ptr(int i0, int i1);
+    const uchar* ptr(int i0, int i1) const;
+
+    //! returns pointer to (i0,i1,i3) submatrix along the dimensions #0, #1, #2
+    uchar* ptr(int i0, int i1, int i2);
+    const uchar* ptr(int i0, int i1, int i2) const;
+
+    //! returns pointer to the matrix element
+    uchar* ptr(const int* idx);
+    //! returns read-only pointer to the matrix element
+    const uchar* ptr(const int* idx) const;
+
+    template<int n> uchar* ptr(const Vec<int, n>& idx);
+    template<int n> const uchar* ptr(const Vec<int, n>& idx) const;
+
+    //! template version of the above method
+    template<typename _Tp> _Tp* ptr(int i0=0);
+    template<typename _Tp> const _Tp* ptr(int i0=0) const;
+
+    template<typename _Tp> _Tp* ptr(int i0, int i1);
+    template<typename _Tp> const _Tp* ptr(int i0, int i1) const;
+
+    template<typename _Tp> _Tp* ptr(int i0, int i1, int i2);
+    template<typename _Tp> const _Tp* ptr(int i0, int i1, int i2) const;
+
+    template<typename _Tp> _Tp* ptr(const int* idx);
+    template<typename _Tp> const _Tp* ptr(const int* idx) const;
+
+    template<typename _Tp, int n> _Tp* ptr(const Vec<int, n>& idx);
+    template<typename _Tp, int n> const _Tp* ptr(const Vec<int, n>& idx) const;
+
+    //! the same as above, with the pointer dereferencing
+    template<typename _Tp> _Tp& at(int i0=0);
+    template<typename _Tp> const _Tp& at(int i0=0) const;
+
+    template<typename _Tp> _Tp& at(int i0, int i1);
+    template<typename _Tp> const _Tp& at(int i0, int i1) const;
+
+    template<typename _Tp> _Tp& at(int i0, int i1, int i2);
+    template<typename _Tp> const _Tp& at(int i0, int i1, int i2) const;
+
+    template<typename _Tp> _Tp& at(const int* idx);
+    template<typename _Tp> const _Tp& at(const int* idx) const;
+
+    template<typename _Tp, int n> _Tp& at(const Vec<int, n>& idx);
+    template<typename _Tp, int n> const _Tp& at(const Vec<int, n>& idx) const;
+
+    //! special versions for 2D arrays (especially convenient for referencing image pixels)
+    template<typename _Tp> _Tp& at(Point pt);
+    template<typename _Tp> const _Tp& at(Point pt) const;
+
+    //! template methods for iteration over matrix elements.
+    // the iterators take care of skipping gaps in the end of rows (if any)
+    template<typename _Tp> MatIterator_<_Tp> begin();
+    template<typename _Tp> MatIterator_<_Tp> end();
+    template<typename _Tp> MatConstIterator_<_Tp> begin() const;
+    template<typename _Tp> MatConstIterator_<_Tp> end() const;
+
+    enum { MAGIC_VAL  = 0x42FF0000, AUTO_STEP = 0, CONTINUOUS_FLAG = CV_MAT_CONT_FLAG, SUBMATRIX_FLAG = CV_SUBMAT_FLAG };
+    enum { MAGIC_MASK = 0xFFFF0000, TYPE_MASK = 0x00000FFF, DEPTH_MASK = 7 };
+
+    /*! includes several bit-fields:
+         - the magic signature
+         - continuity flag
+         - depth
+         - number of channels
+     */
+    int flags;
+    //! the matrix dimensionality, >= 2
+    int dims;
+    //! the number of rows and columns or (-1, -1) when the matrix has more than 2 dimensions
+    int rows, cols;
+    //! pointer to the data
+    uchar* data;
+
+    //! pointer to the reference counter;
+    // when matrix points to user-allocated data, the pointer is NULL
+    int* refcount;
+
+    //! helper fields used in locateROI and adjustROI
+    uchar* datastart;
+    uchar* dataend;
+    uchar* datalimit;
+
+    //! custom allocator
+    MatAllocator* allocator;
+
+    struct CV_EXPORTS MSize
     {
-        dims = 0;
-        copySize(m);
-    }
-}
+        MSize(int* _p);
+        Size operator()() const;
+        const int& operator[](int i) const;
+        int& operator[](int i);
+        operator const int*() const;
+        bool operator == (const MSize& sz) const;
+        bool operator != (const MSize& sz) const;
 
-inline Mat::Mat(int _rows, int _cols, int _type, void* _data, size_t _step)
-    : flags(MAGIC_VAL + (_type & TYPE_MASK)), dims(2), rows(_rows), cols(_cols),
-    data((uchar*)_data), refcount(0), datastart((uchar*)_data), dataend(0),
-    datalimit(0), allocator(0), size(&rows)
-{
-    size_t esz = CV_ELEM_SIZE(_type), minstep = cols*esz;
-    if( _step == AUTO_STEP )
+        int* p;
+    };
+
+    struct CV_EXPORTS MStep
     {
-        _step = minstep;
-        flags |= CONTINUOUS_FLAG;
-    }
-    else
-    {
-        if( rows == 1 ) _step = minstep;
-        CV_DbgAssert( _step >= minstep );
-        flags |= _step == minstep ? CONTINUOUS_FLAG : 0;
-    }
-    step[0] = _step; step[1] = esz;
-    datalimit = datastart + _step*rows;
-    dataend = datalimit - _step + minstep;
-}
+        MStep();
+        MStep(size_t s);
+        const size_t& operator[](int i) const;
+        size_t& operator[](int i);
+        operator size_t() const;
+        MStep& operator = (size_t s);
 
-inline Mat::Mat(Size _sz, int _type, void* _data, size_t _step)
-    : flags(MAGIC_VAL + (_type & TYPE_MASK)), dims(2), rows(_sz.height), cols(_sz.width),
-    data((uchar*)_data), refcount(0), datastart((uchar*)_data), dataend(0),
-    datalimit(0), allocator(0), size(&rows)
-{
-    size_t esz = CV_ELEM_SIZE(_type), minstep = cols*esz;
-    if( _step == AUTO_STEP )
-    {
-        _step = minstep;
-        flags |= CONTINUOUS_FLAG;
-    }
-    else
-    {
-        if( rows == 1 ) _step = minstep;
-        CV_DbgAssert( _step >= minstep );
-        flags |= _step == minstep ? CONTINUOUS_FLAG : 0;
-    }
-    step[0] = _step; step[1] = esz;
-    datalimit = datastart + _step*rows;
-    dataend = datalimit - _step + minstep;
-}
+        size_t* p;
+        size_t buf[2];
+    protected:
+        MStep& operator = (const MStep&);
+    };
 
+    MSize size;
+    MStep step;
 
-template<typename _Tp> inline Mat::Mat(const std::vector<_Tp>& vec, bool copyData)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(2), rows((int)vec.size()), cols(1), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    if(vec.empty())
-        return;
-    if( !copyData )
-    {
-        step[0] = step[1] = sizeof(_Tp);
-        data = datastart = (uchar*)&vec[0];
-        datalimit = dataend = datastart + rows*step[0];
-    }
-    else
-        Mat((int)vec.size(), 1, DataType<_Tp>::type, (uchar*)&vec[0]).copyTo(*this);
-}
+protected:
+};
 
 
-template<typename _Tp, int n> inline Mat::Mat(const Vec<_Tp, n>& vec, bool copyData)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(2), rows(n), cols(1), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    if( !copyData )
-    {
-        step[0] = step[1] = sizeof(_Tp);
-        data = datastart = (uchar*)vec.val;
-        datalimit = dataend = datastart + rows*step[0];
-    }
-    else
-        Mat(n, 1, DataType<_Tp>::type, (void*)vec.val).copyTo(*this);
-}
-
-
-template<typename _Tp, int m, int n> inline Mat::Mat(const Matx<_Tp,m,n>& M, bool copyData)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(2), rows(m), cols(n), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    if( !copyData )
-    {
-        step[0] = cols*sizeof(_Tp);
-        step[1] = sizeof(_Tp);
-        data = datastart = (uchar*)M.val;
-        datalimit = dataend = datastart + rows*step[0];
-    }
-    else
-        Mat(m, n, DataType<_Tp>::type, (uchar*)M.val).copyTo(*this);
-}
-
-
-template<typename _Tp> inline Mat::Mat(const Point_<_Tp>& pt, bool copyData)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(2), rows(2), cols(1), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    if( !copyData )
-    {
-        step[0] = step[1] = sizeof(_Tp);
-        data = datastart = (uchar*)&pt.x;
-        datalimit = dataend = datastart + rows*step[0];
-    }
-    else
-    {
-        create(2, 1, DataType<_Tp>::type);
-        ((_Tp*)data)[0] = pt.x;
-        ((_Tp*)data)[1] = pt.y;
-    }
-}
-
-
-template<typename _Tp> inline Mat::Mat(const Point3_<_Tp>& pt, bool copyData)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(2), rows(3), cols(1), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    if( !copyData )
-    {
-        step[0] = step[1] = sizeof(_Tp);
-        data = datastart = (uchar*)&pt.x;
-        datalimit = dataend = datastart + rows*step[0];
-    }
-    else
-    {
-        create(3, 1, DataType<_Tp>::type);
-        ((_Tp*)data)[0] = pt.x;
-        ((_Tp*)data)[1] = pt.y;
-        ((_Tp*)data)[2] = pt.z;
-    }
-}
-
-
-template<typename _Tp> inline Mat::Mat(const MatCommaInitializer_<_Tp>& commaInitializer)
-    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG),
-    dims(0), rows(0), cols(0), data(0), refcount(0),
-    datastart(0), dataend(0), allocator(0), size(&rows)
-{
-    *this = *commaInitializer;
-}
-
-inline Mat::~Mat()
-{
-    release();
-    if( step.p != step.buf )
-        fastFree(step.p);
-}
-
-inline Mat& Mat::operator = (const Mat& m)
-{
-    if( this != &m )
-    {
-        if( m.refcount )
-            CV_XADD(m.refcount, 1);
-        release();
-        flags = m.flags;
-        if( dims <= 2 && m.dims <= 2 )
-        {
-            dims = m.dims;
-            rows = m.rows;
-            cols = m.cols;
-            step[0] = m.step[0];
-            step[1] = m.step[1];
-        }
-        else
-            copySize(m);
-        data = m.data;
-        datastart = m.datastart;
-        dataend = m.dataend;
-        datalimit = m.datalimit;
-        refcount = m.refcount;
-        allocator = m.allocator;
-    }
-    return *this;
-}
-
-inline Mat Mat::row(int y) const { return Mat(*this, Range(y, y+1), Range::all()); }
-inline Mat Mat::col(int x) const { return Mat(*this, Range::all(), Range(x, x+1)); }
-inline Mat Mat::rowRange(int startrow, int endrow) const
-    { return Mat(*this, Range(startrow, endrow), Range::all()); }
-inline Mat Mat::rowRange(const Range& r) const
-    { return Mat(*this, r, Range::all()); }
-inline Mat Mat::colRange(int startcol, int endcol) const
-    { return Mat(*this, Range::all(), Range(startcol, endcol)); }
-inline Mat Mat::colRange(const Range& r) const
-    { return Mat(*this, Range::all(), r); }
-
-inline Mat Mat::diag(const Mat& d)
-{
-    CV_Assert( d.cols == 1 || d.rows == 1 );
-    int len = d.rows + d.cols - 1;
-    Mat m(len, len, d.type(), Scalar(0)), md = m.diag();
-    if( d.cols == 1 )
-        d.copyTo(md);
-    else
-        transpose(d, md);
-    return m;
-}
-
-inline Mat Mat::clone() const
-{
-    Mat m;
-    copyTo(m);
-    return m;
-}
-
-inline void Mat::assignTo( Mat& m, int _type ) const
-{
-    if( _type < 0 )
-        m = *this;
-    else
-        convertTo(m, _type);
-}
-
-inline void Mat::create(int _rows, int _cols, int _type)
-{
-    _type &= TYPE_MASK;
-    if( dims <= 2 && rows == _rows && cols == _cols && type() == _type && data )
-        return;
-    int sz[] = {_rows, _cols};
-    create(2, sz, _type);
-}
-
-inline void Mat::create(Size _sz, int _type)
-{
-    create(_sz.height, _sz.width, _type);
-}
-
-inline void Mat::addref()
-{ if( refcount ) CV_XADD(refcount, 1); }
-
-inline void Mat::release()
-{
-    if( refcount && CV_XADD(refcount, -1) == 1 )
-        deallocate();
-    data = datastart = dataend = datalimit = 0;
-    size.p[0] = 0;
-    refcount = 0;
-}
-
-inline Mat Mat::operator()( Range _rowRange, Range _colRange ) const
-{
-    return Mat(*this, _rowRange, _colRange);
-}
-
-inline Mat Mat::operator()( const Rect& roi ) const
-{ return Mat(*this, roi); }
-
-inline Mat Mat::operator()(const Range* ranges) const
-{
-    return Mat(*this, ranges);
-}
-
-inline Mat::operator CvMat() const
-{
-    CV_DbgAssert(dims <= 2);
-    CvMat m = cvMat(rows, dims == 1 ? 1 : cols, type(), data);
-    m.step = (int)step[0];
-    m.type = (m.type & ~CONTINUOUS_FLAG) | (flags & CONTINUOUS_FLAG);
-    return m;
-}
-
-inline bool Mat::isContinuous() const { return (flags & CONTINUOUS_FLAG) != 0; }
-inline bool Mat::isSubmatrix() const { return (flags & SUBMATRIX_FLAG) != 0; }
-inline size_t Mat::elemSize() const { return dims > 0 ? step.p[dims-1] : 0; }
-inline size_t Mat::elemSize1() const { return CV_ELEM_SIZE1(flags); }
-inline int Mat::type() const { return CV_MAT_TYPE(flags); }
-inline int Mat::depth() const { return CV_MAT_DEPTH(flags); }
-inline int Mat::channels() const { return CV_MAT_CN(flags); }
-inline size_t Mat::step1(int i) const { return step.p[i]/elemSize1(); }
-inline bool Mat::empty() const { return data == 0 || total() == 0; }
-inline size_t Mat::total() const
-{
-    if( dims <= 2 )
-        return (size_t)rows*cols;
-    size_t p = 1;
-    for( int i = 0; i < dims; i++ )
-        p *= size[i];
-    return p;
-}
-
-inline uchar* Mat::ptr(int y)
-{
-    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
-    return data + step.p[0]*y;
-}
-
-inline const uchar* Mat::ptr(int y) const
-{
-    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
-    return data + step.p[0]*y;
-}
-
-template<typename _Tp> inline _Tp* Mat::ptr(int y)
-{
-    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
-    return (_Tp*)(data + step.p[0]*y);
-}
-
-template<typename _Tp> inline const _Tp* Mat::ptr(int y) const
-{
-    CV_DbgAssert( y == 0 || (data && dims >= 1 && data && (unsigned)y < (unsigned)size.p[0]) );
-    return (const _Tp*)(data + step.p[0]*y);
-}
-
-
-inline uchar* Mat::ptr(int i0, int i1)
-{
-    CV_DbgAssert( dims >= 2 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] );
-    return data + i0*step.p[0] + i1*step.p[1];
-}
-
-inline const uchar* Mat::ptr(int i0, int i1) const
-{
-    CV_DbgAssert( dims >= 2 && data &&
-                 (unsigned)i0 < (unsigned)size.p[0] &&
-                 (unsigned)i1 < (unsigned)size.p[1] );
-    return data + i0*step.p[0] + i1*step.p[1];
-}
-
-template<typename _Tp> inline _Tp* Mat::ptr(int i0, int i1)
-{
-    CV_DbgAssert( dims >= 2 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] );
-    return (_Tp*)(data + i0*step.p[0] + i1*step.p[1]);
-}
-
-template<typename _Tp> inline const _Tp* Mat::ptr(int i0, int i1) const
-{
-    CV_DbgAssert( dims >= 2 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] );
-    return (const _Tp*)(data + i0*step.p[0] + i1*step.p[1]);
-}
-
-inline uchar* Mat::ptr(int i0, int i1, int i2)
-{
-    CV_DbgAssert( dims >= 3 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  (unsigned)i2 < (unsigned)size.p[2] );
-    return data + i0*step.p[0] + i1*step.p[1] + i2*step.p[2];
-}
-
-inline const uchar* Mat::ptr(int i0, int i1, int i2) const
-{
-    CV_DbgAssert( dims >= 3 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  (unsigned)i2 < (unsigned)size.p[2] );
-    return data + i0*step.p[0] + i1*step.p[1] + i2*step.p[2];
-}
-
-template<typename _Tp> inline _Tp* Mat::ptr(int i0, int i1, int i2)
-{
-    CV_DbgAssert( dims >= 3 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  (unsigned)i2 < (unsigned)size.p[2] );
-    return (_Tp*)(data + i0*step.p[0] + i1*step.p[1] + i2*step.p[2]);
-}
-
-template<typename _Tp> inline const _Tp* Mat::ptr(int i0, int i1, int i2) const
-{
-    CV_DbgAssert( dims >= 3 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  (unsigned)i2 < (unsigned)size.p[2] );
-    return (const _Tp*)(data + i0*step.p[0] + i1*step.p[1] + i2*step.p[2]);
-}
-
-inline uchar* Mat::ptr(const int* idx)
-{
-    int i, d = dims;
-    uchar* p = data;
-    CV_DbgAssert( d >= 1 && p );
-    for( i = 0; i < d; i++ )
-    {
-        CV_DbgAssert( (unsigned)idx[i] < (unsigned)size.p[i] );
-        p += idx[i]*step.p[i];
-    }
-    return p;
-}
-
-inline const uchar* Mat::ptr(const int* idx) const
-{
-    int i, d = dims;
-    uchar* p = data;
-    CV_DbgAssert( d >= 1 && p );
-    for( i = 0; i < d; i++ )
-    {
-        CV_DbgAssert( (unsigned)idx[i] < (unsigned)size.p[i] );
-        p += idx[i]*step.p[i];
-    }
-    return p;
-}
-
-template<typename _Tp> inline _Tp& Mat::at(int i0, int i1)
-{
-    CV_DbgAssert( dims <= 2 && data && (unsigned)i0 < (unsigned)size.p[0] &&
-        (unsigned)(i1*DataType<_Tp>::channels) < (unsigned)(size.p[1]*channels()) &&
-        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
-    return ((_Tp*)(data + step.p[0]*i0))[i1];
-}
-
-template<typename _Tp> inline const _Tp& Mat::at(int i0, int i1) const
-{
-    CV_DbgAssert( dims <= 2 && data && (unsigned)i0 < (unsigned)size.p[0] &&
-        (unsigned)(i1*DataType<_Tp>::channels) < (unsigned)(size.p[1]*channels()) &&
-        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
-    return ((const _Tp*)(data + step.p[0]*i0))[i1];
-}
-
-template<typename _Tp> inline _Tp& Mat::at(Point pt)
-{
-    CV_DbgAssert( dims <= 2 && data && (unsigned)pt.y < (unsigned)size.p[0] &&
-        (unsigned)(pt.x*DataType<_Tp>::channels) < (unsigned)(size.p[1]*channels()) &&
-        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
-    return ((_Tp*)(data + step.p[0]*pt.y))[pt.x];
-}
-
-template<typename _Tp> inline const _Tp& Mat::at(Point pt) const
-{
-    CV_DbgAssert( dims <= 2 && data && (unsigned)pt.y < (unsigned)size.p[0] &&
-        (unsigned)(pt.x*DataType<_Tp>::channels) < (unsigned)(size.p[1]*channels()) &&
-        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
-    return ((const _Tp*)(data + step.p[0]*pt.y))[pt.x];
-}
-
-template<typename _Tp> inline _Tp& Mat::at(int i0)
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                 (unsigned)i0 < (unsigned)(size.p[0]*size.p[1]) &&
-                 elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    if( isContinuous() || size.p[0] == 1 )
-        return ((_Tp*)data)[i0];
-    if( size.p[1] == 1 )
-        return *(_Tp*)(data + step.p[0]*i0);
-    int i = i0/cols, j = i0 - i*cols;
-    return ((_Tp*)(data + step.p[0]*i))[j];
-}
-
-template<typename _Tp> inline const _Tp& Mat::at(int i0) const
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                 (unsigned)i0 < (unsigned)(size.p[0]*size.p[1]) &&
-                 elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    if( isContinuous() || size.p[0] == 1 )
-        return ((const _Tp*)data)[i0];
-    if( size.p[1] == 1 )
-        return *(const _Tp*)(data + step.p[0]*i0);
-    int i = i0/cols, j = i0 - i*cols;
-    return ((const _Tp*)(data + step.p[0]*i))[j];
-}
-
-template<typename _Tp> inline _Tp& Mat::at(int i0, int i1, int i2)
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(_Tp*)ptr(i0, i1, i2);
-}
-template<typename _Tp> inline const _Tp& Mat::at(int i0, int i1, int i2) const
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(const _Tp*)ptr(i0, i1, i2);
-}
-template<typename _Tp> inline _Tp& Mat::at(const int* idx)
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(_Tp*)ptr(idx);
-}
-template<typename _Tp> inline const _Tp& Mat::at(const int* idx) const
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(const _Tp*)ptr(idx);
-}
-template<typename _Tp, int n> _Tp& Mat::at(const Vec<int, n>& idx)
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(_Tp*)ptr(idx.val);
-}
-template<typename _Tp, int n> inline const _Tp& Mat::at(const Vec<int, n>& idx) const
-{
-    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
-    return *(const _Tp*)ptr(idx.val);
-}
-
-
-template<typename _Tp> inline MatConstIterator_<_Tp> Mat::begin() const
-{
-    CV_DbgAssert( elemSize() == sizeof(_Tp) );
-    return MatConstIterator_<_Tp>((const Mat_<_Tp>*)this);
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp> Mat::end() const
-{
-    CV_DbgAssert( elemSize() == sizeof(_Tp) );
-    MatConstIterator_<_Tp> it((const Mat_<_Tp>*)this);
-    it += total();
-    return it;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp> Mat::begin()
-{
-    CV_DbgAssert( elemSize() == sizeof(_Tp) );
-    return MatIterator_<_Tp>((Mat_<_Tp>*)this);
-}
-
-template<typename _Tp> inline MatIterator_<_Tp> Mat::end()
-{
-    CV_DbgAssert( elemSize() == sizeof(_Tp) );
-    MatIterator_<_Tp> it((Mat_<_Tp>*)this);
-    it += total();
-    return it;
-}
-
-template<typename _Tp> inline Mat::operator std::vector<_Tp>() const
-{
-    std::vector<_Tp> v;
-    copyTo(v);
-    return v;
-}
-
-template<typename _Tp, int n> inline Mat::operator Vec<_Tp, n>() const
-{
-    CV_Assert( data && dims <= 2 && (rows == 1 || cols == 1) &&
-               rows + cols - 1 == n && channels() == 1 );
-
-    if( isContinuous() && type() == DataType<_Tp>::type )
-        return Vec<_Tp, n>((_Tp*)data);
-    Vec<_Tp, n> v; Mat tmp(rows, cols, DataType<_Tp>::type, v.val);
-    convertTo(tmp, tmp.type());
-    return v;
-}
-
-template<typename _Tp, int m, int n> inline Mat::operator Matx<_Tp, m, n>() const
-{
-    CV_Assert( data && dims <= 2 && rows == m && cols == n && channels() == 1 );
-
-    if( isContinuous() && type() == DataType<_Tp>::type )
-        return Matx<_Tp, m, n>((_Tp*)data);
-    Matx<_Tp, m, n> mtx; Mat tmp(rows, cols, DataType<_Tp>::type, mtx.val);
-    convertTo(tmp, tmp.type());
-    return mtx;
-}
-
-
-template<typename _Tp> inline void Mat::push_back(const _Tp& elem)
-{
-    if( !data )
-    {
-        *this = Mat(1, 1, DataType<_Tp>::type, (void*)&elem).clone();
-        return;
-    }
-    CV_Assert(DataType<_Tp>::type == type() && cols == 1
-              /* && dims == 2 (cols == 1 implies dims == 2) */);
-    uchar* tmp = dataend + step[0];
-    if( !isSubmatrix() && isContinuous() && tmp <= datalimit )
-    {
-        *(_Tp*)(data + (size.p[0]++)*step.p[0]) = elem;
-        dataend = tmp;
-    }
-    else
-        push_back_(&elem);
-}
-
-template<typename _Tp> inline void Mat::push_back(const Mat_<_Tp>& m)
-{
-    push_back((const Mat&)m);
-}
-
-inline Mat::MSize::MSize(int* _p) : p(_p) {}
-inline Size Mat::MSize::operator()() const
-{
-    CV_DbgAssert(p[-1] <= 2);
-    return Size(p[1], p[0]);
-}
-inline const int& Mat::MSize::operator[](int i) const { return p[i]; }
-inline int& Mat::MSize::operator[](int i) { return p[i]; }
-inline Mat::MSize::operator const int*() const { return p; }
-
-inline bool Mat::MSize::operator == (const MSize& sz) const
-{
-    int d = p[-1], dsz = sz.p[-1];
-    if( d != dsz )
-        return false;
-    if( d == 2 )
-        return p[0] == sz.p[0] && p[1] == sz.p[1];
-
-    for( int i = 0; i < d; i++ )
-        if( p[i] != sz.p[i] )
-            return false;
-    return true;
-}
-
-inline bool Mat::MSize::operator != (const MSize& sz) const
-{
-    return !(*this == sz);
-}
-
-inline Mat::MStep::MStep() { p = buf; p[0] = p[1] = 0; }
-inline Mat::MStep::MStep(size_t s) { p = buf; p[0] = s; p[1] = 0; }
-inline const size_t& Mat::MStep::operator[](int i) const { return p[i]; }
-inline size_t& Mat::MStep::operator[](int i) { return p[i]; }
-inline Mat::MStep::operator size_t() const
-{
-    CV_DbgAssert( p == buf );
-    return buf[0];
-}
-inline Mat::MStep& Mat::MStep::operator = (size_t s)
-{
-    CV_DbgAssert( p == buf );
-    buf[0] = s;
-    return *this;
-}
-
-static inline Mat cvarrToMatND(const CvArr* arr, bool copyData=false, int coiMode=0)
-{
-    return cvarrToMat(arr, copyData, true, coiMode);
-}
-
-///////////////////////////////////////////// SVD //////////////////////////////////////////////////////
-
-inline SVD::SVD() {}
-inline SVD::SVD( InputArray m, int flags ) { operator ()(m, flags); }
-inline void SVD::solveZ( InputArray m, OutputArray _dst )
-{
-    Mat mtx = m.getMat();
-    SVD svd(mtx, (mtx.rows >= mtx.cols ? 0 : SVD::FULL_UV));
-    _dst.create(svd.vt.cols, 1, svd.vt.type());
-    Mat dst = _dst.getMat();
-    svd.vt.row(svd.vt.rows-1).reshape(1,svd.vt.cols).copyTo(dst);
-}
-
-template<typename _Tp, int m, int n, int nm> inline void
-    SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w, Matx<_Tp, m, nm>& u, Matx<_Tp, n, nm>& vt )
-{
-    assert( nm == MIN(m, n));
-    Mat _a(a, false), _u(u, false), _w(w, false), _vt(vt, false);
-    SVD::compute(_a, _w, _u, _vt);
-    CV_Assert(_w.data == (uchar*)&w.val[0] && _u.data == (uchar*)&u.val[0] && _vt.data == (uchar*)&vt.val[0]);
-}
-
-template<typename _Tp, int m, int n, int nm> inline void
-SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w )
-{
-    assert( nm == MIN(m, n));
-    Mat _a(a, false), _w(w, false);
-    SVD::compute(_a, _w);
-    CV_Assert(_w.data == (uchar*)&w.val[0]);
-}
-
-template<typename _Tp, int m, int n, int nm, int nb> inline void
-SVD::backSubst( const Matx<_Tp, nm, 1>& w, const Matx<_Tp, m, nm>& u,
-                const Matx<_Tp, n, nm>& vt, const Matx<_Tp, m, nb>& rhs,
-                Matx<_Tp, n, nb>& dst )
-{
-    assert( nm == MIN(m, n));
-    Mat _u(u, false), _w(w, false), _vt(vt, false), _rhs(rhs, false), _dst(dst, false);
-    SVD::backSubst(_w, _u, _vt, _rhs, _dst);
-    CV_Assert(_dst.data == (uchar*)&dst.val[0]);
-}
-
 ///////////////////////////////// Mat_<_Tp> ////////////////////////////////////
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_()
-    : Mat() { flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<_Tp>::type; }
+/*!
+ Template matrix class derived from Mat
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(int _rows, int _cols)
-    : Mat(_rows, _cols, DataType<_Tp>::type) {}
+ The class Mat_ is a "thin" template wrapper on top of cv::Mat. It does not have any extra data fields,
+ nor it or cv::Mat have any virtual methods and thus references or pointers to these two classes
+ can be safely converted one to another. But do it with care, for example:
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(int _rows, int _cols, const _Tp& value)
-    : Mat(_rows, _cols, DataType<_Tp>::type) { *this = value; }
+ \code
+ // create 100x100 8-bit matrix
+ Mat M(100,100,CV_8U);
+ // this will compile fine. no any data conversion will be done.
+ Mat_<float>& M1 = (Mat_<float>&)M;
+ // the program will likely crash at the statement below
+ M1(99,99) = 1.f;
+ \endcode
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(Size _sz)
-    : Mat(_sz.height, _sz.width, DataType<_Tp>::type) {}
+ While cv::Mat is sufficient in most cases, cv::Mat_ can be more convenient if you use a lot of element
+ access operations and if you know matrix type at compile time.
+ Note that cv::Mat::at<_Tp>(int y, int x) and cv::Mat_<_Tp>::operator ()(int y, int x) do absolutely the
+ same thing and run at the same speed, but the latter is certainly shorter:
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(Size _sz, const _Tp& value)
-    : Mat(_sz.height, _sz.width, DataType<_Tp>::type) { *this = value; }
+ \code
+ Mat_<double> M(20,20);
+ for(int i = 0; i < M.rows; i++)
+    for(int j = 0; j < M.cols; j++)
+       M(i,j) = 1./(i+j+1);
+ Mat E, V;
+ eigen(M,E,V);
+ cout << E.at<double>(0,0)/E.at<double>(M.rows-1,0);
+ \endcode
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(int _dims, const int* _sz)
-    : Mat(_dims, _sz, DataType<_Tp>::type) {}
+ It is easy to use Mat_ for multi-channel images/matrices - just pass cv::Vec as cv::Mat_ template parameter:
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(int _dims, const int* _sz, const _Tp& _s)
-    : Mat(_dims, _sz, DataType<_Tp>::type, Scalar(_s)) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Mat_<_Tp>& m, const Range* ranges)
-    : Mat(m, ranges) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Mat& m)
-    : Mat() { flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<_Tp>::type; *this = m; }
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Mat_& m)
-    : Mat(m) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(int _rows, int _cols, _Tp* _data, size_t steps)
-    : Mat(_rows, _cols, DataType<_Tp>::type, _data, steps) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Mat_& m, const Range& _rowRange, const Range& _colRange)
-    : Mat(m, _rowRange, _colRange) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Mat_& m, const Rect& roi)
-    : Mat(m, roi) {}
-
-template<typename _Tp> template<int n> inline
-    Mat_<_Tp>::Mat_(const Vec<typename DataType<_Tp>::channel_type, n>& vec, bool copyData)
-    : Mat(n/DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&vec)
+ \code
+ // allocate 320x240 color image and fill it with green (in RGB space)
+ Mat_<Vec3b> img(240, 320, Vec3b(0,255,0));
+ // now draw a diagonal white line
+ for(int i = 0; i < 100; i++)
+     img(i,i)=Vec3b(255,255,255);
+ // and now modify the 2nd (red) channel of each pixel
+ for(int i = 0; i < img.rows; i++)
+    for(int j = 0; j < img.cols; j++)
+       img(i,j)[2] ^= (uchar)(i ^ j); // img(y,x)[c] accesses c-th channel of the pixel (x,y)
+ \endcode
+*/
+template<typename _Tp> class Mat_ : public Mat
 {
-    CV_Assert(n%DataType<_Tp>::channels == 0);
-    if( copyData )
-        *this = clone();
-}
+public:
+    typedef _Tp value_type;
+    typedef typename DataType<_Tp>::channel_type channel_type;
+    typedef MatIterator_<_Tp> iterator;
+    typedef MatConstIterator_<_Tp> const_iterator;
 
-template<typename _Tp> template<int m, int n> inline
-    Mat_<_Tp>::Mat_(const Matx<typename DataType<_Tp>::channel_type,m,n>& M, bool copyData)
-    : Mat(m, n/DataType<_Tp>::channels, DataType<_Tp>::type, (void*)&M)
+    //! default constructor
+    Mat_();
+    //! equivalent to Mat(_rows, _cols, DataType<_Tp>::type)
+    Mat_(int _rows, int _cols);
+    //! constructor that sets each matrix element to specified value
+    Mat_(int _rows, int _cols, const _Tp& value);
+    //! equivalent to Mat(_size, DataType<_Tp>::type)
+    explicit Mat_(Size _size);
+    //! constructor that sets each matrix element to specified value
+    Mat_(Size _size, const _Tp& value);
+    //! n-dim array constructor
+    Mat_(int _ndims, const int* _sizes);
+    //! n-dim array constructor that sets each matrix element to specified value
+    Mat_(int _ndims, const int* _sizes, const _Tp& value);
+    //! copy/conversion contructor. If m is of different type, it's converted
+    Mat_(const Mat& m);
+    //! copy constructor
+    Mat_(const Mat_& m);
+    //! constructs a matrix on top of user-allocated data. step is in bytes(!!!), regardless of the type
+    Mat_(int _rows, int _cols, _Tp* _data, size_t _step=AUTO_STEP);
+    //! constructs n-dim matrix on top of user-allocated data. steps are in bytes(!!!), regardless of the type
+    Mat_(int _ndims, const int* _sizes, _Tp* _data, const size_t* _steps=0);
+    //! selects a submatrix
+    Mat_(const Mat_& m, const Range& rowRange, const Range& colRange=Range::all());
+    //! selects a submatrix
+    Mat_(const Mat_& m, const Rect& roi);
+    //! selects a submatrix, n-dim version
+    Mat_(const Mat_& m, const Range* ranges);
+    //! from a matrix expression
+    explicit Mat_(const MatExpr& e);
+    //! makes a matrix out of Vec, std::vector, Point_ or Point3_. The matrix will have a single column
+    explicit Mat_(const std::vector<_Tp>& vec, bool copyData=false);
+    template<int n> explicit Mat_(const Vec<typename DataType<_Tp>::channel_type, n>& vec, bool copyData=true);
+    template<int m, int n> explicit Mat_(const Matx<typename DataType<_Tp>::channel_type, m, n>& mtx, bool copyData=true);
+    explicit Mat_(const Point_<typename DataType<_Tp>::channel_type>& pt, bool copyData=true);
+    explicit Mat_(const Point3_<typename DataType<_Tp>::channel_type>& pt, bool copyData=true);
+    explicit Mat_(const MatCommaInitializer_<_Tp>& commaInitializer);
+
+    Mat_& operator = (const Mat& m);
+    Mat_& operator = (const Mat_& m);
+    //! set all the elements to s.
+    Mat_& operator = (const _Tp& s);
+    //! assign a matrix expression
+    Mat_& operator = (const MatExpr& e);
+
+    //! iterators; they are smart enough to skip gaps in the end of rows
+    iterator begin();
+    iterator end();
+    const_iterator begin() const;
+    const_iterator end() const;
+
+    //! equivalent to Mat::create(_rows, _cols, DataType<_Tp>::type)
+    void create(int _rows, int _cols);
+    //! equivalent to Mat::create(_size, DataType<_Tp>::type)
+    void create(Size _size);
+    //! equivalent to Mat::create(_ndims, _sizes, DatType<_Tp>::type)
+    void create(int _ndims, const int* _sizes);
+    //! cross-product
+    Mat_ cross(const Mat_& m) const;
+    //! data type conversion
+    template<typename T2> operator Mat_<T2>() const;
+    //! overridden forms of Mat::row() etc.
+    Mat_ row(int y) const;
+    Mat_ col(int x) const;
+    Mat_ diag(int d=0) const;
+    Mat_ clone() const;
+
+    //! overridden forms of Mat::elemSize() etc.
+    size_t elemSize() const;
+    size_t elemSize1() const;
+    int type() const;
+    int depth() const;
+    int channels() const;
+    size_t step1(int i=0) const;
+    //! returns step()/sizeof(_Tp)
+    size_t stepT(int i=0) const;
+
+    //! overridden forms of Mat::zeros() etc. Data type is omitted, of course
+    static MatExpr zeros(int rows, int cols);
+    static MatExpr zeros(Size size);
+    static MatExpr zeros(int _ndims, const int* _sizes);
+    static MatExpr ones(int rows, int cols);
+    static MatExpr ones(Size size);
+    static MatExpr ones(int _ndims, const int* _sizes);
+    static MatExpr eye(int rows, int cols);
+    static MatExpr eye(Size size);
+
+    //! some more overriden methods
+    Mat_& adjustROI( int dtop, int dbottom, int dleft, int dright );
+    Mat_ operator()( const Range& rowRange, const Range& colRange ) const;
+    Mat_ operator()( const Rect& roi ) const;
+    Mat_ operator()( const Range* ranges ) const;
+
+    //! more convenient forms of row and element access operators
+    _Tp* operator [](int y);
+    const _Tp* operator [](int y) const;
+
+    //! returns reference to the specified element
+    _Tp& operator ()(const int* idx);
+    //! returns read-only reference to the specified element
+    const _Tp& operator ()(const int* idx) const;
+
+    //! returns reference to the specified element
+    template<int n> _Tp& operator ()(const Vec<int, n>& idx);
+    //! returns read-only reference to the specified element
+    template<int n> const _Tp& operator ()(const Vec<int, n>& idx) const;
+
+    //! returns reference to the specified element (1D case)
+    _Tp& operator ()(int idx0);
+    //! returns read-only reference to the specified element (1D case)
+    const _Tp& operator ()(int idx0) const;
+    //! returns reference to the specified element (2D case)
+    _Tp& operator ()(int idx0, int idx1);
+    //! returns read-only reference to the specified element (2D case)
+    const _Tp& operator ()(int idx0, int idx1) const;
+    //! returns reference to the specified element (3D case)
+    _Tp& operator ()(int idx0, int idx1, int idx2);
+    //! returns read-only reference to the specified element (3D case)
+    const _Tp& operator ()(int idx0, int idx1, int idx2) const;
+
+    _Tp& operator ()(Point pt);
+    const _Tp& operator ()(Point pt) const;
+
+    //! conversion to vector.
+    operator std::vector<_Tp>() const;
+    //! conversion to Vec
+    template<int n> operator Vec<typename DataType<_Tp>::channel_type, n>() const;
+    //! conversion to Matx
+    template<int m, int n> operator Matx<typename DataType<_Tp>::channel_type, m, n>() const;
+};
+
+typedef Mat_<uchar> Mat1b;
+typedef Mat_<Vec2b> Mat2b;
+typedef Mat_<Vec3b> Mat3b;
+typedef Mat_<Vec4b> Mat4b;
+
+typedef Mat_<short> Mat1s;
+typedef Mat_<Vec2s> Mat2s;
+typedef Mat_<Vec3s> Mat3s;
+typedef Mat_<Vec4s> Mat4s;
+
+typedef Mat_<ushort> Mat1w;
+typedef Mat_<Vec2w> Mat2w;
+typedef Mat_<Vec3w> Mat3w;
+typedef Mat_<Vec4w> Mat4w;
+
+typedef Mat_<int>   Mat1i;
+typedef Mat_<Vec2i> Mat2i;
+typedef Mat_<Vec3i> Mat3i;
+typedef Mat_<Vec4i> Mat4i;
+
+typedef Mat_<float> Mat1f;
+typedef Mat_<Vec2f> Mat2f;
+typedef Mat_<Vec3f> Mat3f;
+typedef Mat_<Vec4f> Mat4f;
+
+typedef Mat_<double> Mat1d;
+typedef Mat_<Vec2d> Mat2d;
+typedef Mat_<Vec3d> Mat3d;
+typedef Mat_<Vec4d> Mat4d;
+
+
+
+/////////////////////////// multi-dimensional sparse matrix //////////////////////////
+
+/*!
+ Sparse matrix class.
+
+ The class represents multi-dimensional sparse numerical arrays. Such a sparse array can store elements
+ of any type that cv::Mat is able to store. "Sparse" means that only non-zero elements
+ are stored (though, as a result of some operations on a sparse matrix, some of its stored elements
+ can actually become 0. It's user responsibility to detect such elements and delete them using cv::SparseMat::erase().
+ The non-zero elements are stored in a hash table that grows when it's filled enough,
+ so that the search time remains O(1) in average. Elements can be accessed using the following methods:
+
+ <ol>
+ <li>Query operations: cv::SparseMat::ptr() and the higher-level cv::SparseMat::ref(),
+      cv::SparseMat::value() and cv::SparseMat::find, for example:
+ \code
+ const int dims = 5;
+ int size[] = {10, 10, 10, 10, 10};
+ SparseMat sparse_mat(dims, size, CV_32F);
+ for(int i = 0; i < 1000; i++)
+ {
+     int idx[dims];
+     for(int k = 0; k < dims; k++)
+        idx[k] = rand()%sparse_mat.size(k);
+     sparse_mat.ref<float>(idx) += 1.f;
+ }
+ \endcode
+
+ <li>Sparse matrix iterators. Like cv::Mat iterators and unlike cv::Mat iterators, the sparse matrix iterators are STL-style,
+ that is, the iteration is done as following:
+ \code
+ // prints elements of a sparse floating-point matrix and the sum of elements.
+ SparseMatConstIterator_<float>
+        it = sparse_mat.begin<float>(),
+        it_end = sparse_mat.end<float>();
+ double s = 0;
+ int dims = sparse_mat.dims();
+ for(; it != it_end; ++it)
+ {
+     // print element indices and the element value
+     const Node* n = it.node();
+     printf("(")
+     for(int i = 0; i < dims; i++)
+        printf("%3d%c", n->idx[i], i < dims-1 ? ',' : ')');
+     printf(": %f\n", *it);
+     s += *it;
+ }
+ printf("Element sum is %g\n", s);
+ \endcode
+ If you run this loop, you will notice that elements are enumerated
+ in no any logical order (lexicographical etc.),
+ they come in the same order as they stored in the hash table, i.e. semi-randomly.
+
+ You may collect pointers to the nodes and sort them to get the proper ordering.
+ Note, however, that pointers to the nodes may become invalid when you add more
+ elements to the matrix; this is because of possible buffer reallocation.
+
+ <li>A combination of the above 2 methods when you need to process 2 or more sparse
+ matrices simultaneously, e.g. this is how you can compute unnormalized
+ cross-correlation of the 2 floating-point sparse matrices:
+ \code
+ double crossCorr(const SparseMat& a, const SparseMat& b)
+ {
+     const SparseMat *_a = &a, *_b = &b;
+     // if b contains less elements than a,
+     // it's faster to iterate through b
+     if(_a->nzcount() > _b->nzcount())
+        std::swap(_a, _b);
+     SparseMatConstIterator_<float> it = _a->begin<float>(),
+                                    it_end = _a->end<float>();
+     double ccorr = 0;
+     for(; it != it_end; ++it)
+     {
+         // take the next element from the first matrix
+         float avalue = *it;
+         const Node* anode = it.node();
+         // and try to find element with the same index in the second matrix.
+         // since the hash value depends only on the element index,
+         // we reuse hashvalue stored in the node
+         float bvalue = _b->value<float>(anode->idx,&anode->hashval);
+         ccorr += avalue*bvalue;
+     }
+     return ccorr;
+ }
+ \endcode
+ </ol>
+*/
+class CV_EXPORTS SparseMat
 {
-    CV_Assert(n % DataType<_Tp>::channels == 0);
-    if( copyData )
-        *this = clone();
-}
+public:
+    typedef SparseMatIterator iterator;
+    typedef SparseMatConstIterator const_iterator;
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Point_<typename DataType<_Tp>::channel_type>& pt, bool copyData)
-    : Mat(2/DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&pt)
-{
-    CV_Assert(2 % DataType<_Tp>::channels == 0);
-    if( copyData )
-        *this = clone();
-}
+    enum { MAGIC_VAL=0x42FD0000, MAX_DIM=32, HASH_SCALE=0x5bd1e995, HASH_BIT=0x80000000 };
 
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const Point3_<typename DataType<_Tp>::channel_type>& pt, bool copyData)
-    : Mat(3/DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&pt)
-{
-    CV_Assert(3 % DataType<_Tp>::channels == 0);
-    if( copyData )
-        *this = clone();
-}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const MatCommaInitializer_<_Tp>& commaInitializer)
-    : Mat(commaInitializer) {}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const std::vector<_Tp>& vec, bool copyData)
-    : Mat(vec, copyData) {}
-
-template<typename _Tp> inline Mat_<_Tp>& Mat_<_Tp>::operator = (const Mat& m)
-{
-    if( DataType<_Tp>::type == m.type() )
+    //! the sparse matrix header
+    struct CV_EXPORTS Hdr
     {
-        Mat::operator = (m);
-        return *this;
-    }
-    if( DataType<_Tp>::depth == m.depth() )
+        Hdr(int _dims, const int* _sizes, int _type);
+        void clear();
+        int refcount;
+        int dims;
+        int valueOffset;
+        size_t nodeSize;
+        size_t nodeCount;
+        size_t freeList;
+        std::vector<uchar> pool;
+        std::vector<size_t> hashtab;
+        int size[MAX_DIM];
+    };
+
+    //! sparse matrix node - element of a hash table
+    struct CV_EXPORTS Node
     {
-        return (*this = m.reshape(DataType<_Tp>::channels, m.dims, 0));
-    }
-    CV_DbgAssert(DataType<_Tp>::channels == m.channels());
-    m.convertTo(*this, type());
-    return *this;
-}
+        //! hash value
+        size_t hashval;
+        //! index of the next node in the same hash table entry
+        size_t next;
+        //! index of the matrix element
+        int idx[MAX_DIM];
+    };
 
-template<typename _Tp> inline Mat_<_Tp>& Mat_<_Tp>::operator = (const Mat_& m)
+    //! default constructor
+    SparseMat();
+    //! creates matrix of the specified size and type
+    SparseMat(int dims, const int* _sizes, int _type);
+    //! copy constructor
+    SparseMat(const SparseMat& m);
+    //! converts dense 2d matrix to the sparse form
+    /*!
+     \param m the input matrix
+    */
+    explicit SparseMat(const Mat& m);
+    //! converts old-style sparse matrix to the new-style. All the data is copied
+    //SparseMat(const CvSparseMat* m);
+    //! the destructor
+    ~SparseMat();
+
+    //! assignment operator. This is O(1) operation, i.e. no data is copied
+    SparseMat& operator = (const SparseMat& m);
+    //! equivalent to the corresponding constructor
+    SparseMat& operator = (const Mat& m);
+
+    //! creates full copy of the matrix
+    SparseMat clone() const;
+
+    //! copies all the data to the destination matrix. All the previous content of m is erased
+    void copyTo( SparseMat& m ) const;
+    //! converts sparse matrix to dense matrix.
+    void copyTo( Mat& m ) const;
+    //! multiplies all the matrix elements by the specified scale factor alpha and converts the results to the specified data type
+    void convertTo( SparseMat& m, int rtype, double alpha=1 ) const;
+    //! converts sparse matrix to dense n-dim matrix with optional type conversion and scaling.
+    /*!
+      \param rtype The output matrix data type. When it is =-1, the output array will have the same data type as (*this)
+      \param alpha The scale factor
+      \param beta The optional delta added to the scaled values before the conversion
+    */
+    void convertTo( Mat& m, int rtype, double alpha=1, double beta=0 ) const;
+
+    // not used now
+    void assignTo( SparseMat& m, int type=-1 ) const;
+
+    //! reallocates sparse matrix.
+    /*!
+        If the matrix already had the proper size and type,
+        it is simply cleared with clear(), otherwise,
+        the old matrix is released (using release()) and the new one is allocated.
+    */
+    void create(int dims, const int* _sizes, int _type);
+    //! sets all the sparse matrix elements to 0, which means clearing the hash table.
+    void clear();
+    //! manually increments the reference counter to the header.
+    void addref();
+    // decrements the header reference counter. When the counter reaches 0, the header and all the underlying data are deallocated.
+    void release();
+
+    //! converts sparse matrix to the old-style representation; all the elements are copied.
+    //operator CvSparseMat*() const;
+    //! returns the size of each element in bytes (not including the overhead - the space occupied by SparseMat::Node elements)
+    size_t elemSize() const;
+    //! returns elemSize()/channels()
+    size_t elemSize1() const;
+
+    //! returns type of sparse matrix elements
+    int type() const;
+    //! returns the depth of sparse matrix elements
+    int depth() const;
+    //! returns the number of channels
+    int channels() const;
+
+    //! returns the array of sizes, or NULL if the matrix is not allocated
+    const int* size() const;
+    //! returns the size of i-th matrix dimension (or 0)
+    int size(int i) const;
+    //! returns the matrix dimensionality
+    int dims() const;
+    //! returns the number of non-zero elements (=the number of hash table nodes)
+    size_t nzcount() const;
+
+    //! computes the element hash value (1D case)
+    size_t hash(int i0) const;
+    //! computes the element hash value (2D case)
+    size_t hash(int i0, int i1) const;
+    //! computes the element hash value (3D case)
+    size_t hash(int i0, int i1, int i2) const;
+    //! computes the element hash value (nD case)
+    size_t hash(const int* idx) const;
+
+    //@{
+    /*!
+     specialized variants for 1D, 2D, 3D cases and the generic_type one for n-D case.
+
+     return pointer to the matrix element.
+     <ul>
+      <li>if the element is there (it's non-zero), the pointer to it is returned
+      <li>if it's not there and createMissing=false, NULL pointer is returned
+      <li>if it's not there and createMissing=true, then the new element
+        is created and initialized with 0. Pointer to it is returned
+      <li>if the optional hashval pointer is not NULL, the element hash value is
+      not computed, but *hashval is taken instead.
+     </ul>
+    */
+    //! returns pointer to the specified element (1D case)
+    uchar* ptr(int i0, bool createMissing, size_t* hashval=0);
+    //! returns pointer to the specified element (2D case)
+    uchar* ptr(int i0, int i1, bool createMissing, size_t* hashval=0);
+    //! returns pointer to the specified element (3D case)
+    uchar* ptr(int i0, int i1, int i2, bool createMissing, size_t* hashval=0);
+    //! returns pointer to the specified element (nD case)
+    uchar* ptr(const int* idx, bool createMissing, size_t* hashval=0);
+    //@}
+
+    //@{
+    /*!
+     return read-write reference to the specified sparse matrix element.
+
+     ref<_Tp>(i0,...[,hashval]) is equivalent to *(_Tp*)ptr(i0,...,true[,hashval]).
+     The methods always return a valid reference.
+     If the element did not exist, it is created and initialiazed with 0.
+    */
+    //! returns reference to the specified element (1D case)
+    template<typename _Tp> _Tp& ref(int i0, size_t* hashval=0);
+    //! returns reference to the specified element (2D case)
+    template<typename _Tp> _Tp& ref(int i0, int i1, size_t* hashval=0);
+    //! returns reference to the specified element (3D case)
+    template<typename _Tp> _Tp& ref(int i0, int i1, int i2, size_t* hashval=0);
+    //! returns reference to the specified element (nD case)
+    template<typename _Tp> _Tp& ref(const int* idx, size_t* hashval=0);
+    //@}
+
+    //@{
+    /*!
+     return value of the specified sparse matrix element.
+
+     value<_Tp>(i0,...[,hashval]) is equivalent
+
+     \code
+     { const _Tp* p = find<_Tp>(i0,...[,hashval]); return p ? *p : _Tp(); }
+     \endcode
+
+     That is, if the element did not exist, the methods return 0.
+     */
+    //! returns value of the specified element (1D case)
+    template<typename _Tp> _Tp value(int i0, size_t* hashval=0) const;
+    //! returns value of the specified element (2D case)
+    template<typename _Tp> _Tp value(int i0, int i1, size_t* hashval=0) const;
+    //! returns value of the specified element (3D case)
+    template<typename _Tp> _Tp value(int i0, int i1, int i2, size_t* hashval=0) const;
+    //! returns value of the specified element (nD case)
+    template<typename _Tp> _Tp value(const int* idx, size_t* hashval=0) const;
+    //@}
+
+    //@{
+    /*!
+     Return pointer to the specified sparse matrix element if it exists
+
+     find<_Tp>(i0,...[,hashval]) is equivalent to (_const Tp*)ptr(i0,...false[,hashval]).
+
+     If the specified element does not exist, the methods return NULL.
+    */
+    //! returns pointer to the specified element (1D case)
+    template<typename _Tp> const _Tp* find(int i0, size_t* hashval=0) const;
+    //! returns pointer to the specified element (2D case)
+    template<typename _Tp> const _Tp* find(int i0, int i1, size_t* hashval=0) const;
+    //! returns pointer to the specified element (3D case)
+    template<typename _Tp> const _Tp* find(int i0, int i1, int i2, size_t* hashval=0) const;
+    //! returns pointer to the specified element (nD case)
+    template<typename _Tp> const _Tp* find(const int* idx, size_t* hashval=0) const;
+
+    //! erases the specified element (2D case)
+    void erase(int i0, int i1, size_t* hashval=0);
+    //! erases the specified element (3D case)
+    void erase(int i0, int i1, int i2, size_t* hashval=0);
+    //! erases the specified element (nD case)
+    void erase(const int* idx, size_t* hashval=0);
+
+    //@{
+    /*!
+       return the sparse matrix iterator pointing to the first sparse matrix element
+    */
+    //! returns the sparse matrix iterator at the matrix beginning
+    SparseMatIterator begin();
+    //! returns the sparse matrix iterator at the matrix beginning
+    template<typename _Tp> SparseMatIterator_<_Tp> begin();
+    //! returns the read-only sparse matrix iterator at the matrix beginning
+    SparseMatConstIterator begin() const;
+    //! returns the read-only sparse matrix iterator at the matrix beginning
+    template<typename _Tp> SparseMatConstIterator_<_Tp> begin() const;
+    //@}
+    /*!
+       return the sparse matrix iterator pointing to the element following the last sparse matrix element
+    */
+    //! returns the sparse matrix iterator at the matrix end
+    SparseMatIterator end();
+    //! returns the read-only sparse matrix iterator at the matrix end
+    SparseMatConstIterator end() const;
+    //! returns the typed sparse matrix iterator at the matrix end
+    template<typename _Tp> SparseMatIterator_<_Tp> end();
+    //! returns the typed read-only sparse matrix iterator at the matrix end
+    template<typename _Tp> SparseMatConstIterator_<_Tp> end() const;
+
+    //! returns the value stored in the sparse martix node
+    template<typename _Tp> _Tp& value(Node* n);
+    //! returns the value stored in the sparse martix node
+    template<typename _Tp> const _Tp& value(const Node* n) const;
+
+    ////////////// some internal-use methods ///////////////
+    Node* node(size_t nidx);
+    const Node* node(size_t nidx) const;
+
+    uchar* newNode(const int* idx, size_t hashval);
+    void removeNode(size_t hidx, size_t nidx, size_t previdx);
+    void resizeHashTab(size_t newsize);
+
+    int flags;
+    Hdr* hdr;
+};
+
+
+
+///////////////////////////////// SparseMat_<_Tp> ////////////////////////////////////
+
+/*!
+ The Template Sparse Matrix class derived from cv::SparseMat
+
+ The class provides slightly more convenient operations for accessing elements.
+
+ \code
+ SparseMat m;
+ ...
+ SparseMat_<int> m_ = (SparseMat_<int>&)m;
+ m_.ref(1)++; // equivalent to m.ref<int>(1)++;
+ m_.ref(2) += m_(3); // equivalent to m.ref<int>(2) += m.value<int>(3);
+ \endcode
+*/
+template<typename _Tp> class SparseMat_ : public SparseMat
 {
-    Mat::operator=(m);
-    return *this;
-}
+public:
+    typedef SparseMatIterator_<_Tp> iterator;
+    typedef SparseMatConstIterator_<_Tp> const_iterator;
 
-template<typename _Tp> inline Mat_<_Tp>& Mat_<_Tp>::operator = (const _Tp& s)
+    //! the default constructor
+    SparseMat_();
+    //! the full constructor equivelent to SparseMat(dims, _sizes, DataType<_Tp>::type)
+    SparseMat_(int dims, const int* _sizes);
+    //! the copy constructor. If DataType<_Tp>.type != m.type(), the m elements are converted
+    SparseMat_(const SparseMat& m);
+    //! the copy constructor. This is O(1) operation - no data is copied
+    SparseMat_(const SparseMat_& m);
+    //! converts dense matrix to the sparse form
+    SparseMat_(const Mat& m);
+    //! converts the old-style sparse matrix to the C++ class. All the elements are copied
+    //SparseMat_(const CvSparseMat* m);
+    //! the assignment operator. If DataType<_Tp>.type != m.type(), the m elements are converted
+    SparseMat_& operator = (const SparseMat& m);
+    //! the assignment operator. This is O(1) operation - no data is copied
+    SparseMat_& operator = (const SparseMat_& m);
+    //! converts dense matrix to the sparse form
+    SparseMat_& operator = (const Mat& m);
+
+    //! makes full copy of the matrix. All the elements are duplicated
+    SparseMat_ clone() const;
+    //! equivalent to cv::SparseMat::create(dims, _sizes, DataType<_Tp>::type)
+    void create(int dims, const int* _sizes);
+    //! converts sparse matrix to the old-style CvSparseMat. All the elements are copied
+    //operator CvSparseMat*() const;
+
+    //! returns type of the matrix elements
+    int type() const;
+    //! returns depth of the matrix elements
+    int depth() const;
+    //! returns the number of channels in each matrix element
+    int channels() const;
+
+    //! equivalent to SparseMat::ref<_Tp>(i0, hashval)
+    _Tp& ref(int i0, size_t* hashval=0);
+    //! equivalent to SparseMat::ref<_Tp>(i0, i1, hashval)
+    _Tp& ref(int i0, int i1, size_t* hashval=0);
+    //! equivalent to SparseMat::ref<_Tp>(i0, i1, i2, hashval)
+    _Tp& ref(int i0, int i1, int i2, size_t* hashval=0);
+    //! equivalent to SparseMat::ref<_Tp>(idx, hashval)
+    _Tp& ref(const int* idx, size_t* hashval=0);
+
+    //! equivalent to SparseMat::value<_Tp>(i0, hashval)
+    _Tp operator()(int i0, size_t* hashval=0) const;
+    //! equivalent to SparseMat::value<_Tp>(i0, i1, hashval)
+    _Tp operator()(int i0, int i1, size_t* hashval=0) const;
+    //! equivalent to SparseMat::value<_Tp>(i0, i1, i2, hashval)
+    _Tp operator()(int i0, int i1, int i2, size_t* hashval=0) const;
+    //! equivalent to SparseMat::value<_Tp>(idx, hashval)
+    _Tp operator()(const int* idx, size_t* hashval=0) const;
+
+    //! returns sparse matrix iterator pointing to the first sparse matrix element
+    SparseMatIterator_<_Tp> begin();
+    //! returns read-only sparse matrix iterator pointing to the first sparse matrix element
+    SparseMatConstIterator_<_Tp> begin() const;
+    //! returns sparse matrix iterator pointing to the element following the last sparse matrix element
+    SparseMatIterator_<_Tp> end();
+    //! returns read-only sparse matrix iterator pointing to the element following the last sparse matrix element
+    SparseMatConstIterator_<_Tp> end() const;
+};
+
+
+
+////////////////////////////////// MatConstIterator //////////////////////////////////
+
+class CV_EXPORTS MatConstIterator
 {
-    typedef typename DataType<_Tp>::vec_type VT;
-    Mat::operator=(Scalar((const VT&)s));
-    return *this;
-}
+public:
+    typedef uchar* value_type;
+    typedef ptrdiff_t difference_type;
+    typedef const uchar** pointer;
+    typedef uchar* reference;
 
-template<typename _Tp> inline void Mat_<_Tp>::create(int _rows, int _cols)
+#ifndef OPENCV_NOSTL
+    typedef std::random_access_iterator_tag iterator_category;
+#endif
+
+    //! default constructor
+    MatConstIterator();
+    //! constructor that sets the iterator to the beginning of the matrix
+    MatConstIterator(const Mat* _m);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator(const Mat* _m, int _row, int _col=0);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator(const Mat* _m, Point _pt);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator(const Mat* _m, const int* _idx);
+    //! copy constructor
+    MatConstIterator(const MatConstIterator& it);
+
+    //! copy operator
+    MatConstIterator& operator = (const MatConstIterator& it);
+    //! returns the current matrix element
+    uchar* operator *() const;
+    //! returns the i-th matrix element, relative to the current
+    uchar* operator [](ptrdiff_t i) const;
+
+    //! shifts the iterator forward by the specified number of elements
+    MatConstIterator& operator += (ptrdiff_t ofs);
+    //! shifts the iterator backward by the specified number of elements
+    MatConstIterator& operator -= (ptrdiff_t ofs);
+    //! decrements the iterator
+    MatConstIterator& operator --();
+    //! decrements the iterator
+    MatConstIterator operator --(int);
+    //! increments the iterator
+    MatConstIterator& operator ++();
+    //! increments the iterator
+    MatConstIterator operator ++(int);
+    //! returns the current iterator position
+    Point pos() const;
+    //! returns the current iterator position
+    void pos(int* _idx) const;
+
+    ptrdiff_t lpos() const;
+    void seek(ptrdiff_t ofs, bool relative = false);
+    void seek(const int* _idx, bool relative = false);
+
+    const Mat* m;
+    size_t elemSize;
+    uchar* ptr;
+    uchar* sliceStart;
+    uchar* sliceEnd;
+};
+
+
+
+////////////////////////////////// MatConstIterator_ /////////////////////////////////
+
+/*!
+ Matrix read-only iterator
+ */
+template<typename _Tp>
+class MatConstIterator_ : public MatConstIterator
 {
-    Mat::create(_rows, _cols, DataType<_Tp>::type);
-}
+public:
+    typedef _Tp value_type;
+    typedef ptrdiff_t difference_type;
+    typedef const _Tp* pointer;
+    typedef const _Tp& reference;
 
-template<typename _Tp> inline void Mat_<_Tp>::create(Size _sz)
+#ifndef OPENCV_NOSTL
+    typedef std::random_access_iterator_tag iterator_category;
+#endif
+
+    //! default constructor
+    MatConstIterator_();
+    //! constructor that sets the iterator to the beginning of the matrix
+    MatConstIterator_(const Mat_<_Tp>* _m);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator_(const Mat_<_Tp>* _m, int _row, int _col=0);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator_(const Mat_<_Tp>* _m, Point _pt);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatConstIterator_(const Mat_<_Tp>* _m, const int* _idx);
+    //! copy constructor
+    MatConstIterator_(const MatConstIterator_& it);
+
+    //! copy operator
+    MatConstIterator_& operator = (const MatConstIterator_& it);
+    //! returns the current matrix element
+    _Tp operator *() const;
+    //! returns the i-th matrix element, relative to the current
+    _Tp operator [](ptrdiff_t i) const;
+
+    //! shifts the iterator forward by the specified number of elements
+    MatConstIterator_& operator += (ptrdiff_t ofs);
+    //! shifts the iterator backward by the specified number of elements
+    MatConstIterator_& operator -= (ptrdiff_t ofs);
+    //! decrements the iterator
+    MatConstIterator_& operator --();
+    //! decrements the iterator
+    MatConstIterator_ operator --(int);
+    //! increments the iterator
+    MatConstIterator_& operator ++();
+    //! increments the iterator
+    MatConstIterator_ operator ++(int);
+    //! returns the current iterator position
+    Point pos() const;
+};
+
+
+
+//////////////////////////////////// MatIterator_ ////////////////////////////////////
+
+/*!
+ Matrix read-write iterator
+*/
+template<typename _Tp>
+class MatIterator_ : public MatConstIterator_<_Tp>
 {
-    Mat::create(_sz, DataType<_Tp>::type);
-}
+public:
+    typedef _Tp* pointer;
+    typedef _Tp& reference;
 
-template<typename _Tp> inline void Mat_<_Tp>::create(int _dims, const int* _sz)
+#ifndef OPENCV_NOSTL
+    typedef std::random_access_iterator_tag iterator_category;
+#endif
+
+    //! the default constructor
+    MatIterator_();
+    //! constructor that sets the iterator to the beginning of the matrix
+    MatIterator_(Mat_<_Tp>* _m);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatIterator_(Mat_<_Tp>* _m, int _row, int _col=0);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatIterator_(const Mat_<_Tp>* _m, Point _pt);
+    //! constructor that sets the iterator to the specified element of the matrix
+    MatIterator_(const Mat_<_Tp>* _m, const int* _idx);
+    //! copy constructor
+    MatIterator_(const MatIterator_& it);
+    //! copy operator
+    MatIterator_& operator = (const MatIterator_<_Tp>& it );
+
+    //! returns the current matrix element
+    _Tp& operator *() const;
+    //! returns the i-th matrix element, relative to the current
+    _Tp& operator [](ptrdiff_t i) const;
+
+    //! shifts the iterator forward by the specified number of elements
+    MatIterator_& operator += (ptrdiff_t ofs);
+    //! shifts the iterator backward by the specified number of elements
+    MatIterator_& operator -= (ptrdiff_t ofs);
+    //! decrements the iterator
+    MatIterator_& operator --();
+    //! decrements the iterator
+    MatIterator_ operator --(int);
+    //! increments the iterator
+    MatIterator_& operator ++();
+    //! increments the iterator
+    MatIterator_ operator ++(int);
+};
+
+
+
+/////////////////////////////// SparseMatConstIterator ///////////////////////////////
+
+/*!
+ Read-Only Sparse Matrix Iterator.
+ Here is how to use the iterator to compute the sum of floating-point sparse matrix elements:
+
+ \code
+ SparseMatConstIterator it = m.begin(), it_end = m.end();
+ double s = 0;
+ CV_Assert( m.type() == CV_32F );
+ for( ; it != it_end; ++it )
+    s += it.value<float>();
+ \endcode
+*/
+class CV_EXPORTS SparseMatConstIterator
 {
-    Mat::create(_dims, _sz, DataType<_Tp>::type);
-}
+public:
+    //! the default constructor
+    SparseMatConstIterator();
+    //! the full constructor setting the iterator to the first sparse matrix element
+    SparseMatConstIterator(const SparseMat* _m);
+    //! the copy constructor
+    SparseMatConstIterator(const SparseMatConstIterator& it);
+
+    //! the assignment operator
+    SparseMatConstIterator& operator = (const SparseMatConstIterator& it);
+
+    //! template method returning the current matrix element
+    template<typename _Tp> const _Tp& value() const;
+    //! returns the current node of the sparse matrix. it.node->idx is the current element index
+    const SparseMat::Node* node() const;
+
+    //! moves iterator to the previous element
+    SparseMatConstIterator& operator --();
+    //! moves iterator to the previous element
+    SparseMatConstIterator operator --(int);
+    //! moves iterator to the next element
+    SparseMatConstIterator& operator ++();
+    //! moves iterator to the next element
+    SparseMatConstIterator operator ++(int);
+
+    //! moves iterator to the element after the last element
+    void seekEnd();
+
+    const SparseMat* m;
+    size_t hashidx;
+    uchar* ptr;
+};
 
 
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::cross(const Mat_& m) const
-{ return Mat_<_Tp>(Mat::cross(m)); }
 
-template<typename _Tp> template<typename T2> inline Mat_<_Tp>::operator Mat_<T2>() const
-{ return Mat_<T2>(*this); }
+////////////////////////////////// SparseMatIterator /////////////////////////////////
 
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::row(int y) const
-{ return Mat_(*this, Range(y, y+1), Range::all()); }
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::col(int x) const
-{ return Mat_(*this, Range::all(), Range(x, x+1)); }
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::diag(int d) const
-{ return Mat_(Mat::diag(d)); }
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::clone() const
-{ return Mat_(Mat::clone()); }
+/*!
+ Read-write Sparse Matrix Iterator
 
-template<typename _Tp> inline size_t Mat_<_Tp>::elemSize() const
+ The class is similar to cv::SparseMatConstIterator,
+ but can be used for in-place modification of the matrix elements.
+*/
+class CV_EXPORTS SparseMatIterator : public SparseMatConstIterator
 {
-    CV_DbgAssert( Mat::elemSize() == sizeof(_Tp) );
-    return sizeof(_Tp);
-}
+public:
+    //! the default constructor
+    SparseMatIterator();
+    //! the full constructor setting the iterator to the first sparse matrix element
+    SparseMatIterator(SparseMat* _m);
+    //! the full constructor setting the iterator to the specified sparse matrix element
+    SparseMatIterator(SparseMat* _m, const int* idx);
+    //! the copy constructor
+    SparseMatIterator(const SparseMatIterator& it);
 
-template<typename _Tp> inline size_t Mat_<_Tp>::elemSize1() const
+    //! the assignment operator
+    SparseMatIterator& operator = (const SparseMatIterator& it);
+    //! returns read-write reference to the current sparse matrix element
+    template<typename _Tp> _Tp& value() const;
+    //! returns pointer to the current sparse matrix node. it.node->idx is the index of the current element (do not modify it!)
+    SparseMat::Node* node() const;
+
+    //! moves iterator to the next element
+    SparseMatIterator& operator ++();
+    //! moves iterator to the next element
+    SparseMatIterator operator ++(int);
+};
+
+
+
+/////////////////////////////// SparseMatConstIterator_ //////////////////////////////
+
+/*!
+ Template Read-Only Sparse Matrix Iterator Class.
+
+ This is the derived from SparseMatConstIterator class that
+ introduces more convenient operator *() for accessing the current element.
+*/
+template<typename _Tp> class SparseMatConstIterator_ : public SparseMatConstIterator
 {
-    CV_DbgAssert( Mat::elemSize1() == sizeof(_Tp)/DataType<_Tp>::channels );
-    return sizeof(_Tp)/DataType<_Tp>::channels;
-}
-template<typename _Tp> inline int Mat_<_Tp>::type() const
+public:
+
+#ifndef OPENCV_NOSTL
+    typedef std::forward_iterator_tag iterator_category;
+#endif
+
+    //! the default constructor
+    SparseMatConstIterator_();
+    //! the full constructor setting the iterator to the first sparse matrix element
+    SparseMatConstIterator_(const SparseMat_<_Tp>* _m);
+    SparseMatConstIterator_(const SparseMat* _m);
+    //! the copy constructor
+    SparseMatConstIterator_(const SparseMatConstIterator_& it);
+
+    //! the assignment operator
+    SparseMatConstIterator_& operator = (const SparseMatConstIterator_& it);
+    //! the element access operator
+    const _Tp& operator *() const;
+
+    //! moves iterator to the next element
+    SparseMatConstIterator_& operator ++();
+    //! moves iterator to the next element
+    SparseMatConstIterator_ operator ++(int);
+};
+
+
+
+///////////////////////////////// SparseMatIterator_ /////////////////////////////////
+
+/*!
+ Template Read-Write Sparse Matrix Iterator Class.
+
+ This is the derived from cv::SparseMatConstIterator_ class that
+ introduces more convenient operator *() for accessing the current element.
+*/
+template<typename _Tp> class SparseMatIterator_ : public SparseMatConstIterator_<_Tp>
 {
-    CV_DbgAssert( Mat::type() == DataType<_Tp>::type );
-    return DataType<_Tp>::type;
-}
-template<typename _Tp> inline int Mat_<_Tp>::depth() const
+public:
+
+#ifndef OPENCV_NOSTL
+    typedef std::forward_iterator_tag iterator_category;
+#endif
+
+    //! the default constructor
+    SparseMatIterator_();
+    //! the full constructor setting the iterator to the first sparse matrix element
+    SparseMatIterator_(SparseMat_<_Tp>* _m);
+    SparseMatIterator_(SparseMat* _m);
+    //! the copy constructor
+    SparseMatIterator_(const SparseMatIterator_& it);
+
+    //! the assignment operator
+    SparseMatIterator_& operator = (const SparseMatIterator_& it);
+    //! returns the reference to the current element
+    _Tp& operator *() const;
+
+    //! moves the iterator to the next element
+    SparseMatIterator_& operator ++();
+    //! moves the iterator to the next element
+    SparseMatIterator_ operator ++(int);
+};
+
+
+
+/////////////////////////////////// NAryMatIterator //////////////////////////////////
+
+/*!
+ n-Dimensional Dense Matrix Iterator Class.
+
+ The class cv::NAryMatIterator is used for iterating over one or more n-dimensional dense arrays (cv::Mat's).
+
+ The iterator is completely different from cv::Mat_ and cv::SparseMat_ iterators.
+ It iterates through the slices (or planes), not the elements, where "slice" is a continuous part of the arrays.
+
+ Here is the example on how the iterator can be used to normalize 3D histogram:
+
+ \code
+ void normalizeColorHist(Mat& hist)
+ {
+ #if 1
+     // intialize iterator (the style is different from STL).
+     // after initialization the iterator will contain
+     // the number of slices or planes
+     // the iterator will go through
+     Mat* arrays[] = { &hist, 0 };
+     Mat planes[1];
+     NAryMatIterator it(arrays, planes);
+     double s = 0;
+     // iterate through the matrix. on each iteration
+     // it.planes[i] (of type Mat) will be set to the current plane of
+     // i-th n-dim matrix passed to the iterator constructor.
+     for(int p = 0; p < it.nplanes; p++, ++it)
+        s += sum(it.planes[0])[0];
+     it = NAryMatIterator(hist);
+     s = 1./s;
+     for(int p = 0; p < it.nplanes; p++, ++it)
+        it.planes[0] *= s;
+ #elif 1
+     // this is a shorter implementation of the above
+     // using built-in operations on Mat
+     double s = sum(hist)[0];
+     hist.convertTo(hist, hist.type(), 1./s, 0);
+ #else
+     // and this is even shorter one
+     // (assuming that the histogram elements are non-negative)
+     normalize(hist, hist, 1, 0, NORM_L1);
+ #endif
+ }
+ \endcode
+
+ You can iterate through several matrices simultaneously as long as they have the same geometry
+ (dimensionality and all the dimension sizes are the same), which is useful for binary
+ and n-ary operations on such matrices. Just pass those matrices to cv::MatNDIterator.
+ Then, during the iteration it.planes[0], it.planes[1], ... will
+ be the slices of the corresponding matrices
+*/
+class CV_EXPORTS NAryMatIterator
 {
-    CV_DbgAssert( Mat::depth() == DataType<_Tp>::depth );
-    return DataType<_Tp>::depth;
-}
-template<typename _Tp> inline int Mat_<_Tp>::channels() const
-{
-    CV_DbgAssert( Mat::channels() == DataType<_Tp>::channels );
-    return DataType<_Tp>::channels;
-}
-template<typename _Tp> inline size_t Mat_<_Tp>::stepT(int i) const { return step.p[i]/elemSize(); }
-template<typename _Tp> inline size_t Mat_<_Tp>::step1(int i) const { return step.p[i]/elemSize1(); }
+public:
+    //! the default constructor
+    NAryMatIterator();
+    //! the full constructor taking arbitrary number of n-dim matrices
+    NAryMatIterator(const Mat** arrays, uchar** ptrs, int narrays=-1);
+    //! the full constructor taking arbitrary number of n-dim matrices
+    NAryMatIterator(const Mat** arrays, Mat* planes, int narrays=-1);
+    //! the separate iterator initialization method
+    void init(const Mat** arrays, Mat* planes, uchar** ptrs, int narrays=-1);
 
-template<typename _Tp> inline Mat_<_Tp>& Mat_<_Tp>::adjustROI( int dtop, int dbottom, int dleft, int dright )
-{ return (Mat_<_Tp>&)(Mat::adjustROI(dtop, dbottom, dleft, dright));  }
+    //! proceeds to the next plane of every iterated matrix
+    NAryMatIterator& operator ++();
+    //! proceeds to the next plane of every iterated matrix (postfix increment operator)
+    NAryMatIterator operator ++(int);
 
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::operator()( const Range& _rowRange, const Range& _colRange ) const
-{ return Mat_<_Tp>(*this, _rowRange, _colRange); }
-
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::operator()( const Rect& roi ) const
-{ return Mat_<_Tp>(*this, roi); }
-
-template<typename _Tp> inline Mat_<_Tp> Mat_<_Tp>::operator()( const Range* ranges ) const
-{ return Mat_<_Tp>(*this, ranges); }
-
-template<typename _Tp> inline _Tp* Mat_<_Tp>::operator [](int y)
-{ return (_Tp*)ptr(y); }
-template<typename _Tp> inline const _Tp* Mat_<_Tp>::operator [](int y) const
-{ return (const _Tp*)ptr(y); }
-
-template<typename _Tp> inline _Tp& Mat_<_Tp>::operator ()(int i0, int i1)
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  type() == DataType<_Tp>::type );
-    return ((_Tp*)(data + step.p[0]*i0))[i1];
-}
-
-template<typename _Tp> inline const _Tp& Mat_<_Tp>::operator ()(int i0, int i1) const
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                  (unsigned)i0 < (unsigned)size.p[0] &&
-                  (unsigned)i1 < (unsigned)size.p[1] &&
-                  type() == DataType<_Tp>::type );
-    return ((const _Tp*)(data + step.p[0]*i0))[i1];
-}
-
-template<typename _Tp> inline _Tp& Mat_<_Tp>::operator ()(Point pt)
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                  (unsigned)pt.y < (unsigned)size.p[0] &&
-                  (unsigned)pt.x < (unsigned)size.p[1] &&
-                  type() == DataType<_Tp>::type );
-    return ((_Tp*)(data + step.p[0]*pt.y))[pt.x];
-}
-
-template<typename _Tp> inline const _Tp& Mat_<_Tp>::operator ()(Point pt) const
-{
-    CV_DbgAssert( dims <= 2 && data &&
-                  (unsigned)pt.y < (unsigned)size.p[0] &&
-                  (unsigned)pt.x < (unsigned)size.p[1] &&
-                 type() == DataType<_Tp>::type );
-    return ((const _Tp*)(data + step.p[0]*pt.y))[pt.x];
-}
-
-template<typename _Tp> inline _Tp& Mat_<_Tp>::operator ()(const int* idx)
-{
-    return Mat::at<_Tp>(idx);
-}
-
-template<typename _Tp> inline const _Tp& Mat_<_Tp>::operator ()(const int* idx) const
-{
-    return Mat::at<_Tp>(idx);
-}
-
-template<typename _Tp> template<int n> inline _Tp& Mat_<_Tp>::operator ()(const Vec<int, n>& idx)
-{
-    return Mat::at<_Tp>(idx);
-}
-
-template<typename _Tp> template<int n> inline const _Tp& Mat_<_Tp>::operator ()(const Vec<int, n>& idx) const
-{
-    return Mat::at<_Tp>(idx);
-}
-
-template<typename _Tp> inline _Tp& Mat_<_Tp>::operator ()(int i0)
-{
-    return this->at<_Tp>(i0);
-}
-
-template<typename _Tp> inline const _Tp& Mat_<_Tp>::operator ()(int i0) const
-{
-    return this->at<_Tp>(i0);
-}
-
-template<typename _Tp> inline _Tp& Mat_<_Tp>::operator ()(int i0, int i1, int i2)
-{
-    return this->at<_Tp>(i0, i1, i2);
-}
-
-template<typename _Tp> inline const _Tp& Mat_<_Tp>::operator ()(int i0, int i1, int i2) const
-{
-    return this->at<_Tp>(i0, i1, i2);
-}
+    //! the iterated arrays
+    const Mat** arrays;
+    //! the current planes
+    Mat* planes;
+    //! data pointers
+    uchar** ptrs;
+    //! the number of arrays
+    int narrays;
+    //! the number of hyper-planes that the iterator steps through
+    size_t nplanes;
+    //! the size of each segment (in elements)
+    size_t size;
+protected:
+    int iterdepth;
+    size_t idx;
+};
 
 
-template<typename _Tp> inline Mat_<_Tp>::operator std::vector<_Tp>() const
-{
-    std::vector<_Tp> v;
-    copyTo(v);
-    return v;
-}
 
-template<typename _Tp> template<int n> inline Mat_<_Tp>::operator Vec<typename DataType<_Tp>::channel_type, n>() const
-{
-    CV_Assert(n % DataType<_Tp>::channels == 0);
-    return this->Mat::operator Vec<typename DataType<_Tp>::channel_type, n>();
-}
-
-template<typename _Tp> template<int m, int n> inline Mat_<_Tp>::operator Matx<typename DataType<_Tp>::channel_type, m, n>() const
-{
-    CV_Assert(n % DataType<_Tp>::channels == 0);
-    return this->Mat::operator Matx<typename DataType<_Tp>::channel_type, m, n>();
-}
-
-template<typename T1, typename T2, typename Op> inline void
-process( const Mat_<T1>& m1, Mat_<T2>& m2, Op op )
-{
-    int y, x, rows = m1.rows, cols = m1.cols;
-
-    CV_DbgAssert( m1.size() == m2.size() );
-
-    for( y = 0; y < rows; y++ )
-    {
-        const T1* src = m1[y];
-        T2* dst = m2[y];
-
-        for( x = 0; x < cols; x++ )
-            dst[x] = op(src[x]);
-    }
-}
-
-template<typename T1, typename T2, typename T3, typename Op> inline void
-process( const Mat_<T1>& m1, const Mat_<T2>& m2, Mat_<T3>& m3, Op op )
-{
-    int y, x, rows = m1.rows, cols = m1.cols;
-
-    CV_DbgAssert( m1.size() == m2.size() );
-
-    for( y = 0; y < rows; y++ )
-    {
-        const T1* src1 = m1[y];
-        const T2* src2 = m2[y];
-        T3* dst = m3[y];
-
-        for( x = 0; x < cols; x++ )
-            dst[x] = op( src1[x], src2[x] );
-    }
-}
-
-
-/////////////////////////////// Input/Output Arrays /////////////////////////////////
-
-template<typename _Tp> inline _InputArray::_InputArray(const std::vector<_Tp>& vec)
-    : flags(FIXED_TYPE + STD_VECTOR + DataType<_Tp>::type), obj((void*)&vec) {}
-
-template<typename _Tp> inline _InputArray::_InputArray(const std::vector<std::vector<_Tp> >& vec)
-    : flags(FIXED_TYPE + STD_VECTOR_VECTOR + DataType<_Tp>::type), obj((void*)&vec) {}
-
-template<typename _Tp> inline _InputArray::_InputArray(const std::vector<Mat_<_Tp> >& vec)
-    : flags(FIXED_TYPE + STD_VECTOR_MAT + DataType<_Tp>::type), obj((void*)&vec) {}
-
-template<typename _Tp, int m, int n> inline _InputArray::_InputArray(const Matx<_Tp, m, n>& mtx)
-    : flags(FIXED_TYPE + FIXED_SIZE + MATX + DataType<_Tp>::type), obj((void*)&mtx), sz(n, m) {}
-
-template<typename _Tp> inline _InputArray::_InputArray(const _Tp* vec, int n)
-    : flags(FIXED_TYPE + FIXED_SIZE + MATX + DataType<_Tp>::type), obj((void*)vec), sz(n, 1) {}
-
-inline _InputArray::_InputArray(const Scalar& s)
-    : flags(FIXED_TYPE + FIXED_SIZE + MATX + CV_64F), obj((void*)&s), sz(1, 4) {}
-
-template<typename _Tp> inline _InputArray::_InputArray(const Mat_<_Tp>& m)
-    : flags(FIXED_TYPE + MAT + DataType<_Tp>::type), obj((void*)&m) {}
-
-template<typename _Tp> inline _OutputArray::_OutputArray(std::vector<_Tp>& vec)
-    : _InputArray(vec) {}
-template<typename _Tp> inline _OutputArray::_OutputArray(std::vector<std::vector<_Tp> >& vec)
-    : _InputArray(vec) {}
-template<typename _Tp> inline _OutputArray::_OutputArray(std::vector<Mat_<_Tp> >& vec)
-    : _InputArray(vec) {}
-template<typename _Tp> inline _OutputArray::_OutputArray(Mat_<_Tp>& m)
-    : _InputArray(m) {}
-template<typename _Tp, int m, int n> inline _OutputArray::_OutputArray(Matx<_Tp, m, n>& mtx)
-    : _InputArray(mtx) {}
-template<typename _Tp> inline _OutputArray::_OutputArray(_Tp* vec, int n)
-    : _InputArray(vec, n) {}
-
-template<typename _Tp> inline _OutputArray::_OutputArray(const std::vector<_Tp>& vec)
-    : _InputArray(vec) {flags |= FIXED_SIZE;}
-template<typename _Tp> inline _OutputArray::_OutputArray(const std::vector<std::vector<_Tp> >& vec)
-    : _InputArray(vec) {flags |= FIXED_SIZE;}
-template<typename _Tp> inline _OutputArray::_OutputArray(const std::vector<Mat_<_Tp> >& vec)
-    : _InputArray(vec) {flags |= FIXED_SIZE;}
-
-template<typename _Tp> inline _OutputArray::_OutputArray(const Mat_<_Tp>& m)
-    : _InputArray(m) {flags |= FIXED_SIZE;}
-template<typename _Tp, int m, int n> inline _OutputArray::_OutputArray(const Matx<_Tp, m, n>& mtx)
-    : _InputArray(mtx) {}
-template<typename _Tp> inline _OutputArray::_OutputArray(const _Tp* vec, int n)
-    : _InputArray(vec, n) {}
-
-//////////////////////////////////// Matrix Expressions /////////////////////////////////////////
+///////////////////////////////// Matrix Expressions /////////////////////////////////
 
 class CV_EXPORTS MatOp
 {
 public:
-    MatOp() {};
-    virtual ~MatOp() {};
+    MatOp();
+    virtual ~MatOp();
 
     virtual bool elementWise(const MatExpr& expr) const;
     virtual void assign(const MatExpr& expr, Mat& m, int type=-1) const = 0;
@@ -1211,41 +1894,31 @@ public:
 class CV_EXPORTS MatExpr
 {
 public:
-    MatExpr() : op(0), flags(0), a(Mat()), b(Mat()), c(Mat()), alpha(0), beta(0), s(Scalar()) {}
-    MatExpr(const MatOp* _op, int _flags, const Mat& _a=Mat(), const Mat& _b=Mat(),
-            const Mat& _c=Mat(), double _alpha=1, double _beta=1, const Scalar& _s=Scalar())
-        : op(_op), flags(_flags), a(_a), b(_b), c(_c), alpha(_alpha), beta(_beta), s(_s) {}
+    MatExpr();
     explicit MatExpr(const Mat& m);
-    operator Mat() const
-    {
-        Mat m;
-        op->assign(*this, m);
-        return m;
-    }
 
-    template<typename _Tp> operator Mat_<_Tp>() const
-    {
-        Mat_<_Tp> m;
-        op->assign(*this, m, DataType<_Tp>::type);
-        return m;
-    }
+    MatExpr(const MatOp* _op, int _flags, const Mat& _a = Mat(), const Mat& _b = Mat(),
+            const Mat& _c = Mat(), double _alpha = 1, double _beta = 1, const Scalar& _s = Scalar());
+
+    operator Mat() const;
+    template<typename _Tp> operator Mat_<_Tp>() const;
+
+    Size size() const;
+    int type() const;
 
     MatExpr row(int y) const;
     MatExpr col(int x) const;
-    MatExpr diag(int d=0) const;
+    MatExpr diag(int d = 0) const;
     MatExpr operator()( const Range& rowRange, const Range& colRange ) const;
     MatExpr operator()( const Rect& roi ) const;
 
-    Mat cross(const Mat& m) const;
-    double dot(const Mat& m) const;
-
     MatExpr t() const;
     MatExpr inv(int method = DECOMP_LU) const;
     MatExpr mul(const MatExpr& e, double scale=1) const;
     MatExpr mul(const Mat& m, double scale=1) const;
 
-    Size size() const;
-    int type() const;
+    Mat cross(const Mat& m) const;
+    double dot(const Mat& m) const;
 
     const MatOp* op;
     int flags;
@@ -1319,75 +1992,6 @@ CV_EXPORTS MatExpr operator > (const Mat& a, const Mat& b);
 CV_EXPORTS MatExpr operator > (const Mat& a, double s);
 CV_EXPORTS MatExpr operator > (double s, const Mat& a);
 
-CV_EXPORTS MatExpr min(const Mat& a, const Mat& b);
-CV_EXPORTS MatExpr min(const Mat& a, double s);
-CV_EXPORTS MatExpr min(double s, const Mat& a);
-
-CV_EXPORTS MatExpr max(const Mat& a, const Mat& b);
-CV_EXPORTS MatExpr max(const Mat& a, double s);
-CV_EXPORTS MatExpr max(double s, const Mat& a);
-
-template<typename _Tp> static inline MatExpr min(const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    return cv::min((const Mat&)a, (const Mat&)b);
-}
-
-template<typename _Tp> static inline MatExpr min(const Mat_<_Tp>& a, double s)
-{
-    return cv::min((const Mat&)a, s);
-}
-
-template<typename _Tp> static inline MatExpr min(double s, const Mat_<_Tp>& a)
-{
-    return cv::min((const Mat&)a, s);
-}
-
-template<typename _Tp> static inline MatExpr max(const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    return cv::max((const Mat&)a, (const Mat&)b);
-}
-
-template<typename _Tp> static inline MatExpr max(const Mat_<_Tp>& a, double s)
-{
-    return cv::max((const Mat&)a, s);
-}
-
-template<typename _Tp> static inline MatExpr max(double s, const Mat_<_Tp>& a)
-{
-    return cv::max((const Mat&)a, s);
-}
-
-template<typename _Tp> static inline void min(const Mat_<_Tp>& a, const Mat_<_Tp>& b, Mat_<_Tp>& c)
-{
-    cv::min((const Mat&)a, (const Mat&)b, (Mat&)c);
-}
-
-template<typename _Tp> static inline void min(const Mat_<_Tp>& a, double s, Mat_<_Tp>& c)
-{
-    cv::min((const Mat&)a, s, (Mat&)c);
-}
-
-template<typename _Tp> static inline void min(double s, const Mat_<_Tp>& a, Mat_<_Tp>& c)
-{
-    cv::min((const Mat&)a, s, (Mat&)c);
-}
-
-template<typename _Tp> static inline void max(const Mat_<_Tp>& a, const Mat_<_Tp>& b, Mat_<_Tp>& c)
-{
-    cv::max((const Mat&)a, (const Mat&)b, (Mat&)c);
-}
-
-template<typename _Tp> static inline void max(const Mat_<_Tp>& a, double s, Mat_<_Tp>& c)
-{
-    cv::max((const Mat&)a, s, (Mat&)c);
-}
-
-template<typename _Tp> static inline void max(double s, const Mat_<_Tp>& a, Mat_<_Tp>& c)
-{
-    cv::max((const Mat&)a, s, (Mat&)c);
-}
-
-
 CV_EXPORTS MatExpr operator & (const Mat& a, const Mat& b);
 CV_EXPORTS MatExpr operator & (const Mat& a, const Scalar& s);
 CV_EXPORTS MatExpr operator & (const Scalar& s, const Mat& a);
@@ -1402,1204 +2006,19 @@ CV_EXPORTS MatExpr operator ^ (const Scalar& s, const Mat& a);
 
 CV_EXPORTS MatExpr operator ~(const Mat& m);
 
+CV_EXPORTS MatExpr min(const Mat& a, const Mat& b);
+CV_EXPORTS MatExpr min(const Mat& a, double s);
+CV_EXPORTS MatExpr min(double s, const Mat& a);
+
+CV_EXPORTS MatExpr max(const Mat& a, const Mat& b);
+CV_EXPORTS MatExpr max(const Mat& a, double s);
+CV_EXPORTS MatExpr max(double s, const Mat& a);
+
 CV_EXPORTS MatExpr abs(const Mat& m);
 CV_EXPORTS MatExpr abs(const MatExpr& e);
 
-template<typename _Tp> static inline MatExpr abs(const Mat_<_Tp>& m)
-{
-    return cv::abs((const Mat&)m);
-}
+} // cv
 
-////////////////////////////// Augmenting algebraic operations //////////////////////////////////
+#include "opencv2/core/mat.inl.hpp"
 
-inline Mat& Mat::operator = (const MatExpr& e)
-{
-    e.op->assign(e, *this);
-    return *this;
-}
-
-template<typename _Tp> inline Mat_<_Tp>::Mat_(const MatExpr& e)
-{
-    e.op->assign(e, *this, DataType<_Tp>::type);
-}
-
-template<typename _Tp> Mat_<_Tp>& Mat_<_Tp>::operator = (const MatExpr& e)
-{
-    e.op->assign(e, *this, DataType<_Tp>::type);
-    return *this;
-}
-
-static inline Mat& operator += (const Mat& a, const Mat& b)
-{
-    add(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator += (const Mat& a, const Scalar& s)
-{
-    add(a, s, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator += (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    add(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator += (const Mat_<_Tp>& a, const Scalar& s)
-{
-    add(a, s, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator += (const Mat& a, const MatExpr& b)
-{
-    b.op->augAssignAdd(b, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator += (const Mat_<_Tp>& a, const MatExpr& b)
-{
-    b.op->augAssignAdd(b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator -= (const Mat& a, const Mat& b)
-{
-    subtract(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator -= (const Mat& a, const Scalar& s)
-{
-    subtract(a, s, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator -= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    subtract(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator -= (const Mat_<_Tp>& a, const Scalar& s)
-{
-    subtract(a, s, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator -= (const Mat& a, const MatExpr& b)
-{
-    b.op->augAssignSubtract(b, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator -= (const Mat_<_Tp>& a, const MatExpr& b)
-{
-    b.op->augAssignSubtract(b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator *= (const Mat& a, const Mat& b)
-{
-    gemm(a, b, 1, Mat(), 0, (Mat&)a, 0);
-    return (Mat&)a;
-}
-
-static inline Mat& operator *= (const Mat& a, double s)
-{
-    a.convertTo((Mat&)a, -1, s);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator *= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    gemm(a, b, 1, Mat(), 0, (Mat&)a, 0);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator *= (const Mat_<_Tp>& a, double s)
-{
-    a.convertTo((Mat&)a, -1, s);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator *= (const Mat& a, const MatExpr& b)
-{
-    b.op->augAssignMultiply(b, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator *= (const Mat_<_Tp>& a, const MatExpr& b)
-{
-    b.op->augAssignMultiply(b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator /= (const Mat& a, const Mat& b)
-{
-    divide(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator /= (const Mat& a, double s)
-{
-    a.convertTo((Mat&)a, -1, 1./s);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator /= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    divide(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator /= (const Mat_<_Tp>& a, double s)
-{
-    a.convertTo((Mat&)a, -1, 1./s);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator /= (const Mat& a, const MatExpr& b)
-{
-    b.op->augAssignDivide(b, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline
-Mat_<_Tp>& operator /= (const Mat_<_Tp>& a, const MatExpr& b)
-{
-    b.op->augAssignDivide(b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-////////////////////////////// Logical operations ///////////////////////////////
-
-static inline Mat& operator &= (const Mat& a, const Mat& b)
-{
-    bitwise_and(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator &= (const Mat& a, const Scalar& s)
-{
-    bitwise_and(a, s, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator &= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    bitwise_and(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator &= (const Mat_<_Tp>& a, const Scalar& s)
-{
-    bitwise_and(a, s, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator |= (const Mat& a, const Mat& b)
-{
-    bitwise_or(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator |= (const Mat& a, const Scalar& s)
-{
-    bitwise_or(a, s, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator |= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    bitwise_or(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator |= (const Mat_<_Tp>& a, const Scalar& s)
-{
-    bitwise_or(a, s, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-static inline Mat& operator ^= (const Mat& a, const Mat& b)
-{
-    bitwise_xor(a, b, (Mat&)a);
-    return (Mat&)a;
-}
-
-static inline Mat& operator ^= (const Mat& a, const Scalar& s)
-{
-    bitwise_xor(a, s, (Mat&)a);
-    return (Mat&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator ^= (const Mat_<_Tp>& a, const Mat_<_Tp>& b)
-{
-    bitwise_xor(a, b, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-template<typename _Tp> static inline Mat_<_Tp>&
-operator ^= (const Mat_<_Tp>& a, const Scalar& s)
-{
-    bitwise_xor(a, s, (Mat&)a);
-    return (Mat_<_Tp>&)a;
-}
-
-/////////////////////////////// Miscellaneous operations //////////////////////////////
-
-template<typename _Tp> void split(const Mat& src, std::vector<Mat_<_Tp> >& mv)
-{ split(src, (std::vector<Mat>&)mv ); }
-
-//////////////////////////////////////////////////////////////
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::zeros(int rows, int cols)
-{
-    return Mat::zeros(rows, cols, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::zeros(Size sz)
-{
-    return Mat::zeros(sz, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::ones(int rows, int cols)
-{
-    return Mat::ones(rows, cols, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::ones(Size sz)
-{
-    return Mat::ones(sz, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::eye(int rows, int cols)
-{
-    return Mat::eye(rows, cols, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline MatExpr Mat_<_Tp>::eye(Size sz)
-{
-    return Mat::eye(sz, DataType<_Tp>::type);
-}
-
-//////////////////////////////// Iterators & Comma initializers //////////////////////////////////
-
-inline MatConstIterator::MatConstIterator()
-    : m(0), elemSize(0), ptr(0), sliceStart(0), sliceEnd(0) {}
-
-inline MatConstIterator::MatConstIterator(const Mat* _m)
-    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
-{
-    if( m && m->isContinuous() )
-    {
-        sliceStart = m->data;
-        sliceEnd = sliceStart + m->total()*elemSize;
-    }
-    seek((const int*)0);
-}
-
-inline MatConstIterator::MatConstIterator(const Mat* _m, int _row, int _col)
-    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
-{
-    CV_Assert(m && m->dims <= 2);
-    if( m->isContinuous() )
-    {
-        sliceStart = m->data;
-        sliceEnd = sliceStart + m->total()*elemSize;
-    }
-    int idx[]={_row, _col};
-    seek(idx);
-}
-
-inline MatConstIterator::MatConstIterator(const Mat* _m, Point _pt)
-    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
-{
-    CV_Assert(m && m->dims <= 2);
-    if( m->isContinuous() )
-    {
-        sliceStart = m->data;
-        sliceEnd = sliceStart + m->total()*elemSize;
-    }
-    int idx[]={_pt.y, _pt.x};
-    seek(idx);
-}
-
-inline MatConstIterator::MatConstIterator(const MatConstIterator& it)
-    : m(it.m), elemSize(it.elemSize), ptr(it.ptr), sliceStart(it.sliceStart), sliceEnd(it.sliceEnd)
-{}
-
-inline MatConstIterator& MatConstIterator::operator = (const MatConstIterator& it )
-{
-    m = it.m; elemSize = it.elemSize; ptr = it.ptr;
-    sliceStart = it.sliceStart; sliceEnd = it.sliceEnd;
-    return *this;
-}
-
-inline uchar* MatConstIterator::operator *() const { return ptr; }
-
-inline MatConstIterator& MatConstIterator::operator += (ptrdiff_t ofs)
-{
-    if( !m || ofs == 0 )
-        return *this;
-    ptrdiff_t ofsb = ofs*elemSize;
-    ptr += ofsb;
-    if( ptr < sliceStart || sliceEnd <= ptr )
-    {
-        ptr -= ofsb;
-        seek(ofs, true);
-    }
-    return *this;
-}
-
-inline MatConstIterator& MatConstIterator::operator -= (ptrdiff_t ofs)
-{ return (*this += -ofs); }
-
-inline MatConstIterator& MatConstIterator::operator --()
-{
-    if( m && (ptr -= elemSize) < sliceStart )
-    {
-        ptr += elemSize;
-        seek(-1, true);
-    }
-    return *this;
-}
-
-inline MatConstIterator MatConstIterator::operator --(int)
-{
-    MatConstIterator b = *this;
-    *this += -1;
-    return b;
-}
-
-inline MatConstIterator& MatConstIterator::operator ++()
-{
-    if( m && (ptr += elemSize) >= sliceEnd )
-    {
-        ptr -= elemSize;
-        seek(1, true);
-    }
-    return *this;
-}
-
-inline MatConstIterator MatConstIterator::operator ++(int)
-{
-    MatConstIterator b = *this;
-    *this += 1;
-    return b;
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>::MatConstIterator_() {}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>::MatConstIterator_(const Mat_<_Tp>* _m)
-    : MatConstIterator(_m) {}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>::
-    MatConstIterator_(const Mat_<_Tp>* _m, int _row, int _col)
-    : MatConstIterator(_m, _row, _col) {}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>::
-    MatConstIterator_(const Mat_<_Tp>* _m, Point _pt)
-    : MatConstIterator(_m, _pt) {}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>::
-    MatConstIterator_(const MatConstIterator_& it)
-    : MatConstIterator(it) {}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>&
-    MatConstIterator_<_Tp>::operator = (const MatConstIterator_& it )
-{
-    MatConstIterator::operator = (it);
-    return *this;
-}
-
-template<typename _Tp> inline _Tp MatConstIterator_<_Tp>::operator *() const { return *(_Tp*)(this->ptr); }
-
-template<typename _Tp> inline MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator += (ptrdiff_t ofs)
-{
-    MatConstIterator::operator += (ofs);
-    return *this;
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator -= (ptrdiff_t ofs)
-{ return (*this += -ofs); }
-
-template<typename _Tp> inline MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator --()
-{
-    MatConstIterator::operator --();
-    return *this;
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp> MatConstIterator_<_Tp>::operator --(int)
-{
-    MatConstIterator_ b = *this;
-    MatConstIterator::operator --();
-    return b;
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator ++()
-{
-    MatConstIterator::operator ++();
-    return *this;
-}
-
-template<typename _Tp> inline MatConstIterator_<_Tp> MatConstIterator_<_Tp>::operator ++(int)
-{
-    MatConstIterator_ b = *this;
-    MatConstIterator::operator ++();
-    return b;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_() : MatConstIterator_<_Tp>() {}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_(Mat_<_Tp>* _m)
-    : MatConstIterator_<_Tp>(_m) {}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_(Mat_<_Tp>* _m, int _row, int _col)
-    : MatConstIterator_<_Tp>(_m, _row, _col) {}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_(const Mat_<_Tp>* _m, Point _pt)
-    : MatConstIterator_<_Tp>(_m, _pt) {}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_(const Mat_<_Tp>* _m, const int* _idx)
-    : MatConstIterator_<_Tp>(_m, _idx) {}
-
-template<typename _Tp> inline MatIterator_<_Tp>::MatIterator_(const MatIterator_& it)
-    : MatConstIterator_<_Tp>(it) {}
-
-template<typename _Tp> inline MatIterator_<_Tp>& MatIterator_<_Tp>::operator = (const MatIterator_<_Tp>& it )
-{
-    MatConstIterator::operator = (it);
-    return *this;
-}
-
-template<typename _Tp> inline _Tp& MatIterator_<_Tp>::operator *() const { return *(_Tp*)(this->ptr); }
-
-template<typename _Tp> inline MatIterator_<_Tp>& MatIterator_<_Tp>::operator += (ptrdiff_t ofs)
-{
-    MatConstIterator::operator += (ofs);
-    return *this;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp>& MatIterator_<_Tp>::operator -= (ptrdiff_t ofs)
-{
-    MatConstIterator::operator += (-ofs);
-    return *this;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp>& MatIterator_<_Tp>::operator --()
-{
-    MatConstIterator::operator --();
-    return *this;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp> MatIterator_<_Tp>::operator --(int)
-{
-    MatIterator_ b = *this;
-    MatConstIterator::operator --();
-    return b;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp>& MatIterator_<_Tp>::operator ++()
-{
-    MatConstIterator::operator ++();
-    return *this;
-}
-
-template<typename _Tp> inline MatIterator_<_Tp> MatIterator_<_Tp>::operator ++(int)
-{
-    MatIterator_ b = *this;
-    MatConstIterator::operator ++();
-    return b;
-}
-
-template<typename _Tp> inline Point MatConstIterator_<_Tp>::pos() const
-{
-    if( !m )
-        return Point();
-    CV_DbgAssert( m->dims <= 2 );
-    if( m->isContinuous() )
-    {
-        ptrdiff_t ofs = (const _Tp*)ptr - (const _Tp*)m->data;
-        int y = (int)(ofs / m->cols), x = (int)(ofs - (ptrdiff_t)y*m->cols);
-        return Point(x, y);
-    }
-    else
-    {
-        ptrdiff_t ofs = (uchar*)ptr - m->data;
-        int y = (int)(ofs / m->step), x = (int)((ofs - y*m->step)/sizeof(_Tp));
-        return Point(x, y);
-    }
-}
-
-static inline bool
-operator == (const MatConstIterator& a, const MatConstIterator& b)
-{ return a.m == b.m && a.ptr == b.ptr; }
-
-template<typename _Tp> static inline bool
-operator != (const MatConstIterator& a, const MatConstIterator& b)
-{ return !(a == b); }
-
-template<typename _Tp> static inline bool
-operator == (const MatConstIterator_<_Tp>& a, const MatConstIterator_<_Tp>& b)
-{ return a.m == b.m && a.ptr == b.ptr; }
-
-template<typename _Tp> static inline bool
-operator != (const MatConstIterator_<_Tp>& a, const MatConstIterator_<_Tp>& b)
-{ return a.m != b.m || a.ptr != b.ptr; }
-
-template<typename _Tp> static inline bool
-operator == (const MatIterator_<_Tp>& a, const MatIterator_<_Tp>& b)
-{ return a.m == b.m && a.ptr == b.ptr; }
-
-template<typename _Tp> static inline bool
-operator != (const MatIterator_<_Tp>& a, const MatIterator_<_Tp>& b)
-{ return a.m != b.m || a.ptr != b.ptr; }
-
-static inline bool
-operator < (const MatConstIterator& a, const MatConstIterator& b)
-{ return a.ptr < b.ptr; }
-
-static inline bool
-operator > (const MatConstIterator& a, const MatConstIterator& b)
-{ return a.ptr > b.ptr; }
-
-static inline bool
-operator <= (const MatConstIterator& a, const MatConstIterator& b)
-{ return a.ptr <= b.ptr; }
-
-static inline bool
-operator >= (const MatConstIterator& a, const MatConstIterator& b)
-{ return a.ptr >= b.ptr; }
-
-CV_EXPORTS ptrdiff_t operator - (const MatConstIterator& b, const MatConstIterator& a);
-
-static inline MatConstIterator operator + (const MatConstIterator& a, ptrdiff_t ofs)
-{ MatConstIterator b = a; return b += ofs; }
-
-static inline MatConstIterator operator + (ptrdiff_t ofs, const MatConstIterator& a)
-{ MatConstIterator b = a; return b += ofs; }
-
-static inline MatConstIterator operator - (const MatConstIterator& a, ptrdiff_t ofs)
-{ MatConstIterator b = a; return b += -ofs; }
-
-template<typename _Tp> static inline MatConstIterator_<_Tp>
-operator + (const MatConstIterator_<_Tp>& a, ptrdiff_t ofs)
-{ MatConstIterator t = (const MatConstIterator&)a + ofs; return (MatConstIterator_<_Tp>&)t; }
-
-template<typename _Tp> static inline MatConstIterator_<_Tp>
-operator + (ptrdiff_t ofs, const MatConstIterator_<_Tp>& a)
-{ MatConstIterator t = (const MatConstIterator&)a + ofs; return (MatConstIterator_<_Tp>&)t; }
-
-template<typename _Tp> static inline MatConstIterator_<_Tp>
-operator - (const MatConstIterator_<_Tp>& a, ptrdiff_t ofs)
-{ MatConstIterator t = (const MatConstIterator&)a - ofs; return (MatConstIterator_<_Tp>&)t; }
-
-inline uchar* MatConstIterator::operator [](ptrdiff_t i) const
-{ return *(*this + i); }
-
-template<typename _Tp> inline _Tp MatConstIterator_<_Tp>::operator [](ptrdiff_t i) const
-{ return *(_Tp*)MatConstIterator::operator [](i); }
-
-template<typename _Tp> static inline MatIterator_<_Tp>
-operator + (const MatIterator_<_Tp>& a, ptrdiff_t ofs)
-{ MatConstIterator t = (const MatConstIterator&)a + ofs; return (MatIterator_<_Tp>&)t; }
-
-template<typename _Tp> static inline MatIterator_<_Tp>
-operator + (ptrdiff_t ofs, const MatIterator_<_Tp>& a)
-{ MatConstIterator t = (const MatConstIterator&)a + ofs; return (MatIterator_<_Tp>&)t; }
-
-template<typename _Tp> static inline MatIterator_<_Tp>
-operator - (const MatIterator_<_Tp>& a, ptrdiff_t ofs)
-{ MatConstIterator t = (const MatConstIterator&)a - ofs; return (MatIterator_<_Tp>&)t; }
-
-template<typename _Tp> inline _Tp& MatIterator_<_Tp>::operator [](ptrdiff_t i) const
-{ return *(*this + i); }
-
-template<typename _Tp> inline MatConstIterator_<_Tp> Mat_<_Tp>::begin() const
-{ return Mat::begin<_Tp>(); }
-
-template<typename _Tp> inline MatConstIterator_<_Tp> Mat_<_Tp>::end() const
-{ return Mat::end<_Tp>(); }
-
-template<typename _Tp> inline MatIterator_<_Tp> Mat_<_Tp>::begin()
-{ return Mat::begin<_Tp>(); }
-
-template<typename _Tp> inline MatIterator_<_Tp> Mat_<_Tp>::end()
-{ return Mat::end<_Tp>(); }
-
-template<typename _Tp> inline MatCommaInitializer_<_Tp>::MatCommaInitializer_(Mat_<_Tp>* _m) : it(_m) {}
-
-template<typename _Tp> template<typename T2> inline MatCommaInitializer_<_Tp>&
-MatCommaInitializer_<_Tp>::operator , (T2 v)
-{
-    CV_DbgAssert( this->it < ((const Mat_<_Tp>*)this->it.m)->end() );
-    *this->it = _Tp(v); ++this->it;
-    return *this;
-}
-
-template<typename _Tp> inline Mat_<_Tp> MatCommaInitializer_<_Tp>::operator *() const
-{
-    CV_DbgAssert( this->it == ((const Mat_<_Tp>*)this->it.m)->end() );
-    return Mat_<_Tp>(*this->it.m);
-}
-
-template<typename _Tp> inline MatCommaInitializer_<_Tp>::operator Mat_<_Tp>() const
-{
-    CV_DbgAssert( this->it == ((const Mat_<_Tp>*)this->it.m)->end() );
-    return Mat_<_Tp>(*this->it.m);
-}
-
-template<typename _Tp, typename T2> static inline MatCommaInitializer_<_Tp>
-operator << (const Mat_<_Tp>& m, T2 val)
-{
-    MatCommaInitializer_<_Tp> commaInitializer((Mat_<_Tp>*)&m);
-    return (commaInitializer, val);
-}
-
-//////////////////////////////// SparseMat ////////////////////////////////
-
-inline SparseMat::SparseMat()
-: flags(MAGIC_VAL), hdr(0)
-{
-}
-
-inline SparseMat::SparseMat(int _dims, const int* _sizes, int _type)
-: flags(MAGIC_VAL), hdr(0)
-{
-    create(_dims, _sizes, _type);
-}
-
-inline SparseMat::SparseMat(const SparseMat& m)
-: flags(m.flags), hdr(m.hdr)
-{
-    addref();
-}
-
-inline SparseMat::~SparseMat()
-{
-    release();
-}
-
-inline SparseMat& SparseMat::operator = (const SparseMat& m)
-{
-    if( this != &m )
-    {
-        if( m.hdr )
-            CV_XADD(&m.hdr->refcount, 1);
-        release();
-        flags = m.flags;
-        hdr = m.hdr;
-    }
-    return *this;
-}
-
-inline SparseMat& SparseMat::operator = (const Mat& m)
-{ return (*this = SparseMat(m)); }
-
-inline SparseMat SparseMat::clone() const
-{
-    SparseMat temp;
-    this->copyTo(temp);
-    return temp;
-}
-
-
-inline void SparseMat::assignTo( SparseMat& m, int _type ) const
-{
-    if( _type < 0 )
-        m = *this;
-    else
-        convertTo(m, _type);
-}
-
-inline void SparseMat::addref()
-{ if( hdr ) CV_XADD(&hdr->refcount, 1); }
-
-inline void SparseMat::release()
-{
-    if( hdr && CV_XADD(&hdr->refcount, -1) == 1 )
-        delete hdr;
-    hdr = 0;
-}
-
-inline size_t SparseMat::elemSize() const
-{ return CV_ELEM_SIZE(flags); }
-
-inline size_t SparseMat::elemSize1() const
-{ return CV_ELEM_SIZE1(flags); }
-
-inline int SparseMat::type() const
-{ return CV_MAT_TYPE(flags); }
-
-inline int SparseMat::depth() const
-{ return CV_MAT_DEPTH(flags); }
-
-inline int SparseMat::channels() const
-{ return CV_MAT_CN(flags); }
-
-inline const int* SparseMat::size() const
-{
-    return hdr ? hdr->size : 0;
-}
-
-inline int SparseMat::size(int i) const
-{
-    if( hdr )
-    {
-        CV_DbgAssert((unsigned)i < (unsigned)hdr->dims);
-        return hdr->size[i];
-    }
-    return 0;
-}
-
-inline int SparseMat::dims() const
-{
-    return hdr ? hdr->dims : 0;
-}
-
-inline size_t SparseMat::nzcount() const
-{
-    return hdr ? hdr->nodeCount : 0;
-}
-
-inline size_t SparseMat::hash(int i0) const
-{
-    return (size_t)i0;
-}
-
-inline size_t SparseMat::hash(int i0, int i1) const
-{
-    return (size_t)(unsigned)i0*HASH_SCALE + (unsigned)i1;
-}
-
-inline size_t SparseMat::hash(int i0, int i1, int i2) const
-{
-    return ((size_t)(unsigned)i0*HASH_SCALE + (unsigned)i1)*HASH_SCALE + (unsigned)i2;
-}
-
-inline size_t SparseMat::hash(const int* idx) const
-{
-    size_t h = (unsigned)idx[0];
-    if( !hdr )
-        return 0;
-    int i, d = hdr->dims;
-    for( i = 1; i < d; i++ )
-        h = h*HASH_SCALE + (unsigned)idx[i];
-    return h;
-}
-
-template<typename _Tp> inline _Tp& SparseMat::ref(int i0, size_t* hashval)
-{ return *(_Tp*)((SparseMat*)this)->ptr(i0, true, hashval); }
-
-template<typename _Tp> inline _Tp& SparseMat::ref(int i0, int i1, size_t* hashval)
-{ return *(_Tp*)((SparseMat*)this)->ptr(i0, i1, true, hashval); }
-
-template<typename _Tp> inline _Tp& SparseMat::ref(int i0, int i1, int i2, size_t* hashval)
-{ return *(_Tp*)((SparseMat*)this)->ptr(i0, i1, i2, true, hashval); }
-
-template<typename _Tp> inline _Tp& SparseMat::ref(const int* idx, size_t* hashval)
-{ return *(_Tp*)((SparseMat*)this)->ptr(idx, true, hashval); }
-
-template<typename _Tp> inline _Tp SparseMat::value(int i0, size_t* hashval) const
-{
-    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, false, hashval);
-    return p ? *p : _Tp();
-}
-
-template<typename _Tp> inline _Tp SparseMat::value(int i0, int i1, size_t* hashval) const
-{
-    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, i1, false, hashval);
-    return p ? *p : _Tp();
-}
-
-template<typename _Tp> inline _Tp SparseMat::value(int i0, int i1, int i2, size_t* hashval) const
-{
-    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, i1, i2, false, hashval);
-    return p ? *p : _Tp();
-}
-
-template<typename _Tp> inline _Tp SparseMat::value(const int* idx, size_t* hashval) const
-{
-    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(idx, false, hashval);
-    return p ? *p : _Tp();
-}
-
-template<typename _Tp> inline const _Tp* SparseMat::find(int i0, size_t* hashval) const
-{ return (const _Tp*)((SparseMat*)this)->ptr(i0, false, hashval); }
-
-template<typename _Tp> inline const _Tp* SparseMat::find(int i0, int i1, size_t* hashval) const
-{ return (const _Tp*)((SparseMat*)this)->ptr(i0, i1, false, hashval); }
-
-template<typename _Tp> inline const _Tp* SparseMat::find(int i0, int i1, int i2, size_t* hashval) const
-{ return (const _Tp*)((SparseMat*)this)->ptr(i0, i1, i2, false, hashval); }
-
-template<typename _Tp> inline const _Tp* SparseMat::find(const int* idx, size_t* hashval) const
-{ return (const _Tp*)((SparseMat*)this)->ptr(idx, false, hashval); }
-
-template<typename _Tp> inline _Tp& SparseMat::value(Node* n)
-{ return *(_Tp*)((uchar*)n + hdr->valueOffset); }
-
-template<typename _Tp> inline const _Tp& SparseMat::value(const Node* n) const
-{ return *(const _Tp*)((const uchar*)n + hdr->valueOffset); }
-
-inline SparseMat::Node* SparseMat::node(size_t nidx)
-{ return (Node*)&hdr->pool[nidx]; }
-
-inline const SparseMat::Node* SparseMat::node(size_t nidx) const
-{ return (const Node*)&hdr->pool[nidx]; }
-
-inline SparseMatIterator SparseMat::begin()
-{ return SparseMatIterator(this); }
-
-inline SparseMatConstIterator SparseMat::begin() const
-{ return SparseMatConstIterator(this); }
-
-inline SparseMatIterator SparseMat::end()
-{ SparseMatIterator it(this); it.seekEnd(); return it; }
-
-inline SparseMatConstIterator SparseMat::end() const
-{ SparseMatConstIterator it(this); it.seekEnd(); return it; }
-
-template<typename _Tp> inline SparseMatIterator_<_Tp> SparseMat::begin()
-{ return SparseMatIterator_<_Tp>(this); }
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp> SparseMat::begin() const
-{ return SparseMatConstIterator_<_Tp>(this); }
-
-template<typename _Tp> inline SparseMatIterator_<_Tp> SparseMat::end()
-{ SparseMatIterator_<_Tp> it(this); it.seekEnd(); return it; }
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp> SparseMat::end() const
-{ SparseMatConstIterator_<_Tp> it(this); it.seekEnd(); return it; }
-
-
-inline SparseMatConstIterator::SparseMatConstIterator()
-: m(0), hashidx(0), ptr(0)
-{
-}
-
-inline SparseMatConstIterator::SparseMatConstIterator(const SparseMatConstIterator& it)
-: m(it.m), hashidx(it.hashidx), ptr(it.ptr)
-{
-}
-
-static inline bool operator == (const SparseMatConstIterator& it1, const SparseMatConstIterator& it2)
-{ return it1.m == it2.m && it1.ptr == it2.ptr; }
-
-static inline bool operator != (const SparseMatConstIterator& it1, const SparseMatConstIterator& it2)
-{ return !(it1 == it2); }
-
-
-inline SparseMatConstIterator& SparseMatConstIterator::operator = (const SparseMatConstIterator& it)
-{
-    if( this != &it )
-    {
-        m = it.m;
-        hashidx = it.hashidx;
-        ptr = it.ptr;
-    }
-    return *this;
-}
-
-template<typename _Tp> inline const _Tp& SparseMatConstIterator::value() const
-{ return *(_Tp*)ptr; }
-
-inline const SparseMat::Node* SparseMatConstIterator::node() const
-{
-    return ptr && m && m->hdr ?
-        (const SparseMat::Node*)(ptr - m->hdr->valueOffset) : 0;
-}
-
-inline SparseMatConstIterator SparseMatConstIterator::operator ++(int)
-{
-    SparseMatConstIterator it = *this;
-    ++*this;
-    return it;
-}
-
-
-inline void SparseMatConstIterator::seekEnd()
-{
-    if( m && m->hdr )
-    {
-        hashidx = m->hdr->hashtab.size();
-        ptr = 0;
-    }
-}
-
-inline SparseMatIterator::SparseMatIterator()
-{}
-
-inline SparseMatIterator::SparseMatIterator(SparseMat* _m)
-: SparseMatConstIterator(_m)
-{}
-
-inline SparseMatIterator::SparseMatIterator(const SparseMatIterator& it)
-: SparseMatConstIterator(it)
-{
-}
-
-inline SparseMatIterator& SparseMatIterator::operator = (const SparseMatIterator& it)
-{
-    (SparseMatConstIterator&)*this = it;
-    return *this;
-}
-
-template<typename _Tp> inline _Tp& SparseMatIterator::value() const
-{ return *(_Tp*)ptr; }
-
-inline SparseMat::Node* SparseMatIterator::node() const
-{
-    return (SparseMat::Node*)SparseMatConstIterator::node();
-}
-
-inline SparseMatIterator& SparseMatIterator::operator ++()
-{
-    SparseMatConstIterator::operator ++();
-    return *this;
-}
-
-inline SparseMatIterator SparseMatIterator::operator ++(int)
-{
-    SparseMatIterator it = *this;
-    ++*this;
-    return it;
-}
-
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_()
-{ flags = MAGIC_VAL | DataType<_Tp>::type; }
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_(int _dims, const int* _sizes)
-: SparseMat(_dims, _sizes, DataType<_Tp>::type)
-{}
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_(const SparseMat& m)
-{
-    if( m.type() == DataType<_Tp>::type )
-        *this = (const SparseMat_<_Tp>&)m;
-    else
-        m.convertTo(this, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_(const SparseMat_<_Tp>& m)
-{
-    this->flags = m.flags;
-    this->hdr = m.hdr;
-    if( this->hdr )
-        CV_XADD(&this->hdr->refcount, 1);
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_(const Mat& m)
-{
-    SparseMat sm(m);
-    *this = sm;
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>::SparseMat_(const CvSparseMat* m)
-{
-    SparseMat sm(m);
-    *this = sm;
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>&
-SparseMat_<_Tp>::operator = (const SparseMat_<_Tp>& m)
-{
-    if( this != &m )
-    {
-        if( m.hdr ) CV_XADD(&m.hdr->refcount, 1);
-        release();
-        flags = m.flags;
-        hdr = m.hdr;
-    }
-    return *this;
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>&
-SparseMat_<_Tp>::operator = (const SparseMat& m)
-{
-    if( m.type() == DataType<_Tp>::type )
-        return (*this = (const SparseMat_<_Tp>&)m);
-    m.convertTo(*this, DataType<_Tp>::type);
-    return *this;
-}
-
-template<typename _Tp> inline SparseMat_<_Tp>&
-SparseMat_<_Tp>::operator = (const Mat& m)
-{ return (*this = SparseMat(m)); }
-
-template<typename _Tp> inline SparseMat_<_Tp>
-SparseMat_<_Tp>::clone() const
-{
-    SparseMat_<_Tp> m;
-    this->copyTo(m);
-    return m;
-}
-
-template<typename _Tp> inline void
-SparseMat_<_Tp>::create(int _dims, const int* _sizes)
-{
-    SparseMat::create(_dims, _sizes, DataType<_Tp>::type);
-}
-
-template<typename _Tp> inline
-SparseMat_<_Tp>::operator CvSparseMat*() const
-{
-    return SparseMat::operator CvSparseMat*();
-}
-
-template<typename _Tp> inline int SparseMat_<_Tp>::type() const
-{ return DataType<_Tp>::type; }
-
-template<typename _Tp> inline int SparseMat_<_Tp>::depth() const
-{ return DataType<_Tp>::depth; }
-
-template<typename _Tp> inline int SparseMat_<_Tp>::channels() const
-{ return DataType<_Tp>::channels; }
-
-template<typename _Tp> inline _Tp&
-SparseMat_<_Tp>::ref(int i0, size_t* hashval)
-{ return SparseMat::ref<_Tp>(i0, hashval); }
-
-template<typename _Tp> inline _Tp
-SparseMat_<_Tp>::operator()(int i0, size_t* hashval) const
-{ return SparseMat::value<_Tp>(i0, hashval); }
-
-template<typename _Tp> inline _Tp&
-SparseMat_<_Tp>::ref(int i0, int i1, size_t* hashval)
-{ return SparseMat::ref<_Tp>(i0, i1, hashval); }
-
-template<typename _Tp> inline _Tp
-SparseMat_<_Tp>::operator()(int i0, int i1, size_t* hashval) const
-{ return SparseMat::value<_Tp>(i0, i1, hashval); }
-
-template<typename _Tp> inline _Tp&
-SparseMat_<_Tp>::ref(int i0, int i1, int i2, size_t* hashval)
-{ return SparseMat::ref<_Tp>(i0, i1, i2, hashval); }
-
-template<typename _Tp> inline _Tp
-SparseMat_<_Tp>::operator()(int i0, int i1, int i2, size_t* hashval) const
-{ return SparseMat::value<_Tp>(i0, i1, i2, hashval); }
-
-template<typename _Tp> inline _Tp&
-SparseMat_<_Tp>::ref(const int* idx, size_t* hashval)
-{ return SparseMat::ref<_Tp>(idx, hashval); }
-
-template<typename _Tp> inline _Tp
-SparseMat_<_Tp>::operator()(const int* idx, size_t* hashval) const
-{ return SparseMat::value<_Tp>(idx, hashval); }
-
-template<typename _Tp> inline SparseMatIterator_<_Tp> SparseMat_<_Tp>::begin()
-{ return SparseMatIterator_<_Tp>(this); }
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp> SparseMat_<_Tp>::begin() const
-{ return SparseMatConstIterator_<_Tp>(this); }
-
-template<typename _Tp> inline SparseMatIterator_<_Tp> SparseMat_<_Tp>::end()
-{ SparseMatIterator_<_Tp> it(this); it.seekEnd(); return it; }
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp> SparseMat_<_Tp>::end() const
-{ SparseMatConstIterator_<_Tp> it(this); it.seekEnd(); return it; }
-
-template<typename _Tp> inline
-SparseMatConstIterator_<_Tp>::SparseMatConstIterator_()
-{}
-
-template<typename _Tp> inline
-SparseMatConstIterator_<_Tp>::SparseMatConstIterator_(const SparseMat_<_Tp>* _m)
-: SparseMatConstIterator(_m)
-{}
-
-template<typename _Tp> inline
-SparseMatConstIterator_<_Tp>::SparseMatConstIterator_(const SparseMatConstIterator_<_Tp>& it)
-: SparseMatConstIterator(it)
-{}
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp>&
-SparseMatConstIterator_<_Tp>::operator = (const SparseMatConstIterator_<_Tp>& it)
-{ return reinterpret_cast<SparseMatConstIterator_<_Tp>&>
-    (*reinterpret_cast<SparseMatConstIterator*>(this) =
-     reinterpret_cast<const SparseMatConstIterator&>(it)); }
-
-template<typename _Tp> inline const _Tp&
-SparseMatConstIterator_<_Tp>::operator *() const
-{ return *(const _Tp*)this->ptr; }
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp>&
-SparseMatConstIterator_<_Tp>::operator ++()
-{
-    SparseMatConstIterator::operator ++();
-    return *this;
-}
-
-template<typename _Tp> inline SparseMatConstIterator_<_Tp>
-SparseMatConstIterator_<_Tp>::operator ++(int)
-{
-    SparseMatConstIterator it = *this;
-    SparseMatConstIterator::operator ++();
-    return it;
-}
-
-template<typename _Tp> inline
-SparseMatIterator_<_Tp>::SparseMatIterator_()
-{}
-
-template<typename _Tp> inline
-SparseMatIterator_<_Tp>::SparseMatIterator_(SparseMat_<_Tp>* _m)
-: SparseMatConstIterator_<_Tp>(_m)
-{}
-
-template<typename _Tp> inline
-SparseMatIterator_<_Tp>::SparseMatIterator_(const SparseMatIterator_<_Tp>& it)
-: SparseMatConstIterator_<_Tp>(it)
-{}
-
-template<typename _Tp> inline SparseMatIterator_<_Tp>&
-SparseMatIterator_<_Tp>::operator = (const SparseMatIterator_<_Tp>& it)
-{ return reinterpret_cast<SparseMatIterator_<_Tp>&>
-    (*reinterpret_cast<SparseMatConstIterator*>(this) =
-     reinterpret_cast<const SparseMatConstIterator&>(it)); }
-
-template<typename _Tp> inline _Tp&
-SparseMatIterator_<_Tp>::operator *() const
-{ return *(_Tp*)this->ptr; }
-
-template<typename _Tp> inline SparseMatIterator_<_Tp>&
-SparseMatIterator_<_Tp>::operator ++()
-{
-    SparseMatConstIterator::operator ++();
-    return *this;
-}
-
-template<typename _Tp> inline SparseMatIterator_<_Tp>
-SparseMatIterator_<_Tp>::operator ++(int)
-{
-    SparseMatIterator it = *this;
-    SparseMatConstIterator::operator ++();
-    return it;
-}
-
-}
-
-#endif
-#endif
+#endif // __OPENCV_CORE_MAT_HPP__
diff --git a/modules/core/include/opencv2/core/mat.inl.hpp b/modules/core/include/opencv2/core/mat.inl.hpp
new file mode 100644
index 0000000000..026ab695d5
--- /dev/null
+++ b/modules/core/include/opencv2/core/mat.inl.hpp
@@ -0,0 +1,2961 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_MATRIX_OPERATIONS_HPP__
+#define __OPENCV_CORE_MATRIX_OPERATIONS_HPP__
+
+#ifndef __cplusplus
+#  error mat.inl.hpp header must be compiled as C++
+#endif
+
+namespace cv
+{
+
+//////////////////////// Input/Output Arrays ////////////////////////
+
+template<typename _Tp> inline
+_InputArray::_InputArray(const std::vector<_Tp>& vec)
+    : flags(FIXED_TYPE + STD_VECTOR + DataType<_Tp>::type), obj((void*)&vec)
+{}
+
+template<typename _Tp> inline
+_InputArray::_InputArray(const std::vector<std::vector<_Tp> >& vec)
+    : flags(FIXED_TYPE + STD_VECTOR_VECTOR + DataType<_Tp>::type), obj((void*)&vec)
+{}
+
+template<typename _Tp> inline
+_InputArray::_InputArray(const std::vector<Mat_<_Tp> >& vec)
+    : flags(FIXED_TYPE + STD_VECTOR_MAT + DataType<_Tp>::type), obj((void*)&vec)
+{}
+
+template<typename _Tp, int m, int n> inline
+_InputArray::_InputArray(const Matx<_Tp, m, n>& mtx)
+    : flags(FIXED_TYPE + FIXED_SIZE + MATX + DataType<_Tp>::type), obj((void*)&mtx), sz(n, m)
+{}
+
+template<typename _Tp> inline
+_InputArray::_InputArray(const _Tp* vec, int n)
+    : flags(FIXED_TYPE + FIXED_SIZE + MATX + DataType<_Tp>::type), obj((void*)vec), sz(n, 1)
+{}
+
+template<typename _Tp> inline
+_InputArray::_InputArray(const Mat_<_Tp>& m)
+    : flags(FIXED_TYPE + MAT + DataType<_Tp>::type), obj((void*)&m)
+{}
+
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(std::vector<_Tp>& vec)
+    : _InputArray(vec)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(std::vector<std::vector<_Tp> >& vec)
+    : _InputArray(vec)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(std::vector<Mat_<_Tp> >& vec)
+    : _InputArray(vec)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(Mat_<_Tp>& m)
+    : _InputArray(m)
+{}
+
+template<typename _Tp, int m, int n> inline
+_OutputArray::_OutputArray(Matx<_Tp, m, n>& mtx)
+    : _InputArray(mtx)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(_Tp* vec, int n)
+    : _InputArray(vec, n)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(const std::vector<_Tp>& vec)
+    : _InputArray(vec)
+{
+    flags |= FIXED_SIZE;
+}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(const std::vector<std::vector<_Tp> >& vec)
+    : _InputArray(vec)
+{
+    flags |= FIXED_SIZE;
+}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(const std::vector<Mat_<_Tp> >& vec)
+    : _InputArray(vec)
+{
+    flags |= FIXED_SIZE;
+}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(const Mat_<_Tp>& m)
+    : _InputArray(m)
+{
+    flags |= FIXED_SIZE;
+}
+
+template<typename _Tp, int m, int n> inline
+_OutputArray::_OutputArray(const Matx<_Tp, m, n>& mtx)
+    : _InputArray(mtx)
+{}
+
+template<typename _Tp> inline
+_OutputArray::_OutputArray(const _Tp* vec, int n)
+    : _InputArray(vec, n)
+{}
+
+
+
+//////////////////////////////// Mat ////////////////////////////////
+
+inline
+Mat::Mat()
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{}
+
+inline
+Mat::Mat(int _rows, int _cols, int _type)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create(_rows, _cols, _type);
+}
+
+inline
+Mat::Mat(int _rows, int _cols, int _type, const Scalar& _s)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create(_rows, _cols, _type);
+    *this = _s;
+}
+
+inline
+Mat::Mat(Size _sz, int _type)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create( _sz.height, _sz.width, _type );
+}
+
+inline
+Mat::Mat(Size _sz, int _type, const Scalar& _s)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create(_sz.height, _sz.width, _type);
+    *this = _s;
+}
+
+inline
+Mat::Mat(int _dims, const int* _sz, int _type)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create(_dims, _sz, _type);
+}
+
+inline
+Mat::Mat(int _dims, const int* _sz, int _type, const Scalar& _s)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
+{
+    create(_dims, _sz, _type);
+    *this = _s;
+}
+
+inline
+Mat::Mat(const Mat& m)
+    : flags(m.flags), dims(m.dims), rows(m.rows), cols(m.cols), data(m.data), refcount(m.refcount),
+      datastart(m.datastart), dataend(m.dataend), datalimit(m.datalimit), allocator(m.allocator),
+      size(&rows)
+{
+    if( refcount )
+        CV_XADD(refcount, 1);
+    if( m.dims <= 2 )
+    {
+        step[0] = m.step[0]; step[1] = m.step[1];
+    }
+    else
+    {
+        dims = 0;
+        copySize(m);
+    }
+}
+
+inline
+Mat::Mat(int _rows, int _cols, int _type, void* _data, size_t _step)
+    : flags(MAGIC_VAL + (_type & TYPE_MASK)), dims(2), rows(_rows), cols(_cols),
+      data((uchar*)_data), refcount(0), datastart((uchar*)_data), dataend(0), datalimit(0),
+      allocator(0), size(&rows)
+{
+    size_t esz = CV_ELEM_SIZE(_type);
+    size_t minstep = cols * esz;
+    if( _step == AUTO_STEP )
+    {
+        _step = minstep;
+        flags |= CONTINUOUS_FLAG;
+    }
+    else
+    {
+        if( rows == 1 ) _step = minstep;
+        CV_DbgAssert( _step >= minstep );
+        flags |= _step == minstep ? CONTINUOUS_FLAG : 0;
+    }
+    step[0] = _step;
+    step[1] = esz;
+    datalimit = datastart + _step * rows;
+    dataend = datalimit - _step + minstep;
+}
+
+inline
+Mat::Mat(Size _sz, int _type, void* _data, size_t _step)
+    : flags(MAGIC_VAL + (_type & TYPE_MASK)), dims(2), rows(_sz.height), cols(_sz.width),
+      data((uchar*)_data), refcount(0), datastart((uchar*)_data), dataend(0), datalimit(0),
+      allocator(0), size(&rows)
+{
+    size_t esz = CV_ELEM_SIZE(_type);
+    size_t minstep = cols*esz;
+    if( _step == AUTO_STEP )
+    {
+        _step = minstep;
+        flags |= CONTINUOUS_FLAG;
+    }
+    else
+    {
+        if( rows == 1 ) _step = minstep;
+        CV_DbgAssert( _step >= minstep );
+        flags |= _step == minstep ? CONTINUOUS_FLAG : 0;
+    }
+    step[0] = _step;
+    step[1] = esz;
+    datalimit = datastart + _step*rows;
+    dataend = datalimit - _step + minstep;
+}
+
+template<typename _Tp> inline
+Mat::Mat(const std::vector<_Tp>& vec, bool copyData)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(2), rows((int)vec.size()),
+      cols(1), data(0), refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    if(vec.empty())
+        return;
+    if( !copyData )
+    {
+        step[0] = step[1] = sizeof(_Tp);
+        data = datastart = (uchar*)&vec[0];
+        datalimit = dataend = datastart + rows * step[0];
+    }
+    else
+        Mat((int)vec.size(), 1, DataType<_Tp>::type, (uchar*)&vec[0]).copyTo(*this);
+}
+
+template<typename _Tp, int n> inline
+Mat::Mat(const Vec<_Tp, n>& vec, bool copyData)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(2), rows(n), cols(1), data(0),
+      refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    if( !copyData )
+    {
+        step[0] = step[1] = sizeof(_Tp);
+        data = datastart = (uchar*)vec.val;
+        datalimit = dataend = datastart + rows * step[0];
+    }
+    else
+        Mat(n, 1, DataType<_Tp>::type, (void*)vec.val).copyTo(*this);
+}
+
+
+template<typename _Tp, int m, int n> inline
+Mat::Mat(const Matx<_Tp,m,n>& M, bool copyData)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(2), rows(m), cols(n), data(0),
+      refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    if( !copyData )
+    {
+        step[0] = cols * sizeof(_Tp);
+        step[1] = sizeof(_Tp);
+        data = datastart = (uchar*)M.val;
+        datalimit = dataend = datastart + rows * step[0];
+    }
+    else
+        Mat(m, n, DataType<_Tp>::type, (uchar*)M.val).copyTo(*this);
+}
+
+template<typename _Tp> inline
+Mat::Mat(const Point_<_Tp>& pt, bool copyData)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(2), rows(2), cols(1), data(0),
+      refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    if( !copyData )
+    {
+        step[0] = step[1] = sizeof(_Tp);
+        data = datastart = (uchar*)&pt.x;
+        datalimit = dataend = datastart + rows * step[0];
+    }
+    else
+    {
+        create(2, 1, DataType<_Tp>::type);
+        ((_Tp*)data)[0] = pt.x;
+        ((_Tp*)data)[1] = pt.y;
+    }
+}
+
+template<typename _Tp> inline
+Mat::Mat(const Point3_<_Tp>& pt, bool copyData)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(2), rows(3), cols(1), data(0),
+      refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    if( !copyData )
+    {
+        step[0] = step[1] = sizeof(_Tp);
+        data = datastart = (uchar*)&pt.x;
+        datalimit = dataend = datastart + rows * step[0];
+    }
+    else
+    {
+        create(3, 1, DataType<_Tp>::type);
+        ((_Tp*)data)[0] = pt.x;
+        ((_Tp*)data)[1] = pt.y;
+        ((_Tp*)data)[2] = pt.z;
+    }
+}
+
+template<typename _Tp> inline
+Mat::Mat(const MatCommaInitializer_<_Tp>& commaInitializer)
+    : flags(MAGIC_VAL | DataType<_Tp>::type | CV_MAT_CONT_FLAG), dims(0), rows(0), cols(0), data(0),
+      refcount(0), datastart(0), dataend(0), allocator(0), size(&rows)
+{
+    *this = commaInitializer.operator Mat_<_Tp>();
+}
+
+inline
+Mat::~Mat()
+{
+    release();
+    if( step.p != step.buf )
+        fastFree(step.p);
+}
+
+inline
+Mat& Mat::operator = (const Mat& m)
+{
+    if( this != &m )
+    {
+        if( m.refcount )
+            CV_XADD(m.refcount, 1);
+        release();
+        flags = m.flags;
+        if( dims <= 2 && m.dims <= 2 )
+        {
+            dims = m.dims;
+            rows = m.rows;
+            cols = m.cols;
+            step[0] = m.step[0];
+            step[1] = m.step[1];
+        }
+        else
+            copySize(m);
+        data = m.data;
+        datastart = m.datastart;
+        dataend = m.dataend;
+        datalimit = m.datalimit;
+        refcount = m.refcount;
+        allocator = m.allocator;
+    }
+    return *this;
+}
+
+inline
+Mat Mat::row(int y) const
+{
+    return Mat(*this, Range(y, y + 1), Range::all());
+}
+
+inline
+Mat Mat::col(int x) const
+{
+    return Mat(*this, Range::all(), Range(x, x + 1));
+}
+
+inline
+Mat Mat::rowRange(int startrow, int endrow) const
+{
+    return Mat(*this, Range(startrow, endrow), Range::all());
+}
+
+inline
+Mat Mat::rowRange(const Range& r) const
+{
+    return Mat(*this, r, Range::all());
+}
+
+inline
+Mat Mat::colRange(int startcol, int endcol) const
+{
+    return Mat(*this, Range::all(), Range(startcol, endcol));
+}
+
+inline
+Mat Mat::colRange(const Range& r) const
+{
+    return Mat(*this, Range::all(), r);
+}
+
+inline
+Mat Mat::clone() const
+{
+    Mat m;
+    copyTo(m);
+    return m;
+}
+
+inline
+void Mat::assignTo( Mat& m, int _type ) const
+{
+    if( _type < 0 )
+        m = *this;
+    else
+        convertTo(m, _type);
+}
+
+inline
+void Mat::create(int _rows, int _cols, int _type)
+{
+    _type &= TYPE_MASK;
+    if( dims <= 2 && rows == _rows && cols == _cols && type() == _type && data )
+        return;
+    int sz[] = {_rows, _cols};
+    create(2, sz, _type);
+}
+
+inline
+void Mat::create(Size _sz, int _type)
+{
+    create(_sz.height, _sz.width, _type);
+}
+
+inline
+void Mat::addref()
+{
+    if( refcount )
+        CV_XADD(refcount, 1);
+}
+
+inline void Mat::release()
+{
+    if( refcount && CV_XADD(refcount, -1) == 1 )
+        deallocate();
+    data = datastart = dataend = datalimit = 0;
+    size.p[0] = 0;
+    refcount = 0;
+}
+
+inline
+Mat Mat::operator()( Range _rowRange, Range _colRange ) const
+{
+    return Mat(*this, _rowRange, _colRange);
+}
+
+inline
+Mat Mat::operator()( const Rect& roi ) const
+{
+    return Mat(*this, roi);
+}
+
+inline
+Mat Mat::operator()(const Range* ranges) const
+{
+    return Mat(*this, ranges);
+}
+
+inline
+bool Mat::isContinuous() const
+{
+    return (flags & CONTINUOUS_FLAG) != 0;
+}
+
+inline
+bool Mat::isSubmatrix() const
+{
+    return (flags & SUBMATRIX_FLAG) != 0;
+}
+
+inline
+size_t Mat::elemSize() const
+{
+    return dims > 0 ? step.p[dims - 1] : 0;
+}
+
+inline
+size_t Mat::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int Mat::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int Mat::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int Mat::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t Mat::step1(int i) const
+{
+    return step.p[i] / elemSize1();
+}
+
+inline
+bool Mat::empty() const
+{
+    return data == 0 || total() == 0;
+}
+
+inline
+size_t Mat::total() const
+{
+    if( dims <= 2 )
+        return (size_t)rows * cols;
+    size_t p = 1;
+    for( int i = 0; i < dims; i++ )
+        p *= size[i];
+    return p;
+}
+
+inline
+uchar* Mat::ptr(int y)
+{
+    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
+    return data + step.p[0] * y;
+}
+
+inline
+const uchar* Mat::ptr(int y) const
+{
+    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
+    return data + step.p[0] * y;
+}
+
+template<typename _Tp> inline
+_Tp* Mat::ptr(int y)
+{
+    CV_DbgAssert( y == 0 || (data && dims >= 1 && (unsigned)y < (unsigned)size.p[0]) );
+    return (_Tp*)(data + step.p[0] * y);
+}
+
+template<typename _Tp> inline
+const _Tp* Mat::ptr(int y) const
+{
+    CV_DbgAssert( y == 0 || (data && dims >= 1 && data && (unsigned)y < (unsigned)size.p[0]) );
+    return (const _Tp*)(data + step.p[0] * y);
+}
+
+inline
+uchar* Mat::ptr(int i0, int i1)
+{
+    CV_DbgAssert( dims >= 2 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] );
+    return data + i0 * step.p[0] + i1 * step.p[1];
+}
+
+inline
+const uchar* Mat::ptr(int i0, int i1) const
+{
+    CV_DbgAssert( dims >= 2 && data &&
+                 (unsigned)i0 < (unsigned)size.p[0] &&
+                 (unsigned)i1 < (unsigned)size.p[1] );
+    return data + i0 * step.p[0] + i1 * step.p[1];
+}
+
+template<typename _Tp> inline
+_Tp* Mat::ptr(int i0, int i1)
+{
+    CV_DbgAssert( dims >= 2 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] );
+    return (_Tp*)(data + i0 * step.p[0] + i1 * step.p[1]);
+}
+
+template<typename _Tp> inline
+const _Tp* Mat::ptr(int i0, int i1) const
+{
+    CV_DbgAssert( dims >= 2 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] );
+    return (const _Tp*)(data + i0 * step.p[0] + i1 * step.p[1]);
+}
+
+inline
+uchar* Mat::ptr(int i0, int i1, int i2)
+{
+    CV_DbgAssert( dims >= 3 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  (unsigned)i2 < (unsigned)size.p[2] );
+    return data + i0 * step.p[0] + i1 * step.p[1] + i2 * step.p[2];
+}
+
+inline
+const uchar* Mat::ptr(int i0, int i1, int i2) const
+{
+    CV_DbgAssert( dims >= 3 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  (unsigned)i2 < (unsigned)size.p[2] );
+    return data + i0 * step.p[0] + i1 * step.p[1] + i2 * step.p[2];
+}
+
+template<typename _Tp> inline
+_Tp* Mat::ptr(int i0, int i1, int i2)
+{
+    CV_DbgAssert( dims >= 3 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  (unsigned)i2 < (unsigned)size.p[2] );
+    return (_Tp*)(data + i0 * step.p[0] + i1 * step.p[1] + i2 * step.p[2]);
+}
+
+template<typename _Tp> inline
+const _Tp* Mat::ptr(int i0, int i1, int i2) const
+{
+    CV_DbgAssert( dims >= 3 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  (unsigned)i2 < (unsigned)size.p[2] );
+    return (const _Tp*)(data + i0 * step.p[0] + i1 * step.p[1] + i2 * step.p[2]);
+}
+
+inline
+uchar* Mat::ptr(const int* idx)
+{
+    int i, d = dims;
+    uchar* p = data;
+    CV_DbgAssert( d >= 1 && p );
+    for( i = 0; i < d; i++ )
+    {
+        CV_DbgAssert( (unsigned)idx[i] < (unsigned)size.p[i] );
+        p += idx[i] * step.p[i];
+    }
+    return p;
+}
+
+inline
+const uchar* Mat::ptr(const int* idx) const
+{
+    int i, d = dims;
+    uchar* p = data;
+    CV_DbgAssert( d >= 1 && p );
+    for( i = 0; i < d; i++ )
+    {
+        CV_DbgAssert( (unsigned)idx[i] < (unsigned)size.p[i] );
+        p += idx[i] * step.p[i];
+    }
+    return p;
+}
+
+template<typename _Tp> inline
+_Tp& Mat::at(int i0, int i1)
+{
+    CV_DbgAssert( dims <= 2 && data && (unsigned)i0 < (unsigned)size.p[0] &&
+        (unsigned)(i1 * DataType<_Tp>::channels) < (unsigned)(size.p[1] * channels()) &&
+        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
+    return ((_Tp*)(data + step.p[0] * i0))[i1];
+}
+
+template<typename _Tp> inline
+const _Tp& Mat::at(int i0, int i1) const
+{
+    CV_DbgAssert( dims <= 2 && data && (unsigned)i0 < (unsigned)size.p[0] &&
+        (unsigned)(i1 * DataType<_Tp>::channels) < (unsigned)(size.p[1] * channels()) &&
+        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
+    return ((const _Tp*)(data + step.p[0] * i0))[i1];
+}
+
+template<typename _Tp> inline
+_Tp& Mat::at(Point pt)
+{
+    CV_DbgAssert( dims <= 2 && data && (unsigned)pt.y < (unsigned)size.p[0] &&
+        (unsigned)(pt.x * DataType<_Tp>::channels) < (unsigned)(size.p[1] * channels()) &&
+        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
+    return ((_Tp*)(data + step.p[0] * pt.y))[pt.x];
+}
+
+template<typename _Tp> inline
+const _Tp& Mat::at(Point pt) const
+{
+    CV_DbgAssert( dims <= 2 && data && (unsigned)pt.y < (unsigned)size.p[0] &&
+        (unsigned)(pt.x * DataType<_Tp>::channels) < (unsigned)(size.p[1] * channels()) &&
+        CV_ELEM_SIZE1(DataType<_Tp>::depth) == elemSize1());
+    return ((const _Tp*)(data + step.p[0] * pt.y))[pt.x];
+}
+
+template<typename _Tp> inline
+_Tp& Mat::at(int i0)
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                 (unsigned)i0 < (unsigned)(size.p[0] * size.p[1]) &&
+                 elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    if( isContinuous() || size.p[0] == 1 )
+        return ((_Tp*)data)[i0];
+    if( size.p[1] == 1 )
+        return *(_Tp*)(data + step.p[0] * i0);
+    int i = i0 / cols, j = i0 - i * cols;
+    return ((_Tp*)(data + step.p[0] * i))[j];
+}
+
+template<typename _Tp> inline
+const _Tp& Mat::at(int i0) const
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                 (unsigned)i0 < (unsigned)(size.p[0] * size.p[1]) &&
+                 elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    if( isContinuous() || size.p[0] == 1 )
+        return ((const _Tp*)data)[i0];
+    if( size.p[1] == 1 )
+        return *(const _Tp*)(data + step.p[0] * i0);
+    int i = i0 / cols, j = i0 - i * cols;
+    return ((const _Tp*)(data + step.p[0] * i))[j];
+}
+
+template<typename _Tp> inline
+_Tp& Mat::at(int i0, int i1, int i2)
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(_Tp*)ptr(i0, i1, i2);
+}
+
+template<typename _Tp> inline
+const _Tp& Mat::at(int i0, int i1, int i2) const
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(const _Tp*)ptr(i0, i1, i2);
+}
+
+template<typename _Tp> inline
+_Tp& Mat::at(const int* idx)
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(_Tp*)ptr(idx);
+}
+
+template<typename _Tp> inline
+const _Tp& Mat::at(const int* idx) const
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(const _Tp*)ptr(idx);
+}
+
+template<typename _Tp, int n> inline
+_Tp& Mat::at(const Vec<int, n>& idx)
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(_Tp*)ptr(idx.val);
+}
+
+template<typename _Tp, int n> inline
+const _Tp& Mat::at(const Vec<int, n>& idx) const
+{
+    CV_DbgAssert( elemSize() == CV_ELEM_SIZE(DataType<_Tp>::type) );
+    return *(const _Tp*)ptr(idx.val);
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> Mat::begin() const
+{
+    CV_DbgAssert( elemSize() == sizeof(_Tp) );
+    return MatConstIterator_<_Tp>((const Mat_<_Tp>*)this);
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> Mat::end() const
+{
+    CV_DbgAssert( elemSize() == sizeof(_Tp) );
+    MatConstIterator_<_Tp> it((const Mat_<_Tp>*)this);
+    it += total();
+    return it;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> Mat::begin()
+{
+    CV_DbgAssert( elemSize() == sizeof(_Tp) );
+    return MatIterator_<_Tp>((Mat_<_Tp>*)this);
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> Mat::end()
+{
+    CV_DbgAssert( elemSize() == sizeof(_Tp) );
+    MatIterator_<_Tp> it((Mat_<_Tp>*)this);
+    it += total();
+    return it;
+}
+
+template<typename _Tp> inline
+Mat::operator std::vector<_Tp>() const
+{
+    std::vector<_Tp> v;
+    copyTo(v);
+    return v;
+}
+
+template<typename _Tp, int n> inline
+Mat::operator Vec<_Tp, n>() const
+{
+    CV_Assert( data && dims <= 2 && (rows == 1 || cols == 1) &&
+               rows + cols - 1 == n && channels() == 1 );
+
+    if( isContinuous() && type() == DataType<_Tp>::type )
+        return Vec<_Tp, n>((_Tp*)data);
+    Vec<_Tp, n> v;
+    Mat tmp(rows, cols, DataType<_Tp>::type, v.val);
+    convertTo(tmp, tmp.type());
+    return v;
+}
+
+template<typename _Tp, int m, int n> inline
+Mat::operator Matx<_Tp, m, n>() const
+{
+    CV_Assert( data && dims <= 2 && rows == m && cols == n && channels() == 1 );
+
+    if( isContinuous() && type() == DataType<_Tp>::type )
+        return Matx<_Tp, m, n>((_Tp*)data);
+    Matx<_Tp, m, n> mtx;
+    Mat tmp(rows, cols, DataType<_Tp>::type, mtx.val);
+    convertTo(tmp, tmp.type());
+    return mtx;
+}
+
+template<typename _Tp> inline
+void Mat::push_back(const _Tp& elem)
+{
+    if( !data )
+    {
+        *this = Mat(1, 1, DataType<_Tp>::type, (void*)&elem).clone();
+        return;
+    }
+    CV_Assert(DataType<_Tp>::type == type() && cols == 1
+              /* && dims == 2 (cols == 1 implies dims == 2) */);
+    uchar* tmp = dataend + step[0];
+    if( !isSubmatrix() && isContinuous() && tmp <= datalimit )
+    {
+        *(_Tp*)(data + (size.p[0]++) * step.p[0]) = elem;
+        dataend = tmp;
+    }
+    else
+        push_back_(&elem);
+}
+
+template<typename _Tp> inline
+void Mat::push_back(const Mat_<_Tp>& m)
+{
+    push_back((const Mat&)m);
+}
+
+
+
+///////////////////////////// Mat::MSize ////////////////////////////
+
+inline
+Mat::MSize::MSize(int* _p)
+    : p(_p) {}
+
+inline
+Size Mat::MSize::operator()() const
+{
+    CV_DbgAssert(p[-1] <= 2);
+    return Size(p[1], p[0]);
+}
+
+inline
+const int& Mat::MSize::operator[](int i) const
+{
+    return p[i];
+}
+
+inline
+int& Mat::MSize::operator[](int i)
+{
+    return p[i];
+}
+
+inline
+Mat::MSize::operator const int*() const
+{
+    return p;
+}
+
+inline
+bool Mat::MSize::operator == (const MSize& sz) const
+{
+    int d = p[-1];
+    int dsz = sz.p[-1];
+    if( d != dsz )
+        return false;
+    if( d == 2 )
+        return p[0] == sz.p[0] && p[1] == sz.p[1];
+
+    for( int i = 0; i < d; i++ )
+        if( p[i] != sz.p[i] )
+            return false;
+    return true;
+}
+
+inline
+bool Mat::MSize::operator != (const MSize& sz) const
+{
+    return !(*this == sz);
+}
+
+
+
+///////////////////////////// Mat::MStep ////////////////////////////
+
+inline
+Mat::MStep::MStep()
+{
+    p = buf; p[0] = p[1] = 0;
+}
+
+inline
+Mat::MStep::MStep(size_t s)
+{
+    p = buf; p[0] = s; p[1] = 0;
+}
+
+inline
+const size_t& Mat::MStep::operator[](int i) const
+{
+    return p[i];
+}
+
+inline
+size_t& Mat::MStep::operator[](int i)
+{
+    return p[i];
+}
+
+inline Mat::MStep::operator size_t() const
+{
+    CV_DbgAssert( p == buf );
+    return buf[0];
+}
+
+inline Mat::MStep& Mat::MStep::operator = (size_t s)
+{
+    CV_DbgAssert( p == buf );
+    buf[0] = s;
+    return *this;
+}
+
+
+
+////////////////////////////// Mat_<_Tp> ////////////////////////////
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_()
+    : Mat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<_Tp>::type;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(int _rows, int _cols)
+    : Mat(_rows, _cols, DataType<_Tp>::type)
+{
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(int _rows, int _cols, const _Tp& value)
+    : Mat(_rows, _cols, DataType<_Tp>::type)
+{
+    *this = value;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(Size _sz)
+    : Mat(_sz.height, _sz.width, DataType<_Tp>::type)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(Size _sz, const _Tp& value)
+    : Mat(_sz.height, _sz.width, DataType<_Tp>::type)
+{
+    *this = value;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(int _dims, const int* _sz)
+    : Mat(_dims, _sz, DataType<_Tp>::type)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(int _dims, const int* _sz, const _Tp& _s)
+    : Mat(_dims, _sz, DataType<_Tp>::type, Scalar(_s))
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Mat_<_Tp>& m, const Range* ranges)
+    : Mat(m, ranges)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Mat& m)
+    : Mat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<_Tp>::type;
+    *this = m;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Mat_& m)
+    : Mat(m)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(int _rows, int _cols, _Tp* _data, size_t steps)
+    : Mat(_rows, _cols, DataType<_Tp>::type, _data, steps)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Mat_& m, const Range& _rowRange, const Range& _colRange)
+    : Mat(m, _rowRange, _colRange)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Mat_& m, const Rect& roi)
+    : Mat(m, roi)
+{}
+
+template<typename _Tp> template<int n> inline
+Mat_<_Tp>::Mat_(const Vec<typename DataType<_Tp>::channel_type, n>& vec, bool copyData)
+    : Mat(n / DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&vec)
+{
+    CV_Assert(n%DataType<_Tp>::channels == 0);
+    if( copyData )
+        *this = clone();
+}
+
+template<typename _Tp> template<int m, int n> inline
+Mat_<_Tp>::Mat_(const Matx<typename DataType<_Tp>::channel_type, m, n>& M, bool copyData)
+    : Mat(m, n / DataType<_Tp>::channels, DataType<_Tp>::type, (void*)&M)
+{
+    CV_Assert(n % DataType<_Tp>::channels == 0);
+    if( copyData )
+        *this = clone();
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Point_<typename DataType<_Tp>::channel_type>& pt, bool copyData)
+    : Mat(2 / DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&pt)
+{
+    CV_Assert(2 % DataType<_Tp>::channels == 0);
+    if( copyData )
+        *this = clone();
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const Point3_<typename DataType<_Tp>::channel_type>& pt, bool copyData)
+    : Mat(3 / DataType<_Tp>::channels, 1, DataType<_Tp>::type, (void*)&pt)
+{
+    CV_Assert(3 % DataType<_Tp>::channels == 0);
+    if( copyData )
+        *this = clone();
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const MatCommaInitializer_<_Tp>& commaInitializer)
+    : Mat(commaInitializer)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const std::vector<_Tp>& vec, bool copyData)
+    : Mat(vec, copyData)
+{}
+
+template<typename _Tp> inline
+Mat_<_Tp>& Mat_<_Tp>::operator = (const Mat& m)
+{
+    if( DataType<_Tp>::type == m.type() )
+    {
+        Mat::operator = (m);
+        return *this;
+    }
+    if( DataType<_Tp>::depth == m.depth() )
+    {
+        return (*this = m.reshape(DataType<_Tp>::channels, m.dims, 0));
+    }
+    CV_DbgAssert(DataType<_Tp>::channels == m.channels());
+    m.convertTo(*this, type());
+    return *this;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>& Mat_<_Tp>::operator = (const Mat_& m)
+{
+    Mat::operator=(m);
+    return *this;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>& Mat_<_Tp>::operator = (const _Tp& s)
+{
+    typedef typename DataType<_Tp>::vec_type VT;
+    Mat::operator=(Scalar((const VT&)s));
+    return *this;
+}
+
+template<typename _Tp> inline
+void Mat_<_Tp>::create(int _rows, int _cols)
+{
+    Mat::create(_rows, _cols, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+void Mat_<_Tp>::create(Size _sz)
+{
+    Mat::create(_sz, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+void Mat_<_Tp>::create(int _dims, const int* _sz)
+{
+    Mat::create(_dims, _sz, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::cross(const Mat_& m) const
+{
+    return Mat_<_Tp>(Mat::cross(m));
+}
+
+template<typename _Tp> template<typename T2> inline
+Mat_<_Tp>::operator Mat_<T2>() const
+{
+    return Mat_<T2>(*this);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::row(int y) const
+{
+    return Mat_(*this, Range(y, y+1), Range::all());
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::col(int x) const
+{
+    return Mat_(*this, Range::all(), Range(x, x+1));
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::diag(int d) const
+{
+    return Mat_(Mat::diag(d));
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::clone() const
+{
+    return Mat_(Mat::clone());
+}
+
+template<typename _Tp> inline
+size_t Mat_<_Tp>::elemSize() const
+{
+    CV_DbgAssert( Mat::elemSize() == sizeof(_Tp) );
+    return sizeof(_Tp);
+}
+
+template<typename _Tp> inline
+size_t Mat_<_Tp>::elemSize1() const
+{
+    CV_DbgAssert( Mat::elemSize1() == sizeof(_Tp) / DataType<_Tp>::channels );
+    return sizeof(_Tp) / DataType<_Tp>::channels;
+}
+
+template<typename _Tp> inline
+int Mat_<_Tp>::type() const
+{
+    CV_DbgAssert( Mat::type() == DataType<_Tp>::type );
+    return DataType<_Tp>::type;
+}
+
+template<typename _Tp> inline
+int Mat_<_Tp>::depth() const
+{
+    CV_DbgAssert( Mat::depth() == DataType<_Tp>::depth );
+    return DataType<_Tp>::depth;
+}
+
+template<typename _Tp> inline
+int Mat_<_Tp>::channels() const
+{
+    CV_DbgAssert( Mat::channels() == DataType<_Tp>::channels );
+    return DataType<_Tp>::channels;
+}
+
+template<typename _Tp> inline
+size_t Mat_<_Tp>::stepT(int i) const
+{
+    return step.p[i] / elemSize();
+}
+
+template<typename _Tp> inline
+size_t Mat_<_Tp>::step1(int i) const
+{
+    return step.p[i] / elemSize1();
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>& Mat_<_Tp>::adjustROI( int dtop, int dbottom, int dleft, int dright )
+{
+    return (Mat_<_Tp>&)(Mat::adjustROI(dtop, dbottom, dleft, dright));
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::operator()( const Range& _rowRange, const Range& _colRange ) const
+{
+    return Mat_<_Tp>(*this, _rowRange, _colRange);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::operator()( const Rect& roi ) const
+{
+    return Mat_<_Tp>(*this, roi);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp> Mat_<_Tp>::operator()( const Range* ranges ) const
+{
+    return Mat_<_Tp>(*this, ranges);
+}
+
+template<typename _Tp> inline
+_Tp* Mat_<_Tp>::operator [](int y)
+{
+    return (_Tp*)ptr(y);
+}
+
+template<typename _Tp> inline
+const _Tp* Mat_<_Tp>::operator [](int y) const
+{
+    return (const _Tp*)ptr(y);
+}
+
+template<typename _Tp> inline
+_Tp& Mat_<_Tp>::operator ()(int i0, int i1)
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  type() == DataType<_Tp>::type );
+    return ((_Tp*)(data + step.p[0] * i0))[i1];
+}
+
+template<typename _Tp> inline
+const _Tp& Mat_<_Tp>::operator ()(int i0, int i1) const
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                  (unsigned)i0 < (unsigned)size.p[0] &&
+                  (unsigned)i1 < (unsigned)size.p[1] &&
+                  type() == DataType<_Tp>::type );
+    return ((const _Tp*)(data + step.p[0] * i0))[i1];
+}
+
+template<typename _Tp> inline
+_Tp& Mat_<_Tp>::operator ()(Point pt)
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                  (unsigned)pt.y < (unsigned)size.p[0] &&
+                  (unsigned)pt.x < (unsigned)size.p[1] &&
+                  type() == DataType<_Tp>::type );
+    return ((_Tp*)(data + step.p[0] * pt.y))[pt.x];
+}
+
+template<typename _Tp> inline
+const _Tp& Mat_<_Tp>::operator ()(Point pt) const
+{
+    CV_DbgAssert( dims <= 2 && data &&
+                  (unsigned)pt.y < (unsigned)size.p[0] &&
+                  (unsigned)pt.x < (unsigned)size.p[1] &&
+                 type() == DataType<_Tp>::type );
+    return ((const _Tp*)(data + step.p[0] * pt.y))[pt.x];
+}
+
+template<typename _Tp> inline
+_Tp& Mat_<_Tp>::operator ()(const int* idx)
+{
+    return Mat::at<_Tp>(idx);
+}
+
+template<typename _Tp> inline
+const _Tp& Mat_<_Tp>::operator ()(const int* idx) const
+{
+    return Mat::at<_Tp>(idx);
+}
+
+template<typename _Tp> template<int n> inline
+_Tp& Mat_<_Tp>::operator ()(const Vec<int, n>& idx)
+{
+    return Mat::at<_Tp>(idx);
+}
+
+template<typename _Tp> template<int n> inline
+const _Tp& Mat_<_Tp>::operator ()(const Vec<int, n>& idx) const
+{
+    return Mat::at<_Tp>(idx);
+}
+
+template<typename _Tp> inline
+_Tp& Mat_<_Tp>::operator ()(int i0)
+{
+    return this->at<_Tp>(i0);
+}
+
+template<typename _Tp> inline
+const _Tp& Mat_<_Tp>::operator ()(int i0) const
+{
+    return this->at<_Tp>(i0);
+}
+
+template<typename _Tp> inline
+_Tp& Mat_<_Tp>::operator ()(int i0, int i1, int i2)
+{
+    return this->at<_Tp>(i0, i1, i2);
+}
+
+template<typename _Tp> inline
+const _Tp& Mat_<_Tp>::operator ()(int i0, int i1, int i2) const
+{
+    return this->at<_Tp>(i0, i1, i2);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::operator std::vector<_Tp>() const
+{
+    std::vector<_Tp> v;
+    copyTo(v);
+    return v;
+}
+
+template<typename _Tp> template<int n> inline
+Mat_<_Tp>::operator Vec<typename DataType<_Tp>::channel_type, n>() const
+{
+    CV_Assert(n % DataType<_Tp>::channels == 0);
+    return this->Mat::operator Vec<typename DataType<_Tp>::channel_type, n>();
+}
+
+template<typename _Tp> template<int m, int n> inline
+Mat_<_Tp>::operator Matx<typename DataType<_Tp>::channel_type, m, n>() const
+{
+    CV_Assert(n % DataType<_Tp>::channels == 0);
+    return this->Mat::operator Matx<typename DataType<_Tp>::channel_type, m, n>();
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> Mat_<_Tp>::begin() const
+{
+    return Mat::begin<_Tp>();
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> Mat_<_Tp>::end() const
+{
+    return Mat::end<_Tp>();
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> Mat_<_Tp>::begin()
+{
+    return Mat::begin<_Tp>();
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> Mat_<_Tp>::end()
+{
+    return Mat::end<_Tp>();
+}
+
+
+/*template<typename T1, typename T2, typename Op> inline
+void process( const Mat_<T1>& m1, Mat_<T2>& m2, Op op )
+{
+    int y, x, rows = m1.rows, cols = m1.cols;
+
+    CV_DbgAssert( m1.size() == m2.size() );
+
+    for( y = 0; y < rows; y++ )
+    {
+        const T1* src = m1[y];
+        T2* dst = m2[y];
+
+        for( x = 0; x < cols; x++ )
+            dst[x] = op(src[x]);
+    }
+}
+
+template<typename T1, typename T2, typename T3, typename Op> inline
+void process( const Mat_<T1>& m1, const Mat_<T2>& m2, Mat_<T3>& m3, Op op )
+{
+    int y, x, rows = m1.rows, cols = m1.cols;
+
+    CV_DbgAssert( m1.size() == m2.size() );
+
+    for( y = 0; y < rows; y++ )
+    {
+        const T1* src1 = m1[y];
+        const T2* src2 = m2[y];
+        T3* dst = m3[y];
+
+        for( x = 0; x < cols; x++ )
+            dst[x] = op( src1[x], src2[x] );
+    }
+}*/
+
+
+
+///////////////////////////// SparseMat /////////////////////////////
+
+inline
+SparseMat::SparseMat()
+    : flags(MAGIC_VAL), hdr(0)
+{}
+
+inline
+SparseMat::SparseMat(int _dims, const int* _sizes, int _type)
+    : flags(MAGIC_VAL), hdr(0)
+{
+    create(_dims, _sizes, _type);
+}
+
+inline
+SparseMat::SparseMat(const SparseMat& m)
+    : flags(m.flags), hdr(m.hdr)
+{
+    addref();
+}
+
+inline
+SparseMat::~SparseMat()
+{
+    release();
+}
+
+inline
+SparseMat& SparseMat::operator = (const SparseMat& m)
+{
+    if( this != &m )
+    {
+        if( m.hdr )
+            CV_XADD(&m.hdr->refcount, 1);
+        release();
+        flags = m.flags;
+        hdr = m.hdr;
+    }
+    return *this;
+}
+
+inline
+SparseMat& SparseMat::operator = (const Mat& m)
+{
+    return (*this = SparseMat(m));
+}
+
+inline
+SparseMat SparseMat::clone() const
+{
+    SparseMat temp;
+    this->copyTo(temp);
+    return temp;
+}
+
+inline
+void SparseMat::assignTo( SparseMat& m, int _type ) const
+{
+    if( _type < 0 )
+        m = *this;
+    else
+        convertTo(m, _type);
+}
+
+inline
+void SparseMat::addref()
+{
+    if( hdr )
+        CV_XADD(&hdr->refcount, 1);
+}
+
+inline
+void SparseMat::release()
+{
+    if( hdr && CV_XADD(&hdr->refcount, -1) == 1 )
+        delete hdr;
+    hdr = 0;
+}
+
+inline
+size_t SparseMat::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t SparseMat::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int SparseMat::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int SparseMat::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int SparseMat::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+const int* SparseMat::size() const
+{
+    return hdr ? hdr->size : 0;
+}
+
+inline
+int SparseMat::size(int i) const
+{
+    if( hdr )
+    {
+        CV_DbgAssert((unsigned)i < (unsigned)hdr->dims);
+        return hdr->size[i];
+    }
+    return 0;
+}
+
+inline
+int SparseMat::dims() const
+{
+    return hdr ? hdr->dims : 0;
+}
+
+inline
+size_t SparseMat::nzcount() const
+{
+    return hdr ? hdr->nodeCount : 0;
+}
+
+inline
+size_t SparseMat::hash(int i0) const
+{
+    return (size_t)i0;
+}
+
+inline
+size_t SparseMat::hash(int i0, int i1) const
+{
+    return (size_t)(unsigned)i0 * HASH_SCALE + (unsigned)i1;
+}
+
+inline
+size_t SparseMat::hash(int i0, int i1, int i2) const
+{
+    return ((size_t)(unsigned)i0 * HASH_SCALE + (unsigned)i1) * HASH_SCALE + (unsigned)i2;
+}
+
+inline
+size_t SparseMat::hash(const int* idx) const
+{
+    size_t h = (unsigned)idx[0];
+    if( !hdr )
+        return 0;
+    int d = hdr->dims;
+    for(int i = 1; i < d; i++ )
+        h = h * HASH_SCALE + (unsigned)idx[i];
+    return h;
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat::ref(int i0, size_t* hashval)
+{
+    return *(_Tp*)((SparseMat*)this)->ptr(i0, true, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat::ref(int i0, int i1, size_t* hashval)
+{
+    return *(_Tp*)((SparseMat*)this)->ptr(i0, i1, true, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat::ref(int i0, int i1, int i2, size_t* hashval)
+{
+    return *(_Tp*)((SparseMat*)this)->ptr(i0, i1, i2, true, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat::ref(const int* idx, size_t* hashval)
+{
+    return *(_Tp*)((SparseMat*)this)->ptr(idx, true, hashval);
+}
+
+template<typename _Tp> inline
+_Tp SparseMat::value(int i0, size_t* hashval) const
+{
+    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, false, hashval);
+    return p ? *p : _Tp();
+}
+
+template<typename _Tp> inline
+_Tp SparseMat::value(int i0, int i1, size_t* hashval) const
+{
+    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, i1, false, hashval);
+    return p ? *p : _Tp();
+}
+
+template<typename _Tp> inline
+_Tp SparseMat::value(int i0, int i1, int i2, size_t* hashval) const
+{
+    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(i0, i1, i2, false, hashval);
+    return p ? *p : _Tp();
+}
+
+template<typename _Tp> inline
+_Tp SparseMat::value(const int* idx, size_t* hashval) const
+{
+    const _Tp* p = (const _Tp*)((SparseMat*)this)->ptr(idx, false, hashval);
+    return p ? *p : _Tp();
+}
+
+template<typename _Tp> inline
+const _Tp* SparseMat::find(int i0, size_t* hashval) const
+{
+    return (const _Tp*)((SparseMat*)this)->ptr(i0, false, hashval);
+}
+
+template<typename _Tp> inline
+const _Tp* SparseMat::find(int i0, int i1, size_t* hashval) const
+{
+    return (const _Tp*)((SparseMat*)this)->ptr(i0, i1, false, hashval);
+}
+
+template<typename _Tp> inline
+const _Tp* SparseMat::find(int i0, int i1, int i2, size_t* hashval) const
+{
+    return (const _Tp*)((SparseMat*)this)->ptr(i0, i1, i2, false, hashval);
+}
+
+template<typename _Tp> inline
+const _Tp* SparseMat::find(const int* idx, size_t* hashval) const
+{
+    return (const _Tp*)((SparseMat*)this)->ptr(idx, false, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat::value(Node* n)
+{
+    return *(_Tp*)((uchar*)n + hdr->valueOffset);
+}
+
+template<typename _Tp> inline
+const _Tp& SparseMat::value(const Node* n) const
+{
+    return *(const _Tp*)((const uchar*)n + hdr->valueOffset);
+}
+
+inline
+SparseMat::Node* SparseMat::node(size_t nidx)
+{
+    return (Node*)&hdr->pool[nidx];
+}
+
+inline
+const SparseMat::Node* SparseMat::node(size_t nidx) const
+{
+    return (const Node*)&hdr->pool[nidx];
+}
+
+inline
+SparseMatIterator SparseMat::begin()
+{
+    return SparseMatIterator(this);
+}
+
+inline
+SparseMatConstIterator SparseMat::begin() const
+{
+    return SparseMatConstIterator(this);
+}
+
+inline
+SparseMatIterator SparseMat::end()
+{
+    SparseMatIterator it(this);
+    it.seekEnd();
+    return it;
+}
+
+inline
+SparseMatConstIterator SparseMat::end() const
+{
+    SparseMatConstIterator it(this);
+    it.seekEnd();
+    return it;
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp> SparseMat::begin()
+{
+    return SparseMatIterator_<_Tp>(this);
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp> SparseMat::begin() const
+{
+    return SparseMatConstIterator_<_Tp>(this);
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp> SparseMat::end()
+{
+    SparseMatIterator_<_Tp> it(this);
+    it.seekEnd();
+    return it;
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp> SparseMat::end() const
+{
+    SparseMatConstIterator_<_Tp> it(this);
+    it.seekEnd();
+    return it;
+}
+
+
+
+///////////////////////////// SparseMat_ ////////////////////////////
+
+template<typename _Tp> inline
+SparseMat_<_Tp>::SparseMat_()
+{
+    flags = MAGIC_VAL | DataType<_Tp>::type;
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>::SparseMat_(int _dims, const int* _sizes)
+    : SparseMat(_dims, _sizes, DataType<_Tp>::type)
+{}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>::SparseMat_(const SparseMat& m)
+{
+    if( m.type() == DataType<_Tp>::type )
+        *this = (const SparseMat_<_Tp>&)m;
+    else
+        m.convertTo(this, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>::SparseMat_(const SparseMat_<_Tp>& m)
+{
+    this->flags = m.flags;
+    this->hdr = m.hdr;
+    if( this->hdr )
+        CV_XADD(&this->hdr->refcount, 1);
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>::SparseMat_(const Mat& m)
+{
+    SparseMat sm(m);
+    *this = sm;
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>& SparseMat_<_Tp>::operator = (const SparseMat_<_Tp>& m)
+{
+    if( this != &m )
+    {
+        if( m.hdr ) CV_XADD(&m.hdr->refcount, 1);
+        release();
+        flags = m.flags;
+        hdr = m.hdr;
+    }
+    return *this;
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>& SparseMat_<_Tp>::operator = (const SparseMat& m)
+{
+    if( m.type() == DataType<_Tp>::type )
+        return (*this = (const SparseMat_<_Tp>&)m);
+    m.convertTo(*this, DataType<_Tp>::type);
+    return *this;
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp>& SparseMat_<_Tp>::operator = (const Mat& m)
+{
+    return (*this = SparseMat(m));
+}
+
+template<typename _Tp> inline
+SparseMat_<_Tp> SparseMat_<_Tp>::clone() const
+{
+    SparseMat_<_Tp> m;
+    this->copyTo(m);
+    return m;
+}
+
+template<typename _Tp> inline
+void SparseMat_<_Tp>::create(int _dims, const int* _sizes)
+{
+    SparseMat::create(_dims, _sizes, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+int SparseMat_<_Tp>::type() const
+{
+    return DataType<_Tp>::type;
+}
+
+template<typename _Tp> inline
+int SparseMat_<_Tp>::depth() const
+{
+    return DataType<_Tp>::depth;
+}
+
+template<typename _Tp> inline
+int SparseMat_<_Tp>::channels() const
+{
+    return DataType<_Tp>::channels;
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat_<_Tp>::ref(int i0, size_t* hashval)
+{
+    return SparseMat::ref<_Tp>(i0, hashval);
+}
+
+template<typename _Tp> inline
+_Tp SparseMat_<_Tp>::operator()(int i0, size_t* hashval) const
+{
+    return SparseMat::value<_Tp>(i0, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat_<_Tp>::ref(int i0, int i1, size_t* hashval)
+{
+    return SparseMat::ref<_Tp>(i0, i1, hashval);
+}
+
+template<typename _Tp> inline
+_Tp SparseMat_<_Tp>::operator()(int i0, int i1, size_t* hashval) const
+{
+    return SparseMat::value<_Tp>(i0, i1, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat_<_Tp>::ref(int i0, int i1, int i2, size_t* hashval)
+{
+    return SparseMat::ref<_Tp>(i0, i1, i2, hashval);
+}
+
+template<typename _Tp> inline
+_Tp SparseMat_<_Tp>::operator()(int i0, int i1, int i2, size_t* hashval) const
+{
+    return SparseMat::value<_Tp>(i0, i1, i2, hashval);
+}
+
+template<typename _Tp> inline
+_Tp& SparseMat_<_Tp>::ref(const int* idx, size_t* hashval)
+{
+    return SparseMat::ref<_Tp>(idx, hashval);
+}
+
+template<typename _Tp> inline
+_Tp SparseMat_<_Tp>::operator()(const int* idx, size_t* hashval) const
+{
+    return SparseMat::value<_Tp>(idx, hashval);
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp> SparseMat_<_Tp>::begin()
+{
+    return SparseMatIterator_<_Tp>(this);
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp> SparseMat_<_Tp>::begin() const
+{
+    return SparseMatConstIterator_<_Tp>(this);
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp> SparseMat_<_Tp>::end()
+{
+    SparseMatIterator_<_Tp> it(this);
+    it.seekEnd();
+    return it;
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp> SparseMat_<_Tp>::end() const
+{
+    SparseMatConstIterator_<_Tp> it(this);
+    it.seekEnd();
+    return it;
+}
+
+
+
+////////////////////////// MatConstIterator /////////////////////////
+
+inline
+MatConstIterator::MatConstIterator()
+    : m(0), elemSize(0), ptr(0), sliceStart(0), sliceEnd(0)
+{}
+
+inline
+MatConstIterator::MatConstIterator(const Mat* _m)
+    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
+{
+    if( m && m->isContinuous() )
+    {
+        sliceStart = m->data;
+        sliceEnd = sliceStart + m->total()*elemSize;
+    }
+    seek((const int*)0);
+}
+
+inline
+MatConstIterator::MatConstIterator(const Mat* _m, int _row, int _col)
+    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
+{
+    CV_Assert(m && m->dims <= 2);
+    if( m->isContinuous() )
+    {
+        sliceStart = m->data;
+        sliceEnd = sliceStart + m->total()*elemSize;
+    }
+    int idx[] = {_row, _col};
+    seek(idx);
+}
+
+inline
+MatConstIterator::MatConstIterator(const Mat* _m, Point _pt)
+    : m(_m), elemSize(_m->elemSize()), ptr(0), sliceStart(0), sliceEnd(0)
+{
+    CV_Assert(m && m->dims <= 2);
+    if( m->isContinuous() )
+    {
+        sliceStart = m->data;
+        sliceEnd = sliceStart + m->total()*elemSize;
+    }
+    int idx[] = {_pt.y, _pt.x};
+    seek(idx);
+}
+
+inline
+MatConstIterator::MatConstIterator(const MatConstIterator& it)
+    : m(it.m), elemSize(it.elemSize), ptr(it.ptr), sliceStart(it.sliceStart), sliceEnd(it.sliceEnd)
+{}
+
+inline
+MatConstIterator& MatConstIterator::operator = (const MatConstIterator& it )
+{
+    m = it.m; elemSize = it.elemSize; ptr = it.ptr;
+    sliceStart = it.sliceStart; sliceEnd = it.sliceEnd;
+    return *this;
+}
+
+inline
+uchar* MatConstIterator::operator *() const
+{
+    return ptr;
+}
+
+inline MatConstIterator& MatConstIterator::operator += (ptrdiff_t ofs)
+{
+    if( !m || ofs == 0 )
+        return *this;
+    ptrdiff_t ofsb = ofs*elemSize;
+    ptr += ofsb;
+    if( ptr < sliceStart || sliceEnd <= ptr )
+    {
+        ptr -= ofsb;
+        seek(ofs, true);
+    }
+    return *this;
+}
+
+inline
+MatConstIterator& MatConstIterator::operator -= (ptrdiff_t ofs)
+{
+    return (*this += -ofs);
+}
+
+inline
+MatConstIterator& MatConstIterator::operator --()
+{
+    if( m && (ptr -= elemSize) < sliceStart )
+    {
+        ptr += elemSize;
+        seek(-1, true);
+    }
+    return *this;
+}
+
+inline
+MatConstIterator MatConstIterator::operator --(int)
+{
+    MatConstIterator b = *this;
+    *this += -1;
+    return b;
+}
+
+inline
+MatConstIterator& MatConstIterator::operator ++()
+{
+    if( m && (ptr += elemSize) >= sliceEnd )
+    {
+        ptr -= elemSize;
+        seek(1, true);
+    }
+    return *this;
+}
+
+inline MatConstIterator MatConstIterator::operator ++(int)
+{
+    MatConstIterator b = *this;
+    *this += 1;
+    return b;
+}
+
+
+static inline
+bool operator == (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return a.m == b.m && a.ptr == b.ptr;
+}
+
+static inline
+bool operator != (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return !(a == b);
+}
+
+static inline
+bool operator < (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return a.ptr < b.ptr;
+}
+
+static inline
+bool operator > (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return a.ptr > b.ptr;
+}
+
+static inline
+bool operator <= (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return a.ptr <= b.ptr;
+}
+
+static inline
+bool operator >= (const MatConstIterator& a, const MatConstIterator& b)
+{
+    return a.ptr >= b.ptr;
+}
+
+static inline
+ptrdiff_t operator - (const MatConstIterator& b, const MatConstIterator& a)
+{
+    if( a.m != b.m )
+        return ((size_t)(-1) >> 1);
+    if( a.sliceEnd == b.sliceEnd )
+        return (b.ptr - a.ptr)/b.elemSize;
+
+    return b.lpos() - a.lpos();
+}
+
+static inline
+MatConstIterator operator + (const MatConstIterator& a, ptrdiff_t ofs)
+{
+    MatConstIterator b = a;
+    return b += ofs;
+}
+
+static inline
+MatConstIterator operator + (ptrdiff_t ofs, const MatConstIterator& a)
+{
+    MatConstIterator b = a;
+    return b += ofs;
+}
+
+static inline
+MatConstIterator operator - (const MatConstIterator& a, ptrdiff_t ofs)
+{
+    MatConstIterator b = a;
+    return b += -ofs;
+}
+
+
+inline
+uchar* MatConstIterator::operator [](ptrdiff_t i) const
+{
+    return *(*this + i);
+}
+
+
+
+///////////////////////// MatConstIterator_ /////////////////////////
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>::MatConstIterator_()
+{}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>::MatConstIterator_(const Mat_<_Tp>* _m)
+    : MatConstIterator(_m)
+{}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>::MatConstIterator_(const Mat_<_Tp>* _m, int _row, int _col)
+    : MatConstIterator(_m, _row, _col)
+{}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>::MatConstIterator_(const Mat_<_Tp>* _m, Point _pt)
+    : MatConstIterator(_m, _pt)
+{}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>::MatConstIterator_(const MatConstIterator_& it)
+    : MatConstIterator(it)
+{}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator = (const MatConstIterator_& it )
+{
+    MatConstIterator::operator = (it);
+    return *this;
+}
+
+template<typename _Tp> inline
+_Tp MatConstIterator_<_Tp>::operator *() const
+{
+    return *(_Tp*)(this->ptr);
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator += (ptrdiff_t ofs)
+{
+    MatConstIterator::operator += (ofs);
+    return *this;
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator -= (ptrdiff_t ofs)
+{
+    return (*this += -ofs);
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator --()
+{
+    MatConstIterator::operator --();
+    return *this;
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> MatConstIterator_<_Tp>::operator --(int)
+{
+    MatConstIterator_ b = *this;
+    MatConstIterator::operator --();
+    return b;
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp>& MatConstIterator_<_Tp>::operator ++()
+{
+    MatConstIterator::operator ++();
+    return *this;
+}
+
+template<typename _Tp> inline
+MatConstIterator_<_Tp> MatConstIterator_<_Tp>::operator ++(int)
+{
+    MatConstIterator_ b = *this;
+    MatConstIterator::operator ++();
+    return b;
+}
+
+
+template<typename _Tp> inline
+Point MatConstIterator_<_Tp>::pos() const
+{
+    if( !m )
+        return Point();
+    CV_DbgAssert( m->dims <= 2 );
+    if( m->isContinuous() )
+    {
+        ptrdiff_t ofs = (const _Tp*)ptr - (const _Tp*)m->data;
+        int y = (int)(ofs / m->cols);
+        int x = (int)(ofs - (ptrdiff_t)y * m->cols);
+        return Point(x, y);
+    }
+    else
+    {
+        ptrdiff_t ofs = (uchar*)ptr - m->data;
+        int y = (int)(ofs / m->step);
+        int x = (int)((ofs - y * m->step)/sizeof(_Tp));
+        return Point(x, y);
+    }
+}
+
+
+template<typename _Tp> static inline
+bool operator == (const MatConstIterator_<_Tp>& a, const MatConstIterator_<_Tp>& b)
+{
+    return a.m == b.m && a.ptr == b.ptr;
+}
+
+template<typename _Tp> static inline
+bool operator != (const MatConstIterator_<_Tp>& a, const MatConstIterator_<_Tp>& b)
+{
+    return a.m != b.m || a.ptr != b.ptr;
+}
+
+template<typename _Tp> static inline
+MatConstIterator_<_Tp> operator + (const MatConstIterator_<_Tp>& a, ptrdiff_t ofs)
+{
+    MatConstIterator t = (const MatConstIterator&)a + ofs;
+    return (MatConstIterator_<_Tp>&)t;
+}
+
+template<typename _Tp> static inline
+MatConstIterator_<_Tp> operator + (ptrdiff_t ofs, const MatConstIterator_<_Tp>& a)
+{
+    MatConstIterator t = (const MatConstIterator&)a + ofs;
+    return (MatConstIterator_<_Tp>&)t;
+}
+
+template<typename _Tp> static inline
+MatConstIterator_<_Tp> operator - (const MatConstIterator_<_Tp>& a, ptrdiff_t ofs)
+{
+    MatConstIterator t = (const MatConstIterator&)a - ofs;
+    return (MatConstIterator_<_Tp>&)t;
+}
+
+template<typename _Tp> inline
+_Tp MatConstIterator_<_Tp>::operator [](ptrdiff_t i) const
+{
+    return *(_Tp*)MatConstIterator::operator [](i);
+}
+
+
+
+//////////////////////////// MatIterator_ ///////////////////////////
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_()
+    : MatConstIterator_<_Tp>()
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_(Mat_<_Tp>* _m)
+    : MatConstIterator_<_Tp>(_m)
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_(Mat_<_Tp>* _m, int _row, int _col)
+    : MatConstIterator_<_Tp>(_m, _row, _col)
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_(const Mat_<_Tp>* _m, Point _pt)
+    : MatConstIterator_<_Tp>(_m, _pt)
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_(const Mat_<_Tp>* _m, const int* _idx)
+    : MatConstIterator_<_Tp>(_m, _idx)
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>::MatIterator_(const MatIterator_& it)
+    : MatConstIterator_<_Tp>(it)
+{}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>& MatIterator_<_Tp>::operator = (const MatIterator_<_Tp>& it )
+{
+    MatConstIterator::operator = (it);
+    return *this;
+}
+
+template<typename _Tp> inline
+_Tp& MatIterator_<_Tp>::operator *() const
+{
+    return *(_Tp*)(this->ptr);
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>& MatIterator_<_Tp>::operator += (ptrdiff_t ofs)
+{
+    MatConstIterator::operator += (ofs);
+    return *this;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>& MatIterator_<_Tp>::operator -= (ptrdiff_t ofs)
+{
+    MatConstIterator::operator += (-ofs);
+    return *this;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>& MatIterator_<_Tp>::operator --()
+{
+    MatConstIterator::operator --();
+    return *this;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> MatIterator_<_Tp>::operator --(int)
+{
+    MatIterator_ b = *this;
+    MatConstIterator::operator --();
+    return b;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp>& MatIterator_<_Tp>::operator ++()
+{
+    MatConstIterator::operator ++();
+    return *this;
+}
+
+template<typename _Tp> inline
+MatIterator_<_Tp> MatIterator_<_Tp>::operator ++(int)
+{
+    MatIterator_ b = *this;
+    MatConstIterator::operator ++();
+    return b;
+}
+
+template<typename _Tp> inline
+_Tp& MatIterator_<_Tp>::operator [](ptrdiff_t i) const
+{
+    return *(*this + i);
+}
+
+
+template<typename _Tp> static inline
+bool operator == (const MatIterator_<_Tp>& a, const MatIterator_<_Tp>& b)
+{
+    return a.m == b.m && a.ptr == b.ptr;
+}
+
+template<typename _Tp> static inline
+bool operator != (const MatIterator_<_Tp>& a, const MatIterator_<_Tp>& b)
+{
+    return a.m != b.m || a.ptr != b.ptr;
+}
+
+template<typename _Tp> static inline
+MatIterator_<_Tp> operator + (const MatIterator_<_Tp>& a, ptrdiff_t ofs)
+{
+    MatConstIterator t = (const MatConstIterator&)a + ofs;
+    return (MatIterator_<_Tp>&)t;
+}
+
+template<typename _Tp> static inline
+MatIterator_<_Tp> operator + (ptrdiff_t ofs, const MatIterator_<_Tp>& a)
+{
+    MatConstIterator t = (const MatConstIterator&)a + ofs;
+    return (MatIterator_<_Tp>&)t;
+}
+
+template<typename _Tp> static inline
+MatIterator_<_Tp> operator - (const MatIterator_<_Tp>& a, ptrdiff_t ofs)
+{
+    MatConstIterator t = (const MatConstIterator&)a - ofs;
+    return (MatIterator_<_Tp>&)t;
+}
+
+
+
+/////////////////////// SparseMatConstIterator //////////////////////
+
+inline
+SparseMatConstIterator::SparseMatConstIterator()
+    : m(0), hashidx(0), ptr(0)
+{}
+
+inline
+SparseMatConstIterator::SparseMatConstIterator(const SparseMatConstIterator& it)
+    : m(it.m), hashidx(it.hashidx), ptr(it.ptr)
+{}
+
+inline SparseMatConstIterator& SparseMatConstIterator::operator = (const SparseMatConstIterator& it)
+{
+    if( this != &it )
+    {
+        m = it.m;
+        hashidx = it.hashidx;
+        ptr = it.ptr;
+    }
+    return *this;
+}
+
+template<typename _Tp> inline
+const _Tp& SparseMatConstIterator::value() const
+{
+    return *(_Tp*)ptr;
+}
+
+inline
+const SparseMat::Node* SparseMatConstIterator::node() const
+{
+    return (ptr && m && m->hdr) ? (const SparseMat::Node*)(ptr - m->hdr->valueOffset) : 0;
+}
+
+inline
+SparseMatConstIterator SparseMatConstIterator::operator ++(int)
+{
+    SparseMatConstIterator it = *this;
+    ++*this;
+    return it;
+}
+
+inline
+void SparseMatConstIterator::seekEnd()
+{
+    if( m && m->hdr )
+    {
+        hashidx = m->hdr->hashtab.size();
+        ptr = 0;
+    }
+}
+
+
+static inline
+bool operator == (const SparseMatConstIterator& it1, const SparseMatConstIterator& it2)
+{
+    return it1.m == it2.m && it1.ptr == it2.ptr;
+}
+
+static inline
+bool operator != (const SparseMatConstIterator& it1, const SparseMatConstIterator& it2)
+{
+    return !(it1 == it2);
+}
+
+
+
+///////////////////////// SparseMatIterator /////////////////////////
+
+inline
+SparseMatIterator::SparseMatIterator()
+{}
+
+inline
+SparseMatIterator::SparseMatIterator(SparseMat* _m)
+    : SparseMatConstIterator(_m)
+{}
+
+inline
+SparseMatIterator::SparseMatIterator(const SparseMatIterator& it)
+    : SparseMatConstIterator(it)
+{}
+
+inline
+SparseMatIterator& SparseMatIterator::operator = (const SparseMatIterator& it)
+{
+    (SparseMatConstIterator&)*this = it;
+    return *this;
+}
+
+template<typename _Tp> inline
+_Tp& SparseMatIterator::value() const
+{
+    return *(_Tp*)ptr;
+}
+
+inline
+SparseMat::Node* SparseMatIterator::node() const
+{
+    return (SparseMat::Node*)SparseMatConstIterator::node();
+}
+
+inline
+SparseMatIterator& SparseMatIterator::operator ++()
+{
+    SparseMatConstIterator::operator ++();
+    return *this;
+}
+
+inline
+SparseMatIterator SparseMatIterator::operator ++(int)
+{
+    SparseMatIterator it = *this;
+    ++*this;
+    return it;
+}
+
+
+
+////////////////////// SparseMatConstIterator_ //////////////////////
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>::SparseMatConstIterator_()
+{}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>::SparseMatConstIterator_(const SparseMat_<_Tp>* _m)
+    : SparseMatConstIterator(_m)
+{}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>::SparseMatConstIterator_(const SparseMat* _m)
+    : SparseMatConstIterator(_m)
+{
+    CV_Assert( _m->type() == DataType<_Tp>::type );
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>::SparseMatConstIterator_(const SparseMatConstIterator_<_Tp>& it)
+    : SparseMatConstIterator(it)
+{}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>& SparseMatConstIterator_<_Tp>::operator = (const SparseMatConstIterator_<_Tp>& it)
+{
+    return reinterpret_cast<SparseMatConstIterator_<_Tp>&>
+         (*reinterpret_cast<SparseMatConstIterator*>(this) =
+           reinterpret_cast<const SparseMatConstIterator&>(it));
+}
+
+template<typename _Tp> inline
+const _Tp& SparseMatConstIterator_<_Tp>::operator *() const
+{
+    return *(const _Tp*)this->ptr;
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp>& SparseMatConstIterator_<_Tp>::operator ++()
+{
+    SparseMatConstIterator::operator ++();
+    return *this;
+}
+
+template<typename _Tp> inline
+SparseMatConstIterator_<_Tp> SparseMatConstIterator_<_Tp>::operator ++(int)
+{
+    SparseMatConstIterator it = *this;
+    SparseMatConstIterator::operator ++();
+    return it;
+}
+
+
+
+///////////////////////// SparseMatIterator_ ////////////////////////
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>::SparseMatIterator_()
+{}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>::SparseMatIterator_(SparseMat_<_Tp>* _m)
+    : SparseMatConstIterator_<_Tp>(_m)
+{}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>::SparseMatIterator_(SparseMat* _m)
+    : SparseMatConstIterator_<_Tp>(_m)
+{}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>::SparseMatIterator_(const SparseMatIterator_<_Tp>& it)
+    : SparseMatConstIterator_<_Tp>(it)
+{}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>& SparseMatIterator_<_Tp>::operator = (const SparseMatIterator_<_Tp>& it)
+{
+    return reinterpret_cast<SparseMatIterator_<_Tp>&>
+         (*reinterpret_cast<SparseMatConstIterator*>(this) =
+           reinterpret_cast<const SparseMatConstIterator&>(it));
+}
+
+template<typename _Tp> inline
+_Tp& SparseMatIterator_<_Tp>::operator *() const
+{
+    return *(_Tp*)this->ptr;
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp>& SparseMatIterator_<_Tp>::operator ++()
+{
+    SparseMatConstIterator::operator ++();
+    return *this;
+}
+
+template<typename _Tp> inline
+SparseMatIterator_<_Tp> SparseMatIterator_<_Tp>::operator ++(int)
+{
+    SparseMatIterator it = *this;
+    SparseMatConstIterator::operator ++();
+    return it;
+}
+
+
+
+//////////////////////// MatCommaInitializer_ ///////////////////////
+
+template<typename _Tp> inline
+MatCommaInitializer_<_Tp>::MatCommaInitializer_(Mat_<_Tp>* _m)
+    : it(_m)
+{}
+
+template<typename _Tp> template<typename T2> inline
+MatCommaInitializer_<_Tp>& MatCommaInitializer_<_Tp>::operator , (T2 v)
+{
+    CV_DbgAssert( this->it < ((const Mat_<_Tp>*)this->it.m)->end() );
+    *this->it = _Tp(v);
+    ++this->it;
+    return *this;
+}
+
+template<typename _Tp> inline
+MatCommaInitializer_<_Tp>::operator Mat_<_Tp>() const
+{
+    CV_DbgAssert( this->it == ((const Mat_<_Tp>*)this->it.m)->end() );
+    return Mat_<_Tp>(*this->it.m);
+}
+
+
+template<typename _Tp, typename T2> static inline
+MatCommaInitializer_<_Tp> operator << (const Mat_<_Tp>& m, T2 val)
+{
+    MatCommaInitializer_<_Tp> commaInitializer((Mat_<_Tp>*)&m);
+    return (commaInitializer, val);
+}
+
+
+
+///////////////////////// Matrix Expressions ////////////////////////
+
+inline
+Mat& Mat::operator = (const MatExpr& e)
+{
+    e.op->assign(e, *this);
+    return *this;
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>::Mat_(const MatExpr& e)
+{
+    e.op->assign(e, *this, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+Mat_<_Tp>& Mat_<_Tp>::operator = (const MatExpr& e)
+{
+    e.op->assign(e, *this, DataType<_Tp>::type);
+    return *this;
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::zeros(int rows, int cols)
+{
+    return Mat::zeros(rows, cols, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::zeros(Size sz)
+{
+    return Mat::zeros(sz, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::ones(int rows, int cols)
+{
+    return Mat::ones(rows, cols, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::ones(Size sz)
+{
+    return Mat::ones(sz, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::eye(int rows, int cols)
+{
+    return Mat::eye(rows, cols, DataType<_Tp>::type);
+}
+
+template<typename _Tp> inline
+MatExpr Mat_<_Tp>::eye(Size sz)
+{
+    return Mat::eye(sz, DataType<_Tp>::type);
+}
+
+inline
+MatExpr::MatExpr()
+    : op(0), flags(0), a(Mat()), b(Mat()), c(Mat()), alpha(0), beta(0), s()
+{}
+
+inline
+MatExpr::MatExpr(const MatOp* _op, int _flags, const Mat& _a, const Mat& _b,
+                 const Mat& _c, double _alpha, double _beta, const Scalar& _s)
+    : op(_op), flags(_flags), a(_a), b(_b), c(_c), alpha(_alpha), beta(_beta), s(_s)
+{}
+
+inline
+MatExpr::operator Mat() const
+{
+    Mat m;
+    op->assign(*this, m);
+    return m;
+}
+
+template<typename _Tp> inline
+MatExpr::operator Mat_<_Tp>() const
+{
+    Mat_<_Tp> m;
+    op->assign(*this, m, DataType<_Tp>::type);
+    return m;
+}
+
+
+template<typename _Tp> static inline
+MatExpr min(const Mat_<_Tp>& a, const Mat_<_Tp>& b)
+{
+    return cv::min((const Mat&)a, (const Mat&)b);
+}
+
+template<typename _Tp> static inline
+MatExpr min(const Mat_<_Tp>& a, double s)
+{
+    return cv::min((const Mat&)a, s);
+}
+
+template<typename _Tp> static inline
+MatExpr min(double s, const Mat_<_Tp>& a)
+{
+    return cv::min((const Mat&)a, s);
+}
+
+template<typename _Tp> static inline
+MatExpr max(const Mat_<_Tp>& a, const Mat_<_Tp>& b)
+{
+    return cv::max((const Mat&)a, (const Mat&)b);
+}
+
+template<typename _Tp> static inline
+MatExpr max(const Mat_<_Tp>& a, double s)
+{
+    return cv::max((const Mat&)a, s);
+}
+
+template<typename _Tp> static inline
+MatExpr max(double s, const Mat_<_Tp>& a)
+{
+    return cv::max((const Mat&)a, s);
+}
+
+template<typename _Tp> static inline
+MatExpr abs(const Mat_<_Tp>& m)
+{
+    return cv::abs((const Mat&)m);
+}
+
+
+static inline
+Mat& operator += (Mat& a, const MatExpr& b)
+{
+    b.op->augAssignAdd(b, a);
+    return a;
+}
+
+static inline
+const Mat& operator += (const Mat& a, const MatExpr& b)
+{
+    b.op->augAssignAdd(b, (Mat&)a);
+    return a;
+}
+
+template<typename _Tp> static inline
+Mat_<_Tp>& operator += (Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignAdd(b, a);
+    return a;
+}
+
+template<typename _Tp> static inline
+const Mat_<_Tp>& operator += (const Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignAdd(b, (Mat&)a);
+    return a;
+}
+
+static inline
+Mat& operator -= (Mat& a, const MatExpr& b)
+{
+    b.op->augAssignSubtract(b, a);
+    return a;
+}
+
+static inline
+const Mat& operator -= (const Mat& a, const MatExpr& b)
+{
+    b.op->augAssignSubtract(b, (Mat&)a);
+    return a;
+}
+
+template<typename _Tp> static inline
+Mat_<_Tp>& operator -= (Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignSubtract(b, a);
+    return a;
+}
+
+template<typename _Tp> static inline
+const Mat_<_Tp>& operator -= (const Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignSubtract(b, (Mat&)a);
+    return a;
+}
+
+static inline
+Mat& operator *= (Mat& a, const MatExpr& b)
+{
+    b.op->augAssignMultiply(b, a);
+    return a;
+}
+
+static inline
+const Mat& operator *= (const Mat& a, const MatExpr& b)
+{
+    b.op->augAssignMultiply(b, (Mat&)a);
+    return a;
+}
+
+template<typename _Tp> static inline
+Mat_<_Tp>& operator *= (Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignMultiply(b, a);
+    return a;
+}
+
+template<typename _Tp> static inline
+const Mat_<_Tp>& operator *= (const Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignMultiply(b, (Mat&)a);
+    return a;
+}
+
+static inline
+Mat& operator /= (Mat& a, const MatExpr& b)
+{
+    b.op->augAssignDivide(b, a);
+    return a;
+}
+
+static inline
+const Mat& operator /= (const Mat& a, const MatExpr& b)
+{
+    b.op->augAssignDivide(b, (Mat&)a);
+    return a;
+}
+
+template<typename _Tp> static inline
+Mat_<_Tp>& operator /= (Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignDivide(b, a);
+    return a;
+}
+
+template<typename _Tp> static inline
+const Mat_<_Tp>& operator /= (const Mat_<_Tp>& a, const MatExpr& b)
+{
+    b.op->augAssignDivide(b, (Mat&)a);
+    return a;
+}
+
+} //cv
+
+#endif
diff --git a/modules/core/include/opencv2/core/matx.hpp b/modules/core/include/opencv2/core/matx.hpp
new file mode 100644
index 0000000000..86a35cd756
--- /dev/null
+++ b/modules/core/include/opencv2/core/matx.hpp
@@ -0,0 +1,1358 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_MATX_HPP__
+#define __OPENCV_CORE_MATX_HPP__
+
+#ifndef __cplusplus
+#  error matx.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/base.hpp"
+#include "opencv2/core/traits.hpp"
+
+namespace cv
+{
+
+////////////////////////////// Small Matrix ///////////////////////////
+
+/*!
+ A short numerical vector.
+
+ This template class represents short numerical vectors (of 1, 2, 3, 4 ... elements)
+ on which you can perform basic arithmetical operations, access individual elements using [] operator etc.
+ The vectors are allocated on stack, as opposite to std::valarray, std::vector, cv::Mat etc.,
+ which elements are dynamically allocated in the heap.
+
+ The template takes 2 parameters:
+ -# _Tp element type
+ -# cn the number of elements
+
+ In addition to the universal notation like Vec<float, 3>, you can use shorter aliases
+ for the most popular specialized variants of Vec, e.g. Vec3f ~ Vec<float, 3>.
+ */
+
+struct CV_EXPORTS Matx_AddOp {};
+struct CV_EXPORTS Matx_SubOp {};
+struct CV_EXPORTS Matx_ScaleOp {};
+struct CV_EXPORTS Matx_MulOp {};
+struct CV_EXPORTS Matx_DivOp {};
+struct CV_EXPORTS Matx_MatMulOp {};
+struct CV_EXPORTS Matx_TOp {};
+
+template<typename _Tp, int m, int n> class Matx
+{
+public:
+    enum { depth    = DataType<_Tp>::depth,
+           rows     = m,
+           cols     = n,
+           channels = rows*cols,
+           type     = CV_MAKETYPE(depth, channels),
+           shortdim = (m < n ? m : n)
+         };
+
+    typedef _Tp                           value_type;
+    typedef Matx<_Tp, m, n>               mat_type;
+    typedef Matx<_Tp, shortdim, 1> diag_type;
+
+    //! default constructor
+    Matx();
+
+    Matx(_Tp v0); //!< 1x1 matrix
+    Matx(_Tp v0, _Tp v1); //!< 1x2 or 2x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2); //!< 1x3 or 3x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3); //!< 1x4, 2x2 or 4x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4); //!< 1x5 or 5x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5); //!< 1x6, 2x3, 3x2 or 6x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6); //!< 1x7 or 7x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7); //!< 1x8, 2x4, 4x2 or 8x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8); //!< 1x9, 3x3 or 9x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9); //!< 1x10, 2x5 or 5x2 or 10x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
+         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
+         _Tp v8, _Tp v9, _Tp v10, _Tp v11); //!< 1x12, 2x6, 3x4, 4x3, 6x2 or 12x1 matrix
+    Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
+         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
+         _Tp v8, _Tp v9, _Tp v10, _Tp v11,
+         _Tp v12, _Tp v13, _Tp v14, _Tp v15); //!< 1x16, 4x4 or 16x1 matrix
+    explicit Matx(const _Tp* vals); //!< initialize from a plain array
+
+    static Matx all(_Tp alpha);
+    static Matx zeros();
+    static Matx ones();
+    static Matx eye();
+    static Matx diag(const diag_type& d);
+    static Matx randu(_Tp a, _Tp b);
+    static Matx randn(_Tp a, _Tp b);
+
+    //! dot product computed with the default precision
+    _Tp dot(const Matx<_Tp, m, n>& v) const;
+
+    //! dot product computed in double-precision arithmetics
+    double ddot(const Matx<_Tp, m, n>& v) const;
+
+    //! convertion to another data type
+    template<typename T2> operator Matx<T2, m, n>() const;
+
+    //! change the matrix shape
+    template<int m1, int n1> Matx<_Tp, m1, n1> reshape() const;
+
+    //! extract part of the matrix
+    template<int m1, int n1> Matx<_Tp, m1, n1> get_minor(int i, int j) const;
+
+    //! extract the matrix row
+    Matx<_Tp, 1, n> row(int i) const;
+
+    //! extract the matrix column
+    Matx<_Tp, m, 1> col(int i) const;
+
+    //! extract the matrix diagonal
+    diag_type diag() const;
+
+    //! transpose the matrix
+    Matx<_Tp, n, m> t() const;
+
+    //! invert matrix the matrix
+    Matx<_Tp, n, m> inv(int method=DECOMP_LU) const;
+
+    //! solve linear system
+    template<int l> Matx<_Tp, n, l> solve(const Matx<_Tp, m, l>& rhs, int flags=DECOMP_LU) const;
+    Vec<_Tp, n> solve(const Vec<_Tp, m>& rhs, int method) const;
+
+    //! multiply two matrices element-wise
+    Matx<_Tp, m, n> mul(const Matx<_Tp, m, n>& a) const;
+
+    //! divide two matrices element-wise
+    Matx<_Tp, m, n> div(const Matx<_Tp, m, n>& a) const;
+
+    //! element access
+    const _Tp& operator ()(int i, int j) const;
+    _Tp& operator ()(int i, int j);
+
+    //! 1D element access
+    const _Tp& operator ()(int i) const;
+    _Tp& operator ()(int i);
+
+    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_AddOp);
+    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_SubOp);
+    template<typename _T2> Matx(const Matx<_Tp, m, n>& a, _T2 alpha, Matx_ScaleOp);
+    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_MulOp);
+    Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_DivOp);
+    template<int l> Matx(const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b, Matx_MatMulOp);
+    Matx(const Matx<_Tp, n, m>& a, Matx_TOp);
+
+    _Tp val[m*n]; //< matrix elements
+};
+
+/*!
+  \typedef
+*/
+typedef Matx<float, 1, 2> Matx12f;
+typedef Matx<double, 1, 2> Matx12d;
+typedef Matx<float, 1, 3> Matx13f;
+typedef Matx<double, 1, 3> Matx13d;
+typedef Matx<float, 1, 4> Matx14f;
+typedef Matx<double, 1, 4> Matx14d;
+typedef Matx<float, 1, 6> Matx16f;
+typedef Matx<double, 1, 6> Matx16d;
+
+typedef Matx<float, 2, 1> Matx21f;
+typedef Matx<double, 2, 1> Matx21d;
+typedef Matx<float, 3, 1> Matx31f;
+typedef Matx<double, 3, 1> Matx31d;
+typedef Matx<float, 4, 1> Matx41f;
+typedef Matx<double, 4, 1> Matx41d;
+typedef Matx<float, 6, 1> Matx61f;
+typedef Matx<double, 6, 1> Matx61d;
+
+typedef Matx<float, 2, 2> Matx22f;
+typedef Matx<double, 2, 2> Matx22d;
+typedef Matx<float, 2, 3> Matx23f;
+typedef Matx<double, 2, 3> Matx23d;
+typedef Matx<float, 3, 2> Matx32f;
+typedef Matx<double, 3, 2> Matx32d;
+
+typedef Matx<float, 3, 3> Matx33f;
+typedef Matx<double, 3, 3> Matx33d;
+
+typedef Matx<float, 3, 4> Matx34f;
+typedef Matx<double, 3, 4> Matx34d;
+typedef Matx<float, 4, 3> Matx43f;
+typedef Matx<double, 4, 3> Matx43d;
+
+typedef Matx<float, 4, 4> Matx44f;
+typedef Matx<double, 4, 4> Matx44d;
+typedef Matx<float, 6, 6> Matx66f;
+typedef Matx<double, 6, 6> Matx66d;
+
+/*!
+  traits
+*/
+template<typename _Tp, int m, int n> class DataType< Matx<_Tp, m, n> >
+{
+public:
+    typedef Matx<_Tp, m, n>                               value_type;
+    typedef Matx<typename DataType<_Tp>::work_type, m, n> work_type;
+    typedef _Tp                                           channel_type;
+    typedef value_type                                    vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = m * n,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+/*!
+ Comma-separated Matrix Initializer
+*/
+template<typename _Tp, int m, int n> class MatxCommaInitializer
+{
+public:
+    MatxCommaInitializer(Matx<_Tp, m, n>* _mtx);
+    template<typename T2> MatxCommaInitializer<_Tp, m, n>& operator , (T2 val);
+    Matx<_Tp, m, n> operator *() const;
+
+    Matx<_Tp, m, n>* dst;
+    int idx;
+};
+
+/*!
+ Utility methods
+*/
+template<typename _Tp, int m> static double determinant(const Matx<_Tp, m, m>& a);
+template<typename _Tp, int m, int n> static double trace(const Matx<_Tp, m, n>& a);
+template<typename _Tp, int m, int n> static double norm(const Matx<_Tp, m, n>& M);
+template<typename _Tp, int m, int n> static double norm(const Matx<_Tp, m, n>& M, int normType);
+
+
+
+/////////////////////// Vec (used as element of multi-channel images /////////////////////
+
+/*!
+  A short numerical vector.
+
+  This template class represents short numerical vectors (of 1, 2, 3, 4 ... elements)
+  on which you can perform basic arithmetical operations, access individual elements using [] operator etc.
+  The vectors are allocated on stack, as opposite to std::valarray, std::vector, cv::Mat etc.,
+  which elements are dynamically allocated in the heap.
+
+  The template takes 2 parameters:
+  -# _Tp element type
+  -# cn the number of elements
+
+  In addition to the universal notation like Vec<float, 3>, you can use shorter aliases
+  for the most popular specialized variants of Vec, e.g. Vec3f ~ Vec<float, 3>.
+*/
+template<typename _Tp, int cn> class Vec : public Matx<_Tp, cn, 1>
+{
+public:
+    typedef _Tp value_type;
+    enum { depth    = Matx<_Tp, cn, 1>::depth,
+           channels = cn,
+           type     = CV_MAKETYPE(depth, channels)
+         };
+
+    //! default constructor
+    Vec();
+
+    Vec(_Tp v0); //!< 1-element vector constructor
+    Vec(_Tp v0, _Tp v1); //!< 2-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2); //!< 3-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3); //!< 4-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4); //!< 5-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5); //!< 6-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6); //!< 7-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7); //!< 8-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8); //!< 9-element vector constructor
+    Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9); //!< 10-element vector constructor
+    explicit Vec(const _Tp* values);
+
+    Vec(const Vec<_Tp, cn>& v);
+
+    static Vec all(_Tp alpha);
+
+    //! per-element multiplication
+    Vec mul(const Vec<_Tp, cn>& v) const;
+
+    //! conjugation (makes sense for complex numbers and quaternions)
+    Vec conj() const;
+
+    /*!
+      cross product of the two 3D vectors.
+
+      For other dimensionalities the exception is raised
+    */
+    Vec cross(const Vec& v) const;
+    //! convertion to another data type
+    template<typename T2> operator Vec<T2, cn>() const;
+
+    /*! element access */
+    const _Tp& operator [](int i) const;
+    _Tp& operator[](int i);
+    const _Tp& operator ()(int i) const;
+    _Tp& operator ()(int i);
+
+    Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_AddOp);
+    Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_SubOp);
+    template<typename _T2> Vec(const Matx<_Tp, cn, 1>& a, _T2 alpha, Matx_ScaleOp);
+};
+
+/* \typedef
+   Shorter aliases for the most popular specializations of Vec<T,n>
+*/
+typedef Vec<uchar, 2> Vec2b;
+typedef Vec<uchar, 3> Vec3b;
+typedef Vec<uchar, 4> Vec4b;
+
+typedef Vec<short, 2> Vec2s;
+typedef Vec<short, 3> Vec3s;
+typedef Vec<short, 4> Vec4s;
+
+typedef Vec<ushort, 2> Vec2w;
+typedef Vec<ushort, 3> Vec3w;
+typedef Vec<ushort, 4> Vec4w;
+
+typedef Vec<int, 2> Vec2i;
+typedef Vec<int, 3> Vec3i;
+typedef Vec<int, 4> Vec4i;
+typedef Vec<int, 6> Vec6i;
+typedef Vec<int, 8> Vec8i;
+
+typedef Vec<float, 2> Vec2f;
+typedef Vec<float, 3> Vec3f;
+typedef Vec<float, 4> Vec4f;
+typedef Vec<float, 6> Vec6f;
+
+typedef Vec<double, 2> Vec2d;
+typedef Vec<double, 3> Vec3d;
+typedef Vec<double, 4> Vec4d;
+typedef Vec<double, 6> Vec6d;
+
+/*!
+  traits
+*/
+template<typename _Tp, int cn> class DataType< Vec<_Tp, cn> >
+{
+public:
+    typedef Vec<_Tp, cn>                               value_type;
+    typedef Vec<typename DataType<_Tp>::work_type, cn> work_type;
+    typedef _Tp                                        channel_type;
+    typedef value_type                                 vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = cn,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+/*!
+ Comma-separated Vec Initializer
+*/
+template<typename _Tp, int m> class VecCommaInitializer : public MatxCommaInitializer<_Tp, m, 1>
+{
+public:
+    VecCommaInitializer(Vec<_Tp, m>* _vec);
+    template<typename T2> VecCommaInitializer<_Tp, m>& operator , (T2 val);
+    Vec<_Tp, m> operator *() const;
+};
+
+/*!
+ Utility methods
+*/
+template<typename _Tp, int cn> static Vec<_Tp, cn> normalize(const Vec<_Tp, cn>& v);
+
+
+
+///////////////////////////////////// helper classes /////////////////////////////////////
+namespace internal
+{
+
+template<typename _Tp, int m> struct Matx_DetOp
+{
+    double operator ()(const Matx<_Tp, m, m>& a) const
+    {
+        Matx<_Tp, m, m> temp = a;
+        double p = LU(temp.val, m*sizeof(_Tp), m, 0, 0, 0);
+        if( p == 0 )
+            return p;
+        for( int i = 0; i < m; i++ )
+            p *= temp(i, i);
+        return 1./p;
+    }
+};
+
+template<typename _Tp> struct Matx_DetOp<_Tp, 1>
+{
+    double operator ()(const Matx<_Tp, 1, 1>& a) const
+    {
+        return a(0,0);
+    }
+};
+
+template<typename _Tp> struct Matx_DetOp<_Tp, 2>
+{
+    double operator ()(const Matx<_Tp, 2, 2>& a) const
+    {
+        return a(0,0)*a(1,1) - a(0,1)*a(1,0);
+    }
+};
+
+template<typename _Tp> struct Matx_DetOp<_Tp, 3>
+{
+    double operator ()(const Matx<_Tp, 3, 3>& a) const
+    {
+        return a(0,0)*(a(1,1)*a(2,2) - a(2,1)*a(1,2)) -
+            a(0,1)*(a(1,0)*a(2,2) - a(2,0)*a(1,2)) +
+            a(0,2)*(a(1,0)*a(2,1) - a(2,0)*a(1,1));
+    }
+};
+
+template<typename _Tp> Vec<_Tp, 2> inline conjugate(const Vec<_Tp, 2>& v)
+{
+    return Vec<_Tp, 2>(v[0], -v[1]);
+}
+
+template<typename _Tp> Vec<_Tp, 4> inline conjugate(const Vec<_Tp, 4>& v)
+{
+    return Vec<_Tp, 4>(v[0], -v[1], -v[2], -v[3]);
+}
+
+} // internal
+
+
+
+////////////////////////////////// Matx Implementation ///////////////////////////////////
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx()
+{
+    for(int i = 0; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0)
+{
+    val[0] = v0;
+    for(int i = 1; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1)
+{
+    CV_StaticAssert(channels >= 2, "Matx should have at least 2 elaments.");
+    val[0] = v0; val[1] = v1;
+    for(int i = 2; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2)
+{
+    CV_StaticAssert(channels >= 3, "Matx should have at least 3 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2;
+    for(int i = 3; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
+{
+    CV_StaticAssert(channels >= 4, "Matx should have at least 4 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    for(int i = 4; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4)
+{
+    CV_StaticAssert(channels >= 5, "Matx should have at least 5 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3; val[4] = v4;
+    for(int i = 5; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5)
+{
+    CV_StaticAssert(channels >= 6, "Matx should have at least 6 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5;
+    for(int i = 6; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6)
+{
+    CV_StaticAssert(channels >= 7, "Matx should have at least 7 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6;
+    for(int i = 7; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7)
+{
+    CV_StaticAssert(channels >= 8, "Matx should have at least 8 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
+    for(int i = 8; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8)
+{
+    CV_StaticAssert(channels >= 9, "Matx should have at least 9 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
+    val[8] = v8;
+    for(int i = 9; i < channels; i++) val[i] = _Tp(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9)
+{
+    CV_StaticAssert(channels >= 10, "Matx should have at least 10 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
+    val[8] = v8; val[9] = v9;
+    for(int i = 10; i < channels; i++) val[i] = _Tp(0);
+}
+
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9, _Tp v10, _Tp v11)
+{
+    CV_StaticAssert(channels == 12, "Matx should have at least 12 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
+    val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9, _Tp v10, _Tp v11, _Tp v12, _Tp v13, _Tp v14, _Tp v15)
+{
+    CV_StaticAssert(channels == 16, "Matx should have at least 16 elaments.");
+    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
+    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
+    val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;
+    val[12] = v12; val[13] = v13; val[14] = v14; val[15] = v15;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n>::Matx(const _Tp* values)
+{
+    for( int i = 0; i < channels; i++ ) val[i] = values[i];
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n> Matx<_Tp, m, n>::all(_Tp alpha)
+{
+    Matx<_Tp, m, n> M;
+    for( int i = 0; i < m*n; i++ ) M.val[i] = alpha;
+    return M;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::zeros()
+{
+    return all(0);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::ones()
+{
+    return all(1);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::eye()
+{
+    Matx<_Tp,m,n> M;
+    for(int i = 0; i < shortdim; i++)
+        M(i,i) = 1;
+    return M;
+}
+
+template<typename _Tp, int m, int n> inline
+_Tp Matx<_Tp, m, n>::dot(const Matx<_Tp, m, n>& M) const
+{
+    _Tp s = 0;
+    for( int i = 0; i < channels; i++ ) s += val[i]*M.val[i];
+    return s;
+}
+
+template<typename _Tp, int m, int n> inline
+double Matx<_Tp, m, n>::ddot(const Matx<_Tp, m, n>& M) const
+{
+    double s = 0;
+    for( int i = 0; i < channels; i++ ) s += (double)val[i]*M.val[i];
+    return s;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::diag(const typename Matx<_Tp,m,n>::diag_type& d)
+{
+    Matx<_Tp,m,n> M;
+    for(int i = 0; i < shortdim; i++)
+        M(i,i) = d(i, 0);
+    return M;
+}
+
+template<typename _Tp, int m, int n> template<typename T2>
+inline Matx<_Tp, m, n>::operator Matx<T2, m, n>() const
+{
+    Matx<T2, m, n> M;
+    for( int i = 0; i < m*n; i++ ) M.val[i] = saturate_cast<T2>(val[i]);
+    return M;
+}
+
+template<typename _Tp, int m, int n> template<int m1, int n1> inline
+Matx<_Tp, m1, n1> Matx<_Tp, m, n>::reshape() const
+{
+    CV_StaticAssert(m1*n1 == m*n, "Input and destnarion matrices must have the same number of elements");
+    return (const Matx<_Tp, m1, n1>&)*this;
+}
+
+template<typename _Tp, int m, int n>
+template<int m1, int n1> inline
+Matx<_Tp, m1, n1> Matx<_Tp, m, n>::get_minor(int i, int j) const
+{
+    CV_DbgAssert(0 <= i && i+m1 <= m && 0 <= j && j+n1 <= n);
+    Matx<_Tp, m1, n1> s;
+    for( int di = 0; di < m1; di++ )
+        for( int dj = 0; dj < n1; dj++ )
+            s(di, dj) = (*this)(i+di, j+dj);
+    return s;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, 1, n> Matx<_Tp, m, n>::row(int i) const
+{
+    CV_DbgAssert((unsigned)i < (unsigned)m);
+    return Matx<_Tp, 1, n>(&val[i*n]);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, 1> Matx<_Tp, m, n>::col(int j) const
+{
+    CV_DbgAssert((unsigned)j < (unsigned)n);
+    Matx<_Tp, m, 1> v;
+    for( int i = 0; i < m; i++ )
+        v.val[i] = val[i*n + j];
+    return v;
+}
+
+template<typename _Tp, int m, int n> inline
+typename Matx<_Tp, m, n>::diag_type Matx<_Tp, m, n>::diag() const
+{
+    diag_type d;
+    for( int i = 0; i < shortdim; i++ )
+        d.val[i] = val[i*n + i];
+    return d;
+}
+
+template<typename _Tp, int m, int n> inline
+const _Tp& Matx<_Tp, m, n>::operator()(int i, int j) const
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)m && (unsigned)j < (unsigned)n );
+    return this->val[i*n + j];
+}
+
+template<typename _Tp, int m, int n> inline
+_Tp& Matx<_Tp, m, n>::operator ()(int i, int j)
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)m && (unsigned)j < (unsigned)n );
+    return val[i*n + j];
+}
+
+template<typename _Tp, int m, int n> inline
+const _Tp& Matx<_Tp, m, n>::operator ()(int i) const
+{
+    CV_StaticAssert(m == 1 || n == 1, "Single index indexation requires matrix to be a column or a row");
+    CV_DbgAssert( (unsigned)i < (unsigned)(m+n-1) );
+    return val[i];
+}
+
+template<typename _Tp, int m, int n> inline
+_Tp& Matx<_Tp, m, n>::operator ()(int i)
+{
+    CV_StaticAssert(m == 1 || n == 1, "Single index indexation requires matrix to be a column or a row");
+    CV_DbgAssert( (unsigned)i < (unsigned)(m+n-1) );
+    return val[i];
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_AddOp)
+{
+    for( int i = 0; i < channels; i++ )
+        val[i] = saturate_cast<_Tp>(a.val[i] + b.val[i]);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_SubOp)
+{
+    for( int i = 0; i < channels; i++ )
+        val[i] = saturate_cast<_Tp>(a.val[i] - b.val[i]);
+}
+
+template<typename _Tp, int m, int n> template<typename _T2> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, _T2 alpha, Matx_ScaleOp)
+{
+    for( int i = 0; i < channels; i++ )
+        val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_MulOp)
+{
+    for( int i = 0; i < channels; i++ )
+        val[i] = saturate_cast<_Tp>(a.val[i] * b.val[i]);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_DivOp)
+{
+    for( int i = 0; i < channels; i++ )
+        val[i] = saturate_cast<_Tp>(a.val[i] / b.val[i]);
+}
+
+template<typename _Tp, int m, int n> template<int l> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b, Matx_MatMulOp)
+{
+    for( int i = 0; i < m; i++ )
+        for( int j = 0; j < n; j++ )
+        {
+            _Tp s = 0;
+            for( int k = 0; k < l; k++ )
+                s += a(i, k) * b(k, j);
+            val[i*n + j] = s;
+        }
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n>::Matx(const Matx<_Tp, n, m>& a, Matx_TOp)
+{
+    for( int i = 0; i < m; i++ )
+        for( int j = 0; j < n; j++ )
+            val[i*n + j] = a(j, i);
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n> Matx<_Tp, m, n>::mul(const Matx<_Tp, m, n>& a) const
+{
+    return Matx<_Tp, m, n>(*this, a, Matx_MulOp());
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n> Matx<_Tp, m, n>::div(const Matx<_Tp, m, n>& a) const
+{
+    return Matx<_Tp, m, n>(*this, a, Matx_DivOp());
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, n, m> Matx<_Tp, m, n>::t() const
+{
+    return Matx<_Tp, n, m>(*this, Matx_TOp());
+}
+
+template<typename _Tp, int m, int n> inline
+Vec<_Tp, n> Matx<_Tp, m, n>::solve(const Vec<_Tp, m>& rhs, int method) const
+{
+    Matx<_Tp, n, 1> x = solve((const Matx<_Tp, m, 1>&)(rhs), method);
+    return (Vec<_Tp, n>&)(x);
+}
+
+template<typename _Tp, int m> static inline
+double determinant(const Matx<_Tp, m, m>& a)
+{
+    return internal::Matx_DetOp<_Tp, m>()(a);
+}
+
+template<typename _Tp, int m, int n> static inline
+double trace(const Matx<_Tp, m, n>& a)
+{
+    _Tp s = 0;
+    for( int i = 0; i < std::min(m, n); i++ )
+        s += a(i,i);
+    return s;
+}
+
+template<typename _Tp, int m, int n> static inline
+double norm(const Matx<_Tp, m, n>& M)
+{
+    return std::sqrt(normL2Sqr<_Tp, double>(M.val, m*n));
+}
+
+template<typename _Tp, int m, int n> static inline
+double norm(const Matx<_Tp, m, n>& M, int normType)
+{
+    return normType == NORM_INF ? (double)normInf<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n) :
+        normType == NORM_L1 ? (double)normL1<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n) :
+        std::sqrt((double)normL2Sqr<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n));
+}
+
+
+
+//////////////////////////////// matx comma initializer //////////////////////////////////
+
+template<typename _Tp, typename _T2, int m, int n> static inline
+MatxCommaInitializer<_Tp, m, n> operator << (const Matx<_Tp, m, n>& mtx, _T2 val)
+{
+    MatxCommaInitializer<_Tp, m, n> commaInitializer((Matx<_Tp, m, n>*)&mtx);
+    return (commaInitializer, val);
+}
+
+template<typename _Tp, int m, int n> inline
+MatxCommaInitializer<_Tp, m, n>::MatxCommaInitializer(Matx<_Tp, m, n>* _mtx)
+    : dst(_mtx), idx(0)
+{}
+
+template<typename _Tp, int m, int n> template<typename _T2> inline
+MatxCommaInitializer<_Tp, m, n>& MatxCommaInitializer<_Tp, m, n>::operator , (_T2 value)
+{
+    CV_DbgAssert( idx < m*n );
+    dst->val[idx++] = saturate_cast<_Tp>(value);
+    return *this;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, m, n> MatxCommaInitializer<_Tp, m, n>::operator *() const
+{
+    CV_DbgAssert( idx == n*m );
+    return *dst;
+}
+
+
+
+/////////////////////////////////// Vec Implementation ///////////////////////////////////
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec() {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0)
+    : Matx<_Tp, cn, 1>(v0) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1)
+    : Matx<_Tp, cn, 1>(v0, v1) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2)
+    : Matx<_Tp, cn, 1>(v0, v1, v2) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7, v8) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5, _Tp v6, _Tp v7, _Tp v8, _Tp v9)
+    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(const _Tp* values)
+    : Matx<_Tp, cn, 1>(values) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(const Vec<_Tp, cn>& m)
+    : Matx<_Tp, cn, 1>(m.val) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_AddOp op)
+    : Matx<_Tp, cn, 1>(a, b, op) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_SubOp op)
+    : Matx<_Tp, cn, 1>(a, b, op) {}
+
+template<typename _Tp, int cn> template<typename _T2> inline
+Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, _T2 alpha, Matx_ScaleOp op)
+    : Matx<_Tp, cn, 1>(a, alpha, op) {}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn> Vec<_Tp, cn>::all(_Tp alpha)
+{
+    Vec v;
+    for( int i = 0; i < cn; i++ ) v.val[i] = alpha;
+    return v;
+}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn> Vec<_Tp, cn>::mul(const Vec<_Tp, cn>& v) const
+{
+    Vec<_Tp, cn> w;
+    for( int i = 0; i < cn; i++ ) w.val[i] = saturate_cast<_Tp>(this->val[i]*v.val[i]);
+    return w;
+}
+
+template<> inline
+Vec<float, 2> Vec<float, 2>::conj() const
+{
+    return internal::conjugate(*this);
+}
+
+template<> inline
+Vec<double, 2> Vec<double, 2>::conj() const
+{
+    return internal::conjugate(*this);
+}
+
+template<> inline
+Vec<float, 4> Vec<float, 4>::conj() const
+{
+    return internal::conjugate(*this);
+}
+
+template<> inline
+Vec<double, 4> Vec<double, 4>::conj() const
+{
+    return internal::conjugate(*this);
+}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn> Vec<_Tp, cn>::cross(const Vec<_Tp, cn>&) const
+{
+    CV_StaticAssert(cn == 3, "for arbitrary-size vector there is no cross-product defined");
+    return Vec<_Tp, cn>();
+}
+
+template<> inline
+Vec<float, 3> Vec<float, 3>::cross(const Vec<float, 3>& v) const
+{
+    return Vec<float,3>(val[1]*v.val[2] - val[2]*v.val[1],
+                     val[2]*v.val[0] - val[0]*v.val[2],
+                     val[0]*v.val[1] - val[1]*v.val[0]);
+}
+
+template<> inline
+Vec<double, 3> Vec<double, 3>::cross(const Vec<double, 3>& v) const
+{
+    return Vec<double,3>(val[1]*v.val[2] - val[2]*v.val[1],
+                     val[2]*v.val[0] - val[0]*v.val[2],
+                     val[0]*v.val[1] - val[1]*v.val[0]);
+}
+
+template<typename _Tp, int cn> template<typename T2> inline
+Vec<_Tp, cn>::operator Vec<T2, cn>() const
+{
+    Vec<T2, cn> v;
+    for( int i = 0; i < cn; i++ ) v.val[i] = saturate_cast<T2>(this->val[i]);
+    return v;
+}
+
+template<typename _Tp, int cn> inline
+const _Tp& Vec<_Tp, cn>::operator [](int i) const
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)cn );
+    return this->val[i];
+}
+
+template<typename _Tp, int cn> inline
+_Tp& Vec<_Tp, cn>::operator [](int i)
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)cn );
+    return this->val[i];
+}
+
+template<typename _Tp, int cn> inline
+const _Tp& Vec<_Tp, cn>::operator ()(int i) const
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)cn );
+    return this->val[i];
+}
+
+template<typename _Tp, int cn> inline
+_Tp& Vec<_Tp, cn>::operator ()(int i)
+{
+    CV_DbgAssert( (unsigned)i < (unsigned)cn );
+    return this->val[i];
+}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn> normalize(const Vec<_Tp, cn>& v)
+{
+    double nv = norm(v);
+    return v * (nv ? 1./nv : 0.);
+}
+
+
+
+//////////////////////////////// matx comma initializer //////////////////////////////////
+
+
+template<typename _Tp, typename _T2, int cn> static inline
+VecCommaInitializer<_Tp, cn> operator << (const Vec<_Tp, cn>& vec, _T2 val)
+{
+    VecCommaInitializer<_Tp, cn> commaInitializer((Vec<_Tp, cn>*)&vec);
+    return (commaInitializer, val);
+}
+
+template<typename _Tp, int cn> inline
+VecCommaInitializer<_Tp, cn>::VecCommaInitializer(Vec<_Tp, cn>* _vec)
+    : MatxCommaInitializer<_Tp, cn, 1>(_vec)
+{}
+
+template<typename _Tp, int cn> template<typename _T2> inline
+VecCommaInitializer<_Tp, cn>& VecCommaInitializer<_Tp, cn>::operator , (_T2 value)
+{
+    CV_DbgAssert( this->idx < cn );
+    this->dst->val[this->idx++] = saturate_cast<_Tp>(value);
+    return *this;
+}
+
+template<typename _Tp, int cn> inline
+Vec<_Tp, cn> VecCommaInitializer<_Tp, cn>::operator *() const
+{
+    CV_DbgAssert( this->idx == cn );
+    return *this->dst;
+}
+
+
+
+///////////////////////////// Matx out-of-class operators ////////////////////////////////
+
+template<typename _Tp1, typename _Tp2, int m, int n> static inline
+Matx<_Tp1, m, n>& operator += (Matx<_Tp1, m, n>& a, const Matx<_Tp2, m, n>& b)
+{
+    for( int i = 0; i < m*n; i++ )
+        a.val[i] = saturate_cast<_Tp1>(a.val[i] + b.val[i]);
+    return a;
+}
+
+template<typename _Tp1, typename _Tp2, int m, int n> static inline
+Matx<_Tp1, m, n>& operator -= (Matx<_Tp1, m, n>& a, const Matx<_Tp2, m, n>& b)
+{
+    for( int i = 0; i < m*n; i++ )
+        a.val[i] = saturate_cast<_Tp1>(a.val[i] - b.val[i]);
+    return a;
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator + (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
+{
+    return Matx<_Tp, m, n>(a, b, Matx_AddOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator - (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
+{
+    return Matx<_Tp, m, n>(a, b, Matx_SubOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, int alpha)
+{
+    for( int i = 0; i < m*n; i++ )
+        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
+    return a;
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, float alpha)
+{
+    for( int i = 0; i < m*n; i++ )
+        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
+    return a;
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, double alpha)
+{
+    for( int i = 0; i < m*n; i++ )
+        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
+    return a;
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, int alpha)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, float alpha)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, double alpha)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (int alpha, const Matx<_Tp, m, n>& a)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (float alpha, const Matx<_Tp, m, n>& a)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator * (double alpha, const Matx<_Tp, m, n>& a)
+{
+    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Matx<_Tp, m, n> operator - (const Matx<_Tp, m, n>& a)
+{
+    return Matx<_Tp, m, n>(a, -1, Matx_ScaleOp());
+}
+
+template<typename _Tp, int m, int n, int l> static inline
+Matx<_Tp, m, n> operator * (const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b)
+{
+    return Matx<_Tp, m, n>(a, b, Matx_MatMulOp());
+}
+
+template<typename _Tp, int m, int n> static inline
+Vec<_Tp, m> operator * (const Matx<_Tp, m, n>& a, const Vec<_Tp, n>& b)
+{
+    Matx<_Tp, m, 1> c(a, b, Matx_MatMulOp());
+    return (const Vec<_Tp, m>&)(c);
+}
+
+template<typename _Tp, int m, int n> static inline
+bool operator == (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
+{
+    for( int i = 0; i < m*n; i++ )
+        if( a.val[i] != b.val[i] ) return false;
+    return true;
+}
+
+template<typename _Tp, int m, int n> static inline
+bool operator != (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
+{
+    return !(a == b);
+}
+
+
+
+////////////////////////////// Vec out-of-class operators ////////////////////////////////
+
+template<typename _Tp1, typename _Tp2, int cn> static inline
+Vec<_Tp1, cn>& operator += (Vec<_Tp1, cn>& a, const Vec<_Tp2, cn>& b)
+{
+    for( int i = 0; i < cn; i++ )
+        a.val[i] = saturate_cast<_Tp1>(a.val[i] + b.val[i]);
+    return a;
+}
+
+template<typename _Tp1, typename _Tp2, int cn> static inline
+Vec<_Tp1, cn>& operator -= (Vec<_Tp1, cn>& a, const Vec<_Tp2, cn>& b)
+{
+    for( int i = 0; i < cn; i++ )
+        a.val[i] = saturate_cast<_Tp1>(a.val[i] - b.val[i]);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator + (const Vec<_Tp, cn>& a, const Vec<_Tp, cn>& b)
+{
+    return Vec<_Tp, cn>(a, b, Matx_AddOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator - (const Vec<_Tp, cn>& a, const Vec<_Tp, cn>& b)
+{
+    return Vec<_Tp, cn>(a, b, Matx_SubOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, int alpha)
+{
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*alpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, float alpha)
+{
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*alpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, double alpha)
+{
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*alpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, int alpha)
+{
+    double ialpha = 1./alpha;
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, float alpha)
+{
+    float ialpha = 1.f/alpha;
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, double alpha)
+{
+    double ialpha = 1./alpha;
+    for( int i = 0; i < cn; i++ )
+        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
+    return a;
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (const Vec<_Tp, cn>& a, int alpha)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (int alpha, const Vec<_Tp, cn>& a)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (const Vec<_Tp, cn>& a, float alpha)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (float alpha, const Vec<_Tp, cn>& a)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (const Vec<_Tp, cn>& a, double alpha)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator * (double alpha, const Vec<_Tp, cn>& a)
+{
+    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator / (const Vec<_Tp, cn>& a, int alpha)
+{
+    return Vec<_Tp, cn>(a, 1./alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator / (const Vec<_Tp, cn>& a, float alpha)
+{
+    return Vec<_Tp, cn>(a, 1.f/alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator / (const Vec<_Tp, cn>& a, double alpha)
+{
+    return Vec<_Tp, cn>(a, 1./alpha, Matx_ScaleOp());
+}
+
+template<typename _Tp, int cn> static inline
+Vec<_Tp, cn> operator - (const Vec<_Tp, cn>& a)
+{
+    Vec<_Tp,cn> t;
+    for( int i = 0; i < cn; i++ ) t.val[i] = saturate_cast<_Tp>(-a.val[i]);
+    return t;
+}
+
+template<typename _Tp> inline Vec<_Tp, 4> operator * (const Vec<_Tp, 4>& v1, const Vec<_Tp, 4>& v2)
+{
+    return Vec<_Tp, 4>(saturate_cast<_Tp>(v1[0]*v2[0] - v1[1]*v2[1] - v1[2]*v2[2] - v1[3]*v2[3]),
+                       saturate_cast<_Tp>(v1[0]*v2[1] + v1[1]*v2[0] + v1[2]*v2[3] - v1[3]*v2[2]),
+                       saturate_cast<_Tp>(v1[0]*v2[2] - v1[1]*v2[3] + v1[2]*v2[0] + v1[3]*v2[1]),
+                       saturate_cast<_Tp>(v1[0]*v2[3] + v1[1]*v2[2] - v1[2]*v2[1] + v1[3]*v2[0]));
+}
+
+template<typename _Tp> inline Vec<_Tp, 4>& operator *= (Vec<_Tp, 4>& v1, const Vec<_Tp, 4>& v2)
+{
+    v1 = v1 * v2;
+    return v1;
+}
+
+} // cv
+
+#endif // __OPENCV_CORE_MATX_HPP__
diff --git a/modules/core/include/opencv2/core/opengl.hpp b/modules/core/include/opencv2/core/opengl.hpp
index 1643d39e0c..0a4b454c8a 100644
--- a/modules/core/include/opencv2/core/opengl.hpp
+++ b/modules/core/include/opencv2/core/opengl.hpp
@@ -22,7 +22,7 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
@@ -40,10 +40,12 @@
 //
 //M*/
 
-#ifndef __OPENCV_OPENGL_INTEROP_HPP__
-#define __OPENCV_OPENGL_INTEROP_HPP__
+#ifndef __OPENCV_CORE_OPENGL_HPP__
+#define __OPENCV_CORE_OPENGL_HPP__
 
-#ifdef __cplusplus
+#ifndef __cplusplus
+#  error opengl.hpp header must be compiled as C++
+#endif
 
 #include "opencv2/core.hpp"
 
@@ -86,7 +88,7 @@ public:
 
     //! create buffer
     void create(int arows, int acols, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false);
-    void create(Size asize, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false) { create(asize.height, asize.width, atype, target, autoRelease); }
+    void create(Size asize, int atype, Target target = ARRAY_BUFFER, bool autoRelease = false);
 
     //! release memory and delete buffer object
     void release();
@@ -94,11 +96,15 @@ public:
     //! set auto release mode (if true, release will be called in object's destructor)
     void setAutoRelease(bool flag);
 
-    //! copy from host/device memory
+    //! copy from host/device memory (blocking)
     void copyFrom(InputArray arr, Target target = ARRAY_BUFFER, bool autoRelease = false);
+    //! copy from device memory (non blocking)
+    void copyFrom(InputArray arr, gpu::Stream& stream, Target target = ARRAY_BUFFER, bool autoRelease = false);
 
-    //! copy to host/device memory
-    void copyTo(OutputArray arr, Target target = ARRAY_BUFFER, bool autoRelease = false) const;
+    //! copy to host/device memory (blocking)
+    void copyTo(OutputArray arr) const;
+    //! copy to device memory (non blocking)
+    void copyTo(OutputArray arr, gpu::Stream& stream) const;
 
     //! create copy of current buffer
     Buffer clone(Target target = ARRAY_BUFFER, bool autoRelease = false) const;
@@ -113,21 +119,26 @@ public:
     Mat mapHost(Access access);
     void unmapHost();
 
-    //! map to device memory
+    //! map to device memory (blocking)
     gpu::GpuMat mapDevice();
     void unmapDevice();
 
-    int rows() const { return rows_; }
-    int cols() const { return cols_; }
-    Size size() const { return Size(cols_, rows_); }
-    bool empty() const { return rows_ == 0 || cols_ == 0; }
+    //! map to device memory (non blocking)
+    gpu::GpuMat mapDevice(gpu::Stream& stream);
+    void unmapDevice(gpu::Stream& stream);
 
-    int type() const { return type_; }
-    int depth() const { return CV_MAT_DEPTH(type_); }
-    int channels() const { return CV_MAT_CN(type_); }
-    int elemSize() const { return CV_ELEM_SIZE(type_); }
-    int elemSize1() const { return CV_ELEM_SIZE1(type_); }
+    int rows() const;
+    int cols() const;
+    Size size() const;
+    bool empty() const;
 
+    int type() const;
+    int depth() const;
+    int channels() const;
+    int elemSize() const;
+    int elemSize1() const;
+
+    //! get OpenGL opject id
     unsigned int bufId() const;
 
     class Impl;
@@ -167,7 +178,7 @@ public:
 
     //! create texture
     void create(int arows, int acols, Format aformat, bool autoRelease = false);
-    void create(Size asize, Format aformat, bool autoRelease = false) { create(asize.height, asize.width, aformat, autoRelease); }
+    void create(Size asize, Format aformat, bool autoRelease = false);
 
     //! release memory and delete texture object
     void release();
@@ -184,13 +195,14 @@ public:
     //! bind texture to current active texture unit for GL_TEXTURE_2D target
     void bind() const;
 
-    int rows() const { return rows_; }
-    int cols() const { return cols_; }
-    Size size() const { return Size(cols_, rows_); }
-    bool empty() const { return rows_ == 0 || cols_ == 0; }
+    int rows() const;
+    int cols() const;
+    Size size() const;
+    bool empty() const;
 
-    Format format() const { return format_; }
+    Format format() const;
 
+    //! get OpenGL opject id
     unsigned int texId() const;
 
     class Impl;
@@ -226,8 +238,8 @@ public:
 
     void bind() const;
 
-    int size() const { return size_; }
-    bool empty() const { return size_ == 0; }
+    int size() const;
+    bool empty() const;
 
 private:
     int size_;
@@ -262,7 +274,7 @@ enum {
 CV_EXPORTS void render(const Arrays& arr, int mode = POINTS, Scalar color = Scalar::all(255));
 CV_EXPORTS void render(const Arrays& arr, InputArray indices, int mode = POINTS, Scalar color = Scalar::all(255));
 
-}} // namespace cv::gl
+}} // namespace cv::ogl
 
 namespace cv { namespace gpu {
 
@@ -271,13 +283,150 @@ CV_EXPORTS void setGlDevice(int device = 0);
 
 }}
 
-namespace cv {
 
-template <> CV_EXPORTS void Ptr<cv::ogl::Buffer::Impl>::delete_obj();
-template <> CV_EXPORTS void Ptr<cv::ogl::Texture2D::Impl>::delete_obj();
+////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
 
+inline
+cv::ogl::Buffer::Buffer(int arows, int acols, int atype, Target target, bool autoRelease) : rows_(0), cols_(0), type_(0)
+{
+    create(arows, acols, atype, target, autoRelease);
 }
 
-#endif // __cplusplus
+inline
+cv::ogl::Buffer::Buffer(Size asize, int atype, Target target, bool autoRelease) : rows_(0), cols_(0), type_(0)
+{
+    create(asize, atype, target, autoRelease);
+}
 
-#endif // __OPENCV_OPENGL_INTEROP_HPP__
+inline
+void cv::ogl::Buffer::create(Size asize, int atype, Target target, bool autoRelease)
+{
+    create(asize.height, asize.width, atype, target, autoRelease);
+}
+
+inline
+int cv::ogl::Buffer::rows() const
+{
+    return rows_;
+}
+
+inline
+int cv::ogl::Buffer::cols() const
+{
+    return cols_;
+}
+
+inline
+cv::Size cv::ogl::Buffer::size() const
+{
+    return Size(cols_, rows_);
+}
+
+inline
+bool cv::ogl::Buffer::empty() const
+{
+    return rows_ == 0 || cols_ == 0;
+}
+
+inline
+int cv::ogl::Buffer::type() const
+{
+    return type_;
+}
+
+inline
+int cv::ogl::Buffer::depth() const
+{
+    return CV_MAT_DEPTH(type_);
+}
+
+inline
+int cv::ogl::Buffer::channels() const
+{
+    return CV_MAT_CN(type_);
+}
+
+inline
+int cv::ogl::Buffer::elemSize() const
+{
+    return CV_ELEM_SIZE(type_);
+}
+
+inline
+int cv::ogl::Buffer::elemSize1() const
+{
+    return CV_ELEM_SIZE1(type_);
+}
+
+///////
+
+inline
+cv::ogl::Texture2D::Texture2D(int arows, int acols, Format aformat, bool autoRelease) : rows_(0), cols_(0), format_(NONE)
+{
+    create(arows, acols, aformat, autoRelease);
+}
+
+inline
+cv::ogl::Texture2D::Texture2D(Size asize, Format aformat, bool autoRelease) : rows_(0), cols_(0), format_(NONE)
+{
+    create(asize, aformat, autoRelease);
+}
+
+inline
+void cv::ogl::Texture2D::create(Size asize, Format aformat, bool autoRelease)
+{
+    create(asize.height, asize.width, aformat, autoRelease);
+}
+
+inline
+int cv::ogl::Texture2D::rows() const
+{
+    return rows_;
+}
+
+inline
+int cv::ogl::Texture2D::cols() const
+{
+    return cols_;
+}
+
+inline
+cv::Size cv::ogl::Texture2D::size() const
+{
+    return Size(cols_, rows_);
+}
+
+inline
+bool cv::ogl::Texture2D::empty() const
+{
+    return rows_ == 0 || cols_ == 0;
+}
+
+inline
+cv::ogl::Texture2D::Format cv::ogl::Texture2D::format() const
+{
+    return format_;
+}
+
+///////
+
+inline
+cv::ogl::Arrays::Arrays() : size_(0)
+{
+}
+
+inline
+int cv::ogl::Arrays::size() const
+{
+    return size_;
+}
+
+inline
+bool cv::ogl::Arrays::empty() const
+{
+    return size_ == 0;
+}
+
+#endif /* __OPENCV_CORE_OPENGL_HPP__ */
diff --git a/modules/core/include/opencv2/core/operations.hpp b/modules/core/include/opencv2/core/operations.hpp
index c259b892bc..f8aeddfb11 100644
--- a/modules/core/include/opencv2/core/operations.hpp
+++ b/modules/core/include/opencv2/core/operations.hpp
@@ -43,692 +43,21 @@
 #ifndef __OPENCV_CORE_OPERATIONS_HPP__
 #define __OPENCV_CORE_OPERATIONS_HPP__
 
-#ifndef SKIP_INCLUDES
-  #include <string.h>
-  #include <limits.h>
-#endif // SKIP_INCLUDES
-
-
-#ifdef __cplusplus
-
-#include <limits>
-
-#ifdef _MSC_VER
-# pragma warning(push)
-# pragma warning(disable:4127) //conditional expression is constant
+#ifndef __cplusplus
+#  error operations.hpp header must be compiled as C++
 #endif
 
+#include <cstdio>
+
 namespace cv
 {
 
-/////////////// saturate_cast (used in image & signal processing) ///////////////////
+////////////////////////////// Matx methods depending on core API /////////////////////////////
 
-template<typename _Tp> static inline _Tp saturate_cast(uchar v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(schar v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(ushort v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(short v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(unsigned v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(int v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(float v) { return _Tp(v); }
-template<typename _Tp> static inline _Tp saturate_cast(double v) { return _Tp(v); }
-
-template<> inline uchar saturate_cast<uchar>(schar v)
-{ return (uchar)std::max((int)v, 0); }
-template<> inline uchar saturate_cast<uchar>(ushort v)
-{ return (uchar)std::min((unsigned)v, (unsigned)UCHAR_MAX); }
-template<> inline uchar saturate_cast<uchar>(int v)
-{ return (uchar)((unsigned)v <= UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
-template<> inline uchar saturate_cast<uchar>(short v)
-{ return saturate_cast<uchar>((int)v); }
-template<> inline uchar saturate_cast<uchar>(unsigned v)
-{ return (uchar)std::min(v, (unsigned)UCHAR_MAX); }
-template<> inline uchar saturate_cast<uchar>(float v)
-{ int iv = cvRound(v); return saturate_cast<uchar>(iv); }
-template<> inline uchar saturate_cast<uchar>(double v)
-{ int iv = cvRound(v); return saturate_cast<uchar>(iv); }
-
-template<> inline schar saturate_cast<schar>(uchar v)
-{ return (schar)std::min((int)v, SCHAR_MAX); }
-template<> inline schar saturate_cast<schar>(ushort v)
-{ return (schar)std::min((unsigned)v, (unsigned)SCHAR_MAX); }
-template<> inline schar saturate_cast<schar>(int v)
-{
-    return (schar)((unsigned)(v-SCHAR_MIN) <= (unsigned)UCHAR_MAX ?
-                v : v > 0 ? SCHAR_MAX : SCHAR_MIN);
-}
-template<> inline schar saturate_cast<schar>(short v)
-{ return saturate_cast<schar>((int)v); }
-template<> inline schar saturate_cast<schar>(unsigned v)
-{ return (schar)std::min(v, (unsigned)SCHAR_MAX); }
-
-template<> inline schar saturate_cast<schar>(float v)
-{ int iv = cvRound(v); return saturate_cast<schar>(iv); }
-template<> inline schar saturate_cast<schar>(double v)
-{ int iv = cvRound(v); return saturate_cast<schar>(iv); }
-
-template<> inline ushort saturate_cast<ushort>(schar v)
-{ return (ushort)std::max((int)v, 0); }
-template<> inline ushort saturate_cast<ushort>(short v)
-{ return (ushort)std::max((int)v, 0); }
-template<> inline ushort saturate_cast<ushort>(int v)
-{ return (ushort)((unsigned)v <= (unsigned)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
-template<> inline ushort saturate_cast<ushort>(unsigned v)
-{ return (ushort)std::min(v, (unsigned)USHRT_MAX); }
-template<> inline ushort saturate_cast<ushort>(float v)
-{ int iv = cvRound(v); return saturate_cast<ushort>(iv); }
-template<> inline ushort saturate_cast<ushort>(double v)
-{ int iv = cvRound(v); return saturate_cast<ushort>(iv); }
-
-template<> inline short saturate_cast<short>(ushort v)
-{ return (short)std::min((int)v, SHRT_MAX); }
-template<> inline short saturate_cast<short>(int v)
-{
-    return (short)((unsigned)(v - SHRT_MIN) <= (unsigned)USHRT_MAX ?
-            v : v > 0 ? SHRT_MAX : SHRT_MIN);
-}
-template<> inline short saturate_cast<short>(unsigned v)
-{ return (short)std::min(v, (unsigned)SHRT_MAX); }
-template<> inline short saturate_cast<short>(float v)
-{ int iv = cvRound(v); return saturate_cast<short>(iv); }
-template<> inline short saturate_cast<short>(double v)
-{ int iv = cvRound(v); return saturate_cast<short>(iv); }
-
-template<> inline int saturate_cast<int>(float v) { return cvRound(v); }
-template<> inline int saturate_cast<int>(double v) { return cvRound(v); }
-
-// we intentionally do not clip negative numbers, to make -1 become 0xffffffff etc.
-template<> inline unsigned saturate_cast<unsigned>(float v){ return cvRound(v); }
-template<> inline unsigned saturate_cast<unsigned>(double v) { return cvRound(v); }
-
-inline int fast_abs(uchar v) { return v; }
-inline int fast_abs(schar v) { return std::abs((int)v); }
-inline int fast_abs(ushort v) { return v; }
-inline int fast_abs(short v) { return std::abs((int)v); }
-inline int fast_abs(int v) { return std::abs(v); }
-inline float fast_abs(float v) { return std::abs(v); }
-inline double fast_abs(double v) { return std::abs(v); }
-
-//////////////////////////////// Matx /////////////////////////////////
-
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx()
-{
-    for(int i = 0; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0)
-{
-    val[0] = v0;
-    for(int i = 1; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1)
-{
-    assert(channels >= 2);
-    val[0] = v0; val[1] = v1;
-    for(int i = 2; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2)
-{
-    assert(channels >= 3);
-    val[0] = v0; val[1] = v1; val[2] = v2;
-    for(int i = 3; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
-{
-    assert(channels >= 4);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    for(int i = 4; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4)
-{
-    assert(channels >= 5);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3; val[4] = v4;
-    for(int i = 5; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5)
-{
-    assert(channels >= 6);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5;
-    for(int i = 6; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6)
-{
-    assert(channels >= 7);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6;
-    for(int i = 7; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7)
-{
-    assert(channels >= 8);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
-    for(int i = 8; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                                                        _Tp v8)
-{
-    assert(channels >= 9);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
-    val[8] = v8;
-    for(int i = 9; i < channels; i++) val[i] = _Tp(0);
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                                                        _Tp v8, _Tp v9)
-{
-    assert(channels >= 10);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
-    val[8] = v8; val[9] = v9;
-    for(int i = 10; i < channels; i++) val[i] = _Tp(0);
-}
-
-
-template<typename _Tp, int m, int n>
-inline Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                            _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                            _Tp v8, _Tp v9, _Tp v10, _Tp v11)
-{
-    assert(channels == 12);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
-    val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;
-}
-
-template<typename _Tp, int m, int n>
-inline Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                           _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                           _Tp v8, _Tp v9, _Tp v10, _Tp v11,
-                           _Tp v12, _Tp v13, _Tp v14, _Tp v15)
-{
-    assert(channels == 16);
-    val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
-    val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
-    val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;
-    val[12] = v12; val[13] = v13; val[14] = v14; val[15] = v15;
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(const _Tp* values)
-{
-    for( int i = 0; i < channels; i++ ) val[i] = values[i];
-}
-
-template<typename _Tp, int m, int n> inline Matx<_Tp, m, n> Matx<_Tp, m, n>::all(_Tp alpha)
-{
-    Matx<_Tp, m, n> M;
-    for( int i = 0; i < m*n; i++ ) M.val[i] = alpha;
-    return M;
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::zeros()
-{
-    return all(0);
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::ones()
-{
-    return all(1);
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::eye()
-{
-    Matx<_Tp,m,n> M;
-    for(int i = 0; i < MIN(m,n); i++)
-        M(i,i) = 1;
-    return M;
-}
-
-template<typename _Tp, int m, int n> inline _Tp Matx<_Tp, m, n>::dot(const Matx<_Tp, m, n>& M) const
-{
-    _Tp s = 0;
-    for( int i = 0; i < m*n; i++ ) s += val[i]*M.val[i];
-    return s;
-}
-
-
-template<typename _Tp, int m, int n> inline double Matx<_Tp, m, n>::ddot(const Matx<_Tp, m, n>& M) const
-{
-    double s = 0;
-    for( int i = 0; i < m*n; i++ ) s += (double)val[i]*M.val[i];
-    return s;
-}
-
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::diag(const typename Matx<_Tp,m,n>::diag_type& d)
-{
-    Matx<_Tp,m,n> M;
-    for(int i = 0; i < MIN(m,n); i++)
-        M(i,i) = d(i, 0);
-    return M;
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::randu(_Tp a, _Tp b)
-{
-    Matx<_Tp,m,n> M;
-    Mat matM(M, false);
-    cv::randu(matM, Scalar(a), Scalar(b));
-    return M;
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n> Matx<_Tp,m,n>::randn(_Tp a, _Tp b)
-{
-    Matx<_Tp,m,n> M;
-    Mat matM(M, false);
-    cv::randn(matM, Scalar(a), Scalar(b));
-    return M;
-}
-
-template<typename _Tp, int m, int n> template<typename T2>
-inline Matx<_Tp, m, n>::operator Matx<T2, m, n>() const
-{
-    Matx<T2, m, n> M;
-    for( int i = 0; i < m*n; i++ ) M.val[i] = saturate_cast<T2>(val[i]);
-    return M;
-}
-
-
-template<typename _Tp, int m, int n> template<int m1, int n1> inline
-Matx<_Tp, m1, n1> Matx<_Tp, m, n>::reshape() const
-{
-    CV_DbgAssert(m1*n1 == m*n);
-    return (const Matx<_Tp, m1, n1>&)*this;
-}
-
-
-template<typename _Tp, int m, int n>
-template<int m1, int n1> inline
-Matx<_Tp, m1, n1> Matx<_Tp, m, n>::get_minor(int i, int j) const
-{
-    CV_DbgAssert(0 <= i && i+m1 <= m && 0 <= j && j+n1 <= n);
-    Matx<_Tp, m1, n1> s;
-    for( int di = 0; di < m1; di++ )
-        for( int dj = 0; dj < n1; dj++ )
-            s(di, dj) = (*this)(i+di, j+dj);
-    return s;
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, 1, n> Matx<_Tp, m, n>::row(int i) const
-{
-    CV_DbgAssert((unsigned)i < (unsigned)m);
-    return Matx<_Tp, 1, n>(&val[i*n]);
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, m, 1> Matx<_Tp, m, n>::col(int j) const
-{
-    CV_DbgAssert((unsigned)j < (unsigned)n);
-    Matx<_Tp, m, 1> v;
-    for( int i = 0; i < m; i++ )
-        v.val[i] = val[i*n + j];
-    return v;
-}
-
-
-template<typename _Tp, int m, int n> inline
-typename Matx<_Tp, m, n>::diag_type Matx<_Tp, m, n>::diag() const
-{
-    diag_type d;
-    for( int i = 0; i < MIN(m, n); i++ )
-        d.val[i] = val[i*n + i];
-    return d;
-}
-
-
-template<typename _Tp, int m, int n> inline
-const _Tp& Matx<_Tp, m, n>::operator ()(int i, int j) const
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)m && (unsigned)j < (unsigned)n );
-    return this->val[i*n + j];
-}
-
-
-template<typename _Tp, int m, int n> inline
-_Tp& Matx<_Tp, m, n>::operator ()(int i, int j)
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)m && (unsigned)j < (unsigned)n );
-    return val[i*n + j];
-}
-
-
-template<typename _Tp, int m, int n> inline
-const _Tp& Matx<_Tp, m, n>::operator ()(int i) const
-{
-    CV_DbgAssert( (m == 1 || n == 1) && (unsigned)i < (unsigned)(m+n-1) );
-    return val[i];
-}
-
-
-template<typename _Tp, int m, int n> inline
-_Tp& Matx<_Tp, m, n>::operator ()(int i)
-{
-    CV_DbgAssert( (m == 1 || n == 1) && (unsigned)i < (unsigned)(m+n-1) );
-    return val[i];
-}
-
-
-template<typename _Tp1, typename _Tp2, int m, int n> static inline
-Matx<_Tp1, m, n>& operator += (Matx<_Tp1, m, n>& a, const Matx<_Tp2, m, n>& b)
-{
-    for( int i = 0; i < m*n; i++ )
-        a.val[i] = saturate_cast<_Tp1>(a.val[i] + b.val[i]);
-    return a;
-}
-
-
-template<typename _Tp1, typename _Tp2, int m, int n> static inline
-Matx<_Tp1, m, n>& operator -= (Matx<_Tp1, m, n>& a, const Matx<_Tp2, m, n>& b)
+namespace internal
 {
-    for( int i = 0; i < m*n; i++ )
-        a.val[i] = saturate_cast<_Tp1>(a.val[i] - b.val[i]);
-    return a;
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_AddOp)
-{
-    for( int i = 0; i < m*n; i++ )
-        val[i] = saturate_cast<_Tp>(a.val[i] + b.val[i]);
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_SubOp)
-{
-    for( int i = 0; i < m*n; i++ )
-        val[i] = saturate_cast<_Tp>(a.val[i] - b.val[i]);
-}
-
-
-template<typename _Tp, int m, int n> template<typename _T2> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, _T2 alpha, Matx_ScaleOp)
-{
-    for( int i = 0; i < m*n; i++ )
-        val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b, Matx_MulOp)
-{
-    for( int i = 0; i < m*n; i++ )
-        val[i] = saturate_cast<_Tp>(a.val[i] * b.val[i]);
-}
-
-
-template<typename _Tp, int m, int n> template<int l> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b, Matx_MatMulOp)
-{
-    for( int i = 0; i < m; i++ )
-        for( int j = 0; j < n; j++ )
-        {
-            _Tp s = 0;
-            for( int k = 0; k < l; k++ )
-                s += a(i, k) * b(k, j);
-            val[i*n + j] = s;
-        }
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp,m,n>::Matx(const Matx<_Tp, n, m>& a, Matx_TOp)
-{
-    for( int i = 0; i < m; i++ )
-        for( int j = 0; j < n; j++ )
-            val[i*n + j] = a(j, i);
-}
-
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator + (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
-{
-    return Matx<_Tp, m, n>(a, b, Matx_AddOp());
-}
-
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator - (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
-{
-    return Matx<_Tp, m, n>(a, b, Matx_SubOp());
-}
-
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, int alpha)
-{
-    for( int i = 0; i < m*n; i++ )
-        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
-    return a;
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, float alpha)
-{
-    for( int i = 0; i < m*n; i++ )
-        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
-    return a;
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n>& operator *= (Matx<_Tp, m, n>& a, double alpha)
-{
-    for( int i = 0; i < m*n; i++ )
-        a.val[i] = saturate_cast<_Tp>(a.val[i] * alpha);
-    return a;
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, int alpha)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, float alpha)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (const Matx<_Tp, m, n>& a, double alpha)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (int alpha, const Matx<_Tp, m, n>& a)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (float alpha, const Matx<_Tp, m, n>& a)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator * (double alpha, const Matx<_Tp, m, n>& a)
-{
-    return Matx<_Tp, m, n>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int m, int n> static inline
-Matx<_Tp, m, n> operator - (const Matx<_Tp, m, n>& a)
-{
-    return Matx<_Tp, m, n>(a, -1, Matx_ScaleOp());
-}
-
-
-template<typename _Tp, int m, int n, int l> static inline
-Matx<_Tp, m, n> operator * (const Matx<_Tp, m, l>& a, const Matx<_Tp, l, n>& b)
-{
-    return Matx<_Tp, m, n>(a, b, Matx_MatMulOp());
-}
-
-
-template<typename _Tp, int m, int n> static inline
-Vec<_Tp, m> operator * (const Matx<_Tp, m, n>& a, const Vec<_Tp, n>& b)
-{
-    Matx<_Tp, m, 1> c(a, b, Matx_MatMulOp());
-    return reinterpret_cast<const Vec<_Tp, m>&>(c);
-}
-
-
-template<typename _Tp> static inline
-Point_<_Tp> operator * (const Matx<_Tp, 2, 2>& a, const Point_<_Tp>& b)
-{
-    Matx<_Tp, 2, 1> tmp = a*Vec<_Tp,2>(b.x, b.y);
-    return Point_<_Tp>(tmp.val[0], tmp.val[1]);
-}
-
-
-template<typename _Tp> static inline
-Point3_<_Tp> operator * (const Matx<_Tp, 3, 3>& a, const Point3_<_Tp>& b)
-{
-    Matx<_Tp, 3, 1> tmp = a*Vec<_Tp,3>(b.x, b.y, b.z);
-    return Point3_<_Tp>(tmp.val[0], tmp.val[1], tmp.val[2]);
-}
-
-
-template<typename _Tp> static inline
-Point3_<_Tp> operator * (const Matx<_Tp, 3, 3>& a, const Point_<_Tp>& b)
-{
-    Matx<_Tp, 3, 1> tmp = a*Vec<_Tp,3>(b.x, b.y, 1);
-    return Point3_<_Tp>(tmp.val[0], tmp.val[1], tmp.val[2]);
-}
-
-
-template<typename _Tp> static inline
-Matx<_Tp, 4, 1> operator * (const Matx<_Tp, 4, 4>& a, const Point3_<_Tp>& b)
-{
-    return a*Matx<_Tp, 4, 1>(b.x, b.y, b.z, 1);
-}
-
-
-template<typename _Tp> static inline
-Scalar operator * (const Matx<_Tp, 4, 4>& a, const Scalar& b)
-{
-    Matx<double, 4, 1> c(Matx<double, 4, 4>(a), b, Matx_MatMulOp());
-    return reinterpret_cast<const Scalar&>(c);
-}
-
-
-static inline
-Scalar operator * (const Matx<double, 4, 4>& a, const Scalar& b)
-{
-    Matx<double, 4, 1> c(a, b, Matx_MatMulOp());
-    return reinterpret_cast<const Scalar&>(c);
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, m, n> Matx<_Tp, m, n>::mul(const Matx<_Tp, m, n>& a) const
-{
-    return Matx<_Tp, m, n>(*this, a, Matx_MulOp());
-}
-
-
-CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
-CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
-CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
-CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
-
-
-template<typename _Tp, int m> struct CV_EXPORTS Matx_DetOp
-{
-    double operator ()(const Matx<_Tp, m, m>& a) const
-    {
-        Matx<_Tp, m, m> temp = a;
-        double p = LU(temp.val, m*sizeof(_Tp), m, 0, 0, 0);
-        if( p == 0 )
-            return p;
-        for( int i = 0; i < m; i++ )
-            p *= temp(i, i);
-        return 1./p;
-    }
-};
-
-
-template<typename _Tp> struct CV_EXPORTS Matx_DetOp<_Tp, 1>
-{
-    double operator ()(const Matx<_Tp, 1, 1>& a) const
-    {
-        return a(0,0);
-    }
-};
-
-
-template<typename _Tp> struct CV_EXPORTS Matx_DetOp<_Tp, 2>
-{
-    double operator ()(const Matx<_Tp, 2, 2>& a) const
-    {
-        return a(0,0)*a(1,1) - a(0,1)*a(1,0);
-    }
-};
-
-
-template<typename _Tp> struct CV_EXPORTS Matx_DetOp<_Tp, 3>
-{
-    double operator ()(const Matx<_Tp, 3, 3>& a) const
-    {
-        return a(0,0)*(a(1,1)*a(2,2) - a(2,1)*a(1,2)) -
-            a(0,1)*(a(1,0)*a(2,2) - a(2,0)*a(1,2)) +
-            a(0,2)*(a(1,0)*a(2,1) - a(2,0)*a(1,1));
-    }
-};
-
-template<typename _Tp, int m> static inline
-double determinant(const Matx<_Tp, m, m>& a)
-{
-    return Matx_DetOp<_Tp, m>()(a);
-}
-
-
-template<typename _Tp, int m, int n> static inline
-double trace(const Matx<_Tp, m, n>& a)
-{
-    _Tp s = 0;
-    for( int i = 0; i < std::min(m, n); i++ )
-        s += a(i,i);
-    return s;
-}
-
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, n, m> Matx<_Tp, m, n>::t() const
-{
-    return Matx<_Tp, n, m>(*this, Matx_TOp());
-}
-
 
-template<typename _Tp, int m> struct CV_EXPORTS Matx_FastInvOp
+template<typename _Tp, int m> struct Matx_FastInvOp
 {
     bool operator()(const Matx<_Tp, m, m>& a, Matx<_Tp, m, m>& b, int method) const
     {
@@ -745,8 +74,7 @@ template<typename _Tp, int m> struct CV_EXPORTS Matx_FastInvOp
     }
 };
 
-
-template<typename _Tp> struct CV_EXPORTS Matx_FastInvOp<_Tp, 2>
+template<typename _Tp> struct Matx_FastInvOp<_Tp, 2>
 {
     bool operator()(const Matx<_Tp, 2, 2>& a, Matx<_Tp, 2, 2>& b, int) const
     {
@@ -762,8 +90,7 @@ template<typename _Tp> struct CV_EXPORTS Matx_FastInvOp<_Tp, 2>
     }
 };
 
-
-template<typename _Tp> struct CV_EXPORTS Matx_FastInvOp<_Tp, 3>
+template<typename _Tp> struct Matx_FastInvOp<_Tp, 3>
 {
     bool operator()(const Matx<_Tp, 3, 3>& a, Matx<_Tp, 3, 3>& b, int) const
     {
@@ -787,23 +114,7 @@ template<typename _Tp> struct CV_EXPORTS Matx_FastInvOp<_Tp, 3>
 };
 
 
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, n, m> Matx<_Tp, m, n>::inv(int method) const
-{
-    Matx<_Tp, n, m> b;
-    bool ok;
-    if( method == DECOMP_LU || method == DECOMP_CHOLESKY )
-        ok = Matx_FastInvOp<_Tp, m>()(*this, b, method);
-    else
-    {
-        Mat A(*this, false), B(b, false);
-        ok = (invert(A, B, method) != 0);
-    }
-    return ok ? b : Matx<_Tp, n, m>::zeros();
-}
-
-
-template<typename _Tp, int m, int n> struct CV_EXPORTS Matx_FastSolveOp
+template<typename _Tp, int m, int n> struct Matx_FastSolveOp
 {
     bool operator()(const Matx<_Tp, m, m>& a, const Matx<_Tp, m, n>& b,
                     Matx<_Tp, m, n>& x, int method) const
@@ -817,8 +128,7 @@ template<typename _Tp, int m, int n> struct CV_EXPORTS Matx_FastSolveOp
     }
 };
 
-
-template<typename _Tp> struct CV_EXPORTS Matx_FastSolveOp<_Tp, 2, 1>
+template<typename _Tp> struct Matx_FastSolveOp<_Tp, 2, 1>
 {
     bool operator()(const Matx<_Tp, 2, 2>& a, const Matx<_Tp, 2, 1>& b,
                     Matx<_Tp, 2, 1>& x, int) const
@@ -833,8 +143,7 @@ template<typename _Tp> struct CV_EXPORTS Matx_FastSolveOp<_Tp, 2, 1>
     }
 };
 
-
-template<typename _Tp> struct CV_EXPORTS Matx_FastSolveOp<_Tp, 3, 1>
+template<typename _Tp> struct Matx_FastSolveOp<_Tp, 3, 1>
 {
     bool operator()(const Matx<_Tp, 3, 3>& a, const Matx<_Tp, 3, 1>& b,
                     Matx<_Tp, 3, 1>& x, int) const
@@ -858,6 +167,38 @@ template<typename _Tp> struct CV_EXPORTS Matx_FastSolveOp<_Tp, 3, 1>
     }
 };
 
+} // internal
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::randu(_Tp a, _Tp b)
+{
+    Matx<_Tp,m,n> M;
+    cv::randu(M, Scalar(a), Scalar(b));
+    return M;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp,m,n> Matx<_Tp,m,n>::randn(_Tp a, _Tp b)
+{
+    Matx<_Tp,m,n> M;
+    cv::randn(M, Scalar(a), Scalar(b));
+    return M;
+}
+
+template<typename _Tp, int m, int n> inline
+Matx<_Tp, n, m> Matx<_Tp, m, n>::inv(int method) const
+{
+    Matx<_Tp, n, m> b;
+    bool ok;
+    if( method == DECOMP_LU || method == DECOMP_CHOLESKY )
+        ok = internal::Matx_FastInvOp<_Tp, m>()(*this, b, method);
+    else
+    {
+        Mat A(*this, false), B(b, false);
+        ok = (invert(A, B, method) != 0);
+    }
+    return ok ? b : Matx<_Tp, n, m>::zeros();
+}
 
 template<typename _Tp, int m, int n> template<int l> inline
 Matx<_Tp, n, l> Matx<_Tp, m, n>::solve(const Matx<_Tp, m, l>& rhs, int method) const
@@ -865,7 +206,7 @@ Matx<_Tp, n, l> Matx<_Tp, m, n>::solve(const Matx<_Tp, m, l>& rhs, int method) c
     Matx<_Tp, n, l> x;
     bool ok;
     if( method == DECOMP_LU || method == DECOMP_CHOLESKY )
-        ok = Matx_FastSolveOp<_Tp, m, l>()(*this, rhs, x, method);
+        ok = internal::Matx_FastSolveOp<_Tp, m, l>()(*this, rhs, x, method);
     else
     {
         Mat A(*this, false), B(rhs, false), X(x, false);
@@ -875,1606 +216,169 @@ Matx<_Tp, n, l> Matx<_Tp, m, n>::solve(const Matx<_Tp, m, l>& rhs, int method) c
     return ok ? x : Matx<_Tp, n, l>::zeros();
 }
 
-template<typename _Tp, int m, int n> inline
-Vec<_Tp, n> Matx<_Tp, m, n>::solve(const Vec<_Tp, m>& rhs, int method) const
-{
-    Matx<_Tp, n, 1> x = solve(reinterpret_cast<const Matx<_Tp, m, 1>&>(rhs), method);
-    return reinterpret_cast<Vec<_Tp, n>&>(x);
-}
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normL2Sqr(const _Tp* a, int n)
-{
-    _AccTp s = 0;
-    int i=0;
- #if CV_ENABLE_UNROLLED
-    for( ; i <= n - 4; i += 4 )
-    {
-        _AccTp v0 = a[i], v1 = a[i+1], v2 = a[i+2], v3 = a[i+3];
-        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
-    }
-#endif
-    for( ; i < n; i++ )
-    {
-        _AccTp v = a[i];
-        s += v*v;
-    }
-    return s;
-}
-
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normL1(const _Tp* a, int n)
-{
-    _AccTp s = 0;
-    int i = 0;
-#if CV_ENABLE_UNROLLED
-    for(; i <= n - 4; i += 4 )
-    {
-        s += (_AccTp)fast_abs(a[i]) + (_AccTp)fast_abs(a[i+1]) +
-            (_AccTp)fast_abs(a[i+2]) + (_AccTp)fast_abs(a[i+3]);
-    }
-#endif
-    for( ; i < n; i++ )
-        s += fast_abs(a[i]);
-    return s;
-}
-
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normInf(const _Tp* a, int n)
-{
-    _AccTp s = 0;
-    for( int i = 0; i < n; i++ )
-        s = std::max(s, (_AccTp)fast_abs(a[i]));
-    return s;
-}
-
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normL2Sqr(const _Tp* a, const _Tp* b, int n)
-{
-    _AccTp s = 0;
-    int i= 0;
-#if CV_ENABLE_UNROLLED
-    for(; i <= n - 4; i += 4 )
-    {
-        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
-        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
-    }
-#endif
-    for( ; i < n; i++ )
-    {
-        _AccTp v = _AccTp(a[i] - b[i]);
-        s += v*v;
-    }
-    return s;
-}
-
-CV_EXPORTS float normL2Sqr_(const float* a, const float* b, int n);
-CV_EXPORTS float normL1_(const float* a, const float* b, int n);
-CV_EXPORTS int normL1_(const uchar* a, const uchar* b, int n);
-CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n);
-CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n, int cellSize);
-
-template<> inline float normL2Sqr(const float* a, const float* b, int n)
-{
-    if( n >= 8 )
-        return normL2Sqr_(a, b, n);
-    float s = 0;
-    for( int i = 0; i < n; i++ )
-    {
-        float v = a[i] - b[i];
-        s += v*v;
-    }
-    return s;
-}
-
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normL1(const _Tp* a, const _Tp* b, int n)
-{
-    _AccTp s = 0;
-    int i= 0;
-#if CV_ENABLE_UNROLLED
-    for(; i <= n - 4; i += 4 )
-    {
-        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
-        s += std::abs(v0) + std::abs(v1) + std::abs(v2) + std::abs(v3);
-    }
-#endif
-    for( ; i < n; i++ )
-    {
-        _AccTp v = _AccTp(a[i] - b[i]);
-        s += std::abs(v);
-    }
-    return s;
-}
-
-template<> inline float normL1(const float* a, const float* b, int n)
-{
-    if( n >= 8 )
-        return normL1_(a, b, n);
-    float s = 0;
-    for( int i = 0; i < n; i++ )
-    {
-        float v = a[i] - b[i];
-        s += std::abs(v);
-    }
-    return s;
-}
-
-template<> inline int normL1(const uchar* a, const uchar* b, int n)
-{
-    return normL1_(a, b, n);
-}
-
-template<typename _Tp, typename _AccTp> static inline
-_AccTp normInf(const _Tp* a, const _Tp* b, int n)
-{
-    _AccTp s = 0;
-    for( int i = 0; i < n; i++ )
-    {
-        _AccTp v0 = a[i] - b[i];
-        s = std::max(s, std::abs(v0));
-    }
-    return s;
-}
-
-
-template<typename _Tp, int m, int n> static inline
-double norm(const Matx<_Tp, m, n>& M)
-{
-    return std::sqrt(normL2Sqr<_Tp, double>(M.val, m*n));
-}
-
-
-template<typename _Tp, int m, int n> static inline
-double norm(const Matx<_Tp, m, n>& M, int normType)
-{
-    return normType == NORM_INF ? (double)normInf<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n) :
-        normType == NORM_L1 ? (double)normL1<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n) :
-        std::sqrt((double)normL2Sqr<_Tp, typename DataType<_Tp>::work_type>(M.val, m*n));
-}
-
-
-template<typename _Tp, int m, int n> static inline
-bool operator == (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
-{
-    for( int i = 0; i < m*n; i++ )
-        if( a.val[i] != b.val[i] ) return false;
-    return true;
-}
-
-template<typename _Tp, int m, int n> static inline
-bool operator != (const Matx<_Tp, m, n>& a, const Matx<_Tp, m, n>& b)
-{
-    return !(a == b);
-}
-
-
-template<typename _Tp, typename _T2, int m, int n> static inline
-MatxCommaInitializer<_Tp, m, n> operator << (const Matx<_Tp, m, n>& mtx, _T2 val)
-{
-    MatxCommaInitializer<_Tp, m, n> commaInitializer((Matx<_Tp, m, n>*)&mtx);
-    return (commaInitializer, val);
-}
-
-template<typename _Tp, int m, int n> inline
-MatxCommaInitializer<_Tp, m, n>::MatxCommaInitializer(Matx<_Tp, m, n>* _mtx)
-    : dst(_mtx), idx(0)
-{}
-
-template<typename _Tp, int m, int n> template<typename _T2> inline
-MatxCommaInitializer<_Tp, m, n>& MatxCommaInitializer<_Tp, m, n>::operator , (_T2 value)
-{
-    CV_DbgAssert( idx < m*n );
-    dst->val[idx++] = saturate_cast<_Tp>(value);
-    return *this;
-}
-
-template<typename _Tp, int m, int n> inline
-Matx<_Tp, m, n> MatxCommaInitializer<_Tp, m, n>::operator *() const
-{
-    CV_DbgAssert( idx == n*m );
-    return *dst;
-}
-
-/////////////////////////// short vector (Vec) /////////////////////////////
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec()
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0)
-    : Matx<_Tp, cn, 1>(v0)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1)
-    : Matx<_Tp, cn, 1>(v0, v1)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2)
-    : Matx<_Tp, cn, 1>(v0, v1, v2)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4, _Tp v5)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                                                        _Tp v8)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7, v8)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
-                                                        _Tp v4, _Tp v5, _Tp v6, _Tp v7,
-                                                        _Tp v8, _Tp v9)
-    : Matx<_Tp, cn, 1>(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(const _Tp* values)
-    : Matx<_Tp, cn, 1>(values)
-{}
-
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::Vec(const Vec<_Tp, cn>& m)
-    : Matx<_Tp, cn, 1>(m.val)
-{}
-
-template<typename _Tp, int cn> inline
-Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_AddOp op)
-: Matx<_Tp, cn, 1>(a, b, op)
-{}
-
-template<typename _Tp, int cn> inline
-Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, const Matx<_Tp, cn, 1>& b, Matx_SubOp op)
-: Matx<_Tp, cn, 1>(a, b, op)
-{}
-
-template<typename _Tp, int cn> template<typename _T2> inline
-Vec<_Tp, cn>::Vec(const Matx<_Tp, cn, 1>& a, _T2 alpha, Matx_ScaleOp op)
-: Matx<_Tp, cn, 1>(a, alpha, op)
-{}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn> Vec<_Tp, cn>::all(_Tp alpha)
-{
-    Vec v;
-    for( int i = 0; i < cn; i++ ) v.val[i] = alpha;
-    return v;
-}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn> Vec<_Tp, cn>::mul(const Vec<_Tp, cn>& v) const
-{
-    Vec<_Tp, cn> w;
-    for( int i = 0; i < cn; i++ ) w.val[i] = saturate_cast<_Tp>(this->val[i]*v.val[i]);
-    return w;
-}
-
-template<typename _Tp> Vec<_Tp, 2> conjugate(const Vec<_Tp, 2>& v)
-{
-    return Vec<_Tp, 2>(v[0], -v[1]);
-}
-
-template<typename _Tp> Vec<_Tp, 4> conjugate(const Vec<_Tp, 4>& v)
-{
-    return Vec<_Tp, 4>(v[0], -v[1], -v[2], -v[3]);
-}
-
-template<> inline Vec<float, 2> Vec<float, 2>::conj() const
-{
-    return conjugate(*this);
-}
-
-template<> inline Vec<double, 2> Vec<double, 2>::conj() const
-{
-    return conjugate(*this);
-}
-
-template<> inline Vec<float, 4> Vec<float, 4>::conj() const
-{
-    return conjugate(*this);
-}
-
-template<> inline Vec<double, 4> Vec<double, 4>::conj() const
-{
-    return conjugate(*this);
-}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn> Vec<_Tp, cn>::cross(const Vec<_Tp, cn>&) const
-{
-    CV_Error(CV_StsError, "for arbitrary-size vector there is no cross-product defined");
-    return Vec<_Tp, cn>();
-}
-
-template<typename _Tp, int cn> template<typename T2>
-inline Vec<_Tp, cn>::operator Vec<T2, cn>() const
-{
-    Vec<T2, cn> v;
-    for( int i = 0; i < cn; i++ ) v.val[i] = saturate_cast<T2>(this->val[i]);
-    return v;
-}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn>::operator CvScalar() const
-{
-    CvScalar s = {{0,0,0,0}};
-    int i;
-    for( i = 0; i < std::min(cn, 4); i++ ) s.val[i] = this->val[i];
-    for( ; i < 4; i++ ) s.val[i] = 0;
-    return s;
-}
-
-template<typename _Tp, int cn> inline const _Tp& Vec<_Tp, cn>::operator [](int i) const
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)cn );
-    return this->val[i];
-}
-
-template<typename _Tp, int cn> inline _Tp& Vec<_Tp, cn>::operator [](int i)
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)cn );
-    return this->val[i];
-}
 
-template<typename _Tp, int cn> inline const _Tp& Vec<_Tp, cn>::operator ()(int i) const
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)cn );
-    return this->val[i];
-}
 
-template<typename _Tp, int cn> inline _Tp& Vec<_Tp, cn>::operator ()(int i)
-{
-    CV_DbgAssert( (unsigned)i < (unsigned)cn );
-    return this->val[i];
-}
+////////////////////////// Augmenting algebraic & logical operations //////////////////////////
 
-template<typename _Tp1, typename _Tp2, int cn> static inline Vec<_Tp1, cn>&
-operator += (Vec<_Tp1, cn>& a, const Vec<_Tp2, cn>& b)
-{
-    for( int i = 0; i < cn; i++ )
-        a.val[i] = saturate_cast<_Tp1>(a.val[i] + b.val[i]);
-    return a;
-}
+#define CV_MAT_AUG_OPERATOR1(op, cvop, A, B) \
+    static inline A& operator op (A& a, const B& b) { cvop; return a; }
 
-template<typename _Tp1, typename _Tp2, int cn> static inline Vec<_Tp1, cn>&
-operator -= (Vec<_Tp1, cn>& a, const Vec<_Tp2, cn>& b)
-{
-    for( int i = 0; i < cn; i++ )
-        a.val[i] = saturate_cast<_Tp1>(a.val[i] - b.val[i]);
-    return a;
-}
+#define CV_MAT_AUG_OPERATOR(op, cvop, A, B)   \
+    CV_MAT_AUG_OPERATOR1(op, cvop, A, B)      \
+    CV_MAT_AUG_OPERATOR1(op, cvop, const A, B)
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator + (const Vec<_Tp, cn>& a, const Vec<_Tp, cn>& b)
-{
-    return Vec<_Tp, cn>(a, b, Matx_AddOp());
-}
+#define CV_MAT_AUG_OPERATOR_T(op, cvop, A, B)                   \
+    template<typename _Tp> CV_MAT_AUG_OPERATOR1(op, cvop, A, B) \
+    template<typename _Tp> CV_MAT_AUG_OPERATOR1(op, cvop, const A, B)
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator - (const Vec<_Tp, cn>& a, const Vec<_Tp, cn>& b)
-{
-    return Vec<_Tp, cn>(a, b, Matx_SubOp());
-}
+CV_MAT_AUG_OPERATOR  (+=, cv::add(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR  (+=, cv::add(a,b,a), Mat, Scalar)
+CV_MAT_AUG_OPERATOR_T(+=, cv::add(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(+=, cv::add(a,b,a), Mat_<_Tp>, Scalar)
+CV_MAT_AUG_OPERATOR_T(+=, cv::add(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, int alpha)
-{
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*alpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (-=, cv::subtract(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR  (-=, cv::subtract(a,b,a), Mat, Scalar)
+CV_MAT_AUG_OPERATOR_T(-=, cv::subtract(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(-=, cv::subtract(a,b,a), Mat_<_Tp>, Scalar)
+CV_MAT_AUG_OPERATOR_T(-=, cv::subtract(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, float alpha)
-{
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*alpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (*=, cv::gemm(a, b, 1, Mat(), 0, a, 0), Mat, Mat)
+CV_MAT_AUG_OPERATOR_T(*=, cv::gemm(a, b, 1, Mat(), 0, a, 0), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(*=, cv::gemm(a, b, 1, Mat(), 0, a, 0), Mat_<_Tp>, Mat_<_Tp>)
+CV_MAT_AUG_OPERATOR  (*=, a.convertTo(a, -1, b), Mat, double)
+CV_MAT_AUG_OPERATOR_T(*=, a.convertTo(a, -1, b), Mat_<_Tp>, double)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator *= (Vec<_Tp, cn>& a, double alpha)
-{
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*alpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (/=, cv::divide(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR_T(/=, cv::divide(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(/=, cv::divide(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
+CV_MAT_AUG_OPERATOR  (/=, a.convertTo((Mat&)a, -1, 1./b), Mat, double)
+CV_MAT_AUG_OPERATOR_T(/=, a.convertTo((Mat&)a, -1, 1./b), Mat_<_Tp>, double)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, int alpha)
-{
-    double ialpha = 1./alpha;
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (&=, cv::bitwise_and(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR  (&=, cv::bitwise_and(a,b,a), Mat, Scalar)
+CV_MAT_AUG_OPERATOR_T(&=, cv::bitwise_and(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(&=, cv::bitwise_and(a,b,a), Mat_<_Tp>, Scalar)
+CV_MAT_AUG_OPERATOR_T(&=, cv::bitwise_and(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, float alpha)
-{
-    float ialpha = 1.f/alpha;
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (|=, cv::bitwise_or(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR  (|=, cv::bitwise_or(a,b,a), Mat, Scalar)
+CV_MAT_AUG_OPERATOR_T(|=, cv::bitwise_or(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(|=, cv::bitwise_or(a,b,a), Mat_<_Tp>, Scalar)
+CV_MAT_AUG_OPERATOR_T(|=, cv::bitwise_or(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
 
-template<typename _Tp, int cn> static inline
-Vec<_Tp, cn>& operator /= (Vec<_Tp, cn>& a, double alpha)
-{
-    double ialpha = 1./alpha;
-    for( int i = 0; i < cn; i++ )
-        a[i] = saturate_cast<_Tp>(a[i]*ialpha);
-    return a;
-}
+CV_MAT_AUG_OPERATOR  (^=, cv::bitwise_xor(a,b,a), Mat, Mat)
+CV_MAT_AUG_OPERATOR  (^=, cv::bitwise_xor(a,b,a), Mat, Scalar)
+CV_MAT_AUG_OPERATOR_T(^=, cv::bitwise_xor(a,b,a), Mat_<_Tp>, Mat)
+CV_MAT_AUG_OPERATOR_T(^=, cv::bitwise_xor(a,b,a), Mat_<_Tp>, Scalar)
+CV_MAT_AUG_OPERATOR_T(^=, cv::bitwise_xor(a,b,a), Mat_<_Tp>, Mat_<_Tp>)
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (const Vec<_Tp, cn>& a, int alpha)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
+#undef CV_MAT_AUG_OPERATOR_T
+#undef CV_MAT_AUG_OPERATOR
+#undef CV_MAT_AUG_OPERATOR1
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (int alpha, const Vec<_Tp, cn>& a)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (const Vec<_Tp, cn>& a, float alpha)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (float alpha, const Vec<_Tp, cn>& a)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (const Vec<_Tp, cn>& a, double alpha)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
-
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator * (double alpha, const Vec<_Tp, cn>& a)
-{
-    return Vec<_Tp, cn>(a, alpha, Matx_ScaleOp());
-}
+///////////////////////////////////////////// SVD /////////////////////////////////////////////
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator / (const Vec<_Tp, cn>& a, int alpha)
+inline SVD::SVD() {}
+inline SVD::SVD( InputArray m, int flags ) { operator ()(m, flags); }
+inline void SVD::solveZ( InputArray m, OutputArray _dst )
 {
-    return Vec<_Tp, cn>(a, 1./alpha, Matx_ScaleOp());
+    Mat mtx = m.getMat();
+    SVD svd(mtx, (mtx.rows >= mtx.cols ? 0 : SVD::FULL_UV));
+    _dst.create(svd.vt.cols, 1, svd.vt.type());
+    Mat dst = _dst.getMat();
+    svd.vt.row(svd.vt.rows-1).reshape(1,svd.vt.cols).copyTo(dst);
 }
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator / (const Vec<_Tp, cn>& a, float alpha)
+template<typename _Tp, int m, int n, int nm> inline void
+    SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w, Matx<_Tp, m, nm>& u, Matx<_Tp, n, nm>& vt )
 {
-    return Vec<_Tp, cn>(a, 1.f/alpha, Matx_ScaleOp());
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
+    Mat _a(a, false), _u(u, false), _w(w, false), _vt(vt, false);
+    SVD::compute(_a, _w, _u, _vt);
+    CV_Assert(_w.data == (uchar*)&w.val[0] && _u.data == (uchar*)&u.val[0] && _vt.data == (uchar*)&vt.val[0]);
 }
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator / (const Vec<_Tp, cn>& a, double alpha)
+template<typename _Tp, int m, int n, int nm> inline void
+SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w )
 {
-    return Vec<_Tp, cn>(a, 1./alpha, Matx_ScaleOp());
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
+    Mat _a(a, false), _w(w, false);
+    SVD::compute(_a, _w);
+    CV_Assert(_w.data == (uchar*)&w.val[0]);
 }
 
-template<typename _Tp, int cn> static inline Vec<_Tp, cn>
-operator - (const Vec<_Tp, cn>& a)
+template<typename _Tp, int m, int n, int nm, int nb> inline void
+SVD::backSubst( const Matx<_Tp, nm, 1>& w, const Matx<_Tp, m, nm>& u,
+                const Matx<_Tp, n, nm>& vt, const Matx<_Tp, m, nb>& rhs,
+                Matx<_Tp, n, nb>& dst )
 {
-    Vec<_Tp,cn> t;
-    for( int i = 0; i < cn; i++ ) t.val[i] = saturate_cast<_Tp>(-a.val[i]);
-    return t;
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
+    Mat _u(u, false), _w(w, false), _vt(vt, false), _rhs(rhs, false), _dst(dst, false);
+    SVD::backSubst(_w, _u, _vt, _rhs, _dst);
+    CV_Assert(_dst.data == (uchar*)&dst.val[0]);
 }
 
-template<typename _Tp> inline Vec<_Tp, 4> operator * (const Vec<_Tp, 4>& v1, const Vec<_Tp, 4>& v2)
-{
-    return Vec<_Tp, 4>(saturate_cast<_Tp>(v1[0]*v2[0] - v1[1]*v2[1] - v1[2]*v2[2] - v1[3]*v2[3]),
-                       saturate_cast<_Tp>(v1[0]*v2[1] + v1[1]*v2[0] + v1[2]*v2[3] - v1[3]*v2[2]),
-                       saturate_cast<_Tp>(v1[0]*v2[2] - v1[1]*v2[3] + v1[2]*v2[0] + v1[3]*v2[1]),
-                       saturate_cast<_Tp>(v1[0]*v2[3] + v1[1]*v2[2] - v1[2]*v2[1] + v1[3]*v2[0]));
-}
-
-template<typename _Tp> inline Vec<_Tp, 4>& operator *= (Vec<_Tp, 4>& v1, const Vec<_Tp, 4>& v2)
-{
-    v1 = v1 * v2;
-    return v1;
-}
-
-template<> inline Vec<float, 3> Vec<float, 3>::cross(const Vec<float, 3>& v) const
-{
-    return Vec<float,3>(val[1]*v.val[2] - val[2]*v.val[1],
-                     val[2]*v.val[0] - val[0]*v.val[2],
-                     val[0]*v.val[1] - val[1]*v.val[0]);
-}
-
-template<> inline Vec<double, 3> Vec<double, 3>::cross(const Vec<double, 3>& v) const
-{
-    return Vec<double,3>(val[1]*v.val[2] - val[2]*v.val[1],
-                     val[2]*v.val[0] - val[0]*v.val[2],
-                     val[0]*v.val[1] - val[1]*v.val[0]);
-}
-
-template<typename _Tp, int cn> inline Vec<_Tp, cn> normalize(const Vec<_Tp, cn>& v)
-{
-    double nv = norm(v);
-    return v * (nv ? 1./nv : 0.);
-}
-
-template<typename _Tp, typename _T2, int cn> static inline
-VecCommaInitializer<_Tp, cn> operator << (const Vec<_Tp, cn>& vec, _T2 val)
-{
-    VecCommaInitializer<_Tp, cn> commaInitializer((Vec<_Tp, cn>*)&vec);
-    return (commaInitializer, val);
-}
-
-template<typename _Tp, int cn> inline
-VecCommaInitializer<_Tp, cn>::VecCommaInitializer(Vec<_Tp, cn>* _vec)
-    : MatxCommaInitializer<_Tp, cn, 1>(_vec)
-{}
-
-template<typename _Tp, int cn> template<typename _T2> inline
-VecCommaInitializer<_Tp, cn>& VecCommaInitializer<_Tp, cn>::operator , (_T2 value)
-{
-    CV_DbgAssert( this->idx < cn );
-    this->dst->val[this->idx++] = saturate_cast<_Tp>(value);
-    return *this;
-}
-
-template<typename _Tp, int cn> inline
-Vec<_Tp, cn> VecCommaInitializer<_Tp, cn>::operator *() const
-{
-    CV_DbgAssert( this->idx == cn );
-    return *this->dst;
-}
-
-//////////////////////////////// Complex //////////////////////////////
-
-template<typename _Tp> inline Complex<_Tp>::Complex() : re(0), im(0) {}
-template<typename _Tp> inline Complex<_Tp>::Complex( _Tp _re, _Tp _im ) : re(_re), im(_im) {}
-template<typename _Tp> template<typename T2> inline Complex<_Tp>::operator Complex<T2>() const
-{ return Complex<T2>(saturate_cast<T2>(re), saturate_cast<T2>(im)); }
-template<typename _Tp> inline Complex<_Tp> Complex<_Tp>::conj() const
-{ return Complex<_Tp>(re, -im); }
-
-template<typename _Tp> static inline
-bool operator == (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{ return a.re == b.re && a.im == b.im; }
-
-template<typename _Tp> static inline
-bool operator != (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{ return a.re != b.re || a.im != b.im; }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator + (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{ return Complex<_Tp>( a.re + b.re, a.im + b.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator += (Complex<_Tp>& a, const Complex<_Tp>& b)
-{ a.re += b.re; a.im += b.im; return a; }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator - (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{ return Complex<_Tp>( a.re - b.re, a.im - b.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator -= (Complex<_Tp>& a, const Complex<_Tp>& b)
-{ a.re -= b.re; a.im -= b.im; return a; }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator - (const Complex<_Tp>& a)
-{ return Complex<_Tp>(-a.re, -a.im); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator * (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{ return Complex<_Tp>( a.re*b.re - a.im*b.im, a.re*b.im + a.im*b.re ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator * (const Complex<_Tp>& a, _Tp b)
-{ return Complex<_Tp>( a.re*b, a.im*b ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator * (_Tp b, const Complex<_Tp>& a)
-{ return Complex<_Tp>( a.re*b, a.im*b ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator + (const Complex<_Tp>& a, _Tp b)
-{ return Complex<_Tp>( a.re + b, a.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator - (const Complex<_Tp>& a, _Tp b)
-{ return Complex<_Tp>( a.re - b, a.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator + (_Tp b, const Complex<_Tp>& a)
-{ return Complex<_Tp>( a.re + b, a.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator - (_Tp b, const Complex<_Tp>& a)
-{ return Complex<_Tp>( b - a.re, -a.im ); }
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator += (Complex<_Tp>& a, _Tp b)
-{ a.re += b; return a; }
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator -= (Complex<_Tp>& a, _Tp b)
-{ a.re -= b; return a; }
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator *= (Complex<_Tp>& a, _Tp b)
-{ a.re *= b; a.im *= b; return a; }
-
-template<typename _Tp> static inline
-double abs(const Complex<_Tp>& a)
-{ return std::sqrt( (double)a.re*a.re + (double)a.im*a.im); }
-
-template<typename _Tp> static inline
-Complex<_Tp> operator / (const Complex<_Tp>& a, const Complex<_Tp>& b)
-{
-    double t = 1./((double)b.re*b.re + (double)b.im*b.im);
-    return Complex<_Tp>( (_Tp)((a.re*b.re + a.im*b.im)*t),
-                        (_Tp)((-a.re*b.im + a.im*b.re)*t) );
-}
-
-template<typename _Tp> static inline
-Complex<_Tp>& operator /= (Complex<_Tp>& a, const Complex<_Tp>& b)
-{
-    return (a = a / b);
-}
-
-template<typename _Tp> static inline
-Complex<_Tp> operator / (const Complex<_Tp>& a, _Tp b)
-{
-    _Tp t = (_Tp)1/b;
-    return Complex<_Tp>( a.re*t, a.im*t );
-}
-
-template<typename _Tp> static inline
-Complex<_Tp> operator / (_Tp b, const Complex<_Tp>& a)
-{
-    return Complex<_Tp>(b)/a;
-}
-
-template<typename _Tp> static inline
-Complex<_Tp> operator /= (const Complex<_Tp>& a, _Tp b)
-{
-    _Tp t = (_Tp)1/b;
-    a.re *= t; a.im *= t; return a;
-}
-
-//////////////////////////////// 2D Point ////////////////////////////////
-
-template<typename _Tp> inline Point_<_Tp>::Point_() : x(0), y(0) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(_Tp _x, _Tp _y) : x(_x), y(_y) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(const Point_& pt) : x(pt.x), y(pt.y) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(const CvPoint& pt) : x((_Tp)pt.x), y((_Tp)pt.y) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(const CvPoint2D32f& pt)
-    : x(saturate_cast<_Tp>(pt.x)), y(saturate_cast<_Tp>(pt.y)) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(const Size_<_Tp>& sz) : x(sz.width), y(sz.height) {}
-template<typename _Tp> inline Point_<_Tp>::Point_(const Vec<_Tp,2>& v) : x(v[0]), y(v[1]) {}
-template<typename _Tp> inline Point_<_Tp>& Point_<_Tp>::operator = (const Point_& pt)
-{ x = pt.x; y = pt.y; return *this; }
-
-template<typename _Tp> template<typename _Tp2> inline Point_<_Tp>::operator Point_<_Tp2>() const
-{ return Point_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y)); }
-template<typename _Tp> inline Point_<_Tp>::operator CvPoint() const
-{ return cvPoint(saturate_cast<int>(x), saturate_cast<int>(y)); }
-template<typename _Tp> inline Point_<_Tp>::operator CvPoint2D32f() const
-{ return cvPoint2D32f((float)x, (float)y); }
-template<typename _Tp> inline Point_<_Tp>::operator Vec<_Tp, 2>() const
-{ return Vec<_Tp, 2>(x, y); }
-
-template<typename _Tp> inline _Tp Point_<_Tp>::dot(const Point_& pt) const
-{ return saturate_cast<_Tp>(x*pt.x + y*pt.y); }
-template<typename _Tp> inline double Point_<_Tp>::ddot(const Point_& pt) const
-{ return (double)x*pt.x + (double)y*pt.y; }
-
-template<typename _Tp> inline double Point_<_Tp>::cross(const Point_& pt) const
-{ return (double)x*pt.y - (double)y*pt.x; }
-
-template<typename _Tp> static inline Point_<_Tp>&
-operator += (Point_<_Tp>& a, const Point_<_Tp>& b)
-{
-    a.x = saturate_cast<_Tp>(a.x + b.x);
-    a.y = saturate_cast<_Tp>(a.y + b.y);
-    return a;
-}
-
-template<typename _Tp> static inline Point_<_Tp>&
-operator -= (Point_<_Tp>& a, const Point_<_Tp>& b)
-{
-    a.x = saturate_cast<_Tp>(a.x - b.x);
-    a.y = saturate_cast<_Tp>(a.y - b.y);
-    return a;
-}
-
-template<typename _Tp> static inline Point_<_Tp>&
-operator *= (Point_<_Tp>& a, int b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    return a;
-}
-
-template<typename _Tp> static inline Point_<_Tp>&
-operator *= (Point_<_Tp>& a, float b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    return a;
-}
-
-template<typename _Tp> static inline Point_<_Tp>&
-operator *= (Point_<_Tp>& a, double b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    return a;
-}
 
-template<typename _Tp> static inline double norm(const Point_<_Tp>& pt)
-{ return std::sqrt((double)pt.x*pt.x + (double)pt.y*pt.y); }
 
-template<typename _Tp> static inline bool operator == (const Point_<_Tp>& a, const Point_<_Tp>& b)
-{ return a.x == b.x && a.y == b.y; }
-
-template<typename _Tp> static inline bool operator != (const Point_<_Tp>& a, const Point_<_Tp>& b)
-{ return a.x != b.x || a.y != b.y; }
-
-template<typename _Tp> static inline Point_<_Tp> operator + (const Point_<_Tp>& a, const Point_<_Tp>& b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(a.x + b.x), saturate_cast<_Tp>(a.y + b.y) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator - (const Point_<_Tp>& a, const Point_<_Tp>& b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(a.x - b.x), saturate_cast<_Tp>(a.y - b.y) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator - (const Point_<_Tp>& a)
-{ return Point_<_Tp>( saturate_cast<_Tp>(-a.x), saturate_cast<_Tp>(-a.y) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (const Point_<_Tp>& a, int b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (int a, const Point_<_Tp>& b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (const Point_<_Tp>& a, float b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (float a, const Point_<_Tp>& b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (const Point_<_Tp>& a, double b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) ); }
-
-template<typename _Tp> static inline Point_<_Tp> operator * (double a, const Point_<_Tp>& b)
-{ return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) ); }
-
-//////////////////////////////// 3D Point ////////////////////////////////
-
-template<typename _Tp> inline Point3_<_Tp>::Point3_() : x(0), y(0), z(0) {}
-template<typename _Tp> inline Point3_<_Tp>::Point3_(_Tp _x, _Tp _y, _Tp _z) : x(_x), y(_y), z(_z) {}
-template<typename _Tp> inline Point3_<_Tp>::Point3_(const Point3_& pt) : x(pt.x), y(pt.y), z(pt.z) {}
-template<typename _Tp> inline Point3_<_Tp>::Point3_(const Point_<_Tp>& pt) : x(pt.x), y(pt.y), z(_Tp()) {}
-template<typename _Tp> inline Point3_<_Tp>::Point3_(const CvPoint3D32f& pt) :
-    x(saturate_cast<_Tp>(pt.x)), y(saturate_cast<_Tp>(pt.y)), z(saturate_cast<_Tp>(pt.z)) {}
-template<typename _Tp> inline Point3_<_Tp>::Point3_(const Vec<_Tp, 3>& v) : x(v[0]), y(v[1]), z(v[2]) {}
-
-template<typename _Tp> template<typename _Tp2> inline Point3_<_Tp>::operator Point3_<_Tp2>() const
-{ return Point3_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y), saturate_cast<_Tp2>(z)); }
-
-template<typename _Tp> inline Point3_<_Tp>::operator CvPoint3D32f() const
-{ return cvPoint3D32f((float)x, (float)y, (float)z); }
-
-template<typename _Tp> inline Point3_<_Tp>::operator Vec<_Tp, 3>() const
-{ return Vec<_Tp, 3>(x, y, z); }
-
-template<typename _Tp> inline Point3_<_Tp>& Point3_<_Tp>::operator = (const Point3_& pt)
-{ x = pt.x; y = pt.y; z = pt.z; return *this; }
-
-template<typename _Tp> inline _Tp Point3_<_Tp>::dot(const Point3_& pt) const
-{ return saturate_cast<_Tp>(x*pt.x + y*pt.y + z*pt.z); }
-template<typename _Tp> inline double Point3_<_Tp>::ddot(const Point3_& pt) const
-{ return (double)x*pt.x + (double)y*pt.y + (double)z*pt.z; }
-
-template<typename _Tp> inline Point3_<_Tp> Point3_<_Tp>::cross(const Point3_<_Tp>& pt) const
-{
-    return Point3_<_Tp>(y*pt.z - z*pt.y, z*pt.x - x*pt.z, x*pt.y - y*pt.x);
-}
-
-template<typename _Tp> static inline Point3_<_Tp>&
-operator += (Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{
-    a.x = saturate_cast<_Tp>(a.x + b.x);
-    a.y = saturate_cast<_Tp>(a.y + b.y);
-    a.z = saturate_cast<_Tp>(a.z + b.z);
-    return a;
-}
-
-template<typename _Tp> static inline Point3_<_Tp>&
-operator -= (Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{
-    a.x = saturate_cast<_Tp>(a.x - b.x);
-    a.y = saturate_cast<_Tp>(a.y - b.y);
-    a.z = saturate_cast<_Tp>(a.z - b.z);
-    return a;
-}
-
-template<typename _Tp> static inline Point3_<_Tp>&
-operator *= (Point3_<_Tp>& a, int b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    a.z = saturate_cast<_Tp>(a.z*b);
-    return a;
-}
-
-template<typename _Tp> static inline Point3_<_Tp>&
-operator *= (Point3_<_Tp>& a, float b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    a.z = saturate_cast<_Tp>(a.z*b);
-    return a;
-}
-
-template<typename _Tp> static inline Point3_<_Tp>&
-operator *= (Point3_<_Tp>& a, double b)
-{
-    a.x = saturate_cast<_Tp>(a.x*b);
-    a.y = saturate_cast<_Tp>(a.y*b);
-    a.z = saturate_cast<_Tp>(a.z*b);
-    return a;
-}
-
-template<typename _Tp> static inline double norm(const Point3_<_Tp>& pt)
-{ return std::sqrt((double)pt.x*pt.x + (double)pt.y*pt.y + (double)pt.z*pt.z); }
-
-template<typename _Tp> static inline bool operator == (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{ return a.x == b.x && a.y == b.y && a.z == b.z; }
-
-template<typename _Tp> static inline bool operator != (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{ return a.x != b.x || a.y != b.y || a.z != b.z; }
-
-template<typename _Tp> static inline Point3_<_Tp> operator + (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(a.x + b.x),
-                      saturate_cast<_Tp>(a.y + b.y),
-                      saturate_cast<_Tp>(a.z + b.z)); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator - (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(a.x - b.x),
-                        saturate_cast<_Tp>(a.y - b.y),
-                        saturate_cast<_Tp>(a.z - b.z)); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator - (const Point3_<_Tp>& a)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(-a.x),
-                      saturate_cast<_Tp>(-a.y),
-                      saturate_cast<_Tp>(-a.z) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (const Point3_<_Tp>& a, int b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(a.x*b),
-                      saturate_cast<_Tp>(a.y*b),
-                      saturate_cast<_Tp>(a.z*b) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (int a, const Point3_<_Tp>& b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(b.x*a),
-                      saturate_cast<_Tp>(b.y*a),
-                      saturate_cast<_Tp>(b.z*a) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (const Point3_<_Tp>& a, float b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(a.x*b),
-                      saturate_cast<_Tp>(a.y*b),
-                      saturate_cast<_Tp>(a.z*b) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (float a, const Point3_<_Tp>& b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(b.x*a),
-                      saturate_cast<_Tp>(b.y*a),
-                      saturate_cast<_Tp>(b.z*a) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (const Point3_<_Tp>& a, double b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(a.x*b),
-                      saturate_cast<_Tp>(a.y*b),
-                      saturate_cast<_Tp>(a.z*b) ); }
-
-template<typename _Tp> static inline Point3_<_Tp> operator * (double a, const Point3_<_Tp>& b)
-{ return Point3_<_Tp>( saturate_cast<_Tp>(b.x*a),
-                      saturate_cast<_Tp>(b.y*a),
-                      saturate_cast<_Tp>(b.z*a) ); }
-
-//////////////////////////////// Size ////////////////////////////////
-
-template<typename _Tp> inline Size_<_Tp>::Size_()
-    : width(0), height(0) {}
-template<typename _Tp> inline Size_<_Tp>::Size_(_Tp _width, _Tp _height)
-    : width(_width), height(_height) {}
-template<typename _Tp> inline Size_<_Tp>::Size_(const Size_& sz)
-    : width(sz.width), height(sz.height) {}
-template<typename _Tp> inline Size_<_Tp>::Size_(const CvSize& sz)
-    : width(saturate_cast<_Tp>(sz.width)), height(saturate_cast<_Tp>(sz.height)) {}
-template<typename _Tp> inline Size_<_Tp>::Size_(const CvSize2D32f& sz)
-    : width(saturate_cast<_Tp>(sz.width)), height(saturate_cast<_Tp>(sz.height)) {}
-template<typename _Tp> inline Size_<_Tp>::Size_(const Point_<_Tp>& pt) : width(pt.x), height(pt.y) {}
-
-template<typename _Tp> template<typename _Tp2> inline Size_<_Tp>::operator Size_<_Tp2>() const
-{ return Size_<_Tp2>(saturate_cast<_Tp2>(width), saturate_cast<_Tp2>(height)); }
-template<typename _Tp> inline Size_<_Tp>::operator CvSize() const
-{ return cvSize(saturate_cast<int>(width), saturate_cast<int>(height)); }
-template<typename _Tp> inline Size_<_Tp>::operator CvSize2D32f() const
-{ return cvSize2D32f((float)width, (float)height); }
-
-template<typename _Tp> inline Size_<_Tp>& Size_<_Tp>::operator = (const Size_<_Tp>& sz)
-{ width = sz.width; height = sz.height; return *this; }
-template<typename _Tp> static inline Size_<_Tp> operator * (const Size_<_Tp>& a, _Tp b)
-{ return Size_<_Tp>(a.width * b, a.height * b); }
-template<typename _Tp> static inline Size_<_Tp> operator + (const Size_<_Tp>& a, const Size_<_Tp>& b)
-{ return Size_<_Tp>(a.width + b.width, a.height + b.height); }
-template<typename _Tp> static inline Size_<_Tp> operator - (const Size_<_Tp>& a, const Size_<_Tp>& b)
-{ return Size_<_Tp>(a.width - b.width, a.height - b.height); }
-template<typename _Tp> inline _Tp Size_<_Tp>::area() const { return width*height; }
-
-template<typename _Tp> static inline Size_<_Tp>& operator += (Size_<_Tp>& a, const Size_<_Tp>& b)
-{ a.width += b.width; a.height += b.height; return a; }
-template<typename _Tp> static inline Size_<_Tp>& operator -= (Size_<_Tp>& a, const Size_<_Tp>& b)
-{ a.width -= b.width; a.height -= b.height; return a; }
-
-template<typename _Tp> static inline bool operator == (const Size_<_Tp>& a, const Size_<_Tp>& b)
-{ return a.width == b.width && a.height == b.height; }
-template<typename _Tp> static inline bool operator != (const Size_<_Tp>& a, const Size_<_Tp>& b)
-{ return a.width != b.width || a.height != b.height; }
-
-//////////////////////////////// Rect ////////////////////////////////
-
-
-template<typename _Tp> inline Rect_<_Tp>::Rect_() : x(0), y(0), width(0), height(0) {}
-template<typename _Tp> inline Rect_<_Tp>::Rect_(_Tp _x, _Tp _y, _Tp _width, _Tp _height) : x(_x), y(_y), width(_width), height(_height) {}
-template<typename _Tp> inline Rect_<_Tp>::Rect_(const Rect_<_Tp>& r) : x(r.x), y(r.y), width(r.width), height(r.height) {}
-template<typename _Tp> inline Rect_<_Tp>::Rect_(const CvRect& r) : x((_Tp)r.x), y((_Tp)r.y), width((_Tp)r.width), height((_Tp)r.height) {}
-template<typename _Tp> inline Rect_<_Tp>::Rect_(const Point_<_Tp>& org, const Size_<_Tp>& sz) :
-    x(org.x), y(org.y), width(sz.width), height(sz.height) {}
-template<typename _Tp> inline Rect_<_Tp>::Rect_(const Point_<_Tp>& pt1, const Point_<_Tp>& pt2)
-{
-    x = std::min(pt1.x, pt2.x); y = std::min(pt1.y, pt2.y);
-    width = std::max(pt1.x, pt2.x) - x; height = std::max(pt1.y, pt2.y) - y;
-}
-template<typename _Tp> inline Rect_<_Tp>& Rect_<_Tp>::operator = ( const Rect_<_Tp>& r )
-{ x = r.x; y = r.y; width = r.width; height = r.height; return *this; }
-
-template<typename _Tp> inline Point_<_Tp> Rect_<_Tp>::tl() const { return Point_<_Tp>(x,y); }
-template<typename _Tp> inline Point_<_Tp> Rect_<_Tp>::br() const { return Point_<_Tp>(x+width, y+height); }
-
-template<typename _Tp> static inline Rect_<_Tp>& operator += ( Rect_<_Tp>& a, const Point_<_Tp>& b )
-{ a.x += b.x; a.y += b.y; return a; }
-template<typename _Tp> static inline Rect_<_Tp>& operator -= ( Rect_<_Tp>& a, const Point_<_Tp>& b )
-{ a.x -= b.x; a.y -= b.y; return a; }
-
-template<typename _Tp> static inline Rect_<_Tp>& operator += ( Rect_<_Tp>& a, const Size_<_Tp>& b )
-{ a.width += b.width; a.height += b.height; return a; }
-
-template<typename _Tp> static inline Rect_<_Tp>& operator -= ( Rect_<_Tp>& a, const Size_<_Tp>& b )
-{ a.width -= b.width; a.height -= b.height; return a; }
-
-template<typename _Tp> static inline Rect_<_Tp>& operator &= ( Rect_<_Tp>& a, const Rect_<_Tp>& b )
-{
-    _Tp x1 = std::max(a.x, b.x), y1 = std::max(a.y, b.y);
-    a.width = std::min(a.x + a.width, b.x + b.width) - x1;
-    a.height = std::min(a.y + a.height, b.y + b.height) - y1;
-    a.x = x1; a.y = y1;
-    if( a.width <= 0 || a.height <= 0 )
-        a = Rect();
-    return a;
-}
-
-template<typename _Tp> static inline Rect_<_Tp>& operator |= ( Rect_<_Tp>& a, const Rect_<_Tp>& b )
-{
-    _Tp x1 = std::min(a.x, b.x), y1 = std::min(a.y, b.y);
-    a.width = std::max(a.x + a.width, b.x + b.width) - x1;
-    a.height = std::max(a.y + a.height, b.y + b.height) - y1;
-    a.x = x1; a.y = y1;
-    return a;
-}
-
-template<typename _Tp> inline Size_<_Tp> Rect_<_Tp>::size() const { return Size_<_Tp>(width, height); }
-template<typename _Tp> inline _Tp Rect_<_Tp>::area() const { return width*height; }
-
-template<typename _Tp> template<typename _Tp2> inline Rect_<_Tp>::operator Rect_<_Tp2>() const
-{ return Rect_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y),
-                     saturate_cast<_Tp2>(width), saturate_cast<_Tp2>(height)); }
-template<typename _Tp> inline Rect_<_Tp>::operator CvRect() const
-{ return cvRect(saturate_cast<int>(x), saturate_cast<int>(y),
-                saturate_cast<int>(width), saturate_cast<int>(height)); }
-
-template<typename _Tp> inline bool Rect_<_Tp>::contains(const Point_<_Tp>& pt) const
-{ return x <= pt.x && pt.x < x + width && y <= pt.y && pt.y < y + height; }
-
-template<typename _Tp> static inline bool operator == (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
-{
-    return a.x == b.x && a.y == b.y && a.width == b.width && a.height == b.height;
-}
-
-template<typename _Tp> static inline bool operator != (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
-{
-    return a.x != b.x || a.y != b.y || a.width != b.width || a.height != b.height;
-}
-
-template<typename _Tp> static inline Rect_<_Tp> operator + (const Rect_<_Tp>& a, const Point_<_Tp>& b)
-{
-    return Rect_<_Tp>( a.x + b.x, a.y + b.y, a.width, a.height );
-}
-
-template<typename _Tp> static inline Rect_<_Tp> operator - (const Rect_<_Tp>& a, const Point_<_Tp>& b)
-{
-    return Rect_<_Tp>( a.x - b.x, a.y - b.y, a.width, a.height );
-}
-
-template<typename _Tp> static inline Rect_<_Tp> operator + (const Rect_<_Tp>& a, const Size_<_Tp>& b)
-{
-    return Rect_<_Tp>( a.x, a.y, a.width + b.width, a.height + b.height );
-}
-
-template<typename _Tp> static inline Rect_<_Tp> operator & (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
-{
-    Rect_<_Tp> c = a;
-    return c &= b;
-}
-
-template<typename _Tp> static inline Rect_<_Tp> operator | (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
-{
-    Rect_<_Tp> c = a;
-    return c |= b;
-}
-
-template<typename _Tp> inline bool Point_<_Tp>::inside( const Rect_<_Tp>& r ) const
-{
-    return r.contains(*this);
-}
-
-inline RotatedRect::RotatedRect() { angle = 0; }
-inline RotatedRect::RotatedRect(const Point2f& _center, const Size2f& _size, float _angle)
-    : center(_center), size(_size), angle(_angle) {}
-inline RotatedRect::RotatedRect(const CvBox2D& box)
-    : center(box.center), size(box.size), angle(box.angle) {}
-inline RotatedRect::operator CvBox2D() const
-{
-    CvBox2D box; box.center = center; box.size = size; box.angle = angle;
-    return box;
-}
-
-//////////////////////////////// Scalar_ ///////////////////////////////
-
-template<typename _Tp> inline Scalar_<_Tp>::Scalar_()
-{ this->val[0] = this->val[1] = this->val[2] = this->val[3] = 0; }
-
-template<typename _Tp> inline Scalar_<_Tp>::Scalar_(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
-{ this->val[0] = v0; this->val[1] = v1; this->val[2] = v2; this->val[3] = v3; }
-
-template<typename _Tp> inline Scalar_<_Tp>::Scalar_(const CvScalar& s)
-{
-    this->val[0] = saturate_cast<_Tp>(s.val[0]);
-    this->val[1] = saturate_cast<_Tp>(s.val[1]);
-    this->val[2] = saturate_cast<_Tp>(s.val[2]);
-    this->val[3] = saturate_cast<_Tp>(s.val[3]);
-}
-
-template<typename _Tp> inline Scalar_<_Tp>::Scalar_(_Tp v0)
-{ this->val[0] = v0; this->val[1] = this->val[2] = this->val[3] = 0; }
-
-template<typename _Tp> inline Scalar_<_Tp> Scalar_<_Tp>::all(_Tp v0)
-{ return Scalar_<_Tp>(v0, v0, v0, v0); }
-template<typename _Tp> inline Scalar_<_Tp>::operator CvScalar() const
-{ return cvScalar(this->val[0], this->val[1], this->val[2], this->val[3]); }
-
-template<typename _Tp> template<typename T2> inline Scalar_<_Tp>::operator Scalar_<T2>() const
-{
-    return Scalar_<T2>(saturate_cast<T2>(this->val[0]),
-                  saturate_cast<T2>(this->val[1]),
-                  saturate_cast<T2>(this->val[2]),
-                  saturate_cast<T2>(this->val[3]));
-}
-
-template<typename _Tp> static inline Scalar_<_Tp>& operator += (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    a.val[0] = saturate_cast<_Tp>(a.val[0] + b.val[0]);
-    a.val[1] = saturate_cast<_Tp>(a.val[1] + b.val[1]);
-    a.val[2] = saturate_cast<_Tp>(a.val[2] + b.val[2]);
-    a.val[3] = saturate_cast<_Tp>(a.val[3] + b.val[3]);
-    return a;
-}
-
-template<typename _Tp> static inline Scalar_<_Tp>& operator -= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    a.val[0] = saturate_cast<_Tp>(a.val[0] - b.val[0]);
-    a.val[1] = saturate_cast<_Tp>(a.val[1] - b.val[1]);
-    a.val[2] = saturate_cast<_Tp>(a.val[2] - b.val[2]);
-    a.val[3] = saturate_cast<_Tp>(a.val[3] - b.val[3]);
-    return a;
-}
-
-template<typename _Tp> static inline Scalar_<_Tp>& operator *= ( Scalar_<_Tp>& a, _Tp v )
-{
-    a.val[0] = saturate_cast<_Tp>(a.val[0] * v);
-    a.val[1] = saturate_cast<_Tp>(a.val[1] * v);
-    a.val[2] = saturate_cast<_Tp>(a.val[2] * v);
-    a.val[3] = saturate_cast<_Tp>(a.val[3] * v);
-    return a;
-}
-
-template<typename _Tp> inline Scalar_<_Tp> Scalar_<_Tp>::mul(const Scalar_<_Tp>& t, double scale ) const
-{
-    return Scalar_<_Tp>( saturate_cast<_Tp>(this->val[0]*t.val[0]*scale),
-                       saturate_cast<_Tp>(this->val[1]*t.val[1]*scale),
-                       saturate_cast<_Tp>(this->val[2]*t.val[2]*scale),
-                       saturate_cast<_Tp>(this->val[3]*t.val[3]*scale));
-}
-
-template<typename _Tp> static inline bool operator == ( const Scalar_<_Tp>& a, const Scalar_<_Tp>& b )
-{
-    return a.val[0] == b.val[0] && a.val[1] == b.val[1] &&
-        a.val[2] == b.val[2] && a.val[3] == b.val[3];
-}
-
-template<typename _Tp> static inline bool operator != ( const Scalar_<_Tp>& a, const Scalar_<_Tp>& b )
-{
-    return a.val[0] != b.val[0] || a.val[1] != b.val[1] ||
-        a.val[2] != b.val[2] || a.val[3] != b.val[3];
-}
-
-template<typename _Tp> static inline Scalar_<_Tp> operator + (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(a.val[0] + b.val[0]),
-                      saturate_cast<_Tp>(a.val[1] + b.val[1]),
-                      saturate_cast<_Tp>(a.val[2] + b.val[2]),
-                      saturate_cast<_Tp>(a.val[3] + b.val[3]));
-}
-
-template<typename _Tp> static inline Scalar_<_Tp> operator - (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(a.val[0] - b.val[0]),
-                      saturate_cast<_Tp>(a.val[1] - b.val[1]),
-                      saturate_cast<_Tp>(a.val[2] - b.val[2]),
-                      saturate_cast<_Tp>(a.val[3] - b.val[3]));
-}
-
-template<typename _Tp> static inline Scalar_<_Tp> operator * (const Scalar_<_Tp>& a, _Tp alpha)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(a.val[0] * alpha),
-                      saturate_cast<_Tp>(a.val[1] * alpha),
-                      saturate_cast<_Tp>(a.val[2] * alpha),
-                      saturate_cast<_Tp>(a.val[3] * alpha));
-}
-
-template<typename _Tp> static inline Scalar_<_Tp> operator * (_Tp alpha, const Scalar_<_Tp>& a)
-{
-    return a*alpha;
-}
-
-template<typename _Tp> static inline Scalar_<_Tp> operator - (const Scalar_<_Tp>& a)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(-a.val[0]), saturate_cast<_Tp>(-a.val[1]),
-                      saturate_cast<_Tp>(-a.val[2]), saturate_cast<_Tp>(-a.val[3]));
-}
-
-
-template<typename _Tp> static inline Scalar_<_Tp>
-operator * (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(a[0]*b[0] - a[1]*b[1] - a[2]*b[2] - a[3]*b[3]),
-                        saturate_cast<_Tp>(a[0]*b[1] + a[1]*b[0] + a[2]*b[3] - a[3]*b[2]),
-                        saturate_cast<_Tp>(a[0]*b[2] - a[1]*b[3] + a[2]*b[0] + a[3]*b[1]),
-                        saturate_cast<_Tp>(a[0]*b[3] + a[1]*b[2] - a[2]*b[1] + a[3]*b[0]));
-}
-
-template<typename _Tp> static inline Scalar_<_Tp>&
-operator *= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    a = a*b;
-    return a;
-}
-
-template<typename _Tp> inline Scalar_<_Tp> Scalar_<_Tp>::conj() const
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(this->val[0]),
-                        saturate_cast<_Tp>(-this->val[1]),
-                        saturate_cast<_Tp>(-this->val[2]),
-                        saturate_cast<_Tp>(-this->val[3]));
-}
-
-template<typename _Tp> inline bool Scalar_<_Tp>::isReal() const
-{
-    return this->val[1] == 0 && this->val[2] == 0 && this->val[3] == 0;
-}
-
-template<typename _Tp> static inline
-Scalar_<_Tp> operator / (const Scalar_<_Tp>& a, _Tp alpha)
-{
-    return Scalar_<_Tp>(saturate_cast<_Tp>(a.val[0] / alpha),
-                        saturate_cast<_Tp>(a.val[1] / alpha),
-                        saturate_cast<_Tp>(a.val[2] / alpha),
-                        saturate_cast<_Tp>(a.val[3] / alpha));
-}
-
-template<typename _Tp> static inline
-Scalar_<float> operator / (const Scalar_<float>& a, float alpha)
-{
-    float s = 1/alpha;
-    return Scalar_<float>(a.val[0]*s, a.val[1]*s, a.val[2]*s, a.val[3]*s);
-}
-
-template<typename _Tp> static inline
-Scalar_<double> operator / (const Scalar_<double>& a, double alpha)
-{
-    double s = 1/alpha;
-    return Scalar_<double>(a.val[0]*s, a.val[1]*s, a.val[2]*s, a.val[3]*s);
-}
-
-template<typename _Tp> static inline
-Scalar_<_Tp>& operator /= (Scalar_<_Tp>& a, _Tp alpha)
-{
-    a = a/alpha;
-    return a;
-}
-
-template<typename _Tp> static inline
-Scalar_<_Tp> operator / (_Tp a, const Scalar_<_Tp>& b)
-{
-    _Tp s = a/(b[0]*b[0] + b[1]*b[1] + b[2]*b[2] + b[3]*b[3]);
-    return b.conj()*s;
-}
-
-template<typename _Tp> static inline
-Scalar_<_Tp> operator / (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    return a*((_Tp)1/b);
-}
-
-template<typename _Tp> static inline
-Scalar_<_Tp>& operator /= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
-{
-    a = a/b;
-    return a;
-}
-
-//////////////////////////////// Range /////////////////////////////////
-
-inline Range::Range() : start(0), end(0) {}
-inline Range::Range(int _start, int _end) : start(_start), end(_end) {}
-inline Range::Range(const CvSlice& slice) : start(slice.start_index), end(slice.end_index)
-{
-    if( start == 0 && end == CV_WHOLE_SEQ_END_INDEX )
-        *this = Range::all();
-}
-
-inline int Range::size() const { return end - start; }
-inline bool Range::empty() const { return start == end; }
-inline Range Range::all() { return Range(INT_MIN, INT_MAX); }
-
-static inline bool operator == (const Range& r1, const Range& r2)
-{ return r1.start == r2.start && r1.end == r2.end; }
-
-static inline bool operator != (const Range& r1, const Range& r2)
-{ return !(r1 == r2); }
-
-static inline bool operator !(const Range& r)
-{ return r.start == r.end; }
-
-static inline Range operator & (const Range& r1, const Range& r2)
-{
-    Range r(std::max(r1.start, r2.start), std::min(r1.end, r2.end));
-    r.end = std::max(r.end, r.start);
-    return r;
-}
-
-static inline Range& operator &= (Range& r1, const Range& r2)
-{
-    r1 = r1 & r2;
-    return r1;
-}
-
-static inline Range operator + (const Range& r1, int delta)
-{
-    return Range(r1.start + delta, r1.end + delta);
-}
-
-static inline Range operator + (int delta, const Range& r1)
-{
-    return Range(r1.start + delta, r1.end + delta);
-}
-
-static inline Range operator - (const Range& r1, int delta)
-{
-    return r1 + (-delta);
-}
-
-inline Range::operator CvSlice() const
-{ return *this != Range::all() ? cvSlice(start, end) : CV_WHOLE_SEQ; }
-
-
-
-//////////////////////////////// Vector ////////////////////////////////
-
-// template vector class. It is similar to STL's vector,
-// with a few important differences:
-//   1) it can be created on top of user-allocated data w/o copying it
-//   2) vector b = a means copying the header,
-//      not the underlying data (use clone() to make a deep copy)
-template <typename _Tp> class CV_EXPORTS Vector
-{
-public:
-    typedef _Tp value_type;
-    typedef _Tp* iterator;
-    typedef const _Tp* const_iterator;
-    typedef _Tp& reference;
-    typedef const _Tp& const_reference;
-
-    struct CV_EXPORTS Hdr
-    {
-        Hdr() : data(0), datastart(0), refcount(0), size(0), capacity(0) {};
-        _Tp* data;
-        _Tp* datastart;
-        int* refcount;
-        size_t size;
-        size_t capacity;
-    };
-
-    Vector() {}
-    Vector(size_t _size)  { resize(_size); }
-    Vector(size_t _size, const _Tp& val)
-    {
-        resize(_size);
-        for(size_t i = 0; i < _size; i++)
-            hdr.data[i] = val;
-    }
-    Vector(_Tp* _data, size_t _size, bool _copyData=false)
-    { set(_data, _size, _copyData); }
-
-    template<int n> Vector(const Vec<_Tp, n>& vec)
-    { set((_Tp*)&vec.val[0], n, true); }
-
-    Vector(const std::vector<_Tp>& vec, bool _copyData=false)
-    { set(!vec.empty() ? (_Tp*)&vec[0] : 0, vec.size(), _copyData); }
-
-    Vector(const Vector& d) { *this = d; }
-
-    Vector(const Vector& d, const Range& r_)
-    {
-        Range r = r_ == Range::all() ? Range(0, d.size()) : r_;
-        /*if( r == Range::all() )
-            r = Range(0, d.size());*/
-        if( r.size() > 0 && r.start >= 0 && r.end <= d.size() )
-        {
-            if( d.hdr.refcount )
-                CV_XADD(d.hdr.refcount, 1);
-            hdr.refcount = d.hdr.refcount;
-            hdr.datastart = d.hdr.datastart;
-            hdr.data = d.hdr.data + r.start;
-            hdr.capacity = hdr.size = r.size();
-        }
-    }
-
-    Vector<_Tp>& operator = (const Vector& d)
-    {
-        if( this != &d )
-        {
-            if( d.hdr.refcount )
-                CV_XADD(d.hdr.refcount, 1);
-            release();
-            hdr = d.hdr;
-        }
-        return *this;
-    }
-
-    ~Vector()  { release(); }
-
-    Vector<_Tp> clone() const
-    { return hdr.data ? Vector<_Tp>(hdr.data, hdr.size, true) : Vector<_Tp>(); }
-
-    void copyTo(Vector<_Tp>& vec) const
-    {
-        size_t i, sz = size();
-        vec.resize(sz);
-        const _Tp* src = hdr.data;
-        _Tp* dst = vec.hdr.data;
-        for( i = 0; i < sz; i++ )
-            dst[i] = src[i];
-    }
-
-    void copyTo(std::vector<_Tp>& vec) const
-    {
-        size_t i, sz = size();
-        vec.resize(sz);
-        const _Tp* src = hdr.data;
-        _Tp* dst = sz ? &vec[0] : 0;
-        for( i = 0; i < sz; i++ )
-            dst[i] = src[i];
-    }
-
-    operator CvMat() const
-    { return cvMat((int)size(), 1, type(), (void*)hdr.data); }
-
-    _Tp& operator [] (size_t i) { CV_DbgAssert( i < size() ); return hdr.data[i]; }
-    const _Tp& operator [] (size_t i) const { CV_DbgAssert( i < size() ); return hdr.data[i]; }
-    Vector operator() (const Range& r) const { return Vector(*this, r); }
-    _Tp& back() { CV_DbgAssert(!empty()); return hdr.data[hdr.size-1]; }
-    const _Tp& back() const { CV_DbgAssert(!empty()); return hdr.data[hdr.size-1]; }
-    _Tp& front() { CV_DbgAssert(!empty()); return hdr.data[0]; }
-    const _Tp& front() const { CV_DbgAssert(!empty()); return hdr.data[0]; }
-
-    _Tp* begin() { return hdr.data; }
-    _Tp* end() { return hdr.data + hdr.size; }
-    const _Tp* begin() const { return hdr.data; }
-    const _Tp* end() const { return hdr.data + hdr.size; }
-
-    void addref() { if( hdr.refcount ) CV_XADD(hdr.refcount, 1); }
-    void release()
-    {
-        if( hdr.refcount && CV_XADD(hdr.refcount, -1) == 1 )
-        {
-            delete[] hdr.datastart;
-            delete hdr.refcount;
-        }
-        hdr = Hdr();
-    }
-
-    void set(_Tp* _data, size_t _size, bool _copyData=false)
-    {
-        if( !_copyData )
-        {
-            release();
-            hdr.data = hdr.datastart = _data;
-            hdr.size = hdr.capacity = _size;
-            hdr.refcount = 0;
-        }
-        else
-        {
-            reserve(_size);
-            for( size_t i = 0; i < _size; i++ )
-                hdr.data[i] = _data[i];
-            hdr.size = _size;
-        }
-    }
-
-    void reserve(size_t newCapacity)
-    {
-        _Tp* newData;
-        int* newRefcount;
-        size_t i, oldSize = hdr.size;
-        if( (!hdr.refcount || *hdr.refcount == 1) && hdr.capacity >= newCapacity )
-            return;
-        newCapacity = std::max(newCapacity, oldSize);
-        newData = new _Tp[newCapacity];
-        newRefcount = new int(1);
-        for( i = 0; i < oldSize; i++ )
-            newData[i] = hdr.data[i];
-        release();
-        hdr.data = hdr.datastart = newData;
-        hdr.capacity = newCapacity;
-        hdr.size = oldSize;
-        hdr.refcount = newRefcount;
-    }
-
-    void resize(size_t newSize)
-    {
-        size_t i;
-        newSize = std::max(newSize, (size_t)0);
-        if( (!hdr.refcount || *hdr.refcount == 1) && hdr.size == newSize )
-            return;
-        if( newSize > hdr.capacity )
-            reserve(std::max(newSize, std::max((size_t)4, hdr.capacity*2)));
-        for( i = hdr.size; i < newSize; i++ )
-            hdr.data[i] = _Tp();
-        hdr.size = newSize;
-    }
-
-    Vector<_Tp>& push_back(const _Tp& elem)
-    {
-        if( hdr.size == hdr.capacity )
-            reserve( std::max((size_t)4, hdr.capacity*2) );
-        hdr.data[hdr.size++] = elem;
-        return *this;
-    }
-
-    Vector<_Tp>& pop_back()
-    {
-        if( hdr.size > 0 )
-            --hdr.size;
-        return *this;
-    }
-
-    size_t size() const { return hdr.size; }
-    size_t capacity() const { return hdr.capacity; }
-    bool empty() const { return hdr.size == 0; }
-    void clear() { resize(0); }
-    int type() const { return DataType<_Tp>::type; }
-
-protected:
-    Hdr hdr;
-};
-
-
-template<typename _Tp> inline typename DataType<_Tp>::work_type
-dot(const Vector<_Tp>& v1, const Vector<_Tp>& v2)
-{
-    typedef typename DataType<_Tp>::work_type _Tw;
-    size_t i = 0, n = v1.size();
-    assert(v1.size() == v2.size());
-
-    _Tw s = 0;
-    const _Tp *ptr1 = &v1[0], *ptr2 = &v2[0];
-    for( ; i < n; i++ )
-        s += (_Tw)ptr1[i]*ptr2[i];
-
-    return s;
-}
+/////////////////////////////////// Multiply-with-Carry RNG ///////////////////////////////////
 
-// Multiply-with-Carry RNG
-inline RNG::RNG() { state = 0xffffffff; }
+inline RNG::RNG()              { state = 0xffffffff; }
 inline RNG::RNG(uint64 _state) { state = _state ? _state : 0xffffffff; }
+
+inline RNG::operator uchar()    { return (uchar)next(); }
+inline RNG::operator schar()    { return (schar)next(); }
+inline RNG::operator ushort()   { return (ushort)next(); }
+inline RNG::operator short()    { return (short)next(); }
+inline RNG::operator int()      { return (int)next(); }
+inline RNG::operator unsigned() { return next(); }
+inline RNG::operator float()    { return next()*2.3283064365386962890625e-10f; }
+inline RNG::operator double()   { unsigned t = next(); return (((uint64)t << 32) | next()) * 5.4210108624275221700372640043497e-20; }
+
+inline unsigned RNG::operator ()(unsigned N) { return (unsigned)uniform(0,N); }
+inline unsigned RNG::operator ()()           { return next(); }
+
+inline int    RNG::uniform(int a, int b)       { return a == b ? a : (int)(next() % (b - a) + a); }
+inline float  RNG::uniform(float a, float b)   { return ((float)*this)*(b - a) + a; }
+inline double RNG::uniform(double a, double b) { return ((double)*this)*(b - a) + a; }
+
 inline unsigned RNG::next()
 {
-    state = (uint64)(unsigned)state*CV_RNG_COEFF + (unsigned)(state >> 32);
+    state = (uint64)(unsigned)state* /*CV_RNG_COEFF*/ 4164903690U + (unsigned)(state >> 32);
     return (unsigned)state;
 }
 
-inline RNG::operator uchar() { return (uchar)next(); }
-inline RNG::operator schar() { return (schar)next(); }
-inline RNG::operator ushort() { return (ushort)next(); }
-inline RNG::operator short() { return (short)next(); }
-inline RNG::operator unsigned() { return next(); }
-inline unsigned RNG::operator ()(unsigned N) {return (unsigned)uniform(0,N);}
-inline unsigned RNG::operator ()() {return next();}
-inline RNG::operator int() { return (int)next(); }
-// * (2^32-1)^-1
-inline RNG::operator float() { return next()*2.3283064365386962890625e-10f; }
-inline RNG::operator double()
+
+
+///////////////////////////////////////// LineIterator ////////////////////////////////////////
+
+inline
+uchar* LineIterator::operator *()
 {
-    unsigned t = next();
-    return (((uint64)t << 32) | next())*5.4210108624275221700372640043497e-20;
+    return ptr;
 }
-inline int RNG::uniform(int a, int b) { return a == b ? a : (int)(next()%(b - a) + a); }
-inline float RNG::uniform(float a, float b) { return ((float)*this)*(b - a) + a; }
-inline double RNG::uniform(double a, double b) { return ((double)*this)*(b - a) + a; }
 
-inline TermCriteria::TermCriteria() : type(0), maxCount(0), epsilon(0) {}
-inline TermCriteria::TermCriteria(int _type, int _maxCount, double _epsilon)
-    : type(_type), maxCount(_maxCount), epsilon(_epsilon) {}
-inline TermCriteria::TermCriteria(const CvTermCriteria& criteria)
-    : type(criteria.type), maxCount(criteria.max_iter), epsilon(criteria.epsilon) {}
-inline TermCriteria::operator CvTermCriteria() const
-{ return cvTermCriteria(type, maxCount, epsilon); }
-
-inline uchar* LineIterator::operator *() { return ptr; }
-inline LineIterator& LineIterator::operator ++()
+inline
+LineIterator& LineIterator::operator ++()
 {
     int mask = err < 0 ? -1 : 0;
     err += minusDelta + (plusDelta & mask);
     ptr += minusStep + (plusStep & mask);
     return *this;
 }
-inline LineIterator LineIterator::operator ++(int)
+
+inline
+LineIterator LineIterator::operator ++(int)
 {
     LineIterator it = *this;
     ++(*this);
     return it;
 }
-inline Point LineIterator::pos() const
+
+inline
+Point LineIterator::pos() const
 {
     Point p;
     p.y = (int)((ptr - ptr0)/step);
@@ -2482,1525 +386,138 @@ inline Point LineIterator::pos() const
     return p;
 }
 
-/////////////////////////////// AutoBuffer ////////////////////////////////////////
 
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::AutoBuffer()
+//! returns the next unifomly-distributed random number of the specified type
+template<typename _Tp> static inline _Tp randu()
 {
-    ptr = buf;
-    sz = fixed_size;
+  return (_Tp)theRNG();
 }
 
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::AutoBuffer(size_t _size)
+
+
+///////////////////////////////// Formatted output of cv::Mat /////////////////////////////////
+
+static inline
+Ptr<Formatted> format(InputArray mtx, int fmt)
 {
-    ptr = buf;
-    sz = fixed_size;
-    allocate(_size);
+    return Formatter::get(fmt)->format(mtx.getMat());
 }
 
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::AutoBuffer(const AutoBuffer<_Tp, fixed_size>& abuf )
+static inline
+int print(Ptr<Formatted> fmtd, FILE* stream = stdout)
 {
-    ptr = buf;
-    sz = fixed_size;
-    allocate(abuf.size);
-    for( size_t i = 0; i < sz; i++ )
-        ptr[i] = abuf.ptr[i];
+    int written = 0;
+    fmtd->reset();
+    for(const char* str = fmtd->next(); str; str = fmtd->next())
+        written += fputs(str, stream);
+
+    return written;
 }
 
-template<typename _Tp, size_t fixed_size> inline AutoBuffer<_Tp, fixed_size>&
-AutoBuffer<_Tp, fixed_size>::operator = (const AutoBuffer<_Tp, fixed_size>& abuf)
+static inline
+int print(const Mat& mtx, FILE* stream = stdout)
 {
-    if( this != &abuf )
-    {
-        deallocate();
-        allocate(abuf.size);
-        for( size_t i = 0; i < sz; i++ )
-            ptr[i] = abuf.ptr[i];
-    }
-    return *this;
+    return print(Formatter::get()->format(mtx), stream);
 }
 
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::~AutoBuffer()
-{ deallocate(); }
-
-template<typename _Tp, size_t fixed_size> inline void
-AutoBuffer<_Tp, fixed_size>::allocate(size_t _size)
-{
-    if(_size <= sz)
-    {
-        sz = _size;
-        return;
-    }
-    deallocate();
-    if(_size > fixed_size)
-    {
-        ptr = new _Tp[_size];
-        sz = _size;
-    }
-}
-
-template<typename _Tp, size_t fixed_size> inline void
-AutoBuffer<_Tp, fixed_size>::deallocate()
-{
-    if( ptr != buf )
-    {
-        delete[] ptr;
-        ptr = buf;
-        sz = fixed_size;
-    }
-}
-
-template<typename _Tp, size_t fixed_size> inline void
-AutoBuffer<_Tp, fixed_size>::resize(size_t _size)
-{
-    if(_size <= sz)
-    {
-        sz = _size;
-        return;
-    }
-    size_t i, prevsize = sz, minsize = MIN(prevsize, _size);
-    _Tp* prevptr = ptr;
-    
-    ptr = _size > fixed_size ? new _Tp[_size] : buf;
-    sz = _size;
-    
-    if( ptr != prevptr )
-        for( i = 0; i < minsize; i++ )
-            ptr[i] = prevptr[i];
-    for( i = prevsize; i < _size; i++ )
-        ptr[i] = _Tp();
-    
-    if( prevptr != buf )
-        delete[] prevptr;
-}
-
-template<typename _Tp, size_t fixed_size> inline size_t
-AutoBuffer<_Tp, fixed_size>::size() const
-{ return sz; }
-
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::operator _Tp* ()
-{ return ptr; }
-
-template<typename _Tp, size_t fixed_size> inline
-AutoBuffer<_Tp, fixed_size>::operator const _Tp* () const
-{ return ptr; }
-
-
-/////////////////////////////////// Ptr ////////////////////////////////////////
-
-template<typename _Tp> inline Ptr<_Tp>::Ptr() : obj(0), refcount(0) {}
-template<typename _Tp> inline Ptr<_Tp>::Ptr(_Tp* _obj) : obj(_obj)
-{
-    if(obj)
-    {
-        refcount = (int*)fastMalloc(sizeof(*refcount));
-        *refcount = 1;
-    }
-    else
-        refcount = 0;
-}
-
-template<typename _Tp> inline void Ptr<_Tp>::addref()
-{ if( refcount ) CV_XADD(refcount, 1); }
-
-template<typename _Tp> inline void Ptr<_Tp>::release()
-{
-    if( refcount && CV_XADD(refcount, -1) == 1 )
-    {
-        delete_obj();
-        fastFree(refcount);
-    }
-    refcount = 0;
-    obj = 0;
-}
-
-template<typename _Tp> inline void Ptr<_Tp>::delete_obj()
-{
-    if( obj ) delete obj;
-}
-
-template<typename _Tp> inline Ptr<_Tp>::~Ptr() { release(); }
-
-template<typename _Tp> inline Ptr<_Tp>::Ptr(const Ptr<_Tp>& _ptr)
-{
-    obj = _ptr.obj;
-    refcount = _ptr.refcount;
-    addref();
-}
-
-template<typename _Tp> inline Ptr<_Tp>& Ptr<_Tp>::operator = (const Ptr<_Tp>& _ptr)
-{
-    int* _refcount = _ptr.refcount;
-    if( _refcount )
-        CV_XADD(_refcount, 1);
-    release();
-    obj = _ptr.obj;
-    refcount = _refcount;
-    return *this;
-}
-
-template<typename _Tp> inline _Tp* Ptr<_Tp>::operator -> () { return obj; }
-template<typename _Tp> inline const _Tp* Ptr<_Tp>::operator -> () const { return obj; }
-
-template<typename _Tp> inline Ptr<_Tp>::operator _Tp* () { return obj; }
-template<typename _Tp> inline Ptr<_Tp>::operator const _Tp*() const { return obj; }
-
-template<typename _Tp> inline bool Ptr<_Tp>::empty() const { return obj == 0; }
-
-template<typename _Tp> template<typename _Tp2> Ptr<_Tp>::Ptr(const Ptr<_Tp2>& p)
-    : obj(0), refcount(0)
-{
-    if (p.empty())
-        return;
-
-    _Tp* p_casted = dynamic_cast<_Tp*>(p.obj);
-    if (!p_casted)
-        return;
-
-    obj = p_casted;
-    refcount = p.refcount;
-    addref();
-}
-
-template<typename _Tp> template<typename _Tp2> inline Ptr<_Tp2> Ptr<_Tp>::ptr()
-{
-    Ptr<_Tp2> p;
-    if( !obj )
-        return p;
-
-    _Tp2* obj_casted = dynamic_cast<_Tp2*>(obj);
-    if (!obj_casted)
-        return p;
-
-    if( refcount )
-        CV_XADD(refcount, 1);
-
-    p.obj = obj_casted;
-    p.refcount = refcount;
-    return p;
-}
-
-template<typename _Tp> template<typename _Tp2> inline const Ptr<_Tp2> Ptr<_Tp>::ptr() const
-{
-    Ptr<_Tp2> p;
-    if( !obj )
-        return p;
-
-    _Tp2* obj_casted = dynamic_cast<_Tp2*>(obj);
-    if (!obj_casted)
-        return p;
-
-    if( refcount )
-        CV_XADD(refcount, 1);
-
-    p.obj = obj_casted;
-    p.refcount = refcount;
-    return p;
-}
-
-template<class _Tp, class _Tp2> inline bool operator==(const Ptr<_Tp>& a, const Ptr<_Tp2>& b) { return a.refcount == b.refcount; }
-template<class _Tp, class _Tp2> inline bool operator!=(const Ptr<_Tp>& a, const Ptr<_Tp2>& b) { return a.refcount != b.refcount; }
-
-
-
-
-//// specializied implementations of Ptr::delete_obj() for classic OpenCV types
-
-template<> CV_EXPORTS void Ptr<CvMat>::delete_obj();
-template<> CV_EXPORTS void Ptr<IplImage>::delete_obj();
-template<> CV_EXPORTS void Ptr<CvMatND>::delete_obj();
-template<> CV_EXPORTS void Ptr<CvSparseMat>::delete_obj();
-template<> CV_EXPORTS void Ptr<CvMemStorage>::delete_obj();
-template<> CV_EXPORTS void Ptr<CvFileStorage>::delete_obj();
-
-//////////////////////////////////////// XML & YAML I/O ////////////////////////////////////
-
-CV_EXPORTS_W void write( FileStorage& fs, const std::string& name, int value );
-CV_EXPORTS_W void write( FileStorage& fs, const std::string& name, float value );
-CV_EXPORTS_W void write( FileStorage& fs, const std::string& name, double value );
-CV_EXPORTS_W void write( FileStorage& fs, const std::string& name, const std::string& value );
-
-template<typename _Tp> inline void write(FileStorage& fs, const _Tp& value)
-{ write(fs, std::string(), value); }
-
-CV_EXPORTS void writeScalar( FileStorage& fs, int value );
-CV_EXPORTS void writeScalar( FileStorage& fs, float value );
-CV_EXPORTS void writeScalar( FileStorage& fs, double value );
-CV_EXPORTS void writeScalar( FileStorage& fs, const std::string& value );
-
-template<> inline void write( FileStorage& fs, const int& value )
-{
-    writeScalar(fs, value);
-}
-
-template<> inline void write( FileStorage& fs, const float& value )
-{
-    writeScalar(fs, value);
-}
-
-template<> inline void write( FileStorage& fs, const double& value )
-{
-    writeScalar(fs, value);
-}
-
-template<> inline void write( FileStorage& fs, const std::string& value )
-{
-    writeScalar(fs, value);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Point_<_Tp>& pt )
-{
-    write(fs, pt.x);
-    write(fs, pt.y);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Point3_<_Tp>& pt )
-{
-    write(fs, pt.x);
-    write(fs, pt.y);
-    write(fs, pt.z);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Size_<_Tp>& sz )
-{
-    write(fs, sz.width);
-    write(fs, sz.height);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Complex<_Tp>& c )
-{
-    write(fs, c.re);
-    write(fs, c.im);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Rect_<_Tp>& r )
-{
-    write(fs, r.x);
-    write(fs, r.y);
-    write(fs, r.width);
-    write(fs, r.height);
-}
-
-template<typename _Tp, int cn> inline void write(FileStorage& fs, const Vec<_Tp, cn>& v )
-{
-    for(int i = 0; i < cn; i++)
-        write(fs, v.val[i]);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const Scalar_<_Tp>& s )
-{
-    write(fs, s.val[0]);
-    write(fs, s.val[1]);
-    write(fs, s.val[2]);
-    write(fs, s.val[3]);
-}
-
-inline void write(FileStorage& fs, const Range& r )
-{
-    write(fs, r.start);
-    write(fs, r.end);
-}
-
-class CV_EXPORTS WriteStructContext
-{
-public:
-    WriteStructContext(FileStorage& _fs, const std::string& name,
-        int flags, const std::string& typeName=std::string());
-    ~WriteStructContext();
-    FileStorage* fs;
-};
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Point_<_Tp>& pt )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, pt.x);
-    write(fs, pt.y);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Point3_<_Tp>& pt )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, pt.x);
-    write(fs, pt.y);
-    write(fs, pt.z);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Size_<_Tp>& sz )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, sz.width);
-    write(fs, sz.height);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Complex<_Tp>& c )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, c.re);
-    write(fs, c.im);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Rect_<_Tp>& r )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, r.x);
-    write(fs, r.y);
-    write(fs, r.width);
-    write(fs, r.height);
-}
-
-template<typename _Tp, int cn> inline void write(FileStorage& fs, const std::string& name, const Vec<_Tp, cn>& v )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    for(int i = 0; i < cn; i++)
-        write(fs, v.val[i]);
-}
-
-template<typename _Tp> inline void write(FileStorage& fs, const std::string& name, const Scalar_<_Tp>& s )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, s.val[0]);
-    write(fs, s.val[1]);
-    write(fs, s.val[2]);
-    write(fs, s.val[3]);
-}
-
-inline void write(FileStorage& fs, const std::string& name, const Range& r )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+CV_NODE_FLOW);
-    write(fs, r.start);
-    write(fs, r.end);
-}
-
-template<typename _Tp, int numflag> class CV_EXPORTS VecWriterProxy
-{
-public:
-    VecWriterProxy( FileStorage* _fs ) : fs(_fs) {}
-    void operator()(const std::vector<_Tp>& vec) const
-    {
-        size_t i, count = vec.size();
-        for( i = 0; i < count; i++ )
-            write( *fs, vec[i] );
-    }
-    FileStorage* fs;
-};
-
-template<typename _Tp> class CV_EXPORTS VecWriterProxy<_Tp,1>
-{
-public:
-    VecWriterProxy( FileStorage* _fs ) : fs(_fs) {}
-    void operator()(const std::vector<_Tp>& vec) const
-    {
-        int _fmt = DataType<_Tp>::fmt;
-        char fmt[] = { (char)((_fmt>>8)+'1'), (char)_fmt, '\0' };
-        fs->writeRaw( std::string(fmt), !vec.empty() ? (uchar*)&vec[0] : 0, vec.size()*sizeof(_Tp) );
-    }
-    FileStorage* fs;
-};
-
-template<typename _Tp> static inline void write( FileStorage& fs, const std::vector<_Tp>& vec )
-{
-    VecWriterProxy<_Tp, DataType<_Tp>::fmt != 0> w(&fs);
-    w(vec);
-}
-
-template<typename _Tp> static inline void write( FileStorage& fs, const std::string& name,
-                                                const std::vector<_Tp>& vec )
-{
-    WriteStructContext ws(fs, name, CV_NODE_SEQ+(DataType<_Tp>::fmt != 0 ? CV_NODE_FLOW : 0));
-    write(fs, vec);
-}
-
-CV_EXPORTS_W void write( FileStorage& fs, const std::string& name, const Mat& value );
-CV_EXPORTS void write( FileStorage& fs, const std::string& name, const SparseMat& value );
-
-template<typename _Tp> static inline FileStorage& operator << (FileStorage& fs, const _Tp& value)
-{
-    if( !fs.isOpened() )
-        return fs;
-    if( fs.state == FileStorage::NAME_EXPECTED + FileStorage::INSIDE_MAP )
-        CV_Error( CV_StsError, "No element name has been given" );
-    write( fs, fs.elname, value );
-    if( fs.state & FileStorage::INSIDE_MAP )
-        fs.state = FileStorage::NAME_EXPECTED + FileStorage::INSIDE_MAP;
-    return fs;
-}
-
-CV_EXPORTS FileStorage& operator << (FileStorage& fs, const std::string& str);
-
-static inline FileStorage& operator << (FileStorage& fs, const char* str)
-{ return (fs << std::string(str)); }
-
-inline FileNode::FileNode() : fs(0), node(0) {}
-inline FileNode::FileNode(const CvFileStorage* _fs, const CvFileNode* _node)
-    : fs(_fs), node(_node) {}
-
-inline FileNode::FileNode(const FileNode& _node) : fs(_node.fs), node(_node.node) {}
-
-inline int FileNode::type() const { return !node ? NONE : (node->tag & TYPE_MASK); }
-inline bool FileNode::empty() const { return node == 0; }
-inline bool FileNode::isNone() const { return type() == NONE; }
-inline bool FileNode::isSeq() const { return type() == SEQ; }
-inline bool FileNode::isMap() const { return type() == MAP; }
-inline bool FileNode::isInt() const { return type() == INT; }
-inline bool FileNode::isReal() const { return type() == REAL; }
-inline bool FileNode::isString() const { return type() == STR; }
-inline bool FileNode::isNamed() const { return !node ? false : (node->tag & NAMED) != 0; }
-inline size_t FileNode::size() const
-{
-    int t = type();
-    return t == MAP ? (size_t)((CvSet*)node->data.map)->active_count :
-        t == SEQ ? (size_t)node->data.seq->total : (size_t)!isNone();
-}
-
-inline CvFileNode* FileNode::operator *() { return (CvFileNode*)node; }
-inline const CvFileNode* FileNode::operator* () const { return node; }
-
-static inline void read(const FileNode& node, int& value, int default_value)
-{
-    value = !node.node ? default_value :
-    CV_NODE_IS_INT(node.node->tag) ? node.node->data.i :
-    CV_NODE_IS_REAL(node.node->tag) ? cvRound(node.node->data.f) : 0x7fffffff;
-}
-
-static inline void read(const FileNode& node, bool& value, bool default_value)
-{
-    int temp; read(node, temp, (int)default_value);
-    value = temp != 0;
-}
-
-static inline void read(const FileNode& node, uchar& value, uchar default_value)
-{
-    int temp; read(node, temp, (int)default_value);
-    value = saturate_cast<uchar>(temp);
-}
-
-static inline void read(const FileNode& node, schar& value, schar default_value)
-{
-    int temp; read(node, temp, (int)default_value);
-    value = saturate_cast<schar>(temp);
-}
-
-static inline void read(const FileNode& node, ushort& value, ushort default_value)
-{
-    int temp; read(node, temp, (int)default_value);
-    value = saturate_cast<ushort>(temp);
-}
-
-static inline void read(const FileNode& node, short& value, short default_value)
-{
-    int temp; read(node, temp, (int)default_value);
-    value = saturate_cast<short>(temp);
-}
-
-static inline void read(const FileNode& node, float& value, float default_value)
-{
-    value = !node.node ? default_value :
-        CV_NODE_IS_INT(node.node->tag) ? (float)node.node->data.i :
-        CV_NODE_IS_REAL(node.node->tag) ? (float)node.node->data.f : 1e30f;
-}
-
-static inline void read(const FileNode& node, double& value, double default_value)
-{
-    value = !node.node ? default_value :
-        CV_NODE_IS_INT(node.node->tag) ? (double)node.node->data.i :
-        CV_NODE_IS_REAL(node.node->tag) ? node.node->data.f : 1e300;
-}
-
-static inline void read(const FileNode& node, std::string& value, const std::string& default_value)
-{
-    value = !node.node ? default_value : CV_NODE_IS_STRING(node.node->tag) ? std::string(node.node->data.str.ptr) : std::string("");
-}
-
-CV_EXPORTS_W void read(const FileNode& node, Mat& mat, const Mat& default_mat=Mat() );
-CV_EXPORTS void read(const FileNode& node, SparseMat& mat, const SparseMat& default_mat=SparseMat() );
-
-inline FileNode::operator int() const
-{
-    int value;
-    read(*this, value, 0);
-    return value;
-}
-inline FileNode::operator float() const
-{
-    float value;
-    read(*this, value, 0.f);
-    return value;
-}
-inline FileNode::operator double() const
-{
-    double value;
-    read(*this, value, 0.);
-    return value;
-}
-inline FileNode::operator std::string() const
-{
-    std::string value;
-    read(*this, value, value);
-    return value;
-}
-
-inline void FileNode::readRaw( const std::string& fmt, uchar* vec, size_t len ) const
-{
-    begin().readRaw( fmt, vec, len );
-}
-
-template<typename _Tp, int numflag> class CV_EXPORTS VecReaderProxy
-{
-public:
-    VecReaderProxy( FileNodeIterator* _it ) : it(_it) {}
-    void operator()(std::vector<_Tp>& vec, size_t count) const
-    {
-        count = std::min(count, it->remaining);
-        vec.resize(count);
-        for( size_t i = 0; i < count; i++, ++(*it) )
-            read(**it, vec[i], _Tp());
-    }
-    FileNodeIterator* it;
-};
-
-template<typename _Tp> class CV_EXPORTS VecReaderProxy<_Tp,1>
-{
-public:
-    VecReaderProxy( FileNodeIterator* _it ) : it(_it) {}
-    void operator()(std::vector<_Tp>& vec, size_t count) const
-    {
-        size_t remaining = it->remaining, cn = DataType<_Tp>::channels;
-        int _fmt = DataType<_Tp>::fmt;
-        char fmt[] = { (char)((_fmt>>8)+'1'), (char)_fmt, '\0' };
-        size_t remaining1 = remaining/cn;
-        count = count < remaining1 ? count : remaining1;
-        vec.resize(count);
-        it->readRaw( std::string(fmt), !vec.empty() ? (uchar*)&vec[0] : 0, count*sizeof(_Tp) );
-    }
-    FileNodeIterator* it;
-};
-
-template<typename _Tp> static inline void
-read( FileNodeIterator& it, std::vector<_Tp>& vec, size_t maxCount=(size_t)INT_MAX )
-{
-    VecReaderProxy<_Tp, DataType<_Tp>::fmt != 0> r(&it);
-    r(vec, maxCount);
-}
-
-template<typename _Tp> static inline void
-read( const FileNode& node, std::vector<_Tp>& vec, const std::vector<_Tp>& default_value=std::vector<_Tp>() )
-{
-    if(!node.node)
-        vec = default_value;
-    else
-    {
-        FileNodeIterator it = node.begin();
-        read( it, vec );
-    }
-}
-
-inline FileNodeIterator FileNode::begin() const
-{
-    return FileNodeIterator(fs, node);
-}
-
-inline FileNodeIterator FileNode::end() const
-{
-    return FileNodeIterator(fs, node, size());
-}
-
-inline FileNode FileNodeIterator::operator *() const
-{ return FileNode(fs, (const CvFileNode*)reader.ptr); }
-
-inline FileNode FileNodeIterator::operator ->() const
-{ return FileNode(fs, (const CvFileNode*)reader.ptr); }
-
-template<typename _Tp> static inline FileNodeIterator& operator >> (FileNodeIterator& it, _Tp& value)
-{ read( *it, value, _Tp()); return ++it; }
-
 template<typename _Tp> static inline
-FileNodeIterator& operator >> (FileNodeIterator& it, std::vector<_Tp>& vec)
+int print(const std::vector<Point_<_Tp> >& vec, FILE* stream = stdout)
 {
-    VecReaderProxy<_Tp, DataType<_Tp>::fmt != 0> r(&it);
-    r(vec, (size_t)INT_MAX);
-    return it;
+    return print(Formatter::get()->format(Mat(vec)), stream);
 }
 
-template<typename _Tp> static inline void operator >> (const FileNode& n, _Tp& value)
-{ read( n, value, _Tp()); }
-
-template<typename _Tp> static inline void operator >> (const FileNode& n, std::vector<_Tp>& vec)
-{ FileNodeIterator it = n.begin(); it >> vec; }
-
-static inline bool operator == (const FileNodeIterator& it1, const FileNodeIterator& it2)
+template<typename _Tp> static inline
+int print(const std::vector<Point3_<_Tp> >& vec, FILE* stream = stdout)
 {
-    return it1.fs == it2.fs && it1.container == it2.container &&
-        it1.reader.ptr == it2.reader.ptr && it1.remaining == it2.remaining;
+    return print(Formatter::get()->format(Mat(vec)), stream);
 }
 
-static inline bool operator != (const FileNodeIterator& it1, const FileNodeIterator& it2)
+template<typename _Tp, int m, int n> static inline
+int print(const Matx<_Tp, m, n>& matx, FILE* stream = stdout)
 {
-    return !(it1 == it2);
-}
-
-static inline ptrdiff_t operator - (const FileNodeIterator& it1, const FileNodeIterator& it2)
-{
-    return it2.remaining - it1.remaining;
-}
-
-static inline bool operator < (const FileNodeIterator& it1, const FileNodeIterator& it2)
-{
-    return it1.remaining > it2.remaining;
-}
-
-inline FileNode FileStorage::getFirstTopLevelNode() const
-{
-    FileNode r = root();
-    FileNodeIterator it = r.begin();
-    return it != r.end() ? *it : FileNode();
-}
-
-//////////////////////////////////////// Various algorithms ////////////////////////////////////
-
-template<typename _Tp> static inline _Tp gcd(_Tp a, _Tp b)
-{
-    if( a < b )
-        std::swap(a, b);
-    while( b > 0 )
-    {
-        _Tp r = a % b;
-        a = b;
-        b = r;
-    }
-    return a;
-}
-
-/****************************************************************************************\
-
-  Generic implementation of QuickSort algorithm
-  Use it as: vector<_Tp> a; ... sort(a,<less_than_predictor>);
-
-  The current implementation was derived from *BSD system qsort():
-
-    * Copyright (c) 1992, 1993
-    *  The Regents of the University of California.  All rights reserved.
-    *
-    * Redistribution and use in source and binary forms, with or without
-    * modification, are permitted provided that the following conditions
-    * are met:
-    * 1. Redistributions of source code must retain the above copyright
-    *    notice, this list of conditions and the following disclaimer.
-    * 2. Redistributions in binary form must reproduce the above copyright
-    *    notice, this list of conditions and the following disclaimer in the
-    *    documentation and/or other materials provided with the distribution.
-    * 3. All advertising materials mentioning features or use of this software
-    *    must display the following acknowledgement:
-    *  This product includes software developed by the University of
-    *  California, Berkeley and its contributors.
-    * 4. Neither the name of the University nor the names of its contributors
-    *    may be used to endorse or promote products derived from this software
-    *    without specific prior written permission.
-    *
-    * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
-    * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-    * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
-    * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
-    * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-    * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
-    * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
-    * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
-    * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
-    * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
-    * SUCH DAMAGE.
-
-\****************************************************************************************/
-
-template<typename _Tp, class _LT> void sort( std::vector<_Tp>& vec, _LT LT=_LT() )
-{
-    int isort_thresh = 7;
-    int sp = 0;
-
-    struct
-    {
-        _Tp *lb;
-        _Tp *ub;
-    } stack[48];
-
-    size_t total = vec.size();
-
-    if( total <= 1 )
-        return;
-
-    _Tp* arr = &vec[0];
-    stack[0].lb = arr;
-    stack[0].ub = arr + (total - 1);
-
-    while( sp >= 0 )
-    {
-        _Tp* left = stack[sp].lb;
-        _Tp* right = stack[sp--].ub;
-
-        for(;;)
-        {
-            int i, n = (int)(right - left) + 1, m;
-            _Tp* ptr;
-            _Tp* ptr2;
-
-            if( n <= isort_thresh )
-            {
-            insert_sort:
-                for( ptr = left + 1; ptr <= right; ptr++ )
-                {
-                    for( ptr2 = ptr; ptr2 > left && LT(ptr2[0],ptr2[-1]); ptr2--)
-                        std::swap( ptr2[0], ptr2[-1] );
-                }
-                break;
-            }
-            else
-            {
-                _Tp* left0;
-                _Tp* left1;
-                _Tp* right0;
-                _Tp* right1;
-                _Tp* pivot;
-                _Tp* a;
-                _Tp* b;
-                _Tp* c;
-                int swap_cnt = 0;
-
-                left0 = left;
-                right0 = right;
-                pivot = left + (n/2);
-
-                if( n > 40 )
-                {
-                    int d = n / 8;
-                    a = left, b = left + d, c = left + 2*d;
-                    left = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));
-
-                    a = pivot - d, b = pivot, c = pivot + d;
-                    pivot = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));
-
-                    a = right - 2*d, b = right - d, c = right;
-                    right = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))
-                                      : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));
-                }
-
-                a = left, b = pivot, c = right;
-                pivot = LT(*a, *b) ? (LT(*b, *c) ? b : (LT(*a, *c) ? c : a))
-                                   : (LT(*c, *b) ? b : (LT(*a, *c) ? a : c));
-                if( pivot != left0 )
-                {
-                    std::swap( *pivot, *left0 );
-                    pivot = left0;
-                }
-                left = left1 = left0 + 1;
-                right = right1 = right0;
-
-                for(;;)
-                {
-                    while( left <= right && !LT(*pivot, *left) )
-                    {
-                        if( !LT(*left, *pivot) )
-                        {
-                            if( left > left1 )
-                                std::swap( *left1, *left );
-                            swap_cnt = 1;
-                            left1++;
-                        }
-                        left++;
-                    }
-
-                    while( left <= right && !LT(*right, *pivot) )
-                    {
-                        if( !LT(*pivot, *right) )
-                        {
-                            if( right < right1 )
-                                std::swap( *right1, *right );
-                            swap_cnt = 1;
-                            right1--;
-                        }
-                        right--;
-                    }
-
-                    if( left > right )
-                        break;
-                    std::swap( *left, *right );
-                    swap_cnt = 1;
-                    left++;
-                    right--;
-                }
-
-                if( swap_cnt == 0 )
-                {
-                    left = left0, right = right0;
-                    goto insert_sort;
-                }
-
-                n = std::min( (int)(left1 - left0), (int)(left - left1) );
-                for( i = 0; i < n; i++ )
-                    std::swap( left0[i], left[i-n] );
-
-                n = std::min( (int)(right0 - right1), (int)(right1 - right) );
-                for( i = 0; i < n; i++ )
-                    std::swap( left[i], right0[i-n+1] );
-                n = (int)(left - left1);
-                m = (int)(right1 - right);
-                if( n > 1 )
-                {
-                    if( m > 1 )
-                    {
-                        if( n > m )
-                        {
-                            stack[++sp].lb = left0;
-                            stack[sp].ub = left0 + n - 1;
-                            left = right0 - m + 1, right = right0;
-                        }
-                        else
-                        {
-                            stack[++sp].lb = right0 - m + 1;
-                            stack[sp].ub = right0;
-                            left = left0, right = left0 + n - 1;
-                        }
-                    }
-                    else
-                        left = left0, right = left0 + n - 1;
-                }
-                else if( m > 1 )
-                    left = right0 - m + 1, right = right0;
-                else
-                    break;
-            }
-        }
-    }
-}
-
-template<typename _Tp> class CV_EXPORTS LessThan
-{
-public:
-    bool operator()(const _Tp& a, const _Tp& b) const { return a < b; }
-};
-
-template<typename _Tp> class CV_EXPORTS GreaterEq
-{
-public:
-    bool operator()(const _Tp& a, const _Tp& b) const { return a >= b; }
-};
-
-template<typename _Tp> class CV_EXPORTS LessThanIdx
-{
-public:
-    LessThanIdx( const _Tp* _arr ) : arr(_arr) {}
-    bool operator()(int a, int b) const { return arr[a] < arr[b]; }
-    const _Tp* arr;
-};
-
-template<typename _Tp> class CV_EXPORTS GreaterEqIdx
-{
-public:
-    GreaterEqIdx( const _Tp* _arr ) : arr(_arr) {}
-    bool operator()(int a, int b) const { return arr[a] >= arr[b]; }
-    const _Tp* arr;
-};
-
-
-// This function splits the input sequence or set into one or more equivalence classes and
-// returns the vector of labels - 0-based class indexes for each element.
-// predicate(a,b) returns true if the two sequence elements certainly belong to the same class.
-//
-// The algorithm is described in "Introduction to Algorithms"
-// by Cormen, Leiserson and Rivest, the chapter "Data structures for disjoint sets"
-template<typename _Tp, class _EqPredicate> int
-partition( const std::vector<_Tp>& _vec, std::vector<int>& labels,
-           _EqPredicate predicate=_EqPredicate())
-{
-    int i, j, N = (int)_vec.size();
-    const _Tp* vec = &_vec[0];
-
-    const int PARENT=0;
-    const int RANK=1;
-
-    std::vector<int> _nodes(N*2);
-    int (*nodes)[2] = (int(*)[2])&_nodes[0];
-
-    // The first O(N) pass: create N single-vertex trees
-    for(i = 0; i < N; i++)
-    {
-        nodes[i][PARENT]=-1;
-        nodes[i][RANK] = 0;
-    }
-
-    // The main O(N^2) pass: merge connected components
-    for( i = 0; i < N; i++ )
-    {
-        int root = i;
-
-        // find root
-        while( nodes[root][PARENT] >= 0 )
-            root = nodes[root][PARENT];
-
-        for( j = 0; j < N; j++ )
-        {
-            if( i == j || !predicate(vec[i], vec[j]))
-                continue;
-            int root2 = j;
-
-            while( nodes[root2][PARENT] >= 0 )
-                root2 = nodes[root2][PARENT];
-
-            if( root2 != root )
-            {
-                // unite both trees
-                int rank = nodes[root][RANK], rank2 = nodes[root2][RANK];
-                if( rank > rank2 )
-                    nodes[root2][PARENT] = root;
-                else
-                {
-                    nodes[root][PARENT] = root2;
-                    nodes[root2][RANK] += rank == rank2;
-                    root = root2;
-                }
-                assert( nodes[root][PARENT] < 0 );
-
-                int k = j, parent;
-
-                // compress the path from node2 to root
-                while( (parent = nodes[k][PARENT]) >= 0 )
-                {
-                    nodes[k][PARENT] = root;
-                    k = parent;
-                }
-
-                // compress the path from node to root
-                k = i;
-                while( (parent = nodes[k][PARENT]) >= 0 )
-                {
-                    nodes[k][PARENT] = root;
-                    k = parent;
-                }
-            }
-        }
-    }
-
-    // Final O(N) pass: enumerate classes
-    labels.resize(N);
-    int nclasses = 0;
-
-    for( i = 0; i < N; i++ )
-    {
-        int root = i;
-        while( nodes[root][PARENT] >= 0 )
-            root = nodes[root][PARENT];
-        // re-use the rank as the class label
-        if( nodes[root][RANK] >= 0 )
-            nodes[root][RANK] = ~nclasses++;
-        labels[i] = ~nodes[root][RANK];
-    }
-
-    return nclasses;
+    return print(Formatter::get()->format(matx), stream);
 }
 
 
-//////////////////////////////////////////////////////////////////////////////
 
-// bridge C++ => C Seq API
-CV_EXPORTS schar*  seqPush( CvSeq* seq, const void* element=0);
-CV_EXPORTS schar*  seqPushFront( CvSeq* seq, const void* element=0);
-CV_EXPORTS void  seqPop( CvSeq* seq, void* element=0);
-CV_EXPORTS void  seqPopFront( CvSeq* seq, void* element=0);
-CV_EXPORTS void  seqPopMulti( CvSeq* seq, void* elements,
-                              int count, int in_front=0 );
-CV_EXPORTS void  seqRemove( CvSeq* seq, int index );
-CV_EXPORTS void  clearSeq( CvSeq* seq );
-CV_EXPORTS schar*  getSeqElem( const CvSeq* seq, int index );
-CV_EXPORTS void  seqRemoveSlice( CvSeq* seq, CvSlice slice );
-CV_EXPORTS void  seqInsertSlice( CvSeq* seq, int before_index, const CvArr* from_arr );
+////////////////////////////////////////// Algorithm //////////////////////////////////////////
 
-template<typename _Tp> inline Seq<_Tp>::Seq() : seq(0) {}
-template<typename _Tp> inline Seq<_Tp>::Seq( const CvSeq* _seq ) : seq((CvSeq*)_seq)
+template<typename _Tp> inline
+Ptr<_Tp> Algorithm::create(const String& name)
 {
-    CV_Assert(!_seq || _seq->elem_size == sizeof(_Tp));
+    return _create(name).dynamicCast<_Tp>();
 }
 
-template<typename _Tp> inline Seq<_Tp>::Seq( MemStorage& storage,
-                                             int headerSize )
+template<typename _Tp> inline
+void Algorithm::set(const char* _name, const Ptr<_Tp>& value)
 {
-    CV_Assert(headerSize >= (int)sizeof(CvSeq));
-    seq = cvCreateSeq(DataType<_Tp>::type, headerSize, sizeof(_Tp), storage);
-}
-
-template<typename _Tp> inline _Tp& Seq<_Tp>::operator [](int idx)
-{ return *(_Tp*)getSeqElem(seq, idx); }
-
-template<typename _Tp> inline const _Tp& Seq<_Tp>::operator [](int idx) const
-{ return *(_Tp*)getSeqElem(seq, idx); }
-
-template<typename _Tp> inline SeqIterator<_Tp> Seq<_Tp>::begin() const
-{ return SeqIterator<_Tp>(*this); }
-
-template<typename _Tp> inline SeqIterator<_Tp> Seq<_Tp>::end() const
-{ return SeqIterator<_Tp>(*this, true); }
-
-template<typename _Tp> inline size_t Seq<_Tp>::size() const
-{ return seq ? seq->total : 0; }
-
-template<typename _Tp> inline int Seq<_Tp>::type() const
-{ return seq ? CV_MAT_TYPE(seq->flags) : 0; }
-
-template<typename _Tp> inline int Seq<_Tp>::depth() const
-{ return seq ? CV_MAT_DEPTH(seq->flags) : 0; }
-
-template<typename _Tp> inline int Seq<_Tp>::channels() const
-{ return seq ? CV_MAT_CN(seq->flags) : 0; }
-
-template<typename _Tp> inline size_t Seq<_Tp>::elemSize() const
-{ return seq ? seq->elem_size : 0; }
-
-template<typename _Tp> inline size_t Seq<_Tp>::index(const _Tp& elem) const
-{ return cvSeqElemIdx(seq, &elem); }
-
-template<typename _Tp> inline void Seq<_Tp>::push_back(const _Tp& elem)
-{ cvSeqPush(seq, &elem); }
-
-template<typename _Tp> inline void Seq<_Tp>::push_front(const _Tp& elem)
-{ cvSeqPushFront(seq, &elem); }
-
-template<typename _Tp> inline void Seq<_Tp>::push_back(const _Tp* elem, size_t count)
-{ cvSeqPushMulti(seq, elem, (int)count, 0); }
-
-template<typename _Tp> inline void Seq<_Tp>::push_front(const _Tp* elem, size_t count)
-{ cvSeqPushMulti(seq, elem, (int)count, 1); }
-
-template<typename _Tp> inline _Tp& Seq<_Tp>::back()
-{ return *(_Tp*)getSeqElem(seq, -1); }
-
-template<typename _Tp> inline const _Tp& Seq<_Tp>::back() const
-{ return *(const _Tp*)getSeqElem(seq, -1); }
-
-template<typename _Tp> inline _Tp& Seq<_Tp>::front()
-{ return *(_Tp*)getSeqElem(seq, 0); }
-
-template<typename _Tp> inline const _Tp& Seq<_Tp>::front() const
-{ return *(const _Tp*)getSeqElem(seq, 0); }
-
-template<typename _Tp> inline bool Seq<_Tp>::empty() const
-{ return !seq || seq->total == 0; }
-
-template<typename _Tp> inline void Seq<_Tp>::clear()
-{ if(seq) clearSeq(seq); }
-
-template<typename _Tp> inline void Seq<_Tp>::pop_back()
-{ seqPop(seq); }
-
-template<typename _Tp> inline void Seq<_Tp>::pop_front()
-{ seqPopFront(seq); }
-
-template<typename _Tp> inline void Seq<_Tp>::pop_back(_Tp* elem, size_t count)
-{ seqPopMulti(seq, elem, (int)count, 0); }
-
-template<typename _Tp> inline void Seq<_Tp>::pop_front(_Tp* elem, size_t count)
-{ seqPopMulti(seq, elem, (int)count, 1); }
-
-template<typename _Tp> inline void Seq<_Tp>::insert(int idx, const _Tp& elem)
-{ seqInsert(seq, idx, &elem); }
-
-template<typename _Tp> inline void Seq<_Tp>::insert(int idx, const _Tp* elems, size_t count)
-{
-    CvMat m = cvMat(1, count, DataType<_Tp>::type, elems);
-    seqInsertSlice(seq, idx, &m);
-}
-
-template<typename _Tp> inline void Seq<_Tp>::remove(int idx)
-{ seqRemove(seq, idx); }
-
-template<typename _Tp> inline void Seq<_Tp>::remove(const Range& r)
-{ seqRemoveSlice(seq, r); }
-
-template<typename _Tp> inline void Seq<_Tp>::copyTo(std::vector<_Tp>& vec, const Range& range) const
-{
-    size_t len = !seq ? 0 : range == Range::all() ? seq->total : range.end - range.start;
-    vec.resize(len);
-    if( seq && len )
-        cvCvtSeqToArray(seq, &vec[0], range);
-}
-
-template<typename _Tp> inline Seq<_Tp>::operator std::vector<_Tp>() const
-{
-    std::vector<_Tp> vec;
-    copyTo(vec);
-    return vec;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp>::SeqIterator()
-{ memset(this, 0, sizeof(*this)); }
-
-template<typename _Tp> inline SeqIterator<_Tp>::SeqIterator(const Seq<_Tp>& _seq, bool seekEnd)
-{
-    cvStartReadSeq(_seq.seq, this);
-    index = seekEnd ? _seq.seq->total : 0;
-}
-
-template<typename _Tp> inline void SeqIterator<_Tp>::seek(size_t pos)
-{
-    cvSetSeqReaderPos(this, (int)pos, false);
-    index = pos;
-}
-
-template<typename _Tp> inline size_t SeqIterator<_Tp>::tell() const
-{ return index; }
-
-template<typename _Tp> inline _Tp& SeqIterator<_Tp>::operator *()
-{ return *(_Tp*)ptr; }
-
-template<typename _Tp> inline const _Tp& SeqIterator<_Tp>::operator *() const
-{ return *(const _Tp*)ptr; }
-
-template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator ++()
-{
-    CV_NEXT_SEQ_ELEM(sizeof(_Tp), *this);
-    if( ++index >= seq->total*2 )
-        index = 0;
-    return *this;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp> SeqIterator<_Tp>::operator ++(int) const
-{
-    SeqIterator<_Tp> it = *this;
-    ++*this;
-    return it;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator --()
-{
-    CV_PREV_SEQ_ELEM(sizeof(_Tp), *this);
-    if( --index < 0 )
-        index = seq->total*2-1;
-    return *this;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp> SeqIterator<_Tp>::operator --(int) const
-{
-    SeqIterator<_Tp> it = *this;
-    --*this;
-    return it;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator +=(int delta)
-{
-    cvSetSeqReaderPos(this, delta, 1);
-    index += delta;
-    int n = seq->total*2;
-    if( index < 0 )
-        index += n;
-    if( index >= n )
-        index -= n;
-    return *this;
-}
-
-template<typename _Tp> inline SeqIterator<_Tp>& SeqIterator<_Tp>::operator -=(int delta)
-{
-    return (*this += -delta);
-}
-
-template<typename _Tp> inline ptrdiff_t operator - (const SeqIterator<_Tp>& a,
-                                                    const SeqIterator<_Tp>& b)
-{
-    ptrdiff_t delta = a.index - b.index, n = a.seq->total;
-    if( std::abs(static_cast<long>(delta)) > n )
-        delta += delta < 0 ? n : -n;
-    return delta;
-}
-
-template<typename _Tp> inline bool operator == (const SeqIterator<_Tp>& a,
-                                                const SeqIterator<_Tp>& b)
-{
-    return a.seq == b.seq && a.index == b.index;
-}
-
-template<typename _Tp> inline bool operator != (const SeqIterator<_Tp>& a,
-                                                const SeqIterator<_Tp>& b)
-{
-    return !(a == b);
-}
-
-
-template<typename _ClsName> struct CV_EXPORTS RTTIImpl
-{
-public:
-    static int isInstance(const void* ptr)
-    {
-        static _ClsName dummy;
-        static void* dummyp = &dummy;
-        union
-        {
-            const void* p;
-            const void** pp;
-        } a, b;
-        a.p = dummyp;
-        b.p = ptr;
-        return *a.pp == *b.pp;
-    }
-    static void release(void** dbptr)
-    {
-        if(dbptr && *dbptr)
-        {
-            delete (_ClsName*)*dbptr;
-            *dbptr = 0;
-        }
-    }
-    static void* read(CvFileStorage* fs, CvFileNode* n)
-    {
-        FileNode fn(fs, n);
-        _ClsName* obj = new _ClsName;
-        if(obj->read(fn))
-            return obj;
-        delete obj;
-        return 0;
-    }
-
-    static void write(CvFileStorage* _fs, const char* name, const void* ptr, CvAttrList)
-    {
-        if(ptr && _fs)
-        {
-            FileStorage fs(_fs);
-            fs.fs.addref();
-            ((const _ClsName*)ptr)->write(fs, std::string(name));
-        }
-    }
-
-    static void* clone(const void* ptr)
-    {
-        if(!ptr)
-            return 0;
-        return new _ClsName(*(const _ClsName*)ptr);
-    }
-};
-
-
-class CV_EXPORTS Formatter
-{
-public:
-    virtual ~Formatter() {}
-    virtual void write(std::ostream& out, const Mat& m, const int* params=0, int nparams=0) const = 0;
-    virtual void write(std::ostream& out, const void* data, int nelems, int type,
-                       const int* params=0, int nparams=0) const = 0;
-    static const Formatter* get(const char* fmt="");
-    static const Formatter* setDefault(const Formatter* fmt);
-};
-
-
-struct CV_EXPORTS Formatted
-{
-    Formatted(const Mat& m, const Formatter* fmt,
-              const std::vector<int>& params);
-    Formatted(const Mat& m, const Formatter* fmt,
-              const int* params=0);
-    Mat mtx;
-    const Formatter* fmt;
-    std::vector<int> params;
-};
-
-static inline Formatted format(const Mat& mtx, const char* fmt,
-                               const std::vector<int>& params=std::vector<int>())
-{
-    return Formatted(mtx, Formatter::get(fmt), params);
-}
-
-template<typename _Tp> static inline Formatted format(const std::vector<Point_<_Tp> >& vec,
-                                                      const char* fmt, const std::vector<int>& params=std::vector<int>())
-{
-    return Formatted(Mat(vec), Formatter::get(fmt), params);
-}
-
-template<typename _Tp> static inline Formatted format(const std::vector<Point3_<_Tp> >& vec,
-                                                      const char* fmt, const std::vector<int>& params=std::vector<int>())
-{
-    return Formatted(Mat(vec), Formatter::get(fmt), params);
-}
-
-/** \brief prints Mat to the output stream in Matlab notation
- * use like
- @verbatim
- Mat my_mat = Mat::eye(3,3,CV_32F);
- std::cout << my_mat;
- @endverbatim
- */
-static inline std::ostream& operator << (std::ostream& out, const Mat& mtx)
-{
-    Formatter::get()->write(out, mtx);
-    return out;
-}
-
-/** \brief prints Mat to the output stream allows in the specified notation (see format)
- * use like
- @verbatim
- Mat my_mat = Mat::eye(3,3,CV_32F);
- std::cout << my_mat;
- @endverbatim
- */
-static inline std::ostream& operator << (std::ostream& out, const Formatted& fmtd)
-{
-    fmtd.fmt->write(out, fmtd.mtx);
-    return out;
-}
-
-
-template<typename _Tp> static inline std::ostream& operator << (std::ostream& out,
-                                                                const std::vector<Point_<_Tp> >& vec)
-{
-    Formatter::get()->write(out, Mat(vec));
-    return out;
-}
-
-
-template<typename _Tp> static inline std::ostream& operator << (std::ostream& out,
-                                                                const std::vector<Point3_<_Tp> >& vec)
-{
-    Formatter::get()->write(out, Mat(vec));
-    return out;
-}
-
-
-/** Writes a Matx to an output stream.
- */
-template<typename _Tp, int m, int n> inline std::ostream& operator<<(std::ostream& out, const Matx<_Tp, m, n>& matx)
-{
-    out << cv::Mat(matx);
-    return out;
-}
-
-/** Writes a point to an output stream in Matlab notation
- */
-template<typename _Tp> inline std::ostream& operator<<(std::ostream& out, const Point_<_Tp>& p)
-{
-    out << "[" << p.x << ", " << p.y << "]";
-    return out;
-}
-
-/** Writes a point to an output stream in Matlab notation
- */
-template<typename _Tp> inline std::ostream& operator<<(std::ostream& out, const Point3_<_Tp>& p)
-{
-    out << "[" << p.x << ", " << p.y << ", " << p.z << "]";
-    return out;
-}
-
-/** Writes a Vec to an output stream. Format example : [10, 20, 30]
- */
-template<typename _Tp, int n> inline std::ostream& operator<<(std::ostream& out, const Vec<_Tp, n>& vec)
-{
-    out << "[";
-
-    if(Vec<_Tp, n>::depth < CV_32F)
-    {
-        for (int i = 0; i < n - 1; ++i) {
-            out << (int)vec[i] << ", ";
-        }
-        out << (int)vec[n-1] << "]";
-    }
-    else
-    {
-        for (int i = 0; i < n - 1; ++i) {
-            out << vec[i] << ", ";
-        }
-        out << vec[n-1] << "]";
-    }
-
-    return out;
-}
-
-/** Writes a Size_ to an output stream. Format example : [640 x 480]
- */
-template<typename _Tp> inline std::ostream& operator<<(std::ostream& out, const Size_<_Tp>& size)
-{
-    out << "[" << size.width << " x " << size.height << "]";
-    return out;
-}
-
-/** Writes a Rect_ to an output stream. Format example : [640 x 480 from (10, 20)]
- */
-template<typename _Tp> inline std::ostream& operator<<(std::ostream& out, const Rect_<_Tp>& rect)
-{
-    out << "[" << rect.width << " x " << rect.height << " from (" << rect.x << ", " << rect.y << ")]";
-    return out;
-}
-
-
-template<typename _Tp> inline Ptr<_Tp> Algorithm::create(const std::string& name)
-{
-    return _create(name).ptr<_Tp>();
-}
-
-template<typename _Tp>
-inline void Algorithm::set(const char* _name, const Ptr<_Tp>& value)
-{
-    Ptr<Algorithm> algo_ptr = value. template ptr<cv::Algorithm>();
-    if (algo_ptr.empty()) {
-        CV_Error( CV_StsUnsupportedFormat, "unknown/unsupported Ptr type of the second parameter of the method Algorithm::set");
+    Ptr<Algorithm> algo_ptr = value. template dynamicCast<cv::Algorithm>();
+    if (!algo_ptr) {
+        CV_Error( Error::StsUnsupportedFormat, "unknown/unsupported Ptr type of the second parameter of the method Algorithm::set");
     }
     info()->set(this, _name, ParamType<Algorithm>::type, &algo_ptr);
 }
 
-template<typename _Tp>
-inline void Algorithm::set(const std::string& _name, const Ptr<_Tp>& value)
+template<typename _Tp> inline
+void Algorithm::set(const String& _name, const Ptr<_Tp>& value)
 {
     this->set<_Tp>(_name.c_str(), value);
 }
 
-template<typename _Tp>
-inline void Algorithm::setAlgorithm(const char* _name, const Ptr<_Tp>& value)
+template<typename _Tp> inline
+void Algorithm::setAlgorithm(const char* _name, const Ptr<_Tp>& value)
 {
     Ptr<Algorithm> algo_ptr = value. template ptr<cv::Algorithm>();
-    if (algo_ptr.empty()) {
-        CV_Error( CV_StsUnsupportedFormat, "unknown/unsupported Ptr type of the second parameter of the method Algorithm::set");
+    if (!algo_ptr) {
+        CV_Error( Error::StsUnsupportedFormat, "unknown/unsupported Ptr type of the second parameter of the method Algorithm::set");
     }
     info()->set(this, _name, ParamType<Algorithm>::type, &algo_ptr);
 }
 
-template<typename _Tp>
-inline void Algorithm::setAlgorithm(const std::string& _name, const Ptr<_Tp>& value)
+template<typename _Tp> inline
+void Algorithm::setAlgorithm(const String& _name, const Ptr<_Tp>& value)
 {
     this->set<_Tp>(_name.c_str(), value);
 }
 
-template<typename _Tp> inline typename ParamType<_Tp>::member_type Algorithm::get(const std::string& _name) const
+template<typename _Tp> inline
+typename ParamType<_Tp>::member_type Algorithm::get(const String& _name) const
 {
     typename ParamType<_Tp>::member_type value;
     info()->get(this, _name.c_str(), ParamType<_Tp>::type, &value);
     return value;
 }
 
-template<typename _Tp> inline typename ParamType<_Tp>::member_type Algorithm::get(const char* _name) const
+template<typename _Tp> inline
+typename ParamType<_Tp>::member_type Algorithm::get(const char* _name) const
 {
     typename ParamType<_Tp>::member_type value;
     info()->get(this, _name, ParamType<_Tp>::type, &value);
     return value;
 }
 
-template<typename _Tp, typename _Base> inline void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
-                  Ptr<_Tp>& value, bool readOnly, Ptr<_Tp> (Algorithm::*getter)(), void (Algorithm::*setter)(const Ptr<_Tp>&),
-                  const std::string& help)
+template<typename _Tp, typename _Base> inline
+void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter, Ptr<_Tp>& value, bool readOnly,
+                             Ptr<_Tp> (Algorithm::*getter)(), void (Algorithm::*setter)(const Ptr<_Tp>&),
+                             const String& help)
 {
     //TODO: static assert: _Tp inherits from _Base
     addParam_(algo, parameter, ParamType<_Base>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
 }
 
-template<typename _Tp> inline void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
-                  Ptr<_Tp>& value, bool readOnly, Ptr<_Tp> (Algorithm::*getter)(), void (Algorithm::*setter)(const Ptr<_Tp>&),
-                  const std::string& help)
+template<typename _Tp> inline
+void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter, Ptr<_Tp>& value, bool readOnly,
+                             Ptr<_Tp> (Algorithm::*getter)(), void (Algorithm::*setter)(const Ptr<_Tp>&),
+                             const String& help)
 {
     //TODO: static assert: _Tp inherits from Algorithm
     addParam_(algo, parameter, ParamType<Algorithm>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
 }
 
-}
 
-#ifdef _MSC_VER
-# pragma warning(pop)
-#endif
 
-#endif // __cplusplus
+} // cv
+
 #endif
diff --git a/modules/core/include/opencv2/core/persistence.hpp b/modules/core/include/opencv2/core/persistence.hpp
new file mode 100644
index 0000000000..f5687614bb
--- /dev/null
+++ b/modules/core/include/opencv2/core/persistence.hpp
@@ -0,0 +1,870 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_PERSISTENCE_HPP__
+#define __OPENCV_CORE_PERSISTENCE_HPP__
+
+#ifndef __cplusplus
+#  error persistence.hpp header must be compiled as C++
+#endif
+
+// black-box structures used by FileStorage
+typedef struct CvFileStorage CvFileStorage;
+typedef struct CvFileNode CvFileNode;
+
+#include "opencv2/core/types.hpp"
+#include "opencv2/core/mat.hpp"
+
+namespace cv {
+
+////////////////////////// XML & YAML I/O //////////////////////////
+
+class CV_EXPORTS FileNode;
+class CV_EXPORTS FileNodeIterator;
+
+/*!
+ XML/YAML File Storage Class.
+
+ The class describes an object associated with XML or YAML file.
+ It can be used to store data to such a file or read and decode the data.
+
+ The storage is organized as a tree of nested sequences (or lists) and mappings.
+ Sequence is a heterogenious array, which elements are accessed by indices or sequentially using an iterator.
+ Mapping is analogue of std::map or C structure, which elements are accessed by names.
+ The most top level structure is a mapping.
+ Leaves of the file storage tree are integers, floating-point numbers and text strings.
+
+ For example, the following code:
+
+ \code
+ // open file storage for writing. Type of the file is determined from the extension
+ FileStorage fs("test.yml", FileStorage::WRITE);
+ fs << "test_int" << 5 << "test_real" << 3.1 << "test_string" << "ABCDEFGH";
+ fs << "test_mat" << Mat::eye(3,3,CV_32F);
+
+ fs << "test_list" << "[" << 0.0000000000001 << 2 << CV_PI << -3435345 << "2-502 2-029 3egegeg" <<
+ "{:" << "month" << 12 << "day" << 31 << "year" << 1969 << "}" << "]";
+ fs << "test_map" << "{" << "x" << 1 << "y" << 2 << "width" << 100 << "height" << 200 << "lbp" << "[:";
+
+ const uchar arr[] = {0, 1, 1, 0, 1, 1, 0, 1};
+ fs.writeRaw("u", arr, (int)(sizeof(arr)/sizeof(arr[0])));
+
+ fs << "]" << "}";
+ \endcode
+
+ will produce the following file:
+
+ \verbatim
+ %YAML:1.0
+ test_int: 5
+ test_real: 3.1000000000000001e+00
+ test_string: ABCDEFGH
+ test_mat: !!opencv-matrix
+     rows: 3
+     cols: 3
+     dt: f
+     data: [ 1., 0., 0., 0., 1., 0., 0., 0., 1. ]
+ test_list:
+     - 1.0000000000000000e-13
+     - 2
+     - 3.1415926535897931e+00
+     - -3435345
+     - "2-502 2-029 3egegeg"
+     - { month:12, day:31, year:1969 }
+ test_map:
+     x: 1
+     y: 2
+     width: 100
+     height: 200
+     lbp: [ 0, 1, 1, 0, 1, 1, 0, 1 ]
+ \endverbatim
+
+ and to read the file above, the following code can be used:
+
+ \code
+ // open file storage for reading.
+ // Type of the file is determined from the content, not the extension
+ FileStorage fs("test.yml", FileStorage::READ);
+ int test_int = (int)fs["test_int"];
+ double test_real = (double)fs["test_real"];
+ String test_string = (String)fs["test_string"];
+
+ Mat M;
+ fs["test_mat"] >> M;
+
+ FileNode tl = fs["test_list"];
+ CV_Assert(tl.type() == FileNode::SEQ && tl.size() == 6);
+ double tl0 = (double)tl[0];
+ int tl1 = (int)tl[1];
+ double tl2 = (double)tl[2];
+ int tl3 = (int)tl[3];
+ String tl4 = (String)tl[4];
+ CV_Assert(tl[5].type() == FileNode::MAP && tl[5].size() == 3);
+
+ int month = (int)tl[5]["month"];
+ int day = (int)tl[5]["day"];
+ int year = (int)tl[5]["year"];
+
+ FileNode tm = fs["test_map"];
+
+ int x = (int)tm["x"];
+ int y = (int)tm["y"];
+ int width = (int)tm["width"];
+ int height = (int)tm["height"];
+
+ int lbp_val = 0;
+ FileNodeIterator it = tm["lbp"].begin();
+
+ for(int k = 0; k < 8; k++, ++it)
+    lbp_val |= ((int)*it) << k;
+ \endcode
+*/
+class CV_EXPORTS_W FileStorage
+{
+public:
+    //! file storage mode
+    enum
+    {
+        READ        = 0, //! read mode
+        WRITE       = 1, //! write mode
+        APPEND      = 2, //! append mode
+        MEMORY      = 4,
+        FORMAT_MASK = (7<<3),
+        FORMAT_AUTO = 0,
+        FORMAT_XML  = (1<<3),
+        FORMAT_YAML = (2<<3)
+    };
+    enum
+    {
+        UNDEFINED      = 0,
+        VALUE_EXPECTED = 1,
+        NAME_EXPECTED  = 2,
+        INSIDE_MAP     = 4
+    };
+    //! the default constructor
+    CV_WRAP FileStorage();
+    //! the full constructor that opens file storage for reading or writing
+    CV_WRAP FileStorage(const String& source, int flags, const String& encoding=String());
+    //! the constructor that takes pointer to the C FileStorage structure
+    FileStorage(CvFileStorage* fs, bool owning=true);
+    //! the destructor. calls release()
+    virtual ~FileStorage();
+
+    //! opens file storage for reading or writing. The previous storage is closed with release()
+    CV_WRAP virtual bool open(const String& filename, int flags, const String& encoding=String());
+    //! returns true if the object is associated with currently opened file.
+    CV_WRAP virtual bool isOpened() const;
+    //! closes the file and releases all the memory buffers
+    CV_WRAP virtual void release();
+    //! closes the file, releases all the memory buffers and returns the text string
+    CV_WRAP virtual String releaseAndGetString();
+
+    //! returns the first element of the top-level mapping
+    CV_WRAP FileNode getFirstTopLevelNode() const;
+    //! returns the top-level mapping. YAML supports multiple streams
+    CV_WRAP FileNode root(int streamidx=0) const;
+    //! returns the specified element of the top-level mapping
+    FileNode operator[](const String& nodename) const;
+    //! returns the specified element of the top-level mapping
+    CV_WRAP FileNode operator[](const char* nodename) const;
+
+    //! returns pointer to the underlying C FileStorage structure
+    CvFileStorage* operator *() { return fs.get(); }
+    //! returns pointer to the underlying C FileStorage structure
+    const CvFileStorage* operator *() const { return fs.get(); }
+    //! writes one or more numbers of the specified format to the currently written structure
+    void writeRaw( const String& fmt, const uchar* vec, size_t len );
+    //! writes the registered C structure (CvMat, CvMatND, CvSeq). See cvWrite()
+    void writeObj( const String& name, const void* obj );
+
+    //! returns the normalized object name for the specified file name
+    static String getDefaultObjectName(const String& filename);
+
+    Ptr<CvFileStorage> fs; //!< the underlying C FileStorage structure
+    String elname; //!< the currently written element
+    std::vector<char> structs; //!< the stack of written structures
+    int state; //!< the writer state
+};
+
+template<> CV_EXPORTS void DefaultDeleter<CvFileStorage>::operator ()(CvFileStorage* obj) const;
+
+/*!
+ File Storage Node class
+
+ The node is used to store each and every element of the file storage opened for reading -
+ from the primitive objects, such as numbers and text strings, to the complex nodes:
+ sequences, mappings and the registered objects.
+
+ Note that file nodes are only used for navigating file storages opened for reading.
+ When a file storage is opened for writing, no data is stored in memory after it is written.
+*/
+class CV_EXPORTS_W_SIMPLE FileNode
+{
+public:
+    //! type of the file storage node
+    enum
+    {
+        NONE      = 0, //!< empty node
+        INT       = 1, //!< an integer
+        REAL      = 2, //!< floating-point number
+        FLOAT     = REAL, //!< synonym or REAL
+        STR       = 3, //!< text string in UTF-8 encoding
+        STRING    = STR, //!< synonym for STR
+        REF       = 4, //!< integer of size size_t. Typically used for storing complex dynamic structures where some elements reference the others
+        SEQ       = 5, //!< sequence
+        MAP       = 6, //!< mapping
+        TYPE_MASK = 7,
+        FLOW      = 8,  //!< compact representation of a sequence or mapping. Used only by YAML writer
+        USER      = 16, //!< a registered object (e.g. a matrix)
+        EMPTY     = 32, //!< empty structure (sequence or mapping)
+        NAMED     = 64  //!< the node has a name (i.e. it is element of a mapping)
+    };
+    //! the default constructor
+    CV_WRAP FileNode();
+    //! the full constructor wrapping CvFileNode structure.
+    FileNode(const CvFileStorage* fs, const CvFileNode* node);
+    //! the copy constructor
+    FileNode(const FileNode& node);
+    //! returns element of a mapping node
+    FileNode operator[](const String& nodename) const;
+    //! returns element of a mapping node
+    CV_WRAP FileNode operator[](const char* nodename) const;
+    //! returns element of a sequence node
+    CV_WRAP FileNode operator[](int i) const;
+    //! returns type of the node
+    CV_WRAP int type() const;
+
+    //! returns true if the node is empty
+    CV_WRAP bool empty() const;
+    //! returns true if the node is a "none" object
+    CV_WRAP bool isNone() const;
+    //! returns true if the node is a sequence
+    CV_WRAP bool isSeq() const;
+    //! returns true if the node is a mapping
+    CV_WRAP bool isMap() const;
+    //! returns true if the node is an integer
+    CV_WRAP bool isInt() const;
+    //! returns true if the node is a floating-point number
+    CV_WRAP bool isReal() const;
+    //! returns true if the node is a text string
+    CV_WRAP bool isString() const;
+    //! returns true if the node has a name
+    CV_WRAP bool isNamed() const;
+    //! returns the node name or an empty string if the node is nameless
+    CV_WRAP String name() const;
+    //! returns the number of elements in the node, if it is a sequence or mapping, or 1 otherwise.
+    CV_WRAP size_t size() const;
+    //! returns the node content as an integer. If the node stores floating-point number, it is rounded.
+    operator int() const;
+    //! returns the node content as float
+    operator float() const;
+    //! returns the node content as double
+    operator double() const;
+    //! returns the node content as text string
+    operator String() const;
+#ifndef OPENCV_NOSTL
+    operator std::string() const;
+#endif
+
+    //! returns pointer to the underlying file node
+    CvFileNode* operator *();
+    //! returns pointer to the underlying file node
+    const CvFileNode* operator* () const;
+
+    //! returns iterator pointing to the first node element
+    FileNodeIterator begin() const;
+    //! returns iterator pointing to the element following the last node element
+    FileNodeIterator end() const;
+
+    //! reads node elements to the buffer with the specified format
+    void readRaw( const String& fmt, uchar* vec, size_t len ) const;
+    //! reads the registered object and returns pointer to it
+    void* readObj() const;
+
+    // do not use wrapper pointer classes for better efficiency
+    const CvFileStorage* fs;
+    const CvFileNode* node;
+};
+
+
+/*!
+ File Node Iterator
+
+ The class is used for iterating sequences (usually) and mappings.
+ */
+class CV_EXPORTS FileNodeIterator
+{
+public:
+    //! the default constructor
+    FileNodeIterator();
+    //! the full constructor set to the ofs-th element of the node
+    FileNodeIterator(const CvFileStorage* fs, const CvFileNode* node, size_t ofs=0);
+    //! the copy constructor
+    FileNodeIterator(const FileNodeIterator& it);
+    //! returns the currently observed element
+    FileNode operator *() const;
+    //! accesses the currently observed element methods
+    FileNode operator ->() const;
+
+    //! moves iterator to the next node
+    FileNodeIterator& operator ++ ();
+    //! moves iterator to the next node
+    FileNodeIterator operator ++ (int);
+    //! moves iterator to the previous node
+    FileNodeIterator& operator -- ();
+    //! moves iterator to the previous node
+    FileNodeIterator operator -- (int);
+    //! moves iterator forward by the specified offset (possibly negative)
+    FileNodeIterator& operator += (int ofs);
+    //! moves iterator backward by the specified offset (possibly negative)
+    FileNodeIterator& operator -= (int ofs);
+
+    //! reads the next maxCount elements (or less, if the sequence/mapping last element occurs earlier) to the buffer with the specified format
+    FileNodeIterator& readRaw( const String& fmt, uchar* vec,
+                               size_t maxCount=(size_t)INT_MAX );
+
+    struct SeqReader
+    {
+      int          header_size;
+      void*        seq;        /* sequence, beign read; CvSeq      */
+      void*        block;      /* current block;        CvSeqBlock */
+      schar*       ptr;        /* pointer to element be read next */
+      schar*       block_min;  /* pointer to the beginning of block */
+      schar*       block_max;  /* pointer to the end of block */
+      int          delta_index;/* = seq->first->start_index   */
+      schar*       prev_elem;  /* pointer to previous element */
+    };
+
+    const CvFileStorage* fs;
+    const CvFileNode* container;
+    SeqReader reader;
+    size_t remaining;
+};
+
+
+
+/////////////////// XML & YAML I/O implementation //////////////////
+
+CV_EXPORTS void write( FileStorage& fs, const String& name, int value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, float value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, double value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, const String& value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, const Mat& value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, const SparseMat& value );
+CV_EXPORTS void write( FileStorage& fs, const String& name, const std::vector<KeyPoint>& value);
+
+CV_EXPORTS void writeScalar( FileStorage& fs, int value );
+CV_EXPORTS void writeScalar( FileStorage& fs, float value );
+CV_EXPORTS void writeScalar( FileStorage& fs, double value );
+CV_EXPORTS void writeScalar( FileStorage& fs, const String& value );
+
+CV_EXPORTS void read(const FileNode& node, int& value, int default_value);
+CV_EXPORTS void read(const FileNode& node, float& value, float default_value);
+CV_EXPORTS void read(const FileNode& node, double& value, double default_value);
+CV_EXPORTS void read(const FileNode& node, String& value, const String& default_value);
+CV_EXPORTS void read(const FileNode& node, Mat& mat, const Mat& default_mat = Mat() );
+CV_EXPORTS void read(const FileNode& node, SparseMat& mat, const SparseMat& default_mat = SparseMat() );
+CV_EXPORTS void read(const FileNode& node, std::vector<KeyPoint>& keypoints);
+
+template<typename _Tp> static inline void read(const FileNode& node, Point_<_Tp>& value, const Point_<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 2 ? default_value : Point_<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]));
+}
+
+template<typename _Tp> static inline void read(const FileNode& node, Point3_<_Tp>& value, const Point3_<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 3 ? default_value : Point3_<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]),
+                                                            saturate_cast<_Tp>(temp[2]));
+}
+
+template<typename _Tp> static inline void read(const FileNode& node, Size_<_Tp>& value, const Size_<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 2 ? default_value : Size_<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]));
+}
+
+template<typename _Tp> static inline void read(const FileNode& node, Complex<_Tp>& value, const Complex<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 2 ? default_value : Complex<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]));
+}
+
+template<typename _Tp> static inline void read(const FileNode& node, Rect_<_Tp>& value, const Rect_<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 4 ? default_value : Rect_<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]),
+                                                          saturate_cast<_Tp>(temp[2]), saturate_cast<_Tp>(temp[3]));
+}
+
+template<typename _Tp, int cn> static inline void read(const FileNode& node, Vec<_Tp, cn>& value, const Vec<_Tp, cn>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != cn ? default_value : Vec<_Tp, cn>(&temp[0]);
+}
+
+template<typename _Tp> static inline void read(const FileNode& node, Scalar_<_Tp>& value, const Scalar_<_Tp>& default_value)
+{
+    std::vector<_Tp> temp; FileNodeIterator it = node.begin(); it >> temp;
+    value = temp.size() != 4 ? default_value : Scalar_<_Tp>(saturate_cast<_Tp>(temp[0]), saturate_cast<_Tp>(temp[1]),
+                                                            saturate_cast<_Tp>(temp[2]), saturate_cast<_Tp>(temp[3]));
+}
+
+static inline void read(const FileNode& node, Range& value, const Range& default_value)
+{
+    Point2i temp(value.start, value.end); const Point2i default_temp = Point2i(default_value.start, default_value.end);
+    read(node, temp, default_temp);
+    value.start = temp.x; value.end = temp.y;
+}
+
+
+CV_EXPORTS FileStorage& operator << (FileStorage& fs, const String& str);
+
+
+namespace internal
+{
+    class CV_EXPORTS WriteStructContext
+    {
+    public:
+        WriteStructContext(FileStorage& _fs, const String& name, int flags, const String& typeName = String());
+        ~WriteStructContext();
+    private:
+        FileStorage* fs;
+    };
+
+    template<typename _Tp, int numflag> class VecWriterProxy
+    {
+    public:
+        VecWriterProxy( FileStorage* _fs ) : fs(_fs) {}
+        void operator()(const std::vector<_Tp>& vec) const
+        {
+            size_t count = vec.size();
+            for (size_t i = 0; i < count; i++)
+                write(*fs, vec[i]);
+        }
+    private:
+        FileStorage* fs;
+    };
+
+    template<typename _Tp> class VecWriterProxy<_Tp, 1>
+    {
+    public:
+        VecWriterProxy( FileStorage* _fs ) : fs(_fs) {}
+        void operator()(const std::vector<_Tp>& vec) const
+        {
+            int _fmt = DataType<_Tp>::fmt;
+            char fmt[] = { (char)((_fmt >> 8) + '1'), (char)_fmt, '\0' };
+            fs->writeRaw(fmt, !vec.empty() ? (uchar*)&vec[0] : 0, vec.size() * sizeof(_Tp));
+        }
+    private:
+        FileStorage* fs;
+    };
+
+    template<typename _Tp, int numflag> class VecReaderProxy
+    {
+    public:
+        VecReaderProxy( FileNodeIterator* _it ) : it(_it) {}
+        void operator()(std::vector<_Tp>& vec, size_t count) const
+        {
+            count = std::min(count, it->remaining);
+            vec.resize(count);
+            for (size_t i = 0; i < count; i++, ++(*it))
+                read(**it, vec[i], _Tp());
+        }
+    private:
+        FileNodeIterator* it;
+    };
+
+    template<typename _Tp> class VecReaderProxy<_Tp, 1>
+    {
+    public:
+        VecReaderProxy( FileNodeIterator* _it ) : it(_it) {}
+        void operator()(std::vector<_Tp>& vec, size_t count) const
+        {
+            size_t remaining = it->remaining;
+            size_t cn = DataType<_Tp>::channels;
+            int _fmt = DataType<_Tp>::fmt;
+            char fmt[] = { (char)((_fmt >> 8)+'1'), (char)_fmt, '\0' };
+            size_t remaining1 = remaining / cn;
+            count = count < remaining1 ? count : remaining1;
+            vec.resize(count);
+            it->readRaw(fmt, !vec.empty() ? (uchar*)&vec[0] : 0, count*sizeof(_Tp));
+        }
+    private:
+        FileNodeIterator* it;
+    };
+
+} // internal
+
+
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const _Tp& value)
+{
+    write(fs, String(), value);
+}
+
+template<> inline
+void write( FileStorage& fs, const int& value )
+{
+    writeScalar(fs, value);
+}
+
+template<> inline
+void write( FileStorage& fs, const float& value )
+{
+    writeScalar(fs, value);
+}
+
+template<> inline
+void write( FileStorage& fs, const double& value )
+{
+    writeScalar(fs, value);
+}
+
+template<> inline
+void write( FileStorage& fs, const String& value )
+{
+    writeScalar(fs, value);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Point_<_Tp>& pt )
+{
+    write(fs, pt.x);
+    write(fs, pt.y);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Point3_<_Tp>& pt )
+{
+    write(fs, pt.x);
+    write(fs, pt.y);
+    write(fs, pt.z);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Size_<_Tp>& sz )
+{
+    write(fs, sz.width);
+    write(fs, sz.height);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Complex<_Tp>& c )
+{
+    write(fs, c.re);
+    write(fs, c.im);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Rect_<_Tp>& r )
+{
+    write(fs, r.x);
+    write(fs, r.y);
+    write(fs, r.width);
+    write(fs, r.height);
+}
+
+template<typename _Tp, int cn> static inline
+void write(FileStorage& fs, const Vec<_Tp, cn>& v )
+{
+    for(int i = 0; i < cn; i++)
+        write(fs, v.val[i]);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const Scalar_<_Tp>& s )
+{
+    write(fs, s.val[0]);
+    write(fs, s.val[1]);
+    write(fs, s.val[2]);
+    write(fs, s.val[3]);
+}
+
+static inline
+void write(FileStorage& fs, const Range& r )
+{
+    write(fs, r.start);
+    write(fs, r.end);
+}
+
+template<typename _Tp> static inline
+void write( FileStorage& fs, const std::vector<_Tp>& vec )
+{
+    internal::VecWriterProxy<_Tp, DataType<_Tp>::fmt != 0> w(&fs);
+    w(vec);
+}
+
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Point_<_Tp>& pt )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, pt);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Point3_<_Tp>& pt )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, pt);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Size_<_Tp>& sz )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, sz);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Complex<_Tp>& c )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, c);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Rect_<_Tp>& r )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, r);
+}
+
+template<typename _Tp, int cn> static inline
+void write(FileStorage& fs, const String& name, const Vec<_Tp, cn>& v )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, v);
+}
+
+template<typename _Tp> static inline
+void write(FileStorage& fs, const String& name, const Scalar_<_Tp>& s )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, s);
+}
+
+static inline
+void write(FileStorage& fs, const String& name, const Range& r )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+FileNode::FLOW);
+    write(fs, r);
+}
+
+template<typename _Tp> static inline
+void write( FileStorage& fs, const String& name, const std::vector<_Tp>& vec )
+{
+    internal::WriteStructContext ws(fs, name, FileNode::SEQ+(DataType<_Tp>::fmt != 0 ? FileNode::FLOW : 0));
+    write(fs, vec);
+}
+
+
+static inline
+void read(const FileNode& node, bool& value, bool default_value)
+{
+    int temp;
+    read(node, temp, (int)default_value);
+    value = temp != 0;
+}
+
+static inline
+void read(const FileNode& node, uchar& value, uchar default_value)
+{
+    int temp;
+    read(node, temp, (int)default_value);
+    value = saturate_cast<uchar>(temp);
+}
+
+static inline
+void read(const FileNode& node, schar& value, schar default_value)
+{
+    int temp;
+    read(node, temp, (int)default_value);
+    value = saturate_cast<schar>(temp);
+}
+
+static inline
+void read(const FileNode& node, ushort& value, ushort default_value)
+{
+    int temp;
+    read(node, temp, (int)default_value);
+    value = saturate_cast<ushort>(temp);
+}
+
+static inline
+void read(const FileNode& node, short& value, short default_value)
+{
+    int temp;
+    read(node, temp, (int)default_value);
+    value = saturate_cast<short>(temp);
+}
+
+template<typename _Tp> static inline
+void read( FileNodeIterator& it, std::vector<_Tp>& vec, size_t maxCount = (size_t)INT_MAX )
+{
+    internal::VecReaderProxy<_Tp, DataType<_Tp>::fmt != 0> r(&it);
+    r(vec, maxCount);
+}
+
+template<typename _Tp> static inline
+void read( const FileNode& node, std::vector<_Tp>& vec, const std::vector<_Tp>& default_value = std::vector<_Tp>() )
+{
+    if(!node.node)
+        vec = default_value;
+    else
+    {
+        FileNodeIterator it = node.begin();
+        read( it, vec );
+    }
+}
+
+
+template<typename _Tp> static inline
+FileStorage& operator << (FileStorage& fs, const _Tp& value)
+{
+    if( !fs.isOpened() )
+        return fs;
+    if( fs.state == FileStorage::NAME_EXPECTED + FileStorage::INSIDE_MAP )
+        CV_Error( Error::StsError, "No element name has been given" );
+    write( fs, fs.elname, value );
+    if( fs.state & FileStorage::INSIDE_MAP )
+        fs.state = FileStorage::NAME_EXPECTED + FileStorage::INSIDE_MAP;
+    return fs;
+}
+
+static inline
+FileStorage& operator << (FileStorage& fs, const char* str)
+{
+    return (fs << String(str));
+}
+
+static inline
+FileStorage& operator << (FileStorage& fs, char* value)
+{
+    return (fs << String(value));
+}
+
+template<typename _Tp> static inline
+FileNodeIterator& operator >> (FileNodeIterator& it, _Tp& value)
+{
+    read( *it, value, _Tp());
+    return ++it;
+}
+
+template<typename _Tp> static inline
+FileNodeIterator& operator >> (FileNodeIterator& it, std::vector<_Tp>& vec)
+{
+    internal::VecReaderProxy<_Tp, DataType<_Tp>::fmt != 0> r(&it);
+    r(vec, (size_t)INT_MAX);
+    return it;
+}
+
+template<typename _Tp> static inline
+void operator >> (const FileNode& n, _Tp& value)
+{
+    read( n, value, _Tp());
+}
+
+template<typename _Tp> static inline
+void operator >> (const FileNode& n, std::vector<_Tp>& vec)
+{
+    FileNodeIterator it = n.begin();
+    it >> vec;
+}
+
+
+static inline
+bool operator == (const FileNodeIterator& it1, const FileNodeIterator& it2)
+{
+    return it1.fs == it2.fs && it1.container == it2.container &&
+        it1.reader.ptr == it2.reader.ptr && it1.remaining == it2.remaining;
+}
+
+static inline
+bool operator != (const FileNodeIterator& it1, const FileNodeIterator& it2)
+{
+    return !(it1 == it2);
+}
+
+static inline
+ptrdiff_t operator - (const FileNodeIterator& it1, const FileNodeIterator& it2)
+{
+    return it2.remaining - it1.remaining;
+}
+
+static inline
+bool operator < (const FileNodeIterator& it1, const FileNodeIterator& it2)
+{
+    return it1.remaining > it2.remaining;
+}
+
+inline FileNode FileStorage::getFirstTopLevelNode() const { FileNode r = root(); FileNodeIterator it = r.begin(); return it != r.end() ? *it : FileNode(); }
+inline FileNode::FileNode() : fs(0), node(0) {}
+inline FileNode::FileNode(const CvFileStorage* _fs, const CvFileNode* _node) : fs(_fs), node(_node) {}
+inline FileNode::FileNode(const FileNode& _node) : fs(_node.fs), node(_node.node) {}
+inline bool FileNode::empty() const    { return node   == 0;    }
+inline bool FileNode::isNone() const   { return type() == NONE; }
+inline bool FileNode::isSeq() const    { return type() == SEQ;  }
+inline bool FileNode::isMap() const    { return type() == MAP;  }
+inline bool FileNode::isInt() const    { return type() == INT;  }
+inline bool FileNode::isReal() const   { return type() == REAL; }
+inline bool FileNode::isString() const { return type() == STR;  }
+inline CvFileNode* FileNode::operator *() { return (CvFileNode*)node; }
+inline const CvFileNode* FileNode::operator* () const { return node; }
+inline FileNode::operator int() const    { int value;    read(*this, value, 0);     return value; }
+inline FileNode::operator float() const  { float value;  read(*this, value, 0.f);   return value; }
+inline FileNode::operator double() const { double value; read(*this, value, 0.);    return value; }
+inline FileNode::operator String() const { String value; read(*this, value, value); return value; }
+inline FileNodeIterator FileNode::begin() const { return FileNodeIterator(fs, node); }
+inline FileNodeIterator FileNode::end() const   { return FileNodeIterator(fs, node, size()); }
+inline void FileNode::readRaw( const String& fmt, uchar* vec, size_t len ) const { begin().readRaw( fmt, vec, len ); }
+inline FileNode FileNodeIterator::operator *() const  { return FileNode(fs, (const CvFileNode*)reader.ptr); }
+inline FileNode FileNodeIterator::operator ->() const { return FileNode(fs, (const CvFileNode*)reader.ptr); }
+inline String::String(const FileNode& fn): cstr_(0), len_(0) { read(fn, *this, *this); }
+
+} // cv
+
+#endif // __OPENCV_CORE_PERSISTENCE_HPP__
diff --git a/modules/core/include/opencv2/core/private.gpu.hpp b/modules/core/include/opencv2/core/private.gpu.hpp
new file mode 100644
index 0000000000..d26cd0017f
--- /dev/null
+++ b/modules/core/include/opencv2/core/private.gpu.hpp
@@ -0,0 +1,143 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_PRIVATE_GPU_HPP__
+#define __OPENCV_CORE_PRIVATE_GPU_HPP__
+
+#ifndef __OPENCV_BUILD
+#  error this is a private header which should not be used from outside of the OpenCV library
+#endif
+
+#include "cvconfig.h"
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/base.hpp"
+
+#include "opencv2/core/gpu.hpp"
+
+#ifdef HAVE_CUDA
+#  include <cuda.h>
+#  include <cuda_runtime.h>
+#  include <npp.h>
+#  include "opencv2/core/gpu_stream_accessor.hpp"
+#  include "opencv2/core/cuda/common.hpp"
+
+#  define NPP_VERSION (NPP_VERSION_MAJOR * 1000 + NPP_VERSION_MINOR * 100 + NPP_VERSION_BUILD)
+
+#  define CUDART_MINIMUM_REQUIRED_VERSION 4020
+
+#  if (CUDART_VERSION < CUDART_MINIMUM_REQUIRED_VERSION)
+#    error "Insufficient Cuda Runtime library version, please update it."
+#  endif
+
+#  if defined(CUDA_ARCH_BIN_OR_PTX_10)
+#    error "OpenCV GPU module doesn't support NVIDIA compute capability 1.0"
+#  endif
+#endif
+
+namespace cv { namespace gpu {
+    CV_EXPORTS cv::String getNppErrorMessage(int code);
+    CV_EXPORTS cv::String getCudaDriverApiErrorMessage(int code);
+}}
+
+#ifndef HAVE_CUDA
+
+static inline void throw_no_cuda() { CV_Error(cv::Error::GpuNotSupported, "The library is compiled without GPU support"); }
+
+#else // HAVE_CUDA
+
+static inline void throw_no_cuda() { CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform"); }
+
+namespace cv { namespace gpu
+{
+    static inline void checkNppError(int code, const char* file, const int line, const char* func)
+    {
+        if (code < 0)
+            cv::error(cv::Error::GpuApiCallError, getNppErrorMessage(code), func, file, line);
+    }
+
+    static inline void checkCudaDriverApiError(int code, const char* file, const int line, const char* func)
+    {
+        if (code != CUDA_SUCCESS)
+            cv::error(cv::Error::GpuApiCallError, getCudaDriverApiErrorMessage(code), func, file, line);
+    }
+
+    template<int n> struct NPPTypeTraits;
+    template<> struct NPPTypeTraits<CV_8U>  { typedef Npp8u npp_type; };
+    template<> struct NPPTypeTraits<CV_8S>  { typedef Npp8s npp_type; };
+    template<> struct NPPTypeTraits<CV_16U> { typedef Npp16u npp_type; };
+    template<> struct NPPTypeTraits<CV_16S> { typedef Npp16s npp_type; };
+    template<> struct NPPTypeTraits<CV_32S> { typedef Npp32s npp_type; };
+    template<> struct NPPTypeTraits<CV_32F> { typedef Npp32f npp_type; };
+    template<> struct NPPTypeTraits<CV_64F> { typedef Npp64f npp_type; };
+
+    class NppStreamHandler
+    {
+    public:
+        inline explicit NppStreamHandler(cudaStream_t newStream)
+        {
+            oldStream = nppGetStream();
+            nppSetStream(newStream);
+        }
+
+        inline ~NppStreamHandler()
+        {
+            nppSetStream(oldStream);
+        }
+
+    private:
+        cudaStream_t oldStream;
+    };
+}}
+
+#if defined(__GNUC__)
+    #define nppSafeCall(expr)  cv::gpu::checkNppError(expr, __FILE__, __LINE__, __func__)
+    #define cuSafeCall(expr)  cv::gpu::checkCudaDriverApiError(expr, __FILE__, __LINE__, __func__)
+#else /* defined(__CUDACC__) || defined(__MSVC__) */
+    #define nppSafeCall(expr)  cv::gpu::checkNppError(expr, __FILE__, __LINE__, "")
+    #define cuSafeCall(expr)  cv::gpu::checkCudaDriverApiError(expr, __FILE__, __LINE__, "")
+#endif
+
+#endif // HAVE_CUDA
+
+#endif // __OPENCV_CORE_GPU_PRIVATE_HPP__
diff --git a/modules/core/include/opencv2/core/private.hpp b/modules/core/include/opencv2/core/private.hpp
new file mode 100644
index 0000000000..a1cc0e5c89
--- /dev/null
+++ b/modules/core/include/opencv2/core/private.hpp
@@ -0,0 +1,373 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_PRIVATE_HPP__
+#define __OPENCV_CORE_PRIVATE_HPP__
+
+#ifndef __OPENCV_BUILD
+#  error this is a private header which should not be used from outside of the OpenCV library
+#endif
+
+#include "opencv2/core.hpp"
+#include "cvconfig.h"
+
+#ifdef HAVE_EIGEN
+#  if defined __GNUC__ && defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wshadow"
+#  endif
+#  include <Eigen/Core>
+#  include "opencv2/core/eigen.hpp"
+#endif
+
+#ifdef HAVE_TBB
+#  include "tbb/tbb_stddef.h"
+#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#    include "tbb/tbb.h"
+#    include "tbb/task.h"
+#    undef min
+#    undef max
+#  else
+#    undef HAVE_TBB
+#  endif
+#endif
+
+namespace cv
+{
+#ifdef HAVE_TBB
+
+    typedef tbb::blocked_range<int> BlockedRange;
+
+    template<typename Body> static inline
+    void parallel_for( const BlockedRange& range, const Body& body )
+    {
+        tbb::parallel_for(range, body);
+    }
+
+    typedef tbb::split Split;
+
+    template<typename Body> static inline
+    void parallel_reduce( const BlockedRange& range, Body& body )
+    {
+        tbb::parallel_reduce(range, body);
+    }
+
+    typedef tbb::concurrent_vector<Rect> ConcurrentRectVector;
+#else
+    class BlockedRange
+    {
+    public:
+        BlockedRange() : _begin(0), _end(0), _grainsize(0) {}
+        BlockedRange(int b, int e, int g=1) : _begin(b), _end(e), _grainsize(g) {}
+        int begin() const { return _begin; }
+        int end() const { return _end; }
+        int grainsize() const { return _grainsize; }
+
+    protected:
+        int _begin, _end, _grainsize;
+    };
+
+    template<typename Body> static inline
+    void parallel_for( const BlockedRange& range, const Body& body )
+    {
+        body(range);
+    }
+    typedef std::vector<Rect> ConcurrentRectVector;
+
+    class Split {};
+
+    template<typename Body> static inline
+    void parallel_reduce( const BlockedRange& range, Body& body )
+    {
+        body(range);
+    }
+#endif
+
+    // Returns a static string if there is a parallel framework,
+    // NULL otherwise.
+    CV_EXPORTS const char* currentParallelFramework();
+} //namespace cv
+
+#define CV_INIT_ALGORITHM(classname, algname, memberinit) \
+    static inline ::cv::Algorithm* create##classname##_hidden() \
+    { \
+        return new classname; \
+    } \
+    \
+    static inline ::cv::Ptr< ::cv::Algorithm> create##classname##_ptr_hidden() \
+    { \
+        return ::cv::makePtr<classname>(); \
+    } \
+    \
+    static inline ::cv::AlgorithmInfo& classname##_info() \
+    { \
+        static ::cv::AlgorithmInfo classname##_info_var(algname, create##classname##_hidden); \
+        return classname##_info_var; \
+    } \
+    \
+    static ::cv::AlgorithmInfo& classname##_info_auto = classname##_info(); \
+    \
+    ::cv::AlgorithmInfo* classname::info() const \
+    { \
+        static volatile bool initialized = false; \
+        \
+        if( !initialized ) \
+        { \
+            initialized = true; \
+            classname obj; \
+            memberinit; \
+        } \
+        return &classname##_info(); \
+    }
+
+
+
+/****************************************************************************************\
+*                                  Common declarations                                   *
+\****************************************************************************************/
+
+/* the alignment of all the allocated buffers */
+#define  CV_MALLOC_ALIGN    16
+
+#ifdef __GNUC__
+#  define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
+#elif defined _MSC_VER
+#  define CV_DECL_ALIGNED(x) __declspec(align(x))
+#else
+#  define CV_DECL_ALIGNED(x)
+#endif
+
+/* IEEE754 constants and macros */
+#define  CV_TOGGLE_FLT(x) ((x)^((int)(x) < 0 ? 0x7fffffff : 0))
+#define  CV_TOGGLE_DBL(x) ((x)^((int64)(x) < 0 ? CV_BIG_INT(0x7fffffffffffffff) : 0))
+
+static inline void* cvAlignPtr( const void* ptr, int align = 32 )
+{
+    CV_DbgAssert ( (align & (align-1)) == 0 );
+    return (void*)( ((size_t)ptr + align - 1) & ~(size_t)(align-1) );
+}
+
+static inline int cvAlign( int size, int align )
+{
+    CV_DbgAssert( (align & (align-1)) == 0 && size < INT_MAX );
+    return (size + align - 1) & -align;
+}
+
+#ifdef IPL_DEPTH_8U
+static inline cv::Size cvGetMatSize( const CvMat* mat )
+{
+    return cv::Size(mat->cols, mat->rows);
+}
+#endif
+
+namespace cv
+{
+CV_EXPORTS void scalarToRawData(const cv::Scalar& s, void* buf, int type, int unroll_to = 0);
+}
+
+
+/****************************************************************************************\
+*                     Structures and macros for integration with IPP                     *
+\****************************************************************************************/
+
+#ifdef HAVE_IPP
+#  include "ipp.h"
+
+static inline IppiSize ippiSize(int width, int height)
+{
+    IppiSize size = { width, height };
+    return size;
+}
+#endif
+
+#ifndef IPPI_CALL
+#  define IPPI_CALL(func) CV_Assert((func) >= 0)
+#endif
+
+/* IPP-compatible return codes */
+typedef enum CvStatus
+{
+    CV_BADMEMBLOCK_ERR          = -113,
+    CV_INPLACE_NOT_SUPPORTED_ERR= -112,
+    CV_UNMATCHED_ROI_ERR        = -111,
+    CV_NOTFOUND_ERR             = -110,
+    CV_BADCONVERGENCE_ERR       = -109,
+
+    CV_BADDEPTH_ERR             = -107,
+    CV_BADROI_ERR               = -106,
+    CV_BADHEADER_ERR            = -105,
+    CV_UNMATCHED_FORMATS_ERR    = -104,
+    CV_UNSUPPORTED_COI_ERR      = -103,
+    CV_UNSUPPORTED_CHANNELS_ERR = -102,
+    CV_UNSUPPORTED_DEPTH_ERR    = -101,
+    CV_UNSUPPORTED_FORMAT_ERR   = -100,
+
+    CV_BADARG_ERR               = -49,  //ipp comp
+    CV_NOTDEFINED_ERR           = -48,  //ipp comp
+
+    CV_BADCHANNELS_ERR          = -47,  //ipp comp
+    CV_BADRANGE_ERR             = -44,  //ipp comp
+    CV_BADSTEP_ERR              = -29,  //ipp comp
+
+    CV_BADFLAG_ERR              =  -12,
+    CV_DIV_BY_ZERO_ERR          =  -11, //ipp comp
+    CV_BADCOEF_ERR              =  -10,
+
+    CV_BADFACTOR_ERR            =  -7,
+    CV_BADPOINT_ERR             =  -6,
+    CV_BADSCALE_ERR             =  -4,
+    CV_OUTOFMEM_ERR             =  -3,
+    CV_NULLPTR_ERR              =  -2,
+    CV_BADSIZE_ERR              =  -1,
+    CV_NO_ERR                   =   0,
+    CV_OK                       =   CV_NO_ERR
+}
+CvStatus;
+
+
+
+/****************************************************************************************\
+*                                  Auxiliary algorithms                                  *
+\****************************************************************************************/
+
+namespace cv
+{
+
+// This function splits the input sequence or set into one or more equivalence classes and
+// returns the vector of labels - 0-based class indexes for each element.
+// predicate(a,b) returns true if the two sequence elements certainly belong to the same class.
+//
+// The algorithm is described in "Introduction to Algorithms"
+// by Cormen, Leiserson and Rivest, the chapter "Data structures for disjoint sets"
+template<typename _Tp, class _EqPredicate> int
+partition( const std::vector<_Tp>& _vec, std::vector<int>& labels,
+           _EqPredicate predicate=_EqPredicate())
+{
+    int i, j, N = (int)_vec.size();
+    const _Tp* vec = &_vec[0];
+
+    const int PARENT=0;
+    const int RANK=1;
+
+    std::vector<int> _nodes(N*2);
+    int (*nodes)[2] = (int(*)[2])&_nodes[0];
+
+    // The first O(N) pass: create N single-vertex trees
+    for(i = 0; i < N; i++)
+    {
+        nodes[i][PARENT]=-1;
+        nodes[i][RANK] = 0;
+    }
+
+    // The main O(N^2) pass: merge connected components
+    for( i = 0; i < N; i++ )
+    {
+        int root = i;
+
+        // find root
+        while( nodes[root][PARENT] >= 0 )
+            root = nodes[root][PARENT];
+
+        for( j = 0; j < N; j++ )
+        {
+            if( i == j || !predicate(vec[i], vec[j]))
+                continue;
+            int root2 = j;
+
+            while( nodes[root2][PARENT] >= 0 )
+                root2 = nodes[root2][PARENT];
+
+            if( root2 != root )
+            {
+                // unite both trees
+                int rank = nodes[root][RANK], rank2 = nodes[root2][RANK];
+                if( rank > rank2 )
+                    nodes[root2][PARENT] = root;
+                else
+                {
+                    nodes[root][PARENT] = root2;
+                    nodes[root2][RANK] += rank == rank2;
+                    root = root2;
+                }
+                CV_Assert( nodes[root][PARENT] < 0 );
+
+                int k = j, parent;
+
+                // compress the path from node2 to root
+                while( (parent = nodes[k][PARENT]) >= 0 )
+                {
+                    nodes[k][PARENT] = root;
+                    k = parent;
+                }
+
+                // compress the path from node to root
+                k = i;
+                while( (parent = nodes[k][PARENT]) >= 0 )
+                {
+                    nodes[k][PARENT] = root;
+                    k = parent;
+                }
+            }
+        }
+    }
+
+    // Final O(N) pass: enumerate classes
+    labels.resize(N);
+    int nclasses = 0;
+
+    for( i = 0; i < N; i++ )
+    {
+        int root = i;
+        while( nodes[root][PARENT] >= 0 )
+            root = nodes[root][PARENT];
+        // re-use the rank as the class label
+        if( nodes[root][RANK] >= 0 )
+            nodes[root][RANK] = ~nclasses++;
+        labels[i] = ~nodes[root][RANK];
+    }
+
+    return nclasses;
+}
+
+} // namespace cv
+
+#endif // __OPENCV_CORE_PRIVATE_HPP__
diff --git a/modules/core/include/opencv2/core/ptr.inl.hpp b/modules/core/include/opencv2/core/ptr.inl.hpp
new file mode 100644
index 0000000000..989724281f
--- /dev/null
+++ b/modules/core/include/opencv2/core/ptr.inl.hpp
@@ -0,0 +1,338 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the copyright holders or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_PTR_INL_HPP__
+#define __OPENCV_CORE_PTR_INL_HPP__
+
+#include <algorithm>
+
+namespace cv {
+
+template<typename Y>
+void DefaultDeleter<Y>::operator () (Y* p) const
+{
+    delete p;
+}
+
+namespace detail
+{
+
+struct PtrOwner
+{
+    PtrOwner() : refCount(1)
+    {}
+
+    void incRef()
+    {
+        CV_XADD(&refCount, 1);
+    }
+
+    void decRef()
+    {
+        if (CV_XADD(&refCount, -1) == 1) deleteSelf();
+    }
+
+protected:
+    /* This doesn't really need to be virtual, since PtrOwner is never deleted
+       directly, but it doesn't hurt and it helps avoid warnings. */
+    virtual ~PtrOwner()
+    {}
+
+    virtual void deleteSelf() = 0;
+
+private:
+    unsigned int refCount;
+
+    // noncopyable
+    PtrOwner(const PtrOwner&);
+    PtrOwner& operator = (const PtrOwner&);
+};
+
+template<typename Y, typename D>
+struct PtrOwnerImpl : PtrOwner
+{
+    PtrOwnerImpl(Y* p, D d) : owned(p), deleter(d)
+    {}
+
+    void deleteSelf()
+    {
+        deleter(owned);
+        delete this;
+    }
+
+private:
+    Y* owned;
+    D deleter;
+};
+
+
+}
+
+template<typename T>
+Ptr<T>::Ptr() : owner(NULL), stored(NULL)
+{}
+
+template<typename T>
+template<typename Y>
+Ptr<T>::Ptr(Y* p)
+  : owner(p
+      ? new detail::PtrOwnerImpl<Y, DefaultDeleter<Y> >(p, DefaultDeleter<Y>())
+      : NULL),
+    stored(p)
+{}
+
+template<typename T>
+template<typename Y, typename D>
+Ptr<T>::Ptr(Y* p, D d)
+  : owner(p
+      ? new detail::PtrOwnerImpl<Y, D>(p, d)
+      : NULL),
+    stored(p)
+{}
+
+template<typename T>
+Ptr<T>::Ptr(const Ptr& o) : owner(o.owner), stored(o.stored)
+{
+    if (owner) owner->incRef();
+}
+
+template<typename T>
+template<typename Y>
+Ptr<T>::Ptr(const Ptr<Y>& o) : owner(o.owner), stored(o.stored)
+{
+    if (owner) owner->incRef();
+}
+
+template<typename T>
+template<typename Y>
+Ptr<T>::Ptr(const Ptr<Y>& o, T* p) : owner(o.owner), stored(p)
+{
+    if (owner) owner->incRef();
+}
+
+template<typename T>
+Ptr<T>::~Ptr()
+{
+    release();
+}
+
+template<typename T>
+Ptr<T>& Ptr<T>::operator = (const Ptr<T>& o)
+{
+    Ptr(o).swap(*this);
+    return *this;
+}
+
+template<typename T>
+template<typename Y>
+Ptr<T>& Ptr<T>::operator = (const Ptr<Y>& o)
+{
+    Ptr(o).swap(*this);
+    return *this;
+}
+
+template<typename T>
+void Ptr<T>::release()
+{
+    if (owner) owner->decRef();
+    owner = NULL;
+    stored = NULL;
+}
+
+template<typename T>
+template<typename Y>
+void Ptr<T>::reset(Y* p)
+{
+    Ptr(p).swap(*this);
+}
+
+template<typename T>
+template<typename Y, typename D>
+void Ptr<T>::reset(Y* p, D d)
+{
+    Ptr(p, d).swap(*this);
+}
+
+template<typename T>
+void Ptr<T>::swap(Ptr<T>& o)
+{
+    std::swap(owner, o.owner);
+    std::swap(stored, o.stored);
+}
+
+template<typename T>
+T* Ptr<T>::get() const
+{
+    return stored;
+}
+
+template<typename T>
+typename detail::RefOrVoid<T>::type Ptr<T>::operator * () const
+{
+    return *stored;
+}
+
+template<typename T>
+T* Ptr<T>::operator -> () const
+{
+    return stored;
+}
+
+template<typename T>
+Ptr<T>::operator T* () const
+{
+    return stored;
+}
+
+
+template<typename T>
+bool Ptr<T>::empty() const
+{
+    return !stored;
+}
+
+template<typename T>
+template<typename Y>
+Ptr<Y> Ptr<T>::staticCast() const
+{
+    return Ptr<Y>(*this, static_cast<Y*>(stored));
+}
+
+template<typename T>
+template<typename Y>
+Ptr<Y> Ptr<T>::constCast() const
+{
+    return Ptr<Y>(*this, const_cast<Y*>(stored));
+}
+
+template<typename T>
+template<typename Y>
+Ptr<Y> Ptr<T>::dynamicCast() const
+{
+    return Ptr<Y>(*this, dynamic_cast<Y*>(stored));
+}
+
+template<typename T>
+void swap(Ptr<T>& ptr1, Ptr<T>& ptr2){
+    ptr1.swap(ptr2);
+}
+
+template<typename T>
+bool operator == (const Ptr<T>& ptr1, const Ptr<T>& ptr2)
+{
+    return ptr1.get() == ptr2.get();
+}
+
+template<typename T>
+bool operator != (const Ptr<T>& ptr1, const Ptr<T>& ptr2)
+{
+    return ptr1.get() != ptr2.get();
+}
+
+template<typename T>
+Ptr<T> makePtr()
+{
+    return Ptr<T>(new T());
+}
+
+template<typename T, typename A1>
+Ptr<T> makePtr(const A1& a1)
+{
+    return Ptr<T>(new T(a1));
+}
+
+template<typename T, typename A1, typename A2>
+Ptr<T> makePtr(const A1& a1, const A2& a2)
+{
+    return Ptr<T>(new T(a1, a2));
+}
+
+template<typename T, typename A1, typename A2, typename A3>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3)
+{
+    return Ptr<T>(new T(a1, a2, a3));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5, a6));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5, a6, a7));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5, a6, a7, a8));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8, typename A9>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5, a6, a7, a8, a9));
+}
+
+template<typename T, typename A1, typename A2, typename A3, typename A4, typename A5, typename A6, typename A7, typename A8, typename A9, typename A10>
+Ptr<T> makePtr(const A1& a1, const A2& a2, const A3& a3, const A4& a4, const A5& a5, const A6& a6, const A7& a7, const A8& a8, const A9& a9, const A10& a10)
+{
+    return Ptr<T>(new T(a1, a2, a3, a4, a5, a6, a7, a8, a9, a10));
+}
+
+} // namespace cv
+
+#endif // __OPENCV_CORE_PTR_INL_HPP__
diff --git a/modules/core/include/opencv2/core/traits.hpp b/modules/core/include/opencv2/core/traits.hpp
new file mode 100644
index 0000000000..85a5132f0e
--- /dev/null
+++ b/modules/core/include/opencv2/core/traits.hpp
@@ -0,0 +1,282 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_TRAITS_HPP__
+#define __OPENCV_CORE_TRAITS_HPP__
+
+#include "opencv2/core/cvdef.h"
+
+namespace cv
+{
+
+/*!
+   Informative template class for OpenCV "scalars".
+
+   The class is specialized for each primitive numerical type supported by OpenCV (such as unsigned char or float),
+   as well as for more complex types, like cv::Complex<>, std::complex<>, cv::Vec<> etc.
+   The common property of all such types (called "scalars", do not confuse it with cv::Scalar_)
+   is that each of them is basically a tuple of numbers of the same type. Each "scalar" can be represented
+   by the depth id (CV_8U ... CV_64F) and the number of channels.
+   OpenCV matrices, 2D or nD, dense or sparse, can store "scalars",
+   as long as the number of channels does not exceed CV_CN_MAX.
+*/
+template<typename _Tp> class DataType
+{
+public:
+    typedef _Tp         value_type;
+    typedef value_type  work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 1,
+           depth        = -1,
+           channels     = 1,
+           fmt          = 0,
+           type = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<bool>
+{
+public:
+    typedef bool        value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_8U,
+           channels     = 1,
+           fmt          = (int)'u',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<uchar>
+{
+public:
+    typedef uchar       value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_8U,
+           channels     = 1,
+           fmt          = (int)'u',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<schar>
+{
+public:
+    typedef schar       value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_8S,
+           channels     = 1,
+           fmt          = (int)'c',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<char>
+{
+public:
+    typedef schar       value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_8S,
+           channels     = 1,
+           fmt          = (int)'c',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<ushort>
+{
+public:
+    typedef ushort      value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_16U,
+           channels     = 1,
+           fmt          = (int)'w',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<short>
+{
+public:
+    typedef short       value_type;
+    typedef int         work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_16S,
+           channels     = 1,
+           fmt          = (int)'s',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<int>
+{
+public:
+    typedef int         value_type;
+    typedef value_type  work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_32S,
+           channels     = 1,
+           fmt          = (int)'i',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<float>
+{
+public:
+    typedef float       value_type;
+    typedef value_type  work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_32F,
+           channels     = 1,
+           fmt          = (int)'f',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+template<> class DataType<double>
+{
+public:
+    typedef double      value_type;
+    typedef value_type  work_type;
+    typedef value_type  channel_type;
+    typedef value_type  vec_type;
+    enum { generic_type = 0,
+           depth        = CV_64F,
+           channels     = 1,
+           fmt          = (int)'d',
+           type         = CV_MAKETYPE(depth, channels)
+         };
+};
+
+
+/*!
+  A helper class for cv::DataType
+
+  The class is specialized for each fundamental numerical data type supported by OpenCV.
+  It provides DataDepth<T>::value constant.
+*/
+template<typename _Tp> class DataDepth
+{
+public:
+    enum
+    {
+        value = DataType<_Tp>::depth,
+        fmt   = DataType<_Tp>::fmt
+    };
+};
+
+
+
+template<int _depth> class TypeDepth
+{
+    enum { depth = CV_USRTYPE1 };
+    typedef void value_type;
+};
+
+template<> class TypeDepth<CV_8U>
+{
+    enum { depth = CV_8U };
+    typedef uchar value_type;
+};
+
+template<> class TypeDepth<CV_8S>
+{
+    enum { depth = CV_8S };
+    typedef schar value_type;
+};
+
+template<> class TypeDepth<CV_16U>
+{
+    enum { depth = CV_16U };
+    typedef ushort value_type;
+};
+
+template<> class TypeDepth<CV_16S>
+{
+    enum { depth = CV_16S };
+    typedef short value_type;
+};
+
+template<> class TypeDepth<CV_32S>
+{
+    enum { depth = CV_32S };
+    typedef int value_type;
+};
+
+template<> class TypeDepth<CV_32F>
+{
+    enum { depth = CV_32F };
+    typedef float value_type;
+};
+
+template<> class TypeDepth<CV_64F>
+{
+    enum { depth = CV_64F };
+    typedef double value_type;
+};
+
+} // cv
+
+#endif // __OPENCV_CORE_TRAITS_HPP__
diff --git a/modules/core/include/opencv2/core/types.hpp b/modules/core/include/opencv2/core/types.hpp
new file mode 100644
index 0000000000..9252cad14a
--- /dev/null
+++ b/modules/core/include/opencv2/core/types.hpp
@@ -0,0 +1,1925 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_TYPES_HPP__
+#define __OPENCV_CORE_TYPES_HPP__
+
+#ifndef __cplusplus
+#  error types.hpp header must be compiled as C++
+#endif
+
+#include <climits>
+#include <cfloat>
+#include <vector>
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+#include "opencv2/core/matx.hpp"
+
+namespace cv
+{
+
+//////////////////////////////// Complex //////////////////////////////
+
+/*!
+  A complex number class.
+
+  The template class is similar and compatible with std::complex, however it provides slightly
+  more convenient access to the real and imaginary parts using through the simple field access, as opposite
+  to std::complex::real() and std::complex::imag().
+*/
+template<typename _Tp> class Complex
+{
+public:
+
+    //! constructors
+    Complex();
+    Complex( _Tp _re, _Tp _im = 0 );
+
+    //! conversion to another data type
+    template<typename T2> operator Complex<T2>() const;
+    //! conjugation
+    Complex conj() const;
+
+    _Tp re, im; //< the real and the imaginary parts
+};
+
+/*!
+  \typedef
+*/
+typedef Complex<float> Complexf;
+typedef Complex<double> Complexd;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Complex<_Tp> >
+{
+public:
+    typedef Complex<_Tp> value_type;
+    typedef value_type   work_type;
+    typedef _Tp          channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels) };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Point_ ////////////////////////////////
+
+/*!
+  template 2D point class.
+
+  The class defines a point in 2D space. Data type of the point coordinates is specified
+  as a template parameter. There are a few shorter aliases available for user convenience.
+  See cv::Point, cv::Point2i, cv::Point2f and cv::Point2d.
+*/
+template<typename _Tp> class Point_
+{
+public:
+    typedef _Tp value_type;
+
+    // various constructors
+    Point_();
+    Point_(_Tp _x, _Tp _y);
+    Point_(const Point_& pt);
+    Point_(const Size_<_Tp>& sz);
+    Point_(const Vec<_Tp, 2>& v);
+
+    Point_& operator = (const Point_& pt);
+    //! conversion to another data type
+    template<typename _Tp2> operator Point_<_Tp2>() const;
+
+    //! conversion to the old-style C structures
+    operator Vec<_Tp, 2>() const;
+
+    //! dot product
+    _Tp dot(const Point_& pt) const;
+    //! dot product computed in double-precision arithmetics
+    double ddot(const Point_& pt) const;
+    //! cross-product
+    double cross(const Point_& pt) const;
+    //! checks whether the point is inside the specified rectangle
+    bool inside(const Rect_<_Tp>& r) const;
+
+    _Tp x, y; //< the point coordinates
+};
+
+/*!
+  \typedef
+*/
+typedef Point_<int> Point2i;
+typedef Point_<float> Point2f;
+typedef Point_<double> Point2d;
+typedef Point2i Point;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Point_<_Tp> >
+{
+public:
+    typedef Point_<_Tp>                               value_type;
+    typedef Point_<typename DataType<_Tp>::work_type> work_type;
+    typedef _Tp                                       channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Point3_ ////////////////////////////////
+
+/*!
+  template 3D point class.
+
+  The class defines a point in 3D space. Data type of the point coordinates is specified
+  as a template parameter.
+
+  \see cv::Point3i, cv::Point3f and cv::Point3d
+*/
+template<typename _Tp> class Point3_
+{
+public:
+    typedef _Tp value_type;
+
+    // various constructors
+    Point3_();
+    Point3_(_Tp _x, _Tp _y, _Tp _z);
+    Point3_(const Point3_& pt);
+    explicit Point3_(const Point_<_Tp>& pt);
+    Point3_(const Vec<_Tp, 3>& v);
+
+    Point3_& operator = (const Point3_& pt);
+    //! conversion to another data type
+    template<typename _Tp2> operator Point3_<_Tp2>() const;
+    //! conversion to cv::Vec<>
+    operator Vec<_Tp, 3>() const;
+
+    //! dot product
+    _Tp dot(const Point3_& pt) const;
+    //! dot product computed in double-precision arithmetics
+    double ddot(const Point3_& pt) const;
+    //! cross product of the 2 3D points
+    Point3_ cross(const Point3_& pt) const;
+
+    _Tp x, y, z; //< the point coordinates
+};
+
+/*!
+  \typedef
+*/
+typedef Point3_<int> Point3i;
+typedef Point3_<float> Point3f;
+typedef Point3_<double> Point3d;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Point3_<_Tp> >
+{
+public:
+    typedef Point3_<_Tp>                               value_type;
+    typedef Point3_<typename DataType<_Tp>::work_type> work_type;
+    typedef _Tp                                        channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 3,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Size_ ////////////////////////////////
+
+/*!
+  The 2D size class
+
+  The class represents the size of a 2D rectangle, image size, matrix size etc.
+  Normally, cv::Size ~ cv::Size_<int> is used.
+*/
+template<typename _Tp> class Size_
+{
+public:
+    typedef _Tp value_type;
+
+    //! various constructors
+    Size_();
+    Size_(_Tp _width, _Tp _height);
+    Size_(const Size_& sz);
+    Size_(const Point_<_Tp>& pt);
+
+    Size_& operator = (const Size_& sz);
+    //! the area (width*height)
+    _Tp area() const;
+
+    //! conversion of another data type.
+    template<typename _Tp2> operator Size_<_Tp2>() const;
+
+    _Tp width, height; // the width and the height
+};
+
+/*!
+  \typedef
+*/
+typedef Size_<int> Size2i;
+typedef Size_<float> Size2f;
+typedef Size2i Size;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Size_<_Tp> >
+{
+public:
+    typedef Size_<_Tp>                               value_type;
+    typedef Size_<typename DataType<_Tp>::work_type> work_type;
+    typedef _Tp                                      channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Rect_ ////////////////////////////////
+
+/*!
+  The 2D up-right rectangle class
+
+  The class represents a 2D rectangle with coordinates of the specified data type.
+  Normally, cv::Rect ~ cv::Rect_<int> is used.
+*/
+template<typename _Tp> class Rect_
+{
+public:
+    typedef _Tp value_type;
+
+    //! various constructors
+    Rect_();
+    Rect_(_Tp _x, _Tp _y, _Tp _width, _Tp _height);
+    Rect_(const Rect_& r);
+    Rect_(const Point_<_Tp>& org, const Size_<_Tp>& sz);
+    Rect_(const Point_<_Tp>& pt1, const Point_<_Tp>& pt2);
+
+    Rect_& operator = ( const Rect_& r );
+    //! the top-left corner
+    Point_<_Tp> tl() const;
+    //! the bottom-right corner
+    Point_<_Tp> br() const;
+
+    //! size (width, height) of the rectangle
+    Size_<_Tp> size() const;
+    //! area (width*height) of the rectangle
+    _Tp area() const;
+
+    //! conversion to another data type
+    template<typename _Tp2> operator Rect_<_Tp2>() const;
+
+    //! checks whether the rectangle contains the point
+    bool contains(const Point_<_Tp>& pt) const;
+
+    _Tp x, y, width, height; //< the top-left corner, as well as width and height of the rectangle
+};
+
+/*!
+  \typedef
+*/
+typedef Rect_<int> Rect;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Rect_<_Tp> >
+{
+public:
+    typedef Rect_<_Tp>                               value_type;
+    typedef Rect_<typename DataType<_Tp>::work_type> work_type;
+    typedef _Tp                                      channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 4,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+///////////////////////////// RotatedRect /////////////////////////////
+
+/*!
+  The rotated 2D rectangle.
+
+  The class represents rotated (i.e. not up-right) rectangles on a plane.
+  Each rectangle is described by the center point (mass center), length of each side
+  (represented by cv::Size2f structure) and the rotation angle in degrees.
+*/
+class CV_EXPORTS RotatedRect
+{
+public:
+    //! various constructors
+    RotatedRect();
+    RotatedRect(const Point2f& center, const Size2f& size, float angle);
+
+    //! returns 4 vertices of the rectangle
+    void points(Point2f pts[]) const;
+    //! returns the minimal up-right rectangle containing the rotated rectangle
+    Rect boundingRect() const;
+
+    Point2f center; //< the rectangle mass center
+    Size2f size;    //< width and height of the rectangle
+    float angle;    //< the rotation angle. When the angle is 0, 90, 180, 270 etc., the rectangle becomes an up-right rectangle.
+};
+
+/*!
+  traits
+*/
+template<> class DataType< RotatedRect >
+{
+public:
+    typedef RotatedRect  value_type;
+    typedef value_type   work_type;
+    typedef float        channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = (int)sizeof(value_type)/sizeof(channel_type), // 5
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Range /////////////////////////////////
+
+/*!
+   The 2D range class
+
+   This is the class used to specify a continuous subsequence, i.e. part of a contour, or a column span in a matrix.
+*/
+class CV_EXPORTS Range
+{
+public:
+    Range();
+    Range(int _start, int _end);
+    int size() const;
+    bool empty() const;
+    static Range all();
+
+    int start, end;
+};
+
+/*!
+  traits
+*/
+template<> class DataType<Range>
+{
+public:
+    typedef Range      value_type;
+    typedef value_type work_type;
+    typedef int        channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// Scalar_ ///////////////////////////////
+
+/*!
+   The template scalar class.
+
+   This is partially specialized cv::Vec class with the number of elements = 4, i.e. a short vector of four elements.
+   Normally, cv::Scalar ~ cv::Scalar_<double> is used.
+*/
+template<typename _Tp> class Scalar_ : public Vec<_Tp, 4>
+{
+public:
+    //! various constructors
+    Scalar_();
+    Scalar_(_Tp v0, _Tp v1, _Tp v2=0, _Tp v3=0);
+    Scalar_(_Tp v0);
+
+    template<typename _Tp2, int cn>
+    Scalar_(const Vec<_Tp2, cn>& v);
+
+    //! returns a scalar with all elements set to v0
+    static Scalar_<_Tp> all(_Tp v0);
+
+    //! conversion to another data type
+    template<typename T2> operator Scalar_<T2>() const;
+
+    //! per-element product
+    Scalar_<_Tp> mul(const Scalar_<_Tp>& t, double scale=1 ) const;
+
+    // returns (v0, -v1, -v2, -v3)
+    Scalar_<_Tp> conj() const;
+
+    // returns true iff v1 == v2 == v3 == 0
+    bool isReal() const;
+};
+
+/*!
+  \typedef
+*/
+typedef Scalar_<double> Scalar;
+
+/*!
+  traits
+*/
+template<typename _Tp> class DataType< Scalar_<_Tp> >
+{
+public:
+    typedef Scalar_<_Tp>                               value_type;
+    typedef Scalar_<typename DataType<_Tp>::work_type> work_type;
+    typedef _Tp                                        channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 4,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+/////////////////////////////// KeyPoint ////////////////////////////////
+
+/*!
+ The Keypoint Class
+
+ The class instance stores a keypoint, i.e. a point feature found by one of many available keypoint detectors, such as
+ Harris corner detector, cv::FAST, cv::StarDetector, cv::SURF, cv::SIFT, cv::LDetector etc.
+
+ The keypoint is characterized by the 2D position, scale
+ (proportional to the diameter of the neighborhood that needs to be taken into account),
+ orientation and some other parameters. The keypoint neighborhood is then analyzed by another algorithm that builds a descriptor
+ (usually represented as a feature vector). The keypoints representing the same object in different images can then be matched using
+ cv::KDTree or another method.
+*/
+class CV_EXPORTS_W_SIMPLE KeyPoint
+{
+public:
+    //! the default constructor
+    CV_WRAP KeyPoint();
+    //! the full constructor
+    KeyPoint(Point2f _pt, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1);
+    //! another form of the full constructor
+    CV_WRAP KeyPoint(float x, float y, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1);
+
+    size_t hash() const;
+
+    //! converts vector of keypoints to vector of points
+    CV_WRAP static void convert(const std::vector<KeyPoint>& keypoints,
+                                CV_OUT std::vector<Point2f>& points2f,
+                                const std::vector<int>& keypointIndexes=std::vector<int>());
+    //! converts vector of points to the vector of keypoints, where each keypoint is assigned the same size and the same orientation
+    CV_WRAP static void convert(const std::vector<Point2f>& points2f,
+                                CV_OUT std::vector<KeyPoint>& keypoints,
+                                float size=1, float response=1, int octave=0, int class_id=-1);
+
+    //! computes overlap for pair of keypoints;
+    //! overlap is a ratio between area of keypoint regions intersection and
+    //! area of keypoint regions union (now keypoint region is circle)
+    CV_WRAP static float overlap(const KeyPoint& kp1, const KeyPoint& kp2);
+
+    CV_PROP_RW Point2f pt; //!< coordinates of the keypoints
+    CV_PROP_RW float size; //!< diameter of the meaningful keypoint neighborhood
+    CV_PROP_RW float angle; //!< computed orientation of the keypoint (-1 if not applicable);
+                            //!< it's in [0,360) degrees and measured relative to
+                            //!< image coordinate system, ie in clockwise.
+    CV_PROP_RW float response; //!< the response by which the most strong keypoints have been selected. Can be used for the further sorting or subsampling
+    CV_PROP_RW int octave; //!< octave (pyramid layer) from which the keypoint has been extracted
+    CV_PROP_RW int class_id; //!< object class (if the keypoints need to be clustered by an object they belong to)
+};
+
+/*!
+  traits
+*/
+template<> class DataType<KeyPoint>
+{
+public:
+    typedef KeyPoint      value_type;
+    typedef float         work_type;
+    typedef float         channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = (int)(sizeof(value_type)/sizeof(channel_type)), // 7
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+//////////////////////////////// DMatch /////////////////////////////////
+
+/*
+ * Struct for matching: query descriptor index, train descriptor index, train image index and distance between descriptors.
+ */
+class CV_EXPORTS_W_SIMPLE DMatch
+{
+public:
+    CV_WRAP DMatch();
+    CV_WRAP DMatch(int _queryIdx, int _trainIdx, float _distance);
+    CV_WRAP DMatch(int _queryIdx, int _trainIdx, int _imgIdx, float _distance);
+
+    CV_PROP_RW int queryIdx; // query descriptor index
+    CV_PROP_RW int trainIdx; // train descriptor index
+    CV_PROP_RW int imgIdx;   // train image index
+
+    CV_PROP_RW float distance;
+
+    // less is better
+    bool operator<(const DMatch &m) const;
+};
+
+/*!
+  traits
+*/
+template<> class DataType<DMatch>
+{
+public:
+    typedef DMatch      value_type;
+    typedef int         work_type;
+    typedef int         channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = (int)(sizeof(value_type)/sizeof(channel_type)), // 4
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+///////////////////////////// TermCriteria //////////////////////////////
+
+/*!
+ Termination criteria in iterative algorithms
+ */
+class CV_EXPORTS TermCriteria
+{
+public:
+    enum
+    {
+        COUNT=1, //!< the maximum number of iterations or elements to compute
+        MAX_ITER=COUNT, //!< ditto
+        EPS=2 //!< the desired accuracy or change in parameters at which the iterative algorithm stops
+    };
+
+    //! default constructor
+    TermCriteria();
+    //! full constructor
+    TermCriteria(int type, int maxCount, double epsilon);
+
+    int type; //!< the type of termination criteria: COUNT, EPS or COUNT + EPS
+    int maxCount; // the maximum number of iterations/elements
+    double epsilon; // the desired accuracy
+};
+
+
+
+///////////////////////// raster image moments //////////////////////////
+
+class CV_EXPORTS_W_MAP Moments
+{
+public:
+    //! the default constructor
+    Moments();
+    //! the full constructor
+    Moments(double m00, double m10, double m01, double m20, double m11,
+            double m02, double m30, double m21, double m12, double m03 );
+    ////! the conversion from CvMoments
+    //Moments( const CvMoments& moments );
+    ////! the conversion to CvMoments
+    //operator CvMoments() const;
+
+    //! spatial moments
+    CV_PROP_RW double  m00, m10, m01, m20, m11, m02, m30, m21, m12, m03;
+    //! central moments
+    CV_PROP_RW double  mu20, mu11, mu02, mu30, mu21, mu12, mu03;
+    //! central normalized moments
+    CV_PROP_RW double  nu20, nu11, nu02, nu30, nu21, nu12, nu03;
+};
+
+/*!
+  traits
+*/
+template<> class DataType<Moments>
+{
+public:
+    typedef Moments     value_type;
+    typedef double      work_type;
+    typedef double      channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = (int)(sizeof(value_type)/sizeof(channel_type)), // 24
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels)
+         };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+
+
+/////////////////////////////////////////////////////////////////////////
+///////////////////////////// Implementation ////////////////////////////
+/////////////////////////////////////////////////////////////////////////
+
+//////////////////////////////// Complex ////////////////////////////////
+
+template<typename _Tp> inline
+Complex<_Tp>::Complex()
+    : re(0), im(0) {}
+
+template<typename _Tp> inline
+Complex<_Tp>::Complex( _Tp _re, _Tp _im )
+    : re(_re), im(_im) {}
+
+template<typename _Tp> template<typename T2> inline
+Complex<_Tp>::operator Complex<T2>() const
+{
+    return Complex<T2>(saturate_cast<T2>(re), saturate_cast<T2>(im));
+}
+
+template<typename _Tp> inline
+Complex<_Tp> Complex<_Tp>::conj() const
+{
+    return Complex<_Tp>(re, -im);
+}
+
+
+template<typename _Tp> static inline
+bool operator == (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return a.re == b.re && a.im == b.im;
+}
+
+template<typename _Tp> static inline
+bool operator != (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return a.re != b.re || a.im != b.im;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator + (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return Complex<_Tp>( a.re + b.re, a.im + b.im );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator += (Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    a.re += b.re; a.im += b.im;
+    return a;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator - (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return Complex<_Tp>( a.re - b.re, a.im - b.im );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator -= (Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    a.re -= b.re; a.im -= b.im;
+    return a;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator - (const Complex<_Tp>& a)
+{
+    return Complex<_Tp>(-a.re, -a.im);
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator * (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return Complex<_Tp>( a.re*b.re - a.im*b.im, a.re*b.im + a.im*b.re );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator * (const Complex<_Tp>& a, _Tp b)
+{
+    return Complex<_Tp>( a.re*b, a.im*b );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator * (_Tp b, const Complex<_Tp>& a)
+{
+    return Complex<_Tp>( a.re*b, a.im*b );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator + (const Complex<_Tp>& a, _Tp b)
+{
+    return Complex<_Tp>( a.re + b, a.im );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator - (const Complex<_Tp>& a, _Tp b)
+{ return Complex<_Tp>( a.re - b, a.im ); }
+
+template<typename _Tp> static inline
+Complex<_Tp> operator + (_Tp b, const Complex<_Tp>& a)
+{
+    return Complex<_Tp>( a.re + b, a.im );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator - (_Tp b, const Complex<_Tp>& a)
+{
+    return Complex<_Tp>( b - a.re, -a.im );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator += (Complex<_Tp>& a, _Tp b)
+{
+    a.re += b; return a;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator -= (Complex<_Tp>& a, _Tp b)
+{
+    a.re -= b; return a;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator *= (Complex<_Tp>& a, _Tp b)
+{
+    a.re *= b; a.im *= b; return a;
+}
+
+template<typename _Tp> static inline
+double abs(const Complex<_Tp>& a)
+{
+    return std::sqrt( (double)a.re*a.re + (double)a.im*a.im);
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator / (const Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    double t = 1./((double)b.re*b.re + (double)b.im*b.im);
+    return Complex<_Tp>( (_Tp)((a.re*b.re + a.im*b.im)*t),
+                        (_Tp)((-a.re*b.im + a.im*b.re)*t) );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp>& operator /= (Complex<_Tp>& a, const Complex<_Tp>& b)
+{
+    return (a = a / b);
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator / (const Complex<_Tp>& a, _Tp b)
+{
+    _Tp t = (_Tp)1/b;
+    return Complex<_Tp>( a.re*t, a.im*t );
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator / (_Tp b, const Complex<_Tp>& a)
+{
+    return Complex<_Tp>(b)/a;
+}
+
+template<typename _Tp> static inline
+Complex<_Tp> operator /= (const Complex<_Tp>& a, _Tp b)
+{
+    _Tp t = (_Tp)1/b;
+    a.re *= t; a.im *= t; return a;
+}
+
+
+
+//////////////////////////////// 2D Point ///////////////////////////////
+
+template<typename _Tp> inline
+Point_<_Tp>::Point_()
+    : x(0), y(0) {}
+
+template<typename _Tp> inline
+Point_<_Tp>::Point_(_Tp _x, _Tp _y)
+    : x(_x), y(_y) {}
+
+template<typename _Tp> inline
+Point_<_Tp>::Point_(const Point_& pt)
+    : x(pt.x), y(pt.y) {}
+
+template<typename _Tp> inline
+Point_<_Tp>::Point_(const Size_<_Tp>& sz)
+    : x(sz.width), y(sz.height) {}
+
+template<typename _Tp> inline
+Point_<_Tp>::Point_(const Vec<_Tp,2>& v)
+    : x(v[0]), y(v[1]) {}
+
+template<typename _Tp> inline
+Point_<_Tp>& Point_<_Tp>::operator = (const Point_& pt)
+{
+    x = pt.x; y = pt.y;
+    return *this;
+}
+
+template<typename _Tp> template<typename _Tp2> inline
+Point_<_Tp>::operator Point_<_Tp2>() const
+{
+    return Point_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y));
+}
+
+template<typename _Tp> inline
+Point_<_Tp>::operator Vec<_Tp, 2>() const
+{
+    return Vec<_Tp, 2>(x, y);
+}
+
+template<typename _Tp> inline
+_Tp Point_<_Tp>::dot(const Point_& pt) const
+{
+    return saturate_cast<_Tp>(x*pt.x + y*pt.y);
+}
+
+template<typename _Tp> inline
+double Point_<_Tp>::ddot(const Point_& pt) const
+{
+    return (double)x*pt.x + (double)y*pt.y;
+}
+
+template<typename _Tp> inline
+double Point_<_Tp>::cross(const Point_& pt) const
+{
+    return (double)x*pt.y - (double)y*pt.x;
+}
+
+template<typename _Tp> inline bool
+Point_<_Tp>::inside( const Rect_<_Tp>& r ) const
+{
+    return r.contains(*this);
+}
+
+
+template<typename _Tp> static inline
+Point_<_Tp>& operator += (Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    a.x += b.x;
+    a.y += b.y;
+    return a;
+}
+
+template<typename _Tp> static inline
+Point_<_Tp>& operator -= (Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    a.x -= b.x;
+    a.y -= b.y;
+    return a;
+}
+
+template<typename _Tp> static inline
+Point_<_Tp>& operator *= (Point_<_Tp>& a, int b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+Point_<_Tp>& operator *= (Point_<_Tp>& a, float b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+Point_<_Tp>& operator *= (Point_<_Tp>& a, double b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+double norm(const Point_<_Tp>& pt)
+{
+    return std::sqrt((double)pt.x*pt.x + (double)pt.y*pt.y);
+}
+
+template<typename _Tp> static inline
+bool operator == (const Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return a.x == b.x && a.y == b.y;
+}
+
+template<typename _Tp> static inline
+bool operator != (const Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return a.x != b.x || a.y != b.y;
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator + (const Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(a.x + b.x), saturate_cast<_Tp>(a.y + b.y) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator - (const Point_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(a.x - b.x), saturate_cast<_Tp>(a.y - b.y) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator - (const Point_<_Tp>& a)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(-a.x), saturate_cast<_Tp>(-a.y) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (const Point_<_Tp>& a, int b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (int a, const Point_<_Tp>& b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (const Point_<_Tp>& a, float b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (float a, const Point_<_Tp>& b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (const Point_<_Tp>& a, double b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (double a, const Point_<_Tp>& b)
+{
+    return Point_<_Tp>( saturate_cast<_Tp>(b.x*a), saturate_cast<_Tp>(b.y*a) );
+}
+
+template<typename _Tp> static inline
+Point_<_Tp> operator * (const Matx<_Tp, 2, 2>& a, const Point_<_Tp>& b)
+{
+    Matx<_Tp, 2, 1> tmp = a * Vec<_Tp,2>(b.x, b.y);
+    return Point_<_Tp>(tmp.val[0], tmp.val[1]);
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (const Matx<_Tp, 3, 3>& a, const Point_<_Tp>& b)
+{
+    Matx<_Tp, 3, 1> tmp = a * Vec<_Tp,3>(b.x, b.y, 1);
+    return Point3_<_Tp>(tmp.val[0], tmp.val[1], tmp.val[2]);
+}
+
+
+
+//////////////////////////////// 3D Point ///////////////////////////////
+
+template<typename _Tp> inline
+Point3_<_Tp>::Point3_()
+    : x(0), y(0), z(0) {}
+
+template<typename _Tp> inline
+Point3_<_Tp>::Point3_(_Tp _x, _Tp _y, _Tp _z)
+    : x(_x), y(_y), z(_z) {}
+
+template<typename _Tp> inline
+Point3_<_Tp>::Point3_(const Point3_& pt)
+    : x(pt.x), y(pt.y), z(pt.z) {}
+
+template<typename _Tp> inline
+Point3_<_Tp>::Point3_(const Point_<_Tp>& pt)
+    : x(pt.x), y(pt.y), z(_Tp()) {}
+
+template<typename _Tp> inline
+Point3_<_Tp>::Point3_(const Vec<_Tp, 3>& v)
+    : x(v[0]), y(v[1]), z(v[2]) {}
+
+template<typename _Tp> template<typename _Tp2> inline
+Point3_<_Tp>::operator Point3_<_Tp2>() const
+{
+    return Point3_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y), saturate_cast<_Tp2>(z));
+}
+
+template<typename _Tp> inline
+Point3_<_Tp>::operator Vec<_Tp, 3>() const
+{
+    return Vec<_Tp, 3>(x, y, z);
+}
+
+template<typename _Tp> inline
+Point3_<_Tp>& Point3_<_Tp>::operator = (const Point3_& pt)
+{
+    x = pt.x; y = pt.y; z = pt.z;
+    return *this;
+}
+
+template<typename _Tp> inline
+_Tp Point3_<_Tp>::dot(const Point3_& pt) const
+{
+    return saturate_cast<_Tp>(x*pt.x + y*pt.y + z*pt.z);
+}
+
+template<typename _Tp> inline
+double Point3_<_Tp>::ddot(const Point3_& pt) const
+{
+    return (double)x*pt.x + (double)y*pt.y + (double)z*pt.z;
+}
+
+template<typename _Tp> inline
+Point3_<_Tp> Point3_<_Tp>::cross(const Point3_<_Tp>& pt) const
+{
+    return Point3_<_Tp>(y*pt.z - z*pt.y, z*pt.x - x*pt.z, x*pt.y - y*pt.x);
+}
+
+
+template<typename _Tp> static inline
+Point3_<_Tp>& operator += (Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    a.x += b.x;
+    a.y += b.y;
+    a.z += b.z;
+    return a;
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp>& operator -= (Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    a.x -= b.x;
+    a.y -= b.y;
+    a.z -= b.z;
+    return a;
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp>& operator *= (Point3_<_Tp>& a, int b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    a.z = saturate_cast<_Tp>(a.z * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp>& operator *= (Point3_<_Tp>& a, float b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    a.z = saturate_cast<_Tp>(a.z * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp>& operator *= (Point3_<_Tp>& a, double b)
+{
+    a.x = saturate_cast<_Tp>(a.x * b);
+    a.y = saturate_cast<_Tp>(a.y * b);
+    a.z = saturate_cast<_Tp>(a.z * b);
+    return a;
+}
+
+template<typename _Tp> static inline
+double norm(const Point3_<_Tp>& pt)
+{
+    return std::sqrt((double)pt.x*pt.x + (double)pt.y*pt.y + (double)pt.z*pt.z);
+}
+
+template<typename _Tp> static inline
+bool operator == (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    return a.x == b.x && a.y == b.y && a.z == b.z;
+}
+
+template<typename _Tp> static inline
+bool operator != (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    return a.x != b.x || a.y != b.y || a.z != b.z;
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator + (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(a.x + b.x), saturate_cast<_Tp>(a.y + b.y), saturate_cast<_Tp>(a.z + b.z));
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator - (const Point3_<_Tp>& a, const Point3_<_Tp>& b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(a.x - b.x), saturate_cast<_Tp>(a.y - b.y), saturate_cast<_Tp>(a.z - b.z));
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator - (const Point3_<_Tp>& a)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(-a.x), saturate_cast<_Tp>(-a.y), saturate_cast<_Tp>(-a.z) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (const Point3_<_Tp>& a, int b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(a.x*b), saturate_cast<_Tp>(a.y*b), saturate_cast<_Tp>(a.z*b) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (int a, const Point3_<_Tp>& b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(b.x * a), saturate_cast<_Tp>(b.y * a), saturate_cast<_Tp>(b.z * a) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (const Point3_<_Tp>& a, float b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(a.x * b), saturate_cast<_Tp>(a.y * b), saturate_cast<_Tp>(a.z * b) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (float a, const Point3_<_Tp>& b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(b.x * a), saturate_cast<_Tp>(b.y * a), saturate_cast<_Tp>(b.z * a) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (const Point3_<_Tp>& a, double b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(a.x * b), saturate_cast<_Tp>(a.y * b), saturate_cast<_Tp>(a.z * b) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (double a, const Point3_<_Tp>& b)
+{
+    return Point3_<_Tp>( saturate_cast<_Tp>(b.x * a), saturate_cast<_Tp>(b.y * a), saturate_cast<_Tp>(b.z * a) );
+}
+
+template<typename _Tp> static inline
+Point3_<_Tp> operator * (const Matx<_Tp, 3, 3>& a, const Point3_<_Tp>& b)
+{
+    Matx<_Tp, 3, 1> tmp = a * Vec<_Tp,3>(b.x, b.y, b.z);
+    return Point3_<_Tp>(tmp.val[0], tmp.val[1], tmp.val[2]);
+}
+
+template<typename _Tp> static inline
+Matx<_Tp, 4, 1> operator * (const Matx<_Tp, 4, 4>& a, const Point3_<_Tp>& b)
+{
+    return a * Matx<_Tp, 4, 1>(b.x, b.y, b.z, 1);
+}
+
+
+
+////////////////////////////////// Size /////////////////////////////////
+
+template<typename _Tp> inline
+Size_<_Tp>::Size_()
+    : width(0), height(0) {}
+
+template<typename _Tp> inline
+Size_<_Tp>::Size_(_Tp _width, _Tp _height)
+    : width(_width), height(_height) {}
+
+template<typename _Tp> inline
+Size_<_Tp>::Size_(const Size_& sz)
+    : width(sz.width), height(sz.height) {}
+
+template<typename _Tp> inline
+Size_<_Tp>::Size_(const Point_<_Tp>& pt)
+    : width(pt.x), height(pt.y) {}
+
+template<typename _Tp> template<typename _Tp2> inline
+Size_<_Tp>::operator Size_<_Tp2>() const
+{
+    return Size_<_Tp2>(saturate_cast<_Tp2>(width), saturate_cast<_Tp2>(height));
+}
+
+template<typename _Tp> inline
+Size_<_Tp>& Size_<_Tp>::operator = (const Size_<_Tp>& sz)
+{
+    width = sz.width; height = sz.height;
+    return *this;
+}
+
+template<typename _Tp> inline
+_Tp Size_<_Tp>::area() const
+{
+    return width * height;
+}
+
+
+template<typename _Tp> static inline
+Size_<_Tp> operator * (const Size_<_Tp>& a, _Tp b)
+{
+    return Size_<_Tp>(a.width * b, a.height * b);
+}
+
+template<typename _Tp> static inline
+Size_<_Tp> operator + (const Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    return Size_<_Tp>(a.width + b.width, a.height + b.height);
+}
+
+template<typename _Tp> static inline
+Size_<_Tp> operator - (const Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    return Size_<_Tp>(a.width - b.width, a.height - b.height);
+}
+
+template<typename _Tp> static inline
+Size_<_Tp>& operator += (Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    a.width += b.width;
+    a.height += b.height;
+    return a;
+}
+
+template<typename _Tp> static inline
+Size_<_Tp>& operator -= (Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    a.width -= b.width;
+    a.height -= b.height;
+    return a;
+}
+
+template<typename _Tp> static inline
+bool operator == (const Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    return a.width == b.width && a.height == b.height;
+}
+
+template<typename _Tp> static inline
+bool operator != (const Size_<_Tp>& a, const Size_<_Tp>& b)
+{
+    return a.width != b.width || a.height != b.height;
+}
+
+
+
+////////////////////////////////// Rect /////////////////////////////////
+
+template<typename _Tp> inline
+Rect_<_Tp>::Rect_()
+    : x(0), y(0), width(0), height(0) {}
+
+template<typename _Tp> inline
+Rect_<_Tp>::Rect_(_Tp _x, _Tp _y, _Tp _width, _Tp _height)
+    : x(_x), y(_y), width(_width), height(_height) {}
+
+template<typename _Tp> inline
+Rect_<_Tp>::Rect_(const Rect_<_Tp>& r)
+    : x(r.x), y(r.y), width(r.width), height(r.height) {}
+
+template<typename _Tp> inline
+Rect_<_Tp>::Rect_(const Point_<_Tp>& org, const Size_<_Tp>& sz)
+    : x(org.x), y(org.y), width(sz.width), height(sz.height) {}
+
+template<typename _Tp> inline
+Rect_<_Tp>::Rect_(const Point_<_Tp>& pt1, const Point_<_Tp>& pt2)
+{
+    x = std::min(pt1.x, pt2.x);
+    y = std::min(pt1.y, pt2.y);
+    width = std::max(pt1.x, pt2.x) - x;
+    height = std::max(pt1.y, pt2.y) - y;
+}
+
+template<typename _Tp> inline
+Rect_<_Tp>& Rect_<_Tp>::operator = ( const Rect_<_Tp>& r )
+{
+    x = r.x;
+    y = r.y;
+    width = r.width;
+    height = r.height;
+    return *this;
+}
+
+template<typename _Tp> inline
+Point_<_Tp> Rect_<_Tp>::tl() const
+{
+    return Point_<_Tp>(x,y);
+}
+
+template<typename _Tp> inline
+Point_<_Tp> Rect_<_Tp>::br() const
+{
+    return Point_<_Tp>(x + width, y + height);
+}
+
+template<typename _Tp> inline
+Size_<_Tp> Rect_<_Tp>::size() const
+{
+    return Size_<_Tp>(width, height);
+}
+
+template<typename _Tp> inline
+_Tp Rect_<_Tp>::area() const
+{
+    return width * height;
+}
+
+template<typename _Tp> template<typename _Tp2> inline
+Rect_<_Tp>::operator Rect_<_Tp2>() const
+{
+    return Rect_<_Tp2>(saturate_cast<_Tp2>(x), saturate_cast<_Tp2>(y), saturate_cast<_Tp2>(width), saturate_cast<_Tp2>(height));
+}
+
+template<typename _Tp> inline
+bool Rect_<_Tp>::contains(const Point_<_Tp>& pt) const
+{
+    return x <= pt.x && pt.x < x + width && y <= pt.y && pt.y < y + height;
+}
+
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator += ( Rect_<_Tp>& a, const Point_<_Tp>& b )
+{
+    a.x += b.x;
+    a.y += b.y;
+    return a;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator -= ( Rect_<_Tp>& a, const Point_<_Tp>& b )
+{
+    a.x -= b.x;
+    a.y -= b.y;
+    return a;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator += ( Rect_<_Tp>& a, const Size_<_Tp>& b )
+{
+    a.width += b.width;
+    a.height += b.height;
+    return a;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator -= ( Rect_<_Tp>& a, const Size_<_Tp>& b )
+{
+    a.width -= b.width;
+    a.height -= b.height;
+    return a;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator &= ( Rect_<_Tp>& a, const Rect_<_Tp>& b )
+{
+    _Tp x1 = std::max(a.x, b.x);
+    _Tp y1 = std::max(a.y, b.y);
+    a.width = std::min(a.x + a.width, b.x + b.width) - x1;
+    a.height = std::min(a.y + a.height, b.y + b.height) - y1;
+    a.x = x1;
+    a.y = y1;
+    if( a.width <= 0 || a.height <= 0 )
+        a = Rect();
+    return a;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp>& operator |= ( Rect_<_Tp>& a, const Rect_<_Tp>& b )
+{
+    _Tp x1 = std::min(a.x, b.x);
+    _Tp y1 = std::min(a.y, b.y);
+    a.width = std::max(a.x + a.width, b.x + b.width) - x1;
+    a.height = std::max(a.y + a.height, b.y + b.height) - y1;
+    a.x = x1;
+    a.y = y1;
+    return a;
+}
+
+template<typename _Tp> static inline
+bool operator == (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
+{
+    return a.x == b.x && a.y == b.y && a.width == b.width && a.height == b.height;
+}
+
+template<typename _Tp> static inline
+bool operator != (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
+{
+    return a.x != b.x || a.y != b.y || a.width != b.width || a.height != b.height;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp> operator + (const Rect_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return Rect_<_Tp>( a.x + b.x, a.y + b.y, a.width, a.height );
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp> operator - (const Rect_<_Tp>& a, const Point_<_Tp>& b)
+{
+    return Rect_<_Tp>( a.x - b.x, a.y - b.y, a.width, a.height );
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp> operator + (const Rect_<_Tp>& a, const Size_<_Tp>& b)
+{
+    return Rect_<_Tp>( a.x, a.y, a.width + b.width, a.height + b.height );
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp> operator & (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
+{
+    Rect_<_Tp> c = a;
+    return c &= b;
+}
+
+template<typename _Tp> static inline
+Rect_<_Tp> operator | (const Rect_<_Tp>& a, const Rect_<_Tp>& b)
+{
+    Rect_<_Tp> c = a;
+    return c |= b;
+}
+
+
+
+////////////////////////////// RotatedRect //////////////////////////////
+
+inline
+RotatedRect::RotatedRect()
+    : center(), size(), angle(0) {}
+
+inline
+RotatedRect::RotatedRect(const Point2f& _center, const Size2f& _size, float _angle)
+    : center(_center), size(_size), angle(_angle) {}
+
+
+
+///////////////////////////////// Range /////////////////////////////////
+
+inline
+Range::Range()
+    : start(0), end(0) {}
+
+inline
+Range::Range(int _start, int _end)
+    : start(_start), end(_end) {}
+
+inline
+int Range::size() const
+{
+    return end - start;
+}
+
+inline
+bool Range::empty() const
+{
+    return start == end;
+}
+
+inline
+Range Range::all()
+{
+    return Range(INT_MIN, INT_MAX);
+}
+
+
+static inline
+bool operator == (const Range& r1, const Range& r2)
+{
+    return r1.start == r2.start && r1.end == r2.end;
+}
+
+static inline
+bool operator != (const Range& r1, const Range& r2)
+{
+    return !(r1 == r2);
+}
+
+static inline
+bool operator !(const Range& r)
+{
+    return r.start == r.end;
+}
+
+static inline
+Range operator & (const Range& r1, const Range& r2)
+{
+    Range r(std::max(r1.start, r2.start), std::min(r1.end, r2.end));
+    r.end = std::max(r.end, r.start);
+    return r;
+}
+
+static inline
+Range& operator &= (Range& r1, const Range& r2)
+{
+    r1 = r1 & r2;
+    return r1;
+}
+
+static inline
+Range operator + (const Range& r1, int delta)
+{
+    return Range(r1.start + delta, r1.end + delta);
+}
+
+static inline
+Range operator + (int delta, const Range& r1)
+{
+    return Range(r1.start + delta, r1.end + delta);
+}
+
+static inline
+Range operator - (const Range& r1, int delta)
+{
+    return r1 + (-delta);
+}
+
+
+
+///////////////////////////////// Scalar ////////////////////////////////
+
+template<typename _Tp> inline
+Scalar_<_Tp>::Scalar_()
+{
+    this->val[0] = this->val[1] = this->val[2] = this->val[3] = 0;
+}
+
+template<typename _Tp> inline
+Scalar_<_Tp>::Scalar_(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
+{
+    this->val[0] = v0;
+    this->val[1] = v1;
+    this->val[2] = v2;
+    this->val[3] = v3;
+}
+
+template<typename _Tp> template<typename _Tp2, int cn> inline
+Scalar_<_Tp>::Scalar_(const Vec<_Tp2, cn>& v)
+{
+    int i;
+    for( i = 0; i < (cn < 4 ? cn : 4); i++ )
+        this->val[i] = cv::saturate_cast<_Tp>(v.val[i]);
+    for( ; i < 4; i++ )
+        this->val[i] = 0;
+}
+
+template<typename _Tp> inline
+Scalar_<_Tp>::Scalar_(_Tp v0)
+{
+    this->val[0] = v0;
+    this->val[1] = this->val[2] = this->val[3] = 0;
+}
+
+template<typename _Tp> inline
+Scalar_<_Tp> Scalar_<_Tp>::all(_Tp v0)
+{
+    return Scalar_<_Tp>(v0, v0, v0, v0);
+}
+
+
+template<typename _Tp> inline
+Scalar_<_Tp> Scalar_<_Tp>::mul(const Scalar_<_Tp>& t, double scale ) const
+{
+    return Scalar_<_Tp>(saturate_cast<_Tp>(this->val[0] * t.val[0] * scale),
+                        saturate_cast<_Tp>(this->val[1] * t.val[1] * scale),
+                        saturate_cast<_Tp>(this->val[2] * t.val[2] * scale),
+                        saturate_cast<_Tp>(this->val[3] * t.val[3] * scale));
+}
+
+template<typename _Tp> inline
+Scalar_<_Tp> Scalar_<_Tp>::conj() const
+{
+    return Scalar_<_Tp>(saturate_cast<_Tp>( this->val[0]),
+                        saturate_cast<_Tp>(-this->val[1]),
+                        saturate_cast<_Tp>(-this->val[2]),
+                        saturate_cast<_Tp>(-this->val[3]));
+}
+
+template<typename _Tp> inline
+bool Scalar_<_Tp>::isReal() const
+{
+    return this->val[1] == 0 && this->val[2] == 0 && this->val[3] == 0;
+}
+
+
+template<typename _Tp> template<typename T2> inline
+Scalar_<_Tp>::operator Scalar_<T2>() const
+{
+    return Scalar_<T2>(saturate_cast<T2>(this->val[0]),
+                       saturate_cast<T2>(this->val[1]),
+                       saturate_cast<T2>(this->val[2]),
+                       saturate_cast<T2>(this->val[3]));
+}
+
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator += (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    a.val[0] += b.val[0];
+    a.val[1] += b.val[1];
+    a.val[2] += b.val[2];
+    a.val[3] += b.val[3];
+    return a;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator -= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    a.val[0] -= b.val[0];
+    a.val[1] -= b.val[1];
+    a.val[2] -= b.val[2];
+    a.val[3] -= b.val[3];
+    return a;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator *= ( Scalar_<_Tp>& a, _Tp v )
+{
+    a.val[0] *= v;
+    a.val[1] *= v;
+    a.val[2] *= v;
+    a.val[3] *= v;
+    return a;
+}
+
+template<typename _Tp> static inline
+bool operator == ( const Scalar_<_Tp>& a, const Scalar_<_Tp>& b )
+{
+    return a.val[0] == b.val[0] && a.val[1] == b.val[1] &&
+           a.val[2] == b.val[2] && a.val[3] == b.val[3];
+}
+
+template<typename _Tp> static inline
+bool operator != ( const Scalar_<_Tp>& a, const Scalar_<_Tp>& b )
+{
+    return a.val[0] != b.val[0] || a.val[1] != b.val[1] ||
+           a.val[2] != b.val[2] || a.val[3] != b.val[3];
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator + (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    return Scalar_<_Tp>(a.val[0] + b.val[0],
+                        a.val[1] + b.val[1],
+                        a.val[2] + b.val[2],
+                        a.val[3] + b.val[3]);
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator - (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    return Scalar_<_Tp>(saturate_cast<_Tp>(a.val[0] - b.val[0]),
+                        saturate_cast<_Tp>(a.val[1] - b.val[1]),
+                        saturate_cast<_Tp>(a.val[2] - b.val[2]),
+                        saturate_cast<_Tp>(a.val[3] - b.val[3]));
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator * (const Scalar_<_Tp>& a, _Tp alpha)
+{
+    return Scalar_<_Tp>(a.val[0] * alpha,
+                        a.val[1] * alpha,
+                        a.val[2] * alpha,
+                        a.val[3] * alpha);
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator * (_Tp alpha, const Scalar_<_Tp>& a)
+{
+    return a*alpha;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator - (const Scalar_<_Tp>& a)
+{
+    return Scalar_<_Tp>(saturate_cast<_Tp>(-a.val[0]),
+                        saturate_cast<_Tp>(-a.val[1]),
+                        saturate_cast<_Tp>(-a.val[2]),
+                        saturate_cast<_Tp>(-a.val[3]));
+}
+
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator * (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    return Scalar_<_Tp>(saturate_cast<_Tp>(a[0]*b[0] - a[1]*b[1] - a[2]*b[2] - a[3]*b[3]),
+                        saturate_cast<_Tp>(a[0]*b[1] + a[1]*b[0] + a[2]*b[3] - a[3]*b[2]),
+                        saturate_cast<_Tp>(a[0]*b[2] - a[1]*b[3] + a[2]*b[0] + a[3]*b[1]),
+                        saturate_cast<_Tp>(a[0]*b[3] + a[1]*b[2] - a[2]*b[1] + a[3]*b[0]));
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator *= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    a = a * b;
+    return a;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator / (const Scalar_<_Tp>& a, _Tp alpha)
+{
+    return Scalar_<_Tp>(a.val[0] / alpha,
+                        a.val[1] / alpha,
+                        a.val[2] / alpha,
+                        a.val[3] / alpha);
+}
+
+template<typename _Tp> static inline
+Scalar_<float> operator / (const Scalar_<float>& a, float alpha)
+{
+    float s = 1 / alpha;
+    return Scalar_<float>(a.val[0] * s, a.val[1] * s, a.val[2] * s, a.val[3] * s);
+}
+
+template<typename _Tp> static inline
+Scalar_<double> operator / (const Scalar_<double>& a, double alpha)
+{
+    double s = 1 / alpha;
+    return Scalar_<double>(a.val[0] * s, a.val[1] * s, a.val[2] * s, a.val[3] * s);
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator /= (Scalar_<_Tp>& a, _Tp alpha)
+{
+    a = a / alpha;
+    return a;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator / (_Tp a, const Scalar_<_Tp>& b)
+{
+    _Tp s = a / (b[0]*b[0] + b[1]*b[1] + b[2]*b[2] + b[3]*b[3]);
+    return b.conj() * s;
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp> operator / (const Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    return a * ((_Tp)1 / b);
+}
+
+template<typename _Tp> static inline
+Scalar_<_Tp>& operator /= (Scalar_<_Tp>& a, const Scalar_<_Tp>& b)
+{
+    a = a / b;
+    return a;
+}
+
+template<typename _Tp> static inline
+Scalar operator * (const Matx<_Tp, 4, 4>& a, const Scalar& b)
+{
+    Matx<double, 4, 1> c((Matx<double, 4, 4>)a, b, Matx_MatMulOp());
+    return reinterpret_cast<const Scalar&>(c);
+}
+
+template<> inline
+Scalar operator * (const Matx<double, 4, 4>& a, const Scalar& b)
+{
+    Matx<double, 4, 1> c(a, b, Matx_MatMulOp());
+    return reinterpret_cast<const Scalar&>(c);
+}
+
+
+
+//////////////////////////////// KeyPoint ///////////////////////////////
+
+inline
+KeyPoint::KeyPoint()
+    : pt(0,0), size(0), angle(-1), response(0), octave(0), class_id(-1) {}
+
+inline
+KeyPoint::KeyPoint(Point2f _pt, float _size, float _angle, float _response, int _octave, int _class_id)
+    : pt(_pt), size(_size), angle(_angle), response(_response), octave(_octave), class_id(_class_id) {}
+
+inline
+KeyPoint::KeyPoint(float x, float y, float _size, float _angle, float _response, int _octave, int _class_id)
+    : pt(x, y), size(_size), angle(_angle), response(_response), octave(_octave), class_id(_class_id) {}
+
+
+
+///////////////////////////////// DMatch ////////////////////////////////
+
+inline
+DMatch::DMatch()
+    : queryIdx(-1), trainIdx(-1), imgIdx(-1), distance(FLT_MAX) {}
+
+inline
+DMatch::DMatch(int _queryIdx, int _trainIdx, float _distance)
+    : queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(-1), distance(_distance) {}
+
+inline
+DMatch::DMatch(int _queryIdx, int _trainIdx, int _imgIdx, float _distance)
+    : queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(_imgIdx), distance(_distance) {}
+
+inline
+bool DMatch::operator < (const DMatch &m) const
+{
+    return distance < m.distance;
+}
+
+
+
+////////////////////////////// TermCriteria /////////////////////////////
+
+inline
+TermCriteria::TermCriteria()
+    : type(0), maxCount(0), epsilon(0) {}
+
+inline
+TermCriteria::TermCriteria(int _type, int _maxCount, double _epsilon)
+    : type(_type), maxCount(_maxCount), epsilon(_epsilon) {}
+
+} // cv
+
+#endif //__OPENCV_CORE_TYPES_HPP__
diff --git a/modules/core/include/opencv2/core/types_c.h b/modules/core/include/opencv2/core/types_c.h
index 5838c0415d..e4c8cb5a6a 100644
--- a/modules/core/include/opencv2/core/types_c.h
+++ b/modules/core/include/opencv2/core/types_c.h
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -92,6 +93,14 @@
 #  define CVAPI(rettype) CV_EXTERN_C CV_EXPORTS rettype CV_CDECL
 #endif
 
+#ifndef CV_IMPL
+#  define CV_IMPL CV_EXTERN_C
+#endif
+
+#ifdef __cplusplus
+#  include "opencv2/core.hpp"
+#endif
+
 /* CvArr* is used to pass arbitrary
  * array-like data structures
  * into functions where the particular
@@ -177,14 +186,6 @@ enum {
 
 #define CV_SWAP(a,b,t) ((t) = (a), (a) = (b), (b) = (t))
 
-#ifndef MIN
-#  define MIN(a,b)  ((a) > (b) ? (b) : (a))
-#endif
-
-#ifndef MAX
-#  define MAX(a,b)  ((a) < (b) ? (b) : (a))
-#endif
-
 /* min & max without jumps */
 #define  CV_IMIN(a, b)  ((a) ^ (((a)^(b)) & (((a) < (b)) - 1)))
 
@@ -271,7 +272,11 @@ CV_INLINE double cvRandReal( CvRNG* rng )
 #define IPL_BORDER_REFLECT    2
 #define IPL_BORDER_WRAP       3
 
-typedef struct _IplImage
+typedef struct
+#ifdef __cplusplus
+  CV_EXPORTS
+#endif
+_IplImage
 {
     int  nSize;             /* sizeof(IplImage) */
     int  ID;                /* version (=0)*/
@@ -303,6 +308,11 @@ typedef struct _IplImage
     char *imageDataOrigin;  /* Pointer to very origin of image data
                                (not necessarily aligned) -
                                needed for correct deallocation */
+
+#ifdef __cplusplus
+    _IplImage() {}
+    _IplImage(const cv::Mat& m);
+#endif
 }
 IplImage;
 
@@ -413,6 +423,13 @@ typedef struct CvMat
     int cols;
 #endif
 
+
+#ifdef __cplusplus
+    CvMat() {}
+    CvMat(const CvMat& m) { memcpy(this, &m, sizeof(CvMat));}
+    CvMat(const cv::Mat& m);
+#endif
+
 }
 CvMat;
 
@@ -474,6 +491,16 @@ CV_INLINE CvMat cvMat( int rows, int cols, int type, void* data CV_DEFAULT(NULL)
     return m;
 }
 
+#ifdef __cplusplus
+inline CvMat::CvMat(const cv::Mat& m)
+{
+    CV_DbgAssert(m.dims <= 2);
+    *this = cvMat(m.rows, m.dims == 1 ? 1 : m.cols, m.type(), m.data);
+    step = (int)m.step[0];
+    type = (type & ~cv::Mat::CONTINUOUS_FLAG) | (m.flags & cv::Mat::CONTINUOUS_FLAG);
+}
+#endif
+
 
 #define CV_MAT_ELEM_PTR_FAST( mat, row, col, pix_size )  \
     (assert( (unsigned)(row) < (unsigned)(mat).rows &&   \
@@ -540,7 +567,11 @@ CV_INLINE int cvIplDepth( int type )
 #define CV_MAX_DIM            32
 #define CV_MAX_DIM_HEAP       1024
 
-typedef struct CvMatND
+typedef struct
+#ifdef __cplusplus
+  CV_EXPORTS
+#endif
+CvMatND
 {
     int type;
     int dims;
@@ -563,6 +594,11 @@ typedef struct CvMatND
         int step;
     }
     dim[CV_MAX_DIM];
+
+#ifdef __cplusplus
+    CvMatND() {}
+    CvMatND(const cv::Mat& m);
+#endif
 }
 CvMatND;
 
@@ -582,7 +618,11 @@ CvMatND;
 
 struct CvSet;
 
-typedef struct CvSparseMat
+typedef struct
+#ifdef __cplusplus
+  CV_EXPORTS
+#endif
+CvSparseMat
 {
     int type;
     int dims;
@@ -595,9 +635,17 @@ typedef struct CvSparseMat
     int valoffset;
     int idxoffset;
     int size[CV_MAX_DIM];
+
+#ifdef __cplusplus
+    void copyToSparseMat(cv::SparseMat& m) const;
+#endif
 }
 CvSparseMat;
 
+#ifdef __cplusplus
+    CV_EXPORTS CvSparseMat* cvCreateSparseMat(const cv::SparseMat& m);
+#endif
+
 #define CV_IS_SPARSE_MAT_HDR(mat) \
     ((mat) != NULL && \
     (((const CvSparseMat*)(mat))->type & CV_MAGIC_MASK) == CV_SPARSE_MAT_MAGIC_VAL)
@@ -681,6 +729,14 @@ typedef struct CvRect
     int y;
     int width;
     int height;
+
+#ifdef __cplusplus
+    CvRect(int _x = 0, int _y = 0, int w = 0, int h = 0): x(_x), y(_y), width(w), height(h) {}
+    template<typename _Tp>
+    CvRect(const cv::Rect_<_Tp>& r): x(cv::saturate_cast<int>(r.x)), y(cv::saturate_cast<int>(r.y)), width(cv::saturate_cast<int>(r.width)), height(cv::saturate_cast<int>(r.height)) {}
+    template<typename _Tp>
+    operator cv::Rect_<_Tp>() const { return cv::Rect_<_Tp>((_Tp)x, (_Tp)y, (_Tp)width, (_Tp)height); }
+#endif
 }
 CvRect;
 
@@ -728,6 +784,13 @@ typedef struct CvTermCriteria
                      CV_TERMCRIT_EPS */
     int    max_iter;
     double epsilon;
+
+#ifdef __cplusplus
+    CvTermCriteria(int _type = 0, int _iter = 0, double _eps = 0) : type(_type), max_iter(_iter), epsilon(_eps)  {}
+    CvTermCriteria(const cv::TermCriteria& t) : type(t.type), max_iter(t.maxCount), epsilon(t.epsilon)  {}
+    operator cv::TermCriteria() const { return cv::TermCriteria(type, max_iter, epsilon); }
+#endif
+
 }
 CvTermCriteria;
 
@@ -749,6 +812,14 @@ typedef struct CvPoint
 {
     int x;
     int y;
+
+#ifdef __cplusplus
+    CvPoint(int _x = 0, int _y = 0): x(_x), y(_y) {}
+    template<typename _Tp>
+    CvPoint(const cv::Point_<_Tp>& pt): x((int)pt.x), y((int)pt.y) {}
+    template<typename _Tp>
+    operator cv::Point_<_Tp>() const { return cv::Point_<_Tp>(cv::saturate_cast<_Tp>(x), cv::saturate_cast<_Tp>(y)); }
+#endif
 }
 CvPoint;
 
@@ -768,6 +839,14 @@ typedef struct CvPoint2D32f
 {
     float x;
     float y;
+
+#ifdef __cplusplus
+    CvPoint2D32f(float _x = 0, float _y = 0): x(_x), y(_y) {}
+    template<typename _Tp>
+    CvPoint2D32f(const cv::Point_<_Tp>& pt): x((float)pt.x), y((float)pt.y) {}
+    template<typename _Tp>
+    operator cv::Point_<_Tp>() const { return cv::Point_<_Tp>(cv::saturate_cast<_Tp>(x), cv::saturate_cast<_Tp>(y)); }
+#endif
 }
 CvPoint2D32f;
 
@@ -804,6 +883,14 @@ typedef struct CvPoint3D32f
     float x;
     float y;
     float z;
+
+#ifdef __cplusplus
+    CvPoint3D32f(float _x = 0, float _y = 0, float _z = 0): x(_x), y(_y), z(_z) {}
+    template<typename _Tp>
+    CvPoint3D32f(const cv::Point3_<_Tp>& pt): x((float)pt.x), y((float)pt.y), z((float)pt.z) {}
+    template<typename _Tp>
+    operator cv::Point3_<_Tp>() const { return cv::Point3_<_Tp>(cv::saturate_cast<_Tp>(x), cv::saturate_cast<_Tp>(y), cv::saturate_cast<_Tp>(z)); }
+#endif
 }
 CvPoint3D32f;
 
@@ -866,6 +953,14 @@ typedef struct CvSize
 {
     int width;
     int height;
+
+#ifdef __cplusplus
+    CvSize(int w = 0, int h = 0): width(w), height(h) {}
+    template<typename _Tp>
+    CvSize(const cv::Size_<_Tp>& sz): width(cv::saturate_cast<int>(sz.width)), height(cv::saturate_cast<int>(sz.height)) {}
+    template<typename _Tp>
+    operator cv::Size_<_Tp>() const { return cv::Size_<_Tp>(cv::saturate_cast<_Tp>(width), cv::saturate_cast<_Tp>(height)); }
+#endif
 }
 CvSize;
 
@@ -883,6 +978,14 @@ typedef struct CvSize2D32f
 {
     float width;
     float height;
+
+#ifdef __cplusplus
+    CvSize2D32f(float w = 0, float h = 0): width(w), height(h) {}
+    template<typename _Tp>
+    CvSize2D32f(const cv::Size_<_Tp>& sz): width(cv::saturate_cast<float>(sz.width)), height(cv::saturate_cast<float>(sz.height)) {}
+    template<typename _Tp>
+    operator cv::Size_<_Tp>() const { return cv::Size_<_Tp>(cv::saturate_cast<_Tp>(width), cv::saturate_cast<_Tp>(height)); }
+#endif
 }
 CvSize2D32f;
 
@@ -903,6 +1006,12 @@ typedef struct CvBox2D
     CvSize2D32f  size;    /* Box width and length.                       */
     float angle;          /* Angle between the horizontal axis           */
                           /* and the first side (i.e. length) in degrees */
+
+#ifdef __cplusplus
+    CvBox2D(CvPoint2D32f c = CvPoint2D32f(), CvSize2D32f s = CvSize2D32f(), float a = 0) : center(c), size(s), angle(a) {}
+    CvBox2D(const cv::RotatedRect& rr) : center(rr.center), size(rr.size), angle(rr.angle) {}
+    operator cv::RotatedRect() const { return cv::RotatedRect(center, size, angle); }
+#endif
 }
 CvBox2D;
 
@@ -925,10 +1034,18 @@ CvLineIterator;
 
 
 /************************************* CvSlice ******************************************/
+#define CV_WHOLE_SEQ_END_INDEX 0x3fffffff
+#define CV_WHOLE_SEQ  cvSlice(0, CV_WHOLE_SEQ_END_INDEX)
 
 typedef struct CvSlice
 {
     int  start_index, end_index;
+
+#if defined(__cplusplus) && !defined(__CUDACC__)
+    CvSlice(int start = 0, int end = 0) : start_index(start), end_index(end) {}
+    CvSlice(const cv::Range& r) { *this = (r.start != INT_MIN && r.end != INT_MAX) ? CvSlice(r.start, r.end) : CvSlice(0, CV_WHOLE_SEQ_END_INDEX); }
+    operator cv::Range() const { return (start_index == 0 && end_index == CV_WHOLE_SEQ_END_INDEX ) ? cv::Range::all() : cv::Range(start_index, end_index); }
+#endif
 }
 CvSlice;
 
@@ -941,8 +1058,6 @@ CV_INLINE  CvSlice  cvSlice( int start, int end )
     return slice;
 }
 
-#define CV_WHOLE_SEQ_END_INDEX 0x3fffffff
-#define CV_WHOLE_SEQ  cvSlice(0, CV_WHOLE_SEQ_END_INDEX)
 
 
 /************************************* CvScalar *****************************************/
@@ -950,6 +1065,22 @@ CV_INLINE  CvSlice  cvSlice( int start, int end )
 typedef struct CvScalar
 {
     double val[4];
+
+#ifdef __cplusplus
+    CvScalar() {}
+    CvScalar(double d0, double d1 = 0, double d2 = 0, double d3 = 0) { val[0] = d0; val[1] = d1; val[2] = d2; val[3] = d3; }
+    template<typename _Tp>
+    CvScalar(const cv::Scalar_<_Tp>& s) { val[0] = s.val[0]; val[1] = s.val[1]; val[2] = s.val[2]; val[3] = s.val[3]; }
+    template<typename _Tp>
+    operator cv::Scalar_<_Tp>() const { return cv::Scalar_<_Tp>(cv::saturate_cast<_Tp>(val[0]), cv::saturate_cast<_Tp>(val[1]), cv::saturate_cast<_Tp>(val[2]), cv::saturate_cast<_Tp>(val[3])); }
+    template<typename _Tp, int cn>
+    CvScalar(const cv::Vec<_Tp, cn>& v)
+    {
+        int i;
+        for( i = 0; i < (cn < 4 ? cn : 4); i++ ) val[i] = v.val[i];
+        for( ; i < 4; i++ ) val[i] = 0;
+    }
+#endif
 }
 CvScalar;
 
@@ -1343,7 +1474,6 @@ CvSeqWriter;
     int          delta_index;/* = seq->first->start_index   */      \
     schar*       prev_elem;  /* pointer to previous element */
 
-
 typedef struct CvSeqReader
 {
     CV_SEQ_READER_FIELDS()
diff --git a/modules/core/include/opencv2/core/utility.hpp b/modules/core/include/opencv2/core/utility.hpp
new file mode 100644
index 0000000000..c96a0aaf93
--- /dev/null
+++ b/modules/core/include/opencv2/core/utility.hpp
@@ -0,0 +1,466 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_CORE_UTILITY_H__
+#define __OPENCV_CORE_UTILITY_H__
+
+#ifndef __cplusplus
+#  error utility.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core.hpp"
+
+namespace cv
+{
+
+/*!
+ Automatically Allocated Buffer Class
+
+ The class is used for temporary buffers in functions and methods.
+ If a temporary buffer is usually small (a few K's of memory),
+ but its size depends on the parameters, it makes sense to create a small
+ fixed-size array on stack and use it if it's large enough. If the required buffer size
+ is larger than the fixed size, another buffer of sufficient size is allocated dynamically
+ and released after the processing. Therefore, in typical cases, when the buffer size is small,
+ there is no overhead associated with malloc()/free().
+ At the same time, there is no limit on the size of processed data.
+
+ This is what AutoBuffer does. The template takes 2 parameters - type of the buffer elements and
+ the number of stack-allocated elements. Here is how the class is used:
+
+ \code
+ void my_func(const cv::Mat& m)
+ {
+    cv::AutoBuffer<float> buf; // create automatic buffer containing 1000 floats
+
+    buf.allocate(m.rows); // if m.rows <= 1000, the pre-allocated buffer is used,
+                          // otherwise the buffer of "m.rows" floats will be allocated
+                          // dynamically and deallocated in cv::AutoBuffer destructor
+    ...
+ }
+ \endcode
+*/
+template<typename _Tp, size_t fixed_size = 1024/sizeof(_Tp)+8> class AutoBuffer
+{
+public:
+    typedef _Tp value_type;
+
+    //! the default contructor
+    AutoBuffer();
+    //! constructor taking the real buffer size
+    AutoBuffer(size_t _size);
+
+    //! the copy constructor
+    AutoBuffer(const AutoBuffer<_Tp, fixed_size>& buf);
+    //! the assignment operator
+    AutoBuffer<_Tp, fixed_size>& operator = (const AutoBuffer<_Tp, fixed_size>& buf);
+
+    //! destructor. calls deallocate()
+    ~AutoBuffer();
+
+    //! allocates the new buffer of size _size. if the _size is small enough, stack-allocated buffer is used
+    void allocate(size_t _size);
+    //! deallocates the buffer if it was dynamically allocated
+    void deallocate();
+    //! resizes the buffer and preserves the content
+    void resize(size_t _size);
+    //! returns the current buffer size
+    size_t size() const;
+    //! returns pointer to the real buffer, stack-allocated or head-allocated
+    operator _Tp* ();
+    //! returns read-only pointer to the real buffer, stack-allocated or head-allocated
+    operator const _Tp* () const;
+
+protected:
+    //! pointer to the real buffer, can point to buf if the buffer is small enough
+    _Tp* ptr;
+    //! size of the real buffer
+    size_t sz;
+    //! pre-allocated buffer. At least 1 element to confirm C++ standard reqirements
+    _Tp buf[(fixed_size > 0) ? fixed_size : 1];
+};
+
+//! Sets/resets the break-on-error mode.
+
+/*!
+  When the break-on-error mode is set, the default error handler
+  issues a hardware exception, which can make debugging more convenient.
+
+  \return the previous state
+ */
+CV_EXPORTS bool setBreakOnError(bool flag);
+
+extern "C" typedef int (*ErrorCallback)( int status, const char* func_name,
+                                       const char* err_msg, const char* file_name,
+                                       int line, void* userdata );
+
+//! Sets the new error handler and the optional user data.
+
+/*!
+  The function sets the new error handler, called from cv::error().
+
+  \param errCallback the new error handler. If NULL, the default error handler is used.
+  \param userdata the optional user data pointer, passed to the callback.
+  \param prevUserdata the optional output parameter where the previous user data pointer is stored
+
+  \return the previous error handler
+*/
+CV_EXPORTS ErrorCallback redirectError( ErrorCallback errCallback, void* userdata=0, void** prevUserdata=0);
+
+CV_EXPORTS String format( const char* fmt, ... );
+CV_EXPORTS String tempfile( const char* suffix = 0);
+CV_EXPORTS void glob(String pattern, std::vector<String>& result, bool recursive = false);
+CV_EXPORTS void setNumThreads(int nthreads);
+CV_EXPORTS int getNumThreads();
+CV_EXPORTS int getThreadNum();
+
+CV_EXPORTS_W const String& getBuildInformation();
+
+//! Returns the number of ticks.
+
+/*!
+  The function returns the number of ticks since the certain event (e.g. when the machine was turned on).
+  It can be used to initialize cv::RNG or to measure a function execution time by reading the tick count
+  before and after the function call. The granularity of ticks depends on the hardware and OS used. Use
+  cv::getTickFrequency() to convert ticks to seconds.
+*/
+CV_EXPORTS_W int64 getTickCount();
+
+/*!
+  Returns the number of ticks per seconds.
+
+  The function returns the number of ticks (as returned by cv::getTickCount()) per second.
+  The following code computes the execution time in milliseconds:
+
+  \code
+  double exec_time = (double)getTickCount();
+  // do something ...
+  exec_time = ((double)getTickCount() - exec_time)*1000./getTickFrequency();
+  \endcode
+*/
+CV_EXPORTS_W double getTickFrequency();
+
+/*!
+  Returns the number of CPU ticks.
+
+  On platforms where the feature is available, the function returns the number of CPU ticks
+  since the certain event (normally, the system power-on moment). Using this function
+  one can accurately measure the execution time of very small code fragments,
+  for which cv::getTickCount() granularity is not enough.
+*/
+CV_EXPORTS_W int64 getCPUTickCount();
+
+/*!
+  Returns SSE etc. support status
+
+  The function returns true if certain hardware features are available.
+  Currently, the following features are recognized:
+  - CV_CPU_MMX - MMX
+  - CV_CPU_SSE - SSE
+  - CV_CPU_SSE2 - SSE 2
+  - CV_CPU_SSE3 - SSE 3
+  - CV_CPU_SSSE3 - SSSE 3
+  - CV_CPU_SSE4_1 - SSE 4.1
+  - CV_CPU_SSE4_2 - SSE 4.2
+  - CV_CPU_POPCNT - POPCOUNT
+  - CV_CPU_AVX - AVX
+
+  \note {Note that the function output is not static. Once you called cv::useOptimized(false),
+  most of the hardware acceleration is disabled and thus the function will returns false,
+  until you call cv::useOptimized(true)}
+*/
+CV_EXPORTS_W bool checkHardwareSupport(int feature);
+
+//! returns the number of CPUs (including hyper-threading)
+CV_EXPORTS_W int getNumberOfCPUs();
+
+
+/*!
+  Aligns pointer by the certain number of bytes
+
+  This small inline function aligns the pointer by the certian number of bytes by shifting
+  it forward by 0 or a positive offset.
+*/
+template<typename _Tp> static inline _Tp* alignPtr(_Tp* ptr, int n=(int)sizeof(_Tp))
+{
+    return (_Tp*)(((size_t)ptr + n-1) & -n);
+}
+
+/*!
+  Aligns buffer size by the certain number of bytes
+
+  This small inline function aligns a buffer size by the certian number of bytes by enlarging it.
+*/
+static inline size_t alignSize(size_t sz, int n)
+{
+    return (sz + n-1) & -n;
+}
+
+/*!
+  Turns on/off available optimization
+
+  The function turns on or off the optimized code in OpenCV. Some optimization can not be enabled
+  or disabled, but, for example, most of SSE code in OpenCV can be temporarily turned on or off this way.
+
+  \note{Since optimization may imply using special data structures, it may be unsafe
+  to call this function anywhere in the code. Instead, call it somewhere at the top level.}
+*/
+CV_EXPORTS_W void setUseOptimized(bool onoff);
+
+/*!
+  Returns the current optimization status
+
+  The function returns the current optimization status, which is controlled by cv::setUseOptimized().
+*/
+CV_EXPORTS_W bool useOptimized();
+
+static inline size_t getElemSize(int type) { return CV_ELEM_SIZE(type); }
+
+/////////////////////////////// Parallel Primitives //////////////////////////////////
+
+// a base body class
+class CV_EXPORTS ParallelLoopBody
+{
+public:
+    virtual ~ParallelLoopBody();
+    virtual void operator() (const Range& range) const = 0;
+};
+
+CV_EXPORTS void parallel_for_(const Range& range, const ParallelLoopBody& body, double nstripes=-1.);
+
+/////////////////////////// Synchronization Primitives ///////////////////////////////
+
+class CV_EXPORTS Mutex
+{
+public:
+    Mutex();
+    ~Mutex();
+    Mutex(const Mutex& m);
+    Mutex& operator = (const Mutex& m);
+
+    void lock();
+    bool trylock();
+    void unlock();
+
+    struct Impl;
+protected:
+    Impl* impl;
+};
+
+class CV_EXPORTS AutoLock
+{
+public:
+    AutoLock(Mutex& m) : mutex(&m) { mutex->lock(); }
+    ~AutoLock() { mutex->unlock(); }
+protected:
+    Mutex* mutex;
+private:
+    AutoLock(const AutoLock&);
+    AutoLock& operator = (const AutoLock&);
+};
+
+// The CommandLineParser class is designed for command line arguments parsing
+
+class CV_EXPORTS CommandLineParser
+{
+    public:
+    CommandLineParser(int argc, const char* const argv[], const String& keys);
+    CommandLineParser(const CommandLineParser& parser);
+    CommandLineParser& operator = (const CommandLineParser& parser);
+
+    String getPathToApplication() const;
+
+    template <typename T>
+    T get(const String& name, bool space_delete = true) const
+    {
+        T val = T();
+        getByName(name, space_delete, ParamType<T>::type, (void*)&val);
+        return val;
+    }
+
+    template <typename T>
+    T get(int index, bool space_delete = true) const
+    {
+        T val = T();
+        getByIndex(index, space_delete, ParamType<T>::type, (void*)&val);
+        return val;
+    }
+
+    bool has(const String& name) const;
+
+    bool check() const;
+
+    void about(const String& message);
+
+    void printMessage() const;
+    void printErrors() const;
+
+protected:
+    void getByName(const String& name, bool space_delete, int type, void* dst) const;
+    void getByIndex(int index, bool space_delete, int type, void* dst) const;
+
+    struct Impl;
+    Impl* impl;
+};
+
+/////////////////////////////// AutoBuffer implementation ////////////////////////////////////////
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::AutoBuffer()
+{
+    ptr = buf;
+    sz = fixed_size;
+}
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::AutoBuffer(size_t _size)
+{
+    ptr = buf;
+    sz = fixed_size;
+    allocate(_size);
+}
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::AutoBuffer(const AutoBuffer<_Tp, fixed_size>& abuf )
+{
+    ptr = buf;
+    sz = fixed_size;
+    allocate(abuf.size());
+    for( size_t i = 0; i < sz; i++ )
+        ptr[i] = abuf.ptr[i];
+}
+
+template<typename _Tp, size_t fixed_size> inline AutoBuffer<_Tp, fixed_size>&
+AutoBuffer<_Tp, fixed_size>::operator = (const AutoBuffer<_Tp, fixed_size>& abuf)
+{
+    if( this != &abuf )
+    {
+        deallocate();
+        allocate(abuf.size());
+        for( size_t i = 0; i < sz; i++ )
+            ptr[i] = abuf.ptr[i];
+    }
+    return *this;
+}
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::~AutoBuffer()
+{ deallocate(); }
+
+template<typename _Tp, size_t fixed_size> inline void
+AutoBuffer<_Tp, fixed_size>::allocate(size_t _size)
+{
+    if(_size <= sz)
+    {
+        sz = _size;
+        return;
+    }
+    deallocate();
+    if(_size > fixed_size)
+    {
+        ptr = new _Tp[_size];
+        sz = _size;
+    }
+}
+
+template<typename _Tp, size_t fixed_size> inline void
+AutoBuffer<_Tp, fixed_size>::deallocate()
+{
+    if( ptr != buf )
+    {
+        delete[] ptr;
+        ptr = buf;
+        sz = fixed_size;
+    }
+}
+
+template<typename _Tp, size_t fixed_size> inline void
+AutoBuffer<_Tp, fixed_size>::resize(size_t _size)
+{
+    if(_size <= sz)
+    {
+        sz = _size;
+        return;
+    }
+    size_t i, prevsize = sz, minsize = MIN(prevsize, _size);
+    _Tp* prevptr = ptr;
+
+    ptr = _size > fixed_size ? new _Tp[_size] : buf;
+    sz = _size;
+
+    if( ptr != prevptr )
+        for( i = 0; i < minsize; i++ )
+            ptr[i] = prevptr[i];
+    for( i = prevsize; i < _size; i++ )
+        ptr[i] = _Tp();
+
+    if( prevptr != buf )
+        delete[] prevptr;
+}
+
+template<typename _Tp, size_t fixed_size> inline size_t
+AutoBuffer<_Tp, fixed_size>::size() const
+{ return sz; }
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::operator _Tp* ()
+{ return ptr; }
+
+template<typename _Tp, size_t fixed_size> inline
+AutoBuffer<_Tp, fixed_size>::operator const _Tp* () const
+{ return ptr; }
+
+#ifndef OPENCV_NOSTL
+template<> inline std::string CommandLineParser::get<std::string>(int index, bool space_delete) const
+{
+    return get<String>(index, space_delete);
+}
+template<> inline std::string CommandLineParser::get<std::string>(const String& name, bool space_delete) const
+{
+    return get<String>(name, space_delete);
+}
+#endif // OPENCV_NOSTL
+
+} //namespace cv
+
+#endif //__OPENCV_CORE_UTILITY_H__
diff --git a/modules/core/include/opencv2/core/version.hpp b/modules/core/include/opencv2/core/version.hpp
index 916d173bac..203503ecad 100644
--- a/modules/core/include/opencv2/core/version.hpp
+++ b/modules/core/include/opencv2/core/version.hpp
@@ -11,6 +11,7 @@
 //                For Open Source Computer Vision Library
 //
 // Copyright( C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2011-2013, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -47,18 +48,22 @@
 #ifndef __OPENCV_VERSION_HPP__
 #define __OPENCV_VERSION_HPP__
 
-#define CV_VERSION_EPOCH    2
-#define CV_VERSION_MAJOR    4
-#define CV_VERSION_MINOR    9
+#define CV_VERSION_EPOCH    3
+#define CV_VERSION_MAJOR    0
+#define CV_VERSION_MINOR    0
 #define CV_VERSION_REVISION 0
+#define CV_VERSION_STATUS   "-dev"
 
 #define CVAUX_STR_EXP(__A)  #__A
 #define CVAUX_STR(__A)      CVAUX_STR_EXP(__A)
 
+#define CVAUX_STRW_EXP(__A)  L#__A
+#define CVAUX_STRW(__A)      CVAUX_STRW_EXP(__A)
+
 #if CV_VERSION_REVISION
-#  define CV_VERSION        CVAUX_STR(CV_VERSION_EPOCH) "." CVAUX_STR(CV_VERSION_MAJOR) "." CVAUX_STR(CV_VERSION_MINOR) "." CVAUX_STR(CV_VERSION_REVISION)
+#  define CV_VERSION        CVAUX_STR(CV_VERSION_EPOCH) "." CVAUX_STR(CV_VERSION_MAJOR) "." CVAUX_STR(CV_VERSION_MINOR) "." CVAUX_STR(CV_VERSION_REVISION) CV_VERSION_STATUS
 #else
-#  define CV_VERSION        CVAUX_STR(CV_VERSION_EPOCH) "." CVAUX_STR(CV_VERSION_MAJOR) "." CVAUX_STR(CV_VERSION_MINOR)
+#  define CV_VERSION        CVAUX_STR(CV_VERSION_EPOCH) "." CVAUX_STR(CV_VERSION_MAJOR) "." CVAUX_STR(CV_VERSION_MINOR) CV_VERSION_STATUS
 #endif
 
 /* old  style version constants*/
diff --git a/modules/core/perf/perf_abs.cpp b/modules/core/perf/perf_abs.cpp
index 691c6f7c4b..63cb06b638 100644
--- a/modules/core/perf/perf_abs.cpp
+++ b/modules/core/perf/perf_abs.cpp
@@ -24,4 +24,3 @@ PERF_TEST_P(Size_MatType, abs, TYPICAL_MATS_ABS)
 
     SANITY_CHECK(c);
 }
-
diff --git a/modules/core/perf/perf_bitwise.cpp b/modules/core/perf/perf_bitwise.cpp
index 64a8dd8bd9..1308b7bf3e 100644
--- a/modules/core/perf/perf_bitwise.cpp
+++ b/modules/core/perf/perf_bitwise.cpp
@@ -73,4 +73,3 @@ PERF_TEST_P(Size_MatType, bitwise_xor, TYPICAL_MATS_BITW_ARITHM)
 
     SANITY_CHECK(c);
 }
-
diff --git a/modules/core/perf/perf_compare.cpp b/modules/core/perf/perf_compare.cpp
index 5fb755d956..86d07d9c0a 100644
--- a/modules/core/perf/perf_compare.cpp
+++ b/modules/core/perf/perf_compare.cpp
@@ -15,7 +15,7 @@ PERF_TEST_P( Size_MatType_CmpType, compare,
              testing::Combine(
                  testing::Values(::perf::szVGA, ::perf::sz1080p),
                  testing::Values(CV_8UC1, CV_8UC4, CV_8SC1, CV_16UC1, CV_16SC1, CV_32SC1, CV_32FC1),
-                 testing::ValuesIn(CmpType::all())
+                 CmpType::all()
                  )
              )
 {
@@ -38,7 +38,7 @@ PERF_TEST_P( Size_MatType_CmpType, compareScalar,
              testing::Combine(
                  testing::Values(TYPICAL_MAT_SIZES),
                  testing::Values(TYPICAL_MAT_TYPES),
-                 testing::ValuesIn(CmpType::all())
+                 CmpType::all()
                  )
              )
 {
diff --git a/modules/core/perf/perf_main.cpp b/modules/core/perf/perf_main.cpp
index 79c28a645c..7c899c2446 100644
--- a/modules/core/perf/perf_main.cpp
+++ b/modules/core/perf/perf_main.cpp
@@ -1,3 +1,8 @@
 #include "perf_precomp.hpp"
+#ifdef _MSC_VER
+# if _MSC_VER >= 1700
+#  pragma warning(disable:4447) // Disable warning 'main' signature found without threading model
+# endif
+#endif
 
 CV_PERF_TEST_MAIN(core)
diff --git a/modules/core/perf/perf_merge.cpp b/modules/core/perf/perf_merge.cpp
index d82941a92b..e7e8d2fe3f 100644
--- a/modules/core/perf/perf_merge.cpp
+++ b/modules/core/perf/perf_merge.cpp
@@ -34,4 +34,4 @@ PERF_TEST_P( Size_SrcDepth_DstChannels, merge,
     TEST_CYCLE_MULTIRUN(runs) merge( (vector<Mat> &)mv, dst );
 
     SANITY_CHECK(dst, 1e-12);
-}
\ No newline at end of file
+}
diff --git a/modules/core/perf/perf_precomp.cpp b/modules/core/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/core/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/core/perf/perf_reduce.cpp b/modules/core/perf/perf_reduce.cpp
index 81cdb5db57..7b74b0e7e3 100644
--- a/modules/core/perf/perf_reduce.cpp
+++ b/modules/core/perf/perf_reduce.cpp
@@ -16,7 +16,7 @@ PERF_TEST_P(Size_MatType_ROp, reduceR,
             testing::Combine(
                 testing::Values(TYPICAL_MAT_SIZES),
                 testing::Values(TYPICAL_MAT_TYPES),
-                testing::ValuesIn(ROp::all())
+                ROp::all()
                 )
             )
 {
@@ -34,7 +34,8 @@ PERF_TEST_P(Size_MatType_ROp, reduceR,
     declare.in(src, WARMUP_RNG).out(vec);
     declare.time(100);
 
-    TEST_CYCLE() reduce(src, vec, 0, reduceOp, ddepth);
+    int runs = 15;
+    TEST_CYCLE_MULTIRUN(runs) reduce(src, vec, 0, reduceOp, ddepth);
 
     SANITY_CHECK(vec, 1);
 }
@@ -43,7 +44,7 @@ PERF_TEST_P(Size_MatType_ROp, reduceC,
             testing::Combine(
                 testing::Values(TYPICAL_MAT_SIZES),
                 testing::Values(TYPICAL_MAT_TYPES),
-                testing::ValuesIn(ROp::all())
+                ROp::all()
                 )
             )
 {
@@ -65,4 +66,3 @@ PERF_TEST_P(Size_MatType_ROp, reduceC,
 
     SANITY_CHECK(vec, 1);
 }
-
diff --git a/modules/core/perf/perf_stat.cpp b/modules/core/perf/perf_stat.cpp
index b7fc43d120..6b5f0ff524 100644
--- a/modules/core/perf/perf_stat.cpp
+++ b/modules/core/perf/perf_stat.cpp
@@ -33,7 +33,7 @@ PERF_TEST_P(Size_MatType, mean, TYPICAL_MATS)
 
     TEST_CYCLE() s = mean(src);
 
-    SANITY_CHECK(s, 1e-6);
+    SANITY_CHECK(s, 1e-5);
 }
 
 PERF_TEST_P(Size_MatType, mean_mask, TYPICAL_MATS)
@@ -49,7 +49,7 @@ PERF_TEST_P(Size_MatType, mean_mask, TYPICAL_MATS)
 
     TEST_CYCLE() s = mean(src, mask);
 
-    SANITY_CHECK(s, 1e-6);
+    SANITY_CHECK(s, 5e-5);
 }
 
 PERF_TEST_P(Size_MatType, meanStdDev, TYPICAL_MATS)
@@ -83,8 +83,8 @@ PERF_TEST_P(Size_MatType, meanStdDev_mask, TYPICAL_MATS)
 
     TEST_CYCLE() meanStdDev(src, mean, dev, mask);
 
-    SANITY_CHECK(mean, 1e-6);
-    SANITY_CHECK(dev, 1e-6);
+    SANITY_CHECK(mean, 1e-5);
+    SANITY_CHECK(dev, 1e-5);
 }
 
 PERF_TEST_P(Size_MatType, countNonZero, testing::Combine( testing::Values( TYPICAL_MAT_SIZES ), testing::Values( CV_8UC1, CV_8SC1, CV_16UC1, CV_16SC1, CV_32SC1, CV_32FC1, CV_64FC1 ) ))
diff --git a/modules/core/src/algorithm.cpp b/modules/core/src/algorithm.cpp
index 67c71443c5..ff67a5df1e 100644
--- a/modules/core/src/algorithm.cpp
+++ b/modules/core/src/algorithm.cpp
@@ -96,7 +96,7 @@ template<typename _KeyTp, typename _ValueTp> struct sorted_vector
 };
 
 
-template<typename _ValueTp> inline const _ValueTp* findstr(const sorted_vector<std::string, _ValueTp>& vec,
+template<typename _ValueTp> inline const _ValueTp* findstr(const sorted_vector<String, _ValueTp>& vec,
                                                            const char* key)
 {
     if( !key )
@@ -130,7 +130,7 @@ Param::Param()
 
 Param::Param(int _type, bool _readonly, int _offset,
              Algorithm::Getter _getter, Algorithm::Setter _setter,
-             const std::string& _help)
+             const String& _help)
 {
     type = _type;
     readonly = _readonly;
@@ -142,28 +142,28 @@ Param::Param(int _type, bool _readonly, int _offset,
 
 struct CV_EXPORTS AlgorithmInfoData
 {
-    sorted_vector<std::string, Param> params;
-    std::string _name;
+    sorted_vector<String, Param> params;
+    String _name;
 };
 
 
-static sorted_vector<std::string, Algorithm::Constructor>& alglist()
+static sorted_vector<String, Algorithm::Constructor>& alglist()
 {
-    static sorted_vector<std::string, Algorithm::Constructor> alglist_var;
+    static sorted_vector<String, Algorithm::Constructor> alglist_var;
     return alglist_var;
 }
 
-void Algorithm::getList(std::vector<std::string>& algorithms)
+void Algorithm::getList(std::vector<String>& algorithms)
 {
     alglist().get_keys(algorithms);
 }
 
-Ptr<Algorithm> Algorithm::_create(const std::string& name)
+Ptr<Algorithm> Algorithm::_create(const String& name)
 {
     Algorithm::Constructor c = 0;
     if( !alglist().find(name, c) )
         return Ptr<Algorithm>();
-    return c();
+    return Ptr<Algorithm>(c());
 }
 
 Algorithm::Algorithm()
@@ -174,42 +174,42 @@ Algorithm::~Algorithm()
 {
 }
 
-std::string Algorithm::name() const
+String Algorithm::name() const
 {
     return info()->name();
 }
 
-void Algorithm::set(const std::string& parameter, int value)
+void Algorithm::set(const String& parameter, int value)
 {
     info()->set(this, parameter.c_str(), ParamType<int>::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, double value)
+void Algorithm::set(const String& parameter, double value)
 {
     info()->set(this, parameter.c_str(), ParamType<double>::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, bool value)
+void Algorithm::set(const String& parameter, bool value)
 {
     info()->set(this, parameter.c_str(), ParamType<bool>::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, const std::string& value)
+void Algorithm::set(const String& parameter, const String& value)
 {
-    info()->set(this, parameter.c_str(), ParamType<std::string>::type, &value);
+    info()->set(this, parameter.c_str(), ParamType<String>::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, const Mat& value)
+void Algorithm::set(const String& parameter, const Mat& value)
 {
     info()->set(this, parameter.c_str(), ParamType<Mat>::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, const std::vector<Mat>& value)
+void Algorithm::set(const String& parameter, const std::vector<Mat>& value)
 {
     info()->set(this, parameter.c_str(), ParamType<std::vector<Mat> >::type, &value);
 }
 
-void Algorithm::set(const std::string& parameter, const Ptr<Algorithm>& value)
+void Algorithm::set(const String& parameter, const Ptr<Algorithm>& value)
 {
     info()->set(this, parameter.c_str(), ParamType<Algorithm>::type, &value);
 }
@@ -229,9 +229,9 @@ void Algorithm::set(const char* parameter, bool value)
     info()->set(this, parameter, ParamType<bool>::type, &value);
 }
 
-void Algorithm::set(const char* parameter, const std::string& value)
+void Algorithm::set(const char* parameter, const String& value)
 {
-    info()->set(this, parameter, ParamType<std::string>::type, &value);
+    info()->set(this, parameter, ParamType<String>::type, &value);
 }
 
 void Algorithm::set(const char* parameter, const Mat& value)
@@ -250,37 +250,37 @@ void Algorithm::set(const char* parameter, const Ptr<Algorithm>& value)
 }
 
 
-void Algorithm::setInt(const std::string& parameter, int value)
+void Algorithm::setInt(const String& parameter, int value)
 {
     info()->set(this, parameter.c_str(), ParamType<int>::type, &value);
 }
 
-void Algorithm::setDouble(const std::string& parameter, double value)
+void Algorithm::setDouble(const String& parameter, double value)
 {
     info()->set(this, parameter.c_str(), ParamType<double>::type, &value);
 }
 
-void Algorithm::setBool(const std::string& parameter, bool value)
+void Algorithm::setBool(const String& parameter, bool value)
 {
     info()->set(this, parameter.c_str(), ParamType<bool>::type, &value);
 }
 
-void Algorithm::setString(const std::string& parameter, const std::string& value)
+void Algorithm::setString(const String& parameter, const String& value)
 {
-    info()->set(this, parameter.c_str(), ParamType<std::string>::type, &value);
+    info()->set(this, parameter.c_str(), ParamType<String>::type, &value);
 }
 
-void Algorithm::setMat(const std::string& parameter, const Mat& value)
+void Algorithm::setMat(const String& parameter, const Mat& value)
 {
     info()->set(this, parameter.c_str(), ParamType<Mat>::type, &value);
 }
 
-void Algorithm::setMatVector(const std::string& parameter, const std::vector<Mat>& value)
+void Algorithm::setMatVector(const String& parameter, const std::vector<Mat>& value)
 {
     info()->set(this, parameter.c_str(), ParamType<std::vector<Mat> >::type, &value);
 }
 
-void Algorithm::setAlgorithm(const std::string& parameter, const Ptr<Algorithm>& value)
+void Algorithm::setAlgorithm(const String& parameter, const Ptr<Algorithm>& value)
 {
     info()->set(this, parameter.c_str(), ParamType<Algorithm>::type, &value);
 }
@@ -300,9 +300,9 @@ void Algorithm::setBool(const char* parameter, bool value)
     info()->set(this, parameter, ParamType<bool>::type, &value);
 }
 
-void Algorithm::setString(const char* parameter, const std::string& value)
+void Algorithm::setString(const char* parameter, const String& value)
 {
-    info()->set(this, parameter, ParamType<std::string>::type, &value);
+    info()->set(this, parameter, ParamType<String>::type, &value);
 }
 
 void Algorithm::setMat(const char* parameter, const Mat& value)
@@ -322,47 +322,47 @@ void Algorithm::setAlgorithm(const char* parameter, const Ptr<Algorithm>& value)
 
 
 
-int Algorithm::getInt(const std::string& parameter) const
+int Algorithm::getInt(const String& parameter) const
 {
     return get<int>(parameter);
 }
 
-double Algorithm::getDouble(const std::string& parameter) const
+double Algorithm::getDouble(const String& parameter) const
 {
     return get<double>(parameter);
 }
 
-bool Algorithm::getBool(const std::string& parameter) const
+bool Algorithm::getBool(const String& parameter) const
 {
     return get<bool>(parameter);
 }
 
-std::string Algorithm::getString(const std::string& parameter) const
+String Algorithm::getString(const String& parameter) const
 {
-    return get<std::string>(parameter);
+    return get<String>(parameter);
 }
 
-Mat Algorithm::getMat(const std::string& parameter) const
+Mat Algorithm::getMat(const String& parameter) const
 {
     return get<Mat>(parameter);
 }
 
-std::vector<Mat> Algorithm::getMatVector(const std::string& parameter) const
+std::vector<Mat> Algorithm::getMatVector(const String& parameter) const
 {
     return get<std::vector<Mat> >(parameter);
 }
 
-Ptr<Algorithm> Algorithm::getAlgorithm(const std::string& parameter) const
+Ptr<Algorithm> Algorithm::getAlgorithm(const String& parameter) const
 {
     return get<Algorithm>(parameter);
 }
 
-std::string Algorithm::paramHelp(const std::string& parameter) const
+String Algorithm::paramHelp(const String& parameter) const
 {
     return info()->paramHelp(parameter.c_str());
 }
 
-int Algorithm::paramType(const std::string& parameter) const
+int Algorithm::paramType(const String& parameter) const
 {
     return info()->paramType(parameter.c_str());
 }
@@ -372,7 +372,7 @@ int Algorithm::paramType(const char* parameter) const
     return info()->paramType(parameter);
 }
 
-void Algorithm::getParams(std::vector<std::string>& names) const
+void Algorithm::getParams(std::vector<String>& names) const
 {
     info()->getParams(names);
 }
@@ -388,7 +388,7 @@ void Algorithm::read(const FileNode& fn)
 }
 
 
-AlgorithmInfo::AlgorithmInfo(const std::string& _name, Algorithm::Constructor create)
+AlgorithmInfo::AlgorithmInfo(const String& _name, Algorithm::Constructor create)
 {
     data = new AlgorithmInfoData;
     data->_name = _name;
@@ -408,7 +408,7 @@ void AlgorithmInfo::write(const Algorithm* algo, FileStorage& fs) const
     for( i = 0; i < nparams; i++ )
     {
         const Param& p = data->params.vec[i].second;
-        const std::string& pname = data->params.vec[i].first;
+        const String& pname = data->params.vec[i].first;
         if( p.type == Param::INT )
             cv::write(fs, pname, algo->get<int>(pname));
         else if( p.type == Param::BOOLEAN )
@@ -416,14 +416,14 @@ void AlgorithmInfo::write(const Algorithm* algo, FileStorage& fs) const
         else if( p.type == Param::REAL )
             cv::write(fs, pname, algo->get<double>(pname));
         else if( p.type == Param::STRING )
-            cv::write(fs, pname, algo->get<std::string>(pname));
+            cv::write(fs, pname, algo->get<String>(pname));
         else if( p.type == Param::MAT )
             cv::write(fs, pname, algo->get<Mat>(pname));
         else if( p.type == Param::MAT_VECTOR )
             cv::write(fs, pname, algo->get<std::vector<Mat> >(pname));
         else if( p.type == Param::ALGORITHM )
         {
-            WriteStructContext ws(fs, pname, CV_NODE_MAP);
+            internal::WriteStructContext ws(fs, pname, CV_NODE_MAP);
             Ptr<Algorithm> nestedAlgo = algo->get<Algorithm>(pname);
             nestedAlgo->write(fs);
         }
@@ -437,7 +437,7 @@ void AlgorithmInfo::write(const Algorithm* algo, FileStorage& fs) const
             cv::write(fs, pname, algo->getInt(pname));
         else
         {
-            std::string msg = format("unknown/unsupported type of '%s' parameter == %d", pname.c_str(), p.type);
+            String msg = format("unknown/unsupported type of '%s' parameter == %d", pname.c_str(), p.type);
             CV_Error( CV_StsUnsupportedFormat, msg.c_str());
         }
     }
@@ -451,7 +451,7 @@ void AlgorithmInfo::read(Algorithm* algo, const FileNode& fn) const
     for( i = 0; i < nparams; i++ )
     {
         const Param& p = data->params.vec[i].second;
-        const std::string& pname = data->params.vec[i].first;
+        const String& pname = data->params.vec[i].first;
         const FileNode n = fn[pname];
         if( n.empty() )
             continue;
@@ -472,7 +472,7 @@ void AlgorithmInfo::read(Algorithm* algo, const FileNode& fn) const
         }
         else if( p.type == Param::STRING )
         {
-            std::string val = (std::string)n;
+            String val = (String)n;
             info->set(algo, pname.c_str(), p.type, &val, true);
         }
         else if( p.type == Param::MAT )
@@ -489,8 +489,8 @@ void AlgorithmInfo::read(Algorithm* algo, const FileNode& fn) const
         }
         else if( p.type == Param::ALGORITHM )
         {
-            Ptr<Algorithm> nestedAlgo = Algorithm::_create((std::string)n["name"]);
-            CV_Assert( !nestedAlgo.empty() );
+            Ptr<Algorithm> nestedAlgo = Algorithm::_create((String)n["name"]);
+            CV_Assert( nestedAlgo );
             nestedAlgo->read(n);
             info->set(algo, pname.c_str(), p.type, &nestedAlgo, true);
         }
@@ -516,13 +516,13 @@ void AlgorithmInfo::read(Algorithm* algo, const FileNode& fn) const
         }
         else
         {
-            std::string msg = format("unknown/unsupported type of '%s' parameter == %d", pname.c_str(), p.type);
+            String msg = format("unknown/unsupported type of '%s' parameter == %d", pname.c_str(), p.type);
             CV_Error( CV_StsUnsupportedFormat, msg.c_str());
         }
     }
 }
 
-std::string AlgorithmInfo::name() const
+String AlgorithmInfo::name() const
 {
     return data->_name;
 }
@@ -532,7 +532,7 @@ union GetSetParam
     int (Algorithm::*get_int)() const;
     bool (Algorithm::*get_bool)() const;
     double (Algorithm::*get_double)() const;
-    std::string (Algorithm::*get_string)() const;
+    String (Algorithm::*get_string)() const;
     Mat (Algorithm::*get_mat)() const;
     std::vector<Mat> (Algorithm::*get_mat_vector)() const;
     Ptr<Algorithm> (Algorithm::*get_algo)() const;
@@ -544,7 +544,7 @@ union GetSetParam
     void (Algorithm::*set_int)(int);
     void (Algorithm::*set_bool)(bool);
     void (Algorithm::*set_double)(double);
-    void (Algorithm::*set_string)(const std::string&);
+    void (Algorithm::*set_string)(const String&);
     void (Algorithm::*set_mat)(const Mat&);
     void (Algorithm::*set_mat_vector)(const std::vector<Mat>&);
     void (Algorithm::*set_algo)(const Ptr<Algorithm>&);
@@ -554,9 +554,9 @@ union GetSetParam
     void (Algorithm::*set_uchar)(uchar);
 };
 
-static std::string getNameOfType(int argType);
+static String getNameOfType(int argType);
 
-static std::string getNameOfType(int argType)
+static String getNameOfType(int argType)
 {
     switch(argType)
     {
@@ -576,37 +576,37 @@ static std::string getNameOfType(int argType)
     return "";
 }
 
-static std::string getErrorMessageForWrongArgumentInSetter(std::string algoName, std::string paramName, int paramType, int argType)
+static String getErrorMessageForWrongArgumentInSetter(String algoName, String paramName, int paramType, int argType)
 {
-    std::string message = std::string("Argument error: the setter")
+    String message = String("Argument error: the setter")
         + " method was called for the parameter '" + paramName + "' of the algorithm '" + algoName
         +"', the parameter has " + getNameOfType(paramType) + " type, ";
 
     if (paramType == Param::INT || paramType == Param::BOOLEAN || paramType == Param::REAL
             || paramType == Param::FLOAT || paramType == Param::UNSIGNED_INT || paramType == Param::UINT64 || paramType == Param::UCHAR)
     {
-        message += "so it should be set by integer, unsigned integer, uint64, unsigned char, boolean, float or double value, ";
+        message = message + "so it should be set by integer, unsigned integer, uint64, unsigned char, boolean, float or double value, ";
     }
-    message += "but the setter was called with " + getNameOfType(argType) + " value";
+    message = message + "but the setter was called with " + getNameOfType(argType) + " value";
 
     return message;
 }
 
-static std::string getErrorMessageForWrongArgumentInGetter(std::string algoName, std::string paramName, int paramType, int argType)
+static String getErrorMessageForWrongArgumentInGetter(String algoName, String paramName, int paramType, int argType)
 {
-    std::string message = std::string("Argument error: the getter")
+    String message = String("Argument error: the getter")
         + " method was called for the parameter '" + paramName + "' of the algorithm '" + algoName
         +"', the parameter has " + getNameOfType(paramType) + " type, ";
 
     if (paramType == Param::BOOLEAN)
     {
-        message += "so it should be get as integer, unsigned integer, uint64, boolean, unsigned char, float or double value, ";
+        message = message + "so it should be get as integer, unsigned integer, uint64, boolean, unsigned char, float or double value, ";
     }
     else if (paramType == Param::INT || paramType == Param::UNSIGNED_INT || paramType == Param::UINT64 || paramType == Param::UCHAR)
     {
-        message += "so it should be get as integer, unsigned integer, uint64, unsigned char, float or double value, ";
+        message = message + "so it should be get as integer, unsigned integer, uint64, unsigned char, float or double value, ";
     }
-    message += "but the getter was called to get a " + getNameOfType(argType) + " value";
+    message = message + "but the getter was called to get a " + getNameOfType(argType) + " value";
 
     return message;
 }
@@ -630,7 +630,7 @@ void AlgorithmInfo::set(Algorithm* algo, const char* parameter, int argType, con
         if ( !( p->type == Param::INT || p->type == Param::REAL || p->type == Param::BOOLEAN
                 || p->type == Param::UNSIGNED_INT || p->type == Param::UINT64 || p->type == Param::FLOAT || argType == Param::UCHAR) )
         {
-            std::string message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -790,21 +790,21 @@ void AlgorithmInfo::set(Algorithm* algo, const char* parameter, int argType, con
     {
         if( p->type != Param::STRING )
         {
-            std::string message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
-        const std::string& val = *(const std::string*)value;
+        const String& val = *(const String*)value;
         if( p->setter )
             (algo->*f.set_string)(val);
         else
-            *(std::string*)((uchar*)algo + p->offset) = val;
+            *(String*)((uchar*)algo + p->offset) = val;
     }
     else if( argType == Param::MAT )
     {
         if( p->type != Param::MAT )
         {
-            std::string message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -818,7 +818,7 @@ void AlgorithmInfo::set(Algorithm* algo, const char* parameter, int argType, con
     {
         if( p->type != Param::MAT_VECTOR )
         {
-            std::string message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -832,7 +832,7 @@ void AlgorithmInfo::set(Algorithm* algo, const char* parameter, int argType, con
     {
         if( p->type != Param::ALGORITHM )
         {
-            std::string message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInSetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -862,7 +862,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if (!( argType == Param::INT || argType == Param::REAL || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64 || argType == Param::UCHAR))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             int val = p->getter ? (algo->*f.get_int)() : *(int*)((uchar*)algo + p->offset);
@@ -887,7 +887,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if (!( argType == Param::INT || argType == Param::BOOLEAN || argType == Param::REAL || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64 || argType == Param::UCHAR))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             bool val = p->getter ? (algo->*f.get_bool)() : *(bool*)((uchar*)algo + p->offset);
@@ -913,7 +913,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if(!( argType == Param::REAL || argType == Param::FLOAT))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             double val = p->getter ? (algo->*f.get_double)() : *(double*)((uchar*)algo + p->offset);
@@ -929,7 +929,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if(!( argType == Param::REAL || argType == Param::FLOAT))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             float val = p->getter ? (algo->*f.get_float)() : *(float*)((uchar*)algo + p->offset);
@@ -945,7 +945,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if (!( argType == Param::INT || argType == Param::REAL || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64 || argType == Param::UCHAR))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             unsigned int val = p->getter ? (algo->*f.get_uint)() : *(unsigned int*)((uchar*)algo + p->offset);
@@ -969,7 +969,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if (!( argType == Param::INT || argType == Param::REAL || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64 || argType == Param::UCHAR))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
         }
             uint64 val = p->getter ? (algo->*f.get_uint64)() : *(uint64*)((uchar*)algo + p->offset);
@@ -993,7 +993,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
         {
             if (!( argType == Param::INT || argType == Param::REAL || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64 || argType == Param::UCHAR))
             {
-                std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+                String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
                 CV_Error(CV_StsBadArg, message);
             }
             uchar val = p->getter ? (algo->*f.get_uchar)() : *(uchar*)((uchar*)algo + p->offset);
@@ -1021,18 +1021,18 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
     {
         if( p->type != Param::STRING )
         {
-            std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
-        *(std::string*)value = p->getter ? (algo->*f.get_string)() :
-            *(std::string*)((uchar*)algo + p->offset);
+        *(String*)value = p->getter ? (algo->*f.get_string)() :
+            *(String*)((uchar*)algo + p->offset);
     }
     else if( argType == Param::MAT )
     {
         if( p->type != Param::MAT )
         {
-            std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -1043,7 +1043,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
     {
         if( p->type != Param::MAT_VECTOR )
         {
-            std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -1054,7 +1054,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
     {
         if( p->type != Param::ALGORITHM )
         {
-            std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+            String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
             CV_Error(CV_StsBadArg, message);
         }
 
@@ -1063,7 +1063,7 @@ void AlgorithmInfo::get(const Algorithm* algo, const char* parameter, int argTyp
     }
     else
     {
-        std::string message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
+        String message = getErrorMessageForWrongArgumentInGetter(algo->name(), parameter, p->type, argType);
         CV_Error(CV_StsBadArg, message);
     }
 }
@@ -1078,7 +1078,7 @@ int AlgorithmInfo::paramType(const char* parameter) const
 }
 
 
-std::string AlgorithmInfo::paramHelp(const char* parameter) const
+String AlgorithmInfo::paramHelp(const char* parameter) const
 {
     const Param* p = findstr(data->params, parameter);
     if( !p )
@@ -1087,7 +1087,7 @@ std::string AlgorithmInfo::paramHelp(const char* parameter) const
 }
 
 
-void AlgorithmInfo::getParams(std::vector<std::string>& names) const
+void AlgorithmInfo::getParams(std::vector<String>& names) const
 {
     data->params.get_keys(names);
 }
@@ -1096,7 +1096,7 @@ void AlgorithmInfo::getParams(std::vector<std::string>& names) const
 void AlgorithmInfo::addParam_(Algorithm& algo, const char* parameter, int argType,
                               void* value, bool readOnly,
                               Algorithm::Getter getter, Algorithm::Setter setter,
-                              const std::string& help)
+                              const String& help)
 {
     CV_Assert( argType == Param::INT || argType == Param::BOOLEAN ||
                argType == Param::REAL || argType == Param::STRING ||
@@ -1104,7 +1104,7 @@ void AlgorithmInfo::addParam_(Algorithm& algo, const char* parameter, int argTyp
                argType == Param::ALGORITHM
                || argType == Param::FLOAT || argType == Param::UNSIGNED_INT || argType == Param::UINT64
                || argType == Param::UCHAR);
-    data->params.add(std::string(parameter), Param(argType, readOnly,
+    data->params.add(String(parameter), Param(argType, readOnly,
                      (int)((size_t)value - (size_t)(void*)&algo),
                      getter, setter, help));
 }
@@ -1114,7 +1114,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              int& value, bool readOnly,
                              int (Algorithm::*getter)(),
                              void (Algorithm::*setter)(int),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<int>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1124,7 +1124,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              bool& value, bool readOnly,
                              int (Algorithm::*getter)(),
                              void (Algorithm::*setter)(int),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<bool>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1134,19 +1134,19 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              double& value, bool readOnly,
                              double (Algorithm::*getter)(),
                              void (Algorithm::*setter)(double),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<double>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
 }
 
 void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
-                             std::string& value, bool readOnly,
-                             std::string (Algorithm::*getter)(),
-                             void (Algorithm::*setter)(const std::string&),
-                             const std::string& help)
+                             String& value, bool readOnly,
+                             String (Algorithm::*getter)(),
+                             void (Algorithm::*setter)(const String&),
+                             const String& help)
 {
-    addParam_(algo, parameter, ParamType<std::string>::type, &value, readOnly,
+    addParam_(algo, parameter, ParamType<String>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
 }
 
@@ -1154,7 +1154,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              Mat& value, bool readOnly,
                              Mat (Algorithm::*getter)(),
                              void (Algorithm::*setter)(const Mat&),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<Mat>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1164,7 +1164,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              std::vector<Mat>& value, bool readOnly,
                              std::vector<Mat> (Algorithm::*getter)(),
                              void (Algorithm::*setter)(const std::vector<Mat>&),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<std::vector<Mat> >::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1174,7 +1174,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              Ptr<Algorithm>& value, bool readOnly,
                              Ptr<Algorithm> (Algorithm::*getter)(),
                              void (Algorithm::*setter)(const Ptr<Algorithm>&),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<Algorithm>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1184,7 +1184,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              float& value, bool readOnly,
                              float (Algorithm::*getter)(),
                              void (Algorithm::*setter)(float),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<float>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1194,7 +1194,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              unsigned int& value, bool readOnly,
                              unsigned int (Algorithm::*getter)(),
                              void (Algorithm::*setter)(unsigned int),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<unsigned int>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1204,7 +1204,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              uint64& value, bool readOnly,
                              uint64 (Algorithm::*getter)(),
                              void (Algorithm::*setter)(uint64),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<uint64>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
@@ -1214,7 +1214,7 @@ void AlgorithmInfo::addParam(Algorithm& algo, const char* parameter,
                              uchar& value, bool readOnly,
                              uchar (Algorithm::*getter)(),
                              void (Algorithm::*setter)(uchar),
-                             const std::string& help)
+                             const String& help)
 {
     addParam_(algo, parameter, ParamType<uchar>::type, &value, readOnly,
               (Algorithm::Getter)getter, (Algorithm::Setter)setter, help);
diff --git a/modules/core/src/alloc.cpp b/modules/core/src/alloc.cpp
index 1944ed17de..8e218738b3 100644
--- a/modules/core/src/alloc.cpp
+++ b/modules/core/src/alloc.cpp
@@ -94,9 +94,20 @@ void fastFree(void* ptr)
 #define STAT(stmt)
 
 #ifdef WIN32
+#if (_WIN32_WINNT >= 0x0602)
+#include <synchapi.h>
+#endif
+
 struct CriticalSection
 {
-    CriticalSection() { InitializeCriticalSection(&cs); }
+    CriticalSection()
+    {
+#if (_WIN32_WINNT >= 0x0600)
+        InitializeCriticalSectionEx(&cs, 1000, 0);
+#else
+        InitializeCriticalSection(&cs);
+#endif
+    }
     ~CriticalSection() { DeleteCriticalSection(&cs); }
     void lock() { EnterCriticalSection(&cs); }
     void unlock() { LeaveCriticalSection(&cs); }
@@ -683,11 +694,6 @@ void fastFree( void* ptr )
 
 }
 
-CV_IMPL void cvSetMemoryManager( CvAllocFunc, CvFreeFunc, void * )
-{
-    CV_Error( -1, "Custom memory allocator is not supported" );
-}
-
 CV_IMPL void* cvAlloc( size_t size )
 {
     return cv::fastMalloc( size );
diff --git a/modules/core/src/arithm.cpp b/modules/core/src/arithm.cpp
index b76dd7fd40..313d06d881 100644
--- a/modules/core/src/arithm.cpp
+++ b/modules/core/src/arithm.cpp
@@ -1131,23 +1131,33 @@ static void binary_op(InputArray _src1, InputArray _src2, OutputArray _dst,
     }
 }
 
-static BinaryFunc maxTab[] =
+static BinaryFunc* getMaxTab()
 {
-    (BinaryFunc)GET_OPTIMIZED(max8u), (BinaryFunc)GET_OPTIMIZED(max8s),
-    (BinaryFunc)GET_OPTIMIZED(max16u), (BinaryFunc)GET_OPTIMIZED(max16s),
-    (BinaryFunc)GET_OPTIMIZED(max32s),
-    (BinaryFunc)GET_OPTIMIZED(max32f), (BinaryFunc)max64f,
-    0
-};
+    static BinaryFunc maxTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(max8u), (BinaryFunc)GET_OPTIMIZED(max8s),
+        (BinaryFunc)GET_OPTIMIZED(max16u), (BinaryFunc)GET_OPTIMIZED(max16s),
+        (BinaryFunc)GET_OPTIMIZED(max32s),
+        (BinaryFunc)GET_OPTIMIZED(max32f), (BinaryFunc)max64f,
+        0
+    };
 
-static BinaryFunc minTab[] =
+    return maxTab;
+}
+
+static BinaryFunc* getMinTab()
 {
-    (BinaryFunc)GET_OPTIMIZED(min8u), (BinaryFunc)GET_OPTIMIZED(min8s),
-    (BinaryFunc)GET_OPTIMIZED(min16u), (BinaryFunc)GET_OPTIMIZED(min16s),
-    (BinaryFunc)GET_OPTIMIZED(min32s),
-    (BinaryFunc)GET_OPTIMIZED(min32f), (BinaryFunc)min64f,
-    0
-};
+    static BinaryFunc minTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(min8u), (BinaryFunc)GET_OPTIMIZED(min8s),
+        (BinaryFunc)GET_OPTIMIZED(min16u), (BinaryFunc)GET_OPTIMIZED(min16s),
+        (BinaryFunc)GET_OPTIMIZED(min32s),
+        (BinaryFunc)GET_OPTIMIZED(min32f), (BinaryFunc)min64f,
+        0
+    };
+
+    return minTab;
+}
 
 }
 
@@ -1177,37 +1187,26 @@ void cv::bitwise_not(InputArray a, OutputArray c, InputArray mask)
 
 void cv::max( InputArray src1, InputArray src2, OutputArray dst )
 {
-    binary_op(src1, src2, dst, noArray(), maxTab, false );
+    binary_op(src1, src2, dst, noArray(), getMaxTab(), false );
 }
 
 void cv::min( InputArray src1, InputArray src2, OutputArray dst )
 {
-    binary_op(src1, src2, dst, noArray(), minTab, false );
+    binary_op(src1, src2, dst, noArray(), getMinTab(), false );
 }
 
 void cv::max(const Mat& src1, const Mat& src2, Mat& dst)
 {
     OutputArray _dst(dst);
-    binary_op(src1, src2, _dst, noArray(), maxTab, false );
+    binary_op(src1, src2, _dst, noArray(), getMaxTab(), false );
 }
 
 void cv::min(const Mat& src1, const Mat& src2, Mat& dst)
 {
     OutputArray _dst(dst);
-    binary_op(src1, src2, _dst, noArray(), minTab, false );
+    binary_op(src1, src2, _dst, noArray(), getMinTab(), false );
 }
 
-void cv::max(const Mat& src1, double src2, Mat& dst)
-{
-    OutputArray _dst(dst);
-    binary_op(src1, src2, _dst, noArray(), maxTab, false );
-}
-
-void cv::min(const Mat& src1, double src2, Mat& dst)
-{
-    OutputArray _dst(dst);
-    binary_op(src1, src2, _dst, noArray(), minTab, false );
-}
 
 /****************************************************************************************\
 *                                      add/subtract                                      *
@@ -1493,39 +1492,54 @@ static void arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst,
     }
 }
 
-static BinaryFunc addTab[] =
+static BinaryFunc* getAddTab()
 {
-    (BinaryFunc)GET_OPTIMIZED(add8u), (BinaryFunc)GET_OPTIMIZED(add8s),
-    (BinaryFunc)GET_OPTIMIZED(add16u), (BinaryFunc)GET_OPTIMIZED(add16s),
-    (BinaryFunc)GET_OPTIMIZED(add32s),
-    (BinaryFunc)GET_OPTIMIZED(add32f), (BinaryFunc)add64f,
-    0
-};
+    static BinaryFunc addTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(add8u), (BinaryFunc)GET_OPTIMIZED(add8s),
+        (BinaryFunc)GET_OPTIMIZED(add16u), (BinaryFunc)GET_OPTIMIZED(add16s),
+        (BinaryFunc)GET_OPTIMIZED(add32s),
+        (BinaryFunc)GET_OPTIMIZED(add32f), (BinaryFunc)add64f,
+        0
+    };
 
-static BinaryFunc subTab[] =
-{
-    (BinaryFunc)GET_OPTIMIZED(sub8u), (BinaryFunc)GET_OPTIMIZED(sub8s),
-    (BinaryFunc)GET_OPTIMIZED(sub16u), (BinaryFunc)GET_OPTIMIZED(sub16s),
-    (BinaryFunc)GET_OPTIMIZED(sub32s),
-    (BinaryFunc)GET_OPTIMIZED(sub32f), (BinaryFunc)sub64f,
-    0
-};
+    return addTab;
+}
 
-static BinaryFunc absdiffTab[] =
+static BinaryFunc* getSubTab()
 {
-    (BinaryFunc)GET_OPTIMIZED(absdiff8u), (BinaryFunc)GET_OPTIMIZED(absdiff8s),
-    (BinaryFunc)GET_OPTIMIZED(absdiff16u), (BinaryFunc)GET_OPTIMIZED(absdiff16s),
-    (BinaryFunc)GET_OPTIMIZED(absdiff32s),
-    (BinaryFunc)GET_OPTIMIZED(absdiff32f), (BinaryFunc)absdiff64f,
-    0
-};
+    static BinaryFunc subTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(sub8u), (BinaryFunc)GET_OPTIMIZED(sub8s),
+        (BinaryFunc)GET_OPTIMIZED(sub16u), (BinaryFunc)GET_OPTIMIZED(sub16s),
+        (BinaryFunc)GET_OPTIMIZED(sub32s),
+        (BinaryFunc)GET_OPTIMIZED(sub32f), (BinaryFunc)sub64f,
+        0
+    };
+
+    return subTab;
+}
+
+static BinaryFunc* getAbsDiffTab()
+{
+    static BinaryFunc absDiffTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(absdiff8u), (BinaryFunc)GET_OPTIMIZED(absdiff8s),
+        (BinaryFunc)GET_OPTIMIZED(absdiff16u), (BinaryFunc)GET_OPTIMIZED(absdiff16s),
+        (BinaryFunc)GET_OPTIMIZED(absdiff32s),
+        (BinaryFunc)GET_OPTIMIZED(absdiff32f), (BinaryFunc)absdiff64f,
+        0
+    };
+
+    return absDiffTab;
+}
 
 }
 
 void cv::add( InputArray src1, InputArray src2, OutputArray dst,
           InputArray mask, int dtype )
 {
-    arithm_op(src1, src2, dst, mask, dtype, addTab );
+    arithm_op(src1, src2, dst, mask, dtype, getAddTab() );
 }
 
 void cv::subtract( InputArray src1, InputArray src2, OutputArray dst,
@@ -1560,12 +1574,12 @@ void cv::subtract( InputArray src1, InputArray src2, OutputArray dst,
         }
     }
 #endif
-    arithm_op(src1, src2, dst, mask, dtype, subTab );
+    arithm_op(src1, src2, dst, mask, dtype, getSubTab() );
 }
 
 void cv::absdiff( InputArray src1, InputArray src2, OutputArray dst )
 {
-    arithm_op(src1, src2, dst, noArray(), -1, absdiffTab);
+    arithm_op(src1, src2, dst, noArray(), -1, getAbsDiffTab());
 }
 
 /****************************************************************************************\
@@ -1855,46 +1869,60 @@ static void recip64f( const double* src1, size_t step1, const double* src2, size
 }
 
 
-static BinaryFunc mulTab[] =
+static BinaryFunc* getMulTab()
 {
-    (BinaryFunc)mul8u, (BinaryFunc)mul8s, (BinaryFunc)mul16u,
-    (BinaryFunc)mul16s, (BinaryFunc)mul32s, (BinaryFunc)mul32f,
-    (BinaryFunc)mul64f, 0
-};
+    static BinaryFunc mulTab[] =
+    {
+        (BinaryFunc)mul8u, (BinaryFunc)mul8s, (BinaryFunc)mul16u,
+        (BinaryFunc)mul16s, (BinaryFunc)mul32s, (BinaryFunc)mul32f,
+        (BinaryFunc)mul64f, 0
+    };
 
-static BinaryFunc divTab[] =
+    return mulTab;
+}
+
+static BinaryFunc* getDivTab()
 {
-    (BinaryFunc)div8u, (BinaryFunc)div8s, (BinaryFunc)div16u,
-    (BinaryFunc)div16s, (BinaryFunc)div32s, (BinaryFunc)div32f,
-    (BinaryFunc)div64f, 0
-};
+    static BinaryFunc divTab[] =
+    {
+        (BinaryFunc)div8u, (BinaryFunc)div8s, (BinaryFunc)div16u,
+        (BinaryFunc)div16s, (BinaryFunc)div32s, (BinaryFunc)div32f,
+        (BinaryFunc)div64f, 0
+    };
 
-static BinaryFunc recipTab[] =
+    return divTab;
+}
+
+static BinaryFunc* getRecipTab()
 {
-    (BinaryFunc)recip8u, (BinaryFunc)recip8s, (BinaryFunc)recip16u,
-    (BinaryFunc)recip16s, (BinaryFunc)recip32s, (BinaryFunc)recip32f,
-    (BinaryFunc)recip64f, 0
-};
+    static BinaryFunc recipTab[] =
+    {
+        (BinaryFunc)recip8u, (BinaryFunc)recip8s, (BinaryFunc)recip16u,
+        (BinaryFunc)recip16s, (BinaryFunc)recip32s, (BinaryFunc)recip32f,
+        (BinaryFunc)recip64f, 0
+    };
 
+    return recipTab;
+}
 
 }
 
 void cv::multiply(InputArray src1, InputArray src2,
                   OutputArray dst, double scale, int dtype)
 {
-    arithm_op(src1, src2, dst, noArray(), dtype, mulTab, true, &scale);
+    arithm_op(src1, src2, dst, noArray(), dtype, getMulTab(), true, &scale);
 }
 
 void cv::divide(InputArray src1, InputArray src2,
                 OutputArray dst, double scale, int dtype)
 {
-    arithm_op(src1, src2, dst, noArray(), dtype, divTab, true, &scale);
+    arithm_op(src1, src2, dst, noArray(), dtype, getDivTab(), true, &scale);
 }
 
 void cv::divide(double scale, InputArray src2,
                 OutputArray dst, int dtype)
 {
-    arithm_op(src2, src2, dst, noArray(), dtype, recipTab, true, &scale);
+    arithm_op(src2, src2, dst, noArray(), dtype, getRecipTab(), true, &scale);
 }
 
 /****************************************************************************************\
@@ -2037,12 +2065,17 @@ static void addWeighted64f( const double* src1, size_t step1, const double* src2
     addWeighted_<double, double>(src1, step1, src2, step2, dst, step, sz, scalars);
 }
 
-static BinaryFunc addWeightedTab[] =
+static BinaryFunc* getAddWeightedTab()
 {
-    (BinaryFunc)GET_OPTIMIZED(addWeighted8u), (BinaryFunc)GET_OPTIMIZED(addWeighted8s), (BinaryFunc)GET_OPTIMIZED(addWeighted16u),
-    (BinaryFunc)GET_OPTIMIZED(addWeighted16s), (BinaryFunc)GET_OPTIMIZED(addWeighted32s), (BinaryFunc)addWeighted32f,
-    (BinaryFunc)addWeighted64f, 0
-};
+    static BinaryFunc addWeightedTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(addWeighted8u), (BinaryFunc)GET_OPTIMIZED(addWeighted8s), (BinaryFunc)GET_OPTIMIZED(addWeighted16u),
+        (BinaryFunc)GET_OPTIMIZED(addWeighted16s), (BinaryFunc)GET_OPTIMIZED(addWeighted32s), (BinaryFunc)addWeighted32f,
+        (BinaryFunc)addWeighted64f, 0
+    };
+
+    return addWeightedTab;
+}
 
 }
 
@@ -2050,7 +2083,7 @@ void cv::addWeighted( InputArray src1, double alpha, InputArray src2,
                       double beta, double gamma, OutputArray dst, int dtype )
 {
     double scalars[] = {alpha, beta, gamma};
-    arithm_op(src1, src2, dst, noArray(), dtype, addWeightedTab, true, scalars);
+    arithm_op(src1, src2, dst, noArray(), dtype, getAddWeightedTab(), true, scalars);
 }
 
 
@@ -2120,10 +2153,30 @@ cmp_(const T* src1, size_t step1, const T* src2, size_t step2,
     }
 }
 
+#if ARITHM_USE_IPP
+inline static IppCmpOp convert_cmp(int _cmpop)
+{
+    return _cmpop == CMP_EQ ? ippCmpEq :
+        _cmpop == CMP_GT ? ippCmpGreater :
+        _cmpop == CMP_GE ? ippCmpGreaterEq :
+        _cmpop == CMP_LT ? ippCmpLess :
+        _cmpop == CMP_LE ? ippCmpLessEq :
+        (IppCmpOp)-1;
+}
+#endif
 
 static void cmp8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2,
                   uchar* dst, size_t step, Size size, void* _cmpop)
 {
+#if ARITHM_USE_IPP
+    IppCmpOp op = convert_cmp(*(int *)_cmpop);
+    if( op  >= 0 )
+    {
+        fixSteps(size, sizeof(dst[0]), step1, step2, step);
+        if( ippiCompare_8u_C1R(src1, (int)step1, src2, (int)step2, dst, (int)step, (IppiSize&)size, op) >= 0 )
+            return;
+    }
+#endif
   //vz optimized  cmp_(src1, step1, src2, step2, dst, step, size, *(int*)_cmpop);
     int code = *(int*)_cmpop;
     step1 /= sizeof(src1[0]);
@@ -2198,12 +2251,30 @@ static void cmp8s(const schar* src1, size_t step1, const schar* src2, size_t ste
 static void cmp16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2,
                   uchar* dst, size_t step, Size size, void* _cmpop)
 {
+#if ARITHM_USE_IPP
+    IppCmpOp op = convert_cmp(*(int *)_cmpop);
+    if( op  >= 0 )
+    {
+        fixSteps(size, sizeof(dst[0]), step1, step2, step);
+        if( ippiCompare_16u_C1R(src1, (int)step1, src2, (int)step2, dst, (int)step, (IppiSize&)size, op) >= 0 )
+            return;
+    }
+#endif
     cmp_(src1, step1, src2, step2, dst, step, size, *(int*)_cmpop);
 }
 
 static void cmp16s(const short* src1, size_t step1, const short* src2, size_t step2,
                   uchar* dst, size_t step, Size size, void* _cmpop)
 {
+#if ARITHM_USE_IPP
+    IppCmpOp op = convert_cmp(*(int *)_cmpop);
+    if( op  > 0 )
+    {
+        fixSteps(size, sizeof(dst[0]), step1, step2, step);
+        if( ippiCompare_16s_C1R(src1, (int)step1, src2, (int)step2, dst, (int)step, (IppiSize&)size, op) >= 0 )
+            return;
+    }
+#endif
    //vz optimized cmp_(src1, step1, src2, step2, dst, step, size, *(int*)_cmpop);
 
     int code = *(int*)_cmpop;
@@ -2301,6 +2372,15 @@ static void cmp32s(const int* src1, size_t step1, const int* src2, size_t step2,
 static void cmp32f(const float* src1, size_t step1, const float* src2, size_t step2,
                   uchar* dst, size_t step, Size size, void* _cmpop)
 {
+#if ARITHM_USE_IPP
+    IppCmpOp op = convert_cmp(*(int *)_cmpop);
+    if( op  >= 0 )
+    {
+        fixSteps(size, sizeof(dst[0]), step1, step2, step);
+        if( ippiCompare_32f_C1R(src1, (int)step1, src2, (int)step2, dst, (int)step, (IppiSize&)size, op) >= 0 )
+            return;
+    }
+#endif
     cmp_(src1, step1, src2, step2, dst, step, size, *(int*)_cmpop);
 }
 
@@ -2310,15 +2390,19 @@ static void cmp64f(const double* src1, size_t step1, const double* src2, size_t
     cmp_(src1, step1, src2, step2, dst, step, size, *(int*)_cmpop);
 }
 
-static BinaryFunc cmpTab[] =
+static BinaryFunc getCmpFunc(int depth)
 {
-    (BinaryFunc)GET_OPTIMIZED(cmp8u), (BinaryFunc)GET_OPTIMIZED(cmp8s),
-    (BinaryFunc)GET_OPTIMIZED(cmp16u), (BinaryFunc)GET_OPTIMIZED(cmp16s),
-    (BinaryFunc)GET_OPTIMIZED(cmp32s),
-    (BinaryFunc)GET_OPTIMIZED(cmp32f), (BinaryFunc)cmp64f,
-    0
-};
+    static BinaryFunc cmpTab[] =
+    {
+        (BinaryFunc)GET_OPTIMIZED(cmp8u), (BinaryFunc)GET_OPTIMIZED(cmp8s),
+        (BinaryFunc)GET_OPTIMIZED(cmp16u), (BinaryFunc)GET_OPTIMIZED(cmp16s),
+        (BinaryFunc)GET_OPTIMIZED(cmp32s),
+        (BinaryFunc)GET_OPTIMIZED(cmp32f), (BinaryFunc)cmp64f,
+        0
+    };
 
+    return cmpTab[depth];
+}
 
 static double getMinVal(int depth)
 {
@@ -2348,7 +2432,7 @@ void cv::compare(InputArray _src1, InputArray _src2, OutputArray _dst, int op)
         _dst.create(src1.size(), CV_8UC(cn));
         Mat dst = _dst.getMat();
         Size sz = getContinuousSize(src1, src2, dst, src1.channels());
-        cmpTab[src1.depth()](src1.data, src1.step, src2.data, src2.step, dst.data, dst.step, sz, &op);
+        getCmpFunc(src1.depth())(src1.data, src1.step, src2.data, src2.step, dst.data, dst.step, sz, &op);
         return;
     }
 
@@ -2380,7 +2464,7 @@ void cv::compare(InputArray _src1, InputArray _src2, OutputArray _dst, int op)
 
     size_t esz = src1.elemSize();
     size_t blocksize0 = (size_t)(BLOCK_SIZE + esz-1)/esz;
-    BinaryFunc func = cmpTab[depth1];
+    BinaryFunc func = getCmpFunc(depth1);
 
     if( !haveScalar )
     {
@@ -2557,12 +2641,17 @@ static void inRangeReduce(const uchar* src, uchar* dst, size_t len, int cn)
 typedef void (*InRangeFunc)( const uchar* src1, size_t step1, const uchar* src2, size_t step2,
                              const uchar* src3, size_t step3, uchar* dst, size_t step, Size sz );
 
-static InRangeFunc inRangeTab[] =
+static InRangeFunc getInRangeFunc(int depth)
 {
-    (InRangeFunc)GET_OPTIMIZED(inRange8u), (InRangeFunc)GET_OPTIMIZED(inRange8s), (InRangeFunc)GET_OPTIMIZED(inRange16u),
-    (InRangeFunc)GET_OPTIMIZED(inRange16s), (InRangeFunc)GET_OPTIMIZED(inRange32s), (InRangeFunc)GET_OPTIMIZED(inRange32f),
-    (InRangeFunc)inRange64f, 0
-};
+    static InRangeFunc inRangeTab[] =
+    {
+        (InRangeFunc)GET_OPTIMIZED(inRange8u), (InRangeFunc)GET_OPTIMIZED(inRange8s), (InRangeFunc)GET_OPTIMIZED(inRange16u),
+        (InRangeFunc)GET_OPTIMIZED(inRange16s), (InRangeFunc)GET_OPTIMIZED(inRange32s), (InRangeFunc)GET_OPTIMIZED(inRange32f),
+        (InRangeFunc)inRange64f, 0
+    };
+
+    return inRangeTab[depth];
+}
 
 }
 
@@ -2601,7 +2690,7 @@ void cv::inRange(InputArray _src, InputArray _lowerb,
 
     _dst.create(src.dims, src.size, CV_8U);
     Mat dst = _dst.getMat();
-    InRangeFunc func = inRangeTab[depth];
+    InRangeFunc func = getInRangeFunc(depth);
 
     const Mat* arrays_sc[] = { &src, &dst, 0 };
     const Mat* arrays_nosc[] = { &src, &dst, &lb, &ub, 0 };
diff --git a/modules/core/src/array.cpp b/modules/core/src/array.cpp
index 9c024293eb..2ad7b1216c 100644
--- a/modules/core/src/array.cpp
+++ b/modules/core/src/array.cpp
@@ -48,6 +48,12 @@
 
 #include "precomp.hpp"
 
+#define  CV_ORIGIN_TL  0
+#define  CV_ORIGIN_BL  1
+
+/* default image row align (in bytes) */
+#define  CV_DEFAULT_IMAGE_ROW_ALIGN  4
+
 
 static struct
 {
@@ -305,7 +311,8 @@ cvCloneMatND( const CvMatND* src )
     if( src->data.ptr )
     {
         cvCreateData( dst );
-        cv::Mat _src(src), _dst(dst);
+        cv::Mat _src = cv::cvarrToMat(src);
+        cv::Mat _dst = cv::cvarrToMat(dst);
         uchar* data0 = dst->data.ptr;
         _src.copyTo(_dst);
         CV_Assert(_dst.data == data0);
@@ -1210,7 +1217,7 @@ cvGetDimSize( const CvArr* arr, int index )
 CV_IMPL CvSize
 cvGetSize( const CvArr* arr )
 {
-    CvSize size = { 0, 0 };
+    CvSize size;
 
     if( CV_IS_MAT_HDR_Z( arr ))
     {
@@ -1463,28 +1470,28 @@ cvScalarToRawData( const CvScalar* scalar, void* data, int type, int extend_to_1
         while( cn-- )
         {
             int t = cvRound( scalar->val[cn] );
-            ((uchar*)data)[cn] = CV_CAST_8U(t);
+            ((uchar*)data)[cn] = cv::saturate_cast<uchar>(t);
         }
         break;
     case CV_8SC1:
         while( cn-- )
         {
             int t = cvRound( scalar->val[cn] );
-            ((char*)data)[cn] = CV_CAST_8S(t);
+            ((char*)data)[cn] = cv::saturate_cast<schar>(t);
         }
         break;
     case CV_16UC1:
         while( cn-- )
         {
             int t = cvRound( scalar->val[cn] );
-            ((ushort*)data)[cn] = CV_CAST_16U(t);
+            ((ushort*)data)[cn] = cv::saturate_cast<ushort>(t);
         }
         break;
     case CV_16SC1:
         while( cn-- )
         {
             int t = cvRound( scalar->val[cn] );
-            ((short*)data)[cn] = CV_CAST_16S(t);
+            ((short*)data)[cn] = cv::saturate_cast<short>(t);
         }
         break;
     case CV_32SC1:
@@ -1601,19 +1608,19 @@ static void icvSetReal( double value, const void* data, int type )
         switch( type )
         {
         case CV_8U:
-            *(uchar*)data = CV_CAST_8U(ivalue);
+            *(uchar*)data = cv::saturate_cast<uchar>(ivalue);
             break;
         case CV_8S:
-            *(char*)data = CV_CAST_8S(ivalue);
+            *(schar*)data = cv::saturate_cast<schar>(ivalue);
             break;
         case CV_16U:
-            *(ushort*)data = CV_CAST_16U(ivalue);
+            *(ushort*)data = cv::saturate_cast<ushort>(ivalue);
             break;
         case CV_16S:
-            *(short*)data = CV_CAST_16S(ivalue);
+            *(short*)data = cv::saturate_cast<short>(ivalue);
             break;
         case CV_32S:
-            *(int*)data = CV_CAST_32S(ivalue);
+            *(int*)data = cv::saturate_cast<int>(ivalue);
             break;
         }
     }
@@ -1910,7 +1917,7 @@ cvPtrND( const CvArr* arr, const int* idx, int* _type,
 CV_IMPL  CvScalar
 cvGet1D( const CvArr* arr, int idx )
 {
-    CvScalar scalar = {{0,0,0,0}};
+    CvScalar scalar(0);
     int type = 0;
     uchar* ptr;
 
@@ -1945,7 +1952,7 @@ cvGet1D( const CvArr* arr, int idx )
 CV_IMPL  CvScalar
 cvGet2D( const CvArr* arr, int y, int x )
 {
-    CvScalar scalar = {{0,0,0,0}};
+    CvScalar scalar(0);
     int type = 0;
     uchar* ptr;
 
@@ -1979,7 +1986,7 @@ cvGet2D( const CvArr* arr, int y, int x )
 CV_IMPL  CvScalar
 cvGet3D( const CvArr* arr, int z, int y, int x )
 {
-    CvScalar scalar = {{0,0,0,0}};
+    CvScalar scalar(0);
     int type = 0;
     uchar* ptr;
 
@@ -2001,7 +2008,7 @@ cvGet3D( const CvArr* arr, int z, int y, int x )
 CV_IMPL  CvScalar
 cvGetND( const CvArr* arr, const int* idx )
 {
-    CvScalar scalar = {{0,0,0,0}};
+    CvScalar scalar(0);
     int type = 0;
     uchar* ptr;
 
@@ -3052,7 +3059,7 @@ cvResetImageROI( IplImage* image )
 CV_IMPL CvRect
 cvGetImageROI( const IplImage* img )
 {
-    CvRect rect = { 0, 0, 0, 0 };
+    CvRect rect;
     if( !img )
         CV_Error( CV_StsNullPtr, "Null pointer to image" );
 
@@ -3183,22 +3190,22 @@ cvCheckTermCriteria( CvTermCriteria criteria, double default_eps,
 namespace cv
 {
 
-template<> void Ptr<CvMat>::delete_obj()
+template<> void DefaultDeleter<CvMat>::operator ()(CvMat* obj) const
 { cvReleaseMat(&obj); }
 
-template<> void Ptr<IplImage>::delete_obj()
+template<> void DefaultDeleter<IplImage>::operator ()(IplImage* obj) const
 { cvReleaseImage(&obj); }
 
-template<> void Ptr<CvMatND>::delete_obj()
+template<> void DefaultDeleter<CvMatND>::operator ()(CvMatND* obj) const
 { cvReleaseMatND(&obj); }
 
-template<> void Ptr<CvSparseMat>::delete_obj()
+template<> void DefaultDeleter<CvSparseMat>::operator ()(CvSparseMat* obj) const
 { cvReleaseSparseMat(&obj); }
 
-template<> void Ptr<CvMemStorage>::delete_obj()
+template<> void DefaultDeleter<CvMemStorage>::operator ()(CvMemStorage* obj) const
 { cvReleaseMemStorage(&obj); }
 
-template<> void Ptr<CvFileStorage>::delete_obj()
+template<> void DefaultDeleter<CvFileStorage>::operator ()(CvFileStorage* obj) const
 { cvReleaseFileStorage(&obj); }
 
 }
diff --git a/modules/core/src/command_line_parser.cpp b/modules/core/src/command_line_parser.cpp
index 1bffe5af77..b082faee62 100644
--- a/modules/core/src/command_line_parser.cpp
+++ b/modules/core/src/command_line_parser.cpp
@@ -1,7 +1,5 @@
-
 #include "precomp.hpp"
-
-#include <iostream>
+#include <sstream>
 
 namespace cv
 {
@@ -9,9 +7,9 @@ namespace cv
 struct CommandLineParserParams
 {
 public:
-    std::string help_message;
-    std::string def_value;
-    std::vector<std::string> keys;
+    String help_message;
+    String def_value;
+    std::vector<String> keys;
     int number;
 };
 
@@ -19,27 +17,27 @@ public:
 struct CommandLineParser::Impl
 {
     bool error;
-    std::string error_message;
-    std::string about_message;
+    String error_message;
+    String about_message;
 
-    std::string path_to_app;
-    std::string app_name;
+    String path_to_app;
+    String app_name;
 
     std::vector<CommandLineParserParams> data;
 
-    std::vector<std::string> split_range_string(const std::string& str, char fs, char ss) const;
-    std::vector<std::string> split_string(const std::string& str, char symbol = ' ', bool create_empty_item = false) const;
-    std::string cat_string(const std::string& str) const;
+    std::vector<String> split_range_string(const String& str, char fs, char ss) const;
+    std::vector<String> split_string(const String& str, char symbol = ' ', bool create_empty_item = false) const;
+    String cat_string(const String& str) const;
 
-    void apply_params(const std::string& key, const std::string& value);
-    void apply_params(int i, std::string value);
+    void apply_params(const String& key, const String& value);
+    void apply_params(int i, String value);
 
     void sort_params();
     int refcount;
 };
 
 
-static std::string get_type_name(int type)
+static String get_type_name(int type)
 {
     if( type == Param::INT )
         return "int";
@@ -56,9 +54,9 @@ static std::string get_type_name(int type)
     return "unknown";
 }
 
-static void from_str(const std::string& str, int type, void* dst)
+static void from_str(const String& str, int type, void* dst)
 {
-    std::stringstream ss(str);
+    std::stringstream ss(str.c_str());
     if( type == Param::INT )
         ss >> *(int*)dst;
     else if( type == Param::UNSIGNED_INT )
@@ -70,20 +68,20 @@ static void from_str(const std::string& str, int type, void* dst)
     else if( type == Param::REAL )
         ss >> *(double*)dst;
     else if( type == Param::STRING )
-        *(std::string*)dst = str;
+        *(String*)dst = str;
     else
         throw cv::Exception(CV_StsBadArg, "unknown/unsupported parameter type", "", __FILE__, __LINE__);
 
     if (ss.fail())
     {
-        std::string err_msg = "can not convert: [" + str +
+        String err_msg = "can not convert: [" + str +
         + "] to [" + get_type_name(type) + "]";
 
         throw cv::Exception(CV_StsBadArg, err_msg, "", __FILE__, __LINE__);
     }
 }
 
-void CommandLineParser::getByName(const std::string& name, bool space_delete, int type, void* dst) const
+void CommandLineParser::getByName(const String& name, bool space_delete, int type, void* dst) const
 {
     try
     {
@@ -93,7 +91,7 @@ void CommandLineParser::getByName(const std::string& name, bool space_delete, in
             {
                 if (name.compare(impl->data[i].keys[j]) == 0)
                 {
-                    std::string v = impl->data[i].def_value;
+                    String v = impl->data[i].def_value;
                     if (space_delete)
                         v = impl->cat_string(v);
                     from_str(v, type, dst);
@@ -102,12 +100,12 @@ void CommandLineParser::getByName(const std::string& name, bool space_delete, in
             }
         }
         impl->error = true;
-        impl->error_message += "Unknown parametes " + name + "\n";
+        impl->error_message = impl->error_message + "Unknown parametes " + name + "\n";
     }
     catch (std::exception& e)
     {
         impl->error = true;
-        impl->error_message += "Exception: " + std::string(e.what()) + "\n";
+        impl->error_message = impl->error_message + "Exception: " + String(e.what()) + "\n";
     }
 }
 
@@ -120,66 +118,61 @@ void CommandLineParser::getByIndex(int index, bool space_delete, int type, void*
         {
             if (impl->data[i].number == index)
             {
-                std::string v = impl->data[i].def_value;
+                String v = impl->data[i].def_value;
                 if (space_delete == true) v = impl->cat_string(v);
                 from_str(v, type, dst);
                 return;
             }
         }
         impl->error = true;
-        impl->error_message += "Unknown parametes #" + format("%d", index) + "\n";
+        impl->error_message = impl->error_message + "Unknown parametes #" + format("%d", index) + "\n";
     }
     catch(std::exception & e)
     {
         impl->error = true;
-        impl->error_message += "Exception: " + std::string(e.what()) + "\n";
+        impl->error_message = impl->error_message + "Exception: " + String(e.what()) + "\n";
     }
 }
 
 static bool cmp_params(const CommandLineParserParams & p1, const CommandLineParserParams & p2)
 {
+    if (p1.number < p2.number)
+        return true;
+
     if (p1.number > p2.number)
         return false;
 
-    if (p1.number == -1 && p2.number == -1)
-    {
-        if (p1.keys[0].compare(p2.keys[0]) > 0)
-        {
-            return false;
-        }
-    }
-
-    return true;
+    return p1.keys[0].compare(p2.keys[0]) < 0;
 }
 
-CommandLineParser::CommandLineParser(int argc, const char* const argv[], const std::string& keys)
+CommandLineParser::CommandLineParser(int argc, const char* const argv[], const String& keys)
 {
     impl = new Impl;
     impl->refcount = 1;
 
     // path to application
-    size_t pos_s = std::string(argv[0]).find_last_of("/\\");
-    if (pos_s == std::string::npos)
+    size_t pos_s = String(argv[0]).find_last_of("/\\");
+    if (pos_s == String::npos)
     {
         impl->path_to_app = "";
-        impl->app_name = std::string(argv[0]);
+        impl->app_name = String(argv[0]);
     }
     else
     {
-        impl->path_to_app = std::string(argv[0]).substr(0, pos_s);
-        impl->app_name = std::string(argv[0]).substr(pos_s + 1, std::string(argv[0]).length() - pos_s);
+        impl->path_to_app = String(argv[0]).substr(0, pos_s);
+        impl->app_name = String(argv[0]).substr(pos_s + 1, String(argv[0]).length() - pos_s);
     }
 
     impl->error = false;
     impl->error_message = "";
 
     // parse keys
-    std::vector<std::string> k = impl->split_range_string(keys, '{', '}');
+    std::vector<String> k = impl->split_range_string(keys, '{', '}');
 
     int jj = 0;
     for (size_t i = 0; i < k.size(); i++)
     {
-        std::vector<std::string> l = impl->split_string(k[i], '|', true);
+        std::vector<String> l = impl->split_string(k[i], '|', true);
         CommandLineParserParams p;
         p.keys = impl->split_string(l[0]);
         p.def_value = l[1];
@@ -206,11 +199,11 @@ CommandLineParser::CommandLineParser(int argc, const char* const argv[], const s
     jj = 0;
     for (int i = 1; i < argc; i++)
     {
-        std::string s = std::string(argv[i]);
+        String s = String(argv[i]);
 
-        if (s.find('=') != std::string::npos && s.find('=') < s.length())
+        if (s.find('=') != String::npos && s.find('=') < s.length())
         {
-            std::vector<std::string> k_v = impl->split_string(s, '=', true);
+            std::vector<String> k_v = impl->split_string(s, '=', true);
             for (int h = 0; h < 2; h++)
             {
                 if (k_v[0][0] == '-')
@@ -256,12 +249,12 @@ CommandLineParser& CommandLineParser::operator = (const CommandLineParser& parse
     return *this;
 }
 
-void CommandLineParser::about(const std::string& message)
+void CommandLineParser::about(const String& message)
 {
     impl->about_message = message;
 }
 
-void CommandLineParser::Impl::apply_params(const std::string& key, const std::string& value)
+void CommandLineParser::Impl::apply_params(const String& key, const String& value)
 {
     for (size_t i = 0; i < data.size(); i++)
     {
@@ -276,7 +269,7 @@ void CommandLineParser::Impl::apply_params(const std::string& key, const std::st
     }
 }
 
-void CommandLineParser::Impl::apply_params(int i, std::string value)
+void CommandLineParser::Impl::apply_params(int i, String value)
 {
     for (size_t j = 0; j < data.size(); j++)
     {
@@ -292,34 +285,34 @@ void CommandLineParser::Impl::sort_params()
 {
     for (size_t i = 0; i < data.size(); i++)
     {
-        sort(data[i].keys.begin(), data[i].keys.end());
+        std::sort(data[i].keys.begin(), data[i].keys.end());
     }
 
     std::sort (data.begin(), data.end(), cmp_params);
 }
 
-std::string CommandLineParser::Impl::cat_string(const std::string& str) const
+String CommandLineParser::Impl::cat_string(const String& str) const
 {
     int left = 0, right = (int)str.length();
     while( left <= right && str[left] == ' ' )
         left++;
     while( right > left && str[right-1] == ' ' )
         right--;
-    return left >= right ? std::string("") : str.substr(left, right-left);
+    return left >= right ? String("") : str.substr(left, right-left);
 }
 
-std::string CommandLineParser::getPathToApplication() const
+String CommandLineParser::getPathToApplication() const
 {
     return impl->path_to_app;
 }
 
-bool CommandLineParser::has(const std::string& name) const
+bool CommandLineParser::has(const String& name) const
 {
     for (size_t i = 0; i < impl->data.size(); i++)
     {
         for (size_t j = 0; j < impl->data[i].keys.size(); j++)
         {
-            if (name.compare(impl->data[i].keys[j]) == 0 && std::string("true").compare(impl->data[i].def_value) == 0)
+            if (name.compare(impl->data[i].keys[j]) == 0 && String("true").compare(impl->data[i].def_value) == 0)
             {
                 return true;
             }
@@ -337,86 +330,87 @@ void CommandLineParser::printErrors() const
 {
     if (impl->error)
     {
-        std::cout << std::endl << "ERRORS:" << std::endl << impl->error_message << std::endl;
+        printf("\nERRORS:\n%s\n", impl->error_message.c_str());
+        fflush(stdout);
     }
 }
 
 void CommandLineParser::printMessage() const
 {
     if (impl->about_message != "")
-        std::cout << impl->about_message << std::endl;
+        printf("%s\n", impl->about_message.c_str());
 
-    std::cout << "Usage: " << impl->app_name << " [params] ";
+    printf("Usage: %s [params] ", impl->app_name.c_str());
 
     for (size_t i = 0; i < impl->data.size(); i++)
     {
         if (impl->data[i].number > -1)
         {
-            std::string name = impl->data[i].keys[0].substr(1, impl->data[i].keys[0].length() - 1);
-            std::cout << name << " ";
+            String name = impl->data[i].keys[0].substr(1, impl->data[i].keys[0].length() - 1);
+            printf("%s ", name.c_str());
         }
     }
 
-    std::cout << std::endl << std::endl;
+    printf("\n\n");
 
     for (size_t i = 0; i < impl->data.size(); i++)
     {
         if (impl->data[i].number == -1)
         {
-            std::cout << "\t";
+            printf("\t");
             for (size_t j = 0; j < impl->data[i].keys.size(); j++)
             {
-                std::string k = impl->data[i].keys[j];
+                String k = impl->data[i].keys[j];
                 if (k.length() > 1)
                 {
-                    std::cout << "--";
+                    printf("--");
                 }
                 else
                 {
-                    std::cout << "-";
+                    printf("-");
                 }
-                std::cout << k;
+                printf("%s", k.c_str());
 
                 if (j != impl->data[i].keys.size() - 1)
                 {
-                    std::cout << ", ";
+                    printf(", ");
                 }
             }
-            std::string dv = impl->cat_string(impl->data[i].def_value);
+            String dv = impl->cat_string(impl->data[i].def_value);
             if (dv.compare("") != 0)
             {
-                std::cout << " (value:" << dv << ")";
+                printf(" (value:%s)", dv.c_str());
             }
-            std::cout << std::endl << "\t\t" << impl->data[i].help_message << std::endl;
+            printf("\n\t\t%s\n", impl->data[i].help_message.c_str());
         }
     }
-    std::cout << std::endl;
+    printf("\n");
 
     for (size_t i = 0; i < impl->data.size(); i++)
     {
         if (impl->data[i].number != -1)
         {
-            std::cout << "\t";
-            std::string k = impl->data[i].keys[0];
+            printf("\t");
+            String k = impl->data[i].keys[0];
             k = k.substr(1, k.length() - 1);
 
-            std::cout << k;
+            printf("%s", k.c_str());
 
-            std::string dv = impl->cat_string(impl->data[i].def_value);
+            String dv = impl->cat_string(impl->data[i].def_value);
             if (dv.compare("") != 0)
             {
-                std::cout << " (value:" << dv << ")";
+                printf(" (value:%s)", dv.c_str());
             }
-            std::cout << std::endl << "\t\t" << impl->data[i].help_message << std::endl;
+            printf("\n\t\t%s\n", impl->data[i].help_message.c_str());
         }
     }
 }
 
-std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::string& _str, char fs, char ss) const
+std::vector<String> CommandLineParser::Impl::split_range_string(const String& _str, char fs, char ss) const
 {
-    std::string str = _str;
-    std::vector<std::string> vec;
-    std::string word = "";
+    String str = _str;
+    std::vector<String> vec;
+    String word = "";
     bool begin = false;
 
     while (!str.empty())
@@ -426,13 +420,13 @@ std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::
             if (begin == true)
             {
                 throw cv::Exception(CV_StsParseError,
-                         std::string("error in split_range_string(")
+                         String("error in split_range_string(")
                          + str
-                         + std::string(", ")
-                         + std::string(1, fs)
-                         + std::string(", ")
-                         + std::string(1, ss)
-                         + std::string(")"),
+                         + String(", ")
+                         + String(1, fs)
+                         + String(", ")
+                         + String(1, ss)
+                         + String(")"),
                          "", __FILE__, __LINE__
                          );
             }
@@ -446,13 +440,13 @@ std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::
             if (begin == false)
             {
                 throw cv::Exception(CV_StsParseError,
-                         std::string("error in split_range_string(")
+                         String("error in split_range_string(")
                          + str
-                         + std::string(", ")
-                         + std::string(1, fs)
-                         + std::string(", ")
-                         + std::string(1, ss)
-                         + std::string(")"),
+                         + String(", ")
+                         + String(1, fs)
+                         + String(", ")
+                         + String(1, ss)
+                         + String(")"),
                          "", __FILE__, __LINE__
                          );
             }
@@ -462,7 +456,7 @@ std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::
 
         if (begin == true)
         {
-            word += str[0];
+            word = word + str[0];
         }
         str = str.substr(1, str.length() - 1);
     }
@@ -470,13 +464,13 @@ std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::
     if (begin == true)
     {
         throw cv::Exception(CV_StsParseError,
-                 std::string("error in split_range_string(")
+                 String("error in split_range_string(")
                  + str
-                 + std::string(", ")
-                 + std::string(1, fs)
-                 + std::string(", ")
-                 + std::string(1, ss)
-                 + std::string(")"),
+                 + String(", ")
+                 + String(1, fs)
+                 + String(", ")
+                 + String(1, ss)
+                 + String(")"),
                  "", __FILE__, __LINE__
                 );
     }
@@ -484,11 +478,11 @@ std::vector<std::string> CommandLineParser::Impl::split_range_string(const std::
     return vec;
 }
 
-std::vector<std::string> CommandLineParser::Impl::split_string(const std::string& _str, char symbol, bool create_empty_item) const
+std::vector<String> CommandLineParser::Impl::split_string(const String& _str, char symbol, bool create_empty_item) const
 {
-    std::string str = _str;
-    std::vector<std::string> vec;
-    std::string word = "";
+    String str = _str;
+    std::vector<String> vec;
+    String word = "";
 
     while (!str.empty())
     {
@@ -502,7 +496,7 @@ std::vector<std::string> CommandLineParser::Impl::split_string(const std::string
         }
         else
         {
-            word += str[0];
+            word = word + str[0];
         }
         str = str.substr(1, str.length() - 1);
     }
diff --git a/modules/core/src/convert.cpp b/modules/core/src/convert.cpp
index d8ef88151a..5cd6894600 100644
--- a/modules/core/src/convert.cpp
+++ b/modules/core/src/convert.cpp
@@ -194,17 +194,27 @@ static void merge64s(const int64** src, int64* dst, int len, int cn )
 typedef void (*SplitFunc)(const uchar* src, uchar** dst, int len, int cn);
 typedef void (*MergeFunc)(const uchar** src, uchar* dst, int len, int cn);
 
-static SplitFunc splitTab[] =
+static SplitFunc getSplitFunc(int depth)
 {
-    (SplitFunc)GET_OPTIMIZED(split8u), (SplitFunc)GET_OPTIMIZED(split8u), (SplitFunc)GET_OPTIMIZED(split16u), (SplitFunc)GET_OPTIMIZED(split16u),
-    (SplitFunc)GET_OPTIMIZED(split32s), (SplitFunc)GET_OPTIMIZED(split32s), (SplitFunc)GET_OPTIMIZED(split64s), 0
-};
+    static SplitFunc splitTab[] =
+    {
+        (SplitFunc)GET_OPTIMIZED(split8u), (SplitFunc)GET_OPTIMIZED(split8u), (SplitFunc)GET_OPTIMIZED(split16u), (SplitFunc)GET_OPTIMIZED(split16u),
+        (SplitFunc)GET_OPTIMIZED(split32s), (SplitFunc)GET_OPTIMIZED(split32s), (SplitFunc)GET_OPTIMIZED(split64s), 0
+    };
 
-static MergeFunc mergeTab[] =
+    return splitTab[depth];
+}
+
+static MergeFunc getMergeFunc(int depth)
 {
-    (MergeFunc)GET_OPTIMIZED(merge8u), (MergeFunc)GET_OPTIMIZED(merge8u), (MergeFunc)GET_OPTIMIZED(merge16u), (MergeFunc)GET_OPTIMIZED(merge16u),
-    (MergeFunc)GET_OPTIMIZED(merge32s), (MergeFunc)GET_OPTIMIZED(merge32s), (MergeFunc)GET_OPTIMIZED(merge64s), 0
-};
+    static MergeFunc mergeTab[] =
+    {
+        (MergeFunc)GET_OPTIMIZED(merge8u), (MergeFunc)GET_OPTIMIZED(merge8u), (MergeFunc)GET_OPTIMIZED(merge16u), (MergeFunc)GET_OPTIMIZED(merge16u),
+        (MergeFunc)GET_OPTIMIZED(merge32s), (MergeFunc)GET_OPTIMIZED(merge32s), (MergeFunc)GET_OPTIMIZED(merge64s), 0
+    };
+
+    return mergeTab[depth];
+}
 
 }
 
@@ -217,7 +227,7 @@ void cv::split(const Mat& src, Mat* mv)
         return;
     }
 
-    SplitFunc func = splitTab[depth];
+    SplitFunc func = getSplitFunc(depth);
     CV_Assert( func != 0 );
 
     int esz = (int)src.elemSize(), esz1 = (int)src.elemSize1();
@@ -323,7 +333,7 @@ void cv::merge(const Mat* mv, size_t n, OutputArray _dst)
 
     NAryMatIterator it(arrays, ptrs, cn+1);
     int total = (int)it.size, blocksize = cn <= 4 ? total : std::min(total, blocksize0);
-    MergeFunc func = mergeTab[depth];
+    MergeFunc func = getMergeFunc(depth);
 
     for( i = 0; i < it.nplanes; i++, ++it )
     {
@@ -419,12 +429,17 @@ static void mixChannels64s( const int64** src, const int* sdelta,
 typedef void (*MixChannelsFunc)( const uchar** src, const int* sdelta,
         uchar** dst, const int* ddelta, int len, int npairs );
 
-static MixChannelsFunc mixchTab[] =
+static MixChannelsFunc getMixchFunc(int depth)
 {
-    (MixChannelsFunc)mixChannels8u, (MixChannelsFunc)mixChannels8u, (MixChannelsFunc)mixChannels16u,
-    (MixChannelsFunc)mixChannels16u, (MixChannelsFunc)mixChannels32s, (MixChannelsFunc)mixChannels32s,
-    (MixChannelsFunc)mixChannels64s, 0
-};
+    static MixChannelsFunc mixchTab[] =
+    {
+        (MixChannelsFunc)mixChannels8u, (MixChannelsFunc)mixChannels8u, (MixChannelsFunc)mixChannels16u,
+        (MixChannelsFunc)mixChannels16u, (MixChannelsFunc)mixChannels32s, (MixChannelsFunc)mixChannels32s,
+        (MixChannelsFunc)mixChannels64s, 0
+    };
+
+    return mixchTab[depth];
+}
 
 }
 
@@ -479,7 +494,7 @@ void cv::mixChannels( const Mat* src, size_t nsrcs, Mat* dst, size_t ndsts, cons
 
     NAryMatIterator it(arrays, ptrs, (int)(nsrcs + ndsts));
     int total = (int)it.size, blocksize = std::min(total, (int)((BLOCK_SIZE + esz1-1)/esz1));
-    MixChannelsFunc func = mixchTab[depth];
+    MixChannelsFunc func = getMixchFunc(depth);
 
     for( i = 0; i < it.nplanes; i++, ++it )
     {
@@ -505,14 +520,30 @@ void cv::mixChannels( const Mat* src, size_t nsrcs, Mat* dst, size_t ndsts, cons
 }
 
 
-void cv::mixChannels(const std::vector<Mat>& src, std::vector<Mat>& dst,
+void cv::mixChannels(InputArrayOfArrays src, InputOutputArrayOfArrays dst,
                  const int* fromTo, size_t npairs)
 {
-    mixChannels(!src.empty() ? &src[0] : 0, src.size(),
-                !dst.empty() ? &dst[0] : 0, dst.size(), fromTo, npairs);
+    if(npairs == 0)
+        return;
+    bool src_is_mat = src.kind() != _InputArray::STD_VECTOR_MAT &&
+                      src.kind() != _InputArray::STD_VECTOR_VECTOR;
+    bool dst_is_mat = dst.kind() != _InputArray::STD_VECTOR_MAT &&
+                      dst.kind() != _InputArray::STD_VECTOR_VECTOR;
+    int i;
+    int nsrc = src_is_mat ? 1 : (int)src.total();
+    int ndst = dst_is_mat ? 1 : (int)dst.total();
+
+    CV_Assert(nsrc > 0 && ndst > 0);
+    cv::AutoBuffer<Mat> _buf(nsrc + ndst);
+    Mat* buf = _buf;
+    for( i = 0; i < nsrc; i++ )
+        buf[i] = src.getMat(src_is_mat ? -1 : i);
+    for( i = 0; i < ndst; i++ )
+        buf[nsrc + i] = dst.getMat(dst_is_mat ? -1 : i);
+    mixChannels(&buf[0], nsrc, &buf[nsrc], ndst, fromTo, npairs);
 }
 
-void cv::mixChannels(InputArrayOfArrays src, InputArrayOfArrays dst,
+void cv::mixChannels(InputArrayOfArrays src, InputOutputArrayOfArrays dst,
                      const std::vector<int>& fromTo)
 {
     if(fromTo.empty())
@@ -931,104 +962,109 @@ DEF_CVT_FUNC(32s64f, int, double);
 DEF_CVT_FUNC(32f64f, float, double);
 DEF_CPY_FUNC(64s,    int64);
 
-static BinaryFunc cvtScaleAbsTab[] =
+static BinaryFunc getCvtScaleAbsFunc(int depth)
 {
-    (BinaryFunc)cvtScaleAbs8u, (BinaryFunc)cvtScaleAbs8s8u, (BinaryFunc)cvtScaleAbs16u8u,
-    (BinaryFunc)cvtScaleAbs16s8u, (BinaryFunc)cvtScaleAbs32s8u, (BinaryFunc)cvtScaleAbs32f8u,
-    (BinaryFunc)cvtScaleAbs64f8u, 0
-};
+    static BinaryFunc cvtScaleAbsTab[] =
+    {
+        (BinaryFunc)cvtScaleAbs8u, (BinaryFunc)cvtScaleAbs8s8u, (BinaryFunc)cvtScaleAbs16u8u,
+        (BinaryFunc)cvtScaleAbs16s8u, (BinaryFunc)cvtScaleAbs32s8u, (BinaryFunc)cvtScaleAbs32f8u,
+        (BinaryFunc)cvtScaleAbs64f8u, 0
+    };
 
-static BinaryFunc cvtScaleTab[][8] =
-{
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u), (BinaryFunc)GET_OPTIMIZED(cvtScale8s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale16u8u),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale32s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale32f8u),
-        (BinaryFunc)cvtScale64f8u, 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u8s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u8s),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s8s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s8s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f8s),
-        (BinaryFunc)cvtScale64f8s, 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u16u), (BinaryFunc)GET_OPTIMIZED(cvtScale8s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale16u),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale32s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale32f16u),
-        (BinaryFunc)cvtScale64f16u, 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u16s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s16s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u16s),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s16s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f16s),
-        (BinaryFunc)cvtScale64f16s, 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u32s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s32s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u32s),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s32s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f32s),
-        (BinaryFunc)cvtScale64f32s, 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvtScale8u32f), (BinaryFunc)GET_OPTIMIZED(cvtScale8s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale16u32f),
-        (BinaryFunc)GET_OPTIMIZED(cvtScale16s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale32s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale32f),
-        (BinaryFunc)cvtScale64f32f, 0
-    },
-    {
-        (BinaryFunc)cvtScale8u64f, (BinaryFunc)cvtScale8s64f, (BinaryFunc)cvtScale16u64f,
-        (BinaryFunc)cvtScale16s64f, (BinaryFunc)cvtScale32s64f, (BinaryFunc)cvtScale32f64f,
-        (BinaryFunc)cvtScale64f, 0
-    },
-    {
-        0, 0, 0, 0, 0, 0, 0, 0
-    }
-};
-
-static BinaryFunc cvtTab[][8] =
-{
-    {
-        (BinaryFunc)(cvt8u), (BinaryFunc)GET_OPTIMIZED(cvt8s8u), (BinaryFunc)GET_OPTIMIZED(cvt16u8u),
-        (BinaryFunc)GET_OPTIMIZED(cvt16s8u), (BinaryFunc)GET_OPTIMIZED(cvt32s8u), (BinaryFunc)GET_OPTIMIZED(cvt32f8u),
-        (BinaryFunc)GET_OPTIMIZED(cvt64f8u), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u8s), (BinaryFunc)cvt8u, (BinaryFunc)GET_OPTIMIZED(cvt16u8s),
-        (BinaryFunc)GET_OPTIMIZED(cvt16s8s), (BinaryFunc)GET_OPTIMIZED(cvt32s8s), (BinaryFunc)GET_OPTIMIZED(cvt32f8s),
-        (BinaryFunc)GET_OPTIMIZED(cvt64f8s), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u16u), (BinaryFunc)GET_OPTIMIZED(cvt8s16u), (BinaryFunc)cvt16u,
-        (BinaryFunc)GET_OPTIMIZED(cvt16s16u), (BinaryFunc)GET_OPTIMIZED(cvt32s16u), (BinaryFunc)GET_OPTIMIZED(cvt32f16u),
-        (BinaryFunc)GET_OPTIMIZED(cvt64f16u), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u16s), (BinaryFunc)GET_OPTIMIZED(cvt8s16s), (BinaryFunc)GET_OPTIMIZED(cvt16u16s),
-        (BinaryFunc)cvt16u, (BinaryFunc)GET_OPTIMIZED(cvt32s16s), (BinaryFunc)GET_OPTIMIZED(cvt32f16s),
-        (BinaryFunc)GET_OPTIMIZED(cvt64f16s), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u32s), (BinaryFunc)GET_OPTIMIZED(cvt8s32s), (BinaryFunc)GET_OPTIMIZED(cvt16u32s),
-        (BinaryFunc)GET_OPTIMIZED(cvt16s32s), (BinaryFunc)cvt32s, (BinaryFunc)GET_OPTIMIZED(cvt32f32s),
-        (BinaryFunc)GET_OPTIMIZED(cvt64f32s), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u32f), (BinaryFunc)GET_OPTIMIZED(cvt8s32f), (BinaryFunc)GET_OPTIMIZED(cvt16u32f),
-        (BinaryFunc)GET_OPTIMIZED(cvt16s32f), (BinaryFunc)GET_OPTIMIZED(cvt32s32f), (BinaryFunc)cvt32s,
-        (BinaryFunc)GET_OPTIMIZED(cvt64f32f), 0
-    },
-    {
-        (BinaryFunc)GET_OPTIMIZED(cvt8u64f), (BinaryFunc)GET_OPTIMIZED(cvt8s64f), (BinaryFunc)GET_OPTIMIZED(cvt16u64f),
-        (BinaryFunc)GET_OPTIMIZED(cvt16s64f), (BinaryFunc)GET_OPTIMIZED(cvt32s64f), (BinaryFunc)GET_OPTIMIZED(cvt32f64f),
-        (BinaryFunc)(cvt64s), 0
-    },
-    {
-        0, 0, 0, 0, 0, 0, 0, 0
-    }
-};
+    return cvtScaleAbsTab[depth];
+}
 
 BinaryFunc getConvertFunc(int sdepth, int ddepth)
 {
+    static BinaryFunc cvtTab[][8] =
+    {
+        {
+            (BinaryFunc)(cvt8u), (BinaryFunc)GET_OPTIMIZED(cvt8s8u), (BinaryFunc)GET_OPTIMIZED(cvt16u8u),
+            (BinaryFunc)GET_OPTIMIZED(cvt16s8u), (BinaryFunc)GET_OPTIMIZED(cvt32s8u), (BinaryFunc)GET_OPTIMIZED(cvt32f8u),
+            (BinaryFunc)GET_OPTIMIZED(cvt64f8u), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u8s), (BinaryFunc)cvt8u, (BinaryFunc)GET_OPTIMIZED(cvt16u8s),
+            (BinaryFunc)GET_OPTIMIZED(cvt16s8s), (BinaryFunc)GET_OPTIMIZED(cvt32s8s), (BinaryFunc)GET_OPTIMIZED(cvt32f8s),
+            (BinaryFunc)GET_OPTIMIZED(cvt64f8s), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u16u), (BinaryFunc)GET_OPTIMIZED(cvt8s16u), (BinaryFunc)cvt16u,
+            (BinaryFunc)GET_OPTIMIZED(cvt16s16u), (BinaryFunc)GET_OPTIMIZED(cvt32s16u), (BinaryFunc)GET_OPTIMIZED(cvt32f16u),
+            (BinaryFunc)GET_OPTIMIZED(cvt64f16u), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u16s), (BinaryFunc)GET_OPTIMIZED(cvt8s16s), (BinaryFunc)GET_OPTIMIZED(cvt16u16s),
+            (BinaryFunc)cvt16u, (BinaryFunc)GET_OPTIMIZED(cvt32s16s), (BinaryFunc)GET_OPTIMIZED(cvt32f16s),
+            (BinaryFunc)GET_OPTIMIZED(cvt64f16s), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u32s), (BinaryFunc)GET_OPTIMIZED(cvt8s32s), (BinaryFunc)GET_OPTIMIZED(cvt16u32s),
+            (BinaryFunc)GET_OPTIMIZED(cvt16s32s), (BinaryFunc)cvt32s, (BinaryFunc)GET_OPTIMIZED(cvt32f32s),
+            (BinaryFunc)GET_OPTIMIZED(cvt64f32s), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u32f), (BinaryFunc)GET_OPTIMIZED(cvt8s32f), (BinaryFunc)GET_OPTIMIZED(cvt16u32f),
+            (BinaryFunc)GET_OPTIMIZED(cvt16s32f), (BinaryFunc)GET_OPTIMIZED(cvt32s32f), (BinaryFunc)cvt32s,
+            (BinaryFunc)GET_OPTIMIZED(cvt64f32f), 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvt8u64f), (BinaryFunc)GET_OPTIMIZED(cvt8s64f), (BinaryFunc)GET_OPTIMIZED(cvt16u64f),
+            (BinaryFunc)GET_OPTIMIZED(cvt16s64f), (BinaryFunc)GET_OPTIMIZED(cvt32s64f), (BinaryFunc)GET_OPTIMIZED(cvt32f64f),
+            (BinaryFunc)(cvt64s), 0
+        },
+        {
+            0, 0, 0, 0, 0, 0, 0, 0
+        }
+    };
+
     return cvtTab[CV_MAT_DEPTH(ddepth)][CV_MAT_DEPTH(sdepth)];
 }
 
-BinaryFunc getConvertScaleFunc(int sdepth, int ddepth)
+static BinaryFunc getConvertScaleFunc(int sdepth, int ddepth)
 {
+    static BinaryFunc cvtScaleTab[][8] =
+    {
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u), (BinaryFunc)GET_OPTIMIZED(cvtScale8s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale16u8u),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale32s8u), (BinaryFunc)GET_OPTIMIZED(cvtScale32f8u),
+            (BinaryFunc)cvtScale64f8u, 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u8s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u8s),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s8s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s8s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f8s),
+            (BinaryFunc)cvtScale64f8s, 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u16u), (BinaryFunc)GET_OPTIMIZED(cvtScale8s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale16u),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale32s16u), (BinaryFunc)GET_OPTIMIZED(cvtScale32f16u),
+            (BinaryFunc)cvtScale64f16u, 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u16s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s16s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u16s),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s16s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f16s),
+            (BinaryFunc)cvtScale64f16s, 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u32s), (BinaryFunc)GET_OPTIMIZED(cvtScale8s32s), (BinaryFunc)GET_OPTIMIZED(cvtScale16u32s),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s32s), (BinaryFunc)GET_OPTIMIZED(cvtScale32s), (BinaryFunc)GET_OPTIMIZED(cvtScale32f32s),
+            (BinaryFunc)cvtScale64f32s, 0
+        },
+        {
+            (BinaryFunc)GET_OPTIMIZED(cvtScale8u32f), (BinaryFunc)GET_OPTIMIZED(cvtScale8s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale16u32f),
+            (BinaryFunc)GET_OPTIMIZED(cvtScale16s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale32s32f), (BinaryFunc)GET_OPTIMIZED(cvtScale32f),
+            (BinaryFunc)cvtScale64f32f, 0
+        },
+        {
+            (BinaryFunc)cvtScale8u64f, (BinaryFunc)cvtScale8s64f, (BinaryFunc)cvtScale16u64f,
+            (BinaryFunc)cvtScale16s64f, (BinaryFunc)cvtScale32s64f, (BinaryFunc)cvtScale32f64f,
+            (BinaryFunc)cvtScale64f, 0
+        },
+        {
+            0, 0, 0, 0, 0, 0, 0, 0
+        }
+    };
+
     return cvtScaleTab[CV_MAT_DEPTH(ddepth)][CV_MAT_DEPTH(sdepth)];
 }
 
@@ -1041,7 +1077,7 @@ void cv::convertScaleAbs( InputArray _src, OutputArray _dst, double alpha, doubl
     double scale[] = {alpha, beta};
     _dst.create( src.dims, src.size, CV_8UC(cn) );
     Mat dst = _dst.getMat();
-    BinaryFunc func = cvtScaleAbsTab[src.depth()];
+    BinaryFunc func = getCvtScaleAbsFunc(src.depth());
     CV_Assert( func != 0 );
 
     if( src.dims <= 2 )
@@ -1173,10 +1209,9 @@ static LUTFunc lutTab[] =
 
 }
 
-void cv::LUT( InputArray _src, InputArray _lut, OutputArray _dst, int interpolation )
+void cv::LUT( InputArray _src, InputArray _lut, OutputArray _dst )
 {
     Mat src = _src.getMat(), lut = _lut.getMat();
-    CV_Assert( interpolation == 0 );
     int cn = src.channels();
     int lutcn = lut.channels();
 
@@ -1361,4 +1396,4 @@ CV_IMPL void cvNormalize( const CvArr* srcarr, CvArr* dstarr,
     cv::normalize( src, dst, a, b, norm_type, dst.type(), mask );
 }
 
-/* End of file. */
\ No newline at end of file
+/* End of file. */
diff --git a/modules/core/src/copy.cpp b/modules/core/src/copy.cpp
index 8276ffddb6..3d58df7871 100644
--- a/modules/core/src/copy.cpp
+++ b/modules/core/src/copy.cpp
@@ -512,6 +512,231 @@ Mat repeat(const Mat& src, int ny, int nx)
     return dst;
 }
 
+
+} // cv
+
+
+/*
+ Various border types, image boundaries are denoted with '|'
+
+ * BORDER_REPLICATE:     aaaaaa|abcdefgh|hhhhhhh
+ * BORDER_REFLECT:       fedcba|abcdefgh|hgfedcb
+ * BORDER_REFLECT_101:   gfedcb|abcdefgh|gfedcba
+ * BORDER_WRAP:          cdefgh|abcdefgh|abcdefg
+ * BORDER_CONSTANT:      iiiiii|abcdefgh|iiiiiii  with some specified 'i'
+ */
+int cv::borderInterpolate( int p, int len, int borderType )
+{
+    if( (unsigned)p < (unsigned)len )
+        ;
+    else if( borderType == BORDER_REPLICATE )
+        p = p < 0 ? 0 : len - 1;
+    else if( borderType == BORDER_REFLECT || borderType == BORDER_REFLECT_101 )
+    {
+        int delta = borderType == BORDER_REFLECT_101;
+        if( len == 1 )
+            return 0;
+        do
+        {
+            if( p < 0 )
+                p = -p - 1 + delta;
+            else
+                p = len - 1 - (p - len) - delta;
+        }
+        while( (unsigned)p >= (unsigned)len );
+    }
+    else if( borderType == BORDER_WRAP )
+    {
+        if( p < 0 )
+            p -= ((p-len+1)/len)*len;
+        if( p >= len )
+            p %= len;
+    }
+    else if( borderType == BORDER_CONSTANT )
+        p = -1;
+    else
+        CV_Error( CV_StsBadArg, "Unknown/unsupported border type" );
+    return p;
+}
+
+namespace
+{
+
+void copyMakeBorder_8u( const uchar* src, size_t srcstep, cv::Size srcroi,
+                        uchar* dst, size_t dststep, cv::Size dstroi,
+                        int top, int left, int cn, int borderType )
+{
+    const int isz = (int)sizeof(int);
+    int i, j, k, elemSize = 1;
+    bool intMode = false;
+
+    if( (cn | srcstep | dststep | (size_t)src | (size_t)dst) % isz == 0 )
+    {
+        cn /= isz;
+        elemSize = isz;
+        intMode = true;
+    }
+
+    cv::AutoBuffer<int> _tab((dstroi.width - srcroi.width)*cn);
+    int* tab = _tab;
+    int right = dstroi.width - srcroi.width - left;
+    int bottom = dstroi.height - srcroi.height - top;
+
+    for( i = 0; i < left; i++ )
+    {
+        j = cv::borderInterpolate(i - left, srcroi.width, borderType)*cn;
+        for( k = 0; k < cn; k++ )
+            tab[i*cn + k] = j + k;
+    }
+
+    for( i = 0; i < right; i++ )
+    {
+        j = cv::borderInterpolate(srcroi.width + i, srcroi.width, borderType)*cn;
+        for( k = 0; k < cn; k++ )
+            tab[(i+left)*cn + k] = j + k;
+    }
+
+    srcroi.width *= cn;
+    dstroi.width *= cn;
+    left *= cn;
+    right *= cn;
+
+    uchar* dstInner = dst + dststep*top + left*elemSize;
+
+    for( i = 0; i < srcroi.height; i++, dstInner += dststep, src += srcstep )
+    {
+        if( dstInner != src )
+            memcpy(dstInner, src, srcroi.width*elemSize);
+
+        if( intMode )
+        {
+            const int* isrc = (int*)src;
+            int* idstInner = (int*)dstInner;
+            for( j = 0; j < left; j++ )
+                idstInner[j - left] = isrc[tab[j]];
+            for( j = 0; j < right; j++ )
+                idstInner[j + srcroi.width] = isrc[tab[j + left]];
+        }
+        else
+        {
+            for( j = 0; j < left; j++ )
+                dstInner[j - left] = src[tab[j]];
+            for( j = 0; j < right; j++ )
+                dstInner[j + srcroi.width] = src[tab[j + left]];
+        }
+    }
+
+    dstroi.width *= elemSize;
+    dst += dststep*top;
+
+    for( i = 0; i < top; i++ )
+    {
+        j = cv::borderInterpolate(i - top, srcroi.height, borderType);
+        memcpy(dst + (i - top)*dststep, dst + j*dststep, dstroi.width);
+    }
+
+    for( i = 0; i < bottom; i++ )
+    {
+        j = cv::borderInterpolate(i + srcroi.height, srcroi.height, borderType);
+        memcpy(dst + (i + srcroi.height)*dststep, dst + j*dststep, dstroi.width);
+    }
+}
+
+
+void copyMakeConstBorder_8u( const uchar* src, size_t srcstep, cv::Size srcroi,
+                             uchar* dst, size_t dststep, cv::Size dstroi,
+                             int top, int left, int cn, const uchar* value )
+{
+    int i, j;
+    cv::AutoBuffer<uchar> _constBuf(dstroi.width*cn);
+    uchar* constBuf = _constBuf;
+    int right = dstroi.width - srcroi.width - left;
+    int bottom = dstroi.height - srcroi.height - top;
+
+    for( i = 0; i < dstroi.width; i++ )
+    {
+        for( j = 0; j < cn; j++ )
+            constBuf[i*cn + j] = value[j];
+    }
+
+    srcroi.width *= cn;
+    dstroi.width *= cn;
+    left *= cn;
+    right *= cn;
+
+    uchar* dstInner = dst + dststep*top + left;
+
+    for( i = 0; i < srcroi.height; i++, dstInner += dststep, src += srcstep )
+    {
+        if( dstInner != src )
+            memcpy( dstInner, src, srcroi.width );
+        memcpy( dstInner - left, constBuf, left );
+        memcpy( dstInner + srcroi.width, constBuf, right );
+    }
+
+    dst += dststep*top;
+
+    for( i = 0; i < top; i++ )
+        memcpy(dst + (i - top)*dststep, constBuf, dstroi.width);
+
+    for( i = 0; i < bottom; i++ )
+        memcpy(dst + (i + srcroi.height)*dststep, constBuf, dstroi.width);
+}
+
+}
+
+void cv::copyMakeBorder( InputArray _src, OutputArray _dst, int top, int bottom,
+                         int left, int right, int borderType, const Scalar& value )
+{
+    Mat src = _src.getMat();
+    CV_Assert( top >= 0 && bottom >= 0 && left >= 0 && right >= 0 );
+
+    if( src.isSubmatrix() && (borderType & BORDER_ISOLATED) == 0 )
+    {
+        Size wholeSize;
+        Point ofs;
+        src.locateROI(wholeSize, ofs);
+        int dtop = std::min(ofs.y, top);
+        int dbottom = std::min(wholeSize.height - src.rows - ofs.y, bottom);
+        int dleft = std::min(ofs.x, left);
+        int dright = std::min(wholeSize.width - src.cols - ofs.x, right);
+        src.adjustROI(dtop, dbottom, dleft, dright);
+        top -= dtop;
+        left -= dleft;
+        bottom -= dbottom;
+        right -= dright;
+    }
+
+    _dst.create( src.rows + top + bottom, src.cols + left + right, src.type() );
+    Mat dst = _dst.getMat();
+
+    if(top == 0 && left == 0 && bottom == 0 && right == 0)
+    {
+        if(src.data != dst.data || src.step != dst.step)
+            src.copyTo(dst);
+        return;
+    }
+
+    borderType &= ~BORDER_ISOLATED;
+
+    if( borderType != BORDER_CONSTANT )
+        copyMakeBorder_8u( src.data, src.step, src.size(),
+                           dst.data, dst.step, dst.size(),
+                           top, left, (int)src.elemSize(), borderType );
+    else
+    {
+        int cn = src.channels(), cn1 = cn;
+        AutoBuffer<double> buf(cn);
+        if( cn > 4 )
+        {
+            CV_Assert( value[0] == value[1] && value[0] == value[2] && value[0] == value[3] );
+            cn1 = 1;
+        }
+        scalarToRawData(value, buf, CV_MAKETYPE(src.depth(), cn1), cn);
+        copyMakeConstBorder_8u( src.data, src.step, src.size(),
+                                dst.data, dst.step, dst.size(),
+                                top, left, (int)src.elemSize(), (uchar*)(double*)buf );
+    }
 }
 
 /* dst = src */
diff --git a/modules/core/src/cuda/gpu_mat.cu b/modules/core/src/cuda/gpu_mat.cu
new file mode 100644
index 0000000000..0db1584212
--- /dev/null
+++ b/modules/core/src/cuda/gpu_mat.cu
@@ -0,0 +1,486 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifndef HAVE_OPENCV_CUDEV
+
+#error "opencv_cudev is required"
+
+#else
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/cudev.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+
+/////////////////////////////////////////////////////
+/// create
+
+void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
+{
+    CV_DbgAssert( _rows >= 0 && _cols >= 0 );
+
+    _type &= Mat::TYPE_MASK;
+
+    if (rows == _rows && cols == _cols && type() == _type && data)
+        return;
+
+    if (data)
+        release();
+
+    if (_rows > 0 && _cols > 0)
+    {
+        flags = Mat::MAGIC_VAL + _type;
+        rows = _rows;
+        cols = _cols;
+
+        size_t esz = elemSize();
+
+        void* devPtr;
+
+        if (rows > 1 && cols > 1)
+        {
+            CV_CUDEV_SAFE_CALL( cudaMallocPitch(&devPtr, &step, esz * cols, rows) );
+        }
+        else
+        {
+            // Single row or single column must be continuous
+            CV_CUDEV_SAFE_CALL( cudaMalloc(&devPtr, esz * cols * rows) );
+            step = esz * cols;
+        }
+
+        if (esz * cols == step)
+            flags |= Mat::CONTINUOUS_FLAG;
+
+        int64 _nettosize = static_cast<int64>(step) * rows;
+        size_t nettosize = static_cast<size_t>(_nettosize);
+
+        datastart = data = static_cast<uchar*>(devPtr);
+        dataend = data + nettosize;
+
+        refcount = static_cast<int*>(fastMalloc(sizeof(*refcount)));
+        *refcount = 1;
+    }
+}
+
+/////////////////////////////////////////////////////
+/// release
+
+void cv::gpu::GpuMat::release()
+{
+    if (refcount && CV_XADD(refcount, -1) == 1)
+    {
+        cudaFree(datastart);
+        fastFree(refcount);
+    }
+
+    data = datastart = dataend = 0;
+    step = rows = cols = 0;
+    refcount = 0;
+}
+
+/////////////////////////////////////////////////////
+/// upload
+
+void cv::gpu::GpuMat::upload(InputArray arr)
+{
+    Mat mat = arr.getMat();
+
+    CV_DbgAssert( !mat.empty() );
+
+    create(mat.size(), mat.type());
+
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2D(data, step, mat.data, mat.step, cols * elemSize(), rows, cudaMemcpyHostToDevice) );
+}
+
+void cv::gpu::GpuMat::upload(InputArray arr, Stream& _stream)
+{
+    Mat mat = arr.getMat();
+
+    CV_DbgAssert( !mat.empty() );
+
+    create(mat.size(), mat.type());
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2DAsync(data, step, mat.data, mat.step, cols * elemSize(), rows, cudaMemcpyHostToDevice, stream) );
+}
+
+/////////////////////////////////////////////////////
+/// download
+
+void cv::gpu::GpuMat::download(OutputArray _dst) const
+{
+    CV_DbgAssert( !empty() );
+
+    _dst.create(size(), type());
+    Mat dst = _dst.getMat();
+
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2D(dst.data, dst.step, data, step, cols * elemSize(), rows, cudaMemcpyDeviceToHost) );
+}
+
+void cv::gpu::GpuMat::download(OutputArray _dst, Stream& _stream) const
+{
+    CV_DbgAssert( !empty() );
+
+    _dst.create(size(), type());
+    Mat dst = _dst.getMat();
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2DAsync(dst.data, dst.step, data, step, cols * elemSize(), rows, cudaMemcpyDeviceToHost, stream) );
+}
+
+/////////////////////////////////////////////////////
+/// copyTo
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst) const
+{
+    CV_DbgAssert( !empty() );
+
+    _dst.create(size(), type());
+    GpuMat dst = _dst.getGpuMat();
+
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2D(dst.data, dst.step, data, step, cols * elemSize(), rows, cudaMemcpyDeviceToDevice) );
+}
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst, Stream& _stream) const
+{
+    CV_DbgAssert( !empty() );
+
+    _dst.create(size(), type());
+    GpuMat dst = _dst.getGpuMat();
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+    CV_CUDEV_SAFE_CALL( cudaMemcpy2DAsync(dst.data, dst.step, data, step, cols * elemSize(), rows, cudaMemcpyDeviceToDevice, stream) );
+}
+
+namespace
+{
+    template <size_t size> struct CopyToPolicy : DefaultTransformPolicy
+    {
+    };
+    template <> struct CopyToPolicy<4> : DefaultTransformPolicy
+    {
+        enum {
+            shift = 2
+        };
+    };
+    template <> struct CopyToPolicy<8> : DefaultTransformPolicy
+    {
+        enum {
+            shift = 1
+        };
+    };
+
+    template <typename T>
+    void copyWithMask(const GpuMat& src, const GpuMat& dst, const GpuMat& mask, Stream& stream)
+    {
+        gridTransform_< CopyToPolicy<sizeof(typename VecTraits<T>::elem_type)> >(globPtr<T>(src), globPtr<T>(dst), identity<T>(), globPtr<uchar>(mask), stream);
+    }
+}
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst, InputArray _mask, Stream& stream) const
+{
+    CV_DbgAssert( !empty() );
+    CV_DbgAssert( depth() <= CV_64F && channels() <= 4 );
+
+    GpuMat mask = _mask.getGpuMat();
+    CV_DbgAssert( size() == mask.size() && mask.depth() == CV_8U && (mask.channels() == 1 || mask.channels() == channels()) );
+
+    _dst.create(size(), type());
+    GpuMat dst = _dst.getGpuMat();
+
+    typedef void (*func_t)(const GpuMat& src, const GpuMat& dst, const GpuMat& mask, Stream& stream);
+    static const func_t funcs[9][4] =
+    {
+        {0,0,0,0},
+        {copyWithMask<uchar>, copyWithMask<uchar2>, copyWithMask<uchar3>, copyWithMask<uchar4>},
+        {copyWithMask<ushort>, copyWithMask<ushort2>, copyWithMask<ushort3>, copyWithMask<ushort4>},
+        {0,0,0,0},
+        {copyWithMask<int>, copyWithMask<int2>, copyWithMask<int3>, copyWithMask<int4>},
+        {0,0,0,0},
+        {0,0,0,0},
+        {0,0,0,0},
+        {copyWithMask<double>, copyWithMask<double2>, copyWithMask<double3>, copyWithMask<double4>}
+    };
+
+    if (mask.channels() == channels())
+    {
+        const func_t func = funcs[elemSize1()][0];
+        CV_DbgAssert( func != 0 );
+        func(reshape(1), dst.reshape(1), mask.reshape(1), stream);
+    }
+    else
+    {
+        const func_t func = funcs[elemSize1()][channels() - 1];
+        CV_DbgAssert( func != 0 );
+        func(*this, dst, mask, stream);
+    }
+}
+
+/////////////////////////////////////////////////////
+/// setTo
+
+namespace
+{
+    template <typename T>
+    void setToWithOutMask(const GpuMat& mat, Scalar _scalar, Stream& stream)
+    {
+        Scalar_<typename VecTraits<T>::elem_type> scalar = _scalar;
+        gridTransform(constantPtr(VecTraits<T>::make(scalar.val), mat.rows, mat.cols), globPtr<T>(mat), identity<T>(), stream);
+    }
+
+    template <typename T>
+    void setToWithMask(const GpuMat& mat, const GpuMat& mask, Scalar _scalar, Stream& stream)
+    {
+        Scalar_<typename VecTraits<T>::elem_type> scalar = _scalar;
+        gridTransform(constantPtr(VecTraits<T>::make(scalar.val), mat.rows, mat.cols), globPtr<T>(mat), identity<T>(), globPtr<uchar>(mask), stream);
+    }
+}
+
+GpuMat& cv::gpu::GpuMat::setTo(Scalar value, Stream& stream)
+{
+    CV_DbgAssert( !empty() );
+    CV_DbgAssert( depth() <= CV_64F && channels() <= 4 );
+
+    if (value[0] == 0.0 && value[1] == 0.0 && value[2] == 0.0 && value[3] == 0.0)
+    {
+        // Zero fill
+
+        if (stream)
+            CV_CUDEV_SAFE_CALL( cudaMemset2DAsync(data, step, 0, cols * elemSize(), rows, StreamAccessor::getStream(stream)) );
+        else
+            CV_CUDEV_SAFE_CALL( cudaMemset2D(data, step, 0, cols * elemSize(), rows) );
+
+        return *this;
+    }
+
+    if (depth() == CV_8U)
+    {
+        const int cn = channels();
+
+        if (cn == 1
+                || (cn == 2 && value[0] == value[1])
+                || (cn == 3 && value[0] == value[1] && value[0] == value[2])
+                || (cn == 4 && value[0] == value[1] && value[0] == value[2] && value[0] == value[3]))
+        {
+            const int val = cv::saturate_cast<uchar>(value[0]);
+
+            if (stream)
+                CV_CUDEV_SAFE_CALL( cudaMemset2DAsync(data, step, val, cols * elemSize(), rows, StreamAccessor::getStream(stream)) );
+            else
+                CV_CUDEV_SAFE_CALL( cudaMemset2D(data, step, val, cols * elemSize(), rows) );
+
+            return *this;
+        }
+    }
+
+    typedef void (*func_t)(const GpuMat& mat, Scalar scalar, Stream& stream);
+    static const func_t funcs[7][4] =
+    {
+        {setToWithOutMask<uchar>,setToWithOutMask<uchar2>,setToWithOutMask<uchar3>,setToWithOutMask<uchar4>},
+        {setToWithOutMask<schar>,setToWithOutMask<char2>,setToWithOutMask<char3>,setToWithOutMask<char4>},
+        {setToWithOutMask<ushort>,setToWithOutMask<ushort2>,setToWithOutMask<ushort3>,setToWithOutMask<ushort4>},
+        {setToWithOutMask<short>,setToWithOutMask<short2>,setToWithOutMask<short3>,setToWithOutMask<short4>},
+        {setToWithOutMask<int>,setToWithOutMask<int2>,setToWithOutMask<int3>,setToWithOutMask<int4>},
+        {setToWithOutMask<float>,setToWithOutMask<float2>,setToWithOutMask<float3>,setToWithOutMask<float4>},
+        {setToWithOutMask<double>,setToWithOutMask<double2>,setToWithOutMask<double3>,setToWithOutMask<double4>}
+    };
+
+    funcs[depth()][channels() - 1](*this, value, stream);
+
+    return *this;
+}
+
+GpuMat& cv::gpu::GpuMat::setTo(Scalar value, InputArray _mask, Stream& stream)
+{
+    CV_DbgAssert( !empty() );
+    CV_DbgAssert( depth() <= CV_64F && channels() <= 4 );
+
+    GpuMat mask = _mask.getGpuMat();
+
+    CV_DbgAssert( size() == mask.size() && mask.type() == CV_8UC1 );
+
+    typedef void (*func_t)(const GpuMat& mat, const GpuMat& mask, Scalar scalar, Stream& stream);
+    static const func_t funcs[7][4] =
+    {
+        {setToWithMask<uchar>,setToWithMask<uchar2>,setToWithMask<uchar3>,setToWithMask<uchar4>},
+        {setToWithMask<schar>,setToWithMask<char2>,setToWithMask<char3>,setToWithMask<char4>},
+        {setToWithMask<ushort>,setToWithMask<ushort2>,setToWithMask<ushort3>,setToWithMask<ushort4>},
+        {setToWithMask<short>,setToWithMask<short2>,setToWithMask<short3>,setToWithMask<short4>},
+        {setToWithMask<int>,setToWithMask<int2>,setToWithMask<int3>,setToWithMask<int4>},
+        {setToWithMask<float>,setToWithMask<float2>,setToWithMask<float3>,setToWithMask<float4>},
+        {setToWithMask<double>,setToWithMask<double2>,setToWithMask<double3>,setToWithMask<double4>}
+    };
+
+    funcs[depth()][channels() - 1](*this, mask, value, stream);
+
+    return *this;
+}
+
+/////////////////////////////////////////////////////
+/// convertTo
+
+namespace
+{
+    template <typename T> struct ConvertToPolicy : DefaultTransformPolicy
+    {
+    };
+    template <> struct ConvertToPolicy<double> : DefaultTransformPolicy
+    {
+        enum {
+            shift = 1
+        };
+    };
+
+    template <typename T, typename D>
+    void convertToNoScale(const GpuMat& src, const GpuMat& dst, Stream& stream)
+    {
+        typedef typename VecTraits<T>::elem_type src_elem_type;
+        typedef typename VecTraits<D>::elem_type dst_elem_type;
+        typedef typename LargerType<src_elem_type, float>::type larger_elem_type;
+        typedef typename LargerType<float, dst_elem_type>::type scalar_type;
+
+        gridTransform_< ConvertToPolicy<scalar_type> >(globPtr<T>(src), globPtr<D>(dst), saturate_cast_func<T, D>(), stream);
+    }
+
+    template <typename T, typename D, typename S> struct Convertor : unary_function<T, D>
+    {
+        S alpha;
+        S beta;
+
+        __device__ __forceinline__ D operator ()(typename TypeTraits<T>::parameter_type src) const
+        {
+            return cudev::saturate_cast<D>(alpha * src + beta);
+        }
+    };
+
+    template <typename T, typename D>
+    void convertToScale(const GpuMat& src, const GpuMat& dst, double alpha, double beta, Stream& stream)
+    {
+        typedef typename VecTraits<T>::elem_type src_elem_type;
+        typedef typename VecTraits<D>::elem_type dst_elem_type;
+        typedef typename LargerType<src_elem_type, float>::type larger_elem_type;
+        typedef typename LargerType<float, dst_elem_type>::type scalar_type;
+
+        Convertor<T, D, scalar_type> op;
+        op.alpha = cv::saturate_cast<scalar_type>(alpha);
+        op.beta = cv::saturate_cast<scalar_type>(beta);
+
+        gridTransform_< ConvertToPolicy<scalar_type> >(globPtr<T>(src), globPtr<D>(dst), op, stream);
+    }
+}
+
+void cv::gpu::GpuMat::convertTo(OutputArray _dst, int rtype, Stream& stream) const
+{
+    if (rtype < 0)
+        rtype = type();
+    else
+        rtype = CV_MAKE_TYPE(CV_MAT_DEPTH(rtype), channels());
+
+    const int sdepth = depth();
+    const int ddepth = CV_MAT_DEPTH(rtype);
+    if (sdepth == ddepth)
+    {
+        if (stream)
+            copyTo(_dst, stream);
+        else
+            copyTo(_dst);
+
+        return;
+    }
+
+    CV_DbgAssert( sdepth <= CV_64F && ddepth <= CV_64F );
+
+    GpuMat src = *this;
+
+    _dst.create(size(), rtype);
+    GpuMat dst = _dst.getGpuMat();
+
+    typedef void (*func_t)(const GpuMat& src, const GpuMat& dst, Stream& stream);
+    static const func_t funcs[7][7] =
+    {
+        {0, convertToNoScale<uchar, schar>, convertToNoScale<uchar, ushort>, convertToNoScale<uchar, short>, convertToNoScale<uchar, int>, convertToNoScale<uchar, float>, convertToNoScale<uchar, double>},
+        {convertToNoScale<schar, uchar>, 0, convertToNoScale<schar, ushort>, convertToNoScale<schar, short>, convertToNoScale<schar, int>, convertToNoScale<schar, float>, convertToNoScale<schar, double>},
+        {convertToNoScale<ushort, uchar>, convertToNoScale<ushort, schar>, 0, convertToNoScale<ushort, short>, convertToNoScale<ushort, int>, convertToNoScale<ushort, float>, convertToNoScale<ushort, double>},
+        {convertToNoScale<short, uchar>, convertToNoScale<short, schar>, convertToNoScale<short, ushort>, 0, convertToNoScale<short, int>, convertToNoScale<short, float>, convertToNoScale<short, double>},
+        {convertToNoScale<int, uchar>, convertToNoScale<int, schar>, convertToNoScale<int, ushort>, convertToNoScale<int, short>, 0, convertToNoScale<int, float>, convertToNoScale<int, double>},
+        {convertToNoScale<float, uchar>, convertToNoScale<float, schar>, convertToNoScale<float, ushort>, convertToNoScale<float, short>, convertToNoScale<float, int>, 0, convertToNoScale<float, double>},
+        {convertToNoScale<double, uchar>, convertToNoScale<double, schar>, convertToNoScale<double, ushort>, convertToNoScale<double, short>, convertToNoScale<double, int>, convertToNoScale<double, float>, 0}
+    };
+
+    funcs[sdepth][ddepth](reshape(1), dst.reshape(1), stream);
+}
+
+void cv::gpu::GpuMat::convertTo(OutputArray _dst, int rtype, double alpha, double beta, Stream& stream) const
+{
+    if (rtype < 0)
+        rtype = type();
+    else
+        rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), channels());
+
+    const int sdepth = depth();
+    const int ddepth = CV_MAT_DEPTH(rtype);
+
+    GpuMat src = *this;
+
+    _dst.create(size(), rtype);
+    GpuMat dst = _dst.getGpuMat();
+
+    typedef void (*func_t)(const GpuMat& src, const GpuMat& dst, double alpha, double beta, Stream& stream);
+    static const func_t funcs[7][7] =
+    {
+        {convertToScale<uchar, uchar>, convertToScale<uchar, schar>, convertToScale<uchar, ushort>, convertToScale<uchar, short>, convertToScale<uchar, int>, convertToScale<uchar, float>, convertToScale<uchar, double>},
+        {convertToScale<schar, uchar>, convertToScale<schar, schar>, convertToScale<schar, ushort>, convertToScale<schar, short>, convertToScale<schar, int>, convertToScale<schar, float>, convertToScale<schar, double>},
+        {convertToScale<ushort, uchar>, convertToScale<ushort, schar>, convertToScale<ushort, ushort>, convertToScale<ushort, short>, convertToScale<ushort, int>, convertToScale<ushort, float>, convertToScale<ushort, double>},
+        {convertToScale<short, uchar>, convertToScale<short, schar>, convertToScale<short, ushort>, convertToScale<short, short>, convertToScale<short, int>, convertToScale<short, float>, convertToScale<short, double>},
+        {convertToScale<int, uchar>, convertToScale<int, schar>, convertToScale<int, ushort>, convertToScale<int, short>, convertToScale<int, int>, convertToScale<int, float>, convertToScale<int, double>},
+        {convertToScale<float, uchar>, convertToScale<float, schar>, convertToScale<float, ushort>, convertToScale<float, short>, convertToScale<float, int>, convertToScale<float, float>, convertToScale<float, double>},
+        {convertToScale<double, uchar>, convertToScale<double, schar>, convertToScale<double, ushort>, convertToScale<double, short>, convertToScale<double, int>, convertToScale<double, float>, convertToScale<double, double>}
+    };
+
+    funcs[sdepth][ddepth](reshape(1), dst.reshape(1), alpha, beta, stream);
+}
+
+#endif
diff --git a/modules/core/src/cuda/matrix_operations.cu b/modules/core/src/cuda/matrix_operations.cu
deleted file mode 100644
index 60aa073406..0000000000
--- a/modules/core/src/cuda/matrix_operations.cu
+++ /dev/null
@@ -1,383 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/transform.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/type_traits.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    void writeScalar(const uchar*);
-    void writeScalar(const schar*);
-    void writeScalar(const ushort*);
-    void writeScalar(const short int*);
-    void writeScalar(const int*);
-    void writeScalar(const float*);
-    void writeScalar(const double*);
-    void copyToWithMask_gpu(PtrStepSzb src, PtrStepSzb dst, size_t elemSize1, int cn, PtrStepSzb mask, bool colorMask, cudaStream_t stream);
-    void convert_gpu(PtrStepSzb, int, PtrStepSzb, int, double, double, cudaStream_t);
-}}}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct shift_and_sizeof;
-    template <> struct shift_and_sizeof<signed char> { enum { shift = 0 }; };
-    template <> struct shift_and_sizeof<unsigned char> { enum { shift = 0 }; };
-    template <> struct shift_and_sizeof<short> { enum { shift = 1 }; };
-    template <> struct shift_and_sizeof<unsigned short> { enum { shift = 1 }; };
-    template <> struct shift_and_sizeof<int> { enum { shift = 2 }; };
-    template <> struct shift_and_sizeof<float> { enum { shift = 2 }; };
-    template <> struct shift_and_sizeof<double> { enum { shift = 3 }; };
-
-    ///////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////// CopyTo /////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////
-
-    template <typename T> void copyToWithMask(PtrStepSzb src, PtrStepSzb dst, int cn, PtrStepSzb mask, bool colorMask, cudaStream_t stream)
-    {
-        if (colorMask)
-            cv::gpu::device::transform((PtrStepSz<T>)src, (PtrStepSz<T>)dst, identity<T>(), SingleMask(mask), stream);
-        else
-            cv::gpu::device::transform((PtrStepSz<T>)src, (PtrStepSz<T>)dst, identity<T>(), SingleMaskChannels(mask, cn), stream);
-    }
-
-    void copyToWithMask_gpu(PtrStepSzb src, PtrStepSzb dst, size_t elemSize1, int cn, PtrStepSzb mask, bool colorMask, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, int cn, PtrStepSzb mask, bool colorMask, cudaStream_t stream);
-
-        static func_t tab[] =
-        {
-            0,
-            copyToWithMask<unsigned char>,
-            copyToWithMask<unsigned short>,
-            0,
-            copyToWithMask<int>,
-            0,
-            0,
-            0,
-            copyToWithMask<double>
-        };
-
-        tab[elemSize1](src, dst, cn, mask, colorMask, stream);
-    }
-
-    ///////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////// SetTo //////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////
-
-    __constant__ uchar scalar_8u[4];
-    __constant__ schar scalar_8s[4];
-    __constant__ ushort scalar_16u[4];
-    __constant__ short scalar_16s[4];
-    __constant__ int scalar_32s[4];
-    __constant__ float scalar_32f[4];
-    __constant__ double scalar_64f[4];
-
-    template <typename T> __device__ __forceinline__ T readScalar(int i);
-    template <> __device__ __forceinline__ uchar readScalar<uchar>(int i) {return scalar_8u[i];}
-    template <> __device__ __forceinline__ schar readScalar<schar>(int i) {return scalar_8s[i];}
-    template <> __device__ __forceinline__ ushort readScalar<ushort>(int i) {return scalar_16u[i];}
-    template <> __device__ __forceinline__ short readScalar<short>(int i) {return scalar_16s[i];}
-    template <> __device__ __forceinline__ int readScalar<int>(int i) {return scalar_32s[i];}
-    template <> __device__ __forceinline__ float readScalar<float>(int i) {return scalar_32f[i];}
-    template <> __device__ __forceinline__ double readScalar<double>(int i) {return scalar_64f[i];}
-
-    void writeScalar(const uchar* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_8u, vals, sizeof(uchar) * 4) );
-    }
-    void writeScalar(const schar* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_8s, vals, sizeof(schar) * 4) );
-    }
-    void writeScalar(const ushort* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_16u, vals, sizeof(ushort) * 4) );
-    }
-    void writeScalar(const short* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_16s, vals, sizeof(short) * 4) );
-    }
-    void writeScalar(const int* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_32s, vals, sizeof(int) * 4) );
-    }
-    void writeScalar(const float* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_32f, vals, sizeof(float) * 4) );
-    }
-    void writeScalar(const double* vals)
-    {
-        cudaSafeCall( cudaMemcpyToSymbol(scalar_64f, vals, sizeof(double) * 4) );
-    }
-
-    template<typename T>
-    __global__ void set_to_without_mask(T* mat, int cols, int rows, size_t step, int channels)
-    {
-        size_t x = blockIdx.x * blockDim.x + threadIdx.x;
-        size_t y = blockIdx.y * blockDim.y + threadIdx.y;
-
-        if ((x < cols * channels ) && (y < rows))
-        {
-            size_t idx = y * ( step >> shift_and_sizeof<T>::shift ) + x;
-            mat[idx] = readScalar<T>(x % channels);
-        }
-    }
-
-    template<typename T>
-    __global__ void set_to_with_mask(T* mat, const uchar* mask, int cols, int rows, size_t step, int channels, size_t step_mask)
-    {
-        size_t x = blockIdx.x * blockDim.x + threadIdx.x;
-        size_t y = blockIdx.y * blockDim.y + threadIdx.y;
-
-        if ((x < cols * channels ) && (y < rows))
-            if (mask[y * step_mask + x / channels] != 0)
-            {
-                size_t idx = y * ( step >> shift_and_sizeof<T>::shift ) + x;
-                mat[idx] = readScalar<T>(x % channels);
-            }
-    }
-    template <typename T>
-    void set_to_gpu(PtrStepSzb mat, const T* scalar, PtrStepSzb mask, int channels, cudaStream_t stream)
-    {
-        writeScalar(scalar);
-
-        dim3 threadsPerBlock(32, 8, 1);
-        dim3 numBlocks (mat.cols * channels / threadsPerBlock.x + 1, mat.rows / threadsPerBlock.y + 1, 1);
-
-        set_to_with_mask<T><<<numBlocks, threadsPerBlock, 0, stream>>>((T*)mat.data, (uchar*)mask.data, mat.cols, mat.rows, mat.step, channels, mask.step);
-        cudaSafeCall( cudaGetLastError() );
-
-        if (stream == 0)
-            cudaSafeCall ( cudaDeviceSynchronize() );
-    }
-
-    template void set_to_gpu<uchar >(PtrStepSzb mat, const uchar*  scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<schar >(PtrStepSzb mat, const schar*  scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<ushort>(PtrStepSzb mat, const ushort* scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<short >(PtrStepSzb mat, const short*  scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<int   >(PtrStepSzb mat, const int*    scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<float >(PtrStepSzb mat, const float*  scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-    template void set_to_gpu<double>(PtrStepSzb mat, const double* scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-
-    template <typename T>
-    void set_to_gpu(PtrStepSzb mat, const T* scalar, int channels, cudaStream_t stream)
-    {
-        writeScalar(scalar);
-
-        dim3 threadsPerBlock(32, 8, 1);
-        dim3 numBlocks (mat.cols * channels / threadsPerBlock.x + 1, mat.rows / threadsPerBlock.y + 1, 1);
-
-        set_to_without_mask<T><<<numBlocks, threadsPerBlock, 0, stream>>>((T*)mat.data, mat.cols, mat.rows, mat.step, channels);
-        cudaSafeCall( cudaGetLastError() );
-
-        if (stream == 0)
-            cudaSafeCall ( cudaDeviceSynchronize() );
-    }
-
-    template void set_to_gpu<uchar >(PtrStepSzb mat, const uchar*  scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<schar >(PtrStepSzb mat, const schar*  scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<ushort>(PtrStepSzb mat, const ushort* scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<short >(PtrStepSzb mat, const short*  scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<int   >(PtrStepSzb mat, const int*    scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<float >(PtrStepSzb mat, const float*  scalar, int channels, cudaStream_t stream);
-    template void set_to_gpu<double>(PtrStepSzb mat, const double* scalar, int channels, cudaStream_t stream);
-
-    ///////////////////////////////////////////////////////////////////////////
-    //////////////////////////////// ConvertTo ////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////
-
-    template <typename T, typename D, typename S> struct Convertor : unary_function<T, D>
-    {
-        Convertor(S alpha_, S beta_) : alpha(alpha_), beta(beta_) {}
-
-        __device__ __forceinline__ D operator()(typename TypeTraits<T>::ParameterType src) const
-        {
-            return saturate_cast<D>(alpha * src + beta);
-        }
-
-        S alpha, beta;
-    };
-
-    namespace detail
-    {
-        template <size_t src_size, size_t dst_size, typename F> struct ConvertTraitsDispatcher : DefaultTransformFunctorTraits<F>
-        {
-        };
-        template <typename F> struct ConvertTraitsDispatcher<1, 1, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_shift = 8 };
-        };
-        template <typename F> struct ConvertTraitsDispatcher<1, 2, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_shift = 4 };
-        };
-        template <typename F> struct ConvertTraitsDispatcher<1, 4, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_block_dim_y = 8 };
-            enum { smart_shift = 4 };
-        };
-
-        template <typename F> struct ConvertTraitsDispatcher<2, 2, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_shift = 4 };
-        };
-        template <typename F> struct ConvertTraitsDispatcher<2, 4, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_shift = 2 };
-        };
-
-        template <typename F> struct ConvertTraitsDispatcher<4, 2, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_block_dim_y = 8 };
-            enum { smart_shift = 4 };
-        };
-        template <typename F> struct ConvertTraitsDispatcher<4, 4, F> : DefaultTransformFunctorTraits<F>
-        {
-            enum { smart_block_dim_y = 8 };
-            enum { smart_shift = 2 };
-        };
-
-        template <typename F> struct ConvertTraits : ConvertTraitsDispatcher<sizeof(typename F::argument_type), sizeof(typename F::result_type), F>
-        {
-        };
-    }
-
-    template <typename T, typename D, typename S> struct TransformFunctorTraits< Convertor<T, D, S> > : detail::ConvertTraits< Convertor<T, D, S> >
-    {
-    };
-
-    template<typename T, typename D, typename S>
-    void cvt_(PtrStepSzb src, PtrStepSzb dst, double alpha, double beta, cudaStream_t stream)
-    {
-        cudaSafeCall( cudaSetDoubleForDevice(&alpha) );
-        cudaSafeCall( cudaSetDoubleForDevice(&beta) );
-        Convertor<T, D, S> op(static_cast<S>(alpha), static_cast<S>(beta));
-        cv::gpu::device::transform((PtrStepSz<T>)src, (PtrStepSz<D>)dst, op, WithOutMask(), stream);
-    }
-
-#if defined  __clang__
-# pragma clang diagnostic push
-# pragma clang diagnostic ignored "-Wmissing-declarations"
-#endif
-
-    void convert_gpu(PtrStepSzb src, int sdepth, PtrStepSzb dst, int ddepth, double alpha, double beta, cudaStream_t stream)
-    {
-        typedef void (*caller_t)(PtrStepSzb src, PtrStepSzb dst, double alpha, double beta, cudaStream_t stream);
-
-        static const caller_t tab[7][7] =
-        {
-            {
-                cvt_<uchar, uchar, float>,
-                cvt_<uchar, schar, float>,
-                cvt_<uchar, ushort, float>,
-                cvt_<uchar, short, float>,
-                cvt_<uchar, int, float>,
-                cvt_<uchar, float, float>,
-                cvt_<uchar, double, double>
-            },
-            {
-                cvt_<schar, uchar, float>,
-                cvt_<schar, schar, float>,
-                cvt_<schar, ushort, float>,
-                cvt_<schar, short, float>,
-                cvt_<schar, int, float>,
-                cvt_<schar, float, float>,
-                cvt_<schar, double, double>
-            },
-            {
-                cvt_<ushort, uchar, float>,
-                cvt_<ushort, schar, float>,
-                cvt_<ushort, ushort, float>,
-                cvt_<ushort, short, float>,
-                cvt_<ushort, int, float>,
-                cvt_<ushort, float, float>,
-                cvt_<ushort, double, double>
-            },
-            {
-                cvt_<short, uchar, float>,
-                cvt_<short, schar, float>,
-                cvt_<short, ushort, float>,
-                cvt_<short, short, float>,
-                cvt_<short, int, float>,
-                cvt_<short, float, float>,
-                cvt_<short, double, double>
-            },
-            {
-                cvt_<int, uchar, float>,
-                cvt_<int, schar, float>,
-                cvt_<int, ushort, float>,
-                cvt_<int, short, float>,
-                cvt_<int, int, double>,
-                cvt_<int, float, double>,
-                cvt_<int, double, double>
-            },
-            {
-                cvt_<float, uchar, float>,
-                cvt_<float, schar, float>,
-                cvt_<float, ushort, float>,
-                cvt_<float, short, float>,
-                cvt_<float, int, float>,
-                cvt_<float, float, float>,
-                cvt_<float, double, double>
-            },
-            {
-                cvt_<double, uchar, double>,
-                cvt_<double, schar, double>,
-                cvt_<double, ushort, double>,
-                cvt_<double, short, double>,
-                cvt_<double, int, double>,
-                cvt_<double, float, double>,
-                cvt_<double, double, double>
-            }
-        };
-
-        caller_t func = tab[sdepth][ddepth];
-        func(src, dst, alpha, beta, stream);
-    }
-
-#if defined __clang__
-# pragma clang diagnostic pop
-#endif
-}}} // namespace cv { namespace gpu { namespace device
diff --git a/modules/core/src/cudastream.cpp b/modules/core/src/cudastream.cpp
deleted file mode 100644
index 6244af9e5e..0000000000
--- a/modules/core/src/cudastream.cpp
+++ /dev/null
@@ -1,352 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#include "opencv2/core/gpumat.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA)
-#define throw_nogpu() CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support")
-
-cv::gpu::Stream::Stream() { throw_nogpu(); }
-cv::gpu::Stream::~Stream() {}
-cv::gpu::Stream::Stream(const Stream&) { throw_nogpu(); }
-Stream& cv::gpu::Stream::operator=(const Stream&) { throw_nogpu(); return *this; }
-bool cv::gpu::Stream::queryIfComplete() { throw_nogpu(); return false; }
-void cv::gpu::Stream::waitForCompletion() { throw_nogpu(); }
-void cv::gpu::Stream::enqueueDownload(const GpuMat&, Mat&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueDownload(const GpuMat&, CudaMem&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueUpload(const CudaMem&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueUpload(const Mat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueCopy(const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueMemSet(GpuMat&, Scalar) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueMemSet(GpuMat&, Scalar, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueConvert(const GpuMat&, GpuMat&, int, double, double) { throw_nogpu(); }
-void cv::gpu::Stream::enqueueHostCallback(StreamCallback, void*) { throw_nogpu(); }
-Stream& cv::gpu::Stream::Null() { throw_nogpu(); static Stream s; return s; }
-cv::gpu::Stream::operator bool() const { throw_nogpu(); return false; }
-cv::gpu::Stream::Stream(Impl*) { throw_nogpu(); }
-void cv::gpu::Stream::create() { throw_nogpu(); }
-void cv::gpu::Stream::release() { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-#include "opencv2/core/stream_accessor.hpp"
-
-namespace cv { namespace gpu
-{
-    void copyWithMask(const GpuMat& src, GpuMat& dst, const GpuMat& mask, cudaStream_t stream);
-    void convertTo(const GpuMat& src, GpuMat& dst, double alpha, double beta, cudaStream_t stream);
-    void setTo(GpuMat& src, Scalar s, cudaStream_t stream);
-    void setTo(GpuMat& src, Scalar s, const GpuMat& mask, cudaStream_t stream);
-}}
-
-struct Stream::Impl
-{
-    static cudaStream_t getStream(const Impl* impl)
-    {
-        return impl ? impl->stream : 0;
-    }
-
-    cudaStream_t stream;
-    int ref_counter;
-};
-
-cudaStream_t cv::gpu::StreamAccessor::getStream(const Stream& stream)
-{
-    return Stream::Impl::getStream(stream.impl);
-}
-
-cv::gpu::Stream::Stream() : impl(0)
-{
-    create();
-}
-
-cv::gpu::Stream::~Stream()
-{
-    release();
-}
-
-cv::gpu::Stream::Stream(const Stream& stream) : impl(stream.impl)
-{
-    if (impl)
-        CV_XADD(&impl->ref_counter, 1);
-}
-
-Stream& cv::gpu::Stream::operator =(const Stream& stream)
-{
-    if (this != &stream)
-    {
-        release();
-        impl = stream.impl;
-        if (impl)
-            CV_XADD(&impl->ref_counter, 1);
-    }
-
-    return *this;
-}
-
-bool cv::gpu::Stream::queryIfComplete()
-{
-    cudaStream_t stream = Impl::getStream(impl);
-    cudaError_t err = cudaStreamQuery(stream);
-
-    if (err == cudaErrorNotReady || err == cudaSuccess)
-        return err == cudaSuccess;
-
-    cudaSafeCall(err);
-    return false;
-}
-
-void cv::gpu::Stream::waitForCompletion()
-{
-    cudaStream_t stream = Impl::getStream(impl);
-    cudaSafeCall( cudaStreamSynchronize(stream) );
-}
-
-void cv::gpu::Stream::enqueueDownload(const GpuMat& src, Mat& dst)
-{
-    // if not -> allocation will be done, but after that dst will not point to page locked memory
-    CV_Assert( src.size() == dst.size() && src.type() == dst.type() );
-
-    cudaStream_t stream = Impl::getStream(impl);
-    size_t bwidth = src.cols * src.elemSize();
-    cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, cudaMemcpyDeviceToHost, stream) );
-}
-
-void cv::gpu::Stream::enqueueDownload(const GpuMat& src, CudaMem& dst)
-{
-    dst.create(src.size(), src.type(), CudaMem::ALLOC_PAGE_LOCKED);
-
-    cudaStream_t stream = Impl::getStream(impl);
-    size_t bwidth = src.cols * src.elemSize();
-    cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, cudaMemcpyDeviceToHost, stream) );
-}
-
-void cv::gpu::Stream::enqueueUpload(const CudaMem& src, GpuMat& dst)
-{
-    dst.create(src.size(), src.type());
-
-    cudaStream_t stream = Impl::getStream(impl);
-    size_t bwidth = src.cols * src.elemSize();
-    cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, cudaMemcpyHostToDevice, stream) );
-}
-
-void cv::gpu::Stream::enqueueUpload(const Mat& src, GpuMat& dst)
-{
-    dst.create(src.size(), src.type());
-
-    cudaStream_t stream = Impl::getStream(impl);
-    size_t bwidth = src.cols * src.elemSize();
-    cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, cudaMemcpyHostToDevice, stream) );
-}
-
-void cv::gpu::Stream::enqueueCopy(const GpuMat& src, GpuMat& dst)
-{
-    dst.create(src.size(), src.type());
-
-    cudaStream_t stream = Impl::getStream(impl);
-    size_t bwidth = src.cols * src.elemSize();
-    cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, cudaMemcpyDeviceToDevice, stream) );
-}
-
-void cv::gpu::Stream::enqueueMemSet(GpuMat& src, Scalar val)
-{
-    const int sdepth = src.depth();
-
-    if (sdepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    cudaStream_t stream = Impl::getStream(impl);
-
-    if (val[0] == 0.0 && val[1] == 0.0 && val[2] == 0.0 && val[3] == 0.0)
-    {
-        cudaSafeCall( cudaMemset2DAsync(src.data, src.step, 0, src.cols * src.elemSize(), src.rows, stream) );
-        return;
-    }
-
-    if (sdepth == CV_8U)
-    {
-        int cn = src.channels();
-
-        if (cn == 1 || (cn == 2 && val[0] == val[1]) || (cn == 3 && val[0] == val[1] && val[0] == val[2]) || (cn == 4 && val[0] == val[1] && val[0] == val[2] && val[0] == val[3]))
-        {
-            int ival = saturate_cast<uchar>(val[0]);
-            cudaSafeCall( cudaMemset2DAsync(src.data, src.step, ival, src.cols * src.elemSize(), src.rows, stream) );
-            return;
-        }
-    }
-
-    setTo(src, val, stream);
-}
-
-void cv::gpu::Stream::enqueueMemSet(GpuMat& src, Scalar val, const GpuMat& mask)
-{
-    const int sdepth = src.depth();
-
-    if (sdepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    CV_Assert(mask.type() == CV_8UC1);
-
-    cudaStream_t stream = Impl::getStream(impl);
-
-    setTo(src, val, mask, stream);
-}
-
-void cv::gpu::Stream::enqueueConvert(const GpuMat& src, GpuMat& dst, int dtype, double alpha, double beta)
-{
-    if (dtype < 0)
-        dtype = src.type();
-    else
-        dtype = CV_MAKE_TYPE(CV_MAT_DEPTH(dtype), src.channels());
-
-    const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    bool noScale = fabs(alpha - 1) < std::numeric_limits<double>::epsilon()
-                && fabs(beta) < std::numeric_limits<double>::epsilon();
-
-    if (sdepth == ddepth && noScale)
-    {
-        enqueueCopy(src, dst);
-        return;
-    }
-
-    dst.create(src.size(), dtype);
-
-    cudaStream_t stream = Impl::getStream(impl);
-    convertTo(src, dst, alpha, beta, stream);
-}
-
-#if CUDA_VERSION >= 5000
-
-namespace
-{
-    struct CallbackData
-    {
-        cv::gpu::Stream::StreamCallback callback;
-        void* userData;
-        Stream stream;
-    };
-
-    void CUDART_CB cudaStreamCallback(cudaStream_t, cudaError_t status, void* userData)
-    {
-        CallbackData* data = reinterpret_cast<CallbackData*>(userData);
-        data->callback(data->stream, static_cast<int>(status), data->userData);
-        delete data;
-    }
-}
-
-#endif
-
-void cv::gpu::Stream::enqueueHostCallback(StreamCallback callback, void* userData)
-{
-#if CUDA_VERSION >= 5000
-    CallbackData* data = new CallbackData;
-    data->callback = callback;
-    data->userData = userData;
-    data->stream = *this;
-
-    cudaStream_t stream = Impl::getStream(impl);
-
-    cudaSafeCall( cudaStreamAddCallback(stream, cudaStreamCallback, data, 0) );
-#else
-    (void) callback;
-    (void) userData;
-    CV_Error(CV_StsNotImplemented, "This function requires CUDA 5.0");
-#endif
-}
-
-cv::gpu::Stream& cv::gpu::Stream::Null()
-{
-    static Stream s((Impl*) 0);
-    return s;
-}
-
-cv::gpu::Stream::operator bool() const
-{
-    return impl && impl->stream;
-}
-
-cv::gpu::Stream::Stream(Impl* impl_) : impl(impl_)
-{
-}
-
-void cv::gpu::Stream::create()
-{
-    if (impl)
-        release();
-
-    cudaStream_t stream;
-    cudaSafeCall( cudaStreamCreate( &stream ) );
-
-    impl = (Stream::Impl*) fastMalloc(sizeof(Stream::Impl));
-
-    impl->stream = stream;
-    impl->ref_counter = 1;
-}
-
-void cv::gpu::Stream::release()
-{
-    if (impl && CV_XADD(&impl->ref_counter, -1) == 1)
-    {
-        cudaSafeCall( cudaStreamDestroy(impl->stream) );
-        cv::fastFree(impl);
-    }
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/core/src/datastructs.cpp b/modules/core/src/datastructs.cpp
index c927c9e22f..ae2375ad6d 100644
--- a/modules/core/src/datastructs.cpp
+++ b/modules/core/src/datastructs.cpp
@@ -40,6 +40,12 @@
 //M*/
 #include "precomp.hpp"
 
+/* default alignment for dynamic data strucutures, resided in storages. */
+#define  CV_STRUCT_ALIGN    ((int)sizeof(double))
+
+/* default storage block size */
+#define  CV_STORAGE_BLOCK_SIZE   ((1<<16) - 128)
+
 #define ICV_FREE_PTR(storage)  \
     ((schar*)(storage)->top + (storage)->block_size - (storage)->free_space)
 
diff --git a/modules/core/src/drawing.cpp b/modules/core/src/drawing.cpp
index 8850c6ec4c..5cc498256b 100644
--- a/modules/core/src/drawing.cpp
+++ b/modules/core/src/drawing.cpp
@@ -886,12 +886,16 @@ void ellipse2Poly( Point center, Size axes, int angle,
         Point pt;
         pt.x = cvRound( cx + x * alpha - y * beta );
         pt.y = cvRound( cy + x * beta + y * alpha );
-        if( pt != prevPt )
+        if( pt != prevPt ){
             pts.push_back(pt);
+            prevPt = pt;
+        }
     }
 
-    if( pts.size() < 2 )
-        pts.push_back(pts[0]);
+    // If there are no points, it's a zero-size polygon
+    if( pts.size() == 1) {
+        pts.assign(2,center);
+    }
 }
 
 
@@ -1914,7 +1918,7 @@ static const int* getFontData(int fontFace)
 }
 
 
-void putText( Mat& img, const std::string& text, Point org,
+void putText( Mat& img, const String& text, Point org,
               int fontFace, double fontScale, Scalar color,
               int thickness, int line_type, bool bottomLeftOrigin )
 
@@ -1976,7 +1980,7 @@ void putText( Mat& img, const std::string& text, Point org,
     }
 }
 
-Size getTextSize( const std::string& text, int fontFace, double fontScale, int thickness, int* _base_line)
+Size getTextSize( const String& text, int fontFace, double fontScale, int thickness, int* _base_line)
 {
     Size size;
     double view_x = 0;
@@ -2341,7 +2345,7 @@ cvColorToScalar( double packed_color, int type )
         }
         else
         {
-            scalar.val[0] = CV_CAST_8U( icolor );
+            scalar.val[0] = cv::saturate_cast<uchar>( icolor );
             scalar.val[1] = scalar.val[2] = scalar.val[3] = 0;
         }
     }
@@ -2357,7 +2361,7 @@ cvColorToScalar( double packed_color, int type )
         }
         else
         {
-            scalar.val[0] = CV_CAST_8S( icolor );
+            scalar.val[0] = cv::saturate_cast<schar>( icolor );
             scalar.val[1] = scalar.val[2] = scalar.val[3] = 0;
         }
     }
diff --git a/modules/core/src/dxt.cpp b/modules/core/src/dxt.cpp
index b3c2b833b4..a802868df0 100644
--- a/modules/core/src/dxt.cpp
+++ b/modules/core/src/dxt.cpp
@@ -50,6 +50,13 @@ namespace cv
 # pragma warning(disable: 4748)
 #endif
 
+#if defined HAVE_IPP && IPP_VERSION_MAJOR >= 7
+#define USE_IPP_DFT 1
+#else
+#undef USE_IPP_DFT
+#endif
+
+
 /****************************************************************************************\
                                Discrete Fourier Transform
 \****************************************************************************************/
@@ -455,7 +462,7 @@ template<> struct DFT_VecR4<float>
 
 #endif
 
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
 static void ippsDFTFwd_CToC( const Complex<float>* src, Complex<float>* dst,
                              const void* spec, uchar* buf)
 {
@@ -517,7 +524,7 @@ DFT( const Complex<T>* src, Complex<T>* dst, int n,
      int nf, const int* factors, const int* itab,
      const Complex<T>* wave, int tab_size,
      const void*
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
      spec
 #endif
      , Complex<T>* buf,
@@ -537,7 +544,7 @@ DFT( const Complex<T>* src, Complex<T>* dst, int n,
     T scale = (T)_scale;
     int tab_step;
 
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
     if( spec )
     {
         if( !inv )
@@ -957,7 +964,7 @@ DFT( const Complex<T>* src, Complex<T>* dst, int n,
 template<typename T> static void
 RealDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
          const Complex<T>* wave, int tab_size, const void*
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
          spec
 #endif
          ,
@@ -968,11 +975,18 @@ RealDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
     int j, n2 = n >> 1;
     dst += complex_output;
 
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
     if( spec )
     {
         ippsDFTFwd_RToPack( src, dst, spec, (uchar*)buf );
-        goto finalize;
+        if( complex_output )
+        {
+            dst[-1] = dst[0];
+            dst[0] = 0;
+            if( (n & 1) == 0 )
+                dst[n] = 0;
+        }
+        return;
     }
 #endif
     assert( tab_size == n );
@@ -1056,15 +1070,11 @@ RealDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
         }
     }
 
-#ifdef HAVE_IPP
-finalize:
-#endif
     if( complex_output && (n & 1) == 0 )
     {
         dst[-1] = dst[0];
         dst[0] = 0;
-        if( (n & 1) == 0 )
-            dst[n] = 0;
+        dst[n] = 0;
     }
 }
 
@@ -1076,7 +1086,7 @@ template<typename T> static void
 CCSIDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
          const Complex<T>* wave, int tab_size,
          const void*
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
          spec
 #endif
          , Complex<T>* buf,
@@ -1097,7 +1107,7 @@ CCSIDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
         ((T*)src)[1] = src[0];
         src++;
     }
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
     if( spec )
     {
         ippsDFTInv_PackToR( src, dst, spec, (uchar*)buf );
@@ -1225,7 +1235,7 @@ CCSIDFT( const T* src, T* dst, int n, int nf, int* factors, const int* itab,
         }
     }
 
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
 finalize:
 #endif
     if( complex_input )
@@ -1458,6 +1468,10 @@ static void CCSIDFT_64f( const double* src, double* dst, int n, int nf, int* fac
 
 }
 
+#ifdef USE_IPP_DFT
+typedef IppStatus (CV_STDCALL* IppDFTGetSizeFunc)(int, int, IppHintAlgorithm, int*, int*, int*);
+typedef IppStatus (CV_STDCALL* IppDFTInitFunc)(int, int, IppHintAlgorithm, void*, uchar*);
+#endif
 
 void cv::dft( InputArray _src0, OutputArray _dst, int flags, int nonzero_rows )
 {
@@ -1482,8 +1496,8 @@ void cv::dft( InputArray _src0, OutputArray _dst, int flags, int nonzero_rows )
     int elem_size = (int)src.elemSize1(), complex_elem_size = elem_size*2;
     int factors[34];
     bool inplace_transform = false;
-#ifdef HAVE_IPP
-    void *spec_r = 0, *spec_c = 0;
+#ifdef USE_IPP_DFT
+    AutoBuffer<uchar> ippbuf;
     int ipp_norm_flag = !(flags & DFT_SCALE) ? 8 : inv ? 2 : 1;
 #endif
 
@@ -1542,53 +1556,48 @@ void cv::dft( InputArray _src0, OutputArray _dst, int flags, int nonzero_rows )
         }
 
         spec = 0;
-#ifdef HAVE_IPP
+#ifdef USE_IPP_DFT
         if( len*count >= 64 ) // use IPP DFT if available
         {
-            int ipp_sz = 0;
+            int specsize=0, initsize=0, worksize=0;
+            IppDFTGetSizeFunc getSizeFunc = 0;
+            IppDFTInitFunc initFunc = 0;
 
             if( real_transform && stage == 0 )
             {
                 if( depth == CV_32F )
                 {
-                    if( spec_r )
-                        IPPI_CALL( ippsDFTFree_R_32f( (IppsDFTSpec_R_32f*)spec_r ));
-                    IPPI_CALL( ippsDFTInitAlloc_R_32f(
-                        (IppsDFTSpec_R_32f**)&spec_r, len, ipp_norm_flag, ippAlgHintNone ));
-                    IPPI_CALL( ippsDFTGetBufSize_R_32f( (IppsDFTSpec_R_32f*)spec_r, &ipp_sz ));
+                    getSizeFunc = ippsDFTGetSize_R_32f;
+                    initFunc = (IppDFTInitFunc)ippsDFTInit_R_32f;
                 }
                 else
                 {
-                    if( spec_r )
-                        IPPI_CALL( ippsDFTFree_R_64f( (IppsDFTSpec_R_64f*)spec_r ));
-                    IPPI_CALL( ippsDFTInitAlloc_R_64f(
-                        (IppsDFTSpec_R_64f**)&spec_r, len, ipp_norm_flag, ippAlgHintNone ));
-                    IPPI_CALL( ippsDFTGetBufSize_R_64f( (IppsDFTSpec_R_64f*)spec_r, &ipp_sz ));
+                    getSizeFunc = ippsDFTGetSize_R_64f;
+                    initFunc = (IppDFTInitFunc)ippsDFTInit_R_64f;
                 }
-                spec = spec_r;
             }
             else
             {
                 if( depth == CV_32F )
                 {
-                    if( spec_c )
-                        IPPI_CALL( ippsDFTFree_C_32fc( (IppsDFTSpec_C_32fc*)spec_c ));
-                    IPPI_CALL( ippsDFTInitAlloc_C_32fc(
-                        (IppsDFTSpec_C_32fc**)&spec_c, len, ipp_norm_flag, ippAlgHintNone ));
-                    IPPI_CALL( ippsDFTGetBufSize_C_32fc( (IppsDFTSpec_C_32fc*)spec_c, &ipp_sz ));
+                    getSizeFunc = ippsDFTGetSize_C_32fc;
+                    initFunc = (IppDFTInitFunc)ippsDFTInit_C_32fc;
                 }
                 else
                 {
-                    if( spec_c )
-                        IPPI_CALL( ippsDFTFree_C_64fc( (IppsDFTSpec_C_64fc*)spec_c ));
-                    IPPI_CALL( ippsDFTInitAlloc_C_64fc(
-                        (IppsDFTSpec_C_64fc**)&spec_c, len, ipp_norm_flag, ippAlgHintNone ));
-                    IPPI_CALL( ippsDFTGetBufSize_C_64fc( (IppsDFTSpec_C_64fc*)spec_c, &ipp_sz ));
+                    getSizeFunc = ippsDFTGetSize_C_64fc;
+                    initFunc = (IppDFTInitFunc)ippsDFTInit_C_64fc;
                 }
-                spec = spec_c;
             }
-
-            sz += ipp_sz;
+            if( getSizeFunc(len, ipp_norm_flag, ippAlgHintNone, &specsize, &initsize, &worksize) >= 0 )
+            {
+                ippbuf.allocate(specsize + initsize + 64);
+                spec = alignPtr(&ippbuf[0], 32);
+                uchar* initbuf = alignPtr((uchar*)spec + specsize, 32);
+                if( initFunc(len, ipp_norm_flag, ippAlgHintNone, spec, initbuf) < 0 )
+                    spec = 0;
+                sz += worksize;
+            }
         }
         else
 #endif
@@ -1862,24 +1871,6 @@ void cv::dft( InputArray _src0, OutputArray _dst, int flags, int nonzero_rows )
             src = dst;
         }
     }
-
-#ifdef HAVE_IPP
-    if( spec_c )
-    {
-        if( depth == CV_32F )
-            ippsDFTFree_C_32fc( (IppsDFTSpec_C_32fc*)spec_c );
-        else
-            ippsDFTFree_C_64fc( (IppsDFTSpec_C_64fc*)spec_c );
-    }
-
-    if( spec_r )
-    {
-        if( depth == CV_32F )
-            ippsDFTFree_R_32f( (IppsDFTSpec_R_32f*)spec_r );
-        else
-            ippsDFTFree_R_64f( (IppsDFTSpec_R_64f*)spec_r );
-    }
-#endif
 }
 
 
diff --git a/modules/core/src/gl_core_3_1.cpp b/modules/core/src/gl_core_3_1.cpp
index 61006f5cac..48201b4b76 100644
--- a/modules/core/src/gl_core_3_1.cpp
+++ b/modules/core/src/gl_core_3_1.cpp
@@ -1,7 +1,46 @@
-#include <string>
-#include <sstream>
-#include "cvconfig.h"
-#include "opencv2/core.hpp"
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
 #include "gl_core_3_1.hpp"
 
 #ifdef HAVE_OPENGL
@@ -15,8 +54,8 @@
                 image = NSAddImage("/System/Library/Frameworks/OpenGL.framework/Versions/Current/OpenGL", NSADDIMAGE_OPTION_RETURN_ON_ERROR);
 
             // prepend a '_' for the Unix C symbol mangling convention
-            std::string symbolName = "_";
-            symbolName += std::string(name);
+            String symbolName = "_";
+            symbolName += String(name);
 
             NSSymbol symbol = image ? NSLookupSymbolInImage(image, &symbolName[0], NSLOOKUPSYMBOLINIMAGE_OPTION_BIND | NSLOOKUPSYMBOLINIMAGE_OPTION_RETURN_ON_ERROR) : 0;
 
@@ -94,16 +133,14 @@
         void* func =  (void*) CV_GL_GET_PROC_ADDRESS(name);
         if (!func)
         {
-            std::ostringstream msg;
-            msg << "Can't load OpenGL extension [" << name << "]";
-            CV_Error(CV_OpenGlApiCallError, msg.str());
+            CV_Error(cv::Error::OpenGlApiCallError, cv::format("Can't load OpenGL extension [%s]", name) );
         }
         return func;
     }
 #else
     static void* IntGetProcAddress(const char*)
     {
-        CV_Error(CV_OpenGlNotSupported, "The library is compiled without OpenGL support");
+        CV_Error(cv::Error::OpenGlNotSupported, "The library is compiled without OpenGL support");
         return 0;
     }
 #endif
diff --git a/modules/core/src/gl_core_3_1.hpp b/modules/core/src/gl_core_3_1.hpp
index 50dbee66c7..f86390169e 100644
--- a/modules/core/src/gl_core_3_1.hpp
+++ b/modules/core/src/gl_core_3_1.hpp
@@ -1,3 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifndef OPENGL_NOLOAD_STYLE_HPP
 #define OPENGL_NOLOAD_STYLE_HPP
 
diff --git a/modules/core/src/glob.cpp b/modules/core/src/glob.cpp
index 4659c21d89..c75bd2e66b 100644
--- a/modules/core/src/glob.cpp
+++ b/modules/core/src/glob.cpp
@@ -56,19 +56,42 @@ namespace
 
     struct DIR
     {
+#ifdef HAVE_WINRT
+        WIN32_FIND_DATAW data;
+#else
         WIN32_FIND_DATA data;
+#endif
         HANDLE handle;
         dirent ent;
+#ifdef HAVE_WINRT
+        DIR() {};
+        ~DIR()
+        {
+            if (ent.d_name)
+                delete[] ent.d_name;
+        }
+#endif
     };
 
     DIR* opendir(const char* path)
     {
         DIR* dir = new DIR;
         dir->ent.d_name = 0;
-        dir->handle = ::FindFirstFileA((std::string(path) + "\\*").c_str(), &dir->data);
+#ifdef HAVE_WINRT
+        cv::String full_path = cv::String(path) + "\\*";
+        wchar_t wfull_path[MAX_PATH];
+        size_t copied = mbstowcs(wfull_path, full_path.c_str(), MAX_PATH);
+        CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+        dir->handle = ::FindFirstFileExW(wfull_path, FindExInfoStandard,
+                        &dir->data, FindExSearchNameMatch, NULL, 0);
+#else
+        dir->handle = ::FindFirstFileExA((cv::String(path) + "\\*").c_str(),
+            FindExInfoStandard, &dir->data, FindExSearchNameMatch, NULL, 0);
+#endif
         if(dir->handle == INVALID_HANDLE_VALUE)
         {
             /*closedir will do all cleanup*/
+            delete dir;
             return 0;
         }
         return dir;
@@ -76,12 +99,26 @@ namespace
 
     dirent* readdir(DIR* dir)
     {
+#ifdef HAVE_WINRT
         if (dir->ent.d_name != 0)
         {
-            if (::FindNextFile(dir->handle, &dir->data) != TRUE)
+            if (::FindNextFileW(dir->handle, &dir->data) != TRUE)
+                return 0;
+        }
+        size_t asize = wcstombs(NULL, dir->data.cFileName, 0);
+        CV_Assert((asize != 0) && (asize != (size_t)-1));
+        char* aname = new char[asize+1];
+        aname[asize] = 0;
+        wcstombs(aname, dir->data.cFileName, asize);
+        dir->ent.d_name = aname;
+#else
+        if (dir->ent.d_name != 0)
+        {
+            if (::FindNextFileA(dir->handle, &dir->data) != TRUE)
                 return 0;
         }
         dir->ent.d_name = dir->data.cFileName;
+#endif
         return &dir->ent;
     }
 
@@ -100,22 +137,34 @@ const char dir_separators[] = "/";
 const char native_separator = '/';
 #endif
 
-static bool isDir(const std::string& path, DIR* dir)
+static bool isDir(const cv::String& path, DIR* dir)
 {
 #if defined WIN32 || defined _WIN32 || defined WINCE
     DWORD attributes;
+    BOOL status = TRUE;
     if (dir)
         attributes = dir->data.dwFileAttributes;
     else
-        attributes = ::GetFileAttributes(path.c_str());
-
-    return (attributes != INVALID_FILE_ATTRIBUTES) && ((attributes & FILE_ATTRIBUTE_DIRECTORY) != 0);
+    {
+        WIN32_FILE_ATTRIBUTE_DATA all_attrs;
+#ifdef HAVE_WINRT
+        wchar_t wpath[MAX_PATH];
+        size_t copied = mbstowcs(wpath, path.c_str(), MAX_PATH);
+        CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+        status = ::GetFileAttributesExW(wpath, GetFileExInfoStandard, &all_attrs);
 #else
-    struct stat stat_buf;
-    stat( path.c_str(), &stat_buf);
-    int is_dir = S_ISDIR( stat_buf.st_mode);
-    (void)dir;
+        status = ::GetFileAttributesExA(path.c_str(), GetFileExInfoStandard, &all_attrs);
+#endif
+        attributes = all_attrs.dwFileAttributes;
+    }
 
+    return status && ((attributes & FILE_ATTRIBUTE_DIRECTORY) != 0);
+#else
+    (void)dir;
+    struct stat stat_buf;
+    if (0 != stat( path.c_str(), &stat_buf))
+        return false;
+    int is_dir = S_ISDIR( stat_buf.st_mode);
     return is_dir != 0;
 #endif
 }
@@ -168,10 +217,11 @@ static bool wildcmp(const char *string, const char *wild)
     return *wild == 0;
 }
 
-static void glob_rec(const std::string& directory, const std::string& wildchart, std::vector<std::string>& result, bool recursive)
+static void glob_rec(const cv::String& directory, const cv::String& wildchart, std::vector<cv::String>& result, bool recursive)
 {
     DIR *dir;
     struct dirent *ent;
+
     if ((dir = opendir (directory.c_str())) != 0)
     {
         /* find all the files and directories within directory */
@@ -183,7 +233,7 @@ static void glob_rec(const std::string& directory, const std::string& wildchart,
                 if((name[0] == 0) || (name[0] == '.' && name[1] == 0) || (name[0] == '.' && name[1] == '.' && name[2] == 0))
                     continue;
 
-                std::string path = directory + native_separator + name;
+                cv::String path = directory + native_separator + name;
 
                 if (isDir(path, dir))
                 {
@@ -207,14 +257,13 @@ static void glob_rec(const std::string& directory, const std::string& wildchart,
     else CV_Error(CV_StsObjectNotFound, cv::format("could not open directory: %s", directory.c_str()));
 }
 
-void cv::glob(std::string pattern, std::vector<std::string>& result, bool recursive)
+void cv::glob(String pattern, std::vector<String>& result, bool recursive)
 {
     result.clear();
-    std::string path, wildchart;
+    String path, wildchart;
 
     if (isDir(pattern, 0))
     {
-        printf("WE ARE HERE: %s\n", pattern.c_str());
         if(strchr(dir_separators, pattern[pattern.size() - 1]) != 0)
         {
             path = pattern.substr(0, pattern.size() - 1);
@@ -227,7 +276,7 @@ void cv::glob(std::string pattern, std::vector<std::string>& result, bool recurs
     else
     {
         size_t pos = pattern.find_last_of(dir_separators);
-        if (pos == std::string::npos)
+        if (pos == String::npos)
         {
             wildchart = pattern;
             path = ".";
@@ -241,4 +290,4 @@ void cv::glob(std::string pattern, std::vector<std::string>& result, bool recurs
 
     glob_rec(path, wildchart, result, recursive);
     std::sort(result.begin(), result.end());
-}
\ No newline at end of file
+}
diff --git a/modules/core/src/gpu_cuda_mem.cpp b/modules/core/src/gpu_cuda_mem.cpp
new file mode 100644
index 0000000000..52de069cbc
--- /dev/null
+++ b/modules/core/src/gpu_cuda_mem.cpp
@@ -0,0 +1,215 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+namespace
+{
+    size_t alignUpStep(size_t what, size_t alignment)
+    {
+        size_t alignMask = alignment - 1;
+        size_t inverseAlignMask = ~alignMask;
+        size_t res = (what + alignMask) & inverseAlignMask;
+        return res;
+    }
+}
+
+void cv::gpu::CudaMem::create(int rows_, int cols_, int type_)
+{
+#ifndef HAVE_CUDA
+    (void) rows_;
+    (void) cols_;
+    (void) type_;
+    throw_no_cuda();
+#else
+    if (alloc_type == SHARED)
+    {
+        DeviceInfo devInfo;
+        CV_Assert( devInfo.canMapHostMemory() );
+    }
+
+    type_ &= Mat::TYPE_MASK;
+
+    if (rows == rows_ && cols == cols_ && type() == type_ && data)
+        return;
+
+    if (data)
+        release();
+
+    CV_DbgAssert( rows_ >= 0 && cols_ >= 0 );
+
+    if (rows_ > 0 && cols_ > 0)
+    {
+        flags = Mat::MAGIC_VAL + Mat::CONTINUOUS_FLAG + type_;
+        rows = rows_;
+        cols = cols_;
+        step = elemSize() * cols;
+
+        if (alloc_type == SHARED)
+        {
+            DeviceInfo devInfo;
+            step = alignUpStep(step, devInfo.textureAlignment());
+        }
+
+        int64 _nettosize = (int64)step*rows;
+        size_t nettosize = (size_t)_nettosize;
+
+        if (_nettosize != (int64)nettosize)
+            CV_Error(cv::Error::StsNoMem, "Too big buffer is allocated");
+
+        size_t datasize = alignSize(nettosize, (int)sizeof(*refcount));
+
+        void* ptr = 0;
+
+        switch (alloc_type)
+        {
+        case PAGE_LOCKED:    cudaSafeCall( cudaHostAlloc(&ptr, datasize, cudaHostAllocDefault) ); break;
+        case SHARED:         cudaSafeCall( cudaHostAlloc(&ptr, datasize, cudaHostAllocMapped) );  break;
+        case WRITE_COMBINED: cudaSafeCall( cudaHostAlloc(&ptr, datasize, cudaHostAllocWriteCombined) ); break;
+        default:             CV_Error(cv::Error::StsBadFlag, "Invalid alloc type");
+        }
+
+        datastart = data =  (uchar*)ptr;
+        dataend = data + nettosize;
+
+        refcount = (int*)cv::fastMalloc(sizeof(*refcount));
+        *refcount = 1;
+    }
+#endif
+}
+
+CudaMem cv::gpu::CudaMem::reshape(int new_cn, int new_rows) const
+{
+    CudaMem hdr = *this;
+
+    int cn = channels();
+    if (new_cn == 0)
+        new_cn = cn;
+
+    int total_width = cols * cn;
+
+    if ((new_cn > total_width || total_width % new_cn != 0) && new_rows == 0)
+        new_rows = rows * total_width / new_cn;
+
+    if (new_rows != 0 && new_rows != rows)
+    {
+        int total_size = total_width * rows;
+
+        if (!isContinuous())
+            CV_Error(cv::Error::BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
+
+        if ((unsigned)new_rows > (unsigned)total_size)
+            CV_Error(cv::Error::StsOutOfRange, "Bad new number of rows");
+
+        total_width = total_size / new_rows;
+
+        if (total_width * new_rows != total_size)
+            CV_Error(cv::Error::StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
+
+        hdr.rows = new_rows;
+        hdr.step = total_width * elemSize1();
+    }
+
+    int new_width = total_width / new_cn;
+
+    if (new_width * new_cn != total_width)
+        CV_Error(cv::Error::BadNumChannels, "The total width is not divisible by the new number of channels");
+
+    hdr.cols = new_width;
+    hdr.flags = (hdr.flags & ~CV_MAT_CN_MASK) | ((new_cn - 1) << CV_CN_SHIFT);
+
+    return hdr;
+}
+
+void cv::gpu::CudaMem::release()
+{
+#ifdef HAVE_CUDA
+    if (refcount && CV_XADD(refcount, -1) == 1)
+    {
+        cudaFreeHost(datastart);
+        fastFree(refcount);
+    }
+
+    data = datastart = dataend = 0;
+    step = rows = cols = 0;
+    refcount = 0;
+#endif
+}
+
+GpuMat cv::gpu::CudaMem::createGpuMatHeader() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return GpuMat();
+#else
+    CV_Assert( alloc_type == SHARED );
+
+    void *pdev;
+    cudaSafeCall( cudaHostGetDevicePointer(&pdev, data, 0) );
+
+    return GpuMat(rows, cols, type(), pdev, step);
+#endif
+}
+
+void cv::gpu::registerPageLocked(Mat& m)
+{
+#ifndef HAVE_CUDA
+    (void) m;
+    throw_no_cuda();
+#else
+    CV_Assert( m.isContinuous() );
+    cudaSafeCall( cudaHostRegister(m.data, m.step * m.rows, cudaHostRegisterPortable) );
+#endif
+}
+
+void cv::gpu::unregisterPageLocked(Mat& m)
+{
+#ifndef HAVE_CUDA
+    (void) m;
+#else
+    cudaSafeCall( cudaHostUnregister(m.data) );
+#endif
+}
diff --git a/modules/core/src/gpu_info.cpp b/modules/core/src/gpu_info.cpp
new file mode 100644
index 0000000000..5a1e567463
--- /dev/null
+++ b/modules/core/src/gpu_info.cpp
@@ -0,0 +1,1272 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+int cv::gpu::getCudaEnabledDeviceCount()
+{
+#ifndef HAVE_CUDA
+    return 0;
+#else
+    int count;
+    cudaError_t error = cudaGetDeviceCount(&count);
+
+    if (error == cudaErrorInsufficientDriver)
+        return -1;
+
+    if (error == cudaErrorNoDevice)
+        return 0;
+
+    cudaSafeCall( error );
+    return count;
+#endif
+}
+
+void cv::gpu::setDevice(int device)
+{
+#ifndef HAVE_CUDA
+    (void) device;
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaSetDevice(device) );
+#endif
+}
+
+int cv::gpu::getDevice()
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    int device;
+    cudaSafeCall( cudaGetDevice(&device) );
+    return device;
+#endif
+}
+
+void cv::gpu::resetDevice()
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaDeviceReset() );
+#endif
+}
+
+bool cv::gpu::deviceSupports(FeatureSet feature_set)
+{
+#ifndef HAVE_CUDA
+    (void) feature_set;
+    throw_no_cuda();
+    return false;
+#else
+    static int versions[] =
+    {
+        -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
+    };
+    static const int cache_size = static_cast<int>(sizeof(versions) / sizeof(versions[0]));
+
+    const int devId = getDevice();
+
+    int version;
+
+    if (devId < cache_size && versions[devId] >= 0)
+    {
+        version = versions[devId];
+    }
+    else
+    {
+        DeviceInfo dev(devId);
+        version = dev.majorVersion() * 10 + dev.minorVersion();
+        if (devId < cache_size)
+            versions[devId] = version;
+    }
+
+    return TargetArchs::builtWith(feature_set) && (version >= feature_set);
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////
+// TargetArchs
+
+#ifdef HAVE_CUDA
+
+namespace
+{
+    class CudaArch
+    {
+    public:
+        CudaArch();
+
+        bool builtWith(FeatureSet feature_set) const;
+        bool hasPtx(int major, int minor) const;
+        bool hasBin(int major, int minor) const;
+        bool hasEqualOrLessPtx(int major, int minor) const;
+        bool hasEqualOrGreaterPtx(int major, int minor) const;
+        bool hasEqualOrGreaterBin(int major, int minor) const;
+
+    private:
+        static void fromStr(const char* set_as_str, std::vector<int>& arr);
+
+        std::vector<int> bin;
+        std::vector<int> ptx;
+        std::vector<int> features;
+    };
+
+    const CudaArch cudaArch;
+
+    CudaArch::CudaArch()
+    {
+        fromStr(CUDA_ARCH_BIN, bin);
+        fromStr(CUDA_ARCH_PTX, ptx);
+        fromStr(CUDA_ARCH_FEATURES, features);
+    }
+
+    bool CudaArch::builtWith(FeatureSet feature_set) const
+    {
+        return !features.empty() && (features.back() >= feature_set);
+    }
+
+    bool CudaArch::hasPtx(int major, int minor) const
+    {
+        return std::find(ptx.begin(), ptx.end(), major * 10 + minor) != ptx.end();
+    }
+
+    bool CudaArch::hasBin(int major, int minor) const
+    {
+        return std::find(bin.begin(), bin.end(), major * 10 + minor) != bin.end();
+    }
+
+    bool CudaArch::hasEqualOrLessPtx(int major, int minor) const
+    {
+        return !ptx.empty() && (ptx.front() <= major * 10 + minor);
+    }
+
+    bool CudaArch::hasEqualOrGreaterPtx(int major, int minor) const
+    {
+        return !ptx.empty() && (ptx.back() >= major * 10 + minor);
+    }
+
+    bool CudaArch::hasEqualOrGreaterBin(int major, int minor) const
+    {
+        return !bin.empty() && (bin.back() >= major * 10 + minor);
+    }
+
+    void CudaArch::fromStr(const char* set_as_str, std::vector<int>& arr)
+    {
+        arr.clear();
+
+        const size_t len = strlen(set_as_str);
+
+        size_t pos = 0;
+        while (pos < len)
+        {
+            if (isspace(set_as_str[pos]))
+            {
+                ++pos;
+            }
+            else
+            {
+                int cur_value;
+                int chars_read;
+                int args_read = sscanf(set_as_str + pos, "%d%n", &cur_value, &chars_read);
+                CV_Assert( args_read == 1 );
+
+                arr.push_back(cur_value);
+                pos += chars_read;
+            }
+        }
+
+        std::sort(arr.begin(), arr.end());
+    }
+}
+
+#endif
+
+bool cv::gpu::TargetArchs::builtWith(cv::gpu::FeatureSet feature_set)
+{
+#ifndef HAVE_CUDA
+    (void) feature_set;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.builtWith(feature_set);
+#endif
+}
+
+bool cv::gpu::TargetArchs::hasPtx(int major, int minor)
+{
+#ifndef HAVE_CUDA
+    (void) major;
+    (void) minor;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.hasPtx(major, minor);
+#endif
+}
+
+bool cv::gpu::TargetArchs::hasBin(int major, int minor)
+{
+#ifndef HAVE_CUDA
+    (void) major;
+    (void) minor;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.hasBin(major, minor);
+#endif
+}
+
+bool cv::gpu::TargetArchs::hasEqualOrLessPtx(int major, int minor)
+{
+#ifndef HAVE_CUDA
+    (void) major;
+    (void) minor;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.hasEqualOrLessPtx(major, minor);
+#endif
+}
+
+bool cv::gpu::TargetArchs::hasEqualOrGreaterPtx(int major, int minor)
+{
+#ifndef HAVE_CUDA
+    (void) major;
+    (void) minor;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.hasEqualOrGreaterPtx(major, minor);
+#endif
+}
+
+bool cv::gpu::TargetArchs::hasEqualOrGreaterBin(int major, int minor)
+{
+#ifndef HAVE_CUDA
+    (void) major;
+    (void) minor;
+    throw_no_cuda();
+    return false;
+#else
+    return cudaArch.hasEqualOrGreaterBin(major, minor);
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////
+// DeviceInfo
+
+#ifdef HAVE_CUDA
+
+namespace
+{
+    class DeviceProps
+    {
+    public:
+        DeviceProps();
+
+        const cudaDeviceProp* get(int devID) const;
+
+    private:
+        std::vector<cudaDeviceProp> props_;
+    };
+
+    DeviceProps::DeviceProps()
+    {
+        int count = getCudaEnabledDeviceCount();
+
+        if (count > 0)
+        {
+            props_.resize(count);
+
+            for (int devID = 0; devID < count; ++devID)
+            {
+                cudaSafeCall( cudaGetDeviceProperties(&props_[devID], devID) );
+            }
+        }
+    }
+
+    const cudaDeviceProp* DeviceProps::get(int devID) const
+    {
+        CV_Assert( static_cast<size_t>(devID) < props_.size() );
+
+        return &props_[devID];
+    }
+
+    DeviceProps& deviceProps()
+    {
+        static DeviceProps props;
+        return props;
+    }
+}
+
+#endif
+
+const char* cv::gpu::DeviceInfo::name() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return "";
+#else
+    return deviceProps().get(device_id_)->name;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::totalGlobalMem() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->totalGlobalMem;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::sharedMemPerBlock() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->sharedMemPerBlock;
+#endif
+}
+
+int cv::gpu::DeviceInfo::regsPerBlock() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->regsPerBlock;
+#endif
+}
+
+int cv::gpu::DeviceInfo::warpSize() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->warpSize;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::memPitch() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->memPitch;
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxThreadsPerBlock() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxThreadsPerBlock;
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxThreadsDim() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxThreadsDim);
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxGridSize() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxGridSize);
+#endif
+}
+
+int cv::gpu::DeviceInfo::clockRate() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->clockRate;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::totalConstMem() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->totalConstMem;
+#endif
+}
+
+int cv::gpu::DeviceInfo::majorVersion() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->major;
+#endif
+}
+
+int cv::gpu::DeviceInfo::minorVersion() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->minor;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::textureAlignment() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->textureAlignment;
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::texturePitchAlignment() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->texturePitchAlignment;
+#endif
+}
+
+int cv::gpu::DeviceInfo::multiProcessorCount() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->multiProcessorCount;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::kernelExecTimeoutEnabled() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->kernelExecTimeoutEnabled != 0;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::integrated() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->integrated != 0;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::canMapHostMemory() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->canMapHostMemory != 0;
+#endif
+}
+
+DeviceInfo::ComputeMode cv::gpu::DeviceInfo::computeMode() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return ComputeModeDefault;
+#else
+    static const ComputeMode tbl[] =
+    {
+        ComputeModeDefault,
+        ComputeModeExclusive,
+        ComputeModeProhibited,
+        ComputeModeExclusiveProcess
+    };
+
+    return tbl[deviceProps().get(device_id_)->computeMode];
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxTexture1D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxTexture1D;
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxTexture1DMipmap() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    #if CUDA_VERSION >= 5000
+        return deviceProps().get(device_id_)->maxTexture1DMipmap;
+    #else
+        CV_Error(Error::StsNotImplemented, "This function requires CUDA 5.0");
+        return 0;
+    #endif
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxTexture1DLinear() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxTexture1DLinear;
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxTexture2D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxTexture2D);
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxTexture2DMipmap() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    #if CUDA_VERSION >= 5000
+        return Vec2i(deviceProps().get(device_id_)->maxTexture2DMipmap);
+    #else
+        CV_Error(Error::StsNotImplemented, "This function requires CUDA 5.0");
+        return Vec2i();
+    #endif
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxTexture2DLinear() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxTexture2DLinear);
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxTexture2DGather() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxTexture2DGather);
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxTexture3D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxTexture3D);
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxTextureCubemap() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxTextureCubemap;
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxTexture1DLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxTexture1DLayered);
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxTexture2DLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxTexture2DLayered);
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxTextureCubemapLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxTextureCubemapLayered);
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxSurface1D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxSurface1D;
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxSurface2D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxSurface2D);
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxSurface3D() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxSurface3D);
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxSurface1DLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxSurface1DLayered);
+#endif
+}
+
+Vec3i cv::gpu::DeviceInfo::maxSurface2DLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec3i();
+#else
+    return Vec3i(deviceProps().get(device_id_)->maxSurface2DLayered);
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxSurfaceCubemap() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxSurfaceCubemap;
+#endif
+}
+
+Vec2i cv::gpu::DeviceInfo::maxSurfaceCubemapLayered() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return Vec2i();
+#else
+    return Vec2i(deviceProps().get(device_id_)->maxSurfaceCubemapLayered);
+#endif
+}
+
+size_t cv::gpu::DeviceInfo::surfaceAlignment() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->surfaceAlignment;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::concurrentKernels() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->concurrentKernels != 0;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::ECCEnabled() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->ECCEnabled != 0;
+#endif
+}
+
+int cv::gpu::DeviceInfo::pciBusID() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->pciBusID;
+#endif
+}
+
+int cv::gpu::DeviceInfo::pciDeviceID() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->pciDeviceID;
+#endif
+}
+
+int cv::gpu::DeviceInfo::pciDomainID() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->pciDomainID;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::tccDriver() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->tccDriver != 0;
+#endif
+}
+
+int cv::gpu::DeviceInfo::asyncEngineCount() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->asyncEngineCount;
+#endif
+}
+
+bool cv::gpu::DeviceInfo::unifiedAddressing() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    return deviceProps().get(device_id_)->unifiedAddressing != 0;
+#endif
+}
+
+int cv::gpu::DeviceInfo::memoryClockRate() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->memoryClockRate;
+#endif
+}
+
+int cv::gpu::DeviceInfo::memoryBusWidth() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->memoryBusWidth;
+#endif
+}
+
+int cv::gpu::DeviceInfo::l2CacheSize() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->l2CacheSize;
+#endif
+}
+
+int cv::gpu::DeviceInfo::maxThreadsPerMultiProcessor() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return 0;
+#else
+    return deviceProps().get(device_id_)->maxThreadsPerMultiProcessor;
+#endif
+}
+
+void cv::gpu::DeviceInfo::queryMemory(size_t& _totalMemory, size_t& _freeMemory) const
+{
+#ifndef HAVE_CUDA
+    (void) _totalMemory;
+    (void) _freeMemory;
+    throw_no_cuda();
+#else
+    int prevDeviceID = getDevice();
+    if (prevDeviceID != device_id_)
+        setDevice(device_id_);
+
+    cudaSafeCall( cudaMemGetInfo(&_freeMemory, &_totalMemory) );
+
+    if (prevDeviceID != device_id_)
+        setDevice(prevDeviceID);
+#endif
+}
+
+bool cv::gpu::DeviceInfo::isCompatible() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    // Check PTX compatibility
+    if (TargetArchs::hasEqualOrLessPtx(majorVersion(), minorVersion()))
+        return true;
+
+    // Check BIN compatibility
+    for (int i = minorVersion(); i >= 0; --i)
+        if (TargetArchs::hasBin(majorVersion(), i))
+            return true;
+
+    return false;
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////
+// print info
+
+#ifdef HAVE_CUDA
+
+namespace
+{
+    int convertSMVer2Cores(int major, int minor)
+    {
+        // Defines for GPU Architecture types (using the SM version to determine the # of cores per SM
+        typedef struct {
+            int SM; // 0xMm (hexidecimal notation), M = SM Major version, and m = SM minor version
+            int Cores;
+        } SMtoCores;
+
+        SMtoCores gpuArchCoresPerSM[] =  { { 0x10,  8 }, { 0x11,  8 }, { 0x12,  8 }, { 0x13,  8 }, { 0x20, 32 }, { 0x21, 48 }, {0x30, 192}, {0x35, 192}, { -1, -1 }  };
+
+        int index = 0;
+        while (gpuArchCoresPerSM[index].SM != -1)
+        {
+            if (gpuArchCoresPerSM[index].SM == ((major << 4) + minor) )
+                return gpuArchCoresPerSM[index].Cores;
+            index++;
+        }
+
+        return -1;
+    }
+}
+
+#endif
+
+void cv::gpu::printCudaDeviceInfo(int device)
+{
+#ifndef HAVE_CUDA
+    (void) device;
+    throw_no_cuda();
+#else
+    int count = getCudaEnabledDeviceCount();
+    bool valid = (device >= 0) && (device < count);
+
+    int beg = valid ? device   : 0;
+    int end = valid ? device+1 : count;
+
+    printf("*** CUDA Device Query (Runtime API) version (CUDART static linking) *** \n\n");
+    printf("Device count: %d\n", count);
+
+    int driverVersion = 0, runtimeVersion = 0;
+    cudaSafeCall( cudaDriverGetVersion(&driverVersion) );
+    cudaSafeCall( cudaRuntimeGetVersion(&runtimeVersion) );
+
+    const char *computeMode[] = {
+        "Default (multiple host threads can use ::cudaSetDevice() with device simultaneously)",
+        "Exclusive (only one host thread in one process is able to use ::cudaSetDevice() with this device)",
+        "Prohibited (no host thread can use ::cudaSetDevice() with this device)",
+        "Exclusive Process (many threads in one process is able to use ::cudaSetDevice() with this device)",
+        "Unknown",
+        NULL
+    };
+
+    for(int dev = beg; dev < end; ++dev)
+    {
+        cudaDeviceProp prop;
+        cudaSafeCall( cudaGetDeviceProperties(&prop, dev) );
+
+        printf("\nDevice %d: \"%s\"\n", dev, prop.name);
+        printf("  CUDA Driver Version / Runtime Version          %d.%d / %d.%d\n", driverVersion/1000, driverVersion%100, runtimeVersion/1000, runtimeVersion%100);
+        printf("  CUDA Capability Major/Minor version number:    %d.%d\n", prop.major, prop.minor);
+        printf("  Total amount of global memory:                 %.0f MBytes (%llu bytes)\n", (float)prop.totalGlobalMem/1048576.0f, (unsigned long long) prop.totalGlobalMem);
+
+        int cores = convertSMVer2Cores(prop.major, prop.minor);
+        if (cores > 0)
+            printf("  (%2d) Multiprocessors x (%2d) CUDA Cores/MP:     %d CUDA Cores\n", prop.multiProcessorCount, cores, cores * prop.multiProcessorCount);
+
+        printf("  GPU Clock Speed:                               %.2f GHz\n", prop.clockRate * 1e-6f);
+
+        printf("  Max Texture Dimension Size (x,y,z)             1D=(%d), 2D=(%d,%d), 3D=(%d,%d,%d)\n",
+            prop.maxTexture1D, prop.maxTexture2D[0], prop.maxTexture2D[1],
+            prop.maxTexture3D[0], prop.maxTexture3D[1], prop.maxTexture3D[2]);
+        printf("  Max Layered Texture Size (dim) x layers        1D=(%d) x %d, 2D=(%d,%d) x %d\n",
+            prop.maxTexture1DLayered[0], prop.maxTexture1DLayered[1],
+            prop.maxTexture2DLayered[0], prop.maxTexture2DLayered[1], prop.maxTexture2DLayered[2]);
+
+        printf("  Total amount of constant memory:               %u bytes\n", (int)prop.totalConstMem);
+        printf("  Total amount of shared memory per block:       %u bytes\n", (int)prop.sharedMemPerBlock);
+        printf("  Total number of registers available per block: %d\n", prop.regsPerBlock);
+        printf("  Warp size:                                     %d\n", prop.warpSize);
+        printf("  Maximum number of threads per block:           %d\n", prop.maxThreadsPerBlock);
+        printf("  Maximum sizes of each dimension of a block:    %d x %d x %d\n", prop.maxThreadsDim[0], prop.maxThreadsDim[1], prop.maxThreadsDim[2]);
+        printf("  Maximum sizes of each dimension of a grid:     %d x %d x %d\n", prop.maxGridSize[0], prop.maxGridSize[1],  prop.maxGridSize[2]);
+        printf("  Maximum memory pitch:                          %u bytes\n", (int)prop.memPitch);
+        printf("  Texture alignment:                             %u bytes\n", (int)prop.textureAlignment);
+
+        printf("  Concurrent copy and execution:                 %s with %d copy engine(s)\n", (prop.deviceOverlap ? "Yes" : "No"), prop.asyncEngineCount);
+        printf("  Run time limit on kernels:                     %s\n", prop.kernelExecTimeoutEnabled ? "Yes" : "No");
+        printf("  Integrated GPU sharing Host Memory:            %s\n", prop.integrated ? "Yes" : "No");
+        printf("  Support host page-locked memory mapping:       %s\n", prop.canMapHostMemory ? "Yes" : "No");
+
+        printf("  Concurrent kernel execution:                   %s\n", prop.concurrentKernels ? "Yes" : "No");
+        printf("  Alignment requirement for Surfaces:            %s\n", prop.surfaceAlignment ? "Yes" : "No");
+        printf("  Device has ECC support enabled:                %s\n", prop.ECCEnabled ? "Yes" : "No");
+        printf("  Device is using TCC driver mode:               %s\n", prop.tccDriver ? "Yes" : "No");
+        printf("  Device supports Unified Addressing (UVA):      %s\n", prop.unifiedAddressing ? "Yes" : "No");
+        printf("  Device PCI Bus ID / PCI location ID:           %d / %d\n", prop.pciBusID, prop.pciDeviceID );
+        printf("  Compute Mode:\n");
+        printf("      %s \n", computeMode[prop.computeMode]);
+    }
+
+    printf("\n");
+    printf("deviceQuery, CUDA Driver = CUDART");
+    printf(", CUDA Driver Version  = %d.%d", driverVersion / 1000, driverVersion % 100);
+    printf(", CUDA Runtime Version = %d.%d", runtimeVersion/1000, runtimeVersion%100);
+    printf(", NumDevs = %d\n\n", count);
+
+    fflush(stdout);
+#endif
+}
+
+void cv::gpu::printShortCudaDeviceInfo(int device)
+{
+#ifndef HAVE_CUDA
+    (void) device;
+    throw_no_cuda();
+#else
+    int count = getCudaEnabledDeviceCount();
+    bool valid = (device >= 0) && (device < count);
+
+    int beg = valid ? device   : 0;
+    int end = valid ? device+1 : count;
+
+    int driverVersion = 0, runtimeVersion = 0;
+    cudaSafeCall( cudaDriverGetVersion(&driverVersion) );
+    cudaSafeCall( cudaRuntimeGetVersion(&runtimeVersion) );
+
+    for(int dev = beg; dev < end; ++dev)
+    {
+        cudaDeviceProp prop;
+        cudaSafeCall( cudaGetDeviceProperties(&prop, dev) );
+
+        const char *arch_str = prop.major < 2 ? " (not Fermi)" : "";
+        printf("Device %d:  \"%s\"  %.0fMb", dev, prop.name, (float)prop.totalGlobalMem/1048576.0f);
+        printf(", sm_%d%d%s", prop.major, prop.minor, arch_str);
+
+        int cores = convertSMVer2Cores(prop.major, prop.minor);
+        if (cores > 0)
+            printf(", %d cores", cores * prop.multiProcessorCount);
+
+        printf(", Driver/Runtime ver.%d.%d/%d.%d\n", driverVersion/1000, driverVersion%100, runtimeVersion/1000, runtimeVersion%100);
+    }
+
+    fflush(stdout);
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////
+// Error handling
+
+#ifdef HAVE_CUDA
+
+namespace
+{
+    #define error_entry(entry)  { entry, #entry }
+
+    struct ErrorEntry
+    {
+        int code;
+        const char* str;
+    };
+
+    struct ErrorEntryComparer
+    {
+        int code;
+        ErrorEntryComparer(int code_) : code(code_) {}
+        bool operator()(const ErrorEntry& e) const { return e.code == code; }
+    };
+
+    const ErrorEntry npp_errors [] =
+    {
+    #if defined (_MSC_VER)
+        error_entry( NPP_NOT_SUFFICIENT_COMPUTE_CAPABILITY ),
+    #endif
+
+    #if NPP_VERSION < 5500
+        error_entry( NPP_BAD_ARG_ERROR ),
+        error_entry( NPP_COEFF_ERROR ),
+        error_entry( NPP_RECT_ERROR ),
+        error_entry( NPP_QUAD_ERROR ),
+        error_entry( NPP_MEMFREE_ERR ),
+        error_entry( NPP_MEMSET_ERR ),
+        error_entry( NPP_MEM_ALLOC_ERR ),
+        error_entry( NPP_HISTO_NUMBER_OF_LEVELS_ERROR ),
+        error_entry( NPP_MIRROR_FLIP_ERR ),
+        error_entry( NPP_INVALID_INPUT ),
+        error_entry( NPP_POINTER_ERROR ),
+        error_entry( NPP_WARNING ),
+        error_entry( NPP_ODD_ROI_WARNING ),
+    #else
+        error_entry( NPP_INVALID_HOST_POINTER_ERROR ),
+        error_entry( NPP_INVALID_DEVICE_POINTER_ERROR ),
+        error_entry( NPP_LUT_PALETTE_BITSIZE_ERROR ),
+        error_entry( NPP_ZC_MODE_NOT_SUPPORTED_ERROR ),
+        error_entry( NPP_MEMFREE_ERROR ),
+        error_entry( NPP_MEMSET_ERROR ),
+        error_entry( NPP_QUALITY_INDEX_ERROR ),
+        error_entry( NPP_HISTOGRAM_NUMBER_OF_LEVELS_ERROR ),
+        error_entry( NPP_CHANNEL_ORDER_ERROR ),
+        error_entry( NPP_ZERO_MASK_VALUE_ERROR ),
+        error_entry( NPP_QUADRANGLE_ERROR ),
+        error_entry( NPP_RECTANGLE_ERROR ),
+        error_entry( NPP_COEFFICIENT_ERROR ),
+        error_entry( NPP_NUMBER_OF_CHANNELS_ERROR ),
+        error_entry( NPP_COI_ERROR ),
+        error_entry( NPP_DIVISOR_ERROR ),
+        error_entry( NPP_CHANNEL_ERROR ),
+        error_entry( NPP_STRIDE_ERROR ),
+        error_entry( NPP_ANCHOR_ERROR ),
+        error_entry( NPP_MASK_SIZE_ERROR ),
+        error_entry( NPP_MIRROR_FLIP_ERROR ),
+        error_entry( NPP_MOMENT_00_ZERO_ERROR ),
+        error_entry( NPP_THRESHOLD_NEGATIVE_LEVEL_ERROR ),
+        error_entry( NPP_THRESHOLD_ERROR ),
+        error_entry( NPP_CONTEXT_MATCH_ERROR ),
+        error_entry( NPP_FFT_FLAG_ERROR ),
+        error_entry( NPP_FFT_ORDER_ERROR ),
+        error_entry( NPP_SCALE_RANGE_ERROR ),
+        error_entry( NPP_DATA_TYPE_ERROR ),
+        error_entry( NPP_OUT_OFF_RANGE_ERROR ),
+        error_entry( NPP_DIVIDE_BY_ZERO_ERROR ),
+        error_entry( NPP_MEMORY_ALLOCATION_ERR ),
+        error_entry( NPP_RANGE_ERROR ),
+        error_entry( NPP_BAD_ARGUMENT_ERROR ),
+        error_entry( NPP_NO_MEMORY_ERROR ),
+        error_entry( NPP_ERROR_RESERVED ),
+        error_entry( NPP_NO_OPERATION_WARNING ),
+        error_entry( NPP_DIVIDE_BY_ZERO_WARNING ),
+        error_entry( NPP_WRONG_INTERSECTION_ROI_WARNING ),
+    #endif
+
+        error_entry( NPP_NOT_SUPPORTED_MODE_ERROR ),
+        error_entry( NPP_ROUND_MODE_NOT_SUPPORTED_ERROR ),
+        error_entry( NPP_RESIZE_NO_OPERATION_ERROR ),
+        error_entry( NPP_LUT_NUMBER_OF_LEVELS_ERROR ),
+        error_entry( NPP_TEXTURE_BIND_ERROR ),
+        error_entry( NPP_WRONG_INTERSECTION_ROI_ERROR ),
+        error_entry( NPP_NOT_EVEN_STEP_ERROR ),
+        error_entry( NPP_INTERPOLATION_ERROR ),
+        error_entry( NPP_RESIZE_FACTOR_ERROR ),
+        error_entry( NPP_HAAR_CLASSIFIER_PIXEL_MATCH_ERROR ),
+        error_entry( NPP_MEMCPY_ERROR ),
+        error_entry( NPP_ALIGNMENT_ERROR ),
+        error_entry( NPP_STEP_ERROR ),
+        error_entry( NPP_SIZE_ERROR ),
+        error_entry( NPP_NULL_POINTER_ERROR ),
+        error_entry( NPP_CUDA_KERNEL_EXECUTION_ERROR ),
+        error_entry( NPP_NOT_IMPLEMENTED_ERROR ),
+        error_entry( NPP_ERROR ),
+        error_entry( NPP_NO_ERROR ),
+        error_entry( NPP_SUCCESS ),
+        error_entry( NPP_WRONG_INTERSECTION_QUAD_WARNING ),
+        error_entry( NPP_MISALIGNED_DST_ROI_WARNING ),
+        error_entry( NPP_AFFINE_QUAD_INCORRECT_WARNING ),
+        error_entry( NPP_DOUBLE_SIZE_WARNING )
+    };
+
+    const size_t npp_error_num = sizeof(npp_errors) / sizeof(npp_errors[0]);
+
+    const ErrorEntry cu_errors [] =
+    {
+        error_entry( CUDA_SUCCESS                              ),
+        error_entry( CUDA_ERROR_INVALID_VALUE                  ),
+        error_entry( CUDA_ERROR_OUT_OF_MEMORY                  ),
+        error_entry( CUDA_ERROR_NOT_INITIALIZED                ),
+        error_entry( CUDA_ERROR_DEINITIALIZED                  ),
+        error_entry( CUDA_ERROR_PROFILER_DISABLED              ),
+        error_entry( CUDA_ERROR_PROFILER_NOT_INITIALIZED       ),
+        error_entry( CUDA_ERROR_PROFILER_ALREADY_STARTED       ),
+        error_entry( CUDA_ERROR_PROFILER_ALREADY_STOPPED       ),
+        error_entry( CUDA_ERROR_NO_DEVICE                      ),
+        error_entry( CUDA_ERROR_INVALID_DEVICE                 ),
+        error_entry( CUDA_ERROR_INVALID_IMAGE                  ),
+        error_entry( CUDA_ERROR_INVALID_CONTEXT                ),
+        error_entry( CUDA_ERROR_CONTEXT_ALREADY_CURRENT        ),
+        error_entry( CUDA_ERROR_MAP_FAILED                     ),
+        error_entry( CUDA_ERROR_UNMAP_FAILED                   ),
+        error_entry( CUDA_ERROR_ARRAY_IS_MAPPED                ),
+        error_entry( CUDA_ERROR_ALREADY_MAPPED                 ),
+        error_entry( CUDA_ERROR_NO_BINARY_FOR_GPU              ),
+        error_entry( CUDA_ERROR_ALREADY_ACQUIRED               ),
+        error_entry( CUDA_ERROR_NOT_MAPPED                     ),
+        error_entry( CUDA_ERROR_NOT_MAPPED_AS_ARRAY            ),
+        error_entry( CUDA_ERROR_NOT_MAPPED_AS_POINTER          ),
+        error_entry( CUDA_ERROR_ECC_UNCORRECTABLE              ),
+        error_entry( CUDA_ERROR_UNSUPPORTED_LIMIT              ),
+        error_entry( CUDA_ERROR_CONTEXT_ALREADY_IN_USE         ),
+        error_entry( CUDA_ERROR_INVALID_SOURCE                 ),
+        error_entry( CUDA_ERROR_FILE_NOT_FOUND                 ),
+        error_entry( CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND ),
+        error_entry( CUDA_ERROR_SHARED_OBJECT_INIT_FAILED      ),
+        error_entry( CUDA_ERROR_OPERATING_SYSTEM               ),
+        error_entry( CUDA_ERROR_INVALID_HANDLE                 ),
+        error_entry( CUDA_ERROR_NOT_FOUND                      ),
+        error_entry( CUDA_ERROR_NOT_READY                      ),
+        error_entry( CUDA_ERROR_LAUNCH_FAILED                  ),
+        error_entry( CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES        ),
+        error_entry( CUDA_ERROR_LAUNCH_TIMEOUT                 ),
+        error_entry( CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING  ),
+        error_entry( CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED    ),
+        error_entry( CUDA_ERROR_PEER_ACCESS_NOT_ENABLED        ),
+        error_entry( CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE         ),
+        error_entry( CUDA_ERROR_CONTEXT_IS_DESTROYED           ),
+        error_entry( CUDA_ERROR_ASSERT                         ),
+        error_entry( CUDA_ERROR_TOO_MANY_PEERS                 ),
+        error_entry( CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED ),
+        error_entry( CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED     ),
+        error_entry( CUDA_ERROR_UNKNOWN                        )
+    };
+
+    const size_t cu_errors_num = sizeof(cu_errors) / sizeof(cu_errors[0]);
+
+    cv::String getErrorString(int code, const ErrorEntry* errors, size_t n)
+    {
+        size_t idx = std::find_if(errors, errors + n, ErrorEntryComparer(code)) - errors;
+
+        const char* msg = (idx != n) ? errors[idx].str : "Unknown error code";
+        cv::String str = cv::format("%s [Code = %d]", msg, code);
+
+        return str;
+    }
+}
+
+#endif
+
+String cv::gpu::getNppErrorMessage(int code)
+{
+#ifndef HAVE_CUDA
+    (void) code;
+    return String();
+#else
+    return getErrorString(code, npp_errors, npp_error_num);
+#endif
+}
+
+String cv::gpu::getCudaDriverApiErrorMessage(int code)
+{
+#ifndef HAVE_CUDA
+    (void) code;
+    return String();
+#else
+    return getErrorString(code, cu_errors, cu_errors_num);
+#endif
+}
diff --git a/modules/core/src/gpu_mat.cpp b/modules/core/src/gpu_mat.cpp
new file mode 100644
index 0000000000..33a6046fed
--- /dev/null
+++ b/modules/core/src/gpu_mat.cpp
@@ -0,0 +1,444 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+cv::gpu::GpuMat::GpuMat(int rows_, int cols_, int type_, void* data_, size_t step_) :
+    flags(Mat::MAGIC_VAL + (type_ & Mat::TYPE_MASK)), rows(rows_), cols(cols_),
+    step(step_), data((uchar*)data_), refcount(0),
+    datastart((uchar*)data_), dataend((uchar*)data_)
+{
+    size_t minstep = cols * elemSize();
+
+    if (step == Mat::AUTO_STEP)
+    {
+        step = minstep;
+        flags |= Mat::CONTINUOUS_FLAG;
+    }
+    else
+    {
+        if (rows == 1)
+            step = minstep;
+
+        CV_DbgAssert( step >= minstep );
+
+        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
+    }
+
+    dataend += step * (rows - 1) + minstep;
+}
+
+cv::gpu::GpuMat::GpuMat(Size size_, int type_, void* data_, size_t step_) :
+    flags(Mat::MAGIC_VAL + (type_ & Mat::TYPE_MASK)), rows(size_.height), cols(size_.width),
+    step(step_), data((uchar*)data_), refcount(0),
+    datastart((uchar*)data_), dataend((uchar*)data_)
+{
+    size_t minstep = cols * elemSize();
+
+    if (step == Mat::AUTO_STEP)
+    {
+        step = minstep;
+        flags |= Mat::CONTINUOUS_FLAG;
+    }
+    else
+    {
+        if (rows == 1)
+            step = minstep;
+
+        CV_DbgAssert( step >= minstep );
+
+        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
+    }
+    dataend += step * (rows - 1) + minstep;
+}
+
+cv::gpu::GpuMat::GpuMat(const GpuMat& m, Range rowRange_, Range colRange_)
+{
+    flags = m.flags;
+    step = m.step; refcount = m.refcount;
+    data = m.data; datastart = m.datastart; dataend = m.dataend;
+
+    if (rowRange_ == Range::all())
+    {
+        rows = m.rows;
+    }
+    else
+    {
+        CV_Assert( 0 <= rowRange_.start && rowRange_.start <= rowRange_.end && rowRange_.end <= m.rows );
+
+        rows = rowRange_.size();
+        data += step*rowRange_.start;
+    }
+
+    if (colRange_ == Range::all())
+    {
+        cols = m.cols;
+    }
+    else
+    {
+        CV_Assert( 0 <= colRange_.start && colRange_.start <= colRange_.end && colRange_.end <= m.cols );
+
+        cols = colRange_.size();
+        data += colRange_.start*elemSize();
+        flags &= cols < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
+    }
+
+    if (rows == 1)
+        flags |= Mat::CONTINUOUS_FLAG;
+
+    if (refcount)
+        CV_XADD(refcount, 1);
+
+    if (rows <= 0 || cols <= 0)
+        rows = cols = 0;
+}
+
+cv::gpu::GpuMat::GpuMat(const GpuMat& m, Rect roi) :
+    flags(m.flags), rows(roi.height), cols(roi.width),
+    step(m.step), data(m.data + roi.y*step), refcount(m.refcount),
+    datastart(m.datastart), dataend(m.dataend)
+{
+    flags &= roi.width < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
+    data += roi.x * elemSize();
+
+    CV_Assert( 0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols && 0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows );
+
+    if (refcount)
+        CV_XADD(refcount, 1);
+
+    if (rows <= 0 || cols <= 0)
+        rows = cols = 0;
+}
+
+GpuMat cv::gpu::GpuMat::reshape(int new_cn, int new_rows) const
+{
+    GpuMat hdr = *this;
+
+    int cn = channels();
+    if (new_cn == 0)
+        new_cn = cn;
+
+    int total_width = cols * cn;
+
+    if ((new_cn > total_width || total_width % new_cn != 0) && new_rows == 0)
+        new_rows = rows * total_width / new_cn;
+
+    if (new_rows != 0 && new_rows != rows)
+    {
+        int total_size = total_width * rows;
+
+        if (!isContinuous())
+            CV_Error(cv::Error::BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
+
+        if ((unsigned)new_rows > (unsigned)total_size)
+            CV_Error(cv::Error::StsOutOfRange, "Bad new number of rows");
+
+        total_width = total_size / new_rows;
+
+        if (total_width * new_rows != total_size)
+            CV_Error(cv::Error::StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
+
+        hdr.rows = new_rows;
+        hdr.step = total_width * elemSize1();
+    }
+
+    int new_width = total_width / new_cn;
+
+    if (new_width * new_cn != total_width)
+        CV_Error(cv::Error::BadNumChannels, "The total width is not divisible by the new number of channels");
+
+    hdr.cols = new_width;
+    hdr.flags = (hdr.flags & ~CV_MAT_CN_MASK) | ((new_cn - 1) << CV_CN_SHIFT);
+
+    return hdr;
+}
+
+void cv::gpu::GpuMat::locateROI(Size& wholeSize, Point& ofs) const
+{
+    CV_DbgAssert( step > 0 );
+
+    size_t esz = elemSize();
+    ptrdiff_t delta1 = data - datastart;
+    ptrdiff_t delta2 = dataend - datastart;
+
+    if (delta1 == 0)
+    {
+        ofs.x = ofs.y = 0;
+    }
+    else
+    {
+        ofs.y = static_cast<int>(delta1 / step);
+        ofs.x = static_cast<int>((delta1 - step * ofs.y) / esz);
+
+        CV_DbgAssert( data == datastart + ofs.y * step + ofs.x * esz );
+    }
+
+    size_t minstep = (ofs.x + cols) * esz;
+
+    wholeSize.height = std::max(static_cast<int>((delta2 - minstep) / step + 1), ofs.y + rows);
+    wholeSize.width = std::max(static_cast<int>((delta2 - step * (wholeSize.height - 1)) / esz), ofs.x + cols);
+}
+
+GpuMat& cv::gpu::GpuMat::adjustROI(int dtop, int dbottom, int dleft, int dright)
+{
+    Size wholeSize;
+    Point ofs;
+    locateROI(wholeSize, ofs);
+
+    size_t esz = elemSize();
+
+    int row1 = std::max(ofs.y - dtop, 0);
+    int row2 = std::min(ofs.y + rows + dbottom, wholeSize.height);
+
+    int col1 = std::max(ofs.x - dleft, 0);
+    int col2 = std::min(ofs.x + cols + dright, wholeSize.width);
+
+    data += (row1 - ofs.y) * step + (col1 - ofs.x) * esz;
+    rows = row2 - row1;
+    cols = col2 - col1;
+
+    if (esz * cols == step || rows == 1)
+        flags |= Mat::CONTINUOUS_FLAG;
+    else
+        flags &= ~Mat::CONTINUOUS_FLAG;
+
+    return *this;
+}
+
+namespace
+{
+    template <class ObjType>
+    void createContinuousImpl(int rows, int cols, int type, ObjType& obj)
+    {
+        const int area = rows * cols;
+
+        if (obj.empty() || obj.type() != type || !obj.isContinuous() || obj.size().area() < area)
+            obj.create(1, area, type);
+
+        obj = obj.reshape(obj.channels(), rows);
+    }
+}
+
+void cv::gpu::createContinuous(int rows, int cols, int type, OutputArray arr)
+{
+    switch (arr.kind())
+    {
+    case _InputArray::MAT:
+        ::createContinuousImpl(rows, cols, type, arr.getMatRef());
+        break;
+
+    case _InputArray::GPU_MAT:
+        ::createContinuousImpl(rows, cols, type, arr.getGpuMatRef());
+        break;
+
+    case _InputArray::CUDA_MEM:
+        ::createContinuousImpl(rows, cols, type, arr.getCudaMemRef());
+        break;
+
+    default:
+        arr.create(rows, cols, type);
+    }
+}
+
+namespace
+{
+    template <class ObjType>
+    void ensureSizeIsEnoughImpl(int rows, int cols, int type, ObjType& obj)
+    {
+        if (obj.empty() || obj.type() != type || obj.data != obj.datastart)
+        {
+            obj.create(rows, cols, type);
+        }
+        else
+        {
+            const size_t esz = obj.elemSize();
+            const ptrdiff_t delta2 = obj.dataend - obj.datastart;
+
+            const size_t minstep = obj.cols * esz;
+
+            Size wholeSize;
+            wholeSize.height = std::max(static_cast<int>((delta2 - minstep) / static_cast<size_t>(obj.step) + 1), obj.rows);
+            wholeSize.width = std::max(static_cast<int>((delta2 - static_cast<size_t>(obj.step) * (wholeSize.height - 1)) / esz), obj.cols);
+
+            if (wholeSize.height < rows || wholeSize.width < cols)
+            {
+                obj.create(rows, cols, type);
+            }
+            else
+            {
+                obj.cols = cols;
+                obj.rows = rows;
+            }
+        }
+    }
+}
+
+void cv::gpu::ensureSizeIsEnough(int rows, int cols, int type, OutputArray arr)
+{
+    switch (arr.kind())
+    {
+    case _InputArray::MAT:
+        ::ensureSizeIsEnoughImpl(rows, cols, type, arr.getMatRef());
+        break;
+
+    case _InputArray::GPU_MAT:
+        ::ensureSizeIsEnoughImpl(rows, cols, type, arr.getGpuMatRef());
+        break;
+
+    case _InputArray::CUDA_MEM:
+        ::ensureSizeIsEnoughImpl(rows, cols, type, arr.getCudaMemRef());
+        break;
+
+    default:
+        arr.create(rows, cols, type);
+    }
+}
+
+GpuMat cv::gpu::allocMatFromBuf(int rows, int cols, int type, GpuMat& mat)
+{
+    if (!mat.empty() && mat.type() == type && mat.rows >= rows && mat.cols >= cols)
+        return mat(Rect(0, 0, cols, rows));
+
+    return mat = GpuMat(rows, cols, type);
+}
+
+#ifndef HAVE_CUDA
+
+void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
+{
+    (void) _rows;
+    (void) _cols;
+    (void) _type;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::release()
+{
+}
+
+void cv::gpu::GpuMat::upload(InputArray arr)
+{
+    (void) arr;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::upload(InputArray arr, Stream& _stream)
+{
+    (void) arr;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::download(OutputArray _dst) const
+{
+    (void) _dst;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::download(OutputArray _dst, Stream& _stream) const
+{
+    (void) _dst;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst) const
+{
+    (void) _dst;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst, Stream& _stream) const
+{
+    (void) _dst;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::copyTo(OutputArray _dst, InputArray _mask, Stream& _stream) const
+{
+    (void) _dst;
+    (void) _mask;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+GpuMat& cv::gpu::GpuMat::setTo(Scalar s, Stream& _stream)
+{
+    (void) s;
+    (void) _stream;
+    throw_no_cuda();
+    return *this;
+}
+
+GpuMat& cv::gpu::GpuMat::setTo(Scalar s, InputArray _mask, Stream& _stream)
+{
+    (void) s;
+    (void) _mask;
+    (void) _stream;
+    throw_no_cuda();
+    return *this;
+}
+
+void cv::gpu::GpuMat::convertTo(OutputArray _dst, int rtype, Stream& _stream) const
+{
+    (void) _dst;
+    (void) rtype;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+void cv::gpu::GpuMat::convertTo(OutputArray _dst, int rtype, double alpha, double beta, Stream& _stream) const
+{
+    (void) _dst;
+    (void) rtype;
+    (void) alpha;
+    (void) beta;
+    (void) _stream;
+    throw_no_cuda();
+}
+
+#endif
diff --git a/modules/core/src/gpu_stream.cpp b/modules/core/src/gpu_stream.cpp
new file mode 100644
index 0000000000..fcdf480a91
--- /dev/null
+++ b/modules/core/src/gpu_stream.cpp
@@ -0,0 +1,299 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+////////////////////////////////////////////////////////////////
+// Stream
+
+#ifndef HAVE_CUDA
+
+class cv::gpu::Stream::Impl
+{
+public:
+    Impl(void* ptr = 0)
+    {
+        (void) ptr;
+        throw_no_cuda();
+    }
+};
+
+#else
+
+class cv::gpu::Stream::Impl
+{
+public:
+    cudaStream_t stream;
+
+    Impl();
+    Impl(cudaStream_t stream);
+
+    ~Impl();
+};
+
+cv::gpu::Stream::Impl::Impl() : stream(0)
+{
+    cudaSafeCall( cudaStreamCreate(&stream) );
+}
+
+cv::gpu::Stream::Impl::Impl(cudaStream_t stream_) : stream(stream_)
+{
+}
+
+cv::gpu::Stream::Impl::~Impl()
+{
+    if (stream)
+        cudaStreamDestroy(stream);
+}
+
+cudaStream_t cv::gpu::StreamAccessor::getStream(const Stream& stream)
+{
+    return stream.impl_->stream;
+}
+
+#endif
+
+cv::gpu::Stream::Stream()
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+#else
+    impl_ = makePtr<Impl>();
+#endif
+}
+
+bool cv::gpu::Stream::queryIfComplete() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    cudaError_t err = cudaStreamQuery(impl_->stream);
+
+    if (err == cudaErrorNotReady || err == cudaSuccess)
+        return err == cudaSuccess;
+
+    cudaSafeCall(err);
+    return false;
+#endif
+}
+
+void cv::gpu::Stream::waitForCompletion()
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaStreamSynchronize(impl_->stream) );
+#endif
+}
+
+void cv::gpu::Stream::waitEvent(const Event& event)
+{
+#ifndef HAVE_CUDA
+    (void) event;
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaStreamWaitEvent(impl_->stream, EventAccessor::getEvent(event), 0) );
+#endif
+}
+
+#if defined(HAVE_CUDA) && (CUDART_VERSION >= 5000)
+
+namespace
+{
+    struct CallbackData
+    {
+        Stream::StreamCallback callback;
+        void* userData;
+
+        CallbackData(Stream::StreamCallback callback_, void* userData_) : callback(callback_), userData(userData_) {}
+    };
+
+    void CUDART_CB cudaStreamCallback(cudaStream_t, cudaError_t status, void* userData)
+    {
+        CallbackData* data = reinterpret_cast<CallbackData*>(userData);
+        data->callback(static_cast<int>(status), data->userData);
+        delete data;
+    }
+}
+
+#endif
+
+void cv::gpu::Stream::enqueueHostCallback(StreamCallback callback, void* userData)
+{
+#ifndef HAVE_CUDA
+    (void) callback;
+    (void) userData;
+    throw_no_cuda();
+#else
+    #if CUDART_VERSION < 5000
+        (void) callback;
+        (void) userData;
+        CV_Error(cv::Error::StsNotImplemented, "This function requires CUDA 5.0");
+    #else
+        CallbackData* data = new CallbackData(callback, userData);
+
+        cudaSafeCall( cudaStreamAddCallback(impl_->stream, cudaStreamCallback, data, 0) );
+    #endif
+#endif
+}
+
+Stream& cv::gpu::Stream::Null()
+{
+    static Stream s(Ptr<Impl>(new Impl(0)));
+    return s;
+}
+
+cv::gpu::Stream::operator bool_type() const
+{
+#ifndef HAVE_CUDA
+    return 0;
+#else
+    return (impl_->stream != 0) ? &Stream::this_type_does_not_support_comparisons : 0;
+#endif
+}
+
+
+////////////////////////////////////////////////////////////////
+// Stream
+
+#ifndef HAVE_CUDA
+
+class cv::gpu::Event::Impl
+{
+public:
+    Impl(unsigned int)
+    {
+        throw_no_cuda();
+    }
+};
+
+#else
+
+class cv::gpu::Event::Impl
+{
+public:
+    cudaEvent_t event;
+
+    Impl(unsigned int flags);
+    ~Impl();
+};
+
+cv::gpu::Event::Impl::Impl(unsigned int flags) : event(0)
+{
+    cudaSafeCall( cudaEventCreateWithFlags(&event, flags) );
+}
+
+cv::gpu::Event::Impl::~Impl()
+{
+    if (event)
+        cudaEventDestroy(event);
+}
+
+cudaEvent_t cv::gpu::EventAccessor::getEvent(const Event& event)
+{
+    return event.impl_->event;
+}
+
+#endif
+
+cv::gpu::Event::Event(CreateFlags flags)
+{
+#ifndef HAVE_CUDA
+    (void) flags;
+    throw_no_cuda();
+#else
+    impl_ = makePtr<Impl>(flags);
+#endif
+}
+
+void cv::gpu::Event::record(Stream& stream)
+{
+#ifndef HAVE_CUDA
+    (void) stream;
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaEventRecord(impl_->event, StreamAccessor::getStream(stream)) );
+#endif
+}
+
+bool cv::gpu::Event::queryIfComplete() const
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+    return false;
+#else
+    cudaError_t err = cudaEventQuery(impl_->event);
+
+    if (err == cudaErrorNotReady || err == cudaSuccess)
+        return err == cudaSuccess;
+
+    cudaSafeCall(err);
+    return false;
+#endif
+}
+
+void cv::gpu::Event::waitForCompletion()
+{
+#ifndef HAVE_CUDA
+    throw_no_cuda();
+#else
+    cudaSafeCall( cudaEventSynchronize(impl_->event) );
+#endif
+}
+
+float cv::gpu::Event::elapsedTime(const Event& start, const Event& end)
+{
+#ifndef HAVE_CUDA
+    (void) start;
+    (void) end;
+    throw_no_cuda();
+    return 0.0f;
+#else
+    float ms;
+    cudaSafeCall( cudaEventElapsedTime(&ms, start.impl_->event, end.impl_->event) );
+    return ms;
+#endif
+}
diff --git a/modules/core/src/gpumat.cpp b/modules/core/src/gpumat.cpp
deleted file mode 100644
index cc7442a664..0000000000
--- a/modules/core/src/gpumat.cpp
+++ /dev/null
@@ -1,1578 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#include "opencv2/core/gpumat.hpp"
-#include <iostream>
-
-#ifdef HAVE_CUDA
-    #include <cuda_runtime.h>
-    #include <npp.h>
-
-    #define CUDART_MINIMUM_REQUIRED_VERSION 4020
-    #define NPP_MINIMUM_REQUIRED_VERSION 4200
-
-    #if (CUDART_VERSION < CUDART_MINIMUM_REQUIRED_VERSION)
-        #error "Insufficient Cuda Runtime library version, please update it."
-    #endif
-
-    #if (NPP_VERSION_MAJOR * 1000 + NPP_VERSION_MINOR * 100 + NPP_VERSION_BUILD < NPP_MINIMUM_REQUIRED_VERSION)
-        #error "Insufficient NPP version, please update it."
-    #endif
-#endif
-
-using namespace cv;
-using namespace cv::gpu;
-
-#ifndef HAVE_CUDA
-
-#define throw_nogpu CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support")
-
-#else // HAVE_CUDA
-
-namespace
-{
-#if defined(__GNUC__)
-    #define nppSafeCall(expr)  ___nppSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define nppSafeCall(expr)  ___nppSafeCall(expr, __FILE__, __LINE__)
-#endif
-
-    inline void ___nppSafeCall(int err, const char *file, const int line, const char *func = "")
-    {
-        if (err < 0)
-        {
-            std::ostringstream msg;
-            msg << "NPP API Call Error: " << err;
-            cv::gpu::error(msg.str().c_str(), file, line, func);
-        }
-    }
-}
-
-#endif // HAVE_CUDA
-
-//////////////////////////////// Initialization & Info ////////////////////////
-
-#ifndef HAVE_CUDA
-
-int cv::gpu::getCudaEnabledDeviceCount() { return 0; }
-
-void cv::gpu::setDevice(int) { throw_nogpu; }
-int cv::gpu::getDevice() { throw_nogpu; return 0; }
-
-void cv::gpu::resetDevice() { throw_nogpu; }
-
-bool cv::gpu::deviceSupports(FeatureSet) { throw_nogpu; return false; }
-
-bool cv::gpu::TargetArchs::builtWith(FeatureSet) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::has(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasPtx(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasBin(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasEqualOrLessPtx(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasEqualOrGreater(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasEqualOrGreaterPtx(int, int) { throw_nogpu; return false; }
-bool cv::gpu::TargetArchs::hasEqualOrGreaterBin(int, int) { throw_nogpu; return false; }
-
-size_t cv::gpu::DeviceInfo::sharedMemPerBlock() const { throw_nogpu; return 0; }
-void cv::gpu::DeviceInfo::queryMemory(size_t&, size_t&) const { throw_nogpu; }
-size_t cv::gpu::DeviceInfo::freeMemory() const { throw_nogpu; return 0; }
-size_t cv::gpu::DeviceInfo::totalMemory() const { throw_nogpu; return 0; }
-bool cv::gpu::DeviceInfo::supports(FeatureSet) const { throw_nogpu; return false; }
-bool cv::gpu::DeviceInfo::isCompatible() const { throw_nogpu; return false; }
-void cv::gpu::DeviceInfo::query() { throw_nogpu; }
-
-void cv::gpu::printCudaDeviceInfo(int) { throw_nogpu; }
-void cv::gpu::printShortCudaDeviceInfo(int) { throw_nogpu; }
-
-#else // HAVE_CUDA
-
-int cv::gpu::getCudaEnabledDeviceCount()
-{
-    int count;
-    cudaError_t error = cudaGetDeviceCount( &count );
-
-    if (error == cudaErrorInsufficientDriver)
-        return -1;
-
-    if (error == cudaErrorNoDevice)
-        return 0;
-
-    cudaSafeCall( error );
-    return count;
-}
-
-void cv::gpu::setDevice(int device)
-{
-    cudaSafeCall( cudaSetDevice( device ) );
-}
-
-int cv::gpu::getDevice()
-{
-    int device;
-    cudaSafeCall( cudaGetDevice( &device ) );
-    return device;
-}
-
-void cv::gpu::resetDevice()
-{
-    cudaSafeCall( cudaDeviceReset() );
-}
-
-namespace
-{
-    class CudaArch
-    {
-    public:
-        CudaArch();
-
-        bool builtWith(FeatureSet feature_set) const;
-        bool hasPtx(int major, int minor) const;
-        bool hasBin(int major, int minor) const;
-        bool hasEqualOrLessPtx(int major, int minor) const;
-        bool hasEqualOrGreaterPtx(int major, int minor) const;
-        bool hasEqualOrGreaterBin(int major, int minor) const;
-
-    private:
-        static void fromStr(const std::string& set_as_str, std::vector<int>& arr);
-
-        std::vector<int> bin;
-        std::vector<int> ptx;
-        std::vector<int> features;
-    };
-
-    const CudaArch cudaArch;
-
-    CudaArch::CudaArch()
-    {
-        fromStr(CUDA_ARCH_BIN, bin);
-        fromStr(CUDA_ARCH_PTX, ptx);
-        fromStr(CUDA_ARCH_FEATURES, features);
-    }
-
-    bool CudaArch::builtWith(FeatureSet feature_set) const
-    {
-        return !features.empty() && (features.back() >= feature_set);
-    }
-
-    bool CudaArch::hasPtx(int major, int minor) const
-    {
-        return std::find(ptx.begin(), ptx.end(), major * 10 + minor) != ptx.end();
-    }
-
-    bool CudaArch::hasBin(int major, int minor) const
-    {
-        return std::find(bin.begin(), bin.end(), major * 10 + minor) != bin.end();
-    }
-
-    bool CudaArch::hasEqualOrLessPtx(int major, int minor) const
-    {
-        return !ptx.empty() && (ptx.front() <= major * 10 + minor);
-    }
-
-    bool CudaArch::hasEqualOrGreaterPtx(int major, int minor) const
-    {
-        return !ptx.empty() && (ptx.back() >= major * 10 + minor);
-    }
-
-    bool CudaArch::hasEqualOrGreaterBin(int major, int minor) const
-    {
-        return !bin.empty() && (bin.back() >= major * 10 + minor);
-    }
-
-    void CudaArch::fromStr(const std::string& set_as_str, std::vector<int>& arr)
-    {
-        if (set_as_str.find_first_not_of(" ") == std::string::npos)
-            return;
-
-        std::istringstream stream(set_as_str);
-        int cur_value;
-
-        while (!stream.eof())
-        {
-            stream >> cur_value;
-            arr.push_back(cur_value);
-        }
-
-        std::sort(arr.begin(), arr.end());
-    }
-}
-
-bool cv::gpu::TargetArchs::builtWith(cv::gpu::FeatureSet feature_set)
-{
-    return cudaArch.builtWith(feature_set);
-}
-
-bool cv::gpu::TargetArchs::has(int major, int minor)
-{
-    return hasPtx(major, minor) || hasBin(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasPtx(int major, int minor)
-{
-    return cudaArch.hasPtx(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasBin(int major, int minor)
-{
-    return cudaArch.hasBin(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasEqualOrLessPtx(int major, int minor)
-{
-    return cudaArch.hasEqualOrLessPtx(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasEqualOrGreater(int major, int minor)
-{
-    return hasEqualOrGreaterPtx(major, minor) || hasEqualOrGreaterBin(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasEqualOrGreaterPtx(int major, int minor)
-{
-    return cudaArch.hasEqualOrGreaterPtx(major, minor);
-}
-
-bool cv::gpu::TargetArchs::hasEqualOrGreaterBin(int major, int minor)
-{
-    return cudaArch.hasEqualOrGreaterBin(major, minor);
-}
-
-bool cv::gpu::deviceSupports(FeatureSet feature_set)
-{
-    static int versions[] =
-    {
-        -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
-    };
-    static const int cache_size = static_cast<int>(sizeof(versions) / sizeof(versions[0]));
-
-    const int devId = getDevice();
-
-    int version;
-
-    if (devId < cache_size && versions[devId] >= 0)
-        version = versions[devId];
-    else
-    {
-        DeviceInfo dev(devId);
-        version = dev.majorVersion() * 10 + dev.minorVersion();
-        if (devId < cache_size)
-            versions[devId] = version;
-    }
-
-    return TargetArchs::builtWith(feature_set) && (version >= feature_set);
-}
-
-namespace
-{
-    class DeviceProps
-    {
-    public:
-        DeviceProps();
-        ~DeviceProps();
-
-        cudaDeviceProp* get(int devID);
-
-    private:
-        std::vector<cudaDeviceProp*> props_;
-    };
-
-    DeviceProps::DeviceProps()
-    {
-        props_.resize(10, 0);
-    }
-
-    DeviceProps::~DeviceProps()
-    {
-        for (size_t i = 0; i < props_.size(); ++i)
-        {
-            if (props_[i])
-                delete props_[i];
-        }
-        props_.clear();
-    }
-
-    cudaDeviceProp* DeviceProps::get(int devID)
-    {
-        if (devID >= (int) props_.size())
-            props_.resize(devID + 5, 0);
-
-        if (!props_[devID])
-        {
-            props_[devID] = new cudaDeviceProp;
-            cudaSafeCall( cudaGetDeviceProperties(props_[devID], devID) );
-        }
-
-        return props_[devID];
-    }
-
-    DeviceProps deviceProps;
-}
-
-size_t cv::gpu::DeviceInfo::sharedMemPerBlock() const
-{
-    return deviceProps.get(device_id_)->sharedMemPerBlock;
-}
-
-void cv::gpu::DeviceInfo::queryMemory(size_t& _totalMemory, size_t& _freeMemory) const
-{
-    int prevDeviceID = getDevice();
-    if (prevDeviceID != device_id_)
-        setDevice(device_id_);
-
-    cudaSafeCall( cudaMemGetInfo(&_freeMemory, &_totalMemory) );
-
-    if (prevDeviceID != device_id_)
-        setDevice(prevDeviceID);
-}
-
-size_t cv::gpu::DeviceInfo::freeMemory() const
-{
-    size_t _totalMemory, _freeMemory;
-    queryMemory(_totalMemory, _freeMemory);
-    return _freeMemory;
-}
-
-size_t cv::gpu::DeviceInfo::totalMemory() const
-{
-    size_t _totalMemory, _freeMemory;
-    queryMemory(_totalMemory, _freeMemory);
-    return _totalMemory;
-}
-
-bool cv::gpu::DeviceInfo::supports(FeatureSet feature_set) const
-{
-    int version = majorVersion() * 10 + minorVersion();
-    return version >= feature_set;
-}
-
-bool cv::gpu::DeviceInfo::isCompatible() const
-{
-    // Check PTX compatibility
-    if (TargetArchs::hasEqualOrLessPtx(majorVersion(), minorVersion()))
-        return true;
-
-    // Check BIN compatibility
-    for (int i = minorVersion(); i >= 0; --i)
-        if (TargetArchs::hasBin(majorVersion(), i))
-            return true;
-
-    return false;
-}
-
-void cv::gpu::DeviceInfo::query()
-{
-    const cudaDeviceProp* prop = deviceProps.get(device_id_);
-
-    name_ = prop->name;
-    multi_processor_count_ = prop->multiProcessorCount;
-    majorVersion_ = prop->major;
-    minorVersion_ = prop->minor;
-}
-
-namespace
-{
-    int convertSMVer2Cores(int major, int minor)
-    {
-        // Defines for GPU Architecture types (using the SM version to determine the # of cores per SM
-        typedef struct {
-            int SM; // 0xMm (hexidecimal notation), M = SM Major version, and m = SM minor version
-            int Cores;
-        } SMtoCores;
-
-        SMtoCores gpuArchCoresPerSM[] =  { { 0x10,  8 }, { 0x11,  8 }, { 0x12,  8 }, { 0x13,  8 }, { 0x20, 32 }, { 0x21, 48 }, {0x30, 192}, {0x35, 192}, { -1, -1 }  };
-
-        int index = 0;
-        while (gpuArchCoresPerSM[index].SM != -1)
-        {
-            if (gpuArchCoresPerSM[index].SM == ((major << 4) + minor) )
-                return gpuArchCoresPerSM[index].Cores;
-            index++;
-        }
-
-        return -1;
-    }
-}
-
-void cv::gpu::printCudaDeviceInfo(int device)
-{
-    int count = getCudaEnabledDeviceCount();
-    bool valid = (device >= 0) && (device < count);
-
-    int beg = valid ? device   : 0;
-    int end = valid ? device+1 : count;
-
-    printf("*** CUDA Device Query (Runtime API) version (CUDART static linking) *** \n\n");
-    printf("Device count: %d\n", count);
-
-    int driverVersion = 0, runtimeVersion = 0;
-    cudaSafeCall( cudaDriverGetVersion(&driverVersion) );
-    cudaSafeCall( cudaRuntimeGetVersion(&runtimeVersion) );
-
-    const char *computeMode[] = {
-        "Default (multiple host threads can use ::cudaSetDevice() with device simultaneously)",
-        "Exclusive (only one host thread in one process is able to use ::cudaSetDevice() with this device)",
-        "Prohibited (no host thread can use ::cudaSetDevice() with this device)",
-        "Exclusive Process (many threads in one process is able to use ::cudaSetDevice() with this device)",
-        "Unknown",
-        NULL
-    };
-
-    for(int dev = beg; dev < end; ++dev)
-    {
-        cudaDeviceProp prop;
-        cudaSafeCall( cudaGetDeviceProperties(&prop, dev) );
-
-        printf("\nDevice %d: \"%s\"\n", dev, prop.name);
-        printf("  CUDA Driver Version / Runtime Version          %d.%d / %d.%d\n", driverVersion/1000, driverVersion%100, runtimeVersion/1000, runtimeVersion%100);
-        printf("  CUDA Capability Major/Minor version number:    %d.%d\n", prop.major, prop.minor);
-        printf("  Total amount of global memory:                 %.0f MBytes (%llu bytes)\n", (float)prop.totalGlobalMem/1048576.0f, (unsigned long long) prop.totalGlobalMem);
-
-        int cores = convertSMVer2Cores(prop.major, prop.minor);
-        if (cores > 0)
-            printf("  (%2d) Multiprocessors x (%2d) CUDA Cores/MP:     %d CUDA Cores\n", prop.multiProcessorCount, cores, cores * prop.multiProcessorCount);
-
-        printf("  GPU Clock Speed:                               %.2f GHz\n", prop.clockRate * 1e-6f);
-
-        printf("  Max Texture Dimension Size (x,y,z)             1D=(%d), 2D=(%d,%d), 3D=(%d,%d,%d)\n",
-            prop.maxTexture1D, prop.maxTexture2D[0], prop.maxTexture2D[1],
-            prop.maxTexture3D[0], prop.maxTexture3D[1], prop.maxTexture3D[2]);
-        printf("  Max Layered Texture Size (dim) x layers        1D=(%d) x %d, 2D=(%d,%d) x %d\n",
-            prop.maxTexture1DLayered[0], prop.maxTexture1DLayered[1],
-            prop.maxTexture2DLayered[0], prop.maxTexture2DLayered[1], prop.maxTexture2DLayered[2]);
-
-        printf("  Total amount of constant memory:               %u bytes\n", (int)prop.totalConstMem);
-        printf("  Total amount of shared memory per block:       %u bytes\n", (int)prop.sharedMemPerBlock);
-        printf("  Total number of registers available per block: %d\n", prop.regsPerBlock);
-        printf("  Warp size:                                     %d\n", prop.warpSize);
-        printf("  Maximum number of threads per block:           %d\n", prop.maxThreadsPerBlock);
-        printf("  Maximum sizes of each dimension of a block:    %d x %d x %d\n", prop.maxThreadsDim[0], prop.maxThreadsDim[1], prop.maxThreadsDim[2]);
-        printf("  Maximum sizes of each dimension of a grid:     %d x %d x %d\n", prop.maxGridSize[0], prop.maxGridSize[1],  prop.maxGridSize[2]);
-        printf("  Maximum memory pitch:                          %u bytes\n", (int)prop.memPitch);
-        printf("  Texture alignment:                             %u bytes\n", (int)prop.textureAlignment);
-
-        printf("  Concurrent copy and execution:                 %s with %d copy engine(s)\n", (prop.deviceOverlap ? "Yes" : "No"), prop.asyncEngineCount);
-        printf("  Run time limit on kernels:                     %s\n", prop.kernelExecTimeoutEnabled ? "Yes" : "No");
-        printf("  Integrated GPU sharing Host Memory:            %s\n", prop.integrated ? "Yes" : "No");
-        printf("  Support host page-locked memory mapping:       %s\n", prop.canMapHostMemory ? "Yes" : "No");
-
-        printf("  Concurrent kernel execution:                   %s\n", prop.concurrentKernels ? "Yes" : "No");
-        printf("  Alignment requirement for Surfaces:            %s\n", prop.surfaceAlignment ? "Yes" : "No");
-        printf("  Device has ECC support enabled:                %s\n", prop.ECCEnabled ? "Yes" : "No");
-        printf("  Device is using TCC driver mode:               %s\n", prop.tccDriver ? "Yes" : "No");
-        printf("  Device supports Unified Addressing (UVA):      %s\n", prop.unifiedAddressing ? "Yes" : "No");
-        printf("  Device PCI Bus ID / PCI location ID:           %d / %d\n", prop.pciBusID, prop.pciDeviceID );
-        printf("  Compute Mode:\n");
-        printf("      %s \n", computeMode[prop.computeMode]);
-    }
-
-    printf("\n");
-    printf("deviceQuery, CUDA Driver = CUDART");
-    printf(", CUDA Driver Version  = %d.%d", driverVersion / 1000, driverVersion % 100);
-    printf(", CUDA Runtime Version = %d.%d", runtimeVersion/1000, runtimeVersion%100);
-    printf(", NumDevs = %d\n\n", count);
-    fflush(stdout);
-}
-
-void cv::gpu::printShortCudaDeviceInfo(int device)
-{
-    int count = getCudaEnabledDeviceCount();
-    bool valid = (device >= 0) && (device < count);
-
-    int beg = valid ? device   : 0;
-    int end = valid ? device+1 : count;
-
-    int driverVersion = 0, runtimeVersion = 0;
-    cudaSafeCall( cudaDriverGetVersion(&driverVersion) );
-    cudaSafeCall( cudaRuntimeGetVersion(&runtimeVersion) );
-
-    for(int dev = beg; dev < end; ++dev)
-    {
-        cudaDeviceProp prop;
-        cudaSafeCall( cudaGetDeviceProperties(&prop, dev) );
-
-        const char *arch_str = prop.major < 2 ? " (not Fermi)" : "";
-        printf("Device %d:  \"%s\"  %.0fMb", dev, prop.name, (float)prop.totalGlobalMem/1048576.0f);
-        printf(", sm_%d%d%s", prop.major, prop.minor, arch_str);
-
-        int cores = convertSMVer2Cores(prop.major, prop.minor);
-        if (cores > 0)
-            printf(", %d cores", cores * prop.multiProcessorCount);
-
-        printf(", Driver/Runtime ver.%d.%d/%d.%d\n", driverVersion/1000, driverVersion%100, runtimeVersion/1000, runtimeVersion%100);
-    }
-    fflush(stdout);
-}
-
-#endif // HAVE_CUDA
-
-//////////////////////////////// GpuMat ///////////////////////////////
-
-cv::gpu::GpuMat::GpuMat(const GpuMat& m)
-    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend)
-{
-    if (refcount)
-        CV_XADD(refcount, 1);
-}
-
-cv::gpu::GpuMat::GpuMat(int rows_, int cols_, int type_, void* data_, size_t step_) :
-    flags(Mat::MAGIC_VAL + (type_ & TYPE_MASK)), rows(rows_), cols(cols_),
-    step(step_), data((uchar*)data_), refcount(0),
-    datastart((uchar*)data_), dataend((uchar*)data_)
-{
-    size_t minstep = cols * elemSize();
-
-    if (step == Mat::AUTO_STEP)
-    {
-        step = minstep;
-        flags |= Mat::CONTINUOUS_FLAG;
-    }
-    else
-    {
-        if (rows == 1)
-            step = minstep;
-
-        CV_DbgAssert(step >= minstep);
-
-        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
-    }
-    dataend += step * (rows - 1) + minstep;
-}
-
-cv::gpu::GpuMat::GpuMat(Size size_, int type_, void* data_, size_t step_) :
-    flags(Mat::MAGIC_VAL + (type_ & TYPE_MASK)), rows(size_.height), cols(size_.width),
-    step(step_), data((uchar*)data_), refcount(0),
-    datastart((uchar*)data_), dataend((uchar*)data_)
-{
-    size_t minstep = cols * elemSize();
-
-    if (step == Mat::AUTO_STEP)
-    {
-        step = minstep;
-        flags |= Mat::CONTINUOUS_FLAG;
-    }
-    else
-    {
-        if (rows == 1)
-            step = minstep;
-
-        CV_DbgAssert(step >= minstep);
-
-        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
-    }
-    dataend += step * (rows - 1) + minstep;
-}
-
-cv::gpu::GpuMat::GpuMat(const GpuMat& m, Range _rowRange, Range _colRange)
-{
-    flags = m.flags;
-    step = m.step; refcount = m.refcount;
-    data = m.data; datastart = m.datastart; dataend = m.dataend;
-
-    if (_rowRange == Range::all())
-        rows = m.rows;
-    else
-    {
-        CV_Assert(0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows);
-
-        rows = _rowRange.size();
-        data += step*_rowRange.start;
-    }
-
-    if (_colRange == Range::all())
-        cols = m.cols;
-    else
-    {
-        CV_Assert(0 <= _colRange.start && _colRange.start <= _colRange.end && _colRange.end <= m.cols);
-
-        cols = _colRange.size();
-        data += _colRange.start*elemSize();
-        flags &= cols < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
-    }
-
-    if (rows == 1)
-        flags |= Mat::CONTINUOUS_FLAG;
-
-    if (refcount)
-        CV_XADD(refcount, 1);
-
-    if (rows <= 0 || cols <= 0)
-        rows = cols = 0;
-}
-
-cv::gpu::GpuMat::GpuMat(const GpuMat& m, Rect roi) :
-    flags(m.flags), rows(roi.height), cols(roi.width),
-    step(m.step), data(m.data + roi.y*step), refcount(m.refcount),
-    datastart(m.datastart), dataend(m.dataend)
-{
-    flags &= roi.width < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
-    data += roi.x * elemSize();
-
-    CV_Assert(0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols && 0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows);
-
-    if (refcount)
-        CV_XADD(refcount, 1);
-
-    if (rows <= 0 || cols <= 0)
-        rows = cols = 0;
-}
-
-cv::gpu::GpuMat::GpuMat(const Mat& m) :
-    flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0)
-{
-    upload(m);
-}
-
-GpuMat& cv::gpu::GpuMat::operator = (const GpuMat& m)
-{
-    if (this != &m)
-    {
-        GpuMat temp(m);
-        swap(temp);
-    }
-
-    return *this;
-}
-
-void cv::gpu::GpuMat::swap(GpuMat& b)
-{
-    std::swap(flags, b.flags);
-    std::swap(rows, b.rows);
-    std::swap(cols, b.cols);
-    std::swap(step, b.step);
-    std::swap(data, b.data);
-    std::swap(datastart, b.datastart);
-    std::swap(dataend, b.dataend);
-    std::swap(refcount, b.refcount);
-}
-
-void cv::gpu::GpuMat::locateROI(Size& wholeSize, Point& ofs) const
-{
-    size_t esz = elemSize();
-    ptrdiff_t delta1 = data - datastart;
-    ptrdiff_t delta2 = dataend - datastart;
-
-    CV_DbgAssert(step > 0);
-
-    if (delta1 == 0)
-        ofs.x = ofs.y = 0;
-    else
-    {
-        ofs.y = static_cast<int>(delta1 / step);
-        ofs.x = static_cast<int>((delta1 - step * ofs.y) / esz);
-
-        CV_DbgAssert(data == datastart + ofs.y * step + ofs.x * esz);
-    }
-
-    size_t minstep = (ofs.x + cols) * esz;
-
-    wholeSize.height = std::max(static_cast<int>((delta2 - minstep) / step + 1), ofs.y + rows);
-    wholeSize.width = std::max(static_cast<int>((delta2 - step * (wholeSize.height - 1)) / esz), ofs.x + cols);
-}
-
-GpuMat& cv::gpu::GpuMat::adjustROI(int dtop, int dbottom, int dleft, int dright)
-{
-    Size wholeSize;
-    Point ofs;
-    locateROI(wholeSize, ofs);
-
-    size_t esz = elemSize();
-
-    int row1 = std::max(ofs.y - dtop, 0);
-    int row2 = std::min(ofs.y + rows + dbottom, wholeSize.height);
-
-    int col1 = std::max(ofs.x - dleft, 0);
-    int col2 = std::min(ofs.x + cols + dright, wholeSize.width);
-
-    data += (row1 - ofs.y) * step + (col1 - ofs.x) * esz;
-    rows = row2 - row1;
-    cols = col2 - col1;
-
-    if (esz * cols == step || rows == 1)
-        flags |= Mat::CONTINUOUS_FLAG;
-    else
-        flags &= ~Mat::CONTINUOUS_FLAG;
-
-    return *this;
-}
-
-GpuMat cv::gpu::GpuMat::reshape(int new_cn, int new_rows) const
-{
-    GpuMat hdr = *this;
-
-    int cn = channels();
-    if (new_cn == 0)
-        new_cn = cn;
-
-    int total_width = cols * cn;
-
-    if ((new_cn > total_width || total_width % new_cn != 0) && new_rows == 0)
-        new_rows = rows * total_width / new_cn;
-
-    if (new_rows != 0 && new_rows != rows)
-    {
-        int total_size = total_width * rows;
-
-        if (!isContinuous())
-            CV_Error(CV_BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
-
-        if ((unsigned)new_rows > (unsigned)total_size)
-            CV_Error(CV_StsOutOfRange, "Bad new number of rows");
-
-        total_width = total_size / new_rows;
-
-        if (total_width * new_rows != total_size)
-            CV_Error(CV_StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
-
-        hdr.rows = new_rows;
-        hdr.step = total_width * elemSize1();
-    }
-
-    int new_width = total_width / new_cn;
-
-    if (new_width * new_cn != total_width)
-        CV_Error(CV_BadNumChannels, "The total width is not divisible by the new number of channels");
-
-    hdr.cols = new_width;
-    hdr.flags = (hdr.flags & ~CV_MAT_CN_MASK) | ((new_cn - 1) << CV_CN_SHIFT);
-
-    return hdr;
-}
-
-cv::Mat::Mat(const GpuMat& m) : flags(0), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0), datalimit(0), allocator(0), size(&rows)
-{
-    m.download(*this);
-}
-
-void cv::gpu::createContinuous(int rows, int cols, int type, GpuMat& m)
-{
-    int area = rows * cols;
-    if (m.empty() || m.type() != type || !m.isContinuous() || m.size().area() < area)
-        m.create(1, area, type);
-
-    m.cols = cols;
-    m.rows = rows;
-    m.step = m.elemSize() * cols;
-    m.flags |= Mat::CONTINUOUS_FLAG;
-}
-
-void cv::gpu::ensureSizeIsEnough(int rows, int cols, int type, GpuMat& m)
-{
-    if (m.empty() || m.type() != type || m.data != m.datastart)
-        m.create(rows, cols, type);
-    else
-    {
-        const size_t esz = m.elemSize();
-        const ptrdiff_t delta2 = m.dataend - m.datastart;
-
-        const size_t minstep = m.cols * esz;
-
-        Size wholeSize;
-        wholeSize.height = std::max(static_cast<int>((delta2 - minstep) / m.step + 1), m.rows);
-        wholeSize.width = std::max(static_cast<int>((delta2 - m.step * (wholeSize.height - 1)) / esz), m.cols);
-
-        if (wholeSize.height < rows || wholeSize.width < cols)
-            m.create(rows, cols, type);
-        else
-        {
-            m.cols = cols;
-            m.rows = rows;
-        }
-    }
-}
-
-GpuMat cv::gpu::allocMatFromBuf(int rows, int cols, int type, GpuMat &mat)
-{
-    if (!mat.empty() && mat.type() == type && mat.rows >= rows && mat.cols >= cols)
-        return mat(Rect(0, 0, cols, rows));
-    return mat = GpuMat(rows, cols, type);
-}
-
-namespace
-{
-    class GpuFuncTable
-    {
-    public:
-        virtual ~GpuFuncTable() {}
-
-        virtual void copy(const Mat& src, GpuMat& dst) const = 0;
-        virtual void copy(const GpuMat& src, Mat& dst) const = 0;
-        virtual void copy(const GpuMat& src, GpuMat& dst) const = 0;
-
-        virtual void copyWithMask(const GpuMat& src, GpuMat& dst, const GpuMat& mask) const = 0;
-
-        virtual void convert(const GpuMat& src, GpuMat& dst) const = 0;
-        virtual void convert(const GpuMat& src, GpuMat& dst, double alpha, double beta) const = 0;
-
-        virtual void setTo(GpuMat& m, Scalar s, const GpuMat& mask) const = 0;
-
-        virtual void mallocPitch(void** devPtr, size_t* step, size_t width, size_t height) const = 0;
-        virtual void free(void* devPtr) const = 0;
-    };
-}
-
-#ifndef HAVE_CUDA
-
-namespace
-{
-    class EmptyFuncTable : public GpuFuncTable
-    {
-    public:
-        void copy(const Mat&, GpuMat&) const { throw_nogpu; }
-        void copy(const GpuMat&, Mat&) const { throw_nogpu; }
-        void copy(const GpuMat&, GpuMat&) const { throw_nogpu; }
-
-        void copyWithMask(const GpuMat&, GpuMat&, const GpuMat&) const { throw_nogpu; }
-
-        void convert(const GpuMat&, GpuMat&) const { throw_nogpu; }
-        void convert(const GpuMat&, GpuMat&, double, double) const { throw_nogpu; }
-
-        void setTo(GpuMat&, Scalar, const GpuMat&) const { throw_nogpu; }
-
-        void mallocPitch(void**, size_t*, size_t, size_t) const { throw_nogpu; }
-        void free(void*) const {}
-    };
-
-    const GpuFuncTable* gpuFuncTable()
-    {
-        static EmptyFuncTable empty;
-        return &empty;
-    }
-}
-
-#else // HAVE_CUDA
-
-namespace cv { namespace gpu { namespace device
-{
-    void copyToWithMask_gpu(PtrStepSzb src, PtrStepSzb dst, size_t elemSize1, int cn, PtrStepSzb mask, bool colorMask, cudaStream_t stream);
-
-    template <typename T>
-    void set_to_gpu(PtrStepSzb mat, const T* scalar, int channels, cudaStream_t stream);
-
-    template <typename T>
-    void set_to_gpu(PtrStepSzb mat, const T* scalar, PtrStepSzb mask, int channels, cudaStream_t stream);
-
-    void convert_gpu(PtrStepSzb src, int sdepth, PtrStepSzb dst, int ddepth, double alpha, double beta, cudaStream_t stream);
-}}}
-
-namespace
-{
-    template <typename T> void kernelSetCaller(GpuMat& src, Scalar s, cudaStream_t stream)
-    {
-        Scalar_<T> sf = s;
-        cv::gpu::device::set_to_gpu(src, sf.val, src.channels(), stream);
-    }
-
-    template <typename T> void kernelSetCaller(GpuMat& src, Scalar s, const GpuMat& mask, cudaStream_t stream)
-    {
-        Scalar_<T> sf = s;
-        cv::gpu::device::set_to_gpu(src, sf.val, mask, src.channels(), stream);
-    }
-}
-
-
-namespace cv { namespace gpu
-{
-    CV_EXPORTS void copyWithMask(const cv::gpu::GpuMat&, cv::gpu::GpuMat&, const cv::gpu::GpuMat&, CUstream_st*);
-    CV_EXPORTS void convertTo(const cv::gpu::GpuMat&, cv::gpu::GpuMat&);
-    CV_EXPORTS void convertTo(const cv::gpu::GpuMat&, cv::gpu::GpuMat&, double, double, CUstream_st*);
-    CV_EXPORTS void setTo(cv::gpu::GpuMat&, cv::Scalar, CUstream_st*);
-    CV_EXPORTS void setTo(cv::gpu::GpuMat&, cv::Scalar, const cv::gpu::GpuMat&, CUstream_st*);
-    CV_EXPORTS void setTo(cv::gpu::GpuMat&, cv::Scalar);
-    CV_EXPORTS void setTo(cv::gpu::GpuMat&, cv::Scalar, const cv::gpu::GpuMat&);
-}}
-
-
-namespace cv { namespace gpu
-{
-    void copyWithMask(const GpuMat& src, GpuMat& dst, const GpuMat& mask, cudaStream_t stream = 0)
-    {
-        CV_Assert(src.size() == dst.size() && src.type() == dst.type());
-        CV_Assert(src.size() == mask.size() && mask.depth() == CV_8U && (mask.channels() == 1 || mask.channels() == src.channels()));
-
-        cv::gpu::device::copyToWithMask_gpu(src.reshape(1), dst.reshape(1), src.elemSize1(), src.channels(), mask.reshape(1), mask.channels() != 1, stream);
-    }
-
-    void convertTo(const GpuMat& src, GpuMat& dst)
-    {
-        cv::gpu::device::convert_gpu(src.reshape(1), src.depth(), dst.reshape(1), dst.depth(), 1.0, 0.0, 0);
-    }
-
-    void convertTo(const GpuMat& src, GpuMat& dst, double alpha, double beta, cudaStream_t stream = 0)
-    {
-        cv::gpu::device::convert_gpu(src.reshape(1), src.depth(), dst.reshape(1), dst.depth(), alpha, beta, stream);
-    }
-
-    void setTo(GpuMat& src, Scalar s, cudaStream_t stream)
-    {
-        typedef void (*caller_t)(GpuMat& src, Scalar s, cudaStream_t stream);
-
-        static const caller_t callers[] =
-        {
-            kernelSetCaller<uchar>, kernelSetCaller<schar>, kernelSetCaller<ushort>, kernelSetCaller<short>, kernelSetCaller<int>,
-            kernelSetCaller<float>, kernelSetCaller<double>
-        };
-
-        callers[src.depth()](src, s, stream);
-    }
-
-    void setTo(GpuMat& src, Scalar s, const GpuMat& mask, cudaStream_t stream)
-    {
-        typedef void (*caller_t)(GpuMat& src, Scalar s, const GpuMat& mask, cudaStream_t stream);
-
-        static const caller_t callers[] =
-        {
-            kernelSetCaller<uchar>, kernelSetCaller<schar>, kernelSetCaller<ushort>, kernelSetCaller<short>, kernelSetCaller<int>,
-            kernelSetCaller<float>, kernelSetCaller<double>
-        };
-
-        callers[src.depth()](src, s, mask, stream);
-    }
-
-    void setTo(GpuMat& src, Scalar s)
-    {
-        setTo(src, s, 0);
-    }
-
-    void setTo(GpuMat& src, Scalar s, const GpuMat& mask)
-    {
-        setTo(src, s, mask, 0);
-    }
-}}
-
-namespace
-{
-    template<int n> struct NPPTypeTraits;
-    template<> struct NPPTypeTraits<CV_8U>  { typedef Npp8u npp_type; };
-    template<> struct NPPTypeTraits<CV_8S>  { typedef Npp8s npp_type; };
-    template<> struct NPPTypeTraits<CV_16U> { typedef Npp16u npp_type; };
-    template<> struct NPPTypeTraits<CV_16S> { typedef Npp16s npp_type; };
-    template<> struct NPPTypeTraits<CV_32S> { typedef Npp32s npp_type; };
-    template<> struct NPPTypeTraits<CV_32F> { typedef Npp32f npp_type; };
-    template<> struct NPPTypeTraits<CV_64F> { typedef Npp64f npp_type; };
-
-    //////////////////////////////////////////////////////////////////////////
-    // Convert
-
-    template<int SDEPTH, int DDEPTH> struct NppConvertFunc
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-        typedef typename NPPTypeTraits<DDEPTH>::npp_type dst_t;
-
-        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, dst_t* pDst, int nDstStep, NppiSize oSizeROI);
-    };
-    template<int DDEPTH> struct NppConvertFunc<CV_32F, DDEPTH>
-    {
-        typedef typename NPPTypeTraits<DDEPTH>::npp_type dst_t;
-
-        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, dst_t* pDst, int nDstStep, NppiSize oSizeROI, NppRoundMode eRoundMode);
-    };
-
-    template<int SDEPTH, int DDEPTH, typename NppConvertFunc<SDEPTH, DDEPTH>::func_ptr func> struct NppCvt
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-        typedef typename NPPTypeTraits<DDEPTH>::npp_type dst_t;
-
-        static void call(const GpuMat& src, GpuMat& dst)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), dst.ptr<dst_t>(), static_cast<int>(dst.step), sz) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-    template<int DDEPTH, typename NppConvertFunc<CV_32F, DDEPTH>::func_ptr func> struct NppCvt<CV_32F, DDEPTH, func>
-    {
-        typedef typename NPPTypeTraits<DDEPTH>::npp_type dst_t;
-
-        static void call(const GpuMat& src, GpuMat& dst)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            nppSafeCall( func(src.ptr<Npp32f>(), static_cast<int>(src.step), dst.ptr<dst_t>(), static_cast<int>(dst.step), sz, NPP_RND_NEAR) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-
-    //////////////////////////////////////////////////////////////////////////
-    // Set
-
-    template<int SDEPTH, int SCN> struct NppSetFunc
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        typedef NppStatus (*func_ptr)(const src_t values[], src_t* pSrc, int nSrcStep, NppiSize oSizeROI);
-    };
-    template<int SDEPTH> struct NppSetFunc<SDEPTH, 1>
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        typedef NppStatus (*func_ptr)(src_t val, src_t* pSrc, int nSrcStep, NppiSize oSizeROI);
-    };
-    template<int SCN> struct NppSetFunc<CV_8S, SCN>
-    {
-        typedef NppStatus (*func_ptr)(Npp8s values[], Npp8s* pSrc, int nSrcStep, NppiSize oSizeROI);
-    };
-    template<> struct NppSetFunc<CV_8S, 1>
-    {
-        typedef NppStatus (*func_ptr)(Npp8s val, Npp8s* pSrc, int nSrcStep, NppiSize oSizeROI);
-    };
-
-    template<int SDEPTH, int SCN, typename NppSetFunc<SDEPTH, SCN>::func_ptr func> struct NppSet
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        static void call(GpuMat& src, Scalar s)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            Scalar_<src_t> nppS = s;
-
-            nppSafeCall( func(nppS.val, src.ptr<src_t>(), static_cast<int>(src.step), sz) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-    template<int SDEPTH, typename NppSetFunc<SDEPTH, 1>::func_ptr func> struct NppSet<SDEPTH, 1, func>
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        static void call(GpuMat& src, Scalar s)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            Scalar_<src_t> nppS = s;
-
-            nppSafeCall( func(nppS[0], src.ptr<src_t>(), static_cast<int>(src.step), sz) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-
-    template<int SDEPTH, int SCN> struct NppSetMaskFunc
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        typedef NppStatus (*func_ptr)(const src_t values[], src_t* pSrc, int nSrcStep, NppiSize oSizeROI, const Npp8u* pMask, int nMaskStep);
-    };
-    template<int SDEPTH> struct NppSetMaskFunc<SDEPTH, 1>
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        typedef NppStatus (*func_ptr)(src_t val, src_t* pSrc, int nSrcStep, NppiSize oSizeROI, const Npp8u* pMask, int nMaskStep);
-    };
-
-    template<int SDEPTH, int SCN, typename NppSetMaskFunc<SDEPTH, SCN>::func_ptr func> struct NppSetMask
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        static void call(GpuMat& src, Scalar s, const GpuMat& mask)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            Scalar_<src_t> nppS = s;
-
-            nppSafeCall( func(nppS.val, src.ptr<src_t>(), static_cast<int>(src.step), sz, mask.ptr<Npp8u>(), static_cast<int>(mask.step)) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-    template<int SDEPTH, typename NppSetMaskFunc<SDEPTH, 1>::func_ptr func> struct NppSetMask<SDEPTH, 1, func>
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        static void call(GpuMat& src, Scalar s, const GpuMat& mask)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            Scalar_<src_t> nppS = s;
-
-            nppSafeCall( func(nppS[0], src.ptr<src_t>(), static_cast<int>(src.step), sz, mask.ptr<Npp8u>(), static_cast<int>(mask.step)) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-
-    //////////////////////////////////////////////////////////////////////////
-    // CopyMasked
-
-    template<int SDEPTH> struct NppCopyMaskedFunc
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, src_t* pDst, int nDstStep, NppiSize oSizeROI, const Npp8u* pMask, int nMaskStep);
-    };
-
-    template<int SDEPTH, typename NppCopyMaskedFunc<SDEPTH>::func_ptr func> struct NppCopyMasked
-    {
-        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
-
-        static void call(const GpuMat& src, GpuMat& dst, const GpuMat& mask, cudaStream_t /*stream*/)
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), dst.ptr<src_t>(), static_cast<int>(dst.step), sz, mask.ptr<Npp8u>(), static_cast<int>(mask.step)) );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-
-    template <typename T> static inline bool isAligned(const T* ptr, size_t size)
-    {
-        return reinterpret_cast<size_t>(ptr) % size == 0;
-    }
-
-    //////////////////////////////////////////////////////////////////////////
-    // CudaFuncTable
-
-    class CudaFuncTable : public GpuFuncTable
-    {
-    public:
-        void copy(const Mat& src, GpuMat& dst) const
-        {
-            cudaSafeCall( cudaMemcpy2D(dst.data, dst.step, src.data, src.step, src.cols * src.elemSize(), src.rows, cudaMemcpyHostToDevice) );
-        }
-        void copy(const GpuMat& src, Mat& dst) const
-        {
-            cudaSafeCall( cudaMemcpy2D(dst.data, dst.step, src.data, src.step, src.cols * src.elemSize(), src.rows, cudaMemcpyDeviceToHost) );
-        }
-        void copy(const GpuMat& src, GpuMat& dst) const
-        {
-            cudaSafeCall( cudaMemcpy2D(dst.data, dst.step, src.data, src.step, src.cols * src.elemSize(), src.rows, cudaMemcpyDeviceToDevice) );
-        }
-
-        void copyWithMask(const GpuMat& src, GpuMat& dst, const GpuMat& mask) const
-        {
-            CV_Assert(src.depth() <= CV_64F && src.channels() <= 4);
-            CV_Assert(src.size() == dst.size() && src.type() == dst.type());
-            CV_Assert(src.size() == mask.size() && mask.depth() == CV_8U && (mask.channels() == 1 || mask.channels() == src.channels()));
-
-            if (src.depth() == CV_64F)
-            {
-                if (!TargetArchs::builtWith(NATIVE_DOUBLE) || !DeviceInfo().supports(NATIVE_DOUBLE))
-                    CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-            }
-
-            typedef void (*func_t)(const GpuMat& src, GpuMat& dst, const GpuMat& mask, cudaStream_t stream);
-            static const func_t funcs[7][4] =
-            {
-                /*  8U */ {NppCopyMasked<CV_8U , nppiCopy_8u_C1MR >::call, cv::gpu::copyWithMask, NppCopyMasked<CV_8U , nppiCopy_8u_C3MR >::call, NppCopyMasked<CV_8U , nppiCopy_8u_C4MR >::call},
-                /*  8S */ {cv::gpu::copyWithMask                         , cv::gpu::copyWithMask, cv::gpu::copyWithMask                         , cv::gpu::copyWithMask                         },
-                /* 16U */ {NppCopyMasked<CV_16U, nppiCopy_16u_C1MR>::call, cv::gpu::copyWithMask, NppCopyMasked<CV_16U, nppiCopy_16u_C3MR>::call, NppCopyMasked<CV_16U, nppiCopy_16u_C4MR>::call},
-                /* 16S */ {NppCopyMasked<CV_16S, nppiCopy_16s_C1MR>::call, cv::gpu::copyWithMask, NppCopyMasked<CV_16S, nppiCopy_16s_C3MR>::call, NppCopyMasked<CV_16S, nppiCopy_16s_C4MR>::call},
-                /* 32S */ {NppCopyMasked<CV_32S, nppiCopy_32s_C1MR>::call, cv::gpu::copyWithMask, NppCopyMasked<CV_32S, nppiCopy_32s_C3MR>::call, NppCopyMasked<CV_32S, nppiCopy_32s_C4MR>::call},
-                /* 32F */ {NppCopyMasked<CV_32F, nppiCopy_32f_C1MR>::call, cv::gpu::copyWithMask, NppCopyMasked<CV_32F, nppiCopy_32f_C3MR>::call, NppCopyMasked<CV_32F, nppiCopy_32f_C4MR>::call},
-                /* 64F */ {cv::gpu::copyWithMask                         , cv::gpu::copyWithMask, cv::gpu::copyWithMask                         , cv::gpu::copyWithMask                         }
-            };
-
-            const func_t func =  mask.channels() == src.channels() ? funcs[src.depth()][src.channels() - 1] : cv::gpu::copyWithMask;
-
-            func(src, dst, mask, 0);
-        }
-
-        void convert(const GpuMat& src, GpuMat& dst) const
-        {
-            typedef void (*func_t)(const GpuMat& src, GpuMat& dst);
-            static const func_t funcs[7][7][4] =
-            {
-                {
-                    /*  8U ->  8U */ {0, 0, 0, 0},
-                    /*  8U ->  8S */ {cv::gpu::convertTo                                , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /*  8U -> 16U */ {NppCvt<CV_8U, CV_16U, nppiConvert_8u16u_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, NppCvt<CV_8U, CV_16U, nppiConvert_8u16u_C4R>::call},
-                    /*  8U -> 16S */ {NppCvt<CV_8U, CV_16S, nppiConvert_8u16s_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, NppCvt<CV_8U, CV_16S, nppiConvert_8u16s_C4R>::call},
-                    /*  8U -> 32S */ {cv::gpu::convertTo                                , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /*  8U -> 32F */ {NppCvt<CV_8U, CV_32F, nppiConvert_8u32f_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /*  8U -> 64F */ {cv::gpu::convertTo                                , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                }
-                },
-                {
-                    /*  8S ->  8U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /*  8S ->  8S */ {0,0,0,0},
-                    /*  8S -> 16U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /*  8S -> 16S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /*  8S -> 32S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /*  8S -> 32F */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /*  8S -> 64F */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo}
-                },
-                {
-                    /* 16U ->  8U */ {NppCvt<CV_16U, CV_8U , nppiConvert_16u8u_C1R >::call, cv::gpu::convertTo, cv::gpu::convertTo, NppCvt<CV_16U, CV_8U, nppiConvert_16u8u_C4R>::call},
-                    /* 16U ->  8S */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16U -> 16U */ {0,0,0,0},
-                    /* 16U -> 16S */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16U -> 32S */ {NppCvt<CV_16U, CV_32S, nppiConvert_16u32s_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16U -> 32F */ {NppCvt<CV_16U, CV_32F, nppiConvert_16u32f_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16U -> 64F */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                }
-                },
-                {
-                    /* 16S ->  8U */ {NppCvt<CV_16S, CV_8U , nppiConvert_16s8u_C1R >::call, cv::gpu::convertTo, cv::gpu::convertTo, NppCvt<CV_16S, CV_8U, nppiConvert_16s8u_C4R>::call},
-                    /* 16S ->  8S */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16S -> 16U */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16S -> 16S */ {0,0,0,0},
-                    /* 16S -> 32S */ {NppCvt<CV_16S, CV_32S, nppiConvert_16s32s_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16S -> 32F */ {NppCvt<CV_16S, CV_32F, nppiConvert_16s32f_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                },
-                    /* 16S -> 64F */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo                                }
-                },
-                {
-                    /* 32S ->  8U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32S ->  8S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32S -> 16U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32S -> 16S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32S -> 32S */ {0,0,0,0},
-                    /* 32S -> 32F */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32S -> 64F */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo}
-                },
-                {
-                    /* 32F ->  8U */ {NppCvt<CV_32F, CV_8U , nppiConvert_32f8u_C1R >::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32F ->  8S */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32F -> 16U */ {NppCvt<CV_32F, CV_16U, nppiConvert_32f16u_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32F -> 16S */ {NppCvt<CV_32F, CV_16S, nppiConvert_32f16s_C1R>::call, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32F -> 32S */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 32F -> 32F */ {0,0,0,0},
-                    /* 32F -> 64F */ {cv::gpu::convertTo                                  , cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo}
-                },
-                {
-                    /* 64F ->  8U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F ->  8S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F -> 16U */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F -> 16S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F -> 32S */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F -> 32F */ {cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo, cv::gpu::convertTo},
-                    /* 64F -> 64F */ {0,0,0,0}
-                }
-            };
-
-            CV_Assert(src.depth() <= CV_64F && src.channels() <= 4);
-            CV_Assert(dst.depth() <= CV_64F);
-            CV_Assert(src.size() == dst.size() && src.channels() == dst.channels());
-
-            if (src.depth() == CV_64F || dst.depth() == CV_64F)
-            {
-                if (!TargetArchs::builtWith(NATIVE_DOUBLE) || !DeviceInfo().supports(NATIVE_DOUBLE))
-                    CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-            }
-
-            bool aligned = isAligned(src.data, 16) && isAligned(dst.data, 16);
-            if (!aligned)
-            {
-                cv::gpu::convertTo(src, dst);
-                return;
-            }
-
-            const func_t func = funcs[src.depth()][dst.depth()][src.channels() - 1];
-            CV_DbgAssert(func != 0);
-
-            func(src, dst);
-        }
-
-        void convert(const GpuMat& src, GpuMat& dst, double alpha, double beta) const
-        {
-            CV_Assert(src.depth() <= CV_64F && src.channels() <= 4);
-            CV_Assert(dst.depth() <= CV_64F);
-
-            if (src.depth() == CV_64F || dst.depth() == CV_64F)
-            {
-                if (!TargetArchs::builtWith(NATIVE_DOUBLE) || !DeviceInfo().supports(NATIVE_DOUBLE))
-                    CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-            }
-
-            cv::gpu::convertTo(src, dst, alpha, beta);
-        }
-
-        void setTo(GpuMat& m, Scalar s, const GpuMat& mask) const
-        {
-            if (mask.empty())
-            {
-                if (s[0] == 0.0 && s[1] == 0.0 && s[2] == 0.0 && s[3] == 0.0)
-                {
-                    cudaSafeCall( cudaMemset2D(m.data, m.step, 0, m.cols * m.elemSize(), m.rows) );
-                    return;
-                }
-
-                if (m.depth() == CV_8U)
-                {
-                    int cn = m.channels();
-
-                    if (cn == 1 || (cn == 2 && s[0] == s[1]) || (cn == 3 && s[0] == s[1] && s[0] == s[2]) || (cn == 4 && s[0] == s[1] && s[0] == s[2] && s[0] == s[3]))
-                    {
-                        int val = saturate_cast<uchar>(s[0]);
-                        cudaSafeCall( cudaMemset2D(m.data, m.step, val, m.cols * m.elemSize(), m.rows) );
-                        return;
-                    }
-                }
-
-                typedef void (*func_t)(GpuMat& src, Scalar s);
-                static const func_t funcs[7][4] =
-                {
-                    {NppSet<CV_8U , 1, nppiSet_8u_C1R >::call, cv::gpu::setTo                          , cv::gpu::setTo                        , NppSet<CV_8U , 4, nppiSet_8u_C4R >::call},
-                    {NppSet<CV_8S , 1, nppiSet_8s_C1R >::call, NppSet<CV_8S , 2, nppiSet_8s_C2R >::call, NppSet<CV_8S, 3, nppiSet_8s_C3R>::call, NppSet<CV_8S , 4, nppiSet_8s_C4R >::call},
-                    {NppSet<CV_16U, 1, nppiSet_16u_C1R>::call, NppSet<CV_16U, 2, nppiSet_16u_C2R>::call, cv::gpu::setTo                        , NppSet<CV_16U, 4, nppiSet_16u_C4R>::call},
-                    {NppSet<CV_16S, 1, nppiSet_16s_C1R>::call, NppSet<CV_16S, 2, nppiSet_16s_C2R>::call, cv::gpu::setTo                        , NppSet<CV_16S, 4, nppiSet_16s_C4R>::call},
-                    {NppSet<CV_32S, 1, nppiSet_32s_C1R>::call, cv::gpu::setTo                          , cv::gpu::setTo                        , NppSet<CV_32S, 4, nppiSet_32s_C4R>::call},
-                    {NppSet<CV_32F, 1, nppiSet_32f_C1R>::call, cv::gpu::setTo                          , cv::gpu::setTo                        , NppSet<CV_32F, 4, nppiSet_32f_C4R>::call},
-                    {cv::gpu::setTo                          , cv::gpu::setTo                          , cv::gpu::setTo                        , cv::gpu::setTo                          }
-                };
-
-                CV_Assert(m.depth() <= CV_64F && m.channels() <= 4);
-
-                if (m.depth() == CV_64F)
-                {
-                    if (!TargetArchs::builtWith(NATIVE_DOUBLE) || !DeviceInfo().supports(NATIVE_DOUBLE))
-                        CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-                }
-
-                funcs[m.depth()][m.channels() - 1](m, s);
-            }
-            else
-            {
-                typedef void (*func_t)(GpuMat& src, Scalar s, const GpuMat& mask);
-                static const func_t funcs[7][4] =
-                {
-                    {NppSetMask<CV_8U , 1, nppiSet_8u_C1MR >::call, cv::gpu::setTo, cv::gpu::setTo, NppSetMask<CV_8U , 4, nppiSet_8u_C4MR >::call},
-                    {cv::gpu::setTo                               , cv::gpu::setTo, cv::gpu::setTo, cv::gpu::setTo                               },
-                    {NppSetMask<CV_16U, 1, nppiSet_16u_C1MR>::call, cv::gpu::setTo, cv::gpu::setTo, NppSetMask<CV_16U, 4, nppiSet_16u_C4MR>::call},
-                    {NppSetMask<CV_16S, 1, nppiSet_16s_C1MR>::call, cv::gpu::setTo, cv::gpu::setTo, NppSetMask<CV_16S, 4, nppiSet_16s_C4MR>::call},
-                    {NppSetMask<CV_32S, 1, nppiSet_32s_C1MR>::call, cv::gpu::setTo, cv::gpu::setTo, NppSetMask<CV_32S, 4, nppiSet_32s_C4MR>::call},
-                    {NppSetMask<CV_32F, 1, nppiSet_32f_C1MR>::call, cv::gpu::setTo, cv::gpu::setTo, NppSetMask<CV_32F, 4, nppiSet_32f_C4MR>::call},
-                    {cv::gpu::setTo                               , cv::gpu::setTo, cv::gpu::setTo, cv::gpu::setTo                               }
-                };
-
-                CV_Assert(m.depth() <= CV_64F && m.channels() <= 4);
-
-                if (m.depth() == CV_64F)
-                {
-                    if (!TargetArchs::builtWith(NATIVE_DOUBLE) || !DeviceInfo().supports(NATIVE_DOUBLE))
-                        CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-                }
-
-                funcs[m.depth()][m.channels() - 1](m, s, mask);
-            }
-        }
-
-        void mallocPitch(void** devPtr, size_t* step, size_t width, size_t height) const
-        {
-            cudaSafeCall( cudaMallocPitch(devPtr, step, width, height) );
-        }
-
-        void free(void* devPtr) const
-        {
-            cudaFree(devPtr);
-        }
-    };
-
-    const GpuFuncTable* gpuFuncTable()
-    {
-        static CudaFuncTable funcTable;
-        return &funcTable;
-    }
-}
-
-#endif // HAVE_CUDA
-
-void cv::gpu::GpuMat::upload(const Mat& m)
-{
-    CV_DbgAssert(!m.empty());
-
-    create(m.size(), m.type());
-
-    gpuFuncTable()->copy(m, *this);
-}
-
-void cv::gpu::GpuMat::download(Mat& m) const
-{
-    CV_DbgAssert(!empty());
-
-    m.create(size(), type());
-
-    gpuFuncTable()->copy(*this, m);
-}
-
-void cv::gpu::GpuMat::copyTo(GpuMat& m) const
-{
-    CV_DbgAssert(!empty());
-
-    m.create(size(), type());
-
-    gpuFuncTable()->copy(*this, m);
-}
-
-void cv::gpu::GpuMat::copyTo(GpuMat& mat, const GpuMat& mask) const
-{
-    if (mask.empty())
-        copyTo(mat);
-    else
-    {
-        mat.create(size(), type());
-
-        gpuFuncTable()->copyWithMask(*this, mat, mask);
-    }
-}
-
-void cv::gpu::GpuMat::convertTo(GpuMat& dst, int rtype, double alpha, double beta) const
-{
-    bool noScale = fabs(alpha - 1) < std::numeric_limits<double>::epsilon() && fabs(beta) < std::numeric_limits<double>::epsilon();
-
-    if (rtype < 0)
-        rtype = type();
-    else
-        rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), channels());
-
-    int sdepth = depth();
-    int ddepth = CV_MAT_DEPTH(rtype);
-    if (sdepth == ddepth && noScale)
-    {
-        copyTo(dst);
-        return;
-    }
-
-    GpuMat temp;
-    const GpuMat* psrc = this;
-    if (sdepth != ddepth && psrc == &dst)
-    {
-        temp = *this;
-        psrc = &temp;
-    }
-
-    dst.create(size(), rtype);
-
-    if (noScale)
-        gpuFuncTable()->convert(*psrc, dst);
-    else
-        gpuFuncTable()->convert(*psrc, dst, alpha, beta);
-}
-
-GpuMat& cv::gpu::GpuMat::setTo(Scalar s, const GpuMat& mask)
-{
-    CV_Assert(mask.empty() || mask.type() == CV_8UC1);
-    CV_DbgAssert(!empty());
-
-    gpuFuncTable()->setTo(*this, s, mask);
-
-    return *this;
-}
-
-void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
-{
-    _type &= TYPE_MASK;
-
-    if (rows == _rows && cols == _cols && type() == _type && data)
-        return;
-
-    if (data)
-        release();
-
-    CV_DbgAssert(_rows >= 0 && _cols >= 0);
-
-    if (_rows > 0 && _cols > 0)
-    {
-        flags = Mat::MAGIC_VAL + _type;
-        rows = _rows;
-        cols = _cols;
-
-        size_t esz = elemSize();
-
-        void* devPtr;
-        gpuFuncTable()->mallocPitch(&devPtr, &step, esz * cols, rows);
-
-        // Single row must be continuous
-        if (rows == 1)
-            step = esz * cols;
-
-        if (esz * cols == step)
-            flags |= Mat::CONTINUOUS_FLAG;
-
-        int64 _nettosize = static_cast<int64>(step) * rows;
-        size_t nettosize = static_cast<size_t>(_nettosize);
-
-        datastart = data = static_cast<uchar*>(devPtr);
-        dataend = data + nettosize;
-
-        refcount = static_cast<int*>(fastMalloc(sizeof(*refcount)));
-        *refcount = 1;
-    }
-}
-
-void cv::gpu::GpuMat::release()
-{
-    if (refcount && CV_XADD(refcount, -1) == 1)
-    {
-        fastFree(refcount);
-
-        gpuFuncTable()->free(datastart);
-    }
-
-    data = datastart = dataend = 0;
-    step = rows = cols = 0;
-    refcount = 0;
-}
-
-////////////////////////////////////////////////////////////////////////
-// Error handling
-
-void cv::gpu::error(const char *error_string, const char *file, const int line, const char *func)
-{
-    int code = CV_GpuApiCallError;
-
-    if (std::uncaught_exception())
-    {
-        const char* errorStr = cvErrorStr(code);
-        const char* function = func ? func : "unknown function";
-
-        std::cerr << "OpenCV Error: " << errorStr << "(" << error_string << ") in " << function << ", file " << file << ", line " << line;
-        std::cerr.flush();
-    }
-    else
-        cv::error( cv::Exception(code, error_string, func, file, line) );
-}
diff --git a/modules/core/src/lapack.cpp b/modules/core/src/lapack.cpp
index bf376c8753..f6bc7c88c9 100644
--- a/modules/core/src/lapack.cpp
+++ b/modules/core/src/lapack.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include <limits>
 
 #if defined _M_IX86 && defined _MSC_VER && _MSC_VER < 1700
 #pragma float_control(precise, on)
@@ -531,12 +532,12 @@ template<> inline int VBLAS<double>::givensx(double* a, double* b, int n, double
 #endif
 
 template<typename _Tp> void
-JacobiSVDImpl_(_Tp* At, size_t astep, _Tp* _W, _Tp* Vt, size_t vstep, int m, int n, int n1, double minval)
+JacobiSVDImpl_(_Tp* At, size_t astep, _Tp* _W, _Tp* Vt, size_t vstep,
+               int m, int n, int n1, double minval, _Tp eps)
 {
     VBLAS<_Tp> vblas;
     AutoBuffer<double> Wbuf(n);
     double* W = Wbuf;
-    _Tp eps = DBL_EPSILON*10;
     int i, j, k, iter, max_iter = std::max(m, 30);
     _Tp c, s;
     double sd;
@@ -577,10 +578,10 @@ JacobiSVDImpl_(_Tp* At, size_t astep, _Tp* _W, _Tp* Vt, size_t vstep, int m, int
                     continue;
 
                 p *= 2;
-                double beta = a - b, gamma = hypot((double)p, beta), delta;
+                double beta = a - b, gamma = hypot((double)p, beta);
                 if( beta < 0 )
                 {
-                    delta = (gamma - beta)*0.5;
+                    double delta = (gamma - beta)*0.5;
                     s = (_Tp)std::sqrt(delta/gamma);
                     c = (_Tp)(p/(gamma*s*2));
                 }
@@ -588,36 +589,18 @@ JacobiSVDImpl_(_Tp* At, size_t astep, _Tp* _W, _Tp* Vt, size_t vstep, int m, int
                 {
                     c = (_Tp)std::sqrt((gamma + beta)/(gamma*2));
                     s = (_Tp)(p/(gamma*c*2));
-                    delta = p*p*0.5/(gamma + beta);
                 }
 
-                W[i] += delta;
-                W[j] -= delta;
-
-                if( iter % 2 != 0 && W[i] > 0 && W[j] > 0 )
+                a = b = 0;
+                for( k = 0; k < m; k++ )
                 {
-                    k = vblas.givens(Ai, Aj, m, c, s);
+                    _Tp t0 = c*Ai[k] + s*Aj[k];
+                    _Tp t1 = -s*Ai[k] + c*Aj[k];
+                    Ai[k] = t0; Aj[k] = t1;
 
-                    for( ; k < m; k++ )
-                    {
-                        _Tp t0 = c*Ai[k] + s*Aj[k];
-                        _Tp t1 = -s*Ai[k] + c*Aj[k];
-                        Ai[k] = t0; Aj[k] = t1;
-                    }
-                }
-                else
-                {
-                    a = b = 0;
-                    for( k = 0; k < m; k++ )
-                    {
-                        _Tp t0 = c*Ai[k] + s*Aj[k];
-                        _Tp t1 = -s*Ai[k] + c*Aj[k];
-                        Ai[k] = t0; Aj[k] = t1;
-
-                        a += (double)t0*t0; b += (double)t1*t1;
-                    }
-                    W[i] = a; W[j] = b;
+                    a += (double)t0*t0; b += (double)t1*t1;
                 }
+                W[i] = a; W[j] = b;
 
                 changed = true;
 
@@ -729,12 +712,12 @@ JacobiSVDImpl_(_Tp* At, size_t astep, _Tp* _W, _Tp* Vt, size_t vstep, int m, int
 
 static void JacobiSVD(float* At, size_t astep, float* W, float* Vt, size_t vstep, int m, int n, int n1=-1)
 {
-    JacobiSVDImpl_(At, astep, W, Vt, vstep, m, n, !Vt ? 0 : n1 < 0 ? n : n1, FLT_MIN);
+    JacobiSVDImpl_(At, astep, W, Vt, vstep, m, n, !Vt ? 0 : n1 < 0 ? n : n1, FLT_MIN, FLT_EPSILON*2);
 }
 
 static void JacobiSVD(double* At, size_t astep, double* W, double* Vt, size_t vstep, int m, int n, int n1=-1)
 {
-    JacobiSVDImpl_(At, astep, W, Vt, vstep, m, n, !Vt ? 0 : n1 < 0 ? n : n1, DBL_MIN);
+    JacobiSVDImpl_(At, astep, W, Vt, vstep, m, n, !Vt ? 0 : n1 < 0 ? n : n1, DBL_MIN, DBL_EPSILON*10);
 }
 
 /* y[0:m,0:n] += diag(a[0:1,0:m]) * x[0:m,0:n] */
@@ -1482,7 +1465,7 @@ bool cv::solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int meth
 
 /////////////////// finding eigenvalues and eigenvectors of a symmetric matrix ///////////////
 
-bool cv::eigen( InputArray _src, bool computeEvects, OutputArray _evals, OutputArray _evects )
+bool cv::eigen( InputArray _src, OutputArray _evals, OutputArray _evects )
 {
     Mat src = _src.getMat();
     int type = src.type();
@@ -1492,7 +1475,7 @@ bool cv::eigen( InputArray _src, bool computeEvects, OutputArray _evals, OutputA
     CV_Assert (type == CV_32F || type == CV_64F);
 
     Mat v;
-    if( computeEvects )
+    if( _evects.needed() )
     {
         _evects.create(n, n, type);
         v = _evects.getMat();
@@ -1512,16 +1495,6 @@ bool cv::eigen( InputArray _src, bool computeEvects, OutputArray _evals, OutputA
     return ok;
 }
 
-bool cv::eigen( InputArray src, OutputArray evals, int, int )
-{
-    return eigen(src, false, evals, noArray());
-}
-
-bool cv::eigen( InputArray src, OutputArray evals, OutputArray evects, int, int)
-{
-    return eigen(src, true, evals, evects);
-}
-
 namespace cv
 {
 
@@ -1689,7 +1662,6 @@ cvDet( const CvArr* arr )
             if( rows == 3 )
                 return det3(Md);
         }
-        return cv::determinant(cv::Mat(mat));
     }
     return cv::determinant(cv::cvarrToMat(arr));
 }
@@ -1724,13 +1696,13 @@ cvSolve( const CvArr* Aarr, const CvArr* barr, CvArr* xarr, int method )
 
 CV_IMPL void
 cvEigenVV( CvArr* srcarr, CvArr* evectsarr, CvArr* evalsarr, double,
-           int lowindex, int highindex)
+           int, int )
 {
     cv::Mat src = cv::cvarrToMat(srcarr), evals0 = cv::cvarrToMat(evalsarr), evals = evals0;
     if( evectsarr )
     {
         cv::Mat evects0 = cv::cvarrToMat(evectsarr), evects = evects0;
-        eigen(src, evals, evects, lowindex, highindex);
+        eigen(src, evals, evects);
         if( evects0.data != evects.data )
         {
             uchar* p = evects0.data;
@@ -1739,7 +1711,7 @@ cvEigenVV( CvArr* srcarr, CvArr* evectsarr, CvArr* evalsarr, double,
         }
     }
     else
-        eigen(src, evals, lowindex, highindex);
+        eigen(src, evals);
     if( evals0.data != evals.data )
     {
         uchar* p = evals0.data;
@@ -1851,4 +1823,4 @@ cvSVBkSb( const CvArr* warr, const CvArr* uarr,
 
     cv::SVD::backSubst(w, u, v, rhs, dst);
     CV_Assert( dst.data == dst0.data );
-}
\ No newline at end of file
+}
diff --git a/modules/core/src/mathfuncs.cpp b/modules/core/src/mathfuncs.cpp
index 8e76d72355..2df111f0db 100644
--- a/modules/core/src/mathfuncs.cpp
+++ b/modules/core/src/mathfuncs.cpp
@@ -2350,7 +2350,7 @@ int cv::solveCubic( InputArray _coeffs, OutputArray _roots )
                 coeffs.size() == Size(1, n0) ||
                 coeffs.size() == Size(1, n0+1)) );
 
-    _roots.create(n0, 1, ctype, -1, true, DEPTH_MASK_FLT);
+    _roots.create(n0, 1, ctype, -1, true, _OutputArray::DEPTH_MASK_FLT);
     Mat roots = _roots.getMat();
 
     int i = -1, n = 0;
@@ -2482,7 +2482,7 @@ double cv::solvePoly( InputArray _coeffs0, OutputArray _roots0, int maxIters )
 
     int n = coeffs0.cols + coeffs0.rows - 2;
 
-    _roots0.create(n, 1, CV_MAKETYPE(cdepth, 2), -1, true, DEPTH_MASK_FLT);
+    _roots0.create(n, 1, CV_MAKETYPE(cdepth, 2), -1, true, _OutputArray::DEPTH_MASK_FLT);
     Mat roots0 = _roots0.getMat();
 
     AutoBuffer<C> buf(n*2+2);
@@ -2550,7 +2550,9 @@ cvSolveCubic( const CvMat* coeffs, CvMat* roots )
 
 void cvSolvePoly(const CvMat* a, CvMat *r, int maxiter, int)
 {
-    cv::Mat _a = cv::cvarrToMat(a), _r = cv::cvarrToMat(r), _r0 = r;
+    cv::Mat _a = cv::cvarrToMat(a);
+    cv::Mat _r = cv::cvarrToMat(r);
+    cv::Mat _r0 = _r;
     cv::solvePoly(_a, _r, maxiter);
     CV_Assert( _r.data == _r0.data ); // check that the array of roots was not reallocated
 }
diff --git a/modules/core/src/matmul.cpp b/modules/core/src/matmul.cpp
index 6b075a5b8e..7d832cb0fa 100644
--- a/modules/core/src/matmul.cpp
+++ b/modules/core/src/matmul.cpp
@@ -1725,19 +1725,29 @@ diagtransform_64f(const double* src, double* dst, const double* m, int len, int
 
 typedef void (*TransformFunc)( const uchar* src, uchar* dst, const uchar* m, int, int, int );
 
-static TransformFunc transformTab[] =
+static TransformFunc getTransformFunc(int depth)
 {
-    (TransformFunc)transform_8u, (TransformFunc)transform_8s, (TransformFunc)transform_16u,
-    (TransformFunc)transform_16s, (TransformFunc)transform_32s, (TransformFunc)transform_32f,
-    (TransformFunc)transform_64f, 0
-};
+    static TransformFunc transformTab[] =
+    {
+        (TransformFunc)transform_8u, (TransformFunc)transform_8s, (TransformFunc)transform_16u,
+        (TransformFunc)transform_16s, (TransformFunc)transform_32s, (TransformFunc)transform_32f,
+        (TransformFunc)transform_64f, 0
+    };
 
-static TransformFunc diagTransformTab[] =
+    return transformTab[depth];
+}
+
+static TransformFunc getDiagTransformFunc(int depth)
 {
-    (TransformFunc)diagtransform_8u, (TransformFunc)diagtransform_8s, (TransformFunc)diagtransform_16u,
-    (TransformFunc)diagtransform_16s, (TransformFunc)diagtransform_32s, (TransformFunc)diagtransform_32f,
-    (TransformFunc)diagtransform_64f, 0
-};
+    static TransformFunc diagTransformTab[] =
+    {
+        (TransformFunc)diagtransform_8u, (TransformFunc)diagtransform_8s, (TransformFunc)diagtransform_16u,
+        (TransformFunc)diagtransform_16s, (TransformFunc)diagtransform_32s, (TransformFunc)diagtransform_32f,
+        (TransformFunc)diagtransform_64f, 0
+    };
+
+    return diagTransformTab[depth];
+}
 
 }
 
@@ -1800,7 +1810,7 @@ void cv::transform( InputArray _src, OutputArray _dst, InputArray _mtx )
         }
     }
 
-    TransformFunc func = isDiag ? diagTransformTab[depth] : transformTab[depth];
+    TransformFunc func = isDiag ? getDiagTransformFunc(depth): getTransformFunc(depth);
     CV_Assert( func != 0 );
 
     const Mat* arrays[] = {&src, &dst, 0};
@@ -2308,11 +2318,6 @@ double cv::Mahalanobis( InputArray _v1, InputArray _v2, InputArray _icovar )
     return std::sqrt(result);
 }
 
-double cv::Mahalonobis( InputArray _v1, InputArray _v2, InputArray _icovar )
-{
-    return Mahalanobis(_v1, _v2, _icovar);
-}
-
 /****************************************************************************************\
 *                                        MulTransposed                                   *
 \****************************************************************************************/
@@ -2766,19 +2771,24 @@ static double dotProd_64f(const double* src1, const double* src2, int len)
 
 typedef double (*DotProdFunc)(const uchar* src1, const uchar* src2, int len);
 
-static DotProdFunc dotProdTab[] =
+static DotProdFunc getDotProdFunc(int depth)
 {
-    (DotProdFunc)GET_OPTIMIZED(dotProd_8u), (DotProdFunc)GET_OPTIMIZED(dotProd_8s),
-    (DotProdFunc)dotProd_16u, (DotProdFunc)dotProd_16s,
-    (DotProdFunc)dotProd_32s, (DotProdFunc)GET_OPTIMIZED(dotProd_32f),
-    (DotProdFunc)dotProd_64f, 0
-};
+    static DotProdFunc dotProdTab[] =
+    {
+        (DotProdFunc)GET_OPTIMIZED(dotProd_8u), (DotProdFunc)GET_OPTIMIZED(dotProd_8s),
+        (DotProdFunc)dotProd_16u, (DotProdFunc)dotProd_16s,
+        (DotProdFunc)dotProd_32s, (DotProdFunc)GET_OPTIMIZED(dotProd_32f),
+        (DotProdFunc)dotProd_64f, 0
+    };
+
+    return dotProdTab[depth];
+}
 
 double Mat::dot(InputArray _mat) const
 {
     Mat mat = _mat.getMat();
     int cn = channels();
-    DotProdFunc func = dotProdTab[depth()];
+    DotProdFunc func = getDotProdFunc(depth());
     CV_Assert( mat.type() == type() && mat.size == size && func != 0 );
 
     if( isContinuous() && mat.isContinuous() )
@@ -2857,6 +2867,7 @@ PCA& PCA::operator()(InputArray _data, InputArray __mean, int flags, int maxComp
     {
         CV_Assert( _mean.size() == mean_sz );
         _mean.convertTo(mean, ctype);
+        covar_flags |= CV_COVAR_USE_AVG;
     }
 
     calcCovarMatrix( data, covar, mean, covar_flags, ctype );
@@ -2900,6 +2911,57 @@ PCA& PCA::operator()(InputArray _data, InputArray __mean, int flags, int maxComp
     return *this;
 }
 
+void PCA::write(FileStorage& fs ) const
+{
+    CV_Assert( fs.isOpened() );
+
+    fs << "name" << "PCA";
+    fs << "vectors" << eigenvectors;
+    fs << "values" << eigenvalues;
+    fs << "mean" << mean;
+}
+
+void PCA::read(const FileNode& fs)
+{
+    CV_Assert( !fs.empty() );
+    String name = (String)fs["name"];
+    CV_Assert( name == "PCA" );
+
+    cv::read(fs["vectors"], eigenvectors);
+    cv::read(fs["values"], eigenvalues);
+    cv::read(fs["mean"], mean);
+}
+
+template <typename T>
+int computeCumulativeEnergy(const Mat& eigenvalues, double retainedVariance)
+{
+    CV_DbgAssert( eigenvalues.type() == DataType<T>::type );
+
+    Mat g(eigenvalues.size(), DataType<T>::type);
+
+    for(int ig = 0; ig < g.rows; ig++)
+    {
+        g.at<T>(ig, 0) = 0;
+        for(int im = 0; im <= ig; im++)
+        {
+            g.at<T>(ig,0) += eigenvalues.at<T>(im,0);
+        }
+    }
+
+    int L;
+
+    for(L = 0; L < eigenvalues.rows; L++)
+    {
+        double energy = g.at<T>(L, 0) / g.at<T>(g.rows - 1, 0);
+        if(energy > retainedVariance)
+            break;
+    }
+
+    L = std::max(2, L);
+
+    return L;
+}
+
 PCA& PCA::operator()(InputArray _data, InputArray __mean, int flags, double retainedVariance)
 {
     Mat data = _data.getMat(), _mean = __mean.getMat();
@@ -2976,26 +3038,11 @@ PCA& PCA::operator()(InputArray _data, InputArray __mean, int flags, double reta
     }
 
     // compute the cumulative energy content for each eigenvector
-    Mat g(eigenvalues.size(), ctype);
-
-    for(int ig = 0; ig < g.rows; ig++)
-    {
-        g.at<float>(ig,0) = 0;
-        for(int im = 0; im <= ig; im++)
-        {
-            g.at<float>(ig,0) += eigenvalues.at<float>(im,0);
-        }
-    }
-
     int L;
-    for(L = 0; L < eigenvalues.rows; L++)
-    {
-        double energy = g.at<float>(L, 0) / g.at<float>(g.rows - 1, 0);
-        if(energy > retainedVariance)
-            break;
-    }
-
-    L = std::max(2, L);
+    if (ctype == CV_32F)
+        L = computeCumulativeEnergy<float>(eigenvalues, retainedVariance);
+    else
+        L = computeCumulativeEnergy<double>(eigenvalues, retainedVariance);
 
     // use clone() to physically copy the data and thus deallocate the original matrices
     eigenvalues = eigenvalues.rowRange(0,L).clone();
diff --git a/modules/core/src/matop.cpp b/modules/core/src/matop.cpp
index 7d43a8f029..016435650a 100644
--- a/modules/core/src/matop.cpp
+++ b/modules/core/src/matop.cpp
@@ -220,6 +220,10 @@ static inline bool isInitializer(const MatExpr& e) { return e.op == &g_MatOp_Ini
 
 /////////////////////////////////////////////////////////////////////////////////////////////////////
 
+MatOp::MatOp() {}
+MatOp::~MatOp() {}
+
+
 bool MatOp::elementWise(const MatExpr& /*expr*/) const
 {
     return false;
@@ -320,7 +324,7 @@ void MatOp::augAssignXor(const MatExpr& expr, Mat& m) const
 {
     Mat temp;
     expr.op->assign(expr, temp);
-    m /= temp;
+    m ^= temp;
 }
 
 
diff --git a/modules/core/src/matrix.cpp b/modules/core/src/matrix.cpp
index 8336a28683..d2032b2e5c 100644
--- a/modules/core/src/matrix.cpp
+++ b/modules/core/src/matrix.cpp
@@ -41,8 +41,6 @@
 //M*/
 
 #include "precomp.hpp"
-#include "opencv2/core/gpumat.hpp"
-#include "opencv2/core/opengl.hpp"
 
 /****************************************************************************************\
 *                           [scaled] Identity matrix initialization                      *
@@ -183,7 +181,7 @@ static void finalizeHdr(Mat& m)
 void Mat::create(int d, const int* _sizes, int _type)
 {
     int i;
-    CV_Assert(0 <= d && _sizes && d <= CV_MAX_DIM && _sizes);
+    CV_Assert(0 <= d && d <= CV_MAX_DIM && _sizes);
     _type = CV_MAT_TYPE(_type);
 
     if( data && (d == dims || (d == 1 && dims <= 2)) && _type == type() )
@@ -262,9 +260,10 @@ void Mat::deallocate()
 }
 
 
-Mat::Mat(const Mat& m, const Range& _rowRange, const Range& _colRange) : size(&rows)
+Mat::Mat(const Mat& m, const Range& _rowRange, const Range& _colRange)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
 {
-    initEmpty();
     CV_Assert( m.dims >= 2 );
     if( m.dims > 2 )
     {
@@ -335,9 +334,10 @@ Mat::Mat(const Mat& m, const Rect& roi)
 }
 
 
-Mat::Mat(int _dims, const int* _sizes, int _type, void* _data, const size_t* _steps) : size(&rows)
+Mat::Mat(int _dims, const int* _sizes, int _type, void* _data, const size_t* _steps)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
 {
-    initEmpty();
     flags |= CV_MAT_TYPE(_type);
     data = datastart = (uchar*)_data;
     setSize(*this, _dims, _sizes, _steps, true);
@@ -345,9 +345,10 @@ Mat::Mat(int _dims, const int* _sizes, int _type, void* _data, const size_t* _st
 }
 
 
-Mat::Mat(const Mat& m, const Range* ranges) : size(&rows)
+Mat::Mat(const Mat& m, const Range* ranges)
+    : flags(MAGIC_VAL), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0),
+      datalimit(0), allocator(0), size(&rows)
 {
-    initEmpty();
     int i, d = m.dims;
 
     CV_Assert(ranges);
@@ -371,13 +372,14 @@ Mat::Mat(const Mat& m, const Range* ranges) : size(&rows)
 }
 
 
-Mat::Mat(const CvMatND* m, bool copyData) : size(&rows)
+static Mat cvMatNDToMat(const CvMatND* m, bool copyData)
 {
-    initEmpty();
+    Mat thiz;
+
     if( !m )
-        return;
-    data = datastart = m->data.ptr;
-    flags |= CV_MAT_TYPE(m->type);
+        return thiz;
+    thiz.data = thiz.datastart = m->data.ptr;
+    thiz.flags |= CV_MAT_TYPE(m->type);
     int _sizes[CV_MAX_DIM];
     size_t _steps[CV_MAX_DIM];
 
@@ -388,16 +390,107 @@ Mat::Mat(const CvMatND* m, bool copyData) : size(&rows)
         _steps[i] = m->dim[i].step;
     }
 
-    setSize(*this, d, _sizes, _steps);
-    finalizeHdr(*this);
+    setSize(thiz, d, _sizes, _steps);
+    finalizeHdr(thiz);
 
     if( copyData )
     {
-        Mat temp(*this);
-        temp.copyTo(*this);
+        Mat temp(thiz);
+        thiz.release();
+        temp.copyTo(thiz);
     }
+
+    return thiz;
 }
 
+static Mat cvMatToMat(const CvMat* m, bool copyData)
+{
+    Mat thiz;
+
+    if( !m )
+        return thiz;
+
+    if( !copyData )
+    {
+        thiz.flags = Mat::MAGIC_VAL + (m->type & (CV_MAT_TYPE_MASK|CV_MAT_CONT_FLAG));
+        thiz.dims = 2;
+        thiz.rows = m->rows;
+        thiz.cols = m->cols;
+        thiz.data = thiz.datastart = m->data.ptr;
+        size_t esz = CV_ELEM_SIZE(m->type), minstep = thiz.cols*esz, _step = m->step;
+        if( _step == 0 )
+            _step = minstep;
+        thiz.datalimit = thiz.datastart + _step*thiz.rows;
+        thiz.dataend = thiz.datalimit - _step + minstep;
+        thiz.step[0] = _step; thiz.step[1] = esz;
+    }
+    else
+    {
+        thiz.data = thiz.datastart = thiz.dataend = 0;
+        Mat(m->rows, m->cols, m->type, m->data.ptr, m->step).copyTo(thiz);
+    }
+
+    return thiz;
+}
+
+
+static Mat iplImageToMat(const IplImage* img, bool copyData)
+{
+    Mat m;
+
+    if( !img )
+        return m;
+
+    m.dims = 2;
+    CV_DbgAssert(CV_IS_IMAGE(img) && img->imageData != 0);
+
+    int imgdepth = IPL2CV_DEPTH(img->depth);
+    size_t esz;
+    m.step[0] = img->widthStep;
+
+    if(!img->roi)
+    {
+        CV_Assert(img->dataOrder == IPL_DATA_ORDER_PIXEL);
+        m.flags = Mat::MAGIC_VAL + CV_MAKETYPE(imgdepth, img->nChannels);
+        m.rows = img->height;
+        m.cols = img->width;
+        m.datastart = m.data = (uchar*)img->imageData;
+        esz = CV_ELEM_SIZE(m.flags);
+    }
+    else
+    {
+        CV_Assert(img->dataOrder == IPL_DATA_ORDER_PIXEL || img->roi->coi != 0);
+        bool selectedPlane = img->roi->coi && img->dataOrder == IPL_DATA_ORDER_PLANE;
+        m.flags = Mat::MAGIC_VAL + CV_MAKETYPE(imgdepth, selectedPlane ? 1 : img->nChannels);
+        m.rows = img->roi->height;
+        m.cols = img->roi->width;
+        esz = CV_ELEM_SIZE(m.flags);
+        m.data = m.datastart = (uchar*)img->imageData +
+            (selectedPlane ? (img->roi->coi - 1)*m.step*img->height : 0) +
+            img->roi->yOffset*m.step[0] + img->roi->xOffset*esz;
+    }
+    m.datalimit = m.datastart + m.step.p[0]*m.rows;
+    m.dataend = m.datastart + m.step.p[0]*(m.rows-1) + esz*m.cols;
+    m.flags |= (m.cols*esz == m.step.p[0] || m.rows == 1 ? Mat::CONTINUOUS_FLAG : 0);
+    m.step[1] = esz;
+
+    if( copyData )
+    {
+        Mat m2 = m;
+        m.release();
+        if( !img->roi || !img->roi->coi ||
+            img->dataOrder == IPL_DATA_ORDER_PLANE)
+            m2.copyTo(m);
+        else
+        {
+            int ch[] = {img->roi->coi - 1, 0};
+            m.create(m2.rows, m2.cols, m2.type());
+            mixChannels(&m2, 1, &m, 1, ch, 1);
+        }
+    }
+
+    return m;
+}
 
 Mat Mat::diag(int d) const
 {
@@ -433,101 +526,6 @@ Mat Mat::diag(int d) const
     return m;
 }
 
-
-Mat::Mat(const CvMat* m, bool copyData) : size(&rows)
-{
-    initEmpty();
-
-    if( !m )
-        return;
-
-    if( !copyData )
-    {
-        flags = MAGIC_VAL + (m->type & (CV_MAT_TYPE_MASK|CV_MAT_CONT_FLAG));
-        dims = 2;
-        rows = m->rows;
-        cols = m->cols;
-        data = datastart = m->data.ptr;
-        size_t esz = CV_ELEM_SIZE(m->type), minstep = cols*esz, _step = m->step;
-        if( _step == 0 )
-            _step = minstep;
-        datalimit = datastart + _step*rows;
-        dataend = datalimit - _step + minstep;
-        step[0] = _step; step[1] = esz;
-    }
-    else
-    {
-        data = datastart = dataend = 0;
-        Mat(m->rows, m->cols, m->type, m->data.ptr, m->step).copyTo(*this);
-    }
-}
-
-
-Mat::Mat(const IplImage* img, bool copyData) : size(&rows)
-{
-    initEmpty();
-
-    if( !img )
-        return;
-
-    dims = 2;
-    CV_DbgAssert(CV_IS_IMAGE(img) && img->imageData != 0);
-
-    int imgdepth = IPL2CV_DEPTH(img->depth);
-    size_t esz;
-    step[0] = img->widthStep;
-
-    if(!img->roi)
-    {
-        CV_Assert(img->dataOrder == IPL_DATA_ORDER_PIXEL);
-        flags = MAGIC_VAL + CV_MAKETYPE(imgdepth, img->nChannels);
-        rows = img->height; cols = img->width;
-        datastart = data = (uchar*)img->imageData;
-        esz = CV_ELEM_SIZE(flags);
-    }
-    else
-    {
-        CV_Assert(img->dataOrder == IPL_DATA_ORDER_PIXEL || img->roi->coi != 0);
-        bool selectedPlane = img->roi->coi && img->dataOrder == IPL_DATA_ORDER_PLANE;
-        flags = MAGIC_VAL + CV_MAKETYPE(imgdepth, selectedPlane ? 1 : img->nChannels);
-        rows = img->roi->height; cols = img->roi->width;
-        esz = CV_ELEM_SIZE(flags);
-        data = datastart = (uchar*)img->imageData +
-            (selectedPlane ? (img->roi->coi - 1)*step*img->height : 0) +
-            img->roi->yOffset*step[0] + img->roi->xOffset*esz;
-    }
-    datalimit = datastart + step.p[0]*rows;
-    dataend = datastart + step.p[0]*(rows-1) + esz*cols;
-    flags |= (cols*esz == step.p[0] || rows == 1 ? CONTINUOUS_FLAG : 0);
-    step[1] = esz;
-
-    if( copyData )
-    {
-        Mat m = *this;
-        release();
-        if( !img->roi || !img->roi->coi ||
-            img->dataOrder == IPL_DATA_ORDER_PLANE)
-            m.copyTo(*this);
-        else
-        {
-            int ch[] = {img->roi->coi - 1, 0};
-            create(m.rows, m.cols, m.type());
-            mixChannels(&m, 1, this, 1, ch, 1);
-        }
-    }
-}
-
-
-Mat::operator IplImage() const
-{
-    CV_Assert( dims <= 2 );
-    IplImage img;
-    cvInitImageHeader(&img, size(), cvIplDepth(flags), channels());
-    cvSetData(&img, data, (int)step[0]);
-    return img;
-}
-
-
 void Mat::pop_back(size_t nelems)
 {
     CV_Assert( nelems <= (size_t)size.p[0] );
@@ -673,16 +671,16 @@ Mat cvarrToMat(const CvArr* arr, bool copyData,
 {
     if( !arr )
         return Mat();
-    if( CV_IS_MAT(arr) )
-        return Mat((const CvMat*)arr, copyData );
+    if( CV_IS_MAT_HDR_Z(arr) )
+        return cvMatToMat((const CvMat*)arr, copyData);
     if( CV_IS_MATND(arr) )
-        return Mat((const CvMatND*)arr, copyData );
+        return cvMatNDToMat((const CvMatND*)arr, copyData );
     if( CV_IS_IMAGE(arr) )
     {
         const IplImage* iplimg = (const IplImage*)arr;
         if( coiMode == 0 && iplimg->roi && iplimg->roi->coi > 0 )
             CV_Error(CV_BadCOI, "COI is not supported by the function");
-        return Mat(iplimg, copyData);
+        return iplImageToMat(iplimg, copyData);
     }
     if( CV_IS_SEQ(arr) )
     {
@@ -836,6 +834,18 @@ Mat Mat::reshape(int new_cn, int new_rows) const
     return hdr;
 }
 
+Mat Mat::diag(const Mat& d)
+{
+    CV_Assert( d.cols == 1 || d.rows == 1 );
+    int len = d.rows + d.cols - 1;
+    Mat m(len, len, d.type(), Scalar(0));
+    Mat md = m.diag();
+    if( d.cols == 1 )
+        d.copyTo(md);
+    else
+        transpose(d, md);
+    return m;
+}
 
 int Mat::checkVector(int _elemChannels, int _depth, bool _requireContinuous) const
 {
@@ -929,14 +939,15 @@ void scalarToRawData(const Scalar& s, void* _buf, int type, int unroll_to)
 \*************************************************************************************************/
 
 _InputArray::_InputArray() : flags(0), obj(0) {}
-_InputArray::~_InputArray() {}
 _InputArray::_InputArray(const Mat& m) : flags(MAT), obj((void*)&m) {}
 _InputArray::_InputArray(const std::vector<Mat>& vec) : flags(STD_VECTOR_MAT), obj((void*)&vec) {}
 _InputArray::_InputArray(const double& val) : flags(FIXED_TYPE + FIXED_SIZE + MATX + CV_64F), obj((void*)&val), sz(Size(1,1)) {}
 _InputArray::_InputArray(const MatExpr& expr) : flags(FIXED_TYPE + FIXED_SIZE + EXPR), obj((void*)&expr) {}
 _InputArray::_InputArray(const gpu::GpuMat& d_mat) : flags(GPU_MAT), obj((void*)&d_mat) {}
 _InputArray::_InputArray(const ogl::Buffer& buf) : flags(OPENGL_BUFFER), obj((void*)&buf) {}
-_InputArray::_InputArray(const ogl::Texture2D& tex) : flags(OPENGL_TEXTURE), obj((void*)&tex) {}
+_InputArray::_InputArray(const gpu::CudaMem& cuda_mem) : flags(CUDA_MEM), obj((void*)&cuda_mem) {}
+
+_InputArray::~_InputArray() {}
 
 Mat _InputArray::getMat(int i) const
 {
@@ -984,14 +995,42 @@ Mat _InputArray::getMat(int i) const
         return !v.empty() ? Mat(size(i), t, (void*)&v[0]) : Mat();
     }
 
-    CV_Assert( k == STD_VECTOR_MAT );
-    //if( k == STD_VECTOR_MAT )
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
+
+    if( k == STD_VECTOR_MAT )
     {
         const std::vector<Mat>& v = *(const std::vector<Mat>*)obj;
         CV_Assert( 0 <= i && i < (int)v.size() );
 
         return v[i];
     }
+
+    if( k == OPENGL_BUFFER )
+    {
+        CV_Assert( i < 0 );
+        CV_Error(cv::Error::StsNotImplemented, "You should explicitly call mapHost/unmapHost methods for ogl::Buffer object");
+        return Mat();
+    }
+
+    if( k == GPU_MAT )
+    {
+        CV_Assert( i < 0 );
+        CV_Error(cv::Error::StsNotImplemented, "You should explicitly call download method for gpu::GpuMat object");
+        return Mat();
+    }
+
+    CV_Assert( k == CUDA_MEM );
+    //if( k == CUDA_MEM )
+    {
+        CV_Assert( i < 0 );
+
+        const gpu::CudaMem* cuda_mem = (const gpu::CudaMem*)obj;
+
+        return cuda_mem->createMatHeader();
+    }
 }
 
 
@@ -1066,6 +1105,11 @@ void _InputArray::getMatVector(std::vector<Mat>& mv) const
         return;
     }
 
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
+
     CV_Assert( k == STD_VECTOR_MAT );
     //if( k == STD_VECTOR_MAT )
     {
@@ -1080,10 +1124,29 @@ gpu::GpuMat _InputArray::getGpuMat() const
 {
     int k = kind();
 
-    CV_Assert(k == GPU_MAT);
+    if (k == GPU_MAT)
+    {
+        const gpu::GpuMat* d_mat = (const gpu::GpuMat*)obj;
+        return *d_mat;
+    }
 
-    const gpu::GpuMat* d_mat = (const gpu::GpuMat*)obj;
-    return *d_mat;
+    if (k == CUDA_MEM)
+    {
+        const gpu::CudaMem* cuda_mem = (const gpu::CudaMem*)obj;
+        return cuda_mem->createGpuMatHeader();
+    }
+
+    if (k == OPENGL_BUFFER)
+    {
+        CV_Error(cv::Error::StsNotImplemented, "You should explicitly call mapDevice/unmapDevice methods for ogl::Buffer object");
+        return gpu::GpuMat();
+    }
+
+    if (k == NONE)
+        return gpu::GpuMat();
+
+    CV_Error(cv::Error::StsNotImplemented, "getGpuMat is available only for gpu::GpuMat and gpu::CudaMem");
+    return gpu::GpuMat();
 }
 
 ogl::Buffer _InputArray::getOGlBuffer() const
@@ -1096,16 +1159,6 @@ ogl::Buffer _InputArray::getOGlBuffer() const
     return *gl_buf;
 }
 
-ogl::Texture2D _InputArray::getOGlTexture2D() const
-{
-    int k = kind();
-
-    CV_Assert(k == OPENGL_TEXTURE);
-
-    const ogl::Texture2D* gl_tex = (const ogl::Texture2D*)obj;
-    return *gl_tex;
-}
-
 int _InputArray::kind() const
 {
     return flags & KIND_MASK;
@@ -1174,20 +1227,25 @@ Size _InputArray::size(int i) const
         return buf->size();
     }
 
-    if( k == OPENGL_TEXTURE )
-    {
-        CV_Assert( i < 0 );
-        const ogl::Texture2D* tex = (const ogl::Texture2D*)obj;
-        return tex->size();
-    }
-
-    CV_Assert( k == GPU_MAT );
-    //if( k == GPU_MAT )
+    if( k == GPU_MAT )
     {
         CV_Assert( i < 0 );
         const gpu::GpuMat* d_mat = (const gpu::GpuMat*)obj;
         return d_mat->size();
     }
+
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
+
+    CV_Assert( k == CUDA_MEM );
+    //if( k == CUDA_MEM )
+    {
+        CV_Assert( i < 0 );
+        const gpu::CudaMem* cuda_mem = (const gpu::CudaMem*)obj;
+        return cuda_mem->size();
+    }
 }
 
 size_t _InputArray::total(int i) const
@@ -1240,9 +1298,12 @@ int _InputArray::type(int i) const
     if( k == OPENGL_BUFFER )
         return ((const ogl::Buffer*)obj)->type();
 
-    CV_Assert( k == GPU_MAT );
-    //if( k == GPU_MAT )
+    if( k == GPU_MAT )
         return ((const gpu::GpuMat*)obj)->type();
+
+    CV_Assert( k == CUDA_MEM );
+    //if( k == CUDA_MEM )
+        return ((const gpu::CudaMem*)obj)->type();
 }
 
 int _InputArray::depth(int i) const
@@ -1292,29 +1353,34 @@ bool _InputArray::empty() const
     if( k == OPENGL_BUFFER )
         return ((const ogl::Buffer*)obj)->empty();
 
-    if( k == OPENGL_TEXTURE )
-        return ((const ogl::Texture2D*)obj)->empty();
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
 
-    CV_Assert( k == GPU_MAT );
-    //if( k == GPU_MAT )
+    if( k == GPU_MAT )
         return ((const gpu::GpuMat*)obj)->empty();
+
+    CV_Assert( k == CUDA_MEM );
+    //if( k == CUDA_MEM )
+        return ((const gpu::CudaMem*)obj)->empty();
 }
 
 
 _OutputArray::_OutputArray() {}
-_OutputArray::~_OutputArray() {}
 _OutputArray::_OutputArray(Mat& m) : _InputArray(m) {}
 _OutputArray::_OutputArray(std::vector<Mat>& vec) : _InputArray(vec) {}
 _OutputArray::_OutputArray(gpu::GpuMat& d_mat) : _InputArray(d_mat) {}
 _OutputArray::_OutputArray(ogl::Buffer& buf) : _InputArray(buf) {}
-_OutputArray::_OutputArray(ogl::Texture2D& tex) : _InputArray(tex) {}
+_OutputArray::_OutputArray(gpu::CudaMem& cuda_mem) : _InputArray(cuda_mem) {}
 
 _OutputArray::_OutputArray(const Mat& m) : _InputArray(m) {flags |= FIXED_SIZE|FIXED_TYPE;}
 _OutputArray::_OutputArray(const std::vector<Mat>& vec) : _InputArray(vec) {flags |= FIXED_SIZE;}
 _OutputArray::_OutputArray(const gpu::GpuMat& d_mat) : _InputArray(d_mat) {flags |= FIXED_SIZE|FIXED_TYPE;}
 _OutputArray::_OutputArray(const ogl::Buffer& buf) : _InputArray(buf) {flags |= FIXED_SIZE|FIXED_TYPE;}
-_OutputArray::_OutputArray(const ogl::Texture2D& tex) : _InputArray(tex) {flags |= FIXED_SIZE|FIXED_TYPE;}
+_OutputArray::_OutputArray(const gpu::CudaMem& cuda_mem) : _InputArray(cuda_mem) {flags |= FIXED_SIZE|FIXED_TYPE;}
 
+_OutputArray::~_OutputArray() {}
 
 bool _OutputArray::fixedSize() const
 {
@@ -1350,6 +1416,13 @@ void _OutputArray::create(Size _sz, int mtype, int i, bool allowTransposed, int
         ((ogl::Buffer*)obj)->create(_sz, mtype);
         return;
     }
+    if( k == CUDA_MEM && i < 0 && !allowTransposed && fixedDepthMask == 0 )
+    {
+        CV_Assert(!fixedSize() || ((gpu::CudaMem*)obj)->size() == _sz);
+        CV_Assert(!fixedType() || ((gpu::CudaMem*)obj)->type() == mtype);
+        ((gpu::CudaMem*)obj)->create(_sz, mtype);
+        return;
+    }
     int sizes[] = {_sz.height, _sz.width};
     create(2, sizes, mtype, i, allowTransposed, fixedDepthMask);
 }
@@ -1378,6 +1451,13 @@ void _OutputArray::create(int rows, int cols, int mtype, int i, bool allowTransp
         ((ogl::Buffer*)obj)->create(rows, cols, mtype);
         return;
     }
+    if( k == CUDA_MEM && i < 0 && !allowTransposed && fixedDepthMask == 0 )
+    {
+        CV_Assert(!fixedSize() || ((gpu::CudaMem*)obj)->size() == Size(cols, rows));
+        CV_Assert(!fixedType() || ((gpu::CudaMem*)obj)->type() == mtype);
+        ((gpu::CudaMem*)obj)->create(rows, cols, mtype);
+        return;
+    }
     int sizes[] = {rows, cols};
     create(2, sizes, mtype, i, allowTransposed, fixedDepthMask);
 }
@@ -1513,6 +1593,11 @@ void _OutputArray::create(int dims, const int* sizes, int mtype, int i, bool all
         return;
     }
 
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
+
     if( k == NONE )
     {
         CV_Error(CV_StsNullPtr, "create() called for the missing output array" );
@@ -1536,10 +1621,10 @@ void _OutputArray::create(int dims, const int* sizes, int mtype, int i, bool all
                 int _type = CV_MAT_TYPE(flags);
                 for( size_t j = len0; j < len; j++ )
                 {
-                    if( v[i].type() == _type )
+                    if( v[j].type() == _type )
                         continue;
-                    CV_Assert( v[i].empty() );
-                    v[i].flags = (v[i].flags & ~CV_MAT_TYPE_MASK) | _type;
+                    CV_Assert( v[j].empty() );
+                    v[j].flags = (v[j].flags & ~CV_MAT_TYPE_MASK) | _type;
                 }
             }
             return;
@@ -1597,15 +1682,15 @@ void _OutputArray::release() const
         return;
     }
 
-    if( k == OPENGL_BUFFER )
+    if( k == CUDA_MEM )
     {
-        ((ogl::Buffer*)obj)->release();
+        ((gpu::CudaMem*)obj)->release();
         return;
     }
 
-    if( k == OPENGL_TEXTURE )
+    if( k == OPENGL_BUFFER )
     {
-        ((ogl::Texture2D*)obj)->release();
+        ((ogl::Buffer*)obj)->release();
         return;
     }
 
@@ -1624,6 +1709,11 @@ void _OutputArray::release() const
         return;
     }
 
+    if( k == OCL_MAT )
+    {
+        CV_Error(CV_StsNotImplemented, "This method is not implemented for oclMat yet");
+    }
+
     CV_Assert( k == STD_VECTOR_MAT );
     //if( k == STD_VECTOR_MAT )
     {
@@ -1681,11 +1771,11 @@ ogl::Buffer& _OutputArray::getOGlBufferRef() const
     return *(ogl::Buffer*)obj;
 }
 
-ogl::Texture2D& _OutputArray::getOGlTexture2DRef() const
+gpu::CudaMem& _OutputArray::getCudaMemRef() const
 {
     int k = kind();
-    CV_Assert( k == OPENGL_TEXTURE );
-    return *(ogl::Texture2D*)obj;
+    CV_Assert( k == CUDA_MEM );
+    return *(gpu::CudaMem*)obj;
 }
 
 static _OutputArray _none;
@@ -2393,7 +2483,7 @@ template<typename T> static void sort_( const Mat& src, Mat& dst, int flags )
             for( j = 0; j < len; j++ )
                 ptr[j] = ((const T*)(src.data + src.step*j))[i];
         }
-        std::sort( ptr, ptr + len, LessThan<T>() );
+        std::sort( ptr, ptr + len );
         if( sortDescending )
             for( j = 0; j < len/2; j++ )
                 std::swap(ptr[j], ptr[len-1-j]);
@@ -2403,6 +2493,15 @@ template<typename T> static void sort_( const Mat& src, Mat& dst, int flags )
     }
 }
 
+template<typename _Tp> class LessThanIdx
+{
+public:
+    LessThanIdx( const _Tp* _arr ) : arr(_arr) {}
+    bool operator()(int a, int b) const { return arr[a] < arr[b]; }
+    const _Tp* arr;
+};
+
+
 
 template<typename T> static void sortIdx_( const Mat& src, Mat& dst, int flags )
 {
@@ -2929,7 +3028,7 @@ CV_IMPL void cvTranspose( const CvArr* srcarr, CvArr* dstarr )
 
 CV_IMPL void cvCompleteSymm( CvMat* matrix, int LtoR )
 {
-    cv::Mat m(matrix);
+    cv::Mat m = cv::cvarrToMat(matrix);
     cv::completeSymm( m, LtoR != 0 );
 }
 
@@ -3100,17 +3199,6 @@ Mat Mat::reshape(int _cn, int _newndims, const int* _newsz) const
     return Mat();
 }
 
-Mat::operator CvMatND() const
-{
-    CvMatND mat;
-    cvInitMatNDHeader( &mat, dims, size, type(), data );
-    int i, d = dims;
-    for( i = 0; i < d; i++ )
-        mat.dim[i].step = (int)step[i];
-    mat.type |= flags & CONTINUOUS_FLAG;
-    return mat;
-}
-
 NAryMatIterator::NAryMatIterator()
     : arrays(0), planes(0), ptrs(0), narrays(0), nplanes(0), size(0), iterdepth(0), idx(0)
 {
@@ -3412,16 +3500,6 @@ void MatConstIterator::seek(const int* _idx, bool relative)
     seek(ofs, relative);
 }
 
-ptrdiff_t operator - (const MatConstIterator& b, const MatConstIterator& a)
-{
-    if( a.m != b.m )
-        return INT_MAX;
-    if( a.sliceEnd == b.sliceEnd )
-        return (b.ptr - a.ptr)/b.elemSize;
-
-    return b.lpos() - a.lpos();
-}
-
 //////////////////////////////// SparseMat ////////////////////////////////
 
 template<typename T1, typename T2> void
@@ -3448,7 +3526,10 @@ convertScaleData_(const void* _from, void* _to, int cn, double alpha, double bet
             to[i] = saturate_cast<T2>(from[i]*alpha + beta);
 }
 
-ConvertData getConvertElem(int fromType, int toType)
+typedef void (*ConvertData)(const void* from, void* to, int cn);
+typedef void (*ConvertScaleData)(const void* from, void* to, int cn, double alpha, double beta);
+
+static ConvertData getConvertElem(int fromType, int toType)
 {
     static ConvertData tab[][8] =
     {{ convertData_<uchar, uchar>, convertData_<uchar, schar>,
@@ -3493,7 +3574,7 @@ ConvertData getConvertElem(int fromType, int toType)
     return func;
 }
 
-ConvertScaleData getConvertScaleElem(int fromType, int toType)
+static ConvertScaleData getConvertScaleElem(int fromType, int toType)
 {
     static ConvertScaleData tab[][8] =
     {{ convertScaleData_<uchar, uchar>, convertScaleData_<uchar, schar>,
@@ -3621,24 +3702,6 @@ SparseMat::SparseMat(const Mat& m)
     }
 }
 
-SparseMat::SparseMat(const CvSparseMat* m)
-: flags(MAGIC_VAL), hdr(0)
-{
-    CV_Assert(m);
-    create( m->dims, &m->size[0], m->type );
-
-    CvSparseMatIterator it;
-    CvSparseNode* n = cvInitSparseMatIterator(m, &it);
-    size_t esz = elemSize();
-
-    for( ; n != 0; n = cvGetNextSparseNode(&it) )
-    {
-        const int* idx = CV_NODE_IDX(m, n);
-        uchar* to = newNode(idx, hash(idx));
-        copyElem((const uchar*)CV_NODE_VAL(m, n), to, esz);
-    }
-}
-
 void SparseMat::create(int d, const int* _sizes, int _type)
 {
     int i;
@@ -3786,24 +3849,6 @@ void SparseMat::clear()
         hdr->clear();
 }
 
-SparseMat::operator CvSparseMat*() const
-{
-    if( !hdr )
-        return 0;
-    CvSparseMat* m = cvCreateSparseMat(hdr->dims, hdr->size, type());
-
-    SparseMatConstIterator from = begin();
-    size_t i, N = nzcount(), esz = elemSize();
-
-    for( i = 0; i < N; i++, ++from )
-    {
-        const Node* n = from.node();
-        uchar* to = cvPtrND(m, n->idx, 0, -2, 0);
-        copyElem(from.ptr, to, esz);
-    }
-    return m;
-}
-
 uchar* SparseMat::ptr(int i0, bool createMissing, size_t* hashval)
 {
     CV_Assert( hdr && hdr->dims == 1 );
@@ -4257,4 +4302,58 @@ Rect RotatedRect::boundingRect() const
 
 }
 
+// glue
+
+CvMatND::CvMatND(const cv::Mat& m)
+{
+    cvInitMatNDHeader(this, m.dims, m.size, m.type(), m.data );
+    int i, d = m.dims;
+    for( i = 0; i < d; i++ )
+        dim[i].step = (int)m.step[i];
+    type |= m.flags & cv::Mat::CONTINUOUS_FLAG;
+}
+
+_IplImage::_IplImage(const cv::Mat& m)
+{
+    CV_Assert( m.dims <= 2 );
+    cvInitImageHeader(this, m.size(), cvIplDepth(m.flags), m.channels());
+    cvSetData(this, m.data, (int)m.step[0]);
+}
+
+CvSparseMat* cvCreateSparseMat(const cv::SparseMat& sm)
+{
+    if( !sm.hdr )
+        return 0;
+
+    CvSparseMat* m = cvCreateSparseMat(sm.hdr->dims, sm.hdr->size, sm.type());
+
+    cv::SparseMatConstIterator from = sm.begin();
+    size_t i, N = sm.nzcount(), esz = sm.elemSize();
+
+    for( i = 0; i < N; i++, ++from )
+    {
+        const cv::SparseMat::Node* n = from.node();
+        uchar* to = cvPtrND(m, n->idx, 0, -2, 0);
+        cv::copyElem(from.ptr, to, esz);
+    }
+    return m;
+}
+
+void CvSparseMat::copyToSparseMat(cv::SparseMat& m) const
+{
+    m.create( dims, &size[0], type );
+
+    CvSparseMatIterator it;
+    CvSparseNode* n = cvInitSparseMatIterator(this, &it);
+    size_t esz = m.elemSize();
+
+    for( ; n != 0; n = cvGetNextSparseNode(&it) )
+    {
+        const int* idx = CV_NODE_IDX(this, n);
+        uchar* to = m.newNode(idx, m.hash(idx));
+        cv::copyElem((const uchar*)CV_NODE_VAL(this, n), to, esz);
+    }
+}
+
+
 /* End of file. */
diff --git a/modules/core/src/matrix_operations.cpp b/modules/core/src/matrix_operations.cpp
deleted file mode 100644
index 243d56e22a..0000000000
--- a/modules/core/src/matrix_operations.cpp
+++ /dev/null
@@ -1,298 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#include "opencv2/core/gpumat.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-cv::gpu::CudaMem::CudaMem()
-    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
-{
-}
-
-cv::gpu::CudaMem::CudaMem(int _rows, int _cols, int _type, int _alloc_type)
-    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
-{
-    if( _rows > 0 && _cols > 0 )
-        create( _rows, _cols, _type, _alloc_type);
-}
-
-cv::gpu::CudaMem::CudaMem(Size _size, int _type, int _alloc_type)
-    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
-{
-    if( _size.height > 0 && _size.width > 0 )
-        create( _size.height, _size.width, _type, _alloc_type);
-}
-
-cv::gpu::CudaMem::CudaMem(const CudaMem& m)
-    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), alloc_type(m.alloc_type)
-{
-    if( refcount )
-        CV_XADD(refcount, 1);
-}
-
-cv::gpu::CudaMem::CudaMem(const Mat& m, int _alloc_type)
-    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
-{
-    if( m.rows > 0 && m.cols > 0 )
-        create( m.size(), m.type(), _alloc_type);
-
-    Mat tmp = createMatHeader();
-    m.copyTo(tmp);
-}
-
-cv::gpu::CudaMem::~CudaMem()
-{
-    release();
-}
-
-CudaMem& cv::gpu::CudaMem::operator = (const CudaMem& m)
-{
-    if( this != &m )
-    {
-        if( m.refcount )
-            CV_XADD(m.refcount, 1);
-        release();
-        flags = m.flags;
-        rows = m.rows; cols = m.cols;
-        step = m.step; data = m.data;
-        datastart = m.datastart;
-        dataend = m.dataend;
-        refcount = m.refcount;
-        alloc_type = m.alloc_type;
-    }
-    return *this;
-}
-
-CudaMem cv::gpu::CudaMem::clone() const
-{
-    CudaMem m(size(), type(), alloc_type);
-    Mat to = m;
-    Mat from = *this;
-    from.copyTo(to);
-    return m;
-}
-
-void cv::gpu::CudaMem::create(Size _size, int _type, int _alloc_type)
-{
-    create(_size.height, _size.width, _type, _alloc_type);
-}
-
-Mat cv::gpu::CudaMem::createMatHeader() const
-{
-    return Mat(size(), type(), data, step);
-}
-
-cv::gpu::CudaMem::operator Mat() const
-{
-    return createMatHeader();
-}
-
-cv::gpu::CudaMem::operator GpuMat() const
-{
-    return createGpuMatHeader();
-}
-
-bool cv::gpu::CudaMem::isContinuous() const
-{
-    return (flags & Mat::CONTINUOUS_FLAG) != 0;
-}
-
-size_t cv::gpu::CudaMem::elemSize() const
-{
-    return CV_ELEM_SIZE(flags);
-}
-
-size_t cv::gpu::CudaMem::elemSize1() const
-{
-    return CV_ELEM_SIZE1(flags);
-}
-
-int cv::gpu::CudaMem::type() const
-{
-    return CV_MAT_TYPE(flags);
-}
-
-int cv::gpu::CudaMem::depth() const
-{
-    return CV_MAT_DEPTH(flags);
-}
-
-int cv::gpu::CudaMem::channels() const
-{
-    return CV_MAT_CN(flags);
-}
-
-size_t cv::gpu::CudaMem::step1() const
-{
-    return step/elemSize1();
-}
-
-Size cv::gpu::CudaMem::size() const
-{
-    return Size(cols, rows);
-}
-
-bool cv::gpu::CudaMem::empty() const
-{
-    return data == 0;
-}
-
-#if !defined (HAVE_CUDA)
-
-void cv::gpu::registerPageLocked(Mat&) { CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); }
-void cv::gpu::unregisterPageLocked(Mat&) { CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); }
-void cv::gpu::CudaMem::create(int /*_rows*/, int /*_cols*/, int /*_type*/, int /*type_alloc*/)
-{ CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); }
-bool cv::gpu::CudaMem::canMapHostMemory() { CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); return false; }
-void cv::gpu::CudaMem::release() { CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); }
-GpuMat cv::gpu::CudaMem::createGpuMatHeader () const { CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support"); return GpuMat(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-void cv::gpu::registerPageLocked(Mat& m)
-{
-    cudaSafeCall( cudaHostRegister(m.ptr(), m.step * m.rows, cudaHostRegisterPortable) );
-}
-
-void cv::gpu::unregisterPageLocked(Mat& m)
-{
-    cudaSafeCall( cudaHostUnregister(m.ptr()) );
-}
-
-bool cv::gpu::CudaMem::canMapHostMemory()
-{
-    cudaDeviceProp prop;
-    cudaSafeCall( cudaGetDeviceProperties(&prop, getDevice()) );
-    return (prop.canMapHostMemory != 0) ? true : false;
-}
-
-namespace
-{
-    size_t alignUpStep(size_t what, size_t alignment)
-    {
-        size_t alignMask = alignment-1;
-        size_t inverseAlignMask = ~alignMask;
-        size_t res = (what + alignMask) & inverseAlignMask;
-        return res;
-    }
-}
-
-void cv::gpu::CudaMem::create(int _rows, int _cols, int _type, int _alloc_type)
-{
-    if (_alloc_type == ALLOC_ZEROCOPY && !canMapHostMemory())
-            cv::gpu::error("ZeroCopy is not supported by current device", __FILE__, __LINE__);
-
-    _type &= TYPE_MASK;
-    if( rows == _rows && cols == _cols && type() == _type && data )
-        return;
-    if( data )
-        release();
-    CV_DbgAssert( _rows >= 0 && _cols >= 0 );
-    if( _rows > 0 && _cols > 0 )
-    {
-        flags = Mat::MAGIC_VAL + Mat::CONTINUOUS_FLAG + _type;
-        rows = _rows;
-        cols = _cols;
-        step = elemSize()*cols;
-        if (_alloc_type == ALLOC_ZEROCOPY)
-        {
-            cudaDeviceProp prop;
-            cudaSafeCall( cudaGetDeviceProperties(&prop, getDevice()) );
-            step = alignUpStep(step, prop.textureAlignment);
-        }
-        int64 _nettosize = (int64)step*rows;
-        size_t nettosize = (size_t)_nettosize;
-        if( _nettosize != (int64)nettosize )
-            CV_Error(CV_StsNoMem, "Too big buffer is allocated");
-        size_t datasize = alignSize(nettosize, (int)sizeof(*refcount));
-
-        //datastart = data = (uchar*)fastMalloc(datasize + sizeof(*refcount));
-        alloc_type = _alloc_type;
-        void *ptr = 0;
-
-        switch (alloc_type)
-        {
-            case ALLOC_PAGE_LOCKED:    cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocDefault) ); break;
-            case ALLOC_ZEROCOPY:       cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocMapped) );  break;
-            case ALLOC_WRITE_COMBINED: cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocWriteCombined) ); break;
-            default: cv::gpu::error("Invalid alloc type", __FILE__, __LINE__);
-        }
-
-        datastart = data =  (uchar*)ptr;
-        dataend = data + nettosize;
-
-        refcount = (int*)cv::fastMalloc(sizeof(*refcount));
-        *refcount = 1;
-    }
-}
-
-GpuMat cv::gpu::CudaMem::createGpuMatHeader () const
-{
-    GpuMat res;
-    if (alloc_type == ALLOC_ZEROCOPY)
-    {
-        void *pdev;
-        cudaSafeCall( cudaHostGetDevicePointer( &pdev, data, 0 ) );
-        res = GpuMat(rows, cols, type(), pdev, step);
-    }
-    else
-        cv::gpu::error("Zero-copy is not supported or memory was allocated without zero-copy flag", __FILE__, __LINE__);
-
-    return res;
-}
-
-void cv::gpu::CudaMem::release()
-{
-    if( refcount && CV_XADD(refcount, -1) == 1 )
-    {
-        cudaSafeCall( cudaFreeHost(datastart ) );
-        fastFree(refcount);
-    }
-    data = datastart = dataend = 0;
-    step = rows = cols = 0;
-    refcount = 0;
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/core/src/opengl_interop.cpp b/modules/core/src/opengl.cpp
similarity index 83%
rename from modules/core/src/opengl_interop.cpp
rename to modules/core/src/opengl.cpp
index fb4c53f629..36a437a2d6 100644
--- a/modules/core/src/opengl_interop.cpp
+++ b/modules/core/src/opengl.cpp
@@ -41,16 +41,12 @@
 //M*/
 
 #include "precomp.hpp"
-#include "opencv2/core/opengl.hpp"
-#include "opencv2/core/gpumat.hpp"
 
 #ifdef HAVE_OPENGL
-    #include "gl_core_3_1.hpp"
-
-    #ifdef HAVE_CUDA
-        #include <cuda_runtime.h>
-        #include <cuda_gl_interop.h>
-    #endif
+#  include "gl_core_3_1.hpp"
+#  ifdef HAVE_CUDA
+#    include <cuda_gl_interop.h>
+#  endif
 #endif
 
 using namespace cv;
@@ -59,62 +55,62 @@ using namespace cv::gpu;
 namespace
 {
     #ifndef HAVE_OPENGL
-        void throw_nogl() { CV_Error(CV_OpenGlNotSupported, "The library is compiled without OpenGL support"); }
+        inline void throw_no_ogl() { CV_Error(cv::Error::OpenGlNotSupported, "The library is compiled without OpenGL support"); }
     #else
-        void throw_nogl() { CV_Error(CV_OpenGlApiCallError, "OpenGL context doesn't exist"); }
-
-        #ifndef HAVE_CUDA
-            void throw_nocuda() { CV_Error(CV_GpuNotSupported, "The library is compiled without GPU support"); }
-        #else
-            void throw_nocuda() { CV_Error(CV_StsNotImplemented, "The called functionality is disabled for current build or platform"); }
-        #endif
+        inline void throw_no_ogl() { CV_Error(cv::Error::OpenGlApiCallError, "OpenGL context doesn't exist"); }
     #endif
-}
 
-bool cv::ogl::checkError(const char* file, const int line, const char* func)
-{
-#ifndef HAVE_OPENGL
-    (void) file;
-    (void) line;
-    (void) func;
-    return true;
-#else
-    GLenum err = gl::GetError();
-
-    if (err != gl::NO_ERROR_)
+    bool checkError(const char* file, const int line, const char* func = 0)
     {
-        const char* msg;
+    #ifndef HAVE_OPENGL
+        (void) file;
+        (void) line;
+        (void) func;
+        return true;
+    #else
+        GLenum err = gl::GetError();
 
-        switch (err)
+        if (err != gl::NO_ERROR_)
         {
-        case gl::INVALID_ENUM:
-            msg = "An unacceptable value is specified for an enumerated argument";
-            break;
+            const char* msg;
 
-        case gl::INVALID_VALUE:
-            msg = "A numeric argument is out of range";
-            break;
+            switch (err)
+            {
+            case gl::INVALID_ENUM:
+                msg = "An unacceptable value is specified for an enumerated argument";
+                break;
 
-        case gl::INVALID_OPERATION:
-            msg = "The specified operation is not allowed in the current state";
-            break;
+            case gl::INVALID_VALUE:
+                msg = "A numeric argument is out of range";
+                break;
 
-        case gl::OUT_OF_MEMORY:
-            msg = "There is not enough memory left to execute the command";
-            break;
+            case gl::INVALID_OPERATION:
+                msg = "The specified operation is not allowed in the current state";
+                break;
 
-        default:
-            msg = "Unknown error";
-        };
+            case gl::OUT_OF_MEMORY:
+                msg = "There is not enough memory left to execute the command";
+                break;
 
-        cvError(CV_OpenGlApiCallError, func, msg, file, line);
+            default:
+                msg = "Unknown error";
+            };
 
-        return false;
+            cvError(CV_OpenGlApiCallError, func, msg, file, line);
+
+            return false;
+        }
+
+        return true;
+    #endif
     }
 
-    return true;
-#endif
-}
+    #if defined(__GNUC__)
+        #define CV_CheckGlError() CV_DbgAssert( (checkError(__FILE__, __LINE__, __func__)) )
+    #else
+        #define CV_CheckGlError() CV_DbgAssert( (checkError(__FILE__, __LINE__)) )
+    #endif
+} // namespace
 
 #ifdef HAVE_OPENGL
 namespace
@@ -130,11 +126,11 @@ void cv::gpu::setGlDevice(int device)
 {
 #ifndef HAVE_OPENGL
     (void) device;
-    throw_nogl();
+    throw_no_ogl();
 #else
-    #if !defined(HAVE_CUDA) || defined(CUDA_DISABLER)
+    #ifndef HAVE_CUDA
         (void) device;
-        throw_nocuda();
+        throw_no_cuda();
     #else
         cudaSafeCall( cudaGLSetGLDevice(device) );
     #endif
@@ -144,7 +140,7 @@ void cv::gpu::setGlDevice(int device)
 ////////////////////////////////////////////////////////////////////////
 // CudaResource
 
-#if defined(HAVE_OPENGL) && defined(HAVE_CUDA) && !defined(CUDA_DISABLER)
+#if defined(HAVE_OPENGL) && defined(HAVE_CUDA)
 
 namespace
 {
@@ -356,12 +352,13 @@ const Ptr<cv::ogl::Buffer::Impl>& cv::ogl::Buffer::Impl::empty()
     return p;
 }
 
-cv::ogl::Buffer::Impl::Impl() : bufId_(0), autoRelease_(true)
+cv::ogl::Buffer::Impl::Impl() : bufId_(0), autoRelease_(false)
 {
 }
 
 cv::ogl::Buffer::Impl::Impl(GLuint abufId, bool autoRelease) : bufId_(abufId), autoRelease_(autoRelease)
 {
+    CV_Assert( gl::IsBuffer(abufId) == gl::TRUE_ );
 }
 
 cv::ogl::Buffer::Impl::Impl(GLsizeiptr size, const GLvoid* data, GLenum target, bool autoRelease) : bufId_(0), autoRelease_(autoRelease)
@@ -440,36 +437,38 @@ void cv::ogl::Buffer::Impl::unmapHost()
 }
 
 #ifdef HAVE_CUDA
-    void cv::ogl::Buffer::Impl::copyFrom(const void* src, size_t spitch, size_t width, size_t height, cudaStream_t stream)
-    {
-        cudaResource_.registerBuffer(bufId_);
-        cudaResource_.copyFrom(src, spitch, width, height, stream);
-    }
 
-    void cv::ogl::Buffer::Impl::copyTo(void* dst, size_t dpitch, size_t width, size_t height, cudaStream_t stream) const
-    {
-        cudaResource_.registerBuffer(bufId_);
-        cudaResource_.copyTo(dst, dpitch, width, height, stream);
-    }
+void cv::ogl::Buffer::Impl::copyFrom(const void* src, size_t spitch, size_t width, size_t height, cudaStream_t stream)
+{
+    cudaResource_.registerBuffer(bufId_);
+    cudaResource_.copyFrom(src, spitch, width, height, stream);
+}
 
-    void* cv::ogl::Buffer::Impl::mapDevice(cudaStream_t stream)
-    {
-        cudaResource_.registerBuffer(bufId_);
-        return cudaResource_.map(stream);
-    }
+void cv::ogl::Buffer::Impl::copyTo(void* dst, size_t dpitch, size_t width, size_t height, cudaStream_t stream) const
+{
+    cudaResource_.registerBuffer(bufId_);
+    cudaResource_.copyTo(dst, dpitch, width, height, stream);
+}
 
-    void cv::ogl::Buffer::Impl::unmapDevice(cudaStream_t stream)
-    {
-        cudaResource_.unmap(stream);
-    }
-#endif
+void* cv::ogl::Buffer::Impl::mapDevice(cudaStream_t stream)
+{
+    cudaResource_.registerBuffer(bufId_);
+    return cudaResource_.map(stream);
+}
+
+void cv::ogl::Buffer::Impl::unmapDevice(cudaStream_t stream)
+{
+    cudaResource_.unmap(stream);
+}
+
+#endif // HAVE_CUDA
 
 #endif // HAVE_OPENGL
 
 cv::ogl::Buffer::Buffer() : rows_(0), cols_(0), type_(0)
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_ = Impl::empty();
 #endif
@@ -483,9 +482,9 @@ cv::ogl::Buffer::Buffer(int arows, int acols, int atype, unsigned int abufId, bo
     (void) atype;
     (void) abufId;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
-    impl_ = new Impl(abufId, autoRelease);
+    impl_.reset(new Impl(abufId, autoRelease));
     rows_ = arows;
     cols_ = acols;
     type_ = atype;
@@ -499,61 +498,38 @@ cv::ogl::Buffer::Buffer(Size asize, int atype, unsigned int abufId, bool autoRel
     (void) atype;
     (void) abufId;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
-    impl_ = new Impl(abufId, autoRelease);
+    impl_.reset(new Impl(abufId, autoRelease));
     rows_ = asize.height;
     cols_ = asize.width;
     type_ = atype;
 #endif
 }
 
-cv::ogl::Buffer::Buffer(int arows, int acols, int atype, Target target, bool autoRelease) : rows_(0), cols_(0), type_(0)
-{
-    create(arows, acols, atype, target, autoRelease);
-}
-
-cv::ogl::Buffer::Buffer(Size asize, int atype, Target target, bool autoRelease) : rows_(0), cols_(0), type_(0)
-{
-    create(asize, atype, target, autoRelease);
-}
-
 cv::ogl::Buffer::Buffer(InputArray arr, Target target, bool autoRelease) : rows_(0), cols_(0), type_(0)
 {
 #ifndef HAVE_OPENGL
     (void) arr;
     (void) target;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
     switch (kind)
     {
     case _InputArray::OPENGL_BUFFER:
-        {
-            copyFrom(arr, target, autoRelease);
-            break;
-        }
-
-    case _InputArray::OPENGL_TEXTURE:
-        {
-            copyFrom(arr, target, autoRelease);
-            break;
-        }
-
     case _InputArray::GPU_MAT:
-        {
-            copyFrom(arr, target, autoRelease);
-            break;
-        }
+        copyFrom(arr, target, autoRelease);
+        break;
 
     default:
         {
             Mat mat = arr.getMat();
             CV_Assert( mat.isContinuous() );
             const GLsizeiptr asize = mat.rows * mat.cols * mat.elemSize();
-            impl_ = new Impl(asize, mat.data, target, autoRelease);
+            impl_.reset(new Impl(asize, mat.data, target, autoRelease));
             rows_ = mat.rows;
             cols_ = mat.cols;
             type_ = mat.type();
@@ -571,12 +547,12 @@ void cv::ogl::Buffer::create(int arows, int acols, int atype, Target target, boo
     (void) atype;
     (void) target;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     if (rows_ != arows || cols_ != acols || type_ != atype)
     {
         const GLsizeiptr asize = arows * acols * CV_ELEM_SIZE(atype);
-        impl_ = new Impl(asize, 0, target, autoRelease);
+        impl_.reset(new Impl(asize, 0, target, autoRelease));
         rows_ = arows;
         cols_ = acols;
         type_ = atype;
@@ -587,7 +563,7 @@ void cv::ogl::Buffer::create(int arows, int acols, int atype, Target target, boo
 void cv::ogl::Buffer::release()
 {
 #ifdef HAVE_OPENGL
-    if (*impl_.refcount == 1)
+    if (impl_)
         impl_->setAutoRelease(true);
     impl_ = Impl::empty();
     rows_ = 0;
@@ -600,7 +576,7 @@ void cv::ogl::Buffer::setAutoRelease(bool flag)
 {
 #ifndef HAVE_OPENGL
     (void) flag;
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_->setAutoRelease(flag);
 #endif
@@ -612,18 +588,10 @@ void cv::ogl::Buffer::copyFrom(InputArray arr, Target target, bool autoRelease)
     (void) arr;
     (void) target;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
-    if (kind == _InputArray::OPENGL_TEXTURE)
-    {
-        ogl::Texture2D tex = arr.getOGlTexture2D();
-        tex.copyTo(*this);
-        setAutoRelease(autoRelease);
-        return;
-    }
-
     const Size asize = arr.size();
     const int atype = arr.type();
     create(asize, atype, target, autoRelease);
@@ -639,8 +607,8 @@ void cv::ogl::Buffer::copyFrom(InputArray arr, Target target, bool autoRelease)
 
     case _InputArray::GPU_MAT:
         {
-            #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-                throw_nocuda();
+            #ifndef HAVE_CUDA
+                throw_no_cuda();
             #else
                 GpuMat dmat = arr.getGpuMat();
                 impl_->copyFrom(dmat.data, dmat.step, dmat.cols * dmat.elemSize(), dmat.rows);
@@ -659,13 +627,36 @@ void cv::ogl::Buffer::copyFrom(InputArray arr, Target target, bool autoRelease)
 #endif
 }
 
-void cv::ogl::Buffer::copyTo(OutputArray arr, Target target, bool autoRelease) const
+void cv::ogl::Buffer::copyFrom(InputArray arr, gpu::Stream& stream, Target target, bool autoRelease)
 {
 #ifndef HAVE_OPENGL
     (void) arr;
+    (void) stream;
     (void) target;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
+#else
+    #ifndef HAVE_CUDA
+        (void) arr;
+        (void) stream;
+        (void) target;
+        (void) autoRelease;
+        throw_no_cuda();
+    #else
+        GpuMat dmat = arr.getGpuMat();
+
+        create(dmat.size(), dmat.type(), target, autoRelease);
+
+        impl_->copyFrom(dmat.data, dmat.step, dmat.cols * dmat.elemSize(), dmat.rows, gpu::StreamAccessor::getStream(stream));
+    #endif
+#endif
+}
+
+void cv::ogl::Buffer::copyTo(OutputArray arr) const
+{
+#ifndef HAVE_OPENGL
+    (void) arr;
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
@@ -673,20 +664,14 @@ void cv::ogl::Buffer::copyTo(OutputArray arr, Target target, bool autoRelease) c
     {
     case _InputArray::OPENGL_BUFFER:
         {
-            arr.getOGlBufferRef().copyFrom(*this, target, autoRelease);
-            break;
-        }
-
-    case _InputArray::OPENGL_TEXTURE:
-        {
-            arr.getOGlTexture2DRef().copyFrom(*this, autoRelease);
+            arr.getOGlBufferRef().copyFrom(*this);
             break;
         }
 
     case _InputArray::GPU_MAT:
         {
-            #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-                throw_nocuda();
+            #ifndef HAVE_CUDA
+                throw_no_cuda();
             #else
                 GpuMat& dmat = arr.getGpuMatRef();
                 dmat.create(rows_, cols_, type_);
@@ -707,12 +692,31 @@ void cv::ogl::Buffer::copyTo(OutputArray arr, Target target, bool autoRelease) c
 #endif
 }
 
+void cv::ogl::Buffer::copyTo(OutputArray arr, gpu::Stream& stream) const
+{
+#ifndef HAVE_OPENGL
+    (void) arr;
+    (void) stream;
+    throw_no_ogl();
+#else
+    #ifndef HAVE_CUDA
+        (void) arr;
+        (void) stream;
+        throw_no_cuda();
+    #else
+        arr.create(rows_, cols_, type_);
+        GpuMat dmat = arr.getGpuMat();
+        impl_->copyTo(dmat.data, dmat.step, dmat.cols * dmat.elemSize(), dmat.rows, gpu::StreamAccessor::getStream(stream));
+    #endif
+#endif
+}
+
 cv::ogl::Buffer cv::ogl::Buffer::clone(Target target, bool autoRelease) const
 {
 #ifndef HAVE_OPENGL
     (void) target;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
     return cv::ogl::Buffer();
 #else
     ogl::Buffer buf;
@@ -725,7 +729,7 @@ void cv::ogl::Buffer::bind(Target target) const
 {
 #ifndef HAVE_OPENGL
     (void) target;
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_->bind(target);
 #endif
@@ -735,7 +739,7 @@ void cv::ogl::Buffer::unbind(Target target)
 {
 #ifndef HAVE_OPENGL
     (void) target;
-    throw_nogl();
+    throw_no_ogl();
 #else
     gl::BindBuffer(target, 0);
     CV_CheckGlError();
@@ -746,7 +750,7 @@ Mat cv::ogl::Buffer::mapHost(Access access)
 {
 #ifndef HAVE_OPENGL
     (void) access;
-    throw_nogl();
+    throw_no_ogl();
     return Mat();
 #else
     return Mat(rows_, cols_, type_, impl_->mapHost(access));
@@ -756,7 +760,7 @@ Mat cv::ogl::Buffer::mapHost(Access access)
 void cv::ogl::Buffer::unmapHost()
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
     return impl_->unmapHost();
 #endif
@@ -765,11 +769,11 @@ void cv::ogl::Buffer::unmapHost()
 GpuMat cv::ogl::Buffer::mapDevice()
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
     return GpuMat();
 #else
-    #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-        throw_nocuda();
+    #ifndef HAVE_CUDA
+        throw_no_cuda();
         return GpuMat();
     #else
         return GpuMat(rows_, cols_, type_, impl_->mapDevice());
@@ -780,30 +784,58 @@ GpuMat cv::ogl::Buffer::mapDevice()
 void cv::ogl::Buffer::unmapDevice()
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
-    #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-        throw_nocuda();
+    #ifndef HAVE_CUDA
+        throw_no_cuda();
     #else
         impl_->unmapDevice();
     #endif
 #endif
 }
 
+gpu::GpuMat cv::ogl::Buffer::mapDevice(gpu::Stream& stream)
+{
+#ifndef HAVE_OPENGL
+    (void) stream;
+    throw_no_ogl();
+    return GpuMat();
+#else
+    #ifndef HAVE_CUDA
+        (void) stream;
+        throw_no_cuda();
+        return GpuMat();
+    #else
+        return GpuMat(rows_, cols_, type_, impl_->mapDevice(gpu::StreamAccessor::getStream(stream)));
+    #endif
+#endif
+}
+
+void cv::ogl::Buffer::unmapDevice(gpu::Stream& stream)
+{
+#ifndef HAVE_OPENGL
+    (void) stream;
+    throw_no_ogl();
+#else
+    #ifndef HAVE_CUDA
+        (void) stream;
+        throw_no_cuda();
+    #else
+        impl_->unmapDevice(gpu::StreamAccessor::getStream(stream));
+    #endif
+#endif
+}
+
 unsigned int cv::ogl::Buffer::bufId() const
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
     return 0;
 #else
     return impl_->bufId();
 #endif
 }
 
-template <> void cv::Ptr<cv::ogl::Buffer::Impl>::delete_obj()
-{
-    if (obj) delete obj;
-}
 
 //////////////////////////////////////////////////////////////////////////////////////////
 // ogl::Texture
@@ -847,12 +879,13 @@ const Ptr<cv::ogl::Texture2D::Impl> cv::ogl::Texture2D::Impl::empty()
     return p;
 }
 
-cv::ogl::Texture2D::Impl::Impl() : texId_(0), autoRelease_(true)
+cv::ogl::Texture2D::Impl::Impl() : texId_(0), autoRelease_(false)
 {
 }
 
 cv::ogl::Texture2D::Impl::Impl(GLuint atexId, bool autoRelease) : texId_(atexId), autoRelease_(autoRelease)
 {
+    CV_Assert( gl::IsTexture(atexId) == gl::TRUE_ );
 }
 
 cv::ogl::Texture2D::Impl::Impl(GLint internalFormat, GLsizei width, GLsizei height, GLenum format, GLenum type, const GLvoid* pixels, bool autoRelease) : texId_(0), autoRelease_(autoRelease)
@@ -919,7 +952,7 @@ void cv::ogl::Texture2D::Impl::bind() const
 cv::ogl::Texture2D::Texture2D() : rows_(0), cols_(0), format_(NONE)
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_ = Impl::empty();
 #endif
@@ -933,9 +966,9 @@ cv::ogl::Texture2D::Texture2D(int arows, int acols, Format aformat, unsigned int
     (void) aformat;
     (void) atexId;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
-    impl_ = new Impl(atexId, autoRelease);
+    impl_.reset(new Impl(atexId, autoRelease));
     rows_ = arows;
     cols_ = acols;
     format_ = aformat;
@@ -949,31 +982,21 @@ cv::ogl::Texture2D::Texture2D(Size asize, Format aformat, unsigned int atexId, b
     (void) aformat;
     (void) atexId;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
-    impl_ = new Impl(atexId, autoRelease);
+    impl_.reset(new Impl(atexId, autoRelease));
     rows_ = asize.height;
     cols_ = asize.width;
     format_ = aformat;
 #endif
 }
 
-cv::ogl::Texture2D::Texture2D(int arows, int acols, Format aformat, bool autoRelease) : rows_(0), cols_(0), format_(NONE)
-{
-    create(arows, acols, aformat, autoRelease);
-}
-
-cv::ogl::Texture2D::Texture2D(Size asize, Format aformat, bool autoRelease) : rows_(0), cols_(0), format_(NONE)
-{
-    create(asize, aformat, autoRelease);
-}
-
 cv::ogl::Texture2D::Texture2D(InputArray arr, bool autoRelease) : rows_(0), cols_(0), format_(NONE)
 {
 #ifndef HAVE_OPENGL
     (void) arr;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
@@ -1001,20 +1024,20 @@ cv::ogl::Texture2D::Texture2D(InputArray arr, bool autoRelease) : rows_(0), cols
         {
             ogl::Buffer buf = arr.getOGlBuffer();
             buf.bind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
-            impl_ = new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], 0, autoRelease);
+            impl_.reset(new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], 0, autoRelease));
             ogl::Buffer::unbind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
             break;
         }
 
     case _InputArray::GPU_MAT:
         {
-            #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-                throw_nocuda();
+            #ifndef HAVE_CUDA
+                throw_no_cuda();
             #else
                 GpuMat dmat = arr.getGpuMat();
                 ogl::Buffer buf(dmat, ogl::Buffer::PIXEL_UNPACK_BUFFER);
                 buf.bind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
-                impl_ = new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], 0, autoRelease);
+                impl_.reset(new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], 0, autoRelease));
                 ogl::Buffer::unbind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
             #endif
 
@@ -1026,7 +1049,7 @@ cv::ogl::Texture2D::Texture2D(InputArray arr, bool autoRelease) : rows_(0), cols
             Mat mat = arr.getMat();
             CV_Assert( mat.isContinuous() );
             ogl::Buffer::unbind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
-            impl_ = new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], mat.data, autoRelease);
+            impl_.reset(new Impl(internalFormats[cn], asize.width, asize.height, srcFormats[cn], gl_types[depth], mat.data, autoRelease));
             break;
         }
     }
@@ -1044,12 +1067,12 @@ void cv::ogl::Texture2D::create(int arows, int acols, Format aformat, bool autoR
     (void) acols;
     (void) aformat;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     if (rows_ != arows || cols_ != acols || format_ != aformat)
     {
         ogl::Buffer::unbind(ogl::Buffer::PIXEL_UNPACK_BUFFER);
-        impl_ = new Impl(aformat, acols, arows, aformat, gl::FLOAT, 0, autoRelease);
+        impl_.reset(new Impl(aformat, acols, arows, aformat, gl::FLOAT, 0, autoRelease));
         rows_ = arows;
         cols_ = acols;
         format_ = aformat;
@@ -1060,7 +1083,7 @@ void cv::ogl::Texture2D::create(int arows, int acols, Format aformat, bool autoR
 void cv::ogl::Texture2D::release()
 {
 #ifdef HAVE_OPENGL
-    if (*impl_.refcount == 1)
+    if (impl_)
         impl_->setAutoRelease(true);
     impl_ = Impl::empty();
     rows_ = 0;
@@ -1073,7 +1096,7 @@ void cv::ogl::Texture2D::setAutoRelease(bool flag)
 {
 #ifndef HAVE_OPENGL
     (void) flag;
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_->setAutoRelease(flag);
 #endif
@@ -1084,7 +1107,7 @@ void cv::ogl::Texture2D::copyFrom(InputArray arr, bool autoRelease)
 #ifndef HAVE_OPENGL
     (void) arr;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
@@ -1121,8 +1144,8 @@ void cv::ogl::Texture2D::copyFrom(InputArray arr, bool autoRelease)
 
     case _InputArray::GPU_MAT:
         {
-            #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-                throw_nocuda();
+            #ifndef HAVE_CUDA
+                throw_no_cuda();
             #else
                 GpuMat dmat = arr.getGpuMat();
                 ogl::Buffer buf(dmat, ogl::Buffer::PIXEL_UNPACK_BUFFER);
@@ -1151,7 +1174,7 @@ void cv::ogl::Texture2D::copyTo(OutputArray arr, int ddepth, bool autoRelease) c
     (void) arr;
     (void) ddepth;
     (void) autoRelease;
-    throw_nogl();
+    throw_no_ogl();
 #else
     const int kind = arr.kind();
 
@@ -1172,8 +1195,8 @@ void cv::ogl::Texture2D::copyTo(OutputArray arr, int ddepth, bool autoRelease) c
 
     case _InputArray::GPU_MAT:
         {
-            #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-                throw_nocuda();
+            #ifndef HAVE_CUDA
+                throw_no_cuda();
             #else
                 ogl::Buffer buf(rows_, cols_, CV_MAKE_TYPE(ddepth, cn), ogl::Buffer::PIXEL_PACK_BUFFER);
                 buf.bind(ogl::Buffer::PIXEL_PACK_BUFFER);
@@ -1200,7 +1223,7 @@ void cv::ogl::Texture2D::copyTo(OutputArray arr, int ddepth, bool autoRelease) c
 void cv::ogl::Texture2D::bind() const
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
     impl_->bind();
 #endif
@@ -1209,25 +1232,17 @@ void cv::ogl::Texture2D::bind() const
 unsigned int cv::ogl::Texture2D::texId() const
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
     return 0;
 #else
     return impl_->texId();
 #endif
 }
 
-template <> void cv::Ptr<cv::ogl::Texture2D::Impl>::delete_obj()
-{
-    if (obj) delete obj;
-}
 
 ////////////////////////////////////////////////////////////////////////
 // ogl::Arrays
 
-cv::ogl::Arrays::Arrays() : size_(0)
-{
-}
-
 void cv::ogl::Arrays::setVertexArray(InputArray vertex)
 {
     const int cn = vertex.channels();
@@ -1324,7 +1339,7 @@ void cv::ogl::Arrays::setAutoRelease(bool flag)
 void cv::ogl::Arrays::bind() const
 {
 #ifndef HAVE_OPENGL
-    throw_nogl();
+    throw_no_ogl();
 #else
     CV_Assert( texCoord_.empty() || texCoord_.size().area() == size_ );
     CV_Assert( normal_.empty() || normal_.size().area() == size_ );
@@ -1409,7 +1424,7 @@ void cv::ogl::render(const ogl::Texture2D& tex, Rect_<double> wndRect, Rect_<dou
     (void) tex;
     (void) wndRect;
     (void) texRect;
-    throw_nogl();
+    throw_no_ogl();
 #else
     if (!tex.empty())
     {
@@ -1481,7 +1496,7 @@ void cv::ogl::render(const ogl::Arrays& arr, int mode, Scalar color)
     (void) arr;
     (void) mode;
     (void) color;
-    throw_nogl();
+    throw_no_ogl();
 #else
     if (!arr.empty())
     {
@@ -1501,7 +1516,7 @@ void cv::ogl::render(const ogl::Arrays& arr, InputArray indices, int mode, Scala
     (void) indices;
     (void) mode;
     (void) color;
-    throw_nogl();
+    throw_no_ogl();
 #else
     if (!arr.empty() && !indices.empty())
     {
diff --git a/modules/core/src/out.cpp b/modules/core/src/out.cpp
index 5ceee61628..3ed454ebfa 100644
--- a/modules/core/src/out.cpp
+++ b/modules/core/src/out.cpp
@@ -11,7 +11,8 @@
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009-2010, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -41,267 +42,306 @@
 //M*/
 
 #include "precomp.hpp"
-#include <iterator>
+
+namespace
+{
+    class FormattedImpl : public cv::Formatted
+    {
+        enum { STATE_PROLOGUE, STATE_EPILOGUE, STATE_ROW_OPEN, STATE_ROW_CLOSE, STATE_CN_OPEN, STATE_CN_CLOSE, STATE_VALUE, STATE_FINISHED,
+               STATE_LINE_SEPARATOR, STATE_CN_SEPARATOR, STATE_VALUE_SEPARATOR };
+        enum {BRACE_ROW_OPEN = 0, BRACE_ROW_CLOSE = 1, BRACE_ROW_SEP=2, BRACE_CN_OPEN=3, BRACE_CN_CLOSE=4 };
+
+        char floatFormat[8];
+        char buf[32];   // enough for double with precision up to 20
+
+        cv::Mat mtx;
+        int mcn; // == mtx.channels()
+        bool singleLine;
+
+        int state;
+        int row;
+        int col;
+        int cn;
+
+        cv::String prologue;
+        cv::String epilogue;
+        char braces[5];
+
+        void (FormattedImpl::*valueToStr)();
+        void valueToStr8u()  { sprintf(buf, "%3d", (int)mtx.ptr<uchar>(row, col)[cn]); }
+        void valueToStr8s()  { sprintf(buf, "%3d", (int)mtx.ptr<schar>(row, col)[cn]); }
+        void valueToStr16u() { sprintf(buf, "%d", (int)mtx.ptr<ushort>(row, col)[cn]); }
+        void valueToStr16s() { sprintf(buf, "%d", (int)mtx.ptr<short>(row, col)[cn]); }
+        void valueToStr32s() { sprintf(buf, "%d", mtx.ptr<int>(row, col)[cn]); }
+        void valueToStr32f() { sprintf(buf, floatFormat, mtx.ptr<float>(row, col)[cn]); }
+        void valueToStr64f() { sprintf(buf, floatFormat, mtx.ptr<double>(row, col)[cn]); }
+        void valueToStrOther() { buf[0] = 0; }
+
+    public:
+
+        FormattedImpl(cv::String pl, cv::String el, cv::Mat m, char br[5], bool sLine, int precision)
+        {
+            prologue = pl;
+            epilogue = el;
+            mtx = m;
+            mcn = m.channels();
+            memcpy(braces, br, 5);
+            state = STATE_PROLOGUE;
+            singleLine = sLine;
+
+            if (precision < 0)
+            {
+                floatFormat[0] = '%';
+                floatFormat[1] = 'a';
+                floatFormat[2] = 0;
+            }
+            else
+            {
+                sprintf(floatFormat, "%%.%dg", std::min(precision, 20));
+            }
+
+            switch(mtx.depth())
+            {
+                case CV_8U:  valueToStr = &FormattedImpl::valueToStr8u; break;
+                case CV_8S:  valueToStr = &FormattedImpl::valueToStr8s; break;
+                case CV_16U: valueToStr = &FormattedImpl::valueToStr16u; break;
+                case CV_16S: valueToStr = &FormattedImpl::valueToStr16s; break;
+                case CV_32S: valueToStr = &FormattedImpl::valueToStr32s; break;
+                case CV_32F: valueToStr = &FormattedImpl::valueToStr32f; break;
+                case CV_64F: valueToStr = &FormattedImpl::valueToStr64f; break;
+                default:     valueToStr = &FormattedImpl::valueToStrOther; break;
+            }
+        }
+
+        void reset()
+        {
+            state = STATE_PROLOGUE;
+        }
+
+        const char* next()
+        {
+            switch(state)
+            {
+                case STATE_PROLOGUE:
+                    row = 0;
+                    if (mtx.empty())
+                        state = STATE_EPILOGUE;
+                    else
+                        state = STATE_ROW_OPEN;
+                    return prologue.c_str();
+                case STATE_EPILOGUE:
+                    state = STATE_FINISHED;
+                    return epilogue.c_str();
+                case STATE_ROW_OPEN:
+                    col = 0;
+                    state = STATE_CN_OPEN;
+                    {
+                        size_t pos = 0;
+                        if (row > 0)
+                            while(pos < prologue.size() && pos < sizeof(buf) - 2)
+                                buf[pos++] = ' ';
+                        if (braces[BRACE_ROW_OPEN])
+                            buf[pos++] = braces[BRACE_ROW_OPEN];
+                        if(!pos)
+                            return next();
+                        buf[pos] = 0;
+                    }
+                    return buf;
+                case STATE_ROW_CLOSE:
+                    state = STATE_LINE_SEPARATOR;
+                    ++row;
+                    if (braces[BRACE_ROW_CLOSE])
+                    {
+                        buf[0] = braces[BRACE_ROW_CLOSE];
+                        buf[1] = row < mtx.rows ? ',' : '\0';
+                        buf[2] = 0;
+                        return buf;
+                    }
+                    else if(braces[BRACE_ROW_SEP] && row < mtx.rows)
+                    {
+                        buf[0] = braces[BRACE_ROW_SEP];
+                        buf[1] = 0;
+                        return buf;
+                    }
+                    return next();
+                case STATE_CN_OPEN:
+                    cn = 0;
+                    state = STATE_VALUE;
+                    if (mcn > 1 && braces[BRACE_CN_OPEN])
+                    {
+                        buf[0] = braces[BRACE_CN_OPEN];
+                        buf[1] = 0;
+                        return buf;
+                    }
+                    return next();
+                case STATE_CN_CLOSE:
+                    ++col;
+                    if (col >= mtx.cols)
+                        state = STATE_ROW_CLOSE;
+                    else
+                        state = STATE_CN_SEPARATOR;
+                    if (mcn > 1 && braces[BRACE_CN_CLOSE])
+                    {
+                        buf[0] = braces[BRACE_CN_CLOSE];
+                        buf[1] = 0;
+                        return buf;
+                    }
+                    return next();
+                case STATE_VALUE:
+                    (this->*valueToStr)();
+                    if (++cn >= mcn)
+                        state = STATE_CN_CLOSE;
+                    else
+                        state = STATE_VALUE_SEPARATOR;
+                    return buf;
+                case STATE_FINISHED:
+                    return 0;
+                case STATE_LINE_SEPARATOR:
+                    if (row >= mtx.rows)
+                    {
+                        state = STATE_EPILOGUE;
+                        return next();
+                    }
+                    state = STATE_ROW_OPEN;
+                    buf[0] = singleLine ? ' ' : '\n';
+                    buf[1] = 0;
+                    return buf;
+                case STATE_CN_SEPARATOR:
+                    state = STATE_CN_OPEN;
+                    buf[0] = ',';
+                    buf[1] = ' ';
+                    buf[2] = 0;
+                    return buf;
+                case STATE_VALUE_SEPARATOR:
+                    state = STATE_VALUE;
+                    buf[0] = ',';
+                    buf[1] = ' ';
+                    buf[2] = 0;
+                    return buf;
+            }
+            return 0;
+        }
+    };
+
+    class FormatterBase : public cv::Formatter
+    {
+    public:
+        FormatterBase() : prec32f(8), prec64f(16), multiline(true) {}
+
+        void set32fPrecision(int p)
+        {
+            prec32f = p;
+        }
+
+        void set64fPrecision(int p)
+        {
+            prec64f = p;
+        }
+
+        void setMultiline(bool ml)
+        {
+            multiline = ml;
+        }
+
+    protected:
+        int prec32f;
+        int prec64f;
+        int multiline;
+    };
+
+    class MatlabFormatter : public FormatterBase
+    {
+    public:
+
+        cv::Ptr<cv::Formatted> format(const cv::Mat& mtx) const
+        {
+            char braces[5] = {'\0', '\0', ';', '\0', '\0'};
+            return cv::makePtr<FormattedImpl>("[", "]", mtx, &*braces,
+                mtx.cols == 1 || !multiline, mtx.depth() == CV_64F ? prec64f : prec32f );
+        }
+    };
+
+    class PythonFormatter : public FormatterBase
+    {
+    public:
+
+        cv::Ptr<cv::Formatted> format(const cv::Mat& mtx) const
+        {
+            char braces[5] = {'[', ']', '\0', '[', ']'};
+            if (mtx.cols == 1)
+                braces[0] = braces[1] = '\0';
+            return cv::makePtr<FormattedImpl>("[", "]", mtx, &*braces,
+                mtx.cols*mtx.channels() == 1 || !multiline, mtx.depth() == CV_64F ? prec64f : prec32f );
+        }
+    };
+
+    class NumpyFormatter : public FormatterBase
+    {
+    public:
+
+        cv::Ptr<cv::Formatted> format(const cv::Mat& mtx) const
+        {
+            static const char* numpyTypes[] =
+            {
+                "uint8", "int8", "uint16", "int16", "int32", "float32", "float64", "uint64"
+            };
+            char braces[5] = {'[', ']', '\0', '[', ']'};
+            if (mtx.cols == 1)
+                braces[0] = braces[1] = '\0';
+            return cv::makePtr<FormattedImpl>("array([",
+                cv::format("], type='%s')", numpyTypes[mtx.depth()]), mtx, &*braces,
+                mtx.cols*mtx.channels() == 1 || !multiline, mtx.depth() == CV_64F ? prec64f : prec32f );
+        }
+    };
+
+    class CSVFormatter : public FormatterBase
+    {
+    public:
+
+        cv::Ptr<cv::Formatted> format(const cv::Mat& mtx) const
+        {
+            char braces[5] = {'\0', '\0', '\0', '\0', '\0'};
+            return cv::makePtr<FormattedImpl>(cv::String(),
+                mtx.rows > 1 ? cv::String("\n") : cv::String(), mtx, &*braces,
+                mtx.cols*mtx.channels() == 1 || !multiline, mtx.depth() == CV_64F ? prec64f : prec32f );
+        }
+    };
+
+    class CFormatter : public FormatterBase
+    {
+    public:
+
+        cv::Ptr<cv::Formatted> format(const cv::Mat& mtx) const
+        {
+            char braces[5] = {'\0', '\0', ',', '\0', '\0'};
+            return cv::makePtr<FormattedImpl>("{", "}", mtx, &*braces,
+                mtx.cols == 1 || !multiline, mtx.depth() == CV_64F ? prec64f : prec32f );
+        }
+    };
+
+} // namespace
+
 
 namespace cv
 {
+    Formatted::~Formatted() {}
+    Formatter::~Formatter() {}
 
-static inline char getCloseBrace(char c)
-{
-    return c == '[' ? ']' : c == '(' ? ')' : c == '{' ? '}' : '\0';
-}
-
-
-template<typename _Tp> static void writeElems(std::ostream& out, const _Tp* data,
-                                              int nelems, int cn, char obrace, char cbrace)
-{
-    typedef typename DataType<_Tp>::work_type _WTp;
-    nelems *= cn;
-    for(int i = 0; i < nelems; i += cn)
+    Ptr<Formatter> Formatter::get(int fmt)
     {
-        if(cn == 1)
+        switch(fmt)
         {
-            out << (_WTp)data[i] << (i+1 < nelems ? ", " : "");
-            continue;
+            case FMT_MATLAB:
+                return makePtr<MatlabFormatter>();
+            case FMT_CSV:
+                return makePtr<CSVFormatter>();
+            case FMT_PYTHON:
+                return makePtr<PythonFormatter>();
+            case FMT_NUMPY:
+                return makePtr<NumpyFormatter>();
+            case FMT_C:
+                return makePtr<CFormatter>();
         }
-        out << obrace;
-        for(int j = 0; j < cn; j++)
-            out << (_WTp)data[i + j] << (j+1 < cn ? ", " : "");
-        out << cbrace << (i+cn < nelems ? ", " : "");
+        return makePtr<MatlabFormatter>();
     }
-}
-
-
-static void writeElems(std::ostream& out, const void* data, int nelems, int type, char brace)
-{
-    int depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
-    char cbrace = ' ';
-    if(!brace || isspace(brace))
-    {
-        nelems *= cn;
-        cn = 1;
-    }
-    else
-        cbrace = getCloseBrace(brace);
-    if(depth == CV_8U)
-        writeElems(out, (const uchar*)data, nelems, cn, brace, cbrace);
-    else if(depth == CV_8S)
-        writeElems(out, (const schar*)data, nelems, cn, brace, cbrace);
-    else if(depth == CV_16U)
-        writeElems(out, (const ushort*)data, nelems, cn, brace, cbrace);
-    else if(depth == CV_16S)
-        writeElems(out, (const short*)data, nelems, cn, brace, cbrace);
-    else if(depth == CV_32S)
-        writeElems(out, (const int*)data, nelems, cn, brace, cbrace);
-    else if(depth == CV_32F)
-    {
-        std::streamsize pp = out.precision();
-        out.precision(8);
-        writeElems(out, (const float*)data, nelems, cn, brace, cbrace);
-        out.precision(pp);
-    }
-    else if(depth == CV_64F)
-    {
-        std::streamsize pp = out.precision();
-        out.precision(16);
-        writeElems(out, (const double*)data, nelems, cn, brace, cbrace);
-        out.precision(pp);
-    }
-    else
-        CV_Error(CV_StsUnsupportedFormat, "");
-}
-
-
-static void writeMat(std::ostream& out, const Mat& m, char rowsep, char elembrace, bool singleLine)
-{
-    CV_Assert(m.dims <= 2);
-    int type = m.type();
-
-    char crowbrace = getCloseBrace(rowsep);
-    char orowbrace = crowbrace ? rowsep : '\0';
-
-    if( orowbrace || isspace(rowsep) )
-        rowsep = '\0';
-
-    for( int i = 0; i < m.rows; i++ )
-    {
-        if(orowbrace)
-            out << orowbrace;
-        if( m.data )
-            writeElems(out, m.ptr(i), m.cols, type, elembrace);
-        if(orowbrace)
-            out << crowbrace << (i+1 < m.rows ? ", " : "");
-        if(i+1 < m.rows)
-        {
-            if(rowsep)
-                out << rowsep << (singleLine ? " " : "");
-            if(!singleLine)
-                out << "\n  ";
-        }
-    }
-}
-
-class MatlabFormatter : public Formatter
-{
-public:
-    virtual ~MatlabFormatter() {}
-    void write(std::ostream& out, const Mat& m, const int*, int) const
-    {
-        out << "[";
-        writeMat(out, m, ';', ' ', m.cols == 1);
-        out << "]";
-    }
-
-    void write(std::ostream& out, const void* data, int nelems, int type, const int*, int) const
-    {
-        writeElems(out, data, nelems, type, ' ');
-    }
-};
-
-class PythonFormatter : public Formatter
-{
-public:
-    virtual ~PythonFormatter() {}
-    void write(std::ostream& out, const Mat& m, const int*, int) const
-    {
-        out << "[";
-        writeMat(out, m, m.cols > 1 ? '[' : ' ', '[', m.cols*m.channels() == 1);
-        out << "]";
-    }
-
-    void write(std::ostream& out, const void* data, int nelems, int type, const int*, int) const
-    {
-        writeElems(out, data, nelems, type, '[');
-    }
-};
-
-
-class NumpyFormatter : public Formatter
-{
-public:
-    virtual ~NumpyFormatter() {}
-    void write(std::ostream& out, const Mat& m, const int*, int) const
-    {
-        static const char* numpyTypes[] =
-        {
-            "uint8", "int8", "uint16", "int16", "int32", "float32", "float64", "uint64"
-        };
-        out << "array([";
-        writeMat(out, m, m.cols > 1 ? '[' : ' ', '[', m.cols*m.channels() == 1);
-        out << "], type='" << numpyTypes[m.depth()] << "')";
-    }
-
-    void write(std::ostream& out, const void* data, int nelems, int type, const int*, int) const
-    {
-        writeElems(out, data, nelems, type, '[');
-    }
-};
-
-
-class CSVFormatter : public Formatter
-{
-public:
-    virtual ~CSVFormatter() {}
-    void write(std::ostream& out, const Mat& m, const int*, int) const
-    {
-        writeMat(out, m, ' ', ' ', m.cols*m.channels() == 1);
-        if(m.rows > 1)
-            out << "\n";
-    }
-
-    void write(std::ostream& out, const void* data, int nelems, int type, const int*, int) const
-    {
-        writeElems(out, data, nelems, type, ' ');
-    }
-};
-
-
-class CFormatter : public Formatter
-{
-public:
-    virtual ~CFormatter() {}
-    void write(std::ostream& out, const Mat& m, const int*, int) const
-    {
-        out << "{";
-        writeMat(out, m, ',', ' ', m.cols==1);
-        out << "}";
-    }
-
-    void write(std::ostream& out, const void* data, int nelems, int type, const int*, int) const
-    {
-        writeElems(out, data, nelems, type, ' ');
-    }
-};
-
-
-static MatlabFormatter matlabFormatter;
-static PythonFormatter pythonFormatter;
-static NumpyFormatter numpyFormatter;
-static CSVFormatter csvFormatter;
-static CFormatter cFormatter;
-
-static const Formatter* g_defaultFormatter0 = &matlabFormatter;
-static const Formatter* g_defaultFormatter = &matlabFormatter;
-
-static bool my_streq(const char* a, const char* b)
-{
-    size_t i, alen = strlen(a), blen = strlen(b);
-    if( alen != blen )
-        return false;
-    for( i = 0; i < alen; i++ )
-        if( a[i] != b[i] && a[i] - 32 != b[i] )
-            return false;
-    return true;
-}
-
-const Formatter* Formatter::get(const char* fmt)
-{
-    if(!fmt || my_streq(fmt, ""))
-        return g_defaultFormatter;
-    if( my_streq(fmt, "MATLAB"))
-        return &matlabFormatter;
-    if( my_streq(fmt, "CSV"))
-        return &csvFormatter;
-    if( my_streq(fmt, "PYTHON"))
-        return &pythonFormatter;
-    if( my_streq(fmt, "NUMPY"))
-        return &numpyFormatter;
-    if( my_streq(fmt, "C"))
-        return &cFormatter;
-    CV_Error(CV_StsBadArg, "Unknown formatter");
-    return g_defaultFormatter;
-}
-
-const Formatter* Formatter::setDefault(const Formatter* fmt)
-{
-    const Formatter* prevFmt = g_defaultFormatter;
-    if(!fmt)
-        fmt = g_defaultFormatter0;
-    g_defaultFormatter = fmt;
-    return prevFmt;
-}
-
-Formatted::Formatted(const Mat& _m, const Formatter* _fmt,
-                     const std::vector<int>& _params)
-{
-    mtx = _m;
-    fmt = _fmt ? _fmt : Formatter::get();
-    std::copy(_params.begin(), _params.end(), back_inserter(params));
-}
-
-Formatted::Formatted(const Mat& _m, const Formatter* _fmt, const int* _params)
-{
-    mtx = _m;
-    fmt = _fmt ? _fmt : Formatter::get();
-
-    if( _params )
-    {
-        int i, maxParams = 100;
-        for(i = 0; i < maxParams && _params[i] != 0; i+=2)
-            ;
-        std::copy(_params, _params + i, back_inserter(params));
-    }
-}
-
-}
-
+} // cv
diff --git a/modules/core/src/parallel.cpp b/modules/core/src/parallel.cpp
index 0b2a845ac1..27d7ecc038 100644
--- a/modules/core/src/parallel.cpp
+++ b/modules/core/src/parallel.cpp
@@ -110,8 +110,16 @@
     #endif
 #endif
 
-#if defined HAVE_TBB || defined HAVE_CSTRIPES || defined HAVE_OPENMP || defined HAVE_GCD || defined HAVE_CONCURRENCY
-   #define HAVE_PARALLEL_FRAMEWORK
+#if defined HAVE_TBB && TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#  define CV_PARALLEL_FRAMEWORK "tbb"
+#elif defined HAVE_CSTRIPES
+#  define CV_PARALLEL_FRAMEWORK "cstripes"
+#elif defined HAVE_OPENMP
+#  define CV_PARALLEL_FRAMEWORK "openmp"
+#elif defined HAVE_GCD
+#  define CV_PARALLEL_FRAMEWORK "gcd"
+#elif defined HAVE_CONCURRENCY
+#  define CV_PARALLEL_FRAMEWORK "ms-concurrency"
 #endif
 
 namespace cv
@@ -121,7 +129,7 @@ namespace cv
 
 namespace
 {
-#ifdef HAVE_PARALLEL_FRAMEWORK
+#ifdef CV_PARALLEL_FRAMEWORK
     class ParallelLoopBodyWrapper
     {
     public:
@@ -136,9 +144,9 @@ namespace
         {
             cv::Range r;
             r.start = (int)(wholeRange.start +
-                            ((size_t)sr.start*(wholeRange.end - wholeRange.start) + nstripes/2)/nstripes);
+                            ((uint64)sr.start*(wholeRange.end - wholeRange.start) + nstripes/2)/nstripes);
             r.end = sr.end >= nstripes ? wholeRange.end : (int)(wholeRange.start +
-                            ((size_t)sr.end*(wholeRange.end - wholeRange.start) + nstripes/2)/nstripes);
+                            ((uint64)sr.end*(wholeRange.end - wholeRange.start) + nstripes/2)/nstripes);
             (*body)(r);
         }
         cv::Range stripeRange() const { return cv::Range(0, nstripes); }
@@ -218,7 +226,7 @@ public:
 static SchedPtr pplScheduler;
 #endif
 
-#endif // HAVE_PARALLEL_FRAMEWORK
+#endif // CV_PARALLEL_FRAMEWORK
 
 } //namespace
 
@@ -226,7 +234,7 @@ static SchedPtr pplScheduler;
 
 void cv::parallel_for_(const cv::Range& range, const cv::ParallelLoopBody& body, double nstripes)
 {
-#ifdef HAVE_PARALLEL_FRAMEWORK
+#ifdef CV_PARALLEL_FRAMEWORK
 
     if(numThreads != 0)
     {
@@ -281,7 +289,7 @@ void cv::parallel_for_(const cv::Range& range, const cv::ParallelLoopBody& body,
     }
     else
 
-#endif // HAVE_PARALLEL_FRAMEWORK
+#endif // CV_PARALLEL_FRAMEWORK
     {
         (void)nstripes;
         body(range);
@@ -290,7 +298,7 @@ void cv::parallel_for_(const cv::Range& range, const cv::ParallelLoopBody& body,
 
 int cv::getNumThreads(void)
 {
-#ifdef HAVE_PARALLEL_FRAMEWORK
+#ifdef CV_PARALLEL_FRAMEWORK
 
     if(numThreads == 0)
         return 1;
@@ -333,7 +341,7 @@ int cv::getNumThreads(void)
 void cv::setNumThreads( int threads )
 {
     (void)threads;
-#ifdef HAVE_PARALLEL_FRAMEWORK
+#ifdef CV_PARALLEL_FRAMEWORK
     numThreads = threads;
 #endif
 
@@ -445,7 +453,11 @@ int cv::getNumberOfCPUs(void)
 {
 #if defined WIN32 || defined _WIN32
     SYSTEM_INFO sysinfo;
+#if defined(_M_ARM) || defined(_M_X64) || defined(HAVE_WINRT)
+    GetNativeSystemInfo( &sysinfo );
+#else
     GetSystemInfo( &sysinfo );
+#endif
 
     return (int)sysinfo.dwNumberOfProcessors;
 #elif defined ANDROID
@@ -480,6 +492,14 @@ int cv::getNumberOfCPUs(void)
 #endif
 }
 
+const char* cv::currentParallelFramework() {
+#ifdef CV_PARALLEL_FRAMEWORK
+    return CV_PARALLEL_FRAMEWORK;
+#else
+    return NULL;
+#endif
+}
+
 CV_IMPL void cvSetNumThreads(int nt)
 {
     cv::setNumThreads(nt);
diff --git a/modules/core/src/persistence.cpp b/modules/core/src/persistence.cpp
index 1ee511910c..39ce633139 100644
--- a/modules/core/src/persistence.cpp
+++ b/modules/core/src/persistence.cpp
@@ -58,7 +58,6 @@
 #endif
 
 #if USE_ZLIB
-#  undef HAVE_UNISTD_H //to avoid redefinition
 #  ifndef _LFS64_LARGEFILE
 #    define _LFS64_LARGEFILE 0
 #  endif
@@ -116,7 +115,7 @@ static char* icv_itoa( int _val, char* buffer, int /*radix*/ )
     return ptr;
 }
 
-std::string cv::FileStorage::getDefaultObjectName(const std::string& _filename)
+cv::String cv::FileStorage::getDefaultObjectName(const cv::String& _filename)
 {
     static const char* stubname = "unnamed";
     const char* filename = _filename.c_str();
@@ -152,7 +151,7 @@ std::string cv::FileStorage::getDefaultObjectName(const std::string& _filename)
     name = name_buf;
     if( strcmp( name, "_" ) == 0 )
         strcpy( name, stubname );
-    return std::string(name);
+    return String(name);
 }
 
 typedef struct CvGenericHash
@@ -516,7 +515,7 @@ icvFSFlush( CvFileStorage* fs )
 
 
 static void
-icvClose( CvFileStorage* fs, std::string* out )
+icvClose( CvFileStorage* fs, cv::String* out )
 {
     if( out )
         out->clear();
@@ -543,8 +542,7 @@ icvClose( CvFileStorage* fs, std::string* out )
 
     if( fs->outbuf && out )
     {
-        out->resize(fs->outbuf->size());
-        std::copy(fs->outbuf->begin(), fs->outbuf->end(), out->begin());
+        *out = cv::String(fs->outbuf->begin(), fs->outbuf->end());
     }
 }
 
@@ -3247,19 +3245,19 @@ cvReadRawDataSlice( const CvFileStorage* fs, CvSeqReader* reader,
                     switch( elem_type )
                     {
                     case CV_8U:
-                        *(uchar*)data = CV_CAST_8U(ival);
+                        *(uchar*)data = cv::saturate_cast<uchar>(ival);
                         data++;
                         break;
                     case CV_8S:
-                        *(char*)data = CV_CAST_8S(ival);
+                        *(char*)data = cv::saturate_cast<schar>(ival);
                         data++;
                         break;
                     case CV_16U:
-                        *(ushort*)data = CV_CAST_16U(ival);
+                        *(ushort*)data = cv::saturate_cast<ushort>(ival);
                         data += sizeof(ushort);
                         break;
                     case CV_16S:
-                        *(short*)data = CV_CAST_16S(ival);
+                        *(short*)data = cv::saturate_cast<short>(ival);
                         data += sizeof(short);
                         break;
                     case CV_32S:
@@ -3292,22 +3290,22 @@ cvReadRawDataSlice( const CvFileStorage* fs, CvSeqReader* reader,
                     {
                     case CV_8U:
                         ival = cvRound(fval);
-                        *(uchar*)data = CV_CAST_8U(ival);
+                        *(uchar*)data = cv::saturate_cast<uchar>(ival);
                         data++;
                         break;
                     case CV_8S:
                         ival = cvRound(fval);
-                        *(char*)data = CV_CAST_8S(ival);
+                        *(char*)data = cv::saturate_cast<schar>(ival);
                         data++;
                         break;
                     case CV_16U:
                         ival = cvRound(fval);
-                        *(ushort*)data = CV_CAST_16U(ival);
+                        *(ushort*)data = cv::saturate_cast<ushort>(ival);
                         data += sizeof(ushort);
                         break;
                     case CV_16S:
                         ival = cvRound(fval);
-                        *(short*)data = CV_CAST_16S(ival);
+                        *(short*)data = cv::saturate_cast<short>(ival);
                         data += sizeof(short);
                         break;
                     case CV_32S:
@@ -5011,7 +5009,7 @@ cvSave( const char* filename, const void* struct_ptr,
     if( !fs )
         CV_Error( CV_StsError, "Could not open the file storage. Check the path and permissions" );
 
-    std::string name = _name ? std::string(_name) : cv::FileStorage::getDefaultObjectName(filename);
+    cv::String name = _name ? cv::String(_name) : cv::FileStorage::getDefaultObjectName(filename);
 
     if( comment )
         cvWriteComment( fs, comment, 0 );
@@ -5105,7 +5103,7 @@ stop_search:
 namespace cv
 {
 
-static void getElemSize( const std::string& fmt, size_t& elemSize, size_t& cn )
+static void getElemSize( const String& fmt, size_t& elemSize, size_t& cn )
 {
     const char* dt = fmt.c_str();
     cn = 1;
@@ -5125,15 +5123,17 @@ FileStorage::FileStorage()
     state = UNDEFINED;
 }
 
-FileStorage::FileStorage(const std::string& filename, int flags, const std::string& encoding)
+FileStorage::FileStorage(const String& filename, int flags, const String& encoding)
 {
     state = UNDEFINED;
     open( filename, flags, encoding );
 }
 
-FileStorage::FileStorage(CvFileStorage* _fs)
+FileStorage::FileStorage(CvFileStorage* _fs, bool owning)
 {
-    fs = Ptr<CvFileStorage>(_fs);
+    if (owning) fs.reset(_fs);
+    else fs = Ptr<CvFileStorage>(Ptr<CvFileStorage>(), _fs);
+
     state = _fs ? NAME_EXPECTED + INSIDE_MAP : UNDEFINED;
 }
 
@@ -5146,11 +5146,11 @@ FileStorage::~FileStorage()
     }
 }
 
-bool FileStorage::open(const std::string& filename, int flags, const std::string& encoding)
+bool FileStorage::open(const String& filename, int flags, const String& encoding)
 {
     release();
-    fs = Ptr<CvFileStorage>(cvOpenFileStorage( filename.c_str(), 0, flags,
-                                               !encoding.empty() ? encoding.c_str() : 0));
+    fs.reset(cvOpenFileStorage( filename.c_str(), 0, flags,
+                                !encoding.empty() ? encoding.c_str() : 0));
     bool ok = isOpened();
     state = ok ? NAME_EXPECTED + INSIDE_MAP : UNDEFINED;
     return ok;
@@ -5158,7 +5158,7 @@ bool FileStorage::open(const std::string& filename, int flags, const std::string
 
 bool FileStorage::isOpened() const
 {
-    return !fs.empty() && fs.obj->is_opened;
+    return fs && fs->is_opened;
 }
 
 void FileStorage::release()
@@ -5168,12 +5168,11 @@ void FileStorage::release()
     state = UNDEFINED;
 }
 
-std::string FileStorage::releaseAndGetString()
+String FileStorage::releaseAndGetString()
 {
-    std::string buf;
-    buf.reserve(16); // HACK: Work around for compiler bug
-    if( fs.obj && fs.obj->outbuf )
-        icvClose(fs.obj, &buf);
+    String buf;
+    if( fs && fs->outbuf )
+        icvClose(fs, &buf);
 
     release();
     return buf;
@@ -5184,7 +5183,7 @@ FileNode FileStorage::root(int streamidx) const
     return isOpened() ? FileNode(fs, cvGetRootFileNode(fs, streamidx)) : FileNode();
 }
 
-FileStorage& operator << (FileStorage& fs, const std::string& str)
+FileStorage& operator << (FileStorage& fs, const String& str)
 {
     enum { NAME_EXPECTED = FileStorage::NAME_EXPECTED,
         VALUE_EXPECTED = FileStorage::VALUE_EXPECTED,
@@ -5203,7 +5202,7 @@ FileStorage& operator << (FileStorage& fs, const std::string& str)
         fs.state = fs.structs.empty() || fs.structs.back() == '{' ?
             INSIDE_MAP + NAME_EXPECTED : VALUE_EXPECTED;
         cvEndWriteStruct( *fs );
-        fs.elname = std::string();
+        fs.elname = String();
     }
     else if( fs.state == NAME_EXPECTED + INSIDE_MAP )
     {
@@ -5227,12 +5226,12 @@ FileStorage& operator << (FileStorage& fs, const std::string& str)
             }
             cvStartWriteStruct( *fs, fs.elname.size() > 0 ? fs.elname.c_str() : 0,
                 flags, *_str ? _str : 0 );
-            fs.elname = std::string();
+            fs.elname = String();
         }
         else
         {
             write( fs, fs.elname, (_str[0] == '\\' && (_str[1] == '{' || _str[1] == '}' ||
-                _str[1] == '[' || _str[1] == ']')) ? std::string(_str+1) : str );
+                _str[1] == '[' || _str[1] == ']')) ? String(_str+1) : str );
             if( fs.state == INSIDE_MAP + VALUE_EXPECTED )
                 fs.state = INSIDE_MAP + NAME_EXPECTED;
         }
@@ -5243,7 +5242,7 @@ FileStorage& operator << (FileStorage& fs, const std::string& str)
 }
 
 
-void FileStorage::writeRaw( const std::string& fmt, const uchar* vec, size_t len )
+void FileStorage::writeRaw( const String& fmt, const uchar* vec, size_t len )
 {
     if( !isOpened() )
         return;
@@ -5254,7 +5253,7 @@ void FileStorage::writeRaw( const std::string& fmt, const uchar* vec, size_t len
 }
 
 
-void FileStorage::writeObj( const std::string& name, const void* obj )
+void FileStorage::writeObj( const String& name, const void* obj )
 {
     if( !isOpened() )
         return;
@@ -5262,7 +5261,7 @@ void FileStorage::writeObj( const std::string& name, const void* obj )
 }
 
 
-FileNode FileStorage::operator[](const std::string& nodename) const
+FileNode FileStorage::operator[](const String& nodename) const
 {
     return FileNode(fs, cvGetFileNodeByName(fs, 0, nodename.c_str()));
 }
@@ -5272,7 +5271,7 @@ FileNode FileStorage::operator[](const char* nodename) const
     return FileNode(fs, cvGetFileNodeByName(fs, 0, nodename));
 }
 
-FileNode FileNode::operator[](const std::string& nodename) const
+FileNode FileNode::operator[](const String& nodename) const
 {
     return FileNode(fs, cvGetFileNodeByName(fs, node, nodename.c_str()));
 }
@@ -5288,10 +5287,10 @@ FileNode FileNode::operator[](int i) const
         i == 0 ? *this : FileNode();
 }
 
-std::string FileNode::name() const
+String FileNode::name() const
 {
     const char* str;
-    return !node || (str = cvGetFileNodeName(node)) == 0 ? std::string() : std::string(str);
+    return !node || (str = cvGetFileNodeName(node)) == 0 ? String() : String(str);
 }
 
 void* FileNode::readObj() const
@@ -5319,7 +5318,7 @@ FileNodeIterator::FileNodeIterator(const CvFileStorage* _fs,
         container = _node;
         if( !(_node->tag & FileNode::USER) && (node_type == FileNode::SEQ || node_type == FileNode::MAP) )
         {
-            cvStartReadSeq( _node->data.seq, &reader );
+            cvStartReadSeq( _node->data.seq, (CvSeqReader*)&reader );
             remaining = FileNode(_fs, _node).size();
         }
         else
@@ -5352,7 +5351,12 @@ FileNodeIterator& FileNodeIterator::operator ++()
     if( remaining > 0 )
     {
         if( reader.seq )
-            CV_NEXT_SEQ_ELEM( reader.seq->elem_size, reader );
+        {
+            if( ((reader).ptr += (((CvSeq*)reader.seq)->elem_size)) >= (reader).block_max )
+            {
+                cvChangeSeqBlock( (CvSeqReader*)&(reader), 1 );
+            }
+        }
         remaining--;
     }
     return *this;
@@ -5370,7 +5374,12 @@ FileNodeIterator& FileNodeIterator::operator --()
     if( remaining < FileNode(fs, container).size() )
     {
         if( reader.seq )
-            CV_PREV_SEQ_ELEM( reader.seq->elem_size, reader );
+        {
+            if( ((reader).ptr -= (((CvSeq*)reader.seq)->elem_size)) < (reader).block_min )
+            {
+                cvChangeSeqBlock( (CvSeqReader*)&(reader), -1 );
+            }
+        }
         remaining++;
     }
     return *this;
@@ -5396,7 +5405,7 @@ FileNodeIterator& FileNodeIterator::operator += (int ofs)
     }
     remaining -= ofs;
     if( reader.seq )
-        cvSetSeqReaderPos( &reader, ofs, 1 );
+        cvSetSeqReaderPos( (CvSeqReader*)&reader, ofs, 1 );
     return *this;
 }
 
@@ -5406,7 +5415,7 @@ FileNodeIterator& FileNodeIterator::operator -= (int ofs)
 }
 
 
-FileNodeIterator& FileNodeIterator::readRaw( const std::string& fmt, uchar* vec, size_t maxCount )
+FileNodeIterator& FileNodeIterator::readRaw( const String& fmt, uchar* vec, size_t maxCount )
 {
     if( fs && container && remaining > 0 )
     {
@@ -5417,7 +5426,7 @@ FileNodeIterator& FileNodeIterator::readRaw( const std::string& fmt, uchar* vec,
 
         if( reader.seq )
         {
-            cvReadRawDataSlice( fs, &reader, (int)count, vec, fmt.c_str() );
+            cvReadRawDataSlice( fs, (CvSeqReader*)&reader, (int)count, vec, fmt.c_str() );
             remaining -= count*cn;
         }
         else
@@ -5430,16 +5439,16 @@ FileNodeIterator& FileNodeIterator::readRaw( const std::string& fmt, uchar* vec,
 }
 
 
-void write( FileStorage& fs, const std::string& name, int value )
+void write( FileStorage& fs, const String& name, int value )
 { cvWriteInt( *fs, name.size() ? name.c_str() : 0, value ); }
 
-void write( FileStorage& fs, const std::string& name, float value )
+void write( FileStorage& fs, const String& name, float value )
 { cvWriteReal( *fs, name.size() ? name.c_str() : 0, value ); }
 
-void write( FileStorage& fs, const std::string& name, double value )
+void write( FileStorage& fs, const String& name, double value )
 { cvWriteReal( *fs, name.size() ? name.c_str() : 0, value ); }
 
-void write( FileStorage& fs, const std::string& name, const std::string& value )
+void write( FileStorage& fs, const String& name, const String& value )
 { cvWriteString( *fs, name.size() ? name.c_str() : 0, value.c_str() ); }
 
 void writeScalar(FileStorage& fs, int value )
@@ -5451,11 +5460,11 @@ void writeScalar(FileStorage& fs, float value )
 void writeScalar(FileStorage& fs, double value )
 { cvWriteReal( *fs, 0, value ); }
 
-void writeScalar(FileStorage& fs, const std::string& value )
+void writeScalar(FileStorage& fs, const String& value )
 { cvWriteString( *fs, 0, value.c_str() ); }
 
 
-void write( FileStorage& fs, const std::string& name, const Mat& value )
+void write( FileStorage& fs, const String& name, const Mat& value )
 {
     if( value.dims <= 2 )
     {
@@ -5470,21 +5479,24 @@ void write( FileStorage& fs, const std::string& name, const Mat& value )
 }
 
 // TODO: the 4 functions below need to be implemented more efficiently
-void write( FileStorage& fs, const std::string& name, const SparseMat& value )
+void write( FileStorage& fs, const String& name, const SparseMat& value )
 {
-    Ptr<CvSparseMat> mat = (CvSparseMat*)value;
+    Ptr<CvSparseMat> mat(cvCreateSparseMat(value));
     cvWrite( *fs, name.size() ? name.c_str() : 0, mat );
 }
 
 
-WriteStructContext::WriteStructContext(FileStorage& _fs, const std::string& name,
-                   int flags, const std::string& typeName) : fs(&_fs)
+internal::WriteStructContext::WriteStructContext(FileStorage& _fs,
+    const String& name, int flags, const String& typeName) : fs(&_fs)
 {
     cvStartWriteStruct(**fs, !name.empty() ? name.c_str() : 0, flags,
                        !typeName.empty() ? typeName.c_str() : 0);
 }
 
-WriteStructContext::~WriteStructContext() { cvEndWriteStruct(**fs); }
+internal::WriteStructContext::~WriteStructContext()
+{
+    cvEndWriteStruct(**fs);
+}
 
 
 void read( const FileNode& node, Mat& mat, const Mat& default_mat )
@@ -5497,12 +5509,12 @@ void read( const FileNode& node, Mat& mat, const Mat& default_mat )
     void* obj = cvRead((CvFileStorage*)node.fs, (CvFileNode*)*node);
     if(CV_IS_MAT_HDR_Z(obj))
     {
-        Mat((const CvMat*)obj).copyTo(mat);
+        cvarrToMat(obj).copyTo(mat);
         cvReleaseMat((CvMat**)&obj);
     }
     else if(CV_IS_MATND_HDR(obj))
     {
-        Mat((const CvMatND*)obj).copyTo(mat);
+        cvarrToMat(obj).copyTo(mat);
         cvReleaseMatND((CvMatND**)&obj);
     }
     else
@@ -5519,9 +5531,76 @@ void read( const FileNode& node, SparseMat& mat, const SparseMat& default_mat )
         default_mat.copyTo(mat);
         return;
     }
-    Ptr<CvSparseMat> m = (CvSparseMat*)cvRead((CvFileStorage*)node.fs, (CvFileNode*)*node);
-    CV_Assert(CV_IS_SPARSE_MAT(m.obj));
-    SparseMat(m).copyTo(mat);
+    Ptr<CvSparseMat> m((CvSparseMat*)cvRead((CvFileStorage*)node.fs, (CvFileNode*)*node));
+    CV_Assert(CV_IS_SPARSE_MAT(m));
+    m->copyToSparseMat(mat);
+}
+
+void write(FileStorage& fs, const String& objname, const std::vector<KeyPoint>& keypoints)
+{
+    internal::WriteStructContext ws(fs, objname, CV_NODE_SEQ + CV_NODE_FLOW);
+
+    int i, npoints = (int)keypoints.size();
+    for( i = 0; i < npoints; i++ )
+    {
+        const KeyPoint& kpt = keypoints[i];
+        cv::write(fs, kpt.pt.x);
+        cv::write(fs, kpt.pt.y);
+        cv::write(fs, kpt.size);
+        cv::write(fs, kpt.angle);
+        cv::write(fs, kpt.response);
+        cv::write(fs, kpt.octave);
+        cv::write(fs, kpt.class_id);
+    }
+}
+
+
+void read(const FileNode& node, std::vector<KeyPoint>& keypoints)
+{
+    keypoints.resize(0);
+    FileNodeIterator it = node.begin(), it_end = node.end();
+    for( ; it != it_end; )
+    {
+        KeyPoint kpt;
+        it >> kpt.pt.x >> kpt.pt.y >> kpt.size >> kpt.angle >> kpt.response >> kpt.octave >> kpt.class_id;
+        keypoints.push_back(kpt);
+    }
+}
+
+int FileNode::type() const { return !node ? NONE : (node->tag & TYPE_MASK); }
+bool FileNode::isNamed() const { return !node ? false : (node->tag & NAMED) != 0; }
+
+size_t FileNode::size() const
+{
+    int t = type();
+    return t == MAP ? (size_t)((CvSet*)node->data.map)->active_count :
+        t == SEQ ? (size_t)node->data.seq->total : (size_t)!isNone();
+}
+
+void read(const FileNode& node, int& value, int default_value)
+{
+    value = !node.node ? default_value :
+    CV_NODE_IS_INT(node.node->tag) ? node.node->data.i :
+    CV_NODE_IS_REAL(node.node->tag) ? cvRound(node.node->data.f) : 0x7fffffff;
+}
+
+void read(const FileNode& node, float& value, float default_value)
+{
+    value = !node.node ? default_value :
+        CV_NODE_IS_INT(node.node->tag) ? (float)node.node->data.i :
+        CV_NODE_IS_REAL(node.node->tag) ? (float)node.node->data.f : 1e30f;
+}
+
+void read(const FileNode& node, double& value, double default_value)
+{
+    value = !node.node ? default_value :
+        CV_NODE_IS_INT(node.node->tag) ? (double)node.node->data.i :
+        CV_NODE_IS_REAL(node.node->tag) ? node.node->data.f : 1e300;
+}
+
+void read(const FileNode& node, String& value, const String& default_value)
+{
+    value = !node.node ? default_value : CV_NODE_IS_STRING(node.node->tag) ? String(node.node->data.str.ptr) : String();
 }
 
 }
diff --git a/modules/core/src/precomp.hpp b/modules/core/src/precomp.hpp
index 30ce3d17b1..1a59f6e7bd 100644
--- a/modules/core/src/precomp.hpp
+++ b/modules/core/src/precomp.hpp
@@ -43,13 +43,13 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
-#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/core/core_c.h"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/core/opengl.hpp"
+
+#include "opencv2/core/private.hpp"
+#include "opencv2/core/private.gpu.hpp"
 
 #include <assert.h>
 #include <ctype.h>
@@ -66,28 +66,28 @@
 #define GET_OPTIMIZED(func) (func)
 #endif
 
-#ifdef HAVE_CUDA
-#  include <cuda_runtime_api.h>
-#  include "opencv2/core/gpumat.hpp"
-
-#  if defined(__GNUC__)
-#    define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__, __func__)
-#  else
-#    define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__)
-#  endif
-
-static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
-{
-    if (cudaSuccess != err) cv::gpu::error(cudaGetErrorString(err), file, line, func);
-}
-
-#else
-#  define cudaSafeCall(expr)
-#endif
-
 namespace cv
 {
 
+typedef void (*BinaryFunc)(const uchar* src1, size_t step1,
+                       const uchar* src2, size_t step2,
+                       uchar* dst, size_t step, Size sz,
+                       void*);
+
+BinaryFunc getConvertFunc(int sdepth, int ddepth);
+BinaryFunc getCopyMaskFunc(size_t esz);
+
+/* default memory block for sparse array elements */
+#define  CV_SPARSE_MAT_BLOCK     (1<<12)
+
+/* initial hash table size */
+#define  CV_SPARSE_HASH_SIZE0    (1<<10)
+
+/* maximal average node_count/hash_size ratio beyond which hash table is resized */
+#define  CV_SPARSE_HASH_RATIO    3
+
+
+
 // -128.f ... 255.f
 extern const float g_8x32fTab[];
 #define CV_8TO32F(x)  cv::g_8x32fTab[(x)+128]
@@ -194,11 +194,6 @@ extern volatile bool USE_AVX;
 
 enum { BLOCK_SIZE = 1024 };
 
-#ifdef HAVE_IPP
-static inline IppiSize ippiSize(int width, int height) { IppiSize sz = { width, height}; return sz; }
-static inline IppiSize ippiSize(Size _sz)              { IppiSize sz = { _sz.width, _sz.height}; return sz; }
-#endif
-
 #if defined HAVE_IPP && (IPP_VERSION_MAJOR >= 7)
 #define ARITHM_USE_IPP 1
 #define IF_IPP(then_call, else_call) then_call
diff --git a/modules/core/src/rand.cpp b/modules/core/src/rand.cpp
index 3ba7fd684d..ffce63484a 100644
--- a/modules/core/src/rand.cpp
+++ b/modules/core/src/rand.cpp
@@ -54,6 +54,8 @@
     #undef min
     #undef max
     #undef abs
+#else
+    #include <pthread.h>
 #endif
 
 #if defined __SSE2__ || (defined _M_IX86_FP && 2 == _M_IX86_FP)
@@ -726,33 +728,54 @@ void RNG::fill( InputOutputArray _mat, int disttype,
 }
 
 #ifdef WIN32
+
+
+#ifdef HAVE_WINRT
+// using C++11 thread attribute for local thread data
+__declspec( thread ) RNG* rng = NULL;
+
+ void deleteThreadRNGData()
+ {
+    if (rng)
+        delete rng;
+}
+
+RNG& theRNG()
+{
+    if (!rng)
+    {
+        rng =  new RNG;
+    }
+    return *rng;
+}
+#else
 #ifdef WINCE
 #	define TLS_OUT_OF_INDEXES ((DWORD)0xFFFFFFFF)
 #endif
 static DWORD tlsRNGKey = TLS_OUT_OF_INDEXES;
 
-void deleteThreadRNGData()
-{
-    if( tlsRNGKey != TLS_OUT_OF_INDEXES )
-        delete (RNG*)TlsGetValue( tlsRNGKey );
+ void deleteThreadRNGData()
+ {
+     if( tlsRNGKey != TLS_OUT_OF_INDEXES )
+         delete (RNG*)TlsGetValue( tlsRNGKey );
 }
 
 RNG& theRNG()
 {
     if( tlsRNGKey == TLS_OUT_OF_INDEXES )
     {
-        tlsRNGKey = TlsAlloc();
-        CV_Assert(tlsRNGKey != TLS_OUT_OF_INDEXES);
+       tlsRNGKey = TlsAlloc();
+       CV_Assert(tlsRNGKey != TLS_OUT_OF_INDEXES);
     }
     RNG* rng = (RNG*)TlsGetValue( tlsRNGKey );
     if( !rng )
     {
-        rng = new RNG;
-        TlsSetValue( tlsRNGKey, rng );
+       rng = new RNG;
+       TlsSetValue( tlsRNGKey, rng );
     }
     return *rng;
 }
-
+#endif //HAVE_WINRT
 #else
 
 static pthread_key_t tlsRNGKey = 0;
@@ -861,11 +884,6 @@ void cv::randShuffle( InputOutputArray _dst, double iterFactor, RNG* _rng )
     func( dst, rng, iterFactor );
 }
 
-void cv::randShuffle_( InputOutputArray _dst, double iterFactor )
-{
-    randShuffle(_dst, iterFactor);
-}
-
 CV_IMPL void
 cvRandArr( CvRNG* _rng, CvArr* arr, int disttype, CvScalar param1, CvScalar param2 )
 {
diff --git a/modules/core/src/stat.cpp b/modules/core/src/stat.cpp
index b62f10a2a8..86555fcc3b 100644
--- a/modules/core/src/stat.cpp
+++ b/modules/core/src/stat.cpp
@@ -42,6 +42,7 @@
 
 #include "precomp.hpp"
 #include <climits>
+#include <limits>
 
 namespace cv
 {
@@ -199,14 +200,19 @@ static int sum64f( const double* src, const uchar* mask, double* dst, int len, i
 
 typedef int (*SumFunc)(const uchar*, const uchar* mask, uchar*, int, int);
 
-static SumFunc sumTab[] =
+static SumFunc getSumFunc(int depth)
 {
-    (SumFunc)GET_OPTIMIZED(sum8u), (SumFunc)sum8s,
-    (SumFunc)sum16u, (SumFunc)sum16s,
-    (SumFunc)sum32s,
-    (SumFunc)GET_OPTIMIZED(sum32f), (SumFunc)sum64f,
-    0
-};
+    static SumFunc sumTab[] =
+    {
+        (SumFunc)GET_OPTIMIZED(sum8u), (SumFunc)sum8s,
+        (SumFunc)sum16u, (SumFunc)sum16s,
+        (SumFunc)sum32s,
+        (SumFunc)GET_OPTIMIZED(sum32f), (SumFunc)sum64f,
+        0
+    };
+
+    return sumTab[depth];
+}
 
 template<typename T>
 static int countNonZero_(const T* src, int len )
@@ -271,14 +277,18 @@ static int countNonZero64f( const double* src, int len )
 
 typedef int (*CountNonZeroFunc)(const uchar*, int);
 
-static CountNonZeroFunc countNonZeroTab[] =
+static CountNonZeroFunc getCountNonZeroTab(int depth)
 {
-    (CountNonZeroFunc)GET_OPTIMIZED(countNonZero8u), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero8u),
-    (CountNonZeroFunc)GET_OPTIMIZED(countNonZero16u), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero16u),
-    (CountNonZeroFunc)GET_OPTIMIZED(countNonZero32s), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero32f),
-    (CountNonZeroFunc)GET_OPTIMIZED(countNonZero64f), 0
-};
+    static CountNonZeroFunc countNonZeroTab[] =
+    {
+        (CountNonZeroFunc)GET_OPTIMIZED(countNonZero8u), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero8u),
+        (CountNonZeroFunc)GET_OPTIMIZED(countNonZero16u), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero16u),
+        (CountNonZeroFunc)GET_OPTIMIZED(countNonZero32s), (CountNonZeroFunc)GET_OPTIMIZED(countNonZero32f),
+        (CountNonZeroFunc)GET_OPTIMIZED(countNonZero64f), 0
+    };
 
+    return countNonZeroTab[depth];
+}
 
 template<typename T, typename ST, typename SQT>
 static int sumsqr_(const T* src0, const uchar* mask, ST* sum, SQT* sqsum, int len, int cn )
@@ -427,11 +437,16 @@ static int sqsum64f( const double* src, const uchar* mask, double* sum, double*
 
 typedef int (*SumSqrFunc)(const uchar*, const uchar* mask, uchar*, uchar*, int, int);
 
-static SumSqrFunc sumSqrTab[] =
+static SumSqrFunc getSumSqrTab(int depth)
 {
-    (SumSqrFunc)GET_OPTIMIZED(sqsum8u), (SumSqrFunc)sqsum8s, (SumSqrFunc)sqsum16u, (SumSqrFunc)sqsum16s,
-    (SumSqrFunc)sqsum32s, (SumSqrFunc)GET_OPTIMIZED(sqsum32f), (SumSqrFunc)sqsum64f, 0
-};
+    static SumSqrFunc sumSqrTab[] =
+    {
+        (SumSqrFunc)GET_OPTIMIZED(sqsum8u), (SumSqrFunc)sqsum8s, (SumSqrFunc)sqsum16u, (SumSqrFunc)sqsum16s,
+        (SumSqrFunc)sqsum32s, (SumSqrFunc)GET_OPTIMIZED(sqsum32f), (SumSqrFunc)sqsum64f, 0
+    };
+
+    return sumSqrTab[depth];
+}
 
 }
 
@@ -439,7 +454,46 @@ cv::Scalar cv::sum( InputArray _src )
 {
     Mat src = _src.getMat();
     int k, cn = src.channels(), depth = src.depth();
-    SumFunc func = sumTab[depth];
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src.total();
+    int rows = src.size[0], cols = (int)(total_size/rows);
+    if( src.dims == 2 || (src.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) )
+    {
+        IppiSize sz = { cols, rows };
+        int type = src.type();
+        typedef IppStatus (CV_STDCALL* ippiSumFunc)(const void*, int, IppiSize, double *, int);
+        ippiSumFunc ippFunc =
+            type == CV_8UC1 ? (ippiSumFunc)ippiSum_8u_C1R :
+            type == CV_8UC3 ? (ippiSumFunc)ippiSum_8u_C3R :
+            type == CV_8UC4 ? (ippiSumFunc)ippiSum_8u_C4R :
+            type == CV_16UC1 ? (ippiSumFunc)ippiSum_16u_C1R :
+            type == CV_16UC3 ? (ippiSumFunc)ippiSum_16u_C3R :
+            type == CV_16UC4 ? (ippiSumFunc)ippiSum_16u_C4R :
+            type == CV_16SC1 ? (ippiSumFunc)ippiSum_16s_C1R :
+            type == CV_16SC3 ? (ippiSumFunc)ippiSum_16s_C3R :
+            type == CV_16SC4 ? (ippiSumFunc)ippiSum_16s_C4R :
+            type == CV_32FC1 ? (ippiSumFunc)ippiSum_32f_C1R :
+            type == CV_32FC3 ? (ippiSumFunc)ippiSum_32f_C3R :
+            type == CV_32FC4 ? (ippiSumFunc)ippiSum_32f_C4R :
+            0;
+        if( ippFunc )
+        {
+            Ipp64f res[4];
+            if( ippFunc(src.data, src.step[0], sz, res, ippAlgHintAccurate) >= 0 )
+            {
+                Scalar sc;
+                for( int i = 0; i < cn; i++ )
+                {
+                    sc[i] = res[i];
+                }
+                return sc;
+            }
+        }
+    }
+#endif
+
+    SumFunc func = getSumFunc(depth);
 
     CV_Assert( cn <= 4 && func != 0 );
 
@@ -491,7 +545,7 @@ cv::Scalar cv::sum( InputArray _src )
 int cv::countNonZero( InputArray _src )
 {
     Mat src = _src.getMat();
-    CountNonZeroFunc func = countNonZeroTab[src.depth()];
+    CountNonZeroFunc func = getCountNonZeroTab(src.depth());
 
     CV_Assert( src.channels() == 1 && func != 0 );
 
@@ -512,7 +566,82 @@ cv::Scalar cv::mean( InputArray _src, InputArray _mask )
     CV_Assert( mask.empty() || mask.type() == CV_8U );
 
     int k, cn = src.channels(), depth = src.depth();
-    SumFunc func = sumTab[depth];
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src.total();
+    int rows = src.size[0], cols = (int)(total_size/rows);
+    if( src.dims == 2 || (src.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) )
+    {
+        IppiSize sz = { cols, rows };
+        int type = src.type();
+        if( !mask.empty() )
+        {
+            typedef IppStatus (CV_STDCALL* ippiMaskMeanFuncC1)(const void *, int, void *, int, IppiSize, Ipp64f *);
+            ippiMaskMeanFuncC1 ippFuncC1 =
+            type == CV_8UC1 ? (ippiMaskMeanFuncC1)ippiMean_8u_C1MR :
+            type == CV_16UC1 ? (ippiMaskMeanFuncC1)ippiMean_16u_C1MR :
+            type == CV_32FC1 ? (ippiMaskMeanFuncC1)ippiMean_32f_C1MR :
+            0;
+            if( ippFuncC1 )
+            {
+                Ipp64f res;
+                if( ippFuncC1(src.data, src.step[0], mask.data, mask.step[0], sz, &res) >= 0 )
+                {
+                    return Scalar(res);
+                }
+            }
+            typedef IppStatus (CV_STDCALL* ippiMaskMeanFuncC3)(const void *, int, void *, int, IppiSize, int, Ipp64f *);
+            ippiMaskMeanFuncC3 ippFuncC3 =
+            type == CV_8UC3 ? (ippiMaskMeanFuncC3)ippiMean_8u_C3CMR :
+            type == CV_16UC3 ? (ippiMaskMeanFuncC3)ippiMean_16u_C3CMR :
+            type == CV_32FC3 ? (ippiMaskMeanFuncC3)ippiMean_32f_C3CMR :
+            0;
+            if( ippFuncC3 )
+            {
+                Ipp64f res1, res2, res3;
+                if( ippFuncC3(src.data, src.step[0], mask.data, mask.step[0], sz, 1, &res1) >= 0 &&
+                    ippFuncC3(src.data, src.step[0], mask.data, mask.step[0], sz, 2, &res2) >= 0 &&
+                    ippFuncC3(src.data, src.step[0], mask.data, mask.step[0], sz, 3, &res3) >= 0 )
+                {
+                    return Scalar(res1, res2, res3);
+                }
+            }
+        }
+        else
+        {
+            typedef IppStatus (CV_STDCALL* ippiMeanFunc)(const void*, int, IppiSize, double *, int);
+            ippiMeanFunc ippFunc =
+                type == CV_8UC1 ? (ippiMeanFunc)ippiMean_8u_C1R :
+                type == CV_8UC3 ? (ippiMeanFunc)ippiMean_8u_C3R :
+                type == CV_8UC4 ? (ippiMeanFunc)ippiMean_8u_C4R :
+                type == CV_16UC1 ? (ippiMeanFunc)ippiMean_16u_C1R :
+                type == CV_16UC3 ? (ippiMeanFunc)ippiMean_16u_C3R :
+                type == CV_16UC4 ? (ippiMeanFunc)ippiMean_16u_C4R :
+                type == CV_16SC1 ? (ippiMeanFunc)ippiMean_16s_C1R :
+                type == CV_16SC3 ? (ippiMeanFunc)ippiMean_16s_C3R :
+                type == CV_16SC4 ? (ippiMeanFunc)ippiMean_16s_C4R :
+                type == CV_32FC1 ? (ippiMeanFunc)ippiMean_32f_C1R :
+                type == CV_32FC3 ? (ippiMeanFunc)ippiMean_32f_C3R :
+                type == CV_32FC4 ? (ippiMeanFunc)ippiMean_32f_C4R :
+                0;
+            if( ippFunc )
+            {
+                Ipp64f res[4];
+                if( ippFunc(src.data, src.step[0], sz, res, ippAlgHintAccurate) >= 0 )
+                {
+                    Scalar sc;
+                    for( int i = 0; i < cn; i++ )
+                    {
+                        sc[i] = res[i];
+                    }
+                    return sc;
+                }
+            }
+        }
+    }
+#endif
+
+    SumFunc func = getSumFunc(depth);
 
     CV_Assert( cn <= 4 && func != 0 );
 
@@ -571,7 +700,100 @@ void cv::meanStdDev( InputArray _src, OutputArray _mean, OutputArray _sdv, Input
     CV_Assert( mask.empty() || mask.type() == CV_8U );
 
     int k, cn = src.channels(), depth = src.depth();
-    SumSqrFunc func = sumSqrTab[depth];
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src.total();
+    int rows = src.size[0], cols = (int)(total_size/rows);
+    if( src.dims == 2 || (src.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) )
+    {
+        Ipp64f mean_temp[3];
+        Ipp64f stddev_temp[3];
+        Ipp64f *pmean = &mean_temp[0];
+        Ipp64f *pstddev = &stddev_temp[0];
+        Mat mean, stddev;
+        int dcn_mean = -1;
+        if( _mean.needed() )
+        {
+            if( !_mean.fixedSize() )
+                _mean.create(cn, 1, CV_64F, -1, true);
+            mean = _mean.getMat();
+            dcn_mean = (int)mean.total();
+            pmean = (Ipp64f *)mean.data;
+        }
+        int dcn_stddev = -1;
+        if( _sdv.needed() )
+        {
+            if( !_sdv.fixedSize() )
+                _sdv.create(cn, 1, CV_64F, -1, true);
+            stddev = _sdv.getMat();
+            dcn_stddev = (int)stddev.total();
+            pstddev = (Ipp64f *)stddev.data;
+        }
+        for( int k = cn; k < dcn_mean; k++ )
+            pmean[k] = 0;
+        for( int k = cn; k < dcn_stddev; k++ )
+            pstddev[k] = 0;
+        IppiSize sz = { cols, rows };
+        int type = src.type();
+        if( !mask.empty() )
+        {
+            typedef IppStatus (CV_STDCALL* ippiMaskMeanStdDevFuncC1)(const void *, int, void *, int, IppiSize, Ipp64f *, Ipp64f *);
+            ippiMaskMeanStdDevFuncC1 ippFuncC1 =
+            type == CV_8UC1 ? (ippiMaskMeanStdDevFuncC1)ippiMean_StdDev_8u_C1MR :
+            type == CV_16UC1 ? (ippiMaskMeanStdDevFuncC1)ippiMean_StdDev_16u_C1MR :
+            type == CV_32FC1 ? (ippiMaskMeanStdDevFuncC1)ippiMean_StdDev_32f_C1MR :
+            0;
+            if( ippFuncC1 )
+            {
+                if( ippFuncC1(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, pmean, pstddev) >= 0 )
+                    return;
+            }
+            typedef IppStatus (CV_STDCALL* ippiMaskMeanStdDevFuncC3)(const void *, int, void *, int, IppiSize, int, Ipp64f *, Ipp64f *);
+            ippiMaskMeanStdDevFuncC3 ippFuncC3 =
+            type == CV_8UC3 ? (ippiMaskMeanStdDevFuncC3)ippiMean_StdDev_8u_C3CMR :
+            type == CV_16UC3 ? (ippiMaskMeanStdDevFuncC3)ippiMean_StdDev_16u_C3CMR :
+            type == CV_32FC3 ? (ippiMaskMeanStdDevFuncC3)ippiMean_StdDev_32f_C3CMR :
+            0;
+            if( ippFuncC3 )
+            {
+                if( ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 1, &pmean[0], &pstddev[0]) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 2, &pmean[1], &pstddev[1]) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 3, &pmean[2], &pstddev[2]) >= 0 )
+                    return;
+            }
+        }
+        else
+        {
+            typedef IppStatus (CV_STDCALL* ippiMeanStdDevFuncC1)(const void *, int, IppiSize, Ipp64f *, Ipp64f *);
+            ippiMeanStdDevFuncC1 ippFuncC1 =
+            type == CV_8UC1 ? (ippiMeanStdDevFuncC1)ippiMean_StdDev_8u_C1R :
+            type == CV_16UC1 ? (ippiMeanStdDevFuncC1)ippiMean_StdDev_16u_C1R :
+            //type == CV_32FC1 ? (ippiMeanStdDevFuncC1)ippiMean_StdDev_32f_C1R ://Aug 2013: bug in IPP 7.1, 8.0
+            0;
+            if( ippFuncC1 )
+            {
+                if( ippFuncC1(src.data, (int)src.step[0], sz, pmean, pstddev) >= 0 )
+                    return;
+            }
+            typedef IppStatus (CV_STDCALL* ippiMeanStdDevFuncC3)(const void *, int, IppiSize, int, Ipp64f *, Ipp64f *);
+            ippiMeanStdDevFuncC3 ippFuncC3 =
+            type == CV_8UC3 ? (ippiMeanStdDevFuncC3)ippiMean_StdDev_8u_C3CR :
+            type == CV_16UC3 ? (ippiMeanStdDevFuncC3)ippiMean_StdDev_16u_C3CR :
+            type == CV_32FC3 ? (ippiMeanStdDevFuncC3)ippiMean_StdDev_32f_C3CR :
+            0;
+            if( ippFuncC3 )
+            {
+                if( ippFuncC3(src.data, (int)src.step[0], sz, 1, &pmean[0], &pstddev[0]) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], sz, 2, &pmean[1], &pstddev[1]) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], sz, 3, &pmean[2], &pstddev[2]) >= 0 )
+                    return;
+            }
+        }
+    }
+#endif
+
+
+    SumSqrFunc func = getSumSqrTab(depth);
 
     CV_Assert( func != 0 );
 
@@ -745,14 +967,19 @@ static void minMaxIdx_64f(const double* src, const uchar* mask, double* minval,
 
 typedef void (*MinMaxIdxFunc)(const uchar*, const uchar*, int*, int*, size_t*, size_t*, int, size_t);
 
-static MinMaxIdxFunc minmaxTab[] =
+static MinMaxIdxFunc getMinmaxTab(int depth)
 {
-    (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_8u), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_8s),
-    (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_16u), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_16s),
-    (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_32s),
-    (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_32f), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_64f),
-    0
-};
+    static MinMaxIdxFunc minmaxTab[] =
+    {
+        (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_8u), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_8s),
+        (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_16u), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_16s),
+        (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_32s),
+        (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_32f), (MinMaxIdxFunc)GET_OPTIMIZED(minMaxIdx_64f),
+        0
+    };
+
+    return minmaxTab[depth];
+}
 
 static void ofs2idx(const Mat& a, size_t ofs, int* idx)
 {
@@ -785,7 +1012,84 @@ void cv::minMaxIdx(InputArray _src, double* minVal,
 
     CV_Assert( (cn == 1 && (mask.empty() || mask.type() == CV_8U)) ||
                (cn >= 1 && mask.empty() && !minIdx && !maxIdx) );
-    MinMaxIdxFunc func = minmaxTab[depth];
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src.total();
+    int rows = src.size[0], cols = (int)(total_size/rows);
+    if( cn == 1 && ( src.dims == 2 || (src.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size) ) )
+    {
+        IppiSize sz = { cols, rows };
+        int type = src.type();
+        if( !mask.empty() )
+        {
+            typedef IppStatus (CV_STDCALL* ippiMaskMinMaxIndxFuncC1)(const void *, int, const void *, int, IppiSize, Ipp32f *, Ipp32f *, IppiPoint *, IppiPoint *);
+            ippiMaskMinMaxIndxFuncC1 ippFuncC1 =
+            type == CV_8UC1 ? (ippiMaskMinMaxIndxFuncC1)ippiMinMaxIndx_8u_C1MR :
+            type == CV_16UC1 ? (ippiMaskMinMaxIndxFuncC1)ippiMinMaxIndx_16u_C1MR :
+            type == CV_32FC1 ? (ippiMaskMinMaxIndxFuncC1)ippiMinMaxIndx_32f_C1MR :
+            0;
+            if( ippFuncC1 )
+            {
+                Ipp32f min, max;
+                IppiPoint minp, maxp;
+                if( ippFuncC1(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, &min, &max, &minp, &maxp) >= 0 )
+                {
+                    if( minVal )
+                        *minVal = (double)min;
+                    if( maxVal )
+                        *maxVal = (double)max;
+                    if( !minp.x && !minp.y && !maxp.x && !maxp.y && !mask.data[0] )
+                        minp.x = maxp.x = -1;
+                    if( minIdx )
+                    {
+                        size_t minidx = minp.y * cols + minp.x + 1;
+                        ofs2idx(src, minidx, minIdx);
+                    }
+                    if( maxIdx )
+                    {
+                        size_t maxidx = maxp.y * cols + maxp.x + 1;
+                        ofs2idx(src, maxidx, maxIdx);
+                    }
+                    return;
+                }
+            }
+        }
+        else
+        {
+            typedef IppStatus (CV_STDCALL* ippiMinMaxIndxFuncC1)(const void *, int, IppiSize, Ipp32f *, Ipp32f *, IppiPoint *, IppiPoint *);
+            ippiMinMaxIndxFuncC1 ippFuncC1 =
+                type == CV_8UC1 ? (ippiMinMaxIndxFuncC1)ippiMinMaxIndx_8u_C1R :
+                type == CV_16UC1 ? (ippiMinMaxIndxFuncC1)ippiMinMaxIndx_16u_C1R :
+                type == CV_32FC1 ? (ippiMinMaxIndxFuncC1)ippiMinMaxIndx_32f_C1R :
+                0;
+            if( ippFuncC1 )
+            {
+                Ipp32f min, max;
+                IppiPoint minp, maxp;
+                if( ippFuncC1(src.data, (int)src.step[0], sz, &min, &max, &minp, &maxp) >= 0 )
+                {
+                    if( minVal )
+                        *minVal = (double)min;
+                    if( maxVal )
+                        *maxVal = (double)max;
+                    if( minIdx )
+                    {
+                        size_t minidx = minp.y * cols + minp.x + 1;
+                        ofs2idx(src, minidx, minIdx);
+                    }
+                    if( maxIdx )
+                    {
+                        size_t maxidx = maxp.y * cols + maxp.x + 1;
+                        ofs2idx(src, maxidx, maxIdx);
+                    }
+                    return;
+                }
+            }
+        }
+    }
+#endif
+
+    MinMaxIdxFunc func = getMinmaxTab(depth);
     CV_Assert( func != 0 );
 
     const Mat* arrays[] = {&src, &mask, 0};
@@ -999,7 +1303,6 @@ static int normHamming(const uchar* a, int n)
 {
     int i = 0, result = 0;
 #if CV_NEON
-    if (CPU_HAS_NEON_FEATURE)
     {
         uint32x4_t bits = vmovq_n_u32(0);
         for (; i <= n - 16; i += 16) {
@@ -1013,7 +1316,6 @@ static int normHamming(const uchar* a, int n)
         result = vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),0);
         result += vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),2);
     }
-    else
 #endif
         for( ; i <= n - 4; i += 4 )
             result += popCountTable[a[i]] + popCountTable[a[i+1]] +
@@ -1027,7 +1329,6 @@ int normHamming(const uchar* a, const uchar* b, int n)
 {
     int i = 0, result = 0;
 #if CV_NEON
-    if (CPU_HAS_NEON_FEATURE)
     {
         uint32x4_t bits = vmovq_n_u32(0);
         for (; i <= n - 16; i += 16) {
@@ -1043,7 +1344,6 @@ int normHamming(const uchar* a, const uchar* b, int n)
         result = vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),0);
         result += vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),2);
     }
-    else
 #endif
         for( ; i <= n - 4; i += 4 )
             result += popCountTable[a[i] ^ b[i]] + popCountTable[a[i+1] ^ b[i+1]] +
@@ -1111,7 +1411,7 @@ normInf_(const T* src, const uchar* mask, ST* _result, int len, int cn)
             if( mask[i] )
             {
                 for( int k = 0; k < cn; k++ )
-                    result = std::max(result, ST(fast_abs(src[k])));
+                    result = std::max(result, ST(std::abs(src[k])));
             }
     }
     *_result = result;
@@ -1132,7 +1432,7 @@ normL1_(const T* src, const uchar* mask, ST* _result, int len, int cn)
             if( mask[i] )
             {
                 for( int k = 0; k < cn; k++ )
-                    result += fast_abs(src[k]);
+                    result += std::abs(src[k]);
             }
     }
     *_result = result;
@@ -1254,43 +1554,53 @@ CV_DEF_NORM_ALL(64f, double, double, double, double)
 typedef int (*NormFunc)(const uchar*, const uchar*, uchar*, int, int);
 typedef int (*NormDiffFunc)(const uchar*, const uchar*, const uchar*, uchar*, int, int);
 
-static NormFunc normTab[3][8] =
+static NormFunc getNormFunc(int normType, int depth)
 {
+    static NormFunc normTab[3][8] =
     {
-        (NormFunc)GET_OPTIMIZED(normInf_8u), (NormFunc)GET_OPTIMIZED(normInf_8s), (NormFunc)GET_OPTIMIZED(normInf_16u), (NormFunc)GET_OPTIMIZED(normInf_16s),
-        (NormFunc)GET_OPTIMIZED(normInf_32s), (NormFunc)GET_OPTIMIZED(normInf_32f), (NormFunc)normInf_64f, 0
-    },
-    {
-        (NormFunc)GET_OPTIMIZED(normL1_8u), (NormFunc)GET_OPTIMIZED(normL1_8s), (NormFunc)GET_OPTIMIZED(normL1_16u), (NormFunc)GET_OPTIMIZED(normL1_16s),
-        (NormFunc)GET_OPTIMIZED(normL1_32s), (NormFunc)GET_OPTIMIZED(normL1_32f), (NormFunc)normL1_64f, 0
-    },
-    {
-        (NormFunc)GET_OPTIMIZED(normL2_8u), (NormFunc)GET_OPTIMIZED(normL2_8s), (NormFunc)GET_OPTIMIZED(normL2_16u), (NormFunc)GET_OPTIMIZED(normL2_16s),
-        (NormFunc)GET_OPTIMIZED(normL2_32s), (NormFunc)GET_OPTIMIZED(normL2_32f), (NormFunc)normL2_64f, 0
-    }
-};
+        {
+            (NormFunc)GET_OPTIMIZED(normInf_8u), (NormFunc)GET_OPTIMIZED(normInf_8s), (NormFunc)GET_OPTIMIZED(normInf_16u), (NormFunc)GET_OPTIMIZED(normInf_16s),
+            (NormFunc)GET_OPTIMIZED(normInf_32s), (NormFunc)GET_OPTIMIZED(normInf_32f), (NormFunc)normInf_64f, 0
+        },
+        {
+            (NormFunc)GET_OPTIMIZED(normL1_8u), (NormFunc)GET_OPTIMIZED(normL1_8s), (NormFunc)GET_OPTIMIZED(normL1_16u), (NormFunc)GET_OPTIMIZED(normL1_16s),
+            (NormFunc)GET_OPTIMIZED(normL1_32s), (NormFunc)GET_OPTIMIZED(normL1_32f), (NormFunc)normL1_64f, 0
+        },
+        {
+            (NormFunc)GET_OPTIMIZED(normL2_8u), (NormFunc)GET_OPTIMIZED(normL2_8s), (NormFunc)GET_OPTIMIZED(normL2_16u), (NormFunc)GET_OPTIMIZED(normL2_16s),
+            (NormFunc)GET_OPTIMIZED(normL2_32s), (NormFunc)GET_OPTIMIZED(normL2_32f), (NormFunc)normL2_64f, 0
+        }
+    };
 
-static NormDiffFunc normDiffTab[3][8] =
+    return normTab[normType][depth];
+}
+
+static NormDiffFunc getNormDiffFunc(int normType, int depth)
 {
+    static NormDiffFunc normDiffTab[3][8] =
     {
-        (NormDiffFunc)GET_OPTIMIZED(normDiffInf_8u), (NormDiffFunc)normDiffInf_8s,
-        (NormDiffFunc)normDiffInf_16u, (NormDiffFunc)normDiffInf_16s,
-        (NormDiffFunc)normDiffInf_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffInf_32f),
-        (NormDiffFunc)normDiffInf_64f, 0
-    },
-    {
-        (NormDiffFunc)GET_OPTIMIZED(normDiffL1_8u), (NormDiffFunc)normDiffL1_8s,
-        (NormDiffFunc)normDiffL1_16u, (NormDiffFunc)normDiffL1_16s,
-        (NormDiffFunc)normDiffL1_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffL1_32f),
-        (NormDiffFunc)normDiffL1_64f, 0
-    },
-    {
-        (NormDiffFunc)GET_OPTIMIZED(normDiffL2_8u), (NormDiffFunc)normDiffL2_8s,
-        (NormDiffFunc)normDiffL2_16u, (NormDiffFunc)normDiffL2_16s,
-        (NormDiffFunc)normDiffL2_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffL2_32f),
-        (NormDiffFunc)normDiffL2_64f, 0
-    }
-};
+        {
+            (NormDiffFunc)GET_OPTIMIZED(normDiffInf_8u), (NormDiffFunc)normDiffInf_8s,
+            (NormDiffFunc)normDiffInf_16u, (NormDiffFunc)normDiffInf_16s,
+            (NormDiffFunc)normDiffInf_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffInf_32f),
+            (NormDiffFunc)normDiffInf_64f, 0
+        },
+        {
+            (NormDiffFunc)GET_OPTIMIZED(normDiffL1_8u), (NormDiffFunc)normDiffL1_8s,
+            (NormDiffFunc)normDiffL1_16u, (NormDiffFunc)normDiffL1_16s,
+            (NormDiffFunc)normDiffL1_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffL1_32f),
+            (NormDiffFunc)normDiffL1_64f, 0
+        },
+        {
+            (NormDiffFunc)GET_OPTIMIZED(normDiffL2_8u), (NormDiffFunc)normDiffL2_8s,
+            (NormDiffFunc)normDiffL2_16u, (NormDiffFunc)normDiffL2_16s,
+            (NormDiffFunc)normDiffL2_32s, (NormDiffFunc)GET_OPTIMIZED(normDiffL2_32f),
+            (NormDiffFunc)normDiffL2_64f, 0
+        }
+    };
+
+    return normDiffTab[normType][depth];
+}
 
 }
 
@@ -1303,6 +1613,147 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
     CV_Assert( normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR ||
                ((normType == NORM_HAMMING || normType == NORM_HAMMING2) && src.type() == CV_8U) );
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src.total();
+    int rows = src.size[0], cols = (int)(total_size/rows);
+    if( src.dims == 2 || (src.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size)
+        && (normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR) )
+    {
+        IppiSize sz = { cols, rows };
+        int type = src.type();
+        if( !mask.empty() )
+        {
+            typedef IppStatus (CV_STDCALL* ippiMaskNormFuncC1)(const void *, int, const void *, int, IppiSize, Ipp64f *);
+            ippiMaskNormFuncC1 ippFuncC1 =
+                normType == NORM_INF ?
+                (type == CV_8UC1 ? (ippiMaskNormFuncC1)ippiNorm_Inf_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormFuncC1)ippiNorm_Inf_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormFuncC1)ippiNorm_Inf_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormFuncC1)ippiNorm_Inf_32f_C1MR :
+                0) :
+            normType == NORM_L1 ?
+                (type == CV_8UC1 ? (ippiMaskNormFuncC1)ippiNorm_L1_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormFuncC1)ippiNorm_L1_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormFuncC1)ippiNorm_L1_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormFuncC1)ippiNorm_L1_32f_C1MR :
+                0) :
+            normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC1 ? (ippiMaskNormFuncC1)ippiNorm_L2_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormFuncC1)ippiNorm_L2_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormFuncC1)ippiNorm_L2_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormFuncC1)ippiNorm_L2_32f_C1MR :
+                0) : 0;
+            if( ippFuncC1 )
+            {
+                Ipp64f norm;
+                if( ippFuncC1(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, &norm) >= 0 )
+                {
+                    return normType == NORM_L2SQR ? (double)(norm * norm) : (double)norm;
+                }
+            }
+            typedef IppStatus (CV_STDCALL* ippiMaskNormFuncC3)(const void *, int, const void *, int, IppiSize, int, Ipp64f *);
+            ippiMaskNormFuncC3 ippFuncC3 =
+                normType == NORM_INF ?
+                (type == CV_8UC3 ? (ippiMaskNormFuncC3)ippiNorm_Inf_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormFuncC3)ippiNorm_Inf_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormFuncC3)ippiNorm_Inf_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormFuncC3)ippiNorm_Inf_32f_C3CMR :
+                0) :
+            normType == NORM_L1 ?
+                (type == CV_8UC3 ? (ippiMaskNormFuncC3)ippiNorm_L1_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormFuncC3)ippiNorm_L1_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormFuncC3)ippiNorm_L1_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormFuncC3)ippiNorm_L1_32f_C3CMR :
+                0) :
+            normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC3 ? (ippiMaskNormFuncC3)ippiNorm_L2_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormFuncC3)ippiNorm_L2_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormFuncC3)ippiNorm_L2_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormFuncC3)ippiNorm_L2_32f_C3CMR :
+                0) : 0;
+            if( ippFuncC3 )
+            {
+                Ipp64f norm1, norm2, norm3;
+                if( ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 1, &norm1) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 2, &norm2) >= 0 &&
+                    ippFuncC3(src.data, (int)src.step[0], mask.data, (int)mask.step[0], sz, 3, &norm3) >= 0)
+                {
+                    Ipp64f norm =
+                        normType == NORM_INF ? std::max(std::max(norm1, norm2), norm3) :
+                        normType == NORM_L1 ? norm1 + norm2 + norm3 :
+                        normType == NORM_L2 || normType == NORM_L2SQR ? std::sqrt(norm1 * norm1 + norm2 * norm2 + norm3 * norm3) :
+                        0;
+                    return normType == NORM_L2SQR ? (double)(norm * norm) : (double)norm;
+                }
+            }
+        }
+        else
+        {
+            typedef IppStatus (CV_STDCALL* ippiNormFunc)(const void *, int, IppiSize, Ipp64f *, IppHintAlgorithm hint);
+            ippiNormFunc ippFunc =
+                normType == NORM_INF ?
+                (type == CV_8UC1 ? (ippiNormFunc)ippiNorm_Inf_8u_C1R :
+                type == CV_8UC3 ? (ippiNormFunc)ippiNorm_Inf_8u_C3R :
+                type == CV_8UC4 ? (ippiNormFunc)ippiNorm_Inf_8u_C4R :
+                type == CV_16UC1 ? (ippiNormFunc)ippiNorm_Inf_16u_C1R :
+                type == CV_16UC3 ? (ippiNormFunc)ippiNorm_Inf_16u_C3R :
+                type == CV_16UC4 ? (ippiNormFunc)ippiNorm_Inf_16u_C4R :
+                type == CV_16SC1 ? (ippiNormFunc)ippiNorm_Inf_16s_C1R :
+                //type == CV_16SC3 ? (ippiNormFunc)ippiNorm_Inf_16s_C3R : //Aug 2013: problem in IPP 7.1, 8.0 : -32768
+                //type == CV_16SC4 ? (ippiNormFunc)ippiNorm_Inf_16s_C4R : //Aug 2013: problem in IPP 7.1, 8.0 : -32768
+                type == CV_32FC1 ? (ippiNormFunc)ippiNorm_Inf_32f_C1R :
+                type == CV_32FC3 ? (ippiNormFunc)ippiNorm_Inf_32f_C3R :
+                type == CV_32FC4 ? (ippiNormFunc)ippiNorm_Inf_32f_C4R :
+                0) :
+                normType == NORM_L1 ?
+                (type == CV_8UC1 ? (ippiNormFunc)ippiNorm_L1_8u_C1R :
+                type == CV_8UC3 ? (ippiNormFunc)ippiNorm_L1_8u_C3R :
+                type == CV_8UC4 ? (ippiNormFunc)ippiNorm_L1_8u_C4R :
+                type == CV_16UC1 ? (ippiNormFunc)ippiNorm_L1_16u_C1R :
+                type == CV_16UC3 ? (ippiNormFunc)ippiNorm_L1_16u_C3R :
+                type == CV_16UC4 ? (ippiNormFunc)ippiNorm_L1_16u_C4R :
+                type == CV_16SC1 ? (ippiNormFunc)ippiNorm_L1_16s_C1R :
+                type == CV_16SC3 ? (ippiNormFunc)ippiNorm_L1_16s_C3R :
+                type == CV_16SC4 ? (ippiNormFunc)ippiNorm_L1_16s_C4R :
+                type == CV_32FC1 ? (ippiNormFunc)ippiNorm_L1_32f_C1R :
+                type == CV_32FC3 ? (ippiNormFunc)ippiNorm_L1_32f_C3R :
+                type == CV_32FC4 ? (ippiNormFunc)ippiNorm_L1_32f_C4R :
+                0) :
+                normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC1 ? (ippiNormFunc)ippiNorm_L2_8u_C1R :
+                type == CV_8UC3 ? (ippiNormFunc)ippiNorm_L2_8u_C3R :
+                type == CV_8UC4 ? (ippiNormFunc)ippiNorm_L2_8u_C4R :
+                type == CV_16UC1 ? (ippiNormFunc)ippiNorm_L2_16u_C1R :
+                type == CV_16UC3 ? (ippiNormFunc)ippiNorm_L2_16u_C3R :
+                type == CV_16UC4 ? (ippiNormFunc)ippiNorm_L2_16u_C4R :
+                type == CV_16SC1 ? (ippiNormFunc)ippiNorm_L2_16s_C1R :
+                type == CV_16SC3 ? (ippiNormFunc)ippiNorm_L2_16s_C3R :
+                type == CV_16SC4 ? (ippiNormFunc)ippiNorm_L2_16s_C4R :
+                type == CV_32FC1 ? (ippiNormFunc)ippiNorm_L2_32f_C1R :
+                type == CV_32FC3 ? (ippiNormFunc)ippiNorm_L2_32f_C3R :
+                type == CV_32FC4 ? (ippiNormFunc)ippiNorm_L2_32f_C4R :
+                0) : 0;
+            if( ippFunc )
+            {
+                Ipp64f norm_array[4];
+                if( ippFunc(src.data, (int)src.step[0], sz, norm_array, ippAlgHintAccurate) >= 0 )
+                {
+                    Ipp64f norm = (normType == NORM_L2 || normType == NORM_L2SQR) ? norm_array[0] * norm_array[0] : norm_array[0];
+                    for( int i = 1; i < cn; i++ )
+                    {
+                        norm =
+                            normType == NORM_INF ? std::max(norm, norm_array[i]) :
+                            normType == NORM_L1 ? norm + norm_array[i] :
+                            normType == NORM_L2 || normType == NORM_L2SQR ? norm + norm_array[i] * norm_array[i] :
+                            0;
+                    }
+                    return normType == NORM_L2 ? (double)std::sqrt(norm) : (double)norm;
+                }
+            }
+        }
+    }
+#endif
+
     if( src.isContinuous() && mask.empty() )
     {
         size_t len = src.total()*cn;
@@ -1374,7 +1825,7 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
         return result;
     }
 
-    NormFunc func = normTab[normType >> 1][depth];
+    NormFunc func = getNormFunc(normType >> 1, depth);
     CV_Assert( func != 0 );
 
     const Mat* arrays[] = {&src, &mask, 0};
@@ -1441,7 +1892,84 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
 double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _mask )
 {
     if( normType & CV_RELATIVE )
+    {
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+        Mat src1 = _src1.getMat(), src2 = _src2.getMat(), mask = _mask.getMat();
+
+        CV_Assert( src1.size == src2.size && src1.type() == src2.type() );
+
+        normType &= 7;
+        CV_Assert( normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR ||
+                ((normType == NORM_HAMMING || normType == NORM_HAMMING2) && src1.type() == CV_8U) );
+        size_t total_size = src1.total();
+        int rows = src1.size[0], cols = (int)(total_size/rows);
+        if( src1.dims == 2 || (src1.isContinuous() && src2.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size)
+            && (normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR) )
+        {
+            IppiSize sz = { cols, rows };
+            int type = src1.type();
+            if( !mask.empty() )
+            {
+                typedef IppStatus (CV_STDCALL* ippiMaskNormRelFuncC1)(const void *, int, const void *, int, const void *, int, IppiSize, Ipp64f *);
+                ippiMaskNormRelFuncC1 ippFuncC1 =
+                    normType == NORM_INF ?
+                    (type == CV_8UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_Inf_8u_C1MR :
+                    type == CV_8SC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_Inf_8s_C1MR :
+                    type == CV_16UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_Inf_16u_C1MR :
+                    type == CV_32FC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_Inf_32f_C1MR :
+                    0) :
+                    normType == NORM_L1 ?
+                    (type == CV_8UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L1_8u_C1MR :
+                    type == CV_8SC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L1_8s_C1MR :
+                    type == CV_16UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L1_16u_C1MR :
+                    type == CV_32FC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L1_32f_C1MR :
+                    0) :
+                    normType == NORM_L2 || normType == NORM_L2SQR ?
+                    (type == CV_8UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L2_8u_C1MR :
+                    type == CV_8SC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L2_8s_C1MR :
+                    type == CV_16UC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L2_16u_C1MR :
+                    type == CV_32FC1 ? (ippiMaskNormRelFuncC1)ippiNormRel_L2_32f_C1MR :
+                    0) : 0;
+                if( ippFuncC1 )
+                {
+                    Ipp64f norm;
+                    if( ippFuncC1(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], mask.data, (int)mask.step[0], sz, &norm) >= 0 )
+                        return normType == NORM_L2SQR ? (double)(norm * norm) : (double)norm;
+                }
+            }
+            else
+            {
+                typedef IppStatus (CV_STDCALL* ippiNormRelFunc)(const void *, int, const void *, int, IppiSize, Ipp64f *, IppHintAlgorithm hint);
+                ippiNormRelFunc ippFunc =
+                    normType == NORM_INF ?
+                    (type == CV_8UC1 ? (ippiNormRelFunc)ippiNormRel_Inf_8u_C1R :
+                    type == CV_16UC1 ? (ippiNormRelFunc)ippiNormRel_Inf_16u_C1R :
+                    type == CV_16SC1 ? (ippiNormRelFunc)ippiNormRel_Inf_16s_C1R :
+                    type == CV_32FC1 ? (ippiNormRelFunc)ippiNormRel_Inf_32f_C1R :
+                    0) :
+                    normType == NORM_L1 ?
+                    (type == CV_8UC1 ? (ippiNormRelFunc)ippiNormRel_L1_8u_C1R :
+                    type == CV_16UC1 ? (ippiNormRelFunc)ippiNormRel_L1_16u_C1R :
+                    type == CV_16SC1 ? (ippiNormRelFunc)ippiNormRel_L1_16s_C1R :
+                    type == CV_32FC1 ? (ippiNormRelFunc)ippiNormRel_L1_32f_C1R :
+                    0) :
+                    normType == NORM_L2 || normType == NORM_L2SQR ?
+                    (type == CV_8UC1 ? (ippiNormRelFunc)ippiNormRel_L2_8u_C1R :
+                    type == CV_16UC1 ? (ippiNormRelFunc)ippiNormRel_L2_16u_C1R :
+                    type == CV_16SC1 ? (ippiNormRelFunc)ippiNormRel_L2_16s_C1R :
+                    type == CV_32FC1 ? (ippiNormRelFunc)ippiNormRel_L2_32f_C1R :
+                    0) : 0;
+                if( ippFunc )
+                {
+                    Ipp64f norm;
+                    if( ippFunc(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], sz, &norm, ippAlgHintAccurate) >= 0 )
+                        return (double)norm;
+                }
+            }
+        }
+#endif
         return norm(_src1, _src2, normType & ~CV_RELATIVE, _mask)/(norm(_src2, normType, _mask) + DBL_EPSILON);
+    }
 
     Mat src1 = _src1.getMat(), src2 = _src2.getMat(), mask = _mask.getMat();
     int depth = src1.depth(), cn = src1.channels();
@@ -1452,6 +1980,145 @@ double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _m
     CV_Assert( normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR ||
               ((normType == NORM_HAMMING || normType == NORM_HAMMING2) && src1.type() == CV_8U) );
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    size_t total_size = src1.total();
+    int rows = src1.size[0], cols = (int)(total_size/rows);
+    if( src1.dims == 2 || (src1.isContinuous() && src2.isContinuous() && mask.isContinuous() && cols > 0 && (size_t)rows*cols == total_size)
+        && (normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 || normType == NORM_L2SQR) )
+    {
+        IppiSize sz = { cols, rows };
+        int type = src1.type();
+        if( !mask.empty() )
+        {
+            typedef IppStatus (CV_STDCALL* ippiMaskNormDiffFuncC1)(const void *, int, const void *, int, const void *, int, IppiSize, Ipp64f *);
+            ippiMaskNormDiffFuncC1 ippFuncC1 =
+                normType == NORM_INF ?
+                (type == CV_8UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_Inf_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_Inf_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_Inf_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_Inf_32f_C1MR :
+                0) :
+                normType == NORM_L1 ?
+                (type == CV_8UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L1_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L1_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L1_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L1_32f_C1MR :
+                0) :
+                normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L2_8u_C1MR :
+                type == CV_8SC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L2_8s_C1MR :
+                type == CV_16UC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L2_16u_C1MR :
+                type == CV_32FC1 ? (ippiMaskNormDiffFuncC1)ippiNormDiff_L2_32f_C1MR :
+                0) : 0;
+            if( ippFuncC1 )
+            {
+                Ipp64f norm;
+                if( ippFuncC1(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], mask.data, (int)mask.step[0], sz, &norm) >= 0 )
+                    return normType == NORM_L2SQR ? (double)(norm * norm) : (double)norm;
+            }
+            typedef IppStatus (CV_STDCALL* ippiMaskNormDiffFuncC3)(const void *, int, const void *, int, const void *, int, IppiSize, int, Ipp64f *);
+            ippiMaskNormDiffFuncC3 ippFuncC3 =
+                normType == NORM_INF ?
+                (type == CV_8UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_Inf_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_Inf_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_Inf_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_Inf_32f_C3CMR :
+                0) :
+                normType == NORM_L1 ?
+                (type == CV_8UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L1_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L1_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L1_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L1_32f_C3CMR :
+                0) :
+                normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L2_8u_C3CMR :
+                type == CV_8SC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L2_8s_C3CMR :
+                type == CV_16UC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L2_16u_C3CMR :
+                type == CV_32FC3 ? (ippiMaskNormDiffFuncC3)ippiNormDiff_L2_32f_C3CMR :
+                0) : 0;
+            if( ippFuncC3 )
+            {
+                Ipp64f norm1, norm2, norm3;
+                if( ippFuncC3(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], mask.data, (int)mask.step[0], sz, 1, &norm1) >= 0 &&
+                    ippFuncC3(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], mask.data, (int)mask.step[0], sz, 2, &norm2) >= 0 &&
+                    ippFuncC3(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], mask.data, (int)mask.step[0], sz, 3, &norm3) >= 0)
+                {
+                    Ipp64f norm =
+                        normType == NORM_INF ? std::max(std::max(norm1, norm2), norm3) :
+                        normType == NORM_L1 ? norm1 + norm2 + norm3 :
+                        normType == NORM_L2 || normType == NORM_L2SQR ? std::sqrt(norm1 * norm1 + norm2 * norm2 + norm3 * norm3) :
+                        0;
+                    return normType == NORM_L2SQR ? (double)(norm * norm) : (double)norm;
+                }
+            }
+        }
+        else
+        {
+            typedef IppStatus (CV_STDCALL* ippiNormDiffFunc)(const void *, int, const void *, int, IppiSize, Ipp64f *, IppHintAlgorithm hint);
+            ippiNormDiffFunc ippFunc =
+                normType == NORM_INF ?
+                (type == CV_8UC1 ? (ippiNormDiffFunc)ippiNormDiff_Inf_8u_C1R :
+                type == CV_8UC3 ? (ippiNormDiffFunc)ippiNormDiff_Inf_8u_C3R :
+                type == CV_8UC4 ? (ippiNormDiffFunc)ippiNormDiff_Inf_8u_C4R :
+                type == CV_16UC1 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16u_C1R :
+                type == CV_16UC3 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16u_C3R :
+                type == CV_16UC4 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16u_C4R :
+                type == CV_16SC1 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16s_C1R :
+                //type == CV_16SC3 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16s_C3R : //Aug 2013: problem in IPP 7.1, 8.0 : -32768
+                //type == CV_16SC4 ? (ippiNormDiffFunc)ippiNormDiff_Inf_16s_C4R : //Aug 2013: problem in IPP 7.1, 8.0 : -32768
+                type == CV_32FC1 ? (ippiNormDiffFunc)ippiNormDiff_Inf_32f_C1R :
+                type == CV_32FC3 ? (ippiNormDiffFunc)ippiNormDiff_Inf_32f_C3R :
+                type == CV_32FC4 ? (ippiNormDiffFunc)ippiNormDiff_Inf_32f_C4R :
+                0) :
+                normType == NORM_L1 ?
+                (type == CV_8UC1 ? (ippiNormDiffFunc)ippiNormDiff_L1_8u_C1R :
+                type == CV_8UC3 ? (ippiNormDiffFunc)ippiNormDiff_L1_8u_C3R :
+                type == CV_8UC4 ? (ippiNormDiffFunc)ippiNormDiff_L1_8u_C4R :
+                type == CV_16UC1 ? (ippiNormDiffFunc)ippiNormDiff_L1_16u_C1R :
+                type == CV_16UC3 ? (ippiNormDiffFunc)ippiNormDiff_L1_16u_C3R :
+                type == CV_16UC4 ? (ippiNormDiffFunc)ippiNormDiff_L1_16u_C4R :
+                type == CV_16SC1 ? (ippiNormDiffFunc)ippiNormDiff_L1_16s_C1R :
+                type == CV_16SC3 ? (ippiNormDiffFunc)ippiNormDiff_L1_16s_C3R :
+                type == CV_16SC4 ? (ippiNormDiffFunc)ippiNormDiff_L1_16s_C4R :
+                type == CV_32FC1 ? (ippiNormDiffFunc)ippiNormDiff_L1_32f_C1R :
+                type == CV_32FC3 ? (ippiNormDiffFunc)ippiNormDiff_L1_32f_C3R :
+                type == CV_32FC4 ? (ippiNormDiffFunc)ippiNormDiff_L1_32f_C4R :
+                0) :
+                normType == NORM_L2 || normType == NORM_L2SQR ?
+                (type == CV_8UC1 ? (ippiNormDiffFunc)ippiNormDiff_L2_8u_C1R :
+                type == CV_8UC3 ? (ippiNormDiffFunc)ippiNormDiff_L2_8u_C3R :
+                type == CV_8UC4 ? (ippiNormDiffFunc)ippiNormDiff_L2_8u_C4R :
+                type == CV_16UC1 ? (ippiNormDiffFunc)ippiNormDiff_L2_16u_C1R :
+                type == CV_16UC3 ? (ippiNormDiffFunc)ippiNormDiff_L2_16u_C3R :
+                type == CV_16UC4 ? (ippiNormDiffFunc)ippiNormDiff_L2_16u_C4R :
+                type == CV_16SC1 ? (ippiNormDiffFunc)ippiNormDiff_L2_16s_C1R :
+                type == CV_16SC3 ? (ippiNormDiffFunc)ippiNormDiff_L2_16s_C3R :
+                type == CV_16SC4 ? (ippiNormDiffFunc)ippiNormDiff_L2_16s_C4R :
+                type == CV_32FC1 ? (ippiNormDiffFunc)ippiNormDiff_L2_32f_C1R :
+                type == CV_32FC3 ? (ippiNormDiffFunc)ippiNormDiff_L2_32f_C3R :
+                type == CV_32FC4 ? (ippiNormDiffFunc)ippiNormDiff_L2_32f_C4R :
+                0) : 0;
+            if( ippFunc )
+            {
+                Ipp64f norm_array[4];
+                if( ippFunc(src1.data, (int)src1.step[0], src2.data, (int)src2.step[0], sz, norm_array, ippAlgHintAccurate) >= 0 )
+                {
+                    Ipp64f norm = (normType == NORM_L2 || normType == NORM_L2SQR) ? norm_array[0] * norm_array[0] : norm_array[0];
+                    for( int i = 1; i < src1.channels(); i++ )
+                    {
+                        norm =
+                            normType == NORM_INF ? std::max(norm, norm_array[i]) :
+                            normType == NORM_L1 ? norm + norm_array[i] :
+                            normType == NORM_L2 || normType == NORM_L2SQR ? norm + norm_array[i] * norm_array[i] :
+                            0;
+                    }
+                    return normType == NORM_L2 ? (double)std::sqrt(norm) : (double)norm;
+                }
+            }
+        }
+    }
+#endif
+
     if( src1.isContinuous() && src2.isContinuous() && mask.empty() )
     {
         size_t len = src1.total()*src1.channels();
@@ -1515,7 +2182,7 @@ double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _m
         return result;
     }
 
-    NormDiffFunc func = normDiffTab[normType >> 1][depth];
+    NormDiffFunc func = getNormDiffFunc(normType >> 1, depth);
     CV_Assert( func != 0 );
 
     const Mat* arrays[] = {&src1, &src2, &mask, 0};
@@ -1925,6 +2592,14 @@ void cv::findNonZero( InputArray _src, OutputArray _idx )
     }
 }
 
+double cv::PSNR(InputArray _src1, InputArray _src2)
+{
+    Mat src1 = _src1.getMat(), src2 = _src2.getMat();
+    CV_Assert( src1.depth() == CV_8U );
+    double diff = std::sqrt(norm(src1, src2, NORM_L2SQR)/(src1.total()*src1.channels()));
+    return 20*log10(255./(diff+DBL_EPSILON));
+}
+
 
 CV_IMPL CvScalar cvSum( const CvArr* srcarr )
 {
diff --git a/modules/gpu/src/cu_safe_call.h b/modules/core/src/stl.cpp
similarity index 79%
rename from modules/gpu/src/cu_safe_call.h
rename to modules/core/src/stl.cpp
index 6f93adc760..f03c1a2d79 100644
--- a/modules/gpu/src/cu_safe_call.h
+++ b/modules/core/src/stl.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,28 +41,29 @@
 //
 //M*/
 
-#ifndef __CU_SAFE_CALL_H__
-#define __CU_SAFE_CALL_H__
 
 #include "precomp.hpp"
 
-#ifdef HAVE_CUDA
+char* cv::String::allocate(size_t len)
+{
+    size_t totalsize = alignSize(len + 1, (int)sizeof(int));
+    int* data = (int*)cv::fastMalloc(totalsize + sizeof(int));
+    data[0] = 1;
+    cstr_ = (char*)(data + 1);
+    len_ = len;
+    cstr_[len] = 0;
+    return cstr_;
+}
 
-namespace cv { namespace gpu {
-    namespace detail
+
+void cv::String::deallocate()
+{
+    int* data = (int*)cstr_;
+    len_ = 0;
+    cstr_ = 0;
+
+    if(data && 1 == CV_XADD(data-1, -1))
     {
-        std::string cuGetErrString(CUresult res);
-
-        inline void cuSafeCall_impl(CUresult res, const char* file, int line)
-        {
-            if (res != CUDA_SUCCESS)
-                cv::error( cv::Exception(CV_GpuApiCallError, cuGetErrString(res), "unknown function", file, line) );
-        }
+        cv::fastFree(data-1);
     }
-}}
-
-#define cuSafeCall( op ) cv::gpu::detail::cuSafeCall_impl( (op), __FILE__, __LINE__ )
-
-#endif // HAVE_CUDA
-
-#endif // __CU_SAFE_CALL_H__
+}
diff --git a/modules/core/src/system.cpp b/modules/core/src/system.cpp
index 668b868b32..738e863d76 100644
--- a/modules/core/src/system.cpp
+++ b/modules/core/src/system.cpp
@@ -42,11 +42,20 @@
 
 #include "precomp.hpp"
 
+#ifdef _MSC_VER
+# if _MSC_VER >= 1700
+#  pragma warning(disable:4447) // Disable warning 'main' signature found without threading model
+# endif
+#endif
+
 #if defined WIN32 || defined _WIN32 || defined WINCE
 #ifndef _WIN32_WINNT           // This is needed for the declaration of TryEnterCriticalSection in winbase.h with Visual Studio 2005 (and older?)
   #define _WIN32_WINNT 0x0400  // http://msdn.microsoft.com/en-us/library/ms686857(VS.85).aspx
 #endif
 #include <windows.h>
+#if (_WIN32_WINNT >= 0x0602)
+  #include <synchapi.h>
+#endif
 #undef small
 #undef min
 #undef max
@@ -75,6 +84,30 @@
     }
   #endif
 #endif
+
+#ifdef HAVE_WINRT
+#include <wrl/client.h>
+
+std::wstring GetTempPathWinRT()
+{
+    return std::wstring(Windows::Storage::ApplicationData::Current->TemporaryFolder->Path->Data());
+}
+
+std::wstring GetTempFileNameWinRT(std::wstring prefix)
+{
+    wchar_t guidStr[40];
+    GUID g;
+    CoCreateGuid(&g);
+    wchar_t* mask = L"%08x_%04x_%04x_%02x%02x_%02x%02x%02x%02x%02x%02x";
+    swprintf(&guidStr[0], sizeof(guidStr)/sizeof(wchar_t), mask,
+             g.Data1, g.Data2, g.Data3, UINT(g.Data4[0]), UINT(g.Data4[1]),
+             UINT(g.Data4[2]), UINT(g.Data4[3]), UINT(g.Data4[4]),
+             UINT(g.Data4[5]), UINT(g.Data4[6]), UINT(g.Data4[7]));
+
+    return prefix + std::wstring(guidStr);
+}
+
+#endif
 #else
 #include <pthread.h>
 #include <sys/time.h>
@@ -113,7 +146,7 @@ namespace cv
 
 Exception::Exception() { code = 0; line = 0; }
 
-Exception::Exception(int _code, const std::string& _err, const std::string& _func, const std::string& _file, int _line)
+Exception::Exception(int _code, const String& _err, const String& _func, const String& _file, int _line)
 : code(_code), err(_err), func(_func), file(_file), line(_line)
 {
     formatMessage();
@@ -340,31 +373,70 @@ int64 getCPUTickCount(void)
 
 #endif
 
-const std::string& getBuildInformation()
+const String& getBuildInformation()
 {
-    static std::string build_info =
+    static String build_info =
 #include "version_string.inc"
     ;
     return build_info;
 }
 
-std::string format( const char* fmt, ... )
+String format( const char* fmt, ... )
 {
-    char buf[1 << 16];
-    va_list args;
-    va_start( args, fmt );
-    vsprintf( buf, fmt, args );
-    return std::string(buf);
+    char buf[1024];
+
+    va_list va;
+    va_start(va, fmt);
+    int len = vsnprintf(buf, sizeof(buf), fmt, va);
+    va_end(va);
+
+    if (len >= (int)sizeof(buf))
+    {
+        String s(len, '\0');
+        va_start(va, fmt);
+        len = vsnprintf((char*)s.c_str(), len + 1, fmt, va);
+        va_end(va);
+        return s;
+    }
+
+    return String(buf, len);
 }
 
-std::string tempfile( const char* suffix )
+String tempfile( const char* suffix )
 {
+#ifdef HAVE_WINRT
+    std::wstring temp_dir = L"";
+    const wchar_t* opencv_temp_dir = _wgetenv(L"OPENCV_TEMP_PATH");
+    if (opencv_temp_dir)
+        temp_dir = std::wstring(opencv_temp_dir);
+#else
     const char *temp_dir = getenv("OPENCV_TEMP_PATH");
-    std::string fname;
+    String fname;
+#endif
 
 #if defined WIN32 || defined _WIN32
-    char temp_dir2[MAX_PATH + 1] = { 0 };
-    char temp_file[MAX_PATH + 1] = { 0 };
+#ifdef HAVE_WINRT
+    RoInitialize(RO_INIT_MULTITHREADED);
+    std::wstring temp_dir2;
+    if (temp_dir.empty())
+        temp_dir = GetTempPathWinRT();
+
+    std::wstring temp_file;
+    temp_file = GetTempFileNameWinRT(L"ocv");
+    if (temp_file.empty())
+        return std::string();
+
+    temp_file = temp_dir + std::wstring(L"\\") + temp_file;
+    DeleteFileW(temp_file.c_str());
+
+    char aname[MAX_PATH];
+    size_t copied = wcstombs(aname, temp_file.c_str(), MAX_PATH);
+    CV_Assert((copied != MAX_PATH) && (copied != (size_t)-1));
+    fname = std::string(aname);
+    RoUninitialize();
+#else
+    char temp_dir2[MAX_PATH] = { 0 };
+    char temp_file[MAX_PATH] = { 0 };
 
     if (temp_dir == 0 || temp_dir[0] == 0)
     {
@@ -372,11 +444,12 @@ std::string tempfile( const char* suffix )
         temp_dir = temp_dir2;
     }
     if(0 == ::GetTempFileNameA(temp_dir, "ocv", 0, temp_file))
-        return std::string();
+        return String();
 
     DeleteFileA(temp_file);
 
     fname = temp_file;
+#endif
 # else
 #  ifdef ANDROID
     //char defaultTemplate[] = "/mnt/sdcard/__opencv_temp.XXXXXX";
@@ -392,12 +465,12 @@ std::string tempfile( const char* suffix )
         fname = temp_dir;
         char ech = fname[fname.size() - 1];
         if(ech != '/' && ech != '\\')
-            fname += "/";
-        fname += "__opencv_temp.XXXXXX";
+            fname = fname + "/";
+        fname = fname + "__opencv_temp.XXXXXX";
     }
 
     const int fd = mkstemp((char*)fname.c_str());
-    if (fd == -1) return std::string();
+    if (fd == -1) return String();
 
     close(fd);
     remove(fname.c_str());
@@ -453,6 +526,11 @@ void error( const Exception& exc )
     throw exc;
 }
 
+void error(int _code, const String& _err, const char* _func, const char* _file, int _line)
+{
+    error(cv::Exception(_code, _err, _func, _file, _line));
+}
+
 CvErrorCallback
 redirectError( CvErrorCallback errCallback, void* userdata, void** prevUserdata)
 {
@@ -469,40 +547,6 @@ redirectError( CvErrorCallback errCallback, void* userdata, void** prevUserdata)
 
 }
 
-/*CV_IMPL int
-cvGuiBoxReport( int code, const char *func_name, const char *err_msg,
-                const char *file, int line, void* )
-{
-#if (!defined WIN32 && !defined _WIN32) || defined WINCE
-    return cvStdErrReport( code, func_name, err_msg, file, line, 0 );
-#else
-    if( code != CV_StsBackTrace && code != CV_StsAutoTrace )
-    {
-        size_t msg_len = strlen(err_msg ? err_msg : "") + 1024;
-        char* message = (char*)alloca(msg_len);
-        char title[100];
-
-        wsprintf( message, "%s (%s)\nin function %s, %s(%d)\n\n"
-                  "Press \"Abort\" to terminate application.\n"
-                  "Press \"Retry\" to debug (if the app is running under debugger).\n"
-                  "Press \"Ignore\" to continue (this is not safe).\n",
-                  cvErrorStr(code), err_msg ? err_msg : "no description",
-                  func_name, file, line );
-
-        wsprintf( title, "OpenCV GUI Error Handler" );
-
-        int answer = MessageBox( NULL, message, title, MB_ICONERROR|MB_ABORTRETRYIGNORE|MB_SYSTEMMODAL );
-
-        if( answer == IDRETRY )
-        {
-            CV_DBG_BREAK();
-        }
-        return answer != IDIGNORE;
-    }
-    return 0;
-#endif
-}*/
-
 CV_IMPL int cvCheckHardwareSupport(int feature)
 {
     CV_DbgAssert( 0 <= feature && feature <= CV_HARDWARE_MAX_FEATURE );
@@ -659,127 +703,12 @@ cvErrorFromIppStatus( int status )
     }
 }
 
-static CvModuleInfo cxcore_info = { 0, "cxcore", CV_VERSION, 0 };
 
-CvModuleInfo* CvModule::first = 0, *CvModule::last = 0;
+#if defined CVAPI_EXPORTS && defined WIN32 && !defined WINCE
+#ifdef HAVE_WINRT
+    #pragma warning(disable:4447) // Disable warning 'main' signature found without threading model
+#endif
 
-CvModule::CvModule( CvModuleInfo* _info )
-{
-    cvRegisterModule( _info );
-    info = last;
-}
-
-CvModule::~CvModule(void)
-{
-    if( info )
-    {
-        CvModuleInfo* p = first;
-        for( ; p != 0 && p->next != info; p = p->next )
-            ;
-
-        if( p )
-            p->next = info->next;
-
-        if( first == info )
-            first = info->next;
-
-        if( last == info )
-            last = p;
-
-        free( info );
-        info = 0;
-    }
-}
-
-CV_IMPL int
-cvRegisterModule( const CvModuleInfo* module )
-{
-    CV_Assert( module != 0 && module->name != 0 && module->version != 0 );
-
-    size_t name_len = strlen(module->name);
-    size_t version_len = strlen(module->version);
-
-    CvModuleInfo* module_copy = (CvModuleInfo*)malloc( sizeof(*module_copy) +
-                                name_len + 1 + version_len + 1 );
-
-    *module_copy = *module;
-    module_copy->name = (char*)(module_copy + 1);
-    module_copy->version = (char*)(module_copy + 1) + name_len + 1;
-
-    memcpy( (void*)module_copy->name, module->name, name_len + 1 );
-    memcpy( (void*)module_copy->version, module->version, version_len + 1 );
-    module_copy->next = 0;
-
-    if( CvModule::first == 0 )
-        CvModule::first = module_copy;
-    else
-        CvModule::last->next = module_copy;
-
-    CvModule::last = module_copy;
-
-    return 0;
-}
-
-CvModule cxcore_module( &cxcore_info );
-
-CV_IMPL void
-cvGetModuleInfo( const char* name, const char **version, const char **plugin_list )
-{
-    static char joint_verinfo[1024]   = "";
-    static char plugin_list_buf[1024] = "";
-
-    if( version )
-        *version = 0;
-
-    if( plugin_list )
-        *plugin_list = 0;
-
-    CvModuleInfo* module;
-
-    if( version )
-    {
-        if( name )
-        {
-            size_t i, name_len = strlen(name);
-
-            for( module = CvModule::first; module != 0; module = module->next )
-            {
-                if( strlen(module->name) == name_len )
-                {
-                    for( i = 0; i < name_len; i++ )
-                    {
-                        int c0 = toupper(module->name[i]), c1 = toupper(name[i]);
-                        if( c0 != c1 )
-                            break;
-                    }
-                    if( i == name_len )
-                        break;
-                }
-            }
-            if( !module )
-                CV_Error( CV_StsObjectNotFound, "The module is not found" );
-
-            *version = module->version;
-        }
-        else
-        {
-            char* ptr = joint_verinfo;
-
-            for( module = CvModule::first; module != 0; module = module->next )
-            {
-                sprintf( ptr, "%s: %s%s", module->name, module->version, module->next ? ", " : "" );
-                ptr += strlen(ptr);
-            }
-
-            *version = joint_verinfo;
-        }
-    }
-
-    if( plugin_list )
-        *plugin_list = plugin_list_buf;
-}
-
-#if defined BUILD_SHARED_LIBS && defined CVAPI_EXPORTS && defined WIN32 && !defined WINCE
 BOOL WINAPI DllMain( HINSTANCE, DWORD  fdwReason, LPVOID );
 
 BOOL WINAPI DllMain( HINSTANCE, DWORD  fdwReason, LPVOID )
@@ -800,7 +729,15 @@ namespace cv
 
 struct Mutex::Impl
 {
-    Impl() { InitializeCriticalSection(&cs); refcount = 1; }
+    Impl()
+    {
+#if (_WIN32_WINNT >= 0x0600)
+        ::InitializeCriticalSectionEx(&cs, 1000, 0);
+#else
+        ::InitializeCriticalSection(&cs);
+#endif
+        refcount = 1;
+    }
     ~Impl() { DeleteCriticalSection(&cs); }
 
     void lock() { EnterCriticalSection(&cs); }
@@ -893,4 +830,4 @@ bool Mutex::trylock() { return impl->trylock(); }
 
 }
 
-/* End of file. */
\ No newline at end of file
+/* End of file. */
diff --git a/modules/core/src/types.cpp b/modules/core/src/types.cpp
new file mode 100644
index 0000000000..89e0042657
--- /dev/null
+++ b/modules/core/src/types.cpp
@@ -0,0 +1,139 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+namespace cv
+{
+
+size_t KeyPoint::hash() const
+{
+    size_t _Val = 2166136261U, scale = 16777619U;
+    Cv32suf u;
+    u.f = pt.x; _Val = (scale * _Val) ^ u.u;
+    u.f = pt.y; _Val = (scale * _Val) ^ u.u;
+    u.f = size; _Val = (scale * _Val) ^ u.u;
+    u.f = angle; _Val = (scale * _Val) ^ u.u;
+    u.f = response; _Val = (scale * _Val) ^ u.u;
+    _Val = (scale * _Val) ^ ((size_t) octave);
+    _Val = (scale * _Val) ^ ((size_t) class_id);
+    return _Val;
+}
+
+void KeyPoint::convert(const std::vector<KeyPoint>& keypoints, std::vector<Point2f>& points2f,
+                       const std::vector<int>& keypointIndexes)
+{
+    if( keypointIndexes.empty() )
+    {
+        points2f.resize( keypoints.size() );
+        for( size_t i = 0; i < keypoints.size(); i++ )
+            points2f[i] = keypoints[i].pt;
+    }
+    else
+    {
+        points2f.resize( keypointIndexes.size() );
+        for( size_t i = 0; i < keypointIndexes.size(); i++ )
+        {
+            int idx = keypointIndexes[i];
+            if( idx >= 0 )
+                points2f[i] = keypoints[idx].pt;
+            else
+            {
+                CV_Error( CV_StsBadArg, "keypointIndexes has element < 0. TODO: process this case" );
+                //points2f[i] = Point2f(-1, -1);
+            }
+        }
+    }
+}
+
+void KeyPoint::convert( const std::vector<Point2f>& points2f, std::vector<KeyPoint>& keypoints,
+                        float size, float response, int octave, int class_id )
+{
+    keypoints.resize(points2f.size());
+    for( size_t i = 0; i < points2f.size(); i++ )
+        keypoints[i] = KeyPoint(points2f[i], size, -1, response, octave, class_id);
+}
+
+float KeyPoint::overlap( const KeyPoint& kp1, const KeyPoint& kp2 )
+{
+    float a = kp1.size * 0.5f;
+    float b = kp2.size * 0.5f;
+    float a_2 = a * a;
+    float b_2 = b * b;
+
+    Point2f p1 = kp1.pt;
+    Point2f p2 = kp2.pt;
+    float c = (float)norm( p1 - p2 );
+
+    float ovrl = 0.f;
+
+    // one circle is completely encovered by the other => no intersection points!
+    if( std::min( a, b ) + c <= std::max( a, b ) )
+        return std::min( a_2, b_2 ) / std::max( a_2, b_2 );
+
+    if( c < a + b ) // circles intersect
+    {
+        float c_2 = c * c;
+        float cosAlpha = ( b_2 + c_2 - a_2 ) / ( kp2.size * c );
+        float cosBeta  = ( a_2 + c_2 - b_2 ) / ( kp1.size * c );
+        float alpha = acos( cosAlpha );
+        float beta = acos( cosBeta );
+        float sinAlpha = sin(alpha);
+        float sinBeta  = sin(beta);
+
+        float segmentAreaA = a_2 * beta;
+        float segmentAreaB = b_2 * alpha;
+
+        float triangleAreaA = a_2 * sinBeta * cosBeta;
+        float triangleAreaB = b_2 * sinAlpha * cosAlpha;
+
+        float intersectionArea = segmentAreaA + segmentAreaB - triangleAreaA - triangleAreaB;
+        float unionArea = (a_2 + b_2) * (float)CV_PI - intersectionArea;
+
+        ovrl = intersectionArea / unionArea;
+    }
+
+    return ovrl;
+}
+
+} // cv
diff --git a/modules/core/test/test_arithm.cpp b/modules/core/test/test_arithm.cpp
index ebc9eae640..7486e134a7 100644
--- a/modules/core/test/test_arithm.cpp
+++ b/modules/core/test/test_arithm.cpp
@@ -35,7 +35,7 @@ struct BaseElemWiseOp
 
     virtual int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL_BUT_8S, 1,
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1,
                                   ninputs > 1 ? ARITHM_MAX_CHANNELS : 4);
     }
 
@@ -425,7 +425,7 @@ struct CmpOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL_BUT_8S, 1, 1);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1, 1);
     }
 
     double getMaxErr(int)
@@ -455,7 +455,7 @@ struct CmpSOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL_BUT_8S, 1, 1);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1, 1);
     }
     double getMaxErr(int)
     {
@@ -478,7 +478,7 @@ struct CopyOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
     }
     double getMaxErr(int)
     {
@@ -501,7 +501,7 @@ struct SetOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
     }
     double getMaxErr(int)
     {
@@ -718,8 +718,8 @@ struct ConvertScaleOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        int srctype = cvtest::randomType(rng, DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
-        ddepth = cvtest::randomType(rng, DEPTH_MASK_ALL, 1, 1);
+        int srctype = cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL, 1, ARITHM_MAX_CHANNELS);
+        ddepth = cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL, 1, 1);
         return srctype;
     }
     double getMaxErr(int)
@@ -957,7 +957,7 @@ struct ExpOp : public BaseElemWiseOp
     ExpOp() : BaseElemWiseOp(1, FIX_ALPHA+FIX_BETA+FIX_GAMMA, 1, 1, Scalar::all(0)) {};
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_FLT, 1, ARITHM_MAX_CHANNELS);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_FLT, 1, ARITHM_MAX_CHANNELS);
     }
     void getValueRange(int depth, double& minval, double& maxval)
     {
@@ -984,7 +984,7 @@ struct LogOp : public BaseElemWiseOp
     LogOp() : BaseElemWiseOp(1, FIX_ALPHA+FIX_BETA+FIX_GAMMA, 1, 1, Scalar::all(0)) {};
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_FLT, 1, ARITHM_MAX_CHANNELS);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_FLT, 1, ARITHM_MAX_CHANNELS);
     }
     void getValueRange(int depth, double& minval, double& maxval)
     {
@@ -1070,7 +1070,7 @@ struct CartToPolarToCartOp : public BaseElemWiseOp
     }
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_FLT, 1, 1);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_FLT, 1, 1);
     }
     void op(const vector<Mat>& src, Mat& dst, const Mat&)
     {
@@ -1123,7 +1123,7 @@ struct MeanOp : public BaseElemWiseOp
     }
     double getMaxErr(int)
     {
-        return 1e-6;
+        return 1e-5;
     }
 };
 
@@ -1157,7 +1157,7 @@ struct CountNonZeroOp : public BaseElemWiseOp
     {}
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL, 1, 1);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL, 1, 1);
     }
     void op(const vector<Mat>& src, Mat& dst, const Mat& mask)
     {
@@ -1237,7 +1237,7 @@ struct NormOp : public BaseElemWiseOp
     };
     int getRandomType(RNG& rng)
     {
-        int type = cvtest::randomType(rng, DEPTH_MASK_ALL_BUT_8S, 1, 4);
+        int type = cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1, 4);
         for(;;)
         {
             normType = rng.uniform(1, 8);
@@ -1283,7 +1283,7 @@ struct MinMaxLocOp : public BaseElemWiseOp
     };
     int getRandomType(RNG& rng)
     {
-        return cvtest::randomType(rng, DEPTH_MASK_ALL_BUT_8S, 1, 1);
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1, 1);
     }
     void saveOutput(const vector<int>& minidx, const vector<int>& maxidx,
                     double minval, double maxval, Mat& dst)
@@ -1551,3 +1551,16 @@ TEST(Core_Add, AddToColumnWhen4Rows)
 
     ASSERT_EQ(0, countNonZero(m1 - m2));
 }
+
+TEST(Core_round, CvRound)
+{
+    ASSERT_EQ(2, cvRound(2.0));
+    ASSERT_EQ(2, cvRound(2.1));
+    ASSERT_EQ(-2, cvRound(-2.1));
+    ASSERT_EQ(3, cvRound(2.8));
+    ASSERT_EQ(-3, cvRound(-2.8));
+    ASSERT_EQ(2, cvRound(2.5));
+    ASSERT_EQ(4, cvRound(3.5));
+    ASSERT_EQ(-2, cvRound(-2.5));
+    ASSERT_EQ(-4, cvRound(-3.5));
+}
diff --git a/modules/core/test/test_ds.cpp b/modules/core/test/test_ds.cpp
index d79786054a..c71deed06f 100644
--- a/modules/core/test/test_ds.cpp
+++ b/modules/core/test/test_ds.cpp
@@ -358,8 +358,6 @@ Core_DynStructBaseTest::Core_DynStructBaseTest()
     iterations = max_struct_size*2;
     gen = struct_idx = iter = -1;
     test_progress = -1;
-
-    storage = 0;
 }
 
 
@@ -999,7 +997,7 @@ void Core_SeqBaseTest::run( int )
             {
                 t = cvtest::randReal(rng)*(max_log_storage_block_size - min_log_storage_block_size)
                 + min_log_storage_block_size;
-                storage = cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) );
+                storage.reset(cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) ));
             }
 
             iter = struct_idx = -1;
@@ -1083,11 +1081,11 @@ void Core_SeqSortInvTest::run( int )
         {
             struct_idx = iter = -1;
 
-            if( storage.empty() )
+            if( !storage )
             {
                 t = cvtest::randReal(rng)*(max_log_storage_block_size - min_log_storage_block_size)
                 + min_log_storage_block_size;
-                storage = cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) );
+                storage.reset(cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) ));
             }
 
             for( iter = 0; iter < iterations/10; iter++ )
@@ -1384,7 +1382,7 @@ void Core_SetTest::run( int )
         {
             struct_idx = iter = -1;
             t = cvtest::randReal(rng)*(max_log_storage_block_size - min_log_storage_block_size) + min_log_storage_block_size;
-            storage = cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) );
+            storage.reset(cvCreateMemStorage( cvRound( exp(t * CV_LOG2) ) ));
 
             for( int i = 0; i < struct_count; i++ )
             {
@@ -1398,7 +1396,7 @@ void Core_SetTest::run( int )
 
                 cvTsReleaseSimpleSet( (CvTsSimpleSet**)&simple_struct[i] );
                 simple_struct[i] = cvTsCreateSimpleSet( max_struct_size, pure_elem_size );
-                 cxcore_struct[i] = cvCreateSet( 0, sizeof(CvSet), elem_size, storage );
+                cxcore_struct[i] = cvCreateSet( 0, sizeof(CvSet), elem_size, storage );
             }
 
             if( test_set_ops( iterations*100 ) < 0 )
@@ -1811,7 +1809,7 @@ void Core_GraphTest::run( int )
             int block_size = cvRound( exp(t * CV_LOG2) );
             block_size = MAX(block_size, (int)(sizeof(CvGraph) + sizeof(CvMemBlock) + sizeof(CvSeqBlock)));
 
-            storage = cvCreateMemStorage(block_size);
+            storage.reset(cvCreateMemStorage(block_size));
 
             for( i = 0; i < struct_count; i++ )
             {
@@ -1929,7 +1927,7 @@ void Core_GraphScanTest::run( int )
             storage_blocksize = MAX(storage_blocksize, (int)(sizeof(CvGraph) + sizeof(CvMemBlock) + sizeof(CvSeqBlock)));
             storage_blocksize = MAX(storage_blocksize, (int)(sizeof(CvGraphEdge) + sizeof(CvMemBlock) + sizeof(CvSeqBlock)));
             storage_blocksize = MAX(storage_blocksize, (int)(sizeof(CvGraphVtx) + sizeof(CvMemBlock) + sizeof(CvSeqBlock)));
-            storage = cvCreateMemStorage(storage_blocksize);
+            storage.reset(cvCreateMemStorage(storage_blocksize));
 
             if( gen == 0 )
             {
@@ -2118,5 +2116,3 @@ TEST(Core_DS_Seq, sort_invert) { Core_SeqSortInvTest test; test.safe_run(); }
 TEST(Core_DS_Set, basic_operations) { Core_SetTest test; test.safe_run(); }
 TEST(Core_DS_Graph, basic_operations) { Core_GraphTest test; test.safe_run(); }
 TEST(Core_DS_Graph, scan) { Core_GraphScanTest test; test.safe_run(); }
-
-
diff --git a/modules/core/test/test_dxt.cpp b/modules/core/test/test_dxt.cpp
index 16025fa8f2..1c0c7b00bf 100644
--- a/modules/core/test/test_dxt.cpp
+++ b/modules/core/test/test_dxt.cpp
@@ -866,5 +866,3 @@ protected:
 };
 
 TEST(Core_DFT, complex_output) { Core_DFTComplexOutputTest test; test.safe_run(); }
-
-
diff --git a/modules/core/test/test_eigen.cpp b/modules/core/test/test_eigen.cpp
index 046573d26d..21859e59b8 100644
--- a/modules/core/test/test_eigen.cpp
+++ b/modules/core/test/test_eigen.cpp
@@ -294,7 +294,7 @@ bool Core_EigenTest::test_pairs(const cv::Mat& src)
 
     cv::Mat eigen_values, eigen_vectors;
 
-    cv::eigen(src, true, eigen_values, eigen_vectors);
+    cv::eigen(src, eigen_values, eigen_vectors);
 
     if (!check_pair_count(src, eigen_values, eigen_vectors)) return false;
 
@@ -362,8 +362,8 @@ bool Core_EigenTest::test_values(const cv::Mat& src)
 
     if (!test_pairs(src)) return false;
 
-    cv::eigen(src, true, eigen_values_1, eigen_vectors);
-    cv::eigen(src, false, eigen_values_2, eigen_vectors);
+    cv::eigen(src, eigen_values_1, eigen_vectors);
+    cv::eigen(src, eigen_values_2);
 
     if (!check_pair_count(src, eigen_values_2)) return false;
 
diff --git a/modules/core/test/test_io.cpp b/modules/core/test/test_io.cpp
index 7d4e7d9398..ba66567615 100644
--- a/modules/core/test/test_io.cpp
+++ b/modules/core/test/test_io.cpp
@@ -191,7 +191,7 @@ protected:
 
             int real_int = (int)fs["test_int"];
             double real_real = (double)fs["test_real"];
-            string real_string = (string)fs["test_string"];
+            String real_string = (String)fs["test_string"];
 
             if( real_int != test_int ||
                fabs(real_real - test_real) > DBL_EPSILON*(fabs(test_real)+1) ||
@@ -211,7 +211,7 @@ protected:
             vector<int> pt;
 
             if( !m || !CV_IS_MAT(m) || m->rows != test_mat.rows || m->cols != test_mat.cols ||
-               cvtest::cmpEps( Mat(&stub1), Mat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
+               cvtest::cmpEps( cv::cvarrToMat(&stub1), cv::cvarrToMat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
             {
                 ts->printf( cvtest::TS::LOG, "the read matrix is not correct: (%.20g vs %.20g) at (%d,%d)\n",
                             cvGetReal2D(&stub1, pt[0], pt[1]), cvGetReal2D(&_test_stub1, pt[0], pt[1]),
@@ -241,7 +241,7 @@ protected:
             if( !CV_ARE_TYPES_EQ(&stub, &_test_stub) ||
                !CV_ARE_SIZES_EQ(&stub, &_test_stub) ||
                //cvNorm(&stub, &_test_stub, CV_L2) != 0 )
-               cvtest::cmpEps( Mat(&stub1), Mat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
+               cvtest::cmpEps( cv::cvarrToMat(&stub1), cv::cvarrToMat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
             {
                 ts->printf( cvtest::TS::LOG, "readObj method: the read nd matrix is not correct: (%.20g vs %.20g) vs at (%d,%d)\n",
                            cvGetReal2D(&stub1, pt[0], pt[1]), cvGetReal2D(&_test_stub1, pt[0], pt[1]),
@@ -259,7 +259,7 @@ protected:
             if( !CV_ARE_TYPES_EQ(&stub, &_test_stub) ||
                !CV_ARE_SIZES_EQ(&stub, &_test_stub) ||
                //cvNorm(&stub, &_test_stub, CV_L2) != 0 )
-               cvtest::cmpEps( Mat(&stub1), Mat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
+               cvtest::cmpEps( cv::cvarrToMat(&stub1), cv::cvarrToMat(&_test_stub1), &max_diff, 0, &pt, true) < 0 )
             {
                 ts->printf( cvtest::TS::LOG, "C++ method: the read nd matrix is not correct: (%.20g vs %.20g) vs at (%d,%d)\n",
                            cvGetReal2D(&stub1, pt[0], pt[1]), cvGetReal2D(&_test_stub1, pt[1], pt[0]),
@@ -270,16 +270,16 @@ protected:
 
             cvRelease((void**)&m_nd);
 
-            Ptr<CvSparseMat> m_s = (CvSparseMat*)fs["test_sparse_mat"].readObj();
-            Ptr<CvSparseMat> _test_sparse_ = (CvSparseMat*)test_sparse_mat;
-            Ptr<CvSparseMat> _test_sparse = (CvSparseMat*)cvClone(_test_sparse_);
+            Ptr<CvSparseMat> m_s((CvSparseMat*)fs["test_sparse_mat"].readObj());
+            Ptr<CvSparseMat> _test_sparse_(cvCreateSparseMat(test_sparse_mat));
+            Ptr<CvSparseMat> _test_sparse((CvSparseMat*)cvClone(_test_sparse_));
             SparseMat m_s2;
             fs["test_sparse_mat"] >> m_s2;
-            Ptr<CvSparseMat> _m_s2 = (CvSparseMat*)m_s2;
+            Ptr<CvSparseMat> _m_s2(cvCreateSparseMat(m_s2));
 
             if( !m_s || !CV_IS_SPARSE_MAT(m_s) ||
-               !cvTsCheckSparse(m_s, _test_sparse,0) ||
-               !cvTsCheckSparse(_m_s2, _test_sparse,0))
+               !cvTsCheckSparse(m_s, _test_sparse, 0) ||
+               !cvTsCheckSparse(_m_s2, _test_sparse, 0))
             {
                 ts->printf( cvtest::TS::LOG, "the read sparse matrix is not correct\n" );
                 ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_OUTPUT );
@@ -292,7 +292,7 @@ protected:
                (int)tl[1] != 2 ||
                fabs((double)tl[2] - CV_PI) >= DBL_EPSILON ||
                (int)tl[3] != -3435345 ||
-               (string)tl[4] != "2-502 2-029 3egegeg" ||
+               (String)tl[4] != "2-502 2-029 3egegeg" ||
                tl[5].type() != FileNode::MAP || tl[5].size() != 3 ||
                (int)tl[5]["month"] != 12 ||
                (int)tl[5]["day"] != 31 ||
@@ -378,6 +378,7 @@ protected:
 
 TEST(Core_InputOutput, write_read_consistency) { Core_IOTest test; test.safe_run(); }
 
+extern void testFormatter();
 
 class CV_MiscIOTest : public cvtest::BaseTest
 {
@@ -390,7 +391,6 @@ protected:
         try
         {
             string fname = cv::tempfile(".xml");
-            FileStorage fs(fname, FileStorage::WRITE);
             vector<int> mi, mi2, mi3, mi4;
             vector<Mat> mv, mv2, mv3, mv4;
             Mat m(10, 9, CV_32F);
@@ -398,24 +398,59 @@ protected:
             randu(m, 0, 1);
             mi3.push_back(5);
             mv3.push_back(m);
+            Point_<float> p1(1.1f, 2.2f), op1;
+            Point3i p2(3, 4, 5), op2;
+            Size s1(6, 7), os1;
+            Complex<int> c1(9, 10), oc1;
+            Rect r1(11, 12, 13, 14), or1;
+            Vec<int, 5> v1(15, 16, 17, 18, 19), ov1;
+            Scalar sc1(20.0, 21.1, 22.2, 23.3), osc1;
+            Range g1(7, 8), og1;
+
+            FileStorage fs(fname, FileStorage::WRITE);
             fs << "mi" << mi;
             fs << "mv" << mv;
             fs << "mi3" << mi3;
             fs << "mv3" << mv3;
             fs << "empty" << empty;
+            fs << "p1" << p1;
+            fs << "p2" << p2;
+            fs << "s1" << s1;
+            fs << "c1" << c1;
+            fs << "r1" << r1;
+            fs << "v1" << v1;
+            fs << "sc1" << sc1;
+            fs << "g1" << g1;
             fs.release();
+
             fs.open(fname, FileStorage::READ);
             fs["mi"] >> mi2;
             fs["mv"] >> mv2;
             fs["mi3"] >> mi4;
             fs["mv3"] >> mv4;
             fs["empty"] >> empty;
+            fs["p1"] >> op1;
+            fs["p2"] >> op2;
+            fs["s1"] >> os1;
+            fs["c1"] >> oc1;
+            fs["r1"] >> or1;
+            fs["v1"] >> ov1;
+            fs["sc1"] >> osc1;
+            fs["g1"] >> og1;
             CV_Assert( mi2.empty() );
             CV_Assert( mv2.empty() );
             CV_Assert( norm(mi3, mi4, CV_C) == 0 );
             CV_Assert( mv4.size() == 1 );
             double n = norm(mv3[0], mv4[0], CV_C);
             CV_Assert( n == 0 );
+            CV_Assert( op1 == p1 );
+            CV_Assert( op2 == p2 );
+            CV_Assert( os1 == s1 );
+            CV_Assert( oc1 == c1 );
+            CV_Assert( or1 == r1 );
+            CV_Assert( ov1 == v1 );
+            CV_Assert( osc1 == sc1 );
+            CV_Assert( og1 == g1 );
         }
         catch(...)
         {
@@ -454,12 +489,12 @@ protected:
 TEST(Core_InputOutput, huge) { CV_BigMatrixIOTest test; test.safe_run(); }
 */
 
-TEST(Core_globbing, accurasy)
+TEST(Core_globbing, accuracy)
 {
     std::string patternLena    = cvtest::TS::ptr()->get_data_path() + "lena*.*";
     std::string patternLenaPng = cvtest::TS::ptr()->get_data_path() + "lena.png";
 
-    std::vector<std::string> lenas, pngLenas;
+    std::vector<String> lenas, pngLenas;
     cv::glob(patternLena, lenas, true);
     cv::glob(patternLenaPng, pngLenas, true);
 
@@ -469,4 +504,14 @@ TEST(Core_globbing, accurasy)
     {
         ASSERT_NE(std::find(lenas.begin(), lenas.end(), pngLenas[i]), lenas.end());
     }
-}
\ No newline at end of file
+}
+
+TEST(Core_InputOutput, FileStorage)
+{
+    std::string file = cv::tempfile(".xml");
+    cv::FileStorage f(file, cv::FileStorage::WRITE);
+
+    char arr[66];
+    sprintf(arr, "sprintf is hell %d", 666);
+    EXPECT_NO_THROW(f << arr);
+}
diff --git a/modules/core/test/test_main.cpp b/modules/core/test/test_main.cpp
index 6b24993447..d5400e20fd 100644
--- a/modules/core/test/test_main.cpp
+++ b/modules/core/test/test_main.cpp
@@ -1,3 +1,10 @@
+#ifdef _MSC_VER
+# if _MSC_VER >= 1700
+#  pragma warning(disable:4447) // Disable warning 'main' signature found without threading model
+# endif
+#endif
+
+
 #include "test_precomp.hpp"
 
 CV_TEST_MAIN("cv")
diff --git a/modules/core/test/test_mat.cpp b/modules/core/test/test_mat.cpp
index 3693c2dcb1..3c8ae8bf9f 100644
--- a/modules/core/test/test_mat.cpp
+++ b/modules/core/test/test_mat.cpp
@@ -1,5 +1,7 @@
 #include "test_precomp.hpp"
 
+#include <map>
+
 using namespace cv;
 using namespace std;
 
@@ -508,6 +510,32 @@ protected:
             return;
         }
     #endif
+        // Test read and write
+        FileStorage fs( "PCA_store.yml", FileStorage::WRITE );
+        rPCA.write( fs );
+        fs.release();
+
+        PCA lPCA;
+        fs.open( "PCA_store.yml", FileStorage::READ );
+        lPCA.read( fs.root() );
+        err = norm( rPCA.eigenvectors, lPCA.eigenvectors, CV_RELATIVE_L2 );
+        if( err > 0 )
+        {
+            ts->printf( cvtest::TS::LOG, "bad accuracy of write/load functions (YML); err = %f\n", err );
+            ts->set_failed_test_info( cvtest::TS::FAIL_BAD_ACCURACY );
+        }
+        err = norm( rPCA.eigenvalues, lPCA.eigenvalues, CV_RELATIVE_L2 );
+        if( err > 0 )
+        {
+            ts->printf( cvtest::TS::LOG, "bad accuracy of write/load functions (YML); err = %f\n", err );
+            ts->set_failed_test_info( cvtest::TS::FAIL_BAD_ACCURACY );
+        }
+        err = norm( rPCA.mean, lPCA.mean, CV_RELATIVE_L2 );
+        if( err > 0 )
+        {
+            ts->printf( cvtest::TS::LOG, "bad accuracy of write/load functions (YML); err = %f\n", err );
+            ts->set_failed_test_info( cvtest::TS::FAIL_BAD_ACCURACY );
+        }
     }
 };
 
@@ -641,7 +669,7 @@ void Core_ArrayOpTest::run( int /* start_from */)
         cvSetReal3D(&matA, idx1[0], idx1[1], idx1[2], -val0);
         cvSetND(&matB, idx0, val1);
         cvSet3D(&matB, idx1[0], idx1[1], idx1[2], -val1);
-        Ptr<CvMatND> matC = cvCloneMatND(&matB);
+        Ptr<CvMatND> matC(cvCloneMatND(&matB));
 
         if( A.at<float>(idx0[0], idx0[1], idx0[2]) != val0 ||
            A.at<float>(idx1[0], idx1[1], idx1[2]) != -val0 ||
@@ -734,7 +762,7 @@ void Core_ArrayOpTest::run( int /* start_from */)
             }
         }
 
-        Ptr<CvSparseMat> M2 = (CvSparseMat*)M;
+        Ptr<CvSparseMat> M2(cvCreateSparseMat(M));
         MatND Md;
         M.copyTo(Md);
         SparseMat M3; SparseMat(Md).convertTo(M3, Md.type(), 2);
diff --git a/modules/core/test/test_math.cpp b/modules/core/test/test_math.cpp
index e02f78c541..377742386f 100644
--- a/modules/core/test/test_math.cpp
+++ b/modules/core/test/test_math.cpp
@@ -545,7 +545,7 @@ void Core_CrossProductTest::run_func()
 
 void Core_CrossProductTest::prepare_to_validation( int )
 {
-    CvScalar a = {{0,0,0,0}}, b = {{0,0,0,0}}, c = {{0,0,0,0}};
+    CvScalar a(0), b(0), c(0);
 
     if( test_mat[INPUT][0].rows > 1 )
     {
@@ -2292,9 +2292,9 @@ void Core_SolvePolyTest::run( int )
                 cvFlip(&amat, &amat, 0);
                 int nr2;
                 if( cubic_case == 0 )
-                    nr2 = cv::solveCubic(cv::Mat(&amat),umat2);
+                    nr2 = cv::solveCubic(cv::cvarrToMat(&amat),umat2);
                 else
-                    nr2 = cv::solveCubic(cv::Mat_<float>(cv::Mat(&amat)), umat2);
+                    nr2 = cv::solveCubic(cv::Mat_<float>(cv::cvarrToMat(&amat)), umat2);
                 cvFlip(&amat, &amat, 0);
                 if(nr2 > 0)
                     std::sort(ar2.begin(), ar2.begin()+nr2, pred_double());
@@ -2599,6 +2599,35 @@ TEST(Core_Trace, accuracy) { Core_TraceTest test; test.safe_run(); }
 TEST(Core_SolvePoly, accuracy) { Core_SolvePolyTest test; test.safe_run(); }
 TEST(Core_Phase, accuracy) { Core_PhaseTest test; test.safe_run(); }
 
+
+TEST(Core_SVD, flt)
+{
+    float a[] = {
+    1.23377746e+011f, -7.05490125e+010f, -4.18380882e+010f, -11693456.f,
+    -39091328.f, 77492224.f, -7.05490125e+010f, 2.36211143e+011f,
+    -3.51093473e+010f, 70773408.f, -4.83386156e+005f, -129560368.f,
+    -4.18380882e+010f, -3.51093473e+010f, 9.25311222e+010f, -49052424.f,
+    43922752.f, 12176842.f, -11693456.f, 70773408.f, -49052424.f, 8.40836094e+004f,
+    5.17475293e+003f, -1.16122949e+004f, -39091328.f, -4.83386156e+005f,
+    43922752.f, 5.17475293e+003f, 5.16047969e+004f, 5.68887842e+003f, 77492224.f,
+    -129560368.f, 12176842.f, -1.16122949e+004f, 5.68887842e+003f,
+    1.28060578e+005f
+    };
+
+    float b[] = {
+    283751232.f, 2.61604198e+009f, -745033216.f, 2.31125625e+005f,
+    -4.52429188e+005f, -1.37596525e+006f
+    };
+
+    Mat A(6, 6, CV_32F, a);
+    Mat B(6, 1, CV_32F, b);
+    Mat X, B1;
+    solve(A, B, X, DECOMP_SVD);
+    B1 = A*X;
+    EXPECT_LE(norm(B1, B, NORM_L2 + NORM_RELATIVE), FLT_EPSILON*10);
+}
+
+
 // TODO: eigenvv, invsqrt, cbrt, fastarctan, (round, floor, ceil(?)),
 
 
@@ -2726,4 +2755,3 @@ TEST(CovariationMatrixVectorOfMatWithMean, accuracy)
 }
 
 /* End of file. */
-
diff --git a/modules/core/test/test_operations.cpp b/modules/core/test/test_operations.cpp
index 6b36883cfe..8d3341e59d 100644
--- a/modules/core/test/test_operations.cpp
+++ b/modules/core/test/test_operations.cpp
@@ -75,6 +75,7 @@ protected:
     bool TestSparseMat();
     bool TestVec();
     bool TestMatxMultiplication();
+    bool TestMatxElementwiseDivison();
     bool TestSubMatAccess();
     bool TestExp();
     bool TestSVD();
@@ -891,6 +892,28 @@ bool CV_OperationsTest::TestMatxMultiplication()
     return true;
 }
 
+bool CV_OperationsTest::TestMatxElementwiseDivison()
+{
+    try
+    {
+        Matx22f mat(2, 4, 6, 8);
+        Matx22f mat2(2, 2, 2, 2);
+
+        Matx22f res = mat.div(mat2);
+
+        if(res(0, 0) != 1.0) throw test_excep();
+        if(res(0, 1) != 2.0) throw test_excep();
+        if(res(1, 0) != 3.0) throw test_excep();
+        if(res(1, 1) != 4.0) throw test_excep();
+    }
+    catch(const test_excep&)
+    {
+        ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_OUTPUT);
+        return false;
+    }
+    return true;
+}
+
 
 bool CV_OperationsTest::TestVec()
 {
@@ -1109,6 +1132,9 @@ void CV_OperationsTest::run( int /* start_from */)
     if (!TestMatxMultiplication())
         return;
 
+    if (!TestMatxElementwiseDivison())
+        return;
+
     if (!TestSubMatAccess())
         return;
 
diff --git a/modules/core/test/test_precomp.cpp b/modules/core/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/core/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/core/test/test_precomp.hpp b/modules/core/test/test_precomp.hpp
index 6c0f493d4e..d981cea066 100644
--- a/modules/core/test/test_precomp.hpp
+++ b/modules/core/test/test_precomp.hpp
@@ -9,8 +9,10 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/core/core_c.h"
-#include <iostream>
+
+#include "opencv2/core/private.hpp"
 
 #endif
diff --git a/modules/core/test/test_ptr.cpp b/modules/core/test/test_ptr.cpp
new file mode 100644
index 0000000000..c6f793ab16
--- /dev/null
+++ b/modules/core/test/test_ptr.cpp
@@ -0,0 +1,389 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the copyright holders or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+
+namespace {
+
+struct Reporter {
+    Reporter(bool* deleted) : deleted_(deleted)
+    { *deleted_ = false; }
+
+    // the destructor is virtual, so that we can test dynamic_cast later
+    virtual ~Reporter()
+    { *deleted_ = true; }
+
+private:
+    bool* deleted_;
+
+    Reporter(const Reporter&);
+    Reporter& operator = (const Reporter&);
+};
+
+struct ReportingDeleter {
+    ReportingDeleter(bool* deleted) : deleted_(deleted)
+    { *deleted_ = false; }
+
+    void operator()(void*)
+    { *deleted_ = true; }
+
+private:
+    bool* deleted_;
+};
+
+int dummyObject;
+
+}
+
+TEST(Core_Ptr, default_ctor)
+{
+    Ptr<int> p;
+    EXPECT_EQ(NULL, p.get());
+}
+
+TEST(Core_Ptr, owning_ctor)
+{
+    bool deleted = false;
+
+    {
+        Reporter* r = new Reporter(&deleted);
+        Ptr<void> p(r);
+        EXPECT_EQ(r, p.get());
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        Ptr<int> p(&dummyObject, ReportingDeleter(&deleted));
+        EXPECT_EQ(&dummyObject, p.get());
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        Ptr<void> p((void*)0, ReportingDeleter(&deleted));
+        EXPECT_EQ(NULL, p.get());
+    }
+
+    EXPECT_FALSE(deleted);
+}
+
+TEST(Core_Ptr, sharing_ctor)
+{
+    bool deleted = false;
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted));
+        Ptr<Reporter> p2(p1);
+        EXPECT_EQ(p1.get(), p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted));
+        Ptr<void> p2(p1);
+        EXPECT_EQ(p1.get(), p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted));
+        Ptr<int> p2(p1, &dummyObject);
+        EXPECT_EQ(&dummyObject, p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+}
+
+TEST(Core_Ptr, assignment)
+{
+    bool deleted1 = false, deleted2 = false;
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted1));
+        p1 = p1;
+        EXPECT_FALSE(deleted1);
+    }
+
+    EXPECT_TRUE(deleted1);
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted1));
+        Ptr<Reporter> p2(new Reporter(&deleted2));
+        p2 = p1;
+        EXPECT_TRUE(deleted2);
+        EXPECT_EQ(p1.get(), p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted1);
+    }
+
+    EXPECT_TRUE(deleted1);
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted1));
+        Ptr<void> p2(new Reporter(&deleted2));
+        p2 = p1;
+        EXPECT_TRUE(deleted2);
+        EXPECT_EQ(p1.get(), p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted1);
+    }
+
+    EXPECT_TRUE(deleted1);
+}
+
+TEST(Core_Ptr, release)
+{
+    bool deleted = false;
+
+    Ptr<Reporter> p1(new Reporter(&deleted));
+    p1.release();
+    EXPECT_TRUE(deleted);
+    EXPECT_EQ(NULL, p1.get());
+}
+
+TEST(Core_Ptr, reset)
+{
+    bool deleted_old = false, deleted_new = false;
+
+    {
+        Ptr<void> p(new Reporter(&deleted_old));
+        Reporter* r = new Reporter(&deleted_new);
+        p.reset(r);
+        EXPECT_TRUE(deleted_old);
+        EXPECT_EQ(r, p.get());
+    }
+
+    EXPECT_TRUE(deleted_new);
+
+    {
+        Ptr<void> p(new Reporter(&deleted_old));
+        p.reset(&dummyObject, ReportingDeleter(&deleted_new));
+        EXPECT_TRUE(deleted_old);
+        EXPECT_EQ(&dummyObject, p.get());
+    }
+
+    EXPECT_TRUE(deleted_new);
+}
+
+TEST(Core_Ptr, swap)
+{
+    bool deleted1 = false, deleted2 = false;
+
+    {
+        Reporter* r1 = new Reporter(&deleted1);
+        Reporter* r2 = new Reporter(&deleted2);
+        Ptr<Reporter> p1(r1), p2(r2);
+        p1.swap(p2);
+        EXPECT_EQ(r1, p2.get());
+        EXPECT_EQ(r2, p1.get());
+        EXPECT_FALSE(deleted1);
+        EXPECT_FALSE(deleted2);
+        p1.release();
+        EXPECT_TRUE(deleted2);
+    }
+
+    EXPECT_TRUE(deleted1);
+
+    {
+        Reporter* r1 = new Reporter(&deleted1);
+        Reporter* r2 = new Reporter(&deleted2);
+        Ptr<Reporter> p1(r1), p2(r2);
+        swap(p1, p2);
+        EXPECT_EQ(r1, p2.get());
+        EXPECT_EQ(r2, p1.get());
+        EXPECT_FALSE(deleted1);
+        EXPECT_FALSE(deleted2);
+        p1.release();
+        EXPECT_TRUE(deleted2);
+    }
+
+    EXPECT_TRUE(deleted1);
+}
+
+TEST(Core_Ptr, accessors)
+{
+    {
+        Ptr<int> p;
+        EXPECT_EQ(NULL, static_cast<int*>(p));
+        EXPECT_TRUE(p.empty());
+    }
+
+    {
+        Size* s = new Size();
+        Ptr<Size> p(s);
+        EXPECT_EQ(s, static_cast<Size*>(p));
+        EXPECT_EQ(s, &*p);
+        EXPECT_EQ(&s->width, &p->width);
+        EXPECT_FALSE(p.empty());
+    }
+}
+
+namespace {
+
+struct SubReporterBase {
+    virtual ~SubReporterBase() {}
+    int padding;
+};
+
+/* multiple inheritance, so that casts do something interesting */
+struct SubReporter : SubReporterBase, Reporter
+{
+    SubReporter(bool* deleted) : Reporter(deleted)
+    {}
+};
+
+}
+
+TEST(Core_Ptr, casts)
+{
+    bool deleted = false;
+
+    {
+        Ptr<const Reporter> p1(new Reporter(&deleted));
+        Ptr<Reporter> p2 = p1.constCast<Reporter>();
+        EXPECT_EQ(p1.get(), p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        SubReporter* sr = new SubReporter(&deleted);
+        Ptr<Reporter> p1(sr);
+        // This next check isn't really for Ptr itself; it checks that Reporter
+        // is at a non-zero offset within SubReporter, so that the next
+        // check will give us more confidence that the cast actually did something.
+        EXPECT_NE(static_cast<void*>(sr), static_cast<void*>(p1.get()));
+        Ptr<SubReporter> p2 = p1.staticCast<SubReporter>();
+        EXPECT_EQ(sr, p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        SubReporter* sr = new SubReporter(&deleted);
+        Ptr<Reporter> p1(sr);
+        EXPECT_NE(static_cast<void*>(sr), static_cast<void*>(p1.get()));
+        Ptr<void> p2 = p1.dynamicCast<void>();
+        EXPECT_EQ(sr, p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+
+    {
+        Ptr<Reporter> p1(new Reporter(&deleted));
+        Ptr<SubReporter> p2 = p1.dynamicCast<SubReporter>();
+        EXPECT_EQ(NULL, p2.get());
+        p1.release();
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+}
+
+TEST(Core_Ptr, comparisons)
+{
+    Ptr<int> p1, p2(new int), p3(new int);
+    Ptr<int> p4(p2, p3.get());
+
+    // Not using EXPECT_EQ here, since none of them are really "expected" or "actual".
+    EXPECT_TRUE(p1 == p1);
+    EXPECT_TRUE(p2 == p2);
+    EXPECT_TRUE(p2 != p3);
+    EXPECT_TRUE(p2 != p4);
+    EXPECT_TRUE(p3 == p4);
+}
+
+TEST(Core_Ptr, make)
+{
+    bool deleted = true;
+
+    {
+        Ptr<void> p = makePtr<Reporter>(&deleted);
+        EXPECT_FALSE(deleted);
+    }
+
+    EXPECT_TRUE(deleted);
+}
+
+namespace {
+
+struct SpeciallyDeletable
+{
+    SpeciallyDeletable() : deleted(false)
+    {}
+    bool deleted;
+};
+
+}
+
+namespace cv {
+
+template<>
+void DefaultDeleter<SpeciallyDeletable>::operator()(SpeciallyDeletable * obj) const
+{ obj->deleted = true; }
+
+}
+
+TEST(Core_Ptr, specialized_deleter)
+{
+    SpeciallyDeletable sd;
+
+    { Ptr<void> p(&sd); }
+
+    ASSERT_TRUE(sd.deleted);
+}
diff --git a/modules/cudev/CMakeLists.txt b/modules/cudev/CMakeLists.txt
new file mode 100644
index 0000000000..3c7b0592c7
--- /dev/null
+++ b/modules/cudev/CMakeLists.txt
@@ -0,0 +1,22 @@
+if(NOT HAVE_CUDA)
+  ocv_module_disable(cudev)
+endif()
+
+set(the_description "CUDA device layer")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4189 /wd4505 -Wundef -Wmissing-declarations -Wunused-function -Wunused-variable)
+
+ocv_add_module(cudev)
+
+ocv_module_include_directories(opencv_core)
+
+file(GLOB_RECURSE lib_hdrs "include/opencv2/*.hpp")
+file(GLOB         lib_srcs "src/*.cpp")
+
+ocv_set_module_sources(HEADERS ${lib_hdrs} SOURCES ${lib_srcs})
+
+ocv_create_module()
+
+if(BUILD_TESTS)
+  add_subdirectory(test)
+endif()
diff --git a/modules/cudev/include/opencv2/cudev.hpp b/modules/cudev/include/opencv2/cudev.hpp
new file mode 100644
index 0000000000..63bfc95696
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev.hpp
@@ -0,0 +1,112 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_HPP__
+#define __OPENCV_CUDEV_HPP__
+
+#include "cudev/common.hpp"
+
+#include "cudev/util/atomic.hpp"
+#include "cudev/util/limits.hpp"
+#include "cudev/util/saturate_cast.hpp"
+#include "cudev/util/simd_functions.hpp"
+#include "cudev/util/tuple.hpp"
+#include "cudev/util/type_traits.hpp"
+#include "cudev/util/vec_math.hpp"
+#include "cudev/util/vec_traits.hpp"
+
+#include "cudev/functional/color_cvt.hpp"
+#include "cudev/functional/functional.hpp"
+#include "cudev/functional/tuple_adapter.hpp"
+
+#include "cudev/warp/reduce.hpp"
+#include "cudev/warp/scan.hpp"
+#include "cudev/warp/shuffle.hpp"
+#include "cudev/warp/warp.hpp"
+
+#include "cudev/block/block.hpp"
+#include "cudev/block/dynamic_smem.hpp"
+#include "cudev/block/reduce.hpp"
+#include "cudev/block/scan.hpp"
+#include "cudev/block/vec_distance.hpp"
+
+#include "cudev/grid/copy.hpp"
+#include "cudev/grid/glob_reduce.hpp"
+#include "cudev/grid/histogram.hpp"
+#include "cudev/grid/integral.hpp"
+#include "cudev/grid/pyramids.hpp"
+#include "cudev/grid/reduce_to_vec.hpp"
+#include "cudev/grid/split_merge.hpp"
+#include "cudev/grid/transform.hpp"
+#include "cudev/grid/transpose.hpp"
+
+#include "cudev/ptr2d/constant.hpp"
+#include "cudev/ptr2d/deriv.hpp"
+#include "cudev/ptr2d/extrapolation.hpp"
+#include "cudev/ptr2d/glob.hpp"
+#include "cudev/ptr2d/gpumat.hpp"
+#include "cudev/ptr2d/interpolation.hpp"
+#include "cudev/ptr2d/lut.hpp"
+#include "cudev/ptr2d/mask.hpp"
+#include "cudev/ptr2d/remap.hpp"
+#include "cudev/ptr2d/resize.hpp"
+#include "cudev/ptr2d/texture.hpp"
+#include "cudev/ptr2d/traits.hpp"
+#include "cudev/ptr2d/transform.hpp"
+#include "cudev/ptr2d/warping.hpp"
+#include "cudev/ptr2d/zip.hpp"
+
+#include "cudev/expr/binary_func.hpp"
+#include "cudev/expr/binary_op.hpp"
+#include "cudev/expr/color.hpp"
+#include "cudev/expr/deriv.hpp"
+#include "cudev/expr/expr.hpp"
+#include "cudev/expr/per_element_func.hpp"
+#include "cudev/expr/reduction.hpp"
+#include "cudev/expr/unary_func.hpp"
+#include "cudev/expr/unary_op.hpp"
+#include "cudev/expr/warping.hpp"
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/block.hpp b/modules/cudev/include/opencv2/cudev/block/block.hpp
new file mode 100644
index 0000000000..385e1713e5
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/block.hpp
@@ -0,0 +1,127 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_BLOCK_HPP__
+#define __OPENCV_CUDEV_BLOCK_BLOCK_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+struct Block
+{
+    __device__ __forceinline__ static uint blockId()
+    {
+        return (blockIdx.z * gridDim.y + blockIdx.y) * gridDim.x + blockIdx.x;
+    }
+
+    __device__ __forceinline__ static uint blockSize()
+    {
+        return blockDim.x * blockDim.y * blockDim.z;
+    }
+
+    __device__ __forceinline__ static uint threadLineId()
+    {
+        return (threadIdx.z * blockDim.y + threadIdx.y) * blockDim.x + threadIdx.x;
+    }
+};
+
+template <class It, typename T>
+__device__ __forceinline__ static void blockFill(It beg, It end, const T& value)
+{
+    uint STRIDE = Block::blockSize();
+    It t = beg + Block::threadLineId();
+
+    for(; t < end; t += STRIDE)
+        *t = value;
+}
+
+template <class OutIt, typename T>
+__device__ __forceinline__ static void blockYota(OutIt beg, OutIt end, T value)
+{
+    uint STRIDE = Block::blockSize();
+    uint tid = Block::threadLineId();
+    value += tid;
+
+    for(OutIt t = beg + tid; t < end; t += STRIDE, value += STRIDE)
+        *t = value;
+}
+
+template <class InIt, class OutIt>
+__device__ __forceinline__ static void blockCopy(InIt beg, InIt end, OutIt out)
+{
+    uint STRIDE = Block::blockSize();
+    InIt  t = beg + Block::threadLineId();
+    OutIt o = out + (t - beg);
+
+    for(; t < end; t += STRIDE, o += STRIDE)
+        *o = *t;
+}
+
+template <class InIt, class OutIt, class UnOp>
+__device__ __forceinline__ static void blockTransfrom(InIt beg, InIt end, OutIt out, const UnOp& op)
+{
+    uint STRIDE = Block::blockSize();
+    InIt  t = beg + Block::threadLineId();
+    OutIt o = out + (t - beg);
+
+    for(; t < end; t += STRIDE, o += STRIDE)
+        *o = op(*t);
+}
+
+template <class InIt1, class InIt2, class OutIt, class BinOp>
+__device__ __forceinline__ static void blockTransfrom(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, const BinOp& op)
+{
+    uint STRIDE = Block::blockSize();
+    InIt1 t1 = beg1 + Block::threadLineId();
+    InIt2 t2 = beg2 + Block::threadLineId();
+    OutIt o  = out + (t1 - beg1);
+
+    for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, o += STRIDE)
+        *o = op(*t1, *t2);
+}
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/detail/reduce.hpp b/modules/cudev/include/opencv2/cudev/block/detail/reduce.hpp
new file mode 100644
index 0000000000..02803b003a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/detail/reduce.hpp
@@ -0,0 +1,392 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_REDUCE_DETAIL_HPP__
+#define __OPENCV_CUDEV_BLOCK_REDUCE_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/type_traits.hpp"
+#include "../../warp/warp.hpp"
+#include "../../warp/shuffle.hpp"
+
+namespace cv { namespace cudev {
+
+namespace block_reduce_detail
+{
+    // GetType
+
+    template <typename T> struct GetType;
+
+    template <typename T> struct GetType<T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<volatile T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<T&>
+    {
+        typedef T type;
+    };
+
+    // For
+
+    template <int I, int N> struct For
+    {
+        template <class PointerTuple, class ValTuple>
+        __device__ static void loadToSmem(const PointerTuple& smem, const ValTuple& val, uint tid)
+        {
+            get<I>(smem)[tid] = get<I>(val);
+
+            For<I + 1, N>::loadToSmem(smem, val, tid);
+        }
+
+        template <class PointerTuple, class ValTuple>
+        __device__ static void loadFromSmem(const PointerTuple& smem, const ValTuple& val, uint tid)
+        {
+            get<I>(val) = get<I>(smem)[tid];
+
+            For<I + 1, N>::loadFromSmem(smem, val, tid);
+        }
+
+        template <class PointerTuple, class ValTuple, class OpTuple>
+        __device__ static void merge(const PointerTuple& smem, const ValTuple& val, uint tid, uint delta, const OpTuple& op)
+        {
+            typename GetType<typename tuple_element<I, PointerTuple>::type>::type reg = get<I>(smem)[tid + delta];
+            get<I>(smem)[tid] = get<I>(val) = get<I>(op)(get<I>(val), reg);
+
+            For<I + 1, N>::merge(smem, val, tid, delta, op);
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        template <class ValTuple, class OpTuple>
+        __device__ static void mergeShfl(const ValTuple& val, uint delta, uint width, const OpTuple& op)
+        {
+            typename GetType<typename tuple_element<I, ValTuple>::type>::type reg = shfl_down(get<I>(val), delta, width);
+            get<I>(val) = get<I>(op)(get<I>(val), reg);
+
+            For<I + 1, N>::mergeShfl(val, delta, width, op);
+        }
+#endif
+    };
+
+    template <int N> struct For<N, N>
+    {
+        template <class PointerTuple, class ValTuple>
+        __device__ __forceinline__ static void loadToSmem(const PointerTuple&, const ValTuple&, uint)
+        {
+        }
+        template <class PointerTuple, class ValTuple>
+        __device__ __forceinline__ static void loadFromSmem(const PointerTuple&, const ValTuple&, uint)
+        {
+        }
+
+        template <class PointerTuple, class ValTuple, class OpTuple>
+        __device__ __forceinline__ static void merge(const PointerTuple&, const ValTuple&, uint, uint, const OpTuple&)
+        {
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        template <class ValTuple, class OpTuple>
+        __device__ __forceinline__ static void mergeShfl(const ValTuple&, uint, uint, const OpTuple&)
+        {
+        }
+#endif
+    };
+
+    // loadToSmem / loadFromSmem
+
+    template <typename T>
+    __device__ __forceinline__ void loadToSmem(volatile T* smem, T& val, uint tid)
+    {
+        smem[tid] = val;
+    }
+
+    template <typename T>
+    __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& val, uint tid)
+    {
+        val = smem[tid];
+    }
+
+    template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+    __device__ __forceinline__ void loadToSmem(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                               const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                               uint tid)
+    {
+        For<0, tuple_size<tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadToSmem(smem, val, tid);
+    }
+
+    template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+    __device__ __forceinline__ void loadFromSmem(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                                     const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                     uint tid)
+    {
+        For<0, tuple_size<tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadFromSmem(smem, val, tid);
+    }
+
+    // merge
+
+    template <typename T, class Op>
+    __device__ __forceinline__ void merge(volatile T* smem, T& val, uint tid, uint delta, const Op& op)
+    {
+        T reg = smem[tid + delta];
+        smem[tid] = val = op(val, reg);
+    }
+
+    template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void merge(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                          const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                          uint tid,
+                                          uint delta,
+                                          const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        For<0, tuple_size<tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::merge(smem, val, tid, delta, op);
+    }
+
+    // mergeShfl
+
+#if CV_CUDEV_ARCH >= 300
+    template <typename T, class Op>
+    __device__ __forceinline__ void mergeShfl(T& val, uint delta, uint width, const Op& op)
+    {
+        T reg = shfl_down(val, delta, width);
+        val = op(val, reg);
+    }
+
+    template <typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void mergeShfl(const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                              uint delta,
+                                              uint width,
+                                              const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        For<0, tuple_size<tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9> >::value>::mergeShfl(val, delta, width, op);
+    }
+#endif
+
+    // Generic
+
+    template <int N> struct Generic
+    {
+        template <typename Pointer, typename Reference, class Op>
+        __device__ static void reduce(Pointer smem, Reference val, uint tid, Op op)
+        {
+            loadToSmem(smem, val, tid);
+            if (N >= 32)
+                __syncthreads();
+
+            if (N >= 2048)
+            {
+                if (tid < 1024)
+                    merge(smem, val, tid, 1024, op);
+
+                __syncthreads();
+            }
+            if (N >= 1024)
+            {
+                if (tid < 512)
+                    merge(smem, val, tid, 512, op);
+
+                __syncthreads();
+            }
+            if (N >= 512)
+            {
+                if (tid < 256)
+                    merge(smem, val, tid, 256, op);
+
+                __syncthreads();
+            }
+            if (N >= 256)
+            {
+                if (tid < 128)
+                    merge(smem, val, tid, 128, op);
+
+                __syncthreads();
+            }
+            if (N >= 128)
+            {
+                if (tid < 64)
+                    merge(smem, val, tid, 64, op);
+
+                __syncthreads();
+            }
+            if (N >= 64)
+            {
+                if (tid < 32)
+                    merge(smem, val, tid, 32, op);
+            }
+
+            if (tid < 16)
+            {
+                merge(smem, val, tid, 16, op);
+                merge(smem, val, tid, 8, op);
+                merge(smem, val, tid, 4, op);
+                merge(smem, val, tid, 2, op);
+                merge(smem, val, tid, 1, op);
+            }
+        }
+    };
+
+    // Unroll
+
+    template <int I, typename Pointer, typename Reference, class Op> struct Unroll
+    {
+        __device__ static void loop(Pointer smem, Reference val, uint tid, Op op)
+        {
+            merge(smem, val, tid, I, op);
+            Unroll<I / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        __device__ static void loopShfl(Reference val, Op op, uint N)
+        {
+            mergeShfl(val, I, N, op);
+            Unroll<I / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+        }
+#endif
+    };
+
+    template <typename Pointer, typename Reference, class Op> struct Unroll<0, Pointer, Reference, Op>
+    {
+        __device__ __forceinline__ static void loop(Pointer, Reference, uint, Op)
+        {
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        __device__ __forceinline__ static void loopShfl(Reference, Op, uint)
+        {
+        }
+#endif
+    };
+
+    // WarpOptimized
+
+    template <int N> struct WarpOptimized
+    {
+        template <typename Pointer, typename Reference, class Op>
+        __device__ static void reduce(Pointer smem, Reference val, uint tid, Op op)
+        {
+        #if CV_CUDEV_ARCH >= 300
+            (void) smem;
+            (void) tid;
+
+            Unroll<N / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+        #else
+            loadToSmem(smem, val, tid);
+
+            if (tid < N / 2)
+                Unroll<N / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+        #endif
+        }
+    };
+
+    // GenericOptimized32
+
+    template <int N> struct GenericOptimized32
+    {
+        enum { M = N / 32 };
+
+        template <typename Pointer, typename Reference, class Op>
+        __device__ static void reduce(Pointer smem, Reference val, uint tid, Op op)
+        {
+            const uint laneId = Warp::laneId();
+
+        #if CV_CUDEV_ARCH >= 300
+            Unroll<16, Pointer, Reference, Op>::loopShfl(val, op, warpSize);
+
+            if (laneId == 0)
+                loadToSmem(smem, val, tid / 32);
+        #else
+            loadToSmem(smem, val, tid);
+
+            if (laneId < 16)
+                Unroll<16, Pointer, Reference, Op>::loop(smem, val, tid, op);
+
+            __syncthreads();
+
+            if (laneId == 0)
+                loadToSmem(smem, val, tid / 32);
+        #endif
+
+            __syncthreads();
+
+            loadFromSmem(smem, val, tid);
+
+            if (tid < 32)
+            {
+        #if CV_CUDEV_ARCH >= 300
+                Unroll<M / 2, Pointer, Reference, Op>::loopShfl(val, op, M);
+        #else
+                Unroll<M / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+        #endif
+            }
+        }
+    };
+
+    template <int N> struct Dispatcher
+    {
+        typedef typename SelectIf<
+            (N <= 32) && IsPowerOf2<N>::value,
+            WarpOptimized<N>,
+            typename SelectIf<
+                (N <= 1024) && IsPowerOf2<N>::value,
+                GenericOptimized32<N>,
+                Generic<N>
+            >::type
+        >::type reductor;
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/detail/reduce_key_val.hpp b/modules/cudev/include/opencv2/cudev/block/detail/reduce_key_val.hpp
new file mode 100644
index 0000000000..edbe8a586b
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/detail/reduce_key_val.hpp
@@ -0,0 +1,394 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_REDUCE_KEY_VAL_DETAIL_HPP__
+#define __OPENCV_CUDEV_BLOCK_REDUCE_KEY_VAL_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/type_traits.hpp"
+#include "../../warp/warp.hpp"
+
+namespace cv { namespace cudev {
+
+namespace block_reduce_key_val_detail
+{
+    // GetType
+
+    template <typename T> struct GetType;
+
+    template <typename T> struct GetType<T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<volatile T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<T&>
+    {
+        typedef T type;
+    };
+
+    // For
+
+    template <int I, int N> struct For
+    {
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void loadToSmem(const PointerTuple& smem, const ReferenceTuple& data, uint tid)
+        {
+            get<I>(smem)[tid] = get<I>(data);
+
+            For<I + 1, N>::loadToSmem(smem, data, tid);
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void loadFromSmem(const PointerTuple& smem, const ReferenceTuple& data, uint tid)
+        {
+            get<I>(data) = get<I>(smem)[tid];
+
+            For<I + 1, N>::loadFromSmem(smem, data, tid);
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void copy(const PointerTuple& svals, const ReferenceTuple& val, uint tid, uint delta)
+        {
+            get<I>(svals)[tid] = get<I>(val) = get<I>(svals)[tid + delta];
+
+            For<I + 1, N>::copy(svals, val, tid, delta);
+        }
+
+        template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+        __device__ static void merge(const KeyPointerTuple& skeys, const KeyReferenceTuple& key,
+                                     const ValPointerTuple& svals, const ValReferenceTuple& val,
+                                     const CmpTuple& cmp,
+                                     uint tid, uint delta)
+        {
+            typename GetType<typename tuple_element<I, KeyPointerTuple>::type>::type reg = get<I>(skeys)[tid + delta];
+
+            if (get<I>(cmp)(reg, get<I>(key)))
+            {
+                get<I>(skeys)[tid] = get<I>(key) = reg;
+                get<I>(svals)[tid] = get<I>(val) = get<I>(svals)[tid + delta];
+            }
+
+            For<I + 1, N>::merge(skeys, key, svals, val, cmp, tid, delta);
+        }
+    };
+
+    template <int N> struct For<N, N>
+    {
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void loadToSmem(const PointerTuple&, const ReferenceTuple&, uint)
+        {
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void loadFromSmem(const PointerTuple&, const ReferenceTuple&, uint)
+        {
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void copy(const PointerTuple&, const ReferenceTuple&, uint, uint)
+        {
+        }
+
+        template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+        __device__ static void merge(const KeyPointerTuple&, const KeyReferenceTuple&,
+                                     const ValPointerTuple&, const ValReferenceTuple&,
+                                     const CmpTuple&,
+                                     uint, uint)
+        {
+        }
+    };
+
+    // loadToSmem / loadFromSmem
+
+    template <typename T>
+    __device__ __forceinline__ void loadToSmem(volatile T* smem, T& data, uint tid)
+    {
+        smem[tid] = data;
+    }
+
+    template <typename T>
+    __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& data, uint tid)
+    {
+        data = smem[tid];
+    }
+
+    template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+    __device__ __forceinline__ void loadToSmem(const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                               const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                               uint tid)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadToSmem(smem, data, tid);
+    }
+
+    template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+    __device__ __forceinline__ void loadFromSmem(const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                                 const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                                 uint tid)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadFromSmem(smem, data, tid);
+    }
+
+    // copyVals
+
+    template <typename V>
+    __device__ __forceinline__ void copyVals(volatile V* svals, V& val, uint tid, uint delta)
+    {
+        svals[tid] = val = svals[tid + delta];
+    }
+
+    template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+    __device__ __forceinline__ void copyVals(const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                             const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                             uint tid, uint delta)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::copy(svals, val, tid, delta);
+    }
+
+    // merge
+
+    template <typename K, typename V, class Cmp>
+    __device__ void merge(volatile K* skeys, K& key, volatile V* svals, V& val, const Cmp& cmp, uint tid, uint delta)
+    {
+        K reg = skeys[tid + delta];
+
+        if (cmp(reg, key))
+        {
+            skeys[tid] = key = reg;
+            copyVals(svals, val, tid, delta);
+        }
+    }
+
+    template <typename K,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp>
+    __device__ void merge(volatile K* skeys, K& key,
+                          const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                          const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                          const Cmp& cmp, uint tid, uint delta)
+    {
+        K reg = skeys[tid + delta];
+
+        if (cmp(reg, key))
+        {
+            skeys[tid] = key = reg;
+            copyVals(svals, val, tid, delta);
+        }
+    }
+
+    template <typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+              typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+    __device__ __forceinline__ void merge(const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                          const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                          const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                          const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                          const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                          uint tid, uint delta)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::merge(skeys, key, svals, val, cmp, tid, delta);
+    }
+
+    // Generic
+
+    template <int N> struct Generic
+    {
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        __device__ static void reduce(KP skeys, KR key, VP svals, VR val, uint tid, Cmp cmp)
+        {
+            loadToSmem(skeys, key, tid);
+            loadValsToSmem(svals, val, tid);
+            if (N >= 32)
+                __syncthreads();
+
+            if (N >= 2048)
+            {
+                if (tid < 1024)
+                    merge(skeys, key, svals, val, cmp, tid, 1024);
+
+                __syncthreads();
+            }
+            if (N >= 1024)
+            {
+                if (tid < 512)
+                    merge(skeys, key, svals, val, cmp, tid, 512);
+
+                __syncthreads();
+            }
+            if (N >= 512)
+            {
+                if (tid < 256)
+                    merge(skeys, key, svals, val, cmp, tid, 256);
+
+                __syncthreads();
+            }
+            if (N >= 256)
+            {
+                if (tid < 128)
+                    merge(skeys, key, svals, val, cmp, tid, 128);
+
+                __syncthreads();
+            }
+            if (N >= 128)
+            {
+                if (tid < 64)
+                    merge(skeys, key, svals, val, cmp, tid, 64);
+
+                __syncthreads();
+            }
+            if (N >= 64)
+            {
+                if (tid < 32)
+                    merge(skeys, key, svals, val, cmp, tid, 32);
+            }
+
+            if (tid < 16)
+            {
+                merge(skeys, key, svals, val, cmp, tid, 16);
+                merge(skeys, key, svals, val, cmp, tid, 8);
+                merge(skeys, key, svals, val, cmp, tid, 4);
+                merge(skeys, key, svals, val, cmp, tid, 2);
+                merge(skeys, key, svals, val, cmp, tid, 1);
+            }
+        }
+    };
+
+    // Unroll
+
+    template <int I, class KP, class KR, class VP, class VR, class Cmp> struct Unroll
+    {
+        __device__ static void loop(KP skeys, KR key, VP svals, VR val, uint tid, Cmp cmp)
+        {
+            merge(skeys, key, svals, val, cmp, tid, I);
+            Unroll<I / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+        }
+    };
+
+    template <class KP, class KR, class VP, class VR, class Cmp> struct Unroll<0, KP, KR, VP, VR, Cmp>
+    {
+        __device__ __forceinline__ static void loop(KP, KR, VP, VR, uint, Cmp)
+        {
+        }
+    };
+
+    // WarpOptimized
+
+    template <int N> struct WarpOptimized
+    {
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        __device__ static void reduce(KP skeys, KR key, VP svals, VR val, uint tid, Cmp cmp)
+        {
+            loadToSmem(skeys, key, tid);
+            loadToSmem(svals, val, tid);
+
+            if (tid < N / 2)
+                Unroll<N / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+        }
+    };
+
+    // GenericOptimized32
+
+    template <uint N> struct GenericOptimized32
+    {
+        enum { M = N / 32 };
+
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        __device__ static void reduce(KP skeys, KR key, VP svals, VR val, uint tid, Cmp cmp)
+        {
+            const uint laneId = Warp::laneId();
+
+            loadToSmem(skeys, key, tid);
+            loadToSmem(svals, val, tid);
+
+            if (laneId < 16)
+                Unroll<16, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+
+            __syncthreads();
+
+            if (laneId == 0)
+            {
+                loadToSmem(skeys, key, tid / 32);
+                loadToSmem(svals, val, tid / 32);
+            }
+
+            __syncthreads();
+
+            loadFromSmem(skeys, key, tid);
+
+            if (tid < 32)
+            {
+                Unroll<M / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+            }
+        }
+    };
+
+    template <int N> struct Dispatcher
+    {
+        typedef typename SelectIf<
+            (N <= 32) && IsPowerOf2<N>::value,
+            WarpOptimized<N>,
+            typename SelectIf<
+                (N <= 1024) && IsPowerOf2<N>::value,
+                GenericOptimized32<N>,
+                Generic<N>
+            >::type
+        >::type reductor;
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/dynamic_smem.hpp b/modules/cudev/include/opencv2/cudev/block/dynamic_smem.hpp
new file mode 100644
index 0000000000..9f9ba60009
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/dynamic_smem.hpp
@@ -0,0 +1,86 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_DYNAMIC_SMEM_HPP__
+#define __OPENCV_CUDEV_BLOCK_DYNAMIC_SMEM_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+template <class T> struct DynamicSharedMem
+{
+    __device__ __forceinline__ operator T*()
+    {
+        extern __shared__ int __smem[];
+        return (T*) __smem;
+    }
+
+    __device__ __forceinline__ operator const T*() const
+    {
+        extern __shared__ int __smem[];
+        return (T*) __smem;
+    }
+};
+
+// specialize for double to avoid unaligned memory access compile errors
+template <> struct DynamicSharedMem<double>
+{
+    __device__ __forceinline__ operator double*()
+    {
+        extern __shared__ double __smem_d[];
+        return (double*) __smem_d;
+    }
+
+    __device__ __forceinline__ operator const double*() const
+    {
+        extern __shared__ double __smem_d[];
+        return (double*) __smem_d;
+    }
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/reduce.hpp b/modules/cudev/include/opencv2/cudev/block/reduce.hpp
new file mode 100644
index 0000000000..4c9022631b
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/reduce.hpp
@@ -0,0 +1,128 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_REDUCE_HPP__
+#define __OPENCV_CUDEV_BLOCK_REDUCE_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "../warp/reduce.hpp"
+#include "detail/reduce.hpp"
+#include "detail/reduce_key_val.hpp"
+
+namespace cv { namespace cudev {
+
+// blockReduce
+
+template <int N, typename T, class Op>
+__device__ __forceinline__ void blockReduce(volatile T* smem, T& val, uint tid, const Op& op)
+{
+    block_reduce_detail::Dispatcher<N>::reductor::template reduce<volatile T*, T&, const Op&>(smem, val, tid, op);
+}
+
+template <int N,
+          typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+          typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+          class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+__device__ __forceinline__ void blockReduce(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                            const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                            uint tid,
+                                            const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+{
+    block_reduce_detail::Dispatcher<N>::reductor::template reduce<
+            const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>&,
+            const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>&,
+            const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>&>(smem, val, tid, op);
+}
+
+// blockReduceKeyVal
+
+template <int N, typename K, typename V, class Cmp>
+__device__ __forceinline__ void blockReduceKeyVal(volatile K* skeys, K& key, volatile V* svals, V& val, uint tid, const Cmp& cmp)
+{
+    block_reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&, volatile V*, V&, const Cmp&>(skeys, key, svals, val, tid, cmp);
+}
+
+template <int N,
+          typename K,
+          typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+          typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+          class Cmp>
+__device__ __forceinline__ void blockReduceKeyVal(volatile K* skeys, K& key,
+                                                  const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                  const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  uint tid, const Cmp& cmp)
+{
+    block_reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&,
+            const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+            const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+            const Cmp&>(skeys, key, svals, val, tid, cmp);
+}
+
+template <int N,
+          typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+          typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+          typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+          typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+          class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+__device__ __forceinline__ void blockReduceKeyVal(const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                                  const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                  const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                  const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  uint tid,
+                                                  const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp)
+{
+    block_reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<
+            const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>&,
+            const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>&,
+            const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+            const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+            const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>&
+            >(skeys, key, svals, val, tid, cmp);
+}
+
+}}
+
+#endif
diff --git a/modules/gpu/src/thread_wrappers.h b/modules/cudev/include/opencv2/cudev/block/scan.hpp
similarity index 59%
rename from modules/gpu/src/thread_wrappers.h
rename to modules/cudev/include/opencv2/cudev/block/scan.hpp
index 78e675216c..c54dfef9f3 100644
--- a/modules/gpu/src/thread_wrappers.h
+++ b/modules/cudev/include/opencv2/cudev/block/scan.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,73 +41,61 @@
 //
 //M*/
 
-#ifndef __THREAD_WRAPPERS_H__
-#define __THREAD_WRAPPERS_H__
+#pragma once
 
-#include "precomp.hpp"
+#ifndef __OPENCV_CUDEV_BLOCK_SCAN_HPP__
+#define __OPENCV_CUDEV_BLOCK_SCAN_HPP__
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#include "../common.hpp"
+#include "../warp/scan.hpp"
 
-namespace cv { namespace gpu
+namespace cv { namespace cudev {
+
+template <int THREADS_NUM, typename T>
+__device__ T blockScanInclusive(T data, volatile T* smem, uint tid)
 {
-    namespace detail
+    if (THREADS_NUM > WARP_SIZE)
     {
-        class CriticalSection
+        // bottom-level inclusive warp scan
+        T warpResult = warpScanInclusive(data, smem, tid);
+
+        __syncthreads();
+
+        // save top elements of each warp for exclusive warp scan
+        // sync to wait for warp scans to complete (because s_Data is being overwritten)
+        if ((tid & (WARP_SIZE - 1)) == (WARP_SIZE - 1))
         {
-        public:
-            CriticalSection();
-            ~CriticalSection();
+            smem[tid >> LOG_WARP_SIZE] = warpResult;
+        }
 
-            void enter();
-            void leave();
+        __syncthreads();
 
-            class AutoLock
-            {
-            public:
-                explicit AutoLock(CriticalSection& criticalSection) :
-                    criticalSection_(criticalSection)
-                {
-                    criticalSection_.enter();
-                }
-
-                ~AutoLock()
-                {
-                    criticalSection_.leave();
-                }
-
-            private:
-                CriticalSection& criticalSection_;
-            };
-
-        private:
-            CriticalSection(const CriticalSection&);
-            CriticalSection& operator=(const CriticalSection&);
-
-            class Impl;
-            std::auto_ptr<Impl> impl_;
-        };
-
-        class Thread
+        if (tid < (THREADS_NUM / WARP_SIZE))
         {
-        public:
-            explicit Thread(void (*func)(void* userData), void* userData = 0);
-            ~Thread();
+            // grab top warp elements
+            T val = smem[tid];
 
-            void wait();
+            // calculate exclusive scan and write back to shared memory
+            smem[tid] = warpScanExclusive(val, smem, tid);
+        }
 
-            static void sleep(int ms);
-
-        private:
-            Thread(const Thread&);
-            Thread& operator=(const Thread&);
-
-            class Impl;
-            std::auto_ptr<Impl> impl_;
-        };
+        __syncthreads();
 
+        // return updated warp scans with exclusive scan results
+        return warpResult + smem[tid >> LOG_WARP_SIZE];
     }
+    else
+    {
+        return warpScanInclusive(data, smem, tid);
+    }
+}
+
+template <int THREADS_NUM, typename T>
+__device__ __forceinline__ T blockScanExclusive(T data, volatile T* smem, uint tid)
+{
+    return blockScanInclusive<THREADS_NUM>(data, smem, tid) - data;
+}
+
 }}
 
-#endif // HAVE_CUDA
-
-#endif // __THREAD_WRAPPERS_H__
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/block/vec_distance.hpp b/modules/cudev/include/opencv2/cudev/block/vec_distance.hpp
new file mode 100644
index 0000000000..c48e9146ef
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/block/vec_distance.hpp
@@ -0,0 +1,184 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_BLOCK_VEC_DISTANCE_HPP__
+#define __OPENCV_CUDEV_BLOCK_VEC_DISTANCE_HPP__
+
+#include "../common.hpp"
+#include "../functional/functional.hpp"
+#include "../warp/reduce.hpp"
+#include "reduce.hpp"
+
+namespace cv { namespace cudev {
+
+// NormL1
+
+template <typename T> struct NormL1
+{
+    typedef int value_type;
+    typedef uint result_type;
+
+    result_type mySum;
+
+    __device__ __forceinline__ NormL1() : mySum(0) {}
+
+    __device__ __forceinline__ void reduceThread(value_type val1, value_type val2)
+    {
+        mySum = __sad(val1, val2, mySum);
+    }
+
+    __device__ __forceinline__ void reduceWarp(result_type* smem, uint tid)
+    {
+        warpReduce(smem, mySum, tid, plus<result_type>());
+    }
+
+    template <int THREAD_DIM> __device__ __forceinline__ void reduceBlock(result_type* smem, uint tid)
+    {
+        blockReduce<THREAD_DIM>(smem, mySum, tid, plus<result_type>());
+    }
+
+    __device__ __forceinline__ operator result_type() const
+    {
+        return mySum;
+    }
+};
+template <> struct NormL1<float>
+{
+    typedef float value_type;
+    typedef float result_type;
+
+    result_type mySum;
+
+    __device__ __forceinline__ NormL1() : mySum(0.0f) {}
+
+    __device__ __forceinline__ void reduceThread(value_type val1, value_type val2)
+    {
+        mySum += ::fabsf(val1 - val2);
+    }
+
+    __device__ __forceinline__ void reduceWarp(result_type* smem, uint tid)
+    {
+        warpReduce(smem, mySum, tid, plus<result_type>());
+    }
+
+    template <int THREAD_DIM> __device__ __forceinline__ void reduceBlock(result_type* smem, uint tid)
+    {
+        blockReduce<THREAD_DIM>(smem, mySum, tid, plus<result_type>());
+    }
+
+    __device__ __forceinline__ operator result_type() const
+    {
+        return mySum;
+    }
+};
+
+// NormL2
+
+struct NormL2
+{
+    typedef float value_type;
+    typedef float result_type;
+
+    result_type mySum;
+
+    __device__ __forceinline__ NormL2() : mySum(0.0f) {}
+
+    __device__ __forceinline__ void reduceThread(value_type val1, value_type val2)
+    {
+        const float diff = val1 - val2;
+        mySum += diff * diff;
+    }
+
+    __device__ __forceinline__ void reduceWarp(result_type* smem, uint tid)
+    {
+        warpReduce(smem, mySum, tid, plus<result_type>());
+    }
+
+    template <int THREAD_DIM> __device__ __forceinline__ void reduceBlock(result_type* smem, uint tid)
+    {
+        blockReduce<THREAD_DIM>(smem, mySum, tid, plus<result_type>());
+    }
+
+    __device__ __forceinline__ operator result_type() const
+    {
+        return ::sqrtf(mySum);
+    }
+};
+
+// NormHamming
+
+struct NormHamming
+{
+    typedef int value_type;
+    typedef int result_type;
+
+    result_type mySum;
+
+    __device__ __forceinline__ NormHamming() : mySum(0) {}
+
+    __device__ __forceinline__ void reduceThread(value_type val1, value_type val2)
+    {
+        mySum += __popc(val1 ^ val2);
+    }
+
+    __device__ __forceinline__ void reduceWarp(result_type* smem, uint tid)
+    {
+        warpReduce(smem, mySum, tid, plus<result_type>());
+    }
+
+    template <int THREAD_DIM> __device__ __forceinline__ void reduceBlock(result_type* smem, uint tid)
+    {
+        blockReduce<THREAD_DIM>(smem, mySum, tid, plus<result_type>());
+    }
+
+    __device__ __forceinline__ operator result_type() const
+    {
+        return mySum;
+    }
+};
+
+}}
+
+#endif
diff --git a/modules/gpu/src/ffmpeg_video_source.h b/modules/cudev/include/opencv2/cudev/common.hpp
similarity index 65%
rename from modules/gpu/src/ffmpeg_video_source.h
rename to modules/cudev/include/opencv2/cudev/common.hpp
index 41bf0cfd0e..be79901729 100644
--- a/modules/gpu/src/ffmpeg_video_source.h
+++ b/modules/cudev/include/opencv2/cudev/common.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,49 +41,53 @@
 //
 //M*/
 
-#ifndef __FFMPEG_VIDEO_SOURCE_H__
-#define __FFMPEG_VIDEO_SOURCE_H__
+#pragma once
 
-#include "precomp.hpp"
-#include "thread_wrappers.h"
+#ifndef __OPENCV_CUDEV_COMMON_HPP__
+#define __OPENCV_CUDEV_COMMON_HPP__
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#include <cuda_runtime.h>
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/core/gpu_stream_accessor.hpp"
 
-struct InputMediaStream_FFMPEG;
+namespace cv { namespace cudev {
 
-namespace cv { namespace gpu
+using namespace cv::gpu;
+
+// CV_CUDEV_ARCH
+
+#ifndef __CUDA_ARCH__
+#   define CV_CUDEV_ARCH 0
+#else
+#   define CV_CUDEV_ARCH __CUDA_ARCH__
+#endif
+
+// CV_CUDEV_SAFE_CALL
+
+__host__ __forceinline__ void checkCudaError(cudaError_t err, const char* file, const int line, const char* func)
 {
-    namespace detail
-    {
-        class FFmpegVideoSource : public VideoReader_GPU::VideoSource
-        {
-        public:
-            FFmpegVideoSource(const std::string& fname);
-            ~FFmpegVideoSource();
+    if (cudaSuccess != err)
+        cv::error(cv::Error::GpuApiCallError, cudaGetErrorString(err), func, file, line);
+}
 
-            VideoReader_GPU::FormatInfo format() const;
-            void start();
-            void stop();
-            bool isStarted() const;
-            bool hasError() const;
+#ifdef __GNUC__
+#   define CV_CUDEV_SAFE_CALL(expr) cv::cudev::checkCudaError((expr), __FILE__, __LINE__, __func__)
+#else
+#   define CV_CUDEV_SAFE_CALL(expr) cv::cudev::checkCudaError((expr), __FILE__, __LINE__, "")
+#endif
 
-        private:
-            FFmpegVideoSource(const FFmpegVideoSource&);
-            FFmpegVideoSource& operator =(const FFmpegVideoSource&);
+// divUp
 
-            VideoReader_GPU::FormatInfo format_;
+__host__ __device__ __forceinline__ int divUp(int total, int grain)
+{
+    return (total + grain - 1) / grain;
+}
 
-            InputMediaStream_FFMPEG* stream_;
+// math constants
 
-            std::auto_ptr<Thread> thread_;
-            volatile bool stop_;
-            volatile bool hasError_;
+#define CV_PI_F   ((float)CV_PI)
+#define CV_LOG2_F ((float)CV_LOG2)
 
-            static void readLoop(void* userData);
-        };
-    }
 }}
 
-#endif // HAVE_CUDA
-
-#endif // __CUVUD_VIDEO_SOURCE_H__
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/binary_func.hpp b/modules/cudev/include/opencv2/cudev/expr/binary_func.hpp
new file mode 100644
index 0000000000..f35ea2dc3c
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/binary_func.hpp
@@ -0,0 +1,75 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_BINARY_FUNC_HPP__
+#define __OPENCV_CUDEV_EXPR_BINARY_FUNC_HPP__
+
+#include "../common.hpp"
+#include "../util/type_traits.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../functional/functional.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+#define CV_CUDEV_EXPR_BINARY_FUNC(name) \
+    template <class SrcPtr1, class SrcPtr2> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<SrcPtr1>::ptr_type, typename PtrTraits<SrcPtr2>::ptr_type, name ## _func<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type> > > \
+    name ## _(const SrcPtr1& src1, const SrcPtr2& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, name ## _func<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type>())); \
+    }
+
+CV_CUDEV_EXPR_BINARY_FUNC(hypot)
+CV_CUDEV_EXPR_BINARY_FUNC(magnitude)
+CV_CUDEV_EXPR_BINARY_FUNC(atan2)
+CV_CUDEV_EXPR_BINARY_FUNC(absdiff)
+
+#undef CV_CUDEV_EXPR_BINARY_FUNC
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/binary_op.hpp b/modules/cudev/include/opencv2/cudev/expr/binary_op.hpp
new file mode 100644
index 0000000000..f7e9655720
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/binary_op.hpp
@@ -0,0 +1,235 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_BINARY_OP_HPP__
+#define __OPENCV_CUDEV_EXPR_BINARY_OP_HPP__
+
+#include "../common.hpp"
+#include "../util/type_traits.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/texture.hpp"
+#include "../ptr2d/glob.hpp"
+#include "../functional/functional.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+// Binary Operations
+
+#define CV_CUDEV_EXPR_BINOP_INST(op, functor) \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, typename PtrTraits<GpuMat_<T> >::ptr_type, functor<T> > > \
+    operator op(const GpuMat_<T>& src1, const GpuMat_<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, typename PtrTraits<GlobPtrSz<T> >::ptr_type, functor<T> > > \
+    operator op(const GpuMat_<T>& src1, const GlobPtrSz<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, typename PtrTraits<GpuMat_<T> >::ptr_type, functor<T> > > \
+    operator op(const GlobPtrSz<T>& src1, const GpuMat_<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, typename PtrTraits<Texture<T> >::ptr_type, functor<T> > > \
+    operator op(const GpuMat_<T>& src1, const Texture<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, typename PtrTraits<GpuMat_<T> >::ptr_type, functor<T> > > \
+    operator op(const Texture<T>& src1, const GpuMat_<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, typename PtrTraits<Body>::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const GpuMat_<T>& src1, const Expr<Body>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2.body, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, typename PtrTraits<GpuMat_<T> >::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const Expr<Body>& src1, const GpuMat_<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1.body, src2, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, Binder2nd< functor<T> > > > \
+    operator op(const GpuMat_<T>& src, T val) \
+    { \
+        return makeExpr(transformPtr(src, bind2nd(functor<T>(), val))); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, Binder1st< functor<T> > > > \
+    operator op(T val, const GpuMat_<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, bind1st(functor<T>(), val))); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, typename PtrTraits<GlobPtrSz<T> >::ptr_type, functor<T> > > \
+    operator op(const GlobPtrSz<T>& src1, const GlobPtrSz<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, typename PtrTraits<Texture<T> >::ptr_type, functor<T> > > \
+    operator op(const GlobPtrSz<T>& src1, const Texture<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, typename PtrTraits<GlobPtrSz<T> >::ptr_type, functor<T> > > \
+    operator op(const Texture<T>& src1, const GlobPtrSz<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, typename PtrTraits<Body>::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const GlobPtrSz<T>& src1, const Expr<Body>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2.body, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, typename PtrTraits<GlobPtrSz<T> >::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const Expr<Body>& src1, const GlobPtrSz<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1.body, src2, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, Binder2nd< functor<T> > > > \
+    operator op(const GlobPtrSz<T>& src, T val) \
+    { \
+        return makeExpr(transformPtr(src, bind2nd(functor<T>(), val))); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, Binder1st< functor<T> > > > \
+    operator op(T val, const GlobPtrSz<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, bind1st(functor<T>(), val))); \
+    } \
+    template <typename T> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, typename PtrTraits<Texture<T> >::ptr_type, functor<T> > > \
+    operator op(const Texture<T>& src1, const Texture<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2, functor<T>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, typename PtrTraits<Body>::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const Texture<T>& src1, const Expr<Body>& src2) \
+    { \
+        return makeExpr(transformPtr(src1, src2.body, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T, class Body> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, typename PtrTraits<Texture<T> >::ptr_type, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type> > > \
+    operator op(const Expr<Body>& src1, const Texture<T>& src2) \
+    { \
+        return makeExpr(transformPtr(src1.body, src2, functor<typename LargerType<T, typename PtrTraits<Body>::value_type>::type>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, Binder2nd< functor<T> > > > \
+    operator op(const Texture<T>& src, T val) \
+    { \
+        return makeExpr(transformPtr(src, bind2nd(functor<T>(), val))); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, Binder1st< functor<T> > > > \
+    operator op(T val, const Texture<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, bind1st(functor<T>(), val))); \
+    } \
+    template <class Body1, class Body2> \
+    __host__ Expr<BinaryTransformPtrSz<typename PtrTraits<Body1>::ptr_type, typename PtrTraits<Body2>::ptr_type, functor<typename LargerType<typename PtrTraits<Body1>::value_type, typename PtrTraits<Body2>::value_type>::type> > > \
+    operator op(const Expr<Body1>& a, const Expr<Body2>& b) \
+    { \
+        return makeExpr(transformPtr(a.body, b.body, functor<typename LargerType<typename PtrTraits<Body1>::value_type, typename PtrTraits<Body2>::value_type>::type>())); \
+    } \
+    template <class Body> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, Binder2nd< functor<typename Body::value_type> > > > \
+    operator op(const Expr<Body>& a, typename Body::value_type val) \
+    { \
+        return makeExpr(transformPtr(a.body, bind2nd(functor<typename Body::value_type>(), val))); \
+    } \
+    template <class Body> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, Binder1st< functor<typename Body::value_type> > > > \
+    operator op(typename Body::value_type val, const Expr<Body>& a) \
+    { \
+        return makeExpr(transformPtr(a.body, bind1st(functor<typename Body::value_type>(), val))); \
+    }
+
+CV_CUDEV_EXPR_BINOP_INST(+, plus)
+CV_CUDEV_EXPR_BINOP_INST(-, minus)
+CV_CUDEV_EXPR_BINOP_INST(*, multiplies)
+CV_CUDEV_EXPR_BINOP_INST(/, divides)
+CV_CUDEV_EXPR_BINOP_INST(%, modulus)
+
+CV_CUDEV_EXPR_BINOP_INST(==, equal_to)
+CV_CUDEV_EXPR_BINOP_INST(!=, not_equal_to)
+CV_CUDEV_EXPR_BINOP_INST(>, greater)
+CV_CUDEV_EXPR_BINOP_INST(<, less)
+CV_CUDEV_EXPR_BINOP_INST(>=, greater_equal)
+CV_CUDEV_EXPR_BINOP_INST(<=, less_equal)
+
+CV_CUDEV_EXPR_BINOP_INST(&&, logical_and)
+CV_CUDEV_EXPR_BINOP_INST(||, logical_or)
+
+CV_CUDEV_EXPR_BINOP_INST(&, bit_and)
+CV_CUDEV_EXPR_BINOP_INST(|, bit_or)
+CV_CUDEV_EXPR_BINOP_INST(^, bit_xor)
+CV_CUDEV_EXPR_BINOP_INST(<<, bit_lshift)
+CV_CUDEV_EXPR_BINOP_INST(>>, bit_rshift)
+
+#undef CV_CUDEV_EXPR_BINOP_INST
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/color.hpp b/modules/cudev/include/opencv2/cudev/expr/color.hpp
new file mode 100644
index 0000000000..13f07c15a4
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/color.hpp
@@ -0,0 +1,282 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_COLOR_HPP__
+#define __OPENCV_CUDEV_EXPR_COLOR_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../functional/color_cvt.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+#define CV_CUDEV_EXPR_CVTCOLOR_INST(name) \
+    template <class SrcPtr> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, name ## _func<typename VecTraits<typename PtrTraits<SrcPtr>::value_type>::elem_type> > > \
+    name ## _(const SrcPtr& src) \
+    { \
+        return makeExpr(transformPtr(src, name ## _func<typename VecTraits<typename PtrTraits<SrcPtr>::value_type>::elem_type>())); \
+    }
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_RGBA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_GRAY)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_GRAY)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_GRAY)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_GRAY)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(GRAY_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(GRAY_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_YUV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_YUV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_YUV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_YUV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_YUV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_YUV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_YUV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_YUV4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YUV4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_YCrCb)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_YCrCb)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_YCrCb4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_YCrCb4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_YCrCb)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_YCrCb)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_YCrCb4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_YCrCb4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(YCrCb4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_XYZ)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_XYZ)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_XYZ4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_XYZ4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_XYZ)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_XYZ)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_XYZ4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_XYZ4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(XYZ4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HSV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HSV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HSV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HSV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HSV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HSV)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HSV4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HSV4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HSV_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HSV_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HSV4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HSV4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HSV_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HSV_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HSV4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HSV4_FULL)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_RGB_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_RGBA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_RGB_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_RGBA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_BGR_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV_to_BGRA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_BGR_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HSV4_to_BGRA_FULL)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HLS)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HLS)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HLS4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HLS4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HLS)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HLS)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HLS4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HLS4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HLS_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HLS_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_HLS4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_HLS4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HLS_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HLS_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_HLS4_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_HLS4_FULL)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_RGB_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_RGBA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_RGB_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_RGBA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_BGR_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS_to_BGRA_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_BGR_FULL)
+CV_CUDEV_EXPR_CVTCOLOR_INST(HLS4_to_BGRA_FULL)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_Lab4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGB_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGBA_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGB_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGBA_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGR_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGRA_to_Lab)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGR_to_Lab4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGRA_to_Lab4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_LRGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_LRGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_LRGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_LRGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_LBGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_LBGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab_to_LBGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Lab4_to_LBGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGB_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(RGBA_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGR_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(BGRA_to_Luv4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGB_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGBA_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGB_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LRGBA_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGR_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGRA_to_Luv)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGR_to_Luv4)
+CV_CUDEV_EXPR_CVTCOLOR_INST(LBGRA_to_Luv4)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_RGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_RGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_BGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_BGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_BGRA)
+
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_LRGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_LRGB)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_LRGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_LRGBA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_LBGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_LBGR)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv_to_LBGRA)
+CV_CUDEV_EXPR_CVTCOLOR_INST(Luv4_to_LBGRA)
+
+#undef CV_CUDEV_EXPR_CVTCOLOR_INST
+
+}}
+
+#endif
diff --git a/modules/gpu/src/cuvid_video_source.h b/modules/cudev/include/opencv2/cudev/expr/deriv.hpp
similarity index 58%
rename from modules/gpu/src/cuvid_video_source.h
rename to modules/cudev/include/opencv2/cudev/expr/deriv.hpp
index 1c4c0e5e00..822a86b9a3 100644
--- a/modules/gpu/src/cuvid_video_source.h
+++ b/modules/cudev/include/opencv2/cudev/expr/deriv.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,51 +41,81 @@
 //
 //M*/
 
-#ifndef __CUVUD_VIDEO_SOURCE_H__
-#define __CUVUD_VIDEO_SOURCE_H__
+#pragma once
 
-#include "precomp.hpp"
+#ifndef __OPENCV_CUDEV_EXPR_DERIV_HPP__
+#define __OPENCV_CUDEV_EXPR_DERIV_HPP__
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/deriv.hpp"
+#include "expr.hpp"
 
-namespace cv { namespace gpu
+namespace cv { namespace cudev {
+
+// derivX
+
+template <class SrcPtr>
+__host__ Expr<DerivXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+derivX_(const SrcPtr& src)
 {
-    namespace detail
-    {
-        class CuvidVideoSource : public VideoReader_GPU::VideoSource
-        {
-        public:
-            explicit CuvidVideoSource(const std::string& fname);
-            ~CuvidVideoSource() { cuvidDestroyVideoSource(videoSource_); }
+    return makeExpr(derivXPtr(src));
+}
 
-            VideoReader_GPU::FormatInfo format() const;
-            void start();
-            void stop();
-            bool isStarted() const;
-            bool hasError() const;
+// derivY
 
-        private:
-            CuvidVideoSource(const CuvidVideoSource&);
-            CuvidVideoSource& operator =(const CuvidVideoSource&);
+template <class SrcPtr>
+__host__ Expr<DerivYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+derivY_(const SrcPtr& src)
+{
+    return makeExpr(derivYPtr(src));
+}
 
-            // Callback for handling packages of demuxed video data.
-            //
-            // Parameters:
-            //      pUserData - Pointer to user data. We must pass a pointer to a
-            //          VideoSourceData struct here, that contains a valid CUvideoparser
-            //          and FrameQueue.
-            //      pPacket - video-source data packet.
-            //
-            // NOTE: called from a different thread that doesn't not have a cuda context
-            //
-            static int CUDAAPI HandleVideoData(void* pUserData, CUVIDSOURCEDATAPACKET* pPacket);
+// sobelX
+
+template <class SrcPtr>
+__host__ Expr<SobelXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+sobelX_(const SrcPtr& src)
+{
+    return makeExpr(sobelXPtr(src));
+}
+
+// sobelY
+
+template <class SrcPtr>
+__host__ Expr<SobelYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+sobelY_(const SrcPtr& src)
+{
+    return makeExpr(sobelYPtr(src));
+}
+
+// scharrX
+
+template <class SrcPtr>
+__host__ Expr<ScharrXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+scharrX_(const SrcPtr& src)
+{
+    return makeExpr(scharrXPtr(src));
+}
+
+// scharrY
+
+template <class SrcPtr>
+__host__ Expr<ScharrYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+scharrY_(const SrcPtr& src)
+{
+    return makeExpr(scharrYPtr(src));
+}
+
+// laplacian
+
+template <int ksize, class SrcPtr>
+__host__ Expr<LaplacianPtrSz<ksize, typename PtrTraits<SrcPtr>::ptr_type> >
+laplacian_(const SrcPtr& src)
+{
+    return makeExpr(laplacianPtr<ksize>(src));
+}
 
-            CUvideosource videoSource_;
-            VideoReader_GPU::FormatInfo format_;
-        };
-    }
 }}
 
-#endif // defined(HAVE_CUDA) && !defined(__APPLE__)
-
-#endif // __CUVUD_VIDEO_SOURCE_H__
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/expr.hpp b/modules/cudev/include/opencv2/cudev/expr/expr.hpp
new file mode 100644
index 0000000000..46c780b4a2
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/expr.hpp
@@ -0,0 +1,92 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_EXPR_HPP__
+#define __OPENCV_CUDEV_EXPR_EXPR_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Body> struct Expr
+{
+    Body body;
+};
+
+template <class Body>
+__host__ Expr<Body> makeExpr(const Body& body)
+{
+    Expr<Body> e;
+    e.body = body;
+    return e;
+}
+
+template <class Body> struct PtrTraits< Expr<Body> >
+{
+    typedef Expr<Body>                         ptr_sz_type;
+    typedef typename PtrTraits<Body>::ptr_type ptr_type;
+
+    typedef typename ptr_type::value_type value_type;
+
+    __host__ static ptr_type shrinkPtr(const Expr<Body>& expr)
+    {
+        return PtrTraits<Body>::shrinkPtr(expr.body);
+    }
+
+    __host__ static int getRows(const Expr<Body>& expr)
+    {
+        return PtrTraits<Body>::getRows(expr.body);
+    }
+
+    __host__ static int getCols(const Expr<Body>& expr)
+    {
+        return PtrTraits<Body>::getCols(expr.body);
+    }
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/per_element_func.hpp b/modules/cudev/include/opencv2/cudev/expr/per_element_func.hpp
new file mode 100644
index 0000000000..56a067de9d
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/per_element_func.hpp
@@ -0,0 +1,132 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_PER_ELEMENT_FUNC_HPP__
+#define __OPENCV_CUDEV_EXPR_PER_ELEMENT_FUNC_HPP__
+
+#include "../common.hpp"
+#include "../util/type_traits.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../ptr2d/lut.hpp"
+#include "../functional/functional.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+// min/max
+
+template <class SrcPtr1, class SrcPtr2>
+__host__ Expr<BinaryTransformPtrSz<typename PtrTraits<SrcPtr1>::ptr_type, typename PtrTraits<SrcPtr2>::ptr_type, minimum<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type> > >
+min_(const SrcPtr1& src1, const SrcPtr2& src2)
+{
+    return makeExpr(transformPtr(src1, src2, minimum<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type>()));
+}
+
+template <class SrcPtr1, class SrcPtr2>
+__host__ Expr<BinaryTransformPtrSz<typename PtrTraits<SrcPtr1>::ptr_type, typename PtrTraits<SrcPtr2>::ptr_type, maximum<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type> > >
+max_(const SrcPtr1& src1, const SrcPtr2& src2)
+{
+    return makeExpr(transformPtr(src1, src2, maximum<typename LargerType<typename PtrTraits<SrcPtr1>::value_type, typename PtrTraits<SrcPtr2>::value_type>::type>()));
+}
+
+// threshold
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, ThreshBinaryFunc<typename PtrTraits<SrcPtr>::value_type> > >
+threshBinary_(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type thresh, typename PtrTraits<SrcPtr>::value_type maxVal)
+{
+    return makeExpr(transformPtr(src, thresh_binary_func(thresh, maxVal)));
+}
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, ThreshBinaryInvFunc<typename PtrTraits<SrcPtr>::value_type> > >
+threshBinaryInv_(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type thresh, typename PtrTraits<SrcPtr>::value_type maxVal)
+{
+    return makeExpr(transformPtr(src, thresh_binary_inv_func(thresh, maxVal)));
+}
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, ThreshTruncFunc<typename PtrTraits<SrcPtr>::value_type> > >
+threshTrunc_(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type thresh)
+{
+    return makeExpr(transformPtr(src, thresh_trunc_func(thresh)));
+}
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, ThreshToZeroFunc<typename PtrTraits<SrcPtr>::value_type> > >
+threshToZero_(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type thresh)
+{
+    return makeExpr(transformPtr(src, thresh_to_zero_func(thresh)));
+}
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, ThreshToZeroInvFunc<typename PtrTraits<SrcPtr>::value_type> > >
+threshToZeroInv_(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type thresh)
+{
+    return makeExpr(transformPtr(src, thresh_to_zero_inv_func(thresh)));
+}
+
+// cvt
+
+template <typename D, class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, saturate_cast_func<typename PtrTraits<SrcPtr>::value_type, D> > >
+cvt_(const SrcPtr& src)
+{
+    return makeExpr(transformPtr(src, saturate_cast_func<typename PtrTraits<SrcPtr>::value_type, D>()));
+}
+
+// lut
+
+template <class SrcPtr, class TablePtr>
+__host__ Expr<LutPtrSz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<TablePtr>::ptr_type> >
+lut_(const SrcPtr& src, const TablePtr& tbl)
+{
+    return makeExpr(lutPtr(src, tbl));
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/reduction.hpp b/modules/cudev/include/opencv2/cudev/expr/reduction.hpp
new file mode 100644
index 0000000000..0fdde02e53
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/reduction.hpp
@@ -0,0 +1,259 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_REDUCTION_HPP__
+#define __OPENCV_CUDEV_EXPR_REDUCTION_HPP__
+
+#include "../common.hpp"
+#include "../grid/glob_reduce.hpp"
+#include "../grid/histogram.hpp"
+#include "../grid/integral.hpp"
+#include "../grid/reduce_to_vec.hpp"
+#include "../ptr2d/traits.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+// sum
+
+template <class SrcPtr> struct SumExprBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCalcSum(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<SumExprBody<SrcPtr> >
+sum_(const SrcPtr& src)
+{
+    SumExprBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// minVal
+
+template <class SrcPtr> struct FindMinValExprBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridFindMinVal(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<FindMinValExprBody<SrcPtr> >
+minVal_(const SrcPtr& src)
+{
+    FindMinValExprBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// maxVal
+
+template <class SrcPtr> struct FindMaxValExprBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridFindMaxVal(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<FindMaxValExprBody<SrcPtr> >
+maxVal_(const SrcPtr& src)
+{
+    FindMaxValExprBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// minMaxVal
+
+template <class SrcPtr> struct FindMinMaxValExprBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridFindMinMaxVal(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<FindMinMaxValExprBody<SrcPtr> >
+minMaxVal_(const SrcPtr& src)
+{
+    FindMinMaxValExprBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// countNonZero
+
+template <class SrcPtr> struct CountNonZeroExprBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCountNonZero(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<CountNonZeroExprBody<SrcPtr> >
+countNonZero_(const SrcPtr& src)
+{
+    CountNonZeroExprBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// reduceToRow
+
+template <class Reductor, class SrcPtr> struct ReduceToRowBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridReduceToRow<Reductor>(src, dst, stream);
+    }
+};
+
+template <class Reductor, class SrcPtr>
+__host__ Expr<ReduceToRowBody<Reductor, SrcPtr> >
+reduceToRow_(const SrcPtr& src)
+{
+    ReduceToRowBody<Reductor, SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// reduceToColumn
+
+template <class Reductor, class SrcPtr> struct ReduceToColumnBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridReduceToColumn<Reductor>(src, dst, stream);
+    }
+};
+
+template <class Reductor, class SrcPtr>
+__host__ Expr<ReduceToColumnBody<Reductor, SrcPtr> >
+reduceToColumn_(const SrcPtr& src)
+{
+    ReduceToColumnBody<Reductor, SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// histogram
+
+template <int BIN_COUNT, class SrcPtr> struct HistogramBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridHistogram<BIN_COUNT>(src, dst, stream);
+    }
+};
+
+template <int BIN_COUNT, class SrcPtr>
+__host__ Expr<HistogramBody<BIN_COUNT, SrcPtr> >
+histogram_(const SrcPtr& src)
+{
+    HistogramBody<BIN_COUNT, SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// integral
+
+template <class SrcPtr> struct IntegralBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridIntegral(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<IntegralBody<SrcPtr> >
+integral_(const SrcPtr& src)
+{
+    IntegralBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/unary_func.hpp b/modules/cudev/include/opencv2/cudev/expr/unary_func.hpp
new file mode 100644
index 0000000000..a30f6a6f3c
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/unary_func.hpp
@@ -0,0 +1,98 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_UNARY_FUNC_HPP__
+#define __OPENCV_CUDEV_EXPR_UNARY_FUNC_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../functional/functional.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+#define CV_CUDEV_EXPR_UNARY_FUNC(name) \
+    template <class SrcPtr> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, name ## _func<typename PtrTraits<SrcPtr>::value_type> > > \
+    name ## _(const SrcPtr& src) \
+    { \
+        return makeExpr(transformPtr(src, name ## _func<typename PtrTraits<SrcPtr>::value_type>())); \
+    }
+
+CV_CUDEV_EXPR_UNARY_FUNC(abs)
+CV_CUDEV_EXPR_UNARY_FUNC(sqr)
+CV_CUDEV_EXPR_UNARY_FUNC(sqrt)
+CV_CUDEV_EXPR_UNARY_FUNC(exp)
+CV_CUDEV_EXPR_UNARY_FUNC(exp2)
+CV_CUDEV_EXPR_UNARY_FUNC(exp10)
+CV_CUDEV_EXPR_UNARY_FUNC(log)
+CV_CUDEV_EXPR_UNARY_FUNC(log2)
+CV_CUDEV_EXPR_UNARY_FUNC(log10)
+CV_CUDEV_EXPR_UNARY_FUNC(sin)
+CV_CUDEV_EXPR_UNARY_FUNC(cos)
+CV_CUDEV_EXPR_UNARY_FUNC(tan)
+CV_CUDEV_EXPR_UNARY_FUNC(asin)
+CV_CUDEV_EXPR_UNARY_FUNC(acos)
+CV_CUDEV_EXPR_UNARY_FUNC(atan)
+CV_CUDEV_EXPR_UNARY_FUNC(sinh)
+CV_CUDEV_EXPR_UNARY_FUNC(cosh)
+CV_CUDEV_EXPR_UNARY_FUNC(tanh)
+CV_CUDEV_EXPR_UNARY_FUNC(asinh)
+CV_CUDEV_EXPR_UNARY_FUNC(acosh)
+CV_CUDEV_EXPR_UNARY_FUNC(atanh)
+
+#undef CV_CUDEV_EXPR_UNARY_FUNC
+
+template <class SrcPtr>
+__host__ Expr<UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, Binder2nd<pow_func<typename PtrTraits<SrcPtr>::value_type> > > >
+pow_(const SrcPtr& src, float power)
+{
+    return makeExpr(transformPtr(src, bind2nd(pow_func<typename PtrTraits<SrcPtr>::value_type>(), power)));
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/unary_op.hpp b/modules/cudev/include/opencv2/cudev/expr/unary_op.hpp
new file mode 100644
index 0000000000..905013e42f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/unary_op.hpp
@@ -0,0 +1,94 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_UNARY_OP_HPP__
+#define __OPENCV_CUDEV_EXPR_UNARY_OP_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/transform.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/texture.hpp"
+#include "../ptr2d/glob.hpp"
+#include "../functional/functional.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+#define CV_CUDEV_EXPR_UNOP_INST(op, functor) \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GpuMat_<T> >::ptr_type, functor<T> > > \
+    operator op(const GpuMat_<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<GlobPtrSz<T> >::ptr_type, functor<T> > > \
+    operator op(const GlobPtrSz<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, functor<T>())); \
+    } \
+    template <typename T> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Texture<T> >::ptr_type, functor<T> > > \
+    operator op(const Texture<T>& src) \
+    { \
+        return makeExpr(transformPtr(src, functor<T>())); \
+    } \
+    template <class Body> \
+    __host__ Expr<UnaryTransformPtrSz<typename PtrTraits<Body>::ptr_type, functor<typename Body::value_type> > > \
+    operator op(const Expr<Body>& src) \
+    { \
+        return makeExpr(transformPtr(src.body, functor<typename Body::value_type>())); \
+    }
+
+CV_CUDEV_EXPR_UNOP_INST(-, negate)
+CV_CUDEV_EXPR_UNOP_INST(!, logical_not)
+CV_CUDEV_EXPR_UNOP_INST(~, bit_not)
+
+#undef CV_CUDEV_EXPR_UNOP_INST
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/expr/warping.hpp b/modules/cudev/include/opencv2/cudev/expr/warping.hpp
new file mode 100644
index 0000000000..f942a3fb6f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/expr/warping.hpp
@@ -0,0 +1,171 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_EXPR_WARPING_HPP__
+#define __OPENCV_CUDEV_EXPR_WARPING_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/resize.hpp"
+#include "../ptr2d/remap.hpp"
+#include "../ptr2d/warping.hpp"
+#include "../grid/pyramids.hpp"
+#include "../grid/transpose.hpp"
+#include "expr.hpp"
+
+namespace cv { namespace cudev {
+
+// resize
+
+template <class SrcPtr>
+__host__ Expr<ResizePtrSz<typename PtrTraits<SrcPtr>::ptr_type> >
+resize_(const SrcPtr& src, float fx, float fy)
+{
+    return makeExpr(resizePtr(src, fx, fy));
+}
+
+// remap
+
+template <class SrcPtr, class MapPtr>
+__host__ Expr<RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapPtr>::ptr_type> >
+remap_(const SrcPtr& src, const MapPtr& map)
+{
+    return makeExpr(remapPtr(src, map));
+}
+
+template <class SrcPtr, class MapXPtr, class MapYPtr>
+__host__ Expr<RemapPtr2Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapXPtr>::ptr_type, typename PtrTraits<MapYPtr>::ptr_type> >
+remap_(const SrcPtr& src, const MapXPtr& mapx, const MapYPtr& mapy)
+{
+    return makeExpr(remapPtr(src, mapx, mapy));
+}
+
+// warpAffine
+
+template <class SrcPtr>
+__host__ Expr<RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, AffineMapPtr> >
+warpAffine_(const SrcPtr& src, Size dstSize, const GpuMat_<float>& warpMat)
+{
+    return makeExpr(warpAffinePtr(src, dstSize, warpMat));
+}
+
+// warpPerspective
+
+template <class SrcPtr>
+__host__ Expr<RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, PerspectiveMapPtr> >
+warpPerspective_(const SrcPtr& src, Size dstSize, const GpuMat_<float>& warpMat)
+{
+    return makeExpr(warpPerspectivePtr(src, dstSize, warpMat));
+}
+
+// pyrDown
+
+template <class SrcPtr> struct PyrDownBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridPyrDown(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<PyrDownBody<SrcPtr> >
+pyrDown_(const SrcPtr& src)
+{
+    PyrDownBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// pyrUp
+
+template <class SrcPtr> struct PyrUpBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridPyrUp(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<PyrUpBody<SrcPtr> >
+pyrUp_(const SrcPtr& src)
+{
+    PyrUpBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+// transpose
+
+template <class SrcPtr> struct TransposeBody
+{
+    SrcPtr src;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridTranspose(src, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ Expr<TransposeBody<SrcPtr> >
+transpose_(const SrcPtr& src)
+{
+    TransposeBody<SrcPtr> body;
+    body.src = src;
+    return makeExpr(body);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/functional/color_cvt.hpp b/modules/cudev/include/opencv2/cudev/functional/color_cvt.hpp
new file mode 100644
index 0000000000..8be854780a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/functional/color_cvt.hpp
@@ -0,0 +1,474 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_FUNCTIONAL_COLOR_CVT_HPP__
+#define __OPENCV_CUDEV_FUNCTIONAL_COLOR_CVT_HPP__
+
+#include "../common.hpp"
+#include "detail/color_cvt.hpp"
+
+namespace cv { namespace cudev {
+
+// Various 3/4-channel to 3/4-channel RGB transformations
+
+#define CV_CUDEV_RGB2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2RGB<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_RGB2RGB_INST(BGR_to_RGB, 3, 3, 2)
+CV_CUDEV_RGB2RGB_INST(BGR_to_BGRA, 3, 4, 0)
+CV_CUDEV_RGB2RGB_INST(BGR_to_RGBA, 3, 4, 2)
+CV_CUDEV_RGB2RGB_INST(BGRA_to_BGR, 4, 3, 0)
+CV_CUDEV_RGB2RGB_INST(BGRA_to_RGB, 4, 3, 2)
+CV_CUDEV_RGB2RGB_INST(BGRA_to_RGBA, 4, 4, 2)
+
+#undef CV_CUDEV_RGB2RGB_INST
+
+// RGB to Grayscale
+
+#define CV_CUDEV_RGB2GRAY_INST(name, scn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2Gray<SrcDepth, scn, bidx> \
+    { \
+    };
+
+CV_CUDEV_RGB2GRAY_INST(RGB_to_GRAY, 3, 2)
+CV_CUDEV_RGB2GRAY_INST(BGR_to_GRAY, 3, 0)
+CV_CUDEV_RGB2GRAY_INST(RGBA_to_GRAY, 4, 2)
+CV_CUDEV_RGB2GRAY_INST(BGRA_to_GRAY, 4, 0)
+
+#undef CV_CUDEV_RGB2GRAY_INST
+
+// Grayscale to RGB
+
+#define CV_CUDEV_GRAY2RGB_INST(name, dcn) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::Gray2RGB<SrcDepth, dcn> \
+    { \
+    };
+
+CV_CUDEV_GRAY2RGB_INST(GRAY_to_BGR, 3)
+CV_CUDEV_GRAY2RGB_INST(GRAY_to_BGRA, 4)
+
+#undef CV_CUDEV_GRAY2RGB_INST
+
+// RGB to YUV
+
+#define CV_CUDEV_RGB2YUV_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2YUV<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_RGB2YUV_INST(RGB_to_YUV, 3, 3, 2)
+CV_CUDEV_RGB2YUV_INST(RGBA_to_YUV, 4, 3, 2)
+CV_CUDEV_RGB2YUV_INST(RGB_to_YUV4, 3, 4, 2)
+CV_CUDEV_RGB2YUV_INST(RGBA_to_YUV4, 4, 4, 2)
+CV_CUDEV_RGB2YUV_INST(BGR_to_YUV, 3, 3, 0)
+CV_CUDEV_RGB2YUV_INST(BGRA_to_YUV, 4, 3, 0)
+CV_CUDEV_RGB2YUV_INST(BGR_to_YUV4, 3, 4, 0)
+CV_CUDEV_RGB2YUV_INST(BGRA_to_YUV4, 4, 4, 0)
+
+#undef CV_CUDEV_RGB2YUV_INST
+
+// YUV to RGB
+
+#define CV_CUDEV_YUV2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::YUV2RGB<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_YUV2RGB_INST(YUV_to_RGB, 3, 3, 2)
+CV_CUDEV_YUV2RGB_INST(YUV_to_RGBA, 3, 4, 2)
+CV_CUDEV_YUV2RGB_INST(YUV4_to_RGB, 4, 3, 2)
+CV_CUDEV_YUV2RGB_INST(YUV4_to_RGBA, 4, 4, 2)
+CV_CUDEV_YUV2RGB_INST(YUV_to_BGR, 3, 3, 0)
+CV_CUDEV_YUV2RGB_INST(YUV_to_BGRA, 3, 4, 0)
+CV_CUDEV_YUV2RGB_INST(YUV4_to_BGR, 4, 3, 0)
+CV_CUDEV_YUV2RGB_INST(YUV4_to_BGRA, 4, 4, 0)
+
+#undef CV_CUDEV_YUV2RGB_INST
+
+// RGB to YCrCb
+
+#define CV_CUDEV_RGB2YCrCb_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2YCrCb<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_RGB2YCrCb_INST(RGB_to_YCrCb, 3, 3, 2)
+CV_CUDEV_RGB2YCrCb_INST(RGBA_to_YCrCb, 4, 3, 2)
+CV_CUDEV_RGB2YCrCb_INST(RGB_to_YCrCb4, 3, 4, 2)
+CV_CUDEV_RGB2YCrCb_INST(RGBA_to_YCrCb4, 4, 4, 2)
+CV_CUDEV_RGB2YCrCb_INST(BGR_to_YCrCb, 3, 3, 0)
+CV_CUDEV_RGB2YCrCb_INST(BGRA_to_YCrCb, 4, 3, 0)
+CV_CUDEV_RGB2YCrCb_INST(BGR_to_YCrCb4, 3, 4, 0)
+CV_CUDEV_RGB2YCrCb_INST(BGRA_to_YCrCb4, 4, 4, 0)
+
+#undef CV_CUDEV_RGB2YCrCb_INST
+
+// YCrCb to RGB
+
+#define CV_CUDEV_YCrCb2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::YCrCb2RGB<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_YCrCb2RGB_INST(YCrCb_to_RGB, 3, 3, 2)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb_to_RGBA, 3, 4, 2)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb4_to_RGB, 4, 3, 2)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb4_to_RGBA, 4, 4, 2)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb_to_BGR, 3, 3, 0)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb_to_BGRA, 3, 4, 0)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb4_to_BGR, 4, 3, 0)
+CV_CUDEV_YCrCb2RGB_INST(YCrCb4_to_BGRA, 4, 4, 0)
+
+#undef CV_CUDEV_YCrCb2RGB_INST
+
+// RGB to XYZ
+
+#define CV_CUDEV_RGB2XYZ_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2XYZ<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_RGB2XYZ_INST(RGB_to_XYZ, 3, 3, 2)
+CV_CUDEV_RGB2XYZ_INST(RGBA_to_XYZ, 4, 3, 2)
+CV_CUDEV_RGB2XYZ_INST(RGB_to_XYZ4, 3, 4, 2)
+CV_CUDEV_RGB2XYZ_INST(RGBA_to_XYZ4, 4, 4, 2)
+CV_CUDEV_RGB2XYZ_INST(BGR_to_XYZ, 3, 3, 0)
+CV_CUDEV_RGB2XYZ_INST(BGRA_to_XYZ, 4, 3, 0)
+CV_CUDEV_RGB2XYZ_INST(BGR_to_XYZ4, 3, 4, 0)
+CV_CUDEV_RGB2XYZ_INST(BGRA_to_XYZ4, 4, 4, 0)
+
+#undef CV_CUDEV_RGB2XYZ_INST
+
+// XYZ to RGB
+
+#define CV_CUDEV_XYZ2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::XYZ2RGB<SrcDepth, scn, dcn, bidx> \
+    { \
+    };
+
+CV_CUDEV_XYZ2RGB_INST(XYZ_to_RGB, 3, 3, 2)
+CV_CUDEV_XYZ2RGB_INST(XYZ4_to_RGB, 4, 3, 2)
+CV_CUDEV_XYZ2RGB_INST(XYZ_to_RGBA, 3, 4, 2)
+CV_CUDEV_XYZ2RGB_INST(XYZ4_to_RGBA, 4, 4, 2)
+CV_CUDEV_XYZ2RGB_INST(XYZ_to_BGR, 3, 3, 0)
+CV_CUDEV_XYZ2RGB_INST(XYZ4_to_BGR, 4, 3, 0)
+CV_CUDEV_XYZ2RGB_INST(XYZ_to_BGRA, 3, 4, 0)
+CV_CUDEV_XYZ2RGB_INST(XYZ4_to_BGRA, 4, 4, 0)
+
+#undef CV_CUDEV_XYZ2RGB_INST
+
+// RGB to HSV
+
+#define CV_CUDEV_RGB2HSV_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2HSV<SrcDepth, scn, dcn, bidx, 180> \
+    { \
+    }; \
+    template <typename SrcDepth> struct name ## _FULL ## _func : cv::cudev::color_cvt_detail::RGB2HSV<SrcDepth, scn, dcn, bidx, 256> \
+    { \
+    }; \
+    template <> struct name ## _func<float> : cv::cudev::color_cvt_detail::RGB2HSV<float, scn, dcn, bidx, 360> \
+    { \
+    }; \
+    template <> struct name ## _FULL ## _func<float> : cv::cudev::color_cvt_detail::RGB2HSV<float, scn, dcn, bidx, 360> \
+    { \
+    };
+
+CV_CUDEV_RGB2HSV_INST(RGB_to_HSV, 3, 3, 2)
+CV_CUDEV_RGB2HSV_INST(RGBA_to_HSV, 4, 3, 2)
+CV_CUDEV_RGB2HSV_INST(RGB_to_HSV4, 3, 4, 2)
+CV_CUDEV_RGB2HSV_INST(RGBA_to_HSV4, 4, 4, 2)
+CV_CUDEV_RGB2HSV_INST(BGR_to_HSV, 3, 3, 0)
+CV_CUDEV_RGB2HSV_INST(BGRA_to_HSV, 4, 3, 0)
+CV_CUDEV_RGB2HSV_INST(BGR_to_HSV4, 3, 4, 0)
+CV_CUDEV_RGB2HSV_INST(BGRA_to_HSV4, 4, 4, 0)
+
+#undef CV_CUDEV_RGB2HSV_INST
+
+// HSV to RGB
+
+#define CV_CUDEV_HSV2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::HSV2RGB<SrcDepth, scn, dcn, bidx, 180> \
+    { \
+    }; \
+    template <typename SrcDepth> struct name ## _FULL ## _func : cv::cudev::color_cvt_detail::HSV2RGB<SrcDepth, scn, dcn, bidx, 255> \
+    { \
+    }; \
+    template <> struct name ## _func<float> : cv::cudev::color_cvt_detail::HSV2RGB<float, scn, dcn, bidx, 360> \
+    { \
+    }; \
+    template <> struct name ## _FULL ## _func<float> : cv::cudev::color_cvt_detail::HSV2RGB<float, scn, dcn, bidx, 360> \
+    { \
+    };
+
+CV_CUDEV_HSV2RGB_INST(HSV_to_RGB, 3, 3, 2)
+CV_CUDEV_HSV2RGB_INST(HSV_to_RGBA, 3, 4, 2)
+CV_CUDEV_HSV2RGB_INST(HSV4_to_RGB, 4, 3, 2)
+CV_CUDEV_HSV2RGB_INST(HSV4_to_RGBA, 4, 4, 2)
+CV_CUDEV_HSV2RGB_INST(HSV_to_BGR, 3, 3, 0)
+CV_CUDEV_HSV2RGB_INST(HSV_to_BGRA, 3, 4, 0)
+CV_CUDEV_HSV2RGB_INST(HSV4_to_BGR, 4, 3, 0)
+CV_CUDEV_HSV2RGB_INST(HSV4_to_BGRA, 4, 4, 0)
+
+#undef CV_CUDEV_HSV2RGB_INST
+
+// RGB to HLS
+
+#define CV_CUDEV_RGB2HLS_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2HLS<SrcDepth, scn, dcn, bidx, 180> \
+    { \
+    }; \
+    template <typename SrcDepth> struct name ## _FULL ## _func : cv::cudev::color_cvt_detail::RGB2HLS<SrcDepth, scn, dcn, bidx, 256> \
+    { \
+    }; \
+    template <> struct name ## _func<float> : cv::cudev::color_cvt_detail::RGB2HLS<float, scn, dcn, bidx, 360> \
+    { \
+    }; \
+    template <> struct name ## _FULL ## _func<float> : cv::cudev::color_cvt_detail::RGB2HLS<float, scn, dcn, bidx, 360> \
+    { \
+    };
+
+CV_CUDEV_RGB2HLS_INST(RGB_to_HLS, 3, 3, 2)
+CV_CUDEV_RGB2HLS_INST(RGBA_to_HLS, 4, 3, 2)
+CV_CUDEV_RGB2HLS_INST(RGB_to_HLS4, 3, 4, 2)
+CV_CUDEV_RGB2HLS_INST(RGBA_to_HLS4, 4, 4, 2)
+CV_CUDEV_RGB2HLS_INST(BGR_to_HLS, 3, 3, 0)
+CV_CUDEV_RGB2HLS_INST(BGRA_to_HLS, 4, 3, 0)
+CV_CUDEV_RGB2HLS_INST(BGR_to_HLS4, 3, 4, 0)
+CV_CUDEV_RGB2HLS_INST(BGRA_to_HLS4, 4, 4, 0)
+
+#undef CV_CUDEV_RGB2HLS_INST
+
+// HLS to RGB
+
+#define CV_CUDEV_HLS2RGB_INST(name, scn, dcn, bidx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::HLS2RGB<SrcDepth, scn, dcn, bidx, 180> \
+    { \
+    }; \
+    template <typename SrcDepth> struct name ## _FULL ## _func : cv::cudev::color_cvt_detail::HLS2RGB<SrcDepth, scn, dcn, bidx, 255> \
+    { \
+    }; \
+    template <> struct name ## _func<float> : cv::cudev::color_cvt_detail::HLS2RGB<float, scn, dcn, bidx, 360> \
+    { \
+    }; \
+    template <> struct name ## _FULL ## _func<float> : cv::cudev::color_cvt_detail::HLS2RGB<float, scn, dcn, bidx, 360> \
+    { \
+    };
+
+CV_CUDEV_HLS2RGB_INST(HLS_to_RGB, 3, 3, 2)
+CV_CUDEV_HLS2RGB_INST(HLS_to_RGBA, 3, 4, 2)
+CV_CUDEV_HLS2RGB_INST(HLS4_to_RGB, 4, 3, 2)
+CV_CUDEV_HLS2RGB_INST(HLS4_to_RGBA, 4, 4, 2)
+CV_CUDEV_HLS2RGB_INST(HLS_to_BGR, 3, 3, 0)
+CV_CUDEV_HLS2RGB_INST(HLS_to_BGRA, 3, 4, 0)
+CV_CUDEV_HLS2RGB_INST(HLS4_to_BGR, 4, 3, 0)
+CV_CUDEV_HLS2RGB_INST(HLS4_to_BGRA, 4, 4, 0)
+
+#undef CV_CUDEV_HLS2RGB_INST
+
+// RGB to Lab
+
+#define CV_CUDEV_RGB2Lab_INST(name, scn, dcn, sRGB, blueIdx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2Lab<SrcDepth, scn, dcn, sRGB, blueIdx> \
+    { \
+    };
+
+CV_CUDEV_RGB2Lab_INST(RGB_to_Lab, 3, 3, true, 2)
+CV_CUDEV_RGB2Lab_INST(RGBA_to_Lab, 4, 3, true, 2)
+CV_CUDEV_RGB2Lab_INST(RGB_to_Lab4, 3, 4, true, 2)
+CV_CUDEV_RGB2Lab_INST(RGBA_to_Lab4, 4, 4, true, 2)
+CV_CUDEV_RGB2Lab_INST(BGR_to_Lab, 3, 3, true, 0)
+CV_CUDEV_RGB2Lab_INST(BGRA_to_Lab, 4, 3, true, 0)
+CV_CUDEV_RGB2Lab_INST(BGR_to_Lab4, 3, 4, true, 0)
+CV_CUDEV_RGB2Lab_INST(BGRA_to_Lab4, 4, 4, true, 0)
+
+CV_CUDEV_RGB2Lab_INST(LRGB_to_Lab, 3, 3, false, 2)
+CV_CUDEV_RGB2Lab_INST(LRGBA_to_Lab, 4, 3, false, 2)
+CV_CUDEV_RGB2Lab_INST(LRGB_to_Lab4, 3, 4, false, 2)
+CV_CUDEV_RGB2Lab_INST(LRGBA_to_Lab4, 4, 4, false, 2)
+CV_CUDEV_RGB2Lab_INST(LBGR_to_Lab, 3, 3, false, 0)
+CV_CUDEV_RGB2Lab_INST(LBGRA_to_Lab, 4, 3, false, 0)
+CV_CUDEV_RGB2Lab_INST(LBGR_to_Lab4, 3, 4, false, 0)
+CV_CUDEV_RGB2Lab_INST(LBGRA_to_Lab4, 4, 4, false, 0)
+
+#undef CV_CUDEV_RGB2Lab_INST
+
+// Lab to RGB
+
+#define CV_CUDEV_Lab2RGB_INST(name, scn, dcn, sRGB, blueIdx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::Lab2RGB<SrcDepth, scn, dcn, sRGB, blueIdx> \
+    { \
+    };
+
+CV_CUDEV_Lab2RGB_INST(Lab_to_RGB, 3, 3, true, 2)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_RGB, 4, 3, true, 2)
+CV_CUDEV_Lab2RGB_INST(Lab_to_RGBA, 3, 4, true, 2)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_RGBA, 4, 4, true, 2)
+CV_CUDEV_Lab2RGB_INST(Lab_to_BGR, 3, 3, true, 0)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_BGR, 4, 3, true, 0)
+CV_CUDEV_Lab2RGB_INST(Lab_to_BGRA, 3, 4, true, 0)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_BGRA, 4, 4, true, 0)
+
+CV_CUDEV_Lab2RGB_INST(Lab_to_LRGB, 3, 3, false, 2)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_LRGB, 4, 3, false, 2)
+CV_CUDEV_Lab2RGB_INST(Lab_to_LRGBA, 3, 4, false, 2)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_LRGBA, 4, 4, false, 2)
+CV_CUDEV_Lab2RGB_INST(Lab_to_LBGR, 3, 3, false, 0)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_LBGR, 4, 3, false, 0)
+CV_CUDEV_Lab2RGB_INST(Lab_to_LBGRA, 3, 4, false, 0)
+CV_CUDEV_Lab2RGB_INST(Lab4_to_LBGRA, 4, 4, false, 0)
+
+#undef CV_CUDEV_Lab2RGB_INST
+
+// RGB to Luv
+
+#define CV_CUDEV_RGB2Luv_INST(name, scn, dcn, sRGB, blueIdx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::RGB2Luv<SrcDepth, scn, dcn, sRGB, blueIdx> \
+    { \
+    };
+
+CV_CUDEV_RGB2Luv_INST(RGB_to_Luv, 3, 3, true, 2)
+CV_CUDEV_RGB2Luv_INST(RGBA_to_Luv, 4, 3, true, 2)
+CV_CUDEV_RGB2Luv_INST(RGB_to_Luv4, 3, 4, true, 2)
+CV_CUDEV_RGB2Luv_INST(RGBA_to_Luv4, 4, 4, true, 2)
+CV_CUDEV_RGB2Luv_INST(BGR_to_Luv, 3, 3, true, 0)
+CV_CUDEV_RGB2Luv_INST(BGRA_to_Luv, 4, 3, true, 0)
+CV_CUDEV_RGB2Luv_INST(BGR_to_Luv4, 3, 4, true, 0)
+CV_CUDEV_RGB2Luv_INST(BGRA_to_Luv4, 4, 4, true, 0)
+
+CV_CUDEV_RGB2Luv_INST(LRGB_to_Luv, 3, 3, false, 2)
+CV_CUDEV_RGB2Luv_INST(LRGBA_to_Luv, 4, 3, false, 2)
+CV_CUDEV_RGB2Luv_INST(LRGB_to_Luv4, 3, 4, false, 2)
+CV_CUDEV_RGB2Luv_INST(LRGBA_to_Luv4, 4, 4, false, 2)
+CV_CUDEV_RGB2Luv_INST(LBGR_to_Luv, 3, 3, false, 0)
+CV_CUDEV_RGB2Luv_INST(LBGRA_to_Luv, 4, 3, false, 0)
+CV_CUDEV_RGB2Luv_INST(LBGR_to_Luv4, 3, 4, false, 0)
+CV_CUDEV_RGB2Luv_INST(LBGRA_to_Luv4, 4, 4, false, 0)
+
+#undef CV_CUDEV_RGB2Luv_INST
+
+// Luv to RGB
+
+#define CV_CUDEV_Luv2RGB_INST(name, scn, dcn, sRGB, blueIdx) \
+    template <typename SrcDepth> struct name ## _func : cv::cudev::color_cvt_detail::Luv2RGB<SrcDepth, scn, dcn, sRGB, blueIdx> \
+    { \
+    };
+
+CV_CUDEV_Luv2RGB_INST(Luv_to_RGB, 3, 3, true, 2)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_RGB, 4, 3, true, 2)
+CV_CUDEV_Luv2RGB_INST(Luv_to_RGBA, 3, 4, true, 2)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_RGBA, 4, 4, true, 2)
+CV_CUDEV_Luv2RGB_INST(Luv_to_BGR, 3, 3, true, 0)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_BGR, 4, 3, true, 0)
+CV_CUDEV_Luv2RGB_INST(Luv_to_BGRA, 3, 4, true, 0)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_BGRA, 4, 4, true, 0)
+
+CV_CUDEV_Luv2RGB_INST(Luv_to_LRGB, 3, 3, false, 2)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_LRGB, 4, 3, false, 2)
+CV_CUDEV_Luv2RGB_INST(Luv_to_LRGBA, 3, 4, false, 2)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_LRGBA, 4, 4, false, 2)
+CV_CUDEV_Luv2RGB_INST(Luv_to_LBGR, 3, 3, false, 0)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_LBGR, 4, 3, false, 0)
+CV_CUDEV_Luv2RGB_INST(Luv_to_LBGRA, 3, 4, false, 0)
+CV_CUDEV_Luv2RGB_INST(Luv4_to_LBGRA, 4, 4, false, 0)
+
+#undef CV_CUDEV_Luv2RGB_INST
+
+// 24/32-bit RGB to 16-bit (565 or 555) RGB
+
+#define CV_CUDEV_RGB2RGB5x5_INST(name, scn, bidx, green_bits) \
+    typedef cv::cudev::color_cvt_detail::RGB2RGB5x5<scn, bidx, green_bits> name ## _func;
+
+CV_CUDEV_RGB2RGB5x5_INST(BGR_to_BGR555, 3, 0, 5)
+CV_CUDEV_RGB2RGB5x5_INST(BGR_to_BGR565, 3, 0, 6)
+CV_CUDEV_RGB2RGB5x5_INST(RGB_to_BGR555, 3, 2, 5)
+CV_CUDEV_RGB2RGB5x5_INST(RGB_to_BGR565, 3, 2, 6)
+CV_CUDEV_RGB2RGB5x5_INST(BGRA_to_BGR555, 4, 0, 5)
+CV_CUDEV_RGB2RGB5x5_INST(BGRA_to_BGR565, 4, 0, 6)
+CV_CUDEV_RGB2RGB5x5_INST(RGBA_to_BGR555, 4, 2, 5)
+CV_CUDEV_RGB2RGB5x5_INST(RGBA_to_BGR565, 4, 2, 6)
+
+#undef CV_CUDEV_RGB2RGB5x5_INST
+
+// 16-bit (565 or 555) RGB to 24/32-bit RGB
+
+#define CV_CUDEV_RGB5x52RGB_INST(name, dcn, bidx, green_bits) \
+    typedef cv::cudev::color_cvt_detail::RGB5x52RGB<dcn, bidx, green_bits> name ## _func;
+
+CV_CUDEV_RGB5x52RGB_INST(BGR555_to_RGB, 3, 2, 5)
+CV_CUDEV_RGB5x52RGB_INST(BGR565_to_RGB, 3, 2, 6)
+CV_CUDEV_RGB5x52RGB_INST(BGR555_to_BGR, 3, 0, 5)
+CV_CUDEV_RGB5x52RGB_INST(BGR565_to_BGR, 3, 0, 6)
+CV_CUDEV_RGB5x52RGB_INST(BGR555_to_RGBA, 4, 2, 5)
+CV_CUDEV_RGB5x52RGB_INST(BGR565_to_RGBA, 4, 2, 6)
+CV_CUDEV_RGB5x52RGB_INST(BGR555_to_BGRA, 4, 0, 5)
+CV_CUDEV_RGB5x52RGB_INST(BGR565_to_BGRA, 4, 0, 6)
+
+#undef CV_CUDEV_RGB5x52RGB_INST
+
+// Grayscale to 16-bit (565 or 555) RGB
+
+#define CV_CUDEV_GRAY2RGB5x5_INST(name, green_bits) \
+    typedef cv::cudev::color_cvt_detail::Gray2RGB5x5<green_bits> name ## _func;
+
+CV_CUDEV_GRAY2RGB5x5_INST(GRAY_to_BGR555, 5)
+CV_CUDEV_GRAY2RGB5x5_INST(GRAY_to_BGR565, 6)
+
+#undef CV_CUDEV_GRAY2RGB5x5_INST
+
+// 16-bit (565 or 555) RGB to Grayscale
+
+#define CV_CUDEV_RGB5x52GRAY_INST(name, green_bits) \
+    typedef cv::cudev::color_cvt_detail::RGB5x52Gray<green_bits> name ## _func;
+
+CV_CUDEV_RGB5x52GRAY_INST(BGR555_to_GRAY, 5)
+CV_CUDEV_RGB5x52GRAY_INST(BGR565_to_GRAY, 6)
+
+#undef CV_CUDEV_RGB5x52GRAY_INST
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/functional/detail/color_cvt.hpp b/modules/cudev/include/opencv2/cudev/functional/detail/color_cvt.hpp
new file mode 100644
index 0000000000..9b20fcc2c6
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/functional/detail/color_cvt.hpp
@@ -0,0 +1,1279 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_FUNCTIONAL_COLOR_CVT_DETAIL_HPP__
+#define __OPENCV_CUDEV_FUNCTIONAL_COLOR_CVT_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/limits.hpp"
+#include "../functional.hpp"
+
+namespace cv { namespace cudev {
+
+namespace color_cvt_detail
+{
+    // utility
+
+    #define CV_CUDEV_DESCALE(x, n) (((x) + (1 << ((n)-1))) >> (n))
+
+    template <typename T> struct ColorChannel
+    {
+        __device__ __forceinline__ static T max() { return numeric_limits<T>::max(); }
+        __device__ __forceinline__ static T half() { return (T)(max()/2 + 1); }
+    };
+
+    template <> struct ColorChannel<float>
+    {
+        __device__ __forceinline__ static float max() { return 1.f; }
+        __device__ __forceinline__ static float half() { return 0.5f; }
+    };
+
+    template <typename T> __device__ __forceinline__ void setAlpha(typename MakeVec<T, 3>::type& vec, T val)
+    {
+    }
+
+    template <typename T> __device__ __forceinline__ void setAlpha(typename MakeVec<T, 4>::type& vec, T val)
+    {
+        vec.w = val;
+    }
+
+    template <typename T> __device__ __forceinline__ T getAlpha(const typename MakeVec<T, 3>::type& vec)
+    {
+        return ColorChannel<T>::max();
+    }
+
+    template <typename T> __device__ __forceinline__ T getAlpha(const typename MakeVec<T, 4>::type& vec)
+    {
+        return vec.w;
+    }
+
+    enum
+    {
+        yuv_shift  = 14,
+        xyz_shift  = 12,
+        R2Y        = 4899,
+        G2Y        = 9617,
+        B2Y        = 1868,
+        BLOCK_SIZE = 256
+    };
+
+    // Various 3/4-channel to 3/4-channel RGB transformations
+
+    template <typename T, int scn, int dcn, int bidx> struct RGB2RGB
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = bidx == 0 ? src.x : src.z;
+            dst.y = src.y;
+            dst.z = bidx == 0 ? src.z : src.x;
+            setAlpha(dst, getAlpha<T>(src));
+
+            return dst;
+        }
+    };
+
+    // 24/32-bit RGB to 16-bit (565 or 555) RGB
+
+    template <int scn, int bidx, int green_bits> struct RGB2RGB5x5;
+
+    template <int scn, int bidx> struct RGB2RGB5x5<scn, bidx, 6>
+            : unary_function<typename MakeVec<uchar, scn>::type, ushort>
+    {
+        __device__ ushort operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+            return (ushort) ((b >> 3) | ((g & ~3) << 3) | ((r & ~7) << 8));
+        }
+    };
+
+    template <int bidx> struct RGB2RGB5x5<3, bidx, 5>
+            : unary_function<uchar3, ushort>
+    {
+        __device__ ushort operator ()(const uchar3& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+            return (ushort) ((b >> 3) | ((g & ~7) << 2) | ((r & ~7) << 7));
+        }
+    };
+
+    template <int bidx> struct RGB2RGB5x5<4, bidx, 5>
+            : unary_function<uchar4, ushort>
+    {
+        __device__ ushort operator ()(const uchar4& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+            const int a = src.w;
+            return (ushort) ((b >> 3) | ((g & ~7) << 2) | ((r & ~7) << 7) | (a * 0x8000));
+        }
+    };
+
+    // 16-bit (565 or 555) RGB to 24/32-bit RGB
+
+    template <int dcn, int bidx, int green_bits> struct RGB5x52RGB;
+
+    template <int bidx> struct RGB5x52RGB<3, bidx, 5>
+            : unary_function<ushort, uchar3>
+    {
+        __device__ uchar3 operator ()(ushort src) const
+        {
+            const int b = src << 3;
+            const int r = (src >> 7) & ~7;
+
+            uchar3 dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = (src >> 2) & ~7;
+            dst.z = bidx == 0 ? r : b;
+
+            return dst;
+        }
+    };
+
+    template <int bidx> struct RGB5x52RGB<4, bidx, 5>
+            : unary_function<ushort, uchar4>
+    {
+        __device__ uchar4 operator ()(ushort src) const
+        {
+            const int b = src << 3;
+            const int r = (src >> 7) & ~7;
+
+            uchar4 dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = (src >> 2) & ~7;
+            dst.z = bidx == 0 ? r : b;
+            dst.w = (src & 0x8000) * 0xffu;
+
+            return dst;
+        }
+    };
+
+    template <int bidx> struct RGB5x52RGB<3, bidx, 6>
+            : unary_function<ushort, uchar3>
+    {
+        __device__ uchar3 operator ()(ushort src) const
+        {
+            const int b = src << 3;
+            const int r = (src >> 8) & ~7;
+
+            uchar3 dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = (src >> 3) & ~3;
+            dst.z = bidx == 0 ? r : b;
+
+            return dst;
+        }
+    };
+
+    template <int bidx> struct RGB5x52RGB<4, bidx, 6>
+            : unary_function<ushort, uchar4>
+    {
+        __device__ uchar4 operator ()(ushort src) const
+        {
+            const int b = src << 3;
+            const int r = (src >> 8) & ~7;
+
+            uchar4 dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = (src >> 3) & ~3;
+            dst.z = bidx == 0 ? r : b;
+            dst.w = 255;
+
+            return dst;
+        }
+    };
+
+    // Grayscale to RGB
+
+    template <typename T, int dcn> struct Gray2RGB
+            : unary_function<T, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(T src) const
+        {
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.z = dst.y = dst.x = src;
+            setAlpha(dst, ColorChannel<T>::max());
+
+            return dst;
+        }
+    };
+
+    // Grayscale to 16-bit (565 or 555) RGB
+
+    template <int green_bits> struct Gray2RGB5x5;
+
+    template <> struct Gray2RGB5x5<5>
+            : unary_function<uchar, ushort>
+    {
+        __device__ ushort operator ()(uchar src) const
+        {
+            return (ushort) (src | (src << 5) | (src << 10));
+        }
+    };
+
+    template <> struct Gray2RGB5x5<6>
+            : unary_function<uchar, ushort>
+    {
+        __device__ ushort operator ()(uchar src) const
+        {
+            return (ushort) ((src >> 3) | ((src & ~3) << 3) | ((src & ~7) << 8));
+        }
+    };
+
+    // 16-bit (565 or 555) RGB to Grayscale
+
+    template <int green_bits> struct RGB5x52Gray;
+
+    template <> struct RGB5x52Gray<5>
+            : unary_function<ushort, uchar>
+    {
+        __device__ uchar operator ()(ushort src) const
+        {
+            return (uchar) CV_CUDEV_DESCALE(((src << 3) & 0xf8) * B2Y + ((src >> 2) & 0xf8) * G2Y + ((src >> 7) & 0xf8) * R2Y, yuv_shift);
+        }
+    };
+
+    template <> struct RGB5x52Gray<6>
+            : unary_function<ushort, uchar>
+    {
+        __device__ uchar operator ()(ushort src) const
+        {
+            return (uchar) CV_CUDEV_DESCALE(((src << 3) & 0xf8) * B2Y + ((src >> 3) & 0xfc) * G2Y + ((src >> 8) & 0xf8) * R2Y, yuv_shift);
+        }
+    };
+
+    // RGB to Grayscale
+
+    template <typename T, int scn, int bidx> struct RGB2Gray
+            : unary_function<typename MakeVec<T, scn>::type, T>
+    {
+        __device__ T operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+            return (T) CV_CUDEV_DESCALE(b * B2Y + g * G2Y + r * R2Y, yuv_shift);
+        }
+    };
+
+    template <int scn, int bidx> struct RGB2Gray<float, scn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, float>
+    {
+        __device__ float operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+            return b * 0.114f + g * 0.587f + r * 0.299f;
+        }
+    };
+
+    // RGB to YUV
+
+    __constant__ float c_RGB2YUVCoeffs_f[5] = { 0.114f, 0.587f, 0.299f, 0.492f, 0.877f };
+    __constant__ int   c_RGB2YUVCoeffs_i[5] = { B2Y, G2Y, R2Y, 8061, 14369 };
+
+    template <typename T, int scn, int dcn, int bidx> struct RGB2YUV
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+
+            const int delta = ColorChannel<T>::half() * (1 << yuv_shift);
+
+            const int Y = CV_CUDEV_DESCALE(b * c_RGB2YUVCoeffs_i[2] + g * c_RGB2YUVCoeffs_i[1] + r * c_RGB2YUVCoeffs_i[0], yuv_shift);
+            const int Cr = CV_CUDEV_DESCALE((r - Y) * c_RGB2YUVCoeffs_i[3] + delta, yuv_shift);
+            const int Cb = CV_CUDEV_DESCALE((b - Y) * c_RGB2YUVCoeffs_i[4] + delta, yuv_shift);
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = saturate_cast<T>(Y);
+            dst.y = saturate_cast<T>(Cr);
+            dst.z = saturate_cast<T>(Cb);
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct RGB2YUV<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = b * c_RGB2YUVCoeffs_f[2] + g * c_RGB2YUVCoeffs_f[1] + r * c_RGB2YUVCoeffs_f[0];
+            dst.y = (r - dst.x) * c_RGB2YUVCoeffs_f[3] + ColorChannel<float>::half();
+            dst.z = (b - dst.x) * c_RGB2YUVCoeffs_f[4] + ColorChannel<float>::half();
+
+            return dst;
+        }
+    };
+
+    // YUV to RGB
+
+    __constant__ float c_YUV2RGBCoeffs_f[5] = { 2.032f, -0.395f, -0.581f, 1.140f };
+    __constant__ int   c_YUV2RGBCoeffs_i[5] = { 33292, -6472, -9519, 18678 };
+
+    template <typename T, int scn, int dcn, int bidx> struct YUV2RGB
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = src.x + CV_CUDEV_DESCALE((src.z - ColorChannel<T>::half()) * c_YUV2RGBCoeffs_i[3], yuv_shift);
+            const int g = src.x + CV_CUDEV_DESCALE((src.z - ColorChannel<T>::half()) * c_YUV2RGBCoeffs_i[2] + (src.y - ColorChannel<T>::half()) * c_YUV2RGBCoeffs_i[1], yuv_shift);
+            const int r = src.x + CV_CUDEV_DESCALE((src.y - ColorChannel<T>::half()) * c_YUV2RGBCoeffs_i[0], yuv_shift);
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = saturate_cast<T>(bidx == 0 ? b : r);
+            dst.y = saturate_cast<T>(g);
+            dst.z = saturate_cast<T>(bidx == 0 ? r : b);
+            setAlpha(dst, ColorChannel<T>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct YUV2RGB<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = src.x + (src.z - ColorChannel<float>::half()) * c_YUV2RGBCoeffs_f[3];
+            const float g = src.x + (src.z - ColorChannel<float>::half()) * c_YUV2RGBCoeffs_f[2] + (src.y - ColorChannel<float>::half()) * c_YUV2RGBCoeffs_f[1];
+            const float r = src.x + (src.y - ColorChannel<float>::half()) * c_YUV2RGBCoeffs_f[0];
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = g;
+            dst.z = bidx == 0 ? r : b;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to YCrCb
+
+    __constant__ float c_RGB2YCrCbCoeffs_f[5] = { 0.299f, 0.587f, 0.114f, 0.713f, 0.564f };
+    __constant__ int   c_RGB2YCrCbCoeffs_i[5] = { R2Y, G2Y, B2Y, 11682, 9241 };
+
+    template <typename T, int scn, int dcn, int bidx> struct RGB2YCrCb
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+
+            const int delta = ColorChannel<T>::half() * (1 << yuv_shift);
+
+            const int Y = CV_CUDEV_DESCALE(b * c_RGB2YCrCbCoeffs_i[2] + g * c_RGB2YCrCbCoeffs_i[1] + r * c_RGB2YCrCbCoeffs_i[0], yuv_shift);
+            const int Cr = CV_CUDEV_DESCALE((r - Y) * c_RGB2YCrCbCoeffs_i[3] + delta, yuv_shift);
+            const int Cb = CV_CUDEV_DESCALE((b - Y) * c_RGB2YCrCbCoeffs_i[4] + delta, yuv_shift);
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = saturate_cast<T>(Y);
+            dst.y = saturate_cast<T>(Cr);
+            dst.z = saturate_cast<T>(Cb);
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct RGB2YCrCb<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = b * c_RGB2YCrCbCoeffs_f[2] + g * c_RGB2YCrCbCoeffs_f[1] + r * c_RGB2YCrCbCoeffs_f[0];
+            dst.y = (r - dst.x) * c_RGB2YCrCbCoeffs_f[3] + ColorChannel<float>::half();
+            dst.z = (b - dst.x) * c_RGB2YCrCbCoeffs_f[4] + ColorChannel<float>::half();
+
+            return dst;
+        }
+    };
+
+    // YCrCb to RGB
+
+    __constant__ float c_YCrCb2RGBCoeffs_f[5] = {1.403f, -0.714f, -0.344f, 1.773f};
+    __constant__ int   c_YCrCb2RGBCoeffs_i[5] = {22987, -11698, -5636, 29049};
+
+    template <typename T, int scn, int dcn, int bidx> struct YCrCb2RGB
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = src.x + CV_CUDEV_DESCALE((src.z - ColorChannel<T>::half()) * c_YCrCb2RGBCoeffs_i[3], yuv_shift);
+            const int g = src.x + CV_CUDEV_DESCALE((src.z - ColorChannel<T>::half()) * c_YCrCb2RGBCoeffs_i[2] + (src.y - ColorChannel<T>::half()) * c_YCrCb2RGBCoeffs_i[1], yuv_shift);
+            const int r = src.x + CV_CUDEV_DESCALE((src.y - ColorChannel<T>::half()) * c_YCrCb2RGBCoeffs_i[0], yuv_shift);
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = saturate_cast<T>(bidx == 0 ? b : r);
+            dst.y = saturate_cast<T>(g);
+            dst.z = saturate_cast<T>(bidx == 0 ? r : b);
+            setAlpha(dst, ColorChannel<T>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct YCrCb2RGB<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = src.x + (src.z - ColorChannel<float>::half()) * c_YCrCb2RGBCoeffs_f[3];
+            const float g = src.x + (src.z - ColorChannel<float>::half()) * c_YCrCb2RGBCoeffs_f[2] + (src.y - ColorChannel<float>::half()) * c_YCrCb2RGBCoeffs_f[1];
+            const float r = src.x + (src.y - ColorChannel<float>::half()) * c_YCrCb2RGBCoeffs_f[0];
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = g;
+            dst.z = bidx == 0 ? r : b;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to XYZ
+
+    __constant__ float c_RGB2XYZ_D65f[9] = { 0.412453f, 0.357580f, 0.180423f, 0.212671f, 0.715160f, 0.072169f, 0.019334f, 0.119193f, 0.950227f };
+    __constant__ int   c_RGB2XYZ_D65i[9] = { 1689, 1465, 739, 871, 2929, 296, 79, 488, 3892 };
+
+    template <typename T, int scn, int dcn, int bidx> struct RGB2XYZ
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.z = saturate_cast<T>(CV_CUDEV_DESCALE(r * c_RGB2XYZ_D65i[6] + g * c_RGB2XYZ_D65i[7] + b * c_RGB2XYZ_D65i[8], xyz_shift));
+            dst.x = saturate_cast<T>(CV_CUDEV_DESCALE(r * c_RGB2XYZ_D65i[0] + g * c_RGB2XYZ_D65i[1] + b * c_RGB2XYZ_D65i[2], xyz_shift));
+            dst.y = saturate_cast<T>(CV_CUDEV_DESCALE(r * c_RGB2XYZ_D65i[3] + g * c_RGB2XYZ_D65i[4] + b * c_RGB2XYZ_D65i[5], xyz_shift));
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct RGB2XYZ<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = r * c_RGB2XYZ_D65f[0] + g * c_RGB2XYZ_D65f[1] + b * c_RGB2XYZ_D65f[2];
+            dst.y = r * c_RGB2XYZ_D65f[3] + g * c_RGB2XYZ_D65f[4] + b * c_RGB2XYZ_D65f[5];
+            dst.z = r * c_RGB2XYZ_D65f[6] + g * c_RGB2XYZ_D65f[7] + b * c_RGB2XYZ_D65f[8];
+
+            return dst;
+        }
+    };
+
+    // XYZ to RGB
+
+    __constant__ float c_XYZ2sRGB_D65f[9] = { 3.240479f, -1.53715f, -0.498535f, -0.969256f, 1.875991f, 0.041556f, 0.055648f, -0.204043f, 1.057311f };
+    __constant__ int   c_XYZ2sRGB_D65i[9] = { 13273, -6296, -2042, -3970, 7684, 170, 228, -836, 4331 };
+
+    template <typename T, int scn, int dcn, int bidx> struct XYZ2RGB
+            : unary_function<typename MakeVec<T, scn>::type, typename MakeVec<T, dcn>::type>
+    {
+        __device__ typename MakeVec<T, dcn>::type operator ()(const typename MakeVec<T, scn>::type& src) const
+        {
+            const int b = CV_CUDEV_DESCALE(src.x * c_XYZ2sRGB_D65i[6] + src.y * c_XYZ2sRGB_D65i[7] + src.z * c_XYZ2sRGB_D65i[8], xyz_shift);
+            const int g = CV_CUDEV_DESCALE(src.x * c_XYZ2sRGB_D65i[3] + src.y * c_XYZ2sRGB_D65i[4] + src.z * c_XYZ2sRGB_D65i[5], xyz_shift);
+            const int r = CV_CUDEV_DESCALE(src.x * c_XYZ2sRGB_D65i[0] + src.y * c_XYZ2sRGB_D65i[1] + src.z * c_XYZ2sRGB_D65i[2], xyz_shift);
+
+            typename MakeVec<T, dcn>::type dst;
+
+            dst.x = saturate_cast<T>(bidx == 0 ? b : r);
+            dst.y = saturate_cast<T>(g);
+            dst.z = saturate_cast<T>(bidx == 0 ? r : b);
+            setAlpha(dst, ColorChannel<T>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx> struct XYZ2RGB<float, scn, dcn, bidx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float b = src.x * c_XYZ2sRGB_D65f[6] + src.y * c_XYZ2sRGB_D65f[7] + src.z * c_XYZ2sRGB_D65f[8];
+            const float g = src.x * c_XYZ2sRGB_D65f[3] + src.y * c_XYZ2sRGB_D65f[4] + src.z * c_XYZ2sRGB_D65f[5];
+            const float r = src.x * c_XYZ2sRGB_D65f[0] + src.y * c_XYZ2sRGB_D65f[1] + src.z * c_XYZ2sRGB_D65f[2];
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = g;
+            dst.z = bidx == 0 ? r : b;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to HSV
+
+    __constant__ int c_HsvDivTable   [256] = {0, 1044480, 522240, 348160, 261120, 208896, 174080, 149211, 130560, 116053, 104448, 94953, 87040, 80345, 74606, 69632, 65280, 61440, 58027, 54973, 52224, 49737, 47476, 45412, 43520, 41779, 40172, 38684, 37303, 36017, 34816, 33693, 32640, 31651, 30720, 29842, 29013, 28229, 27486, 26782, 26112, 25475, 24869, 24290, 23738, 23211, 22706, 22223, 21760, 21316, 20890, 20480, 20086, 19707, 19342, 18991, 18651, 18324, 18008, 17703, 17408, 17123, 16846, 16579, 16320, 16069, 15825, 15589, 15360, 15137, 14921, 14711, 14507, 14308, 14115, 13926, 13743, 13565, 13391, 13221, 13056, 12895, 12738, 12584, 12434, 12288, 12145, 12006, 11869, 11736, 11605, 11478, 11353, 11231, 11111, 10995, 10880, 10768, 10658, 10550, 10445, 10341, 10240, 10141, 10043, 9947, 9854, 9761, 9671, 9582, 9495, 9410, 9326, 9243, 9162, 9082, 9004, 8927, 8852, 8777, 8704, 8632, 8561, 8492, 8423, 8356, 8290, 8224, 8160, 8097, 8034, 7973, 7913, 7853, 7795, 7737, 7680, 7624, 7569, 7514, 7461, 7408, 7355, 7304, 7253, 7203, 7154, 7105, 7057, 7010, 6963, 6917, 6872, 6827, 6782, 6739, 6695, 6653, 6611, 6569, 6528, 6487, 6447, 6408, 6369, 6330, 6292, 6254, 6217, 6180, 6144, 6108, 6073, 6037, 6003, 5968, 5935, 5901, 5868, 5835, 5803, 5771, 5739, 5708, 5677, 5646, 5615, 5585, 5556, 5526, 5497, 5468, 5440, 5412, 5384, 5356, 5329, 5302, 5275, 5249, 5222, 5196, 5171, 5145, 5120, 5095, 5070, 5046, 5022, 4998, 4974, 4950, 4927, 4904, 4881, 4858, 4836, 4813, 4791, 4769, 4748, 4726, 4705, 4684, 4663, 4642, 4622, 4601, 4581, 4561, 4541, 4522, 4502, 4483, 4464, 4445, 4426, 4407, 4389, 4370, 4352, 4334, 4316, 4298, 4281, 4263, 4246, 4229, 4212, 4195, 4178, 4161, 4145, 4128, 4112, 4096};
+    __constant__ int c_HsvDivTable180[256] = {0, 122880, 61440, 40960, 30720, 24576, 20480, 17554, 15360, 13653, 12288, 11171, 10240, 9452, 8777, 8192, 7680, 7228, 6827, 6467, 6144, 5851, 5585, 5343, 5120, 4915, 4726, 4551, 4389, 4237, 4096, 3964, 3840, 3724, 3614, 3511, 3413, 3321, 3234, 3151, 3072, 2997, 2926, 2858, 2793, 2731, 2671, 2614, 2560, 2508, 2458, 2409, 2363, 2318, 2276, 2234, 2194, 2156, 2119, 2083, 2048, 2014, 1982, 1950, 1920, 1890, 1862, 1834, 1807, 1781, 1755, 1731, 1707, 1683, 1661, 1638, 1617, 1596, 1575, 1555, 1536, 1517, 1499, 1480, 1463, 1446, 1429, 1412, 1396, 1381, 1365, 1350, 1336, 1321, 1307, 1293, 1280, 1267, 1254, 1241, 1229, 1217, 1205, 1193, 1182, 1170, 1159, 1148, 1138, 1127, 1117, 1107, 1097, 1087, 1078, 1069, 1059, 1050, 1041, 1033, 1024, 1016, 1007, 999, 991, 983, 975, 968, 960, 953, 945, 938, 931, 924, 917, 910, 904, 897, 890, 884, 878, 871, 865, 859, 853, 847, 842, 836, 830, 825, 819, 814, 808, 803, 798, 793, 788, 783, 778, 773, 768, 763, 759, 754, 749, 745, 740, 736, 731, 727, 723, 719, 714, 710, 706, 702, 698, 694, 690, 686, 683, 679, 675, 671, 668, 664, 661, 657, 654, 650, 647, 643, 640, 637, 633, 630, 627, 624, 621, 617, 614, 611, 608, 605, 602, 599, 597, 594, 591, 588, 585, 582, 580, 577, 574, 572, 569, 566, 564, 561, 559, 556, 554, 551, 549, 546, 544, 541, 539, 537, 534, 532, 530, 527, 525, 523, 521, 518, 516, 514, 512, 510, 508, 506, 504, 502, 500, 497, 495, 493, 492, 490, 488, 486, 484, 482};
+    __constant__ int c_HsvDivTable256[256] = {0, 174763, 87381, 58254, 43691, 34953, 29127, 24966, 21845, 19418, 17476, 15888, 14564, 13443, 12483, 11651, 10923, 10280, 9709, 9198, 8738, 8322, 7944, 7598, 7282, 6991, 6722, 6473, 6242, 6026, 5825, 5638, 5461, 5296, 5140, 4993, 4855, 4723, 4599, 4481, 4369, 4263, 4161, 4064, 3972, 3884, 3799, 3718, 3641, 3567, 3495, 3427, 3361, 3297, 3236, 3178, 3121, 3066, 3013, 2962, 2913, 2865, 2819, 2774, 2731, 2689, 2648, 2608, 2570, 2533, 2497, 2461, 2427, 2394, 2362, 2330, 2300, 2270, 2241, 2212, 2185, 2158, 2131, 2106, 2081, 2056, 2032, 2009, 1986, 1964, 1942, 1920, 1900, 1879, 1859, 1840, 1820, 1802, 1783, 1765, 1748, 1730, 1713, 1697, 1680, 1664, 1649, 1633, 1618, 1603, 1589, 1574, 1560, 1547, 1533, 1520, 1507, 1494, 1481, 1469, 1456, 1444, 1432, 1421, 1409, 1398, 1387, 1376, 1365, 1355, 1344, 1334, 1324, 1314, 1304, 1295, 1285, 1276, 1266, 1257, 1248, 1239, 1231, 1222, 1214, 1205, 1197, 1189, 1181, 1173, 1165, 1157, 1150, 1142, 1135, 1128, 1120, 1113, 1106, 1099, 1092, 1085, 1079, 1072, 1066, 1059, 1053, 1046, 1040, 1034, 1028, 1022, 1016, 1010, 1004, 999, 993, 987, 982, 976, 971, 966, 960, 955, 950, 945, 940, 935, 930, 925, 920, 915, 910, 906, 901, 896, 892, 887, 883, 878, 874, 869, 865, 861, 857, 853, 848, 844, 840, 836, 832, 828, 824, 820, 817, 813, 809, 805, 802, 798, 794, 791, 787, 784, 780, 777, 773, 770, 767, 763, 760, 757, 753, 750, 747, 744, 741, 737, 734, 731, 728, 725, 722, 719, 716, 713, 710, 708, 705, 702, 699, 696, 694, 691, 688, 685};
+
+    template <typename T, int scn, int dcn, int bidx, int hr> struct RGB2HSV;
+
+    template <int scn, int dcn, int bidx, int hr> struct RGB2HSV<uchar, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            const int hsv_shift = 12;
+            const int* hdiv_table = hr == 180 ? c_HsvDivTable180 : c_HsvDivTable256;
+
+            const int b = bidx == 0 ? src.x : src.z;
+            const int g = src.y;
+            const int r = bidx == 0 ? src.z : src.x;
+
+            int h, s, v = b;
+            int vmin = b, diff;
+            int vr, vg;
+
+            v = ::max(v, g);
+            v = ::max(v, r);
+            vmin = ::min(vmin, g);
+            vmin = ::min(vmin, r);
+
+            diff = v - vmin;
+            vr = (v == r) * -1;
+            vg = (v == g) * -1;
+
+            s = (diff * c_HsvDivTable[v] + (1 << (hsv_shift-1))) >> hsv_shift;
+            h = (vr & (g - b)) + (~vr & ((vg & (b - r + 2 * diff)) + ((~vg) & (r - g + 4 * diff))));
+            h = (h * hdiv_table[diff] + (1 << (hsv_shift-1))) >> hsv_shift;
+            h += (h < 0) * hr;
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(h);
+            dst.y = saturate_cast<uchar>(s);
+            dst.z = saturate_cast<uchar>(v);
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx, int hr> struct RGB2HSV<float, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float hscale = hr * (1.f / 360.f);
+
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+
+            float h, s, v;
+            float vmin, diff;
+
+            v = vmin = r;
+            v = ::fmax(v, g);
+            v = ::fmax(v, b);
+            vmin = ::fmin(vmin, g);
+            vmin = ::fmin(vmin, b);
+
+            diff = v - vmin;
+            s = diff / (float)(::fabs(v) + numeric_limits<float>::epsilon());
+            diff = (float)(60. / (diff + numeric_limits<float>::epsilon()));
+
+            h  = (v == r) * (g - b) * diff;
+            h += (v != r && v == g) * ((b - r) * diff + 120.f);
+            h += (v != r && v != g) * ((r - g) * diff + 240.f);
+            h += (h < 0) * 360.f;
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = h * hscale;
+            dst.y = s;
+            dst.z = v;
+
+            return dst;
+        }
+    };
+
+    // HSV to RGB
+
+    __constant__ int c_HsvSectorData[6][3] = { {1,3,0}, {1,0,2}, {3,0,1}, {0,2,1}, {0,1,3}, {2,1,0} };
+
+    template <typename T, int scn, int dcn, int bidx, int hr> struct HSV2RGB;
+
+    template <int scn, int dcn, int bidx, int hr> struct HSV2RGB<float, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float hscale = 6.f / hr;
+
+            float h = src.x, s = src.y, v = src.z;
+            float b = v, g = v, r = v;
+
+            if (s != 0)
+            {
+                h *= hscale;
+
+                if( h < 0 )
+                    do h += 6; while( h < 0 );
+                else if( h >= 6 )
+                    do h -= 6; while( h >= 6 );
+
+                int sector = __float2int_rd(h);
+                h -= sector;
+
+                if ( (unsigned)sector >= 6u )
+                {
+                    sector = 0;
+                    h = 0.f;
+                }
+
+                float tab[4];
+                tab[0] = v;
+                tab[1] = v * (1.f - s);
+                tab[2] = v * (1.f - s * h);
+                tab[3] = v * (1.f - s * (1.f - h));
+
+                b = tab[c_HsvSectorData[sector][0]];
+                g = tab[c_HsvSectorData[sector][1]];
+                r = tab[c_HsvSectorData[sector][2]];
+            }
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = g;
+            dst.z = bidx == 0 ? r : b;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx, int hr> struct HSV2RGB<uchar, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x;
+            buf.y = src.y * (1.f / 255.f);
+            buf.z = src.z * (1.f / 255.f);
+
+            HSV2RGB<float, 3, 3, bidx, hr> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x * 255.f);
+            dst.y = saturate_cast<uchar>(buf.y * 255.f);
+            dst.z = saturate_cast<uchar>(buf.z * 255.f);
+            setAlpha(dst, ColorChannel<uchar>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to HLS
+
+    template <typename T, int scn, int dcn, int bidx, int hr> struct RGB2HLS;
+
+    template <int scn, int dcn, int bidx, int hr> struct RGB2HLS<float, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float hscale = hr * (1.f / 360.f);
+
+            const float b = bidx == 0 ? src.x : src.z;
+            const float g = src.y;
+            const float r = bidx == 0 ? src.z : src.x;
+
+            float h = 0.f, s = 0.f, l;
+            float vmin, vmax, diff;
+
+            vmax = vmin = r;
+            vmax = ::fmax(vmax, g);
+            vmax = ::fmax(vmax, b);
+            vmin = ::fmin(vmin, g);
+            vmin = ::fmin(vmin, b);
+
+            diff = vmax - vmin;
+            l = (vmax + vmin) * 0.5f;
+
+            if (diff > numeric_limits<float>::epsilon())
+            {
+                s = (l < 0.5f) * diff / (vmax + vmin);
+                s += (l >= 0.5f) * diff / (2.0f - vmax - vmin);
+
+                diff = 60.f / diff;
+
+                h  = (vmax == r) * (g - b) * diff;
+                h += (vmax != r && vmax == g) * ((b - r) * diff + 120.f);
+                h += (vmax != r && vmax != g) * ((r - g) * diff + 240.f);
+                h += (h < 0.f) * 360.f;
+            }
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = h * hscale;
+            dst.y = l;
+            dst.z = s;
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx, int hr> struct RGB2HLS<uchar, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x * (1.f / 255.f);
+            buf.y = src.y * (1.f / 255.f);
+            buf.z = src.z * (1.f / 255.f);
+
+            RGB2HLS<float, 3, 3, bidx, hr> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x);
+            dst.y = saturate_cast<uchar>(buf.y * 255.f);
+            dst.z = saturate_cast<uchar>(buf.z * 255.f);
+
+            return dst;
+        }
+    };
+
+    // HLS to RGB
+
+    __constant__ int c_HlsSectorData[6][3] = { {1,3,0}, {1,0,2}, {3,0,1}, {0,2,1}, {0,1,3}, {2,1,0} };
+
+    template <typename T, int scn, int dcn, int bidx, int hr> struct HLS2RGB;
+
+    template <int scn, int dcn, int bidx, int hr> struct HLS2RGB<float, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float hscale = 6.0f / hr;
+
+            float h = src.x, l = src.y, s = src.z;
+            float b = l, g = l, r = l;
+
+            if (s != 0)
+            {
+                float p2  = (l <= 0.5f) * l * (1 + s);
+                      p2 += (l > 0.5f) * (l + s - l * s);
+                float p1 = 2 * l - p2;
+
+                h *= hscale;
+
+                if( h < 0 )
+                    do h += 6; while( h < 0 );
+                else if( h >= 6 )
+                    do h -= 6; while( h >= 6 );
+
+                int sector;
+                sector = __float2int_rd(h);
+
+                h -= sector;
+
+                float tab[4];
+                tab[0] = p2;
+                tab[1] = p1;
+                tab[2] = p1 + (p2 - p1) * (1 - h);
+                tab[3] = p1 + (p2 - p1) * h;
+
+                b = tab[c_HlsSectorData[sector][0]];
+                g = tab[c_HlsSectorData[sector][1]];
+                r = tab[c_HlsSectorData[sector][2]];
+            }
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = bidx == 0 ? b : r;
+            dst.y = g;
+            dst.z = bidx == 0 ? r : b;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, int bidx, int hr> struct HLS2RGB<uchar, scn, dcn, bidx, hr>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x;
+            buf.y = src.y * (1.f / 255.f);
+            buf.z = src.z * (1.f / 255.f);
+
+            HLS2RGB<float, 3, 3, bidx, hr> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x * 255.f);
+            dst.y = saturate_cast<uchar>(buf.y * 255.f);
+            dst.z = saturate_cast<uchar>(buf.z * 255.f);
+            setAlpha(dst, ColorChannel<uchar>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to Lab
+
+    enum
+    {
+        LAB_CBRT_TAB_SIZE = 1024,
+        GAMMA_TAB_SIZE = 1024,
+        lab_shift = xyz_shift,
+        gamma_shift = 3,
+        lab_shift2 = (lab_shift + gamma_shift),
+        LAB_CBRT_TAB_SIZE_B = (256 * 3 / 2 * (1 << gamma_shift))
+    };
+
+    __constant__ ushort c_sRGBGammaTab_b[] = {0,1,1,2,2,3,4,4,5,6,6,7,8,8,9,10,11,11,12,13,14,15,16,17,19,20,21,22,24,25,26,28,29,31,33,34,36,38,40,41,43,45,47,49,51,54,56,58,60,63,65,68,70,73,75,78,81,83,86,89,92,95,98,101,105,108,111,115,118,121,125,129,132,136,140,144,147,151,155,160,164,168,172,176,181,185,190,194,199,204,209,213,218,223,228,233,239,244,249,255,260,265,271,277,282,288,294,300,306,312,318,324,331,337,343,350,356,363,370,376,383,390,397,404,411,418,426,433,440,448,455,463,471,478,486,494,502,510,518,527,535,543,552,560,569,578,586,595,604,613,622,631,641,650,659,669,678,688,698,707,717,727,737,747,757,768,778,788,799,809,820,831,842,852,863,875,886,897,908,920,931,943,954,966,978,990,1002,1014,1026,1038,1050,1063,1075,1088,1101,1113,1126,1139,1152,1165,1178,1192,1205,1218,1232,1245,1259,1273,1287,1301,1315,1329,1343,1357,1372,1386,1401,1415,1430,1445,1460,1475,1490,1505,1521,1536,1551,1567,1583,1598,1614,1630,1646,1662,1678,1695,1711,1728,1744,1761,1778,1794,1811,1828,1846,1863,1880,1897,1915,1933,1950,1968,1986,2004,2022,2040};
+    __constant__ float c_sRGBGammaTab[] = {0,7.55853e-05,0.,-7.51331e-13,7.55853e-05,7.55853e-05,-2.25399e-12,3.75665e-12,0.000151171,7.55853e-05,9.01597e-12,-6.99932e-12,0.000226756,7.55853e-05,-1.1982e-11,2.41277e-12,0.000302341,7.55853e-05,-4.74369e-12,1.19001e-11,0.000377927,7.55853e-05,3.09568e-11,-2.09095e-11,0.000453512,7.55853e-05,-3.17718e-11,1.35303e-11,0.000529097,7.55853e-05,8.81905e-12,-4.10782e-12,0.000604683,7.55853e-05,-3.50439e-12,2.90097e-12,0.000680268,7.55853e-05,5.19852e-12,-7.49607e-12,0.000755853,7.55853e-05,-1.72897e-11,2.70833e-11,0.000831439,7.55854e-05,6.39602e-11,-4.26295e-11,0.000907024,7.55854e-05,-6.39282e-11,2.70193e-11,0.000982609,7.55853e-05,1.71298e-11,-7.24017e-12,0.00105819,7.55853e-05,-4.59077e-12,1.94137e-12,0.00113378,7.55853e-05,1.23333e-12,-5.25291e-13,0.00120937,7.55853e-05,-3.42545e-13,1.59799e-13,0.00128495,7.55853e-05,1.36852e-13,-1.13904e-13,0.00136054,7.55853e-05,-2.04861e-13,2.95818e-13,0.00143612,7.55853e-05,6.82594e-13,-1.06937e-12,0.00151171,7.55853e-05,-2.52551e-12,3.98166e-12,0.00158729,7.55853e-05,9.41946e-12,-1.48573e-11,0.00166288,7.55853e-05,-3.51523e-11,5.54474e-11,0.00173846,7.55854e-05,1.3119e-10,-9.0517e-11,0.00181405,7.55854e-05,-1.40361e-10,7.37899e-11,0.00188963,7.55853e-05,8.10085e-11,-8.82272e-11,0.00196522,7.55852e-05,-1.83673e-10,1.62704e-10,0.0020408,7.55853e-05,3.04438e-10,-2.13341e-10,0.00211639,7.55853e-05,-3.35586e-10,2.25e-10,0.00219197,7.55853e-05,3.39414e-10,-2.20997e-10,0.00226756,7.55853e-05,-3.23576e-10,1.93326e-10,0.00234315,7.55853e-05,2.564e-10,-8.66446e-11,0.00241873,7.55855e-05,-3.53328e-12,-7.9578e-11,0.00249432,7.55853e-05,-2.42267e-10,1.72126e-10,0.0025699,7.55853e-05,2.74111e-10,-1.43265e-10,0.00264549,7.55854e-05,-1.55683e-10,-6.47292e-11,0.00272107,7.55849e-05,-3.4987e-10,8.67842e-10,0.00279666,7.55868e-05,2.25366e-09,-3.8723e-09,0.00287224,7.55797e-05,-9.36325e-09,1.5087e-08,0.00294783,7.56063e-05,3.58978e-08,-5.69415e-08,0.00302341,7.55072e-05,-1.34927e-07,2.13144e-07,0.003099,7.58768e-05,5.04507e-07,1.38713e-07,0.00317552,7.7302e-05,9.20646e-07,-1.55186e-07,0.00325359,7.86777e-05,4.55087e-07,4.26813e-08,0.00333276,7.97159e-05,5.83131e-07,-1.06495e-08,0.00341305,8.08502e-05,5.51182e-07,3.87467e-09,0.00349446,8.19642e-05,5.62806e-07,-1.92586e-10,0.00357698,8.30892e-05,5.62228e-07,1.0866e-09,0.00366063,8.4217e-05,5.65488e-07,5.02818e-10,0.00374542,8.53494e-05,5.66997e-07,8.60211e-10,0.00383133,8.6486e-05,5.69577e-07,7.13044e-10,0.00391839,8.76273e-05,5.71716e-07,4.78527e-10,0.00400659,8.87722e-05,5.73152e-07,1.09818e-09,0.00409594,8.99218e-05,5.76447e-07,2.50964e-10,0.00418644,9.10754e-05,5.772e-07,1.15762e-09,0.00427809,9.22333e-05,5.80672e-07,2.40865e-10,0.0043709,9.33954e-05,5.81395e-07,1.13854e-09,0.00446488,9.45616e-05,5.84811e-07,3.27267e-10,0.00456003,9.57322e-05,5.85792e-07,8.1197e-10,0.00465635,9.69062e-05,5.88228e-07,6.15823e-10,0.00475384,9.80845e-05,5.90076e-07,9.15747e-10,0.00485252,9.92674e-05,5.92823e-07,3.778e-10,0.00495238,0.000100454,5.93956e-07,8.32623e-10,0.00505343,0.000101645,5.96454e-07,4.82695e-10,0.00515567,0.000102839,5.97902e-07,9.61904e-10,0.00525911,0.000104038,6.00788e-07,3.26281e-10,0.00536375,0.00010524,6.01767e-07,9.926e-10,0.00546959,0.000106447,6.04745e-07,3.59933e-10,0.00557664,0.000107657,6.05824e-07,8.2728e-10,0.0056849,0.000108871,6.08306e-07,5.21898e-10,0.00579438,0.00011009,6.09872e-07,8.10492e-10,0.00590508,0.000111312,6.12303e-07,4.27046e-10,0.00601701,0.000112538,6.13585e-07,7.40878e-10,0.00613016,0.000113767,6.15807e-07,8.00469e-10,0.00624454,0.000115001,6.18209e-07,2.48178e-10,0.00636016,0.000116238,6.18953e-07,1.00073e-09,0.00647702,0.000117479,6.21955e-07,4.05654e-10,0.00659512,0.000118724,6.23172e-07,6.36192e-10,0.00671447,0.000119973,6.25081e-07,7.74927e-10,0.00683507,0.000121225,6.27406e-07,4.54975e-10,0.00695692,0.000122481,6.28771e-07,6.64841e-10,0.00708003,0.000123741,6.30765e-07,6.10972e-10,0.00720441,0.000125004,6.32598e-07,6.16543e-10,0.00733004,0.000126271,6.34448e-07,6.48204e-10,0.00745695,0.000127542,6.36392e-07,5.15835e-10,0.00758513,0.000128816,6.3794e-07,5.48103e-10,0.00771458,0.000130094,6.39584e-07,1.01706e-09,0.00784532,0.000131376,6.42635e-07,4.0283e-11,0.00797734,0.000132661,6.42756e-07,6.84471e-10,0.00811064,0.000133949,6.4481e-07,9.47144e-10,0.00824524,0.000135241,6.47651e-07,1.83472e-10,0.00838112,0.000136537,6.48201e-07,1.11296e-09,0.00851831,0.000137837,6.5154e-07,2.13163e-11,0.0086568,0.00013914,6.51604e-07,6.64462e-10,0.00879659,0.000140445,6.53598e-07,1.04613e-09,0.00893769,0.000141756,6.56736e-07,-1.92377e-10,0.0090801,0.000143069,6.56159e-07,1.58601e-09,0.00922383,0.000144386,6.60917e-07,-5.63754e-10,0.00936888,0.000145706,6.59226e-07,1.60033e-09,0.00951524,0.000147029,6.64027e-07,-2.49543e-10,0.00966294,0.000148356,6.63278e-07,1.26043e-09,0.00981196,0.000149687,6.67059e-07,-1.35572e-10,0.00996231,0.00015102,6.66653e-07,1.14458e-09,0.010114,0.000152357,6.70086e-07,2.13864e-10,0.010267,0.000153698,6.70728e-07,7.93856e-10,0.0104214,0.000155042,6.73109e-07,3.36077e-10,0.0105771,0.000156389,6.74118e-07,6.55765e-10,0.0107342,0.000157739,6.76085e-07,7.66211e-10,0.0108926,0.000159094,6.78384e-07,4.66116e-12,0.0110524,0.000160451,6.78398e-07,1.07775e-09,0.0112135,0.000161811,6.81631e-07,3.41023e-10,0.011376,0.000163175,6.82654e-07,3.5205e-10,0.0115398,0.000164541,6.8371e-07,1.04473e-09,0.0117051,0.000165912,6.86844e-07,1.25757e-10,0.0118717,0.000167286,6.87222e-07,3.14818e-10,0.0120396,0.000168661,6.88166e-07,1.40886e-09,0.012209,0.000170042,6.92393e-07,-3.62244e-10,0.0123797,0.000171425,6.91306e-07,9.71397e-10,0.0125518,0.000172811,6.9422e-07,2.02003e-10,0.0127253,0.0001742,6.94826e-07,1.01448e-09,0.0129002,0.000175593,6.97869e-07,3.96653e-10,0.0130765,0.00017699,6.99059e-07,1.92927e-10,0.0132542,0.000178388,6.99638e-07,6.94305e-10,0.0134333,0.00017979,7.01721e-07,7.55108e-10,0.0136138,0.000181195,7.03986e-07,1.05918e-11,0.0137957,0.000182603,7.04018e-07,1.06513e-09,0.013979,0.000184015,7.07214e-07,3.85512e-10,0.0141637,0.00018543,7.0837e-07,1.86769e-10,0.0143499,0.000186848,7.0893e-07,7.30116e-10,0.0145374,0.000188268,7.11121e-07,6.17983e-10,0.0147264,0.000189692,7.12975e-07,5.23282e-10,0.0149168,0.000191119,7.14545e-07,8.28398e-11,0.0151087,0.000192549,7.14793e-07,1.0081e-09,0.0153019,0.000193981,7.17817e-07,5.41244e-10,0.0154966,0.000195418,7.19441e-07,-3.7907e-10,0.0156928,0.000196856,7.18304e-07,1.90641e-09,0.0158903,0.000198298,7.24023e-07,-7.27387e-10,0.0160893,0.000199744,7.21841e-07,1.00317e-09,0.0162898,0.000201191,7.24851e-07,4.39949e-10,0.0164917,0.000202642,7.2617e-07,9.6234e-10,0.0166951,0.000204097,7.29057e-07,-5.64019e-10,0.0168999,0.000205554,7.27365e-07,1.29374e-09,0.0171062,0.000207012,7.31247e-07,9.77025e-10,0.017314,0.000208478,7.34178e-07,-1.47651e-09,0.0175232,0.000209942,7.29748e-07,3.06636e-09,0.0177338,0.00021141,7.38947e-07,-1.47573e-09,0.017946,0.000212884,7.3452e-07,9.7386e-10,0.0181596,0.000214356,7.37442e-07,1.30562e-09,0.0183747,0.000215835,7.41358e-07,-6.08376e-10,0.0185913,0.000217315,7.39533e-07,1.12785e-09,0.0188093,0.000218798,7.42917e-07,-1.77711e-10,0.0190289,0.000220283,7.42384e-07,1.44562e-09,0.0192499,0.000221772,7.46721e-07,-1.68825e-11,0.0194724,0.000223266,7.4667e-07,4.84533e-10,0.0196964,0.000224761,7.48124e-07,-5.85298e-11,0.0199219,0.000226257,7.47948e-07,1.61217e-09,0.0201489,0.000227757,7.52785e-07,-8.02136e-10,0.0203775,0.00022926,7.50378e-07,1.59637e-09,0.0206075,0.000230766,7.55167e-07,4.47168e-12,0.020839,0.000232276,7.55181e-07,2.48387e-10,0.021072,0.000233787,7.55926e-07,8.6474e-10,0.0213066,0.000235302,7.5852e-07,1.78299e-11,0.0215426,0.000236819,7.58573e-07,9.26567e-10,0.0217802,0.000238339,7.61353e-07,1.34529e-12,0.0220193,0.000239862,7.61357e-07,9.30659e-10,0.0222599,0.000241387,7.64149e-07,1.34529e-12,0.0225021,0.000242915,7.64153e-07,9.26567e-10,0.0227458,0.000244447,7.66933e-07,1.76215e-11,0.022991,0.00024598,7.66986e-07,8.65536e-10,0.0232377,0.000247517,7.69582e-07,2.45677e-10,0.023486,0.000249057,7.70319e-07,1.44193e-11,0.0237358,0.000250598,7.70363e-07,1.55918e-09,0.0239872,0.000252143,7.7504e-07,-6.63173e-10,0.0242401,0.000253691,7.73051e-07,1.09357e-09,0.0244946,0.000255241,7.76331e-07,1.41919e-11,0.0247506,0.000256793,7.76374e-07,7.12248e-10,0.0250082,0.000258348,7.78511e-07,8.62049e-10,0.0252673,0.000259908,7.81097e-07,-4.35061e-10,0.025528,0.000261469,7.79792e-07,8.7825e-10,0.0257902,0.000263031,7.82426e-07,6.47181e-10,0.0260541,0.000264598,7.84368e-07,2.58448e-10,0.0263194,0.000266167,7.85143e-07,1.81558e-10,0.0265864,0.000267738,7.85688e-07,8.78041e-10,0.0268549,0.000269312,7.88322e-07,3.15102e-11,0.027125,0.000270889,7.88417e-07,8.58525e-10,0.0273967,0.000272468,7.90992e-07,2.59812e-10,0.02767,0.000274051,7.91772e-07,-3.5224e-11,0.0279448,0.000275634,7.91666e-07,1.74377e-09,0.0282212,0.000277223,7.96897e-07,-1.35196e-09,0.0284992,0.000278813,7.92841e-07,1.80141e-09,0.0287788,0.000280404,7.98246e-07,-2.65629e-10,0.0290601,0.000281999,7.97449e-07,1.12374e-09,0.0293428,0.000283598,8.0082e-07,-5.04106e-10,0.0296272,0.000285198,7.99308e-07,8.92764e-10,0.0299132,0.000286799,8.01986e-07,6.58379e-10,0.0302008,0.000288405,8.03961e-07,1.98971e-10,0.0304901,0.000290014,8.04558e-07,4.08382e-10,0.0307809,0.000291624,8.05783e-07,3.01839e-11,0.0310733,0.000293236,8.05874e-07,1.33343e-09,0.0313673,0.000294851,8.09874e-07,2.2419e-10,0.031663,0.000296472,8.10547e-07,-3.67606e-10,0.0319603,0.000298092,8.09444e-07,1.24624e-09,0.0322592,0.000299714,8.13182e-07,-8.92025e-10,0.0325597,0.000301338,8.10506e-07,2.32183e-09,0.0328619,0.000302966,8.17472e-07,-9.44719e-10,0.0331657,0.000304598,8.14638e-07,1.45703e-09,0.0334711,0.000306232,8.19009e-07,-1.15805e-09,0.0337781,0.000307866,8.15535e-07,3.17507e-09,0.0340868,0.000309507,8.2506e-07,-4.09161e-09,0.0343971,0.000311145,8.12785e-07,5.74079e-09,0.0347091,0.000312788,8.30007e-07,-3.97034e-09,0.0350227,0.000314436,8.18096e-07,2.68985e-09,0.035338,0.00031608,8.26166e-07,6.61676e-10,0.0356549,0.000317734,8.28151e-07,-1.61123e-09,0.0359734,0.000319386,8.23317e-07,2.05786e-09,0.0362936,0.000321038,8.29491e-07,8.30388e-10,0.0366155,0.0003227,8.31982e-07,-1.65424e-09,0.036939,0.000324359,8.27019e-07,2.06129e-09,0.0372642,0.000326019,8.33203e-07,8.59719e-10,0.0375911,0.000327688,8.35782e-07,-1.77488e-09,0.0379196,0.000329354,8.30458e-07,2.51464e-09,0.0382498,0.000331023,8.38002e-07,-8.33135e-10,0.0385817,0.000332696,8.35502e-07,8.17825e-10,0.0389152,0.00033437,8.37956e-07,1.28718e-09,0.0392504,0.00033605,8.41817e-07,-2.2413e-09,0.0395873,0.000337727,8.35093e-07,3.95265e-09,0.0399258,0.000339409,8.46951e-07,-2.39332e-09,0.0402661,0.000341095,8.39771e-07,1.89533e-09,0.040608,0.000342781,8.45457e-07,-1.46271e-09,0.0409517,0.000344467,8.41069e-07,3.95554e-09,0.041297,0.000346161,8.52936e-07,-3.18369e-09,0.041644,0.000347857,8.43385e-07,1.32873e-09,0.0419927,0.000349548,8.47371e-07,1.59402e-09,0.0423431,0.000351248,8.52153e-07,-2.54336e-10,0.0426952,0.000352951,8.5139e-07,-5.76676e-10,0.043049,0.000354652,8.4966e-07,2.56114e-09,0.0434045,0.000356359,8.57343e-07,-2.21744e-09,0.0437617,0.000358067,8.50691e-07,2.58344e-09,0.0441206,0.000359776,8.58441e-07,-6.65826e-10,0.0444813,0.000361491,8.56444e-07,7.99218e-11,0.0448436,0.000363204,8.56684e-07,3.46063e-10,0.0452077,0.000364919,8.57722e-07,2.26116e-09,0.0455734,0.000366641,8.64505e-07,-1.94005e-09,0.045941,0.000368364,8.58685e-07,1.77384e-09,0.0463102,0.000370087,8.64007e-07,-1.43005e-09,0.0466811,0.000371811,8.59717e-07,3.94634e-09,0.0470538,0.000373542,8.71556e-07,-3.17946e-09,0.0474282,0.000375276,8.62017e-07,1.32104e-09,0.0478043,0.000377003,8.6598e-07,1.62045e-09,0.0481822,0.00037874,8.70842e-07,-3.52297e-10,0.0485618,0.000380481,8.69785e-07,-2.11211e-10,0.0489432,0.00038222,8.69151e-07,1.19716e-09,0.0493263,0.000383962,8.72743e-07,-8.52026e-10,0.0497111,0.000385705,8.70187e-07,2.21092e-09,0.0500977,0.000387452,8.76819e-07,-5.41339e-10,0.050486,0.000389204,8.75195e-07,-4.5361e-11,0.0508761,0.000390954,8.75059e-07,7.22669e-10,0.0512679,0.000392706,8.77227e-07,8.79936e-10,0.0516615,0.000394463,8.79867e-07,-5.17048e-10,0.0520568,0.000396222,8.78316e-07,1.18833e-09,0.0524539,0.000397982,8.81881e-07,-5.11022e-10,0.0528528,0.000399744,8.80348e-07,8.55683e-10,0.0532534,0.000401507,8.82915e-07,8.13562e-10,0.0536558,0.000403276,8.85356e-07,-3.84603e-10,0.05406,0.000405045,8.84202e-07,7.24962e-10,0.0544659,0.000406816,8.86377e-07,1.20986e-09,0.0548736,0.000408592,8.90006e-07,-1.83896e-09,0.0552831,0.000410367,8.84489e-07,2.42071e-09,0.0556944,0.000412143,8.91751e-07,-3.93413e-10,0.0561074,0.000413925,8.90571e-07,-8.46967e-10,0.0565222,0.000415704,8.8803e-07,3.78122e-09,0.0569388,0.000417491,8.99374e-07,-3.1021e-09,0.0573572,0.000419281,8.90068e-07,1.17658e-09,0.0577774,0.000421064,8.93597e-07,2.12117e-09,0.0581993,0.000422858,8.99961e-07,-2.21068e-09,0.0586231,0.000424651,8.93329e-07,2.9961e-09,0.0590486,0.000426447,9.02317e-07,-2.32311e-09,0.059476,0.000428244,8.95348e-07,2.57122e-09,0.0599051,0.000430043,9.03062e-07,-5.11098e-10,0.0603361,0.000431847,9.01528e-07,-5.27166e-10,0.0607688,0.000433649,8.99947e-07,2.61984e-09,0.0612034,0.000435457,9.07806e-07,-2.50141e-09,0.0616397,0.000437265,9.00302e-07,3.66045e-09,0.0620779,0.000439076,9.11283e-07,-4.68977e-09,0.0625179,0.000440885,8.97214e-07,7.64783e-09,0.0629597,0.000442702,9.20158e-07,-7.27499e-09,0.0634033,0.000444521,8.98333e-07,6.55113e-09,0.0638487,0.000446337,9.17986e-07,-4.02844e-09,0.0642959,0.000448161,9.05901e-07,2.11196e-09,0.064745,0.000449979,9.12236e-07,3.03125e-09,0.0651959,0.000451813,9.2133e-07,-6.78648e-09,0.0656486,0.000453635,9.00971e-07,9.21375e-09,0.0661032,0.000455464,9.28612e-07,-7.71684e-09,0.0665596,0.000457299,9.05462e-07,6.7522e-09,0.0670178,0.00045913,9.25718e-07,-4.3907e-09,0.0674778,0.000460968,9.12546e-07,3.36e-09,0.0679397,0.000462803,9.22626e-07,-1.59876e-09,0.0684034,0.000464644,9.1783e-07,3.0351e-09,0.068869,0.000466488,9.26935e-07,-3.09101e-09,0.0693364,0.000468333,9.17662e-07,1.8785e-09,0.0698057,0.000470174,9.23298e-07,3.02733e-09,0.0702768,0.00047203,9.3238e-07,-6.53722e-09,0.0707497,0.000473875,9.12768e-07,8.22054e-09,0.0712245,0.000475725,9.37429e-07,-3.99325e-09,0.0717012,0.000477588,9.2545e-07,3.01839e-10,0.0721797,0.00047944,9.26355e-07,2.78597e-09,0.0726601,0.000481301,9.34713e-07,-3.99507e-09,0.0731423,0.000483158,9.22728e-07,5.7435e-09,0.0736264,0.000485021,9.39958e-07,-4.07776e-09,0.0741123,0.000486888,9.27725e-07,3.11695e-09,0.0746002,0.000488753,9.37076e-07,-9.39394e-10,0.0750898,0.000490625,9.34258e-07,6.4055e-10,0.0755814,0.000492495,9.3618e-07,-1.62265e-09,0.0760748,0.000494363,9.31312e-07,5.84995e-09,0.0765701,0.000496243,9.48861e-07,-6.87601e-09,0.0770673,0.00049812,9.28233e-07,6.75296e-09,0.0775664,0.000499997,9.48492e-07,-5.23467e-09,0.0780673,0.000501878,9.32788e-07,6.73523e-09,0.0785701,0.000503764,9.52994e-07,-6.80514e-09,0.0790748,0.000505649,9.32578e-07,5.5842e-09,0.0795814,0.000507531,9.49331e-07,-6.30583e-10,0.0800899,0.000509428,9.47439e-07,-3.0618e-09,0.0806003,0.000511314,9.38254e-07,5.4273e-09,0.0811125,0.000513206,9.54536e-07,-3.74627e-09,0.0816267,0.000515104,9.43297e-07,2.10713e-09,0.0821427,0.000516997,9.49618e-07,2.76839e-09,0.0826607,0.000518905,9.57924e-07,-5.73006e-09,0.0831805,0.000520803,9.40733e-07,5.25072e-09,0.0837023,0.0005227,9.56486e-07,-3.71718e-10,0.084226,0.000524612,9.5537e-07,-3.76404e-09,0.0847515,0.000526512,9.44078e-07,7.97735e-09,0.085279,0.000528424,9.6801e-07,-5.79367e-09,0.0858084,0.000530343,9.50629e-07,2.96268e-10,0.0863397,0.000532245,9.51518e-07,4.6086e-09,0.0868729,0.000534162,9.65344e-07,-3.82947e-09,0.087408,0.000536081,9.53856e-07,3.25861e-09,0.087945,0.000537998,9.63631e-07,-1.7543e-09,0.088484,0.00053992,9.58368e-07,3.75849e-09,0.0890249,0.000541848,9.69644e-07,-5.82891e-09,0.0895677,0.00054377,9.52157e-07,4.65593e-09,0.0901124,0.000545688,9.66125e-07,2.10643e-09,0.0906591,0.000547627,9.72444e-07,-5.63099e-09,0.0912077,0.000549555,9.55551e-07,5.51627e-09,0.0917582,0.000551483,9.721e-07,-1.53292e-09,0.0923106,0.000553422,9.67501e-07,6.15311e-10,0.092865,0.000555359,9.69347e-07,-9.28291e-10,0.0934213,0.000557295,9.66562e-07,3.09774e-09,0.0939796,0.000559237,9.75856e-07,-4.01186e-09,0.0945398,0.000561177,9.6382e-07,5.49892e-09,0.095102,0.000563121,9.80317e-07,-3.08258e-09,0.0956661,0.000565073,9.71069e-07,-6.19176e-10,0.0962321,0.000567013,9.69212e-07,5.55932e-09,0.0968001,0.000568968,9.8589e-07,-6.71704e-09,0.09737,0.00057092,9.65738e-07,6.40762e-09,0.0979419,0.00057287,9.84961e-07,-4.0122e-09,0.0985158,0.000574828,9.72925e-07,2.19059e-09,0.0990916,0.000576781,9.79496e-07,2.70048e-09,0.0996693,0.000578748,9.87598e-07,-5.54193e-09,0.100249,0.000580706,9.70972e-07,4.56597e-09,0.100831,0.000582662,9.8467e-07,2.17923e-09,0.101414,0.000584638,9.91208e-07,-5.83232e-09,0.102,0.000586603,9.73711e-07,6.24884e-09,0.102588,0.000588569,9.92457e-07,-4.26178e-09,0.103177,0.000590541,9.79672e-07,3.34781e-09,0.103769,0.00059251,9.89715e-07,-1.67904e-09,0.104362,0.000594485,9.84678e-07,3.36839e-09,0.104958,0.000596464,9.94783e-07,-4.34397e-09,0.105555,0.000598441,9.81751e-07,6.55696e-09,0.106155,0.000600424,1.00142e-06,-6.98272e-09,0.106756,0.000602406,9.80474e-07,6.4728e-09,0.107359,0.000604386,9.99893e-07,-4.00742e-09,0.107965,0.000606374,9.8787e-07,2.10654e-09,0.108572,0.000608356,9.9419e-07,3.0318e-09,0.109181,0.000610353,1.00329e-06,-6.7832e-09,0.109793,0.00061234,9.82936e-07,9.1998e-09,0.110406,0.000614333,1.01054e-06,-7.6642e-09,0.111021,0.000616331,9.87543e-07,6.55579e-09,0.111639,0.000618326,1.00721e-06,-3.65791e-09,0.112258,0.000620329,9.96236e-07,6.25467e-10,0.112879,0.000622324,9.98113e-07,1.15593e-09,0.113503,0.000624323,1.00158e-06,2.20158e-09,0.114128,0.000626333,1.00819e-06,-2.51191e-09,0.114755,0.000628342,1.00065e-06,3.95517e-10,0.115385,0.000630345,1.00184e-06,9.29807e-10,0.116016,0.000632351,1.00463e-06,3.33599e-09,0.116649,0.00063437,1.01463e-06,-6.82329e-09,0.117285,0.000636379,9.94163e-07,9.05595e-09,0.117922,0.000638395,1.02133e-06,-7.04862e-09,0.118562,0.000640416,1.00019e-06,4.23737e-09,0.119203,0.000642429,1.0129e-06,-2.45033e-09,0.119847,0.000644448,1.00555e-06,5.56395e-09,0.120492,0.000646475,1.02224e-06,-4.9043e-09,0.121139,0.000648505,1.00753e-06,-8.47952e-10,0.121789,0.000650518,1.00498e-06,8.29622e-09,0.122441,0.000652553,1.02987e-06,-9.98538e-09,0.123094,0.000654582,9.99914e-07,9.2936e-09,0.12375,0.00065661,1.02779e-06,-4.83707e-09,0.124407,0.000658651,1.01328e-06,2.60411e-09,0.125067,0.000660685,1.0211e-06,-5.57945e-09,0.125729,0.000662711,1.00436e-06,1.22631e-08,0.126392,0.000664756,1.04115e-06,-1.36704e-08,0.127058,0.000666798,1.00014e-06,1.26161e-08,0.127726,0.000668836,1.03798e-06,-6.99155e-09,0.128396,0.000670891,1.01701e-06,4.48836e-10,0.129068,0.000672926,1.01836e-06,5.19606e-09,0.129742,0.000674978,1.03394e-06,-6.3319e-09,0.130418,0.000677027,1.01495e-06,5.2305e-09,0.131096,0.000679073,1.03064e-06,3.11123e-10,0.131776,0.000681135,1.03157e-06,-6.47511e-09,0.132458,0.000683179,1.01215e-06,1.06882e-08,0.133142,0.000685235,1.04421e-06,-6.47519e-09,0.133829,0.000687304,1.02479e-06,3.11237e-10,0.134517,0.000689355,1.02572e-06,5.23035e-09,0.135207,0.000691422,1.04141e-06,-6.3316e-09,0.1359,0.000693486,1.02242e-06,5.19484e-09,0.136594,0.000695546,1.038e-06,4.53497e-10,0.137291,0.000697623,1.03936e-06,-7.00891e-09,0.137989,0.000699681,1.01834e-06,1.2681e-08,0.13869,0.000701756,1.05638e-06,-1.39128e-08,0.139393,0.000703827,1.01464e-06,1.31679e-08,0.140098,0.000705896,1.05414e-06,-8.95659e-09,0.140805,0.000707977,1.02727e-06,7.75742e-09,0.141514,0.000710055,1.05055e-06,-7.17182e-09,0.142225,0.000712135,1.02903e-06,6.02862e-09,0.142938,0.000714211,1.04712e-06,-2.04163e-09,0.143653,0.000716299,1.04099e-06,2.13792e-09,0.144371,0.000718387,1.04741e-06,-6.51009e-09,0.14509,0.000720462,1.02787e-06,9.00123e-09,0.145812,0.000722545,1.05488e-06,3.07523e-10,0.146535,0.000724656,1.0558e-06,-1.02312e-08,0.147261,0.000726737,1.02511e-06,1.0815e-08,0.147989,0.000728819,1.05755e-06,-3.22681e-09,0.148719,0.000730925,1.04787e-06,2.09244e-09,0.14945,0.000733027,1.05415e-06,-5.143e-09,0.150185,0.00073512,1.03872e-06,3.57844e-09,0.150921,0.000737208,1.04946e-06,5.73027e-09,0.151659,0.000739324,1.06665e-06,-1.15983e-08,0.152399,0.000741423,1.03185e-06,1.08605e-08,0.153142,0.000743519,1.06443e-06,-2.04106e-09,0.153886,0.000745642,1.05831e-06,-2.69642e-09,0.154633,0.00074775,1.05022e-06,-2.07425e-09,0.155382,0.000749844,1.044e-06,1.09934e-08,0.156133,0.000751965,1.07698e-06,-1.20972e-08,0.156886,0.000754083,1.04069e-06,7.59288e-09,0.157641,0.000756187,1.06347e-06,-3.37305e-09,0.158398,0.000758304,1.05335e-06,5.89921e-09,0.159158,0.000760428,1.07104e-06,-5.32248e-09,0.159919,0.000762554,1.05508e-06,4.8927e-10,0.160683,0.000764666,1.05654e-06,3.36547e-09,0.161448,0.000766789,1.06664e-06,9.50081e-10,0.162216,0.000768925,1.06949e-06,-7.16568e-09,0.162986,0.000771043,1.04799e-06,1.28114e-08,0.163758,0.000773177,1.08643e-06,-1.42774e-08,0.164533,0.000775307,1.0436e-06,1.44956e-08,0.165309,0.000777438,1.08708e-06,-1.39025e-08,0.166087,0.00077957,1.04538e-06,1.13118e-08,0.166868,0.000781695,1.07931e-06,-1.54224e-09,0.167651,0.000783849,1.07468e-06,-5.14312e-09,0.168436,0.000785983,1.05925e-06,7.21381e-09,0.169223,0.000788123,1.0809e-06,-8.81096e-09,0.170012,0.000790259,1.05446e-06,1.31289e-08,0.170803,0.000792407,1.09385e-06,-1.39022e-08,0.171597,0.000794553,1.05214e-06,1.26775e-08,0.172392,0.000796695,1.09018e-06,-7.00557e-09,0.17319,0.000798855,1.06916e-06,4.43796e-10,0.17399,0.000800994,1.07049e-06,5.23031e-09,0.174792,0.000803151,1.08618e-06,-6.46397e-09,0.175596,0.000805304,1.06679e-06,5.72444e-09,0.176403,0.000807455,1.08396e-06,-1.53254e-09,0.177211,0.000809618,1.07937e-06,4.05673e-10,0.178022,0.000811778,1.08058e-06,-9.01916e-11,0.178835,0.000813939,1.08031e-06,-4.49821e-11,0.17965,0.000816099,1.08018e-06,2.70234e-10,0.180467,0.00081826,1.08099e-06,-1.03603e-09,0.181286,0.000820419,1.07788e-06,3.87392e-09,0.182108,0.000822587,1.0895e-06,4.41522e-10,0.182932,0.000824767,1.09083e-06,-5.63997e-09,0.183758,0.000826932,1.07391e-06,7.21707e-09,0.184586,0.000829101,1.09556e-06,-8.32718e-09,0.185416,0.000831267,1.07058e-06,1.11907e-08,0.186248,0.000833442,1.10415e-06,-6.63336e-09,0.187083,0.00083563,1.08425e-06,4.41484e-10,0.187919,0.0008378,1.08557e-06,4.86754e-09,0.188758,0.000839986,1.10017e-06,-5.01041e-09,0.189599,0.000842171,1.08514e-06,2.72811e-10,0.190443,0.000844342,1.08596e-06,3.91916e-09,0.191288,0.000846526,1.09772e-06,-1.04819e-09,0.192136,0.000848718,1.09457e-06,2.73531e-10,0.192985,0.000850908,1.0954e-06,-4.58916e-11,0.193837,0.000853099,1.09526e-06,-9.01158e-11,0.194692,0.000855289,1.09499e-06,4.06506e-10,0.195548,0.00085748,1.09621e-06,-1.53595e-09,0.196407,0.000859668,1.0916e-06,5.73717e-09,0.197267,0.000861869,1.10881e-06,-6.51164e-09,0.19813,0.000864067,1.08928e-06,5.40831e-09,0.198995,0.000866261,1.1055e-06,-2.20401e-10,0.199863,0.000868472,1.10484e-06,-4.52652e-09,0.200732,0.000870668,1.09126e-06,3.42508e-09,0.201604,0.000872861,1.10153e-06,5.72762e-09,0.202478,0.000875081,1.11872e-06,-1.14344e-08,0.203354,0.000877284,1.08441e-06,1.02076e-08,0.204233,0.000879484,1.11504e-06,4.06355e-10,0.205113,0.000881715,1.11626e-06,-1.18329e-08,0.205996,0.000883912,1.08076e-06,1.71227e-08,0.206881,0.000886125,1.13213e-06,-1.19546e-08,0.207768,0.000888353,1.09626e-06,8.93465e-10,0.208658,0.000890548,1.09894e-06,8.38062e-09,0.209549,0.000892771,1.12408e-06,-4.61353e-09,0.210443,0.000895006,1.11024e-06,-4.82756e-09,0.211339,0.000897212,1.09576e-06,9.02245e-09,0.212238,0.00089943,1.12283e-06,-1.45997e-09,0.213138,0.000901672,1.11845e-06,-3.18255e-09,0.214041,0.000903899,1.1089e-06,-7.11073e-10,0.214946,0.000906115,1.10677e-06,6.02692e-09,0.215853,0.000908346,1.12485e-06,-8.49548e-09,0.216763,0.00091057,1.09936e-06,1.30537e-08,0.217675,0.000912808,1.13852e-06,-1.3917e-08,0.218588,0.000915044,1.09677e-06,1.28121e-08,0.219505,0.000917276,1.13521e-06,-7.5288e-09,0.220423,0.000919523,1.11262e-06,2.40205e-09,0.221344,0.000921756,1.11983e-06,-2.07941e-09,0.222267,0.000923989,1.11359e-06,5.91551e-09,0.223192,0.000926234,1.13134e-06,-6.68149e-09,0.224119,0.000928477,1.11129e-06,5.90929e-09,0.225049,0.000930717,1.12902e-06,-2.05436e-09,0.22598,0.000932969,1.12286e-06,2.30807e-09,0.226915,0.000935222,1.12978e-06,-7.17796e-09,0.227851,0.00093746,1.10825e-06,1.15028e-08,0.228789,0.000939711,1.14276e-06,-9.03083e-09,0.22973,0.000941969,1.11566e-06,9.71932e-09,0.230673,0.00094423,1.14482e-06,-1.49452e-08,0.231619,0.000946474,1.09998e-06,2.02591e-08,0.232566,0.000948735,1.16076e-06,-2.13879e-08,0.233516,0.000950993,1.0966e-06,2.05888e-08,0.234468,0.000953247,1.15837e-06,-1.62642e-08,0.235423,0.000955515,1.10957e-06,1.46658e-08,0.236379,0.000957779,1.15357e-06,-1.25966e-08,0.237338,0.000960048,1.11578e-06,5.91793e-09,0.238299,0.000962297,1.13353e-06,3.82602e-09,0.239263,0.000964576,1.14501e-06,-6.3208e-09,0.240229,0.000966847,1.12605e-06,6.55613e-09,0.241197,0.000969119,1.14572e-06,-5.00268e-09,0.242167,0.000971395,1.13071e-06,-1.44659e-09,0.243139,0.000973652,1.12637e-06,1.07891e-08,0.244114,0.000975937,1.15874e-06,-1.19073e-08,0.245091,0.000978219,1.12302e-06,7.03782e-09,0.246071,0.000980486,1.14413e-06,-1.34276e-09,0.247052,0.00098277,1.1401e-06,-1.66669e-09,0.248036,0.000985046,1.1351e-06,8.00935e-09,0.249022,0.00098734,1.15913e-06,-1.54694e-08,0.250011,0.000989612,1.11272e-06,2.4066e-08,0.251002,0.000991909,1.18492e-06,-2.11901e-08,0.251995,0.000994215,1.12135e-06,1.08973e-09,0.25299,0.000996461,1.12462e-06,1.68311e-08,0.253988,0.000998761,1.17511e-06,-8.8094e-09,0.254987,0.00100109,1.14868e-06,-1.13958e-08,0.25599,0.00100335,1.1145e-06,2.45902e-08,0.256994,0.00100565,1.18827e-06,-2.73603e-08,0.258001,0.00100795,1.10618e-06,2.52464e-08,0.25901,0.00101023,1.18192e-06,-1.40207e-08,0.260021,0.00101256,1.13986e-06,1.03387e-09,0.261035,0.00101484,1.14296e-06,9.8853e-09,0.262051,0.00101715,1.17262e-06,-1.07726e-08,0.263069,0.00101947,1.1403e-06,3.40272e-09,0.26409,0.00102176,1.15051e-06,-2.83827e-09,0.265113,0.00102405,1.142e-06,7.95039e-09,0.266138,0.00102636,1.16585e-06,8.39047e-10,0.267166,0.00102869,1.16836e-06,-1.13066e-08,0.268196,0.00103099,1.13444e-06,1.4585e-08,0.269228,0.00103331,1.1782e-06,-1.72314e-08,0.270262,0.00103561,1.1265e-06,2.45382e-08,0.271299,0.00103794,1.20012e-06,-2.13166e-08,0.272338,0.00104028,1.13617e-06,1.12364e-09,0.273379,0.00104255,1.13954e-06,1.68221e-08,0.274423,0.00104488,1.19001e-06,-8.80736e-09,0.275469,0.00104723,1.16358e-06,-1.13948e-08,0.276518,0.00104953,1.1294e-06,2.45839e-08,0.277568,0.00105186,1.20315e-06,-2.73361e-08,0.278621,0.00105418,1.12114e-06,2.51559e-08,0.279677,0.0010565,1.19661e-06,-1.36832e-08,0.280734,0.00105885,1.15556e-06,-2.25706e-10,0.281794,0.00106116,1.15488e-06,1.45862e-08,0.282857,0.00106352,1.19864e-06,-2.83167e-08,0.283921,0.00106583,1.11369e-06,3.90759e-08,0.284988,0.00106817,1.23092e-06,-3.85801e-08,0.286058,0.00107052,1.11518e-06,2.58375e-08,0.287129,0.00107283,1.19269e-06,-5.16498e-09,0.288203,0.0010752,1.1772e-06,-5.17768e-09,0.28928,0.00107754,1.16167e-06,-3.92671e-09,0.290358,0.00107985,1.14988e-06,2.08846e-08,0.29144,0.00108221,1.21254e-06,-2.00072e-08,0.292523,0.00108458,1.15252e-06,-4.60659e-10,0.293609,0.00108688,1.15114e-06,2.18499e-08,0.294697,0.00108925,1.21669e-06,-2.73343e-08,0.295787,0.0010916,1.13468e-06,2.78826e-08,0.29688,0.00109395,1.21833e-06,-2.45915e-08,0.297975,0.00109632,1.14456e-06,1.08787e-08,0.299073,0.00109864,1.17719e-06,1.08788e-08,0.300172,0.00110102,1.20983e-06,-2.45915e-08,0.301275,0.00110337,1.13605e-06,2.78828e-08,0.302379,0.00110573,1.2197e-06,-2.73348e-08,0.303486,0.00110808,1.1377e-06,2.18518e-08,0.304595,0.00111042,1.20325e-06,-4.67556e-10,0.305707,0.00111283,1.20185e-06,-1.99816e-08,0.306821,0.00111517,1.14191e-06,2.07891e-08,0.307937,0.00111752,1.20427e-06,-3.57026e-09,0.309056,0.00111992,1.19356e-06,-6.50797e-09,0.310177,0.00112228,1.17404e-06,-2.00165e-10,0.3113,0.00112463,1.17344e-06,7.30874e-09,0.312426,0.001127,1.19536e-06,7.67424e-10,0.313554,0.00112939,1.19767e-06,-1.03784e-08,0.314685,0.00113176,1.16653e-06,1.09437e-08,0.315818,0.00113412,1.19936e-06,-3.59406e-09,0.316953,0.00113651,1.18858e-06,3.43251e-09,0.318091,0.0011389,1.19888e-06,-1.0136e-08,0.319231,0.00114127,1.16847e-06,7.30915e-09,0.320374,0.00114363,1.1904e-06,1.07018e-08,0.321518,0.00114604,1.2225e-06,-2.03137e-08,0.322666,0.00114842,1.16156e-06,1.09484e-08,0.323815,0.00115078,1.19441e-06,6.32224e-09,0.324967,0.00115319,1.21337e-06,-6.43509e-09,0.326122,0.00115559,1.19407e-06,-1.03842e-08,0.327278,0.00115795,1.16291e-06,1.81697e-08,0.328438,0.00116033,1.21742e-06,-2.6901e-09,0.329599,0.00116276,1.20935e-06,-7.40939e-09,0.330763,0.00116515,1.18713e-06,2.52533e-09,0.331929,0.00116754,1.1947e-06,-2.69191e-09,0.333098,0.00116992,1.18663e-06,8.24218e-09,0.334269,0.00117232,1.21135e-06,-4.74377e-10,0.335443,0.00117474,1.20993e-06,-6.34471e-09,0.336619,0.00117714,1.1909e-06,-3.94922e-09,0.337797,0.00117951,1.17905e-06,2.21417e-08,0.338978,0.00118193,1.24547e-06,-2.50128e-08,0.340161,0.00118435,1.17043e-06,1.8305e-08,0.341346,0.00118674,1.22535e-06,-1.84048e-08,0.342534,0.00118914,1.17013e-06,2.55121e-08,0.343725,0.00119156,1.24667e-06,-2.40389e-08,0.344917,0.00119398,1.17455e-06,1.10389e-08,0.346113,0.00119636,1.20767e-06,9.68574e-09,0.34731,0.0011988,1.23673e-06,-1.99797e-08,0.34851,0.00120122,1.17679e-06,1.06284e-08,0.349713,0.0012036,1.20867e-06,7.26868e-09,0.350917,0.00120604,1.23048e-06,-9.90072e-09,0.352125,0.00120847,1.20078e-06,2.53177e-09,0.353334,0.00121088,1.20837e-06,-2.26199e-10,0.354546,0.0012133,1.20769e-06,-1.62705e-09,0.355761,0.00121571,1.20281e-06,6.73435e-09,0.356978,0.00121813,1.22302e-06,4.49207e-09,0.358197,0.00122059,1.23649e-06,-2.47027e-08,0.359419,0.00122299,1.16238e-06,3.47142e-08,0.360643,0.00122542,1.26653e-06,-2.47472e-08,0.36187,0.00122788,1.19229e-06,4.66965e-09,0.363099,0.00123028,1.20629e-06,6.06872e-09,0.36433,0.00123271,1.2245e-06,8.57729e-10,0.365564,0.00123516,1.22707e-06,-9.49952e-09,0.366801,0.00123759,1.19858e-06,7.33792e-09,0.36804,0.00124001,1.22059e-06,9.95025e-09,0.369281,0.00124248,1.25044e-06,-1.73366e-08,0.370525,0.00124493,1.19843e-06,-2.08464e-10,0.371771,0.00124732,1.1978e-06,1.81704e-08,0.373019,0.00124977,1.25232e-06,-1.28683e-08,0.37427,0.00125224,1.21371e-06,3.50042e-09,0.375524,0.00125468,1.22421e-06,-1.1335e-09,0.37678,0.00125712,1.22081e-06,1.03345e-09,0.378038,0.00125957,1.22391e-06,-3.00023e-09,0.379299,0.00126201,1.21491e-06,1.09676e-08,0.380562,0.00126447,1.24781e-06,-1.10676e-08,0.381828,0.00126693,1.21461e-06,3.50042e-09,0.383096,0.00126937,1.22511e-06,-2.93403e-09,0.384366,0.00127181,1.21631e-06,8.23574e-09,0.385639,0.00127427,1.24102e-06,-2.06607e-10,0.386915,0.00127675,1.2404e-06,-7.40935e-09,0.388193,0.00127921,1.21817e-06,4.1761e-11,0.389473,0.00128165,1.21829e-06,7.24223e-09,0.390756,0.0012841,1.24002e-06,7.91564e-10,0.392042,0.00128659,1.2424e-06,-1.04086e-08,0.393329,0.00128904,1.21117e-06,1.10405e-08,0.39462,0.0012915,1.24429e-06,-3.951e-09,0.395912,0.00129397,1.23244e-06,4.7634e-09,0.397208,0.00129645,1.24673e-06,-1.51025e-08,0.398505,0.0012989,1.20142e-06,2.58443e-08,0.399805,0.00130138,1.27895e-06,-2.86702e-08,0.401108,0.00130385,1.19294e-06,2.92318e-08,0.402413,0.00130632,1.28064e-06,-2.86524e-08,0.403721,0.0013088,1.19468e-06,2.57731e-08,0.405031,0.00131127,1.272e-06,-1.48355e-08,0.406343,0.00131377,1.2275e-06,3.76652e-09,0.407658,0.00131623,1.23879e-06,-2.30784e-10,0.408976,0.00131871,1.2381e-06,-2.84331e-09,0.410296,0.00132118,1.22957e-06,1.16041e-08,0.411618,0.00132367,1.26438e-06,-1.37708e-08,0.412943,0.00132616,1.22307e-06,1.36768e-08,0.41427,0.00132865,1.2641e-06,-1.1134e-08,0.4156,0.00133114,1.2307e-06,1.05714e-09,0.416933,0.00133361,1.23387e-06,6.90538e-09,0.418267,0.00133609,1.25459e-06,1.12372e-09,0.419605,0.00133861,1.25796e-06,-1.14002e-08,0.420945,0.00134109,1.22376e-06,1.46747e-08,0.422287,0.00134358,1.26778e-06,-1.7496e-08,0.423632,0.00134606,1.21529e-06,2.5507e-08,0.424979,0.00134857,1.29182e-06,-2.49272e-08,0.426329,0.00135108,1.21703e-06,1.45972e-08,0.427681,0.00135356,1.26083e-06,-3.65935e-09,0.429036,0.00135607,1.24985e-06,4.00178e-11,0.430393,0.00135857,1.24997e-06,3.49917e-09,0.431753,0.00136108,1.26047e-06,-1.40366e-08,0.433116,0.00136356,1.21836e-06,2.28448e-08,0.43448,0.00136606,1.28689e-06,-1.77378e-08,0.435848,0.00136858,1.23368e-06,1.83043e-08,0.437218,0.0013711,1.28859e-06,-2.56769e-08,0.43859,0.0013736,1.21156e-06,2.47987e-08,0.439965,0.0013761,1.28595e-06,-1.39133e-08,0.441342,0.00137863,1.24421e-06,1.05202e-09,0.442722,0.00138112,1.24737e-06,9.70507e-09,0.444104,0.00138365,1.27649e-06,-1.00698e-08,0.445489,0.00138617,1.24628e-06,7.72123e-10,0.446877,0.00138867,1.24859e-06,6.98132e-09,0.448267,0.00139118,1.26954e-06,1.10477e-09,0.449659,0.00139373,1.27285e-06,-1.14003e-08,0.451054,0.00139624,1.23865e-06,1.4694e-08,0.452452,0.00139876,1.28273e-06,-1.75734e-08,0.453852,0.00140127,1.23001e-06,2.5797e-08,0.455254,0.00140381,1.3074e-06,-2.60097e-08,0.456659,0.00140635,1.22937e-06,1.86371e-08,0.458067,0.00140886,1.28529e-06,-1.8736e-08,0.459477,0.00141137,1.22908e-06,2.65048e-08,0.46089,0.00141391,1.30859e-06,-2.76784e-08,0.462305,0.00141645,1.22556e-06,2.46043e-08,0.463722,0.00141897,1.29937e-06,-1.11341e-08,0.465143,0.00142154,1.26597e-06,-9.87033e-09,0.466565,0.00142404,1.23636e-06,2.08131e-08,0.467991,0.00142657,1.2988e-06,-1.37773e-08,0.469419,0.00142913,1.25746e-06,4.49378e-09,0.470849,0.00143166,1.27094e-06,-4.19781e-09,0.472282,0.00143419,1.25835e-06,1.22975e-08,0.473717,0.00143674,1.29524e-06,-1.51902e-08,0.475155,0.00143929,1.24967e-06,1.86608e-08,0.476596,0.00144184,1.30566e-06,-2.96506e-08,0.478039,0.00144436,1.2167e-06,4.03368e-08,0.479485,0.00144692,1.33771e-06,-4.22896e-08,0.480933,0.00144947,1.21085e-06,3.94148e-08,0.482384,0.00145201,1.32909e-06,-2.59626e-08,0.483837,0.00145459,1.2512e-06,4.83124e-09,0.485293,0.0014571,1.2657e-06,6.63757e-09,0.486751,0.00145966,1.28561e-06,-1.57911e-09,0.488212,0.00146222,1.28087e-06,-3.21468e-10,0.489676,0.00146478,1.27991e-06,2.86517e-09,0.491142,0.00146735,1.2885e-06,-1.11392e-08,0.49261,0.00146989,1.25508e-06,1.18893e-08,0.494081,0.00147244,1.29075e-06,-6.61574e-09,0.495555,0.001475,1.27091e-06,1.45736e-08,0.497031,0.00147759,1.31463e-06,-2.18759e-08,0.49851,0.00148015,1.249e-06,1.33252e-08,0.499992,0.00148269,1.28897e-06,-1.62277e-09,0.501476,0.00148526,1.28411e-06,-6.83421e-09,0.502962,0.00148781,1.2636e-06,2.89596e-08,0.504451,0.00149042,1.35048e-06,-4.93997e-08,0.505943,0.00149298,1.20228e-06,4.94299e-08,0.507437,0.00149553,1.35057e-06,-2.91107e-08,0.508934,0.00149814,1.26324e-06,7.40848e-09,0.510434,0.00150069,1.28547e-06,-5.23187e-10,0.511936,0.00150326,1.2839e-06,-5.31585e-09,0.51344,0.00150581,1.26795e-06,2.17866e-08,0.514947,0.00150841,1.33331e-06,-2.22257e-08,0.516457,0.00151101,1.26663e-06,7.51178e-09,0.517969,0.00151357,1.28917e-06,-7.82128e-09,0.519484,0.00151613,1.2657e-06,2.37733e-08,0.521002,0.00151873,1.33702e-06,-2.76674e-08,0.522522,0.00152132,1.25402e-06,2.72917e-08,0.524044,0.00152391,1.3359e-06,-2.18949e-08,0.525569,0.00152652,1.27021e-06,6.83372e-10,0.527097,0.00152906,1.27226e-06,1.91613e-08,0.528628,0.00153166,1.32974e-06,-1.77241e-08,0.53016,0.00153427,1.27657e-06,-7.86963e-09,0.531696,0.0015368,1.25296e-06,4.92027e-08,0.533234,0.00153945,1.40057e-06,-6.9732e-08,0.534775,0.00154204,1.19138e-06,5.09114e-08,0.536318,0.00154458,1.34411e-06,-1.4704e-08,0.537864,0.00154722,1.3e-06,7.9048e-09,0.539413,0.00154984,1.32371e-06,-1.69152e-08,0.540964,0.00155244,1.27297e-06,1.51355e-10,0.542517,0.00155499,1.27342e-06,1.63099e-08,0.544074,0.00155758,1.32235e-06,-5.78647e-09,0.545633,0.00156021,1.30499e-06,6.83599e-09,0.547194,0.00156284,1.3255e-06,-2.15575e-08,0.548758,0.00156543,1.26083e-06,1.97892e-08,0.550325,0.00156801,1.32019e-06,2.00525e-09,0.551894,0.00157065,1.32621e-06,-2.78103e-08,0.553466,0.00157322,1.24278e-06,4.96314e-08,0.555041,0.00157586,1.39167e-06,-5.1506e-08,0.556618,0.00157849,1.23716e-06,3.71835e-08,0.558198,0.00158107,1.34871e-06,-3.76233e-08,0.55978,0.00158366,1.23584e-06,5.37052e-08,0.561365,0.00158629,1.39695e-06,-5.79884e-08,0.562953,0.00158891,1.22299e-06,5.90392e-08,0.564543,0.00159153,1.4001e-06,-5.89592e-08,0.566136,0.00159416,1.22323e-06,5.7588e-08,0.567731,0.00159678,1.39599e-06,-5.21835e-08,0.569329,0.00159941,1.23944e-06,3.19369e-08,0.57093,0.00160199,1.33525e-06,-1.59594e-08,0.572533,0.00160461,1.28737e-06,3.19006e-08,0.574139,0.00160728,1.38307e-06,-5.20383e-08,0.575748,0.00160989,1.22696e-06,5.70431e-08,0.577359,0.00161251,1.39809e-06,-5.69247e-08,0.578973,0.00161514,1.22731e-06,5.14463e-08,0.580589,0.00161775,1.38165e-06,-2.9651e-08,0.582208,0.00162042,1.2927e-06,7.55339e-09,0.58383,0.00162303,1.31536e-06,-5.62636e-10,0.585455,0.00162566,1.31367e-06,-5.30281e-09,0.587081,0.00162827,1.29776e-06,2.17738e-08,0.588711,0.00163093,1.36309e-06,-2.21875e-08,0.590343,0.00163359,1.29652e-06,7.37164e-09,0.591978,0.00163621,1.31864e-06,-7.29907e-09,0.593616,0.00163882,1.29674e-06,2.18247e-08,0.595256,0.00164148,1.36221e-06,-2.03952e-08,0.596899,0.00164414,1.30103e-06,1.51241e-10,0.598544,0.00164675,1.30148e-06,1.97902e-08,0.600192,0.00164941,1.36085e-06,-1.97074e-08,0.601843,0.00165207,1.30173e-06,-5.65175e-10,0.603496,0.00165467,1.30004e-06,2.1968e-08,0.605152,0.00165734,1.36594e-06,-2.77024e-08,0.606811,0.00165999,1.28283e-06,2.92369e-08,0.608472,0.00166264,1.37054e-06,-2.96407e-08,0.610136,0.00166529,1.28162e-06,2.97215e-08,0.611803,0.00166795,1.37079e-06,-2.96408e-08,0.613472,0.0016706,1.28186e-06,2.92371e-08,0.615144,0.00167325,1.36957e-06,-2.77031e-08,0.616819,0.00167591,1.28647e-06,2.19708e-08,0.618496,0.00167855,1.35238e-06,-5.75407e-10,0.620176,0.00168125,1.35065e-06,-1.9669e-08,0.621858,0.00168389,1.29164e-06,1.96468e-08,0.623544,0.00168653,1.35058e-06,6.86403e-10,0.625232,0.00168924,1.35264e-06,-2.23924e-08,0.626922,0.00169187,1.28547e-06,2.92788e-08,0.628615,0.00169453,1.3733e-06,-3.51181e-08,0.630311,0.00169717,1.26795e-06,5.15889e-08,0.63201,0.00169987,1.42272e-06,-5.2028e-08,0.633711,0.00170255,1.26663e-06,3.73139e-08,0.635415,0.0017052,1.37857e-06,-3.76227e-08,0.637121,0.00170784,1.2657e-06,5.35722e-08,0.63883,0.00171054,1.42642e-06,-5.74567e-08,0.640542,0.00171322,1.25405e-06,5.70456e-08,0.642257,0.0017159,1.42519e-06,-5.15163e-08,0.643974,0.00171859,1.27064e-06,2.98103e-08,0.645694,0.00172122,1.36007e-06,-8.12016e-09,0.647417,0.00172392,1.33571e-06,2.67039e-09,0.649142,0.0017266,1.34372e-06,-2.56152e-09,0.65087,0.00172928,1.33604e-06,7.57571e-09,0.6526,0.00173197,1.35876e-06,-2.77413e-08,0.654334,0.00173461,1.27554e-06,4.3785e-08,0.65607,0.00173729,1.40689e-06,-2.81896e-08,0.657808,0.00174002,1.32233e-06,9.36893e-09,0.65955,0.00174269,1.35043e-06,-9.28617e-09,0.661294,0.00174536,1.32257e-06,2.77757e-08,0.66304,0.00174809,1.4059e-06,-4.2212e-08,0.66479,0.00175078,1.27926e-06,2.1863e-08,0.666542,0.0017534,1.34485e-06,1.43648e-08,0.668297,0.00175613,1.38795e-06,-1.97177e-08,0.670054,0.00175885,1.3288e-06,4.90115e-09,0.671814,0.00176152,1.3435e-06,1.13232e-10,0.673577,0.00176421,1.34384e-06,-5.3542e-09,0.675343,0.00176688,1.32778e-06,2.13035e-08,0.677111,0.0017696,1.39169e-06,-2.02553e-08,0.678882,0.00177232,1.33092e-06,1.13005e-10,0.680656,0.00177499,1.33126e-06,1.98031e-08,0.682432,0.00177771,1.39067e-06,-1.97211e-08,0.684211,0.00178043,1.33151e-06,-5.2349e-10,0.685993,0.00178309,1.32994e-06,2.18151e-08,0.687777,0.00178582,1.39538e-06,-2.71325e-08,0.689564,0.00178853,1.31398e-06,2.71101e-08,0.691354,0.00179124,1.39531e-06,-2.17035e-08,0.693147,0.00179396,1.3302e-06,9.92865e-11,0.694942,0.00179662,1.3305e-06,2.13063e-08,0.69674,0.00179935,1.39442e-06,-2.57198e-08,0.698541,0.00180206,1.31726e-06,2.19682e-08,0.700344,0.00180476,1.38317e-06,-2.54852e-09,0.70215,0.00180752,1.37552e-06,-1.17741e-08,0.703959,0.00181023,1.3402e-06,-9.95999e-09,0.705771,0.00181288,1.31032e-06,5.16141e-08,0.707585,0.00181566,1.46516e-06,-7.72869e-08,0.709402,0.00181836,1.2333e-06,7.87197e-08,0.711222,0.00182106,1.46946e-06,-5.87781e-08,0.713044,0.00182382,1.29312e-06,3.71834e-08,0.714869,0.00182652,1.40467e-06,-3.03511e-08,0.716697,0.00182924,1.31362e-06,2.46161e-08,0.718528,0.00183194,1.38747e-06,-8.5087e-09,0.720361,0.00183469,1.36194e-06,9.41892e-09,0.722197,0.00183744,1.3902e-06,-2.91671e-08,0.724036,0.00184014,1.3027e-06,4.76448e-08,0.725878,0.00184288,1.44563e-06,-4.22028e-08,0.727722,0.00184565,1.31902e-06,1.95682e-09,0.729569,0.00184829,1.3249e-06,3.43754e-08,0.731419,0.00185104,1.42802e-06,-2.0249e-08,0.733271,0.00185384,1.36727e-06,-1.29838e-08,0.735126,0.00185654,1.32832e-06,1.25794e-08,0.736984,0.00185923,1.36606e-06,2.22711e-08,0.738845,0.00186203,1.43287e-06,-4.20594e-08,0.740708,0.00186477,1.3067e-06,2.67571e-08,0.742574,0.00186746,1.38697e-06,-5.36424e-09,0.744443,0.00187022,1.37087e-06,-5.30023e-09,0.746315,0.00187295,1.35497e-06,2.65653e-08,0.748189,0.00187574,1.43467e-06,-4.13564e-08,0.750066,0.00187848,1.3106e-06,1.9651e-08,0.751946,0.00188116,1.36955e-06,2.23572e-08,0.753828,0.00188397,1.43663e-06,-4.9475e-08,0.755714,0.00188669,1.2882e-06,5.63335e-08,0.757602,0.00188944,1.4572e-06,-5.66499e-08,0.759493,0.00189218,1.28725e-06,5.10567e-08,0.761386,0.00189491,1.44042e-06,-2.83677e-08,0.763283,0.00189771,1.35532e-06,2.80962e-09,0.765182,0.00190042,1.36375e-06,1.71293e-08,0.767083,0.0019032,1.41513e-06,-1.17221e-08,0.768988,0.001906,1.37997e-06,-2.98453e-08,0.770895,0.00190867,1.29043e-06,7.14987e-08,0.772805,0.00191146,1.50493e-06,-7.73354e-08,0.774718,0.00191424,1.27292e-06,5.90292e-08,0.776634,0.00191697,1.45001e-06,-3.9572e-08,0.778552,0.00191975,1.33129e-06,3.9654e-08,0.780473,0.00192253,1.45026e-06,-5.94395e-08,0.782397,0.00192525,1.27194e-06,7.88945e-08,0.784324,0.00192803,1.50862e-06,-7.73249e-08,0.786253,0.00193082,1.27665e-06,5.15913e-08,0.788185,0.00193352,1.43142e-06,-9.83099e-09,0.79012,0.00193636,1.40193e-06,-1.22672e-08,0.792058,0.00193912,1.36513e-06,-7.05275e-10,0.793999,0.00194185,1.36301e-06,1.50883e-08,0.795942,0.00194462,1.40828e-06,-4.33147e-11,0.797888,0.00194744,1.40815e-06,-1.49151e-08,0.799837,0.00195021,1.3634e-06,9.93244e-11,0.801788,0.00195294,1.3637e-06,1.45179e-08,0.803743,0.00195571,1.40725e-06,1.43363e-09,0.8057,0.00195853,1.41155e-06,-2.02525e-08,0.80766,0.00196129,1.35079e-06,1.99718e-08,0.809622,0.00196405,1.41071e-06,-3.01649e-11,0.811588,0.00196687,1.41062e-06,-1.9851e-08,0.813556,0.00196964,1.35107e-06,1.98296e-08,0.815527,0.0019724,1.41056e-06,1.37485e-10,0.817501,0.00197522,1.41097e-06,-2.03796e-08,0.819477,0.00197798,1.34983e-06,2.17763e-08,0.821457,0.00198074,1.41516e-06,-7.12085e-09,0.823439,0.00198355,1.3938e-06,6.70707e-09,0.825424,0.00198636,1.41392e-06,-1.97074e-08,0.827412,0.00198913,1.35479e-06,1.25179e-08,0.829402,0.00199188,1.39235e-06,2.92405e-08,0.831396,0.00199475,1.48007e-06,-6.98755e-08,0.833392,0.0019975,1.27044e-06,7.14477e-08,0.835391,0.00200026,1.48479e-06,-3.71014e-08,0.837392,0.00200311,1.37348e-06,1.73533e-08,0.839397,0.00200591,1.42554e-06,-3.23118e-08,0.841404,0.00200867,1.32861e-06,5.2289e-08,0.843414,0.00201148,1.48547e-06,-5.76348e-08,0.845427,0.00201428,1.31257e-06,5.9041e-08,0.847443,0.00201708,1.48969e-06,-5.93197e-08,0.849461,0.00201988,1.31173e-06,5.90289e-08,0.851482,0.00202268,1.48882e-06,-5.75864e-08,0.853507,0.00202549,1.31606e-06,5.21075e-08,0.855533,0.00202828,1.47238e-06,-3.16344e-08,0.857563,0.00203113,1.37748e-06,1.48257e-08,0.859596,0.00203393,1.42196e-06,-2.76684e-08,0.861631,0.00203669,1.33895e-06,3.62433e-08,0.863669,0.00203947,1.44768e-06,1.90463e-09,0.86571,0.00204237,1.45339e-06,-4.38617e-08,0.867754,0.00204515,1.32181e-06,5.43328e-08,0.8698,0.00204796,1.48481e-06,-5.42603e-08,0.87185,0.00205076,1.32203e-06,4.34989e-08,0.873902,0.00205354,1.45252e-06,-5.26029e-10,0.875957,0.00205644,1.45095e-06,-4.13949e-08,0.878015,0.00205922,1.32676e-06,4.68962e-08,0.880075,0.00206201,1.46745e-06,-2.69807e-08,0.882139,0.00206487,1.38651e-06,1.42181e-09,0.884205,0.00206764,1.39077e-06,2.12935e-08,0.886274,0.00207049,1.45465e-06,-2.69912e-08,0.888346,0.00207332,1.37368e-06,2.70664e-08,0.890421,0.00207615,1.45488e-06,-2.16698e-08,0.892498,0.00207899,1.38987e-06,8.14756e-12,0.894579,0.00208177,1.38989e-06,2.16371e-08,0.896662,0.00208462,1.45481e-06,-2.6952e-08,0.898748,0.00208744,1.37395e-06,2.65663e-08,0.900837,0.00209027,1.45365e-06,-1.97084e-08,0.902928,0.00209312,1.39452e-06,-7.33731e-09,0.905023,0.00209589,1.37251e-06,4.90578e-08,0.90712,0.00209878,1.51968e-06,-6.96845e-08,0.90922,0.00210161,1.31063e-06,5.08664e-08,0.911323,0.00210438,1.46323e-06,-1.45717e-08,0.913429,0.00210727,1.41952e-06,7.42038e-09,0.915538,0.00211013,1.44178e-06,-1.51097e-08,0.917649,0.00211297,1.39645e-06,-6.58618e-09,0.919764,0.00211574,1.37669e-06,4.14545e-08,0.921881,0.00211862,1.50105e-06,-4.00222e-08,0.924001,0.0021215,1.38099e-06,-5.7518e-10,0.926124,0.00212426,1.37926e-06,4.23229e-08,0.92825,0.00212714,1.50623e-06,-4.9507e-08,0.930378,0.00213001,1.35771e-06,3.64958e-08,0.93251,0.00213283,1.4672e-06,-3.68713e-08,0.934644,0.00213566,1.35658e-06,5.13848e-08,0.936781,0.00213852,1.51074e-06,-4.94585e-08,0.938921,0.0021414,1.36236e-06,2.72399e-08,0.941064,0.0021442,1.44408e-06,1.0372e-10,0.943209,0.00214709,1.44439e-06,-2.76547e-08,0.945358,0.0021499,1.36143e-06,5.09106e-08,0.947509,0.00215277,1.51416e-06,-5.67784e-08,0.949663,0.00215563,1.34382e-06,5.69935e-08,0.95182,0.00215849,1.5148e-06,-5.19861e-08,0.95398,0.00216136,1.35885e-06,3.17417e-08,0.956143,0.00216418,1.45407e-06,-1.53758e-08,0.958309,0.00216704,1.40794e-06,2.97615e-08,0.960477,0.00216994,1.49723e-06,-4.40657e-08,0.962649,0.00217281,1.36503e-06,2.72919e-08,0.964823,0.00217562,1.44691e-06,-5.49729e-09,0.967,0.0021785,1.43041e-06,-5.30273e-09,0.96918,0.00218134,1.41451e-06,2.67084e-08,0.971363,0.00218425,1.49463e-06,-4.19265e-08,0.973548,0.00218711,1.36885e-06,2.17881e-08,0.975737,0.00218992,1.43422e-06,1.43789e-08,0.977928,0.00219283,1.47735e-06,-1.96989e-08,0.980122,0.00219572,1.41826e-06,4.81221e-09,0.98232,0.00219857,1.43269e-06,4.50048e-10,0.98452,0.00220144,1.43404e-06,-6.61237e-09,0.986722,0.00220429,1.41421e-06,2.59993e-08,0.988928,0.0022072,1.4922e-06,-3.77803e-08,0.991137,0.00221007,1.37886e-06,5.9127e-09,0.993348,0.00221284,1.3966e-06,1.33339e-07,0.995563,0.00221604,1.79662e-06,-5.98872e-07,0.99778,0.00222015,0.,0.};
+
+    __device__ static int LabCbrt_b(int i)
+    {
+        float x = i * (1.f / (255.f * (1 << gamma_shift)));
+        return (1 << lab_shift2) * (x < 0.008856f ? x * 7.787f + 0.13793103448275862f : ::cbrtf(x));
+    }
+
+    __device__ static float splineInterpolate(float x, const float* tab, int n)
+    {
+        int ix = ::min(::max(int(x), 0), n-1);
+        x -= ix;
+        tab += ix * 4;
+        return ((tab[3] * x + tab[2]) * x + tab[1]) * x + tab[0];
+    }
+
+    template <typename T, int scn, int dcn, bool srgb, int blueIdx> struct RGB2Lab;
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct RGB2Lab<uchar, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            const int Lscale = (116 * 255 + 50) / 100;
+            const int Lshift = -((16 * 255 * (1 << lab_shift2) + 50) / 100);
+
+            int B = blueIdx == 0 ? src.x : src.z;
+            int G = src.y;
+            int R = blueIdx == 0 ? src.z : src.x;
+
+            if (srgb)
+            {
+                B = c_sRGBGammaTab_b[B];
+                G = c_sRGBGammaTab_b[G];
+                R = c_sRGBGammaTab_b[R];
+            }
+            else
+            {
+                B <<= 3;
+                G <<= 3;
+                R <<= 3;
+            }
+
+            int fX = LabCbrt_b(CV_CUDEV_DESCALE(B * 778 + G * 1541 + R * 1777, lab_shift));
+            int fY = LabCbrt_b(CV_CUDEV_DESCALE(B * 296 + G * 2929 + R * 871, lab_shift));
+            int fZ = LabCbrt_b(CV_CUDEV_DESCALE(B * 3575 + G * 448 + R * 73, lab_shift));
+
+            int L = CV_CUDEV_DESCALE(Lscale * fY + Lshift, lab_shift2);
+            int a = CV_CUDEV_DESCALE(500 * (fX - fY) + 128 * (1 << lab_shift2), lab_shift2);
+            int b = CV_CUDEV_DESCALE(200 * (fY - fZ) + 128 * (1 << lab_shift2), lab_shift2);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(L);
+            dst.y = saturate_cast<uchar>(a);
+            dst.z = saturate_cast<uchar>(b);
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct RGB2Lab<float, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float _1_3 = 1.0f / 3.0f;
+            const float _a = 16.0f / 116.0f;
+
+            float B = blueIdx == 0 ? src.x : src.z;
+            float G = src.y;
+            float R = blueIdx == 0 ? src.z : src.x;
+
+            if (srgb)
+            {
+                B = splineInterpolate(B * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+                G = splineInterpolate(G * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+                R = splineInterpolate(R * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+            }
+
+            float X = B * 0.189828f + G * 0.376219f + R * 0.433953f;
+            float Y = B * 0.072169f + G * 0.715160f + R * 0.212671f;
+            float Z = B * 0.872766f + G * 0.109477f + R * 0.017758f;
+
+            float FX = X > 0.008856f ? ::powf(X, _1_3) : (7.787f * X + _a);
+            float FY = Y > 0.008856f ? ::powf(Y, _1_3) : (7.787f * Y + _a);
+            float FZ = Z > 0.008856f ? ::powf(Z, _1_3) : (7.787f * Z + _a);
+
+            float L = Y > 0.008856f ? (116.f * FY - 16.f) : (903.3f * Y);
+            float a = 500.f * (FX - FY);
+            float b = 200.f * (FY - FZ);
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = L;
+            dst.y = a;
+            dst.z = b;
+
+            return dst;
+        }
+    };
+
+    // Lab to RGB
+
+    __constant__ float c_sRGBInvGammaTab[] = {0,0.0126255,0.,-8.33961e-06,0.0126172,0.0126005,-2.50188e-05,4.1698e-05,0.0252344,0.0126756,0.000100075,-0.000158451,0.0378516,0.0124004,-0.000375277,-0.000207393,0.0496693,0.0110276,-0.000997456,0.00016837,0.0598678,0.00953783,-0.000492346,2.07235e-05,0.068934,0.00861531,-0.000430176,3.62876e-05,0.0771554,0.00786382,-0.000321313,1.87625e-05,0.0847167,0.00727748,-0.000265025,1.53594e-05,0.0917445,0.00679351,-0.000218947,1.10545e-05,0.0983301,0.00638877,-0.000185784,8.66984e-06,0.104542,0.00604322,-0.000159774,6.82996e-06,0.110432,0.00574416,-0.000139284,5.51008e-06,0.116042,0.00548212,-0.000122754,4.52322e-06,0.121406,0.00525018,-0.000109184,3.75557e-06,0.126551,0.00504308,-9.79177e-05,3.17134e-06,0.131499,0.00485676,-8.84037e-05,2.68469e-06,0.13627,0.004688,-8.03496e-05,2.31725e-06,0.14088,0.00453426,-7.33978e-05,2.00868e-06,0.145343,0.00439349,-6.73718e-05,1.74775e-06,0.149671,0.00426399,-6.21286e-05,1.53547e-06,0.153875,0.00414434,-5.75222e-05,1.364e-06,0.157963,0.00403338,-5.34301e-05,1.20416e-06,0.161944,0.00393014,-4.98177e-05,1.09114e-06,0.165825,0.00383377,-4.65443e-05,9.57987e-07,0.169613,0.00374356,-4.36703e-05,8.88359e-07,0.173314,0.00365888,-4.10052e-05,7.7849e-07,0.176933,0.00357921,-3.86697e-05,7.36254e-07,0.180474,0.00350408,-3.6461e-05,6.42534e-07,0.183942,0.00343308,-3.45334e-05,6.12614e-07,0.187342,0.00336586,-3.26955e-05,5.42894e-07,0.190675,0.00330209,-3.10669e-05,5.08967e-07,0.193947,0.00324149,-2.954e-05,4.75977e-07,0.197159,0.00318383,-2.8112e-05,4.18343e-07,0.200315,0.00312887,-2.6857e-05,4.13651e-07,0.203418,0.00307639,-2.5616e-05,3.70847e-07,0.206469,0.00302627,-2.45035e-05,3.3813e-07,0.209471,0.00297828,-2.34891e-05,3.32999e-07,0.212426,0.0029323,-2.24901e-05,2.96826e-07,0.215336,0.00288821,-2.15996e-05,2.82736e-07,0.218203,0.00284586,-2.07514e-05,2.70961e-07,0.221029,0.00280517,-1.99385e-05,2.42744e-07,0.223814,0.00276602,-1.92103e-05,2.33277e-07,0.226561,0.0027283,-1.85105e-05,2.2486e-07,0.229271,0.00269195,-1.78359e-05,2.08383e-07,0.231945,0.00265691,-1.72108e-05,1.93305e-07,0.234585,0.00262307,-1.66308e-05,1.80687e-07,0.237192,0.00259035,-1.60888e-05,1.86632e-07,0.239766,0.00255873,-1.55289e-05,1.60569e-07,0.24231,0.00252815,-1.50472e-05,1.54566e-07,0.244823,0.00249852,-1.45835e-05,1.59939e-07,0.247307,0.00246983,-1.41037e-05,1.29549e-07,0.249763,0.00244202,-1.3715e-05,1.41429e-07,0.252191,0.00241501,-1.32907e-05,1.39198e-07,0.254593,0.00238885,-1.28731e-05,1.06444e-07,0.256969,0.00236342,-1.25538e-05,1.2048e-07,0.25932,0.00233867,-1.21924e-05,1.26892e-07,0.261647,0.00231467,-1.18117e-05,8.72084e-08,0.26395,0.00229131,-1.15501e-05,1.20323e-07,0.26623,0.00226857,-1.11891e-05,8.71514e-08,0.268487,0.00224645,-1.09276e-05,9.73165e-08,0.270723,0.00222489,-1.06357e-05,8.98259e-08,0.272937,0.00220389,-1.03662e-05,7.98218e-08,0.275131,0.00218339,-1.01267e-05,9.75254e-08,0.277304,0.00216343,-9.83416e-06,6.65195e-08,0.279458,0.00214396,-9.63461e-06,8.34313e-08,0.281592,0.00212494,-9.38431e-06,7.65919e-08,0.283708,0.00210641,-9.15454e-06,5.7236e-08,0.285805,0.00208827,-8.98283e-06,8.18939e-08,0.287885,0.00207055,-8.73715e-06,6.2224e-08,0.289946,0.00205326,-8.55047e-06,5.66388e-08,0.291991,0.00203633,-8.38056e-06,6.88491e-08,0.294019,0.00201978,-8.17401e-06,5.53955e-08,0.296031,0.00200359,-8.00782e-06,6.71971e-08,0.298027,0.00198778,-7.80623e-06,3.34439e-08,0.300007,0.00197227,-7.7059e-06,6.7248e-08,0.301971,0.00195706,-7.50416e-06,5.51915e-08,0.303921,0.00194221,-7.33858e-06,3.98124e-08,0.305856,0.00192766,-7.21915e-06,5.37795e-08,0.307776,0.00191338,-7.05781e-06,4.30919e-08,0.309683,0.00189939,-6.92853e-06,4.20744e-08,0.311575,0.00188566,-6.80231e-06,5.68321e-08,0.313454,0.00187223,-6.63181e-06,2.86195e-08,0.31532,0.00185905,-6.54595e-06,3.73075e-08,0.317172,0.00184607,-6.43403e-06,6.05684e-08,0.319012,0.00183338,-6.25233e-06,1.84426e-08,0.320839,0.00182094,-6.197e-06,4.44757e-08,0.322654,0.00180867,-6.06357e-06,4.20729e-08,0.324456,0.00179667,-5.93735e-06,2.56511e-08,0.326247,0.00178488,-5.8604e-06,3.41368e-08,0.328026,0.00177326,-5.75799e-06,4.64177e-08,0.329794,0.00176188,-5.61874e-06,1.86107e-08,0.33155,0.0017507,-5.5629e-06,2.81511e-08,0.333295,0.00173966,-5.47845e-06,4.75987e-08,0.335029,0.00172884,-5.33565e-06,1.98726e-08,0.336753,0.00171823,-5.27604e-06,2.19226e-08,0.338466,0.00170775,-5.21027e-06,4.14483e-08,0.340169,0.00169745,-5.08592e-06,2.09017e-08,0.341861,0.00168734,-5.02322e-06,2.39561e-08,0.343543,0.00167737,-4.95135e-06,3.22852e-08,0.345216,0.00166756,-4.85449e-06,2.57173e-08,0.346878,0.00165793,-4.77734e-06,1.38569e-08,0.348532,0.00164841,-4.73577e-06,3.80634e-08,0.350175,0.00163906,-4.62158e-06,1.27043e-08,0.35181,0.00162985,-4.58347e-06,3.03279e-08,0.353435,0.00162078,-4.49249e-06,1.49961e-08,0.355051,0.00161184,-4.4475e-06,2.88977e-08,0.356659,0.00160303,-4.3608e-06,1.84241e-08,0.358257,0.00159436,-4.30553e-06,1.6616e-08,0.359848,0.0015858,-4.25568e-06,3.43218e-08,0.361429,0.00157739,-4.15272e-06,-4.89172e-09,0.363002,0.00156907,-4.16739e-06,4.48498e-08,0.364567,0.00156087,-4.03284e-06,4.30676e-09,0.366124,0.00155282,-4.01992e-06,2.73303e-08,0.367673,0.00154486,-3.93793e-06,5.58036e-09,0.369214,0.001537,-3.92119e-06,3.97554e-08,0.370747,0.00152928,-3.80193e-06,-1.55904e-08,0.372272,0.00152163,-3.8487e-06,5.24081e-08,0.37379,0.00151409,-3.69147e-06,-1.52272e-08,0.375301,0.00150666,-3.73715e-06,3.83028e-08,0.376804,0.0014993,-3.62225e-06,1.10278e-08,0.378299,0.00149209,-3.58916e-06,6.99326e-09,0.379788,0.00148493,-3.56818e-06,2.06038e-08,0.381269,0.00147786,-3.50637e-06,2.98009e-08,0.382744,0.00147093,-3.41697e-06,-2.05978e-08,0.384211,0.00146404,-3.47876e-06,5.25899e-08,0.385672,0.00145724,-3.32099e-06,-1.09471e-08,0.387126,0.00145056,-3.35383e-06,2.10009e-08,0.388573,0.00144392,-3.29083e-06,1.63501e-08,0.390014,0.00143739,-3.24178e-06,3.00641e-09,0.391448,0.00143091,-3.23276e-06,3.12282e-08,0.392875,0.00142454,-3.13908e-06,-8.70932e-09,0.394297,0.00141824,-3.16521e-06,3.34114e-08,0.395712,0.00141201,-3.06497e-06,-5.72754e-09,0.397121,0.00140586,-3.08215e-06,1.9301e-08,0.398524,0.00139975,-3.02425e-06,1.7931e-08,0.39992,0.00139376,-2.97046e-06,-1.61822e-09,0.401311,0.00138781,-2.97531e-06,1.83442e-08,0.402696,0.00138192,-2.92028e-06,1.76485e-08,0.404075,0.00137613,-2.86733e-06,4.68617e-10,0.405448,0.00137039,-2.86593e-06,1.02794e-08,0.406816,0.00136469,-2.83509e-06,1.80179e-08,0.408178,0.00135908,-2.78104e-06,7.05594e-09,0.409534,0.00135354,-2.75987e-06,1.33633e-08,0.410885,0.00134806,-2.71978e-06,-9.04568e-10,0.41223,0.00134261,-2.72249e-06,2.0057e-08,0.41357,0.00133723,-2.66232e-06,1.00841e-08,0.414905,0.00133194,-2.63207e-06,-7.88835e-10,0.416234,0.00132667,-2.63444e-06,2.28734e-08,0.417558,0.00132147,-2.56582e-06,-1.29785e-09,0.418877,0.00131633,-2.56971e-06,1.21205e-08,0.420191,0.00131123,-2.53335e-06,1.24202e-08,0.421499,0.0013062,-2.49609e-06,-2.19681e-09,0.422803,0.0013012,-2.50268e-06,2.61696e-08,0.424102,0.00129628,-2.42417e-06,-1.30747e-08,0.425396,0.00129139,-2.46339e-06,2.6129e-08,0.426685,0.00128654,-2.38501e-06,-2.03454e-09,0.427969,0.00128176,-2.39111e-06,1.18115e-08,0.429248,0.00127702,-2.35567e-06,1.43932e-08,0.430523,0.00127235,-2.31249e-06,-9.77965e-09,0.431793,0.00126769,-2.34183e-06,2.47253e-08,0.433058,0.00126308,-2.26766e-06,2.85278e-10,0.434319,0.00125855,-2.2668e-06,3.93614e-09,0.435575,0.00125403,-2.25499e-06,1.37722e-08,0.436827,0.00124956,-2.21368e-06,5.79803e-10,0.438074,0.00124513,-2.21194e-06,1.37112e-08,0.439317,0.00124075,-2.1708e-06,4.17973e-09,0.440556,0.00123642,-2.15826e-06,-6.27703e-10,0.44179,0.0012321,-2.16015e-06,2.81332e-08,0.44302,0.00122787,-2.07575e-06,-2.24985e-08,0.444246,0.00122365,-2.14324e-06,3.20586e-08,0.445467,0.00121946,-2.04707e-06,-1.6329e-08,0.446685,0.00121532,-2.09605e-06,3.32573e-08,0.447898,0.00121122,-1.99628e-06,-2.72927e-08,0.449107,0.00120715,-2.07816e-06,4.6111e-08,0.450312,0.00120313,-1.93983e-06,-3.79416e-08,0.451514,0.00119914,-2.05365e-06,4.60507e-08,0.452711,0.00119517,-1.9155e-06,-2.7052e-08,0.453904,0.00119126,-1.99666e-06,3.23551e-08,0.455093,0.00118736,-1.89959e-06,-1.29613e-08,0.456279,0.00118352,-1.93848e-06,1.94905e-08,0.45746,0.0011797,-1.88e-06,-5.39588e-09,0.458638,0.00117593,-1.89619e-06,2.09282e-09,0.459812,0.00117214,-1.88991e-06,2.68267e-08,0.460982,0.00116844,-1.80943e-06,-1.99925e-08,0.462149,0.00116476,-1.86941e-06,2.3341e-08,0.463312,0.00116109,-1.79939e-06,-1.37674e-08,0.464471,0.00115745,-1.84069e-06,3.17287e-08,0.465627,0.00115387,-1.7455e-06,-2.37407e-08,0.466779,0.00115031,-1.81673e-06,3.34315e-08,0.467927,0.00114677,-1.71643e-06,-2.05786e-08,0.469073,0.00114328,-1.77817e-06,1.90802e-08,0.470214,0.00113978,-1.72093e-06,3.86247e-09,0.471352,0.00113635,-1.70934e-06,-4.72759e-09,0.472487,0.00113292,-1.72352e-06,1.50478e-08,0.473618,0.00112951,-1.67838e-06,4.14108e-09,0.474746,0.00112617,-1.66595e-06,-1.80986e-09,0.47587,0.00112283,-1.67138e-06,3.09816e-09,0.476991,0.0011195,-1.66209e-06,1.92198e-08,0.478109,0.00111623,-1.60443e-06,-2.03726e-08,0.479224,0.00111296,-1.66555e-06,3.2468e-08,0.480335,0.00110973,-1.56814e-06,-2.00922e-08,0.481443,0.00110653,-1.62842e-06,1.80983e-08,0.482548,0.00110333,-1.57413e-06,7.30362e-09,0.48365,0.0011002,-1.55221e-06,-1.75107e-08,0.484749,0.00109705,-1.60475e-06,3.29373e-08,0.485844,0.00109393,-1.50594e-06,-2.48315e-08,0.486937,0.00109085,-1.58043e-06,3.65865e-08,0.488026,0.0010878,-1.47067e-06,-3.21078e-08,0.489112,0.00108476,-1.56699e-06,3.22397e-08,0.490195,0.00108172,-1.47027e-06,-7.44391e-09,0.491276,0.00107876,-1.49261e-06,-2.46428e-09,0.492353,0.00107577,-1.5e-06,1.73011e-08,0.493427,0.00107282,-1.4481e-06,-7.13552e-09,0.494499,0.0010699,-1.4695e-06,1.1241e-08,0.495567,0.001067,-1.43578e-06,-8.02637e-09,0.496633,0.0010641,-1.45986e-06,2.08645e-08,0.497695,0.00106124,-1.39726e-06,-1.58271e-08,0.498755,0.0010584,-1.44475e-06,1.26415e-08,0.499812,0.00105555,-1.40682e-06,2.48655e-08,0.500866,0.00105281,-1.33222e-06,-5.24988e-08,0.501918,0.00104999,-1.48972e-06,6.59206e-08,0.502966,0.00104721,-1.29196e-06,-3.237e-08,0.504012,0.00104453,-1.38907e-06,3.95479e-09,0.505055,0.00104176,-1.3772e-06,1.65509e-08,0.506096,0.00103905,-1.32755e-06,-1.05539e-08,0.507133,0.00103637,-1.35921e-06,2.56648e-08,0.508168,0.00103373,-1.28222e-06,-3.25007e-08,0.509201,0.00103106,-1.37972e-06,4.47336e-08,0.51023,0.00102844,-1.24552e-06,-2.72245e-08,0.511258,0.00102587,-1.32719e-06,4.55952e-09,0.512282,0.00102323,-1.31352e-06,8.98645e-09,0.513304,0.00102063,-1.28656e-06,1.90992e-08,0.514323,0.00101811,-1.22926e-06,-2.57786e-08,0.51534,0.00101557,-1.30659e-06,2.44104e-08,0.516355,0.00101303,-1.23336e-06,-1.22581e-08,0.517366,0.00101053,-1.27014e-06,2.4622e-08,0.518376,0.00100806,-1.19627e-06,-2.66253e-08,0.519383,0.00100559,-1.27615e-06,2.22744e-08,0.520387,0.00100311,-1.20932e-06,-2.8679e-09,0.521389,0.00100068,-1.21793e-06,-1.08029e-08,0.522388,0.000998211,-1.25034e-06,4.60795e-08,0.523385,0.000995849,-1.1121e-06,-5.4306e-08,0.52438,0.000993462,-1.27502e-06,5.19354e-08,0.525372,0.000991067,-1.11921e-06,-3.42262e-08,0.526362,0.000988726,-1.22189e-06,2.53646e-08,0.52735,0.000986359,-1.14579e-06,-7.62782e-09,0.528335,0.000984044,-1.16868e-06,5.14668e-09,0.529318,0.000981722,-1.15324e-06,-1.29589e-08,0.530298,0.000979377,-1.19211e-06,4.66888e-08,0.531276,0.000977133,-1.05205e-06,-5.45868e-08,0.532252,0.000974865,-1.21581e-06,5.24495e-08,0.533226,0.000972591,-1.05846e-06,-3.60019e-08,0.534198,0.000970366,-1.16647e-06,3.19537e-08,0.535167,0.000968129,-1.07061e-06,-3.2208e-08,0.536134,0.000965891,-1.16723e-06,3.72738e-08,0.537099,0.000963668,-1.05541e-06,2.32205e-09,0.538061,0.000961564,-1.04844e-06,-4.65618e-08,0.539022,0.000959328,-1.18813e-06,6.47159e-08,0.53998,0.000957146,-9.93979e-07,-3.3488e-08,0.540936,0.000955057,-1.09444e-06,9.63166e-09,0.54189,0.000952897,-1.06555e-06,-5.03871e-09,0.542842,0.000950751,-1.08066e-06,1.05232e-08,0.543792,0.000948621,-1.04909e-06,2.25503e-08,0.544739,0.000946591,-9.81444e-07,-4.11195e-08,0.545685,0.000944504,-1.1048e-06,2.27182e-08,0.546628,0.000942363,-1.03665e-06,9.85146e-09,0.54757,0.000940319,-1.00709e-06,-2.51938e-09,0.548509,0.000938297,-1.01465e-06,2.25858e-10,0.549446,0.000936269,-1.01397e-06,1.61598e-09,0.550381,0.000934246,-1.00913e-06,-6.68983e-09,0.551315,0.000932207,-1.0292e-06,2.51434e-08,0.552246,0.000930224,-9.53765e-07,-3.42793e-08,0.553175,0.000928214,-1.0566e-06,5.23688e-08,0.554102,0.000926258,-8.99497e-07,-5.59865e-08,0.555028,0.000924291,-1.06746e-06,5.23679e-08,0.555951,0.000922313,-9.10352e-07,-3.42763e-08,0.556872,0.00092039,-1.01318e-06,2.51326e-08,0.557792,0.000918439,-9.37783e-07,-6.64954e-09,0.558709,0.000916543,-9.57732e-07,1.46554e-09,0.559625,0.000914632,-9.53335e-07,7.87281e-10,0.560538,0.000912728,-9.50973e-07,-4.61466e-09,0.56145,0.000910812,-9.64817e-07,1.76713e-08,0.56236,0.000908935,-9.11804e-07,-6.46564e-09,0.563268,0.000907092,-9.312e-07,8.19121e-09,0.564174,0.000905255,-9.06627e-07,-2.62992e-08,0.565078,0.000903362,-9.85524e-07,3.74007e-08,0.565981,0.000901504,-8.73322e-07,-4.0942e-09,0.566882,0.000899745,-8.85605e-07,-2.1024e-08,0.56778,0.00089791,-9.48677e-07,2.85854e-08,0.568677,0.000896099,-8.62921e-07,-3.3713e-08,0.569573,0.000894272,-9.64059e-07,4.6662e-08,0.570466,0.000892484,-8.24073e-07,-3.37258e-08,0.571358,0.000890734,-9.25251e-07,2.86365e-08,0.572247,0.00088897,-8.39341e-07,-2.12155e-08,0.573135,0.000887227,-9.02988e-07,-3.37913e-09,0.574022,0.000885411,-9.13125e-07,3.47319e-08,0.574906,0.000883689,-8.08929e-07,-1.63394e-08,0.575789,0.000882022,-8.57947e-07,-2.8979e-08,0.57667,0.00088022,-9.44885e-07,7.26509e-08,0.57755,0.000878548,-7.26932e-07,-8.28106e-08,0.578427,0.000876845,-9.75364e-07,7.97774e-08,0.579303,0.000875134,-7.36032e-07,-5.74849e-08,0.580178,0.00087349,-9.08486e-07,3.09529e-08,0.58105,0.000871765,-8.15628e-07,-6.72206e-09,0.581921,0.000870114,-8.35794e-07,-4.06451e-09,0.582791,0.00086843,-8.47987e-07,2.29799e-08,0.583658,0.000866803,-7.79048e-07,-2.82503e-08,0.584524,0.00086516,-8.63799e-07,3.04167e-08,0.585388,0.000863524,-7.72548e-07,-3.38119e-08,0.586251,0.000861877,-8.73984e-07,4.52264e-08,0.587112,0.000860265,-7.38305e-07,-2.78842e-08,0.587972,0.000858705,-8.21958e-07,6.70567e-09,0.58883,0.000857081,-8.01841e-07,1.06161e-09,0.589686,0.000855481,-7.98656e-07,-1.09521e-08,0.590541,0.00085385,-8.31512e-07,4.27468e-08,0.591394,0.000852316,-7.03272e-07,-4.08257e-08,0.592245,0.000850787,-8.25749e-07,1.34677e-09,0.593095,0.000849139,-8.21709e-07,3.54387e-08,0.593944,0.000847602,-7.15393e-07,-2.38924e-08,0.59479,0.0008461,-7.8707e-07,5.26143e-10,0.595636,0.000844527,-7.85491e-07,2.17879e-08,0.596479,0.000843021,-7.20127e-07,-2.80733e-08,0.597322,0.000841497,-8.04347e-07,3.09005e-08,0.598162,0.000839981,-7.11646e-07,-3.5924e-08,0.599002,0.00083845,-8.19418e-07,5.3191e-08,0.599839,0.000836971,-6.59845e-07,-5.76307e-08,0.600676,0.000835478,-8.32737e-07,5.81227e-08,0.60151,0.000833987,-6.58369e-07,-5.56507e-08,0.602344,0.000832503,-8.25321e-07,4.52706e-08,0.603175,0.000830988,-6.89509e-07,-6.22236e-09,0.604006,0.000829591,-7.08176e-07,-2.03811e-08,0.604834,0.000828113,-7.6932e-07,2.8142e-08,0.605662,0.000826659,-6.84894e-07,-3.25822e-08,0.606488,0.000825191,-7.8264e-07,4.25823e-08,0.607312,0.000823754,-6.54893e-07,-1.85376e-08,0.608135,0.000822389,-7.10506e-07,-2.80365e-08,0.608957,0.000820883,-7.94616e-07,7.1079e-08,0.609777,0.000819507,-5.81379e-07,-7.74655e-08,0.610596,0.000818112,-8.13775e-07,5.9969e-08,0.611413,0.000816665,-6.33868e-07,-4.32013e-08,0.612229,0.000815267,-7.63472e-07,5.32313e-08,0.613044,0.0008139,-6.03778e-07,-5.05148e-08,0.613857,0.000812541,-7.55323e-07,2.96187e-08,0.614669,0.000811119,-6.66466e-07,-8.35545e-09,0.615479,0.000809761,-6.91533e-07,3.80301e-09,0.616288,0.00080839,-6.80124e-07,-6.85666e-09,0.617096,0.000807009,-7.00694e-07,2.36237e-08,0.617903,0.000805678,-6.29822e-07,-2.80336e-08,0.618708,0.000804334,-7.13923e-07,2.8906e-08,0.619511,0.000802993,-6.27205e-07,-2.79859e-08,0.620314,0.000801655,-7.11163e-07,2.34329e-08,0.621114,0.000800303,-6.40864e-07,-6.14108e-09,0.621914,0.000799003,-6.59287e-07,1.13151e-09,0.622712,0.000797688,-6.55893e-07,1.61507e-09,0.62351,0.000796381,-6.51048e-07,-7.59186e-09,0.624305,0.000795056,-6.73823e-07,2.87524e-08,0.6251,0.000793794,-5.87566e-07,-4.7813e-08,0.625893,0.000792476,-7.31005e-07,4.32901e-08,0.626685,0.000791144,-6.01135e-07,-6.13814e-09,0.627475,0.000789923,-6.19549e-07,-1.87376e-08,0.628264,0.000788628,-6.75762e-07,2.14837e-08,0.629052,0.000787341,-6.11311e-07,-7.59265e-09,0.629839,0.000786095,-6.34089e-07,8.88692e-09,0.630625,0.000784854,-6.07428e-07,-2.7955e-08,0.631409,0.000783555,-6.91293e-07,4.33285e-08,0.632192,0.000782302,-5.61307e-07,-2.61497e-08,0.632973,0.000781101,-6.39757e-07,1.6658e-09,0.633754,0.000779827,-6.34759e-07,1.94866e-08,0.634533,0.000778616,-5.76299e-07,-2.00076e-08,0.635311,0.000777403,-6.36322e-07,9.39091e-10,0.636088,0.000776133,-6.33505e-07,1.62512e-08,0.636863,0.000774915,-5.84751e-07,-6.33937e-09,0.637638,0.000773726,-6.03769e-07,9.10609e-09,0.638411,0.000772546,-5.76451e-07,-3.00849e-08,0.639183,0.000771303,-6.66706e-07,5.1629e-08,0.639953,0.000770125,-5.11819e-07,-5.7222e-08,0.640723,0.000768929,-6.83485e-07,5.80497e-08,0.641491,0.000767736,-5.09336e-07,-5.57674e-08,0.642259,0.000766551,-6.76638e-07,4.58105e-08,0.643024,0.000765335,-5.39206e-07,-8.26541e-09,0.643789,0.000764231,-5.64002e-07,-1.27488e-08,0.644553,0.000763065,-6.02249e-07,-3.44168e-10,0.645315,0.00076186,-6.03281e-07,1.41254e-08,0.646077,0.000760695,-5.60905e-07,3.44727e-09,0.646837,0.000759584,-5.50563e-07,-2.79144e-08,0.647596,0.000758399,-6.34307e-07,4.86057e-08,0.648354,0.000757276,-4.88489e-07,-4.72989e-08,0.64911,0.000756158,-6.30386e-07,2.13807e-08,0.649866,0.000754961,-5.66244e-07,2.13808e-08,0.65062,0.000753893,-5.02102e-07,-4.7299e-08,0.651374,0.000752746,-6.43999e-07,4.86059e-08,0.652126,0.000751604,-4.98181e-07,-2.79154e-08,0.652877,0.000750524,-5.81927e-07,3.45089e-09,0.653627,0.000749371,-5.71575e-07,1.41119e-08,0.654376,0.00074827,-5.29239e-07,-2.93748e-10,0.655123,0.00074721,-5.3012e-07,-1.29368e-08,0.65587,0.000746111,-5.68931e-07,-7.56355e-09,0.656616,0.000744951,-5.91621e-07,4.3191e-08,0.65736,0.000743897,-4.62048e-07,-4.59911e-08,0.658103,0.000742835,-6.00022e-07,2.15642e-08,0.658846,0.0007417,-5.35329e-07,1.93389e-08,0.659587,0.000740687,-4.77312e-07,-3.93152e-08,0.660327,0.000739615,-5.95258e-07,1.87126e-08,0.661066,0.00073848,-5.3912e-07,2.40695e-08,0.661804,0.000737474,-4.66912e-07,-5.53859e-08,0.662541,0.000736374,-6.33069e-07,7.82648e-08,0.663277,0.000735343,-3.98275e-07,-7.88593e-08,0.664012,0.00073431,-6.34853e-07,5.83585e-08,0.664745,0.000733215,-4.59777e-07,-3.53656e-08,0.665478,0.000732189,-5.65874e-07,2.34994e-08,0.66621,0.000731128,-4.95376e-07,9.72743e-10,0.66694,0.00073014,-4.92458e-07,-2.73903e-08,0.66767,0.000729073,-5.74629e-07,4.89839e-08,0.668398,0.000728071,-4.27677e-07,-4.93359e-08,0.669126,0.000727068,-5.75685e-07,2.91504e-08,0.669853,0.000726004,-4.88234e-07,-7.66109e-09,0.670578,0.000725004,-5.11217e-07,1.49392e-09,0.671303,0.000723986,-5.06735e-07,1.68533e-09,0.672026,0.000722978,-5.01679e-07,-8.23525e-09,0.672749,0.00072195,-5.26385e-07,3.12556e-08,0.67347,0.000720991,-4.32618e-07,-5.71825e-08,0.674191,0.000719954,-6.04166e-07,7.8265e-08,0.67491,0.00071898,-3.69371e-07,-7.70634e-08,0.675628,0.00071801,-6.00561e-07,5.11747e-08,0.676346,0.000716963,-4.47037e-07,-8.42615e-09,0.677062,0.000716044,-4.72315e-07,-1.747e-08,0.677778,0.000715046,-5.24725e-07,1.87015e-08,0.678493,0.000714053,-4.68621e-07,2.26856e-09,0.679206,0.000713123,-4.61815e-07,-2.77758e-08,0.679919,0.000712116,-5.45142e-07,4.92298e-08,0.68063,0.000711173,-3.97453e-07,-4.99339e-08,0.681341,0.000710228,-5.47255e-07,3.12967e-08,0.682051,0.000709228,-4.53365e-07,-1.56481e-08,0.68276,0.000708274,-5.00309e-07,3.12958e-08,0.683467,0.000707367,-4.06422e-07,-4.99303e-08,0.684174,0.000706405,-5.56213e-07,4.9216e-08,0.68488,0.00070544,-4.08565e-07,-2.77245e-08,0.685585,0.00070454,-4.91738e-07,2.07748e-09,0.686289,0.000703562,-4.85506e-07,1.94146e-08,0.686992,0.00070265,-4.27262e-07,-2.01314e-08,0.687695,0.000701735,-4.87656e-07,1.50616e-09,0.688396,0.000700764,-4.83137e-07,1.41067e-08,0.689096,0.00069984,-4.40817e-07,1.67168e-09,0.689795,0.000698963,-4.35802e-07,-2.07934e-08,0.690494,0.000698029,-4.98182e-07,2.18972e-08,0.691192,0.000697099,-4.32491e-07,-7.19092e-09,0.691888,0.000696212,-4.54064e-07,6.86642e-09,0.692584,0.000695325,-4.33464e-07,-2.02747e-08,0.693279,0.000694397,-4.94288e-07,1.46279e-08,0.693973,0.000693452,-4.50405e-07,2.13678e-08,0.694666,0.000692616,-3.86301e-07,-4.04945e-08,0.695358,0.000691721,-5.07785e-07,2.14009e-08,0.696049,0.00069077,-4.43582e-07,1.44955e-08,0.69674,0.000689926,-4.00096e-07,-1.97783e-08,0.697429,0.000689067,-4.5943e-07,5.01296e-09,0.698118,0.000688163,-4.44392e-07,-2.73521e-10,0.698805,0.000687273,-4.45212e-07,-3.91893e-09,0.699492,0.000686371,-4.56969e-07,1.59493e-08,0.700178,0.000685505,-4.09121e-07,-2.73351e-10,0.700863,0.000684686,-4.09941e-07,-1.4856e-08,0.701548,0.000683822,-4.54509e-07,9.25979e-11,0.702231,0.000682913,-4.54231e-07,1.44855e-08,0.702913,0.000682048,-4.10775e-07,1.56992e-09,0.703595,0.000681231,-4.06065e-07,-2.07652e-08,0.704276,0.000680357,-4.68361e-07,2.18864e-08,0.704956,0.000679486,-4.02701e-07,-7.17595e-09,0.705635,0.000678659,-4.24229e-07,6.81748e-09,0.706313,0.000677831,-4.03777e-07,-2.0094e-08,0.70699,0.000676963,-4.64059e-07,1.39538e-08,0.707667,0.000676077,-4.22197e-07,2.38835e-08,0.708343,0.000675304,-3.50547e-07,-4.98831e-08,0.709018,0.000674453,-5.00196e-07,5.64395e-08,0.709692,0.000673622,-3.30878e-07,-5.66657e-08,0.710365,0.00067279,-5.00875e-07,5.1014e-08,0.711037,0.000671942,-3.47833e-07,-2.81809e-08,0.711709,0.000671161,-4.32376e-07,2.10513e-09,0.712379,0.000670303,-4.2606e-07,1.97604e-08,0.713049,0.00066951,-3.66779e-07,-2.15422e-08,0.713718,0.000668712,-4.31406e-07,6.8038e-09,0.714387,0.000667869,-4.10994e-07,-5.67295e-09,0.715054,0.00066703,-4.28013e-07,1.5888e-08,0.715721,0.000666222,-3.80349e-07,1.72576e-09,0.716387,0.000665467,-3.75172e-07,-2.27911e-08,0.717052,0.000664648,-4.43545e-07,2.9834e-08,0.717716,0.00066385,-3.54043e-07,-3.69401e-08,0.718379,0.000663031,-4.64864e-07,5.83219e-08,0.719042,0.000662277,-2.89898e-07,-7.71382e-08,0.719704,0.000661465,-5.21313e-07,7.14171e-08,0.720365,0.000660637,-3.07061e-07,-2.97161e-08,0.721025,0.000659934,-3.96209e-07,-1.21575e-08,0.721685,0.000659105,-4.32682e-07,1.87412e-08,0.722343,0.000658296,-3.76458e-07,-3.2029e-09,0.723001,0.000657533,-3.86067e-07,-5.9296e-09,0.723659,0.000656743,-4.03856e-07,2.69213e-08,0.724315,0.000656016,-3.23092e-07,-4.21511e-08,0.724971,0.000655244,-4.49545e-07,2.24737e-08,0.725625,0.000654412,-3.82124e-07,1.18611e-08,0.726279,0.000653683,-3.46541e-07,-1.03132e-08,0.726933,0.000652959,-3.7748e-07,-3.02128e-08,0.727585,0.000652114,-4.68119e-07,7.15597e-08,0.728237,0.000651392,-2.5344e-07,-7.72119e-08,0.728888,0.000650654,-4.85075e-07,5.8474e-08,0.729538,0.000649859,-3.09654e-07,-3.74746e-08,0.730188,0.000649127,-4.22077e-07,3.18197e-08,0.730837,0.000648379,-3.26618e-07,-3.01997e-08,0.731485,0.000647635,-4.17217e-07,2.93747e-08,0.732132,0.000646888,-3.29093e-07,-2.76943e-08,0.732778,0.000646147,-4.12176e-07,2.17979e-08,0.733424,0.000645388,-3.46783e-07,1.07292e-10,0.734069,0.000644695,-3.46461e-07,-2.22271e-08,0.734713,0.000643935,-4.13142e-07,2.91963e-08,0.735357,0.000643197,-3.25553e-07,-3.49536e-08,0.736,0.000642441,-4.30414e-07,5.10133e-08,0.736642,0.000641733,-2.77374e-07,-4.98904e-08,0.737283,0.000641028,-4.27045e-07,2.93392e-08,0.737924,0.000640262,-3.39028e-07,-7.86156e-09,0.738564,0.000639561,-3.62612e-07,2.10703e-09,0.739203,0.000638842,-3.56291e-07,-5.6653e-10,0.739842,0.000638128,-3.57991e-07,1.59086e-10,0.740479,0.000637412,-3.57513e-07,-6.98321e-11,0.741116,0.000636697,-3.57723e-07,1.20214e-10,0.741753,0.000635982,-3.57362e-07,-4.10987e-10,0.742388,0.000635266,-3.58595e-07,1.5237e-09,0.743023,0.000634553,-3.54024e-07,-5.68376e-09,0.743657,0.000633828,-3.71075e-07,2.12113e-08,0.744291,0.00063315,-3.07441e-07,-1.95569e-08,0.744924,0.000632476,-3.66112e-07,-2.58816e-09,0.745556,0.000631736,-3.73877e-07,2.99096e-08,0.746187,0.000631078,-2.84148e-07,-5.74454e-08,0.746818,0.000630337,-4.56484e-07,8.06629e-08,0.747448,0.000629666,-2.14496e-07,-8.63922e-08,0.748077,0.000628978,-4.73672e-07,8.60918e-08,0.748706,0.000628289,-2.15397e-07,-7.91613e-08,0.749334,0.000627621,-4.5288e-07,5.17393e-08,0.749961,0.00062687,-2.97663e-07,-8.58662e-09,0.750588,0.000626249,-3.23422e-07,-1.73928e-08,0.751214,0.00062555,-3.75601e-07,1.85532e-08,0.751839,0.000624855,-3.19941e-07,2.78479e-09,0.752463,0.000624223,-3.11587e-07,-2.96923e-08,0.753087,0.000623511,-4.00664e-07,5.63799e-08,0.75371,0.000622879,-2.31524e-07,-7.66179e-08,0.754333,0.000622186,-4.61378e-07,7.12778e-08,0.754955,0.000621477,-2.47545e-07,-2.96794e-08,0.755576,0.000620893,-3.36583e-07,-1.21648e-08,0.756196,0.000620183,-3.73077e-07,1.87339e-08,0.756816,0.000619493,-3.16875e-07,-3.16622e-09,0.757435,0.00061885,-3.26374e-07,-6.0691e-09,0.758054,0.000618179,-3.44581e-07,2.74426e-08,0.758672,0.000617572,-2.62254e-07,-4.40968e-08,0.759289,0.000616915,-3.94544e-07,2.97352e-08,0.759906,0.000616215,-3.05338e-07,-1.52393e-08,0.760522,0.000615559,-3.51056e-07,3.12221e-08,0.761137,0.000614951,-2.5739e-07,-5.00443e-08,0.761751,0.000614286,-4.07523e-07,4.9746e-08,0.762365,0.00061362,-2.58285e-07,-2.97303e-08,0.762979,0.000613014,-3.47476e-07,9.57079e-09,0.763591,0.000612348,-3.18764e-07,-8.55287e-09,0.764203,0.000611685,-3.44422e-07,2.46407e-08,0.764815,0.00061107,-2.705e-07,-3.04053e-08,0.765426,0.000610437,-3.61716e-07,3.73759e-08,0.766036,0.000609826,-2.49589e-07,-5.94935e-08,0.766645,0.000609149,-4.28069e-07,8.13889e-08,0.767254,0.000608537,-1.83902e-07,-8.72483e-08,0.767862,0.000607907,-4.45647e-07,8.87901e-08,0.76847,0.000607282,-1.79277e-07,-8.90983e-08,0.769077,0.000606656,-4.46572e-07,8.87892e-08,0.769683,0.000606029,-1.80204e-07,-8.72446e-08,0.770289,0.000605407,-4.41938e-07,8.13752e-08,0.770894,0.000604768,-1.97812e-07,-5.94423e-08,0.771498,0.000604194,-3.76139e-07,3.71848e-08,0.772102,0.000603553,-2.64585e-07,-2.96922e-08,0.772705,0.000602935,-3.53661e-07,2.19793e-08,0.773308,0.000602293,-2.87723e-07,1.37955e-09,0.77391,0.000601722,-2.83585e-07,-2.74976e-08,0.774512,0.000601072,-3.66077e-07,4.9006e-08,0.775112,0.000600487,-2.19059e-07,-4.93171e-08,0.775712,0.000599901,-3.67011e-07,2.90531e-08,0.776312,0.000599254,-2.79851e-07,-7.29081e-09,0.776911,0.000598673,-3.01724e-07,1.10077e-10,0.777509,0.00059807,-3.01393e-07,6.85053e-09,0.778107,0.000597487,-2.80842e-07,-2.75123e-08,0.778704,0.000596843,-3.63379e-07,4.35939e-08,0.779301,0.000596247,-2.32597e-07,-2.7654e-08,0.779897,0.000595699,-3.15559e-07,7.41741e-09,0.780492,0.00059509,-2.93307e-07,-2.01562e-09,0.781087,0.000594497,-2.99354e-07,6.45059e-10,0.781681,0.000593901,-2.97418e-07,-5.64635e-10,0.782275,0.000593304,-2.99112e-07,1.61347e-09,0.782868,0.000592711,-2.94272e-07,-5.88926e-09,0.78346,0.000592105,-3.1194e-07,2.19436e-08,0.784052,0.000591546,-2.46109e-07,-2.22805e-08,0.784643,0.000590987,-3.1295e-07,7.57368e-09,0.785234,0.000590384,-2.90229e-07,-8.01428e-09,0.785824,0.00058978,-3.14272e-07,2.44834e-08,0.786414,0.000589225,-2.40822e-07,-3.03148e-08,0.787003,0.000588652,-3.31766e-07,3.7171e-08,0.787591,0.0005881,-2.20253e-07,-5.87646e-08,0.788179,0.000587483,-3.96547e-07,7.86782e-08,0.788766,0.000586926,-1.60512e-07,-7.71342e-08,0.789353,0.000586374,-3.91915e-07,5.10444e-08,0.789939,0.000585743,-2.38782e-07,-7.83422e-09,0.790524,0.000585242,-2.62284e-07,-1.97076e-08,0.791109,0.000584658,-3.21407e-07,2.70598e-08,0.791693,0.000584097,-2.40228e-07,-2.89269e-08,0.792277,0.000583529,-3.27008e-07,2.90431e-08,0.792861,0.000582963,-2.39879e-07,-2.76409e-08,0.793443,0.0005824,-3.22802e-07,2.1916e-08,0.794025,0.00058182,-2.57054e-07,-4.18368e-10,0.794607,0.000581305,-2.58309e-07,-2.02425e-08,0.795188,0.000580727,-3.19036e-07,2.17838e-08,0.795768,0.000580155,-2.53685e-07,-7.28814e-09,0.796348,0.000579625,-2.75549e-07,7.36871e-09,0.796928,0.000579096,-2.53443e-07,-2.21867e-08,0.797506,0.000578523,-3.20003e-07,2.17736e-08,0.798085,0.000577948,-2.54683e-07,-5.30296e-09,0.798662,0.000577423,-2.70592e-07,-5.61698e-10,0.799239,0.00057688,-2.72277e-07,7.54977e-09,0.799816,0.000576358,-2.49627e-07,-2.96374e-08,0.800392,0.00057577,-3.38539e-07,5.1395e-08,0.800968,0.000575247,-1.84354e-07,-5.67335e-08,0.801543,0.000574708,-3.54555e-07,5.63297e-08,0.802117,0.000574168,-1.85566e-07,-4.93759e-08,0.802691,0.000573649,-3.33693e-07,2.19646e-08,0.803264,0.000573047,-2.678e-07,2.1122e-08,0.803837,0.000572575,-2.04433e-07,-4.68482e-08,0.804409,0.000572026,-3.44978e-07,4.70613e-08,0.804981,0.000571477,-2.03794e-07,-2.21877e-08,0.805552,0.000571003,-2.70357e-07,-1.79153e-08,0.806123,0.000570408,-3.24103e-07,3.42443e-08,0.806693,0.000569863,-2.2137e-07,1.47556e-10,0.807263,0.000569421,-2.20928e-07,-3.48345e-08,0.807832,0.000568874,-3.25431e-07,1.99812e-08,0.808401,0.000568283,-2.65487e-07,1.45143e-08,0.808969,0.000567796,-2.21945e-07,-1.84338e-08,0.809536,0.000567297,-2.77246e-07,-3.83608e-10,0.810103,0.000566741,-2.78397e-07,1.99683e-08,0.81067,0.000566244,-2.18492e-07,-1.98848e-08,0.811236,0.000565747,-2.78146e-07,-3.38976e-11,0.811801,0.000565191,-2.78248e-07,2.00204e-08,0.812366,0.000564695,-2.18187e-07,-2.04429e-08,0.812931,0.000564197,-2.79516e-07,2.1467e-09,0.813495,0.000563644,-2.73076e-07,1.18561e-08,0.814058,0.000563134,-2.37507e-07,1.00334e-08,0.814621,0.000562689,-2.07407e-07,-5.19898e-08,0.815183,0.000562118,-3.63376e-07,7.87163e-08,0.815745,0.000561627,-1.27227e-07,-8.40616e-08,0.816306,0.000561121,-3.79412e-07,7.87163e-08,0.816867,0.000560598,-1.43263e-07,-5.19898e-08,0.817428,0.000560156,-2.99233e-07,1.00335e-08,0.817988,0.000559587,-2.69132e-07,1.18559e-08,0.818547,0.000559085,-2.33564e-07,2.14764e-09,0.819106,0.000558624,-2.27122e-07,-2.04464e-08,0.819664,0.000558108,-2.88461e-07,2.00334e-08,0.820222,0.000557591,-2.28361e-07,-8.24277e-11,0.820779,0.000557135,-2.28608e-07,-1.97037e-08,0.821336,0.000556618,-2.87719e-07,1.92925e-08,0.821893,0.000556101,-2.29841e-07,2.13831e-09,0.822448,0.000555647,-2.23427e-07,-2.78458e-08,0.823004,0.000555117,-3.06964e-07,4.96402e-08,0.823559,0.000554652,-1.58043e-07,-5.15058e-08,0.824113,0.000554181,-3.12561e-07,3.71737e-08,0.824667,0.000553668,-2.0104e-07,-3.75844e-08,0.82522,0.000553153,-3.13793e-07,5.35592e-08,0.825773,0.000552686,-1.53115e-07,-5.74431e-08,0.826326,0.000552207,-3.25444e-07,5.7004e-08,0.826878,0.000551728,-1.54433e-07,-5.13635e-08,0.827429,0.000551265,-3.08523e-07,2.92406e-08,0.82798,0.000550735,-2.20801e-07,-5.99424e-09,0.828531,0.000550276,-2.38784e-07,-5.26363e-09,0.829081,0.000549782,-2.54575e-07,2.70488e-08,0.82963,0.000549354,-1.73429e-07,-4.33268e-08,0.83018,0.000548878,-3.03409e-07,2.7049e-08,0.830728,0.000548352,-2.22262e-07,-5.26461e-09,0.831276,0.000547892,-2.38056e-07,-5.99057e-09,0.831824,0.000547397,-2.56027e-07,2.92269e-08,0.832371,0.000546973,-1.68347e-07,-5.13125e-08,0.832918,0.000546482,-3.22284e-07,5.68139e-08,0.833464,0.000546008,-1.51843e-07,-5.67336e-08,0.83401,0.000545534,-3.22043e-07,5.09113e-08,0.834555,0.000545043,-1.6931e-07,-2.77022e-08,0.8351,0.000544621,-2.52416e-07,2.92924e-10,0.835644,0.000544117,-2.51537e-07,2.65305e-08,0.836188,0.000543694,-1.71946e-07,-4.68105e-08,0.836732,0.00054321,-3.12377e-07,4.15021e-08,0.837275,0.000542709,-1.87871e-07,1.13355e-11,0.837817,0.000542334,-1.87837e-07,-4.15474e-08,0.838359,0.000541833,-3.12479e-07,4.69691e-08,0.838901,0.000541349,-1.71572e-07,-2.71196e-08,0.839442,0.000540925,-2.52931e-07,1.90462e-09,0.839983,0.000540425,-2.47217e-07,1.95011e-08,0.840523,0.000539989,-1.88713e-07,-2.03045e-08,0.841063,0.00053955,-2.49627e-07,2.11216e-09,0.841602,0.000539057,-2.4329e-07,1.18558e-08,0.842141,0.000538606,-2.07723e-07,1.00691e-08,0.842679,0.000538221,-1.77516e-07,-5.21324e-08,0.843217,0.00053771,-3.33913e-07,7.92513e-08,0.843755,0.00053728,-9.6159e-08,-8.60587e-08,0.844292,0.000536829,-3.54335e-07,8.61696e-08,0.844828,0.000536379,-9.58263e-08,-7.98057e-08,0.845364,0.000535948,-3.35243e-07,5.42394e-08,0.8459,0.00053544,-1.72525e-07,-1.79426e-08,0.846435,0.000535041,-2.26353e-07,1.75308e-08,0.84697,0.000534641,-1.73761e-07,-5.21806e-08,0.847505,0.000534137,-3.30302e-07,7.19824e-08,0.848038,0.000533692,-1.14355e-07,-5.69349e-08,0.848572,0.000533293,-2.8516e-07,3.65479e-08,0.849105,0.000532832,-1.75516e-07,-2.96519e-08,0.849638,0.000532392,-2.64472e-07,2.2455e-08,0.85017,0.000531931,-1.97107e-07,-5.63451e-10,0.850702,0.000531535,-1.98797e-07,-2.02011e-08,0.851233,0.000531077,-2.59401e-07,2.17634e-08,0.851764,0.000530623,-1.94111e-07,-7.24794e-09,0.852294,0.000530213,-2.15854e-07,7.22832e-09,0.852824,0.000529803,-1.94169e-07,-2.16653e-08,0.853354,0.00052935,-2.59165e-07,1.98283e-08,0.853883,0.000528891,-1.9968e-07,1.95678e-09,0.854412,0.000528497,-1.9381e-07,-2.76554e-08,0.85494,0.000528027,-2.76776e-07,4.90603e-08,0.855468,0.00052762,-1.29596e-07,-4.93764e-08,0.855995,0.000527213,-2.77725e-07,2.92361e-08,0.856522,0.000526745,-1.90016e-07,-7.96341e-09,0.857049,0.000526341,-2.13907e-07,2.61752e-09,0.857575,0.000525922,-2.06054e-07,-2.50665e-09,0.8581,0.000525502,-2.13574e-07,7.40906e-09,0.858626,0.000525097,-1.91347e-07,-2.71296e-08,0.859151,0.000524633,-2.72736e-07,4.15048e-08,0.859675,0.000524212,-1.48221e-07,-1.96802e-08,0.860199,0.000523856,-2.07262e-07,-2.23886e-08,0.860723,0.000523375,-2.74428e-07,4.96299e-08,0.861246,0.000522975,-1.25538e-07,-5.69216e-08,0.861769,0.000522553,-2.96303e-07,5.88473e-08,0.862291,0.000522137,-1.19761e-07,-5.92584e-08,0.862813,0.00052172,-2.97536e-07,5.8977e-08,0.863334,0.000521301,-1.20605e-07,-5.74403e-08,0.863855,0.000520888,-2.92926e-07,5.15751e-08,0.864376,0.000520457,-1.38201e-07,-2.96506e-08,0.864896,0.000520091,-2.27153e-07,7.42277e-09,0.865416,0.000519659,-2.04885e-07,-4.05057e-11,0.865936,0.00051925,-2.05006e-07,-7.26074e-09,0.866455,0.000518818,-2.26788e-07,2.90835e-08,0.866973,0.000518451,-1.39538e-07,-4.94686e-08,0.867492,0.000518024,-2.87944e-07,4.95814e-08,0.868009,0.000517597,-1.39199e-07,-2.96479e-08,0.868527,0.000517229,-2.28143e-07,9.40539e-09,0.869044,0.000516801,-1.99927e-07,-7.9737e-09,0.86956,0.000516378,-2.23848e-07,2.24894e-08,0.870077,0.000515997,-1.5638e-07,-2.23793e-08,0.870592,0.000515617,-2.23517e-07,7.42302e-09,0.871108,0.000515193,-2.01248e-07,-7.31283e-09,0.871623,0.000514768,-2.23187e-07,2.18283e-08,0.872137,0.000514387,-1.57702e-07,-2.03959e-08,0.872652,0.000514011,-2.1889e-07,1.50711e-10,0.873165,0.000513573,-2.18437e-07,1.97931e-08,0.873679,0.000513196,-1.59058e-07,-1.97183e-08,0.874192,0.000512819,-2.18213e-07,-5.24324e-10,0.874704,0.000512381,-2.19786e-07,2.18156e-08,0.875217,0.000512007,-1.54339e-07,-2.71336e-08,0.875728,0.000511616,-2.3574e-07,2.71141e-08,0.87624,0.000511226,-1.54398e-07,-2.17182e-08,0.876751,0.000510852,-2.19552e-07,1.54131e-10,0.877262,0.000510414,-2.1909e-07,2.11017e-08,0.877772,0.000510039,-1.55785e-07,-2.49562e-08,0.878282,0.000509652,-2.30654e-07,1.91183e-08,0.878791,0.000509248,-1.73299e-07,8.08751e-09,0.8793,0.000508926,-1.49036e-07,-5.14684e-08,0.879809,0.000508474,-3.03441e-07,7.85766e-08,0.880317,0.000508103,-6.77112e-08,-8.40242e-08,0.880825,0.000507715,-3.19784e-07,7.87063e-08,0.881333,0.000507312,-8.36649e-08,-5.19871e-08,0.88184,0.000506988,-2.39626e-07,1.00327e-08,0.882346,0.000506539,-2.09528e-07,1.18562e-08,0.882853,0.000506156,-1.73959e-07,2.14703e-09,0.883359,0.000505814,-1.67518e-07,-2.04444e-08,0.883864,0.000505418,-2.28851e-07,2.00258e-08,0.88437,0.00050502,-1.68774e-07,-5.42855e-11,0.884874,0.000504682,-1.68937e-07,-1.98087e-08,0.885379,0.000504285,-2.28363e-07,1.96842e-08,0.885883,0.000503887,-1.6931e-07,6.76342e-10,0.886387,0.000503551,-1.67281e-07,-2.23896e-08,0.88689,0.000503149,-2.3445e-07,2.92774e-08,0.887393,0.000502768,-1.46618e-07,-3.51152e-08,0.887896,0.00050237,-2.51963e-07,5.15787e-08,0.888398,0.00050202,-9.72271e-08,-5.19903e-08,0.8889,0.00050167,-2.53198e-07,3.71732e-08,0.889401,0.000501275,-1.41678e-07,-3.70978e-08,0.889902,0.00050088,-2.52972e-07,5.16132e-08,0.890403,0.000500529,-9.81321e-08,-5.01459e-08,0.890903,0.000500183,-2.4857e-07,2.9761e-08,0.891403,0.000499775,-1.59287e-07,-9.29351e-09,0.891903,0.000499428,-1.87167e-07,7.41301e-09,0.892402,0.000499076,-1.64928e-07,-2.03585e-08,0.892901,0.000498685,-2.26004e-07,1.44165e-08,0.893399,0.000498276,-1.82754e-07,2.22974e-08,0.893898,0.000497978,-1.15862e-07,-4.40013e-08,0.894395,0.000497614,-2.47866e-07,3.44985e-08,0.894893,0.000497222,-1.44371e-07,-3.43882e-08,0.89539,0.00049683,-2.47535e-07,4.34497e-08,0.895886,0.000496465,-1.17186e-07,-2.02012e-08,0.896383,0.00049617,-1.7779e-07,-2.22497e-08,0.896879,0.000495748,-2.44539e-07,4.95952e-08,0.897374,0.000495408,-9.57532e-08,-5.69217e-08,0.89787,0.000495045,-2.66518e-07,5.88823e-08,0.898364,0.000494689,-8.98713e-08,-5.93983e-08,0.898859,0.000494331,-2.68066e-07,5.95017e-08,0.899353,0.000493973,-8.95613e-08,-5.9399e-08,0.899847,0.000493616,-2.67758e-07,5.8885e-08,0.90034,0.000493257,-9.11033e-08,-5.69317e-08,0.900833,0.000492904,-2.61898e-07,4.96326e-08,0.901326,0.000492529,-1.13001e-07,-2.23893e-08,0.901819,0.000492236,-1.80169e-07,-1.968e-08,0.902311,0.000491817,-2.39209e-07,4.15047e-08,0.902802,0.000491463,-1.14694e-07,-2.71296e-08,0.903293,0.000491152,-1.96083e-07,7.409e-09,0.903784,0.000490782,-1.73856e-07,-2.50645e-09,0.904275,0.000490427,-1.81376e-07,2.61679e-09,0.904765,0.000490072,-1.73525e-07,-7.96072e-09,0.905255,0.000489701,-1.97407e-07,2.92261e-08,0.905745,0.000489394,-1.09729e-07,-4.93389e-08,0.906234,0.000489027,-2.57746e-07,4.89204e-08,0.906723,0.000488658,-1.10985e-07,-2.71333e-08,0.907211,0.000488354,-1.92385e-07,8.30861e-12,0.907699,0.00048797,-1.9236e-07,2.71001e-08,0.908187,0.000487666,-1.1106e-07,-4.88041e-08,0.908675,0.000487298,-2.57472e-07,4.89069e-08,0.909162,0.000486929,-1.10751e-07,-2.76143e-08,0.909649,0.000486625,-1.93594e-07,1.9457e-09,0.910135,0.000486244,-1.87757e-07,1.98315e-08,0.910621,0.000485928,-1.28262e-07,-2.16671e-08,0.911107,0.000485606,-1.93264e-07,7.23216e-09,0.911592,0.000485241,-1.71567e-07,-7.26152e-09,0.912077,0.000484877,-1.93352e-07,2.18139e-08,0.912562,0.000484555,-1.2791e-07,-2.03895e-08,0.913047,0.000484238,-1.89078e-07,1.39494e-10,0.913531,0.000483861,-1.8866e-07,1.98315e-08,0.914014,0.000483543,-1.29165e-07,-1.98609e-08,0.914498,0.000483225,-1.88748e-07,7.39912e-12,0.914981,0.000482847,-1.88726e-07,1.98313e-08,0.915463,0.000482529,-1.29232e-07,-1.9728e-08,0.915946,0.000482212,-1.88416e-07,-5.24035e-10,0.916428,0.000481833,-1.89988e-07,2.18241e-08,0.916909,0.000481519,-1.24516e-07,-2.71679e-08,0.917391,0.000481188,-2.06019e-07,2.72427e-08,0.917872,0.000480858,-1.24291e-07,-2.21985e-08,0.918353,0.000480543,-1.90886e-07,1.94644e-09,0.918833,0.000480167,-1.85047e-07,1.44127e-08,0.919313,0.00047984,-1.41809e-07,7.39438e-12,0.919793,0.000479556,-1.41787e-07,-1.44423e-08,0.920272,0.000479229,-1.85114e-07,-1.84291e-09,0.920751,0.000478854,-1.90642e-07,2.18139e-08,0.92123,0.000478538,-1.25201e-07,-2.58081e-08,0.921708,0.00047821,-2.02625e-07,2.18139e-08,0.922186,0.00047787,-1.37183e-07,-1.84291e-09,0.922664,0.00047759,-1.42712e-07,-1.44423e-08,0.923141,0.000477262,-1.86039e-07,7.34701e-12,0.923618,0.00047689,-1.86017e-07,1.44129e-08,0.924095,0.000476561,-1.42778e-07,1.94572e-09,0.924572,0.000476281,-1.36941e-07,-2.21958e-08,0.925048,0.000475941,-2.03528e-07,2.72327e-08,0.925523,0.000475615,-1.2183e-07,-2.71304e-08,0.925999,0.00047529,-2.03221e-07,2.16843e-08,0.926474,0.000474949,-1.38168e-07,-2.16005e-12,0.926949,0.000474672,-1.38175e-07,-2.16756e-08,0.927423,0.000474331,-2.03202e-07,2.71001e-08,0.927897,0.000474006,-1.21902e-07,-2.71201e-08,0.928371,0.000473681,-2.03262e-07,2.17757e-08,0.928845,0.00047334,-1.37935e-07,-3.78028e-10,0.929318,0.000473063,-1.39069e-07,-2.02636e-08,0.929791,0.000472724,-1.9986e-07,2.18276e-08,0.930263,0.000472389,-1.34377e-07,-7.44231e-09,0.930736,0.000472098,-1.56704e-07,7.94165e-09,0.931208,0.000471809,-1.32879e-07,-2.43243e-08,0.931679,0.00047147,-2.05851e-07,2.97508e-08,0.932151,0.000471148,-1.16599e-07,-3.50742e-08,0.932622,0.000470809,-2.21822e-07,5.09414e-08,0.933092,0.000470518,-6.89976e-08,-4.94821e-08,0.933563,0.000470232,-2.17444e-07,2.77775e-08,0.934033,0.00046988,-1.34111e-07,-2.02351e-09,0.934502,0.000469606,-1.40182e-07,-1.96835e-08,0.934972,0.000469267,-1.99232e-07,2.11529e-08,0.935441,0.000468932,-1.35774e-07,-5.32332e-09,0.93591,0.000468644,-1.51743e-07,1.40413e-10,0.936378,0.000468341,-1.51322e-07,4.76166e-09,0.936846,0.000468053,-1.37037e-07,-1.9187e-08,0.937314,0.000467721,-1.94598e-07,1.23819e-08,0.937782,0.000467369,-1.57453e-07,2.92642e-08,0.938249,0.000467142,-6.96601e-08,-6.98342e-08,0.938716,0.000466793,-2.79163e-07,7.12586e-08,0.939183,0.000466449,-6.53869e-08,-3.63863e-08,0.939649,0.000466209,-1.74546e-07,1.46818e-08,0.940115,0.000465904,-1.305e-07,-2.2341e-08,0.940581,0.000465576,-1.97523e-07,1.50774e-08,0.941046,0.000465226,-1.52291e-07,2.16359e-08,0.941511,0.000464986,-8.73832e-08,-4.20162e-08,0.941976,0.000464685,-2.13432e-07,2.72198e-08,0.942441,0.00046434,-1.31773e-07,-7.2581e-09,0.942905,0.000464055,-1.53547e-07,1.81263e-09,0.943369,0.000463753,-1.48109e-07,7.58386e-12,0.943832,0.000463457,-1.48086e-07,-1.84298e-09,0.944296,0.000463155,-1.53615e-07,7.36433e-09,0.944759,0.00046287,-1.31522e-07,-2.76143e-08,0.945221,0.000462524,-2.14365e-07,4.34883e-08,0.945684,0.000462226,-8.39003e-08,-2.71297e-08,0.946146,0.000461977,-1.65289e-07,5.42595e-09,0.946608,0.000461662,-1.49012e-07,5.42593e-09,0.947069,0.000461381,-1.32734e-07,-2.71297e-08,0.94753,0.000461034,-2.14123e-07,4.34881e-08,0.947991,0.000460736,-8.36585e-08,-2.76134e-08,0.948452,0.000460486,-1.66499e-07,7.36083e-09,0.948912,0.000460175,-1.44416e-07,-1.82993e-09,0.949372,0.000459881,-1.49906e-07,-4.11073e-11,0.949832,0.000459581,-1.50029e-07,1.99434e-09,0.950291,0.000459287,-1.44046e-07,-7.93627e-09,0.950751,0.000458975,-1.67855e-07,2.97507e-08,0.951209,0.000458728,-7.86029e-08,-5.1462e-08,0.951668,0.000458417,-2.32989e-07,5.6888e-08,0.952126,0.000458121,-6.2325e-08,-5.68806e-08,0.952584,0.000457826,-2.32967e-07,5.14251e-08,0.953042,0.000457514,-7.86914e-08,-2.96107e-08,0.953499,0.000457268,-1.67523e-07,7.41296e-09,0.953956,0.000456955,-1.45285e-07,-4.11262e-11,0.954413,0.000456665,-1.45408e-07,-7.24847e-09,0.95487,0.000456352,-1.67153e-07,2.9035e-08,0.955326,0.000456105,-8.00484e-08,-4.92869e-08,0.955782,0.000455797,-2.27909e-07,4.89032e-08,0.956238,0.000455488,-8.11994e-08,-2.71166e-08,0.956693,0.000455244,-1.62549e-07,-4.13678e-11,0.957148,0.000454919,-1.62673e-07,2.72821e-08,0.957603,0.000454675,-8.0827e-08,-4.94824e-08,0.958057,0.000454365,-2.29274e-07,5.14382e-08,0.958512,0.000454061,-7.49597e-08,-3.7061e-08,0.958965,0.0004538,-1.86143e-07,3.72013e-08,0.959419,0.000453539,-7.45389e-08,-5.21396e-08,0.959873,0.000453234,-2.30958e-07,5.21476e-08,0.960326,0.000452928,-7.45146e-08,-3.72416e-08,0.960778,0.000452667,-1.8624e-07,3.72143e-08,0.961231,0.000452407,-7.45967e-08,-5.20109e-08,0.961683,0.000452101,-2.30629e-07,5.16199e-08,0.962135,0.000451795,-7.57696e-08,-3.52595e-08,0.962587,0.000451538,-1.81548e-07,2.98133e-08,0.963038,0.000451264,-9.2108e-08,-2.43892e-08,0.963489,0.000451007,-1.65276e-07,8.13892e-09,0.96394,0.000450701,-1.40859e-07,-8.16647e-09,0.964391,0.000450394,-1.65358e-07,2.45269e-08,0.964841,0.000450137,-9.17775e-08,-3.03367e-08,0.965291,0.000449863,-1.82787e-07,3.7215e-08,0.965741,0.000449609,-7.11424e-08,-5.89188e-08,0.96619,0.00044929,-2.47899e-07,7.92509e-08,0.966639,0.000449032,-1.01462e-08,-7.92707e-08,0.967088,0.000448773,-2.47958e-07,5.90181e-08,0.967537,0.000448455,-7.0904e-08,-3.75925e-08,0.967985,0.0004482,-1.83681e-07,3.17471e-08,0.968433,0.000447928,-8.84401e-08,-2.97913e-08,0.968881,0.000447662,-1.77814e-07,2.78133e-08,0.969329,0.000447389,-9.4374e-08,-2.18572e-08,0.969776,0.000447135,-1.59946e-07,1.10134e-11,0.970223,0.000446815,-1.59913e-07,2.18132e-08,0.97067,0.000446561,-9.44732e-08,-2.76591e-08,0.971116,0.000446289,-1.7745e-07,2.92185e-08,0.971562,0.000446022,-8.97948e-08,-2.96104e-08,0.972008,0.000445753,-1.78626e-07,2.96185e-08,0.972454,0.000445485,-8.97706e-08,-2.92588e-08,0.972899,0.000445218,-1.77547e-07,2.78123e-08,0.973344,0.000444946,-9.41103e-08,-2.23856e-08,0.973789,0.000444691,-1.61267e-07,2.12559e-09,0.974233,0.000444374,-1.5489e-07,1.38833e-08,0.974678,0.000444106,-1.13241e-07,1.94591e-09,0.975122,0.000443886,-1.07403e-07,-2.16669e-08,0.975565,0.000443606,-1.72404e-07,2.5117e-08,0.976009,0.000443336,-9.70526e-08,-1.91963e-08,0.976452,0.000443085,-1.54642e-07,-7.93627e-09,0.976895,0.000442752,-1.7845e-07,5.09414e-08,0.977338,0.000442548,-2.56262e-08,-7.66201e-08,0.97778,0.000442266,-2.55486e-07,7.67249e-08,0.978222,0.000441986,-2.53118e-08,-5.14655e-08,0.978664,0.000441781,-1.79708e-07,9.92773e-09,0.979106,0.000441451,-1.49925e-07,1.17546e-08,0.979547,0.000441186,-1.14661e-07,2.65868e-09,0.979988,0.000440965,-1.06685e-07,-2.23893e-08,0.980429,0.000440684,-1.73853e-07,2.72939e-08,0.980869,0.000440419,-9.19716e-08,-2.71816e-08,0.98131,0.000440153,-1.73516e-07,2.18278e-08,0.98175,0.000439872,-1.08033e-07,-5.24833e-10,0.982189,0.000439654,-1.09607e-07,-1.97284e-08,0.982629,0.000439376,-1.68793e-07,1.98339e-08,0.983068,0.000439097,-1.09291e-07,-2.62901e-12,0.983507,0.000438879,-1.09299e-07,-1.98234e-08,0.983946,0.000438601,-1.68769e-07,1.96916e-08,0.984384,0.000438322,-1.09694e-07,6.6157e-10,0.984823,0.000438105,-1.0771e-07,-2.23379e-08,0.985261,0.000437823,-1.74723e-07,2.90855e-08,0.985698,0.00043756,-8.74669e-08,-3.43992e-08,0.986136,0.000437282,-1.90665e-07,4.89068e-08,0.986573,0.000437048,-4.39442e-08,-4.20188e-08,0.98701,0.000436834,-1.7e-07,-4.11073e-11,0.987446,0.000436494,-1.70124e-07,4.21832e-08,0.987883,0.00043628,-4.35742e-08,-4.94824e-08,0.988319,0.000436044,-1.92021e-07,3.6537e-08,0.988755,0.00043577,-8.24102e-08,-3.70611e-08,0.989191,0.000435494,-1.93593e-07,5.21026e-08,0.989626,0.000435263,-3.72855e-08,-5.21402e-08,0.990061,0.000435032,-1.93706e-07,3.7249e-08,0.990496,0.000434756,-8.19592e-08,-3.72512e-08,0.990931,0.000434481,-1.93713e-07,5.21511e-08,0.991365,0.00043425,-3.72595e-08,-5.21439e-08,0.991799,0.000434019,-1.93691e-07,3.72152e-08,0.992233,0.000433743,-8.20456e-08,-3.71123e-08,0.992667,0.000433468,-1.93382e-07,5.16292e-08,0.9931,0.000433236,-3.84947e-08,-5.01953e-08,0.993533,0.000433008,-1.89081e-07,2.99427e-08,0.993966,0.00043272,-9.92525e-08,-9.9708e-09,0.994399,0.000432491,-1.29165e-07,9.94051e-09,0.994831,0.000432263,-9.93434e-08,-2.97912e-08,0.995263,0.000431975,-1.88717e-07,4.96198e-08,0.995695,0.000431746,-3.98578e-08,-4.94785e-08,0.996127,0.000431518,-1.88293e-07,2.9085e-08,0.996558,0.000431229,-1.01038e-07,-7.25675e-09,0.996989,0.000431005,-1.22809e-07,-5.79945e-11,0.99742,0.000430759,-1.22983e-07,7.48873e-09,0.997851,0.000430536,-1.00516e-07,-2.98969e-08,0.998281,0.000430245,-1.90207e-07,5.24942e-08,0.998711,0.000430022,-3.27246e-08,-6.08706e-08,0.999141,0.000429774,-2.15336e-07,7.17788e-08,0.999571,0.000429392,0.,0.};
+
+    template <typename T, int scn, int dcn, bool srgb, int blueIdx> struct Lab2RGB;
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct Lab2RGB<float, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float lThresh = 0.008856f * 903.3f;
+            const float fThresh = 7.787f * 0.008856f + 16.0f / 116.0f;
+
+            float Y, fy;
+
+            if (src.x <= lThresh)
+            {
+                Y = src.x / 903.3f;
+                fy = 7.787f * Y + 16.0f / 116.0f;
+            }
+            else
+            {
+                fy = (src.x + 16.0f) / 116.0f;
+                Y = fy * fy * fy;
+            }
+
+            float X = src.y / 500.0f + fy;
+            float Z = fy - src.z / 200.0f;
+
+            if (X <= fThresh)
+                X = (X - 16.0f / 116.0f) / 7.787f;
+            else
+                X = X * X * X;
+
+            if (Z <= fThresh)
+                Z = (Z - 16.0f / 116.0f) / 7.787f;
+            else
+                Z = Z * Z * Z;
+
+            float B = 0.052891f * X - 0.204043f * Y + 1.151152f * Z;
+            float G = -0.921235f * X + 1.875991f * Y + 0.045244f * Z;
+            float R = 3.079933f * X - 1.537150f * Y - 0.542782f * Z;
+
+            if (srgb)
+            {
+                B = splineInterpolate(B * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+                G = splineInterpolate(G * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+                R = splineInterpolate(R * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+            }
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = blueIdx == 0 ? B : R;
+            dst.y = G;
+            dst.z = blueIdx == 0 ? R : B;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct Lab2RGB<uchar, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x * (100.f / 255.f);
+            buf.y = src.y - 128;
+            buf.z = src.z - 128;
+
+            Lab2RGB<float, 3, 3, srgb, blueIdx> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x * 255.f);
+            dst.y = saturate_cast<uchar>(buf.y * 255.f);
+            dst.z = saturate_cast<uchar>(buf.z * 255.f);
+            setAlpha(dst, ColorChannel<uchar>::max());
+
+            return dst;
+        }
+    };
+
+    // RGB to Luv
+
+    __constant__ float c_LabCbrtTab[] = {0.137931,0.0114066,0.,1.18859e-07,0.149338,0.011407,3.56578e-07,-5.79396e-07,0.160745,0.0114059,-1.38161e-06,2.16892e-06,0.172151,0.0114097,5.12516e-06,-8.0814e-06,0.183558,0.0113957,-1.9119e-05,3.01567e-05,0.194965,0.0114479,7.13509e-05,-0.000112545,0.206371,0.011253,-0.000266285,-0.000106493,0.217252,0.0104009,-0.000585765,7.32149e-05,0.22714,0.00944906,-0.00036612,1.21917e-05,0.236235,0.0087534,-0.000329545,2.01753e-05,0.244679,0.00815483,-0.000269019,1.24435e-05,0.252577,0.00765412,-0.000231689,1.05618e-05,0.26001,0.00722243,-0.000200003,8.26662e-06,0.267041,0.00684723,-0.000175203,6.76746e-06,0.27372,0.00651712,-0.000154901,5.61192e-06,0.280088,0.00622416,-0.000138065,4.67009e-06,0.286179,0.00596204,-0.000124055,3.99012e-06,0.292021,0.0057259,-0.000112085,3.36032e-06,0.297638,0.00551181,-0.000102004,2.95338e-06,0.30305,0.00531666,-9.31435e-05,2.52875e-06,0.308277,0.00513796,-8.55572e-05,2.22022e-06,0.313331,0.00497351,-7.88966e-05,1.97163e-06,0.318228,0.00482163,-7.29817e-05,1.7248e-06,0.322978,0.00468084,-6.78073e-05,1.55998e-06,0.327593,0.0045499,-6.31274e-05,1.36343e-06,0.332081,0.00442774,-5.90371e-05,1.27136e-06,0.336451,0.00431348,-5.5223e-05,1.09111e-06,0.34071,0.00420631,-5.19496e-05,1.0399e-06,0.344866,0.00410553,-4.88299e-05,9.18347e-07,0.348923,0.00401062,-4.60749e-05,8.29942e-07,0.352889,0.00392096,-4.35851e-05,7.98478e-07,0.356767,0.00383619,-4.11896e-05,6.84917e-07,0.360562,0.00375586,-3.91349e-05,6.63976e-07,0.36428,0.00367959,-3.7143e-05,5.93086e-07,0.367923,0.00360708,-3.53637e-05,5.6976e-07,0.371495,0.00353806,-3.36544e-05,4.95533e-07,0.375,0.00347224,-3.21678e-05,4.87951e-07,0.378441,0.00340937,-3.0704e-05,4.4349e-07,0.38182,0.00334929,-2.93735e-05,4.20297e-07,0.38514,0.0032918,-2.81126e-05,3.7872e-07,0.388404,0.00323671,-2.69764e-05,3.596e-07,0.391614,0.00318384,-2.58976e-05,3.5845e-07,0.394772,0.00313312,-2.48223e-05,2.92765e-07,0.397881,0.00308435,-2.3944e-05,3.18232e-07,0.400942,0.00303742,-2.29893e-05,2.82046e-07,0.403957,0.00299229,-2.21432e-05,2.52315e-07,0.406927,0.00294876,-2.13862e-05,2.58416e-07,0.409855,0.00290676,-2.0611e-05,2.33939e-07,0.412741,0.00286624,-1.99092e-05,2.36342e-07,0.415587,0.00282713,-1.92001e-05,1.916e-07,0.418396,0.00278931,-1.86253e-05,2.1915e-07,0.421167,0.00275271,-1.79679e-05,1.83498e-07,0.423901,0.00271733,-1.74174e-05,1.79343e-07,0.426602,0.00268303,-1.68794e-05,1.72013e-07,0.429268,0.00264979,-1.63633e-05,1.75686e-07,0.431901,0.00261759,-1.58363e-05,1.3852e-07,0.434503,0.00258633,-1.54207e-05,1.64304e-07,0.437074,0.00255598,-1.49278e-05,1.28136e-07,0.439616,0.00252651,-1.45434e-05,1.57618e-07,0.442128,0.0024979,-1.40705e-05,1.0566e-07,0.444612,0.00247007,-1.37535e-05,1.34998e-07,0.447068,0.00244297,-1.33485e-05,1.29207e-07,0.449498,0.00241666,-1.29609e-05,9.32347e-08,0.451902,0.00239102,-1.26812e-05,1.23703e-07,0.45428,0.00236603,-1.23101e-05,9.74072e-08,0.456634,0.0023417,-1.20179e-05,1.12518e-07,0.458964,0.002318,-1.16803e-05,7.83681e-08,0.46127,0.00229488,-1.14452e-05,1.10452e-07,0.463554,0.00227232,-1.11139e-05,7.58719e-08,0.465815,0.00225032,-1.08863e-05,9.2699e-08,0.468055,0.00222882,-1.06082e-05,8.97738e-08,0.470273,0.00220788,-1.03388e-05,5.4845e-08,0.47247,0.00218736,-1.01743e-05,1.0808e-07,0.474648,0.00216734,-9.85007e-06,4.9277e-08,0.476805,0.00214779,-9.70224e-06,8.22408e-08,0.478943,0.00212863,-9.45551e-06,6.87942e-08,0.481063,0.00210993,-9.24913e-06,5.98144e-08,0.483163,0.00209161,-9.06969e-06,7.93789e-08,0.485246,0.00207371,-8.83155e-06,3.99032e-08,0.487311,0.00205616,-8.71184e-06,8.88325e-08,0.489358,0.002039,-8.44534e-06,2.20004e-08,0.491389,0.00202218,-8.37934e-06,9.13872e-08,0.493403,0.0020057,-8.10518e-06,2.96829e-08,0.495401,0.00198957,-8.01613e-06,5.81028e-08,0.497382,0.00197372,-7.84183e-06,6.5731e-08,0.499348,0.00195823,-7.64463e-06,3.66019e-08,0.501299,0.00194305,-7.53483e-06,2.62811e-08,0.503234,0.00192806,-7.45598e-06,9.66907e-08,0.505155,0.00191344,-7.16591e-06,4.18928e-09,0.507061,0.00189912,-7.15334e-06,6.53665e-08,0.508953,0.00188501,-6.95724e-06,3.23686e-08,0.510831,0.00187119,-6.86014e-06,4.35774e-08,0.512696,0.0018576,-6.72941e-06,3.17406e-08,0.514547,0.00184424,-6.63418e-06,6.78785e-08,0.516384,0.00183117,-6.43055e-06,-5.23126e-09,0.518209,0.0018183,-6.44624e-06,7.22562e-08,0.520021,0.00180562,-6.22947e-06,1.42292e-08,0.52182,0.0017932,-6.18679e-06,4.9641e-08,0.523607,0.00178098,-6.03786e-06,2.56259e-08,0.525382,0.00176898,-5.96099e-06,2.66696e-08,0.527145,0.00175714,-5.88098e-06,4.65094e-08,0.528897,0.00174552,-5.74145e-06,2.57114e-08,0.530637,0.00173411,-5.66431e-06,2.94588e-08,0.532365,0.00172287,-5.57594e-06,3.52667e-08,0.534082,0.00171182,-5.47014e-06,8.28868e-09,0.535789,0.00170091,-5.44527e-06,5.07871e-08,0.537484,0.00169017,-5.29291e-06,2.69817e-08,0.539169,0.00167967,-5.21197e-06,2.01009e-08,0.540844,0.0016693,-5.15166e-06,1.18237e-08,0.542508,0.00165903,-5.11619e-06,5.18135e-08,0.544162,0.00164896,-4.96075e-06,1.9341e-08,0.545806,0.00163909,-4.90273e-06,-9.96867e-09,0.54744,0.00162926,-4.93263e-06,8.01382e-08,0.549064,0.00161963,-4.69222e-06,-1.25601e-08,0.550679,0.00161021,-4.7299e-06,2.97067e-08,0.552285,0.00160084,-4.64078e-06,1.29426e-08,0.553881,0.0015916,-4.60195e-06,3.77327e-08,0.555468,0.00158251,-4.48875e-06,1.49412e-08,0.557046,0.00157357,-4.44393e-06,2.17118e-08,0.558615,0.00156475,-4.3788e-06,1.74206e-08,0.560176,0.00155605,-4.32653e-06,2.78152e-08,0.561727,0.00154748,-4.24309e-06,-9.47239e-09,0.563271,0.00153896,-4.27151e-06,6.9679e-08,0.564805,0.00153063,-4.06247e-06,-3.08246e-08,0.566332,0.00152241,-4.15494e-06,5.36188e-08,0.56785,0.00151426,-3.99409e-06,-4.83594e-09,0.56936,0.00150626,-4.00859e-06,2.53293e-08,0.570863,0.00149832,-3.93261e-06,2.27286e-08,0.572357,0.00149052,-3.86442e-06,2.96541e-09,0.573844,0.0014828,-3.85552e-06,2.50147e-08,0.575323,0.00147516,-3.78048e-06,1.61842e-08,0.576794,0.00146765,-3.73193e-06,2.94582e-08,0.578258,0.00146028,-3.64355e-06,-1.48076e-08,0.579715,0.00145295,-3.68798e-06,2.97724e-08,0.581164,0.00144566,-3.59866e-06,1.49272e-08,0.582606,0.00143851,-3.55388e-06,2.97285e-08,0.584041,0.00143149,-3.46469e-06,-1.46323e-08,0.585469,0.00142451,-3.50859e-06,2.88004e-08,0.58689,0.00141758,-3.42219e-06,1.864e-08,0.588304,0.00141079,-3.36627e-06,1.58482e-08,0.589712,0.00140411,-3.31872e-06,-2.24279e-08,0.591112,0.00139741,-3.38601e-06,7.38639e-08,0.592507,0.00139085,-3.16441e-06,-3.46088e-08,0.593894,0.00138442,-3.26824e-06,4.96675e-09,0.595275,0.0013779,-3.25334e-06,7.4346e-08,0.59665,0.00137162,-3.0303e-06,-6.39319e-08,0.598019,0.00136536,-3.2221e-06,6.21725e-08,0.599381,0.00135911,-3.03558e-06,-5.94423e-09,0.600737,0.00135302,-3.05341e-06,2.12091e-08,0.602087,0.00134697,-2.98979e-06,-1.92876e-08,0.603431,0.00134094,-3.04765e-06,5.5941e-08,0.604769,0.00133501,-2.87983e-06,-2.56622e-08,0.606101,0.00132917,-2.95681e-06,4.67078e-08,0.607427,0.0013234,-2.81669e-06,-4.19592e-08,0.608748,0.00131764,-2.94257e-06,6.15243e-08,0.610062,0.00131194,-2.75799e-06,-2.53244e-08,0.611372,0.00130635,-2.83397e-06,3.97739e-08,0.612675,0.0013008,-2.71465e-06,-1.45618e-08,0.613973,0.00129533,-2.75833e-06,1.84733e-08,0.615266,0.00128986,-2.70291e-06,2.73606e-10,0.616553,0.00128446,-2.70209e-06,4.00367e-08,0.617835,0.00127918,-2.58198e-06,-4.12113e-08,0.619111,0.00127389,-2.70561e-06,6.52039e-08,0.620383,0.00126867,-2.51e-06,-4.07901e-08,0.621649,0.00126353,-2.63237e-06,3.83516e-08,0.62291,0.00125838,-2.51732e-06,6.59315e-09,0.624166,0.00125337,-2.49754e-06,-5.11939e-09,0.625416,0.00124836,-2.5129e-06,1.38846e-08,0.626662,0.00124337,-2.47124e-06,9.18514e-09,0.627903,0.00123846,-2.44369e-06,8.97952e-09,0.629139,0.0012336,-2.41675e-06,1.45012e-08,0.63037,0.00122881,-2.37325e-06,-7.37949e-09,0.631597,0.00122404,-2.39538e-06,1.50169e-08,0.632818,0.00121929,-2.35033e-06,6.91648e-09,0.634035,0.00121461,-2.32958e-06,1.69219e-08,0.635248,0.00121,-2.27882e-06,-1.49997e-08,0.636455,0.0012054,-2.32382e-06,4.30769e-08,0.637659,0.00120088,-2.19459e-06,-3.80986e-08,0.638857,0.00119638,-2.30888e-06,4.97134e-08,0.640051,0.00119191,-2.15974e-06,-4.15463e-08,0.641241,0.00118747,-2.28438e-06,5.68667e-08,0.642426,0.00118307,-2.11378e-06,-7.10641e-09,0.643607,0.00117882,-2.1351e-06,-2.8441e-08,0.644784,0.00117446,-2.22042e-06,6.12658e-08,0.645956,0.00117021,-2.03663e-06,-3.78083e-08,0.647124,0.00116602,-2.15005e-06,3.03627e-08,0.648288,0.00116181,-2.05896e-06,-2.40379e-08,0.649448,0.00115762,-2.13108e-06,6.57887e-08,0.650603,0.00115356,-1.93371e-06,-6.03028e-08,0.651755,0.00114951,-2.11462e-06,5.62134e-08,0.652902,0.00114545,-1.94598e-06,-4.53417e-08,0.654046,0.00114142,-2.082e-06,6.55489e-08,0.655185,0.00113745,-1.88536e-06,-3.80396e-08,0.656321,0.00113357,-1.99948e-06,2.70049e-08,0.657452,0.00112965,-1.91846e-06,-1.03755e-08,0.65858,0.00112578,-1.94959e-06,1.44973e-08,0.659704,0.00112192,-1.9061e-06,1.1991e-08,0.660824,0.00111815,-1.87012e-06,-2.85634e-09,0.66194,0.0011144,-1.87869e-06,-5.65782e-10,0.663053,0.00111064,-1.88039e-06,5.11947e-09,0.664162,0.0011069,-1.86503e-06,3.96924e-08,0.665267,0.00110328,-1.74595e-06,-4.46795e-08,0.666368,0.00109966,-1.87999e-06,1.98161e-08,0.667466,0.00109596,-1.82054e-06,2.502e-08,0.66856,0.00109239,-1.74548e-06,-6.86593e-10,0.669651,0.0010889,-1.74754e-06,-2.22739e-08,0.670738,0.00108534,-1.81437e-06,3.01776e-08,0.671821,0.0010818,-1.72383e-06,2.07732e-08,0.672902,0.00107841,-1.66151e-06,-5.36658e-08,0.673978,0.00107493,-1.82251e-06,7.46802e-08,0.675051,0.00107151,-1.59847e-06,-6.62411e-08,0.676121,0.00106811,-1.79719e-06,7.10748e-08,0.677188,0.00106473,-1.58397e-06,-3.92441e-08,0.678251,0.00106145,-1.7017e-06,2.62973e-08,0.679311,0.00105812,-1.62281e-06,-6.34035e-09,0.680367,0.00105486,-1.64183e-06,-9.36249e-10,0.68142,0.00105157,-1.64464e-06,1.00854e-08,0.68247,0.00104831,-1.61438e-06,2.01995e-08,0.683517,0.00104514,-1.55378e-06,-3.1279e-08,0.68456,0.00104194,-1.64762e-06,4.53114e-08,0.685601,0.00103878,-1.51169e-06,-3.07573e-08,0.686638,0.00103567,-1.60396e-06,1.81133e-08,0.687672,0.00103251,-1.54962e-06,1.79085e-08,0.688703,0.00102947,-1.49589e-06,-3.01428e-08,0.689731,0.00102639,-1.58632e-06,4.30583e-08,0.690756,0.00102334,-1.45715e-06,-2.28814e-08,0.691778,0.00102036,-1.52579e-06,-1.11373e-08,0.692797,0.00101727,-1.5592e-06,6.74305e-08,0.693812,0.00101436,-1.35691e-06,-7.97709e-08,0.694825,0.0010114,-1.59622e-06,7.28391e-08,0.695835,0.00100843,-1.37771e-06,-3.27715e-08,0.696842,0.00100558,-1.47602e-06,-1.35807e-09,0.697846,0.00100262,-1.48009e-06,3.82037e-08,0.698847,0.000999775,-1.36548e-06,-3.22474e-08,0.699846,0.000996948,-1.46223e-06,3.11809e-08,0.700841,0.000994117,-1.36868e-06,-3.28714e-08,0.701834,0.000991281,-1.4673e-06,4.07001e-08,0.702824,0.000988468,-1.3452e-06,-1.07197e-08,0.703811,0.000985746,-1.37736e-06,2.17866e-09,0.704795,0.000982998,-1.37082e-06,2.00521e-09,0.705777,0.000980262,-1.3648e-06,-1.01996e-08,0.706756,0.000977502,-1.3954e-06,3.87931e-08,0.707732,0.000974827,-1.27902e-06,-2.57632e-08,0.708706,0.000972192,-1.35631e-06,4.65513e-09,0.709676,0.000969493,-1.34235e-06,7.14257e-09,0.710645,0.00096683,-1.32092e-06,2.63791e-08,0.71161,0.000964267,-1.24178e-06,-5.30543e-08,0.712573,0.000961625,-1.40095e-06,6.66289e-08,0.713533,0.000959023,-1.20106e-06,-3.46474e-08,0.714491,0.000956517,-1.305e-06,1.23559e-08,0.715446,0.000953944,-1.26793e-06,-1.47763e-08,0.716399,0.000951364,-1.31226e-06,4.67494e-08,0.717349,0.000948879,-1.17201e-06,-5.3012e-08,0.718297,0.000946376,-1.33105e-06,4.60894e-08,0.719242,0.000943852,-1.19278e-06,-1.21366e-08,0.720185,0.00094143,-1.22919e-06,2.45673e-09,0.721125,0.000938979,-1.22182e-06,2.30966e-09,0.722063,0.000936543,-1.21489e-06,-1.16954e-08,0.722998,0.000934078,-1.24998e-06,4.44718e-08,0.723931,0.000931711,-1.11656e-06,-4.69823e-08,0.724861,0.000929337,-1.25751e-06,2.4248e-08,0.725789,0.000926895,-1.18477e-06,9.5949e-09,0.726715,0.000924554,-1.15598e-06,-3.02286e-09,0.727638,0.000922233,-1.16505e-06,2.49649e-09,0.72856,0.00091991,-1.15756e-06,-6.96321e-09,0.729478,0.000917575,-1.17845e-06,2.53564e-08,0.730395,0.000915294,-1.10238e-06,-3.48578e-08,0.731309,0.000912984,-1.20695e-06,5.44704e-08,0.732221,0.000910734,-1.04354e-06,-6.38144e-08,0.73313,0.000908455,-1.23499e-06,8.15781e-08,0.734038,0.00090623,-9.90253e-07,-8.3684e-08,0.734943,0.000903999,-1.2413e-06,7.43441e-08,0.735846,0.000901739,-1.01827e-06,-3.48787e-08,0.736746,0.000899598,-1.12291e-06,5.56596e-09,0.737645,0.000897369,-1.10621e-06,1.26148e-08,0.738541,0.000895194,-1.06837e-06,3.57935e-09,0.739435,0.000893068,-1.05763e-06,-2.69322e-08,0.740327,0.000890872,-1.13842e-06,4.45448e-08,0.741217,0.000888729,-1.00479e-06,-3.20376e-08,0.742105,0.000886623,-1.1009e-06,2.40011e-08,0.74299,0.000884493,-1.0289e-06,-4.36209e-09,0.743874,0.000882422,-1.04199e-06,-6.55268e-09,0.744755,0.000880319,-1.06164e-06,3.05728e-08,0.745634,0.000878287,-9.69926e-07,-5.61338e-08,0.746512,0.000876179,-1.13833e-06,7.4753e-08,0.747387,0.000874127,-9.14068e-07,-6.40644e-08,0.74826,0.000872106,-1.10626e-06,6.22955e-08,0.749131,0.000870081,-9.19375e-07,-6.59083e-08,0.75,0.000868044,-1.1171e-06,8.21284e-08,0.750867,0.000866056,-8.70714e-07,-8.37915e-08,0.751732,0.000864064,-1.12209e-06,7.42237e-08,0.752595,0.000862042,-8.99418e-07,-3.42894e-08,0.753456,0.00086014,-1.00229e-06,3.32955e-09,0.754315,0.000858146,-9.92297e-07,2.09712e-08,0.755173,0.000856224,-9.29384e-07,-2.76096e-08,0.756028,0.000854282,-1.01221e-06,2.98627e-08,0.756881,0.000852348,-9.22625e-07,-3.22365e-08,0.757733,0.000850406,-1.01933e-06,3.94786e-08,0.758582,0.000848485,-9.00898e-07,-6.46833e-09,0.75943,0.000846664,-9.20303e-07,-1.36052e-08,0.760275,0.000844783,-9.61119e-07,1.28447e-09,0.761119,0.000842864,-9.57266e-07,8.4674e-09,0.761961,0.000840975,-9.31864e-07,2.44506e-08,0.762801,0.000839185,-8.58512e-07,-4.6665e-08,0.763639,0.000837328,-9.98507e-07,4.30001e-08,0.764476,0.00083546,-8.69507e-07,-6.12609e-09,0.76531,0.000833703,-8.87885e-07,-1.84959e-08,0.766143,0.000831871,-9.43372e-07,2.05052e-08,0.766974,0.000830046,-8.81857e-07,-3.92026e-09,0.767803,0.000828271,-8.93618e-07,-4.82426e-09,0.768631,0.000826469,-9.0809e-07,2.32172e-08,0.769456,0.000824722,-8.38439e-07,-2.84401e-08,0.77028,0.00082296,-9.23759e-07,3.09386e-08,0.771102,0.000821205,-8.30943e-07,-3.57099e-08,0.771922,0.000819436,-9.38073e-07,5.22963e-08,0.772741,0.000817717,-7.81184e-07,-5.42658e-08,0.773558,0.000815992,-9.43981e-07,4.55579e-08,0.774373,0.000814241,-8.07308e-07,-8.75656e-09,0.775186,0.0008126,-8.33578e-07,-1.05315e-08,0.775998,0.000810901,-8.65172e-07,-8.72188e-09,0.776808,0.000809145,-8.91338e-07,4.54191e-08,0.777616,0.000807498,-7.5508e-07,-5.37454e-08,0.778423,0.000805827,-9.16317e-07,5.03532e-08,0.779228,0.000804145,-7.65257e-07,-2.84584e-08,0.780031,0.000802529,-8.50632e-07,3.87579e-09,0.780833,0.00080084,-8.39005e-07,1.29552e-08,0.781633,0.0007992,-8.00139e-07,3.90804e-09,0.782432,0.000797612,-7.88415e-07,-2.85874e-08,0.783228,0.000795949,-8.74177e-07,5.0837e-08,0.784023,0.000794353,-7.21666e-07,-5.55513e-08,0.784817,0.000792743,-8.8832e-07,5.21587e-08,0.785609,0.000791123,-7.31844e-07,-3.38744e-08,0.786399,0.000789558,-8.33467e-07,2.37342e-08,0.787188,0.000787962,-7.62264e-07,-1.45775e-09,0.787975,0.000786433,-7.66638e-07,-1.79034e-08,0.788761,0.000784846,-8.20348e-07,1.34665e-08,0.789545,0.000783246,-7.79948e-07,2.3642e-08,0.790327,0.000781757,-7.09022e-07,-4.84297e-08,0.791108,0.000780194,-8.54311e-07,5.08674e-08,0.791888,0.000778638,-7.01709e-07,-3.58303e-08,0.792666,0.000777127,-8.092e-07,3.28493e-08,0.793442,0.000775607,-7.10652e-07,-3.59624e-08,0.794217,0.000774078,-8.1854e-07,5.13959e-08,0.79499,0.000772595,-6.64352e-07,-5.04121e-08,0.795762,0.000771115,-8.15588e-07,3.10431e-08,0.796532,0.000769577,-7.22459e-07,-1.41557e-08,0.797301,0.00076809,-7.64926e-07,2.55795e-08,0.798069,0.000766636,-6.88187e-07,-2.85578e-08,0.798835,0.000765174,-7.73861e-07,2.90472e-08,0.799599,0.000763714,-6.86719e-07,-2.80262e-08,0.800362,0.000762256,-7.70798e-07,2.34531e-08,0.801123,0.000760785,-7.00438e-07,-6.18144e-09,0.801884,0.000759366,-7.18983e-07,1.27263e-09,0.802642,0.000757931,-7.15165e-07,1.09101e-09,0.803399,0.000756504,-7.11892e-07,-5.63675e-09,0.804155,0.000755064,-7.28802e-07,2.14559e-08,0.80491,0.00075367,-6.64434e-07,-2.05821e-08,0.805663,0.00075228,-7.26181e-07,1.26812e-09,0.806414,0.000750831,-7.22377e-07,1.55097e-08,0.807164,0.000749433,-6.75848e-07,-3.70216e-09,0.807913,0.00074807,-6.86954e-07,-7.0105e-10,0.80866,0.000746694,-6.89057e-07,6.5063e-09,0.809406,0.000745336,-6.69538e-07,-2.53242e-08,0.810151,0.000743921,-7.45511e-07,3.51858e-08,0.810894,0.000742535,-6.39953e-07,3.79034e-09,0.811636,0.000741267,-6.28582e-07,-5.03471e-08,0.812377,0.000739858,-7.79624e-07,7.83886e-08,0.813116,0.000738534,-5.44458e-07,-8.43935e-08,0.813854,0.000737192,-7.97638e-07,8.03714e-08,0.81459,0.000735838,-5.56524e-07,-5.82784e-08,0.815325,0.00073455,-7.31359e-07,3.35329e-08,0.816059,0.000733188,-6.3076e-07,-1.62486e-08,0.816792,0.000731878,-6.79506e-07,3.14614e-08,0.817523,0.000730613,-5.85122e-07,-4.99925e-08,0.818253,0.000729293,-7.35099e-07,4.92994e-08,0.818982,0.000727971,-5.87201e-07,-2.79959e-08,0.819709,0.000726712,-6.71189e-07,3.07959e-09,0.820435,0.000725379,-6.6195e-07,1.56777e-08,0.82116,0.000724102,-6.14917e-07,-6.18564e-09,0.821883,0.000722854,-6.33474e-07,9.06488e-09,0.822606,0.000721614,-6.06279e-07,-3.00739e-08,0.823327,0.000720311,-6.96501e-07,5.16262e-08,0.824046,0.000719073,-5.41623e-07,-5.72214e-08,0.824765,0.000717818,-7.13287e-07,5.80503e-08,0.825482,0.000716566,-5.39136e-07,-5.57703e-08,0.826198,0.00071532,-7.06447e-07,4.58215e-08,0.826912,0.000714045,-5.68983e-07,-8.30636e-09,0.827626,0.000712882,-5.93902e-07,-1.25961e-08,0.828338,0.000711656,-6.3169e-07,-9.13985e-10,0.829049,0.00071039,-6.34432e-07,1.62519e-08,0.829759,0.00070917,-5.85676e-07,-4.48904e-09,0.830468,0.000707985,-5.99143e-07,1.70418e-09,0.831175,0.000706792,-5.9403e-07,-2.32768e-09,0.831881,0.000705597,-6.01014e-07,7.60648e-09,0.832586,0.000704418,-5.78194e-07,-2.80982e-08,0.83329,0.000703177,-6.62489e-07,4.51817e-08,0.833993,0.000701988,-5.26944e-07,-3.34192e-08,0.834694,0.000700834,-6.27201e-07,2.88904e-08,0.835394,0.000699666,-5.4053e-07,-2.25378e-08,0.836093,0.000698517,-6.08143e-07,1.65589e-09,0.836791,0.000697306,-6.03176e-07,1.59142e-08,0.837488,0.000696147,-5.55433e-07,-5.70801e-09,0.838184,0.000695019,-5.72557e-07,6.91792e-09,0.838878,0.000693895,-5.51803e-07,-2.19637e-08,0.839571,0.000692725,-6.17694e-07,2.13321e-08,0.840263,0.000691554,-5.53698e-07,-3.75996e-09,0.840954,0.000690435,-5.64978e-07,-6.29219e-09,0.841644,0.000689287,-5.83855e-07,2.89287e-08,0.842333,0.000688206,-4.97068e-07,-4.98181e-08,0.843021,0.000687062,-6.46523e-07,5.11344e-08,0.843707,0.000685922,-4.9312e-07,-3.55102e-08,0.844393,0.00068483,-5.9965e-07,3.13019e-08,0.845077,0.000683724,-5.05745e-07,-3.00925e-08,0.84576,0.000682622,-5.96022e-07,2.94636e-08,0.846442,0.000681519,-5.07631e-07,-2.81572e-08,0.847123,0.000680419,-5.92103e-07,2.35606e-08,0.847803,0.000679306,-5.21421e-07,-6.48045e-09,0.848482,0.000678243,-5.40863e-07,2.36124e-09,0.849159,0.000677169,-5.33779e-07,-2.96461e-09,0.849836,0.000676092,-5.42673e-07,9.49728e-09,0.850512,0.000675035,-5.14181e-07,-3.50245e-08,0.851186,0.000673902,-6.19254e-07,7.09959e-08,0.851859,0.000672876,-4.06267e-07,-7.01453e-08,0.852532,0.000671853,-6.16703e-07,3.07714e-08,0.853203,0.000670712,-5.24388e-07,6.66423e-09,0.853873,0.000669684,-5.04396e-07,2.17629e-09,0.854542,0.000668681,-4.97867e-07,-1.53693e-08,0.855211,0.000667639,-5.43975e-07,-3.03752e-10,0.855878,0.000666551,-5.44886e-07,1.65844e-08,0.856544,0.000665511,-4.95133e-07,-6.42907e-09,0.857209,0.000664501,-5.1442e-07,9.13195e-09,0.857873,0.0006635,-4.87024e-07,-3.00987e-08,0.858536,0.000662435,-5.7732e-07,5.16584e-08,0.859198,0.000661436,-4.22345e-07,-5.73255e-08,0.859859,0.000660419,-5.94322e-07,5.84343e-08,0.860518,0.000659406,-4.19019e-07,-5.72022e-08,0.861177,0.000658396,-5.90626e-07,5.11653e-08,0.861835,0.000657368,-4.3713e-07,-2.82495e-08,0.862492,0.000656409,-5.21878e-07,2.22788e-09,0.863148,0.000655372,-5.15195e-07,1.9338e-08,0.863803,0.0006544,-4.5718e-07,-1.99754e-08,0.864457,0.000653425,-5.17107e-07,9.59024e-10,0.86511,0.000652394,-5.1423e-07,1.61393e-08,0.865762,0.000651414,-4.65812e-07,-5.91149e-09,0.866413,0.000650465,-4.83546e-07,7.50665e-09,0.867063,0.00064952,-4.61026e-07,-2.4115e-08,0.867712,0.000648526,-5.33371e-07,2.93486e-08,0.86836,0.000647547,-4.45325e-07,-3.36748e-08,0.869007,0.000646555,-5.4635e-07,4.57461e-08,0.869653,0.0006456,-4.09112e-07,-3.01002e-08,0.870298,0.000644691,-4.99412e-07,1.50501e-08,0.870942,0.000643738,-4.54262e-07,-3.01002e-08,0.871585,0.000642739,-5.44563e-07,4.57461e-08,0.872228,0.000641787,-4.07324e-07,-3.36748e-08,0.872869,0.000640871,-5.08349e-07,2.93486e-08,0.873509,0.000639943,-4.20303e-07,-2.4115e-08,0.874149,0.00063903,-4.92648e-07,7.50655e-09,0.874787,0.000638067,-4.70128e-07,-5.91126e-09,0.875425,0.000637109,-4.87862e-07,1.61385e-08,0.876062,0.000636182,-4.39447e-07,9.61961e-10,0.876697,0.000635306,-4.36561e-07,-1.99863e-08,0.877332,0.000634373,-4.9652e-07,1.93785e-08,0.877966,0.000633438,-4.38384e-07,2.07697e-09,0.878599,0.000632567,-4.32153e-07,-2.76864e-08,0.879231,0.00063162,-5.15212e-07,4.90641e-08,0.879862,0.000630737,-3.6802e-07,-4.93606e-08,0.880493,0.000629852,-5.16102e-07,2.9169e-08,0.881122,0.000628908,-4.28595e-07,-7.71083e-09,0.881751,0.000628027,-4.51727e-07,1.6744e-09,0.882378,0.000627129,-4.46704e-07,1.01317e-09,0.883005,0.000626239,-4.43665e-07,-5.72703e-09,0.883631,0.000625334,-4.60846e-07,2.1895e-08,0.884255,0.000624478,-3.95161e-07,-2.22481e-08,0.88488,0.000623621,-4.61905e-07,7.4928e-09,0.885503,0.00062272,-4.39427e-07,-7.72306e-09,0.886125,0.000621818,-4.62596e-07,2.33995e-08,0.886746,0.000620963,-3.92398e-07,-2.62704e-08,0.887367,0.000620099,-4.71209e-07,2.20775e-08,0.887987,0.000619223,-4.04976e-07,-2.43496e-09,0.888605,0.000618406,-4.12281e-07,-1.23377e-08,0.889223,0.000617544,-4.49294e-07,-7.81876e-09,0.88984,0.000616622,-4.72751e-07,4.36128e-08,0.890457,0.000615807,-3.41912e-07,-4.7423e-08,0.891072,0.000614981,-4.84181e-07,2.68698e-08,0.891687,0.000614093,-4.03572e-07,-4.51384e-10,0.8923,0.000613285,-4.04926e-07,-2.50643e-08,0.892913,0.0006124,-4.80119e-07,4.11038e-08,0.893525,0.000611563,-3.56808e-07,-2.01414e-08,0.894136,0.000610789,-4.17232e-07,-2.01426e-08,0.894747,0.000609894,-4.7766e-07,4.11073e-08,0.895356,0.000609062,-3.54338e-07,-2.50773e-08,0.895965,0.000608278,-4.2957e-07,-4.02954e-10,0.896573,0.000607418,-4.30779e-07,2.66891e-08,0.89718,0.000606636,-3.50711e-07,-4.67489e-08,0.897786,0.000605795,-4.90958e-07,4.10972e-08,0.898391,0.000604936,-3.67666e-07,1.56948e-09,0.898996,0.000604205,-3.62958e-07,-4.73751e-08,0.8996,0.000603337,-5.05083e-07,6.87214e-08,0.900202,0.000602533,-2.98919e-07,-4.86966e-08,0.900805,0.000601789,-4.45009e-07,6.85589e-09,0.901406,0.00060092,-4.24441e-07,2.1273e-08,0.902007,0.000600135,-3.60622e-07,-3.23434e-08,0.902606,0.000599317,-4.57652e-07,4.84959e-08,0.903205,0.000598547,-3.12164e-07,-4.24309e-08,0.903803,0.000597795,-4.39457e-07,2.01844e-09,0.904401,0.000596922,-4.33402e-07,3.43571e-08,0.904997,0.000596159,-3.30331e-07,-2.02374e-08,0.905593,0.000595437,-3.91043e-07,-1.30123e-08,0.906188,0.000594616,-4.3008e-07,1.26819e-08,0.906782,0.000593794,-3.92034e-07,2.18894e-08,0.907376,0.000593076,-3.26366e-07,-4.06349e-08,0.907968,0.000592301,-4.4827e-07,2.1441e-08,0.90856,0.000591469,-3.83947e-07,1.44754e-08,0.909151,0.000590744,-3.40521e-07,-1.97379e-08,0.909742,0.000590004,-3.99735e-07,4.87161e-09,0.910331,0.000589219,-3.8512e-07,2.51532e-10,0.91092,0.00058845,-3.84366e-07,-5.87776e-09,0.911508,0.000587663,-4.01999e-07,2.32595e-08,0.912096,0.000586929,-3.3222e-07,-2.75554e-08,0.912682,0.000586182,-4.14887e-07,2.73573e-08,0.913268,0.000585434,-3.32815e-07,-2.22692e-08,0.913853,0.000584702,-3.99622e-07,2.11486e-09,0.914437,0.000583909,-3.93278e-07,1.38098e-08,0.915021,0.000583164,-3.51848e-07,2.25042e-09,0.915604,0.000582467,-3.45097e-07,-2.28115e-08,0.916186,0.000581708,-4.13531e-07,2.93911e-08,0.916767,0.000580969,-3.25358e-07,-3.51481e-08,0.917348,0.000580213,-4.30803e-07,5.15967e-08,0.917928,0.000579506,-2.76012e-07,-5.20296e-08,0.918507,0.000578798,-4.32101e-07,3.73124e-08,0.919085,0.000578046,-3.20164e-07,-3.76154e-08,0.919663,0.000577293,-4.3301e-07,5.35447e-08,0.92024,0.000576587,-2.72376e-07,-5.7354e-08,0.920816,0.000575871,-4.44438e-07,5.66621e-08,0.921391,0.000575152,-2.74452e-07,-5.00851e-08,0.921966,0.000574453,-4.24707e-07,2.4469e-08,0.92254,0.000573677,-3.513e-07,1.18138e-08,0.923114,0.000573009,-3.15859e-07,-1.21195e-08,0.923686,0.000572341,-3.52217e-07,-2.29403e-08,0.924258,0.000571568,-4.21038e-07,4.4276e-08,0.924829,0.000570859,-2.8821e-07,-3.49546e-08,0.9254,0.000570178,-3.93074e-07,3.59377e-08,0.92597,0.000569499,-2.85261e-07,-4.91915e-08,0.926539,0.000568781,-4.32835e-07,4.16189e-08,0.927107,0.00056804,-3.07979e-07,1.92523e-09,0.927675,0.00056743,-3.02203e-07,-4.93198e-08,0.928242,0.000566678,-4.50162e-07,7.61447e-08,0.928809,0.000566006,-2.21728e-07,-7.6445e-08,0.929374,0.000565333,-4.51063e-07,5.08216e-08,0.929939,0.000564583,-2.98599e-07,-7.63212e-09,0.930503,0.000563963,-3.21495e-07,-2.02931e-08,0.931067,0.000563259,-3.82374e-07,2.92001e-08,0.93163,0.000562582,-2.94774e-07,-3.69025e-08,0.932192,0.000561882,-4.05482e-07,5.88053e-08,0.932754,0.000561247,-2.29066e-07,-7.91094e-08,0.933315,0.000560552,-4.66394e-07,7.88184e-08,0.933875,0.000559856,-2.29939e-07,-5.73501e-08,0.934434,0.000559224,-4.01989e-07,3.13727e-08,0.934993,0.000558514,-3.07871e-07,-8.53611e-09,0.935551,0.000557873,-3.33479e-07,2.77175e-09,0.936109,0.000557214,-3.25164e-07,-2.55091e-09,0.936666,0.000556556,-3.32817e-07,7.43188e-09,0.937222,0.000555913,-3.10521e-07,-2.71766e-08,0.937778,0.00055521,-3.92051e-07,4.167e-08,0.938333,0.000554551,-2.67041e-07,-2.02941e-08,0.938887,0.000553956,-3.27923e-07,-2.00984e-08,0.93944,0.00055324,-3.88218e-07,4.10828e-08,0.939993,0.000552587,-2.6497e-07,-2.50237e-08,0.940546,0.000551982,-3.40041e-07,-5.92583e-10,0.941097,0.0005513,-3.41819e-07,2.7394e-08,0.941648,0.000550698,-2.59637e-07,-4.93788e-08,0.942199,0.000550031,-4.07773e-07,5.09119e-08,0.942748,0.000549368,-2.55038e-07,-3.50595e-08,0.943297,0.000548753,-3.60216e-07,2.97214e-08,0.943846,0.000548122,-2.71052e-07,-2.42215e-08,0.944394,0.000547507,-3.43716e-07,7.55985e-09,0.944941,0.000546842,-3.21037e-07,-6.01796e-09,0.945487,0.000546182,-3.3909e-07,1.65119e-08,0.946033,0.000545553,-2.89555e-07,-4.2498e-10,0.946578,0.000544973,-2.9083e-07,-1.4812e-08,0.947123,0.000544347,-3.35266e-07,6.83068e-11,0.947667,0.000543676,-3.35061e-07,1.45388e-08,0.94821,0.00054305,-2.91444e-07,1.38123e-09,0.948753,0.000542471,-2.87301e-07,-2.00637e-08,0.949295,0.000541836,-3.47492e-07,1.92688e-08,0.949837,0.000541199,-2.89685e-07,2.59298e-09,0.950378,0.000540628,-2.81906e-07,-2.96407e-08,0.950918,0.000539975,-3.70829e-07,5.63652e-08,0.951458,0.000539402,-2.01733e-07,-7.66107e-08,0.951997,0.000538769,-4.31565e-07,7.12638e-08,0.952535,0.00053812,-2.17774e-07,-2.96305e-08,0.953073,0.000537595,-3.06665e-07,-1.23464e-08,0.95361,0.000536945,-3.43704e-07,1.94114e-08,0.954147,0.000536316,-2.8547e-07,-5.69451e-09,0.954683,0.000535728,-3.02554e-07,3.36666e-09,0.955219,0.000535133,-2.92454e-07,-7.77208e-09,0.955753,0.000534525,-3.1577e-07,2.77216e-08,0.956288,0.000533976,-2.32605e-07,-4.35097e-08,0.956821,0.00053338,-3.63134e-07,2.7108e-08,0.957354,0.000532735,-2.8181e-07,-5.31772e-09,0.957887,0.000532156,-2.97764e-07,-5.83718e-09,0.958419,0.000531543,-3.15275e-07,2.86664e-08,0.95895,0.000530998,-2.29276e-07,-4.9224e-08,0.959481,0.000530392,-3.76948e-07,4.90201e-08,0.960011,0.000529785,-2.29887e-07,-2.76471e-08,0.96054,0.000529243,-3.12829e-07,1.96385e-09,0.961069,0.000528623,-3.06937e-07,1.97917e-08,0.961598,0.000528068,-2.47562e-07,-2.15261e-08,0.962125,0.000527508,-3.1214e-07,6.70795e-09,0.962653,0.000526904,-2.92016e-07,-5.30573e-09,0.963179,0.000526304,-3.07934e-07,1.4515e-08,0.963705,0.000525732,-2.64389e-07,6.85048e-09,0.964231,0.000525224,-2.43837e-07,-4.19169e-08,0.964756,0.00052461,-3.69588e-07,4.1608e-08,0.96528,0.000523996,-2.44764e-07,-5.30598e-09,0.965804,0.000523491,-2.60682e-07,-2.03841e-08,0.966327,0.000522908,-3.21834e-07,2.72378e-08,0.966849,0.000522346,-2.40121e-07,-2.89625e-08,0.967371,0.000521779,-3.27008e-07,2.90075e-08,0.967893,0.000521212,-2.39986e-07,-2.74629e-08,0.968414,0.00052065,-3.22374e-07,2.12396e-08,0.968934,0.000520069,-2.58656e-07,2.10922e-09,0.969454,0.000519558,-2.52328e-07,-2.96765e-08,0.969973,0.000518964,-3.41357e-07,5.6992e-08,0.970492,0.000518452,-1.70382e-07,-7.90821e-08,0.97101,0.000517874,-4.07628e-07,8.05224e-08,0.971528,0.000517301,-1.66061e-07,-6.41937e-08,0.972045,0.000516776,-3.58642e-07,5.70429e-08,0.972561,0.00051623,-1.87513e-07,-4.47686e-08,0.973077,0.00051572,-3.21819e-07,2.82237e-09,0.973593,0.000515085,-3.13352e-07,3.34792e-08,0.974108,0.000514559,-2.12914e-07,-1.75298e-08,0.974622,0.000514081,-2.65503e-07,-2.29648e-08,0.975136,0.000513481,-3.34398e-07,4.97843e-08,0.975649,0.000512961,-1.85045e-07,-5.6963e-08,0.976162,0.00051242,-3.55934e-07,5.88585e-08,0.976674,0.000511885,-1.79359e-07,-5.92616e-08,0.977185,0.000511348,-3.57143e-07,5.89785e-08,0.977696,0.000510811,-1.80208e-07,-5.74433e-08,0.978207,0.000510278,-3.52538e-07,5.15854e-08,0.978717,0.000509728,-1.97781e-07,-2.9689e-08,0.979226,0.000509243,-2.86848e-07,7.56591e-09,0.979735,0.000508692,-2.64151e-07,-5.74649e-10,0.980244,0.000508162,-2.65875e-07,-5.26732e-09,0.980752,0.000507615,-2.81677e-07,2.16439e-08,0.981259,0.000507116,-2.16745e-07,-2.17037e-08,0.981766,0.000506618,-2.81856e-07,5.56636e-09,0.982272,0.000506071,-2.65157e-07,-5.61689e-10,0.982778,0.000505539,-2.66842e-07,-3.31963e-09,0.983283,0.000504995,-2.76801e-07,1.38402e-08,0.983788,0.000504483,-2.3528e-07,7.56339e-09,0.984292,0.000504035,-2.1259e-07,-4.40938e-08,0.984796,0.000503478,-3.44871e-07,4.96026e-08,0.985299,0.000502937,-1.96064e-07,-3.51071e-08,0.985802,0.000502439,-3.01385e-07,3.12212e-08,0.986304,0.00050193,-2.07721e-07,-3.0173e-08,0.986806,0.000501424,-2.9824e-07,2.9866e-08,0.987307,0.000500917,-2.08642e-07,-2.96865e-08,0.987808,0.000500411,-2.97702e-07,2.92753e-08,0.988308,0.000499903,-2.09876e-07,-2.78101e-08,0.988807,0.0004994,-2.93306e-07,2.23604e-08,0.989307,0.000498881,-2.26225e-07,-2.02681e-09,0.989805,0.000498422,-2.32305e-07,-1.42531e-08,0.990303,0.000497915,-2.75065e-07,-5.65232e-10,0.990801,0.000497363,-2.76761e-07,1.65141e-08,0.991298,0.000496859,-2.27218e-07,-5.88639e-09,0.991795,0.000496387,-2.44878e-07,7.0315e-09,0.992291,0.000495918,-2.23783e-07,-2.22396e-08,0.992787,0.000495404,-2.90502e-07,2.23224e-08,0.993282,0.00049489,-2.23535e-07,-7.44543e-09,0.993776,0.000494421,-2.45871e-07,7.45924e-09,0.994271,0.000493951,-2.23493e-07,-2.23915e-08,0.994764,0.000493437,-2.90668e-07,2.25021e-08,0.995257,0.000492923,-2.23161e-07,-8.01218e-09,0.99575,0.000492453,-2.47198e-07,9.54669e-09,0.996242,0.000491987,-2.18558e-07,-3.01746e-08,0.996734,0.000491459,-3.09082e-07,5.1547e-08,0.997225,0.000490996,-1.54441e-07,-5.68039e-08,0.997716,0.000490517,-3.24853e-07,5.64594e-08,0.998206,0.000490036,-1.55474e-07,-4.98245e-08,0.998696,0.000489576,-3.04948e-07,2.36292e-08,0.999186,0.000489037,-2.3406e-07,1.49121e-08,0.999674,0.000488613,-1.89324e-07,-2.3673e-08,1.00016,0.000488164,-2.60343e-07,2.01754e-08,1.00065,0.000487704,-1.99816e-07,-5.70288e-08,1.00114,0.000487133,-3.70903e-07,8.87303e-08,1.00162,0.000486657,-1.04712e-07,-5.94737e-08,1.00211,0.000486269,-2.83133e-07,2.99553e-08,1.0026,0.000485793,-1.93267e-07,-6.03474e-08,1.00308,0.000485225,-3.74309e-07,9.2225e-08,1.00357,0.000484754,-9.76345e-08,-7.0134e-08,1.00405,0.000484348,-3.08036e-07,6.91016e-08,1.00454,0.000483939,-1.00731e-07,-8.70633e-08,1.00502,0.000483476,-3.61921e-07,4.07328e-08,1.0055,0.000482875,-2.39723e-07,4.33413e-08,1.00599,0.000482525,-1.09699e-07,-9.48886e-08,1.00647,0.000482021,-3.94365e-07,9.77947e-08,1.00695,0.000481526,-1.00981e-07,-5.78713e-08,1.00743,0.00048115,-2.74595e-07,1.44814e-08,1.00791,0.000480645,-2.31151e-07,-5.42665e-11,1.00839,0.000480182,-2.31314e-07,-1.42643e-08,1.00887,0.000479677,-2.74106e-07,5.71115e-08,1.00935,0.0004793,-1.02772e-07,-9.49724e-08,1.00983,0.000478809,-3.87689e-07,8.43596e-08,1.01031,0.000478287,-1.3461e-07,-4.04755e-09,1.01079,0.000478006,-1.46753e-07,-6.81694e-08,1.01127,0.000477508,-3.51261e-07,3.83067e-08,1.01174,0.00047692,-2.36341e-07,3.41521e-08,1.01222,0.00047655,-1.33885e-07,-5.57058e-08,1.0127,0.000476115,-3.01002e-07,6.94616e-08,1.01317,0.000475721,-9.26174e-08,-1.02931e-07,1.01365,0.000475227,-4.01412e-07,1.03846e-07,1.01412,0.000474736,-8.98751e-08,-7.40321e-08,1.0146,0.000474334,-3.11971e-07,7.30735e-08,1.01507,0.00047393,-9.27508e-08,-9.90527e-08,1.01554,0.000473447,-3.89909e-07,8.47188e-08,1.01602,0.000472921,-1.35753e-07,-1.40381e-09,1.01649,0.000472645,-1.39964e-07,-7.91035e-08,1.01696,0.000472128,-3.77275e-07,7.93993e-08,1.01744,0.000471612,-1.39077e-07,-7.52607e-11,1.01791,0.000471334,-1.39302e-07,-7.90983e-08,1.01838,0.000470818,-3.76597e-07,7.80499e-08,1.01885,0.000470299,-1.42448e-07,5.31733e-09,1.01932,0.00047003,-1.26496e-07,-9.93193e-08,1.01979,0.000469479,-4.24453e-07,1.53541e-07,1.02026,0.00046909,3.617e-08,-1.57217e-07,1.02073,0.000468691,-4.35482e-07,1.177e-07,1.02119,0.000468173,-8.23808e-08,-7.51659e-08,1.02166,0.000467783,-3.07878e-07,6.37538e-08,1.02213,0.000467358,-1.16617e-07,-6.064e-08,1.0226,0.000466943,-2.98537e-07,5.9597e-08,1.02306,0.000466525,-1.19746e-07,-5.85386e-08,1.02353,0.00046611,-2.95362e-07,5.53482e-08,1.024,0.000465685,-1.29317e-07,-4.36449e-08,1.02446,0.000465296,-2.60252e-07,2.20268e-11,1.02493,0.000464775,-2.60186e-07,4.35568e-08,1.02539,0.000464386,-1.29516e-07,-5.50398e-08,1.02586,0.000463961,-2.94635e-07,5.73932e-08,1.02632,0.000463544,-1.22456e-07,-5.53236e-08,1.02678,0.000463133,-2.88426e-07,4.46921e-08,1.02725,0.000462691,-1.5435e-07,-4.23534e-09,1.02771,0.000462369,-1.67056e-07,-2.77507e-08,1.02817,0.000461952,-2.50308e-07,-3.97101e-09,1.02863,0.000461439,-2.62221e-07,4.36348e-08,1.02909,0.000461046,-1.31317e-07,-5.13589e-08,1.02955,0.000460629,-2.85394e-07,4.25913e-08,1.03001,0.000460186,-1.5762e-07,2.0285e-10,1.03047,0.000459871,-1.57011e-07,-4.34027e-08,1.03093,0.000459427,-2.87219e-07,5.41987e-08,1.03139,0.000459015,-1.24623e-07,-5.4183e-08,1.03185,0.000458604,-2.87172e-07,4.33239e-08,1.03231,0.000458159,-1.572e-07,9.65817e-11,1.03277,0.000457845,-1.56911e-07,-4.37103e-08,1.03323,0.0004574,-2.88041e-07,5.55351e-08,1.03368,0.000456991,-1.21436e-07,-5.9221e-08,1.03414,0.00045657,-2.99099e-07,6.21394e-08,1.0346,0.000456158,-1.1268e-07,-7.01275e-08,1.03505,0.000455723,-3.23063e-07,9.91614e-08,1.03551,0.000455374,-2.55788e-08,-8.80996e-08,1.03596,0.000455058,-2.89878e-07,1.48184e-08,1.03642,0.000454523,-2.45422e-07,2.88258e-08,1.03687,0.000454119,-1.58945e-07,-1.09125e-08,1.03733,0.000453768,-1.91682e-07,1.48241e-08,1.03778,0.000453429,-1.4721e-07,-4.83838e-08,1.03823,0.00045299,-2.92361e-07,5.95019e-08,1.03869,0.000452584,-1.13856e-07,-7.04146e-08,1.03914,0.000452145,-3.25099e-07,1.02947e-07,1.03959,0.000451803,-1.62583e-08,-1.02955e-07,1.04004,0.000451462,-3.25123e-07,7.04544e-08,1.04049,0.000451023,-1.1376e-07,-5.96534e-08,1.04094,0.000450616,-2.9272e-07,4.89499e-08,1.04139,0.000450178,-1.45871e-07,-1.69369e-08,1.04184,0.000449835,-1.96681e-07,1.87977e-08,1.04229,0.000449498,-1.40288e-07,-5.82539e-08,1.04274,0.000449043,-3.1505e-07,9.50087e-08,1.04319,0.000448698,-3.00238e-08,-8.33623e-08,1.04364,0.000448388,-2.80111e-07,2.20363e-11,1.04409,0.000447828,-2.80045e-07,8.32742e-08,1.04454,0.000447517,-3.02221e-08,-9.47002e-08,1.04498,0.000447173,-3.14323e-07,5.7108e-08,1.04543,0.000446716,-1.42999e-07,-1.45225e-08,1.04588,0.000446386,-1.86566e-07,9.82022e-10,1.04632,0.000446016,-1.8362e-07,1.05944e-08,1.04677,0.00044568,-1.51837e-07,-4.33597e-08,1.04721,0.000445247,-2.81916e-07,4.36352e-08,1.04766,0.000444814,-1.51011e-07,-1.19717e-08,1.0481,0.000444476,-1.86926e-07,4.25158e-09,1.04855,0.000444115,-1.74171e-07,-5.03461e-09,1.04899,0.000443751,-1.89275e-07,1.58868e-08,1.04944,0.00044342,-1.41614e-07,-5.85127e-08,1.04988,0.000442961,-3.17152e-07,9.89548e-08,1.05032,0.000442624,-2.0288e-08,-9.88878e-08,1.05076,0.000442287,-3.16951e-07,5.81779e-08,1.05121,0.000441827,-1.42418e-07,-1.46144e-08,1.05165,0.000441499,-1.86261e-07,2.79892e-10,1.05209,0.000441127,-1.85421e-07,1.34949e-08,1.05253,0.000440797,-1.44937e-07,-5.42594e-08,1.05297,0.000440344,-3.07715e-07,8.43335e-08,1.05341,0.000439982,-5.47146e-08,-4.46558e-08,1.05385,0.000439738,-1.88682e-07,-2.49193e-08,1.05429,0.000439286,-2.6344e-07,2.5124e-08,1.05473,0.000438835,-1.88068e-07,4.36328e-08,1.05517,0.000438589,-5.71699e-08,-8.04459e-08,1.05561,0.000438234,-2.98508e-07,3.97324e-08,1.05605,0.000437756,-1.79311e-07,4.07258e-08,1.05648,0.000437519,-5.71332e-08,-8.34263e-08,1.05692,0.000437155,-3.07412e-07,5.45608e-08,1.05736,0.000436704,-1.4373e-07,-1.56078e-08,1.05779,0.000436369,-1.90553e-07,7.87043e-09,1.05823,0.000436012,-1.66942e-07,-1.58739e-08,1.05867,0.00043563,-2.14563e-07,5.56251e-08,1.0591,0.000435368,-4.76881e-08,-8.74172e-08,1.05954,0.000435011,-3.0994e-07,5.56251e-08,1.05997,0.000434558,-1.43064e-07,-1.58739e-08,1.06041,0.000434224,-1.90686e-07,7.87042e-09,1.06084,0.000433866,-1.67075e-07,-1.56078e-08,1.06127,0.000433485,-2.13898e-07,5.45609e-08,1.06171,0.000433221,-5.02157e-08,-8.34263e-08,1.06214,0.00043287,-3.00495e-07,4.07258e-08,1.06257,0.000432391,-1.78317e-07,3.97325e-08,1.063,0.000432154,-5.91198e-08,-8.04464e-08,1.06344,0.000431794,-3.00459e-07,4.36347e-08,1.06387,0.000431324,-1.69555e-07,2.5117e-08,1.0643,0.000431061,-9.42041e-08,-2.48934e-08,1.06473,0.000430798,-1.68884e-07,-4.47527e-08,1.06516,0.000430326,-3.03142e-07,8.46951e-08,1.06559,0.000429973,-4.90573e-08,-5.56089e-08,1.06602,0.000429708,-2.15884e-07,1.85314e-08,1.06645,0.000429332,-1.6029e-07,-1.85166e-08,1.06688,0.000428956,-2.1584e-07,5.5535e-08,1.06731,0.000428691,-4.92347e-08,-8.44142e-08,1.06774,0.000428339,-3.02477e-07,4.37032e-08,1.06816,0.000427865,-1.71368e-07,2.88107e-08,1.06859,0.000427609,-8.49356e-08,-3.97367e-08,1.06902,0.00042732,-2.04146e-07,1.09267e-08,1.06945,0.000426945,-1.71365e-07,-3.97023e-09,1.06987,0.00042659,-1.83276e-07,4.9542e-09,1.0703,0.000426238,-1.68414e-07,-1.58466e-08,1.07073,0.000425854,-2.15953e-07,5.84321e-08,1.07115,0.000425597,-4.0657e-08,-9.86725e-08,1.07158,0.00042522,-3.36674e-07,9.78392e-08,1.072,0.00042484,-4.31568e-08,-5.42658e-08,1.07243,0.000424591,-2.05954e-07,1.45377e-11,1.07285,0.000424179,-2.0591e-07,5.42076e-08,1.07328,0.00042393,-4.32877e-08,-9.76357e-08,1.0737,0.00042355,-3.36195e-07,9.79165e-08,1.07412,0.000423172,-4.24451e-08,-5.56118e-08,1.07455,0.00042292,-2.09281e-07,5.32143e-09,1.07497,0.000422518,-1.93316e-07,3.43261e-08,1.07539,0.000422234,-9.0338e-08,-2.34165e-08,1.07581,0.000421983,-1.60588e-07,-5.98692e-08,1.07623,0.000421482,-3.40195e-07,1.43684e-07,1.07666,0.000421233,9.08574e-08,-1.5724e-07,1.07708,0.000420943,-3.80862e-07,1.27647e-07,1.0775,0.000420564,2.0791e-09,-1.1493e-07,1.07792,0.000420223,-3.4271e-07,9.36534e-08,1.07834,0.000419819,-6.17499e-08,-2.12653e-08,1.07876,0.000419632,-1.25546e-07,-8.59219e-09,1.07918,0.000419355,-1.51322e-07,-6.35752e-08,1.0796,0.000418861,-3.42048e-07,1.43684e-07,1.08002,0.000418608,8.90034e-08,-1.53532e-07,1.08043,0.000418326,-3.71593e-07,1.12817e-07,1.08085,0.000417921,-3.31414e-08,-5.93184e-08,1.08127,0.000417677,-2.11097e-07,5.24697e-09,1.08169,0.00041727,-1.95356e-07,3.83305e-08,1.0821,0.000416995,-8.03642e-08,-3.93597e-08,1.08252,0.000416716,-1.98443e-07,-1.0094e-10,1.08294,0.000416319,-1.98746e-07,3.97635e-08,1.08335,0.00041604,-7.94557e-08,-3.97437e-08,1.08377,0.000415762,-1.98687e-07,1.94215e-12,1.08419,0.000415365,-1.98681e-07,3.97359e-08,1.0846,0.000415087,-7.94732e-08,-3.97362e-08,1.08502,0.000414809,-1.98682e-07,-4.31063e-13,1.08543,0.000414411,-1.98683e-07,3.97379e-08,1.08584,0.000414133,-7.94694e-08,-3.97418e-08,1.08626,0.000413855,-1.98695e-07,2.00563e-11,1.08667,0.000413458,-1.98635e-07,3.96616e-08,1.08709,0.000413179,-7.965e-08,-3.9457e-08,1.0875,0.000412902,-1.98021e-07,-1.04281e-09,1.08791,0.000412502,-2.01149e-07,4.36282e-08,1.08832,0.000412231,-7.02648e-08,-5.42608e-08,1.08874,0.000411928,-2.33047e-07,5.42057e-08,1.08915,0.000411624,-7.04301e-08,-4.33527e-08,1.08956,0.000411353,-2.00488e-07,-4.07378e-12,1.08997,0.000410952,-2.005e-07,4.3369e-08,1.09038,0.000410681,-7.03934e-08,-5.42627e-08,1.09079,0.000410378,-2.33182e-07,5.44726e-08,1.0912,0.000410075,-6.97637e-08,-4.44186e-08,1.09161,0.000409802,-2.03019e-07,3.99235e-09,1.09202,0.000409408,-1.91042e-07,2.84491e-08,1.09243,0.000409111,-1.05695e-07,1.42043e-09,1.09284,0.000408904,-1.01434e-07,-3.41308e-08,1.09325,0.000408599,-2.03826e-07,1.58937e-08,1.09366,0.000408239,-1.56145e-07,-2.94438e-08,1.09406,0.000407838,-2.44476e-07,1.01881e-07,1.09447,0.000407655,6.11676e-08,-1.39663e-07,1.09488,0.000407358,-3.57822e-07,9.91432e-08,1.09529,0.00040694,-6.03921e-08,-1.84912e-08,1.09569,0.000406764,-1.15866e-07,-2.51785e-08,1.0961,0.000406457,-1.91401e-07,-4.03115e-12,1.09651,0.000406074,-1.91413e-07,2.51947e-08,1.09691,0.000405767,-1.15829e-07,1.84346e-08,1.09732,0.00040559,-6.05254e-08,-9.89332e-08,1.09772,0.000405172,-3.57325e-07,1.3888e-07,1.09813,0.000404874,5.93136e-08,-9.8957e-08,1.09853,0.000404696,-2.37557e-07,1.853e-08,1.09894,0.000404277,-1.81968e-07,2.48372e-08,1.09934,0.000403987,-1.07456e-07,1.33047e-09,1.09975,0.000403776,-1.03465e-07,-3.01591e-08,1.10015,0.000403479,-1.93942e-07,9.66054e-11,1.10055,0.000403091,-1.93652e-07,2.97727e-08,1.10096,0.000402793,-1.04334e-07,2.19273e-11,1.10136,0.000402585,-1.04268e-07,-2.98604e-08,1.10176,0.000402287,-1.93849e-07,2.10325e-10,1.10216,0.0004019,-1.93218e-07,2.90191e-08,1.10256,0.0004016,-1.06161e-07,2.92264e-09,1.10297,0.000401397,-9.73931e-08,-4.07096e-08,1.10337,0.00040108,-2.19522e-07,4.07067e-08,1.10377,0.000400763,-9.7402e-08,-2.90783e-09,1.10417,0.000400559,-1.06126e-07,-2.90754e-08,1.10457,0.00040026,-1.93352e-07,9.00021e-14,1.10497,0.000399873,-1.93351e-07,2.9075e-08,1.10537,0.000399574,-1.06126e-07,2.90902e-09,1.10577,0.00039937,-9.73992e-08,-4.07111e-08,1.10617,0.000399053,-2.19533e-07,4.07262e-08,1.10657,0.000398736,-9.73541e-08,-2.98424e-09,1.10697,0.000398533,-1.06307e-07,-2.87892e-08,1.10736,0.000398234,-1.92674e-07,-1.06824e-09,1.10776,0.000397845,-1.95879e-07,3.30622e-08,1.10816,0.000397552,-9.66926e-08,-1.19712e-08,1.10856,0.000397323,-1.32606e-07,1.48225e-08,1.10895,0.000397102,-8.81387e-08,-4.73187e-08,1.10935,0.000396784,-2.30095e-07,5.52429e-08,1.10975,0.00039649,-6.4366e-08,-5.44437e-08,1.11014,0.000396198,-2.27697e-07,4.33226e-08,1.11054,0.000395872,-9.77293e-08,3.62656e-10,1.11094,0.000395678,-9.66414e-08,-4.47732e-08,1.11133,0.00039535,-2.30961e-07,5.95208e-08,1.11173,0.000395067,-5.23985e-08,-7.41008e-08,1.11212,0.00039474,-2.74701e-07,1.17673e-07,1.11252,0.000394543,7.83181e-08,-1.58172e-07,1.11291,0.000394225,-3.96199e-07,1.57389e-07,1.1133,0.000393905,7.59679e-08,-1.13756e-07,1.1137,0.000393716,-2.653e-07,5.92165e-08,1.11409,0.000393363,-8.76507e-08,-3.90074e-09,1.11449,0.000393176,-9.93529e-08,-4.36136e-08,1.11488,0.000392846,-2.30194e-07,5.91457e-08,1.11527,0.000392563,-5.27564e-08,-7.376e-08,1.11566,0.000392237,-2.74037e-07,1.16685e-07,1.11606,0.000392039,7.60189e-08,-1.54562e-07,1.11645,0.000391727,-3.87667e-07,1.43935e-07,1.11684,0.000391384,4.4137e-08,-6.35487e-08,1.11723,0.000391281,-1.46509e-07,-8.94896e-09,1.11762,0.000390961,-1.73356e-07,-1.98647e-08,1.11801,0.000390555,-2.3295e-07,8.8408e-08,1.1184,0.000390354,3.22736e-08,-9.53486e-08,1.11879,0.000390133,-2.53772e-07,5.45677e-08,1.11918,0.000389789,-9.0069e-08,-3.71296e-09,1.11957,0.000389598,-1.01208e-07,-3.97159e-08,1.11996,0.000389276,-2.20355e-07,4.33671e-08,1.12035,0.000388966,-9.02542e-08,-1.45431e-08,1.12074,0.000388741,-1.33883e-07,1.48052e-08,1.12113,0.000388518,-8.94678e-08,-4.46778e-08,1.12152,0.000388205,-2.23501e-07,4.46966e-08,1.12191,0.000387892,-8.94114e-08,-1.48992e-08,1.12229,0.000387669,-1.34109e-07,1.49003e-08,1.12268,0.000387445,-8.94082e-08,-4.47019e-08,1.12307,0.000387132,-2.23514e-07,4.4698e-08,1.12345,0.000386819,-8.942e-08,-1.48806e-08,1.12384,0.000386596,-1.34062e-07,1.48245e-08,1.12423,0.000386372,-8.95885e-08,-4.44172e-08,1.12461,0.00038606,-2.2284e-07,4.36351e-08,1.125,0.000385745,-9.19348e-08,-1.09139e-08,1.12539,0.000385528,-1.24677e-07,2.05584e-11,1.12577,0.000385279,-1.24615e-07,1.08317e-08,1.12616,0.000385062,-9.21198e-08,-4.33473e-08,1.12654,0.000384748,-2.22162e-07,4.33481e-08,1.12693,0.000384434,-9.21174e-08,-1.08356e-08,1.12731,0.000384217,-1.24624e-07,-5.50907e-12,1.12769,0.000383968,-1.24641e-07,1.08577e-08,1.12808,0.000383751,-9.20679e-08,-4.34252e-08,1.12846,0.000383437,-2.22343e-07,4.36337e-08,1.12884,0.000383123,-9.14422e-08,-1.19005e-08,1.12923,0.000382904,-1.27144e-07,3.96813e-09,1.12961,0.000382662,-1.15239e-07,-3.97207e-09,1.12999,0.000382419,-1.27155e-07,1.19201e-08,1.13038,0.000382201,-9.1395e-08,-4.37085e-08,1.13076,0.000381887,-2.2252e-07,4.37046e-08,1.13114,0.000381573,-9.14068e-08,-1.19005e-08,1.13152,0.000381355,-1.27108e-07,3.89734e-09,1.1319,0.000381112,-1.15416e-07,-3.68887e-09,1.13228,0.00038087,-1.26483e-07,1.08582e-08,1.13266,0.00038065,-9.39083e-08,-3.97438e-08,1.13304,0.000380343,-2.1314e-07,2.89076e-08,1.13342,0.000380003,-1.26417e-07,4.33225e-08,1.1338,0.00037988,3.55072e-09,-8.29883e-08,1.13418,0.000379638,-2.45414e-07,5.0212e-08,1.13456,0.000379298,-9.47781e-08,1.34964e-09,1.13494,0.000379113,-9.07292e-08,-5.56105e-08,1.13532,0.000378764,-2.57561e-07,1.01883e-07,1.1357,0.000378555,4.80889e-08,-1.13504e-07,1.13608,0.000378311,-2.92423e-07,1.13713e-07,1.13646,0.000378067,4.87176e-08,-1.02931e-07,1.13683,0.000377856,-2.60076e-07,5.95923e-08,1.13721,0.000377514,-8.12988e-08,-1.62288e-08,1.13759,0.000377303,-1.29985e-07,5.32278e-09,1.13797,0.000377059,-1.14017e-07,-5.06237e-09,1.13834,0.000376816,-1.29204e-07,1.49267e-08,1.13872,0.000376602,-8.44237e-08,-5.46444e-08,1.1391,0.000376269,-2.48357e-07,8.44417e-08,1.13947,0.000376026,4.96815e-09,-4.47039e-08,1.13985,0.000375902,-1.29143e-07,-2.48355e-08,1.14023,0.000375569,-2.0365e-07,2.48368e-08,1.1406,0.000375236,-1.2914e-07,4.46977e-08,1.14098,0.000375112,4.95341e-09,-8.44184e-08,1.14135,0.000374869,-2.48302e-07,5.45572e-08,1.14173,0.000374536,-8.463e-08,-1.46013e-08,1.1421,0.000374323,-1.28434e-07,3.8478e-09,1.14247,0.000374077,-1.1689e-07,-7.89941e-10,1.14285,0.000373841,-1.1926e-07,-6.88042e-10,1.14322,0.0003736,-1.21324e-07,3.54213e-09,1.1436,0.000373368,-1.10698e-07,-1.34805e-08,1.14397,0.000373107,-1.51139e-07,5.03798e-08,1.14434,0.000372767,0.,0.};
+
+    template <typename T, int scn, int dcn, bool srgb, int blueIdx> struct RGB2Luv;
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct RGB2Luv<float, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float _d = 1.f / (0.950456f + 15 + 1.088754f * 3);
+            const float _un = 13 * (4 * 0.950456f * _d);
+            const float _vn = 13 * (9 * _d);
+
+            float B = blueIdx == 0 ? src.x : src.z;
+            float G = src.y;
+            float R = blueIdx == 0 ? src.z : src.x;
+
+            if (srgb)
+            {
+                B = splineInterpolate(B * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+                G = splineInterpolate(G * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+                R = splineInterpolate(R * GAMMA_TAB_SIZE, c_sRGBGammaTab, GAMMA_TAB_SIZE);
+            }
+
+            float X = R * 0.412453f + G * 0.357580f + B * 0.180423f;
+            float Y = R * 0.212671f + G * 0.715160f + B * 0.072169f;
+            float Z = R * 0.019334f + G * 0.119193f + B * 0.950227f;
+
+            float L = splineInterpolate(Y * (LAB_CBRT_TAB_SIZE / 1.5f), c_LabCbrtTab, LAB_CBRT_TAB_SIZE);
+            L = 116.f * L - 16.f;
+
+            const float d = (4 * 13) / ::fmaxf(X + 15 * Y + 3 * Z, numeric_limits<float>::epsilon());
+            float u = L * (X * d - _un);
+            float v = L * ((9 * 0.25f) * Y * d - _vn);
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = L;
+            dst.y = u;
+            dst.z = v;
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct RGB2Luv<uchar, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x * (1.f / 255.f);
+            buf.y = src.y * (1.f / 255.f);
+            buf.z = src.z * (1.f / 255.f);
+
+            RGB2Luv<float, 3, 3, srgb, blueIdx> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x * 2.55f);
+            dst.y = saturate_cast<uchar>(buf.y * 0.72033898305084743f + 96.525423728813564f);
+            dst.z = saturate_cast<uchar>(buf.z * 0.99609375f + 139.453125f);
+
+            return dst;
+        }
+    };
+
+    // Luv to RGB
+
+    template <typename T, int scn, int dcn, bool srgb, int blueIdx> struct Luv2RGB;
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct Luv2RGB<float, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<float, scn>::type, typename MakeVec<float, dcn>::type>
+    {
+        __device__ typename MakeVec<float, dcn>::type operator ()(const typename MakeVec<float, scn>::type& src) const
+        {
+            const float _d = 1.f / (0.950456f + 15 + 1.088754f * 3);
+            const float _un = 4 * 0.950456f * _d;
+            const float _vn = 9 * _d;
+
+            float L = src.x;
+            float u = src.y;
+            float v = src.z;
+
+            float Y = (L + 16.f) * (1.f / 116.f);
+            Y = Y * Y * Y;
+
+            float d = (1.f / 13.f) / L;
+            u = u * d + _un;
+            v = v * d + _vn;
+
+            float iv = 1.f / v;
+            float X = 2.25f * u * Y * iv;
+            float Z = (12 - 3 * u - 20 * v) * Y * 0.25f * iv;
+
+            float B = 0.055648f * X - 0.204043f * Y + 1.057311f * Z;
+            float G = -0.969256f * X + 1.875991f * Y + 0.041556f * Z;
+            float R = 3.240479f * X - 1.537150f * Y - 0.498535f * Z;
+
+            if (srgb)
+            {
+                B = splineInterpolate(B * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+                G = splineInterpolate(G * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+                R = splineInterpolate(R * GAMMA_TAB_SIZE, c_sRGBInvGammaTab, GAMMA_TAB_SIZE);
+            }
+
+            typename MakeVec<float, dcn>::type dst;
+
+            dst.x = blueIdx == 0 ? B : R;
+            dst.y = G;
+            dst.z = blueIdx == 0 ? R : B;
+            setAlpha(dst, ColorChannel<float>::max());
+
+            return dst;
+        }
+    };
+
+    template <int scn, int dcn, bool srgb, int blueIdx> struct Luv2RGB<uchar, scn, dcn, srgb, blueIdx>
+            : unary_function<typename MakeVec<uchar, scn>::type, typename MakeVec<uchar, dcn>::type>
+    {
+        __device__ typename MakeVec<uchar, dcn>::type operator ()(const typename MakeVec<uchar, scn>::type& src) const
+        {
+            float3 buf;
+
+            buf.x = src.x * (100.f / 255.f);
+            buf.y = src.y * 1.388235294117647f - 134.f;
+            buf.z = src.z * 1.003921568627451f - 140.f;
+
+            Luv2RGB<float, 3, 3, srgb, blueIdx> cvtf;
+            buf = cvtf(buf);
+
+            typename MakeVec<uchar, dcn>::type dst;
+
+            dst.x = saturate_cast<uchar>(buf.x * 255.f);
+            dst.y = saturate_cast<uchar>(buf.y * 255.f);
+            dst.z = saturate_cast<uchar>(buf.z * 255.f);
+            setAlpha(dst, ColorChannel<uchar>::max());
+
+            return dst;
+        }
+    };
+
+    #undef CV_CUDEV_DESCALE
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/functional/functional.hpp b/modules/cudev/include/opencv2/cudev/functional/functional.hpp
new file mode 100644
index 0000000000..3ac532819a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/functional/functional.hpp
@@ -0,0 +1,854 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_FUNCTIONAL_FUNCTIONAL_HPP__
+#define __OPENCV_CUDEV_FUNCTIONAL_FUNCTIONAL_HPP__
+
+#include "../common.hpp"
+#include "../util/saturate_cast.hpp"
+#include "../util/vec_traits.hpp"
+#include "../util/vec_math.hpp"
+#include "../util/type_traits.hpp"
+
+namespace cv { namespace cudev {
+
+// Function Objects
+
+template <typename _Arg, typename _Result> struct unary_function
+{
+    typedef _Arg    argument_type;
+    typedef _Result result_type;
+};
+
+template <typename _Arg1, typename _Arg2, typename _Result> struct binary_function
+{
+    typedef _Arg1   first_argument_type;
+    typedef _Arg2   second_argument_type;
+    typedef _Result result_type;
+};
+
+// Arithmetic Operations
+
+template <typename T> struct plus : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return saturate_cast<T>(a + b);
+    }
+};
+
+template <typename T> struct minus : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return saturate_cast<T>(a - b);
+    }
+};
+
+template <typename T> struct multiplies : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return saturate_cast<T>(a * b);
+    }
+};
+
+template <typename T> struct divides : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return saturate_cast<T>(a / b);
+    }
+};
+
+template <typename T> struct modulus : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return saturate_cast<T>(a % b);
+    }
+};
+
+template <typename T> struct negate : unary_function<T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a) const
+    {
+        return saturate_cast<T>(-a);
+    }
+};
+
+// Comparison Operations
+
+template <typename T> struct equal_to : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a == b;
+    }
+};
+
+template <typename T> struct not_equal_to : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a != b;
+    }
+};
+
+template <typename T> struct greater : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a > b;
+    }
+};
+
+template <typename T> struct less : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a < b;
+    }
+};
+
+template <typename T> struct greater_equal : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a >= b;
+    }
+};
+
+template <typename T> struct less_equal : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a <= b;
+    }
+};
+
+// Logical Operations
+
+template <typename T> struct logical_and : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a && b;
+    }
+};
+
+template <typename T> struct logical_or : binary_function<T, T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a,
+                                                typename TypeTraits<T>::parameter_type b) const
+    {
+        return a || b;
+    }
+};
+
+template <typename T> struct logical_not : unary_function<T, typename MakeVec<uchar, VecTraits<T>::cn>::type>
+{
+    __device__ __forceinline__ typename MakeVec<uchar, VecTraits<T>::cn>::type
+                                    operator ()(typename TypeTraits<T>::parameter_type a) const
+    {
+        return !a;
+    }
+};
+
+// Bitwise Operations
+
+template <typename T> struct bit_and : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return a & b;
+    }
+};
+
+template <typename T> struct bit_or : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return a | b;
+    }
+};
+
+template <typename T> struct bit_xor : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return a ^ b;
+    }
+};
+
+template <typename T> struct bit_not : unary_function<T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type v) const
+    {
+        return ~v;
+    }
+};
+
+template <typename T> struct bit_lshift : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return a << b;
+    }
+};
+
+template <typename T> struct bit_rshift : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return a >> b;
+    }
+};
+
+// Generalized Identity Operations
+
+template <typename T> struct identity : unary_function<T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type x) const
+    {
+        return x;
+    }
+};
+
+template <typename T1, typename T2> struct project1st : binary_function<T1, T2, T1>
+{
+    __device__ __forceinline__ T1
+                    operator ()(typename TypeTraits<T1>::parameter_type lhs,
+                                typename TypeTraits<T2>::parameter_type) const
+    {
+        return lhs;
+    }
+};
+
+template <typename T1, typename T2> struct project2nd : binary_function<T1, T2, T2>
+{
+    __device__ __forceinline__ T2
+                    operator ()(typename TypeTraits<T1>::parameter_type,
+                                typename TypeTraits<T2>::parameter_type rhs) const
+    {
+        return rhs;
+    }
+};
+
+// Min/Max Operations
+
+template <typename T> struct maximum : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return max(a, b);
+    }
+};
+
+template <typename T> struct minimum : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a,
+                                             typename TypeTraits<T>::parameter_type b) const
+    {
+        return min(a, b);
+    }
+};
+
+#define CV_CUDEV_MINMAX_INST(type, maxop, minop) \
+    template <> struct maximum<type> : binary_function<type, type, type> \
+    { \
+        __device__ __forceinline__ type operator ()(type a, type b) const {return maxop(a, b);} \
+    }; \
+    template <> struct minimum<type> : binary_function<type, type, type> \
+    { \
+        __device__ __forceinline__ type operator ()(type a, type b) const {return minop(a, b);} \
+    };
+
+
+CV_CUDEV_MINMAX_INST(uchar, ::max, ::min)
+CV_CUDEV_MINMAX_INST(schar, ::max, ::min)
+CV_CUDEV_MINMAX_INST(ushort, ::max, ::min)
+CV_CUDEV_MINMAX_INST(short, ::max, ::min)
+CV_CUDEV_MINMAX_INST(int, ::max, ::min)
+CV_CUDEV_MINMAX_INST(uint, ::max, ::min)
+CV_CUDEV_MINMAX_INST(float, ::fmaxf, ::fminf)
+CV_CUDEV_MINMAX_INST(double, ::fmax, ::fmin)
+
+#undef CV_CUDEV_MINMAX_INST
+
+// abs_func
+
+template <typename T> struct abs_func : unary_function<T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type x) const
+    {
+        return abs(x);
+    }
+};
+
+template <> struct abs_func<uchar> : unary_function<uchar, uchar>
+{
+    __device__ __forceinline__ uchar operator ()(uchar x) const
+    {
+        return x;
+    }
+};
+
+template <> struct abs_func<schar> : unary_function<schar, schar>
+{
+    __device__ __forceinline__ schar operator ()(schar x) const
+    {
+        return ::abs((int) x);
+    }
+};
+
+template <> struct abs_func<ushort> : unary_function<ushort, ushort>
+{
+    __device__ __forceinline__ ushort operator ()(ushort x) const
+    {
+        return x;
+    }
+};
+
+template <> struct abs_func<short> : unary_function<short, short>
+{
+    __device__ __forceinline__ short operator ()(short x) const
+    {
+        return ::abs((int) x);
+    }
+};
+
+template <> struct abs_func<uint> : unary_function<uint, uint>
+{
+    __device__ __forceinline__ uint operator ()(uint x) const
+    {
+        return x;
+    }
+};
+
+template <> struct abs_func<int> : unary_function<int, int>
+{
+    __device__ __forceinline__ int operator ()(int x) const
+    {
+        return ::abs(x);
+    }
+};
+
+template <> struct abs_func<float> : unary_function<float, float>
+{
+    __device__ __forceinline__ float operator ()(float x) const
+    {
+        return ::fabsf(x);
+    }
+};
+
+template <> struct abs_func<double> : unary_function<double, double>
+{
+    __device__ __forceinline__ double operator ()(double x) const
+    {
+        return ::fabs(x);
+    }
+};
+
+// absdiff_func
+
+template <typename T> struct absdiff_func : binary_function<T, T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type a, typename TypeTraits<T>::parameter_type b) const
+    {
+        abs_func<T> f;
+        return f(a - b);
+    }
+};
+
+// Math functions
+
+template <typename T> struct sqr_func : unary_function<T, T>
+{
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type x) const
+    {
+        return x * x;
+    }
+};
+
+namespace functional_detail
+{
+    template <typename T> struct FloatType
+    {
+        typedef typename MakeVec<
+            typename LargerType<float, typename VecTraits<T>::elem_type>::type,
+            VecTraits<T>::cn
+        >::type type;
+    };
+}
+
+#define CV_CUDEV_UNARY_FUNCTION_INST(name, func) \
+    template <typename T> struct name ## _func : unary_function<T, typename functional_detail::FloatType<T>::type> \
+    { \
+        __device__ __forceinline__ typename functional_detail::FloatType<T>::type operator ()(typename TypeTraits<T>::parameter_type a) const \
+        { \
+            return name(a); \
+        } \
+    }; \
+    template <> struct name ## _func<uchar> : unary_function<uchar, float> \
+    { \
+        __device__ __forceinline__ float operator ()(uchar a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<schar> : unary_function<schar, float> \
+    { \
+        __device__ __forceinline__ float operator ()(schar a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<ushort> : unary_function<ushort, float> \
+    { \
+        __device__ __forceinline__ float operator ()(ushort a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<short> : unary_function<short, float> \
+    { \
+        __device__ __forceinline__ float operator ()(short a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<uint> : unary_function<uint, float> \
+    { \
+        __device__ __forceinline__ float operator ()(uint a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<int> : unary_function<int, float> \
+    { \
+        __device__ __forceinline__ float operator ()(int a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<float> : unary_function<float, float> \
+    { \
+        __device__ __forceinline__ float operator ()(float a) const \
+        { \
+            return func ## f(a); \
+        } \
+    }; \
+    template <> struct name ## _func<double> : unary_function<double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double a) const \
+        { \
+            return func(a); \
+        } \
+    };
+
+CV_CUDEV_UNARY_FUNCTION_INST(sqrt, ::sqrt)
+CV_CUDEV_UNARY_FUNCTION_INST(exp, ::exp)
+CV_CUDEV_UNARY_FUNCTION_INST(exp2, ::exp2)
+CV_CUDEV_UNARY_FUNCTION_INST(exp10, ::exp10)
+CV_CUDEV_UNARY_FUNCTION_INST(log, ::log)
+CV_CUDEV_UNARY_FUNCTION_INST(log2, ::log2)
+CV_CUDEV_UNARY_FUNCTION_INST(log10, ::log10)
+CV_CUDEV_UNARY_FUNCTION_INST(sin, ::sin)
+CV_CUDEV_UNARY_FUNCTION_INST(cos, ::cos)
+CV_CUDEV_UNARY_FUNCTION_INST(tan, ::tan)
+CV_CUDEV_UNARY_FUNCTION_INST(asin, ::asin)
+CV_CUDEV_UNARY_FUNCTION_INST(acos, ::acos)
+CV_CUDEV_UNARY_FUNCTION_INST(atan, ::atan)
+CV_CUDEV_UNARY_FUNCTION_INST(sinh, ::sinh)
+CV_CUDEV_UNARY_FUNCTION_INST(cosh, ::cosh)
+CV_CUDEV_UNARY_FUNCTION_INST(tanh, ::tanh)
+CV_CUDEV_UNARY_FUNCTION_INST(asinh, ::asinh)
+CV_CUDEV_UNARY_FUNCTION_INST(acosh, ::acosh)
+CV_CUDEV_UNARY_FUNCTION_INST(atanh, ::atanh)
+
+#undef CV_CUDEV_UNARY_FUNCTION_INST
+
+#define CV_CUDEV_BINARY_FUNCTION_INST(name, func) \
+    template <typename T> struct name ## _func : binary_function<T, T, typename functional_detail::FloatType<T>::type> \
+    { \
+        __device__ __forceinline__ typename functional_detail::FloatType<T>::type operator ()(typename TypeTraits<T>::parameter_type a, typename TypeTraits<T>::parameter_type b) const \
+        { \
+            return name(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<uchar> : binary_function<uchar, uchar, float> \
+    { \
+        __device__ __forceinline__ float operator ()(uchar a, uchar b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<schar> : binary_function<schar, schar, float> \
+    { \
+        __device__ __forceinline__ float operator ()(schar a, schar b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<ushort> : binary_function<ushort, ushort, float> \
+    { \
+        __device__ __forceinline__ float operator ()(ushort a, ushort b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<short> : binary_function<short, short, float> \
+    { \
+        __device__ __forceinline__ float operator ()(short a, short b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<uint> : binary_function<uint, uint, float> \
+    { \
+        __device__ __forceinline__ float operator ()(uint a, uint b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<int> : binary_function<int, int, float> \
+    { \
+        __device__ __forceinline__ float operator ()(int a, int b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<float> : binary_function<float, float, float> \
+    { \
+        __device__ __forceinline__ float operator ()(float a, float b) const \
+        { \
+            return func ## f(a, b); \
+        } \
+    }; \
+    template <> struct name ## _func<double> : binary_function<double, double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double a, double b) const \
+        { \
+            return func(a, b); \
+        } \
+    };
+
+CV_CUDEV_BINARY_FUNCTION_INST(hypot, ::hypot)
+CV_CUDEV_BINARY_FUNCTION_INST(atan2, ::atan2)
+
+#undef CV_CUDEV_BINARY_FUNCTION_INST
+
+template <typename T> struct magnitude_func : binary_function<T, T, typename functional_detail::FloatType<T>::type>
+{
+    __device__ __forceinline__ typename functional_detail::FloatType<T>::type operator ()(typename TypeTraits<T>::parameter_type a, typename TypeTraits<T>::parameter_type b) const
+    {
+        sqrt_func<typename functional_detail::FloatType<T>::type> f;
+        return f(a * a + b * b);
+    }
+};
+
+template <typename T> struct pow_func : binary_function<T, float, float>
+{
+    __device__ __forceinline__ float operator ()(T val, float power) const
+    {
+        return ::powf(val, power);
+    }
+};
+template <> struct pow_func<double> : binary_function<double, double, double>
+{
+    __device__ __forceinline__ double operator ()(double val, double power) const
+    {
+        return ::pow(val, power);
+    }
+};
+
+// Saturate Cast Functor
+
+template <typename T, typename D> struct saturate_cast_func : unary_function<T, D>
+{
+    __device__ __forceinline__ D operator ()(typename TypeTraits<T>::parameter_type v) const
+    {
+        return saturate_cast<D>(v);
+    }
+};
+
+// Threshold Functors
+
+template <typename T> struct ThreshBinaryFunc : unary_function<T, T>
+{
+    T thresh;
+    T maxVal;
+
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type src) const
+    {
+        return saturate_cast<T>(src > thresh) * maxVal;
+    }
+};
+
+template <typename T>
+__host__ __device__ ThreshBinaryFunc<T> thresh_binary_func(T thresh, T maxVal)
+{
+    ThreshBinaryFunc<T> f;
+    f.thresh = thresh;
+    f.maxVal = maxVal;
+    return f;
+}
+
+template <typename T> struct ThreshBinaryInvFunc : unary_function<T, T>
+{
+    T thresh;
+    T maxVal;
+
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type src) const
+    {
+        return saturate_cast<T>(src <= thresh) * maxVal;
+    }
+};
+
+template <typename T>
+__host__ __device__ ThreshBinaryInvFunc<T> thresh_binary_inv_func(T thresh, T maxVal)
+{
+    ThreshBinaryInvFunc<T> f;
+    f.thresh = thresh;
+    f.maxVal = maxVal;
+    return f;
+}
+
+template <typename T> struct ThreshTruncFunc : unary_function<T, T>
+{
+    T thresh;
+
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type src) const
+    {
+        minimum<T> minOp;
+        return minOp(src, thresh);
+    }
+};
+
+template <typename T>
+__host__ __device__ ThreshTruncFunc<T> thresh_trunc_func(T thresh)
+{
+    ThreshTruncFunc<T> f;
+    f.thresh = thresh;
+    return f;
+}
+
+template <typename T> struct ThreshToZeroFunc : unary_function<T, T>
+{
+    T thresh;
+
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type src) const
+    {
+        return saturate_cast<T>(src > thresh) * src;
+    }
+};
+
+template <typename T>
+__host__ __device__ ThreshToZeroFunc<T> thresh_to_zero_func(T thresh)
+{
+    ThreshToZeroFunc<T> f;
+    f.thresh = thresh;
+    return f;
+}
+
+template <typename T> struct ThreshToZeroInvFunc : unary_function<T, T>
+{
+    T thresh;
+
+    __device__ __forceinline__ T operator ()(typename TypeTraits<T>::parameter_type src) const
+    {
+        return saturate_cast<T>(src <= thresh) * src;
+    }
+};
+
+template <typename T>
+__host__ __device__ ThreshToZeroInvFunc<T> thresh_to_zero_inv_func(T thresh)
+{
+    ThreshToZeroInvFunc<T> f;
+    f.thresh = thresh;
+    return f;
+}
+
+// Function Object Adaptors
+
+template <class Predicate> struct UnaryNegate : unary_function<typename Predicate::argument_type, typename Predicate::result_type>
+{
+    Predicate pred;
+
+    __device__ __forceinline__ typename Predicate::result_type operator ()(
+            typename TypeTraits<typename Predicate::argument_type>::parameter_type x) const
+    {
+        return !pred(x);
+    }
+};
+
+template <class Predicate>
+__host__ __device__ UnaryNegate<Predicate> not1(const Predicate& pred)
+{
+    UnaryNegate<Predicate> n;
+    n.pred = pred;
+    return n;
+}
+
+template <class Predicate> struct BinaryNegate : binary_function<typename Predicate::first_argument_type, typename Predicate::second_argument_type, typename Predicate::result_type>
+{
+    Predicate pred;
+
+    __device__ __forceinline__ typename Predicate::result_type operator ()(
+            typename TypeTraits<typename Predicate::first_argument_type>::parameter_type x,
+            typename TypeTraits<typename Predicate::second_argument_type>::parameter_type y) const
+    {
+        return !pred(x, y);
+    }
+};
+
+template <class Predicate>
+__host__ __device__ BinaryNegate<Predicate> not2(const Predicate& pred)
+{
+    BinaryNegate<Predicate> n;
+    n.pred = pred;
+    return n;
+}
+
+template <class Op> struct Binder1st : unary_function<typename Op::second_argument_type, typename Op::result_type>
+{
+    Op op;
+    typename Op::first_argument_type arg1;
+
+    __device__ __forceinline__ typename Op::result_type operator ()(
+            typename TypeTraits<typename Op::second_argument_type>::parameter_type a) const
+    {
+        return op(arg1, a);
+    }
+};
+
+template <class Op>
+__host__ __device__ Binder1st<Op> bind1st(const Op& op, const typename Op::first_argument_type& arg1)
+{
+    Binder1st<Op> b;
+    b.op = op;
+    b.arg1 = arg1;
+    return b;
+}
+
+template <class Op> struct Binder2nd : unary_function<typename Op::first_argument_type, typename Op::result_type>
+{
+    Op op;
+    typename Op::second_argument_type arg2;
+
+    __device__ __forceinline__ typename Op::result_type operator ()(
+            typename TypeTraits<typename Op::first_argument_type>::parameter_type a) const
+    {
+        return op(a, arg2);
+    }
+};
+
+template <class Op>
+__host__ __device__ Binder2nd<Op> bind2nd(const Op& op, const typename Op::second_argument_type& arg2)
+{
+    Binder2nd<Op> b;
+    b.op = op;
+    b.arg2 = arg2;
+    return b;
+}
+
+// Functor Traits
+
+template <typename F> struct IsUnaryFunction
+{
+    typedef char Yes;
+    struct No {Yes a[2];};
+
+    template <typename T, typename D> static Yes check(unary_function<T, D>);
+    static No check(...);
+
+    static F makeF();
+
+    enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+};
+
+template <typename F> struct IsBinaryFunction
+{
+    typedef char Yes;
+    struct No {Yes a[2];};
+
+    template <typename T1, typename T2, typename D> static Yes check(binary_function<T1, T2, D>);
+    static No check(...);
+
+    static F makeF();
+
+    enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/functional/tuple_adapter.hpp b/modules/cudev/include/opencv2/cudev/functional/tuple_adapter.hpp
new file mode 100644
index 0000000000..d3a40db0ea
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/functional/tuple_adapter.hpp
@@ -0,0 +1,98 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_FUNCTIONAL_TUPLE_ADAPTER_HPP__
+#define __OPENCV_CUDEV_FUNCTIONAL_TUPLE_ADAPTER_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Op, int n> struct UnaryTupleAdapter
+{
+    typedef typename Op::result_type result_type;
+
+    Op op;
+
+    template <class Tuple>
+    __device__ __forceinline__ typename Op::result_type operator ()(const Tuple& t) const
+    {
+        return op(get<n>(t));
+    }
+};
+
+template <int n, class Op>
+__host__ __device__ UnaryTupleAdapter<Op, n> unaryTupleAdapter(const Op& op)
+{
+    UnaryTupleAdapter<Op, n> a;
+    a.op = op;
+    return a;
+}
+
+template <class Op, int n0, int n1> struct BinaryTupleAdapter
+{
+    typedef typename Op::result_type result_type;
+
+    Op op;
+
+    template <class Tuple>
+    __device__ __forceinline__ typename Op::result_type operator ()(const Tuple& t) const
+    {
+        return op(get<n0>(t), get<n1>(t));
+    }
+};
+
+template <int n0, int n1, class Op>
+__host__ __device__ BinaryTupleAdapter<Op, n0, n1> binaryTupleAdapter(const Op& op)
+{
+    BinaryTupleAdapter<Op, n0, n1> a;
+    a.op = op;
+    return a;
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/copy.hpp b/modules/cudev/include/opencv2/cudev/grid/copy.hpp
new file mode 100644
index 0000000000..d7d3ea8343
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/copy.hpp
@@ -0,0 +1,452 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_COPY_HPP__
+#define __OPENCV_CUDEV_GRID_COPY_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/glob.hpp"
+#include "../ptr2d/mask.hpp"
+#include "../ptr2d/zip.hpp"
+#include "detail/copy.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridCopy_(const SrcPtr& src, GpuMat_<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(rows, cols);
+
+    grid_copy_detail::copy<Policy>(shrinkPtr(src), shrinkPtr(dst), shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridCopy_(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_copy_detail::copy<Policy>(shrinkPtr(src), shrinkPtr(dst), shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridCopy_(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(rows, cols);
+
+    grid_copy_detail::copy<Policy>(shrinkPtr(src), shrinkPtr(dst), WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridCopy_(const SrcPtr& src, const GlobPtrSz<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+
+    grid_copy_detail::copy<Policy>(shrinkPtr(src), shrinkPtr(dst), WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3, class MaskPtr>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(get<3>(dst)) == rows && getCols(get<3>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<SrcPtrTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(get<3>(dst)) == rows && getCols(get<3>(dst)) == cols );
+
+    grid_copy_detail::copy_tuple<Policy>(shrinkPtr(src),
+                                         shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+// Default Policy
+
+struct DefaultCopyPolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8
+    };
+};
+
+template <class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridCopy(const SrcPtr& src, GpuMat_<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridCopy(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridCopy(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridCopy(const SrcPtr& src, const GlobPtrSz<DstType>& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3, class MaskPtr>
+__host__ void gridCopy(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+template <class SrcPtrTuple, typename D0, typename D1, typename D2, typename D3>
+__host__ void gridCopy_(const SrcPtrTuple& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, Stream& stream = Stream::Null())
+{
+    gridCopy_<DefaultCopyPolicy>(src, dst, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/copy.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/copy.hpp
new file mode 100644
index 0000000000..707b842f9f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/copy.hpp
@@ -0,0 +1,132 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_COPY_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_COPY_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_copy_detail
+{
+    template <class SrcPtr, typename DstType, class MaskPtr>
+    __global__ void copy(const SrcPtr src, GlobPtr<DstType> dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = saturate_cast<DstType>(src(y, x));
+    }
+
+    template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+    __host__ void copy(const SrcPtr& src, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        copy<<<grid, block, 0, stream>>>(src, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+
+    template <int count> struct Unroll
+    {
+        template <class SrcPtrTuple, class DstPtrTuple>
+        __device__ static void copy(const SrcPtrTuple& src, DstPtrTuple& dst, const int y, const int x)
+        {
+            typedef typename tuple_element<count - 1, DstPtrTuple>::type dst_ptr_type;
+            typedef typename PtrTraits<dst_ptr_type>::value_type dst_type;
+
+            get<count - 1>(dst)(y, x) = saturate_cast<dst_type>(get<count - 1>(src)(y, x));
+            Unroll<count - 1>::copy(src, dst, y, x);
+        }
+    };
+    template <> struct Unroll<0>
+    {
+        template <class SrcPtrTuple, class DstPtrTuple>
+        __device__ __forceinline__ static void copy(const SrcPtrTuple&, DstPtrTuple&, const int, const int)
+        {
+        }
+    };
+
+    template <class SrcPtrTuple, class DstPtrTuple, class MaskPtr>
+    __global__ void copy_tuple(const SrcPtrTuple src, DstPtrTuple dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        Unroll<tuple_size<SrcPtrTuple>::value>::copy(src, dst, y, x);
+    }
+
+    template <class Policy, class SrcPtrTuple, class DstPtrTuple, class MaskPtr>
+    __host__ void copy_tuple(const SrcPtrTuple& src, const DstPtrTuple& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        copy_tuple<<<grid, block, 0, stream>>>(src, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/glob_reduce.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/glob_reduce.hpp
new file mode 100644
index 0000000000..c9d715982c
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/glob_reduce.hpp
@@ -0,0 +1,475 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_GLOB_REDUCE_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_GLOB_REDUCE_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/atomic.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../util/type_traits.hpp"
+#include "../../util/limits.hpp"
+#include "../../block/reduce.hpp"
+#include "../../functional/functional.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_glob_reduce_detail
+{
+    // Unroll
+
+    template <int cn> struct Unroll;
+
+    template <> struct Unroll<1>
+    {
+        template <int BLOCK_SIZE, typename R>
+        __device__ __forceinline__ static volatile R* smem(R* ptr)
+        {
+            return ptr;
+        }
+
+        template <typename R>
+        __device__ __forceinline__ static R& res(R& val)
+        {
+            return val;
+        }
+
+        template <class Op>
+        __device__ __forceinline__ static const Op& op(const Op& aop)
+        {
+            return aop;
+        }
+    };
+
+    template <> struct Unroll<2>
+    {
+        template <int BLOCK_SIZE, typename R>
+        __device__ __forceinline__ static tuple<volatile R*, volatile R*> smem(R* ptr)
+        {
+            return smem_tuple(ptr, ptr + BLOCK_SIZE);
+        }
+
+        template <typename R>
+        __device__ __forceinline__ static tuple<typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&> res(R& val)
+        {
+            return tie(val.x, val.y);
+        }
+
+        template <class Op>
+        __device__ __forceinline__ static tuple<Op, Op> op(const Op& aop)
+        {
+            return make_tuple(aop, aop);
+        }
+    };
+
+    template <> struct Unroll<3>
+    {
+        template <int BLOCK_SIZE, typename R>
+        __device__ __forceinline__ static tuple<volatile R*, volatile R*, volatile R*> smem(R* ptr)
+        {
+            return smem_tuple(ptr, ptr + BLOCK_SIZE, ptr + 2 * BLOCK_SIZE);
+        }
+
+        template <typename R>
+        __device__ __forceinline__ static tuple<typename VecTraits<R>::elem_type&,
+                                                typename VecTraits<R>::elem_type&,
+                                                typename VecTraits<R>::elem_type&> res(R& val)
+        {
+            return tie(val.x, val.y, val.z);
+        }
+
+        template <class Op>
+        __device__ __forceinline__ static tuple<Op, Op, Op> op(const Op& aop)
+        {
+            return make_tuple(aop, aop, aop);
+        }
+    };
+
+    template <> struct Unroll<4>
+    {
+        template <int BLOCK_SIZE, typename R>
+        __device__ __forceinline__ static tuple<volatile R*, volatile R*, volatile R*, volatile R*> smem(R* ptr)
+        {
+            return smem_tuple(ptr, ptr + BLOCK_SIZE, ptr + 2 * BLOCK_SIZE, ptr + 3 * BLOCK_SIZE);
+        }
+
+        template <typename R>
+        __device__ __forceinline__ static tuple<typename VecTraits<R>::elem_type&,
+                                                typename VecTraits<R>::elem_type&,
+                                                typename VecTraits<R>::elem_type&,
+                                                typename VecTraits<R>::elem_type&> res(R& val)
+        {
+            return tie(val.x, val.y, val.z, val.w);
+        }
+
+        template <class Op>
+        __device__ __forceinline__ static tuple<Op, Op, Op, Op> op(const Op& aop)
+        {
+            return make_tuple(aop, aop, aop, aop);
+        }
+    };
+
+    // AtomicUnroll
+
+    template <typename R, int cn> struct AtomicUnroll;
+
+    template <typename R> struct AtomicUnroll<R, 1>
+    {
+        __device__ __forceinline__ static void add(R* ptr, R val)
+        {
+            atomicAdd(ptr, val);
+        }
+
+        __device__ __forceinline__ static void min(R* ptr, R val)
+        {
+            atomicMin(ptr, val);
+        }
+
+        __device__ __forceinline__ static void max(R* ptr, R val)
+        {
+            atomicMax(ptr, val);
+        }
+    };
+
+    template <typename R> struct AtomicUnroll<R, 2>
+    {
+        typedef typename MakeVec<R, 2>::type val_type;
+
+        __device__ __forceinline__ static void add(R* ptr, val_type val)
+        {
+            atomicAdd(ptr, val.x);
+            atomicAdd(ptr + 1, val.y);
+        }
+
+        __device__ __forceinline__ static void min(R* ptr, val_type val)
+        {
+            atomicMin(ptr, val.x);
+            atomicMin(ptr + 1, val.y);
+        }
+
+        __device__ __forceinline__ static void max(R* ptr, val_type val)
+        {
+            atomicMax(ptr, val.x);
+            atomicMax(ptr + 1, val.y);
+        }
+    };
+
+    template <typename R> struct AtomicUnroll<R, 3>
+    {
+        typedef typename MakeVec<R, 3>::type val_type;
+
+        __device__ __forceinline__ static void add(R* ptr, val_type val)
+        {
+            atomicAdd(ptr, val.x);
+            atomicAdd(ptr + 1, val.y);
+            atomicAdd(ptr + 2, val.z);
+        }
+
+        __device__ __forceinline__ static void min(R* ptr, val_type val)
+        {
+            atomicMin(ptr, val.x);
+            atomicMin(ptr + 1, val.y);
+            atomicMin(ptr + 2, val.z);
+        }
+
+        __device__ __forceinline__ static void max(R* ptr, val_type val)
+        {
+            atomicMax(ptr, val.x);
+            atomicMax(ptr + 1, val.y);
+            atomicMax(ptr + 2, val.z);
+        }
+    };
+
+    template <typename R> struct AtomicUnroll<R, 4>
+    {
+        typedef typename MakeVec<R, 4>::type val_type;
+
+        __device__ __forceinline__ static void add(R* ptr, val_type val)
+        {
+            atomicAdd(ptr, val.x);
+            atomicAdd(ptr + 1, val.y);
+            atomicAdd(ptr + 2, val.z);
+            atomicAdd(ptr + 3, val.w);
+        }
+
+        __device__ __forceinline__ static void min(R* ptr, val_type val)
+        {
+            atomicMin(ptr, val.x);
+            atomicMin(ptr + 1, val.y);
+            atomicMin(ptr + 2, val.z);
+            atomicMin(ptr + 3, val.w);
+        }
+
+        __device__ __forceinline__ static void max(R* ptr, val_type val)
+        {
+            atomicMax(ptr, val.x);
+            atomicMax(ptr + 1, val.y);
+            atomicMax(ptr + 2, val.z);
+            atomicMax(ptr + 3, val.w);
+        }
+    };
+
+    // SumReductor
+
+    template <typename src_type, typename work_type> struct SumReductor
+    {
+        typedef typename VecTraits<work_type>::elem_type work_elem_type;
+        enum { cn = VecTraits<src_type>::cn };
+
+        work_type sum;
+
+        __device__ __forceinline__ SumReductor()
+        {
+            sum = VecTraits<work_type>::all(0);
+        }
+
+        __device__ __forceinline__ void reduceVal(typename TypeTraits<src_type>::parameter_type srcVal)
+        {
+            sum = sum + saturate_cast<work_type>(srcVal);
+        }
+
+        template <int BLOCK_SIZE>
+        __device__ void reduceGrid(work_elem_type* result, int tid)
+        {
+            __shared__ work_elem_type smem[BLOCK_SIZE * cn];
+
+            blockReduce<BLOCK_SIZE>(Unroll<cn>::template smem<BLOCK_SIZE>(smem), Unroll<cn>::res(sum), tid, Unroll<cn>::op(plus<work_elem_type>()));
+
+            if (tid == 0)
+                AtomicUnroll<work_elem_type, cn>::add(result, sum);
+        }
+    };
+
+    // MinMaxReductor
+
+    template <typename T> struct minop : minimum<T>
+    {
+        __device__ __forceinline__ static T initial()
+        {
+            return numeric_limits<T>::max();
+        }
+
+        __device__ __forceinline__ static void atomic(T* result, T myval)
+        {
+            atomicMin(result, myval);
+        }
+    };
+
+    template <typename T> struct maxop : maximum<T>
+    {
+        __device__ __forceinline__ static T initial()
+        {
+            return -numeric_limits<T>::max();
+        }
+
+        __device__ __forceinline__ static void atomic(T* result, T myval)
+        {
+            atomicMax(result, myval);
+        }
+    };
+
+    struct both
+    {
+    };
+
+    template <class Op, typename src_type, typename work_type> struct MinMaxReductor
+    {
+        work_type myval;
+
+        __device__ __forceinline__ MinMaxReductor()
+        {
+            myval = Op::initial();
+        }
+
+        __device__ __forceinline__ void reduceVal(typename TypeTraits<src_type>::parameter_type srcVal)
+        {
+            Op op;
+
+            myval = op(myval, srcVal);
+        }
+
+        template <int BLOCK_SIZE>
+        __device__ void reduceGrid(work_type* result, int tid)
+        {
+            __shared__ work_type smem[BLOCK_SIZE];
+
+            Op op;
+
+            blockReduce<BLOCK_SIZE>(smem, myval, tid, op);
+
+            if (tid == 0)
+                Op::atomic(result, myval);
+        }
+    };
+
+    template <typename src_type, typename work_type> struct MinMaxReductor<both, src_type, work_type>
+    {
+        work_type mymin;
+        work_type mymax;
+
+        __device__ __forceinline__ MinMaxReductor()
+        {
+            mymin = numeric_limits<work_type>::max();
+            mymax = -numeric_limits<work_type>::max();
+        }
+
+        __device__ __forceinline__ void reduceVal(typename TypeTraits<src_type>::parameter_type srcVal)
+        {
+            minimum<work_type> minOp;
+            maximum<work_type> maxOp;
+
+            mymin = minOp(mymin, srcVal);
+            mymax = maxOp(mymax, srcVal);
+        }
+
+        template <int BLOCK_SIZE>
+        __device__ void reduceGrid(work_type* result, int tid)
+        {
+            __shared__ work_type sminval[BLOCK_SIZE];
+            __shared__ work_type smaxval[BLOCK_SIZE];
+
+            minimum<work_type> minOp;
+            maximum<work_type> maxOp;
+
+            blockReduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), tie(mymin, mymax), tid, make_tuple(minOp, maxOp));
+
+            if (tid == 0)
+            {
+                atomicMin(result, mymin);
+                atomicMax(result + 1, mymax);
+            }
+        }
+    };
+
+    // glob_reduce
+
+    template <class Reductor, int BLOCK_SIZE, int PATCH_X, int PATCH_Y, class SrcPtr, typename ResType, class MaskPtr>
+    __global__ void glob_reduce(const SrcPtr src, ResType* result, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x0 = blockIdx.x * blockDim.x * PATCH_X + threadIdx.x;
+        const int y0 = blockIdx.y * blockDim.y * PATCH_Y + threadIdx.y;
+
+        Reductor reductor;
+
+        for (int i = 0, y = y0; i < PATCH_Y && y < rows; ++i, y += blockDim.y)
+        {
+            for (int j = 0, x = x0; j < PATCH_X && x < cols; ++j, x += blockDim.x)
+            {
+                if (mask(y, x))
+                {
+                    reductor.reduceVal(src(y, x));
+                }
+            }
+        }
+
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+
+        reductor.template reduceGrid<BLOCK_SIZE>(result, tid);
+    }
+
+    template <class Reductor, class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void glob_reduce(const SrcPtr& src, ResType* result, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x * Policy::patch_size_x), divUp(rows, block.y * Policy::patch_size_y));
+
+        const int BLOCK_SIZE = Policy::block_size_x * Policy::block_size_y;
+
+        glob_reduce<Reductor, BLOCK_SIZE, Policy::patch_size_x, Policy::patch_size_y><<<grid, block, 0, stream>>>(src, result, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+
+    // callers
+
+    template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void sum(const SrcPtr& src, ResType* result, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        const int cn = VecTraits<src_type>::cn;
+        typedef typename MakeVec<ResType, cn>::type work_type;
+
+        glob_reduce<SumReductor<src_type, work_type>, Policy>(src, result, mask, rows, cols, stream);
+    }
+
+    template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void minVal(const SrcPtr& src, ResType* result, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        const int cn = VecTraits<src_type>::cn;
+        typedef typename MakeVec<ResType, cn>::type work_type;
+
+        glob_reduce<MinMaxReductor<minop<work_type>, src_type, work_type>, Policy>(src, result, mask, rows, cols, stream);
+    }
+
+    template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void maxVal(const SrcPtr& src, ResType* result, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        const int cn = VecTraits<src_type>::cn;
+        typedef typename MakeVec<ResType, cn>::type work_type;
+
+        glob_reduce<MinMaxReductor<maxop<work_type>, src_type, work_type>, Policy>(src, result, mask, rows, cols, stream);
+    }
+
+    template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void minMaxVal(const SrcPtr& src, ResType* result, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        const int cn = VecTraits<src_type>::cn;
+        typedef typename MakeVec<ResType, cn>::type work_type;
+
+        glob_reduce<MinMaxReductor<both, src_type, work_type>, Policy>(src, result, mask, rows, cols, stream);
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/gpu/src/video_parser.h b/modules/cudev/include/opencv2/cudev/grid/detail/histogram.hpp
similarity index 52%
rename from modules/gpu/src/video_parser.h
rename to modules/cudev/include/opencv2/cudev/grid/detail/histogram.hpp
index 44a53710bf..4c2675098d 100644
--- a/modules/gpu/src/video_parser.h
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/histogram.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,61 +41,69 @@
 //
 //M*/
 
-#ifndef __VIDEO_PARSER_H__
-#define __VIDEO_PARSER_H__
+#pragma once
 
-#include "precomp.hpp"
+#ifndef __OPENCV_CUDEV_GRID_HISTOGRAM_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_HISTOGRAM_DETAIL_HPP__
 
-#include "frame_queue.h"
-#include "video_decoder.h"
+#include "../../common.hpp"
+#include "../../util/atomic.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+namespace cv { namespace cudev {
 
-namespace cv { namespace gpu
+namespace grid_histogram_detail
 {
-    namespace detail
+    template <int BIN_COUNT, int BLOCK_SIZE, class SrcPtr, typename ResType, class MaskPtr>
+    __global__ void histogram(const SrcPtr src, ResType* hist, const MaskPtr mask, const int rows, const int cols)
     {
-        class VideoParser
+        __shared__ ResType smem[BIN_COUNT];
+
+        const int y = blockIdx.x * blockDim.y + threadIdx.y;
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+
+        for (int i = tid; i < BIN_COUNT; i += BLOCK_SIZE)
+            smem[i] = 0;
+
+        __syncthreads();
+
+        if (y < rows)
         {
-        public:
-            VideoParser(VideoDecoder* videoDecoder, FrameQueue* frameQueue);
-
-            ~VideoParser()
+            for (int x = threadIdx.x; x < cols; x += blockDim.x)
             {
-                cuvidDestroyVideoParser(parser_);
+                if (mask(y, x))
+                {
+                    const uint data = src(y, x);
+                    atomicAdd(&smem[data % BIN_COUNT], 1);
+                }
             }
+        }
 
-            bool parseVideoData(const unsigned char* data, size_t size, bool endOfStream);
+        __syncthreads();
 
-            bool hasError() const { return hasError_; }
-
-        private:
-            VideoParser(const VideoParser&);
-            VideoParser& operator =(const VideoParser&);
-
-            VideoDecoder* videoDecoder_;
-            FrameQueue* frameQueue_;
-            CUvideoparser parser_;
-            int unparsedPackets_;
-            volatile bool hasError_;
-
-            // Called when the decoder encounters a video format change (or initial sequence header)
-            // This particular implementation of the callback returns 0 in case the video format changes
-            // to something different than the original format. Returning 0 causes a stop of the app.
-            static int CUDAAPI HandleVideoSequence(void* pUserData, CUVIDEOFORMAT* pFormat);
-
-            // Called by the video parser to decode a single picture
-            // Since the parser will deliver data as fast as it can, we need to make sure that the picture
-            // index we're attempting to use for decode is no longer used for display
-            static int CUDAAPI HandlePictureDecode(void* pUserData, CUVIDPICPARAMS* pPicParams);
-
-            // Called by the video parser to display a video frame (in the case of field pictures, there may be
-            // 2 decode calls per 1 display call, since two fields make up one frame)
-            static int CUDAAPI HandlePictureDisplay(void* pUserData, CUVIDPARSERDISPINFO* pPicParams);
-        };
+        for (int i = tid; i < BIN_COUNT; i += BLOCK_SIZE)
+        {
+            const ResType histVal = smem[i];
+            if (histVal > 0)
+                atomicAdd(hist + i, histVal);
+        }
     }
+
+    template <int BIN_COUNT, class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void histogram(const SrcPtr& src, ResType* hist, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(rows, block.y));
+
+        const int BLOCK_SIZE = Policy::block_size_x * Policy::block_size_y;
+
+        histogram<BIN_COUNT, BLOCK_SIZE><<<grid, block, 0, stream>>>(src, hist, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
 }}
 
-#endif // HAVE_CUDA
-
-#endif // __VIDEO_PARSER_H__
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/integral.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/integral.hpp
new file mode 100644
index 0000000000..b06d7ddbb6
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/integral.hpp
@@ -0,0 +1,626 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_INTEGRAL_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_INTEGRAL_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../warp/shuffle.hpp"
+#include "../../block/scan.hpp"
+#include "../../ptr2d/glob.hpp"
+
+namespace cv { namespace cudev {
+
+namespace integral_detail
+{
+    // horizontal_pass
+
+    template <int NUM_SCAN_THREADS, class SrcPtr, typename D>
+    __global__ void horizontal_pass(const SrcPtr src, GlobPtr<D> dst, const int cols)
+    {
+        __shared__ D smem[NUM_SCAN_THREADS * 2];
+        __shared__ D carryElem;
+
+        carryElem = 0;
+
+        __syncthreads();
+
+        D* dst_row = dst.row(blockIdx.x);
+
+        int numBuckets = divUp(cols, NUM_SCAN_THREADS);
+        int offsetX = 0;
+
+        while (numBuckets--)
+        {
+            const int curElemOffs = offsetX + threadIdx.x;
+
+            D curElem = 0.0f;
+
+            if (curElemOffs < cols)
+                curElem = src(blockIdx.x, curElemOffs);
+
+            const D curScanElem = blockScanInclusive<NUM_SCAN_THREADS>(curElem, smem, threadIdx.x);
+
+            if (curElemOffs < cols)
+                dst_row[curElemOffs] = carryElem + curScanElem;
+
+            offsetX += NUM_SCAN_THREADS;
+
+            __syncthreads();
+
+            if (threadIdx.x == NUM_SCAN_THREADS - 1)
+            {
+                carryElem += curScanElem;
+            }
+
+            __syncthreads();
+        }
+    }
+
+    template <int NUM_SCAN_THREADS, typename T, typename D>
+    __global__ void horizontal_pass(const GlobPtr<T> src, GlobPtr<D> dst, const int cols)
+    {
+        __shared__ D smem[NUM_SCAN_THREADS * 2];
+        __shared__ D carryElem;
+
+        carryElem = 0;
+
+        __syncthreads();
+
+        const T* src_row = src.row(blockIdx.x);
+        D* dst_row = dst.row(blockIdx.x);
+
+        int numBuckets = divUp(cols, NUM_SCAN_THREADS);
+        int offsetX = 0;
+
+        while (numBuckets--)
+        {
+            const int curElemOffs = offsetX + threadIdx.x;
+
+            D curElem = 0.0f;
+
+            if (curElemOffs < cols)
+                curElem = src_row[curElemOffs];
+
+            const D curScanElem = blockScanInclusive<NUM_SCAN_THREADS>(curElem, smem, threadIdx.x);
+
+            if (curElemOffs < cols)
+                dst_row[curElemOffs] = carryElem + curScanElem;
+
+            offsetX += NUM_SCAN_THREADS;
+
+            __syncthreads();
+
+            if (threadIdx.x == NUM_SCAN_THREADS - 1)
+            {
+                carryElem += curScanElem;
+            }
+
+            __syncthreads();
+        }
+    }
+
+    template <class SrcPtr, typename D>
+    __host__ void horizontal_pass(const SrcPtr& src, const GlobPtr<D>& dst, int rows, int cols, cudaStream_t stream)
+    {
+        const int NUM_SCAN_THREADS = 256;
+
+        const dim3 block(NUM_SCAN_THREADS);
+        const dim3 grid(rows);
+
+        horizontal_pass<NUM_SCAN_THREADS><<<grid, block, 0, stream>>>(src, dst, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+    }
+
+    // horisontal_pass_8u_shfl
+
+    __device__ static uchar4 int_to_uchar4(unsigned int in)
+    {
+        uchar4 bytes;
+        bytes.x = (in & 0x000000ff) >>  0;
+        bytes.y = (in & 0x0000ff00) >>  8;
+        bytes.z = (in & 0x00ff0000) >> 16;
+        bytes.w = (in & 0xff000000) >> 24;
+        return bytes;
+    }
+
+    __global__ static void horisontal_pass_8u_shfl_kernel(const GlobPtr<uint4> img, GlobPtr<uint4> integral)
+    {
+    #if CV_CUDEV_ARCH >= 300
+        __shared__ int sums[128];
+
+        const int id = threadIdx.x;
+        const int lane_id = id % warpSize;
+        const int warp_id = id / warpSize;
+
+        const uint4 data = img(blockIdx.x, id);
+
+        const uchar4 a = int_to_uchar4(data.x);
+        const uchar4 b = int_to_uchar4(data.y);
+        const uchar4 c = int_to_uchar4(data.z);
+        const uchar4 d = int_to_uchar4(data.w);
+
+        int result[16];
+
+        result[0]  =              a.x;
+        result[1]  = result[0]  + a.y;
+        result[2]  = result[1]  + a.z;
+        result[3]  = result[2]  + a.w;
+
+        result[4]  = result[3]  + b.x;
+        result[5]  = result[4]  + b.y;
+        result[6]  = result[5]  + b.z;
+        result[7]  = result[6]  + b.w;
+
+        result[8]  = result[7]  + c.x;
+        result[9]  = result[8]  + c.y;
+        result[10] = result[9]  + c.z;
+        result[11] = result[10] + c.w;
+
+        result[12] = result[11] + d.x;
+        result[13] = result[12] + d.y;
+        result[14] = result[13] + d.z;
+        result[15] = result[14] + d.w;
+
+        int sum = result[15];
+
+        // the prefix sum for each thread's 16 value is computed,
+        // now the final sums (result[15]) need to be shared
+        // with the other threads and add.  To do this,
+        // the shfl_up() instruction is used and a shuffle scan
+        // operation is performed to distribute the sums to the correct
+        // threads
+        #pragma unroll
+        for (int i = 1; i < 32; i *= 2)
+        {
+            const int n = shfl_up(sum, i, 32);
+
+            if (lane_id >= i)
+            {
+                #pragma unroll
+                for (int k = 0; k < 16; ++k)
+                    result[k] += n;
+
+                sum += n;
+            }
+        }
+
+        // Now the final sum for the warp must be shared
+        // between warps.  This is done by each warp
+        // having a thread store to shared memory, then
+        // having some other warp load the values and
+        // compute a prefix sum, again by using shfl_up.
+        // The results are uniformly added back to the warps.
+        // last thread in the warp holding sum of the warp
+        // places that in shared
+        if (threadIdx.x % warpSize == warpSize - 1)
+            sums[warp_id] = result[15];
+
+        __syncthreads();
+
+        if (warp_id == 0)
+        {
+            int warp_sum = sums[lane_id];
+
+            #pragma unroll
+            for (int i = 1; i <= 32; i *= 2)
+            {
+                const int n = shfl_up(warp_sum, i, 32);
+
+                if (lane_id >= i)
+                    warp_sum += n;
+            }
+
+            sums[lane_id] = warp_sum;
+        }
+
+        __syncthreads();
+
+        int blockSum = 0;
+
+        // fold in unused warp
+        if (warp_id > 0)
+        {
+            blockSum = sums[warp_id - 1];
+
+            #pragma unroll
+            for (int k = 0; k < 16; ++k)
+                result[k] += blockSum;
+        }
+
+        // assemble result
+        // Each thread has 16 values to write, which are
+        // now integer data (to avoid overflow).  Instead of
+        // each thread writing consecutive uint4s, the
+        // approach shown here experiments using
+        // the shuffle command to reformat the data
+        // inside the registers so that each thread holds
+        // consecutive data to be written so larger contiguous
+        // segments can be assembled for writing.
+
+        /*
+            For example data that needs to be written as
+
+            GMEM[16] <- x0 x1 x2 x3 y0 y1 y2 y3 z0 z1 z2 z3 w0 w1 w2 w3
+            but is stored in registers (r0..r3), in four threads (0..3) as:
+
+            threadId   0  1  2  3
+              r0      x0 y0 z0 w0
+              r1      x1 y1 z1 w1
+              r2      x2 y2 z2 w2
+              r3      x3 y3 z3 w3
+
+              after apply shfl_xor operations to move data between registers r1..r3:
+
+            threadId  00 01 10 11
+                      x0 y0 z0 w0
+             xor(01)->y1 x1 w1 z1
+             xor(10)->z2 w2 x2 y2
+             xor(11)->w3 z3 y3 x3
+
+             and now x0..x3, and z0..z3 can be written out in order by all threads.
+
+             In the current code, each register above is actually representing
+             four integers to be written as uint4's to GMEM.
+        */
+
+        result[4]  = shfl_xor(result[4] , 1, 32);
+        result[5]  = shfl_xor(result[5] , 1, 32);
+        result[6]  = shfl_xor(result[6] , 1, 32);
+        result[7]  = shfl_xor(result[7] , 1, 32);
+
+        result[8]  = shfl_xor(result[8] , 2, 32);
+        result[9]  = shfl_xor(result[9] , 2, 32);
+        result[10] = shfl_xor(result[10], 2, 32);
+        result[11] = shfl_xor(result[11], 2, 32);
+
+        result[12] = shfl_xor(result[12], 3, 32);
+        result[13] = shfl_xor(result[13], 3, 32);
+        result[14] = shfl_xor(result[14], 3, 32);
+        result[15] = shfl_xor(result[15], 3, 32);
+
+        uint4* integral_row = integral.row(blockIdx.x);
+        uint4 output;
+
+        ///////
+
+        if (threadIdx.x % 4 == 0)
+            output = make_uint4(result[0], result[1], result[2], result[3]);
+
+        if (threadIdx.x % 4 == 1)
+            output = make_uint4(result[4], result[5], result[6], result[7]);
+
+        if (threadIdx.x % 4 == 2)
+            output = make_uint4(result[8], result[9], result[10], result[11]);
+
+        if (threadIdx.x % 4 == 3)
+            output = make_uint4(result[12], result[13], result[14], result[15]);
+
+        integral_row[threadIdx.x % 4 + (threadIdx.x / 4) * 16] = output;
+
+        ///////
+
+        if (threadIdx.x % 4 == 2)
+            output = make_uint4(result[0], result[1], result[2], result[3]);
+
+        if (threadIdx.x % 4 == 3)
+            output = make_uint4(result[4], result[5], result[6], result[7]);
+
+        if (threadIdx.x % 4 == 0)
+            output = make_uint4(result[8], result[9], result[10], result[11]);
+
+        if (threadIdx.x % 4 == 1)
+            output = make_uint4(result[12], result[13], result[14], result[15]);
+
+        integral_row[(threadIdx.x + 2) % 4 + (threadIdx.x / 4) * 16 + 8] = output;
+
+        // continuning from the above example,
+        // this use of shfl_xor() places the y0..y3 and w0..w3 data
+        // in order.
+
+        #pragma unroll
+        for (int i = 0; i < 16; ++i)
+            result[i] = shfl_xor(result[i], 1, 32);
+
+        if (threadIdx.x % 4 == 0)
+            output = make_uint4(result[0], result[1], result[2], result[3]);
+
+        if (threadIdx.x % 4 == 1)
+            output = make_uint4(result[4], result[5], result[6], result[7]);
+
+        if (threadIdx.x % 4 == 2)
+            output = make_uint4(result[8], result[9], result[10], result[11]);
+
+        if (threadIdx.x % 4 == 3)
+            output = make_uint4(result[12], result[13], result[14], result[15]);
+
+        integral_row[threadIdx.x % 4 + (threadIdx.x / 4) * 16 + 4] = output;
+
+        ///////
+
+        if (threadIdx.x % 4 == 2)
+            output = make_uint4(result[0], result[1], result[2], result[3]);
+
+        if (threadIdx.x % 4 == 3)
+            output = make_uint4(result[4], result[5], result[6], result[7]);
+
+        if (threadIdx.x % 4 == 0)
+            output = make_uint4(result[8], result[9], result[10], result[11]);
+
+        if (threadIdx.x % 4 == 1)
+            output = make_uint4(result[12], result[13], result[14], result[15]);
+
+        integral_row[(threadIdx.x + 2) % 4 + (threadIdx.x / 4) * 16 + 12] = output;
+    #endif
+    }
+
+    __host__ static void horisontal_pass_8u_shfl(const GlobPtr<uchar> src, GlobPtr<uint> integral, int rows, int cols, cudaStream_t stream)
+    {
+        // each thread handles 16 values, use 1 block/row
+        // save, because step is actually can't be less 512 bytes
+        const int block = cols / 16;
+
+        // launch 1 block / row
+        const int grid = rows;
+
+        CV_CUDEV_SAFE_CALL( cudaFuncSetCacheConfig(horisontal_pass_8u_shfl_kernel, cudaFuncCachePreferL1) );
+
+        GlobPtr<uint4> src4 = globPtr((uint4*) src.data, src.step);
+        GlobPtr<uint4> integral4 = globPtr((uint4*) integral.data, integral.step);
+
+        horisontal_pass_8u_shfl_kernel<<<grid, block, 0, stream>>>(src4, integral4);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+    }
+
+    // vertical
+
+    template <typename T>
+    __global__ void vertical_pass(GlobPtr<T> integral, const int rows, const int cols)
+    {
+    #if CV_CUDEV_ARCH >= 300
+        __shared__ T sums[32][9];
+
+        const int tidx = blockIdx.x * blockDim.x + threadIdx.x;
+        const int lane_id = tidx % 8;
+
+        sums[threadIdx.x][threadIdx.y] = 0;
+        __syncthreads();
+
+        T stepSum = 0;
+
+        int numBuckets = divUp(rows, blockDim.y);
+        int y = threadIdx.y;
+
+        while (numBuckets--)
+        {
+            T* p = integral.row(y) + tidx;
+
+            T sum = (tidx < cols) && (y < rows) ? *p : 0;
+
+            y += blockDim.y;
+
+            sums[threadIdx.x][threadIdx.y] = sum;
+            __syncthreads();
+
+            // place into SMEM
+            // shfl scan reduce the SMEM, reformating so the column
+            // sums are computed in a warp
+            // then read out properly
+            const int j = threadIdx.x % 8;
+            const int k = threadIdx.x / 8 + threadIdx.y * 4;
+
+            T partial_sum = sums[k][j];
+
+            for (int i = 1; i <= 8; i *= 2)
+            {
+                T n = shfl_up(partial_sum, i, 32);
+
+                if (lane_id >= i)
+                    partial_sum += n;
+            }
+
+            sums[k][j] = partial_sum;
+            __syncthreads();
+
+            if (threadIdx.y > 0)
+                sum += sums[threadIdx.x][threadIdx.y - 1];
+
+            if (tidx < cols)
+            {
+                sum += stepSum;
+                stepSum += sums[threadIdx.x][blockDim.y - 1];
+                *p = sum;
+            }
+
+            __syncthreads();
+        }
+    #else
+        __shared__ T smem[32][32];
+        __shared__ T prevVals[32];
+
+        volatile T* smem_row = &smem[0][0] + 64 * threadIdx.y;
+
+        if (threadIdx.y == 0)
+            prevVals[threadIdx.x] = 0;
+
+        __syncthreads();
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+
+        int numBuckets = divUp(rows, 8 * 4);
+        int offsetY = 0;
+
+        while (numBuckets--)
+        {
+            const int curRowOffs = offsetY + threadIdx.y;
+
+            T curElems[4];
+            T temp[4];
+
+            // load patch
+
+            smem[threadIdx.y +  0][threadIdx.x] = 0.0f;
+            smem[threadIdx.y +  8][threadIdx.x] = 0.0f;
+            smem[threadIdx.y + 16][threadIdx.x] = 0.0f;
+            smem[threadIdx.y + 24][threadIdx.x] = 0.0f;
+
+            if (x < cols)
+            {
+                for (int i = 0; i < 4; ++i)
+                {
+                    if (curRowOffs + i * 8 < rows)
+                        smem[threadIdx.y + i * 8][threadIdx.x] = integral(curRowOffs + i * 8, x);
+                }
+            }
+
+            __syncthreads();
+
+            // reduce
+
+            curElems[0] = smem[threadIdx.x][threadIdx.y     ];
+            curElems[1] = smem[threadIdx.x][threadIdx.y +  8];
+            curElems[2] = smem[threadIdx.x][threadIdx.y + 16];
+            curElems[3] = smem[threadIdx.x][threadIdx.y + 24];
+
+            __syncthreads();
+
+            temp[0] = curElems[0] = warpScanInclusive(curElems[0], smem_row, threadIdx.x);
+            temp[1] = curElems[1] = warpScanInclusive(curElems[1], smem_row, threadIdx.x);
+            temp[2] = curElems[2] = warpScanInclusive(curElems[2], smem_row, threadIdx.x);
+            temp[3] = curElems[3] = warpScanInclusive(curElems[3], smem_row, threadIdx.x);
+
+            curElems[0] += prevVals[threadIdx.y     ];
+            curElems[1] += prevVals[threadIdx.y +  8];
+            curElems[2] += prevVals[threadIdx.y + 16];
+            curElems[3] += prevVals[threadIdx.y + 24];
+
+            __syncthreads();
+
+            if (threadIdx.x == 31)
+            {
+                prevVals[threadIdx.y     ] += temp[0];
+                prevVals[threadIdx.y +  8] += temp[1];
+                prevVals[threadIdx.y + 16] += temp[2];
+                prevVals[threadIdx.y + 24] += temp[3];
+            }
+
+            smem[threadIdx.y     ][threadIdx.x] = curElems[0];
+            smem[threadIdx.y +  8][threadIdx.x] = curElems[1];
+            smem[threadIdx.y + 16][threadIdx.x] = curElems[2];
+            smem[threadIdx.y + 24][threadIdx.x] = curElems[3];
+
+            __syncthreads();
+
+            // store patch
+
+            if (x < cols)
+            {
+                // read 4 value from source
+                for (int i = 0; i < 4; ++i)
+                {
+                    if (curRowOffs + i * 8 < rows)
+                        integral(curRowOffs + i * 8, x) = smem[threadIdx.x][threadIdx.y + i * 8];
+                }
+            }
+
+            __syncthreads();
+
+            offsetY += 8 * 4;
+        }
+    #endif
+    }
+
+    template <typename T>
+    __host__ void vertical_pass(const GlobPtr<T>& integral, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(cols, block.x));
+
+        vertical_pass<<<grid, block, 0, stream>>>(integral, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+    }
+
+    // integral
+
+    template <class SrcPtr, typename D>
+    __host__ void integral(const SrcPtr& src, const GlobPtr<D>& dst, int rows, int cols, cudaStream_t stream)
+    {
+        horizontal_pass(src, dst, rows, cols, stream);
+        vertical_pass(dst, rows, cols, stream);
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+
+    __host__ static void integral(const GlobPtr<uchar> src, GlobPtr<uint> dst, int rows, int cols, cudaStream_t stream)
+    {
+        if (deviceSupports(FEATURE_SET_COMPUTE_30)
+            && (cols % 16 == 0)
+            && reinterpret_cast<intptr_t>(src.data) % 32 == 0
+            && reinterpret_cast<intptr_t>(dst.data) % 32 == 0)
+        {
+            horisontal_pass_8u_shfl(src, dst, rows, cols, stream);
+        }
+        else
+        {
+            horizontal_pass(src, dst, rows, cols, stream);
+        }
+
+        vertical_pass(dst, rows, cols, stream);
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+
+    __host__ static void integral(const GlobPtr<uchar> src, GlobPtr<int> dst, int rows, int cols, cudaStream_t stream)
+    {
+        GlobPtr<uint> dstui = globPtr((uint*) dst.data, dst.step);
+        integral(src, dstui, rows, cols, stream);
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/pyr_down.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/pyr_down.hpp
new file mode 100644
index 0000000000..b14792cf4e
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/pyr_down.hpp
@@ -0,0 +1,201 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_PYR_DOWN_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_PYR_DOWN_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/type_traits.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace pyramids_detail
+{
+    template <class Brd, class SrcPtr, typename DstType>
+    __global__ void pyrDown(const SrcPtr src, GlobPtr<DstType> dst, const int src_rows, const int src_cols, const int dst_cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        __shared__ work_type smem[256 + 4];
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y;
+
+        const int src_y = 2 * y;
+
+        if (src_y >= 2 && src_y < src_rows - 2 && x >= 2 && x < src_cols - 2)
+        {
+            {
+                work_type sum;
+
+                sum =       0.0625f * src(src_y - 2, x);
+                sum = sum + 0.25f   * src(src_y - 1, x);
+                sum = sum + 0.375f  * src(src_y    , x);
+                sum = sum + 0.25f   * src(src_y + 1, x);
+                sum = sum + 0.0625f * src(src_y + 2, x);
+
+                smem[2 + threadIdx.x] = sum;
+            }
+
+            if (threadIdx.x < 2)
+            {
+                const int left_x = x - 2;
+
+                work_type sum;
+
+                sum =       0.0625f * src(src_y - 2, left_x);
+                sum = sum + 0.25f   * src(src_y - 1, left_x);
+                sum = sum + 0.375f  * src(src_y    , left_x);
+                sum = sum + 0.25f   * src(src_y + 1, left_x);
+                sum = sum + 0.0625f * src(src_y + 2, left_x);
+
+                smem[threadIdx.x] = sum;
+            }
+
+            if (threadIdx.x > 253)
+            {
+                const int right_x = x + 2;
+
+                work_type sum;
+
+                sum =       0.0625f * src(src_y - 2, right_x);
+                sum = sum + 0.25f   * src(src_y - 1, right_x);
+                sum = sum + 0.375f  * src(src_y    , right_x);
+                sum = sum + 0.25f   * src(src_y + 1, right_x);
+                sum = sum + 0.0625f * src(src_y + 2, right_x);
+
+                smem[4 + threadIdx.x] = sum;
+            }
+        }
+        else
+        {
+            {
+                work_type sum;
+
+                sum =       0.0625f * src(Brd::idx_low(src_y - 2, src_rows) , Brd::idx_high(x, src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_low(src_y - 1, src_rows) , Brd::idx_high(x, src_cols));
+                sum = sum + 0.375f  * src(src_y                             , Brd::idx_high(x, src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_high(src_y + 1, src_rows), Brd::idx_high(x, src_cols));
+                sum = sum + 0.0625f * src(Brd::idx_high(src_y + 2, src_rows), Brd::idx_high(x, src_cols));
+
+                smem[2 + threadIdx.x] = sum;
+            }
+
+            if (threadIdx.x < 2)
+            {
+                const int left_x = x - 2;
+
+                work_type sum;
+
+                sum =       0.0625f * src(Brd::idx_low(src_y - 2, src_rows) , Brd::idx_low(Brd::idx_high(left_x, src_cols), src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_low(src_y - 1, src_rows) , Brd::idx_low(Brd::idx_high(left_x, src_cols), src_cols));
+                sum = sum + 0.375f  * src(src_y                             , Brd::idx_low(Brd::idx_high(left_x, src_cols), src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_high(src_y + 1, src_rows), Brd::idx_low(Brd::idx_high(left_x, src_cols), src_cols));
+                sum = sum + 0.0625f * src(Brd::idx_high(src_y + 2, src_rows), Brd::idx_low(Brd::idx_high(left_x, src_cols), src_cols));
+
+                smem[threadIdx.x] = sum;
+            }
+
+            if (threadIdx.x > 253)
+            {
+                const int right_x = x + 2;
+
+                work_type sum;
+
+                sum =       0.0625f * src(Brd::idx_low(src_y - 2, src_rows) , Brd::idx_high(right_x, src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_low(src_y - 1, src_rows) , Brd::idx_high(right_x, src_cols));
+                sum = sum + 0.375f  * src(src_y                             , Brd::idx_high(right_x, src_cols));
+                sum = sum + 0.25f   * src(Brd::idx_high(src_y + 1, src_rows), Brd::idx_high(right_x, src_cols));
+                sum = sum + 0.0625f * src(Brd::idx_high(src_y + 2, src_rows), Brd::idx_high(right_x, src_cols));
+
+                smem[4 + threadIdx.x] = sum;
+            }
+        }
+
+        __syncthreads();
+
+        if (threadIdx.x < 128)
+        {
+            const int tid2 = threadIdx.x * 2;
+
+            work_type sum;
+
+            sum =       0.0625f * smem[2 + tid2 - 2];
+            sum = sum + 0.25f   * smem[2 + tid2 - 1];
+            sum = sum + 0.375f  * smem[2 + tid2    ];
+            sum = sum + 0.25f   * smem[2 + tid2 + 1];
+            sum = sum + 0.0625f * smem[2 + tid2 + 2];
+
+            const int dst_x = (blockIdx.x * blockDim.x + tid2) / 2;
+
+            if (dst_x < dst_cols)
+                dst(y, dst_x) = saturate_cast<DstType>(sum);
+        }
+    }
+
+    template <class Brd, class SrcPtr, typename DstType>
+    __host__ void pyrDown(const SrcPtr& src, const GlobPtr<DstType>& dst, int src_rows, int src_cols, int dst_rows, int dst_cols, cudaStream_t stream)
+    {
+        const dim3 block(256);
+        const dim3 grid(divUp(src_cols, block.x), dst_rows);
+
+        pyrDown<Brd><<<grid, block, 0, stream>>>(src, dst, src_rows, src_cols, dst_cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/pyr_up.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/pyr_up.hpp
new file mode 100644
index 0000000000..16c8b2091e
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/pyr_up.hpp
@@ -0,0 +1,172 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_PYR_UP_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_PYR_UP_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/type_traits.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace pyramids_detail
+{
+    template <class SrcPtr, typename DstType>
+    __global__ void pyrUp(const SrcPtr src, GlobPtr<DstType> dst, const int src_rows, const int src_cols, const int dst_rows, const int dst_cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        __shared__ work_type s_srcPatch[10][10];
+        __shared__ work_type s_dstPatch[20][16];
+
+        if (threadIdx.x < 10 && threadIdx.y < 10)
+        {
+            int srcx = static_cast<int>((blockIdx.x * blockDim.x) / 2 + threadIdx.x) - 1;
+            int srcy = static_cast<int>((blockIdx.y * blockDim.y) / 2 + threadIdx.y) - 1;
+
+            srcx = ::abs(srcx);
+            srcx = ::min(src_cols - 1, srcx);
+
+            srcy = ::abs(srcy);
+            srcy = ::min(src_rows - 1, srcy);
+
+            s_srcPatch[threadIdx.y][threadIdx.x] = saturate_cast<work_type>(src(srcy, srcx));
+        }
+
+        __syncthreads();
+
+        work_type sum = VecTraits<work_type>::all(0);
+
+        const int evenFlag = static_cast<int>((threadIdx.x & 1) == 0);
+        const int oddFlag  = static_cast<int>((threadIdx.x & 1) != 0);
+        const bool eveny = ((threadIdx.y & 1) == 0);
+        const int tidx = threadIdx.x;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (threadIdx.y >> 1)][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (threadIdx.y >> 1)][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (threadIdx.y >> 1)][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (threadIdx.y >> 1)][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (threadIdx.y >> 1)][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[2 + threadIdx.y][threadIdx.x] = sum;
+
+        if (threadIdx.y < 2)
+        {
+            sum = VecTraits<work_type>::all(0);
+
+            if (eveny)
+            {
+                sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
+                sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
+                sum = sum + (evenFlag * 0.375f ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
+                sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
+                sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
+            }
+
+            s_dstPatch[threadIdx.y][threadIdx.x] = sum;
+        }
+
+        if (threadIdx.y > 13)
+        {
+            sum = VecTraits<work_type>::all(0);
+
+            if (eveny)
+            {
+                sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
+                sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
+                sum = sum + (evenFlag * 0.375f ) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
+                sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
+                sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
+            }
+
+            s_dstPatch[4 + threadIdx.y][threadIdx.x] = sum;
+        }
+
+        __syncthreads();
+
+        sum = VecTraits<work_type>::all(0);
+
+        const int tidy = threadIdx.y;
+
+        sum = sum + 0.0625f * s_dstPatch[2 + tidy - 2][threadIdx.x];
+        sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][threadIdx.x];
+        sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][threadIdx.x];
+        sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][threadIdx.x];
+        sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][threadIdx.x];
+
+        if (x < dst_cols && y < dst_rows)
+            dst(y, x) = saturate_cast<DstType>(4.0f * sum);
+    }
+
+    template <class SrcPtr, typename DstType>
+    __host__ void pyrUp(const SrcPtr& src, const GlobPtr<DstType>& dst, int src_rows, int src_cols, int dst_rows, int dst_cols, cudaStream_t stream)
+    {
+        const dim3 block(16, 16);
+        const dim3 grid(divUp(dst_cols, block.x), divUp(dst_rows, block.y));
+
+        pyrUp<<<grid, block, 0, stream>>>(src, dst, src_rows, src_cols, dst_rows, dst_cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_column.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_column.hpp
new file mode 100644
index 0000000000..b257e75039
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_column.hpp
@@ -0,0 +1,106 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_REDUCE_TO_COLUMN_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_REDUCE_TO_COLUMN_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../block/reduce.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_reduce_to_vec_detail
+{
+    template <class Reductor, int BLOCK_SIZE, class SrcPtr, typename ResType, class MaskPtr>
+    __global__ void reduceToColumn(const SrcPtr src, ResType* dst, const MaskPtr mask, const int cols)
+    {
+        typedef typename Reductor::work_type work_type;
+
+        __shared__ work_type smem[BLOCK_SIZE];
+
+        const int y = blockIdx.x;
+
+        work_type myVal = Reductor::initialValue();
+
+        Reductor op;
+
+        for (int x = threadIdx.x; x < cols; x += BLOCK_SIZE)
+        {
+            if (mask(y, x))
+            {
+                myVal = op(myVal, saturate_cast<work_type>(src(y, x)));
+            }
+        }
+
+        blockReduce<BLOCK_SIZE>(smem, myVal, threadIdx.x, op);
+
+        if (threadIdx.x == 0)
+            dst[y] = saturate_cast<ResType>(Reductor::result(myVal, cols));
+    }
+
+    template <class Reductor, class Policy, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void reduceToColumn(const SrcPtr& src, ResType* dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const int BLOCK_SIZE_X = Policy::block_size_x;
+        const int BLOCK_SIZE_Y = Policy::block_size_y;
+
+        const int BLOCK_SIZE = BLOCK_SIZE_X * BLOCK_SIZE_Y;
+
+        const dim3 block(BLOCK_SIZE);
+        const dim3 grid(rows);
+
+        reduceToColumn<Reductor, BLOCK_SIZE><<<grid, block, 0, stream>>>(src, dst, mask, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_row.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_row.hpp
new file mode 100644
index 0000000000..154004f6e4
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/reduce_to_row.hpp
@@ -0,0 +1,118 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_REDUCE_TO_ROW_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_REDUCE_TO_ROW_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../block/reduce.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_reduce_to_vec_detail
+{
+    template <class Reductor, int BLOCK_SIZE_X, int BLOCK_SIZE_Y, class SrcPtr, typename ResType, class MaskPtr>
+    __global__ void reduceToRow(const SrcPtr src, ResType* dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename Reductor::work_type work_type;
+
+        __shared__ work_type smem[BLOCK_SIZE_X * BLOCK_SIZE_Y];
+
+        const int x = blockIdx.x * BLOCK_SIZE_X + threadIdx.x;
+
+        work_type myVal = Reductor::initialValue();
+
+        Reductor op;
+
+        if (x < cols)
+        {
+            for (int y = threadIdx.y; y < rows; y += BLOCK_SIZE_Y)
+            {
+                if (mask(y, x))
+                {
+                    myVal = op(myVal, saturate_cast<work_type>(src(y, x)));
+                }
+            }
+        }
+
+        smem[threadIdx.x * BLOCK_SIZE_Y + threadIdx.y] = myVal;
+
+        __syncthreads();
+
+        volatile work_type* srow = smem + threadIdx.y * BLOCK_SIZE_X;
+
+        myVal = srow[threadIdx.x];
+        blockReduce<BLOCK_SIZE_X>(srow, myVal, threadIdx.x, op);
+
+        if (threadIdx.x == 0)
+            srow[0] = myVal;
+
+        __syncthreads();
+
+        if (threadIdx.y == 0 && x < cols)
+            dst[x] = saturate_cast<ResType>(Reductor::result(smem[threadIdx.x * BLOCK_SIZE_X], rows));
+    }
+
+    template <class Reductor, class SrcPtr, typename ResType, class MaskPtr>
+    __host__ void reduceToRow(const SrcPtr& src, ResType* dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const int BLOCK_SIZE_X = 16;
+        const int BLOCK_SIZE_Y = 16;
+
+        const dim3 block(BLOCK_SIZE_X, BLOCK_SIZE_Y);
+        const dim3 grid(divUp(cols, block.x));
+
+        reduceToRow<Reductor, BLOCK_SIZE_X, BLOCK_SIZE_Y><<<grid, block, 0, stream>>>(src, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/split_merge.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/split_merge.hpp
new file mode 100644
index 0000000000..0bd76ac464
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/split_merge.hpp
@@ -0,0 +1,282 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_SPLIT_MERGE_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_SPLIT_MERGE_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_split_merge_detail
+{
+    // merge
+
+    template <class Src1Ptr, class Src2Ptr, typename DstType, class MaskPtr>
+    __global__ void mergeC2(const Src1Ptr src1, const Src2Ptr src2, GlobPtr<DstType> dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename VecTraits<DstType>::elem_type dst_elem_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = VecTraits<DstType>::make(
+                    saturate_cast<dst_elem_type>(src1(y, x)),
+                    saturate_cast<dst_elem_type>(src2(y, x))
+                    );
+    }
+
+    template <class Policy, class Src1Ptr, class Src2Ptr, typename DstType, class MaskPtr>
+    __host__ void mergeC2(const Src1Ptr& src1, const Src2Ptr& src2, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        mergeC2<<<grid, block, 0, stream>>>(src1, src2, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+
+    template <class Src1Ptr, class Src2Ptr, class Src3Ptr, typename DstType, class MaskPtr>
+    __global__ void mergeC3(const Src1Ptr src1, const Src2Ptr src2, const Src3Ptr src3, GlobPtr<DstType> dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename VecTraits<DstType>::elem_type dst_elem_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = VecTraits<DstType>::make(
+                    saturate_cast<dst_elem_type>(src1(y, x)),
+                    saturate_cast<dst_elem_type>(src2(y, x)),
+                    saturate_cast<dst_elem_type>(src3(y, x))
+                    );
+    }
+
+    template <class Policy, class Src1Ptr, class Src2Ptr, class Src3Ptr, typename DstType, class MaskPtr>
+    __host__ void mergeC3(const Src1Ptr& src1, const Src2Ptr& src2, const Src3Ptr& src3, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        mergeC3<<<grid, block, 0, stream>>>(src1, src2, src3, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+
+    template <class Src1Ptr, class Src2Ptr, class Src3Ptr, class Src4Ptr, typename DstType, class MaskPtr>
+    __global__ void mergeC4(const Src1Ptr src1, const Src2Ptr src2, const Src3Ptr src3, const Src4Ptr src4, GlobPtr<DstType> dst, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename VecTraits<DstType>::elem_type dst_elem_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = VecTraits<DstType>::make(
+                    saturate_cast<dst_elem_type>(src1(y, x)),
+                    saturate_cast<dst_elem_type>(src2(y, x)),
+                    saturate_cast<dst_elem_type>(src3(y, x)),
+                    saturate_cast<dst_elem_type>(src4(y, x))
+                    );
+    }
+
+    template <class Policy, class Src1Ptr, class Src2Ptr, class Src3Ptr, class Src4Ptr, typename DstType, class MaskPtr>
+    __host__ void mergeC4(const Src1Ptr& src1, const Src2Ptr& src2, const Src3Ptr& src3, const Src4Ptr& src4, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        mergeC4<<<grid, block, 0, stream>>>(src1, src2, src3, src4, dst, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+
+    template <int cn, class Policy> struct MergeImpl;
+
+    template <class Policy> struct MergeImpl<2, Policy>
+    {
+        template <class SrcPtrTuple, typename DstType, class MaskPtr>
+        __host__ static void merge(const SrcPtrTuple& src, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            mergeC2<Policy>(get<0>(src), get<1>(src), dst, mask, rows, cols, stream);
+        }
+    };
+
+    template <class Policy> struct MergeImpl<3, Policy>
+    {
+        template <class SrcPtrTuple, typename DstType, class MaskPtr>
+        __host__ static void merge(const SrcPtrTuple& src, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            mergeC3<Policy>(get<0>(src), get<1>(src), get<2>(src), dst, mask, rows, cols, stream);
+        }
+    };
+
+    template <class Policy> struct MergeImpl<4, Policy>
+    {
+        template <class SrcPtrTuple, typename DstType, class MaskPtr>
+        __host__ static void merge(const SrcPtrTuple& src, const GlobPtr<DstType>& dst, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            mergeC4<Policy>(get<0>(src), get<1>(src), get<2>(src), get<3>(src), dst, mask, rows, cols, stream);
+        }
+    };
+
+    // split
+
+    template <class SrcPtr, typename DstType, class MaskPtr>
+    __global__ void split(const SrcPtr src, GlobPtr<DstType> dst1, GlobPtr<DstType> dst2, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        const src_type src_value = src(y, x);
+
+        dst1(y, x) = src_value.x;
+        dst2(y, x) = src_value.y;
+    }
+
+    template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+    __host__ void split(const SrcPtr& src, const GlobPtr<DstType>& dst1, const GlobPtr<DstType>& dst2, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        split<<<grid, block, 0, stream>>>(src, dst1, dst2, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+
+    template <class SrcPtr, typename DstType, class MaskPtr>
+    __global__ void split(const SrcPtr src, GlobPtr<DstType> dst1, GlobPtr<DstType> dst2, GlobPtr<DstType> dst3, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        const src_type src_value = src(y, x);
+
+        dst1(y, x) = src_value.x;
+        dst2(y, x) = src_value.y;
+        dst3(y, x) = src_value.z;
+    }
+
+    template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+    __host__ void split(const SrcPtr& src, const GlobPtr<DstType>& dst1, const GlobPtr<DstType>& dst2, const GlobPtr<DstType>& dst3, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        split<<<grid, block, 0, stream>>>(src, dst1, dst2, dst3, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+
+    template <class SrcPtr, typename DstType, class MaskPtr>
+    __global__ void split(const SrcPtr src, GlobPtr<DstType> dst1, GlobPtr<DstType> dst2, GlobPtr<DstType> dst3, GlobPtr<DstType> dst4, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        const src_type src_value = src(y, x);
+
+        dst1(y, x) = src_value.x;
+        dst2(y, x) = src_value.y;
+        dst3(y, x) = src_value.z;
+        dst4(y, x) = src_value.w;
+    }
+
+    template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+    __host__ void split(const SrcPtr& src, const GlobPtr<DstType>& dst1, const GlobPtr<DstType>& dst2, const GlobPtr<DstType>& dst3, const GlobPtr<DstType>& dst4, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        split<<<grid, block, 0, stream>>>(src, dst1, dst2, dst3, dst4, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/transform.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/transform.hpp
new file mode 100644
index 0000000000..0e0df0794a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/transform.hpp
@@ -0,0 +1,423 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_TRANSFORM_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_TRANSFORM_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../util/vec_traits.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace grid_transform_detail
+{
+    // OpUnroller
+
+    template <int cn> struct OpUnroller;
+
+    template <> struct OpUnroller<1>
+    {
+        template <typename T, typename D, class UnOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T& src, D& dst, const UnOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src.x);
+        }
+
+        template <typename T1, typename T2, typename D, class BinOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T1& src1, const T2& src2, D& dst, const BinOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src1.x, src2.x);
+        }
+    };
+
+    template <> struct OpUnroller<2>
+    {
+        template <typename T, typename D, class UnOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T& src, D& dst, const UnOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src.y);
+        }
+
+        template <typename T1, typename T2, typename D, class BinOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T1& src1, const T2& src2, D& dst, const BinOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src1.x, src2.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src1.y, src2.y);
+        }
+    };
+
+    template <> struct OpUnroller<3>
+    {
+        template <typename T, typename D, class UnOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T& src, D& dst, const UnOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src.y);
+            if (mask(y, x_shifted + 2))
+                dst.z = op(src.z);
+        }
+
+        template <typename T1, typename T2, typename D, class BinOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T1& src1, const T2& src2, D& dst, const BinOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src1.x, src2.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src1.y, src2.y);
+            if (mask(y, x_shifted + 2))
+                dst.z = op(src1.z, src2.z);
+        }
+    };
+
+    template <> struct OpUnroller<4>
+    {
+        template <typename T, typename D, class UnOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T& src, D& dst, const UnOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src.y);
+            if (mask(y, x_shifted + 2))
+                dst.z = op(src.z);
+            if (mask(y, x_shifted + 3))
+                dst.w = op(src.w);
+        }
+
+        template <typename T1, typename T2, typename D, class BinOp, class MaskPtr>
+        __device__ __forceinline__ static void unroll(const T1& src1, const T2& src2, D& dst, const BinOp& op, const MaskPtr& mask, int x_shifted, int y)
+        {
+            if (mask(y, x_shifted))
+                dst.x = op(src1.x, src2.x);
+            if (mask(y, x_shifted + 1))
+                dst.y = op(src1.y, src2.y);
+            if (mask(y, x_shifted + 2))
+                dst.z = op(src1.z, src2.z);
+            if (mask(y, x_shifted + 3))
+                dst.w = op(src1.w, src2.w);
+        }
+    };
+
+    // transformSimple
+
+    template <class SrcPtr, typename DstType, class UnOp, class MaskPtr>
+    __global__ void transformSimple(const SrcPtr src, GlobPtr<DstType> dst, const UnOp op, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = saturate_cast<DstType>(op(src(y, x)));
+    }
+
+    template <class SrcPtr1, class SrcPtr2, typename DstType, class BinOp, class MaskPtr>
+    __global__ void transformSimple(const SrcPtr1 src1, const SrcPtr2 src2, GlobPtr<DstType> dst, const BinOp op, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        dst(y, x) = saturate_cast<DstType>(op(src1(y, x), src2(y, x)));
+    }
+
+    // transformSmart
+
+    template <int SHIFT, typename SrcType, typename DstType, class UnOp, class MaskPtr>
+    __global__ void transformSmart(const GlobPtr<SrcType> src_, GlobPtr<DstType> dst_, const UnOp op, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename MakeVec<SrcType, SHIFT>::type read_type;
+        typedef typename MakeVec<DstType, SHIFT>::type write_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+        const int x_shifted = x * SHIFT;
+
+        if (y < rows)
+        {
+            const SrcType* src = src_.row(y);
+            DstType* dst = dst_.row(y);
+
+            if (x_shifted + SHIFT - 1 < cols)
+            {
+                const read_type src_n_el = ((const read_type*)src)[x];
+                write_type dst_n_el = ((const write_type*)dst)[x];
+
+                OpUnroller<SHIFT>::unroll(src_n_el, dst_n_el, op, mask, x_shifted, y);
+
+                ((write_type*)dst)[x] = dst_n_el;
+            }
+            else
+            {
+                for (int real_x = x_shifted; real_x < cols; ++real_x)
+                {
+                    if (mask(y, real_x))
+                        dst[real_x] = op(src[real_x]);
+                }
+            }
+        }
+    }
+
+    template <int SHIFT, typename SrcType1, typename SrcType2, typename DstType, class BinOp, class MaskPtr>
+    __global__ void transformSmart(const GlobPtr<SrcType1> src1_, const GlobPtr<SrcType2> src2_, PtrStep<DstType> dst_, const BinOp op, const MaskPtr mask, const int rows, const int cols)
+    {
+        typedef typename MakeVec<SrcType1, SHIFT>::type read_type1;
+        typedef typename MakeVec<SrcType2, SHIFT>::type read_type2;
+        typedef typename MakeVec<DstType, SHIFT>::type write_type;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+        const int x_shifted = x * SHIFT;
+
+        if (y < rows)
+        {
+            const SrcType1* src1 = src1_.row(y);
+            const SrcType2* src2 = src2_.row(y);
+            DstType* dst = dst_.row(y);
+
+            if (x_shifted + SHIFT - 1 < cols)
+            {
+                const read_type1 src1_n_el = ((const read_type1*)src1)[x];
+                const read_type2 src2_n_el = ((const read_type2*)src2)[x];
+                write_type dst_n_el = ((const write_type*)dst)[x];
+
+                OpUnroller<SHIFT>::unroll(src1_n_el, src2_n_el, dst_n_el, op, mask, x_shifted, y);
+
+                ((write_type*)dst)[x] = dst_n_el;
+            }
+            else
+            {
+                for (int real_x = x_shifted; real_x < cols; ++real_x)
+                {
+                    if (mask(y, real_x))
+                        dst[real_x] = op(src1[real_x], src2[real_x]);
+                }
+            }
+        }
+    }
+
+    // TransformDispatcher
+
+    template <bool UseSmart, class Policy> struct TransformDispatcher;
+
+    template <class Policy> struct TransformDispatcher<false, Policy>
+    {
+        template <class SrcPtr, typename DstType, class UnOp, class MaskPtr>
+        __host__ static void call(const SrcPtr& src, const GlobPtr<DstType>& dst, const UnOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            const dim3 block(Policy::block_size_x, Policy::block_size_y);
+            const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+            transformSimple<<<grid, block, 0, stream>>>(src, dst, op, mask, rows, cols);
+            CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+            if (stream == 0)
+                CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+        }
+
+        template <class SrcPtr1, class SrcPtr2, typename DstType, class BinOp, class MaskPtr>
+        __host__ static void call(const SrcPtr1& src1, const SrcPtr2& src2, const GlobPtr<DstType>& dst, const BinOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            const dim3 block(Policy::block_size_x, Policy::block_size_y);
+            const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+            transformSimple<<<grid, block, 0, stream>>>(src1, src2, dst, op, mask, rows, cols);
+            CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+            if (stream == 0)
+                CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+        }
+    };
+
+    template <class Policy> struct TransformDispatcher<true, Policy>
+    {
+        template <typename T>
+        __host__ static bool isAligned(const T* ptr, size_t size)
+        {
+            return reinterpret_cast<size_t>(ptr) % size == 0;
+        }
+
+        __host__ static bool isAligned(size_t step, size_t size)
+        {
+            return step % size == 0;
+        }
+
+        template <typename SrcType, typename DstType, class UnOp, class MaskPtr>
+        __host__ static void call(const GlobPtr<SrcType>& src, const GlobPtr<DstType>& dst, const UnOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            if (Policy::shift == 1 ||
+                !isAligned(src.data, Policy::shift * sizeof(SrcType)) || !isAligned(src.step, Policy::shift * sizeof(SrcType)) ||
+                !isAligned(dst.data, Policy::shift * sizeof(DstType)) || !isAligned(dst.step, Policy::shift * sizeof(DstType)))
+            {
+                TransformDispatcher<false, Policy>::call(src, dst, op, mask, rows, cols, stream);
+                return;
+            }
+
+            const dim3 block(Policy::block_size_x, Policy::block_size_y);
+            const dim3 grid(divUp(cols, block.x * Policy::shift), divUp(rows, block.y));
+
+            transformSmart<Policy::shift><<<grid, block, 0, stream>>>(src, dst, op, mask, rows, cols);
+            CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+            if (stream == 0)
+                CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+        }
+
+        template <typename SrcType1, typename SrcType2, typename DstType, class BinOp, class MaskPtr>
+        __host__ static void call(const GlobPtr<SrcType1>& src1, const GlobPtr<SrcType2>& src2, const GlobPtr<DstType>& dst, const BinOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+        {
+            if (Policy::shift == 1 ||
+                !isAligned(src1.data, Policy::shift * sizeof(SrcType1)) || !isAligned(src1.step, Policy::shift * sizeof(SrcType1)) ||
+                !isAligned(src2.data, Policy::shift * sizeof(SrcType2)) || !isAligned(src2.step, Policy::shift * sizeof(SrcType2)) ||
+                !isAligned(dst.data,  Policy::shift * sizeof(DstType))  || !isAligned(dst.step,  Policy::shift * sizeof(DstType)))
+            {
+                TransformDispatcher<false, Policy>::call(src1, src2, dst, op, mask, rows, cols, stream);
+                return;
+            }
+
+            const dim3 block(Policy::block_size_x, Policy::block_size_y);
+            const dim3 grid(divUp(cols, block.x * Policy::shift), divUp(rows, block.y));
+
+            transformSmart<Policy::shift><<<grid, block, 0, stream>>>(src1, src2, dst, op, mask, rows, cols);
+            CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+            if (stream == 0)
+                CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+        }
+    };
+
+    template <class Policy, class SrcPtr, typename DstType, class UnOp, class MaskPtr>
+    __host__ void transform(const SrcPtr& src, const GlobPtr<DstType>& dst, const UnOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        TransformDispatcher<false, Policy>::call(src, dst, op, mask, rows, cols, stream);
+    }
+
+    template <class Policy, class SrcPtr1, class SrcPtr2, typename DstType, class BinOp, class MaskPtr>
+    __host__ void transform(const SrcPtr1& src1, const SrcPtr2& src2, const GlobPtr<DstType>& dst, const BinOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        TransformDispatcher<false, Policy>::call(src1, src2, dst, op, mask, rows, cols, stream);
+    }
+
+    template <class Policy, typename SrcType, typename DstType, class UnOp, class MaskPtr>
+    __host__ void transform(const GlobPtr<SrcType>& src, const GlobPtr<DstType>& dst, const UnOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        TransformDispatcher<VecTraits<SrcType>::cn == 1 && VecTraits<DstType>::cn == 1 && Policy::shift != 1, Policy>::call(src, dst, op, mask, rows, cols, stream);
+    }
+
+    template <class Policy, typename SrcType1, typename SrcType2, typename DstType, class BinOp, class MaskPtr>
+    __host__ void transform(const GlobPtr<SrcType1>& src1, const GlobPtr<SrcType2>& src2, const GlobPtr<DstType>& dst, const BinOp& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        TransformDispatcher<VecTraits<SrcType1>::cn == 1 && VecTraits<SrcType2>::cn == 1 && VecTraits<DstType>::cn == 1 && Policy::shift != 1, Policy>::call(src1, src2, dst, op, mask, rows, cols, stream);
+    }
+
+    // transform_tuple
+
+    template <int count> struct Unroll
+    {
+        template <class SrcVal, class DstPtrTuple, class OpTuple>
+        __device__ static void transform(const SrcVal& srcVal, DstPtrTuple& dst, const OpTuple& op, int y, int x)
+        {
+            typedef typename tuple_element<count - 1, DstPtrTuple>::type dst_ptr_type;
+            typedef typename PtrTraits<dst_ptr_type>::value_type dst_type;
+
+            get<count - 1>(dst)(y, x) = saturate_cast<dst_type>(get<count - 1>(op)(srcVal));
+            Unroll<count - 1>::transform(srcVal, dst, op, y, x);
+        }
+    };
+    template <> struct Unroll<0>
+    {
+        template <class SrcVal, class DstPtrTuple, class OpTuple>
+        __device__ __forceinline__ static void transform(const SrcVal&, DstPtrTuple&, const OpTuple&, int, int)
+        {
+        }
+    };
+
+    template <class SrcPtr, class DstPtrTuple, class OpTuple, class MaskPtr>
+    __global__ void transform_tuple(const SrcPtr src, DstPtrTuple dst, const OpTuple op, const MaskPtr mask, const int rows, const int cols)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= cols || y >= rows || !mask(y, x))
+            return;
+
+        typename PtrTraits<SrcPtr>::value_type srcVal = src(y, x);
+
+        Unroll<tuple_size<DstPtrTuple>::value>::transform(srcVal, dst, op, y, x);
+    }
+
+    template <class Policy, class SrcPtrTuple, class DstPtrTuple, class OpTuple, class MaskPtr>
+    __host__ void transform_tuple(const SrcPtrTuple& src, const DstPtrTuple& dst, const OpTuple& op, const MaskPtr& mask, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(Policy::block_size_x, Policy::block_size_y);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        transform_tuple<<<grid, block, 0, stream>>>(src, dst, op, mask, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/detail/transpose.hpp b/modules/cudev/include/opencv2/cudev/grid/detail/transpose.hpp
new file mode 100644
index 0000000000..3854a17402
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/detail/transpose.hpp
@@ -0,0 +1,130 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_TRANSPOSE_DETAIL_HPP__
+#define __OPENCV_CUDEV_GRID_TRANSPOSE_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/saturate_cast.hpp"
+#include "../../ptr2d/glob.hpp"
+#include "../../ptr2d/traits.hpp"
+
+namespace cv { namespace cudev {
+
+namespace transpose_detail
+{
+    const int TRANSPOSE_TILE_DIM   = 16;
+    const int TRANSPOSE_BLOCK_ROWS = 16;
+
+    template <class SrcPtr, typename DstType>
+    __global__ void transpose(const SrcPtr src, GlobPtr<DstType> dst, const int rows, const int cols)
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+
+        __shared__ src_type tile[TRANSPOSE_TILE_DIM][TRANSPOSE_TILE_DIM + 1];
+
+        int blockIdx_x, blockIdx_y;
+
+        // do diagonal reordering
+        if (gridDim.x == gridDim.y)
+        {
+            blockIdx_y = blockIdx.x;
+            blockIdx_x = (blockIdx.x + blockIdx.y) % gridDim.x;
+        }
+        else
+        {
+            int bid = blockIdx.x + gridDim.x * blockIdx.y;
+            blockIdx_y = bid % gridDim.y;
+            blockIdx_x = ((bid / gridDim.y) + blockIdx_y) % gridDim.x;
+        }
+
+        int xIndex = blockIdx_x * TRANSPOSE_TILE_DIM + threadIdx.x;
+        int yIndex = blockIdx_y * TRANSPOSE_TILE_DIM + threadIdx.y;
+
+        if (xIndex < cols)
+        {
+            for (int i = 0; i < TRANSPOSE_TILE_DIM; i += TRANSPOSE_BLOCK_ROWS)
+            {
+                if (yIndex + i < rows)
+                {
+                    tile[threadIdx.y + i][threadIdx.x] = src(yIndex + i, xIndex);
+                }
+            }
+        }
+
+        __syncthreads();
+
+        xIndex = blockIdx_y * TRANSPOSE_TILE_DIM + threadIdx.x;
+        yIndex = blockIdx_x * TRANSPOSE_TILE_DIM + threadIdx.y;
+
+        if (xIndex < rows)
+        {
+            for (int i = 0; i < TRANSPOSE_TILE_DIM; i += TRANSPOSE_BLOCK_ROWS)
+            {
+                if (yIndex + i < cols)
+                {
+                    dst(yIndex + i, xIndex) = saturate_cast<DstType>(tile[threadIdx.x][threadIdx.y + i]);
+                }
+            }
+        }
+    }
+
+    template <class SrcPtr, typename DstType>
+    __host__ void transpose(const SrcPtr& src, const GlobPtr<DstType>& dst, int rows, int cols, cudaStream_t stream)
+    {
+        const dim3 block(TRANSPOSE_TILE_DIM, TRANSPOSE_TILE_DIM);
+        const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
+
+        transpose<<<grid, block, 0, stream>>>(src, dst, rows, cols);
+        CV_CUDEV_SAFE_CALL( cudaGetLastError() );
+
+        if (stream == 0)
+            CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() );
+    }
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/glob_reduce.hpp b/modules/cudev/include/opencv2/cudev/grid/glob_reduce.hpp
new file mode 100644
index 0000000000..8d3e0826d1
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/glob_reduce.hpp
@@ -0,0 +1,314 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_GLOB_REDUCE_HPP__
+#define __OPENCV_CUDEV_GRID_GLOB_REDUCE_HPP__
+
+#include <limits>
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/mask.hpp"
+#include "../ptr2d/transform.hpp"
+#include "detail/glob_reduce.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridCalcSum_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(0, stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_glob_reduce_detail::sum<Policy>(shrinkPtr(src),
+                                         dst[0],
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType>
+__host__ void gridCalcSum_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(0, stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    grid_glob_reduce_detail::sum<Policy>(shrinkPtr(src),
+                                         dst[0],
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMinVal_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(Scalar::all(std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_glob_reduce_detail::minVal<Policy>(shrinkPtr(src),
+                                            dst[0],
+                                            shrinkPtr(mask),
+                                            rows, cols,
+                                            StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType>
+__host__ void gridFindMinVal_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(Scalar::all(std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    grid_glob_reduce_detail::minVal<Policy>(shrinkPtr(src),
+                                            dst[0],
+                                            WithOutMask(),
+                                            rows, cols,
+                                            StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMaxVal_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(Scalar::all(-std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_glob_reduce_detail::maxVal<Policy>(shrinkPtr(src),
+                                            dst[0],
+                                            shrinkPtr(mask),
+                                            rows, cols,
+                                            StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType>
+__host__ void gridFindMaxVal_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(Scalar::all(-std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    grid_glob_reduce_detail::maxVal<Policy>(shrinkPtr(src),
+                                            dst[0],
+                                            WithOutMask(),
+                                            rows, cols,
+                                            StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMinMaxVal_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    dst.create(1, 2);
+    dst.col(0).setTo(Scalar::all(std::numeric_limits<ResType>::max()), stream);
+    dst.col(1).setTo(Scalar::all(-std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_glob_reduce_detail::minMaxVal<Policy>(shrinkPtr(src),
+                                               dst[0],
+                                               shrinkPtr(mask),
+                                               rows, cols,
+                                               StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType>
+__host__ void gridFindMinMaxVal_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    dst.create(1, 2);
+    dst.col(0).setTo(Scalar::all(std::numeric_limits<ResType>::max()), stream);
+    dst.col(1).setTo(Scalar::all(-std::numeric_limits<ResType>::max()), stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    grid_glob_reduce_detail::minMaxVal<Policy>(shrinkPtr(src),
+                                               dst[0],
+                                               WithOutMask(),
+                                               rows, cols,
+                                               StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridCountNonZero_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(0, stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    typedef typename PtrTraits<SrcPtr>::value_type src_type;
+    not_equal_to<src_type> ne_op;
+    const src_type zero = VecTraits<src_type>::all(0);
+
+    grid_glob_reduce_detail::sum<Policy>(shrinkPtr(transformPtr(src, bind2nd(ne_op, zero))),
+                                         dst[0],
+                                         shrinkPtr(mask),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename ResType>
+__host__ void gridCountNonZero_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    dst.create(1, 1);
+    dst.setTo(0, stream);
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    typedef typename PtrTraits<SrcPtr>::value_type src_type;
+    not_equal_to<src_type> ne_op;
+    const src_type zero = VecTraits<src_type>::all(0);
+
+    grid_glob_reduce_detail::sum<Policy>(shrinkPtr(transformPtr(src, bind2nd(ne_op, zero))),
+                                         dst[0],
+                                         WithOutMask(),
+                                         rows, cols,
+                                         StreamAccessor::getStream(stream));
+}
+
+// default policy
+
+struct DefaultGlobReducePolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8,
+
+        patch_size_x = 4,
+        patch_size_y = 4
+    };
+};
+
+template <class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridCalcSum(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCalcSum_<DefaultGlobReducePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename ResType>
+__host__ void gridCalcSum(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridCalcSum_<DefaultGlobReducePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMinVal(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridFindMinVal_<DefaultGlobReducePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename ResType>
+__host__ void gridFindMinVal(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridFindMinVal_<DefaultGlobReducePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMaxVal(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridFindMaxVal_<DefaultGlobReducePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename ResType>
+__host__ void gridFindMaxVal(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridFindMaxVal_<DefaultGlobReducePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridFindMinMaxVal(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridFindMinMaxVal_<DefaultGlobReducePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename ResType>
+__host__ void gridFindMinMaxVal(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridFindMinMaxVal_<DefaultGlobReducePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridCountNonZero(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridCountNonZero_<DefaultGlobReducePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename ResType>
+__host__ void gridCountNonZero(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridCountNonZero_<DefaultGlobReducePolicy>(src, dst, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/histogram.hpp b/modules/cudev/include/opencv2/cudev/grid/histogram.hpp
new file mode 100644
index 0000000000..b81b57febc
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/histogram.hpp
@@ -0,0 +1,115 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_HISTOGRAM_HPP__
+#define __OPENCV_CUDEV_GRID_HISTOGRAM_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/mask.hpp"
+#include "detail/histogram.hpp"
+
+namespace cv { namespace cudev {
+
+template <int BIN_COUNT, class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridHistogram_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(1, BIN_COUNT);
+    dst.setTo(0, stream);
+
+    grid_histogram_detail::histogram<BIN_COUNT, Policy>(shrinkPtr(src),
+                                                        dst[0],
+                                                        shrinkPtr(mask),
+                                                        rows, cols,
+                                                        StreamAccessor::getStream(stream));
+}
+
+template <int BIN_COUNT, class Policy, class SrcPtr, typename ResType>
+__host__ void gridHistogram_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(1, BIN_COUNT);
+    dst.setTo(0, stream);
+
+    grid_histogram_detail::histogram<BIN_COUNT, Policy>(shrinkPtr(src),
+                                                        dst[0],
+                                                        WithOutMask(),
+                                                        rows, cols,
+                                                        StreamAccessor::getStream(stream));
+}
+
+// default policy
+
+struct DefaultHistogramPolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8
+    };
+};
+
+template <int BIN_COUNT, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridHistogram(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridHistogram_<BIN_COUNT, DefaultHistogramPolicy>(src, dst, mask, stream);
+}
+
+template <int BIN_COUNT, class SrcPtr, typename ResType>
+__host__ void gridHistogram(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridHistogram_<BIN_COUNT, DefaultHistogramPolicy>(src, dst, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/integral.hpp b/modules/cudev/include/opencv2/cudev/grid/integral.hpp
new file mode 100644
index 0000000000..d948c1267f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/integral.hpp
@@ -0,0 +1,69 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_INTEGRAL_HPP__
+#define __OPENCV_CUDEV_GRID_INTEGRAL_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "detail/integral.hpp"
+
+namespace cv { namespace cudev {
+
+template <class SrcPtr, typename DstType>
+__host__ void gridIntegral(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(rows, cols);
+
+    integral_detail::integral(shrinkPtr(src), shrinkPtr(dst), rows, cols, StreamAccessor::getStream(stream));
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/pyramids.hpp b/modules/cudev/include/opencv2/cudev/grid/pyramids.hpp
new file mode 100644
index 0000000000..99833bd3f7
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/pyramids.hpp
@@ -0,0 +1,88 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_PYRAMIDS_HPP__
+#define __OPENCV_CUDEV_GRID_PYRAMIDS_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/extrapolation.hpp"
+#include "detail/pyr_down.hpp"
+#include "detail/pyr_up.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Brd, class SrcPtr, typename DstType>
+__host__ void gridPyrDown_(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(divUp(rows, 2), divUp(cols, 2));
+
+    pyramids_detail::pyrDown<Brd>(shrinkPtr(src), shrinkPtr(dst), rows, cols, dst.rows, dst.cols, StreamAccessor::getStream(stream));
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridPyrDown(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    gridPyrDown_<BrdReflect101>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridPyrUp(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(rows * 2, cols * 2);
+
+    pyramids_detail::pyrUp(shrinkPtr(src), shrinkPtr(dst), rows, cols, dst.rows, dst.cols, StreamAccessor::getStream(stream));
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/reduce_to_vec.hpp b/modules/cudev/include/opencv2/cudev/grid/reduce_to_vec.hpp
new file mode 100644
index 0000000000..f9e3512428
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/reduce_to_vec.hpp
@@ -0,0 +1,209 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_REDUCE_TO_VEC_HPP__
+#define __OPENCV_CUDEV_GRID_REDUCE_TO_VEC_HPP__
+
+#include "../common.hpp"
+#include "../util/vec_traits.hpp"
+#include "../util/limits.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/mask.hpp"
+#include "../functional/functional.hpp"
+#include "detail/reduce_to_column.hpp"
+#include "detail/reduce_to_row.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T> struct Sum : plus<T>
+{
+    typedef T work_type;
+
+    __device__ __forceinline__ static T initialValue()
+    {
+        return VecTraits<T>::all(0);
+    }
+
+    __device__ __forceinline__ static T result(T r, int)
+    {
+        return r;
+    }
+};
+
+template <typename T> struct Avg : plus<T>
+{
+    typedef T work_type;
+
+    __device__ __forceinline__ static T initialValue()
+    {
+        return VecTraits<T>::all(0);
+    }
+
+    __device__ __forceinline__ static T result(T r, int sz)
+    {
+        return r / sz;
+    }
+};
+
+template <typename T> struct Min : minimum<T>
+{
+    typedef T work_type;
+
+    __device__ __forceinline__ static T initialValue()
+    {
+        return VecTraits<T>::all(numeric_limits<typename VecTraits<T>::elem_type>::max());
+    }
+
+    __device__ __forceinline__ static T result(T r, int)
+    {
+        return r;
+    }
+};
+
+template <typename T> struct Max : maximum<T>
+{
+    typedef T work_type;
+
+    __device__ __forceinline__ static T initialValue()
+    {
+        return VecTraits<T>::all(-numeric_limits<typename VecTraits<T>::elem_type>::max());
+    }
+
+    __device__ __forceinline__ static T result(T r, int)
+    {
+        return r;
+    }
+};
+
+template <class Reductor, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridReduceToRow(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(1, cols);
+
+    grid_reduce_to_vec_detail::reduceToRow<Reductor>(shrinkPtr(src),
+                                                     dst[0],
+                                                     shrinkPtr(mask),
+                                                     rows, cols,
+                                                     StreamAccessor::getStream(stream));
+}
+
+template <class Reductor, class SrcPtr, typename ResType>
+__host__ void gridReduceToRow(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(1, cols);
+
+    grid_reduce_to_vec_detail::reduceToRow<Reductor>(shrinkPtr(src),
+                                                     dst[0],
+                                                     WithOutMask(),
+                                                     rows, cols,
+                                                     StreamAccessor::getStream(stream));
+}
+
+template <class Reductor, class Policy, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridReduceToColumn_(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    createContinuous(rows, 1, DataType<ResType>::type, dst);
+
+    grid_reduce_to_vec_detail::reduceToColumn<Reductor, Policy>(shrinkPtr(src),
+                                                                dst[0],
+                                                                shrinkPtr(mask),
+                                                                rows, cols,
+                                                                StreamAccessor::getStream(stream));
+}
+
+template <class Reductor, class Policy, class SrcPtr, typename ResType>
+__host__ void gridReduceToColumn_(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    createContinuous(rows, 1, DataType<ResType>::type, dst);
+
+    grid_reduce_to_vec_detail::reduceToColumn<Reductor, Policy>(shrinkPtr(src),
+                                                                dst[0],
+                                                                WithOutMask(),
+                                                                rows, cols,
+                                                                StreamAccessor::getStream(stream));
+}
+
+// default policy
+
+struct DefaultReduceToVecPolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8
+    };
+};
+
+template <class Reductor, class SrcPtr, typename ResType, class MaskPtr>
+__host__ void gridReduceToColumn(const SrcPtr& src, GpuMat_<ResType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridReduceToColumn_<Reductor, DefaultReduceToVecPolicy>(src, dst, mask, stream);
+}
+
+template <class Reductor, class SrcPtr, typename ResType>
+__host__ void gridReduceToColumn(const SrcPtr& src, GpuMat_<ResType>& dst, Stream& stream = Stream::Null())
+{
+    gridReduceToColumn_<Reductor, DefaultReduceToVecPolicy>(src, dst, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/split_merge.hpp b/modules/cudev/include/opencv2/cudev/grid/split_merge.hpp
new file mode 100644
index 0000000000..78584aad84
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/split_merge.hpp
@@ -0,0 +1,407 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_SPLIT_MERGE_HPP__
+#define __OPENCV_CUDEV_GRID_SPLIT_MERGE_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "../util/vec_traits.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/mask.hpp"
+#include "detail/split_merge.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Policy, class SrcPtrTuple, typename DstType, class MaskPtr>
+__host__ void gridMerge_(const SrcPtrTuple& src, GpuMat_<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<DstType>::cn == tuple_size<SrcPtrTuple>::value, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(rows, cols);
+
+    grid_split_merge_detail::MergeImpl<VecTraits<DstType>::cn, Policy>::merge(shrinkPtr(src),
+                                                                              shrinkPtr(dst),
+                                                                              shrinkPtr(mask),
+                                                                              rows, cols,
+                                                                              StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtrTuple, typename DstType>
+__host__ void gridMerge_(const SrcPtrTuple& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<DstType>::cn == tuple_size<SrcPtrTuple>::value, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(rows, cols);
+
+    grid_split_merge_detail::MergeImpl<VecTraits<DstType>::cn, Policy>::merge(shrinkPtr(src),
+                                                                              shrinkPtr(dst),
+                                                                              WithOutMask(),
+                                                                              rows, cols,
+                                                                              StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[2], const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[2], Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)), shrinkPtr(get<2>(dst)),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[3], const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+    dst[2].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]), shrinkPtr(dst[2]),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)), shrinkPtr(get<2>(dst)),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[3], Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+    dst[2].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]), shrinkPtr(dst[2]),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)), shrinkPtr(get<2>(dst)), shrinkPtr(get<3>(dst)),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[4], const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+    dst[2].create(rows, cols);
+    dst[4].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]), shrinkPtr(dst[2]), shrinkPtr(dst[4]),
+                                           shrinkPtr(mask),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(get<0>(dst)), shrinkPtr(get<1>(dst)), shrinkPtr(get<2>(dst)), shrinkPtr(get<3>(dst)),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType>
+__host__ void gridSplit_(const SrcPtr& src, GpuMat_<DstType> (&dst)[4], Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( VecTraits<typename PtrTraits<SrcPtr>::value_type>::cn == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst[0].create(rows, cols);
+    dst[1].create(rows, cols);
+    dst[2].create(rows, cols);
+    dst[4].create(rows, cols);
+
+    grid_split_merge_detail::split<Policy>(shrinkPtr(src),
+                                           shrinkPtr(dst[0]), shrinkPtr(dst[1]), shrinkPtr(dst[2]), shrinkPtr(dst[4]),
+                                           WithOutMask(),
+                                           rows, cols,
+                                           StreamAccessor::getStream(stream));
+}
+
+// Default Policy
+
+struct DefaultSplitMergePolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8
+    };
+};
+
+template <class SrcPtrTuple, typename DstType, class MaskPtr>
+__host__ void gridMerge(const SrcPtrTuple& src, GpuMat_<DstType>& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridMerge_<DefaultSplitMergePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtrTuple, typename DstType>
+__host__ void gridMerge(const SrcPtrTuple& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    gridMerge_<DefaultSplitMergePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType, class MaskPtr>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType>
+__host__ void gridSplit(const SrcPtr& src, const tuple< GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>&, GpuMat_<DstType>& >& dst, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, stream);
+}
+
+template <class SrcPtr, typename DstType, int COUNT, class MaskPtr>
+__host__ void gridSplit(const SrcPtr& src, GpuMat_<DstType> (&dst)[COUNT], const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, mask, stream);
+}
+
+template <class SrcPtr, typename DstType, int COUNT>
+__host__ void gridSplit(const SrcPtr& src, GpuMat_<DstType> (&dst)[COUNT], Stream& stream = Stream::Null())
+{
+    gridSplit_<DefaultSplitMergePolicy>(src, dst, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/transform.hpp b/modules/cudev/include/opencv2/cudev/grid/transform.hpp
new file mode 100644
index 0000000000..0da5e2d8a4
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/transform.hpp
@@ -0,0 +1,541 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_TRANSFORM_HPP__
+#define __OPENCV_CUDEV_GRID_TRANSFORM_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "../ptr2d/glob.hpp"
+#include "../ptr2d/mask.hpp"
+#include "../ptr2d/zip.hpp"
+#include "detail/transform.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Policy, class SrcPtr, typename DstType, class UnOp, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, GpuMat_<DstType>& dst, const UnOp& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(rows, cols);
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src), shrinkPtr(dst), op, shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class UnOp, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const UnOp& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src), shrinkPtr(dst), op, shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class UnOp>
+__host__ void gridTransform_(const SrcPtr& src, GpuMat_<DstType>& dst, const UnOp& op, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(rows, cols);
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src), shrinkPtr(dst), op, WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename DstType, class UnOp>
+__host__ void gridTransform_(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const UnOp& op, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src), shrinkPtr(dst), op, WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr1, class SrcPtr2, typename DstType, class BinOp, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr1& src1, const SrcPtr2& src2, GpuMat_<DstType>& dst, const BinOp& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src1);
+    const int cols = getCols(src1);
+
+    CV_Assert( getRows(src2) == rows && getCols(src2) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    dst.create(rows, cols);
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src1), shrinkPtr(src2), shrinkPtr(dst), op, shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr1, class SrcPtr2, typename DstType, class BinOp, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr1& src1, const SrcPtr2& src2, const GlobPtrSz<DstType>& dst, const BinOp& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src1);
+    const int cols = getCols(src1);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+    CV_Assert( getRows(src2) == rows && getCols(src2) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src1), shrinkPtr(src2), shrinkPtr(dst), op, shrinkPtr(mask), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr1, class SrcPtr2, typename DstType, class BinOp>
+__host__ void gridTransform_(const SrcPtr1& src1, const SrcPtr2& src2, GpuMat_<DstType>& dst, const BinOp& op, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src1);
+    const int cols = getCols(src1);
+
+    CV_Assert( getRows(src2) == rows && getCols(src2) == cols );
+
+    dst.create(rows, cols);
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src1), shrinkPtr(src2), shrinkPtr(dst), op, WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr1, class SrcPtr2, typename DstType, class BinOp>
+__host__ void gridTransform_(const SrcPtr1& src1, const SrcPtr2& src2, GlobPtrSz<DstType>& dst, const BinOp& op, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src1);
+    const int cols = getCols(src1);
+
+    CV_Assert( getRows(dst) == rows && getCols(dst) == cols );
+    CV_Assert( getRows(src2) == rows && getCols(src2) == cols );
+
+    grid_transform_detail::transform<Policy>(shrinkPtr(src1), shrinkPtr(src2), shrinkPtr(dst), op, WithOutMask(), rows, cols, StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 2, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 3, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple, class MaskPtr>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(get<3>(dst)) == rows && getCols(get<3>(dst)) == cols );
+    CV_Assert( getRows(mask) == rows && getCols(mask) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                                   op,
+                                                   shrinkPtr(mask),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    get<0>(dst).create(rows, cols);
+    get<1>(dst).create(rows, cols);
+    get<2>(dst).create(rows, cols);
+    get<3>(dst).create(rows, cols);
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+template <class Policy, class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple>
+__host__ void gridTransform_(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    CV_StaticAssert( tuple_size<OpTuple>::value == 4, "" );
+
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    CV_Assert( getRows(get<0>(dst)) == rows && getCols(get<0>(dst)) == cols );
+    CV_Assert( getRows(get<1>(dst)) == rows && getCols(get<1>(dst)) == cols );
+    CV_Assert( getRows(get<2>(dst)) == rows && getCols(get<2>(dst)) == cols );
+    CV_Assert( getRows(get<3>(dst)) == rows && getCols(get<3>(dst)) == cols );
+
+    grid_transform_detail::transform_tuple<Policy>(shrinkPtr(src),
+                                                   shrinkPtr(zipPtr(get<0>(dst), get<1>(dst), get<2>(dst), get<3>(dst))),
+                                                   op,
+                                                   WithOutMask(),
+                                                   rows, cols,
+                                                   StreamAccessor::getStream(stream));
+}
+
+// Default Policy
+
+struct DefaultTransformPolicy
+{
+    enum {
+        block_size_x = 32,
+        block_size_y = 8,
+        shift = 4
+    };
+};
+
+template <class SrcPtr, typename DstType, class Op, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, GpuMat_<DstType>& dst, const Op& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename DstType, class Op, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const Op& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename DstType, class Op>
+__host__ void gridTransform(const SrcPtr& src, GpuMat_<DstType>& dst, const Op& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename DstType, class Op>
+__host__ void gridTransform(const SrcPtr& src, const GlobPtrSz<DstType>& dst, const Op& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr1, class SrcPtr2, typename DstType, class Op, class MaskPtr>
+__host__ void gridTransform(const SrcPtr1& src1, const SrcPtr1& src2, GpuMat_<DstType>& dst, const Op& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src1, src2, dst, op, mask, stream);
+}
+
+template <class SrcPtr1, class SrcPtr2, typename DstType, class Op, class MaskPtr>
+__host__ void gridTransform(const SrcPtr1& src1, const SrcPtr1& src2, const GlobPtrSz<DstType>& dst, const Op& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src1, src2, dst, op, mask, stream);
+}
+
+template <class SrcPtr1, class SrcPtr2, typename DstType, class Op>
+__host__ void gridTransform(const SrcPtr1& src1, const SrcPtr1& src2, GpuMat_<DstType>& dst, const Op& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src1, src2, dst, op, stream);
+}
+
+template <class SrcPtr1, class SrcPtr2, typename DstType, class Op>
+__host__ void gridTransform(const SrcPtr1& src1, const SrcPtr1& src2, const GlobPtrSz<DstType>& dst, const Op& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src1, src2, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple, class MaskPtr>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const OpTuple& op, const MaskPtr& mask, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, mask, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GpuMat_<D0>&, GpuMat_<D1>&, GpuMat_<D2>&, GpuMat_<D3>& >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+template <class SrcPtr, typename D0, typename D1, typename D2, typename D3, class OpTuple>
+__host__ void gridTransform(const SrcPtr& src, const tuple< GlobPtrSz<D0>, GlobPtrSz<D1>, GlobPtrSz<D2>, GlobPtrSz<D3> >& dst, const OpTuple& op, Stream& stream = Stream::Null())
+{
+    gridTransform_<DefaultTransformPolicy>(src, dst, op, stream);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/grid/transpose.hpp b/modules/cudev/include/opencv2/cudev/grid/transpose.hpp
new file mode 100644
index 0000000000..5e8f24868f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/grid/transpose.hpp
@@ -0,0 +1,69 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_GRID_TRANSPOSE_HPP__
+#define __OPENCV_CUDEV_GRID_TRANSPOSE_HPP__
+
+#include "../common.hpp"
+#include "../ptr2d/traits.hpp"
+#include "../ptr2d/gpumat.hpp"
+#include "detail/transpose.hpp"
+
+namespace cv { namespace cudev {
+
+template <class SrcPtr, typename DstType>
+__host__ void gridTranspose(const SrcPtr& src, GpuMat_<DstType>& dst, Stream& stream = Stream::Null())
+{
+    const int rows = getRows(src);
+    const int cols = getCols(src);
+
+    dst.create(cols, rows);
+
+    transpose_detail::transpose(shrinkPtr(src), shrinkPtr(dst), rows, cols, StreamAccessor::getStream(stream));
+}
+
+}}
+
+#endif
diff --git a/modules/video/src/simpleflow.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/constant.hpp
similarity index 67%
rename from modules/video/src/simpleflow.hpp
rename to modules/cudev/include/opencv2/cudev/ptr2d/constant.hpp
index 363fdea168..d3c56e7713 100644
--- a/modules/video/src/simpleflow.hpp
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/constant.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,47 +41,53 @@
 //
 //M*/
 
-#ifndef __OPENCV_SIMPLEFLOW_H__
-#define __OPENCV_SIMPLEFLOW_H__
+#pragma once
 
-#include <vector>
+#ifndef __OPENCV_CUDEV_PTR2D_CONSTANT_HPP__
+#define __OPENCV_CUDEV_PTR2D_CONSTANT_HPP__
 
-#define MASK_TRUE_VALUE 255
-#define UNKNOWN_FLOW_THRESH 1e9
+#include "../common.hpp"
+#include "traits.hpp"
 
-namespace cv {
+namespace cv { namespace cudev {
 
-inline static float dist(const Vec3b& p1, const Vec3b& p2) {
-  return (float)((p1[0] - p2[0]) * (p1[0] - p2[0]) +
-         (p1[1] - p2[1]) * (p1[1] - p2[1]) +
-         (p1[2] - p2[2]) * (p1[2] - p2[2]));
+template <typename T> struct ConstantPtr
+{
+    typedef T   value_type;
+    typedef int index_type;
+
+    T value;
+
+    __device__ __forceinline__ T operator ()(int, int) const { return value; }
+};
+
+template <typename T> struct ConstantPtrSz : ConstantPtr<T>
+{
+    int rows, cols;
+};
+
+template <typename T>
+__host__ ConstantPtr<T> constantPtr(T value)
+{
+    ConstantPtr<T> p;
+    p.value = value;
+    return p;
 }
 
-inline static float dist(const Vec2f& p1, const Vec2f& p2) {
-  return (p1[0] - p2[0]) * (p1[0] - p2[0]) +
-         (p1[1] - p2[1]) * (p1[1] - p2[1]);
+template <typename T> ConstantPtrSz<T>
+__host__ constantPtr(T value, int rows, int cols)
+{
+    ConstantPtrSz<T> p;
+    p.value = value;
+    p.rows = rows;
+    p.cols = cols;
+    return p;
 }
 
-inline static float dist(const Point2f& p1, const Point2f& p2) {
-  return (p1.x - p2.x) * (p1.x - p2.x) +
-         (p1.y - p2.y) * (p1.y - p2.y);
-}
+template <typename T> struct PtrTraits< ConstantPtrSz<T> > : PtrTraitsBase< ConstantPtrSz<T>, ConstantPtr<T> >
+{
+};
 
-inline static float dist(float x1, float y1, float x2, float y2) {
-  return (x1 - x2) * (x1 - x2) +
-         (y1 - y2) * (y1 - y2);
-}
-
-inline static int dist(int x1, int y1, int x2, int y2) {
-  return (x1 - x2) * (x1 - x2) +
-         (y1 - y2) * (y1 - y2);
-}
-
-template<class T>
-inline static T min(T t1, T t2, T t3) {
-  return (t1 <= t2 && t1 <= t3) ? t1 : min(t2, t3);
-}
-
-}
+}}
 
 #endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/deriv.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/deriv.hpp
new file mode 100644
index 0000000000..097007400f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/deriv.hpp
@@ -0,0 +1,393 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_DERIV_HPP__
+#define __OPENCV_CUDEV_PTR2D_DERIV_HPP__
+
+#include "../common.hpp"
+#include "../grid/copy.hpp"
+#include "traits.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+// derivX
+
+template <class SrcPtr> struct DerivXPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        return src(y, x + 1) - src(y, x - 1);
+    }
+};
+
+template <class SrcPtr> struct DerivXPtrSz : DerivXPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ DerivXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> derivXPtr(const SrcPtr& src)
+{
+    DerivXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< DerivXPtrSz<SrcPtr> > : PtrTraitsBase<DerivXPtrSz<SrcPtr>, DerivXPtr<SrcPtr> >
+{
+};
+
+// derivY
+
+template <class SrcPtr> struct DerivYPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        return src(y + 1, x) - src(y - 1, x);
+    }
+};
+
+template <class SrcPtr> struct DerivYPtrSz : DerivYPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ DerivYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> derivYPtr(const SrcPtr& src)
+{
+    DerivYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< DerivYPtrSz<SrcPtr> > : PtrTraitsBase<DerivYPtrSz<SrcPtr>, DerivYPtr<SrcPtr> >
+{
+};
+
+// sobelX
+
+template <class SrcPtr> struct SobelXPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        typename PtrTraits<SrcPtr>::value_type vals[6] =
+        {
+            src(y - 1, x - 1), src(y - 1, x + 1),
+            src(y    , x - 1), src(y    , x + 1),
+            src(y + 1, x - 1), src(y + 1, x + 1),
+        };
+
+        return (vals[1] - vals[0]) + 2 * (vals[3] - vals[2]) + (vals[5] - vals[4]);
+    }
+};
+
+template <class SrcPtr> struct SobelXPtrSz : SobelXPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ SobelXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> sobelXPtr(const SrcPtr& src)
+{
+    SobelXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< SobelXPtrSz<SrcPtr> > : PtrTraitsBase<SobelXPtrSz<SrcPtr>, SobelXPtr<SrcPtr> >
+{
+};
+
+// sobelY
+
+template <class SrcPtr> struct SobelYPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        typename PtrTraits<SrcPtr>::value_type vals[6] =
+        {
+            src(y - 1, x - 1), src(y - 1, x), src(y - 1, x + 1),
+            src(y + 1, x - 1), src(y + 1, x), src(y + 1, x + 1)
+        };
+
+        return (vals[3] - vals[0]) + 2 * (vals[4] - vals[1]) + (vals[5] - vals[2]);
+    }
+};
+
+template <class SrcPtr> struct SobelYPtrSz : SobelYPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ SobelYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> sobelYPtr(const SrcPtr& src)
+{
+    SobelYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< SobelYPtrSz<SrcPtr> > : PtrTraitsBase<SobelYPtrSz<SrcPtr>, SobelYPtr<SrcPtr> >
+{
+};
+
+// scharrX
+
+template <class SrcPtr> struct ScharrXPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        typename PtrTraits<SrcPtr>::value_type vals[6] =
+        {
+            src(y - 1, x - 1), src(y - 1, x + 1),
+            src(y    , x - 1), src(y    , x + 1),
+            src(y + 1, x - 1), src(y + 1, x + 1),
+        };
+
+        return 3 * (vals[1] - vals[0]) + 10 * (vals[3] - vals[2]) + 3 * (vals[5] - vals[4]);
+    }
+};
+
+template <class SrcPtr> struct ScharrXPtrSz : ScharrXPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ ScharrXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> scharrXPtr(const SrcPtr& src)
+{
+    ScharrXPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< ScharrXPtrSz<SrcPtr> > : PtrTraitsBase<ScharrXPtrSz<SrcPtr>, ScharrXPtr<SrcPtr> >
+{
+};
+
+// scharrY
+
+template <class SrcPtr> struct ScharrYPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        typename PtrTraits<SrcPtr>::value_type vals[6] =
+        {
+            src(y - 1, x - 1), src(y - 1, x), src(y - 1, x + 1),
+            src(y + 1, x - 1), src(y + 1, x), src(y + 1, x + 1)
+        };
+
+        return 3 * (vals[3] - vals[0]) + 10 * (vals[4] - vals[1]) + 3 * (vals[5] - vals[2]);
+    }
+};
+
+template <class SrcPtr> struct ScharrYPtrSz : ScharrYPtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ ScharrYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> scharrYPtr(const SrcPtr& src)
+{
+    ScharrYPtrSz<typename PtrTraits<SrcPtr>::ptr_type> s;
+    s.src = shrinkPtr(src);
+    s.rows = getRows(src);
+    s.cols = getCols(src);
+    return s;
+}
+
+template <class SrcPtr> struct PtrTraits< ScharrYPtrSz<SrcPtr> > : PtrTraitsBase<ScharrYPtrSz<SrcPtr>, ScharrYPtr<SrcPtr> >
+{
+};
+
+// laplacian
+
+template <int ksize, class SrcPtr> struct LaplacianPtr;
+
+template <class SrcPtr> struct LaplacianPtr<1, SrcPtr>
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        typename PtrTraits<SrcPtr>::value_type vals[5] =
+        {
+                           src(y - 1, x),
+            src(y, x - 1), src(y    , x), src(y, x + 1),
+                           src(y + 1, x)
+        };
+
+        return (vals[0] + vals[1] + vals[3] + vals[4]) - 4 * vals[2];
+    }
+};
+
+template <class SrcPtr> struct LaplacianPtr<3, SrcPtr>
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+   SrcPtr src;
+
+   __device__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+   {
+       typename PtrTraits<SrcPtr>::value_type vals[5] =
+       {
+           src(y - 1, x - 1),            src(y - 1, x + 1),
+                              src(y, x),
+           src(y + 1, x - 1),            src(y + 1, x + 1)
+       };
+
+       return 2 * (vals[0] + vals[1] + vals[3] + vals[4]) - 8 * vals[2];
+   }
+};
+
+template <int ksize, class SrcPtr> struct LaplacianPtrSz : LaplacianPtr<ksize, SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <int ksize, class SrcPtr>
+__host__ LaplacianPtrSz<ksize, typename PtrTraits<SrcPtr>::ptr_type> laplacianPtr(const SrcPtr& src)
+{
+    LaplacianPtrSz<ksize, typename PtrTraits<SrcPtr>::ptr_type> ptr;
+    ptr.src = shrinkPtr(src);
+    ptr.rows = getRows(src);
+    ptr.cols = getCols(src);
+    return ptr;
+}
+
+template <int ksize, class SrcPtr> struct PtrTraits< LaplacianPtrSz<ksize, SrcPtr> > : PtrTraitsBase<LaplacianPtrSz<ksize, SrcPtr>, LaplacianPtr<ksize, SrcPtr> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/detail/gpumat.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/detail/gpumat.hpp
new file mode 100644
index 0000000000..e378c52372
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/detail/gpumat.hpp
@@ -0,0 +1,361 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_GPUMAT_DETAIL_HPP__
+#define __OPENCV_CUDEV_PTR2D_GPUMAT_DETAIL_HPP__
+
+#include "../gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_()
+    : GpuMat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<T>::type;
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(int arows, int acols)
+    : GpuMat(arows, acols, DataType<T>::type)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(Size asize)
+    : GpuMat(asize.height, asize.width, DataType<T>::type)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(int arows, int acols, Scalar val)
+    : GpuMat(arows, acols, DataType<T>::type, val)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(Size asize, Scalar val)
+    : GpuMat(asize.height, asize.width, DataType<T>::type, val)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(const GpuMat_& m)
+    : GpuMat(m)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(const GpuMat& m)
+    : GpuMat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<T>::type;
+
+    if (DataType<T>::type == m.type())
+    {
+        GpuMat::operator =(m);
+        return;
+    }
+
+    if (DataType<T>::depth == m.depth())
+    {
+        GpuMat::operator =(m.reshape(DataType<T>::channels, m.rows));
+        return;
+    }
+
+    CV_Assert( DataType<T>::channels == m.channels() );
+    m.convertTo(*this, type());
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(int arows, int acols, T* adata, size_t astep)
+    : GpuMat(arows, acols, DataType<T>::type, adata, astep)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(Size asize, T* adata, size_t astep)
+    : GpuMat(asize.height, asize.width, DataType<T>::type, adata, astep)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(const GpuMat_& m, Range arowRange, Range acolRange)
+    : GpuMat(m, arowRange, acolRange)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(const GpuMat_& m, Rect roi)
+    : GpuMat(m, roi)
+{
+}
+
+template <typename T>
+__host__ GpuMat_<T>::GpuMat_(InputArray arr)
+    : GpuMat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<T>::type;
+    upload(arr);
+}
+
+template <typename T>
+__host__ GpuMat_<T>& GpuMat_<T>::operator =(const GpuMat_& m)
+{
+    GpuMat::operator =(m);
+    return *this;
+}
+
+template <typename T>
+__host__ void GpuMat_<T>::create(int arows, int acols)
+{
+    GpuMat::create(arows, acols, DataType<T>::type);
+}
+
+template <typename T>
+__host__ void GpuMat_<T>::create(Size asize)
+{
+    GpuMat::create(asize, DataType<T>::type);
+}
+
+template <typename T>
+__host__ void GpuMat_<T>::swap(GpuMat_& mat)
+{
+    GpuMat::swap(mat);
+}
+
+template <typename T>
+__host__ void GpuMat_<T>::upload(InputArray arr)
+{
+    CV_Assert( arr.type() == DataType<T>::type );
+    GpuMat::upload(arr);
+}
+
+template <typename T>
+__host__ void GpuMat_<T>::upload(InputArray arr, Stream& stream)
+{
+    CV_Assert( arr.type() == DataType<T>::type );
+    GpuMat::upload(arr, stream);
+}
+
+template <typename T>
+__host__ GpuMat_<T>::operator GlobPtrSz<T>() const
+{
+    return globPtr((T*) data, step, rows, cols);
+}
+
+template <typename T>
+__host__ GpuMat_<T>::operator GlobPtr<T>() const
+{
+    return globPtr((T*) data, step);
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::clone() const
+{
+    return GpuMat_(GpuMat::clone());
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::row(int y) const
+{
+    return GpuMat_(*this, Range(y, y+1), Range::all());
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::col(int x) const
+{
+    return GpuMat_(*this, Range::all(), Range(x, x+1));
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::rowRange(int startrow, int endrow) const
+{
+    return GpuMat_(*this, Range(startrow, endrow), Range::all());
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::rowRange(Range r) const
+{
+    return GpuMat_(*this, r, Range::all());
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::colRange(int startcol, int endcol) const
+{
+    return GpuMat_(*this, Range::all(), Range(startcol, endcol));
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::colRange(Range r) const
+{
+    return GpuMat_(*this, Range::all(), r);
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::operator ()(Range _rowRange, Range _colRange) const
+{
+    return GpuMat_(*this, _rowRange, _colRange);
+}
+
+template <typename T>
+__host__ GpuMat_<T> GpuMat_<T>::operator ()(Rect roi) const
+{
+    return GpuMat_(*this, roi);
+}
+
+template <typename T>
+__host__ GpuMat_<T>& GpuMat_<T>::adjustROI(int dtop, int dbottom, int dleft, int dright)
+{
+    return (GpuMat_<T>&)(GpuMat::adjustROI(dtop, dbottom, dleft, dright));
+}
+
+template <typename T>
+__host__ size_t GpuMat_<T>::elemSize() const
+{
+    CV_DbgAssert( GpuMat::elemSize() == sizeof(T) );
+    return sizeof(T);
+}
+
+template <typename T>
+__host__ size_t GpuMat_<T>::elemSize1() const
+{
+    CV_DbgAssert( GpuMat::elemSize1() == sizeof(T) / DataType<T>::channels );
+    return sizeof(T) / DataType<T>::channels;
+}
+
+template <typename T>
+__host__ int GpuMat_<T>::type() const
+{
+    CV_DbgAssert( GpuMat::type() == DataType<T>::type );
+    return DataType<T>::type;
+}
+
+template <typename T>
+__host__ int GpuMat_<T>::depth() const
+{
+    CV_DbgAssert( GpuMat::depth() == DataType<T>::depth );
+    return DataType<T>::depth;
+}
+
+template <typename T>
+__host__ int GpuMat_<T>::channels() const
+{
+    CV_DbgAssert( GpuMat::channels() == DataType<T>::channels );
+    return DataType<T>::channels;
+}
+
+template <typename T>
+__host__ size_t GpuMat_<T>::stepT() const
+{
+    return step / elemSize();
+}
+
+template <typename T>
+__host__ size_t GpuMat_<T>::step1() const
+{
+    return step / elemSize1();
+}
+
+template <typename T>
+__host__ T* GpuMat_<T>::operator [](int y)
+{
+    return (T*)ptr(y);
+}
+
+template <typename T>
+__host__ const T* GpuMat_<T>::operator [](int y) const
+{
+    return (const T*)ptr(y);
+}
+
+template <typename T> template <class Body>
+__host__ GpuMat_<T>::GpuMat_(const Expr<Body>& expr)
+    : GpuMat()
+{
+    flags = (flags & ~CV_MAT_TYPE_MASK) | DataType<T>::type;
+    *this = expr;
+}
+
+template <typename T> template <class Body>
+__host__ GpuMat_<T>& GpuMat_<T>::operator =(const Expr<Body>& expr)
+{
+    expr.body.assignTo(*this);
+    return *this;
+}
+
+template <typename T> template <class Body>
+__host__ GpuMat_<T>& GpuMat_<T>::assign(const Expr<Body>& expr, Stream& stream)
+{
+    expr.body.assignTo(*this, stream);
+    return *this;
+}
+
+}}
+
+// Input / Output Arrays
+
+namespace cv {
+
+template<typename _Tp>
+__host__ _InputArray::_InputArray(const cudev::GpuMat_<_Tp>& m)
+    : flags(FIXED_TYPE + GPU_MAT + DataType<_Tp>::type), obj((void*)&m)
+{}
+
+template<typename _Tp>
+__host__ _OutputArray::_OutputArray(cudev::GpuMat_<_Tp>& m)
+    : _InputArray(m)
+{}
+
+template<typename _Tp>
+__host__ _OutputArray::_OutputArray(const cudev::GpuMat_<_Tp>& m)
+    : _InputArray(m)
+{
+    flags |= FIXED_SIZE;
+}
+
+}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/extrapolation.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/extrapolation.hpp
new file mode 100644
index 0000000000..c90e79174f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/extrapolation.hpp
@@ -0,0 +1,219 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_EXTRAPOLATION_HPP__
+#define __OPENCV_CUDEV_PTR2D_EXTRAPOLATION_HPP__
+
+#include "../common.hpp"
+#include "../util/vec_traits.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+// BrdConstant
+
+template <class SrcPtr> struct BrdConstant
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+    int rows, cols;
+    typename PtrTraits<SrcPtr>::value_type val;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        return (x >= 0 && x < cols && y >= 0 && y < rows) ? src(y, x) : val;
+    }
+};
+
+template <class SrcPtr>
+__host__ BrdConstant<typename PtrTraits<SrcPtr>::ptr_type> brdConstant(const SrcPtr& src, typename PtrTraits<SrcPtr>::value_type val)
+{
+    BrdConstant<typename PtrTraits<SrcPtr>::ptr_type> b;
+    b.src = shrinkPtr(src);
+    b.rows = getRows(src);
+    b.cols = getCols(src);
+    b.val = val;
+    return b;
+}
+
+template <class SrcPtr>
+__host__ BrdConstant<typename PtrTraits<SrcPtr>::ptr_type> brdConstant(const SrcPtr& src)
+{
+    return brdConstant(src, VecTraits<typename PtrTraits<SrcPtr>::value_type>::all(0));
+}
+
+// BrdBase
+
+template <class BrdImpl, class SrcPtr> struct BrdBase
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef int                                    index_type;
+
+    SrcPtr src;
+    int rows, cols;
+
+    __device__ __forceinline__ int idx_row(int y) const
+    {
+        return BrdImpl::idx_low(BrdImpl::idx_high(y, rows), rows);
+    }
+
+    __device__ __forceinline__ int idx_col(int x) const
+    {
+        return BrdImpl::idx_low(BrdImpl::idx_high(x, cols), cols);
+    }
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(int y, int x) const
+    {
+        return src(idx_row(y), idx_col(x));
+    }
+};
+
+// BrdReplicate
+
+struct BrdReplicate
+{
+    __device__ __forceinline__ static int idx_low(int i, int len)
+    {
+        return ::max(i, 0);
+    }
+
+    __device__ __forceinline__ static int idx_high(int i, int len)
+    {
+        return ::min(i, len - 1);
+    }
+};
+
+template <class SrcPtr>
+__host__ BrdBase<BrdReplicate, typename PtrTraits<SrcPtr>::ptr_type> brdReplicate(const SrcPtr& src)
+{
+    BrdBase<BrdReplicate, typename PtrTraits<SrcPtr>::ptr_type> b;
+    b.src = shrinkPtr(src);
+    b.rows = getRows(src);
+    b.cols = getCols(src);
+    return b;
+}
+
+// BrdReflect101
+
+struct BrdReflect101
+{
+    __device__ __forceinline__ static int idx_low(int i, int len)
+    {
+        return ::abs(i) % len;
+    }
+
+    __device__ __forceinline__ static int idx_high(int i, int len)
+    {
+        const int last_ind = len - 1;
+        return ::abs(last_ind - ::abs(last_ind - i)) % len;
+    }
+};
+
+template <class SrcPtr>
+__host__ BrdBase<BrdReflect101, typename PtrTraits<SrcPtr>::ptr_type> brdReflect101(const SrcPtr& src)
+{
+    BrdBase<BrdReflect101, typename PtrTraits<SrcPtr>::ptr_type> b;
+    b.src = shrinkPtr(src);
+    b.rows = getRows(src);
+    b.cols = getCols(src);
+    return b;
+}
+
+// BrdReflect
+
+struct BrdReflect
+{
+    __device__ __forceinline__ static int idx_low(int i, int len)
+    {
+        return (::abs(i) - (i < 0)) % len;
+    }
+
+    __device__ __forceinline__ static int idx_high(int i, int len)
+    {
+        const int last_ind = len - 1;
+        return (last_ind - ::abs(last_ind - i) + (i > last_ind));
+    }
+};
+
+template <class SrcPtr>
+__host__ BrdBase<BrdReflect, typename PtrTraits<SrcPtr>::ptr_type> brdReflect(const SrcPtr& src)
+{
+    BrdBase<BrdReflect, typename PtrTraits<SrcPtr>::ptr_type> b;
+    b.src = shrinkPtr(src);
+    b.rows = getRows(src);
+    b.cols = getCols(src);
+    return b;
+}
+
+// BrdWrap
+
+struct BrdWrap
+{
+    __device__ __forceinline__ static int idx_low(int i, int len)
+    {
+        return (i >= 0) * i + (i < 0) * (i - ((i - len + 1) / len) * len);
+    }
+
+    __device__ __forceinline__ static int idx_high(int i, int len)
+    {
+        return (i < len) * i + (i >= len) * (i % len);
+    }
+};
+
+template <class SrcPtr>
+__host__ BrdBase<BrdWrap, typename PtrTraits<SrcPtr>::ptr_type> brdWrap(const SrcPtr& src)
+{
+    BrdBase<BrdWrap, typename PtrTraits<SrcPtr>::ptr_type> b;
+    b.src = shrinkPtr(src);
+    b.rows = getRows(src);
+    b.cols = getCols(src);
+    return b;
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/glob.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/glob.hpp
new file mode 100644
index 0000000000..7304a8c7f5
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/glob.hpp
@@ -0,0 +1,111 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_GLOB_HPP__
+#define __OPENCV_CUDEV_PTR2D_GLOB_HPP__
+
+#include "../common.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T> struct GlobPtr
+{
+    typedef T   value_type;
+    typedef int index_type;
+
+    T* data;
+    size_t step;
+
+    __device__ __forceinline__       T* row(int y)       { return (      T*)( (      uchar*)data + y * step); }
+    __device__ __forceinline__ const T* row(int y) const { return (const T*)( (const uchar*)data + y * step); }
+
+    __device__ __forceinline__       T& operator ()(int y, int x)       { return row(y)[x]; }
+    __device__ __forceinline__ const T& operator ()(int y, int x) const { return row(y)[x]; }
+};
+
+template <typename T> struct GlobPtrSz : GlobPtr<T>
+{
+    int rows, cols;
+};
+
+template <typename T>
+__host__ GlobPtr<T> globPtr(T* data, size_t step)
+{
+    GlobPtr<T> p;
+    p.data = data;
+    p.step = step;
+    return p;
+}
+
+template <typename T>
+__host__ GlobPtrSz<T> globPtr(T* data, size_t step, int rows, int cols)
+{
+    GlobPtrSz<T> p;
+    p.data = data;
+    p.step = step;
+    p.rows = rows;
+    p.cols = cols;
+    return p;
+}
+
+template <typename T>
+__host__ GlobPtrSz<T> globPtr(const GpuMat& mat)
+{
+    GlobPtrSz<T> p;
+    p.data = (T*) mat.data;
+    p.step = mat.step;
+    p.rows = mat.rows;
+    p.cols = mat.cols;
+    return p;
+}
+
+template <typename T> struct PtrTraits< GlobPtrSz<T> > : PtrTraitsBase<GlobPtrSz<T>, GlobPtr<T> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/gpumat.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/gpumat.hpp
new file mode 100644
index 0000000000..50bb0874f1
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/gpumat.hpp
@@ -0,0 +1,161 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_GPUMAT_HPP__
+#define __OPENCV_CUDEV_PTR2D_GPUMAT_HPP__
+
+#include "../common.hpp"
+#include "../util/vec_traits.hpp"
+#include "../expr/expr.hpp"
+#include "glob.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T>
+class GpuMat_ : public GpuMat
+{
+public:
+    typedef T value_type;
+
+    //! default constructor
+    __host__ GpuMat_();
+
+    //! constructs GpuMat of the specified size
+    __host__ GpuMat_(int arows, int acols);
+    __host__ explicit GpuMat_(Size asize);
+
+    //! constucts GpuMat and fills it with the specified value
+    __host__ GpuMat_(int arows, int acols, Scalar val);
+    __host__ GpuMat_(Size asize, Scalar val);
+
+    //! copy constructor
+    __host__ GpuMat_(const GpuMat_& m);
+
+    //! copy/conversion contructor. If m is of different type, it's converted
+    __host__ explicit GpuMat_(const GpuMat& m);
+
+    //! constructs a matrix on top of user-allocated data. step is in bytes(!!!), regardless of the type
+    __host__ GpuMat_(int arows, int acols, T* adata, size_t astep = Mat::AUTO_STEP);
+    __host__ GpuMat_(Size asize, T* adata, size_t astep = Mat::AUTO_STEP);
+
+    //! selects a submatrix
+    __host__ GpuMat_(const GpuMat_& m, Range arowRange, Range acolRange);
+    __host__ GpuMat_(const GpuMat_& m, Rect roi);
+
+    //! builds GpuMat from host memory (Blocking call)
+    __host__ explicit GpuMat_(InputArray arr);
+
+    //! assignment operators
+    __host__ GpuMat_& operator =(const GpuMat_& m);
+
+    //! allocates new GpuMat data unless the GpuMat already has specified size and type
+    __host__ void create(int arows, int acols);
+    __host__ void create(Size asize);
+
+    //! swaps with other smart pointer
+    __host__ void swap(GpuMat_& mat);
+
+    //! pefroms upload data to GpuMat (Blocking call)
+    __host__ void upload(InputArray arr);
+
+    //! pefroms upload data to GpuMat (Non-Blocking call)
+    __host__ void upload(InputArray arr, Stream& stream);
+
+    //! convert to GlobPtr
+    __host__ operator GlobPtrSz<T>() const;
+    __host__ operator GlobPtr<T>() const;
+
+    //! overridden forms of GpuMat::row() etc.
+    __host__ GpuMat_ clone() const;
+    __host__ GpuMat_ row(int y) const;
+    __host__ GpuMat_ col(int x) const;
+    __host__ GpuMat_ rowRange(int startrow, int endrow) const;
+    __host__ GpuMat_ rowRange(Range r) const;
+    __host__ GpuMat_ colRange(int startcol, int endcol) const;
+    __host__ GpuMat_ colRange(Range r) const;
+    __host__ GpuMat_ operator ()(Range rowRange, Range colRange) const;
+    __host__ GpuMat_ operator ()(Rect roi) const;
+    __host__ GpuMat_& adjustROI(int dtop, int dbottom, int dleft, int dright);
+
+    //! overridden forms of GpuMat::elemSize() etc.
+    __host__ size_t elemSize() const;
+    __host__ size_t elemSize1() const;
+    __host__ int type() const;
+    __host__ int depth() const;
+    __host__ int channels() const;
+    __host__ size_t step1() const;
+
+    //! returns step()/sizeof(T)
+    __host__ size_t stepT() const;
+
+    //! more convenient forms of row and element access operators
+    __host__ T* operator [](int y);
+    __host__ const T* operator [](int y) const;
+
+    //! expression templates
+    template <class Body> __host__ GpuMat_(const Expr<Body>& expr);
+    template <class Body> __host__ GpuMat_& operator =(const Expr<Body>& expr);
+    template <class Body> __host__ GpuMat_& assign(const Expr<Body>& expr, Stream& stream);
+};
+
+//! creates alternative GpuMat header for the same data, with different
+//! number of channels and/or different number of rows. see cvReshape.
+template <int cn, typename T>
+__host__ GpuMat_<typename MakeVec<typename VecTraits<T>::elem_type, cn>::type>
+reshape_(const GpuMat_<T>& mat, int rows = 0)
+{
+    GpuMat_<typename MakeVec<typename VecTraits<T>::elem_type, cn>::type> dst(mat.reshape(cn, rows));
+    return dst;
+}
+
+template <typename T> struct PtrTraits< GpuMat_<T> > : PtrTraitsBase<GpuMat_<T>, GlobPtr<T> >
+{
+};
+
+}}
+
+#include "detail/gpumat.hpp"
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/interpolation.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/interpolation.hpp
new file mode 100644
index 0000000000..e86d7191ed
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/interpolation.hpp
@@ -0,0 +1,385 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_INTERPOLATION_HPP__
+#define __OPENCV_CUDEV_PTR2D_INTERPOLATION_HPP__
+
+#include "../common.hpp"
+#include "../util/vec_traits.hpp"
+#include "../util/saturate_cast.hpp"
+#include "../util/type_traits.hpp"
+#include "../util/limits.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+// Nearest
+
+template <class SrcPtr> struct NearestInterPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef float                                  index_type;
+
+    SrcPtr src;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(float y, float x) const
+    {
+        return src(__float2int_rn(y), __float2int_rn(x));
+    }
+};
+
+template <class SrcPtr> struct NearestInterPtrSz : NearestInterPtr<SrcPtr>
+{
+    int rows, cols;
+};
+
+template <class SrcPtr>
+__host__ NearestInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> interNearest(const SrcPtr& src)
+{
+    NearestInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> i;
+    i.src = shrinkPtr(src);
+    i.rows = getRows(src);
+    i.cols = getCols(src);
+    return i;
+}
+
+template <class SrcPtr> struct PtrTraits< NearestInterPtrSz<SrcPtr> > : PtrTraitsBase<NearestInterPtrSz<SrcPtr>, NearestInterPtr<SrcPtr> >
+{
+};
+
+// Linear
+
+template <typename SrcPtr> struct LinearInterPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef float                                  index_type;
+
+    SrcPtr src;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(float y, float x) const
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        work_type out = VecTraits<work_type>::all(0);
+
+        const int x1 = __float2int_rd(x);
+        const int y1 = __float2int_rd(y);
+        const int x2 = x1 + 1;
+        const int y2 = y1 + 1;
+
+        typename PtrTraits<SrcPtr>::value_type src_reg = src(y1, x1);
+        out = out + src_reg * static_cast<work_elem_type>((x2 - x) * (y2 - y));
+
+        src_reg = src(y1, x2);
+        out = out + src_reg * static_cast<work_elem_type>((x - x1) * (y2 - y));
+
+        src_reg = src(y2, x1);
+        out = out + src_reg * static_cast<work_elem_type>((x2 - x) * (y - y1));
+
+        src_reg = src(y2, x2);
+        out = out + src_reg * static_cast<work_elem_type>((x - x1) * (y - y1));
+
+        return saturate_cast<typename PtrTraits<SrcPtr>::value_type>(out);
+    }
+};
+
+template <class SrcPtr> struct LinearInterPtrSz : LinearInterPtr<SrcPtr>
+{
+    int rows, cols;
+};
+
+template <class SrcPtr>
+__host__ LinearInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> interLinear(const SrcPtr& src)
+{
+    LinearInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> i;
+    i.src = shrinkPtr(src);
+    i.rows = getRows(src);
+    i.cols = getCols(src);
+    return i;
+}
+
+template <class SrcPtr> struct PtrTraits< LinearInterPtrSz<SrcPtr> > : PtrTraitsBase<LinearInterPtrSz<SrcPtr>, LinearInterPtr<SrcPtr> >
+{
+};
+
+// Cubic
+
+template <typename SrcPtr> struct CubicInterPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef float                                  index_type;
+
+    SrcPtr src;
+
+    __device__ static float bicubicCoeff(float x_)
+    {
+        float x = ::fabsf(x_);
+        if (x <= 1.0f)
+        {
+            return x * x * (1.5f * x - 2.5f) + 1.0f;
+        }
+        else if (x < 2.0f)
+        {
+            return x * (x * (-0.5f * x + 2.5f) - 4.0f) + 2.0f;
+        }
+        else
+        {
+            return 0.0f;
+        }
+    }
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(float y, float x) const
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        const float xmin = ::ceilf(x - 2.0f);
+        const float xmax = ::floorf(x + 2.0f);
+
+        const float ymin = ::ceilf(y - 2.0f);
+        const float ymax = ::floorf(y + 2.0f);
+
+        work_type sum = VecTraits<work_type>::all(0);
+        float wsum = 0.0f;
+
+        for (float cy = ymin; cy <= ymax; cy += 1.0f)
+        {
+            for (float cx = xmin; cx <= xmax; cx += 1.0f)
+            {
+                typename PtrTraits<SrcPtr>::value_type src_reg = src(__float2int_rd(cy), __float2int_rd(cx));
+                const float w = bicubicCoeff(x - cx) * bicubicCoeff(y - cy);
+
+                sum = sum + static_cast<work_elem_type>(w) * src_reg;
+                wsum += w;
+            }
+        }
+
+        work_type res = (wsum > numeric_limits<float>::epsilon()) ? VecTraits<work_type>::all(0) : sum / static_cast<work_elem_type>(wsum);
+
+        return saturate_cast<typename PtrTraits<SrcPtr>::value_type>(res);
+    }
+};
+
+template <class SrcPtr> struct CubicInterPtrSz : CubicInterPtr<SrcPtr>
+{
+    int rows, cols;
+};
+
+template <class SrcPtr>
+__host__ CubicInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> interCubic(const SrcPtr& src)
+{
+    CubicInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> i;
+    i.src = shrinkPtr(src);
+    i.rows = getRows(src);
+    i.cols = getCols(src);
+    return i;
+}
+
+template <class SrcPtr> struct PtrTraits< CubicInterPtrSz<SrcPtr> > : PtrTraitsBase<CubicInterPtrSz<SrcPtr>, CubicInterPtr<SrcPtr> >
+{
+};
+
+// IntegerArea
+
+template <typename SrcPtr> struct IntegerAreaInterPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef float                                  index_type;
+
+    SrcPtr src;
+    int area_width, area_height;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(float y, float x) const
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        const int sx1 = __float2int_rd(x);
+        const int sx2 = sx1 + area_width;
+
+        const int sy1 = __float2int_rd(y);
+        const int sy2 = sy1 + area_height;
+
+        work_type out = VecTraits<work_type>::all(0);
+
+        for (int dy = sy1; dy < sy2; ++dy)
+        {
+            for (int dx = sx1; dx < sx2; ++dx)
+            {
+                out = out + saturate_cast<work_type>(src(dy, dx));
+            }
+        }
+
+        const work_elem_type scale = 1.0f / (area_width * area_height);
+
+        return saturate_cast<typename PtrTraits<SrcPtr>::value_type>(out * scale);
+    }
+};
+
+template <class SrcPtr> struct IntegerAreaInterPtrSz : IntegerAreaInterPtr<SrcPtr>
+{
+    int rows, cols;
+};
+
+template <class SrcPtr>
+__host__ IntegerAreaInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> interArea(const SrcPtr& src, Size areaSize)
+{
+    IntegerAreaInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> i;
+    i.src = shrinkPtr(src);
+    i.area_width = areaSize.width;
+    i.area_height = areaSize.height;
+    i.rows = getRows(src);
+    i.cols = getCols(src);
+    return i;
+}
+
+template <class SrcPtr> struct PtrTraits< IntegerAreaInterPtrSz<SrcPtr> > : PtrTraitsBase<IntegerAreaInterPtrSz<SrcPtr>, IntegerAreaInterPtr<SrcPtr> >
+{
+};
+
+// CommonArea
+
+template <typename SrcPtr> struct CommonAreaInterPtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef float                                  index_type;
+
+    SrcPtr src;
+    float area_width, area_height;
+
+    __device__ typename PtrTraits<SrcPtr>::value_type operator ()(float y, float x) const
+    {
+        typedef typename PtrTraits<SrcPtr>::value_type src_type;
+        typedef typename VecTraits<src_type>::elem_type src_elem_type;
+        typedef typename LargerType<float, src_elem_type>::type work_elem_type;
+        typedef typename MakeVec<work_elem_type, VecTraits<src_type>::cn>::type work_type;
+
+        const float fsx1 = x;
+        const float fsx2 = fsx1 + area_width;
+
+        const int sx1 = __float2int_rd(fsx1);
+        const int sx2 = __float2int_ru(fsx2);
+
+        const float fsy1 = y;
+        const float fsy2 = fsy1 + area_height;
+
+        const int sy1 = __float2int_rd(fsy1);
+        const int sy2 = __float2int_ru(fsy2);
+
+        work_type out = VecTraits<work_type>::all(0);
+
+        for (int dy = sy1; dy < sy2; ++dy)
+        {
+            for (int dx = sx1; dx < sx2; ++dx)
+                out = out + saturate_cast<work_type>(src(dy, dx));
+
+            if (sx1 > fsx1)
+                out = out + saturate_cast<work_type>(src(dy, sx1 - 1)) * static_cast<work_elem_type>(sx1 - fsx1);
+
+            if (sx2 < fsx2)
+                out = out + saturate_cast<work_type>(src(dy, sx2)) * static_cast<work_elem_type>(fsx2 - sx2);
+        }
+
+        if (sy1 > fsy1)
+        {
+            for (int dx = sx1; dx < sx2; ++dx)
+                out = out + saturate_cast<work_type>(src(sy1 - 1, dx)) * static_cast<work_elem_type>(sy1 - fsy1);
+        }
+
+        if (sy2 < fsy2)
+        {
+            for (int dx = sx1; dx < sx2; ++dx)
+                out = out + saturate_cast<work_type>(src(sy2, dx)) * static_cast<work_elem_type>(fsy2 - sy2);
+        }
+
+        if ((sy1 > fsy1) && (sx1 > fsx1))
+            out = out + saturate_cast<work_type>(src(sy1 - 1, sx1 - 1)) * static_cast<work_elem_type>((sy1 - fsy1) * (sx1 - fsx1));
+
+        if ((sy1 > fsy1) && (sx2 < fsx2))
+            out = out + saturate_cast<work_type>(src(sy1 - 1, sx2)) * static_cast<work_elem_type>((sy1 - fsy1) * (fsx2 - sx2));
+
+        if ((sy2 < fsy2) && (sx2 < fsx2))
+            out = out + saturate_cast<work_type>(src(sy2, sx2)) * static_cast<work_elem_type>((fsy2 - sy2) * (fsx2 - sx2));
+
+        if ((sy2 < fsy2) && (sx1 > fsx1))
+            out = out + saturate_cast<work_type>(src(sy2, sx1 - 1)) * static_cast<work_elem_type>((fsy2 - sy2) * (sx1 - fsx1));
+
+        const work_elem_type scale = 1.0f / (area_width * area_height);
+
+        return saturate_cast<typename PtrTraits<SrcPtr>::value_type>(out * scale);
+    }
+};
+
+template <class SrcPtr> struct CommonAreaInterPtrSz : CommonAreaInterPtr<SrcPtr>
+{
+    int rows, cols;
+};
+
+template <class SrcPtr>
+__host__ CommonAreaInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> interArea(const SrcPtr& src, Size2f areaSize)
+{
+    CommonAreaInterPtrSz<typename PtrTraits<SrcPtr>::ptr_type> i;
+    i.src = shrinkPtr(src);
+    i.area_width = areaSize.width;
+    i.area_height = areaSize.height;
+    i.rows = getRows(src);
+    i.cols = getCols(src);
+    return i;
+}
+
+template <class SrcPtr> struct PtrTraits< CommonAreaInterPtrSz<SrcPtr> > : PtrTraitsBase<CommonAreaInterPtrSz<SrcPtr>, CommonAreaInterPtr<SrcPtr> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/lut.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/lut.hpp
new file mode 100644
index 0000000000..e47719c6ec
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/lut.hpp
@@ -0,0 +1,100 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_LUT_HPP__
+#define __OPENCV_CUDEV_PTR2D_LUT_HPP__
+
+#include "../common.hpp"
+#include "../grid/copy.hpp"
+#include "traits.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+template <class SrcPtr, class TablePtr> struct LutPtr
+{
+    typedef typename PtrTraits<TablePtr>::value_type value_type;
+    typedef typename PtrTraits<SrcPtr>::index_type   index_type;
+
+    SrcPtr src;
+    TablePtr tbl;
+
+    __device__ __forceinline__ typename PtrTraits<TablePtr>::value_type operator ()(typename PtrTraits<SrcPtr>::index_type y, typename PtrTraits<SrcPtr>::index_type x) const
+    {
+        return tbl(0, src(y, x));
+    }
+};
+
+template <class SrcPtr, class TablePtr> struct LutPtrSz : LutPtr<SrcPtr, TablePtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr, class TablePtr>
+__host__ LutPtrSz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<TablePtr>::ptr_type> lutPtr(const SrcPtr& src, const TablePtr& tbl)
+{
+    CV_Assert( getRows(tbl) == 1 );
+
+    LutPtrSz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<TablePtr>::ptr_type> ptr;
+    ptr.src = shrinkPtr(src);
+    ptr.tbl = shrinkPtr(tbl);
+    ptr.rows = getRows(src);
+    ptr.cols = getCols(src);
+    return ptr;
+}
+
+template <class SrcPtr, class TablePtr> struct PtrTraits< LutPtrSz<SrcPtr, TablePtr> > : PtrTraitsBase<LutPtrSz<SrcPtr, TablePtr>, LutPtr<SrcPtr, TablePtr> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/mask.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/mask.hpp
new file mode 100644
index 0000000000..1ff3371931
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/mask.hpp
@@ -0,0 +1,67 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_MASK_HPP__
+#define __OPENCV_CUDEV_PTR2D_MASK_HPP__
+
+#include "../common.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+struct WithOutMask
+{
+    typedef bool value_type;
+    typedef int  index_type;
+
+    __device__ __forceinline__ bool operator ()(int, int) const
+    {
+        return true;
+    }
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/remap.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/remap.hpp
new file mode 100644
index 0000000000..db2669a40a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/remap.hpp
@@ -0,0 +1,154 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_REMAP_HPP__
+#define __OPENCV_CUDEV_PTR2D_REMAP_HPP__
+
+#include "opencv2/core/base.hpp"
+#include "../common.hpp"
+#include "../grid/copy.hpp"
+#include "traits.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+template <class SrcPtr, class MapPtr> struct RemapPtr1
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef typename PtrTraits<MapPtr>::index_type index_type;
+
+    SrcPtr src;
+    MapPtr map;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(typename PtrTraits<MapPtr>::index_type y, typename PtrTraits<MapPtr>::index_type x) const
+    {
+        const typename PtrTraits<MapPtr>::value_type coord = map(y, x);
+        return src(coord.y, coord.x);
+    }
+};
+
+template <class SrcPtr, class MapXPtr, class MapYPtr> struct RemapPtr2
+{
+    typedef typename PtrTraits<SrcPtr>::value_type  value_type;
+    typedef typename PtrTraits<MapXPtr>::index_type index_type;
+
+    SrcPtr src;
+    MapXPtr mapx;
+    MapYPtr mapy;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(typename PtrTraits<MapXPtr>::index_type y, typename PtrTraits<MapXPtr>::index_type x) const
+    {
+        const typename PtrTraits<MapXPtr>::value_type nx = mapx(y, x);
+        const typename PtrTraits<MapYPtr>::value_type ny = mapy(y, x);
+        return src(ny, nx);
+    }
+};
+
+template <class SrcPtr, class MapPtr> struct RemapPtr1Sz : RemapPtr1<SrcPtr, MapPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr, class MapXPtr, class MapYPtr> struct RemapPtr2Sz : RemapPtr2<SrcPtr, MapXPtr, MapYPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr, class MapPtr>
+__host__ RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapPtr>::ptr_type>
+remapPtr(const SrcPtr& src, const MapPtr& map)
+{
+    const int rows = getRows(map);
+    const int cols = getCols(map);
+
+    RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapPtr>::ptr_type> r;
+    r.src = shrinkPtr(src);
+    r.map = shrinkPtr(map);
+    r.rows = rows;
+    r.cols = cols;
+    return r;
+}
+
+template <class SrcPtr, class MapXPtr, class MapYPtr>
+__host__ RemapPtr2Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapXPtr>::ptr_type, typename PtrTraits<MapYPtr>::ptr_type>
+remapPtr(const SrcPtr& src, const MapXPtr& mapx, const MapYPtr& mapy)
+{
+    const int rows = getRows(mapx);
+    const int cols = getCols(mapx);
+
+    CV_Assert( getRows(mapy) == rows && getCols(mapy) == cols );
+
+    RemapPtr2Sz<typename PtrTraits<SrcPtr>::ptr_type, typename PtrTraits<MapXPtr>::ptr_type, typename PtrTraits<MapYPtr>::ptr_type> r;
+    r.src = shrinkPtr(src);
+    r.mapx = shrinkPtr(mapx);
+    r.mapy = shrinkPtr(mapy);
+    r.rows = rows;
+    r.cols = cols;
+    return r;
+}
+
+template <class SrcPtr, class MapPtr> struct PtrTraits< RemapPtr1Sz<SrcPtr, MapPtr> > : PtrTraitsBase<RemapPtr1Sz<SrcPtr, MapPtr>, RemapPtr1<SrcPtr, MapPtr> >
+{
+};
+
+template <class SrcPtr, class MapXPtr, class MapYPtr> struct PtrTraits< RemapPtr2Sz<SrcPtr, MapXPtr, MapYPtr> > : PtrTraitsBase<RemapPtr2Sz<SrcPtr, MapXPtr, MapYPtr>, RemapPtr2<SrcPtr, MapXPtr, MapYPtr> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/resize.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/resize.hpp
new file mode 100644
index 0000000000..10a4bad906
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/resize.hpp
@@ -0,0 +1,103 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_RESIZE_HPP__
+#define __OPENCV_CUDEV_PTR2D_RESIZE_HPP__
+
+#include "opencv2/core/base.hpp"
+#include "../common.hpp"
+#include "../grid/copy.hpp"
+#include "traits.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+template <class SrcPtr> struct ResizePtr
+{
+    typedef typename PtrTraits<SrcPtr>::value_type value_type;
+    typedef typename PtrTraits<SrcPtr>::index_type index_type;
+
+    SrcPtr src;
+    float fx, fy;
+
+    __device__ __forceinline__ typename PtrTraits<SrcPtr>::value_type operator ()(typename PtrTraits<SrcPtr>::index_type y, typename PtrTraits<SrcPtr>::index_type x) const
+    {
+        const float yn = static_cast<float>(y * fy);
+        const float xn = static_cast<float>(x * fx);
+
+        return src(yn, xn);
+    }
+};
+
+template <class SrcPtr> struct ResizePtrSz : ResizePtr<SrcPtr>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr>
+__host__ ResizePtrSz<typename PtrTraits<SrcPtr>::ptr_type> resizePtr(const SrcPtr& src, float fx, float fy)
+{
+    ResizePtrSz<typename PtrTraits<SrcPtr>::ptr_type> r;
+    r.src = shrinkPtr(src);
+    r.fx = 1.0f / fx;
+    r.fy = 1.0f / fy;
+    r.rows = cv::saturate_cast<int>(getRows(src) * fy);
+    r.cols = cv::saturate_cast<int>(getCols(src) * fx);
+    return r;
+}
+
+template <class SrcPtr> struct PtrTraits< ResizePtrSz<SrcPtr> > : PtrTraitsBase<ResizePtrSz<SrcPtr>, ResizePtr<SrcPtr> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/texture.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/texture.hpp
new file mode 100644
index 0000000000..b01a2c7b84
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/texture.hpp
@@ -0,0 +1,148 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_TEXTURE_HPP__
+#define __OPENCV_CUDEV_PTR2D_TEXTURE_HPP__
+
+#include <cstring>
+#include "../common.hpp"
+#include "glob.hpp"
+#include "gpumat.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T> struct TexturePtr
+{
+    typedef T     value_type;
+    typedef float index_type;
+
+    cudaTextureObject_t texObj;
+
+    __device__ __forceinline__ T operator ()(float y, float x) const
+    {
+        return tex2D<T>(texObj, x, y);
+    }
+};
+
+template <typename T> struct Texture : TexturePtr<T>
+{
+    int rows, cols;
+
+    __host__ explicit Texture(const GlobPtrSz<T>& mat,
+                              bool normalizedCoords = false,
+                              cudaTextureFilterMode filterMode = cudaFilterModePoint,
+                              cudaTextureAddressMode addressMode = cudaAddressModeClamp)
+    {
+        CV_Assert( deviceSupports(FEATURE_SET_COMPUTE_30) );
+
+        rows = mat.rows;
+        cols = mat.cols;
+
+        cudaResourceDesc texRes;
+        std::memset(&texRes, 0, sizeof(texRes));
+        texRes.resType = cudaResourceTypePitch2D;
+        texRes.res.pitch2D.devPtr = mat.data;
+        texRes.res.pitch2D.height = mat.rows;
+        texRes.res.pitch2D.width = mat.cols;
+        texRes.res.pitch2D.pitchInBytes = mat.step;
+        texRes.res.pitch2D.desc = cudaCreateChannelDesc<T>();
+
+        cudaTextureDesc texDescr;
+        std::memset(&texDescr, 0, sizeof(texDescr));
+        texDescr.addressMode[0] = addressMode;
+        texDescr.addressMode[1] = addressMode;
+        texDescr.addressMode[2] = addressMode;
+        texDescr.filterMode = filterMode;
+        texDescr.readMode = cudaReadModeElementType;
+        texDescr.normalizedCoords = normalizedCoords;
+
+        CV_CUDEV_SAFE_CALL( cudaCreateTextureObject(&this->texObj, &texRes, &texDescr, 0) );
+    }
+
+    __host__ explicit Texture(const GpuMat_<T>& mat,
+                              bool normalizedCoords = false,
+                              cudaTextureFilterMode filterMode = cudaFilterModePoint,
+                              cudaTextureAddressMode addressMode = cudaAddressModeClamp)
+    {
+        CV_Assert( deviceSupports(FEATURE_SET_COMPUTE_30) );
+
+        rows = mat.rows;
+        cols = mat.cols;
+
+        cudaResourceDesc texRes;
+        std::memset(&texRes, 0, sizeof(texRes));
+        texRes.resType = cudaResourceTypePitch2D;
+        texRes.res.pitch2D.devPtr = mat.data;
+        texRes.res.pitch2D.height = mat.rows;
+        texRes.res.pitch2D.width = mat.cols;
+        texRes.res.pitch2D.pitchInBytes = mat.step;
+        texRes.res.pitch2D.desc = cudaCreateChannelDesc<T>();
+
+        cudaTextureDesc texDescr;
+        std::memset(&texDescr, 0, sizeof(texDescr));
+        texDescr.addressMode[0] = addressMode;
+        texDescr.addressMode[1] = addressMode;
+        texDescr.addressMode[2] = addressMode;
+        texDescr.filterMode = filterMode;
+        texDescr.readMode = cudaReadModeElementType;
+        texDescr.normalizedCoords = normalizedCoords;
+
+        CV_CUDEV_SAFE_CALL( cudaCreateTextureObject(&this->texObj, &texRes, &texDescr, 0) );
+    }
+
+    __host__ ~Texture()
+    {
+        cudaDestroyTextureObject(this->texObj);
+    }
+};
+
+template <typename T> struct PtrTraits< Texture<T> > : PtrTraitsBase<Texture<T>, TexturePtr<T> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/traits.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/traits.hpp
new file mode 100644
index 0000000000..7fb4b32b17
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/traits.hpp
@@ -0,0 +1,101 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_TRAITS_HPP__
+#define __OPENCV_CUDEV_PTR2D_TRAITS_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+template <class Ptr2DSz, class Ptr2D> struct PtrTraitsBase
+{
+    typedef Ptr2DSz ptr_sz_type;
+    typedef Ptr2D   ptr_type;
+
+    typedef typename Ptr2D::value_type value_type;
+    typedef typename Ptr2D::index_type index_type;
+
+    __host__ static Ptr2D shrinkPtr(const Ptr2DSz& ptr)
+    {
+        return ptr;
+    }
+
+    __host__ static int getRows(const Ptr2DSz& ptr)
+    {
+        return ptr.rows;
+    }
+
+    __host__ static int getCols(const Ptr2DSz& ptr)
+    {
+        return ptr.cols;
+    }
+};
+
+template <class Ptr2DSz> struct PtrTraits : PtrTraitsBase<Ptr2DSz, Ptr2DSz>
+{
+};
+
+template <class Ptr2DSz>
+__host__ typename PtrTraits<Ptr2DSz>::ptr_type shrinkPtr(const Ptr2DSz& ptr)
+{
+    return PtrTraits<Ptr2DSz>::shrinkPtr(ptr);
+}
+
+template <class Ptr2DSz>
+__host__ int getRows(const Ptr2DSz& ptr)
+{
+    return PtrTraits<Ptr2DSz>::getRows(ptr);
+}
+
+template <class Ptr2DSz>
+__host__ int getCols(const Ptr2DSz& ptr)
+{
+    return PtrTraits<Ptr2DSz>::getCols(ptr);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/transform.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/transform.hpp
new file mode 100644
index 0000000000..f540e75213
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/transform.hpp
@@ -0,0 +1,151 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_TRANSFORM_HPP__
+#define __OPENCV_CUDEV_PTR2D_TRANSFORM_HPP__
+
+#include "../common.hpp"
+#include "../grid/copy.hpp"
+#include "traits.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+// UnaryTransformPtr
+
+template <class SrcPtr, class Op> struct UnaryTransformPtr
+{
+    typedef typename Op::result_type               value_type;
+    typedef typename PtrTraits<SrcPtr>::index_type index_type;
+
+    SrcPtr src;
+    Op op;
+
+    __device__ __forceinline__ typename Op::result_type operator ()(typename PtrTraits<SrcPtr>::index_type y, typename PtrTraits<SrcPtr>::index_type x) const
+    {
+        return op(src(y, x));
+    }
+};
+
+template <class SrcPtr, class Op> struct UnaryTransformPtrSz : UnaryTransformPtr<SrcPtr, Op>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class SrcPtr, class Op>
+__host__ UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, Op>
+transformPtr(const SrcPtr& src, const Op& op)
+{
+    UnaryTransformPtrSz<typename PtrTraits<SrcPtr>::ptr_type, Op> ptr;
+    ptr.src = shrinkPtr(src);
+    ptr.op = op;
+    ptr.rows = getRows(src);
+    ptr.cols = getCols(src);
+    return ptr;
+}
+
+template <class SrcPtr, class Op> struct PtrTraits< UnaryTransformPtrSz<SrcPtr, Op> > : PtrTraitsBase<UnaryTransformPtrSz<SrcPtr, Op>, UnaryTransformPtr<SrcPtr, Op> >
+{
+};
+
+// BinaryTransformPtr
+
+template <class Src1Ptr, class Src2Ptr, class Op> struct BinaryTransformPtr
+{
+    typedef typename Op::result_type                value_type;
+    typedef typename PtrTraits<Src1Ptr>::index_type index_type;
+
+    Src1Ptr src1;
+    Src2Ptr src2;
+    Op op;
+
+    __device__ __forceinline__ typename Op::result_type operator ()(typename PtrTraits<Src1Ptr>::index_type y, typename PtrTraits<Src1Ptr>::index_type x) const
+    {
+        return op(src1(y, x), src2(y, x));
+    }
+};
+
+template <class Src1Ptr, class Src2Ptr, class Op> struct BinaryTransformPtrSz : BinaryTransformPtr<Src1Ptr, Src2Ptr, Op>
+{
+    int rows, cols;
+
+    template <typename T>
+    __host__ void assignTo(GpuMat_<T>& dst, Stream& stream = Stream::Null()) const
+    {
+        gridCopy(*this, dst, stream);
+    }
+};
+
+template <class Src1Ptr, class Src2Ptr, class Op>
+__host__ BinaryTransformPtrSz<typename PtrTraits<Src1Ptr>::ptr_type, typename PtrTraits<Src2Ptr>::ptr_type, Op>
+transformPtr(const Src1Ptr& src1, const Src2Ptr& src2, const Op& op)
+{
+    const int rows = getRows(src1);
+    const int cols = getCols(src1);
+
+    CV_Assert( getRows(src2) == rows && getCols(src2) == cols );
+
+    BinaryTransformPtrSz<typename PtrTraits<Src1Ptr>::ptr_type, typename PtrTraits<Src2Ptr>::ptr_type, Op> ptr;
+    ptr.src1 = shrinkPtr(src1);
+    ptr.src2 = shrinkPtr(src2);
+    ptr.op = op;
+    ptr.rows = rows;
+    ptr.cols = cols;
+    return ptr;
+}
+
+template <class Src1Ptr, class Src2Ptr, class Op> struct PtrTraits< BinaryTransformPtrSz<Src1Ptr, Src2Ptr, Op> > : PtrTraitsBase<BinaryTransformPtrSz<Src1Ptr, Src2Ptr, Op>, BinaryTransformPtr<Src1Ptr, Src2Ptr, Op> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/warping.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/warping.hpp
new file mode 100644
index 0000000000..80e5fbeef0
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/warping.hpp
@@ -0,0 +1,152 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_WARPING_HPP__
+#define __OPENCV_CUDEV_PTR2D_WARPING_HPP__
+
+#include "../common.hpp"
+#include "traits.hpp"
+#include "remap.hpp"
+#include "gpumat.hpp"
+
+namespace cv { namespace cudev {
+
+// affine
+
+struct AffineMapPtr
+{
+    typedef float2 value_type;
+    typedef float  index_type;
+
+    const float* warpMat;
+
+    __device__ __forceinline__ float2 operator ()(float y, float x) const
+    {
+        const float xcoo = warpMat[0] * x + warpMat[1] * y + warpMat[2];
+        const float ycoo = warpMat[3] * x + warpMat[4] * y + warpMat[5];
+
+        return make_float2(xcoo, ycoo);
+    }
+};
+
+struct AffineMapPtrSz : AffineMapPtr
+{
+    int rows, cols;
+};
+
+template <> struct PtrTraits<AffineMapPtrSz> : PtrTraitsBase<AffineMapPtrSz, AffineMapPtr>
+{
+};
+
+__host__ static AffineMapPtrSz affineMap(Size dstSize, const GpuMat_<float>& warpMat)
+{
+    CV_Assert( warpMat.rows == 2 && warpMat.cols == 3 );
+    CV_Assert( warpMat.isContinuous() );
+
+    AffineMapPtrSz map;
+    map.warpMat = warpMat[0];
+    map.rows = dstSize.height;
+    map.cols = dstSize.width;
+    return map;
+}
+
+template <class SrcPtr>
+__host__ RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, AffineMapPtr>
+warpAffinePtr(const SrcPtr& src, Size dstSize, const GpuMat_<float>& warpMat)
+{
+    return remapPtr(src, affineMap(dstSize, warpMat));
+}
+
+// perspective
+
+struct PerspectiveMapPtr
+{
+    typedef float2 value_type;
+    typedef float  index_type;
+
+    const float* warpMat;
+
+    __device__ __forceinline__ float2 operator ()(float y, float x) const
+    {
+        const float coeff = 1.0f / (warpMat[6] * x + warpMat[7] * y + warpMat[8]);
+
+        const float xcoo = coeff * (warpMat[0] * x + warpMat[1] * y + warpMat[2]);
+        const float ycoo = coeff * (warpMat[3] * x + warpMat[4] * y + warpMat[5]);
+
+        return make_float2(xcoo, ycoo);
+    }
+};
+
+struct PerspectiveMapPtrSz : PerspectiveMapPtr
+{
+    int rows, cols;
+};
+
+template <> struct PtrTraits<PerspectiveMapPtrSz> : PtrTraitsBase<PerspectiveMapPtrSz, PerspectiveMapPtr>
+{
+};
+
+__host__ static PerspectiveMapPtrSz perspectiveMap(Size dstSize, const GpuMat_<float>& warpMat)
+{
+    CV_Assert( warpMat.rows == 3 && warpMat.cols == 3 );
+    CV_Assert( warpMat.isContinuous() );
+
+    PerspectiveMapPtrSz map;
+    map.warpMat = warpMat[0];
+    map.rows = dstSize.height;
+    map.cols = dstSize.width;
+    return map;
+}
+
+template <class SrcPtr>
+__host__ RemapPtr1Sz<typename PtrTraits<SrcPtr>::ptr_type, PerspectiveMapPtr>
+warpPerspectivePtr(const SrcPtr& src, Size dstSize, const GpuMat_<float>& warpMat)
+{
+    return remapPtr(src, perspectiveMap(dstSize, warpMat));
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/ptr2d/zip.hpp b/modules/cudev/include/opencv2/cudev/ptr2d/zip.hpp
new file mode 100644
index 0000000000..934939f624
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/ptr2d/zip.hpp
@@ -0,0 +1,173 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_PTR2D_ZIP_HPP__
+#define __OPENCV_CUDEV_PTR2D_ZIP_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "traits.hpp"
+
+namespace cv { namespace cudev {
+
+template <class PtrTuple> struct ZipPtr;
+
+template <class Ptr0, class Ptr1> struct ZipPtr< tuple<Ptr0, Ptr1> > : tuple<Ptr0, Ptr1>
+{
+    typedef tuple<typename PtrTraits<Ptr0>::value_type,
+                  typename PtrTraits<Ptr1>::value_type> value_type;
+    typedef typename PtrTraits<Ptr0>::index_type        index_type;
+
+    __host__ __device__ __forceinline__ ZipPtr() {}
+    __host__ __device__ __forceinline__ ZipPtr(const tuple<Ptr0, Ptr1>& t) : tuple<Ptr0, Ptr1>(t) {}
+
+    __device__ __forceinline__ value_type operator ()(index_type y, index_type x) const
+    {
+        return make_tuple(cv::cudev::get<0>(*this)(y, x), cv::cudev::get<1>(*this)(y, x));
+    }
+};
+
+template <class Ptr0, class Ptr1, class Ptr2> struct ZipPtr< tuple<Ptr0, Ptr1, Ptr2> > : tuple<Ptr0, Ptr1, Ptr2>
+{
+    typedef tuple<typename PtrTraits<Ptr0>::value_type,
+                  typename PtrTraits<Ptr1>::value_type,
+                  typename PtrTraits<Ptr2>::value_type> value_type;
+    typedef typename PtrTraits<Ptr0>::index_type        index_type;
+
+    __host__ __device__ __forceinline__ ZipPtr() {}
+    __host__ __device__ __forceinline__ ZipPtr(const tuple<Ptr0, Ptr1, Ptr2>& t) : tuple<Ptr0, Ptr1, Ptr2>(t) {}
+
+    __device__ __forceinline__ value_type operator ()(index_type y, index_type x) const
+    {
+        return make_tuple(cv::cudev::get<0>(*this)(y, x), cv::cudev::get<1>(*this)(y, x), cv::cudev::get<2>(*this)(y, x));
+    }
+};
+
+template <class Ptr0, class Ptr1, class Ptr2, class Ptr3> struct ZipPtr< tuple<Ptr0, Ptr1, Ptr2, Ptr3> > : tuple<Ptr0, Ptr1, Ptr2, Ptr3>
+{
+    typedef tuple<typename PtrTraits<Ptr0>::value_type,
+                  typename PtrTraits<Ptr1>::value_type,
+                  typename PtrTraits<Ptr2>::value_type,
+                  typename PtrTraits<Ptr3>::value_type> value_type;
+    typedef typename PtrTraits<Ptr0>::index_type        index_type;
+
+    __host__ __device__ __forceinline__ ZipPtr() {}
+    __host__ __device__ __forceinline__ ZipPtr(const tuple<Ptr0, Ptr1, Ptr2, Ptr3>& t) : tuple<Ptr0, Ptr1, Ptr2, Ptr3>(t) {}
+
+    __device__ __forceinline__ value_type operator ()(index_type y, index_type x) const
+    {
+        return make_tuple(cv::cudev::get<0>(*this)(y, x), cv::cudev::get<1>(*this)(y, x), cv::cudev::get<2>(*this)(y, x), cv::cudev::get<3>(*this)(y, x));
+    }
+};
+
+template <class PtrTuple> struct ZipPtrSz : ZipPtr<PtrTuple>
+{
+    int rows, cols;
+
+    __host__ __device__ __forceinline__ ZipPtrSz() {}
+    __host__ __device__ __forceinline__ ZipPtrSz(const PtrTuple& t) : ZipPtr<PtrTuple>(t) {}
+};
+
+template <class Ptr0, class Ptr1>
+__host__ ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type> >
+zipPtr(const Ptr0& ptr0, const Ptr1& ptr1)
+{
+    const int rows = getRows(ptr0);
+    const int cols = getCols(ptr0);
+
+    CV_Assert( getRows(ptr1) == rows && getCols(ptr1) == cols );
+
+    ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type> >
+            z(make_tuple(shrinkPtr(ptr0), shrinkPtr(ptr1)));
+    z.rows = rows;
+    z.cols = cols;
+
+    return z;
+}
+
+template <class Ptr0, class Ptr1, class Ptr2>
+__host__ ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type, typename PtrTraits<Ptr2>::ptr_type> >
+zipPtr(const Ptr0& ptr0, const Ptr1& ptr1, const Ptr2& ptr2)
+{
+    const int rows = getRows(ptr0);
+    const int cols = getCols(ptr0);
+
+    CV_Assert( getRows(ptr1) == rows && getCols(ptr1) == cols );
+    CV_Assert( getRows(ptr2) == rows && getCols(ptr2) == cols );
+
+    ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type, typename PtrTraits<Ptr2>::ptr_type> >
+            z(make_tuple(shrinkPtr(ptr0), shrinkPtr(ptr1), shrinkPtr(ptr2)));
+    z.rows = rows;
+    z.cols = cols;
+
+    return z;
+}
+
+template <class Ptr0, class Ptr1, class Ptr2, class Ptr3>
+__host__ ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type, typename PtrTraits<Ptr2>::ptr_type, typename PtrTraits<Ptr3>::ptr_type> >
+zipPtr(const Ptr0& ptr0, const Ptr1& ptr1, const Ptr2& ptr2, const Ptr3& ptr3)
+{
+    const int rows = getRows(ptr0);
+    const int cols = getCols(ptr0);
+
+    CV_Assert( getRows(ptr1) == rows && getCols(ptr1) == cols );
+    CV_Assert( getRows(ptr2) == rows && getCols(ptr2) == cols );
+    CV_Assert( getRows(ptr3) == rows && getCols(ptr3) == cols );
+
+    ZipPtrSz< tuple<typename PtrTraits<Ptr0>::ptr_type, typename PtrTraits<Ptr1>::ptr_type, typename PtrTraits<Ptr2>::ptr_type, typename PtrTraits<Ptr3>::ptr_type> >
+            z(make_tuple(shrinkPtr(ptr0), shrinkPtr(ptr1), shrinkPtr(ptr2), shrinkPtr(ptr3)));
+    z.rows = rows;
+    z.cols = cols;
+
+    return z;
+}
+
+template <class PtrTuple> struct PtrTraits< ZipPtrSz<PtrTuple> > : PtrTraitsBase<ZipPtrSz<PtrTuple>, ZipPtr<PtrTuple> >
+{
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/atomic.hpp b/modules/cudev/include/opencv2/cudev/util/atomic.hpp
new file mode 100644
index 0000000000..f650c68ec2
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/atomic.hpp
@@ -0,0 +1,156 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_ATOMIC_HPP__
+#define __OPENCV_CUDEV_UTIL_ATOMIC_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+// atomicAdd
+
+__device__ __forceinline__ int atomicAdd(int* address, int val)
+{
+    return ::atomicAdd(address, val);
+}
+
+__device__ __forceinline__ uint atomicAdd(uint* address, uint val)
+{
+    return ::atomicAdd(address, val);
+}
+
+__device__ __forceinline__ float atomicAdd(float* address, float val)
+{
+    return ::atomicAdd(address, val);
+}
+
+__device__ static double atomicAdd(double* address, double val)
+{
+    unsigned long long int* address_as_ull = (unsigned long long int*) address;
+    unsigned long long int old = *address_as_ull, assumed;
+    do {
+        assumed = old;
+        old = ::atomicCAS(address_as_ull, assumed,
+            __double_as_longlong(val + __longlong_as_double(assumed)));
+    } while (assumed != old);
+    return __longlong_as_double(old);
+}
+
+// atomicMin
+
+__device__ __forceinline__ int atomicMin(int* address, int val)
+{
+    return ::atomicMin(address, val);
+}
+
+__device__ __forceinline__ uint atomicMin(uint* address, uint val)
+{
+    return ::atomicMin(address, val);
+}
+
+__device__ static float atomicMin(float* address, float val)
+{
+    int* address_as_i = (int*) address;
+    int old = *address_as_i, assumed;
+    do {
+        assumed = old;
+        old = ::atomicCAS(address_as_i, assumed,
+            __float_as_int(::fminf(val, __int_as_float(assumed))));
+    } while (assumed != old);
+    return __int_as_float(old);
+}
+
+__device__ static double atomicMin(double* address, double val)
+{
+    unsigned long long int* address_as_ull = (unsigned long long int*) address;
+    unsigned long long int old = *address_as_ull, assumed;
+    do {
+        assumed = old;
+        old = ::atomicCAS(address_as_ull, assumed,
+            __double_as_longlong(::fmin(val, __longlong_as_double(assumed))));
+    } while (assumed != old);
+    return __longlong_as_double(old);
+}
+
+// atomicMax
+
+__device__ __forceinline__ int atomicMax(int* address, int val)
+{
+    return ::atomicMax(address, val);
+}
+
+__device__ __forceinline__ uint atomicMax(uint* address, uint val)
+{
+    return ::atomicMax(address, val);
+}
+
+__device__ static float atomicMax(float* address, float val)
+{
+    int* address_as_i = (int*) address;
+    int old = *address_as_i, assumed;
+    do {
+        assumed = old;
+        old = ::atomicCAS(address_as_i, assumed,
+            __float_as_int(::fmaxf(val, __int_as_float(assumed))));
+    } while (assumed != old);
+    return __int_as_float(old);
+}
+
+__device__ static double atomicMax(double* address, double val)
+{
+    unsigned long long int* address_as_ull = (unsigned long long int*) address;
+    unsigned long long int old = *address_as_ull, assumed;
+    do {
+        assumed = old;
+        old = ::atomicCAS(address_as_ull, assumed,
+            __double_as_longlong(::fmax(val, __longlong_as_double(assumed))));
+    } while (assumed != old);
+    return __longlong_as_double(old);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/detail/tuple.hpp b/modules/cudev/include/opencv2/cudev/util/detail/tuple.hpp
new file mode 100644
index 0000000000..2fc84946bb
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/detail/tuple.hpp
@@ -0,0 +1,175 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_TUPLE_DETAIL_HPP__
+#define __OPENCV_CUDEV_UTIL_TUPLE_DETAIL_HPP__
+
+#include <thrust/tuple.h>
+
+namespace cv { namespace cudev {
+
+namespace tuple_detail
+{
+    using thrust::tuple;
+    using thrust::tuple_size;
+    using thrust::get;
+    using thrust::tuple_element;
+    using thrust::make_tuple;
+    using thrust::tie;
+
+    template <class Tuple, int SIZE, template <typename T> class CvtOp> struct ConvertTuple;
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 2, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 3, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 4, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 5, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 6, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<5, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 7, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<5, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<6, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 8, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<5, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<6, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<7, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 9, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<5, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<6, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<7, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<8, Tuple>::type>::type
+        > type;
+    };
+
+    template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple<Tuple, 10, CvtOp>
+    {
+        typedef tuple<
+            typename CvtOp<typename tuple_element<0, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<1, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<2, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<3, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<4, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<5, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<6, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<7, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<8, Tuple>::type>::type,
+            typename CvtOp<typename tuple_element<9, Tuple>::type>::type
+        > type;
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/detail/type_traits.hpp b/modules/cudev/include/opencv2/cudev/util/detail/type_traits.hpp
new file mode 100644
index 0000000000..d711642836
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/detail/type_traits.hpp
@@ -0,0 +1,238 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_TYPE_TRAITS_DETAIL_HPP__
+#define __OPENCV_CUDEV_UTIL_TYPE_TRAITS_DETAIL_HPP__
+
+#include "../../common.hpp"
+
+namespace cv { namespace cudev {
+
+namespace type_traits_detail
+{
+    template <typename T> struct IsSignedIntergral { enum {value = 0}; };
+    template <> struct IsSignedIntergral<schar> { enum {value = 1}; };
+    template <> struct IsSignedIntergral<short> { enum {value = 1}; };
+    template <> struct IsSignedIntergral<int> { enum {value = 1}; };
+
+    template <typename T> struct IsUnsignedIntegral { enum {value = 0}; };
+    template <> struct IsUnsignedIntegral<uchar> { enum {value = 1}; };
+    template <> struct IsUnsignedIntegral<ushort> { enum {value = 1}; };
+    template <> struct IsUnsignedIntegral<uint> { enum {value = 1}; };
+
+    template <typename T> struct IsIntegral { enum {value = IsSignedIntergral<T>::value || IsUnsignedIntegral<T>::value}; };
+    template <> struct IsIntegral<char> { enum {value = 1}; };
+    template <> struct IsIntegral<bool> { enum {value = 1}; };
+
+    template <typename T> struct IsFloat { enum {value = 0}; };
+    template <> struct IsFloat<float> { enum {value = 1}; };
+    template <> struct IsFloat<double> { enum {value = 1}; };
+
+    template <typename T> struct IsVec { enum {value = 0}; };
+    template <> struct IsVec<uchar1> { enum {value = 1}; };
+    template <> struct IsVec<uchar2> { enum {value = 1}; };
+    template <> struct IsVec<uchar3> { enum {value = 1}; };
+    template <> struct IsVec<uchar4> { enum {value = 1}; };
+    template <> struct IsVec<char1> { enum {value = 1}; };
+    template <> struct IsVec<char2> { enum {value = 1}; };
+    template <> struct IsVec<char3> { enum {value = 1}; };
+    template <> struct IsVec<char4> { enum {value = 1}; };
+    template <> struct IsVec<ushort1> { enum {value = 1}; };
+    template <> struct IsVec<ushort2> { enum {value = 1}; };
+    template <> struct IsVec<ushort3> { enum {value = 1}; };
+    template <> struct IsVec<ushort4> { enum {value = 1}; };
+    template <> struct IsVec<short1> { enum {value = 1}; };
+    template <> struct IsVec<short2> { enum {value = 1}; };
+    template <> struct IsVec<short3> { enum {value = 1}; };
+    template <> struct IsVec<short4> { enum {value = 1}; };
+    template <> struct IsVec<uint1> { enum {value = 1}; };
+    template <> struct IsVec<uint2> { enum {value = 1}; };
+    template <> struct IsVec<uint3> { enum {value = 1}; };
+    template <> struct IsVec<uint4> { enum {value = 1}; };
+    template <> struct IsVec<int1> { enum {value = 1}; };
+    template <> struct IsVec<int2> { enum {value = 1}; };
+    template <> struct IsVec<int3> { enum {value = 1}; };
+    template <> struct IsVec<int4> { enum {value = 1}; };
+    template <> struct IsVec<float1> { enum {value = 1}; };
+    template <> struct IsVec<float2> { enum {value = 1}; };
+    template <> struct IsVec<float3> { enum {value = 1}; };
+    template <> struct IsVec<float4> { enum {value = 1}; };
+    template <> struct IsVec<double1> { enum {value = 1}; };
+    template <> struct IsVec<double2> { enum {value = 1}; };
+    template <> struct IsVec<double3> { enum {value = 1}; };
+    template <> struct IsVec<double4> { enum {value = 1}; };
+
+    template <class U> struct AddParameterType { typedef const U& type; };
+    template <class U> struct AddParameterType<U&> { typedef U& type; };
+    template <> struct AddParameterType<void> { typedef void type; };
+
+    // ReferenceTraits
+
+    template <class U> struct ReferenceTraits
+    {
+        enum { value = 0 };
+        typedef U type;
+    };
+    template <class U> struct ReferenceTraits<U&>
+    {
+        enum { value = 1 };
+        typedef U type;
+    };
+
+    // PointerTraits
+
+    template <class U> struct PointerTraits
+    {
+        enum { value = 0 };
+        typedef void type;
+    };
+    template <class U> struct PointerTraits<U*>
+    {
+        enum { value = 1 };
+        typedef U type;
+    };
+    template <class U> struct PointerTraits<U*&>
+    {
+        enum { value = 1 };
+        typedef U type;
+    };
+
+    // UnConst
+
+    template <class U> struct UnConst
+    {
+        typedef U type;
+        enum { value = 0 };
+    };
+    template <class U> struct UnConst<const U>
+    {
+        typedef U type;
+        enum { value = 1 };
+    };
+    template <class U> struct UnConst<const U&>
+    {
+        typedef U& type;
+        enum { value = 1 };
+    };
+
+    // UnVolatile
+
+    template <class U> struct UnVolatile
+    {
+        typedef U type;
+        enum { value = 0 };
+    };
+    template <class U> struct UnVolatile<volatile U>
+    {
+        typedef U type;
+        enum { value = 1 };
+    };
+    template <class U> struct UnVolatile<volatile U&>
+    {
+        typedef U& type;
+        enum { value = 1 };
+    };
+
+    // IsSimpleParameter
+
+    template <typename T> struct IsSimpleParameter
+    {
+        enum { value = IsIntegral<T>::value
+               || IsFloat<T>::value
+               || PointerTraits<typename ReferenceTraits<T>::type>::value};
+    };
+
+    // LargerDepth
+
+    template <bool, typename ThenType, typename ElseType> struct SelectIf
+    {
+        typedef ThenType type;
+    };
+    template <typename ThenType, typename ElseType> struct SelectIf<false, ThenType, ElseType>
+    {
+        typedef ElseType type;
+    };
+
+    template <typename A, typename B> struct LargerDepth
+    {
+        typedef typename SelectIf<sizeof(A) >= sizeof(B), A, B>::type type;
+    };
+    template <typename A> struct LargerDepth<A, float>
+    {
+        typedef float type;
+    };
+    template <typename A> struct LargerDepth<float, A>
+    {
+        typedef float type;
+    };
+    template <typename A> struct LargerDepth<A, double>
+    {
+        typedef double type;
+    };
+    template <typename A> struct LargerDepth<double, A>
+    {
+        typedef double type;
+    };
+    template <> struct LargerDepth<float, float>
+    {
+        typedef float type;
+    };
+    template <> struct LargerDepth<float, double>
+    {
+        typedef double type;
+    };
+    template <> struct LargerDepth<double, float>
+    {
+        typedef double type;
+    };
+    template <> struct LargerDepth<double, double>
+    {
+        typedef double type;
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/limits.hpp b/modules/cudev/include/opencv2/cudev/util/limits.hpp
new file mode 100644
index 0000000000..58faca6b56
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/limits.hpp
@@ -0,0 +1,124 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_LIMITS_HPP__
+#define __OPENCV_CUDEV_UTIL_LIMITS_HPP__
+
+#include <limits.h>
+#include <float.h>
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+template <class T> struct numeric_limits;
+
+template <> struct numeric_limits<bool>
+{
+    __device__ __forceinline__ static bool min() { return false; }
+    __device__ __forceinline__ static bool max() { return true;  }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<schar>
+{
+    __device__ __forceinline__ static schar min() { return SCHAR_MIN; }
+    __device__ __forceinline__ static schar max() { return SCHAR_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<uchar>
+{
+    __device__ __forceinline__ static uchar min() { return 0; }
+    __device__ __forceinline__ static uchar max() { return UCHAR_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<short>
+{
+    __device__ __forceinline__ static short min() { return SHRT_MIN; }
+    __device__ __forceinline__ static short max() { return SHRT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<ushort>
+{
+    __device__ __forceinline__ static ushort min() { return 0; }
+    __device__ __forceinline__ static ushort max() { return USHRT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<int>
+{
+    __device__ __forceinline__ static int min() { return INT_MIN; }
+    __device__ __forceinline__ static int max() { return INT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<uint>
+{
+    __device__ __forceinline__ static uint min() { return 0; }
+    __device__ __forceinline__ static uint max() { return UINT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<float>
+{
+    __device__ __forceinline__ static float min() { return FLT_MIN; }
+    __device__ __forceinline__ static float max() { return FLT_MAX; }
+    __device__ __forceinline__ static float epsilon() { return FLT_EPSILON; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<double>
+{
+    __device__ __forceinline__ static double min() { return DBL_MIN; }
+    __device__ __forceinline__ static double max() { return DBL_MAX; }
+    __device__ __forceinline__ static double epsilon() { return DBL_EPSILON; }
+    static const bool is_signed = true;
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/saturate_cast.hpp b/modules/cudev/include/opencv2/cudev/util/saturate_cast.hpp
new file mode 100644
index 0000000000..faa12e32f2
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/saturate_cast.hpp
@@ -0,0 +1,264 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_SATURATE_CAST_HPP__
+#define __OPENCV_CUDEV_UTIL_SATURATE_CAST_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T> __device__ __forceinline__ T saturate_cast(uchar v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(schar v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(ushort v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(short v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(uint v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(int v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(float v) { return T(v); }
+template <typename T> __device__ __forceinline__ T saturate_cast(double v) { return T(v); }
+
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(schar v)
+{
+    uint res = 0;
+    int vi = v;
+    asm("cvt.sat.u8.s8 %0, %1;" : "=r"(res) : "r"(vi));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(short v)
+{
+    uint res = 0;
+    asm("cvt.sat.u8.s16 %0, %1;" : "=r"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(ushort v)
+{
+    uint res = 0;
+    asm("cvt.sat.u8.u16 %0, %1;" : "=r"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(int v)
+{
+    uint res = 0;
+    asm("cvt.sat.u8.s32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(uint v)
+{
+    uint res = 0;
+    asm("cvt.sat.u8.u32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(float v)
+{
+    uint res = 0;
+    asm("cvt.rni.sat.u8.f32 %0, %1;" : "=r"(res) : "f"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uchar saturate_cast<uchar>(double v)
+{
+    uint res = 0;
+    asm("cvt.rni.sat.u8.f64 %0, %1;" : "=r"(res) : "d"(v));
+    return res;
+}
+
+template <> __device__ __forceinline__ schar saturate_cast<schar>(uchar v)
+{
+    uint res = 0;
+    uint vi = v;
+    asm("cvt.sat.s8.u8 %0, %1;" : "=r"(res) : "r"(vi));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(short v)
+{
+    uint res = 0;
+    asm("cvt.sat.s8.s16 %0, %1;" : "=r"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(ushort v)
+{
+    uint res = 0;
+    asm("cvt.sat.s8.u16 %0, %1;" : "=r"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(int v)
+{
+    uint res = 0;
+    asm("cvt.sat.s8.s32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(uint v)
+{
+    uint res = 0;
+    asm("cvt.sat.s8.u32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(float v)
+{
+    uint res = 0;
+    asm("cvt.rni.sat.s8.f32 %0, %1;" : "=r"(res) : "f"(v));
+    return res;
+}
+template <> __device__ __forceinline__ schar saturate_cast<schar>(double v)
+{
+    uint res = 0;
+    asm("cvt.rni.sat.s8.f64 %0, %1;" : "=r"(res) : "d"(v));
+    return res;
+}
+
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(schar v)
+{
+    ushort res = 0;
+    int vi = v;
+    asm("cvt.sat.u16.s8 %0, %1;" : "=h"(res) : "r"(vi));
+    return res;
+}
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(short v)
+{
+    ushort res = 0;
+    asm("cvt.sat.u16.s16 %0, %1;" : "=h"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(int v)
+{
+    ushort res = 0;
+    asm("cvt.sat.u16.s32 %0, %1;" : "=h"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(uint v)
+{
+    ushort res = 0;
+    asm("cvt.sat.u16.u32 %0, %1;" : "=h"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(float v)
+{
+    ushort res = 0;
+    asm("cvt.rni.sat.u16.f32 %0, %1;" : "=h"(res) : "f"(v));
+    return res;
+}
+template <> __device__ __forceinline__ ushort saturate_cast<ushort>(double v)
+{
+    ushort res = 0;
+    asm("cvt.rni.sat.u16.f64 %0, %1;" : "=h"(res) : "d"(v));
+    return res;
+}
+
+template <> __device__ __forceinline__ short saturate_cast<short>(ushort v)
+{
+    short res = 0;
+    asm("cvt.sat.s16.u16 %0, %1;" : "=h"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ short saturate_cast<short>(int v)
+{
+    short res = 0;
+    asm("cvt.sat.s16.s32 %0, %1;" : "=h"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ short saturate_cast<short>(uint v)
+{
+    short res = 0;
+    asm("cvt.sat.s16.u32 %0, %1;" : "=h"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ short saturate_cast<short>(float v)
+{
+    short res = 0;
+    asm("cvt.rni.sat.s16.f32 %0, %1;" : "=h"(res) : "f"(v));
+    return res;
+}
+template <> __device__ __forceinline__ short saturate_cast<short>(double v)
+{
+    short res = 0;
+    asm("cvt.rni.sat.s16.f64 %0, %1;" : "=h"(res) : "d"(v));
+    return res;
+}
+
+template <> __device__ __forceinline__ int saturate_cast<int>(uint v)
+{
+    int res = 0;
+    asm("cvt.sat.s32.u32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ int saturate_cast<int>(float v)
+{
+    return __float2int_rn(v);
+}
+template <> __device__ __forceinline__ int saturate_cast<int>(double v)
+{
+    return __double2int_rn(v);
+}
+
+template <> __device__ __forceinline__ uint saturate_cast<uint>(schar v)
+{
+    uint res = 0;
+    int vi = v;
+    asm("cvt.sat.u32.s8 %0, %1;" : "=r"(res) : "r"(vi));
+    return res;
+}
+template <> __device__ __forceinline__ uint saturate_cast<uint>(short v)
+{
+    uint res = 0;
+    asm("cvt.sat.u32.s16 %0, %1;" : "=r"(res) : "h"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uint saturate_cast<uint>(int v)
+{
+    uint res = 0;
+    asm("cvt.sat.u32.s32 %0, %1;" : "=r"(res) : "r"(v));
+    return res;
+}
+template <> __device__ __forceinline__ uint saturate_cast<uint>(float v)
+{
+    return __float2uint_rn(v);
+}
+template <> __device__ __forceinline__ uint saturate_cast<uint>(double v)
+{
+    return __double2uint_rn(v);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/simd_functions.hpp b/modules/cudev/include/opencv2/cudev/util/simd_functions.hpp
new file mode 100644
index 0000000000..db63f51800
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/simd_functions.hpp
@@ -0,0 +1,913 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/*
+ * Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ *
+ *   Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ *
+ *   Neither the name of NVIDIA Corporation nor the names of its contributors
+ *   may be used to endorse or promote products derived from this software
+ *   without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_SIMD_FUNCTIONS_HPP__
+#define __OPENCV_CUDEV_UTIL_SIMD_FUNCTIONS_HPP__
+
+#include "../common.hpp"
+
+/*
+  This header file contains inline functions that implement intra-word SIMD
+  operations, that are hardware accelerated on sm_3x (Kepler) GPUs. Efficient
+  emulation code paths are provided for earlier architectures (sm_1x, sm_2x)
+  to make the code portable across all GPUs supported by CUDA. The following
+  functions are currently implemented:
+
+  vadd2(a,b)      per-halfword unsigned addition, with wrap-around: a + b
+  vsub2(a,b)      per-halfword unsigned subtraction, with wrap-around: a - b
+  vabsdiff2(a,b)  per-halfword unsigned absolute difference: |a - b|
+  vavg2(a,b)      per-halfword unsigned average: (a + b) / 2
+  vavrg2(a,b)     per-halfword unsigned rounded average: (a + b + 1) / 2
+  vseteq2(a,b)    per-halfword unsigned comparison: a == b ? 1 : 0
+  vcmpeq2(a,b)    per-halfword unsigned comparison: a == b ? 0xffff : 0
+  vsetge2(a,b)    per-halfword unsigned comparison: a >= b ? 1 : 0
+  vcmpge2(a,b)    per-halfword unsigned comparison: a >= b ? 0xffff : 0
+  vsetgt2(a,b)    per-halfword unsigned comparison: a > b ? 1 : 0
+  vcmpgt2(a,b)    per-halfword unsigned comparison: a > b ? 0xffff : 0
+  vsetle2(a,b)    per-halfword unsigned comparison: a <= b ? 1 : 0
+  vcmple2(a,b)    per-halfword unsigned comparison: a <= b ? 0xffff : 0
+  vsetlt2(a,b)    per-halfword unsigned comparison: a < b ? 1 : 0
+  vcmplt2(a,b)    per-halfword unsigned comparison: a < b ? 0xffff : 0
+  vsetne2(a,b)    per-halfword unsigned comparison: a != b ? 1 : 0
+  vcmpne2(a,b)    per-halfword unsigned comparison: a != b ? 0xffff : 0
+  vmax2(a,b)      per-halfword unsigned maximum: max(a, b)
+  vmin2(a,b)      per-halfword unsigned minimum: min(a, b)
+
+  vadd4(a,b)      per-byte unsigned addition, with wrap-around: a + b
+  vsub4(a,b)      per-byte unsigned subtraction, with wrap-around: a - b
+  vabsdiff4(a,b)  per-byte unsigned absolute difference: |a - b|
+  vavg4(a,b)      per-byte unsigned average: (a + b) / 2
+  vavrg4(a,b)     per-byte unsigned rounded average: (a + b + 1) / 2
+  vseteq4(a,b)    per-byte unsigned comparison: a == b ? 1 : 0
+  vcmpeq4(a,b)    per-byte unsigned comparison: a == b ? 0xff : 0
+  vsetge4(a,b)    per-byte unsigned comparison: a >= b ? 1 : 0
+  vcmpge4(a,b)    per-byte unsigned comparison: a >= b ? 0xff : 0
+  vsetgt4(a,b)    per-byte unsigned comparison: a > b ? 1 : 0
+  vcmpgt4(a,b)    per-byte unsigned comparison: a > b ? 0xff : 0
+  vsetle4(a,b)    per-byte unsigned comparison: a <= b ? 1 : 0
+  vcmple4(a,b)    per-byte unsigned comparison: a <= b ? 0xff : 0
+  vsetlt4(a,b)    per-byte unsigned comparison: a < b ? 1 : 0
+  vcmplt4(a,b)    per-byte unsigned comparison: a < b ? 0xff : 0
+  vsetne4(a,b)    per-byte unsigned comparison: a != b ? 1: 0
+  vcmpne4(a,b)    per-byte unsigned comparison: a != b ? 0xff: 0
+  vmax4(a,b)      per-byte unsigned maximum: max(a, b)
+  vmin4(a,b)      per-byte unsigned minimum: min(a, b)
+*/
+
+namespace cv { namespace cudev {
+
+// 2
+
+__device__ __forceinline__ uint vadd2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vadd2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vadd.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vadd.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s;
+    s = a ^ b;          // sum bits
+    r = a + b;          // actual sum
+    s = s ^ r;          // determine carry-ins for each bit position
+    s = s & 0x00010000; // carry-in to high word (= carry-out from low word)
+    r = r - s;          // subtract out carry-out from low word
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsub2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vsub2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vsub.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vsub.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s;
+    s = a ^ b;          // sum bits
+    r = a - b;          // actual sum
+    s = s ^ r;          // determine carry-ins for each bit position
+    s = s & 0x00010000; // borrow to high word
+    r = r + s;          // compensate for borrow from low word
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vabsdiff2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vabsdiff2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vabsdiff.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vabsdiff.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s, t, u, v;
+    s = a & 0x0000ffff; // extract low halfword
+    r = b & 0x0000ffff; // extract low halfword
+    u = ::max(r, s);    // maximum of low halfwords
+    v = ::min(r, s);    // minimum of low halfwords
+    s = a & 0xffff0000; // extract high halfword
+    r = b & 0xffff0000; // extract high halfword
+    t = ::max(r, s);    // maximum of high halfwords
+    s = ::min(r, s);    // minimum of high halfwords
+    r = u | t;          // maximum of both halfwords
+    s = v | s;          // minimum of both halfwords
+    r = r - s;          // |a - b| = max(a,b) - min(a,b);
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vavg2(uint a, uint b)
+{
+    uint r, s;
+
+    // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+    // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+    s = a ^ b;
+    r = a & b;
+    s = s & 0xfffefffe; // ensure shift doesn't cross halfword boundaries
+    s = s >> 1;
+    s = r + s;
+
+    return s;
+}
+
+__device__ __forceinline__ uint vavrg2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vavrg2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+    // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+    uint s;
+    s = a ^ b;
+    r = a | b;
+    s = s & 0xfffefffe; // ensure shift doesn't cross half-word boundaries
+    s = s >> 1;
+    r = r - s;
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vseteq2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset2.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    uint c;
+    r = a ^ b;          // 0x0000 if a == b
+    c = r | 0x80008000; // set msbs, to catch carry out
+    r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+    c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+    c = r & ~c;         // msb = 1, if r was 0x0000
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpeq2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vseteq2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    r = a ^ b;          // 0x0000 if a == b
+    c = r | 0x80008000; // set msbs, to catch carry out
+    r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+    c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+    c = r & ~c;         // msb = 1, if r was 0x0000
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetge2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset2.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpge2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetge2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetgt2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset2.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+    c = c & 0x80008000; // msbs = carry-outs
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpgt2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetgt2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+    c = c & 0x80008000; // msbs = carry-outs
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetle2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset2.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmple2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetle2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetlt2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset2.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmplt2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetlt2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+    c = c & 0x80008000; // msb = carry-outs
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetne2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm ("vset2.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    uint c;
+    r = a ^ b;          // 0x0000 if a == b
+    c = r | 0x80008000; // set msbs, to catch carry out
+    c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+    c = r | c;          // msb = 1, if r was not 0x0000
+    c = c & 0x80008000; // extract msbs
+    r = c >> 15;        // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpne2(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetne2(a, b);
+    c = r << 16;        // convert bool
+    r = c - r;          //  into mask
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    r = a ^ b;          // 0x0000 if a == b
+    c = r | 0x80008000; // set msbs, to catch carry out
+    c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+    c = r | c;          // msb = 1, if r was not 0x0000
+    c = c & 0x80008000; // extract msbs
+    r = c >> 15;        // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vmax2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vmax2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vmax.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmax.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s, t, u;
+    r = a & 0x0000ffff; // extract low halfword
+    s = b & 0x0000ffff; // extract low halfword
+    t = ::max(r, s);    // maximum of low halfwords
+    r = a & 0xffff0000; // extract high halfword
+    s = b & 0xffff0000; // extract high halfword
+    u = ::max(r, s);    // maximum of high halfwords
+    r = t | u;          // combine halfword maximums
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vmin2(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vmin2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vmin.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmin.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s, t, u;
+    r = a & 0x0000ffff; // extract low halfword
+    s = b & 0x0000ffff; // extract low halfword
+    t = ::min(r, s);    // minimum of low halfwords
+    r = a & 0xffff0000; // extract high halfword
+    s = b & 0xffff0000; // extract high halfword
+    u = ::min(r, s);    // minimum of high halfwords
+    r = t | u;          // combine halfword minimums
+#endif
+
+    return r;
+}
+
+// 4
+
+__device__ __forceinline__ uint vadd4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vadd4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vadd.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vadd.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vadd.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vadd.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s, t;
+    s = a ^ b;          // sum bits
+    r = a & 0x7f7f7f7f; // clear msbs
+    t = b & 0x7f7f7f7f; // clear msbs
+    s = s & 0x80808080; // msb sum bits
+    r = r + t;          // add without msbs, record carry-out in msbs
+    r = r ^ s;          // sum of msb sum and carry-in bits, w/o carry-out
+#endif /* CV_CUDEV_ARCH >= 300 */
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsub4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vsub4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vsub.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vsub.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vsub.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vsub.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s, t;
+    s = a ^ ~b;         // inverted sum bits
+    r = a | 0x80808080; // set msbs
+    t = b & 0x7f7f7f7f; // clear msbs
+    s = s & 0x80808080; // inverted msb sum bits
+    r = r - t;          // subtract w/o msbs, record inverted borrows in msb
+    r = r ^ s;          // combine inverted msb sum bits and borrows
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vavg4(uint a, uint b)
+{
+    uint r, s;
+
+    // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+    // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+    s = a ^ b;
+    r = a & b;
+    s = s & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+    s = s >> 1;
+    s = r + s;
+
+    return s;
+}
+
+__device__ __forceinline__ uint vavrg4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vavrg4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+    // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+    uint c;
+    c = a ^ b;
+    r = a | b;
+    c = c & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+    c = c >> 1;
+    r = r - c;
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vseteq4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    uint c;
+    r = a ^ b;          // 0x00 if a == b
+    c = r | 0x80808080; // set msbs, to catch carry out
+    r = r ^ c;          // extract msbs, msb = 1 if r < 0x80
+    c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+    c = r & ~c;         // msb = 1, if r was 0x00
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpeq4(uint a, uint b)
+{
+    uint r, t;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vseteq4(a, b);
+    t = r << 8;         // convert bool
+    r = t - r;          //  to mask
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    t = a ^ b;          // 0x00 if a == b
+    r = t | 0x80808080; // set msbs, to catch carry out
+    t = t ^ r;          // extract msbs, msb = 1 if t < 0x80
+    r = r - 0x01010101; // msb = 0, if t was 0x00 or 0x80
+    r = t & ~r;         // msb = 1, if t was 0x00
+    t = r >> 7;         // build mask
+    t = r - t;          //  from
+    r = t | r;          //   msbs
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetle4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+    c = c & 0x80808080; // msb = carry-outs
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmple4(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetle4(a, b);
+    c = r << 8;         // convert bool
+    r = c - r;          //  to mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+    c = c & 0x80808080; // msbs = carry-outs
+    r = c >> 7;         // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetlt4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+    c = c & 0x80808080; // msb = carry-outs
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmplt4(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetlt4(a, b);
+    c = r << 8;         // convert bool
+    r = c - r;          //  to mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(a));
+    c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+    c = c & 0x80808080; // msbs = carry-outs
+    r = c >> 7;         // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetge4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavrg4(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+    c = c & 0x80808080; // msb = carry-outs
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpge4(uint a, uint b)
+{
+    uint r, s;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetge4(a, b);
+    s = r << 8;         // convert bool
+    r = s - r;          //  to mask
+#else
+    asm ("not.b32 %0,%0;" : "+r"(b));
+    r = vavrg4 (a, b);  // (a + ~b + 1) / 2 = (a - b) / 2
+    r = r & 0x80808080; // msb = carry-outs
+    s = r >> 7;         // build mask
+    s = r - s;          //  from
+    r = s | r;          //   msbs
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetgt4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint c;
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+    c = c & 0x80808080; // msb = carry-outs
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpgt4(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetgt4(a, b);
+    c = r << 8;         // convert bool
+    r = c - r;          //  to mask
+#else
+    asm("not.b32 %0, %0;" : "+r"(b));
+    c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+    c = c & 0x80808080; // msb = carry-outs
+    r = c >> 7;         // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vsetne4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vset4.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    uint c;
+    r = a ^ b;          // 0x00 if a == b
+    c = r | 0x80808080; // set msbs, to catch carry out
+    c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+    c = r | c;          // msb = 1, if r was not 0x00
+    c = c & 0x80808080; // extract msbs
+    r = c >> 7;         // convert to bool
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vcmpne4(uint a, uint b)
+{
+    uint r, c;
+
+#if CV_CUDEV_ARCH >= 300
+    r = vsetne4(a, b);
+    c = r << 8;         // convert bool
+    r = c - r;          //  to mask
+#else
+    // inspired by Alan Mycroft's null-byte detection algorithm:
+    // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+    r = a ^ b;          // 0x00 if a == b
+    c = r | 0x80808080; // set msbs, to catch carry out
+    c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+    c = r | c;          // msb = 1, if r was not 0x00
+    c = c & 0x80808080; // extract msbs
+    r = c >> 7;         // convert
+    r = c - r;          //  msbs to
+    r = c | r;          //   mask
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vabsdiff4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vabsdiff4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vabsdiff.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vabsdiff.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vabsdiff.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vabsdiff.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s;
+    s = vcmpge4(a, b);  // mask = 0xff if a >= b
+    r = a ^ b;          //
+    s = (r &  s) ^ b;   // select a when a >= b, else select b => max(a,b)
+    r = s ^ r;          // select a when b >= a, else select b => min(a,b)
+    r = s - r;          // |a - b| = max(a,b) - min(a,b);
+#endif
+
+    return r;
+}
+
+__device__ __forceinline__ uint vmax4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vmax4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vmax.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmax.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmax.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmax.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s;
+    s = vcmpge4(a, b);  // mask = 0xff if a >= b
+    r = a & s;          // select a when b >= a
+    s = b & ~s;         // select b when b < a
+    r = r | s;          // combine byte selections
+#endif
+
+    return r;           // byte-wise unsigned maximum
+}
+
+__device__ __forceinline__ uint vmin4(uint a, uint b)
+{
+    uint r = 0;
+
+#if CV_CUDEV_ARCH >= 300
+    asm("vmin4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#elif CV_CUDEV_ARCH >= 200
+    asm("vmin.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmin.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmin.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    asm("vmin.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+#else
+    uint s;
+    s = vcmpge4(b, a);  // mask = 0xff if a >= b
+    r = a & s;          // select a when b >= a
+    s = b & ~s;         // select b when b < a
+    r = r | s;          // combine byte selections
+#endif
+
+    return r;
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/tuple.hpp b/modules/cudev/include/opencv2/cudev/util/tuple.hpp
new file mode 100644
index 0000000000..b015ff3445
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/tuple.hpp
@@ -0,0 +1,80 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_TUPLE_HPP__
+#define __OPENCV_CUDEV_UTIL_TUPLE_HPP__
+
+#include "../common.hpp"
+#include "detail/tuple.hpp"
+
+namespace cv { namespace cudev {
+
+using tuple_detail::tuple;
+using tuple_detail::tuple_size;
+using tuple_detail::get;
+using tuple_detail::tuple_element;
+using tuple_detail::make_tuple;
+using tuple_detail::tie;
+
+template <typename T> struct TupleTraits
+{
+    enum { is_tuple = 0 };
+    enum { size = 1 };
+};
+template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9>
+struct TupleTraits< tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >
+{
+    enum { is_tuple = 1 };
+    enum { size = tuple_size< tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value };
+};
+
+template <class Tuple, template <typename T> class CvtOp> struct ConvertTuple
+{
+    typedef typename tuple_detail::ConvertTuple<Tuple, tuple_size<Tuple>::value, CvtOp>::type type;
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/type_traits.hpp b/modules/cudev/include/opencv2/cudev/util/type_traits.hpp
new file mode 100644
index 0000000000..ca800c0b7a
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/type_traits.hpp
@@ -0,0 +1,169 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_TYPE_TRAITS_HPP__
+#define __OPENCV_CUDEV_UTIL_TYPE_TRAITS_HPP__
+
+#include "../common.hpp"
+#include "vec_traits.hpp"
+#include "detail/type_traits.hpp"
+
+namespace cv { namespace cudev {
+
+// NullType
+
+struct NullType {};
+
+// Int2Type
+
+template <int A> struct Int2Type
+{
+   enum { value = A };
+};
+
+// ArrayWrapper
+
+template <typename T, int COUNT> struct ArrayWrapper
+{
+    T array[COUNT];
+};
+
+// Log2 (compile time calculation)
+
+template <int N, int CURRENT_VAL = N, int COUNT = 0> struct Log2
+{
+    enum { value = Log2<N, (CURRENT_VAL >> 1), COUNT + 1>::VALUE };
+};
+template <int N, int COUNT> struct Log2<N, 0, COUNT>
+{
+    enum { value = (1 << (COUNT - 1) < N) ? COUNT : COUNT - 1 };
+};
+
+// IsPowerOf2
+
+template <int N> struct IsPowerOf2
+{
+    enum { value = ((N != 0) && !(N & (N - 1))) };
+};
+
+// SelectIf
+
+template <bool, typename ThenType, typename ElseType> struct SelectIf
+{
+    typedef ThenType type;
+};
+template <typename ThenType, typename ElseType> struct SelectIf<false, ThenType, ElseType>
+{
+    typedef ElseType type;
+};
+
+// EnableIf
+
+template <bool, typename T = void> struct EnableIf {};
+template <typename T> struct EnableIf<true, T> { typedef T type; };
+
+// DisableIf
+
+template <bool, typename T = void> struct DisableIf {};
+template <typename T> struct DisableIf<false, T> { typedef T type; };
+
+// TypesEquals
+
+template <typename A, typename B> struct TypesEquals
+{
+    enum { value = 0 };
+};
+template <typename A> struct TypesEquals<A, A>
+{
+    enum { value = 1 };
+};
+
+// TypeTraits
+
+template <typename T> struct TypeTraits
+{
+    typedef typename type_traits_detail::UnConst<T>::type                                                non_const_type;
+    typedef typename type_traits_detail::UnVolatile<T>::type                                             non_volatile_type;
+    typedef typename type_traits_detail::UnVolatile<typename type_traits_detail::UnConst<T>::type>::type unqualified_type;
+    typedef typename type_traits_detail::PointerTraits<unqualified_type>::type                           pointee_type;
+    typedef typename type_traits_detail::ReferenceTraits<T>::type                                        referred_type;
+
+    enum { is_const          = type_traits_detail::UnConst<T>::value };
+    enum { is_volatile       = type_traits_detail::UnVolatile<T>::value };
+
+    enum { is_reference      = type_traits_detail::ReferenceTraits<unqualified_type>::value };
+    enum { is_pointer        = type_traits_detail::PointerTraits<typename type_traits_detail::ReferenceTraits<unqualified_type>::type>::value };
+
+    enum { is_unsigned_int   = type_traits_detail::IsUnsignedIntegral<unqualified_type>::value };
+    enum { is_signed_int     = type_traits_detail::IsSignedIntergral<unqualified_type>::value };
+    enum { is_integral       = type_traits_detail::IsIntegral<unqualified_type>::value };
+    enum { is_float          = type_traits_detail::IsFloat<unqualified_type>::value };
+    enum { is_scalar         = is_integral || is_float };
+    enum { is_vec            = type_traits_detail::IsVec<unqualified_type>::value };
+
+    typedef typename SelectIf<type_traits_detail::IsSimpleParameter<unqualified_type>::value,
+        T, typename type_traits_detail::AddParameterType<T>::type>::type parameter_type;
+};
+
+// LargerType
+
+template <typename A, typename B> struct LargerType
+{
+    typedef typename SelectIf<
+        unsigned(VecTraits<A>::cn) != unsigned(VecTraits<B>::cn),
+        void,
+        typename MakeVec<
+            typename type_traits_detail::LargerDepth<
+                typename VecTraits<A>::elem_type,
+                typename VecTraits<B>::elem_type
+            >::type,
+            VecTraits<A>::cn
+        >::type
+    >::type type;
+};
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/vec_math.hpp b/modules/cudev/include/opencv2/cudev/util/vec_math.hpp
new file mode 100644
index 0000000000..3ce265e277
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/vec_math.hpp
@@ -0,0 +1,923 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_VEC_MATH_HPP__
+#define __OPENCV_CUDEV_UTIL_VEC_MATH_HPP__
+
+#include "vec_traits.hpp"
+#include "saturate_cast.hpp"
+
+namespace cv { namespace cudev {
+
+// saturate_cast
+
+namespace vec_math_detail
+{
+    template <int cn, typename VecD> struct SatCastHelper;
+
+    template <typename VecD> struct SatCastHelper<1, VecD>
+    {
+        template <typename VecS> __device__ __forceinline__ static VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x));
+        }
+    };
+
+    template <typename VecD> struct SatCastHelper<2, VecD>
+    {
+        template <typename VecS> __device__ __forceinline__ static VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y));
+        }
+    };
+
+    template <typename VecD> struct SatCastHelper<3, VecD>
+    {
+        template <typename VecS> __device__ __forceinline__ static VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z));
+        }
+    };
+
+    template <typename VecD> struct SatCastHelper<4, VecD>
+    {
+        template <typename VecS> __device__ __forceinline__ static VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z), saturate_cast<D>(v.w));
+        }
+    };
+}
+
+template<typename T> __device__ __forceinline__ T saturate_cast(const uchar1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const char1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const ushort1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const short1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const uint1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const int1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const float1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const double1& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+
+template<typename T> __device__ __forceinline__ T saturate_cast(const uchar2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const char2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const ushort2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const short2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const uint2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const int2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const float2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const double2& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+
+template<typename T> __device__ __forceinline__ T saturate_cast(const uchar3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const char3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const ushort3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const short3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const uint3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const int3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const float3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const double3& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+
+template<typename T> __device__ __forceinline__ T saturate_cast(const uchar4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const char4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const ushort4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const short4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const uint4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const int4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const float4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+template<typename T> __device__ __forceinline__ T saturate_cast(const double4& v) { return vec_math_detail::SatCastHelper<VecTraits<T>::cn, T>::cast(v); }
+
+// unary operators
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(op (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(op (a.x), op (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(op (a.x), op (a.y), op (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(op (a.x), op (a.y), op (a.z), op (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_OP
+
+// unary functions
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x), func (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x), func (a.y), func (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x), func (a.y), func (a.z), func (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::abs, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, /*::abs*/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabsf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabs, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrt, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::exp, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::log, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::cosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanh, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC
+
+// binary operators (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op b.x, a.y op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op b.x, a.y op b.y, a.z op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op b.x, a.y op b.y, a.z op b.z, a.w op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_OP
+
+// binary operators (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(op, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op s); \
+    } \
+    __device__ __forceinline__ output_type ## 1 operator op(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(s op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op s, a.y op s); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(s op b.x, s op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op s, a.y op s, a.z op s); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(s op b.x, s op b.y, s op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op s, a.y op s, a.z op s, a.w op s); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(s op b.x, s op b.y, s op b.z, s op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uint, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP
+
+// binary function (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x, b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x, b.x), func (a.y, b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z), func (a.w, b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmaxf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmax, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fminf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fmin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypot, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC
+
+// binary function (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(func_name, func, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) a.x, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 1 func_name(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) s, (output_type) b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s), func ((output_type) a.w, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z), func ((output_type) s, (output_type) b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, double, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/util/vec_traits.hpp b/modules/cudev/include/opencv2/cudev/util/vec_traits.hpp
new file mode 100644
index 0000000000..bdff958d87
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/util/vec_traits.hpp
@@ -0,0 +1,320 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_UTIL_VEC_TRAITS_HPP__
+#define __OPENCV_CUDEV_UTIL_VEC_TRAITS_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+// MakeVec
+
+template<typename T, int CN> struct MakeVec;
+
+#define CV_CUDEV_MAKE_VEC_INST(elem_type) \
+    template<> struct MakeVec<elem_type, 1> { typedef elem_type      type; }; \
+    template<> struct MakeVec<elem_type, 2> { typedef elem_type ## 2 type; }; \
+    template<> struct MakeVec<elem_type, 3> { typedef elem_type ## 3 type; }; \
+    template<> struct MakeVec<elem_type, 4> { typedef elem_type ## 4 type; };
+
+CV_CUDEV_MAKE_VEC_INST(uchar)
+CV_CUDEV_MAKE_VEC_INST(ushort)
+CV_CUDEV_MAKE_VEC_INST(short)
+CV_CUDEV_MAKE_VEC_INST(int)
+CV_CUDEV_MAKE_VEC_INST(uint)
+CV_CUDEV_MAKE_VEC_INST(float)
+CV_CUDEV_MAKE_VEC_INST(double)
+
+#undef CV_CUDEV_MAKE_VEC_INST
+
+template<> struct MakeVec<schar, 1> { typedef char  type; };
+template<> struct MakeVec<schar, 2> { typedef char2 type; };
+template<> struct MakeVec<schar, 3> { typedef char3 type; };
+template<> struct MakeVec<schar, 4> { typedef char4 type; };
+
+template<> struct MakeVec<bool, 1> { typedef uchar  type; };
+template<> struct MakeVec<bool, 2> { typedef uchar2 type; };
+template<> struct MakeVec<bool, 3> { typedef uchar3 type; };
+template<> struct MakeVec<bool, 4> { typedef uchar4 type; };
+
+// VecTraits
+
+template<typename T> struct VecTraits;
+
+#define CV_CUDEV_VEC_TRAITS_INST(type) \
+    template <> struct VecTraits<type> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        __host__ __device__ __forceinline__ static type all(type v) {return v;} \
+        __host__ __device__ __forceinline__ static type make(type x) {return x;} \
+        __host__ __device__ __forceinline__ static type make(const type* v) {return *v;} \
+    }; \
+    template <> struct VecTraits<type ## 1> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        __host__ __device__ __forceinline__ static type ## 1 all(type v) {return make_ ## type ## 1(v);} \
+        __host__ __device__ __forceinline__ static type ## 1 make(type x) {return make_ ## type ## 1(x);} \
+        __host__ __device__ __forceinline__ static type ## 1 make(const type* v) {return make_ ## type ## 1(*v);} \
+    }; \
+    template <> struct VecTraits<type ## 2> \
+    { \
+        typedef type elem_type; \
+        enum {cn=2}; \
+        __host__ __device__ __forceinline__ static type ## 2 all(type v) {return make_ ## type ## 2(v, v);} \
+        __host__ __device__ __forceinline__ static type ## 2 make(type x, type y) {return make_ ## type ## 2(x, y);} \
+        __host__ __device__ __forceinline__ static type ## 2 make(const type* v) {return make_ ## type ## 2(v[0], v[1]);} \
+    }; \
+    template <> struct VecTraits<type ## 3> \
+    { \
+        typedef type elem_type; \
+        enum {cn=3}; \
+        __host__ __device__ __forceinline__ static type ## 3 all(type v) {return make_ ## type ## 3(v, v, v);} \
+        __host__ __device__ __forceinline__ static type ## 3 make(type x, type y, type z) {return make_ ## type ## 3(x, y, z);} \
+        __host__ __device__ __forceinline__ static type ## 3 make(const type* v) {return make_ ## type ## 3(v[0], v[1], v[2]);} \
+    }; \
+    template <> struct VecTraits<type ## 4> \
+    { \
+        typedef type elem_type; \
+        enum {cn=4}; \
+        __host__ __device__ __forceinline__ static type ## 4 all(type v) {return make_ ## type ## 4(v, v, v, v);} \
+        __host__ __device__ __forceinline__ static type ## 4 make(type x, type y, type z, type w) {return make_ ## type ## 4(x, y, z, w);} \
+        __host__ __device__ __forceinline__ static type ## 4 make(const type* v) {return make_ ## type ## 4(v[0], v[1], v[2], v[3]);} \
+    };
+
+CV_CUDEV_VEC_TRAITS_INST(uchar)
+CV_CUDEV_VEC_TRAITS_INST(ushort)
+CV_CUDEV_VEC_TRAITS_INST(short)
+CV_CUDEV_VEC_TRAITS_INST(int)
+CV_CUDEV_VEC_TRAITS_INST(uint)
+CV_CUDEV_VEC_TRAITS_INST(float)
+CV_CUDEV_VEC_TRAITS_INST(double)
+
+#undef CV_CUDEV_VEC_TRAITS_INST
+
+template<> struct VecTraits<schar>
+{
+    typedef schar elem_type;
+    enum {cn=1};
+    __host__ __device__ __forceinline__ static schar all(schar v) {return v;}
+    __host__ __device__ __forceinline__ static schar make(schar x) {return x;}
+    __host__ __device__ __forceinline__ static schar make(const schar* x) {return *x;}
+};
+template<> struct VecTraits<char1>
+{
+    typedef schar elem_type;
+    enum {cn=1};
+    __host__ __device__ __forceinline__ static char1 all(schar v) {return make_char1(v);}
+    __host__ __device__ __forceinline__ static char1 make(schar x) {return make_char1(x);}
+    __host__ __device__ __forceinline__ static char1 make(const schar* v) {return make_char1(v[0]);}
+};
+template<> struct VecTraits<char2>
+{
+    typedef schar elem_type;
+    enum {cn=2};
+    __host__ __device__ __forceinline__ static char2 all(schar v) {return make_char2(v, v);}
+    __host__ __device__ __forceinline__ static char2 make(schar x, schar y) {return make_char2(x, y);}
+    __host__ __device__ __forceinline__ static char2 make(const schar* v) {return make_char2(v[0], v[1]);}
+};
+template<> struct VecTraits<char3>
+{
+    typedef schar elem_type;
+    enum {cn=3};
+    __host__ __device__ __forceinline__ static char3 all(schar v) {return make_char3(v, v, v);}
+    __host__ __device__ __forceinline__ static char3 make(schar x, schar y, schar z) {return make_char3(x, y, z);}
+    __host__ __device__ __forceinline__ static char3 make(const schar* v) {return make_char3(v[0], v[1], v[2]);}
+};
+template<> struct VecTraits<char4>
+{
+    typedef schar elem_type;
+    enum {cn=4};
+    __host__ __device__ __forceinline__ static char4 all(schar v) {return make_char4(v, v, v, v);}
+    __host__ __device__ __forceinline__ static char4 make(schar x, schar y, schar z, schar w) {return make_char4(x, y, z, w);}
+    __host__ __device__ __forceinline__ static char4 make(const schar* v) {return make_char4(v[0], v[1], v[2], v[3]);}
+};
+
+}}
+
+// DataType
+
+namespace cv {
+
+template <> class DataType<uint>
+{
+public:
+    typedef uint         value_type;
+    typedef value_type   work_type;
+    typedef value_type   channel_type;
+    typedef value_type   vec_type;
+    enum { generic_type = 0,
+           depth        = CV_32S,
+           channels     = 1,
+           fmt          = (int)'i',
+           type         = CV_MAKE_TYPE(depth, channels)
+         };
+};
+
+#define CV_CUDEV_DATA_TYPE_INST(_depth_type, _channel_num) \
+    template <> class DataType< _depth_type ## _channel_num > \
+    { \
+    public: \
+        typedef _depth_type ## _channel_num     value_type; \
+        typedef value_type                      work_type; \
+        typedef _depth_type                     channel_type; \
+        typedef value_type                      vec_type; \
+        enum { generic_type = 0, \
+               depth        = DataType<channel_type>::depth, \
+               channels     = _channel_num, \
+               fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8), \
+               type         = CV_MAKE_TYPE(depth, channels) \
+             }; \
+    };
+
+CV_CUDEV_DATA_TYPE_INST(uchar, 1)
+CV_CUDEV_DATA_TYPE_INST(uchar, 2)
+CV_CUDEV_DATA_TYPE_INST(uchar, 3)
+CV_CUDEV_DATA_TYPE_INST(uchar, 4)
+
+CV_CUDEV_DATA_TYPE_INST(ushort, 1)
+CV_CUDEV_DATA_TYPE_INST(ushort, 2)
+CV_CUDEV_DATA_TYPE_INST(ushort, 3)
+CV_CUDEV_DATA_TYPE_INST(ushort, 4)
+
+CV_CUDEV_DATA_TYPE_INST(short, 1)
+CV_CUDEV_DATA_TYPE_INST(short, 2)
+CV_CUDEV_DATA_TYPE_INST(short, 3)
+CV_CUDEV_DATA_TYPE_INST(short, 4)
+
+CV_CUDEV_DATA_TYPE_INST(int, 1)
+CV_CUDEV_DATA_TYPE_INST(int, 2)
+CV_CUDEV_DATA_TYPE_INST(int, 3)
+CV_CUDEV_DATA_TYPE_INST(int, 4)
+
+CV_CUDEV_DATA_TYPE_INST(uint, 1)
+CV_CUDEV_DATA_TYPE_INST(uint, 2)
+CV_CUDEV_DATA_TYPE_INST(uint, 3)
+CV_CUDEV_DATA_TYPE_INST(uint, 4)
+
+CV_CUDEV_DATA_TYPE_INST(float, 1)
+CV_CUDEV_DATA_TYPE_INST(float, 2)
+CV_CUDEV_DATA_TYPE_INST(float, 3)
+CV_CUDEV_DATA_TYPE_INST(float, 4)
+
+CV_CUDEV_DATA_TYPE_INST(double, 1)
+CV_CUDEV_DATA_TYPE_INST(double, 2)
+CV_CUDEV_DATA_TYPE_INST(double, 3)
+CV_CUDEV_DATA_TYPE_INST(double, 4)
+
+#undef CV_CUDEV_DATA_TYPE_INST
+
+template<> class DataType<char1>
+{
+public:
+    typedef char1      value_type;
+    typedef value_type work_type;
+    typedef schar      channel_type;
+    typedef value_type vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 1,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKE_TYPE(depth, channels)
+         };
+};
+
+template<> class DataType<char2>
+{
+public:
+    typedef char2      value_type;
+    typedef value_type work_type;
+    typedef schar      channel_type;
+    typedef value_type vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKE_TYPE(depth, channels)
+         };
+};
+
+template<> class DataType<char3>
+{
+public:
+    typedef char3      value_type;
+    typedef value_type work_type;
+    typedef schar      channel_type;
+    typedef value_type vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 3,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKE_TYPE(depth, channels)
+         };
+};
+
+template<> class DataType<char4>
+{
+public:
+    typedef char4      value_type;
+    typedef value_type work_type;
+    typedef schar      channel_type;
+    typedef value_type vec_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 4,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKE_TYPE(depth, channels)
+         };
+};
+
+}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/detail/reduce.hpp b/modules/cudev/include/opencv2/cudev/warp/detail/reduce.hpp
new file mode 100644
index 0000000000..7f9faf9cac
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/detail/reduce.hpp
@@ -0,0 +1,222 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_REDUCE_DETAIL_HPP__
+#define __OPENCV_CUDEV_WARP_REDUCE_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+#include "../../warp/shuffle.hpp"
+
+namespace cv { namespace cudev {
+
+namespace warp_reduce_detail
+{
+    // GetType
+
+    template <typename T> struct GetType;
+
+    template <typename T> struct GetType<T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<volatile T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<T&>
+    {
+        typedef T type;
+    };
+
+    // For
+
+    template <int I, int N> struct For
+    {
+        template <class PointerTuple, class ValTuple>
+        __device__ static void loadToSmem(const PointerTuple& smem, const ValTuple& val, uint tid)
+        {
+            get<I>(smem)[tid] = get<I>(val);
+
+            For<I + 1, N>::loadToSmem(smem, val, tid);
+        }
+
+        template <class PointerTuple, class ValTuple, class OpTuple>
+        __device__ static void merge(const PointerTuple& smem, const ValTuple& val, uint tid, uint delta, const OpTuple& op)
+        {
+            typename GetType<typename tuple_element<I, PointerTuple>::type>::type reg = get<I>(smem)[tid + delta];
+            get<I>(smem)[tid] = get<I>(val) = get<I>(op)(get<I>(val), reg);
+
+            For<I + 1, N>::merge(smem, val, tid, delta, op);
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        template <class ValTuple, class OpTuple>
+        __device__ static void mergeShfl(const ValTuple& val, uint delta, uint width, const OpTuple& op)
+        {
+            typename GetType<typename tuple_element<I, ValTuple>::type>::type reg = shfl_down(get<I>(val), delta, width);
+            get<I>(val) = get<I>(op)(get<I>(val), reg);
+
+            For<I + 1, N>::mergeShfl(val, delta, width, op);
+        }
+#endif
+    };
+
+    template <int N> struct For<N, N>
+    {
+        template <class PointerTuple, class ValTuple>
+        __device__ __forceinline__ static void loadToSmem(const PointerTuple&, const ValTuple&, uint)
+        {
+        }
+
+        template <class PointerTuple, class ValTuple, class OpTuple>
+        __device__ __forceinline__ static void merge(const PointerTuple&, const ValTuple&, uint, uint, const OpTuple&)
+        {
+        }
+
+#if CV_CUDEV_ARCH >= 300
+        template <class ValTuple, class OpTuple>
+        __device__ __forceinline__ static void mergeShfl(const ValTuple&, uint, uint, const OpTuple&)
+        {
+        }
+#endif
+    };
+
+    // loadToSmem
+
+    template <typename T>
+    __device__ __forceinline__ void loadToSmem(volatile T* smem, T& val, uint tid)
+    {
+        smem[tid] = val;
+    }
+
+    template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+    __device__ __forceinline__ void loadToSmem(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                               const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                               uint tid)
+    {
+        For<0, tuple_size<tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadToSmem(smem, val, tid);
+    }
+
+    // merge
+
+    template <typename T, class Op>
+    __device__ __forceinline__ void merge(volatile T* smem, T& val, uint tid, uint delta, const Op& op)
+    {
+        T reg = smem[tid + delta];
+        smem[tid] = val = op(val, reg);
+    }
+
+    template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void merge(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                          const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                          uint tid,
+                                          uint delta,
+                                          const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        For<0, tuple_size<tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::merge(smem, val, tid, delta, op);
+    }
+
+    // mergeShfl
+
+#if CV_CUDEV_ARCH >= 300
+    template <typename T, class Op>
+    __device__ __forceinline__ void mergeShfl(T& val, uint delta, uint width, const Op& op)
+    {
+        T reg = shfl_down(val, delta, width);
+        val = op(val, reg);
+    }
+
+    template <typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void mergeShfl(const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                              uint delta,
+                                              uint width,
+                                              const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        For<0, tuple_size<tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9> >::value>::mergeShfl(val, delta, width, op);
+    }
+#endif
+
+    // WarpReductor
+
+    struct WarpReductor
+    {
+        template <typename Pointer, typename Reference, class Op>
+        __device__ static void reduce(Pointer smem, Reference val, uint tid, Op op)
+        {
+        #if CV_CUDEV_ARCH >= 300
+            (void) smem;
+            (void) tid;
+
+            mergeShfl(val, 16, 32, op);
+            mergeShfl(val, 8, 32, op);
+            mergeShfl(val, 4, 32, op);
+            mergeShfl(val, 2, 32, op);
+            mergeShfl(val, 1, 32, op);
+        #else
+            loadToSmem(smem, val, tid);
+
+            if (tid < 16)
+            {
+                merge(smem, val, tid, 16, op);
+                merge(smem, val, tid, 8, op);
+                merge(smem, val, tid, 4, op);
+                merge(smem, val, tid, 2, op);
+                merge(smem, val, tid, 1, op);
+            }
+        #endif
+        }
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/detail/reduce_key_val.hpp b/modules/cudev/include/opencv2/cudev/warp/detail/reduce_key_val.hpp
new file mode 100644
index 0000000000..7701795757
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/detail/reduce_key_val.hpp
@@ -0,0 +1,239 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_REDUCE_KEY_VAL_DETAIL_HPP__
+#define __OPENCV_CUDEV_WARP_REDUCE_KEY_VAL_DETAIL_HPP__
+
+#include "../../common.hpp"
+#include "../../util/tuple.hpp"
+
+namespace cv { namespace cudev {
+
+namespace warp_reduce_key_val_detail
+{
+    // GetType
+
+    template <typename T> struct GetType;
+
+    template <typename T> struct GetType<T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<volatile T*>
+    {
+        typedef T type;
+    };
+
+    template <typename T> struct GetType<T&>
+    {
+        typedef T type;
+    };
+
+    // For
+
+    template <int I, int N> struct For
+    {
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void loadToSmem(const PointerTuple& smem, const ReferenceTuple& data, uint tid)
+        {
+            get<I>(smem)[tid] = get<I>(data);
+
+            For<I + 1, N>::loadToSmem(smem, data, tid);
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ static void copy(const PointerTuple& svals, const ReferenceTuple& val, uint tid, uint delta)
+        {
+            get<I>(svals)[tid] = get<I>(val) = get<I>(svals)[tid + delta];
+
+            For<I + 1, N>::copy(svals, val, tid, delta);
+        }
+
+        template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+        __device__ static void merge(const KeyPointerTuple& skeys, const KeyReferenceTuple& key,
+                                     const ValPointerTuple& svals, const ValReferenceTuple& val,
+                                     const CmpTuple& cmp,
+                                     uint tid, uint delta)
+        {
+            typename GetType<typename tuple_element<I, KeyPointerTuple>::type>::type reg = get<I>(skeys)[tid + delta];
+
+            if (get<I>(cmp)(reg, get<I>(key)))
+            {
+                get<I>(skeys)[tid] = get<I>(key) = reg;
+                get<I>(svals)[tid] = get<I>(val) = get<I>(svals)[tid + delta];
+            }
+
+            For<I + 1, N>::merge(skeys, key, svals, val, cmp, tid, delta);
+        }
+    };
+
+    template <int N> struct For<N, N>
+    {
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ __forceinline__ static void loadToSmem(const PointerTuple&, const ReferenceTuple&, uint)
+        {
+        }
+
+        template <class PointerTuple, class ReferenceTuple>
+        __device__ __forceinline__ static void copy(const PointerTuple&, const ReferenceTuple&, uint, uint)
+        {
+        }
+
+        template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+        __device__ __forceinline__ static void merge(const KeyPointerTuple&, const KeyReferenceTuple&,
+                                                     const ValPointerTuple&, const ValReferenceTuple&,
+                                                     const CmpTuple&,
+                                                     uint, uint)
+        {
+        }
+    };
+
+    // loadToSmem
+
+    template <typename T>
+    __device__ __forceinline__ void loadToSmem(volatile T* smem, T& data, uint tid)
+    {
+        smem[tid] = data;
+    }
+
+    template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+    __device__ __forceinline__ void loadToSmem(const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                               const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                               uint tid)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadToSmem(smem, data, tid);
+    }
+
+    // copyVals
+
+    template <typename V>
+    __device__ __forceinline__ void copyVals(volatile V* svals, V& val, uint tid, uint delta)
+    {
+        svals[tid] = val = svals[tid + delta];
+    }
+
+    template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+    __device__ __forceinline__ void copyVals(const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                             const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                             uint tid, uint delta)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::copy(svals, val, tid, delta);
+    }
+
+    // merge
+
+    template <typename K, typename V, class Cmp>
+    __device__ void merge(volatile K* skeys, K& key, volatile V* svals, V& val, const Cmp& cmp, uint tid, uint delta)
+    {
+        K reg = skeys[tid + delta];
+
+        if (cmp(reg, key))
+        {
+            skeys[tid] = key = reg;
+            copyVals(svals, val, tid, delta);
+        }
+    }
+
+    template <typename K,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp>
+    __device__ void merge(volatile K* skeys, K& key,
+                          const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                          const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                          const Cmp& cmp, uint tid, uint delta)
+    {
+        K reg = skeys[tid + delta];
+
+        if (cmp(reg, key))
+        {
+            skeys[tid] = key = reg;
+            copyVals(svals, val, tid, delta);
+        }
+    }
+
+    template <typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+              typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+    __device__ __forceinline__ void merge(const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                          const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                          const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                          const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                          const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                          uint tid, uint delta)
+    {
+        For<0, tuple_size<tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::merge(skeys, key, svals, val, cmp, tid, delta);
+    }
+
+    // WarpReductor
+
+    struct WarpReductor
+    {
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        __device__ static void reduce(KP skeys, KR key, VP svals, VR val, uint tid, Cmp cmp)
+        {
+            loadToSmem(skeys, key, tid);
+            loadToSmem(svals, val, tid);
+
+            if (tid < 16)
+            {
+                merge(skeys, key, svals, val, cmp, tid, 16);
+                merge(skeys, key, svals, val, cmp, tid, 8);
+                merge(skeys, key, svals, val, cmp, tid, 4);
+                merge(skeys, key, svals, val, cmp, tid, 2);
+                merge(skeys, key, svals, val, cmp, tid, 1);
+            }
+        }
+    };
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/reduce.hpp b/modules/cudev/include/opencv2/cudev/warp/reduce.hpp
new file mode 100644
index 0000000000..089ef92d0f
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/reduce.hpp
@@ -0,0 +1,206 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_REDUCE_HPP__
+#define __OPENCV_CUDEV_WARP_REDUCE_HPP__
+
+#include "../common.hpp"
+#include "../util/tuple.hpp"
+#include "detail/reduce.hpp"
+#include "detail/reduce_key_val.hpp"
+
+namespace cv { namespace cudev {
+
+// warpReduce
+
+template <typename T, class Op>
+__device__ __forceinline__ void warpReduce(volatile T* smem, T& val, uint tid, const Op& op)
+{
+    warp_reduce_detail::WarpReductor::template reduce<volatile T*, T&, const Op&>(smem, val, tid, op);
+}
+
+template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+          typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+          class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+__device__ __forceinline__ void warpReduce(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                           const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                           uint tid,
+                                           const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+{
+    warp_reduce_detail::WarpReductor::template reduce<
+            const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>&,
+            const tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>&,
+            const tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>&>(smem, val, tid, op);
+}
+
+// warpReduceKeyVal
+
+template <typename K, typename V, class Cmp>
+__device__ __forceinline__ void warpReduceKeyVal(volatile K* skeys, K& key, volatile V* svals, V& val, uint tid, const Cmp& cmp)
+{
+    warp_reduce_key_val_detail::WarpReductor::template reduce<volatile K*, K&, volatile V*, V&, const Cmp&>(skeys, key, svals, val, tid, cmp);
+}
+
+template <typename K,
+          typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+          typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+          class Cmp>
+__device__ __forceinline__ void warpReduceKeyVal(volatile K* skeys, K& key,
+                                                 const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 uint tid, const Cmp& cmp)
+{
+    warp_reduce_key_val_detail::WarpReductor::template reduce<volatile K*, K&,
+            const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+            const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+            const Cmp&>(skeys, key, svals, val, tid, cmp);
+}
+
+template <typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+          typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+          typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+          typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+          class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+__device__ __forceinline__ void warpReduceKeyVal(const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                                 const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                 const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 uint tid,
+                                                 const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp)
+{
+    warp_reduce_key_val_detail::WarpReductor::template reduce<
+            const tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>&,
+            const tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>&,
+            const tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+            const tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+            const tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>&
+            >(skeys, key, svals, val, tid, cmp);
+}
+
+// smem_tuple
+
+template <typename T0>
+__device__ __forceinline__
+tuple<volatile T0*>
+smem_tuple(T0* t0)
+{
+    return make_tuple((volatile T0*) t0);
+}
+
+template <typename T0, typename T1>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*>
+smem_tuple(T0* t0, T1* t1)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1);
+}
+
+template <typename T0, typename T1, typename T2>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*>
+smem_tuple(T0* t0, T1* t1, T2* t2)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2);
+}
+
+template <typename T0, typename T1, typename T2, typename T3>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8);
+}
+
+template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9>
+__device__ __forceinline__
+tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*, volatile T9*>
+smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8, T9* t9)
+{
+    return make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8, (volatile T9*) t9);
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/scan.hpp b/modules/cudev/include/opencv2/cudev/warp/scan.hpp
new file mode 100644
index 0000000000..acd032fb00
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/scan.hpp
@@ -0,0 +1,99 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_SCAN_HPP__
+#define __OPENCV_CUDEV_WARP_SCAN_HPP__
+
+#include "../common.hpp"
+#include "warp.hpp"
+#include "shuffle.hpp"
+
+namespace cv { namespace cudev {
+
+template <typename T>
+__device__ T warpScanInclusive(T data, volatile T* smem, uint tid)
+{
+#if CV_CUDEV_ARCH >= 300
+    (void) smem;
+    (void) tid;
+
+    const uint laneId = Warp::laneId();
+
+    // scan on shufl functions
+    #pragma unroll
+    for (int i = 1; i <= (WARP_SIZE / 2); i *= 2)
+    {
+        const T val = shfl_up(data, i);
+        if (laneId >= i)
+              data += val;
+    }
+
+    return data;
+#else
+    uint pos = 2 * tid - (tid & (WARP_SIZE - 1));
+    smem[pos] = 0;
+
+    pos += WARP_SIZE;
+    smem[pos] = data;
+
+    smem[pos] += smem[pos - 1];
+    smem[pos] += smem[pos - 2];
+    smem[pos] += smem[pos - 4];
+    smem[pos] += smem[pos - 8];
+    smem[pos] += smem[pos - 16];
+
+    return smem[pos];
+#endif
+}
+
+template <typename T>
+__device__ __forceinline__ T warpScanExclusive(T data, volatile T* smem, uint tid)
+{
+    return warpScanInclusive(data, smem, tid) - data;
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/shuffle.hpp b/modules/cudev/include/opencv2/cudev/warp/shuffle.hpp
new file mode 100644
index 0000000000..a6aae5b902
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/shuffle.hpp
@@ -0,0 +1,424 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_SHUFFLE_HPP__
+#define __OPENCV_CUDEV_WARP_SHUFFLE_HPP__
+
+#include "../common.hpp"
+#include "../util/vec_traits.hpp"
+
+namespace cv { namespace cudev {
+
+#if CV_CUDEV_ARCH >= 300
+
+// shfl
+
+__device__ __forceinline__ uchar shfl(uchar val, int srcLane, int width = warpSize)
+{
+    return (uchar) __shfl((int) val, srcLane, width);
+}
+
+__device__ __forceinline__ schar shfl(schar val, int srcLane, int width = warpSize)
+{
+    return (schar) __shfl((int) val, srcLane, width);
+}
+
+__device__ __forceinline__ ushort shfl(ushort val, int srcLane, int width = warpSize)
+{
+    return (ushort) __shfl((int) val, srcLane, width);
+}
+
+__device__ __forceinline__ short shfl(short val, int srcLane, int width = warpSize)
+{
+    return (short) __shfl((int) val, srcLane, width);
+}
+
+__device__ __forceinline__ int shfl(int val, int srcLane, int width = warpSize)
+{
+    return __shfl(val, srcLane, width);
+}
+
+__device__ __forceinline__ uint shfl(uint val, int srcLane, int width = warpSize)
+{
+    return (uint) __shfl((int) val, srcLane, width);
+}
+
+__device__ __forceinline__ float shfl(float val, int srcLane, int width = warpSize)
+{
+    return __shfl(val, srcLane, width);
+}
+
+__device__ double shfl(double val, int srcLane, int width = warpSize)
+{
+    int lo = __double2loint(val);
+    int hi = __double2hiint(val);
+
+    lo = __shfl(lo, srcLane, width);
+    hi = __shfl(hi, srcLane, width);
+
+    return __hiloint2double(hi, lo);
+}
+
+#define CV_CUDEV_SHFL_VEC_INST(input_type) \
+    __device__ __forceinline__ input_type ## 1 shfl(const input_type ## 1 & val, int srcLane, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 1>::make( \
+                        shfl(val.x, srcLane, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 2 shfl(const input_type ## 2 & val, int srcLane, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 2>::make( \
+                        shfl(val.x, srcLane, width), \
+                        shfl(val.y, srcLane, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 3 shfl(const input_type ## 3 & val, int srcLane, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 3>::make( \
+                        shfl(val.x, srcLane, width), \
+                        shfl(val.y, srcLane, width), \
+                        shfl(val.z, srcLane, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 4 shfl(const input_type ## 4 & val, int srcLane, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 4>::make( \
+                        shfl(val.x, srcLane, width), \
+                        shfl(val.y, srcLane, width), \
+                        shfl(val.z, srcLane, width), \
+                        shfl(val.w, srcLane, width) \
+                    ); \
+    }
+
+CV_CUDEV_SHFL_VEC_INST(uchar)
+CV_CUDEV_SHFL_VEC_INST(char)
+CV_CUDEV_SHFL_VEC_INST(ushort)
+CV_CUDEV_SHFL_VEC_INST(short)
+CV_CUDEV_SHFL_VEC_INST(uint)
+CV_CUDEV_SHFL_VEC_INST(int)
+CV_CUDEV_SHFL_VEC_INST(float)
+CV_CUDEV_SHFL_VEC_INST(double)
+
+#undef CV_CUDEV_SHFL_VEC_INST
+
+// shfl_up
+
+__device__ __forceinline__ uchar shfl_up(uchar val, uint delta, int width = warpSize)
+{
+    return (uchar) __shfl_up((int) val, delta, width);
+}
+
+__device__ __forceinline__ schar shfl_up(schar val, uint delta, int width = warpSize)
+{
+    return (schar) __shfl_up((int) val, delta, width);
+}
+
+__device__ __forceinline__ ushort shfl_up(ushort val, uint delta, int width = warpSize)
+{
+    return (ushort) __shfl_up((int) val, delta, width);
+}
+
+__device__ __forceinline__ short shfl_up(short val, uint delta, int width = warpSize)
+{
+    return (short) __shfl_up((int) val, delta, width);
+}
+
+__device__ __forceinline__ int shfl_up(int val, uint delta, int width = warpSize)
+{
+    return __shfl_up(val, delta, width);
+}
+
+__device__ __forceinline__ uint shfl_up(uint val, uint delta, int width = warpSize)
+{
+    return (uint) __shfl_up((int) val, delta, width);
+}
+
+__device__ __forceinline__ float shfl_up(float val, uint delta, int width = warpSize)
+{
+    return __shfl_up(val, delta, width);
+}
+
+__device__ double shfl_up(double val, uint delta, int width = warpSize)
+{
+    int lo = __double2loint(val);
+    int hi = __double2hiint(val);
+
+    lo = __shfl_up(lo, delta, width);
+    hi = __shfl_up(hi, delta, width);
+
+    return __hiloint2double(hi, lo);
+}
+
+#define CV_CUDEV_SHFL_UP_VEC_INST(input_type) \
+    __device__ __forceinline__ input_type ## 1 shfl_up(const input_type ## 1 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 1>::make( \
+                        shfl_up(val.x, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 2 shfl_up(const input_type ## 2 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 2>::make( \
+                        shfl_up(val.x, delta, width), \
+                        shfl_up(val.y, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 3 shfl_up(const input_type ## 3 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 3>::make( \
+                        shfl_up(val.x, delta, width), \
+                        shfl_up(val.y, delta, width), \
+                        shfl_up(val.z, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 4 shfl_up(const input_type ## 4 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 4>::make( \
+                        shfl_up(val.x, delta, width), \
+                        shfl_up(val.y, delta, width), \
+                        shfl_up(val.z, delta, width), \
+                        shfl_up(val.w, delta, width) \
+                    ); \
+    }
+
+CV_CUDEV_SHFL_UP_VEC_INST(uchar)
+CV_CUDEV_SHFL_UP_VEC_INST(char)
+CV_CUDEV_SHFL_UP_VEC_INST(ushort)
+CV_CUDEV_SHFL_UP_VEC_INST(short)
+CV_CUDEV_SHFL_UP_VEC_INST(uint)
+CV_CUDEV_SHFL_UP_VEC_INST(int)
+CV_CUDEV_SHFL_UP_VEC_INST(float)
+CV_CUDEV_SHFL_UP_VEC_INST(double)
+
+#undef CV_CUDEV_SHFL_UP_VEC_INST
+
+// shfl_down
+
+__device__ __forceinline__ uchar shfl_down(uchar val, uint delta, int width = warpSize)
+{
+    return (uchar) __shfl_down((int) val, delta, width);
+}
+
+__device__ __forceinline__ schar shfl_down(schar val, uint delta, int width = warpSize)
+{
+    return (schar) __shfl_down((int) val, delta, width);
+}
+
+__device__ __forceinline__ ushort shfl_down(ushort val, uint delta, int width = warpSize)
+{
+    return (ushort) __shfl_down((int) val, delta, width);
+}
+
+__device__ __forceinline__ short shfl_down(short val, uint delta, int width = warpSize)
+{
+    return (short) __shfl_down((int) val, delta, width);
+}
+
+__device__ __forceinline__ int shfl_down(int val, uint delta, int width = warpSize)
+{
+    return __shfl_down(val, delta, width);
+}
+
+__device__ __forceinline__ uint shfl_down(uint val, uint delta, int width = warpSize)
+{
+    return (uint) __shfl_down((int) val, delta, width);
+}
+
+__device__ __forceinline__ float shfl_down(float val, uint delta, int width = warpSize)
+{
+    return __shfl_down(val, delta, width);
+}
+
+__device__ double shfl_down(double val, uint delta, int width = warpSize)
+{
+    int lo = __double2loint(val);
+    int hi = __double2hiint(val);
+
+    lo = __shfl_down(lo, delta, width);
+    hi = __shfl_down(hi, delta, width);
+
+    return __hiloint2double(hi, lo);
+}
+
+#define CV_CUDEV_SHFL_DOWN_VEC_INST(input_type) \
+    __device__ __forceinline__ input_type ## 1 shfl_down(const input_type ## 1 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 1>::make( \
+                        shfl_down(val.x, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 2 shfl_down(const input_type ## 2 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 2>::make( \
+                        shfl_down(val.x, delta, width), \
+                        shfl_down(val.y, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 3 shfl_down(const input_type ## 3 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 3>::make( \
+                        shfl_down(val.x, delta, width), \
+                        shfl_down(val.y, delta, width), \
+                        shfl_down(val.z, delta, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 4 shfl_down(const input_type ## 4 & val, uint delta, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 4>::make( \
+                        shfl_down(val.x, delta, width), \
+                        shfl_down(val.y, delta, width), \
+                        shfl_down(val.z, delta, width), \
+                        shfl_down(val.w, delta, width) \
+                    ); \
+    }
+
+CV_CUDEV_SHFL_DOWN_VEC_INST(uchar)
+CV_CUDEV_SHFL_DOWN_VEC_INST(char)
+CV_CUDEV_SHFL_DOWN_VEC_INST(ushort)
+CV_CUDEV_SHFL_DOWN_VEC_INST(short)
+CV_CUDEV_SHFL_DOWN_VEC_INST(uint)
+CV_CUDEV_SHFL_DOWN_VEC_INST(int)
+CV_CUDEV_SHFL_DOWN_VEC_INST(float)
+CV_CUDEV_SHFL_DOWN_VEC_INST(double)
+
+#undef CV_CUDEV_SHFL_DOWN_VEC_INST
+
+// shfl_xor
+
+__device__ __forceinline__ uchar shfl_xor(uchar val, int laneMask, int width = warpSize)
+{
+    return (uchar) __shfl_xor((int) val, laneMask, width);
+}
+
+__device__ __forceinline__ schar shfl_xor(schar val, int laneMask, int width = warpSize)
+{
+    return (schar) __shfl_xor((int) val, laneMask, width);
+}
+
+__device__ __forceinline__ ushort shfl_xor(ushort val, int laneMask, int width = warpSize)
+{
+    return (ushort) __shfl_xor((int) val, laneMask, width);
+}
+
+__device__ __forceinline__ short shfl_xor(short val, int laneMask, int width = warpSize)
+{
+    return (short) __shfl_xor((int) val, laneMask, width);
+}
+
+__device__ __forceinline__ int shfl_xor(int val, int laneMask, int width = warpSize)
+{
+    return __shfl_xor(val, laneMask, width);
+}
+
+__device__ __forceinline__ uint shfl_xor(uint val, int laneMask, int width = warpSize)
+{
+    return (uint) __shfl_xor((int) val, laneMask, width);
+}
+
+__device__ __forceinline__ float shfl_xor(float val, int laneMask, int width = warpSize)
+{
+    return __shfl_xor(val, laneMask, width);
+}
+
+__device__ double shfl_xor(double val, int laneMask, int width = warpSize)
+{
+    int lo = __double2loint(val);
+    int hi = __double2hiint(val);
+
+    lo = __shfl_xor(lo, laneMask, width);
+    hi = __shfl_xor(hi, laneMask, width);
+
+    return __hiloint2double(hi, lo);
+}
+
+#define CV_CUDEV_SHFL_XOR_VEC_INST(input_type) \
+    __device__ __forceinline__ input_type ## 1 shfl_xor(const input_type ## 1 & val, int laneMask, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 1>::make( \
+                        shfl_xor(val.x, laneMask, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 2 shfl_xor(const input_type ## 2 & val, int laneMask, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 2>::make( \
+                        shfl_xor(val.x, laneMask, width), \
+                        shfl_xor(val.y, laneMask, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 3 shfl_xor(const input_type ## 3 & val, int laneMask, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 3>::make( \
+                        shfl_xor(val.x, laneMask, width), \
+                        shfl_xor(val.y, laneMask, width), \
+                        shfl_xor(val.z, laneMask, width) \
+                    ); \
+    } \
+    __device__ __forceinline__ input_type ## 4 shfl_xor(const input_type ## 4 & val, int laneMask, int width = warpSize) \
+    { \
+        return VecTraits<input_type ## 4>::make( \
+                        shfl_xor(val.x, laneMask, width), \
+                        shfl_xor(val.y, laneMask, width), \
+                        shfl_xor(val.z, laneMask, width), \
+                        shfl_xor(val.w, laneMask, width) \
+                    ); \
+    }
+
+CV_CUDEV_SHFL_XOR_VEC_INST(uchar)
+CV_CUDEV_SHFL_XOR_VEC_INST(char)
+CV_CUDEV_SHFL_XOR_VEC_INST(ushort)
+CV_CUDEV_SHFL_XOR_VEC_INST(short)
+CV_CUDEV_SHFL_XOR_VEC_INST(uint)
+CV_CUDEV_SHFL_XOR_VEC_INST(int)
+CV_CUDEV_SHFL_XOR_VEC_INST(float)
+CV_CUDEV_SHFL_XOR_VEC_INST(double)
+
+#undef CV_CUDEV_SHFL_XOR_VEC_INST
+
+#endif // CV_CUDEV_ARCH >= 300
+
+}}
+
+#endif
diff --git a/modules/cudev/include/opencv2/cudev/warp/warp.hpp b/modules/cudev/include/opencv2/cudev/warp/warp.hpp
new file mode 100644
index 0000000000..c7649880f3
--- /dev/null
+++ b/modules/cudev/include/opencv2/cudev/warp/warp.hpp
@@ -0,0 +1,122 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma once
+
+#ifndef __OPENCV_CUDEV_WARP_WARP_HPP__
+#define __OPENCV_CUDEV_WARP_WARP_HPP__
+
+#include "../common.hpp"
+
+namespace cv { namespace cudev {
+
+enum
+{
+    LOG_WARP_SIZE = 5,
+    WARP_SIZE     = 1 << LOG_WARP_SIZE
+};
+
+struct Warp
+{
+    __device__ __forceinline__ static uint laneId()
+    {
+        uint ret;
+        asm("mov.u32 %0, %laneid;" : "=r"(ret));
+        return ret;
+    }
+
+    __device__ __forceinline__ static uint warpId()
+    {
+        const uint tid = (threadIdx.z * blockDim.y + threadIdx.y) * blockDim.x + threadIdx.x;
+        return tid / WARP_SIZE;
+    }
+};
+
+template <class It, typename T>
+__device__ __forceinline__ void warpFill(It beg, It end, const T& value)
+{
+    for(It t = beg + Warp::laneId(); t < end; t += WARP_SIZE)
+        *t = value;
+}
+
+template <class InIt, class OutIt>
+__device__ __forceinline__ OutIt warpCopy(InIt beg, InIt end, OutIt out)
+{
+    for(InIt t = beg + Warp::laneId(); t < end; t += WARP_SIZE, out += WARP_SIZE)
+        *out = *t;
+    return out;
+}
+
+template <class InIt, class OutIt, class UnOp>
+__device__ __forceinline__ OutIt warpTransform(InIt beg, InIt end, OutIt out, const UnOp& op)
+{
+    for(InIt t = beg + Warp::laneId(); t < end; t += WARP_SIZE, out += WARP_SIZE)
+        *out = op(*t);
+    return out;
+}
+
+template <class InIt1, class InIt2, class OutIt, class BinOp>
+__device__ __forceinline__ OutIt warpTransform(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, const BinOp& op)
+{
+    uint lane = Warp::laneId();
+
+    InIt1 t1 = beg1 + lane;
+    InIt2 t2 = beg2 + lane;
+    for(; t1 < end1; t1 += WARP_SIZE, t2 += WARP_SIZE, out += WARP_SIZE)
+        *out = op(*t1, *t2);
+    return out;
+}
+
+template<typename OutIt, typename T>
+__device__ __forceinline__ void warpYota(OutIt beg, OutIt end, T value)
+{
+    uint lane = Warp::laneId();
+    value += lane;
+
+    for(OutIt t = beg + lane; t < end; t += WARP_SIZE, value += WARP_SIZE)
+        *t = value;
+}
+
+}}
+
+#endif
diff --git a/modules/cudev/src/stub.cpp b/modules/cudev/src/stub.cpp
new file mode 100644
index 0000000000..ec060adedc
--- /dev/null
+++ b/modules/cudev/src/stub.cpp
@@ -0,0 +1,11 @@
+#include <opencv2/core/cvdef.h>
+
+namespace cv { namespace cudev {
+
+CV_EXPORTS void stubFunc();
+
+}}
+
+void cv::cudev::stubFunc()
+{
+}
diff --git a/modules/cudev/test/CMakeLists.txt b/modules/cudev/test/CMakeLists.txt
new file mode 100644
index 0000000000..438e0a64c0
--- /dev/null
+++ b/modules/cudev/test/CMakeLists.txt
@@ -0,0 +1,48 @@
+set(test_deps opencv_cudev opencv_core opencv_imgproc opencv_highgui opencv_ts ${OPENCV_MODULE_opencv_ts_DEPS})
+
+ocv_check_dependencies(${test_deps})
+
+if(OCV_DEPENDENCIES_FOUND)
+  set(the_target "opencv_test_${name}")
+
+  ocv_module_include_directories("${test_deps}" "${the_module}")
+
+  file(GLOB test_srcs "${CMAKE_CURRENT_SOURCE_DIR}/*.cpp" "${CMAKE_CURRENT_SOURCE_DIR}/*.cu")
+  file(GLOB test_hdrs "${CMAKE_CURRENT_SOURCE_DIR}/*.hpp")
+  source_group("Src" FILES ${test_srcs})
+  source_group("Include" FILES ${test_hdrs})
+  set(OPENCV_TEST_${the_module}_SOURCES ${test_srcs} ${test_hdrs})
+
+  foreach(var CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_RELEASE CMAKE_CXX_FLAGS_DEBUG)
+    # we remove /EHa as it generates warnings under windows
+    string(REPLACE "/EHa" "" ${var} "${${var}}")
+
+    # we remove -ggdb3 flag as it leads to preprocessor errors when compiling CUDA files (CUDA 4.1)
+    string(REPLACE "-ggdb3" "" ${var} "${${var}}")
+
+    # we remove -Wsign-promo as it generates warnings under linux
+    string(REPLACE "-Wsign-promo" "" ${var} "${${var}}")
+
+    # we remove -fvisibility-inlines-hidden because it's used for C++ compiler
+    # but NVCC uses C compiler by default
+    string(REPLACE "-fvisibility-inlines-hidden" "" ${var} "${${var}}")
+  endforeach()
+
+  CUDA_ADD_EXECUTABLE(${the_target} ${OPENCV_TEST_${the_module}_SOURCES})
+  target_link_libraries(${the_target} ${test_deps} ${OPENCV_LINKER_LIBS} ${CUDA_LIBRARIES})
+  add_dependencies(opencv_tests ${the_target})
+
+  # Additional target properties
+  set_target_properties(${the_target} PROPERTIES
+    DEBUG_POSTFIX "${OPENCV_DEBUG_POSTFIX}"
+    RUNTIME_OUTPUT_DIRECTORY "${EXECUTABLE_OUTPUT_PATH}"
+  )
+
+  if(ENABLE_SOLUTION_FOLDERS)
+    set_target_properties(${the_target} PROPERTIES FOLDER "tests accuracy")
+  endif()
+
+  enable_testing()
+  get_target_property(LOC ${the_target} LOCATION)
+  add_test(${the_target} "${LOC}")
+endif()
diff --git a/modules/cudev/test/test_arithm_func.cu b/modules/cudev/test/test_arithm_func.cu
new file mode 100644
index 0000000000..bb73b04529
--- /dev/null
+++ b/modules/cudev/test/test_arithm_func.cu
@@ -0,0 +1,168 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// SqrtTest
+
+template <typename T>
+class SqrtTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst = sqrt_(d_src);
+
+        Mat dst_gold;
+        cv::sqrt(src, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_expr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = sqrt_(d_src1 * d_src2);
+
+        Mat dst_gold;
+        cv::multiply(src1, src2, dst_gold);
+        cv::sqrt(dst_gold, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(SqrtTest, float);
+
+TYPED_TEST(SqrtTest, GpuMat)
+{
+    SqrtTest<TypeParam>::test_gpumat();
+}
+
+TYPED_TEST(SqrtTest, Expr)
+{
+    SqrtTest<TypeParam>::test_expr();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// MagnitudeTest
+
+template <typename T>
+class MagnitudeTest : public ::testing::Test
+{
+public:
+    void test_accuracy()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst1 = hypot_(d_src1, d_src2);
+        GpuMat_<T> dst2 = magnitude_(d_src1, d_src2);
+        GpuMat_<T> dst3 = sqrt_(sqr_(d_src1) + sqr_(d_src2));
+
+        EXPECT_MAT_NEAR(dst1, dst2, 1e-4);
+        EXPECT_MAT_NEAR(dst2, dst3, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(MagnitudeTest, float);
+
+TYPED_TEST(MagnitudeTest, Accuracy)
+{
+    MagnitudeTest<TypeParam>::test_accuracy();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// PowTest
+
+template <typename T>
+class PowTest : public ::testing::Test
+{
+public:
+    void test_accuracy()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst1 = pow_(d_src, 0.5);
+        GpuMat_<T> dst2 = sqrt_(d_src);
+
+        EXPECT_MAT_NEAR(dst1, dst2, 1e-5);
+    }
+};
+
+TYPED_TEST_CASE(PowTest, float);
+
+TYPED_TEST(PowTest, Accuracy)
+{
+    PowTest<TypeParam>::test_accuracy();
+}
diff --git a/modules/cudev/test/test_arithm_op.cu b/modules/cudev/test/test_arithm_op.cu
new file mode 100644
index 0000000000..a904c54c7f
--- /dev/null
+++ b/modules/cudev/test/test_arithm_op.cu
@@ -0,0 +1,395 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+typedef ::testing::Types<uchar, ushort, short, int, float> AllTypes;
+typedef ::testing::Types<short, int, float> SignedTypes;
+
+////////////////////////////////////////////////////////////////////////////////
+// UnaryMinusTest
+
+template <typename T>
+class UnaryMinusTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst = -d_src;
+
+        Mat dst_gold;
+        src.convertTo(dst_gold, src.depth(), -1);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_globptr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+        GlobPtrSz<T> d_src_ptr = d_src;
+
+        GpuMat_<T> dst = -d_src_ptr;
+
+        Mat dst_gold;
+        src.convertTo(dst_gold, src.depth(), -1);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_texptr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+        Texture<T> tex_src(d_src);
+
+        GpuMat_<T> dst = -tex_src;
+
+        Mat dst_gold;
+        src.convertTo(dst_gold, src.depth(), -1);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_expr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = -(d_src1 + d_src2);
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+        dst_gold.convertTo(dst_gold, dst_gold.depth(), -1);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(UnaryMinusTest, SignedTypes);
+
+TYPED_TEST(UnaryMinusTest, GpuMat)
+{
+    UnaryMinusTest<TypeParam>::test_gpumat();
+}
+
+TYPED_TEST(UnaryMinusTest, GlobPtrSz)
+{
+    UnaryMinusTest<TypeParam>::test_globptr();
+}
+
+TYPED_TEST(UnaryMinusTest, TexturePtr)
+{
+    UnaryMinusTest<TypeParam>::test_texptr();
+}
+
+TYPED_TEST(UnaryMinusTest, Expr)
+{
+    UnaryMinusTest<TypeParam>::test_expr();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// PlusTest
+
+template <typename T>
+class PlusTest : public ::testing::Test
+{
+public:
+    void test_gpumat_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = d_src1 + d_src2;
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_texptr_scalar()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+        Texture<T> tex_src(d_src);
+
+        GpuMat_<T> dst = tex_src + static_cast<T>(5);
+
+        Mat dst_gold;
+        cv::add(src, 5, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_expr_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+        Mat src3 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2), d_src3(src3);
+
+        GpuMat_<T> dst = d_src1 + d_src2 + d_src3;
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+        cv::add(dst_gold, src3, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_scalar_expr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = static_cast<T>(5) + (d_src1 + d_src2);
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+        cv::add(dst_gold, 5, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(PlusTest, AllTypes);
+
+TYPED_TEST(PlusTest, GpuMat_GpuMat)
+{
+    PlusTest<TypeParam>::test_gpumat_gpumat();
+}
+
+TYPED_TEST(PlusTest, TexturePtr_Scalar)
+{
+    PlusTest<TypeParam>::test_texptr_scalar();
+}
+
+TYPED_TEST(PlusTest, Expr_GpuMat)
+{
+    PlusTest<TypeParam>::test_expr_gpumat();
+}
+
+TYPED_TEST(PlusTest, Scalar_Expr)
+{
+    PlusTest<TypeParam>::test_scalar_expr();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// MinusTest
+
+template <typename T>
+class MinusTest : public ::testing::Test
+{
+public:
+    void test_gpumat_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = d_src1 - d_src2;
+
+        Mat dst_gold;
+        cv::subtract(src1, src2, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_texptr_scalar()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+        Texture<T> tex_src(d_src);
+
+        GpuMat_<T> dst = tex_src - static_cast<T>(5);
+
+        Mat dst_gold;
+        cv::subtract(src, 5, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_expr_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+        Mat src3 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2), d_src3(src3);
+
+        GpuMat_<T> dst = (d_src1 + d_src2) - d_src3;
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+        cv::subtract(dst_gold, src3, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_scalar_expr()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = static_cast<T>(5) - (d_src1 + d_src2);
+
+        Mat dst_gold;
+        cv::add(src1, src2, dst_gold);
+        cv::subtract(5, dst_gold, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(MinusTest, SignedTypes);
+
+TYPED_TEST(MinusTest, GpuMat_GpuMat)
+{
+    MinusTest<TypeParam>::test_gpumat_gpumat();
+}
+
+TYPED_TEST(MinusTest, TexturePtr_Scalar)
+{
+    MinusTest<TypeParam>::test_texptr_scalar();
+}
+
+TYPED_TEST(MinusTest, Expr_GpuMat)
+{
+    MinusTest<TypeParam>::test_expr_gpumat();
+}
+
+TYPED_TEST(MinusTest, Scalar_Expr)
+{
+    MinusTest<TypeParam>::test_scalar_expr();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// AbsDiffTest
+
+template <typename T>
+class AbsDiffTest : public ::testing::Test
+{
+public:
+    void test_accuracy()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst1 = absdiff_(d_src1, d_src2);
+        GpuMat_<T> dst2 = abs_(d_src1 - d_src2);
+
+        EXPECT_MAT_NEAR(dst1, dst2, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(AbsDiffTest, SignedTypes);
+
+TYPED_TEST(AbsDiffTest, Accuracy)
+{
+    AbsDiffTest<TypeParam>::test_accuracy();
+}
diff --git a/modules/cudev/test/test_bitwize_op.cu b/modules/cudev/test/test_bitwize_op.cu
new file mode 100644
index 0000000000..908d46d66a
--- /dev/null
+++ b/modules/cudev/test/test_bitwize_op.cu
@@ -0,0 +1,146 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+typedef ::testing::Types<uchar, ushort, short, int> IntTypes;
+
+////////////////////////////////////////////////////////////////////////////////
+// BitNotTest
+
+template <typename T>
+class BitNotTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst = ~d_src;
+
+        Mat dst_gold;
+        cv::bitwise_not(src, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(BitNotTest, IntTypes);
+
+TYPED_TEST(BitNotTest, GpuMat)
+{
+    BitNotTest<TypeParam>::test_gpumat();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// BitAndTest
+
+template <typename T>
+class BitAndTest : public ::testing::Test
+{
+public:
+    void test_gpumat_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = d_src1 & d_src2;
+
+        Mat dst_gold;
+        cv::bitwise_and(src1, src2, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(BitAndTest, IntTypes);
+
+TYPED_TEST(BitAndTest, GpuMat_GpuMat)
+{
+    BitAndTest<TypeParam>::test_gpumat_gpumat();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// LShiftTest
+
+template <typename T>
+class LShiftTest : public ::testing::Test
+{
+public:
+    void test_accuracy()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst1 = d_src << 2;
+        GpuMat_<T> dst2 = d_src * 4;
+
+        EXPECT_MAT_NEAR(dst1, dst2, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(LShiftTest, int);
+
+TYPED_TEST(LShiftTest, Accuracy)
+{
+    LShiftTest<TypeParam>::test_accuracy();
+}
diff --git a/modules/cudev/test/test_cmp_op.cu b/modules/cudev/test/test_cmp_op.cu
new file mode 100644
index 0000000000..4d557b6528
--- /dev/null
+++ b/modules/cudev/test/test_cmp_op.cu
@@ -0,0 +1,151 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+typedef ::testing::Types<uchar, ushort, short, int, float> AllTypes;
+
+////////////////////////////////////////////////////////////////////////////////
+// LessTest
+
+template <typename T>
+class LessTest : public ::testing::Test
+{
+public:
+    void test_gpumat_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<uchar> dst = (d_src1 < d_src2) * 255;
+
+        Mat dst_gold;
+        cv::compare(src1, src2, dst_gold, CMP_LT);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(LessTest, AllTypes);
+
+TYPED_TEST(LessTest, GpuMat_GpuMat)
+{
+    LessTest<TypeParam>::test_gpumat_gpumat();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// MinTest
+
+template <typename T>
+class MinTest : public ::testing::Test
+{
+public:
+    void test_gpumat_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, type);
+        Mat src2 = randomMat(size, type);
+
+        GpuMat_<T> d_src1(src1), d_src2(src2);
+
+        GpuMat_<T> dst = min_(d_src1, d_src2);
+
+        Mat dst_gold;
+        cv::min(src1, src2, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(MinTest, AllTypes);
+
+TYPED_TEST(MinTest, GpuMat_GpuMat)
+{
+    MinTest<TypeParam>::test_gpumat_gpumat();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// ThreshBinaryTest
+
+typedef ::testing::Types<uchar, short, float> ThreshTypes;
+
+template <typename T>
+class ThreshBinaryTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst = threshBinary_(d_src, 128, 0);
+
+        Mat dst_gold;
+        cv::threshold(src, dst_gold, 128, 0, THRESH_BINARY);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(ThreshBinaryTest, ThreshTypes);
+
+TYPED_TEST(ThreshBinaryTest, GpuMat)
+{
+    ThreshBinaryTest<TypeParam>::test_gpumat();
+}
diff --git a/modules/cudev/test/test_color_cvt.cu b/modules/cudev/test/test_color_cvt.cu
new file mode 100644
index 0000000000..70d904b49e
--- /dev/null
+++ b/modules/cudev/test/test_color_cvt.cu
@@ -0,0 +1,180 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+namespace cv {
+
+enum {
+    COLOR_BGR2BGR = COLOR_BGR2RGB,
+    COLOR_BGR2LRGB = COLOR_BGR2RGB,
+    COLOR_BGR2LBGR = COLOR_BGR2RGB
+};
+
+}
+
+#define CVT_COLOR_TEST(src_space, dst_space, src_cn, dst_cn) \
+    TEST(CvtColor, src_space ## _to_ ## dst_space) \
+    { \
+        const Size size = randomSize(100, 400); \
+        Mat bgrb = randomMat(size, CV_8UC3); \
+        Mat srcb; \
+        cv::cvtColor(bgrb, srcb, COLOR_BGR ## 2 ## src_space, src_cn); \
+        GpuMat_<SelectIf<src_cn == 1, uchar, uchar ## src_cn>::type> d_srcb(srcb); \
+        GpuMat_<SelectIf<dst_cn == 1, uchar, uchar ## dst_cn>::type> dstb = src_space ## _to_ ## dst_space ## _(d_srcb); \
+        Mat dstb_gold; \
+        cv::cvtColor(srcb, dstb_gold, COLOR_ ## src_space ## 2 ## dst_space); \
+        EXPECT_MAT_NEAR(dstb_gold, dstb, 1.0); \
+        Mat bgrf = randomMat(size, CV_32FC3, 0, 1); \
+        Mat srcf; \
+        cv::cvtColor(bgrf, srcf, COLOR_BGR ## 2 ## src_space, src_cn); \
+        GpuMat_<SelectIf<src_cn == 1, float, float ## src_cn>::type> d_srcf(srcf); \
+        GpuMat_<SelectIf<dst_cn == 1, float, float ## dst_cn>::type> dstf = src_space ## _to_ ## dst_space ## _(d_srcf); \
+        Mat dstf_gold; \
+        cv::cvtColor(srcf, dstf_gold, COLOR_ ## src_space ## 2 ## dst_space); \
+        EXPECT_MAT_NEAR(dstf_gold, dstf, 1.0); \
+    }
+
+// RGB <-> BGR
+
+CVT_COLOR_TEST(BGR, RGB, 3, 3)
+CVT_COLOR_TEST(BGR, BGRA, 3, 4)
+CVT_COLOR_TEST(BGR, RGBA, 3, 4)
+CVT_COLOR_TEST(BGRA, BGR, 4, 3)
+CVT_COLOR_TEST(BGRA, RGB, 4, 3)
+CVT_COLOR_TEST(BGRA, RGBA, 4, 4)
+
+// RGB <-> Gray
+
+CVT_COLOR_TEST(BGR, GRAY, 3, 1)
+CVT_COLOR_TEST(RGB, GRAY, 3, 1)
+CVT_COLOR_TEST(BGRA, GRAY, 4, 1)
+CVT_COLOR_TEST(RGBA, GRAY, 4, 1)
+
+CVT_COLOR_TEST(GRAY, BGR, 1, 3)
+CVT_COLOR_TEST(GRAY, BGRA, 1, 4)
+
+// RGB <-> YUV
+
+CVT_COLOR_TEST(RGB, YUV, 3, 3)
+CVT_COLOR_TEST(BGR, YUV, 3, 3)
+
+CVT_COLOR_TEST(YUV, RGB, 3, 3)
+CVT_COLOR_TEST(YUV, BGR, 3, 3)
+
+// RGB <-> YCrCb
+
+CVT_COLOR_TEST(RGB, YCrCb, 3, 3)
+CVT_COLOR_TEST(BGR, YCrCb, 3, 3)
+
+CVT_COLOR_TEST(YCrCb, RGB, 3, 3)
+CVT_COLOR_TEST(YCrCb, BGR, 3, 3)
+
+// RGB <-> XYZ
+
+CVT_COLOR_TEST(RGB, XYZ, 3, 3)
+CVT_COLOR_TEST(BGR, XYZ, 3, 3)
+
+CVT_COLOR_TEST(XYZ, RGB, 3, 3)
+CVT_COLOR_TEST(XYZ, BGR, 3, 3)
+
+// RGB <-> HSV
+
+CVT_COLOR_TEST(RGB, HSV, 3, 3)
+CVT_COLOR_TEST(BGR, HSV, 3, 3)
+
+CVT_COLOR_TEST(HSV, RGB, 3, 3)
+CVT_COLOR_TEST(HSV, BGR, 3, 3)
+
+CVT_COLOR_TEST(RGB, HSV_FULL, 3, 3)
+CVT_COLOR_TEST(BGR, HSV_FULL, 3, 3)
+
+CVT_COLOR_TEST(HSV, RGB_FULL, 3, 3)
+CVT_COLOR_TEST(HSV, BGR_FULL, 3, 3)
+
+// RGB <-> HLS
+
+CVT_COLOR_TEST(RGB, HLS, 3, 3)
+CVT_COLOR_TEST(BGR, HLS, 3, 3)
+
+CVT_COLOR_TEST(HLS, RGB, 3, 3)
+CVT_COLOR_TEST(HLS, BGR, 3, 3)
+
+CVT_COLOR_TEST(RGB, HLS_FULL, 3, 3)
+CVT_COLOR_TEST(BGR, HLS_FULL, 3, 3)
+
+CVT_COLOR_TEST(HLS, RGB_FULL, 3, 3)
+CVT_COLOR_TEST(HLS, BGR_FULL, 3, 3)
+
+// RGB <-> Lab
+
+CVT_COLOR_TEST(RGB, Lab, 3, 3)
+CVT_COLOR_TEST(BGR, Lab, 3, 3)
+
+CVT_COLOR_TEST(Lab, RGB, 3, 3)
+CVT_COLOR_TEST(Lab, BGR, 3, 3)
+
+CVT_COLOR_TEST(LRGB, Lab, 3, 3)
+CVT_COLOR_TEST(LBGR, Lab, 3, 3)
+
+CVT_COLOR_TEST(Lab, LRGB, 3, 3)
+CVT_COLOR_TEST(Lab, LBGR, 3, 3)
+
+// RGB <-> Luv
+
+CVT_COLOR_TEST(RGB, Luv, 3, 3)
+CVT_COLOR_TEST(BGR, Luv, 3, 3)
+
+CVT_COLOR_TEST(Luv, RGB, 3, 3)
+CVT_COLOR_TEST(Luv, BGR, 3, 3)
+
+CVT_COLOR_TEST(LRGB, Luv, 3, 3)
+CVT_COLOR_TEST(LBGR, Luv, 3, 3)
+
+CVT_COLOR_TEST(Luv, LRGB, 3, 3)
+CVT_COLOR_TEST(Luv, LBGR, 3, 3)
diff --git a/modules/cudev/test/test_cvt.cu b/modules/cudev/test/test_cvt.cu
new file mode 100644
index 0000000000..bdfa493bdc
--- /dev/null
+++ b/modules/cudev/test/test_cvt.cu
@@ -0,0 +1,83 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+typedef ::testing::Types<uchar, ushort, short, int, float> AllTypes;
+
+////////////////////////////////////////////////////////////////////////////////
+// CvtTest
+
+template <typename T>
+class CvtTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+
+        GpuMat_<T> d_src(src);
+
+        GpuMat_<T> dst = cvt_<T>(cvt_<float>(d_src) * 2.0f - 10.0f);
+
+        Mat dst_gold;
+        src.convertTo(dst_gold, src.depth(), 2, -10);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(CvtTest, AllTypes);
+
+TYPED_TEST(CvtTest, GpuMat)
+{
+    CvtTest<TypeParam>::test_gpumat();
+}
diff --git a/modules/cudev/test/test_deriv.cu b/modules/cudev/test/test_deriv.cu
new file mode 100644
index 0000000000..8ef9fb76aa
--- /dev/null
+++ b/modules/cudev/test/test_deriv.cu
@@ -0,0 +1,109 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+TEST(Sobel, Accuracy)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+    Texture<uchar> tex_src(d_src);
+
+    GpuMat_<short> dx = sobelX_(cvt_<int>(tex_src));
+    GpuMat_<short> dy = sobelY_(cvt_<int>(tex_src));
+
+    Mat dx_gold, dy_gold;
+    cv::Sobel(src, dx_gold, CV_16S, 1, 0, 3, 1, 0, BORDER_REPLICATE);
+    cv::Sobel(src, dy_gold, CV_16S, 0, 1, 3, 1, 0, BORDER_REPLICATE);
+
+    EXPECT_MAT_NEAR(dx_gold, dx, 0.0);
+    EXPECT_MAT_NEAR(dy_gold, dy, 0.0);
+}
+
+TEST(Scharr, Accuracy)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+    Texture<uchar> tex_src(d_src);
+
+    GpuMat_<short> dx = scharrX_(cvt_<int>(tex_src));
+    GpuMat_<short> dy = scharrY_(cvt_<int>(tex_src));
+
+    Mat dx_gold, dy_gold;
+    cv::Scharr(src, dx_gold, CV_16S, 1, 0, 1, 0, BORDER_REPLICATE);
+    cv::Scharr(src, dy_gold, CV_16S, 0, 1, 1, 0, BORDER_REPLICATE);
+
+    EXPECT_MAT_NEAR(dx_gold, dx, 0.0);
+    EXPECT_MAT_NEAR(dy_gold, dy, 0.0);
+}
+
+TEST(Laplacian, Accuracy)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+    Texture<uchar> tex_src(d_src);
+
+    GpuMat_<short> dst1 = laplacian_<1>(cvt_<int>(tex_src));
+    GpuMat_<short> dst3 = laplacian_<3>(cvt_<int>(tex_src));
+
+    Mat dst1_gold, dst3_gold;
+    cv::Laplacian(src, dst1_gold, CV_16S, 1, 1, 0, BORDER_REPLICATE);
+    cv::Laplacian(src, dst3_gold, CV_16S, 3, 1, 0, BORDER_REPLICATE);
+
+    EXPECT_MAT_NEAR(dst1_gold, dst1, 0.0);
+    EXPECT_MAT_NEAR(dst3_gold, dst3, 0.0);
+}
diff --git a/modules/cudev/test/test_integral.cu b/modules/cudev/test/test_integral.cu
new file mode 100644
index 0000000000..190fc354ac
--- /dev/null
+++ b/modules/cudev/test/test_integral.cu
@@ -0,0 +1,103 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+TEST(Integral, _8u)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uint> dst = integral_(d_src);
+
+    Mat dst_gold;
+    cv::integral(src, dst_gold);
+
+    dst_gold = dst_gold(Rect(1, 1, size.width, size.height));
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(Integral, _32f)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+
+    GpuMat_<float> d_src(src);
+
+    GpuMat_<float> dst = integral_(d_src);
+
+    Mat dst_gold;
+    cv::integral(src, dst_gold, CV_32F);
+
+    dst_gold = dst_gold(Rect(1, 1, size.width, size.height));
+
+    ASSERT_PRED_FORMAT2(cvtest::MatComparator(1e-5, 0), dst_gold, Mat(dst));
+}
+
+TEST(Integral, _8u_opt)
+{
+    const Size size(640, 480);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uint> dst = integral_(d_src);
+
+    Mat dst_gold;
+    cv::integral(src, dst_gold);
+
+    dst_gold = dst_gold(Rect(1, 1, size.width, size.height));
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 0.0);
+}
diff --git a/modules/cudev/test/test_lut.cu b/modules/cudev/test/test_lut.cu
new file mode 100644
index 0000000000..d2548ec24b
--- /dev/null
+++ b/modules/cudev/test/test_lut.cu
@@ -0,0 +1,82 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// LutTest
+
+template <typename T>
+class LutTest : public ::testing::Test
+{
+public:
+    void test_gpumat()
+    {
+        const Size size = randomSize(100, 400);
+        const int type = DataType<T>::type;
+
+        Mat src = randomMat(size, type);
+        Mat tbl = randomMat(Size(256, 1), type);
+
+        GpuMat_<T> d_src(src), d_tbl(tbl);
+
+        GpuMat_<T> dst = lut_(d_src, d_tbl);
+
+        Mat dst_gold;
+        cv::LUT(src, tbl, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(LutTest, uchar);
+
+TYPED_TEST(LutTest, GpuMat)
+{
+    LutTest<TypeParam>::test_gpumat();
+}
diff --git a/modules/cudev/test/test_main.cpp b/modules/cudev/test/test_main.cpp
new file mode 100644
index 0000000000..fc7f8a3b3a
--- /dev/null
+++ b/modules/cudev/test/test_main.cpp
@@ -0,0 +1,46 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/ts.hpp"
+
+CV_TEST_MAIN("cv")
diff --git a/modules/cudev/test/test_precomp.hpp b/modules/cudev/test/test_precomp.hpp
new file mode 100644
index 0000000000..18e7cc1ce2
--- /dev/null
+++ b/modules/cudev/test/test_precomp.hpp
@@ -0,0 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/cudev.hpp"
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/cudev/test/test_pyramids.cu b/modules/cudev/test/test_pyramids.cu
new file mode 100644
index 0000000000..c196c923ca
--- /dev/null
+++ b/modules/cudev/test/test_pyramids.cu
@@ -0,0 +1,81 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+TEST(PyrDown, _8uc1)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = pyrDown_(d_src);
+
+    Mat dst_gold;
+    cv::pyrDown(src, dst_gold);
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 1.0);
+}
+
+TEST(PyrUp, _32fc4)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC4);
+
+    GpuMat_<float4> d_src(src);
+
+    GpuMat_<float4> dst = pyrDown_(d_src);
+
+    Mat dst_gold;
+    cv::pyrDown(src, dst_gold);
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 1e-4);
+}
diff --git a/modules/cudev/test/test_reduction.cu b/modules/cudev/test/test_reduction.cu
new file mode 100644
index 0000000000..22cadbebc4
--- /dev/null
+++ b/modules/cudev/test/test_reduction.cu
@@ -0,0 +1,312 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+TEST(Sum, GpuMat)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<float> dst = sum_(d_src);
+    float res;
+    dst.download(_OutputArray(&res, 1));
+
+    Scalar dst_gold = cv::sum(src);
+
+    ASSERT_FLOAT_EQ(static_cast<float>(dst_gold[0]), res);
+}
+
+TEST(Sum, Expr)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src1 = randomMat(size, CV_32FC1, 0, 1);
+    Mat src2 = randomMat(size, CV_32FC1, 0, 1);
+
+    GpuMat_<float> d_src1(src1), d_src2(src2);
+
+    GpuMat_<float> dst = sum_(abs_(d_src1 - d_src2));
+    float res;
+    dst.download(_OutputArray(&res, 1));
+
+    Scalar dst_gold = cv::norm(src1, src2, NORM_L1);
+
+    ASSERT_FLOAT_EQ(static_cast<float>(dst_gold[0]), res);
+}
+
+TEST(MinVal, GpuMat)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<float> dst = minVal_(d_src);
+    float res;
+    dst.download(_OutputArray(&res, 1));
+
+    double res_gold;
+    cv::minMaxLoc(src, &res_gold, 0);
+
+    ASSERT_FLOAT_EQ(static_cast<float>(res_gold), res);
+}
+
+TEST(MaxVal, Expr)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src1 = randomMat(size, CV_32SC1);
+    Mat src2 = randomMat(size, CV_32SC1);
+
+    GpuMat_<int> d_src1(src1), d_src2(src2);
+
+    GpuMat_<float> dst = maxVal_(abs_(d_src1 - d_src2));
+    float res;
+    dst.download(_OutputArray(&res, 1));
+
+    double res_gold = cv::norm(src1, src2, NORM_INF);
+
+    ASSERT_FLOAT_EQ(static_cast<float>(res_gold), res);
+}
+
+TEST(MinMaxVal, GpuMat)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<float> dst = minMaxVal_(d_src);
+    float res[2];
+    dst.download(Mat(1, 2, CV_32FC1, res));
+
+    double res_gold[2];
+    cv::minMaxLoc(src, &res_gold[0], &res_gold[1]);
+
+    ASSERT_FLOAT_EQ(static_cast<float>(res_gold[0]), res[0]);
+    ASSERT_FLOAT_EQ(static_cast<float>(res_gold[1]), res[1]);
+}
+
+TEST(NonZeroCount, Accuracy)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1, 0, 5);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<int> dst1 = countNonZero_(d_src);
+    GpuMat_<int> dst2 = sum_(cvt_<int>(d_src) != 0);
+
+    EXPECT_MAT_NEAR(dst1, dst2, 0.0);
+}
+
+TEST(ReduceToRow, Sum)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<int> dst = reduceToRow_<Sum<int> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 0, REDUCE_SUM, CV_32S);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(ReduceToRow, Avg)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<float> dst = reduceToRow_<Avg<float> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 0, REDUCE_AVG, CV_32F);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-4);
+}
+
+TEST(ReduceToRow, Min)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = reduceToRow_<Min<uchar> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 0, REDUCE_MIN);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(ReduceToRow, Max)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = reduceToRow_<Max<uchar> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 0, REDUCE_MAX);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(ReduceToColumn, Sum)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<int> dst = reduceToColumn_<Sum<int> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 1, REDUCE_SUM, CV_32S);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(ReduceToColumn, Avg)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<float> dst = reduceToColumn_<Avg<float> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 1, REDUCE_AVG, CV_32F);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-4);
+}
+
+TEST(ReduceToColumn, Min)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = reduceToColumn_<Min<uchar> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 1, REDUCE_MIN);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(ReduceToColumn, Max)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = reduceToColumn_<Max<uchar> >(d_src);
+
+    Mat dst_gold;
+    cv::reduce(src, dst_gold, 1, REDUCE_MAX);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+static void calcHistGold(const cv::Mat& src, cv::Mat& hist)
+{
+    hist.create(1, 256, CV_32SC1);
+    hist.setTo(cv::Scalar::all(0));
+
+    int* hist_row = hist.ptr<int>();
+    for (int y = 0; y < src.rows; ++y)
+    {
+        const uchar* src_row = src.ptr(y);
+
+        for (int x = 0; x < src.cols; ++x)
+            ++hist_row[src_row[x]];
+    }
+}
+
+TEST(Histogram, GpuMat)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<int> dst = histogram_<256>(d_src);
+
+    Mat dst_gold;
+    calcHistGold(src, dst_gold);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
diff --git a/modules/cudev/test/test_split_merge.cu b/modules/cudev/test/test_split_merge.cu
new file mode 100644
index 0000000000..3af24f5882
--- /dev/null
+++ b/modules/cudev/test/test_split_merge.cu
@@ -0,0 +1,180 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+typedef ::testing::Types<uchar, ushort, short, int, float> AllTypes;
+
+////////////////////////////////////////////////////////////////////////////////
+// MergeTest
+
+template <typename T>
+class MergeTest : public ::testing::Test
+{
+public:
+    void test_c2()
+    {
+        const Size size = randomSize(100, 400);
+
+        const int src_type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, src_type);
+        Mat src2 = randomMat(size, src_type);
+
+        GpuMat_<T> d_src1(src1);
+        GpuMat_<T> d_src2(src2);
+
+        GpuMat_<typename MakeVec<T, 2>::type> dst;
+        gridMerge(zipPtr(d_src1, d_src2), dst);
+
+        Mat dst_gold;
+        Mat srcs[] = {src1, src2};
+        cv::merge(srcs, 2, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+
+    void test_c3()
+    {
+        const Size size = randomSize(100, 400);
+
+        const int src_type = DataType<T>::type;
+
+        Mat src1 = randomMat(size, src_type);
+        Mat src2 = randomMat(size, src_type);
+        Mat src3 = randomMat(size, src_type);
+
+        GpuMat_<T> d_src1(src1);
+        GpuMat_<T> d_src2(src2);
+        GpuMat_<T> d_src3(src3);
+
+        GpuMat_<typename MakeVec<T, 3>::type> dst;
+        gridMerge(zipPtr(d_src1, d_src2, d_src3), dst);
+
+        Mat dst_gold;
+        Mat srcs[] = {src1, src2, src3};
+        cv::merge(srcs, 3, dst_gold);
+
+        ASSERT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(MergeTest, AllTypes);
+
+TYPED_TEST(MergeTest, C2)
+{
+    MergeTest<TypeParam>::test_c2();
+}
+
+TYPED_TEST(MergeTest, C3)
+{
+    MergeTest<TypeParam>::test_c3();
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// SplitTest
+
+template <typename T>
+class SplitTest : public ::testing::Test
+{
+public:
+    void test_c3()
+    {
+        const Size size = randomSize(100, 400);
+
+        const int src_type = CV_MAKE_TYPE(DataType<T>::depth, 3);
+
+        Mat src = randomMat(size, src_type);
+
+        GpuMat_<typename MakeVec<T, 3>::type> d_src(src);
+
+        GpuMat_<T> dst1, dst2, dst3;
+        gridSplit(d_src, tie(dst1, dst2, dst3));
+
+        std::vector<Mat> dst;
+        cv::split(src, dst);
+
+        ASSERT_MAT_NEAR(dst[0], dst1, 0.0);
+        ASSERT_MAT_NEAR(dst[1], dst2, 0.0);
+        ASSERT_MAT_NEAR(dst[2], dst3, 0.0);
+    }
+
+    void test_c4()
+    {
+        const Size size = randomSize(100, 400);
+
+        const int src_type = CV_MAKE_TYPE(DataType<T>::depth, 4);
+
+        Mat src = randomMat(size, src_type);
+
+        GpuMat_<typename MakeVec<T, 4>::type> d_src(src);
+
+        GpuMat_<T> dst1, dst2, dst3, dst4;
+        gridSplit(d_src, tie(dst1, dst2, dst3, dst4));
+
+        std::vector<Mat> dst;
+        cv::split(src, dst);
+
+        ASSERT_MAT_NEAR(dst[0], dst1, 0.0);
+        ASSERT_MAT_NEAR(dst[1], dst2, 0.0);
+        ASSERT_MAT_NEAR(dst[2], dst3, 0.0);
+        ASSERT_MAT_NEAR(dst[3], dst4, 0.0);
+    }
+};
+
+TYPED_TEST_CASE(SplitTest, AllTypes);
+
+TYPED_TEST(SplitTest, C3)
+{
+    SplitTest<TypeParam>::test_c3();
+}
+
+TYPED_TEST(SplitTest, C4)
+{
+    SplitTest<TypeParam>::test_c4();
+}
diff --git a/modules/cudev/test/test_warp.cu b/modules/cudev/test/test_warp.cu
new file mode 100644
index 0000000000..8777867bfc
--- /dev/null
+++ b/modules/cudev/test/test_warp.cu
@@ -0,0 +1,256 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+// remap
+
+enum { HALF_SIZE=0, UPSIDE_DOWN, REFLECTION_X, REFLECTION_BOTH };
+
+static void generateMap(Mat& mapx, Mat& mapy, int remapMode)
+{
+    for (int j = 0; j < mapx.rows; ++j)
+    {
+        for (int i = 0; i < mapx.cols; ++i)
+        {
+            switch (remapMode)
+            {
+            case HALF_SIZE:
+                if (i > mapx.cols*0.25 && i < mapx.cols*0.75 && j > mapx.rows*0.25 && j < mapx.rows*0.75)
+                {
+                    mapx.at<float>(j,i) = 2.f * (i - mapx.cols * 0.25f) + 0.5f;
+                    mapy.at<float>(j,i) = 2.f * (j - mapx.rows * 0.25f) + 0.5f;
+                }
+                else
+                {
+                    mapx.at<float>(j,i) = 0.f;
+                    mapy.at<float>(j,i) = 0.f;
+                }
+                break;
+            case UPSIDE_DOWN:
+                mapx.at<float>(j,i) = static_cast<float>(i);
+                mapy.at<float>(j,i) = static_cast<float>(mapx.rows - j);
+                break;
+            case REFLECTION_X:
+                mapx.at<float>(j,i) = static_cast<float>(mapx.cols - i);
+                mapy.at<float>(j,i) = static_cast<float>(j);
+                break;
+            case REFLECTION_BOTH:
+                mapx.at<float>(j,i) = static_cast<float>(mapx.cols - i);
+                mapy.at<float>(j,i) = static_cast<float>(mapx.rows - j);
+                break;
+            } // end of switch
+        }
+    }
+}
+
+static void test_remap(int remapMode)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+
+    Mat mapx(size, CV_32FC1);
+    Mat mapy(size, CV_32FC1);
+    generateMap(mapx, mapy, remapMode);
+
+    GpuMat_<float> d_src(src);
+    GpuMat_<float> d_mapx(mapx);
+    GpuMat_<float> d_mapy(mapy);
+
+    GpuMat_<float> dst = remap_(interNearest(brdReplicate(d_src)), d_mapx, d_mapy);
+
+    Mat dst_gold;
+    cv::remap(src, dst_gold, mapx, mapy, INTER_NEAREST, BORDER_REPLICATE);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(Remap, HALF_SIZE)
+{
+    test_remap(HALF_SIZE);
+}
+
+TEST(Remap, UPSIDE_DOWN)
+{
+    test_remap(UPSIDE_DOWN);
+}
+
+TEST(Remap, REFLECTION_X)
+{
+    test_remap(REFLECTION_X);
+}
+
+TEST(Remap, REFLECTION_BOTH)
+{
+    test_remap(REFLECTION_BOTH);
+}
+
+// resize
+
+TEST(Resize, Upscale)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+
+    GpuMat_<float> d_src(src);
+    Texture<float> tex_src(d_src);
+
+    GpuMat_<float> dst1 = resize_(interCubic(tex_src), 2, 2);
+
+    Mat mapx(size.height * 2, size.width * 2, CV_32FC1);
+    Mat mapy(size.height * 2, size.width * 2, CV_32FC1);
+
+    for (int y = 0; y < mapx.rows; ++y)
+    {
+        for (int x = 0; x < mapx.cols; ++x)
+        {
+            mapx.at<float>(y, x) = static_cast<float>(x / 2);
+            mapy.at<float>(y, x) = static_cast<float>(y / 2);
+        }
+    }
+
+    GpuMat_<float> d_mapx(mapx);
+    GpuMat_<float> d_mapy(mapy);
+
+    GpuMat_<float> dst2 = remap_(interCubic(brdReplicate(d_src)), d_mapx, d_mapy);
+
+    EXPECT_MAT_NEAR(dst1, dst2, 0.0);
+}
+
+TEST(Resize, Downscale)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+    const float fx = 1.0f / 3.0f;
+    const float fy = 1.0f / 3.0f;
+
+    GpuMat_<float> d_src(src);
+    Texture<float> tex_src(d_src);
+
+    GpuMat_<float> dst1 = resize_(interArea(tex_src, Size(3, 3)), fx, fy);
+
+    Mat mapx(cv::saturate_cast<int>(size.height * fy), cv::saturate_cast<int>(size.width * fx), CV_32FC1);
+    Mat mapy(cv::saturate_cast<int>(size.height * fy), cv::saturate_cast<int>(size.width * fx), CV_32FC1);
+
+    for (int y = 0; y < mapx.rows; ++y)
+    {
+        for (int x = 0; x < mapx.cols; ++x)
+        {
+            mapx.at<float>(y, x) = x / fx;
+            mapy.at<float>(y, x) = y / fy;
+        }
+    }
+
+    GpuMat_<float> d_mapx(mapx);
+    GpuMat_<float> d_mapy(mapy);
+
+    GpuMat_<float> dst2 = remap_(interArea(brdReplicate(d_src), Size(3, 3)), d_mapx, d_mapy);
+
+    EXPECT_MAT_NEAR(dst1, dst2, 0.0);
+}
+
+// warpAffine & warpPerspective
+
+Mat createAffineTransfomMatrix(Size srcSize, float angle, bool perspective)
+{
+    cv::Mat M(perspective ? 3 : 2, 3, CV_32FC1);
+
+    {
+        M.at<float>(0, 0) = std::cos(angle); M.at<float>(0, 1) = -std::sin(angle); M.at<float>(0, 2) = static_cast<float>(srcSize.width / 2);
+        M.at<float>(1, 0) = std::sin(angle); M.at<float>(1, 1) =  std::cos(angle); M.at<float>(1, 2) = 0.0f;
+    }
+    if (perspective)
+    {
+        M.at<float>(2, 0) = 0.0f           ; M.at<float>(2, 1) =  0.0f           ; M.at<float>(2, 2) = 1.0f;
+    }
+
+    return M;
+}
+
+TEST(WarpAffine, Rotation)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+    Mat M = createAffineTransfomMatrix(size, static_cast<float>(CV_PI / 4), false);
+
+    GpuMat_<float> d_src(src);
+    GpuMat_<float> d_M;
+    createContinuous(M.size(), M.type(), d_M);
+    d_M.upload(M);
+
+    GpuMat_<float> dst = warpAffine_(interNearest(brdConstant(d_src)), size, d_M);
+
+    Mat dst_gold;
+    cv::warpAffine(src, dst_gold, M, size, INTER_NEAREST | WARP_INVERSE_MAP);
+
+    EXPECT_MAT_SIMILAR(dst_gold, dst, 1e-3);
+}
+
+TEST(WarpPerspective, Rotation)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC1, 0, 1);
+    Mat M = createAffineTransfomMatrix(size, static_cast<float>(CV_PI / 4), true);
+
+    GpuMat_<float> d_src(src);
+    GpuMat_<float> d_M;
+    createContinuous(M.size(), M.type(), d_M);
+    d_M.upload(M);
+
+    GpuMat_<float> dst = warpPerspective_(interNearest(brdConstant(d_src)), size, d_M);
+
+    Mat dst_gold;
+    cv::warpPerspective(src, dst_gold, M, size, INTER_NEAREST | WARP_INVERSE_MAP);
+
+    EXPECT_MAT_SIMILAR(dst_gold, dst, 1e-3);
+}
diff --git a/modules/cudev/test/transpose.cu b/modules/cudev/test/transpose.cu
new file mode 100644
index 0000000000..a1e477c225
--- /dev/null
+++ b/modules/cudev/test/transpose.cu
@@ -0,0 +1,81 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::cudev;
+using namespace cvtest;
+
+TEST(Transpose, _8uc1)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_8UC1);
+
+    GpuMat_<uchar> d_src(src);
+
+    GpuMat_<uchar> dst = transpose_(d_src);
+
+    Mat dst_gold;
+    cv::transpose(src, dst_gold);
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+TEST(Transpose, _32fc3)
+{
+    const Size size = randomSize(100, 400);
+
+    Mat src = randomMat(size, CV_32FC3);
+
+    GpuMat_<float3> d_src(src);
+
+    GpuMat_<float3> dst = transpose_(d_src);
+
+    Mat dst_gold;
+    cv::transpose(src, dst_gold);
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 0.0);
+}
diff --git a/modules/features2d/CMakeLists.txt b/modules/features2d/CMakeLists.txt
index 7d36a58ac1..0b080cfb9a 100644
--- a/modules/features2d/CMakeLists.txt
+++ b/modules/features2d/CMakeLists.txt
@@ -1,3 +1,2 @@
 set(the_description "2D Features Framework")
 ocv_define_module(features2d opencv_imgproc opencv_flann OPTIONAL opencv_highgui)
-
diff --git a/modules/features2d/doc/common_interfaces_of_descriptor_extractors.rst b/modules/features2d/doc/common_interfaces_of_descriptor_extractors.rst
index 63e56f6786..83688c0a31 100644
--- a/modules/features2d/doc/common_interfaces_of_descriptor_extractors.rst
+++ b/modules/features2d/doc/common_interfaces_of_descriptor_extractors.rst
@@ -9,7 +9,10 @@ represented as vectors in a multidimensional space. All objects that implement t
 descriptor extractors inherit the
 :ocv:class:`DescriptorExtractor` interface.
 
+.. note::
 
+   * An example explaining keypoint extraction can be found at opencv_source_code/samples/cpp/descriptor_extractor_matcher.cpp
+   * An example on descriptor evaluation can be found at opencv_source_code/samples/cpp/detector_descriptor_evaluation.cpp
 
 DescriptorExtractor
 -------------------
@@ -33,7 +36,7 @@ Abstract base class for computing descriptors for image keypoints. ::
         virtual int descriptorSize() const = 0;
         virtual int descriptorType() const = 0;
 
-        static Ptr<DescriptorExtractor> create( const string& descriptorExtractorType );
+        static Ptr<DescriptorExtractor> create( const String& descriptorExtractorType );
 
     protected:
         ...
@@ -57,6 +60,8 @@ Computes the descriptors for a set of keypoints detected in an image (first vari
 
 .. ocv:function:: void DescriptorExtractor::compute( const vector<Mat>& images, vector<vector<KeyPoint> >& keypoints, vector<Mat>& descriptors ) const
 
+.. ocv:pyfunction:: cv2.DescriptorExtractor_create.compute(image, keypoints[, descriptors]) -> keypoints, descriptors
+
     :param image: Image.
 
     :param images: Image set.
@@ -70,7 +75,9 @@ DescriptorExtractor::create
 -------------------------------
 Creates a descriptor extractor by name.
 
-.. ocv:function:: Ptr<DescriptorExtractor>  DescriptorExtractor::create( const string& descriptorExtractorType )
+.. ocv:function:: Ptr<DescriptorExtractor>  DescriptorExtractor::create( const String& descriptorExtractorType )
+
+.. ocv:pyfunction:: cv2.DescriptorExtractor_create(descriptorExtractorType) -> retval
 
     :param descriptorExtractorType: Descriptor extractor type.
 
@@ -78,9 +85,10 @@ The current implementation supports the following types of a descriptor extracto
 
  * ``"SIFT"`` -- :ocv:class:`SIFT`
  * ``"SURF"`` -- :ocv:class:`SURF`
- * ``"ORB"`` -- :ocv:class:`ORB`
- * ``"BRISK"`` -- :ocv:class:`BRISK`
  * ``"BRIEF"`` -- :ocv:class:`BriefDescriptorExtractor`
+ * ``"BRISK"`` -- :ocv:class:`BRISK`
+ * ``"ORB"`` -- :ocv:class:`ORB`
+ * ``"FREAK"`` -- :ocv:class:`FREAK`
 
 A combined format is also supported: descriptor extractor adapter name ( ``"Opponent"`` --
 :ocv:class:`OpponentColorDescriptorExtractor` ) + descriptor extractor name (see above),
@@ -137,4 +145,6 @@ Strecha C., Fua P. *BRIEF: Binary Robust Independent Elementary Features* ,
         ...
     };
 
+.. note::
 
+   * A complete BRIEF extractor sample can be found at opencv_source_code/samples/cpp/brief_match_test.cpp
diff --git a/modules/features2d/doc/common_interfaces_of_descriptor_matchers.rst b/modules/features2d/doc/common_interfaces_of_descriptor_matchers.rst
index 8596ae43db..2c2cf28f8d 100644
--- a/modules/features2d/doc/common_interfaces_of_descriptor_matchers.rst
+++ b/modules/features2d/doc/common_interfaces_of_descriptor_matchers.rst
@@ -9,34 +9,11 @@ that are represented as vectors in a multidimensional space. All objects that im
 descriptor matchers inherit the
 :ocv:class:`DescriptorMatcher` interface.
 
-DMatch
-------
-.. ocv:struct:: DMatch
-
-Class for matching keypoint descriptors: query descriptor index,
-train descriptor index, train image index, and distance between descriptors. ::
-
-    struct DMatch
-    {
-        DMatch() : queryIdx(-1), trainIdx(-1), imgIdx(-1),
-                   distance(std::numeric_limits<float>::max()) {}
-        DMatch( int _queryIdx, int _trainIdx, float _distance ) :
-                queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(-1),
-                distance(_distance) {}
-        DMatch( int _queryIdx, int _trainIdx, int _imgIdx, float _distance ) :
-                queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(_imgIdx),
-                distance(_distance) {}
-
-        int queryIdx; // query descriptor index
-        int trainIdx; // train descriptor index
-        int imgIdx;   // train image index
-
-        float distance;
-
-        // less is better
-        bool operator<( const DMatch &m ) const;
-    };
+.. note::
 
+   * An example explaining keypoint matching can be found at opencv_source_code/samples/cpp/descriptor_extractor_matcher.cpp
+   * An example on descriptor matching evaluation can be found at opencv_source_code/samples/cpp/detector_descriptor_matcher_evaluation.cpp
+   * An example on one to many image matching can be found at opencv_source_code/samples/cpp/matching_to_many_images.cpp
 
 DescriptorMatcher
 -----------------
@@ -88,7 +65,7 @@ with an image set. ::
 
         virtual Ptr<DescriptorMatcher> clone( bool emptyTrainData=false ) const = 0;
 
-        static Ptr<DescriptorMatcher> create( const string& descriptorMatcherType );
+        static Ptr<DescriptorMatcher> create( const String& descriptorMatcherType );
 
     protected:
         vector<Mat> trainDescCollection;
@@ -217,7 +194,7 @@ For each query descriptor, finds the training descriptors not farther than the s
 
     :param compactResult: Parameter used when the mask (or masks) is not empty. If  ``compactResult``  is false, the  ``matches``  vector has the same size as  ``queryDescriptors``  rows. If  ``compactResult``  is true, the  ``matches``  vector does not contain matches for fully masked-out query descriptors.
 
-    :param maxDistance: Threshold for the distance between matched descriptors.
+    :param maxDistance: Threshold for the distance between matched descriptors. Distance means here metric distance (e.g. Hamming distance), not the distance between coordinates (which is measured in Pixels)!
 
 For each query descriptor, the methods find such training descriptors that the distance between the query descriptor and the training descriptor is equal or smaller than ``maxDistance``. Found matches are returned in the distance increasing order.
 
@@ -237,7 +214,7 @@ DescriptorMatcher::create
 -----------------------------
 Creates a descriptor matcher of a given type with the default parameters (using default constructor).
 
-.. ocv:function:: Ptr<DescriptorMatcher> DescriptorMatcher::create( const string& descriptorMatcherType )
+.. ocv:function:: Ptr<DescriptorMatcher> DescriptorMatcher::create( const String& descriptorMatcherType )
 
     :param descriptorMatcherType: Descriptor matcher type. Now the following matcher types are supported:
 
@@ -299,4 +276,3 @@ Flann-based descriptor matcher. This matcher trains :ocv:class:`flann::Index_` o
     };
 
 ..
-
diff --git a/modules/features2d/doc/common_interfaces_of_feature_detectors.rst b/modules/features2d/doc/common_interfaces_of_feature_detectors.rst
index 81c72d3a81..434585d1e4 100644
--- a/modules/features2d/doc/common_interfaces_of_feature_detectors.rst
+++ b/modules/features2d/doc/common_interfaces_of_feature_detectors.rst
@@ -8,64 +8,9 @@ between different algorithms solving the same problem. All objects that implemen
 inherit the
 :ocv:class:`FeatureDetector` interface.
 
-KeyPoint
---------
-.. ocv:class:: KeyPoint
-
-  Data structure for salient point detectors.
-
-  .. ocv:member:: Point2f pt
-
-     coordinates of the keypoint
-
-  .. ocv:member:: float size
-
-     diameter of the meaningful keypoint neighborhood
-
-  .. ocv:member:: float angle
-
-     computed orientation of the keypoint (-1 if not applicable). Its possible values are in a range [0,360) degrees. It is measured relative to image coordinate system (y-axis is directed downward), ie in clockwise.
-
-  .. ocv:member:: float response
-
-     the response by which the most strong keypoints have been selected. Can be used for further sorting or subsampling
-
-  .. ocv:member:: int octave
-
-     octave (pyramid layer) from which the keypoint has been extracted
-
-  .. ocv:member:: int class_id
-
-     object id that can be used to clustered keypoints by an object they belong to
-
-KeyPoint::KeyPoint
-------------------
-The keypoint constructors
-
-.. ocv:function:: KeyPoint::KeyPoint()
-
-.. ocv:function:: KeyPoint::KeyPoint(Point2f _pt, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1)
-
-.. ocv:function:: KeyPoint::KeyPoint(float x, float y, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1)
-
-.. ocv:pyfunction:: cv2.KeyPoint([x, y, _size[, _angle[, _response[, _octave[, _class_id]]]]]) -> <KeyPoint object>
-
-    :param x: x-coordinate of the keypoint
-
-    :param y: y-coordinate of the keypoint
-
-    :param _pt: x & y coordinates of the keypoint
-
-    :param _size: keypoint diameter
-
-    :param _angle: keypoint orientation
-
-    :param _response: keypoint detector response on the keypoint (that is, strength of the keypoint)
-
-    :param _octave: pyramid octave in which the keypoint has been detected
-
-    :param _class_id: object id
+.. note::
 
+   * An example explaining keypoint detection can be found at opencv_source_code/samples/cpp/descriptor_extractor_matcher.cpp
 
 FeatureDetector
 ---------------
@@ -88,7 +33,7 @@ Abstract base class for 2D image feature detectors. ::
         virtual void read(const FileNode&);
         virtual void write(FileStorage&) const;
 
-        static Ptr<FeatureDetector> create( const string& detectorType );
+        static Ptr<FeatureDetector> create( const String& detectorType );
 
     protected:
     ...
@@ -102,6 +47,8 @@ Detects keypoints in an image (first variant) or image set (second variant).
 
 .. ocv:function:: void FeatureDetector::detect( const vector<Mat>& images, vector<vector<KeyPoint> >& keypoints, const vector<Mat>& masks=vector<Mat>() ) const
 
+.. ocv:pyfunction:: cv2.FeatureDetector_create.detect(image[, mask]) -> keypoints
+
     :param image: Image.
 
     :param images: Image set.
@@ -116,7 +63,9 @@ FeatureDetector::create
 -----------------------
 Creates a feature detector by its name.
 
-.. ocv:function:: Ptr<FeatureDetector> FeatureDetector::create( const string& detectorType )
+.. ocv:function:: Ptr<FeatureDetector> FeatureDetector::create( const String& detectorType )
+
+.. ocv:pyfunction:: cv2.FeatureDetector_create(detectorType) -> retval
 
     :param detectorType: Feature detector type.
 
@@ -220,7 +169,7 @@ StarFeatureDetector
 -------------------
 .. ocv:class:: StarFeatureDetector : public FeatureDetector
 
-The class implements the keypoint detector introduced by K. Konolige, synonym of ``StarDetector``.  ::
+The class implements the keypoint detector introduced by [Agrawal08]_, synonym of ``StarDetector``.  ::
 
     class StarFeatureDetector : public FeatureDetector
     {
@@ -234,6 +183,9 @@ The class implements the keypoint detector introduced by K. Konolige, synonym of
         ...
     };
 
+.. [Agrawal08] Agrawal, M., Konolige, K., & Blas, M. R. (2008). Censure: Center surround extremas for realtime feature detection and matching. In Computer Vision–ECCV 2008 (pp. 102-115). Springer Berlin Heidelberg.
+
+
 DenseFeatureDetector
 --------------------
 .. ocv:class:: DenseFeatureDetector : public FeatureDetector
@@ -439,7 +391,7 @@ Class providing an interface for adjusting parameters of a feature detector. Thi
         virtual void tooMany(int max, int n_detected) = 0;
         virtual bool good() const = 0;
         virtual Ptr<AdjusterAdapter> clone() const = 0;
-        static Ptr<AdjusterAdapter> create( const string& detectorType );
+        static Ptr<AdjusterAdapter> create( const String& detectorType );
      };
 
 
@@ -500,7 +452,7 @@ AdjusterAdapter::create
 -----------------------
 Creates an adjuster adapter by name
 
-.. ocv:function:: Ptr<AdjusterAdapter> AdjusterAdapter::create( const string& detectorType )
+.. ocv:function:: Ptr<AdjusterAdapter> AdjusterAdapter::create( const String& detectorType )
 
     Creates an adjuster adapter by name ``detectorType``. The detector name is the same as in :ocv:func:`FeatureDetector::create`, but now supports ``"FAST"``, ``"STAR"``, and ``"SURF"`` only.
 
diff --git a/modules/features2d/doc/common_interfaces_of_generic_descriptor_matchers.rst b/modules/features2d/doc/common_interfaces_of_generic_descriptor_matchers.rst
index a306c66067..5a7f952bc2 100644
--- a/modules/features2d/doc/common_interfaces_of_generic_descriptor_matchers.rst
+++ b/modules/features2d/doc/common_interfaces_of_generic_descriptor_matchers.rst
@@ -11,7 +11,11 @@ Every descriptor with the
 :ocv:class:`VectorDescriptorMatcher` ).
 There are descriptors such as the One-way descriptor and Ferns that have the ``GenericDescriptorMatcher`` interface implemented but do not support ``DescriptorExtractor``.
 
+.. note::
 
+   * An example explaining keypoint description can be found at opencv_source_code/samples/cpp/descriptor_extractor_matcher.cpp
+   * An example on descriptor matching evaluation can be found at opencv_source_code/samples/cpp/detector_descriptor_matcher_evaluation.cpp
+   * An example on one to many image matching can be found at opencv_source_code/samples/cpp/matching_to_many_images.cpp
 
 GenericDescriptorMatcher
 ------------------------
@@ -270,5 +274,3 @@ Example: ::
 
     VectorDescriptorMatcher matcher( new SurfDescriptorExtractor,
                                      new BruteForceMatcher<L2<float> > );
-
-
diff --git a/modules/features2d/doc/drawing_function_of_keypoints_and_matches.rst b/modules/features2d/doc/drawing_function_of_keypoints_and_matches.rst
index 2669ab9f27..68c68fc6c4 100644
--- a/modules/features2d/doc/drawing_function_of_keypoints_and_matches.rst
+++ b/modules/features2d/doc/drawing_function_of_keypoints_and_matches.rst
@@ -11,6 +11,10 @@ Draws the found matches of keypoints from two images.
 
 .. ocv:function:: void drawMatches( const Mat& img1, const vector<KeyPoint>& keypoints1, const Mat& img2, const vector<KeyPoint>& keypoints2, const vector<vector<DMatch> >& matches1to2, Mat& outImg, const Scalar& matchColor=Scalar::all(-1), const Scalar& singlePointColor=Scalar::all(-1), const vector<vector<char> >& matchesMask=vector<vector<char> >(), int flags=DrawMatchesFlags::DEFAULT )
 
+.. ocv:pyfunction:: cv2.drawMatches(img1, keypoints1, img2, keypoints2, matches1to2[, outImg[, matchColor[, singlePointColor[, matchesMask[, flags]]]]]) -> outImg
+
+.. ocv:pyfunction:: cv2.drawMatchesKnn(img1, keypoints1, img2, keypoints2, matches1to2[, outImg[, matchColor[, singlePointColor[, matchesMask[, flags]]]]]) -> outImg
+
 
     :param img1: First source image.
 
@@ -67,6 +71,8 @@ Draws keypoints.
 
 .. ocv:function:: void drawKeypoints( const Mat& image, const vector<KeyPoint>& keypoints, Mat& outImage, const Scalar& color=Scalar::all(-1), int flags=DrawMatchesFlags::DEFAULT )
 
+.. ocv:pyfunction:: cv2.drawKeypoints(image, keypoints[, outImage[, color[, flags]]]) -> outImage
+
     :param image: Source image.
 
     :param keypoints: Keypoints from the source image.
@@ -77,3 +83,4 @@ Draws keypoints.
 
     :param flags: Flags setting drawing features. Possible  ``flags``  bit values are defined by  ``DrawMatchesFlags``. See details above in  :ocv:func:`drawMatches` .
 
+.. note:: For Python API, flags are modified as `cv2.DRAW_MATCHES_FLAGS_DEFAULT`, `cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS`, `cv2.DRAW_MATCHES_FLAGS_DRAW_OVER_OUTIMG`, `cv2.DRAW_MATCHES_FLAGS_NOT_DRAW_SINGLE_POINTS`
diff --git a/modules/features2d/doc/feature_detection_and_description.rst b/modules/features2d/doc/feature_detection_and_description.rst
index 659fa4c3b2..a6fe7c8fad 100644
--- a/modules/features2d/doc/feature_detection_and_description.rst
+++ b/modules/features2d/doc/feature_detection_and_description.rst
@@ -3,6 +3,10 @@ Feature Detection and Description
 
 .. highlight:: cpp
 
+.. note::
+
+   * An example explaining keypoint detection and description can be found at opencv_source_code/samples/cpp/descriptor_extractor_matcher.cpp
+
 FAST
 ----
 Detects corners using the FAST algorithm
@@ -10,6 +14,11 @@ Detects corners using the FAST algorithm
 .. ocv:function:: void FAST( InputArray image, vector<KeyPoint>& keypoints, int threshold, bool nonmaxSupression=true )
 .. ocv:function:: void FAST( InputArray image, vector<KeyPoint>& keypoints, int threshold, bool nonmaxSupression, int type )
 
+.. ocv:pyfunction:: cv2.FastFeatureDetector([, threshold[, nonmaxSuppression]]) -> <FastFeatureDetector object>
+.. ocv:pyfunction:: cv2.FastFeatureDetector(threshold, nonmaxSuppression, type) -> <FastFeatureDetector object>
+.. ocv:pyfunction:: cv2.FastFeatureDetector.detect(image[, mask]) -> keypoints
+
+
     :param image: grayscale image where keypoints (corners) are detected.
 
     :param keypoints: keypoints detected on the image.
@@ -22,6 +31,9 @@ Detects corners using the FAST algorithm
 
 Detects corners using the FAST algorithm by [Rosten06]_.
 
+..note:: In Python API, types are given as ``cv2.FAST_FEATURE_DETECTOR_TYPE_5_8``, ``cv2.FAST_FEATURE_DETECTOR_TYPE_7_12`` and  ``cv2.FAST_FEATURE_DETECTOR_TYPE_9_16``. For corner detection, use ``cv2.FAST.detect()`` method.
+
+
 .. [Rosten06] E. Rosten. Machine Learning for High-speed Corner Detection, 2006.
 
 
@@ -48,7 +60,11 @@ Maximally stable extremal region extractor. ::
     };
 
 The class encapsulates all the parameters of the MSER extraction algorithm (see
-http://en.wikipedia.org/wiki/Maximally_stable_extremal_regions). Also see http://opencv.willowgarage.com/wiki/documentation/cpp/features2d/MSER for useful comments and parameters description.
+http://en.wikipedia.org/wiki/Maximally_stable_extremal_regions). Also see http://code.opencv.org/projects/opencv/wiki/MSER for useful comments and parameters description.
+
+.. note::
+
+   * (Python) A complete example showing the use of the MSER detector can be found at opencv_source_code/samples/python2/mser.py
 
 
 ORB
@@ -65,6 +81,9 @@ The ORB constructor
 
 .. ocv:function:: ORB::ORB(int nfeatures = 500, float scaleFactor = 1.2f, int nlevels = 8, int edgeThreshold = 31, int firstLevel = 0, int WTA_K=2, int scoreType=ORB::HARRIS_SCORE, int patchSize=31)
 
+.. ocv:pyfunction:: cv2.ORB([, nfeatures[, scaleFactor[, nlevels[, edgeThreshold[, firstLevel[, WTA_K[, scoreType[, patchSize]]]]]]]]) -> <ORB object>
+
+
     :param nfeatures: The maximum number of features to retain.
 
     :param scaleFactor: Pyramid decimation ratio, greater than 1. ``scaleFactor==2`` means the classical pyramid, where each next level has 4x less pixels than the previous, but such a big scale factor will degrade feature matching scores dramatically. On the other hand, too close to 1 scale factor will mean that to cover certain scale range you will need more pyramid levels and so the speed will suffer.
@@ -87,6 +106,11 @@ Finds keypoints in an image and computes their descriptors
 
 .. ocv:function:: void ORB::operator()(InputArray image, InputArray mask, vector<KeyPoint>& keypoints, OutputArray descriptors, bool useProvidedKeypoints=false ) const
 
+.. ocv:pyfunction:: cv2.ORB.detect(image[, mask]) -> keypoints
+.. ocv:pyfunction:: cv2.ORB.compute(image, keypoints[, descriptors]) -> keypoints, descriptors
+.. ocv:pyfunction:: cv2.ORB.detectAndCompute(image, mask[, descriptors[, useProvidedKeypoints]]) -> keypoints, descriptors
+
+
     :param image: The input 8-bit grayscale image.
 
     :param mask: The operation mask.
@@ -97,6 +121,7 @@ Finds keypoints in an image and computes their descriptors
 
     :param useProvidedKeypoints: If it is true, then the method will use the provided vector of keypoints instead of detecting them.
 
+
 BRISK
 -----
 .. ocv:class:: BRISK : public Feature2D
@@ -111,6 +136,8 @@ The BRISK constructor
 
 .. ocv:function:: BRISK::BRISK(int thresh=30, int octaves=3, float patternScale=1.0f)
 
+.. ocv:pyfunction:: cv2.BRISK([, thresh[, octaves[, patternScale]]]) -> <BRISK object>
+
     :param thresh: FAST/AGAST detection threshold score.
 
     :param octaves: detection octaves. Use 0 to do single scale.
@@ -123,6 +150,8 @@ The BRISK constructor for a custom pattern
 
 .. ocv:function:: BRISK::BRISK(std::vector<float> &radiusList, std::vector<int> &numberList, float dMax=5.85f, float dMin=8.2f, std::vector<int> indexChange=std::vector<int>())
 
+.. ocv:pyfunction:: cv2.BRISK(radiusList, numberList[, dMax[, dMin[, indexChange]]]) -> <BRISK object>
+
     :param radiusList: defines the radii (in pixels) where the samples around a keypoint are taken (for keypoint scale 1).
 
     :param numberList: defines the number of sampling points on the sampling circle. Must be the same size as radiusList..
@@ -139,6 +168,10 @@ Finds keypoints in an image and computes their descriptors
 
 .. ocv:function:: void BRISK::operator()(InputArray image, InputArray mask, vector<KeyPoint>& keypoints, OutputArray descriptors, bool useProvidedKeypoints=false ) const
 
+.. ocv:pyfunction:: cv2.BRISK.detect(image[, mask]) -> keypoints
+.. ocv:pyfunction:: cv2.BRISK.compute(image, keypoints[, descriptors]) -> keypoints, descriptors
+.. ocv:pyfunction:: cv2.BRISK.detectAndCompute(image, mask[, descriptors[, useProvidedKeypoints]]) -> keypoints, descriptors
+
     :param image: The input 8-bit grayscale image.
 
     :param mask: The operation mask.
@@ -157,6 +190,10 @@ Class implementing the FREAK (*Fast Retina Keypoint*) keypoint descriptor, descr
 
 .. [AOV12] A. Alahi, R. Ortiz, and P. Vandergheynst. FREAK: Fast Retina Keypoint. In IEEE Conference on Computer Vision and Pattern Recognition, 2012. CVPR 2012 Open Source Award Winner.
 
+.. note::
+
+   * An example on how to use the FREAK descriptor can be found at opencv_source_code/samples/cpp/freak_demo.cpp
+
 FREAK::FREAK
 ------------
 The FREAK constructor
diff --git a/modules/features2d/doc/object_categorization.rst b/modules/features2d/doc/object_categorization.rst
index 73089ce7ad..644634fd06 100644
--- a/modules/features2d/doc/object_categorization.rst
+++ b/modules/features2d/doc/object_categorization.rst
@@ -5,6 +5,12 @@ Object Categorization
 
 This section describes approaches based on local 2D features and used to categorize objects.
 
+.. note::
+
+   * A complete Bag-Of-Words sample can be found at opencv_source_code/samples/cpp/bagofwords_classification.cpp
+
+   * (Python) An example using the features2D framework to perform object categorization can be found at opencv_source_code/samples/python2/find_obj.py
+
 BOWTrainer
 ----------
 .. ocv:class:: BOWTrainer
@@ -198,4 +204,3 @@ BOWImgDescriptorExtractor::descriptorType
 Returns an image descriptor type.
 
 .. ocv:function:: int BOWImgDescriptorExtractor::descriptorType() const
-
diff --git a/modules/features2d/include/opencv2/features2d.hpp b/modules/features2d/include/opencv2/features2d.hpp
index 8f32445ca2..af687e3085 100644
--- a/modules/features2d/include/opencv2/features2d.hpp
+++ b/modules/features2d/include/opencv2/features2d.hpp
@@ -46,72 +46,15 @@
 #include "opencv2/core.hpp"
 #include "opencv2/flann/miniflann.hpp"
 
-#ifdef __cplusplus
-#include <limits>
-
 namespace cv
 {
 
 CV_EXPORTS bool initModule_features2d();
 
-/*!
- The Keypoint Class
-
- The class instance stores a keypoint, i.e. a point feature found by one of many available keypoint detectors, such as
- Harris corner detector, cv::FAST, cv::StarDetector, cv::SURF, cv::SIFT, cv::LDetector etc.
-
- The keypoint is characterized by the 2D position, scale
- (proportional to the diameter of the neighborhood that needs to be taken into account),
- orientation and some other parameters. The keypoint neighborhood is then analyzed by another algorithm that builds a descriptor
- (usually represented as a feature vector). The keypoints representing the same object in different images can then be matched using
- cv::KDTree or another method.
-*/
-class CV_EXPORTS_W_SIMPLE KeyPoint
-{
-public:
-    //! the default constructor
-    CV_WRAP KeyPoint() : pt(0,0), size(0), angle(-1), response(0), octave(0), class_id(-1) {}
-    //! the full constructor
-    KeyPoint(Point2f _pt, float _size, float _angle=-1,
-            float _response=0, int _octave=0, int _class_id=-1)
-            : pt(_pt), size(_size), angle(_angle),
-            response(_response), octave(_octave), class_id(_class_id) {}
-    //! another form of the full constructor
-    CV_WRAP KeyPoint(float x, float y, float _size, float _angle=-1,
-            float _response=0, int _octave=0, int _class_id=-1)
-            : pt(x, y), size(_size), angle(_angle),
-            response(_response), octave(_octave), class_id(_class_id) {}
-
-    size_t hash() const;
-
-    //! converts vector of keypoints to vector of points
-    static void convert(const std::vector<KeyPoint>& keypoints,
-                        CV_OUT std::vector<Point2f>& points2f,
-                        const std::vector<int>& keypointIndexes=std::vector<int>());
-    //! converts vector of points to the vector of keypoints, where each keypoint is assigned the same size and the same orientation
-    static void convert(const std::vector<Point2f>& points2f,
-                        CV_OUT std::vector<KeyPoint>& keypoints,
-                        float size=1, float response=1, int octave=0, int class_id=-1);
-
-    //! computes overlap for pair of keypoints;
-    //! overlap is a ratio between area of keypoint regions intersection and
-    //! area of keypoint regions union (now keypoint region is circle)
-    static float overlap(const KeyPoint& kp1, const KeyPoint& kp2);
-
-    CV_PROP_RW Point2f pt; //!< coordinates of the keypoints
-    CV_PROP_RW float size; //!< diameter of the meaningful keypoint neighborhood
-    CV_PROP_RW float angle; //!< computed orientation of the keypoint (-1 if not applicable);
-                            //!< it's in [0,360) degrees and measured relative to
-                            //!< image coordinate system, ie in clockwise.
-    CV_PROP_RW float response; //!< the response by which the most strong keypoints have been selected. Can be used for the further sorting or subsampling
-    CV_PROP_RW int octave; //!< octave (pyramid layer) from which the keypoint has been extracted
-    CV_PROP_RW int class_id; //!< object class (if the keypoints need to be clustered by an object they belong to)
-};
-
-//! writes vector of keypoints to the file storage
-CV_EXPORTS void write(FileStorage& fs, const std::string& name, const std::vector<KeyPoint>& keypoints);
-//! reads vector of keypoints from the specified file storage node
-CV_EXPORTS void read(const FileNode& node, CV_OUT std::vector<KeyPoint>& keypoints);
+// //! writes vector of keypoints to the file storage
+// CV_EXPORTS void write(FileStorage& fs, const String& name, const std::vector<KeyPoint>& keypoints);
+// //! reads vector of keypoints from the specified file storage node
+// CV_EXPORTS void read(const FileNode& node, CV_OUT std::vector<KeyPoint>& keypoints);
 
 /*
  * A class filters a vector of keypoints.
@@ -179,7 +122,7 @@ public:
     CV_WRAP virtual bool empty() const;
 
     // Create feature detector by detector name.
-    CV_WRAP static Ptr<FeatureDetector> create( const std::string& detectorType );
+    CV_WRAP static Ptr<FeatureDetector> create( const String& detectorType );
 
 protected:
     virtual void detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask=Mat() ) const = 0;
@@ -229,7 +172,7 @@ public:
 
     CV_WRAP virtual bool empty() const;
 
-    CV_WRAP static Ptr<DescriptorExtractor> create( const std::string& descriptorExtractorType );
+    CV_WRAP static Ptr<DescriptorExtractor> create( const String& descriptorExtractorType );
 
 protected:
     virtual void computeImpl( const Mat& image, std::vector<KeyPoint>& keypoints, Mat& descriptors ) const = 0;
@@ -263,8 +206,10 @@ public:
                                      OutputArray descriptors,
                                      bool useProvidedKeypoints=false ) const = 0;
 
+    CV_WRAP void compute( const Mat& image, CV_OUT CV_IN_OUT std::vector<KeyPoint>& keypoints, CV_OUT Mat& descriptors ) const;
+
     // Create feature detector and descriptor extractor by name.
-    CV_WRAP static Ptr<Feature2D> create( const std::string& name );
+    CV_WRAP static Ptr<Feature2D> create( const String& name );
 };
 
 /*!
@@ -590,11 +535,11 @@ protected:
 };
 
 
-class CV_EXPORTS GFTTDetector : public FeatureDetector
+class CV_EXPORTS_W GFTTDetector : public FeatureDetector
 {
 public:
-    GFTTDetector( int maxCorners=1000, double qualityLevel=0.01, double minDistance=1,
-                  int blockSize=3, bool useHarrisDetector=false, double k=0.04 );
+    CV_WRAP GFTTDetector( int maxCorners=1000, double qualityLevel=0.01, double minDistance=1,
+                          int blockSize=3, bool useHarrisDetector=false, double k=0.04 );
     AlgorithmInfo* info() const;
 
 protected:
@@ -701,7 +646,7 @@ public:
      * gridRows            Grid rows count.
      * gridCols            Grid column count.
      */
-    CV_WRAP GridAdaptedFeatureDetector( const Ptr<FeatureDetector>& detector=0,
+    CV_WRAP GridAdaptedFeatureDetector( const Ptr<FeatureDetector>& detector=Ptr<FeatureDetector>(),
                                         int maxTotalKeypoints=1000,
                                         int gridRows=4, int gridCols=4 );
 
@@ -765,7 +710,7 @@ public:
 
     virtual Ptr<AdjusterAdapter> clone() const = 0;
 
-    static Ptr<AdjusterAdapter> create( const std::string& detectorType );
+    static Ptr<AdjusterAdapter> create( const String& detectorType );
 };
 /** \brief an adaptively adjusting detector that iteratively detects until the desired number
  * of features are detected.
@@ -1016,7 +961,7 @@ struct CV_EXPORTS Hamming
 
 typedef Hamming HammingLUT;
 
-template<int cellsize> struct CV_EXPORTS HammingMultilevel
+template<int cellsize> struct HammingMultilevel
 {
     enum { normType = NORM_HAMMING + (cellsize>1) };
     typedef unsigned char ValueType;
@@ -1028,33 +973,6 @@ template<int cellsize> struct CV_EXPORTS HammingMultilevel
     }
 };
 
-/****************************************************************************************\
-*                                      DMatch                                            *
-\****************************************************************************************/
-/*
- * Struct for matching: query descriptor index, train descriptor index, train image index and distance between descriptors.
- */
-struct CV_EXPORTS_W_SIMPLE DMatch
-{
-    CV_WRAP DMatch() : queryIdx(-1), trainIdx(-1), imgIdx(-1), distance(FLT_MAX) {}
-    CV_WRAP DMatch( int _queryIdx, int _trainIdx, float _distance ) :
-            queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(-1), distance(_distance) {}
-    CV_WRAP DMatch( int _queryIdx, int _trainIdx, int _imgIdx, float _distance ) :
-            queryIdx(_queryIdx), trainIdx(_trainIdx), imgIdx(_imgIdx), distance(_distance) {}
-
-    CV_PROP_RW int queryIdx; // query descriptor index
-    CV_PROP_RW int trainIdx; // train descriptor index
-    CV_PROP_RW int imgIdx;   // train image index
-
-    CV_PROP_RW float distance;
-
-    // less is better
-    bool operator<( const DMatch &m ) const
-    {
-        return distance < m.distance;
-    }
-};
-
 /****************************************************************************************\
 *                                  DescriptorMatcher                                     *
 \****************************************************************************************/
@@ -1141,7 +1059,7 @@ public:
     // but with empty train data.
     virtual Ptr<DescriptorMatcher> clone( bool emptyTrainData=false ) const = 0;
 
-    CV_WRAP static Ptr<DescriptorMatcher> create( const std::string& descriptorMatcherType );
+    CV_WRAP static Ptr<DescriptorMatcher> create( const String& descriptorMatcherType );
 protected:
     /*
      * Class to work with descriptors from several images as with one merged matrix.
@@ -1225,8 +1143,8 @@ protected:
 class CV_EXPORTS_W FlannBasedMatcher : public DescriptorMatcher
 {
 public:
-    CV_WRAP FlannBasedMatcher( const Ptr<flann::IndexParams>& indexParams=new flann::KDTreeIndexParams(),
-                       const Ptr<flann::SearchParams>& searchParams=new flann::SearchParams() );
+    CV_WRAP FlannBasedMatcher( const Ptr<flann::IndexParams>& indexParams=makePtr<flann::KDTreeIndexParams>(),
+                       const Ptr<flann::SearchParams>& searchParams=makePtr<flann::SearchParams>() );
 
     virtual void add( const std::vector<Mat>& descriptors );
     virtual void clear();
@@ -1367,8 +1285,8 @@ public:
     // but with empty train data.
     virtual Ptr<GenericDescriptorMatcher> clone( bool emptyTrainData=false ) const = 0;
 
-    static Ptr<GenericDescriptorMatcher> create( const std::string& genericDescritptorMatcherType,
-                                                 const std::string &paramsFilename=std::string() );
+    static Ptr<GenericDescriptorMatcher> create( const String& genericDescritptorMatcherType,
+                                                 const String &paramsFilename=String() );
 
 protected:
     // In fact the matching is implemented only by the following two methods. These methods suppose
@@ -1486,15 +1404,15 @@ CV_EXPORTS_W void drawKeypoints( const Mat& image, const std::vector<KeyPoint>&
                                const Scalar& color=Scalar::all(-1), int flags=DrawMatchesFlags::DEFAULT );
 
 // Draws matches of keypints from two images on output image.
-CV_EXPORTS void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
+CV_EXPORTS_W void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
                              const Mat& img2, const std::vector<KeyPoint>& keypoints2,
-                             const std::vector<DMatch>& matches1to2, Mat& outImg,
+                             const std::vector<DMatch>& matches1to2, CV_OUT Mat& outImg,
                              const Scalar& matchColor=Scalar::all(-1), const Scalar& singlePointColor=Scalar::all(-1),
                              const std::vector<char>& matchesMask=std::vector<char>(), int flags=DrawMatchesFlags::DEFAULT );
 
-CV_EXPORTS void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
+CV_EXPORTS_AS(drawMatchesKnn) void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
                              const Mat& img2, const std::vector<KeyPoint>& keypoints2,
-                             const std::vector<std::vector<DMatch> >& matches1to2, Mat& outImg,
+                             const std::vector<std::vector<DMatch> >& matches1to2, CV_OUT Mat& outImg,
                              const Scalar& matchColor=Scalar::all(-1), const Scalar& singlePointColor=Scalar::all(-1),
                              const std::vector<std::vector<char> >& matchesMask=std::vector<std::vector<char> >(), int flags=DrawMatchesFlags::DEFAULT );
 
@@ -1603,8 +1521,4 @@ protected:
 
 } /* namespace cv */
 
-#endif /* __cplusplus */
-
 #endif
-
-/* End of file. */
diff --git a/modules/features2d/include/opencv2/features2d/features2d.hpp b/modules/features2d/include/opencv2/features2d/features2d.hpp
index c825295ecd..e81df0ad08 100644
--- a/modules/features2d/include/opencv2/features2d/features2d.hpp
+++ b/modules/features2d/include/opencv2/features2d/features2d.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/features2d.hpp"
\ No newline at end of file
+#include "opencv2/features2d.hpp"
diff --git a/modules/features2d/perf/perf_fast.cpp b/modules/features2d/perf/perf_fast.cpp
index 465bef7688..25b82bd412 100644
--- a/modules/features2d/perf/perf_fast.cpp
+++ b/modules/features2d/perf/perf_fast.cpp
@@ -18,7 +18,7 @@ typedef perf::TestBaseWithParam<File_Type_t> fast;
 
 PERF_TEST_P(fast, detect, testing::Combine(
                             testing::Values(FAST_IMAGES),
-                            testing::ValuesIn(FastType::all())
+                            FastType::all()
                           ))
 {
     string filename = getDataPath(get<0>(GetParam()));
@@ -41,4 +41,3 @@ PERF_TEST_P(fast, detect, testing::Combine(
 
     SANITY_CHECK_KEYPOINTS(points);
 }
-
diff --git a/modules/features2d/perf/perf_precomp.cpp b/modules/features2d/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/features2d/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/features2d/src/bagofwords.cpp b/modules/features2d/src/bagofwords.cpp
index b27b85123e..a257f6cbe4 100644
--- a/modules/features2d/src/bagofwords.cpp
+++ b/modules/features2d/src/bagofwords.cpp
@@ -147,7 +147,7 @@ void BOWImgDescriptorExtractor::compute( const Mat& image, std::vector<KeyPoint>
     int clusterCount = descriptorSize(); // = vocabulary.rows
 
     // Compute descriptors for the image.
-    Mat descriptors = _descriptors ? *_descriptors : Mat();
+    Mat descriptors;
     dextractor->compute( image, keypoints, descriptors );
 
     // Match keypoint descriptors to cluster center (to vocabulary)
@@ -176,6 +176,11 @@ void BOWImgDescriptorExtractor::compute( const Mat& image, std::vector<KeyPoint>
 
     // Normalize image descriptor.
     imgDescriptor /= descriptors.rows;
+
+    // Add the descriptors of image keypoints
+    if (_descriptors) {
+        *_descriptors = descriptors.clone();
+    }
 }
 
 int BOWImgDescriptorExtractor::descriptorSize() const
diff --git a/modules/features2d/src/blobdetector.cpp b/modules/features2d/src/blobdetector.cpp
index 3f50f3c1af..92d50ee53c 100644
--- a/modules/features2d/src/blobdetector.cpp
+++ b/modules/features2d/src/blobdetector.cpp
@@ -42,6 +42,7 @@
 
 #include "precomp.hpp"
 #include <iterator>
+#include <limits>
 
 //#define DEBUG_BLOB_DETECTOR
 
@@ -169,7 +170,7 @@ void SimpleBlobDetector::findBlobs(const cv::Mat &image, const cv::Mat &binaryIm
 
     std::vector < std::vector<Point> > contours;
     Mat tmpBinaryImage = binaryImage.clone();
-    findContours(tmpBinaryImage, contours, CV_RETR_LIST, CV_CHAIN_APPROX_NONE);
+    findContours(tmpBinaryImage, contours, RETR_LIST, CHAIN_APPROX_NONE);
 
 #ifdef DEBUG_BLOB_DETECTOR
     //  Mat keypointsImage;
@@ -260,8 +261,11 @@ void SimpleBlobDetector::findBlobs(const cv::Mat &image, const cv::Mat &binaryIm
             center.radius = (dists[(dists.size() - 1) / 2] + dists[dists.size() / 2]) / 2.;
         }
 
+        if(moms.m00 == 0.0)
+            continue;
         centers.push_back(center);
 
+
 #ifdef DEBUG_BLOB_DETECTOR
         //    circle( keypointsImage, center.location, 1, Scalar(0,0,255), 1 );
 #endif
@@ -278,7 +282,7 @@ void SimpleBlobDetector::detectImpl(const cv::Mat& image, std::vector<cv::KeyPoi
     keypoints.clear();
     Mat grayscaleImage;
     if (image.channels() == 3)
-        cvtColor(image, grayscaleImage, CV_BGR2GRAY);
+        cvtColor(image, grayscaleImage, COLOR_BGR2GRAY);
     else
         grayscaleImage = image;
 
diff --git a/modules/features2d/src/brief.cpp b/modules/features2d/src/brief.cpp
index 95ed868155..a429451b46 100644
--- a/modules/features2d/src/brief.cpp
+++ b/modules/features2d/src/brief.cpp
@@ -111,7 +111,7 @@ BriefDescriptorExtractor::BriefDescriptorExtractor(int bytes) :
             test_fn_ = pixelTests64;
             break;
         default:
-            CV_Error(CV_StsBadArg, "bytes must be 16, 32, or 64");
+            CV_Error(Error::StsBadArg, "bytes must be 16, 32, or 64");
     }
 }
 
@@ -140,7 +140,7 @@ void BriefDescriptorExtractor::read( const FileNode& fn)
             test_fn_ = pixelTests64;
             break;
         default:
-            CV_Error(CV_StsBadArg, "descriptorSize must be 16, 32, or 64");
+            CV_Error(Error::StsBadArg, "descriptorSize must be 16, 32, or 64");
     }
     bytes_ = dSize;
 }
@@ -156,7 +156,7 @@ void BriefDescriptorExtractor::computeImpl(const Mat& image, std::vector<KeyPoin
     Mat sum;
 
     Mat grayImage = image;
-    if( image.type() != CV_8U ) cvtColor( image, grayImage, CV_BGR2GRAY );
+    if( image.type() != CV_8U ) cvtColor( image, grayImage, COLOR_BGR2GRAY );
 
     ///TODO allow the user to pass in a precomputed integral image
     //if(image.type() == CV_32S)
diff --git a/modules/features2d/src/brisk.cpp b/modules/features2d/src/brisk.cpp
index 6709449fa5..76bded666b 100644
--- a/modules/features2d/src/brisk.cpp
+++ b/modules/features2d/src/brisk.cpp
@@ -236,7 +236,7 @@ BRISK::generateKernel(std::vector<float> &radiusList, std::vector<int> &numberLi
 
   // get the total number of points
   const int rings = (int)radiusList.size();
-  assert(radiusList.size()!=0&&radiusList.size()==numberList.size());
+  CV_Assert(radiusList.size() != 0 && radiusList.size() == numberList.size());
   points_ = 0; // remember the total number of points
   for (int ring = 0; ring < rings; ring++)
   {
@@ -336,7 +336,7 @@ BRISK::generateKernel(std::vector<float> &radiusList, std::vector<int> &numberLi
       else if (norm_sq < dMax_sq)
       {
         // save to short pairs
-        assert(noShortPairs_<indSize);
+        CV_Assert(noShortPairs_ < indSize);
         // make sure the user passes something sensible
         BriskShortPair& shortPair = shortPairs_[indexChange[noShortPairs_]];
         shortPair.j = j;
@@ -525,7 +525,11 @@ BRISK::operator()( InputArray _image, InputArray _mask, std::vector<KeyPoint>& k
   bool doOrientation=true;
   if (useProvidedKeypoints)
     doOrientation = false;
-  computeDescriptorsAndOrOrientation(_image, _mask, keypoints, _descriptors, true, doOrientation,
+
+  // If the user specified cv::noArray(), this will yield false. Otherwise it will return true.
+  bool doDescriptors = _descriptors.needed();
+
+  computeDescriptorsAndOrOrientation(_image, _mask, keypoints, _descriptors, doDescriptors, doOrientation,
                                        useProvidedKeypoints);
 }
 
@@ -536,7 +540,7 @@ BRISK::computeDescriptorsAndOrOrientation(InputArray _image, InputArray _mask, s
 {
   Mat image = _image.getMat(), mask = _mask.getMat();
   if( image.type() != CV_8UC1 )
-      cvtColor(image, image, CV_BGR2GRAY);
+      cvtColor(image, image, COLOR_BGR2GRAY);
 
   if (!useProvidedKeypoints)
   {
@@ -729,7 +733,7 @@ BRISK::computeKeypointsNoOrientation(InputArray _image, InputArray _mask, std::v
 {
   Mat image = _image.getMat(), mask = _mask.getMat();
   if( image.type() != CV_8UC1 )
-      cvtColor(_image, image, CV_BGR2GRAY);
+      cvtColor(_image, image, COLOR_BGR2GRAY);
 
   BriskScaleSpace briskScaleSpace(octaves);
   briskScaleSpace.constructPyramid(image);
@@ -912,7 +916,7 @@ BriskScaleSpace::getKeypoints(const int threshold_, std::vector<cv::KeyPoint>& k
 inline int
 BriskScaleSpace::getScoreAbove(const int layer, const int x_layer, const int y_layer) const
 {
-  assert(layer<layers_-1);
+  CV_Assert(layer < layers_-1);
   const BriskLayer& l = pyramid_[layer + 1];
   if (layer % 2 == 0)
   { // octave
@@ -955,7 +959,7 @@ BriskScaleSpace::getScoreAbove(const int layer, const int x_layer, const int y_l
 inline int
 BriskScaleSpace::getScoreBelow(const int layer, const int x_layer, const int y_layer) const
 {
-  assert(layer);
+  CV_Assert(layer);
   const BriskLayer& l = pyramid_[layer - 1];
   int sixth_x;
   int quarter_x;
@@ -1343,7 +1347,7 @@ BriskScaleSpace::getScoreMaxAbove(const int layer, const int x_layer, const int
   float y1;
 
   // the layer above
-  assert(layer+1<layers_);
+  CV_Assert(layer + 1 < layers_);
   const BriskLayer& layerAbove = pyramid_[layer + 1];
 
   if (layer % 2 == 0)
@@ -1539,7 +1543,7 @@ BriskScaleSpace::getScoreMaxBelow(const int layer, const int x_layer, const int
   }
 
   // the layer below
-  assert(layer>0);
+  CV_Assert(layer > 0);
   const BriskLayer& layerBelow = pyramid_[layer - 1];
 
   // check the first row
@@ -1999,7 +2003,7 @@ BriskLayer::BriskLayer(const cv::Mat& img_in, float scale_in, float offset_in)
   scale_ = scale_in;
   offset_ = offset_in;
   // create an agast detector
-  fast_9_16_ = new FastFeatureDetector(1, true, FastFeatureDetector::TYPE_9_16);
+  fast_9_16_ = makePtr<FastFeatureDetector>(1, true, FastFeatureDetector::TYPE_9_16);
   makeOffsets(pixel_5_8_, (int)img_.step, 8);
   makeOffsets(pixel_9_16_, (int)img_.step, 16);
 }
@@ -2021,7 +2025,7 @@ BriskLayer::BriskLayer(const BriskLayer& layer, int mode)
     offset_ = 0.5f * scale_ - 0.5f;
   }
   scores_ = cv::Mat::zeros(img_.rows, img_.cols, CV_8U);
-  fast_9_16_ = new FastFeatureDetector(1, false, FastFeatureDetector::TYPE_9_16);
+  fast_9_16_ = makePtr<FastFeatureDetector>(1, false, FastFeatureDetector::TYPE_9_16);
   makeOffsets(pixel_5_8_, (int)img_.step, 8);
   makeOffsets(pixel_9_16_, (int)img_.step, 16);
 }
@@ -2109,7 +2113,7 @@ BriskLayer::getAgastScore(float xf, float yf, int threshold_in, float scale_in)
 inline int
 BriskLayer::value(const cv::Mat& mat, float xf, float yf, float scale_in) const
 {
-  assert(!mat.empty());
+  CV_Assert(!mat.empty());
   // get the position
   const int x = cvFloor(xf);
   const int y = cvFloor(yf);
@@ -2216,8 +2220,8 @@ inline void
 BriskLayer::halfsample(const cv::Mat& srcimg, cv::Mat& dstimg)
 {
   // make sure the destination image is of the right size:
-  assert(srcimg.cols/2==dstimg.cols);
-  assert(srcimg.rows/2==dstimg.rows);
+  CV_Assert(srcimg.cols / 2 == dstimg.cols);
+  CV_Assert(srcimg.rows / 2 == dstimg.rows);
 
   // handle non-SSE case
   resize(srcimg, dstimg, dstimg.size(), 0, 0, INTER_AREA);
@@ -2227,8 +2231,8 @@ inline void
 BriskLayer::twothirdsample(const cv::Mat& srcimg, cv::Mat& dstimg)
 {
   // make sure the destination image is of the right size:
-  assert((srcimg.cols/3)*2==dstimg.cols);
-  assert((srcimg.rows/3)*2==dstimg.rows);
+  CV_Assert((srcimg.cols / 3) * 2 == dstimg.cols);
+  CV_Assert((srcimg.rows / 3) * 2 == dstimg.rows);
 
   resize(srcimg, dstimg, dstimg.size(), 0, 0, INTER_AREA);
 }
diff --git a/modules/features2d/src/descriptors.cpp b/modules/features2d/src/descriptors.cpp
index 7f87bd5eba..b79768ae15 100644
--- a/modules/features2d/src/descriptors.cpp
+++ b/modules/features2d/src/descriptors.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include <limits>
 
 namespace cv
 {
@@ -92,18 +93,24 @@ void DescriptorExtractor::removeBorderKeypoints( std::vector<KeyPoint>& keypoint
     KeyPointsFilter::runByImageBorder( keypoints, imageSize, borderSize );
 }
 
-Ptr<DescriptorExtractor> DescriptorExtractor::create(const std::string& descriptorExtractorType)
+Ptr<DescriptorExtractor> DescriptorExtractor::create(const String& descriptorExtractorType)
 {
     if( descriptorExtractorType.find("Opponent") == 0 )
     {
-        size_t pos = std::string("Opponent").size();
-        std::string type = descriptorExtractorType.substr(pos);
-        return new OpponentColorDescriptorExtractor(DescriptorExtractor::create(type));
+        size_t pos = String("Opponent").size();
+        String type = descriptorExtractorType.substr(pos);
+        return makePtr<OpponentColorDescriptorExtractor>(DescriptorExtractor::create(type));
     }
 
     return Algorithm::create<DescriptorExtractor>("Feature2D." + descriptorExtractorType);
 }
 
+
+CV_WRAP void Feature2D::compute( const Mat& image, CV_OUT CV_IN_OUT std::vector<KeyPoint>& keypoints, CV_OUT Mat& descriptors ) const
+{
+   DescriptorExtractor::compute(image, keypoints, descriptors);
+}
+
 /////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
 /****************************************************************************************\
@@ -112,13 +119,13 @@ Ptr<DescriptorExtractor> DescriptorExtractor::create(const std::string& descript
 OpponentColorDescriptorExtractor::OpponentColorDescriptorExtractor( const Ptr<DescriptorExtractor>& _descriptorExtractor ) :
         descriptorExtractor(_descriptorExtractor)
 {
-    CV_Assert( !descriptorExtractor.empty() );
+    CV_Assert( descriptorExtractor );
 }
 
 static void convertBGRImageToOpponentColorSpace( const Mat& bgrImage, std::vector<Mat>& opponentChannels )
 {
     if( bgrImage.type() != CV_8UC3 )
-        CV_Error( CV_StsBadArg, "input image must be an BGR image of type CV_8UC3" );
+        CV_Error( Error::StsBadArg, "input image must be an BGR image of type CV_8UC3" );
 
     // Prepare opponent color space storage matrices.
     opponentChannels.resize( 3 );
@@ -242,7 +249,7 @@ int OpponentColorDescriptorExtractor::descriptorType() const
 
 bool OpponentColorDescriptorExtractor::empty() const
 {
-    return descriptorExtractor.empty() || (DescriptorExtractor*)(descriptorExtractor)->empty();
+    return !descriptorExtractor || descriptorExtractor->empty();
 }
 
 }
diff --git a/modules/features2d/src/detectors.cpp b/modules/features2d/src/detectors.cpp
index 19175f1b68..63a882dd3f 100644
--- a/modules/features2d/src/detectors.cpp
+++ b/modules/features2d/src/detectors.cpp
@@ -86,23 +86,23 @@ void FeatureDetector::removeInvalidPoints( const Mat& mask, std::vector<KeyPoint
     KeyPointsFilter::runByPixelsMask( keypoints, mask );
 }
 
-Ptr<FeatureDetector> FeatureDetector::create( const std::string& detectorType )
+Ptr<FeatureDetector> FeatureDetector::create( const String& detectorType )
 {
     if( detectorType.find("Grid") == 0 )
     {
-        return new GridAdaptedFeatureDetector(FeatureDetector::create(
+        return makePtr<GridAdaptedFeatureDetector>(FeatureDetector::create(
                                 detectorType.substr(strlen("Grid"))));
     }
 
     if( detectorType.find("Pyramid") == 0 )
     {
-        return new PyramidAdaptedFeatureDetector(FeatureDetector::create(
+        return makePtr<PyramidAdaptedFeatureDetector>(FeatureDetector::create(
                                 detectorType.substr(strlen("Pyramid"))));
     }
 
     if( detectorType.find("Dynamic") == 0 )
     {
-        return new DynamicAdaptedFeatureDetector(AdjusterAdapter::create(
+        return makePtr<DynamicAdaptedFeatureDetector>(AdjusterAdapter::create(
                                 detectorType.substr(strlen("Dynamic"))));
     }
 
@@ -128,7 +128,7 @@ GFTTDetector::GFTTDetector( int _nfeatures, double _qualityLevel,
 void GFTTDetector::detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask) const
 {
     Mat grayImage = image;
-    if( image.type() != CV_8U ) cvtColor( image, grayImage, CV_BGR2GRAY );
+    if( image.type() != CV_8U ) cvtColor( image, grayImage, COLOR_BGR2GRAY );
 
     std::vector<Point2f> corners;
     goodFeaturesToTrack( grayImage, corners, nfeatures, qualityLevel, minDistance, mask,
@@ -190,7 +190,7 @@ GridAdaptedFeatureDetector::GridAdaptedFeatureDetector( const Ptr<FeatureDetecto
 
 bool GridAdaptedFeatureDetector::empty() const
 {
-    return detector.empty() || (FeatureDetector*)detector->empty();
+    return !detector || detector->empty();
 }
 
 struct ResponseComparator
@@ -212,7 +212,7 @@ static void keepStrongest( int N, std::vector<KeyPoint>& keypoints )
 }
 
 namespace {
-class GridAdaptedFeatureDetectorInvoker
+class GridAdaptedFeatureDetectorInvoker : public ParallelLoopBody
 {
 private:
     int gridRows_, gridCols_;
@@ -221,29 +221,24 @@ private:
     const Mat& image_;
     const Mat& mask_;
     const Ptr<FeatureDetector>& detector_;
-#ifdef HAVE_TBB
-    tbb::mutex* kptLock_;
-#endif
+    Mutex* kptLock_;
 
     GridAdaptedFeatureDetectorInvoker& operator=(const GridAdaptedFeatureDetectorInvoker&); // to quiet MSVC
 
 public:
 
-    GridAdaptedFeatureDetectorInvoker(const Ptr<FeatureDetector>& detector, const Mat& image, const Mat& mask, std::vector<KeyPoint>& keypoints, int maxPerCell, int gridRows, int gridCols
-#ifdef HAVE_TBB
-        , tbb::mutex* kptLock
-#endif
-        ) : gridRows_(gridRows), gridCols_(gridCols), maxPerCell_(maxPerCell),
-            keypoints_(keypoints), image_(image), mask_(mask), detector_(detector)
-#ifdef HAVE_TBB
-            , kptLock_(kptLock)
-#endif
+    GridAdaptedFeatureDetectorInvoker(const Ptr<FeatureDetector>& detector, const Mat& image, const Mat& mask,
+                                      std::vector<KeyPoint>& keypoints, int maxPerCell, int gridRows, int gridCols,
+                                      cv::Mutex* kptLock)
+        : gridRows_(gridRows), gridCols_(gridCols), maxPerCell_(maxPerCell),
+          keypoints_(keypoints), image_(image), mask_(mask), detector_(detector),
+          kptLock_(kptLock)
     {
     }
 
-    void operator() (const BlockedRange& range) const
+    void operator() (const Range& range) const
     {
-        for (int i = range.begin(); i < range.end(); ++i)
+        for (int i = range.start; i < range.end; ++i)
         {
             int celly = i / gridCols_;
             int cellx = i - celly * gridCols_;
@@ -268,9 +263,8 @@ public:
                 it->pt.x += col_range.start;
                 it->pt.y += row_range.start;
             }
-#ifdef HAVE_TBB
-            tbb::mutex::scoped_lock join_keypoints(*kptLock_);
-#endif
+
+            cv::AutoLock join_keypoints(*kptLock_);
             keypoints_.insert( keypoints_.end(), sub_keypoints.begin(), sub_keypoints.end() );
         }
     }
@@ -287,13 +281,9 @@ void GridAdaptedFeatureDetector::detectImpl( const Mat& image, std::vector<KeyPo
     keypoints.reserve(maxTotalKeypoints);
     int maxPerCell = maxTotalKeypoints / (gridRows * gridCols);
 
-#ifdef HAVE_TBB
-    tbb::mutex kptLock;
-    cv::parallel_for(cv::BlockedRange(0, gridRows * gridCols),
+    cv::Mutex kptLock;
+    cv::parallel_for_(cv::Range(0, gridRows * gridCols),
         GridAdaptedFeatureDetectorInvoker(detector, image, mask, keypoints, maxPerCell, gridRows, gridCols, &kptLock));
-#else
-    GridAdaptedFeatureDetectorInvoker(detector, image, mask, keypoints, maxPerCell, gridRows, gridCols)(cv::BlockedRange(0, gridRows * gridCols));
-#endif
 }
 
 /*
@@ -305,7 +295,7 @@ PyramidAdaptedFeatureDetector::PyramidAdaptedFeatureDetector( const Ptr<FeatureD
 
 bool PyramidAdaptedFeatureDetector::empty() const
 {
-    return detector.empty() || (FeatureDetector*)detector->empty();
+    return !detector || detector->empty();
 }
 
 void PyramidAdaptedFeatureDetector::detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask ) const
@@ -346,7 +336,7 @@ void PyramidAdaptedFeatureDetector::detectImpl( const Mat& image, std::vector<Ke
             src = dst;
 
             if( !mask.empty() )
-                resize( dilated_mask, src_mask, src.size(), 0, 0, CV_INTER_AREA );
+                resize( dilated_mask, src_mask, src.size(), 0, 0, INTER_AREA );
         }
     }
 
diff --git a/modules/features2d/src/draw.cpp b/modules/features2d/src/draw.cpp
index 16466577f8..61c2f817a1 100644
--- a/modules/features2d/src/draw.cpp
+++ b/modules/features2d/src/draw.cpp
@@ -60,7 +60,7 @@ static inline void _drawKeypoint( Mat& img, const KeyPoint& p, const Scalar& col
         int radius = cvRound(p.size/2 * draw_multiplier); // KeyPoint::size is a diameter
 
         // draw the circles around keypoints with the keypoints size
-        circle( img, center, radius, color, 1, CV_AA, draw_shift_bits );
+        circle( img, center, radius, color, 1, LINE_AA, draw_shift_bits );
 
         // draw orientation of the keypoint, if it is applicable
         if( p.angle != -1 )
@@ -69,14 +69,14 @@ static inline void _drawKeypoint( Mat& img, const KeyPoint& p, const Scalar& col
             Point orient( cvRound(cos(srcAngleRad)*radius ),
                           cvRound(sin(srcAngleRad)*radius )
                         );
-            line( img, center, center+orient, color, 1, CV_AA, draw_shift_bits );
+            line( img, center, center+orient, color, 1, LINE_AA, draw_shift_bits );
         }
 #if 0
         else
         {
             // draw center with R=1
             int radius = 1 * draw_multiplier;
-            circle( img, center, radius, color, 1, CV_AA, draw_shift_bits );
+            circle( img, center, radius, color, 1, LINE_AA, draw_shift_bits );
         }
 #endif
     }
@@ -84,7 +84,7 @@ static inline void _drawKeypoint( Mat& img, const KeyPoint& p, const Scalar& col
     {
         // draw center with R=3
         int radius = 3 * draw_multiplier;
-        circle( img, center, radius, color, 1, CV_AA, draw_shift_bits );
+        circle( img, center, radius, color, 1, LINE_AA, draw_shift_bits );
     }
 }
 
@@ -99,11 +99,11 @@ void drawKeypoints( const Mat& image, const std::vector<KeyPoint>& keypoints, Ma
         }
         else if( image.type() == CV_8UC1 )
         {
-            cvtColor( image, outImage, CV_GRAY2BGR );
+            cvtColor( image, outImage, COLOR_GRAY2BGR );
         }
         else
         {
-            CV_Error( CV_StsBadArg, "Incorrect type of input image.\n" );
+            CV_Error( Error::StsBadArg, "Incorrect type of input image.\n" );
         }
     }
 
@@ -129,7 +129,7 @@ static void _prepareImgAndDrawKeypoints( const Mat& img1, const std::vector<KeyP
     if( flags & DrawMatchesFlags::DRAW_OVER_OUTIMG )
     {
         if( size.width > outImg.cols || size.height > outImg.rows )
-            CV_Error( CV_StsBadSize, "outImg has size less than need to draw img1 and img2 together" );
+            CV_Error( Error::StsBadSize, "outImg has size less than need to draw img1 and img2 together" );
         outImg1 = outImg( Rect(0, 0, img1.cols, img1.rows) );
         outImg2 = outImg( Rect(img1.cols, 0, img2.cols, img2.rows) );
     }
@@ -141,12 +141,12 @@ static void _prepareImgAndDrawKeypoints( const Mat& img1, const std::vector<KeyP
         outImg2 = outImg( Rect(img1.cols, 0, img2.cols, img2.rows) );
 
         if( img1.type() == CV_8U )
-            cvtColor( img1, outImg1, CV_GRAY2BGR );
+            cvtColor( img1, outImg1, COLOR_GRAY2BGR );
         else
             img1.copyTo( outImg1 );
 
         if( img2.type() == CV_8U )
-            cvtColor( img2, outImg2, CV_GRAY2BGR );
+            cvtColor( img2, outImg2, COLOR_GRAY2BGR );
         else
             img2.copyTo( outImg2 );
     }
@@ -179,7 +179,7 @@ static inline void _drawMatch( Mat& outImg, Mat& outImg1, Mat& outImg2 ,
     line( outImg,
           Point(cvRound(pt1.x*draw_multiplier), cvRound(pt1.y*draw_multiplier)),
           Point(cvRound(dpt2.x*draw_multiplier), cvRound(dpt2.y*draw_multiplier)),
-          color, 1, CV_AA, draw_shift_bits );
+          color, 1, LINE_AA, draw_shift_bits );
 }
 
 void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
@@ -189,7 +189,7 @@ void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
                   const std::vector<char>& matchesMask, int flags )
 {
     if( !matchesMask.empty() && matchesMask.size() != matches1to2.size() )
-        CV_Error( CV_StsBadSize, "matchesMask must have the same size as matches1to2" );
+        CV_Error( Error::StsBadSize, "matchesMask must have the same size as matches1to2" );
 
     Mat outImg1, outImg2;
     _prepareImgAndDrawKeypoints( img1, keypoints1, img2, keypoints2,
@@ -218,7 +218,7 @@ void drawMatches( const Mat& img1, const std::vector<KeyPoint>& keypoints1,
                   const std::vector<std::vector<char> >& matchesMask, int flags )
 {
     if( !matchesMask.empty() && matchesMask.size() != matches1to2.size() )
-        CV_Error( CV_StsBadSize, "matchesMask must have the same size as matches1to2" );
+        CV_Error( Error::StsBadSize, "matchesMask must have the same size as matches1to2" );
 
     Mat outImg1, outImg2;
     _prepareImgAndDrawKeypoints( img1, keypoints1, img2, keypoints2,
diff --git a/modules/features2d/src/dynamic.cpp b/modules/features2d/src/dynamic.cpp
index 523a3e1f82..6bd6ab4de8 100644
--- a/modules/features2d/src/dynamic.cpp
+++ b/modules/features2d/src/dynamic.cpp
@@ -51,7 +51,7 @@ DynamicAdaptedFeatureDetector::DynamicAdaptedFeatureDetector(const Ptr<AdjusterA
 
 bool DynamicAdaptedFeatureDetector::empty() const
 {
-    return adjuster_.empty() || adjuster_->empty();
+    return !adjuster_ || adjuster_->empty();
 }
 
 void DynamicAdaptedFeatureDetector::detectImpl(const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask) const
@@ -124,7 +124,7 @@ bool FastAdjuster::good() const
 
 Ptr<AdjusterAdapter> FastAdjuster::clone() const
 {
-    Ptr<AdjusterAdapter> cloned_obj = new FastAdjuster( init_thresh_, nonmax_, min_thresh_, max_thresh_ );
+    Ptr<AdjusterAdapter> cloned_obj(new FastAdjuster( init_thresh_, nonmax_, min_thresh_, max_thresh_ ));
     return cloned_obj;
 }
 
@@ -158,7 +158,7 @@ bool StarAdjuster::good() const
 
 Ptr<AdjusterAdapter> StarAdjuster::clone() const
 {
-    Ptr<AdjusterAdapter> cloned_obj = new StarAdjuster( init_thresh_, min_thresh_, max_thresh_ );
+    Ptr<AdjusterAdapter> cloned_obj(new StarAdjuster( init_thresh_, min_thresh_, max_thresh_ ));
     return cloned_obj;
 }
 
@@ -195,25 +195,25 @@ bool SurfAdjuster::good() const
 
 Ptr<AdjusterAdapter> SurfAdjuster::clone() const
 {
-    Ptr<AdjusterAdapter> cloned_obj = new SurfAdjuster( init_thresh_, min_thresh_, max_thresh_ );
+    Ptr<AdjusterAdapter> cloned_obj(new SurfAdjuster( init_thresh_, min_thresh_, max_thresh_ ));
     return cloned_obj;
 }
 
-Ptr<AdjusterAdapter> AdjusterAdapter::create( const std::string& detectorType )
+Ptr<AdjusterAdapter> AdjusterAdapter::create( const String& detectorType )
 {
     Ptr<AdjusterAdapter> adapter;
 
     if( !detectorType.compare( "FAST" ) )
     {
-        adapter = new FastAdjuster();
+        adapter = makePtr<FastAdjuster>();
     }
     else if( !detectorType.compare( "STAR" ) )
     {
-        adapter = new StarAdjuster();
+        adapter = makePtr<StarAdjuster>();
     }
     else if( !detectorType.compare( "SURF" ) )
     {
-        adapter = new SurfAdjuster();
+        adapter = makePtr<SurfAdjuster>();
     }
 
     return adapter;
diff --git a/modules/features2d/src/evaluation.cpp b/modules/features2d/src/evaluation.cpp
index 591e1c20bc..5bde951d60 100644
--- a/modules/features2d/src/evaluation.cpp
+++ b/modules/features2d/src/evaluation.cpp
@@ -185,7 +185,7 @@ void EllipticKeyPoint::convert( const std::vector<KeyPoint>& src, std::vector<El
         for( size_t i = 0; i < src.size(); i++ )
         {
             float rad = src[i].size/2;
-            assert( rad );
+            CV_Assert( rad );
             float fac = 1.f/(rad*rad);
             dst[i] = EllipticKeyPoint( src[i].pt, Scalar(fac, 0, fac) );
         }
@@ -210,7 +210,7 @@ void EllipticKeyPoint::calcProjection( const std::vector<EllipticKeyPoint>& src,
 {
     if( !src.empty() )
     {
-        assert( !H.empty() && H.cols == 3 && H.rows == 3);
+        CV_Assert( !H.empty() && H.cols == 3 && H.rows == 3);
         dst.resize(src.size());
         std::vector<EllipticKeyPoint>::const_iterator srcIt = src.begin();
         std::vector<EllipticKeyPoint>::iterator       dstIt = dst.begin();
@@ -257,7 +257,7 @@ struct IntersectAreaCounter
     {
         CV_Assert( miny < maxy );
         CV_Assert( dr > FLT_EPSILON );
-        
+
         int temp_bua = bua, temp_bna = bna;
         for( int i = range.begin(); i != range.end(); i++ )
         {
@@ -461,8 +461,8 @@ void cv::evaluateFeatureDetector( const Mat& img1, const Mat& img2, const Mat& H
     keypoints1 = _keypoints1 != 0 ? _keypoints1 : &buf1;
     keypoints2 = _keypoints2 != 0 ? _keypoints2 : &buf2;
 
-    if( (keypoints1->empty() || keypoints2->empty()) && fdetector.empty() )
-        CV_Error( CV_StsBadArg, "fdetector must not be empty when keypoints1 or keypoints2 is empty" );
+    if( (keypoints1->empty() || keypoints2->empty()) && !fdetector )
+        CV_Error( Error::StsBadArg, "fdetector must not be empty when keypoints1 or keypoints2 is empty" );
 
     if( keypoints1->empty() )
         fdetector->detect( img1, *keypoints1 );
@@ -573,15 +573,15 @@ void cv::evaluateGenericDescriptorMatcher( const Mat& img1, const Mat& img2, con
     correctMatches1to2Mask = _correctMatches1to2Mask != 0 ? _correctMatches1to2Mask : &buf2;
 
     if( keypoints1.empty() )
-        CV_Error( CV_StsBadArg, "keypoints1 must not be empty" );
+        CV_Error( Error::StsBadArg, "keypoints1 must not be empty" );
 
-    if( matches1to2->empty() && dmatcher.empty() )
-        CV_Error( CV_StsBadArg, "dmatch must not be empty when matches1to2 is empty" );
+    if( matches1to2->empty() && !dmatcher )
+        CV_Error( Error::StsBadArg, "dmatch must not be empty when matches1to2 is empty" );
 
     bool computeKeypoints2ByPrj = keypoints2.empty();
     if( computeKeypoints2ByPrj )
     {
-        assert(0);
+        CV_Error(Error::StsNotImplemented, "");
         // TODO: add computing keypoints2 from keypoints1 using H1to2
     }
 
diff --git a/modules/features2d/src/fast.cpp b/modules/features2d/src/fast.cpp
index 8306e3eb24..9e2181c140 100644
--- a/modules/features2d/src/fast.cpp
+++ b/modules/features2d/src/fast.cpp
@@ -100,61 +100,61 @@ void FAST_t(InputArray _img, std::vector<KeyPoint>& keypoints, int threshold, bo
     #if CV_SSE2
             if( patternSize == 16 )
             {
-            for(; j < img.cols - 16 - 3; j += 16, ptr += 16)
-            {
-                __m128i m0, m1;
-                __m128i v0 = _mm_loadu_si128((const __m128i*)ptr);
-                __m128i v1 = _mm_xor_si128(_mm_subs_epu8(v0, t), delta);
-                v0 = _mm_xor_si128(_mm_adds_epu8(v0, t), delta);
-
-                __m128i x0 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[0])), delta);
-                __m128i x1 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[quarterPatternSize])), delta);
-                __m128i x2 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[2*quarterPatternSize])), delta);
-                __m128i x3 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[3*quarterPatternSize])), delta);
-                m0 = _mm_and_si128(_mm_cmpgt_epi8(x0, v0), _mm_cmpgt_epi8(x1, v0));
-                m1 = _mm_and_si128(_mm_cmpgt_epi8(v1, x0), _mm_cmpgt_epi8(v1, x1));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x1, v0), _mm_cmpgt_epi8(x2, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x1), _mm_cmpgt_epi8(v1, x2)));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x2, v0), _mm_cmpgt_epi8(x3, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x2), _mm_cmpgt_epi8(v1, x3)));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x3, v0), _mm_cmpgt_epi8(x0, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x3), _mm_cmpgt_epi8(v1, x0)));
-                m0 = _mm_or_si128(m0, m1);
-                int mask = _mm_movemask_epi8(m0);
-                if( mask == 0 )
-                    continue;
-                if( (mask & 255) == 0 )
+                for(; j < img.cols - 16 - 3; j += 16, ptr += 16)
                 {
-                    j -= 8;
-                    ptr -= 8;
-                    continue;
-                }
+                    __m128i m0, m1;
+                    __m128i v0 = _mm_loadu_si128((const __m128i*)ptr);
+                    __m128i v1 = _mm_xor_si128(_mm_subs_epu8(v0, t), delta);
+                    v0 = _mm_xor_si128(_mm_adds_epu8(v0, t), delta);
 
-                __m128i c0 = _mm_setzero_si128(), c1 = c0, max0 = c0, max1 = c0;
-                for( k = 0; k < N; k++ )
-                {
-                    __m128i x = _mm_xor_si128(_mm_loadu_si128((const __m128i*)(ptr + pixel[k])), delta);
-                    m0 = _mm_cmpgt_epi8(x, v0);
-                    m1 = _mm_cmpgt_epi8(v1, x);
-
-                    c0 = _mm_and_si128(_mm_sub_epi8(c0, m0), m0);
-                    c1 = _mm_and_si128(_mm_sub_epi8(c1, m1), m1);
-
-                    max0 = _mm_max_epu8(max0, c0);
-                    max1 = _mm_max_epu8(max1, c1);
-                }
-
-                max0 = _mm_max_epu8(max0, max1);
-                int m = _mm_movemask_epi8(_mm_cmpgt_epi8(max0, K16));
-
-                for( k = 0; m > 0 && k < 16; k++, m >>= 1 )
-                    if(m & 1)
+                    __m128i x0 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[0])), delta);
+                    __m128i x1 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[quarterPatternSize])), delta);
+                    __m128i x2 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[2*quarterPatternSize])), delta);
+                    __m128i x3 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[3*quarterPatternSize])), delta);
+                    m0 = _mm_and_si128(_mm_cmpgt_epi8(x0, v0), _mm_cmpgt_epi8(x1, v0));
+                    m1 = _mm_and_si128(_mm_cmpgt_epi8(v1, x0), _mm_cmpgt_epi8(v1, x1));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x1, v0), _mm_cmpgt_epi8(x2, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x1), _mm_cmpgt_epi8(v1, x2)));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x2, v0), _mm_cmpgt_epi8(x3, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x2), _mm_cmpgt_epi8(v1, x3)));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x3, v0), _mm_cmpgt_epi8(x0, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x3), _mm_cmpgt_epi8(v1, x0)));
+                    m0 = _mm_or_si128(m0, m1);
+                    int mask = _mm_movemask_epi8(m0);
+                    if( mask == 0 )
+                        continue;
+                    if( (mask & 255) == 0 )
                     {
-                        cornerpos[ncorners++] = j+k;
-                        if(nonmax_suppression)
-                            curr[j+k] = (uchar)cornerScore<patternSize>(ptr+k, pixel, threshold);
+                        j -= 8;
+                        ptr -= 8;
+                        continue;
                     }
-            }
+
+                    __m128i c0 = _mm_setzero_si128(), c1 = c0, max0 = c0, max1 = c0;
+                    for( k = 0; k < N; k++ )
+                    {
+                        __m128i x = _mm_xor_si128(_mm_loadu_si128((const __m128i*)(ptr + pixel[k])), delta);
+                        m0 = _mm_cmpgt_epi8(x, v0);
+                        m1 = _mm_cmpgt_epi8(v1, x);
+
+                        c0 = _mm_and_si128(_mm_sub_epi8(c0, m0), m0);
+                        c1 = _mm_and_si128(_mm_sub_epi8(c1, m1), m1);
+
+                        max0 = _mm_max_epu8(max0, c0);
+                        max1 = _mm_max_epu8(max1, c1);
+                    }
+
+                    max0 = _mm_max_epu8(max0, max1);
+                    int m = _mm_movemask_epi8(_mm_cmpgt_epi8(max0, K16));
+
+                    for( k = 0; m > 0 && k < 16; k++, m >>= 1 )
+                        if(m & 1)
+                        {
+                            cornerpos[ncorners++] = j+k;
+                            if(nonmax_suppression)
+                                curr[j+k] = (uchar)cornerScore<patternSize>(ptr+k, pixel, threshold);
+                        }
+                }
             }
     #endif
             for( ; j < img.cols - 3; j++, ptr++ )
@@ -286,7 +286,7 @@ FastFeatureDetector::FastFeatureDetector( int _threshold, bool _nonmaxSuppressio
 void FastFeatureDetector::detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask ) const
 {
     Mat grayImage = image;
-    if( image.type() != CV_8U ) cvtColor( image, grayImage, CV_BGR2GRAY );
+    if( image.type() != CV_8U ) cvtColor( image, grayImage, COLOR_BGR2GRAY );
     FAST( grayImage, keypoints, threshold, nonmaxSuppression, type );
     KeyPointsFilter::runByPixelsMask( keypoints, mask );
 }
diff --git a/modules/features2d/src/fast_score.cpp b/modules/features2d/src/fast_score.cpp
index 423b1f950d..de697b7c90 100644
--- a/modules/features2d/src/fast_score.cpp
+++ b/modules/features2d/src/fast_score.cpp
@@ -357,4 +357,3 @@ int cornerScore<8>(const uchar* ptr, const int pixel[], int threshold)
 }
 
 } // namespace cv
-
diff --git a/modules/features2d/src/features2d_init.cpp b/modules/features2d/src/features2d_init.cpp
index 85640585f5..780a67b87a 100644
--- a/modules/features2d/src/features2d_init.cpp
+++ b/modules/features2d/src/features2d_init.cpp
@@ -44,7 +44,7 @@
 
 using namespace cv;
 
-Ptr<Feature2D> Feature2D::create( const std::string& feature2DType )
+Ptr<Feature2D> Feature2D::create( const String& feature2DType )
 {
     return Algorithm::create<Feature2D>("Feature2D." + feature2DType);
 }
diff --git a/modules/features2d/src/freak.cpp b/modules/features2d/src/freak.cpp
index 6c9093f7ce..086a2e2e71 100644
--- a/modules/features2d/src/freak.cpp
+++ b/modules/features2d/src/freak.cpp
@@ -206,7 +206,7 @@ void FREAK::buildPattern()
                  descriptionPairs[i] = allPairs[selectedPairs0.at(i)];
         }
         else {
-            CV_Error(CV_StsVecLengthErr, "Input vector does not match the required size");
+            CV_Error(Error::StsVecLengthErr, "Input vector does not match the required size");
         }
     }
     else { // default selected pairs
@@ -548,7 +548,7 @@ std::vector<int> FREAK::selectPairs(const std::vector<Mat>& images
             int idxB = pairStat[m].idx;
             double corr(0);
             // compute correlation between 2 pairs
-            corr = fabs(compareHist(descriptorsFloat.col(idxA), descriptorsFloat.col(idxB), CV_COMP_CORREL));
+            corr = fabs(compareHist(descriptorsFloat.col(idxA), descriptorsFloat.col(idxB), HISTCMP_CORREL));
 
             if( corr > corrMax ) {
                 corrMax = corr;
@@ -575,7 +575,7 @@ std::vector<int> FREAK::selectPairs(const std::vector<Mat>& images
     else {
         if( verbose )
             std::cout << "correlation threshold too small (restrictive)" << std::endl;
-        CV_Error(CV_StsError, "correlation threshold too small (restrictive)");
+        CV_Error(Error::StsError, "correlation threshold too small (restrictive)");
     }
     extAll = false;
     return idxBestPairs;
diff --git a/modules/features2d/src/keypoint.cpp b/modules/features2d/src/keypoint.cpp
index a4b960d1b3..0cf7ae0571 100644
--- a/modules/features2d/src/keypoint.cpp
+++ b/modules/features2d/src/keypoint.cpp
@@ -44,129 +44,6 @@
 namespace cv
 {
 
-size_t KeyPoint::hash() const
-{
-    size_t _Val = 2166136261U, scale = 16777619U;
-    Cv32suf u;
-    u.f = pt.x; _Val = (scale * _Val) ^ u.u;
-    u.f = pt.y; _Val = (scale * _Val) ^ u.u;
-    u.f = size; _Val = (scale * _Val) ^ u.u;
-    u.f = angle; _Val = (scale * _Val) ^ u.u;
-    u.f = response; _Val = (scale * _Val) ^ u.u;
-    _Val = (scale * _Val) ^ ((size_t) octave);
-    _Val = (scale * _Val) ^ ((size_t) class_id);
-    return _Val;
-}
-
-void write(FileStorage& fs, const std::string& objname, const std::vector<KeyPoint>& keypoints)
-{
-    WriteStructContext ws(fs, objname, CV_NODE_SEQ + CV_NODE_FLOW);
-
-    int i, npoints = (int)keypoints.size();
-    for( i = 0; i < npoints; i++ )
-    {
-        const KeyPoint& kpt = keypoints[i];
-        write(fs, kpt.pt.x);
-        write(fs, kpt.pt.y);
-        write(fs, kpt.size);
-        write(fs, kpt.angle);
-        write(fs, kpt.response);
-        write(fs, kpt.octave);
-        write(fs, kpt.class_id);
-    }
-}
-
-
-void read(const FileNode& node, std::vector<KeyPoint>& keypoints)
-{
-    keypoints.resize(0);
-    FileNodeIterator it = node.begin(), it_end = node.end();
-    for( ; it != it_end; )
-    {
-        KeyPoint kpt;
-        it >> kpt.pt.x >> kpt.pt.y >> kpt.size >> kpt.angle >> kpt.response >> kpt.octave >> kpt.class_id;
-        keypoints.push_back(kpt);
-    }
-}
-
-
-void KeyPoint::convert(const std::vector<KeyPoint>& keypoints, std::vector<Point2f>& points2f,
-                       const std::vector<int>& keypointIndexes)
-{
-    if( keypointIndexes.empty() )
-    {
-        points2f.resize( keypoints.size() );
-        for( size_t i = 0; i < keypoints.size(); i++ )
-            points2f[i] = keypoints[i].pt;
-    }
-    else
-    {
-        points2f.resize( keypointIndexes.size() );
-        for( size_t i = 0; i < keypointIndexes.size(); i++ )
-        {
-            int idx = keypointIndexes[i];
-            if( idx >= 0 )
-                points2f[i] = keypoints[idx].pt;
-            else
-            {
-                CV_Error( CV_StsBadArg, "keypointIndexes has element < 0. TODO: process this case" );
-                //points2f[i] = Point2f(-1, -1);
-            }
-        }
-    }
-}
-
-void KeyPoint::convert( const std::vector<Point2f>& points2f, std::vector<KeyPoint>& keypoints,
-                        float size, float response, int octave, int class_id )
-{
-    keypoints.resize(points2f.size());
-    for( size_t i = 0; i < points2f.size(); i++ )
-        keypoints[i] = KeyPoint(points2f[i], size, -1, response, octave, class_id);
-}
-
-float KeyPoint::overlap( const KeyPoint& kp1, const KeyPoint& kp2 )
-{
-    float a = kp1.size * 0.5f;
-    float b = kp2.size * 0.5f;
-    float a_2 = a * a;
-    float b_2 = b * b;
-
-    Point2f p1 = kp1.pt;
-    Point2f p2 = kp2.pt;
-    float c = (float)norm( p1 - p2 );
-
-    float ovrl = 0.f;
-
-    // one circle is completely encovered by the other => no intersection points!
-    if( std::min( a, b ) + c <= std::max( a, b ) )
-        return std::min( a_2, b_2 ) / std::max( a_2, b_2 );
-
-    if( c < a + b ) // circles intersect
-    {
-        float c_2 = c * c;
-        float cosAlpha = ( b_2 + c_2 - a_2 ) / ( kp2.size * c );
-        float cosBeta  = ( a_2 + c_2 - b_2 ) / ( kp1.size * c );
-        float alpha = acos( cosAlpha );
-        float beta = acos( cosBeta );
-        float sinAlpha = sin(alpha);
-        float sinBeta  = sin(beta);
-
-        float segmentAreaA = a_2 * beta;
-        float segmentAreaB = b_2 * alpha;
-
-        float triangleAreaA = a_2 * sinBeta * cosBeta;
-        float triangleAreaB = b_2 * sinAlpha * cosAlpha;
-
-        float intersectionArea = segmentAreaA + segmentAreaB - triangleAreaA - triangleAreaB;
-        float unionArea = (a_2 + b_2) * (float)CV_PI - intersectionArea;
-
-        ovrl = intersectionArea / unionArea;
-    }
-
-    return ovrl;
-}
-
-
 struct KeypointResponseGreaterThanThreshold
 {
     KeypointResponseGreaterThanThreshold(float _value) :
@@ -192,7 +69,7 @@ struct KeypointResponseGreater
 void KeyPointsFilter::retainBest(std::vector<KeyPoint>& keypoints, int n_points)
 {
     //this is only necessary if the keypoints size is greater than the number of desired points.
-    if( n_points > 0 && keypoints.size() > (size_t)n_points )
+    if( n_points >= 0 && keypoints.size() > (size_t)n_points )
     {
         if (n_points==0)
         {
diff --git a/modules/features2d/src/matchers.cpp b/modules/features2d/src/matchers.cpp
index 658c902f93..087c6a78bd 100644
--- a/modules/features2d/src/matchers.cpp
+++ b/modules/features2d/src/matchers.cpp
@@ -40,8 +40,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include <limits>
 
-#include "opencv2/core/internal.hpp"
 #if defined(HAVE_EIGEN) && EIGEN_WORLD_VERSION == 2
 #include <Eigen/Array>
 #endif
@@ -113,7 +113,7 @@ void DescriptorMatcher::DescriptorCollection::set( const std::vector<Mat>& descr
         dim = descriptors[0].cols;
         type = descriptors[0].type();
     }
-    assert( dim > 0 );
+    CV_Assert( dim > 0 );
 
     int count = startIdxs[imageCount-1] + descriptors[imageCount-1].rows;
 
@@ -326,7 +326,7 @@ BFMatcher::BFMatcher( int _normType, bool _crossCheck )
 
 Ptr<DescriptorMatcher> BFMatcher::clone( bool emptyTrainData ) const
 {
-    BFMatcher* matcher = new BFMatcher(normType, crossCheck);
+    Ptr<BFMatcher> matcher = makePtr<BFMatcher>(normType, crossCheck);
     if( !emptyTrainData )
     {
         matcher->trainDescCollection.resize(trainDescCollection.size());
@@ -456,36 +456,36 @@ void BFMatcher::radiusMatchImpl( const Mat& queryDescriptors, std::vector<std::v
 /*
  * Factory function for DescriptorMatcher creating
  */
-Ptr<DescriptorMatcher> DescriptorMatcher::create( const std::string& descriptorMatcherType )
+Ptr<DescriptorMatcher> DescriptorMatcher::create( const String& descriptorMatcherType )
 {
-    DescriptorMatcher* dm = 0;
+    Ptr<DescriptorMatcher> dm;
     if( !descriptorMatcherType.compare( "FlannBased" ) )
     {
-        dm = new FlannBasedMatcher();
+        dm = makePtr<FlannBasedMatcher>();
     }
     else if( !descriptorMatcherType.compare( "BruteForce" ) ) // L2
     {
-        dm = new BFMatcher(NORM_L2);
+        dm = makePtr<BFMatcher>(int(NORM_L2)); // anonymous enums can't be template parameters
     }
     else if( !descriptorMatcherType.compare( "BruteForce-SL2" ) ) // Squared L2
     {
-        dm = new BFMatcher(NORM_L2SQR);
+        dm = makePtr<BFMatcher>(int(NORM_L2SQR));
     }
     else if( !descriptorMatcherType.compare( "BruteForce-L1" ) )
     {
-        dm = new BFMatcher(NORM_L1);
+        dm = makePtr<BFMatcher>(int(NORM_L1));
     }
     else if( !descriptorMatcherType.compare("BruteForce-Hamming") ||
              !descriptorMatcherType.compare("BruteForce-HammingLUT") )
     {
-        dm = new BFMatcher(NORM_HAMMING);
+        dm = makePtr<BFMatcher>(int(NORM_HAMMING));
     }
     else if( !descriptorMatcherType.compare("BruteForce-Hamming(2)") )
     {
-        dm = new BFMatcher(NORM_HAMMING2);
+        dm = makePtr<BFMatcher>(int(NORM_HAMMING2));
     }
     else
-        CV_Error( CV_StsBadArg, "Unknown matcher name" );
+        CV_Error( Error::StsBadArg, "Unknown matcher name" );
 
     return dm;
 }
@@ -497,8 +497,8 @@ Ptr<DescriptorMatcher> DescriptorMatcher::create( const std::string& descriptorM
 FlannBasedMatcher::FlannBasedMatcher( const Ptr<flann::IndexParams>& _indexParams, const Ptr<flann::SearchParams>& _searchParams )
     : indexParams(_indexParams), searchParams(_searchParams), addedDescCount(0)
 {
-    CV_Assert( !_indexParams.empty() );
-    CV_Assert( !_searchParams.empty() );
+    CV_Assert( _indexParams );
+    CV_Assert( _searchParams );
 }
 
 void FlannBasedMatcher::add( const std::vector<Mat>& descriptors )
@@ -522,17 +522,17 @@ void FlannBasedMatcher::clear()
 
 void FlannBasedMatcher::train()
 {
-    if( flannIndex.empty() || mergedDescriptors.size() < addedDescCount )
+    if( !flannIndex || mergedDescriptors.size() < addedDescCount )
     {
         mergedDescriptors.set( trainDescCollection );
-        flannIndex = new flann::Index( mergedDescriptors.getDescriptors(), *indexParams );
+        flannIndex = makePtr<flann::Index>( mergedDescriptors.getDescriptors(), *indexParams );
     }
 }
 
 void FlannBasedMatcher::read( const FileNode& fn)
 {
-     if (indexParams.empty())
-         indexParams = new flann::IndexParams();
+     if (!indexParams)
+         indexParams = makePtr<flann::IndexParams>();
 
      FileNode ip = fn["indexParams"];
      CV_Assert(ip.type() == FileNode::SEQ);
@@ -540,7 +540,7 @@ void FlannBasedMatcher::read( const FileNode& fn)
      for(int i = 0; i < (int)ip.size(); ++i)
      {
         CV_Assert(ip[i].type() == FileNode::MAP);
-        std::string _name =  (std::string)ip[i]["name"];
+        String _name =  (String)ip[i]["name"];
         int type =  (int)ip[i]["type"];
 
         switch(type)
@@ -559,7 +559,7 @@ void FlannBasedMatcher::read( const FileNode& fn)
             indexParams->setDouble(_name, (double) ip[i]["value"]);
             break;
         case CV_USRTYPE1:
-            indexParams->setString(_name, (std::string) ip[i]["value"]);
+            indexParams->setString(_name, (String) ip[i]["value"]);
             break;
         case CV_MAKETYPE(CV_USRTYPE1,2):
             indexParams->setBool(_name, (int) ip[i]["value"] != 0);
@@ -570,8 +570,8 @@ void FlannBasedMatcher::read( const FileNode& fn)
         };
      }
 
-     if (searchParams.empty())
-         searchParams = new flann::SearchParams();
+     if (!searchParams)
+         searchParams = makePtr<flann::SearchParams>();
 
      FileNode sp = fn["searchParams"];
      CV_Assert(sp.type() == FileNode::SEQ);
@@ -579,7 +579,7 @@ void FlannBasedMatcher::read( const FileNode& fn)
      for(int i = 0; i < (int)sp.size(); ++i)
      {
         CV_Assert(sp[i].type() == FileNode::MAP);
-        std::string _name =  (std::string)sp[i]["name"];
+        String _name =  (String)sp[i]["name"];
         int type =  (int)sp[i]["type"];
 
         switch(type)
@@ -598,7 +598,7 @@ void FlannBasedMatcher::read( const FileNode& fn)
             searchParams->setDouble(_name, (double) ip[i]["value"]);
             break;
         case CV_USRTYPE1:
-            searchParams->setString(_name, (std::string) ip[i]["value"]);
+            searchParams->setString(_name, (String) ip[i]["value"]);
             break;
         case CV_MAKETYPE(CV_USRTYPE1,2):
             searchParams->setBool(_name, (int) ip[i]["value"] != 0);
@@ -618,9 +618,9 @@ void FlannBasedMatcher::write( FileStorage& fs) const
 
      if (indexParams)
      {
-         std::vector<std::string> names;
+         std::vector<String> names;
          std::vector<int> types;
-         std::vector<std::string> strValues;
+         std::vector<String> strValues;
          std::vector<double> numValues;
 
          indexParams->getAll(names, types, strValues, numValues);
@@ -669,9 +669,9 @@ void FlannBasedMatcher::write( FileStorage& fs) const
 
      if (searchParams)
      {
-         std::vector<std::string> names;
+         std::vector<String> names;
          std::vector<int> types;
-         std::vector<std::string> strValues;
+         std::vector<String> strValues;
          std::vector<double> numValues;
 
          searchParams->getAll(names, types, strValues, numValues);
@@ -725,10 +725,10 @@ bool FlannBasedMatcher::isMaskSupported() const
 
 Ptr<DescriptorMatcher> FlannBasedMatcher::clone( bool emptyTrainData ) const
 {
-    FlannBasedMatcher* matcher = new FlannBasedMatcher(indexParams, searchParams);
+    Ptr<FlannBasedMatcher> matcher = makePtr<FlannBasedMatcher>(indexParams, searchParams);
     if( !emptyTrainData )
     {
-        CV_Error( CV_StsNotImplemented, "deep clone functionality is not implemented, because "
+        CV_Error( Error::StsNotImplemented, "deep clone functionality is not implemented, because "
                   "Flann::Index has not copy constructor or clone method ");
         //matcher->flannIndex;
         matcher->addedDescCount = addedDescCount;
@@ -1060,13 +1060,13 @@ bool GenericDescriptorMatcher::empty() const
 /*
  * Factory function for GenericDescriptorMatch creating
  */
-Ptr<GenericDescriptorMatcher> GenericDescriptorMatcher::create( const std::string& genericDescritptorMatcherType,
-                                                                const std::string &paramsFilename )
+Ptr<GenericDescriptorMatcher> GenericDescriptorMatcher::create( const String& genericDescritptorMatcherType,
+                                                                const String &paramsFilename )
 {
     Ptr<GenericDescriptorMatcher> descriptorMatcher =
         Algorithm::create<GenericDescriptorMatcher>("DescriptorMatcher." + genericDescritptorMatcherType);
 
-    if( !paramsFilename.empty() && !descriptorMatcher.empty() )
+    if( !paramsFilename.empty() && descriptorMatcher )
     {
         FileStorage fs = FileStorage( paramsFilename, FileStorage::READ );
         if( fs.isOpened() )
@@ -1086,7 +1086,7 @@ VectorDescriptorMatcher::VectorDescriptorMatcher( const Ptr<DescriptorExtractor>
                                                   const Ptr<DescriptorMatcher>& _matcher )
                                 : extractor( _extractor ), matcher( _matcher )
 {
-    CV_Assert( !extractor.empty() && !matcher.empty() );
+    CV_Assert( extractor && matcher );
 }
 
 VectorDescriptorMatcher::~VectorDescriptorMatcher()
@@ -1152,14 +1152,14 @@ void VectorDescriptorMatcher::write (FileStorage& fs) const
 
 bool VectorDescriptorMatcher::empty() const
 {
-    return extractor.empty() || extractor->empty() ||
-           matcher.empty() || matcher->empty();
+    return !extractor || extractor->empty() ||
+           !matcher || matcher->empty();
 }
 
 Ptr<GenericDescriptorMatcher> VectorDescriptorMatcher::clone( bool emptyTrainData ) const
 {
     // TODO clone extractor
-    return new VectorDescriptorMatcher( extractor, matcher->clone(emptyTrainData) );
+    return makePtr<VectorDescriptorMatcher>( extractor, matcher->clone(emptyTrainData) );
 }
 
 }
diff --git a/modules/features2d/src/mser.cpp b/modules/features2d/src/mser.cpp
index 6581612dfc..a68a5a5b7a 100644
--- a/modules/features2d/src/mser.cpp
+++ b/modules/features2d/src/mser.cpp
@@ -40,6 +40,8 @@
  */
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include <limits>
 
 namespace cv
 {
@@ -1054,10 +1056,11 @@ static int preprocessMSER_8UC3( MSCRNode* node,
     return Ne;
 }
 
-#define cmp_mscr_edge(edge1, edge2) \
-    ((edge1).chi < (edge2).chi)
-
-static CV_IMPLEMENT_QSORT( QuickSortMSCREdge, MSCREdge, cmp_mscr_edge )
+class LessThanEdge
+{
+public:
+    bool operator()(const MSCREdge& a, const MSCREdge& b) const { return a.chi < b.chi; }
+};
 
 // to find the root of one region
 static MSCRNode* findMSCR( MSCRNode* x )
@@ -1112,7 +1115,7 @@ extractMSER_8UC3( CvMat* src,
     CvMat* dy = cvCreateMat( src->rows-1, src->cols, CV_64FC1 );
     Ne = preprocessMSER_8UC3( map, edge, &emean, src, mask, dx, dy, Ne, params.edgeBlurSize );
     emean = emean / (double)Ne;
-    QuickSortMSCREdge( edge, Ne, 0 );
+    std::sort(edge, edge + Ne, LessThanEdge());
     MSCREdge* edge_ub = edge+Ne;
     MSCREdge* edgeptr = edge;
     TempMSCR* mscrptr = mscr;
diff --git a/modules/features2d/src/orb.cpp b/modules/features2d/src/orb.cpp
index 33c23d1610..7ed3ff03c3 100644
--- a/modules/features2d/src/orb.cpp
+++ b/modules/features2d/src/orb.cpp
@@ -240,7 +240,7 @@ static void computeOrbDescriptor(const KeyPoint& kpt,
         }
     }
     else
-        CV_Error( CV_StsBadSize, "Wrong WTA_K. It can be only 2, 3 or 4." );
+        CV_Error( Error::StsBadSize, "Wrong WTA_K. It can be only 2, 3 or 4." );
 
     #undef GET_VALUE
 }
@@ -738,7 +738,7 @@ void ORB::operator()( InputArray _image, InputArray _mask, std::vector<KeyPoint>
 
     Mat image = _image.getMat(), mask = _mask.getMat();
     if( image.type() != CV_8UC1 )
-        cvtColor(_image, image, CV_BGR2GRAY);
+        cvtColor(_image, image, COLOR_BGR2GRAY);
 
     int levelsNum = this->nlevels;
 
diff --git a/modules/features2d/src/precomp.cpp b/modules/features2d/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/features2d/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/features2d/src/precomp.hpp b/modules/features2d/src/precomp.hpp
index 62936b48f3..3c9073ec52 100644
--- a/modules/features2d/src/precomp.hpp
+++ b/modules/features2d/src/precomp.hpp
@@ -43,14 +43,11 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/features2d.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
 
 #include <algorithm>
 
diff --git a/modules/features2d/src/stardetector.cpp b/modules/features2d/src/stardetector.cpp
index 3dbea0ac13..02b999b620 100644
--- a/modules/features2d/src/stardetector.cpp
+++ b/modules/features2d/src/stardetector.cpp
@@ -114,7 +114,7 @@ StarDetectorComputeResponses( const Mat& img, Mat& responses, Mat& sizes, int ma
 #if CV_SSE2
     __m128 invSizes4[MAX_PATTERN][2];
     __m128 sizes1_4[MAX_PATTERN];
-    Cv32suf absmask;
+    union { int i; float f; } absmask;
     absmask.i = 0x7fffffff;
     volatile bool useSIMD = cv::checkHardwareSupport(CV_CPU_SSE2);
 #endif
@@ -429,7 +429,7 @@ StarDetector::StarDetector(int _maxSize, int _responseThreshold,
 void StarDetector::detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask ) const
 {
     Mat grayImage = image;
-    if( image.type() != CV_8U ) cvtColor( image, grayImage, CV_BGR2GRAY );
+    if( image.type() != CV_8U ) cvtColor( image, grayImage, COLOR_BGR2GRAY );
 
     (*this)(grayImage, keypoints);
     KeyPointsFilter::runByPixelsMask( keypoints, mask );
diff --git a/modules/features2d/test/test_brisk.cpp b/modules/features2d/test/test_brisk.cpp
index c76afff570..38e07c3f6f 100644
--- a/modules/features2d/test/test_brisk.cpp
+++ b/modules/features2d/test/test_brisk.cpp
@@ -69,8 +69,8 @@ void CV_BRISKTest::run( int )
     }
 
   Mat gray1, gray2;
-  cvtColor(image1, gray1, CV_BGR2GRAY);
-  cvtColor(image2, gray2, CV_BGR2GRAY);
+  cvtColor(image1, gray1, COLOR_BGR2GRAY);
+  cvtColor(image2, gray2, COLOR_BGR2GRAY);
 
   Ptr<FeatureDetector> detector = Algorithm::create<FeatureDetector>("Feature2D.BRISK");
 
@@ -93,4 +93,3 @@ void CV_BRISKTest::run( int )
 }
 
 TEST(Features2d_BRISK, regression) { CV_BRISKTest test; test.safe_run(); }
-
diff --git a/modules/features2d/test/test_descriptors_regression.cpp b/modules/features2d/test/test_descriptors_regression.cpp
index 193545162b..08eb59eae2 100644
--- a/modules/features2d/test/test_descriptors_regression.cpp
+++ b/modules/features2d/test/test_descriptors_regression.cpp
@@ -80,12 +80,16 @@ static Mat readMatFromBin( const string& filename )
         size_t elements_read4 = fread( (void*)&dataSize, sizeof(int), 1, f );
         CV_Assert(elements_read1 == 1 && elements_read2 == 1 && elements_read3 == 1 && elements_read4 == 1);
 
-        uchar* data = (uchar*)cvAlloc(dataSize);
-        size_t elements_read = fread( (void*)data, 1, dataSize, f );
+        size_t step = dataSize / rows / CV_ELEM_SIZE(type);
+        CV_Assert(step >= (size_t)cols);
+
+        Mat m = Mat( rows, step, type).colRange(0, cols);
+
+        size_t elements_read = fread( m.ptr(), 1, dataSize, f );
         CV_Assert(elements_read == (size_t)(dataSize));
         fclose(f);
 
-        return Mat( rows, cols, type, data );
+        return m;
     }
     return Mat();
 }
@@ -137,7 +141,7 @@ protected:
 
     void emptyDataTest()
     {
-        assert( !dextractor.empty() );
+        assert( dextractor );
 
         // One image.
         Mat image;
@@ -182,7 +186,7 @@ protected:
 
     void regressionTest()
     {
-        assert( !dextractor.empty() );
+        assert( dextractor );
 
         // Read the test image.
         string imgFilename =  string(ts->get_data_path()) + FEATURES2D_DIR + "/" + IMAGE_FILENAME;
@@ -205,7 +209,7 @@ protected:
             double t = (double)getTickCount();
             dextractor->compute( img, keypoints, calcDescriptors );
             t = getTickCount() - t;
-            ts->printf(cvtest::TS::LOG, "\nAverage time of computing one descriptor = %g ms.\n", t/((double)cvGetTickFrequency()*1000.)/calcDescriptors.rows);
+            ts->printf(cvtest::TS::LOG, "\nAverage time of computing one descriptor = %g ms.\n", t/((double)getTickFrequency()*1000.)/calcDescriptors.rows);
 
             if( calcDescriptors.rows != (int)keypoints.size() )
             {
@@ -263,7 +267,7 @@ protected:
     void run(int)
     {
         createDescriptorExtractor();
-        if( dextractor.empty() )
+        if( !dextractor )
         {
             ts->printf(cvtest::TS::LOG, "Descriptor extractor is empty.\n");
             ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
diff --git a/modules/features2d/test/test_detectors_regression.cpp b/modules/features2d/test/test_detectors_regression.cpp
index 9a88c42f7b..8f34913a9a 100644
--- a/modules/features2d/test/test_detectors_regression.cpp
+++ b/modules/features2d/test/test_detectors_regression.cpp
@@ -230,7 +230,7 @@ void CV_FeatureDetectorTest::regressionTest()
 
 void CV_FeatureDetectorTest::run( int /*start_from*/ )
 {
-    if( fdetector.empty() )
+    if( !fdetector )
     {
         ts->printf( cvtest::TS::LOG, "Feature detector is empty.\n" );
         ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
diff --git a/modules/features2d/test/test_fast.cpp b/modules/features2d/test/test_fast.cpp
index cee9152bff..d500ce549c 100644
--- a/modules/features2d/test/test_fast.cpp
+++ b/modules/features2d/test/test_fast.cpp
@@ -71,8 +71,8 @@ void CV_FastTest::run( int )
     }
 
     Mat gray1, gray2;
-    cvtColor(image1, gray1, CV_BGR2GRAY);
-    cvtColor(image2, gray2, CV_BGR2GRAY);
+    cvtColor(image1, gray1, COLOR_BGR2GRAY);
+    cvtColor(image2, gray2, COLOR_BGR2GRAY);
 
     vector<KeyPoint> keypoints1;
     vector<KeyPoint> keypoints2;
@@ -82,13 +82,13 @@ void CV_FastTest::run( int )
     for(size_t i = 0; i < keypoints1.size(); ++i)
     {
         const KeyPoint& kp = keypoints1[i];
-        cv::circle(image1, kp.pt, cvRound(kp.size/2), CV_RGB(255, 0, 0));
+        cv::circle(image1, kp.pt, cvRound(kp.size/2), Scalar(255, 0, 0));
     }
 
     for(size_t i = 0; i < keypoints2.size(); ++i)
     {
         const KeyPoint& kp = keypoints2[i];
-        cv::circle(image2, kp.pt, cvRound(kp.size/2), CV_RGB(255, 0, 0));
+        cv::circle(image2, kp.pt, cvRound(kp.size/2), Scalar(255, 0, 0));
     }
 
     Mat kps1(1, (int)(keypoints1.size() * sizeof(KeyPoint)), CV_8U, &keypoints1[0]);
@@ -135,4 +135,3 @@ void CV_FastTest::run( int )
 }
 
 TEST(Features2d_FAST, regression) { CV_FastTest test; test.safe_run(); }
-
diff --git a/modules/features2d/test/test_keypoints.cpp b/modules/features2d/test/test_keypoints.cpp
index 7bc5c6e9a2..e15d4fa17f 100644
--- a/modules/features2d/test/test_keypoints.cpp
+++ b/modules/features2d/test/test_keypoints.cpp
@@ -62,7 +62,7 @@ protected:
     virtual void run(int)
     {
         cv::initModule_features2d();
-        CV_Assert(!detector.empty());
+        CV_Assert(detector);
         string imgFilename = string(ts->get_data_path()) + FEATURES2D_DIR + "/" + IMAGE_FILENAME;
 
         // Read the test image.
@@ -166,5 +166,3 @@ TEST(Features2d_Detector_Keypoints_Dense, validation)
     CV_FeatureDetectorKeypointsTest test(Algorithm::create<FeatureDetector>("Feature2D.Dense"));
     test.safe_run();
 }
-
-
diff --git a/modules/features2d/test/test_mser.cpp b/modules/features2d/test/test_mser.cpp
index b30400a502..6b1b4ef639 100644
--- a/modules/features2d/test/test_mser.cpp
+++ b/modules/features2d/test/test_mser.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 
 #include <vector>
 #include <string>
@@ -204,4 +205,3 @@ void CV_MserTest::run(int)
 }
 
 TEST(Features2d_MSER, DISABLED_regression) { CV_MserTest test; test.safe_run(); }
-
diff --git a/modules/features2d/test/test_precomp.hpp b/modules/features2d/test/test_precomp.hpp
index 979e7c78cb..49bc1dfd18 100644
--- a/modules/features2d/test/test_precomp.hpp
+++ b/modules/features2d/test/test_precomp.hpp
@@ -11,7 +11,6 @@
 
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/features2d.hpp"
 #include "opencv2/highgui.hpp"
 #include <iostream>
diff --git a/modules/features2d/test/test_rotation_and_scale_invariance.cpp b/modules/features2d/test/test_rotation_and_scale_invariance.cpp
index 1f67da17be..adfe42804b 100644
--- a/modules/features2d/test/test_rotation_and_scale_invariance.cpp
+++ b/modules/features2d/test/test_rotation_and_scale_invariance.cpp
@@ -196,7 +196,7 @@ public:
         minKeyPointMatchesRatio(_minKeyPointMatchesRatio),
         minAngleInliersRatio(_minAngleInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
+        CV_Assert(featureDetector);
     }
 
 protected:
@@ -217,7 +217,7 @@ protected:
         vector<KeyPoint> keypoints0;
         featureDetector->detect(image0, keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
         const int maxAngle = 360, angleStep = 15;
         for(int angle = 0; angle < maxAngle; angle += angleStep)
@@ -246,7 +246,7 @@ protected:
                 float angle0 = keypoints0[matches[m].queryIdx].angle;
                 float angle1 = keypoints1[matches[m].trainIdx].angle;
                 if(angle0 == -1 || angle1 == -1)
-                    CV_Error(CV_StsBadArg, "Given FeatureDetector is not rotation invariant, it can not be tested here.\n");
+                    CV_Error(Error::StsBadArg, "Given FeatureDetector is not rotation invariant, it can not be tested here.\n");
                 CV_Assert(angle0 >= 0.f && angle0 < 360.f);
                 CV_Assert(angle1 >= 0.f && angle1 < 360.f);
 
@@ -307,8 +307,8 @@ public:
         normType(_normType),
         minDescInliersRatio(_minDescInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
-        CV_Assert(!descriptorExtractor.empty());
+        CV_Assert(featureDetector);
+        CV_Assert(descriptorExtractor);
     }
 
 protected:
@@ -330,7 +330,7 @@ protected:
         Mat descriptors0;
         featureDetector->detect(image0, keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
         descriptorExtractor->compute(image0, keypoints0, descriptors0);
 
         BFMatcher bfmatcher(normType);
@@ -392,7 +392,7 @@ public:
         minKeyPointMatchesRatio(_minKeyPointMatchesRatio),
         minScaleInliersRatio(_minScaleInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
+        CV_Assert(featureDetector);
     }
 
 protected:
@@ -413,7 +413,7 @@ protected:
         vector<KeyPoint> keypoints0;
         featureDetector->detect(image0, keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
         for(int scaleIdx = 1; scaleIdx <= 3; scaleIdx++)
         {
@@ -424,7 +424,7 @@ protected:
             vector<KeyPoint> keypoints1, osiKeypoints1; // osi - original size image
             featureDetector->detect(image1, keypoints1);
             if(keypoints1.size() < 15)
-                CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+                CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
             if(keypoints1.size() > keypoints0.size())
             {
@@ -510,8 +510,8 @@ public:
         normType(_normType),
         minDescInliersRatio(_minDescInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
-        CV_Assert(!descriptorExtractor.empty());
+        CV_Assert(featureDetector);
+        CV_Assert(descriptorExtractor);
     }
 
 protected:
@@ -532,7 +532,7 @@ protected:
         vector<KeyPoint> keypoints0;
         featureDetector->detect(image0, keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
         Mat descriptors0;
         descriptorExtractor->compute(image0, keypoints0, descriptors0);
 
diff --git a/modules/flann/CMakeLists.txt b/modules/flann/CMakeLists.txt
index 645e8ce857..a6326c40a0 100644
--- a/modules/flann/CMakeLists.txt
+++ b/modules/flann/CMakeLists.txt
@@ -1,3 +1,2 @@
 set(the_description "Clustering and Search in Multi-Dimensional Spaces")
 ocv_define_module(flann opencv_core)
-
diff --git a/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.rst b/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.rst
index 07b0b2a467..e058e8886e 100644
--- a/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.rst
+++ b/modules/flann/doc/flann_fast_approximate_nearest_neighbor_search.rst
@@ -138,7 +138,7 @@ The method constructs a fast search structure from a set of features using the s
 
             struct SavedIndexParams : public IndexParams
             {
-                SavedIndexParams( std::string filename );
+                SavedIndexParams( String filename );
             };
 
 
@@ -199,7 +199,7 @@ flann::Index_<T>::save
 ------------------------------
 Saves the index to a file.
 
-.. ocv:function:: void flann::Index_<T>::save(std::string filename)
+.. ocv:function:: void flann::Index_<T>::save(String filename)
 
     :param filename: The file to save the index to
 
diff --git a/modules/flann/include/opencv2/flann.hpp b/modules/flann/include/opencv2/flann.hpp
index d5eedcc0f0..36ca8c7c3a 100644
--- a/modules/flann/include/opencv2/flann.hpp
+++ b/modules/flann/include/opencv2/flann.hpp
@@ -43,12 +43,9 @@
 #ifndef _OPENCV_FLANN_HPP_
 #define _OPENCV_FLANN_HPP_
 
-#ifdef __cplusplus
-
-#include "opencv2/core/types_c.h"
 #include "opencv2/core.hpp"
-#include "opencv2/flann/flann_base.hpp"
 #include "opencv2/flann/miniflann.hpp"
+#include "opencv2/flann/flann_base.hpp"
 
 namespace cvflann
 {
@@ -112,7 +109,7 @@ public:
         int radiusSearch(const Mat& query, Mat& indices, Mat& dists,
                          DistanceType radius, const ::cvflann::SearchParams& params);
 
-        void save(std::string filename) { nnIndex->save(filename); }
+        void save(String filename) { nnIndex->save(filename); }
 
         int veclen() const { return nnIndex->veclen(); }
 
@@ -244,7 +241,7 @@ public:
     int radiusSearch(const std::vector<ElementType>& query, std::vector<int>& indices, std::vector<DistanceType>& dists, DistanceType radius, const ::cvflann::SearchParams& params);
     int radiusSearch(const Mat& query, Mat& indices, Mat& dists, DistanceType radius, const ::cvflann::SearchParams& params);
 
-    void save(std::string filename)
+    void save(String filename)
         {
             if (nnIndex_L1) nnIndex_L1->save(filename);
             if (nnIndex_L2) nnIndex_L2->save(filename);
@@ -422,6 +419,4 @@ FLANN_DEPRECATED int hierarchicalClustering(const Mat& features, Mat& centers, c
 
 } } // namespace cv::flann
 
-#endif // __cplusplus
-
 #endif
diff --git a/modules/flann/include/opencv2/flann/any.h b/modules/flann/include/opencv2/flann/any.h
index dc0b9481a2..070d5a421f 100644
--- a/modules/flann/include/opencv2/flann/any.h
+++ b/modules/flann/include/opencv2/flann/any.h
@@ -106,6 +106,11 @@ template<> inline void big_any_policy<flann_algorithm_t>::print(std::ostream& ou
     out << int(*reinterpret_cast<flann_algorithm_t const*>(*src));
 }
 
+template<> inline void big_any_policy<cv::String>::print(std::ostream& out, void* const* src)
+{
+    out << (*reinterpret_cast<cv::String const*>(*src)).c_str();
+}
+
 template<typename T>
 struct choose_policy
 {
@@ -252,8 +257,7 @@ public:
     const T& cast() const
     {
         if (policy->type() != typeid(T)) throw anyimpl::bad_any_cast();
-        void* obj = const_cast<void*>(object);
-        T* r = reinterpret_cast<T*>(policy->get_value(&obj));
+        T* r = reinterpret_cast<T*>(policy->get_value(const_cast<void **>(&object)));
         return *r;
     }
 
diff --git a/modules/flann/include/opencv2/flann/dist.h b/modules/flann/include/opencv2/flann/dist.h
index bfd37a9e23..dd4cc421d1 100644
--- a/modules/flann/include/opencv2/flann/dist.h
+++ b/modules/flann/include/opencv2/flann/dist.h
@@ -43,8 +43,12 @@ typedef unsigned __int64 uint64_t;
 
 #include "defines.h"
 
+#if (defined WIN32 || defined _WIN32) && defined(_M_ARM)
+# include <Intrin.h>
+#endif
+
 #ifdef __ARM_NEON__
-#include "arm_neon.h"
+# include "arm_neon.h"
 #endif
 
 namespace cvflann
@@ -421,7 +425,6 @@ struct Hamming
     ResultType operator()(Iterator1 a, Iterator2 b, size_t size, ResultType /*worst_dist*/ = -1) const
     {
         ResultType result = 0;
-#ifdef __GNUC__
 #ifdef __ARM_NEON__
         {
             uint32x4_t bits = vmovq_n_u32(0);
@@ -438,7 +441,7 @@ struct Hamming
             result = vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),0);
             result += vgetq_lane_s32 (vreinterpretq_s32_u64(bitSet2),2);
         }
-#else
+#elif __GNUC__
         {
             //for portability just use unsigned long -- and use the __builtin_popcountll (see docs for __builtin_popcountll)
             typedef unsigned long long pop_t;
@@ -458,8 +461,8 @@ struct Hamming
                 result += __builtin_popcountll(a_final ^ b_final);
             }
         }
-#endif //NEON
-#else
+#else // NO NEON and NOT GNUC
+        typedef unsigned long long pop_t;
         HammingLUT lut;
         result = lut(reinterpret_cast<const unsigned char*> (a),
                      reinterpret_cast<const unsigned char*> (b), size * sizeof(pop_t));
diff --git a/modules/flann/include/opencv2/flann/flann.hpp b/modules/flann/include/opencv2/flann/flann.hpp
index f40aaf79f3..227683f979 100644
--- a/modules/flann/include/opencv2/flann/flann.hpp
+++ b/modules/flann/include/opencv2/flann/flann.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/flann.hpp"
\ No newline at end of file
+#include "opencv2/flann.hpp"
diff --git a/modules/flann/include/opencv2/flann/flann_base.hpp b/modules/flann/include/opencv2/flann/flann_base.hpp
index b5ba7d79e2..98c33cf6c0 100644
--- a/modules/flann/include/opencv2/flann/flann_base.hpp
+++ b/modules/flann/include/opencv2/flann/flann_base.hpp
@@ -32,7 +32,6 @@
 #define OPENCV_FLANN_BASE_HPP_
 
 #include <vector>
-#include <string>
 #include <cassert>
 #include <cstdio>
 
@@ -62,7 +61,7 @@ inline void log_verbosity(int level)
  */
 struct SavedIndexParams : public IndexParams
 {
-    SavedIndexParams(std::string filename)
+    SavedIndexParams(cv::String filename)
     {
         (* this)["algorithm"] = FLANN_INDEX_SAVED;
         (*this)["filename"] = filename;
@@ -71,7 +70,7 @@ struct SavedIndexParams : public IndexParams
 
 
 template<typename Distance>
-NNIndex<Distance>* load_saved_index(const Matrix<typename Distance::ElementType>& dataset, const std::string& filename, Distance distance)
+NNIndex<Distance>* load_saved_index(const Matrix<typename Distance::ElementType>& dataset, const cv::String& filename, Distance distance)
 {
     typedef typename Distance::ElementType ElementType;
 
@@ -111,7 +110,7 @@ public:
         loaded_ = false;
 
         if (index_type == FLANN_INDEX_SAVED) {
-            nnIndex_ = load_saved_index<Distance>(features, get_param<std::string>(params,"filename"), distance);
+            nnIndex_ = load_saved_index<Distance>(features, get_param<cv::String>(params,"filename"), distance);
             loaded_ = true;
         }
         else {
@@ -134,7 +133,7 @@ public:
         }
     }
 
-    void save(std::string filename)
+    void save(cv::String filename)
     {
         FILE* fout = fopen(filename.c_str(), "wb");
         if (fout == NULL) {
diff --git a/modules/flann/include/opencv2/flann/general.h b/modules/flann/include/opencv2/flann/general.h
index 87e7e2f288..9d5402a6f2 100644
--- a/modules/flann/include/opencv2/flann/general.h
+++ b/modules/flann/include/opencv2/flann/general.h
@@ -31,19 +31,17 @@
 #ifndef OPENCV_FLANN_GENERAL_H_
 #define OPENCV_FLANN_GENERAL_H_
 
-#include "defines.h"
-#include <stdexcept>
-#include <cassert>
+#include "opencv2/core.hpp"
 
 namespace cvflann
 {
 
-class FLANNException : public std::runtime_error
+class FLANNException : public cv::Exception
 {
 public:
-    FLANNException(const char* message) : std::runtime_error(message) { }
+    FLANNException(const char* message) : cv::Exception(0, message, "", __FILE__, __LINE__) { }
 
-    FLANNException(const std::string& message) : std::runtime_error(message) { }
+    FLANNException(const cv::String& message) : cv::Exception(0, message, "", __FILE__, __LINE__) { }
 };
 
 }
diff --git a/modules/flann/include/opencv2/flann/ground_truth.h b/modules/flann/include/opencv2/flann/ground_truth.h
index 69d978ba07..fd8f3ae7f0 100644
--- a/modules/flann/include/opencv2/flann/ground_truth.h
+++ b/modules/flann/include/opencv2/flann/ground_truth.h
@@ -42,7 +42,6 @@ template <typename Distance>
 void find_nearest(const Matrix<typename Distance::ElementType>& dataset, typename Distance::ElementType* query, int* matches, int nn,
                   int skip = 0, Distance distance = Distance())
 {
-    typedef typename Distance::ElementType ElementType;
     typedef typename Distance::ResultType DistanceType;
     int n = nn + skip;
 
diff --git a/modules/flann/include/opencv2/flann/hdf5.h b/modules/flann/include/opencv2/flann/hdf5.h
index ef3e999738..80d23b9770 100644
--- a/modules/flann/include/opencv2/flann/hdf5.h
+++ b/modules/flann/include/opencv2/flann/hdf5.h
@@ -73,7 +73,7 @@ hid_t get_hdf5_type<double>() { return H5T_NATIVE_DOUBLE; }
 #define CHECK_ERROR(x,y) if ((x)<0) throw FLANNException((y));
 
 template<typename T>
-void save_to_file(const cvflann::Matrix<T>& dataset, const std::string& filename, const std::string& name)
+void save_to_file(const cvflann::Matrix<T>& dataset, const String& filename, const String& name)
 {
 
 #if H5Eset_auto_vers == 2
@@ -125,7 +125,7 @@ void save_to_file(const cvflann::Matrix<T>& dataset, const std::string& filename
 
 
 template<typename T>
-void load_from_file(cvflann::Matrix<T>& dataset, const std::string& filename, const std::string& name)
+void load_from_file(cvflann::Matrix<T>& dataset, const String& filename, const String& name)
 {
     herr_t status;
     hid_t file_id = H5Fopen(filename.c_str(), H5F_ACC_RDWR, H5P_DEFAULT);
@@ -166,7 +166,7 @@ namespace mpi
  * @param name Name of dataset inside file
  */
 template<typename T>
-void load_from_file(cvflann::Matrix<T>& dataset, const std::string& filename, const std::string& name)
+void load_from_file(cvflann::Matrix<T>& dataset, const String& filename, const String& name)
 {
     MPI_Comm comm  = MPI_COMM_WORLD;
     MPI_Info info  = MPI_INFO_NULL;
diff --git a/modules/flann/include/opencv2/flann/hierarchical_clustering_index.h b/modules/flann/include/opencv2/flann/hierarchical_clustering_index.h
index ce2d622450..b8b16941f2 100644
--- a/modules/flann/include/opencv2/flann/hierarchical_clustering_index.h
+++ b/modules/flann/include/opencv2/flann/hierarchical_clustering_index.h
@@ -32,7 +32,6 @@
 #define OPENCV_FLANN_HIERARCHICAL_CLUSTERING_INDEX_H_
 
 #include <algorithm>
-#include <string>
 #include <map>
 #include <cassert>
 #include <limits>
diff --git a/modules/flann/include/opencv2/flann/kmeans_index.h b/modules/flann/include/opencv2/flann/kmeans_index.h
index 3fea956a74..c3b1fd5b8f 100644
--- a/modules/flann/include/opencv2/flann/kmeans_index.h
+++ b/modules/flann/include/opencv2/flann/kmeans_index.h
@@ -32,7 +32,6 @@
 #define OPENCV_FLANN_KMEANS_INDEX_H_
 
 #include <algorithm>
-#include <string>
 #include <map>
 #include <cassert>
 #include <limits>
diff --git a/modules/flann/include/opencv2/flann/lsh_table.h b/modules/flann/include/opencv2/flann/lsh_table.h
index a30642a480..9b3ac09911 100644
--- a/modules/flann/include/opencv2/flann/lsh_table.h
+++ b/modules/flann/include/opencv2/flann/lsh_table.h
@@ -266,9 +266,7 @@ private:
         const size_t key_size_upper_bound = std::min(sizeof(BucketKey) * CHAR_BIT + 1, sizeof(size_t) * CHAR_BIT);
         if (key_size < key_size_lower_bound || key_size >= key_size_upper_bound)
         {
-            std::stringstream errorMessage;
-            errorMessage << "Invalid key_size (=" << key_size << "). Valid values for your system are " << key_size_lower_bound << " <= key_size < " << key_size_upper_bound << ".";
-            CV_Error(CV_StsBadArg, errorMessage.str());
+            CV_Error(cv::Error::StsBadArg, cv::format("Invalid key_size (=%d). Valid values for your system are %d <= key_size < %d.", (int)key_size, (int)key_size_lower_bound, (int)key_size_upper_bound));
         }
 
         speed_level_ = kHash;
diff --git a/modules/flann/include/opencv2/flann/miniflann.hpp b/modules/flann/include/opencv2/flann/miniflann.hpp
index 475335ca39..f2acc23bff 100644
--- a/modules/flann/include/opencv2/flann/miniflann.hpp
+++ b/modules/flann/include/opencv2/flann/miniflann.hpp
@@ -43,8 +43,6 @@
 #ifndef _OPENCV_MINIFLANN_HPP_
 #define _OPENCV_MINIFLANN_HPP_
 
-#ifdef __cplusplus
-
 #include "opencv2/core.hpp"
 #include "opencv2/flann/defines.h"
 
@@ -59,20 +57,20 @@ struct CV_EXPORTS IndexParams
     IndexParams();
     ~IndexParams();
 
-    std::string getString(const std::string& key, const std::string& defaultVal=std::string()) const;
-    int getInt(const std::string& key, int defaultVal=-1) const;
-    double getDouble(const std::string& key, double defaultVal=-1) const;
+    String getString(const String& key, const String& defaultVal=String()) const;
+    int getInt(const String& key, int defaultVal=-1) const;
+    double getDouble(const String& key, double defaultVal=-1) const;
 
-    void setString(const std::string& key, const std::string& value);
-    void setInt(const std::string& key, int value);
-    void setDouble(const std::string& key, double value);
-    void setFloat(const std::string& key, float value);
-    void setBool(const std::string& key, bool value);
+    void setString(const String& key, const String& value);
+    void setInt(const String& key, int value);
+    void setDouble(const String& key, double value);
+    void setFloat(const String& key, float value);
+    void setBool(const String& key, bool value);
     void setAlgorithm(int value);
 
-    void getAll(std::vector<std::string>& names,
+    void getAll(std::vector<String>& names,
                 std::vector<int>& types,
-                std::vector<std::string>& strValues,
+                std::vector<String>& strValues,
                 std::vector<double>& numValues) const;
 
     void* params;
@@ -119,7 +117,7 @@ struct CV_EXPORTS LshIndexParams : public IndexParams
 
 struct CV_EXPORTS SavedIndexParams : public IndexParams
 {
-    SavedIndexParams(const std::string& filename);
+    SavedIndexParams(const String& filename);
 };
 
 struct CV_EXPORTS SearchParams : public IndexParams
@@ -142,8 +140,8 @@ public:
                              OutputArray dists, double radius, int maxResults,
                              const SearchParams& params=SearchParams());
 
-    CV_WRAP virtual void save(const std::string& filename) const;
-    CV_WRAP virtual bool load(InputArray features, const std::string& filename);
+    CV_WRAP virtual void save(const String& filename) const;
+    CV_WRAP virtual bool load(InputArray features, const String& filename);
     CV_WRAP virtual void release();
     CV_WRAP cvflann::flann_distance_t getDistance() const;
     CV_WRAP cvflann::flann_algorithm_t getAlgorithm() const;
@@ -157,6 +155,4 @@ protected:
 
 } } // namespace cv::flann
 
-#endif // __cplusplus
-
 #endif
diff --git a/modules/flann/include/opencv2/flann/nn_index.h b/modules/flann/include/opencv2/flann/nn_index.h
index d14e83a92c..381d4bc3cb 100644
--- a/modules/flann/include/opencv2/flann/nn_index.h
+++ b/modules/flann/include/opencv2/flann/nn_index.h
@@ -31,8 +31,6 @@
 #ifndef OPENCV_FLANN_NNINDEX_H
 #define OPENCV_FLANN_NNINDEX_H
 
-#include <string>
-
 #include "general.h"
 #include "matrix.h"
 #include "result_set.h"
diff --git a/modules/flann/include/opencv2/flann/params.h b/modules/flann/include/opencv2/flann/params.h
index fc2a906198..2f1d5e7caf 100644
--- a/modules/flann/include/opencv2/flann/params.h
+++ b/modules/flann/include/opencv2/flann/params.h
@@ -39,7 +39,7 @@
 namespace cvflann
 {
 
-typedef std::map<std::string, any> IndexParams;
+typedef std::map<cv::String, any> IndexParams;
 
 struct SearchParams : public IndexParams
 {
@@ -56,7 +56,7 @@ struct SearchParams : public IndexParams
 
 
 template<typename T>
-T get_param(const IndexParams& params, std::string name, const T& default_value)
+T get_param(const IndexParams& params, cv::String name, const T& default_value)
 {
     IndexParams::const_iterator it = params.find(name);
     if (it != params.end()) {
@@ -68,14 +68,14 @@ T get_param(const IndexParams& params, std::string name, const T& default_value)
 }
 
 template<typename T>
-T get_param(const IndexParams& params, std::string name)
+T get_param(const IndexParams& params, cv::String name)
 {
     IndexParams::const_iterator it = params.find(name);
     if (it != params.end()) {
         return it->second.cast<T>();
     }
     else {
-        throw FLANNException(std::string("Missing parameter '")+name+std::string("' in the parameters given"));
+        throw FLANNException(cv::String("Missing parameter '")+name+cv::String("' in the parameters given"));
     }
 }
 
diff --git a/modules/flann/include/opencv2/flann/random.h b/modules/flann/include/opencv2/flann/random.h
index 2a67352da3..a3cf5ec53d 100644
--- a/modules/flann/include/opencv2/flann/random.h
+++ b/modules/flann/include/opencv2/flann/random.h
@@ -131,5 +131,3 @@ public:
 }
 
 #endif //OPENCV_FLANN_RANDOM_H
-
-
diff --git a/modules/flann/include/opencv2/flann/result_set.h b/modules/flann/include/opencv2/flann/result_set.h
index 7bb709b76e..3adad46596 100644
--- a/modules/flann/include/opencv2/flann/result_set.h
+++ b/modules/flann/include/opencv2/flann/result_set.h
@@ -540,4 +540,3 @@ private:
 }
 
 #endif //OPENCV_FLANN_RESULTSET_H
-
diff --git a/modules/flann/src/flann.cpp b/modules/flann/src/flann.cpp
index 43d0e1f2a0..7ebb7eda7a 100644
--- a/modules/flann/src/flann.cpp
+++ b/modules/flann/src/flann.cpp
@@ -54,4 +54,4 @@ namespace cvflann
     }
 
     void dummyfunc() {}
-}
\ No newline at end of file
+}
diff --git a/modules/flann/src/miniflann.cpp b/modules/flann/src/miniflann.cpp
index a7e476fdee..efa58927d3 100644
--- a/modules/flann/src/miniflann.cpp
+++ b/modules/flann/src/miniflann.cpp
@@ -26,7 +26,7 @@ IndexParams::IndexParams()
 }
 
 template<typename T>
-T getParam(const IndexParams& _p, const std::string& key, const T& defaultVal=T())
+T getParam(const IndexParams& _p, const String& key, const T& defaultVal=T())
 {
     ::cvflann::IndexParams& p = get_params(_p);
     ::cvflann::IndexParams::const_iterator it = p.find(key);
@@ -36,49 +36,49 @@ T getParam(const IndexParams& _p, const std::string& key, const T& defaultVal=T(
 }
 
 template<typename T>
-void setParam(IndexParams& _p, const std::string& key, const T& value)
+void setParam(IndexParams& _p, const String& key, const T& value)
 {
     ::cvflann::IndexParams& p = get_params(_p);
     p[key] = value;
 }
 
-std::string IndexParams::getString(const std::string& key, const std::string& defaultVal) const
+String IndexParams::getString(const String& key, const String& defaultVal) const
 {
     return getParam(*this, key, defaultVal);
 }
 
-int IndexParams::getInt(const std::string& key, int defaultVal) const
+int IndexParams::getInt(const String& key, int defaultVal) const
 {
     return getParam(*this, key, defaultVal);
 }
 
-double IndexParams::getDouble(const std::string& key, double defaultVal) const
+double IndexParams::getDouble(const String& key, double defaultVal) const
 {
     return getParam(*this, key, defaultVal);
 }
 
 
-void IndexParams::setString(const std::string& key, const std::string& value)
+void IndexParams::setString(const String& key, const String& value)
 {
     setParam(*this, key, value);
 }
 
-void IndexParams::setInt(const std::string& key, int value)
+void IndexParams::setInt(const String& key, int value)
 {
     setParam(*this, key, value);
 }
 
-void IndexParams::setDouble(const std::string& key, double value)
+void IndexParams::setDouble(const String& key, double value)
 {
     setParam(*this, key, value);
 }
 
-void IndexParams::setFloat(const std::string& key, float value)
+void IndexParams::setFloat(const String& key, float value)
 {
     setParam(*this, key, value);
 }
 
-void IndexParams::setBool(const std::string& key, bool value)
+void IndexParams::setBool(const String& key, bool value)
 {
     setParam(*this, key, value);
 }
@@ -88,9 +88,9 @@ void IndexParams::setAlgorithm(int value)
     setParam(*this, "algorithm", (cvflann::flann_algorithm_t)value);
 }
 
-void IndexParams::getAll(std::vector<std::string>& names,
+void IndexParams::getAll(std::vector<String>& names,
             std::vector<int>& types,
-            std::vector<std::string>& strValues,
+            std::vector<String>& strValues,
             std::vector<double>& numValues) const
 {
     names.clear();
@@ -106,7 +106,7 @@ void IndexParams::getAll(std::vector<std::string>& names,
         names.push_back(it->first);
         try
         {
-            std::string val = it->second.cast<std::string>();
+            String val = it->second.cast<String>();
             types.push_back(CV_USRTYPE1);
             strValues.push_back(val);
             numValues.push_back(-1);
@@ -285,9 +285,9 @@ LshIndexParams::LshIndexParams(int table_number, int key_size, int multi_probe_l
     p["multi_probe_level"] = multi_probe_level;
 }
 
-SavedIndexParams::SavedIndexParams(const std::string& _filename)
+SavedIndexParams::SavedIndexParams(const String& _filename)
 {
-    std::string filename = _filename;
+    String filename = _filename;
     ::cvflann::IndexParams& p = get_params(*this);
 
     p["algorithm"] = FLANN_INDEX_SAVED;
@@ -312,9 +312,9 @@ buildIndex_(void*& index, const Mat& data, const IndexParams& params, const Dist
 {
     typedef typename Distance::ElementType ElementType;
     if(DataType<ElementType>::type != data.type())
-        CV_Error_(CV_StsUnsupportedFormat, ("type=%d\n", data.type()));
+        CV_Error_(Error::StsUnsupportedFormat, ("type=%d\n", data.type()));
     if(!data.isContinuous())
-        CV_Error(CV_StsBadArg, "Only continuous arrays are supported");
+        CV_Error(Error::StsBadArg, "Only continuous arrays are supported");
 
     ::cvflann::Matrix<ElementType> dataset((ElementType*)data.data, data.rows, data.cols);
     IndexType* _index = new IndexType(dataset, get_params(params), dist);
@@ -357,7 +357,7 @@ void Index::build(InputArray _data, const IndexParams& params, flann_distance_t
     algo = getParam<flann_algorithm_t>(params, "algorithm", FLANN_INDEX_LINEAR);
     if( algo == FLANN_INDEX_SAVED )
     {
-        load(_data, getParam<std::string>(params, "filename", std::string()));
+        load(_data, getParam<String>(params, "filename", String()));
         return;
     }
 
@@ -400,7 +400,7 @@ void Index::build(InputArray _data, const IndexParams& params, flann_distance_t
         break;
 #endif
     default:
-        CV_Error(CV_StsBadArg, "Unknown/unsupported distance type");
+        CV_Error(Error::StsBadArg, "Unknown/unsupported distance type");
     }
 }
 
@@ -453,7 +453,7 @@ void Index::release()
             break;
 #endif
         default:
-            CV_Error(CV_StsBadArg, "Unknown/unsupported distance type");
+            CV_Error(Error::StsBadArg, "Unknown/unsupported distance type");
     }
     index = 0;
 }
@@ -585,7 +585,7 @@ void Index::knnSearch(InputArray _query, OutputArray _indices,
         break;
 #endif
     default:
-        CV_Error(CV_StsBadArg, "Unknown/unsupported distance type");
+        CV_Error(Error::StsBadArg, "Unknown/unsupported distance type");
     }
 }
 
@@ -599,7 +599,7 @@ int Index::radiusSearch(InputArray _query, OutputArray _indices,
     createIndicesDists( _indices, _dists, indices, dists, query.rows, maxResults, INT_MAX, dtype );
 
     if( algo == FLANN_INDEX_LSH )
-        CV_Error( CV_StsNotImplemented, "LSH index does not support radiusSearch operation" );
+        CV_Error( Error::StsNotImplemented, "LSH index does not support radiusSearch operation" );
 
     switch( distType )
     {
@@ -623,7 +623,7 @@ int Index::radiusSearch(InputArray _query, OutputArray _indices,
         return runRadiusSearch< ::cvflann::KL_Divergence<float> >(index, query, indices, dists, radius, params);
 #endif
     default:
-        CV_Error(CV_StsBadArg, "Unknown/unsupported distance type");
+        CV_Error(Error::StsBadArg, "Unknown/unsupported distance type");
     }
     return -1;
 }
@@ -654,11 +654,11 @@ template<typename Distance> void saveIndex(const Index* index0, const void* inde
     saveIndex_< ::cvflann::Index<Distance> >(index0, index, fout);
 }
 
-void Index::save(const std::string& filename) const
+void Index::save(const String& filename) const
 {
     FILE* fout = fopen(filename.c_str(), "wb");
     if (fout == NULL)
-        CV_Error_( CV_StsError, ("Can not open file %s for writing FLANN index\n", filename.c_str()) );
+        CV_Error_( Error::StsError, ("Can not open file %s for writing FLANN index\n", filename.c_str()) );
 
     switch( distType )
     {
@@ -691,7 +691,7 @@ void Index::save(const std::string& filename) const
     default:
         fclose(fout);
         fout = 0;
-        CV_Error(CV_StsBadArg, "Unknown/unsupported distance type");
+        CV_Error(Error::StsBadArg, "Unknown/unsupported distance type");
     }
     if( fout )
         fclose(fout);
@@ -720,7 +720,7 @@ bool loadIndex(Index* index0, void*& index, const Mat& data, FILE* fin, const Di
     return loadIndex_<Distance, ::cvflann::Index<Distance> >(index0, index, data, fin, dist);
 }
 
-bool Index::load(InputArray _data, const std::string& filename)
+bool Index::load(InputArray _data, const String& filename)
 {
     Mat data = _data.getMat();
     bool ok = true;
diff --git a/modules/flann/src/precomp.cpp b/modules/flann/src/precomp.cpp
deleted file mode 100644
index c149df18f9..0000000000
--- a/modules/flann/src/precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "precomp.hpp"
diff --git a/modules/flann/src/precomp.hpp b/modules/flann/src/precomp.hpp
index 0d2647932b..1c41542857 100644
--- a/modules/flann/src/precomp.hpp
+++ b/modules/flann/src/precomp.hpp
@@ -5,11 +5,8 @@
 #include <cstdarg>
 #include <sstream>
 
-#ifdef HAVE_CVCONFIG_H
-# include "cvconfig.h"
-#endif
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/utility.hpp"
 
 #include "opencv2/flann/miniflann.hpp"
 #include "opencv2/flann/dist.h"
@@ -23,5 +20,6 @@
 #include "opencv2/flann/all_indices.h"
 #include "opencv2/flann/flann_base.hpp"
 
-#endif
+#include "opencv2/core/private.hpp"
 
+#endif
diff --git a/modules/flann/test/test_lshtable_badarg.cpp b/modules/flann/test/test_lshtable_badarg.cpp
index 9e42335c2c..3b776668d1 100644
--- a/modules/flann/test/test_lshtable_badarg.cpp
+++ b/modules/flann/test/test_lshtable_badarg.cpp
@@ -74,13 +74,13 @@ void CV_LshTableBadArgTest::run( int /* start_from */ )
 
     int errors = 0;
     caller.key_size = 0;
-    errors += run_test_case(CV_StsBadArg, "key_size is zero", caller);
+    errors += run_test_case(Error::StsBadArg, "key_size is zero", caller);
 
     caller.key_size = static_cast<int>(sizeof(size_t) * CHAR_BIT);
-    errors += run_test_case(CV_StsBadArg, "key_size is too big", caller);
+    errors += run_test_case(Error::StsBadArg, "key_size is too big", caller);
 
     caller.key_size += cvtest::randInt(rng) % 100;
-    errors += run_test_case(CV_StsBadArg, "key_size is too big", caller);
+    errors += run_test_case(Error::StsBadArg, "key_size is too big", caller);
 
     if (errors != 0)
         ts->set_failed_test_info(cvtest::TS::FAIL_MISMATCH);
diff --git a/modules/flann/test/test_precomp.cpp b/modules/flann/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/flann/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/flann/test/test_precomp.hpp b/modules/flann/test/test_precomp.hpp
index b5cd174610..cbee8957fd 100644
--- a/modules/flann/test/test_precomp.hpp
+++ b/modules/flann/test/test_precomp.hpp
@@ -10,7 +10,6 @@
 #define __OPENCV_TEST_PRECOMP_HPP__
 
 #include "opencv2/ts.hpp"
-#include "opencv2/core.hpp"
 #include "opencv2/flann.hpp"
 #include <iostream>
 
diff --git a/modules/gpu/CMakeLists.txt b/modules/gpu/CMakeLists.txt
index 26f806fb89..146056839a 100644
--- a/modules/gpu/CMakeLists.txt
+++ b/modules/gpu/CMakeLists.txt
@@ -3,107 +3,10 @@ if(ANDROID OR IOS)
 endif()
 
 set(the_description "GPU-accelerated Computer Vision")
-ocv_add_module(gpu opencv_imgproc opencv_calib3d opencv_objdetect opencv_video opencv_nonfree opencv_photo opencv_legacy)
 
-ocv_module_include_directories("${CMAKE_CURRENT_SOURCE_DIR}/src/cuda")
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4100 /wd4324 /wd4512 /wd4515 -Wundef -Wmissing-declarations -Wshadow -Wunused-parameter)
 
-file(GLOB lib_hdrs               "include/opencv2/*.hpp"                       "include/opencv2/${name}/*.hpp"               "include/opencv2/${name}/*.h")
-file(GLOB lib_device_hdrs        "include/opencv2/${name}/device/*.hpp"        "include/opencv2/${name}/device/*.h")
-file(GLOB lib_device_hdrs_detail "include/opencv2/${name}/device/detail/*.hpp" "include/opencv2/${name}/device/detail/*.h")
-file(GLOB lib_int_hdrs           "src/*.hpp"      "src/*.h")
-file(GLOB lib_cuda_hdrs          "src/cuda/*.hpp" "src/cuda/*.h")
-file(GLOB lib_srcs               "src/*.cpp")
-file(GLOB lib_cuda               "src/cuda/*.cu*")
-
-source_group("Include"        FILES ${lib_hdrs})
-source_group("Src\\Host"      FILES ${lib_srcs} ${lib_int_hdrs})
-source_group("Src\\Cuda"      FILES ${lib_cuda} ${lib_cuda_hdrs})
-source_group("Device"         FILES ${lib_device_hdrs})
-source_group("Device\\Detail" FILES ${lib_device_hdrs_detail})
-
-if(HAVE_CUDA)
-  file(GLOB_RECURSE ncv_srcs "src/nvidia/*.cpp" "src/nvidia/*.h*")
-  file(GLOB_RECURSE ncv_cuda "src/nvidia/*.cu")
-  set(ncv_files ${ncv_srcs} ${ncv_cuda})
-
-  source_group("Src\\NVidia" FILES ${ncv_files})
-  ocv_include_directories("src/nvidia" "src/nvidia/core" "src/nvidia/NPP_staging" ${CUDA_INCLUDE_DIRS})
-  ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wmissing-declarations /wd4211 /wd4201 /wd4100 /wd4505 /wd4408)
-  string(REPLACE "-Wsign-promo" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
-  #set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler;/EHsc-;")
-
-  if(MSVC)
-    if(NOT ENABLE_NOISY_WARNINGS)
-      foreach(var CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_RELEASE CMAKE_CXX_FLAGS_DEBUG)
-        string(REPLACE "/W4" "/W3" ${var} "${${var}}")
-      endforeach()
-
-      set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler /wd4251)
-    endif()
-  endif()
-
-  ocv_cuda_compile(cuda_objs ${lib_cuda} ${ncv_cuda})
-
-  set(cuda_link_libs ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
-
-  if(WITH_NVCUVID)
-    set(cuda_link_libs ${cuda_link_libs} ${CUDA_nvcuvid_LIBRARY})
-  endif()
-
-  if(WIN32)
-    find_cuda_helper_libs(nvcuvenc)
-    set(cuda_link_libs ${cuda_link_libs} ${CUDA_nvcuvenc_LIBRARY})
-  endif()
-
-  if(WITH_FFMPEG)
-    set(cuda_link_libs ${cuda_link_libs} ${HIGHGUI_LIBRARIES})
-  endif()
-else()
-  set(lib_cuda "")
-  set(cuda_objs "")
-  set(cuda_link_libs "")
-  set(ncv_files "")
-endif()
-
-ocv_set_module_sources(
-  HEADERS ${lib_hdrs} ${lib_device_hdrs} ${lib_device_hdrs_detail}
-  SOURCES ${lib_int_hdrs} ${lib_cuda_hdrs} ${lib_srcs} ${lib_cuda} ${ncv_files} ${cuda_objs}
-  )
-
-ocv_create_module(${cuda_link_libs})
-
-if(HAVE_CUDA)
-  if(HAVE_CUFFT)
-    CUDA_ADD_CUFFT_TO_TARGET(${the_module})
-  endif()
-
-  if(HAVE_CUBLAS)
-    CUDA_ADD_CUBLAS_TO_TARGET(${the_module})
-  endif()
-
-  install(FILES src/nvidia/NPP_staging/NPP_staging.hpp  src/nvidia/core/NCV.hpp
-    DESTINATION ${OPENCV_INCLUDE_INSTALL_PATH}/opencv2/${name}
-    COMPONENT main)
-endif()
-
-ocv_add_precompiled_headers(${the_module})
-
-################################################################################################################
-################################      GPU Module Tests     #####################################################
-################################################################################################################
-file(GLOB test_srcs "test/*.cpp")
-file(GLOB test_hdrs "test/*.hpp" "test/*.h")
-
-set(nvidia "")
-if(HAVE_CUDA)
-  file(GLOB nvidia "test/nvidia/*.cpp" "test/nvidia/*.hpp" "test/nvidia/*.h")
-  set(nvidia FILES "Src\\\\\\\\NVidia" ${nvidia}) # 8 ugly backslashes :'(
-endif()
-
-ocv_add_accuracy_tests(FILES "Include" ${test_hdrs}
-                       FILES "Src" ${test_srcs}
-                       ${nvidia})
-ocv_add_perf_tests()
+ocv_define_module(gpu opencv_calib3d opencv_objdetect opencv_gpuarithm opencv_gpuwarping OPTIONAL opencv_gpulegacy)
 
 if(HAVE_CUDA)
   add_subdirectory(perf4au)
diff --git a/modules/gpu/app/nv_perf_test/CMakeLists.txt b/modules/gpu/app/nv_perf_test/CMakeLists.txt
deleted file mode 100644
index c13f5ef46b..0000000000
--- a/modules/gpu/app/nv_perf_test/CMakeLists.txt
+++ /dev/null
@@ -1,10 +0,0 @@
-cmake_minimum_required(VERSION 2.8.3)
-
-project(nv_perf_test)
-
-find_package(OpenCV REQUIRED)
-include_directories(${OpenCV_INCLUDE_DIR})
-
-add_executable(${PROJECT_NAME} main.cpp)
-
-target_link_libraries(${PROJECT_NAME} ${OpenCV_LIBS})
diff --git a/modules/gpu/app/nv_perf_test/im1_1280x800.jpg b/modules/gpu/app/nv_perf_test/im1_1280x800.jpg
deleted file mode 100644
index bdbbd4aee9..0000000000
Binary files a/modules/gpu/app/nv_perf_test/im1_1280x800.jpg and /dev/null differ
diff --git a/modules/gpu/app/nv_perf_test/im2_1280x800.jpg b/modules/gpu/app/nv_perf_test/im2_1280x800.jpg
deleted file mode 100644
index ae49640a95..0000000000
Binary files a/modules/gpu/app/nv_perf_test/im2_1280x800.jpg and /dev/null differ
diff --git a/modules/gpu/app/nv_perf_test/main.cpp b/modules/gpu/app/nv_perf_test/main.cpp
deleted file mode 100644
index 4bdf387edf..0000000000
--- a/modules/gpu/app/nv_perf_test/main.cpp
+++ /dev/null
@@ -1,487 +0,0 @@
-#include <cstdio>
-#define HAVE_CUDA 1
-#include <opencv2/core.hpp>
-#include <opencv2/gpu.hpp>
-#include <opencv2/highgui.hpp>
-#include <opencv2/video.hpp>
-#include <opencv2/legacy.hpp>
-#include <opencv2/ts.hpp>
-
-static void printOsInfo()
-{
-#if defined _WIN32
-#   if defined _WIN64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#elif defined linux
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#elif defined __APPLE__
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#endif
-}
-
-static void printCudaInfo()
-{
-    const int deviceCount = cv::gpu::getCudaEnabledDeviceCount();
-
-    printf("[----------]\n"); fflush(stdout);
-    printf("[ GPU INFO ] \tCUDA device count:: %d.\n", deviceCount); fflush(stdout);
-    printf("[----------]\n"); fflush(stdout);
-
-    for (int i = 0; i < deviceCount; ++i)
-    {
-        cv::gpu::DeviceInfo info(i);
-
-        printf("[----------]\n"); fflush(stdout);
-        printf("[ DEVICE   ] \t# %d %s.\n", i, info.name().c_str()); fflush(stdout);
-        printf("[          ] \tCompute capability: %d.%d\n", info.majorVersion(), info.minorVersion()); fflush(stdout);
-        printf("[          ] \tMulti Processor Count:  %d\n", info.multiProcessorCount()); fflush(stdout);
-        printf("[          ] \tTotal memory: %d Mb\n", static_cast<int>(static_cast<int>(info.totalMemory() / 1024.0) / 1024.0)); fflush(stdout);
-        printf("[          ] \tFree  memory: %d Mb\n", static_cast<int>(static_cast<int>(info.freeMemory()  / 1024.0) / 1024.0)); fflush(stdout);
-        if (!info.isCompatible())
-            printf("[ GPU INFO ] \tThis device is NOT compatible with current GPU module build\n");
-        printf("[----------]\n"); fflush(stdout);
-    }
-}
-
-int main(int argc, char* argv[])
-{
-    printOsInfo();
-    printCudaInfo();
-
-    perf::Regression::Init("nv_perf_test");
-    perf::TestBase::Init(argc, argv);
-    testing::InitGoogleTest(&argc, argv);
-
-    return RUN_ALL_TESTS();
-}
-
-#define DEF_PARAM_TEST(name, ...) typedef ::perf::TestBaseWithParam< std::tr1::tuple< __VA_ARGS__ > > name
-#define DEF_PARAM_TEST_1(name, param_type) typedef ::perf::TestBaseWithParam< param_type > name
-
-//////////////////////////////////////////////////////////
-// HoughLinesP
-
-DEF_PARAM_TEST_1(Image, std::string);
-
-GPU_PERF_TEST_P(Image, HoughLinesP, testing::Values(std::string("im1_1280x800.jpg")))
-{
-    declare.time(30.0);
-
-    std::string fileName = GetParam();
-
-    const float rho = 1.f;
-    const float theta = 1.f;
-    const int threshold = 40;
-    const int minLineLenght = 20;
-    const int maxLineGap = 5;
-
-    cv::Mat image = cv::imread(fileName, cv::IMREAD_GRAYSCALE);
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_image(image);
-        cv::gpu::GpuMat d_lines;
-        cv::gpu::HoughLinesBuf d_buf;
-
-        cv::gpu::HoughLinesP(d_image, d_lines, d_buf, rho, theta, minLineLenght, maxLineGap);
-
-        TEST_CYCLE()
-        {
-            cv::gpu::HoughLinesP(d_image, d_lines, d_buf, rho, theta, minLineLenght, maxLineGap);
-        }
-    }
-    else
-    {
-        cv::Mat mask;
-        cv::Canny(image, mask, 50, 100);
-
-        std::vector<cv::Vec4i> lines;
-        cv::HoughLinesP(mask, lines, rho, theta, threshold, minLineLenght, maxLineGap);
-
-        TEST_CYCLE()
-        {
-            cv::HoughLinesP(mask, lines, rho, theta, threshold, minLineLenght, maxLineGap);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
-
-//////////////////////////////////////////////////////////
-// GoodFeaturesToTrack
-
-DEF_PARAM_TEST(Image_Depth, std::string, perf::MatDepth);
-
-GPU_PERF_TEST_P(Image_Depth, GoodFeaturesToTrack,
-                testing::Combine(
-                testing::Values(std::string("im1_1280x800.jpg")),
-                testing::Values(CV_8U, CV_16U)
-                ))
-{
-    declare.time(60);
-
-    const std::string fileName = std::tr1::get<0>(GetParam());
-    const int depth = std::tr1::get<1>(GetParam());
-
-    const int maxCorners = 5000;
-    const double qualityLevel = 0.05;
-    const int minDistance = 5;
-    const int blockSize = 3;
-    const bool useHarrisDetector = true;
-    const double k = 0.05;
-
-    cv::Mat src = cv::imread(fileName, cv::IMREAD_GRAYSCALE);
-    if (src.empty())
-        FAIL() << "Unable to load source image [" << fileName << "]";
-
-    if (depth != CV_8U)
-        src.convertTo(src, depth);
-
-    cv::Mat mask(src.size(), CV_8UC1, cv::Scalar::all(1));
-    mask(cv::Rect(0, 0, 100, 100)).setTo(cv::Scalar::all(0));
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GoodFeaturesToTrackDetector_GPU d_detector(maxCorners, qualityLevel, minDistance, blockSize, useHarrisDetector, k);
-
-        cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_mask(mask);
-        cv::gpu::GpuMat d_pts;
-
-        d_detector(d_src, d_pts, d_mask);
-
-        TEST_CYCLE()
-        {
-            d_detector(d_src, d_pts, d_mask);
-        }
-    }
-    else
-    {
-        if (depth != CV_8U)
-            FAIL() << "Unsupported depth";
-
-        cv::Mat pts;
-
-        cv::goodFeaturesToTrack(src, pts, maxCorners, qualityLevel, minDistance, mask, blockSize, useHarrisDetector, k);
-
-        TEST_CYCLE()
-        {
-            cv::goodFeaturesToTrack(src, pts, maxCorners, qualityLevel, minDistance, mask, blockSize, useHarrisDetector, k);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
-
-//////////////////////////////////////////////////////////
-// OpticalFlowPyrLKSparse
-
-typedef std::pair<std::string, std::string> string_pair;
-
-DEF_PARAM_TEST(ImagePair_Depth_GraySource, string_pair, perf::MatDepth, bool);
-
-GPU_PERF_TEST_P(ImagePair_Depth_GraySource, OpticalFlowPyrLKSparse,
-                testing::Combine(
-                    testing::Values(string_pair("im1_1280x800.jpg", "im2_1280x800.jpg")),
-                    testing::Values(CV_8U, CV_16U),
-                    testing::Bool()
-                    ))
-{
-    declare.time(60);
-
-    const string_pair fileNames = std::tr1::get<0>(GetParam());
-    const int depth = std::tr1::get<1>(GetParam());
-    const bool graySource = std::tr1::get<2>(GetParam());
-
-    // PyrLK params
-    const cv::Size winSize(15, 15);
-    const int maxLevel = 5;
-    const cv::TermCriteria criteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, 30, 0.01);
-
-    // GoodFeaturesToTrack params
-    const int maxCorners = 5000;
-    const double qualityLevel = 0.05;
-    const int minDistance = 5;
-    const int blockSize = 3;
-    const bool useHarrisDetector = true;
-    const double k = 0.05;
-
-    cv::Mat src1 = cv::imread(fileNames.first, graySource ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    if (src1.empty())
-        FAIL() << "Unable to load source image [" << fileNames.first << "]";
-
-    cv::Mat src2 = cv::imread(fileNames.second, graySource ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    if (src2.empty())
-        FAIL() << "Unable to load source image [" << fileNames.second << "]";
-
-    cv::Mat gray_src;
-    if (graySource)
-        gray_src = src1;
-    else
-        cv::cvtColor(src1, gray_src, cv::COLOR_BGR2GRAY);
-
-    cv::Mat pts;
-    cv::goodFeaturesToTrack(gray_src, pts, maxCorners, qualityLevel, minDistance, cv::noArray(), blockSize, useHarrisDetector, k);
-
-    if (depth != CV_8U)
-    {
-        src1.convertTo(src1, depth);
-        src2.convertTo(src2, depth);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_src1(src1);
-        cv::gpu::GpuMat d_src2(src2);
-        cv::gpu::GpuMat d_pts(pts.reshape(2, 1));
-        cv::gpu::GpuMat d_nextPts;
-        cv::gpu::GpuMat d_status;
-
-        cv::gpu::PyrLKOpticalFlow d_pyrLK;
-        d_pyrLK.winSize = winSize;
-        d_pyrLK.maxLevel = maxLevel;
-        d_pyrLK.iters = criteria.maxCount;
-        d_pyrLK.useInitialFlow = false;
-
-        d_pyrLK.sparse(d_src1, d_src2, d_pts, d_nextPts, d_status);
-
-        TEST_CYCLE()
-        {
-            d_pyrLK.sparse(d_src1, d_src2, d_pts, d_nextPts, d_status);
-        }
-    }
-    else
-    {
-        if (depth != CV_8U)
-            FAIL() << "Unsupported depth";
-
-        cv::Mat nextPts;
-        cv::Mat status;
-
-        cv::calcOpticalFlowPyrLK(src1, src2, pts, nextPts, status, cv::noArray(), winSize, maxLevel, criteria);
-
-        TEST_CYCLE()
-        {
-            cv::calcOpticalFlowPyrLK(src1, src2, pts, nextPts, status, cv::noArray(), winSize, maxLevel, criteria);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
-
-//////////////////////////////////////////////////////////
-// OpticalFlowFarneback
-
-DEF_PARAM_TEST(ImagePair_Depth, string_pair, perf::MatDepth);
-
-GPU_PERF_TEST_P(ImagePair_Depth, OpticalFlowFarneback,
-                testing::Combine(
-                    testing::Values(string_pair("im1_1280x800.jpg", "im2_1280x800.jpg")),
-                    testing::Values(CV_8U, CV_16U)
-                    ))
-{
-    declare.time(500);
-
-    const string_pair fileNames = std::tr1::get<0>(GetParam());
-    const int depth = std::tr1::get<1>(GetParam());
-
-    const double pyrScale = 0.5;
-    const int numLevels = 6;
-    const int winSize = 7;
-    const int numIters = 15;
-    const int polyN = 7;
-    const double polySigma = 1.5;
-    const int flags = cv::OPTFLOW_USE_INITIAL_FLOW;
-
-    cv::Mat src1 = cv::imread(fileNames.first, cv::IMREAD_GRAYSCALE);
-    if (src1.empty())
-        FAIL() << "Unable to load source image [" << fileNames.first << "]";
-
-    cv::Mat src2 = cv::imread(fileNames.second, cv::IMREAD_GRAYSCALE);
-    if (src2.empty())
-        FAIL() << "Unable to load source image [" << fileNames.second << "]";
-
-    if (depth != CV_8U)
-    {
-        src1.convertTo(src1, depth);
-        src2.convertTo(src2, depth);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_src1(src1);
-        cv::gpu::GpuMat d_src2(src2);
-        cv::gpu::GpuMat d_u(src1.size(), CV_32FC1, cv::Scalar::all(0));
-        cv::gpu::GpuMat d_v(src1.size(), CV_32FC1, cv::Scalar::all(0));
-
-        cv::gpu::FarnebackOpticalFlow d_farneback;
-        d_farneback.pyrScale = pyrScale;
-        d_farneback.numLevels = numLevels;
-        d_farneback.winSize = winSize;
-        d_farneback.numIters = numIters;
-        d_farneback.polyN = polyN;
-        d_farneback.polySigma = polySigma;
-        d_farneback.flags = flags;
-
-        d_farneback(d_src1, d_src2, d_u, d_v);
-
-        TEST_CYCLE_N(10)
-        {
-            d_farneback(d_src1, d_src2, d_u, d_v);
-        }
-    }
-    else
-    {
-        if (depth != CV_8U)
-            FAIL() << "Unsupported depth";
-
-        cv::Mat flow(src1.size(), CV_32FC2, cv::Scalar::all(0));
-
-        cv::calcOpticalFlowFarneback(src1, src2, flow, pyrScale, numLevels, winSize, numIters, polyN, polySigma, flags);
-
-        TEST_CYCLE_N(10)
-        {
-            cv::calcOpticalFlowFarneback(src1, src2, flow, pyrScale, numLevels, winSize, numIters, polyN, polySigma, flags);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
-
-//////////////////////////////////////////////////////////
-// OpticalFlowBM
-
-void calcOpticalFlowBM(const cv::Mat& prev, const cv::Mat& curr,
-                       cv::Size bSize, cv::Size shiftSize, cv::Size maxRange, int usePrevious,
-                       cv::Mat& velx, cv::Mat& vely)
-{
-    cv::Size sz((curr.cols - bSize.width + shiftSize.width)/shiftSize.width, (curr.rows - bSize.height + shiftSize.height)/shiftSize.height);
-
-    velx.create(sz, CV_32FC1);
-    vely.create(sz, CV_32FC1);
-
-    CvMat cvprev = prev;
-    CvMat cvcurr = curr;
-
-    CvMat cvvelx = velx;
-    CvMat cvvely = vely;
-
-    cvCalcOpticalFlowBM(&cvprev, &cvcurr, bSize, shiftSize, maxRange, usePrevious, &cvvelx, &cvvely);
-}
-
-DEF_PARAM_TEST(ImagePair_BlockSize_ShiftSize_MaxRange, string_pair, cv::Size, cv::Size, cv::Size);
-
-GPU_PERF_TEST_P(ImagePair_BlockSize_ShiftSize_MaxRange, OpticalFlowBM,
-                testing::Combine(
-                    testing::Values(string_pair("im1_1280x800.jpg", "im2_1280x800.jpg")),
-                    testing::Values(cv::Size(16, 16)),
-                    testing::Values(cv::Size(2, 2)),
-                    testing::Values(cv::Size(16, 16))
-                    ))
-{
-    declare.time(3000);
-
-    const string_pair fileNames = std::tr1::get<0>(GetParam());
-    const cv::Size block_size = std::tr1::get<1>(GetParam());
-    const cv::Size shift_size = std::tr1::get<2>(GetParam());
-    const cv::Size max_range = std::tr1::get<3>(GetParam());
-
-    cv::Mat src1 = cv::imread(fileNames.first, cv::IMREAD_GRAYSCALE);
-    if (src1.empty())
-        FAIL() << "Unable to load source image [" << fileNames.first << "]";
-
-    cv::Mat src2 = cv::imread(fileNames.second, cv::IMREAD_GRAYSCALE);
-    if (src2.empty())
-        FAIL() << "Unable to load source image [" << fileNames.second << "]";
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_src1(src1);
-        cv::gpu::GpuMat d_src2(src2);
-        cv::gpu::GpuMat d_velx, d_vely, buf;
-
-        cv::gpu::calcOpticalFlowBM(d_src1, d_src2, block_size, shift_size, max_range, false, d_velx, d_vely, buf);
-
-        TEST_CYCLE_N(10)
-        {
-            cv::gpu::calcOpticalFlowBM(d_src1, d_src2, block_size, shift_size, max_range, false, d_velx, d_vely, buf);
-        }
-    }
-    else
-    {
-        cv::Mat velx, vely;
-
-        calcOpticalFlowBM(src1, src2, block_size, shift_size, max_range, false, velx, vely);
-
-        TEST_CYCLE_N(10)
-        {
-            calcOpticalFlowBM(src1, src2, block_size, shift_size, max_range, false, velx, vely);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
-
-GPU_PERF_TEST_P(ImagePair_BlockSize_ShiftSize_MaxRange, FastOpticalFlowBM,
-                testing::Combine(
-                    testing::Values(string_pair("im1_1280x800.jpg", "im2_1280x800.jpg")),
-                    testing::Values(cv::Size(16, 16)),
-                    testing::Values(cv::Size(1, 1)),
-                    testing::Values(cv::Size(16, 16))
-                    ))
-{
-    declare.time(3000);
-
-    const string_pair fileNames = std::tr1::get<0>(GetParam());
-    const cv::Size block_size = std::tr1::get<1>(GetParam());
-    const cv::Size shift_size = std::tr1::get<2>(GetParam());
-    const cv::Size max_range = std::tr1::get<3>(GetParam());
-
-    cv::Mat src1 = cv::imread(fileNames.first, cv::IMREAD_GRAYSCALE);
-    if (src1.empty())
-        FAIL() << "Unable to load source image [" << fileNames.first << "]";
-
-    cv::Mat src2 = cv::imread(fileNames.second, cv::IMREAD_GRAYSCALE);
-    if (src2.empty())
-        FAIL() << "Unable to load source image [" << fileNames.second << "]";
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_src1(src1);
-        cv::gpu::GpuMat d_src2(src2);
-        cv::gpu::GpuMat d_velx, d_vely;
-
-        cv::gpu::FastOpticalFlowBM fastBM;
-
-        fastBM(d_src1, d_src2, d_velx, d_vely, max_range.width, block_size.width);
-
-        TEST_CYCLE_N(10)
-        {
-            fastBM(d_src1, d_src2, d_velx, d_vely, max_range.width, block_size.width);
-        }
-    }
-    else
-    {
-        cv::Mat velx, vely;
-
-        calcOpticalFlowBM(src1, src2, block_size, shift_size, max_range, false, velx, vely);
-
-        TEST_CYCLE_N(10)
-        {
-            calcOpticalFlowBM(src1, src2, block_size, shift_size, max_range, false, velx, vely);
-        }
-    }
-
-    SANITY_CHECK(0);
-}
diff --git a/modules/gpu/doc/calib3d.rst b/modules/gpu/doc/calib3d.rst
new file mode 100644
index 0000000000..faa6c0fec2
--- /dev/null
+++ b/modules/gpu/doc/calib3d.rst
@@ -0,0 +1,36 @@
+Camera Calibration and 3D Reconstruction
+========================================
+
+.. highlight:: cpp
+
+
+
+gpu::solvePnPRansac
+-------------------
+Finds the object pose from 3D-2D point correspondences.
+
+.. ocv:function:: void gpu::solvePnPRansac(const Mat& object, const Mat& image, const Mat& camera_mat, const Mat& dist_coef, Mat& rvec, Mat& tvec, bool use_extrinsic_guess=false, int num_iters=100, float max_dist=8.0, int min_inlier_count=100, vector<int>* inliers=NULL)
+
+    :param object: Single-row matrix of object points.
+
+    :param image: Single-row matrix of image points.
+
+    :param camera_mat: 3x3 matrix of intrinsic camera parameters.
+
+    :param dist_coef: Distortion coefficients. See :ocv:func:`undistortPoints` for details.
+
+    :param rvec: Output 3D rotation vector.
+
+    :param tvec: Output 3D translation vector.
+
+    :param use_extrinsic_guess: Flag to indicate that the function must use ``rvec`` and ``tvec`` as an initial transformation guess. It is not supported for now.
+
+    :param num_iters: Maximum number of RANSAC iterations.
+
+    :param max_dist: Euclidean distance threshold to detect whether point is inlier or not.
+
+    :param min_inlier_count: Flag to indicate that the function must stop if greater or equal number of inliers is achieved. It is not supported for now.
+
+    :param inliers: Output vector of inlier indices.
+
+.. seealso:: :ocv:func:`solvePnPRansac`
diff --git a/modules/gpu/doc/camera_calibration_and_3d_reconstruction.rst b/modules/gpu/doc/camera_calibration_and_3d_reconstruction.rst
deleted file mode 100644
index 587c253d2a..0000000000
--- a/modules/gpu/doc/camera_calibration_and_3d_reconstruction.rst
+++ /dev/null
@@ -1,499 +0,0 @@
-Camera Calibration and 3D Reconstruction
-========================================
-
-.. highlight:: cpp
-
-
-
-gpu::StereoBM_GPU
------------------
-.. ocv:class:: gpu::StereoBM_GPU
-
-Class computing stereo correspondence (disparity map) using the block matching algorithm. ::
-
-    class StereoBM_GPU
-    {
-    public:
-        enum { BASIC_PRESET = 0, PREFILTER_XSOBEL = 1 };
-
-        enum { DEFAULT_NDISP = 64, DEFAULT_WINSZ = 19 };
-
-        StereoBM_GPU();
-        StereoBM_GPU(int preset, int ndisparities = DEFAULT_NDISP,
-                     int winSize = DEFAULT_WINSZ);
-
-        void operator() (const GpuMat& left, const GpuMat& right,
-                         GpuMat& disparity, Stream& stream = Stream::Null());
-
-        static bool checkIfGpuCallReasonable();
-
-        int preset;
-        int ndisp;
-        int winSize;
-
-        float avergeTexThreshold;
-
-        ...
-    };
-
-
-The class also performs pre- and post-filtering steps: Sobel pre-filtering (if ``PREFILTER_XSOBEL`` flag is set) and low textureness filtering (if ``averageTexThreshols > 0`` ). If ``avergeTexThreshold = 0`` , low textureness filtering is disabled. Otherwise, the disparity is set to 0 in each point ``(x, y)`` , where for the left image
-
-.. math::
-    \sum HorizontalGradiensInWindow(x, y, winSize) < (winSize \cdot winSize) \cdot avergeTexThreshold
-
-This means that the input left image is low textured.
-
-
-
-gpu::StereoBM_GPU::StereoBM_GPU
------------------------------------
-Enables :ocv:class:`gpu::StereoBM_GPU` constructors.
-
-.. ocv:function:: gpu::StereoBM_GPU::StereoBM_GPU()
-
-.. ocv:function:: gpu::StereoBM_GPU::StereoBM_GPU(int preset, int ndisparities = DEFAULT_NDISP, int winSize = DEFAULT_WINSZ)
-
-    :param preset: Parameter presetting:
-
-        * **BASIC_PRESET** Basic mode without pre-processing.
-
-        * **PREFILTER_XSOBEL** Sobel pre-filtering mode.
-
-    :param ndisparities: Number of disparities. It must be a multiple of 8 and less or equal to 256.
-
-    :param winSize: Block size.
-
-
-
-gpu::StereoBM_GPU::operator ()
-----------------------------------
-Enables the stereo correspondence operator that finds the disparity for the specified rectified stereo pair.
-
-.. ocv:function:: void gpu::StereoBM_GPU::operator ()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null())
-
-    :param left: Left image. Only  ``CV_8UC1``  type is supported.
-
-    :param right: Right image with the same size and the same type as the left one.
-
-    :param disparity: Output disparity map. It is a  ``CV_8UC1``  image with the same size as the input images.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::StereoBM_GPU::checkIfGpuCallReasonable
------------------------------------------------
-Uses a heuristic method to estimate whether the current GPU is faster than the CPU in this algorithm. It queries the currently active device.
-
-.. ocv:function:: bool gpu::StereoBM_GPU::checkIfGpuCallReasonable()
-
-
-
-gpu::StereoBeliefPropagation
-----------------------------
-.. ocv:class:: gpu::StereoBeliefPropagation
-
-Class computing stereo correspondence using the belief propagation algorithm. ::
-
-    class StereoBeliefPropagation
-    {
-    public:
-        enum { DEFAULT_NDISP  = 64 };
-        enum { DEFAULT_ITERS  = 5  };
-        enum { DEFAULT_LEVELS = 5  };
-
-        static void estimateRecommendedParams(int width, int height,
-            int& ndisp, int& iters, int& levels);
-
-        explicit StereoBeliefPropagation(int ndisp = DEFAULT_NDISP,
-            int iters  = DEFAULT_ITERS,
-            int levels = DEFAULT_LEVELS,
-            int msg_type = CV_32F);
-        StereoBeliefPropagation(int ndisp, int iters, int levels,
-            float max_data_term, float data_weight,
-            float max_disc_term, float disc_single_jump,
-            int msg_type = CV_32F);
-
-        void operator()(const GpuMat& left, const GpuMat& right,
-                        GpuMat& disparity, Stream& stream = Stream::Null());
-        void operator()(const GpuMat& data, GpuMat& disparity, Stream& stream = Stream::Null());
-
-        int ndisp;
-
-        int iters;
-        int levels;
-
-        float max_data_term;
-        float data_weight;
-        float max_disc_term;
-        float disc_single_jump;
-
-        int msg_type;
-
-        ...
-    };
-
-The class implements algorithm described in [Felzenszwalb2006]_ . It can compute own data cost (using a truncated linear model) or use a user-provided data cost.
-
-.. note::
-
-    ``StereoBeliefPropagation`` requires a lot of memory for message storage:
-
-    .. math::
-
-        width \_ step  \cdot height  \cdot ndisp  \cdot 4  \cdot (1 + 0.25)
-
-    and for data cost storage:
-
-    .. math::
-
-        width\_step \cdot height \cdot ndisp \cdot (1 + 0.25 + 0.0625 +  \dotsm + \frac{1}{4^{levels}})
-
-    ``width_step`` is the number of bytes in a line including padding.
-
-
-
-gpu::StereoBeliefPropagation::StereoBeliefPropagation
----------------------------------------------------------
-Enables the :ocv:class:`gpu::StereoBeliefPropagation` constructors.
-
-.. ocv:function:: gpu::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp = DEFAULT_NDISP, int iters = DEFAULT_ITERS, int levels = DEFAULT_LEVELS, int msg_type = CV_32F)
-
-.. ocv:function:: gpu::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp, int iters, int levels, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump, int msg_type = CV_32F)
-
-    :param ndisp: Number of disparities.
-
-    :param iters: Number of BP iterations on each level.
-
-    :param levels: Number of levels.
-
-    :param max_data_term: Threshold for data cost truncation.
-
-    :param data_weight: Data weight.
-
-    :param max_disc_term: Threshold for discontinuity truncation.
-
-    :param disc_single_jump: Discontinuity single jump.
-
-    :param msg_type: Type for messages.  ``CV_16SC1``  and  ``CV_32FC1`` types are supported.
-
-``StereoBeliefPropagation`` uses a truncated linear model for the data cost and discontinuity terms:
-
-.. math::
-
-    DataCost = data \_ weight  \cdot \min ( \lvert Img_Left(x,y)-Img_Right(x-d,y)  \rvert , max \_ data \_ term)
-
-.. math::
-
-    DiscTerm =  \min (disc \_ single \_ jump  \cdot \lvert f_1-f_2  \rvert , max \_ disc \_ term)
-
-For more details, see [Felzenszwalb2006]_.
-
-By default, :ocv:class:`gpu::StereoBeliefPropagation` uses floating-point arithmetics and the ``CV_32FC1`` type for messages. But it can also use fixed-point arithmetics and the ``CV_16SC1`` message type for better performance. To avoid an overflow in this case, the parameters must satisfy the following requirement:
-
-.. math::
-
-    10  \cdot 2^{levels-1}  \cdot max \_ data \_ term < SHRT \_ MAX
-
-
-
-gpu::StereoBeliefPropagation::estimateRecommendedParams
------------------------------------------------------------
-Uses a heuristic method to compute the recommended parameters ( ``ndisp``, ``iters`` and ``levels`` ) for the specified image size ( ``width`` and ``height`` ).
-
-.. ocv:function:: void gpu::StereoBeliefPropagation::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels)
-
-
-
-gpu::StereoBeliefPropagation::operator ()
----------------------------------------------
-Enables the stereo correspondence operator that finds the disparity for the specified rectified stereo pair or data cost.
-
-.. ocv:function:: void gpu::StereoBeliefPropagation::operator ()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::StereoBeliefPropagation::operator ()(const GpuMat& data, GpuMat& disparity, Stream& stream = Stream::Null())
-
-    :param left: Left image. ``CV_8UC1`` , ``CV_8UC3``  and  ``CV_8UC4``  types are supported.
-
-    :param right: Right image with the same size and the same type as the left one.
-
-    :param data: User-specified data cost, a matrix of ``msg_type`` type and ``Size(<image columns>*ndisp, <image rows>)`` size.
-
-    :param disparity: Output disparity map. If  ``disparity``  is empty, the output type is  ``CV_16SC1`` . Otherwise, the type is retained.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::StereoConstantSpaceBP
---------------------------
-.. ocv:class:: gpu::StereoConstantSpaceBP
-
-Class computing stereo correspondence using the constant space belief propagation algorithm. ::
-
-    class StereoConstantSpaceBP
-    {
-    public:
-        enum { DEFAULT_NDISP    = 128 };
-        enum { DEFAULT_ITERS    = 8   };
-        enum { DEFAULT_LEVELS   = 4   };
-        enum { DEFAULT_NR_PLANE = 4   };
-
-        static void estimateRecommendedParams(int width, int height,
-            int& ndisp, int& iters, int& levels, int& nr_plane);
-
-        explicit StereoConstantSpaceBP(int ndisp = DEFAULT_NDISP,
-            int iters    = DEFAULT_ITERS,
-            int levels   = DEFAULT_LEVELS,
-            int nr_plane = DEFAULT_NR_PLANE,
-            int msg_type = CV_32F);
-        StereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane,
-            float max_data_term, float data_weight,
-            float max_disc_term, float disc_single_jump,
-            int min_disp_th = 0,
-            int msg_type = CV_32F);
-
-        void operator()(const GpuMat& left, const GpuMat& right,
-                        GpuMat& disparity, Stream& stream = Stream::Null());
-
-        int ndisp;
-
-        int iters;
-        int levels;
-
-        int nr_plane;
-
-        float max_data_term;
-        float data_weight;
-        float max_disc_term;
-        float disc_single_jump;
-
-        int min_disp_th;
-
-        int msg_type;
-
-        bool use_local_init_data_cost;
-
-        ...
-    };
-
-
-The class implements algorithm described in [Yang2010]_. ``StereoConstantSpaceBP`` supports both local minimum and global minimum data cost initialization algorithms. For more details, see the paper mentioned above. By default, a local algorithm is used. To enable a global algorithm, set ``use_local_init_data_cost`` to ``false`` .
-
-
-
-gpu::StereoConstantSpaceBP::StereoConstantSpaceBP
------------------------------------------------------
-Enables the :ocv:class:`gpu::StereoConstantSpaceBP` constructors.
-
-.. ocv:function:: gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp = DEFAULT_NDISP, int iters = DEFAULT_ITERS, int levels = DEFAULT_LEVELS, int nr_plane = DEFAULT_NR_PLANE, int msg_type = CV_32F)
-
-.. ocv:function:: gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump, int min_disp_th = 0, int msg_type = CV_32F)
-
-    :param ndisp: Number of disparities.
-
-    :param iters: Number of BP iterations on each level.
-
-    :param levels: Number of levels.
-
-    :param nr_plane: Number of disparity levels on the first level.
-
-    :param max_data_term: Truncation of data cost.
-
-    :param data_weight: Data weight.
-
-    :param max_disc_term: Truncation of discontinuity.
-
-    :param disc_single_jump: Discontinuity single jump.
-
-    :param min_disp_th: Minimal disparity threshold.
-
-    :param msg_type: Type for messages.  ``CV_16SC1``  and  ``CV_32FC1`` types are supported.
-
-``StereoConstantSpaceBP`` uses a truncated linear model for the data cost and discontinuity terms:
-
-.. math::
-
-    DataCost = data \_ weight  \cdot \min ( \lvert I_2-I_1  \rvert , max \_ data \_ term)
-
-.. math::
-
-    DiscTerm =  \min (disc \_ single \_ jump  \cdot \lvert f_1-f_2  \rvert , max \_ disc \_ term)
-
-For more details, see [Yang2010]_.
-
-By default, ``StereoConstantSpaceBP`` uses floating-point arithmetics and the ``CV_32FC1`` type for messages. But it can also use fixed-point arithmetics and the ``CV_16SC1`` message type for better performance. To avoid an overflow in this case, the parameters must satisfy the following requirement:
-
-.. math::
-
-    10  \cdot 2^{levels-1}  \cdot max \_ data \_ term < SHRT \_ MAX
-
-
-
-gpu::StereoConstantSpaceBP::estimateRecommendedParams
----------------------------------------------------------
-Uses a heuristic method to compute parameters (ndisp, iters, levelsand nrplane) for the specified image size (widthand height).
-
-.. ocv:function:: void gpu::StereoConstantSpaceBP::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane)
-
-
-
-gpu::StereoConstantSpaceBP::operator ()
--------------------------------------------
-Enables the stereo correspondence operator that finds the disparity for the specified rectified stereo pair.
-
-.. ocv:function:: void gpu::StereoConstantSpaceBP::operator ()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null())
-
-    :param left: Left image. ``CV_8UC1`` , ``CV_8UC3``  and  ``CV_8UC4``  types are supported.
-
-    :param right: Right image with the same size and the same type as the left one.
-
-    :param disparity: Output disparity map. If  ``disparity``  is empty, the output type is  ``CV_16SC1`` . Otherwise, the output type is  ``disparity.type()`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::DisparityBilateralFilter
------------------------------
-.. ocv:class:: gpu::DisparityBilateralFilter
-
-Class refining a disparity map using joint bilateral filtering. ::
-
-    class CV_EXPORTS DisparityBilateralFilter
-    {
-    public:
-        enum { DEFAULT_NDISP  = 64 };
-        enum { DEFAULT_RADIUS = 3 };
-        enum { DEFAULT_ITERS  = 1 };
-
-        explicit DisparityBilateralFilter(int ndisp = DEFAULT_NDISP,
-            int radius = DEFAULT_RADIUS, int iters = DEFAULT_ITERS);
-
-        DisparityBilateralFilter(int ndisp, int radius, int iters,
-            float edge_threshold, float max_disc_threshold,
-            float sigma_range);
-
-        void operator()(const GpuMat& disparity, const GpuMat& image,
-                        GpuMat& dst, Stream& stream = Stream::Null());
-
-        ...
-    };
-
-
-The class implements [Yang2010]_ algorithm.
-
-
-
-gpu::DisparityBilateralFilter::DisparityBilateralFilter
------------------------------------------------------------
-Enables the :ocv:class:`gpu::DisparityBilateralFilter` constructors.
-
-.. ocv:function:: gpu::DisparityBilateralFilter::DisparityBilateralFilter(int ndisp = DEFAULT_NDISP, int radius = DEFAULT_RADIUS, int iters = DEFAULT_ITERS)
-
-.. ocv:function:: gpu::DisparityBilateralFilter::DisparityBilateralFilter(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold, float sigma_range)
-
-    :param ndisp: Number of disparities.
-
-    :param radius: Filter radius.
-
-    :param iters: Number of iterations.
-
-    :param edge_threshold: Threshold for edges.
-
-    :param max_disc_threshold: Constant to reject outliers.
-
-    :param sigma_range: Filter range.
-
-
-
-gpu::DisparityBilateralFilter::operator ()
-----------------------------------------------
-Refines a disparity map using joint bilateral filtering.
-
-.. ocv:function:: void gpu::DisparityBilateralFilter::operator ()(const GpuMat& disparity, const GpuMat& image, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param disparity: Input disparity map.  ``CV_8UC1``  and  ``CV_16SC1``  types are supported.
-
-    :param image: Input image. ``CV_8UC1``  and  ``CV_8UC3``  types are supported.
-
-    :param dst: Destination disparity map. It has the same size and type as  ``disparity`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::drawColorDisp
-----------------------
-Colors a disparity image.
-
-.. ocv:function:: void gpu::drawColorDisp(const GpuMat& src_disp, GpuMat& dst_disp, int ndisp, Stream& stream = Stream::Null())
-
-    :param src_disp: Source disparity image.  ``CV_8UC1``  and  ``CV_16SC1``  types are supported.
-
-    :param dst_disp: Output disparity image. It has the same size as  ``src_disp`` . The  type is ``CV_8UC4``  in  ``BGRA``  format (alpha = 255).
-
-    :param ndisp: Number of disparities.
-
-    :param stream: Stream for the asynchronous version.
-
-This function draws a colored disparity map by converting disparity values from ``[0..ndisp)`` interval first to ``HSV`` color space (where different disparity values correspond to different hues) and then converting the pixels to ``RGB`` for visualization.
-
-
-
-gpu::reprojectImageTo3D
----------------------------
-Reprojects a disparity image to 3D space.
-
-.. ocv:function:: void gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, int dst_cn = 4, Stream& stream = Stream::Null())
-
-    :param disp: Input disparity image.  ``CV_8U``  and  ``CV_16S``  types are supported.
-
-    :param xyzw: Output 3- or 4-channel floating-point image of the same size as  ``disp`` . Each element of  ``xyzw(x,y)``  contains 3D coordinates ``(x,y,z)`` or ``(x,y,z,1)``  of the point  ``(x,y)`` , computed from the disparity map.
-
-    :param Q: :math:`4 \times 4`  perspective transformation matrix that can be obtained via  :ocv:func:`stereoRectify` .
-
-    :param dst_cn: The number of channels for output image. Can be 3 or 4.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`reprojectImageTo3D`
-
-
-
-gpu::solvePnPRansac
--------------------
-Finds the object pose from 3D-2D point correspondences.
-
-.. ocv:function:: void gpu::solvePnPRansac(const Mat& object, const Mat& image, const Mat& camera_mat, const Mat& dist_coef, Mat& rvec, Mat& tvec, bool use_extrinsic_guess=false, int num_iters=100, float max_dist=8.0, int min_inlier_count=100, vector<int>* inliers=NULL)
-
-    :param object: Single-row matrix of object points.
-
-    :param image: Single-row matrix of image points.
-
-    :param camera_mat: 3x3 matrix of intrinsic camera parameters.
-
-    :param dist_coef: Distortion coefficients. See :ocv:func:`undistortPoints` for details.
-
-    :param rvec: Output 3D rotation vector.
-
-    :param tvec: Output 3D translation vector.
-
-    :param use_extrinsic_guess: Flag to indicate that the function must use ``rvec`` and ``tvec`` as an initial transformation guess. It is not supported for now.
-
-    :param num_iters: Maximum number of RANSAC iterations.
-
-    :param max_dist: Euclidean distance threshold to detect whether point is inlier or not.
-
-    :param min_inlier_count: Flag to indicate that the function must stop if greater or equal number of inliers is achieved. It is not supported for now.
-
-    :param inliers: Output vector of inlier indices.
-
-.. seealso:: :ocv:func:`solvePnPRansac`
-
-
-
-.. [Felzenszwalb2006] Pedro F. Felzenszwalb algorithm [Pedro F. Felzenszwalb and Daniel P. Huttenlocher. *Efficient belief propagation for early vision*. International Journal of Computer Vision, 70(1), October 2006
-
-.. [Yang2010] Q. Yang, L. Wang, and N. Ahuja. *A constant-space belief propagation algorithm for stereo matching*. In CVPR, 2010.
diff --git a/modules/gpu/doc/data_structures.rst b/modules/gpu/doc/data_structures.rst
index 5699723203..2a21710356 100644
--- a/modules/gpu/doc/data_structures.rst
+++ b/modules/gpu/doc/data_structures.rst
@@ -6,32 +6,22 @@ Data Structures
 
 
 gpu::PtrStepSz
----------------
+--------------
 .. ocv:class:: gpu::PtrStepSz
 
 Lightweight class encapsulating pitched memory on a GPU and passed to nvcc-compiled code (CUDA kernels). Typically, it is used internally by OpenCV and by users who write device code. You can call its members from both host and device code. ::
 
-    template <typename T> struct PtrStepSz
+    template <typename T> struct PtrStepSz : public PtrStep<T>
     {
-        int cols;
-        int rows;
-        T* data;
-        size_t step;
-
-        PtrStepSz() : cols(0), rows(0), data(0), step(0){};
-        PtrStepSz(int rows, int cols, T *data, size_t step);
+        __CV_GPU_HOST_DEVICE__ PtrStepSz() : cols(0), rows(0) {}
+        __CV_GPU_HOST_DEVICE__ PtrStepSz(int rows_, int cols_, T* data_, size_t step_)
+            : PtrStep<T>(data_, step_), cols(cols_), rows(rows_) {}
 
         template <typename U>
-        explicit PtrStepSz(const PtrStepSz<U>& d);
+        explicit PtrStepSz(const PtrStepSz<U>& d) : PtrStep<T>((T*)d.data, d.step), cols(d.cols), rows(d.rows){}
 
-        typedef T elem_type;
-        enum { elem_size = sizeof(elem_type) };
-
-        __CV_GPU_HOST_DEVICE__ size_t elemSize() const;
-
-        /* returns pointer to the beginning of the given image row */
-        __CV_GPU_HOST_DEVICE__ T* ptr(int y = 0);
-        __CV_GPU_HOST_DEVICE__ const T* ptr(int y = 0) const;
+        int cols;
+        int rows;
     };
 
     typedef PtrStepSz<unsigned char> PtrStepSzb;
@@ -41,32 +31,32 @@ Lightweight class encapsulating pitched memory on a GPU and passed to nvcc-compi
 
 
 gpu::PtrStep
---------------
+------------
 .. ocv:class:: gpu::PtrStep
 
 Structure similar to :ocv:class:`gpu::PtrStepSz` but containing only a pointer and row step. Width and height fields are excluded due to performance reasons. The structure is intended for internal use or for users who write device code. ::
 
-    template<typename T> struct PtrStep
+    template <typename T> struct PtrStep : public DevPtr<T>
     {
-            T* data;
-            size_t step;
+        __CV_GPU_HOST_DEVICE__ PtrStep() : step(0) {}
+        __CV_GPU_HOST_DEVICE__ PtrStep(T* data_, size_t step_) : DevPtr<T>(data_), step(step_) {}
 
-            PtrStep();
-            PtrStep(const PtrStepSz<T>& mem);
+        //! stride between two consecutive rows in bytes. Step is stored always and everywhere in bytes!!!
+        size_t step;
 
-            typedef T elem_type;
-            enum { elem_size = sizeof(elem_type) };
+        __CV_GPU_HOST_DEVICE__       T* ptr(int y = 0)       { return (      T*)( (      char*)DevPtr<T>::data + y * step); }
+        __CV_GPU_HOST_DEVICE__ const T* ptr(int y = 0) const { return (const T*)( (const char*)DevPtr<T>::data + y * step); }
 
-            __CV_GPU_HOST_DEVICE__ size_t elemSize() const;
-            __CV_GPU_HOST_DEVICE__ T* ptr(int y = 0);
-            __CV_GPU_HOST_DEVICE__ const T* ptr(int y = 0) const;
+        __CV_GPU_HOST_DEVICE__       T& operator ()(int y, int x)       { return ptr(y)[x]; }
+        __CV_GPU_HOST_DEVICE__ const T& operator ()(int y, int x) const { return ptr(y)[x]; }
     };
 
-    typedef PtrStep<unsigned char> PtrStep;
+    typedef PtrStep<unsigned char> PtrStepb;
     typedef PtrStep<float> PtrStepf;
     typedef PtrStep<int> PtrStepi;
 
 
+
 gpu::GpuMat
 -----------
 .. ocv:class:: gpu::GpuMat
@@ -89,28 +79,31 @@ Beware that the latter limitation may lead to overloaded matrix operators that c
             //! default constructor
             GpuMat();
 
+            //! constructs GpuMat of the specified size and type
             GpuMat(int rows, int cols, int type);
             GpuMat(Size size, int type);
 
             .....
 
-            //! builds GpuMat from Mat. Blocks uploading to device.
-            explicit GpuMat (const Mat& m);
+            //! builds GpuMat from host memory (Blocking call)
+            explicit GpuMat(InputArray arr);
 
             //! returns lightweight PtrStepSz structure for passing
             //to nvcc-compiled code. Contains size, data ptr and step.
             template <class T> operator PtrStepSz<T>() const;
             template <class T> operator PtrStep<T>() const;
 
-            //! blocks uploading data to GpuMat.
-            void upload(const cv::Mat& m);
-            void upload(const CudaMem& m, Stream& stream);
+            //! pefroms upload data to GpuMat (Blocking call)
+            void upload(InputArray arr);
 
-            //! downloads data from device to host memory. Blocking calls.
-            void download(cv::Mat& m) const;
+            //! pefroms upload data to GpuMat (Non-Blocking call)
+            void upload(InputArray arr, Stream& stream);
 
-            //! download async
-            void download(CudaMem& m, Stream& stream) const;
+            //! pefroms download data from device to host memory (Blocking call)
+            void download(OutputArray dst) const;
+
+            //! pefroms download data from device to host memory (Non-Blocking call)
+            void download(OutputArray dst, Stream& stream) const;
     };
 
 
@@ -121,16 +114,10 @@ Beware that the latter limitation may lead to overloaded matrix operators that c
 
 
 gpu::createContinuous
--------------------------
-Creates a continuous matrix in the GPU memory.
+---------------------
+Creates a continuous matrix.
 
-.. ocv:function:: void gpu::createContinuous(int rows, int cols, int type, GpuMat& m)
-
-.. ocv:function:: GpuMat gpu::createContinuous(int rows, int cols, int type)
-
-.. ocv:function:: void gpu::createContinuous(Size size, int type, GpuMat& m)
-
-.. ocv:function:: GpuMat gpu::createContinuous(Size size, int type)
+.. ocv:function:: void gpu::createContinuous(int rows, int cols, int type, OutputArray arr)
 
     :param rows: Row count.
 
@@ -138,36 +125,92 @@ Creates a continuous matrix in the GPU memory.
 
     :param type: Type of the matrix.
 
-    :param m: Destination matrix. This parameter changes only if it has a proper type and area ( :math:`\texttt{rows} \times \texttt{cols}` ).
+    :param arr: Destination matrix. This parameter changes only if it has a proper type and area ( :math:`\texttt{rows} \times \texttt{cols}` ).
 
 Matrix is called continuous if its elements are stored continuously, that is, without gaps at the end of each row.
 
 
 
 gpu::ensureSizeIsEnough
----------------------------
+-----------------------
 Ensures that the size of a matrix is big enough and the matrix has a proper type.
 
-.. ocv:function:: void gpu::ensureSizeIsEnough(int rows, int cols, int type, GpuMat& m)
-
-.. ocv:function:: void gpu::ensureSizeIsEnough(Size size, int type, GpuMat& m)
+.. ocv:function:: void gpu::ensureSizeIsEnough(int rows, int cols, int type, OutputArray arr)
 
     :param rows: Minimum desired number of rows.
 
     :param cols: Minimum desired number of columns.
 
-    :param size: Rows and columns passed as a structure.
-
     :param type: Desired matrix type.
 
-    :param m: Destination matrix.
+    :param arr: Destination matrix.
 
 The function does not reallocate memory if the matrix has proper attributes already.
 
 
 
+gpu::CudaMem
+------------
+.. ocv:class:: gpu::CudaMem
+
+Class with reference counting wrapping special memory type allocation functions from CUDA. Its interface is also :ocv:func:`Mat`-like but with additional memory type parameters.
+
+* **PAGE_LOCKED** sets a page locked memory type used commonly for fast and asynchronous uploading/downloading data from/to GPU.
+* **SHARED** specifies a zero copy memory allocation that enables mapping the host memory to GPU address space, if supported.
+* **WRITE_COMBINED**  sets the write combined buffer that is not cached by CPU. Such buffers are used to supply GPU with data when GPU only reads it. The advantage is a better CPU cache utilization.
+
+.. note:: Allocation size of such memory types is usually limited. For more details, see *CUDA 2.2 Pinned Memory APIs* document or *CUDA C Programming Guide*.
+
+::
+
+    class CV_EXPORTS CudaMem
+    {
+    public:
+        enum AllocType { PAGE_LOCKED = 1, SHARED = 2, WRITE_COMBINED = 4 };
+
+        explicit CudaMem(AllocType alloc_type = PAGE_LOCKED);
+
+        CudaMem(int rows, int cols, int type, AllocType alloc_type = PAGE_LOCKED);
+        CudaMem(Size size, int type, AllocType alloc_type = PAGE_LOCKED);
+
+        //! creates from host memory with coping data
+        explicit CudaMem(InputArray arr, AllocType alloc_type = PAGE_LOCKED);
+
+        ......
+
+        //! returns matrix header with disabled reference counting for CudaMem data.
+        Mat createMatHeader() const;
+
+        //! maps host memory into device address space and returns GpuMat header for it. Throws exception if not supported by hardware.
+        GpuMat createGpuMatHeader() const;
+
+        ......
+
+        AllocType alloc_type;
+    };
+
+
+
+gpu::CudaMem::createMatHeader
+-----------------------------
+Creates a header without reference counting to :ocv:class:`gpu::CudaMem` data.
+
+.. ocv:function:: Mat gpu::CudaMem::createMatHeader() const
+
+
+
+gpu::CudaMem::createGpuMatHeader
+--------------------------------
+Maps CPU memory to GPU address space and creates the :ocv:class:`gpu::GpuMat` header without reference counting for it.
+
+.. ocv:function:: GpuMat gpu::CudaMem::createGpuMatHeader() const
+
+This can be done only if memory was allocated with the ``SHARED`` flag and if it is supported by the hardware. Laptops often share video and CPU memory, so address spaces can be mapped, which eliminates an extra copy.
+
+
+
 gpu::registerPageLocked
--------------------------------
+-----------------------
 Page-locks the memory of matrix and maps it for the device(s).
 
 .. ocv:function:: void gpu::registerPageLocked(Mat& m)
@@ -177,7 +220,7 @@ Page-locks the memory of matrix and maps it for the device(s).
 
 
 gpu::unregisterPageLocked
--------------------------------
+-------------------------
 Unmaps the memory of matrix and makes it pageable again.
 
 .. ocv:function:: void gpu::unregisterPageLocked(Mat& m)
@@ -186,83 +229,11 @@ Unmaps the memory of matrix and makes it pageable again.
 
 
 
-gpu::CudaMem
-------------
-.. ocv:class:: gpu::CudaMem
-
-Class with reference counting wrapping special memory type allocation functions from CUDA. Its interface is also
-:ocv:func:`Mat`-like but with additional memory type parameters.
-
-* **ALLOC_PAGE_LOCKED** sets a page locked memory type used commonly for fast and asynchronous uploading/downloading data from/to GPU.
-* **ALLOC_ZEROCOPY** specifies a zero copy memory allocation that enables mapping the host memory to GPU address space, if supported.
-* **ALLOC_WRITE_COMBINED**  sets the write combined buffer that is not cached by CPU. Such buffers are used to supply GPU with data when GPU only reads it. The advantage is a better CPU cache utilization.
-
-.. note:: Allocation size of such memory types is usually limited. For more details, see *CUDA 2.2 Pinned Memory APIs* document or *CUDA C Programming Guide*.
-
-::
-
-    class CV_EXPORTS CudaMem
-    {
-    public:
-            enum  { ALLOC_PAGE_LOCKED = 1, ALLOC_ZEROCOPY = 2,
-                     ALLOC_WRITE_COMBINED = 4 };
-
-            CudaMem(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
-
-            //! creates from cv::Mat with coping data
-            explicit CudaMem(const Mat& m, int alloc_type = ALLOC_PAGE_LOCKED);
-
-             ......
-
-            void create(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
-
-            //! returns matrix header with disabled ref. counting for CudaMem data.
-            Mat createMatHeader() const;
-            operator Mat() const;
-
-            //! maps host memory into device address space
-            GpuMat createGpuMatHeader() const;
-            operator GpuMat() const;
-
-            //if host memory can be mapped to gpu address space;
-            static bool canMapHostMemory();
-
-            int alloc_type;
-    };
-
-
-
-gpu::CudaMem::createMatHeader
----------------------------------
-Creates a header without reference counting to :ocv:class:`gpu::CudaMem` data.
-
-.. ocv:function:: Mat gpu::CudaMem::createMatHeader() const
-
-
-
-gpu::CudaMem::createGpuMatHeader
-------------------------------------
-Maps CPU memory to GPU address space and creates the :ocv:class:`gpu::GpuMat` header without reference counting for it.
-
-.. ocv:function:: GpuMat gpu::CudaMem::createGpuMatHeader() const
-
-This can be done only if memory was allocated with the ``ALLOC_ZEROCOPY`` flag and if it is supported by the hardware. Laptops often share video and CPU memory, so address spaces can be mapped, which eliminates an extra copy.
-
-
-
-gpu::CudaMem::canMapHostMemory
-----------------------------------
-Returns ``true`` if the current hardware supports address space mapping and ``ALLOC_ZEROCOPY`` memory allocation.
-
-.. ocv:function:: static bool gpu::CudaMem::canMapHostMemory()
-
-
-
 gpu::Stream
 -----------
 .. ocv:class:: gpu::Stream
 
-This class encapsulates a queue of asynchronous calls. Some functions have overloads with the additional ``gpu::Stream`` parameter. The overloads do initialization work (allocate output buffers, upload constants, and so on), start the GPU kernel, and return before results are ready. You can check whether all operations are complete via :ocv:func:`gpu::Stream::queryIfComplete`. You can asynchronously upload/download data from/to page-locked buffers, using the :ocv:class:`gpu::CudaMem` or :ocv:class:`Mat` header that points to a region of :ocv:class:`gpu::CudaMem`.
+This class encapsulates a queue of asynchronous calls.
 
 .. note:: Currently, you may face problems if an operation is enqueued twice with different data. Some functions use the constant GPU memory, and next call may update the memory before the previous one has been finished. But calling different operations asynchronously is safe because each operation has its own constant buffer. Memory copy/upload/download/set operations to the buffers you hold are also safe.
 
@@ -272,30 +243,24 @@ This class encapsulates a queue of asynchronous calls. Some functions have overl
     {
     public:
         Stream();
-        ~Stream();
 
-        Stream(const Stream&);
-        Stream& operator=(const Stream&);
+        //! queries an asynchronous stream for completion status
+        bool queryIfComplete() const;
 
-        bool queryIfComplete();
+        //! waits for stream tasks to complete
         void waitForCompletion();
 
-        void enqueueDownload(const GpuMat& src, CudaMem& dst);
-        void enqueueDownload(const GpuMat& src, Mat& dst);
+        //! makes a compute stream wait on an event
+        void waitEvent(const Event& event);
 
-        void enqueueUpload(const CudaMem& src, GpuMat& dst);
-        void enqueueUpload(const Mat& src, GpuMat& dst);
-
-        void enqueueCopy(const GpuMat& src, GpuMat& dst);
-
-        void enqueueMemSet(const GpuMat& src, Scalar val);
-        void enqueueMemSet(const GpuMat& src, Scalar val, const GpuMat& mask);
-
-        void enqueueConvert(const GpuMat& src, GpuMat& dst, int type,
-                            double a = 1, double b = 0);
-
-        typedef void (*StreamCallback)(Stream& stream, int status, void* userData);
+        //! adds a callback to be called on the host after all currently enqueued items in the stream have completed
         void enqueueHostCallback(StreamCallback callback, void* userData);
+
+        //! return Stream object for default CUDA stream
+        static Stream& Null();
+
+        //! returns true if stream object is not default (!= 0)
+        operator bool_type() const;
     };
 
 
@@ -316,53 +281,11 @@ Blocks the current CPU thread until all operations in the stream are complete.
 
 
 
-gpu::Stream::enqueueDownload
-----------------------------
-Copies data from device to host.
+gpu::Stream::waitEvent
+----------------------
+Makes a compute stream wait on an event.
 
-.. ocv:function:: void gpu::Stream::enqueueDownload(const GpuMat& src, CudaMem& dst)
-
-.. ocv:function:: void gpu::Stream::enqueueDownload(const GpuMat& src, Mat& dst)
-
-.. note:: ``cv::Mat`` must point to page locked memory (i.e. to ``CudaMem`` data or to its subMat) or must be registered with :ocv:func:`gpu::registerPageLocked` .
-
-
-
-gpu::Stream::enqueueUpload
---------------------------
-Copies data from host to device.
-
-.. ocv:function:: void gpu::Stream::enqueueUpload(const CudaMem& src, GpuMat& dst)
-
-.. ocv:function:: void gpu::Stream::enqueueUpload(const Mat& src, GpuMat& dst)
-
-.. note:: ``cv::Mat`` must point to page locked memory (i.e. to ``CudaMem`` data or to its subMat) or must be registered with :ocv:func:`gpu::registerPageLocked` .
-
-
-
-gpu::Stream::enqueueCopy
-------------------------
-Copies data from device to device.
-
-.. ocv:function:: void gpu::Stream::enqueueCopy(const GpuMat& src, GpuMat& dst)
-
-
-
-gpu::Stream::enqueueMemSet
---------------------------
-Initializes or sets device memory to a value.
-
-.. ocv:function:: void gpu::Stream::enqueueMemSet( GpuMat& src, Scalar val )
-
-.. ocv:function:: void gpu::Stream::enqueueMemSet( GpuMat& src, Scalar val, const GpuMat& mask )
-
-
-
-gpu::Stream::enqueueConvert
----------------------------
-Converts matrix type, ex from float to uchar depending on type.
-
-.. ocv:function:: void gpu::Stream::enqueueConvert( const GpuMat& src, GpuMat& dst, int dtype, double a=1, double b=0 )
+.. ocv:function:: void gpu::Stream::waitEvent(const Event& event)
 
 
 
@@ -386,4 +309,3 @@ Class that enables getting ``cudaStream_t`` from :ocv:class:`gpu::Stream` and is
     {
         CV_EXPORTS static cudaStream_t getStream(const Stream& stream);
     };
-
diff --git a/modules/gpu/doc/gpu.rst b/modules/gpu/doc/gpu.rst
index b21e2abac8..2a0358e017 100644
--- a/modules/gpu/doc/gpu.rst
+++ b/modules/gpu/doc/gpu.rst
@@ -8,12 +8,5 @@ gpu. GPU-accelerated Computer Vision
     introduction
     initalization_and_information
     data_structures
-    operations_on_matrices
-    per_element_operations
-    image_processing
-    matrix_reductions
     object_detection
-    feature_detection_and_description
-    image_filtering
-    camera_calibration_and_3d_reconstruction
-    video
+    calib3d
diff --git a/modules/gpu/doc/image_filtering.rst b/modules/gpu/doc/image_filtering.rst
deleted file mode 100644
index 348a42510e..0000000000
--- a/modules/gpu/doc/image_filtering.rst
+++ /dev/null
@@ -1,719 +0,0 @@
-Image Filtering
-===============
-
-.. highlight:: cpp
-
-Functions and classes described in this section are used to perform various linear or non-linear filtering operations on 2D images.
-
-
-
-gpu::BaseRowFilter_GPU
-----------------------
-.. ocv:class:: gpu::BaseRowFilter_GPU
-
-Base class for linear or non-linear filters that processes rows of 2D arrays. Such filters are used for the "horizontal" filtering passes in separable filters. ::
-
-    class BaseRowFilter_GPU
-    {
-    public:
-        BaseRowFilter_GPU(int ksize_, int anchor_);
-        virtual ~BaseRowFilter_GPU() {}
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-        int ksize, anchor;
-    };
-
-
-.. note:: This class does not allocate memory for a destination image. Usually this class is used inside :ocv:class:`gpu::FilterEngine_GPU`.
-
-
-
-gpu::BaseColumnFilter_GPU
--------------------------
-.. ocv:class:: gpu::BaseColumnFilter_GPU
-
-Base class for linear or non-linear filters that processes columns of 2D arrays. Such filters are used for the "vertical" filtering passes in separable filters. ::
-
-    class BaseColumnFilter_GPU
-    {
-    public:
-        BaseColumnFilter_GPU(int ksize_, int anchor_);
-        virtual ~BaseColumnFilter_GPU() {}
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-        int ksize, anchor;
-    };
-
-
-.. note:: This class does not allocate memory for a destination image. Usually this class is used inside :ocv:class:`gpu::FilterEngine_GPU`.
-
-
-
-gpu::BaseFilter_GPU
--------------------
-.. ocv:class:: gpu::BaseFilter_GPU
-
-Base class for non-separable 2D filters. ::
-
-    class CV_EXPORTS BaseFilter_GPU
-    {
-    public:
-        BaseFilter_GPU(const Size& ksize_, const Point& anchor_);
-        virtual ~BaseFilter_GPU() {}
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-        Size ksize;
-        Point anchor;
-    };
-
-
-.. note:: This class does not allocate memory for a destination image. Usually this class is used inside :ocv:class:`gpu::FilterEngine_GPU`.
-
-
-
-gpu::FilterEngine_GPU
----------------------
-.. ocv:class:: gpu::FilterEngine_GPU
-
-Base class for the Filter Engine. ::
-
-    class CV_EXPORTS FilterEngine_GPU
-    {
-    public:
-        virtual ~FilterEngine_GPU() {}
-
-        virtual void apply(const GpuMat& src, GpuMat& dst,
-                           Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null()) = 0;
-    };
-
-
-The class can be used to apply an arbitrary filtering operation to an image. It contains all the necessary intermediate buffers. Pointers to the initialized ``FilterEngine_GPU`` instances are returned by various ``create*Filter_GPU`` functions (see below), and they are used inside high-level functions such as :ocv:func:`gpu::filter2D`, :ocv:func:`gpu::erode`, :ocv:func:`gpu::Sobel` , and others.
-
-By using ``FilterEngine_GPU`` instead of functions you can avoid unnecessary memory allocation for intermediate buffers and get better performance: ::
-
-    while (...)
-    {
-        gpu::GpuMat src = getImg();
-        gpu::GpuMat dst;
-        // Allocate and release buffers at each iterations
-        gpu::GaussianBlur(src, dst, ksize, sigma1);
-    }
-
-    // Allocate buffers only once
-    cv::Ptr<gpu::FilterEngine_GPU> filter =
-        gpu::createGaussianFilter_GPU(CV_8UC4, ksize, sigma1);
-    while (...)
-    {
-        gpu::GpuMat src = getImg();
-        gpu::GpuMat dst;
-        filter->apply(src, dst, cv::Rect(0, 0, src.cols, src.rows));
-    }
-    // Release buffers only once
-    filter.release();
-
-
-``FilterEngine_GPU`` can process a rectangular sub-region of an image. By default, if ``roi == Rect(0,0,-1,-1)`` , ``FilterEngine_GPU`` processes the inner region of an image ( ``Rect(anchor.x, anchor.y, src_size.width - ksize.width, src_size.height - ksize.height)`` ) because some filters do not check whether indices are outside the image for better performance. See below to understand which filters support processing the whole image and which do not and identify image type limitations.
-
-.. note:: The GPU filters do not support the in-place mode.
-
-.. seealso:: :ocv:class:`gpu::BaseRowFilter_GPU`, :ocv:class:`gpu::BaseColumnFilter_GPU`, :ocv:class:`gpu::BaseFilter_GPU`, :ocv:func:`gpu::createFilter2D_GPU`, :ocv:func:`gpu::createSeparableFilter_GPU`, :ocv:func:`gpu::createBoxFilter_GPU`, :ocv:func:`gpu::createMorphologyFilter_GPU`, :ocv:func:`gpu::createLinearFilter_GPU`, :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`gpu::createDerivFilter_GPU`, :ocv:func:`gpu::createGaussianFilter_GPU`
-
-
-
-gpu::createFilter2D_GPU
----------------------------
-Creates a non-separable filter engine with the specified filter.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createFilter2D_GPU( const Ptr<BaseFilter_GPU>& filter2D, int srcType, int dstType)
-
-    :param filter2D: Non-separable 2D filter.
-
-    :param srcType: Input image type. It must be supported by  ``filter2D`` .
-
-    :param dstType: Output image type. It must be supported by  ``filter2D`` .
-
-Usually this function is used inside such high-level functions as :ocv:func:`gpu::createLinearFilter_GPU`, :ocv:func:`gpu::createBoxFilter_GPU`.
-
-
-
-gpu::createSeparableFilter_GPU
-----------------------------------
-Creates a separable filter engine with the specified filters.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createSeparableFilter_GPU( const Ptr<BaseRowFilter_GPU>& rowFilter, const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType)
-
-    :param rowFilter: "Horizontal" 1D filter.
-
-    :param columnFilter: "Vertical" 1D filter.
-
-    :param srcType: Input image type. It must be supported by  ``rowFilter`` .
-
-    :param bufType: Buffer image type. It must be supported by  ``rowFilter``  and  ``columnFilter`` .
-
-    :param dstType: Output image type. It must be supported by  ``columnFilter`` .
-
-Usually this function is used inside such high-level functions as :ocv:func:`gpu::createSeparableLinearFilter_GPU`.
-
-
-
-gpu::getRowSumFilter_GPU
-----------------------------
-Creates a horizontal 1D box filter.
-
-.. ocv:function:: Ptr<BaseRowFilter_GPU> gpu::getRowSumFilter_GPU(int srcType, int sumType, int ksize, int anchor = -1)
-
-    :param srcType: Input image type. Only ``CV_8UC1`` type is supported for now.
-
-    :param sumType: Output image type. Only ``CV_32FC1`` type is supported for now.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-
-
-gpu::getColumnSumFilter_GPU
--------------------------------
-Creates a vertical 1D box filter.
-
-.. ocv:function:: Ptr<BaseColumnFilter_GPU> gpu::getColumnSumFilter_GPU(int sumType, int dstType, int ksize, int anchor = -1)
-
-    :param sumType: Input image type. Only ``CV_8UC1`` type is supported for now.
-
-    :param dstType: Output image type. Only ``CV_32FC1`` type is supported for now.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-
-
-gpu::createBoxFilter_GPU
-----------------------------
-Creates a normalized 2D box filter.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createBoxFilter_GPU(int srcType, int dstType, const Size& ksize, const Point& anchor = Point(-1,-1))
-
-.. ocv:function:: Ptr<BaseFilter_GPU> gpu::getBoxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1, -1))
-
-    :param srcType: Input image type supporting ``CV_8UC1`` and ``CV_8UC4`` .
-
-    :param dstType: Output image type.  It supports only the same values as the source type.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value ``Point(-1, -1)`` means that the anchor is at the kernel center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`boxFilter`
-
-
-
-gpu::boxFilter
-------------------
-Smooths the image using the normalized box filter.
-
-.. ocv:function:: void gpu::boxFilter(const GpuMat& src, GpuMat& dst, int ddepth, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null())
-
-    :param src: Input image. ``CV_8UC1`` and ``CV_8UC4`` source types are supported.
-
-    :param dst: Output image type. The size and type is the same as ``src`` .
-
-    :param ddepth: Output image depth. If -1, the output image has the same depth as the input one. The only values allowed here are ``CV_8U`` and -1.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value ``Point(-1, -1)`` means that the anchor is at the kernel center.
-
-    :param stream: Stream for the asynchronous version.
-
-.. note::    This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`boxFilter`
-
-
-
-gpu::blur
--------------
-Acts as a synonym for the normalized box filter.
-
-.. ocv:function:: void gpu::blur(const GpuMat& src, GpuMat& dst, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null())
-
-    :param src: Input image.  ``CV_8UC1``  and  ``CV_8UC4``  source types are supported.
-
-    :param dst: Output image type with the same size and type as  ``src`` .
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value Point(-1, -1) means that the anchor is at the kernel center.
-
-    :param stream: Stream for the asynchronous version.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`blur`, :ocv:func:`gpu::boxFilter`
-
-
-
-gpu::createMorphologyFilter_GPU
------------------------------------
-Creates a 2D morphological filter.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Point& anchor = Point(-1,-1), int iterations = 1)
-
-.. ocv:function:: Ptr<BaseFilter_GPU> gpu::getMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Size& ksize, Point anchor=Point(-1,-1))
-
-    :param op: Morphology operation id. Only ``MORPH_ERODE`` and ``MORPH_DILATE`` are supported.
-
-    :param type: Input/output image type. Only  ``CV_8UC1``  and  ``CV_8UC4``  are supported.
-
-    :param kernel: 2D 8-bit structuring element for the morphological operation.
-
-    :param ksize: Size of a horizontal or vertical structuring element used for separable morphological operations.
-
-    :param anchor: Anchor position within the structuring element. Negative values mean that the anchor is at the center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`createMorphologyFilter`
-
-
-
-gpu::erode
---------------
-Erodes an image by using a specific structuring element.
-
-.. ocv:function:: void gpu::erode( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor=Point(-1, -1), int iterations=1 )
-
-.. ocv:function:: void gpu::erode( const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf, Point anchor=Point(-1, -1), int iterations=1, Stream& stream=Stream::Null() )
-
-    :param src: Source image. Only  ``CV_8UC1``  and  ``CV_8UC4``  types are supported.
-
-    :param dst: Destination image with the same size and type as  ``src`` .
-
-    :param kernel: Structuring element used for erosion. If  ``kernel=Mat()``, a  3x3 rectangular structuring element is used.
-
-    :param anchor: Position of an anchor within the element. The default value  ``(-1, -1)``  means that the anchor is at the element center.
-
-    :param iterations: Number of times erosion to be applied.
-
-    :param stream: Stream for the asynchronous version.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`erode`
-
-
-
-gpu::dilate
----------------
-Dilates an image by using a specific structuring element.
-
-.. ocv:function:: void gpu::dilate( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor=Point(-1, -1), int iterations=1 )
-
-.. ocv:function:: void gpu::dilate( const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf, Point anchor=Point(-1, -1), int iterations=1, Stream& stream=Stream::Null() )
-
-    :param src: Source image. ``CV_8UC1`` and ``CV_8UC4`` source types are supported.
-
-    :param dst: Destination image with the same size and type as ``src``.
-
-    :param kernel: Structuring element used for dilation. If  ``kernel=Mat()``, a  3x3 rectangular structuring element is used.
-
-    :param anchor: Position of an anchor within the element. The default value  ``(-1, -1)``  means that the anchor is at the element center.
-
-    :param iterations: Number of times dilation to be applied.
-
-    :param stream: Stream for the asynchronous version.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`dilate`
-
-
-
-gpu::morphologyEx
----------------------
-Applies an advanced morphological operation to an image.
-
-.. ocv:function::  void gpu::morphologyEx( const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, Point anchor=Point(-1, -1), int iterations=1 )
-
-.. ocv:function:: void gpu::morphologyEx( const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, GpuMat& buf1, GpuMat& buf2, Point anchor=Point(-1, -1), int iterations=1, Stream& stream=Stream::Null() )
-
-    :param src: Source image.  ``CV_8UC1``  and  ``CV_8UC4``  source types are supported.
-
-    :param dst: Destination image with the same size and type as  ``src`` .
-
-    :param op: Type of morphological operation. The following types are possible:
-
-        * **MORPH_OPEN** opening
-
-        * **MORPH_CLOSE** closing
-
-        * **MORPH_GRADIENT** morphological gradient
-
-        * **MORPH_TOPHAT** "top hat"
-
-        * **MORPH_BLACKHAT** "black hat"
-
-    :param kernel: Structuring element.
-
-    :param anchor: Position of an anchor within the element. The default value ``Point(-1, -1)`` means that the anchor is at the element center.
-
-    :param iterations: Number of times erosion and dilation to be applied.
-
-    :param stream: Stream for the asynchronous version.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`morphologyEx`
-
-
-
-gpu::createLinearFilter_GPU
--------------------------------
-Creates a non-separable linear filter.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createLinearFilter_GPU(int srcType, int dstType, const Mat& kernel, Point anchor = Point(-1,-1), int borderType = BORDER_DEFAULT)
-
-    :param srcType: Input image type. Supports  ``CV_8U``  ,  ``CV_16U``  and  ``CV_32F``  one and four channel image.
-
-    :param dstType: Output image type. The same type as ``src`` is supported.
-
-    :param kernel: 2D array of filter coefficients. Floating-point coefficients will be converted to fixed-point representation before the actual processing. Supports size up to 16. For larger kernels use :ocv:func:`gpu::convolve`.
-
-    :param anchor: Anchor point. The default value Point(-1, -1) means that the anchor is at the kernel center.
-
-    :param borderType: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
-
-.. seealso:: :ocv:func:`createLinearFilter`
-
-
-
-gpu::filter2D
------------------
-Applies the non-separable 2D linear filter to an image.
-
-.. ocv:function:: void gpu::filter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernel, Point anchor=Point(-1,-1), int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null())
-
-    :param src: Source image. Supports  ``CV_8U``  ,  ``CV_16U``  and  ``CV_32F``  one and four channel image.
-
-    :param dst: Destination image. The size and the number of channels is the same as  ``src`` .
-
-    :param ddepth: Desired depth of the destination image. If it is negative, it is the same as  ``src.depth()`` . It supports only the same depth as the source image depth.
-
-    :param kernel: 2D array of filter coefficients.
-
-    :param anchor: Anchor of the kernel that indicates the relative position of a filtered point within the kernel. The anchor resides within the kernel. The special default value (-1,-1) means that the anchor is at the kernel center.
-
-    :param borderType: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`filter2D`, :ocv:func:`gpu::convolve`
-
-
-
-gpu::Laplacian
-------------------
-Applies the Laplacian operator to an image.
-
-.. ocv:function:: void gpu::Laplacian(const GpuMat& src, GpuMat& dst, int ddepth, int ksize = 1, double scale = 1, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null())
-
-    :param src: Source image. ``CV_8UC1``  and  ``CV_8UC4``  source types are supported.
-
-    :param dst: Destination image. The size and number of channels is the same as  ``src`` .
-
-    :param ddepth: Desired depth of the destination image. It supports only the same depth as the source image depth.
-
-    :param ksize: Aperture size used to compute the second-derivative filters (see :ocv:func:`getDerivKernels`). It must be positive and odd. Only  ``ksize``  = 1 and  ``ksize``  = 3 are supported.
-
-    :param scale: Optional scale factor for the computed Laplacian values. By default, no scaling is applied (see  :ocv:func:`getDerivKernels` ).
-
-    :param borderType: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-.. seealso:: :ocv:func:`Laplacian`, :ocv:func:`gpu::filter2D`
-
-
-
-gpu::getLinearRowFilter_GPU
--------------------------------
-Creates a primitive row filter with the specified kernel.
-
-.. ocv:function:: Ptr<BaseRowFilter_GPU> gpu::getLinearRowFilter_GPU( int srcType, int bufType, const Mat& rowKernel, int anchor=-1, int borderType=BORDER_DEFAULT )
-
-    :param srcType: Source array type. Only  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param bufType: Intermediate buffer type with as many channels as  ``srcType`` .
-
-    :param rowKernel: Filter coefficients. Support kernels with ``size <= 16`` .
-
-    :param anchor: Anchor position within the kernel. Negative values mean that the anchor is positioned at the aperture center.
-
-    :param borderType: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate`. For details on limitations, see below.
-
-There are two versions of the algorithm: NPP and OpenCV.
-
-    * NPP version is called when ``srcType == CV_8UC1`` or ``srcType == CV_8UC4`` and ``bufType == srcType`` . Otherwise, the OpenCV version is called. NPP supports only ``BORDER_CONSTANT`` border type and does not check indices outside the image.
-
-    * OpenCV version supports only ``CV_32F`` buffer depth and ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , and ``BORDER_CONSTANT`` border types. It checks indices outside the image.
-
-.. seealso:: :ocv:func:`createSeparableLinearFilter` .
-
-
-
-gpu::getLinearColumnFilter_GPU
-----------------------------------
-Creates a primitive column filter with the specified kernel.
-
-.. ocv:function:: Ptr<BaseColumnFilter_GPU> gpu::getLinearColumnFilter_GPU( int bufType, int dstType, const Mat& columnKernel, int anchor=-1, int borderType=BORDER_DEFAULT )
-
-    :param bufType: Intermediate buffer type with as many channels as  ``dstType`` .
-
-    :param dstType: Destination array type. ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1`` destination types are supported.
-
-    :param columnKernel: Filter coefficients. Support kernels with ``size <= 16`` .
-
-    :param anchor: Anchor position within the kernel. Negative values mean that the anchor is positioned at the aperture center.
-
-    :param borderType: Pixel extrapolation method. For details, see  :ocv:func:`borderInterpolate` . For details on limitations, see below.
-
-There are two versions of the algorithm: NPP and OpenCV.
-
-    * NPP version is called when ``dstType == CV_8UC1`` or ``dstType == CV_8UC4`` and ``bufType == dstType`` . Otherwise, the OpenCV version is called. NPP supports only ``BORDER_CONSTANT`` border type and does not check indices outside the image.
-
-    * OpenCV version supports only ``CV_32F`` buffer depth and ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , and ``BORDER_CONSTANT`` border types. It checks indices outside image.
-
-.. seealso:: :ocv:func:`gpu::getLinearRowFilter_GPU`, :ocv:func:`createSeparableLinearFilter`
-
-
-
-gpu::createSeparableLinearFilter_GPU
-----------------------------------------
-Creates a separable linear filter engine.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel, const Mat& columnKernel, const Point& anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1)
-
-    :param srcType: Source array type.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dstType: Destination array type.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  destination types are supported.
-
-    :param rowKernel: Horizontal filter coefficients. Support kernels with ``size <= 16`` .
-
-    :param columnKernel: Vertical filter coefficients. Support kernels with ``size <= 16`` .
-
-    :param anchor: Anchor position within the kernel. Negative values mean that anchor is positioned at the aperture center.
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction For details, see  :ocv:func:`borderInterpolate`. For details on limitations, see :ocv:func:`gpu::getLinearRowFilter_GPU`, cpp:ocv:func:`gpu::getLinearColumnFilter_GPU`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-.. seealso:: :ocv:func:`gpu::getLinearRowFilter_GPU`, :ocv:func:`gpu::getLinearColumnFilter_GPU`, :ocv:func:`createSeparableLinearFilter`
-
-
-
-gpu::sepFilter2D
---------------------
-Applies a separable 2D linear filter to an image.
-
-.. ocv:function:: void gpu::sepFilter2D( const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY, Point anchor=Point(-1,-1), int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1 )
-
-.. ocv:function:: void gpu::sepFilter2D( const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY, GpuMat& buf, Point anchor=Point(-1,-1), int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1, Stream& stream=Stream::Null() )
-
-
-    :param src: Source image.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dst: Destination image with the same size and number of channels as  ``src`` .
-
-    :param ddepth: Destination image depth.  ``CV_8U`` , ``CV_16S`` , ``CV_32S`` , and  ``CV_32F`` are supported.
-
-    :param kernelX: Horizontal filter coefficients.
-
-    :param kernelY: Vertical filter coefficients.
-
-    :param anchor: Anchor position within the kernel. The default value ``(-1, 1)`` means that the anchor is at the kernel center.
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`sepFilter2D`
-
-
-
-gpu::createDerivFilter_GPU
-------------------------------
-Creates a filter engine for the generalized Sobel operator.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize, int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1)
-
-    :param srcType: Source image type.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dstType: Destination image type with as many channels as  ``srcType`` ,  ``CV_8U`` , ``CV_16S`` , ``CV_32S`` , and  ``CV_32F``  depths are supported.
-
-    :param dx: Derivative order in respect of x.
-
-    :param dy: Derivative order in respect of y.
-
-    :param ksize: Aperture size. See  :ocv:func:`getDerivKernels` for details.
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-.. seealso:: :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`createDerivFilter`
-
-
-
-gpu::Sobel
---------------
-Applies the generalized Sobel operator to an image.
-
-.. ocv:function:: void gpu::Sobel( const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, int ksize=3, double scale=1, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1 )
-
-.. ocv:function:: void gpu::Sobel( const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, int ksize=3, double scale=1, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1, Stream& stream=Stream::Null() )
-
-    :param src: Source image.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dst: Destination image with the same size and number of channels as source image.
-
-    :param ddepth: Destination image depth.  ``CV_8U`` , ``CV_16S`` , ``CV_32S`` , and  ``CV_32F`` are supported.
-
-    :param dx: Derivative order in respect of x.
-
-    :param dy: Derivative order in respect of y.
-
-    :param ksize: Size of the extended Sobel kernel. Possible values are 1, 3, 5 or 7.
-
-    :param scale: Optional scale factor for the computed derivative values. By default, no scaling is applied. For details, see  :ocv:func:`getDerivKernels` .
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`Sobel`
-
-
-
-gpu::Scharr
----------------
-Calculates the first x- or y- image derivative using the Scharr operator.
-
-.. ocv:function:: void gpu::Scharr( const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, double scale=1, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1 )
-
-.. ocv:function:: void gpu::Scharr( const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, double scale=1, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1, Stream& stream=Stream::Null() )
-
-    :param src: Source image.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dst: Destination image with the same size and number of channels as  ``src`` has.
-
-    :param ddepth: Destination image depth.  ``CV_8U`` , ``CV_16S`` , ``CV_32S`` , and  ``CV_32F`` are supported.
-
-    :param dx: Order of the derivative x.
-
-    :param dy: Order of the derivative y.
-
-    :param scale: Optional scale factor for the computed derivative values. By default, no scaling is applied. See  :ocv:func:`getDerivKernels`  for details.
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`Scharr`
-
-
-
-gpu::createGaussianFilter_GPU
----------------------------------
-Creates a Gaussian filter engine.
-
-.. ocv:function:: Ptr<FilterEngine_GPU> gpu::createGaussianFilter_GPU( int type, Size ksize, double sigma1, double sigma2=0, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1 )
-
-    :param type: Source and destination image type.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1`` are supported.
-
-    :param ksize: Aperture size. See  :ocv:func:`getGaussianKernel` for details.
-
-    :param sigma1: Gaussian sigma in the horizontal direction. See  :ocv:func:`getGaussianKernel` for details.
-
-    :param sigma2: Gaussian sigma in the vertical direction. If 0, then  :math:`\texttt{sigma2}\leftarrow\texttt{sigma1}` .
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-.. seealso:: :ocv:func:`gpu::createSeparableLinearFilter_GPU`, :ocv:func:`createGaussianFilter`
-
-
-
-gpu::GaussianBlur
----------------------
-Smooths an image using the Gaussian filter.
-
-.. ocv:function:: void gpu::GaussianBlur( const GpuMat& src, GpuMat& dst, Size ksize, double sigma1, double sigma2=0, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1 )
-
-.. ocv:function:: void gpu::GaussianBlur( const GpuMat& src, GpuMat& dst, Size ksize, GpuMat& buf, double sigma1, double sigma2=0, int rowBorderType=BORDER_DEFAULT, int columnBorderType=-1, Stream& stream=Stream::Null() )
-
-    :param src: Source image.  ``CV_8UC1`` , ``CV_8UC4`` , ``CV_16SC1`` , ``CV_16SC2`` , ``CV_16SC3`` , ``CV_32SC1`` , ``CV_32FC1``  source types are supported.
-
-    :param dst: Destination image with the same size and type as  ``src`` .
-
-    :param ksize: Gaussian kernel size.  ``ksize.width``  and  ``ksize.height``  can differ but they both must be positive and odd. If they are zeros, they are computed from  ``sigma1``  and  ``sigma2`` .
-
-    :param sigma1: Gaussian kernel standard deviation in X direction.
-
-    :param sigma2: Gaussian kernel standard deviation in Y direction. If  ``sigma2``  is zero, it is set to be equal to  ``sigma1`` . If they are both zeros, they are computed from  ``ksize.width``  and  ``ksize.height``, respectively. See  :ocv:func:`getGaussianKernel` for details. To fully control the result regardless of possible future modification of all this semantics, you are recommended to specify all of  ``ksize`` , ``sigma1`` , and  ``sigma2`` .
-
-    :param rowBorderType: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
-
-    :param columnBorderType: Pixel extrapolation method in the horizontal direction.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::createGaussianFilter_GPU`, :ocv:func:`GaussianBlur`
-
-
-
-gpu::getMaxFilter_GPU
--------------------------
-Creates the maximum filter.
-
-.. ocv:function:: Ptr<BaseFilter_GPU> gpu::getMaxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1))
-
-    :param srcType: Input image type. Only  ``CV_8UC1``  and  ``CV_8UC4`` are supported.
-
-    :param dstType: Output image type. It supports only the same type as the source type.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
-
-
-
-gpu::getMinFilter_GPU
--------------------------
-Creates the minimum filter.
-
-.. ocv:function:: Ptr<BaseFilter_GPU> gpu::getMinFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1))
-
-    :param srcType: Input image type. Only  ``CV_8UC1``  and  ``CV_8UC4`` are supported.
-
-    :param dstType: Output image type. It supports only the same type as the source type.
-
-    :param ksize: Kernel size.
-
-    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
-
-.. note:: This filter does not check out-of-border accesses, so only a proper sub-matrix of a bigger matrix has to be passed to it.
diff --git a/modules/gpu/doc/image_processing.rst b/modules/gpu/doc/image_processing.rst
deleted file mode 100644
index 7b404c832a..0000000000
--- a/modules/gpu/doc/image_processing.rst
+++ /dev/null
@@ -1,1087 +0,0 @@
-Image Processing
-================
-
-.. highlight:: cpp
-
-
-
-gpu::meanShiftFiltering
----------------------------
-Performs mean-shift filtering for each point of the source image.
-
-.. ocv:function:: void gpu::meanShiftFiltering( const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria=TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1), Stream& stream=Stream::Null() )
-
-    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
-
-    :param dst: Destination image containing the color of mapped points. It has the same size and type as  ``src`` .
-
-    :param sp: Spatial window radius.
-
-    :param sr: Color window radius.
-
-    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
-
-It maps each point of the source image into another point. As a result, you have a new color and new position of each point.
-
-
-
-gpu::meanShiftProc
-----------------------
-Performs a mean-shift procedure and stores information about processed points (their colors and positions) in two images.
-
-.. ocv:function:: void gpu::meanShiftProc( const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria=TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1), Stream& stream=Stream::Null() )
-
-    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
-
-    :param dstr: Destination image containing the color of mapped points. The size and type is the same as  ``src`` .
-
-    :param dstsp: Destination image containing the position of mapped points. The size is the same as  ``src`` size. The type is  ``CV_16SC2`` .
-
-    :param sp: Spatial window radius.
-
-    :param sr: Color window radius.
-
-    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
-
-.. seealso:: :ocv:func:`gpu::meanShiftFiltering`
-
-
-
-gpu::meanShiftSegmentation
-------------------------------
-Performs a mean-shift segmentation of the source image and eliminates small segments.
-
-.. ocv:function:: void gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize, TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1))
-
-    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
-
-    :param dst: Segmented image with the same size and type as  ``src`` .
-
-    :param sp: Spatial window radius.
-
-    :param sr: Color window radius.
-
-    :param minsize: Minimum segment size. Smaller segments are merged.
-
-    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
-
-
-
-gpu::integral
------------------
-Computes an integral image.
-
-.. ocv:function:: void gpu::integral(const GpuMat& src, GpuMat& sum, Stream& stream = Stream::Null())
-
-    :param src: Source image. Only  ``CV_8UC1`` images are supported for now.
-
-    :param sum: Integral image containing 32-bit unsigned integer values packed into  ``CV_32SC1`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`integral`
-
-
-
-gpu::sqrIntegral
---------------------
-Computes a squared integral image.
-
-.. ocv:function:: void gpu::sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& stream = Stream::Null())
-
-    :param src: Source image. Only  ``CV_8UC1`` images are supported for now.
-
-    :param sqsum: Squared integral image containing 64-bit unsigned integer values packed into  ``CV_64FC1`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::columnSum
-------------------
-Computes a vertical (column) sum.
-
-.. ocv:function:: void gpu::columnSum(const GpuMat& src, GpuMat& sum)
-
-    :param src: Source image. Only  ``CV_32FC1`` images are supported for now.
-
-    :param sum: Destination image of the  ``CV_32FC1`` type.
-
-
-
-gpu::cornerHarris
----------------------
-Computes the Harris cornerness criteria at each image pixel.
-
-.. ocv:function:: void gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType=BORDER_REFLECT101)
-
-    :param src: Source image. Only  ``CV_8UC1`` and  ``CV_32FC1`` images are supported for now.
-
-    :param dst: Destination image containing cornerness values. It has the same size as ``src`` and ``CV_32FC1`` type.
-
-    :param blockSize: Neighborhood size.
-
-    :param ksize: Aperture parameter for the Sobel operator.
-
-    :param k: Harris detector free parameter.
-
-    :param borderType: Pixel extrapolation method. Only  ``BORDER_REFLECT101`` and  ``BORDER_REPLICATE`` are supported for now.
-
-.. seealso:: :ocv:func:`cornerHarris`
-
-
-
-gpu::cornerMinEigenVal
---------------------------
-Computes the minimum eigen value of a 2x2 derivative covariation matrix at each pixel (the cornerness criteria).
-
-.. ocv:function:: void gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType=BORDER_REFLECT101)
-
-.. ocv:function:: void gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType=BORDER_REFLECT101)
-
-.. ocv:function:: void gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize, int borderType=BORDER_REFLECT101, Stream& stream = Stream::Null())
-
-    :param src: Source image. Only  ``CV_8UC1`` and  ``CV_32FC1`` images are supported for now.
-
-    :param dst: Destination image containing cornerness values. The size is the same. The type is  ``CV_32FC1`` .
-
-    :param blockSize: Neighborhood size.
-
-    :param ksize: Aperture parameter for the Sobel operator.
-
-    :param borderType: Pixel extrapolation method. Only ``BORDER_REFLECT101`` and ``BORDER_REPLICATE`` are supported for now.
-
-.. seealso:: :ocv:func:`cornerMinEigenVal`
-
-
-
-gpu::mulSpectrums
----------------------
-Performs a per-element multiplication of two Fourier spectrums.
-
-.. ocv:function:: void gpu::mulSpectrums( const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, bool conjB=false, Stream& stream=Stream::Null() )
-
-    :param a: First spectrum.
-
-    :param b: Second spectrum with the same size and type as  ``a`` .
-
-    :param c: Destination spectrum.
-
-    :param flags: Mock parameter used for CPU/GPU interfaces similarity.
-
-    :param conjB: Optional flag to specify if the second spectrum needs to be conjugated before the multiplication.
-
-    Only full (not packed) ``CV_32FC2`` complex spectrums in the interleaved format are supported for now.
-
-.. seealso:: :ocv:func:`mulSpectrums`
-
-
-
-gpu::mulAndScaleSpectrums
------------------------------
-Performs a per-element multiplication of two Fourier spectrums and scales the result.
-
-.. ocv:function:: void gpu::mulAndScaleSpectrums( const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, float scale, bool conjB=false, Stream& stream=Stream::Null() )
-
-    :param a: First spectrum.
-
-    :param b: Second spectrum with the same size and type as  ``a`` .
-
-    :param c: Destination spectrum.
-
-    :param flags: Mock parameter used for CPU/GPU interfaces similarity.
-
-    :param scale: Scale constant.
-
-    :param conjB: Optional flag to specify if the second spectrum needs to be conjugated before the multiplication.
-
-    Only full (not packed) ``CV_32FC2`` complex spectrums in the interleaved format are supported for now.
-
-.. seealso:: :ocv:func:`mulSpectrums`
-
-
-
-gpu::dft
-------------
-Performs a forward or inverse discrete Fourier transform (1D or 2D) of the floating point matrix.
-
-.. ocv:function:: void gpu::dft( const GpuMat& src, GpuMat& dst, Size dft_size, int flags=0, Stream& stream=Stream::Null() )
-
-    :param src: Source matrix (real or complex).
-
-    :param dst: Destination matrix (real or complex).
-
-    :param dft_size: Size of a discrete Fourier transform.
-
-    :param flags: Optional flags:
-
-        * **DFT_ROWS** transforms each individual row of the source matrix.
-
-        * **DFT_SCALE** scales the result: divide it by the number of elements in the transform (obtained from  ``dft_size`` ).
-
-        * **DFT_INVERSE** inverts DFT. Use for complex-complex cases (real-complex and complex-real cases are always forward and inverse, respectively).
-
-        * **DFT_REAL_OUTPUT** specifies the output as real. The source matrix is the result of real-complex transform, so the destination matrix must be real.
-
-Use to handle real matrices ( ``CV32FC1`` ) and complex matrices in the interleaved format ( ``CV32FC2`` ).
-
-The source matrix should be continuous, otherwise reallocation and data copying is performed. The function chooses an operation mode depending on the flags, size, and channel count of the source matrix:
-
-    * If the source matrix is complex and the output is not specified as real, the destination matrix is complex and has the ``dft_size``    size and ``CV_32FC2``    type. The destination matrix contains a full result of the DFT (forward or inverse).
-
-    * If the source matrix is complex and the output is specified as real, the function assumes that its input is the result of the forward transform (see the next item). The destination matrix has the ``dft_size`` size and ``CV_32FC1`` type. It contains the result of the inverse DFT.
-
-    * If the source matrix is real (its type is ``CV_32FC1`` ), forward DFT is performed. The result of the DFT is packed into complex ( ``CV_32FC2`` ) matrix. So, the width of the destination matrix is ``dft_size.width / 2 + 1`` . But if the source is a single column, the height is reduced instead of the width.
-
-.. seealso:: :ocv:func:`dft`
-
-
-gpu::ConvolveBuf
-----------------
-.. ocv:struct:: gpu::ConvolveBuf
-
-Class providing a memory buffer for :ocv:func:`gpu::convolve` function, plus it allows to adjust some specific parameters. ::
-
-    struct CV_EXPORTS ConvolveBuf
-    {
-        Size result_size;
-        Size block_size;
-        Size user_block_size;
-        Size dft_size;
-        int spect_len;
-
-        GpuMat image_spect, templ_spect, result_spect;
-        GpuMat image_block, templ_block, result_data;
-
-        void create(Size image_size, Size templ_size);
-        static Size estimateBlockSize(Size result_size, Size templ_size);
-    };
-
-You can use field `user_block_size` to set specific block size for :ocv:func:`gpu::convolve` function. If you leave its default value `Size(0,0)` then automatic estimation of block size will be used (which is optimized for speed). By varying `user_block_size` you can reduce memory requirements at the cost of speed.
-
-gpu::ConvolveBuf::create
-------------------------
-.. ocv:function:: gpu::ConvolveBuf::create(Size image_size, Size templ_size)
-
-Constructs a buffer for :ocv:func:`gpu::convolve` function with respective arguments.
-
-
-gpu::convolve
------------------
-Computes a convolution (or cross-correlation) of two images.
-
-.. ocv:function:: void gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr=false)
-
-.. ocv:function:: void gpu::convolve( const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr, ConvolveBuf& buf, Stream& stream=Stream::Null() )
-
-    :param image: Source image. Only  ``CV_32FC1`` images are supported for now.
-
-    :param templ: Template image. The size is not greater than the  ``image`` size. The type is the same as  ``image`` .
-
-    :param result: Result image. If  ``image`` is  *W x H*  and ``templ`` is  *w x h*, then  ``result`` must be *W-w+1 x H-h+1*.
-
-    :param ccorr: Flags to evaluate cross-correlation instead of convolution.
-
-    :param buf: Optional buffer to avoid extra memory allocations and to adjust some specific parameters. See :ocv:struct:`gpu::ConvolveBuf`.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::filter2D`
-
-gpu::MatchTemplateBuf
----------------------
-.. ocv:struct:: gpu::MatchTemplateBuf
-
-Class providing memory buffers for :ocv:func:`gpu::matchTemplate` function, plus it allows to adjust some specific parameters. ::
-
-    struct CV_EXPORTS MatchTemplateBuf
-    {
-        Size user_block_size;
-        GpuMat imagef, templf;
-        std::vector<GpuMat> images;
-        std::vector<GpuMat> image_sums;
-        std::vector<GpuMat> image_sqsums;
-    };
-
-You can use field `user_block_size` to set specific block size for :ocv:func:`gpu::matchTemplate` function. If you leave its default value `Size(0,0)` then automatic estimation of block size will be used (which is optimized for speed). By varying `user_block_size` you can reduce memory requirements at the cost of speed.
-
-gpu::matchTemplate
-----------------------
-Computes a proximity map for a raster template and an image where the template is searched for.
-
-.. ocv:function:: void gpu::matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, Stream &stream = Stream::Null())
-
-.. ocv:function:: void gpu::matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, MatchTemplateBuf &buf, Stream& stream = Stream::Null())
-
-    :param image: Source image.  ``CV_32F`` and  ``CV_8U`` depth images (1..4 channels) are supported for now.
-
-    :param templ: Template image with the size and type the same as  ``image`` .
-
-    :param result: Map containing comparison results ( ``CV_32FC1`` ). If  ``image`` is  *W x H*  and ``templ`` is  *w x h*, then  ``result`` must be *W-w+1 x H-h+1*.
-
-    :param method: Specifies the way to compare the template with the image.
-
-    :param buf: Optional buffer to avoid extra memory allocations and to adjust some specific parameters. See :ocv:struct:`gpu::MatchTemplateBuf`.
-
-    :param stream: Stream for the asynchronous version.
-
-    The following methods are supported for the ``CV_8U`` depth images for now:
-
-    * ``CV_TM_SQDIFF``
-    * ``CV_TM_SQDIFF_NORMED``
-    * ``CV_TM_CCORR``
-    * ``CV_TM_CCORR_NORMED``
-    * ``CV_TM_CCOEFF``
-    * ``CV_TM_CCOEFF_NORMED``
-
-    The following methods are supported for the ``CV_32F`` images for now:
-
-    * ``CV_TM_SQDIFF``
-    * ``CV_TM_CCORR``
-
-.. seealso:: :ocv:func:`matchTemplate`
-
-
-gpu::remap
---------------
-Applies a generic geometrical transformation to an image.
-
-.. ocv:function:: void gpu::remap( const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
-
-    :param src: Source image.
-
-    :param dst: Destination image with the size the same as  ``xmap`` and the type the same as  ``src`` .
-
-    :param xmap: X values. Only  ``CV_32FC1`` type is supported.
-
-    :param ymap: Y values. Only  ``CV_32FC1`` type is supported.
-
-    :param interpolation: Interpolation method (see  :ocv:func:`resize` ). ``INTER_NEAREST`` , ``INTER_LINEAR`` and ``INTER_CUBIC`` are supported for now.
-
-    :param borderMode: Pixel extrapolation method (see  :ocv:func:`borderInterpolate` ). ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
-
-    :param borderValue: Value used in case of a constant border. By default, it is 0.
-
-    :param stream: Stream for the asynchronous version.
-
-The function transforms the source image using the specified map:
-
-.. math::
-
-    \texttt{dst} (x,y) =  \texttt{src} (xmap(x,y), ymap(x,y))
-
-Values of pixels with non-integer coordinates are computed using the bilinear interpolation.
-
-.. seealso:: :ocv:func:`remap`
-
-
-
-gpu::cvtColor
------------------
-Converts an image from one color space to another.
-
-.. ocv:function:: void gpu::cvtColor(const GpuMat& src, GpuMat& dst, int code, int dcn = 0, Stream& stream = Stream::Null())
-
-    :param src: Source image with  ``CV_8U`` , ``CV_16U`` , or  ``CV_32F`` depth and 1, 3, or 4 channels.
-
-    :param dst: Destination image with the same size and depth as  ``src`` .
-
-    :param code: Color space conversion code. For details, see  :ocv:func:`cvtColor` . Conversion to/from Luv and Bayer color spaces is not supported.
-
-    :param dcn: Number of channels in the destination image. If the parameter is 0, the number of the channels is derived automatically from  ``src`` and the  ``code`` .
-
-    :param stream: Stream for the asynchronous version.
-
-3-channel color spaces (like ``HSV``, ``XYZ``, and so on) can be stored in a 4-channel image for better performance.
-
-.. seealso:: :ocv:func:`cvtColor`
-
-
-
-gpu::swapChannels
------------------
-Exchanges the color channels of an image in-place.
-
-.. ocv:function:: void gpu::swapChannels(GpuMat& image, const int dstOrder[4], Stream& stream = Stream::Null())
-
-    :param image: Source image. Supports only ``CV_8UC4`` type.
-
-    :param dstOrder: Integer array describing how channel values are permutated. The n-th entry of the array contains the number of the channel that is stored in the n-th channel of the output image. E.g. Given an RGBA image, aDstOrder = [3,2,1,0] converts this to ABGR channel order.
-
-    :param stream: Stream for the asynchronous version.
-
-The methods support arbitrary permutations of the original channels, including replication.
-
-
-
-gpu::threshold
-------------------
-Applies a fixed-level threshold to each array element.
-
-.. ocv:function:: double gpu::threshold(const GpuMat& src, GpuMat& dst, double thresh, double maxval, int type, Stream& stream = Stream::Null())
-
-    :param src: Source array (single-channel).
-
-    :param dst: Destination array with the same size and type as  ``src`` .
-
-    :param thresh: Threshold value.
-
-    :param maxval: Maximum value to use with  ``THRESH_BINARY`` and  ``THRESH_BINARY_INV`` threshold types.
-
-    :param type: Threshold type. For details, see  :ocv:func:`threshold` . The ``THRESH_OTSU`` threshold type is not supported.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`threshold`
-
-
-
-gpu::resize
----------------
-Resizes an image.
-
-.. ocv:function:: void gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
-
-    :param src: Source image.
-
-    :param dst: Destination image  with the same type as  ``src`` . The size is ``dsize`` (when it is non-zero) or the size is computed from  ``src.size()`` , ``fx`` , and  ``fy`` .
-
-    :param dsize: Destination image size. If it is zero, it is computed as:
-
-        .. math::
-            \texttt{dsize = Size(round(fx*src.cols), round(fy*src.rows))}
-
-        Either  ``dsize`` or both  ``fx`` and  ``fy`` must be non-zero.
-
-    :param fx: Scale factor along the horizontal axis. If it is zero, it is computed as:
-
-        .. math::
-
-            \texttt{(double)dsize.width/src.cols}
-
-    :param fy: Scale factor along the vertical axis. If it is zero, it is computed as:
-
-        .. math::
-
-            \texttt{(double)dsize.height/src.rows}
-
-    :param interpolation: Interpolation method. ``INTER_NEAREST`` , ``INTER_LINEAR`` and ``INTER_CUBIC`` are supported for now.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`resize`
-
-
-
-gpu::warpAffine
--------------------
-Applies an affine transformation to an image.
-
-.. ocv:function:: void gpu::warpAffine( const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags=INTER_LINEAR, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
-
-    :param src: Source image.  ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
-
-    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
-
-    :param M: *2x3*  transformation matrix.
-
-    :param dsize: Size of the destination image.
-
-    :param flags: Combination of interpolation methods (see  :ocv:func:`resize`) and the optional flag  ``WARP_INVERSE_MAP`` specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ). Only ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` interpolation methods are supported.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`warpAffine`
-
-
-
-gpu::buildWarpAffineMaps
-------------------------
-Builds transformation maps for affine transformation.
-
-.. ocv:function:: void gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null())
-
-    :param M: *2x3*  transformation matrix.
-
-    :param inverse: Flag  specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ).
-
-    :param dsize: Size of the destination image.
-
-    :param xmap: X values with  ``CV_32FC1`` type.
-
-    :param ymap: Y values with  ``CV_32FC1`` type.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::warpAffine` , :ocv:func:`gpu::remap`
-
-
-
-gpu::warpPerspective
-------------------------
-Applies a perspective transformation to an image.
-
-.. ocv:function:: void gpu::warpPerspective( const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags=INTER_LINEAR, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
-
-    :param src: Source image. ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
-
-    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
-
-    :param M: *3x3* transformation matrix.
-
-    :param dsize: Size of the destination image.
-
-    :param flags: Combination of interpolation methods (see  :ocv:func:`resize` ) and the optional flag  ``WARP_INVERSE_MAP`` specifying that  ``M`` is the inverse transformation ( ``dst => src`` ). Only  ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` interpolation methods are supported.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`warpPerspective`
-
-
-
-gpu::buildWarpPerspectiveMaps
------------------------------
-Builds transformation maps for perspective transformation.
-
-.. ocv:function:: void gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null())
-
-    :param M: *3x3*  transformation matrix.
-
-    :param inverse: Flag  specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ).
-
-    :param dsize: Size of the destination image.
-
-    :param xmap: X values with  ``CV_32FC1`` type.
-
-    :param ymap: Y values with  ``CV_32FC1`` type.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::warpPerspective` , :ocv:func:`gpu::remap`
-
-
-
-gpu::rotate
----------------
-Rotates an image around the origin (0,0) and then shifts it.
-
-.. ocv:function:: void gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift = 0, double yShift = 0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
-
-    :param src: Source image. Supports 1, 3 or 4 channels images with ``CV_8U`` , ``CV_16U`` or ``CV_32F`` depth.
-
-    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
-
-    :param dsize: Size of the destination image.
-
-    :param angle: Angle of rotation in degrees.
-
-    :param xShift: Shift along the horizontal axis.
-
-    :param yShift: Shift along the vertical axis.
-
-    :param interpolation: Interpolation method. Only  ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` are supported.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`gpu::warpAffine`
-
-
-
-gpu::copyMakeBorder
------------------------
-Forms a border around an image.
-
-.. ocv:function:: void gpu::copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, int borderType, const Scalar& value = Scalar(), Stream& stream = Stream::Null())
-
-    :param src: Source image. ``CV_8UC1`` , ``CV_8UC4`` , ``CV_32SC1`` , and  ``CV_32FC1`` types are supported.
-
-    :param dst: Destination image with the same type as  ``src``. The size is  ``Size(src.cols+left+right, src.rows+top+bottom)`` .
-
-    :param top:
-
-    :param bottom:
-
-    :param left:
-
-    :param right: Number of pixels in each direction from the source image rectangle to extrapolate. For example:  ``top=1, bottom=1, left=1, right=1`` mean that 1 pixel-wide border needs to be built.
-
-    :param borderType: Border type. See  :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
-
-    :param value: Border value.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`copyMakeBorder`
-
-
-
-gpu::rectStdDev
--------------------
-Computes a standard deviation of integral images.
-
-.. ocv:function:: void gpu::rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& stream = Stream::Null())
-
-    :param src: Source image. Only the ``CV_32SC1`` type is supported.
-
-    :param sqr: Squared source image. Only  the ``CV_32FC1`` type is supported.
-
-    :param dst: Destination image with the same type and size as  ``src`` .
-
-    :param rect: Rectangular window.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::evenLevels
--------------------
-Computes levels with even distribution.
-
-.. ocv:function:: void gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel)
-
-    :param levels: Destination array.  ``levels`` has 1 row, ``nLevels`` columns, and the ``CV_32SC1`` type.
-
-    :param nLevels: Number of computed levels.  ``nLevels`` must be at least 2.
-
-    :param lowerLevel: Lower boundary value of the lowest level.
-
-    :param upperLevel: Upper boundary value of the greatest level.
-
-
-
-gpu::histEven
------------------
-Calculates a histogram with evenly distributed bins.
-
-.. ocv:function:: void gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::histEven( const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::histEven( const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream=Stream::Null() )
-
-    :param src: Source image. ``CV_8U``, ``CV_16U``, or ``CV_16S`` depth and 1 or 4 channels are supported. For a four-channel image, all channels are processed separately.
-
-    :param hist: Destination histogram with one row, ``histSize`` columns, and the ``CV_32S`` type.
-
-    :param histSize: Size of the histogram.
-
-    :param lowerLevel: Lower boundary of lowest-level bin.
-
-    :param upperLevel: Upper boundary of highest-level bin.
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::histRange
-------------------
-Calculates a histogram with bins determined by the ``levels`` array.
-
-.. ocv:function:: void gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream = Stream::Null())
-
-    :param src: Source image. ``CV_8U`` , ``CV_16U`` , or  ``CV_16S`` depth and 1 or 4 channels are supported. For a four-channel image, all channels are processed separately.
-
-    :param hist: Destination histogram with one row, ``(levels.cols-1)`` columns, and the  ``CV_32SC1`` type.
-
-    :param levels: Number of levels in the histogram.
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::calcHist
-------------------
-Calculates histogram for one channel 8-bit image.
-
-.. ocv:function:: void gpu::calcHist(const GpuMat& src, GpuMat& hist, Stream& stream = Stream::Null())
-
-    :param src: Source image.
-
-    :param hist: Destination histogram with one row, 256 columns, and the  ``CV_32SC1`` type.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::equalizeHist
-------------------
-Equalizes the histogram of a grayscale image.
-
-.. ocv:function:: void gpu::equalizeHist(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, GpuMat& buf, Stream& stream = Stream::Null())
-
-    :param src: Source image.
-
-    :param dst: Destination image.
-
-    :param hist: Destination histogram with one row, 256 columns, and the  ``CV_32SC1`` type.
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`equalizeHist`
-
-
-
-gpu::buildWarpPlaneMaps
------------------------
-Builds plane warping maps.
-
-.. ocv:function:: void gpu::buildWarpPlaneMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, const Mat & T, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::buildWarpCylindricalMaps
------------------------------
-Builds cylindrical warping maps.
-
-.. ocv:function:: void gpu::buildWarpCylindricalMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::buildWarpSphericalMaps
----------------------------
-Builds spherical warping maps.
-
-.. ocv:function:: void gpu::buildWarpSphericalMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::pyrDown
--------------------
-Smoothes an image and downsamples it.
-
-.. ocv:function:: void gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source image.
-
-    :param dst: Destination image. Will have ``Size((src.cols+1)/2, (src.rows+1)/2)`` size and the same type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`pyrDown`
-
-
-
-gpu::pyrUp
--------------------
-Upsamples an image and then smoothes it.
-
-.. ocv:function:: void gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source image.
-
-    :param dst: Destination image. Will have ``Size(src.cols*2, src.rows*2)`` size and the same type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`pyrUp`
-
-
-
-gpu::blendLinear
--------------------
-Performs linear blending of two images.
-
-.. ocv:function:: void gpu::blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat& weights1, const GpuMat& weights2, GpuMat& result, Stream& stream = Stream::Null())
-
-    :param img1: First image. Supports only ``CV_8U`` and ``CV_32F`` depth.
-
-    :param img2: Second image. Must have the same size and the same type as ``img1`` .
-
-    :param weights1: Weights for first image. Must have tha same size as ``img1`` . Supports only ``CV_32F`` type.
-
-    :param weights2: Weights for second image. Must have tha same size as ``img2`` . Supports only ``CV_32F`` type.
-
-    :param result: Destination image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-gpu::bilateralFilter
---------------------
-Performs bilateral filtering of passed image
-
-.. ocv:function:: void gpu::bilateralFilter( const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode=BORDER_DEFAULT, Stream& stream=Stream::Null() )
-
-    :param src: Source image. Supports only (channles != 2 && depth() != CV_8S && depth() != CV_32S && depth() != CV_64F).
-
-    :param dst: Destination imagwe.
-
-    :param kernel_size: Kernel window size.
-
-    :param sigma_color: Filter sigma in the color space.
-
-    :param sigma_spatial:  Filter sigma in the coordinate space.
-
-    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso::
-
-    :ocv:func:`bilateralFilter`,
-
-
-gpu::nonLocalMeans
--------------------
-Performs pure non local means denoising without any simplification, and thus it is not fast.
-
-.. ocv:function:: void gpu::nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null())
-
-    :param src: Source image. Supports only CV_8UC1, CV_8UC2 and CV_8UC3.
-
-    :param dst: Destination image.
-
-    :param h: Filter sigma regulating filter strength for color.
-
-    :param search_window: Size of search window.
-
-    :param block_size: Size of block used for computing weights.
-
-    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso::
-
-    :ocv:func:`fastNlMeansDenoising`
-
-gpu::FastNonLocalMeansDenoising
--------------------------------
-.. ocv:class:: gpu::FastNonLocalMeansDenoising
-
-    ::
-
-        class FastNonLocalMeansDenoising
-        {
-        public:
-            //! Simple method, recommended for grayscale images (though it supports multichannel images)
-            void simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
-            //! Processes luminance and color components separatelly
-            void labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
-        };
-
-The class implements fast approximate Non Local Means Denoising algorithm.
-
-gpu::FastNonLocalMeansDenoising::simpleMethod()
------------------------------------------------
-Perform image denoising using Non-local Means Denoising algorithm http://www.ipol.im/pub/algo/bcm_non_local_means_denoising with several computational optimizations. Noise expected to be a gaussian white noise
-
-.. ocv:function:: void gpu::FastNonLocalMeansDenoising::simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
-
-    :param src: Input 8-bit 1-channel, 2-channel or 3-channel image.
-
-    :param dst: Output image with the same size and type as  ``src`` .
-
-    :param h: Parameter regulating filter strength. Big h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise
-
-    :param search_window: Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater search_window - greater denoising time. Recommended value 21 pixels
-
-    :param block_size: Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels
-
-    :param stream: Stream for the asynchronous invocations.
-
-This function expected to be applied to grayscale images. For colored images look at ``FastNonLocalMeansDenoising::labMethod``.
-
-.. seealso::
-
-    :ocv:func:`fastNlMeansDenoising`
-
-gpu::FastNonLocalMeansDenoising::labMethod()
---------------------------------------------
-Modification of ``FastNonLocalMeansDenoising::simpleMethod`` for color images
-
-.. ocv:function:: void gpu::FastNonLocalMeansDenoising::labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
-
-    :param src: Input 8-bit 3-channel image.
-
-    :param dst: Output image with the same size and type as  ``src`` .
-
-    :param h_luminance: Parameter regulating filter strength. Big h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise
-
-    :param float: The same as h but for color components. For most images value equals 10 will be enought to remove colored noise and do not distort colors
-
-    :param search_window: Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater search_window - greater denoising time. Recommended value 21 pixels
-
-    :param block_size: Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels
-
-    :param stream: Stream for the asynchronous invocations.
-
-The function converts image to CIELAB colorspace and then separately denoise L and AB components with given h parameters using ``FastNonLocalMeansDenoising::simpleMethod`` function.
-
-.. seealso::
-
-    :ocv:func:`fastNlMeansDenoisingColored`
-
-gpu::alphaComp
--------------------
-Composites two images using alpha opacity values contained in each image.
-
-.. ocv:function:: void gpu::alphaComp(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, int alpha_op, Stream& stream = Stream::Null())
-
-    :param img1: First image. Supports ``CV_8UC4`` , ``CV_16UC4`` , ``CV_32SC4`` and ``CV_32FC4`` types.
-
-    :param img2: Second image. Must have the same size and the same type as ``img1`` .
-
-    :param dst: Destination image.
-
-    :param alpha_op: Flag specifying the alpha-blending operation:
-
-            * **ALPHA_OVER**
-            * **ALPHA_IN**
-            * **ALPHA_OUT**
-            * **ALPHA_ATOP**
-            * **ALPHA_XOR**
-            * **ALPHA_PLUS**
-            * **ALPHA_OVER_PREMUL**
-            * **ALPHA_IN_PREMUL**
-            * **ALPHA_OUT_PREMUL**
-            * **ALPHA_ATOP_PREMUL**
-            * **ALPHA_XOR_PREMUL**
-            * **ALPHA_PLUS_PREMUL**
-            * **ALPHA_PREMUL**
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::Canny
--------------------
-Finds edges in an image using the [Canny86]_ algorithm.
-
-.. ocv:function:: void gpu::Canny(const GpuMat& image, GpuMat& edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false)
-
-.. ocv:function:: void gpu::Canny(const GpuMat& image, CannyBuf& buf, GpuMat& edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false)
-
-.. ocv:function:: void gpu::Canny(const GpuMat& dx, const GpuMat& dy, GpuMat& edges, double low_thresh, double high_thresh, bool L2gradient = false)
-
-.. ocv:function:: void gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& edges, double low_thresh, double high_thresh, bool L2gradient = false)
-
-    :param image: Single-channel 8-bit input image.
-
-    :param dx: First derivative of image in the vertical direction. Support only ``CV_32S`` type.
-
-    :param dy: First derivative of image in the horizontal direction. Support only ``CV_32S`` type.
-
-    :param edges: Output edge map. It has the same size and type as  ``image`` .
-
-    :param low_thresh: First threshold for the hysteresis procedure.
-
-    :param high_thresh: Second threshold for the hysteresis procedure.
-
-    :param apperture_size: Aperture size for the  :ocv:func:`Sobel`  operator.
-
-    :param L2gradient: Flag indicating whether a more accurate  :math:`L_2`  norm  :math:`=\sqrt{(dI/dx)^2 + (dI/dy)^2}`  should be used to compute the image gradient magnitude ( ``L2gradient=true`` ), or a faster default  :math:`L_1`  norm  :math:`=|dI/dx|+|dI/dy|`  is enough ( ``L2gradient=false`` ).
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-.. seealso:: :ocv:func:`Canny`
-
-
-
-gpu::HoughLines
----------------
-Finds lines in a binary image using the classical Hough transform.
-
-.. ocv:function:: void gpu::HoughLines(const GpuMat& src, GpuMat& lines, float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096)
-
-.. ocv:function:: void gpu::HoughLines(const GpuMat& src, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096)
-
-    :param src: 8-bit, single-channel binary source image.
-
-    :param lines: Output vector of lines. Each line is represented by a two-element vector  :math:`(\rho, \theta)` .  :math:`\rho`  is the distance from the coordinate origin  :math:`(0,0)`  (top-left corner of the image).  :math:`\theta`  is the line rotation angle in radians ( :math:`0 \sim \textrm{vertical line}, \pi/2 \sim \textrm{horizontal line}` ).
-
-    :param rho: Distance resolution of the accumulator in pixels.
-
-    :param theta: Angle resolution of the accumulator in radians.
-
-    :param threshold: Accumulator threshold parameter. Only those lines are returned that get enough votes ( :math:`>\texttt{threshold}` ).
-
-    :param doSort: Performs lines sort by votes.
-
-    :param maxLines: Maximum number of output lines.
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-.. seealso:: :ocv:func:`HoughLines`
-
-
-
-gpu::HoughLinesDownload
------------------------
-Downloads results from :ocv:func:`gpu::HoughLines` to host memory.
-
-.. ocv:function:: void gpu::HoughLinesDownload(const GpuMat& d_lines, OutputArray h_lines, OutputArray h_votes = noArray())
-
-    :param d_lines: Result of :ocv:func:`gpu::HoughLines` .
-
-    :param h_lines: Output host array.
-
-    :param h_votes: Optional output array for line's votes.
-
-.. seealso:: :ocv:func:`gpu::HoughLines`
-
-
-
-gpu::HoughCircles
------------------
-Finds circles in a grayscale image using the Hough transform.
-
-.. ocv:function:: void gpu::HoughCircles(const GpuMat& src, GpuMat& circles, int method, float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096)
-
-.. ocv:function:: void gpu::HoughCircles(const GpuMat& src, GpuMat& circles, HoughCirclesBuf& buf, int method, float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096)
-
-    :param src: 8-bit, single-channel grayscale input image.
-
-    :param circles: Output vector of found circles. Each vector is encoded as a 3-element floating-point vector  :math:`(x, y, radius)` .
-
-    :param method: Detection method to use. Currently, the only implemented method is  ``CV_HOUGH_GRADIENT`` , which is basically  *21HT* , described in  [Yuen90]_.
-
-    :param dp: Inverse ratio of the accumulator resolution to the image resolution. For example, if  ``dp=1`` , the accumulator has the same resolution as the input image. If  ``dp=2`` , the accumulator has half as big width and height.
-
-    :param minDist: Minimum distance between the centers of the detected circles. If the parameter is too small, multiple neighbor circles may be falsely detected in addition to a true one. If it is too large, some circles may be missed.
-
-    :param cannyThreshold: The higher threshold of the two passed to  the :ocv:func:`gpu::Canny`  edge detector (the lower one is twice smaller).
-
-    :param votesThreshold: The accumulator threshold for the circle centers at the detection stage. The smaller it is, the more false circles may be detected.
-
-    :param minRadius: Minimum circle radius.
-
-    :param maxRadius: Maximum circle radius.
-
-    :param maxCircles: Maximum number of output circles.
-
-    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
-
-.. seealso:: :ocv:func:`HoughCircles`
-
-
-
-gpu::HoughCirclesDownload
--------------------------
-Downloads results from :ocv:func:`gpu::HoughCircles` to host memory.
-
-.. ocv:function:: void gpu::HoughCirclesDownload(const GpuMat& d_circles, OutputArray h_circles)
-
-    :param d_circles: Result of :ocv:func:`gpu::HoughCircles` .
-
-    :param h_circles: Output host array.
-
-.. seealso:: :ocv:func:`gpu::HoughCircles`
diff --git a/modules/gpu/doc/initalization_and_information.rst b/modules/gpu/doc/initalization_and_information.rst
index fc7236dcf1..abfc0860cf 100644
--- a/modules/gpu/doc/initalization_and_information.rst
+++ b/modules/gpu/doc/initalization_and_information.rst
@@ -107,23 +107,186 @@ Class providing functionality for querying the specified GPU properties. ::
     class CV_EXPORTS DeviceInfo
     {
     public:
+        //! creates DeviceInfo object for the current GPU
         DeviceInfo();
+
+        //! creates DeviceInfo object for the given GPU
         DeviceInfo(int device_id);
 
-        string name() const;
+        //! ASCII string identifying device
+        const char* name() const;
 
+        //! global memory available on device in bytes
+        size_t totalGlobalMem() const;
+
+        //! shared memory available per block in bytes
+        size_t sharedMemPerBlock() const;
+
+        //! 32-bit registers available per block
+        int regsPerBlock() const;
+
+        //! warp size in threads
+        int warpSize() const;
+
+        //! maximum pitch in bytes allowed by memory copies
+        size_t memPitch() const;
+
+        //! maximum number of threads per block
+        int maxThreadsPerBlock() const;
+
+        //! maximum size of each dimension of a block
+        Vec3i maxThreadsDim() const;
+
+        //! maximum size of each dimension of a grid
+        Vec3i maxGridSize() const;
+
+        //! clock frequency in kilohertz
+        int clockRate() const;
+
+        //! constant memory available on device in bytes
+        size_t totalConstMem() const;
+
+        //! major compute capability
         int majorVersion() const;
+
+        //! minor compute capability
         int minorVersion() const;
 
+        //! alignment requirement for textures
+        size_t textureAlignment() const;
+
+        //! pitch alignment requirement for texture references bound to pitched memory
+        size_t texturePitchAlignment() const;
+
+        //! number of multiprocessors on device
         int multiProcessorCount() const;
 
+        //! specified whether there is a run time limit on kernels
+        bool kernelExecTimeoutEnabled() const;
+
+        //! device is integrated as opposed to discrete
+        bool integrated() const;
+
+        //! device can map host memory with cudaHostAlloc/cudaHostGetDevicePointer
+        bool canMapHostMemory() const;
+
+        enum ComputeMode
+        {
+            ComputeModeDefault,         /**< default compute mode (Multiple threads can use ::cudaSetDevice() with this device) */
+            ComputeModeExclusive,       /**< compute-exclusive-thread mode (Only one thread in one process will be able to use ::cudaSetDevice() with this device) */
+            ComputeModeProhibited,      /**< compute-prohibited mode (No threads can use ::cudaSetDevice() with this device) */
+            ComputeModeExclusiveProcess /**< compute-exclusive-process mode (Many threads in one process will be able to use ::cudaSetDevice() with this device) */
+        };
+
+        //! compute mode
+        ComputeMode computeMode() const;
+
+        //! maximum 1D texture size
+        int maxTexture1D() const;
+
+        //! maximum 1D mipmapped texture size
+        int maxTexture1DMipmap() const;
+
+        //! maximum size for 1D textures bound to linear memory
+        int maxTexture1DLinear() const;
+
+        //! maximum 2D texture dimensions
+        Vec2i maxTexture2D() const;
+
+        //! maximum 2D mipmapped texture dimensions
+        Vec2i maxTexture2DMipmap() const;
+
+        //! maximum dimensions (width, height, pitch) for 2D textures bound to pitched memory
+        Vec3i maxTexture2DLinear() const;
+
+        //! maximum 2D texture dimensions if texture gather operations have to be performed
+        Vec2i maxTexture2DGather() const;
+
+        //! maximum 3D texture dimensions
+        Vec3i maxTexture3D() const;
+
+        //! maximum Cubemap texture dimensions
+        int maxTextureCubemap() const;
+
+        //! maximum 1D layered texture dimensions
+        Vec2i maxTexture1DLayered() const;
+
+        //! maximum 2D layered texture dimensions
+        Vec3i maxTexture2DLayered() const;
+
+        //! maximum Cubemap layered texture dimensions
+        Vec2i maxTextureCubemapLayered() const;
+
+        //! maximum 1D surface size
+        int maxSurface1D() const;
+
+        //! maximum 2D surface dimensions
+        Vec2i maxSurface2D() const;
+
+        //! maximum 3D surface dimensions
+        Vec3i maxSurface3D() const;
+
+        //! maximum 1D layered surface dimensions
+        Vec2i maxSurface1DLayered() const;
+
+        //! maximum 2D layered surface dimensions
+        Vec3i maxSurface2DLayered() const;
+
+        //! maximum Cubemap surface dimensions
+        int maxSurfaceCubemap() const;
+
+        //! maximum Cubemap layered surface dimensions
+        Vec2i maxSurfaceCubemapLayered() const;
+
+        //! alignment requirements for surfaces
+        size_t surfaceAlignment() const;
+
+        //! device can possibly execute multiple kernels concurrently
+        bool concurrentKernels() const;
+
+        //! device has ECC support enabled
+        bool ECCEnabled() const;
+
+        //! PCI bus ID of the device
+        int pciBusID() const;
+
+        //! PCI device ID of the device
+        int pciDeviceID() const;
+
+        //! PCI domain ID of the device
+        int pciDomainID() const;
+
+        //! true if device is a Tesla device using TCC driver, false otherwise
+        bool tccDriver() const;
+
+        //! number of asynchronous engines
+        int asyncEngineCount() const;
+
+        //! device shares a unified address space with the host
+        bool unifiedAddressing() const;
+
+        //! peak memory clock frequency in kilohertz
+        int memoryClockRate() const;
+
+        //! global memory bus width in bits
+        int memoryBusWidth() const;
+
+        //! size of L2 cache in bytes
+        int l2CacheSize() const;
+
+        //! maximum resident threads per multiprocessor
+        int maxThreadsPerMultiProcessor() const;
+
+        //! gets free and total device memory
+        void queryMemory(size_t& totalMemory, size_t& freeMemory) const;
         size_t freeMemory() const;
         size_t totalMemory() const;
 
-        bool supports(FeatureSet feature) const;
-        bool isCompatible() const;
+        //! checks whether device supports the given feature
+        bool supports(FeatureSet feature_set) const;
 
-        int deviceID() const;
+        //! checks whether the GPU module can be run on the given device
+        bool isCompatible() const;
     };
 
 
@@ -146,7 +309,7 @@ gpu::DeviceInfo::name
 ---------------------
 Returns the device name.
 
-.. ocv:function:: string gpu::DeviceInfo::name() const
+.. ocv:function:: const char* gpu::DeviceInfo::name() const
 
 
 
@@ -166,14 +329,6 @@ Returns the minor compute capability version.
 
 
 
-gpu::DeviceInfo::multiProcessorCount
-------------------------------------
-Returns the number of streaming multiprocessors.
-
-.. ocv:function:: int gpu::DeviceInfo::multiProcessorCount()
-
-
-
 gpu::DeviceInfo::freeMemory
 ---------------------------
 Returns the amount of free memory in bytes.
@@ -194,7 +349,7 @@ gpu::DeviceInfo::supports
 -------------------------
 Provides information on GPU feature support.
 
-.. ocv:function:: bool gpu::DeviceInfo::supports( FeatureSet feature_set ) const
+.. ocv:function:: bool gpu::DeviceInfo::supports(FeatureSet feature_set) const
 
     :param feature_set: Features to be checked. See :ocv:enum:`gpu::FeatureSet`.
 
diff --git a/modules/gpu/doc/introduction.rst b/modules/gpu/doc/introduction.rst
index ef34c369b9..a1237629c5 100644
--- a/modules/gpu/doc/introduction.rst
+++ b/modules/gpu/doc/introduction.rst
@@ -60,4 +60,3 @@ With this algorithm, a dual GPU gave a 180
 %
 performance increase comparing to the single Fermi GPU. For a source code example, see
 http://code.opencv.org/projects/opencv/repository/revisions/master/entry/samples/gpu/.
-
diff --git a/modules/gpu/doc/object_detection.rst b/modules/gpu/doc/object_detection.rst
index fd0ac867e2..52755608bf 100644
--- a/modules/gpu/doc/object_detection.rst
+++ b/modules/gpu/doc/object_detection.rst
@@ -62,7 +62,12 @@ The class implements Histogram of Oriented Gradients ([Dalal2005]_) object detec
 
 Interfaces of all methods are kept similar to the ``CPU HOG`` descriptor and detector analogues as much as possible.
 
+.. note::
 
+   * An example applying the HOG descriptor for people detection can be found at opencv_source_code/samples/cpp/peopledetect.cpp
+   * A GPU example applying the HOG descriptor for people detection can be found at opencv_source_code/samples/gpu/hog.cpp
+
+   * (Python) An example applying the HOG descriptor for people detection can be found at opencv_source_code/samples/python2/peopledetect.py
 
 gpu::HOGDescriptor::HOGDescriptor
 -------------------------------------
@@ -209,11 +214,11 @@ Cascade classifier class used for object detection. Supports HAAR and LBP cascad
     {
     public:
             CascadeClassifier_GPU();
-            CascadeClassifier_GPU(const string& filename);
+            CascadeClassifier_GPU(const String& filename);
             ~CascadeClassifier_GPU();
 
             bool empty() const;
-            bool load(const string& filename);
+            bool load(const String& filename);
             void release();
 
             /* Returns number of detected objects */
@@ -229,13 +234,16 @@ Cascade classifier class used for object detection. Supports HAAR and LBP cascad
             Size getClassifierSize() const;
     };
 
+.. note::
 
+   * A cascade classifier example can be found at opencv_source_code/samples/gpu/cascadeclassifier.cpp
+   * A Nvidea API specific cascade classifier example can be found at opencv_source_code/samples/gpu/cascadeclassifier_nvidia_api.cpp
 
 gpu::CascadeClassifier_GPU::CascadeClassifier_GPU
 -----------------------------------------------------
 Loads the classifier from a file. Cascade type is detected automatically by constructor parameter.
 
-.. ocv:function:: gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const string& filename)
+.. ocv:function:: gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const String& filename)
 
     :param filename: Name of the file from which the classifier is loaded. Only the old ``haar`` classifier (trained by the ``haar`` training application) and NVIDIA's ``nvbin`` are supported for HAAR and only new type of OpenCV XML cascade supported for LBP.
 
@@ -253,7 +261,7 @@ gpu::CascadeClassifier_GPU::load
 ------------------------------------
 Loads the classifier from a file. The previous content is destroyed.
 
-.. ocv:function:: bool gpu::CascadeClassifier_GPU::load(const string& filename)
+.. ocv:function:: bool gpu::CascadeClassifier_GPU::load(const String& filename)
 
     :param filename: Name of the file from which the classifier is loaded. Only the old ``haar`` classifier (trained by the ``haar`` training application) and NVIDIA's ``nvbin`` are supported for HAAR and only new type of OpenCV XML cascade supported for LBP.
 
diff --git a/modules/gpu/doc/operations_on_matrices.rst b/modules/gpu/doc/operations_on_matrices.rst
deleted file mode 100644
index d1762f442a..0000000000
--- a/modules/gpu/doc/operations_on_matrices.rst
+++ /dev/null
@@ -1,274 +0,0 @@
-Operations on Matrices
-======================
-
-.. highlight:: cpp
-
-
-
-gpu::gemm
-------------------
-Performs generalized matrix multiplication.
-
-.. ocv:function:: void gpu::gemm(const GpuMat& src1, const GpuMat& src2, double alpha, const GpuMat& src3, double beta, GpuMat& dst, int flags = 0, Stream& stream = Stream::Null())
-
-    :param src1: First multiplied input matrix that should have  ``CV_32FC1`` , ``CV_64FC1`` , ``CV_32FC2`` , or  ``CV_64FC2``  type.
-
-    :param src2: Second multiplied input matrix of the same type as  ``src1`` .
-
-    :param alpha: Weight of the matrix product.
-
-    :param src3: Third optional delta matrix added to the matrix product. It should have the same type as  ``src1``  and  ``src2`` .
-
-    :param beta: Weight of  ``src3`` .
-
-    :param dst: Destination matrix. It has the proper size and the same type as input matrices.
-
-    :param flags: Operation flags:
-
-            * **GEMM_1_T** transpose  ``src1``
-            * **GEMM_2_T** transpose  ``src2``
-            * **GEMM_3_T** transpose  ``src3``
-
-    :param stream: Stream for the asynchronous version.
-
-The function performs generalized matrix multiplication similar to the ``gemm`` functions in BLAS level 3. For example, ``gemm(src1, src2, alpha, src3, beta, dst, GEMM_1_T + GEMM_3_T)`` corresponds to
-
-.. math::
-
-    \texttt{dst} =  \texttt{alpha} \cdot \texttt{src1} ^T  \cdot \texttt{src2} +  \texttt{beta} \cdot \texttt{src3} ^T
-
-.. note:: Transposition operation doesn't support  ``CV_64FC2``  input type.
-
-.. seealso:: :ocv:func:`gemm`
-
-
-
-gpu::transpose
-------------------
-Transposes a matrix.
-
-.. ocv:function:: void gpu::transpose( const GpuMat& src1, GpuMat& dst, Stream& stream=Stream::Null() )
-
-    :param src1: Source matrix. 1-, 4-, 8-byte element sizes are supported for now (CV_8UC1, CV_8UC4, CV_16UC2, CV_32FC1, etc).
-
-    :param dst: Destination matrix.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`transpose`
-
-
-
-gpu::flip
--------------
-Flips a 2D matrix around vertical, horizontal, or both axes.
-
-.. ocv:function:: void gpu::flip( const GpuMat& a, GpuMat& b, int flipCode, Stream& stream=Stream::Null() )
-
-    :param a: Source matrix. Supports 1, 3 and 4 channels images with ``CV_8U``, ``CV_16U``, ``CV_32S`` or ``CV_32F`` depth.
-
-    :param b: Destination matrix.
-
-    :param flipCode: Flip mode for the source:
-
-        * ``0`` Flips around x-axis.
-
-        * ``>0`` Flips around y-axis.
-
-        * ``<0`` Flips around both axes.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`flip`
-
-
-
-gpu::LUT
-------------
-Transforms the source matrix into the destination matrix using the given look-up table: ``dst(I) = lut(src(I))``
-
-.. ocv:function:: void gpu::LUT(const GpuMat& src, const Mat& lut, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix.  ``CV_8UC1``  and  ``CV_8UC3``  matrices are supported for now.
-
-    :param lut: Look-up table of 256 elements. It is a continuous ``CV_8U`` matrix.
-
-    :param dst: Destination matrix with the same depth as  ``lut``  and the same number of channels as  ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`LUT`
-
-
-
-gpu::merge
---------------
-Makes a multi-channel matrix out of several single-channel matrices.
-
-.. ocv:function:: void gpu::merge(const GpuMat* src, size_t n, GpuMat& dst, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::merge(const vector<GpuMat>& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Array/vector of source matrices.
-
-    :param n: Number of source matrices.
-
-    :param dst: Destination matrix.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`merge`
-
-
-
-gpu::split
---------------
-Copies each plane of a multi-channel matrix into an array.
-
-.. ocv:function:: void gpu::split(const GpuMat& src, GpuMat* dst, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::split(const GpuMat& src, vector<GpuMat>& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix.
-
-    :param dst: Destination array/vector of single-channel matrices.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`split`
-
-
-
-gpu::magnitude
-------------------
-Computes magnitudes of complex matrix elements.
-
-.. ocv:function:: void gpu::magnitude( const GpuMat& xy, GpuMat& magnitude, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::magnitude(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null())
-
-    :param xy: Source complex matrix in the interleaved format ( ``CV_32FC2`` ).
-
-    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
-
-    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
-
-    :param magnitude: Destination matrix of float magnitudes ( ``CV_32FC1`` ).
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`magnitude`
-
-
-
-gpu::magnitudeSqr
----------------------
-Computes squared magnitudes of complex matrix elements.
-
-.. ocv:function:: void gpu::magnitudeSqr( const GpuMat& xy, GpuMat& magnitude, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::magnitudeSqr(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null())
-
-    :param xy: Source complex matrix in the interleaved format ( ``CV_32FC2`` ).
-
-    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
-
-    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
-
-    :param magnitude: Destination matrix of float magnitude squares ( ``CV_32FC1`` ).
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::phase
---------------
-Computes polar angles of complex matrix elements.
-
-.. ocv:function:: void gpu::phase(const GpuMat& x, const GpuMat& y, GpuMat& angle, bool angleInDegrees=false, Stream& stream = Stream::Null())
-
-    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
-
-    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
-
-    :param angle: Destination matrix of angles ( ``CV_32FC1`` ).
-
-    :param angleInDegrees: Flag for angles that must be evaluated in degrees.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`phase`
-
-
-
-gpu::cartToPolar
---------------------
-Converts Cartesian coordinates into polar.
-
-.. ocv:function:: void gpu::cartToPolar(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, GpuMat& angle, bool angleInDegrees=false, Stream& stream = Stream::Null())
-
-    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
-
-    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
-
-    :param magnitude: Destination matrix of float magnitudes ( ``CV_32FC1`` ).
-
-    :param angle: Destination matrix of angles ( ``CV_32FC1`` ).
-
-    :param angleInDegrees: Flag for angles that must be evaluated in degrees.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`cartToPolar`
-
-
-
-gpu::polarToCart
---------------------
-Converts polar coordinates into Cartesian.
-
-.. ocv:function:: void gpu::polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees=false, Stream& stream = Stream::Null())
-
-    :param magnitude: Source matrix containing magnitudes ( ``CV_32FC1`` ).
-
-    :param angle: Source matrix containing angles ( ``CV_32FC1`` ).
-
-    :param x: Destination matrix of real components ( ``CV_32FC1`` ).
-
-    :param y: Destination matrix of imaginary components ( ``CV_32FC1`` ).
-
-    :param angleInDegrees: Flag that indicates angles in degrees.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`polarToCart`
-
-
-
-gpu::normalize
---------------
-Normalizes the norm or value range of an array.
-
-.. ocv:function:: void gpu::normalize(const GpuMat& src, GpuMat& dst, double alpha = 1, double beta = 0, int norm_type = NORM_L2, int dtype = -1, const GpuMat& mask = GpuMat())
-
-.. ocv:function:: void gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf)
-
-    :param src: input array.
-
-    :param dst: output array of the same size as  ``src`` .
-
-    :param alpha: norm value to normalize to or the lower range boundary in case of the range normalization.
-
-    :param beta: upper range boundary in case of the range normalization; it is not used for the norm normalization.
-
-    :param normType: normalization type (see the details below).
-
-    :param dtype: when negative, the output array has the same type as ``src``; otherwise, it has the same number of channels as  ``src`` and the depth ``=CV_MAT_DEPTH(dtype)``.
-
-    :param mask: optional operation mask.
-
-    :param norm_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
-
-    :param cvt_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
-
-.. seealso:: :ocv:func:`normalize`
diff --git a/modules/gpu/doc/per_element_operations.rst b/modules/gpu/doc/per_element_operations.rst
deleted file mode 100644
index 2670ba3233..0000000000
--- a/modules/gpu/doc/per_element_operations.rst
+++ /dev/null
@@ -1,445 +0,0 @@
-Per-element Operations
-=======================
-
-.. highlight:: cpp
-
-
-
-gpu::add
-------------
-Computes a matrix-matrix or matrix-scalar sum.
-
-.. ocv:function:: void gpu::add( const GpuMat& a, const GpuMat& b, GpuMat& c, const GpuMat& mask=GpuMat(), int dtype=-1, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::add( const GpuMat& a, const Scalar& sc, GpuMat& c, const GpuMat& mask=GpuMat(), int dtype=-1, Stream& stream=Stream::Null() )
-
-    :param a: First source matrix.
-
-    :param b: Second source matrix to be added to ``a`` . Matrix should have the same size and type as ``a`` .
-
-    :param sc: A scalar to be added to ``a`` .
-
-    :param c: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``a`` depth.
-
-    :param mask: Optional operation mask, 8-bit single channel array, that specifies elements of the destination array to be changed.
-
-    :param dtype: Optional depth of the output array.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`add`
-
-
-
-gpu::subtract
------------------
-Computes a matrix-matrix or matrix-scalar difference.
-
-.. ocv:function:: void gpu::subtract( const GpuMat& a, const GpuMat& b, GpuMat& c, const GpuMat& mask=GpuMat(), int dtype=-1, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::subtract( const GpuMat& a, const Scalar& sc, GpuMat& c, const GpuMat& mask=GpuMat(), int dtype=-1, Stream& stream=Stream::Null() )
-
-    :param a: First source matrix.
-
-    :param b: Second source matrix to be added to ``a`` . Matrix should have the same size and type as ``a`` .
-
-    :param sc: A scalar to be added to ``a`` .
-
-    :param c: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``a`` depth.
-
-    :param mask: Optional operation mask, 8-bit single channel array, that specifies elements of the destination array to be changed.
-
-    :param dtype: Optional depth of the output array.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`subtract`
-
-
-
-gpu::multiply
------------------
-Computes a matrix-matrix or matrix-scalar per-element product.
-
-.. ocv:function:: void gpu::multiply( const GpuMat& a, const GpuMat& b, GpuMat& c, double scale=1, int dtype=-1, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::multiply( const GpuMat& a, const Scalar& sc, GpuMat& c, double scale=1, int dtype=-1, Stream& stream=Stream::Null() )
-
-    :param a: First source matrix.
-
-    :param b: Second source matrix to be multiplied by ``a`` elements.
-
-    :param sc: A scalar to be multiplied by ``a`` elements.
-
-    :param c: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``a`` depth.
-
-    :param scale: Optional scale factor.
-
-    :param dtype: Optional depth of the output array.
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`multiply`
-
-
-
-gpu::divide
------------
-Computes a matrix-matrix or matrix-scalar division.
-
-.. ocv:function:: void gpu::divide( const GpuMat& a, const GpuMat& b, GpuMat& c, double scale=1, int dtype=-1, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::divide(const GpuMat& a, const Scalar& sc, GpuMat& c, double scale = 1, int dtype = -1, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::divide( double scale, const GpuMat& b, GpuMat& c, int dtype=-1, Stream& stream=Stream::Null() )
-
-    :param a: First source matrix or a scalar.
-
-    :param b: Second source matrix. The ``a`` elements are divided by it.
-
-    :param sc: A scalar to be divided by the elements of ``a`` matrix.
-
-    :param c: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``a`` depth.
-
-    :param scale: Optional scale factor.
-
-    :param dtype: Optional depth of the output array.
-
-    :param stream: Stream for the asynchronous version.
-
-This function, in contrast to :ocv:func:`divide`, uses a round-down rounding mode.
-
-.. seealso:: :ocv:func:`divide`
-
-
-gpu::addWeighted
-----------------
-Computes the weighted sum of two arrays.
-
-.. ocv:function:: void gpu::addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2, double beta, double gamma, GpuMat& dst, int dtype = -1, Stream& stream = Stream::Null())
-
-    :param src1: First source array.
-
-    :param alpha: Weight for the first array elements.
-
-    :param src2: Second source array of the same size and channel number as  ``src1`` .
-
-    :param beta: Weight for the second array elements.
-
-    :param dst: Destination array that has the same size and number of channels as the input arrays.
-
-    :param gamma: Scalar added to each sum.
-
-    :param dtype: Optional depth of the destination array. When both input arrays have the same depth, ``dtype`` can be set to ``-1``, which will be equivalent to ``src1.depth()``.
-
-    :param stream: Stream for the asynchronous version.
-
-The function ``addWeighted`` calculates the weighted sum of two arrays as follows:
-
-.. math::
-
-    \texttt{dst} (I)= \texttt{saturate} ( \texttt{src1} (I)* \texttt{alpha} +  \texttt{src2} (I)* \texttt{beta} +  \texttt{gamma} )
-
-where ``I`` is a multi-dimensional index of array elements. In case of multi-channel arrays, each channel is processed independently.
-
-.. seealso:: :ocv:func:`addWeighted`
-
-
-
-gpu::abs
-------------
-Computes an absolute value of each matrix element.
-
-.. ocv:function:: void gpu::abs(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix. Supports ``CV_16S`` and ``CV_32F`` depth.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`abs`
-
-
-
-gpu::sqr
-------------
-Computes a square value of each matrix element.
-
-.. ocv:function:: void gpu::sqr(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix. Supports ``CV_8U`` , ``CV_16U`` , ``CV_16S`` and ``CV_32F`` depth.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::sqrt
-------------
-Computes a square root of each matrix element.
-
-.. ocv:function:: void gpu::sqrt(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix. Supports ``CV_8U`` , ``CV_16U`` , ``CV_16S`` and ``CV_32F`` depth.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`sqrt`
-
-
-
-gpu::exp
-------------
-Computes an exponent of each matrix element.
-
-.. ocv:function:: void gpu::exp( const GpuMat& a, GpuMat& b, Stream& stream=Stream::Null() )
-
-    :param a: Source matrix. Supports ``CV_8U`` , ``CV_16U`` , ``CV_16S`` and ``CV_32F`` depth.
-
-    :param b: Destination matrix with the same size and type as ``a`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`exp`
-
-
-
-gpu::log
-------------
-Computes a natural logarithm of absolute value of each matrix element.
-
-.. ocv:function:: void gpu::log( const GpuMat& a, GpuMat& b, Stream& stream=Stream::Null() )
-
-    :param a: Source matrix. Supports ``CV_8U`` , ``CV_16U`` , ``CV_16S`` and ``CV_32F`` depth.
-
-    :param b: Destination matrix with the same size and type as ``a`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`log`
-
-
-
-gpu::pow
-------------
-Raises every matrix element to a power.
-
-.. ocv:function:: void gpu::pow(const GpuMat& src, double power, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src: Source matrix. Supports all type, except ``CV_64F`` depth.
-
-    :param power: Exponent of power.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-The function ``pow`` raises every element of the input matrix to ``p`` :
-
-.. math::
-
-    \texttt{dst} (I) =  \fork{\texttt{src}(I)^p}{if \texttt{p} is integer}{|\texttt{src}(I)|^p}{otherwise}
-
-.. seealso:: :ocv:func:`pow`
-
-
-
-gpu::absdiff
-----------------
-Computes per-element absolute difference of two matrices (or of a matrix and scalar).
-
-.. ocv:function:: void gpu::absdiff( const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::absdiff( const GpuMat& a, const Scalar& s, GpuMat& c, Stream& stream=Stream::Null() )
-
-    :param a: First source matrix.
-
-    :param b: Second source matrix to be added to ``a`` .
-
-    :param s: A scalar to be added to ``a`` .
-
-    :param c: Destination matrix with the same size and type as ``a`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`absdiff`
-
-
-
-gpu::compare
-----------------
-Compares elements of two matrices.
-
-.. ocv:function:: void gpu::compare( const GpuMat& a, const GpuMat& b, GpuMat& c, int cmpop, Stream& stream=Stream::Null() )
-
-.. ocv:function:: void gpu::compare(const GpuMat& a, Scalar sc, GpuMat& c, int cmpop, Stream& stream = Stream::Null())
-
-    :param a: First source matrix.
-
-    :param b: Second source matrix with the same size and type as ``a`` .
-
-    :param sc: A scalar to be compared with ``a`` .
-
-    :param c: Destination matrix with the same size as ``a`` and the ``CV_8UC1`` type.
-
-    :param cmpop: Flag specifying the relation between the elements to be checked:
-
-            * **CMP_EQ:** ``a(.) == b(.)``
-            * **CMP_GT:** ``a(.) < b(.)``
-            * **CMP_GE:** ``a(.) <= b(.)``
-            * **CMP_LT:** ``a(.) < b(.)``
-            * **CMP_LE:** ``a(.) <= b(.)``
-            * **CMP_NE:** ``a(.) != b(.)``
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`compare`
-
-
-
-gpu::bitwise_not
---------------------
-Performs a per-element bitwise inversion.
-
-.. ocv:function:: void gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null())
-
-    :param src: Source matrix.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param mask: Optional operation mask. 8-bit single channel image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::bitwise_or
--------------------
-Performs a per-element bitwise disjunction of two matrices or of matrix and scalar.
-
-.. ocv:function:: void gpu::bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null())
-.. ocv:function:: void gpu::bitwise_or(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src1: First source matrix.
-
-    :param src2: Second source matrix with the same size and type as ``src1`` .
-
-    :param dst: Destination matrix with the same size and type as ``src1`` .
-
-    :param mask: Optional operation mask. 8-bit single channel image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::bitwise_and
---------------------
-Performs a per-element bitwise conjunction of two matrices or of matrix and scalar.
-
-.. ocv:function:: void gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null())
-.. ocv:function:: void gpu::bitwise_and(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src1: First source matrix.
-
-    :param src2: Second source matrix with the same size and type as ``src1`` .
-
-    :param dst: Destination matrix with the same size and type as ``src1`` .
-
-    :param mask: Optional operation mask. 8-bit single channel image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::bitwise_xor
---------------------
-Performs a per-element bitwise ``exclusive or`` operation of two matrices of matrix and scalar.
-
-.. ocv:function:: void gpu::bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null())
-.. ocv:function:: void gpu::bitwise_xor(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src1: First source matrix.
-
-    :param src2: Second source matrix with the same size and type as ``src1`` .
-
-    :param dst: Destination matrix with the same size and type as ``src1`` .
-
-    :param mask: Optional operation mask. 8-bit single channel image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::rshift
---------------------
-Performs pixel by pixel right shift of an image by a constant value.
-
-.. ocv:function:: void gpu::rshift( const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream=Stream::Null() )
-
-    :param src: Source matrix. Supports 1, 3 and 4 channels images with integers elements.
-
-    :param sc: Constant values, one per channel.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::lshift
---------------------
-Performs pixel by pixel right left of an image by a constant value.
-
-.. ocv:function:: void gpu::lshift( const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream=Stream::Null() )
-
-    :param src: Source matrix. Supports 1, 3 and 4 channels images with ``CV_8U`` , ``CV_16U`` or ``CV_32S`` depth.
-
-    :param sc: Constant values, one per channel.
-
-    :param dst: Destination matrix with the same size and type as ``src`` .
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::min
-------------
-Computes the per-element minimum of two matrices (or a matrix and a scalar).
-
-.. ocv:function:: void gpu::min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::min(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src1: First source matrix.
-
-    :param src2: Second source matrix or a scalar to compare ``src1`` elements with.
-
-    :param dst: Destination matrix with the same size and type as ``src1`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`min`
-
-
-
-gpu::max
-------------
-Computes the per-element maximum of two matrices (or a matrix and a scalar).
-
-.. ocv:function:: void gpu::max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null())
-
-.. ocv:function:: void gpu::max(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null())
-
-    :param src1: First source matrix.
-
-    :param src2: Second source matrix or a scalar to compare ``src1`` elements with.
-
-    :param dst: Destination matrix with the same size and type as ``src1`` .
-
-    :param stream: Stream for the asynchronous version.
-
-.. seealso:: :ocv:func:`max`
diff --git a/modules/gpu/doc/video.rst b/modules/gpu/doc/video.rst
deleted file mode 100644
index fc5b1fb6cb..0000000000
--- a/modules/gpu/doc/video.rst
+++ /dev/null
@@ -1,1213 +0,0 @@
-Video Analysis
-==============
-
-.. highlight:: cpp
-
-
-
-gpu::BroxOpticalFlow
---------------------
-.. ocv:class:: gpu::BroxOpticalFlow
-
-Class computing the optical flow for two images using Brox et al Optical Flow algorithm ([Brox2004]_). ::
-
-    class BroxOpticalFlow
-    {
-    public:
-        BroxOpticalFlow(float alpha_, float gamma_, float scale_factor_, int inner_iterations_, int outer_iterations_, int solver_iterations_);
-
-        //! Compute optical flow
-        //! frame0 - source frame (supports only CV_32FC1 type)
-        //! frame1 - frame to track (with the same size and type as frame0)
-        //! u      - flow horizontal component (along x axis)
-        //! v      - flow vertical component (along y axis)
-        void operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& stream = Stream::Null());
-
-        //! flow smoothness
-        float alpha;
-
-        //! gradient constancy importance
-        float gamma;
-
-        //! pyramid scale factor
-        float scale_factor;
-
-        //! number of lagged non-linearity iterations (inner loop)
-        int inner_iterations;
-
-        //! number of warping iterations (number of pyramid levels)
-        int outer_iterations;
-
-        //! number of linear system solver iterations
-        int solver_iterations;
-
-        GpuMat buf;
-    };
-
-
-
-gpu::GoodFeaturesToTrackDetector_GPU
-------------------------------------
-.. ocv:class:: gpu::GoodFeaturesToTrackDetector_GPU
-
-Class used for strong corners detection on an image. ::
-
-    class GoodFeaturesToTrackDetector_GPU
-    {
-    public:
-        explicit GoodFeaturesToTrackDetector_GPU(int maxCorners_ = 1000, double qualityLevel_ = 0.01, double minDistance_ = 0.0,
-            int blockSize_ = 3, bool useHarrisDetector_ = false, double harrisK_ = 0.04);
-
-        void operator ()(const GpuMat& image, GpuMat& corners, const GpuMat& mask = GpuMat());
-
-        int maxCorners;
-        double qualityLevel;
-        double minDistance;
-
-        int blockSize;
-        bool useHarrisDetector;
-        double harrisK;
-
-        void releaseMemory();
-    };
-
-The class finds the most prominent corners in the image.
-
-.. seealso:: :ocv:func:`goodFeaturesToTrack`
-
-
-
-gpu::GoodFeaturesToTrackDetector_GPU::GoodFeaturesToTrackDetector_GPU
----------------------------------------------------------------------
-Constructor.
-
-.. ocv:function:: gpu::GoodFeaturesToTrackDetector_GPU::GoodFeaturesToTrackDetector_GPU(int maxCorners = 1000, double qualityLevel = 0.01, double minDistance = 0.0, int blockSize = 3, bool useHarrisDetector = false, double harrisK = 0.04)
-
-    :param maxCorners: Maximum number of corners to return. If there are more corners than are found, the strongest of them is returned.
-
-    :param qualityLevel: Parameter characterizing the minimal accepted quality of image corners. The parameter value is multiplied by the best corner quality measure, which is the minimal eigenvalue (see  :ocv:func:`gpu::cornerMinEigenVal` ) or the Harris function response (see  :ocv:func:`gpu::cornerHarris` ). The corners with the quality measure less than the product are rejected. For example, if the best corner has the quality measure = 1500, and the  ``qualityLevel=0.01`` , then all the corners with the quality measure less than 15 are rejected.
-
-    :param minDistance: Minimum possible Euclidean distance between the returned corners.
-
-    :param blockSize: Size of an average block for computing a derivative covariation matrix over each pixel neighborhood. See  :ocv:func:`cornerEigenValsAndVecs` .
-
-    :param useHarrisDetector: Parameter indicating whether to use a Harris detector (see :ocv:func:`gpu::cornerHarris`) or :ocv:func:`gpu::cornerMinEigenVal`.
-
-    :param harrisK: Free parameter of the Harris detector.
-
-
-
-gpu::GoodFeaturesToTrackDetector_GPU::operator ()
--------------------------------------------------
-Finds the most prominent corners in the image.
-
-.. ocv:function:: void gpu::GoodFeaturesToTrackDetector_GPU::operator ()(const GpuMat& image, GpuMat& corners, const GpuMat& mask = GpuMat())
-
-    :param image: Input 8-bit, single-channel image.
-
-    :param corners: Output vector of detected corners (it will be one row matrix with CV_32FC2 type).
-
-    :param mask: Optional region of interest. If the image is not empty (it needs to have the type  ``CV_8UC1``  and the same size as  ``image`` ), it  specifies the region in which the corners are detected.
-
-.. seealso:: :ocv:func:`goodFeaturesToTrack`
-
-
-
-gpu::GoodFeaturesToTrackDetector_GPU::releaseMemory
----------------------------------------------------
-Releases inner buffers memory.
-
-.. ocv:function:: void gpu::GoodFeaturesToTrackDetector_GPU::releaseMemory()
-
-
-
-gpu::FarnebackOpticalFlow
--------------------------
-.. ocv:class:: gpu::FarnebackOpticalFlow
-
-Class computing a dense optical flow using the Gunnar Farneback’s algorithm. ::
-
-    class CV_EXPORTS FarnebackOpticalFlow
-    {
-    public:
-        FarnebackOpticalFlow()
-        {
-            numLevels = 5;
-            pyrScale = 0.5;
-            fastPyramids = false;
-            winSize = 13;
-            numIters = 10;
-            polyN = 5;
-            polySigma = 1.1;
-            flags = 0;
-        }
-
-        int numLevels;
-        double pyrScale;
-        bool fastPyramids;
-        int winSize;
-        int numIters;
-        int polyN;
-        double polySigma;
-        int flags;
-
-        void operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null());
-
-        void releaseMemory();
-
-    private:
-        /* hidden */
-    };
-
-
-
-gpu::FarnebackOpticalFlow::operator ()
---------------------------------------
-Computes a dense optical flow using the Gunnar Farneback’s algorithm.
-
-.. ocv:function:: void gpu::FarnebackOpticalFlow::operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null())
-
-    :param frame0: First 8-bit gray-scale input image
-    :param frame1: Second 8-bit gray-scale input image
-    :param flowx: Flow horizontal component
-    :param flowy: Flow vertical component
-    :param s: Stream
-
-.. seealso:: :ocv:func:`calcOpticalFlowFarneback`
-
-
-
-gpu::FarnebackOpticalFlow::releaseMemory
-----------------------------------------
-Releases unused auxiliary memory buffers.
-
-.. ocv:function:: void gpu::FarnebackOpticalFlow::releaseMemory()
-
-
-
-gpu::PyrLKOpticalFlow
----------------------
-.. ocv:class:: gpu::PyrLKOpticalFlow
-
-Class used for calculating an optical flow. ::
-
-    class PyrLKOpticalFlow
-    {
-    public:
-        PyrLKOpticalFlow();
-
-        void sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts,
-            GpuMat& status, GpuMat* err = 0);
-
-        void dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0);
-
-        Size winSize;
-        int maxLevel;
-        int iters;
-        bool useInitialFlow;
-
-        void releaseMemory();
-    };
-
-The class can calculate an optical flow for a sparse feature set or dense optical flow using the iterative Lucas-Kanade method with pyramids.
-
-.. seealso:: :ocv:func:`calcOpticalFlowPyrLK`
-
-
-
-gpu::PyrLKOpticalFlow::sparse
------------------------------
-Calculate an optical flow for a sparse feature set.
-
-.. ocv:function:: void gpu::PyrLKOpticalFlow::sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts, GpuMat& status, GpuMat* err = 0)
-
-    :param prevImg: First 8-bit input image (supports both grayscale and color images).
-
-    :param nextImg: Second input image of the same size and the same type as  ``prevImg`` .
-
-    :param prevPts: Vector of 2D points for which the flow needs to be found. It must be one row matrix with CV_32FC2 type.
-
-    :param nextPts: Output vector of 2D points (with single-precision floating-point coordinates) containing the calculated new positions of input features in the second image. When ``useInitialFlow`` is true, the vector must have the same size as in the input.
-
-    :param status: Output status vector (CV_8UC1 type). Each element of the vector is set to 1 if the flow for the corresponding features has been found. Otherwise, it is set to 0.
-
-    :param err: Output vector (CV_32FC1 type) that contains the difference between patches around the original and moved points or min eigen value if ``getMinEigenVals`` is checked. It can be NULL, if not needed.
-
-.. seealso:: :ocv:func:`calcOpticalFlowPyrLK`
-
-
-
-gpu::PyrLKOpticalFlow::dense
------------------------------
-Calculate dense optical flow.
-
-.. ocv:function:: void gpu::PyrLKOpticalFlow::dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0)
-
-    :param prevImg: First 8-bit grayscale input image.
-
-    :param nextImg: Second input image of the same size and the same type as  ``prevImg`` .
-
-    :param u: Horizontal component of the optical flow of the same size as input images, 32-bit floating-point, single-channel
-
-    :param v: Vertical component of the optical flow of the same size as input images, 32-bit floating-point, single-channel
-
-    :param err: Output vector (CV_32FC1 type) that contains the difference between patches around the original and moved points or min eigen value if ``getMinEigenVals`` is checked. It can be NULL, if not needed.
-
-
-
-gpu::PyrLKOpticalFlow::releaseMemory
-------------------------------------
-Releases inner buffers memory.
-
-.. ocv:function:: void gpu::PyrLKOpticalFlow::releaseMemory()
-
-
-
-gpu::interpolateFrames
-----------------------
-Interpolates frames (images) using provided optical flow (displacement field).
-
-.. ocv:function:: void gpu::interpolateFrames(const GpuMat& frame0, const GpuMat& frame1, const GpuMat& fu, const GpuMat& fv, const GpuMat& bu, const GpuMat& bv, float pos, GpuMat& newFrame, GpuMat& buf, Stream& stream = Stream::Null())
-
-    :param frame0: First frame (32-bit floating point images, single channel).
-
-    :param frame1: Second frame. Must have the same type and size as ``frame0`` .
-
-    :param fu: Forward horizontal displacement.
-
-    :param fv: Forward vertical displacement.
-
-    :param bu: Backward horizontal displacement.
-
-    :param bv: Backward vertical displacement.
-
-    :param pos: New frame position.
-
-    :param newFrame: Output image.
-
-    :param buf: Temporary buffer, will have width x 6*height size, CV_32FC1 type and contain 6 GpuMat: occlusion masks for first frame, occlusion masks for second, interpolated forward horizontal flow, interpolated forward vertical flow, interpolated backward horizontal flow, interpolated backward vertical flow.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::FGDStatModel
------------------
-.. ocv:class:: gpu::FGDStatModel
-
-Class used for background/foreground segmentation. ::
-
-    class FGDStatModel
-    {
-    public:
-        struct Params
-        {
-            ...
-        };
-
-        explicit FGDStatModel(int out_cn = 3);
-        explicit FGDStatModel(const cv::gpu::GpuMat& firstFrame, const Params& params = Params(), int out_cn = 3);
-
-        ~FGDStatModel();
-
-        void create(const cv::gpu::GpuMat& firstFrame, const Params& params = Params());
-        void release();
-
-        int update(const cv::gpu::GpuMat& curFrame);
-
-        //8UC3 or 8UC4 reference background image
-        cv::gpu::GpuMat background;
-
-        //8UC1 foreground image
-        cv::gpu::GpuMat foreground;
-
-        std::vector< std::vector<cv::Point> > foreground_regions;
-    };
-
-  The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [FGD2003]_.
-
-  The results are available through the class fields:
-
-    .. ocv:member:: cv::gpu::GpuMat background
-
-        The output background image.
-
-    .. ocv:member:: cv::gpu::GpuMat foreground
-
-        The output foreground mask as an 8-bit binary image.
-
-    .. ocv:member:: cv::gpu::GpuMat foreground_regions
-
-        The output foreground regions calculated by :ocv:func:`findContours`.
-
-
-
-gpu::FGDStatModel::FGDStatModel
--------------------------------
-Constructors.
-
-.. ocv:function:: gpu::FGDStatModel::FGDStatModel(int out_cn = 3)
-.. ocv:function:: gpu::FGDStatModel::FGDStatModel(const cv::gpu::GpuMat& firstFrame, const Params& params = Params(), int out_cn = 3)
-
-    :param firstFrame: First frame from video stream. Supports 3- and 4-channels input ( ``CV_8UC3`` and ``CV_8UC4`` ).
-
-    :param params: Algorithm's parameters. See [FGD2003]_ for explanation.
-
-    :param out_cn: Channels count in output result and inner buffers. Can be 3 or 4. 4-channels version requires more memory, but works a bit faster.
-
-.. seealso:: :ocv:func:`gpu::FGDStatModel::create`
-
-
-
-gpu::FGDStatModel::create
--------------------------
-Initializes background model.
-
-.. ocv:function:: void gpu::FGDStatModel::create(const cv::gpu::GpuMat& firstFrame, const Params& params = Params())
-
-    :param firstFrame: First frame from video stream. Supports 3- and 4-channels input ( ``CV_8UC3`` and ``CV_8UC4`` ).
-
-    :param params: Algorithm's parameters. See [FGD2003]_ for explanation.
-
-
-
-gpu::FGDStatModel::release
---------------------------
-Releases all inner buffer's memory.
-
-.. ocv:function:: void gpu::FGDStatModel::release()
-
-
-
-gpu::FGDStatModel::update
---------------------------
-Updates the background model and returns foreground regions count.
-
-.. ocv:function:: int gpu::FGDStatModel::update(const cv::gpu::GpuMat& curFrame)
-
-    :param curFrame: Next video frame.
-
-
-
-gpu::MOG_GPU
-------------
-.. ocv:class:: gpu::MOG_GPU
-
-Gaussian Mixture-based Backbround/Foreground Segmentation Algorithm. ::
-
-    class MOG_GPU
-    {
-    public:
-        MOG_GPU(int nmixtures = -1);
-
-        void initialize(Size frameSize, int frameType);
-
-        void operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate = 0.0f, Stream& stream = Stream::Null());
-
-        void getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const;
-
-        void release();
-
-        int history;
-        float varThreshold;
-        float backgroundRatio;
-        float noiseSigma;
-    };
-
-The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [MOG2001]_.
-
-.. seealso:: :ocv:class:`BackgroundSubtractorMOG`
-
-
-
-gpu::MOG_GPU::MOG_GPU
----------------------
-The constructor.
-
-.. ocv:function:: gpu::MOG_GPU::MOG_GPU(int nmixtures = -1)
-
-    :param nmixtures: Number of Gaussian mixtures.
-
-Default constructor sets all parameters to default values.
-
-
-
-gpu::MOG_GPU::operator()
-------------------------
-Updates the background model and returns the foreground mask.
-
-.. ocv:function:: void gpu::MOG_GPU::operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate = 0.0f, Stream& stream = Stream::Null())
-
-    :param frame: Next video frame.
-
-    :param fgmask: The output foreground mask as an 8-bit binary image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::MOG_GPU::getBackgroundImage
---------------------------------
-Computes a background image.
-
-.. ocv:function:: void gpu::MOG_GPU::getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const
-
-    :param backgroundImage: The output background image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::MOG_GPU::release
----------------------
-Releases all inner buffer's memory.
-
-.. ocv:function:: void gpu::MOG_GPU::release()
-
-
-
-gpu::MOG2_GPU
--------------
-.. ocv:class:: gpu::MOG2_GPU
-
-Gaussian Mixture-based Background/Foreground Segmentation Algorithm. ::
-
-    class MOG2_GPU
-    {
-    public:
-        MOG2_GPU(int nmixtures = -1);
-
-        void initialize(Size frameSize, int frameType);
-
-        void operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate = 0.0f, Stream& stream = Stream::Null());
-
-        void getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const;
-
-        void release();
-
-        // parameters
-        ...
-    };
-
-  The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [MOG2004]_.
-
-  Here are important members of the class that control the algorithm, which you can set after constructing the class instance:
-
-    .. ocv:member:: float backgroundRatio
-
-        Threshold defining whether the component is significant enough to be included into the background model ( corresponds to ``TB=1-cf`` from the paper??which paper??). ``cf=0.1 => TB=0.9`` is default. For ``alpha=0.001``, it means that the mode should exist for approximately 105 frames before it is considered foreground.
-
-    .. ocv:member:: float varThreshold
-
-        Threshold for the squared Mahalanobis distance that helps decide when a sample is close to the existing components (corresponds to ``Tg``). If it is not close to any component, a new component is generated. ``3 sigma => Tg=3*3=9`` is default. A smaller ``Tg`` value generates more components. A higher ``Tg`` value may result in a small number of components but they can grow too large.
-
-    .. ocv:member:: float fVarInit
-
-        Initial variance for the newly generated components. It affects the speed of adaptation. The parameter value is based on your estimate of the typical standard deviation from the images. OpenCV uses 15 as a reasonable value.
-
-    .. ocv:member:: float fVarMin
-
-        Parameter used to further control the variance.
-
-    .. ocv:member:: float fVarMax
-
-        Parameter used to further control the variance.
-
-    .. ocv:member:: float fCT
-
-        Complexity reduction parameter. This parameter defines the number of samples needed to accept to prove the component exists. ``CT=0.05`` is a default value for all the samples. By setting ``CT=0`` you get an algorithm very similar to the standard Stauffer&Grimson algorithm.
-
-    .. ocv:member:: uchar nShadowDetection
-
-        The value for marking shadow pixels in the output foreground mask. Default value is 127.
-
-    .. ocv:member:: float fTau
-
-        Shadow threshold. The shadow is detected if the pixel is a darker version of the background. ``Tau`` is a threshold defining how much darker the shadow can be. ``Tau= 0.5`` means that if a pixel is more than twice darker then it is not shadow. See [ShadowDetect2003]_.
-
-    .. ocv:member:: bool bShadowDetection
-
-        Parameter defining whether shadow detection should be enabled.
-
-.. seealso:: :ocv:class:`BackgroundSubtractorMOG2`
-
-
-
-gpu::MOG2_GPU::MOG2_GPU
------------------------
-The constructor.
-
-.. ocv:function:: gpu::MOG2_GPU::MOG2_GPU(int nmixtures = -1)
-
-    :param nmixtures: Number of Gaussian mixtures.
-
-Default constructor sets all parameters to default values.
-
-
-
-gpu::MOG2_GPU::operator()
--------------------------
-Updates the background model and returns the foreground mask.
-
-.. ocv:function:: void gpu::MOG2_GPU::operator()( const GpuMat& frame, GpuMat& fgmask, float learningRate=-1.0f, Stream& stream=Stream::Null() )
-
-    :param frame: Next video frame.
-
-    :param fgmask: The output foreground mask as an 8-bit binary image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::MOG2_GPU::getBackgroundImage
----------------------------------
-Computes a background image.
-
-.. ocv:function:: void gpu::MOG2_GPU::getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const
-
-    :param backgroundImage: The output background image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::MOG2_GPU::release
-----------------------
-Releases all inner buffer's memory.
-
-.. ocv:function:: void gpu::MOG2_GPU::release()
-
-
-
-gpu::VIBE_GPU
--------------
-.. ocv:class:: gpu::VIBE_GPU
-
-Class used for background/foreground segmentation. ::
-
-    class VIBE_GPU
-    {
-    public:
-        explicit VIBE_GPU(unsigned long rngSeed = 1234567);
-
-        void initialize(const GpuMat& firstFrame, Stream& stream = Stream::Null());
-
-        void operator()(const GpuMat& frame, GpuMat& fgmask, Stream& stream = Stream::Null());
-
-        void release();
-
-        ...
-    };
-
-The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [VIBE2011]_.
-
-
-
-gpu::VIBE_GPU::VIBE_GPU
------------------------
-The constructor.
-
-.. ocv:function:: gpu::VIBE_GPU::VIBE_GPU(unsigned long rngSeed = 1234567)
-
-    :param rngSeed: Value used to initiate a random sequence.
-
-Default constructor sets all parameters to default values.
-
-
-
-gpu::VIBE_GPU::initialize
--------------------------
-Initialize background model and allocates all inner buffers.
-
-.. ocv:function:: void gpu::VIBE_GPU::initialize(const GpuMat& firstFrame, Stream& stream = Stream::Null())
-
-    :param firstFrame: First frame from video sequence.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::VIBE_GPU::operator()
--------------------------
-Updates the background model and returns the foreground mask
-
-.. ocv:function:: void gpu::VIBE_GPU::operator()(const GpuMat& frame, GpuMat& fgmask, Stream& stream = Stream::Null())
-
-    :param frame: Next video frame.
-
-    :param fgmask: The output foreground mask as an 8-bit binary image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::VIBE_GPU::release
-----------------------
-Releases all inner buffer's memory.
-
-.. ocv:function:: void gpu::VIBE_GPU::release()
-
-
-
-gpu::GMG_GPU
-------------
-.. ocv:class:: gpu::GMG_GPU
-
-  Class used for background/foreground segmentation. ::
-
-    class GMG_GPU_GPU
-    {
-    public:
-        GMG_GPU();
-
-        void initialize(Size frameSize, float min = 0.0f, float max = 255.0f);
-
-        void operator ()(const GpuMat& frame, GpuMat& fgmask, float learningRate = -1.0f, Stream& stream = Stream::Null());
-
-        void release();
-
-        int    maxFeatures;
-        float  learningRate;
-        int    numInitializationFrames;
-        int    quantizationLevels;
-        float  backgroundPrior;
-        float  decisionThreshold;
-        int    smoothingRadius;
-
-        ...
-    };
-
-  The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [GMG2012]_.
-
-  Here are important members of the class that control the algorithm, which you can set after constructing the class instance:
-
-    .. ocv:member:: int maxFeatures
-
-        Total number of distinct colors to maintain in histogram.
-
-    .. ocv:member:: float learningRate
-
-        Set between 0.0 and 1.0, determines how quickly features are "forgotten" from histograms.
-
-    .. ocv:member:: int numInitializationFrames
-
-        Number of frames of video to use to initialize histograms.
-
-    .. ocv:member:: int quantizationLevels
-
-        Number of discrete levels in each channel to be used in histograms.
-
-    .. ocv:member:: float backgroundPrior
-
-        Prior probability that any given pixel is a background pixel. A sensitivity parameter.
-
-    .. ocv:member:: float decisionThreshold
-
-        Value above which pixel is determined to be FG.
-
-    .. ocv:member:: float smoothingRadius
-
-        Smoothing radius, in pixels, for cleaning up FG image.
-
-
-
-gpu::GMG_GPU::GMG_GPU
----------------------
-The default constructor.
-
-.. ocv:function:: gpu::GMG_GPU::GMG_GPU()
-
-Default constructor sets all parameters to default values.
-
-
-
-gpu::GMG_GPU::initialize
-------------------------
-Initialize background model and allocates all inner buffers.
-
-.. ocv:function:: void gpu::GMG_GPU::initialize(Size frameSize, float min = 0.0f, float max = 255.0f)
-
-    :param frameSize: Input frame size.
-
-    :param min: Minimum value taken on by pixels in image sequence. Usually 0.
-
-    :param max: Maximum value taken on by pixels in image sequence, e.g. 1.0 or 255.
-
-
-
-gpu::GMG_GPU::operator()
-------------------------
-Updates the background model and returns the foreground mask
-
-.. ocv:function:: void gpu::GMG_GPU::operator ()( const GpuMat& frame, GpuMat& fgmask, float learningRate=-1.0f, Stream& stream=Stream::Null() )
-
-    :param frame: Next video frame.
-
-    :param fgmask: The output foreground mask as an 8-bit binary image.
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::GMG_GPU::release
----------------------
-Releases all inner buffer's memory.
-
-.. ocv:function:: void gpu::GMG_GPU::release()
-
-
-
-gpu::VideoWriter_GPU
----------------------
-Video writer class.
-
-.. ocv:class:: gpu::VideoWriter_GPU
-
-The class uses H264 video codec.
-
-.. note:: Currently only Windows platform is supported.
-
-
-
-gpu::VideoWriter_GPU::VideoWriter_GPU
--------------------------------------
-Constructors.
-
-.. ocv:function:: gpu::VideoWriter_GPU::VideoWriter_GPU()
-.. ocv:function:: gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR)
-.. ocv:function:: gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
-.. ocv:function:: gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR)
-.. ocv:function:: gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
-
-    :param fileName: Name of the output video file. Only AVI file format is supported.
-
-    :param frameSize: Size of the input video frames.
-
-    :param fps: Framerate of the created video stream.
-
-    :param params: Encoder parameters. See :ocv:struct:`gpu::VideoWriter_GPU::EncoderParams` .
-
-    :param format: Surface format of input frames ( ``SF_UYVY`` , ``SF_YUY2`` , ``SF_YV12`` , ``SF_NV12`` , ``SF_IYUV`` , ``SF_BGR`` or ``SF_GRAY``). BGR or gray frames will be converted to YV12 format before encoding, frames with other formats will be used as is.
-
-    :param encoderCallback: Callbacks for video encoder. See :ocv:class:`gpu::VideoWriter_GPU::EncoderCallBack` . Use it if you want to work with raw video stream.
-
-The constructors initialize video writer. FFMPEG is used to write videos. User can implement own multiplexing with :ocv:class:`gpu::VideoWriter_GPU::EncoderCallBack` .
-
-
-
-gpu::VideoWriter_GPU::open
---------------------------
-Initializes or reinitializes video writer.
-
-.. ocv:function:: void gpu::VideoWriter_GPU::open(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR)
-.. ocv:function:: void gpu::VideoWriter_GPU::open(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
-.. ocv:function:: void gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR)
-.. ocv:function:: void gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
-
-The method opens video writer. Parameters are the same as in the constructor :ocv:func:`gpu::VideoWriter_GPU::VideoWriter_GPU` . The method throws :ocv:class:`Exception` if error occurs.
-
-
-
-gpu::VideoWriter_GPU::isOpened
-------------------------------
-Returns true if video writer has been successfully initialized.
-
-.. ocv:function:: bool gpu::VideoWriter_GPU::isOpened() const
-
-
-
-gpu::VideoWriter_GPU::close
----------------------------
-Releases the video writer.
-
-.. ocv:function:: void gpu::VideoWriter_GPU::close()
-
-
-
-gpu::VideoWriter_GPU::write
----------------------------
-Writes the next video frame.
-
-.. ocv:function:: void gpu::VideoWriter_GPU::write(const cv::gpu::GpuMat& image, bool lastFrame = false)
-
-    :param image: The written frame.
-
-    :param lastFrame: Indicates that it is end of stream. The parameter can be ignored.
-
-The method write the specified image to video file. The image must have the same size and the same surface format as has been specified when opening the video writer.
-
-
-
-gpu::VideoWriter_GPU::EncoderParams
------------------------------------
-.. ocv:struct:: gpu::VideoWriter_GPU::EncoderParams
-
-Different parameters for CUDA video encoder. ::
-
-    struct EncoderParams
-    {
-        int       P_Interval;      //    NVVE_P_INTERVAL,
-        int       IDR_Period;      //    NVVE_IDR_PERIOD,
-        int       DynamicGOP;      //    NVVE_DYNAMIC_GOP,
-        int       RCType;          //    NVVE_RC_TYPE,
-        int       AvgBitrate;      //    NVVE_AVG_BITRATE,
-        int       PeakBitrate;     //    NVVE_PEAK_BITRATE,
-        int       QP_Level_Intra;  //    NVVE_QP_LEVEL_INTRA,
-        int       QP_Level_InterP; //    NVVE_QP_LEVEL_INTER_P,
-        int       QP_Level_InterB; //    NVVE_QP_LEVEL_INTER_B,
-        int       DeblockMode;     //    NVVE_DEBLOCK_MODE,
-        int       ProfileLevel;    //    NVVE_PROFILE_LEVEL,
-        int       ForceIntra;      //    NVVE_FORCE_INTRA,
-        int       ForceIDR;        //    NVVE_FORCE_IDR,
-        int       ClearStat;       //    NVVE_CLEAR_STAT,
-        int       DIMode;          //    NVVE_SET_DEINTERLACE,
-        int       Presets;         //    NVVE_PRESETS,
-        int       DisableCabac;    //    NVVE_DISABLE_CABAC,
-        int       NaluFramingType; //    NVVE_CONFIGURE_NALU_FRAMING_TYPE
-        int       DisableSPSPPS;   //    NVVE_DISABLE_SPS_PPS
-
-        EncoderParams();
-        explicit EncoderParams(const std::string& configFile);
-
-        void load(const std::string& configFile);
-        void save(const std::string& configFile) const;
-    };
-
-
-
-gpu::VideoWriter_GPU::EncoderParams::EncoderParams
---------------------------------------------------
-Constructors.
-
-.. ocv:function:: gpu::VideoWriter_GPU::EncoderParams::EncoderParams()
-.. ocv:function:: gpu::VideoWriter_GPU::EncoderParams::EncoderParams(const std::string& configFile)
-
-    :param configFile: Config file name.
-
-Creates default parameters or reads parameters from config file.
-
-
-
-gpu::VideoWriter_GPU::EncoderParams::load
------------------------------------------
-Reads parameters from config file.
-
-.. ocv:function:: void gpu::VideoWriter_GPU::EncoderParams::load(const std::string& configFile)
-
-    :param configFile: Config file name.
-
-
-
-gpu::VideoWriter_GPU::EncoderParams::save
------------------------------------------
-Saves parameters to config file.
-
-.. ocv:function:: void gpu::VideoWriter_GPU::EncoderParams::save(const std::string& configFile) const
-
-    :param configFile: Config file name.
-
-
-
-gpu::VideoWriter_GPU::EncoderCallBack
--------------------------------------
-.. ocv:class:: gpu::VideoWriter_GPU::EncoderCallBack
-
-Callbacks for CUDA video encoder. ::
-
-    class EncoderCallBack
-    {
-    public:
-        enum PicType
-        {
-            IFRAME = 1,
-            PFRAME = 2,
-            BFRAME = 3
-        };
-
-        virtual ~EncoderCallBack() {}
-
-        virtual unsigned char* acquireBitStream(int* bufferSize) = 0;
-        virtual void releaseBitStream(unsigned char* data, int size) = 0;
-        virtual void onBeginFrame(int frameNumber, PicType picType) = 0;
-        virtual void onEndFrame(int frameNumber, PicType picType) = 0;
-    };
-
-
-
-gpu::VideoWriter_GPU::EncoderCallBack::acquireBitStream
--------------------------------------------------------
-Callback function to signal the start of bitstream that is to be encoded.
-
-.. ocv:function:: virtual uchar* gpu::VideoWriter_GPU::EncoderCallBack::acquireBitStream(int* bufferSize) = 0
-
-Callback must allocate buffer for CUDA encoder and return pointer to it and it's size.
-
-
-
-gpu::VideoWriter_GPU::EncoderCallBack::releaseBitStream
--------------------------------------------------------
-Callback function to signal that the encoded bitstream is ready to be written to file.
-
-.. ocv:function:: virtual void gpu::VideoWriter_GPU::EncoderCallBack::releaseBitStream(unsigned char* data, int size) = 0
-
-
-
-gpu::VideoWriter_GPU::EncoderCallBack::onBeginFrame
----------------------------------------------------
-Callback function to signal that the encoding operation on the frame has started.
-
-.. ocv:function:: virtual void gpu::VideoWriter_GPU::EncoderCallBack::onBeginFrame(int frameNumber, PicType picType) = 0
-
-    :param picType: Specify frame type (I-Frame, P-Frame or B-Frame).
-
-
-
-gpu::VideoWriter_GPU::EncoderCallBack::onEndFrame
--------------------------------------------------
-Callback function signals that the encoding operation on the frame has finished.
-
-.. ocv:function:: virtual void gpu::VideoWriter_GPU::EncoderCallBack::onEndFrame(int frameNumber, PicType picType) = 0
-
-    :param picType: Specify frame type (I-Frame, P-Frame or B-Frame).
-
-
-
-gpu::VideoReader_GPU
---------------------
-Class for reading video from files.
-
-.. ocv:class:: gpu::VideoReader_GPU
-
-.. note:: Currently only Windows and Linux platforms are supported.
-
-
-
-gpu::VideoReader_GPU::Codec
----------------------------
-
-Video codecs supported by :ocv:class:`gpu::VideoReader_GPU` .
-
-.. ocv:enum:: gpu::VideoReader_GPU::Codec
-
-  .. ocv:emember:: MPEG1 = 0
-  .. ocv:emember:: MPEG2
-  .. ocv:emember:: MPEG4
-  .. ocv:emember:: VC1
-  .. ocv:emember:: H264
-  .. ocv:emember:: JPEG
-  .. ocv:emember:: H264_SVC
-  .. ocv:emember:: H264_MVC
-
-  .. ocv:emember:: Uncompressed_YUV420 = (('I'<<24)|('Y'<<16)|('U'<<8)|('V'))
-
-        Y,U,V (4:2:0)
-
-  .. ocv:emember:: Uncompressed_YV12   = (('Y'<<24)|('V'<<16)|('1'<<8)|('2'))
-
-        Y,V,U (4:2:0)
-
-  .. ocv:emember:: Uncompressed_NV12   = (('N'<<24)|('V'<<16)|('1'<<8)|('2'))
-
-        Y,UV  (4:2:0)
-
-  .. ocv:emember:: Uncompressed_YUYV   = (('Y'<<24)|('U'<<16)|('Y'<<8)|('V'))
-
-        YUYV/YUY2 (4:2:2)
-
-  .. ocv:emember:: Uncompressed_UYVY   = (('U'<<24)|('Y'<<16)|('V'<<8)|('Y'))
-
-        UYVY (4:2:2)
-
-
-gpu::VideoReader_GPU::ChromaFormat
-----------------------------------
-
-Chroma formats supported by :ocv:class:`gpu::VideoReader_GPU` .
-
-.. ocv:enum:: gpu::VideoReader_GPU::ChromaFormat
-
-  .. ocv:emember:: Monochrome = 0
-  .. ocv:emember:: YUV420
-  .. ocv:emember:: YUV422
-  .. ocv:emember:: YUV444
-
-
-gpu::VideoReader_GPU::FormatInfo
---------------------------------
-.. ocv:struct:: gpu::VideoReader_GPU::FormatInfo
-
-Struct providing information about video file format. ::
-
-    struct FormatInfo
-    {
-        Codec codec;
-        ChromaFormat chromaFormat;
-        int width;
-        int height;
-    };
-
-
-gpu::VideoReader_GPU::VideoReader_GPU
--------------------------------------
-Constructors.
-
-.. ocv:function:: gpu::VideoReader_GPU::VideoReader_GPU()
-.. ocv:function:: gpu::VideoReader_GPU::VideoReader_GPU(const std::string& filename)
-.. ocv:function:: gpu::VideoReader_GPU::VideoReader_GPU(const cv::Ptr<VideoSource>& source)
-
-    :param filename: Name of the input video file.
-
-    :param source: Video file parser implemented by user.
-
-The constructors initialize video reader. FFMPEG is used to read videos. User can implement own demultiplexing with :ocv:class:`gpu::VideoReader_GPU::VideoSource` .
-
-
-
-gpu::VideoReader_GPU::open
---------------------------
-Initializes or reinitializes video reader.
-
-.. ocv:function:: void gpu::VideoReader_GPU::open(const std::string& filename)
-.. ocv:function:: void gpu::VideoReader_GPU::open(const cv::Ptr<VideoSource>& source)
-
-The method opens video reader. Parameters are the same as in the constructor :ocv:func:`gpu::VideoReader_GPU::VideoReader_GPU` . The method throws :ocv:class:`Exception` if error occurs.
-
-
-
-gpu::VideoReader_GPU::isOpened
-------------------------------
-Returns true if video reader has been successfully initialized.
-
-.. ocv:function:: bool gpu::VideoReader_GPU::isOpened() const
-
-
-
-gpu::VideoReader_GPU::close
----------------------------
-Releases the video reader.
-
-.. ocv:function:: void gpu::VideoReader_GPU::close()
-
-
-
-gpu::VideoReader_GPU::read
---------------------------
-Grabs, decodes and returns the next video frame.
-
-.. ocv:function:: bool gpu::VideoReader_GPU::read(GpuMat& image)
-
-If no frames has been grabbed (there are no more frames in video file), the methods return ``false`` . The method throws :ocv:class:`Exception` if error occurs.
-
-
-
-gpu::VideoReader_GPU::format
-----------------------------
-Returns information about video file format.
-
-.. ocv:function:: FormatInfo gpu::VideoReader_GPU::format() const
-
-The method throws :ocv:class:`Exception` if video reader wasn't initialized.
-
-
-
-gpu::VideoReader_GPU::dumpFormat
---------------------------------
-Dump information about video file format to specified stream.
-
-.. ocv:function:: void gpu::VideoReader_GPU::dumpFormat(std::ostream& st)
-
-    :param st: Output stream.
-
-The method throws :ocv:class:`Exception` if video reader wasn't initialized.
-
-
-
-gpu::VideoReader_GPU::VideoSource
------------------------------------
-.. ocv:class:: gpu::VideoReader_GPU::VideoSource
-
-Interface for video demultiplexing. ::
-
-    class VideoSource
-    {
-    public:
-        VideoSource();
-        virtual ~VideoSource() {}
-
-        virtual FormatInfo format() const = 0;
-        virtual void start() = 0;
-        virtual void stop() = 0;
-        virtual bool isStarted() const = 0;
-        virtual bool hasError() const = 0;
-
-    protected:
-        bool parseVideoData(const unsigned char* data, size_t size, bool endOfStream = false);
-    };
-
-User can implement own demultiplexing by implementing this interface.
-
-
-
-gpu::VideoReader_GPU::VideoSource::format
------------------------------------------
-Returns information about video file format.
-
-.. ocv:function:: virtual FormatInfo gpu::VideoReader_GPU::VideoSource::format() const = 0
-
-
-
-gpu::VideoReader_GPU::VideoSource::start
-----------------------------------------
-Starts processing.
-
-.. ocv:function:: virtual void gpu::VideoReader_GPU::VideoSource::start() = 0
-
-Implementation must create own thread with video processing and call periodic :ocv:func:`gpu::VideoReader_GPU::VideoSource::parseVideoData` .
-
-
-
-gpu::VideoReader_GPU::VideoSource::stop
----------------------------------------
-Stops processing.
-
-.. ocv:function:: virtual void gpu::VideoReader_GPU::VideoSource::stop() = 0
-
-
-
-gpu::VideoReader_GPU::VideoSource::isStarted
---------------------------------------------
-Returns ``true`` if processing was successfully started.
-
-.. ocv:function:: virtual bool gpu::VideoReader_GPU::VideoSource::isStarted() const = 0
-
-
-
-gpu::VideoReader_GPU::VideoSource::hasError
--------------------------------------------
-Returns ``true`` if error occured during processing.
-
-.. ocv:function:: virtual bool gpu::VideoReader_GPU::VideoSource::hasError() const = 0
-
-
-
-gpu::VideoReader_GPU::VideoSource::parseVideoData
--------------------------------------------------
-Parse next video frame. Implementation must call this method after new frame was grabbed.
-
-.. ocv:function:: bool gpu::VideoReader_GPU::VideoSource::parseVideoData(const uchar* data, size_t size, bool endOfStream = false)
-
-    :param data: Pointer to frame data. Can be ``NULL`` if ``endOfStream`` if ``true`` .
-
-    :param size: Size in bytes of current frame.
-
-    :param endOfStream: Indicates that it is end of stream.
-
-
-
-.. [Brox2004] T. Brox, A. Bruhn, N. Papenberg, J. Weickert. *High accuracy optical flow estimation based on a theory for warping*. ECCV 2004.
-.. [FGD2003] Liyuan Li, Weimin Huang, Irene Y.H. Gu, and Qi Tian. *Foreground Object Detection from Videos Containing Complex Background*. ACM MM2003 9p, 2003.
-.. [MOG2001] P. KadewTraKuPong and R. Bowden. *An improved adaptive background mixture model for real-time tracking with shadow detection*. Proc. 2nd European Workshop on Advanced Video-Based Surveillance Systems, 2001
-.. [MOG2004] Z. Zivkovic. *Improved adaptive Gausian mixture model for background subtraction*. International Conference Pattern Recognition, UK, August, 2004
-.. [ShadowDetect2003] Prati, Mikic, Trivedi and Cucchiarra. *Detecting Moving Shadows...*. IEEE PAMI, 2003
-.. [VIBE2011] O. Barnich and M. Van D Roogenbroeck. *ViBe: A universal background subtraction algorithm for video sequences*. IEEE Transactions on Image Processing, 20(6) :1709-1724, June 2011
-.. [GMG2012] A. Godbehere, A. Matsukawa and K. Goldberg. *Visual Tracking of Human Visitors under Variable-Lighting Conditions for a Responsive Audio Art Installation*. American Control Conference, Montreal, June 2012
diff --git a/modules/gpu/include/opencv2/gpu.hpp b/modules/gpu/include/opencv2/gpu.hpp
index 21a03dc209..1b0120ba25 100644
--- a/modules/gpu/include/opencv2/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu.hpp
@@ -22,7 +22,7 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
@@ -43,1051 +43,56 @@
 #ifndef __OPENCV_GPU_HPP__
 #define __OPENCV_GPU_HPP__
 
-#ifndef SKIP_INCLUDES
-#include <vector>
-#include <memory>
-#include <iosfwd>
+#ifndef __cplusplus
+#  error gpu.hpp header must be compiled as C++
 #endif
 
-#include "opencv2/core/gpumat.hpp"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/objdetect.hpp"
-#include "opencv2/features2d.hpp"
+#include "opencv2/core/gpu.hpp"
+
+#if !defined(__OPENCV_BUILD) && !defined(OPENCV_GPU_SKIP_INCLUDE)
+    #include "opencv2/opencv_modules.hpp"
+
+    #ifdef HAVE_OPENCV_GPUARITHM
+        #include "opencv2/gpuarithm.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUWARPING
+        #include "opencv2/gpuwarping.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUFILTERS
+        #include "opencv2/gpufilters.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUIMGPROC
+        #include "opencv2/gpuimgproc.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUFEATURES2D
+        #include "opencv2/gpufeatures2d.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUOPTFLOW
+        #include "opencv2/gpuoptflow.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUBGSEGM
+        #include "opencv2/gpubgsegm.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUSTEREO
+        #include "opencv2/gpustereo.hpp"
+    #endif
+
+    #ifdef HAVE_OPENCV_GPUCODEC
+        #include "opencv2/gpucodec.hpp"
+    #endif
+#endif
 
 namespace cv { namespace gpu {
-//////////////////////////////// Filter Engine ////////////////////////////////
-
-/*!
-The Base Class for 1D or Row-wise Filters
-
-This is the base class for linear or non-linear filters that process 1D data.
-In particular, such filters are used for the "horizontal" filtering parts in separable filters.
-*/
-class CV_EXPORTS BaseRowFilter_GPU
-{
-public:
-    BaseRowFilter_GPU(int ksize_, int anchor_) : ksize(ksize_), anchor(anchor_) {}
-    virtual ~BaseRowFilter_GPU() {}
-    virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-    int ksize, anchor;
-};
-
-/*!
-The Base Class for Column-wise Filters
-
-This is the base class for linear or non-linear filters that process columns of 2D arrays.
-Such filters are used for the "vertical" filtering parts in separable filters.
-*/
-class CV_EXPORTS BaseColumnFilter_GPU
-{
-public:
-    BaseColumnFilter_GPU(int ksize_, int anchor_) : ksize(ksize_), anchor(anchor_) {}
-    virtual ~BaseColumnFilter_GPU() {}
-    virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-    int ksize, anchor;
-};
-
-/*!
-The Base Class for Non-Separable 2D Filters.
-
-This is the base class for linear or non-linear 2D filters.
-*/
-class CV_EXPORTS BaseFilter_GPU
-{
-public:
-    BaseFilter_GPU(const Size& ksize_, const Point& anchor_) : ksize(ksize_), anchor(anchor_) {}
-    virtual ~BaseFilter_GPU() {}
-    virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
-    Size ksize;
-    Point anchor;
-};
-
-/*!
-The Base Class for Filter Engine.
-
-The class can be used to apply an arbitrary filtering operation to an image.
-It contains all the necessary intermediate buffers.
-*/
-class CV_EXPORTS FilterEngine_GPU
-{
-public:
-    virtual ~FilterEngine_GPU() {}
-
-    virtual void apply(const GpuMat& src, GpuMat& dst, Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null()) = 0;
-};
-
-//! returns the non-separable filter engine with the specified filter
-CV_EXPORTS Ptr<FilterEngine_GPU> createFilter2D_GPU(const Ptr<BaseFilter_GPU>& filter2D, int srcType, int dstType);
-
-//! returns the separable filter engine with the specified filters
-CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter,
-    const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType);
-CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter,
-    const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType, GpuMat& buf);
-
-//! returns horizontal 1D box filter
-//! supports only CV_8UC1 source type and CV_32FC1 sum type
-CV_EXPORTS Ptr<BaseRowFilter_GPU> getRowSumFilter_GPU(int srcType, int sumType, int ksize, int anchor = -1);
-
-//! returns vertical 1D box filter
-//! supports only CV_8UC1 sum type and CV_32FC1 dst type
-CV_EXPORTS Ptr<BaseColumnFilter_GPU> getColumnSumFilter_GPU(int sumType, int dstType, int ksize, int anchor = -1);
-
-//! returns 2D box filter
-//! supports CV_8UC1 and CV_8UC4 source type, dst type must be the same as source type
-CV_EXPORTS Ptr<BaseFilter_GPU> getBoxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1, -1));
-
-//! returns box filter engine
-CV_EXPORTS Ptr<FilterEngine_GPU> createBoxFilter_GPU(int srcType, int dstType, const Size& ksize,
-    const Point& anchor = Point(-1,-1));
-
-//! returns 2D morphological filter
-//! only MORPH_ERODE and MORPH_DILATE are supported
-//! supports CV_8UC1 and CV_8UC4 types
-//! kernel must have CV_8UC1 type, one rows and cols == ksize.width * ksize.height
-CV_EXPORTS Ptr<BaseFilter_GPU> getMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Size& ksize,
-    Point anchor=Point(-1,-1));
-
-//! returns morphological filter engine. Only MORPH_ERODE and MORPH_DILATE are supported.
-CV_EXPORTS Ptr<FilterEngine_GPU> createMorphologyFilter_GPU(int op, int type, const Mat& kernel,
-    const Point& anchor = Point(-1,-1), int iterations = 1);
-CV_EXPORTS Ptr<FilterEngine_GPU> createMorphologyFilter_GPU(int op, int type, const Mat& kernel, GpuMat& buf,
-    const Point& anchor = Point(-1,-1), int iterations = 1);
-
-//! returns 2D filter with the specified kernel
-//! supports CV_8U, CV_16U and CV_32F one and four channel image
-CV_EXPORTS Ptr<BaseFilter_GPU> getLinearFilter_GPU(int srcType, int dstType, const Mat& kernel, Point anchor = Point(-1, -1), int borderType = BORDER_DEFAULT);
-
-//! returns the non-separable linear filter engine
-CV_EXPORTS Ptr<FilterEngine_GPU> createLinearFilter_GPU(int srcType, int dstType, const Mat& kernel,
-    Point anchor = Point(-1,-1), int borderType = BORDER_DEFAULT);
-
-//! returns the primitive row filter with the specified kernel.
-//! supports only CV_8UC1, CV_8UC4, CV_16SC1, CV_16SC2, CV_32SC1, CV_32FC1 source type.
-//! there are two version of algorithm: NPP and OpenCV.
-//! NPP calls when srcType == CV_8UC1 or srcType == CV_8UC4 and bufType == srcType,
-//! otherwise calls OpenCV version.
-//! NPP supports only BORDER_CONSTANT border type.
-//! OpenCV version supports only CV_32F as buffer depth and
-//! BORDER_REFLECT101, BORDER_REPLICATE and BORDER_CONSTANT border types.
-CV_EXPORTS Ptr<BaseRowFilter_GPU> getLinearRowFilter_GPU(int srcType, int bufType, const Mat& rowKernel,
-    int anchor = -1, int borderType = BORDER_DEFAULT);
-
-//! returns the primitive column filter with the specified kernel.
-//! supports only CV_8UC1, CV_8UC4, CV_16SC1, CV_16SC2, CV_32SC1, CV_32FC1 dst type.
-//! there are two version of algorithm: NPP and OpenCV.
-//! NPP calls when dstType == CV_8UC1 or dstType == CV_8UC4 and bufType == dstType,
-//! otherwise calls OpenCV version.
-//! NPP supports only BORDER_CONSTANT border type.
-//! OpenCV version supports only CV_32F as buffer depth and
-//! BORDER_REFLECT101, BORDER_REPLICATE and BORDER_CONSTANT border types.
-CV_EXPORTS Ptr<BaseColumnFilter_GPU> getLinearColumnFilter_GPU(int bufType, int dstType, const Mat& columnKernel,
-    int anchor = -1, int borderType = BORDER_DEFAULT);
-
-//! returns the separable linear filter engine
-CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel,
-    const Mat& columnKernel, const Point& anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT,
-    int columnBorderType = -1);
-CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel,
-    const Mat& columnKernel, GpuMat& buf, const Point& anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT,
-    int columnBorderType = -1);
-
-//! returns filter engine for the generalized Sobel operator
-CV_EXPORTS Ptr<FilterEngine_GPU> createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize,
-                                                       int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS Ptr<FilterEngine_GPU> createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize, GpuMat& buf,
-                                                       int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-
-//! returns the Gaussian filter engine
-CV_EXPORTS Ptr<FilterEngine_GPU> createGaussianFilter_GPU(int type, Size ksize, double sigma1, double sigma2 = 0,
-                                                          int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS Ptr<FilterEngine_GPU> createGaussianFilter_GPU(int type, Size ksize, GpuMat& buf, double sigma1, double sigma2 = 0,
-                                                          int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-
-//! returns maximum filter
-CV_EXPORTS Ptr<BaseFilter_GPU> getMaxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1));
-
-//! returns minimum filter
-CV_EXPORTS Ptr<BaseFilter_GPU> getMinFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1));
-
-//! smooths the image using the normalized box filter
-//! supports CV_8UC1, CV_8UC4 types
-CV_EXPORTS void boxFilter(const GpuMat& src, GpuMat& dst, int ddepth, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null());
-
-//! a synonym for normalized box filter
-static inline void blur(const GpuMat& src, GpuMat& dst, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null())
-{
-    boxFilter(src, dst, -1, ksize, anchor, stream);
-}
-
-//! erodes the image (applies the local minimum operator)
-CV_EXPORTS void erode(const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1);
-CV_EXPORTS void erode(const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf,
-                      Point anchor = Point(-1, -1), int iterations = 1,
-                      Stream& stream = Stream::Null());
-
-//! dilates the image (applies the local maximum operator)
-CV_EXPORTS void dilate(const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1);
-CV_EXPORTS void dilate(const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf,
-                       Point anchor = Point(-1, -1), int iterations = 1,
-                       Stream& stream = Stream::Null());
-
-//! applies an advanced morphological operation to the image
-CV_EXPORTS void morphologyEx(const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1);
-CV_EXPORTS void morphologyEx(const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, GpuMat& buf1, GpuMat& buf2,
-                             Point anchor = Point(-1, -1), int iterations = 1, Stream& stream = Stream::Null());
-
-//! applies non-separable 2D linear filter to the image
-CV_EXPORTS void filter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernel, Point anchor=Point(-1,-1), int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null());
-
-//! applies separable 2D linear filter to the image
-CV_EXPORTS void sepFilter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY,
-                            Point anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS void sepFilter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY, GpuMat& buf,
-                            Point anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1,
-                            Stream& stream = Stream::Null());
-
-//! applies generalized Sobel operator to the image
-CV_EXPORTS void Sobel(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, int ksize = 3, double scale = 1,
-                      int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS void Sobel(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, int ksize = 3, double scale = 1,
-                      int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
-
-//! applies the vertical or horizontal Scharr operator to the image
-CV_EXPORTS void Scharr(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, double scale = 1,
-                       int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS void Scharr(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, double scale = 1,
-                       int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
-
-//! smooths the image using Gaussian filter.
-CV_EXPORTS void GaussianBlur(const GpuMat& src, GpuMat& dst, Size ksize, double sigma1, double sigma2 = 0,
-                             int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
-CV_EXPORTS void GaussianBlur(const GpuMat& src, GpuMat& dst, Size ksize, GpuMat& buf, double sigma1, double sigma2 = 0,
-                             int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
-
-//! applies Laplacian operator to the image
-//! supports only ksize = 1 and ksize = 3
-CV_EXPORTS void Laplacian(const GpuMat& src, GpuMat& dst, int ddepth, int ksize = 1, double scale = 1, int borderType = BORDER_DEFAULT, Stream& stream = Stream::Null());
-
-
-////////////////////////////// Arithmetics ///////////////////////////////////
-
-//! implements generalized matrix product algorithm GEMM from BLAS
-CV_EXPORTS void gemm(const GpuMat& src1, const GpuMat& src2, double alpha,
-    const GpuMat& src3, double beta, GpuMat& dst, int flags = 0, Stream& stream = Stream::Null());
-
-//! transposes the matrix
-//! supports matrix with element size = 1, 4 and 8 bytes (CV_8UC1, CV_8UC4, CV_16UC2, CV_32FC1, etc)
-CV_EXPORTS void transpose(const GpuMat& src1, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! reverses the order of the rows, columns or both in a matrix
-//! supports 1, 3 and 4 channels images with CV_8U, CV_16U, CV_32S or CV_32F depth
-CV_EXPORTS void flip(const GpuMat& a, GpuMat& b, int flipCode, Stream& stream = Stream::Null());
-
-//! transforms 8-bit unsigned integers using lookup table: dst(i)=lut(src(i))
-//! destination array will have the depth type as lut and the same channels number as source
-//! supports CV_8UC1, CV_8UC3 types
-CV_EXPORTS void LUT(const GpuMat& src, const Mat& lut, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! makes multi-channel array out of several single-channel arrays
-CV_EXPORTS void merge(const GpuMat* src, size_t n, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! makes multi-channel array out of several single-channel arrays
-CV_EXPORTS void merge(const std::vector<GpuMat>& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! copies each plane of a multi-channel array to a dedicated array
-CV_EXPORTS void split(const GpuMat& src, GpuMat* dst, Stream& stream = Stream::Null());
-
-//! copies each plane of a multi-channel array to a dedicated array
-CV_EXPORTS void split(const GpuMat& src, std::vector<GpuMat>& dst, Stream& stream = Stream::Null());
-
-//! computes magnitude of complex (x(i).re, x(i).im) vector
-//! supports only CV_32FC2 type
-CV_EXPORTS void magnitude(const GpuMat& xy, GpuMat& magnitude, Stream& stream = Stream::Null());
-
-//! computes squared magnitude of complex (x(i).re, x(i).im) vector
-//! supports only CV_32FC2 type
-CV_EXPORTS void magnitudeSqr(const GpuMat& xy, GpuMat& magnitude, Stream& stream = Stream::Null());
-
-//! computes magnitude of each (x(i), y(i)) vector
-//! supports only floating-point source
-CV_EXPORTS void magnitude(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null());
-
-//! computes squared magnitude of each (x(i), y(i)) vector
-//! supports only floating-point source
-CV_EXPORTS void magnitudeSqr(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null());
-
-//! computes angle (angle(i)) of each (x(i), y(i)) vector
-//! supports only floating-point source
-CV_EXPORTS void phase(const GpuMat& x, const GpuMat& y, GpuMat& angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
-
-//! converts Cartesian coordinates to polar
-//! supports only floating-point source
-CV_EXPORTS void cartToPolar(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, GpuMat& angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
-
-//! converts polar coordinates to Cartesian
-//! supports only floating-point source
-CV_EXPORTS void polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees = false, Stream& stream = Stream::Null());
-
-//! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
-CV_EXPORTS void normalize(const GpuMat& src, GpuMat& dst, double alpha = 1, double beta = 0,
-                          int norm_type = NORM_L2, int dtype = -1, const GpuMat& mask = GpuMat());
-CV_EXPORTS void normalize(const GpuMat& src, GpuMat& dst, double a, double b,
-                          int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf);
-
-
-//////////////////////////// Per-element operations ////////////////////////////////////
-
-//! adds one matrix to another (c = a + b)
-CV_EXPORTS void add(const GpuMat& a, const GpuMat& b, GpuMat& c, const GpuMat& mask = GpuMat(), int dtype = -1, Stream& stream = Stream::Null());
-//! adds scalar to a matrix (c = a + s)
-CV_EXPORTS void add(const GpuMat& a, const Scalar& sc, GpuMat& c, const GpuMat& mask = GpuMat(), int dtype = -1, Stream& stream = Stream::Null());
-
-//! subtracts one matrix from another (c = a - b)
-CV_EXPORTS void subtract(const GpuMat& a, const GpuMat& b, GpuMat& c, const GpuMat& mask = GpuMat(), int dtype = -1, Stream& stream = Stream::Null());
-//! subtracts scalar from a matrix (c = a - s)
-CV_EXPORTS void subtract(const GpuMat& a, const Scalar& sc, GpuMat& c, const GpuMat& mask = GpuMat(), int dtype = -1, Stream& stream = Stream::Null());
-
-//! computes element-wise weighted product of the two arrays (c = scale * a * b)
-CV_EXPORTS void multiply(const GpuMat& a, const GpuMat& b, GpuMat& c, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
-//! weighted multiplies matrix to a scalar (c = scale * a * s)
-CV_EXPORTS void multiply(const GpuMat& a, const Scalar& sc, GpuMat& c, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
-
-//! computes element-wise weighted quotient of the two arrays (c = a / b)
-CV_EXPORTS void divide(const GpuMat& a, const GpuMat& b, GpuMat& c, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
-//! computes element-wise weighted quotient of matrix and scalar (c = a / s)
-CV_EXPORTS void divide(const GpuMat& a, const Scalar& sc, GpuMat& c, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
-//! computes element-wise weighted reciprocal of an array (dst = scale/src2)
-CV_EXPORTS void divide(double scale, const GpuMat& b, GpuMat& c, int dtype = -1, Stream& stream = Stream::Null());
-
-//! computes the weighted sum of two arrays (dst = alpha*src1 + beta*src2 + gamma)
-CV_EXPORTS void addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2, double beta, double gamma, GpuMat& dst,
-                            int dtype = -1, Stream& stream = Stream::Null());
-
-//! adds scaled array to another one (dst = alpha*src1 + src2)
-static inline void scaleAdd(const GpuMat& src1, double alpha, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null())
-{
-    addWeighted(src1, alpha, src2, 1.0, 0.0, dst, -1, stream);
-}
-
-//! computes element-wise absolute difference of two arrays (c = abs(a - b))
-CV_EXPORTS void absdiff(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
-//! computes element-wise absolute difference of array and scalar (c = abs(a - s))
-CV_EXPORTS void absdiff(const GpuMat& a, const Scalar& s, GpuMat& c, Stream& stream = Stream::Null());
-
-//! computes absolute value of each matrix element
-//! supports CV_16S and CV_32F depth
-CV_EXPORTS void abs(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes square of each pixel in an image
-//! supports CV_8U, CV_16U, CV_16S and CV_32F depth
-CV_EXPORTS void sqr(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes square root of each pixel in an image
-//! supports CV_8U, CV_16U, CV_16S and CV_32F depth
-CV_EXPORTS void sqrt(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes exponent of each matrix element (b = e**a)
-//! supports CV_8U, CV_16U, CV_16S and CV_32F depth
-CV_EXPORTS void exp(const GpuMat& a, GpuMat& b, Stream& stream = Stream::Null());
-
-//! computes natural logarithm of absolute value of each matrix element: b = log(abs(a))
-//! supports CV_8U, CV_16U, CV_16S and CV_32F depth
-CV_EXPORTS void log(const GpuMat& a, GpuMat& b, Stream& stream = Stream::Null());
-
-//! computes power of each matrix element:
-//    (dst(i,j) = pow(     src(i,j) , power), if src.type() is integer
-//    (dst(i,j) = pow(fabs(src(i,j)), power), otherwise
-//! supports all, except depth == CV_64F
-CV_EXPORTS void pow(const GpuMat& src, double power, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! compares elements of two arrays (c = a <cmpop> b)
-CV_EXPORTS void compare(const GpuMat& a, const GpuMat& b, GpuMat& c, int cmpop, Stream& stream = Stream::Null());
-CV_EXPORTS void compare(const GpuMat& a, Scalar sc, GpuMat& c, int cmpop, Stream& stream = Stream::Null());
-
-//! performs per-elements bit-wise inversion
-CV_EXPORTS void bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
-
-//! calculates per-element bit-wise disjunction of two arrays
-CV_EXPORTS void bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
-//! calculates per-element bit-wise disjunction of array and scalar
-//! supports 1, 3 and 4 channels images with CV_8U, CV_16U or CV_32S depth
-CV_EXPORTS void bitwise_or(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! calculates per-element bit-wise conjunction of two arrays
-CV_EXPORTS void bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
-//! calculates per-element bit-wise conjunction of array and scalar
-//! supports 1, 3 and 4 channels images with CV_8U, CV_16U or CV_32S depth
-CV_EXPORTS void bitwise_and(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! calculates per-element bit-wise "exclusive or" operation
-CV_EXPORTS void bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
-//! calculates per-element bit-wise "exclusive or" of array and scalar
-//! supports 1, 3 and 4 channels images with CV_8U, CV_16U or CV_32S depth
-CV_EXPORTS void bitwise_xor(const GpuMat& src1, const Scalar& sc, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! pixel by pixel right shift of an image by a constant value
-//! supports 1, 3 and 4 channels images with integers elements
-CV_EXPORTS void rshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! pixel by pixel left shift of an image by a constant value
-//! supports 1, 3 and 4 channels images with CV_8U, CV_16U or CV_32S depth
-CV_EXPORTS void lshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes per-element minimum of two arrays (dst = min(src1, src2))
-CV_EXPORTS void min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes per-element minimum of array and scalar (dst = min(src1, src2))
-CV_EXPORTS void min(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes per-element maximum of two arrays (dst = max(src1, src2))
-CV_EXPORTS void max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! computes per-element maximum of array and scalar (dst = max(src1, src2))
-CV_EXPORTS void max(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null());
-
-enum { ALPHA_OVER, ALPHA_IN, ALPHA_OUT, ALPHA_ATOP, ALPHA_XOR, ALPHA_PLUS, ALPHA_OVER_PREMUL, ALPHA_IN_PREMUL, ALPHA_OUT_PREMUL,
-       ALPHA_ATOP_PREMUL, ALPHA_XOR_PREMUL, ALPHA_PLUS_PREMUL, ALPHA_PREMUL};
-
-//! Composite two images using alpha opacity values contained in each image
-//! Supports CV_8UC4, CV_16UC4, CV_32SC4 and CV_32FC4 types
-CV_EXPORTS void alphaComp(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, int alpha_op, Stream& stream = Stream::Null());
-
-
-////////////////////////////// Image processing //////////////////////////////
-
-//! DST[x,y] = SRC[xmap[x,y],ymap[x,y]]
-//! supports only CV_32FC1 map type
-CV_EXPORTS void remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap,
-                      int interpolation, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(),
-                      Stream& stream = Stream::Null());
-
-//! Does mean shift filtering on GPU.
-CV_EXPORTS void meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr,
-                                   TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
-                                   Stream& stream = Stream::Null());
-
-//! Does mean shift procedure on GPU.
-CV_EXPORTS void meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr,
-                              TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
-                              Stream& stream = Stream::Null());
-
-//! Does mean shift segmentation with elimination of small regions.
-CV_EXPORTS void meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize,
-                                      TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
-
-//! Does coloring of disparity image: [0..ndisp) -> [0..240, 1, 1] in HSV.
-//! Supported types of input disparity: CV_8U, CV_16S.
-//! Output disparity has CV_8UC4 type in BGRA format (alpha = 255).
-CV_EXPORTS void drawColorDisp(const GpuMat& src_disp, GpuMat& dst_disp, int ndisp, Stream& stream = Stream::Null());
-
-//! Reprojects disparity image to 3D space.
-//! Supports CV_8U and CV_16S types of input disparity.
-//! The output is a 3- or 4-channel floating-point matrix.
-//! Each element of this matrix will contain the 3D coordinates of the point (x,y,z,1), computed from the disparity map.
-//! Q is the 4x4 perspective transformation matrix that can be obtained with cvStereoRectify.
-CV_EXPORTS void reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, int dst_cn = 4, Stream& stream = Stream::Null());
-
-//! converts image from one color space to another
-CV_EXPORTS void cvtColor(const GpuMat& src, GpuMat& dst, int code, int dcn = 0, Stream& stream = Stream::Null());
-
-//! swap channels
-//! dstOrder - Integer array describing how channel values are permutated. The n-th entry
-//!            of the array contains the number of the channel that is stored in the n-th channel of
-//!            the output image. E.g. Given an RGBA image, aDstOrder = [3,2,1,0] converts this to ABGR
-//!            channel order.
-CV_EXPORTS void swapChannels(GpuMat& image, const int dstOrder[4], Stream& stream = Stream::Null());
-
-//! Routines for correcting image color gamma
-CV_EXPORTS void gammaCorrection(const GpuMat& src, GpuMat& dst, bool forward = true, Stream& stream = Stream::Null());
-
-//! applies fixed threshold to the image
-CV_EXPORTS double threshold(const GpuMat& src, GpuMat& dst, double thresh, double maxval, int type, Stream& stream = Stream::Null());
-
-//! resizes the image
-//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_AREA
-CV_EXPORTS void resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
-
-//! warps the image using affine transformation
-//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
-CV_EXPORTS void warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR,
-    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
-
-CV_EXPORTS void buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null());
-
-//! warps the image using perspective transformation
-//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
-CV_EXPORTS void warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR,
-    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
-
-CV_EXPORTS void buildWarpPerspectiveMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null());
-
-//! builds plane warping maps
-CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, const Mat &T, float scale,
-                                   GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
-
-//! builds cylindrical warping maps
-CV_EXPORTS void buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
-                                         GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
-
-//! builds spherical warping maps
-CV_EXPORTS void buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
-                                       GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
-
-//! rotates an image around the origin (0,0) and then shifts it
-//! supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
-//! supports 1, 3 or 4 channels images with CV_8U, CV_16U or CV_32F depth
-CV_EXPORTS void rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift = 0, double yShift = 0,
-                       int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
-
-//! copies 2D array to a larger destination array and pads borders with user-specifiable constant
-CV_EXPORTS void copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, int borderType,
-                               const Scalar& value = Scalar(), Stream& stream = Stream::Null());
-
-//! computes the integral image
-//! sum will have CV_32S type, but will contain unsigned int values
-//! supports only CV_8UC1 source type
-CV_EXPORTS void integral(const GpuMat& src, GpuMat& sum, Stream& stream = Stream::Null());
-//! buffered version
-CV_EXPORTS void integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer, Stream& stream = Stream::Null());
-
-//! computes squared integral image
-//! result matrix will have 64F type, but will contain 64U values
-//! supports source images of 8UC1 type only
-CV_EXPORTS void sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& stream = Stream::Null());
-
-//! computes vertical sum, supports only CV_32FC1 images
-CV_EXPORTS void columnSum(const GpuMat& src, GpuMat& sum);
-
-//! computes the standard deviation of integral images
-//! supports only CV_32SC1 source type and CV_32FC1 sqr type
-//! output will have CV_32FC1 type
-CV_EXPORTS void rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& stream = Stream::Null());
-
-//! computes Harris cornerness criteria at each image pixel
-CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType = BORDER_REFLECT101);
-CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType = BORDER_REFLECT101);
-CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize, double k,
-                             int borderType = BORDER_REFLECT101, Stream& stream = Stream::Null());
-
-//! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria
-CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType=BORDER_REFLECT101);
-CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType=BORDER_REFLECT101);
-CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize,
-    int borderType=BORDER_REFLECT101, Stream& stream = Stream::Null());
-
-//! performs per-element multiplication of two full (not packed) Fourier spectrums
-//! supports 32FC2 matrixes only (interleaved format)
-CV_EXPORTS void mulSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, bool conjB=false, Stream& stream = Stream::Null());
-
-//! performs per-element multiplication of two full (not packed) Fourier spectrums
-//! supports 32FC2 matrixes only (interleaved format)
-CV_EXPORTS void mulAndScaleSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, float scale, bool conjB=false, Stream& stream = Stream::Null());
-
-//! Performs a forward or inverse discrete Fourier transform (1D or 2D) of floating point matrix.
-//! Param dft_size is the size of DFT transform.
-//!
-//! If the source matrix is not continous, then additional copy will be done,
-//! so to avoid copying ensure the source matrix is continous one. If you want to use
-//! preallocated output ensure it is continuous too, otherwise it will be reallocated.
-//!
-//! Being implemented via CUFFT real-to-complex transform result contains only non-redundant values
-//! in CUFFT's format. Result as full complex matrix for such kind of transform cannot be retrieved.
-//!
-//! For complex-to-real transform it is assumed that the source matrix is packed in CUFFT's format.
-CV_EXPORTS void dft(const GpuMat& src, GpuMat& dst, Size dft_size, int flags=0, Stream& stream = Stream::Null());
-
-struct CV_EXPORTS ConvolveBuf
-{
-    Size result_size;
-    Size block_size;
-    Size user_block_size;
-    Size dft_size;
-    int spect_len;
-
-    GpuMat image_spect, templ_spect, result_spect;
-    GpuMat image_block, templ_block, result_data;
-
-    void create(Size image_size, Size templ_size);
-    static Size estimateBlockSize(Size result_size, Size templ_size);
-};
-
-
-//! computes convolution (or cross-correlation) of two images using discrete Fourier transform
-//! supports source images of 32FC1 type only
-//! result matrix will have 32FC1 type
-CV_EXPORTS void convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr = false);
-CV_EXPORTS void convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr, ConvolveBuf& buf, Stream& stream = Stream::Null());
-
-struct CV_EXPORTS MatchTemplateBuf
-{
-    Size user_block_size;
-    GpuMat imagef, templf;
-    std::vector<GpuMat> images;
-    std::vector<GpuMat> image_sums;
-    std::vector<GpuMat> image_sqsums;
-};
-
-//! computes the proximity map for the raster template and the image where the template is searched for
-CV_EXPORTS void matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, Stream &stream = Stream::Null());
-
-//! computes the proximity map for the raster template and the image where the template is searched for
-CV_EXPORTS void matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, MatchTemplateBuf &buf, Stream& stream = Stream::Null());
-
-//! smoothes the source image and downsamples it
-CV_EXPORTS void pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! upsamples the source image and then smoothes it
-CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-
-//! performs linear blending of two images
-//! to avoid accuracy errors sum of weigths shouldn't be very close to zero
-CV_EXPORTS void blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat& weights1, const GpuMat& weights2,
-                            GpuMat& result, Stream& stream = Stream::Null());
-
-//! Performa bilateral filtering of passsed image
-CV_EXPORTS void bilateralFilter(const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial,
-                                int borderMode = BORDER_DEFAULT, Stream& stream = Stream::Null());
-
-//! Brute force non-local means algorith (slow but universal)
-CV_EXPORTS void nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null());
-
-//! Fast (but approximate)version of non-local means algorith similar to CPU function (running sums technique)
-class CV_EXPORTS FastNonLocalMeansDenoising
-{
-public:
-    //! Simple method, recommended for grayscale images (though it supports multichannel images)
-    void simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null());
-
-    //! Processes luminance and color components separatelly
-    void labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null());
-
-private:
-
-    GpuMat buffer, extended_src_buffer;
-    GpuMat lab, l, ab;
-};
-
-struct CV_EXPORTS CannyBuf
-{
-    void create(const Size& image_size, int apperture_size = 3);
-    void release();
-
-    GpuMat dx, dy;
-    GpuMat mag;
-    GpuMat map;
-    GpuMat st1, st2;
-    Ptr<FilterEngine_GPU> filterDX, filterDY;
-};
-
-CV_EXPORTS void Canny(const GpuMat& image, GpuMat& edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false);
-CV_EXPORTS void Canny(const GpuMat& image, CannyBuf& buf, GpuMat& edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false);
-CV_EXPORTS void Canny(const GpuMat& dx, const GpuMat& dy, GpuMat& edges, double low_thresh, double high_thresh, bool L2gradient = false);
-CV_EXPORTS void Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& edges, double low_thresh, double high_thresh, bool L2gradient = false);
-
-class CV_EXPORTS ImagePyramid
-{
-public:
-    inline ImagePyramid() : nLayers_(0) {}
-    inline ImagePyramid(const GpuMat& img, int nLayers, Stream& stream = Stream::Null())
-    {
-        build(img, nLayers, stream);
-    }
-
-    void build(const GpuMat& img, int nLayers, Stream& stream = Stream::Null());
-
-    void getLayer(GpuMat& outImg, Size outRoi, Stream& stream = Stream::Null()) const;
-
-    inline void release()
-    {
-        layer0_.release();
-        pyramid_.clear();
-        nLayers_ = 0;
-    }
-
-private:
-    GpuMat layer0_;
-    std::vector<GpuMat> pyramid_;
-    int nLayers_;
-};
-
-//! HoughLines
-
-struct HoughLinesBuf
-{
-    GpuMat accum;
-    GpuMat list;
-};
-
-CV_EXPORTS void HoughLines(const GpuMat& src, GpuMat& lines, float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096);
-CV_EXPORTS void HoughLines(const GpuMat& src, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096);
-CV_EXPORTS void HoughLinesDownload(const GpuMat& d_lines, OutputArray h_lines, OutputArray h_votes = noArray());
-
-//! HoughLinesP
-
-//! finds line segments in the black-n-white image using probabalistic Hough transform
-CV_EXPORTS void HoughLinesP(const GpuMat& image, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int minLineLength, int maxLineGap, int maxLines = 4096);
-
-//! HoughCircles
-
-struct HoughCirclesBuf
-{
-    GpuMat edges;
-    GpuMat accum;
-    GpuMat list;
-    CannyBuf cannyBuf;
-};
-
-CV_EXPORTS void HoughCircles(const GpuMat& src, GpuMat& circles, int method, float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096);
-CV_EXPORTS void HoughCircles(const GpuMat& src, GpuMat& circles, HoughCirclesBuf& buf, int method, float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096);
-CV_EXPORTS void HoughCirclesDownload(const GpuMat& d_circles, OutputArray h_circles);
-
-//! finds arbitrary template in the grayscale image using Generalized Hough Transform
-//! Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
-//! Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
-class CV_EXPORTS GeneralizedHough_GPU : public cv::Algorithm
-{
-public:
-    static Ptr<GeneralizedHough_GPU> create(int method);
-
-    virtual ~GeneralizedHough_GPU();
-
-    //! set template to search
-    void setTemplate(const GpuMat& templ, int cannyThreshold = 100, Point templCenter = Point(-1, -1));
-    void setTemplate(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Point templCenter = Point(-1, -1));
-
-    //! find template on image
-    void detect(const GpuMat& image, GpuMat& positions, int cannyThreshold = 100);
-    void detect(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, GpuMat& positions);
-
-    void download(const GpuMat& d_positions, OutputArray h_positions, OutputArray h_votes = noArray());
-
-    void release();
-
-protected:
-    virtual void setTemplateImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Point templCenter) = 0;
-    virtual void detectImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, GpuMat& positions) = 0;
-    virtual void releaseImpl() = 0;
-
-private:
-    GpuMat edges_;
-    CannyBuf cannyBuf_;
-};
-
-////////////////////////////// Matrix reductions //////////////////////////////
-
-//! computes mean value and standard deviation of all or selected array elements
-//! supports only CV_8UC1 type
-CV_EXPORTS void meanStdDev(const GpuMat& mtx, Scalar& mean, Scalar& stddev);
-//! buffered version
-CV_EXPORTS void meanStdDev(const GpuMat& mtx, Scalar& mean, Scalar& stddev, GpuMat& buf);
-
-//! computes norm of array
-//! supports NORM_INF, NORM_L1, NORM_L2
-//! supports all matrices except 64F
-CV_EXPORTS double norm(const GpuMat& src1, int normType=NORM_L2);
-CV_EXPORTS double norm(const GpuMat& src1, int normType, GpuMat& buf);
-CV_EXPORTS double norm(const GpuMat& src1, int normType, const GpuMat& mask, GpuMat& buf);
-
-//! computes norm of the difference between two arrays
-//! supports NORM_INF, NORM_L1, NORM_L2
-//! supports only CV_8UC1 type
-CV_EXPORTS double norm(const GpuMat& src1, const GpuMat& src2, int normType=NORM_L2);
-
-//! computes sum of array elements
-//! supports only single channel images
-CV_EXPORTS Scalar sum(const GpuMat& src);
-CV_EXPORTS Scalar sum(const GpuMat& src, GpuMat& buf);
-CV_EXPORTS Scalar sum(const GpuMat& src, const GpuMat& mask, GpuMat& buf);
-
-//! computes sum of array elements absolute values
-//! supports only single channel images
-CV_EXPORTS Scalar absSum(const GpuMat& src);
-CV_EXPORTS Scalar absSum(const GpuMat& src, GpuMat& buf);
-CV_EXPORTS Scalar absSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf);
-
-//! computes squared sum of array elements
-//! supports only single channel images
-CV_EXPORTS Scalar sqrSum(const GpuMat& src);
-CV_EXPORTS Scalar sqrSum(const GpuMat& src, GpuMat& buf);
-CV_EXPORTS Scalar sqrSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf);
-
-//! finds global minimum and maximum array elements and returns their values
-CV_EXPORTS void minMax(const GpuMat& src, double* minVal, double* maxVal=0, const GpuMat& mask=GpuMat());
-CV_EXPORTS void minMax(const GpuMat& src, double* minVal, double* maxVal, const GpuMat& mask, GpuMat& buf);
-
-//! finds global minimum and maximum array elements and returns their values with locations
-CV_EXPORTS void minMaxLoc(const GpuMat& src, double* minVal, double* maxVal=0, Point* minLoc=0, Point* maxLoc=0,
-                          const GpuMat& mask=GpuMat());
-CV_EXPORTS void minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc,
-                          const GpuMat& mask, GpuMat& valbuf, GpuMat& locbuf);
-
-//! counts non-zero array elements
-CV_EXPORTS int countNonZero(const GpuMat& src);
-CV_EXPORTS int countNonZero(const GpuMat& src, GpuMat& buf);
-
-//! reduces a matrix to a vector
-CV_EXPORTS void reduce(const GpuMat& mtx, GpuMat& vec, int dim, int reduceOp, int dtype = -1, Stream& stream = Stream::Null());
-
-
-///////////////////////////// Calibration 3D //////////////////////////////////
-
-CV_EXPORTS void transformPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
-                                GpuMat& dst, Stream& stream = Stream::Null());
-
-CV_EXPORTS void projectPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
-                              const Mat& camera_mat, const Mat& dist_coef, GpuMat& dst,
-                              Stream& stream = Stream::Null());
-
-CV_EXPORTS void solvePnPRansac(const Mat& object, const Mat& image, const Mat& camera_mat,
-                               const Mat& dist_coef, Mat& rvec, Mat& tvec, bool use_extrinsic_guess=false,
-                               int num_iters=100, float max_dist=8.0, int min_inlier_count=100,
-                               std::vector<int>* inliers=NULL);
-
-//////////////////////////////// Image Labeling ////////////////////////////////
-
-//!performs labeling via graph cuts of a 2D regular 4-connected graph.
-CV_EXPORTS void graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& bottom, GpuMat& labels,
-                         GpuMat& buf, Stream& stream = Stream::Null());
-
-//!performs labeling via graph cuts of a 2D regular 8-connected graph.
-CV_EXPORTS void graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& topLeft, GpuMat& topRight,
-                         GpuMat& bottom, GpuMat& bottomLeft, GpuMat& bottomRight,
-                         GpuMat& labels,
-                         GpuMat& buf, Stream& stream = Stream::Null());
-
-//! compute mask for Generalized Flood fill componetns labeling.
-CV_EXPORTS void connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scalar& lo, const cv::Scalar& hi, Stream& stream = Stream::Null());
-
-//! performs connected componnents labeling.
-CV_EXPORTS void labelComponents(const GpuMat& mask, GpuMat& components, int flags = 0, Stream& stream = Stream::Null());
-
-////////////////////////////////// Histograms //////////////////////////////////
-
-//! Compute levels with even distribution. levels will have 1 row and nLevels cols and CV_32SC1 type.
-CV_EXPORTS void evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel);
-//! Calculates histogram with evenly distributed bins for signle channel source.
-//! Supports CV_8UC1, CV_16UC1 and CV_16SC1 source types.
-//! Output hist will have one row and histSize cols and CV_32SC1 type.
-CV_EXPORTS void histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());
-CV_EXPORTS void histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());
-//! Calculates histogram with evenly distributed bins for four-channel source.
-//! All channels of source are processed separately.
-//! Supports CV_8UC4, CV_16UC4 and CV_16SC4 source types.
-//! Output hist[i] will have one row and histSize[i] cols and CV_32SC1 type.
-CV_EXPORTS void histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());
-CV_EXPORTS void histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());
-//! Calculates histogram with bins determined by levels array.
-//! levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
-//! Supports CV_8UC1, CV_16UC1, CV_16SC1 and CV_32FC1 source types.
-//! Output hist will have one row and (levels.cols-1) cols and CV_32SC1 type.
-CV_EXPORTS void histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream = Stream::Null());
-CV_EXPORTS void histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream = Stream::Null());
-//! Calculates histogram with bins determined by levels array.
-//! All levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
-//! All channels of source are processed separately.
-//! Supports CV_8UC4, CV_16UC4, CV_16SC4 and CV_32FC4 source types.
-//! Output hist[i] will have one row and (levels[i].cols-1) cols and CV_32SC1 type.
-CV_EXPORTS void histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream = Stream::Null());
-CV_EXPORTS void histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream = Stream::Null());
-
-//! Calculates histogram for 8u one channel image
-//! Output hist will have one row, 256 cols and CV32SC1 type.
-CV_EXPORTS void calcHist(const GpuMat& src, GpuMat& hist, Stream& stream = Stream::Null());
-
-//! normalizes the grayscale image brightness and contrast by normalizing its histogram
-CV_EXPORTS void equalizeHist(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
-CV_EXPORTS void equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, GpuMat& buf, Stream& stream = Stream::Null());
-
-//////////////////////////////// StereoBM_GPU ////////////////////////////////
-
-class CV_EXPORTS StereoBM_GPU
-{
-public:
-    enum { BASIC_PRESET = 0, PREFILTER_XSOBEL = 1 };
-
-    enum { DEFAULT_NDISP = 64, DEFAULT_WINSZ = 19 };
-
-    //! the default constructor
-    StereoBM_GPU();
-    //! the full constructor taking the camera-specific preset, number of disparities and the SAD window size. ndisparities must be multiple of 8.
-    StereoBM_GPU(int preset, int ndisparities = DEFAULT_NDISP, int winSize = DEFAULT_WINSZ);
-
-    //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair
-    //! Output disparity has CV_8U type.
-    void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
-
-    //! Some heuristics that tries to estmate
-    // if current GPU will be faster than CPU in this algorithm.
-    // It queries current active device.
-    static bool checkIfGpuCallReasonable();
-
-    int preset;
-    int ndisp;
-    int winSize;
-
-    // If avergeTexThreshold  == 0 => post procesing is disabled
-    // If avergeTexThreshold != 0 then disparity is set 0 in each point (x,y) where for left image
-    // SumOfHorizontalGradiensInWindow(x, y, winSize) < (winSize * winSize) * avergeTexThreshold
-    // i.e. input left image is low textured.
-    float avergeTexThreshold;
-
-private:
-    GpuMat minSSD, leBuf, riBuf;
-};
-
-////////////////////////// StereoBeliefPropagation ///////////////////////////
-// "Efficient Belief Propagation for Early Vision"
-// P.Felzenszwalb
-
-class CV_EXPORTS StereoBeliefPropagation
-{
-public:
-    enum { DEFAULT_NDISP  = 64 };
-    enum { DEFAULT_ITERS  = 5  };
-    enum { DEFAULT_LEVELS = 5  };
-
-    static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels);
-
-    //! the default constructor
-    explicit StereoBeliefPropagation(int ndisp  = DEFAULT_NDISP,
-                                     int iters  = DEFAULT_ITERS,
-                                     int levels = DEFAULT_LEVELS,
-                                     int msg_type = CV_32F);
-
-    //! the full constructor taking the number of disparities, number of BP iterations on each level,
-    //! number of levels, truncation of data cost, data weight,
-    //! truncation of discontinuity cost and discontinuity single jump
-    //! DataTerm = data_weight * min(fabs(I2-I1), max_data_term)
-    //! DiscTerm = min(disc_single_jump * fabs(f1-f2), max_disc_term)
-    //! please see paper for more details
-    StereoBeliefPropagation(int ndisp, int iters, int levels,
-        float max_data_term, float data_weight,
-        float max_disc_term, float disc_single_jump,
-        int msg_type = CV_32F);
-
-    //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair,
-    //! if disparity is empty output type will be CV_16S else output type will be disparity.type().
-    void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
-
-
-    //! version for user specified data term
-    void operator()(const GpuMat& data, GpuMat& disparity, Stream& stream = Stream::Null());
-
-    int ndisp;
-
-    int iters;
-    int levels;
-
-    float max_data_term;
-    float data_weight;
-    float max_disc_term;
-    float disc_single_jump;
-
-    int msg_type;
-private:
-    GpuMat u, d, l, r, u2, d2, l2, r2;
-    std::vector<GpuMat> datas;
-    GpuMat out;
-};
-
-/////////////////////////// StereoConstantSpaceBP ///////////////////////////
-// "A Constant-Space Belief Propagation Algorithm for Stereo Matching"
-// Qingxiong Yang, Liang Wang, Narendra Ahuja
-// http://vision.ai.uiuc.edu/~qyang6/
-
-class CV_EXPORTS StereoConstantSpaceBP
-{
-public:
-    enum { DEFAULT_NDISP    = 128 };
-    enum { DEFAULT_ITERS    = 8   };
-    enum { DEFAULT_LEVELS   = 4   };
-    enum { DEFAULT_NR_PLANE = 4   };
-
-    static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane);
-
-    //! the default constructor
-    explicit StereoConstantSpaceBP(int ndisp    = DEFAULT_NDISP,
-                                   int iters    = DEFAULT_ITERS,
-                                   int levels   = DEFAULT_LEVELS,
-                                   int nr_plane = DEFAULT_NR_PLANE,
-                                   int msg_type = CV_32F);
-
-    //! the full constructor taking the number of disparities, number of BP iterations on each level,
-    //! number of levels, number of active disparity on the first level, truncation of data cost, data weight,
-    //! truncation of discontinuity cost, discontinuity single jump and minimum disparity threshold
-    StereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane,
-        float max_data_term, float data_weight, float max_disc_term, float disc_single_jump,
-        int min_disp_th = 0,
-        int msg_type = CV_32F);
-
-    //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair,
-    //! if disparity is empty output type will be CV_16S else output type will be disparity.type().
-    void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
-
-    int ndisp;
-
-    int iters;
-    int levels;
-
-    int nr_plane;
-
-    float max_data_term;
-    float data_weight;
-    float max_disc_term;
-    float disc_single_jump;
-
-    int min_disp_th;
-
-    int msg_type;
-
-    bool use_local_init_data_cost;
-private:
-    GpuMat messages_buffers;
-
-    GpuMat temp;
-    GpuMat out;
-};
-
-/////////////////////////// DisparityBilateralFilter ///////////////////////////
-// Disparity map refinement using joint bilateral filtering given a single color image.
-// Qingxiong Yang, Liang Wang, Narendra Ahuja
-// http://vision.ai.uiuc.edu/~qyang6/
-
-class CV_EXPORTS DisparityBilateralFilter
-{
-public:
-    enum { DEFAULT_NDISP  = 64 };
-    enum { DEFAULT_RADIUS = 3 };
-    enum { DEFAULT_ITERS  = 1 };
-
-    //! the default constructor
-    explicit DisparityBilateralFilter(int ndisp = DEFAULT_NDISP, int radius = DEFAULT_RADIUS, int iters = DEFAULT_ITERS);
-
-    //! the full constructor taking the number of disparities, filter radius,
-    //! number of iterations, truncation of data continuity, truncation of disparity continuity
-    //! and filter range sigma
-    DisparityBilateralFilter(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold, float sigma_range);
-
-    //! the disparity map refinement operator. Refine disparity map using joint bilateral filtering given a single color image.
-    //! disparity must have CV_8U or CV_16S type, image must have CV_8UC1 or CV_8UC3 type.
-    void operator()(const GpuMat& disparity, const GpuMat& image, GpuMat& dst, Stream& stream = Stream::Null());
-
-private:
-    int ndisp;
-    int radius;
-    int iters;
-
-    float edge_threshold;
-    float max_disc_threshold;
-    float sigma_range;
-
-    GpuMat table_color;
-    GpuMat table_space;
-};
-
 
 //////////////// HOG (Histogram-of-Oriented-Gradients) Descriptor and Object Detector //////////////
+
 struct CV_EXPORTS HOGConfidence
 {
    double scale;
@@ -1178,195 +183,18 @@ protected:
     std::vector<GpuMat> image_scales;
 };
 
+//////////////////////////// CascadeClassifier ////////////////////////////
 
-////////////////////////////////// BruteForceMatcher //////////////////////////////////
-
-class CV_EXPORTS BFMatcher_GPU
-{
-public:
-    explicit BFMatcher_GPU(int norm = cv::NORM_L2);
-
-    // Add descriptors to train descriptor collection
-    void add(const std::vector<GpuMat>& descCollection);
-
-    // Get train descriptors collection
-    const std::vector<GpuMat>& getTrainDescriptors() const;
-
-    // Clear train descriptors collection
-    void clear();
-
-    // Return true if there are not train descriptors in collection
-    bool empty() const;
-
-    // Return true if the matcher supports mask in match methods
-    bool isMaskSupported() const;
-
-    // Find one best match for each query descriptor
-    void matchSingle(const GpuMat& query, const GpuMat& train,
-        GpuMat& trainIdx, GpuMat& distance,
-        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
-
-    // Download trainIdx and distance and convert it to CPU vector with DMatch
-    static void matchDownload(const GpuMat& trainIdx, const GpuMat& distance, std::vector<DMatch>& matches);
-    // Convert trainIdx and distance to vector with DMatch
-    static void matchConvert(const Mat& trainIdx, const Mat& distance, std::vector<DMatch>& matches);
-
-    // Find one best match for each query descriptor
-    void match(const GpuMat& query, const GpuMat& train, std::vector<DMatch>& matches, const GpuMat& mask = GpuMat());
-
-    // Make gpu collection of trains and masks in suitable format for matchCollection function
-    void makeGpuCollection(GpuMat& trainCollection, GpuMat& maskCollection, const std::vector<GpuMat>& masks = std::vector<GpuMat>());
-
-    // Find one best match from train collection for each query descriptor
-    void matchCollection(const GpuMat& query, const GpuMat& trainCollection,
-        GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance,
-        const GpuMat& masks = GpuMat(), Stream& stream = Stream::Null());
-
-    // Download trainIdx, imgIdx and distance and convert it to vector with DMatch
-    static void matchDownload(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance, std::vector<DMatch>& matches);
-    // Convert trainIdx, imgIdx and distance to vector with DMatch
-    static void matchConvert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance, std::vector<DMatch>& matches);
-
-    // Find one best match from train collection for each query descriptor.
-    void match(const GpuMat& query, std::vector<DMatch>& matches, const std::vector<GpuMat>& masks = std::vector<GpuMat>());
-
-    // Find k best matches for each query descriptor (in increasing order of distances)
-    void knnMatchSingle(const GpuMat& query, const GpuMat& train,
-        GpuMat& trainIdx, GpuMat& distance, GpuMat& allDist, int k,
-        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
-
-    // Download trainIdx and distance and convert it to vector with DMatch
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    static void knnMatchDownload(const GpuMat& trainIdx, const GpuMat& distance,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-    // Convert trainIdx and distance to vector with DMatch
-    static void knnMatchConvert(const Mat& trainIdx, const Mat& distance,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-
-    // Find k best matches for each query descriptor (in increasing order of distances).
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    void knnMatch(const GpuMat& query, const GpuMat& train,
-        std::vector< std::vector<DMatch> >& matches, int k, const GpuMat& mask = GpuMat(),
-        bool compactResult = false);
-
-    // Find k best matches from train collection for each query descriptor (in increasing order of distances)
-    void knnMatch2Collection(const GpuMat& query, const GpuMat& trainCollection,
-        GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance,
-        const GpuMat& maskCollection = GpuMat(), Stream& stream = Stream::Null());
-
-    // Download trainIdx and distance and convert it to vector with DMatch
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    static void knnMatch2Download(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-    // Convert trainIdx and distance to vector with DMatch
-    static void knnMatch2Convert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-
-    // Find k best matches  for each query descriptor (in increasing order of distances).
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    void knnMatch(const GpuMat& query, std::vector< std::vector<DMatch> >& matches, int k,
-        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false);
-
-    // Find best matches for each query descriptor which have distance less than maxDistance.
-    // nMatches.at<int>(0, queryIdx) will contain matches count for queryIdx.
-    // carefully nMatches can be greater than trainIdx.cols - it means that matcher didn't find all matches,
-    // because it didn't have enough memory.
-    // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nTrain / 100), 10),
-    // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
-    // Matches doesn't sorted.
-    void radiusMatchSingle(const GpuMat& query, const GpuMat& train,
-        GpuMat& trainIdx, GpuMat& distance, GpuMat& nMatches, float maxDistance,
-        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
-
-    // Download trainIdx, nMatches and distance and convert it to vector with DMatch.
-    // matches will be sorted in increasing order of distances.
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    static void radiusMatchDownload(const GpuMat& trainIdx, const GpuMat& distance, const GpuMat& nMatches,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-    // Convert trainIdx, nMatches and distance to vector with DMatch.
-    static void radiusMatchConvert(const Mat& trainIdx, const Mat& distance, const Mat& nMatches,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-
-    // Find best matches for each query descriptor which have distance less than maxDistance
-    // in increasing order of distances).
-    void radiusMatch(const GpuMat& query, const GpuMat& train,
-        std::vector< std::vector<DMatch> >& matches, float maxDistance,
-        const GpuMat& mask = GpuMat(), bool compactResult = false);
-
-    // Find best matches for each query descriptor which have distance less than maxDistance.
-    // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nQuery / 100), 10),
-    // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
-    // Matches doesn't sorted.
-    void radiusMatchCollection(const GpuMat& query, GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance, GpuMat& nMatches, float maxDistance,
-        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), Stream& stream = Stream::Null());
-
-    // Download trainIdx, imgIdx, nMatches and distance and convert it to vector with DMatch.
-    // matches will be sorted in increasing order of distances.
-    // compactResult is used when mask is not empty. If compactResult is false matches
-    // vector will have the same size as queryDescriptors rows. If compactResult is true
-    // matches vector will not contain matches for fully masked out query descriptors.
-    static void radiusMatchDownload(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance, const GpuMat& nMatches,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-    // Convert trainIdx, nMatches and distance to vector with DMatch.
-    static void radiusMatchConvert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance, const Mat& nMatches,
-        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
-
-    // Find best matches from train collection for each query descriptor which have distance less than
-    // maxDistance (in increasing order of distances).
-    void radiusMatch(const GpuMat& query, std::vector< std::vector<DMatch> >& matches, float maxDistance,
-        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false);
-
-    int norm;
-
-private:
-    std::vector<GpuMat> trainDescCollection;
-};
-
-template <class Distance>
-class CV_EXPORTS BruteForceMatcher_GPU;
-
-template <typename T>
-class CV_EXPORTS BruteForceMatcher_GPU< L1<T> > : public BFMatcher_GPU
-{
-public:
-    explicit BruteForceMatcher_GPU() : BFMatcher_GPU(NORM_L1) {}
-    explicit BruteForceMatcher_GPU(L1<T> /*d*/) : BFMatcher_GPU(NORM_L1) {}
-};
-template <typename T>
-class CV_EXPORTS BruteForceMatcher_GPU< L2<T> > : public BFMatcher_GPU
-{
-public:
-    explicit BruteForceMatcher_GPU() : BFMatcher_GPU(NORM_L2) {}
-    explicit BruteForceMatcher_GPU(L2<T> /*d*/) : BFMatcher_GPU(NORM_L2) {}
-};
-template <> class CV_EXPORTS BruteForceMatcher_GPU< Hamming > : public BFMatcher_GPU
-{
-public:
-    explicit BruteForceMatcher_GPU() : BFMatcher_GPU(NORM_HAMMING) {}
-    explicit BruteForceMatcher_GPU(Hamming /*d*/) : BFMatcher_GPU(NORM_HAMMING) {}
-};
-
-////////////////////////////////// CascadeClassifier_GPU //////////////////////////////////////////
 // The cascade classifier class for object detection: supports old haar and new lbp xlm formats and nvbin for haar cascades olny.
 class CV_EXPORTS CascadeClassifier_GPU
 {
 public:
     CascadeClassifier_GPU();
-    CascadeClassifier_GPU(const std::string& filename);
+    CascadeClassifier_GPU(const String& filename);
     ~CascadeClassifier_GPU();
 
     bool empty() const;
-    bool load(const std::string& filename);
+    bool load(const String& filename);
     void release();
 
     /* returns number of detected objects */
@@ -1386,1059 +214,39 @@ private:
     friend class CascadeClassifier_GPU_LBP;
 };
 
-////////////////////////////////// SURF //////////////////////////////////////////
+//////////////////////////// Labeling ////////////////////////////
 
-class CV_EXPORTS SURF_GPU
-{
-public:
-    enum KeypointLayout
-    {
-        X_ROW = 0,
-        Y_ROW,
-        LAPLACIAN_ROW,
-        OCTAVE_ROW,
-        SIZE_ROW,
-        ANGLE_ROW,
-        HESSIAN_ROW,
-        ROWS_COUNT
-    };
+//!performs labeling via graph cuts of a 2D regular 4-connected graph.
+CV_EXPORTS void graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& bottom, GpuMat& labels,
+                         GpuMat& buf, Stream& stream = Stream::Null());
 
-    //! the default constructor
-    SURF_GPU();
-    //! the full constructor taking all the necessary parameters
-    explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
-         int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
+//!performs labeling via graph cuts of a 2D regular 8-connected graph.
+CV_EXPORTS void graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& topLeft, GpuMat& topRight,
+                         GpuMat& bottom, GpuMat& bottomLeft, GpuMat& bottomRight,
+                         GpuMat& labels,
+                         GpuMat& buf, Stream& stream = Stream::Null());
 
-    //! returns the descriptor size in float's (64 or 128)
-    int descriptorSize() const;
+//! compute mask for Generalized Flood fill componetns labeling.
+CV_EXPORTS void connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scalar& lo, const cv::Scalar& hi, Stream& stream = Stream::Null());
 
-    //! upload host keypoints to device memory
-    static void uploadKeypoints(const std::vector<KeyPoint>& keypoints, GpuMat& keypointsGPU);
-    //! download keypoints from device to host memory
-    static void downloadKeypoints(const GpuMat& keypointsGPU, std::vector<KeyPoint>& keypoints);
+//! performs connected componnents labeling.
+CV_EXPORTS void labelComponents(const GpuMat& mask, GpuMat& components, int flags = 0, Stream& stream = Stream::Null());
 
-    //! download descriptors from device to host memory
-    static void downloadDescriptors(const GpuMat& descriptorsGPU, std::vector<float>& descriptors);
+//////////////////////////// Calib3d ////////////////////////////
 
-    //! finds the keypoints using fast hessian detector used in SURF
-    //! supports CV_8UC1 images
-    //! keypoints will have nFeature cols and 6 rows
-    //! keypoints.ptr<float>(X_ROW)[i] will contain x coordinate of i'th feature
-    //! keypoints.ptr<float>(Y_ROW)[i] will contain y coordinate of i'th feature
-    //! keypoints.ptr<float>(LAPLACIAN_ROW)[i] will contain laplacian sign of i'th feature
-    //! keypoints.ptr<float>(OCTAVE_ROW)[i] will contain octave of i'th feature
-    //! keypoints.ptr<float>(SIZE_ROW)[i] will contain size of i'th feature
-    //! keypoints.ptr<float>(ANGLE_ROW)[i] will contain orientation of i'th feature
-    //! keypoints.ptr<float>(HESSIAN_ROW)[i] will contain response of i'th feature
-    void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints);
-    //! finds the keypoints and computes their descriptors.
-    //! Optionally it can compute descriptors for the user-provided keypoints and recompute keypoints direction
-    void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors,
-        bool useProvidedKeypoints = false);
+CV_EXPORTS void transformPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
+                                GpuMat& dst, Stream& stream = Stream::Null());
 
-    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
-    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
-        bool useProvidedKeypoints = false);
+CV_EXPORTS void projectPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
+                              const Mat& camera_mat, const Mat& dist_coef, GpuMat& dst,
+                              Stream& stream = Stream::Null());
 
-    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, std::vector<float>& descriptors,
-        bool useProvidedKeypoints = false);
+CV_EXPORTS void solvePnPRansac(const Mat& object, const Mat& image, const Mat& camera_mat,
+                               const Mat& dist_coef, Mat& rvec, Mat& tvec, bool use_extrinsic_guess=false,
+                               int num_iters=100, float max_dist=8.0, int min_inlier_count=100,
+                               std::vector<int>* inliers=NULL);
 
-    void releaseMemory();
-
-    // SURF parameters
-    double hessianThreshold;
-    int nOctaves;
-    int nOctaveLayers;
-    bool extended;
-    bool upright;
-
-    //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
-    float keypointsRatio;
-
-    GpuMat sum, mask1, maskSum, intBuffer;
-
-    GpuMat det, trace;
-
-    GpuMat maxPosBuffer;
-};
-
-////////////////////////////////// FAST //////////////////////////////////////////
-
-class CV_EXPORTS FAST_GPU
-{
-public:
-    enum
-    {
-        LOCATION_ROW = 0,
-        RESPONSE_ROW,
-        ROWS_COUNT
-    };
-
-    // all features have same size
-    static const int FEATURE_SIZE = 7;
-
-    explicit FAST_GPU(int threshold, bool nonmaxSupression = true, double keypointsRatio = 0.05);
-
-    //! finds the keypoints using FAST detector
-    //! supports only CV_8UC1 images
-    void operator ()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints);
-    void operator ()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
-
-    //! download keypoints from device to host memory
-    static void downloadKeypoints(const GpuMat& d_keypoints, std::vector<KeyPoint>& keypoints);
-
-    //! convert keypoints to KeyPoint vector
-    static void convertKeypoints(const Mat& h_keypoints, std::vector<KeyPoint>& keypoints);
-
-    //! release temporary buffer's memory
-    void release();
-
-    bool nonmaxSupression;
-
-    int threshold;
-
-    //! max keypoints = keypointsRatio * img.size().area()
-    double keypointsRatio;
-
-    //! find keypoints and compute it's response if nonmaxSupression is true
-    //! return count of detected keypoints
-    int calcKeyPointsLocation(const GpuMat& image, const GpuMat& mask);
-
-    //! get final array of keypoints
-    //! performs nonmax supression if needed
-    //! return final count of keypoints
-    int getKeyPoints(GpuMat& keypoints);
-
-private:
-    GpuMat kpLoc_;
-    int count_;
-
-    GpuMat score_;
-
-    GpuMat d_keypoints_;
-};
-
-////////////////////////////////// ORB //////////////////////////////////////////
-
-class CV_EXPORTS ORB_GPU
-{
-public:
-    enum
-    {
-        X_ROW = 0,
-        Y_ROW,
-        RESPONSE_ROW,
-        ANGLE_ROW,
-        OCTAVE_ROW,
-        SIZE_ROW,
-        ROWS_COUNT
-    };
-
-    enum
-    {
-        DEFAULT_FAST_THRESHOLD = 20
-    };
-
-    //! Constructor
-    explicit ORB_GPU(int nFeatures = 500, float scaleFactor = 1.2f, int nLevels = 8, int edgeThreshold = 31,
-                     int firstLevel = 0, int WTA_K = 2, int scoreType = 0, int patchSize = 31);
-
-    //! Compute the ORB features on an image
-    //! image - the image to compute the features (supports only CV_8UC1 images)
-    //! mask - the mask to apply
-    //! keypoints - the resulting keypoints
-    void operator()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
-    void operator()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints);
-
-    //! Compute the ORB features and descriptors on an image
-    //! image - the image to compute the features (supports only CV_8UC1 images)
-    //! mask - the mask to apply
-    //! keypoints - the resulting keypoints
-    //! descriptors - descriptors array
-    void operator()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors);
-    void operator()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors);
-
-    //! download keypoints from device to host memory
-    static void downloadKeyPoints(const GpuMat& d_keypoints, std::vector<KeyPoint>& keypoints);
-    //! convert keypoints to KeyPoint vector
-    static void convertKeyPoints(const Mat& d_keypoints, std::vector<KeyPoint>& keypoints);
-
-    //! returns the descriptor size in bytes
-    inline int descriptorSize() const { return kBytes; }
-
-    inline void setFastParams(int threshold, bool nonmaxSupression = true)
-    {
-        fastDetector_.threshold = threshold;
-        fastDetector_.nonmaxSupression = nonmaxSupression;
-    }
-
-    //! release temporary buffer's memory
-    void release();
-
-    //! if true, image will be blurred before descriptors calculation
-    bool blurForDescriptor;
-
-private:
-    enum { kBytes = 32 };
-
-    void buildScalePyramids(const GpuMat& image, const GpuMat& mask);
-
-    void computeKeyPointsPyramid();
-
-    void computeDescriptors(GpuMat& descriptors);
-
-    void mergeKeyPoints(GpuMat& keypoints);
-
-    int nFeatures_;
-    float scaleFactor_;
-    int nLevels_;
-    int edgeThreshold_;
-    int firstLevel_;
-    int WTA_K_;
-    int scoreType_;
-    int patchSize_;
-
-    // The number of desired features per scale
-    std::vector<size_t> n_features_per_level_;
-
-    // Points to compute BRIEF descriptors from
-    GpuMat pattern_;
-
-    std::vector<GpuMat> imagePyr_;
-    std::vector<GpuMat> maskPyr_;
-
-    GpuMat buf_;
-
-    std::vector<GpuMat> keyPointsPyr_;
-    std::vector<int> keyPointsCount_;
-
-    FAST_GPU fastDetector_;
-
-    Ptr<FilterEngine_GPU> blurFilter;
-
-    GpuMat d_keypoints_;
-};
-
-////////////////////////////////// Optical Flow //////////////////////////////////////////
-
-class CV_EXPORTS BroxOpticalFlow
-{
-public:
-    BroxOpticalFlow(float alpha_, float gamma_, float scale_factor_, int inner_iterations_, int outer_iterations_, int solver_iterations_) :
-        alpha(alpha_), gamma(gamma_), scale_factor(scale_factor_),
-        inner_iterations(inner_iterations_), outer_iterations(outer_iterations_), solver_iterations(solver_iterations_)
-    {
-    }
-
-    //! Compute optical flow
-    //! frame0 - source frame (supports only CV_32FC1 type)
-    //! frame1 - frame to track (with the same size and type as frame0)
-    //! u      - flow horizontal component (along x axis)
-    //! v      - flow vertical component (along y axis)
-    void operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& stream = Stream::Null());
-
-    //! flow smoothness
-    float alpha;
-
-    //! gradient constancy importance
-    float gamma;
-
-    //! pyramid scale factor
-    float scale_factor;
-
-    //! number of lagged non-linearity iterations (inner loop)
-    int inner_iterations;
-
-    //! number of warping iterations (number of pyramid levels)
-    int outer_iterations;
-
-    //! number of linear system solver iterations
-    int solver_iterations;
-
-    GpuMat buf;
-};
-
-class CV_EXPORTS GoodFeaturesToTrackDetector_GPU
-{
-public:
-    explicit GoodFeaturesToTrackDetector_GPU(int maxCorners = 1000, double qualityLevel = 0.01, double minDistance = 0.0,
-        int blockSize = 3, bool useHarrisDetector = false, double harrisK = 0.04);
-
-    //! return 1 rows matrix with CV_32FC2 type
-    void operator ()(const GpuMat& image, GpuMat& corners, const GpuMat& mask = GpuMat());
-
-    int maxCorners;
-    double qualityLevel;
-    double minDistance;
-
-    int blockSize;
-    bool useHarrisDetector;
-    double harrisK;
-
-    void releaseMemory()
-    {
-        Dx_.release();
-        Dy_.release();
-        buf_.release();
-        eig_.release();
-        minMaxbuf_.release();
-        tmpCorners_.release();
-    }
-
-private:
-    GpuMat Dx_;
-    GpuMat Dy_;
-    GpuMat buf_;
-    GpuMat eig_;
-    GpuMat minMaxbuf_;
-    GpuMat tmpCorners_;
-};
-
-inline GoodFeaturesToTrackDetector_GPU::GoodFeaturesToTrackDetector_GPU(int maxCorners_, double qualityLevel_, double minDistance_,
-        int blockSize_, bool useHarrisDetector_, double harrisK_)
-{
-    maxCorners = maxCorners_;
-    qualityLevel = qualityLevel_;
-    minDistance = minDistance_;
-    blockSize = blockSize_;
-    useHarrisDetector = useHarrisDetector_;
-    harrisK = harrisK_;
-}
-
-
-class CV_EXPORTS PyrLKOpticalFlow
-{
-public:
-    PyrLKOpticalFlow();
-
-    void sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts,
-        GpuMat& status, GpuMat* err = 0);
-
-    void dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0);
-
-    void releaseMemory();
-
-    Size winSize;
-    int maxLevel;
-    int iters;
-    bool useInitialFlow;
-
-private:
-    std::vector<GpuMat> prevPyr_;
-    std::vector<GpuMat> nextPyr_;
-
-    GpuMat buf_;
-
-    GpuMat uPyr_[2];
-    GpuMat vPyr_[2];
-};
-
-
-class CV_EXPORTS FarnebackOpticalFlow
-{
-public:
-    FarnebackOpticalFlow()
-    {
-        numLevels = 5;
-        pyrScale = 0.5;
-        fastPyramids = false;
-        winSize = 13;
-        numIters = 10;
-        polyN = 5;
-        polySigma = 1.1;
-        flags = 0;
-    }
-
-    int numLevels;
-    double pyrScale;
-    bool fastPyramids;
-    int winSize;
-    int numIters;
-    int polyN;
-    double polySigma;
-    int flags;
-
-    void operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null());
-
-    void releaseMemory()
-    {
-        frames_[0].release();
-        frames_[1].release();
-        pyrLevel_[0].release();
-        pyrLevel_[1].release();
-        M_.release();
-        bufM_.release();
-        R_[0].release();
-        R_[1].release();
-        blurredFrame_[0].release();
-        blurredFrame_[1].release();
-        pyramid0_.clear();
-        pyramid1_.clear();
-    }
-
-private:
-    void prepareGaussian(
-            int n, double sigma, float *g, float *xg, float *xxg,
-            double &ig11, double &ig03, double &ig33, double &ig55);
-
-    void setPolynomialExpansionConsts(int n, double sigma);
-
-    void updateFlow_boxFilter(
-            const GpuMat& R0, const GpuMat& R1, GpuMat& flowx, GpuMat &flowy,
-            GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[]);
-
-    void updateFlow_gaussianBlur(
-            const GpuMat& R0, const GpuMat& R1, GpuMat& flowx, GpuMat& flowy,
-            GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[]);
-
-    GpuMat frames_[2];
-    GpuMat pyrLevel_[2], M_, bufM_, R_[2], blurredFrame_[2];
-    std::vector<GpuMat> pyramid0_, pyramid1_;
-};
-
-
-// Implementation of the Zach, Pock and Bischof Dual TV-L1 Optical Flow method
-//
-// see reference:
-//   [1] C. Zach, T. Pock and H. Bischof, "A Duality Based Approach for Realtime TV-L1 Optical Flow".
-//   [2] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
-class CV_EXPORTS OpticalFlowDual_TVL1_GPU
-{
-public:
-    OpticalFlowDual_TVL1_GPU();
-
-    void operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy);
-
-    void collectGarbage();
-
-    /**
-     * Time step of the numerical scheme.
-     */
-    double tau;
-
-    /**
-     * Weight parameter for the data term, attachment parameter.
-     * This is the most relevant parameter, which determines the smoothness of the output.
-     * The smaller this parameter is, the smoother the solutions we obtain.
-     * It depends on the range of motions of the images, so its value should be adapted to each image sequence.
-     */
-    double lambda;
-
-    /**
-     * Weight parameter for (u - v)^2, tightness parameter.
-     * It serves as a link between the attachment and the regularization terms.
-     * In theory, it should have a small value in order to maintain both parts in correspondence.
-     * The method is stable for a large range of values of this parameter.
-     */
-    double theta;
-
-    /**
-     * Number of scales used to create the pyramid of images.
-     */
-    int nscales;
-
-    /**
-     * Number of warpings per scale.
-     * Represents the number of times that I1(x+u0) and grad( I1(x+u0) ) are computed per scale.
-     * This is a parameter that assures the stability of the method.
-     * It also affects the running time, so it is a compromise between speed and accuracy.
-     */
-    int warps;
-
-    /**
-     * Stopping criterion threshold used in the numerical scheme, which is a trade-off between precision and running time.
-     * A small value will yield more accurate solutions at the expense of a slower convergence.
-     */
-    double epsilon;
-
-    /**
-     * Stopping criterion iterations number used in the numerical scheme.
-     */
-    int iterations;
-
-    bool useInitialFlow;
-
-private:
-    void procOneScale(const GpuMat& I0, const GpuMat& I1, GpuMat& u1, GpuMat& u2);
-
-    std::vector<GpuMat> I0s;
-    std::vector<GpuMat> I1s;
-    std::vector<GpuMat> u1s;
-    std::vector<GpuMat> u2s;
-
-    GpuMat I1x_buf;
-    GpuMat I1y_buf;
-
-    GpuMat I1w_buf;
-    GpuMat I1wx_buf;
-    GpuMat I1wy_buf;
-
-    GpuMat grad_buf;
-    GpuMat rho_c_buf;
-
-    GpuMat p11_buf;
-    GpuMat p12_buf;
-    GpuMat p21_buf;
-    GpuMat p22_buf;
-
-    GpuMat diff_buf;
-    GpuMat norm_buf;
-};
-
-
-//! Calculates optical flow for 2 images using block matching algorithm */
-CV_EXPORTS void calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr,
-                                  Size block_size, Size shift_size, Size max_range, bool use_previous,
-                                  GpuMat& velx, GpuMat& vely, GpuMat& buf,
-                                  Stream& stream = Stream::Null());
-
-class CV_EXPORTS FastOpticalFlowBM
-{
-public:
-    void operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy, int search_window = 21, int block_window = 7, Stream& s = Stream::Null());
-
-private:
-    GpuMat buffer;
-    GpuMat extended_I0;
-    GpuMat extended_I1;
-};
-
-
-//! Interpolate frames (images) using provided optical flow (displacement field).
-//! frame0   - frame 0 (32-bit floating point images, single channel)
-//! frame1   - frame 1 (the same type and size)
-//! fu       - forward horizontal displacement
-//! fv       - forward vertical displacement
-//! bu       - backward horizontal displacement
-//! bv       - backward vertical displacement
-//! pos      - new frame position
-//! newFrame - new frame
-//! buf      - temporary buffer, will have width x 6*height size, CV_32FC1 type and contain 6 GpuMat;
-//!            occlusion masks            0, occlusion masks            1,
-//!            interpolated forward flow  0, interpolated forward flow  1,
-//!            interpolated backward flow 0, interpolated backward flow 1
-//!
-CV_EXPORTS void interpolateFrames(const GpuMat& frame0, const GpuMat& frame1,
-                                  const GpuMat& fu, const GpuMat& fv,
-                                  const GpuMat& bu, const GpuMat& bv,
-                                  float pos, GpuMat& newFrame, GpuMat& buf,
-                                  Stream& stream = Stream::Null());
-
-CV_EXPORTS void createOpticalFlowNeedleMap(const GpuMat& u, const GpuMat& v, GpuMat& vertex, GpuMat& colors);
-
-
-//////////////////////// Background/foreground segmentation ////////////////////////
-
-// Foreground Object Detection from Videos Containing Complex Background.
-// Liyuan Li, Weimin Huang, Irene Y.H. Gu, and Qi Tian.
-// ACM MM2003 9p
-class CV_EXPORTS FGDStatModel
-{
-public:
-    struct CV_EXPORTS Params
-    {
-        int Lc;  // Quantized levels per 'color' component. Power of two, typically 32, 64 or 128.
-        int N1c; // Number of color vectors used to model normal background color variation at a given pixel.
-        int N2c; // Number of color vectors retained at given pixel.  Must be > N1c, typically ~ 5/3 of N1c.
-        // Used to allow the first N1c vectors to adapt over time to changing background.
-
-        int Lcc;  // Quantized levels per 'color co-occurrence' component.  Power of two, typically 16, 32 or 64.
-        int N1cc; // Number of color co-occurrence vectors used to model normal background color variation at a given pixel.
-        int N2cc; // Number of color co-occurrence vectors retained at given pixel.  Must be > N1cc, typically ~ 5/3 of N1cc.
-        // Used to allow the first N1cc vectors to adapt over time to changing background.
-
-        bool is_obj_without_holes; // If TRUE we ignore holes within foreground blobs. Defaults to TRUE.
-        int perform_morphing;     // Number of erode-dilate-erode foreground-blob cleanup iterations.
-        // These erase one-pixel junk blobs and merge almost-touching blobs. Default value is 1.
-
-        float alpha1; // How quickly we forget old background pixel values seen. Typically set to 0.1.
-        float alpha2; // "Controls speed of feature learning". Depends on T. Typical value circa 0.005.
-        float alpha3; // Alternate to alpha2, used (e.g.) for quicker initial convergence. Typical value 0.1.
-
-        float delta;   // Affects color and color co-occurrence quantization, typically set to 2.
-        float T;       // A percentage value which determines when new features can be recognized as new background. (Typically 0.9).
-        float minArea; // Discard foreground blobs whose bounding box is smaller than this threshold.
-
-        // default Params
-        Params();
-    };
-
-    // out_cn - channels count in output result (can be 3 or 4)
-    // 4-channels require more memory, but a bit faster
-    explicit FGDStatModel(int out_cn = 3);
-    explicit FGDStatModel(const cv::gpu::GpuMat& firstFrame, const Params& params = Params(), int out_cn = 3);
-
-    ~FGDStatModel();
-
-    void create(const cv::gpu::GpuMat& firstFrame, const Params& params = Params());
-    void release();
-
-    int update(const cv::gpu::GpuMat& curFrame);
-
-    //8UC3 or 8UC4 reference background image
-    cv::gpu::GpuMat background;
-
-    //8UC1 foreground image
-    cv::gpu::GpuMat foreground;
-
-    std::vector< std::vector<cv::Point> > foreground_regions;
-
-private:
-    FGDStatModel(const FGDStatModel&);
-    FGDStatModel& operator=(const FGDStatModel&);
-
-    class Impl;
-    std::auto_ptr<Impl> impl_;
-};
-
-/*!
- Gaussian Mixture-based Backbround/Foreground Segmentation Algorithm
-
- The class implements the following algorithm:
- "An improved adaptive background mixture model for real-time tracking with shadow detection"
- P. KadewTraKuPong and R. Bowden,
- Proc. 2nd European Workshp on Advanced Video-Based Surveillance Systems, 2001."
- http://personal.ee.surrey.ac.uk/Personal/R.Bowden/publications/avbs01/avbs01.pdf
-*/
-class CV_EXPORTS MOG_GPU
-{
-public:
-    //! the default constructor
-    MOG_GPU(int nmixtures = -1);
-
-    //! re-initiaization method
-    void initialize(Size frameSize, int frameType);
-
-    //! the update operator
-    void operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate = 0.0f, Stream& stream = Stream::Null());
-
-    //! computes a background image which are the mean of all background gaussians
-    void getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const;
-
-    //! releases all inner buffers
-    void release();
-
-    int history;
-    float varThreshold;
-    float backgroundRatio;
-    float noiseSigma;
-
-private:
-    int nmixtures_;
-
-    Size frameSize_;
-    int frameType_;
-    int nframes_;
-
-    GpuMat weight_;
-    GpuMat sortKey_;
-    GpuMat mean_;
-    GpuMat var_;
-};
-
-/*!
- The class implements the following algorithm:
- "Improved adaptive Gausian mixture model for background subtraction"
- Z.Zivkovic
- International Conference Pattern Recognition, UK, August, 2004.
- http://www.zoranz.net/Publications/zivkovic2004ICPR.pdf
-*/
-class CV_EXPORTS MOG2_GPU
-{
-public:
-    //! the default constructor
-    MOG2_GPU(int nmixtures = -1);
-
-    //! re-initiaization method
-    void initialize(Size frameSize, int frameType);
-
-    //! the update operator
-    void operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate = -1.0f, Stream& stream = Stream::Null());
-
-    //! computes a background image which are the mean of all background gaussians
-    void getBackgroundImage(GpuMat& backgroundImage, Stream& stream = Stream::Null()) const;
-
-    //! releases all inner buffers
-    void release();
-
-    // parameters
-    // you should call initialize after parameters changes
-
-    int history;
-
-    //! here it is the maximum allowed number of mixture components.
-    //! Actual number is determined dynamically per pixel
-    float varThreshold;
-    // threshold on the squared Mahalanobis distance to decide if it is well described
-    // by the background model or not. Related to Cthr from the paper.
-    // This does not influence the update of the background. A typical value could be 4 sigma
-    // and that is varThreshold=4*4=16; Corresponds to Tb in the paper.
-
-    /////////////////////////
-    // less important parameters - things you might change but be carefull
-    ////////////////////////
-
-    float backgroundRatio;
-    // corresponds to fTB=1-cf from the paper
-    // TB - threshold when the component becomes significant enough to be included into
-    // the background model. It is the TB=1-cf from the paper. So I use cf=0.1 => TB=0.
-    // For alpha=0.001 it means that the mode should exist for approximately 105 frames before
-    // it is considered foreground
-    // float noiseSigma;
-    float varThresholdGen;
-
-    //correspondts to Tg - threshold on the squared Mahalan. dist. to decide
-    //when a sample is close to the existing components. If it is not close
-    //to any a new component will be generated. I use 3 sigma => Tg=3*3=9.
-    //Smaller Tg leads to more generated components and higher Tg might make
-    //lead to small number of components but they can grow too large
-    float fVarInit;
-    float fVarMin;
-    float fVarMax;
-
-    //initial variance  for the newly generated components.
-    //It will will influence the speed of adaptation. A good guess should be made.
-    //A simple way is to estimate the typical standard deviation from the images.
-    //I used here 10 as a reasonable value
-    // min and max can be used to further control the variance
-    float fCT; //CT - complexity reduction prior
-    //this is related to the number of samples needed to accept that a component
-    //actually exists. We use CT=0.05 of all the samples. By setting CT=0 you get
-    //the standard Stauffer&Grimson algorithm (maybe not exact but very similar)
-
-    //shadow detection parameters
-    bool bShadowDetection; //default 1 - do shadow detection
-    unsigned char nShadowDetection; //do shadow detection - insert this value as the detection result - 127 default value
-    float fTau;
-    // Tau - shadow threshold. The shadow is detected if the pixel is darker
-    //version of the background. Tau is a threshold on how much darker the shadow can be.
-    //Tau= 0.5 means that if pixel is more than 2 times darker then it is not shadow
-    //See: Prati,Mikic,Trivedi,Cucchiarra,"Detecting Moving Shadows...",IEEE PAMI,2003.
-
-private:
-    int nmixtures_;
-
-    Size frameSize_;
-    int frameType_;
-    int nframes_;
-
-    GpuMat weight_;
-    GpuMat variance_;
-    GpuMat mean_;
-
-    GpuMat bgmodelUsedModes_; //keep track of number of modes per pixel
-};
-
-/*!
- * The class implements the following algorithm:
- * "ViBe: A universal background subtraction algorithm for video sequences"
- * O. Barnich and M. Van D Roogenbroeck
- * IEEE Transactions on Image Processing, 20(6) :1709-1724, June 2011
- */
-class CV_EXPORTS VIBE_GPU
-{
-public:
-    //! the default constructor
-    explicit VIBE_GPU(unsigned long rngSeed = 1234567);
-
-    //! re-initiaization method
-    void initialize(const GpuMat& firstFrame, Stream& stream = Stream::Null());
-
-    //! the update operator
-    void operator()(const GpuMat& frame, GpuMat& fgmask, Stream& stream = Stream::Null());
-
-    //! releases all inner buffers
-    void release();
-
-    int nbSamples;         // number of samples per pixel
-    int reqMatches;        // #_min
-    int radius;            // R
-    int subsamplingFactor; // amount of random subsampling
-
-private:
-    Size frameSize_;
-
-    unsigned long rngSeed_;
-    GpuMat randStates_;
-
-    GpuMat samples_;
-};
-
-/**
- * Background Subtractor module. Takes a series of images and returns a sequence of mask (8UC1)
- * images of the same size, where 255 indicates Foreground and 0 represents Background.
- * This class implements an algorithm described in "Visual Tracking of Human Visitors under
- * Variable-Lighting Conditions for a Responsive Audio Art Installation," A. Godbehere,
- * A. Matsukawa, K. Goldberg, American Control Conference, Montreal, June 2012.
- */
-class CV_EXPORTS GMG_GPU
-{
-public:
-    GMG_GPU();
-
-    /**
-     * Validate parameters and set up data structures for appropriate frame size.
-     * @param frameSize Input frame size
-     * @param min       Minimum value taken on by pixels in image sequence. Usually 0
-     * @param max       Maximum value taken on by pixels in image sequence. e.g. 1.0 or 255
-     */
-    void initialize(Size frameSize, float min = 0.0f, float max = 255.0f);
-
-    /**
-     * Performs single-frame background subtraction and builds up a statistical background image
-     * model.
-     * @param frame        Input frame
-     * @param fgmask       Output mask image representing foreground and background pixels
-     * @param stream       Stream for the asynchronous version
-     */
-    void operator ()(const GpuMat& frame, GpuMat& fgmask, float learningRate = -1.0f, Stream& stream = Stream::Null());
-
-    //! Releases all inner buffers
-    void release();
-
-    //! Total number of distinct colors to maintain in histogram.
-    int maxFeatures;
-
-    //! Set between 0.0 and 1.0, determines how quickly features are "forgotten" from histograms.
-    float learningRate;
-
-    //! Number of frames of video to use to initialize histograms.
-    int numInitializationFrames;
-
-    //! Number of discrete levels in each channel to be used in histograms.
-    int quantizationLevels;
-
-    //! Prior probability that any given pixel is a background pixel. A sensitivity parameter.
-    float backgroundPrior;
-
-    //! Value above which pixel is determined to be FG.
-    float decisionThreshold;
-
-    //! Smoothing radius, in pixels, for cleaning up FG image.
-    int smoothingRadius;
-
-    //! Perform background model update.
-    bool updateBackgroundModel;
-
-private:
-    float maxVal_, minVal_;
-
-    Size frameSize_;
-
-    int frameNum_;
-
-    GpuMat nfeatures_;
-    GpuMat colors_;
-    GpuMat weights_;
-
-    Ptr<FilterEngine_GPU> boxFilter_;
-    GpuMat buf_;
-};
-
-////////////////////////////////// Video Encoding //////////////////////////////////
-
-// Works only under Windows
-// Supports olny H264 video codec and AVI files
-class CV_EXPORTS VideoWriter_GPU
-{
-public:
-    struct EncoderParams;
-
-    // Callbacks for video encoder, use it if you want to work with raw video stream
-    class EncoderCallBack;
-
-    enum SurfaceFormat
-    {
-        SF_UYVY = 0,
-        SF_YUY2,
-        SF_YV12,
-        SF_NV12,
-        SF_IYUV,
-        SF_BGR,
-        SF_GRAY = SF_BGR
-    };
-
-    VideoWriter_GPU();
-    VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR);
-    VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
-    VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR);
-    VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
-    ~VideoWriter_GPU();
-
-    // all methods throws cv::Exception if error occurs
-    void open(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR);
-    void open(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
-    void open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format = SF_BGR);
-    void open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
-
-    bool isOpened() const;
-    void close();
-
-    void write(const cv::gpu::GpuMat& image, bool lastFrame = false);
-
-    struct CV_EXPORTS EncoderParams
-    {
-        int       P_Interval;      //    NVVE_P_INTERVAL,
-        int       IDR_Period;      //    NVVE_IDR_PERIOD,
-        int       DynamicGOP;      //    NVVE_DYNAMIC_GOP,
-        int       RCType;          //    NVVE_RC_TYPE,
-        int       AvgBitrate;      //    NVVE_AVG_BITRATE,
-        int       PeakBitrate;     //    NVVE_PEAK_BITRATE,
-        int       QP_Level_Intra;  //    NVVE_QP_LEVEL_INTRA,
-        int       QP_Level_InterP; //    NVVE_QP_LEVEL_INTER_P,
-        int       QP_Level_InterB; //    NVVE_QP_LEVEL_INTER_B,
-        int       DeblockMode;     //    NVVE_DEBLOCK_MODE,
-        int       ProfileLevel;    //    NVVE_PROFILE_LEVEL,
-        int       ForceIntra;      //    NVVE_FORCE_INTRA,
-        int       ForceIDR;        //    NVVE_FORCE_IDR,
-        int       ClearStat;       //    NVVE_CLEAR_STAT,
-        int       DIMode;          //    NVVE_SET_DEINTERLACE,
-        int       Presets;         //    NVVE_PRESETS,
-        int       DisableCabac;    //    NVVE_DISABLE_CABAC,
-        int       NaluFramingType; //    NVVE_CONFIGURE_NALU_FRAMING_TYPE
-        int       DisableSPSPPS;   //    NVVE_DISABLE_SPS_PPS
-
-        EncoderParams();
-        explicit EncoderParams(const std::string& configFile);
-
-        void load(const std::string& configFile);
-        void save(const std::string& configFile) const;
-    };
-
-    EncoderParams getParams() const;
-
-    class CV_EXPORTS EncoderCallBack
-    {
-    public:
-        enum PicType
-        {
-            IFRAME = 1,
-            PFRAME = 2,
-            BFRAME = 3
-        };
-
-        virtual ~EncoderCallBack() {}
-
-        // callback function to signal the start of bitstream that is to be encoded
-        // must return pointer to buffer
-        virtual uchar* acquireBitStream(int* bufferSize) = 0;
-
-        // callback function to signal that the encoded bitstream is ready to be written to file
-        virtual void releaseBitStream(unsigned char* data, int size) = 0;
-
-        // callback function to signal that the encoding operation on the frame has started
-        virtual void onBeginFrame(int frameNumber, PicType picType) = 0;
-
-        // callback function signals that the encoding operation on the frame has finished
-        virtual void onEndFrame(int frameNumber, PicType picType) = 0;
-    };
-
-private:
-    VideoWriter_GPU(const VideoWriter_GPU&);
-    VideoWriter_GPU& operator=(const VideoWriter_GPU&);
-
-    class Impl;
-    std::auto_ptr<Impl> impl_;
-};
-
-
-////////////////////////////////// Video Decoding //////////////////////////////////////////
-
-namespace detail
-{
-    class FrameQueue;
-    class VideoParser;
-}
-
-class CV_EXPORTS VideoReader_GPU
-{
-public:
-    enum Codec
-    {
-        MPEG1 = 0,
-        MPEG2,
-        MPEG4,
-        VC1,
-        H264,
-        JPEG,
-        H264_SVC,
-        H264_MVC,
-
-        Uncompressed_YUV420 = (('I'<<24)|('Y'<<16)|('U'<<8)|('V')),   // Y,U,V (4:2:0)
-        Uncompressed_YV12   = (('Y'<<24)|('V'<<16)|('1'<<8)|('2')),   // Y,V,U (4:2:0)
-        Uncompressed_NV12   = (('N'<<24)|('V'<<16)|('1'<<8)|('2')),   // Y,UV  (4:2:0)
-        Uncompressed_YUYV   = (('Y'<<24)|('U'<<16)|('Y'<<8)|('V')),   // YUYV/YUY2 (4:2:2)
-        Uncompressed_UYVY   = (('U'<<24)|('Y'<<16)|('V'<<8)|('Y')),   // UYVY (4:2:2)
-    };
-
-    enum ChromaFormat
-    {
-        Monochrome=0,
-        YUV420,
-        YUV422,
-        YUV444,
-    };
-
-    struct FormatInfo
-    {
-        Codec codec;
-        ChromaFormat chromaFormat;
-        int width;
-        int height;
-    };
-
-    class VideoSource;
-
-    VideoReader_GPU();
-    explicit VideoReader_GPU(const std::string& filename);
-    explicit VideoReader_GPU(const cv::Ptr<VideoSource>& source);
-
-    ~VideoReader_GPU();
-
-    void open(const std::string& filename);
-    void open(const cv::Ptr<VideoSource>& source);
-    bool isOpened() const;
-
-    void close();
-
-    bool read(GpuMat& image);
-
-    FormatInfo format() const;
-    void dumpFormat(std::ostream& st);
-
-    class CV_EXPORTS VideoSource
-    {
-    public:
-        VideoSource() : frameQueue_(0), videoParser_(0) {}
-        virtual ~VideoSource() {}
-
-        virtual FormatInfo format() const = 0;
-        virtual void start() = 0;
-        virtual void stop() = 0;
-        virtual bool isStarted() const = 0;
-        virtual bool hasError() const = 0;
-
-        void setFrameQueue(detail::FrameQueue* frameQueue) { frameQueue_ = frameQueue; }
-        void setVideoParser(detail::VideoParser* videoParser) { videoParser_ = videoParser; }
-
-    protected:
-        bool parseVideoData(const uchar* data, size_t size, bool endOfStream = false);
-
-    private:
-        VideoSource(const VideoSource&);
-        VideoSource& operator =(const VideoSource&);
-
-        detail::FrameQueue* frameQueue_;
-        detail::VideoParser* videoParser_;
-    };
-
-private:
-    VideoReader_GPU(const VideoReader_GPU&);
-    VideoReader_GPU& operator =(const VideoReader_GPU&);
-
-    class Impl;
-    std::auto_ptr<Impl> impl_;
-};
+//////////////////////////// VStab ////////////////////////////
 
 //! removes points (CV_32FC2, single row matrix) with zero mask value
 CV_EXPORTS void compactPoints(GpuMat &points0, GpuMat &points1, const GpuMat &mask);
@@ -2447,8 +255,6 @@ CV_EXPORTS void calcWobbleSuppressionMaps(
         int left, int idx, int right, Size size, const Mat &ml, const Mat &mr,
         GpuMat &mapx, GpuMat &mapy);
 
-} // namespace gpu
-
-} // namespace cv
+}} // namespace cv { namespace gpu {
 
 #endif /* __OPENCV_GPU_HPP__ */
diff --git a/modules/gpu/include/opencv2/gpu/device/emulation.hpp b/modules/gpu/include/opencv2/gpu/device/emulation.hpp
deleted file mode 100644
index b6fba230e7..0000000000
--- a/modules/gpu/include/opencv2/gpu/device/emulation.hpp
+++ /dev/null
@@ -1,138 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef OPENCV_GPU_EMULATION_HPP_
-#define OPENCV_GPU_EMULATION_HPP_
-
-#include "warp_reduce.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    struct Emulation
-    {
-
-        static __device__ __forceinline__ int syncthreadsOr(int pred)
-        {
-#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 200)
-                // just campilation stab
-                return 0;
-#else
-                return __syncthreads_or(pred);
-#endif
-        }
-
-        template<int CTA_SIZE>
-        static __forceinline__ __device__ int Ballot(int predicate)
-        {
-#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ >= 200)
-            return __ballot(predicate);
-#else
-            __shared__ volatile int cta_buffer[CTA_SIZE];
-
-            int tid = threadIdx.x;
-            cta_buffer[tid] = predicate ? (1 << (tid & 31)) : 0;
-            return warp_reduce(cta_buffer);
-#endif
-        }
-
-        struct smem
-        {
-            enum { TAG_MASK = (1U << ( (sizeof(unsigned int) << 3) - 5U)) - 1U };
-
-            template<typename T>
-            static __device__ __forceinline__ T atomicInc(T* address, T val)
-            {
-#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
-                T count;
-                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
-                do
-                {
-                    count = *address & TAG_MASK;
-                    count = tag | (count + 1);
-                    *address = count;
-                } while (*address != count);
-
-                return (count & TAG_MASK) - 1;
-#else
-                return ::atomicInc(address, val);
-#endif
-            }
-
-            template<typename T>
-            static __device__ __forceinline__ T atomicAdd(T* address, T val)
-            {
-#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
-                T count;
-                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
-                do
-                {
-                    count = *address & TAG_MASK;
-                    count = tag | (count + val);
-                    *address = count;
-                } while (*address != count);
-
-                return (count & TAG_MASK) - val;
-#else
-                return ::atomicAdd(address, val);
-#endif
-            }
-
-            template<typename T>
-            static __device__ __forceinline__ T atomicMin(T* address, T val)
-            {
-#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
-                T count = ::min(*address, val);
-                do
-                {
-                    *address = count;
-                } while (*address > count);
-
-                return count;
-#else
-                return ::atomicMin(address, val);
-#endif
-            }
-        };
-    };
-}}} // namespace cv { namespace gpu { namespace device
-
-#endif /* OPENCV_GPU_EMULATION_HPP_ */
diff --git a/modules/gpu/include/opencv2/gpu/device/limits.hpp b/modules/gpu/include/opencv2/gpu/device/limits.hpp
deleted file mode 100644
index b040f199d6..0000000000
--- a/modules/gpu/include/opencv2/gpu/device/limits.hpp
+++ /dev/null
@@ -1,235 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_GPU_LIMITS_GPU_HPP__
-#define __OPENCV_GPU_LIMITS_GPU_HPP__
-
-#include <limits>
-#include "common.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    template<class T> struct numeric_limits
-    {
-        typedef T type;
-        __device__ __forceinline__ static type min()  { return type(); };
-        __device__ __forceinline__ static type max() { return type(); };
-        __device__ __forceinline__ static type epsilon() { return type(); }
-        __device__ __forceinline__ static type round_error() { return type(); }
-        __device__ __forceinline__ static type denorm_min()  { return type(); }
-        __device__ __forceinline__ static type infinity() { return type(); }
-        __device__ __forceinline__ static type quiet_NaN() { return type(); }
-        __device__ __forceinline__ static type signaling_NaN() { return T(); }
-        static const bool is_signed;
-    };
-
-    template<> struct numeric_limits<bool>
-    {
-        typedef bool type;
-        __device__ __forceinline__ static type min() { return false; };
-        __device__ __forceinline__ static type max() { return true;  };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = false;
-    };
-
-    template<> struct numeric_limits<char>
-    {
-        typedef char type;
-        __device__ __forceinline__ static type min() { return CHAR_MIN; };
-        __device__ __forceinline__ static type max() { return CHAR_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = (char)-1 == -1;
-    };
-
-    template<> struct numeric_limits<signed char>
-    {
-        typedef char type;
-        __device__ __forceinline__ static type min() { return SCHAR_MIN; };
-        __device__ __forceinline__ static type max() { return SCHAR_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = (signed char)-1 == -1;
-    };
-
-    template<> struct numeric_limits<unsigned char>
-    {
-        typedef unsigned char type;
-        __device__ __forceinline__ static type min() { return 0; };
-        __device__ __forceinline__ static type max() { return UCHAR_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = false;
-    };
-
-    template<> struct numeric_limits<short>
-    {
-        typedef short type;
-        __device__ __forceinline__ static type min() { return SHRT_MIN; };
-        __device__ __forceinline__ static type max() { return SHRT_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = true;
-    };
-
-    template<> struct numeric_limits<unsigned short>
-    {
-        typedef unsigned short type;
-        __device__ __forceinline__ static type min() { return 0; };
-        __device__ __forceinline__ static type max() { return USHRT_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = false;
-    };
-
-    template<> struct numeric_limits<int>
-    {
-        typedef int type;
-        __device__ __forceinline__ static type min() { return INT_MIN; };
-        __device__ __forceinline__ static type max() { return INT_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = true;
-    };
-
-
-    template<> struct numeric_limits<unsigned int>
-    {
-        typedef unsigned int type;
-        __device__ __forceinline__ static type min() { return 0; };
-        __device__ __forceinline__ static type max() { return UINT_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = false;
-    };
-
-    template<> struct numeric_limits<long>
-    {
-        typedef long type;
-        __device__ __forceinline__ static type min() { return LONG_MIN; };
-        __device__ __forceinline__ static type max() { return LONG_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = true;
-    };
-
-    template<> struct numeric_limits<unsigned long>
-    {
-        typedef unsigned long type;
-        __device__ __forceinline__ static type min() { return 0; };
-        __device__ __forceinline__ static type max() { return ULONG_MAX; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = false;
-    };
-
-    template<> struct numeric_limits<float>
-    {
-        typedef float type;
-        __device__ __forceinline__ static type min() { return 1.175494351e-38f/*FLT_MIN*/; };
-        __device__ __forceinline__ static type max() { return 3.402823466e+38f/*FLT_MAX*/; };
-        __device__ __forceinline__ static type epsilon() { return 1.192092896e-07f/*FLT_EPSILON*/; };
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = true;
-    };
-
-    template<> struct numeric_limits<double>
-    {
-        typedef double type;
-        __device__ __forceinline__ static type min() { return 2.2250738585072014e-308/*DBL_MIN*/; };
-        __device__ __forceinline__ static type max() { return 1.7976931348623158e+308/*DBL_MAX*/; };
-        __device__ __forceinline__ static type epsilon();
-        __device__ __forceinline__ static type round_error();
-        __device__ __forceinline__ static type denorm_min();
-        __device__ __forceinline__ static type infinity();
-        __device__ __forceinline__ static type quiet_NaN();
-        __device__ __forceinline__ static type signaling_NaN();
-        static const bool is_signed = true;
-    };
-}}} // namespace cv { namespace gpu { namespace device {
-
-#endif // __OPENCV_GPU_LIMITS_GPU_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/device/vec_math.hpp b/modules/gpu/include/opencv2/gpu/device/vec_math.hpp
deleted file mode 100644
index 1c46dc0c33..0000000000
--- a/modules/gpu/include/opencv2/gpu/device/vec_math.hpp
+++ /dev/null
@@ -1,330 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_GPU_VECMATH_HPP__
-#define __OPENCV_GPU_VECMATH_HPP__
-
-#include "saturate_cast.hpp"
-#include "vec_traits.hpp"
-#include "functional.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace vec_math_detail
-    {
-        template <int cn, typename VecD> struct SatCastHelper;
-        template <typename VecD> struct SatCastHelper<1, VecD>
-        {
-            template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
-            {
-                typedef typename VecTraits<VecD>::elem_type D;
-                return VecTraits<VecD>::make(saturate_cast<D>(v.x));
-            }
-        };
-        template <typename VecD> struct SatCastHelper<2, VecD>
-        {
-            template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
-            {
-                typedef typename VecTraits<VecD>::elem_type D;
-                return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y));
-            }
-        };
-        template <typename VecD> struct SatCastHelper<3, VecD>
-        {
-            template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
-            {
-                typedef typename VecTraits<VecD>::elem_type D;
-                return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z));
-            }
-        };
-        template <typename VecD> struct SatCastHelper<4, VecD>
-        {
-            template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
-            {
-                typedef typename VecTraits<VecD>::elem_type D;
-                return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z), saturate_cast<D>(v.w));
-            }
-        };
-
-        template <typename VecD, typename VecS> static __device__ __forceinline__ VecD saturate_cast_caller(const VecS& v)
-        {
-            return SatCastHelper<VecTraits<VecD>::cn, VecD>::cast(v);
-        }
-    }
-
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uchar1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const char1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const ushort1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const short1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uint1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const int1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const float1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const double1& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uchar2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const char2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const ushort2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const short2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uint2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const int2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const float2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const double2& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uchar3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const char3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const ushort3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const short3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uint3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const int3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const float3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const double3& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uchar4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const char4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const ushort4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const short4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const uint4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const int4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const float4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-    template<typename _Tp> static __device__ __forceinline__ _Tp saturate_cast(const double4& v) {return vec_math_detail::saturate_cast_caller<_Tp>(v);}
-
-#define OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, op, func) \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 1>::vec_type op(const type ## 1 & a) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 1>::vec_type>::make(f(a.x)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 2>::vec_type op(const type ## 2 & a) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 2>::vec_type>::make(f(a.x), f(a.y)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 3>::vec_type op(const type ## 3 & a) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 3>::vec_type>::make(f(a.x), f(a.y), f(a.z)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 4>::vec_type op(const type ## 4 & a) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 4>::vec_type>::make(f(a.x), f(a.y), f(a.z), f(a.w)); \
-    }
-
-    namespace vec_math_detail
-    {
-        template <typename T1, typename T2> struct BinOpTraits
-        {
-            typedef int argument_type;
-        };
-        template <typename T> struct BinOpTraits<T, T>
-        {
-            typedef T argument_type;
-        };
-        template <typename T> struct BinOpTraits<T, double>
-        {
-            typedef double argument_type;
-        };
-        template <typename T> struct BinOpTraits<double, T>
-        {
-            typedef double argument_type;
-        };
-        template <> struct BinOpTraits<double, double>
-        {
-            typedef double argument_type;
-        };
-        template <typename T> struct BinOpTraits<T, float>
-        {
-            typedef float argument_type;
-        };
-        template <typename T> struct BinOpTraits<float, T>
-        {
-            typedef float argument_type;
-        };
-        template <> struct BinOpTraits<float, float>
-        {
-            typedef float argument_type;
-        };
-        template <> struct BinOpTraits<double, float>
-        {
-            typedef double argument_type;
-        };
-        template <> struct BinOpTraits<float, double>
-        {
-            typedef double argument_type;
-        };
-    }
-
-#define OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, op, func) \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 1>::vec_type op(const type ## 1 & a, const type ## 1 & b) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 1>::vec_type>::make(f(a.x, b.x)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 1>::vec_type op(const type ## 1 & v, T s) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 1>::vec_type>::make(f(v.x, s)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 1>::vec_type op(T s, const type ## 1 & v) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 1>::vec_type>::make(f(s, v.x)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 2>::vec_type op(const type ## 2 & a, const type ## 2 & b) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 2>::vec_type>::make(f(a.x, b.x), f(a.y, b.y)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 2>::vec_type op(const type ## 2 & v, T s) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 2>::vec_type>::make(f(v.x, s), f(v.y, s)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 2>::vec_type op(T s, const type ## 2 & v) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 2>::vec_type>::make(f(s, v.x), f(s, v.y)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 3>::vec_type op(const type ## 3 & a, const type ## 3 & b) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 3>::vec_type>::make(f(a.x, b.x), f(a.y, b.y), f(a.z, b.z)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 3>::vec_type op(const type ## 3 & v, T s) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 3>::vec_type>::make(f(v.x, s), f(v.y, s), f(v.z, s)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 3>::vec_type op(T s, const type ## 3 & v) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 3>::vec_type>::make(f(s, v.x), f(s, v.y), f(s, v.z)); \
-    } \
-    __device__ __forceinline__ TypeVec<func<type>::result_type, 4>::vec_type op(const type ## 4 & a, const type ## 4 & b) \
-    { \
-        func<type> f; \
-        return VecTraits<TypeVec<func<type>::result_type, 4>::vec_type>::make(f(a.x, b.x), f(a.y, b.y), f(a.z, b.z), f(a.w, b.w)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 4>::vec_type op(const type ## 4 & v, T s) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<type, T>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 4>::vec_type>::make(f(v.x, s), f(v.y, s), f(v.z, s), f(v.w, s)); \
-    } \
-    template <typename T> \
-    __device__ __forceinline__ typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 4>::vec_type op(T s, const type ## 4 & v) \
-    { \
-        func<typename vec_math_detail::BinOpTraits<T, type>::argument_type> f; \
-        return VecTraits<typename TypeVec<typename func<typename vec_math_detail::BinOpTraits<type, T>::argument_type>::result_type, 4>::vec_type>::make(f(s, v.x), f(s, v.y), f(s, v.z), f(s, v.w)); \
-    }
-
-#define OPENCV_GPU_IMPLEMENT_VEC_OP(type) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator +, plus) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator -, minus) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator *, multiplies) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator /, divides) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP (type, operator -, negate) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator ==, equal_to) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator !=, not_equal_to) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator > , greater) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator < , less) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator >=, greater_equal) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator <=, less_equal) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator &&, logical_and) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator ||, logical_or) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP (type, operator ! , logical_not) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, max, maximum) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, min, minimum) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, abs, abs_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, sqrt, sqrt_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, exp, exp_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, exp2, exp2_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, exp10, exp10_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, log, log_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, log2, log2_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, log10, log10_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, sin, sin_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, cos, cos_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, tan, tan_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, asin, asin_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, acos, acos_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, atan, atan_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, sinh, sinh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, cosh, cosh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, tanh, tanh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, asinh, asinh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, acosh, acosh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, atanh, atanh_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, hypot, hypot_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, atan2, atan2_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, pow, pow_func) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, hypot_sqr, hypot_sqr_func)
-
-#define OPENCV_GPU_IMPLEMENT_VEC_INT_OP(type) \
-    OPENCV_GPU_IMPLEMENT_VEC_OP(type) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator &, bit_and) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator |, bit_or) \
-    OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, operator ^, bit_xor) \
-    OPENCV_GPU_IMPLEMENT_VEC_UNOP (type, operator ~, bit_not)
-
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(uchar)
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(char)
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(ushort)
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(short)
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(int)
-    OPENCV_GPU_IMPLEMENT_VEC_INT_OP(uint)
-    OPENCV_GPU_IMPLEMENT_VEC_OP(float)
-    OPENCV_GPU_IMPLEMENT_VEC_OP(double)
-
-    #undef OPENCV_GPU_IMPLEMENT_VEC_UNOP
-    #undef OPENCV_GPU_IMPLEMENT_VEC_BINOP
-    #undef OPENCV_GPU_IMPLEMENT_VEC_OP
-    #undef OPENCV_GPU_IMPLEMENT_VEC_INT_OP
-}}} // namespace cv { namespace gpu { namespace device
-
-#endif // __OPENCV_GPU_VECMATH_HPP__
diff --git a/modules/gpu/include/opencv2/gpu/gpu.hpp b/modules/gpu/include/opencv2/gpu/gpu.hpp
index d82211bf38..e044ad1e8a 100644
--- a/modules/gpu/include/opencv2/gpu/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu/gpu.hpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/gpu.hpp"
\ No newline at end of file
+#include "opencv2/gpu.hpp"
diff --git a/modules/gpu/include/opencv2/gpu/gpumat.hpp b/modules/gpu/include/opencv2/gpu/gpumat.hpp
deleted file mode 100644
index 0033cbe961..0000000000
--- a/modules/gpu/include/opencv2/gpu/gpumat.hpp
+++ /dev/null
@@ -1,43 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "opencv2/core/gpumat.hpp"
diff --git a/modules/gpu/misc/mark_nvidia.py b/modules/gpu/misc/mark_nvidia.py
index e8cc3e8417..80dc7f9cee 100755
--- a/modules/gpu/misc/mark_nvidia.py
+++ b/modules/gpu/misc/mark_nvidia.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys, re
 
@@ -257,4 +257,3 @@ if __name__ == "__main__":
     outputFile = open(sys.argv[2], 'w')
     outputFile.writelines(lines)
     outputFile.close()
-
diff --git a/modules/gpu/perf/perf_calib3d.cpp b/modules/gpu/perf/perf_calib3d.cpp
index b174d9a12e..185d9cd684 100644
--- a/modules/gpu/perf/perf_calib3d.cpp
+++ b/modules/gpu/perf/perf_calib3d.cpp
@@ -1,183 +1,53 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
-
-//////////////////////////////////////////////////////////////////////
-// StereoBM
-
-typedef std::tr1::tuple<string, string> pair_string;
-DEF_PARAM_TEST_1(ImagePair, pair_string);
-
-PERF_TEST_P(ImagePair, Calib3D_StereoBM,
-            Values(pair_string("gpu/perf/aloe.png", "gpu/perf/aloeR.png")))
-{
-    declare.time(5.0);
-
-    const cv::Mat imgLeft = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(imgLeft.empty());
-
-    const cv::Mat imgRight = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(imgRight.empty());
-
-    const int preset = 0;
-    const int ndisp = 256;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::StereoBM_GPU d_bm(preset, ndisp);
-
-        const cv::gpu::GpuMat d_imgLeft(imgLeft);
-        const cv::gpu::GpuMat d_imgRight(imgRight);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() d_bm(d_imgLeft, d_imgRight, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::StereoBM bm(preset, ndisp);
-
-        cv::Mat dst;
-
-        TEST_CYCLE() bm(imgLeft, imgRight, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// StereoBeliefPropagation
-
-PERF_TEST_P(ImagePair, Calib3D_StereoBeliefPropagation,
-            Values(pair_string("gpu/stereobp/aloe-L.png", "gpu/stereobp/aloe-R.png")))
-{
-    declare.time(10.0);
-
-    const cv::Mat imgLeft = readImage(GET_PARAM(0));
-    ASSERT_FALSE(imgLeft.empty());
-
-    const cv::Mat imgRight = readImage(GET_PARAM(1));
-    ASSERT_FALSE(imgRight.empty());
-
-    const int ndisp = 64;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::StereoBeliefPropagation d_bp(ndisp);
-
-        const cv::gpu::GpuMat d_imgLeft(imgLeft);
-        const cv::gpu::GpuMat d_imgRight(imgRight);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() d_bp(d_imgLeft, d_imgRight, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// StereoConstantSpaceBP
-
-PERF_TEST_P(ImagePair, Calib3D_StereoConstantSpaceBP,
-            Values(pair_string("gpu/stereobm/aloe-L.png", "gpu/stereobm/aloe-R.png")))
-{
-    declare.time(10.0);
-
-    const cv::Mat imgLeft = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(imgLeft.empty());
-
-    const cv::Mat imgRight = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(imgRight.empty());
-
-    const int ndisp = 128;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::StereoConstantSpaceBP d_csbp(ndisp);
-
-        const cv::gpu::GpuMat d_imgLeft(imgLeft);
-        const cv::gpu::GpuMat d_imgRight(imgRight);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() d_csbp(d_imgLeft, d_imgRight, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// DisparityBilateralFilter
-
-PERF_TEST_P(ImagePair, Calib3D_DisparityBilateralFilter,
-            Values(pair_string("gpu/stereobm/aloe-L.png", "gpu/stereobm/aloe-disp.png")))
-{
-    const cv::Mat img = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    const cv::Mat disp = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(disp.empty());
-
-    const int ndisp = 128;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::DisparityBilateralFilter d_filter(ndisp);
-
-        const cv::gpu::GpuMat d_img(img);
-        const cv::gpu::GpuMat d_disp(disp);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() d_filter(d_disp, d_img, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// TransformPoints
+using namespace perf;
 
 DEF_PARAM_TEST_1(Count, int);
 
-PERF_TEST_P(Count, Calib3D_TransformPoints,
-            Values(5000, 10000, 20000))
-{
-    const int count = GetParam();
-
-    cv::Mat src(1, count, CV_32FC3);
-    declare.in(src, WARMUP_RNG);
-
-    const cv::Mat rvec = cv::Mat::ones(1, 3, CV_32FC1);
-    const cv::Mat tvec = cv::Mat::ones(1, 3, CV_32FC1);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::transformPoints(d_src, rvec, tvec, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // ProjectPoints
 
@@ -263,66 +133,3 @@ PERF_TEST_P(Count, Calib3D_SolvePnPRansac,
         CPU_SANITY_CHECK(tvec, 1e-6);
     }
 }
-
-//////////////////////////////////////////////////////////////////////
-// ReprojectImageTo3D
-
-PERF_TEST_P(Sz_Depth, Calib3D_ReprojectImageTo3D,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16S)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    cv::Mat Q(4, 4, CV_32FC1);
-    cv::randu(Q, 0.1, 1.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::reprojectImageTo3D(d_src, dst, Q);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::reprojectImageTo3D(src, dst, Q);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// DrawColorDisp
-
-PERF_TEST_P(Sz_Depth, Calib3D_DrawColorDisp,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16S)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::drawColorDisp(d_src, dst, 255);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
diff --git a/modules/gpu/perf/perf_imgproc.cpp b/modules/gpu/perf/perf_imgproc.cpp
deleted file mode 100644
index be6eb4877c..0000000000
--- a/modules/gpu/perf/perf_imgproc.cpp
+++ /dev/null
@@ -1,1775 +0,0 @@
-#include "perf_precomp.hpp"
-
-using namespace std;
-using namespace testing;
-using namespace perf;
-
-//////////////////////////////////////////////////////////////////////
-// Remap
-
-enum { HALF_SIZE=0, UPSIDE_DOWN, REFLECTION_X, REFLECTION_BOTH };
-CV_ENUM(RemapMode, HALF_SIZE, UPSIDE_DOWN, REFLECTION_X, REFLECTION_BOTH);
-#define ALL_REMAP_MODES ValuesIn(RemapMode::all())
-
-void generateMap(cv::Mat& map_x, cv::Mat& map_y, int remapMode)
-{
-    for (int j = 0; j < map_x.rows; ++j)
-    {
-        for (int i = 0; i < map_x.cols; ++i)
-        {
-            switch (remapMode)
-            {
-            case HALF_SIZE:
-                if (i > map_x.cols*0.25 && i < map_x.cols*0.75 && j > map_x.rows*0.25 && j < map_x.rows*0.75)
-                {
-                    map_x.at<float>(j,i) = 2.f * (i - map_x.cols * 0.25f) + 0.5f;
-                    map_y.at<float>(j,i) = 2.f * (j - map_x.rows * 0.25f) + 0.5f;
-                }
-                else
-                {
-                    map_x.at<float>(j,i) = 0.f;
-                    map_y.at<float>(j,i) = 0.f;
-                }
-                break;
-            case UPSIDE_DOWN:
-                map_x.at<float>(j,i) = static_cast<float>(i);
-                map_y.at<float>(j,i) = static_cast<float>(map_x.rows - j);
-                break;
-            case REFLECTION_X:
-                map_x.at<float>(j,i) = static_cast<float>(map_x.cols - i);
-                map_y.at<float>(j,i) = static_cast<float>(j);
-                break;
-            case REFLECTION_BOTH:
-                map_x.at<float>(j,i) = static_cast<float>(map_x.cols - i);
-                map_y.at<float>(j,i) = static_cast<float>(map_x.rows - j);
-                break;
-            } // end of switch
-        }
-    }
-}
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border_Mode, cv::Size, MatDepth, MatCn, Interpolation, BorderMode, RemapMode);
-
-PERF_TEST_P(Sz_Depth_Cn_Inter_Border_Mode, ImgProc_Remap,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
-                    ALL_BORDER_MODES,
-                    ALL_REMAP_MODES))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = GET_PARAM(3);
-    const int borderMode = GET_PARAM(4);
-    const int remapMode = GET_PARAM(5);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    cv::Mat xmap(size, CV_32FC1);
-    cv::Mat ymap(size, CV_32FC1);
-    generateMap(xmap, ymap, remapMode);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        const cv::gpu::GpuMat d_xmap(xmap);
-        const cv::gpu::GpuMat d_ymap(ymap);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::remap(d_src, dst, d_xmap, d_ymap, interpolation, borderMode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::remap(src, dst, xmap, ymap, interpolation, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Resize
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Scale, cv::Size, MatDepth, MatCn, Interpolation, double);
-
-PERF_TEST_P(Sz_Depth_Cn_Inter_Scale, ImgProc_Resize,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
-                    Values(0.5, 0.3, 2.0)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = GET_PARAM(3);
-    const double f = GET_PARAM(4);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::resize(d_src, dst, cv::Size(), f, f, interpolation);
-
-        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::resize(src, dst, cv::Size(), f, f, interpolation);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// ResizeArea
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Scale, cv::Size, MatDepth, MatCn, double);
-
-PERF_TEST_P(Sz_Depth_Cn_Scale, ImgProc_ResizeArea,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(0.2, 0.1, 0.05)))
-{
-    declare.time(1.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = cv::INTER_AREA;
-    const double f = GET_PARAM(3);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::resize(d_src, dst, cv::Size(), f, f, interpolation);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::resize(src, dst, cv::Size(), f, f, interpolation);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// WarpAffine
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border, cv::Size, MatDepth, MatCn, Interpolation, BorderMode);
-
-PERF_TEST_P(Sz_Depth_Cn_Inter_Border, ImgProc_WarpAffine,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
-                    ALL_BORDER_MODES))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = GET_PARAM(3);
-    const int borderMode = GET_PARAM(4);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const double aplha = CV_PI / 4;
-    const double mat[2 * 3] =
-    {
-        std::cos(aplha), -std::sin(aplha), src.cols / 2,
-        std::sin(aplha),  std::cos(aplha), 0
-    };
-    const cv::Mat M(2, 3, CV_64F, (void*) mat);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::warpAffine(d_src, dst, M, size, interpolation, borderMode);
-
-        GPU_SANITY_CHECK(dst, 1);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::warpAffine(src, dst, M, size, interpolation, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// WarpPerspective
-
-PERF_TEST_P(Sz_Depth_Cn_Inter_Border, ImgProc_WarpPerspective,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
-                    ALL_BORDER_MODES))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = GET_PARAM(3);
-    const int borderMode = GET_PARAM(4);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const double aplha = CV_PI / 4;
-    double mat[3][3] = { {std::cos(aplha), -std::sin(aplha), src.cols / 2},
-                         {std::sin(aplha),  std::cos(aplha), 0},
-                         {0.0,              0.0,             1.0}};
-    const cv::Mat M(3, 3, CV_64F, (void*) mat);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::warpPerspective(d_src, dst, M, size, interpolation, borderMode);
-
-        GPU_SANITY_CHECK(dst, 1);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::warpPerspective(src, dst, M, size, interpolation, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CopyMakeBorder
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Border, cv::Size, MatDepth, MatCn, BorderMode);
-
-PERF_TEST_P(Sz_Depth_Cn_Border, ImgProc_CopyMakeBorder,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    ALL_BORDER_MODES))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int borderMode = GET_PARAM(3);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::copyMakeBorder(d_src, dst, 5, 5, 5, 5, borderMode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::copyMakeBorder(src, dst, 5, 5, 5, 5, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Threshold
-
-CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
-#define ALL_THRESH_OPS ValuesIn(ThreshOp::all())
-
-DEF_PARAM_TEST(Sz_Depth_Op, cv::Size, MatDepth, ThreshOp);
-
-PERF_TEST_P(Sz_Depth_Op, ImgProc_Threshold,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-            Values(CV_8U, CV_16U, CV_32F, CV_64F),
-            ALL_THRESH_OPS))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int threshOp = GET_PARAM(2);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::threshold(d_src, dst, 100.0, 255.0, threshOp);
-
-        GPU_SANITY_CHECK(dst, 1e-10);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::threshold(src, dst, 100.0, 255.0, threshOp);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Integral
-
-PERF_TEST_P(Sz, ImgProc_Integral,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_8UC1);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::integralBuffered(d_src, dst, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::integral(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// IntegralSqr
-
-PERF_TEST_P(Sz, ImgProc_IntegralSqr,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_8UC1);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::sqrIntegral(d_src, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// HistEvenC1
-
-PERF_TEST_P(Sz_Depth, ImgProc_HistEvenC1,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_16S)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::histEven(d_src, dst, d_buf, 30, 0, 180);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        const int hbins = 30;
-        const float hranges[] = {0.0f, 180.0f};
-        const int histSize[] = {hbins};
-        const float* ranges[] = {hranges};
-        const int channels[] = {0};
-
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::calcHist(&src, 1, channels, cv::Mat(), dst, 1, histSize, ranges);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// HistEvenC4
-
-PERF_TEST_P(Sz_Depth, ImgProc_HistEvenC4,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_16S)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, CV_MAKE_TYPE(depth, 4));
-    declare.in(src, WARMUP_RNG);
-
-    int histSize[] = {30, 30, 30, 30};
-    int lowerLevel[] = {0, 0, 0, 0};
-    int upperLevel[] = {180, 180, 180, 180};
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_hist[4];
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::histEven(d_src, d_hist, d_buf, histSize, lowerLevel, upperLevel);
-
-        cv::Mat cpu_hist0, cpu_hist1, cpu_hist2, cpu_hist3;
-        d_hist[0].download(cpu_hist0);
-        d_hist[1].download(cpu_hist1);
-        d_hist[2].download(cpu_hist2);
-        d_hist[3].download(cpu_hist3);
-        SANITY_CHECK(cpu_hist0);
-        SANITY_CHECK(cpu_hist1);
-        SANITY_CHECK(cpu_hist2);
-        SANITY_CHECK(cpu_hist3);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CalcHist
-
-PERF_TEST_P(Sz, ImgProc_CalcHist,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_8UC1);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::calcHist(d_src, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// EqualizeHist
-
-PERF_TEST_P(Sz, ImgProc_EqualizeHist,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_8UC1);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_hist;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::equalizeHist(d_src, dst, d_hist, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::equalizeHist(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// ColumnSum
-
-PERF_TEST_P(Sz, ImgProc_ColumnSum,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_32FC1);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::columnSum(d_src, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Canny
-
-DEF_PARAM_TEST(Image_AppertureSz_L2gradient, string, int, bool);
-
-PERF_TEST_P(Image_AppertureSz_L2gradient, ImgProc_Canny,
-            Combine(Values("perf/800x600.png", "perf/1280x1024.png", "perf/1680x1050.png"),
-                    Values(3, 5),
-                    Bool()))
-{
-    const string fileName = GET_PARAM(0);
-    const int apperture_size = GET_PARAM(1);
-    const bool useL2gradient = GET_PARAM(2);
-
-    const cv::Mat image = readImage(fileName, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    const double low_thresh = 50.0;
-    const double high_thresh = 100.0;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_image(image);
-        cv::gpu::GpuMat dst;
-        cv::gpu::CannyBuf d_buf;
-
-        TEST_CYCLE() cv::gpu::Canny(d_image, d_buf, dst, low_thresh, high_thresh, apperture_size, useL2gradient);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::Canny(image, dst, low_thresh, high_thresh, apperture_size, useL2gradient);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MeanShiftFiltering
-
-DEF_PARAM_TEST_1(Image, string);
-
-PERF_TEST_P(Image, ImgProc_MeanShiftFiltering,
-            Values<string>("gpu/meanshift/cones.png"))
-{
-    declare.time(15.0);
-
-    const cv::Mat img = readImage(GetParam());
-    ASSERT_FALSE(img.empty());
-
-    cv::Mat rgba;
-    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
-
-    const int sp = 50;
-    const int sr = 50;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(rgba);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::meanShiftFiltering(d_src, dst, sp, sr);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::pyrMeanShiftFiltering(img, dst, sp, sr);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MeanShiftProc
-
-PERF_TEST_P(Image, ImgProc_MeanShiftProc,
-            Values<string>("gpu/meanshift/cones.png"))
-{
-    declare.time(5.0);
-
-    const cv::Mat img = readImage(GetParam());
-    ASSERT_FALSE(img.empty());
-
-    cv::Mat rgba;
-    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
-
-    const int sp = 50;
-    const int sr = 50;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(rgba);
-        cv::gpu::GpuMat dstr;
-        cv::gpu::GpuMat dstsp;
-
-        TEST_CYCLE() cv::gpu::meanShiftProc(d_src, dstr, dstsp, sp, sr);
-
-        GPU_SANITY_CHECK(dstr);
-        GPU_SANITY_CHECK(dstsp);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MeanShiftSegmentation
-
-PERF_TEST_P(Image, ImgProc_MeanShiftSegmentation,
-            Values<string>("gpu/meanshift/cones.png"))
-{
-    declare.time(5.0);
-
-    const cv::Mat img = readImage(GetParam());
-    ASSERT_FALSE(img.empty());
-
-    cv::Mat rgba;
-    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
-
-    const int sp = 10;
-    const int sr = 10;
-    const int minsize = 20;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(rgba);
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::gpu::meanShiftSegmentation(d_src, dst, sp, sr, minsize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// BlendLinear
-
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_BlendLinear,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat img1(size, type);
-    cv::Mat img2(size, type);
-    declare.in(img1, img2, WARMUP_RNG);
-
-    const cv::Mat weights1(size, CV_32FC1, cv::Scalar::all(0.5));
-    const cv::Mat weights2(size, CV_32FC1, cv::Scalar::all(0.5));
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_img1(img1);
-        const cv::gpu::GpuMat d_img2(img2);
-        const cv::gpu::GpuMat d_weights1(weights1);
-        const cv::gpu::GpuMat d_weights2(weights2);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::blendLinear(d_img1, d_img2, d_weights1, d_weights2, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Convolve
-
-DEF_PARAM_TEST(Sz_KernelSz_Ccorr, cv::Size, int, bool);
-
-PERF_TEST_P(Sz_KernelSz_Ccorr, ImgProc_Convolve,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(17, 27, 32, 64),
-                    Bool()))
-{
-    declare.time(10.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int templ_size = GET_PARAM(1);
-    const bool ccorr = GET_PARAM(2);
-
-    const cv::Mat image(size, CV_32FC1);
-    const cv::Mat templ(templ_size, templ_size, CV_32FC1);
-    declare.in(image, templ, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_image = cv::gpu::createContinuous(size, CV_32FC1);
-        d_image.upload(image);
-
-        cv::gpu::GpuMat d_templ = cv::gpu::createContinuous(templ_size, templ_size, CV_32FC1);
-        d_templ.upload(templ);
-
-        cv::gpu::GpuMat dst;
-        cv::gpu::ConvolveBuf d_buf;
-
-        TEST_CYCLE() cv::gpu::convolve(d_image, d_templ, dst, ccorr, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        if (ccorr)
-            FAIL_NO_CPU();
-
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::filter2D(image, dst, image.depth(), templ);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate8U
-
-CV_ENUM(TemplateMethod, cv::TM_SQDIFF, cv::TM_SQDIFF_NORMED, cv::TM_CCORR, cv::TM_CCORR_NORMED, cv::TM_CCOEFF, cv::TM_CCOEFF_NORMED)
-#define ALL_TEMPLATE_METHODS ValuesIn(TemplateMethod::all())
-
-DEF_PARAM_TEST(Sz_TemplateSz_Cn_Method, cv::Size, cv::Size, MatCn, TemplateMethod);
-
-PERF_TEST_P(Sz_TemplateSz_Cn_Method, ImgProc_MatchTemplate8U,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
-                    GPU_CHANNELS_1_3_4,
-                    ALL_TEMPLATE_METHODS))
-{
-    const cv::Size size = GET_PARAM(0);
-    const cv::Size templ_size = GET_PARAM(1);
-    const int cn = GET_PARAM(2);
-    const int method = GET_PARAM(3);
-
-    cv::Mat image(size, CV_MAKE_TYPE(CV_8U, cn));
-    cv::Mat templ(templ_size, CV_MAKE_TYPE(CV_8U, cn));
-    declare.in(image, templ, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_image(image);
-        const cv::gpu::GpuMat d_templ(templ);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::matchTemplate(d_image, d_templ, dst, method);
-
-        GPU_SANITY_CHECK(dst, 1e-5, ERROR_RELATIVE);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::matchTemplate(image, templ, dst, method);
-
-        CPU_SANITY_CHECK(dst);
-    }
-};
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate32F
-
-PERF_TEST_P(Sz_TemplateSz_Cn_Method, ImgProc_MatchTemplate32F,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
-                    GPU_CHANNELS_1_3_4,
-                    Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const cv::Size templ_size = GET_PARAM(1);
-    const int cn = GET_PARAM(2);
-    int method = GET_PARAM(3);
-
-    cv::Mat image(size, CV_MAKE_TYPE(CV_32F, cn));
-    cv::Mat templ(templ_size, CV_MAKE_TYPE(CV_32F, cn));
-    declare.in(image, templ, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_image(image);
-        const cv::gpu::GpuMat d_templ(templ);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::matchTemplate(d_image, d_templ, dst, method);
-
-        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::matchTemplate(image, templ, dst, method);
-
-        CPU_SANITY_CHECK(dst);
-    }
-};
-
-//////////////////////////////////////////////////////////////////////
-// MulSpectrums
-
-CV_FLAGS(DftFlags, 0, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
-
-DEF_PARAM_TEST(Sz_Flags, cv::Size, DftFlags);
-
-PERF_TEST_P(Sz_Flags, ImgProc_MulSpectrums,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(0, DftFlags(cv::DFT_ROWS))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int flag = GET_PARAM(1);
-
-    cv::Mat a(size, CV_32FC2);
-    cv::Mat b(size, CV_32FC2);
-    declare.in(a, b, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_a(a);
-        const cv::gpu::GpuMat d_b(b);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::mulSpectrums(d_a, d_b, dst, flag);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::mulSpectrums(a, b, dst, flag);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MulAndScaleSpectrums
-
-PERF_TEST_P(Sz, ImgProc_MulAndScaleSpectrums,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    const float scale = 1.f / size.area();
-
-    cv::Mat src1(size, CV_32FC2);
-    cv::Mat src2(size, CV_32FC2);
-    declare.in(src1,src2, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src1(src1);
-        const cv::gpu::GpuMat d_src2(src2);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::mulAndScaleSpectrums(d_src1, d_src2, dst, cv::DFT_ROWS, scale, false);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Dft
-
-PERF_TEST_P(Sz_Flags, ImgProc_Dft,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(0, DftFlags(cv::DFT_ROWS), DftFlags(cv::DFT_INVERSE))))
-{
-    declare.time(10.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int flag = GET_PARAM(1);
-
-    cv::Mat src(size, CV_32FC2);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::dft(d_src, dst, size, flag);
-
-        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::dft(src, dst, flag);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CornerHarris
-
-DEF_PARAM_TEST(Image_Type_Border_BlockSz_ApertureSz, string, MatType, BorderMode, int, int);
-
-PERF_TEST_P(Image_Type_Border_BlockSz_ApertureSz, ImgProc_CornerHarris,
-            Combine(Values<string>("gpu/stereobm/aloe-L.png"),
-                    Values(CV_8UC1, CV_32FC1),
-                    Values(BorderMode(cv::BORDER_REFLECT101), BorderMode(cv::BORDER_REPLICATE), BorderMode(cv::BORDER_REFLECT)),
-                    Values(3, 5, 7),
-                    Values(0, 3, 5, 7)))
-{
-    const string fileName = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int borderMode = GET_PARAM(2);
-    const int blockSize = GET_PARAM(3);
-    const int apertureSize = GET_PARAM(4);
-
-    cv::Mat img = readImage(fileName, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    img.convertTo(img, type, type == CV_32F ? 1.0 / 255.0 : 1.0);
-
-    const double k = 0.5;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_img(img);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_Dx;
-        cv::gpu::GpuMat d_Dy;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::cornerHarris(d_img, dst, d_Dx, d_Dy, d_buf, blockSize, apertureSize, k, borderMode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::cornerHarris(img, dst, blockSize, apertureSize, k, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CornerMinEigenVal
-
-PERF_TEST_P(Image_Type_Border_BlockSz_ApertureSz, ImgProc_CornerMinEigenVal,
-            Combine(Values<string>("gpu/stereobm/aloe-L.png"),
-                    Values(CV_8UC1, CV_32FC1),
-                    Values(BorderMode(cv::BORDER_REFLECT101), BorderMode(cv::BORDER_REPLICATE), BorderMode(cv::BORDER_REFLECT)),
-                    Values(3, 5, 7),
-                    Values(0, 3, 5, 7)))
-{
-    const string fileName = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int borderMode = GET_PARAM(2);
-    const int blockSize = GET_PARAM(3);
-    const int apertureSize = GET_PARAM(4);
-
-    cv::Mat img = readImage(fileName, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    img.convertTo(img, type, type == CV_32F ? 1.0 / 255.0 : 1.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_img(img);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_Dx;
-        cv::gpu::GpuMat d_Dy;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::cornerMinEigenVal(d_img, dst, d_Dx, d_Dy, d_buf, blockSize, apertureSize, borderMode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::cornerMinEigenVal(img, dst, blockSize, apertureSize, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// BuildWarpPlaneMaps
-
-PERF_TEST_P(Sz, ImgProc_BuildWarpPlaneMaps,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
-    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
-    const cv::Mat T = cv::Mat::zeros(1, 3, CV_32F);
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat map_x;
-        cv::gpu::GpuMat map_y;
-
-        TEST_CYCLE() cv::gpu::buildWarpPlaneMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, T, 1.0, map_x, map_y);
-
-        GPU_SANITY_CHECK(map_x);
-        GPU_SANITY_CHECK(map_y);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// BuildWarpCylindricalMaps
-
-PERF_TEST_P(Sz, ImgProc_BuildWarpCylindricalMaps,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
-    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat map_x;
-        cv::gpu::GpuMat map_y;
-
-        TEST_CYCLE() cv::gpu::buildWarpCylindricalMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, 1.0, map_x, map_y);
-
-        GPU_SANITY_CHECK(map_x);
-        GPU_SANITY_CHECK(map_y);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// BuildWarpSphericalMaps
-
-PERF_TEST_P(Sz, ImgProc_BuildWarpSphericalMaps,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
-    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat map_x;
-        cv::gpu::GpuMat map_y;
-
-        TEST_CYCLE() cv::gpu::buildWarpSphericalMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, 1.0, map_x, map_y);
-
-        GPU_SANITY_CHECK(map_x);
-        GPU_SANITY_CHECK(map_y);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Rotate
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Inter, cv::Size, MatDepth, MatCn, Interpolation);
-
-PERF_TEST_P(Sz_Depth_Cn_Inter, ImgProc_Rotate,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int interpolation = GET_PARAM(3);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::rotate(d_src, dst, size, 30.0, 0, 0, interpolation);
-
-        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// PyrDown
-
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_PyrDown,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::pyrDown(d_src, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::pyrDown(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// PyrUp
-
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_PyrUp,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::pyrUp(d_src, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::pyrUp(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CvtColor
-
-DEF_PARAM_TEST(Sz_Depth_Code, cv::Size, MatDepth, CvtColorInfo);
-
-PERF_TEST_P(Sz_Depth_Code, ImgProc_CvtColor,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_32F),
-                    Values(CvtColorInfo(4, 4, cv::COLOR_RGBA2BGRA),
-                           CvtColorInfo(4, 1, cv::COLOR_BGRA2GRAY),
-                           CvtColorInfo(1, 4, cv::COLOR_GRAY2BGRA),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2XYZ),
-                           CvtColorInfo(3, 3, cv::COLOR_XYZ2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2YCrCb),
-                           CvtColorInfo(3, 3, cv::COLOR_YCrCb2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2YUV),
-                           CvtColorInfo(3, 3, cv::COLOR_YUV2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2HSV),
-                           CvtColorInfo(3, 3, cv::COLOR_HSV2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2HLS),
-                           CvtColorInfo(3, 3, cv::COLOR_HLS2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2Lab),
-                           CvtColorInfo(3, 3, cv::COLOR_LBGR2Lab),
-                           CvtColorInfo(3, 3, cv::COLOR_BGR2Luv),
-                           CvtColorInfo(3, 3, cv::COLOR_LBGR2Luv),
-                           CvtColorInfo(3, 3, cv::COLOR_Lab2BGR),
-                           CvtColorInfo(3, 3, cv::COLOR_Lab2LBGR),
-                           CvtColorInfo(3, 3, cv::COLOR_Luv2RGB),
-                           CvtColorInfo(3, 3, cv::COLOR_Luv2LRGB))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const CvtColorInfo info = GET_PARAM(2);
-
-    cv::Mat src(size, CV_MAKETYPE(depth, info.scn));
-    cv::randu(src, 0, depth == CV_8U ? 255.0 : 1.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::cvtColor(d_src, dst, info.code, info.dcn);
-
-        GPU_SANITY_CHECK(dst, 1e-4);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::cvtColor(src, dst, info.code, info.dcn);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-PERF_TEST_P(Sz_Depth_Code, ImgProc_CvtColorBayer,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U),
-                    Values(CvtColorInfo(1, 3, cv::COLOR_BayerBG2BGR),
-                           CvtColorInfo(1, 3, cv::COLOR_BayerGB2BGR),
-                           CvtColorInfo(1, 3, cv::COLOR_BayerRG2BGR),
-                           CvtColorInfo(1, 3, cv::COLOR_BayerGR2BGR))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const CvtColorInfo info = GET_PARAM(2);
-
-    cv::Mat src(size, CV_MAKETYPE(depth, info.scn));
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::cvtColor(d_src, dst, info.code, info.dcn);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::cvtColor(src, dst, info.code, info.dcn);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// SwapChannels
-
-PERF_TEST_P(Sz, ImgProc_SwapChannels,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
-
-    cv::Mat src(size, CV_8UC4);
-    declare.in(src, WARMUP_RNG);
-
-    const int dstOrder[] = {2, 1, 0, 3};
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat dst(src);
-
-        TEST_CYCLE() cv::gpu::swapChannels(dst, dstOrder);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// AlphaComp
-
-CV_ENUM(AlphaOp, cv::gpu::ALPHA_OVER, cv::gpu::ALPHA_IN, cv::gpu::ALPHA_OUT, cv::gpu::ALPHA_ATOP, cv::gpu::ALPHA_XOR, cv::gpu::ALPHA_PLUS, cv::gpu::ALPHA_OVER_PREMUL, cv::gpu::ALPHA_IN_PREMUL, cv::gpu::ALPHA_OUT_PREMUL, cv::gpu::ALPHA_ATOP_PREMUL, cv::gpu::ALPHA_XOR_PREMUL, cv::gpu::ALPHA_PLUS_PREMUL, cv::gpu::ALPHA_PREMUL)
-#define ALL_ALPHA_OPS ValuesIn(AlphaOp::all())
-
-DEF_PARAM_TEST(Sz_Type_Op, cv::Size, MatType, AlphaOp);
-
-PERF_TEST_P(Sz_Type_Op, ImgProc_AlphaComp,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8UC4, CV_16UC4, CV_32SC4, CV_32FC4),
-                    ALL_ALPHA_OPS))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int alpha_op = GET_PARAM(2);
-
-    cv::Mat img1(size, type);
-    cv::Mat img2(size, type);
-    declare.in(img1, img2, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_img1(img1);
-        const cv::gpu::GpuMat d_img2(img2);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::alphaComp(d_img1, d_img2, dst, alpha_op);
-
-        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// ImagePyramidBuild
-
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidBuild,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const int nLayers = 5;
-    const cv::Size dstSize(size.width / 2 + 10, size.height / 2 + 10);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-
-        cv::gpu::ImagePyramid d_pyr;
-
-        TEST_CYCLE() d_pyr.build(d_src, nLayers);
-
-        cv::gpu::GpuMat dst;
-        d_pyr.getLayer(dst, dstSize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// ImagePyramidGetLayer
-
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidGetLayer,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const int nLayers = 3;
-    const cv::Size dstSize(size.width / 2 + 10, size.height / 2 + 10);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        cv::gpu::ImagePyramid d_pyr(d_src, nLayers);
-
-        TEST_CYCLE() d_pyr.getLayer(dst, dstSize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// HoughLines
-
-namespace
-{
-    struct Vec4iComparator
-    {
-        bool operator()(const cv::Vec4i& a, const cv::Vec4i b) const
-        {
-            if (a[0] != b[0]) return a[0] < b[0];
-            else if(a[1] != b[1]) return a[1] < b[1];
-            else if(a[2] != b[2]) return a[2] < b[2];
-            else return a[3] < b[3];
-        }
-    };
-    struct Vec3fComparator
-    {
-        bool operator()(const cv::Vec3f& a, const cv::Vec3f b) const
-        {
-            if(a[0] != b[0]) return a[0] < b[0];
-            else if(a[1] != b[1]) return a[1] < b[1];
-            else return a[2] < b[2];
-        }
-    };
-    struct Vec2fComparator
-    {
-        bool operator()(const cv::Vec2f& a, const cv::Vec2f b) const
-        {
-            if(a[0] != b[0]) return a[0] < b[0];
-            else return a[1] < b[1];
-        }
-    };
-}
-
-PERF_TEST_P(Sz, ImgProc_HoughLines,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    declare.time(30.0);
-
-    const cv::Size size = GetParam();
-
-    const float rho = 1.0f;
-    const float theta = static_cast<float>(CV_PI / 180.0);
-    const int threshold = 300;
-
-    cv::Mat src(size, CV_8UC1, cv::Scalar::all(0));
-    cv::line(src, cv::Point(0, 100), cv::Point(src.cols, 100), cv::Scalar::all(255), 1);
-    cv::line(src, cv::Point(0, 200), cv::Point(src.cols, 200), cv::Scalar::all(255), 1);
-    cv::line(src, cv::Point(0, 400), cv::Point(src.cols, 400), cv::Scalar::all(255), 1);
-    cv::line(src, cv::Point(100, 0), cv::Point(100, src.rows), cv::Scalar::all(255), 1);
-    cv::line(src, cv::Point(200, 0), cv::Point(200, src.rows), cv::Scalar::all(255), 1);
-    cv::line(src, cv::Point(400, 0), cv::Point(400, src.rows), cv::Scalar::all(255), 1);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_lines;
-        cv::gpu::HoughLinesBuf d_buf;
-
-        TEST_CYCLE() cv::gpu::HoughLines(d_src, d_lines, d_buf, rho, theta, threshold);
-
-        cv::Mat gpu_lines(d_lines.row(0));
-        cv::Vec2f* begin = gpu_lines.ptr<cv::Vec2f>(0);
-        cv::Vec2f* end = begin + gpu_lines.cols;
-        std::sort(begin, end, Vec2fComparator());
-        SANITY_CHECK(gpu_lines);
-    }
-    else
-    {
-        std::vector<cv::Vec2f> cpu_lines;
-
-        TEST_CYCLE() cv::HoughLines(src, cpu_lines, rho, theta, threshold);
-
-        SANITY_CHECK(cpu_lines);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// HoughLinesP
-
-DEF_PARAM_TEST_1(Image, std::string);
-
-PERF_TEST_P(Image, ImgProc_HoughLinesP,
-            testing::Values("cv/shared/pic5.png", "stitching/a1.png"))
-{
-    declare.time(30.0);
-
-    const std::string fileName = getDataPath(GetParam());
-
-    const float rho = 1.0f;
-    const float theta = static_cast<float>(CV_PI / 180.0);
-    const int threshold = 100;
-    const int minLineLenght = 50;
-    const int maxLineGap = 5;
-
-    const cv::Mat image = cv::imread(fileName, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    cv::Mat mask;
-    cv::Canny(image, mask, 50, 100);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_mask(mask);
-        cv::gpu::GpuMat d_lines;
-        cv::gpu::HoughLinesBuf d_buf;
-
-        TEST_CYCLE() cv::gpu::HoughLinesP(d_mask, d_lines, d_buf, rho, theta, minLineLenght, maxLineGap);
-
-        cv::Mat gpu_lines(d_lines);
-        cv::Vec4i* begin = gpu_lines.ptr<cv::Vec4i>();
-        cv::Vec4i* end = begin + gpu_lines.cols;
-        std::sort(begin, end, Vec4iComparator());
-        SANITY_CHECK(gpu_lines);
-    }
-    else
-    {
-        std::vector<cv::Vec4i> cpu_lines;
-
-        TEST_CYCLE() cv::HoughLinesP(mask, cpu_lines, rho, theta, threshold, minLineLenght, maxLineGap);
-
-        SANITY_CHECK(cpu_lines);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// HoughCircles
-
-DEF_PARAM_TEST(Sz_Dp_MinDist, cv::Size, float, float);
-
-PERF_TEST_P(Sz_Dp_MinDist, ImgProc_HoughCircles,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(1.0f, 2.0f, 4.0f),
-                    Values(1.0f)))
-{
-    declare.time(30.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const float dp = GET_PARAM(1);
-    const float minDist = GET_PARAM(2);
-
-    const int minRadius = 10;
-    const int maxRadius = 30;
-    const int cannyThreshold = 100;
-    const int votesThreshold = 15;
-
-    cv::Mat src(size, CV_8UC1, cv::Scalar::all(0));
-    cv::circle(src, cv::Point(100, 100), 20, cv::Scalar::all(255), -1);
-    cv::circle(src, cv::Point(200, 200), 25, cv::Scalar::all(255), -1);
-    cv::circle(src, cv::Point(200, 100), 25, cv::Scalar::all(255), -1);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_circles;
-        cv::gpu::HoughCirclesBuf d_buf;
-
-        TEST_CYCLE() cv::gpu::HoughCircles(d_src, d_circles, d_buf, CV_HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
-
-        cv::Mat gpu_circles(d_circles);
-        cv::Vec3f* begin = gpu_circles.ptr<cv::Vec3f>(0);
-        cv::Vec3f* end = begin + gpu_circles.cols;
-        std::sort(begin, end, Vec3fComparator());
-        SANITY_CHECK(gpu_circles);
-    }
-    else
-    {
-        std::vector<cv::Vec3f> cpu_circles;
-
-        TEST_CYCLE() cv::HoughCircles(src, cpu_circles, CV_HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
-
-        SANITY_CHECK(cpu_circles);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// GeneralizedHough
-
-CV_FLAGS(GHMethod, cv::GHT_POSITION, cv::GHT_SCALE, cv::GHT_ROTATION);
-
-DEF_PARAM_TEST(Method_Sz, GHMethod, cv::Size);
-
-PERF_TEST_P(Method_Sz, ImgProc_GeneralizedHough,
-            Combine(Values(GHMethod(cv::GHT_POSITION), GHMethod(cv::GHT_POSITION | cv::GHT_SCALE), GHMethod(cv::GHT_POSITION | cv::GHT_ROTATION), GHMethod(cv::GHT_POSITION | cv::GHT_SCALE | cv::GHT_ROTATION)),
-                    GPU_TYPICAL_MAT_SIZES))
-{
-    declare.time(10);
-
-    const int method = GET_PARAM(0);
-    const cv::Size imageSize = GET_PARAM(1);
-
-    const cv::Mat templ = readImage("cv/shared/templ.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(templ.empty());
-
-    cv::Mat image(imageSize, CV_8UC1, cv::Scalar::all(0));
-    templ.copyTo(image(cv::Rect(50, 50, templ.cols, templ.rows)));
-
-    cv::RNG rng(123456789);
-    const int objCount = rng.uniform(5, 15);
-    for (int i = 0; i < objCount; ++i)
-    {
-        double scale = rng.uniform(0.7, 1.3);
-        bool rotate = 1 == rng.uniform(0, 2);
-
-        cv::Mat obj;
-        cv::resize(templ, obj, cv::Size(), scale, scale);
-        if (rotate)
-            obj = obj.t();
-
-        cv::Point pos;
-
-        pos.x = rng.uniform(0, image.cols - obj.cols);
-        pos.y = rng.uniform(0, image.rows - obj.rows);
-
-        cv::Mat roi = image(cv::Rect(pos, obj.size()));
-        cv::add(roi, obj, roi);
-    }
-
-    cv::Mat edges;
-    cv::Canny(image, edges, 50, 100);
-
-    cv::Mat dx, dy;
-    cv::Sobel(image, dx, CV_32F, 1, 0);
-    cv::Sobel(image, dy, CV_32F, 0, 1);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_edges(edges);
-        const cv::gpu::GpuMat d_dx(dx);
-        const cv::gpu::GpuMat d_dy(dy);
-        cv::gpu::GpuMat posAndVotes;
-
-        cv::Ptr<cv::gpu::GeneralizedHough_GPU> d_hough = cv::gpu::GeneralizedHough_GPU::create(method);
-        if (method & cv::GHT_ROTATION)
-        {
-            d_hough->set("maxAngle", 90.0);
-            d_hough->set("angleStep", 2.0);
-        }
-
-        d_hough->setTemplate(cv::gpu::GpuMat(templ));
-
-        TEST_CYCLE() d_hough->detect(d_edges, d_dx, d_dy, posAndVotes);
-
-        const cv::gpu::GpuMat positions(1, posAndVotes.cols, CV_32FC4, posAndVotes.data);
-        GPU_SANITY_CHECK(positions);
-    }
-    else
-    {
-        cv::Mat positions;
-
-        cv::Ptr<cv::GeneralizedHough> hough = cv::GeneralizedHough::create(method);
-        if (method & cv::GHT_ROTATION)
-        {
-            hough->set("maxAngle", 90.0);
-            hough->set("angleStep", 2.0);
-        }
-
-        hough->setTemplate(templ);
-
-        TEST_CYCLE() hough->detect(edges, dx, dy, positions);
-
-        CPU_SANITY_CHECK(positions);
-    }
-}
diff --git a/modules/gpu/perf/perf_labeling.cpp b/modules/gpu/perf/perf_labeling.cpp
index f3ad12c949..0484da9d59 100644
--- a/modules/gpu/perf/perf_labeling.cpp
+++ b/modules/gpu/perf/perf_labeling.cpp
@@ -1,7 +1,50 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
+using namespace perf;
 
 DEF_PARAM_TEST_1(Image, string);
 
diff --git a/modules/gpu/perf/perf_main.cpp b/modules/gpu/perf/perf_main.cpp
index 312b744482..802d559afb 100644
--- a/modules/gpu/perf/perf_main.cpp
+++ b/modules/gpu/perf/perf_main.cpp
@@ -1,70 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
-static void printOsInfo()
-{
-#if defined _WIN32
-#   if defined _WIN64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x64.\n[----------]\n"), fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x32.\n[----------]\n"), fflush(stdout);
-#   endif
-#elif defined linux
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x64.\n[----------]\n"), fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x32.\n[----------]\n"), fflush(stdout);
-#   endif
-#elif defined __APPLE__
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x64.\n[----------]\n"), fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x32.\n[----------]\n"), fflush(stdout);
-#   endif
-#endif
+using namespace perf;
 
-}
-
-static void printCudaInfo()
-{
-    printOsInfo();
-#ifndef HAVE_CUDA
-    printf("[----------]\n[ GPU INFO ] \tOpenCV was built without CUDA support.\n[----------]\n"), fflush(stdout);
-#else
-    int driver;
-    cudaDriverGetVersion(&driver);
-
-    printf("[----------]\n"), fflush(stdout);
-    printf("[ GPU INFO ] \tCUDA Driver  version: %d.\n", driver), fflush(stdout);
-    printf("[ GPU INFO ] \tCUDA Runtime version: %d.\n", CUDART_VERSION), fflush(stdout);
-    printf("[----------]\n"), fflush(stdout);
-
-    printf("[----------]\n"), fflush(stdout);
-    printf("[ GPU INFO ] \tGPU module was compiled for the following GPU archs.\n"), fflush(stdout);
-    printf("[      BIN ] \t%s.\n", CUDA_ARCH_BIN), fflush(stdout);
-    printf("[      PTX ] \t%s.\n", CUDA_ARCH_PTX), fflush(stdout);
-    printf("[----------]\n"), fflush(stdout);
-
-    printf("[----------]\n"), fflush(stdout);
-    int deviceCount = cv::gpu::getCudaEnabledDeviceCount();
-    printf("[ GPU INFO ] \tCUDA device count:: %d.\n", deviceCount), fflush(stdout);
-    printf("[----------]\n"), fflush(stdout);
-
-    for (int i = 0; i < deviceCount; ++i)
-    {
-        cv::gpu::DeviceInfo info(i);
-
-        printf("[----------]\n"), fflush(stdout);
-        printf("[ DEVICE   ] \t# %d %s.\n", i, info.name().c_str()), fflush(stdout);
-        printf("[          ] \tCompute capability: %d.%d\n", (int)info.majorVersion(), (int)info.minorVersion()), fflush(stdout);
-        printf("[          ] \tMulti Processor Count:  %d\n", info.multiProcessorCount()), fflush(stdout);
-        printf("[          ] \tTotal memory: %d Mb\n", static_cast<int>(static_cast<int>(info.totalMemory() / 1024.0) / 1024.0)), fflush(stdout);
-        printf("[          ] \tFree  memory: %d Mb\n", static_cast<int>(static_cast<int>(info.freeMemory()  / 1024.0) / 1024.0)), fflush(stdout);
-        if (!info.isCompatible())
-            printf("[ GPU INFO ] \tThis device is NOT compatible with current GPU module build\n");
-        printf("[----------]\n"), fflush(stdout);
-    }
-
-#endif
-}
-
-CV_PERF_TEST_MAIN(gpu, printCudaInfo())
+CV_PERF_TEST_CUDA_MAIN(gpu)
diff --git a/modules/gpu/perf/perf_matop.cpp b/modules/gpu/perf/perf_matop.cpp
index 1696e3a7eb..f80ba1b087 100644
--- a/modules/gpu/perf/perf_matop.cpp
+++ b/modules/gpu/perf/perf_matop.cpp
@@ -1,7 +1,50 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
+using namespace perf;
 
 //////////////////////////////////////////////////////////////////////
 // SetTo
diff --git a/modules/gpu/perf/perf_objdetect.cpp b/modules/gpu/perf/perf_objdetect.cpp
index 969ac10762..1516d6b99b 100644
--- a/modules/gpu/perf/perf_objdetect.cpp
+++ b/modules/gpu/perf/perf_objdetect.cpp
@@ -1,7 +1,50 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
+using namespace perf;
 
 ///////////////////////////////////////////////////////////////
 // HOG
@@ -18,6 +61,8 @@ PERF_TEST_P(Image, ObjDetect_HOG,
                            "gpu/caltech/image_00000527_0.png",
                            "gpu/caltech/image_00000574_0.png"))
 {
+    declare.time(300.0);
+
     const cv::Mat img = readImage(GetParam(), cv::IMREAD_GRAYSCALE);
     ASSERT_FALSE(img.empty());
 
diff --git a/modules/gpu/perf/perf_precomp.cpp b/modules/gpu/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/gpu/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/gpu/perf/perf_precomp.hpp b/modules/gpu/perf/perf_precomp.hpp
index 71fe9e7d07..5bfb14b52e 100644
--- a/modules/gpu/perf/perf_precomp.hpp
+++ b/modules/gpu/perf/perf_precomp.hpp
@@ -1,3 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifdef __GNUC__
 #  pragma GCC diagnostic ignored "-Wmissing-declarations"
 #  if defined __clang__ || defined __APPLE__
@@ -9,28 +51,12 @@
 #ifndef __OPENCV_PERF_PRECOMP_HPP__
 #define __OPENCV_PERF_PRECOMP_HPP__
 
-#include <cstdio>
-#include <iostream>
-
-#include "cvconfig.h"
-
-#ifdef HAVE_CUDA
-#include <cuda_runtime.h>
-#endif
-
 #include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-#include "opencv2/core.hpp"
-#include "opencv2/highgui.hpp"
 #include "opencv2/gpu.hpp"
 #include "opencv2/calib3d.hpp"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/video.hpp"
-#include "opencv2/nonfree.hpp"
-#include "opencv2/legacy.hpp"
-#include "opencv2/photo.hpp"
-
-#include "utility.hpp"
+#include "opencv2/objdetect.hpp"
 
 #ifdef GTEST_CREATE_SHARED_LIBRARY
 #error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
diff --git a/modules/gpu/perf/perf_video.cpp b/modules/gpu/perf/perf_video.cpp
deleted file mode 100644
index 61c2687ca9..0000000000
--- a/modules/gpu/perf/perf_video.cpp
+++ /dev/null
@@ -1,1094 +0,0 @@
-#include "perf_precomp.hpp"
-
-using namespace std;
-using namespace testing;
-using namespace perf;
-
-namespace cv
-{
-    template<> void Ptr<CvBGStatModel>::delete_obj()
-    {
-        cvReleaseBGStatModel(&obj);
-    }
-}
-
-//////////////////////////////////////////////////////
-// InterpolateFrames
-
-typedef pair<string, string> pair_string;
-
-DEF_PARAM_TEST_1(ImagePair, pair_string);
-
-PERF_TEST_P(ImagePair, Video_InterpolateFrames,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
-    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat d_fu, d_fv;
-        cv::gpu::GpuMat d_bu, d_bv;
-
-        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
-                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
-
-        d_flow(d_frame0, d_frame1, d_fu, d_fv);
-        d_flow(d_frame1, d_frame0, d_bu, d_bv);
-
-        cv::gpu::GpuMat newFrame;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::interpolateFrames(d_frame0, d_frame1, d_fu, d_fv, d_bu, d_bv, 0.5f, newFrame, d_buf);
-
-        GPU_SANITY_CHECK(newFrame);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// CreateOpticalFlowNeedleMap
-
-PERF_TEST_P(ImagePair, Video_CreateOpticalFlowNeedleMap,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
-    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u;
-        cv::gpu::GpuMat v;
-
-        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
-                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
-
-        d_flow(d_frame0, d_frame1, u, v);
-
-        cv::gpu::GpuMat vertex, colors;
-
-        TEST_CYCLE() cv::gpu::createOpticalFlowNeedleMap(u, v, vertex, colors);
-
-        GPU_SANITY_CHECK(vertex);
-        GPU_SANITY_CHECK(colors);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// GoodFeaturesToTrack
-
-DEF_PARAM_TEST(Image_MinDistance, string, double);
-
-PERF_TEST_P(Image_MinDistance, Video_GoodFeaturesToTrack,
-            Combine(Values<string>("gpu/perf/aloe.png"),
-                    Values(0.0, 3.0)))
-{
-    const string fileName = GET_PARAM(0);
-    const double minDistance = GET_PARAM(1);
-
-    const cv::Mat image = readImage(fileName, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    const int maxCorners = 8000;
-    const double qualityLevel = 0.01;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GoodFeaturesToTrackDetector_GPU d_detector(maxCorners, qualityLevel, minDistance);
-
-        const cv::gpu::GpuMat d_image(image);
-        cv::gpu::GpuMat pts;
-
-        TEST_CYCLE() d_detector(d_image, pts);
-
-        GPU_SANITY_CHECK(pts);
-    }
-    else
-    {
-        cv::Mat pts;
-
-        TEST_CYCLE() cv::goodFeaturesToTrack(image, pts, maxCorners, qualityLevel, minDistance);
-
-        CPU_SANITY_CHECK(pts);
-    }
-}
-
-//////////////////////////////////////////////////////
-// BroxOpticalFlow
-
-PERF_TEST_P(ImagePair, Video_BroxOpticalFlow,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    declare.time(10);
-
-    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
-    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u;
-        cv::gpu::GpuMat v;
-
-        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
-                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
-
-        TEST_CYCLE() d_flow(d_frame0, d_frame1, u, v);
-
-        GPU_SANITY_CHECK(u);
-        GPU_SANITY_CHECK(v);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// PyrLKOpticalFlowSparse
-
-DEF_PARAM_TEST(ImagePair_Gray_NPts_WinSz_Levels_Iters, pair_string, bool, int, int, int, int);
-
-PERF_TEST_P(ImagePair_Gray_NPts_WinSz_Levels_Iters, Video_PyrLKOpticalFlowSparse,
-            Combine(Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")),
-                    Bool(),
-                    Values(8000),
-                    Values(21),
-                    Values(1, 3),
-                    Values(1, 30)))
-{
-    declare.time(20.0);
-
-    const pair_string imagePair = GET_PARAM(0);
-    const bool useGray = GET_PARAM(1);
-    const int points = GET_PARAM(2);
-    const int winSize = GET_PARAM(3);
-    const int levels = GET_PARAM(4);
-    const int iters = GET_PARAM(5);
-
-    const cv::Mat frame0 = readImage(imagePair.first, useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(imagePair.second, useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    ASSERT_FALSE(frame1.empty());
-
-    cv::Mat gray_frame;
-    if (useGray)
-        gray_frame = frame0;
-    else
-        cv::cvtColor(frame0, gray_frame, cv::COLOR_BGR2GRAY);
-
-    cv::Mat pts;
-    cv::goodFeaturesToTrack(gray_frame, pts, points, 0.01, 0.0);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_pts(pts.reshape(2, 1));
-
-        cv::gpu::PyrLKOpticalFlow d_pyrLK;
-        d_pyrLK.winSize = cv::Size(winSize, winSize);
-        d_pyrLK.maxLevel = levels - 1;
-        d_pyrLK.iters = iters;
-
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat nextPts;
-        cv::gpu::GpuMat status;
-
-        TEST_CYCLE() d_pyrLK.sparse(d_frame0, d_frame1, d_pts, nextPts, status);
-
-        GPU_SANITY_CHECK(nextPts);
-        GPU_SANITY_CHECK(status);
-    }
-    else
-    {
-        cv::Mat nextPts;
-        cv::Mat status;
-
-        TEST_CYCLE()
-        {
-            cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, cv::noArray(),
-                                     cv::Size(winSize, winSize), levels - 1,
-                                     cv::TermCriteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, iters, 0.01));
-        }
-
-        CPU_SANITY_CHECK(nextPts);
-        CPU_SANITY_CHECK(status);
-    }
-}
-
-//////////////////////////////////////////////////////
-// PyrLKOpticalFlowDense
-
-DEF_PARAM_TEST(ImagePair_WinSz_Levels_Iters, pair_string, int, int, int);
-
-PERF_TEST_P(ImagePair_WinSz_Levels_Iters, Video_PyrLKOpticalFlowDense,
-            Combine(Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")),
-                    Values(3, 5, 7, 9, 13, 17, 21),
-                    Values(1, 3),
-                    Values(1, 10)))
-{
-    declare.time(30);
-
-    const pair_string imagePair = GET_PARAM(0);
-    const int winSize = GET_PARAM(1);
-    const int levels = GET_PARAM(2);
-    const int iters = GET_PARAM(3);
-
-    const cv::Mat frame0 = readImage(imagePair.first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(imagePair.second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u;
-        cv::gpu::GpuMat v;
-
-        cv::gpu::PyrLKOpticalFlow d_pyrLK;
-        d_pyrLK.winSize = cv::Size(winSize, winSize);
-        d_pyrLK.maxLevel = levels - 1;
-        d_pyrLK.iters = iters;
-
-        TEST_CYCLE() d_pyrLK.dense(d_frame0, d_frame1, u, v);
-
-        GPU_SANITY_CHECK(u);
-        GPU_SANITY_CHECK(v);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// FarnebackOpticalFlow
-
-PERF_TEST_P(ImagePair, Video_FarnebackOpticalFlow,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    declare.time(10);
-
-    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    const int numLevels = 5;
-    const double pyrScale = 0.5;
-    const int winSize = 13;
-    const int numIters = 10;
-    const int polyN = 5;
-    const double polySigma = 1.1;
-    const int flags = 0;
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u;
-        cv::gpu::GpuMat v;
-
-        cv::gpu::FarnebackOpticalFlow d_farneback;
-        d_farneback.numLevels = numLevels;
-        d_farneback.pyrScale = pyrScale;
-        d_farneback.winSize = winSize;
-        d_farneback.numIters = numIters;
-        d_farneback.polyN = polyN;
-        d_farneback.polySigma = polySigma;
-        d_farneback.flags = flags;
-
-        TEST_CYCLE() d_farneback(d_frame0, d_frame1, u, v);
-
-        GPU_SANITY_CHECK(u, 1e-4);
-        GPU_SANITY_CHECK(v, 1e-4);
-    }
-    else
-    {
-        cv::Mat flow;
-
-        TEST_CYCLE() cv::calcOpticalFlowFarneback(frame0, frame1, flow, pyrScale, numLevels, winSize, numIters, polyN, polySigma, flags);
-
-        CPU_SANITY_CHECK(flow);
-    }
-}
-
-//////////////////////////////////////////////////////
-// OpticalFlowDual_TVL1
-
-PERF_TEST_P(ImagePair, Video_OpticalFlowDual_TVL1,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    declare.time(20);
-
-    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u;
-        cv::gpu::GpuMat v;
-
-        cv::gpu::OpticalFlowDual_TVL1_GPU d_alg;
-
-        TEST_CYCLE() d_alg(d_frame0, d_frame1, u, v);
-
-        GPU_SANITY_CHECK(u, 1e-4);
-        GPU_SANITY_CHECK(v, 1e-4);
-    }
-    else
-    {
-        cv::Mat flow;
-
-        cv::Ptr<cv::DenseOpticalFlow> alg = cv::createOptFlow_DualTVL1();
-
-        TEST_CYCLE() alg->calc(frame0, frame1, flow);
-
-        CPU_SANITY_CHECK(flow);
-    }
-}
-
-//////////////////////////////////////////////////////
-// OpticalFlowBM
-
-void calcOpticalFlowBM(const cv::Mat& prev, const cv::Mat& curr,
-                       cv::Size bSize, cv::Size shiftSize, cv::Size maxRange, int usePrevious,
-                       cv::Mat& velx, cv::Mat& vely)
-{
-    cv::Size sz((curr.cols - bSize.width + shiftSize.width)/shiftSize.width, (curr.rows - bSize.height + shiftSize.height)/shiftSize.height);
-
-    velx.create(sz, CV_32FC1);
-    vely.create(sz, CV_32FC1);
-
-    CvMat cvprev = prev;
-    CvMat cvcurr = curr;
-
-    CvMat cvvelx = velx;
-    CvMat cvvely = vely;
-
-    cvCalcOpticalFlowBM(&cvprev, &cvcurr, bSize, shiftSize, maxRange, usePrevious, &cvvelx, &cvvely);
-}
-
-PERF_TEST_P(ImagePair, Video_OpticalFlowBM,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    declare.time(400);
-
-    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    const cv::Size block_size(16, 16);
-    const cv::Size shift_size(1, 1);
-    const cv::Size max_range(16, 16);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u, v, buf;
-
-        TEST_CYCLE() cv::gpu::calcOpticalFlowBM(d_frame0, d_frame1, block_size, shift_size, max_range, false, u, v, buf);
-
-        GPU_SANITY_CHECK(u);
-        GPU_SANITY_CHECK(v);
-    }
-    else
-    {
-        cv::Mat u, v;
-
-        TEST_CYCLE() calcOpticalFlowBM(frame0, frame1, block_size, shift_size, max_range, false, u, v);
-
-        CPU_SANITY_CHECK(u);
-        CPU_SANITY_CHECK(v);
-    }
-}
-
-PERF_TEST_P(ImagePair, Video_FastOpticalFlowBM,
-            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
-{
-    declare.time(400);
-
-    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame0.empty());
-
-    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(frame1.empty());
-
-    const cv::Size block_size(16, 16);
-    const cv::Size shift_size(1, 1);
-    const cv::Size max_range(16, 16);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_frame0(frame0);
-        const cv::gpu::GpuMat d_frame1(frame1);
-        cv::gpu::GpuMat u, v;
-
-        cv::gpu::FastOpticalFlowBM fastBM;
-
-        TEST_CYCLE() fastBM(d_frame0, d_frame1, u, v, max_range.width, block_size.width);
-
-        GPU_SANITY_CHECK(u, 2);
-        GPU_SANITY_CHECK(v, 2);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// FGDStatModel
-
-DEF_PARAM_TEST_1(Video, string);
-
-PERF_TEST_P(Video, Video_FGDStatModel,
-            Values(string("gpu/video/768x576.avi")))
-{
-    declare.time(60);
-
-    const string inputFile = perf::TestBase::getDataPath(GetParam());
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-    cap >> frame;
-    ASSERT_FALSE(frame.empty());
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_frame(frame);
-
-        cv::gpu::FGDStatModel d_model(4);
-        d_model.create(d_frame);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            d_frame.upload(frame);
-
-            startTimer(); next();
-            d_model.update(d_frame);
-            stopTimer();
-        }
-
-        const cv::gpu::GpuMat background = d_model.background;
-        const cv::gpu::GpuMat foreground = d_model.foreground;
-
-        GPU_SANITY_CHECK(background, 1e-2, ERROR_RELATIVE);
-        GPU_SANITY_CHECK(foreground, 1e-2, ERROR_RELATIVE);
-    }
-    else
-    {
-        IplImage ipl_frame = frame;
-        cv::Ptr<CvBGStatModel> model(cvCreateFGDStatModel(&ipl_frame));
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            ipl_frame = frame;
-
-            startTimer(); next();
-            cvUpdateBGStatModel(&ipl_frame, model);
-            stopTimer();
-        }
-
-        const cv::Mat background = model->background;
-        const cv::Mat foreground = model->foreground;
-
-        CPU_SANITY_CHECK(background);
-        CPU_SANITY_CHECK(foreground);
-    }
-}
-
-//////////////////////////////////////////////////////
-// MOG
-
-DEF_PARAM_TEST(Video_Cn_LearningRate, string, MatCn, double);
-
-PERF_TEST_P(Video_Cn_LearningRate, Video_MOG,
-            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
-                    GPU_CHANNELS_1_3_4,
-                    Values(0.0, 0.01)))
-{
-    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
-    const int cn = GET_PARAM(1);
-    const float learningRate = static_cast<float>(GET_PARAM(2));
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-
-    cap >> frame;
-    ASSERT_FALSE(frame.empty());
-
-    if (cn != 3)
-    {
-        cv::Mat temp;
-        if (cn == 1)
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-        else
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-        cv::swap(temp, frame);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_frame(frame);
-        cv::gpu::MOG_GPU d_mog;
-        cv::gpu::GpuMat foreground;
-
-        d_mog(d_frame, foreground, learningRate);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            d_frame.upload(frame);
-
-            startTimer(); next();
-            d_mog(d_frame, foreground, learningRate);
-            stopTimer();
-        }
-
-        GPU_SANITY_CHECK(foreground);
-    }
-    else
-    {
-        cv::BackgroundSubtractorMOG mog;
-        cv::Mat foreground;
-
-        mog(frame, foreground, learningRate);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            startTimer(); next();
-            mog(frame, foreground, learningRate);
-            stopTimer();
-        }
-
-        CPU_SANITY_CHECK(foreground);
-    }
-}
-
-//////////////////////////////////////////////////////
-// MOG2
-
-DEF_PARAM_TEST(Video_Cn, string, int);
-
-PERF_TEST_P(Video_Cn, Video_MOG2,
-            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
-                    GPU_CHANNELS_1_3_4))
-{
-    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
-    const int cn = GET_PARAM(1);
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-
-    cap >> frame;
-    ASSERT_FALSE(frame.empty());
-
-    if (cn != 3)
-    {
-        cv::Mat temp;
-        if (cn == 1)
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-        else
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-        cv::swap(temp, frame);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::MOG2_GPU d_mog2;
-        d_mog2.bShadowDetection = false;
-
-        cv::gpu::GpuMat d_frame(frame);
-        cv::gpu::GpuMat foreground;
-
-        d_mog2(d_frame, foreground);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            d_frame.upload(frame);
-
-            startTimer(); next();
-            d_mog2(d_frame, foreground);
-            stopTimer();
-        }
-
-        GPU_SANITY_CHECK(foreground);
-    }
-    else
-    {
-        cv::BackgroundSubtractorMOG2 mog2;
-        mog2.set("detectShadows", false);
-
-        cv::Mat foreground;
-
-        mog2(frame, foreground);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            startTimer(); next();
-            mog2(frame, foreground);
-            stopTimer();
-        }
-
-        CPU_SANITY_CHECK(foreground);
-    }
-}
-
-//////////////////////////////////////////////////////
-// MOG2GetBackgroundImage
-
-PERF_TEST_P(Video_Cn, Video_MOG2GetBackgroundImage,
-            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
-                    GPU_CHANNELS_1_3_4))
-{
-    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
-    const int cn = GET_PARAM(1);
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_frame;
-        cv::gpu::MOG2_GPU d_mog2;
-        cv::gpu::GpuMat d_foreground;
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            d_frame.upload(frame);
-
-            d_mog2(d_frame, d_foreground);
-        }
-
-        cv::gpu::GpuMat background;
-
-        TEST_CYCLE() d_mog2.getBackgroundImage(background);
-
-        GPU_SANITY_CHECK(background, 1);
-    }
-    else
-    {
-        cv::BackgroundSubtractorMOG2 mog2;
-        cv::Mat foreground;
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            mog2(frame, foreground);
-        }
-
-        cv::Mat background;
-
-        TEST_CYCLE() mog2.getBackgroundImage(background);
-
-        CPU_SANITY_CHECK(background);
-    }
-}
-
-//////////////////////////////////////////////////////
-// VIBE
-
-PERF_TEST_P(Video_Cn, Video_VIBE,
-            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
-                    GPU_CHANNELS_1_3_4))
-{
-    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
-    const int cn = GET_PARAM(1);
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-    cap >> frame;
-    ASSERT_FALSE(frame.empty());
-
-    if (cn != 3)
-    {
-        cv::Mat temp;
-        if (cn == 1)
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-        else
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-        cv::swap(temp, frame);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_frame(frame);
-        cv::gpu::VIBE_GPU vibe;
-        cv::gpu::GpuMat foreground;
-
-        vibe(d_frame, foreground);
-
-        for (int i = 0; i < 10; ++i)
-        {
-            cap >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            d_frame.upload(frame);
-
-            startTimer(); next();
-            vibe(d_frame, foreground);
-            stopTimer();
-        }
-
-        GPU_SANITY_CHECK(foreground);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////
-// GMG
-
-DEF_PARAM_TEST(Video_Cn_MaxFeatures, string, MatCn, int);
-
-PERF_TEST_P(Video_Cn_MaxFeatures, Video_GMG,
-            Combine(Values(string("gpu/video/768x576.avi")),
-                    GPU_CHANNELS_1_3_4,
-                    Values(20, 40, 60)))
-{
-    const std::string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
-    const int cn = GET_PARAM(1);
-    const int maxFeatures = GET_PARAM(2);
-
-    cv::VideoCapture cap(inputFile);
-    ASSERT_TRUE(cap.isOpened());
-
-    cv::Mat frame;
-    cap >> frame;
-    ASSERT_FALSE(frame.empty());
-
-    if (cn != 3)
-    {
-        cv::Mat temp;
-        if (cn == 1)
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-        else
-            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-        cv::swap(temp, frame);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::GpuMat d_frame(frame);
-        cv::gpu::GpuMat foreground;
-
-        cv::gpu::GMG_GPU d_gmg;
-        d_gmg.maxFeatures = maxFeatures;
-
-        d_gmg(d_frame, foreground);
-
-        for (int i = 0; i < 150; ++i)
-        {
-            cap >> frame;
-            if (frame.empty())
-            {
-                cap.release();
-                cap.open(inputFile);
-                cap >> frame;
-            }
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            d_frame.upload(frame);
-
-            startTimer(); next();
-            d_gmg(d_frame, foreground);
-            stopTimer();
-        }
-
-        GPU_SANITY_CHECK(foreground);
-    }
-    else
-    {
-        cv::Mat foreground;
-        cv::Mat zeros(frame.size(), CV_8UC1, cv::Scalar::all(0));
-
-        cv::BackgroundSubtractorGMG gmg;
-        gmg.set("maxFeatures", maxFeatures);
-        gmg.initialize(frame.size(), 0.0, 255.0);
-
-        gmg(frame, foreground);
-
-        for (int i = 0; i < 150; ++i)
-        {
-            cap >> frame;
-            if (frame.empty())
-            {
-                cap.release();
-                cap.open(inputFile);
-                cap >> frame;
-            }
-
-            if (cn != 3)
-            {
-                cv::Mat temp;
-                if (cn == 1)
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
-                else
-                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
-                cv::swap(temp, frame);
-            }
-
-            startTimer(); next();
-            gmg(frame, foreground);
-            stopTimer();
-        }
-
-        CPU_SANITY_CHECK(foreground);
-    }
-}
-
-#ifdef HAVE_NVCUVID
-
-//////////////////////////////////////////////////////
-// VideoReader
-
-PERF_TEST_P(Video, Video_VideoReader, Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"))
-{
-    declare.time(20);
-
-    const string inputFile = perf::TestBase::getDataPath(GetParam());
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::VideoReader_GPU d_reader(inputFile);
-        ASSERT_TRUE( d_reader.isOpened() );
-
-        cv::gpu::GpuMat frame;
-
-        TEST_CYCLE_N(10) d_reader.read(frame);
-
-        GPU_SANITY_CHECK(frame);
-    }
-    else
-    {
-        cv::VideoCapture reader(inputFile);
-        ASSERT_TRUE( reader.isOpened() );
-
-        cv::Mat frame;
-
-        TEST_CYCLE_N(10) reader >> frame;
-
-        CPU_SANITY_CHECK(frame);
-    }
-}
-
-//////////////////////////////////////////////////////
-// VideoWriter
-
-#ifdef WIN32
-
-PERF_TEST_P(Video, Video_VideoWriter, Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"))
-{
-    declare.time(30);
-
-    const string inputFile = perf::TestBase::getDataPath(GetParam());
-    const string outputFile = cv::tempfile(".avi");
-
-    const double FPS = 25.0;
-
-    cv::VideoCapture reader(inputFile);
-    ASSERT_TRUE( reader.isOpened() );
-
-    cv::Mat frame;
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::VideoWriter_GPU d_writer;
-
-        cv::gpu::GpuMat d_frame;
-
-        for (int i = 0; i < 10; ++i)
-        {
-            reader >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            d_frame.upload(frame);
-
-            if (!d_writer.isOpened())
-                d_writer.open(outputFile, frame.size(), FPS);
-
-            startTimer(); next();
-            d_writer.write(d_frame);
-            stopTimer();
-        }
-    }
-    else
-    {
-        cv::VideoWriter writer;
-
-        for (int i = 0; i < 10; ++i)
-        {
-            reader >> frame;
-            ASSERT_FALSE(frame.empty());
-
-            if (!writer.isOpened())
-                writer.open(outputFile, CV_FOURCC('X', 'V', 'I', 'D'), FPS, frame.size());
-
-            startTimer(); next();
-            writer.write(frame);
-            stopTimer();
-        }
-    }
-
-    SANITY_CHECK(frame);
-}
-
-#endif // WIN32
-
-#endif // HAVE_NVCUVID
diff --git a/modules/gpu/perf/utility.cpp b/modules/gpu/perf/utility.cpp
deleted file mode 100644
index 16c61e0c7d..0000000000
--- a/modules/gpu/perf/utility.cpp
+++ /dev/null
@@ -1,184 +0,0 @@
-#include "perf_precomp.hpp"
-
-using namespace std;
-using namespace cv;
-
-Mat readImage(const string& fileName, int flags)
-{
-    return imread(perf::TestBase::getDataPath(fileName), flags);
-}
-
-void PrintTo(const CvtColorInfo& info, ostream* os)
-{
-    static const char* str[] =
-    {
-        "BGR2BGRA",
-        "BGRA2BGR",
-        "BGR2RGBA",
-        "RGBA2BGR",
-        "BGR2RGB",
-        "BGRA2RGBA",
-
-        "BGR2GRAY",
-        "RGB2GRAY",
-        "GRAY2BGR",
-        "GRAY2BGRA",
-        "BGRA2GRAY",
-        "RGBA2GRAY",
-
-        "BGR2BGR565",
-        "RGB2BGR565",
-        "BGR5652BGR",
-        "BGR5652RGB",
-        "BGRA2BGR565",
-        "RGBA2BGR565",
-        "BGR5652BGRA",
-        "BGR5652RGBA",
-
-        "GRAY2BGR565",
-        "BGR5652GRAY",
-
-        "BGR2BGR555",
-        "RGB2BGR555",
-        "BGR5552BGR",
-        "BGR5552RGB",
-        "BGRA2BGR555",
-        "RGBA2BGR555",
-        "BGR5552BGRA",
-        "BGR5552RGBA",
-
-        "GRAY2BGR555",
-        "BGR5552GRAY",
-
-        "BGR2XYZ",
-        "RGB2XYZ",
-        "XYZ2BGR",
-        "XYZ2RGB",
-
-        "BGR2YCrCb",
-        "RGB2YCrCb",
-        "YCrCb2BGR",
-        "YCrCb2RGB",
-
-        "BGR2HSV",
-        "RGB2HSV",
-
-        "",
-        "",
-
-        "BGR2Lab",
-        "RGB2Lab",
-
-        "BayerBG2BGR",
-        "BayerGB2BGR",
-        "BayerRG2BGR",
-        "BayerGR2BGR",
-
-        "BGR2Luv",
-        "RGB2Luv",
-
-        "BGR2HLS",
-        "RGB2HLS",
-
-        "HSV2BGR",
-        "HSV2RGB",
-
-        "Lab2BGR",
-        "Lab2RGB",
-        "Luv2BGR",
-        "Luv2RGB",
-
-        "HLS2BGR",
-        "HLS2RGB",
-
-        "BayerBG2BGR_VNG",
-        "BayerGB2BGR_VNG",
-        "BayerRG2BGR_VNG",
-        "BayerGR2BGR_VNG",
-
-        "BGR2HSV_FULL",
-        "RGB2HSV_FULL",
-        "BGR2HLS_FULL",
-        "RGB2HLS_FULL",
-
-        "HSV2BGR_FULL",
-        "HSV2RGB_FULL",
-        "HLS2BGR_FULL",
-        "HLS2RGB_FULL",
-
-        "LBGR2Lab",
-        "LRGB2Lab",
-        "LBGR2Luv",
-        "LRGB2Luv",
-
-        "Lab2LBGR",
-        "Lab2LRGB",
-        "Luv2LBGR",
-        "Luv2LRGB",
-
-        "BGR2YUV",
-        "RGB2YUV",
-        "YUV2BGR",
-        "YUV2RGB",
-
-        "BayerBG2GRAY",
-        "BayerGB2GRAY",
-        "BayerRG2GRAY",
-        "BayerGR2GRAY",
-
-        //YUV 4:2:0 formats family
-        "YUV2RGB_NV12",
-        "YUV2BGR_NV12",
-        "YUV2RGB_NV21",
-        "YUV2BGR_NV21",
-
-        "YUV2RGBA_NV12",
-        "YUV2BGRA_NV12",
-        "YUV2RGBA_NV21",
-        "YUV2BGRA_NV21",
-
-        "YUV2RGB_YV12",
-        "YUV2BGR_YV12",
-        "YUV2RGB_IYUV",
-        "YUV2BGR_IYUV",
-
-        "YUV2RGBA_YV12",
-        "YUV2BGRA_YV12",
-        "YUV2RGBA_IYUV",
-        "YUV2BGRA_IYUV",
-
-        "YUV2GRAY_420",
-
-        //YUV 4:2:2 formats family
-        "YUV2RGB_UYVY",
-        "YUV2BGR_UYVY",
-        "YUV2RGB_VYUY",
-        "YUV2BGR_VYUY",
-
-        "YUV2RGBA_UYVY",
-        "YUV2BGRA_UYVY",
-        "YUV2RGBA_VYUY",
-        "YUV2BGRA_VYUY",
-
-        "YUV2RGB_YUY2",
-        "YUV2BGR_YUY2",
-        "YUV2RGB_YVYU",
-        "YUV2BGR_YVYU",
-
-        "YUV2RGBA_YUY2",
-        "YUV2BGRA_YUY2",
-        "YUV2RGBA_YVYU",
-        "YUV2BGRA_YVYU",
-
-        "YUV2GRAY_UYVY",
-        "YUV2GRAY_YUY2",
-
-        // alpha premultiplication
-        "RGBA2mRGBA",
-        "mRGBA2RGBA",
-
-        "COLORCVT_MAX"
-    };
-
-    *os << str[info.code];
-}
diff --git a/modules/gpu/perf/utility.hpp b/modules/gpu/perf/utility.hpp
deleted file mode 100644
index 18c85854ac..0000000000
--- a/modules/gpu/perf/utility.hpp
+++ /dev/null
@@ -1,63 +0,0 @@
-#ifndef __OPENCV_PERF_GPU_UTILITY_HPP__
-#define __OPENCV_PERF_GPU_UTILITY_HPP__
-
-#include "opencv2/core.hpp"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/ts/ts_perf.hpp"
-
-cv::Mat readImage(const std::string& fileName, int flags = cv::IMREAD_COLOR);
-
-using perf::MatType;
-using perf::MatDepth;
-
-CV_ENUM(BorderMode, cv::BORDER_REFLECT101, cv::BORDER_REPLICATE, cv::BORDER_CONSTANT, cv::BORDER_REFLECT, cv::BORDER_WRAP)
-#define ALL_BORDER_MODES testing::ValuesIn(BorderMode::all())
-
-CV_ENUM(Interpolation, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::INTER_AREA)
-#define ALL_INTERPOLATIONS testing::ValuesIn(Interpolation::all())
-
-CV_ENUM(NormType, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_HAMMING, cv::NORM_MINMAX)
-
-enum { Gray = 1, TwoChannel = 2, BGR = 3, BGRA = 4 };
-CV_ENUM(MatCn, Gray, TwoChannel, BGR, BGRA)
-#define GPU_CHANNELS_1_3_4 testing::Values(MatCn(Gray), MatCn(BGR), MatCn(BGRA))
-#define GPU_CHANNELS_1_3 testing::Values(MatCn(Gray), MatCn(BGR))
-
-struct CvtColorInfo
-{
-    int scn;
-    int dcn;
-    int code;
-
-    CvtColorInfo() {}
-    explicit CvtColorInfo(int scn_, int dcn_, int code_) : scn(scn_), dcn(dcn_), code(code_) {}
-};
-void PrintTo(const CvtColorInfo& info, std::ostream* os);
-
-#define GET_PARAM(k) std::tr1::get< k >(GetParam())
-
-#define DEF_PARAM_TEST(name, ...) typedef ::perf::TestBaseWithParam< std::tr1::tuple< __VA_ARGS__ > > name
-#define DEF_PARAM_TEST_1(name, param_type) typedef ::perf::TestBaseWithParam< param_type > name
-
-DEF_PARAM_TEST_1(Sz, cv::Size);
-typedef perf::Size_MatType Sz_Type;
-DEF_PARAM_TEST(Sz_Depth, cv::Size, MatDepth);
-DEF_PARAM_TEST(Sz_Depth_Cn, cv::Size, MatDepth, MatCn);
-
-#define GPU_TYPICAL_MAT_SIZES testing::Values(perf::sz720p, perf::szSXGA, perf::sz1080p)
-
-#define FAIL_NO_CPU() FAIL() << "No such CPU implementation analogy"
-
-#define GPU_SANITY_CHECK(mat, ...) \
-    do{ \
-        cv::Mat gpu_##mat(mat); \
-        SANITY_CHECK(gpu_##mat, ## __VA_ARGS__); \
-    } while(0)
-
-#define CPU_SANITY_CHECK(mat, ...) \
-    do{ \
-        cv::Mat cpu_##mat(mat); \
-        SANITY_CHECK(cpu_##mat, ## __VA_ARGS__); \
-    } while(0)
-
-#endif // __OPENCV_PERF_GPU_UTILITY_HPP__
diff --git a/modules/gpu/perf4au/CMakeLists.txt b/modules/gpu/perf4au/CMakeLists.txt
index 7452203826..cb395069d0 100644
--- a/modules/gpu/perf4au/CMakeLists.txt
+++ b/modules/gpu/perf4au/CMakeLists.txt
@@ -1,4 +1,4 @@
-set(PERF4AU_REQUIRED_DEPS opencv_core opencv_imgproc opencv_highgui opencv_video opencv_legacy opencv_gpu opencv_ts)
+set(PERF4AU_REQUIRED_DEPS opencv_core opencv_imgproc opencv_highgui opencv_video opencv_legacy opencv_ml opencv_ts opencv_gpufilters opencv_gpuimgproc opencv_gpuoptflow)
 
 ocv_check_dependencies(${PERF4AU_REQUIRED_DEPS})
 
@@ -25,4 +25,3 @@ if(WIN32)
         set_target_properties(${the_target} PROPERTIES LINK_FLAGS "/NODEFAULTLIB:atlthunk.lib /NODEFAULTLIB:atlsd.lib /DEBUG")
     endif()
 endif()
-
diff --git a/modules/gpu/perf4au/main.cpp b/modules/gpu/perf4au/main.cpp
index f4a04432b5..707251c42f 100644
--- a/modules/gpu/perf4au/main.cpp
+++ b/modules/gpu/perf4au/main.cpp
@@ -1,75 +1,63 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include <cstdio>
-#ifdef HAVE_CVCONFIG_H
+
 #include "cvconfig.h"
-#endif
 #include "opencv2/core.hpp"
-#include "opencv2/gpu.hpp"
+#include "opencv2/gpuimgproc.hpp"
+#include "opencv2/gpuoptflow.hpp"
 #include "opencv2/highgui.hpp"
 #include "opencv2/video.hpp"
 #include "opencv2/legacy.hpp"
 #include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-static void printOsInfo()
-{
-#if defined _WIN32
-#   if defined _WIN64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#elif defined linux
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#elif defined __APPLE__
-#   if defined _LP64
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x64.\n[----------]\n"); fflush(stdout);
-#   else
-        printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x32.\n[----------]\n"); fflush(stdout);
-#   endif
-#endif
-}
+static const char * impls[] = {
+    "cuda",
+    "plain"
+};
 
-static void printCudaInfo()
-{
-    const int deviceCount = cv::gpu::getCudaEnabledDeviceCount();
-
-    printf("[----------]\n"); fflush(stdout);
-    printf("[ GPU INFO ] \tCUDA device count:: %d.\n", deviceCount); fflush(stdout);
-    printf("[----------]\n"); fflush(stdout);
-
-    for (int i = 0; i < deviceCount; ++i)
-    {
-        cv::gpu::DeviceInfo info(i);
-
-        printf("[----------]\n"); fflush(stdout);
-        printf("[ DEVICE   ] \t# %d %s.\n", i, info.name().c_str()); fflush(stdout);
-        printf("[          ] \tCompute capability: %d.%d\n", info.majorVersion(), info.minorVersion()); fflush(stdout);
-        printf("[          ] \tMulti Processor Count:  %d\n", info.multiProcessorCount()); fflush(stdout);
-        printf("[          ] \tTotal memory: %d Mb\n", static_cast<int>(static_cast<int>(info.totalMemory() / 1024.0) / 1024.0)); fflush(stdout);
-        printf("[          ] \tFree  memory: %d Mb\n", static_cast<int>(static_cast<int>(info.freeMemory()  / 1024.0) / 1024.0)); fflush(stdout);
-        if (!info.isCompatible())
-            printf("[ GPU INFO ] \tThis device is NOT compatible with current GPU module build\n");
-        printf("[----------]\n"); fflush(stdout);
-    }
-}
-
-int main(int argc, char* argv[])
-{
-    printOsInfo();
-    printCudaInfo();
-
-    perf::Regression::Init("nv_perf_test");
-    perf::TestBase::Init(argc, argv);
-    testing::InitGoogleTest(&argc, argv);
-
-    return RUN_ALL_TESTS();
-}
-
-#define DEF_PARAM_TEST(name, ...) typedef ::perf::TestBaseWithParam< std::tr1::tuple< __VA_ARGS__ > > name
-#define DEF_PARAM_TEST_1(name, param_type) typedef ::perf::TestBaseWithParam< param_type > name
+CV_PERF_TEST_MAIN_WITH_IMPLS(gpu_perf4au, impls, perf::printCudaInfo())
 
 //////////////////////////////////////////////////////////
 // HoughLinesP
@@ -94,13 +82,14 @@ PERF_TEST_P(Image, HoughLinesP, testing::Values(std::string("im1_1280x800.jpg"))
     {
         cv::gpu::GpuMat d_image(image);
         cv::gpu::GpuMat d_lines;
-        cv::gpu::HoughLinesBuf d_buf;
 
-        cv::gpu::HoughLinesP(d_image, d_lines, d_buf, rho, theta, minLineLenght, maxLineGap);
+        cv::Ptr<cv::gpu::HoughSegmentDetector> hough = cv::gpu::createHoughSegmentDetector(rho, theta, minLineLenght, maxLineGap);
+
+        hough->detect(d_image, d_lines);
 
         TEST_CYCLE()
         {
-            cv::gpu::HoughLinesP(d_image, d_lines, d_buf, rho, theta, minLineLenght, maxLineGap);
+            hough->detect(d_image, d_lines);
         }
     }
     else
@@ -155,17 +144,17 @@ PERF_TEST_P(Image_Depth, GoodFeaturesToTrack,
 
     if (PERF_RUN_GPU())
     {
-        cv::gpu::GoodFeaturesToTrackDetector_GPU d_detector(maxCorners, qualityLevel, minDistance, blockSize, useHarrisDetector, k);
+        cv::Ptr<cv::gpu::CornersDetector> detector = cv::gpu::createGoodFeaturesToTrackDetector(src.type(), maxCorners, qualityLevel, minDistance, blockSize, useHarrisDetector, k);
 
         cv::gpu::GpuMat d_src(src);
         cv::gpu::GpuMat d_mask(mask);
         cv::gpu::GpuMat d_pts;
 
-        d_detector(d_src, d_pts, d_mask);
+        detector->detect(d_src, d_pts, d_mask);
 
         TEST_CYCLE()
         {
-            d_detector(d_src, d_pts, d_mask);
+            detector->detect(d_src, d_pts, d_mask);
         }
     }
     else
diff --git a/modules/gpu/src/arithm.cpp b/modules/gpu/src/arithm.cpp
deleted file mode 100644
index 851ac938ef..0000000000
--- a/modules/gpu/src/arithm.cpp
+++ /dev/null
@@ -1,576 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::gemm(const GpuMat&, const GpuMat&, double, const GpuMat&, double, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::transpose(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::flip(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::LUT(const GpuMat&, const Mat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::magnitude(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::magnitudeSqr(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::magnitude(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::magnitudeSqr(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::phase(const GpuMat&, const GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
-void cv::gpu::cartToPolar(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
-void cv::gpu::polarToCart(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
-void cv::gpu::normalize(const GpuMat&, GpuMat&, double, double, int, int, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::normalize(const GpuMat&, GpuMat&, double, double, int, int, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-////////////////////////////////////////////////////////////////////////
-// gemm
-
-void cv::gpu::gemm(const GpuMat& src1, const GpuMat& src2, double alpha, const GpuMat& src3, double beta, GpuMat& dst, int flags, Stream& stream)
-{
-#ifndef HAVE_CUBLAS
-    (void)src1;
-    (void)src2;
-    (void)alpha;
-    (void)src3;
-    (void)beta;
-    (void)dst;
-    (void)flags;
-    (void)stream;
-    CV_Error(CV_StsNotImplemented, "The library was build without CUBLAS");
-#else
-    // CUBLAS works with column-major matrices
-
-    CV_Assert(src1.type() == CV_32FC1 || src1.type() == CV_32FC2 || src1.type() == CV_64FC1 || src1.type() == CV_64FC2);
-    CV_Assert(src2.type() == src1.type() && (src3.empty() || src3.type() == src1.type()));
-
-    if (src1.depth() == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    bool tr1 = (flags & GEMM_1_T) != 0;
-    bool tr2 = (flags & GEMM_2_T) != 0;
-    bool tr3 = (flags & GEMM_3_T) != 0;
-
-    if (src1.type() == CV_64FC2)
-    {
-        if (tr1 || tr2 || tr3)
-            CV_Error(CV_StsNotImplemented, "transpose operation doesn't implemented for CV_64FC2 type");
-    }
-
-    Size src1Size = tr1 ? Size(src1.rows, src1.cols) : src1.size();
-    Size src2Size = tr2 ? Size(src2.rows, src2.cols) : src2.size();
-    Size src3Size = tr3 ? Size(src3.rows, src3.cols) : src3.size();
-    Size dstSize(src2Size.width, src1Size.height);
-
-    CV_Assert(src1Size.width == src2Size.height);
-    CV_Assert(src3.empty() || src3Size == dstSize);
-
-    dst.create(dstSize, src1.type());
-
-    if (beta != 0)
-    {
-        if (src3.empty())
-        {
-            if (stream)
-                stream.enqueueMemSet(dst, Scalar::all(0));
-            else
-                dst.setTo(Scalar::all(0));
-        }
-        else
-        {
-            if (tr3)
-            {
-                transpose(src3, dst, stream);
-            }
-            else
-            {
-                if (stream)
-                    stream.enqueueCopy(src3, dst);
-                else
-                    src3.copyTo(dst);
-            }
-        }
-    }
-
-    cublasHandle_t handle;
-    cublasSafeCall( cublasCreate_v2(&handle) );
-
-    cublasSafeCall( cublasSetStream_v2(handle, StreamAccessor::getStream(stream)) );
-
-    cublasSafeCall( cublasSetPointerMode_v2(handle, CUBLAS_POINTER_MODE_HOST) );
-
-    const float alphaf = static_cast<float>(alpha);
-    const float betaf = static_cast<float>(beta);
-
-    const cuComplex alphacf = make_cuComplex(alphaf, 0);
-    const cuComplex betacf = make_cuComplex(betaf, 0);
-
-    const cuDoubleComplex alphac = make_cuDoubleComplex(alpha, 0);
-    const cuDoubleComplex betac = make_cuDoubleComplex(beta, 0);
-
-    cublasOperation_t transa = tr2 ? CUBLAS_OP_T : CUBLAS_OP_N;
-    cublasOperation_t transb = tr1 ? CUBLAS_OP_T : CUBLAS_OP_N;
-
-    switch (src1.type())
-    {
-    case CV_32FC1:
-        cublasSafeCall( cublasSgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
-            &alphaf,
-            src2.ptr<float>(), static_cast<int>(src2.step / sizeof(float)),
-            src1.ptr<float>(), static_cast<int>(src1.step / sizeof(float)),
-            &betaf,
-            dst.ptr<float>(), static_cast<int>(dst.step / sizeof(float))) );
-        break;
-
-    case CV_64FC1:
-        cublasSafeCall( cublasDgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
-            &alpha,
-            src2.ptr<double>(), static_cast<int>(src2.step / sizeof(double)),
-            src1.ptr<double>(), static_cast<int>(src1.step / sizeof(double)),
-            &beta,
-            dst.ptr<double>(), static_cast<int>(dst.step / sizeof(double))) );
-        break;
-
-    case CV_32FC2:
-        cublasSafeCall( cublasCgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
-            &alphacf,
-            src2.ptr<cuComplex>(), static_cast<int>(src2.step / sizeof(cuComplex)),
-            src1.ptr<cuComplex>(), static_cast<int>(src1.step / sizeof(cuComplex)),
-            &betacf,
-            dst.ptr<cuComplex>(), static_cast<int>(dst.step / sizeof(cuComplex))) );
-        break;
-
-    case CV_64FC2:
-        cublasSafeCall( cublasZgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
-            &alphac,
-            src2.ptr<cuDoubleComplex>(), static_cast<int>(src2.step / sizeof(cuDoubleComplex)),
-            src1.ptr<cuDoubleComplex>(), static_cast<int>(src1.step / sizeof(cuDoubleComplex)),
-            &betac,
-            dst.ptr<cuDoubleComplex>(), static_cast<int>(dst.step / sizeof(cuDoubleComplex))) );
-        break;
-    }
-
-    cublasSafeCall( cublasDestroy_v2(handle) );
-#endif
-}
-
-////////////////////////////////////////////////////////////////////////
-// transpose
-
-void cv::gpu::transpose(const GpuMat& src, GpuMat& dst, Stream& s)
-{
-    CV_Assert(src.elemSize() == 1 || src.elemSize() == 4 || src.elemSize() == 8);
-
-    dst.create( src.cols, src.rows, src.type() );
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    if (src.elemSize() == 1)
-    {
-        NppStreamHandler h(stream);
-
-        NppiSize sz;
-        sz.width  = src.cols;
-        sz.height = src.rows;
-
-        nppSafeCall( nppiTranspose_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-            dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz) );
-    }
-    else if (src.elemSize() == 4)
-    {
-        NppStStreamHandler h(stream);
-
-        NcvSize32u sz;
-        sz.width  = src.cols;
-        sz.height = src.rows;
-
-        ncvSafeCall( nppiStTranspose_32u_C1R(const_cast<Ncv32u*>(src.ptr<Ncv32u>()), static_cast<int>(src.step),
-            dst.ptr<Ncv32u>(), static_cast<int>(dst.step), sz) );
-    }
-    else // if (src.elemSize() == 8)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-
-        NppStStreamHandler h(stream);
-
-        NcvSize32u sz;
-        sz.width  = src.cols;
-        sz.height = src.rows;
-
-        ncvSafeCall( nppiStTranspose_64u_C1R(const_cast<Ncv64u*>(src.ptr<Ncv64u>()), static_cast<int>(src.step),
-            dst.ptr<Ncv64u>(), static_cast<int>(dst.step), sz) );
-    }
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-////////////////////////////////////////////////////////////////////////
-// flip
-
-namespace
-{
-    template<int DEPTH> struct NppTypeTraits;
-    template<> struct NppTypeTraits<CV_8U>  { typedef Npp8u npp_t; };
-    template<> struct NppTypeTraits<CV_8S>  { typedef Npp8s npp_t; };
-    template<> struct NppTypeTraits<CV_16U> { typedef Npp16u npp_t; };
-    template<> struct NppTypeTraits<CV_16S> { typedef Npp16s npp_t; };
-    template<> struct NppTypeTraits<CV_32S> { typedef Npp32s npp_t; };
-    template<> struct NppTypeTraits<CV_32F> { typedef Npp32f npp_t; };
-    template<> struct NppTypeTraits<CV_64F> { typedef Npp64f npp_t; };
-
-    template <int DEPTH> struct NppMirrorFunc
-    {
-        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
-
-        typedef NppStatus (*func_t)(const npp_t* pSrc, int nSrcStep, npp_t* pDst, int nDstStep, NppiSize oROI, NppiAxis flip);
-    };
-
-    template <int DEPTH, typename NppMirrorFunc<DEPTH>::func_t func> struct NppMirror
-    {
-        typedef typename NppMirrorFunc<DEPTH>::npp_t npp_t;
-
-        static void call(const GpuMat& src, GpuMat& dst, int flipCode, cudaStream_t stream)
-        {
-            NppStreamHandler h(stream);
-
-            NppiSize sz;
-            sz.width  = src.cols;
-            sz.height = src.rows;
-
-            nppSafeCall( func(src.ptr<npp_t>(), static_cast<int>(src.step),
-                dst.ptr<npp_t>(), static_cast<int>(dst.step), sz,
-                (flipCode == 0 ? NPP_HORIZONTAL_AXIS : (flipCode > 0 ? NPP_VERTICAL_AXIS : NPP_BOTH_AXIS))) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-void cv::gpu::flip(const GpuMat& src, GpuMat& dst, int flipCode, Stream& stream)
-{
-    typedef void (*func_t)(const GpuMat& src, GpuMat& dst, int flipCode, cudaStream_t stream);
-    static const func_t funcs[6][4] =
-    {
-        {NppMirror<CV_8U, nppiMirror_8u_C1R>::call, 0, NppMirror<CV_8U, nppiMirror_8u_C3R>::call, NppMirror<CV_8U, nppiMirror_8u_C4R>::call},
-        {0,0,0,0},
-        {NppMirror<CV_16U, nppiMirror_16u_C1R>::call, 0, NppMirror<CV_16U, nppiMirror_16u_C3R>::call, NppMirror<CV_16U, nppiMirror_16u_C4R>::call},
-        {0,0,0,0},
-        {NppMirror<CV_32S, nppiMirror_32s_C1R>::call, 0, NppMirror<CV_32S, nppiMirror_32s_C3R>::call, NppMirror<CV_32S, nppiMirror_32s_C4R>::call},
-        {NppMirror<CV_32F, nppiMirror_32f_C1R>::call, 0, NppMirror<CV_32F, nppiMirror_32f_C3R>::call, NppMirror<CV_32F, nppiMirror_32f_C4R>::call}
-    };
-
-    CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S || src.depth() == CV_32F);
-    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
-
-    dst.create(src.size(), src.type());
-
-    funcs[src.depth()][src.channels() - 1](src, dst, flipCode, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// LUT
-
-void cv::gpu::LUT(const GpuMat& src, const Mat& lut, GpuMat& dst, Stream& s)
-{
-    class LevelsInit
-    {
-    public:
-        Npp32s pLevels[256];
-        const Npp32s* pLevels3[3];
-        int nValues3[3];
-
-#if (CUDA_VERSION > 4020)
-        GpuMat d_pLevels;
-#endif
-
-        LevelsInit()
-        {
-            nValues3[0] = nValues3[1] = nValues3[2] = 256;
-            for (int i = 0; i < 256; ++i)
-                pLevels[i] = i;
-
-
-#if (CUDA_VERSION <= 4020)
-            pLevels3[0] = pLevels3[1] = pLevels3[2] = pLevels;
-#else
-            d_pLevels.upload(Mat(1, 256, CV_32S, pLevels));
-            pLevels3[0] = pLevels3[1] = pLevels3[2] = d_pLevels.ptr<Npp32s>();
-#endif
-        }
-    };
-    static LevelsInit lvls;
-
-    int cn = src.channels();
-
-    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC3);
-    CV_Assert(lut.depth() == CV_8U && (lut.channels() == 1 || lut.channels() == cn) && lut.rows * lut.cols == 256 && lut.isContinuous());
-
-    dst.create(src.size(), CV_MAKETYPE(lut.depth(), cn));
-
-    NppiSize sz;
-    sz.height = src.rows;
-    sz.width = src.cols;
-
-    Mat nppLut;
-    lut.convertTo(nppLut, CV_32S);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    NppStreamHandler h(stream);
-
-    if (src.type() == CV_8UC1)
-    {
-#if (CUDA_VERSION <= 4020)
-        nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-            dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, nppLut.ptr<Npp32s>(), lvls.pLevels, 256) );
-#else
-        GpuMat d_nppLut(Mat(1, 256, CV_32S, nppLut.data));
-        nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-            dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, d_nppLut.ptr<Npp32s>(), lvls.d_pLevels.ptr<Npp32s>(), 256) );
-#endif
-    }
-    else
-    {
-        const Npp32s* pValues3[3];
-
-        Mat nppLut3[3];
-        if (nppLut.channels() == 1)
-        {
-#if (CUDA_VERSION <= 4020)
-            pValues3[0] = pValues3[1] = pValues3[2] = nppLut.ptr<Npp32s>();
-#else
-            GpuMat d_nppLut(Mat(1, 256, CV_32S, nppLut.data));
-            pValues3[0] = pValues3[1] = pValues3[2] = d_nppLut.ptr<Npp32s>();
-#endif
-        }
-        else
-        {
-            cv::split(nppLut, nppLut3);
-
-#if (CUDA_VERSION <= 4020)
-            pValues3[0] = nppLut3[0].ptr<Npp32s>();
-            pValues3[1] = nppLut3[1].ptr<Npp32s>();
-            pValues3[2] = nppLut3[2].ptr<Npp32s>();
-#else
-            GpuMat d_nppLut0(Mat(1, 256, CV_32S, nppLut3[0].data));
-            GpuMat d_nppLut1(Mat(1, 256, CV_32S, nppLut3[1].data));
-            GpuMat d_nppLut2(Mat(1, 256, CV_32S, nppLut3[2].data));
-
-            pValues3[0] = d_nppLut0.ptr<Npp32s>();
-            pValues3[1] = d_nppLut1.ptr<Npp32s>();
-            pValues3[2] = d_nppLut2.ptr<Npp32s>();
-#endif
-        }
-
-        nppSafeCall( nppiLUT_Linear_8u_C3R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-            dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, pValues3, lvls.pLevels3, lvls.nValues3) );
-    }
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-////////////////////////////////////////////////////////////////////////
-// NPP magnitide
-
-namespace
-{
-    typedef NppStatus (*nppMagnitude_t)(const Npp32fc* pSrc, int nSrcStep, Npp32f* pDst, int nDstStep, NppiSize oSizeROI);
-
-    inline void npp_magnitude(const GpuMat& src, GpuMat& dst, nppMagnitude_t func, cudaStream_t stream)
-    {
-        CV_Assert(src.type() == CV_32FC2);
-
-        dst.create(src.size(), CV_32FC1);
-
-        NppiSize sz;
-        sz.width = src.cols;
-        sz.height = src.rows;
-
-        NppStreamHandler h(stream);
-
-        nppSafeCall( func(src.ptr<Npp32fc>(), static_cast<int>(src.step), dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz) );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-}
-
-void cv::gpu::magnitude(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    npp_magnitude(src, dst, nppiMagnitude_32fc32f_C1R, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::magnitudeSqr(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    npp_magnitude(src, dst, nppiMagnitudeSqr_32fc32f_C1R, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// Polar <-> Cart
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace mathfunc
-    {
-        void cartToPolar_gpu(PtrStepSzf x, PtrStepSzf y, PtrStepSzf mag, bool magSqr, PtrStepSzf angle, bool angleInDegrees, cudaStream_t stream);
-        void polarToCart_gpu(PtrStepSzf mag, PtrStepSzf angle, PtrStepSzf x, PtrStepSzf y, bool angleInDegrees, cudaStream_t stream);
-    }
-}}}
-
-namespace
-{
-    inline void cartToPolar_caller(const GpuMat& x, const GpuMat& y, GpuMat* mag, bool magSqr, GpuMat* angle, bool angleInDegrees, cudaStream_t stream)
-    {
-        using namespace ::cv::gpu::device::mathfunc;
-
-        CV_Assert(x.size() == y.size() && x.type() == y.type());
-        CV_Assert(x.depth() == CV_32F);
-
-        if (mag)
-            mag->create(x.size(), x.type());
-        if (angle)
-            angle->create(x.size(), x.type());
-
-        GpuMat x1cn = x.reshape(1);
-        GpuMat y1cn = y.reshape(1);
-        GpuMat mag1cn = mag ? mag->reshape(1) : GpuMat();
-        GpuMat angle1cn = angle ? angle->reshape(1) : GpuMat();
-
-        cartToPolar_gpu(x1cn, y1cn, mag1cn, magSqr, angle1cn, angleInDegrees, stream);
-    }
-
-    inline void polarToCart_caller(const GpuMat& mag, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees, cudaStream_t stream)
-    {
-        using namespace ::cv::gpu::device::mathfunc;
-
-        CV_Assert((mag.empty() || mag.size() == angle.size()) && mag.type() == angle.type());
-        CV_Assert(mag.depth() == CV_32F);
-
-        x.create(mag.size(), mag.type());
-        y.create(mag.size(), mag.type());
-
-        GpuMat mag1cn = mag.reshape(1);
-        GpuMat angle1cn = angle.reshape(1);
-        GpuMat x1cn = x.reshape(1);
-        GpuMat y1cn = y.reshape(1);
-
-        polarToCart_gpu(mag1cn, angle1cn, x1cn, y1cn, angleInDegrees, stream);
-    }
-}
-
-void cv::gpu::magnitude(const GpuMat& x, const GpuMat& y, GpuMat& dst, Stream& stream)
-{
-    cartToPolar_caller(x, y, &dst, false, 0, false, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::magnitudeSqr(const GpuMat& x, const GpuMat& y, GpuMat& dst, Stream& stream)
-{
-    cartToPolar_caller(x, y, &dst, true, 0, false, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::phase(const GpuMat& x, const GpuMat& y, GpuMat& angle, bool angleInDegrees, Stream& stream)
-{
-    cartToPolar_caller(x, y, 0, false, &angle, angleInDegrees, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::cartToPolar(const GpuMat& x, const GpuMat& y, GpuMat& mag, GpuMat& angle, bool angleInDegrees, Stream& stream)
-{
-    cartToPolar_caller(x, y, &mag, false, &angle, angleInDegrees, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees, Stream& stream)
-{
-    polarToCart_caller(magnitude, angle, x, y, angleInDegrees, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// normalize
-
-void cv::gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask)
-{
-    GpuMat norm_buf;
-    GpuMat cvt_buf;
-    normalize(src, dst, a, b, norm_type, dtype, mask, norm_buf, cvt_buf);
-}
-
-void cv::gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf)
-{
-    double scale = 1, shift = 0;
-    if (norm_type == NORM_MINMAX)
-    {
-        double smin = 0, smax = 0;
-        double dmin = std::min(a, b), dmax = std::max(a, b);
-        minMax(src, &smin, &smax, mask, norm_buf);
-        scale = (dmax - dmin) * (smax - smin > std::numeric_limits<double>::epsilon() ? 1.0 / (smax - smin) : 0.0);
-        shift = dmin - smin * scale;
-    }
-    else if (norm_type == NORM_L2 || norm_type == NORM_L1 || norm_type == NORM_INF)
-    {
-        scale = norm(src, norm_type, mask, norm_buf);
-        scale = scale > std::numeric_limits<double>::epsilon() ? a / scale : 0.0;
-        shift = 0;
-    }
-    else
-    {
-        CV_Error(CV_StsBadArg, "Unknown/unsupported norm type");
-    }
-
-    if (mask.empty())
-    {
-        src.convertTo(dst, dtype, scale, shift);
-    }
-    else
-    {
-        src.convertTo(cvt_buf, dtype, scale, shift);
-        cvt_buf.copyTo(dst, mask);
-    }
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/bgfg_gmg.cpp b/modules/gpu/src/bgfg_gmg.cpp
deleted file mode 100644
index 46f3ba2890..0000000000
--- a/modules/gpu/src/bgfg_gmg.cpp
+++ /dev/null
@@ -1,168 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-cv::gpu::GMG_GPU::GMG_GPU() { throw_nogpu(); }
-void cv::gpu::GMG_GPU::initialize(cv::Size, float, float) { throw_nogpu(); }
-void cv::gpu::GMG_GPU::operator ()(const cv::gpu::GpuMat&, cv::gpu::GpuMat&, float, cv::gpu::Stream&) { throw_nogpu(); }
-void cv::gpu::GMG_GPU::release() {}
-
-#else
-
-namespace cv { namespace gpu { namespace device {
-    namespace bgfg_gmg
-    {
-        void loadConstants(int width, int height, float minVal, float maxVal, int quantizationLevels, float backgroundPrior,
-                           float decisionThreshold, int maxFeatures, int numInitializationFrames);
-
-        template <typename SrcT>
-        void update_gpu(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures,
-                        int frameNum,  float learningRate, bool updateBackgroundModel, cudaStream_t stream);
-    }
-}}}
-
-cv::gpu::GMG_GPU::GMG_GPU()
-{
-    maxFeatures = 64;
-    learningRate = 0.025f;
-    numInitializationFrames = 120;
-    quantizationLevels = 16;
-    backgroundPrior = 0.8f;
-    decisionThreshold = 0.8f;
-    smoothingRadius = 7;
-    updateBackgroundModel = true;
-}
-
-void cv::gpu::GMG_GPU::initialize(cv::Size frameSize, float min, float max)
-{
-    using namespace cv::gpu::device::bgfg_gmg;
-
-    CV_Assert(min < max);
-    CV_Assert(maxFeatures > 0);
-    CV_Assert(learningRate >= 0.0f && learningRate <= 1.0f);
-    CV_Assert(numInitializationFrames >= 1);
-    CV_Assert(quantizationLevels >= 1 && quantizationLevels <= 255);
-    CV_Assert(backgroundPrior >= 0.0f && backgroundPrior <= 1.0f);
-
-    minVal_ = min;
-    maxVal_ = max;
-
-    frameSize_ = frameSize;
-
-    frameNum_ = 0;
-
-    nfeatures_.create(frameSize_, CV_32SC1);
-    colors_.create(maxFeatures * frameSize_.height, frameSize_.width, CV_32SC1);
-    weights_.create(maxFeatures * frameSize_.height, frameSize_.width, CV_32FC1);
-
-    nfeatures_.setTo(cv::Scalar::all(0));
-
-    if (smoothingRadius > 0)
-        boxFilter_ = cv::gpu::createBoxFilter_GPU(CV_8UC1, CV_8UC1, cv::Size(smoothingRadius, smoothingRadius));
-
-    loadConstants(frameSize_.width, frameSize_.height, minVal_, maxVal_, quantizationLevels, backgroundPrior, decisionThreshold, maxFeatures, numInitializationFrames);
-}
-
-void cv::gpu::GMG_GPU::operator ()(const cv::gpu::GpuMat& frame, cv::gpu::GpuMat& fgmask, float newLearningRate, cv::gpu::Stream& stream)
-{
-    using namespace cv::gpu::device::bgfg_gmg;
-
-    typedef void (*func_t)(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures,
-                           int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream);
-    static const func_t funcs[6][4] =
-    {
-        {update_gpu<uchar>, 0, update_gpu<uchar3>, update_gpu<uchar4>},
-        {0,0,0,0},
-        {update_gpu<ushort>, 0, update_gpu<ushort3>, update_gpu<ushort4>},
-        {0,0,0,0},
-        {0,0,0,0},
-        {update_gpu<float>, 0, update_gpu<float3>, update_gpu<float4>}
-    };
-
-    CV_Assert(frame.depth() == CV_8U || frame.depth() == CV_16U || frame.depth() == CV_32F);
-    CV_Assert(frame.channels() == 1 || frame.channels() == 3 || frame.channels() == 4);
-
-    if (newLearningRate != -1.0f)
-    {
-        CV_Assert(newLearningRate >= 0.0f && newLearningRate <= 1.0f);
-        learningRate = newLearningRate;
-    }
-
-    if (frame.size() != frameSize_)
-        initialize(frame.size(), 0.0f, frame.depth() == CV_8U ? 255.0f : frame.depth() == CV_16U ? std::numeric_limits<ushort>::max() : 1.0f);
-
-    fgmask.create(frameSize_, CV_8UC1);
-    if (stream)
-        stream.enqueueMemSet(fgmask, cv::Scalar::all(0));
-    else
-        fgmask.setTo(cv::Scalar::all(0));
-
-    funcs[frame.depth()][frame.channels() - 1](frame, fgmask, colors_, weights_, nfeatures_, frameNum_, learningRate, updateBackgroundModel, cv::gpu::StreamAccessor::getStream(stream));
-
-    // medianBlur
-    if (smoothingRadius > 0)
-    {
-        boxFilter_->apply(fgmask, buf_, cv::Rect(0,0,-1,-1), stream);
-        int minCount = (smoothingRadius * smoothingRadius + 1) / 2;
-        double thresh = 255.0 * minCount / (smoothingRadius * smoothingRadius);
-        cv::gpu::threshold(buf_, fgmask, thresh, 255.0, cv::THRESH_BINARY, stream);
-    }
-
-    // keep track of how many frames we have processed
-    ++frameNum_;
-}
-
-void cv::gpu::GMG_GPU::release()
-{
-    frameSize_ = Size();
-
-    nfeatures_.release();
-    colors_.release();
-    weights_.release();
-    boxFilter_.release();
-    buf_.release();
-}
-
-#endif
diff --git a/modules/gpu/src/bgfg_mog.cpp b/modules/gpu/src/bgfg_mog.cpp
deleted file mode 100644
index 05b4ffb9c8..0000000000
--- a/modules/gpu/src/bgfg_mog.cpp
+++ /dev/null
@@ -1,279 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-cv::gpu::MOG_GPU::MOG_GPU(int) { throw_nogpu(); }
-void cv::gpu::MOG_GPU::initialize(cv::Size, int) { throw_nogpu(); }
-void cv::gpu::MOG_GPU::operator()(const cv::gpu::GpuMat&, cv::gpu::GpuMat&, float, Stream&) { throw_nogpu(); }
-void cv::gpu::MOG_GPU::getBackgroundImage(GpuMat&, Stream&) const { throw_nogpu(); }
-void cv::gpu::MOG_GPU::release() {}
-
-cv::gpu::MOG2_GPU::MOG2_GPU(int) { throw_nogpu(); }
-void cv::gpu::MOG2_GPU::initialize(cv::Size, int) { throw_nogpu(); }
-void cv::gpu::MOG2_GPU::operator()(const GpuMat&, GpuMat&, float, Stream&) { throw_nogpu(); }
-void cv::gpu::MOG2_GPU::getBackgroundImage(GpuMat&, Stream&) const { throw_nogpu(); }
-void cv::gpu::MOG2_GPU::release() {}
-
-#else
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace mog
-    {
-        void mog_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzf weight, PtrStepSzf sortKey, PtrStepSzb mean, PtrStepSzb var,
-                     int nmixtures, float varThreshold, float learningRate, float backgroundRatio, float noiseSigma,
-                     cudaStream_t stream);
-        void getBackgroundImage_gpu(int cn, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, int nmixtures, float backgroundRatio, cudaStream_t stream);
-
-        void loadConstants(int nmixtures, float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau, unsigned char shadowVal);
-        void mog2_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean, float alphaT, float prune, bool detectShadows, cudaStream_t stream);
-        void getBackgroundImage2_gpu(int cn, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream);
-    }
-}}}
-
-namespace mog
-{
-    const int defaultNMixtures = 5;
-    const int defaultHistory = 200;
-    const float defaultBackgroundRatio = 0.7f;
-    const float defaultVarThreshold = 2.5f * 2.5f;
-    const float defaultNoiseSigma = 30.0f * 0.5f;
-    const float defaultInitialWeight = 0.05f;
-}
-
-cv::gpu::MOG_GPU::MOG_GPU(int nmixtures) :
-    frameSize_(0, 0), frameType_(0), nframes_(0)
-{
-    nmixtures_ = std::min(nmixtures > 0 ? nmixtures : mog::defaultNMixtures, 8);
-    history = mog::defaultHistory;
-    varThreshold = mog::defaultVarThreshold;
-    backgroundRatio = mog::defaultBackgroundRatio;
-    noiseSigma = mog::defaultNoiseSigma;
-}
-
-void cv::gpu::MOG_GPU::initialize(cv::Size frameSize, int frameType)
-{
-    CV_Assert(frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4);
-
-    frameSize_ = frameSize;
-    frameType_ = frameType;
-
-    int ch = CV_MAT_CN(frameType);
-    int work_ch = ch;
-
-    // for each gaussian mixture of each pixel bg model we store
-    // the mixture sort key (w/sum_of_variances), the mixture weight (w),
-    // the mean (nchannels values) and
-    // the diagonal covariance matrix (another nchannels values)
-
-    weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
-    sortKey_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
-    mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
-    var_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
-
-    weight_.setTo(cv::Scalar::all(0));
-    sortKey_.setTo(cv::Scalar::all(0));
-    mean_.setTo(cv::Scalar::all(0));
-    var_.setTo(cv::Scalar::all(0));
-
-    nframes_ = 0;
-}
-
-void cv::gpu::MOG_GPU::operator()(const cv::gpu::GpuMat& frame, cv::gpu::GpuMat& fgmask, float learningRate, Stream& stream)
-{
-    using namespace cv::gpu::device::mog;
-
-    CV_Assert(frame.depth() == CV_8U);
-
-    int ch = frame.channels();
-    int work_ch = ch;
-
-    if (nframes_ == 0 || learningRate >= 1.0 || frame.size() != frameSize_ || work_ch != mean_.channels())
-        initialize(frame.size(), frame.type());
-
-    fgmask.create(frameSize_, CV_8UC1);
-
-    ++nframes_;
-    learningRate = learningRate >= 0.0f && nframes_ > 1 ? learningRate : 1.0f / std::min(nframes_, history);
-    CV_Assert(learningRate >= 0.0f);
-
-    mog_gpu(frame, ch, fgmask, weight_, sortKey_, mean_, var_, nmixtures_,
-            varThreshold, learningRate, backgroundRatio, noiseSigma,
-            StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::MOG_GPU::getBackgroundImage(GpuMat& backgroundImage, Stream& stream) const
-{
-    using namespace cv::gpu::device::mog;
-
-    backgroundImage.create(frameSize_, frameType_);
-
-    getBackgroundImage_gpu(backgroundImage.channels(), weight_, mean_, backgroundImage, nmixtures_, backgroundRatio, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::MOG_GPU::release()
-{
-    frameSize_ = Size(0, 0);
-    frameType_ = 0;
-    nframes_ = 0;
-
-    weight_.release();
-    sortKey_.release();
-    mean_.release();
-    var_.release();
-}
-
-/////////////////////////////////////////////////////////////////
-// MOG2
-
-namespace mog2
-{
-    // default parameters of gaussian background detection algorithm
-    const int defaultHistory = 500; // Learning rate; alpha = 1/defaultHistory2
-    const float defaultVarThreshold = 4.0f * 4.0f;
-    const int defaultNMixtures = 5; // maximal number of Gaussians in mixture
-    const float defaultBackgroundRatio = 0.9f; // threshold sum of weights for background test
-    const float defaultVarThresholdGen = 3.0f * 3.0f;
-    const float defaultVarInit = 15.0f; // initial variance for new components
-    const float defaultVarMax = 5.0f * defaultVarInit;
-    const float defaultVarMin = 4.0f;
-
-    // additional parameters
-    const float defaultfCT = 0.05f; // complexity reduction prior constant 0 - no reduction of number of components
-    const unsigned char defaultnShadowDetection = 127; // value to use in the segmentation mask for shadows, set 0 not to do shadow detection
-    const float defaultfTau = 0.5f; // Tau - shadow threshold, see the paper for explanation
-}
-
-cv::gpu::MOG2_GPU::MOG2_GPU(int nmixtures) :
-    frameSize_(0, 0), frameType_(0), nframes_(0)
-{
-    nmixtures_ = nmixtures > 0 ? nmixtures : mog2::defaultNMixtures;
-
-    history = mog2::defaultHistory;
-    varThreshold = mog2::defaultVarThreshold;
-    bShadowDetection = true;
-
-    backgroundRatio = mog2::defaultBackgroundRatio;
-    fVarInit = mog2::defaultVarInit;
-    fVarMax  = mog2::defaultVarMax;
-    fVarMin = mog2::defaultVarMin;
-
-    varThresholdGen = mog2::defaultVarThresholdGen;
-    fCT = mog2::defaultfCT;
-    nShadowDetection =  mog2::defaultnShadowDetection;
-    fTau = mog2::defaultfTau;
-}
-
-void cv::gpu::MOG2_GPU::initialize(cv::Size frameSize, int frameType)
-{
-    using namespace cv::gpu::device::mog;
-
-    CV_Assert(frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4);
-
-    frameSize_ = frameSize;
-    frameType_ = frameType;
-    nframes_ = 0;
-
-    int ch = CV_MAT_CN(frameType);
-    int work_ch = ch;
-
-    // for each gaussian mixture of each pixel bg model we store ...
-    // the mixture weight (w),
-    // the mean (nchannels values) and
-    // the covariance
-    weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
-    variance_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
-    mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
-
-    //make the array for keeping track of the used modes per pixel - all zeros at start
-    bgmodelUsedModes_.create(frameSize_, CV_8UC1);
-    bgmodelUsedModes_.setTo(cv::Scalar::all(0));
-
-    loadConstants(nmixtures_, varThreshold, backgroundRatio, varThresholdGen, fVarInit, fVarMin, fVarMax, fTau, nShadowDetection);
-}
-
-void cv::gpu::MOG2_GPU::operator()(const GpuMat& frame, GpuMat& fgmask, float learningRate, Stream& stream)
-{
-    using namespace cv::gpu::device::mog;
-
-    int ch = frame.channels();
-    int work_ch = ch;
-
-    if (nframes_ == 0 || learningRate >= 1.0f || frame.size() != frameSize_ || work_ch != mean_.channels())
-        initialize(frame.size(), frame.type());
-
-    fgmask.create(frameSize_, CV_8UC1);
-    fgmask.setTo(cv::Scalar::all(0));
-
-    ++nframes_;
-    learningRate = learningRate >= 0.0f && nframes_ > 1 ? learningRate : 1.0f / std::min(2 * nframes_, history);
-    CV_Assert(learningRate >= 0.0f);
-
-    mog2_gpu(frame, frame.channels(), fgmask, bgmodelUsedModes_, weight_, variance_, mean_, learningRate, -learningRate * fCT, bShadowDetection, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::MOG2_GPU::getBackgroundImage(GpuMat& backgroundImage, Stream& stream) const
-{
-    using namespace cv::gpu::device::mog;
-
-    backgroundImage.create(frameSize_, frameType_);
-
-    getBackgroundImage2_gpu(backgroundImage.channels(), bgmodelUsedModes_, weight_, mean_, backgroundImage, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::MOG2_GPU::release()
-{
-    frameSize_ = Size(0, 0);
-    frameType_ = 0;
-    nframes_ = 0;
-
-    weight_.release();
-    variance_.release();
-    mean_.release();
-
-    bgmodelUsedModes_.release();
-}
-
-#endif
diff --git a/modules/gpu/src/bgfg_vibe.cpp b/modules/gpu/src/bgfg_vibe.cpp
deleted file mode 100644
index ff7e58082e..0000000000
--- a/modules/gpu/src/bgfg_vibe.cpp
+++ /dev/null
@@ -1,137 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-cv::gpu::VIBE_GPU::VIBE_GPU(unsigned long) { throw_nogpu(); }
-void cv::gpu::VIBE_GPU::initialize(const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::VIBE_GPU::operator()(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::VIBE_GPU::release() {}
-
-#else
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace vibe
-    {
-        void loadConstants(int nbSamples, int reqMatches, int radius, int subsamplingFactor);
-
-        void init_gpu(PtrStepSzb frame, int cn, PtrStepSzb samples, PtrStepSz<unsigned int> randStates, cudaStream_t stream);
-
-        void update_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb samples, PtrStepSz<unsigned int> randStates, cudaStream_t stream);
-    }
-}}}
-
-namespace
-{
-    const int defaultNbSamples = 20;
-    const int defaultReqMatches = 2;
-    const int defaultRadius = 20;
-    const int defaultSubsamplingFactor = 16;
-}
-
-cv::gpu::VIBE_GPU::VIBE_GPU(unsigned long rngSeed) :
-    frameSize_(0, 0), rngSeed_(rngSeed)
-{
-    nbSamples = defaultNbSamples;
-    reqMatches = defaultReqMatches;
-    radius = defaultRadius;
-    subsamplingFactor = defaultSubsamplingFactor;
-}
-
-void cv::gpu::VIBE_GPU::initialize(const GpuMat& firstFrame, Stream& s)
-{
-    using namespace cv::gpu::device::vibe;
-
-    CV_Assert(firstFrame.type() == CV_8UC1 || firstFrame.type() == CV_8UC3 || firstFrame.type() == CV_8UC4);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    loadConstants(nbSamples, reqMatches, radius, subsamplingFactor);
-
-    frameSize_ = firstFrame.size();
-
-    if (randStates_.size() != frameSize_)
-    {
-        cv::RNG rng(rngSeed_);
-        cv::Mat h_randStates(frameSize_, CV_8UC4);
-        rng.fill(h_randStates, cv::RNG::UNIFORM, 0, 255);
-        randStates_.upload(h_randStates);
-    }
-
-    int ch = firstFrame.channels();
-    int sample_ch = ch == 1 ? 1 : 4;
-
-    samples_.create(nbSamples * frameSize_.height, frameSize_.width, CV_8UC(sample_ch));
-
-    init_gpu(firstFrame, ch, samples_, randStates_, stream);
-}
-
-void cv::gpu::VIBE_GPU::operator()(const GpuMat& frame, GpuMat& fgmask, Stream& s)
-{
-    using namespace cv::gpu::device::vibe;
-
-    CV_Assert(frame.depth() == CV_8U);
-
-    int ch = frame.channels();
-    int sample_ch = ch == 1 ? 1 : 4;
-
-    if (frame.size() != frameSize_ || sample_ch != samples_.channels())
-        initialize(frame);
-
-    fgmask.create(frameSize_, CV_8UC1);
-
-    update_gpu(frame, ch, fgmask, samples_, randStates_, StreamAccessor::getStream(s));
-}
-
-void cv::gpu::VIBE_GPU::release()
-{
-    frameSize_ = Size(0, 0);
-
-    randStates_.release();
-
-    samples_.release();
-}
-
-#endif
diff --git a/modules/gpu/src/bilateral_filter.cpp b/modules/gpu/src/bilateral_filter.cpp
deleted file mode 100644
index 6aa0f525d3..0000000000
--- a/modules/gpu/src/bilateral_filter.cpp
+++ /dev/null
@@ -1,157 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-cv::gpu::DisparityBilateralFilter::DisparityBilateralFilter(int, int, int) { throw_nogpu(); }
-cv::gpu::DisparityBilateralFilter::DisparityBilateralFilter(int, int, int, float, float, float) { throw_nogpu(); }
-
-void cv::gpu::DisparityBilateralFilter::operator()(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace disp_bilateral_filter
-    {
-        void disp_load_constants(float* table_color, PtrStepSzf table_space, int ndisp, int radius, short edge_disc, short max_disc);
-
-        template<typename T>
-        void disp_bilateral_filter(PtrStepSz<T> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream);
-    }
-}}}
-
-using namespace ::cv::gpu::device::disp_bilateral_filter;
-
-namespace
-{
-    const float DEFAULT_EDGE_THRESHOLD = 0.1f;
-    const float DEFAULT_MAX_DISC_THRESHOLD = 0.2f;
-    const float DEFAULT_SIGMA_RANGE = 10.0f;
-
-    inline void calc_color_weighted_table(GpuMat& table_color, float sigma_range, int len)
-    {
-        Mat cpu_table_color(1, len, CV_32F);
-
-        float* line = cpu_table_color.ptr<float>();
-
-        for(int i = 0; i < len; i++)
-            line[i] = static_cast<float>(std::exp(-double(i * i) / (2 * sigma_range * sigma_range)));
-
-        table_color.upload(cpu_table_color);
-    }
-
-    inline void calc_space_weighted_filter(GpuMat& table_space, int win_size, float dist_space)
-    {
-        int half = (win_size >> 1);
-
-        Mat cpu_table_space(half + 1, half + 1, CV_32F);
-
-        for (int y = 0; y <= half; ++y)
-        {
-            float* row = cpu_table_space.ptr<float>(y);
-            for (int x = 0; x <= half; ++x)
-                row[x] = exp(-sqrt(float(y * y) + float(x * x)) / dist_space);
-        }
-
-        table_space.upload(cpu_table_space);
-    }
-
-    template <typename T>
-    void disp_bilateral_filter_operator(int ndisp, int radius, int iters, float edge_threshold,float max_disc_threshold,
-                                   GpuMat& table_color, GpuMat& table_space,
-                                   const GpuMat& disp, const GpuMat& img, GpuMat& dst, Stream& stream)
-    {
-        short edge_disc = std::max<short>(short(1), short(ndisp * edge_threshold + 0.5));
-        short max_disc = short(ndisp * max_disc_threshold + 0.5);
-
-        disp_load_constants(table_color.ptr<float>(), table_space, ndisp, radius, edge_disc, max_disc);
-
-        if (&dst != &disp)
-        {
-            if (stream)
-                stream.enqueueCopy(disp, dst);
-            else
-                disp.copyTo(dst);
-        }
-
-        disp_bilateral_filter<T>(dst, img, img.channels(), iters, StreamAccessor::getStream(stream));
-    }
-
-    typedef void (*bilateral_filter_operator_t)(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold,
-                                                GpuMat& table_color, GpuMat& table_space,
-                                                const GpuMat& disp, const GpuMat& img, GpuMat& dst, Stream& stream);
-
-    const bilateral_filter_operator_t operators[] =
-        {disp_bilateral_filter_operator<unsigned char>, 0, 0, disp_bilateral_filter_operator<short>, 0, 0, 0, 0};
-}
-
-cv::gpu::DisparityBilateralFilter::DisparityBilateralFilter(int ndisp_, int radius_, int iters_)
-    : ndisp(ndisp_), radius(radius_), iters(iters_), edge_threshold(DEFAULT_EDGE_THRESHOLD), max_disc_threshold(DEFAULT_MAX_DISC_THRESHOLD),
-      sigma_range(DEFAULT_SIGMA_RANGE)
-{
-    calc_color_weighted_table(table_color, sigma_range, 255);
-    calc_space_weighted_filter(table_space, radius * 2 + 1, radius + 1.0f);
-}
-
-cv::gpu::DisparityBilateralFilter::DisparityBilateralFilter(int ndisp_, int radius_, int iters_, float edge_threshold_,
-                                                     float max_disc_threshold_, float sigma_range_)
-    : ndisp(ndisp_), radius(radius_), iters(iters_), edge_threshold(edge_threshold_), max_disc_threshold(max_disc_threshold_),
-      sigma_range(sigma_range_)
-{
-    calc_color_weighted_table(table_color, sigma_range, 255);
-    calc_space_weighted_filter(table_space, radius * 2 + 1, radius + 1.0f);
-}
-
-void cv::gpu::DisparityBilateralFilter::operator()(const GpuMat& disp, const GpuMat& img, GpuMat& dst, Stream& stream)
-{
-    CV_DbgAssert(0 < ndisp && 0 < radius && 0 < iters);
-    CV_Assert(disp.rows == img.rows && disp.cols == img.cols && (disp.type() == CV_8U || disp.type() == CV_16S) && (img.type() == CV_8UC1 || img.type() == CV_8UC3));
-    operators[disp.type()](ndisp, radius, iters, edge_threshold, max_disc_threshold, table_color, table_space, disp, img, dst, stream);
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/calib3d.cpp b/modules/gpu/src/calib3d.cpp
index ab7f63f455..631174218c 100644
--- a/modules/gpu/src/calib3d.cpp
+++ b/modules/gpu/src/calib3d.cpp
@@ -47,15 +47,15 @@ using namespace cv::gpu;
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::transformPoints(const GpuMat&, const Mat&, const Mat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::transformPoints(const GpuMat&, const Mat&, const Mat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::projectPoints(const GpuMat&, const Mat&, const Mat&, const Mat&, const Mat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::projectPoints(const GpuMat&, const Mat&, const Mat&, const Mat&, const Mat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::solvePnPRansac(const Mat&, const Mat&, const Mat&, const Mat&, Mat&, Mat&, bool, int, float, int, std::vector<int>*) { throw_nogpu(); }
+void cv::gpu::solvePnPRansac(const Mat&, const Mat&, const Mat&, const Mat&, Mat&, Mat&, bool, int, float, int, std::vector<int>*) { throw_no_cuda(); }
 
 #else
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace transform_points
     {
@@ -78,7 +78,7 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-using namespace ::cv::gpu::device;
+using namespace ::cv::gpu::cudev;
 
 namespace
 {
@@ -150,7 +150,7 @@ namespace
     }
 
     // Computes rotation, translation pair for small subsets if the input data
-    class TransformHypothesesGenerator
+    class TransformHypothesesGenerator : public ParallelLoopBody
     {
     public:
         TransformHypothesesGenerator(const Mat& object_, const Mat& image_, const Mat& dist_coef_,
@@ -160,7 +160,7 @@ namespace
                   num_points(num_points_), subset_size(subset_size_), rot_matrices(rot_matrices_),
                   transl_vectors(transl_vectors_) {}
 
-        void operator()(const BlockedRange& range) const
+        void operator()(const Range& range) const
         {
             // Input data for generation of the current hypothesis
             std::vector<int> subset_indices(subset_size);
@@ -172,7 +172,7 @@ namespace
             Mat rot_mat(3, 3, CV_64F);
             Mat transl_vec(1, 3, CV_64F);
 
-            for (int iter = range.begin(); iter < range.end(); ++iter)
+            for (int iter = range.start; iter < range.end; ++iter)
             {
                 selectRandom(subset_size, num_points, subset_indices);
                 for (int i = 0; i < subset_size; ++i)
@@ -238,7 +238,7 @@ void cv::gpu::solvePnPRansac(const Mat& object, const Mat& image, const Mat& cam
     // Generate set of hypotheses using small subsets of the input data
     TransformHypothesesGenerator body(object, image_normalized, empty_dist_coef, eye_camera_mat,
                                       num_points, subset_size, rot_matrices, transl_vectors);
-    parallel_for(BlockedRange(0, num_iters), body);
+    parallel_for_(Range(0, num_iters), body);
 
     // Compute scores (i.e. number of inliers) for each hypothesis
     GpuMat d_object(object);
@@ -252,7 +252,7 @@ void cv::gpu::solvePnPRansac(const Mat& object, const Mat& image, const Mat& cam
     // Find the best hypothesis index
     Point best_idx;
     double best_score;
-    minMaxLoc(d_hypothesis_scores, NULL, &best_score, NULL, &best_idx);
+    gpu::minMaxLoc(d_hypothesis_scores, NULL, &best_score, NULL, &best_idx);
     int num_inliers = static_cast<int>(best_score);
 
     // Extract the best hypothesis data
@@ -290,5 +290,3 @@ void cv::gpu::solvePnPRansac(const Mat& object, const Mat& image, const Mat& cam
 }
 
 #endif
-
-
diff --git a/modules/gpu/src/cascadeclassifier.cpp b/modules/gpu/src/cascadeclassifier.cpp
index e82ee9d336..fc5eafbf4d 100644
--- a/modules/gpu/src/cascadeclassifier.cpp
+++ b/modules/gpu/src/cascadeclassifier.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -41,23 +41,22 @@
 //M*/
 
 #include "precomp.hpp"
-#include <vector>
-#include <iostream>
+#include "opencv2/objdetect/objdetect_c.h"
 
 using namespace cv;
 using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU()               { throw_nogpu(); }
-cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const std::string&)  { throw_nogpu(); }
-cv::gpu::CascadeClassifier_GPU::~CascadeClassifier_GPU()              { throw_nogpu(); }
-bool cv::gpu::CascadeClassifier_GPU::empty() const                    { throw_nogpu(); return true; }
-bool cv::gpu::CascadeClassifier_GPU::load(const std::string&)              { throw_nogpu(); return true; }
-Size cv::gpu::CascadeClassifier_GPU::getClassifierSize() const        { throw_nogpu(); return Size();}
-void cv::gpu::CascadeClassifier_GPU::release()                        { throw_nogpu(); }
-int cv::gpu::CascadeClassifier_GPU::detectMultiScale( const GpuMat&, GpuMat&, double, int, Size)       {throw_nogpu(); return -1;}
-int cv::gpu::CascadeClassifier_GPU::detectMultiScale( const GpuMat&, GpuMat&, Size, Size, double, int) {throw_nogpu(); return -1;}
+cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU()               { throw_no_cuda(); }
+cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const String&)  { throw_no_cuda(); }
+cv::gpu::CascadeClassifier_GPU::~CascadeClassifier_GPU()              { throw_no_cuda(); }
+bool cv::gpu::CascadeClassifier_GPU::empty() const                    { throw_no_cuda(); return true; }
+bool cv::gpu::CascadeClassifier_GPU::load(const String&)              { throw_no_cuda(); return true; }
+Size cv::gpu::CascadeClassifier_GPU::getClassifierSize() const        { throw_no_cuda(); return Size();}
+void cv::gpu::CascadeClassifier_GPU::release()                        { throw_no_cuda(); }
+int cv::gpu::CascadeClassifier_GPU::detectMultiScale( const GpuMat&, GpuMat&, double, int, Size)       {throw_no_cuda(); return -1;}
+int cv::gpu::CascadeClassifier_GPU::detectMultiScale( const GpuMat&, GpuMat&, Size, Size, double, int) {throw_no_cuda(); return -1;}
 
 #else
 
@@ -71,9 +70,40 @@ public:
                       bool findLargestObject, bool visualizeInPlace, cv::Size ncvMinSize, cv::Size maxObjectSize) = 0;
 
     virtual cv::Size getClassifierCvSize() const = 0;
-    virtual bool read(const std::string& classifierAsXml) = 0;
+    virtual bool read(const String& classifierAsXml) = 0;
 };
 
+#ifndef HAVE_OPENCV_GPULEGACY
+
+struct cv::gpu::CascadeClassifier_GPU::HaarCascade : cv::gpu::CascadeClassifier_GPU::CascadeClassifierImpl
+{
+public:
+    HaarCascade()
+    {
+        throw_no_cuda();
+    }
+
+    unsigned int process(const GpuMat&, GpuMat&, float, int, bool, bool, cv::Size, cv::Size)
+    {
+        throw_no_cuda();
+        return 0;
+    }
+
+    cv::Size getClassifierCvSize() const
+    {
+        throw_no_cuda();
+        return cv::Size();
+    }
+
+    bool read(const String&)
+    {
+        throw_no_cuda();
+        return false;
+    }
+};
+
+#else
+
 struct cv::gpu::CascadeClassifier_GPU::HaarCascade : cv::gpu::CascadeClassifier_GPU::CascadeClassifierImpl
 {
 public:
@@ -82,7 +112,7 @@ public:
         ncvSetDebugOutputHandler(NCVDebugOutputHandler);
     }
 
-    bool read(const std::string& filename)
+    bool read(const String& filename)
     {
         ncvSafeCall( load(filename) );
         return true;
@@ -169,16 +199,16 @@ public:
     cv::Size getClassifierCvSize() const { return cv::Size(haar.ClassifierSize.width, haar.ClassifierSize.height); }
 
 private:
-    static void NCVDebugOutputHandler(const std::string &msg) { CV_Error(CV_GpuApiCallError, msg.c_str()); }
+    static void NCVDebugOutputHandler(const String &msg) { CV_Error(cv::Error::GpuApiCallError, msg.c_str()); }
 
-    NCVStatus load(const std::string& classifierFile)
+    NCVStatus load(const String& classifierFile)
     {
         int devId = cv::gpu::getDevice();
         ncvAssertCUDAReturn(cudaGetDeviceProperties(&devProp, devId), NCV_CUDA_ERROR);
 
         // Load the classifier from file (assuming its size is about 1 mb) using a simple allocator
-        gpuCascadeAllocator = new NCVMemNativeAllocator(NCVMemoryTypeDevice, static_cast<int>(devProp.textureAlignment));
-        cpuCascadeAllocator = new NCVMemNativeAllocator(NCVMemoryTypeHostPinned, static_cast<int>(devProp.textureAlignment));
+        gpuCascadeAllocator = makePtr<NCVMemNativeAllocator>(NCVMemoryTypeDevice, static_cast<int>(devProp.textureAlignment));
+        cpuCascadeAllocator = makePtr<NCVMemNativeAllocator>(NCVMemoryTypeHostPinned, static_cast<int>(devProp.textureAlignment));
 
         ncvAssertPrintReturn(gpuCascadeAllocator->isInitialized(), "Error creating cascade GPU allocator", NCV_CUDA_ERROR);
         ncvAssertPrintReturn(cpuCascadeAllocator->isInitialized(), "Error creating cascade CPU allocator", NCV_CUDA_ERROR);
@@ -187,9 +217,9 @@ private:
         ncvStat = ncvHaarGetClassifierSize(classifierFile, haarNumStages, haarNumNodes, haarNumFeatures);
         ncvAssertPrintReturn(ncvStat == NCV_SUCCESS, "Error reading classifier size (check the file)", NCV_FILE_ERROR);
 
-        h_haarStages   = new NCVVectorAlloc<HaarStage64>(*cpuCascadeAllocator, haarNumStages);
-        h_haarNodes    = new NCVVectorAlloc<HaarClassifierNode128>(*cpuCascadeAllocator, haarNumNodes);
-        h_haarFeatures = new NCVVectorAlloc<HaarFeature64>(*cpuCascadeAllocator, haarNumFeatures);
+        h_haarStages.reset  (new NCVVectorAlloc<HaarStage64>(*cpuCascadeAllocator, haarNumStages));
+        h_haarNodes.reset   (new NCVVectorAlloc<HaarClassifierNode128>(*cpuCascadeAllocator, haarNumNodes));
+        h_haarFeatures.reset(new NCVVectorAlloc<HaarFeature64>(*cpuCascadeAllocator, haarNumFeatures));
 
         ncvAssertPrintReturn(h_haarStages->isMemAllocated(), "Error in cascade CPU allocator", NCV_CUDA_ERROR);
         ncvAssertPrintReturn(h_haarNodes->isMemAllocated(), "Error in cascade CPU allocator", NCV_CUDA_ERROR);
@@ -198,9 +228,9 @@ private:
         ncvStat = ncvHaarLoadFromFile_host(classifierFile, haar, *h_haarStages, *h_haarNodes, *h_haarFeatures);
         ncvAssertPrintReturn(ncvStat == NCV_SUCCESS, "Error loading classifier", NCV_FILE_ERROR);
 
-        d_haarStages   = new NCVVectorAlloc<HaarStage64>(*gpuCascadeAllocator, haarNumStages);
-        d_haarNodes    = new NCVVectorAlloc<HaarClassifierNode128>(*gpuCascadeAllocator, haarNumNodes);
-        d_haarFeatures = new NCVVectorAlloc<HaarFeature64>(*gpuCascadeAllocator, haarNumFeatures);
+        d_haarStages.reset  (new NCVVectorAlloc<HaarStage64>(*gpuCascadeAllocator, haarNumStages));
+        d_haarNodes.reset   (new NCVVectorAlloc<HaarClassifierNode128>(*gpuCascadeAllocator, haarNumNodes));
+        d_haarFeatures.reset(new NCVVectorAlloc<HaarFeature64>(*gpuCascadeAllocator, haarNumFeatures));
 
         ncvAssertPrintReturn(d_haarStages->isMemAllocated(), "Error in cascade GPU allocator", NCV_CUDA_ERROR);
         ncvAssertPrintReturn(d_haarNodes->isMemAllocated(), "Error in cascade GPU allocator", NCV_CUDA_ERROR);
@@ -249,8 +279,8 @@ private:
         ncvAssertReturnNcvStat(ncvStat);
         ncvAssertCUDAReturn(cudaStreamSynchronize(0), NCV_CUDA_ERROR);
 
-        gpuAllocator = new NCVMemStackAllocator(NCVMemoryTypeDevice, gpuCounter.maxSize(), static_cast<int>(devProp.textureAlignment));
-        cpuAllocator = new NCVMemStackAllocator(NCVMemoryTypeHostPinned, cpuCounter.maxSize(), static_cast<int>(devProp.textureAlignment));
+        gpuAllocator = makePtr<NCVMemStackAllocator>(NCVMemoryTypeDevice, gpuCounter.maxSize(), static_cast<int>(devProp.textureAlignment));
+        cpuAllocator = makePtr<NCVMemStackAllocator>(NCVMemoryTypeHostPinned, cpuCounter.maxSize(), static_cast<int>(devProp.textureAlignment));
 
         ncvAssertPrintReturn(gpuAllocator->isInitialized(), "Error creating GPU memory allocator", NCV_CUDA_ERROR);
         ncvAssertPrintReturn(cpuAllocator->isInitialized(), "Error creating CPU memory allocator", NCV_CUDA_ERROR);
@@ -283,6 +313,8 @@ private:
     virtual ~HaarCascade(){}
 };
 
+#endif
+
 cv::Size operator -(const cv::Size& a, const cv::Size& b)
 {
     return cv::Size(a.width - b.width, a.height - b.height);
@@ -340,7 +372,7 @@ struct PyrLavel
     cv::Size sWindow;
 };
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace lbp
     {
@@ -405,7 +437,7 @@ public:
         GpuMat dclassified(1, 1, CV_32S);
         cudaSafeCall( cudaMemcpy(dclassified.ptr(), &classified, sizeof(int), cudaMemcpyHostToDevice) );
 
-        PyrLavel level(0, 1.0f, image.size(), NxM, minObjectSize);
+        PyrLavel level(0, scaleFactor, image.size(), NxM, minObjectSize);
 
         while (level.isFeasible(maxObjectSize))
         {
@@ -425,8 +457,8 @@ public:
                 GpuMat buff = integralBuffer;
 
                 // generate integral for scale
-                gpu::resize(image, src, level.sFrame, 0, 0, CV_INTER_LINEAR);
-                gpu::integralBuffered(src, sint, buff);
+                gpu::resize(image, src, level.sFrame, 0, 0, cv::INTER_LINEAR);
+                gpu::integral(src, sint, buff);
 
                 // calculate job
                 int totalWidth = level.workArea.width / step;
@@ -441,7 +473,7 @@ public:
                 acc += level.sFrame.width + 1;
             }
 
-            device::lbp::classifyPyramid(image.cols, image.rows, NxM.width - 1, NxM.height - 1, iniScale, scaleFactor, total, stage_mat, stage_mat.cols / sizeof(Stage), nodes_mat,
+            cudev::lbp::classifyPyramid(image.cols, image.rows, NxM.width - 1, NxM.height - 1, iniScale, scaleFactor, total, stage_mat, stage_mat.cols / sizeof(Stage), nodes_mat,
                 leaves_mat, subsets_mat, features_mat, subsetSize, candidates, dclassified.ptr<unsigned int>(), integral);
         }
 
@@ -449,7 +481,7 @@ public:
             return 0;
 
         cudaSafeCall( cudaMemcpy(&classified, dclassified.ptr(), sizeof(int), cudaMemcpyDeviceToHost) );
-        device::lbp::connectedConmonents(candidates, classified, objects, groupThreshold, grouping_eps, dclassified.ptr<unsigned int>());
+        cudev::lbp::connectedConmonents(candidates, classified, objects, groupThreshold, grouping_eps, dclassified.ptr<unsigned int>());
 
         cudaSafeCall( cudaMemcpy(&classified, dclassified.ptr(), sizeof(int), cudaMemcpyDeviceToHost) );
         cudaSafeCall( cudaDeviceSynchronize() );
@@ -458,7 +490,7 @@ public:
 
     virtual cv::Size getClassifierCvSize() const { return NxM; }
 
-    bool read(const std::string& classifierAsXml)
+    bool read(const String& classifierAsXml)
     {
         FileStorage fs(classifierAsXml, FileStorage::READ);
         return fs.isOpened() ? read(fs.getFirstTopLevelNode()) : false;
@@ -476,6 +508,8 @@ private:
             resuzeBuffer.create(frame, CV_8UC1);
 
             integral.create(frame.height + 1, integralFactor * (frame.width + 1), CV_32SC1);
+
+#ifdef HAVE_OPENCV_GPULEGACY
             NcvSize32u roiSize;
             roiSize.width = frame.width;
             roiSize.height = frame.height;
@@ -486,6 +520,7 @@ private:
             Ncv32u bufSize;
             ncvSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize, prop) );
             integralBuffer.create(1, bufSize, CV_8UC1);
+#endif
 
             candidates.create(1 , frame.width >> 1, CV_32SC4);
         }
@@ -512,10 +547,10 @@ private:
         const char *GPU_CC_FEATURES         = "features";
         const char *GPU_CC_RECT             = "rect";
 
-        std::string stageTypeStr = (std::string)root[GPU_CC_STAGE_TYPE];
+        String stageTypeStr = (String)root[GPU_CC_STAGE_TYPE];
         CV_Assert(stageTypeStr == GPU_CC_BOOST);
 
-        std::string featureTypeStr = (std::string)root[GPU_CC_FEATURE_TYPE];
+        String featureTypeStr = (String)root[GPU_CC_FEATURE_TYPE];
         CV_Assert(featureTypeStr == GPU_CC_LBP);
 
         NxM.width =  (int)root[GPU_CC_WIDTH];
@@ -662,7 +697,7 @@ private:
 cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU()
 : findLargestObject(false), visualizeInPlace(false), impl(0) {}
 
-cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const std::string& filename)
+cv::gpu::CascadeClassifier_GPU::CascadeClassifier_GPU(const String& filename)
 : findLargestObject(false), visualizeInPlace(false), impl(0) { load(filename); }
 
 cv::gpu::CascadeClassifier_GPU::~CascadeClassifier_GPU() { release(); }
@@ -688,12 +723,12 @@ int cv::gpu::CascadeClassifier_GPU::detectMultiScale(const GpuMat& image, GpuMat
     return impl->process(image, objectsBuf, (float)scaleFactor, minNeighbors, findLargestObject, visualizeInPlace, minSize, maxObjectSize);
 }
 
-bool cv::gpu::CascadeClassifier_GPU::load(const std::string& filename)
+bool cv::gpu::CascadeClassifier_GPU::load(const String& filename)
 {
     release();
 
-    std::string fext = filename.substr(filename.find_last_of(".") + 1);
-    std::transform(fext.begin(), fext.end(), fext.begin(), ::tolower);
+    String fext = filename.substr(filename.find_last_of(".") + 1);
+    fext = fext.toLowerCase();
 
     if (fext == "nvbin")
     {
@@ -710,7 +745,7 @@ bool cv::gpu::CascadeClassifier_GPU::load(const std::string& filename)
     }
 
     const char *GPU_CC_LBP = "LBP";
-    std::string featureTypeStr = (std::string)fs.getFirstTopLevelNode()["featureType"];
+    String featureTypeStr = (String)fs.getFirstTopLevelNode()["featureType"];
     if (featureTypeStr == GPU_CC_LBP)
         impl = new LbpCascade();
     else
@@ -721,240 +756,3 @@ bool cv::gpu::CascadeClassifier_GPU::load(const std::string& filename)
 }
 
 #endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-
-#if defined (HAVE_CUDA)
-
-struct RectConvert
-{
-    Rect operator()(const NcvRect32u& nr) const { return Rect(nr.x, nr.y, nr.width, nr.height); }
-    NcvRect32u operator()(const Rect& nr) const
-    {
-        NcvRect32u rect;
-        rect.x = nr.x;
-        rect.y = nr.y;
-        rect.width = nr.width;
-        rect.height = nr.height;
-        return rect;
-    }
-};
-
-void groupRectangles(std::vector<NcvRect32u> &hypotheses, int groupThreshold, double eps, std::vector<Ncv32u> *weights)
-{
-    std::vector<Rect> rects(hypotheses.size());
-    std::transform(hypotheses.begin(), hypotheses.end(), rects.begin(), RectConvert());
-
-    if (weights)
-    {
-        std::vector<int> weights_int;
-        weights_int.assign(weights->begin(), weights->end());
-        cv::groupRectangles(rects, weights_int, groupThreshold, eps);
-    }
-    else
-    {
-        cv::groupRectangles(rects, groupThreshold, eps);
-    }
-    std::transform(rects.begin(), rects.end(), hypotheses.begin(), RectConvert());
-    hypotheses.resize(rects.size());
-}
-
-NCVStatus loadFromXML(const std::string &filename,
-                      HaarClassifierCascadeDescriptor &haar,
-                      std::vector<HaarStage64> &haarStages,
-                      std::vector<HaarClassifierNode128> &haarClassifierNodes,
-                      std::vector<HaarFeature64> &haarFeatures)
-{
-    NCVStatus ncvStat;
-
-    haar.NumStages = 0;
-    haar.NumClassifierRootNodes = 0;
-    haar.NumClassifierTotalNodes = 0;
-    haar.NumFeatures = 0;
-    haar.ClassifierSize.width = 0;
-    haar.ClassifierSize.height = 0;
-    haar.bHasStumpsOnly = true;
-    haar.bNeedsTiltedII = false;
-    Ncv32u curMaxTreeDepth;
-
-    std::vector<char> xmlFileCont;
-
-    std::vector<HaarClassifierNode128> h_TmpClassifierNotRootNodes;
-    haarStages.resize(0);
-    haarClassifierNodes.resize(0);
-    haarFeatures.resize(0);
-
-    Ptr<CvHaarClassifierCascade> oldCascade = (CvHaarClassifierCascade*)cvLoad(filename.c_str(), 0, 0, 0);
-    if (oldCascade.empty())
-    {
-        return NCV_HAAR_XML_LOADING_EXCEPTION;
-    }
-
-    haar.ClassifierSize.width = oldCascade->orig_window_size.width;
-    haar.ClassifierSize.height = oldCascade->orig_window_size.height;
-
-    int stagesCound = oldCascade->count;
-    for(int s = 0; s < stagesCound; ++s) // by stages
-    {
-        HaarStage64 curStage;
-        curStage.setStartClassifierRootNodeOffset(static_cast<Ncv32u>(haarClassifierNodes.size()));
-
-        curStage.setStageThreshold(oldCascade->stage_classifier[s].threshold);
-
-        int treesCount = oldCascade->stage_classifier[s].count;
-        for(int t = 0; t < treesCount; ++t) // by trees
-        {
-            Ncv32u nodeId = 0;
-            CvHaarClassifier* tree = &oldCascade->stage_classifier[s].classifier[t];
-
-            int nodesCount = tree->count;
-            for(int n = 0; n < nodesCount; ++n)  //by features
-            {
-                CvHaarFeature* feature = &tree->haar_feature[n];
-
-                HaarClassifierNode128 curNode;
-                curNode.setThreshold(tree->threshold[n]);
-
-                NcvBool bIsLeftNodeLeaf = false;
-                NcvBool bIsRightNodeLeaf = false;
-
-                HaarClassifierNodeDescriptor32 nodeLeft;
-                if ( tree->left[n] <= 0 )
-                {
-                    Ncv32f leftVal = tree->alpha[-tree->left[n]];
-                    ncvStat = nodeLeft.create(leftVal);
-                    ncvAssertReturn(ncvStat == NCV_SUCCESS, ncvStat);
-                    bIsLeftNodeLeaf = true;
-                }
-                else
-                {
-                    Ncv32u leftNodeOffset = tree->left[n];
-                    nodeLeft.create((Ncv32u)(h_TmpClassifierNotRootNodes.size() + leftNodeOffset - 1));
-                    haar.bHasStumpsOnly = false;
-                }
-                curNode.setLeftNodeDesc(nodeLeft);
-
-                HaarClassifierNodeDescriptor32 nodeRight;
-                if ( tree->right[n] <= 0 )
-                {
-                    Ncv32f rightVal = tree->alpha[-tree->right[n]];
-                    ncvStat = nodeRight.create(rightVal);
-                    ncvAssertReturn(ncvStat == NCV_SUCCESS, ncvStat);
-                    bIsRightNodeLeaf = true;
-                }
-                else
-                {
-                    Ncv32u rightNodeOffset = tree->right[n];
-                    nodeRight.create((Ncv32u)(h_TmpClassifierNotRootNodes.size() + rightNodeOffset - 1));
-                    haar.bHasStumpsOnly = false;
-                }
-                curNode.setRightNodeDesc(nodeRight);
-
-                Ncv32u tiltedVal = feature->tilted;
-                haar.bNeedsTiltedII = (tiltedVal != 0);
-
-                Ncv32u featureId = 0;
-                for(int l = 0; l < CV_HAAR_FEATURE_MAX; ++l) //by rects
-                {
-                    Ncv32u rectX = feature->rect[l].r.x;
-                    Ncv32u rectY = feature->rect[l].r.y;
-                    Ncv32u rectWidth = feature->rect[l].r.width;
-                    Ncv32u rectHeight = feature->rect[l].r.height;
-
-                    Ncv32f rectWeight = feature->rect[l].weight;
-
-                    if (rectWeight == 0/* && rectX == 0 &&rectY == 0 && rectWidth == 0 && rectHeight == 0*/)
-                        break;
-
-                    HaarFeature64 curFeature;
-                    ncvStat = curFeature.setRect(rectX, rectY, rectWidth, rectHeight, haar.ClassifierSize.width, haar.ClassifierSize.height);
-                    curFeature.setWeight(rectWeight);
-                    ncvAssertReturn(NCV_SUCCESS == ncvStat, ncvStat);
-                    haarFeatures.push_back(curFeature);
-
-                    featureId++;
-                }
-
-                HaarFeatureDescriptor32 tmpFeatureDesc;
-                ncvStat = tmpFeatureDesc.create(haar.bNeedsTiltedII, bIsLeftNodeLeaf, bIsRightNodeLeaf,
-                    featureId, static_cast<Ncv32u>(haarFeatures.size()) - featureId);
-                ncvAssertReturn(NCV_SUCCESS == ncvStat, ncvStat);
-                curNode.setFeatureDesc(tmpFeatureDesc);
-
-                if (!nodeId)
-                {
-                    //root node
-                    haarClassifierNodes.push_back(curNode);
-                    curMaxTreeDepth = 1;
-                }
-                else
-                {
-                    //other node
-                    h_TmpClassifierNotRootNodes.push_back(curNode);
-                    curMaxTreeDepth++;
-                }
-
-                nodeId++;
-            }
-        }
-
-        curStage.setNumClassifierRootNodes(treesCount);
-        haarStages.push_back(curStage);
-    }
-
-    //fill in cascade stats
-    haar.NumStages = static_cast<Ncv32u>(haarStages.size());
-    haar.NumClassifierRootNodes = static_cast<Ncv32u>(haarClassifierNodes.size());
-    haar.NumClassifierTotalNodes = static_cast<Ncv32u>(haar.NumClassifierRootNodes + h_TmpClassifierNotRootNodes.size());
-    haar.NumFeatures = static_cast<Ncv32u>(haarFeatures.size());
-
-    //merge root and leaf nodes in one classifiers array
-    Ncv32u offsetRoot = static_cast<Ncv32u>(haarClassifierNodes.size());
-    for (Ncv32u i=0; i<haarClassifierNodes.size(); i++)
-    {
-        HaarFeatureDescriptor32 featureDesc = haarClassifierNodes[i].getFeatureDesc();
-
-        HaarClassifierNodeDescriptor32 nodeLeft = haarClassifierNodes[i].getLeftNodeDesc();
-        if (!featureDesc.isLeftNodeLeaf())
-        {
-            Ncv32u newOffset = nodeLeft.getNextNodeOffset() + offsetRoot;
-            nodeLeft.create(newOffset);
-        }
-        haarClassifierNodes[i].setLeftNodeDesc(nodeLeft);
-
-        HaarClassifierNodeDescriptor32 nodeRight = haarClassifierNodes[i].getRightNodeDesc();
-        if (!featureDesc.isRightNodeLeaf())
-        {
-            Ncv32u newOffset = nodeRight.getNextNodeOffset() + offsetRoot;
-            nodeRight.create(newOffset);
-        }
-        haarClassifierNodes[i].setRightNodeDesc(nodeRight);
-    }
-
-    for (Ncv32u i=0; i<h_TmpClassifierNotRootNodes.size(); i++)
-    {
-        HaarFeatureDescriptor32 featureDesc = h_TmpClassifierNotRootNodes[i].getFeatureDesc();
-
-        HaarClassifierNodeDescriptor32 nodeLeft = h_TmpClassifierNotRootNodes[i].getLeftNodeDesc();
-        if (!featureDesc.isLeftNodeLeaf())
-        {
-            Ncv32u newOffset = nodeLeft.getNextNodeOffset() + offsetRoot;
-            nodeLeft.create(newOffset);
-        }
-        h_TmpClassifierNotRootNodes[i].setLeftNodeDesc(nodeLeft);
-
-        HaarClassifierNodeDescriptor32 nodeRight = h_TmpClassifierNotRootNodes[i].getRightNodeDesc();
-        if (!featureDesc.isRightNodeLeaf())
-        {
-            Ncv32u newOffset = nodeRight.getNextNodeOffset() + offsetRoot;
-            nodeRight.create(newOffset);
-        }
-        h_TmpClassifierNotRootNodes[i].setRightNodeDesc(nodeRight);
-
-        haarClassifierNodes.push_back(h_TmpClassifierNotRootNodes[i]);
-    }
-
-    return NCV_SUCCESS;
-}
-
-#endif /* HAVE_CUDA */
diff --git a/modules/gpu/src/color.cpp b/modules/gpu/src/color.cpp
deleted file mode 100644
index 05d4609001..0000000000
--- a/modules/gpu/src/color.cpp
+++ /dev/null
@@ -1,1887 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::cvtColor(const GpuMat&, GpuMat&, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::swapChannels(GpuMat&, const int[], Stream&) { throw_nogpu(); }
-void cv::gpu::gammaCorrection(const GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-#include "cvt_color_internal.h"
-
-namespace cv { namespace gpu {
-    namespace device
-    {
-        template <int cn>
-        void Bayer2BGR_8u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-        template <int cn>
-        void Bayer2BGR_16u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-    }
-}}
-
-using namespace ::cv::gpu::device;
-
-namespace
-{
-    typedef void (*gpu_func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
-    void bgr_to_rgb(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgr_to_rgb_8u, 0, bgr_to_rgb_16u, 0, 0, bgr_to_rgb_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 3));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_bgra(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgr_to_bgra_8u, 0, bgr_to_bgra_16u, 0, 0, bgr_to_bgra_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 4));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_rgba(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgr_to_rgba_8u, 0, bgr_to_rgba_16u, 0, 0, bgr_to_rgba_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 4));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_bgr(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgra_to_bgr_8u, 0, bgra_to_bgr_16u, 0, 0, bgra_to_bgr_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 3));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_rgb(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgra_to_rgb_8u, 0, bgra_to_rgb_16u, 0, 0, bgra_to_rgb_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 3));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_rgba(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgra_to_rgba_8u, 0, bgra_to_rgba_16u, 0, 0, bgra_to_rgba_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 4));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_bgr555(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::bgr_to_bgr555(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_bgr565(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::bgr_to_bgr565(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_bgr555(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::rgb_to_bgr555(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_bgr565(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::rgb_to_bgr565(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_bgr555(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::bgra_to_bgr555(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_bgr565(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::bgra_to_bgr565(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgba_to_bgr555(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::rgba_to_bgr555(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgba_to_bgr565(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::rgba_to_bgr565(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr555_to_rgb(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC3);
-
-        device::bgr555_to_rgb(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr565_to_rgb(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC3);
-
-        device::bgr565_to_rgb(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr555_to_bgr(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC3);
-
-        device::bgr555_to_bgr(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr565_to_bgr(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC3);
-
-        device::bgr565_to_bgr(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr555_to_rgba(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC4);
-
-        device::bgr555_to_rgba(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr565_to_rgba(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC4);
-
-        device::bgr565_to_rgba(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr555_to_bgra(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC4);
-
-        device::bgr555_to_bgra(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr565_to_bgra(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC4);
-
-        device::bgr565_to_bgra(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void gray_to_bgr(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {gray_to_bgr_8u, 0, gray_to_bgr_16u, 0, 0, gray_to_bgr_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 1);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 3));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void gray_to_bgra(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {gray_to_bgra_8u, 0, gray_to_bgra_16u, 0, 0, gray_to_bgra_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 1);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 4));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void gray_to_bgr555(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 1);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::gray_to_bgr555(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void gray_to_bgr565(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 1);
-
-        dst.create(src.size(), CV_8UC2);
-
-        device::gray_to_bgr565(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr555_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC1);
-
-        device::bgr555_to_gray(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr565_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        CV_Assert(src.depth() == CV_8U);
-        CV_Assert(src.channels() == 2);
-
-        dst.create(src.size(), CV_8UC1);
-
-        device::bgr565_to_gray(src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {rgb_to_gray_8u, 0, rgb_to_gray_16u, 0, 0, rgb_to_gray_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 1));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgr_to_gray_8u, 0, bgr_to_gray_16u, 0, 0, bgr_to_gray_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 1));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgba_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {rgba_to_gray_8u, 0, rgba_to_gray_16u, 0, 0, rgba_to_gray_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 1));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgra_to_gray(const GpuMat& src, GpuMat& dst, int, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[] = {bgra_to_gray_8u, 0, bgra_to_gray_16u, 0, 0, bgra_to_gray_32f};
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), 1));
-
-        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_yuv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_yuv_8u, 0, rgb_to_yuv_16u, 0, 0, rgb_to_yuv_32f},
-                {rgba_to_yuv_8u, 0, rgba_to_yuv_16u, 0, 0, rgba_to_yuv_32f}
-            },
-            {
-                {rgb_to_yuv4_8u, 0, rgb_to_yuv4_16u, 0, 0, rgb_to_yuv4_32f},
-                {rgba_to_yuv4_8u, 0, rgba_to_yuv4_16u, 0, 0, rgba_to_yuv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_yuv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_yuv_8u, 0, bgr_to_yuv_16u, 0, 0, bgr_to_yuv_32f},
-                {bgra_to_yuv_8u, 0, bgra_to_yuv_16u, 0, 0, bgra_to_yuv_32f}
-            },
-            {
-                {bgr_to_yuv4_8u, 0, bgr_to_yuv4_16u, 0, 0, bgr_to_yuv4_32f},
-                {bgra_to_yuv4_8u, 0, bgra_to_yuv4_16u, 0, 0, bgra_to_yuv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void yuv_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {yuv_to_rgb_8u, 0, yuv_to_rgb_16u, 0, 0, yuv_to_rgb_32f},
-                {yuv4_to_rgb_8u, 0, yuv4_to_rgb_16u, 0, 0, yuv4_to_rgb_32f}
-            },
-            {
-                {yuv_to_rgba_8u, 0, yuv_to_rgba_16u, 0, 0, yuv_to_rgba_32f},
-                {yuv4_to_rgba_8u, 0, yuv4_to_rgba_16u, 0, 0, yuv4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void yuv_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {yuv_to_bgr_8u, 0, yuv_to_bgr_16u, 0, 0, yuv_to_bgr_32f},
-                {yuv4_to_bgr_8u, 0, yuv4_to_bgr_16u, 0, 0, yuv4_to_bgr_32f}
-            },
-            {
-                {yuv_to_bgra_8u, 0, yuv_to_bgra_16u, 0, 0, yuv_to_bgra_32f},
-                {yuv4_to_bgra_8u, 0, yuv4_to_bgra_16u, 0, 0, yuv4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_YCrCb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_YCrCb_8u, 0, rgb_to_YCrCb_16u, 0, 0, rgb_to_YCrCb_32f},
-                {rgba_to_YCrCb_8u, 0, rgba_to_YCrCb_16u, 0, 0, rgba_to_YCrCb_32f}
-            },
-            {
-                {rgb_to_YCrCb4_8u, 0, rgb_to_YCrCb4_16u, 0, 0, rgb_to_YCrCb4_32f},
-                {rgba_to_YCrCb4_8u, 0, rgba_to_YCrCb4_16u, 0, 0, rgba_to_YCrCb4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_YCrCb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_YCrCb_8u, 0, bgr_to_YCrCb_16u, 0, 0, bgr_to_YCrCb_32f},
-                {bgra_to_YCrCb_8u, 0, bgra_to_YCrCb_16u, 0, 0, bgra_to_YCrCb_32f}
-            },
-            {
-                {bgr_to_YCrCb4_8u, 0, bgr_to_YCrCb4_16u, 0, 0, bgr_to_YCrCb4_32f},
-                {bgra_to_YCrCb4_8u, 0, bgra_to_YCrCb4_16u, 0, 0, bgra_to_YCrCb4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void YCrCb_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {YCrCb_to_rgb_8u, 0, YCrCb_to_rgb_16u, 0, 0, YCrCb_to_rgb_32f},
-                {YCrCb4_to_rgb_8u, 0, YCrCb4_to_rgb_16u, 0, 0, YCrCb4_to_rgb_32f}
-            },
-            {
-                {YCrCb_to_rgba_8u, 0, YCrCb_to_rgba_16u, 0, 0, YCrCb_to_rgba_32f},
-                {YCrCb4_to_rgba_8u, 0, YCrCb4_to_rgba_16u, 0, 0, YCrCb4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void YCrCb_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {YCrCb_to_bgr_8u, 0, YCrCb_to_bgr_16u, 0, 0, YCrCb_to_bgr_32f},
-                {YCrCb4_to_bgr_8u, 0, YCrCb4_to_bgr_16u, 0, 0, YCrCb4_to_bgr_32f}
-            },
-            {
-                {YCrCb_to_bgra_8u, 0, YCrCb_to_bgra_16u, 0, 0, YCrCb_to_bgra_32f},
-                {YCrCb4_to_bgra_8u, 0, YCrCb4_to_bgra_16u, 0, 0, YCrCb4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_xyz(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_xyz_8u, 0, rgb_to_xyz_16u, 0, 0, rgb_to_xyz_32f},
-                {rgba_to_xyz_8u, 0, rgba_to_xyz_16u, 0, 0, rgba_to_xyz_32f}
-            },
-            {
-                {rgb_to_xyz4_8u, 0, rgb_to_xyz4_16u, 0, 0, rgb_to_xyz4_32f},
-                {rgba_to_xyz4_8u, 0, rgba_to_xyz4_16u, 0, 0, rgba_to_xyz4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_xyz(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_xyz_8u, 0, bgr_to_xyz_16u, 0, 0, bgr_to_xyz_32f},
-                {bgra_to_xyz_8u, 0, bgra_to_xyz_16u, 0, 0, bgra_to_xyz_32f}
-            },
-            {
-                {bgr_to_xyz4_8u, 0, bgr_to_xyz4_16u, 0, 0, bgr_to_xyz4_32f},
-                {bgra_to_xyz4_8u, 0, bgra_to_xyz4_16u, 0, 0, bgra_to_xyz4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void xyz_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {xyz_to_rgb_8u, 0, xyz_to_rgb_16u, 0, 0, xyz_to_rgb_32f},
-                {xyz4_to_rgb_8u, 0, xyz4_to_rgb_16u, 0, 0, xyz4_to_rgb_32f}
-            },
-            {
-                {xyz_to_rgba_8u, 0, xyz_to_rgba_16u, 0, 0, xyz_to_rgba_32f},
-                {xyz4_to_rgba_8u, 0, xyz4_to_rgba_16u, 0, 0, xyz4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void xyz_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {xyz_to_bgr_8u, 0, xyz_to_bgr_16u, 0, 0, xyz_to_bgr_32f},
-                {xyz4_to_bgr_8u, 0, xyz4_to_bgr_16u, 0, 0, xyz4_to_bgr_32f}
-            },
-            {
-                {xyz_to_bgra_8u, 0, xyz_to_bgra_16u, 0, 0, xyz_to_bgra_32f},
-                {xyz4_to_bgra_8u, 0, xyz4_to_bgra_16u, 0, 0, xyz4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_hsv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_hsv_8u, 0, 0, 0, 0, rgb_to_hsv_32f},
-                {rgba_to_hsv_8u, 0, 0, 0, 0, rgba_to_hsv_32f},
-            },
-            {
-                {rgb_to_hsv4_8u, 0, 0, 0, 0, rgb_to_hsv4_32f},
-                {rgba_to_hsv4_8u, 0, 0, 0, 0, rgba_to_hsv4_32f},
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_hsv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_hsv_8u, 0, 0, 0, 0, bgr_to_hsv_32f},
-                {bgra_to_hsv_8u, 0, 0, 0, 0, bgra_to_hsv_32f}
-            },
-            {
-                {bgr_to_hsv4_8u, 0, 0, 0, 0, bgr_to_hsv4_32f},
-                {bgra_to_hsv4_8u, 0, 0, 0, 0, bgra_to_hsv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hsv_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hsv_to_rgb_8u, 0, 0, 0, 0, hsv_to_rgb_32f},
-                {hsv4_to_rgb_8u, 0, 0, 0, 0, hsv4_to_rgb_32f}
-            },
-            {
-                {hsv_to_rgba_8u, 0, 0, 0, 0, hsv_to_rgba_32f},
-                {hsv4_to_rgba_8u, 0, 0, 0, 0, hsv4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hsv_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hsv_to_bgr_8u, 0, 0, 0, 0, hsv_to_bgr_32f},
-                {hsv4_to_bgr_8u, 0, 0, 0, 0, hsv4_to_bgr_32f}
-            },
-            {
-                {hsv_to_bgra_8u, 0, 0, 0, 0, hsv_to_bgra_32f},
-                {hsv4_to_bgra_8u, 0, 0, 0, 0, hsv4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_hls(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_hls_8u, 0, 0, 0, 0, rgb_to_hls_32f},
-                {rgba_to_hls_8u, 0, 0, 0, 0, rgba_to_hls_32f},
-            },
-            {
-                {rgb_to_hls4_8u, 0, 0, 0, 0, rgb_to_hls4_32f},
-                {rgba_to_hls4_8u, 0, 0, 0, 0, rgba_to_hls4_32f},
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_hls(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_hls_8u, 0, 0, 0, 0, bgr_to_hls_32f},
-                {bgra_to_hls_8u, 0, 0, 0, 0, bgra_to_hls_32f}
-            },
-            {
-                {bgr_to_hls4_8u, 0, 0, 0, 0, bgr_to_hls4_32f},
-                {bgra_to_hls4_8u, 0, 0, 0, 0, bgra_to_hls4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hls_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hls_to_rgb_8u, 0, 0, 0, 0, hls_to_rgb_32f},
-                {hls4_to_rgb_8u, 0, 0, 0, 0, hls4_to_rgb_32f}
-            },
-            {
-                {hls_to_rgba_8u, 0, 0, 0, 0, hls_to_rgba_32f},
-                {hls4_to_rgba_8u, 0, 0, 0, 0, hls4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hls_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hls_to_bgr_8u, 0, 0, 0, 0, hls_to_bgr_32f},
-                {hls4_to_bgr_8u, 0, 0, 0, 0, hls4_to_bgr_32f}
-            },
-            {
-                {hls_to_bgra_8u, 0, 0, 0, 0, hls_to_bgra_32f},
-                {hls4_to_bgra_8u, 0, 0, 0, 0, hls4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_hsv_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_hsv_full_8u, 0, 0, 0, 0, rgb_to_hsv_full_32f},
-                {rgba_to_hsv_full_8u, 0, 0, 0, 0, rgba_to_hsv_full_32f},
-            },
-            {
-                {rgb_to_hsv4_full_8u, 0, 0, 0, 0, rgb_to_hsv4_full_32f},
-                {rgba_to_hsv4_full_8u, 0, 0, 0, 0, rgba_to_hsv4_full_32f},
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_hsv_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_hsv_full_8u, 0, 0, 0, 0, bgr_to_hsv_full_32f},
-                {bgra_to_hsv_full_8u, 0, 0, 0, 0, bgra_to_hsv_full_32f}
-            },
-            {
-                {bgr_to_hsv4_full_8u, 0, 0, 0, 0, bgr_to_hsv4_full_32f},
-                {bgra_to_hsv4_full_8u, 0, 0, 0, 0, bgra_to_hsv4_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hsv_to_rgb_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hsv_to_rgb_full_8u, 0, 0, 0, 0, hsv_to_rgb_full_32f},
-                {hsv4_to_rgb_full_8u, 0, 0, 0, 0, hsv4_to_rgb_full_32f}
-            },
-            {
-                {hsv_to_rgba_full_8u, 0, 0, 0, 0, hsv_to_rgba_full_32f},
-                {hsv4_to_rgba_full_8u, 0, 0, 0, 0, hsv4_to_rgba_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hsv_to_bgr_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hsv_to_bgr_full_8u, 0, 0, 0, 0, hsv_to_bgr_full_32f},
-                {hsv4_to_bgr_full_8u, 0, 0, 0, 0, hsv4_to_bgr_full_32f}
-            },
-            {
-                {hsv_to_bgra_full_8u, 0, 0, 0, 0, hsv_to_bgra_full_32f},
-                {hsv4_to_bgra_full_8u, 0, 0, 0, 0, hsv4_to_bgra_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_hls_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {rgb_to_hls_full_8u, 0, 0, 0, 0, rgb_to_hls_full_32f},
-                {rgba_to_hls_full_8u, 0, 0, 0, 0, rgba_to_hls_full_32f},
-            },
-            {
-                {rgb_to_hls4_full_8u, 0, 0, 0, 0, rgb_to_hls4_full_32f},
-                {rgba_to_hls4_full_8u, 0, 0, 0, 0, rgba_to_hls4_full_32f},
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_hls_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {bgr_to_hls_full_8u, 0, 0, 0, 0, bgr_to_hls_full_32f},
-                {bgra_to_hls_full_8u, 0, 0, 0, 0, bgra_to_hls_full_32f}
-            },
-            {
-                {bgr_to_hls4_full_8u, 0, 0, 0, 0, bgr_to_hls4_full_32f},
-                {bgra_to_hls4_full_8u, 0, 0, 0, 0, bgra_to_hls4_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hls_to_rgb_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hls_to_rgb_full_8u, 0, 0, 0, 0, hls_to_rgb_full_32f},
-                {hls4_to_rgb_full_8u, 0, 0, 0, 0, hls4_to_rgb_full_32f}
-            },
-            {
-                {hls_to_rgba_full_8u, 0, 0, 0, 0, hls_to_rgba_full_32f},
-                {hls4_to_rgba_full_8u, 0, 0, 0, 0, hls4_to_rgba_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void hls_to_bgr_full(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][6] =
-        {
-            {
-                {hls_to_bgr_full_8u, 0, 0, 0, 0, hls_to_bgr_full_32f},
-                {hls4_to_bgr_full_8u, 0, 0, 0, 0, hls4_to_bgr_full_32f}
-            },
-            {
-                {hls_to_bgra_full_8u, 0, 0, 0, 0, hls_to_bgra_full_32f},
-                {hls4_to_bgra_full_8u, 0, 0, 0, 0, hls4_to_bgra_full_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_lab(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {bgr_to_lab_8u, bgr_to_lab_32f},
-                {bgra_to_lab_8u, bgra_to_lab_32f}
-            },
-            {
-                {bgr_to_lab4_8u, bgr_to_lab4_32f},
-                {bgra_to_lab4_8u, bgra_to_lab4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_lab(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {rgb_to_lab_8u, rgb_to_lab_32f},
-                {rgba_to_lab_8u, rgba_to_lab_32f}
-            },
-            {
-                {rgb_to_lab4_8u, rgb_to_lab4_32f},
-                {rgba_to_lab4_8u, rgba_to_lab4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lbgr_to_lab(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lbgr_to_lab_8u, lbgr_to_lab_32f},
-                {lbgra_to_lab_8u, lbgra_to_lab_32f}
-            },
-            {
-                {lbgr_to_lab4_8u, lbgr_to_lab4_32f},
-                {lbgra_to_lab4_8u, lbgra_to_lab4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lrgb_to_lab(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lrgb_to_lab_8u, lrgb_to_lab_32f},
-                {lrgba_to_lab_8u, lrgba_to_lab_32f}
-            },
-            {
-                {lrgb_to_lab4_8u, lrgb_to_lab4_32f},
-                {lrgba_to_lab4_8u, lrgba_to_lab4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lab_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lab_to_bgr_8u, lab_to_bgr_32f},
-                {lab4_to_bgr_8u, lab4_to_bgr_32f}
-            },
-            {
-                {lab_to_bgra_8u, lab_to_bgra_32f},
-                {lab4_to_bgra_8u, lab4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lab_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lab_to_rgb_8u, lab_to_rgb_32f},
-                {lab4_to_rgb_8u, lab4_to_rgb_32f}
-            },
-            {
-                {lab_to_rgba_8u, lab_to_rgba_32f},
-                {lab4_to_rgba_8u, lab4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lab_to_lbgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lab_to_lbgr_8u, lab_to_lbgr_32f},
-                {lab4_to_lbgr_8u, lab4_to_lbgr_32f}
-            },
-            {
-                {lab_to_lbgra_8u, lab_to_lbgra_32f},
-                {lab4_to_lbgra_8u, lab4_to_lbgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lab_to_lrgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lab_to_lrgb_8u, lab_to_lrgb_32f},
-                {lab4_to_lrgb_8u, lab4_to_lrgb_32f}
-            },
-            {
-                {lab_to_lrgba_8u, lab_to_lrgba_32f},
-                {lab4_to_lrgba_8u, lab4_to_lrgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void bgr_to_luv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {bgr_to_luv_8u, bgr_to_luv_32f},
-                {bgra_to_luv_8u, bgra_to_luv_32f}
-            },
-            {
-                {bgr_to_luv4_8u, bgr_to_luv4_32f},
-                {bgra_to_luv4_8u, bgra_to_luv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgb_to_luv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {rgb_to_luv_8u, rgb_to_luv_32f},
-                {rgba_to_luv_8u, rgba_to_luv_32f}
-            },
-            {
-                {rgb_to_luv4_8u, rgb_to_luv4_32f},
-                {rgba_to_luv4_8u, rgba_to_luv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lbgr_to_luv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lbgr_to_luv_8u, lbgr_to_luv_32f},
-                {lbgra_to_luv_8u, lbgra_to_luv_32f}
-            },
-            {
-                {lbgr_to_luv4_8u, lbgr_to_luv4_32f},
-                {lbgra_to_luv4_8u, lbgra_to_luv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void lrgb_to_luv(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {lrgb_to_luv_8u, lrgb_to_luv_32f},
-                {lrgba_to_luv_8u, lrgba_to_luv_32f}
-            },
-            {
-                {lrgb_to_luv4_8u, lrgb_to_luv4_32f},
-                {lrgba_to_luv4_8u, lrgba_to_luv4_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void luv_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {luv_to_bgr_8u, luv_to_bgr_32f},
-                {luv4_to_bgr_8u, luv4_to_bgr_32f}
-            },
-            {
-                {luv_to_bgra_8u, luv_to_bgra_32f},
-                {luv4_to_bgra_8u, luv4_to_bgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void luv_to_rgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {luv_to_rgb_8u, luv_to_rgb_32f},
-                {luv4_to_rgb_8u, luv4_to_rgb_32f}
-            },
-            {
-                {luv_to_rgba_8u, luv_to_rgba_32f},
-                {luv4_to_rgba_8u, luv4_to_rgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void luv_to_lbgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {luv_to_lbgr_8u, luv_to_lbgr_32f},
-                {luv4_to_lbgr_8u, luv4_to_lbgr_32f}
-            },
-            {
-                {luv_to_lbgra_8u, luv_to_lbgra_32f},
-                {luv4_to_lbgra_8u, luv4_to_lbgra_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void luv_to_lrgb(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        using namespace cv::gpu::device;
-        static const gpu_func_t funcs[2][2][2] =
-        {
-            {
-                {luv_to_lrgb_8u, luv_to_lrgb_32f},
-                {luv4_to_lrgb_8u, luv4_to_lrgb_32f}
-            },
-            {
-                {luv_to_lrgba_8u, luv_to_lrgba_32f},
-                {luv4_to_lrgba_8u, luv4_to_lrgba_32f}
-            }
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.depth() == CV_8U || src.depth() == CV_32F);
-        CV_Assert(src.channels() == 3 || src.channels() == 4);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
-
-        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
-    }
-
-    void rgba_to_mbgra(const GpuMat& src, GpuMat& dst, int, Stream& st)
-    {
-    #if (CUDA_VERSION < 5000)
-        (void)src;
-        (void)dst;
-        (void)st;
-        CV_Error( CV_StsBadFlag, "Unknown/unsupported color conversion code" );
-    #else
-        CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4);
-
-        dst.create(src.size(), src.type());
-
-        cudaStream_t stream = StreamAccessor::getStream(st);
-        NppStreamHandler h(stream);
-
-        NppiSize oSizeROI;
-        oSizeROI.width = src.cols;
-        oSizeROI.height = src.rows;
-
-        if (src.depth() == CV_8U)
-            nppSafeCall( nppiAlphaPremul_8u_AC4R(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), oSizeROI) );
-        else
-            nppSafeCall( nppiAlphaPremul_16u_AC4R(src.ptr<Npp16u>(), static_cast<int>(src.step), dst.ptr<Npp16u>(), static_cast<int>(dst.step), oSizeROI) );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    #endif
-    }
-
-    void bayer_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, bool blue_last, bool start_with_green, Stream& stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-        static const func_t funcs[3][4] =
-        {
-            {0,0,Bayer2BGR_8u_gpu<3>, Bayer2BGR_8u_gpu<4>},
-            {0,0,0,0},
-            {0,0,Bayer2BGR_16u_gpu<3>, Bayer2BGR_16u_gpu<4>}
-        };
-
-        if (dcn <= 0) dcn = 3;
-
-        CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1);
-        CV_Assert(src.rows > 2 && src.cols > 2);
-        CV_Assert(dcn == 3 || dcn == 4);
-
-        dst.create(src.size(), CV_MAKETYPE(src.depth(), dcn));
-
-        funcs[src.depth()][dcn - 1](src, dst, blue_last, start_with_green, StreamAccessor::getStream(stream));
-    }
-
-    void bayerBG_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        bayer_to_bgr(src, dst, dcn, false, false, stream);
-    }
-
-    void bayerGB_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        bayer_to_bgr(src, dst, dcn, false, true, stream);
-    }
-
-    void bayerRG_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        bayer_to_bgr(src, dst, dcn, true, false, stream);
-    }
-
-    void bayerGR_to_bgr(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream)
-    {
-        bayer_to_bgr(src, dst, dcn, true, true, stream);
-    }
-}
-
-void cv::gpu::cvtColor(const GpuMat& src, GpuMat& dst, int code, int dcn, Stream& stream)
-{
-    typedef void (*func_t)(const GpuMat& src, GpuMat& dst, int dcn, Stream& stream);
-    static const func_t funcs[] =
-    {
-        bgr_to_bgra,            // CV_BGR2BGRA    =0
-        bgra_to_bgr,            // CV_BGRA2BGR    =1
-        bgr_to_rgba,            // CV_BGR2RGBA    =2
-        bgra_to_rgb,            // CV_RGBA2BGR    =3
-        bgr_to_rgb,             // CV_BGR2RGB     =4
-        bgra_to_rgba,           // CV_BGRA2RGBA   =5
-
-        bgr_to_gray,            // CV_BGR2GRAY    =6
-        rgb_to_gray,            // CV_RGB2GRAY    =7
-        gray_to_bgr,            // CV_GRAY2BGR    =8
-        gray_to_bgra,           // CV_GRAY2BGRA   =9
-        bgra_to_gray,           // CV_BGRA2GRAY   =10
-        rgba_to_gray,           // CV_RGBA2GRAY   =11
-
-        bgr_to_bgr565,          // CV_BGR2BGR565  =12
-        rgb_to_bgr565,          // CV_RGB2BGR565  =13
-        bgr565_to_bgr,          // CV_BGR5652BGR  =14
-        bgr565_to_rgb,          // CV_BGR5652RGB  =15
-        bgra_to_bgr565,         // CV_BGRA2BGR565 =16
-        rgba_to_bgr565,         // CV_RGBA2BGR565 =17
-        bgr565_to_bgra,         // CV_BGR5652BGRA =18
-        bgr565_to_rgba,         // CV_BGR5652RGBA =19
-
-        gray_to_bgr565,         // CV_GRAY2BGR565 =20
-        bgr565_to_gray,         // CV_BGR5652GRAY =21
-
-        bgr_to_bgr555,          // CV_BGR2BGR555  =22
-        rgb_to_bgr555,          // CV_RGB2BGR555  =23
-        bgr555_to_bgr,          // CV_BGR5552BGR  =24
-        bgr555_to_rgb,          // CV_BGR5552RGB  =25
-        bgra_to_bgr555,         // CV_BGRA2BGR555 =26
-        rgba_to_bgr555,         // CV_RGBA2BGR555 =27
-        bgr555_to_bgra,         // CV_BGR5552BGRA =28
-        bgr555_to_rgba,         // CV_BGR5552RGBA =29
-
-        gray_to_bgr555,         // CV_GRAY2BGR555 =30
-        bgr555_to_gray,         // CV_BGR5552GRAY =31
-
-        bgr_to_xyz,             // CV_BGR2XYZ     =32
-        rgb_to_xyz,             // CV_RGB2XYZ     =33
-        xyz_to_bgr,             // CV_XYZ2BGR     =34
-        xyz_to_rgb,             // CV_XYZ2RGB     =35
-
-        bgr_to_YCrCb,           // CV_BGR2YCrCb   =36
-        rgb_to_YCrCb,           // CV_RGB2YCrCb   =37
-        YCrCb_to_bgr,           // CV_YCrCb2BGR   =38
-        YCrCb_to_rgb,           // CV_YCrCb2RGB   =39
-
-        bgr_to_hsv,             // CV_BGR2HSV     =40
-        rgb_to_hsv,             // CV_RGB2HSV     =41
-
-        0,                      //                =42
-        0,                      //                =43
-
-        bgr_to_lab,             // CV_BGR2Lab     =44
-        rgb_to_lab,             // CV_RGB2Lab     =45
-
-        bayerBG_to_bgr,         // CV_BayerBG2BGR =46
-        bayerGB_to_bgr,         // CV_BayerGB2BGR =47
-        bayerRG_to_bgr,         // CV_BayerRG2BGR =48
-        bayerGR_to_bgr,         // CV_BayerGR2BGR =49
-
-        bgr_to_luv,             // CV_BGR2Luv     =50
-        rgb_to_luv,             // CV_RGB2Luv     =51
-
-        bgr_to_hls,             // CV_BGR2HLS     =52
-        rgb_to_hls,             // CV_RGB2HLS     =53
-
-        hsv_to_bgr,             // CV_HSV2BGR     =54
-        hsv_to_rgb,             // CV_HSV2RGB     =55
-
-        lab_to_bgr,             // CV_Lab2BGR     =56
-        lab_to_rgb,             // CV_Lab2RGB     =57
-        luv_to_bgr,             // CV_Luv2BGR     =58
-        luv_to_rgb,             // CV_Luv2RGB     =59
-
-        hls_to_bgr,             // CV_HLS2BGR     =60
-        hls_to_rgb,             // CV_HLS2RGB     =61
-
-        0,                      // CV_BayerBG2BGR_VNG =62
-        0,                      // CV_BayerGB2BGR_VNG =63
-        0,                      // CV_BayerRG2BGR_VNG =64
-        0,                      // CV_BayerGR2BGR_VNG =65
-
-        bgr_to_hsv_full,        // CV_BGR2HSV_FULL = 66
-        rgb_to_hsv_full,        // CV_RGB2HSV_FULL = 67
-        bgr_to_hls_full,        // CV_BGR2HLS_FULL = 68
-        rgb_to_hls_full,        // CV_RGB2HLS_FULL = 69
-
-        hsv_to_bgr_full,        // CV_HSV2BGR_FULL = 70
-        hsv_to_rgb_full,        // CV_HSV2RGB_FULL = 71
-        hls_to_bgr_full,        // CV_HLS2BGR_FULL = 72
-        hls_to_rgb_full,        // CV_HLS2RGB_FULL = 73
-
-        lbgr_to_lab,            // CV_LBGR2Lab     = 74
-        lrgb_to_lab,            // CV_LRGB2Lab     = 75
-        lbgr_to_luv,            // CV_LBGR2Luv     = 76
-        lrgb_to_luv,            // CV_LRGB2Luv     = 77
-
-        lab_to_lbgr,            // CV_Lab2LBGR     = 78
-        lab_to_lrgb,            // CV_Lab2LRGB     = 79
-        luv_to_lbgr,            // CV_Luv2LBGR     = 80
-        luv_to_lrgb,            // CV_Luv2LRGB     = 81
-
-        bgr_to_yuv,             // CV_BGR2YUV      = 82
-        rgb_to_yuv,             // CV_RGB2YUV      = 83
-        yuv_to_bgr,             // CV_YUV2BGR      = 84
-        yuv_to_rgb,             // CV_YUV2RGB      = 85
-
-        0,                      // CV_BayerBG2GRAY = 86
-        0,                      // CV_BayerGB2GRAY = 87
-        0,                      // CV_BayerRG2GRAY = 88
-        0,                      // CV_BayerGR2GRAY = 89
-
-        //YUV 4:2:0 formats family
-        0,                      // CV_YUV2RGB_NV12 = 90,
-        0,                      // CV_YUV2BGR_NV12 = 91,
-        0,                      // CV_YUV2RGB_NV21 = 92,
-        0,                      // CV_YUV2BGR_NV21 = 93,
-
-        0,                      // CV_YUV2RGBA_NV12 = 94,
-        0,                      // CV_YUV2BGRA_NV12 = 95,
-        0,                      // CV_YUV2RGBA_NV21 = 96,
-        0,                      // CV_YUV2BGRA_NV21 = 97,
-
-        0,                      // CV_YUV2RGB_YV12 = 98,
-        0,                      // CV_YUV2BGR_YV12 = 99,
-        0,                      // CV_YUV2RGB_IYUV = 100,
-        0,                      // CV_YUV2BGR_IYUV = 101,
-
-        0,                      // CV_YUV2RGBA_YV12 = 102,
-        0,                      // CV_YUV2BGRA_YV12 = 103,
-        0,                      // CV_YUV2RGBA_IYUV = 104,
-        0,                      // CV_YUV2BGRA_IYUV = 105,
-
-        0,                      // CV_YUV2GRAY_420 = 106,
-
-        //YUV 4:2:2 formats family
-        0,                      // CV_YUV2RGB_UYVY = 107,
-        0,                      // CV_YUV2BGR_UYVY = 108,
-        0,                      // //CV_YUV2RGB_VYUY = 109,
-        0,                      // //CV_YUV2BGR_VYUY = 110,
-
-        0,                      // CV_YUV2RGBA_UYVY = 111,
-        0,                      // CV_YUV2BGRA_UYVY = 112,
-        0,                      // //CV_YUV2RGBA_VYUY = 113,
-        0,                      // //CV_YUV2BGRA_VYUY = 114,
-
-        0,                      // CV_YUV2RGB_YUY2 = 115,
-        0,                      // CV_YUV2BGR_YUY2 = 116,
-        0,                      // CV_YUV2RGB_YVYU = 117,
-        0,                      // CV_YUV2BGR_YVYU = 118,
-
-        0,                      // CV_YUV2RGBA_YUY2 = 119,
-        0,                      // CV_YUV2BGRA_YUY2 = 120,
-        0,                      // CV_YUV2RGBA_YVYU = 121,
-        0,                      // CV_YUV2BGRA_YVYU = 122,
-
-        0,                      // CV_YUV2GRAY_UYVY = 123,
-        0,                      // CV_YUV2GRAY_YUY2 = 124,
-
-        // alpha premultiplication
-        rgba_to_mbgra,          // CV_RGBA2mRGBA = 125,
-        0,                      // CV_mRGBA2RGBA = 126,
-
-        0,                      // CV_COLORCVT_MAX  = 127
-    };
-
-    CV_Assert(code < 128);
-
-    func_t func = funcs[code];
-
-    if (func == 0)
-        CV_Error( CV_StsBadFlag, "Unknown/unsupported color conversion code" );
-
-    func(src, dst, dcn, stream);
-}
-
-void cv::gpu::swapChannels(GpuMat& image, const int dstOrder[4], Stream& s)
-{
-    CV_Assert(image.type() == CV_8UC4);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    NppStreamHandler h(stream);
-
-    NppiSize sz;
-    sz.width  = image.cols;
-    sz.height = image.rows;
-
-    nppSafeCall( nppiSwapChannels_8u_C4IR(image.ptr<Npp8u>(), static_cast<int>(image.step), sz, dstOrder) );
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-void cv::gpu::gammaCorrection(const GpuMat& src, GpuMat& dst, bool forward, Stream& stream)
-{
-#if (CUDA_VERSION < 5000)
-    (void)src;
-    (void)dst;
-    (void)forward;
-    (void)stream;
-    CV_Error( CV_StsNotImplemented, "This function works only with CUDA 5.0 or higher" );
-#else
-    typedef NppStatus (*func_t)(const Npp8u* pSrc, int nSrcStep, Npp8u* pDst, int nDstStep, NppiSize oSizeROI);
-    typedef NppStatus (*func_inplace_t)(Npp8u* pSrcDst, int nSrcDstStep, NppiSize oSizeROI);
-
-    static const func_t funcs[2][5] =
-    {
-        {0, 0, 0, nppiGammaInv_8u_C3R, nppiGammaInv_8u_AC4R},
-        {0, 0, 0, nppiGammaFwd_8u_C3R, nppiGammaFwd_8u_AC4R}
-    };
-    static const func_inplace_t funcs_inplace[2][5] =
-    {
-        {0, 0, 0, nppiGammaInv_8u_C3IR, nppiGammaInv_8u_AC4IR},
-        {0, 0, 0, nppiGammaFwd_8u_C3IR, nppiGammaFwd_8u_AC4IR}
-    };
-
-    CV_Assert(src.type() == CV_8UC3 || src.type() == CV_8UC4);
-
-    dst.create(src.size(), src.type());
-
-    NppStreamHandler h(StreamAccessor::getStream(stream));
-
-    NppiSize oSizeROI;
-    oSizeROI.width = src.cols;
-    oSizeROI.height = src.rows;
-
-    if (dst.data == src.data)
-        funcs_inplace[forward][src.channels()](dst.ptr<Npp8u>(), static_cast<int>(src.step), oSizeROI);
-    else
-        funcs[forward][src.channels()](src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), oSizeROI);
-
-#endif
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/cu_safe_call.cpp b/modules/gpu/src/cu_safe_call.cpp
deleted file mode 100644
index 7218873dff..0000000000
--- a/modules/gpu/src/cu_safe_call.cpp
+++ /dev/null
@@ -1,139 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "cu_safe_call.h"
-
-#ifdef HAVE_CUDA
-
-namespace
-{
-    #define error_entry(entry)  { entry, #entry }
-
-    struct ErrorEntry
-    {
-        int code;
-        std::string str;
-    };
-
-    class ErrorEntryComparer
-    {
-    public:
-        inline ErrorEntryComparer(int code) : code_(code) {}
-
-        inline bool operator()(const ErrorEntry& e) const { return e.code == code_; }
-
-    private:
-        int code_;
-    };
-
-    std::string getErrorString(int code, const ErrorEntry* errors, size_t n)
-    {
-        size_t idx = std::find_if(errors, errors + n, ErrorEntryComparer(code)) - errors;
-
-        const std::string& msg = (idx != n) ? errors[idx].str : std::string("Unknown error code");
-
-        std::ostringstream ostr;
-        ostr << msg << " [Code = " << code << "]";
-
-        return ostr.str();
-    }
-
-    const ErrorEntry cu_errors [] =
-    {
-        error_entry( CUDA_SUCCESS                              ),
-        error_entry( CUDA_ERROR_INVALID_VALUE                  ),
-        error_entry( CUDA_ERROR_OUT_OF_MEMORY                  ),
-        error_entry( CUDA_ERROR_NOT_INITIALIZED                ),
-        error_entry( CUDA_ERROR_DEINITIALIZED                  ),
-        error_entry( CUDA_ERROR_PROFILER_DISABLED              ),
-        error_entry( CUDA_ERROR_PROFILER_NOT_INITIALIZED       ),
-        error_entry( CUDA_ERROR_PROFILER_ALREADY_STARTED       ),
-        error_entry( CUDA_ERROR_PROFILER_ALREADY_STOPPED       ),
-        error_entry( CUDA_ERROR_NO_DEVICE                      ),
-        error_entry( CUDA_ERROR_INVALID_DEVICE                 ),
-        error_entry( CUDA_ERROR_INVALID_IMAGE                  ),
-        error_entry( CUDA_ERROR_INVALID_CONTEXT                ),
-        error_entry( CUDA_ERROR_CONTEXT_ALREADY_CURRENT        ),
-        error_entry( CUDA_ERROR_MAP_FAILED                     ),
-        error_entry( CUDA_ERROR_UNMAP_FAILED                   ),
-        error_entry( CUDA_ERROR_ARRAY_IS_MAPPED                ),
-        error_entry( CUDA_ERROR_ALREADY_MAPPED                 ),
-        error_entry( CUDA_ERROR_NO_BINARY_FOR_GPU              ),
-        error_entry( CUDA_ERROR_ALREADY_ACQUIRED               ),
-        error_entry( CUDA_ERROR_NOT_MAPPED                     ),
-        error_entry( CUDA_ERROR_NOT_MAPPED_AS_ARRAY            ),
-        error_entry( CUDA_ERROR_NOT_MAPPED_AS_POINTER          ),
-        error_entry( CUDA_ERROR_ECC_UNCORRECTABLE              ),
-        error_entry( CUDA_ERROR_UNSUPPORTED_LIMIT              ),
-        error_entry( CUDA_ERROR_CONTEXT_ALREADY_IN_USE         ),
-        error_entry( CUDA_ERROR_INVALID_SOURCE                 ),
-        error_entry( CUDA_ERROR_FILE_NOT_FOUND                 ),
-        error_entry( CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND ),
-        error_entry( CUDA_ERROR_SHARED_OBJECT_INIT_FAILED      ),
-        error_entry( CUDA_ERROR_OPERATING_SYSTEM               ),
-        error_entry( CUDA_ERROR_INVALID_HANDLE                 ),
-        error_entry( CUDA_ERROR_NOT_FOUND                      ),
-        error_entry( CUDA_ERROR_NOT_READY                      ),
-        error_entry( CUDA_ERROR_LAUNCH_FAILED                  ),
-        error_entry( CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES        ),
-        error_entry( CUDA_ERROR_LAUNCH_TIMEOUT                 ),
-        error_entry( CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING  ),
-        error_entry( CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED    ),
-        error_entry( CUDA_ERROR_PEER_ACCESS_NOT_ENABLED        ),
-        error_entry( CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE         ),
-        error_entry( CUDA_ERROR_CONTEXT_IS_DESTROYED           ),
-        error_entry( CUDA_ERROR_ASSERT                         ),
-        error_entry( CUDA_ERROR_TOO_MANY_PEERS                 ),
-        error_entry( CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED ),
-        error_entry( CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED     ),
-        error_entry( CUDA_ERROR_UNKNOWN                        )
-    };
-
-    const size_t cu_errors_num = sizeof(cu_errors) / sizeof(cu_errors[0]);
-}
-
-std::string cv::gpu::detail::cuGetErrString(CUresult res)
-{
-    return getErrorString(res, cu_errors, cu_errors_num);
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/cuda/NV12ToARGB.cu b/modules/gpu/src/cuda/NV12ToARGB.cu
deleted file mode 100644
index 8ad6cfdfae..0000000000
--- a/modules/gpu/src/cuda/NV12ToARGB.cu
+++ /dev/null
@@ -1,212 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * Please refer to the NVIDIA end user license agreement (EULA) associated
- * with this source code for terms and conditions that govern your use of
- * this software. Any use, reproduction, disclosure, or distribution of
- * this software and related documentation outside the terms of the EULA
- * is strictly prohibited.
- *
- */
-
-/*
-    NV12ToARGB color space conversion CUDA kernel
-
-    This sample uses CUDA to perform a simple NV12 (YUV 4:2:0 planar)
-    source and converts to output in ARGB format
-*/
-
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-
-namespace cv { namespace gpu { namespace device {
-    namespace video_decoding
-    {
-        __constant__ uint constAlpha = ((uint)0xff << 24);
-
-        __constant__ float constHueColorSpaceMat[9];
-
-        void loadHueCSC(float hueCSC[9])
-        {
-            cudaSafeCall( cudaMemcpyToSymbol(constHueColorSpaceMat, hueCSC, 9 * sizeof(float)) );
-        }
-
-        __device__ void YUV2RGB(const uint* yuvi, float* red, float* green, float* blue)
-        {
-            float luma, chromaCb, chromaCr;
-
-            // Prepare for hue adjustment
-            luma     = (float)yuvi[0];
-            chromaCb = (float)((int)yuvi[1] - 512.0f);
-            chromaCr = (float)((int)yuvi[2] - 512.0f);
-
-           // Convert YUV To RGB with hue adjustment
-           *red   = (luma     * constHueColorSpaceMat[0]) +
-                    (chromaCb * constHueColorSpaceMat[1]) +
-                    (chromaCr * constHueColorSpaceMat[2]);
-
-           *green = (luma     * constHueColorSpaceMat[3]) +
-                    (chromaCb * constHueColorSpaceMat[4]) +
-                    (chromaCr * constHueColorSpaceMat[5]);
-
-           *blue  = (luma     * constHueColorSpaceMat[6]) +
-                    (chromaCb * constHueColorSpaceMat[7]) +
-                    (chromaCr * constHueColorSpaceMat[8]);
-        }
-
-        __device__ uint RGBAPACK_10bit(float red, float green, float blue, uint alpha)
-        {
-            uint ARGBpixel = 0;
-
-            // Clamp final 10 bit results
-            red   = ::fmin(::fmax(red,   0.0f), 1023.f);
-            green = ::fmin(::fmax(green, 0.0f), 1023.f);
-            blue  = ::fmin(::fmax(blue,  0.0f), 1023.f);
-
-            // Convert to 8 bit unsigned integers per color component
-            ARGBpixel = (((uint)blue  >> 2) |
-                        (((uint)green >> 2) << 8)  |
-                        (((uint)red   >> 2) << 16) |
-                        (uint)alpha);
-
-            return ARGBpixel;
-        }
-
-        // CUDA kernel for outputing the final ARGB output from NV12
-
-        #define COLOR_COMPONENT_BIT_SIZE 10
-        #define COLOR_COMPONENT_MASK     0x3FF
-
-        __global__ void NV12ToARGB(uchar* srcImage, size_t nSourcePitch,
-                                   uint* dstImage, size_t nDestPitch,
-                                   uint width, uint height)
-        {
-            // Pad borders with duplicate pixels, and we multiply by 2 because we process 2 pixels per thread
-            const int x = blockIdx.x * (blockDim.x << 1) + (threadIdx.x << 1);
-            const int y = blockIdx.y *  blockDim.y       +  threadIdx.y;
-
-            if (x >= width || y >= height)
-                return;
-
-            // Read 2 Luma components at a time, so we don't waste processing since CbCr are decimated this way.
-            // if we move to texture we could read 4 luminance values
-
-            uint yuv101010Pel[2];
-
-            yuv101010Pel[0] = (srcImage[y * nSourcePitch + x    ]) << 2;
-            yuv101010Pel[1] = (srcImage[y * nSourcePitch + x + 1]) << 2;
-
-            const size_t chromaOffset = nSourcePitch * height;
-
-            const int y_chroma = y >> 1;
-
-            if (y & 1)  // odd scanline ?
-            {
-                uint chromaCb = srcImage[chromaOffset + y_chroma * nSourcePitch + x    ];
-                uint chromaCr = srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1];
-
-                if (y_chroma < ((height >> 1) - 1)) // interpolate chroma vertically
-                {
-                    chromaCb = (chromaCb + srcImage[chromaOffset + (y_chroma + 1) * nSourcePitch + x    ] + 1) >> 1;
-                    chromaCr = (chromaCr + srcImage[chromaOffset + (y_chroma + 1) * nSourcePitch + x + 1] + 1) >> 1;
-                }
-
-                yuv101010Pel[0] |= (chromaCb << ( COLOR_COMPONENT_BIT_SIZE       + 2));
-                yuv101010Pel[0] |= (chromaCr << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
-
-                yuv101010Pel[1] |= (chromaCb << ( COLOR_COMPONENT_BIT_SIZE       + 2));
-                yuv101010Pel[1] |= (chromaCr << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
-            }
-            else
-            {
-                yuv101010Pel[0] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x    ] << ( COLOR_COMPONENT_BIT_SIZE       + 2));
-                yuv101010Pel[0] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1] << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
-
-                yuv101010Pel[1] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x    ] << ( COLOR_COMPONENT_BIT_SIZE       + 2));
-                yuv101010Pel[1] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1] << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
-            }
-
-            // this steps performs the color conversion
-            uint yuvi[6];
-            float red[2], green[2], blue[2];
-
-            yuvi[0] =  (yuv101010Pel[0] &   COLOR_COMPONENT_MASK    );
-            yuvi[1] = ((yuv101010Pel[0] >>  COLOR_COMPONENT_BIT_SIZE)       & COLOR_COMPONENT_MASK);
-            yuvi[2] = ((yuv101010Pel[0] >> (COLOR_COMPONENT_BIT_SIZE << 1)) & COLOR_COMPONENT_MASK);
-
-            yuvi[3] =  (yuv101010Pel[1] &   COLOR_COMPONENT_MASK    );
-            yuvi[4] = ((yuv101010Pel[1] >>  COLOR_COMPONENT_BIT_SIZE)       & COLOR_COMPONENT_MASK);
-            yuvi[5] = ((yuv101010Pel[1] >> (COLOR_COMPONENT_BIT_SIZE << 1)) & COLOR_COMPONENT_MASK);
-
-            // YUV to RGB Transformation conversion
-            YUV2RGB(&yuvi[0], &red[0], &green[0], &blue[0]);
-            YUV2RGB(&yuvi[3], &red[1], &green[1], &blue[1]);
-
-            // Clamp the results to RGBA
-
-            const size_t dstImagePitch = nDestPitch >> 2;
-
-            dstImage[y * dstImagePitch + x     ] = RGBAPACK_10bit(red[0], green[0], blue[0], constAlpha);
-            dstImage[y * dstImagePitch + x + 1 ] = RGBAPACK_10bit(red[1], green[1], blue[1], constAlpha);
-        }
-
-        void NV12ToARGB_gpu(const PtrStepb decodedFrame, PtrStepSz<uint> interopFrame, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(interopFrame.cols, 2 * block.x), divUp(interopFrame.rows, block.y));
-
-            NV12ToARGB<<<grid, block, 0, stream>>>(decodedFrame.data, decodedFrame.step, interopFrame.data, interopFrame.step,
-                interopFrame.cols, interopFrame.rows);
-
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    }
-}}}
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/cuda/bgfg_vibe.cu b/modules/gpu/src/cuda/bgfg_vibe.cu
deleted file mode 100644
index 72ea7f0ae7..0000000000
--- a/modules/gpu/src/cuda/bgfg_vibe.cu
+++ /dev/null
@@ -1,258 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace vibe
-    {
-        __constant__ int c_nbSamples;
-        __constant__ int c_reqMatches;
-        __constant__ int c_radius;
-        __constant__ int c_subsamplingFactor;
-
-        void loadConstants(int nbSamples, int reqMatches, int radius, int subsamplingFactor)
-        {
-            cudaSafeCall( cudaMemcpyToSymbol(c_nbSamples, &nbSamples, sizeof(int)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_reqMatches, &reqMatches, sizeof(int)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_radius, &radius, sizeof(int)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_subsamplingFactor, &subsamplingFactor, sizeof(int)) );
-        }
-
-        __device__ __forceinline__ uint nextRand(uint& state)
-        {
-            const unsigned int CV_RNG_COEFF = 4164903690U;
-            state = state * CV_RNG_COEFF + (state >> 16);
-            return state;
-        }
-
-        __constant__ int c_xoff[9] = {-1,  0,  1, -1, 1, -1, 0, 1, 0};
-        __constant__ int c_yoff[9] = {-1, -1, -1,  0, 0,  1, 1, 1, 0};
-
-        __device__ __forceinline__ int2 chooseRandomNeighbor(int x, int y, uint& randState, int count = 8)
-        {
-            int idx = nextRand(randState) % count;
-
-            return make_int2(x + c_xoff[idx], y + c_yoff[idx]);
-        }
-
-        __device__ __forceinline__ uchar cvt(uchar val)
-        {
-            return val;
-        }
-        __device__ __forceinline__ uchar4 cvt(const uchar3& val)
-        {
-            return make_uchar4(val.x, val.y, val.z, 0);
-        }
-        __device__ __forceinline__ uchar4 cvt(const uchar4& val)
-        {
-            return val;
-        }
-
-        template <typename SrcT, typename SampleT>
-        __global__ void init(const PtrStepSz<SrcT> frame, PtrStep<SampleT> samples, PtrStep<uint> randStates)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= frame.cols || y >= frame.rows)
-                return;
-
-            uint localState = randStates(y, x);
-
-            for (int k = 0; k < c_nbSamples; ++k)
-            {
-                int2 np = chooseRandomNeighbor(x, y, localState, 9);
-
-                np.x = ::max(0, ::min(np.x, frame.cols - 1));
-                np.y = ::max(0, ::min(np.y, frame.rows - 1));
-
-                SrcT pix = frame(np.y, np.x);
-
-                samples(k * frame.rows + y, x) = cvt(pix);
-            }
-
-            randStates(y, x) = localState;
-        }
-
-        template <typename SrcT, typename SampleT>
-        void init_caller(PtrStepSzb frame, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(frame.cols, block.x), divUp(frame.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(init<SrcT, SampleT>, cudaFuncCachePreferL1) );
-
-            init<SrcT, SampleT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, (PtrStepSz<SampleT>) samples, randStates);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void init_gpu(PtrStepSzb frame, int cn, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream)
-        {
-            typedef void (*func_t)(PtrStepSzb frame, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream);
-            static const func_t funcs[] =
-            {
-                0, init_caller<uchar, uchar>, 0, init_caller<uchar3, uchar4>, init_caller<uchar4, uchar4>
-            };
-
-            funcs[cn](frame, samples, randStates, stream);
-        }
-
-        __device__ __forceinline__ int calcDist(uchar a, uchar b)
-        {
-            return ::abs(a - b);
-        }
-        __device__ __forceinline__ int calcDist(const uchar3& a, const uchar4& b)
-        {
-            return (::abs(a.x - b.x) + ::abs(a.y - b.y) + ::abs(a.z - b.z)) / 3;
-        }
-        __device__ __forceinline__ int calcDist(const uchar4& a, const uchar4& b)
-        {
-            return (::abs(a.x - b.x) + ::abs(a.y - b.y) + ::abs(a.z - b.z)) / 3;
-        }
-
-        template <typename SrcT, typename SampleT>
-        __global__ void update(const PtrStepSz<SrcT> frame, PtrStepb fgmask, PtrStep<SampleT> samples, PtrStep<uint> randStates)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= frame.cols || y >= frame.rows)
-                return;
-
-            uint localState = randStates(y, x);
-
-            SrcT imgPix = frame(y, x);
-
-            // comparison with the model
-
-            int count = 0;
-            for (int k = 0; (count < c_reqMatches) && (k < c_nbSamples); ++k)
-            {
-                SampleT samplePix = samples(k * frame.rows + y, x);
-
-                int distance = calcDist(imgPix, samplePix);
-
-                if (distance < c_radius)
-                    ++count;
-            }
-
-            // pixel classification according to reqMatches
-
-            fgmask(y, x) = (uchar) (-(count < c_reqMatches));
-
-            if (count >= c_reqMatches)
-            {
-                // the pixel belongs to the background
-
-                // gets a random number between 0 and subsamplingFactor-1
-                int randomNumber = nextRand(localState) % c_subsamplingFactor;
-
-                // update of the current pixel model
-                if (randomNumber == 0)
-                {
-                    // random subsampling
-
-                    int k = nextRand(localState) % c_nbSamples;
-
-                    samples(k * frame.rows + y, x) = cvt(imgPix);
-                }
-
-                // update of a neighboring pixel model
-                randomNumber = nextRand(localState) % c_subsamplingFactor;
-
-                if (randomNumber == 0)
-                {
-                    // random subsampling
-
-                    // chooses a neighboring pixel randomly
-                    int2 np = chooseRandomNeighbor(x, y, localState);
-
-                    np.x = ::max(0, ::min(np.x, frame.cols - 1));
-                    np.y = ::max(0, ::min(np.y, frame.rows - 1));
-
-                    // chooses the value to be replaced randomly
-                    int k = nextRand(localState) % c_nbSamples;
-
-                    samples(k * frame.rows + np.y, np.x) = cvt(imgPix);
-                }
-            }
-
-            randStates(y, x) = localState;
-        }
-
-        template <typename SrcT, typename SampleT>
-        void update_caller(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(frame.cols, block.x), divUp(frame.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(update<SrcT, SampleT>, cudaFuncCachePreferL1) );
-
-            update<SrcT, SampleT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, (PtrStepSz<SampleT>) samples, randStates);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void update_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream)
-        {
-            typedef void (*func_t)(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb samples, PtrStepSz<uint> randStates, cudaStream_t stream);
-            static const func_t funcs[] =
-            {
-                0, update_caller<uchar, uchar>, 0, update_caller<uchar3, uchar4>, update_caller<uchar4, uchar4>
-            };
-
-            funcs[cn](frame, fgmask, samples, randStates, stream);
-        }
-    }
-}}}
-
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/cuda/calib3d.cu b/modules/gpu/src/cuda/calib3d.cu
index 0fd482c41a..9adc7806f6 100644
--- a/modules/gpu/src/cuda/calib3d.cu
+++ b/modules/gpu/src/cuda/calib3d.cu
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/transform.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     #define SOLVE_PNP_RANSAC_MAX_NUM_ITERS 200
 
@@ -67,8 +67,8 @@ namespace cv { namespace gpu { namespace device
                         crot1.x * p.x + crot1.y * p.y + crot1.z * p.z + ctransl.y,
                         crot2.x * p.x + crot2.y * p.y + crot2.z * p.z + ctransl.z);
             }
-            __device__ __forceinline__ TransformOp() {}
-            __device__ __forceinline__ TransformOp(const TransformOp&) {}
+            __host__ __device__ __forceinline__ TransformOp() {}
+            __host__ __device__ __forceinline__ TransformOp(const TransformOp&) {}
         };
 
         void call(const PtrStepSz<float3> src, const float* rot,
@@ -79,7 +79,7 @@ namespace cv { namespace gpu { namespace device
             cudaSafeCall(cudaMemcpyToSymbol(crot1, rot + 3, sizeof(float) * 3));
             cudaSafeCall(cudaMemcpyToSymbol(crot2, rot + 6, sizeof(float) * 3));
             cudaSafeCall(cudaMemcpyToSymbol(ctransl, transl, sizeof(float) * 3));
-            cv::gpu::device::transform(src, dst, TransformOp(), WithOutMask(), stream);
+            cv::gpu::cudev::transform(src, dst, TransformOp(), WithOutMask(), stream);
         }
     } // namespace transform_points
 
@@ -106,8 +106,8 @@ namespace cv { namespace gpu { namespace device
                         (cproj0.x * t.x + cproj0.y * t.y) / t.z + cproj0.z,
                         (cproj1.x * t.x + cproj1.y * t.y) / t.z + cproj1.z);
             }
-            __device__ __forceinline__ ProjectOp() {}
-            __device__ __forceinline__ ProjectOp(const ProjectOp&) {}
+            __host__ __device__ __forceinline__ ProjectOp() {}
+            __host__ __device__ __forceinline__ ProjectOp(const ProjectOp&) {}
         };
 
         void call(const PtrStepSz<float3> src, const float* rot,
@@ -120,7 +120,7 @@ namespace cv { namespace gpu { namespace device
             cudaSafeCall(cudaMemcpyToSymbol(ctransl, transl, sizeof(float) * 3));
             cudaSafeCall(cudaMemcpyToSymbol(cproj0, proj, sizeof(float) * 3));
             cudaSafeCall(cudaMemcpyToSymbol(cproj1, proj + 3, sizeof(float) * 3));
-            cv::gpu::device::transform(src, dst, ProjectOp(), WithOutMask(), stream);
+            cv::gpu::cudev::transform(src, dst, ProjectOp(), WithOutMask(), stream);
         }
     } // namespace project_points
 
@@ -187,7 +187,7 @@ namespace cv { namespace gpu { namespace device
             cudaSafeCall( cudaDeviceSynchronize() );
         }
     } // namespace solvepnp_ransac
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/ccomponetns.cu b/modules/gpu/src/cuda/ccomponetns.cu
index 0533522bed..58ceb99cd8 100644
--- a/modules/gpu/src/cuda/ccomponetns.cu
+++ b/modules/gpu/src/cuda/ccomponetns.cu
@@ -1,31 +1,31 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
-// By downloading, copying, installing or using the software you agree to this license.
-// If you do not agree to this license, do not download, install,
-// copy or use the software.
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
 //
 //
-//                          License Agreement
-//               For Open Source Computer Vision Library
+//                           License Agreement
+//                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
-//  * Redistributions of source code must retain the above copyright notice,
-//    this list of conditions and the following disclaimer.
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
 //
-//  * Redistributions in binary form must reproduce the above copyright notice,
-//    this list of conditions and the following disclaimer in the documentation
-//    and/or other materials provided with the distribution.
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
 //
-//  * The name of the copyright holders may not be used to endorse or promote products
-//    derived from this software without specific prior written permission.
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
 // any express or implied warranties, including, but not limited to, the implied
@@ -37,19 +37,20 @@
 // and on any theory of liability, whether in contract, strict liability,
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
+//
 //M*/
 
 #if !defined CUDA_DISABLER
 
-#include <opencv2/gpu/device/common.hpp>
-#include <opencv2/gpu/device/vec_traits.hpp>
-#include <opencv2/gpu/device/vec_math.hpp>
-#include <opencv2/gpu/device/emulation.hpp>
+#include <opencv2/core/cuda/common.hpp>
+#include <opencv2/core/cuda/vec_traits.hpp>
+#include <opencv2/core/cuda/vec_math.hpp>
+#include <opencv2/core/cuda/emulation.hpp>
 
 #include <iostream>
 #include <stdio.h>
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace ccl
     {
@@ -152,7 +153,7 @@ namespace cv { namespace gpu { namespace device
 
             template<typename I> __device__ __forceinline__ bool operator() (const I& a, const I& b) const
             {
-                I d = a - b;
+                I d = saturate_cast<I>(a - b);
                 return lo.x <= d.x && d.x <= hi.x &&
                        lo.y <= d.y && d.y <= hi.y &&
                        lo.z <= d.z && d.z <= hi.z;
@@ -168,7 +169,7 @@ namespace cv { namespace gpu { namespace device
 
             template<typename I> __device__ __forceinline__ bool operator() (const I& a, const I& b) const
             {
-                I d = a - b;
+                I d = saturate_cast<I>(a - b);
                 return lo.x <= d.x && d.x <= hi.x &&
                        lo.y <= d.y && d.y <= hi.y &&
                        lo.z <= d.z && d.z <= hi.z &&
diff --git a/modules/gpu/src/cuda/debayer.cu b/modules/gpu/src/cuda/debayer.cu
deleted file mode 100644
index 57322ed81f..0000000000
--- a/modules/gpu/src/cuda/debayer.cu
+++ /dev/null
@@ -1,331 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include <opencv2/gpu/device/common.hpp>
-#include <opencv2/gpu/device/vec_traits.hpp>
-#include <opencv2/gpu/device/vec_math.hpp>
-#include <opencv2/gpu/device/limits.hpp>
-
-namespace cv { namespace gpu {
-    namespace device
-    {
-        template <typename D>
-        __global__ void Bayer2BGR_8u(const PtrStepb src, PtrStepSz<D> dst, const bool blue_last, const bool start_with_green)
-        {
-            const int s_x = blockIdx.x * blockDim.x + threadIdx.x;
-            int s_y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (s_y >= dst.rows || (s_x << 2) >= dst.cols)
-                return;
-
-            s_y = ::min(::max(s_y, 1), dst.rows - 2);
-
-            uchar4 patch[3][3];
-            patch[0][1] = ((const uchar4*) src.ptr(s_y - 1))[s_x];
-            patch[0][0] = ((const uchar4*) src.ptr(s_y - 1))[::max(s_x - 1, 0)];
-            patch[0][2] = ((const uchar4*) src.ptr(s_y - 1))[::min(s_x + 1, ((dst.cols + 3) >> 2) - 1)];
-
-            patch[1][1] = ((const uchar4*) src.ptr(s_y))[s_x];
-            patch[1][0] = ((const uchar4*) src.ptr(s_y))[::max(s_x - 1, 0)];
-            patch[1][2] = ((const uchar4*) src.ptr(s_y))[::min(s_x + 1, ((dst.cols + 3) >> 2) - 1)];
-
-            patch[2][1] = ((const uchar4*) src.ptr(s_y + 1))[s_x];
-            patch[2][0] = ((const uchar4*) src.ptr(s_y + 1))[::max(s_x - 1, 0)];
-            patch[2][2] = ((const uchar4*) src.ptr(s_y + 1))[::min(s_x + 1, ((dst.cols + 3) >> 2) - 1)];
-
-            D res0 = VecTraits<D>::all(numeric_limits<uchar>::max());
-            D res1 = VecTraits<D>::all(numeric_limits<uchar>::max());
-            D res2 = VecTraits<D>::all(numeric_limits<uchar>::max());
-            D res3 = VecTraits<D>::all(numeric_limits<uchar>::max());
-
-            if ((s_y & 1) ^ start_with_green)
-            {
-                const int t0 = (patch[0][1].x + patch[2][1].x + 1) >> 1;
-                const int t1 = (patch[1][0].w + patch[1][1].y + 1) >> 1;
-
-                const int t2 = (patch[0][1].x + patch[0][1].z + patch[2][1].x + patch[2][1].z + 2) >> 2;
-                const int t3 = (patch[0][1].y + patch[1][1].x + patch[1][1].z + patch[2][1].y + 2) >> 2;
-
-                const int t4 = (patch[0][1].z + patch[2][1].z + 1) >> 1;
-                const int t5 = (patch[1][1].y + patch[1][1].w + 1) >> 1;
-
-                const int t6 = (patch[0][1].z + patch[0][2].x + patch[2][1].z + patch[2][2].x + 2) >> 2;
-                const int t7 = (patch[0][1].w + patch[1][1].z + patch[1][2].x + patch[2][1].w + 2) >> 2;
-
-                if ((s_y & 1) ^ blue_last)
-                {
-                    res0.x = t1;
-                    res0.y = patch[1][1].x;
-                    res0.z = t0;
-
-                    res1.x = patch[1][1].y;
-                    res1.y = t3;
-                    res1.z = t2;
-
-                    res2.x = t5;
-                    res2.y = patch[1][1].z;
-                    res2.z = t4;
-
-                    res3.x = patch[1][1].w;
-                    res3.y = t7;
-                    res3.z = t6;
-                }
-                else
-                {
-                    res0.x = t0;
-                    res0.y = patch[1][1].x;
-                    res0.z = t1;
-
-                    res1.x = t2;
-                    res1.y = t3;
-                    res1.z = patch[1][1].y;
-
-                    res2.x = t4;
-                    res2.y = patch[1][1].z;
-                    res2.z = t5;
-
-                    res3.x = t6;
-                    res3.y = t7;
-                    res3.z = patch[1][1].w;
-                }
-            }
-            else
-            {
-                const int t0 = (patch[0][0].w + patch[0][1].y + patch[2][0].w + patch[2][1].y + 2) >> 2;
-                const int t1 = (patch[0][1].x + patch[1][0].w + patch[1][1].y + patch[2][1].x + 2) >> 2;
-
-                const int t2 = (patch[0][1].y + patch[2][1].y + 1) >> 1;
-                const int t3 = (patch[1][1].x + patch[1][1].z + 1) >> 1;
-
-                const int t4 = (patch[0][1].y + patch[0][1].w + patch[2][1].y + patch[2][1].w + 2) >> 2;
-                const int t5 = (patch[0][1].z + patch[1][1].y + patch[1][1].w + patch[2][1].z + 2) >> 2;
-
-                const int t6 = (patch[0][1].w + patch[2][1].w + 1) >> 1;
-                const int t7 = (patch[1][1].z + patch[1][2].x + 1) >> 1;
-
-                if ((s_y & 1) ^ blue_last)
-                {
-                    res0.x = patch[1][1].x;
-                    res0.y = t1;
-                    res0.z = t0;
-
-                    res1.x = t3;
-                    res1.y = patch[1][1].y;
-                    res1.z = t2;
-
-                    res2.x = patch[1][1].z;
-                    res2.y = t5;
-                    res2.z = t4;
-
-                    res3.x = t7;
-                    res3.y = patch[1][1].w;
-                    res3.z = t6;
-                }
-                else
-                {
-                    res0.x = t0;
-                    res0.y = t1;
-                    res0.z = patch[1][1].x;
-
-                    res1.x = t2;
-                    res1.y = patch[1][1].y;
-                    res1.z = t3;
-
-                    res2.x = t4;
-                    res2.y = t5;
-                    res2.z = patch[1][1].z;
-
-                    res3.x = t6;
-                    res3.y = patch[1][1].w;
-                    res3.z = t7;
-                }
-            }
-
-            const int d_x = (blockIdx.x * blockDim.x + threadIdx.x) << 2;
-            const int d_y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            dst(d_y, d_x) = res0;
-            if (d_x + 1 < dst.cols)
-                dst(d_y, d_x + 1) = res1;
-            if (d_x + 2 < dst.cols)
-                dst(d_y, d_x + 2) = res2;
-            if (d_x + 3 < dst.cols)
-                dst(d_y, d_x + 3) = res3;
-        }
-
-        template <typename D>
-        __global__ void Bayer2BGR_16u(const PtrStepb src, PtrStepSz<D> dst, const bool blue_last, const bool start_with_green)
-        {
-            const int s_x = blockIdx.x * blockDim.x + threadIdx.x;
-            int s_y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (s_y >= dst.rows || (s_x << 1) >= dst.cols)
-                return;
-
-            s_y = ::min(::max(s_y, 1), dst.rows - 2);
-
-            ushort2 patch[3][3];
-            patch[0][1] = ((const ushort2*) src.ptr(s_y - 1))[s_x];
-            patch[0][0] = ((const ushort2*) src.ptr(s_y - 1))[::max(s_x - 1, 0)];
-            patch[0][2] = ((const ushort2*) src.ptr(s_y - 1))[::min(s_x + 1, ((dst.cols + 1) >> 1) - 1)];
-
-            patch[1][1] = ((const ushort2*) src.ptr(s_y))[s_x];
-            patch[1][0] = ((const ushort2*) src.ptr(s_y))[::max(s_x - 1, 0)];
-            patch[1][2] = ((const ushort2*) src.ptr(s_y))[::min(s_x + 1, ((dst.cols + 1) >> 1) - 1)];
-
-            patch[2][1] = ((const ushort2*) src.ptr(s_y + 1))[s_x];
-            patch[2][0] = ((const ushort2*) src.ptr(s_y + 1))[::max(s_x - 1, 0)];
-            patch[2][2] = ((const ushort2*) src.ptr(s_y + 1))[::min(s_x + 1, ((dst.cols + 1) >> 1) - 1)];
-
-            D res0 = VecTraits<D>::all(numeric_limits<ushort>::max());
-            D res1 = VecTraits<D>::all(numeric_limits<ushort>::max());
-
-            if ((s_y & 1) ^ start_with_green)
-            {
-                const int t0 = (patch[0][1].x + patch[2][1].x + 1) >> 1;
-                const int t1 = (patch[1][0].y + patch[1][1].y + 1) >> 1;
-
-                const int t2 = (patch[0][1].x + patch[0][2].x + patch[2][1].x + patch[2][2].x + 2) >> 2;
-                const int t3 = (patch[0][1].y + patch[1][1].x + patch[1][2].x + patch[2][1].y + 2) >> 2;
-
-                if ((s_y & 1) ^ blue_last)
-                {
-                    res0.x = t1;
-                    res0.y = patch[1][1].x;
-                    res0.z = t0;
-
-                    res1.x = patch[1][1].y;
-                    res1.y = t3;
-                    res1.z = t2;
-                }
-                else
-                {
-                    res0.x = t0;
-                    res0.y = patch[1][1].x;
-                    res0.z = t1;
-
-                    res1.x = t2;
-                    res1.y = t3;
-                    res1.z = patch[1][1].y;
-                }
-            }
-            else
-            {
-                const int t0 = (patch[0][0].y + patch[0][1].y + patch[2][0].y + patch[2][1].y + 2) >> 2;
-                const int t1 = (patch[0][1].x + patch[1][0].y + patch[1][1].y + patch[2][1].x + 2) >> 2;
-
-                const int t2 = (patch[0][1].y + patch[2][1].y + 1) >> 1;
-                const int t3 = (patch[1][1].x + patch[1][2].x + 1) >> 1;
-
-                if ((s_y & 1) ^ blue_last)
-                {
-                    res0.x = patch[1][1].x;
-                    res0.y = t1;
-                    res0.z = t0;
-
-                    res1.x = t3;
-                    res1.y = patch[1][1].y;
-                    res1.z = t2;
-                }
-                else
-                {
-                    res0.x = t0;
-                    res0.y = t1;
-                    res0.z = patch[1][1].x;
-
-                    res1.x = t2;
-                    res1.y = patch[1][1].y;
-                    res1.z = t3;
-                }
-            }
-
-            const int d_x = (blockIdx.x * blockDim.x + threadIdx.x) << 1;
-            const int d_y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            dst(d_y, d_x) = res0;
-            if (d_x + 1 < dst.cols)
-                dst(d_y, d_x + 1) = res1;
-        }
-
-        template <int cn>
-        void Bayer2BGR_8u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream)
-        {
-            typedef typename TypeVec<uchar, cn>::vec_type dst_t;
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(dst.cols, 4 * block.x), divUp(dst.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(Bayer2BGR_8u<dst_t>, cudaFuncCachePreferL1) );
-
-            Bayer2BGR_8u<dst_t><<<grid, block, 0, stream>>>(src, (PtrStepSz<dst_t>)dst, blue_last, start_with_green);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-        template <int cn>
-        void Bayer2BGR_16u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream)
-        {
-            typedef typename TypeVec<ushort, cn>::vec_type dst_t;
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(dst.cols, 2 * block.x), divUp(dst.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(Bayer2BGR_16u<dst_t>, cudaFuncCachePreferL1) );
-
-            Bayer2BGR_16u<dst_t><<<grid, block, 0, stream>>>(src, (PtrStepSz<dst_t>)dst, blue_last, start_with_green);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        template void Bayer2BGR_8u_gpu<3>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-        template void Bayer2BGR_8u_gpu<4>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-        template void Bayer2BGR_16u_gpu<3>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-        template void Bayer2BGR_16u_gpu<4>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
-    }
-}}
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/cuda/element_operations.cu b/modules/gpu/src/cuda/element_operations.cu
deleted file mode 100644
index 5165b352ae..0000000000
--- a/modules/gpu/src/cuda/element_operations.cu
+++ /dev/null
@@ -1,3074 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/transform.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-
-using namespace cv::gpu;
-using namespace cv::gpu::device;
-
-namespace arithm
-{
-    template <size_t src_size, size_t dst_size> struct ArithmFuncTraits
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 1 };
-    };
-
-    template <> struct ArithmFuncTraits<1, 1>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<1, 2>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<1, 4>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-
-    template <> struct ArithmFuncTraits<2, 1>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<2, 2>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<2, 4>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-
-    template <> struct ArithmFuncTraits<4, 1>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<4, 2>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-    template <> struct ArithmFuncTraits<4, 4>
-    {
-        enum { simple_block_dim_x = 32 };
-        enum { simple_block_dim_y = 8 };
-
-        enum { smart_block_dim_x = 32 };
-        enum { smart_block_dim_y = 8 };
-        enum { smart_shift = 4 };
-    };
-}
-
-//////////////////////////////////////////////////////////////////////////
-// addMat
-
-namespace arithm
-{
-    template <typename T, typename D> struct VAdd4;
-    template <> struct VAdd4<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd4() {}
-        __device__ __forceinline__ VAdd4(const VAdd4<uint, uint>& other) {}
-    };
-    template <> struct VAdd4<int, uint> : binary_function<int, int, uint>
-    {
-        __device__ __forceinline__ uint operator ()(int a, int b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd4.u32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.u32.s32.s32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.s32.s32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.s32.s32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.s32.s32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd4() {}
-        __device__ __forceinline__ VAdd4(const VAdd4<int, uint>& other) {}
-    };
-    template <> struct VAdd4<uint, int> : binary_function<uint, uint, int>
-    {
-        __device__ __forceinline__ int operator ()(uint a, uint b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd4.s32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.s32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd4() {}
-        __device__ __forceinline__ VAdd4(const VAdd4<uint, int>& other) {}
-    };
-    template <> struct VAdd4<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd4.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.s32.s32.s32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.s32.s32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.s32.s32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.s32.s32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd4() {}
-        __device__ __forceinline__ VAdd4(const VAdd4<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct VAdd2;
-    template <> struct VAdd2<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd2() {}
-        __device__ __forceinline__ VAdd2(const VAdd2<uint, uint>& other) {}
-    };
-    template <> struct VAdd2<uint, int> : binary_function<uint, uint, int>
-    {
-        __device__ __forceinline__ int operator ()(uint a, uint b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd2.s32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.s32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd2() {}
-        __device__ __forceinline__ VAdd2(const VAdd2<uint, int>& other) {}
-    };
-    template <> struct VAdd2<int, uint> : binary_function<int, int, uint>
-    {
-        __device__ __forceinline__ uint operator ()(int a, int b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd2.u32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.u32.s32.s32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.u32.s32.s32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd2() {}
-        __device__ __forceinline__ VAdd2(const VAdd2<int, uint>& other) {}
-    };
-    template <> struct VAdd2<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vadd2.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vadd.s32.s32.s32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vadd.s32.s32.s32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAdd2() {}
-        __device__ __forceinline__ VAdd2(const VAdd2<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct AddMat : binary_function<T, T, D>
-    {
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return saturate_cast<D>(a + b);
-        }
-
-        __device__ __forceinline__ AddMat() {}
-        __device__ __forceinline__ AddMat(const AddMat& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VAdd4<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VAdd2<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::AddMat<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename D>
-    void vadd4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, VAdd4<T, D>(), WithOutMask(), stream);
-    }
-
-    template void vadd4<uint, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd4<uint, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd4<int, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd4<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void vadd2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, VAdd2<T, D>(), WithOutMask(), stream);
-    }
-
-    template void vadd2<uint, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd2<uint, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd2<int, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vadd2<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void addMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, AddMat<T, D>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, AddMat<T, D>(), WithOutMask(), stream);
-    }
-
-    template void addMat<uchar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<uchar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void addMat<schar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<schar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addMat<ushort, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<ushort, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<ushort, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<ushort, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<ushort, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<ushort, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<ushort, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addMat<short, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<short, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<short, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<short, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<short, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<short, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<short, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addMat<int, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<int, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<int, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<int, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<int, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<int, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addMat<float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<float, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addMat<double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addMat<double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addMat<double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// addScalar
-
-namespace arithm
-{
-    template <typename T, typename S, typename D> struct AddScalar : unary_function<T, D>
-    {
-        S val;
-
-        explicit AddScalar(S val_) : val(val_) {}
-
-        __device__ __forceinline__ D operator ()(T a) const
-        {
-            return saturate_cast<D>(a + val);
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::AddScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename S, typename D>
-    void addScalar(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        AddScalar<T, S, D> op(static_cast<S>(val));
-
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void addScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void addScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void addScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void addScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void addScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// subMat
-
-namespace arithm
-{
-    template <typename T, typename D> struct VSub4;
-    template <> struct VSub4<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub4() {}
-        __device__ __forceinline__ VSub4(const VSub4<uint, uint>& other) {}
-    };
-    template <> struct VSub4<int, uint> : binary_function<int, int, uint>
-    {
-        __device__ __forceinline__ uint operator ()(int a, int b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub4.u32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.u32.s32.s32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.s32.s32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.s32.s32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.s32.s32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub4() {}
-        __device__ __forceinline__ VSub4(const VSub4<int, uint>& other) {}
-    };
-    template <> struct VSub4<uint, int> : binary_function<uint, uint, int>
-    {
-        __device__ __forceinline__ int operator ()(uint a, uint b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub4.s32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.s32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub4() {}
-        __device__ __forceinline__ VSub4(const VSub4<uint, int>& other) {}
-    };
-    template <> struct VSub4<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub4.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.s32.s32.s32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.s32.s32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.s32.s32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.s32.s32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub4() {}
-        __device__ __forceinline__ VSub4(const VSub4<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct VSub2;
-    template <> struct VSub2<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub2() {}
-        __device__ __forceinline__ VSub2(const VSub2<uint, uint>& other) {}
-    };
-    template <> struct VSub2<uint, int> : binary_function<uint, uint, int>
-    {
-        __device__ __forceinline__ int operator ()(uint a, uint b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub2.s32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.s32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub2() {}
-        __device__ __forceinline__ VSub2(const VSub2<uint, int>& other) {}
-    };
-    template <> struct VSub2<int, uint> : binary_function<int, int, uint>
-    {
-        __device__ __forceinline__ uint operator ()(int a, int b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub2.u32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.u32.s32.s32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.u32.s32.s32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub2() {}
-        __device__ __forceinline__ VSub2(const VSub2<int, uint>& other) {}
-    };
-    template <> struct VSub2<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vsub2.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vsub.s32.s32.s32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vsub.s32.s32.s32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VSub2() {}
-        __device__ __forceinline__ VSub2(const VSub2<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct SubMat : binary_function<T, T, D>
-    {
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return saturate_cast<D>(a - b);
-        }
-
-        __device__ __forceinline__ SubMat() {}
-        __device__ __forceinline__ SubMat(const SubMat& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VSub4<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VSub2<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::SubMat<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename D>
-    void vsub4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, VSub4<T, D>(), WithOutMask(), stream);
-    }
-
-    template void vsub4<uint, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub4<uint, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub4<int, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub4<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void vsub2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, VSub2<T, D>(), WithOutMask(), stream);
-    }
-
-    template void vsub2<uint, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub2<uint, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub2<int, uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vsub2<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void subMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, SubMat<T, D>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, SubMat<T, D>(), WithOutMask(), stream);
-    }
-
-    template void subMat<uchar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<uchar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void subMat<schar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<schar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subMat<ushort, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<ushort, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<ushort, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<ushort, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<ushort, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<ushort, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<ushort, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subMat<short, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<short, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<short, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<short, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<short, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<short, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<short, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subMat<int, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<int, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<int, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<int, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<int, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<int, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subMat<float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<float, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subMat<double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subMat<double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subMat<double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// subScalar
-
-namespace arithm
-{
-    template <typename T, typename S, typename D>
-    void subScalar(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        AddScalar<T, S, D> op(-static_cast<S>(val));
-
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void subScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void subScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    //template void subScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    //template void subScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void subScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// mulMat
-
-namespace arithm
-{
-    struct Mul_8uc4_32f : binary_function<uint, float, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, float b) const
-        {
-            uint res = 0;
-
-            res |= (saturate_cast<uchar>((0xffu & (a      )) * b)      );
-            res |= (saturate_cast<uchar>((0xffu & (a >>  8)) * b) <<  8);
-            res |= (saturate_cast<uchar>((0xffu & (a >> 16)) * b) << 16);
-            res |= (saturate_cast<uchar>((0xffu & (a >> 24)) * b) << 24);
-
-            return res;
-        }
-
-        __device__ __forceinline__ Mul_8uc4_32f() {}
-        __device__ __forceinline__ Mul_8uc4_32f(const Mul_8uc4_32f& other) {}
-    };
-
-    struct Mul_16sc4_32f : binary_function<short4, float, short4>
-    {
-        __device__ __forceinline__ short4 operator ()(short4 a, float b) const
-        {
-            return make_short4(saturate_cast<short>(a.x * b), saturate_cast<short>(a.y * b),
-                               saturate_cast<short>(a.z * b), saturate_cast<short>(a.w * b));
-        }
-
-        __device__ __forceinline__ Mul_16sc4_32f() {}
-        __device__ __forceinline__ Mul_16sc4_32f(const Mul_16sc4_32f& other) {}
-    };
-
-    template <typename T, typename D> struct Mul : binary_function<T, T, D>
-    {
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return saturate_cast<D>(a * b);
-        }
-
-        __device__ __forceinline__ Mul() {}
-        __device__ __forceinline__ Mul(const Mul& other) {}
-    };
-
-    template <typename T, typename S, typename D> struct MulScale : binary_function<T, T, D>
-    {
-        S scale;
-
-        explicit MulScale(S scale_) : scale(scale_) {}
-
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return saturate_cast<D>(scale * a * b);
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <> struct TransformFunctorTraits<arithm::Mul_8uc4_32f> : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
-    {
-    };
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::Mul<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::MulScale<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    void mulMat_8uc4_32f(PtrStepSz<uint> src1, PtrStepSzf src2, PtrStepSz<uint> dst, cudaStream_t stream)
-    {
-        transform(src1, src2, dst, Mul_8uc4_32f(), WithOutMask(), stream);
-    }
-
-    void mulMat_16sc4_32f(PtrStepSz<short4> src1, PtrStepSzf src2, PtrStepSz<short4> dst, cudaStream_t stream)
-    {
-        transform(src1, src2, dst, Mul_16sc4_32f(), WithOutMask(), stream);
-    }
-
-    template <typename T, typename S, typename D>
-    void mulMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream)
-    {
-        if (scale == 1)
-        {
-            Mul<T, D> op;
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-        }
-        else
-        {
-            MulScale<T, S, D> op(static_cast<S>(scale));
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-        }
-    }
-
-    template void mulMat<uchar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<uchar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    template void mulMat<schar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<schar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void mulMat<ushort, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<ushort, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<ushort, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<ushort, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<ushort, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<ushort, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<ushort, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void mulMat<short, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<short, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<short, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<short, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<short, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<short, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<short, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void mulMat<int, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<int, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<int, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<int, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<int, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<int, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<int, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void mulMat<float, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<float, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<float, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<float, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<float, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<float, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<float, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void mulMat<double, double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<double, double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<double, double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<double, double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<double, double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void mulMat<double, double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void mulMat<double, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// mulScalar
-
-namespace arithm
-{
-    template <typename T, typename S, typename D> struct MulScalar : unary_function<T, D>
-    {
-        S val;
-
-        explicit MulScalar(S val_) : val(val_) {}
-
-        __device__ __forceinline__ D operator ()(T a) const
-        {
-            return saturate_cast<D>(a * val);
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::MulScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename S, typename D>
-    void mulScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
-    {
-        MulScalar<T, S, D> op(static_cast<S>(val));
-        transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void mulScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    template void mulScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void mulScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void mulScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void mulScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void mulScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void mulScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void mulScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void mulScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// divMat
-
-namespace arithm
-{
-    struct Div_8uc4_32f : binary_function<uint, float, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, float b) const
-        {
-            uint res = 0;
-
-            if (b != 0)
-            {
-                b = 1.0f / b;
-                res |= (saturate_cast<uchar>((0xffu & (a      )) * b)      );
-                res |= (saturate_cast<uchar>((0xffu & (a >>  8)) * b) <<  8);
-                res |= (saturate_cast<uchar>((0xffu & (a >> 16)) * b) << 16);
-                res |= (saturate_cast<uchar>((0xffu & (a >> 24)) * b) << 24);
-            }
-
-            return res;
-        }
-    };
-
-    struct Div_16sc4_32f : binary_function<short4, float, short4>
-    {
-        __device__ __forceinline__ short4 operator ()(short4 a, float b) const
-        {
-            return b != 0 ? make_short4(saturate_cast<short>(a.x / b), saturate_cast<short>(a.y / b),
-                                        saturate_cast<short>(a.z / b), saturate_cast<short>(a.w / b))
-                          : make_short4(0,0,0,0);
-        }
-    };
-
-    template <typename T, typename D> struct Div : binary_function<T, T, D>
-    {
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return b != 0 ? saturate_cast<D>(a / b) : 0;
-        }
-
-        __device__ __forceinline__ Div() {}
-        __device__ __forceinline__ Div(const Div& other) {}
-    };
-    template <typename T> struct Div<T, float> : binary_function<T, T, float>
-    {
-        __device__ __forceinline__ float operator ()(T a, T b) const
-        {
-            return b != 0 ? static_cast<float>(a) / b : 0;
-        }
-
-        __device__ __forceinline__ Div() {}
-        __device__ __forceinline__ Div(const Div& other) {}
-    };
-    template <typename T> struct Div<T, double> : binary_function<T, T, double>
-    {
-        __device__ __forceinline__ double operator ()(T a, T b) const
-        {
-            return b != 0 ? static_cast<double>(a) / b : 0;
-        }
-
-        __device__ __forceinline__ Div() {}
-        __device__ __forceinline__ Div(const Div& other) {}
-    };
-
-    template <typename T, typename S, typename D> struct DivScale : binary_function<T, T, D>
-    {
-        S scale;
-
-        explicit DivScale(S scale_) : scale(scale_) {}
-
-        __device__ __forceinline__ D operator ()(T a, T b) const
-        {
-            return b != 0 ? saturate_cast<D>(scale * a / b) : 0;
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <> struct TransformFunctorTraits<arithm::Div_8uc4_32f> : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
-    {
-    };
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::Div<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::DivScale<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    void divMat_8uc4_32f(PtrStepSz<uint> src1, PtrStepSzf src2, PtrStepSz<uint> dst, cudaStream_t stream)
-    {
-        transform(src1, src2, dst, Div_8uc4_32f(), WithOutMask(), stream);
-    }
-
-    void divMat_16sc4_32f(PtrStepSz<short4> src1, PtrStepSzf src2, PtrStepSz<short4> dst, cudaStream_t stream)
-    {
-        transform(src1, src2, dst, Div_16sc4_32f(), WithOutMask(), stream);
-    }
-
-    template <typename T, typename S, typename D>
-    void divMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream)
-    {
-        if (scale == 1)
-        {
-            Div<T, D> op;
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-        }
-        else
-        {
-            DivScale<T, S, D> op(static_cast<S>(scale));
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-        }
-    }
-
-    template void divMat<uchar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<uchar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    template void divMat<schar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<schar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void divMat<ushort, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<ushort, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<ushort, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<ushort, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<ushort, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<ushort, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<ushort, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void divMat<short, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<short, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<short, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<short, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<short, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<short, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<short, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void divMat<int, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<int, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<int, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<int, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<int, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<int, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<int, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void divMat<float, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<float, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<float, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<float, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<float, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<float, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<float, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-
-    //template void divMat<double, double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<double, double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<double, double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<double, double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<double, double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    //template void divMat<double, double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-    template void divMat<double, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// divScalar
-
-namespace arithm
-{
-    template <typename T, typename S, typename D>
-    void divScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
-    {
-        MulScalar<T, S, D> op(static_cast<S>(1.0 / val));
-        transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void divScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    template void divScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// divInv
-
-namespace arithm
-{
-    template <typename T, typename S, typename D> struct DivInv : unary_function<T, D>
-    {
-        S val;
-
-        explicit DivInv(S val_) : val(val_) {}
-
-        __device__ __forceinline__ D operator ()(T a) const
-        {
-            return a != 0 ? saturate_cast<D>(val / a) : 0;
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::DivInv<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename S, typename D>
-    void divInv(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
-    {
-        DivInv<T, S, D> op(static_cast<S>(val));
-        transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void divInv<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    template void divInv<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divInv<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divInv<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divInv<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divInv<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-
-    //template void divInv<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    //template void divInv<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    template void divInv<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// absDiffMat
-
-namespace arithm
-{
-    template <typename T, typename D> struct VAbsDiff4;
-    template <> struct VAbsDiff4<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vabsdiff4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vabsdiff.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAbsDiff4() {}
-        __device__ __forceinline__ VAbsDiff4(const VAbsDiff4<uint, uint>& other) {}
-    };
-    template <> struct VAbsDiff4<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vabsdiff4.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vabsdiff.s32.s32.s32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.s32.s32.s32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.s32.s32.s32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.s32.s32.s32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAbsDiff4() {}
-        __device__ __forceinline__ VAbsDiff4(const VAbsDiff4<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct VAbsDiff2;
-    template <> struct VAbsDiff2<uint, uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vabsdiff2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vabsdiff.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAbsDiff2() {}
-        __device__ __forceinline__ VAbsDiff2(const VAbsDiff2<uint, uint>& other) {}
-    };
-    template <> struct VAbsDiff2<int, int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vabsdiff2.s32.s32.s32.sat %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vabsdiff.s32.s32.s32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vabsdiff.s32.s32.s32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VAbsDiff2() {}
-        __device__ __forceinline__ VAbsDiff2(const VAbsDiff2<int, int>& other) {}
-    };
-
-    ////////////////////////////////////
-
-    __device__ __forceinline__ int _abs(int a)
-    {
-        return ::abs(a);
-    }
-    __device__ __forceinline__ float _abs(float a)
-    {
-        return ::fabsf(a);
-    }
-    __device__ __forceinline__ double _abs(double a)
-    {
-        return ::fabs(a);
-    }
-
-    template <typename T> struct AbsDiffMat : binary_function<T, T, T>
-    {
-        __device__ __forceinline__ T operator ()(T a, T b) const
-        {
-            return saturate_cast<T>(_abs(a - b));
-        }
-
-        __device__ __forceinline__ AbsDiffMat() {}
-        __device__ __forceinline__ AbsDiffMat(const AbsDiffMat& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VAbsDiff4<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T, typename D> struct TransformFunctorTraits< arithm::VAbsDiff2<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct TransformFunctorTraits< arithm::AbsDiffMat<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void vabsDiff4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VAbsDiff4<T, T>(), WithOutMask(), stream);
-    }
-
-    template void vabsDiff4<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vabsDiff4<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T>
-    void vabsDiff2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VAbsDiff2<T, T>(), WithOutMask(), stream);
-    }
-
-    template void vabsDiff2<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vabsDiff2<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T>
-    void absDiffMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, AbsDiffMat<T>(), WithOutMask(), stream);
-    }
-
-    template void absDiffMat<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// absDiffScalar
-
-namespace arithm
-{
-    template <typename T, typename S> struct AbsDiffScalar : unary_function<T, T>
-    {
-        S val;
-
-        explicit AbsDiffScalar(S val_) : val(val_) {}
-
-        __device__ __forceinline__ T operator ()(T a) const
-        {
-            abs_func<S> f;
-            return saturate_cast<T>(f(a - val));
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T, typename S> struct TransformFunctorTraits< arithm::AbsDiffScalar<T, S> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T, typename S>
-    void absDiffScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
-    {
-        AbsDiffScalar<T, S> op(static_cast<S>(val));
-
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, op, WithOutMask(), stream);
-    }
-
-    template void absDiffScalar<uchar, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<schar, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<ushort, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<short, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<int, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<float, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void absDiffScalar<double, double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// absMat
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< abs_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void absMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, abs_func<T>(), WithOutMask(), stream);
-    }
-
-    template void absMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void absMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// sqrMat
-
-namespace arithm
-{
-    template <typename T> struct Sqr : unary_function<T, T>
-    {
-        __device__ __forceinline__ T operator ()(T x) const
-        {
-            return saturate_cast<T>(x * x);
-        }
-
-        __device__ __forceinline__ Sqr() {}
-        __device__ __forceinline__ Sqr(const Sqr& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< arithm::Sqr<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void sqrMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, Sqr<T>(), WithOutMask(), stream);
-    }
-
-    template void sqrMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// sqrtMat
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< sqrt_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void sqrtMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, sqrt_func<T>(), WithOutMask(), stream);
-    }
-
-    template void sqrtMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void sqrtMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// logMat
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< log_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void logMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, log_func<T>(), WithOutMask(), stream);
-    }
-
-    template void logMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void logMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// expMat
-
-namespace arithm
-{
-    template <typename T> struct Exp : unary_function<T, T>
-    {
-        __device__ __forceinline__ T operator ()(T x) const
-        {
-            exp_func<T> f;
-            return saturate_cast<T>(f(x));
-        }
-
-        __device__ __forceinline__ Exp() {}
-        __device__ __forceinline__ Exp(const Exp& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< arithm::Exp<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T>
-    void expMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, Exp<T>(), WithOutMask(), stream);
-    }
-
-    template void expMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    template void expMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////////////////
-// cmpMat
-
-namespace arithm
-{
-    template <class Op, typename T>
-    struct Cmp : binary_function<T, T, uchar>
-    {
-        __device__ __forceinline__ uchar operator()(T a, T b) const
-        {
-            Op op;
-            return -op(a, b);
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <class Op, typename T> struct TransformFunctorTraits< arithm::Cmp<Op, T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(uchar)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <template <typename> class Op, typename T>
-    void cmpMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        Cmp<Op<T>, T> op;
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, dst, op, WithOutMask(), stream);
-    }
-
-    template <typename T> void cmpMatEq(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        cmpMat<equal_to, T>(src1, src2, dst, stream);
-    }
-    template <typename T> void cmpMatNe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        cmpMat<not_equal_to, T>(src1, src2, dst, stream);
-    }
-    template <typename T> void cmpMatLt(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        cmpMat<less, T>(src1, src2, dst, stream);
-    }
-    template <typename T> void cmpMatLe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        cmpMat<less_equal, T>(src1, src2, dst, stream);
-    }
-
-    template void cmpMatEq<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatEq<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpMatNe<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatNe<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpMatLt<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLt<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpMatLe<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void cmpMatLe<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////////////////
-// cmpScalar
-
-namespace arithm
-{
-#define TYPE_VEC(type, cn) typename TypeVec<type, cn>::vec_type
-
-    template <class Op, typename T, int cn> struct CmpScalar;
-    template <class Op, typename T>
-    struct CmpScalar<Op, T, 1> : unary_function<T, uchar>
-    {
-        const T val;
-
-        __host__ explicit CmpScalar(T val_) : val(val_) {}
-
-        __device__ __forceinline__ uchar operator()(T src) const
-        {
-            Cmp<Op, T> op;
-            return op(src, val);
-        }
-    };
-    template <class Op, typename T>
-    struct CmpScalar<Op, T, 2> : unary_function<TYPE_VEC(T, 2), TYPE_VEC(uchar, 2)>
-    {
-        const TYPE_VEC(T, 2) val;
-
-        __host__ explicit CmpScalar(TYPE_VEC(T, 2) val_) : val(val_) {}
-
-        __device__ __forceinline__ TYPE_VEC(uchar, 2) operator()(const TYPE_VEC(T, 2) & src) const
-        {
-            Cmp<Op, T> op;
-            return VecTraits<TYPE_VEC(uchar, 2)>::make(op(src.x, val.x), op(src.y, val.y));
-        }
-    };
-    template <class Op, typename T>
-    struct CmpScalar<Op, T, 3> : unary_function<TYPE_VEC(T, 3), TYPE_VEC(uchar, 3)>
-    {
-        const TYPE_VEC(T, 3) val;
-
-        __host__ explicit CmpScalar(TYPE_VEC(T, 3) val_) : val(val_) {}
-
-        __device__ __forceinline__ TYPE_VEC(uchar, 3) operator()(const TYPE_VEC(T, 3) & src) const
-        {
-            Cmp<Op, T> op;
-            return VecTraits<TYPE_VEC(uchar, 3)>::make(op(src.x, val.x), op(src.y, val.y), op(src.z, val.z));
-        }
-    };
-    template <class Op, typename T>
-    struct CmpScalar<Op, T, 4> : unary_function<TYPE_VEC(T, 4), TYPE_VEC(uchar, 4)>
-    {
-        const TYPE_VEC(T, 4) val;
-
-        __host__ explicit CmpScalar(TYPE_VEC(T, 4) val_) : val(val_) {}
-
-        __device__ __forceinline__ TYPE_VEC(uchar, 4) operator()(const TYPE_VEC(T, 4) & src) const
-        {
-            Cmp<Op, T> op;
-            return VecTraits<TYPE_VEC(uchar, 4)>::make(op(src.x, val.x), op(src.y, val.y), op(src.z, val.z), op(src.w, val.w));
-        }
-    };
-
-#undef TYPE_VEC
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <class Op, typename T> struct TransformFunctorTraits< arithm::CmpScalar<Op, T, 1> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(uchar)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <template <typename> class Op, typename T, int cn>
-    void cmpScalar(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef typename TypeVec<T, cn>::vec_type src_t;
-        typedef typename TypeVec<uchar, cn>::vec_type dst_t;
-
-        T sval[] = {static_cast<T>(val[0]), static_cast<T>(val[1]), static_cast<T>(val[2]), static_cast<T>(val[3])};
-        src_t val1 = VecTraits<src_t>::make(sval);
-
-        CmpScalar<Op<T>, T, cn> op(val1);
-        transform((PtrStepSz<src_t>) src, (PtrStepSz<dst_t>) dst, op, WithOutMask(), stream);
-    }
-
-    template <typename T> void cmpScalarEq(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<equal_to, T, 1>,
-            cmpScalar<equal_to, T, 2>,
-            cmpScalar<equal_to, T, 3>,
-            cmpScalar<equal_to, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-    template <typename T> void cmpScalarNe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<not_equal_to, T, 1>,
-            cmpScalar<not_equal_to, T, 2>,
-            cmpScalar<not_equal_to, T, 3>,
-            cmpScalar<not_equal_to, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-    template <typename T> void cmpScalarLt(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<less, T, 1>,
-            cmpScalar<less, T, 2>,
-            cmpScalar<less, T, 3>,
-            cmpScalar<less, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-    template <typename T> void cmpScalarLe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<less_equal, T, 1>,
-            cmpScalar<less_equal, T, 2>,
-            cmpScalar<less_equal, T, 3>,
-            cmpScalar<less_equal, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-    template <typename T> void cmpScalarGt(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<greater, T, 1>,
-            cmpScalar<greater, T, 2>,
-            cmpScalar<greater, T, 3>,
-            cmpScalar<greater, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-    template <typename T> void cmpScalarGe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            0,
-            cmpScalar<greater_equal, T, 1>,
-            cmpScalar<greater_equal, T, 2>,
-            cmpScalar<greater_equal, T, 3>,
-            cmpScalar<greater_equal, T, 4>
-        };
-
-        funcs[cn](src, val, dst, stream);
-    }
-
-    template void cmpScalarEq<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarEq<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpScalarNe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarNe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpScalarLt<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLt<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpScalarLe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarLe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpScalarGt<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGt<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-
-    template void cmpScalarGe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-    template void cmpScalarGe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////////////////
-// bitMat
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< bit_not<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< bit_and<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< bit_or<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< bit_xor<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T> void bitMatNot(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, bit_not<T>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, bit_not<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void bitMatAnd(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_and<T>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_and<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void bitMatOr(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_or<T>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_or<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void bitMatXor(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
-    {
-        if (mask.data)
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_xor<T>(), mask, stream);
-        else
-            transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_xor<T>(), WithOutMask(), stream);
-    }
-
-    template void bitMatNot<uchar>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatNot<ushort>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatNot<uint>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void bitMatAnd<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatAnd<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatAnd<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void bitMatOr<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatOr<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatOr<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-
-    template void bitMatXor<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatXor<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-    template void bitMatXor<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////////////////
-// bitScalar
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< binder2nd< bit_and<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< binder2nd< bit_or<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< binder2nd< bit_xor<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T> void bitScalarAnd(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::device::bind2nd(bit_and<T>(), src2), WithOutMask(), stream);
-    }
-
-    template <typename T> void bitScalarOr(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::device::bind2nd(bit_or<T>(), src2), WithOutMask(), stream);
-    }
-
-    template <typename T> void bitScalarXor(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::device::bind2nd(bit_xor<T>(), src2), WithOutMask(), stream);
-    }
-
-    template void bitScalarAnd<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarAnd<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarAnd<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarAnd<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void bitScalarOr<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarOr<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarOr<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarOr<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void bitScalarXor<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarXor<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarXor<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-    template void bitScalarXor<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// min
-
-namespace arithm
-{
-    template <typename T> struct VMin4;
-    template <> struct VMin4<uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmin4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmin.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMin4() {}
-        __device__ __forceinline__ VMin4(const VMin4& other) {}
-    };
-    template <> struct VMin4<int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmin4.s32.s32.s32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmin.s32.s32.s32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.s32.s32.s32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.s32.s32.s32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.s32.s32.s32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMin4() {}
-        __device__ __forceinline__ VMin4(const VMin4& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct VMin2;
-    template <> struct VMin2<uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmin2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmin.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMin2() {}
-        __device__ __forceinline__ VMin2(const VMin2& other) {}
-    };
-    template <> struct VMin2<int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmin2.s32.s32.s32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmin.s32.s32.s32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmin.s32.s32.s32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMin2() {}
-        __device__ __forceinline__ VMin2(const VMin2& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< arithm::VMin4<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct TransformFunctorTraits< arithm::VMin2<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct TransformFunctorTraits< minimum<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< binder2nd< minimum<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T> void vmin4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VMin4<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void vmin2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VMin2<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void minMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, minimum<T>(), WithOutMask(), stream);
-    }
-
-    template void vmin4<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vmin4<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void vmin2<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vmin2<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void minMat<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T> void minScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::device::bind2nd(minimum<T>(), src2), WithOutMask(), stream);
-    }
-
-    template void minScalar<uchar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<schar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<ushort>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<short >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<int   >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<float >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void minScalar<double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// max
-
-namespace arithm
-{
-    template <typename T> struct VMax4;
-    template <> struct VMax4<uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmax4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmax.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMax4() {}
-        __device__ __forceinline__ VMax4(const VMax4& other) {}
-    };
-    template <> struct VMax4<int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmax4.s32.s32.s32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmax.s32.s32.s32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.s32.s32.s32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.s32.s32.s32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.s32.s32.s32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMax4() {}
-        __device__ __forceinline__ VMax4(const VMax4& other) {}
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct VMax2;
-    template <> struct VMax2<uint> : binary_function<uint, uint, uint>
-    {
-        __device__ __forceinline__ uint operator ()(uint a, uint b) const
-        {
-            uint res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmax2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmax.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMax2() {}
-        __device__ __forceinline__ VMax2(const VMax2& other) {}
-    };
-    template <> struct VMax2<int> : binary_function<int, int, int>
-    {
-        __device__ __forceinline__ int operator ()(int a, int b) const
-        {
-            int res = 0;
-
-        #if __CUDA_ARCH__ >= 300
-            asm("vmax2.s32.s32.s32 %0, %1, %2, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #elif __CUDA_ARCH__ >= 200
-            asm("vmax.s32.s32.s32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-            asm("vmax.s32.s32.s32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(res) : "r"(a), "r"(b), "r"(res));
-        #endif
-
-            return res;
-        }
-
-        __device__ __forceinline__ VMax2() {}
-        __device__ __forceinline__ VMax2(const VMax2& other) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< arithm::VMax4<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct TransformFunctorTraits< arithm::VMax2<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    ////////////////////////////////////
-
-    template <typename T> struct TransformFunctorTraits< maximum<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< binder2nd< maximum<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T> void vmax4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VMax4<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void vmax2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, VMax2<T>(), WithOutMask(), stream);
-    }
-
-    template <typename T> void maxMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, maximum<T>(), WithOutMask(), stream);
-    }
-
-    template void vmax4<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vmax4<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void vmax2<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void vmax2<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template void maxMat<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T> void maxScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::device::bind2nd(maximum<T>(), src2), WithOutMask(), stream);
-    }
-
-    template void maxScalar<uchar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<schar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<ushort>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<short >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<int   >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<float >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    template void maxScalar<double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// threshold
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< thresh_binary_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< thresh_binary_inv_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< thresh_trunc_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< thresh_to_zero_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-
-    template <typename T> struct TransformFunctorTraits< thresh_to_zero_inv_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <template <typename> class Op, typename T>
-    void threshold_caller(PtrStepSz<T> src, PtrStepSz<T> dst, T thresh, T maxVal, cudaStream_t stream)
-    {
-        Op<T> op(thresh, maxVal);
-        transform(src, dst, op, WithOutMask(), stream);
-    }
-
-    template <typename T>
-    void threshold(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream)
-    {
-        typedef void (*caller_t)(PtrStepSz<T> src, PtrStepSz<T> dst, T thresh, T maxVal, cudaStream_t stream);
-
-        static const caller_t callers[] =
-        {
-            threshold_caller<thresh_binary_func, T>,
-            threshold_caller<thresh_binary_inv_func, T>,
-            threshold_caller<thresh_trunc_func, T>,
-            threshold_caller<thresh_to_zero_func, T>,
-            threshold_caller<thresh_to_zero_inv_func, T>
-        };
-
-        callers[type]((PtrStepSz<T>) src, (PtrStepSz<T>) dst, static_cast<T>(thresh), static_cast<T>(maxVal), stream);
-    }
-
-    template void threshold<uchar>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<schar>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<ushort>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<short>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<int>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<float>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-    template void threshold<double>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// pow
-
-namespace arithm
-{
-    template<typename T, bool Signed = numeric_limits<T>::is_signed> struct PowOp : unary_function<T, T>
-    {
-        float power;
-
-        PowOp(double power_) : power(static_cast<float>(power_)) {}
-
-        __device__ __forceinline__ T operator()(T e) const
-        {
-            return saturate_cast<T>(__powf((float)e, power));
-        }
-    };
-    template<typename T> struct PowOp<T, true> : unary_function<T, T>
-    {
-        float power;
-
-        PowOp(double power_) : power(static_cast<float>(power_)) {}
-
-        __device__ __forceinline__ T operator()(T e) const
-        {
-            T res = saturate_cast<T>(__powf((float)e, power));
-
-            if ((e < 0) && (1 & static_cast<int>(power)))
-                res *= -1;
-
-            return res;
-        }
-    };
-    template<> struct PowOp<float> : unary_function<float, float>
-    {
-        const float power;
-
-        PowOp(double power_) : power(static_cast<float>(power_)) {}
-
-        __device__ __forceinline__ float operator()(float e) const
-        {
-            return __powf(::fabs(e), power);
-        }
-    };
-    template<> struct PowOp<double> : unary_function<double, double>
-    {
-        double power;
-
-        PowOp(double power_) : power(power_) {}
-
-        __device__ __forceinline__ double operator()(double e) const
-        {
-            return ::pow(::fabs(e), power);
-        }
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T> struct TransformFunctorTraits< arithm::PowOp<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template<typename T>
-    void pow(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream)
-    {
-        transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, PowOp<T>(power), WithOutMask(), stream);
-    }
-
-    template void pow<uchar>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<schar>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<short>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<ushort>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<int>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<float>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    template void pow<double>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-}
-
-//////////////////////////////////////////////////////////////////////////
-// addWeighted
-
-namespace arithm
-{
-    template <typename T> struct UseDouble_
-    {
-        enum {value = 0};
-    };
-    template <> struct UseDouble_<double>
-    {
-        enum {value = 1};
-    };
-    template <typename T1, typename T2, typename D> struct UseDouble
-    {
-        enum {value = (UseDouble_<T1>::value || UseDouble_<T2>::value || UseDouble_<D>::value)};
-    };
-
-    template <typename T1, typename T2, typename D, bool useDouble> struct AddWeighted_;
-    template <typename T1, typename T2, typename D> struct AddWeighted_<T1, T2, D, false> : binary_function<T1, T2, D>
-    {
-        float alpha;
-        float beta;
-        float gamma;
-
-        AddWeighted_(double alpha_, double beta_, double gamma_) : alpha(static_cast<float>(alpha_)), beta(static_cast<float>(beta_)), gamma(static_cast<float>(gamma_)) {}
-
-        __device__ __forceinline__ D operator ()(T1 a, T2 b) const
-        {
-            return saturate_cast<D>(a * alpha + b * beta + gamma);
-        }
-    };
-    template <typename T1, typename T2, typename D> struct AddWeighted_<T1, T2, D, true> : binary_function<T1, T2, D>
-    {
-        double alpha;
-        double beta;
-        double gamma;
-
-        AddWeighted_(double alpha_, double beta_, double gamma_) : alpha(alpha_), beta(beta_), gamma(gamma_) {}
-
-        __device__ __forceinline__ D operator ()(T1 a, T2 b) const
-        {
-            return saturate_cast<D>(a * alpha + b * beta + gamma);
-        }
-    };
-    template <typename T1, typename T2, typename D> struct AddWeighted : AddWeighted_<T1, T2, D, UseDouble<T1, T2, D>::value>
-    {
-        AddWeighted(double alpha_, double beta_, double gamma_) : AddWeighted_<T1, T2, D, UseDouble<T1, T2, D>::value>(alpha_, beta_, gamma_) {}
-    };
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    template <typename T1, typename T2, typename D, size_t src1_size, size_t src2_size, size_t dst_size> struct AddWeightedTraits : DefaultTransformFunctorTraits< arithm::AddWeighted<T1, T2, D> >
-    {
-    };
-    template <typename T1, typename T2, typename D, size_t src_size, size_t dst_size> struct AddWeightedTraits<T1, T2, D, src_size, src_size, dst_size> : arithm::ArithmFuncTraits<src_size, dst_size>
-    {
-    };
-
-    template <typename T1, typename T2, typename D> struct TransformFunctorTraits< arithm::AddWeighted<T1, T2, D> > : AddWeightedTraits<T1, T2, D, sizeof(T1), sizeof(T2), sizeof(D)>
-    {
-    };
-}}}
-
-namespace arithm
-{
-    template <typename T1, typename T2, typename D>
-    void addWeighted(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream)
-    {
-        AddWeighted<T1, T2, D> op(alpha, beta, gamma);
-
-        transform((PtrStepSz<T1>) src1, (PtrStepSz<T2>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
-    }
-
-    template void addWeighted<uchar, uchar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, uchar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, schar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, schar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<uchar, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<uchar, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<schar, schar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, schar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<schar, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<schar, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<schar, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<schar, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<schar, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<schar, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<ushort, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<ushort, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<ushort, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<ushort, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<ushort, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<ushort, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<short, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<short, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<short, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<short, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<short, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<int, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<int, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<int, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<int, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<float, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-    template void addWeighted<float, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<float, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-
-
-
-    template void addWeighted<double, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-    template void addWeighted<double, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
-}
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/global_motion.cu b/modules/gpu/src/cuda/global_motion.cu
index b62ca2c9fd..5685c6750c 100644
--- a/modules/gpu/src/cuda/global_motion.cu
+++ b/modules/gpu/src/cuda/global_motion.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -29,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -46,9 +45,9 @@
 #include <thrust/device_ptr.h>
 #include <thrust/remove.h>
 #include <thrust/functional.h>
-#include "opencv2/gpu/device/common.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device { namespace globmotion {
+namespace cv { namespace gpu { namespace cudev { namespace globmotion {
 
 __constant__ float cml[9];
 __constant__ float cmr[9];
diff --git a/modules/gpu/src/cuda/hog.cu b/modules/gpu/src/cuda/hog.cu
index 523e5bf645..48d656a744 100644
--- a/modules/gpu/src/cuda/hog.cu
+++ b/modules/gpu/src/cuda/hog.cu
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/warp_shuffle.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     // Other values are not supported
     #define CELL_WIDTH 8
@@ -316,7 +316,7 @@ namespace cv { namespace gpu { namespace device
             else if (nthreads == 512)
                 normalize_hists_kernel_many_blocks<512, nblocks><<<grid, threads>>>(block_hist_size, img_block_width, block_hists, threshold);
             else
-                cv::gpu::error("normalize_hists: histogram's size is too big, try to decrease number of bins", __FILE__, __LINE__, "normalize_hists");
+                CV_Error(cv::Error::StsBadArg, "normalize_hists: histogram's size is too big, try to decrease number of bins");
 
             cudaSafeCall( cudaGetLastError() );
 
@@ -808,7 +808,7 @@ namespace cv { namespace gpu { namespace device
         void resize_8UC1(const PtrStepSzb& src, PtrStepSzb dst) { resize_for_hog<uchar> (src, dst, resize8UC1_tex); }
         void resize_8UC4(const PtrStepSzb& src, PtrStepSzb dst) { resize_for_hog<uchar4>(src, dst, resize8UC4_tex); }
     } // namespace hog
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/hough.cu b/modules/gpu/src/cuda/hough.cu
deleted file mode 100644
index 695a47def4..0000000000
--- a/modules/gpu/src/cuda/hough.cu
+++ /dev/null
@@ -1,1709 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include <thrust/device_ptr.h>
-#include <thrust/sort.h>
-
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/emulation.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/dynamic_smem.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        __device__ int g_counter;
-
-        ////////////////////////////////////////////////////////////////////////
-        // buildPointList
-
-        template <int PIXELS_PER_THREAD>
-        __global__ void buildPointList(const PtrStepSzb src, unsigned int* list)
-        {
-            __shared__ unsigned int s_queues[4][32 * PIXELS_PER_THREAD];
-            __shared__ int s_qsize[4];
-            __shared__ int s_globStart[4];
-
-            const int x = blockIdx.x * blockDim.x * PIXELS_PER_THREAD + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (threadIdx.x == 0)
-                s_qsize[threadIdx.y] = 0;
-            __syncthreads();
-
-            if (y < src.rows)
-            {
-                // fill the queue
-                const uchar* srcRow = src.ptr(y);
-                for (int i = 0, xx = x; i < PIXELS_PER_THREAD && xx < src.cols; ++i, xx += blockDim.x)
-                {
-                    if (srcRow[xx])
-                    {
-                        const unsigned int val = (y << 16) | xx;
-                        const int qidx = Emulation::smem::atomicAdd(&s_qsize[threadIdx.y], 1);
-                        s_queues[threadIdx.y][qidx] = val;
-                    }
-                }
-            }
-
-            __syncthreads();
-
-            // let one thread reserve the space required in the global list
-            if (threadIdx.x == 0 && threadIdx.y == 0)
-            {
-                // find how many items are stored in each list
-                int totalSize = 0;
-                for (int i = 0; i < blockDim.y; ++i)
-                {
-                    s_globStart[i] = totalSize;
-                    totalSize += s_qsize[i];
-                }
-
-                // calculate the offset in the global list
-                const int globalOffset = atomicAdd(&g_counter, totalSize);
-                for (int i = 0; i < blockDim.y; ++i)
-                    s_globStart[i] += globalOffset;
-            }
-
-            __syncthreads();
-
-            // copy local queues to global queue
-            const int qsize = s_qsize[threadIdx.y];
-            int gidx = s_globStart[threadIdx.y] + threadIdx.x;
-            for(int i = threadIdx.x; i < qsize; i += blockDim.x, gidx += blockDim.x)
-                list[gidx] = s_queues[threadIdx.y][i];
-        }
-
-        int buildPointList_gpu(PtrStepSzb src, unsigned int* list)
-        {
-            const int PIXELS_PER_THREAD = 16;
-
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 4);
-            const dim3 grid(divUp(src.cols, block.x * PIXELS_PER_THREAD), divUp(src.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(buildPointList<PIXELS_PER_THREAD>, cudaFuncCachePreferShared) );
-
-            buildPointList<PIXELS_PER_THREAD><<<grid, block>>>(src, list);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // linesAccum
-
-        __global__ void linesAccumGlobal(const unsigned int* list, const int count, PtrStepi accum, const float irho, const float theta, const int numrho)
-        {
-            const int n = blockIdx.x;
-            const float ang = n * theta;
-
-            float sinVal;
-            float cosVal;
-            sincosf(ang, &sinVal, &cosVal);
-            sinVal *= irho;
-            cosVal *= irho;
-
-            const int shift = (numrho - 1) / 2;
-
-            int* accumRow = accum.ptr(n + 1);
-            for (int i = threadIdx.x; i < count; i += blockDim.x)
-            {
-                const unsigned int val = list[i];
-
-                const int x = (val & 0xFFFF);
-                const int y = (val >> 16) & 0xFFFF;
-
-                int r = __float2int_rn(x * cosVal + y * sinVal);
-                r += shift;
-
-                ::atomicAdd(accumRow + r + 1, 1);
-            }
-        }
-
-        __global__ void linesAccumShared(const unsigned int* list, const int count, PtrStepi accum, const float irho, const float theta, const int numrho)
-        {
-            int* smem = DynamicSharedMem<int>();
-
-            for (int i = threadIdx.x; i < numrho + 1; i += blockDim.x)
-                smem[i] = 0;
-
-            __syncthreads();
-
-            const int n = blockIdx.x;
-            const float ang = n * theta;
-
-            float sinVal;
-            float cosVal;
-            sincosf(ang, &sinVal, &cosVal);
-            sinVal *= irho;
-            cosVal *= irho;
-
-            const int shift = (numrho - 1) / 2;
-
-            for (int i = threadIdx.x; i < count; i += blockDim.x)
-            {
-                const unsigned int val = list[i];
-
-                const int x = (val & 0xFFFF);
-                const int y = (val >> 16) & 0xFFFF;
-
-                int r = __float2int_rn(x * cosVal + y * sinVal);
-                r += shift;
-
-                Emulation::smem::atomicAdd(&smem[r + 1], 1);
-            }
-
-            __syncthreads();
-
-            int* accumRow = accum.ptr(n + 1);
-            for (int i = threadIdx.x; i < numrho + 1; i += blockDim.x)
-                accumRow[i] = smem[i];
-        }
-
-        void linesAccum_gpu(const unsigned int* list, int count, PtrStepSzi accum, float rho, float theta, size_t sharedMemPerBlock, bool has20)
-        {
-            const dim3 block(has20 ? 1024 : 512);
-            const dim3 grid(accum.rows - 2);
-
-            size_t smemSize = (accum.cols - 1) * sizeof(int);
-
-            if (smemSize < sharedMemPerBlock - 1000)
-                linesAccumShared<<<grid, block, smemSize>>>(list, count, accum, 1.0f / rho, theta, accum.cols - 2);
-            else
-                linesAccumGlobal<<<grid, block>>>(list, count, accum, 1.0f / rho, theta, accum.cols - 2);
-
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // linesGetResult
-
-        __global__ void linesGetResult(const PtrStepSzi accum, float2* out, int* votes, const int maxSize, const float rho, const float theta, const int threshold, const int numrho)
-        {
-            const int r = blockIdx.x * blockDim.x + threadIdx.x;
-            const int n = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (r >= accum.cols - 2 || n >= accum.rows - 2)
-                return;
-
-            const int curVotes = accum(n + 1, r + 1);
-
-            if (curVotes > threshold &&
-                curVotes >  accum(n + 1, r) &&
-                curVotes >= accum(n + 1, r + 2) &&
-                curVotes >  accum(n, r + 1) &&
-                curVotes >= accum(n + 2, r + 1))
-            {
-                const float radius = (r - (numrho - 1) * 0.5f) * rho;
-                const float angle = n * theta;
-
-                const int ind = ::atomicAdd(&g_counter, 1);
-                if (ind < maxSize)
-                {
-                    out[ind] = make_float2(radius, angle);
-                    votes[ind] = curVotes;
-                }
-            }
-        }
-
-        int linesGetResult_gpu(PtrStepSzi accum, float2* out, int* votes, int maxSize, float rho, float theta, int threshold, bool doSort)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(linesGetResult, cudaFuncCachePreferL1) );
-
-            linesGetResult<<<grid, block>>>(accum, out, votes, maxSize, rho, theta, threshold, accum.cols - 2);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            if (doSort && totalCount > 0)
-            {
-                thrust::device_ptr<float2> outPtr(out);
-                thrust::device_ptr<int> votesPtr(votes);
-                thrust::sort_by_key(votesPtr, votesPtr + totalCount, outPtr, thrust::greater<int>());
-            }
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // houghLinesProbabilistic
-
-        texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_mask(false, cudaFilterModePoint, cudaAddressModeClamp);
-
-        __global__ void houghLinesProbabilistic(const PtrStepSzi accum,
-                                                int4* out, const int maxSize,
-                                                const float rho, const float theta,
-                                                const int lineGap, const int lineLength,
-                                                const int rows, const int cols)
-        {
-            const int r = blockIdx.x * blockDim.x + threadIdx.x;
-            const int n = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (r >= accum.cols - 2 || n >= accum.rows - 2)
-                return;
-
-            const int curVotes = accum(n + 1, r + 1);
-
-            if (curVotes >= lineLength &&
-                curVotes > accum(n, r) &&
-                curVotes > accum(n, r + 1) &&
-                curVotes > accum(n, r + 2) &&
-                curVotes > accum(n + 1, r) &&
-                curVotes > accum(n + 1, r + 2) &&
-                curVotes > accum(n + 2, r) &&
-                curVotes > accum(n + 2, r + 1) &&
-                curVotes > accum(n + 2, r + 2))
-            {
-                const float radius = (r - (accum.cols - 2 - 1) * 0.5f) * rho;
-                const float angle = n * theta;
-
-                float cosa;
-                float sina;
-                sincosf(angle, &sina, &cosa);
-
-                float2 p0 = make_float2(cosa * radius, sina * radius);
-                float2 dir = make_float2(-sina, cosa);
-
-                float2 pb[4] = {make_float2(-1, -1), make_float2(-1, -1), make_float2(-1, -1), make_float2(-1, -1)};
-                float a;
-
-                if (dir.x != 0)
-                {
-                    a = -p0.x / dir.x;
-                    pb[0].x = 0;
-                    pb[0].y = p0.y + a * dir.y;
-
-                    a = (cols - 1 - p0.x) / dir.x;
-                    pb[1].x = cols - 1;
-                    pb[1].y = p0.y + a * dir.y;
-                }
-                if (dir.y != 0)
-                {
-                    a = -p0.y / dir.y;
-                    pb[2].x = p0.x + a * dir.x;
-                    pb[2].y = 0;
-
-                    a = (rows - 1 - p0.y) / dir.y;
-                    pb[3].x = p0.x + a * dir.x;
-                    pb[3].y = rows - 1;
-                }
-
-                if (pb[0].x == 0 && (pb[0].y >= 0 && pb[0].y < rows))
-                {
-                    p0 = pb[0];
-                    if (dir.x < 0)
-                        dir = -dir;
-                }
-                else if (pb[1].x == cols - 1 && (pb[0].y >= 0 && pb[0].y < rows))
-                {
-                    p0 = pb[1];
-                    if (dir.x > 0)
-                        dir = -dir;
-                }
-                else if (pb[2].y == 0 && (pb[2].x >= 0 && pb[2].x < cols))
-                {
-                    p0 = pb[2];
-                    if (dir.y < 0)
-                        dir = -dir;
-                }
-                else if (pb[3].y == rows - 1 && (pb[3].x >= 0 && pb[3].x < cols))
-                {
-                    p0 = pb[3];
-                    if (dir.y > 0)
-                        dir = -dir;
-                }
-
-                float2 d;
-                if (::fabsf(dir.x) > ::fabsf(dir.y))
-                {
-                    d.x = dir.x > 0 ? 1 : -1;
-                    d.y = dir.y / ::fabsf(dir.x);
-                }
-                else
-                {
-                    d.x = dir.x / ::fabsf(dir.y);
-                    d.y = dir.y > 0 ? 1 : -1;
-                }
-
-                float2 line_end[2];
-                int gap;
-                bool inLine = false;
-
-                float2 p1 = p0;
-                if (p1.x < 0 || p1.x >= cols || p1.y < 0 || p1.y >= rows)
-                    return;
-
-                for (;;)
-                {
-                    if (tex2D(tex_mask, p1.x, p1.y))
-                    {
-                        gap = 0;
-
-                        if (!inLine)
-                        {
-                            line_end[0] = p1;
-                            line_end[1] = p1;
-                            inLine = true;
-                        }
-                        else
-                        {
-                            line_end[1] = p1;
-                        }
-                    }
-                    else if (inLine)
-                    {
-                        if (++gap > lineGap)
-                        {
-                            bool good_line = ::abs(line_end[1].x - line_end[0].x) >= lineLength ||
-                                             ::abs(line_end[1].y - line_end[0].y) >= lineLength;
-
-                            if (good_line)
-                            {
-                                const int ind = ::atomicAdd(&g_counter, 1);
-                                if (ind < maxSize)
-                                    out[ind] = make_int4(line_end[0].x, line_end[0].y, line_end[1].x, line_end[1].y);
-                            }
-
-                            gap = 0;
-                            inLine = false;
-                        }
-                    }
-
-                    p1 = p1 + d;
-                    if (p1.x < 0 || p1.x >= cols || p1.y < 0 || p1.y >= rows)
-                    {
-                        if (inLine)
-                        {
-                            bool good_line = ::abs(line_end[1].x - line_end[0].x) >= lineLength ||
-                                             ::abs(line_end[1].y - line_end[0].y) >= lineLength;
-
-                            if (good_line)
-                            {
-                                const int ind = ::atomicAdd(&g_counter, 1);
-                                if (ind < maxSize)
-                                    out[ind] = make_int4(line_end[0].x, line_end[0].y, line_end[1].x, line_end[1].y);
-                            }
-
-                        }
-                        break;
-                    }
-                }
-            }
-        }
-
-        int houghLinesProbabilistic_gpu(PtrStepSzb mask, PtrStepSzi accum, int4* out, int maxSize, float rho, float theta, int lineGap, int lineLength)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
-
-            bindTexture(&tex_mask, mask);
-
-            houghLinesProbabilistic<<<grid, block>>>(accum,
-                                                     out, maxSize,
-                                                     rho, theta,
-                                                     lineGap, lineLength,
-                                                     mask.rows, mask.cols);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // circlesAccumCenters
-
-        __global__ void circlesAccumCenters(const unsigned int* list, const int count, const PtrStepi dx, const PtrStepi dy,
-                                            PtrStepi accum, const int width, const int height, const int minRadius, const int maxRadius, const float idp)
-        {
-            const int SHIFT = 10;
-            const int ONE = 1 << SHIFT;
-
-            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
-
-            if (tid >= count)
-                return;
-
-            const unsigned int val = list[tid];
-
-            const int x = (val & 0xFFFF);
-            const int y = (val >> 16) & 0xFFFF;
-
-            const int vx = dx(y, x);
-            const int vy = dy(y, x);
-
-            if (vx == 0 && vy == 0)
-                return;
-
-            const float mag = ::sqrtf(vx * vx + vy * vy);
-
-            const int x0 = __float2int_rn((x * idp) * ONE);
-            const int y0 = __float2int_rn((y * idp) * ONE);
-
-            int sx = __float2int_rn((vx * idp) * ONE / mag);
-            int sy = __float2int_rn((vy * idp) * ONE / mag);
-
-            // Step from minRadius to maxRadius in both directions of the gradient
-            for (int k1 = 0; k1 < 2; ++k1)
-            {
-                int x1 = x0 + minRadius * sx;
-                int y1 = y0 + minRadius * sy;
-
-                for (int r = minRadius; r <= maxRadius; x1 += sx, y1 += sy, ++r)
-                {
-                    const int x2 = x1 >> SHIFT;
-                    const int y2 = y1 >> SHIFT;
-
-                    if (x2 < 0 || x2 >= width || y2 < 0 || y2 >= height)
-                        break;
-
-                    ::atomicAdd(accum.ptr(y2 + 1) + x2 + 1, 1);
-                }
-
-                sx = -sx;
-                sy = -sy;
-            }
-        }
-
-        void circlesAccumCenters_gpu(const unsigned int* list, int count, PtrStepi dx, PtrStepi dy, PtrStepSzi accum, int minRadius, int maxRadius, float idp)
-        {
-            const dim3 block(256);
-            const dim3 grid(divUp(count, block.x));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(circlesAccumCenters, cudaFuncCachePreferL1) );
-
-            circlesAccumCenters<<<grid, block>>>(list, count, dx, dy, accum, accum.cols - 2, accum.rows - 2, minRadius, maxRadius, idp);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // buildCentersList
-
-        __global__ void buildCentersList(const PtrStepSzi accum, unsigned int* centers, const int threshold)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < accum.cols - 2 && y < accum.rows - 2)
-            {
-                const int top = accum(y, x + 1);
-
-                const int left = accum(y + 1, x);
-                const int cur = accum(y + 1, x + 1);
-                const int right = accum(y + 1, x + 2);
-
-                const int bottom = accum(y + 2, x + 1);
-
-                if (cur > threshold && cur > top && cur >= bottom && cur >  left && cur >= right)
-                {
-                    const unsigned int val = (y << 16) | x;
-                    const int idx = ::atomicAdd(&g_counter, 1);
-                    centers[idx] = val;
-                }
-            }
-        }
-
-        int buildCentersList_gpu(PtrStepSzi accum, unsigned int* centers, int threshold)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(buildCentersList, cudaFuncCachePreferL1) );
-
-            buildCentersList<<<grid, block>>>(accum, centers, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // circlesAccumRadius
-
-        __global__ void circlesAccumRadius(const unsigned int* centers, const unsigned int* list, const int count,
-                                           float3* circles, const int maxCircles, const float dp,
-                                           const int minRadius, const int maxRadius, const int histSize, const int threshold)
-        {
-            int* smem = DynamicSharedMem<int>();
-
-            for (int i = threadIdx.x; i < histSize + 2; i += blockDim.x)
-                smem[i] = 0;
-            __syncthreads();
-
-            unsigned int val = centers[blockIdx.x];
-
-            float cx = (val & 0xFFFF);
-            float cy = (val >> 16) & 0xFFFF;
-
-            cx = (cx + 0.5f) * dp;
-            cy = (cy + 0.5f) * dp;
-
-            for (int i = threadIdx.x; i < count; i += blockDim.x)
-            {
-                val = list[i];
-
-                const int x = (val & 0xFFFF);
-                const int y = (val >> 16) & 0xFFFF;
-
-                const float rad = ::sqrtf((cx - x) * (cx - x) + (cy - y) * (cy - y));
-                if (rad >= minRadius && rad <= maxRadius)
-                {
-                    const int r = __float2int_rn(rad - minRadius);
-
-                    Emulation::smem::atomicAdd(&smem[r + 1], 1);
-                }
-            }
-
-            __syncthreads();
-
-            for (int i = threadIdx.x; i < histSize; i += blockDim.x)
-            {
-                const int curVotes = smem[i + 1];
-
-                if (curVotes >= threshold && curVotes > smem[i] && curVotes >= smem[i + 2])
-                {
-                    const int ind = ::atomicAdd(&g_counter, 1);
-                    if (ind < maxCircles)
-                        circles[ind] = make_float3(cx, cy, i + minRadius);
-                }
-            }
-        }
-
-        int circlesAccumRadius_gpu(const unsigned int* centers, int centersCount, const unsigned int* list, int count,
-                                   float3* circles, int maxCircles, float dp, int minRadius, int maxRadius, int threshold, bool has20)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(has20 ? 1024 : 512);
-            const dim3 grid(centersCount);
-
-            const int histSize = maxRadius - minRadius + 1;
-            size_t smemSize = (histSize + 2) * sizeof(int);
-
-            circlesAccumRadius<<<grid, block, smemSize>>>(centers, list, count, circles, maxCircles, dp, minRadius, maxRadius, histSize, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxCircles);
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // Generalized Hough
-
-        template <typename T, int PIXELS_PER_THREAD>
-        __global__ void buildEdgePointList(const PtrStepSzb edges, const PtrStep<T> dx, const PtrStep<T> dy, unsigned int* coordList, float* thetaList)
-        {
-            __shared__ unsigned int s_coordLists[4][32 * PIXELS_PER_THREAD];
-            __shared__ float s_thetaLists[4][32 * PIXELS_PER_THREAD];
-            __shared__ int s_sizes[4];
-            __shared__ int s_globStart[4];
-
-            const int x = blockIdx.x * blockDim.x * PIXELS_PER_THREAD + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (threadIdx.x == 0)
-                s_sizes[threadIdx.y] = 0;
-            __syncthreads();
-
-            if (y < edges.rows)
-            {
-                // fill the queue
-                const uchar* edgesRow = edges.ptr(y);
-                const T* dxRow = dx.ptr(y);
-                const T* dyRow = dy.ptr(y);
-
-                for (int i = 0, xx = x; i < PIXELS_PER_THREAD && xx < edges.cols; ++i, xx += blockDim.x)
-                {
-                    const T dxVal = dxRow[xx];
-                    const T dyVal = dyRow[xx];
-
-                    if (edgesRow[xx] && (dxVal != 0 || dyVal != 0))
-                    {
-                        const unsigned int coord = (y << 16) | xx;
-
-                        float theta = ::atan2f(dyVal, dxVal);
-                        if (theta < 0)
-                            theta += 2.0f * CV_PI_F;
-
-                        const int qidx = Emulation::smem::atomicAdd(&s_sizes[threadIdx.y], 1);
-
-                        s_coordLists[threadIdx.y][qidx] = coord;
-                        s_thetaLists[threadIdx.y][qidx] = theta;
-                    }
-                }
-            }
-
-            __syncthreads();
-
-            // let one thread reserve the space required in the global list
-            if (threadIdx.x == 0 && threadIdx.y == 0)
-            {
-                // find how many items are stored in each list
-                int totalSize = 0;
-                for (int i = 0; i < blockDim.y; ++i)
-                {
-                    s_globStart[i] = totalSize;
-                    totalSize += s_sizes[i];
-                }
-
-                // calculate the offset in the global list
-                const int globalOffset = atomicAdd(&g_counter, totalSize);
-                for (int i = 0; i < blockDim.y; ++i)
-                    s_globStart[i] += globalOffset;
-            }
-
-            __syncthreads();
-
-            // copy local queues to global queue
-            const int qsize = s_sizes[threadIdx.y];
-            int gidx = s_globStart[threadIdx.y] + threadIdx.x;
-            for(int i = threadIdx.x; i < qsize; i += blockDim.x, gidx += blockDim.x)
-            {
-                coordList[gidx] = s_coordLists[threadIdx.y][i];
-                thetaList[gidx] = s_thetaLists[threadIdx.y][i];
-            }
-        }
-
-        template <typename T>
-        int buildEdgePointList_gpu(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList)
-        {
-            const int PIXELS_PER_THREAD = 8;
-
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 4);
-            const dim3 grid(divUp(edges.cols, block.x * PIXELS_PER_THREAD), divUp(edges.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(buildEdgePointList<T, PIXELS_PER_THREAD>, cudaFuncCachePreferShared) );
-
-            buildEdgePointList<T, PIXELS_PER_THREAD><<<grid, block>>>(edges, (PtrStepSz<T>) dx, (PtrStepSz<T>) dy, coordList, thetaList);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            return totalCount;
-        }
-
-        template int buildEdgePointList_gpu<short>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
-        template int buildEdgePointList_gpu<int>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
-        template int buildEdgePointList_gpu<float>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
-
-        __global__ void buildRTable(const unsigned int* coordList, const float* thetaList, const int pointsCount,
-                                    PtrStep<short2> r_table, int* r_sizes, int maxSize,
-                                    const short2 templCenter, const float thetaScale)
-        {
-            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
-
-            if (tid >= pointsCount)
-                return;
-
-            const unsigned int coord = coordList[tid];
-            short2 p;
-            p.x = (coord & 0xFFFF);
-            p.y = (coord >> 16) & 0xFFFF;
-
-            const float theta = thetaList[tid];
-            const int n = __float2int_rn(theta * thetaScale);
-
-            const int ind = ::atomicAdd(r_sizes + n, 1);
-            if (ind < maxSize)
-                r_table(n, ind) = p - templCenter;
-        }
-
-        void buildRTable_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                             PtrStepSz<short2> r_table, int* r_sizes,
-                             short2 templCenter, int levels)
-        {
-            const dim3 block(256);
-            const dim3 grid(divUp(pointsCount, block.x));
-
-            const float thetaScale = levels / (2.0f * CV_PI_F);
-
-            buildRTable<<<grid, block>>>(coordList, thetaList, pointsCount, r_table, r_sizes, r_table.cols, templCenter, thetaScale);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // GHT_Ballard_Pos
-
-        __global__ void GHT_Ballard_Pos_calcHist(const unsigned int* coordList, const float* thetaList, const int pointsCount,
-                                                 const PtrStep<short2> r_table, const int* r_sizes,
-                                                 PtrStepSzi hist,
-                                                 const float idp, const float thetaScale)
-        {
-            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
-
-            if (tid >= pointsCount)
-                return;
-
-            const unsigned int coord = coordList[tid];
-            short2 p;
-            p.x = (coord & 0xFFFF);
-            p.y = (coord >> 16) & 0xFFFF;
-
-            const float theta = thetaList[tid];
-            const int n = __float2int_rn(theta * thetaScale);
-
-            const short2* r_row = r_table.ptr(n);
-            const int r_row_size = r_sizes[n];
-
-            for (int j = 0; j < r_row_size; ++j)
-            {
-                short2 c = p - r_row[j];
-
-                c.x = __float2int_rn(c.x * idp);
-                c.y = __float2int_rn(c.y * idp);
-
-                if (c.x >= 0 && c.x < hist.cols - 2 && c.y >= 0 && c.y < hist.rows - 2)
-                    ::atomicAdd(hist.ptr(c.y + 1) + c.x + 1, 1);
-            }
-        }
-
-        void GHT_Ballard_Pos_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                          PtrStepSz<short2> r_table, const int* r_sizes,
-                                          PtrStepSzi hist,
-                                          float dp, int levels)
-        {
-            const dim3 block(256);
-            const dim3 grid(divUp(pointsCount, block.x));
-
-            const float idp = 1.0f / dp;
-            const float thetaScale = levels / (2.0f * CV_PI_F);
-
-            GHT_Ballard_Pos_calcHist<<<grid, block>>>(coordList, thetaList, pointsCount, r_table, r_sizes, hist, idp, thetaScale);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Ballard_Pos_findPosInHist(const PtrStepSzi hist, float4* out, int3* votes, const int maxSize, const float dp, const int threshold)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= hist.cols - 2 || y >= hist.rows - 2)
-                return;
-
-            const int curVotes = hist(y + 1, x + 1);
-
-            if (curVotes > threshold &&
-                curVotes >  hist(y + 1, x) &&
-                curVotes >= hist(y + 1, x + 2) &&
-                curVotes >  hist(y, x + 1) &&
-                curVotes >= hist(y + 2, x + 1))
-            {
-                const int ind = ::atomicAdd(&g_counter, 1);
-
-                if (ind < maxSize)
-                {
-                    out[ind] = make_float4(x * dp, y * dp, 1.0f, 0.0f);
-                    votes[ind] = make_int3(curVotes, 0, 0);
-                }
-            }
-        }
-
-        int GHT_Ballard_Pos_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int maxSize, float dp, int threshold)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(hist.cols - 2, block.x), divUp(hist.rows - 2, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(GHT_Ballard_Pos_findPosInHist, cudaFuncCachePreferL1) );
-
-            GHT_Ballard_Pos_findPosInHist<<<grid, block>>>(hist, out, votes, maxSize, dp, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // GHT_Ballard_PosScale
-
-        __global__ void GHT_Ballard_PosScale_calcHist(const unsigned int* coordList, const float* thetaList,
-                                                      PtrStep<short2> r_table, const int* r_sizes,
-                                                      PtrStepi hist, const int rows, const int cols,
-                                                      const float minScale, const float scaleStep, const int scaleRange,
-                                                      const float idp, const float thetaScale)
-        {
-            const unsigned int coord = coordList[blockIdx.x];
-            float2 p;
-            p.x = (coord & 0xFFFF);
-            p.y = (coord >> 16) & 0xFFFF;
-
-            const float theta = thetaList[blockIdx.x];
-            const int n = __float2int_rn(theta * thetaScale);
-
-            const short2* r_row = r_table.ptr(n);
-            const int r_row_size = r_sizes[n];
-
-            for (int j = 0; j < r_row_size; ++j)
-            {
-                const float2 d = saturate_cast<float2>(r_row[j]);
-
-                for (int s = threadIdx.x; s < scaleRange; s += blockDim.x)
-                {
-                    const float scale = minScale + s * scaleStep;
-
-                    float2 c = p - scale * d;
-
-                    c.x *= idp;
-                    c.y *= idp;
-
-                    if (c.x >= 0 && c.x < cols && c.y >= 0 && c.y < rows)
-                        ::atomicAdd(hist.ptr((s + 1) * (rows + 2) + __float2int_rn(c.y + 1)) + __float2int_rn(c.x + 1), 1);
-                }
-            }
-        }
-
-        void GHT_Ballard_PosScale_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                               PtrStepSz<short2> r_table, const int* r_sizes,
-                                               PtrStepi hist, int rows, int cols,
-                                               float minScale, float scaleStep, int scaleRange,
-                                               float dp, int levels)
-        {
-            const dim3 block(256);
-            const dim3 grid(pointsCount);
-
-            const float idp = 1.0f / dp;
-            const float thetaScale = levels / (2.0f * CV_PI_F);
-
-            GHT_Ballard_PosScale_calcHist<<<grid, block>>>(coordList, thetaList,
-                                                           r_table, r_sizes,
-                                                           hist, rows, cols,
-                                                           minScale, scaleStep, scaleRange,
-                                                           idp, thetaScale);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Ballard_PosScale_findPosInHist(const PtrStepi hist, const int rows, const int cols, const int scaleRange,
-                                                           float4* out, int3* votes, const int maxSize,
-                                                           const float minScale, const float scaleStep, const float dp, const int threshold)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= cols || y >= rows)
-                return;
-
-            for (int s = 0; s < scaleRange; ++s)
-            {
-                const float scale = minScale + s * scaleStep;
-
-                const int prevScaleIdx = (s) * (rows + 2);
-                const int curScaleIdx = (s + 1) * (rows + 2);
-                const int nextScaleIdx = (s + 2) * (rows + 2);
-
-                const int curVotes = hist(curScaleIdx + y + 1, x + 1);
-
-                if (curVotes > threshold &&
-                    curVotes >  hist(curScaleIdx + y + 1, x) &&
-                    curVotes >= hist(curScaleIdx + y + 1, x + 2) &&
-                    curVotes >  hist(curScaleIdx + y, x + 1) &&
-                    curVotes >= hist(curScaleIdx + y + 2, x + 1) &&
-                    curVotes >  hist(prevScaleIdx + y + 1, x + 1) &&
-                    curVotes >= hist(nextScaleIdx + y + 1, x + 1))
-                {
-                    const int ind = ::atomicAdd(&g_counter, 1);
-
-                    if (ind < maxSize)
-                    {
-                        out[ind] = make_float4(x * dp, y * dp, scale, 0.0f);
-                        votes[ind] = make_int3(curVotes, curVotes, 0);
-                    }
-                }
-            }
-        }
-
-        int GHT_Ballard_PosScale_findPosInHist_gpu(PtrStepi hist, int rows, int cols, int scaleRange, float4* out, int3* votes, int maxSize,
-                                                   float minScale, float scaleStep, float dp, int threshold)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(GHT_Ballard_PosScale_findPosInHist, cudaFuncCachePreferL1) );
-
-            GHT_Ballard_PosScale_findPosInHist<<<grid, block>>>(hist, rows, cols, scaleRange, out, votes, maxSize, minScale, scaleStep, dp, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // GHT_Ballard_PosRotation
-
-        __global__ void GHT_Ballard_PosRotation_calcHist(const unsigned int* coordList, const float* thetaList,
-                                                         PtrStep<short2> r_table, const int* r_sizes,
-                                                         PtrStepi hist, const int rows, const int cols,
-                                                         const float minAngle, const float angleStep, const int angleRange,
-                                                         const float idp, const float thetaScale)
-        {
-            const unsigned int coord = coordList[blockIdx.x];
-            float2 p;
-            p.x = (coord & 0xFFFF);
-            p.y = (coord >> 16) & 0xFFFF;
-
-            const float thetaVal = thetaList[blockIdx.x];
-
-            for (int a = threadIdx.x; a < angleRange; a += blockDim.x)
-            {
-                const float angle = (minAngle + a * angleStep) * (CV_PI_F / 180.0f);
-                float sinA, cosA;
-                sincosf(angle, &sinA, &cosA);
-
-                float theta = thetaVal - angle;
-                if (theta < 0)
-                    theta += 2.0f * CV_PI_F;
-
-                const int n = __float2int_rn(theta * thetaScale);
-
-                const short2* r_row = r_table.ptr(n);
-                const int r_row_size = r_sizes[n];
-
-                for (int j = 0; j < r_row_size; ++j)
-                {
-                    const float2 d = saturate_cast<float2>(r_row[j]);
-
-                    const float2 dr = make_float2(d.x * cosA - d.y * sinA, d.x * sinA + d.y * cosA);
-
-                    float2 c = make_float2(p.x - dr.x, p.y - dr.y);
-                    c.x *= idp;
-                    c.y *= idp;
-
-                    if (c.x >= 0 && c.x < cols && c.y >= 0 && c.y < rows)
-                        ::atomicAdd(hist.ptr((a + 1) * (rows + 2) + __float2int_rn(c.y + 1)) + __float2int_rn(c.x + 1), 1);
-                }
-            }
-        }
-
-        void GHT_Ballard_PosRotation_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                  PtrStepSz<short2> r_table, const int* r_sizes,
-                                                  PtrStepi hist, int rows, int cols,
-                                                  float minAngle, float angleStep, int angleRange,
-                                                  float dp, int levels)
-        {
-            const dim3 block(256);
-            const dim3 grid(pointsCount);
-
-            const float idp = 1.0f / dp;
-            const float thetaScale = levels / (2.0f * CV_PI_F);
-
-            GHT_Ballard_PosRotation_calcHist<<<grid, block>>>(coordList, thetaList,
-                                                              r_table, r_sizes,
-                                                              hist, rows, cols,
-                                                              minAngle, angleStep, angleRange,
-                                                              idp, thetaScale);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Ballard_PosRotation_findPosInHist(const PtrStepi hist, const int rows, const int cols, const int angleRange,
-                                                              float4* out, int3* votes, const int maxSize,
-                                                              const float minAngle, const float angleStep, const float dp, const int threshold)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= cols || y >= rows)
-                return;
-
-            for (int a = 0; a < angleRange; ++a)
-            {
-                const float angle = minAngle + a * angleStep;
-
-                const int prevAngleIdx = (a) * (rows + 2);
-                const int curAngleIdx = (a + 1) * (rows + 2);
-                const int nextAngleIdx = (a + 2) * (rows + 2);
-
-                const int curVotes = hist(curAngleIdx + y + 1, x + 1);
-
-                if (curVotes > threshold &&
-                    curVotes >  hist(curAngleIdx + y + 1, x) &&
-                    curVotes >= hist(curAngleIdx + y + 1, x + 2) &&
-                    curVotes >  hist(curAngleIdx + y, x + 1) &&
-                    curVotes >= hist(curAngleIdx + y + 2, x + 1) &&
-                    curVotes >  hist(prevAngleIdx + y + 1, x + 1) &&
-                    curVotes >= hist(nextAngleIdx + y + 1, x + 1))
-                {
-                    const int ind = ::atomicAdd(&g_counter, 1);
-
-                    if (ind < maxSize)
-                    {
-                        out[ind] = make_float4(x * dp, y * dp, 1.0f, angle);
-                        votes[ind] = make_int3(curVotes, 0, curVotes);
-                    }
-                }
-            }
-        }
-
-        int GHT_Ballard_PosRotation_findPosInHist_gpu(PtrStepi hist, int rows, int cols, int angleRange, float4* out, int3* votes, int maxSize,
-                                                      float minAngle, float angleStep, float dp, int threshold)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(cols, block.x), divUp(rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(GHT_Ballard_PosRotation_findPosInHist, cudaFuncCachePreferL1) );
-
-            GHT_Ballard_PosRotation_findPosInHist<<<grid, block>>>(hist, rows, cols, angleRange, out, votes, maxSize, minAngle, angleStep, dp, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            return totalCount;
-        }
-
-        ////////////////////////////////////////////////////////////////////////
-        // GHT_Guil_Full
-
-        struct FeatureTable
-        {
-            uchar* p1_pos_data;
-            size_t p1_pos_step;
-
-            uchar* p1_theta_data;
-            size_t p1_theta_step;
-
-            uchar* p2_pos_data;
-            size_t p2_pos_step;
-
-            uchar* d12_data;
-            size_t d12_step;
-
-            uchar* r1_data;
-            size_t r1_step;
-
-            uchar* r2_data;
-            size_t r2_step;
-        };
-
-        __constant__ FeatureTable c_templFeatures;
-        __constant__ FeatureTable c_imageFeatures;
-
-        void GHT_Guil_Full_setTemplFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2)
-        {
-            FeatureTable tbl;
-
-            tbl.p1_pos_data = p1_pos.data;
-            tbl.p1_pos_step = p1_pos.step;
-
-            tbl.p1_theta_data = p1_theta.data;
-            tbl.p1_theta_step = p1_theta.step;
-
-            tbl.p2_pos_data = p2_pos.data;
-            tbl.p2_pos_step = p2_pos.step;
-
-            tbl.d12_data = d12.data;
-            tbl.d12_step = d12.step;
-
-            tbl.r1_data = r1.data;
-            tbl.r1_step = r1.step;
-
-            tbl.r2_data = r2.data;
-            tbl.r2_step = r2.step;
-
-            cudaSafeCall( cudaMemcpyToSymbol(c_templFeatures, &tbl, sizeof(FeatureTable)) );
-        }
-        void GHT_Guil_Full_setImageFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2)
-        {
-            FeatureTable tbl;
-
-            tbl.p1_pos_data = p1_pos.data;
-            tbl.p1_pos_step = p1_pos.step;
-
-            tbl.p1_theta_data = p1_theta.data;
-            tbl.p1_theta_step = p1_theta.step;
-
-            tbl.p2_pos_data = p2_pos.data;
-            tbl.p2_pos_step = p2_pos.step;
-
-            tbl.d12_data = d12.data;
-            tbl.d12_step = d12.step;
-
-            tbl.r1_data = r1.data;
-            tbl.r1_step = r1.step;
-
-            tbl.r2_data = r2.data;
-            tbl.r2_step = r2.step;
-
-            cudaSafeCall( cudaMemcpyToSymbol(c_imageFeatures, &tbl, sizeof(FeatureTable)) );
-        }
-
-        struct TemplFeatureTable
-        {
-            static __device__ float2* p1_pos(int n)
-            {
-                return (float2*)(c_templFeatures.p1_pos_data + n * c_templFeatures.p1_pos_step);
-            }
-            static __device__ float* p1_theta(int n)
-            {
-                return (float*)(c_templFeatures.p1_theta_data + n * c_templFeatures.p1_theta_step);
-            }
-            static __device__ float2* p2_pos(int n)
-            {
-                return (float2*)(c_templFeatures.p2_pos_data + n * c_templFeatures.p2_pos_step);
-            }
-
-            static __device__ float* d12(int n)
-            {
-                return (float*)(c_templFeatures.d12_data + n * c_templFeatures.d12_step);
-            }
-
-            static __device__ float2* r1(int n)
-            {
-                return (float2*)(c_templFeatures.r1_data + n * c_templFeatures.r1_step);
-            }
-            static __device__ float2* r2(int n)
-            {
-                return (float2*)(c_templFeatures.r2_data + n * c_templFeatures.r2_step);
-            }
-        };
-        struct ImageFeatureTable
-        {
-            static __device__ float2* p1_pos(int n)
-            {
-                return (float2*)(c_imageFeatures.p1_pos_data + n * c_imageFeatures.p1_pos_step);
-            }
-            static __device__ float* p1_theta(int n)
-            {
-                return (float*)(c_imageFeatures.p1_theta_data + n * c_imageFeatures.p1_theta_step);
-            }
-            static __device__ float2* p2_pos(int n)
-            {
-                return (float2*)(c_imageFeatures.p2_pos_data + n * c_imageFeatures.p2_pos_step);
-            }
-
-            static __device__ float* d12(int n)
-            {
-                return (float*)(c_imageFeatures.d12_data + n * c_imageFeatures.d12_step);
-            }
-
-            static __device__ float2* r1(int n)
-            {
-                return (float2*)(c_imageFeatures.r1_data + n * c_imageFeatures.r1_step);
-            }
-            static __device__ float2* r2(int n)
-            {
-                return (float2*)(c_imageFeatures.r2_data + n * c_imageFeatures.r2_step);
-            }
-        };
-
-        __device__ float clampAngle(float a)
-        {
-            float res = a;
-
-            while (res > 2.0f * CV_PI_F)
-                res -= 2.0f * CV_PI_F;
-            while (res < 0.0f)
-                res += 2.0f * CV_PI_F;
-
-            return res;
-        }
-
-        __device__ bool angleEq(float a, float b, float eps)
-        {
-            return (::fabs(clampAngle(a - b)) <= eps);
-        }
-
-        template <class FT, bool isTempl>
-        __global__ void GHT_Guil_Full_buildFeatureList(const unsigned int* coordList, const float* thetaList, const int pointsCount,
-                                                       int* sizes, const int maxSize,
-                                                       const float xi, const float angleEpsilon, const float alphaScale,
-                                                       const float2 center, const float maxDist)
-        {
-            const float p1_theta = thetaList[blockIdx.x];
-            const unsigned int coord1 = coordList[blockIdx.x];
-            float2 p1_pos;
-            p1_pos.x = (coord1 & 0xFFFF);
-            p1_pos.y = (coord1 >> 16) & 0xFFFF;
-
-            for (int i = threadIdx.x; i < pointsCount; i += blockDim.x)
-            {
-                const float p2_theta = thetaList[i];
-                const unsigned int coord2 = coordList[i];
-                float2 p2_pos;
-                p2_pos.x = (coord2 & 0xFFFF);
-                p2_pos.y = (coord2 >> 16) & 0xFFFF;
-
-                if (angleEq(p1_theta - p2_theta, xi, angleEpsilon))
-                {
-                    const float2 d = p1_pos - p2_pos;
-
-                    float alpha12 = clampAngle(::atan2(d.y, d.x) - p1_theta);
-                    float d12 = ::sqrtf(d.x * d.x + d.y * d.y);
-
-                    if (d12 > maxDist)
-                        continue;
-
-                    float2 r1 = p1_pos - center;
-                    float2 r2 = p2_pos - center;
-
-                    const int n = __float2int_rn(alpha12 * alphaScale);
-
-                    const int ind = ::atomicAdd(sizes + n, 1);
-
-                    if (ind < maxSize)
-                    {
-                        if (!isTempl)
-                        {
-                            FT::p1_pos(n)[ind] = p1_pos;
-                            FT::p2_pos(n)[ind] = p2_pos;
-                        }
-
-                        FT::p1_theta(n)[ind] = p1_theta;
-
-                        FT::d12(n)[ind] = d12;
-
-                        if (isTempl)
-                        {
-                            FT::r1(n)[ind] = r1;
-                            FT::r2(n)[ind] = r2;
-                        }
-                    }
-                }
-            }
-        }
-
-        template <class FT, bool isTempl>
-        void GHT_Guil_Full_buildFeatureList_caller(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                   int* sizes, int maxSize,
-                                                   float xi, float angleEpsilon, int levels,
-                                                   float2 center, float maxDist)
-        {
-            const dim3 block(256);
-            const dim3 grid(pointsCount);
-
-            const float alphaScale = levels / (2.0f * CV_PI_F);
-
-            GHT_Guil_Full_buildFeatureList<FT, isTempl><<<grid, block>>>(coordList, thetaList, pointsCount,
-                                                                         sizes, maxSize,
-                                                                         xi * (CV_PI_F / 180.0f), angleEpsilon * (CV_PI_F / 180.0f), alphaScale,
-                                                                         center, maxDist);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            thrust::device_ptr<int> sizesPtr(sizes);
-            thrust::transform(sizesPtr, sizesPtr + levels + 1, sizesPtr, device::bind2nd(device::minimum<int>(), maxSize));
-        }
-
-        void GHT_Guil_Full_buildTemplFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                     int* sizes, int maxSize,
-                                                     float xi, float angleEpsilon, int levels,
-                                                     float2 center, float maxDist)
-        {
-            GHT_Guil_Full_buildFeatureList_caller<TemplFeatureTable, true>(coordList, thetaList, pointsCount,
-                                                                           sizes, maxSize,
-                                                                           xi, angleEpsilon, levels,
-                                                                           center, maxDist);
-        }
-        void GHT_Guil_Full_buildImageFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                     int* sizes, int maxSize,
-                                                     float xi, float angleEpsilon, int levels,
-                                                     float2 center, float maxDist)
-        {
-            GHT_Guil_Full_buildFeatureList_caller<ImageFeatureTable, false>(coordList, thetaList, pointsCount,
-                                                                            sizes, maxSize,
-                                                                            xi, angleEpsilon, levels,
-                                                                            center, maxDist);
-        }
-
-        __global__ void GHT_Guil_Full_calcOHist(const int* templSizes, const int* imageSizes, int* OHist,
-                                                const float minAngle, const float maxAngle, const float iAngleStep, const int angleRange)
-        {
-            extern __shared__ int s_OHist[];
-            for (int i = threadIdx.x; i <= angleRange; i += blockDim.x)
-                s_OHist[i] = 0;
-            __syncthreads();
-
-            const int tIdx = blockIdx.x;
-            const int level = blockIdx.y;
-
-            const int tSize = templSizes[level];
-
-            if (tIdx < tSize)
-            {
-                const int imSize = imageSizes[level];
-
-                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx];
-
-                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
-                {
-                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
-
-                    const float angle = clampAngle(im_p1_theta - t_p1_theta);
-
-                    if (angle >= minAngle && angle <= maxAngle)
-                    {
-                        const int n = __float2int_rn((angle - minAngle) * iAngleStep);
-                        Emulation::smem::atomicAdd(&s_OHist[n], 1);
-                    }
-                }
-            }
-            __syncthreads();
-
-            for (int i = threadIdx.x; i <= angleRange; i += blockDim.x)
-                ::atomicAdd(OHist + i, s_OHist[i]);
-        }
-
-        void GHT_Guil_Full_calcOHist_gpu(const int* templSizes, const int* imageSizes, int* OHist,
-                                         float minAngle, float maxAngle, float angleStep, int angleRange,
-                                         int levels, int tMaxSize)
-        {
-            const dim3 block(256);
-            const dim3 grid(tMaxSize, levels + 1);
-
-            minAngle *= (CV_PI_F / 180.0f);
-            maxAngle *= (CV_PI_F / 180.0f);
-            angleStep *= (CV_PI_F / 180.0f);
-
-            const size_t smemSize = (angleRange + 1) * sizeof(float);
-
-            GHT_Guil_Full_calcOHist<<<grid, block, smemSize>>>(templSizes, imageSizes, OHist,
-                                                               minAngle, maxAngle, 1.0f / angleStep, angleRange);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Guil_Full_calcSHist(const int* templSizes, const int* imageSizes, int* SHist,
-                                                const float angle, const float angleEpsilon,
-                                                const float minScale, const float maxScale, const float iScaleStep, const int scaleRange)
-        {
-            extern __shared__ int s_SHist[];
-            for (int i = threadIdx.x; i <= scaleRange; i += blockDim.x)
-                s_SHist[i] = 0;
-            __syncthreads();
-
-            const int tIdx = blockIdx.x;
-            const int level = blockIdx.y;
-
-            const int tSize = templSizes[level];
-
-            if (tIdx < tSize)
-            {
-                const int imSize = imageSizes[level];
-
-                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx] + angle;
-                const float t_d12 = TemplFeatureTable::d12(level)[tIdx] + angle;
-
-                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
-                {
-                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
-                    const float im_d12 = ImageFeatureTable::d12(level)[i];
-
-                    if (angleEq(im_p1_theta, t_p1_theta, angleEpsilon))
-                    {
-                        const float scale = im_d12 / t_d12;
-
-                        if (scale >= minScale && scale <= maxScale)
-                        {
-                            const int s = __float2int_rn((scale - minScale) * iScaleStep);
-                            Emulation::smem::atomicAdd(&s_SHist[s], 1);
-                        }
-                    }
-                }
-            }
-            __syncthreads();
-
-            for (int i = threadIdx.x; i <= scaleRange; i += blockDim.x)
-                ::atomicAdd(SHist + i, s_SHist[i]);
-        }
-
-        void GHT_Guil_Full_calcSHist_gpu(const int* templSizes, const int* imageSizes, int* SHist,
-                                         float angle, float angleEpsilon,
-                                         float minScale, float maxScale, float iScaleStep, int scaleRange,
-                                         int levels, int tMaxSize)
-        {
-            const dim3 block(256);
-            const dim3 grid(tMaxSize, levels + 1);
-
-            angle *= (CV_PI_F / 180.0f);
-            angleEpsilon *= (CV_PI_F / 180.0f);
-
-            const size_t smemSize = (scaleRange + 1) * sizeof(float);
-
-            GHT_Guil_Full_calcSHist<<<grid, block, smemSize>>>(templSizes, imageSizes, SHist,
-                                                               angle, angleEpsilon,
-                                                               minScale, maxScale, iScaleStep, scaleRange);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Guil_Full_calcPHist(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
-                                                const float angle, const float sinVal, const float cosVal, const float angleEpsilon, const float scale,
-                                                const float idp)
-        {
-            const int tIdx = blockIdx.x;
-            const int level = blockIdx.y;
-
-            const int tSize = templSizes[level];
-
-            if (tIdx < tSize)
-            {
-                const int imSize = imageSizes[level];
-
-                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx] + angle;
-
-                float2 r1 = TemplFeatureTable::r1(level)[tIdx];
-                float2 r2 = TemplFeatureTable::r2(level)[tIdx];
-
-                r1 = r1 * scale;
-                r2 = r2 * scale;
-
-                r1 = make_float2(cosVal * r1.x - sinVal * r1.y, sinVal * r1.x + cosVal * r1.y);
-                r2 = make_float2(cosVal * r2.x - sinVal * r2.y, sinVal * r2.x + cosVal * r2.y);
-
-                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
-                {
-                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
-
-                    const float2 im_p1_pos = ImageFeatureTable::p1_pos(level)[i];
-                    const float2 im_p2_pos = ImageFeatureTable::p2_pos(level)[i];
-
-                    if (angleEq(im_p1_theta, t_p1_theta, angleEpsilon))
-                    {
-                        float2 c1, c2;
-
-                        c1 = im_p1_pos - r1;
-                        c1 = c1 * idp;
-
-                        c2 = im_p2_pos - r2;
-                        c2 = c2 * idp;
-
-                        if (::fabs(c1.x - c2.x) > 1 || ::fabs(c1.y - c2.y) > 1)
-                            continue;
-
-                        if (c1.y >= 0 && c1.y < PHist.rows - 2 && c1.x >= 0 && c1.x < PHist.cols - 2)
-                            ::atomicAdd(PHist.ptr(__float2int_rn(c1.y) + 1) + __float2int_rn(c1.x) + 1, 1);
-                    }
-                }
-            }
-        }
-
-        void GHT_Guil_Full_calcPHist_gpu(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
-                                         float angle, float angleEpsilon, float scale,
-                                         float dp,
-                                         int levels, int tMaxSize)
-        {
-            const dim3 block(256);
-            const dim3 grid(tMaxSize, levels + 1);
-
-            angle *= (CV_PI_F / 180.0f);
-            angleEpsilon *= (CV_PI_F / 180.0f);
-
-            const float sinVal = ::sinf(angle);
-            const float cosVal = ::cosf(angle);
-
-            cudaSafeCall( cudaFuncSetCacheConfig(GHT_Guil_Full_calcPHist, cudaFuncCachePreferL1) );
-
-            GHT_Guil_Full_calcPHist<<<grid, block>>>(templSizes, imageSizes, PHist,
-                                                     angle, sinVal, cosVal, angleEpsilon, scale,
-                                                     1.0f / dp);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        __global__ void GHT_Guil_Full_findPosInHist(const PtrStepSzi hist, float4* out, int3* votes, const int maxSize,
-                                                    const float angle, const int angleVotes, const float scale, const int scaleVotes,
-                                                    const float dp, const int threshold)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= hist.cols - 2 || y >= hist.rows - 2)
-                return;
-
-            const int curVotes = hist(y + 1, x + 1);
-
-            if (curVotes > threshold &&
-                curVotes >  hist(y + 1, x) &&
-                curVotes >= hist(y + 1, x + 2) &&
-                curVotes >  hist(y, x + 1) &&
-                curVotes >= hist(y + 2, x + 1))
-            {
-                const int ind = ::atomicAdd(&g_counter, 1);
-
-                if (ind < maxSize)
-                {
-                    out[ind] = make_float4(x * dp, y * dp, scale, angle);
-                    votes[ind] = make_int3(curVotes, scaleVotes, angleVotes);
-                }
-            }
-        }
-
-        int GHT_Guil_Full_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int curSize, int maxSize,
-                                             float angle, int angleVotes, float scale, int scaleVotes,
-                                             float dp, int threshold)
-        {
-            void* counterPtr;
-            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
-
-            cudaSafeCall( cudaMemcpy(counterPtr, &curSize, sizeof(int), cudaMemcpyHostToDevice) );
-
-            const dim3 block(32, 8);
-            const dim3 grid(divUp(hist.cols - 2, block.x), divUp(hist.rows - 2, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(GHT_Guil_Full_findPosInHist, cudaFuncCachePreferL1) );
-
-            GHT_Guil_Full_findPosInHist<<<grid, block>>>(hist, out, votes, maxSize,
-                                                         angle, angleVotes, scale, scaleVotes,
-                                                         dp, threshold);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-            int totalCount;
-            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
-
-            totalCount = ::min(totalCount, maxSize);
-
-            return totalCount;
-        }
-    }
-}}}
-
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/imgproc.cu b/modules/gpu/src/cuda/imgproc.cu
deleted file mode 100644
index b23e0a6653..0000000000
--- a/modules/gpu/src/cuda/imgproc.cu
+++ /dev/null
@@ -1,1007 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        /////////////////////////////////// MeanShiftfiltering ///////////////////////////////////////////////
-
-        texture<uchar4, 2> tex_meanshift;
-
-        __device__ short2 do_mean_shift(int x0, int y0, unsigned char* out,
-                                        size_t out_step, int cols, int rows,
-                                        int sp, int sr, int maxIter, float eps)
-        {
-            int isr2 = sr*sr;
-            uchar4 c = tex2D(tex_meanshift, x0, y0 );
-
-            // iterate meanshift procedure
-            for( int iter = 0; iter < maxIter; iter++ )
-            {
-                int count = 0;
-                int s0 = 0, s1 = 0, s2 = 0, sx = 0, sy = 0;
-                float icount;
-
-                //mean shift: process pixels in window (p-sigmaSp)x(p+sigmaSp)
-                int minx = x0-sp;
-                int miny = y0-sp;
-                int maxx = x0+sp;
-                int maxy = y0+sp;
-
-                for( int y = miny; y <= maxy; y++)
-                {
-                    int rowCount = 0;
-                    for( int x = minx; x <= maxx; x++ )
-                    {
-                        uchar4 t = tex2D( tex_meanshift, x, y );
-
-                        int norm2 = (t.x - c.x) * (t.x - c.x) + (t.y - c.y) * (t.y - c.y) + (t.z - c.z) * (t.z - c.z);
-                        if( norm2 <= isr2 )
-                        {
-                            s0 += t.x; s1 += t.y; s2 += t.z;
-                            sx += x; rowCount++;
-                        }
-                    }
-                    count += rowCount;
-                    sy += y*rowCount;
-                }
-
-                if( count == 0 )
-                    break;
-
-                icount = 1.f/count;
-                int x1 = __float2int_rz(sx*icount);
-                int y1 = __float2int_rz(sy*icount);
-                s0 = __float2int_rz(s0*icount);
-                s1 = __float2int_rz(s1*icount);
-                s2 = __float2int_rz(s2*icount);
-
-                int norm2 = (s0 - c.x) * (s0 - c.x) + (s1 - c.y) * (s1 - c.y) + (s2 - c.z) * (s2 - c.z);
-
-                bool stopFlag = (x0 == x1 && y0 == y1) || (::abs(x1-x0) + ::abs(y1-y0) + norm2 <= eps);
-
-                x0 = x1; y0 = y1;
-                c.x = s0; c.y = s1; c.z = s2;
-
-                if( stopFlag )
-                    break;
-            }
-
-            int base = (blockIdx.y * blockDim.y + threadIdx.y) * out_step + (blockIdx.x * blockDim.x + threadIdx.x) * 4 * sizeof(uchar);
-            *(uchar4*)(out + base) = c;
-
-            return make_short2((short)x0, (short)y0);
-        }
-
-        __global__ void meanshift_kernel(unsigned char* out, size_t out_step, int cols, int rows, int sp, int sr, int maxIter, float eps )
-        {
-            int x0 = blockIdx.x * blockDim.x + threadIdx.x;
-            int y0 = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if( x0 < cols && y0 < rows )
-                do_mean_shift(x0, y0, out, out_step, cols, rows, sp, sr, maxIter, eps);
-        }
-
-        __global__ void meanshiftproc_kernel(unsigned char* outr, size_t outrstep,
-                                             unsigned char* outsp, size_t outspstep,
-                                             int cols, int rows,
-                                             int sp, int sr, int maxIter, float eps)
-        {
-            int x0 = blockIdx.x * blockDim.x + threadIdx.x;
-            int y0 = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if( x0 < cols && y0 < rows )
-            {
-                int basesp = (blockIdx.y * blockDim.y + threadIdx.y) * outspstep + (blockIdx.x * blockDim.x + threadIdx.x) * 2 * sizeof(short);
-                *(short2*)(outsp + basesp) = do_mean_shift(x0, y0, outr, outrstep, cols, rows, sp, sr, maxIter, eps);
-            }
-        }
-
-        void meanShiftFiltering_gpu(const PtrStepSzb& src, PtrStepSzb dst, int sp, int sr, int maxIter, float eps, cudaStream_t stream)
-        {
-            dim3 grid(1, 1, 1);
-            dim3 threads(32, 8, 1);
-            grid.x = divUp(src.cols, threads.x);
-            grid.y = divUp(src.rows, threads.y);
-
-            cudaChannelFormatDesc desc = cudaCreateChannelDesc<uchar4>();
-            cudaSafeCall( cudaBindTexture2D( 0, tex_meanshift, src.data, desc, src.cols, src.rows, src.step ) );
-
-            meanshift_kernel<<< grid, threads, 0, stream >>>( dst.data, dst.step, dst.cols, dst.rows, sp, sr, maxIter, eps );
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-
-            //cudaSafeCall( cudaUnbindTexture( tex_meanshift ) );
-        }
-
-        void meanShiftProc_gpu(const PtrStepSzb& src, PtrStepSzb dstr, PtrStepSzb dstsp, int sp, int sr, int maxIter, float eps, cudaStream_t stream)
-        {
-            dim3 grid(1, 1, 1);
-            dim3 threads(32, 8, 1);
-            grid.x = divUp(src.cols, threads.x);
-            grid.y = divUp(src.rows, threads.y);
-
-            cudaChannelFormatDesc desc = cudaCreateChannelDesc<uchar4>();
-            cudaSafeCall( cudaBindTexture2D( 0, tex_meanshift, src.data, desc, src.cols, src.rows, src.step ) );
-
-            meanshiftproc_kernel<<< grid, threads, 0, stream >>>( dstr.data, dstr.step, dstsp.data, dstsp.step, dstr.cols, dstr.rows, sp, sr, maxIter, eps );
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-
-            //cudaSafeCall( cudaUnbindTexture( tex_meanshift ) );
-        }
-
-        /////////////////////////////////// drawColorDisp ///////////////////////////////////////////////
-
-        template <typename T>
-        __device__ unsigned int cvtPixel(T d, int ndisp, float S = 1, float V = 1)
-        {
-            unsigned int H = ((ndisp-d) * 240)/ndisp;
-
-            unsigned int hi = (H/60) % 6;
-            float f = H/60.f - H/60;
-            float p = V * (1 - S);
-            float q = V * (1 - f * S);
-            float t = V * (1 - (1 - f) * S);
-
-            float3 res;
-
-            if (hi == 0) //R = V,	G = t,	B = p
-            {
-                res.x = p;
-                res.y = t;
-                res.z = V;
-            }
-
-            if (hi == 1) // R = q,	G = V,	B = p
-            {
-                res.x = p;
-                res.y = V;
-                res.z = q;
-            }
-
-            if (hi == 2) // R = p,	G = V,	B = t
-            {
-                res.x = t;
-                res.y = V;
-                res.z = p;
-            }
-
-            if (hi == 3) // R = p,	G = q,	B = V
-            {
-                res.x = V;
-                res.y = q;
-                res.z = p;
-            }
-
-            if (hi == 4) // R = t,	G = p,	B = V
-            {
-                res.x = V;
-                res.y = p;
-                res.z = t;
-            }
-
-            if (hi == 5) // R = V,	G = p,	B = q
-            {
-                res.x = q;
-                res.y = p;
-                res.z = V;
-            }
-            const unsigned int b = (unsigned int)(::max(0.f, ::min(res.x, 1.f)) * 255.f);
-            const unsigned int g = (unsigned int)(::max(0.f, ::min(res.y, 1.f)) * 255.f);
-            const unsigned int r = (unsigned int)(::max(0.f, ::min(res.z, 1.f)) * 255.f);
-            const unsigned int a = 255U;
-
-            return (a << 24) + (r << 16) + (g << 8) + b;
-        }
-
-        __global__ void drawColorDisp(uchar* disp, size_t disp_step, uchar* out_image, size_t out_step, int width, int height, int ndisp)
-        {
-            const int x = (blockIdx.x * blockDim.x + threadIdx.x) << 2;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if(x < width && y < height)
-            {
-                uchar4 d4 = *(uchar4*)(disp + y * disp_step + x);
-
-                uint4 res;
-                res.x = cvtPixel(d4.x, ndisp);
-                res.y = cvtPixel(d4.y, ndisp);
-                res.z = cvtPixel(d4.z, ndisp);
-                res.w = cvtPixel(d4.w, ndisp);
-
-                uint4* line = (uint4*)(out_image + y * out_step);
-                line[x >> 2] = res;
-            }
-        }
-
-        __global__ void drawColorDisp(short* disp, size_t disp_step, uchar* out_image, size_t out_step, int width, int height, int ndisp)
-        {
-            const int x = (blockIdx.x * blockDim.x + threadIdx.x) << 1;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if(x < width && y < height)
-            {
-                short2 d2 = *(short2*)(disp + y * disp_step + x);
-
-                uint2 res;
-                res.x = cvtPixel(d2.x, ndisp);
-                res.y = cvtPixel(d2.y, ndisp);
-
-                uint2* line = (uint2*)(out_image + y * out_step);
-                line[x >> 1] = res;
-            }
-        }
-
-
-        void drawColorDisp_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream)
-        {
-            dim3 threads(16, 16, 1);
-            dim3 grid(1, 1, 1);
-            grid.x = divUp(src.cols, threads.x << 2);
-            grid.y = divUp(src.rows, threads.y);
-
-            drawColorDisp<<<grid, threads, 0, stream>>>(src.data, src.step, dst.data, dst.step, src.cols, src.rows, ndisp);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void drawColorDisp_gpu(const PtrStepSz<short>& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream)
-        {
-            dim3 threads(32, 8, 1);
-            dim3 grid(1, 1, 1);
-            grid.x = divUp(src.cols, threads.x << 1);
-            grid.y = divUp(src.rows, threads.y);
-
-            drawColorDisp<<<grid, threads, 0, stream>>>(src.data, src.step / sizeof(short), dst.data, dst.step, src.cols, src.rows, ndisp);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        /////////////////////////////////// reprojectImageTo3D ///////////////////////////////////////////////
-
-        __constant__ float cq[16];
-
-        template <typename T, typename D>
-        __global__ void reprojectImageTo3D(const PtrStepSz<T> disp, PtrStep<D> xyz)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (y >= disp.rows || x >= disp.cols)
-                return;
-
-            const float qx = x * cq[ 0] + y * cq[ 1] + cq[ 3];
-            const float qy = x * cq[ 4] + y * cq[ 5] + cq[ 7];
-            const float qz = x * cq[ 8] + y * cq[ 9] + cq[11];
-            const float qw = x * cq[12] + y * cq[13] + cq[15];
-
-            const T d = disp(y, x);
-
-            const float iW = 1.f / (qw + cq[14] * d);
-
-            D v = VecTraits<D>::all(1.0f);
-            v.x = (qx + cq[2] * d) * iW;
-            v.y = (qy + cq[6] * d) * iW;
-            v.z = (qz + cq[10] * d) * iW;
-
-            xyz(y, x) = v;
-        }
-
-        template <typename T, typename D>
-        void reprojectImageTo3D_gpu(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(disp.cols, block.x), divUp(disp.rows, block.y));
-
-            cudaSafeCall( cudaMemcpyToSymbol(cq, q, 16 * sizeof(float)) );
-
-            reprojectImageTo3D<T, D><<<grid, block, 0, stream>>>((PtrStepSz<T>)disp, (PtrStepSz<D>)xyz);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        template void reprojectImageTo3D_gpu<uchar, float3>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-        template void reprojectImageTo3D_gpu<uchar, float4>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-        template void reprojectImageTo3D_gpu<short, float3>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-        template void reprojectImageTo3D_gpu<short, float4>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-
-        /////////////////////////////////////////// Corner Harris /////////////////////////////////////////////////
-
-        texture<float, cudaTextureType2D, cudaReadModeElementType> harrisDxTex(0, cudaFilterModePoint, cudaAddressModeClamp);
-        texture<float, cudaTextureType2D, cudaReadModeElementType> harrisDyTex(0, cudaFilterModePoint, cudaAddressModeClamp);
-
-        __global__ void cornerHarris_kernel(const int block_size, const float k, PtrStepSzf dst)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                float a = 0.f;
-                float b = 0.f;
-                float c = 0.f;
-
-                const int ibegin = y - (block_size / 2);
-                const int jbegin = x - (block_size / 2);
-                const int iend = ibegin + block_size;
-                const int jend = jbegin + block_size;
-
-                for (int i = ibegin; i < iend; ++i)
-                {
-                    for (int j = jbegin; j < jend; ++j)
-                    {
-                        float dx = tex2D(harrisDxTex, j, i);
-                        float dy = tex2D(harrisDyTex, j, i);
-
-                        a += dx * dx;
-                        b += dx * dy;
-                        c += dy * dy;
-                    }
-                }
-
-                dst(y, x) = a * c - b * b - k * (a + c) * (a + c);
-            }
-        }
-
-        template <typename BR, typename BC>
-        __global__ void cornerHarris_kernel(const int block_size, const float k, PtrStepSzf dst, const BR border_row, const BC border_col)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                float a = 0.f;
-                float b = 0.f;
-                float c = 0.f;
-
-                const int ibegin = y - (block_size / 2);
-                const int jbegin = x - (block_size / 2);
-                const int iend = ibegin + block_size;
-                const int jend = jbegin + block_size;
-
-                for (int i = ibegin; i < iend; ++i)
-                {
-                    const int y = border_col.idx_row(i);
-
-                    for (int j = jbegin; j < jend; ++j)
-                    {
-                        const int x = border_row.idx_col(j);
-
-                        float dx = tex2D(harrisDxTex, x, y);
-                        float dy = tex2D(harrisDyTex, x, y);
-
-                        a += dx * dx;
-                        b += dx * dy;
-                        c += dy * dy;
-                    }
-                }
-
-                dst(y, x) = a * c - b * b - k * (a + c) * (a + c);
-            }
-        }
-
-        void cornerHarris_gpu(int block_size, float k, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(Dx.cols, block.x), divUp(Dx.rows, block.y));
-
-            bindTexture(&harrisDxTex, Dx);
-            bindTexture(&harrisDyTex, Dy);
-
-            switch (border_type)
-            {
-            case BORDER_REFLECT101_GPU:
-                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst, BrdRowReflect101<void>(Dx.cols), BrdColReflect101<void>(Dx.rows));
-                break;
-
-            case BORDER_REFLECT_GPU:
-                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst, BrdRowReflect<void>(Dx.cols), BrdColReflect<void>(Dx.rows));
-                break;
-
-            case BORDER_REPLICATE_GPU:
-                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst);
-                break;
-            }
-
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        /////////////////////////////////////////// Corner Min Eigen Val /////////////////////////////////////////////////
-
-        texture<float, cudaTextureType2D, cudaReadModeElementType> minEigenValDxTex(0, cudaFilterModePoint, cudaAddressModeClamp);
-        texture<float, cudaTextureType2D, cudaReadModeElementType> minEigenValDyTex(0, cudaFilterModePoint, cudaAddressModeClamp);
-
-        __global__ void cornerMinEigenVal_kernel(const int block_size, PtrStepSzf dst)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                float a = 0.f;
-                float b = 0.f;
-                float c = 0.f;
-
-                const int ibegin = y - (block_size / 2);
-                const int jbegin = x - (block_size / 2);
-                const int iend = ibegin + block_size;
-                const int jend = jbegin + block_size;
-
-                for (int i = ibegin; i < iend; ++i)
-                {
-                    for (int j = jbegin; j < jend; ++j)
-                    {
-                        float dx = tex2D(minEigenValDxTex, j, i);
-                        float dy = tex2D(minEigenValDyTex, j, i);
-
-                        a += dx * dx;
-                        b += dx * dy;
-                        c += dy * dy;
-                    }
-                }
-
-                a *= 0.5f;
-                c *= 0.5f;
-
-                dst(y, x) = (a + c) - sqrtf((a - c) * (a - c) + b * b);
-            }
-        }
-
-
-        template <typename BR, typename BC>
-        __global__ void cornerMinEigenVal_kernel(const int block_size, PtrStepSzf dst, const BR border_row, const BC border_col)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                float a = 0.f;
-                float b = 0.f;
-                float c = 0.f;
-
-                const int ibegin = y - (block_size / 2);
-                const int jbegin = x - (block_size / 2);
-                const int iend = ibegin + block_size;
-                const int jend = jbegin + block_size;
-
-                for (int i = ibegin; i < iend; ++i)
-                {
-                    int y = border_col.idx_row(i);
-
-                    for (int j = jbegin; j < jend; ++j)
-                    {
-                        int x = border_row.idx_col(j);
-
-                        float dx = tex2D(minEigenValDxTex, x, y);
-                        float dy = tex2D(minEigenValDyTex, x, y);
-
-                        a += dx * dx;
-                        b += dx * dy;
-                        c += dy * dy;
-                    }
-                }
-
-                a *= 0.5f;
-                c *= 0.5f;
-
-                dst(y, x) = (a + c) - sqrtf((a - c) * (a - c) + b * b);
-            }
-        }
-
-        void cornerMinEigenVal_gpu(int block_size, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(Dx.cols, block.x), divUp(Dx.rows, block.y));
-
-            bindTexture(&minEigenValDxTex, Dx);
-            bindTexture(&minEigenValDyTex, Dy);
-
-            switch (border_type)
-            {
-            case BORDER_REFLECT101_GPU:
-                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst, BrdRowReflect101<void>(Dx.cols), BrdColReflect101<void>(Dx.rows));
-                break;
-
-            case BORDER_REFLECT_GPU:
-                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst, BrdRowReflect<void>(Dx.cols), BrdColReflect<void>(Dx.rows));
-                break;
-
-            case BORDER_REPLICATE_GPU:
-                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst);
-                break;
-            }
-
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall(cudaDeviceSynchronize());
-        }
-
-        ////////////////////////////// Column Sum //////////////////////////////////////
-
-        __global__ void column_sumKernel_32F(int cols, int rows, const PtrStepb src, const PtrStepb dst)
-        {
-            int x = blockIdx.x * blockDim.x + threadIdx.x;
-
-            if (x < cols)
-            {
-                const unsigned char* src_data = src.data + x * sizeof(float);
-                unsigned char* dst_data = dst.data + x * sizeof(float);
-
-                float sum = 0.f;
-                for (int y = 0; y < rows; ++y)
-                {
-                    sum += *(const float*)src_data;
-                    *(float*)dst_data = sum;
-                    src_data += src.step;
-                    dst_data += dst.step;
-                }
-            }
-        }
-
-
-        void columnSum_32F(const PtrStepSzb src, const PtrStepSzb dst)
-        {
-            dim3 threads(256);
-            dim3 grid(divUp(src.cols, threads.x));
-
-            column_sumKernel_32F<<<grid, threads>>>(src.cols, src.rows, src, dst);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-
-        //////////////////////////////////////////////////////////////////////////
-        // mulSpectrums
-
-        __global__ void mulSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < c.cols && y < c.rows)
-            {
-                c.ptr(y)[x] = cuCmulf(a.ptr(y)[x], b.ptr(y)[x]);
-            }
-        }
-
-
-        void mulSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream)
-        {
-            dim3 threads(256);
-            dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
-
-            mulSpectrumsKernel<<<grid, threads, 0, stream>>>(a, b, c);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-
-        //////////////////////////////////////////////////////////////////////////
-        // mulSpectrums_CONJ
-
-        __global__ void mulSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < c.cols && y < c.rows)
-            {
-                c.ptr(y)[x] = cuCmulf(a.ptr(y)[x], cuConjf(b.ptr(y)[x]));
-            }
-        }
-
-
-        void mulSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream)
-        {
-            dim3 threads(256);
-            dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
-
-            mulSpectrumsKernel_CONJ<<<grid, threads, 0, stream>>>(a, b, c);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-
-        //////////////////////////////////////////////////////////////////////////
-        // mulAndScaleSpectrums
-
-        __global__ void mulAndScaleSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < c.cols && y < c.rows)
-            {
-                cufftComplex v = cuCmulf(a.ptr(y)[x], b.ptr(y)[x]);
-                c.ptr(y)[x] = make_cuFloatComplex(cuCrealf(v) * scale, cuCimagf(v) * scale);
-            }
-        }
-
-
-        void mulAndScaleSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream)
-        {
-            dim3 threads(256);
-            dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
-
-            mulAndScaleSpectrumsKernel<<<grid, threads, 0, stream>>>(a, b, scale, c);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-
-        //////////////////////////////////////////////////////////////////////////
-        // mulAndScaleSpectrums_CONJ
-
-        __global__ void mulAndScaleSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x < c.cols && y < c.rows)
-            {
-                cufftComplex v = cuCmulf(a.ptr(y)[x], cuConjf(b.ptr(y)[x]));
-                c.ptr(y)[x] = make_cuFloatComplex(cuCrealf(v) * scale, cuCimagf(v) * scale);
-            }
-        }
-
-
-        void mulAndScaleSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream)
-        {
-            dim3 threads(256);
-            dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
-
-            mulAndScaleSpectrumsKernel_CONJ<<<grid, threads, 0, stream>>>(a, b, scale, c);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        //////////////////////////////////////////////////////////////////////////
-        // buildWarpMaps
-
-        // TODO use intrinsics like __sinf and so on
-
-        namespace build_warp_maps
-        {
-
-            __constant__ float ck_rinv[9];
-            __constant__ float cr_kinv[9];
-            __constant__ float ct[3];
-            __constant__ float cscale;
-        }
-
-
-        class PlaneMapper
-        {
-        public:
-            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
-            {
-                using namespace build_warp_maps;
-
-                float x_ = u / cscale - ct[0];
-                float y_ = v / cscale - ct[1];
-
-                float z;
-                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * (1 - ct[2]);
-                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * (1 - ct[2]);
-                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * (1 - ct[2]);
-
-                x /= z;
-                y /= z;
-            }
-        };
-
-
-        class CylindricalMapper
-        {
-        public:
-            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
-            {
-                using namespace build_warp_maps;
-
-                u /= cscale;
-                float x_ = ::sinf(u);
-                float y_ = v / cscale;
-                float z_ = ::cosf(u);
-
-                float z;
-                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * z_;
-                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * z_;
-                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * z_;
-
-                if (z > 0) { x /= z; y /= z; }
-                else x = y = -1;
-            }
-        };
-
-
-        class SphericalMapper
-        {
-        public:
-            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
-            {
-                using namespace build_warp_maps;
-
-                v /= cscale;
-                u /= cscale;
-
-                float sinv = ::sinf(v);
-                float x_ = sinv * ::sinf(u);
-                float y_ = -::cosf(v);
-                float z_ = sinv * ::cosf(u);
-
-                float z;
-                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * z_;
-                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * z_;
-                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * z_;
-
-                if (z > 0) { x /= z; y /= z; }
-                else x = y = -1;
-            }
-        };
-
-
-        template <typename Mapper>
-        __global__ void buildWarpMapsKernel(int tl_u, int tl_v, int cols, int rows,
-                                            PtrStepf map_x, PtrStepf map_y)
-        {
-            int du = blockIdx.x * blockDim.x + threadIdx.x;
-            int dv = blockIdx.y * blockDim.y + threadIdx.y;
-            if (du < cols && dv < rows)
-            {
-                float u = tl_u + du;
-                float v = tl_v + dv;
-                float x, y;
-                Mapper::mapBackward(u, v, x, y);
-                map_x.ptr(dv)[du] = x;
-                map_y.ptr(dv)[du] = y;
-            }
-        }
-
-
-        void buildWarpPlaneMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                const float k_rinv[9], const float r_kinv[9], const float t[3],
-                                float scale, cudaStream_t stream)
-        {
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ct, t, 3*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
-
-            int cols = map_x.cols;
-            int rows = map_x.rows;
-
-            dim3 threads(32, 8);
-            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
-
-            buildWarpMapsKernel<PlaneMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
-            cudaSafeCall(cudaGetLastError());
-            if (stream == 0)
-                cudaSafeCall(cudaDeviceSynchronize());
-        }
-
-
-        void buildWarpCylindricalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                      const float k_rinv[9], const float r_kinv[9], float scale,
-                                      cudaStream_t stream)
-        {
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
-
-            int cols = map_x.cols;
-            int rows = map_x.rows;
-
-            dim3 threads(32, 8);
-            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
-
-            buildWarpMapsKernel<CylindricalMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
-            cudaSafeCall(cudaGetLastError());
-            if (stream == 0)
-                cudaSafeCall(cudaDeviceSynchronize());
-        }
-
-
-        void buildWarpSphericalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                    const float k_rinv[9], const float r_kinv[9], float scale,
-                                    cudaStream_t stream)
-        {
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
-            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
-
-            int cols = map_x.cols;
-            int rows = map_x.rows;
-
-            dim3 threads(32, 8);
-            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
-
-            buildWarpMapsKernel<SphericalMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
-            cudaSafeCall(cudaGetLastError());
-            if (stream == 0)
-                cudaSafeCall(cudaDeviceSynchronize());
-        }
-
-        //////////////////////////////////////////////////////////////////////////
-        // filter2D
-
-        #define FILTER2D_MAX_KERNEL_SIZE 16
-
-        __constant__ float c_filter2DKernel[FILTER2D_MAX_KERNEL_SIZE * FILTER2D_MAX_KERNEL_SIZE];
-
-        template <class SrcT, typename D>
-        __global__ void filter2D(const SrcT src, PtrStepSz<D> dst, const int kWidth, const int kHeight, const int anchorX, const int anchorY)
-        {
-            typedef typename TypeVec<float, VecTraits<D>::cn>::vec_type sum_t;
-
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= dst.cols || y >= dst.rows)
-                return;
-
-            sum_t res = VecTraits<sum_t>::all(0);
-            int kInd = 0;
-
-            for (int i = 0; i < kHeight; ++i)
-            {
-                for (int j = 0; j < kWidth; ++j)
-                    res = res + src(y - anchorY + i, x - anchorX + j) * c_filter2DKernel[kInd++];
-            }
-
-            dst(y, x) = saturate_cast<D>(res);
-        }
-
-        template <typename T, typename D, template <typename> class Brd> struct Filter2DCaller;
-
-        #define IMPLEMENT_FILTER2D_TEX_READER(type) \
-            texture< type , cudaTextureType2D, cudaReadModeElementType> tex_filter2D_ ## type (0, cudaFilterModePoint, cudaAddressModeClamp); \
-            struct tex_filter2D_ ## type ## _reader \
-            { \
-                typedef type elem_type; \
-                typedef int index_type; \
-                const int xoff; \
-                const int yoff; \
-                tex_filter2D_ ## type ## _reader (int xoff_, int yoff_) : xoff(xoff_), yoff(yoff_) {} \
-                __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const \
-                { \
-                    return tex2D(tex_filter2D_ ## type , x + xoff, y + yoff); \
-                } \
-            }; \
-            template <typename D, template <typename> class Brd> struct Filter2DCaller< type , D, Brd> \
-            { \
-                static void call(const PtrStepSz< type > srcWhole, int xoff, int yoff, PtrStepSz<D> dst, \
-                    int kWidth, int kHeight, int anchorX, int anchorY, const float* borderValue, cudaStream_t stream) \
-                { \
-                    typedef typename TypeVec<float, VecTraits< type >::cn>::vec_type work_type; \
-                    dim3 block(16, 16); \
-                    dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \
-                    bindTexture(&tex_filter2D_ ## type , srcWhole); \
-                    tex_filter2D_ ## type ##_reader texSrc(xoff, yoff); \
-                    Brd<work_type> brd(dst.rows, dst.cols, VecTraits<work_type>::make(borderValue)); \
-                    BorderReader< tex_filter2D_ ## type ##_reader, Brd<work_type> > brdSrc(texSrc, brd); \
-                    filter2D<<<grid, block, 0, stream>>>(brdSrc, dst, kWidth, kHeight, anchorX, anchorY); \
-                    cudaSafeCall( cudaGetLastError() ); \
-                    if (stream == 0) \
-                        cudaSafeCall( cudaDeviceSynchronize() ); \
-                } \
-            };
-
-        IMPLEMENT_FILTER2D_TEX_READER(uchar);
-        IMPLEMENT_FILTER2D_TEX_READER(uchar4);
-
-        IMPLEMENT_FILTER2D_TEX_READER(ushort);
-        IMPLEMENT_FILTER2D_TEX_READER(ushort4);
-
-        IMPLEMENT_FILTER2D_TEX_READER(float);
-        IMPLEMENT_FILTER2D_TEX_READER(float4);
-
-        #undef IMPLEMENT_FILTER2D_TEX_READER
-
-        template <typename T, typename D>
-        void filter2D_gpu(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst,
-                          int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel,
-                          int borderMode, const float* borderValue, cudaStream_t stream)
-        {
-            typedef void (*func_t)(const PtrStepSz<T> srcWhole, int xoff, int yoff, PtrStepSz<D> dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* borderValue, cudaStream_t stream);
-            static const func_t funcs[] =
-            {
-                Filter2DCaller<T, D, BrdReflect101>::call,
-                Filter2DCaller<T, D, BrdReplicate>::call,
-                Filter2DCaller<T, D, BrdConstant>::call,
-                Filter2DCaller<T, D, BrdReflect>::call,
-                Filter2DCaller<T, D, BrdWrap>::call
-            };
-
-            if (stream == 0)
-                cudaSafeCall( cudaMemcpyToSymbol(c_filter2DKernel, kernel, kWidth * kHeight * sizeof(float), 0, cudaMemcpyDeviceToDevice) );
-            else
-                cudaSafeCall( cudaMemcpyToSymbolAsync(c_filter2DKernel, kernel, kWidth * kHeight * sizeof(float), 0, cudaMemcpyDeviceToDevice, stream) );
-
-            funcs[borderMode](static_cast< PtrStepSz<T> >(srcWhole), ofsX, ofsY, static_cast< PtrStepSz<D> >(dst), kWidth, kHeight, anchorX, anchorY, borderValue, stream);
-        }
-
-        template void filter2D_gpu<uchar, uchar>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-        template void filter2D_gpu<uchar4, uchar4>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-        template void filter2D_gpu<ushort, ushort>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-        template void filter2D_gpu<ushort4, ushort4>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-        template void filter2D_gpu<float, float>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-        template void filter2D_gpu<float4, float4>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel, int borderMode, const float* borderValue, cudaStream_t stream);
-    } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device {
-
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/lbp.cu b/modules/gpu/src/cuda/lbp.cu
index a09aa1e79b..fb6267f2fe 100644
--- a/modules/gpu/src/cuda/lbp.cu
+++ b/modules/gpu/src/cuda/lbp.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -43,10 +43,10 @@
 #if !defined CUDA_DISABLER
 
 #include "lbp.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace lbp
     {
diff --git a/modules/gpu/src/cuda/lbp.hpp b/modules/gpu/src/cuda/lbp.hpp
index 0c8a03e334..1bfdc610cb 100644
--- a/modules/gpu/src/cuda/lbp.hpp
+++ b/modules/gpu/src/cuda/lbp.hpp
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -43,10 +43,10 @@
 #ifndef __OPENCV_GPU_DEVICE_LBP_HPP_
 #define __OPENCV_GPU_DEVICE_LBP_HPP_
 
-#include "internal_shared.hpp"
-#include <opencv2/gpu/device/emulation.hpp>
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
 
-namespace cv { namespace gpu { namespace device {
+namespace cv { namespace gpu { namespace cudev {
 
 namespace lbp {
 
@@ -72,10 +72,10 @@ namespace lbp {
 
         __device__ __forceinline__ bool operator()(const int4& r1, const int4& r2) const
         {
-            float delta = eps * (min(r1.z, r2.z) + min(r1.w, r2.w)) * 0.5f;
+            float delta = eps * (::min(r1.z, r2.z) + ::min(r1.w, r2.w)) * 0.5f;
 
-            return abs(r1.x - r2.x) <= delta && abs(r1.y - r2.y) <= delta
-                && abs(r1.x + r1.z - r2.x - r2.z) <= delta && abs(r1.y + r1.w - r2.y - r2.w) <= delta;
+            return ::abs(r1.x - r2.x) <= delta && ::abs(r1.y - r2.y) <= delta
+                && ::abs(r1.x + r1.z - r2.x - r2.z) <= delta && ::abs(r1.y + r1.w - r2.y - r2.w) <= delta;
         }
         float eps;
     };
@@ -109,4 +109,4 @@ namespace lbp {
 
 } } }// namespaces
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/gpu/src/cuda/matrix_reductions.cu b/modules/gpu/src/cuda/matrix_reductions.cu
deleted file mode 100644
index 745daca1db..0000000000
--- a/modules/gpu/src/cuda/matrix_reductions.cu
+++ /dev/null
@@ -1,1366 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/type_traits.hpp"
-
-using namespace cv::gpu;
-using namespace cv::gpu::device;
-
-namespace detail
-{
-    __device__ __forceinline__ int cvAtomicAdd(int* address, int val)
-    {
-        return ::atomicAdd(address, val);
-    }
-    __device__ __forceinline__ unsigned int cvAtomicAdd(unsigned int* address, unsigned int val)
-    {
-        return ::atomicAdd(address, val);
-    }
-    __device__ __forceinline__ float cvAtomicAdd(float* address, float val)
-    {
-    #if __CUDA_ARCH__ >= 200
-        return ::atomicAdd(address, val);
-    #else
-        int* address_as_i = (int*) address;
-        int old = *address_as_i, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_i, assumed,
-                __float_as_int(val + __int_as_float(assumed)));
-        } while (assumed != old);
-        return __int_as_float(old);
-    #endif
-    }
-    __device__ __forceinline__ double cvAtomicAdd(double* address, double val)
-    {
-    #if __CUDA_ARCH__ >= 130
-        unsigned long long int* address_as_ull = (unsigned long long int*) address;
-        unsigned long long int old = *address_as_ull, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_ull, assumed,
-                __double_as_longlong(val + __longlong_as_double(assumed)));
-        } while (assumed != old);
-        return __longlong_as_double(old);
-    #else
-        (void) address;
-        (void) val;
-        return 0.0;
-    #endif
-    }
-
-    __device__ __forceinline__ int cvAtomicMin(int* address, int val)
-    {
-        return ::atomicMin(address, val);
-    }
-    __device__ __forceinline__ float cvAtomicMin(float* address, float val)
-    {
-    #if __CUDA_ARCH__ >= 120
-        int* address_as_i = (int*) address;
-        int old = *address_as_i, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_i, assumed,
-                __float_as_int(::fminf(val, __int_as_float(assumed))));
-        } while (assumed != old);
-        return __int_as_float(old);
-    #else
-        (void) address;
-        (void) val;
-        return 0.0f;
-    #endif
-    }
-    __device__ __forceinline__ double cvAtomicMin(double* address, double val)
-    {
-    #if __CUDA_ARCH__ >= 130
-        unsigned long long int* address_as_ull = (unsigned long long int*) address;
-        unsigned long long int old = *address_as_ull, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_ull, assumed,
-                __double_as_longlong(::fmin(val, __longlong_as_double(assumed))));
-        } while (assumed != old);
-        return __longlong_as_double(old);
-    #else
-        (void) address;
-        (void) val;
-        return 0.0;
-    #endif
-    }
-
-    __device__ __forceinline__ int cvAtomicMax(int* address, int val)
-    {
-        return ::atomicMax(address, val);
-    }
-    __device__ __forceinline__ float cvAtomicMax(float* address, float val)
-    {
-    #if __CUDA_ARCH__ >= 120
-        int* address_as_i = (int*) address;
-        int old = *address_as_i, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_i, assumed,
-                __float_as_int(::fmaxf(val, __int_as_float(assumed))));
-        } while (assumed != old);
-        return __int_as_float(old);
-    #else
-        (void) address;
-        (void) val;
-        return 0.0f;
-    #endif
-    }
-    __device__ __forceinline__ double cvAtomicMax(double* address, double val)
-    {
-    #if __CUDA_ARCH__ >= 130
-        unsigned long long int* address_as_ull = (unsigned long long int*) address;
-        unsigned long long int old = *address_as_ull, assumed;
-        do {
-            assumed = old;
-            old = ::atomicCAS(address_as_ull, assumed,
-                __double_as_longlong(::fmax(val, __longlong_as_double(assumed))));
-        } while (assumed != old);
-        return __longlong_as_double(old);
-    #else
-        (void) address;
-        (void) val;
-        return 0.0;
-    #endif
-    }
-}
-
-namespace detail
-{
-    template <int cn> struct Unroll;
-    template <> struct Unroll<1>
-    {
-        template <int BLOCK_SIZE, typename R>
-        static __device__ __forceinline__ volatile R* smem_tuple(R* smem)
-        {
-            return smem;
-        }
-
-        template <typename R>
-        static __device__ __forceinline__ R& tie(R& val)
-        {
-            return val;
-        }
-
-        template <class Op>
-        static __device__ __forceinline__ const Op& op(const Op& op)
-        {
-            return op;
-        }
-    };
-    template <> struct Unroll<2>
-    {
-        template <int BLOCK_SIZE, typename R>
-        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*> smem_tuple(R* smem)
-        {
-            return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE);
-        }
-
-        template <typename R>
-        static __device__ __forceinline__ thrust::tuple<typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&> tie(R& val)
-        {
-            return thrust::tie(val.x, val.y);
-        }
-
-        template <class Op>
-        static __device__ __forceinline__ const thrust::tuple<Op, Op> op(const Op& op)
-        {
-            return thrust::make_tuple(op, op);
-        }
-    };
-    template <> struct Unroll<3>
-    {
-        template <int BLOCK_SIZE, typename R>
-        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*, volatile R*> smem_tuple(R* smem)
-        {
-            return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE);
-        }
-
-        template <typename R>
-        static __device__ __forceinline__ thrust::tuple<typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&> tie(R& val)
-        {
-            return thrust::tie(val.x, val.y, val.z);
-        }
-
-        template <class Op>
-        static __device__ __forceinline__ const thrust::tuple<Op, Op, Op> op(const Op& op)
-        {
-            return thrust::make_tuple(op, op, op);
-        }
-    };
-    template <> struct Unroll<4>
-    {
-        template <int BLOCK_SIZE, typename R>
-        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*, volatile R*, volatile R*> smem_tuple(R* smem)
-        {
-            return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE);
-        }
-
-        template <typename R>
-        static __device__ __forceinline__ thrust::tuple<typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&, typename VecTraits<R>::elem_type&> tie(R& val)
-        {
-            return thrust::tie(val.x, val.y, val.z, val.w);
-        }
-
-        template <class Op>
-        static __device__ __forceinline__ const thrust::tuple<Op, Op, Op, Op> op(const Op& op)
-        {
-            return thrust::make_tuple(op, op, op, op);
-        }
-    };
-}
-
-/////////////////////////////////////////////////////////////
-// sum
-
-namespace sum
-{
-    __device__ unsigned int blocks_finished = 0;
-
-    template <typename R, int cn> struct AtomicAdd;
-    template <typename R> struct AtomicAdd<R, 1>
-    {
-        static __device__ void run(R* ptr, R val)
-        {
-            detail::cvAtomicAdd(ptr, val);
-        }
-    };
-    template <typename R> struct AtomicAdd<R, 2>
-    {
-        typedef typename TypeVec<R, 2>::vec_type val_type;
-
-        static __device__ void run(R* ptr, val_type val)
-        {
-            detail::cvAtomicAdd(ptr, val.x);
-            detail::cvAtomicAdd(ptr + 1, val.y);
-        }
-    };
-    template <typename R> struct AtomicAdd<R, 3>
-    {
-        typedef typename TypeVec<R, 3>::vec_type val_type;
-
-        static __device__ void run(R* ptr, val_type val)
-        {
-            detail::cvAtomicAdd(ptr, val.x);
-            detail::cvAtomicAdd(ptr + 1, val.y);
-            detail::cvAtomicAdd(ptr + 2, val.z);
-        }
-    };
-    template <typename R> struct AtomicAdd<R, 4>
-    {
-        typedef typename TypeVec<R, 4>::vec_type val_type;
-
-        static __device__ void run(R* ptr, val_type val)
-        {
-            detail::cvAtomicAdd(ptr, val.x);
-            detail::cvAtomicAdd(ptr + 1, val.y);
-            detail::cvAtomicAdd(ptr + 2, val.z);
-            detail::cvAtomicAdd(ptr + 3, val.w);
-        }
-    };
-
-    template <int BLOCK_SIZE, typename R, int cn>
-    struct GlobalReduce
-    {
-        typedef typename TypeVec<R, cn>::vec_type result_type;
-
-        static __device__ void run(result_type& sum, result_type* result, int tid, int bid, R* smem)
-        {
-        #if __CUDA_ARCH__ >= 200
-            if (tid == 0)
-                AtomicAdd<R, cn>::run((R*) result, sum);
-        #else
-            __shared__ bool is_last;
-
-            if (tid == 0)
-            {
-                result[bid] = sum;
-
-                __threadfence();
-
-                unsigned int ticket = ::atomicAdd(&blocks_finished, 1);
-                is_last = (ticket == gridDim.x * gridDim.y - 1);
-            }
-
-            __syncthreads();
-
-            if (is_last)
-            {
-                sum = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<result_type>::all(0);
-
-                device::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(sum), tid, detail::Unroll<cn>::op(plus<R>()));
-
-                if (tid == 0)
-                {
-                    result[0] = sum;
-                    blocks_finished = 0;
-                }
-            }
-        #endif
-        }
-    };
-
-    template <int BLOCK_SIZE, typename src_type, typename result_type, class Mask, class Op>
-    __global__ void kernel(const PtrStepSz<src_type> src, result_type* result, const Mask mask, const Op op, const int twidth, const int theight)
-    {
-        typedef typename VecTraits<src_type>::elem_type T;
-        typedef typename VecTraits<result_type>::elem_type R;
-        const int cn = VecTraits<src_type>::cn;
-
-        __shared__ R smem[BLOCK_SIZE * cn];
-
-        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
-        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
-
-        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
-        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
-
-        result_type sum = VecTraits<result_type>::all(0);
-
-        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
-        {
-            const src_type* ptr = src.ptr(y);
-
-            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
-            {
-                if (mask(y, x))
-                {
-                    const src_type srcVal = ptr[x];
-                    sum = sum + op(saturate_cast<result_type>(srcVal));
-                }
-            }
-        }
-
-        device::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(sum), tid, detail::Unroll<cn>::op(plus<R>()));
-
-        GlobalReduce<BLOCK_SIZE, R, cn>::run(sum, result, tid, bid, smem);
-    }
-
-    const int threads_x = 32;
-    const int threads_y = 8;
-
-    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
-    {
-        block = dim3(threads_x, threads_y);
-
-        grid = dim3(divUp(cols, block.x * block.y),
-                    divUp(rows, block.y * block.x));
-
-        grid.x = ::min(grid.x, block.x);
-        grid.y = ::min(grid.y, block.y);
-    }
-
-    void getBufSize(int cols, int rows, int cn, int& bufcols, int& bufrows)
-    {
-        dim3 block, grid;
-        getLaunchCfg(cols, rows, block, grid);
-
-        bufcols = grid.x * grid.y * sizeof(double) * cn;
-        bufrows = 1;
-    }
-
-    template <typename T, typename R, int cn, template <typename> class Op>
-    void caller(PtrStepSzb src_, void* buf_, double* out, PtrStepSzb mask)
-    {
-        typedef typename TypeVec<T, cn>::vec_type src_type;
-        typedef typename TypeVec<R, cn>::vec_type result_type;
-
-        PtrStepSz<src_type> src(src_);
-        result_type* buf = (result_type*) buf_;
-
-        dim3 block, grid;
-        getLaunchCfg(src.cols, src.rows, block, grid);
-
-        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
-        const int theight = divUp(divUp(src.rows, grid.y), block.y);
-
-        Op<result_type> op;
-
-        if (mask.data)
-            kernel<threads_x * threads_y><<<grid, block>>>(src, buf, SingleMask(mask), op, twidth, theight);
-        else
-            kernel<threads_x * threads_y><<<grid, block>>>(src, buf, WithOutMask(), op, twidth, theight);
-        cudaSafeCall( cudaGetLastError() );
-
-        cudaSafeCall( cudaDeviceSynchronize() );
-
-        R result[4] = {0, 0, 0, 0};
-        cudaSafeCall( cudaMemcpy(&result, buf, sizeof(result_type), cudaMemcpyDeviceToHost) );
-
-        out[0] = result[0];
-        out[1] = result[1];
-        out[2] = result[2];
-        out[3] = result[3];
-    }
-
-    template <typename T> struct SumType;
-    template <> struct SumType<uchar> { typedef unsigned int R; };
-    template <> struct SumType<schar> { typedef int R; };
-    template <> struct SumType<ushort> { typedef unsigned int R; };
-    template <> struct SumType<short> { typedef int R; };
-    template <> struct SumType<int> { typedef int R; };
-    template <> struct SumType<float> { typedef float R; };
-    template <> struct SumType<double> { typedef double R; };
-
-    template <typename T, int cn>
-    void run(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
-    {
-        typedef typename SumType<T>::R R;
-        caller<T, R, cn, identity>(src, buf, out, mask);
-    }
-
-    template void run<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void run<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void run<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template <typename T, int cn>
-    void runAbs(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
-    {
-        typedef typename SumType<T>::R R;
-        caller<T, R, cn, abs_func>(src, buf, out, mask);
-    }
-
-    template void runAbs<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runAbs<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runAbs<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template <typename T> struct Sqr : unary_function<T, T>
-    {
-        __device__ __forceinline__ T operator ()(T x) const
-        {
-            return x * x;
-        }
-    };
-
-    template <typename T, int cn>
-    void runSqr(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
-    {
-        caller<T, double, cn, Sqr>(src, buf, out, mask);
-    }
-
-    template void runSqr<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-
-    template void runSqr<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-    template void runSqr<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
-}
-
-/////////////////////////////////////////////////////////////
-// minMax
-
-namespace minMax
-{
-    __device__ unsigned int blocks_finished = 0;
-
-    // To avoid shared bank conflicts we convert each value into value of
-    // appropriate type (32 bits minimum)
-    template <typename T> struct MinMaxTypeTraits;
-    template <> struct MinMaxTypeTraits<uchar> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<schar> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<ushort> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<short> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<int> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<float> { typedef float best_type; };
-    template <> struct MinMaxTypeTraits<double> { typedef double best_type; };
-
-    template <int BLOCK_SIZE, typename R>
-    struct GlobalReduce
-    {
-        static __device__ void run(R& mymin, R& mymax, R* minval, R* maxval, int tid, int bid, R* sminval, R* smaxval)
-        {
-        #if __CUDA_ARCH__ >= 200
-            if (tid == 0)
-            {
-                detail::cvAtomicMin(minval, mymin);
-                detail::cvAtomicMax(maxval, mymax);
-            }
-        #else
-            __shared__ bool is_last;
-
-            if (tid == 0)
-            {
-                minval[bid] = mymin;
-                maxval[bid] = mymax;
-
-                __threadfence();
-
-                unsigned int ticket = ::atomicAdd(&blocks_finished, 1);
-                is_last = (ticket == gridDim.x * gridDim.y - 1);
-            }
-
-            __syncthreads();
-
-            if (is_last)
-            {
-                int idx = ::min(tid, gridDim.x * gridDim.y - 1);
-
-                mymin = minval[idx];
-                mymax = maxval[idx];
-
-                const minimum<R> minOp;
-                const maximum<R> maxOp;
-                device::reduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax), tid, thrust::make_tuple(minOp, maxOp));
-
-                if (tid == 0)
-                {
-                    minval[0] = mymin;
-                    maxval[0] = mymax;
-
-                    blocks_finished = 0;
-                }
-            }
-        #endif
-        }
-    };
-
-    template <int BLOCK_SIZE, typename T, typename R, class Mask>
-    __global__ void kernel(const PtrStepSz<T> src, const Mask mask, R* minval, R* maxval, const int twidth, const int theight)
-    {
-        __shared__ R sminval[BLOCK_SIZE];
-        __shared__ R smaxval[BLOCK_SIZE];
-
-        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
-        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
-
-        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
-        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
-
-        R mymin = numeric_limits<R>::max();
-        R mymax = -numeric_limits<R>::max();
-
-        const minimum<R> minOp;
-        const maximum<R> maxOp;
-
-        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
-        {
-            const T* ptr = src.ptr(y);
-
-            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
-            {
-                if (mask(y, x))
-                {
-                    const R srcVal = ptr[x];
-
-                    mymin = minOp(mymin, srcVal);
-                    mymax = maxOp(mymax, srcVal);
-                }
-            }
-        }
-
-        device::reduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax), tid, thrust::make_tuple(minOp, maxOp));
-
-        GlobalReduce<BLOCK_SIZE, R>::run(mymin, mymax, minval, maxval, tid, bid, sminval, smaxval);
-    }
-
-    const int threads_x = 32;
-    const int threads_y = 8;
-
-    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
-    {
-        block = dim3(threads_x, threads_y);
-
-        grid = dim3(divUp(cols, block.x * block.y),
-                    divUp(rows, block.y * block.x));
-
-        grid.x = ::min(grid.x, block.x);
-        grid.y = ::min(grid.y, block.y);
-    }
-
-    void getBufSize(int cols, int rows, int& bufcols, int& bufrows)
-    {
-        dim3 block, grid;
-        getLaunchCfg(cols, rows, block, grid);
-
-        bufcols = grid.x * grid.y * sizeof(double);
-        bufrows = 2;
-    }
-
-    __global__ void setDefaultKernel(int* minval_buf, int* maxval_buf)
-    {
-        *minval_buf = numeric_limits<int>::max();
-        *maxval_buf = numeric_limits<int>::min();
-    }
-    __global__ void setDefaultKernel(float* minval_buf, float* maxval_buf)
-    {
-        *minval_buf = numeric_limits<float>::max();
-        *maxval_buf = -numeric_limits<float>::max();
-    }
-    __global__ void setDefaultKernel(double* minval_buf, double* maxval_buf)
-    {
-        *minval_buf = numeric_limits<double>::max();
-        *maxval_buf = -numeric_limits<double>::max();
-    }
-
-    template <typename R>
-    void setDefault(R* minval_buf, R* maxval_buf)
-    {
-        setDefaultKernel<<<1, 1>>>(minval_buf, maxval_buf);
-    }
-
-    template <typename T>
-    void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf)
-    {
-        typedef typename MinMaxTypeTraits<T>::best_type R;
-
-        dim3 block, grid;
-        getLaunchCfg(src.cols, src.rows, block, grid);
-
-        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
-        const int theight = divUp(divUp(src.rows, grid.y), block.y);
-
-        R* minval_buf = (R*) buf.ptr(0);
-        R* maxval_buf = (R*) buf.ptr(1);
-
-        setDefault(minval_buf, maxval_buf);
-
-        if (mask.data)
-            kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, SingleMask(mask), minval_buf, maxval_buf, twidth, theight);
-        else
-            kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, WithOutMask(), minval_buf, maxval_buf, twidth, theight);
-
-        cudaSafeCall( cudaGetLastError() );
-
-        cudaSafeCall( cudaDeviceSynchronize() );
-
-        R minval_, maxval_;
-        cudaSafeCall( cudaMemcpy(&minval_, minval_buf, sizeof(R), cudaMemcpyDeviceToHost) );
-        cudaSafeCall( cudaMemcpy(&maxval_, maxval_buf, sizeof(R), cudaMemcpyDeviceToHost) );
-        *minval = minval_;
-        *maxval = maxval_;
-    }
-
-    template void run<uchar >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<schar >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<ushort>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<short >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<int   >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<float >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-    template void run<double>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
-}
-
-/////////////////////////////////////////////////////////////
-// minMaxLoc
-
-namespace minMaxLoc
-{
-    // To avoid shared bank conflicts we convert each value into value of
-    // appropriate type (32 bits minimum)
-    template <typename T> struct MinMaxTypeTraits;
-    template <> struct MinMaxTypeTraits<unsigned char> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<signed char> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<unsigned short> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<short> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<int> { typedef int best_type; };
-    template <> struct MinMaxTypeTraits<float> { typedef float best_type; };
-    template <> struct MinMaxTypeTraits<double> { typedef double best_type; };
-
-    template <int BLOCK_SIZE, typename T, class Mask>
-    __global__ void kernel_pass_1(const PtrStepSz<T> src, const Mask mask, T* minval, T* maxval, unsigned int* minloc, unsigned int* maxloc, const int twidth, const int theight)
-    {
-        typedef typename MinMaxTypeTraits<T>::best_type work_type;
-
-        __shared__ work_type sminval[BLOCK_SIZE];
-        __shared__ work_type smaxval[BLOCK_SIZE];
-        __shared__ unsigned int sminloc[BLOCK_SIZE];
-        __shared__ unsigned int smaxloc[BLOCK_SIZE];
-
-        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
-        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
-
-        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
-        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
-
-        work_type mymin = numeric_limits<work_type>::max();
-        work_type mymax = -numeric_limits<work_type>::max();
-        unsigned int myminloc = 0;
-        unsigned int mymaxloc = 0;
-
-        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
-        {
-            const T* ptr = src.ptr(y);
-
-            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
-            {
-                if (mask(y, x))
-                {
-                    const work_type srcVal = ptr[x];
-
-                    if (srcVal < mymin)
-                    {
-                        mymin = srcVal;
-                        myminloc = y * src.cols + x;
-                    }
-
-                    if (srcVal > mymax)
-                    {
-                        mymax = srcVal;
-                        mymaxloc = y * src.cols + x;
-                    }
-                }
-            }
-        }
-
-        reduceKeyVal<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax),
-                                 smem_tuple(sminloc, smaxloc), thrust::tie(myminloc, mymaxloc),
-                                 tid,
-                                 thrust::make_tuple(less<work_type>(), greater<work_type>()));
-
-        if (tid == 0)
-        {
-            minval[bid] = (T) mymin;
-            maxval[bid] = (T) mymax;
-            minloc[bid] = myminloc;
-            maxloc[bid] = mymaxloc;
-        }
-    }
-    template <int BLOCK_SIZE, typename T>
-    __global__ void kernel_pass_2(T* minval, T* maxval, unsigned int* minloc, unsigned int* maxloc, int count)
-    {
-        typedef typename MinMaxTypeTraits<T>::best_type work_type;
-
-        __shared__ work_type sminval[BLOCK_SIZE];
-        __shared__ work_type smaxval[BLOCK_SIZE];
-        __shared__ unsigned int sminloc[BLOCK_SIZE];
-        __shared__ unsigned int smaxloc[BLOCK_SIZE];
-
-        unsigned int idx = ::min(threadIdx.x, count - 1);
-
-        work_type mymin = minval[idx];
-        work_type mymax = maxval[idx];
-        unsigned int myminloc = minloc[idx];
-        unsigned int mymaxloc = maxloc[idx];
-
-        reduceKeyVal<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax),
-                                 smem_tuple(sminloc, smaxloc), thrust::tie(myminloc, mymaxloc),
-                                 threadIdx.x,
-                                 thrust::make_tuple(less<work_type>(), greater<work_type>()));
-
-        if (threadIdx.x == 0)
-        {
-            minval[0] = (T) mymin;
-            maxval[0] = (T) mymax;
-            minloc[0] = myminloc;
-            maxloc[0] = mymaxloc;
-        }
-    }
-
-    const int threads_x = 32;
-    const int threads_y = 8;
-
-    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
-    {
-        block = dim3(threads_x, threads_y);
-
-        grid = dim3(divUp(cols, block.x * block.y),
-                    divUp(rows, block.y * block.x));
-
-        grid.x = ::min(grid.x, block.x);
-        grid.y = ::min(grid.y, block.y);
-    }
-
-    void getBufSize(int cols, int rows, size_t elem_size, int& b1cols, int& b1rows, int& b2cols, int& b2rows)
-    {
-        dim3 block, grid;
-        getLaunchCfg(cols, rows, block, grid);
-
-        // For values
-        b1cols = (int)(grid.x * grid.y * elem_size);
-        b1rows = 2;
-
-        // For locations
-        b2cols = grid.x * grid.y * sizeof(int);
-        b2rows = 2;
-    }
-
-    template <typename T>
-    void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf)
-    {
-        dim3 block, grid;
-        getLaunchCfg(src.cols, src.rows, block, grid);
-
-        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
-        const int theight = divUp(divUp(src.rows, grid.y), block.y);
-
-        T* minval_buf = (T*) valbuf.ptr(0);
-        T* maxval_buf = (T*) valbuf.ptr(1);
-        unsigned int* minloc_buf = locbuf.ptr(0);
-        unsigned int* maxloc_buf = locbuf.ptr(1);
-
-        if (mask.data)
-            kernel_pass_1<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, SingleMask(mask), minval_buf, maxval_buf, minloc_buf, maxloc_buf, twidth, theight);
-        else
-            kernel_pass_1<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, WithOutMask(), minval_buf, maxval_buf, minloc_buf, maxloc_buf, twidth, theight);
-
-        cudaSafeCall( cudaGetLastError() );
-
-        kernel_pass_2<threads_x * threads_y><<<1, threads_x * threads_y>>>(minval_buf, maxval_buf, minloc_buf, maxloc_buf, grid.x * grid.y);
-        cudaSafeCall( cudaGetLastError() );
-
-        cudaSafeCall( cudaDeviceSynchronize() );
-
-        T minval_, maxval_;
-        cudaSafeCall( cudaMemcpy(&minval_, minval_buf, sizeof(T), cudaMemcpyDeviceToHost) );
-        cudaSafeCall( cudaMemcpy(&maxval_, maxval_buf, sizeof(T), cudaMemcpyDeviceToHost) );
-        *minval = minval_;
-        *maxval = maxval_;
-
-        unsigned int minloc_, maxloc_;
-        cudaSafeCall( cudaMemcpy(&minloc_, minloc_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost) );
-        cudaSafeCall( cudaMemcpy(&maxloc_, maxloc_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost) );
-        minloc[1] = minloc_ / src.cols; minloc[0] = minloc_ - minloc[1] * src.cols;
-        maxloc[1] = maxloc_ / src.cols; maxloc[0] = maxloc_ - maxloc[1] * src.cols;
-    }
-
-    template void run<unsigned char >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<signed char >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<unsigned short>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<short >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<int   >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<float >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-    template void run<double>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
-}
-
-/////////////////////////////////////////////////////////////
-// countNonZero
-
-namespace countNonZero
-{
-    __device__ unsigned int blocks_finished = 0;
-
-    template <int BLOCK_SIZE, typename T>
-    __global__ void kernel(const PtrStepSz<T> src, unsigned int* count, const int twidth, const int theight)
-    {
-        __shared__ unsigned int scount[BLOCK_SIZE];
-
-        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
-        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
-
-        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
-
-        unsigned int mycount = 0;
-
-        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
-        {
-            const T* ptr = src.ptr(y);
-
-            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
-            {
-                const T srcVal = ptr[x];
-
-                mycount += (srcVal != 0);
-            }
-        }
-
-        device::reduce<BLOCK_SIZE>(scount, mycount, tid, plus<unsigned int>());
-
-    #if __CUDA_ARCH__ >= 200
-        if (tid == 0)
-            ::atomicAdd(count, mycount);
-    #else
-        __shared__ bool is_last;
-        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
-
-        if (tid == 0)
-        {
-            count[bid] = mycount;
-
-            __threadfence();
-
-            unsigned int ticket = ::atomicInc(&blocks_finished, gridDim.x * gridDim.y);
-            is_last = (ticket == gridDim.x * gridDim.y - 1);
-        }
-
-        __syncthreads();
-
-        if (is_last)
-        {
-            mycount = tid < gridDim.x * gridDim.y ? count[tid] : 0;
-
-            device::reduce<BLOCK_SIZE>(scount, mycount, tid, plus<unsigned int>());
-
-            if (tid == 0)
-            {
-                count[0] = mycount;
-
-                blocks_finished = 0;
-            }
-        }
-    #endif
-    }
-
-    const int threads_x = 32;
-    const int threads_y = 8;
-
-    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
-    {
-        block = dim3(threads_x, threads_y);
-
-        grid = dim3(divUp(cols, block.x * block.y),
-                    divUp(rows, block.y * block.x));
-
-        grid.x = ::min(grid.x, block.x);
-        grid.y = ::min(grid.y, block.y);
-    }
-
-    void getBufSize(int cols, int rows, int& bufcols, int& bufrows)
-    {
-        dim3 block, grid;
-        getLaunchCfg(cols, rows, block, grid);
-
-        bufcols = grid.x * grid.y * sizeof(int);
-        bufrows = 1;
-    }
-
-    template <typename T>
-    int run(const PtrStepSzb src, PtrStep<unsigned int> buf)
-    {
-        dim3 block, grid;
-        getLaunchCfg(src.cols, src.rows, block, grid);
-
-        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
-        const int theight = divUp(divUp(src.rows, grid.y), block.y);
-
-        unsigned int* count_buf = buf.ptr(0);
-
-        cudaSafeCall( cudaMemset(count_buf, 0, sizeof(unsigned int)) );
-
-        kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, count_buf, twidth, theight);
-        cudaSafeCall( cudaGetLastError() );
-
-        cudaSafeCall( cudaDeviceSynchronize() );
-
-        unsigned int count;
-        cudaSafeCall(cudaMemcpy(&count, count_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost));
-
-        return count;
-    }
-
-    template int run<uchar >(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<schar >(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<ushort>(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<short >(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<int   >(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<float >(const PtrStepSzb src, PtrStep<unsigned int> buf);
-    template int run<double>(const PtrStepSzb src, PtrStep<unsigned int> buf);
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// reduce
-
-namespace reduce
-{
-    struct Sum
-    {
-        template <typename T>
-        __device__ __forceinline__ T startValue() const
-        {
-            return VecTraits<T>::all(0);
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T operator ()(T a, T b) const
-        {
-            return a + b;
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T result(T r, double) const
-        {
-            return r;
-        }
-
-        __device__ __forceinline__ Sum() {}
-        __device__ __forceinline__ Sum(const Sum&) {}
-    };
-
-    struct Avg
-    {
-        template <typename T>
-        __device__ __forceinline__ T startValue() const
-        {
-            return VecTraits<T>::all(0);
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T operator ()(T a, T b) const
-        {
-            return a + b;
-        }
-
-        template <typename T>
-        __device__ __forceinline__ typename TypeVec<double, VecTraits<T>::cn>::vec_type result(T r, double sz) const
-        {
-            return r / sz;
-        }
-
-        __device__ __forceinline__ Avg() {}
-        __device__ __forceinline__ Avg(const Avg&) {}
-    };
-
-    struct Min
-    {
-        template <typename T>
-        __device__ __forceinline__ T startValue() const
-        {
-            return VecTraits<T>::all(numeric_limits<typename VecTraits<T>::elem_type>::max());
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T operator ()(T a, T b) const
-        {
-            minimum<T> minOp;
-            return minOp(a, b);
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T result(T r, double) const
-        {
-            return r;
-        }
-
-        __device__ __forceinline__ Min() {}
-        __device__ __forceinline__ Min(const Min&) {}
-    };
-
-    struct Max
-    {
-        template <typename T>
-        __device__ __forceinline__ T startValue() const
-        {
-            return VecTraits<T>::all(-numeric_limits<typename VecTraits<T>::elem_type>::max());
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T operator ()(T a, T b) const
-        {
-            maximum<T> maxOp;
-            return maxOp(a, b);
-        }
-
-        template <typename T>
-        __device__ __forceinline__ T result(T r, double) const
-        {
-            return r;
-        }
-
-        __device__ __forceinline__ Max() {}
-        __device__ __forceinline__ Max(const Max&) {}
-    };
-
-    ///////////////////////////////////////////////////////////
-
-    template <typename T, typename S, typename D, class Op>
-    __global__ void rowsKernel(const PtrStepSz<T> src, D* dst, const Op op)
-    {
-        __shared__ S smem[16 * 16];
-
-        const int x = blockIdx.x * 16 + threadIdx.x;
-
-        S myVal = op.template startValue<S>();
-
-        if (x < src.cols)
-        {
-            for (int y = threadIdx.y; y < src.rows; y += 16)
-            {
-                S srcVal = src(y, x);
-                myVal = op(myVal, srcVal);
-            }
-        }
-
-        smem[threadIdx.x * 16 + threadIdx.y] = myVal;
-
-        __syncthreads();
-
-        volatile S* srow = smem + threadIdx.y * 16;
-
-        myVal = srow[threadIdx.x];
-        device::reduce<16>(srow, myVal, threadIdx.x, op);
-
-        if (threadIdx.x == 0)
-            srow[0] = myVal;
-
-        __syncthreads();
-
-        if (threadIdx.y == 0 && x < src.cols)
-            dst[x] = (D) op.result(smem[threadIdx.x * 16], src.rows);
-    }
-
-    template <typename T, typename S, typename D, class Op>
-    void rowsCaller(PtrStepSz<T> src, D* dst, cudaStream_t stream)
-    {
-        const dim3 block(16, 16);
-        const dim3 grid(divUp(src.cols, block.x));
-
-        Op op;
-        rowsKernel<T, S, D, Op><<<grid, block, 0, stream>>>(src, dst, op);
-        cudaSafeCall( cudaGetLastError() );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-
-    template <typename T, typename S, typename D>
-    void rows(PtrStepSzb src, void* dst, int op, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSz<T> src, D* dst, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            rowsCaller<T, S, D, Sum>,
-            rowsCaller<T, S, D, Avg>,
-            rowsCaller<T, S, D, Max>,
-            rowsCaller<T, S, D, Min>
-        };
-
-        funcs[op]((PtrStepSz<T>) src, (D*) dst, stream);
-    }
-
-    template void rows<unsigned char, int, unsigned char>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned char, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned char, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned char, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    template void rows<unsigned short, int, unsigned short>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned short, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned short, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<unsigned short, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    template void rows<short, int, short>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<short, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<short, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<short, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    template void rows<int, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<int, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<int, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    template void rows<float, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-    template void rows<float, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    template void rows<double, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
-
-    ///////////////////////////////////////////////////////////
-
-    template <int BLOCK_SIZE, typename T, typename S, typename D, int cn, class Op>
-    __global__ void colsKernel(const PtrStepSz<typename TypeVec<T, cn>::vec_type> src, typename TypeVec<D, cn>::vec_type* dst, const Op op)
-    {
-        typedef typename TypeVec<T, cn>::vec_type src_type;
-        typedef typename TypeVec<S, cn>::vec_type work_type;
-        typedef typename TypeVec<D, cn>::vec_type dst_type;
-
-        __shared__ S smem[BLOCK_SIZE * cn];
-
-        const int y = blockIdx.x;
-
-        const src_type* srcRow = src.ptr(y);
-
-        work_type myVal = op.template startValue<work_type>();
-
-        for (int x = threadIdx.x; x < src.cols; x += BLOCK_SIZE)
-            myVal = op(myVal, saturate_cast<work_type>(srcRow[x]));
-
-        device::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(myVal), threadIdx.x, detail::Unroll<cn>::op(op));
-
-        if (threadIdx.x == 0)
-            dst[y] = saturate_cast<dst_type>(op.result(myVal, src.cols));
-    }
-
-    template <typename T, typename S, typename D, int cn, class Op> void colsCaller(PtrStepSzb src, void* dst, cudaStream_t stream)
-    {
-        const int BLOCK_SIZE = 256;
-
-        const dim3 block(BLOCK_SIZE);
-        const dim3 grid(src.rows);
-
-        Op op;
-        colsKernel<BLOCK_SIZE, T, S, D, cn, Op><<<grid, block, 0, stream>>>((PtrStepSz<typename TypeVec<T, cn>::vec_type>) src, (typename TypeVec<D, cn>::vec_type*) dst, op);
-        cudaSafeCall( cudaGetLastError() );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-
-    }
-
-    template <typename T, typename S, typename D> void cols(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream)
-    {
-        typedef void (*func_t)(PtrStepSzb src, void* dst, cudaStream_t stream);
-        static const func_t funcs[5][4] =
-        {
-            {0,0,0,0},
-            {colsCaller<T, S, D, 1, Sum>, colsCaller<T, S, D, 1, Avg>, colsCaller<T, S, D, 1, Max>, colsCaller<T, S, D, 1, Min>},
-            {colsCaller<T, S, D, 2, Sum>, colsCaller<T, S, D, 2, Avg>, colsCaller<T, S, D, 2, Max>, colsCaller<T, S, D, 2, Min>},
-            {colsCaller<T, S, D, 3, Sum>, colsCaller<T, S, D, 3, Avg>, colsCaller<T, S, D, 3, Max>, colsCaller<T, S, D, 3, Min>},
-            {colsCaller<T, S, D, 4, Sum>, colsCaller<T, S, D, 4, Avg>, colsCaller<T, S, D, 4, Max>, colsCaller<T, S, D, 4, Min>},
-        };
-
-        funcs[cn][op](src, dst, stream);
-    }
-
-    template void cols<unsigned char, int, unsigned char>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned char, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned char, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned char, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-
-    template void cols<unsigned short, int, unsigned short>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned short, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned short, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<unsigned short, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-
-    template void cols<short, int, short>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<short, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<short, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<short, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-
-    template void cols<int, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<int, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<int, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-
-    template void cols<float, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-    template void cols<float, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-
-    template void cols<double, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
-}
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/resize.cu b/modules/gpu/src/cuda/resize.cu
deleted file mode 100644
index 85d8e3fea5..0000000000
--- a/modules/gpu/src/cuda/resize.cu
+++ /dev/null
@@ -1,302 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/filters.hpp"
-#include <cfloat>
-#include <opencv2/gpu/device/scan.hpp>
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename Ptr2D, typename T> __global__ void resize(const Ptr2D src, float fx, float fy, PtrStepSz<T> dst)
-        {
-            const int x = blockDim.x * blockIdx.x + threadIdx.x;
-            const int y = blockDim.y * blockIdx.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                const float xcoo = x * fx;
-                const float ycoo = y * fy;
-
-                dst(y, x) = saturate_cast<T>(src(ycoo, xcoo));
-            }
-        }
-
-        template <typename Ptr2D, typename T> __global__ void resize_area(const Ptr2D src, float fx, float fy, PtrStepSz<T> dst)
-        {
-            const int x = blockDim.x * blockIdx.x + threadIdx.x;
-            const int y = blockDim.y * blockIdx.y + threadIdx.y;
-
-            if (x < dst.cols && y < dst.rows)
-            {
-                dst(y, x) = saturate_cast<T>(src(y, x));
-            }
-        }
-
-        template <template <typename> class Filter, typename T> struct ResizeDispatcherStream
-        {
-            static void call(PtrStepSz<T> src, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream)
-            {
-                dim3 block(32, 8);
-                dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
-
-                BrdReplicate<T> brd(src.rows, src.cols);
-                BorderReader< PtrStep<T>, BrdReplicate<T> > brdSrc(src, brd);
-                Filter< BorderReader< PtrStep<T>, BrdReplicate<T> > > filteredSrc(brdSrc, fx, fy);
-
-                resize<<<grid, block, 0, stream>>>(filteredSrc, fx, fy, dst);
-                cudaSafeCall( cudaGetLastError() );
-            }
-        };
-
-        template <typename T> struct ResizeDispatcherStream<AreaFilter, T>
-        {
-            static void call(PtrStepSz<T> src, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream)
-            {
-                dim3 block(32, 8);
-                dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
-
-                BrdConstant<T> brd(src.rows, src.cols);
-                BorderReader< PtrStep<T>, BrdConstant<T> > brdSrc(src, brd);
-                AreaFilter< BorderReader< PtrStep<T>, BrdConstant<T> > > filteredSrc(brdSrc, fx, fy);
-                resize_area<<<grid, block, 0, stream>>>(filteredSrc, fx, fy, dst);
-                cudaSafeCall( cudaGetLastError() );
-                if (stream == 0)
-                    cudaSafeCall( cudaDeviceSynchronize() );
-            }
-        };
-
-        template <typename T> struct ResizeDispatcherStream<IntegerAreaFilter, T>
-        {
-            static void call(PtrStepSz<T> src, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream)
-            {
-                dim3 block(32, 8);
-                dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
-                BrdConstant<T> brd(src.rows, src.cols);
-                BorderReader< PtrStep<T>, BrdConstant<T> > brdSrc(src, brd);
-                IntegerAreaFilter< BorderReader< PtrStep<T>, BrdConstant<T> > > filteredSrc(brdSrc, fx, fy);
-                resize_area<<<grid, block, 0, stream>>>(filteredSrc, fx, fy, dst);
-                cudaSafeCall( cudaGetLastError() );
-                if (stream == 0)
-                    cudaSafeCall( cudaDeviceSynchronize() );
-            }
-        };
-
-        template <template <typename> class Filter, typename T> struct ResizeDispatcherNonStream
-        {
-            static void call(PtrStepSz<T> src, PtrStepSz<T> srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSz<T> dst)
-            {
-                (void)srcWhole;
-                (void)xoff;
-                (void)yoff;
-
-                dim3 block(32, 8);
-                dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
-
-                BrdReplicate<T> brd(src.rows, src.cols);
-                BorderReader< PtrStep<T>, BrdReplicate<T> > brdSrc(src, brd);
-                Filter< BorderReader< PtrStep<T>, BrdReplicate<T> > > filteredSrc(brdSrc);
-
-                resize<<<grid, block>>>(filteredSrc, fx, fy, dst);
-                cudaSafeCall( cudaGetLastError() );
-
-                cudaSafeCall( cudaDeviceSynchronize() );
-            }
-        };
-
-        #define OPENCV_GPU_IMPLEMENT_RESIZE_TEX(type) \
-            texture< type , cudaTextureType2D> tex_resize_ ## type (0, cudaFilterModePoint, cudaAddressModeClamp); \
-            struct tex_resize_ ## type ## _reader \
-            { \
-                typedef type elem_type; \
-                typedef int index_type; \
-                const int xoff; \
-                const int yoff; \
-                __host__ tex_resize_ ## type ## _reader(int xoff_, int yoff_) : xoff(xoff_), yoff(yoff_) {} \
-                __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const \
-                { \
-                    return tex2D(tex_resize_ ## type, x + xoff, y + yoff); \
-                } \
-            }; \
-            template <template <typename> class Filter> struct ResizeDispatcherNonStream<Filter, type > \
-            { \
-                static void call(PtrStepSz< type > src, PtrStepSz< type > srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSz< type > dst) \
-                { \
-                    dim3 block(32, 8); \
-                    dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \
-                    bindTexture(&tex_resize_ ## type, srcWhole); \
-                    tex_resize_ ## type ## _reader texSrc(xoff, yoff); \
-                    if (srcWhole.cols == src.cols && srcWhole.rows == src.rows) \
-                    { \
-                        Filter<tex_resize_ ## type ## _reader> filteredSrc(texSrc); \
-                        resize<<<grid, block>>>(filteredSrc, fx, fy, dst); \
-                    } \
-                    else \
-                    { \
-                        BrdReplicate< type > brd(src.rows, src.cols); \
-                        BorderReader<tex_resize_ ## type ## _reader, BrdReplicate< type > > brdSrc(texSrc, brd); \
-                        Filter< BorderReader<tex_resize_ ## type ## _reader, BrdReplicate< type > > > filteredSrc(brdSrc); \
-                        resize<<<grid, block>>>(filteredSrc, fx, fy, dst); \
-                    } \
-                    cudaSafeCall( cudaGetLastError() ); \
-                    cudaSafeCall( cudaDeviceSynchronize() ); \
-                } \
-            };
-
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(uchar)
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(uchar4)
-
-        //OPENCV_GPU_IMPLEMENT_RESIZE_TEX(schar)
-        //OPENCV_GPU_IMPLEMENT_RESIZE_TEX(char4)
-
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(ushort)
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(ushort4)
-
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(short)
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(short4)
-
-        //OPENCV_GPU_IMPLEMENT_RESIZE_TEX(int)
-        //OPENCV_GPU_IMPLEMENT_RESIZE_TEX(int4)
-
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(float)
-        OPENCV_GPU_IMPLEMENT_RESIZE_TEX(float4)
-
-        #undef OPENCV_GPU_IMPLEMENT_RESIZE_TEX
-
-        template <template <typename> class Filter, typename T> struct ResizeDispatcher
-        {
-            static void call(PtrStepSz<T> src, PtrStepSz<T> srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream)
-            {
-                if (stream == 0)
-                    ResizeDispatcherNonStream<Filter, T>::call(src, srcWhole, xoff, yoff, fx, fy, dst);
-                else
-                    ResizeDispatcherStream<Filter, T>::call(src, fx, fy, dst, stream);
-            }
-        };
-
-        template <typename T> struct ResizeDispatcher<AreaFilter, T>
-        {
-            static void call(PtrStepSz<T> src, PtrStepSz<T> srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream)
-            {
-                (void)srcWhole;
-                (void)xoff;
-                (void)yoff;
-                int iscale_x = (int)round(fx);
-                int iscale_y = (int)round(fy);
-
-                if( std::abs(fx - iscale_x) < FLT_MIN && std::abs(fy - iscale_y) < FLT_MIN)
-                    ResizeDispatcherStream<IntegerAreaFilter, T>::call(src, fx, fy, dst, stream);
-                else
-                    ResizeDispatcherStream<AreaFilter, T>::call(src, fx, fy, dst, stream);
-            }
-        };
-
-        template <typename T> void resize_gpu(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy,
-            PtrStepSzb dst, int interpolation, cudaStream_t stream)
-        {
-            typedef void (*caller_t)(PtrStepSz<T> src, PtrStepSz<T> srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSz<T> dst, cudaStream_t stream);
-
-            static const caller_t callers[4] =
-            {
-                ResizeDispatcher<PointFilter, T>::call,
-                ResizeDispatcher<LinearFilter, T>::call,
-                ResizeDispatcher<CubicFilter, T>::call,
-                ResizeDispatcher<AreaFilter, T>::call
-            };
-            // chenge to linear if area interpolation upscaling
-            if (interpolation == 3 && (fx <= 1.f || fy <= 1.f))
-                interpolation = 1;
-
-            callers[interpolation](static_cast< PtrStepSz<T> >(src), static_cast< PtrStepSz<T> >(srcWhole), xoff, yoff, fx, fy,
-                static_cast< PtrStepSz<T> >(dst), stream);
-        }
-
-        template void resize_gpu<uchar >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<uchar2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<uchar3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<uchar4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        //template void resize_gpu<schar>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<char2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<char3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<char4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        template void resize_gpu<ushort >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<ushort2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<ushort3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<ushort4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        template void resize_gpu<short >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<short2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<short3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<short4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        //template void resize_gpu<int >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<int2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<int3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<int4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        template void resize_gpu<float >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        //template void resize_gpu<float2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<float3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-        template void resize_gpu<float4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        template<typename T> struct scan_traits{};
-
-        template<> struct scan_traits<uchar>
-        {
-            typedef float scan_line_type;
-        };
-
-    } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
-
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/cuda/rgb_to_yv12.cu b/modules/gpu/src/cuda/rgb_to_yv12.cu
deleted file mode 100644
index 5fd0217dc9..0000000000
--- a/modules/gpu/src/cuda/rgb_to_yv12.cu
+++ /dev/null
@@ -1,175 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace video_encoding
-    {
-        __device__ __forceinline__ void rgbtoy(const uchar b, const uchar g, const uchar r, uchar& y)
-        {
-            y = static_cast<uchar>(((int)(30 * r) + (int)(59 * g) + (int)(11 * b)) / 100);
-        }
-
-        __device__ __forceinline__ void rgbtoyuv(const uchar b, const uchar g, const uchar r, uchar& y, uchar& u, uchar& v)
-        {
-            rgbtoy(b, g, r, y);
-            u = static_cast<uchar>(((int)(-17 * r) - (int)(33 * g) + (int)(50 * b) + 12800) / 100);
-            v = static_cast<uchar>(((int)(50 * r) - (int)(42 * g) - (int)(8 * b) + 12800) / 100);
-        }
-
-        __global__ void Gray_to_YV12(const PtrStepSzb src, PtrStepb dst)
-        {
-            const int x = (blockIdx.x * blockDim.x + threadIdx.x) * 2;
-            const int y = (blockIdx.y * blockDim.y + threadIdx.y) * 2;
-
-            if (x + 1 >= src.cols || y + 1 >= src.rows)
-                return;
-
-            // get pointers to the data
-            const size_t planeSize = src.rows * dst.step;
-            PtrStepb y_plane(dst.data, dst.step);
-            PtrStepb u_plane(y_plane.data + planeSize, dst.step / 2);
-            PtrStepb v_plane(u_plane.data + (planeSize / 4), dst.step / 2);
-
-            uchar pix;
-            uchar y_val, u_val, v_val;
-
-            pix = src(y, x);
-            rgbtoy(pix, pix, pix, y_val);
-            y_plane(y, x) = y_val;
-
-            pix = src(y, x + 1);
-            rgbtoy(pix, pix, pix, y_val);
-            y_plane(y, x + 1) = y_val;
-
-            pix = src(y + 1, x);
-            rgbtoy(pix, pix, pix, y_val);
-            y_plane(y + 1, x) = y_val;
-
-            pix = src(y + 1, x + 1);
-            rgbtoyuv(pix, pix, pix, y_val, u_val, v_val);
-            y_plane(y + 1, x + 1) = y_val;
-            u_plane(y / 2, x / 2) = u_val;
-            v_plane(y / 2, x / 2) = v_val;
-        }
-
-        template <typename T>
-        __global__ void BGR_to_YV12(const PtrStepSz<T> src, PtrStepb dst)
-        {
-            const int x = (blockIdx.x * blockDim.x + threadIdx.x) * 2;
-            const int y = (blockIdx.y * blockDim.y + threadIdx.y) * 2;
-
-            if (x + 1 >= src.cols || y + 1 >= src.rows)
-                return;
-
-            // get pointers to the data
-            const size_t planeSize = src.rows * dst.step;
-            PtrStepb y_plane(dst.data, dst.step);
-            PtrStepb u_plane(y_plane.data + planeSize, dst.step / 2);
-            PtrStepb v_plane(u_plane.data + (planeSize / 4), dst.step / 2);
-
-            T pix;
-            uchar y_val, u_val, v_val;
-
-            pix = src(y, x);
-            rgbtoy(pix.z, pix.y, pix.x, y_val);
-            y_plane(y, x) = y_val;
-
-            pix = src(y, x + 1);
-            rgbtoy(pix.z, pix.y, pix.x, y_val);
-            y_plane(y, x + 1) = y_val;
-
-            pix = src(y + 1, x);
-            rgbtoy(pix.z, pix.y, pix.x, y_val);
-            y_plane(y + 1, x) = y_val;
-
-            pix = src(y + 1, x + 1);
-            rgbtoyuv(pix.z, pix.y, pix.x, y_val, u_val, v_val);
-            y_plane(y + 1, x + 1) = y_val;
-            u_plane(y / 2, x / 2) = u_val;
-            v_plane(y / 2, x / 2) = v_val;
-        }
-
-        void Gray_to_YV12_caller(const PtrStepSzb src, PtrStepb dst)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(src.cols, block.x * 2), divUp(src.rows, block.y * 2));
-
-            Gray_to_YV12<<<grid, block>>>(src, dst);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-        template <int cn>
-        void BGR_to_YV12_caller(const PtrStepSzb src, PtrStepb dst)
-        {
-            typedef typename TypeVec<uchar, cn>::vec_type src_t;
-
-            dim3 block(32, 8);
-            dim3 grid(divUp(src.cols, block.x * 2), divUp(src.rows, block.y * 2));
-
-            BGR_to_YV12<<<grid, block>>>(static_cast< PtrStepSz<src_t> >(src), dst);
-            cudaSafeCall( cudaGetLastError() );
-
-            cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void YV12_gpu(const PtrStepSzb src, int cn, PtrStepSzb dst)
-        {
-            typedef void (*func_t)(const PtrStepSzb src, PtrStepb dst);
-
-            static const func_t funcs[] =
-            {
-                0, Gray_to_YV12_caller, 0, BGR_to_YV12_caller<3>, BGR_to_YV12_caller<4>
-            };
-
-            funcs[cn](src, dst);
-        }
-    }
-}}}
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/cuda/safe_call.hpp b/modules/gpu/src/cuda/safe_call.hpp
deleted file mode 100644
index 7fae710820..0000000000
--- a/modules/gpu/src/cuda/safe_call.hpp
+++ /dev/null
@@ -1,95 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_CUDA_SAFE_CALL_HPP__
-#define __OPENCV_CUDA_SAFE_CALL_HPP__
-
-#include <cuda_runtime_api.h>
-#include <cufft.h>
-#include <cublas.h>
-#include "NCV.hpp"
-
-#if defined(__GNUC__)
-    #define nppSafeCall(expr)  ___nppSafeCall(expr, __FILE__, __LINE__, __func__)
-    #define ncvSafeCall(expr)  ___ncvSafeCall(expr, __FILE__, __LINE__, __func__)
-    #define cufftSafeCall(expr)  ___cufftSafeCall(expr, __FILE__, __LINE__, __func__)
-    #define cublasSafeCall(expr)  ___cublasSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define nppSafeCall(expr)  ___nppSafeCall(expr, __FILE__, __LINE__)
-    #define ncvSafeCall(expr)  ___ncvSafeCall(expr, __FILE__, __LINE__)
-    #define cufftSafeCall(expr)  ___cufftSafeCall(expr, __FILE__, __LINE__)
-    #define cublasSafeCall(expr)  ___cublasSafeCall(expr, __FILE__, __LINE__)
-#endif
-
-namespace cv { namespace gpu
-{
-    void nppError(int err, const char *file, const int line, const char *func = "");
-    void ncvError(int err, const char *file, const int line, const char *func = "");
-    void cufftError(int err, const char *file, const int line, const char *func = "");
-    void cublasError(int err, const char *file, const int line, const char *func = "");
-}}
-
-static inline void ___nppSafeCall(int err, const char *file, const int line, const char *func = "")
-{
-    if (err < 0)
-        cv::gpu::nppError(err, file, line, func);
-}
-
-static inline void ___ncvSafeCall(int err, const char *file, const int line, const char *func = "")
-{
-    if (NCV_SUCCESS != err)
-        cv::gpu::ncvError(err, file, line, func);
-}
-
-static inline void ___cufftSafeCall(cufftResult_t err, const char *file, const int line, const char *func = "")
-{
-    if (CUFFT_SUCCESS != err)
-        cv::gpu::cufftError(err, file, line, func);
-}
-
-static inline void ___cublasSafeCall(cublasStatus_t err, const char *file, const int line, const char *func = "")
-{
-    if (CUBLAS_STATUS_SUCCESS != err)
-        cv::gpu::cublasError(err, file, line, func);
-}
-
-#endif /* __OPENCV_CUDA_SAFE_CALL_HPP__ */
diff --git a/modules/gpu/src/cuvid_video_source.cpp b/modules/gpu/src/cuvid_video_source.cpp
deleted file mode 100644
index 7d45b8fee9..0000000000
--- a/modules/gpu/src/cuvid_video_source.cpp
+++ /dev/null
@@ -1,63 +0,0 @@
-#include "cuvid_video_source.h"
-#include "cu_safe_call.h"
-
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
-
-cv::gpu::detail::CuvidVideoSource::CuvidVideoSource(const std::string& fname)
-{
-    CUVIDSOURCEPARAMS params;
-    std::memset(&params, 0, sizeof(CUVIDSOURCEPARAMS));
-
-    // Fill parameter struct
-    params.pUserData = this;                        // will be passed to data handlers
-    params.pfnVideoDataHandler = HandleVideoData;   // our local video-handler callback
-    params.pfnAudioDataHandler = 0;
-
-    // now create the actual source
-    CUresult res = cuvidCreateVideoSource(&videoSource_, fname.c_str(), &params);
-    if (res == CUDA_ERROR_INVALID_SOURCE)
-        throw std::runtime_error("Unsupported video source");
-    cuSafeCall( res );
-
-    CUVIDEOFORMAT vidfmt;
-    cuSafeCall( cuvidGetSourceVideoFormat(videoSource_, &vidfmt, 0) );
-
-    format_.codec = static_cast<VideoReader_GPU::Codec>(vidfmt.codec);
-    format_.chromaFormat = static_cast<VideoReader_GPU::ChromaFormat>(vidfmt.chroma_format);
-    format_.width = vidfmt.coded_width;
-    format_.height = vidfmt.coded_height;
-}
-
-cv::gpu::VideoReader_GPU::FormatInfo cv::gpu::detail::CuvidVideoSource::format() const
-{
-    return format_;
-}
-
-void cv::gpu::detail::CuvidVideoSource::start()
-{
-    cuSafeCall( cuvidSetVideoSourceState(videoSource_, cudaVideoState_Started) );
-}
-
-void cv::gpu::detail::CuvidVideoSource::stop()
-{
-    cuSafeCall( cuvidSetVideoSourceState(videoSource_, cudaVideoState_Stopped) );
-}
-
-bool cv::gpu::detail::CuvidVideoSource::isStarted() const
-{
-    return (cuvidGetVideoSourceState(videoSource_) == cudaVideoState_Started);
-}
-
-bool cv::gpu::detail::CuvidVideoSource::hasError() const
-{
-    return (cuvidGetVideoSourceState(videoSource_) == cudaVideoState_Error);
-}
-
-int CUDAAPI cv::gpu::detail::CuvidVideoSource::HandleVideoData(void* userData, CUVIDSOURCEDATAPACKET* packet)
-{
-    CuvidVideoSource* thiz = static_cast<CuvidVideoSource*>(userData);
-
-    return thiz->parseVideoData(packet->payload, packet->payload_size, (packet->flags & CUVID_PKT_ENDOFSTREAM) != 0);
-}
-
-#endif // defined(HAVE_CUDA) && !defined(__APPLE__)
diff --git a/modules/gpu/src/error.cpp b/modules/gpu/src/error.cpp
deleted file mode 100644
index e01be8010a..0000000000
--- a/modules/gpu/src/error.cpp
+++ /dev/null
@@ -1,248 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#ifdef HAVE_CUDA
-
-namespace
-{
-    #define error_entry(entry)  { entry, #entry }
-
-    struct ErrorEntry
-    {
-        int code;
-        std::string str;
-    };
-
-    struct ErrorEntryComparer
-    {
-        int code;
-        ErrorEntryComparer(int code_) : code(code_) {};
-        bool operator()(const ErrorEntry& e) const { return e.code == code; }
-    };
-
-    std::string getErrorString(int code, const ErrorEntry* errors, size_t n)
-    {
-        size_t idx = std::find_if(errors, errors + n, ErrorEntryComparer(code)) - errors;
-
-        const std::string& msg = (idx != n) ? errors[idx].str : std::string("Unknown error code");
-
-        std::ostringstream ostr;
-        ostr << msg << " [Code = " << code << "]";
-
-        return ostr.str();
-    }
-
-    //////////////////////////////////////////////////////////////////////////
-    // NPP errors
-
-    const ErrorEntry npp_errors [] =
-    {
-        error_entry( NPP_NOT_SUPPORTED_MODE_ERROR ),
-        error_entry( NPP_ROUND_MODE_NOT_SUPPORTED_ERROR ),
-        error_entry( NPP_RESIZE_NO_OPERATION_ERROR ),
-
-#if defined (_MSC_VER)
-        error_entry( NPP_NOT_SUFFICIENT_COMPUTE_CAPABILITY ),
-#endif
-
-        error_entry( NPP_BAD_ARG_ERROR ),
-        error_entry( NPP_LUT_NUMBER_OF_LEVELS_ERROR ),
-        error_entry( NPP_TEXTURE_BIND_ERROR ),
-        error_entry( NPP_COEFF_ERROR ),
-        error_entry( NPP_RECT_ERROR ),
-        error_entry( NPP_QUAD_ERROR ),
-        error_entry( NPP_WRONG_INTERSECTION_ROI_ERROR ),
-        error_entry( NPP_NOT_EVEN_STEP_ERROR ),
-        error_entry( NPP_INTERPOLATION_ERROR ),
-        error_entry( NPP_RESIZE_FACTOR_ERROR ),
-        error_entry( NPP_HAAR_CLASSIFIER_PIXEL_MATCH_ERROR ),
-        error_entry( NPP_MEMFREE_ERR ),
-        error_entry( NPP_MEMSET_ERR ),
-        error_entry( NPP_MEMCPY_ERROR ),
-        error_entry( NPP_MEM_ALLOC_ERR ),
-        error_entry( NPP_HISTO_NUMBER_OF_LEVELS_ERROR ),
-        error_entry( NPP_MIRROR_FLIP_ERR ),
-        error_entry( NPP_INVALID_INPUT ),
-        error_entry( NPP_ALIGNMENT_ERROR ),
-        error_entry( NPP_STEP_ERROR ),
-        error_entry( NPP_SIZE_ERROR ),
-        error_entry( NPP_POINTER_ERROR ),
-        error_entry( NPP_NULL_POINTER_ERROR ),
-        error_entry( NPP_CUDA_KERNEL_EXECUTION_ERROR ),
-        error_entry( NPP_NOT_IMPLEMENTED_ERROR ),
-        error_entry( NPP_ERROR ),
-        error_entry( NPP_NO_ERROR ),
-        error_entry( NPP_SUCCESS ),
-        error_entry( NPP_WARNING ),
-        error_entry( NPP_WRONG_INTERSECTION_QUAD_WARNING ),
-        error_entry( NPP_MISALIGNED_DST_ROI_WARNING ),
-        error_entry( NPP_AFFINE_QUAD_INCORRECT_WARNING ),
-        error_entry( NPP_DOUBLE_SIZE_WARNING ),
-        error_entry( NPP_ODD_ROI_WARNING )
-    };
-
-    const size_t npp_error_num = sizeof(npp_errors) / sizeof(npp_errors[0]);
-
-    //////////////////////////////////////////////////////////////////////////
-    // NCV errors
-
-    const ErrorEntry ncv_errors [] =
-    {
-        error_entry( NCV_SUCCESS ),
-        error_entry( NCV_UNKNOWN_ERROR ),
-        error_entry( NCV_CUDA_ERROR ),
-        error_entry( NCV_NPP_ERROR ),
-        error_entry( NCV_FILE_ERROR ),
-        error_entry( NCV_NULL_PTR ),
-        error_entry( NCV_INCONSISTENT_INPUT ),
-        error_entry( NCV_TEXTURE_BIND_ERROR ),
-        error_entry( NCV_DIMENSIONS_INVALID ),
-        error_entry( NCV_INVALID_ROI ),
-        error_entry( NCV_INVALID_STEP ),
-        error_entry( NCV_INVALID_SCALE ),
-        error_entry( NCV_INVALID_SCALE ),
-        error_entry( NCV_ALLOCATOR_NOT_INITIALIZED ),
-        error_entry( NCV_ALLOCATOR_BAD_ALLOC ),
-        error_entry( NCV_ALLOCATOR_BAD_DEALLOC ),
-        error_entry( NCV_ALLOCATOR_INSUFFICIENT_CAPACITY ),
-        error_entry( NCV_ALLOCATOR_DEALLOC_ORDER ),
-        error_entry( NCV_ALLOCATOR_BAD_REUSE ),
-        error_entry( NCV_MEM_COPY_ERROR ),
-        error_entry( NCV_MEM_RESIDENCE_ERROR ),
-        error_entry( NCV_MEM_INSUFFICIENT_CAPACITY ),
-        error_entry( NCV_HAAR_INVALID_PIXEL_STEP ),
-        error_entry( NCV_HAAR_TOO_MANY_FEATURES_IN_CLASSIFIER ),
-        error_entry( NCV_HAAR_TOO_MANY_FEATURES_IN_CASCADE ),
-        error_entry( NCV_HAAR_TOO_LARGE_FEATURES ),
-        error_entry( NCV_HAAR_XML_LOADING_EXCEPTION ),
-        error_entry( NCV_NOIMPL_HAAR_TILTED_FEATURES ),
-        error_entry( NCV_WARNING_HAAR_DETECTIONS_VECTOR_OVERFLOW ),
-        error_entry( NPPST_SUCCESS ),
-        error_entry( NPPST_ERROR ),
-        error_entry( NPPST_CUDA_KERNEL_EXECUTION_ERROR ),
-        error_entry( NPPST_NULL_POINTER_ERROR ),
-        error_entry( NPPST_TEXTURE_BIND_ERROR ),
-        error_entry( NPPST_MEMCPY_ERROR ),
-        error_entry( NPPST_MEM_ALLOC_ERR ),
-        error_entry( NPPST_MEMFREE_ERR ),
-        error_entry( NPPST_INVALID_ROI ),
-        error_entry( NPPST_INVALID_STEP ),
-        error_entry( NPPST_INVALID_SCALE ),
-        error_entry( NPPST_MEM_INSUFFICIENT_BUFFER ),
-        error_entry( NPPST_MEM_RESIDENCE_ERROR ),
-        error_entry( NPPST_MEM_INTERNAL_ERROR )
-    };
-
-    const size_t ncv_error_num = sizeof(ncv_errors) / sizeof(ncv_errors[0]);
-
-    //////////////////////////////////////////////////////////////////////////
-    // CUFFT errors
-
-    const ErrorEntry cufft_errors[] =
-    {
-        error_entry( CUFFT_INVALID_PLAN ),
-        error_entry( CUFFT_ALLOC_FAILED ),
-        error_entry( CUFFT_INVALID_TYPE ),
-        error_entry( CUFFT_INVALID_VALUE ),
-        error_entry( CUFFT_INTERNAL_ERROR ),
-        error_entry( CUFFT_EXEC_FAILED ),
-        error_entry( CUFFT_SETUP_FAILED ),
-        error_entry( CUFFT_INVALID_SIZE ),
-        error_entry( CUFFT_UNALIGNED_DATA )
-    };
-
-    const int cufft_error_num = sizeof(cufft_errors) / sizeof(cufft_errors[0]);
-
-    //////////////////////////////////////////////////////////////////////////
-    // CUBLAS errors
-
-    const ErrorEntry cublas_errors[] =
-    {
-        error_entry( CUBLAS_STATUS_SUCCESS ),
-        error_entry( CUBLAS_STATUS_NOT_INITIALIZED ),
-        error_entry( CUBLAS_STATUS_ALLOC_FAILED ),
-        error_entry( CUBLAS_STATUS_INVALID_VALUE ),
-        error_entry( CUBLAS_STATUS_ARCH_MISMATCH ),
-        error_entry( CUBLAS_STATUS_MAPPING_ERROR ),
-        error_entry( CUBLAS_STATUS_EXECUTION_FAILED ),
-        error_entry( CUBLAS_STATUS_INTERNAL_ERROR )
-    };
-
-    const int cublas_error_num = sizeof(cublas_errors) / sizeof(cublas_errors[0]);
-}
-
-namespace cv
-{
-    namespace gpu
-    {
-        void nppError(int code, const char *file, const int line, const char *func)
-        {
-            std::string msg = getErrorString(code, npp_errors, npp_error_num);
-            cv::gpu::error(msg.c_str(), file, line, func);
-        }
-
-        void ncvError(int code, const char *file, const int line, const char *func)
-        {
-            std::string msg = getErrorString(code, ncv_errors, ncv_error_num);
-            cv::gpu::error(msg.c_str(), file, line, func);
-        }
-
-        void cufftError(int code, const char *file, const int line, const char *func)
-        {
-            std::string msg = getErrorString(code, cufft_errors, cufft_error_num);
-            cv::gpu::error(msg.c_str(), file, line, func);
-        }
-
-        void cublasError(int code, const char *file, const int line, const char *func)
-        {
-            std::string msg = getErrorString(code, cublas_errors, cublas_error_num);
-            cv::gpu::error(msg.c_str(), file, line, func);
-        }
-    }
-}
-
-#endif
diff --git a/modules/gpu/src/fgd_bgfg.cpp b/modules/gpu/src/fgd_bgfg.cpp
deleted file mode 100644
index 1576f69d1f..0000000000
--- a/modules/gpu/src/fgd_bgfg.cpp
+++ /dev/null
@@ -1,752 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-class cv::gpu::FGDStatModel::Impl
-{
-};
-
-cv::gpu::FGDStatModel::Params::Params() { throw_nogpu(); }
-
-cv::gpu::FGDStatModel::FGDStatModel(int) { throw_nogpu(); }
-cv::gpu::FGDStatModel::FGDStatModel(const cv::gpu::GpuMat&, const Params&, int) { throw_nogpu(); }
-cv::gpu::FGDStatModel::~FGDStatModel() {}
-void cv::gpu::FGDStatModel::create(const cv::gpu::GpuMat&, const Params&) { throw_nogpu(); }
-void cv::gpu::FGDStatModel::release() {}
-int cv::gpu::FGDStatModel::update(const cv::gpu::GpuMat&) { throw_nogpu(); return 0; }
-
-#else
-
-#include "fgd_bgfg_common.hpp"
-
-namespace
-{
-    class BGPixelStat
-    {
-    public:
-        void create(cv::Size size, const cv::gpu::FGDStatModel::Params& params, int out_cn);
-        void release();
-
-        void setTrained();
-
-        operator bgfg::BGPixelStat();
-
-    private:
-        cv::gpu::GpuMat Pbc_;
-        cv::gpu::GpuMat Pbcc_;
-        cv::gpu::GpuMat is_trained_st_model_;
-        cv::gpu::GpuMat is_trained_dyn_model_;
-
-        cv::gpu::GpuMat ctable_Pv_;
-        cv::gpu::GpuMat ctable_Pvb_;
-        cv::gpu::GpuMat ctable_v_;
-
-        cv::gpu::GpuMat cctable_Pv_;
-        cv::gpu::GpuMat cctable_Pvb_;
-        cv::gpu::GpuMat cctable_v1_;
-        cv::gpu::GpuMat cctable_v2_;
-    };
-
-    void BGPixelStat::create(cv::Size size, const cv::gpu::FGDStatModel::Params& params, int out_cn)
-    {
-        cv::gpu::ensureSizeIsEnough(size, CV_32FC1, Pbc_);
-        Pbc_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(size, CV_32FC1, Pbcc_);
-        Pbcc_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(size, CV_8UC1, is_trained_st_model_);
-        is_trained_st_model_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(size, CV_8UC1, is_trained_dyn_model_);
-        is_trained_dyn_model_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_32FC1, ctable_Pv_);
-        ctable_Pv_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_32FC1, ctable_Pvb_);
-        ctable_Pvb_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_8UC(out_cn), ctable_v_);
-        ctable_v_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_32FC1, cctable_Pv_);
-        cctable_Pv_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_32FC1, cctable_Pvb_);
-        cctable_Pvb_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_8UC(out_cn), cctable_v1_);
-        cctable_v1_.setTo(cv::Scalar::all(0));
-
-        cv::gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_8UC(out_cn), cctable_v2_);
-        cctable_v2_.setTo(cv::Scalar::all(0));
-    }
-
-    void BGPixelStat::release()
-    {
-        Pbc_.release();
-        Pbcc_.release();
-        is_trained_st_model_.release();
-        is_trained_dyn_model_.release();
-
-        ctable_Pv_.release();
-        ctable_Pvb_.release();
-        ctable_v_.release();
-
-        cctable_Pv_.release();
-        cctable_Pvb_.release();
-        cctable_v1_.release();
-        cctable_v2_.release();
-    }
-
-    void BGPixelStat::setTrained()
-    {
-        is_trained_st_model_.setTo(cv::Scalar::all(1));
-        is_trained_dyn_model_.setTo(cv::Scalar::all(1));
-    }
-
-    BGPixelStat::operator bgfg::BGPixelStat()
-    {
-        bgfg::BGPixelStat stat;
-
-        stat.rows_ = Pbc_.rows;
-
-        stat.Pbc_data_ = Pbc_.data;
-        stat.Pbc_step_ = Pbc_.step;
-
-        stat.Pbcc_data_ = Pbcc_.data;
-        stat.Pbcc_step_ = Pbcc_.step;
-
-        stat.is_trained_st_model_data_ = is_trained_st_model_.data;
-        stat.is_trained_st_model_step_ = is_trained_st_model_.step;
-
-        stat.is_trained_dyn_model_data_ = is_trained_dyn_model_.data;
-        stat.is_trained_dyn_model_step_ = is_trained_dyn_model_.step;
-
-        stat.ctable_Pv_data_ = ctable_Pv_.data;
-        stat.ctable_Pv_step_ = ctable_Pv_.step;
-
-        stat.ctable_Pvb_data_ = ctable_Pvb_.data;
-        stat.ctable_Pvb_step_ = ctable_Pvb_.step;
-
-        stat.ctable_v_data_ = ctable_v_.data;
-        stat.ctable_v_step_ = ctable_v_.step;
-
-        stat.cctable_Pv_data_ = cctable_Pv_.data;
-        stat.cctable_Pv_step_ = cctable_Pv_.step;
-
-        stat.cctable_Pvb_data_ = cctable_Pvb_.data;
-        stat.cctable_Pvb_step_ = cctable_Pvb_.step;
-
-        stat.cctable_v1_data_ = cctable_v1_.data;
-        stat.cctable_v1_step_ = cctable_v1_.step;
-
-        stat.cctable_v2_data_ = cctable_v2_.data;
-        stat.cctable_v2_step_ = cctable_v2_.step;
-
-        return stat;
-    }
-}
-
-class cv::gpu::FGDStatModel::Impl
-{
-public:
-    Impl(cv::gpu::GpuMat& background, cv::gpu::GpuMat& foreground, std::vector< std::vector<cv::Point> >& foreground_regions, int out_cn);
-    ~Impl();
-
-    void create(const cv::gpu::GpuMat& firstFrame, const cv::gpu::FGDStatModel::Params& params);
-    void release();
-
-    int update(const cv::gpu::GpuMat& curFrame);
-
-private:
-    Impl(const Impl&);
-    Impl& operator=(const Impl&);
-
-    int out_cn_;
-
-    cv::gpu::FGDStatModel::Params params_;
-
-    cv::gpu::GpuMat& background_;
-    cv::gpu::GpuMat& foreground_;
-    std::vector< std::vector<cv::Point> >& foreground_regions_;
-
-    cv::Mat h_foreground_;
-
-    cv::gpu::GpuMat prevFrame_;
-    cv::gpu::GpuMat Ftd_;
-    cv::gpu::GpuMat Fbd_;
-    BGPixelStat stat_;
-
-    cv::gpu::GpuMat hist_;
-    cv::gpu::GpuMat histBuf_;
-
-    cv::gpu::GpuMat countBuf_;
-
-    cv::gpu::GpuMat buf_;
-    cv::gpu::GpuMat filterBuf_;
-    cv::gpu::GpuMat filterBrd_;
-
-    cv::Ptr<cv::gpu::FilterEngine_GPU> dilateFilter_;
-    cv::Ptr<cv::gpu::FilterEngine_GPU> erodeFilter_;
-
-    CvMemStorage* storage_;
-};
-
-cv::gpu::FGDStatModel::Impl::Impl(cv::gpu::GpuMat& background, cv::gpu::GpuMat& foreground, std::vector< std::vector<cv::Point> >& foreground_regions, int out_cn) :
-    out_cn_(out_cn), background_(background), foreground_(foreground), foreground_regions_(foreground_regions)
-{
-    CV_Assert( out_cn_ == 3 || out_cn_ == 4 );
-
-    storage_ = cvCreateMemStorage();
-    CV_Assert( storage_ != 0 );
-}
-
-cv::gpu::FGDStatModel::Impl::~Impl()
-{
-    cvReleaseMemStorage(&storage_);
-}
-
-namespace
-{
-    void copyChannels(const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst, int dst_cn = -1)
-    {
-        const int src_cn = src.channels();
-
-        if (dst_cn < 0)
-            dst_cn = src_cn;
-
-        cv::gpu::ensureSizeIsEnough(src.size(), CV_MAKE_TYPE(src.depth(), dst_cn), dst);
-
-        if (src_cn == dst_cn)
-            src.copyTo(dst);
-        else
-        {
-            static const int cvt_codes[4][4] =
-            {
-                {-1, -1, cv::COLOR_GRAY2BGR, cv::COLOR_GRAY2BGRA},
-                {-1, -1, -1, -1},
-                {cv::COLOR_BGR2GRAY, -1, -1, cv::COLOR_BGR2BGRA},
-                {cv::COLOR_BGRA2GRAY, -1, cv::COLOR_BGRA2BGR, -1}
-            };
-
-            const int cvt_code = cvt_codes[src_cn - 1][dst_cn - 1];
-            CV_DbgAssert( cvt_code >= 0 );
-
-            cv::gpu::cvtColor(src, dst, cvt_code, dst_cn);
-        }
-    }
-}
-
-void cv::gpu::FGDStatModel::Impl::create(const cv::gpu::GpuMat& firstFrame, const cv::gpu::FGDStatModel::Params& params)
-{
-    CV_Assert(firstFrame.type() == CV_8UC3 || firstFrame.type() == CV_8UC4);
-
-    params_ = params;
-
-    cv::gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, foreground_);
-
-    copyChannels(firstFrame, background_, out_cn_);
-
-    copyChannels(firstFrame, prevFrame_);
-
-    cv::gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, Ftd_);
-    cv::gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, Fbd_);
-
-    stat_.create(firstFrame.size(), params_, out_cn_);
-    bgfg::setBGPixelStat(stat_);
-
-    if (params_.perform_morphing > 0)
-    {
-        cv::Mat kernel = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(1 + params_.perform_morphing * 2, 1 + params_.perform_morphing * 2));
-        cv::Point anchor(params_.perform_morphing, params_.perform_morphing);
-
-        dilateFilter_ = cv::gpu::createMorphologyFilter_GPU(cv::MORPH_DILATE, CV_8UC1, kernel, filterBuf_, anchor);
-        erodeFilter_ = cv::gpu::createMorphologyFilter_GPU(cv::MORPH_ERODE, CV_8UC1, kernel, filterBuf_, anchor);
-    }
-}
-
-void cv::gpu::FGDStatModel::Impl::release()
-{
-    background_.release();
-    foreground_.release();
-
-    prevFrame_.release();
-    Ftd_.release();
-    Fbd_.release();
-    stat_.release();
-
-    hist_.release();
-    histBuf_.release();
-
-    countBuf_.release();
-
-    buf_.release();
-    filterBuf_.release();
-    filterBrd_.release();
-}
-
-/////////////////////////////////////////////////////////////////////////
-// changeDetection
-
-namespace
-{
-    void calcDiffHistogram(const cv::gpu::GpuMat& prevFrame, const cv::gpu::GpuMat& curFrame, cv::gpu::GpuMat& hist, cv::gpu::GpuMat& histBuf)
-    {
-        typedef void (*func_t)(cv::gpu::PtrStepSzb prevFrame, cv::gpu::PtrStepSzb curFrame, unsigned int* hist0, unsigned int* hist1, unsigned int* hist2, unsigned int* partialBuf0, unsigned int* partialBuf1, unsigned int* partialBuf2, bool cc20, cudaStream_t stream);
-        static const func_t funcs[4][4] =
-        {
-            {0,0,0,0},
-            {0,0,0,0},
-            {0,0,bgfg::calcDiffHistogram_gpu<uchar3, uchar3>,bgfg::calcDiffHistogram_gpu<uchar3, uchar4>},
-            {0,0,bgfg::calcDiffHistogram_gpu<uchar4, uchar3>,bgfg::calcDiffHistogram_gpu<uchar4, uchar4>}
-        };
-
-        hist.create(3, 256, CV_32SC1);
-        histBuf.create(3, bgfg::PARTIAL_HISTOGRAM_COUNT * bgfg::HISTOGRAM_BIN_COUNT, CV_32SC1);
-
-        funcs[prevFrame.channels() - 1][curFrame.channels() - 1](
-                    prevFrame, curFrame,
-                    hist.ptr<unsigned int>(0), hist.ptr<unsigned int>(1), hist.ptr<unsigned int>(2),
-                    histBuf.ptr<unsigned int>(0), histBuf.ptr<unsigned int>(1), histBuf.ptr<unsigned int>(2),
-                    cv::gpu::deviceSupports(cv::gpu::FEATURE_SET_COMPUTE_20), 0);
-    }
-
-    void calcRelativeVariance(unsigned int hist[3 * 256], double relativeVariance[3][bgfg::HISTOGRAM_BIN_COUNT])
-    {
-        std::memset(relativeVariance, 0, 3 * bgfg::HISTOGRAM_BIN_COUNT * sizeof(double));
-
-        for (int thres = bgfg::HISTOGRAM_BIN_COUNT - 2; thres >= 0; --thres)
-        {
-            cv::Vec3d sum(0.0, 0.0, 0.0);
-            cv::Vec3d sqsum(0.0, 0.0, 0.0);
-            cv::Vec3i count(0, 0, 0);
-
-            for (int j = thres; j < bgfg::HISTOGRAM_BIN_COUNT; ++j)
-            {
-                sum[0]   += static_cast<double>(j) * hist[j];
-                sqsum[0] += static_cast<double>(j * j) * hist[j];
-                count[0] += hist[j];
-
-                sum[1]   += static_cast<double>(j) * hist[j + 256];
-                sqsum[1] += static_cast<double>(j * j) * hist[j + 256];
-                count[1] += hist[j + 256];
-
-                sum[2]   += static_cast<double>(j) * hist[j + 512];
-                sqsum[2] += static_cast<double>(j * j) * hist[j + 512];
-                count[2] += hist[j + 512];
-            }
-
-            count[0] = std::max(count[0], 1);
-            count[1] = std::max(count[1], 1);
-            count[2] = std::max(count[2], 1);
-
-            cv::Vec3d my(
-                sum[0] / count[0],
-                sum[1] / count[1],
-                sum[2] / count[2]
-            );
-
-            relativeVariance[0][thres] = std::sqrt(sqsum[0] / count[0] - my[0] * my[0]);
-            relativeVariance[1][thres] = std::sqrt(sqsum[1] / count[1] - my[1] * my[1]);
-            relativeVariance[2][thres] = std::sqrt(sqsum[2] / count[2] - my[2] * my[2]);
-        }
-    }
-
-    void calcDiffThreshMask(const cv::gpu::GpuMat& prevFrame, const cv::gpu::GpuMat& curFrame, cv::Vec3d bestThres, cv::gpu::GpuMat& changeMask)
-    {
-        typedef void (*func_t)(cv::gpu::PtrStepSzb prevFrame, cv::gpu::PtrStepSzb curFrame, uchar3 bestThres, cv::gpu::PtrStepSzb changeMask, cudaStream_t stream);
-        static const func_t funcs[4][4] =
-        {
-            {0,0,0,0},
-            {0,0,0,0},
-            {0,0,bgfg::calcDiffThreshMask_gpu<uchar3, uchar3>,bgfg::calcDiffThreshMask_gpu<uchar3, uchar4>},
-            {0,0,bgfg::calcDiffThreshMask_gpu<uchar4, uchar3>,bgfg::calcDiffThreshMask_gpu<uchar4, uchar4>}
-        };
-
-        changeMask.setTo(cv::Scalar::all(0));
-
-        funcs[prevFrame.channels() - 1][curFrame.channels() - 1](prevFrame, curFrame, make_uchar3((uchar)bestThres[0], (uchar)bestThres[1], (uchar)bestThres[2]), changeMask, 0);
-    }
-
-    // performs change detection for Foreground detection algorithm
-    void changeDetection(const cv::gpu::GpuMat& prevFrame, const cv::gpu::GpuMat& curFrame, cv::gpu::GpuMat& changeMask, cv::gpu::GpuMat& hist, cv::gpu::GpuMat& histBuf)
-    {
-        calcDiffHistogram(prevFrame, curFrame, hist, histBuf);
-
-        unsigned int histData[3 * 256];
-        cv::Mat h_hist(3, 256, CV_32SC1, histData);
-        hist.download(h_hist);
-
-        double relativeVariance[3][bgfg::HISTOGRAM_BIN_COUNT];
-        calcRelativeVariance(histData, relativeVariance);
-
-        // Find maximum:
-        cv::Vec3d bestThres(10.0, 10.0, 10.0);
-        for (int i = 0; i < bgfg::HISTOGRAM_BIN_COUNT; ++i)
-        {
-            bestThres[0] = std::max(bestThres[0], relativeVariance[0][i]);
-            bestThres[1] = std::max(bestThres[1], relativeVariance[1][i]);
-            bestThres[2] = std::max(bestThres[2], relativeVariance[2][i]);
-        }
-
-        calcDiffThreshMask(prevFrame, curFrame, bestThres, changeMask);
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////
-// bgfgClassification
-
-namespace
-{
-    int bgfgClassification(const cv::gpu::GpuMat& prevFrame, const cv::gpu::GpuMat& curFrame,
-                           const cv::gpu::GpuMat& Ftd, const cv::gpu::GpuMat& Fbd,
-                           cv::gpu::GpuMat& foreground, cv::gpu::GpuMat& countBuf,
-                           const cv::gpu::FGDStatModel::Params& params, int out_cn)
-    {
-        typedef void (*func_t)(cv::gpu::PtrStepSzb prevFrame, cv::gpu::PtrStepSzb curFrame, cv::gpu::PtrStepSzb Ftd, cv::gpu::PtrStepSzb Fbd, cv::gpu::PtrStepSzb foreground,
-                               int deltaC, int deltaCC, float alpha2, int N1c, int N1cc, cudaStream_t stream);
-        static const func_t funcs[4][4][4] =
-        {
-            {
-                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0},
-                {0,0,bgfg::bgfgClassification_gpu<uchar3, uchar3, uchar3>,bgfg::bgfgClassification_gpu<uchar3, uchar3, uchar4>},
-                {0,0,bgfg::bgfgClassification_gpu<uchar3, uchar4, uchar3>,bgfg::bgfgClassification_gpu<uchar3, uchar4, uchar4>}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0},
-                {0,0,bgfg::bgfgClassification_gpu<uchar4, uchar3, uchar3>,bgfg::bgfgClassification_gpu<uchar4, uchar3, uchar4>},
-                {0,0,bgfg::bgfgClassification_gpu<uchar4, uchar4, uchar3>,bgfg::bgfgClassification_gpu<uchar4, uchar4, uchar4>}
-            }
-        };
-
-        const int deltaC  = cvRound(params.delta * 256 / params.Lc);
-        const int deltaCC = cvRound(params.delta * 256 / params.Lcc);
-
-        funcs[prevFrame.channels() - 1][curFrame.channels() - 1][out_cn - 1](prevFrame, curFrame, Ftd, Fbd, foreground, deltaC, deltaCC, params.alpha2, params.N1c, params.N1cc, 0);
-
-        int count = cv::gpu::countNonZero(foreground, countBuf);
-
-        cv::gpu::multiply(foreground, cv::Scalar::all(255), foreground);
-
-        return count;
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////
-// smoothForeground
-
-namespace
-{
-    void morphology(const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst, cv::gpu::GpuMat& filterBrd, int brd, cv::Ptr<cv::gpu::FilterEngine_GPU>& filter, cv::Scalar brdVal)
-    {
-        cv::gpu::copyMakeBorder(src, filterBrd, brd, brd, brd, brd, cv::BORDER_CONSTANT, brdVal);
-        filter->apply(filterBrd(cv::Rect(brd, brd, src.cols, src.rows)), dst, cv::Rect(0, 0, src.cols, src.rows));
-    }
-
-    void smoothForeground(cv::gpu::GpuMat& foreground, cv::gpu::GpuMat& filterBrd, cv::gpu::GpuMat& buf,
-                          cv::Ptr<cv::gpu::FilterEngine_GPU>& erodeFilter, cv::Ptr<cv::gpu::FilterEngine_GPU>& dilateFilter,
-                          const cv::gpu::FGDStatModel::Params& params)
-    {
-        const int brd = params.perform_morphing;
-
-        const cv::Scalar erodeBrdVal = cv::Scalar::all(UCHAR_MAX);
-        const cv::Scalar dilateBrdVal = cv::Scalar::all(0);
-
-        // MORPH_OPEN
-        morphology(foreground, buf, filterBrd, brd, erodeFilter, erodeBrdVal);
-        morphology(buf, foreground, filterBrd, brd, dilateFilter, dilateBrdVal);
-
-        // MORPH_CLOSE
-        morphology(foreground, buf, filterBrd, brd, dilateFilter, dilateBrdVal);
-        morphology(buf, foreground, filterBrd, brd, erodeFilter, erodeBrdVal);
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////
-// findForegroundRegions
-
-namespace
-{
-    void seqToContours(CvSeq* _ccontours, CvMemStorage* storage, cv::OutputArrayOfArrays _contours)
-    {
-        cv::Seq<CvSeq*> all_contours(cvTreeToNodeSeq(_ccontours, sizeof(CvSeq), storage));
-
-        size_t total = all_contours.size();
-
-        _contours.create((int) total, 1, 0, -1, true);
-
-        cv::SeqIterator<CvSeq*> it = all_contours.begin();
-        for (size_t i = 0; i < total; ++i, ++it)
-        {
-            CvSeq* c = *it;
-            ((CvContour*)c)->color = (int)i;
-            _contours.create((int)c->total, 1, CV_32SC2, (int)i, true);
-            cv::Mat ci = _contours.getMat((int)i);
-            CV_Assert( ci.isContinuous() );
-            cvCvtSeqToArray(c, ci.data);
-        }
-    }
-
-    int findForegroundRegions(cv::gpu::GpuMat& d_foreground, cv::Mat& h_foreground, std::vector< std::vector<cv::Point> >& foreground_regions,
-                              CvMemStorage* storage, const cv::gpu::FGDStatModel::Params& params)
-    {
-        int region_count = 0;
-
-        // Discard under-size foreground regions:
-
-        d_foreground.download(h_foreground);
-        IplImage ipl_foreground = h_foreground;
-        CvSeq* first_seq = 0;
-
-        cvFindContours(&ipl_foreground, storage, &first_seq, sizeof(CvContour), CV_RETR_LIST);
-
-        for (CvSeq* seq = first_seq; seq; seq = seq->h_next)
-        {
-            CvContour* cnt = reinterpret_cast<CvContour*>(seq);
-
-            if (cnt->rect.width * cnt->rect.height < params.minArea || (params.is_obj_without_holes && CV_IS_SEQ_HOLE(seq)))
-            {
-                // Delete under-size contour:
-                CvSeq* prev_seq = seq->h_prev;
-                if (prev_seq)
-                {
-                    prev_seq->h_next = seq->h_next;
-
-                    if (seq->h_next)
-                        seq->h_next->h_prev = prev_seq;
-                }
-                else
-                {
-                    first_seq = seq->h_next;
-
-                    if (seq->h_next)
-                        seq->h_next->h_prev = NULL;
-                }
-            }
-            else
-            {
-                region_count++;
-            }
-        }
-
-        seqToContours(first_seq, storage, foreground_regions);
-        h_foreground.setTo(0);
-
-        cv::drawContours(h_foreground, foreground_regions, -1, cv::Scalar::all(255), -1);
-
-        d_foreground.upload(h_foreground);
-
-        return region_count;
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////
-// updateBackgroundModel
-
-namespace
-{
-    void updateBackgroundModel(const cv::gpu::GpuMat& prevFrame, const cv::gpu::GpuMat& curFrame, const cv::gpu::GpuMat& Ftd, const cv::gpu::GpuMat& Fbd,
-                               const cv::gpu::GpuMat& foreground, cv::gpu::GpuMat& background,
-                               const cv::gpu::FGDStatModel::Params& params)
-    {
-        typedef void (*func_t)(cv::gpu::PtrStepSzb prevFrame, cv::gpu::PtrStepSzb curFrame, cv::gpu::PtrStepSzb Ftd, cv::gpu::PtrStepSzb Fbd,
-                               cv::gpu::PtrStepSzb foreground, cv::gpu::PtrStepSzb background,
-                               int deltaC, int deltaCC, float alpha1, float alpha2, float alpha3, int N1c, int N1cc, int N2c, int N2cc, float T, cudaStream_t stream);
-        static const func_t funcs[4][4][4] =
-        {
-            {
-                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0},
-                {0,0,bgfg::updateBackgroundModel_gpu<uchar3, uchar3, uchar3>,bgfg::updateBackgroundModel_gpu<uchar3, uchar3, uchar4>},
-                {0,0,bgfg::updateBackgroundModel_gpu<uchar3, uchar4, uchar3>,bgfg::updateBackgroundModel_gpu<uchar3, uchar4, uchar4>}
-            },
-            {
-                {0,0,0,0}, {0,0,0,0},
-                {0,0,bgfg::updateBackgroundModel_gpu<uchar4, uchar3, uchar3>,bgfg::updateBackgroundModel_gpu<uchar4, uchar3, uchar4>},
-                {0,0,bgfg::updateBackgroundModel_gpu<uchar4, uchar4, uchar3>,bgfg::updateBackgroundModel_gpu<uchar4, uchar4, uchar4>}
-            }
-        };
-
-        const int deltaC  = cvRound(params.delta * 256 / params.Lc);
-        const int deltaCC = cvRound(params.delta * 256 / params.Lcc);
-
-        funcs[prevFrame.channels() - 1][curFrame.channels() - 1][background.channels() - 1](
-                    prevFrame, curFrame, Ftd, Fbd, foreground, background,
-                    deltaC, deltaCC, params.alpha1, params.alpha2, params.alpha3, params.N1c, params.N1cc, params.N2c, params.N2cc, params.T,
-                    0);
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////
-// Impl::update
-
-int cv::gpu::FGDStatModel::Impl::update(const cv::gpu::GpuMat& curFrame)
-{
-    CV_Assert(curFrame.type() == CV_8UC3 || curFrame.type() == CV_8UC4);
-    CV_Assert(curFrame.size() == prevFrame_.size());
-
-    cvClearMemStorage(storage_);
-    foreground_regions_.clear();
-    foreground_.setTo(cv::Scalar::all(0));
-
-    changeDetection(prevFrame_, curFrame, Ftd_, hist_, histBuf_);
-    changeDetection(background_, curFrame, Fbd_, hist_, histBuf_);
-
-    int FG_pixels_count = bgfgClassification(prevFrame_, curFrame, Ftd_, Fbd_, foreground_, countBuf_, params_, out_cn_);
-
-    if (params_.perform_morphing > 0)
-        smoothForeground(foreground_, filterBrd_, buf_, erodeFilter_, dilateFilter_, params_);
-
-    int region_count = 0;
-    if (params_.minArea > 0 || params_.is_obj_without_holes)
-        region_count = findForegroundRegions(foreground_, h_foreground_, foreground_regions_, storage_, params_);
-
-    // Check ALL BG update condition:
-    const double BGFG_FGD_BG_UPDATE_TRESH = 0.5;
-    if (static_cast<double>(FG_pixels_count) / Ftd_.size().area() > BGFG_FGD_BG_UPDATE_TRESH)
-        stat_.setTrained();
-
-    updateBackgroundModel(prevFrame_, curFrame, Ftd_, Fbd_, foreground_, background_, params_);
-
-    copyChannels(curFrame, prevFrame_);
-
-    return region_count;
-}
-
-namespace
-{
-    // Default parameters of foreground detection algorithm:
-    const int BGFG_FGD_LC  = 128;
-    const int BGFG_FGD_N1C = 15;
-    const int BGFG_FGD_N2C = 25;
-
-    const int BGFG_FGD_LCC   = 64;
-    const int BGFG_FGD_N1CC = 25;
-    const int BGFG_FGD_N2CC = 40;
-
-    // Background reference image update parameter:
-    const float BGFG_FGD_ALPHA_1 = 0.1f;
-
-    // stat model update parameter
-    // 0.002f ~ 1K frame(~45sec), 0.005 ~ 18sec (if 25fps and absolutely static BG)
-    const float BGFG_FGD_ALPHA_2 = 0.005f;
-
-    // start value for alpha parameter (to fast initiate statistic model)
-    const float BGFG_FGD_ALPHA_3 = 0.1f;
-
-    const float BGFG_FGD_DELTA = 2.0f;
-
-    const float BGFG_FGD_T = 0.9f;
-
-    const float BGFG_FGD_MINAREA= 15.0f;
-}
-
-cv::gpu::FGDStatModel::Params::Params()
-{
-    Lc      = BGFG_FGD_LC;
-    N1c     = BGFG_FGD_N1C;
-    N2c     = BGFG_FGD_N2C;
-
-    Lcc     = BGFG_FGD_LCC;
-    N1cc    = BGFG_FGD_N1CC;
-    N2cc    = BGFG_FGD_N2CC;
-
-    delta   = BGFG_FGD_DELTA;
-
-    alpha1  = BGFG_FGD_ALPHA_1;
-    alpha2  = BGFG_FGD_ALPHA_2;
-    alpha3  = BGFG_FGD_ALPHA_3;
-
-    T       = BGFG_FGD_T;
-    minArea = BGFG_FGD_MINAREA;
-
-    is_obj_without_holes = true;
-    perform_morphing     = 1;
-}
-
-cv::gpu::FGDStatModel::FGDStatModel(int out_cn)
-{
-    impl_.reset(new Impl(background, foreground, foreground_regions, out_cn));
-}
-
-cv::gpu::FGDStatModel::FGDStatModel(const cv::gpu::GpuMat& firstFrame, const Params& params, int out_cn)
-{
-    impl_.reset(new Impl(background, foreground, foreground_regions, out_cn));
-    create(firstFrame, params);
-}
-
-cv::gpu::FGDStatModel::~FGDStatModel()
-{
-}
-
-void cv::gpu::FGDStatModel::create(const cv::gpu::GpuMat& firstFrame, const Params& params)
-{
-    impl_->create(firstFrame, params);
-}
-
-void cv::gpu::FGDStatModel::release()
-{
-    impl_->release();
-}
-
-int cv::gpu::FGDStatModel::update(const cv::gpu::GpuMat& curFrame)
-{
-    return impl_->update(curFrame);
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/filtering.cpp b/modules/gpu/src/filtering.cpp
deleted file mode 100644
index 8905eaed64..0000000000
--- a/modules/gpu/src/filtering.cpp
+++ /dev/null
@@ -1,1339 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-Ptr<FilterEngine_GPU> cv::gpu::createFilter2D_GPU(const Ptr<BaseFilter_GPU>&, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>&, const Ptr<BaseColumnFilter_GPU>&, int, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>&, const Ptr<BaseColumnFilter_GPU>&, int, int, int, GpuMat&) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<BaseRowFilter_GPU> cv::gpu::getRowSumFilter_GPU(int, int, int, int) { throw_nogpu(); return Ptr<BaseRowFilter_GPU>(0); }
-Ptr<BaseColumnFilter_GPU> cv::gpu::getColumnSumFilter_GPU(int, int, int, int) { throw_nogpu(); return Ptr<BaseColumnFilter_GPU>(0); }
-Ptr<BaseFilter_GPU> cv::gpu::getBoxFilter_GPU(int, int, const Size&, Point) { throw_nogpu(); return Ptr<BaseFilter_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createBoxFilter_GPU(int, int, const Size&, const Point&) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<BaseFilter_GPU> cv::gpu::getMorphologyFilter_GPU(int, int, const Mat&, const Size&, Point) { throw_nogpu(); return Ptr<BaseFilter_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createMorphologyFilter_GPU(int, int, const Mat&, const Point&, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createMorphologyFilter_GPU(int, int, const Mat&, GpuMat&, const Point&, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<BaseFilter_GPU> cv::gpu::getLinearFilter_GPU(int, int, const Mat&, Point, int) { throw_nogpu(); return Ptr<BaseFilter_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createLinearFilter_GPU(int, int, const Mat&, Point, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<BaseRowFilter_GPU> cv::gpu::getLinearRowFilter_GPU(int, int, const Mat&, int, int) { throw_nogpu(); return Ptr<BaseRowFilter_GPU>(0); }
-Ptr<BaseColumnFilter_GPU> cv::gpu::getLinearColumnFilter_GPU(int, int, const Mat&, int, int) { throw_nogpu(); return Ptr<BaseColumnFilter_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableLinearFilter_GPU(int, int, const Mat&, const Mat&, const Point&, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableLinearFilter_GPU(int, int, const Mat&, const Mat&, GpuMat&, const Point&, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createDerivFilter_GPU(int, int, int, int, int, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createDerivFilter_GPU(int, int, int, int, int, GpuMat&, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createGaussianFilter_GPU(int, Size, double, double, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<FilterEngine_GPU> cv::gpu::createGaussianFilter_GPU(int, Size, GpuMat&, double, double, int, int) { throw_nogpu(); return Ptr<FilterEngine_GPU>(0); }
-Ptr<BaseFilter_GPU> cv::gpu::getMaxFilter_GPU(int, int, const Size&, Point) { throw_nogpu(); return Ptr<BaseFilter_GPU>(0); }
-Ptr<BaseFilter_GPU> cv::gpu::getMinFilter_GPU(int, int, const Size&, Point) { throw_nogpu(); return Ptr<BaseFilter_GPU>(0); }
-
-void cv::gpu::boxFilter(const GpuMat&, GpuMat&, int, Size, Point, Stream&) { throw_nogpu(); }
-void cv::gpu::erode(const GpuMat&, GpuMat&, const Mat&, Point, int) { throw_nogpu(); }
-void cv::gpu::erode(const GpuMat&, GpuMat&, const Mat&, GpuMat&, Point, int, Stream&) { throw_nogpu(); }
-void cv::gpu::dilate(const GpuMat&, GpuMat&, const Mat&, Point, int) { throw_nogpu(); }
-void cv::gpu::dilate(const GpuMat&, GpuMat&, const Mat&, GpuMat&, Point, int, Stream&) { throw_nogpu(); }
-void cv::gpu::morphologyEx(const GpuMat&, GpuMat&, int, const Mat&, Point, int) { throw_nogpu(); }
-void cv::gpu::morphologyEx(const GpuMat&, GpuMat&, int, const Mat&, GpuMat&, GpuMat&, Point, int, Stream&) { throw_nogpu(); }
-void cv::gpu::filter2D(const GpuMat&, GpuMat&, int, const Mat&, Point, int, Stream&) { throw_nogpu(); }
-void cv::gpu::sepFilter2D(const GpuMat&, GpuMat&, int, const Mat&, const Mat&, Point, int, int) { throw_nogpu(); }
-void cv::gpu::sepFilter2D(const GpuMat&, GpuMat&, int, const Mat&, const Mat&, GpuMat&, Point, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::Sobel(const GpuMat&, GpuMat&, int, int, int, int, double, int, int) { throw_nogpu(); }
-void cv::gpu::Sobel(const GpuMat&, GpuMat&, int, int, int, GpuMat&, int, double, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::Scharr(const GpuMat&, GpuMat&, int, int, int, double, int, int) { throw_nogpu(); }
-void cv::gpu::Scharr(const GpuMat&, GpuMat&, int, int, int, GpuMat&, double, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::GaussianBlur(const GpuMat&, GpuMat&, Size, double, double, int, int) { throw_nogpu(); }
-void cv::gpu::GaussianBlur(const GpuMat&, GpuMat&, Size, GpuMat&, double, double, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::Laplacian(const GpuMat&, GpuMat&, int, int, double, int, Stream&) { throw_nogpu(); }
-
-#else
-
-namespace
-{
-    inline void normalizeAnchor(int& anchor, int ksize)
-    {
-        if (anchor < 0)
-            anchor = ksize >> 1;
-
-        CV_Assert(0 <= anchor && anchor < ksize);
-    }
-
-    inline void normalizeAnchor(Point& anchor, const Size& ksize)
-    {
-        normalizeAnchor(anchor.x, ksize.width);
-        normalizeAnchor(anchor.y, ksize.height);
-    }
-
-    inline void normalizeROI(Rect& roi, const Size& ksize, const Point& anchor, const Size& src_size)
-    {
-        if (roi == Rect(0,0,-1,-1))
-            roi = Rect(anchor.x, anchor.y, src_size.width - ksize.width, src_size.height - ksize.height);
-
-        CV_Assert(roi.x >= 0 && roi.y >= 0 && roi.width <= src_size.width && roi.height <= src_size.height);
-    }
-
-    inline void normalizeKernel(const Mat& kernel, GpuMat& gpu_krnl, int type = CV_8U, int* nDivisor = 0, bool reverse = false)
-    {
-        int scale = nDivisor && (kernel.depth() == CV_32F || kernel.depth() == CV_64F) ? 256 : 1;
-        if (nDivisor) *nDivisor = scale;
-
-        Mat temp(kernel.size(), type);
-        kernel.convertTo(temp, type, scale);
-        Mat cont_krnl = temp.reshape(1, 1);
-
-        if (reverse)
-        {
-            int count = cont_krnl.cols >> 1;
-            for (int i = 0; i < count; ++i)
-            {
-                std::swap(cont_krnl.at<int>(0, i), cont_krnl.at<int>(0, cont_krnl.cols - 1 - i));
-            }
-        }
-
-        gpu_krnl.upload(cont_krnl);
-    }
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Filter2D
-
-namespace
-{
-    struct Filter2DEngine_GPU : public FilterEngine_GPU
-    {
-        Filter2DEngine_GPU(const Ptr<BaseFilter_GPU>& filter2D_, int srcType_, int dstType_) :
-            filter2D(filter2D_), srcType(srcType_), dstType(dstType_)
-        {}
-
-        virtual void apply(const GpuMat& src, GpuMat& dst, Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null())
-        {
-            CV_Assert(src.type() == srcType);
-
-            Size src_size = src.size();
-
-            dst.create(src_size, dstType);
-
-            if (roi.size() != src_size)
-            {
-                if (stream)
-                    stream.enqueueMemSet(dst, Scalar::all(0));
-                else
-                    dst.setTo(Scalar::all(0));
-            }
-
-            normalizeROI(roi, filter2D->ksize, filter2D->anchor, src_size);
-
-            GpuMat srcROI = src(roi);
-            GpuMat dstROI = dst(roi);
-
-            (*filter2D)(srcROI, dstROI, stream);
-        }
-
-        Ptr<BaseFilter_GPU> filter2D;
-        int srcType, dstType;
-    };
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createFilter2D_GPU(const Ptr<BaseFilter_GPU>& filter2D, int srcType, int dstType)
-{
-    return Ptr<FilterEngine_GPU>(new Filter2DEngine_GPU(filter2D, srcType, dstType));
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// SeparableFilter
-
-namespace
-{
-    struct SeparableFilterEngine_GPU : public FilterEngine_GPU
-    {
-        SeparableFilterEngine_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter_, const Ptr<BaseColumnFilter_GPU>& columnFilter_,
-                                  int srcType_, int bufType_, int dstType_) :
-            rowFilter(rowFilter_), columnFilter(columnFilter_),
-            srcType(srcType_), bufType(bufType_), dstType(dstType_)
-        {
-            ksize = Size(rowFilter->ksize, columnFilter->ksize);
-            anchor = Point(rowFilter->anchor, columnFilter->anchor);
-
-            pbuf = &buf;
-        }
-
-        SeparableFilterEngine_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter_, const Ptr<BaseColumnFilter_GPU>& columnFilter_,
-                                  int srcType_, int bufType_, int dstType_,
-                                  GpuMat& buf_) :
-            rowFilter(rowFilter_), columnFilter(columnFilter_),
-            srcType(srcType_), bufType(bufType_), dstType(dstType_)
-        {
-            ksize = Size(rowFilter->ksize, columnFilter->ksize);
-            anchor = Point(rowFilter->anchor, columnFilter->anchor);
-
-            pbuf = &buf_;
-        }
-
-        virtual void apply(const GpuMat& src, GpuMat& dst, Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null())
-        {
-            CV_Assert(src.type() == srcType);
-
-            Size src_size = src.size();
-
-            dst.create(src_size, dstType);
-
-            if (roi.size() != src_size)
-            {
-                if (stream)
-                    stream.enqueueMemSet(dst, Scalar::all(0));
-                else
-                    dst.setTo(Scalar::all(0));
-            }
-
-            ensureSizeIsEnough(src_size, bufType, *pbuf);
-
-            normalizeROI(roi, ksize, anchor, src_size);
-
-            GpuMat srcROI = src(roi);
-            GpuMat dstROI = dst(roi);
-            GpuMat bufROI = (*pbuf)(roi);
-
-            (*rowFilter)(srcROI, bufROI, stream);
-            (*columnFilter)(bufROI, dstROI, stream);
-        }
-
-        Ptr<BaseRowFilter_GPU> rowFilter;
-        Ptr<BaseColumnFilter_GPU> columnFilter;
-
-        int srcType, bufType, dstType;
-
-        Size ksize;
-        Point anchor;
-
-        GpuMat buf;
-        GpuMat* pbuf;
-    };
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter,
-    const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType)
-{
-    return Ptr<FilterEngine_GPU>(new SeparableFilterEngine_GPU(rowFilter, columnFilter, srcType, bufType, dstType));
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter,
-    const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType, GpuMat& buf)
-{
-    return Ptr<FilterEngine_GPU>(new SeparableFilterEngine_GPU(rowFilter, columnFilter, srcType, bufType, dstType, buf));
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// 1D Sum Filter
-
-namespace
-{
-    struct NppRowSumFilter : public BaseRowFilter_GPU
-    {
-        NppRowSumFilter(int ksize_, int anchor_) : BaseRowFilter_GPU(ksize_, anchor_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( nppiSumWindowRow_8u32f_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-                dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, ksize, anchor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-Ptr<BaseRowFilter_GPU> cv::gpu::getRowSumFilter_GPU(int srcType, int sumType, int ksize, int anchor)
-{
-    CV_Assert(srcType == CV_8UC1 && sumType == CV_32FC1);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseRowFilter_GPU>(new NppRowSumFilter(ksize, anchor));
-}
-
-namespace
-{
-    struct NppColumnSumFilter : public BaseColumnFilter_GPU
-    {
-        NppColumnSumFilter(int ksize_, int anchor_) : BaseColumnFilter_GPU(ksize_, anchor_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( nppiSumWindowColumn_8u32f_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
-                dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, ksize, anchor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-Ptr<BaseColumnFilter_GPU> cv::gpu::getColumnSumFilter_GPU(int sumType, int dstType, int ksize, int anchor)
-{
-    CV_Assert(sumType == CV_8UC1 && dstType == CV_32FC1);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseColumnFilter_GPU>(new NppColumnSumFilter(ksize, anchor));
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Box Filter
-
-namespace
-{
-    typedef NppStatus (*nppFilterBox_t)(const Npp8u * pSrc, Npp32s nSrcStep, Npp8u * pDst, Npp32s nDstStep, NppiSize oSizeROI,
-        NppiSize oMaskSize, NppiPoint oAnchor);
-
-    struct NPPBoxFilter : public BaseFilter_GPU
-    {
-        NPPBoxFilter(const Size& ksize_, const Point& anchor_, nppFilterBox_t func_) : BaseFilter_GPU(ksize_, anchor_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-            NppiSize oKernelSize;
-            oKernelSize.height = ksize.height;
-            oKernelSize.width = ksize.width;
-            NppiPoint oAnchor;
-            oAnchor.x = anchor.x;
-            oAnchor.y = anchor.y;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step),
-                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, oKernelSize, oAnchor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        nppFilterBox_t func;
-    };
-}
-
-Ptr<BaseFilter_GPU> cv::gpu::getBoxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor)
-{
-    static const nppFilterBox_t nppFilterBox_callers[] = {0, nppiFilterBox_8u_C1R, 0, 0, nppiFilterBox_8u_C4R};
-
-    CV_Assert((srcType == CV_8UC1 || srcType == CV_8UC4) && dstType == srcType);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseFilter_GPU>(new NPPBoxFilter(ksize, anchor, nppFilterBox_callers[CV_MAT_CN(srcType)]));
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createBoxFilter_GPU(int srcType, int dstType, const Size& ksize, const Point& anchor)
-{
-    Ptr<BaseFilter_GPU> boxFilter = getBoxFilter_GPU(srcType, dstType, ksize, anchor);
-    return createFilter2D_GPU(boxFilter, srcType, dstType);
-}
-
-void cv::gpu::boxFilter(const GpuMat& src, GpuMat& dst, int ddepth, Size ksize, Point anchor, Stream& stream)
-{
-    int sdepth = src.depth(), cn = src.channels();
-    if( ddepth < 0 )
-        ddepth = sdepth;
-
-    dst.create(src.size(), CV_MAKETYPE(ddepth, cn));
-
-    Ptr<FilterEngine_GPU> f = createBoxFilter_GPU(src.type(), dst.type(), ksize, anchor);
-    f->apply(src, dst, Rect(0,0,-1,-1), stream);
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Morphology Filter
-
-namespace
-{
-    typedef NppStatus (*nppMorfFilter_t)(const Npp8u*, Npp32s, Npp8u*, Npp32s, NppiSize, const Npp8u*, NppiSize, NppiPoint);
-
-    struct NPPMorphFilter : public BaseFilter_GPU
-    {
-        NPPMorphFilter(const Size& ksize_, const Point& anchor_, const GpuMat& kernel_, nppMorfFilter_t func_) :
-            BaseFilter_GPU(ksize_, anchor_), kernel(kernel_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-            NppiSize oKernelSize;
-            oKernelSize.height = ksize.height;
-            oKernelSize.width = ksize.width;
-            NppiPoint oAnchor;
-            oAnchor.x = anchor.x;
-            oAnchor.y = anchor.y;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step),
-                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, kernel.ptr<Npp8u>(), oKernelSize, oAnchor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        GpuMat kernel;
-        nppMorfFilter_t func;
-    };
-}
-
-Ptr<BaseFilter_GPU> cv::gpu::getMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Size& ksize, Point anchor)
-{
-    static const nppMorfFilter_t nppMorfFilter_callers[2][5] =
-    {
-        {0, nppiErode_8u_C1R, 0, 0, nppiErode_8u_C4R },
-        {0, nppiDilate_8u_C1R, 0, 0, nppiDilate_8u_C4R }
-    };
-
-    CV_Assert(op == MORPH_ERODE || op == MORPH_DILATE);
-    CV_Assert(type == CV_8UC1 || type == CV_8UC4);
-
-    GpuMat gpu_krnl;
-    normalizeKernel(kernel, gpu_krnl);
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseFilter_GPU>(new NPPMorphFilter(ksize, anchor, gpu_krnl, nppMorfFilter_callers[op][CV_MAT_CN(type)]));
-}
-
-namespace
-{
-    struct MorphologyFilterEngine_GPU : public FilterEngine_GPU
-    {
-        MorphologyFilterEngine_GPU(const Ptr<BaseFilter_GPU>& filter2D_, int type_, int iters_) :
-            filter2D(filter2D_), type(type_), iters(iters_)
-        {
-            pbuf = &buf;
-        }
-
-        MorphologyFilterEngine_GPU(const Ptr<BaseFilter_GPU>& filter2D_, int type_, int iters_, GpuMat& buf_) :
-            filter2D(filter2D_), type(type_), iters(iters_)
-        {
-            pbuf = &buf_;
-        }
-
-        virtual void apply(const GpuMat& src, GpuMat& dst, Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null())
-        {
-            CV_Assert(src.type() == type);
-
-            Size src_size = src.size();
-
-            dst.create(src_size, type);
-
-            if (roi.size() != src_size)
-            {
-                if (stream)
-                    stream.enqueueMemSet(dst, Scalar::all(0));
-                else
-                    dst.setTo(Scalar::all(0));
-            }
-
-            normalizeROI(roi, filter2D->ksize, filter2D->anchor, src_size);
-
-            if (iters > 1)
-                pbuf->create(src_size, type);
-
-            GpuMat srcROI = src(roi);
-            GpuMat dstROI = dst(roi);
-
-            (*filter2D)(srcROI, dstROI, stream);
-
-            for(int i = 1; i < iters; ++i)
-            {
-                dst.swap((*pbuf));
-
-                dstROI = dst(roi);
-                GpuMat bufROI = (*pbuf)(roi);
-
-                (*filter2D)(bufROI, dstROI, stream);
-            }
-        }
-
-        Ptr<BaseFilter_GPU> filter2D;
-
-        int type;
-        int iters;
-
-        GpuMat buf;
-        GpuMat* pbuf;
-    };
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Point& anchor, int iterations)
-{
-    CV_Assert(iterations > 0);
-
-    Size ksize = kernel.size();
-
-    Ptr<BaseFilter_GPU> filter2D = getMorphologyFilter_GPU(op, type, kernel, ksize, anchor);
-
-    return Ptr<FilterEngine_GPU>(new MorphologyFilterEngine_GPU(filter2D, type, iterations));
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createMorphologyFilter_GPU(int op, int type, const Mat& kernel, GpuMat& buf, const Point& anchor, int iterations)
-{
-    CV_Assert(iterations > 0);
-
-    Size ksize = kernel.size();
-
-    Ptr<BaseFilter_GPU> filter2D = getMorphologyFilter_GPU(op, type, kernel, ksize, anchor);
-
-    return Ptr<FilterEngine_GPU>(new MorphologyFilterEngine_GPU(filter2D, type, iterations, buf));
-}
-
-namespace
-{
-    void morphOp(int op, const GpuMat& src, GpuMat& dst, const Mat& _kernel, GpuMat& buf, Point anchor, int iterations, Stream& stream = Stream::Null())
-    {
-        Mat kernel;
-        Size ksize = _kernel.data ? _kernel.size() : Size(3, 3);
-
-        normalizeAnchor(anchor, ksize);
-
-        if (iterations == 0 || _kernel.rows * _kernel.cols == 1)
-        {
-            if (stream)
-                stream.enqueueCopy(src, dst);
-            else
-                src.copyTo(dst);
-            return;
-        }
-
-        dst.create(src.size(), src.type());
-
-        if (!_kernel.data)
-        {
-            kernel = getStructuringElement(MORPH_RECT, Size(1 + iterations * 2, 1 + iterations * 2));
-            anchor = Point(iterations, iterations);
-            iterations = 1;
-        }
-        else if (iterations > 1 && countNonZero(_kernel) == _kernel.rows * _kernel.cols)
-        {
-            anchor = Point(anchor.x * iterations, anchor.y * iterations);
-            kernel = getStructuringElement(MORPH_RECT,
-                                           Size(ksize.width + (iterations - 1) * (ksize.width - 1),
-                                                ksize.height + (iterations - 1) * (ksize.height - 1)),
-                                           anchor);
-            iterations = 1;
-        }
-        else
-            kernel = _kernel;
-
-        Ptr<FilterEngine_GPU> f = createMorphologyFilter_GPU(op, src.type(), kernel, buf, anchor, iterations);
-
-        f->apply(src, dst, Rect(0,0,-1,-1), stream);
-    }
-
-    void morphOp(int op, const GpuMat& src, GpuMat& dst, const Mat& _kernel, Point anchor, int iterations)
-    {
-        GpuMat buf;
-        morphOp(op, src, dst, _kernel, buf, anchor, iterations);
-    }
-}
-
-void cv::gpu::erode( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor, int iterations)
-{
-    morphOp(MORPH_ERODE, src, dst, kernel, anchor, iterations);
-}
-
-void cv::gpu::erode( const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf, Point anchor, int iterations, Stream& stream)
-{
-    morphOp(MORPH_ERODE, src, dst, kernel, buf, anchor, iterations, stream);
-}
-
-void cv::gpu::dilate( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor, int iterations)
-{
-    morphOp(MORPH_DILATE, src, dst, kernel, anchor, iterations);
-}
-
-void cv::gpu::dilate( const GpuMat& src, GpuMat& dst, const Mat& kernel, GpuMat& buf, Point anchor, int iterations, Stream& stream)
-{
-    morphOp(MORPH_DILATE, src, dst, kernel, buf, anchor, iterations, stream);
-}
-
-void cv::gpu::morphologyEx(const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, Point anchor, int iterations)
-{
-    GpuMat buf1;
-    GpuMat buf2;
-    morphologyEx(src, dst, op, kernel, buf1, buf2, anchor, iterations);
-}
-
-void cv::gpu::morphologyEx(const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, GpuMat& buf1, GpuMat& buf2, Point anchor, int iterations, Stream& stream)
-{
-    switch( op )
-    {
-    case MORPH_ERODE:   erode(src, dst, kernel, buf1, anchor, iterations, stream); break;
-    case MORPH_DILATE: dilate(src, dst, kernel, buf1, anchor, iterations, stream); break;
-    case MORPH_OPEN:
-        erode(src, buf2, kernel, buf1, anchor, iterations, stream);
-        dilate(buf2, dst, kernel, buf1, anchor, iterations, stream);
-        break;
-    case CV_MOP_CLOSE:
-        dilate(src, buf2, kernel, buf1, anchor, iterations, stream);
-        erode(buf2, dst, kernel, buf1, anchor, iterations, stream);
-        break;
-    case CV_MOP_GRADIENT:
-        erode(src, buf2, kernel, buf1, anchor, iterations, stream);
-        dilate(src, dst, kernel, buf1, anchor, iterations, stream);
-        subtract(dst, buf2, dst, GpuMat(), -1, stream);
-        break;
-    case CV_MOP_TOPHAT:
-        erode(src, dst, kernel, buf1, anchor, iterations, stream);
-        dilate(dst, buf2, kernel, buf1, anchor, iterations, stream);
-        subtract(src, buf2, dst, GpuMat(), -1, stream);
-        break;
-    case CV_MOP_BLACKHAT:
-        dilate(src, dst, kernel, buf1, anchor, iterations, stream);
-        erode(dst, buf2, kernel, buf1, anchor, iterations, stream);
-        subtract(buf2, src, dst, GpuMat(), -1, stream);
-        break;
-    default:
-        CV_Error(CV_StsBadArg, "unknown morphological operation");
-    }
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Linear Filter
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T, typename D>
-        void filter2D_gpu(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst,
-                          int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel,
-                          int borderMode, const float* borderValue, cudaStream_t stream);
-    }
-}}}
-
-namespace
-{
-    typedef NppStatus (*nppFilter2D_t)(const Npp8u * pSrc, Npp32s nSrcStep, Npp8u * pDst, Npp32s nDstStep, NppiSize oSizeROI,
-        const Npp32s * pKernel, NppiSize oKernelSize, NppiPoint oAnchor, Npp32s nDivisor);
-
-    struct NPPLinearFilter : public BaseFilter_GPU
-    {
-        NPPLinearFilter(const Size& ksize_, const Point& anchor_, const GpuMat& kernel_, Npp32s nDivisor_, nppFilter2D_t func_) :
-            BaseFilter_GPU(ksize_, anchor_), kernel(kernel_), nDivisor(nDivisor_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-            NppiSize oKernelSize;
-            oKernelSize.height = ksize.height;
-            oKernelSize.width = ksize.width;
-            NppiPoint oAnchor;
-            oAnchor.x = anchor.x;
-            oAnchor.y = anchor.y;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz,
-                kernel.ptr<Npp32s>(), oKernelSize, oAnchor, nDivisor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        GpuMat kernel;
-        Npp32s nDivisor;
-        nppFilter2D_t func;
-    };
-
-    typedef void (*gpuFilter2D_t)(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst,
-                                   int kWidth, int kHeight, int anchorX, int anchorY, const float* kernel,
-                                   int borderMode, const float* borderValue, cudaStream_t stream);
-
-    struct GpuFilter2D : public BaseFilter_GPU
-    {
-        GpuFilter2D(Size ksize_, Point anchor_, gpuFilter2D_t func_, const GpuMat& kernel_, int brd_type_) :
-            BaseFilter_GPU(ksize_, anchor_), func(func_), kernel(kernel_), brd_type(brd_type_)
-        {
-        }
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
-        {
-            using namespace cv::gpu::device::imgproc;
-
-            Point ofs;
-            Size wholeSize;
-            src.locateROI(wholeSize, ofs);
-            GpuMat srcWhole(wholeSize, src.type(), src.datastart);
-
-            static const Scalar_<float> zero = Scalar_<float>::all(0.0f);
-            func(srcWhole, ofs.x, ofs.y, dst, ksize.width, ksize.height, anchor.x, anchor.y, kernel.ptr<float>(), brd_type, zero.val, StreamAccessor::getStream(stream));
-        }
-
-        gpuFilter2D_t func;
-        GpuMat kernel;
-        int brd_type;
-    };
-}
-
-Ptr<BaseFilter_GPU> cv::gpu::getLinearFilter_GPU(int srcType, int dstType, const Mat& kernel, Point anchor, int brd_type)
-{
-    using namespace cv::gpu::device::imgproc;
-
-    int sdepth = CV_MAT_DEPTH(srcType);
-    int scn = CV_MAT_CN(srcType);
-
-    CV_Assert(sdepth == CV_8U || sdepth == CV_16U || sdepth == CV_32F);
-    CV_Assert(scn == 1 || scn == 4);
-    CV_Assert(dstType == srcType);
-    CV_Assert(brd_type == BORDER_REFLECT101 || brd_type == BORDER_REPLICATE || brd_type == BORDER_CONSTANT || brd_type == BORDER_REFLECT || brd_type == BORDER_WRAP);
-
-    Size ksize = kernel.size();
-
-#if 0
-    if ((srcType == CV_8UC1 || srcType == CV_8UC4) && brd_type == BORDER_CONSTANT)
-    {
-        static const nppFilter2D_t cppFilter2D_callers[] = {0, nppiFilter_8u_C1R, 0, 0, nppiFilter_8u_C4R};
-
-        GpuMat gpu_krnl;
-        int nDivisor;
-        normalizeKernel(kernel, gpu_krnl, CV_32S, &nDivisor, true);
-
-        normalizeAnchor(anchor, ksize);
-
-        return Ptr<BaseFilter_GPU>(new NPPLinearFilter(ksize, anchor, gpu_krnl, nDivisor, cppFilter2D_callers[CV_MAT_CN(srcType)]));
-    }
-#endif
-
-    CV_Assert(ksize.width * ksize.height <= 16 * 16);
-
-    int gpuBorderType;
-    CV_Assert( tryConvertToGpuBorderType(brd_type, gpuBorderType) );
-
-    GpuMat gpu_krnl;
-    normalizeKernel(kernel, gpu_krnl, CV_32F);
-
-    normalizeAnchor(anchor, ksize);
-
-    gpuFilter2D_t func = 0;
-
-    switch (srcType)
-    {
-    case CV_8UC1:
-        func = filter2D_gpu<uchar, uchar>;
-        break;
-    case CV_8UC4:
-        func = filter2D_gpu<uchar4, uchar4>;
-        break;
-    case CV_16UC1:
-        func = filter2D_gpu<ushort, ushort>;
-        break;
-    case CV_16UC4:
-        func = filter2D_gpu<ushort4, ushort4>;
-        break;
-    case CV_32FC1:
-        func = filter2D_gpu<float, float>;
-        break;
-    case CV_32FC4:
-        func = filter2D_gpu<float4, float4>;
-        break;
-    }
-
-    return Ptr<BaseFilter_GPU>(new GpuFilter2D(ksize, anchor, func, gpu_krnl, gpuBorderType));
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createLinearFilter_GPU(int srcType, int dstType, const Mat& kernel, Point anchor, int borderType)
-{
-    Ptr<BaseFilter_GPU> linearFilter = getLinearFilter_GPU(srcType, dstType, kernel, anchor, borderType);
-
-    return createFilter2D_GPU(linearFilter, srcType, dstType);
-}
-
-void cv::gpu::filter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernel, Point anchor, int borderType, Stream& stream)
-{
-    if (ddepth < 0)
-        ddepth = src.depth();
-
-    int dst_type = CV_MAKE_TYPE(ddepth, src.channels());
-
-    Ptr<FilterEngine_GPU> f = createLinearFilter_GPU(src.type(), dst_type, kernel, anchor, borderType);
-
-    dst.create(src.size(), dst_type);
-
-    f->apply(src, dst, Rect(0, 0, src.cols, src.rows), stream);
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Separable Linear Filter
-
-namespace filter
-{
-    template <typename T, typename D>
-    void linearRow(PtrStepSzb src, PtrStepSzb dst, const float* kernel, int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void linearColumn(PtrStepSzb src, PtrStepSzb dst, const float* kernel, int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
-}
-
-namespace
-{
-    typedef NppStatus (*nppFilter1D_t)(const Npp8u * pSrc, Npp32s nSrcStep, Npp8u * pDst, Npp32s nDstStep, NppiSize oROI,
-        const Npp32s * pKernel, Npp32s nMaskSize, Npp32s nAnchor, Npp32s nDivisor);
-
-    typedef void (*gpuFilter1D_t)(PtrStepSzb src, PtrStepSzb dst, const float kernel[], int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
-
-    struct NppLinearRowFilter : public BaseRowFilter_GPU
-    {
-        NppLinearRowFilter(int ksize_, int anchor_, const GpuMat& kernel_, Npp32s nDivisor_, nppFilter1D_t func_) :
-            BaseRowFilter_GPU(ksize_, anchor_), kernel(kernel_), nDivisor(nDivisor_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz,
-                kernel.ptr<Npp32s>(), ksize, anchor, nDivisor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        GpuMat kernel;
-        Npp32s nDivisor;
-        nppFilter1D_t func;
-    };
-
-    struct GpuLinearRowFilter : public BaseRowFilter_GPU
-    {
-        GpuLinearRowFilter(int ksize_, int anchor_, const GpuMat& kernel_, gpuFilter1D_t func_, int brd_type_) :
-            BaseRowFilter_GPU(ksize_, anchor_), kernel(kernel_), func(func_), brd_type(brd_type_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            DeviceInfo devInfo;
-            int cc = devInfo.majorVersion() * 10 + devInfo.minorVersion();
-            func(src, dst, kernel.ptr<float>(), ksize, anchor, brd_type, cc, StreamAccessor::getStream(s));
-        }
-
-        GpuMat kernel;
-        gpuFilter1D_t func;
-        int brd_type;
-    };
-}
-
-Ptr<BaseRowFilter_GPU> cv::gpu::getLinearRowFilter_GPU(int srcType, int bufType, const Mat& rowKernel, int anchor, int borderType)
-{
-    static const gpuFilter1D_t funcs[7][4] =
-    {
-        {filter::linearRow<uchar, float>, 0, filter::linearRow<uchar3, float3>, filter::linearRow<uchar4, float4>},
-        {0, 0, 0, 0},
-        {filter::linearRow<ushort, float>, 0, filter::linearRow<ushort3, float3>, filter::linearRow<ushort4, float4>},
-        {filter::linearRow<short, float>, 0, filter::linearRow<short3, float3>, filter::linearRow<short4, float4>},
-        {filter::linearRow<int, float>, 0, filter::linearRow<int3, float3>, filter::linearRow<int4, float4>},
-        {filter::linearRow<float, float>, 0, filter::linearRow<float3, float3>, filter::linearRow<float4, float4>},
-        {0, 0, 0, 0}
-    };
-    static const nppFilter1D_t npp_funcs[] =
-    {
-        0, nppiFilterRow_8u_C1R, 0, 0, nppiFilterRow_8u_C4R
-    };
-
-    if ((bufType == srcType) && (srcType == CV_8UC1 || srcType == CV_8UC4))
-    {
-        CV_Assert( borderType == BORDER_CONSTANT );
-
-        GpuMat gpu_row_krnl;
-        int nDivisor;
-        normalizeKernel(rowKernel, gpu_row_krnl, CV_32S, &nDivisor, true);
-
-        const int ksize = gpu_row_krnl.cols;
-        normalizeAnchor(anchor, ksize);
-
-        return Ptr<BaseRowFilter_GPU>(new NppLinearRowFilter(ksize, anchor, gpu_row_krnl, nDivisor, npp_funcs[CV_MAT_CN(srcType)]));
-    }
-
-    CV_Assert( borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP );
-
-    int gpuBorderType;
-    CV_Assert( tryConvertToGpuBorderType(borderType, gpuBorderType) );
-
-    const int sdepth = CV_MAT_DEPTH(srcType);
-    const int cn = CV_MAT_CN(srcType);
-    CV_Assert( sdepth <= CV_64F && cn <= 4 );
-    CV_Assert( CV_MAT_DEPTH(bufType) == CV_32F && CV_MAT_CN(bufType) == cn );
-
-    const gpuFilter1D_t func = funcs[sdepth][cn - 1];
-    CV_Assert( func != 0 );
-
-    GpuMat gpu_row_krnl;
-    normalizeKernel(rowKernel, gpu_row_krnl, CV_32F);
-
-    const int ksize = gpu_row_krnl.cols;
-    CV_Assert( ksize > 0 && ksize <= 32 );
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseRowFilter_GPU>(new GpuLinearRowFilter(ksize, anchor, gpu_row_krnl, func, gpuBorderType));
-}
-
-namespace
-{
-    struct NppLinearColumnFilter : public BaseColumnFilter_GPU
-    {
-        NppLinearColumnFilter(int ksize_, int anchor_, const GpuMat& kernel_, Npp32s nDivisor_, nppFilter1D_t func_) :
-            BaseColumnFilter_GPU(ksize_, anchor_), kernel(kernel_), nDivisor(nDivisor_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz,
-                kernel.ptr<Npp32s>(), ksize, anchor, nDivisor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        GpuMat kernel;
-        Npp32s nDivisor;
-        nppFilter1D_t func;
-    };
-
-    struct GpuLinearColumnFilter : public BaseColumnFilter_GPU
-    {
-        GpuLinearColumnFilter(int ksize_, int anchor_, const GpuMat& kernel_, gpuFilter1D_t func_, int brd_type_) :
-            BaseColumnFilter_GPU(ksize_, anchor_), kernel(kernel_), func(func_), brd_type(brd_type_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            DeviceInfo devInfo;
-            int cc = devInfo.majorVersion() * 10 + devInfo.minorVersion();
-            if (ksize > 16 && cc < 20)
-                CV_Error(CV_StsNotImplemented, "column linear filter doesn't implemented for kernel size > 16 for device with compute capabilities less than 2.0");
-
-            func(src, dst, kernel.ptr<float>(), ksize, anchor, brd_type, cc, StreamAccessor::getStream(s));
-        }
-
-        GpuMat kernel;
-        gpuFilter1D_t func;
-        int brd_type;
-    };
-}
-
-Ptr<BaseColumnFilter_GPU> cv::gpu::getLinearColumnFilter_GPU(int bufType, int dstType, const Mat& columnKernel, int anchor, int borderType)
-{
-    static const gpuFilter1D_t funcs[7][4] =
-    {
-        {filter::linearColumn<float, uchar>, 0, filter::linearColumn<float3, uchar3>, filter::linearColumn<float4, uchar4>},
-        {0, 0, 0, 0},
-        {filter::linearColumn<float, ushort>, 0, filter::linearColumn<float3, ushort3>, filter::linearColumn<float4, ushort4>},
-        {filter::linearColumn<float, short>, 0, filter::linearColumn<float3, short3>, filter::linearColumn<float4, short4>},
-        {filter::linearColumn<float, int>, 0, filter::linearColumn<float3, int3>, filter::linearColumn<float4, int4>},
-        {filter::linearColumn<float, float>, 0, filter::linearColumn<float3, float3>, filter::linearColumn<float4, float4>},
-        {0, 0, 0, 0}
-    };
-    static const nppFilter1D_t npp_funcs[] =
-    {
-        0, nppiFilterColumn_8u_C1R, 0, 0, nppiFilterColumn_8u_C4R
-    };
-
-    if ((bufType == dstType) && (bufType == CV_8UC1 || bufType == CV_8UC4))
-    {
-        CV_Assert( borderType == BORDER_CONSTANT );
-
-        GpuMat gpu_col_krnl;
-        int nDivisor;
-        normalizeKernel(columnKernel, gpu_col_krnl, CV_32S, &nDivisor, true);
-
-        const int ksize = gpu_col_krnl.cols;
-        normalizeAnchor(anchor, ksize);
-
-        return Ptr<BaseColumnFilter_GPU>(new NppLinearColumnFilter(ksize, anchor, gpu_col_krnl, nDivisor, npp_funcs[CV_MAT_CN(bufType)]));
-    }
-
-    CV_Assert( borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP );
-
-    int gpuBorderType;
-    CV_Assert( tryConvertToGpuBorderType(borderType, gpuBorderType) );
-
-    const int ddepth = CV_MAT_DEPTH(dstType);
-    const int cn = CV_MAT_CN(dstType);
-    CV_Assert( ddepth <= CV_64F && cn <= 4 );
-    CV_Assert( CV_MAT_DEPTH(bufType) == CV_32F && CV_MAT_CN(bufType) == cn );
-
-    gpuFilter1D_t func = funcs[ddepth][cn - 1];
-    CV_Assert( func != 0 );
-
-    GpuMat gpu_col_krnl;
-    normalizeKernel(columnKernel, gpu_col_krnl, CV_32F);
-
-    const int ksize = gpu_col_krnl.cols;
-    CV_Assert(ksize > 0 && ksize <= 32);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseColumnFilter_GPU>(new GpuLinearColumnFilter(ksize, anchor, gpu_col_krnl, func, gpuBorderType));
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel, const Mat& columnKernel,
-    const Point& anchor, int rowBorderType, int columnBorderType)
-{
-    if (columnBorderType < 0)
-        columnBorderType = rowBorderType;
-
-    int cn = CV_MAT_CN(srcType);
-    int bdepth = CV_32F;
-    int bufType = CV_MAKETYPE(bdepth, cn);
-
-    Ptr<BaseRowFilter_GPU> rowFilter = getLinearRowFilter_GPU(srcType, bufType, rowKernel, anchor.x, rowBorderType);
-    Ptr<BaseColumnFilter_GPU> columnFilter = getLinearColumnFilter_GPU(bufType, dstType, columnKernel, anchor.y, columnBorderType);
-
-    return createSeparableFilter_GPU(rowFilter, columnFilter, srcType, bufType, dstType);
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel, const Mat& columnKernel, GpuMat& buf,
-    const Point& anchor, int rowBorderType, int columnBorderType)
-{
-    if (columnBorderType < 0)
-        columnBorderType = rowBorderType;
-
-    int cn = CV_MAT_CN(srcType);
-    int bdepth = CV_32F;
-    int bufType = CV_MAKETYPE(bdepth, cn);
-
-    Ptr<BaseRowFilter_GPU> rowFilter = getLinearRowFilter_GPU(srcType, bufType, rowKernel, anchor.x, rowBorderType);
-    Ptr<BaseColumnFilter_GPU> columnFilter = getLinearColumnFilter_GPU(bufType, dstType, columnKernel, anchor.y, columnBorderType);
-
-    return createSeparableFilter_GPU(rowFilter, columnFilter, srcType, bufType, dstType, buf);
-}
-
-void cv::gpu::sepFilter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY,
-                          Point anchor, int rowBorderType, int columnBorderType)
-{
-    if( ddepth < 0 )
-        ddepth = src.depth();
-
-    dst.create(src.size(), CV_MAKETYPE(ddepth, src.channels()));
-
-    Ptr<FilterEngine_GPU> f = createSeparableLinearFilter_GPU(src.type(), dst.type(), kernelX, kernelY, anchor, rowBorderType, columnBorderType);
-    f->apply(src, dst, Rect(0, 0, src.cols, src.rows));
-}
-
-void cv::gpu::sepFilter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY, GpuMat& buf,
-                          Point anchor, int rowBorderType, int columnBorderType,
-                          Stream& stream)
-{
-    if( ddepth < 0 )
-        ddepth = src.depth();
-
-    dst.create(src.size(), CV_MAKETYPE(ddepth, src.channels()));
-
-    Ptr<FilterEngine_GPU> f = createSeparableLinearFilter_GPU(src.type(), dst.type(), kernelX, kernelY, buf, anchor, rowBorderType, columnBorderType);
-    f->apply(src, dst, Rect(0, 0, src.cols, src.rows), stream);
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Deriv Filter
-
-Ptr<FilterEngine_GPU> cv::gpu::createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize, int rowBorderType, int columnBorderType)
-{
-    Mat kx, ky;
-    getDerivKernels(kx, ky, dx, dy, ksize, false, CV_32F);
-    return createSeparableLinearFilter_GPU(srcType, dstType, kx, ky, Point(-1,-1), rowBorderType, columnBorderType);
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize, GpuMat& buf, int rowBorderType, int columnBorderType)
-{
-    Mat kx, ky;
-    getDerivKernels(kx, ky, dx, dy, ksize, false, CV_32F);
-    return createSeparableLinearFilter_GPU(srcType, dstType, kx, ky, buf, Point(-1,-1), rowBorderType, columnBorderType);
-}
-
-void cv::gpu::Sobel(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, int ksize, double scale, int rowBorderType, int columnBorderType)
-{
-    GpuMat buf;
-    Sobel(src, dst, ddepth, dx, dy, buf, ksize, scale, rowBorderType, columnBorderType);
-}
-
-void cv::gpu::Sobel(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, int ksize, double scale, int rowBorderType, int columnBorderType, Stream& stream)
-{
-    Mat kx, ky;
-    getDerivKernels(kx, ky, dx, dy, ksize, false, CV_32F);
-
-    if (scale != 1)
-    {
-        // usually the smoothing part is the slowest to compute,
-        // so try to scale it instead of the faster differenciating part
-        if (dx == 0)
-            kx *= scale;
-        else
-            ky *= scale;
-    }
-
-    sepFilter2D(src, dst, ddepth, kx, ky, buf, Point(-1,-1), rowBorderType, columnBorderType, stream);
-}
-
-void cv::gpu::Scharr(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, double scale, int rowBorderType, int columnBorderType)
-{
-    GpuMat buf;
-    Scharr(src, dst, ddepth, dx, dy, buf, scale, rowBorderType, columnBorderType);
-}
-
-void cv::gpu::Scharr(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, GpuMat& buf, double scale, int rowBorderType, int columnBorderType, Stream& stream)
-{
-    Mat kx, ky;
-    getDerivKernels(kx, ky, dx, dy, -1, false, CV_32F);
-
-    if( scale != 1 )
-    {
-        // usually the smoothing part is the slowest to compute,
-        // so try to scale it instead of the faster differenciating part
-        if( dx == 0 )
-            kx *= scale;
-        else
-            ky *= scale;
-    }
-
-    sepFilter2D(src, dst, ddepth, kx, ky, buf, Point(-1,-1), rowBorderType, columnBorderType, stream);
-}
-
-void cv::gpu::Laplacian(const GpuMat& src, GpuMat& dst, int ddepth, int ksize, double scale, int borderType, Stream& stream)
-{
-    CV_Assert(ksize == 1 || ksize == 3);
-
-    static const int K[2][9] =
-    {
-        {0, 1, 0, 1, -4, 1, 0, 1, 0},
-        {2, 0, 2, 0, -8, 0, 2, 0, 2}
-    };
-    Mat kernel(3, 3, CV_32S, (void*)K[ksize == 3]);
-    if (scale != 1)
-        kernel *= scale;
-
-    filter2D(src, dst, ddepth, kernel, Point(-1,-1), borderType, stream);
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Gaussian Filter
-
-Ptr<FilterEngine_GPU> cv::gpu::createGaussianFilter_GPU(int type, Size ksize, double sigma1, double sigma2, int rowBorderType, int columnBorderType)
-{
-    int depth = CV_MAT_DEPTH(type);
-
-    if (sigma2 <= 0)
-        sigma2 = sigma1;
-
-    // automatic detection of kernel size from sigma
-    if (ksize.width <= 0 && sigma1 > 0)
-        ksize.width = cvRound(sigma1 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
-    if (ksize.height <= 0 && sigma2 > 0)
-        ksize.height = cvRound(sigma2 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
-
-    CV_Assert( ksize.width > 0 && ksize.width % 2 == 1 && ksize.height > 0 && ksize.height % 2 == 1 );
-
-    sigma1 = std::max(sigma1, 0.0);
-    sigma2 = std::max(sigma2, 0.0);
-
-    Mat kx = getGaussianKernel( ksize.width, sigma1, std::max(depth, CV_32F) );
-    Mat ky;
-    if( ksize.height == ksize.width && std::abs(sigma1 - sigma2) < DBL_EPSILON )
-        ky = kx;
-    else
-        ky = getGaussianKernel( ksize.height, sigma2, std::max(depth, CV_32F) );
-
-    return createSeparableLinearFilter_GPU(type, type, kx, ky, Point(-1,-1), rowBorderType, columnBorderType);
-}
-
-Ptr<FilterEngine_GPU> cv::gpu::createGaussianFilter_GPU(int type, Size ksize, GpuMat& buf, double sigma1, double sigma2, int rowBorderType, int columnBorderType)
-{
-    int depth = CV_MAT_DEPTH(type);
-
-    if (sigma2 <= 0)
-        sigma2 = sigma1;
-
-    // automatic detection of kernel size from sigma
-    if (ksize.width <= 0 && sigma1 > 0)
-        ksize.width = cvRound(sigma1 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
-    if (ksize.height <= 0 && sigma2 > 0)
-        ksize.height = cvRound(sigma2 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
-
-    CV_Assert( ksize.width > 0 && ksize.width % 2 == 1 && ksize.height > 0 && ksize.height % 2 == 1 );
-
-    sigma1 = std::max(sigma1, 0.0);
-    sigma2 = std::max(sigma2, 0.0);
-
-    Mat kx = getGaussianKernel( ksize.width, sigma1, std::max(depth, CV_32F) );
-    Mat ky;
-    if( ksize.height == ksize.width && std::abs(sigma1 - sigma2) < DBL_EPSILON )
-        ky = kx;
-    else
-        ky = getGaussianKernel( ksize.height, sigma2, std::max(depth, CV_32F) );
-
-    return createSeparableLinearFilter_GPU(type, type, kx, ky, buf, Point(-1,-1), rowBorderType, columnBorderType);
-}
-
-void cv::gpu::GaussianBlur(const GpuMat& src, GpuMat& dst, Size ksize, double sigma1, double sigma2, int rowBorderType, int columnBorderType)
-{
-    if (ksize.width == 1 && ksize.height == 1)
-    {
-        src.copyTo(dst);
-        return;
-    }
-
-    dst.create(src.size(), src.type());
-
-    Ptr<FilterEngine_GPU> f = createGaussianFilter_GPU(src.type(), ksize, sigma1, sigma2, rowBorderType, columnBorderType);
-    f->apply(src, dst, Rect(0, 0, src.cols, src.rows));
-}
-
-void cv::gpu::GaussianBlur(const GpuMat& src, GpuMat& dst, Size ksize, GpuMat& buf, double sigma1, double sigma2, int rowBorderType, int columnBorderType, Stream& stream)
-{
-    if (ksize.width == 1 && ksize.height == 1)
-    {
-        src.copyTo(dst);
-        return;
-    }
-
-    dst.create(src.size(), src.type());
-
-    Ptr<FilterEngine_GPU> f = createGaussianFilter_GPU(src.type(), ksize, buf, sigma1, sigma2, rowBorderType, columnBorderType);
-    f->apply(src, dst, Rect(0, 0, src.cols, src.rows), stream);
-}
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Image Rank Filter
-
-namespace
-{
-    typedef NppStatus (*nppFilterRank_t)(const Npp8u * pSrc, Npp32s nSrcStep, Npp8u * pDst, Npp32s nDstStep, NppiSize oSizeROI,
-        NppiSize oMaskSize, NppiPoint oAnchor);
-
-    struct NPPRankFilter : public BaseFilter_GPU
-    {
-        NPPRankFilter(const Size& ksize_, const Point& anchor_, nppFilterRank_t func_) : BaseFilter_GPU(ksize_, anchor_), func(func_) {}
-
-        virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& s = Stream::Null())
-        {
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-            NppiSize oKernelSize;
-            oKernelSize.height = ksize.height;
-            oKernelSize.width = ksize.width;
-            NppiPoint oAnchor;
-            oAnchor.x = anchor.x;
-            oAnchor.y = anchor.y;
-
-            cudaStream_t stream = StreamAccessor::getStream(s);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, oKernelSize, oAnchor) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        nppFilterRank_t func;
-    };
-}
-
-Ptr<BaseFilter_GPU> cv::gpu::getMaxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor)
-{
-    static const nppFilterRank_t nppFilterRank_callers[] = {0, nppiFilterMax_8u_C1R, 0, 0, nppiFilterMax_8u_C4R};
-
-    CV_Assert((srcType == CV_8UC1 || srcType == CV_8UC4) && dstType == srcType);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseFilter_GPU>(new NPPRankFilter(ksize, anchor, nppFilterRank_callers[CV_MAT_CN(srcType)]));
-}
-
-Ptr<BaseFilter_GPU> cv::gpu::getMinFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor)
-{
-    static const nppFilterRank_t nppFilterRank_callers[] = {0, nppiFilterMin_8u_C1R, 0, 0, nppiFilterMin_8u_C4R};
-
-    CV_Assert((srcType == CV_8UC1 || srcType == CV_8UC4) && dstType == srcType);
-
-    normalizeAnchor(anchor, ksize);
-
-    return Ptr<BaseFilter_GPU>(new NPPRankFilter(ksize, anchor, nppFilterRank_callers[CV_MAT_CN(srcType)]));
-}
-
-#endif
diff --git a/modules/gpu/src/gftt.cpp b/modules/gpu/src/gftt.cpp
deleted file mode 100644
index 2c50ed69d2..0000000000
--- a/modules/gpu/src/gftt.cpp
+++ /dev/null
@@ -1,169 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::GoodFeaturesToTrackDetector_GPU::operator ()(const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace gfft
-    {
-        int findCorners_gpu(PtrStepSzf eig, float threshold, PtrStepSzb mask, float2* corners, int max_count);
-        void sortCorners_gpu(PtrStepSzf eig, float2* corners, int count);
-    }
-}}}
-
-void cv::gpu::GoodFeaturesToTrackDetector_GPU::operator ()(const GpuMat& image, GpuMat& corners, const GpuMat& mask)
-{
-    using namespace cv::gpu::device::gfft;
-
-    CV_Assert(qualityLevel > 0 && minDistance >= 0 && maxCorners >= 0);
-    CV_Assert(mask.empty() || (mask.type() == CV_8UC1 && mask.size() == image.size()));
-
-    ensureSizeIsEnough(image.size(), CV_32F, eig_);
-
-    if (useHarrisDetector)
-        cornerHarris(image, eig_, Dx_, Dy_, buf_, blockSize, 3, harrisK);
-    else
-        cornerMinEigenVal(image, eig_, Dx_, Dy_, buf_, blockSize, 3);
-
-    double maxVal = 0;
-    minMax(eig_, 0, &maxVal, GpuMat(), minMaxbuf_);
-
-    ensureSizeIsEnough(1, std::max(1000, static_cast<int>(image.size().area() * 0.05)), CV_32FC2, tmpCorners_);
-
-    int total = findCorners_gpu(eig_, static_cast<float>(maxVal * qualityLevel), mask, tmpCorners_.ptr<float2>(), tmpCorners_.cols);
-
-    if (total == 0)
-    {
-        corners.release();
-        return;
-    }
-
-    sortCorners_gpu(eig_, tmpCorners_.ptr<float2>(), total);
-
-    if (minDistance < 1)
-        tmpCorners_.colRange(0, maxCorners > 0 ? std::min(maxCorners, total) : total).copyTo(corners);
-    else
-    {
-        std::vector<Point2f> tmp(total);
-        Mat tmpMat(1, total, CV_32FC2, (void*)&tmp[0]);
-        tmpCorners_.colRange(0, total).download(tmpMat);
-
-        std::vector<Point2f> tmp2;
-        tmp2.reserve(total);
-
-        const int cell_size = cvRound(minDistance);
-        const int grid_width = (image.cols + cell_size - 1) / cell_size;
-        const int grid_height = (image.rows + cell_size - 1) / cell_size;
-
-        std::vector< std::vector<Point2f> > grid(grid_width * grid_height);
-
-        for (int i = 0; i < total; ++i)
-        {
-            Point2f p = tmp[i];
-
-            bool good = true;
-
-            int x_cell = static_cast<int>(p.x / cell_size);
-            int y_cell = static_cast<int>(p.y / cell_size);
-
-            int x1 = x_cell - 1;
-            int y1 = y_cell - 1;
-            int x2 = x_cell + 1;
-            int y2 = y_cell + 1;
-
-            // boundary check
-            x1 = std::max(0, x1);
-            y1 = std::max(0, y1);
-            x2 = std::min(grid_width - 1, x2);
-            y2 = std::min(grid_height - 1, y2);
-
-            for (int yy = y1; yy <= y2; yy++)
-            {
-                for (int xx = x1; xx <= x2; xx++)
-                {
-                    std::vector<Point2f>& m = grid[yy * grid_width + xx];
-
-                    if (!m.empty())
-                    {
-                        for(size_t j = 0; j < m.size(); j++)
-                        {
-                            float dx = p.x - m[j].x;
-                            float dy = p.y - m[j].y;
-
-                            if (dx * dx + dy * dy < minDistance * minDistance)
-                            {
-                                good = false;
-                                goto break_out;
-                            }
-                        }
-                    }
-                }
-            }
-
-            break_out:
-
-            if(good)
-            {
-                grid[y_cell * grid_width + x_cell].push_back(p);
-
-                tmp2.push_back(p);
-
-                if (maxCorners > 0 && tmp2.size() == static_cast<size_t>(maxCorners))
-                    break;
-            }
-        }
-
-        corners.upload(Mat(1, static_cast<int>(tmp2.size()), CV_32FC2, &tmp2[0]));
-    }
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/global_motion.cpp b/modules/gpu/src/global_motion.cpp
index 36c917744e..a9cc66954d 100644
--- a/modules/gpu/src/global_motion.cpp
+++ b/modules/gpu/src/global_motion.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,13 +47,13 @@ using namespace cv::gpu;
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::compactPoints(GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
+void cv::gpu::compactPoints(GpuMat&, GpuMat&, const GpuMat&) { throw_no_cuda(); }
 void cv::gpu::calcWobbleSuppressionMaps(
-        int, int, int, Size, const Mat&, const Mat&, GpuMat&, GpuMat&) { throw_nogpu(); }
+        int, int, int, Size, const Mat&, const Mat&, GpuMat&, GpuMat&) { throw_no_cuda(); }
 
 #else
 
-namespace cv { namespace gpu { namespace device { namespace globmotion {
+namespace cv { namespace gpu { namespace cudev { namespace globmotion {
 
     int compactPoints(int N, float *points0, float *points1, const uchar *mask);
 
@@ -70,7 +70,7 @@ void cv::gpu::compactPoints(GpuMat &points0, GpuMat &points1, const GpuMat &mask
     CV_Assert(points0.cols == mask.cols && points1.cols == mask.cols);
 
     int npoints = points0.cols;
-    int remaining = cv::gpu::device::globmotion::compactPoints(
+    int remaining = cv::gpu::cudev::globmotion::compactPoints(
             npoints, (float*)points0.data, (float*)points1.data, mask.data);
 
     points0 = points0.colRange(0, remaining);
@@ -88,7 +88,7 @@ void cv::gpu::calcWobbleSuppressionMaps(
     mapx.create(size, CV_32F);
     mapy.create(size, CV_32F);
 
-    cv::gpu::device::globmotion::calcWobbleSuppressionMaps(
+    cv::gpu::cudev::globmotion::calcWobbleSuppressionMaps(
                 left, idx, right, size.width, size.height,
                 ml.ptr<float>(), mr.ptr<float>(), mapx, mapy);
 }
diff --git a/modules/gpu/src/graphcuts.cpp b/modules/gpu/src/graphcuts.cpp
index 49230e65a9..40ccd04710 100644
--- a/modules/gpu/src/graphcuts.cpp
+++ b/modules/gpu/src/graphcuts.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -44,15 +44,15 @@
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::connectivityMask(const GpuMat&, GpuMat&, const cv::Scalar&, const cv::Scalar&, Stream&) { throw_nogpu(); }
-void cv::gpu::labelComponents(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
+void cv::gpu::connectivityMask(const GpuMat&, GpuMat&, const cv::Scalar&, const cv::Scalar&, Stream&) { throw_no_cuda(); }
+void cv::gpu::labelComponents(const GpuMat&, GpuMat&, int, Stream&) { throw_no_cuda(); }
 
 #else /* !defined (HAVE_CUDA) */
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace ccl
     {
@@ -81,12 +81,12 @@ void cv::gpu::connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scal
 
     static const func_t suppotLookup[8][4] =
     {   //    1,    2,     3,     4
-        { device::ccl::computeEdges<uchar>,  0,  device::ccl::computeEdges<uchar3>,  device::ccl::computeEdges<uchar4>  },// CV_8U
+        { cudev::ccl::computeEdges<uchar>,  0,  cudev::ccl::computeEdges<uchar3>,  cudev::ccl::computeEdges<uchar4>  },// CV_8U
         { 0,                                 0,  0,                                  0                                  },// CV_16U
-        { device::ccl::computeEdges<ushort>, 0,  device::ccl::computeEdges<ushort3>, device::ccl::computeEdges<ushort4> },// CV_8S
+        { cudev::ccl::computeEdges<ushort>, 0,  cudev::ccl::computeEdges<ushort3>, cudev::ccl::computeEdges<ushort4> },// CV_8S
         { 0,                                 0,  0,                                  0                                  },// CV_16S
-        { device::ccl::computeEdges<int>,    0,  0,                                  0                                  },// CV_32S
-        { device::ccl::computeEdges<float>,  0,  0,                                  0                                  },// CV_32F
+        { cudev::ccl::computeEdges<int>,    0,  0,                                  0                                  },// CV_32S
+        { cudev::ccl::computeEdges<float>,  0,  0,                                  0                                  },// CV_32F
         { 0,                                 0,  0,                                  0                                  },// CV_64F
         { 0,                                 0,  0,                                  0                                  } // CV_USRTYPE1
     };
@@ -107,12 +107,12 @@ void cv::gpu::labelComponents(const GpuMat& mask, GpuMat& components, int flags,
     CV_Assert(!mask.empty() && mask.type() == CV_8U);
 
     if (!deviceSupports(SHARED_ATOMICS))
-        CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
+        CV_Error(cv::Error::StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
 
     components.create(mask.size(), CV_32SC1);
 
     cudaStream_t stream = StreamAccessor::getStream(s);
-    device::ccl::labelComponents(mask, components, flags, stream);
+    cudev::ccl::labelComponents(mask, components, flags, stream);
 }
 
 namespace
diff --git a/modules/gpu/src/hog.cpp b/modules/gpu/src/hog.cpp
index a3af032a1b..a599fa8f1b 100644
--- a/modules/gpu/src/hog.cpp
+++ b/modules/gpu/src/hog.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -44,25 +44,25 @@
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::HOGDescriptor::HOGDescriptor(Size, Size, Size, Size, int, double, double, bool, int) { throw_nogpu(); }
-size_t cv::gpu::HOGDescriptor::getDescriptorSize() const { throw_nogpu(); return 0; }
-size_t cv::gpu::HOGDescriptor::getBlockHistogramSize() const { throw_nogpu(); return 0; }
-double cv::gpu::HOGDescriptor::getWinSigma() const { throw_nogpu(); return 0; }
-bool cv::gpu::HOGDescriptor::checkDetectorSize() const { throw_nogpu(); return false; }
-void cv::gpu::HOGDescriptor::setSVMDetector(const std::vector<float>&) { throw_nogpu(); }
-void cv::gpu::HOGDescriptor::detect(const GpuMat&, std::vector<Point>&, double, Size, Size) { throw_nogpu(); }
-void cv::gpu::HOGDescriptor::detectMultiScale(const GpuMat&, std::vector<Rect>&, double, Size, Size, double, int) { throw_nogpu(); }
-void cv::gpu::HOGDescriptor::computeBlockHistograms(const GpuMat&) { throw_nogpu(); }
-void cv::gpu::HOGDescriptor::getDescriptors(const GpuMat&, Size, GpuMat&, int) { throw_nogpu(); }
-std::vector<float> cv::gpu::HOGDescriptor::getDefaultPeopleDetector() { throw_nogpu(); return std::vector<float>(); }
-std::vector<float> cv::gpu::HOGDescriptor::getPeopleDetector48x96() { throw_nogpu(); return std::vector<float>(); }
-std::vector<float> cv::gpu::HOGDescriptor::getPeopleDetector64x128() { throw_nogpu(); return std::vector<float>(); }
-void cv::gpu::HOGDescriptor::computeConfidence(const GpuMat&, std::vector<Point>&, double, Size, Size, std::vector<Point>&, std::vector<double>&) { throw_nogpu(); }
-void cv::gpu::HOGDescriptor::computeConfidenceMultiScale(const GpuMat&, std::vector<Rect>&, double, Size, Size, std::vector<HOGConfidence>&, int) { throw_nogpu(); }
+cv::gpu::HOGDescriptor::HOGDescriptor(Size, Size, Size, Size, int, double, double, bool, int) { throw_no_cuda(); }
+size_t cv::gpu::HOGDescriptor::getDescriptorSize() const { throw_no_cuda(); return 0; }
+size_t cv::gpu::HOGDescriptor::getBlockHistogramSize() const { throw_no_cuda(); return 0; }
+double cv::gpu::HOGDescriptor::getWinSigma() const { throw_no_cuda(); return 0; }
+bool cv::gpu::HOGDescriptor::checkDetectorSize() const { throw_no_cuda(); return false; }
+void cv::gpu::HOGDescriptor::setSVMDetector(const std::vector<float>&) { throw_no_cuda(); }
+void cv::gpu::HOGDescriptor::detect(const GpuMat&, std::vector<Point>&, double, Size, Size) { throw_no_cuda(); }
+void cv::gpu::HOGDescriptor::detectMultiScale(const GpuMat&, std::vector<Rect>&, double, Size, Size, double, int) { throw_no_cuda(); }
+void cv::gpu::HOGDescriptor::computeBlockHistograms(const GpuMat&) { throw_no_cuda(); }
+void cv::gpu::HOGDescriptor::getDescriptors(const GpuMat&, Size, GpuMat&, int) { throw_no_cuda(); }
+std::vector<float> cv::gpu::HOGDescriptor::getDefaultPeopleDetector() { throw_no_cuda(); return std::vector<float>(); }
+std::vector<float> cv::gpu::HOGDescriptor::getPeopleDetector48x96() { throw_no_cuda(); return std::vector<float>(); }
+std::vector<float> cv::gpu::HOGDescriptor::getPeopleDetector64x128() { throw_no_cuda(); return std::vector<float>(); }
+void cv::gpu::HOGDescriptor::computeConfidence(const GpuMat&, std::vector<Point>&, double, Size, Size, std::vector<Point>&, std::vector<double>&) { throw_no_cuda(); }
+void cv::gpu::HOGDescriptor::computeConfidenceMultiScale(const GpuMat&, std::vector<Rect>&, double, Size, Size, std::vector<HOGConfidence>&, int) { throw_no_cuda(); }
 
 #else
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace hog
     {
@@ -102,7 +102,7 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-using namespace ::cv::gpu::device;
+using namespace ::cv::gpu::cudev;
 
 cv::gpu::HOGDescriptor::HOGDescriptor(Size win_size_, Size block_size_, Size block_stride_, Size cell_size_,
                                       int nbins_, double win_sigma_, double threshold_L2hys_, bool gamma_correction_, int nlevels_)
@@ -260,7 +260,7 @@ void cv::gpu::HOGDescriptor::getDescriptors(const GpuMat& img, Size win_stride,
                                     win_stride.height, win_stride.width, img.rows, img.cols, block_hists.ptr<float>(), descriptors);
         break;
     default:
-        CV_Error(CV_StsBadArg, "Unknown descriptor format");
+        CV_Error(cv::Error::StsBadArg, "Unknown descriptor format");
     }
 }
 
@@ -351,7 +351,7 @@ void cv::gpu::HOGDescriptor::computeConfidenceMultiScale(const GpuMat& img, std:
 
         Size scaled_win_size(cvRound(win_size.width * scale), cvRound(win_size.height * scale));
         for (size_t j = 0; j < locations.size(); j++)
-            all_candidates.push_back(Rect(Point2d((CvPoint)locations[j]) * scale, scaled_win_size));
+            all_candidates.push_back(Rect(Point2d(locations[j]) * scale, scaled_win_size));
     }
 
     found_locations.assign(all_candidates.begin(), all_candidates.end());
@@ -443,7 +443,7 @@ void cv::gpu::HOGDescriptor::detectMultiScale(const GpuMat& img, std::vector<Rec
         detect(smaller_img, locations, hit_threshold, win_stride, padding);
         Size scaled_win_size(cvRound(win_size.width * scale), cvRound(win_size.height * scale));
         for (size_t j = 0; j < locations.size(); j++)
-            all_candidates.push_back(Rect(Point2d((CvPoint)locations[j]) * scale, scaled_win_size));
+            all_candidates.push_back(Rect(Point2d(locations[j]) * scale, scaled_win_size));
     }
 
     found_locations.assign(all_candidates.begin(), all_candidates.end());
diff --git a/modules/gpu/src/hough.cpp b/modules/gpu/src/hough.cpp
deleted file mode 100644
index a1e7033436..0000000000
--- a/modules/gpu/src/hough.cpp
+++ /dev/null
@@ -1,1421 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::HoughLines(const GpuMat&, GpuMat&, float, float, int, bool, int) { throw_nogpu(); }
-void cv::gpu::HoughLines(const GpuMat&, GpuMat&, HoughLinesBuf&, float, float, int, bool, int) { throw_nogpu(); }
-void cv::gpu::HoughLinesDownload(const GpuMat&, OutputArray, OutputArray) { throw_nogpu(); }
-
-void cv::gpu::HoughLinesP(const GpuMat&, GpuMat&, HoughLinesBuf&, float, float, int, int, int) { throw_nogpu(); }
-
-void cv::gpu::HoughCircles(const GpuMat&, GpuMat&, int, float, float, int, int, int, int, int) { throw_nogpu(); }
-void cv::gpu::HoughCircles(const GpuMat&, GpuMat&, HoughCirclesBuf&, int, float, float, int, int, int, int, int) { throw_nogpu(); }
-void cv::gpu::HoughCirclesDownload(const GpuMat&, OutputArray) { throw_nogpu(); }
-
-Ptr<GeneralizedHough_GPU> cv::gpu::GeneralizedHough_GPU::create(int) { throw_nogpu(); return Ptr<GeneralizedHough_GPU>(); }
-cv::gpu::GeneralizedHough_GPU::~GeneralizedHough_GPU() {}
-void cv::gpu::GeneralizedHough_GPU::setTemplate(const GpuMat&, int, Point) { throw_nogpu(); }
-void cv::gpu::GeneralizedHough_GPU::setTemplate(const GpuMat&, const GpuMat&, const GpuMat&, Point) { throw_nogpu(); }
-void cv::gpu::GeneralizedHough_GPU::detect(const GpuMat&, GpuMat&, int) { throw_nogpu(); }
-void cv::gpu::GeneralizedHough_GPU::detect(const GpuMat&, const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::GeneralizedHough_GPU::download(const GpuMat&, OutputArray, OutputArray) { throw_nogpu(); }
-void cv::gpu::GeneralizedHough_GPU::release() {}
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        int buildPointList_gpu(PtrStepSzb src, unsigned int* list);
-    }
-}}}
-
-//////////////////////////////////////////////////////////
-// HoughLines
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        void linesAccum_gpu(const unsigned int* list, int count, PtrStepSzi accum, float rho, float theta, size_t sharedMemPerBlock, bool has20);
-        int linesGetResult_gpu(PtrStepSzi accum, float2* out, int* votes, int maxSize, float rho, float theta, int threshold, bool doSort);
-    }
-}}}
-
-void cv::gpu::HoughLines(const GpuMat& src, GpuMat& lines, float rho, float theta, int threshold, bool doSort, int maxLines)
-{
-    HoughLinesBuf buf;
-    HoughLines(src, lines, buf, rho, theta, threshold, doSort, maxLines);
-}
-
-void cv::gpu::HoughLines(const GpuMat& src, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int threshold, bool doSort, int maxLines)
-{
-    using namespace cv::gpu::device::hough;
-
-    CV_Assert(src.type() == CV_8UC1);
-    CV_Assert(src.cols < std::numeric_limits<unsigned short>::max());
-    CV_Assert(src.rows < std::numeric_limits<unsigned short>::max());
-
-    ensureSizeIsEnough(1, src.size().area(), CV_32SC1, buf.list);
-    unsigned int* srcPoints = buf.list.ptr<unsigned int>();
-
-    const int pointsCount = buildPointList_gpu(src, srcPoints);
-    if (pointsCount == 0)
-    {
-        lines.release();
-        return;
-    }
-
-    const int numangle = cvRound(CV_PI / theta);
-    const int numrho = cvRound(((src.cols + src.rows) * 2 + 1) / rho);
-    CV_Assert(numangle > 0 && numrho > 0);
-
-    ensureSizeIsEnough(numangle + 2, numrho + 2, CV_32SC1, buf.accum);
-    buf.accum.setTo(Scalar::all(0));
-
-    DeviceInfo devInfo;
-    linesAccum_gpu(srcPoints, pointsCount, buf.accum, rho, theta, devInfo.sharedMemPerBlock(), devInfo.supports(FEATURE_SET_COMPUTE_20));
-
-    ensureSizeIsEnough(2, maxLines, CV_32FC2, lines);
-
-    int linesCount = linesGetResult_gpu(buf.accum, lines.ptr<float2>(0), lines.ptr<int>(1), maxLines, rho, theta, threshold, doSort);
-    if (linesCount > 0)
-        lines.cols = linesCount;
-    else
-        lines.release();
-}
-
-void cv::gpu::HoughLinesDownload(const GpuMat& d_lines, OutputArray h_lines_, OutputArray h_votes_)
-{
-    if (d_lines.empty())
-    {
-        h_lines_.release();
-        if (h_votes_.needed())
-            h_votes_.release();
-        return;
-    }
-
-    CV_Assert(d_lines.rows == 2 && d_lines.type() == CV_32FC2);
-
-    h_lines_.create(1, d_lines.cols, CV_32FC2);
-    Mat h_lines = h_lines_.getMat();
-    d_lines.row(0).download(h_lines);
-
-    if (h_votes_.needed())
-    {
-        h_votes_.create(1, d_lines.cols, CV_32SC1);
-        Mat h_votes = h_votes_.getMat();
-        GpuMat d_votes(1, d_lines.cols, CV_32SC1, const_cast<int*>(d_lines.ptr<int>(1)));
-        d_votes.download(h_votes);
-    }
-}
-
-//////////////////////////////////////////////////////////
-// HoughLinesP
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        int houghLinesProbabilistic_gpu(PtrStepSzb mask, PtrStepSzi accum, int4* out, int maxSize, float rho, float theta, int lineGap, int lineLength);
-    }
-}}}
-
-void cv::gpu::HoughLinesP(const GpuMat& src, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int minLineLength, int maxLineGap, int maxLines)
-{
-    using namespace cv::gpu::device::hough;
-
-    CV_Assert( src.type() == CV_8UC1 );
-    CV_Assert( src.cols < std::numeric_limits<unsigned short>::max() );
-    CV_Assert( src.rows < std::numeric_limits<unsigned short>::max() );
-
-    ensureSizeIsEnough(1, src.size().area(), CV_32SC1, buf.list);
-    unsigned int* srcPoints = buf.list.ptr<unsigned int>();
-
-    const int pointsCount = buildPointList_gpu(src, srcPoints);
-    if (pointsCount == 0)
-    {
-        lines.release();
-        return;
-    }
-
-    const int numangle = cvRound(CV_PI / theta);
-    const int numrho = cvRound(((src.cols + src.rows) * 2 + 1) / rho);
-    CV_Assert( numangle > 0 && numrho > 0 );
-
-    ensureSizeIsEnough(numangle + 2, numrho + 2, CV_32SC1, buf.accum);
-    buf.accum.setTo(Scalar::all(0));
-
-    DeviceInfo devInfo;
-    linesAccum_gpu(srcPoints, pointsCount, buf.accum, rho, theta, devInfo.sharedMemPerBlock(), devInfo.supports(FEATURE_SET_COMPUTE_20));
-
-    ensureSizeIsEnough(1, maxLines, CV_32SC4, lines);
-
-    int linesCount = houghLinesProbabilistic_gpu(src, buf.accum, lines.ptr<int4>(), maxLines, rho, theta, maxLineGap, minLineLength);
-
-    if (linesCount > 0)
-        lines.cols = linesCount;
-    else
-        lines.release();
-}
-
-//////////////////////////////////////////////////////////
-// HoughCircles
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        void circlesAccumCenters_gpu(const unsigned int* list, int count, PtrStepi dx, PtrStepi dy, PtrStepSzi accum, int minRadius, int maxRadius, float idp);
-        int buildCentersList_gpu(PtrStepSzi accum, unsigned int* centers, int threshold);
-        int circlesAccumRadius_gpu(const unsigned int* centers, int centersCount, const unsigned int* list, int count,
-                                   float3* circles, int maxCircles, float dp, int minRadius, int maxRadius, int threshold, bool has20);
-    }
-}}}
-
-void cv::gpu::HoughCircles(const GpuMat& src, GpuMat& circles, int method, float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles)
-{
-    HoughCirclesBuf buf;
-    HoughCircles(src, circles, buf, method, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius, maxCircles);
-}
-
-void cv::gpu::HoughCircles(const GpuMat& src, GpuMat& circles, HoughCirclesBuf& buf, int method,
-                           float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles)
-{
-    using namespace cv::gpu::device::hough;
-
-    CV_Assert(src.type() == CV_8UC1);
-    CV_Assert(src.cols < std::numeric_limits<unsigned short>::max());
-    CV_Assert(src.rows < std::numeric_limits<unsigned short>::max());
-    CV_Assert(method == CV_HOUGH_GRADIENT);
-    CV_Assert(dp > 0);
-    CV_Assert(minRadius > 0 && maxRadius > minRadius);
-    CV_Assert(cannyThreshold > 0);
-    CV_Assert(votesThreshold > 0);
-    CV_Assert(maxCircles > 0);
-
-    const float idp = 1.0f / dp;
-
-    cv::gpu::Canny(src, buf.cannyBuf, buf.edges, std::max(cannyThreshold / 2, 1), cannyThreshold);
-
-    ensureSizeIsEnough(2, src.size().area(), CV_32SC1, buf.list);
-    unsigned int* srcPoints = buf.list.ptr<unsigned int>(0);
-    unsigned int* centers = buf.list.ptr<unsigned int>(1);
-
-    const int pointsCount = buildPointList_gpu(buf.edges, srcPoints);
-    if (pointsCount == 0)
-    {
-        circles.release();
-        return;
-    }
-
-    ensureSizeIsEnough(cvCeil(src.rows * idp) + 2, cvCeil(src.cols * idp) + 2, CV_32SC1, buf.accum);
-    buf.accum.setTo(Scalar::all(0));
-
-    circlesAccumCenters_gpu(srcPoints, pointsCount, buf.cannyBuf.dx, buf.cannyBuf.dy, buf.accum, minRadius, maxRadius, idp);
-
-    int centersCount = buildCentersList_gpu(buf.accum, centers, votesThreshold);
-    if (centersCount == 0)
-    {
-        circles.release();
-        return;
-    }
-
-    if (minDist > 1)
-    {
-        cv::AutoBuffer<ushort2> oldBuf_(centersCount);
-        cv::AutoBuffer<ushort2> newBuf_(centersCount);
-        int newCount = 0;
-
-        ushort2* oldBuf = oldBuf_;
-        ushort2* newBuf = newBuf_;
-
-        cudaSafeCall( cudaMemcpy(oldBuf, centers, centersCount * sizeof(ushort2), cudaMemcpyDeviceToHost) );
-
-        const int cellSize = cvRound(minDist);
-        const int gridWidth = (src.cols + cellSize - 1) / cellSize;
-        const int gridHeight = (src.rows + cellSize - 1) / cellSize;
-
-        std::vector< std::vector<ushort2> > grid(gridWidth * gridHeight);
-
-        const float minDist2 = minDist * minDist;
-
-        for (int i = 0; i < centersCount; ++i)
-        {
-            ushort2 p = oldBuf[i];
-
-            bool good = true;
-
-            int xCell = static_cast<int>(p.x / cellSize);
-            int yCell = static_cast<int>(p.y / cellSize);
-
-            int x1 = xCell - 1;
-            int y1 = yCell - 1;
-            int x2 = xCell + 1;
-            int y2 = yCell + 1;
-
-            // boundary check
-            x1 = std::max(0, x1);
-            y1 = std::max(0, y1);
-            x2 = std::min(gridWidth - 1, x2);
-            y2 = std::min(gridHeight - 1, y2);
-
-            for (int yy = y1; yy <= y2; ++yy)
-            {
-                for (int xx = x1; xx <= x2; ++xx)
-                {
-                    std::vector<ushort2>& m = grid[yy * gridWidth + xx];
-
-                    for(size_t j = 0; j < m.size(); ++j)
-                    {
-                        float dx = (float)(p.x - m[j].x);
-                        float dy = (float)(p.y - m[j].y);
-
-                        if (dx * dx + dy * dy < minDist2)
-                        {
-                            good = false;
-                            goto break_out;
-                        }
-                    }
-                }
-            }
-
-            break_out:
-
-            if(good)
-            {
-                grid[yCell * gridWidth + xCell].push_back(p);
-
-                newBuf[newCount++] = p;
-            }
-        }
-
-        cudaSafeCall( cudaMemcpy(centers, newBuf, newCount * sizeof(unsigned int), cudaMemcpyHostToDevice) );
-        centersCount = newCount;
-    }
-
-    ensureSizeIsEnough(1, maxCircles, CV_32FC3, circles);
-
-    const int circlesCount = circlesAccumRadius_gpu(centers, centersCount, srcPoints, pointsCount, circles.ptr<float3>(), maxCircles,
-                                                    dp, minRadius, maxRadius, votesThreshold, deviceSupports(FEATURE_SET_COMPUTE_20));
-
-    if (circlesCount > 0)
-        circles.cols = circlesCount;
-    else
-        circles.release();
-}
-
-void cv::gpu::HoughCirclesDownload(const GpuMat& d_circles, cv::OutputArray h_circles_)
-{
-    if (d_circles.empty())
-    {
-        h_circles_.release();
-        return;
-    }
-
-    CV_Assert(d_circles.rows == 1 && d_circles.type() == CV_32FC3);
-
-    h_circles_.create(1, d_circles.cols, CV_32FC3);
-    Mat h_circles = h_circles_.getMat();
-    d_circles.download(h_circles);
-}
-
-//////////////////////////////////////////////////////////
-// GeneralizedHough
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace hough
-    {
-        template <typename T>
-        int buildEdgePointList_gpu(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
-        void buildRTable_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                             PtrStepSz<short2> r_table, int* r_sizes,
-                             short2 templCenter, int levels);
-
-        void GHT_Ballard_Pos_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                          PtrStepSz<short2> r_table, const int* r_sizes,
-                                          PtrStepSzi hist,
-                                          float dp, int levels);
-        int GHT_Ballard_Pos_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int maxSize, float dp, int threshold);
-
-        void GHT_Ballard_PosScale_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                               PtrStepSz<short2> r_table, const int* r_sizes,
-                                               PtrStepi hist, int rows, int cols,
-                                               float minScale, float scaleStep, int scaleRange,
-                                               float dp, int levels);
-        int GHT_Ballard_PosScale_findPosInHist_gpu(PtrStepi hist, int rows, int cols, int scaleRange, float4* out, int3* votes, int maxSize,
-                                                   float minScale, float scaleStep, float dp, int threshold);
-
-        void GHT_Ballard_PosRotation_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                  PtrStepSz<short2> r_table, const int* r_sizes,
-                                                  PtrStepi hist, int rows, int cols,
-                                                  float minAngle, float angleStep, int angleRange,
-                                                  float dp, int levels);
-        int GHT_Ballard_PosRotation_findPosInHist_gpu(PtrStepi hist, int rows, int cols, int angleRange, float4* out, int3* votes, int maxSize,
-                                                      float minAngle, float angleStep, float dp, int threshold);
-
-        void GHT_Guil_Full_setTemplFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
-        void GHT_Guil_Full_setImageFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
-        void GHT_Guil_Full_buildTemplFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                     int* sizes, int maxSize,
-                                                     float xi, float angleEpsilon, int levels,
-                                                     float2 center, float maxDist);
-        void GHT_Guil_Full_buildImageFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                                     int* sizes, int maxSize,
-                                                     float xi, float angleEpsilon, int levels,
-                                                     float2 center, float maxDist);
-        void GHT_Guil_Full_calcOHist_gpu(const int* templSizes, const int* imageSizes, int* OHist,
-                                         float minAngle, float maxAngle, float angleStep, int angleRange,
-                                         int levels, int tMaxSize);
-        void GHT_Guil_Full_calcSHist_gpu(const int* templSizes, const int* imageSizes, int* SHist,
-                                         float angle, float angleEpsilon,
-                                         float minScale, float maxScale, float iScaleStep, int scaleRange,
-                                         int levels, int tMaxSize);
-        void GHT_Guil_Full_calcPHist_gpu(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
-                                         float angle, float angleEpsilon, float scale,
-                                         float dp,
-                                         int levels, int tMaxSize);
-        int GHT_Guil_Full_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int curSize, int maxSize,
-                                             float angle, int angleVotes, float scale, int scaleVotes,
-                                             float dp, int threshold);
-    }
-}}}
-
-namespace
-{
-    /////////////////////////////////////
-    // Common
-
-    template <typename T, class A> void releaseVector(std::vector<T, A>& v)
-    {
-        std::vector<T, A> empty;
-        empty.swap(v);
-    }
-
-    class GHT_Pos : public GeneralizedHough_GPU
-    {
-    public:
-        GHT_Pos();
-
-    protected:
-        void setTemplateImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Point templCenter);
-        void detectImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, GpuMat& positions);
-        void releaseImpl();
-
-        virtual void processTempl() = 0;
-        virtual void processImage() = 0;
-
-        void buildEdgePointList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy);
-        void filterMinDist();
-        void convertTo(GpuMat& positions);
-
-        int maxSize;
-        double minDist;
-
-        Size templSize;
-        Point templCenter;
-        GpuMat templEdges;
-        GpuMat templDx;
-        GpuMat templDy;
-
-        Size imageSize;
-        GpuMat imageEdges;
-        GpuMat imageDx;
-        GpuMat imageDy;
-
-        GpuMat edgePointList;
-
-        GpuMat outBuf;
-        int posCount;
-
-        std::vector<float4> oldPosBuf;
-        std::vector<int3> oldVoteBuf;
-        std::vector<float4> newPosBuf;
-        std::vector<int3> newVoteBuf;
-        std::vector<int> indexies;
-    };
-
-    GHT_Pos::GHT_Pos()
-    {
-        maxSize = 10000;
-        minDist = 1.0;
-    }
-
-    void GHT_Pos::setTemplateImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Point templCenter_)
-    {
-        templSize = edges.size();
-        templCenter = templCenter_;
-
-        ensureSizeIsEnough(templSize, edges.type(), templEdges);
-        ensureSizeIsEnough(templSize, dx.type(), templDx);
-        ensureSizeIsEnough(templSize, dy.type(), templDy);
-
-        edges.copyTo(templEdges);
-        dx.copyTo(templDx);
-        dy.copyTo(templDy);
-
-        processTempl();
-    }
-
-    void GHT_Pos::detectImpl(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, GpuMat& positions)
-    {
-        imageSize = edges.size();
-
-        ensureSizeIsEnough(imageSize, edges.type(), imageEdges);
-        ensureSizeIsEnough(imageSize, dx.type(), imageDx);
-        ensureSizeIsEnough(imageSize, dy.type(), imageDy);
-
-        edges.copyTo(imageEdges);
-        dx.copyTo(imageDx);
-        dy.copyTo(imageDy);
-
-        posCount = 0;
-
-        processImage();
-
-        if (posCount == 0)
-            positions.release();
-        else
-        {
-            if (minDist > 1)
-                filterMinDist();
-            convertTo(positions);
-        }
-    }
-
-    void GHT_Pos::releaseImpl()
-    {
-        templSize = Size();
-        templCenter = Point(-1, -1);
-        templEdges.release();
-        templDx.release();
-        templDy.release();
-
-        imageSize = Size();
-        imageEdges.release();
-        imageDx.release();
-        imageDy.release();
-
-        edgePointList.release();
-
-        outBuf.release();
-        posCount = 0;
-
-        releaseVector(oldPosBuf);
-        releaseVector(oldVoteBuf);
-        releaseVector(newPosBuf);
-        releaseVector(newVoteBuf);
-        releaseVector(indexies);
-    }
-
-    void GHT_Pos::buildEdgePointList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy)
-    {
-        using namespace cv::gpu::device::hough;
-
-        typedef int (*func_t)(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
-        static const func_t funcs[] =
-        {
-            0,
-            0,
-            0,
-            buildEdgePointList_gpu<short>,
-            buildEdgePointList_gpu<int>,
-            buildEdgePointList_gpu<float>,
-            0
-        };
-
-        CV_Assert(edges.type() == CV_8UC1);
-        CV_Assert(dx.size() == edges.size());
-        CV_Assert(dy.type() == dx.type() && dy.size() == edges.size());
-
-        const func_t func = funcs[dx.depth()];
-        CV_Assert(func != 0);
-
-        edgePointList.cols = (int) (edgePointList.step / sizeof(int));
-        ensureSizeIsEnough(2, edges.size().area(), CV_32SC1, edgePointList);
-
-        edgePointList.cols = func(edges, dx, dy, edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1));
-    }
-
-    #define votes_cmp_gt(l1, l2) (aux[l1].x > aux[l2].x)
-    static CV_IMPLEMENT_QSORT_EX( sortIndexies, int, votes_cmp_gt, const int3* )
-
-    void GHT_Pos::filterMinDist()
-    {
-        oldPosBuf.resize(posCount);
-        oldVoteBuf.resize(posCount);
-
-        cudaSafeCall( cudaMemcpy(&oldPosBuf[0], outBuf.ptr(0), posCount * sizeof(float4), cudaMemcpyDeviceToHost) );
-        cudaSafeCall( cudaMemcpy(&oldVoteBuf[0], outBuf.ptr(1), posCount * sizeof(int3), cudaMemcpyDeviceToHost) );
-
-        indexies.resize(posCount);
-        for (int i = 0; i < posCount; ++i)
-            indexies[i] = i;
-        sortIndexies(&indexies[0], posCount, &oldVoteBuf[0]);
-
-        newPosBuf.clear();
-        newVoteBuf.clear();
-        newPosBuf.reserve(posCount);
-        newVoteBuf.reserve(posCount);
-
-        const int cellSize = cvRound(minDist);
-        const int gridWidth = (imageSize.width + cellSize - 1) / cellSize;
-        const int gridHeight = (imageSize.height + cellSize - 1) / cellSize;
-
-        std::vector< std::vector<Point2f> > grid(gridWidth * gridHeight);
-
-        const double minDist2 = minDist * minDist;
-
-        for (int i = 0; i < posCount; ++i)
-        {
-            const int ind = indexies[i];
-
-            Point2f p(oldPosBuf[ind].x, oldPosBuf[ind].y);
-
-            bool good = true;
-
-            const int xCell = static_cast<int>(p.x / cellSize);
-            const int yCell = static_cast<int>(p.y / cellSize);
-
-            int x1 = xCell - 1;
-            int y1 = yCell - 1;
-            int x2 = xCell + 1;
-            int y2 = yCell + 1;
-
-            // boundary check
-            x1 = std::max(0, x1);
-            y1 = std::max(0, y1);
-            x2 = std::min(gridWidth - 1, x2);
-            y2 = std::min(gridHeight - 1, y2);
-
-            for (int yy = y1; yy <= y2; ++yy)
-            {
-                for (int xx = x1; xx <= x2; ++xx)
-                {
-                    const std::vector<Point2f>& m = grid[yy * gridWidth + xx];
-
-                    for(size_t j = 0; j < m.size(); ++j)
-                    {
-                        const Point2f d = p - m[j];
-
-                        if (d.ddot(d) < minDist2)
-                        {
-                            good = false;
-                            goto break_out;
-                        }
-                    }
-                }
-            }
-
-            break_out:
-
-            if(good)
-            {
-                grid[yCell * gridWidth + xCell].push_back(p);
-
-                newPosBuf.push_back(oldPosBuf[ind]);
-                newVoteBuf.push_back(oldVoteBuf[ind]);
-            }
-        }
-
-        posCount = static_cast<int>(newPosBuf.size());
-        cudaSafeCall( cudaMemcpy(outBuf.ptr(0), &newPosBuf[0], posCount * sizeof(float4), cudaMemcpyHostToDevice) );
-        cudaSafeCall( cudaMemcpy(outBuf.ptr(1), &newVoteBuf[0], posCount * sizeof(int3), cudaMemcpyHostToDevice) );
-    }
-
-    void GHT_Pos::convertTo(GpuMat& positions)
-    {
-        ensureSizeIsEnough(2, posCount, CV_32FC4, positions);
-        GpuMat(2, posCount, CV_32FC4, outBuf.data, outBuf.step).copyTo(positions);
-    }
-
-    /////////////////////////////////////
-    // POSITION Ballard
-
-    class GHT_Ballard_Pos : public GHT_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Ballard_Pos();
-
-    protected:
-        void releaseImpl();
-
-        void processTempl();
-        void processImage();
-
-        virtual void calcHist();
-        virtual void findPosInHist();
-
-        int levels;
-        int votesThreshold;
-        double dp;
-
-        GpuMat r_table;
-        GpuMat r_sizes;
-
-        GpuMat hist;
-    };
-
-    CV_INIT_ALGORITHM(GHT_Ballard_Pos, "GeneralizedHough_GPU.POSITION",
-                      obj.info()->addParam(obj, "maxSize", obj.maxSize, false, 0, 0,
-                                           "Maximal size of inner buffers.");
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution."));
-
-    GHT_Ballard_Pos::GHT_Ballard_Pos()
-    {
-        levels = 360;
-        votesThreshold = 100;
-        dp = 1.0;
-    }
-
-    void GHT_Ballard_Pos::releaseImpl()
-    {
-        GHT_Pos::releaseImpl();
-
-        r_table.release();
-        r_sizes.release();
-
-        hist.release();
-    }
-
-    void GHT_Ballard_Pos::processTempl()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(levels > 0);
-
-        buildEdgePointList(templEdges, templDx, templDy);
-
-        ensureSizeIsEnough(levels + 1, maxSize, CV_16SC2, r_table);
-        ensureSizeIsEnough(1, levels + 1, CV_32SC1, r_sizes);
-        r_sizes.setTo(Scalar::all(0));
-
-        if (edgePointList.cols > 0)
-        {
-            buildRTable_gpu(edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1), edgePointList.cols,
-                            r_table, r_sizes.ptr<int>(), make_short2(templCenter.x, templCenter.y), levels);
-            min(r_sizes, maxSize, r_sizes);
-        }
-    }
-
-    void GHT_Ballard_Pos::processImage()
-    {
-        calcHist();
-        findPosInHist();
-    }
-
-    void GHT_Ballard_Pos::calcHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(levels > 0 && r_table.rows == (levels + 1) && r_sizes.cols == (levels + 1));
-        CV_Assert(dp > 0.0);
-
-        const double idp = 1.0 / dp;
-
-        buildEdgePointList(imageEdges, imageDx, imageDy);
-
-        ensureSizeIsEnough(cvCeil(imageSize.height * idp) + 2, cvCeil(imageSize.width * idp) + 2, CV_32SC1, hist);
-        hist.setTo(Scalar::all(0));
-
-        if (edgePointList.cols > 0)
-        {
-            GHT_Ballard_Pos_calcHist_gpu(edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1), edgePointList.cols,
-                                         r_table, r_sizes.ptr<int>(),
-                                         hist,
-                                         (float)dp, levels);
-        }
-    }
-
-    void GHT_Ballard_Pos::findPosInHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(votesThreshold > 0);
-
-        ensureSizeIsEnough(2, maxSize, CV_32FC4, outBuf);
-
-        posCount = GHT_Ballard_Pos_findPosInHist_gpu(hist, outBuf.ptr<float4>(0), outBuf.ptr<int3>(1), maxSize, (float)dp, votesThreshold);
-    }
-
-    /////////////////////////////////////
-    // POSITION & SCALE
-
-    class GHT_Ballard_PosScale : public GHT_Ballard_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Ballard_PosScale();
-
-    protected:
-        void calcHist();
-        void findPosInHist();
-
-        double minScale;
-        double maxScale;
-        double scaleStep;
-    };
-
-    CV_INIT_ALGORITHM(GHT_Ballard_PosScale, "GeneralizedHough_GPU.POSITION_SCALE",
-                      obj.info()->addParam(obj, "maxSize", obj.maxSize, false, 0, 0,
-                                           "Maximal size of inner buffers.");
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "minScale", obj.minScale, false, 0, 0,
-                                           "Minimal scale to detect.");
-                      obj.info()->addParam(obj, "maxScale", obj.maxScale, false, 0, 0,
-                                           "Maximal scale to detect.");
-                      obj.info()->addParam(obj, "scaleStep", obj.scaleStep, false, 0, 0,
-                                           "Scale step."));
-
-    GHT_Ballard_PosScale::GHT_Ballard_PosScale()
-    {
-        minScale = 0.5;
-        maxScale = 2.0;
-        scaleStep = 0.05;
-    }
-
-    void GHT_Ballard_PosScale::calcHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(levels > 0 && r_table.rows == (levels + 1) && r_sizes.cols == (levels + 1));
-        CV_Assert(dp > 0.0);
-        CV_Assert(minScale > 0.0 && minScale < maxScale);
-        CV_Assert(scaleStep > 0.0);
-
-        const double idp = 1.0 / dp;
-        const int scaleRange = cvCeil((maxScale - minScale) / scaleStep);
-        const int rows = cvCeil(imageSize.height * idp);
-        const int cols = cvCeil(imageSize.width * idp);
-
-        buildEdgePointList(imageEdges, imageDx, imageDy);
-
-        ensureSizeIsEnough((scaleRange + 2) * (rows + 2), cols + 2, CV_32SC1, hist);
-        hist.setTo(Scalar::all(0));
-
-        if (edgePointList.cols > 0)
-        {
-            GHT_Ballard_PosScale_calcHist_gpu(edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1), edgePointList.cols,
-                                              r_table, r_sizes.ptr<int>(),
-                                              hist, rows, cols,
-                                              (float)minScale, (float)scaleStep, scaleRange, (float)dp, levels);
-        }
-    }
-
-    void GHT_Ballard_PosScale::findPosInHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(votesThreshold > 0);
-
-        const double idp = 1.0 / dp;
-        const int scaleRange = cvCeil((maxScale - minScale) / scaleStep);
-        const int rows = cvCeil(imageSize.height * idp);
-        const int cols = cvCeil(imageSize.width * idp);
-
-        ensureSizeIsEnough(2, maxSize, CV_32FC4, outBuf);
-
-        posCount =  GHT_Ballard_PosScale_findPosInHist_gpu(hist, rows, cols, scaleRange, outBuf.ptr<float4>(0), outBuf.ptr<int3>(1), maxSize, (float)minScale, (float)scaleStep, (float)dp, votesThreshold);
-    }
-
-    /////////////////////////////////////
-    // POSITION & Rotation
-
-    class GHT_Ballard_PosRotation : public GHT_Ballard_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Ballard_PosRotation();
-
-    protected:
-        void calcHist();
-        void findPosInHist();
-
-        double minAngle;
-        double maxAngle;
-        double angleStep;
-    };
-
-    CV_INIT_ALGORITHM(GHT_Ballard_PosRotation, "GeneralizedHough_GPU.POSITION_ROTATION",
-                      obj.info()->addParam(obj, "maxSize", obj.maxSize, false, 0, 0,
-                                           "Maximal size of inner buffers.");
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "minAngle", obj.minAngle, false, 0, 0,
-                                           "Minimal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "maxAngle", obj.maxAngle, false, 0, 0,
-                                           "Maximal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "angleStep", obj.angleStep, false, 0, 0,
-                                           "Angle step in degrees."));
-
-    GHT_Ballard_PosRotation::GHT_Ballard_PosRotation()
-    {
-        minAngle = 0.0;
-        maxAngle = 360.0;
-        angleStep = 1.0;
-    }
-
-    void GHT_Ballard_PosRotation::calcHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(levels > 0 && r_table.rows == (levels + 1) && r_sizes.cols == (levels + 1));
-        CV_Assert(dp > 0.0);
-        CV_Assert(minAngle >= 0.0 && minAngle < maxAngle && maxAngle <= 360.0);
-        CV_Assert(angleStep > 0.0 && angleStep < 360.0);
-
-        const double idp = 1.0 / dp;
-        const int angleRange = cvCeil((maxAngle - minAngle) / angleStep);
-        const int rows = cvCeil(imageSize.height * idp);
-        const int cols = cvCeil(imageSize.width * idp);
-
-        buildEdgePointList(imageEdges, imageDx, imageDy);
-
-        ensureSizeIsEnough((angleRange + 2) * (rows + 2), cols + 2, CV_32SC1, hist);
-        hist.setTo(Scalar::all(0));
-
-        if (edgePointList.cols > 0)
-        {
-            GHT_Ballard_PosRotation_calcHist_gpu(edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1), edgePointList.cols,
-                                                 r_table, r_sizes.ptr<int>(),
-                                                 hist, rows, cols,
-                                                 (float)minAngle, (float)angleStep, angleRange, (float)dp, levels);
-        }
-    }
-
-    void GHT_Ballard_PosRotation::findPosInHist()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(votesThreshold > 0);
-
-        const double idp = 1.0 / dp;
-        const int angleRange = cvCeil((maxAngle - minAngle) / angleStep);
-        const int rows = cvCeil(imageSize.height * idp);
-        const int cols = cvCeil(imageSize.width * idp);
-
-        ensureSizeIsEnough(2, maxSize, CV_32FC4, outBuf);
-
-        posCount = GHT_Ballard_PosRotation_findPosInHist_gpu(hist, rows, cols, angleRange, outBuf.ptr<float4>(0), outBuf.ptr<int3>(1), maxSize, (float)minAngle, (float)angleStep, (float)dp, votesThreshold);
-    }
-
-    /////////////////////////////////////////
-    // POSITION & SCALE & ROTATION
-
-    double toRad(double a)
-    {
-        return a * CV_PI / 180.0;
-    }
-
-    double clampAngle(double a)
-    {
-        double res = a;
-
-        while (res > 360.0)
-            res -= 360.0;
-        while (res < 0)
-            res += 360.0;
-
-        return res;
-    }
-
-    bool angleEq(double a, double b, double eps = 1.0)
-    {
-        return (fabs(clampAngle(a - b)) <= eps);
-    }
-
-    class GHT_Guil_Full : public GHT_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Guil_Full();
-
-    protected:
-        void releaseImpl();
-
-        void processTempl();
-        void processImage();
-
-        struct Feature
-        {
-            GpuMat p1_pos;
-            GpuMat p1_theta;
-            GpuMat p2_pos;
-
-            GpuMat d12;
-
-            GpuMat r1;
-            GpuMat r2;
-
-            GpuMat sizes;
-            int maxSize;
-
-            void create(int levels, int maxCapacity, bool isTempl);
-            void release();
-        };
-
-        typedef void (*set_func_t)(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
-        typedef void (*build_func_t)(const unsigned int* coordList, const float* thetaList, int pointsCount,
-                                     int* sizes, int maxSize,
-                                     float xi, float angleEpsilon, int levels,
-                                     float2 center, float maxDist);
-
-        void buildFeatureList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Feature& features,
-                              set_func_t set_func, build_func_t build_func, bool isTempl, Point2d center = Point2d());
-
-        void calcOrientation();
-        void calcScale(double angle);
-        void calcPosition(double angle, int angleVotes, double scale, int scaleVotes);
-
-        double xi;
-        int levels;
-        double angleEpsilon;
-
-        double minAngle;
-        double maxAngle;
-        double angleStep;
-        int angleThresh;
-
-        double minScale;
-        double maxScale;
-        double scaleStep;
-        int scaleThresh;
-
-        double dp;
-        int posThresh;
-
-        Feature templFeatures;
-        Feature imageFeatures;
-
-        std::vector< std::pair<double, int> > angles;
-        std::vector< std::pair<double, int> > scales;
-
-        GpuMat hist;
-        std::vector<int> h_buf;
-    };
-
-    CV_INIT_ALGORITHM(GHT_Guil_Full, "GeneralizedHough_GPU.POSITION_SCALE_ROTATION",
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "maxSize", obj.maxSize, false, 0, 0,
-                                           "Maximal size of inner buffers.");
-                      obj.info()->addParam(obj, "xi", obj.xi, false, 0, 0,
-                                           "Angle difference in degrees between two points in feature.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "Feature table levels.");
-                      obj.info()->addParam(obj, "angleEpsilon", obj.angleEpsilon, false, 0, 0,
-                                           "Maximal difference between angles that treated as equal.");
-                      obj.info()->addParam(obj, "minAngle", obj.minAngle, false, 0, 0,
-                                           "Minimal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "maxAngle", obj.maxAngle, false, 0, 0,
-                                           "Maximal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "angleStep", obj.angleStep, false, 0, 0,
-                                           "Angle step in degrees.");
-                      obj.info()->addParam(obj, "angleThresh", obj.angleThresh, false, 0, 0,
-                                           "Angle threshold.");
-                      obj.info()->addParam(obj, "minScale", obj.minScale, false, 0, 0,
-                                           "Minimal scale to detect.");
-                      obj.info()->addParam(obj, "maxScale", obj.maxScale, false, 0, 0,
-                                           "Maximal scale to detect.");
-                      obj.info()->addParam(obj, "scaleStep", obj.scaleStep, false, 0, 0,
-                                           "Scale step.");
-                      obj.info()->addParam(obj, "scaleThresh", obj.scaleThresh, false, 0, 0,
-                                           "Scale threshold.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "posThresh", obj.posThresh, false, 0, 0,
-                                           "Position threshold."));
-
-    GHT_Guil_Full::GHT_Guil_Full()
-    {
-        maxSize = 1000;
-        xi = 90.0;
-        levels = 360;
-        angleEpsilon = 1.0;
-
-        minAngle = 0.0;
-        maxAngle = 360.0;
-        angleStep = 1.0;
-        angleThresh = 15000;
-
-        minScale = 0.5;
-        maxScale = 2.0;
-        scaleStep = 0.05;
-        scaleThresh = 1000;
-
-        dp = 1.0;
-        posThresh = 100;
-    }
-
-    void GHT_Guil_Full::releaseImpl()
-    {
-        GHT_Pos::releaseImpl();
-
-        templFeatures.release();
-        imageFeatures.release();
-
-        releaseVector(angles);
-        releaseVector(scales);
-
-        hist.release();
-        releaseVector(h_buf);
-    }
-
-    void GHT_Guil_Full::processTempl()
-    {
-        using namespace cv::gpu::device::hough;
-
-        buildFeatureList(templEdges, templDx, templDy, templFeatures,
-            GHT_Guil_Full_setTemplFeatures, GHT_Guil_Full_buildTemplFeatureList_gpu,
-            true, templCenter);
-
-        h_buf.resize(templFeatures.sizes.cols);
-        cudaSafeCall( cudaMemcpy(&h_buf[0], templFeatures.sizes.data, h_buf.size() * sizeof(int), cudaMemcpyDeviceToHost) );
-        templFeatures.maxSize = *max_element(h_buf.begin(), h_buf.end());
-    }
-
-    void GHT_Guil_Full::processImage()
-    {
-        using namespace cv::gpu::device::hough;
-
-        CV_Assert(levels > 0);
-        CV_Assert(templFeatures.sizes.cols == levels + 1);
-        CV_Assert(minAngle >= 0.0 && minAngle < maxAngle && maxAngle <= 360.0);
-        CV_Assert(angleStep > 0.0 && angleStep < 360.0);
-        CV_Assert(angleThresh > 0);
-        CV_Assert(minScale > 0.0 && minScale < maxScale);
-        CV_Assert(scaleStep > 0.0);
-        CV_Assert(scaleThresh > 0);
-        CV_Assert(dp > 0.0);
-        CV_Assert(posThresh > 0);
-
-        const double iAngleStep = 1.0 / angleStep;
-        const int angleRange = cvCeil((maxAngle - minAngle) * iAngleStep);
-
-        const double iScaleStep = 1.0 / scaleStep;
-        const int scaleRange = cvCeil((maxScale - minScale) * iScaleStep);
-
-        const double idp = 1.0 / dp;
-        const int histRows = cvCeil(imageSize.height * idp);
-        const int histCols = cvCeil(imageSize.width * idp);
-
-        ensureSizeIsEnough(histRows + 2, std::max(angleRange + 1, std::max(scaleRange + 1, histCols + 2)), CV_32SC1, hist);
-        h_buf.resize(std::max(angleRange + 1, scaleRange + 1));
-
-        ensureSizeIsEnough(2, maxSize, CV_32FC4, outBuf);
-
-        buildFeatureList(imageEdges, imageDx, imageDy, imageFeatures,
-            GHT_Guil_Full_setImageFeatures, GHT_Guil_Full_buildImageFeatureList_gpu,
-            false);
-
-        calcOrientation();
-
-        for (size_t i = 0; i < angles.size(); ++i)
-        {
-            const double angle = angles[i].first;
-            const int angleVotes = angles[i].second;
-
-            calcScale(angle);
-
-            for (size_t j = 0; j < scales.size(); ++j)
-            {
-                const double scale = scales[j].first;
-                const int scaleVotes = scales[j].second;
-
-                calcPosition(angle, angleVotes, scale, scaleVotes);
-            }
-        }
-    }
-
-    void GHT_Guil_Full::Feature::create(int levels, int maxCapacity, bool isTempl)
-    {
-        if (!isTempl)
-        {
-            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, p1_pos);
-            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, p2_pos);
-        }
-
-        ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC1, p1_theta);
-
-        ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC1, d12);
-
-        if (isTempl)
-        {
-            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, r1);
-            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, r2);
-        }
-
-        ensureSizeIsEnough(1, levels + 1, CV_32SC1, sizes);
-        sizes.setTo(Scalar::all(0));
-
-        maxSize = 0;
-    }
-
-    void GHT_Guil_Full::Feature::release()
-    {
-        p1_pos.release();
-        p1_theta.release();
-        p2_pos.release();
-
-        d12.release();
-
-        r1.release();
-        r2.release();
-
-        sizes.release();
-
-        maxSize = 0;
-    }
-
-    void GHT_Guil_Full::buildFeatureList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Feature& features,
-                                         set_func_t set_func, build_func_t build_func, bool isTempl, Point2d center)
-    {
-        CV_Assert(levels > 0);
-
-        const double maxDist = sqrt((double) templSize.width * templSize.width + templSize.height * templSize.height) * maxScale;
-
-        features.create(levels, maxSize, isTempl);
-        set_func(features.p1_pos, features.p1_theta, features.p2_pos, features.d12, features.r1, features.r2);
-
-        buildEdgePointList(edges, dx, dy);
-
-        if (edgePointList.cols > 0)
-        {
-            build_func(edgePointList.ptr<unsigned int>(0), edgePointList.ptr<float>(1), edgePointList.cols,
-                features.sizes.ptr<int>(), maxSize, (float)xi, (float)angleEpsilon, levels, make_float2((float)center.x, (float)center.y), (float)maxDist);
-        }
-    }
-
-    void GHT_Guil_Full::calcOrientation()
-    {
-        using namespace cv::gpu::device::hough;
-
-        const double iAngleStep = 1.0 / angleStep;
-        const int angleRange = cvCeil((maxAngle - minAngle) * iAngleStep);
-
-        hist.setTo(Scalar::all(0));
-        GHT_Guil_Full_calcOHist_gpu(templFeatures.sizes.ptr<int>(), imageFeatures.sizes.ptr<int>(0),
-            hist.ptr<int>(), (float)minAngle, (float)maxAngle, (float)angleStep, angleRange, levels, templFeatures.maxSize);
-        cudaSafeCall( cudaMemcpy(&h_buf[0], hist.data, h_buf.size() * sizeof(int), cudaMemcpyDeviceToHost) );
-
-        angles.clear();
-
-        for (int n = 0; n < angleRange; ++n)
-        {
-            if (h_buf[n] >= angleThresh)
-            {
-                const double angle = minAngle + n * angleStep;
-                angles.push_back(std::make_pair(angle, h_buf[n]));
-            }
-        }
-    }
-
-    void GHT_Guil_Full::calcScale(double angle)
-    {
-        using namespace cv::gpu::device::hough;
-
-        const double iScaleStep = 1.0 / scaleStep;
-        const int scaleRange = cvCeil((maxScale - minScale) * iScaleStep);
-
-        hist.setTo(Scalar::all(0));
-        GHT_Guil_Full_calcSHist_gpu(templFeatures.sizes.ptr<int>(), imageFeatures.sizes.ptr<int>(0),
-            hist.ptr<int>(), (float)angle, (float)angleEpsilon, (float)minScale, (float)maxScale, (float)iScaleStep, scaleRange, levels, templFeatures.maxSize);
-        cudaSafeCall( cudaMemcpy(&h_buf[0], hist.data, h_buf.size() * sizeof(int), cudaMemcpyDeviceToHost) );
-
-        scales.clear();
-
-        for (int s = 0; s < scaleRange; ++s)
-        {
-            if (h_buf[s] >= scaleThresh)
-            {
-                const double scale = minScale + s * scaleStep;
-                scales.push_back(std::make_pair(scale, h_buf[s]));
-            }
-        }
-    }
-
-    void GHT_Guil_Full::calcPosition(double angle, int angleVotes, double scale, int scaleVotes)
-    {
-        using namespace cv::gpu::device::hough;
-
-        hist.setTo(Scalar::all(0));
-        GHT_Guil_Full_calcPHist_gpu(templFeatures.sizes.ptr<int>(), imageFeatures.sizes.ptr<int>(0),
-            hist,(float) (float)angle, (float)angleEpsilon, (float)scale, (float)dp, levels, templFeatures.maxSize);
-
-        posCount = GHT_Guil_Full_findPosInHist_gpu(hist, outBuf.ptr<float4>(0), outBuf.ptr<int3>(1),
-            posCount, maxSize, (float)angle, angleVotes, (float)scale, scaleVotes, (float)dp, posThresh);
-    }
-}
-
-Ptr<GeneralizedHough_GPU> cv::gpu::GeneralizedHough_GPU::create(int method)
-{
-    switch (method)
-    {
-    case GHT_POSITION:
-        CV_Assert( !GHT_Ballard_Pos_info_auto.name().empty() );
-        return new GHT_Ballard_Pos();
-
-    case (GHT_POSITION | GHT_SCALE):
-        CV_Assert( !GHT_Ballard_PosScale_info_auto.name().empty() );
-        return new GHT_Ballard_PosScale();
-
-    case (GHT_POSITION | GHT_ROTATION):
-        CV_Assert( !GHT_Ballard_PosRotation_info_auto.name().empty() );
-        return new GHT_Ballard_PosRotation();
-
-    case (GHT_POSITION | GHT_SCALE | GHT_ROTATION):
-        CV_Assert( !GHT_Guil_Full_info_auto.name().empty() );
-        return new GHT_Guil_Full();
-    }
-
-    CV_Error(CV_StsBadArg, "Unsupported method");
-    return Ptr<GeneralizedHough_GPU>();
-}
-
-cv::gpu::GeneralizedHough_GPU::~GeneralizedHough_GPU()
-{
-}
-
-void cv::gpu::GeneralizedHough_GPU::setTemplate(const GpuMat& templ, int cannyThreshold, Point templCenter)
-{
-    CV_Assert(templ.type() == CV_8UC1);
-    CV_Assert(cannyThreshold > 0);
-
-    ensureSizeIsEnough(templ.size(), CV_8UC1, edges_);
-    Canny(templ, cannyBuf_, edges_, cannyThreshold / 2, cannyThreshold);
-
-    if (templCenter == Point(-1, -1))
-        templCenter = Point(templ.cols / 2, templ.rows / 2);
-
-    setTemplateImpl(edges_, cannyBuf_.dx, cannyBuf_.dy, templCenter);
-}
-
-void cv::gpu::GeneralizedHough_GPU::setTemplate(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Point templCenter)
-{
-    if (templCenter == Point(-1, -1))
-        templCenter = Point(edges.cols / 2, edges.rows / 2);
-
-    setTemplateImpl(edges, dx, dy, templCenter);
-}
-
-void cv::gpu::GeneralizedHough_GPU::detect(const GpuMat& image, GpuMat& positions, int cannyThreshold)
-{
-    CV_Assert(image.type() == CV_8UC1);
-    CV_Assert(cannyThreshold > 0);
-
-    ensureSizeIsEnough(image.size(), CV_8UC1, edges_);
-    Canny(image, cannyBuf_, edges_, cannyThreshold / 2, cannyThreshold);
-
-    detectImpl(edges_, cannyBuf_.dx, cannyBuf_.dy, positions);
-}
-
-void cv::gpu::GeneralizedHough_GPU::detect(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, GpuMat& positions)
-{
-    detectImpl(edges, dx, dy, positions);
-}
-
-void cv::gpu::GeneralizedHough_GPU::download(const GpuMat& d_positions, OutputArray h_positions_, OutputArray h_votes_)
-{
-    if (d_positions.empty())
-    {
-        h_positions_.release();
-        if (h_votes_.needed())
-            h_votes_.release();
-        return;
-    }
-
-    CV_Assert(d_positions.rows == 2 && d_positions.type() == CV_32FC4);
-
-    h_positions_.create(1, d_positions.cols, CV_32FC4);
-    Mat h_positions = h_positions_.getMat();
-    d_positions.row(0).download(h_positions);
-
-    if (h_votes_.needed())
-    {
-        h_votes_.create(1, d_positions.cols, CV_32SC3);
-        Mat h_votes = h_votes_.getMat();
-        GpuMat d_votes(1, d_positions.cols, CV_32SC3, const_cast<int3*>(d_positions.ptr<int3>(1)));
-        d_votes.download(h_votes);
-    }
-}
-
-void cv::gpu::GeneralizedHough_GPU::release()
-{
-    edges_.release();
-    cannyBuf_.release();
-    releaseImpl();
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/imgproc.cpp b/modules/gpu/src/imgproc.cpp
deleted file mode 100644
index 63342231d0..0000000000
--- a/modules/gpu/src/imgproc.cpp
+++ /dev/null
@@ -1,1545 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::meanShiftFiltering(const GpuMat&, GpuMat&, int, int, TermCriteria, Stream&) { throw_nogpu(); }
-void cv::gpu::meanShiftProc(const GpuMat&, GpuMat&, GpuMat&, int, int, TermCriteria, Stream&) { throw_nogpu(); }
-void cv::gpu::drawColorDisp(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::reprojectImageTo3D(const GpuMat&, GpuMat&, const Mat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::copyMakeBorder(const GpuMat&, GpuMat&, int, int, int, int, int, const Scalar&, Stream&) { throw_nogpu(); }
-void cv::gpu::buildWarpPlaneMaps(Size, Rect, const Mat&, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::buildWarpCylindricalMaps(Size, Rect, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::buildWarpSphericalMaps(Size, Rect, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::integral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::integralBuffered(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::sqrIntegral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::columnSum(const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::rectStdDev(const GpuMat&, const GpuMat&, GpuMat&, const Rect&, Stream&) { throw_nogpu(); }
-void cv::gpu::evenLevels(GpuMat&, int, int, int) { throw_nogpu(); }
-void cv::gpu::histEven(const GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::histEven(const GpuMat&, GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::histEven(const GpuMat&, GpuMat*, int*, int*, int*, Stream&) { throw_nogpu(); }
-void cv::gpu::histEven(const GpuMat&, GpuMat*, GpuMat&, int*, int*, int*, Stream&) { throw_nogpu(); }
-void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, Stream&) { throw_nogpu(); }
-void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::calcHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
-void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
-void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
-void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
-void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::mulSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, bool, Stream&) { throw_nogpu(); }
-void cv::gpu::mulAndScaleSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, float, bool, Stream&) { throw_nogpu(); }
-void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int, Stream&) { throw_nogpu(); }
-void cv::gpu::ConvolveBuf::create(Size, Size) { throw_nogpu(); }
-void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool) { throw_nogpu(); }
-void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&, Stream&) { throw_nogpu(); }
-void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
-void cv::gpu::Canny(const GpuMat&, CannyBuf&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
-void cv::gpu::Canny(const GpuMat&, const GpuMat&, GpuMat&, double, double, bool) { throw_nogpu(); }
-void cv::gpu::Canny(const GpuMat&, const GpuMat&, CannyBuf&, GpuMat&, double, double, bool) { throw_nogpu(); }
-void cv::gpu::CannyBuf::create(const Size&, int) { throw_nogpu(); }
-void cv::gpu::CannyBuf::release() { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-////////////////////////////////////////////////////////////////////////
-// meanShiftFiltering_GPU
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void meanShiftFiltering_gpu(const PtrStepSzb& src, PtrStepSzb dst, int sp, int sr, int maxIter, float eps, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria, Stream& stream)
-{
-    using namespace ::cv::gpu::device::imgproc;
-
-    if( src.empty() )
-        CV_Error( CV_StsBadArg, "The input image is empty" );
-
-    if( src.depth() != CV_8U || src.channels() != 4 )
-        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
-
-    dst.create( src.size(), CV_8UC4 );
-
-    if( !(criteria.type & TermCriteria::MAX_ITER) )
-        criteria.maxCount = 5;
-
-    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
-
-    float eps;
-    if( !(criteria.type & TermCriteria::EPS) )
-        eps = 1.f;
-    eps = (float)std::max(criteria.epsilon, 0.0);
-
-    meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// meanShiftProc_GPU
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void meanShiftProc_gpu(const PtrStepSzb& src, PtrStepSzb dstr, PtrStepSzb dstsp, int sp, int sr, int maxIter, float eps, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria, Stream& stream)
-{
-    using namespace ::cv::gpu::device::imgproc;
-
-    if( src.empty() )
-        CV_Error( CV_StsBadArg, "The input image is empty" );
-
-    if( src.depth() != CV_8U || src.channels() != 4 )
-        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
-
-    dstr.create( src.size(), CV_8UC4 );
-    dstsp.create( src.size(), CV_16SC2 );
-
-    if( !(criteria.type & TermCriteria::MAX_ITER) )
-        criteria.maxCount = 5;
-
-    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
-
-    float eps;
-    if( !(criteria.type & TermCriteria::EPS) )
-        eps = 1.f;
-    eps = (float)std::max(criteria.epsilon, 0.0);
-
-    meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// drawColorDisp
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void drawColorDisp_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream);
-        void drawColorDisp_gpu(const PtrStepSz<short>& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream);
-    }
-}}}
-
-namespace
-{
-    template <typename T>
-    void drawColorDisp_caller(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream)
-    {
-        using namespace ::cv::gpu::device::imgproc;
-
-        dst.create(src.size(), CV_8UC4);
-
-        drawColorDisp_gpu((PtrStepSz<T>)src, dst, ndisp, stream);
-    }
-
-    typedef void (*drawColorDisp_caller_t)(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream);
-
-    const drawColorDisp_caller_t drawColorDisp_callers[] = {drawColorDisp_caller<unsigned char>, 0, 0, drawColorDisp_caller<short>, 0, 0, 0, 0};
-}
-
-void cv::gpu::drawColorDisp(const GpuMat& src, GpuMat& dst, int ndisp, Stream& stream)
-{
-    CV_Assert(src.type() == CV_8U || src.type() == CV_16S);
-
-    drawColorDisp_callers[src.type()](src, dst, ndisp, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// reprojectImageTo3D
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T, typename D>
-        void reprojectImageTo3D_gpu(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyz, const Mat& Q, int dst_cn, Stream& stream)
-{
-    using namespace cv::gpu::device::imgproc;
-
-    typedef void (*func_t)(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
-    static const func_t funcs[2][4] =
-    {
-        {reprojectImageTo3D_gpu<uchar, float3>, 0, 0, reprojectImageTo3D_gpu<short, float3>},
-        {reprojectImageTo3D_gpu<uchar, float4>, 0, 0, reprojectImageTo3D_gpu<short, float4>}
-    };
-
-    CV_Assert(disp.type() == CV_8U || disp.type() == CV_16S);
-    CV_Assert(Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4 && Q.isContinuous());
-    CV_Assert(dst_cn == 3 || dst_cn == 4);
-
-    xyz.create(disp.size(), CV_MAKE_TYPE(CV_32F, dst_cn));
-
-    funcs[dst_cn == 4][disp.type()](disp, xyz, Q.ptr<float>(), StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// copyMakeBorder
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T, int cn> void copyMakeBorder_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderMode, const T* borderValue, cudaStream_t stream);
-    }
-}}}
-
-namespace
-{
-    template <typename T, int cn> void copyMakeBorder_caller(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderType, const Scalar& value, cudaStream_t stream)
-    {
-        using namespace ::cv::gpu::device::imgproc;
-
-        Scalar_<T> val(saturate_cast<T>(value[0]), saturate_cast<T>(value[1]), saturate_cast<T>(value[2]), saturate_cast<T>(value[3]));
-
-        copyMakeBorder_gpu<T, cn>(src, dst, top, left, borderType, val.val, stream);
-    }
-}
-
-#if defined __GNUC__ && __GNUC__ > 2 && __GNUC_MINOR__  > 4
-typedef Npp32s __attribute__((__may_alias__)) Npp32s_a;
-#else
-typedef Npp32s Npp32s_a;
-#endif
-
-void cv::gpu::copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, int borderType, const Scalar& value, Stream& s)
-{
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    CV_Assert(borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP);
-
-    dst.create(src.rows + top + bottom, src.cols + left + right, src.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    if (borderType == BORDER_CONSTANT && (src.type() == CV_8UC1 || src.type() == CV_8UC4 || src.type() == CV_32SC1 || src.type() == CV_32FC1))
-    {
-        NppiSize srcsz;
-        srcsz.width  = src.cols;
-        srcsz.height = src.rows;
-
-        NppiSize dstsz;
-        dstsz.width  = dst.cols;
-        dstsz.height = dst.rows;
-
-        NppStreamHandler h(stream);
-
-        switch (src.type())
-        {
-        case CV_8UC1:
-            {
-                Npp8u nVal = saturate_cast<Npp8u>(value[0]);
-                nppSafeCall( nppiCopyConstBorder_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
-                    dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
-                break;
-            }
-        case CV_8UC4:
-            {
-                Npp8u nVal[] = {saturate_cast<Npp8u>(value[0]), saturate_cast<Npp8u>(value[1]), saturate_cast<Npp8u>(value[2]), saturate_cast<Npp8u>(value[3])};
-                nppSafeCall( nppiCopyConstBorder_8u_C4R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
-                    dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
-                break;
-            }
-        case CV_32SC1:
-            {
-                Npp32s nVal = saturate_cast<Npp32s>(value[0]);
-                nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
-                    dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
-                break;
-            }
-        case CV_32FC1:
-            {
-                Npp32f val = saturate_cast<Npp32f>(value[0]);
-                Npp32s nVal = *(reinterpret_cast<Npp32s_a*>(&val));
-                nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
-                    dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
-                break;
-            }
-        }
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-    else
-    {
-        typedef void (*caller_t)(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderType, const Scalar& value, cudaStream_t stream);
-        static const caller_t callers[6][4] =
-        {
-            {   copyMakeBorder_caller<uchar, 1>  ,    copyMakeBorder_caller<uchar, 2>   ,    copyMakeBorder_caller<uchar, 3>  ,    copyMakeBorder_caller<uchar, 4>},
-            {0/*copyMakeBorder_caller<schar, 1>*/, 0/*copyMakeBorder_caller<schar, 2>*/ , 0/*copyMakeBorder_caller<schar, 3>*/, 0/*copyMakeBorder_caller<schar, 4>*/},
-            {   copyMakeBorder_caller<ushort, 1> , 0/*copyMakeBorder_caller<ushort, 2>*/,    copyMakeBorder_caller<ushort, 3> ,    copyMakeBorder_caller<ushort, 4>},
-            {   copyMakeBorder_caller<short, 1>  , 0/*copyMakeBorder_caller<short, 2>*/ ,    copyMakeBorder_caller<short, 3>  ,    copyMakeBorder_caller<short, 4>},
-            {0/*copyMakeBorder_caller<int,   1>*/, 0/*copyMakeBorder_caller<int,   2>*/ , 0/*copyMakeBorder_caller<int,   3>*/, 0/*copyMakeBorder_caller<int  , 4>*/},
-            {   copyMakeBorder_caller<float, 1>  , 0/*copyMakeBorder_caller<float, 2>*/ ,    copyMakeBorder_caller<float, 3>  ,    copyMakeBorder_caller<float ,4>}
-        };
-
-        caller_t func = callers[src.depth()][src.channels() - 1];
-        CV_Assert(func != 0);
-
-        int gpuBorderType;
-        CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
-
-        func(src, dst, top, left, gpuBorderType, value, stream);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// buildWarpPlaneMaps
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void buildWarpPlaneMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                const float k_rinv[9], const float r_kinv[9], const float t[3], float scale,
-                                cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, const Mat &T,
-                                 float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream)
-{
-    (void)src_size;
-    using namespace ::cv::gpu::device::imgproc;
-
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
-    CV_Assert((T.size() == Size(3,1) || T.size() == Size(1,3)) && T.type() == CV_32F && T.isContinuous());
-
-    Mat K_Rinv = K * R.t();
-    Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
-    CV_Assert(R_Kinv.isContinuous());
-
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
-    device::imgproc::buildWarpPlaneMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(),
-                       T.ptr<float>(), scale, StreamAccessor::getStream(stream));
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// buildWarpCylyndricalMaps
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void buildWarpCylindricalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                      const float k_rinv[9], const float r_kinv[9], float scale,
-                                      cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
-                                       GpuMat& map_x, GpuMat& map_y, Stream& stream)
-{
-    (void)src_size;
-    using namespace ::cv::gpu::device::imgproc;
-
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
-
-    Mat K_Rinv = K * R.t();
-    Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
-    CV_Assert(R_Kinv.isContinuous());
-
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
-    device::imgproc::buildWarpCylindricalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
-}
-
-
-//////////////////////////////////////////////////////////////////////////////
-// buildWarpSphericalMaps
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void buildWarpSphericalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
-                                    const float k_rinv[9], const float r_kinv[9], float scale,
-                                    cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
-                                     GpuMat& map_x, GpuMat& map_y, Stream& stream)
-{
-    (void)src_size;
-    using namespace ::cv::gpu::device::imgproc;
-
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
-
-    Mat K_Rinv = K * R.t();
-    Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
-    CV_Assert(R_Kinv.isContinuous());
-
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
-    device::imgproc::buildWarpSphericalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// rotate
-
-namespace
-{
-    template<int DEPTH> struct NppTypeTraits;
-    template<> struct NppTypeTraits<CV_8U>  { typedef Npp8u npp_t; };
-    template<> struct NppTypeTraits<CV_8S>  { typedef Npp8s npp_t; };
-    template<> struct NppTypeTraits<CV_16U> { typedef Npp16u npp_t; };
-    template<> struct NppTypeTraits<CV_16S> { typedef Npp16s npp_t; };
-    template<> struct NppTypeTraits<CV_32S> { typedef Npp32s npp_t; };
-    template<> struct NppTypeTraits<CV_32F> { typedef Npp32f npp_t; };
-    template<> struct NppTypeTraits<CV_64F> { typedef Npp64f npp_t; };
-
-    template <int DEPTH> struct NppRotateFunc
-    {
-        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
-
-        typedef NppStatus (*func_t)(const npp_t* pSrc, NppiSize oSrcSize, int nSrcStep, NppiRect oSrcROI,
-                                    npp_t* pDst, int nDstStep, NppiRect oDstROI,
-                                    double nAngle, double nShiftX, double nShiftY, int eInterpolation);
-    };
-
-    template <int DEPTH, typename NppRotateFunc<DEPTH>::func_t func> struct NppRotate
-    {
-        typedef typename NppRotateFunc<DEPTH>::npp_t npp_t;
-
-        static void call(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, cudaStream_t stream)
-        {
-            (void)dsize;
-            static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
-
-            NppStreamHandler h(stream);
-
-            NppiSize srcsz;
-            srcsz.height = src.rows;
-            srcsz.width = src.cols;
-            NppiRect srcroi;
-            srcroi.x = srcroi.y = 0;
-            srcroi.height = src.rows;
-            srcroi.width = src.cols;
-            NppiRect dstroi;
-            dstroi.x = dstroi.y = 0;
-            dstroi.height = dst.rows;
-            dstroi.width = dst.cols;
-
-            nppSafeCall( func(src.ptr<npp_t>(), srcsz, static_cast<int>(src.step), srcroi,
-                dst.ptr<npp_t>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& stream)
-{
-    typedef void (*func_t)(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, cudaStream_t stream);
-
-    static const func_t funcs[6][4] =
-    {
-        {NppRotate<CV_8U, nppiRotate_8u_C1R>::call, 0, NppRotate<CV_8U, nppiRotate_8u_C3R>::call, NppRotate<CV_8U, nppiRotate_8u_C4R>::call},
-        {0,0,0,0},
-        {NppRotate<CV_16U, nppiRotate_16u_C1R>::call, 0, NppRotate<CV_16U, nppiRotate_16u_C3R>::call, NppRotate<CV_16U, nppiRotate_16u_C4R>::call},
-        {0,0,0,0},
-        {0,0,0,0},
-        {NppRotate<CV_32F, nppiRotate_32f_C1R>::call, 0, NppRotate<CV_32F, nppiRotate_32f_C3R>::call, NppRotate<CV_32F, nppiRotate_32f_C4R>::call}
-    };
-
-    CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
-    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-
-    dst.create(dsize, src.type());
-    dst.setTo(Scalar::all(0));
-
-    funcs[src.depth()][src.channels() - 1](src, dst, dsize, angle, xShift, yShift, interpolation, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// integral
-
-void cv::gpu::integral(const GpuMat& src, GpuMat& sum, Stream& s)
-{
-    GpuMat buffer;
-    integralBuffered(src, sum, buffer, s);
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void shfl_integral_gpu(const PtrStepSzb& img, PtrStepSz<unsigned int> integral, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer, Stream& s)
-{
-    CV_Assert(src.type() == CV_8UC1);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    cv::Size whole;
-    cv::Point offset;
-
-    src.locateROI(whole, offset);
-
-    if (deviceSupports(WARP_SHUFFLE_FUNCTIONS) && src.cols <= 2048
-        && offset.x % 16 == 0 && ((src.cols + 63) / 64) * 64 <= (static_cast<int>(src.step) - offset.x))
-    {
-        ensureSizeIsEnough(((src.rows + 7) / 8) * 8, ((src.cols + 63) / 64) * 64, CV_32SC1, buffer);
-
-        cv::gpu::device::imgproc::shfl_integral_gpu(src, buffer, stream);
-
-        sum.create(src.rows + 1, src.cols + 1, CV_32SC1);
-        if (s)
-            s.enqueueMemSet(sum, Scalar::all(0));
-        else
-            sum.setTo(Scalar::all(0));
-
-        GpuMat inner = sum(Rect(1, 1, src.cols, src.rows));
-        GpuMat res = buffer(Rect(0, 0, src.cols, src.rows));
-
-        if (s)
-            s.enqueueCopy(res, inner);
-        else
-            res.copyTo(inner);
-    }
-    else
-    {
-        sum.create(src.rows + 1, src.cols + 1, CV_32SC1);
-
-        NcvSize32u roiSize;
-        roiSize.width = src.cols;
-        roiSize.height = src.rows;
-
-        cudaDeviceProp prop;
-        cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
-
-        Ncv32u bufSize;
-        ncvSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize, prop) );
-        ensureSizeIsEnough(1, bufSize, CV_8UC1, buffer);
-
-
-        NppStStreamHandler h(stream);
-
-        ncvSafeCall( nppiStIntegral_8u32u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>()), static_cast<int>(src.step),
-            sum.ptr<Ncv32u>(), static_cast<int>(sum.step), roiSize, buffer.ptr<Ncv8u>(), bufSize, prop) );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// sqrIntegral
-
-void cv::gpu::sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& s)
-{
-    CV_Assert(src.type() == CV_8U);
-
-    NcvSize32u roiSize;
-    roiSize.width = src.cols;
-    roiSize.height = src.rows;
-
-    cudaDeviceProp prop;
-    cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
-
-    Ncv32u bufSize;
-    ncvSafeCall(nppiStSqrIntegralGetSize_8u64u(roiSize, &bufSize, prop));
-    GpuMat buf(1, bufSize, CV_8U);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    NppStStreamHandler h(stream);
-
-    sqsum.create(src.rows + 1, src.cols + 1, CV_64F);
-    ncvSafeCall(nppiStSqrIntegral_8u64u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>(0)), static_cast<int>(src.step),
-            sqsum.ptr<Ncv64u>(0), static_cast<int>(sqsum.step), roiSize, buf.ptr<Ncv8u>(0), bufSize, prop));
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// columnSum
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void columnSum_32F(const PtrStepSzb src, const PtrStepSzb dst);
-    }
-}}}
-
-void cv::gpu::columnSum(const GpuMat& src, GpuMat& dst)
-{
-    using namespace ::cv::gpu::device::imgproc;
-
-    CV_Assert(src.type() == CV_32F);
-
-    dst.create(src.size(), CV_32F);
-
-    device::imgproc::columnSum_32F(src, dst);
-}
-
-void cv::gpu::rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& s)
-{
-    CV_Assert(src.type() == CV_32SC1 && sqr.type() == CV_64FC1);
-
-    dst.create(src.size(), CV_32FC1);
-
-    NppiSize sz;
-    sz.width = src.cols;
-    sz.height = src.rows;
-
-    NppiRect nppRect;
-    nppRect.height = rect.height;
-    nppRect.width = rect.width;
-    nppRect.x = rect.x;
-    nppRect.y = rect.y;
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    NppStreamHandler h(stream);
-
-    nppSafeCall( nppiRectStdDev_32s32f_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), sqr.ptr<Npp64f>(), static_cast<int>(sqr.step),
-                dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, nppRect) );
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-
-////////////////////////////////////////////////////////////////////////
-// Histogram
-
-namespace
-{
-    typedef NppStatus (*get_buf_size_c1_t)(NppiSize oSizeROI, int nLevels, int* hpBufferSize);
-    typedef NppStatus (*get_buf_size_c4_t)(NppiSize oSizeROI, int nLevels[], int* hpBufferSize);
-
-    template<int SDEPTH> struct NppHistogramEvenFuncC1
-    {
-        typedef typename NppTypeTraits<SDEPTH>::npp_t src_t;
-
-    typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s * pHist,
-            int nLevels, Npp32s nLowerLevel, Npp32s nUpperLevel, Npp8u * pBuffer);
-    };
-    template<int SDEPTH> struct NppHistogramEvenFuncC4
-    {
-        typedef typename NppTypeTraits<SDEPTH>::npp_t src_t;
-
-        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI,
-            Npp32s * pHist[4], int nLevels[4], Npp32s nLowerLevel[4], Npp32s nUpperLevel[4], Npp8u * pBuffer);
-    };
-
-    template<int SDEPTH, typename NppHistogramEvenFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
-    struct NppHistogramEvenC1
-    {
-        typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;
-
-        static void hist(const GpuMat& src, GpuMat& hist, GpuMat& buffer, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)
-        {
-            int levels = histSize + 1;
-            hist.create(1, histSize, CV_32S);
-
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            int buf_size;
-            get_buf_size(sz, levels, &buf_size);
-
-            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels,
-                lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-    template<int SDEPTH, typename NppHistogramEvenFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
-    struct NppHistogramEvenC4
-    {
-        typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;
-
-        static void hist(const GpuMat& src, GpuMat hist[4], GpuMat& buffer, int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)
-        {
-            int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};
-            hist[0].create(1, histSize[0], CV_32S);
-            hist[1].create(1, histSize[1], CV_32S);
-            hist[2].create(1, histSize[2], CV_32S);
-            hist[3].create(1, histSize[3], CV_32S);
-
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
-
-            int buf_size;
-            get_buf_size(sz, levels, &buf_size);
-
-            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, levels, lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-
-    template<int SDEPTH> struct NppHistogramRangeFuncC1
-    {
-        typedef typename NppTypeTraits<SDEPTH>::npp_t src_t;
-        typedef Npp32s level_t;
-        enum {LEVEL_TYPE_CODE=CV_32SC1};
-
-        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
-            const Npp32s* pLevels, int nLevels, Npp8u* pBuffer);
-    };
-    template<> struct NppHistogramRangeFuncC1<CV_32F>
-    {
-        typedef Npp32f src_t;
-        typedef Npp32f level_t;
-        enum {LEVEL_TYPE_CODE=CV_32FC1};
-
-        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
-            const Npp32f* pLevels, int nLevels, Npp8u* pBuffer);
-    };
-    template<int SDEPTH> struct NppHistogramRangeFuncC4
-    {
-        typedef typename NppTypeTraits<SDEPTH>::npp_t src_t;
-        typedef Npp32s level_t;
-        enum {LEVEL_TYPE_CODE=CV_32SC1};
-
-        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
-            const Npp32s* pLevels[4], int nLevels[4], Npp8u* pBuffer);
-    };
-    template<> struct NppHistogramRangeFuncC4<CV_32F>
-    {
-        typedef Npp32f src_t;
-        typedef Npp32f level_t;
-        enum {LEVEL_TYPE_CODE=CV_32FC1};
-
-        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
-            const Npp32f* pLevels[4], int nLevels[4], Npp8u* pBuffer);
-    };
-
-    template<int SDEPTH, typename NppHistogramRangeFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
-    struct NppHistogramRangeC1
-    {
-        typedef typename NppHistogramRangeFuncC1<SDEPTH>::src_t src_t;
-        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
-        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
-
-        static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buffer, cudaStream_t stream)
-        {
-            CV_Assert(levels.type() == LEVEL_TYPE_CODE && levels.rows == 1);
-
-            hist.create(1, levels.cols - 1, CV_32S);
-
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            int buf_size;
-            get_buf_size(sz, levels.cols, &buf_size);
-
-            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels.ptr<level_t>(), levels.cols, buffer.ptr<Npp8u>()) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-    template<int SDEPTH, typename NppHistogramRangeFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
-    struct NppHistogramRangeC4
-    {
-        typedef typename NppHistogramRangeFuncC4<SDEPTH>::src_t src_t;
-        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
-        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
-
-        static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buffer, cudaStream_t stream)
-        {
-            CV_Assert(levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1);
-            CV_Assert(levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1);
-            CV_Assert(levels[2].type() == LEVEL_TYPE_CODE && levels[2].rows == 1);
-            CV_Assert(levels[3].type() == LEVEL_TYPE_CODE && levels[3].rows == 1);
-
-            hist[0].create(1, levels[0].cols - 1, CV_32S);
-            hist[1].create(1, levels[1].cols - 1, CV_32S);
-            hist[2].create(1, levels[2].cols - 1, CV_32S);
-            hist[3].create(1, levels[3].cols - 1, CV_32S);
-
-            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
-            int nLevels[] = {levels[0].cols, levels[1].cols, levels[2].cols, levels[3].cols};
-            const level_t* pLevels[] = {levels[0].ptr<level_t>(), levels[1].ptr<level_t>(), levels[2].ptr<level_t>(), levels[3].ptr<level_t>()};
-
-            NppiSize sz;
-            sz.width = src.cols;
-            sz.height = src.rows;
-
-            int buf_size;
-            get_buf_size(sz, nLevels, &buf_size);
-
-            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
-
-            NppStreamHandler h(stream);
-
-            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, pLevels, nLevels, buffer.ptr<Npp8u>()) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-void cv::gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel)
-{
-    Mat host_levels(1, nLevels, CV_32SC1);
-    nppSafeCall( nppiEvenLevelsHost_32s(host_levels.ptr<Npp32s>(), nLevels, lowerLevel, upperLevel) );
-    levels.upload(host_levels);
-}
-
-void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream)
-{
-    GpuMat buf;
-    histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
-}
-
-void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream)
-{
-    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );
-
-    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, GpuMat& buf, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);
-    static const hist_t hist_callers[] =
-    {
-        NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,
-        0,
-        NppHistogramEvenC1<CV_16U, nppiHistogramEven_16u_C1R, nppiHistogramEvenGetBufferSize_16u_C1R>::hist,
-        NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist
-    };
-
-    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
-{
-    GpuMat buf;
-    histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
-}
-
-void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
-{
-    CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );
-
-    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int levels[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream);
-    static const hist_t hist_callers[] =
-    {
-        NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,
-        0,
-        NppHistogramEvenC4<CV_16U, nppiHistogramEven_16u_C4R, nppiHistogramEvenGetBufferSize_16u_C4R>::hist,
-        NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist
-    };
-
-    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream)
-{
-    GpuMat buf;
-    histRange(src, hist, levels, buf, stream);
-}
-
-void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream)
-{
-    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1);
-
-    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, cudaStream_t stream);
-    static const hist_t hist_callers[] =
-    {
-        NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,
-        0,
-        NppHistogramRangeC1<CV_16U, nppiHistogramRange_16u_C1R, nppiHistogramRangeGetBufferSize_16u_C1R>::hist,
-        NppHistogramRangeC1<CV_16S, nppiHistogramRange_16s_C1R, nppiHistogramRangeGetBufferSize_16s_C1R>::hist,
-        0,
-        NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist
-    };
-
-    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream)
-{
-    GpuMat buf;
-    histRange(src, hist, levels, buf, stream);
-}
-
-void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream)
-{
-    CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4);
-
-    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, cudaStream_t stream);
-    static const hist_t hist_callers[] =
-    {
-        NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,
-        0,
-        NppHistogramRangeC4<CV_16U, nppiHistogramRange_16u_C4R, nppiHistogramRangeGetBufferSize_16u_C4R>::hist,
-        NppHistogramRangeC4<CV_16S, nppiHistogramRange_16s_C4R, nppiHistogramRangeGetBufferSize_16s_C4R>::hist,
-        0,
-        NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist
-    };
-
-    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
-}
-
-namespace hist
-{
-    void histogram256(PtrStepSzb src, int* hist, cudaStream_t stream);
-    void equalizeHist(PtrStepSzb src, PtrStepSzb dst, const int* lut, cudaStream_t stream);
-}
-
-void cv::gpu::calcHist(const GpuMat& src, GpuMat& hist, Stream& stream)
-{
-    CV_Assert(src.type() == CV_8UC1);
-
-    hist.create(1, 256, CV_32SC1);
-    hist.setTo(Scalar::all(0));
-
-    hist::histogram256(src, hist.ptr<int>(), StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    GpuMat hist;
-    GpuMat buf;
-    equalizeHist(src, dst, hist, buf, stream);
-}
-
-void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, GpuMat& buf, Stream& s)
-{
-    CV_Assert(src.type() == CV_8UC1);
-
-    dst.create(src.size(), src.type());
-
-    int intBufSize;
-    nppSafeCall( nppsIntegralGetBufferSize_32s(256, &intBufSize) );
-
-    ensureSizeIsEnough(1, intBufSize + 256 * sizeof(int), CV_8UC1, buf);
-
-    GpuMat intBuf(1, intBufSize, CV_8UC1, buf.ptr());
-    GpuMat lut(1, 256, CV_32S, buf.ptr() + intBufSize);
-
-    calcHist(src, hist, s);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    NppStreamHandler h(stream);
-
-    nppSafeCall( nppsIntegral_32s(hist.ptr<Npp32s>(), lut.ptr<Npp32s>(), 256, intBuf.ptr<Npp8u>()) );
-
-    hist::equalizeHist(src, dst, lut.ptr<int>(), stream);
-}
-
-////////////////////////////////////////////////////////////////////////
-// cornerHarris & minEgenVal
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void cornerHarris_gpu(int block_size, float k, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream);
-        void cornerMinEigenVal_gpu(int block_size, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream);
-    }
-}}}
-
-namespace
-{
-    void extractCovData(const GpuMat& src, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize, int borderType, Stream& stream)
-    {
-        double scale = static_cast<double>(1 << ((ksize > 0 ? ksize : 3) - 1)) * blockSize;
-
-        if (ksize < 0)
-            scale *= 2.;
-
-        if (src.depth() == CV_8U)
-            scale *= 255.;
-
-        scale = 1./scale;
-
-        Dx.create(src.size(), CV_32F);
-        Dy.create(src.size(), CV_32F);
-
-        if (ksize > 0)
-        {
-            Sobel(src, Dx, CV_32F, 1, 0, buf, ksize, scale, borderType, -1, stream);
-            Sobel(src, Dy, CV_32F, 0, 1, buf, ksize, scale, borderType, -1, stream);
-        }
-        else
-        {
-            Scharr(src, Dx, CV_32F, 1, 0, buf, scale, borderType, -1, stream);
-            Scharr(src, Dy, CV_32F, 0, 1, buf, scale, borderType, -1, stream);
-        }
-    }
-}
-
-void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType)
-{
-    GpuMat Dx, Dy;
-    cornerHarris(src, dst, Dx, Dy, blockSize, ksize, k, borderType);
-}
-
-void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType)
-{
-    GpuMat buf;
-    cornerHarris(src, dst, Dx, Dy, buf, blockSize, ksize, k, borderType);
-}
-
-void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize, double k, int borderType, Stream& stream)
-{
-    using namespace cv::gpu::device::imgproc;
-
-    CV_Assert(borderType == cv::BORDER_REFLECT101 || borderType == cv::BORDER_REPLICATE || borderType == cv::BORDER_REFLECT);
-
-    int gpuBorderType;
-    CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
-
-    extractCovData(src, Dx, Dy, buf, blockSize, ksize, borderType, stream);
-
-    dst.create(src.size(), CV_32F);
-
-    cornerHarris_gpu(blockSize, static_cast<float>(k), Dx, Dy, dst, gpuBorderType, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType)
-{
-    GpuMat Dx, Dy;
-    cornerMinEigenVal(src, dst, Dx, Dy, blockSize, ksize, borderType);
-}
-
-void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
-{
-    GpuMat buf;
-    cornerMinEigenVal(src, dst, Dx, Dy, buf, blockSize, ksize, borderType);
-}
-
-void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, GpuMat& buf, int blockSize, int ksize, int borderType, Stream& stream)
-{
-    using namespace ::cv::gpu::device::imgproc;
-
-    CV_Assert(borderType == cv::BORDER_REFLECT101 || borderType == cv::BORDER_REPLICATE || borderType == cv::BORDER_REFLECT);
-
-    int gpuBorderType;
-    CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
-
-    extractCovData(src, Dx, Dy, buf, blockSize, ksize, borderType, stream);
-
-    dst.create(src.size(), CV_32F);
-
-    cornerMinEigenVal_gpu(blockSize, Dx, Dy, dst, gpuBorderType, StreamAccessor::getStream(stream));
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// mulSpectrums
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void mulSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream);
-
-        void mulSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::mulSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, bool conjB, Stream& stream)
-{
-    (void)flags;
-    using namespace ::cv::gpu::device::imgproc;
-
-    typedef void (*Caller)(const PtrStep<cufftComplex>, const PtrStep<cufftComplex>, PtrStepSz<cufftComplex>, cudaStream_t stream);
-
-    static Caller callers[] = { device::imgproc::mulSpectrums, device::imgproc::mulSpectrums_CONJ };
-
-    CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
-    CV_Assert(a.size() == b.size());
-
-    c.create(a.size(), CV_32FC2);
-
-    Caller caller = callers[(int)conjB];
-    caller(a, b, c, StreamAccessor::getStream(stream));
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// mulAndScaleSpectrums
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        void mulAndScaleSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream);
-
-        void mulAndScaleSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::mulAndScaleSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, float scale, bool conjB, Stream& stream)
-{
-    (void)flags;
-    using namespace ::cv::gpu::device::imgproc;
-
-    typedef void (*Caller)(const PtrStep<cufftComplex>, const PtrStep<cufftComplex>, float scale, PtrStepSz<cufftComplex>, cudaStream_t stream);
-    static Caller callers[] = { device::imgproc::mulAndScaleSpectrums, device::imgproc::mulAndScaleSpectrums_CONJ };
-
-    CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
-    CV_Assert(a.size() == b.size());
-
-    c.create(a.size(), CV_32FC2);
-
-    Caller caller = callers[(int)conjB];
-    caller(a, b, scale, c, StreamAccessor::getStream(stream));
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// dft
-
-void cv::gpu::dft(const GpuMat& src, GpuMat& dst, Size dft_size, int flags, Stream& stream)
-{
-#ifndef HAVE_CUFFT
-
-    OPENCV_GPU_UNUSED(src);
-    OPENCV_GPU_UNUSED(dst);
-    OPENCV_GPU_UNUSED(dft_size);
-    OPENCV_GPU_UNUSED(flags);
-    OPENCV_GPU_UNUSED(stream);
-
-    throw_nogpu();
-
-#else
-
-    CV_Assert(src.type() == CV_32F || src.type() == CV_32FC2);
-
-    // We don't support unpacked output (in the case of real input)
-    CV_Assert(!(flags & DFT_COMPLEX_OUTPUT));
-
-    bool is_1d_input = (dft_size.height == 1) || (dft_size.width == 1);
-    int is_row_dft = flags & DFT_ROWS;
-    int is_scaled_dft = flags & DFT_SCALE;
-    int is_inverse = flags & DFT_INVERSE;
-    bool is_complex_input = src.channels() == 2;
-    bool is_complex_output = !(flags & DFT_REAL_OUTPUT);
-
-    // We don't support real-to-real transform
-    CV_Assert(is_complex_input || is_complex_output);
-
-    GpuMat src_data;
-
-    // Make sure here we work with the continuous input,
-    // as CUFFT can't handle gaps
-    src_data = src;
-    createContinuous(src.rows, src.cols, src.type(), src_data);
-    if (src_data.data != src.data)
-        src.copyTo(src_data);
-
-    Size dft_size_opt = dft_size;
-    if (is_1d_input && !is_row_dft)
-    {
-        // If the source matrix is single column handle it as single row
-        dft_size_opt.width = std::max(dft_size.width, dft_size.height);
-        dft_size_opt.height = std::min(dft_size.width, dft_size.height);
-    }
-
-    cufftType dft_type = CUFFT_R2C;
-    if (is_complex_input)
-        dft_type = is_complex_output ? CUFFT_C2C : CUFFT_C2R;
-
-    CV_Assert(dft_size_opt.width > 1);
-
-    cufftHandle plan;
-    if (is_1d_input || is_row_dft)
-        cufftPlan1d(&plan, dft_size_opt.width, dft_type, dft_size_opt.height);
-    else
-        cufftPlan2d(&plan, dft_size_opt.height, dft_size_opt.width, dft_type);
-
-    cufftSafeCall( cufftSetStream(plan, StreamAccessor::getStream(stream)) );
-
-    if (is_complex_input)
-    {
-        if (is_complex_output)
-        {
-            createContinuous(dft_size, CV_32FC2, dst);
-            cufftSafeCall(cufftExecC2C(
-                    plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftComplex>(),
-                    is_inverse ? CUFFT_INVERSE : CUFFT_FORWARD));
-        }
-        else
-        {
-            createContinuous(dft_size, CV_32F, dst);
-            cufftSafeCall(cufftExecC2R(
-                    plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftReal>()));
-        }
-    }
-    else
-    {
-        // We could swap dft_size for efficiency. Here we must reflect it
-        if (dft_size == dft_size_opt)
-            createContinuous(Size(dft_size.width / 2 + 1, dft_size.height), CV_32FC2, dst);
-        else
-            createContinuous(Size(dft_size.width, dft_size.height / 2 + 1), CV_32FC2, dst);
-
-        cufftSafeCall(cufftExecR2C(
-                plan, src_data.ptr<cufftReal>(), dst.ptr<cufftComplex>()));
-    }
-
-    cufftSafeCall(cufftDestroy(plan));
-
-    if (is_scaled_dft)
-        multiply(dst, Scalar::all(1. / dft_size.area()), dst, 1, -1, stream);
-
-#endif
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// convolve
-
-void cv::gpu::ConvolveBuf::create(Size image_size, Size templ_size)
-{
-    result_size = Size(image_size.width - templ_size.width + 1,
-                       image_size.height - templ_size.height + 1);
-
-    block_size = user_block_size;
-    if (user_block_size.width == 0 || user_block_size.height == 0)
-        block_size = estimateBlockSize(result_size, templ_size);
-
-    dft_size.width = 1 << int(ceil(std::log(block_size.width + templ_size.width - 1.) / std::log(2.)));
-    dft_size.height = 1 << int(ceil(std::log(block_size.height + templ_size.height - 1.) / std::log(2.)));
-
-    // CUFFT has hard-coded kernels for power-of-2 sizes (up to 8192),
-    // see CUDA Toolkit 4.1 CUFFT Library Programming Guide
-    if (dft_size.width > 8192)
-        dft_size.width = getOptimalDFTSize(block_size.width + templ_size.width - 1);
-    if (dft_size.height > 8192)
-        dft_size.height = getOptimalDFTSize(block_size.height + templ_size.height - 1);
-
-    // To avoid wasting time doing small DFTs
-    dft_size.width = std::max(dft_size.width, 512);
-    dft_size.height = std::max(dft_size.height, 512);
-
-    createContinuous(dft_size, CV_32F, image_block);
-    createContinuous(dft_size, CV_32F, templ_block);
-    createContinuous(dft_size, CV_32F, result_data);
-
-    spect_len = dft_size.height * (dft_size.width / 2 + 1);
-    createContinuous(1, spect_len, CV_32FC2, image_spect);
-    createContinuous(1, spect_len, CV_32FC2, templ_spect);
-    createContinuous(1, spect_len, CV_32FC2, result_spect);
-
-    // Use maximum result matrix block size for the estimated DFT block size
-    block_size.width = std::min(dft_size.width - templ_size.width + 1, result_size.width);
-    block_size.height = std::min(dft_size.height - templ_size.height + 1, result_size.height);
-}
-
-
-Size cv::gpu::ConvolveBuf::estimateBlockSize(Size result_size, Size /*templ_size*/)
-{
-    int width = (result_size.width + 2) / 3;
-    int height = (result_size.height + 2) / 3;
-    width = std::min(width, result_size.width);
-    height = std::min(height, result_size.height);
-    return Size(width, height);
-}
-
-
-void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr)
-{
-    ConvolveBuf buf;
-    convolve(image, templ, result, ccorr, buf);
-}
-
-void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result, bool ccorr, ConvolveBuf& buf, Stream& stream)
-{
-    using namespace ::cv::gpu::device::imgproc;
-
-#ifndef HAVE_CUFFT
-    throw_nogpu();
-#else
-    StaticAssert<sizeof(float) == sizeof(cufftReal)>::check();
-    StaticAssert<sizeof(float) * 2 == sizeof(cufftComplex)>::check();
-
-    CV_Assert(image.type() == CV_32F);
-    CV_Assert(templ.type() == CV_32F);
-
-    buf.create(image.size(), templ.size());
-    result.create(buf.result_size, CV_32F);
-
-    Size& block_size = buf.block_size;
-    Size& dft_size = buf.dft_size;
-
-    GpuMat& image_block = buf.image_block;
-    GpuMat& templ_block = buf.templ_block;
-    GpuMat& result_data = buf.result_data;
-
-    GpuMat& image_spect = buf.image_spect;
-    GpuMat& templ_spect = buf.templ_spect;
-    GpuMat& result_spect = buf.result_spect;
-
-    cufftHandle planR2C, planC2R;
-    cufftSafeCall(cufftPlan2d(&planC2R, dft_size.height, dft_size.width, CUFFT_C2R));
-    cufftSafeCall(cufftPlan2d(&planR2C, dft_size.height, dft_size.width, CUFFT_R2C));
-
-    cufftSafeCall( cufftSetStream(planR2C, StreamAccessor::getStream(stream)) );
-    cufftSafeCall( cufftSetStream(planC2R, StreamAccessor::getStream(stream)) );
-
-    GpuMat templ_roi(templ.size(), CV_32F, templ.data, templ.step);
-    copyMakeBorder(templ_roi, templ_block, 0, templ_block.rows - templ_roi.rows, 0,
-                   templ_block.cols - templ_roi.cols, 0, Scalar(), stream);
-
-    cufftSafeCall(cufftExecR2C(planR2C, templ_block.ptr<cufftReal>(),
-                               templ_spect.ptr<cufftComplex>()));
-
-    // Process all blocks of the result matrix
-    for (int y = 0; y < result.rows; y += block_size.height)
-    {
-        for (int x = 0; x < result.cols; x += block_size.width)
-        {
-            Size image_roi_size(std::min(x + dft_size.width, image.cols) - x,
-                                std::min(y + dft_size.height, image.rows) - y);
-            GpuMat image_roi(image_roi_size, CV_32F, (void*)(image.ptr<float>(y) + x),
-                             image.step);
-            copyMakeBorder(image_roi, image_block, 0, image_block.rows - image_roi.rows,
-                           0, image_block.cols - image_roi.cols, 0, Scalar(), stream);
-
-            cufftSafeCall(cufftExecR2C(planR2C, image_block.ptr<cufftReal>(),
-                                       image_spect.ptr<cufftComplex>()));
-            mulAndScaleSpectrums(image_spect, templ_spect, result_spect, 0,
-                                 1.f / dft_size.area(), ccorr, stream);
-            cufftSafeCall(cufftExecC2R(planC2R, result_spect.ptr<cufftComplex>(),
-                                       result_data.ptr<cufftReal>()));
-
-            Size result_roi_size(std::min(x + block_size.width, result.cols) - x,
-                                 std::min(y + block_size.height, result.rows) - y);
-            GpuMat result_roi(result_roi_size, result.type(),
-                              (void*)(result.ptr<float>(y) + x), result.step);
-            GpuMat result_block(result_roi_size, result_data.type(),
-                                result_data.ptr(), result_data.step);
-
-            if (stream)
-                stream.enqueueCopy(result_block, result_roi);
-            else
-                result_block.copyTo(result_roi);
-        }
-    }
-
-    cufftSafeCall(cufftDestroy(planR2C));
-    cufftSafeCall(cufftDestroy(planC2R));
-#endif
-}
-
-
-//////////////////////////////////////////////////////////////////////////////
-// Canny
-
-void cv::gpu::CannyBuf::create(const Size& image_size, int apperture_size)
-{
-    if (apperture_size > 0)
-    {
-        ensureSizeIsEnough(image_size, CV_32SC1, dx);
-        ensureSizeIsEnough(image_size, CV_32SC1, dy);
-
-        if (apperture_size != 3)
-        {
-            filterDX = createDerivFilter_GPU(CV_8UC1, CV_32S, 1, 0, apperture_size, BORDER_REPLICATE);
-            filterDY = createDerivFilter_GPU(CV_8UC1, CV_32S, 0, 1, apperture_size, BORDER_REPLICATE);
-        }
-    }
-
-    ensureSizeIsEnough(image_size, CV_32FC1, mag);
-    ensureSizeIsEnough(image_size, CV_32SC1, map);
-
-    ensureSizeIsEnough(1, image_size.area(), CV_16UC2, st1);
-    ensureSizeIsEnough(1, image_size.area(), CV_16UC2, st2);
-}
-
-void cv::gpu::CannyBuf::release()
-{
-    dx.release();
-    dy.release();
-    mag.release();
-    map.release();
-    st1.release();
-    st2.release();
-}
-
-namespace canny
-{
-    void calcMagnitude(PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, bool L2Grad);
-    void calcMagnitude(PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, bool L2Grad);
-
-    void calcMap(PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, PtrStepSzi map, float low_thresh, float high_thresh);
-
-    void edgesHysteresisLocal(PtrStepSzi map, ushort2* st1);
-
-    void edgesHysteresisGlobal(PtrStepSzi map, ushort2* st1, ushort2* st2);
-
-    void getEdges(PtrStepSzi map, PtrStepSzb dst);
-}
-
-namespace
-{
-    void CannyCaller(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& dst, float low_thresh, float high_thresh)
-    {
-        using namespace canny;
-
-        buf.map.setTo(Scalar::all(0));
-        calcMap(dx, dy, buf.mag, buf.map, low_thresh, high_thresh);
-
-        edgesHysteresisLocal(buf.map, buf.st1.ptr<ushort2>());
-
-        edgesHysteresisGlobal(buf.map, buf.st1.ptr<ushort2>(), buf.st2.ptr<ushort2>());
-
-        getEdges(buf.map, dst);
-    }
-}
-
-void cv::gpu::Canny(const GpuMat& src, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
-{
-    CannyBuf buf;
-    Canny(src, buf, dst, low_thresh, high_thresh, apperture_size, L2gradient);
-}
-
-void cv::gpu::Canny(const GpuMat& src, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
-{
-    using namespace canny;
-
-    CV_Assert(src.type() == CV_8UC1);
-
-    if (!deviceSupports(SHARED_ATOMICS))
-        CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics");
-
-    if( low_thresh > high_thresh )
-        std::swap( low_thresh, high_thresh);
-
-    dst.create(src.size(), CV_8U);
-    buf.create(src.size(), apperture_size);
-
-    if (apperture_size == 3)
-    {
-        Size wholeSize;
-        Point ofs;
-        src.locateROI(wholeSize, ofs);
-        GpuMat srcWhole(wholeSize, src.type(), src.datastart, src.step);
-
-        calcMagnitude(srcWhole, ofs.x, ofs.y, buf.dx, buf.dy, buf.mag, L2gradient);
-    }
-    else
-    {
-        buf.filterDX->apply(src, buf.dx, Rect(0, 0, src.cols, src.rows));
-        buf.filterDY->apply(src, buf.dy, Rect(0, 0, src.cols, src.rows));
-
-        calcMagnitude(buf.dx, buf.dy, buf.mag, L2gradient);
-    }
-
-    CannyCaller(buf.dx, buf.dy, buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
-}
-
-void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
-{
-    CannyBuf buf;
-    Canny(dx, dy, buf, dst, low_thresh, high_thresh, L2gradient);
-}
-
-void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
-{
-    using namespace canny;
-
-    CV_Assert(TargetArchs::builtWith(SHARED_ATOMICS) && DeviceInfo().supports(SHARED_ATOMICS));
-    CV_Assert(dx.type() == CV_32SC1 && dy.type() == CV_32SC1 && dx.size() == dy.size());
-
-    if( low_thresh > high_thresh )
-        std::swap( low_thresh, high_thresh);
-
-    dst.create(dx.size(), CV_8U);
-    buf.create(dx.size(), -1);
-
-    calcMagnitude(dx, dy, buf.mag, L2gradient);
-
-    CannyCaller(dx, dy, buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/match_template.cpp b/modules/gpu/src/match_template.cpp
deleted file mode 100644
index 26dcd4b9ef..0000000000
--- a/modules/gpu/src/match_template.cpp
+++ /dev/null
@@ -1,439 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::matchTemplate(const GpuMat&, const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-
-#else
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace match_template
-    {
-        void matchTemplateNaive_CCORR_8U(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
-        void matchTemplateNaive_CCORR_32F(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
-
-        void matchTemplateNaive_SQDIFF_8U(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
-        void matchTemplateNaive_SQDIFF_32F(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
-
-        void matchTemplatePrepared_SQDIFF_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum, unsigned long long templ_sqsum, PtrStepSzf result,
-            int cn, cudaStream_t stream);
-
-        void matchTemplatePrepared_SQDIFF_NORMED_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum, unsigned long long templ_sqsum, PtrStepSzf result,
-            int cn, cudaStream_t stream);
-
-        void matchTemplatePrepared_CCOFF_8U(int w, int h, const PtrStepSz<unsigned int> image_sum, unsigned int templ_sum, PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_8UC2(
-            int w, int h,
-            const PtrStepSz<unsigned int> image_sum_r,
-            const PtrStepSz<unsigned int> image_sum_g,
-            unsigned int templ_sum_r,
-            unsigned int templ_sum_g,
-            PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_8UC3(
-                int w, int h,
-                const PtrStepSz<unsigned int> image_sum_r,
-                const PtrStepSz<unsigned int> image_sum_g,
-                const PtrStepSz<unsigned int> image_sum_b,
-                unsigned int templ_sum_r,
-                unsigned int templ_sum_g,
-                unsigned int templ_sum_b,
-                PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_8UC4(
-                int w, int h,
-                const PtrStepSz<unsigned int> image_sum_r,
-                const PtrStepSz<unsigned int> image_sum_g,
-                const PtrStepSz<unsigned int> image_sum_b,
-                const PtrStepSz<unsigned int> image_sum_a,
-                unsigned int templ_sum_r,
-                unsigned int templ_sum_g,
-                unsigned int templ_sum_b,
-                unsigned int templ_sum_a,
-                PtrStepSzf result, cudaStream_t stream);
-
-
-        void matchTemplatePrepared_CCOFF_NORMED_8U(
-                int w, int h, const PtrStepSz<unsigned int> image_sum,
-                const PtrStepSz<unsigned long long> image_sqsum,
-                unsigned int templ_sum, unsigned long long templ_sqsum,
-                PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_NORMED_8UC2(
-                int w, int h,
-                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
-                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
-                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
-                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
-                PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_NORMED_8UC3(
-                int w, int h,
-                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
-                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
-                const PtrStepSz<unsigned int> image_sum_b, const PtrStepSz<unsigned long long> image_sqsum_b,
-                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
-                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
-                unsigned int templ_sum_b, unsigned long long templ_sqsum_b,
-                PtrStepSzf result, cudaStream_t stream);
-        void matchTemplatePrepared_CCOFF_NORMED_8UC4(
-                int w, int h,
-                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
-                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
-                const PtrStepSz<unsigned int> image_sum_b, const PtrStepSz<unsigned long long> image_sqsum_b,
-                const PtrStepSz<unsigned int> image_sum_a, const PtrStepSz<unsigned long long> image_sqsum_a,
-                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
-                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
-                unsigned int templ_sum_b, unsigned long long templ_sqsum_b,
-                unsigned int templ_sum_a, unsigned long long templ_sqsum_a,
-                PtrStepSzf result, cudaStream_t stream);
-
-        void normalize_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum,
-                          unsigned long long templ_sqsum, PtrStepSzf result, int cn, cudaStream_t stream);
-
-        void extractFirstChannel_32F(const PtrStepSzb image, PtrStepSzf result, int cn, cudaStream_t stream);
-    }
-}}}
-
-using namespace ::cv::gpu::device::match_template;
-
-namespace
-{
-
-    // Evaluates optimal template's area threshold. If
-    // template's area is less  than the threshold, we use naive match
-    // template version, otherwise FFT-based (if available)
-    int getTemplateThreshold(int method, int depth)
-    {
-        switch (method)
-        {
-        case CV_TM_CCORR:
-            if (depth == CV_32F) return 250;
-            if (depth == CV_8U) return 300;
-            break;
-        case CV_TM_SQDIFF:
-            if (depth == CV_8U) return 300;
-            break;
-        }
-        CV_Error(CV_StsBadArg, "getTemplateThreshold: unsupported match template mode");
-        return 0;
-    }
-
-
-    void matchTemplate_CCORR_32F(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-        if (templ.size().area() < getTemplateThreshold(CV_TM_CCORR, CV_32F))
-        {
-            matchTemplateNaive_CCORR_32F(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
-            return;
-        }
-
-        ConvolveBuf convolve_buf;
-        convolve_buf.user_block_size = buf.user_block_size;
-
-        if (image.channels() == 1)
-            convolve(image.reshape(1), templ.reshape(1), result, true, convolve_buf, stream);
-        else
-        {
-            GpuMat result_;
-            convolve(image.reshape(1), templ.reshape(1), result_, true, convolve_buf, stream);
-            extractFirstChannel_32F(result_, result, image.channels(), StreamAccessor::getStream(stream));
-        }
-    }
-
-
-    void matchTemplate_CCORR_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        if (templ.size().area() < getTemplateThreshold(CV_TM_CCORR, CV_8U))
-        {
-            result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-            matchTemplateNaive_CCORR_8U(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
-            return;
-        }
-
-        if (stream)
-        {
-            stream.enqueueConvert(image, buf.imagef, CV_32F);
-            stream.enqueueConvert(templ, buf.templf, CV_32F);
-        }
-        else
-        {
-            image.convertTo(buf.imagef, CV_32F);
-            templ.convertTo(buf.templf, CV_32F);
-        }
-        matchTemplate_CCORR_32F(buf.imagef, buf.templf, result, buf, stream);
-    }
-
-
-    void matchTemplate_CCORR_NORMED_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        matchTemplate_CCORR_8U(image, templ, result, buf, stream);
-
-        buf.image_sqsums.resize(1);
-        sqrIntegral(image.reshape(1), buf.image_sqsums[0], stream);
-
-        unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
-        normalize_8U(templ.cols, templ.rows, buf.image_sqsums[0], templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
-    }
-
-
-    void matchTemplate_SQDIFF_32F(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        (void)buf;
-        result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-        matchTemplateNaive_SQDIFF_32F(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
-    }
-
-
-    void matchTemplate_SQDIFF_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        if (templ.size().area() < getTemplateThreshold(CV_TM_SQDIFF, CV_8U))
-        {
-            result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-            matchTemplateNaive_SQDIFF_8U(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
-            return;
-        }
-
-        buf.image_sqsums.resize(1);
-        sqrIntegral(image.reshape(1), buf.image_sqsums[0], stream);
-
-        unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
-
-        matchTemplate_CCORR_8U(image, templ, result, buf, stream);
-        matchTemplatePrepared_SQDIFF_8U(templ.cols, templ.rows, buf.image_sqsums[0], templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
-    }
-
-
-    void matchTemplate_SQDIFF_NORMED_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        buf.image_sqsums.resize(1);
-        sqrIntegral(image.reshape(1), buf.image_sqsums[0], stream);
-
-        unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
-
-        matchTemplate_CCORR_8U(image, templ, result, buf, stream);
-        matchTemplatePrepared_SQDIFF_NORMED_8U(templ.cols, templ.rows, buf.image_sqsums[0], templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
-    }
-
-
-    void matchTemplate_CCOFF_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        matchTemplate_CCORR_8U(image, templ, result, buf, stream);
-
-        if (image.channels() == 1)
-        {
-            buf.image_sums.resize(1);
-            integral(image, buf.image_sums[0], stream);
-
-            unsigned int templ_sum = (unsigned int)sum(templ)[0];
-            matchTemplatePrepared_CCOFF_8U(templ.cols, templ.rows, buf.image_sums[0], templ_sum, result, StreamAccessor::getStream(stream));
-        }
-        else
-        {
-            split(image, buf.images);
-            buf.image_sums.resize(buf.images.size());
-            for (int i = 0; i < image.channels(); ++i)
-                integral(buf.images[i], buf.image_sums[i], stream);
-
-            Scalar templ_sum = sum(templ);
-
-            switch (image.channels())
-            {
-            case 2:
-                matchTemplatePrepared_CCOFF_8UC2(
-                        templ.cols, templ.rows, buf.image_sums[0], buf.image_sums[1],
-                        (unsigned int)templ_sum[0], (unsigned int)templ_sum[1],
-                        result, StreamAccessor::getStream(stream));
-                break;
-            case 3:
-                matchTemplatePrepared_CCOFF_8UC3(
-                        templ.cols, templ.rows, buf.image_sums[0], buf.image_sums[1], buf.image_sums[2],
-                        (unsigned int)templ_sum[0], (unsigned int)templ_sum[1], (unsigned int)templ_sum[2],
-                        result, StreamAccessor::getStream(stream));
-                break;
-            case 4:
-                matchTemplatePrepared_CCOFF_8UC4(
-                        templ.cols, templ.rows, buf.image_sums[0], buf.image_sums[1], buf.image_sums[2], buf.image_sums[3],
-                        (unsigned int)templ_sum[0], (unsigned int)templ_sum[1], (unsigned int)templ_sum[2],
-                        (unsigned int)templ_sum[3], result, StreamAccessor::getStream(stream));
-                break;
-            default:
-                CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
-            }
-        }
-    }
-
-
-    void matchTemplate_CCOFF_NORMED_8U(
-            const GpuMat& image, const GpuMat& templ, GpuMat& result, MatchTemplateBuf &buf, Stream& stream)
-    {
-        if (stream)
-        {
-            stream.enqueueConvert(image, buf.imagef, CV_32F);
-            stream.enqueueConvert(templ, buf.templf, CV_32F);
-        }
-        else
-        {
-            image.convertTo(buf.imagef, CV_32F);
-            templ.convertTo(buf.templf, CV_32F);
-        }
-
-        matchTemplate_CCORR_32F(buf.imagef, buf.templf, result, buf, stream);
-
-        if (image.channels() == 1)
-        {
-            buf.image_sums.resize(1);
-            integral(image, buf.image_sums[0], stream);
-            buf.image_sqsums.resize(1);
-            sqrIntegral(image, buf.image_sqsums[0], stream);
-
-            unsigned int templ_sum = (unsigned int)sum(templ)[0];
-            unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ)[0];
-
-            matchTemplatePrepared_CCOFF_NORMED_8U(
-                    templ.cols, templ.rows, buf.image_sums[0], buf.image_sqsums[0],
-                    templ_sum, templ_sqsum, result, StreamAccessor::getStream(stream));
-        }
-        else
-        {
-            split(image, buf.images);
-            buf.image_sums.resize(buf.images.size());
-            buf.image_sqsums.resize(buf.images.size());
-            for (int i = 0; i < image.channels(); ++i)
-            {
-                integral(buf.images[i], buf.image_sums[i], stream);
-                sqrIntegral(buf.images[i], buf.image_sqsums[i], stream);
-            }
-
-            Scalar templ_sum = sum(templ);
-            Scalar templ_sqsum = sqrSum(templ);
-
-            switch (image.channels())
-            {
-            case 2:
-                matchTemplatePrepared_CCOFF_NORMED_8UC2(
-                        templ.cols, templ.rows,
-                        buf.image_sums[0], buf.image_sqsums[0],
-                        buf.image_sums[1], buf.image_sqsums[1],
-                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
-                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
-                        result, StreamAccessor::getStream(stream));
-                break;
-            case 3:
-                matchTemplatePrepared_CCOFF_NORMED_8UC3(
-                        templ.cols, templ.rows,
-                        buf.image_sums[0], buf.image_sqsums[0],
-                        buf.image_sums[1], buf.image_sqsums[1],
-                        buf.image_sums[2], buf.image_sqsums[2],
-                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
-                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
-                        (unsigned int)templ_sum[2], (unsigned long long)templ_sqsum[2],
-                        result, StreamAccessor::getStream(stream));
-                break;
-            case 4:
-                matchTemplatePrepared_CCOFF_NORMED_8UC4(
-                        templ.cols, templ.rows,
-                        buf.image_sums[0], buf.image_sqsums[0],
-                        buf.image_sums[1], buf.image_sqsums[1],
-                        buf.image_sums[2], buf.image_sqsums[2],
-                        buf.image_sums[3], buf.image_sqsums[3],
-                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
-                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
-                        (unsigned int)templ_sum[2], (unsigned long long)templ_sqsum[2],
-                        (unsigned int)templ_sum[3], (unsigned long long)templ_sqsum[3],
-                        result, StreamAccessor::getStream(stream));
-                break;
-            default:
-                CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
-            }
-        }
-    }
-}
-
-
-void cv::gpu::matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method, Stream& stream)
-{
-    MatchTemplateBuf buf;
-    matchTemplate(image, templ, result, method, buf, stream);
-}
-
-
-void cv::gpu::matchTemplate(
-        const GpuMat& image, const GpuMat& templ, GpuMat& result, int method,
-        MatchTemplateBuf &buf, Stream& stream)
-{
-    CV_Assert(image.type() == templ.type());
-    CV_Assert(image.cols >= templ.cols && image.rows >= templ.rows);
-
-    typedef void (*Caller)(const GpuMat&, const GpuMat&, GpuMat&, MatchTemplateBuf&, Stream& stream);
-
-    static const Caller callers8U[] = { ::matchTemplate_SQDIFF_8U, ::matchTemplate_SQDIFF_NORMED_8U,
-                                        ::matchTemplate_CCORR_8U, ::matchTemplate_CCORR_NORMED_8U,
-                                        ::matchTemplate_CCOFF_8U, ::matchTemplate_CCOFF_NORMED_8U };
-    static const Caller callers32F[] = { ::matchTemplate_SQDIFF_32F, 0,
-                                         ::matchTemplate_CCORR_32F, 0, 0, 0 };
-
-    const Caller* callers = 0;
-    switch (image.depth())
-    {
-        case CV_8U: callers = callers8U; break;
-        case CV_32F: callers = callers32F; break;
-        default: CV_Error(CV_StsBadArg, "matchTemplate: unsupported data type");
-    }
-
-    Caller caller = callers[method];
-    CV_Assert(caller);
-    caller(image, templ, result, buf, stream);
-}
-
-#endif
diff --git a/modules/gpu/src/precomp.cpp b/modules/gpu/src/precomp.cpp
deleted file mode 100644
index 8e0d0db19f..0000000000
--- a/modules/gpu/src/precomp.cpp
+++ /dev/null
@@ -1,75 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-bool cv::gpu::tryConvertToGpuBorderType(int cpuBorderType, int& gpuBorderType)
-{
-#if !defined (HAVE_CUDA)
-    (void)cpuBorderType;
-    (void)gpuBorderType;
-#else
-    switch (cpuBorderType)
-    {
-    case cv::BORDER_REFLECT101:
-        gpuBorderType = cv::gpu::BORDER_REFLECT101_GPU;
-        return true;
-    case cv::BORDER_REPLICATE:
-        gpuBorderType = cv::gpu::BORDER_REPLICATE_GPU;
-        return true;
-    case cv::BORDER_CONSTANT:
-        gpuBorderType = cv::gpu::BORDER_CONSTANT_GPU;
-        return true;
-    case cv::BORDER_REFLECT:
-        gpuBorderType = cv::gpu::BORDER_REFLECT_GPU;
-        return true;
-    case cv::BORDER_WRAP:
-        gpuBorderType = cv::gpu::BORDER_WRAP_GPU;
-        return true;
-    default:
-        return false;
-    };
-#endif
-    return false;
-}
-
-/* End of file. */
-
diff --git a/modules/gpu/src/precomp.hpp b/modules/gpu/src/precomp.hpp
index caa46cfc8a..cfd136aa30 100644
--- a/modules/gpu/src/precomp.hpp
+++ b/modules/gpu/src/precomp.hpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -43,106 +43,18 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#if defined _MSC_VER && _MSC_VER >= 1200
-    #pragma warning( disable: 4251 4710 4711 4514 4996 )
-#endif
-
-#ifdef HAVE_CVCONFIG_H
-    #include "cvconfig.h"
-#endif
-
-#include <cstring>
-#include <iostream>
-#include <limits>
-#include <vector>
-#include <algorithm>
-#include <sstream>
-#include <exception>
-#include <iterator>
-#include <functional>
-#include <utility>
-#include <deque>
-#include <stdexcept>
-#include <memory>
-#include <string>
-
 #include "opencv2/gpu.hpp"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/gpuwarping.hpp"
 #include "opencv2/calib3d.hpp"
-#include "opencv2/core/internal.hpp"
-#include "opencv2/video.hpp"
+#include "opencv2/objdetect.hpp"
 
-#if defined WIN32 || defined WINCE
-    #include <windows.h>
-    #undef small
-    #undef min
-    #undef max
-    #undef abs
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPULEGACY
+#  include "opencv2/gpulegacy/private.hpp"
 #endif
 
-#define OPENCV_GPU_UNUSED(x) (void)x
-
-#ifdef HAVE_CUDA
-
-    #include <cuda.h>
-    #include <cuda_runtime.h>
-    #include <npp.h>
-
-    #ifdef HAVE_CUFFT
-        #include <cufft.h>
-    #endif
-
-    #ifdef HAVE_CUBLAS
-        #include <cublas.h>
-    #endif
-
-    #ifdef HAVE_NVCUVID
-        #include <nvcuvid.h>
-
-        #ifdef WIN32
-            #include <NVEncoderAPI.h>
-        #endif
-    #endif
-
-    #include "internal_shared.hpp"
-    #include "opencv2/core/stream_accessor.hpp"
-
-    #include "nvidia/core/NCV.hpp"
-    #include "nvidia/NPP_staging/NPP_staging.hpp"
-    #include "nvidia/NCVHaarObjectDetection.hpp"
-    #include "nvidia/NCVBroxOpticalFlow.hpp"
-
-    #define CUDART_MINIMUM_REQUIRED_VERSION 4010
-    #define NPP_MINIMUM_REQUIRED_VERSION 4100
-
-    #if (CUDART_VERSION < CUDART_MINIMUM_REQUIRED_VERSION)
-        #error "Insufficient Cuda Runtime library version, please update it."
-    #endif
-
-    #if (NPP_VERSION_MAJOR * 1000 + NPP_VERSION_MINOR * 100 + NPP_VERSION_BUILD < NPP_MINIMUM_REQUIRED_VERSION)
-        #error "Insufficient NPP version, please update it."
-    #endif
-
-    #if defined(CUDA_ARCH_BIN_OR_PTX_10)
-        #error "OpenCV GPU module doesn't support NVIDIA compute capability 1.0"
-    #endif
-
-    static inline void throw_nogpu() { CV_Error(CV_StsNotImplemented, "The called functionality is disabled for current build or platform"); }
-
-#else /* defined(HAVE_CUDA) */
-
-    static inline void throw_nogpu() { CV_Error(CV_GpuNotSupported, "The library is compiled without GPU support"); }
-
-#endif /* defined(HAVE_CUDA) */
-
-
-namespace cv { namespace gpu
-{
-    // Converts CPU border extrapolation mode into GPU internal analogue.
-    // Returns true if the GPU analogue exists, false otherwise.
-    bool tryConvertToGpuBorderType(int cpuBorderType, int& gpuBorderType);
-
-}}
-
 #endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/src/pyramids.cpp b/modules/gpu/src/pyramids.cpp
deleted file mode 100644
index 85fb99040c..0000000000
--- a/modules/gpu/src/pyramids.cpp
+++ /dev/null
@@ -1,249 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-void cv::gpu::pyrDown(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::pyrUp(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::ImagePyramid::build(const GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::ImagePyramid::getLayer(GpuMat&, Size, Stream&) const { throw_nogpu(); }
-
-#else // HAVE_CUDA
-
-//////////////////////////////////////////////////////////////////////////////
-// pyrDown
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T> void pyrDown_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    using namespace cv::gpu::device::imgproc;
-
-    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
-    static const func_t funcs[6][4] =
-    {
-        {pyrDown_gpu<uchar>      , 0 /*pyrDown_gpu<uchar2>*/ , pyrDown_gpu<uchar3>      , pyrDown_gpu<uchar4>      },
-        {0 /*pyrDown_gpu<schar>*/, 0 /*pyrDown_gpu<schar2>*/ , 0 /*pyrDown_gpu<schar3>*/, 0 /*pyrDown_gpu<schar4>*/},
-        {pyrDown_gpu<ushort>     , 0 /*pyrDown_gpu<ushort2>*/, pyrDown_gpu<ushort3>     , pyrDown_gpu<ushort4>     },
-        {pyrDown_gpu<short>      , 0 /*pyrDown_gpu<short2>*/ , pyrDown_gpu<short3>      , pyrDown_gpu<short4>      },
-        {0 /*pyrDown_gpu<int>*/  , 0 /*pyrDown_gpu<int2>*/   , 0 /*pyrDown_gpu<int3>*/  , 0 /*pyrDown_gpu<int4>*/  },
-        {pyrDown_gpu<float>      , 0 /*pyrDown_gpu<float2>*/ , pyrDown_gpu<float3>      , pyrDown_gpu<float4>      }
-    };
-
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-
-    const func_t func = funcs[src.depth()][src.channels() - 1];
-    CV_Assert(func != 0);
-
-    dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
-
-    func(src, dst, StreamAccessor::getStream(stream));
-}
-
-
-//////////////////////////////////////////////////////////////////////////////
-// pyrUp
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T> void pyrUp_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    using namespace cv::gpu::device::imgproc;
-
-    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
-    static const func_t funcs[6][4] =
-    {
-        {pyrUp_gpu<uchar>      , 0 /*pyrUp_gpu<uchar2>*/ , pyrUp_gpu<uchar3>      , pyrUp_gpu<uchar4>      },
-        {0 /*pyrUp_gpu<schar>*/, 0 /*pyrUp_gpu<schar2>*/ , 0 /*pyrUp_gpu<schar3>*/, 0 /*pyrUp_gpu<schar4>*/},
-        {pyrUp_gpu<ushort>     , 0 /*pyrUp_gpu<ushort2>*/, pyrUp_gpu<ushort3>     , pyrUp_gpu<ushort4>     },
-        {pyrUp_gpu<short>      , 0 /*pyrUp_gpu<short2>*/ , pyrUp_gpu<short3>      , pyrUp_gpu<short4>      },
-        {0 /*pyrUp_gpu<int>*/  , 0 /*pyrUp_gpu<int2>*/   , 0 /*pyrUp_gpu<int3>*/  , 0 /*pyrUp_gpu<int4>*/  },
-        {pyrUp_gpu<float>      , 0 /*pyrUp_gpu<float2>*/ , pyrUp_gpu<float3>      , pyrUp_gpu<float4>      }
-    };
-
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-
-    const func_t func = funcs[src.depth()][src.channels() - 1];
-    CV_Assert(func != 0);
-
-    dst.create(src.rows * 2, src.cols * 2, src.type());
-
-    func(src, dst, StreamAccessor::getStream(stream));
-}
-
-
-//////////////////////////////////////////////////////////////////////////////
-// ImagePyramid
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace pyramid
-    {
-        template <typename T> void kernelDownsampleX2_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template <typename T> void kernelInterpolateFrom1_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::ImagePyramid::build(const GpuMat& img, int numLayers, Stream& stream)
-{
-    using namespace cv::gpu::device::pyramid;
-
-    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
-    static const func_t funcs[6][4] =
-    {
-        {kernelDownsampleX2_gpu<uchar1>       , 0 /*kernelDownsampleX2_gpu<uchar2>*/ , kernelDownsampleX2_gpu<uchar3>      , kernelDownsampleX2_gpu<uchar4>      },
-        {0 /*kernelDownsampleX2_gpu<char1>*/  , 0 /*kernelDownsampleX2_gpu<char2>*/  , 0 /*kernelDownsampleX2_gpu<char3>*/ , 0 /*kernelDownsampleX2_gpu<char4>*/ },
-        {kernelDownsampleX2_gpu<ushort1>      , 0 /*kernelDownsampleX2_gpu<ushort2>*/, kernelDownsampleX2_gpu<ushort3>     , kernelDownsampleX2_gpu<ushort4>     },
-        {0 /*kernelDownsampleX2_gpu<short1>*/ , 0 /*kernelDownsampleX2_gpu<short2>*/ , 0 /*kernelDownsampleX2_gpu<short3>*/, 0 /*kernelDownsampleX2_gpu<short4>*/},
-        {0 /*kernelDownsampleX2_gpu<int1>*/   , 0 /*kernelDownsampleX2_gpu<int2>*/   , 0 /*kernelDownsampleX2_gpu<int3>*/  , 0 /*kernelDownsampleX2_gpu<int4>*/  },
-        {kernelDownsampleX2_gpu<float1>       , 0 /*kernelDownsampleX2_gpu<float2>*/ , kernelDownsampleX2_gpu<float3>      , kernelDownsampleX2_gpu<float4>      }
-    };
-
-    CV_Assert(img.depth() <= CV_32F && img.channels() <= 4);
-
-    const func_t func = funcs[img.depth()][img.channels() - 1];
-    CV_Assert(func != 0);
-
-    layer0_ = img;
-    Size szLastLayer = img.size();
-    nLayers_ = 1;
-
-    if (numLayers <= 0)
-        numLayers = 255; //it will cut-off when any of the dimensions goes 1
-
-    pyramid_.resize(numLayers);
-
-    for (int i = 0; i < numLayers - 1; ++i)
-    {
-        Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
-
-        if (szCurLayer.width == 0 || szCurLayer.height == 0)
-            break;
-
-        ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
-        nLayers_++;
-
-        const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
-
-        func(prevLayer, pyramid_[i], StreamAccessor::getStream(stream));
-
-        szLastLayer = szCurLayer;
-    }
-}
-
-void cv::gpu::ImagePyramid::getLayer(GpuMat& outImg, Size outRoi, Stream& stream) const
-{
-    using namespace cv::gpu::device::pyramid;
-
-    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
-    static const func_t funcs[6][4] =
-    {
-        {kernelInterpolateFrom1_gpu<uchar1>      , 0 /*kernelInterpolateFrom1_gpu<uchar2>*/ , kernelInterpolateFrom1_gpu<uchar3>      , kernelInterpolateFrom1_gpu<uchar4>      },
-        {0 /*kernelInterpolateFrom1_gpu<char1>*/ , 0 /*kernelInterpolateFrom1_gpu<char2>*/  , 0 /*kernelInterpolateFrom1_gpu<char3>*/ , 0 /*kernelInterpolateFrom1_gpu<char4>*/ },
-        {kernelInterpolateFrom1_gpu<ushort1>     , 0 /*kernelInterpolateFrom1_gpu<ushort2>*/, kernelInterpolateFrom1_gpu<ushort3>     , kernelInterpolateFrom1_gpu<ushort4>     },
-        {0 /*kernelInterpolateFrom1_gpu<short1>*/, 0 /*kernelInterpolateFrom1_gpu<short2>*/ , 0 /*kernelInterpolateFrom1_gpu<short3>*/, 0 /*kernelInterpolateFrom1_gpu<short4>*/},
-        {0 /*kernelInterpolateFrom1_gpu<int1>*/  , 0 /*kernelInterpolateFrom1_gpu<int2>*/   , 0 /*kernelInterpolateFrom1_gpu<int3>*/  , 0 /*kernelInterpolateFrom1_gpu<int4>*/  },
-        {kernelInterpolateFrom1_gpu<float1>      , 0 /*kernelInterpolateFrom1_gpu<float2>*/ , kernelInterpolateFrom1_gpu<float3>      , kernelInterpolateFrom1_gpu<float4>      }
-    };
-
-    CV_Assert(outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0);
-
-    ensureSizeIsEnough(outRoi, layer0_.type(), outImg);
-
-    const func_t func = funcs[outImg.depth()][outImg.channels() - 1];
-    CV_Assert(func != 0);
-
-    if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
-    {
-        if (stream)
-            stream.enqueueCopy(layer0_, outImg);
-        else
-            layer0_.copyTo(outImg);
-    }
-
-    float lastScale = 1.0f;
-    float curScale;
-    GpuMat lastLayer = layer0_;
-    GpuMat curLayer;
-
-    for (int i = 0; i < nLayers_ - 1; ++i)
-    {
-        curScale = lastScale * 0.5f;
-        curLayer = pyramid_[i];
-
-        if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
-        {
-            if (stream)
-                stream.enqueueCopy(curLayer, outImg);
-            else
-                curLayer.copyTo(outImg);
-        }
-
-        if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
-            break;
-
-        lastScale = curScale;
-        lastLayer = curLayer;
-    }
-
-    func(lastLayer, outImg, StreamAccessor::getStream(stream));
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/resize.cpp b/modules/gpu/src/resize.cpp
deleted file mode 100644
index 6850135428..0000000000
--- a/modules/gpu/src/resize.cpp
+++ /dev/null
@@ -1,162 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-
-void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation, Stream& s)
-{
-    (void)src;
-    (void)dst;
-    (void)dsize;
-    (void)fx;
-    (void)fy;
-    (void)interpolation;
-    (void)s;
-
-    throw_nogpu();
-}
-
-#else // HAVE_CUDA
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace imgproc
-    {
-        template <typename T>
-        void resize_gpu(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy,
-                        PtrStepSzb dst, int interpolation, cudaStream_t stream);
-    }
-}}}
-
-void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation, Stream& s)
-{
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR
-            || interpolation == INTER_CUBIC || interpolation == INTER_AREA);
-    CV_Assert(!(dsize == Size()) || (fx > 0 && fy > 0));
-
-    if (dsize == Size())
-        dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));
-    else
-    {
-        fx = static_cast<double>(dsize.width) / src.cols;
-        fy = static_cast<double>(dsize.height) / src.rows;
-    }
-    if (dsize != dst.size())
-        dst.create(dsize, src.type());
-
-    if (dsize == src.size())
-    {
-        if (s)
-            s.enqueueCopy(src, dst);
-        else
-            src.copyTo(dst);
-        return;
-    }
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    Size wholeSize;
-    Point ofs;
-    src.locateROI(wholeSize, ofs);
-
-    bool useNpp = (src.type() == CV_8UC1 || src.type() == CV_8UC4);
-    useNpp = useNpp && (interpolation == INTER_NEAREST || interpolation == INTER_LINEAR);
-
-    if (useNpp)
-    {
-        typedef NppStatus (*func_t)(const Npp8u * pSrc, NppiSize oSrcSize, int nSrcStep, NppiRect oSrcROI, Npp8u * pDst, int nDstStep, NppiSize dstROISize,
-                                    double xFactor, double yFactor, int eInterpolation);
-
-        const func_t funcs[4] = { nppiResize_8u_C1R, 0, 0, nppiResize_8u_C4R };
-
-        static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC, 0, NPPI_INTER_LANCZOS};
-
-        NppiSize srcsz;
-        srcsz.width  = wholeSize.width;
-        srcsz.height = wholeSize.height;
-
-        NppiRect srcrect;
-        srcrect.x = ofs.x;
-        srcrect.y = ofs.y;
-        srcrect.width  = src.cols;
-        srcrect.height = src.rows;
-
-        NppiSize dstsz;
-        dstsz.width  = dst.cols;
-        dstsz.height = dst.rows;
-
-        NppStreamHandler h(stream);
-
-        nppSafeCall( funcs[src.channels() - 1](src.datastart, srcsz, static_cast<int>(src.step), srcrect,
-                dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, fx, fy, npp_inter[interpolation]) );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-    else
-    {
-        using namespace ::cv::gpu::device::imgproc;
-
-        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float fx, float fy, PtrStepSzb dst, int interpolation, cudaStream_t stream);
-
-        static const func_t funcs[6][4] =
-        {
-            {resize_gpu<uchar>      , 0 /*resize_gpu<uchar2>*/ , resize_gpu<uchar3>     , resize_gpu<uchar4>     },
-            {0 /*resize_gpu<schar>*/, 0 /*resize_gpu<char2>*/  , 0 /*resize_gpu<char3>*/, 0 /*resize_gpu<char4>*/},
-            {resize_gpu<ushort>     , 0 /*resize_gpu<ushort2>*/, resize_gpu<ushort3>    , resize_gpu<ushort4>    },
-            {resize_gpu<short>      , 0 /*resize_gpu<short2>*/ , resize_gpu<short3>     , resize_gpu<short4>     },
-            {0 /*resize_gpu<int>*/  , 0 /*resize_gpu<int2>*/   , 0 /*resize_gpu<int3>*/ , 0 /*resize_gpu<int4>*/ },
-            {resize_gpu<float>      , 0 /*resize_gpu<float2>*/ , resize_gpu<float3>     , resize_gpu<float4>     }
-        };
-
-        const func_t func = funcs[src.depth()][src.channels() - 1];
-        CV_Assert(func != 0);
-
-        func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y,
-            static_cast<float>(1.0 / fx), static_cast<float>(1.0 / fy), dst, interpolation, stream);
-    }
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/speckle_filtering.cpp b/modules/gpu/src/speckle_filtering.cpp
deleted file mode 100644
index 571cbf657b..0000000000
--- a/modules/gpu/src/speckle_filtering.cpp
+++ /dev/null
@@ -1,168 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-
-namespace cv {
-// TODO: conflicts with calib3d.hpp : filterSpeckles, should be removed ?
-
-//! Speckle filtering - filters small connected components on diparity image.
-//! It sets pixel (x,y) to newVal if it coresponds to small CC with size < maxSpeckleSize.
-//! Threshold for border between CC is diffThreshold;
-  CV_EXPORTS void filterSpeckles(Mat& img, uchar newVal, int maxSpeckleSize, uchar diffThreshold, Mat& buf);
-}
-
-typedef Point_<short> Point2s;
-
-void cv::filterSpeckles( Mat& img, uchar newVal, int maxSpeckleSize, uchar maxDiff, Mat& _buf)
-{
-    int MaxD = 1024;
-    int WinSz = 64;
-
-    int bufSize0 = (MaxD + 2)*sizeof(int) + (img.rows+WinSz+2)*MaxD*sizeof(int) +
-        (img.rows + WinSz + 2)*sizeof(int) +
-        (img.rows+WinSz+2)*MaxD*(WinSz+1)*sizeof(uchar) + 256;
-    int bufSize1 = (img.cols + 9 + 2) * sizeof(int) + 256;
-    int bufSz = std::max(bufSize0 * 1, bufSize1 * 2);
-
-    _buf.create(1, bufSz, CV_8U);
-
-    CV_Assert( img.type() == CV_8U );
-
-    int width = img.cols, height = img.rows, npixels = width*height;
-    size_t bufSize = npixels*(int)(sizeof(Point2s) + sizeof(int) + sizeof(uchar));
-    if( !_buf.isContinuous() || !_buf.data || _buf.cols*_buf.rows*_buf.elemSize() < bufSize )
-        _buf.create(1, (int)bufSize, CV_8U);
-
-    uchar* buf = _buf.data;
-    int i, j, dstep = (int)(img.step/sizeof(uchar));
-    int* labels = (int*)buf;
-    buf += npixels*sizeof(labels[0]);
-    Point2s* wbuf = (Point2s*)buf;
-    buf += npixels*sizeof(wbuf[0]);
-    uchar* rtype = (uchar*)buf;
-    int curlabel = 0;
-
-    // clear out label assignments
-    memset(labels, 0, npixels*sizeof(labels[0]));
-
-    for( i = 0; i < height; i++ )
-    {
-        uchar* ds = img.ptr<uchar>(i);
-        int* ls = labels + width*i;
-
-        for( j = 0; j < width; j++ )
-        {
-            if( ds[j] != newVal )	// not a bad disparity
-            {
-                if( ls[j] )		// has a label, check for bad label
-                {
-                    if( rtype[ls[j]] ) // small region, zero out disparity
-                        ds[j] = (uchar)newVal;
-                }
-                // no label, assign and propagate
-                else
-                {
-                    Point2s* ws = wbuf;	// initialize wavefront
-                    Point2s p((short)j, (short)i);	// current pixel
-                    curlabel++;	// next label
-                    int count = 0;	// current region size
-                    ls[j] = curlabel;
-
-                    // wavefront propagation
-                    while( ws >= wbuf ) // wavefront not empty
-                    {
-                        count++;
-                        // put neighbors onto wavefront
-                        uchar* dpp = &img.at<uchar>(p.y, p.x);
-                        uchar dp = *dpp;
-                        int* lpp = labels + width*p.y + p.x;
-
-                        if( p.x < width-1 && !lpp[+1] && dpp[+1] != newVal && std::abs(dp - dpp[+1]) <= maxDiff )
-                        {
-                            lpp[+1] = curlabel;
-                            *ws++ = Point2s(p.x+1, p.y);
-                        }
-
-                        if( p.x > 0 && !lpp[-1] && dpp[-1] != newVal && std::abs(dp - dpp[-1]) <= maxDiff )
-                        {
-                            lpp[-1] = curlabel;
-                            *ws++ = Point2s(p.x-1, p.y);
-                        }
-
-                        if( p.y < height-1 && !lpp[+width] && dpp[+dstep] != newVal && std::abs(dp - dpp[+dstep]) <= maxDiff )
-                        {
-                            lpp[+width] = curlabel;
-                            *ws++ = Point2s(p.x, p.y+1);
-                        }
-
-                        if( p.y > 0 && !lpp[-width] && dpp[-dstep] != newVal && std::abs(dp - dpp[-dstep]) <= maxDiff )
-                        {
-                            lpp[-width] = curlabel;
-                            *ws++ = Point2s(p.x, p.y-1);
-                        }
-
-                        // pop most recent and propagate
-                        // NB: could try least recent, maybe better convergence
-                        p = *--ws;
-                    }
-
-                    // assign label type
-                    if( count <= maxSpeckleSize )	// speckle region
-                    {
-                        //printf("count = %d\n", count);
-                        rtype[ls[j]] = 1;	// small region label
-                        ds[j] = (uchar)newVal;
-                    }
-                    else
-                    {
-                        //printf("count = %d\n", count);
-                        rtype[ls[j]] = 0;	// large region label
-                    }
-                }
-            }
-        }
-    }
-}
-
diff --git a/modules/gpu/src/split_merge.cpp b/modules/gpu/src/split_merge.cpp
deleted file mode 100644
index 20631e0ec7..0000000000
--- a/modules/gpu/src/split_merge.cpp
+++ /dev/null
@@ -1,171 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::merge(const GpuMat* /*src*/, size_t /*count*/, GpuMat& /*dst*/, Stream& /*stream*/) { throw_nogpu(); }
-void cv::gpu::merge(const std::vector<GpuMat>& /*src*/, GpuMat& /*dst*/, Stream& /*stream*/) { throw_nogpu(); }
-void cv::gpu::split(const GpuMat& /*src*/, GpuMat* /*dst*/, Stream& /*stream*/) { throw_nogpu(); }
-void cv::gpu::split(const GpuMat& /*src*/, std::vector<GpuMat>& /*dst*/, Stream& /*stream*/) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace split_merge
-    {
-        void merge_caller(const PtrStepSzb* src, PtrStepSzb& dst, int total_channels, size_t elem_size, const cudaStream_t& stream);
-        void split_caller(const PtrStepSzb& src, PtrStepSzb* dst, int num_channels, size_t elem_size1, const cudaStream_t& stream);
-    }
-}}}
-
-namespace
-{
-    void merge(const GpuMat* src, size_t n, GpuMat& dst, const cudaStream_t& stream)
-    {
-        using namespace ::cv::gpu::device::split_merge;
-
-        CV_Assert(src);
-        CV_Assert(n > 0);
-
-        int depth = src[0].depth();
-        Size size = src[0].size();
-
-        if (depth == CV_64F)
-        {
-            if (!deviceSupports(NATIVE_DOUBLE))
-                CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-        }
-
-        bool single_channel_only = true;
-        int total_channels = 0;
-
-        for (size_t i = 0; i < n; ++i)
-        {
-            CV_Assert(src[i].size() == size);
-            CV_Assert(src[i].depth() == depth);
-            single_channel_only = single_channel_only && src[i].channels() == 1;
-            total_channels += src[i].channels();
-        }
-
-        CV_Assert(single_channel_only);
-        CV_Assert(total_channels <= 4);
-
-        if (total_channels == 1)
-            src[0].copyTo(dst);
-        else
-        {
-            dst.create(size, CV_MAKETYPE(depth, total_channels));
-
-            PtrStepSzb src_as_devmem[4];
-            for(size_t i = 0; i < n; ++i)
-                src_as_devmem[i] = src[i];
-
-            PtrStepSzb dst_as_devmem(dst);
-            merge_caller(src_as_devmem, dst_as_devmem, total_channels, CV_ELEM_SIZE(depth), stream);
-        }
-    }
-
-    void split(const GpuMat& src, GpuMat* dst, const cudaStream_t& stream)
-    {
-        using namespace ::cv::gpu::device::split_merge;
-
-        CV_Assert(dst);
-
-        int depth = src.depth();
-        int num_channels = src.channels();
-
-        if (depth == CV_64F)
-        {
-            if (!deviceSupports(NATIVE_DOUBLE))
-                CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-        }
-
-        if (num_channels == 1)
-        {
-            src.copyTo(dst[0]);
-            return;
-        }
-
-        for (int i = 0; i < num_channels; ++i)
-            dst[i].create(src.size(), depth);
-
-        CV_Assert(num_channels <= 4);
-
-        PtrStepSzb dst_as_devmem[4];
-        for (int i = 0; i < num_channels; ++i)
-            dst_as_devmem[i] = dst[i];
-
-        PtrStepSzb src_as_devmem(src);
-        split_caller(src_as_devmem, dst_as_devmem, num_channels, src.elemSize1(), stream);
-    }
-}
-
-void cv::gpu::merge(const GpuMat* src, size_t n, GpuMat& dst, Stream& stream)
-{
-    ::merge(src, n, dst, StreamAccessor::getStream(stream));
-}
-
-
-void cv::gpu::merge(const std::vector<GpuMat>& src, GpuMat& dst, Stream& stream)
-{
-    ::merge(&src[0], src.size(), dst, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::split(const GpuMat& src, GpuMat* dst, Stream& stream)
-{
-    ::split(src, dst, StreamAccessor::getStream(stream));
-}
-
-void cv::gpu::split(const GpuMat& src, std::vector<GpuMat>& dst, Stream& stream)
-{
-    dst.resize(src.channels());
-    if(src.channels() > 0)
-        ::split(src, &dst[0], StreamAccessor::getStream(stream));
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/stereobm.cpp b/modules/gpu/src/stereobm.cpp
deleted file mode 100644
index 43a5af17d7..0000000000
--- a/modules/gpu/src/stereobm.cpp
+++ /dev/null
@@ -1,140 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-cv::gpu::StereoBM_GPU::StereoBM_GPU() { throw_nogpu(); }
-cv::gpu::StereoBM_GPU::StereoBM_GPU(int, int, int) { throw_nogpu(); }
-
-bool cv::gpu::StereoBM_GPU::checkIfGpuCallReasonable() { throw_nogpu(); return false; }
-void cv::gpu::StereoBM_GPU::operator() ( const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace stereobm
-    {
-        void stereoBM_GPU(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& disp, int ndisp, int winsz, const PtrStepSz<unsigned int>& minSSD_buf, cudaStream_t & stream);
-        void prefilter_xsobel(const PtrStepSzb& input, const PtrStepSzb& output, int prefilterCap /*= 31*/, cudaStream_t & stream);
-        void postfilter_textureness(const PtrStepSzb& input, int winsz, float avgTexturenessThreshold, const PtrStepSzb& disp, cudaStream_t & stream);
-    }
-}}}
-
-namespace
-{
-    const float defaultAvgTexThreshold = 3;
-}
-
-cv::gpu::StereoBM_GPU::StereoBM_GPU()
-    : preset(BASIC_PRESET), ndisp(DEFAULT_NDISP), winSize(DEFAULT_WINSZ), avergeTexThreshold(defaultAvgTexThreshold)
-{
-}
-
-cv::gpu::StereoBM_GPU::StereoBM_GPU(int preset_, int ndisparities_, int winSize_)
-    : preset(preset_), ndisp(ndisparities_), winSize(winSize_), avergeTexThreshold(defaultAvgTexThreshold)
-{
-    const int max_supported_ndisp = 1 << (sizeof(unsigned char) * 8);
-    CV_Assert(0 < ndisp && ndisp <= max_supported_ndisp);
-    CV_Assert(ndisp % 8 == 0);
-    CV_Assert(winSize % 2 == 1);
-}
-
-bool cv::gpu::StereoBM_GPU::checkIfGpuCallReasonable()
-{
-    if (0 == getCudaEnabledDeviceCount())
-        return false;
-
-    DeviceInfo device_info;
-
-    if (device_info.majorVersion() > 1 || device_info.multiProcessorCount() > 16)
-        return true;
-
-    return false;
-}
-
-namespace
-{
-    void stereo_bm_gpu_operator( GpuMat& minSSD,  GpuMat& leBuf, GpuMat&  riBuf,  int preset, int ndisp, int winSize, float avergeTexThreshold, const GpuMat& left, const GpuMat& right, GpuMat& disparity, cudaStream_t stream)
-    {
-        using namespace ::cv::gpu::device::stereobm;
-
-        CV_Assert(left.rows == right.rows && left.cols == right.cols);
-        CV_Assert(left.type() == CV_8UC1);
-        CV_Assert(right.type() == CV_8UC1);
-
-        disparity.create(left.size(), CV_8U);
-        minSSD.create(left.size(), CV_32S);
-
-        GpuMat le_for_bm =  left;
-        GpuMat ri_for_bm = right;
-
-        if (preset == StereoBM_GPU::PREFILTER_XSOBEL)
-        {
-            leBuf.create( left.size(),  left.type());
-            riBuf.create(right.size(), right.type());
-
-            prefilter_xsobel( left, leBuf, 31, stream);
-            prefilter_xsobel(right, riBuf, 31, stream);
-
-            le_for_bm = leBuf;
-            ri_for_bm = riBuf;
-        }
-
-        stereoBM_GPU(le_for_bm, ri_for_bm, disparity, ndisp, winSize, minSSD, stream);
-
-        if (avergeTexThreshold)
-            postfilter_textureness(le_for_bm, winSize, avergeTexThreshold, disparity, stream);
-    }
-}
-
-void cv::gpu::StereoBM_GPU::operator() ( const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream)
-{
-    stereo_bm_gpu_operator(minSSD, leBuf, riBuf, preset, ndisp, winSize, avergeTexThreshold, left, right, disparity, StreamAccessor::getStream(stream));
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/stereobp.cpp b/modules/gpu/src/stereobp.cpp
deleted file mode 100644
index 210158d071..0000000000
--- a/modules/gpu/src/stereobp.cpp
+++ /dev/null
@@ -1,368 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::StereoBeliefPropagation::estimateRecommendedParams(int, int, int&, int&, int&) { throw_nogpu(); }
-
-cv::gpu::StereoBeliefPropagation::StereoBeliefPropagation(int, int, int, int) { throw_nogpu(); }
-cv::gpu::StereoBeliefPropagation::StereoBeliefPropagation(int, int, int, float, float, float, float, int) { throw_nogpu(); }
-
-void cv::gpu::StereoBeliefPropagation::operator()(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-void cv::gpu::StereoBeliefPropagation::operator()(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace stereobp
-    {
-        void load_constants(int ndisp, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump);
-        template<typename T, typename D>
-        void comp_data_gpu(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& data, cudaStream_t stream);
-        template<typename T>
-        void data_step_down_gpu(int dst_cols, int dst_rows, int src_rows, const PtrStepSzb& src, const PtrStepSzb& dst, cudaStream_t stream);
-        template <typename T>
-        void level_up_messages_gpu(int dst_idx, int dst_cols, int dst_rows, int src_rows, PtrStepSzb* mus, PtrStepSzb* mds, PtrStepSzb* mls, PtrStepSzb* mrs, cudaStream_t stream);
-        template <typename T>
-        void calc_all_iterations_gpu(int cols, int rows, int iters, const PtrStepSzb& u, const PtrStepSzb& d,
-            const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, cudaStream_t stream);
-        template <typename T>
-        void output_gpu(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data,
-            const PtrStepSz<short>& disp, cudaStream_t stream);
-    }
-}}}
-
-using namespace ::cv::gpu::device::stereobp;
-
-namespace
-{
-    const float DEFAULT_MAX_DATA_TERM = 10.0f;
-    const float DEFAULT_DATA_WEIGHT = 0.07f;
-    const float DEFAULT_MAX_DISC_TERM = 1.7f;
-    const float DEFAULT_DISC_SINGLE_JUMP = 1.0f;
-}
-
-void cv::gpu::StereoBeliefPropagation::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels)
-{
-    ndisp = width / 4;
-    if ((ndisp & 1) != 0)
-        ndisp++;
-
-    int mm = std::max(width, height);
-    iters = mm / 100 + 2;
-
-    levels = (int)(::log(static_cast<double>(mm)) + 1) * 4 / 5;
-    if (levels == 0) levels++;
-}
-
-cv::gpu::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp_, int iters_, int levels_, int msg_type_)
-    : ndisp(ndisp_), iters(iters_), levels(levels_),
-      max_data_term(DEFAULT_MAX_DATA_TERM), data_weight(DEFAULT_DATA_WEIGHT),
-      max_disc_term(DEFAULT_MAX_DISC_TERM), disc_single_jump(DEFAULT_DISC_SINGLE_JUMP),
-      msg_type(msg_type_), datas(levels_)
-{
-}
-
-cv::gpu::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp_, int iters_, int levels_, float max_data_term_, float data_weight_, float max_disc_term_, float disc_single_jump_, int msg_type_)
-    : ndisp(ndisp_), iters(iters_), levels(levels_),
-      max_data_term(max_data_term_), data_weight(data_weight_),
-      max_disc_term(max_disc_term_), disc_single_jump(disc_single_jump_),
-      msg_type(msg_type_), datas(levels_)
-{
-}
-
-namespace
-{
-    class StereoBeliefPropagationImpl
-    {
-    public:
-        StereoBeliefPropagationImpl(StereoBeliefPropagation& rthis_,
-                                    GpuMat& u_, GpuMat& d_, GpuMat& l_, GpuMat& r_,
-                                    GpuMat& u2_, GpuMat& d2_, GpuMat& l2_, GpuMat& r2_,
-                                    std::vector<GpuMat>& datas_, GpuMat& out_)
-            : rthis(rthis_), u(u_), d(d_), l(l_), r(r_), u2(u2_), d2(d2_), l2(l2_), r2(r2_), datas(datas_), out(out_),
-              zero(Scalar::all(0)), scale(rthis_.msg_type == CV_32F ? 1.0f : 10.0f)
-        {
-            CV_Assert(0 < rthis.ndisp && 0 < rthis.iters && 0 < rthis.levels);
-            CV_Assert(rthis.msg_type == CV_32F || rthis.msg_type == CV_16S);
-            CV_Assert(rthis.msg_type == CV_32F || (1 << (rthis.levels - 1)) * scale * rthis.max_data_term < std::numeric_limits<short>::max());
-        }
-
-        void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disp, Stream& stream)
-        {
-            typedef void (*comp_data_t)(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& data, cudaStream_t stream);
-            static const comp_data_t comp_data_callers[2][5] =
-            {
-                {0, comp_data_gpu<unsigned char, short>, 0, comp_data_gpu<uchar3, short>, comp_data_gpu<uchar4, short>},
-                {0, comp_data_gpu<unsigned char, float>, 0, comp_data_gpu<uchar3, float>, comp_data_gpu<uchar4, float>}
-            };
-
-            CV_Assert(left.size() == right.size() && left.type() == right.type());
-            CV_Assert(left.type() == CV_8UC1 || left.type() == CV_8UC3 || left.type() == CV_8UC4);
-
-            rows = left.rows;
-            cols = left.cols;
-
-            int divisor = (int)pow(2.f, rthis.levels - 1.0f);
-            int lowest_cols = cols / divisor;
-            int lowest_rows = rows / divisor;
-            const int min_image_dim_size = 2;
-            CV_Assert(std::min(lowest_cols, lowest_rows) > min_image_dim_size);
-
-            init(stream);
-
-            datas[0].create(rows * rthis.ndisp, cols, rthis.msg_type);
-
-            comp_data_callers[rthis.msg_type == CV_32F][left.channels()](left, right, datas[0], StreamAccessor::getStream(stream));
-
-            calcBP(disp, stream);
-        }
-
-        void operator()(const GpuMat& data, GpuMat& disp, Stream& stream)
-        {
-            CV_Assert((data.type() == rthis.msg_type) && (data.rows % rthis.ndisp == 0));
-
-            rows = data.rows / rthis.ndisp;
-            cols = data.cols;
-
-            int divisor = (int)pow(2.f, rthis.levels - 1.0f);
-            int lowest_cols = cols / divisor;
-            int lowest_rows = rows / divisor;
-            const int min_image_dim_size = 2;
-            CV_Assert(std::min(lowest_cols, lowest_rows) > min_image_dim_size);
-
-            init(stream);
-
-            datas[0] = data;
-
-            calcBP(disp, stream);
-        }
-    private:
-        void init(Stream& stream)
-        {
-            u.create(rows * rthis.ndisp, cols, rthis.msg_type);
-            d.create(rows * rthis.ndisp, cols, rthis.msg_type);
-            l.create(rows * rthis.ndisp, cols, rthis.msg_type);
-            r.create(rows * rthis.ndisp, cols, rthis.msg_type);
-
-            if (rthis.levels & 1)
-            {
-                //can clear less area
-                if (stream)
-                {
-                    stream.enqueueMemSet(u, zero);
-                    stream.enqueueMemSet(d, zero);
-                    stream.enqueueMemSet(l, zero);
-                    stream.enqueueMemSet(r, zero);
-                }
-                else
-                {
-                    u.setTo(zero);
-                    d.setTo(zero);
-                    l.setTo(zero);
-                    r.setTo(zero);
-                }
-            }
-
-            if (rthis.levels > 1)
-            {
-                int less_rows = (rows + 1) / 2;
-                int less_cols = (cols + 1) / 2;
-
-                u2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
-                d2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
-                l2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
-                r2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
-
-                if ((rthis.levels & 1) == 0)
-                {
-                    if (stream)
-                    {
-                        stream.enqueueMemSet(u2, zero);
-                        stream.enqueueMemSet(d2, zero);
-                        stream.enqueueMemSet(l2, zero);
-                        stream.enqueueMemSet(r2, zero);
-                    }
-                    else
-                    {
-                        u2.setTo(zero);
-                        d2.setTo(zero);
-                        l2.setTo(zero);
-                        r2.setTo(zero);
-                    }
-                }
-            }
-
-            load_constants(rthis.ndisp, rthis.max_data_term, scale * rthis.data_weight, scale * rthis.max_disc_term, scale * rthis.disc_single_jump);
-
-            datas.resize(rthis.levels);
-
-            cols_all.resize(rthis.levels);
-            rows_all.resize(rthis.levels);
-
-            cols_all[0] = cols;
-            rows_all[0] = rows;
-        }
-
-        void calcBP(GpuMat& disp, Stream& stream)
-        {
-            typedef void (*data_step_down_t)(int dst_cols, int dst_rows, int src_rows, const PtrStepSzb& src, const PtrStepSzb& dst, cudaStream_t stream);
-            static const data_step_down_t data_step_down_callers[2] =
-            {
-                data_step_down_gpu<short>, data_step_down_gpu<float>
-            };
-
-            typedef void (*level_up_messages_t)(int dst_idx, int dst_cols, int dst_rows, int src_rows, PtrStepSzb* mus, PtrStepSzb* mds, PtrStepSzb* mls, PtrStepSzb* mrs, cudaStream_t stream);
-            static const level_up_messages_t level_up_messages_callers[2] =
-            {
-                level_up_messages_gpu<short>, level_up_messages_gpu<float>
-            };
-
-            typedef void (*calc_all_iterations_t)(int cols, int rows, int iters, const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, cudaStream_t stream);
-            static const calc_all_iterations_t calc_all_iterations_callers[2] =
-            {
-                calc_all_iterations_gpu<short>, calc_all_iterations_gpu<float>
-            };
-
-            typedef void (*output_t)(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, const PtrStepSz<short>& disp, cudaStream_t stream);
-            static const output_t output_callers[2] =
-            {
-                output_gpu<short>, output_gpu<float>
-            };
-
-            const int funcIdx = rthis.msg_type == CV_32F;
-
-            cudaStream_t cudaStream = StreamAccessor::getStream(stream);
-
-            for (int i = 1; i < rthis.levels; ++i)
-            {
-                cols_all[i] = (cols_all[i-1] + 1) / 2;
-                rows_all[i] = (rows_all[i-1] + 1) / 2;
-
-                datas[i].create(rows_all[i] * rthis.ndisp, cols_all[i], rthis.msg_type);
-
-                data_step_down_callers[funcIdx](cols_all[i], rows_all[i], rows_all[i-1], datas[i-1], datas[i], cudaStream);
-            }
-
-            PtrStepSzb mus[] = {u, u2};
-            PtrStepSzb mds[] = {d, d2};
-            PtrStepSzb mrs[] = {r, r2};
-            PtrStepSzb mls[] = {l, l2};
-
-            int mem_idx = (rthis.levels & 1) ? 0 : 1;
-
-            for (int i = rthis.levels - 1; i >= 0; --i)
-            {
-                // for lower level we have already computed messages by setting to zero
-                if (i != rthis.levels - 1)
-                    level_up_messages_callers[funcIdx](mem_idx, cols_all[i], rows_all[i], rows_all[i+1], mus, mds, mls, mrs, cudaStream);
-
-                calc_all_iterations_callers[funcIdx](cols_all[i], rows_all[i], rthis.iters, mus[mem_idx], mds[mem_idx], mls[mem_idx], mrs[mem_idx], datas[i], cudaStream);
-
-                mem_idx = (mem_idx + 1) & 1;
-            }
-
-            if (disp.empty())
-                disp.create(rows, cols, CV_16S);
-
-            out = ((disp.type() == CV_16S) ? disp : (out.create(rows, cols, CV_16S), out));
-
-            if (stream)
-                stream.enqueueMemSet(out, zero);
-            else
-                out.setTo(zero);
-
-            output_callers[funcIdx](u, d, l, r, datas.front(), out, cudaStream);
-
-            if (disp.type() != CV_16S)
-            {
-                if (stream)
-                    stream.enqueueConvert(out, disp, disp.type());
-                else
-                    out.convertTo(disp, disp.type());
-            }
-        }
-
-        StereoBeliefPropagation& rthis;
-
-        GpuMat& u;
-        GpuMat& d;
-        GpuMat& l;
-        GpuMat& r;
-
-        GpuMat& u2;
-        GpuMat& d2;
-        GpuMat& l2;
-        GpuMat& r2;
-
-        std::vector<GpuMat>& datas;
-        GpuMat& out;
-
-        const Scalar zero;
-        const float scale;
-
-        int rows, cols;
-
-        std::vector<int> cols_all, rows_all;
-    };
-}
-
-void cv::gpu::StereoBeliefPropagation::operator()(const GpuMat& left, const GpuMat& right, GpuMat& disp, Stream& stream)
-{
-    StereoBeliefPropagationImpl impl(*this, u, d, l, r, u2, d2, l2, r2, datas, out);
-    impl(left, right, disp, stream);
-}
-
-void cv::gpu::StereoBeliefPropagation::operator()(const GpuMat& data, GpuMat& disp, Stream& stream)
-{
-    StereoBeliefPropagationImpl impl(*this, u, d, l, r, u2, d2, l2, r2, datas, out);
-    impl(data, disp, stream);
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/stereocsbp.cpp b/modules/gpu/src/stereocsbp.cpp
deleted file mode 100644
index 797184f1af..0000000000
--- a/modules/gpu/src/stereocsbp.cpp
+++ /dev/null
@@ -1,310 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-using namespace cv;
-using namespace cv::gpu;
-
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::StereoConstantSpaceBP::estimateRecommendedParams(int, int, int&, int&, int&, int&) { throw_nogpu(); }
-
-cv::gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int, int, int, int, int) { throw_nogpu(); }
-cv::gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int, int, int, int, float, float, float, float, int, int) { throw_nogpu(); }
-
-void cv::gpu::StereoConstantSpaceBP::operator()(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-#else /* !defined (HAVE_CUDA) */
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace stereocsbp
-    {
-        void load_constants(int ndisp, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump, int min_disp_th,
-            const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& temp);
-
-        template<class T>
-        void init_data_cost(int rows, int cols, T* disp_selected_pyr, T* data_cost_selected, size_t msg_step,
-                    int h, int w, int level, int nr_plane, int ndisp, int channels, bool use_local_init_data_cost, cudaStream_t stream);
-
-        template<class T>
-        void compute_data_cost(const T* disp_selected_pyr, T* data_cost, size_t msg_step,
-                               int rows, int cols, int h, int w, int h2, int level, int nr_plane, int channels, cudaStream_t stream);
-
-        template<class T>
-        void init_message(T* u_new, T* d_new, T* l_new, T* r_new,
-                          const T* u_cur, const T* d_cur, const T* l_cur, const T* r_cur,
-                          T* selected_disp_pyr_new, const T* selected_disp_pyr_cur,
-                          T* data_cost_selected, const T* data_cost, size_t msg_step,
-                          int h, int w, int nr_plane, int h2, int w2, int nr_plane2, cudaStream_t stream);
-
-        template<class T>
-        void calc_all_iterations(T* u, T* d, T* l, T* r, const T* data_cost_selected,
-            const T* selected_disp_pyr_cur, size_t msg_step, int h, int w, int nr_plane, int iters, cudaStream_t stream);
-
-        template<class T>
-        void compute_disp(const T* u, const T* d, const T* l, const T* r, const T* data_cost_selected, const T* disp_selected, size_t msg_step,
-            const PtrStepSz<short>& disp, int nr_plane, cudaStream_t stream);
-    }
-}}}
-
-using namespace ::cv::gpu::device::stereocsbp;
-
-namespace
-{
-    const float DEFAULT_MAX_DATA_TERM = 30.0f;
-    const float DEFAULT_DATA_WEIGHT = 1.0f;
-    const float DEFAULT_MAX_DISC_TERM = 160.0f;
-    const float DEFAULT_DISC_SINGLE_JUMP = 10.0f;
-}
-
-void cv::gpu::StereoConstantSpaceBP::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane)
-{
-    ndisp = (int) ((float) width / 3.14f);
-    if ((ndisp & 1) != 0)
-        ndisp++;
-
-    int mm = std::max(width, height);
-    iters = mm / 100 + ((mm > 1200)? - 4 : 4);
-
-    levels = (int)::log(static_cast<double>(mm)) * 2 / 3;
-    if (levels == 0) levels++;
-
-    nr_plane = (int) ((float) ndisp / std::pow(2.0, levels + 1));
-}
-
-cv::gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp_, int iters_, int levels_, int nr_plane_,
-                                                      int msg_type_)
-
-    : ndisp(ndisp_), iters(iters_), levels(levels_), nr_plane(nr_plane_),
-      max_data_term(DEFAULT_MAX_DATA_TERM), data_weight(DEFAULT_DATA_WEIGHT),
-      max_disc_term(DEFAULT_MAX_DISC_TERM), disc_single_jump(DEFAULT_DISC_SINGLE_JUMP), min_disp_th(0),
-      msg_type(msg_type_), use_local_init_data_cost(true)
-{
-    CV_Assert(msg_type_ == CV_32F || msg_type_ == CV_16S);
-}
-
-cv::gpu::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp_, int iters_, int levels_, int nr_plane_,
-                                                      float max_data_term_, float data_weight_, float max_disc_term_, float disc_single_jump_,
-                                                      int min_disp_th_, int msg_type_)
-    : ndisp(ndisp_), iters(iters_), levels(levels_), nr_plane(nr_plane_),
-      max_data_term(max_data_term_), data_weight(data_weight_),
-      max_disc_term(max_disc_term_), disc_single_jump(disc_single_jump_), min_disp_th(min_disp_th_),
-      msg_type(msg_type_), use_local_init_data_cost(true)
-{
-    CV_Assert(msg_type_ == CV_32F || msg_type_ == CV_16S);
-}
-
-template<class T>
-static void csbp_operator(StereoConstantSpaceBP& rthis, GpuMat& mbuf, GpuMat& temp, GpuMat& out, const GpuMat& left, const GpuMat& right, GpuMat& disp, Stream& stream)
-{
-    CV_DbgAssert(0 < rthis.ndisp && 0 < rthis.iters && 0 < rthis.levels && 0 < rthis.nr_plane
-        && left.rows == right.rows && left.cols == right.cols && left.type() == right.type());
-
-    CV_Assert(rthis.levels <= 8 && (left.type() == CV_8UC1 || left.type() == CV_8UC3 || left.type() == CV_8UC4));
-
-    const Scalar zero = Scalar::all(0);
-
-    cudaStream_t cudaStream = StreamAccessor::getStream(stream);
-
-    ////////////////////////////////////////////////////////////////////////////////////////////
-    // Init
-
-    int rows = left.rows;
-    int cols = left.cols;
-
-    rthis.levels = std::min(rthis.levels, int(log((double)rthis.ndisp) / log(2.0)));
-    int levels = rthis.levels;
-
-    // compute sizes
-    AutoBuffer<int> buf(levels * 3);
-    int* cols_pyr = buf;
-    int* rows_pyr = cols_pyr + levels;
-    int* nr_plane_pyr = rows_pyr + levels;
-
-    cols_pyr[0]     = cols;
-    rows_pyr[0]     = rows;
-    nr_plane_pyr[0] = rthis.nr_plane;
-
-    for (int i = 1; i < levels; i++)
-    {
-        cols_pyr[i]     = cols_pyr[i-1] / 2;
-        rows_pyr[i]     = rows_pyr[i-1] / 2;
-        nr_plane_pyr[i] = nr_plane_pyr[i-1] * 2;
-    }
-
-
-    GpuMat u[2], d[2], l[2], r[2], disp_selected_pyr[2], data_cost, data_cost_selected;
-
-
-    //allocate buffers
-    int buffers_count = 10; // (up + down + left + right + disp_selected_pyr) * 2
-    buffers_count += 2; //  data_cost has twice more rows than other buffers, what's why +2, not +1;
-    buffers_count += 1; //  data_cost_selected
-    mbuf.create(rows * rthis.nr_plane * buffers_count, cols, DataType<T>::type);
-
-    data_cost          = mbuf.rowRange(0, rows * rthis.nr_plane * 2);
-    data_cost_selected = mbuf.rowRange(data_cost.rows, data_cost.rows + rows * rthis.nr_plane);
-
-    for(int k = 0; k < 2; ++k) // in/out
-    {
-        GpuMat sub1 = mbuf.rowRange(data_cost.rows + data_cost_selected.rows, mbuf.rows);
-        GpuMat sub2 = sub1.rowRange((k+0)*sub1.rows/2, (k+1)*sub1.rows/2);
-
-        GpuMat *buf_ptrs[] = { &u[k], &d[k], &l[k], &r[k], &disp_selected_pyr[k] };
-        for(int _r = 0; _r < 5; ++_r)
-        {
-            *buf_ptrs[_r] = sub2.rowRange(_r * sub2.rows/5, (_r+1) * sub2.rows/5);
-            assert(buf_ptrs[_r]->cols == cols && buf_ptrs[_r]->rows == rows * rthis.nr_plane);
-        }
-    };
-
-    size_t elem_step = mbuf.step / sizeof(T);
-
-    Size temp_size = data_cost.size();
-    if ((size_t)temp_size.area() < elem_step * rows_pyr[levels - 1] * rthis.ndisp)
-        temp_size = Size(static_cast<int>(elem_step), rows_pyr[levels - 1] * rthis.ndisp);
-
-    temp.create(temp_size, DataType<T>::type);
-
-    ////////////////////////////////////////////////////////////////////////////
-    // Compute
-
-    load_constants(rthis.ndisp, rthis.max_data_term, rthis.data_weight, rthis.max_disc_term, rthis.disc_single_jump, rthis.min_disp_th, left, right, temp);
-
-    if (stream)
-    {
-        stream.enqueueMemSet(l[0], zero);
-        stream.enqueueMemSet(d[0], zero);
-        stream.enqueueMemSet(r[0], zero);
-        stream.enqueueMemSet(u[0], zero);
-
-        stream.enqueueMemSet(l[1], zero);
-        stream.enqueueMemSet(d[1], zero);
-        stream.enqueueMemSet(r[1], zero);
-        stream.enqueueMemSet(u[1], zero);
-
-        stream.enqueueMemSet(data_cost, zero);
-        stream.enqueueMemSet(data_cost_selected, zero);
-    }
-    else
-    {
-        l[0].setTo(zero);
-        d[0].setTo(zero);
-        r[0].setTo(zero);
-        u[0].setTo(zero);
-
-        l[1].setTo(zero);
-        d[1].setTo(zero);
-        r[1].setTo(zero);
-        u[1].setTo(zero);
-
-        data_cost.setTo(zero);
-        data_cost_selected.setTo(zero);
-    }
-
-    int cur_idx = 0;
-
-    for (int i = levels - 1; i >= 0; i--)
-    {
-        if (i == levels - 1)
-        {
-            init_data_cost(left.rows, left.cols, disp_selected_pyr[cur_idx].ptr<T>(), data_cost_selected.ptr<T>(),
-                elem_step, rows_pyr[i], cols_pyr[i], i, nr_plane_pyr[i], rthis.ndisp, left.channels(), rthis.use_local_init_data_cost, cudaStream);
-        }
-        else
-        {
-            compute_data_cost(disp_selected_pyr[cur_idx].ptr<T>(), data_cost.ptr<T>(), elem_step,
-                left.rows, left.cols, rows_pyr[i], cols_pyr[i], rows_pyr[i+1], i, nr_plane_pyr[i+1], left.channels(), cudaStream);
-
-            int new_idx = (cur_idx + 1) & 1;
-
-            init_message(u[new_idx].ptr<T>(), d[new_idx].ptr<T>(), l[new_idx].ptr<T>(), r[new_idx].ptr<T>(),
-                         u[cur_idx].ptr<T>(), d[cur_idx].ptr<T>(), l[cur_idx].ptr<T>(), r[cur_idx].ptr<T>(),
-                         disp_selected_pyr[new_idx].ptr<T>(), disp_selected_pyr[cur_idx].ptr<T>(),
-                         data_cost_selected.ptr<T>(), data_cost.ptr<T>(), elem_step, rows_pyr[i],
-                         cols_pyr[i], nr_plane_pyr[i], rows_pyr[i+1], cols_pyr[i+1], nr_plane_pyr[i+1], cudaStream);
-
-            cur_idx = new_idx;
-        }
-
-        calc_all_iterations(u[cur_idx].ptr<T>(), d[cur_idx].ptr<T>(), l[cur_idx].ptr<T>(), r[cur_idx].ptr<T>(),
-                            data_cost_selected.ptr<T>(), disp_selected_pyr[cur_idx].ptr<T>(), elem_step,
-                            rows_pyr[i], cols_pyr[i], nr_plane_pyr[i], rthis.iters, cudaStream);
-    }
-
-    if (disp.empty())
-        disp.create(rows, cols, CV_16S);
-
-    out = ((disp.type() == CV_16S) ? disp : (out.create(rows, cols, CV_16S), out));
-
-    if (stream)
-        stream.enqueueMemSet(out, zero);
-    else
-        out.setTo(zero);
-
-    compute_disp(u[cur_idx].ptr<T>(), d[cur_idx].ptr<T>(), l[cur_idx].ptr<T>(), r[cur_idx].ptr<T>(),
-                 data_cost_selected.ptr<T>(), disp_selected_pyr[cur_idx].ptr<T>(), elem_step, out, nr_plane_pyr[0], cudaStream);
-
-    if (disp.type() != CV_16S)
-    {
-        if (stream)
-            stream.enqueueConvert(out, disp, disp.type());
-        else
-            out.convertTo(disp, disp.type());
-    }
-}
-
-
-typedef void (*csbp_operator_t)(StereoConstantSpaceBP& rthis, GpuMat& mbuf,
-                                     GpuMat& temp, GpuMat& out, const GpuMat& left, const GpuMat& right, GpuMat& disp, Stream& stream);
-
-const static csbp_operator_t operators[] = {0, 0, 0, csbp_operator<short>, 0, csbp_operator<float>, 0, 0};
-
-void cv::gpu::StereoConstantSpaceBP::operator()(const GpuMat& left, const GpuMat& right, GpuMat& disp, Stream& stream)
-{
-    CV_Assert(msg_type == CV_32F || msg_type == CV_16S);
-    operators[msg_type](*this, messages_buffers, temp, out, left, right, disp, stream);
-}
-
-#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/thread_wrappers.cpp b/modules/gpu/src/thread_wrappers.cpp
deleted file mode 100644
index f8099aca91..0000000000
--- a/modules/gpu/src/thread_wrappers.cpp
+++ /dev/null
@@ -1,254 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "thread_wrappers.h"
-
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
-
-#ifdef WIN32
-    #define NOMINMAX
-    #include <windows.h>
-#else
-    #include <pthread.h>
-    #include <unistd.h>
-#endif
-
-#ifdef WIN32
-    class cv::gpu::detail::CriticalSection::Impl
-    {
-    public:
-        Impl()
-        {
-            InitializeCriticalSection(&criticalSection_);
-        }
-
-        ~Impl()
-        {
-            DeleteCriticalSection(&criticalSection_);
-        }
-
-        void enter()
-        {
-            EnterCriticalSection(&criticalSection_);
-        }
-
-        void leave()
-        {
-            LeaveCriticalSection(&criticalSection_);
-        }
-
-    private:
-        CRITICAL_SECTION criticalSection_;
-    };
-#else
-    class cv::gpu::detail::CriticalSection::Impl
-    {
-    public:
-        Impl()
-        {
-            pthread_mutexattr_t mutex_attribute;
-            pthread_mutexattr_init(&mutex_attribute);
-            pthread_mutexattr_settype(&mutex_attribute, PTHREAD_MUTEX_RECURSIVE);
-            pthread_mutex_init(&mutex_, 0);
-            pthread_mutexattr_destroy(&mutex_attribute);
-        }
-
-        ~Impl()
-        {
-            pthread_mutex_destroy(&mutex_);
-        }
-
-        void enter()
-        {
-            pthread_mutex_lock(&mutex_);
-        }
-
-        void leave()
-        {
-            pthread_mutex_unlock(&mutex_);
-        }
-
-    private:
-        pthread_mutex_t mutex_;
-    };
-#endif
-
-cv::gpu::detail::CriticalSection::CriticalSection() :
-    impl_(new Impl)
-{
-}
-
-cv::gpu::detail::CriticalSection::~CriticalSection()
-{
-}
-
-void cv::gpu::detail::CriticalSection::enter()
-{
-    impl_->enter();
-}
-
-void cv::gpu::detail::CriticalSection::leave()
-{
-    impl_->leave();
-}
-
-#ifdef WIN32
-    namespace
-    {
-        struct UserData
-        {
-            void (*func)(void* userData);
-            void* param;
-        };
-
-        DWORD WINAPI WinThreadFunction(LPVOID lpParam)
-        {
-            UserData* userData = static_cast<UserData*>(lpParam);
-
-            userData->func(userData->param);
-
-            return 0;
-        }
-    }
-
-    class cv::gpu::detail::Thread::Impl
-    {
-    public:
-        Impl(void (*func)(void* userData), void* userData)
-        {
-            userData_.func = func;
-            userData_.param = userData;
-
-            thread_ = CreateThread(
-                NULL,                   // default security attributes
-                0,                      // use default stack size
-                WinThreadFunction,      // thread function name
-                &userData_,             // argument to thread function
-                0,                      // use default creation flags
-                &threadId_);            // returns the thread identifier
-        }
-
-        ~Impl()
-        {
-            CloseHandle(thread_);
-        }
-
-        void wait()
-        {
-            WaitForSingleObject(thread_, INFINITE);
-        }
-
-    private:
-        UserData userData_;
-        HANDLE thread_;
-        DWORD threadId_;
-    };
-#else
-    namespace
-    {
-        struct UserData
-        {
-            void (*func)(void* userData);
-            void* param;
-        };
-
-        void* PThreadFunction(void* lpParam)
-        {
-            UserData* userData = static_cast<UserData*>(lpParam);
-
-            userData->func(userData->param);
-
-            return 0;
-        }
-    }
-
-    class cv::gpu::detail::Thread::Impl
-    {
-    public:
-        Impl(void (*func)(void* userData), void* userData)
-        {
-            userData_.func = func;
-            userData_.param = userData;
-
-            pthread_create(&thread_, NULL, PThreadFunction, &userData_);
-        }
-
-        ~Impl()
-        {
-            pthread_detach(thread_);
-        }
-
-        void wait()
-        {
-            pthread_join(thread_, NULL);
-        }
-
-    private:
-        pthread_t thread_;
-        UserData userData_;
-    };
-#endif
-
-cv::gpu::detail::Thread::Thread(void (*func)(void* userData), void* userData) :
-    impl_(new Impl(func, userData))
-{
-}
-
-cv::gpu::detail::Thread::~Thread()
-{
-}
-
-void cv::gpu::detail::Thread::wait()
-{
-    impl_->wait();
-}
-
-void cv::gpu::detail::Thread::sleep(int ms)
-{
-#ifdef WIN32
-    ::Sleep(ms);
-#else
-    ::usleep(ms * 1000);
-#endif
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/video_decoder.h b/modules/gpu/src/video_decoder.h
deleted file mode 100644
index 634643b889..0000000000
--- a/modules/gpu/src/video_decoder.h
+++ /dev/null
@@ -1,117 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __VIDEO_DECODER_H__
-#define __VIDEO_DECODER_H__
-
-#include "precomp.hpp"
-#include "cu_safe_call.h"
-
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
-
-namespace cv { namespace gpu
-{
-    namespace detail
-    {
-        class VideoDecoder
-        {
-        public:
-            VideoDecoder(const VideoReader_GPU::FormatInfo& videoFormat, CUvideoctxlock lock) : lock_(lock), decoder_(0)
-            {
-                create(videoFormat);
-            }
-
-            ~VideoDecoder()
-            {
-                release();
-            }
-
-            void create(const VideoReader_GPU::FormatInfo& videoFormat);
-            void release();
-
-            // Get the code-type currently used.
-            cudaVideoCodec codec() const { return createInfo_.CodecType; }
-            unsigned long maxDecodeSurfaces() const { return createInfo_.ulNumDecodeSurfaces; }
-
-            unsigned long frameWidth() const { return createInfo_.ulWidth; }
-            unsigned long frameHeight() const { return createInfo_.ulHeight; }
-
-            unsigned long targetWidth() const { return createInfo_.ulTargetWidth; }
-            unsigned long targetHeight() const { return createInfo_.ulTargetHeight; }
-
-            cudaVideoChromaFormat chromaFormat() const { return createInfo_.ChromaFormat; }
-
-            bool decodePicture(CUVIDPICPARAMS* picParams)
-            {
-                return cuvidDecodePicture(decoder_, picParams) == CUDA_SUCCESS;
-            }
-
-            cv::gpu::GpuMat mapFrame(int picIdx, CUVIDPROCPARAMS& videoProcParams)
-            {
-                CUdeviceptr ptr;
-                unsigned int pitch;
-
-                cuSafeCall( cuvidMapVideoFrame(decoder_, picIdx, &ptr, &pitch, &videoProcParams) );
-
-                return GpuMat(targetHeight() * 3 / 2, targetWidth(), CV_8UC1, (void*) ptr, pitch);
-            }
-
-            void unmapFrame(cv::gpu::GpuMat& frame)
-            {
-                cuSafeCall( cuvidUnmapVideoFrame(decoder_, (CUdeviceptr) frame.data) );
-                frame.release();
-            }
-
-        private:
-            VideoDecoder(const VideoDecoder&);
-            VideoDecoder& operator =(const VideoDecoder&);
-
-            CUvideoctxlock lock_;
-            CUVIDDECODECREATEINFO createInfo_;
-            CUvideodecoder        decoder_;
-        };
-    }
-}}
-
-#endif // HAVE_CUDA
-
-#endif // __VIDEO_DECODER_H__
diff --git a/modules/gpu/src/video_reader.cpp b/modules/gpu/src/video_reader.cpp
deleted file mode 100644
index 3224902c6d..0000000000
--- a/modules/gpu/src/video_reader.cpp
+++ /dev/null
@@ -1,401 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                          License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined(HAVE_CUDA) || defined(CUDA_DISABLER) || !defined(HAVE_NVCUVID)
-
-class cv::gpu::VideoReader_GPU::Impl
-{
-};
-
-cv::gpu::VideoReader_GPU::VideoReader_GPU() { throw_nogpu(); }
-cv::gpu::VideoReader_GPU::VideoReader_GPU(const std::string&) { throw_nogpu(); }
-cv::gpu::VideoReader_GPU::VideoReader_GPU(const cv::Ptr<VideoSource>&) { throw_nogpu(); }
-cv::gpu::VideoReader_GPU::~VideoReader_GPU() { }
-void cv::gpu::VideoReader_GPU::open(const std::string&) { throw_nogpu(); }
-void cv::gpu::VideoReader_GPU::open(const cv::Ptr<VideoSource>&) { throw_nogpu(); }
-bool cv::gpu::VideoReader_GPU::isOpened() const { return false; }
-void cv::gpu::VideoReader_GPU::close() { }
-bool cv::gpu::VideoReader_GPU::read(GpuMat&) { throw_nogpu(); return false; }
-cv::gpu::VideoReader_GPU::FormatInfo cv::gpu::VideoReader_GPU::format() const { throw_nogpu(); FormatInfo format_ = {MPEG1,Monochrome,0,0}; return format_; }
-bool cv::gpu::VideoReader_GPU::VideoSource::parseVideoData(const unsigned char*, size_t, bool) { throw_nogpu(); return false; }
-void cv::gpu::VideoReader_GPU::dumpFormat(std::ostream&) { throw_nogpu(); }
-
-#else // HAVE_CUDA
-
-#include "frame_queue.h"
-#include "video_decoder.h"
-#include "video_parser.h"
-
-#include "cuvid_video_source.h"
-#include "ffmpeg_video_source.h"
-
-#include "cu_safe_call.h"
-
-class cv::gpu::VideoReader_GPU::Impl
-{
-public:
-    explicit Impl(const cv::Ptr<cv::gpu::VideoReader_GPU::VideoSource>& source);
-    ~Impl();
-
-    bool grab(cv::gpu::GpuMat& frame);
-
-    cv::gpu::VideoReader_GPU::FormatInfo format() const { return videoSource_->format(); }
-
-private:
-    Impl(const Impl&);
-    Impl& operator =(const Impl&);
-
-    cv::Ptr<cv::gpu::VideoReader_GPU::VideoSource> videoSource_;
-
-    std::auto_ptr<cv::gpu::detail::FrameQueue> frameQueue_;
-    std::auto_ptr<cv::gpu::detail::VideoDecoder> videoDecoder_;
-    std::auto_ptr<cv::gpu::detail::VideoParser> videoParser_;
-
-    CUvideoctxlock lock_;
-
-    std::deque< std::pair<CUVIDPARSERDISPINFO, CUVIDPROCPARAMS> > frames_;
-};
-
-cv::gpu::VideoReader_GPU::Impl::Impl(const cv::Ptr<VideoSource>& source) :
-    videoSource_(source),
-    lock_(0)
-{
-    // init context
-    GpuMat temp(1, 1, CV_8UC1);
-    temp.release();
-
-    DeviceInfo devInfo;
-    CV_Assert( devInfo.supports(FEATURE_SET_COMPUTE_11) );
-
-    CUcontext ctx;
-    cuSafeCall( cuCtxGetCurrent(&ctx) );
-    cuSafeCall( cuvidCtxLockCreate(&lock_, ctx) );
-
-    frameQueue_.reset(new detail::FrameQueue);
-    videoDecoder_.reset(new detail::VideoDecoder(videoSource_->format(), lock_));
-    videoParser_.reset(new detail::VideoParser(videoDecoder_.get(), frameQueue_.get()));
-
-    videoSource_->setFrameQueue(frameQueue_.get());
-    videoSource_->setVideoParser(videoParser_.get());
-
-    videoSource_->start();
-}
-
-cv::gpu::VideoReader_GPU::Impl::~Impl()
-{
-    frameQueue_->endDecode();
-    videoSource_->stop();
-}
-
-namespace cv { namespace gpu { namespace device {
-    namespace video_decoding
-    {
-        void loadHueCSC(float hueCSC[9]);
-        void NV12ToARGB_gpu(const PtrStepb decodedFrame, PtrStepSz<unsigned int> interopFrame, cudaStream_t stream = 0);
-    }
-}}}
-
-namespace
-{
-    class VideoCtxAutoLock
-    {
-    public:
-        VideoCtxAutoLock(CUvideoctxlock lock) : m_lock(lock) { cuSafeCall( cuvidCtxLock(m_lock, 0) ); }
-        ~VideoCtxAutoLock() { cuvidCtxUnlock(m_lock, 0); }
-
-    private:
-        CUvideoctxlock m_lock;
-    };
-
-    enum ColorSpace
-    {
-        ITU601 = 1,
-        ITU709 = 2
-    };
-
-    void setColorSpaceMatrix(ColorSpace CSC, float hueCSC[9], float hue)
-    {
-        float hueSin = std::sin(hue);
-        float hueCos = std::cos(hue);
-
-        if (CSC == ITU601)
-        {
-            //CCIR 601
-            hueCSC[0] = 1.1644f;
-            hueCSC[1] = hueSin * 1.5960f;
-            hueCSC[2] = hueCos * 1.5960f;
-            hueCSC[3] = 1.1644f;
-            hueCSC[4] = (hueCos * -0.3918f) - (hueSin * 0.8130f);
-            hueCSC[5] = (hueSin *  0.3918f) - (hueCos * 0.8130f);
-            hueCSC[6] = 1.1644f;
-            hueCSC[7] = hueCos *  2.0172f;
-            hueCSC[8] = hueSin * -2.0172f;
-        }
-        else if (CSC == ITU709)
-        {
-            //CCIR 709
-            hueCSC[0] = 1.0f;
-            hueCSC[1] = hueSin * 1.57480f;
-            hueCSC[2] = hueCos * 1.57480f;
-            hueCSC[3] = 1.0;
-            hueCSC[4] = (hueCos * -0.18732f) - (hueSin * 0.46812f);
-            hueCSC[5] = (hueSin *  0.18732f) - (hueCos * 0.46812f);
-            hueCSC[6] = 1.0f;
-            hueCSC[7] = hueCos *  1.85560f;
-            hueCSC[8] = hueSin * -1.85560f;
-        }
-    }
-
-    void cudaPostProcessFrame(const cv::gpu::GpuMat& decodedFrame, cv::gpu::GpuMat& interopFrame, int width, int height)
-    {
-        using namespace cv::gpu::device::video_decoding;
-
-        static bool updateCSC = true;
-        static float hueColorSpaceMat[9];
-
-        // Upload the Color Space Conversion Matrices
-        if (updateCSC)
-        {
-            const ColorSpace colorSpace = ITU601;
-            const float hue = 0.0f;
-
-            // CCIR 601/709
-            setColorSpaceMatrix(colorSpace, hueColorSpaceMat, hue);
-
-            updateCSC = false;
-        }
-
-        // Final Stage: NV12toARGB color space conversion
-
-        interopFrame.create(height, width, CV_8UC4);
-
-        loadHueCSC(hueColorSpaceMat);
-
-        NV12ToARGB_gpu(decodedFrame, interopFrame);
-    }
-}
-
-bool cv::gpu::VideoReader_GPU::Impl::grab(GpuMat& frame)
-{
-    if (videoSource_->hasError() || videoParser_->hasError())
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported video source");
-
-    if (!videoSource_->isStarted() || frameQueue_->isEndOfDecode())
-        return false;
-
-    if (frames_.empty())
-    {
-        CUVIDPARSERDISPINFO displayInfo;
-
-        for (;;)
-        {
-            if (frameQueue_->dequeue(displayInfo))
-                break;
-
-            if (videoSource_->hasError() || videoParser_->hasError())
-                CV_Error(CV_StsUnsupportedFormat, "Unsupported video source");
-
-            if (frameQueue_->isEndOfDecode())
-                return false;
-
-            // Wait a bit
-            detail::Thread::sleep(1);
-        }
-
-        bool isProgressive = displayInfo.progressive_frame != 0;
-        const int num_fields = isProgressive ? 1 : 2 + displayInfo.repeat_first_field;
-
-        for (int active_field = 0; active_field < num_fields; ++active_field)
-        {
-            CUVIDPROCPARAMS videoProcParams;
-            std::memset(&videoProcParams, 0, sizeof(CUVIDPROCPARAMS));
-
-            videoProcParams.progressive_frame = displayInfo.progressive_frame;
-            videoProcParams.second_field      = active_field;
-            videoProcParams.top_field_first   = displayInfo.top_field_first;
-            videoProcParams.unpaired_field    = (num_fields == 1);
-
-            frames_.push_back(std::make_pair(displayInfo, videoProcParams));
-        }
-    }
-
-    if (frames_.empty())
-        return false;
-
-    std::pair<CUVIDPARSERDISPINFO, CUVIDPROCPARAMS> frameInfo = frames_.front();
-    frames_.pop_front();
-
-    {
-        VideoCtxAutoLock autoLock(lock_);
-
-        // map decoded video frame to CUDA surface
-        cv::gpu::GpuMat decodedFrame = videoDecoder_->mapFrame(frameInfo.first.picture_index, frameInfo.second);
-
-        // perform post processing on the CUDA surface (performs colors space conversion and post processing)
-        // comment this out if we inclue the line of code seen above
-        cudaPostProcessFrame(decodedFrame, frame, videoDecoder_->targetWidth(), videoDecoder_->targetHeight());
-
-        // unmap video frame
-        // unmapFrame() synchronizes with the VideoDecode API (ensures the frame has finished decoding)
-        videoDecoder_->unmapFrame(decodedFrame);
-    }
-
-    // release the frame, so it can be re-used in decoder
-    if (frames_.empty())
-        frameQueue_->releaseFrame(frameInfo.first);
-
-    return true;
-}
-
-////////////////////////////////////////////////////////////////////////////
-
-cv::gpu::VideoReader_GPU::VideoReader_GPU()
-{
-}
-
-cv::gpu::VideoReader_GPU::VideoReader_GPU(const std::string& filename)
-{
-    open(filename);
-}
-
-cv::gpu::VideoReader_GPU::VideoReader_GPU(const cv::Ptr<VideoSource>& source)
-{
-    open(source);
-}
-
-cv::gpu::VideoReader_GPU::~VideoReader_GPU()
-{
-    close();
-}
-
-void cv::gpu::VideoReader_GPU::open(const std::string& filename)
-{
-    CV_Assert( !filename.empty() );
-
-#ifndef __APPLE__
-    try
-    {
-        cv::Ptr<VideoSource> source(new detail::CuvidVideoSource(filename));
-        open(source);
-    }
-    catch (const std::runtime_error&)
-#endif
-    {
-        cv::Ptr<VideoSource> source(new cv::gpu::detail::FFmpegVideoSource(filename));
-        open(source);
-    }
-}
-
-void cv::gpu::VideoReader_GPU::open(const cv::Ptr<VideoSource>& source)
-{
-    CV_Assert( !source.empty() );
-    close();
-    impl_.reset(new Impl(source));
-}
-
-bool cv::gpu::VideoReader_GPU::isOpened() const
-{
-    return impl_.get() != 0;
-}
-
-void cv::gpu::VideoReader_GPU::close()
-{
-    impl_.reset();
-}
-
-bool cv::gpu::VideoReader_GPU::read(GpuMat& image)
-{
-    if (!isOpened())
-        return false;
-
-    if (!impl_->grab(image))
-    {
-        close();
-        return false;
-    }
-
-    return true;
-}
-
-cv::gpu::VideoReader_GPU::FormatInfo cv::gpu::VideoReader_GPU::format() const
-{
-    CV_Assert( isOpened() );
-    return impl_->format();
-}
-
-bool cv::gpu::VideoReader_GPU::VideoSource::parseVideoData(const unsigned char* data, size_t size, bool endOfStream)
-{
-    return videoParser_->parseVideoData(data, size, endOfStream);
-}
-
-void cv::gpu::VideoReader_GPU::dumpFormat(std::ostream& st)
-{
-    static const char* codecs[] =
-    {
-        "MPEG1",
-        "MPEG2",
-        "MPEG4",
-        "VC1",
-        "H264",
-        "JPEG",
-        "H264_SVC",
-        "H264_MVC"
-    };
-
-    static const char* chromas[] =
-    {
-        "Monochrome",
-        "YUV420",
-        "YUV422",
-        "YUV444"
-    };
-
-    FormatInfo _format = this->format();
-
-    st << "Frame Size    : " << _format.width << "x" << _format.height << std::endl;
-    st << "Codec         : " << (_format.codec <= H264_MVC ? codecs[_format.codec] : "Uncompressed YUV") << std::endl;
-    st << "Chroma Format : " << chromas[_format.chromaFormat] << std::endl;
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/src/video_writer.cpp b/modules/gpu/src/video_writer.cpp
deleted file mode 100644
index fe44a16f7c..0000000000
--- a/modules/gpu/src/video_writer.cpp
+++ /dev/null
@@ -1,1004 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-#if !defined(HAVE_CUDA) || defined(CUDA_DISABLER) || !defined(HAVE_NVCUVID) || !defined(WIN32)
-
-class cv::gpu::VideoWriter_GPU::Impl
-{
-};
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU() { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string&, cv::Size, double, SurfaceFormat) { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string&, cv::Size, double, const EncoderParams&, SurfaceFormat) { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>&, cv::Size, double, SurfaceFormat) { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>&, cv::Size, double, const EncoderParams&, SurfaceFormat) { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::~VideoWriter_GPU() {}
-void cv::gpu::VideoWriter_GPU::open(const std::string&, cv::Size, double, SurfaceFormat) { throw_nogpu(); }
-void cv::gpu::VideoWriter_GPU::open(const std::string&, cv::Size, double, const EncoderParams&, SurfaceFormat) { throw_nogpu(); }
-void cv::gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>&, cv::Size, double, SurfaceFormat) { throw_nogpu(); }
-void cv::gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>&, cv::Size, double, const EncoderParams&, SurfaceFormat) { throw_nogpu(); }
-bool cv::gpu::VideoWriter_GPU::isOpened() const { return false; }
-void cv::gpu::VideoWriter_GPU::close() {}
-void cv::gpu::VideoWriter_GPU::write(const cv::gpu::GpuMat&, bool) { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::EncoderParams cv::gpu::VideoWriter_GPU::getParams() const { EncoderParams params; throw_nogpu(); return params; }
-
-cv::gpu::VideoWriter_GPU::EncoderParams::EncoderParams() { throw_nogpu(); }
-cv::gpu::VideoWriter_GPU::EncoderParams::EncoderParams(const std::string&) { throw_nogpu(); }
-void cv::gpu::VideoWriter_GPU::EncoderParams::load(const std::string&) { throw_nogpu(); }
-void cv::gpu::VideoWriter_GPU::EncoderParams::save(const std::string&) const { throw_nogpu(); }
-
-#else // !defined HAVE_CUDA || !defined WIN32
-
-#ifdef HAVE_FFMPEG
-    #include "../src/cap_ffmpeg_impl.hpp"
-#else
-    #include "../src/cap_ffmpeg_api.hpp"
-#endif
-
-
-///////////////////////////////////////////////////////////////////////////
-// VideoWriter_GPU::Impl
-
-namespace
-{
-    class NVEncoderWrapper
-    {
-    public:
-        NVEncoderWrapper() : encoder_(0)
-        {
-            int err;
-
-            err = NVGetHWEncodeCaps();
-            if (err)
-                CV_Error(CV_GpuNotSupported, "No CUDA capability present");
-
-            // Create the Encoder API Interface
-            err = NVCreateEncoder(&encoder_);
-            CV_Assert( err == 0 );
-        }
-
-        ~NVEncoderWrapper()
-        {
-            if (encoder_)
-                NVDestroyEncoder(encoder_);
-        }
-
-        operator NVEncoder() const
-        {
-            return encoder_;
-        }
-
-    private:
-        NVEncoder encoder_;
-    };
-
-    enum CodecType
-    {
-        MPEG1, //not supported yet
-        MPEG2, //not supported yet
-        MPEG4, //not supported yet
-        H264
-    };
-}
-
-class cv::gpu::VideoWriter_GPU::Impl
-{
-public:
-    Impl(const cv::Ptr<EncoderCallBack>& callback, cv::Size frameSize, double fps, SurfaceFormat format, CodecType codec = H264);
-    Impl(const cv::Ptr<EncoderCallBack>& callback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format, CodecType codec = H264);
-
-    void write(const cv::gpu::GpuMat& image, bool lastFrame);
-
-    EncoderParams getParams() const;
-
-private:
-    Impl(const Impl&);
-    Impl& operator=(const Impl&);
-
-    void initEncoder(double fps);
-    void setEncodeParams(const EncoderParams& params);
-    void initGpuMemory();
-    void initCallBacks();
-    void createHWEncoder();
-
-    cv::Ptr<EncoderCallBack> callback_;
-    cv::Size frameSize_;
-
-    CodecType codec_;
-    SurfaceFormat inputFormat_;
-    NVVE_SurfaceFormat surfaceFormat_;
-
-    NVEncoderWrapper encoder_;
-
-    cv::gpu::GpuMat videoFrame_;
-    CUvideoctxlock cuCtxLock_;
-
-    // CallBacks
-
-    static unsigned char* NVENCAPI HandleAcquireBitStream(int* pBufferSize, void* pUserdata);
-    static void NVENCAPI HandleReleaseBitStream(int nBytesInBuffer, unsigned char* cb, void* pUserdata);
-    static void NVENCAPI HandleOnBeginFrame(const NVVE_BeginFrameInfo* pbfi, void* pUserdata);
-    static void NVENCAPI HandleOnEndFrame(const NVVE_EndFrameInfo* pefi, void* pUserdata);
-};
-
-cv::gpu::VideoWriter_GPU::Impl::Impl(const cv::Ptr<EncoderCallBack>& callback, cv::Size frameSize, double fps, SurfaceFormat format, CodecType codec) :
-    callback_(callback),
-    frameSize_(frameSize),
-    codec_(codec),
-    inputFormat_(format),
-    cuCtxLock_(0)
-{
-    surfaceFormat_ = inputFormat_ == SF_BGR ? YV12 : static_cast<NVVE_SurfaceFormat>(inputFormat_);
-
-    initEncoder(fps);
-
-    initGpuMemory();
-
-    initCallBacks();
-
-    createHWEncoder();
-}
-
-cv::gpu::VideoWriter_GPU::Impl::Impl(const cv::Ptr<EncoderCallBack>& callback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format, CodecType codec) :
-    callback_(callback),
-    frameSize_(frameSize),
-    codec_(codec),
-    inputFormat_(format),
-    cuCtxLock_(0)
-{
-    surfaceFormat_ = inputFormat_ == SF_BGR ? YV12 : static_cast<NVVE_SurfaceFormat>(inputFormat_);
-
-    initEncoder(fps);
-
-    setEncodeParams(params);
-
-    initGpuMemory();
-
-    initCallBacks();
-
-    createHWEncoder();
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::initEncoder(double fps)
-{
-    int err;
-
-    // Set codec
-
-    static const unsigned long codecs_id[] =
-    {
-        NV_CODEC_TYPE_MPEG1, NV_CODEC_TYPE_MPEG2, NV_CODEC_TYPE_MPEG4, NV_CODEC_TYPE_H264, NV_CODEC_TYPE_VC1
-    };
-    err = NVSetCodec(encoder_, codecs_id[codec_]);
-    if (err)
-        CV_Error(CV_StsNotImplemented, "Codec format is not supported");
-
-    // Set default params
-
-    err = NVSetDefaultParam(encoder_);
-    CV_Assert( err == 0 );
-
-    // Set some common params
-
-    int inputSize[] = { frameSize_.width, frameSize_.height };
-    err = NVSetParamValue(encoder_, NVVE_IN_SIZE, &inputSize);
-    CV_Assert( err == 0 );
-    err = NVSetParamValue(encoder_, NVVE_OUT_SIZE, &inputSize);
-    CV_Assert( err == 0 );
-
-    int aspectRatio[] = { frameSize_.width, frameSize_.height, ASPECT_RATIO_DAR };
-    err = NVSetParamValue(encoder_, NVVE_ASPECT_RATIO, &aspectRatio);
-    CV_Assert( err == 0 );
-
-    // FPS
-
-    int frame_rate = static_cast<int>(fps + 0.5);
-    int frame_rate_base = 1;
-    while (fabs(static_cast<double>(frame_rate) / frame_rate_base) - fps > 0.001)
-    {
-        frame_rate_base *= 10;
-        frame_rate = static_cast<int>(fps*frame_rate_base + 0.5);
-    }
-    int FrameRate[] = { frame_rate, frame_rate_base };
-    err = NVSetParamValue(encoder_, NVVE_FRAME_RATE, &FrameRate);
-    CV_Assert( err == 0 );
-
-    // Select device for encoding
-
-    int gpuID = cv::gpu::getDevice();
-    err = NVSetParamValue(encoder_, NVVE_FORCE_GPU_SELECTION, &gpuID);
-    CV_Assert( err == 0 );
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::setEncodeParams(const EncoderParams& params)
-{
-    int err;
-
-    int P_Interval = params.P_Interval;
-    err = NVSetParamValue(encoder_, NVVE_P_INTERVAL, &P_Interval);
-    CV_Assert( err == 0 );
-
-    int IDR_Period = params.IDR_Period;
-    err = NVSetParamValue(encoder_, NVVE_IDR_PERIOD, &IDR_Period);
-    CV_Assert( err == 0 );
-
-    int DynamicGOP = params.DynamicGOP;
-    err = NVSetParamValue(encoder_, NVVE_DYNAMIC_GOP, &DynamicGOP);
-    CV_Assert( err == 0 );
-
-    NVVE_RateCtrlType RCType = static_cast<NVVE_RateCtrlType>(params.RCType);
-    err = NVSetParamValue(encoder_, NVVE_RC_TYPE, &RCType);
-    CV_Assert( err == 0 );
-
-    int AvgBitrate = params.AvgBitrate;
-    err = NVSetParamValue(encoder_, NVVE_AVG_BITRATE, &AvgBitrate);
-    CV_Assert( err == 0 );
-
-    int PeakBitrate = params.PeakBitrate;
-    err = NVSetParamValue(encoder_, NVVE_PEAK_BITRATE, &PeakBitrate);
-    CV_Assert( err == 0 );
-
-    int QP_Level_Intra = params.QP_Level_Intra;
-    err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTRA, &QP_Level_Intra);
-    CV_Assert( err == 0 );
-
-    int QP_Level_InterP = params.QP_Level_InterP;
-    err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTER_P, &QP_Level_InterP);
-    CV_Assert( err == 0 );
-
-    int QP_Level_InterB = params.QP_Level_InterB;
-    err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTER_B, &QP_Level_InterB);
-    CV_Assert( err == 0 );
-
-    int DeblockMode = params.DeblockMode;
-    err = NVSetParamValue(encoder_, NVVE_DEBLOCK_MODE, &DeblockMode);
-    CV_Assert( err == 0 );
-
-    int ProfileLevel = params.ProfileLevel;
-    err = NVSetParamValue(encoder_, NVVE_PROFILE_LEVEL, &ProfileLevel);
-    CV_Assert( err == 0 );
-
-    int ForceIntra = params.ForceIntra;
-    err = NVSetParamValue(encoder_, NVVE_FORCE_INTRA, &ForceIntra);
-    CV_Assert( err == 0 );
-
-    int ForceIDR = params.ForceIDR;
-    err = NVSetParamValue(encoder_, NVVE_FORCE_IDR, &ForceIDR);
-    CV_Assert( err == 0 );
-
-    int ClearStat = params.ClearStat;
-    err = NVSetParamValue(encoder_, NVVE_CLEAR_STAT, &ClearStat);
-    CV_Assert( err == 0 );
-
-    NVVE_DI_MODE DIMode = static_cast<NVVE_DI_MODE>(params.DIMode);
-    err = NVSetParamValue(encoder_, NVVE_SET_DEINTERLACE, &DIMode);
-    CV_Assert( err == 0 );
-
-    if (params.Presets != -1)
-    {
-        NVVE_PRESETS_TARGET Presets = static_cast<NVVE_PRESETS_TARGET>(params.Presets);
-        err = NVSetParamValue(encoder_, NVVE_PRESETS, &Presets);
-        CV_Assert ( err == 0 );
-    }
-
-    int DisableCabac = params.DisableCabac;
-    err = NVSetParamValue(encoder_, NVVE_DISABLE_CABAC, &DisableCabac);
-    CV_Assert ( err == 0 );
-
-    int NaluFramingType = params.NaluFramingType;
-    err = NVSetParamValue(encoder_, NVVE_CONFIGURE_NALU_FRAMING_TYPE, &NaluFramingType);
-    CV_Assert ( err == 0 );
-
-    int DisableSPSPPS = params.DisableSPSPPS;
-    err = NVSetParamValue(encoder_, NVVE_DISABLE_SPS_PPS, &DisableSPSPPS);
-    CV_Assert ( err == 0 );
-}
-
-cv::gpu::VideoWriter_GPU::EncoderParams cv::gpu::VideoWriter_GPU::Impl::getParams() const
-{
-    int err;
-
-    EncoderParams params;
-
-    int P_Interval;
-    err = NVGetParamValue(encoder_, NVVE_P_INTERVAL, &P_Interval);
-    CV_Assert( err == 0 );
-    params.P_Interval = P_Interval;
-
-    int IDR_Period;
-    err = NVGetParamValue(encoder_, NVVE_IDR_PERIOD, &IDR_Period);
-    CV_Assert( err == 0 );
-    params.IDR_Period = IDR_Period;
-
-    int DynamicGOP;
-    err = NVGetParamValue(encoder_, NVVE_DYNAMIC_GOP, &DynamicGOP);
-    CV_Assert( err == 0 );
-    params.DynamicGOP = DynamicGOP;
-
-    NVVE_RateCtrlType RCType;
-    err = NVGetParamValue(encoder_, NVVE_RC_TYPE, &RCType);
-    CV_Assert( err == 0 );
-    params.RCType = RCType;
-
-    int AvgBitrate;
-    err = NVGetParamValue(encoder_, NVVE_AVG_BITRATE, &AvgBitrate);
-    CV_Assert( err == 0 );
-    params.AvgBitrate = AvgBitrate;
-
-    int PeakBitrate;
-    err = NVGetParamValue(encoder_, NVVE_PEAK_BITRATE, &PeakBitrate);
-    CV_Assert( err == 0 );
-    params.PeakBitrate = PeakBitrate;
-
-    int QP_Level_Intra;
-    err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTRA, &QP_Level_Intra);
-    CV_Assert( err == 0 );
-    params.QP_Level_Intra = QP_Level_Intra;
-
-    int QP_Level_InterP;
-    err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTER_P, &QP_Level_InterP);
-    CV_Assert( err == 0 );
-    params.QP_Level_InterP = QP_Level_InterP;
-
-    int QP_Level_InterB;
-    err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTER_B, &QP_Level_InterB);
-    CV_Assert( err == 0 );
-    params.QP_Level_InterB = QP_Level_InterB;
-
-    int DeblockMode;
-    err = NVGetParamValue(encoder_, NVVE_DEBLOCK_MODE, &DeblockMode);
-    CV_Assert( err == 0 );
-    params.DeblockMode = DeblockMode;
-
-    int ProfileLevel;
-    err = NVGetParamValue(encoder_, NVVE_PROFILE_LEVEL, &ProfileLevel);
-    CV_Assert( err == 0 );
-    params.ProfileLevel = ProfileLevel;
-
-    int ForceIntra;
-    err = NVGetParamValue(encoder_, NVVE_FORCE_INTRA, &ForceIntra);
-    CV_Assert( err == 0 );
-    params.ForceIntra = ForceIntra;
-
-    int ForceIDR;
-    err = NVGetParamValue(encoder_, NVVE_FORCE_IDR, &ForceIDR);
-    CV_Assert( err == 0 );
-    params.ForceIDR = ForceIDR;
-
-    int ClearStat;
-    err = NVGetParamValue(encoder_, NVVE_CLEAR_STAT, &ClearStat);
-    CV_Assert( err == 0 );
-    params.ClearStat = ClearStat;
-
-    NVVE_DI_MODE DIMode;
-    err = NVGetParamValue(encoder_, NVVE_SET_DEINTERLACE, &DIMode);
-    CV_Assert( err == 0 );
-    params.DIMode = DIMode;
-
-    params.Presets = -1;
-
-    int DisableCabac;
-    err = NVGetParamValue(encoder_, NVVE_DISABLE_CABAC, &DisableCabac);
-    CV_Assert ( err == 0 );
-    params.DisableCabac = DisableCabac;
-
-    int NaluFramingType;
-    err = NVGetParamValue(encoder_, NVVE_CONFIGURE_NALU_FRAMING_TYPE, &NaluFramingType);
-    CV_Assert ( err == 0 );
-    params.NaluFramingType = NaluFramingType;
-
-    int DisableSPSPPS;
-    err = NVGetParamValue(encoder_, NVVE_DISABLE_SPS_PPS, &DisableSPSPPS);
-    CV_Assert ( err == 0 );
-    params.DisableSPSPPS = DisableSPSPPS;
-
-    return params;
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::initGpuMemory()
-{
-    int err;
-    CUresult cuRes;
-
-    // initialize context
-    cv::gpu::GpuMat temp(1, 1, CV_8U);
-    temp.release();
-
-    static const int bpp[] =
-    {
-        16, // UYVY, 4:2:2
-        16, // YUY2, 4:2:2
-        12, // YV12, 4:2:0
-        12, // NV12, 4:2:0
-        12, // IYUV, 4:2:0
-    };
-
-    CUcontext cuContext;
-    cuRes = cuCtxGetCurrent(&cuContext);
-    CV_Assert( cuRes == CUDA_SUCCESS );
-
-    // Allocate the CUDA memory Pitched Surface
-    if (surfaceFormat_ == UYVY || surfaceFormat_ == YUY2)
-        videoFrame_.create(frameSize_.height, (frameSize_.width * bpp[surfaceFormat_]) / 8, CV_8UC1);
-    else
-        videoFrame_.create((frameSize_.height * bpp[surfaceFormat_]) / 8, frameSize_.width, CV_8UC1);
-
-    // Create the Video Context Lock (used for synchronization)
-    cuRes = cuvidCtxLockCreate(&cuCtxLock_, cuContext);
-    CV_Assert( cuRes == CUDA_SUCCESS );
-
-    // If we are using GPU Device Memory with NVCUVENC, it is necessary to create a
-    // CUDA Context with a Context Lock cuvidCtxLock.  The Context Lock needs to be passed to NVCUVENC
-
-    int iUseDeviceMem = 1;
-    err = NVSetParamValue(encoder_, NVVE_DEVICE_MEMORY_INPUT, &iUseDeviceMem);
-    CV_Assert ( err == 0 );
-
-    err = NVSetParamValue(encoder_, NVVE_DEVICE_CTX_LOCK, &cuCtxLock_);
-    CV_Assert ( err == 0 );
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::initCallBacks()
-{
-    NVVE_CallbackParams cb;
-    memset(&cb, 0, sizeof(NVVE_CallbackParams));
-
-    cb.pfnacquirebitstream = HandleAcquireBitStream;
-    cb.pfnonbeginframe     = HandleOnBeginFrame;
-    cb.pfnonendframe       = HandleOnEndFrame;
-    cb.pfnreleasebitstream = HandleReleaseBitStream;
-
-    NVRegisterCB(encoder_, cb, this);
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::createHWEncoder()
-{
-    int err;
-
-    // Create the NVIDIA HW resources for Encoding on NVIDIA hardware
-    err = NVCreateHWEncoder(encoder_);
-    CV_Assert( err == 0 );
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace video_encoding
-    {
-        void YV12_gpu(const PtrStepSzb src, int cn, PtrStepSzb dst);
-    }
-}}}
-
-namespace
-{
-    // UYVY/YUY2 are both 4:2:2 formats (16bpc)
-    // Luma, U, V are interleaved, chroma is subsampled (w/2,h)
-    void copyUYVYorYUY2Frame(cv::Size frameSize, const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst)
-    {
-        CUresult res;
-
-        // Source is YUVY/YUY2 4:2:2, the YUV data in a packed and interleaved
-
-        // YUV Copy setup
-        CUDA_MEMCPY2D stCopyYUV422;
-        memset((void*)&stCopyYUV422, 0, sizeof(stCopyYUV422));
-        stCopyYUV422.srcXInBytes          = 0;
-        stCopyYUV422.srcY                 = 0;
-        stCopyYUV422.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyYUV422.srcHost              = 0;
-        stCopyYUV422.srcDevice            = (CUdeviceptr) src.data;
-        stCopyYUV422.srcArray             = 0;
-        stCopyYUV422.srcPitch             = src.step;
-
-        stCopyYUV422.dstXInBytes          = 0;
-        stCopyYUV422.dstY                 = 0;
-        stCopyYUV422.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyYUV422.dstHost              = 0;
-        stCopyYUV422.dstDevice            = (CUdeviceptr) dst.data;
-        stCopyYUV422.dstArray             = 0;
-        stCopyYUV422.dstPitch             = dst.step;
-
-        stCopyYUV422.WidthInBytes         = frameSize.width * 2;
-        stCopyYUV422.Height               = frameSize.height;
-
-        // DMA Luma/Chroma
-        res = cuMemcpy2D(&stCopyYUV422);
-        CV_Assert( res == CUDA_SUCCESS );
-    }
-
-    // YV12/IYUV are both 4:2:0 planar formats (12bpc)
-    // Luma, U, V chroma planar (12bpc), chroma is subsampled (w/2,h/2)
-    void copyYV12orIYUVFrame(cv::Size frameSize, const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst)
-    {
-        CUresult res;
-
-        // Source is YV12/IYUV, this native format is converted to NV12 format by the video encoder
-
-        // (1) luma copy setup
-        CUDA_MEMCPY2D stCopyLuma;
-        memset((void*)&stCopyLuma, 0, sizeof(stCopyLuma));
-        stCopyLuma.srcXInBytes          = 0;
-        stCopyLuma.srcY                 = 0;
-        stCopyLuma.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyLuma.srcHost              = 0;
-        stCopyLuma.srcDevice            = (CUdeviceptr) src.data;
-        stCopyLuma.srcArray             = 0;
-        stCopyLuma.srcPitch             = src.step;
-
-        stCopyLuma.dstXInBytes          = 0;
-        stCopyLuma.dstY                 = 0;
-        stCopyLuma.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyLuma.dstHost              = 0;
-        stCopyLuma.dstDevice            = (CUdeviceptr) dst.data;
-        stCopyLuma.dstArray             = 0;
-        stCopyLuma.dstPitch             = dst.step;
-
-        stCopyLuma.WidthInBytes         = frameSize.width;
-        stCopyLuma.Height               = frameSize.height;
-
-        // (2) chroma copy setup, U/V can be done together
-        CUDA_MEMCPY2D stCopyChroma;
-        memset((void*)&stCopyChroma, 0, sizeof(stCopyChroma));
-        stCopyChroma.srcXInBytes        = 0;
-        stCopyChroma.srcY               = frameSize.height << 1; // U/V chroma offset
-        stCopyChroma.srcMemoryType      = CU_MEMORYTYPE_DEVICE;
-        stCopyChroma.srcHost            = 0;
-        stCopyChroma.srcDevice          = (CUdeviceptr) src.data;
-        stCopyChroma.srcArray           = 0;
-        stCopyChroma.srcPitch           = src.step >> 1; // chroma is subsampled by 2 (but it has U/V are next to each other)
-
-        stCopyChroma.dstXInBytes        = 0;
-        stCopyChroma.dstY               = frameSize.height << 1; // chroma offset (srcY*srcPitch now points to the chroma planes)
-        stCopyChroma.dstMemoryType      = CU_MEMORYTYPE_DEVICE;
-        stCopyChroma.dstHost            = 0;
-        stCopyChroma.dstDevice          = (CUdeviceptr) dst.data;
-        stCopyChroma.dstArray           = 0;
-        stCopyChroma.dstPitch           = dst.step >> 1;
-
-        stCopyChroma.WidthInBytes       = frameSize.width >> 1;
-        stCopyChroma.Height             = frameSize.height; // U/V are sent together
-
-        // DMA Luma
-        res = cuMemcpy2D(&stCopyLuma);
-        CV_Assert( res == CUDA_SUCCESS );
-
-        // DMA Chroma channels (UV side by side)
-        res = cuMemcpy2D(&stCopyChroma);
-        CV_Assert( res == CUDA_SUCCESS );
-    }
-
-    // NV12 is 4:2:0 format (12bpc)
-    // Luma followed by U/V chroma interleaved (12bpc), chroma is subsampled (w/2,h/2)
-    void copyNV12Frame(cv::Size frameSize, const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst)
-    {
-        CUresult res;
-
-        // Source is NV12 in pitch linear memory
-        // Because we are assume input is NV12 (if we take input in the native format), the encoder handles NV12 as a native format in pitch linear memory
-
-        // Luma/Chroma can be done in a single transfer
-        CUDA_MEMCPY2D stCopyNV12;
-        memset((void*)&stCopyNV12, 0, sizeof(stCopyNV12));
-        stCopyNV12.srcXInBytes          = 0;
-        stCopyNV12.srcY                 = 0;
-        stCopyNV12.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyNV12.srcHost              = 0;
-        stCopyNV12.srcDevice            = (CUdeviceptr) src.data;
-        stCopyNV12.srcArray             = 0;
-        stCopyNV12.srcPitch             = src.step;
-
-        stCopyNV12.dstXInBytes          = 0;
-        stCopyNV12.dstY                 = 0;
-        stCopyNV12.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
-        stCopyNV12.dstHost              = 0;
-        stCopyNV12.dstDevice            = (CUdeviceptr) dst.data;
-        stCopyNV12.dstArray             = 0;
-        stCopyNV12.dstPitch             = dst.step;
-
-        stCopyNV12.WidthInBytes         = frameSize.width;
-        stCopyNV12.Height               =(frameSize.height * 3) >> 1;
-
-        // DMA Luma/Chroma
-        res = cuMemcpy2D(&stCopyNV12);
-        CV_Assert( res == CUDA_SUCCESS );
-    }
-}
-
-void cv::gpu::VideoWriter_GPU::Impl::write(const cv::gpu::GpuMat& frame, bool lastFrame)
-{
-    if (inputFormat_ == SF_BGR)
-    {
-        CV_Assert( frame.size() == frameSize_ );
-        CV_Assert( frame.type() == CV_8UC1 || frame.type() == CV_8UC3 || frame.type() == CV_8UC4 );
-    }
-    else
-    {
-        CV_Assert( frame.size() == videoFrame_.size() );
-        CV_Assert( frame.type() == videoFrame_.type() );
-    }
-
-    NVVE_EncodeFrameParams efparams;
-    efparams.Width = frameSize_.width;
-    efparams.Height = frameSize_.height;
-    efparams.Pitch = static_cast<int>(videoFrame_.step);
-    efparams.SurfFmt = surfaceFormat_;
-    efparams.PictureStruc = FRAME_PICTURE;
-    efparams.topfieldfirst =  0;
-    efparams.repeatFirstField = 0;
-    efparams.progressiveFrame = (surfaceFormat_ == NV12) ? 1 : 0;
-    efparams.bLast = lastFrame;
-    efparams.picBuf = 0; // Must be set to NULL in order to support device memory input
-
-    // Don't forget we need to lock/unlock between memcopies
-    CUresult res = cuvidCtxLock(cuCtxLock_, 0);
-    CV_Assert( res == CUDA_SUCCESS );
-
-    if (inputFormat_ == SF_BGR)
-        cv::gpu::device::video_encoding::YV12_gpu(frame, frame.channels(), videoFrame_);
-    else
-    {
-        switch (surfaceFormat_)
-        {
-        case UYVY: // UYVY (4:2:2)
-        case YUY2: // YUY2 (4:2:2)
-            copyUYVYorYUY2Frame(frameSize_, frame, videoFrame_);
-            break;
-
-        case YV12: // YV12 (4:2:0), Y V U
-        case IYUV: // IYUV (4:2:0), Y U V
-            copyYV12orIYUVFrame(frameSize_, frame, videoFrame_);
-            break;
-
-        case NV12: // NV12 (4:2:0)
-            copyNV12Frame(frameSize_, frame, videoFrame_);
-            break;
-        }
-    }
-
-    res = cuvidCtxUnlock(cuCtxLock_, 0);
-    CV_Assert( res == CUDA_SUCCESS );
-
-    int err = NVEncodeFrame(encoder_, &efparams, 0, videoFrame_.data);
-    CV_Assert( err == 0 );
-}
-
-unsigned char* NVENCAPI cv::gpu::VideoWriter_GPU::Impl::HandleAcquireBitStream(int* pBufferSize, void* pUserdata)
-{
-    Impl* thiz = static_cast<Impl*>(pUserdata);
-
-    return thiz->callback_->acquireBitStream(pBufferSize);
-}
-
-void NVENCAPI cv::gpu::VideoWriter_GPU::Impl::HandleReleaseBitStream(int nBytesInBuffer, unsigned char* cb, void* pUserdata)
-{
-    Impl* thiz = static_cast<Impl*>(pUserdata);
-
-    thiz->callback_->releaseBitStream(cb, nBytesInBuffer);
-}
-
-void NVENCAPI cv::gpu::VideoWriter_GPU::Impl::HandleOnBeginFrame(const NVVE_BeginFrameInfo* pbfi, void* pUserdata)
-{
-    Impl* thiz = static_cast<Impl*>(pUserdata);
-
-    thiz->callback_->onBeginFrame(pbfi->nFrameNumber, static_cast<EncoderCallBack::PicType>(pbfi->nPicType));
-}
-
-void NVENCAPI cv::gpu::VideoWriter_GPU::Impl::HandleOnEndFrame(const NVVE_EndFrameInfo* pefi, void* pUserdata)
-{
-    Impl* thiz = static_cast<Impl*>(pUserdata);
-
-    thiz->callback_->onEndFrame(pefi->nFrameNumber, static_cast<EncoderCallBack::PicType>(pefi->nPicType));
-}
-
-///////////////////////////////////////////////////////////////////////////
-// FFMPEG
-
-class EncoderCallBackFFMPEG : public cv::gpu::VideoWriter_GPU::EncoderCallBack
-{
-public:
-    EncoderCallBackFFMPEG(const std::string& fileName, cv::Size frameSize, double fps);
-    ~EncoderCallBackFFMPEG();
-
-    unsigned char* acquireBitStream(int* bufferSize);
-    void releaseBitStream(unsigned char* data, int size);
-    void onBeginFrame(int frameNumber, PicType picType);
-    void onEndFrame(int frameNumber, PicType picType);
-
-private:
-    EncoderCallBackFFMPEG(const EncoderCallBackFFMPEG&);
-    EncoderCallBackFFMPEG& operator=(const EncoderCallBackFFMPEG&);
-
-    struct OutputMediaStream_FFMPEG* stream_;
-    std::vector<uchar> buf_;
-    bool isKeyFrame_;
-};
-
-namespace
-{
-    Create_OutputMediaStream_FFMPEG_Plugin create_OutputMediaStream_FFMPEG_p = 0;
-    Release_OutputMediaStream_FFMPEG_Plugin release_OutputMediaStream_FFMPEG_p = 0;
-    Write_OutputMediaStream_FFMPEG_Plugin write_OutputMediaStream_FFMPEG_p = 0;
-
-    bool init_MediaStream_FFMPEG()
-    {
-        static bool initialized = 0;
-
-        if (!initialized)
-        {
-            #if defined WIN32 || defined _WIN32
-                const char* module_name = "opencv_ffmpeg"
-                    CVAUX_STR(CV_VERSION_EPOCH) CVAUX_STR(CV_VERSION_MAJOR) CVAUX_STR(CV_VERSION_MINOR)
-                #if (defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__)
-                    "_64"
-                #endif
-                    ".dll";
-
-                static HMODULE cvFFOpenCV = LoadLibrary(module_name);
-
-                if (cvFFOpenCV)
-                {
-                    create_OutputMediaStream_FFMPEG_p =
-                        (Create_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "create_OutputMediaStream_FFMPEG");
-                    release_OutputMediaStream_FFMPEG_p =
-                        (Release_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "release_OutputMediaStream_FFMPEG");
-                    write_OutputMediaStream_FFMPEG_p =
-                        (Write_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "write_OutputMediaStream_FFMPEG");
-
-                    initialized = create_OutputMediaStream_FFMPEG_p != 0 && release_OutputMediaStream_FFMPEG_p != 0 && write_OutputMediaStream_FFMPEG_p != 0;
-                }
-            #elif defined HAVE_FFMPEG
-                create_OutputMediaStream_FFMPEG_p = create_OutputMediaStream_FFMPEG;
-                release_OutputMediaStream_FFMPEG_p = release_OutputMediaStream_FFMPEG;
-                write_OutputMediaStream_FFMPEG_p = write_OutputMediaStream_FFMPEG;
-
-                initialized = true;
-            #endif
-        }
-
-        return initialized;
-    }
-}
-
-EncoderCallBackFFMPEG::EncoderCallBackFFMPEG(const std::string& fileName, cv::Size frameSize, double fps) :
-    stream_(0), isKeyFrame_(false)
-{
-    int buf_size = std::max(frameSize.area() * 4, 1024 * 1024);
-    buf_.resize(buf_size);
-
-    CV_Assert( init_MediaStream_FFMPEG() );
-
-    stream_ = create_OutputMediaStream_FFMPEG_p(fileName.c_str(), frameSize.width, frameSize.height, fps);
-    CV_Assert( stream_ != 0 );
-}
-
-EncoderCallBackFFMPEG::~EncoderCallBackFFMPEG()
-{
-    release_OutputMediaStream_FFMPEG_p(stream_);
-}
-
-unsigned char* EncoderCallBackFFMPEG::acquireBitStream(int* bufferSize)
-{
-    *bufferSize = static_cast<int>(buf_.size());
-    return &buf_[0];
-}
-
-void EncoderCallBackFFMPEG::releaseBitStream(unsigned char* data, int size)
-{
-    write_OutputMediaStream_FFMPEG_p(stream_, data, size, isKeyFrame_);
-}
-
-void EncoderCallBackFFMPEG::onBeginFrame(int frameNumber, PicType picType)
-{
-    isKeyFrame_ = picType == IFRAME;
-}
-
-void EncoderCallBackFFMPEG::onEndFrame(int frameNumber, PicType picType)
-{
-}
-
-///////////////////////////////////////////////////////////////////////////
-// VideoWriter_GPU
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU()
-{
-}
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format)
-{
-    open(fileName, frameSize, fps, format);
-}
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
-{
-    open(fileName, frameSize, fps, params, format);
-}
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format)
-{
-    open(encoderCallback, frameSize, fps, format);
-}
-
-cv::gpu::VideoWriter_GPU::VideoWriter_GPU(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
-{
-    open(encoderCallback, frameSize, fps, params, format);
-}
-
-cv::gpu::VideoWriter_GPU::~VideoWriter_GPU()
-{
-    close();
-}
-
-void cv::gpu::VideoWriter_GPU::open(const std::string& fileName, cv::Size frameSize, double fps, SurfaceFormat format)
-{
-    close();
-    cv::Ptr<EncoderCallBack> encoderCallback(new EncoderCallBackFFMPEG(fileName, frameSize, fps));
-    open(encoderCallback, frameSize, fps, format);
-}
-
-void cv::gpu::VideoWriter_GPU::open(const std::string& fileName, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
-{
-    close();
-    cv::Ptr<EncoderCallBack> encoderCallback(new EncoderCallBackFFMPEG(fileName, frameSize, fps));
-    open(encoderCallback, frameSize, fps, params, format);
-}
-
-void cv::gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, SurfaceFormat format)
-{
-    close();
-    impl_.reset(new Impl(encoderCallback, frameSize, fps, format));
-}
-
-void cv::gpu::VideoWriter_GPU::open(const cv::Ptr<EncoderCallBack>& encoderCallback, cv::Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
-{
-    close();
-    impl_.reset(new Impl(encoderCallback, frameSize, fps, params, format));
-}
-
-bool cv::gpu::VideoWriter_GPU::isOpened() const
-{
-    return impl_.get() != 0;
-}
-
-void cv::gpu::VideoWriter_GPU::close()
-{
-    impl_.reset();
-}
-
-void cv::gpu::VideoWriter_GPU::write(const cv::gpu::GpuMat& image, bool lastFrame)
-{
-    CV_Assert( isOpened() );
-
-    impl_->write(image, lastFrame);
-}
-
-cv::gpu::VideoWriter_GPU::EncoderParams cv::gpu::VideoWriter_GPU::getParams() const
-{
-    CV_Assert( isOpened() );
-
-    return impl_->getParams();
-}
-
-///////////////////////////////////////////////////////////////////////////
-// VideoWriter_GPU::EncoderParams
-
-cv::gpu::VideoWriter_GPU::EncoderParams::EncoderParams()
-{
-    P_Interval = 3;
-    IDR_Period = 15;
-    DynamicGOP = 0;
-    RCType = 1;
-    AvgBitrate = 4000000;
-    PeakBitrate = 10000000;
-    QP_Level_Intra = 25;
-    QP_Level_InterP = 28;
-    QP_Level_InterB = 31;
-    DeblockMode = 1;
-    ProfileLevel = 65357;
-    ForceIntra = 0;
-    ForceIDR = 0;
-    ClearStat = 0;
-    DIMode = 1;
-    Presets = 2;
-    DisableCabac = 0;
-    NaluFramingType = 0;
-    DisableSPSPPS = 0;
-}
-
-cv::gpu::VideoWriter_GPU::EncoderParams::EncoderParams(const std::string& configFile)
-{
-    load(configFile);
-}
-
-void cv::gpu::VideoWriter_GPU::EncoderParams::load(const std::string& configFile)
-{
-    cv::FileStorage fs(configFile, cv::FileStorage::READ);
-    CV_Assert( fs.isOpened() );
-
-    cv::read(fs["P_Interval"     ], P_Interval, 3);
-    cv::read(fs["IDR_Period"     ], IDR_Period, 15);
-    cv::read(fs["DynamicGOP"     ], DynamicGOP, 0);
-    cv::read(fs["RCType"         ], RCType, 1);
-    cv::read(fs["AvgBitrate"     ], AvgBitrate, 4000000);
-    cv::read(fs["PeakBitrate"    ], PeakBitrate, 10000000);
-    cv::read(fs["QP_Level_Intra" ], QP_Level_Intra, 25);
-    cv::read(fs["QP_Level_InterP"], QP_Level_InterP, 28);
-    cv::read(fs["QP_Level_InterB"], QP_Level_InterB, 31);
-    cv::read(fs["DeblockMode"    ], DeblockMode, 1);
-    cv::read(fs["ProfileLevel"   ], ProfileLevel, 65357);
-    cv::read(fs["ForceIntra"     ], ForceIntra, 0);
-    cv::read(fs["ForceIDR"       ], ForceIDR, 0);
-    cv::read(fs["ClearStat"      ], ClearStat, 0);
-    cv::read(fs["DIMode"         ], DIMode, 1);
-    cv::read(fs["Presets"        ], Presets, 2);
-    cv::read(fs["DisableCabac"   ], DisableCabac, 0);
-    cv::read(fs["NaluFramingType"], NaluFramingType, 0);
-    cv::read(fs["DisableSPSPPS"  ], DisableSPSPPS, 0);
-}
-
-void cv::gpu::VideoWriter_GPU::EncoderParams::save(const std::string& configFile) const
-{
-    cv::FileStorage fs(configFile, cv::FileStorage::WRITE);
-    CV_Assert( fs.isOpened() );
-
-    cv::write(fs, "P_Interval"     , P_Interval);
-    cv::write(fs, "IDR_Period"     , IDR_Period);
-    cv::write(fs, "DynamicGOP"     , DynamicGOP);
-    cv::write(fs, "RCType"         , RCType);
-    cv::write(fs, "AvgBitrate"     , AvgBitrate);
-    cv::write(fs, "PeakBitrate"    , PeakBitrate);
-    cv::write(fs, "QP_Level_Intra" , QP_Level_Intra);
-    cv::write(fs, "QP_Level_InterP", QP_Level_InterP);
-    cv::write(fs, "QP_Level_InterB", QP_Level_InterB);
-    cv::write(fs, "DeblockMode"    , DeblockMode);
-    cv::write(fs, "ProfileLevel"   , ProfileLevel);
-    cv::write(fs, "ForceIntra"     , ForceIntra);
-    cv::write(fs, "ForceIDR"       , ForceIDR);
-    cv::write(fs, "ClearStat"      , ClearStat);
-    cv::write(fs, "DIMode"         , DIMode);
-    cv::write(fs, "Presets"        , Presets);
-    cv::write(fs, "DisableCabac"   , DisableCabac);
-    cv::write(fs, "NaluFramingType", NaluFramingType);
-    cv::write(fs, "DisableSPSPPS"  , DisableSPSPPS);
-}
-
-#endif // !defined HAVE_CUDA || !defined WIN32
diff --git a/modules/gpu/test/main_test_nvidia.h b/modules/gpu/test/main_test_nvidia.h
deleted file mode 100644
index 173cbdbce1..0000000000
--- a/modules/gpu/test/main_test_nvidia.h
+++ /dev/null
@@ -1,25 +0,0 @@
-#ifndef __main_test_nvidia_h__
-#define __main_test_nvidia_h__
-
-enum OutputLevel
-{
-    OutputLevelNone,
-    OutputLevelCompact,
-    OutputLevelFull
-};
-
-extern OutputLevel nvidiaTestOutputLevel;
-
-bool nvidia_NPPST_Integral_Image(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NPPST_Squared_Integral_Image(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NPPST_RectStdDev(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NPPST_Resize(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NPPST_Vector_Operations(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NPPST_Transpose(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NCV_Vector_Operations(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NCV_Haar_Cascade_Loader(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NCV_Haar_Cascade_Application(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NCV_Hypotheses_Filtration(const std::string& test_data_path, OutputLevel outputLevel);
-bool nvidia_NCV_Visualization(const std::string& test_data_path, OutputLevel outputLevel);
-
-#endif
diff --git a/modules/gpu/test/nvidia/TestCompact.h b/modules/gpu/test/nvidia/TestCompact.h
deleted file mode 100644
index 8bb178428e..0000000000
--- a/modules/gpu/test/nvidia/TestCompact.h
+++ /dev/null
@@ -1,41 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testhypothesescompact_h_
-#define _testhypothesescompact_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestCompact : public NCVTestProvider
-{
-public:
-
-    TestCompact(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
-                          Ncv32u length, Ncv32u badElem, Ncv32u badElemPercentage);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestCompact(const TestCompact&);
-    TestCompact& operator=(const TestCompact&);
-
-
-    NCVTestSourceProvider<Ncv32u> &src;
-    Ncv32u length;
-    Ncv32u badElem;
-    Ncv32u badElemPercentage;
-};
-
-#endif // _testhypothesescompact_h_
diff --git a/modules/gpu/test/nvidia/TestDrawRects.h b/modules/gpu/test/nvidia/TestDrawRects.h
deleted file mode 100644
index 79aa1b51bf..0000000000
--- a/modules/gpu/test/nvidia/TestDrawRects.h
+++ /dev/null
@@ -1,44 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testdrawrects_h_
-#define _testdrawrects_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-template <class T>
-class TestDrawRects : public NCVTestProvider
-{
-public:
-
-    TestDrawRects(std::string testName, NCVTestSourceProvider<T> &src, NCVTestSourceProvider<Ncv32u> &src32u,
-                  Ncv32u width, Ncv32u height, Ncv32u numRects, T color);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-
-    TestDrawRects(const TestDrawRects&);
-    TestDrawRects& operator=(const TestDrawRects&);
-
-    NCVTestSourceProvider<T> &src;
-    NCVTestSourceProvider<Ncv32u> &src32u;
-    Ncv32u width;
-    Ncv32u height;
-    Ncv32u numRects;
-    T color;
-};
-
-#endif // _testdrawrects_h_
diff --git a/modules/gpu/test/nvidia/TestHaarCascadeApplication.h b/modules/gpu/test/nvidia/TestHaarCascadeApplication.h
deleted file mode 100644
index efc25cb591..0000000000
--- a/modules/gpu/test/nvidia/TestHaarCascadeApplication.h
+++ /dev/null
@@ -1,41 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testhaarcascadeapplication_h_
-#define _testhaarcascadeapplication_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestHaarCascadeApplication : public NCVTestProvider
-{
-public:
-
-    TestHaarCascadeApplication(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
-                               std::string cascadeName, Ncv32u width, Ncv32u height);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestHaarCascadeApplication(const TestHaarCascadeApplication&);
-    TestHaarCascadeApplication& operator=(const TestHaarCascadeApplication&);
-
-
-    NCVTestSourceProvider<Ncv8u> &src;
-    std::string cascadeName;
-    Ncv32u width;
-    Ncv32u height;
-};
-
-#endif // _testhaarcascadeapplication_h_
diff --git a/modules/gpu/test/nvidia/TestHaarCascadeLoader.h b/modules/gpu/test/nvidia/TestHaarCascadeLoader.h
deleted file mode 100644
index 44054d3484..0000000000
--- a/modules/gpu/test/nvidia/TestHaarCascadeLoader.h
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testhaarcascadeloader_h_
-#define _testhaarcascadeloader_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestHaarCascadeLoader : public NCVTestProvider
-{
-public:
-
-    TestHaarCascadeLoader(std::string testName, std::string cascadeName);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-
-    std::string cascadeName;
-};
-
-#endif // _testhaarcascadeloader_h_
diff --git a/modules/gpu/test/nvidia/TestHypothesesFilter.h b/modules/gpu/test/nvidia/TestHypothesesFilter.h
deleted file mode 100644
index 2084b3bd4b..0000000000
--- a/modules/gpu/test/nvidia/TestHypothesesFilter.h
+++ /dev/null
@@ -1,44 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testhypothesesfilter_h_
-#define _testhypothesesfilter_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestHypothesesFilter : public NCVTestProvider
-{
-public:
-
-    TestHypothesesFilter(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
-                         Ncv32u numDstRects, Ncv32u minNeighbors, Ncv32f eps);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-
-    TestHypothesesFilter(const TestHypothesesFilter&);
-    TestHypothesesFilter& operator=(const TestHypothesesFilter&);
-
-    NCVTestSourceProvider<Ncv32u> &src;
-    Ncv32u numDstRects;
-    Ncv32u minNeighbors;
-    Ncv32f eps;
-
-    Ncv32u canvasWidth;
-    Ncv32u canvasHeight;
-};
-
-#endif // _testhypothesesfilter_h_
diff --git a/modules/gpu/test/nvidia/TestHypothesesGrow.h b/modules/gpu/test/nvidia/TestHypothesesGrow.h
deleted file mode 100644
index d2f2fdaace..0000000000
--- a/modules/gpu/test/nvidia/TestHypothesesGrow.h
+++ /dev/null
@@ -1,46 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testhypothesesgrow_h_
-#define _testhypothesesgrow_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestHypothesesGrow : public NCVTestProvider
-{
-public:
-
-    TestHypothesesGrow(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
-                       Ncv32u rectWidth, Ncv32u rectHeight, Ncv32f rectScale,
-                       Ncv32u maxLenSrc, Ncv32u lenSrc, Ncv32u maxLenDst, Ncv32u lenDst);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestHypothesesGrow(const TestHypothesesGrow&);
-    TestHypothesesGrow& operator=(const TestHypothesesGrow&);
-
-
-    NCVTestSourceProvider<Ncv32u> &src;
-    Ncv32u rectWidth;
-    Ncv32u rectHeight;
-    Ncv32f rectScale;
-    Ncv32u maxLenSrc;
-    Ncv32u lenSrc;
-    Ncv32u maxLenDst;
-    Ncv32u lenDst;
-};
-
-#endif // _testhypothesesgrow_h_
diff --git a/modules/gpu/test/nvidia/TestIntegralImage.h b/modules/gpu/test/nvidia/TestIntegralImage.h
deleted file mode 100644
index 4dba575fe9..0000000000
--- a/modules/gpu/test/nvidia/TestIntegralImage.h
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testintegralimage_h_
-#define _testintegralimage_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-template <class T_in, class T_out>
-class TestIntegralImage : public NCVTestProvider
-{
-public:
-
-    TestIntegralImage(std::string testName, NCVTestSourceProvider<T_in> &src,
-                      Ncv32u width, Ncv32u height);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestIntegralImage(const TestIntegralImage&);
-    TestIntegralImage& operator=(const TestIntegralImage&);
-
-    NCVTestSourceProvider<T_in> &src;
-    Ncv32u width;
-    Ncv32u height;
-};
-
-#endif // _testintegralimage_h_
diff --git a/modules/gpu/test/nvidia/TestIntegralImageSquared.h b/modules/gpu/test/nvidia/TestIntegralImageSquared.h
deleted file mode 100644
index 8d20d732db..0000000000
--- a/modules/gpu/test/nvidia/TestIntegralImageSquared.h
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testintegralimagesquared_h_
-#define _testintegralimagesquared_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestIntegralImageSquared : public NCVTestProvider
-{
-public:
-
-    TestIntegralImageSquared(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
-                             Ncv32u width, Ncv32u height);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestIntegralImageSquared(const TestIntegralImageSquared&);
-    TestIntegralImageSquared& operator=(const TestIntegralImageSquared&);
-
-    NCVTestSourceProvider<Ncv8u> &src;
-    Ncv32u width;
-    Ncv32u height;
-};
-
-#endif // _testintegralimagesquared_h_
diff --git a/modules/gpu/test/nvidia/TestRectStdDev.h b/modules/gpu/test/nvidia/TestRectStdDev.h
deleted file mode 100644
index d1f2e759e4..0000000000
--- a/modules/gpu/test/nvidia/TestRectStdDev.h
+++ /dev/null
@@ -1,44 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testrectstddev_h_
-#define _testrectstddev_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-class TestRectStdDev : public NCVTestProvider
-{
-public:
-
-    TestRectStdDev(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
-                   Ncv32u width, Ncv32u height, NcvRect32u rect, Ncv32f scaleFactor,
-                   NcvBool bTextureCache);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestRectStdDev(const TestRectStdDev&);
-    TestRectStdDev& operator=(const TestRectStdDev&);
-
-    NCVTestSourceProvider<Ncv8u> &src;
-    Ncv32u width;
-    Ncv32u height;
-    NcvRect32u rect;
-    Ncv32f scaleFactor;
-
-    NcvBool bTextureCache;
-};
-
-#endif // _testrectstddev_h_
diff --git a/modules/gpu/test/nvidia/TestResize.h b/modules/gpu/test/nvidia/TestResize.h
deleted file mode 100644
index ee6eb617e8..0000000000
--- a/modules/gpu/test/nvidia/TestResize.h
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testresize_h_
-#define _testresize_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-template <class T>
-class TestResize : public NCVTestProvider
-{
-public:
-
-    TestResize(std::string testName, NCVTestSourceProvider<T> &src,
-               Ncv32u width, Ncv32u height, Ncv32u scaleFactor, NcvBool bTextureCache);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-    TestResize(const TestResize&);
-    TestResize& operator=(const TestResize&);
-
-    NCVTestSourceProvider<T> &src;
-    Ncv32u width;
-    Ncv32u height;
-    Ncv32u scaleFactor;
-
-    NcvBool bTextureCache;
-};
-
-#endif // _testresize_h_
diff --git a/modules/gpu/test/nvidia/TestTranspose.h b/modules/gpu/test/nvidia/TestTranspose.h
deleted file mode 100644
index 2f8ccf41c2..0000000000
--- a/modules/gpu/test/nvidia/TestTranspose.h
+++ /dev/null
@@ -1,41 +0,0 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-#ifndef _testtranspose_h_
-#define _testtranspose_h_
-
-#include "NCVTest.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-
-template <class T>
-class TestTranspose : public NCVTestProvider
-{
-public:
-
-    TestTranspose(std::string testName, NCVTestSourceProvider<T> &src,
-                  Ncv32u width, Ncv32u height);
-
-    virtual bool init();
-    virtual bool process();
-    virtual bool deinit();
-    virtual bool toString(std::ofstream &strOut);
-
-private:
-
-    TestTranspose(const TestTranspose&);
-    TestTranspose& operator=(const TestTranspose&);
-
-    NCVTestSourceProvider<T> &src;
-    Ncv32u width;
-    Ncv32u height;
-};
-
-#endif // _testtranspose_h_
diff --git a/modules/gpu/test/test_calib3d.cpp b/modules/gpu/test/test_calib3d.cpp
index 318de8d898..3ad19dcbea 100644
--- a/modules/gpu/test/test_calib3d.cpp
+++ b/modules/gpu/test/test_calib3d.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,122 +44,7 @@
 
 #ifdef HAVE_CUDA
 
-//////////////////////////////////////////////////////////////////////////
-// StereoBM
-
-struct StereoBM : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(StereoBM, Regression)
-{
-    cv::Mat left_image  = readImage("stereobm/aloe-L.png", cv::IMREAD_GRAYSCALE);
-    cv::Mat right_image = readImage("stereobm/aloe-R.png", cv::IMREAD_GRAYSCALE);
-    cv::Mat disp_gold   = readImage("stereobm/aloe-disp.png", cv::IMREAD_GRAYSCALE);
-
-    ASSERT_FALSE(left_image.empty());
-    ASSERT_FALSE(right_image.empty());
-    ASSERT_FALSE(disp_gold.empty());
-
-    cv::gpu::StereoBM_GPU bm(0, 128, 19);
-    cv::gpu::GpuMat disp;
-
-    bm(loadMat(left_image), loadMat(right_image), disp);
-
-    EXPECT_MAT_NEAR(disp_gold, disp, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Calib3D, StereoBM, ALL_DEVICES);
-
-//////////////////////////////////////////////////////////////////////////
-// StereoBeliefPropagation
-
-struct StereoBeliefPropagation : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(StereoBeliefPropagation, Regression)
-{
-    cv::Mat left_image  = readImage("stereobp/aloe-L.png");
-    cv::Mat right_image = readImage("stereobp/aloe-R.png");
-    cv::Mat disp_gold   = readImage("stereobp/aloe-disp.png", cv::IMREAD_GRAYSCALE);
-
-    ASSERT_FALSE(left_image.empty());
-    ASSERT_FALSE(right_image.empty());
-    ASSERT_FALSE(disp_gold.empty());
-
-    cv::gpu::StereoBeliefPropagation bp(64, 8, 2, 25, 0.1f, 15, 1, CV_16S);
-    cv::gpu::GpuMat disp;
-
-    bp(loadMat(left_image), loadMat(right_image), disp);
-
-    cv::Mat h_disp(disp);
-    h_disp.convertTo(h_disp, disp_gold.depth());
-
-    EXPECT_MAT_NEAR(disp_gold, h_disp, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Calib3D, StereoBeliefPropagation, ALL_DEVICES);
-
-//////////////////////////////////////////////////////////////////////////
-// StereoConstantSpaceBP
-
-struct StereoConstantSpaceBP : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(StereoConstantSpaceBP, Regression)
-{
-    cv::Mat left_image  = readImage("csstereobp/aloe-L.png");
-    cv::Mat right_image = readImage("csstereobp/aloe-R.png");
-
-    cv::Mat disp_gold;
-
-    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
-        disp_gold = readImage("csstereobp/aloe-disp.png", cv::IMREAD_GRAYSCALE);
-    else
-        disp_gold = readImage("csstereobp/aloe-disp_CC1X.png", cv::IMREAD_GRAYSCALE);
-
-    ASSERT_FALSE(left_image.empty());
-    ASSERT_FALSE(right_image.empty());
-    ASSERT_FALSE(disp_gold.empty());
-
-    cv::gpu::StereoConstantSpaceBP csbp(128, 16, 4, 4);
-    cv::gpu::GpuMat disp;
-
-    csbp(loadMat(left_image), loadMat(right_image), disp);
-
-    cv::Mat h_disp(disp);
-    h_disp.convertTo(h_disp, disp_gold.depth());
-
-    EXPECT_MAT_NEAR(disp_gold, h_disp, 1.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Calib3D, StereoConstantSpaceBP, ALL_DEVICES);
+using namespace cvtest;
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 // transformPoints
@@ -301,45 +187,4 @@ GPU_TEST_P(SolvePnPRansac, Accuracy)
 
 INSTANTIATE_TEST_CASE_P(GPU_Calib3D, SolvePnPRansac, ALL_DEVICES);
 
-////////////////////////////////////////////////////////////////////////////////
-// reprojectImageTo3D
-
-PARAM_TEST_CASE(ReprojectImageTo3D, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(ReprojectImageTo3D, Accuracy)
-{
-    cv::Mat disp = randomMat(size, depth, 5.0, 30.0);
-    cv::Mat Q = randomMat(cv::Size(4, 4), CV_32FC1, 0.1, 1.0);
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::reprojectImageTo3D(loadMat(disp, useRoi), dst, Q, 3);
-
-    cv::Mat dst_gold;
-    cv::reprojectImageTo3D(disp, dst_gold, Q, false);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 1e-5);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Calib3D, ReprojectImageTo3D, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatDepth(CV_8U), MatDepth(CV_16S)),
-    WHOLE_SUBMAT));
-
 #endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_copy_make_border.cpp b/modules/gpu/test/test_copy_make_border.cpp
deleted file mode 100644
index 0b59fe2d8a..0000000000
--- a/modules/gpu/test/test_copy_make_border.cpp
+++ /dev/null
@@ -1,103 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "test_precomp.hpp"
-
-#ifdef HAVE_CUDA
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(Border, int)
-}
-
-PARAM_TEST_CASE(CopyMakeBorder, cv::gpu::DeviceInfo, cv::Size, MatType, Border, BorderType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    int border;
-    int borderType;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        border = GET_PARAM(3);
-        borderType = GET_PARAM(4);
-        useRoi = GET_PARAM(5);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(CopyMakeBorder, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-    cv::Scalar val = randomScalar(0, 255);
-
-    cv::gpu::GpuMat dst = createMat(cv::Size(size.width + 2 * border, size.height + 2 * border), type, useRoi);
-    cv::gpu::copyMakeBorder(loadMat(src, useRoi), dst, border, border, border, border, borderType, val);
-
-    cv::Mat dst_gold;
-    cv::copyMakeBorder(src, dst_gold, border, border, border, border, borderType, val);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CopyMakeBorder, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1),
-                    MatType(CV_8UC3),
-                    MatType(CV_8UC4),
-                    MatType(CV_16UC1),
-                    MatType(CV_16UC3),
-                    MatType(CV_16UC4),
-                    MatType(CV_32FC1),
-                    MatType(CV_32FC3),
-                    MatType(CV_32FC4)),
-    testing::Values(Border(1), Border(10), Border(50)),
-    ALL_BORDER_TYPES,
-    WHOLE_SUBMAT));
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_denoising.cpp b/modules/gpu/test/test_denoising.cpp
deleted file mode 100644
index fe0c548c71..0000000000
--- a/modules/gpu/test/test_denoising.cpp
+++ /dev/null
@@ -1,182 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "test_precomp.hpp"
-
-#ifdef HAVE_CUDA
-
-////////////////////////////////////////////////////////
-// BilateralFilter
-
-PARAM_TEST_CASE(BilateralFilter, cv::gpu::DeviceInfo, cv::Size, MatType)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    int kernel_size;
-    float sigma_color;
-    float sigma_spatial;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-
-        kernel_size = 5;
-        sigma_color = 10.f;
-        sigma_spatial = 3.5f;
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(BilateralFilter, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-
-    src.convertTo(src, type);
-    cv::gpu::GpuMat dst;
-
-    cv::gpu::bilateralFilter(loadMat(src), dst, kernel_size, sigma_color, sigma_spatial);
-
-    cv::Mat dst_gold;
-    cv::bilateralFilter(src, dst_gold, kernel_size, sigma_color, sigma_spatial);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-3 : 1.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Denoising, BilateralFilter, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(cv::Size(128, 128), cv::Size(113, 113), cv::Size(639, 481)),
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_32FC1), MatType(CV_32FC3))
-    ));
-
-
-////////////////////////////////////////////////////////
-// Brute Force Non local means
-
-struct BruteForceNonLocalMeans: testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(BruteForceNonLocalMeans, Regression)
-{
-    using cv::gpu::GpuMat;
-
-    cv::Mat bgr  = readImage("denoising/lena_noised_gaussian_sigma=20_multi_0.png", cv::IMREAD_COLOR);
-    ASSERT_FALSE(bgr.empty());
-
-    cv::Mat gray;
-    cv::cvtColor(bgr, gray, CV_BGR2GRAY);
-
-    GpuMat dbgr, dgray;
-    cv::gpu::nonLocalMeans(GpuMat(bgr),  dbgr, 20);
-    cv::gpu::nonLocalMeans(GpuMat(gray), dgray, 20);
-
-#if 0
-    dumpImage("denoising/nlm_denoised_lena_bgr.png", cv::Mat(dbgr));
-    dumpImage("denoising/nlm_denoised_lena_gray.png", cv::Mat(dgray));
-#endif
-
-    cv::Mat bgr_gold  = readImage("denoising/nlm_denoised_lena_bgr.png", cv::IMREAD_COLOR);
-    cv::Mat gray_gold  = readImage("denoising/nlm_denoised_lena_gray.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(bgr_gold.empty() || gray_gold.empty());
-
-    EXPECT_MAT_NEAR(bgr_gold, dbgr, 1e-4);
-    EXPECT_MAT_NEAR(gray_gold, dgray, 1e-4);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Denoising, BruteForceNonLocalMeans, ALL_DEVICES);
-
-////////////////////////////////////////////////////////
-// Fast Force Non local means
-
-struct FastNonLocalMeans: testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(FastNonLocalMeans, Regression)
-{
-    using cv::gpu::GpuMat;
-
-    cv::Mat bgr  = readImage("denoising/lena_noised_gaussian_sigma=20_multi_0.png", cv::IMREAD_COLOR);
-    ASSERT_FALSE(bgr.empty());
-
-    cv::Mat gray;
-    cv::cvtColor(bgr, gray, CV_BGR2GRAY);
-
-    GpuMat dbgr, dgray;
-    cv::gpu::FastNonLocalMeansDenoising fnlmd;
-
-    fnlmd.simpleMethod(GpuMat(gray),  dgray, 20);
-    fnlmd.labMethod(GpuMat(bgr),  dbgr, 20, 10);
-
-#if 0
-    dumpImage("denoising/fnlm_denoised_lena_bgr.png", cv::Mat(dbgr));
-    dumpImage("denoising/fnlm_denoised_lena_gray.png", cv::Mat(dgray));
-#endif
-
-    cv::Mat bgr_gold  = readImage("denoising/fnlm_denoised_lena_bgr.png", cv::IMREAD_COLOR);
-    cv::Mat gray_gold  = readImage("denoising/fnlm_denoised_lena_gray.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(bgr_gold.empty() || gray_gold.empty());
-
-    EXPECT_MAT_NEAR(bgr_gold, dbgr, 1);
-    EXPECT_MAT_NEAR(gray_gold, dgray, 1);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Denoising, FastNonLocalMeans, ALL_DEVICES);
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_global_motion.cpp b/modules/gpu/test/test_global_motion.cpp
index 48fc428cee..a156f93107 100644
--- a/modules/gpu/test/test_global_motion.cpp
+++ b/modules/gpu/test/test_global_motion.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
diff --git a/modules/gpu/test/test_gpumat.cpp b/modules/gpu/test/test_gpumat.cpp
index 9ece87caa3..6fd4d69b43 100644
--- a/modules/gpu/test/test_gpumat.cpp
+++ b/modules/gpu/test/test_gpumat.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -44,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ////////////////////////////////////////////////////////////////////////////////
 // SetTo
 
@@ -88,7 +90,7 @@ GPU_TEST_P(SetTo, SameVal)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -113,7 +115,7 @@ GPU_TEST_P(SetTo, DifferentVal)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -140,7 +142,7 @@ GPU_TEST_P(SetTo, Masked)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -208,7 +210,7 @@ GPU_TEST_P(CopyTo, Masked)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -267,7 +269,7 @@ GPU_TEST_P(ConvertTo, WithOutScaling)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -299,7 +301,7 @@ GPU_TEST_P(ConvertTo, WithScaling)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
diff --git a/modules/gpu/test/test_imgproc.cpp b/modules/gpu/test/test_imgproc.cpp
deleted file mode 100644
index 41a94299bf..0000000000
--- a/modules/gpu/test/test_imgproc.cpp
+++ /dev/null
@@ -1,1154 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "test_precomp.hpp"
-
-#ifdef HAVE_CUDA
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// Integral
-
-PARAM_TEST_CASE(Integral, cv::gpu::DeviceInfo, cv::Size, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        useRoi = GET_PARAM(2);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Integral, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_8UC1);
-
-    cv::gpu::GpuMat dst = createMat(cv::Size(src.cols + 1, src.rows + 1), CV_32SC1, useRoi);
-    cv::gpu::integral(loadMat(src, useRoi), dst);
-
-    cv::Mat dst_gold;
-    cv::integral(src, dst_gold, CV_32S);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Integral, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    WHOLE_SUBMAT));
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// HistEven
-
-struct HistEven : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(HistEven, Accuracy)
-{
-    cv::Mat img = readImage("stereobm/aloe-L.png");
-    ASSERT_FALSE(img.empty());
-
-    cv::Mat hsv;
-    cv::cvtColor(img, hsv, CV_BGR2HSV);
-
-    int hbins = 30;
-    float hranges[] = {0.0f, 180.0f};
-
-    std::vector<cv::gpu::GpuMat> srcs;
-    cv::gpu::split(loadMat(hsv), srcs);
-
-    cv::gpu::GpuMat hist;
-    cv::gpu::histEven(srcs[0], hist, hbins, (int)hranges[0], (int)hranges[1]);
-
-    cv::MatND histnd;
-    int histSize[] = {hbins};
-    const float* ranges[] = {hranges};
-    int channels[] = {0};
-    cv::calcHist(&hsv, 1, channels, cv::Mat(), histnd, 1, histSize, ranges);
-
-    cv::Mat hist_gold = histnd;
-    hist_gold = hist_gold.t();
-    hist_gold.convertTo(hist_gold, CV_32S);
-
-    EXPECT_MAT_NEAR(hist_gold, hist, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, HistEven, ALL_DEVICES);
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// CalcHist
-
-namespace
-{
-    void calcHistGold(const cv::Mat& src, cv::Mat& hist)
-    {
-        hist.create(1, 256, CV_32SC1);
-        hist.setTo(cv::Scalar::all(0));
-
-        int* hist_row = hist.ptr<int>();
-        for (int y = 0; y < src.rows; ++y)
-        {
-            const uchar* src_row = src.ptr(y);
-
-            for (int x = 0; x < src.cols; ++x)
-                ++hist_row[src_row[x]];
-        }
-    }
-}
-
-PARAM_TEST_CASE(CalcHist, cv::gpu::DeviceInfo, cv::Size)
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    cv::Size size;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(CalcHist, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_8UC1);
-
-    cv::gpu::GpuMat hist;
-    cv::gpu::calcHist(loadMat(src), hist);
-
-    cv::Mat hist_gold;
-    calcHistGold(src, hist_gold);
-
-    EXPECT_MAT_NEAR(hist_gold, hist, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CalcHist, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES));
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// EqualizeHist
-
-PARAM_TEST_CASE(EqualizeHist, cv::gpu::DeviceInfo, cv::Size)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(EqualizeHist, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_8UC1);
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::equalizeHist(loadMat(src), dst);
-
-    cv::Mat dst_gold;
-    cv::equalizeHist(src, dst_gold);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 3.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, EqualizeHist, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES));
-
-////////////////////////////////////////////////////////////////////////
-// ColumnSum
-
-PARAM_TEST_CASE(ColumnSum, cv::gpu::DeviceInfo, cv::Size)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(ColumnSum, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_32FC1);
-
-    cv::gpu::GpuMat d_dst;
-    cv::gpu::columnSum(loadMat(src), d_dst);
-
-    cv::Mat dst(d_dst);
-
-    for (int j = 0; j < src.cols; ++j)
-    {
-        float gold = src.at<float>(0, j);
-        float res = dst.at<float>(0, j);
-        ASSERT_NEAR(res, gold, 1e-5);
-    }
-
-    for (int i = 1; i < src.rows; ++i)
-    {
-        for (int j = 0; j < src.cols; ++j)
-        {
-            float gold = src.at<float>(i, j) += src.at<float>(i - 1, j);
-            float res = dst.at<float>(i, j);
-            ASSERT_NEAR(res, gold, 1e-5);
-        }
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, ColumnSum, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES));
-
-////////////////////////////////////////////////////////
-// Canny
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(AppertureSize, int);
-    IMPLEMENT_PARAM_CLASS(L2gradient, bool);
-}
-
-PARAM_TEST_CASE(Canny, cv::gpu::DeviceInfo, AppertureSize, L2gradient, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    int apperture_size;
-    bool useL2gradient;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        apperture_size = GET_PARAM(1);
-        useL2gradient = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Canny, Accuracy)
-{
-    cv::Mat img = readImage("stereobm/aloe-L.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    double low_thresh = 50.0;
-    double high_thresh = 100.0;
-
-    if (!supportFeature(devInfo, cv::gpu::SHARED_ATOMICS))
-    {
-        try
-        {
-        cv::gpu::GpuMat edges;
-        cv::gpu::Canny(loadMat(img), edges, low_thresh, high_thresh, apperture_size, useL2gradient);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat edges;
-        cv::gpu::Canny(loadMat(img, useRoi), edges, low_thresh, high_thresh, apperture_size, useL2gradient);
-
-        cv::Mat edges_gold;
-        cv::Canny(img, edges_gold, low_thresh, high_thresh, apperture_size, useL2gradient);
-
-        EXPECT_MAT_SIMILAR(edges_gold, edges, 2e-2);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Canny, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(AppertureSize(3), AppertureSize(5)),
-    testing::Values(L2gradient(false), L2gradient(true)),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// MeanShift
-
-struct MeanShift : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    cv::Mat img;
-
-    int spatialRad;
-    int colorRad;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        img = readImageType("meanshift/cones.png", CV_8UC4);
-        ASSERT_FALSE(img.empty());
-
-        spatialRad = 30;
-        colorRad = 30;
-    }
-};
-
-GPU_TEST_P(MeanShift, Filtering)
-{
-    cv::Mat img_template;
-    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
-        img_template = readImage("meanshift/con_result.png");
-    else
-        img_template = readImage("meanshift/con_result_CC1X.png");
-    ASSERT_FALSE(img_template.empty());
-
-    cv::gpu::GpuMat d_dst;
-    cv::gpu::meanShiftFiltering(loadMat(img), d_dst, spatialRad, colorRad);
-
-    ASSERT_EQ(CV_8UC4, d_dst.type());
-
-    cv::Mat dst(d_dst);
-
-    cv::Mat result;
-    cv::cvtColor(dst, result, CV_BGRA2BGR);
-
-    EXPECT_MAT_NEAR(img_template, result, 0.0);
-}
-
-GPU_TEST_P(MeanShift, Proc)
-{
-    cv::FileStorage fs;
-    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
-        fs.open(std::string(cvtest::TS::ptr()->get_data_path()) + "meanshift/spmap.yaml", cv::FileStorage::READ);
-    else
-        fs.open(std::string(cvtest::TS::ptr()->get_data_path()) + "meanshift/spmap_CC1X.yaml", cv::FileStorage::READ);
-    ASSERT_TRUE(fs.isOpened());
-
-    cv::Mat spmap_template;
-    fs["spmap"] >> spmap_template;
-    ASSERT_FALSE(spmap_template.empty());
-
-    cv::gpu::GpuMat rmap_filtered;
-    cv::gpu::meanShiftFiltering(loadMat(img), rmap_filtered, spatialRad, colorRad);
-
-    cv::gpu::GpuMat rmap;
-    cv::gpu::GpuMat spmap;
-    cv::gpu::meanShiftProc(loadMat(img), rmap, spmap, spatialRad, colorRad);
-
-    ASSERT_EQ(CV_8UC4, rmap.type());
-
-    EXPECT_MAT_NEAR(rmap_filtered, rmap, 0.0);
-    EXPECT_MAT_NEAR(spmap_template, spmap, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MeanShift, ALL_DEVICES);
-
-////////////////////////////////////////////////////////////////////////////////
-// MeanShiftSegmentation
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(MinSize, int);
-}
-
-PARAM_TEST_CASE(MeanShiftSegmentation, cv::gpu::DeviceInfo, MinSize)
-{
-    cv::gpu::DeviceInfo devInfo;
-    int minsize;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        minsize = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MeanShiftSegmentation, Regression)
-{
-    cv::Mat img = readImageType("meanshift/cones.png", CV_8UC4);
-    ASSERT_FALSE(img.empty());
-
-    std::ostringstream path;
-    path << "meanshift/cones_segmented_sp10_sr10_minsize" << minsize;
-    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
-        path << ".png";
-    else
-        path << "_CC1X.png";
-    cv::Mat dst_gold = readImage(path.str());
-    ASSERT_FALSE(dst_gold.empty());
-
-    cv::Mat dst;
-    cv::gpu::meanShiftSegmentation(loadMat(img), dst, 10, 10, minsize);
-
-    cv::Mat dst_rgb;
-    cv::cvtColor(dst, dst_rgb, CV_BGRA2BGR);
-
-    EXPECT_MAT_SIMILAR(dst_gold, dst_rgb, 1e-3);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MeanShiftSegmentation, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(MinSize(0), MinSize(4), MinSize(20), MinSize(84), MinSize(340), MinSize(1364))));
-
-////////////////////////////////////////////////////////////////////////////
-// Blend
-
-namespace
-{
-    template <typename T>
-    void blendLinearGold(const cv::Mat& img1, const cv::Mat& img2, const cv::Mat& weights1, const cv::Mat& weights2, cv::Mat& result_gold)
-    {
-        result_gold.create(img1.size(), img1.type());
-
-        int cn = img1.channels();
-
-        for (int y = 0; y < img1.rows; ++y)
-        {
-            const float* weights1_row = weights1.ptr<float>(y);
-            const float* weights2_row = weights2.ptr<float>(y);
-            const T* img1_row = img1.ptr<T>(y);
-            const T* img2_row = img2.ptr<T>(y);
-            T* result_gold_row = result_gold.ptr<T>(y);
-
-            for (int x = 0; x < img1.cols * cn; ++x)
-            {
-                float w1 = weights1_row[x / cn];
-                float w2 = weights2_row[x / cn];
-                result_gold_row[x] = static_cast<T>((img1_row[x] * w1 + img2_row[x] * w2) / (w1 + w2 + 1e-5f));
-            }
-        }
-    }
-}
-
-PARAM_TEST_CASE(Blend, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Blend, Accuracy)
-{
-    int depth = CV_MAT_DEPTH(type);
-
-    cv::Mat img1 = randomMat(size, type, 0.0, depth == CV_8U ? 255.0 : 1.0);
-    cv::Mat img2 = randomMat(size, type, 0.0, depth == CV_8U ? 255.0 : 1.0);
-    cv::Mat weights1 = randomMat(size, CV_32F, 0, 1);
-    cv::Mat weights2 = randomMat(size, CV_32F, 0, 1);
-
-    cv::gpu::GpuMat result;
-    cv::gpu::blendLinear(loadMat(img1, useRoi), loadMat(img2, useRoi), loadMat(weights1, useRoi), loadMat(weights2, useRoi), result);
-
-    cv::Mat result_gold;
-    if (depth == CV_8U)
-        blendLinearGold<uchar>(img1, img2, weights1, weights2, result_gold);
-    else
-        blendLinearGold<float>(img1, img2, weights1, weights2, result_gold);
-
-    EXPECT_MAT_NEAR(result_gold, result, CV_MAT_DEPTH(type) == CV_8U ? 1.0 : 1e-5);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Blend, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////
-// Convolve
-
-namespace
-{
-    void convolveDFT(const cv::Mat& A, const cv::Mat& B, cv::Mat& C, bool ccorr = false)
-    {
-        // reallocate the output array if needed
-        C.create(std::abs(A.rows - B.rows) + 1, std::abs(A.cols - B.cols) + 1, A.type());
-        cv::Size dftSize;
-
-        // compute the size of DFT transform
-        dftSize.width = cv::getOptimalDFTSize(A.cols + B.cols - 1);
-        dftSize.height = cv::getOptimalDFTSize(A.rows + B.rows - 1);
-
-        // allocate temporary buffers and initialize them with 0s
-        cv::Mat tempA(dftSize, A.type(), cv::Scalar::all(0));
-        cv::Mat tempB(dftSize, B.type(), cv::Scalar::all(0));
-
-        // copy A and B to the top-left corners of tempA and tempB, respectively
-        cv::Mat roiA(tempA, cv::Rect(0, 0, A.cols, A.rows));
-        A.copyTo(roiA);
-        cv::Mat roiB(tempB, cv::Rect(0, 0, B.cols, B.rows));
-        B.copyTo(roiB);
-
-        // now transform the padded A & B in-place;
-        // use "nonzeroRows" hint for faster processing
-        cv::dft(tempA, tempA, 0, A.rows);
-        cv::dft(tempB, tempB, 0, B.rows);
-
-        // multiply the spectrums;
-        // the function handles packed spectrum representations well
-        cv::mulSpectrums(tempA, tempB, tempA, 0, ccorr);
-
-        // transform the product back from the frequency domain.
-        // Even though all the result rows will be non-zero,
-        // you need only the first C.rows of them, and thus you
-        // pass nonzeroRows == C.rows
-        cv::dft(tempA, tempA, cv::DFT_INVERSE + cv::DFT_SCALE, C.rows);
-
-        // now copy the result back to C.
-        tempA(cv::Rect(0, 0, C.cols, C.rows)).copyTo(C);
-    }
-
-    IMPLEMENT_PARAM_CLASS(KSize, int);
-    IMPLEMENT_PARAM_CLASS(Ccorr, bool);
-}
-
-PARAM_TEST_CASE(Convolve, cv::gpu::DeviceInfo, cv::Size, KSize, Ccorr)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int ksize;
-    bool ccorr;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        ksize = GET_PARAM(2);
-        ccorr = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Convolve, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_32FC1, 0.0, 100.0);
-    cv::Mat kernel = randomMat(cv::Size(ksize, ksize), CV_32FC1, 0.0, 1.0);
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::convolve(loadMat(src), loadMat(kernel), dst, ccorr);
-
-    cv::Mat dst_gold;
-    convolveDFT(src, kernel, dst_gold, ccorr);
-
-    EXPECT_MAT_NEAR(dst, dst_gold, 1e-1);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Convolve, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(KSize(3), KSize(7), KSize(11), KSize(17), KSize(19), KSize(23), KSize(45)),
-    testing::Values(Ccorr(false), Ccorr(true))));
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate8U
-
-CV_ENUM(TemplateMethod, cv::TM_SQDIFF, cv::TM_SQDIFF_NORMED, cv::TM_CCORR, cv::TM_CCORR_NORMED, cv::TM_CCOEFF, cv::TM_CCOEFF_NORMED)
-#define ALL_TEMPLATE_METHODS testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_SQDIFF_NORMED), TemplateMethod(cv::TM_CCORR), TemplateMethod(cv::TM_CCORR_NORMED), TemplateMethod(cv::TM_CCOEFF), TemplateMethod(cv::TM_CCOEFF_NORMED))
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(TemplateSize, cv::Size);
-}
-
-PARAM_TEST_CASE(MatchTemplate8U, cv::gpu::DeviceInfo, cv::Size, TemplateSize, Channels, TemplateMethod)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    cv::Size templ_size;
-    int cn;
-    int method;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        templ_size = GET_PARAM(2);
-        cn = GET_PARAM(3);
-        method = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MatchTemplate8U, Accuracy)
-{
-    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_8U, cn));
-    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_8U, cn));
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::matchTemplate(loadMat(image), loadMat(templ), dst, method);
-
-    cv::Mat dst_gold;
-    cv::matchTemplate(image, templ, dst_gold, method);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, templ_size.area() * 1e-1);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate8U, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
-    testing::Values(Channels(1), Channels(3), Channels(4)),
-    ALL_TEMPLATE_METHODS));
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate32F
-
-PARAM_TEST_CASE(MatchTemplate32F, cv::gpu::DeviceInfo, cv::Size, TemplateSize, Channels, TemplateMethod)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    cv::Size templ_size;
-    int cn;
-    int method;
-
-    int n, m, h, w;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        templ_size = GET_PARAM(2);
-        cn = GET_PARAM(3);
-        method = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MatchTemplate32F, Regression)
-{
-    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_32F, cn));
-    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_32F, cn));
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::matchTemplate(loadMat(image), loadMat(templ), dst, method);
-
-    cv::Mat dst_gold;
-    cv::matchTemplate(image, templ, dst_gold, method);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, templ_size.area() * 1e-1);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate32F, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
-    testing::Values(Channels(1), Channels(3), Channels(4)),
-    testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))));
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplateBlackSource
-
-PARAM_TEST_CASE(MatchTemplateBlackSource, cv::gpu::DeviceInfo, TemplateMethod)
-{
-    cv::gpu::DeviceInfo devInfo;
-    int method;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        method = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MatchTemplateBlackSource, Accuracy)
-{
-    cv::Mat image = readImage("matchtemplate/black.png");
-    ASSERT_FALSE(image.empty());
-
-    cv::Mat pattern = readImage("matchtemplate/cat.png");
-    ASSERT_FALSE(pattern.empty());
-
-    cv::gpu::GpuMat d_dst;
-    cv::gpu::matchTemplate(loadMat(image), loadMat(pattern), d_dst, method);
-
-    cv::Mat dst(d_dst);
-
-    double maxValue;
-    cv::Point maxLoc;
-    cv::minMaxLoc(dst, NULL, &maxValue, NULL, &maxLoc);
-
-    cv::Point maxLocGold = cv::Point(284, 12);
-
-    ASSERT_EQ(maxLocGold, maxLoc);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplateBlackSource, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(TemplateMethod(cv::TM_CCOEFF_NORMED), TemplateMethod(cv::TM_CCORR_NORMED))));
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate_CCOEF_NORMED
-
-PARAM_TEST_CASE(MatchTemplate_CCOEF_NORMED, cv::gpu::DeviceInfo, std::pair<std::string, std::string>)
-{
-    cv::gpu::DeviceInfo devInfo;
-    std::string imageName;
-    std::string patternName;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        imageName = GET_PARAM(1).first;
-        patternName = GET_PARAM(1).second;
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MatchTemplate_CCOEF_NORMED, Accuracy)
-{
-    cv::Mat image = readImage(imageName);
-    ASSERT_FALSE(image.empty());
-
-    cv::Mat pattern = readImage(patternName);
-    ASSERT_FALSE(pattern.empty());
-
-    cv::gpu::GpuMat d_dst;
-    cv::gpu::matchTemplate(loadMat(image), loadMat(pattern), d_dst, CV_TM_CCOEFF_NORMED);
-
-    cv::Mat dst(d_dst);
-
-    cv::Point minLoc, maxLoc;
-    double minVal, maxVal;
-    cv::minMaxLoc(dst, &minVal, &maxVal, &minLoc, &maxLoc);
-
-    cv::Mat dstGold;
-    cv::matchTemplate(image, pattern, dstGold, CV_TM_CCOEFF_NORMED);
-
-    double minValGold, maxValGold;
-    cv::Point minLocGold, maxLocGold;
-    cv::minMaxLoc(dstGold, &minValGold, &maxValGold, &minLocGold, &maxLocGold);
-
-    ASSERT_EQ(minLocGold, minLoc);
-    ASSERT_EQ(maxLocGold, maxLoc);
-    ASSERT_LE(maxVal, 1.0);
-    ASSERT_GE(minVal, -1.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate_CCOEF_NORMED, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(std::make_pair(std::string("matchtemplate/source-0.png"), std::string("matchtemplate/target-0.png")))));
-
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate_CanFindBigTemplate
-
-struct MatchTemplate_CanFindBigTemplate : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MatchTemplate_CanFindBigTemplate, SQDIFF_NORMED)
-{
-    cv::Mat scene = readImage("matchtemplate/scene.png");
-    ASSERT_FALSE(scene.empty());
-
-    cv::Mat templ = readImage("matchtemplate/template.png");
-    ASSERT_FALSE(templ.empty());
-
-    cv::gpu::GpuMat d_result;
-    cv::gpu::matchTemplate(loadMat(scene), loadMat(templ), d_result, CV_TM_SQDIFF_NORMED);
-
-    cv::Mat result(d_result);
-
-    double minVal;
-    cv::Point minLoc;
-    cv::minMaxLoc(result, &minVal, 0, &minLoc, 0);
-
-    ASSERT_GE(minVal, 0);
-    ASSERT_LT(minVal, 1e-3);
-    ASSERT_EQ(344, minLoc.x);
-    ASSERT_EQ(0, minLoc.y);
-}
-
-GPU_TEST_P(MatchTemplate_CanFindBigTemplate, SQDIFF)
-{
-    cv::Mat scene = readImage("matchtemplate/scene.png");
-    ASSERT_FALSE(scene.empty());
-
-    cv::Mat templ = readImage("matchtemplate/template.png");
-    ASSERT_FALSE(templ.empty());
-
-    cv::gpu::GpuMat d_result;
-    cv::gpu::matchTemplate(loadMat(scene), loadMat(templ), d_result, CV_TM_SQDIFF);
-
-    cv::Mat result(d_result);
-
-    double minVal;
-    cv::Point minLoc;
-    cv::minMaxLoc(result, &minVal, 0, &minLoc, 0);
-
-    ASSERT_GE(minVal, 0);
-    ASSERT_EQ(344, minLoc.x);
-    ASSERT_EQ(0, minLoc.y);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate_CanFindBigTemplate, ALL_DEVICES);
-
-////////////////////////////////////////////////////////////////////////////
-// MulSpectrums
-
-CV_FLAGS(DftFlags, 0, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
-
-PARAM_TEST_CASE(MulSpectrums, cv::gpu::DeviceInfo, cv::Size, DftFlags)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int flag;
-
-    cv::Mat a, b;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        flag = GET_PARAM(2);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        a = randomMat(size, CV_32FC2);
-        b = randomMat(size, CV_32FC2);
-    }
-};
-
-GPU_TEST_P(MulSpectrums, Simple)
-{
-    cv::gpu::GpuMat c;
-    cv::gpu::mulSpectrums(loadMat(a), loadMat(b), c, flag, false);
-
-    cv::Mat c_gold;
-    cv::mulSpectrums(a, b, c_gold, flag, false);
-
-    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
-}
-
-GPU_TEST_P(MulSpectrums, Scaled)
-{
-    float scale = 1.f / size.area();
-
-    cv::gpu::GpuMat c;
-    cv::gpu::mulAndScaleSpectrums(loadMat(a), loadMat(b), c, flag, scale, false);
-
-    cv::Mat c_gold;
-    cv::mulSpectrums(a, b, c_gold, flag, false);
-    c_gold.convertTo(c_gold, c_gold.type(), scale);
-
-    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MulSpectrums, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(DftFlags(0), DftFlags(cv::DFT_ROWS))));
-
-////////////////////////////////////////////////////////////////////////////
-// Dft
-
-struct Dft : testing::TestWithParam<cv::gpu::DeviceInfo>
-{
-    cv::gpu::DeviceInfo devInfo;
-
-    virtual void SetUp()
-    {
-        devInfo = GetParam();
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-namespace
-{
-    void testC2C(const std::string& hint, int cols, int rows, int flags, bool inplace)
-    {
-        SCOPED_TRACE(hint);
-
-        cv::Mat a = randomMat(cv::Size(cols, rows), CV_32FC2, 0.0, 10.0);
-
-        cv::Mat b_gold;
-        cv::dft(a, b_gold, flags);
-
-        cv::gpu::GpuMat d_b;
-        cv::gpu::GpuMat d_b_data;
-        if (inplace)
-        {
-            d_b_data.create(1, a.size().area(), CV_32FC2);
-            d_b = cv::gpu::GpuMat(a.rows, a.cols, CV_32FC2, d_b_data.ptr(), a.cols * d_b_data.elemSize());
-        }
-        cv::gpu::dft(loadMat(a), d_b, cv::Size(cols, rows), flags);
-
-        EXPECT_TRUE(!inplace || d_b.ptr() == d_b_data.ptr());
-        ASSERT_EQ(CV_32F, d_b.depth());
-        ASSERT_EQ(2, d_b.channels());
-        EXPECT_MAT_NEAR(b_gold, cv::Mat(d_b), rows * cols * 1e-4);
-    }
-}
-
-GPU_TEST_P(Dft, C2C)
-{
-    int cols = randomInt(2, 100);
-    int rows = randomInt(2, 100);
-
-    for (int i = 0; i < 2; ++i)
-    {
-        bool inplace = i != 0;
-
-        testC2C("no flags", cols, rows, 0, inplace);
-        testC2C("no flags 0 1", cols, rows + 1, 0, inplace);
-        testC2C("no flags 1 0", cols, rows + 1, 0, inplace);
-        testC2C("no flags 1 1", cols + 1, rows, 0, inplace);
-        testC2C("DFT_INVERSE", cols, rows, cv::DFT_INVERSE, inplace);
-        testC2C("DFT_ROWS", cols, rows, cv::DFT_ROWS, inplace);
-        testC2C("single col", 1, rows, 0, inplace);
-        testC2C("single row", cols, 1, 0, inplace);
-        testC2C("single col inversed", 1, rows, cv::DFT_INVERSE, inplace);
-        testC2C("single row inversed", cols, 1, cv::DFT_INVERSE, inplace);
-        testC2C("single row DFT_ROWS", cols, 1, cv::DFT_ROWS, inplace);
-        testC2C("size 1 2", 1, 2, 0, inplace);
-        testC2C("size 2 1", 2, 1, 0, inplace);
-    }
-}
-
-namespace
-{
-    void testR2CThenC2R(const std::string& hint, int cols, int rows, bool inplace)
-    {
-        SCOPED_TRACE(hint);
-
-        cv::Mat a = randomMat(cv::Size(cols, rows), CV_32FC1, 0.0, 10.0);
-
-        cv::gpu::GpuMat d_b, d_c;
-        cv::gpu::GpuMat d_b_data, d_c_data;
-        if (inplace)
-        {
-            if (a.cols == 1)
-            {
-                d_b_data.create(1, (a.rows / 2 + 1) * a.cols, CV_32FC2);
-                d_b = cv::gpu::GpuMat(a.rows / 2 + 1, a.cols, CV_32FC2, d_b_data.ptr(), a.cols * d_b_data.elemSize());
-            }
-            else
-            {
-                d_b_data.create(1, a.rows * (a.cols / 2 + 1), CV_32FC2);
-                d_b = cv::gpu::GpuMat(a.rows, a.cols / 2 + 1, CV_32FC2, d_b_data.ptr(), (a.cols / 2 + 1) * d_b_data.elemSize());
-            }
-            d_c_data.create(1, a.size().area(), CV_32F);
-            d_c = cv::gpu::GpuMat(a.rows, a.cols, CV_32F, d_c_data.ptr(), a.cols * d_c_data.elemSize());
-        }
-
-        cv::gpu::dft(loadMat(a), d_b, cv::Size(cols, rows), 0);
-        cv::gpu::dft(d_b, d_c, cv::Size(cols, rows), cv::DFT_REAL_OUTPUT | cv::DFT_SCALE);
-
-        EXPECT_TRUE(!inplace || d_b.ptr() == d_b_data.ptr());
-        EXPECT_TRUE(!inplace || d_c.ptr() == d_c_data.ptr());
-        ASSERT_EQ(CV_32F, d_c.depth());
-        ASSERT_EQ(1, d_c.channels());
-
-        cv::Mat c(d_c);
-        EXPECT_MAT_NEAR(a, c, rows * cols * 1e-5);
-    }
-}
-
-GPU_TEST_P(Dft, R2CThenC2R)
-{
-    int cols = randomInt(2, 100);
-    int rows = randomInt(2, 100);
-
-    testR2CThenC2R("sanity", cols, rows, false);
-    testR2CThenC2R("sanity 0 1", cols, rows + 1, false);
-    testR2CThenC2R("sanity 1 0", cols + 1, rows, false);
-    testR2CThenC2R("sanity 1 1", cols + 1, rows + 1, false);
-    testR2CThenC2R("single col", 1, rows, false);
-    testR2CThenC2R("single col 1", 1, rows + 1, false);
-    testR2CThenC2R("single row", cols, 1, false);
-    testR2CThenC2R("single row 1", cols + 1, 1, false);
-
-    testR2CThenC2R("sanity", cols, rows, true);
-    testR2CThenC2R("sanity 0 1", cols, rows + 1, true);
-    testR2CThenC2R("sanity 1 0", cols + 1, rows, true);
-    testR2CThenC2R("sanity 1 1", cols + 1, rows + 1, true);
-    testR2CThenC2R("single row", cols, 1, true);
-    testR2CThenC2R("single row 1", cols + 1, 1, true);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Dft, ALL_DEVICES);
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// CornerHarris
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(BlockSize, int);
-    IMPLEMENT_PARAM_CLASS(ApertureSize, int);
-}
-
-PARAM_TEST_CASE(CornerHarris, cv::gpu::DeviceInfo, MatType, BorderType, BlockSize, ApertureSize)
-{
-    cv::gpu::DeviceInfo devInfo;
-    int type;
-    int borderType;
-    int blockSize;
-    int apertureSize;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        type = GET_PARAM(1);
-        borderType = GET_PARAM(2);
-        blockSize = GET_PARAM(3);
-        apertureSize = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(CornerHarris, Accuracy)
-{
-    cv::Mat src = readImageType("stereobm/aloe-L.png", type);
-    ASSERT_FALSE(src.empty());
-
-    double k = randomDouble(0.1, 0.9);
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::cornerHarris(loadMat(src), dst, blockSize, apertureSize, k, borderType);
-
-    cv::Mat dst_gold;
-    cv::cornerHarris(src, dst_gold, blockSize, apertureSize, k, borderType);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.02);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CornerHarris, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_32FC1)),
-    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_REFLECT)),
-    testing::Values(BlockSize(3), BlockSize(5), BlockSize(7)),
-    testing::Values(ApertureSize(0), ApertureSize(3), ApertureSize(5), ApertureSize(7))));
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-// cornerMinEigen
-
-PARAM_TEST_CASE(CornerMinEigen, cv::gpu::DeviceInfo, MatType, BorderType, BlockSize, ApertureSize)
-{
-    cv::gpu::DeviceInfo devInfo;
-    int type;
-    int borderType;
-    int blockSize;
-    int apertureSize;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        type = GET_PARAM(1);
-        borderType = GET_PARAM(2);
-        blockSize = GET_PARAM(3);
-        apertureSize = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(CornerMinEigen, Accuracy)
-{
-    cv::Mat src = readImageType("stereobm/aloe-L.png", type);
-    ASSERT_FALSE(src.empty());
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::cornerMinEigenVal(loadMat(src), dst, blockSize, apertureSize, borderType);
-
-    cv::Mat dst_gold;
-    cv::cornerMinEigenVal(src, dst_gold, blockSize, apertureSize, borderType);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.02);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CornerMinEigen, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_32FC1)),
-    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_REFLECT)),
-    testing::Values(BlockSize(3), BlockSize(5), BlockSize(7)),
-    testing::Values(ApertureSize(0), ApertureSize(3), ApertureSize(5), ApertureSize(7))));
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_labeling.cpp b/modules/gpu/test/test_labeling.cpp
index b4fd08a26f..4a1927c392 100644
--- a/modules/gpu/test/test_labeling.cpp
+++ b/modules/gpu/test/test_labeling.cpp
@@ -1,31 +1,31 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
-// By downloading, copying, installing or using the software you agree to this license.
-// If you do not agree to this license, do not download, install,
-// copy or use the software.
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
 //
 //
-//                          License Agreement
-//               For Open Source Computer Vision Library
+//                           License Agreement
+//                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
-//  * Redistributions of source code must retain the above copyright notice,
-//    this list of conditions and the following disclaimer.
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
 //
-//  * Redistributions in binary form must reproduce the above copyright notice,
-//    this list of conditions and the following disclaimer in the documentation
-//    and/or other materials provided with the distribution.
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
 //
-//  * The name of the copyright holders may not be used to endorse or promote products
-//    derived from this software without specific prior written permission.
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
 // any express or implied warranties, including, but not limited to, the implied
@@ -37,6 +37,7 @@
 // and on any theory of liability, whether in contract, strict liability,
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
+//
 //M*/
 
 #include "test_precomp.hpp"
@@ -169,9 +170,9 @@ struct Labeling : testing::TestWithParam<cv::gpu::DeviceInfo>
 GPU_TEST_P(Labeling, DISABLED_ConnectedComponents)
 {
     cv::Mat image;
-    cvtColor(loat_image(), image, CV_BGR2GRAY);
+    cvtColor(loat_image(), image, cv::COLOR_BGR2GRAY);
 
-    cv::threshold(image, image, 150, 255, CV_THRESH_BINARY);
+    cv::threshold(image, image, 150, 255, cv::THRESH_BINARY);
 
     ASSERT_TRUE(image.type() == CV_8UC1);
 
diff --git a/modules/core/src/precomp.cpp b/modules/gpu/test/test_main.cpp
similarity index 97%
rename from modules/core/src/precomp.cpp
rename to modules/gpu/test/test_main.cpp
index e540cc5e8a..eea3d7c008 100644
--- a/modules/core/src/precomp.cpp
+++ b/modules/gpu/test/test_main.cpp
@@ -40,6 +40,6 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
-/* End of file. */
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpu/test/test_objdetect.cpp b/modules/gpu/test/test_objdetect.cpp
index fd610d9b50..b91c5dcf4d 100644
--- a/modules/gpu/test/test_objdetect.cpp
+++ b/modules/gpu/test/test_objdetect.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 //#define DUMP
 
 struct HOG : testing::TestWithParam<cv::gpu::DeviceInfo>, cv::gpu::HOGDescriptor
@@ -189,11 +192,11 @@ GPU_TEST_P(HOG, Detect)
 
     // Test on color image
     cv::Mat img;
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     testDetect(img);
 
     // Test on gray image
-    cv::cvtColor(img_rgb, img, CV_BGR2GRAY);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2GRAY);
     testDetect(img);
 
     f.close();
@@ -207,7 +210,7 @@ GPU_TEST_P(HOG, GetDescriptors)
 
     // Convert to C4
     cv::Mat img;
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
 
     cv::gpu::GpuMat d_img(img);
 
@@ -247,38 +250,38 @@ GPU_TEST_P(HOG, GetDescriptors)
 
     img_rgb = readImage("hog/positive1.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     // Everything is fine with interpolation for left top subimage
     ASSERT_EQ(0.0, cv::norm((cv::Mat)block_hists, (cv::Mat)descriptors.rowRange(0, 1)));
 
     img_rgb = readImage("hog/positive2.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     compare_inner_parts(cv::Mat(block_hists), cv::Mat(descriptors.rowRange(1, 2)));
 
     img_rgb = readImage("hog/negative1.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     compare_inner_parts(cv::Mat(block_hists), cv::Mat(descriptors.rowRange(2, 3)));
 
     img_rgb = readImage("hog/negative2.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     compare_inner_parts(cv::Mat(block_hists), cv::Mat(descriptors.rowRange(3, 4)));
 
     img_rgb = readImage("hog/positive3.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     compare_inner_parts(cv::Mat(block_hists), cv::Mat(descriptors.rowRange(4, 5)));
 
     img_rgb = readImage("hog/negative3.png");
     ASSERT_TRUE(!img_rgb.empty());
-    cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+    cv::cvtColor(img_rgb, img, cv::COLOR_BGR2BGRA);
     computeBlockHistograms(cv::gpu::GpuMat(img));
     compare_inner_parts(cv::Mat(block_hists), cv::Mat(descriptors.rowRange(5, 6)));
 }
@@ -382,7 +385,7 @@ GPU_TEST_P(LBP_classify, Accuracy)
     cv::Mat image = cv::imread(imagePath);
     image = image.colRange(0, image.cols/2);
     cv::Mat grey;
-    cvtColor(image, grey, CV_BGR2GRAY);
+    cvtColor(image, grey, cv::COLOR_BGR2GRAY);
     ASSERT_FALSE(image.empty());
 
     std::vector<cv::Rect> rects;
@@ -391,7 +394,7 @@ GPU_TEST_P(LBP_classify, Accuracy)
 
     std::vector<cv::Rect>::iterator it = rects.begin();
     for (; it != rects.end(); ++it)
-        cv::rectangle(markedImage, *it, CV_RGB(0, 0, 255));
+        cv::rectangle(markedImage, *it, cv::Scalar(255, 0, 0));
 
     cv::gpu::CascadeClassifier_GPU gpuClassifier;
     ASSERT_TRUE(gpuClassifier.load(classifierXmlPath));
diff --git a/modules/gpu/test/test_opengl.cpp b/modules/gpu/test/test_opengl.cpp
index 626851e30e..61660814ce 100644
--- a/modules/gpu/test/test_opengl.cpp
+++ b/modules/gpu/test/test_opengl.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #if defined(HAVE_CUDA) && defined(HAVE_OPENGL)
 
+using namespace cvtest;
+
 /////////////////////////////////////////////
 // Buffer
 
@@ -123,25 +126,6 @@ GPU_TEST_P(Buffer, ConstructorFromBuffer)
     EXPECT_EQ(buf_gold.type(), buf.type());
 }
 
-GPU_TEST_P(Buffer, ConstructorFromTexture2D)
-{
-    const int depth = CV_MAT_DEPTH(type);
-    const int cn = CV_MAT_CN(type);
-
-    if (depth != CV_32F || cn == 2)
-        return;
-
-    cv::Mat gold = randomMat(size, type, 0, 1.0);
-    cv::ogl::Texture2D tex_gold(gold, true);
-
-    cv::ogl::Buffer buf(tex_gold, cv::ogl::Buffer::PIXEL_PACK_BUFFER, true);
-
-    cv::Mat bufData;
-    buf.copyTo(bufData);
-
-    EXPECT_MAT_NEAR(gold, bufData, 1e-2);
-}
-
 GPU_TEST_P(Buffer, Create)
 {
     cv::ogl::Buffer buf;
@@ -195,26 +179,6 @@ GPU_TEST_P(Buffer, CopyFromBuffer)
     EXPECT_MAT_NEAR(gold, bufData, 0);
 }
 
-GPU_TEST_P(Buffer, CopyFromTexture2D)
-{
-    const int depth = CV_MAT_DEPTH(type);
-    const int cn = CV_MAT_CN(type);
-
-    if (depth != CV_32F || cn == 2)
-        return;
-
-    cv::Mat gold = randomMat(size, type, 0, 1.0);
-    cv::ogl::Texture2D tex_gold(gold, true);
-
-    cv::ogl::Buffer buf;
-    buf.copyFrom(tex_gold, cv::ogl::Buffer::ARRAY_BUFFER, true);
-
-    cv::Mat bufData;
-    buf.copyTo(bufData);
-
-    EXPECT_MAT_NEAR(gold, bufData, 1e-2);
-}
-
 GPU_TEST_P(Buffer, CopyToGpuMat)
 {
     cv::Mat gold = randomMat(size, type);
@@ -234,7 +198,8 @@ GPU_TEST_P(Buffer, CopyToBuffer)
     cv::ogl::Buffer buf(gold, cv::ogl::Buffer::ARRAY_BUFFER, true);
 
     cv::ogl::Buffer dst;
-    buf.copyTo(dst, cv::ogl::Buffer::ARRAY_BUFFER, true);
+    buf.copyTo(dst);
+    dst.setAutoRelease(true);
 
     EXPECT_NE(buf.bufId(), dst.bufId());
 
@@ -244,27 +209,6 @@ GPU_TEST_P(Buffer, CopyToBuffer)
     EXPECT_MAT_NEAR(gold, bufData, 0);
 }
 
-GPU_TEST_P(Buffer, CopyToTexture2D)
-{
-    const int depth = CV_MAT_DEPTH(type);
-    const int cn = CV_MAT_CN(type);
-
-    if (depth != CV_32F || cn == 2)
-        return;
-
-    cv::Mat gold = randomMat(size, type, 0, 1.0);
-
-    cv::ogl::Buffer buf(gold, cv::ogl::Buffer::PIXEL_PACK_BUFFER, true);
-
-    cv::ogl::Texture2D tex;
-    buf.copyTo(tex, cv::ogl::Buffer::PIXEL_PACK_BUFFER, true);
-
-    cv::Mat texData;
-    tex.copyTo(texData);
-
-    EXPECT_MAT_NEAR(gold, texData, 1e-2);
-}
-
 GPU_TEST_P(Buffer, Clone)
 {
     cv::Mat gold = randomMat(size, type);
diff --git a/modules/gpu/test/test_precomp.hpp b/modules/gpu/test/test_precomp.hpp
index 93c1df04d5..0a9c1df3dc 100644
--- a/modules/gpu/test/test_precomp.hpp
+++ b/modules/gpu/test/test_precomp.hpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -50,39 +51,17 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
-#include <cmath>
-#include <ctime>
-#include <cstdio>
-#include <iostream>
 #include <fstream>
-#include <functional>
-#include <sstream>
-#include <string>
-#include <limits>
-#include <algorithm>
-#include <iterator>
-#include <stdexcept>
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpu.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/opengl.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/objdetect.hpp"
 
 #include "cvconfig.h"
 
-#ifdef HAVE_CUDA
-    #include <cuda.h>
-    #include <cuda_runtime.h>
-
-    #include "opencv2/core.hpp"
-    #include "opencv2/core/opengl.hpp"
-    #include "opencv2/highgui.hpp"
-    #include "opencv2/calib3d.hpp"
-    #include "opencv2/imgproc.hpp"
-    #include "opencv2/video.hpp"
-    #include "opencv2/ts.hpp"
-    #include "opencv2/gpu.hpp"
-    #include "opencv2/nonfree.hpp"
-    #include "opencv2/legacy.hpp"
-
-    #include "utility.hpp"
-    #include "interpolation.hpp"
-    #include "main_test_nvidia.h"
-#endif
-
 #endif
diff --git a/modules/gpu/test/test_stream.cpp b/modules/gpu/test/test_stream.cpp
index 4adac41292..4ce29db750 100644
--- a/modules/gpu/test/test_stream.cpp
+++ b/modules/gpu/test/test_stream.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -44,7 +44,11 @@
 
 #ifdef HAVE_CUDA
 
-#if CUDA_VERSION >= 5000
+#include <cuda_runtime.h>
+
+#if CUDART_VERSION >= 5000
+
+using namespace cvtest;
 
 struct Async : testing::TestWithParam<cv::gpu::DeviceInfo>
 {
@@ -59,20 +63,21 @@ struct Async : testing::TestWithParam<cv::gpu::DeviceInfo>
         cv::gpu::DeviceInfo devInfo = GetParam();
         cv::gpu::setDevice(devInfo.deviceID());
 
+        src = cv::gpu::CudaMem(cv::gpu::CudaMem::PAGE_LOCKED);
+
         cv::Mat m = randomMat(cv::Size(128, 128), CV_8UC1);
-        src.create(m.size(), m.type(), cv::gpu::CudaMem::ALLOC_PAGE_LOCKED);
-        m.copyTo(src.createMatHeader());
+        m.copyTo(src);
     }
 };
 
-void checkMemSet(cv::gpu::Stream&, int status, void* userData)
+void checkMemSet(int status, void* userData)
 {
     ASSERT_EQ(cudaSuccess, status);
 
     Async* test = reinterpret_cast<Async*>(userData);
 
-    cv::Mat src = test->src;
-    cv::Mat dst = test->dst;
+    cv::gpu::CudaMem src = test->src;
+    cv::gpu::CudaMem dst = test->dst;
 
     cv::Mat dst_gold = cv::Mat::zeros(src.size(), src.type());
 
@@ -85,8 +90,8 @@ GPU_TEST_P(Async, MemSet)
 
     d_dst.upload(src);
 
-    stream.enqueueMemSet(d_dst, cv::Scalar::all(0));
-    stream.enqueueDownload(d_dst, dst);
+    d_dst.setTo(cv::Scalar::all(0), stream);
+    d_dst.download(dst, stream);
 
     Async* test = this;
     stream.enqueueHostCallback(checkMemSet, test);
@@ -94,17 +99,17 @@ GPU_TEST_P(Async, MemSet)
     stream.waitForCompletion();
 }
 
-void checkConvert(cv::gpu::Stream&, int status, void* userData)
+void checkConvert(int status, void* userData)
 {
     ASSERT_EQ(cudaSuccess, status);
 
     Async* test = reinterpret_cast<Async*>(userData);
 
-    cv::Mat src = test->src;
-    cv::Mat dst = test->dst;
+    cv::gpu::CudaMem src = test->src;
+    cv::gpu::CudaMem dst = test->dst;
 
     cv::Mat dst_gold;
-    src.convertTo(dst_gold, CV_32S);
+    src.createMatHeader().convertTo(dst_gold, CV_32S);
 
     ASSERT_MAT_NEAR(dst_gold, dst, 0);
 }
@@ -113,9 +118,9 @@ GPU_TEST_P(Async, Convert)
 {
     cv::gpu::Stream stream;
 
-    stream.enqueueUpload(src, d_src);
-    stream.enqueueConvert(d_src, d_dst, CV_32S);
-    stream.enqueueDownload(d_dst, dst);
+    d_src.upload(src, stream);
+    d_src.convertTo(d_dst, CV_32S, stream);
+    d_dst.download(dst, stream);
 
     Async* test = this;
     stream.enqueueHostCallback(checkConvert, test);
@@ -125,6 +130,6 @@ GPU_TEST_P(Async, Convert)
 
 INSTANTIATE_TEST_CASE_P(GPU_Stream, Async, ALL_DEVICES);
 
-#endif
+#endif // CUDART_VERSION >= 5000
 
 #endif // HAVE_CUDA
diff --git a/modules/gpu/test/utility.cpp b/modules/gpu/test/utility.cpp
deleted file mode 100644
index 88f7963242..0000000000
--- a/modules/gpu/test/utility.cpp
+++ /dev/null
@@ -1,407 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "test_precomp.hpp"
-
-#ifdef HAVE_CUDA
-
-using namespace std;
-using namespace cv;
-using namespace cv::gpu;
-using namespace cvtest;
-using namespace testing;
-using namespace testing::internal;
-
-//////////////////////////////////////////////////////////////////////
-// random generators
-
-int randomInt(int minVal, int maxVal)
-{
-    RNG& rng = TS::ptr()->get_rng();
-    return rng.uniform(minVal, maxVal);
-}
-
-double randomDouble(double minVal, double maxVal)
-{
-    RNG& rng = TS::ptr()->get_rng();
-    return rng.uniform(minVal, maxVal);
-}
-
-Size randomSize(int minVal, int maxVal)
-{
-    return Size(randomInt(minVal, maxVal), randomInt(minVal, maxVal));
-}
-
-Scalar randomScalar(double minVal, double maxVal)
-{
-    return Scalar(randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal));
-}
-
-Mat randomMat(Size size, int type, double minVal, double maxVal)
-{
-    return randomMat(TS::ptr()->get_rng(), size, type, minVal, maxVal, false);
-}
-
-//////////////////////////////////////////////////////////////////////
-// GpuMat create
-
-GpuMat createMat(Size size, int type, bool useRoi)
-{
-    Size size0 = size;
-
-    if (useRoi)
-    {
-        size0.width += randomInt(5, 15);
-        size0.height += randomInt(5, 15);
-    }
-
-    GpuMat d_m(size0, type);
-
-    if (size0 != size)
-        d_m = d_m(Rect((size0.width - size.width) / 2, (size0.height - size.height) / 2, size.width, size.height));
-
-    return d_m;
-}
-
-GpuMat loadMat(const Mat& m, bool useRoi)
-{
-    GpuMat d_m = createMat(m.size(), m.type(), useRoi);
-    d_m.upload(m);
-    return d_m;
-}
-
-//////////////////////////////////////////////////////////////////////
-// Image load
-
-Mat readImage(const std::string& fileName, int flags)
-{
-    return imread(TS::ptr()->get_data_path() + fileName, flags);
-}
-
-Mat readImageType(const std::string& fname, int type)
-{
-    Mat src = readImage(fname, CV_MAT_CN(type) == 1 ? IMREAD_GRAYSCALE : IMREAD_COLOR);
-    if (CV_MAT_CN(type) == 4)
-    {
-        Mat temp;
-        cvtColor(src, temp, COLOR_BGR2BGRA);
-        swap(src, temp);
-    }
-    src.convertTo(src, CV_MAT_DEPTH(type), CV_MAT_DEPTH(type) == CV_32F ? 1.0 / 255.0 : 1.0);
-    return src;
-}
-
-//////////////////////////////////////////////////////////////////////
-// Gpu devices
-
-bool supportFeature(const DeviceInfo& info, FeatureSet feature)
-{
-    return TargetArchs::builtWith(feature) && info.supports(feature);
-}
-
-DeviceManager& DeviceManager::instance()
-{
-    static DeviceManager obj;
-    return obj;
-}
-
-void DeviceManager::load(int i)
-{
-    devices_.clear();
-    devices_.reserve(1);
-
-    std::ostringstream msg;
-
-    if (i < 0 || i >= getCudaEnabledDeviceCount())
-    {
-        msg << "Incorrect device number - " << i;
-        throw runtime_error(msg.str());
-    }
-
-    DeviceInfo info(i);
-
-    if (!info.isCompatible())
-    {
-        msg << "Device " << i << " [" << info.name() << "] is NOT compatible with current GPU module build";
-        throw runtime_error(msg.str());
-    }
-
-    devices_.push_back(info);
-}
-
-void DeviceManager::loadAll()
-{
-    int deviceCount = getCudaEnabledDeviceCount();
-
-    devices_.clear();
-    devices_.reserve(deviceCount);
-
-    for (int i = 0; i < deviceCount; ++i)
-    {
-        DeviceInfo info(i);
-        if (info.isCompatible())
-        {
-            devices_.push_back(info);
-        }
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Additional assertion
-
-namespace
-{
-    template <typename T, typename OutT> std::string printMatValImpl(const Mat& m, Point p)
-    {
-        const int cn = m.channels();
-
-        std::ostringstream ostr;
-        ostr << "(";
-
-        p.x /= cn;
-
-        ostr << static_cast<OutT>(m.at<T>(p.y, p.x * cn));
-        for (int c = 1; c < m.channels(); ++c)
-        {
-            ostr << ", " << static_cast<OutT>(m.at<T>(p.y, p.x * cn + c));
-        }
-        ostr << ")";
-
-        return ostr.str();
-    }
-
-    std::string printMatVal(const Mat& m, Point p)
-    {
-        typedef std::string (*func_t)(const Mat& m, Point p);
-
-        static const func_t funcs[] =
-        {
-            printMatValImpl<uchar, int>, printMatValImpl<schar, int>, printMatValImpl<ushort, int>, printMatValImpl<short, int>,
-            printMatValImpl<int, int>, printMatValImpl<float, float>, printMatValImpl<double, double>
-        };
-
-        return funcs[m.depth()](m, p);
-    }
-}
-
-void minMaxLocGold(const Mat& src, double* minVal_, double* maxVal_, Point* minLoc_, Point* maxLoc_, const Mat& mask)
-{
-    if (src.depth() != CV_8S)
-    {
-        minMaxLoc(src, minVal_, maxVal_, minLoc_, maxLoc_, mask);
-        return;
-    }
-
-    // OpenCV's minMaxLoc doesn't support CV_8S type
-    double minVal = numeric_limits<double>::max();
-    Point minLoc(-1, -1);
-
-    double maxVal = -numeric_limits<double>::max();
-    Point maxLoc(-1, -1);
-
-    for (int y = 0; y < src.rows; ++y)
-    {
-        const schar* src_row = src.ptr<schar>(y);
-        const uchar* mask_row = mask.empty() ? 0 : mask.ptr<uchar>(y);
-
-        for (int x = 0; x < src.cols; ++x)
-        {
-            if (!mask_row || mask_row[x])
-            {
-                schar val = src_row[x];
-
-                if (val < minVal)
-                {
-                    minVal = val;
-                    minLoc = cv::Point(x, y);
-                }
-
-                if (val > maxVal)
-                {
-                    maxVal = val;
-                    maxLoc = cv::Point(x, y);
-                }
-            }
-        }
-    }
-
-    if (minVal_) *minVal_ = minVal;
-    if (maxVal_) *maxVal_ = maxVal;
-
-    if (minLoc_) *minLoc_ = minLoc;
-    if (maxLoc_) *maxLoc_ = maxLoc;
-}
-
-Mat getMat(InputArray arr)
-{
-    if (arr.kind() == _InputArray::GPU_MAT)
-    {
-        Mat m;
-        arr.getGpuMat().download(m);
-        return m;
-    }
-
-    return arr.getMat();
-}
-
-AssertionResult assertMatNear(const char* expr1, const char* expr2, const char* eps_expr, InputArray m1_, InputArray m2_, double eps)
-{
-    Mat m1 = getMat(m1_);
-    Mat m2 = getMat(m2_);
-
-    if (m1.size() != m2.size())
-    {
-        return AssertionFailure() << "Matrices \"" << expr1 << "\" and \"" << expr2 << "\" have different sizes : \""
-                                  << expr1 << "\" [" << PrintToString(m1.size()) << "] vs \""
-                                  << expr2 << "\" [" << PrintToString(m2.size()) << "]";
-    }
-
-    if (m1.type() != m2.type())
-    {
-        return AssertionFailure() << "Matrices \"" << expr1 << "\" and \"" << expr2 << "\" have different types : \""
-                                  << expr1 << "\" [" << PrintToString(MatType(m1.type())) << "] vs \""
-                                  << expr2 << "\" [" << PrintToString(MatType(m2.type())) << "]";
-    }
-
-    Mat diff;
-    absdiff(m1.reshape(1), m2.reshape(1), diff);
-
-    double maxVal = 0.0;
-    Point maxLoc;
-    minMaxLocGold(diff, 0, &maxVal, 0, &maxLoc);
-
-    if (maxVal > eps)
-    {
-        return AssertionFailure() << "The max difference between matrices \"" << expr1 << "\" and \"" << expr2
-                                  << "\" is " << maxVal << " at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ")"
-                                  << ", which exceeds \"" << eps_expr << "\", where \""
-                                  << expr1 << "\" at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ") evaluates to " << printMatVal(m1, maxLoc) << ", \""
-                                  << expr2 << "\" at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ") evaluates to " << printMatVal(m2, maxLoc) << ", \""
-                                  << eps_expr << "\" evaluates to " << eps;
-    }
-
-    return AssertionSuccess();
-}
-
-double checkSimilarity(InputArray m1, InputArray m2)
-{
-    Mat diff;
-    matchTemplate(getMat(m1), getMat(m2), diff, CV_TM_CCORR_NORMED);
-    return std::abs(diff.at<float>(0, 0) - 1.f);
-}
-
-//////////////////////////////////////////////////////////////////////
-// Helper structs for value-parameterized tests
-
-vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end)
-{
-    vector<MatType> v;
-
-    v.reserve((depth_end - depth_start + 1) * (cn_end - cn_start + 1));
-
-    for (int depth = depth_start; depth <= depth_end; ++depth)
-    {
-        for (int cn = cn_start; cn <= cn_end; ++cn)
-        {
-            v.push_back(MatType(CV_MAKE_TYPE(depth, cn)));
-        }
-    }
-
-    return v;
-}
-
-const vector<MatType>& all_types()
-{
-    static vector<MatType> v = types(CV_8U, CV_64F, 1, 4);
-
-    return v;
-}
-
-void cv::gpu::PrintTo(const DeviceInfo& info, ostream* os)
-{
-    (*os) << info.name();
-}
-
-void PrintTo(const UseRoi& useRoi, std::ostream* os)
-{
-    if (useRoi)
-        (*os) << "sub matrix";
-    else
-        (*os) << "whole matrix";
-}
-
-void PrintTo(const Inverse& inverse, std::ostream* os)
-{
-    if (inverse)
-        (*os) << "inverse";
-    else
-        (*os) << "direct";
-}
-
-//////////////////////////////////////////////////////////////////////
-// Other
-
-void dumpImage(const std::string& fileName, const Mat& image)
-{
-    imwrite(TS::ptr()->get_data_path() + fileName, image);
-}
-
-void showDiff(InputArray gold_, InputArray actual_, double eps)
-{
-    Mat gold = getMat(gold_);
-    Mat actual = getMat(actual_);
-
-    Mat diff;
-    absdiff(gold, actual, diff);
-    threshold(diff, diff, eps, 255.0, cv::THRESH_BINARY);
-
-    namedWindow("gold", WINDOW_NORMAL);
-    namedWindow("actual", WINDOW_NORMAL);
-    namedWindow("diff", WINDOW_NORMAL);
-
-    imshow("gold", gold);
-    imshow("actual", actual);
-    imshow("diff", diff);
-
-    waitKey();
-}
-
-#endif // HAVE_CUDA
diff --git a/modules/gpu/test/utility.hpp b/modules/gpu/test/utility.hpp
deleted file mode 100644
index fa0cd93e7c..0000000000
--- a/modules/gpu/test/utility.hpp
+++ /dev/null
@@ -1,331 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_GPU_TEST_UTILITY_HPP__
-#define __OPENCV_GPU_TEST_UTILITY_HPP__
-
-#include "opencv2/core.hpp"
-#include "opencv2/core/gpumat.hpp"
-#include "opencv2/highgui.hpp"
-#include "opencv2/ts.hpp"
-
-//////////////////////////////////////////////////////////////////////
-// random generators
-
-int randomInt(int minVal, int maxVal);
-double randomDouble(double minVal, double maxVal);
-cv::Size randomSize(int minVal, int maxVal);
-cv::Scalar randomScalar(double minVal, double maxVal);
-cv::Mat randomMat(cv::Size size, int type, double minVal = 0.0, double maxVal = 255.0);
-
-//////////////////////////////////////////////////////////////////////
-// GpuMat create
-
-cv::gpu::GpuMat createMat(cv::Size size, int type, bool useRoi = false);
-cv::gpu::GpuMat loadMat(const cv::Mat& m, bool useRoi = false);
-
-//////////////////////////////////////////////////////////////////////
-// Image load
-
-//! read image from testdata folder
-cv::Mat readImage(const std::string& fileName, int flags = cv::IMREAD_COLOR);
-
-//! read image from testdata folder and convert it to specified type
-cv::Mat readImageType(const std::string& fname, int type);
-
-//////////////////////////////////////////////////////////////////////
-// Gpu devices
-
-//! return true if device supports specified feature and gpu module was built with support the feature.
-bool supportFeature(const cv::gpu::DeviceInfo& info, cv::gpu::FeatureSet feature);
-
-class DeviceManager
-{
-public:
-    static DeviceManager& instance();
-
-    void load(int i);
-    void loadAll();
-
-    const std::vector<cv::gpu::DeviceInfo>& values() const { return devices_; }
-
-private:
-    std::vector<cv::gpu::DeviceInfo> devices_;
-};
-
-#define ALL_DEVICES testing::ValuesIn(DeviceManager::instance().values())
-
-//////////////////////////////////////////////////////////////////////
-// Additional assertion
-
-void minMaxLocGold(const cv::Mat& src, double* minVal_, double* maxVal_ = 0, cv::Point* minLoc_ = 0, cv::Point* maxLoc_ = 0, const cv::Mat& mask = cv::Mat());
-
-cv::Mat getMat(cv::InputArray arr);
-
-testing::AssertionResult assertMatNear(const char* expr1, const char* expr2, const char* eps_expr, cv::InputArray m1, cv::InputArray m2, double eps);
-
-#define EXPECT_MAT_NEAR(m1, m2, eps) EXPECT_PRED_FORMAT3(assertMatNear, m1, m2, eps)
-#define ASSERT_MAT_NEAR(m1, m2, eps) ASSERT_PRED_FORMAT3(assertMatNear, m1, m2, eps)
-
-#define EXPECT_SCALAR_NEAR(s1, s2, eps) \
-    { \
-        EXPECT_NEAR(s1[0], s2[0], eps); \
-        EXPECT_NEAR(s1[1], s2[1], eps); \
-        EXPECT_NEAR(s1[2], s2[2], eps); \
-        EXPECT_NEAR(s1[3], s2[3], eps); \
-    }
-#define ASSERT_SCALAR_NEAR(s1, s2, eps) \
-    { \
-        ASSERT_NEAR(s1[0], s2[0], eps); \
-        ASSERT_NEAR(s1[1], s2[1], eps); \
-        ASSERT_NEAR(s1[2], s2[2], eps); \
-        ASSERT_NEAR(s1[3], s2[3], eps); \
-    }
-
-#define EXPECT_POINT2_NEAR(p1, p2, eps) \
-    { \
-        EXPECT_NEAR(p1.x, p2.x, eps); \
-        EXPECT_NEAR(p1.y, p2.y, eps); \
-    }
-#define ASSERT_POINT2_NEAR(p1, p2, eps) \
-    { \
-        ASSERT_NEAR(p1.x, p2.x, eps); \
-        ASSERT_NEAR(p1.y, p2.y, eps); \
-    }
-
-#define EXPECT_POINT3_NEAR(p1, p2, eps) \
-    { \
-        EXPECT_NEAR(p1.x, p2.x, eps); \
-        EXPECT_NEAR(p1.y, p2.y, eps); \
-        EXPECT_NEAR(p1.z, p2.z, eps); \
-    }
-#define ASSERT_POINT3_NEAR(p1, p2, eps) \
-    { \
-        ASSERT_NEAR(p1.x, p2.x, eps); \
-        ASSERT_NEAR(p1.y, p2.y, eps); \
-        ASSERT_NEAR(p1.z, p2.z, eps); \
-    }
-
-double checkSimilarity(cv::InputArray m1, cv::InputArray m2);
-
-#define EXPECT_MAT_SIMILAR(mat1, mat2, eps) \
-    { \
-        ASSERT_EQ(mat1.type(), mat2.type()); \
-        ASSERT_EQ(mat1.size(), mat2.size()); \
-        EXPECT_LE(checkSimilarity(mat1, mat2), eps); \
-    }
-#define ASSERT_MAT_SIMILAR(mat1, mat2, eps) \
-    { \
-        ASSERT_EQ(mat1.type(), mat2.type()); \
-        ASSERT_EQ(mat1.size(), mat2.size()); \
-        ASSERT_LE(checkSimilarity(mat1, mat2), eps); \
-    }
-
-//////////////////////////////////////////////////////////////////////
-// Helper structs for value-parameterized tests
-
-#define GPU_TEST_P(test_case_name, test_name) \
-  class GTEST_TEST_CLASS_NAME_(test_case_name, test_name) \
-      : public test_case_name { \
-   public: \
-    GTEST_TEST_CLASS_NAME_(test_case_name, test_name)() {} \
-    virtual void TestBody(); \
-   private: \
-    void UnsafeTestBody(); \
-    static int AddToRegistry() { \
-      ::testing::UnitTest::GetInstance()->parameterized_test_registry(). \
-          GetTestCasePatternHolder<test_case_name>(\
-              #test_case_name, __FILE__, __LINE__)->AddTestPattern(\
-                  #test_case_name, \
-                  #test_name, \
-                  new ::testing::internal::TestMetaFactory< \
-                      GTEST_TEST_CLASS_NAME_(test_case_name, test_name)>()); \
-      return 0; \
-    } \
-    static int gtest_registering_dummy_; \
-    GTEST_DISALLOW_COPY_AND_ASSIGN_(\
-        GTEST_TEST_CLASS_NAME_(test_case_name, test_name)); \
-  }; \
-  int GTEST_TEST_CLASS_NAME_(test_case_name, \
-                             test_name)::gtest_registering_dummy_ = \
-      GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::AddToRegistry(); \
-  void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::TestBody() \
-  { \
-    try \
-    { \
-      UnsafeTestBody(); \
-    } \
-    catch (...) \
-    { \
-      cv::gpu::resetDevice(); \
-      throw; \
-    } \
-  } \
-  void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::UnsafeTestBody()
-
-#define PARAM_TEST_CASE(name, ...) struct name : testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
-#define GET_PARAM(k) std::tr1::get< k >(GetParam())
-
-namespace cv { namespace gpu
-{
-    void PrintTo(const DeviceInfo& info, std::ostream* os);
-}}
-
-#define DIFFERENT_SIZES testing::Values(cv::Size(128, 128), cv::Size(113, 113))
-
-// Depth
-
-using perf::MatDepth;
-
-#define ALL_DEPTH testing::Values(MatDepth(CV_8U), MatDepth(CV_8S), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32S), MatDepth(CV_32F), MatDepth(CV_64F))
-
-#define DEPTH_PAIRS testing::Values(std::make_pair(MatDepth(CV_8U), MatDepth(CV_8U)),   \
-                                    std::make_pair(MatDepth(CV_8U), MatDepth(CV_16U)),  \
-                                    std::make_pair(MatDepth(CV_8U), MatDepth(CV_16S)),  \
-                                    std::make_pair(MatDepth(CV_8U), MatDepth(CV_32S)),  \
-                                    std::make_pair(MatDepth(CV_8U), MatDepth(CV_32F)),  \
-                                    std::make_pair(MatDepth(CV_8U), MatDepth(CV_64F)),  \
-                                                                                        \
-                                    std::make_pair(MatDepth(CV_16U), MatDepth(CV_16U)), \
-                                    std::make_pair(MatDepth(CV_16U), MatDepth(CV_32S)), \
-                                    std::make_pair(MatDepth(CV_16U), MatDepth(CV_32F)), \
-                                    std::make_pair(MatDepth(CV_16U), MatDepth(CV_64F)), \
-                                                                                        \
-                                    std::make_pair(MatDepth(CV_16S), MatDepth(CV_16S)), \
-                                    std::make_pair(MatDepth(CV_16S), MatDepth(CV_32S)), \
-                                    std::make_pair(MatDepth(CV_16S), MatDepth(CV_32F)), \
-                                    std::make_pair(MatDepth(CV_16S), MatDepth(CV_64F)), \
-                                                                                        \
-                                    std::make_pair(MatDepth(CV_32S), MatDepth(CV_32S)), \
-                                    std::make_pair(MatDepth(CV_32S), MatDepth(CV_32F)), \
-                                    std::make_pair(MatDepth(CV_32S), MatDepth(CV_64F)), \
-                                                                                        \
-                                    std::make_pair(MatDepth(CV_32F), MatDepth(CV_32F)), \
-                                    std::make_pair(MatDepth(CV_32F), MatDepth(CV_64F)), \
-                                                                                        \
-                                    std::make_pair(MatDepth(CV_64F), MatDepth(CV_64F)))
-
-// Type
-
-using perf::MatType;
-
-//! return vector with types from specified range.
-std::vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end);
-
-//! return vector with all types (depth: CV_8U-CV_64F, channels: 1-4).
-const std::vector<MatType>& all_types();
-
-#define ALL_TYPES testing::ValuesIn(all_types())
-#define TYPES(depth_start, depth_end, cn_start, cn_end) testing::ValuesIn(types(depth_start, depth_end, cn_start, cn_end))
-
-// ROI
-
-class UseRoi
-{
-public:
-    inline UseRoi(bool val = false) : val_(val) {}
-
-    inline operator bool() const { return val_; }
-
-private:
-    bool val_;
-};
-
-void PrintTo(const UseRoi& useRoi, std::ostream* os);
-
-#define WHOLE_SUBMAT testing::Values(UseRoi(false), UseRoi(true))
-
-// Direct/Inverse
-
-class Inverse
-{
-public:
-    inline Inverse(bool val = false) : val_(val) {}
-
-    inline operator bool() const { return val_; }
-
-private:
-    bool val_;
-};
-
-void PrintTo(const Inverse& useRoi, std::ostream* os);
-
-#define DIRECT_INVERSE testing::Values(Inverse(false), Inverse(true))
-
-// Param class
-
-#define IMPLEMENT_PARAM_CLASS(name, type) \
-    class name \
-    { \
-    public: \
-        name ( type arg = type ()) : val_(arg) {} \
-        operator type () const {return val_;} \
-    private: \
-        type val_; \
-    }; \
-    inline void PrintTo( name param, std::ostream* os) \
-    { \
-        *os << #name <<  "(" << testing::PrintToString(static_cast< type >(param)) << ")"; \
-    }
-
-IMPLEMENT_PARAM_CLASS(Channels, int)
-
-#define ALL_CHANNELS testing::Values(Channels(1), Channels(2), Channels(3), Channels(4))
-#define IMAGE_CHANNELS testing::Values(Channels(1), Channels(3), Channels(4))
-
-// Flags and enums
-
-CV_ENUM(NormCode, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_TYPE_MASK, cv::NORM_RELATIVE, cv::NORM_MINMAX)
-
-CV_ENUM(Interpolation, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::INTER_AREA)
-
-CV_ENUM(BorderType, cv::BORDER_REFLECT101, cv::BORDER_REPLICATE, cv::BORDER_CONSTANT, cv::BORDER_REFLECT, cv::BORDER_WRAP)
-#define ALL_BORDER_TYPES testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_CONSTANT), BorderType(cv::BORDER_REFLECT), BorderType(cv::BORDER_WRAP))
-
-CV_FLAGS(WarpFlags, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::WARP_INVERSE_MAP)
-
-//////////////////////////////////////////////////////////////////////
-// Other
-
-void dumpImage(const std::string& fileName, const cv::Mat& image);
-void showDiff(cv::InputArray gold, cv::InputArray actual, double eps);
-
-#endif // __OPENCV_GPU_TEST_UTILITY_HPP__
diff --git a/modules/gpuarithm/CMakeLists.txt b/modules/gpuarithm/CMakeLists.txt
new file mode 100644
index 0000000000..7cd9a458b9
--- /dev/null
+++ b/modules/gpuarithm/CMakeLists.txt
@@ -0,0 +1,29 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpuarithm)
+endif()
+
+set(the_description "GPU-accelerated Operations on Matrices")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_add_module(gpuarithm opencv_core OPTIONAL opencv_gpulegacy)
+
+ocv_module_include_directories()
+ocv_glob_module_sources()
+
+set(extra_libs "")
+
+if(HAVE_CUBLAS)
+  list(APPEND extra_libs ${CUDA_cublas_LIBRARY})
+endif()
+
+if(HAVE_CUFFT)
+  list(APPEND extra_libs ${CUDA_cufft_LIBRARY})
+endif()
+
+ocv_create_module(${extra_libs})
+
+ocv_add_precompiled_headers(${the_module})
+
+ocv_add_accuracy_tests(DEPENDS_ON opencv_imgproc)
+ocv_add_perf_tests(DEPENDS_ON opencv_imgproc)
diff --git a/modules/gpuarithm/doc/arithm.rst b/modules/gpuarithm/doc/arithm.rst
new file mode 100644
index 0000000000..09b7220c09
--- /dev/null
+++ b/modules/gpuarithm/doc/arithm.rst
@@ -0,0 +1,168 @@
+Arithm Operations on Matrices
+=============================
+
+.. highlight:: cpp
+
+
+
+gpu::gemm
+---------
+Performs generalized matrix multiplication.
+
+.. ocv:function:: void gpu::gemm(InputArray src1, InputArray src2, double alpha, InputArray src3, double beta, OutputArray dst, int flags = 0, Stream& stream = Stream::Null())
+
+    :param src1: First multiplied input matrix that should have  ``CV_32FC1`` , ``CV_64FC1`` , ``CV_32FC2`` , or  ``CV_64FC2``  type.
+
+    :param src2: Second multiplied input matrix of the same type as  ``src1`` .
+
+    :param alpha: Weight of the matrix product.
+
+    :param src3: Third optional delta matrix added to the matrix product. It should have the same type as  ``src1``  and  ``src2`` .
+
+    :param beta: Weight of  ``src3`` .
+
+    :param dst: Destination matrix. It has the proper size and the same type as input matrices.
+
+    :param flags: Operation flags:
+
+            * **GEMM_1_T** transpose  ``src1``
+            * **GEMM_2_T** transpose  ``src2``
+            * **GEMM_3_T** transpose  ``src3``
+
+    :param stream: Stream for the asynchronous version.
+
+The function performs generalized matrix multiplication similar to the ``gemm`` functions in BLAS level 3. For example, ``gemm(src1, src2, alpha, src3, beta, dst, GEMM_1_T + GEMM_3_T)`` corresponds to
+
+.. math::
+
+    \texttt{dst} =  \texttt{alpha} \cdot \texttt{src1} ^T  \cdot \texttt{src2} +  \texttt{beta} \cdot \texttt{src3} ^T
+
+.. note:: Transposition operation doesn't support  ``CV_64FC2``  input type.
+
+.. seealso:: :ocv:func:`gemm`
+
+
+
+gpu::mulSpectrums
+-----------------
+Performs a per-element multiplication of two Fourier spectrums.
+
+.. ocv:function:: void gpu::mulSpectrums(InputArray src1, InputArray src2, OutputArray dst, int flags, bool conjB=false, Stream& stream = Stream::Null())
+
+    :param src1: First spectrum.
+
+    :param src2: Second spectrum with the same size and type as  ``a`` .
+
+    :param dst: Destination spectrum.
+
+    :param flags: Mock parameter used for CPU/GPU interfaces similarity.
+
+    :param conjB: Optional flag to specify if the second spectrum needs to be conjugated before the multiplication.
+
+    :param stream: Stream for the asynchronous version.
+
+Only full (not packed) ``CV_32FC2`` complex spectrums in the interleaved format are supported for now.
+
+.. seealso:: :ocv:func:`mulSpectrums`
+
+
+
+gpu::mulAndScaleSpectrums
+-------------------------
+Performs a per-element multiplication of two Fourier spectrums and scales the result.
+
+.. ocv:function:: void gpu::mulAndScaleSpectrums(InputArray src1, InputArray src2, OutputArray dst, int flags, float scale, bool conjB=false, Stream& stream = Stream::Null())
+
+    :param src1: First spectrum.
+
+    :param src2: Second spectrum with the same size and type as  ``a`` .
+
+    :param dst: Destination spectrum.
+
+    :param flags: Mock parameter used for CPU/GPU interfaces similarity.
+
+    :param scale: Scale constant.
+
+    :param conjB: Optional flag to specify if the second spectrum needs to be conjugated before the multiplication.
+
+Only full (not packed) ``CV_32FC2`` complex spectrums in the interleaved format are supported for now.
+
+.. seealso:: :ocv:func:`mulSpectrums`
+
+
+
+gpu::dft
+--------
+Performs a forward or inverse discrete Fourier transform (1D or 2D) of the floating point matrix.
+
+.. ocv:function:: void gpu::dft(InputArray src, OutputArray dst, Size dft_size, int flags=0, Stream& stream = Stream::Null())
+
+    :param src: Source matrix (real or complex).
+
+    :param dst: Destination matrix (real or complex).
+
+    :param dft_size: Size of a discrete Fourier transform.
+
+    :param flags: Optional flags:
+
+        * **DFT_ROWS** transforms each individual row of the source matrix.
+
+        * **DFT_SCALE** scales the result: divide it by the number of elements in the transform (obtained from  ``dft_size`` ).
+
+        * **DFT_INVERSE** inverts DFT. Use for complex-complex cases (real-complex and complex-real cases are always forward and inverse, respectively).
+
+        * **DFT_REAL_OUTPUT** specifies the output as real. The source matrix is the result of real-complex transform, so the destination matrix must be real.
+
+Use to handle real matrices ( ``CV32FC1`` ) and complex matrices in the interleaved format ( ``CV32FC2`` ).
+
+The source matrix should be continuous, otherwise reallocation and data copying is performed. The function chooses an operation mode depending on the flags, size, and channel count of the source matrix:
+
+    * If the source matrix is complex and the output is not specified as real, the destination matrix is complex and has the ``dft_size``    size and ``CV_32FC2``    type. The destination matrix contains a full result of the DFT (forward or inverse).
+
+    * If the source matrix is complex and the output is specified as real, the function assumes that its input is the result of the forward transform (see the next item). The destination matrix has the ``dft_size`` size and ``CV_32FC1`` type. It contains the result of the inverse DFT.
+
+    * If the source matrix is real (its type is ``CV_32FC1`` ), forward DFT is performed. The result of the DFT is packed into complex ( ``CV_32FC2`` ) matrix. So, the width of the destination matrix is ``dft_size.width / 2 + 1`` . But if the source is a single column, the height is reduced instead of the width.
+
+.. seealso:: :ocv:func:`dft`
+
+
+
+gpu::Convolution
+----------------
+.. ocv:class:: gpu::Convolution : public Algorithm
+
+Base class for convolution (or cross-correlation) operator. ::
+
+    class CV_EXPORTS Convolution : public Algorithm
+    {
+    public:
+        virtual void convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null()) = 0;
+    };
+
+
+
+gpu::Convolution::convolve
+---------------------------
+Computes a convolution (or cross-correlation) of two images.
+
+.. ocv:function:: void gpu::Convolution::convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null())
+
+    :param image: Source image. Only  ``CV_32FC1`` images are supported for now.
+
+    :param templ: Template image. The size is not greater than the  ``image`` size. The type is the same as  ``image`` .
+
+    :param result: Result image. If  ``image`` is  *W x H*  and ``templ`` is  *w x h*, then  ``result`` must be *W-w+1 x H-h+1*.
+
+    :param ccorr: Flags to evaluate cross-correlation instead of convolution.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createConvolution
+----------------------
+Creates implementation for :ocv:class:`gpu::Convolution` .
+
+.. ocv:function:: Ptr<Convolution> createConvolution(Size user_block_size = Size())
+
+    :param user_block_size: Block size. If you leave default value `Size(0,0)` then automatic estimation of block size will be used (which is optimized for speed). By varying `user_block_size` you can reduce memory requirements at the cost of speed.
diff --git a/modules/gpuarithm/doc/core.rst b/modules/gpuarithm/doc/core.rst
new file mode 100644
index 0000000000..624ea3e7b3
--- /dev/null
+++ b/modules/gpuarithm/doc/core.rst
@@ -0,0 +1,150 @@
+Core Operations on Matrices
+===========================
+
+.. highlight:: cpp
+
+
+
+gpu::merge
+----------
+Makes a multi-channel matrix out of several single-channel matrices.
+
+.. ocv:function:: void gpu::merge(const GpuMat* src, size_t n, OutputArray dst, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::merge(const std::vector<GpuMat>& src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Array/vector of source matrices.
+
+    :param n: Number of source matrices.
+
+    :param dst: Destination matrix.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`merge`
+
+
+
+gpu::split
+----------
+Copies each plane of a multi-channel matrix into an array.
+
+.. ocv:function:: void gpu::split(InputArray src, GpuMat* dst, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::split(InputArray src, vector<GpuMat>& dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination array/vector of single-channel matrices.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`split`
+
+
+
+gpu::transpose
+--------------
+Transposes a matrix.
+
+.. ocv:function:: void gpu::transpose(InputArray src1, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src1: Source matrix. 1-, 4-, 8-byte element sizes are supported for now.
+
+    :param dst: Destination matrix.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`transpose`
+
+
+
+gpu::flip
+---------
+Flips a 2D matrix around vertical, horizontal, or both axes.
+
+.. ocv:function:: void gpu::flip(InputArray src, OutputArray dst, int flipCode, Stream& stream = Stream::Null())
+
+    :param src: Source matrix. Supports 1, 3 and 4 channels images with ``CV_8U``, ``CV_16U``, ``CV_32S`` or ``CV_32F`` depth.
+
+    :param dst: Destination matrix.
+
+    :param flipCode: Flip mode for the source:
+
+        * ``0`` Flips around x-axis.
+
+        * ``> 0`` Flips around y-axis.
+
+        * ``< 0`` Flips around both axes.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`flip`
+
+
+
+gpu::LookUpTable
+----------------
+.. ocv:class:: gpu::LookUpTable : public Algorithm
+
+Base class for transform using lookup table. ::
+
+    class CV_EXPORTS LookUpTable : public Algorithm
+    {
+    public:
+        virtual void transform(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+    };
+
+.. seealso:: :ocv:func:`LUT`
+
+
+
+gpu::LookUpTable::transform
+---------------------------
+Transforms the source matrix into the destination matrix using the given look-up table: ``dst(I) = lut(src(I))`` .
+
+.. ocv:function:: void gpu::LookUpTable::transform(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.  ``CV_8UC1``  and  ``CV_8UC3``  matrices are supported for now.
+
+    :param dst: Destination matrix.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createLookUpTable
+----------------------
+Creates implementation for :ocv:class:`gpu::LookUpTable` .
+
+.. ocv:function:: Ptr<LookUpTable> createLookUpTable(InputArray lut)
+
+    :param lut: Look-up table of 256 elements. It is a continuous ``CV_8U`` matrix.
+
+
+
+gpu::copyMakeBorder
+-----------------------
+Forms a border around an image.
+
+.. ocv:function:: void gpu::copyMakeBorder(InputArray src, OutputArray dst, int top, int bottom, int left, int right, int borderType, Scalar value = Scalar(), Stream& stream = Stream::Null())
+
+    :param src: Source image. ``CV_8UC1`` , ``CV_8UC4`` , ``CV_32SC1`` , and ``CV_32FC1`` types are supported.
+
+    :param dst: Destination image with the same type as  ``src``. The size is ``Size(src.cols+left+right, src.rows+top+bottom)`` .
+
+    :param top:
+
+    :param bottom:
+
+    :param left:
+
+    :param right: Number of pixels in each direction from the source image rectangle to extrapolate. For example:  ``top=1, bottom=1, left=1, right=1`` mean that 1 pixel-wide border needs to be built.
+
+    :param borderType: Border type. See  :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param value: Border value.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`copyMakeBorder`
diff --git a/modules/gpuarithm/doc/element_operations.rst b/modules/gpuarithm/doc/element_operations.rst
new file mode 100644
index 0000000000..eb616c1c39
--- /dev/null
+++ b/modules/gpuarithm/doc/element_operations.rst
@@ -0,0 +1,543 @@
+Per-element Operations
+======================
+
+.. highlight:: cpp
+
+
+
+gpu::add
+--------
+Computes a matrix-matrix or matrix-scalar sum.
+
+.. ocv:function:: void gpu::add(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), int dtype = -1, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar. Matrix should have the same size and type as ``src1`` .
+
+    :param dst: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``src1`` depth.
+
+    :param mask: Optional operation mask, 8-bit single channel array, that specifies elements of the destination array to be changed.
+
+    :param dtype: Optional depth of the output array.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`add`
+
+
+
+gpu::subtract
+-------------
+Computes a matrix-matrix or matrix-scalar difference.
+
+.. ocv:function:: void gpu::subtract(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), int dtype = -1, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar. Matrix should have the same size and type as ``src1`` .
+
+    :param dst: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``src1`` depth.
+
+    :param mask: Optional operation mask, 8-bit single channel array, that specifies elements of the destination array to be changed.
+
+    :param dtype: Optional depth of the output array.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`subtract`
+
+
+
+gpu::multiply
+-------------
+Computes a matrix-matrix or matrix-scalar per-element product.
+
+.. ocv:function:: void gpu::multiply(InputArray src1, InputArray src2, OutputArray dst, double scale = 1, int dtype = -1, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``src1`` depth.
+
+    :param scale: Optional scale factor.
+
+    :param dtype: Optional depth of the output array.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`multiply`
+
+
+
+gpu::divide
+-----------
+Computes a matrix-matrix or matrix-scalar division.
+
+.. ocv:function:: void gpu::divide(InputArray src1, InputArray src2, OutputArray dst, double scale = 1, int dtype = -1, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::divide(double src1, InputArray src2, OutputArray dst, int dtype = -1, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or a scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and number of channels as the input array(s). The depth is defined by ``dtype`` or ``src1`` depth.
+
+    :param scale: Optional scale factor.
+
+    :param dtype: Optional depth of the output array.
+
+    :param stream: Stream for the asynchronous version.
+
+This function, in contrast to :ocv:func:`divide`, uses a round-down rounding mode.
+
+.. seealso:: :ocv:func:`divide`
+
+
+
+gpu::absdiff
+------------
+Computes per-element absolute difference of two matrices (or of a matrix and scalar).
+
+.. ocv:function:: void gpu::absdiff(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`absdiff`
+
+
+
+gpu::abs
+--------
+Computes an absolute value of each matrix element.
+
+.. ocv:function:: void gpu::abs(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`abs`
+
+
+
+gpu::sqr
+--------
+Computes a square value of each matrix element.
+
+.. ocv:function:: void gpu::sqr(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::sqrt
+---------
+Computes a square root of each matrix element.
+
+.. ocv:function:: void gpu::sqrt(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`sqrt`
+
+
+
+gpu::exp
+--------
+Computes an exponent of each matrix element.
+
+.. ocv:function:: void gpu::exp(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`exp`
+
+
+
+gpu::log
+--------
+Computes a natural logarithm of absolute value of each matrix element.
+
+.. ocv:function:: void gpu::log(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`log`
+
+
+
+gpu::pow
+--------
+Raises every matrix element to a power.
+
+.. ocv:function:: void gpu::pow(InputArray src, double power, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param power: Exponent of power.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+The function ``pow`` raises every element of the input matrix to ``power`` :
+
+.. math::
+
+    \texttt{dst} (I) =  \fork{\texttt{src}(I)^power}{if \texttt{power} is integer}{|\texttt{src}(I)|^power}{otherwise}
+
+.. seealso:: :ocv:func:`pow`
+
+
+
+gpu::compare
+------------
+Compares elements of two matrices (or of a matrix and scalar).
+
+.. ocv:function:: void gpu::compare(InputArray src1, InputArray src2, OutputArray dst, int cmpop, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param cmpop: Flag specifying the relation between the elements to be checked:
+
+            * **CMP_EQ:** ``a(.) == b(.)``
+            * **CMP_GT:** ``a(.) < b(.)``
+            * **CMP_GE:** ``a(.) <= b(.)``
+            * **CMP_LT:** ``a(.) < b(.)``
+            * **CMP_LE:** ``a(.) <= b(.)``
+            * **CMP_NE:** ``a(.) != b(.)``
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`compare`
+
+
+
+gpu::bitwise_not
+----------------
+Performs a per-element bitwise inversion.
+
+.. ocv:function:: void gpu::bitwise_not(InputArray src, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null())
+
+    :param src: Source matrix.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param mask: Optional operation mask. 8-bit single channel image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::bitwise_or
+---------------
+Performs a per-element bitwise disjunction of two matrices (or of matrix and scalar).
+
+.. ocv:function:: void gpu::bitwise_or(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param mask: Optional operation mask. 8-bit single channel image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::bitwise_and
+----------------
+Performs a per-element bitwise conjunction of two matrices (or of matrix and scalar).
+
+.. ocv:function:: void gpu::bitwise_and(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param mask: Optional operation mask. 8-bit single channel image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::bitwise_xor
+----------------
+Performs a per-element bitwise ``exclusive or`` operation of two matrices (or of matrix and scalar).
+
+.. ocv:function:: void gpu::bitwise_xor(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param mask: Optional operation mask. 8-bit single channel image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::rshift
+-----------
+Performs pixel by pixel right shift of an image by a constant value.
+
+.. ocv:function:: void gpu::rshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix. Supports 1, 3 and 4 channels images with integers elements.
+
+    :param val: Constant values, one per channel.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::lshift
+-----------
+Performs pixel by pixel right left of an image by a constant value.
+
+.. ocv:function:: void gpu::lshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source matrix. Supports 1, 3 and 4 channels images with ``CV_8U`` , ``CV_16U`` or ``CV_32S`` depth.
+
+    :param val: Constant values, one per channel.
+
+    :param dst: Destination matrix with the same size and type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::min
+--------
+Computes the per-element minimum of two matrices (or a matrix and a scalar).
+
+.. ocv:function:: void gpu::min(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`min`
+
+
+
+gpu::max
+--------
+Computes the per-element maximum of two matrices (or a matrix and a scalar).
+
+.. ocv:function:: void gpu::max(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src1: First source matrix or scalar.
+
+    :param src2: Second source matrix or scalar.
+
+    :param dst: Destination matrix that has the same size and type as the input array(s).
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`max`
+
+
+
+gpu::addWeighted
+----------------
+Computes the weighted sum of two arrays.
+
+.. ocv:function:: void gpu::addWeighted(InputArray src1, double alpha, InputArray src2, double beta, double gamma, OutputArray dst, int dtype = -1, Stream& stream = Stream::Null())
+
+    :param src1: First source array.
+
+    :param alpha: Weight for the first array elements.
+
+    :param src2: Second source array of the same size and channel number as  ``src1`` .
+
+    :param beta: Weight for the second array elements.
+
+    :param dst: Destination array that has the same size and number of channels as the input arrays.
+
+    :param gamma: Scalar added to each sum.
+
+    :param dtype: Optional depth of the destination array. When both input arrays have the same depth, ``dtype`` can be set to ``-1``, which will be equivalent to ``src1.depth()``.
+
+    :param stream: Stream for the asynchronous version.
+
+The function ``addWeighted`` calculates the weighted sum of two arrays as follows:
+
+.. math::
+
+    \texttt{dst} (I)= \texttt{saturate} ( \texttt{src1} (I)* \texttt{alpha} +  \texttt{src2} (I)* \texttt{beta} +  \texttt{gamma} )
+
+where ``I`` is a multi-dimensional index of array elements. In case of multi-channel arrays, each channel is processed independently.
+
+.. seealso:: :ocv:func:`addWeighted`
+
+
+
+gpu::threshold
+--------------
+Applies a fixed-level threshold to each array element.
+
+.. ocv:function:: double gpu::threshold(InputArray src, OutputArray dst, double thresh, double maxval, int type, Stream& stream = Stream::Null())
+
+    :param src: Source array (single-channel).
+
+    :param dst: Destination array with the same size and type as  ``src`` .
+
+    :param thresh: Threshold value.
+
+    :param maxval: Maximum value to use with  ``THRESH_BINARY`` and  ``THRESH_BINARY_INV`` threshold types.
+
+    :param type: Threshold type. For details, see  :ocv:func:`threshold` . The ``THRESH_OTSU`` threshold type is not supported.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`threshold`
+
+
+
+gpu::magnitude
+--------------
+Computes magnitudes of complex matrix elements.
+
+.. ocv:function:: void gpu::magnitude(InputArray xy, OutputArray magnitude, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::magnitude(InputArray x, InputArray y, OutputArray magnitude, Stream& stream = Stream::Null())
+
+    :param xy: Source complex matrix in the interleaved format ( ``CV_32FC2`` ).
+
+    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
+
+    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
+
+    :param magnitude: Destination matrix of float magnitudes ( ``CV_32FC1`` ).
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`magnitude`
+
+
+
+gpu::magnitudeSqr
+-----------------
+Computes squared magnitudes of complex matrix elements.
+
+.. ocv:function:: void gpu::magnitudeSqr(InputArray xy, OutputArray magnitude, Stream& stream=Stream::Null() )
+
+.. ocv:function:: void gpu::magnitudeSqr(InputArray x, InputArray y, OutputArray magnitude, Stream& stream = Stream::Null())
+
+    :param xy: Source complex matrix in the interleaved format ( ``CV_32FC2`` ).
+
+    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
+
+    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
+
+    :param magnitude: Destination matrix of float magnitude squares ( ``CV_32FC1`` ).
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::phase
+----------
+Computes polar angles of complex matrix elements.
+
+.. ocv:function:: void gpu::phase(InputArray x, InputArray y, OutputArray angle, bool angleInDegrees = false, Stream& stream = Stream::Null())
+
+    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
+
+    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
+
+    :param angle: Destination matrix of angles ( ``CV_32FC1`` ).
+
+    :param angleInDegrees: Flag for angles that must be evaluated in degrees.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`phase`
+
+
+
+gpu::cartToPolar
+----------------
+Converts Cartesian coordinates into polar.
+
+.. ocv:function:: void gpu::cartToPolar(InputArray x, InputArray y, OutputArray magnitude, OutputArray angle, bool angleInDegrees = false, Stream& stream = Stream::Null())
+
+    :param x: Source matrix containing real components ( ``CV_32FC1`` ).
+
+    :param y: Source matrix containing imaginary components ( ``CV_32FC1`` ).
+
+    :param magnitude: Destination matrix of float magnitudes ( ``CV_32FC1`` ).
+
+    :param angle: Destination matrix of angles ( ``CV_32FC1`` ).
+
+    :param angleInDegrees: Flag for angles that must be evaluated in degrees.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`cartToPolar`
+
+
+
+gpu::polarToCart
+----------------
+Converts polar coordinates into Cartesian.
+
+.. ocv:function:: void gpu::polarToCart(InputArray magnitude, InputArray angle, OutputArray x, OutputArray y, bool angleInDegrees = false, Stream& stream = Stream::Null())
+
+    :param magnitude: Source matrix containing magnitudes ( ``CV_32FC1`` ).
+
+    :param angle: Source matrix containing angles ( ``CV_32FC1`` ).
+
+    :param x: Destination matrix of real components ( ``CV_32FC1`` ).
+
+    :param y: Destination matrix of imaginary components ( ``CV_32FC1`` ).
+
+    :param angleInDegrees: Flag that indicates angles in degrees.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`polarToCart`
diff --git a/modules/gpuarithm/doc/gpuarithm.rst b/modules/gpuarithm/doc/gpuarithm.rst
new file mode 100644
index 0000000000..92a31ea43c
--- /dev/null
+++ b/modules/gpuarithm/doc/gpuarithm.rst
@@ -0,0 +1,11 @@
+*************************************************
+gpuarithm. GPU-accelerated Operations on Matrices
+*************************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    core
+    element_operations
+    reductions
+    arithm
diff --git a/modules/gpu/doc/matrix_reductions.rst b/modules/gpuarithm/doc/reductions.rst
similarity index 51%
rename from modules/gpu/doc/matrix_reductions.rst
rename to modules/gpuarithm/doc/reductions.rst
index e9229f8a81..b34c2d860d 100644
--- a/modules/gpu/doc/matrix_reductions.rst
+++ b/modules/gpuarithm/doc/reductions.rst
@@ -5,36 +5,17 @@ Matrix Reductions
 
 
 
-gpu::meanStdDev
--------------------
-Computes a mean value and a standard deviation of matrix elements.
-
-.. ocv:function:: void gpu::meanStdDev(const GpuMat& mtx, Scalar& mean, Scalar& stddev)
-.. ocv:function:: void gpu::meanStdDev(const GpuMat& mtx, Scalar& mean, Scalar& stddev, GpuMat& buf)
-
-    :param mtx: Source matrix.  ``CV_8UC1``  matrices are supported for now.
-
-    :param mean: Mean value.
-
-    :param stddev: Standard deviation value.
-
-    :param buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
-
-.. seealso:: :ocv:func:`meanStdDev`
-
-
-
 gpu::norm
--------------
+---------
 Returns the norm of a matrix (or difference of two matrices).
 
-.. ocv:function:: double gpu::norm(const GpuMat& src1, int normType=NORM_L2)
+.. ocv:function:: double gpu::norm(InputArray src1, int normType)
 
-.. ocv:function:: double gpu::norm(const GpuMat& src1, int normType, GpuMat& buf)
+.. ocv:function:: double gpu::norm(InputArray src1, int normType, GpuMat& buf)
 
-.. ocv:function:: double gpu::norm(const GpuMat& src1, int normType, const GpuMat& mask, GpuMat& buf)
+.. ocv:function:: double gpu::norm(InputArray src1, int normType, InputArray mask, GpuMat& buf)
 
-.. ocv:function:: double gpu::norm(const GpuMat& src1, const GpuMat& src2, int normType=NORM_L2)
+.. ocv:function:: double gpu::norm(InputArray src1, InputArray src2, int normType=NORM_L2)
 
     :param src1: Source matrix. Any matrices except 64F are supported.
 
@@ -51,14 +32,14 @@ Returns the norm of a matrix (or difference of two matrices).
 
 
 gpu::sum
-------------
+--------
 Returns the sum of matrix elements.
 
-.. ocv:function:: Scalar gpu::sum(const GpuMat& src)
+.. ocv:function:: Scalar gpu::sum(InputArray src)
 
-.. ocv:function:: Scalar gpu::sum(const GpuMat& src, GpuMat& buf)
+.. ocv:function:: Scalar gpu::sum(InputArray src, GpuMat& buf)
 
-.. ocv:function:: Scalar gpu::sum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
+.. ocv:function:: Scalar gpu::sum(InputArray src, InputArray mask, GpuMat& buf)
 
     :param src: Source image of any depth except for ``CV_64F`` .
 
@@ -71,14 +52,14 @@ Returns the sum of matrix elements.
 
 
 gpu::absSum
----------------
+-----------
 Returns the sum of absolute values for matrix elements.
 
-.. ocv:function:: Scalar gpu::absSum(const GpuMat& src)
+.. ocv:function:: Scalar gpu::absSum(InputArray src)
 
-.. ocv:function:: Scalar gpu::absSum(const GpuMat& src, GpuMat& buf)
+.. ocv:function:: Scalar gpu::absSum(InputArray src, GpuMat& buf)
 
-.. ocv:function:: Scalar gpu::absSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
+.. ocv:function:: Scalar gpu::absSum(InputArray src, InputArray mask, GpuMat& buf)
 
     :param src: Source image of any depth except for ``CV_64F`` .
 
@@ -89,14 +70,14 @@ Returns the sum of absolute values for matrix elements.
 
 
 gpu::sqrSum
----------------
+-----------
 Returns the squared sum of matrix elements.
 
-.. ocv:function:: Scalar gpu::sqrSum(const GpuMat& src)
+.. ocv:function:: Scalar gpu::sqrSum(InputArray src)
 
-.. ocv:function:: Scalar gpu::sqrSum(const GpuMat& src, GpuMat& buf)
+.. ocv:function:: Scalar gpu::sqrSum(InputArray src, GpuMat& buf)
 
-.. ocv:function:: Scalar gpu::sqrSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
+.. ocv:function:: Scalar gpu::sqrSum(InputArray src, InputArray mask, GpuMat& buf)
 
     :param src: Source image of any depth except for ``CV_64F`` .
 
@@ -107,12 +88,12 @@ Returns the squared sum of matrix elements.
 
 
 gpu::minMax
----------------
+-----------
 Finds global minimum and maximum matrix elements and returns their values.
 
-.. ocv:function:: void gpu::minMax(const GpuMat& src, double* minVal, double* maxVal=0, const GpuMat& mask=GpuMat())
+.. ocv:function:: void gpu::minMax(InputArray src, double* minVal, double* maxVal=0, InputArray mask=noArray())
 
-.. ocv:function:: void gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const GpuMat& mask, GpuMat& buf)
+.. ocv:function:: void gpu::minMax(InputArray src, double* minVal, double* maxVal, InputArray mask, GpuMat& buf)
 
     :param src: Single-channel source image.
 
@@ -131,12 +112,12 @@ The function does not work with ``CV_64F`` images on GPUs with the compute capab
 
 
 gpu::minMaxLoc
-------------------
+--------------
 Finds global minimum and maximum matrix elements and returns their values with locations.
 
-.. ocv:function:: void gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal=0, Point* minLoc=0, Point* maxLoc=0, const GpuMat& mask=GpuMat())
+.. ocv:function:: void gpu::minMaxLoc(InputArray src, double* minVal, double* maxVal=0, Point* minLoc=0, Point* maxLoc=0, InputArray mask=noArray())
 
-.. ocv:function:: void gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc, const GpuMat& mask, GpuMat& valbuf, GpuMat& locbuf)
+.. ocv:function:: void gpu::minMaxLoc(InputArray src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc, InputArray mask, GpuMat& valbuf, GpuMat& locbuf)
 
     :param src: Single-channel source image.
 
@@ -161,12 +142,12 @@ Finds global minimum and maximum matrix elements and returns their values with l
 
 
 gpu::countNonZero
----------------------
+-----------------
 Counts non-zero matrix elements.
 
-.. ocv:function:: int gpu::countNonZero(const GpuMat& src)
+.. ocv:function:: int gpu::countNonZero(InputArray src)
 
-.. ocv:function:: int gpu::countNonZero(const GpuMat& src, GpuMat& buf)
+.. ocv:function:: int gpu::countNonZero(InputArray src, GpuMat& buf)
 
     :param src: Single-channel source image.
 
@@ -182,7 +163,7 @@ gpu::reduce
 -----------
 Reduces a matrix to a vector.
 
-.. ocv:function:: void gpu::reduce(const GpuMat& mtx, GpuMat& vec, int dim, int reduceOp, int dtype = -1, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::reduce(InputArray mtx, OutputArray vec, int dim, int reduceOp, int dtype = -1, Stream& stream = Stream::Null())
 
     :param mtx: Source 2D matrix.
 
@@ -202,6 +183,113 @@ Reduces a matrix to a vector.
 
     :param dtype: When it is negative, the destination vector will have the same type as the source matrix. Otherwise, its type will be  ``CV_MAKE_TYPE(CV_MAT_DEPTH(dtype), mtx.channels())`` .
 
+    :param stream: Stream for the asynchronous version.
+
 The function ``reduce`` reduces the matrix to a vector by treating the matrix rows/columns as a set of 1D vectors and performing the specified operation on the vectors until a single row/column is obtained. For example, the function can be used to compute horizontal and vertical projections of a raster image. In case of ``CV_REDUCE_SUM`` and ``CV_REDUCE_AVG`` , the output may have a larger element bit-depth to preserve accuracy. And multi-channel arrays are also supported in these two reduction modes.
 
 .. seealso:: :ocv:func:`reduce`
+
+
+
+gpu::meanStdDev
+---------------
+Computes a mean value and a standard deviation of matrix elements.
+
+.. ocv:function:: void gpu::meanStdDev(InputArray mtx, Scalar& mean, Scalar& stddev)
+.. ocv:function:: void gpu::meanStdDev(InputArray mtx, Scalar& mean, Scalar& stddev, GpuMat& buf)
+
+    :param mtx: Source matrix.  ``CV_8UC1``  matrices are supported for now.
+
+    :param mean: Mean value.
+
+    :param stddev: Standard deviation value.
+
+    :param buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+.. seealso:: :ocv:func:`meanStdDev`
+
+
+
+gpu::rectStdDev
+---------------
+Computes a standard deviation of integral images.
+
+.. ocv:function:: void gpu::rectStdDev(InputArray src, InputArray sqr, OutputArray dst, Rect rect, Stream& stream = Stream::Null())
+
+    :param src: Source image. Only the ``CV_32SC1`` type is supported.
+
+    :param sqr: Squared source image. Only  the ``CV_32FC1`` type is supported.
+
+    :param dst: Destination image with the same type and size as  ``src`` .
+
+    :param rect: Rectangular window.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::normalize
+--------------
+Normalizes the norm or value range of an array.
+
+.. ocv:function:: void gpu::normalize(InputArray src, OutputArray dst, double alpha = 1, double beta = 0, int norm_type = NORM_L2, int dtype = -1, InputArray mask = noArray())
+
+.. ocv:function:: void gpu::normalize(InputArray src, OutputArray dst, double alpha, double beta, int norm_type, int dtype, InputArray mask, GpuMat& norm_buf, GpuMat& cvt_buf)
+
+    :param src: Input array.
+
+    :param dst: Output array of the same size as  ``src`` .
+
+    :param alpha: Norm value to normalize to or the lower range boundary in case of the range normalization.
+
+    :param beta: Upper range boundary in case of the range normalization; it is not used for the norm normalization.
+
+    :param normType: Normalization type ( ``NORM_MINMAX`` , ``NORM_L2`` , ``NORM_L1`` or ``NORM_INF`` ).
+
+    :param dtype: When negative, the output array has the same type as ``src``; otherwise, it has the same number of channels as  ``src`` and the depth ``=CV_MAT_DEPTH(dtype)``.
+
+    :param mask: Optional operation mask.
+
+    :param norm_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+    :param cvt_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+.. seealso:: :ocv:func:`normalize`
+
+
+
+gpu::integral
+-------------
+Computes an integral image.
+
+.. ocv:function:: void gpu::integral(InputArray src, OutputArray sum, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::integral(InputArray src, OutputArray sum, GpuMat& buffer, Stream& stream = Stream::Null())
+
+    :param src: Source image. Only  ``CV_8UC1`` images are supported for now.
+
+    :param sum: Integral image containing 32-bit unsigned integer values packed into  ``CV_32SC1`` .
+
+    :param buffer: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`integral`
+
+
+
+gpu::sqrIntegral
+----------------
+Computes a squared integral image.
+
+.. ocv:function:: void gpu::sqrIntegral(InputArray src, OutputArray sqsum, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::sqrIntegral(InputArray src, OutputArray sqsum, GpuMat& buf, Stream& stream = Stream::Null())
+
+    :param src: Source image. Only  ``CV_8UC1`` images are supported for now.
+
+    :param sqsum: Squared integral image containing 64-bit unsigned integer values packed into  ``CV_64FC1`` .
+
+    :param buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+    :param stream: Stream for the asynchronous version.
diff --git a/modules/gpuarithm/include/opencv2/gpuarithm.hpp b/modules/gpuarithm/include/opencv2/gpuarithm.hpp
new file mode 100644
index 0000000000..dabb4e2ab0
--- /dev/null
+++ b/modules/gpuarithm/include/opencv2/gpuarithm.hpp
@@ -0,0 +1,374 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUARITHM_HPP__
+#define __OPENCV_GPUARITHM_HPP__
+
+#ifndef __cplusplus
+#  error gpuarithm.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpu {
+
+//! adds one matrix to another (dst = src1 + src2)
+CV_EXPORTS void add(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), int dtype = -1, Stream& stream = Stream::Null());
+
+//! subtracts one matrix from another (dst = src1 - src2)
+CV_EXPORTS void subtract(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), int dtype = -1, Stream& stream = Stream::Null());
+
+//! computes element-wise weighted product of the two arrays (dst = scale * src1 * src2)
+CV_EXPORTS void multiply(InputArray src1, InputArray src2, OutputArray dst, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
+
+//! computes element-wise weighted quotient of the two arrays (dst = scale * (src1 / src2))
+CV_EXPORTS void divide(InputArray src1, InputArray src2, OutputArray dst, double scale = 1, int dtype = -1, Stream& stream = Stream::Null());
+
+//! computes element-wise weighted reciprocal of an array (dst = scale/src2)
+static inline void divide(double src1, InputArray src2, OutputArray dst, int dtype = -1, Stream& stream = Stream::Null())
+{
+    divide(src1, src2, dst, 1.0, dtype, stream);
+}
+
+//! computes element-wise absolute difference of two arrays (dst = abs(src1 - src2))
+CV_EXPORTS void absdiff(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes absolute value of each matrix element
+CV_EXPORTS void abs(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes square of each pixel in an image
+CV_EXPORTS void sqr(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes square root of each pixel in an image
+CV_EXPORTS void sqrt(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes exponent of each matrix element
+CV_EXPORTS void exp(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes natural logarithm of absolute value of each matrix element
+CV_EXPORTS void log(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes power of each matrix element:
+//!    (dst(i,j) = pow(     src(i,j) , power), if src.type() is integer
+//!    (dst(i,j) = pow(fabs(src(i,j)), power), otherwise
+CV_EXPORTS void pow(InputArray src, double power, OutputArray dst, Stream& stream = Stream::Null());
+
+//! compares elements of two arrays (dst = src1 <cmpop> src2)
+CV_EXPORTS void compare(InputArray src1, InputArray src2, OutputArray dst, int cmpop, Stream& stream = Stream::Null());
+
+//! performs per-elements bit-wise inversion
+CV_EXPORTS void bitwise_not(InputArray src, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null());
+
+//! calculates per-element bit-wise disjunction of two arrays
+CV_EXPORTS void bitwise_or(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null());
+
+//! calculates per-element bit-wise conjunction of two arrays
+CV_EXPORTS void bitwise_and(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null());
+
+//! calculates per-element bit-wise "exclusive or" operation
+CV_EXPORTS void bitwise_xor(InputArray src1, InputArray src2, OutputArray dst, InputArray mask = noArray(), Stream& stream = Stream::Null());
+
+//! pixel by pixel right shift of an image by a constant value
+//! supports 1, 3 and 4 channels images with integers elements
+CV_EXPORTS void rshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null());
+
+//! pixel by pixel left shift of an image by a constant value
+//! supports 1, 3 and 4 channels images with CV_8U, CV_16U or CV_32S depth
+CV_EXPORTS void lshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes per-element minimum of two arrays (dst = min(src1, src2))
+CV_EXPORTS void min(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes per-element maximum of two arrays (dst = max(src1, src2))
+CV_EXPORTS void max(InputArray src1, InputArray src2, OutputArray dst, Stream& stream = Stream::Null());
+
+//! computes the weighted sum of two arrays (dst = alpha*src1 + beta*src2 + gamma)
+CV_EXPORTS void addWeighted(InputArray src1, double alpha, InputArray src2, double beta, double gamma, OutputArray dst,
+                            int dtype = -1, Stream& stream = Stream::Null());
+
+//! adds scaled array to another one (dst = alpha*src1 + src2)
+static inline void scaleAdd(InputArray src1, double alpha, InputArray src2, OutputArray dst, Stream& stream = Stream::Null())
+{
+    addWeighted(src1, alpha, src2, 1.0, 0.0, dst, -1, stream);
+}
+
+//! applies fixed threshold to the image
+CV_EXPORTS double threshold(InputArray src, OutputArray dst, double thresh, double maxval, int type, Stream& stream = Stream::Null());
+
+//! computes magnitude of complex (x(i).re, x(i).im) vector
+//! supports only CV_32FC2 type
+CV_EXPORTS void magnitude(InputArray xy, OutputArray magnitude, Stream& stream = Stream::Null());
+
+//! computes squared magnitude of complex (x(i).re, x(i).im) vector
+//! supports only CV_32FC2 type
+CV_EXPORTS void magnitudeSqr(InputArray xy, OutputArray magnitude, Stream& stream = Stream::Null());
+
+//! computes magnitude of each (x(i), y(i)) vector
+//! supports only floating-point source
+CV_EXPORTS void magnitude(InputArray x, InputArray y, OutputArray magnitude, Stream& stream = Stream::Null());
+
+//! computes squared magnitude of each (x(i), y(i)) vector
+//! supports only floating-point source
+CV_EXPORTS void magnitudeSqr(InputArray x, InputArray y, OutputArray magnitude, Stream& stream = Stream::Null());
+
+//! computes angle of each (x(i), y(i)) vector
+//! supports only floating-point source
+CV_EXPORTS void phase(InputArray x, InputArray y, OutputArray angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
+
+//! converts Cartesian coordinates to polar
+//! supports only floating-point source
+CV_EXPORTS void cartToPolar(InputArray x, InputArray y, OutputArray magnitude, OutputArray angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
+
+//! converts polar coordinates to Cartesian
+//! supports only floating-point source
+CV_EXPORTS void polarToCart(InputArray magnitude, InputArray angle, OutputArray x, OutputArray y, bool angleInDegrees = false, Stream& stream = Stream::Null());
+
+//! makes multi-channel array out of several single-channel arrays
+CV_EXPORTS void merge(const GpuMat* src, size_t n, OutputArray dst, Stream& stream = Stream::Null());
+CV_EXPORTS void merge(const std::vector<GpuMat>& src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! copies each plane of a multi-channel array to a dedicated array
+CV_EXPORTS void split(InputArray src, GpuMat* dst, Stream& stream = Stream::Null());
+CV_EXPORTS void split(InputArray src, std::vector<GpuMat>& dst, Stream& stream = Stream::Null());
+
+//! transposes the matrix
+//! supports matrix with element size = 1, 4 and 8 bytes (CV_8UC1, CV_8UC4, CV_16UC2, CV_32FC1, etc)
+CV_EXPORTS void transpose(InputArray src1, OutputArray dst, Stream& stream = Stream::Null());
+
+//! reverses the order of the rows, columns or both in a matrix
+//! supports 1, 3 and 4 channels images with CV_8U, CV_16U, CV_32S or CV_32F depth
+CV_EXPORTS void flip(InputArray src, OutputArray dst, int flipCode, Stream& stream = Stream::Null());
+
+//! transforms 8-bit unsigned integers using lookup table: dst(i)=lut(src(i))
+//! destination array will have the depth type as lut and the same channels number as source
+//! supports CV_8UC1, CV_8UC3 types
+class CV_EXPORTS LookUpTable : public Algorithm
+{
+public:
+    virtual void transform(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+};
+
+CV_EXPORTS Ptr<LookUpTable> createLookUpTable(InputArray lut);
+
+//! copies 2D array to a larger destination array and pads borders with user-specifiable constant
+CV_EXPORTS void copyMakeBorder(InputArray src, OutputArray dst, int top, int bottom, int left, int right, int borderType,
+                               Scalar value = Scalar(), Stream& stream = Stream::Null());
+
+//! computes norm of array
+//! supports NORM_INF, NORM_L1, NORM_L2
+//! supports all matrices except 64F
+CV_EXPORTS double norm(InputArray src1, int normType, InputArray mask, GpuMat& buf);
+static inline double norm(InputArray src, int normType)
+{
+    GpuMat buf;
+    return norm(src, normType, GpuMat(), buf);
+}
+static inline double norm(InputArray src, int normType, GpuMat& buf)
+{
+    return norm(src, normType, GpuMat(), buf);
+}
+
+//! computes norm of the difference between two arrays
+//! supports NORM_INF, NORM_L1, NORM_L2
+//! supports only CV_8UC1 type
+CV_EXPORTS double norm(InputArray src1, InputArray src2, GpuMat& buf, int normType=NORM_L2);
+static inline double norm(InputArray src1, InputArray src2, int normType=NORM_L2)
+{
+    GpuMat buf;
+    return norm(src1, src2, buf, normType);
+}
+
+//! computes sum of array elements
+//! supports only single channel images
+CV_EXPORTS Scalar sum(InputArray src, InputArray mask, GpuMat& buf);
+static inline Scalar sum(InputArray src)
+{
+    GpuMat buf;
+    return sum(src, GpuMat(), buf);
+}
+static inline Scalar sum(InputArray src, GpuMat& buf)
+{
+    return sum(src, GpuMat(), buf);
+}
+
+//! computes sum of array elements absolute values
+//! supports only single channel images
+CV_EXPORTS Scalar absSum(InputArray src, InputArray mask, GpuMat& buf);
+static inline Scalar absSum(InputArray src)
+{
+    GpuMat buf;
+    return absSum(src, GpuMat(), buf);
+}
+static inline Scalar absSum(InputArray src, GpuMat& buf)
+{
+    return absSum(src, GpuMat(), buf);
+}
+
+//! computes squared sum of array elements
+//! supports only single channel images
+CV_EXPORTS Scalar sqrSum(InputArray src, InputArray mask, GpuMat& buf);
+static inline Scalar sqrSum(InputArray src)
+{
+    GpuMat buf;
+    return sqrSum(src, GpuMat(), buf);
+}
+static inline Scalar sqrSum(InputArray src, GpuMat& buf)
+{
+    return sqrSum(src, GpuMat(), buf);
+}
+
+//! finds global minimum and maximum array elements and returns their values
+CV_EXPORTS void minMax(InputArray src, double* minVal, double* maxVal, InputArray mask, GpuMat& buf);
+static inline void minMax(InputArray src, double* minVal, double* maxVal=0, InputArray mask=noArray())
+{
+    GpuMat buf;
+    minMax(src, minVal, maxVal, mask, buf);
+}
+
+//! finds global minimum and maximum array elements and returns their values with locations
+CV_EXPORTS void minMaxLoc(InputArray src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc,
+                          InputArray mask, GpuMat& valbuf, GpuMat& locbuf);
+static inline void minMaxLoc(InputArray src, double* minVal, double* maxVal=0, Point* minLoc=0, Point* maxLoc=0,
+                             InputArray mask=noArray())
+{
+    GpuMat valBuf, locBuf;
+    minMaxLoc(src, minVal, maxVal, minLoc, maxLoc, mask, valBuf, locBuf);
+}
+
+//! counts non-zero array elements
+CV_EXPORTS int countNonZero(InputArray src, GpuMat& buf);
+static inline int countNonZero(const GpuMat& src)
+{
+    GpuMat buf;
+    return countNonZero(src, buf);
+}
+
+//! reduces a matrix to a vector
+CV_EXPORTS void reduce(InputArray mtx, OutputArray vec, int dim, int reduceOp, int dtype = -1, Stream& stream = Stream::Null());
+
+//! computes mean value and standard deviation of all or selected array elements
+//! supports only CV_8UC1 type
+CV_EXPORTS void meanStdDev(InputArray mtx, Scalar& mean, Scalar& stddev, GpuMat& buf);
+static inline void meanStdDev(InputArray src, Scalar& mean, Scalar& stddev)
+{
+    GpuMat buf;
+    meanStdDev(src, mean, stddev, buf);
+}
+
+//! computes the standard deviation of integral images
+//! supports only CV_32SC1 source type and CV_32FC1 sqr type
+//! output will have CV_32FC1 type
+CV_EXPORTS void rectStdDev(InputArray src, InputArray sqr, OutputArray dst, Rect rect, Stream& stream = Stream::Null());
+
+//! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
+CV_EXPORTS void normalize(InputArray src, OutputArray dst, double alpha, double beta,
+                          int norm_type, int dtype, InputArray mask, GpuMat& norm_buf, GpuMat& cvt_buf);
+static inline void normalize(InputArray src, OutputArray dst, double alpha = 1, double beta = 0,
+                             int norm_type = NORM_L2, int dtype = -1, InputArray mask = noArray())
+{
+    GpuMat norm_buf;
+    GpuMat cvt_buf;
+    normalize(src, dst, alpha, beta, norm_type, dtype, mask, norm_buf, cvt_buf);
+}
+
+//! computes the integral image
+//! sum will have CV_32S type, but will contain unsigned int values
+//! supports only CV_8UC1 source type
+CV_EXPORTS void integral(InputArray src, OutputArray sum, GpuMat& buffer, Stream& stream = Stream::Null());
+static inline void integralBuffered(InputArray src, OutputArray sum, GpuMat& buffer, Stream& stream = Stream::Null())
+{
+    integral(src, sum, buffer, stream);
+}
+static inline void integral(InputArray src, OutputArray sum, Stream& stream = Stream::Null())
+{
+    GpuMat buffer;
+    integral(src, sum, buffer, stream);
+}
+
+//! computes squared integral image
+//! result matrix will have 64F type, but will contain 64U values
+//! supports source images of 8UC1 type only
+CV_EXPORTS void sqrIntegral(InputArray src, OutputArray sqsum, GpuMat& buf, Stream& stream = Stream::Null());
+static inline void sqrIntegral(InputArray src, OutputArray sqsum, Stream& stream = Stream::Null())
+{
+    GpuMat buffer;
+    sqrIntegral(src, sqsum, buffer, stream);
+}
+
+CV_EXPORTS void gemm(InputArray src1, InputArray src2, double alpha,
+                     InputArray src3, double beta, OutputArray dst, int flags = 0, Stream& stream = Stream::Null());
+
+//! performs per-element multiplication of two full (not packed) Fourier spectrums
+//! supports 32FC2 matrixes only (interleaved format)
+CV_EXPORTS void mulSpectrums(InputArray src1, InputArray src2, OutputArray dst, int flags, bool conjB=false, Stream& stream = Stream::Null());
+
+//! performs per-element multiplication of two full (not packed) Fourier spectrums
+//! supports 32FC2 matrixes only (interleaved format)
+CV_EXPORTS void mulAndScaleSpectrums(InputArray src1, InputArray src2, OutputArray dst, int flags, float scale, bool conjB=false, Stream& stream = Stream::Null());
+
+//! Performs a forward or inverse discrete Fourier transform (1D or 2D) of floating point matrix.
+//! Param dft_size is the size of DFT transform.
+//!
+//! If the source matrix is not continous, then additional copy will be done,
+//! so to avoid copying ensure the source matrix is continous one. If you want to use
+//! preallocated output ensure it is continuous too, otherwise it will be reallocated.
+//!
+//! Being implemented via CUFFT real-to-complex transform result contains only non-redundant values
+//! in CUFFT's format. Result as full complex matrix for such kind of transform cannot be retrieved.
+//!
+//! For complex-to-real transform it is assumed that the source matrix is packed in CUFFT's format.
+CV_EXPORTS void dft(InputArray src, OutputArray dst, Size dft_size, int flags=0, Stream& stream = Stream::Null());
+
+//! computes convolution (or cross-correlation) of two images using discrete Fourier transform
+//! supports source images of 32FC1 type only
+//! result matrix will have 32FC1 type
+class CV_EXPORTS Convolution : public Algorithm
+{
+public:
+    virtual void convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null()) = 0;
+};
+
+CV_EXPORTS Ptr<Convolution> createConvolution(Size user_block_size = Size());
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUARITHM_HPP__ */
diff --git a/modules/gpuarithm/perf/perf_arithm.cpp b/modules/gpuarithm/perf/perf_arithm.cpp
new file mode 100644
index 0000000000..b18c8a8c0e
--- /dev/null
+++ b/modules/gpuarithm/perf/perf_arithm.cpp
@@ -0,0 +1,307 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// GEMM
+
+CV_FLAGS(GemmFlags, 0, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T)
+#define ALL_GEMM_FLAGS Values(GemmFlags(0), GemmFlags(cv::GEMM_1_T), GemmFlags(cv::GEMM_2_T), GemmFlags(cv::GEMM_3_T), \
+                              GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_3_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T | cv::GEMM_3_T))
+
+DEF_PARAM_TEST(Sz_Type_Flags, cv::Size, MatType, GemmFlags);
+
+PERF_TEST_P(Sz_Type_Flags, GEMM,
+            Combine(Values(cv::Size(512, 512), cv::Size(1024, 1024)),
+                    Values(CV_32FC1, CV_32FC2, CV_64FC1),
+                    ALL_GEMM_FLAGS))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int flags = GET_PARAM(2);
+
+    cv::Mat src1(size, type);
+    declare.in(src1, WARMUP_RNG);
+
+    cv::Mat src2(size, type);
+    declare.in(src2, WARMUP_RNG);
+
+    cv::Mat src3(size, type);
+    declare.in(src3, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        declare.time(5.0);
+
+        const cv::gpu::GpuMat d_src1(src1);
+        const cv::gpu::GpuMat d_src2(src2);
+        const cv::gpu::GpuMat d_src3(src3);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::gemm(d_src1, d_src2, 1.0, d_src3, 1.0, dst, flags);
+
+        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        declare.time(50.0);
+
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::gemm(src1, src2, 1.0, src3, 1.0, dst, flags);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MulSpectrums
+
+CV_FLAGS(DftFlags, 0, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
+
+DEF_PARAM_TEST(Sz_Flags, cv::Size, DftFlags);
+
+PERF_TEST_P(Sz_Flags, MulSpectrums,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(0, DftFlags(cv::DFT_ROWS))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int flag = GET_PARAM(1);
+
+    cv::Mat a(size, CV_32FC2);
+    cv::Mat b(size, CV_32FC2);
+    declare.in(a, b, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_a(a);
+        const cv::gpu::GpuMat d_b(b);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::mulSpectrums(d_a, d_b, dst, flag);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::mulSpectrums(a, b, dst, flag);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MulAndScaleSpectrums
+
+PERF_TEST_P(Sz, MulAndScaleSpectrums,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    const float scale = 1.f / size.area();
+
+    cv::Mat src1(size, CV_32FC2);
+    cv::Mat src2(size, CV_32FC2);
+    declare.in(src1,src2, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src1(src1);
+        const cv::gpu::GpuMat d_src2(src2);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::mulAndScaleSpectrums(d_src1, d_src2, dst, cv::DFT_ROWS, scale, false);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Dft
+
+PERF_TEST_P(Sz_Flags, Dft,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(0, DftFlags(cv::DFT_ROWS), DftFlags(cv::DFT_INVERSE))))
+{
+    declare.time(10.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int flag = GET_PARAM(1);
+
+    cv::Mat src(size, CV_32FC2);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::dft(d_src, dst, size, flag);
+
+        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::dft(src, dst, flag);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Convolve
+
+DEF_PARAM_TEST(Sz_KernelSz_Ccorr, cv::Size, int, bool);
+
+PERF_TEST_P(Sz_KernelSz_Ccorr, Convolve,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(17, 27, 32, 64),
+                    Bool()))
+{
+    declare.time(10.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int templ_size = GET_PARAM(1);
+    const bool ccorr = GET_PARAM(2);
+
+    const cv::Mat image(size, CV_32FC1);
+    const cv::Mat templ(templ_size, templ_size, CV_32FC1);
+    declare.in(image, templ, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat d_image = cv::gpu::createContinuous(size, CV_32FC1);
+        d_image.upload(image);
+
+        cv::gpu::GpuMat d_templ = cv::gpu::createContinuous(templ_size, templ_size, CV_32FC1);
+        d_templ.upload(templ);
+
+        cv::Ptr<cv::gpu::Convolution> convolution = cv::gpu::createConvolution();
+
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() convolution->convolve(d_image, d_templ, dst, ccorr);
+
+        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        if (ccorr)
+            FAIL_NO_CPU();
+
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::filter2D(image, dst, image.depth(), templ);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Integral
+
+PERF_TEST_P(Sz, Integral,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+        cv::gpu::GpuMat d_buf;
+
+        TEST_CYCLE() cv::gpu::integral(d_src, dst, d_buf);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::integral(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// IntegralSqr
+
+PERF_TEST_P(Sz, IntegralSqr,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst, buf;
+
+        TEST_CYCLE() cv::gpu::sqrIntegral(d_src, dst, buf);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuarithm/perf/perf_core.cpp b/modules/gpuarithm/perf/perf_core.cpp
new file mode 100644
index 0000000000..0add472ca3
--- /dev/null
+++ b/modules/gpuarithm/perf/perf_core.cpp
@@ -0,0 +1,321 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+#define ARITHM_MAT_DEPTH Values(CV_8U, CV_16U, CV_32F, CV_64F)
+
+//////////////////////////////////////////////////////////////////////
+// Merge
+
+PERF_TEST_P(Sz_Depth_Cn, Merge,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    ARITHM_MAT_DEPTH,
+                    Values(2, 3, 4)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    std::vector<cv::Mat> src(channels);
+    for (int i = 0; i < channels; ++i)
+    {
+        src[i].create(size, depth);
+        declare.in(src[i], WARMUP_RNG);
+    }
+
+    if (PERF_RUN_GPU())
+    {
+        std::vector<cv::gpu::GpuMat> d_src(channels);
+        for (int i = 0; i < channels; ++i)
+            d_src[i].upload(src[i]);
+
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::merge(d_src, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-10);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::merge(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Split
+
+PERF_TEST_P(Sz_Depth_Cn, Split,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    ARITHM_MAT_DEPTH,
+                    Values(2, 3, 4)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    cv::Mat src(size, CV_MAKE_TYPE(depth, channels));
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        std::vector<cv::gpu::GpuMat> dst;
+
+        TEST_CYCLE() cv::gpu::split(d_src, dst);
+
+        const cv::gpu::GpuMat& dst0 = dst[0];
+        const cv::gpu::GpuMat& dst1 = dst[1];
+
+        GPU_SANITY_CHECK(dst0, 1e-10);
+        GPU_SANITY_CHECK(dst1, 1e-10);
+    }
+    else
+    {
+        std::vector<cv::Mat> dst;
+
+        TEST_CYCLE() cv::split(src, dst);
+
+        const cv::Mat& dst0 = dst[0];
+        const cv::Mat& dst1 = dst[1];
+
+        CPU_SANITY_CHECK(dst0);
+        CPU_SANITY_CHECK(dst1);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Transpose
+
+PERF_TEST_P(Sz_Type, Transpose,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8UC1, CV_8UC4, CV_16UC2, CV_16SC2, CV_32SC1, CV_32SC2, CV_64FC1)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::transpose(d_src, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-10);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::transpose(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Flip
+
+enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
+CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Code, cv::Size, MatDepth, MatCn, FlipCode);
+
+PERF_TEST_P(Sz_Depth_Cn_Code, Flip,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    FlipCode::all()))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int flipCode = GET_PARAM(3);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::flip(d_src, dst, flipCode);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::flip(src, dst, flipCode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// LutOneChannel
+
+PERF_TEST_P(Sz_Type, LutOneChannel,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8UC1, CV_8UC3)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    cv::Mat lut(1, 256, CV_8UC1);
+    declare.in(lut, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::LookUpTable> lutAlg = cv::gpu::createLookUpTable(lut);
+
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() lutAlg->transform(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::LUT(src, lut, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// LutMultiChannel
+
+PERF_TEST_P(Sz_Type, LutMultiChannel,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values<MatType>(CV_8UC3)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    cv::Mat lut(1, 256, CV_MAKE_TYPE(CV_8U, src.channels()));
+    declare.in(lut, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::LookUpTable> lutAlg = cv::gpu::createLookUpTable(lut);
+
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() lutAlg->transform(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::LUT(src, lut, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// CopyMakeBorder
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Border, cv::Size, MatDepth, MatCn, BorderMode);
+
+PERF_TEST_P(Sz_Depth_Cn_Border, CopyMakeBorder,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    ALL_BORDER_MODES))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int borderMode = GET_PARAM(3);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::copyMakeBorder(d_src, dst, 5, 5, 5, 5, borderMode);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::copyMakeBorder(src, dst, 5, 5, 5, 5, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpu/perf/perf_core.cpp b/modules/gpuarithm/perf/perf_element_operations.cpp
similarity index 62%
rename from modules/gpu/perf/perf_core.cpp
rename to modules/gpuarithm/perf/perf_element_operations.cpp
index 22840f9f46..1370da5804 100644
--- a/modules/gpu/perf/perf_core.cpp
+++ b/modules/gpuarithm/perf/perf_element_operations.cpp
@@ -1,3 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
@@ -6,93 +48,10 @@ using namespace perf;
 
 #define ARITHM_MAT_DEPTH Values(CV_8U, CV_16U, CV_32F, CV_64F)
 
-//////////////////////////////////////////////////////////////////////
-// Merge
-
-PERF_TEST_P(Sz_Depth_Cn, Core_Merge,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    ARITHM_MAT_DEPTH,
-                    Values(2, 3, 4)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    std::vector<cv::Mat> src(channels);
-    for (int i = 0; i < channels; ++i)
-    {
-        src[i].create(size, depth);
-        declare.in(src[i], WARMUP_RNG);
-    }
-
-    if (PERF_RUN_GPU())
-    {
-        std::vector<cv::gpu::GpuMat> d_src(channels);
-        for (int i = 0; i < channels; ++i)
-            d_src[i].upload(src[i]);
-
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::merge(d_src, dst);
-
-        GPU_SANITY_CHECK(dst, 1e-10);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::merge(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Split
-
-PERF_TEST_P(Sz_Depth_Cn, Core_Split,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    ARITHM_MAT_DEPTH,
-                    Values(2, 3, 4)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    cv::Mat src(size, CV_MAKE_TYPE(depth, channels));
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        std::vector<cv::gpu::GpuMat> dst;
-
-        TEST_CYCLE() cv::gpu::split(d_src, dst);
-
-        const cv::gpu::GpuMat& dst0 = dst[0];
-        const cv::gpu::GpuMat& dst1 = dst[1];
-
-        GPU_SANITY_CHECK(dst0, 1e-10);
-        GPU_SANITY_CHECK(dst1, 1e-10);
-    }
-    else
-    {
-        std::vector<cv::Mat> dst;
-
-        TEST_CYCLE() cv::split(src, dst);
-
-        const cv::Mat& dst0 = dst[0];
-        const cv::Mat& dst1 = dst[1];
-
-        CPU_SANITY_CHECK(dst0);
-        CPU_SANITY_CHECK(dst1);
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // AddMat
 
-PERF_TEST_P(Sz_Depth, Core_AddMat,
+PERF_TEST_P(Sz_Depth, AddMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -128,7 +87,7 @@ PERF_TEST_P(Sz_Depth, Core_AddMat,
 //////////////////////////////////////////////////////////////////////
 // AddScalar
 
-PERF_TEST_P(Sz_Depth, Core_AddScalar,
+PERF_TEST_P(Sz_Depth, AddScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -163,7 +122,7 @@ PERF_TEST_P(Sz_Depth, Core_AddScalar,
 //////////////////////////////////////////////////////////////////////
 // SubtractMat
 
-PERF_TEST_P(Sz_Depth, Core_SubtractMat,
+PERF_TEST_P(Sz_Depth, SubtractMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -199,7 +158,7 @@ PERF_TEST_P(Sz_Depth, Core_SubtractMat,
 //////////////////////////////////////////////////////////////////////
 // SubtractScalar
 
-PERF_TEST_P(Sz_Depth, Core_SubtractScalar,
+PERF_TEST_P(Sz_Depth, SubtractScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -234,7 +193,7 @@ PERF_TEST_P(Sz_Depth, Core_SubtractScalar,
 //////////////////////////////////////////////////////////////////////
 // MultiplyMat
 
-PERF_TEST_P(Sz_Depth, Core_MultiplyMat,
+PERF_TEST_P(Sz_Depth, MultiplyMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -270,7 +229,7 @@ PERF_TEST_P(Sz_Depth, Core_MultiplyMat,
 //////////////////////////////////////////////////////////////////////
 // MultiplyScalar
 
-PERF_TEST_P(Sz_Depth, Core_MultiplyScalar,
+PERF_TEST_P(Sz_Depth, MultiplyScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -305,7 +264,7 @@ PERF_TEST_P(Sz_Depth, Core_MultiplyScalar,
 //////////////////////////////////////////////////////////////////////
 // DivideMat
 
-PERF_TEST_P(Sz_Depth, Core_DivideMat,
+PERF_TEST_P(Sz_Depth, DivideMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -341,7 +300,7 @@ PERF_TEST_P(Sz_Depth, Core_DivideMat,
 //////////////////////////////////////////////////////////////////////
 // DivideScalar
 
-PERF_TEST_P(Sz_Depth, Core_DivideScalar,
+PERF_TEST_P(Sz_Depth, DivideScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -376,7 +335,7 @@ PERF_TEST_P(Sz_Depth, Core_DivideScalar,
 //////////////////////////////////////////////////////////////////////
 // DivideScalarInv
 
-PERF_TEST_P(Sz_Depth, Core_DivideScalarInv,
+PERF_TEST_P(Sz_Depth, DivideScalarInv,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -411,7 +370,7 @@ PERF_TEST_P(Sz_Depth, Core_DivideScalarInv,
 //////////////////////////////////////////////////////////////////////
 // AbsDiffMat
 
-PERF_TEST_P(Sz_Depth, Core_AbsDiffMat,
+PERF_TEST_P(Sz_Depth, AbsDiffMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -447,7 +406,7 @@ PERF_TEST_P(Sz_Depth, Core_AbsDiffMat,
 //////////////////////////////////////////////////////////////////////
 // AbsDiffScalar
 
-PERF_TEST_P(Sz_Depth, Core_AbsDiffScalar,
+PERF_TEST_P(Sz_Depth, AbsDiffScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH))
 {
@@ -482,7 +441,7 @@ PERF_TEST_P(Sz_Depth, Core_AbsDiffScalar,
 //////////////////////////////////////////////////////////////////////
 // Abs
 
-PERF_TEST_P(Sz_Depth, Core_Abs,
+PERF_TEST_P(Sz_Depth, Abs,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_16S, CV_32F)))
 {
@@ -510,7 +469,7 @@ PERF_TEST_P(Sz_Depth, Core_Abs,
 //////////////////////////////////////////////////////////////////////
 // Sqr
 
-PERF_TEST_P(Sz_Depth, Core_Sqr,
+PERF_TEST_P(Sz_Depth, Sqr,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16S, CV_32F)))
 {
@@ -538,7 +497,7 @@ PERF_TEST_P(Sz_Depth, Core_Sqr,
 //////////////////////////////////////////////////////////////////////
 // Sqrt
 
-PERF_TEST_P(Sz_Depth, Core_Sqrt,
+PERF_TEST_P(Sz_Depth, Sqrt,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16S, CV_32F)))
 {
@@ -570,7 +529,7 @@ PERF_TEST_P(Sz_Depth, Core_Sqrt,
 //////////////////////////////////////////////////////////////////////
 // Log
 
-PERF_TEST_P(Sz_Depth, Core_Log,
+PERF_TEST_P(Sz_Depth, Log,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16S, CV_32F)))
 {
@@ -602,7 +561,7 @@ PERF_TEST_P(Sz_Depth, Core_Log,
 //////////////////////////////////////////////////////////////////////
 // Exp
 
-PERF_TEST_P(Sz_Depth, Core_Exp,
+PERF_TEST_P(Sz_Depth, Exp,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16S, CV_32F)))
 {
@@ -636,7 +595,7 @@ PERF_TEST_P(Sz_Depth, Core_Exp,
 
 DEF_PARAM_TEST(Sz_Depth_Power, cv::Size, MatDepth, double);
 
-PERF_TEST_P(Sz_Depth_Power, Core_Pow,
+PERF_TEST_P(Sz_Depth_Power, Pow,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16S, CV_32F),
                     Values(0.3, 2.0, 2.4)))
@@ -671,14 +630,13 @@ PERF_TEST_P(Sz_Depth_Power, Core_Pow,
 // CompareMat
 
 CV_ENUM(CmpCode, cv::CMP_EQ, cv::CMP_GT, cv::CMP_GE, cv::CMP_LT, cv::CMP_LE, cv::CMP_NE)
-#define ALL_CMP_CODES ValuesIn(CmpCode::all())
 
 DEF_PARAM_TEST(Sz_Depth_Code, cv::Size, MatDepth, CmpCode);
 
-PERF_TEST_P(Sz_Depth_Code, Core_CompareMat,
+PERF_TEST_P(Sz_Depth_Code, CompareMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH,
-                    ALL_CMP_CODES))
+                    CmpCode::all()))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -713,10 +671,10 @@ PERF_TEST_P(Sz_Depth_Code, Core_CompareMat,
 //////////////////////////////////////////////////////////////////////
 // CompareScalar
 
-PERF_TEST_P(Sz_Depth_Code, Core_CompareScalar,
+PERF_TEST_P(Sz_Depth_Code, CompareScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     ARITHM_MAT_DEPTH,
-                    ALL_CMP_CODES))
+                    CmpCode::all()))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -750,7 +708,7 @@ PERF_TEST_P(Sz_Depth_Code, Core_CompareScalar,
 //////////////////////////////////////////////////////////////////////
 // BitwiseNot
 
-PERF_TEST_P(Sz_Depth, Core_BitwiseNot,
+PERF_TEST_P(Sz_Depth, BitwiseNot,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S)))
 {
@@ -782,7 +740,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseNot,
 //////////////////////////////////////////////////////////////////////
 // BitwiseAndMat
 
-PERF_TEST_P(Sz_Depth, Core_BitwiseAndMat,
+PERF_TEST_P(Sz_Depth, BitwiseAndMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S)))
 {
@@ -818,7 +776,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseAndMat,
 //////////////////////////////////////////////////////////////////////
 // BitwiseAndScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseAndScalar,
+PERF_TEST_P(Sz_Depth_Cn, BitwiseAndScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S),
                     GPU_CHANNELS_1_3_4))
@@ -858,7 +816,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseAndScalar,
 //////////////////////////////////////////////////////////////////////
 // BitwiseOrMat
 
-PERF_TEST_P(Sz_Depth, Core_BitwiseOrMat,
+PERF_TEST_P(Sz_Depth, BitwiseOrMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S)))
 {
@@ -894,7 +852,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseOrMat,
 //////////////////////////////////////////////////////////////////////
 // BitwiseOrScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseOrScalar,
+PERF_TEST_P(Sz_Depth_Cn, BitwiseOrScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S),
                     GPU_CHANNELS_1_3_4))
@@ -934,7 +892,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseOrScalar,
 //////////////////////////////////////////////////////////////////////
 // BitwiseXorMat
 
-PERF_TEST_P(Sz_Depth, Core_BitwiseXorMat,
+PERF_TEST_P(Sz_Depth, BitwiseXorMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S)))
 {
@@ -970,7 +928,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseXorMat,
 //////////////////////////////////////////////////////////////////////
 // BitwiseXorScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseXorScalar,
+PERF_TEST_P(Sz_Depth_Cn, BitwiseXorScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S),
                     GPU_CHANNELS_1_3_4))
@@ -1010,7 +968,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseXorScalar,
 //////////////////////////////////////////////////////////////////////
 // RShift
 
-PERF_TEST_P(Sz_Depth_Cn, Core_RShift,
+PERF_TEST_P(Sz_Depth_Cn, RShift,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S),
                     GPU_CHANNELS_1_3_4))
@@ -1044,7 +1002,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_RShift,
 //////////////////////////////////////////////////////////////////////
 // LShift
 
-PERF_TEST_P(Sz_Depth_Cn, Core_LShift,
+PERF_TEST_P(Sz_Depth_Cn, LShift,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32S),
                     GPU_CHANNELS_1_3_4))
@@ -1078,7 +1036,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_LShift,
 //////////////////////////////////////////////////////////////////////
 // MinMat
 
-PERF_TEST_P(Sz_Depth, Core_MinMat,
+PERF_TEST_P(Sz_Depth, MinMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32F)))
 {
@@ -1114,7 +1072,7 @@ PERF_TEST_P(Sz_Depth, Core_MinMat,
 //////////////////////////////////////////////////////////////////////
 // MinScalar
 
-PERF_TEST_P(Sz_Depth, Core_MinScalar,
+PERF_TEST_P(Sz_Depth, MinScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32F)))
 {
@@ -1149,7 +1107,7 @@ PERF_TEST_P(Sz_Depth, Core_MinScalar,
 //////////////////////////////////////////////////////////////////////
 // MaxMat
 
-PERF_TEST_P(Sz_Depth, Core_MaxMat,
+PERF_TEST_P(Sz_Depth, MaxMat,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32F)))
 {
@@ -1185,7 +1143,7 @@ PERF_TEST_P(Sz_Depth, Core_MaxMat,
 //////////////////////////////////////////////////////////////////////
 // MaxScalar
 
-PERF_TEST_P(Sz_Depth, Core_MaxScalar,
+PERF_TEST_P(Sz_Depth, MaxScalar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32F)))
 {
@@ -1222,7 +1180,7 @@ PERF_TEST_P(Sz_Depth, Core_MaxScalar,
 
 DEF_PARAM_TEST(Sz_3Depth, cv::Size, MatDepth, MatDepth, MatDepth);
 
-PERF_TEST_P(Sz_3Depth, Core_AddWeighted,
+PERF_TEST_P(Sz_3Depth, AddWeighted,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8U, CV_16U, CV_32F, CV_64F),
                     Values(CV_8U, CV_16U, CV_32F, CV_64F),
@@ -1259,207 +1217,10 @@ PERF_TEST_P(Sz_3Depth, Core_AddWeighted,
     }
 }
 
-//////////////////////////////////////////////////////////////////////
-// GEMM
-
-CV_FLAGS(GemmFlags, 0, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T)
-#define ALL_GEMM_FLAGS Values(0, CV_GEMM_A_T, CV_GEMM_B_T, CV_GEMM_C_T, CV_GEMM_A_T | CV_GEMM_B_T, CV_GEMM_A_T | CV_GEMM_C_T, CV_GEMM_A_T | CV_GEMM_B_T | CV_GEMM_C_T)
-
-DEF_PARAM_TEST(Sz_Type_Flags, cv::Size, MatType, GemmFlags);
-
-PERF_TEST_P(Sz_Type_Flags, Core_GEMM,
-            Combine(Values(cv::Size(512, 512), cv::Size(1024, 1024)),
-                    Values(CV_32FC1, CV_32FC2, CV_64FC1),
-                    ALL_GEMM_FLAGS))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int flags = GET_PARAM(2);
-
-    cv::Mat src1(size, type);
-    declare.in(src1, WARMUP_RNG);
-
-    cv::Mat src2(size, type);
-    declare.in(src2, WARMUP_RNG);
-
-    cv::Mat src3(size, type);
-    declare.in(src3, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        declare.time(5.0);
-
-        const cv::gpu::GpuMat d_src1(src1);
-        const cv::gpu::GpuMat d_src2(src2);
-        const cv::gpu::GpuMat d_src3(src3);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::gemm(d_src1, d_src2, 1.0, d_src3, 1.0, dst, flags);
-
-        GPU_SANITY_CHECK(dst, 1e-6);
-    }
-    else
-    {
-        declare.time(50.0);
-
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::gemm(src1, src2, 1.0, src3, 1.0, dst, flags);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Transpose
-
-PERF_TEST_P(Sz_Type, Core_Transpose,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8UC1, CV_8UC4, CV_16UC2, CV_16SC2, CV_32SC1, CV_32SC2, CV_64FC1)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::transpose(d_src, dst);
-
-        GPU_SANITY_CHECK(dst, 1e-10);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::transpose(src, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Flip
-
-enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
-CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
-#define ALL_FLIP_CODES ValuesIn(FlipCode::all())
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Code, cv::Size, MatDepth, MatCn, FlipCode);
-
-PERF_TEST_P(Sz_Depth_Cn_Code, Core_Flip,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4,
-                    ALL_FLIP_CODES))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int flipCode = GET_PARAM(3);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::flip(d_src, dst, flipCode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::flip(src, dst, flipCode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// LutOneChannel
-
-PERF_TEST_P(Sz_Type, Core_LutOneChannel,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8UC1, CV_8UC3)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    cv::Mat lut(1, 256, CV_8UC1);
-    declare.in(lut, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::LUT(d_src, lut, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::LUT(src, lut, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// LutMultiChannel
-
-PERF_TEST_P(Sz_Type, Core_LutMultiChannel,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values<MatType>(CV_8UC3)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    cv::Mat lut(1, 256, CV_MAKE_TYPE(CV_8U, src.channels()));
-    declare.in(lut, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::LUT(d_src, lut, dst);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::LUT(src, lut, dst);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // MagnitudeComplex
 
-PERF_TEST_P(Sz, Core_MagnitudeComplex,
+PERF_TEST_P(Sz, MagnitudeComplex,
             GPU_TYPICAL_MAT_SIZES)
 {
     const cv::Size size = GetParam();
@@ -1492,7 +1253,7 @@ PERF_TEST_P(Sz, Core_MagnitudeComplex,
 //////////////////////////////////////////////////////////////////////
 // MagnitudeSqrComplex
 
-PERF_TEST_P(Sz, Core_MagnitudeSqrComplex,
+PERF_TEST_P(Sz, MagnitudeSqrComplex,
             GPU_TYPICAL_MAT_SIZES)
 {
     const cv::Size size = GetParam();
@@ -1518,7 +1279,7 @@ PERF_TEST_P(Sz, Core_MagnitudeSqrComplex,
 //////////////////////////////////////////////////////////////////////
 // Magnitude
 
-PERF_TEST_P(Sz, Core_Magnitude,
+PERF_TEST_P(Sz, Magnitude,
             GPU_TYPICAL_MAT_SIZES)
 {
     const cv::Size size = GetParam();
@@ -1552,7 +1313,7 @@ PERF_TEST_P(Sz, Core_Magnitude,
 //////////////////////////////////////////////////////////////////////
 // MagnitudeSqr
 
-PERF_TEST_P(Sz, Core_MagnitudeSqr,
+PERF_TEST_P(Sz, MagnitudeSqr,
             GPU_TYPICAL_MAT_SIZES)
 {
     const cv::Size size = GetParam();
@@ -1584,7 +1345,7 @@ PERF_TEST_P(Sz, Core_MagnitudeSqr,
 
 DEF_PARAM_TEST(Sz_AngleInDegrees, cv::Size, bool);
 
-PERF_TEST_P(Sz_AngleInDegrees, Core_Phase,
+PERF_TEST_P(Sz_AngleInDegrees, Phase,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Bool()))
 {
@@ -1620,7 +1381,7 @@ PERF_TEST_P(Sz_AngleInDegrees, Core_Phase,
 //////////////////////////////////////////////////////////////////////
 // CartToPolar
 
-PERF_TEST_P(Sz_AngleInDegrees, Core_CartToPolar,
+PERF_TEST_P(Sz_AngleInDegrees, CartToPolar,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Bool()))
 {
@@ -1660,7 +1421,7 @@ PERF_TEST_P(Sz_AngleInDegrees, Core_CartToPolar,
 //////////////////////////////////////////////////////////////////////
 // PolarToCart
 
-PERF_TEST_P(Sz_AngleInDegrees, Core_PolarToCart,
+PERF_TEST_P(Sz_AngleInDegrees, PolarToCart,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Bool()))
 {
@@ -1698,411 +1459,38 @@ PERF_TEST_P(Sz_AngleInDegrees, Core_PolarToCart,
 }
 
 //////////////////////////////////////////////////////////////////////
-// MeanStdDev
+// Threshold
 
-PERF_TEST_P(Sz, Core_MeanStdDev,
-            GPU_TYPICAL_MAT_SIZES)
-{
-    const cv::Size size = GetParam();
+CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
 
-    cv::Mat src(size, CV_8UC1);
-    declare.in(src, WARMUP_RNG);
+DEF_PARAM_TEST(Sz_Depth_Op, cv::Size, MatDepth, ThreshOp);
 
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        cv::Scalar gpu_mean;
-        cv::Scalar gpu_stddev;
-
-        TEST_CYCLE() cv::gpu::meanStdDev(d_src, gpu_mean, gpu_stddev, d_buf);
-
-        SANITY_CHECK(gpu_mean);
-        SANITY_CHECK(gpu_stddev);
-    }
-    else
-    {
-        cv::Scalar cpu_mean;
-        cv::Scalar cpu_stddev;
-
-        TEST_CYCLE() cv::meanStdDev(src, cpu_mean, cpu_stddev);
-
-        SANITY_CHECK(cpu_mean);
-        SANITY_CHECK(cpu_stddev);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Norm
-
-DEF_PARAM_TEST(Sz_Depth_Norm, cv::Size, MatDepth, NormType);
-
-PERF_TEST_P(Sz_Depth_Norm, Core_Norm,
+PERF_TEST_P(Sz_Depth_Op, Threshold,
             Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32S, CV_32F),
-                    Values(NormType(cv::NORM_INF), NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
+            Values(CV_8U, CV_16U, CV_32F, CV_64F),
+            ThreshOp::all()))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
-    const int normType = GET_PARAM(2);
+    const int threshOp = GET_PARAM(2);
 
     cv::Mat src(size, depth);
     declare.in(src, WARMUP_RNG);
 
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        double gpu_dst;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::norm(d_src, normType, d_buf);
-
-        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
-    }
-    else
-    {
-        double cpu_dst;
-
-        TEST_CYCLE() cpu_dst = cv::norm(src, normType);
-
-        SANITY_CHECK(cpu_dst, 1e-6, ERROR_RELATIVE);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// NormDiff
-
-DEF_PARAM_TEST(Sz_Norm, cv::Size, NormType);
-
-PERF_TEST_P(Sz_Norm, Core_NormDiff,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(NormType(cv::NORM_INF), NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int normType = GET_PARAM(1);
-
-    cv::Mat src1(size, CV_8UC1);
-    declare.in(src1, WARMUP_RNG);
-
-    cv::Mat src2(size, CV_8UC1);
-    declare.in(src2, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src1(src1);
-        const cv::gpu::GpuMat d_src2(src2);
-        double gpu_dst;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::norm(d_src1, d_src2, normType);
-
-        SANITY_CHECK(gpu_dst);
-
-    }
-    else
-    {
-        double cpu_dst;
-
-        TEST_CYCLE() cpu_dst = cv::norm(src1, src2, normType);
-
-        SANITY_CHECK(cpu_dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Sum
-
-PERF_TEST_P(Sz_Depth_Cn, Core_Sum,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        cv::Scalar gpu_dst;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::sum(d_src, d_buf);
-
-        SANITY_CHECK(gpu_dst, 1e-5, ERROR_RELATIVE);
-    }
-    else
-    {
-        cv::Scalar cpu_dst;
-
-        TEST_CYCLE() cpu_dst = cv::sum(src);
-
-        SANITY_CHECK(cpu_dst, 1e-6, ERROR_RELATIVE);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// SumAbs
-
-PERF_TEST_P(Sz_Depth_Cn, Core_SumAbs,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        cv::Scalar gpu_dst;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::absSum(d_src, d_buf);
-
-        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// SumSqr
-
-PERF_TEST_P(Sz_Depth_Cn, Core_SumSqr,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values<MatDepth>(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        cv::Scalar gpu_dst;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::sqrSum(d_src, d_buf);
-
-        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MinMax
-
-PERF_TEST_P(Sz_Depth, Core_MinMax,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        double gpu_minVal, gpu_maxVal;
-
-        TEST_CYCLE() cv::gpu::minMax(d_src, &gpu_minVal, &gpu_maxVal, cv::gpu::GpuMat(), d_buf);
-
-        SANITY_CHECK(gpu_minVal, 1e-10);
-        SANITY_CHECK(gpu_maxVal, 1e-10);
-    }
-    else
-    {
-        double cpu_minVal, cpu_maxVal;
-
-        TEST_CYCLE() cv::minMaxLoc(src, &cpu_minVal, &cpu_maxVal);
-
-        SANITY_CHECK(cpu_minVal);
-        SANITY_CHECK(cpu_maxVal);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MinMaxLoc
-
-PERF_TEST_P(Sz_Depth, Core_MinMaxLoc,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_valbuf, d_locbuf;
-        double gpu_minVal, gpu_maxVal;
-        cv::Point gpu_minLoc, gpu_maxLoc;
-
-        TEST_CYCLE() cv::gpu::minMaxLoc(d_src, &gpu_minVal, &gpu_maxVal, &gpu_minLoc, &gpu_maxLoc, cv::gpu::GpuMat(), d_valbuf, d_locbuf);
-
-        SANITY_CHECK(gpu_minVal, 1e-10);
-        SANITY_CHECK(gpu_maxVal, 1e-10);
-    }
-    else
-    {
-        double cpu_minVal, cpu_maxVal;
-        cv::Point cpu_minLoc, cpu_maxLoc;
-
-        TEST_CYCLE() cv::minMaxLoc(src, &cpu_minVal, &cpu_maxVal, &cpu_minLoc, &cpu_maxLoc);
-
-        SANITY_CHECK(cpu_minVal);
-        SANITY_CHECK(cpu_maxVal);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// CountNonZero
-
-PERF_TEST_P(Sz_Depth, Core_CountNonZero,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-
-    cv::Mat src(size, depth);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat d_buf;
-        int gpu_dst = 0;
-
-        TEST_CYCLE() gpu_dst = cv::gpu::countNonZero(d_src, d_buf);
-
-        SANITY_CHECK(gpu_dst);
-    }
-    else
-    {
-        int cpu_dst = 0;
-
-        TEST_CYCLE() cpu_dst = cv::countNonZero(src);
-
-        SANITY_CHECK(cpu_dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Reduce
-
-CV_ENUM(ReduceCode, CV_REDUCE_SUM, CV_REDUCE_AVG, CV_REDUCE_MAX, CV_REDUCE_MIN)
-#define ALL_REDUCE_CODES ValuesIn(ReduceCode::all())
-
-enum {Rows = 0, Cols = 1};
-CV_ENUM(ReduceDim, Rows, Cols)
-#define ALL_REDUCE_DIMS ValuesIn(ReduceDim::all())
-
-DEF_PARAM_TEST(Sz_Depth_Cn_Code_Dim, cv::Size, MatDepth, MatCn, ReduceCode, ReduceDim);
-
-PERF_TEST_P(Sz_Depth_Cn_Code_Dim, Core_Reduce,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_16S, CV_32F),
-                    Values(1, 2, 3, 4),
-                    ALL_REDUCE_CODES,
-                    ALL_REDUCE_DIMS))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int reduceOp = GET_PARAM(3);
-    const int dim = GET_PARAM(4);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
     if (PERF_RUN_GPU())
     {
         const cv::gpu::GpuMat d_src(src);
         cv::gpu::GpuMat dst;
 
-        TEST_CYCLE() cv::gpu::reduce(d_src, dst, dim, reduceOp);
+        TEST_CYCLE() cv::gpu::threshold(d_src, dst, 100.0, 255.0, threshOp);
 
-        GPU_SANITY_CHECK(dst);
+        GPU_SANITY_CHECK(dst, 1e-10);
     }
     else
     {
         cv::Mat dst;
 
-        TEST_CYCLE() cv::reduce(src, dst, dim, reduceOp);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-//////////////////////////////////////////////////////////////////////
-// Normalize
-
-DEF_PARAM_TEST(Sz_Depth_NormType, cv::Size, MatDepth, NormType);
-
-PERF_TEST_P(Sz_Depth_NormType, Core_Normalize,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F, CV_64F),
-                    Values(NormType(cv::NORM_INF),
-                           NormType(cv::NORM_L1),
-                           NormType(cv::NORM_L2),
-                           NormType(cv::NORM_MINMAX))))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int norm_type = GET_PARAM(2);
-
-    const double alpha = 1;
-    const double beta = 0;
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_norm_buf, d_cvt_buf;
-
-        TEST_CYCLE() cv::gpu::normalize(d_src, dst, alpha, beta, norm_type, type, cv::gpu::GpuMat(), d_norm_buf, d_cvt_buf);
-
-        GPU_SANITY_CHECK(dst, 1e-6);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::normalize(src, dst, alpha, beta, norm_type, type);
+        TEST_CYCLE() cv::threshold(src, dst, 100.0, 255.0, threshOp);
 
         CPU_SANITY_CHECK(dst);
     }
diff --git a/modules/gpuarithm/perf/perf_main.cpp b/modules/gpuarithm/perf/perf_main.cpp
new file mode 100644
index 0000000000..552cf5c162
--- /dev/null
+++ b/modules/gpuarithm/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpuarithm)
diff --git a/modules/gpu/include/opencv2/gpu/device/static_check.hpp b/modules/gpuarithm/perf/perf_precomp.hpp
similarity index 78%
rename from modules/gpu/include/opencv2/gpu/device/static_check.hpp
rename to modules/gpuarithm/perf/perf_precomp.hpp
index e77691b7b2..0b50ab0f9a 100644
--- a/modules/gpu/include/opencv2/gpu/device/static_check.hpp
+++ b/modules/gpuarithm/perf/perf_precomp.hpp
@@ -40,28 +40,26 @@
 //
 //M*/
 
-#ifndef __OPENCV_GPU_GPU_DEVICE_STATIC_CHECK_HPP__
-#define __OPENCV_GPU_GPU_DEVICE_STATIC_CHECK_HPP__
-
-#if defined(__CUDACC__)
-    #define __OPENCV_GPU_HOST_DEVICE__ __host__ __device__ __forceinline__
-#else
-    #define __OPENCV_GPU_HOST_DEVICE__
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
 #endif
 
-namespace cv { namespace gpu
-{
-    namespace device
-    {
-        template<bool expr> struct Static {};
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
 
-        template<> struct Static<true>
-        {
-            __OPENCV_GPU_HOST_DEVICE__ static void check() {};
-        };
-    }
-}}
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-#undef __OPENCV_GPU_HOST_DEVICE__
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
 
-#endif /* __OPENCV_GPU_GPU_DEVICE_STATIC_CHECK_HPP__ */
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpuarithm/perf/perf_reductions.cpp b/modules/gpuarithm/perf/perf_reductions.cpp
new file mode 100644
index 0000000000..c541ce0e28
--- /dev/null
+++ b/modules/gpuarithm/perf/perf_reductions.cpp
@@ -0,0 +1,467 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// Norm
+
+DEF_PARAM_TEST(Sz_Depth_Norm, cv::Size, MatDepth, NormType);
+
+PERF_TEST_P(Sz_Depth_Norm, Norm,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32S, CV_32F),
+                    Values(NormType(cv::NORM_INF), NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int normType = GET_PARAM(2);
+
+    cv::Mat src(size, depth);
+    if (depth == CV_8U)
+        cv::randu(src, 0, 254);
+    else
+        declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        double gpu_dst;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::norm(d_src, normType, d_buf);
+
+        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        double cpu_dst;
+
+        TEST_CYCLE() cpu_dst = cv::norm(src, normType);
+
+        SANITY_CHECK(cpu_dst, 1e-6, ERROR_RELATIVE);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// NormDiff
+
+DEF_PARAM_TEST(Sz_Norm, cv::Size, NormType);
+
+PERF_TEST_P(Sz_Norm, NormDiff,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(NormType(cv::NORM_INF), NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int normType = GET_PARAM(1);
+
+    cv::Mat src1(size, CV_8UC1);
+    declare.in(src1, WARMUP_RNG);
+
+    cv::Mat src2(size, CV_8UC1);
+    declare.in(src2, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src1(src1);
+        const cv::gpu::GpuMat d_src2(src2);
+        cv::gpu::GpuMat d_buf;
+        double gpu_dst;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::norm(d_src1, d_src2, d_buf, normType);
+
+        SANITY_CHECK(gpu_dst);
+
+    }
+    else
+    {
+        double cpu_dst;
+
+        TEST_CYCLE() cpu_dst = cv::norm(src1, src2, normType);
+
+        SANITY_CHECK(cpu_dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Sum
+
+PERF_TEST_P(Sz_Depth_Cn, Sum,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        cv::Scalar gpu_dst;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::sum(d_src, d_buf);
+
+        SANITY_CHECK(gpu_dst, 1e-5, ERROR_RELATIVE);
+    }
+    else
+    {
+        cv::Scalar cpu_dst;
+
+        TEST_CYCLE() cpu_dst = cv::sum(src);
+
+        SANITY_CHECK(cpu_dst, 1e-6, ERROR_RELATIVE);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// SumAbs
+
+PERF_TEST_P(Sz_Depth_Cn, SumAbs,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        cv::Scalar gpu_dst;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::absSum(d_src, d_buf);
+
+        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// SumSqr
+
+PERF_TEST_P(Sz_Depth_Cn, SumSqr,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values<MatDepth>(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        cv::Scalar gpu_dst;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::sqrSum(d_src, d_buf);
+
+        SANITY_CHECK(gpu_dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MinMax
+
+PERF_TEST_P(Sz_Depth, MinMax,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, depth);
+    if (depth == CV_8U)
+        cv::randu(src, 0, 254);
+    else
+        declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        double gpu_minVal, gpu_maxVal;
+
+        TEST_CYCLE() cv::gpu::minMax(d_src, &gpu_minVal, &gpu_maxVal, cv::gpu::GpuMat(), d_buf);
+
+        SANITY_CHECK(gpu_minVal, 1e-10);
+        SANITY_CHECK(gpu_maxVal, 1e-10);
+    }
+    else
+    {
+        double cpu_minVal, cpu_maxVal;
+
+        TEST_CYCLE() cv::minMaxLoc(src, &cpu_minVal, &cpu_maxVal);
+
+        SANITY_CHECK(cpu_minVal);
+        SANITY_CHECK(cpu_maxVal);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MinMaxLoc
+
+PERF_TEST_P(Sz_Depth, MinMaxLoc,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, depth);
+    if (depth == CV_8U)
+        cv::randu(src, 0, 254);
+    else
+        declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_valbuf, d_locbuf;
+        double gpu_minVal, gpu_maxVal;
+        cv::Point gpu_minLoc, gpu_maxLoc;
+
+        TEST_CYCLE() cv::gpu::minMaxLoc(d_src, &gpu_minVal, &gpu_maxVal, &gpu_minLoc, &gpu_maxLoc, cv::gpu::GpuMat(), d_valbuf, d_locbuf);
+
+        SANITY_CHECK(gpu_minVal, 1e-10);
+        SANITY_CHECK(gpu_maxVal, 1e-10);
+    }
+    else
+    {
+        double cpu_minVal, cpu_maxVal;
+        cv::Point cpu_minLoc, cpu_maxLoc;
+
+        TEST_CYCLE() cv::minMaxLoc(src, &cpu_minVal, &cpu_maxVal, &cpu_minLoc, &cpu_maxLoc);
+
+        SANITY_CHECK(cpu_minVal);
+        SANITY_CHECK(cpu_maxVal);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// CountNonZero
+
+PERF_TEST_P(Sz_Depth, CountNonZero,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F, CV_64F)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, depth);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        int gpu_dst = 0;
+
+        TEST_CYCLE() gpu_dst = cv::gpu::countNonZero(d_src, d_buf);
+
+        SANITY_CHECK(gpu_dst);
+    }
+    else
+    {
+        int cpu_dst = 0;
+
+        TEST_CYCLE() cpu_dst = cv::countNonZero(src);
+
+        SANITY_CHECK(cpu_dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Reduce
+
+CV_ENUM(ReduceCode, REDUCE_SUM, REDUCE_AVG, REDUCE_MAX, REDUCE_MIN)
+
+enum {Rows = 0, Cols = 1};
+CV_ENUM(ReduceDim, Rows, Cols)
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Code_Dim, cv::Size, MatDepth, MatCn, ReduceCode, ReduceDim);
+
+PERF_TEST_P(Sz_Depth_Cn_Code_Dim, Reduce,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_16S, CV_32F),
+                    Values(1, 2, 3, 4),
+                    ReduceCode::all(),
+                    ReduceDim::all()))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int reduceOp = GET_PARAM(3);
+    const int dim = GET_PARAM(4);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::reduce(d_src, dst, dim, reduceOp);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::reduce(src, dst, dim, reduceOp);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Normalize
+
+DEF_PARAM_TEST(Sz_Depth_NormType, cv::Size, MatDepth, NormType);
+
+PERF_TEST_P(Sz_Depth_NormType, Normalize,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F, CV_64F),
+                    Values(NormType(cv::NORM_INF),
+                           NormType(cv::NORM_L1),
+                           NormType(cv::NORM_L2),
+                           NormType(cv::NORM_MINMAX))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int norm_type = GET_PARAM(2);
+
+    const double alpha = 1;
+    const double beta = 0;
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+        cv::gpu::GpuMat d_norm_buf, d_cvt_buf;
+
+        TEST_CYCLE() cv::gpu::normalize(d_src, dst, alpha, beta, norm_type, type, cv::gpu::GpuMat(), d_norm_buf, d_cvt_buf);
+
+        GPU_SANITY_CHECK(dst, 1e-6);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::normalize(src, dst, alpha, beta, norm_type, type);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MeanStdDev
+
+PERF_TEST_P(Sz, MeanStdDev,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_buf;
+        cv::Scalar gpu_mean;
+        cv::Scalar gpu_stddev;
+
+        TEST_CYCLE() cv::gpu::meanStdDev(d_src, gpu_mean, gpu_stddev, d_buf);
+
+        SANITY_CHECK(gpu_mean);
+        SANITY_CHECK(gpu_stddev);
+    }
+    else
+    {
+        cv::Scalar cpu_mean;
+        cv::Scalar cpu_stddev;
+
+        TEST_CYCLE() cv::meanStdDev(src, cpu_mean, cpu_stddev);
+
+        SANITY_CHECK(cpu_mean);
+        SANITY_CHECK(cpu_stddev);
+    }
+}
diff --git a/modules/gpuarithm/src/arithm.cpp b/modules/gpuarithm/src/arithm.cpp
new file mode 100644
index 0000000000..78071ba8a7
--- /dev/null
+++ b/modules/gpuarithm/src/arithm.cpp
@@ -0,0 +1,636 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::gemm(InputArray, InputArray, double, InputArray, double, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::mulSpectrums(InputArray, InputArray, OutputArray, int, bool, Stream&) { throw_no_cuda(); }
+void cv::gpu::mulAndScaleSpectrums(InputArray, InputArray, OutputArray, int, float, bool, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::dft(InputArray, OutputArray, Size, int, Stream&) { throw_no_cuda(); }
+
+Ptr<Convolution> cv::gpu::createConvolution(Size) { throw_no_cuda(); return Ptr<Convolution>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace
+{
+    #define error_entry(entry)  { entry, #entry }
+
+    struct ErrorEntry
+    {
+        int code;
+        const char* str;
+    };
+
+    struct ErrorEntryComparer
+    {
+        int code;
+        ErrorEntryComparer(int code_) : code(code_) {}
+        bool operator()(const ErrorEntry& e) const { return e.code == code; }
+    };
+
+    String getErrorString(int code, const ErrorEntry* errors, size_t n)
+    {
+        size_t idx = std::find_if(errors, errors + n, ErrorEntryComparer(code)) - errors;
+
+        const char* msg = (idx != n) ? errors[idx].str : "Unknown error code";
+        String str = cv::format("%s [Code = %d]", msg, code);
+
+        return str;
+    }
+}
+
+#ifdef HAVE_CUBLAS
+    namespace
+    {
+        const ErrorEntry cublas_errors[] =
+        {
+            error_entry( CUBLAS_STATUS_SUCCESS ),
+            error_entry( CUBLAS_STATUS_NOT_INITIALIZED ),
+            error_entry( CUBLAS_STATUS_ALLOC_FAILED ),
+            error_entry( CUBLAS_STATUS_INVALID_VALUE ),
+            error_entry( CUBLAS_STATUS_ARCH_MISMATCH ),
+            error_entry( CUBLAS_STATUS_MAPPING_ERROR ),
+            error_entry( CUBLAS_STATUS_EXECUTION_FAILED ),
+            error_entry( CUBLAS_STATUS_INTERNAL_ERROR )
+        };
+
+        const size_t cublas_error_num = sizeof(cublas_errors) / sizeof(cublas_errors[0]);
+
+        static inline void ___cublasSafeCall(cublasStatus_t err, const char* file, const int line, const char* func)
+        {
+            if (CUBLAS_STATUS_SUCCESS != err)
+            {
+                String msg = getErrorString(err, cublas_errors, cublas_error_num);
+                cv::error(cv::Error::GpuApiCallError, msg, func, file, line);
+            }
+        }
+    }
+
+    #if defined(__GNUC__)
+        #define cublasSafeCall(expr)  ___cublasSafeCall(expr, __FILE__, __LINE__, __func__)
+    #else /* defined(__CUDACC__) || defined(__MSVC__) */
+        #define cublasSafeCall(expr)  ___cublasSafeCall(expr, __FILE__, __LINE__, "")
+    #endif
+#endif // HAVE_CUBLAS
+
+#ifdef HAVE_CUFFT
+    namespace
+    {
+        //////////////////////////////////////////////////////////////////////////
+        // CUFFT errors
+
+        const ErrorEntry cufft_errors[] =
+        {
+            error_entry( CUFFT_INVALID_PLAN ),
+            error_entry( CUFFT_ALLOC_FAILED ),
+            error_entry( CUFFT_INVALID_TYPE ),
+            error_entry( CUFFT_INVALID_VALUE ),
+            error_entry( CUFFT_INTERNAL_ERROR ),
+            error_entry( CUFFT_EXEC_FAILED ),
+            error_entry( CUFFT_SETUP_FAILED ),
+            error_entry( CUFFT_INVALID_SIZE ),
+            error_entry( CUFFT_UNALIGNED_DATA )
+        };
+
+        const int cufft_error_num = sizeof(cufft_errors) / sizeof(cufft_errors[0]);
+
+        void ___cufftSafeCall(int err, const char* file, const int line, const char* func)
+        {
+            if (CUFFT_SUCCESS != err)
+            {
+                String msg = getErrorString(err, cufft_errors, cufft_error_num);
+                cv::error(cv::Error::GpuApiCallError, msg, func, file, line);
+            }
+        }
+    }
+
+    #if defined(__GNUC__)
+        #define cufftSafeCall(expr)  ___cufftSafeCall(expr, __FILE__, __LINE__, __func__)
+    #else /* defined(__CUDACC__) || defined(__MSVC__) */
+        #define cufftSafeCall(expr)  ___cufftSafeCall(expr, __FILE__, __LINE__, "")
+    #endif
+
+#endif
+
+////////////////////////////////////////////////////////////////////////
+// gemm
+
+void cv::gpu::gemm(InputArray _src1, InputArray _src2, double alpha, InputArray _src3, double beta, OutputArray _dst, int flags, Stream& stream)
+{
+#ifndef HAVE_CUBLAS
+    (void) _src1;
+    (void) _src2;
+    (void) alpha;
+    (void) _src3;
+    (void) beta;
+    (void) _dst;
+    (void) flags;
+    (void) stream;
+    CV_Error(Error::StsNotImplemented, "The library was build without CUBLAS");
+#else
+    // CUBLAS works with column-major matrices
+
+    GpuMat src1 = _src1.getGpuMat();
+    GpuMat src2 = _src2.getGpuMat();
+    GpuMat src3 = _src3.getGpuMat();
+
+    CV_Assert( src1.type() == CV_32FC1 || src1.type() == CV_32FC2 || src1.type() == CV_64FC1 || src1.type() == CV_64FC2 );
+    CV_Assert( src2.type() == src1.type() && (src3.empty() || src3.type() == src1.type()) );
+
+    if (src1.depth() == CV_64F)
+    {
+        if (!deviceSupports(NATIVE_DOUBLE))
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+    }
+
+    bool tr1 = (flags & GEMM_1_T) != 0;
+    bool tr2 = (flags & GEMM_2_T) != 0;
+    bool tr3 = (flags & GEMM_3_T) != 0;
+
+    if (src1.type() == CV_64FC2)
+    {
+        if (tr1 || tr2 || tr3)
+            CV_Error(cv::Error::StsNotImplemented, "transpose operation doesn't implemented for CV_64FC2 type");
+    }
+
+    Size src1Size = tr1 ? Size(src1.rows, src1.cols) : src1.size();
+    Size src2Size = tr2 ? Size(src2.rows, src2.cols) : src2.size();
+    Size src3Size = tr3 ? Size(src3.rows, src3.cols) : src3.size();
+    Size dstSize(src2Size.width, src1Size.height);
+
+    CV_Assert( src1Size.width == src2Size.height );
+    CV_Assert( src3.empty() || src3Size == dstSize );
+
+    _dst.create(dstSize, src1.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    if (beta != 0)
+    {
+        if (src3.empty())
+        {
+            dst.setTo(Scalar::all(0), stream);
+        }
+        else
+        {
+            if (tr3)
+            {
+                gpu::transpose(src3, dst, stream);
+            }
+            else
+            {
+                src3.copyTo(dst, stream);
+            }
+        }
+    }
+
+    cublasHandle_t handle;
+    cublasSafeCall( cublasCreate_v2(&handle) );
+
+    cublasSafeCall( cublasSetStream_v2(handle, StreamAccessor::getStream(stream)) );
+
+    cublasSafeCall( cublasSetPointerMode_v2(handle, CUBLAS_POINTER_MODE_HOST) );
+
+    const float alphaf = static_cast<float>(alpha);
+    const float betaf = static_cast<float>(beta);
+
+    const cuComplex alphacf = make_cuComplex(alphaf, 0);
+    const cuComplex betacf = make_cuComplex(betaf, 0);
+
+    const cuDoubleComplex alphac = make_cuDoubleComplex(alpha, 0);
+    const cuDoubleComplex betac = make_cuDoubleComplex(beta, 0);
+
+    cublasOperation_t transa = tr2 ? CUBLAS_OP_T : CUBLAS_OP_N;
+    cublasOperation_t transb = tr1 ? CUBLAS_OP_T : CUBLAS_OP_N;
+
+    switch (src1.type())
+    {
+    case CV_32FC1:
+        cublasSafeCall( cublasSgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
+            &alphaf,
+            src2.ptr<float>(), static_cast<int>(src2.step / sizeof(float)),
+            src1.ptr<float>(), static_cast<int>(src1.step / sizeof(float)),
+            &betaf,
+            dst.ptr<float>(), static_cast<int>(dst.step / sizeof(float))) );
+        break;
+
+    case CV_64FC1:
+        cublasSafeCall( cublasDgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
+            &alpha,
+            src2.ptr<double>(), static_cast<int>(src2.step / sizeof(double)),
+            src1.ptr<double>(), static_cast<int>(src1.step / sizeof(double)),
+            &beta,
+            dst.ptr<double>(), static_cast<int>(dst.step / sizeof(double))) );
+        break;
+
+    case CV_32FC2:
+        cublasSafeCall( cublasCgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
+            &alphacf,
+            src2.ptr<cuComplex>(), static_cast<int>(src2.step / sizeof(cuComplex)),
+            src1.ptr<cuComplex>(), static_cast<int>(src1.step / sizeof(cuComplex)),
+            &betacf,
+            dst.ptr<cuComplex>(), static_cast<int>(dst.step / sizeof(cuComplex))) );
+        break;
+
+    case CV_64FC2:
+        cublasSafeCall( cublasZgemm_v2(handle, transa, transb, tr2 ? src2.rows : src2.cols, tr1 ? src1.cols : src1.rows, tr2 ? src2.cols : src2.rows,
+            &alphac,
+            src2.ptr<cuDoubleComplex>(), static_cast<int>(src2.step / sizeof(cuDoubleComplex)),
+            src1.ptr<cuDoubleComplex>(), static_cast<int>(src1.step / sizeof(cuDoubleComplex)),
+            &betac,
+            dst.ptr<cuDoubleComplex>(), static_cast<int>(dst.step / sizeof(cuDoubleComplex))) );
+        break;
+    }
+
+    cublasSafeCall( cublasDestroy_v2(handle) );
+#endif
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// mulSpectrums
+
+#ifdef HAVE_CUFFT
+
+namespace cv { namespace gpu { namespace cudev
+{
+    void mulSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream);
+
+    void mulSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream);
+}}}
+
+#endif
+
+void cv::gpu::mulSpectrums(InputArray _src1, InputArray _src2, OutputArray _dst, int flags, bool conjB, Stream& stream)
+{
+#ifndef HAVE_CUFFT
+    (void) _src1;
+    (void) _src2;
+    (void) _dst;
+    (void) flags;
+    (void) conjB;
+    (void) stream;
+    throw_no_cuda();
+#else
+    (void) flags;
+
+    typedef void (*Caller)(const PtrStep<cufftComplex>, const PtrStep<cufftComplex>, PtrStepSz<cufftComplex>, cudaStream_t stream);
+    static Caller callers[] = { cudev::mulSpectrums, cudev::mulSpectrums_CONJ };
+
+    GpuMat src1 = _src1.getGpuMat();
+    GpuMat src2 = _src2.getGpuMat();
+
+    CV_Assert( src1.type() == src2.type() && src1.type() == CV_32FC2 );
+    CV_Assert( src1.size() == src2.size() );
+
+    _dst.create(src1.size(), CV_32FC2);
+    GpuMat dst = _dst.getGpuMat();
+
+    Caller caller = callers[(int)conjB];
+    caller(src1, src2, dst, StreamAccessor::getStream(stream));
+#endif
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// mulAndScaleSpectrums
+
+#ifdef HAVE_CUFFT
+
+namespace cv { namespace gpu { namespace cudev
+{
+    void mulAndScaleSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream);
+
+    void mulAndScaleSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream);
+}}}
+
+#endif
+
+void cv::gpu::mulAndScaleSpectrums(InputArray _src1, InputArray _src2, OutputArray _dst, int flags, float scale, bool conjB, Stream& stream)
+{
+#ifndef HAVE_CUFFT
+    (void) _src1;
+    (void) _src2;
+    (void) _dst;
+    (void) flags;
+    (void) scale;
+    (void) conjB;
+    (void) stream;
+    throw_no_cuda();
+#else
+    (void)flags;
+
+    typedef void (*Caller)(const PtrStep<cufftComplex>, const PtrStep<cufftComplex>, float scale, PtrStepSz<cufftComplex>, cudaStream_t stream);
+    static Caller callers[] = { cudev::mulAndScaleSpectrums, cudev::mulAndScaleSpectrums_CONJ };
+
+    GpuMat src1 = _src1.getGpuMat();
+    GpuMat src2 = _src2.getGpuMat();
+
+    CV_Assert( src1.type() == src2.type() && src1.type() == CV_32FC2);
+    CV_Assert( src1.size() == src2.size() );
+
+    _dst.create(src1.size(), CV_32FC2);
+    GpuMat dst = _dst.getGpuMat();
+
+    Caller caller = callers[(int)conjB];
+    caller(src1, src2, scale, dst, StreamAccessor::getStream(stream));
+#endif
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// dft
+
+void cv::gpu::dft(InputArray _src, OutputArray _dst, Size dft_size, int flags, Stream& stream)
+{
+#ifndef HAVE_CUFFT
+    (void) _src;
+    (void) _dst;
+    (void) dft_size;
+    (void) flags;
+    (void) stream;
+    throw_no_cuda();
+#else
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_32FC1 || src.type() == CV_32FC2 );
+
+    // We don't support unpacked output (in the case of real input)
+    CV_Assert( !(flags & DFT_COMPLEX_OUTPUT) );
+
+    const bool is_1d_input       = (dft_size.height == 1) || (dft_size.width == 1);
+    const bool is_row_dft        = (flags & DFT_ROWS) != 0;
+    const bool is_scaled_dft     = (flags & DFT_SCALE) != 0;
+    const bool is_inverse        = (flags & DFT_INVERSE) != 0;
+    const bool is_complex_input  = src.channels() == 2;
+    const bool is_complex_output = !(flags & DFT_REAL_OUTPUT);
+
+    // We don't support real-to-real transform
+    CV_Assert( is_complex_input || is_complex_output );
+
+    GpuMat src_cont = src;
+
+    // Make sure here we work with the continuous input,
+    // as CUFFT can't handle gaps
+    createContinuous(src.rows, src.cols, src.type(), src_cont);
+    if (src_cont.data != src.data)
+        src.copyTo(src_cont, stream);
+
+    Size dft_size_opt = dft_size;
+    if (is_1d_input && !is_row_dft)
+    {
+        // If the source matrix is single column handle it as single row
+        dft_size_opt.width = std::max(dft_size.width, dft_size.height);
+        dft_size_opt.height = std::min(dft_size.width, dft_size.height);
+    }
+
+    CV_Assert( dft_size_opt.width > 1 );
+
+    cufftType dft_type = CUFFT_R2C;
+    if (is_complex_input)
+        dft_type = is_complex_output ? CUFFT_C2C : CUFFT_C2R;
+
+    cufftHandle plan;
+    if (is_1d_input || is_row_dft)
+        cufftSafeCall( cufftPlan1d(&plan, dft_size_opt.width, dft_type, dft_size_opt.height) );
+    else
+        cufftSafeCall( cufftPlan2d(&plan, dft_size_opt.height, dft_size_opt.width, dft_type) );
+
+    cufftSafeCall( cufftSetStream(plan, StreamAccessor::getStream(stream)) );
+
+    if (is_complex_input)
+    {
+        if (is_complex_output)
+        {
+            createContinuous(dft_size, CV_32FC2, _dst);
+            GpuMat dst = _dst.getGpuMat();
+
+            cufftSafeCall(cufftExecC2C(
+                    plan, src_cont.ptr<cufftComplex>(), dst.ptr<cufftComplex>(),
+                    is_inverse ? CUFFT_INVERSE : CUFFT_FORWARD));
+        }
+        else
+        {
+            createContinuous(dft_size, CV_32F, _dst);
+            GpuMat dst = _dst.getGpuMat();
+
+            cufftSafeCall(cufftExecC2R(
+                    plan, src_cont.ptr<cufftComplex>(), dst.ptr<cufftReal>()));
+        }
+    }
+    else
+    {
+        // We could swap dft_size for efficiency. Here we must reflect it
+        if (dft_size == dft_size_opt)
+            createContinuous(Size(dft_size.width / 2 + 1, dft_size.height), CV_32FC2, _dst);
+        else
+            createContinuous(Size(dft_size.width, dft_size.height / 2 + 1), CV_32FC2, _dst);
+
+        GpuMat dst = _dst.getGpuMat();
+
+        cufftSafeCall(cufftExecR2C(
+                plan, src_cont.ptr<cufftReal>(), dst.ptr<cufftComplex>()));
+    }
+
+    cufftSafeCall( cufftDestroy(plan) );
+
+    if (is_scaled_dft)
+        gpu::multiply(_dst, Scalar::all(1. / dft_size.area()), _dst, 1, -1, stream);
+
+#endif
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// Convolution
+
+#ifdef HAVE_CUFFT
+
+namespace
+{
+    class ConvolutionImpl : public Convolution
+    {
+    public:
+        explicit ConvolutionImpl(Size user_block_size_) : user_block_size(user_block_size_) {}
+
+        void convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null());
+
+    private:
+        void create(Size image_size, Size templ_size);
+        static Size estimateBlockSize(Size result_size);
+
+        Size result_size;
+        Size block_size;
+        Size user_block_size;
+        Size dft_size;
+        int spect_len;
+
+        GpuMat image_spect, templ_spect, result_spect;
+        GpuMat image_block, templ_block, result_data;
+    };
+
+    void ConvolutionImpl::create(Size image_size, Size templ_size)
+    {
+        result_size = Size(image_size.width - templ_size.width + 1,
+                           image_size.height - templ_size.height + 1);
+
+        block_size = user_block_size;
+        if (user_block_size.width == 0 || user_block_size.height == 0)
+            block_size = estimateBlockSize(result_size);
+
+        dft_size.width = 1 << int(ceil(std::log(block_size.width + templ_size.width - 1.) / std::log(2.)));
+        dft_size.height = 1 << int(ceil(std::log(block_size.height + templ_size.height - 1.) / std::log(2.)));
+
+        // CUFFT has hard-coded kernels for power-of-2 sizes (up to 8192),
+        // see CUDA Toolkit 4.1 CUFFT Library Programming Guide
+        if (dft_size.width > 8192)
+            dft_size.width = getOptimalDFTSize(block_size.width + templ_size.width - 1);
+        if (dft_size.height > 8192)
+            dft_size.height = getOptimalDFTSize(block_size.height + templ_size.height - 1);
+
+        // To avoid wasting time doing small DFTs
+        dft_size.width = std::max(dft_size.width, 512);
+        dft_size.height = std::max(dft_size.height, 512);
+
+        createContinuous(dft_size, CV_32F, image_block);
+        createContinuous(dft_size, CV_32F, templ_block);
+        createContinuous(dft_size, CV_32F, result_data);
+
+        spect_len = dft_size.height * (dft_size.width / 2 + 1);
+        createContinuous(1, spect_len, CV_32FC2, image_spect);
+        createContinuous(1, spect_len, CV_32FC2, templ_spect);
+        createContinuous(1, spect_len, CV_32FC2, result_spect);
+
+        // Use maximum result matrix block size for the estimated DFT block size
+        block_size.width = std::min(dft_size.width - templ_size.width + 1, result_size.width);
+        block_size.height = std::min(dft_size.height - templ_size.height + 1, result_size.height);
+    }
+
+    Size ConvolutionImpl::estimateBlockSize(Size result_size)
+    {
+        int width = (result_size.width + 2) / 3;
+        int height = (result_size.height + 2) / 3;
+        width = std::min(width, result_size.width);
+        height = std::min(height, result_size.height);
+        return Size(width, height);
+    }
+
+    void ConvolutionImpl::convolve(InputArray _image, InputArray _templ, OutputArray _result, bool ccorr, Stream& _stream)
+    {
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.type() == CV_32FC1 );
+        CV_Assert( templ.type() == CV_32FC1 );
+
+        create(image.size(), templ.size());
+
+        _result.create(result_size, CV_32FC1);
+        GpuMat result = _result.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        cufftHandle planR2C, planC2R;
+        cufftSafeCall( cufftPlan2d(&planC2R, dft_size.height, dft_size.width, CUFFT_C2R) );
+        cufftSafeCall( cufftPlan2d(&planR2C, dft_size.height, dft_size.width, CUFFT_R2C) );
+
+        cufftSafeCall( cufftSetStream(planR2C, stream) );
+        cufftSafeCall( cufftSetStream(planC2R, stream) );
+
+        GpuMat templ_roi(templ.size(), CV_32FC1, templ.data, templ.step);
+        gpu::copyMakeBorder(templ_roi, templ_block, 0, templ_block.rows - templ_roi.rows, 0,
+                            templ_block.cols - templ_roi.cols, 0, Scalar(), _stream);
+
+        cufftSafeCall( cufftExecR2C(planR2C, templ_block.ptr<cufftReal>(), templ_spect.ptr<cufftComplex>()) );
+
+        // Process all blocks of the result matrix
+        for (int y = 0; y < result.rows; y += block_size.height)
+        {
+            for (int x = 0; x < result.cols; x += block_size.width)
+            {
+                Size image_roi_size(std::min(x + dft_size.width, image.cols) - x,
+                                    std::min(y + dft_size.height, image.rows) - y);
+                GpuMat image_roi(image_roi_size, CV_32F, (void*)(image.ptr<float>(y) + x),
+                                 image.step);
+                gpu::copyMakeBorder(image_roi, image_block, 0, image_block.rows - image_roi.rows,
+                                    0, image_block.cols - image_roi.cols, 0, Scalar(), _stream);
+
+                cufftSafeCall(cufftExecR2C(planR2C, image_block.ptr<cufftReal>(),
+                                           image_spect.ptr<cufftComplex>()));
+                gpu::mulAndScaleSpectrums(image_spect, templ_spect, result_spect, 0,
+                                          1.f / dft_size.area(), ccorr, _stream);
+                cufftSafeCall(cufftExecC2R(planC2R, result_spect.ptr<cufftComplex>(),
+                                           result_data.ptr<cufftReal>()));
+
+                Size result_roi_size(std::min(x + block_size.width, result.cols) - x,
+                                     std::min(y + block_size.height, result.rows) - y);
+                GpuMat result_roi(result_roi_size, result.type(),
+                                  (void*)(result.ptr<float>(y) + x), result.step);
+                GpuMat result_block(result_roi_size, result_data.type(),
+                                    result_data.ptr(), result_data.step);
+
+                result_block.copyTo(result_roi, _stream);
+            }
+        }
+
+        cufftSafeCall( cufftDestroy(planR2C) );
+        cufftSafeCall( cufftDestroy(planC2R) );
+    }
+}
+
+#endif
+
+Ptr<Convolution> cv::gpu::createConvolution(Size user_block_size)
+{
+#ifndef HAVE_CUFFT
+    (void) user_block_size;
+    CV_Error(Error::StsNotImplemented, "The library was build without CUFFT");
+    return Ptr<Convolution>();
+#else
+    return makePtr<ConvolutionImpl>(user_block_size);
+#endif
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuarithm/src/core.cpp b/modules/gpuarithm/src/core.cpp
new file mode 100644
index 0000000000..05ea5741d8
--- /dev/null
+++ b/modules/gpuarithm/src/core.cpp
@@ -0,0 +1,612 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::merge(const GpuMat*, size_t, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::merge(const std::vector<GpuMat>&, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::split(InputArray, GpuMat*, Stream&) { throw_no_cuda(); }
+void cv::gpu::split(InputArray, std::vector<GpuMat>&, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::transpose(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::flip(InputArray, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+Ptr<LookUpTable> cv::gpu::createLookUpTable(InputArray) { throw_no_cuda(); return Ptr<LookUpTable>(); }
+
+void cv::gpu::copyMakeBorder(InputArray, OutputArray, int, int, int, int, int, Scalar, Stream&) { throw_no_cuda(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+////////////////////////////////////////////////////////////////////////
+// merge/split
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace split_merge
+    {
+        void merge(const PtrStepSzb* src, PtrStepSzb& dst, int total_channels, size_t elem_size, const cudaStream_t& stream);
+        void split(const PtrStepSzb& src, PtrStepSzb* dst, int num_channels, size_t elem_size1, const cudaStream_t& stream);
+    }
+}}}
+
+namespace
+{
+    void merge_caller(const GpuMat* src, size_t n, OutputArray _dst, Stream& stream)
+    {
+        CV_Assert( src != 0 );
+        CV_Assert( n > 0 && n <= 4 );
+
+        const int depth = src[0].depth();
+        const Size size = src[0].size();
+
+        for (size_t i = 0; i < n; ++i)
+        {
+            CV_Assert( src[i].size() == size );
+            CV_Assert( src[i].depth() == depth );
+            CV_Assert( src[i].channels() == 1 );
+        }
+
+        if (depth == CV_64F)
+        {
+            if (!deviceSupports(NATIVE_DOUBLE))
+                CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+        }
+
+        if (n == 1)
+        {
+            src[0].copyTo(_dst, stream);
+        }
+        else
+        {
+            _dst.create(size, CV_MAKE_TYPE(depth, (int)n));
+            GpuMat dst = _dst.getGpuMat();
+
+            PtrStepSzb src_as_devmem[4];
+            for(size_t i = 0; i < n; ++i)
+                src_as_devmem[i] = src[i];
+
+            PtrStepSzb dst_as_devmem(dst);
+            cv::gpu::cudev::split_merge::merge(src_as_devmem, dst_as_devmem, (int)n, CV_ELEM_SIZE(depth), StreamAccessor::getStream(stream));
+        }
+    }
+
+    void split_caller(const GpuMat& src, GpuMat* dst, Stream& stream)
+    {
+        CV_Assert( dst != 0 );
+
+        const int depth = src.depth();
+        const int num_channels = src.channels();
+
+        CV_Assert( num_channels <= 4 );
+
+        if (depth == CV_64F)
+        {
+            if (!deviceSupports(NATIVE_DOUBLE))
+                CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+        }
+
+        if (num_channels == 1)
+        {
+            src.copyTo(dst[0], stream);
+            return;
+        }
+
+        for (int i = 0; i < num_channels; ++i)
+            dst[i].create(src.size(), depth);
+
+        PtrStepSzb dst_as_devmem[4];
+        for (int i = 0; i < num_channels; ++i)
+            dst_as_devmem[i] = dst[i];
+
+        PtrStepSzb src_as_devmem(src);
+        cv::gpu::cudev::split_merge::split(src_as_devmem, dst_as_devmem, num_channels, src.elemSize1(), StreamAccessor::getStream(stream));
+    }
+}
+
+void cv::gpu::merge(const GpuMat* src, size_t n, OutputArray dst, Stream& stream)
+{
+    merge_caller(src, n, dst, stream);
+}
+
+
+void cv::gpu::merge(const std::vector<GpuMat>& src, OutputArray dst, Stream& stream)
+{
+    merge_caller(&src[0], src.size(), dst, stream);
+}
+
+void cv::gpu::split(InputArray _src, GpuMat* dst, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+    split_caller(src, dst, stream);
+}
+
+void cv::gpu::split(InputArray _src, std::vector<GpuMat>& dst, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+    dst.resize(src.channels());
+    if(src.channels() > 0)
+        split_caller(src, &dst[0], stream);
+}
+
+////////////////////////////////////////////////////////////////////////
+// transpose
+
+namespace arithm
+{
+    template <typename T> void transpose(PtrStepSz<T> src, PtrStepSz<T> dst, cudaStream_t stream);
+}
+
+void cv::gpu::transpose(InputArray _src, OutputArray _dst, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.elemSize() == 1 || src.elemSize() == 4 || src.elemSize() == 8 );
+
+    _dst.create( src.cols, src.rows, src.type() );
+    GpuMat dst = _dst.getGpuMat();
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    if (src.elemSize() == 1)
+    {
+        NppStreamHandler h(stream);
+
+        NppiSize sz;
+        sz.width  = src.cols;
+        sz.height = src.rows;
+
+        nppSafeCall( nppiTranspose_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
+            dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+    else if (src.elemSize() == 4)
+    {
+        arithm::transpose<int>(src, dst, stream);
+    }
+    else // if (src.elemSize() == 8)
+    {
+        if (!deviceSupports(NATIVE_DOUBLE))
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+
+        arithm::transpose<double>(src, dst, stream);
+    }
+}
+
+////////////////////////////////////////////////////////////////////////
+// flip
+
+namespace
+{
+    template<int DEPTH> struct NppTypeTraits;
+    template<> struct NppTypeTraits<CV_8U>  { typedef Npp8u npp_t; };
+    template<> struct NppTypeTraits<CV_8S>  { typedef Npp8s npp_t; };
+    template<> struct NppTypeTraits<CV_16U> { typedef Npp16u npp_t; };
+    template<> struct NppTypeTraits<CV_16S> { typedef Npp16s npp_t; };
+    template<> struct NppTypeTraits<CV_32S> { typedef Npp32s npp_t; };
+    template<> struct NppTypeTraits<CV_32F> { typedef Npp32f npp_t; };
+    template<> struct NppTypeTraits<CV_64F> { typedef Npp64f npp_t; };
+
+    template <int DEPTH> struct NppMirrorFunc
+    {
+        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
+
+        typedef NppStatus (*func_t)(const npp_t* pSrc, int nSrcStep, npp_t* pDst, int nDstStep, NppiSize oROI, NppiAxis flip);
+    };
+
+    template <int DEPTH, typename NppMirrorFunc<DEPTH>::func_t func> struct NppMirror
+    {
+        typedef typename NppMirrorFunc<DEPTH>::npp_t npp_t;
+
+        static void call(const GpuMat& src, GpuMat& dst, int flipCode, cudaStream_t stream)
+        {
+            NppStreamHandler h(stream);
+
+            NppiSize sz;
+            sz.width  = src.cols;
+            sz.height = src.rows;
+
+            nppSafeCall( func(src.ptr<npp_t>(), static_cast<int>(src.step),
+                dst.ptr<npp_t>(), static_cast<int>(dst.step), sz,
+                (flipCode == 0 ? NPP_HORIZONTAL_AXIS : (flipCode > 0 ? NPP_VERTICAL_AXIS : NPP_BOTH_AXIS))) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+}
+
+void cv::gpu::flip(InputArray _src, OutputArray _dst, int flipCode, Stream& stream)
+{
+    typedef void (*func_t)(const GpuMat& src, GpuMat& dst, int flipCode, cudaStream_t stream);
+    static const func_t funcs[6][4] =
+    {
+        {NppMirror<CV_8U, nppiMirror_8u_C1R>::call, 0, NppMirror<CV_8U, nppiMirror_8u_C3R>::call, NppMirror<CV_8U, nppiMirror_8u_C4R>::call},
+        {0,0,0,0},
+        {NppMirror<CV_16U, nppiMirror_16u_C1R>::call, 0, NppMirror<CV_16U, nppiMirror_16u_C3R>::call, NppMirror<CV_16U, nppiMirror_16u_C4R>::call},
+        {0,0,0,0},
+        {NppMirror<CV_32S, nppiMirror_32s_C1R>::call, 0, NppMirror<CV_32S, nppiMirror_32s_C3R>::call, NppMirror<CV_32S, nppiMirror_32s_C4R>::call},
+        {NppMirror<CV_32F, nppiMirror_32f_C1R>::call, 0, NppMirror<CV_32F, nppiMirror_32f_C3R>::call, NppMirror<CV_32F, nppiMirror_32f_C4R>::call}
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S || src.depth() == CV_32F);
+    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    funcs[src.depth()][src.channels() - 1](src, dst, flipCode, StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// LUT
+
+#if (CUDA_VERSION >= 5000)
+
+namespace
+{
+    class LookUpTableImpl : public LookUpTable
+    {
+    public:
+        LookUpTableImpl(InputArray lut);
+
+        void transform(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        int lut_cn;
+
+        int nValues3[3];
+        const Npp32s* pValues3[3];
+        const Npp32s* pLevels3[3];
+
+        GpuMat d_pLevels;
+        GpuMat d_nppLut;
+        GpuMat d_nppLut3[3];
+    };
+
+    LookUpTableImpl::LookUpTableImpl(InputArray _lut)
+    {
+        nValues3[0] = nValues3[1] = nValues3[2] = 256;
+
+        Npp32s pLevels[256];
+        for (int i = 0; i < 256; ++i)
+            pLevels[i] = i;
+
+        d_pLevels.upload(Mat(1, 256, CV_32S, pLevels));
+        pLevels3[0] = pLevels3[1] = pLevels3[2] = d_pLevels.ptr<Npp32s>();
+
+        GpuMat lut;
+        if (_lut.kind() == _InputArray::GPU_MAT)
+        {
+            lut = _lut.getGpuMat();
+        }
+        else
+        {
+            Mat hLut = _lut.getMat();
+            CV_Assert( hLut.total() == 256 && hLut.isContinuous() );
+            lut.upload(Mat(1, 256, hLut.type(), hLut.data));
+        }
+
+        lut_cn = lut.channels();
+
+        CV_Assert( lut.depth() == CV_8U );
+        CV_Assert( lut.rows == 1 && lut.cols == 256 );
+
+        lut.convertTo(d_nppLut, CV_32S);
+
+        if (lut_cn == 1)
+        {
+            pValues3[0] = pValues3[1] = pValues3[2] = d_nppLut.ptr<Npp32s>();
+        }
+        else
+        {
+            gpu::split(d_nppLut, d_nppLut3);
+
+            pValues3[0] = d_nppLut3[0].ptr<Npp32s>();
+            pValues3[1] = d_nppLut3[1].ptr<Npp32s>();
+            pValues3[2] = d_nppLut3[2].ptr<Npp32s>();
+        }
+    }
+
+    void LookUpTableImpl::transform(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        const int cn = src.channels();
+
+        CV_Assert( src.type() == CV_8UC1 || src.type() == CV_8UC3 );
+        CV_Assert( lut_cn == 1 || lut_cn == cn );
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        NppStreamHandler h(stream);
+
+        NppiSize sz;
+        sz.height = src.rows;
+        sz.width = src.cols;
+
+        if (src.type() == CV_8UC1)
+        {
+            nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
+                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, d_nppLut.ptr<Npp32s>(), d_pLevels.ptr<Npp32s>(), 256) );
+        }
+        else
+        {
+            nppSafeCall( nppiLUT_Linear_8u_C3R(src.ptr<Npp8u>(), static_cast<int>(src.step),
+                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, pValues3, pLevels3, nValues3) );
+        }
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+#else //  (CUDA_VERSION >= 5000)
+
+namespace
+{
+    class LookUpTableImpl : public LookUpTable
+    {
+    public:
+        LookUpTableImpl(InputArray lut);
+
+        void transform(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        int lut_cn;
+
+        Npp32s pLevels[256];
+        int nValues3[3];
+        const Npp32s* pValues3[3];
+        const Npp32s* pLevels3[3];
+
+        Mat nppLut;
+        Mat nppLut3[3];
+    };
+
+    LookUpTableImpl::LookUpTableImpl(InputArray _lut)
+    {
+        nValues3[0] = nValues3[1] = nValues3[2] = 256;
+
+        for (int i = 0; i < 256; ++i)
+            pLevels[i] = i;
+        pLevels3[0] = pLevels3[1] = pLevels3[2] = pLevels;
+
+        Mat lut;
+        if (_lut.kind() == _InputArray::GPU_MAT)
+        {
+            lut = Mat(_lut.getGpuMat());
+        }
+        else
+        {
+            Mat hLut = _lut.getMat();
+            CV_Assert( hLut.total() == 256 && hLut.isContinuous() );
+            lut = hLut;
+        }
+
+        lut_cn = lut.channels();
+
+        CV_Assert( lut.depth() == CV_8U );
+        CV_Assert( lut.rows == 1 && lut.cols == 256 );
+
+        lut.convertTo(nppLut, CV_32S);
+
+        if (lut_cn == 1)
+        {
+            pValues3[0] = pValues3[1] = pValues3[2] = nppLut.ptr<Npp32s>();
+        }
+        else
+        {
+            cv::split(nppLut, nppLut3);
+
+            pValues3[0] = nppLut3[0].ptr<Npp32s>();
+            pValues3[1] = nppLut3[1].ptr<Npp32s>();
+            pValues3[2] = nppLut3[2].ptr<Npp32s>();
+        }
+    }
+
+    void LookUpTableImpl::transform(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        const int cn = src.channels();
+
+        CV_Assert( src.type() == CV_8UC1 || src.type() == CV_8UC3 );
+        CV_Assert( lut_cn == 1 || lut_cn == cn );
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        NppStreamHandler h(stream);
+
+        NppiSize sz;
+        sz.height = src.rows;
+        sz.width = src.cols;
+
+        if (src.type() == CV_8UC1)
+        {
+            nppSafeCall( nppiLUT_Linear_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step),
+                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, nppLut.ptr<Npp32s>(), pLevels, 256) );
+        }
+        else
+        {
+            nppSafeCall( nppiLUT_Linear_8u_C3R(src.ptr<Npp8u>(), static_cast<int>(src.step),
+                dst.ptr<Npp8u>(), static_cast<int>(dst.step), sz, pValues3, pLevels3, nValues3) );
+        }
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+#endif //  (CUDA_VERSION >= 5000)
+
+Ptr<LookUpTable> cv::gpu::createLookUpTable(InputArray lut)
+{
+    return makePtr<LookUpTableImpl>(lut);
+}
+
+////////////////////////////////////////////////////////////////////////
+// copyMakeBorder
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        template <typename T, int cn> void copyMakeBorder_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderMode, const T* borderValue, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    template <typename T, int cn> void copyMakeBorder_caller(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderType, const Scalar& value, cudaStream_t stream)
+    {
+        using namespace ::cv::gpu::cudev::imgproc;
+
+        Scalar_<T> val(saturate_cast<T>(value[0]), saturate_cast<T>(value[1]), saturate_cast<T>(value[2]), saturate_cast<T>(value[3]));
+
+        copyMakeBorder_gpu<T, cn>(src, dst, top, left, borderType, val.val, stream);
+    }
+}
+
+#if defined __GNUC__ && __GNUC__ > 2 && __GNUC_MINOR__  > 4
+typedef Npp32s __attribute__((__may_alias__)) Npp32s_a;
+#else
+typedef Npp32s Npp32s_a;
+#endif
+
+void cv::gpu::copyMakeBorder(InputArray _src, OutputArray _dst, int top, int bottom, int left, int right, int borderType, Scalar value, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( borderType == BORDER_REFLECT_101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP );
+
+    _dst.create(src.rows + top + bottom, src.cols + left + right, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    if (borderType == BORDER_CONSTANT && (src.type() == CV_8UC1 || src.type() == CV_8UC4 || src.type() == CV_32SC1 || src.type() == CV_32FC1))
+    {
+        NppiSize srcsz;
+        srcsz.width  = src.cols;
+        srcsz.height = src.rows;
+
+        NppiSize dstsz;
+        dstsz.width  = dst.cols;
+        dstsz.height = dst.rows;
+
+        NppStreamHandler h(stream);
+
+        switch (src.type())
+        {
+        case CV_8UC1:
+            {
+                Npp8u nVal = saturate_cast<Npp8u>(value[0]);
+                nppSafeCall( nppiCopyConstBorder_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
+                    dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
+                break;
+            }
+        case CV_8UC4:
+            {
+                Npp8u nVal[] = {saturate_cast<Npp8u>(value[0]), saturate_cast<Npp8u>(value[1]), saturate_cast<Npp8u>(value[2]), saturate_cast<Npp8u>(value[3])};
+                nppSafeCall( nppiCopyConstBorder_8u_C4R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
+                    dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
+                break;
+            }
+        case CV_32SC1:
+            {
+                Npp32s nVal = saturate_cast<Npp32s>(value[0]);
+                nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
+                    dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
+                break;
+            }
+        case CV_32FC1:
+            {
+                Npp32f val = saturate_cast<Npp32f>(value[0]);
+                Npp32s nVal = *(reinterpret_cast<Npp32s_a*>(&val));
+                nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
+                    dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
+                break;
+            }
+        }
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+    else
+    {
+        typedef void (*caller_t)(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderType, const Scalar& value, cudaStream_t stream);
+        static const caller_t callers[6][4] =
+        {
+            {   copyMakeBorder_caller<uchar, 1>  ,    copyMakeBorder_caller<uchar, 2>   ,    copyMakeBorder_caller<uchar, 3>  ,    copyMakeBorder_caller<uchar, 4>},
+            {0/*copyMakeBorder_caller<schar, 1>*/, 0/*copyMakeBorder_caller<schar, 2>*/ , 0/*copyMakeBorder_caller<schar, 3>*/, 0/*copyMakeBorder_caller<schar, 4>*/},
+            {   copyMakeBorder_caller<ushort, 1> , 0/*copyMakeBorder_caller<ushort, 2>*/,    copyMakeBorder_caller<ushort, 3> ,    copyMakeBorder_caller<ushort, 4>},
+            {   copyMakeBorder_caller<short, 1>  , 0/*copyMakeBorder_caller<short, 2>*/ ,    copyMakeBorder_caller<short, 3>  ,    copyMakeBorder_caller<short, 4>},
+            {0/*copyMakeBorder_caller<int,   1>*/, 0/*copyMakeBorder_caller<int,   2>*/ , 0/*copyMakeBorder_caller<int,   3>*/, 0/*copyMakeBorder_caller<int  , 4>*/},
+            {   copyMakeBorder_caller<float, 1>  , 0/*copyMakeBorder_caller<float, 2>*/ ,    copyMakeBorder_caller<float, 3>  ,    copyMakeBorder_caller<float ,4>}
+        };
+
+        caller_t func = callers[src.depth()][src.channels() - 1];
+        CV_Assert(func != 0);
+
+        func(src, dst, top, left, borderType, value, stream);
+    }
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuarithm/src/cuda/absdiff_mat.cu b/modules/gpuarithm/src/cuda/absdiff_mat.cu
new file mode 100644
index 0000000000..d47068ee03
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/absdiff_mat.cu
@@ -0,0 +1,147 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct VAbsDiff4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vabsdiff4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VAbsDiff4() {}
+        __host__ __device__ __forceinline__ VAbsDiff4(const VAbsDiff4&) {}
+    };
+
+    struct VAbsDiff2 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vabsdiff2(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VAbsDiff2() {}
+        __host__ __device__ __forceinline__ VAbsDiff2(const VAbsDiff2&) {}
+    };
+
+    __device__ __forceinline__ int _abs(int a)
+    {
+        return ::abs(a);
+    }
+    __device__ __forceinline__ float _abs(float a)
+    {
+        return ::fabsf(a);
+    }
+    __device__ __forceinline__ double _abs(double a)
+    {
+        return ::fabs(a);
+    }
+
+    template <typename T> struct AbsDiffMat : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(T a, T b) const
+        {
+            return saturate_cast<T>(_abs(a - b));
+        }
+
+        __host__ __device__ __forceinline__ AbsDiffMat() {}
+        __host__ __device__ __forceinline__ AbsDiffMat(const AbsDiffMat&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VAbsDiff4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <> struct TransformFunctorTraits< arithm::VAbsDiff2 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< arithm::AbsDiffMat<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void absDiffMat_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VAbsDiff4(), WithOutMask(), stream);
+    }
+
+    void absDiffMat_v2(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VAbsDiff2(), WithOutMask(), stream);
+    }
+
+    template <typename T>
+    void absDiffMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, AbsDiffMat<T>(), WithOutMask(), stream);
+    }
+
+    template void absDiffMat<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/absdiff_scalar.cu b/modules/gpuarithm/src/cuda/absdiff_scalar.cu
new file mode 100644
index 0000000000..5a89791f86
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/absdiff_scalar.cu
@@ -0,0 +1,98 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T, typename S> struct AbsDiffScalar : unary_function<T, T>
+    {
+        S val;
+
+        __host__ explicit AbsDiffScalar(S val_) : val(val_) {}
+
+        __device__ __forceinline__ T operator ()(T a) const
+        {
+            abs_func<S> f;
+            return saturate_cast<T>(f(a - val));
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename S> struct TransformFunctorTraits< arithm::AbsDiffScalar<T, S> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T, typename S>
+    void absDiffScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
+    {
+        AbsDiffScalar<T, S> op(static_cast<S>(val));
+
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, op, WithOutMask(), stream);
+    }
+
+    template void absDiffScalar<uchar, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<schar, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<ushort, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<short, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<int, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<float, float>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void absDiffScalar<double, double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/add_mat.cu b/modules/gpuarithm/src/cuda/add_mat.cu
new file mode 100644
index 0000000000..3b1bc1f385
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/add_mat.cu
@@ -0,0 +1,185 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct VAdd4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vadd4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VAdd4() {}
+        __host__ __device__ __forceinline__ VAdd4(const VAdd4&) {}
+    };
+
+    struct VAdd2 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vadd2(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VAdd2() {}
+        __host__ __device__ __forceinline__ VAdd2(const VAdd2&) {}
+    };
+
+    template <typename T, typename D> struct AddMat : binary_function<T, T, D>
+    {
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return saturate_cast<D>(a + b);
+        }
+
+        __host__ __device__ __forceinline__ AddMat() {}
+        __host__ __device__ __forceinline__ AddMat(const AddMat&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VAdd4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <> struct TransformFunctorTraits< arithm::VAdd2 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T, typename D> struct TransformFunctorTraits< arithm::AddMat<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void addMat_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VAdd4(), WithOutMask(), stream);
+    }
+
+    void addMat_v2(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VAdd2(), WithOutMask(), stream);
+    }
+
+    template <typename T, typename D>
+    void addMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, AddMat<T, D>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, AddMat<T, D>(), WithOutMask(), stream);
+    }
+
+    template void addMat<uchar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<uchar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void addMat<schar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<schar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addMat<ushort, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<ushort, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<ushort, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<ushort, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<ushort, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<ushort, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<ushort, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addMat<short, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<short, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<short, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<short, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<short, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<short, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<short, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addMat<int, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<int, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<int, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<int, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<int, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<int, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addMat<float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<float, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addMat<double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addMat<double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addMat<double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/add_scalar.cu b/modules/gpuarithm/src/cuda/add_scalar.cu
new file mode 100644
index 0000000000..3362c2b930
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/add_scalar.cu
@@ -0,0 +1,148 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T, typename S, typename D> struct AddScalar : unary_function<T, D>
+    {
+        S val;
+
+        __host__ explicit AddScalar(S val_) : val(val_) {}
+
+        __device__ __forceinline__ D operator ()(T a) const
+        {
+            return saturate_cast<D>(a + val);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::AddScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T, typename S, typename D>
+    void addScalar(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        AddScalar<T, S, D> op(static_cast<S>(val));
+
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+    }
+
+    template void addScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void addScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void addScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void addScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void addScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/add_weighted.cu b/modules/gpuarithm/src/cuda/add_weighted.cu
new file mode 100644
index 0000000000..bf632d68f2
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/add_weighted.cu
@@ -0,0 +1,364 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T> struct UseDouble_
+    {
+        enum {value = 0};
+    };
+    template <> struct UseDouble_<double>
+    {
+        enum {value = 1};
+    };
+    template <typename T1, typename T2, typename D> struct UseDouble
+    {
+        enum {value = (UseDouble_<T1>::value || UseDouble_<T2>::value || UseDouble_<D>::value)};
+    };
+
+    template <typename T1, typename T2, typename D, bool useDouble> struct AddWeighted_;
+    template <typename T1, typename T2, typename D> struct AddWeighted_<T1, T2, D, false> : binary_function<T1, T2, D>
+    {
+        float alpha;
+        float beta;
+        float gamma;
+
+        __host__ AddWeighted_(double alpha_, double beta_, double gamma_) : alpha(static_cast<float>(alpha_)), beta(static_cast<float>(beta_)), gamma(static_cast<float>(gamma_)) {}
+
+        __device__ __forceinline__ D operator ()(T1 a, T2 b) const
+        {
+            return saturate_cast<D>(a * alpha + b * beta + gamma);
+        }
+    };
+    template <typename T1, typename T2, typename D> struct AddWeighted_<T1, T2, D, true> : binary_function<T1, T2, D>
+    {
+        double alpha;
+        double beta;
+        double gamma;
+
+        __host__ AddWeighted_(double alpha_, double beta_, double gamma_) : alpha(alpha_), beta(beta_), gamma(gamma_) {}
+
+        __device__ __forceinline__ D operator ()(T1 a, T2 b) const
+        {
+            return saturate_cast<D>(a * alpha + b * beta + gamma);
+        }
+    };
+    template <typename T1, typename T2, typename D> struct AddWeighted : AddWeighted_<T1, T2, D, UseDouble<T1, T2, D>::value>
+    {
+        AddWeighted(double alpha_, double beta_, double gamma_) : AddWeighted_<T1, T2, D, UseDouble<T1, T2, D>::value>(alpha_, beta_, gamma_) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T1, typename T2, typename D, size_t src1_size, size_t src2_size, size_t dst_size> struct AddWeightedTraits : DefaultTransformFunctorTraits< arithm::AddWeighted<T1, T2, D> >
+    {
+    };
+    template <typename T1, typename T2, typename D, size_t src_size, size_t dst_size> struct AddWeightedTraits<T1, T2, D, src_size, src_size, dst_size> : arithm::ArithmFuncTraits<src_size, dst_size>
+    {
+    };
+
+    template <typename T1, typename T2, typename D> struct TransformFunctorTraits< arithm::AddWeighted<T1, T2, D> > : AddWeightedTraits<T1, T2, D, sizeof(T1), sizeof(T2), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T1, typename T2, typename D>
+    void addWeighted(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream)
+    {
+        AddWeighted<T1, T2, D> op(alpha, beta, gamma);
+
+        cudev::transform((PtrStepSz<T1>) src1, (PtrStepSz<T2>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+    }
+
+    template void addWeighted<uchar, uchar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, uchar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, schar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, schar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<uchar, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<uchar, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<schar, schar, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, schar, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<schar, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<schar, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<schar, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<schar, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<schar, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<schar, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<ushort, ushort, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, ushort, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<ushort, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<ushort, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<ushort, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<ushort, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<ushort, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<short, short, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, short, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<short, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<short, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<short, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<short, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<int, int, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, int, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<int, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<int, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<int, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<float, float, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, float, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+    template void addWeighted<float, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<float, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+
+
+
+    template void addWeighted<double, double, uchar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, schar>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, ushort>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, short>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, int>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, float>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+    template void addWeighted<double, double, double>(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuarithm/src/cuda/arithm_func_traits.hpp b/modules/gpuarithm/src/cuda/arithm_func_traits.hpp
new file mode 100644
index 0000000000..051fd9ff08
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/arithm_func_traits.hpp
@@ -0,0 +1,145 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __ARITHM_FUNC_TRAITS_HPP__
+#define __ARITHM_FUNC_TRAITS_HPP__
+
+#include <cstddef>
+
+namespace arithm
+{
+    template <size_t src_size, size_t dst_size> struct ArithmFuncTraits
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 1 };
+    };
+
+    template <> struct ArithmFuncTraits<1, 1>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<1, 2>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<1, 4>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+
+    template <> struct ArithmFuncTraits<2, 1>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<2, 2>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<2, 4>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+
+    template <> struct ArithmFuncTraits<4, 1>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<4, 2>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+    template <> struct ArithmFuncTraits<4, 4>
+    {
+        enum { simple_block_dim_x = 32 };
+        enum { simple_block_dim_y = 8 };
+
+        enum { smart_block_dim_x = 32 };
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+}
+
+#endif // __ARITHM_FUNC_TRAITS_HPP__
diff --git a/modules/gpuarithm/src/cuda/bitwise_mat.cu b/modules/gpuarithm/src/cuda/bitwise_mat.cu
new file mode 100644
index 0000000000..40222214ea
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/bitwise_mat.cu
@@ -0,0 +1,126 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< bit_not<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< bit_and<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< bit_or<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< bit_xor<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T> void bitMatNot(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, bit_not<T>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, bit_not<T>(), WithOutMask(), stream);
+    }
+
+    template <typename T> void bitMatAnd(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_and<T>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_and<T>(), WithOutMask(), stream);
+    }
+
+    template <typename T> void bitMatOr(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_or<T>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_or<T>(), WithOutMask(), stream);
+    }
+
+    template <typename T> void bitMatXor(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_xor<T>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, bit_xor<T>(), WithOutMask(), stream);
+    }
+
+    template void bitMatNot<uchar>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatNot<ushort>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatNot<uint>(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void bitMatAnd<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatAnd<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatAnd<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void bitMatOr<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatOr<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatOr<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void bitMatXor<uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatXor<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void bitMatXor<uint>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/bitwise_scalar.cu b/modules/gpuarithm/src/cuda/bitwise_scalar.cu
new file mode 100644
index 0000000000..145885024a
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/bitwise_scalar.cu
@@ -0,0 +1,104 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< binder2nd< bit_and<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< binder2nd< bit_or<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< binder2nd< bit_xor<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T> void bitScalarAnd(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::cudev::bind2nd(bit_and<T>(), src2), WithOutMask(), stream);
+    }
+
+    template <typename T> void bitScalarOr(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::cudev::bind2nd(bit_or<T>(), src2), WithOutMask(), stream);
+    }
+
+    template <typename T> void bitScalarXor(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::cudev::bind2nd(bit_xor<T>(), src2), WithOutMask(), stream);
+    }
+
+    template void bitScalarAnd<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarAnd<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarAnd<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarAnd<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template void bitScalarOr<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarOr<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarOr<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarOr<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template void bitScalarXor<uchar>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarXor<ushort>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarXor<int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+    template void bitScalarXor<unsigned int>(PtrStepSzb src1, uint src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/cmp_mat.cu b/modules/gpuarithm/src/cuda/cmp_mat.cu
new file mode 100644
index 0000000000..6602edf62f
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/cmp_mat.cu
@@ -0,0 +1,206 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct VCmpEq4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vcmpeq4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VCmpEq4() {}
+        __host__ __device__ __forceinline__ VCmpEq4(const VCmpEq4&) {}
+    };
+    struct VCmpNe4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vcmpne4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VCmpNe4() {}
+        __host__ __device__ __forceinline__ VCmpNe4(const VCmpNe4&) {}
+    };
+    struct VCmpLt4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vcmplt4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VCmpLt4() {}
+        __host__ __device__ __forceinline__ VCmpLt4(const VCmpLt4&) {}
+    };
+    struct VCmpLe4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vcmple4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VCmpLe4() {}
+        __host__ __device__ __forceinline__ VCmpLe4(const VCmpLe4&) {}
+    };
+
+    template <class Op, typename T>
+    struct Cmp : binary_function<T, T, uchar>
+    {
+        __device__ __forceinline__ uchar operator()(T a, T b) const
+        {
+            Op op;
+            return -op(a, b);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VCmpEq4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+    template <> struct TransformFunctorTraits< arithm::VCmpNe4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+    template <> struct TransformFunctorTraits< arithm::VCmpLt4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+    template <> struct TransformFunctorTraits< arithm::VCmpLe4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <class Op, typename T> struct TransformFunctorTraits< arithm::Cmp<Op, T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(uchar)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void cmpMatEq_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VCmpEq4(), WithOutMask(), stream);
+    }
+    void cmpMatNe_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VCmpNe4(), WithOutMask(), stream);
+    }
+    void cmpMatLt_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VCmpLt4(), WithOutMask(), stream);
+    }
+    void cmpMatLe_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VCmpLe4(), WithOutMask(), stream);
+    }
+
+    template <template <typename> class Op, typename T>
+    void cmpMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        Cmp<Op<T>, T> op;
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, dst, op, WithOutMask(), stream);
+    }
+
+    template <typename T> void cmpMatEq(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cmpMat<equal_to, T>(src1, src2, dst, stream);
+    }
+    template <typename T> void cmpMatNe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cmpMat<not_equal_to, T>(src1, src2, dst, stream);
+    }
+    template <typename T> void cmpMatLt(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cmpMat<less, T>(src1, src2, dst, stream);
+    }
+    template <typename T> void cmpMatLe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cmpMat<less_equal, T>(src1, src2, dst, stream);
+    }
+
+    template void cmpMatEq<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatEq<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpMatNe<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatNe<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpMatLt<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLt<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpMatLe<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void cmpMatLe<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/cmp_scalar.cu b/modules/gpuarithm/src/cuda/cmp_scalar.cu
new file mode 100644
index 0000000000..678f253ea8
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/cmp_scalar.cu
@@ -0,0 +1,284 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <class Op, typename T>
+    struct Cmp : binary_function<T, T, uchar>
+    {
+        __device__ __forceinline__ uchar operator()(T a, T b) const
+        {
+            Op op;
+            return -op(a, b);
+        }
+    };
+
+#define TYPE_VEC(type, cn) typename TypeVec<type, cn>::vec_type
+
+    template <class Op, typename T, int cn> struct CmpScalar;
+    template <class Op, typename T>
+    struct CmpScalar<Op, T, 1> : unary_function<T, uchar>
+    {
+        T val;
+
+        __host__ explicit CmpScalar(T val_) : val(val_) {}
+
+        __device__ __forceinline__ uchar operator()(T src) const
+        {
+            Cmp<Op, T> op;
+            return op(src, val);
+        }
+    };
+    template <class Op, typename T>
+    struct CmpScalar<Op, T, 2> : unary_function<TYPE_VEC(T, 2), TYPE_VEC(uchar, 2)>
+    {
+        TYPE_VEC(T, 2) val;
+
+        __host__ explicit CmpScalar(TYPE_VEC(T, 2) val_) : val(val_) {}
+
+        __device__ __forceinline__ TYPE_VEC(uchar, 2) operator()(const TYPE_VEC(T, 2) & src) const
+        {
+            Cmp<Op, T> op;
+            return VecTraits<TYPE_VEC(uchar, 2)>::make(op(src.x, val.x), op(src.y, val.y));
+        }
+    };
+    template <class Op, typename T>
+    struct CmpScalar<Op, T, 3> : unary_function<TYPE_VEC(T, 3), TYPE_VEC(uchar, 3)>
+    {
+        TYPE_VEC(T, 3) val;
+
+        __host__ explicit CmpScalar(TYPE_VEC(T, 3) val_) : val(val_) {}
+
+        __device__ __forceinline__ TYPE_VEC(uchar, 3) operator()(const TYPE_VEC(T, 3) & src) const
+        {
+            Cmp<Op, T> op;
+            return VecTraits<TYPE_VEC(uchar, 3)>::make(op(src.x, val.x), op(src.y, val.y), op(src.z, val.z));
+        }
+    };
+    template <class Op, typename T>
+    struct CmpScalar<Op, T, 4> : unary_function<TYPE_VEC(T, 4), TYPE_VEC(uchar, 4)>
+    {
+        TYPE_VEC(T, 4) val;
+
+        __host__ explicit CmpScalar(TYPE_VEC(T, 4) val_) : val(val_) {}
+
+        __device__ __forceinline__ TYPE_VEC(uchar, 4) operator()(const TYPE_VEC(T, 4) & src) const
+        {
+            Cmp<Op, T> op;
+            return VecTraits<TYPE_VEC(uchar, 4)>::make(op(src.x, val.x), op(src.y, val.y), op(src.z, val.z), op(src.w, val.w));
+        }
+    };
+
+#undef TYPE_VEC
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <class Op, typename T> struct TransformFunctorTraits< arithm::CmpScalar<Op, T, 1> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(uchar)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <template <typename> class Op, typename T, int cn>
+    void cmpScalar(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef typename TypeVec<T, cn>::vec_type src_t;
+        typedef typename TypeVec<uchar, cn>::vec_type dst_t;
+
+        T sval[] = {static_cast<T>(val[0]), static_cast<T>(val[1]), static_cast<T>(val[2]), static_cast<T>(val[3])};
+        src_t val1 = VecTraits<src_t>::make(sval);
+
+        CmpScalar<Op<T>, T, cn> op(val1);
+        cudev::transform((PtrStepSz<src_t>) src, (PtrStepSz<dst_t>) dst, op, WithOutMask(), stream);
+    }
+
+    template <typename T> void cmpScalarEq(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<equal_to, T, 1>,
+            cmpScalar<equal_to, T, 2>,
+            cmpScalar<equal_to, T, 3>,
+            cmpScalar<equal_to, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+    template <typename T> void cmpScalarNe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<not_equal_to, T, 1>,
+            cmpScalar<not_equal_to, T, 2>,
+            cmpScalar<not_equal_to, T, 3>,
+            cmpScalar<not_equal_to, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+    template <typename T> void cmpScalarLt(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<less, T, 1>,
+            cmpScalar<less, T, 2>,
+            cmpScalar<less, T, 3>,
+            cmpScalar<less, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+    template <typename T> void cmpScalarLe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<less_equal, T, 1>,
+            cmpScalar<less_equal, T, 2>,
+            cmpScalar<less_equal, T, 3>,
+            cmpScalar<less_equal, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+    template <typename T> void cmpScalarGt(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<greater, T, 1>,
+            cmpScalar<greater, T, 2>,
+            cmpScalar<greater, T, 3>,
+            cmpScalar<greater, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+    template <typename T> void cmpScalarGe(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, double val[4], PtrStepSzb dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0,
+            cmpScalar<greater_equal, T, 1>,
+            cmpScalar<greater_equal, T, 2>,
+            cmpScalar<greater_equal, T, 3>,
+            cmpScalar<greater_equal, T, 4>
+        };
+
+        funcs[cn](src, val, dst, stream);
+    }
+
+    template void cmpScalarEq<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarEq<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpScalarNe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarNe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpScalarLt<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLt<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpScalarLe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarLe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpScalarGt<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGt<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+
+    template void cmpScalarGe<uchar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<schar >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<ushort>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<short >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<int   >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<float >(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+    template void cmpScalarGe<double>(PtrStepSzb src, int cn, double val[4], PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpu/src/cuda/copy_make_border.cu b/modules/gpuarithm/src/cuda/copy_make_border.cu
similarity index 97%
rename from modules/gpu/src/cuda/copy_make_border.cu
rename to modules/gpuarithm/src/cuda/copy_make_border.cu
index d9898d75ed..d772e09ede 100644
--- a/modules/gpu/src/cuda/copy_make_border.cu
+++ b/modules/gpuarithm/src/cuda/copy_make_border.cu
@@ -42,10 +42,10 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -86,11 +86,11 @@ namespace cv { namespace gpu { namespace device
 
             static const caller_t callers[5] =
             {
-                CopyMakeBorderDispatcher<BrdReflect101, vec_type>::call,
-                CopyMakeBorderDispatcher<BrdReplicate, vec_type>::call,
                 CopyMakeBorderDispatcher<BrdConstant, vec_type>::call,
+                CopyMakeBorderDispatcher<BrdReplicate, vec_type>::call,
                 CopyMakeBorderDispatcher<BrdReflect, vec_type>::call,
-                CopyMakeBorderDispatcher<BrdWrap, vec_type>::call
+                CopyMakeBorderDispatcher<BrdWrap, vec_type>::call,
+                CopyMakeBorderDispatcher<BrdReflect101, vec_type>::call
             };
 
             callers[borderMode](PtrStepSz<vec_type>(src), PtrStepSz<vec_type>(dst), top, left, borderValue, stream);
@@ -126,6 +126,6 @@ namespace cv { namespace gpu { namespace device
         template void copyMakeBorder_gpu<float, 3>(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderMode, const float* borderValue, cudaStream_t stream);
         template void copyMakeBorder_gpu<float, 4>(const PtrStepSzb& src, const PtrStepSzb& dst, int top, int left, int borderMode, const float* borderValue, cudaStream_t stream);
     } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuarithm/src/cuda/countnonzero.cu b/modules/gpuarithm/src/cuda/countnonzero.cu
new file mode 100644
index 0000000000..beab82b4b0
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/countnonzero.cu
@@ -0,0 +1,176 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace countNonZero
+{
+    __device__ unsigned int blocks_finished = 0;
+
+    template <int BLOCK_SIZE, typename T>
+    __global__ void kernel(const PtrStepSz<T> src, unsigned int* count, const int twidth, const int theight)
+    {
+        __shared__ unsigned int scount[BLOCK_SIZE];
+
+        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
+        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
+
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+
+        unsigned int mycount = 0;
+
+        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
+        {
+            const T* ptr = src.ptr(y);
+
+            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
+            {
+                const T srcVal = ptr[x];
+
+                mycount += (srcVal != 0);
+            }
+        }
+
+        cudev::reduce<BLOCK_SIZE>(scount, mycount, tid, plus<unsigned int>());
+
+    #if __CUDA_ARCH__ >= 200
+        if (tid == 0)
+            ::atomicAdd(count, mycount);
+    #else
+        __shared__ bool is_last;
+        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
+
+        if (tid == 0)
+        {
+            count[bid] = mycount;
+
+            __threadfence();
+
+            unsigned int ticket = ::atomicInc(&blocks_finished, gridDim.x * gridDim.y);
+            is_last = (ticket == gridDim.x * gridDim.y - 1);
+        }
+
+        __syncthreads();
+
+        if (is_last)
+        {
+            mycount = tid < gridDim.x * gridDim.y ? count[tid] : 0;
+
+            cudev::reduce<BLOCK_SIZE>(scount, mycount, tid, plus<unsigned int>());
+
+            if (tid == 0)
+            {
+                count[0] = mycount;
+
+                blocks_finished = 0;
+            }
+        }
+    #endif
+    }
+
+    const int threads_x = 32;
+    const int threads_y = 8;
+
+    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
+    {
+        block = dim3(threads_x, threads_y);
+
+        grid = dim3(divUp(cols, block.x * block.y),
+                    divUp(rows, block.y * block.x));
+
+        grid.x = ::min(grid.x, block.x);
+        grid.y = ::min(grid.y, block.y);
+    }
+
+    void getBufSize(int cols, int rows, int& bufcols, int& bufrows)
+    {
+        dim3 block, grid;
+        getLaunchCfg(cols, rows, block, grid);
+
+        bufcols = grid.x * grid.y * sizeof(int);
+        bufrows = 1;
+    }
+
+    template <typename T>
+    int run(const PtrStepSzb src, PtrStep<unsigned int> buf)
+    {
+        dim3 block, grid;
+        getLaunchCfg(src.cols, src.rows, block, grid);
+
+        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
+        const int theight = divUp(divUp(src.rows, grid.y), block.y);
+
+        unsigned int* count_buf = buf.ptr(0);
+
+        cudaSafeCall( cudaMemset(count_buf, 0, sizeof(unsigned int)) );
+
+        kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, count_buf, twidth, theight);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+
+        unsigned int count;
+        cudaSafeCall(cudaMemcpy(&count, count_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost));
+
+        return count;
+    }
+
+    template int run<uchar >(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<schar >(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<ushort>(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<short >(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<int   >(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<float >(const PtrStepSzb src, PtrStep<unsigned int> buf);
+    template int run<double>(const PtrStepSzb src, PtrStep<unsigned int> buf);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/div_mat.cu b/modules/gpuarithm/src/cuda/div_mat.cu
new file mode 100644
index 0000000000..aab6638900
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/div_mat.cu
@@ -0,0 +1,230 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct Div_8uc4_32f : binary_function<uint, float, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, float b) const
+        {
+            uint res = 0;
+
+            if (b != 0)
+            {
+                b = 1.0f / b;
+                res |= (saturate_cast<uchar>((0xffu & (a      )) * b)      );
+                res |= (saturate_cast<uchar>((0xffu & (a >>  8)) * b) <<  8);
+                res |= (saturate_cast<uchar>((0xffu & (a >> 16)) * b) << 16);
+                res |= (saturate_cast<uchar>((0xffu & (a >> 24)) * b) << 24);
+            }
+
+            return res;
+        }
+    };
+
+    struct Div_16sc4_32f : binary_function<short4, float, short4>
+    {
+        __device__ __forceinline__ short4 operator ()(short4 a, float b) const
+        {
+            return b != 0 ? make_short4(saturate_cast<short>(a.x / b), saturate_cast<short>(a.y / b),
+                                        saturate_cast<short>(a.z / b), saturate_cast<short>(a.w / b))
+                          : make_short4(0,0,0,0);
+        }
+    };
+
+    template <typename T, typename D> struct Div : binary_function<T, T, D>
+    {
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return b != 0 ? saturate_cast<D>(a / b) : 0;
+        }
+
+        __host__ __device__ __forceinline__ Div() {}
+        __host__ __device__ __forceinline__ Div(const Div&) {}
+    };
+    template <typename T> struct Div<T, float> : binary_function<T, T, float>
+    {
+        __device__ __forceinline__ float operator ()(T a, T b) const
+        {
+            return b != 0 ? static_cast<float>(a) / b : 0;
+        }
+
+        __host__ __device__ __forceinline__ Div() {}
+        __host__ __device__ __forceinline__ Div(const Div&) {}
+    };
+    template <typename T> struct Div<T, double> : binary_function<T, T, double>
+    {
+        __device__ __forceinline__ double operator ()(T a, T b) const
+        {
+            return b != 0 ? static_cast<double>(a) / b : 0;
+        }
+
+        __host__ __device__ __forceinline__ Div() {}
+        __host__ __device__ __forceinline__ Div(const Div&) {}
+    };
+
+    template <typename T, typename S, typename D> struct DivScale : binary_function<T, T, D>
+    {
+        S scale;
+
+        __host__ explicit DivScale(S scale_) : scale(scale_) {}
+
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return b != 0 ? saturate_cast<D>(scale * a / b) : 0;
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits<arithm::Div_8uc4_32f> : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T, typename D> struct TransformFunctorTraits< arithm::Div<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::DivScale<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void divMat_8uc4_32f(PtrStepSz<uint> src1, PtrStepSzf src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, Div_8uc4_32f(), WithOutMask(), stream);
+    }
+
+    void divMat_16sc4_32f(PtrStepSz<short4> src1, PtrStepSzf src2, PtrStepSz<short4> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, Div_16sc4_32f(), WithOutMask(), stream);
+    }
+
+    template <typename T, typename S, typename D>
+    void divMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream)
+    {
+        if (scale == 1)
+        {
+            Div<T, D> op;
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+        else
+        {
+            DivScale<T, S, D> op(static_cast<S>(scale));
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+    }
+
+    template void divMat<uchar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<uchar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    template void divMat<schar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<schar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void divMat<ushort, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<ushort, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<ushort, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<ushort, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<ushort, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<ushort, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<ushort, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void divMat<short, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<short, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<short, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<short, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<short, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<short, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<short, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void divMat<int, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<int, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<int, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<int, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<int, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<int, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<int, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void divMat<float, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<float, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<float, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<float, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<float, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<float, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<float, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void divMat<double, double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<double, double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<double, double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<double, double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<double, double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void divMat<double, double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void divMat<double, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/div_scalar.cu b/modules/gpuarithm/src/cuda/div_scalar.cu
new file mode 100644
index 0000000000..464c4adf87
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/div_scalar.cu
@@ -0,0 +1,168 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T, typename S, typename D> struct DivScalar : unary_function<T, D>
+    {
+        S val;
+
+        __host__ explicit DivScalar(S val_) : val(val_) {}
+
+        __device__ __forceinline__ D operator ()(T a) const
+        {
+            return saturate_cast<D>(a / val);
+        }
+    };
+
+    template <typename T, typename S, typename D> struct DivScalarInv : unary_function<T, D>
+    {
+        S val;
+
+        explicit DivScalarInv(S val_) : val(val_) {}
+
+        __device__ __forceinline__ D operator ()(T a) const
+        {
+            return a != 0 ? saturate_cast<D>(val / a) : 0;
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::DivScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::DivScalarInv<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T, typename S, typename D>
+    void divScalar(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream)
+    {
+        if (inv)
+        {
+            DivScalarInv<T, S, D> op(static_cast<S>(val));
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+        else
+        {
+            DivScalar<T, S, D> op(static_cast<S>(val));
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+    }
+
+    template void divScalar<uchar, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<uchar, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    template void divScalar<schar, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<schar, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void divScalar<ushort, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<ushort, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<ushort, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<ushort, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<ushort, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<ushort, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<ushort, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void divScalar<short, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<short, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<short, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<short, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<short, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<short, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<short, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void divScalar<int, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<int, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<int, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<int, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<int, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<int, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<int, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void divScalar<float, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<float, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<float, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<float, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<float, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<float, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<float, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void divScalar<double, double, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<double, double, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<double, double, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<double, double, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<double, double, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    //template void divScalar<double, double, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+    template void divScalar<double, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpu/src/cuda/integral_image.cu b/modules/gpuarithm/src/cuda/integral.cu
similarity index 98%
rename from modules/gpu/src/cuda/integral_image.cu
rename to modules/gpuarithm/src/cuda/integral.cu
index 73cefa9a0b..d8276b2c9a 100644
--- a/modules/gpu/src/cuda/integral_image.cu
+++ b/modules/gpuarithm/src/cuda/integral.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -469,4 +469,4 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuarithm/src/cuda/math.cu b/modules/gpuarithm/src/cuda/math.cu
new file mode 100644
index 0000000000..1f2e7b8a14
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/math.cu
@@ -0,0 +1,302 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/type_traits.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+//////////////////////////////////////////////////////////////////////////
+// absMat
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< abs_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T>
+    void absMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, abs_func<T>(), WithOutMask(), stream);
+    }
+
+    template void absMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void absMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// sqrMat
+
+namespace arithm
+{
+    template <typename T> struct Sqr : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(T x) const
+        {
+            return saturate_cast<T>(x * x);
+        }
+
+        __host__ __device__ __forceinline__ Sqr() {}
+        __host__ __device__ __forceinline__ Sqr(const Sqr&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< arithm::Sqr<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T>
+    void sqrMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, Sqr<T>(), WithOutMask(), stream);
+    }
+
+    template void sqrMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// sqrtMat
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< sqrt_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T>
+    void sqrtMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, sqrt_func<T>(), WithOutMask(), stream);
+    }
+
+    template void sqrtMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void sqrtMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// logMat
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< log_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T>
+    void logMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, log_func<T>(), WithOutMask(), stream);
+    }
+
+    template void logMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void logMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// expMat
+
+namespace arithm
+{
+    template <typename T> struct Exp : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(T x) const
+        {
+            exp_func<T> f;
+            return saturate_cast<T>(f(x));
+        }
+
+        __host__ __device__ __forceinline__ Exp() {}
+        __host__ __device__ __forceinline__ Exp(const Exp&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< arithm::Exp<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T>
+    void expMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, Exp<T>(), WithOutMask(), stream);
+    }
+
+    template void expMat<uchar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<schar>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<ushort>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<short>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<int>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<float>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    template void expMat<double>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// pow
+
+namespace arithm
+{
+    template<typename T, bool Signed = numeric_limits<T>::is_signed> struct PowOp : unary_function<T, T>
+    {
+        float power;
+
+        __host__ explicit PowOp(double power_) : power(static_cast<float>(power_)) {}
+
+        __device__ __forceinline__ T operator()(T e) const
+        {
+            return saturate_cast<T>(__powf((float)e, power));
+        }
+    };
+    template<typename T> struct PowOp<T, true> : unary_function<T, T>
+    {
+        float power;
+
+        __host__ explicit PowOp(double power_) : power(static_cast<float>(power_)) {}
+
+        __device__ __forceinline__ T operator()(T e) const
+        {
+            T res = saturate_cast<T>(__powf((float)e, power));
+
+            if ((e < 0) && (1 & static_cast<int>(power)))
+                res *= -1;
+
+            return res;
+        }
+    };
+    template<> struct PowOp<float> : unary_function<float, float>
+    {
+        float power;
+
+        __host__ explicit PowOp(double power_) : power(static_cast<float>(power_)) {}
+
+        __device__ __forceinline__ float operator()(float e) const
+        {
+            return __powf(::fabs(e), power);
+        }
+    };
+    template<> struct PowOp<double> : unary_function<double, double>
+    {
+        double power;
+
+        __host__ explicit PowOp(double power_) : power(power_) {}
+
+        __device__ __forceinline__ double operator()(double e) const
+        {
+            return ::pow(::fabs(e), power);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< arithm::PowOp<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template<typename T>
+    void pow(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src, (PtrStepSz<T>) dst, PowOp<T>(power), WithOutMask(), stream);
+    }
+
+    template void pow<uchar>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<schar>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<short>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<ushort>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<int>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<float>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    template void pow<double>(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/minmax.cu b/modules/gpuarithm/src/cuda/minmax.cu
new file mode 100644
index 0000000000..1bdeb7ddd6
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/minmax.cu
@@ -0,0 +1,247 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace minMax
+{
+    __device__ unsigned int blocks_finished = 0;
+
+    // To avoid shared bank conflicts we convert each value into value of
+    // appropriate type (32 bits minimum)
+    template <typename T> struct MinMaxTypeTraits;
+    template <> struct MinMaxTypeTraits<uchar> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<schar> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<ushort> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<short> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<int> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<float> { typedef float best_type; };
+    template <> struct MinMaxTypeTraits<double> { typedef double best_type; };
+
+    template <int BLOCK_SIZE, typename R>
+    struct GlobalReduce
+    {
+        static __device__ void run(R& mymin, R& mymax, R* minval, R* maxval, int tid, int bid, R* sminval, R* smaxval)
+        {
+        #if __CUDA_ARCH__ >= 200
+            if (tid == 0)
+            {
+                Emulation::glob::atomicMin(minval, mymin);
+                Emulation::glob::atomicMax(maxval, mymax);
+            }
+        #else
+            __shared__ bool is_last;
+
+            if (tid == 0)
+            {
+                minval[bid] = mymin;
+                maxval[bid] = mymax;
+
+                __threadfence();
+
+                unsigned int ticket = ::atomicAdd(&blocks_finished, 1);
+                is_last = (ticket == gridDim.x * gridDim.y - 1);
+            }
+
+            __syncthreads();
+
+            if (is_last)
+            {
+                int idx = ::min(tid, gridDim.x * gridDim.y - 1);
+
+                mymin = minval[idx];
+                mymax = maxval[idx];
+
+                const minimum<R> minOp;
+                const maximum<R> maxOp;
+                cudev::reduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax), tid, thrust::make_tuple(minOp, maxOp));
+
+                if (tid == 0)
+                {
+                    minval[0] = mymin;
+                    maxval[0] = mymax;
+
+                    blocks_finished = 0;
+                }
+            }
+        #endif
+        }
+    };
+
+    template <int BLOCK_SIZE, typename T, typename R, class Mask>
+    __global__ void kernel(const PtrStepSz<T> src, const Mask mask, R* minval, R* maxval, const int twidth, const int theight)
+    {
+        __shared__ R sminval[BLOCK_SIZE];
+        __shared__ R smaxval[BLOCK_SIZE];
+
+        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
+        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
+
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
+
+        R mymin = numeric_limits<R>::max();
+        R mymax = -numeric_limits<R>::max();
+
+        const minimum<R> minOp;
+        const maximum<R> maxOp;
+
+        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
+        {
+            const T* ptr = src.ptr(y);
+
+            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
+            {
+                if (mask(y, x))
+                {
+                    const R srcVal = ptr[x];
+
+                    mymin = minOp(mymin, srcVal);
+                    mymax = maxOp(mymax, srcVal);
+                }
+            }
+        }
+
+        cudev::reduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax), tid, thrust::make_tuple(minOp, maxOp));
+
+        GlobalReduce<BLOCK_SIZE, R>::run(mymin, mymax, minval, maxval, tid, bid, sminval, smaxval);
+    }
+
+    const int threads_x = 32;
+    const int threads_y = 8;
+
+    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
+    {
+        block = dim3(threads_x, threads_y);
+
+        grid = dim3(divUp(cols, block.x * block.y),
+                    divUp(rows, block.y * block.x));
+
+        grid.x = ::min(grid.x, block.x);
+        grid.y = ::min(grid.y, block.y);
+    }
+
+    void getBufSize(int cols, int rows, int& bufcols, int& bufrows)
+    {
+        dim3 block, grid;
+        getLaunchCfg(cols, rows, block, grid);
+
+        bufcols = grid.x * grid.y * sizeof(double);
+        bufrows = 2;
+    }
+
+    __global__ void setDefaultKernel(int* minval_buf, int* maxval_buf)
+    {
+        *minval_buf = numeric_limits<int>::max();
+        *maxval_buf = numeric_limits<int>::min();
+    }
+    __global__ void setDefaultKernel(float* minval_buf, float* maxval_buf)
+    {
+        *minval_buf = numeric_limits<float>::max();
+        *maxval_buf = -numeric_limits<float>::max();
+    }
+    __global__ void setDefaultKernel(double* minval_buf, double* maxval_buf)
+    {
+        *minval_buf = numeric_limits<double>::max();
+        *maxval_buf = -numeric_limits<double>::max();
+    }
+
+    template <typename R>
+    void setDefault(R* minval_buf, R* maxval_buf)
+    {
+        setDefaultKernel<<<1, 1>>>(minval_buf, maxval_buf);
+    }
+
+    template <typename T>
+    void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf)
+    {
+        typedef typename MinMaxTypeTraits<T>::best_type R;
+
+        dim3 block, grid;
+        getLaunchCfg(src.cols, src.rows, block, grid);
+
+        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
+        const int theight = divUp(divUp(src.rows, grid.y), block.y);
+
+        R* minval_buf = (R*) buf.ptr(0);
+        R* maxval_buf = (R*) buf.ptr(1);
+
+        setDefault(minval_buf, maxval_buf);
+
+        if (mask.data)
+            kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, SingleMask(mask), minval_buf, maxval_buf, twidth, theight);
+        else
+            kernel<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, WithOutMask(), minval_buf, maxval_buf, twidth, theight);
+
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+
+        R minval_, maxval_;
+        cudaSafeCall( cudaMemcpy(&minval_, minval_buf, sizeof(R), cudaMemcpyDeviceToHost) );
+        cudaSafeCall( cudaMemcpy(&maxval_, maxval_buf, sizeof(R), cudaMemcpyDeviceToHost) );
+        *minval = minval_;
+        *maxval = maxval_;
+    }
+
+    template void run<uchar >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<schar >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<ushort>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<short >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<int   >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<float >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+    template void run<double>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/minmax_mat.cu b/modules/gpuarithm/src/cuda/minmax_mat.cu
new file mode 100644
index 0000000000..0bf5a468d9
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/minmax_mat.cu
@@ -0,0 +1,228 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+//////////////////////////////////////////////////////////////////////////
+// min
+
+namespace arithm
+{
+    struct VMin4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vmin4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VMin4() {}
+        __host__ __device__ __forceinline__ VMin4(const VMin4&) {}
+    };
+
+    struct VMin2 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vmin2(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VMin2() {}
+        __host__ __device__ __forceinline__ VMin2(const VMin2&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VMin4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <> struct TransformFunctorTraits< arithm::VMin2 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< minimum<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< binder2nd< minimum<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void minMat_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VMin4(), WithOutMask(), stream);
+    }
+
+    void minMat_v2(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VMin2(), WithOutMask(), stream);
+    }
+
+    template <typename T> void minMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, minimum<T>(), WithOutMask(), stream);
+    }
+
+    template void minMat<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template <typename T> void minScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::cudev::bind2nd(minimum<T>(), src2), WithOutMask(), stream);
+    }
+
+    template void minScalar<uchar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<schar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<ushort>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<short >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<int   >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<float >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void minScalar<double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+//////////////////////////////////////////////////////////////////////////
+// max
+
+namespace arithm
+{
+    struct VMax4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vmax4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VMax4() {}
+        __host__ __device__ __forceinline__ VMax4(const VMax4&) {}
+    };
+
+    struct VMax2 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vmax2(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VMax2() {}
+        __host__ __device__ __forceinline__ VMax2(const VMax2&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VMax4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <> struct TransformFunctorTraits< arithm::VMax2 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< maximum<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< binder2nd< maximum<T> > > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void maxMat_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VMax4(), WithOutMask(), stream);
+    }
+
+    void maxMat_v2(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VMax2(), WithOutMask(), stream);
+    }
+
+    template <typename T> void maxMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<T>) dst, maximum<T>(), WithOutMask(), stream);
+    }
+
+    template void maxMat<uchar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<schar >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<short >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<int   >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<float >(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxMat<double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+
+    template <typename T> void maxScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream)
+    {
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) dst, cv::gpu::cudev::bind2nd(maximum<T>(), src2), WithOutMask(), stream);
+    }
+
+    template void maxScalar<uchar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<schar >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<ushort>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<short >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<int   >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<float >(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+    template void maxScalar<double>(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/minmaxloc.cu b/modules/gpuarithm/src/cuda/minmaxloc.cu
new file mode 100644
index 0000000000..fbd7029861
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/minmaxloc.cu
@@ -0,0 +1,236 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace minMaxLoc
+{
+    // To avoid shared bank conflicts we convert each value into value of
+    // appropriate type (32 bits minimum)
+    template <typename T> struct MinMaxTypeTraits;
+    template <> struct MinMaxTypeTraits<unsigned char> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<signed char> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<unsigned short> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<short> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<int> { typedef int best_type; };
+    template <> struct MinMaxTypeTraits<float> { typedef float best_type; };
+    template <> struct MinMaxTypeTraits<double> { typedef double best_type; };
+
+    template <int BLOCK_SIZE, typename T, class Mask>
+    __global__ void kernel_pass_1(const PtrStepSz<T> src, const Mask mask, T* minval, T* maxval, unsigned int* minloc, unsigned int* maxloc, const int twidth, const int theight)
+    {
+        typedef typename MinMaxTypeTraits<T>::best_type work_type;
+
+        __shared__ work_type sminval[BLOCK_SIZE];
+        __shared__ work_type smaxval[BLOCK_SIZE];
+        __shared__ unsigned int sminloc[BLOCK_SIZE];
+        __shared__ unsigned int smaxloc[BLOCK_SIZE];
+
+        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
+        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
+
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
+
+        work_type mymin = numeric_limits<work_type>::max();
+        work_type mymax = -numeric_limits<work_type>::max();
+        unsigned int myminloc = 0;
+        unsigned int mymaxloc = 0;
+
+        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
+        {
+            const T* ptr = src.ptr(y);
+
+            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
+            {
+                if (mask(y, x))
+                {
+                    const work_type srcVal = ptr[x];
+
+                    if (srcVal < mymin)
+                    {
+                        mymin = srcVal;
+                        myminloc = y * src.cols + x;
+                    }
+
+                    if (srcVal > mymax)
+                    {
+                        mymax = srcVal;
+                        mymaxloc = y * src.cols + x;
+                    }
+                }
+            }
+        }
+
+        reduceKeyVal<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax),
+                                 smem_tuple(sminloc, smaxloc), thrust::tie(myminloc, mymaxloc),
+                                 tid,
+                                 thrust::make_tuple(less<work_type>(), greater<work_type>()));
+
+        if (tid == 0)
+        {
+            minval[bid] = (T) mymin;
+            maxval[bid] = (T) mymax;
+            minloc[bid] = myminloc;
+            maxloc[bid] = mymaxloc;
+        }
+    }
+    template <int BLOCK_SIZE, typename T>
+    __global__ void kernel_pass_2(T* minval, T* maxval, unsigned int* minloc, unsigned int* maxloc, int count)
+    {
+        typedef typename MinMaxTypeTraits<T>::best_type work_type;
+
+        __shared__ work_type sminval[BLOCK_SIZE];
+        __shared__ work_type smaxval[BLOCK_SIZE];
+        __shared__ unsigned int sminloc[BLOCK_SIZE];
+        __shared__ unsigned int smaxloc[BLOCK_SIZE];
+
+        unsigned int idx = ::min(threadIdx.x, count - 1);
+
+        work_type mymin = minval[idx];
+        work_type mymax = maxval[idx];
+        unsigned int myminloc = minloc[idx];
+        unsigned int mymaxloc = maxloc[idx];
+
+        reduceKeyVal<BLOCK_SIZE>(smem_tuple(sminval, smaxval), thrust::tie(mymin, mymax),
+                                 smem_tuple(sminloc, smaxloc), thrust::tie(myminloc, mymaxloc),
+                                 threadIdx.x,
+                                 thrust::make_tuple(less<work_type>(), greater<work_type>()));
+
+        if (threadIdx.x == 0)
+        {
+            minval[0] = (T) mymin;
+            maxval[0] = (T) mymax;
+            minloc[0] = myminloc;
+            maxloc[0] = mymaxloc;
+        }
+    }
+
+    const int threads_x = 32;
+    const int threads_y = 8;
+
+    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
+    {
+        block = dim3(threads_x, threads_y);
+
+        grid = dim3(divUp(cols, block.x * block.y),
+                    divUp(rows, block.y * block.x));
+
+        grid.x = ::min(grid.x, block.x);
+        grid.y = ::min(grid.y, block.y);
+    }
+
+    void getBufSize(int cols, int rows, size_t elem_size, int& b1cols, int& b1rows, int& b2cols, int& b2rows)
+    {
+        dim3 block, grid;
+        getLaunchCfg(cols, rows, block, grid);
+
+        // For values
+        b1cols = (int)(grid.x * grid.y * elem_size);
+        b1rows = 2;
+
+        // For locations
+        b2cols = grid.x * grid.y * sizeof(int);
+        b2rows = 2;
+    }
+
+    template <typename T>
+    void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf)
+    {
+        dim3 block, grid;
+        getLaunchCfg(src.cols, src.rows, block, grid);
+
+        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
+        const int theight = divUp(divUp(src.rows, grid.y), block.y);
+
+        T* minval_buf = (T*) valbuf.ptr(0);
+        T* maxval_buf = (T*) valbuf.ptr(1);
+        unsigned int* minloc_buf = locbuf.ptr(0);
+        unsigned int* maxloc_buf = locbuf.ptr(1);
+
+        if (mask.data)
+            kernel_pass_1<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, SingleMask(mask), minval_buf, maxval_buf, minloc_buf, maxloc_buf, twidth, theight);
+        else
+            kernel_pass_1<threads_x * threads_y><<<grid, block>>>((PtrStepSz<T>) src, WithOutMask(), minval_buf, maxval_buf, minloc_buf, maxloc_buf, twidth, theight);
+
+        cudaSafeCall( cudaGetLastError() );
+
+        kernel_pass_2<threads_x * threads_y><<<1, threads_x * threads_y>>>(minval_buf, maxval_buf, minloc_buf, maxloc_buf, grid.x * grid.y);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+
+        T minval_, maxval_;
+        cudaSafeCall( cudaMemcpy(&minval_, minval_buf, sizeof(T), cudaMemcpyDeviceToHost) );
+        cudaSafeCall( cudaMemcpy(&maxval_, maxval_buf, sizeof(T), cudaMemcpyDeviceToHost) );
+        *minval = minval_;
+        *maxval = maxval_;
+
+        unsigned int minloc_, maxloc_;
+        cudaSafeCall( cudaMemcpy(&minloc_, minloc_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost) );
+        cudaSafeCall( cudaMemcpy(&maxloc_, maxloc_buf, sizeof(unsigned int), cudaMemcpyDeviceToHost) );
+        minloc[1] = minloc_ / src.cols; minloc[0] = minloc_ - minloc[1] * src.cols;
+        maxloc[1] = maxloc_ / src.cols; maxloc[0] = maxloc_ - maxloc[1] * src.cols;
+    }
+
+    template void run<unsigned char >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<signed char >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<unsigned short>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<short >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<int   >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<float >(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+    template void run<double>(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/mul_mat.cu b/modules/gpuarithm/src/cuda/mul_mat.cu
new file mode 100644
index 0000000000..25bc8a5970
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/mul_mat.cu
@@ -0,0 +1,211 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct Mul_8uc4_32f : binary_function<uint, float, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, float b) const
+        {
+            uint res = 0;
+
+            res |= (saturate_cast<uchar>((0xffu & (a      )) * b)      );
+            res |= (saturate_cast<uchar>((0xffu & (a >>  8)) * b) <<  8);
+            res |= (saturate_cast<uchar>((0xffu & (a >> 16)) * b) << 16);
+            res |= (saturate_cast<uchar>((0xffu & (a >> 24)) * b) << 24);
+
+            return res;
+        }
+
+        __host__ __device__ __forceinline__ Mul_8uc4_32f() {}
+        __host__ __device__ __forceinline__ Mul_8uc4_32f(const Mul_8uc4_32f&) {}
+    };
+
+    struct Mul_16sc4_32f : binary_function<short4, float, short4>
+    {
+        __device__ __forceinline__ short4 operator ()(short4 a, float b) const
+        {
+            return make_short4(saturate_cast<short>(a.x * b), saturate_cast<short>(a.y * b),
+                               saturate_cast<short>(a.z * b), saturate_cast<short>(a.w * b));
+        }
+
+        __host__ __device__ __forceinline__ Mul_16sc4_32f() {}
+        __host__ __device__ __forceinline__ Mul_16sc4_32f(const Mul_16sc4_32f&) {}
+    };
+
+    template <typename T, typename D> struct Mul : binary_function<T, T, D>
+    {
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return saturate_cast<D>(a * b);
+        }
+
+        __host__ __device__ __forceinline__ Mul() {}
+        __host__ __device__ __forceinline__ Mul(const Mul&) {}
+    };
+
+    template <typename T, typename S, typename D> struct MulScale : binary_function<T, T, D>
+    {
+        S scale;
+
+        __host__ explicit MulScale(S scale_) : scale(scale_) {}
+
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return saturate_cast<D>(scale * a * b);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits<arithm::Mul_8uc4_32f> : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T, typename D> struct TransformFunctorTraits< arithm::Mul<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::MulScale<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void mulMat_8uc4_32f(PtrStepSz<uint> src1, PtrStepSzf src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, Mul_8uc4_32f(), WithOutMask(), stream);
+    }
+
+    void mulMat_16sc4_32f(PtrStepSz<short4> src1, PtrStepSzf src2, PtrStepSz<short4> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, Mul_16sc4_32f(), WithOutMask(), stream);
+    }
+
+    template <typename T, typename S, typename D>
+    void mulMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream)
+    {
+        if (scale == 1)
+        {
+            Mul<T, D> op;
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+        else
+        {
+            MulScale<T, S, D> op(static_cast<S>(scale));
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+        }
+    }
+
+    template void mulMat<uchar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<uchar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    template void mulMat<schar, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<schar, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void mulMat<ushort, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<ushort, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<ushort, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<ushort, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<ushort, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<ushort, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<ushort, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void mulMat<short, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<short, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<short, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<short, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<short, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<short, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<short, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void mulMat<int, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<int, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<int, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<int, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<int, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<int, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<int, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void mulMat<float, float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<float, float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<float, float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<float, float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<float, float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<float, float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<float, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+
+    //template void mulMat<double, double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<double, double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<double, double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<double, double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<double, double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    //template void mulMat<double, double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    template void mulMat<double, double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/mul_scalar.cu b/modules/gpuarithm/src/cuda/mul_scalar.cu
new file mode 100644
index 0000000000..6546550275
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/mul_scalar.cu
@@ -0,0 +1,144 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T, typename S, typename D> struct MulScalar : unary_function<T, D>
+    {
+        S val;
+
+        __host__ explicit MulScalar(S val_) : val(val_) {}
+
+        __device__ __forceinline__ D operator ()(T a) const
+        {
+            return saturate_cast<D>(a * val);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::MulScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T, typename S, typename D>
+    void mulScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream)
+    {
+        MulScalar<T, S, D> op(static_cast<S>(val));
+        cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+    }
+
+    template void mulScalar<uchar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<uchar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    template void mulScalar<schar, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<schar, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void mulScalar<ushort, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<ushort, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<ushort, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<ushort, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<ushort, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<ushort, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<ushort, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void mulScalar<short, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<short, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<short, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<short, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<short, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<short, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<short, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void mulScalar<int, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<int, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<int, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<int, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<int, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<int, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<int, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void mulScalar<float, float, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<float, float, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<float, float, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<float, float, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<float, float, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<float, float, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<float, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+
+    //template void mulScalar<double, double, uchar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<double, double, schar>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<double, double, ushort>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<double, double, short>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<double, double, int>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    //template void mulScalar<double, double, float>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    template void mulScalar<double, double, double>(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/mul_spectrums.cu b/modules/gpuarithm/src/cuda/mul_spectrums.cu
new file mode 100644
index 0000000000..1b58b8ca7f
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/mul_spectrums.cu
@@ -0,0 +1,171 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "cvconfig.h"
+
+#ifdef HAVE_CUFFT
+
+#include <cufft.h>
+
+#include "opencv2/core/cuda/common.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    //////////////////////////////////////////////////////////////////////////
+    // mulSpectrums
+
+    __global__ void mulSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x < c.cols && y < c.rows)
+        {
+            c.ptr(y)[x] = cuCmulf(a.ptr(y)[x], b.ptr(y)[x]);
+        }
+    }
+
+
+    void mulSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream)
+    {
+        dim3 threads(256);
+        dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
+
+        mulSpectrumsKernel<<<grid, threads, 0, stream>>>(a, b, c);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+
+    //////////////////////////////////////////////////////////////////////////
+    // mulSpectrums_CONJ
+
+    __global__ void mulSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x < c.cols && y < c.rows)
+        {
+            c.ptr(y)[x] = cuCmulf(a.ptr(y)[x], cuConjf(b.ptr(y)[x]));
+        }
+    }
+
+
+    void mulSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c, cudaStream_t stream)
+    {
+        dim3 threads(256);
+        dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
+
+        mulSpectrumsKernel_CONJ<<<grid, threads, 0, stream>>>(a, b, c);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+
+    //////////////////////////////////////////////////////////////////////////
+    // mulAndScaleSpectrums
+
+    __global__ void mulAndScaleSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x < c.cols && y < c.rows)
+        {
+            cufftComplex v = cuCmulf(a.ptr(y)[x], b.ptr(y)[x]);
+            c.ptr(y)[x] = make_cuFloatComplex(cuCrealf(v) * scale, cuCimagf(v) * scale);
+        }
+    }
+
+
+    void mulAndScaleSpectrums(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream)
+    {
+        dim3 threads(256);
+        dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
+
+        mulAndScaleSpectrumsKernel<<<grid, threads, 0, stream>>>(a, b, scale, c);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+
+    //////////////////////////////////////////////////////////////////////////
+    // mulAndScaleSpectrums_CONJ
+
+    __global__ void mulAndScaleSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x < c.cols && y < c.rows)
+        {
+            cufftComplex v = cuCmulf(a.ptr(y)[x], cuConjf(b.ptr(y)[x]));
+            c.ptr(y)[x] = make_cuFloatComplex(cuCrealf(v) * scale, cuCimagf(v) * scale);
+        }
+    }
+
+
+    void mulAndScaleSpectrums_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c, cudaStream_t stream)
+    {
+        dim3 threads(256);
+        dim3 grid(divUp(c.cols, threads.x), divUp(c.rows, threads.y));
+
+        mulAndScaleSpectrumsKernel_CONJ<<<grid, threads, 0, stream>>>(a, b, scale, c);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}}} // namespace cv { namespace gpu { namespace cudev
+
+#endif // HAVE_CUFFT
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/mathfunc.cu b/modules/gpuarithm/src/cuda/polar_cart.cu
similarity index 98%
rename from modules/gpu/src/cuda/mathfunc.cu
rename to modules/gpuarithm/src/cuda/polar_cart.cu
index 3ae8fdc7b6..88626c5d2e 100644
--- a/modules/gpu/src/cuda/mathfunc.cu
+++ b/modules/gpuarithm/src/cuda/polar_cart.cu
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace mathfunc
     {
@@ -212,6 +212,6 @@ namespace cv { namespace gpu { namespace device
             callers[mag.data == 0](mag, angle, x, y, angleInDegrees, stream);
         }
     } // namespace mathfunc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpuarithm/src/cuda/reduce.cu b/modules/gpuarithm/src/cuda/reduce.cu
new file mode 100644
index 0000000000..8588a3b234
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/reduce.cu
@@ -0,0 +1,340 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+
+#include "unroll_detail.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace reduce
+{
+    struct Sum
+    {
+        template <typename T>
+        __device__ __forceinline__ T startValue() const
+        {
+            return VecTraits<T>::all(0);
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T operator ()(T a, T b) const
+        {
+            return a + b;
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T result(T r, int) const
+        {
+            return r;
+        }
+
+        __host__ __device__ __forceinline__ Sum() {}
+        __host__ __device__ __forceinline__ Sum(const Sum&) {}
+    };
+
+    template <typename T> struct OutputType
+    {
+        typedef float type;
+    };
+    template <> struct OutputType<double>
+    {
+        typedef double type;
+    };
+
+    struct Avg
+    {
+        template <typename T>
+        __device__ __forceinline__ T startValue() const
+        {
+            return VecTraits<T>::all(0);
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T operator ()(T a, T b) const
+        {
+            return a + b;
+        }
+
+        template <typename T>
+        __device__ __forceinline__ typename TypeVec<typename OutputType<typename VecTraits<T>::elem_type>::type, VecTraits<T>::cn>::vec_type result(T r, float sz) const
+        {
+            return r / sz;
+        }
+
+        __host__ __device__ __forceinline__ Avg() {}
+        __host__ __device__ __forceinline__ Avg(const Avg&) {}
+    };
+
+    struct Min
+    {
+        template <typename T>
+        __device__ __forceinline__ T startValue() const
+        {
+            return VecTraits<T>::all(numeric_limits<typename VecTraits<T>::elem_type>::max());
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T operator ()(T a, T b) const
+        {
+            minimum<T> minOp;
+            return minOp(a, b);
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T result(T r, int) const
+        {
+            return r;
+        }
+
+        __host__ __device__ __forceinline__ Min() {}
+        __host__ __device__ __forceinline__ Min(const Min&) {}
+    };
+
+    struct Max
+    {
+        template <typename T>
+        __device__ __forceinline__ T startValue() const
+        {
+            return VecTraits<T>::all(-numeric_limits<typename VecTraits<T>::elem_type>::max());
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T operator ()(T a, T b) const
+        {
+            maximum<T> maxOp;
+            return maxOp(a, b);
+        }
+
+        template <typename T>
+        __device__ __forceinline__ T result(T r, int) const
+        {
+            return r;
+        }
+
+        __host__ __device__ __forceinline__ Max() {}
+        __host__ __device__ __forceinline__ Max(const Max&) {}
+    };
+
+    ///////////////////////////////////////////////////////////
+
+    template <typename T, typename S, typename D, class Op>
+    __global__ void rowsKernel(const PtrStepSz<T> src, D* dst, const Op op)
+    {
+        __shared__ S smem[16 * 16];
+
+        const int x = blockIdx.x * 16 + threadIdx.x;
+
+        S myVal = op.template startValue<S>();
+
+        if (x < src.cols)
+        {
+            for (int y = threadIdx.y; y < src.rows; y += 16)
+            {
+                S srcVal = src(y, x);
+                myVal = op(myVal, srcVal);
+            }
+        }
+
+        smem[threadIdx.x * 16 + threadIdx.y] = myVal;
+
+        __syncthreads();
+
+        volatile S* srow = smem + threadIdx.y * 16;
+
+        myVal = srow[threadIdx.x];
+        cudev::reduce<16>(srow, myVal, threadIdx.x, op);
+
+        if (threadIdx.x == 0)
+            srow[0] = myVal;
+
+        __syncthreads();
+
+        if (threadIdx.y == 0 && x < src.cols)
+            dst[x] = (D) op.result(smem[threadIdx.x * 16], src.rows);
+    }
+
+    template <typename T, typename S, typename D, class Op>
+    void rowsCaller(PtrStepSz<T> src, D* dst, cudaStream_t stream)
+    {
+        const dim3 block(16, 16);
+        const dim3 grid(divUp(src.cols, block.x));
+
+        Op op;
+        rowsKernel<T, S, D, Op><<<grid, block, 0, stream>>>(src, dst, op);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <typename T, typename S, typename D>
+    void rows(PtrStepSzb src, void* dst, int op, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSz<T> src, D* dst, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            rowsCaller<T, S, D, Sum>,
+            rowsCaller<T, S, D, Avg>,
+            rowsCaller<T, S, D, Max>,
+            rowsCaller<T, S, D, Min>
+        };
+
+        funcs[op]((PtrStepSz<T>) src, (D*) dst, stream);
+    }
+
+    template void rows<unsigned char, int, unsigned char>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned char, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned char, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned char, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    template void rows<unsigned short, int, unsigned short>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned short, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned short, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<unsigned short, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    template void rows<short, int, short>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<short, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<short, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<short, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    template void rows<int, int, int>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<int, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<int, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    template void rows<float, float, float>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+    template void rows<float, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    template void rows<double, double, double>(PtrStepSzb src, void* dst, int op, cudaStream_t stream);
+
+    ///////////////////////////////////////////////////////////
+
+    template <int BLOCK_SIZE, typename T, typename S, typename D, int cn, class Op>
+    __global__ void colsKernel(const PtrStepSz<typename TypeVec<T, cn>::vec_type> src, typename TypeVec<D, cn>::vec_type* dst, const Op op)
+    {
+        typedef typename TypeVec<T, cn>::vec_type src_type;
+        typedef typename TypeVec<S, cn>::vec_type work_type;
+        typedef typename TypeVec<D, cn>::vec_type dst_type;
+
+        __shared__ S smem[BLOCK_SIZE * cn];
+
+        const int y = blockIdx.x;
+
+        const src_type* srcRow = src.ptr(y);
+
+        work_type myVal = op.template startValue<work_type>();
+
+        for (int x = threadIdx.x; x < src.cols; x += BLOCK_SIZE)
+            myVal = op(myVal, saturate_cast<work_type>(srcRow[x]));
+
+        cudev::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(myVal), threadIdx.x, detail::Unroll<cn>::op(op));
+
+        if (threadIdx.x == 0)
+            dst[y] = saturate_cast<dst_type>(op.result(myVal, src.cols));
+    }
+
+    template <typename T, typename S, typename D, int cn, class Op> void colsCaller(PtrStepSzb src, void* dst, cudaStream_t stream)
+    {
+        const int BLOCK_SIZE = 256;
+
+        const dim3 block(BLOCK_SIZE);
+        const dim3 grid(src.rows);
+
+        Op op;
+        colsKernel<BLOCK_SIZE, T, S, D, cn, Op><<<grid, block, 0, stream>>>((PtrStepSz<typename TypeVec<T, cn>::vec_type>) src, (typename TypeVec<D, cn>::vec_type*) dst, op);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+    }
+
+    template <typename T, typename S, typename D> void cols(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, void* dst, cudaStream_t stream);
+        static const func_t funcs[5][4] =
+        {
+            {0,0,0,0},
+            {colsCaller<T, S, D, 1, Sum>, colsCaller<T, S, D, 1, Avg>, colsCaller<T, S, D, 1, Max>, colsCaller<T, S, D, 1, Min>},
+            {colsCaller<T, S, D, 2, Sum>, colsCaller<T, S, D, 2, Avg>, colsCaller<T, S, D, 2, Max>, colsCaller<T, S, D, 2, Min>},
+            {colsCaller<T, S, D, 3, Sum>, colsCaller<T, S, D, 3, Avg>, colsCaller<T, S, D, 3, Max>, colsCaller<T, S, D, 3, Min>},
+            {colsCaller<T, S, D, 4, Sum>, colsCaller<T, S, D, 4, Avg>, colsCaller<T, S, D, 4, Max>, colsCaller<T, S, D, 4, Min>},
+        };
+
+        funcs[cn][op](src, dst, stream);
+    }
+
+    template void cols<unsigned char, int, unsigned char>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned char, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned char, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned char, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+
+    template void cols<unsigned short, int, unsigned short>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned short, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned short, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<unsigned short, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+
+    template void cols<short, int, short>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<short, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<short, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<short, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+
+    template void cols<int, int, int>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<int, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<int, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+
+    template void cols<float, float, float>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+    template void cols<float, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+
+    template void cols<double, double, double>(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
+}
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/split_merge.cu b/modules/gpuarithm/src/cuda/split_merge.cu
similarity index 96%
rename from modules/gpu/src/cuda/split_merge.cu
rename to modules/gpuarithm/src/cuda/split_merge.cu
index 834b283f0c..388441c634 100644
--- a/modules/gpu/src/cuda/split_merge.cu
+++ b/modules/gpuarithm/src/cuda/split_merge.cu
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace split_merge
     {
@@ -278,7 +278,7 @@ namespace cv { namespace gpu { namespace device
         }
 
 
-        void merge_caller(const PtrStepSzb* src, PtrStepSzb& dst,
+        void merge(const PtrStepSzb* src, PtrStepSzb& dst,
                                      int total_channels, size_t elem_size,
                                      const cudaStream_t& stream)
         {
@@ -293,7 +293,7 @@ namespace cv { namespace gpu { namespace device
             MergeFunction merge_func = merge_func_tbl[merge_func_id];
 
             if (merge_func == 0)
-                cv::gpu::error("Unsupported channel count or data type", __FILE__, __LINE__, "merge_caller");
+                CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported channel count or data type");
 
             merge_func(src, dst, stream);
         }
@@ -487,7 +487,7 @@ namespace cv { namespace gpu { namespace device
         }
 
 
-        void split_caller(const PtrStepSzb& src, PtrStepSzb* dst, int num_channels, size_t elem_size1, const cudaStream_t& stream)
+        void split(const PtrStepSzb& src, PtrStepSzb* dst, int num_channels, size_t elem_size1, const cudaStream_t& stream)
         {
             static SplitFunction split_func_tbl[] =
             {
@@ -500,12 +500,12 @@ namespace cv { namespace gpu { namespace device
             SplitFunction split_func = split_func_tbl[split_func_id];
 
             if (split_func == 0)
-                cv::gpu::error("Unsupported channel count or data type", __FILE__, __LINE__, "split_caller");
+                CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported channel count or data type");
 
             split_func(src, dst, stream);
         }
     } // namespace split_merge
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpuarithm/src/cuda/sub_mat.cu b/modules/gpuarithm/src/cuda/sub_mat.cu
new file mode 100644
index 0000000000..077eafb356
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/sub_mat.cu
@@ -0,0 +1,185 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    struct VSub4 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vsub4(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VSub4() {}
+        __host__ __device__ __forceinline__ VSub4(const VSub4&) {}
+    };
+
+    struct VSub2 : binary_function<uint, uint, uint>
+    {
+        __device__ __forceinline__ uint operator ()(uint a, uint b) const
+        {
+            return vsub2(a, b);
+        }
+
+        __host__ __device__ __forceinline__ VSub2() {}
+        __host__ __device__ __forceinline__ VSub2(const VSub2&) {}
+    };
+
+    template <typename T, typename D> struct SubMat : binary_function<T, T, D>
+    {
+        __device__ __forceinline__ D operator ()(T a, T b) const
+        {
+            return saturate_cast<D>(a - b);
+        }
+
+        __host__ __device__ __forceinline__ SubMat() {}
+        __host__ __device__ __forceinline__ SubMat(const SubMat&) {}
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits< arithm::VSub4 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <> struct TransformFunctorTraits< arithm::VSub2 > : arithm::ArithmFuncTraits<sizeof(uint), sizeof(uint)>
+    {
+    };
+
+    template <typename T, typename D> struct TransformFunctorTraits< arithm::SubMat<T, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    void subMat_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VSub4(), WithOutMask(), stream);
+    }
+
+    void subMat_v2(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream)
+    {
+        cudev::transform(src1, src2, dst, VSub2(), WithOutMask(), stream);
+    }
+
+    template <typename T, typename D>
+    void subMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, SubMat<T, D>(), mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<T>) src2, (PtrStepSz<D>) dst, SubMat<T, D>(), WithOutMask(), stream);
+    }
+
+    template void subMat<uchar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<uchar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void subMat<schar, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<schar, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subMat<ushort, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<ushort, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<ushort, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<ushort, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<ushort, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<ushort, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<ushort, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subMat<short, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<short, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<short, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<short, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<short, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<short, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<short, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subMat<int, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<int, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<int, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<int, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<int, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<int, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<int, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subMat<float, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<float, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<float, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<float, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<float, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<float, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<float, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subMat<double, uchar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<double, schar>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<double, ushort>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<double, short>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<double, int>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subMat<double, float>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subMat<double, double>(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/sub_scalar.cu b/modules/gpuarithm/src/cuda/sub_scalar.cu
new file mode 100644
index 0000000000..619ab4310f
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/sub_scalar.cu
@@ -0,0 +1,149 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    template <typename T, typename S, typename D> struct SubScalar : unary_function<T, D>
+    {
+        S val;
+        int scale;
+
+        __host__ SubScalar(S val_, int scale_) : val(val_), scale(scale_) {}
+
+        __device__ __forceinline__ D operator ()(T a) const
+        {
+            return saturate_cast<D>(scale * (a - val));
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename S, typename D> struct TransformFunctorTraits< arithm::SubScalar<T, S, D> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(D)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <typename T, typename S, typename D>
+    void subScalar(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream)
+    {
+        SubScalar<T, S, D> op(static_cast<S>(val), inv ? -1 : 1);
+
+        if (mask.data)
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, mask, stream);
+        else
+            cudev::transform((PtrStepSz<T>) src1, (PtrStepSz<D>) dst, op, WithOutMask(), stream);
+    }
+
+    template void subScalar<uchar, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<uchar, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    template void subScalar<schar, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<schar, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subScalar<ushort, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<ushort, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<ushort, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<ushort, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<ushort, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<ushort, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<ushort, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subScalar<short, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<short, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<short, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<short, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<short, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<short, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<short, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subScalar<int, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<int, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<int, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<int, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<int, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<int, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<int, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subScalar<float, float, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<float, float, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<float, float, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<float, float, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<float, float, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<float, float, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<float, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+
+    //template void subScalar<double, double, uchar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<double, double, schar>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<double, double, ushort>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<double, double, short>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<double, double, int>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    //template void subScalar<double, double, float>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    template void subScalar<double, double, double>(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/sum.cu b/modules/gpuarithm/src/cuda/sum.cu
new file mode 100644
index 0000000000..2af7692061
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/sum.cu
@@ -0,0 +1,381 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+
+#include "unroll_detail.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace sum
+{
+    __device__ unsigned int blocks_finished = 0;
+
+    template <typename R, int cn> struct AtomicAdd;
+    template <typename R> struct AtomicAdd<R, 1>
+    {
+        static __device__ void run(R* ptr, R val)
+        {
+            Emulation::glob::atomicAdd(ptr, val);
+        }
+    };
+    template <typename R> struct AtomicAdd<R, 2>
+    {
+        typedef typename TypeVec<R, 2>::vec_type val_type;
+
+        static __device__ void run(R* ptr, val_type val)
+        {
+            Emulation::glob::atomicAdd(ptr, val.x);
+            Emulation::glob::atomicAdd(ptr + 1, val.y);
+        }
+    };
+    template <typename R> struct AtomicAdd<R, 3>
+    {
+        typedef typename TypeVec<R, 3>::vec_type val_type;
+
+        static __device__ void run(R* ptr, val_type val)
+        {
+            Emulation::glob::atomicAdd(ptr, val.x);
+            Emulation::glob::atomicAdd(ptr + 1, val.y);
+            Emulation::glob::atomicAdd(ptr + 2, val.z);
+        }
+    };
+    template <typename R> struct AtomicAdd<R, 4>
+    {
+        typedef typename TypeVec<R, 4>::vec_type val_type;
+
+        static __device__ void run(R* ptr, val_type val)
+        {
+            Emulation::glob::atomicAdd(ptr, val.x);
+            Emulation::glob::atomicAdd(ptr + 1, val.y);
+            Emulation::glob::atomicAdd(ptr + 2, val.z);
+            Emulation::glob::atomicAdd(ptr + 3, val.w);
+        }
+    };
+
+    template <int BLOCK_SIZE, typename R, int cn>
+    struct GlobalReduce
+    {
+        typedef typename TypeVec<R, cn>::vec_type result_type;
+
+        static __device__ void run(result_type& sum, result_type* result, int tid, int bid, R* smem)
+        {
+        #if __CUDA_ARCH__ >= 200
+            if (tid == 0)
+                AtomicAdd<R, cn>::run((R*) result, sum);
+        #else
+            __shared__ bool is_last;
+
+            if (tid == 0)
+            {
+                result[bid] = sum;
+
+                __threadfence();
+
+                unsigned int ticket = ::atomicAdd(&blocks_finished, 1);
+                is_last = (ticket == gridDim.x * gridDim.y - 1);
+            }
+
+            __syncthreads();
+
+            if (is_last)
+            {
+                sum = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<result_type>::all(0);
+
+                cudev::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(sum), tid, detail::Unroll<cn>::op(plus<R>()));
+
+                if (tid == 0)
+                {
+                    result[0] = sum;
+                    blocks_finished = 0;
+                }
+            }
+        #endif
+        }
+    };
+
+    template <int BLOCK_SIZE, typename src_type, typename result_type, class Mask, class Op>
+    __global__ void kernel(const PtrStepSz<src_type> src, result_type* result, const Mask mask, const Op op, const int twidth, const int theight)
+    {
+        typedef typename VecTraits<src_type>::elem_type T;
+        typedef typename VecTraits<result_type>::elem_type R;
+        const int cn = VecTraits<src_type>::cn;
+
+        __shared__ R smem[BLOCK_SIZE * cn];
+
+        const int x0 = blockIdx.x * blockDim.x * twidth + threadIdx.x;
+        const int y0 = blockIdx.y * blockDim.y * theight + threadIdx.y;
+
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+        const int bid = blockIdx.y * gridDim.x + blockIdx.x;
+
+        result_type sum = VecTraits<result_type>::all(0);
+
+        for (int i = 0, y = y0; i < theight && y < src.rows; ++i, y += blockDim.y)
+        {
+            const src_type* ptr = src.ptr(y);
+
+            for (int j = 0, x = x0; j < twidth && x < src.cols; ++j, x += blockDim.x)
+            {
+                if (mask(y, x))
+                {
+                    const src_type srcVal = ptr[x];
+                    sum = sum + op(saturate_cast<result_type>(srcVal));
+                }
+            }
+        }
+
+        cudev::reduce<BLOCK_SIZE>(detail::Unroll<cn>::template smem_tuple<BLOCK_SIZE>(smem), detail::Unroll<cn>::tie(sum), tid, detail::Unroll<cn>::op(plus<R>()));
+
+        GlobalReduce<BLOCK_SIZE, R, cn>::run(sum, result, tid, bid, smem);
+    }
+
+    const int threads_x = 32;
+    const int threads_y = 8;
+
+    void getLaunchCfg(int cols, int rows, dim3& block, dim3& grid)
+    {
+        block = dim3(threads_x, threads_y);
+
+        grid = dim3(divUp(cols, block.x * block.y),
+                    divUp(rows, block.y * block.x));
+
+        grid.x = ::min(grid.x, block.x);
+        grid.y = ::min(grid.y, block.y);
+    }
+
+    void getBufSize(int cols, int rows, int cn, int& bufcols, int& bufrows)
+    {
+        dim3 block, grid;
+        getLaunchCfg(cols, rows, block, grid);
+
+        bufcols = grid.x * grid.y * sizeof(double) * cn;
+        bufrows = 1;
+    }
+
+    template <typename T, typename R, int cn, template <typename> class Op>
+    void caller(PtrStepSzb src_, void* buf_, double* out, PtrStepSzb mask)
+    {
+        typedef typename TypeVec<T, cn>::vec_type src_type;
+        typedef typename TypeVec<R, cn>::vec_type result_type;
+
+        PtrStepSz<src_type> src(src_);
+        result_type* buf = (result_type*) buf_;
+
+        dim3 block, grid;
+        getLaunchCfg(src.cols, src.rows, block, grid);
+
+        const int twidth = divUp(divUp(src.cols, grid.x), block.x);
+        const int theight = divUp(divUp(src.rows, grid.y), block.y);
+
+        Op<result_type> op;
+
+        if (mask.data)
+            kernel<threads_x * threads_y><<<grid, block>>>(src, buf, SingleMask(mask), op, twidth, theight);
+        else
+            kernel<threads_x * threads_y><<<grid, block>>>(src, buf, WithOutMask(), op, twidth, theight);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+
+        R result[4] = {0, 0, 0, 0};
+        cudaSafeCall( cudaMemcpy(&result, buf, sizeof(result_type), cudaMemcpyDeviceToHost) );
+
+        out[0] = result[0];
+        out[1] = result[1];
+        out[2] = result[2];
+        out[3] = result[3];
+    }
+
+    template <typename T> struct SumType;
+    template <> struct SumType<uchar> { typedef unsigned int R; };
+    template <> struct SumType<schar> { typedef int R; };
+    template <> struct SumType<ushort> { typedef unsigned int R; };
+    template <> struct SumType<short> { typedef int R; };
+    template <> struct SumType<int> { typedef int R; };
+    template <> struct SumType<float> { typedef float R; };
+    template <> struct SumType<double> { typedef double R; };
+
+    template <typename T, int cn>
+    void run(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
+    {
+        typedef typename SumType<T>::R R;
+        caller<T, R, cn, identity>(src, buf, out, mask);
+    }
+
+    template void run<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void run<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void run<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template <typename T, int cn>
+    void runAbs(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
+    {
+        typedef typename SumType<T>::R R;
+        caller<T, R, cn, abs_func>(src, buf, out, mask);
+    }
+
+    template void runAbs<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runAbs<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runAbs<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template <typename T> struct Sqr : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(T x) const
+        {
+            return x * x;
+        }
+    };
+
+    template <typename T, int cn>
+    void runSqr(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask)
+    {
+        caller<T, double, cn, Sqr>(src, buf, out, mask);
+    }
+
+    template void runSqr<uchar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<uchar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<uchar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<uchar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<schar, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<schar, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<schar, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<schar, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<ushort, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<ushort, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<ushort, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<ushort, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<short, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<short, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<short, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<short, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<int, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<int, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<int, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<int, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<float, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<float, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<float, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<float, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+
+    template void runSqr<double, 1>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<double, 2>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<double, 3>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+    template void runSqr<double, 4>(PtrStepSzb src, void* buf, double* out, PtrStepSzb mask);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/threshold.cu b/modules/gpuarithm/src/cuda/threshold.cu
new file mode 100644
index 0000000000..73ce8cee7a
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/threshold.cu
@@ -0,0 +1,114 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/simd_functions.hpp"
+
+#include "arithm_func_traits.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct TransformFunctorTraits< thresh_binary_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< thresh_binary_inv_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< thresh_trunc_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< thresh_to_zero_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+
+    template <typename T> struct TransformFunctorTraits< thresh_to_zero_inv_func<T> > : arithm::ArithmFuncTraits<sizeof(T), sizeof(T)>
+    {
+    };
+}}}
+
+namespace arithm
+{
+    template <template <typename> class Op, typename T>
+    void threshold_caller(PtrStepSz<T> src, PtrStepSz<T> dst, T thresh, T maxVal, cudaStream_t stream)
+    {
+        Op<T> op(thresh, maxVal);
+        cudev::transform(src, dst, op, WithOutMask(), stream);
+    }
+
+    template <typename T>
+    void threshold(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream)
+    {
+        typedef void (*caller_t)(PtrStepSz<T> src, PtrStepSz<T> dst, T thresh, T maxVal, cudaStream_t stream);
+
+        static const caller_t callers[] =
+        {
+            threshold_caller<thresh_binary_func, T>,
+            threshold_caller<thresh_binary_inv_func, T>,
+            threshold_caller<thresh_trunc_func, T>,
+            threshold_caller<thresh_to_zero_func, T>,
+            threshold_caller<thresh_to_zero_inv_func, T>
+        };
+
+        callers[type]((PtrStepSz<T>) src, (PtrStepSz<T>) dst, static_cast<T>(thresh), static_cast<T>(maxVal), stream);
+    }
+
+    template void threshold<uchar>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<schar>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<ushort>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<short>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<int>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<float>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+    template void threshold<double>(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/transpose.cu b/modules/gpuarithm/src/cuda/transpose.cu
new file mode 100644
index 0000000000..b51dc201a3
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/transpose.cu
@@ -0,0 +1,122 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace arithm
+{
+    const int TRANSPOSE_TILE_DIM   = 16;
+    const int TRANSPOSE_BLOCK_ROWS = 16;
+
+    template <typename T>
+    __global__ void transposeKernel(const PtrStepSz<T> src, PtrStep<T> dst)
+    {
+        __shared__ T tile[TRANSPOSE_TILE_DIM][TRANSPOSE_TILE_DIM + 1];
+
+        int blockIdx_x, blockIdx_y;
+
+        // do diagonal reordering
+        if (gridDim.x == gridDim.y)
+        {
+            blockIdx_y = blockIdx.x;
+            blockIdx_x = (blockIdx.x + blockIdx.y) % gridDim.x;
+        }
+        else
+        {
+            int bid = blockIdx.x + gridDim.x * blockIdx.y;
+            blockIdx_y = bid % gridDim.y;
+            blockIdx_x = ((bid / gridDim.y) + blockIdx_y) % gridDim.x;
+        }
+
+        int xIndex = blockIdx_x * TRANSPOSE_TILE_DIM + threadIdx.x;
+        int yIndex = blockIdx_y * TRANSPOSE_TILE_DIM + threadIdx.y;
+
+        if (xIndex < src.cols)
+        {
+            for (int i = 0; i < TRANSPOSE_TILE_DIM; i += TRANSPOSE_BLOCK_ROWS)
+            {
+                if (yIndex + i < src.rows)
+                {
+                    tile[threadIdx.y + i][threadIdx.x] = src(yIndex + i, xIndex);
+                }
+            }
+        }
+
+        __syncthreads();
+
+        xIndex = blockIdx_y * TRANSPOSE_TILE_DIM + threadIdx.x;
+        yIndex = blockIdx_x * TRANSPOSE_TILE_DIM + threadIdx.y;
+
+        if (xIndex < src.rows)
+        {
+            for (int i = 0; i < TRANSPOSE_TILE_DIM; i += TRANSPOSE_BLOCK_ROWS)
+            {
+                if (yIndex + i < src.cols)
+                {
+                    dst(yIndex + i, xIndex) = tile[threadIdx.x][threadIdx.y + i];
+                }
+            }
+        }
+    }
+
+    template <typename T> void transpose(PtrStepSz<T> src, PtrStepSz<T> dst, cudaStream_t stream)
+    {
+        const dim3 block(TRANSPOSE_TILE_DIM, TRANSPOSE_TILE_DIM);
+        const dim3 grid(divUp(src.cols, block.x), divUp(src.rows, block.y));
+
+        transposeKernel<<<grid, block, 0, stream>>>(src, dst);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void transpose<int>(PtrStepSz<int> src, PtrStepSz<int> dst, cudaStream_t stream);
+    template void transpose<double>(PtrStepSz<double> src, PtrStepSz<double> dst, cudaStream_t stream);
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuarithm/src/cuda/unroll_detail.hpp b/modules/gpuarithm/src/cuda/unroll_detail.hpp
new file mode 100644
index 0000000000..993b10be5a
--- /dev/null
+++ b/modules/gpuarithm/src/cuda/unroll_detail.hpp
@@ -0,0 +1,135 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __UNROLL_DETAIL_HPP__
+#define __UNROLL_DETAIL_HPP__
+
+#include <thrust/tuple.h>
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+
+namespace detail
+{
+    template <int cn> struct Unroll;
+    template <> struct Unroll<1>
+    {
+        template <int BLOCK_SIZE, typename R>
+        static __device__ __forceinline__ volatile R* smem_tuple(R* smem)
+        {
+            return smem;
+        }
+
+        template <typename R>
+        static __device__ __forceinline__ R& tie(R& val)
+        {
+            return val;
+        }
+
+        template <class Op>
+        static __device__ __forceinline__ const Op& op(const Op& op)
+        {
+            return op;
+        }
+    };
+    template <> struct Unroll<2>
+    {
+        template <int BLOCK_SIZE, typename R>
+        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*> smem_tuple(R* smem)
+        {
+            return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE);
+        }
+
+        template <typename R>
+        static __device__ __forceinline__ thrust::tuple<typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&> tie(R& val)
+        {
+            return thrust::tie(val.x, val.y);
+        }
+
+        template <class Op>
+        static __device__ __forceinline__ const thrust::tuple<Op, Op> op(const Op& op)
+        {
+            return thrust::make_tuple(op, op);
+        }
+    };
+    template <> struct Unroll<3>
+    {
+        template <int BLOCK_SIZE, typename R>
+        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*, volatile R*> smem_tuple(R* smem)
+        {
+            return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE);
+        }
+
+        template <typename R>
+        static __device__ __forceinline__ thrust::tuple<typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&> tie(R& val)
+        {
+            return thrust::tie(val.x, val.y, val.z);
+        }
+
+        template <class Op>
+        static __device__ __forceinline__ const thrust::tuple<Op, Op, Op> op(const Op& op)
+        {
+            return thrust::make_tuple(op, op, op);
+        }
+    };
+    template <> struct Unroll<4>
+    {
+        template <int BLOCK_SIZE, typename R>
+        static __device__ __forceinline__ thrust::tuple<volatile R*, volatile R*, volatile R*, volatile R*> smem_tuple(R* smem)
+        {
+            return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE);
+        }
+
+        template <typename R>
+        static __device__ __forceinline__ thrust::tuple<typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&, typename cv::gpu::cudev::VecTraits<R>::elem_type&> tie(R& val)
+        {
+            return thrust::tie(val.x, val.y, val.z, val.w);
+        }
+
+        template <class Op>
+        static __device__ __forceinline__ const thrust::tuple<Op, Op, Op, Op> op(const Op& op)
+        {
+            return thrust::make_tuple(op, op, op, op);
+        }
+    };
+}
+
+#endif // __UNROLL_DETAIL_HPP__
diff --git a/modules/gpu/src/element_operations.cpp b/modules/gpuarithm/src/element_operations.cpp
similarity index 55%
rename from modules/gpu/src/element_operations.cpp
rename to modules/gpuarithm/src/element_operations.cpp
index eedb31364a..20473de381 100644
--- a/modules/gpu/src/element_operations.cpp
+++ b/modules/gpuarithm/src/element_operations.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,44 +47,119 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::gpu::add(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::add(const GpuMat&, const Scalar&, GpuMat&, const GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::subtract(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::subtract(const GpuMat&, const Scalar&, GpuMat&, const GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::multiply(const GpuMat&, const GpuMat&, GpuMat&, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::multiply(const GpuMat&, const Scalar&, GpuMat&, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::divide(const GpuMat&, const GpuMat&, GpuMat&, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::divide(const GpuMat&, const Scalar&, GpuMat&, double, int, Stream&) { throw_nogpu(); }
-void cv::gpu::divide(double, const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::absdiff(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::absdiff(const GpuMat&, const Scalar&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::abs(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::sqr(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::sqrt(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::exp(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::log(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::compare(const GpuMat&, const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::compare(const GpuMat&, Scalar, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_not(const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_or(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_or(const GpuMat&, const Scalar&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_and(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_and(const GpuMat&, const Scalar&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_xor(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::bitwise_xor(const GpuMat&, const Scalar&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::rshift(const GpuMat&, Scalar_<int>, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::lshift(const GpuMat&, Scalar_<int>, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::min(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::min(const GpuMat&, double, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::max(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::max(const GpuMat&, double, GpuMat&, Stream&) { throw_nogpu(); }
-double cv::gpu::threshold(const GpuMat&, GpuMat&, double, double, int, Stream&) {throw_nogpu(); return 0.0;}
-void cv::gpu::pow(const GpuMat&, double, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::alphaComp(const GpuMat&, const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
-void cv::gpu::addWeighted(const GpuMat&, double, const GpuMat&, double, double, GpuMat&, int, Stream&) { throw_nogpu(); }
+void cv::gpu::add(InputArray, InputArray, OutputArray, InputArray, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::subtract(InputArray, InputArray, OutputArray, InputArray, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::multiply(InputArray, InputArray, OutputArray, double, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::divide(InputArray, InputArray, OutputArray, double, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::absdiff(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::abs(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::sqr(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::sqrt(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::exp(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::log(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::pow(InputArray, double, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::compare(InputArray, InputArray, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::bitwise_not(InputArray, OutputArray, InputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::bitwise_or(InputArray, InputArray, OutputArray, InputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::bitwise_and(InputArray, InputArray, OutputArray, InputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::bitwise_xor(InputArray, InputArray, OutputArray, InputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::rshift(InputArray, Scalar_<int>, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::lshift(InputArray, Scalar_<int>, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::min(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::max(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::addWeighted(InputArray, double, InputArray, double, double, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+double cv::gpu::threshold(InputArray, OutputArray, double, double, int, Stream&) {throw_no_cuda(); return 0.0;}
+
+void cv::gpu::magnitude(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::magnitude(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::magnitudeSqr(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::magnitudeSqr(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::phase(InputArray, InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::gpu::cartToPolar(InputArray, InputArray, OutputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::gpu::polarToCart(InputArray, InputArray, OutputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
 
 #else
 
+////////////////////////////////////////////////////////////////////////
+// arithm_op
+
+namespace
+{
+    typedef void (*mat_mat_func_t)(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, double scale, Stream& stream, int op);
+    typedef void (*mat_scalar_func_t)(const GpuMat& src, Scalar val, bool inv, GpuMat& dst, const GpuMat& mask, double scale, Stream& stream, int op);
+
+    void arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst, InputArray _mask, double scale, int dtype, Stream& stream,
+                   mat_mat_func_t mat_mat_func, mat_scalar_func_t mat_scalar_func, int op = 0)
+    {
+        const int kind1 = _src1.kind();
+        const int kind2 = _src2.kind();
+
+        const bool isScalar1 = (kind1 == _InputArray::MATX);
+        const bool isScalar2 = (kind2 == _InputArray::MATX);
+        CV_Assert( !isScalar1 || !isScalar2 );
+
+        GpuMat src1;
+        if (!isScalar1)
+            src1 = _src1.getGpuMat();
+
+        GpuMat src2;
+        if (!isScalar2)
+            src2 = _src2.getGpuMat();
+
+        Mat scalar;
+        if (isScalar1)
+            scalar = _src1.getMat();
+        else if (isScalar2)
+            scalar = _src2.getMat();
+
+        Scalar val;
+        if (!scalar.empty())
+        {
+            CV_Assert( scalar.total() <= 4 );
+            scalar.convertTo(Mat_<double>(scalar.rows, scalar.cols, &val[0]), CV_64F);
+        }
+
+        GpuMat mask = _mask.getGpuMat();
+
+        const int sdepth = src1.empty() ? src2.depth() : src1.depth();
+        const int cn = src1.empty() ? src2.channels() : src1.channels();
+        const Size size = src1.empty() ? src2.size() : src1.size();
+
+        if (dtype < 0)
+            dtype = sdepth;
+
+        const int ddepth = CV_MAT_DEPTH(dtype);
+
+        CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
+        CV_Assert( !scalar.empty() || (src2.type() == src1.type() && src2.size() == src1.size()) );
+        CV_Assert( mask.empty() || (cn == 1 && mask.size() == size && mask.type() == CV_8UC1) );
+
+        if (sdepth == CV_64F || ddepth == CV_64F)
+        {
+            if (!deviceSupports(NATIVE_DOUBLE))
+                CV_Error(Error::StsUnsupportedFormat, "The device doesn't support double");
+        }
+
+        _dst.create(size, CV_MAKE_TYPE(ddepth, cn));
+        GpuMat dst = _dst.getGpuMat();
+
+        if (isScalar1)
+            mat_scalar_func(src2, val, true, dst, mask, scale, stream, op);
+        else if (isScalar2)
+            mat_scalar_func(src1, val, false, dst, mask, scale, stream, op);
+        else
+            mat_mat_func(src1, src2, dst, mask, scale, stream, op);
+    }
+}
+
+
 ////////////////////////////////////////////////////////////////////////
 // Basic arithmetical operations (add subtract multiply divide)
 
@@ -263,170 +338,94 @@ namespace
 
 namespace arithm
 {
-    template <typename T, typename D>
-    void vadd4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void vadd2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    void addMat_v4(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    void addMat_v2(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
 
     template <typename T, typename D>
     void addMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::add(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, int dtype, Stream& s)
+static void addMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, double, Stream& _stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            addMat<unsigned char, unsigned char>,
-            addMat<unsigned char, signed char>,
-            addMat<unsigned char, unsigned short>,
-            addMat<unsigned char, short>,
-            addMat<unsigned char, int>,
-            addMat<unsigned char, float>,
-            addMat<unsigned char, double>
+            arithm::addMat<unsigned char, unsigned char>,
+            arithm::addMat<unsigned char, signed char>,
+            arithm::addMat<unsigned char, unsigned short>,
+            arithm::addMat<unsigned char, short>,
+            arithm::addMat<unsigned char, int>,
+            arithm::addMat<unsigned char, float>,
+            arithm::addMat<unsigned char, double>
         },
         {
-            addMat<signed char, unsigned char>,
-            addMat<signed char, signed char>,
-            addMat<signed char, unsigned short>,
-            addMat<signed char, short>,
-            addMat<signed char, int>,
-            addMat<signed char, float>,
-            addMat<signed char, double>
+            arithm::addMat<signed char, unsigned char>,
+            arithm::addMat<signed char, signed char>,
+            arithm::addMat<signed char, unsigned short>,
+            arithm::addMat<signed char, short>,
+            arithm::addMat<signed char, int>,
+            arithm::addMat<signed char, float>,
+            arithm::addMat<signed char, double>
         },
         {
-            0 /*addMat<unsigned short, unsigned char>*/,
-            0 /*addMat<unsigned short, signed char>*/,
-            addMat<unsigned short, unsigned short>,
-            addMat<unsigned short, short>,
-            addMat<unsigned short, int>,
-            addMat<unsigned short, float>,
-            addMat<unsigned short, double>
+            0 /*arithm::addMat<unsigned short, unsigned char>*/,
+            0 /*arithm::addMat<unsigned short, signed char>*/,
+            arithm::addMat<unsigned short, unsigned short>,
+            arithm::addMat<unsigned short, short>,
+            arithm::addMat<unsigned short, int>,
+            arithm::addMat<unsigned short, float>,
+            arithm::addMat<unsigned short, double>
         },
         {
-            0 /*addMat<short, unsigned char>*/,
-            0 /*addMat<short, signed char>*/,
-            addMat<short, unsigned short>,
-            addMat<short, short>,
-            addMat<short, int>,
-            addMat<short, float>,
-            addMat<short, double>
+            0 /*arithm::addMat<short, unsigned char>*/,
+            0 /*arithm::addMat<short, signed char>*/,
+            arithm::addMat<short, unsigned short>,
+            arithm::addMat<short, short>,
+            arithm::addMat<short, int>,
+            arithm::addMat<short, float>,
+            arithm::addMat<short, double>
         },
         {
-            0 /*addMat<int, unsigned char>*/,
-            0 /*addMat<int, signed char>*/,
-            0 /*addMat<int, unsigned short>*/,
-            0 /*addMat<int, short>*/,
-            addMat<int, int>,
-            addMat<int, float>,
-            addMat<int, double>
+            0 /*arithm::addMat<int, unsigned char>*/,
+            0 /*arithm::addMat<int, signed char>*/,
+            0 /*arithm::addMat<int, unsigned short>*/,
+            0 /*arithm::addMat<int, short>*/,
+            arithm::addMat<int, int>,
+            arithm::addMat<int, float>,
+            arithm::addMat<int, double>
         },
         {
-            0 /*addMat<float, unsigned char>*/,
-            0 /*addMat<float, signed char>*/,
-            0 /*addMat<float, unsigned short>*/,
-            0 /*addMat<float, short>*/,
-            0 /*addMat<float, int>*/,
-            addMat<float, float>,
-            addMat<float, double>
+            0 /*arithm::addMat<float, unsigned char>*/,
+            0 /*arithm::addMat<float, signed char>*/,
+            0 /*arithm::addMat<float, unsigned short>*/,
+            0 /*arithm::addMat<float, short>*/,
+            0 /*arithm::addMat<float, int>*/,
+            arithm::addMat<float, float>,
+            arithm::addMat<float, double>
         },
         {
-            0 /*addMat<double, unsigned char>*/,
-            0 /*addMat<double, signed char>*/,
-            0 /*addMat<double, unsigned short>*/,
-            0 /*addMat<double, short>*/,
-            0 /*addMat<double, int>*/,
-            0 /*addMat<double, float>*/,
-            addMat<double, double>
+            0 /*arithm::addMat<double, unsigned char>*/,
+            0 /*arithm::addMat<double, signed char>*/,
+            0 /*arithm::addMat<double, unsigned short>*/,
+            0 /*arithm::addMat<double, short>*/,
+            0 /*arithm::addMat<double, int>*/,
+            0 /*arithm::addMat<double, float>*/,
+            arithm::addMat<double, double>
         }
     };
 
-    typedef void (*vfunc_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    static const vfunc_t vfuncs4[4][4] =
-    {
-        {
-            vadd4<unsigned int, unsigned int>,
-            vadd4<unsigned int, int>,
-            0,
-            0
-        },
-        {
-            vadd4<int, unsigned int>,
-            vadd4<int, int>,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        }
-    };
-    static const vfunc_t vfuncs2[4][4] =
-    {
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            vadd2<unsigned int, unsigned int>,
-            vadd2<unsigned int, int>
-        },
-        {
-            0,
-            0,
-            vadd2<int, unsigned int>,
-            vadd2<int, int>
-        }
-    };
-
-    if (dtype < 0)
-        dtype = src1.depth();
-
     const int sdepth = src1.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src1.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-    CV_Assert( mask.empty() || (cn == 1 && mask.size() == src1.size() && mask.type() == CV_8U) );
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
     PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
     PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-    if (mask.empty() && sdepth < CV_32S && ddepth < CV_32S)
+    if (mask.empty() && (sdepth == CV_8U || sdepth == CV_16U) && ddepth == sdepth)
     {
         const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
         const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
@@ -434,31 +433,27 @@ void cv::gpu::add(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const Gpu
 
         const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
 
-        if (deviceSupports(FEATURE_SET_COMPUTE_20) && isAllAligned)
+        if (isAllAligned)
         {
-            const vfunc_t vfunc4 = vfuncs4[sdepth][ddepth];
-            const vfunc_t vfunc2 = vfuncs2[sdepth][ddepth];
-
-            if (vfunc4 != 0 && (src1_.cols & 3) == 0)
+            if (sdepth == CV_8U && (src1_.cols & 3) == 0)
             {
                 const int vcols = src1_.cols >> 2;
 
-                vfunc4(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::addMat_v4(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                  stream);
 
                 return;
             }
-
-            if (vfunc2 != 0 && (src1_.cols & 1) == 0)
+            else if (sdepth == CV_16U && (src1_.cols & 1) == 0)
             {
                 const int vcols = src1_.cols >> 1;
 
-                vfunc2(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::addMat_v2(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                  stream);
 
                 return;
             }
@@ -468,7 +463,7 @@ void cv::gpu::add(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const Gpu
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
     func(src1_, src2_, dst_, mask, stream);
 }
@@ -479,75 +474,73 @@ namespace arithm
     void addScalar(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::add(const GpuMat& src, const Scalar& sc, GpuMat& dst, const GpuMat& mask, int dtype, Stream& s)
+static void addScalar(const GpuMat& src, Scalar val, bool, GpuMat& dst, const GpuMat& mask, double, Stream& _stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            addScalar<unsigned char, float, unsigned char>,
-            addScalar<unsigned char, float, signed char>,
-            addScalar<unsigned char, float, unsigned short>,
-            addScalar<unsigned char, float, short>,
-            addScalar<unsigned char, float, int>,
-            addScalar<unsigned char, float, float>,
-            addScalar<unsigned char, double, double>
+            arithm::addScalar<unsigned char, float, unsigned char>,
+            arithm::addScalar<unsigned char, float, signed char>,
+            arithm::addScalar<unsigned char, float, unsigned short>,
+            arithm::addScalar<unsigned char, float, short>,
+            arithm::addScalar<unsigned char, float, int>,
+            arithm::addScalar<unsigned char, float, float>,
+            arithm::addScalar<unsigned char, double, double>
         },
         {
-            addScalar<signed char, float, unsigned char>,
-            addScalar<signed char, float, signed char>,
-            addScalar<signed char, float, unsigned short>,
-            addScalar<signed char, float, short>,
-            addScalar<signed char, float, int>,
-            addScalar<signed char, float, float>,
-            addScalar<signed char, double, double>
+            arithm::addScalar<signed char, float, unsigned char>,
+            arithm::addScalar<signed char, float, signed char>,
+            arithm::addScalar<signed char, float, unsigned short>,
+            arithm::addScalar<signed char, float, short>,
+            arithm::addScalar<signed char, float, int>,
+            arithm::addScalar<signed char, float, float>,
+            arithm::addScalar<signed char, double, double>
         },
         {
-            0 /*addScalar<unsigned short, float, unsigned char>*/,
-            0 /*addScalar<unsigned short, float, signed char>*/,
-            addScalar<unsigned short, float, unsigned short>,
-            addScalar<unsigned short, float, short>,
-            addScalar<unsigned short, float, int>,
-            addScalar<unsigned short, float, float>,
-            addScalar<unsigned short, double, double>
+            0 /*arithm::addScalar<unsigned short, float, unsigned char>*/,
+            0 /*arithm::addScalar<unsigned short, float, signed char>*/,
+            arithm::addScalar<unsigned short, float, unsigned short>,
+            arithm::addScalar<unsigned short, float, short>,
+            arithm::addScalar<unsigned short, float, int>,
+            arithm::addScalar<unsigned short, float, float>,
+            arithm::addScalar<unsigned short, double, double>
         },
         {
-            0 /*addScalar<short, float, unsigned char>*/,
-            0 /*addScalar<short, float, signed char>*/,
-            addScalar<short, float, unsigned short>,
-            addScalar<short, float, short>,
-            addScalar<short, float, int>,
-            addScalar<short, float, float>,
-            addScalar<short, double, double>
+            0 /*arithm::addScalar<short, float, unsigned char>*/,
+            0 /*arithm::addScalar<short, float, signed char>*/,
+            arithm::addScalar<short, float, unsigned short>,
+            arithm::addScalar<short, float, short>,
+            arithm::addScalar<short, float, int>,
+            arithm::addScalar<short, float, float>,
+            arithm::addScalar<short, double, double>
         },
         {
-            0 /*addScalar<int, float, unsigned char>*/,
-            0 /*addScalar<int, float, signed char>*/,
-            0 /*addScalar<int, float, unsigned short>*/,
-            0 /*addScalar<int, float, short>*/,
-            addScalar<int, float, int>,
-            addScalar<int, float, float>,
-            addScalar<int, double, double>
+            0 /*arithm::addScalar<int, float, unsigned char>*/,
+            0 /*arithm::addScalar<int, float, signed char>*/,
+            0 /*arithm::addScalar<int, float, unsigned short>*/,
+            0 /*arithm::addScalar<int, float, short>*/,
+            arithm::addScalar<int, float, int>,
+            arithm::addScalar<int, float, float>,
+            arithm::addScalar<int, double, double>
         },
         {
-            0 /*addScalar<float, float, unsigned char>*/,
-            0 /*addScalar<float, float, signed char>*/,
-            0 /*addScalar<float, float, unsigned short>*/,
-            0 /*addScalar<float, float, short>*/,
-            0 /*addScalar<float, float, int>*/,
-            addScalar<float, float, float>,
-            addScalar<float, double, double>
+            0 /*arithm::addScalar<float, float, unsigned char>*/,
+            0 /*arithm::addScalar<float, float, signed char>*/,
+            0 /*arithm::addScalar<float, float, unsigned short>*/,
+            0 /*arithm::addScalar<float, float, short>*/,
+            0 /*arithm::addScalar<float, float, int>*/,
+            arithm::addScalar<float, float, float>,
+            arithm::addScalar<float, double, double>
         },
         {
-            0 /*addScalar<double, double, unsigned char>*/,
-            0 /*addScalar<double, double, signed char>*/,
-            0 /*addScalar<double, double, unsigned short>*/,
-            0 /*addScalar<double, double, short>*/,
-            0 /*addScalar<double, double, int>*/,
-            0 /*addScalar<double, double, float>*/,
-            addScalar<double, double, double>
+            0 /*arithm::addScalar<double, double, unsigned char>*/,
+            0 /*arithm::addScalar<double, double, signed char>*/,
+            0 /*arithm::addScalar<double, double, unsigned short>*/,
+            0 /*arithm::addScalar<double, double, short>*/,
+            0 /*arithm::addScalar<double, double, int>*/,
+            0 /*arithm::addScalar<double, double, float>*/,
+            arithm::addScalar<double, double, double>
         }
     };
 
@@ -563,31 +556,16 @@ void cv::gpu::add(const GpuMat& src, const Scalar& sc, GpuMat& dst, const GpuMat
         {0                                                    , 0                                                     , 0                                                    , 0                                                    }
     };
 
-    if (dtype < 0)
-        dtype = src.depth();
-
     const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( cn <= 4 );
-    CV_Assert( mask.empty() || (cn == 1 && mask.size() == src.size() && mask.type() == CV_8U) );
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     const npp_func_t npp_func = npp_funcs[sdepth][cn - 1];
     if (ddepth == sdepth && cn > 1 && npp_func != 0)
     {
-        npp_func(src, sc, dst, stream);
+        npp_func(src, val, dst, stream);
         return;
     }
 
@@ -596,9 +574,14 @@ void cv::gpu::add(const GpuMat& src, const Scalar& sc, GpuMat& dst, const GpuMat
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
-    func(src, sc.val[0], dst, mask, stream);
+    func(src, val[0], dst, mask, stream);
+}
+
+void cv::gpu::add(InputArray src1, InputArray src2, OutputArray dst, InputArray mask, int dtype, Stream& stream)
+{
+    arithm_op(src1, src2, dst, mask, 1.0, dtype, stream, addMat, addScalar);
 }
 
 ////////////////////////////////////////////////////////////////////////
@@ -606,170 +589,94 @@ void cv::gpu::add(const GpuMat& src, const Scalar& sc, GpuMat& dst, const GpuMat
 
 namespace arithm
 {
-    template <typename T, typename D>
-    void vsub4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T, typename D>
-    void vsub2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    void subMat_v4(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    void subMat_v2(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
 
     template <typename T, typename D>
     void subMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::subtract(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, int dtype, Stream& s)
+static void subMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, double, Stream& _stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            subMat<unsigned char, unsigned char>,
-            subMat<unsigned char, signed char>,
-            subMat<unsigned char, unsigned short>,
-            subMat<unsigned char, short>,
-            subMat<unsigned char, int>,
-            subMat<unsigned char, float>,
-            subMat<unsigned char, double>
+            arithm::subMat<unsigned char, unsigned char>,
+            arithm::subMat<unsigned char, signed char>,
+            arithm::subMat<unsigned char, unsigned short>,
+            arithm::subMat<unsigned char, short>,
+            arithm::subMat<unsigned char, int>,
+            arithm::subMat<unsigned char, float>,
+            arithm::subMat<unsigned char, double>
         },
         {
-            subMat<signed char, unsigned char>,
-            subMat<signed char, signed char>,
-            subMat<signed char, unsigned short>,
-            subMat<signed char, short>,
-            subMat<signed char, int>,
-            subMat<signed char, float>,
-            subMat<signed char, double>
+            arithm::subMat<signed char, unsigned char>,
+            arithm::subMat<signed char, signed char>,
+            arithm::subMat<signed char, unsigned short>,
+            arithm::subMat<signed char, short>,
+            arithm::subMat<signed char, int>,
+            arithm::subMat<signed char, float>,
+            arithm::subMat<signed char, double>
         },
         {
-            0 /*subMat<unsigned short, unsigned char>*/,
-            0 /*subMat<unsigned short, signed char>*/,
-            subMat<unsigned short, unsigned short>,
-            subMat<unsigned short, short>,
-            subMat<unsigned short, int>,
-            subMat<unsigned short, float>,
-            subMat<unsigned short, double>
+            0 /*arithm::subMat<unsigned short, unsigned char>*/,
+            0 /*arithm::subMat<unsigned short, signed char>*/,
+            arithm::subMat<unsigned short, unsigned short>,
+            arithm::subMat<unsigned short, short>,
+            arithm::subMat<unsigned short, int>,
+            arithm::subMat<unsigned short, float>,
+            arithm::subMat<unsigned short, double>
         },
         {
-            0 /*subMat<short, unsigned char>*/,
-            0 /*subMat<short, signed char>*/,
-            subMat<short, unsigned short>,
-            subMat<short, short>,
-            subMat<short, int>,
-            subMat<short, float>,
-            subMat<short, double>
+            0 /*arithm::subMat<short, unsigned char>*/,
+            0 /*arithm::subMat<short, signed char>*/,
+            arithm::subMat<short, unsigned short>,
+            arithm::subMat<short, short>,
+            arithm::subMat<short, int>,
+            arithm::subMat<short, float>,
+            arithm::subMat<short, double>
         },
         {
-            0 /*subMat<int, unsigned char>*/,
-            0 /*subMat<int, signed char>*/,
-            0 /*subMat<int, unsigned short>*/,
-            0 /*subMat<int, short>*/,
-            subMat<int, int>,
-            subMat<int, float>,
-            subMat<int, double>
+            0 /*arithm::subMat<int, unsigned char>*/,
+            0 /*arithm::subMat<int, signed char>*/,
+            0 /*arithm::subMat<int, unsigned short>*/,
+            0 /*arithm::subMat<int, short>*/,
+            arithm::subMat<int, int>,
+            arithm::subMat<int, float>,
+            arithm::subMat<int, double>
         },
         {
-            0 /*subMat<float, unsigned char>*/,
-            0 /*subMat<float, signed char>*/,
-            0 /*subMat<float, unsigned short>*/,
-            0 /*subMat<float, short>*/,
-            0 /*subMat<float, int>*/,
-            subMat<float, float>,
-            subMat<float, double>
+            0 /*arithm::subMat<float, unsigned char>*/,
+            0 /*arithm::subMat<float, signed char>*/,
+            0 /*arithm::subMat<float, unsigned short>*/,
+            0 /*arithm::subMat<float, short>*/,
+            0 /*arithm::subMat<float, int>*/,
+            arithm::subMat<float, float>,
+            arithm::subMat<float, double>
         },
         {
-            0 /*subMat<double, unsigned char>*/,
-            0 /*subMat<double, signed char>*/,
-            0 /*subMat<double, unsigned short>*/,
-            0 /*subMat<double, short>*/,
-            0 /*subMat<double, int>*/,
-            0 /*subMat<double, float>*/,
-            subMat<double, double>
+            0 /*arithm::subMat<double, unsigned char>*/,
+            0 /*arithm::subMat<double, signed char>*/,
+            0 /*arithm::subMat<double, unsigned short>*/,
+            0 /*arithm::subMat<double, short>*/,
+            0 /*arithm::subMat<double, int>*/,
+            0 /*arithm::subMat<double, float>*/,
+            arithm::subMat<double, double>
         }
     };
 
-    typedef void (*vfunc_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    static const vfunc_t vfuncs4[4][4] =
-    {
-        {
-            vsub4<unsigned int, unsigned int>,
-            vsub4<unsigned int, int>,
-            0,
-            0
-        },
-        {
-            vsub4<int, unsigned int>,
-            vsub4<int, int>,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        }
-    };
-    static const vfunc_t vfuncs2[4][4] =
-    {
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            0,
-            0
-        },
-        {
-            0,
-            0,
-            vsub2<unsigned int, unsigned int>,
-            vsub2<unsigned int, int>
-        },
-        {
-            0,
-            0,
-            vsub2<int, unsigned int>,
-            vsub2<int, int>
-        }
-    };
-
-    if (dtype < 0)
-        dtype = src1.depth();
-
     const int sdepth = src1.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src1.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-    CV_Assert( mask.empty() || (cn == 1 && mask.size() == src1.size() && mask.type() == CV_8U) );
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
     PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
     PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-    if (mask.empty() && sdepth < CV_32S && ddepth < CV_32S)
+    if (mask.empty() && (sdepth == CV_8U || sdepth == CV_16U) && ddepth == sdepth)
     {
         const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
         const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
@@ -777,31 +684,27 @@ void cv::gpu::subtract(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cons
 
         const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
 
-        if (deviceSupports(FEATURE_SET_COMPUTE_20) && isAllAligned)
+        if (isAllAligned)
         {
-            const vfunc_t vfunc4 = vfuncs4[sdepth][ddepth];
-            const vfunc_t vfunc2 = vfuncs2[sdepth][ddepth];
-
-            if (vfunc4 != 0 && (src1_.cols & 3) == 0)
+            if (sdepth == CV_8U && (src1_.cols & 3) == 0)
             {
                 const int vcols = src1_.cols >> 2;
 
-                vfunc4(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::subMat_v4(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                  stream);
 
                 return;
             }
-
-            if (vfunc2 != 0 && (src1_.cols & 1) == 0)
+            else if (sdepth == CV_16U && (src1_.cols & 1) == 0)
             {
                 const int vcols = src1_.cols >> 1;
 
-                vfunc2(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::subMat_v2(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                  PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                  stream);
 
                 return;
             }
@@ -811,7 +714,7 @@ void cv::gpu::subtract(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cons
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
     func(src1_, src2_, dst_, mask, stream);
 }
@@ -819,78 +722,76 @@ void cv::gpu::subtract(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cons
 namespace arithm
 {
     template <typename T, typename S, typename D>
-    void subScalar(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    void subScalar(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::subtract(const GpuMat& src, const Scalar& sc, GpuMat& dst, const GpuMat& mask, int dtype, Stream& s)
+static void subScalar(const GpuMat& src, Scalar val, bool inv, GpuMat& dst, const GpuMat& mask, double, Stream& _stream, int)
 {
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    typedef void (*func_t)(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            subScalar<unsigned char, float, unsigned char>,
-            subScalar<unsigned char, float, signed char>,
-            subScalar<unsigned char, float, unsigned short>,
-            subScalar<unsigned char, float, short>,
-            subScalar<unsigned char, float, int>,
-            subScalar<unsigned char, float, float>,
-            subScalar<unsigned char, double, double>
+            arithm::subScalar<unsigned char, float, unsigned char>,
+            arithm::subScalar<unsigned char, float, signed char>,
+            arithm::subScalar<unsigned char, float, unsigned short>,
+            arithm::subScalar<unsigned char, float, short>,
+            arithm::subScalar<unsigned char, float, int>,
+            arithm::subScalar<unsigned char, float, float>,
+            arithm::subScalar<unsigned char, double, double>
         },
         {
-            subScalar<signed char, float, unsigned char>,
-            subScalar<signed char, float, signed char>,
-            subScalar<signed char, float, unsigned short>,
-            subScalar<signed char, float, short>,
-            subScalar<signed char, float, int>,
-            subScalar<signed char, float, float>,
-            subScalar<signed char, double, double>
+            arithm::subScalar<signed char, float, unsigned char>,
+            arithm::subScalar<signed char, float, signed char>,
+            arithm::subScalar<signed char, float, unsigned short>,
+            arithm::subScalar<signed char, float, short>,
+            arithm::subScalar<signed char, float, int>,
+            arithm::subScalar<signed char, float, float>,
+            arithm::subScalar<signed char, double, double>
         },
         {
-            0 /*subScalar<unsigned short, float, unsigned char>*/,
-            0 /*subScalar<unsigned short, float, signed char>*/,
-            subScalar<unsigned short, float, unsigned short>,
-            subScalar<unsigned short, float, short>,
-            subScalar<unsigned short, float, int>,
-            subScalar<unsigned short, float, float>,
-            subScalar<unsigned short, double, double>
+            0 /*arithm::subScalar<unsigned short, float, unsigned char>*/,
+            0 /*arithm::subScalar<unsigned short, float, signed char>*/,
+            arithm::subScalar<unsigned short, float, unsigned short>,
+            arithm::subScalar<unsigned short, float, short>,
+            arithm::subScalar<unsigned short, float, int>,
+            arithm::subScalar<unsigned short, float, float>,
+            arithm::subScalar<unsigned short, double, double>
         },
         {
-            0 /*subScalar<short, float, unsigned char>*/,
-            0 /*subScalar<short, float, signed char>*/,
-            subScalar<short, float, unsigned short>,
-            subScalar<short, float, short>,
-            subScalar<short, float, int>,
-            subScalar<short, float, float>,
-            subScalar<short, double, double>
+            0 /*arithm::subScalar<short, float, unsigned char>*/,
+            0 /*arithm::subScalar<short, float, signed char>*/,
+            arithm::subScalar<short, float, unsigned short>,
+            arithm::subScalar<short, float, short>,
+            arithm::subScalar<short, float, int>,
+            arithm::subScalar<short, float, float>,
+            arithm::subScalar<short, double, double>
         },
         {
-            0 /*subScalar<int, float, unsigned char>*/,
-            0 /*subScalar<int, float, signed char>*/,
-            0 /*subScalar<int, float, unsigned short>*/,
-            0 /*subScalar<int, float, short>*/,
-            subScalar<int, float, int>,
-            subScalar<int, float, float>,
-            subScalar<int, double, double>
+            0 /*arithm::subScalar<int, float, unsigned char>*/,
+            0 /*arithm::subScalar<int, float, signed char>*/,
+            0 /*arithm::subScalar<int, float, unsigned short>*/,
+            0 /*arithm::subScalar<int, float, short>*/,
+            arithm::subScalar<int, float, int>,
+            arithm::subScalar<int, float, float>,
+            arithm::subScalar<int, double, double>
         },
         {
-            0 /*subScalar<float, float, unsigned char>*/,
-            0 /*subScalar<float, float, signed char>*/,
-            0 /*subScalar<float, float, unsigned short>*/,
-            0 /*subScalar<float, float, short>*/,
-            0 /*subScalar<float, float, int>*/,
-            subScalar<float, float, float>,
-            subScalar<float, double, double>
+            0 /*arithm::subScalar<float, float, unsigned char>*/,
+            0 /*arithm::subScalar<float, float, signed char>*/,
+            0 /*arithm::subScalar<float, float, unsigned short>*/,
+            0 /*arithm::subScalar<float, float, short>*/,
+            0 /*arithm::subScalar<float, float, int>*/,
+            arithm::subScalar<float, float, float>,
+            arithm::subScalar<float, double, double>
         },
         {
-            0 /*subScalar<double, double, unsigned char>*/,
-            0 /*subScalar<double, double, signed char>*/,
-            0 /*subScalar<double, double, unsigned short>*/,
-            0 /*subScalar<double, double, short>*/,
-            0 /*subScalar<double, double, int>*/,
-            0 /*subScalar<double, double, float>*/,
-            subScalar<double, double, double>
+            0 /*arithm::subScalar<double, double, unsigned char>*/,
+            0 /*arithm::subScalar<double, double, signed char>*/,
+            0 /*arithm::subScalar<double, double, unsigned short>*/,
+            0 /*arithm::subScalar<double, double, short>*/,
+            0 /*arithm::subScalar<double, double, int>*/,
+            0 /*arithm::subScalar<double, double, float>*/,
+            arithm::subScalar<double, double, double>
         }
     };
 
@@ -906,31 +807,16 @@ void cv::gpu::subtract(const GpuMat& src, const Scalar& sc, GpuMat& dst, const G
         {0                                                    , 0                                                     , 0                                                    , 0                                                    }
     };
 
-    if (dtype < 0)
-        dtype = src.depth();
-
     const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( cn <= 4 );
-    CV_Assert( mask.empty() || (cn == 1 && mask.size() == src.size() && mask.type() == CV_8U) );
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     const npp_func_t npp_func = npp_funcs[sdepth][cn - 1];
-    if (ddepth == sdepth && cn > 1 && npp_func != 0)
+    if (ddepth == sdepth && cn > 1 && npp_func != 0 && !inv)
     {
-        npp_func(src, sc, dst, stream);
+        npp_func(src, val, dst, stream);
         return;
     }
 
@@ -939,9 +825,14 @@ void cv::gpu::subtract(const GpuMat& src, const Scalar& sc, GpuMat& dst, const G
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
-    func(src, sc.val[0], dst, mask, stream);
+    func(src, val[0], inv, dst, mask, stream);
+}
+
+void cv::gpu::subtract(InputArray src1, InputArray src2, OutputArray dst, InputArray mask, int dtype, Stream& stream)
+{
+    arithm_op(src1, src2, dst, mask, 1.0, dtype, stream, subMat, subScalar);
 }
 
 ////////////////////////////////////////////////////////////////////////
@@ -957,127 +848,92 @@ namespace arithm
     void mulMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
 }
 
-void cv::gpu::multiply(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, double scale, int dtype, Stream& s)
+static void mulMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat&, double scale, Stream& _stream, int)
 {
-    using namespace arithm;
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    if (src1.type() == CV_8UC4 && src2.type() == CV_32FC1)
+    typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    static const func_t funcs[7][7] =
     {
-        CV_Assert( src1.size() == src2.size() );
-
-        dst.create(src1.size(), src1.type());
-
-        mulMat_8uc4_32f(src1, src2, dst, stream);
-    }
-    else if (src1.type() == CV_16SC4 && src2.type() == CV_32FC1)
-    {
-        CV_Assert( src1.size() == src2.size() );
-
-        dst.create(src1.size(), src1.type());
-
-        mulMat_16sc4_32f(src1, src2, dst, stream);
-    }
-    else
-    {
-        typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-        static const func_t funcs[7][7] =
         {
-            {
-                mulMat<unsigned char, float, unsigned char>,
-                mulMat<unsigned char, float, signed char>,
-                mulMat<unsigned char, float, unsigned short>,
-                mulMat<unsigned char, float, short>,
-                mulMat<unsigned char, float, int>,
-                mulMat<unsigned char, float, float>,
-                mulMat<unsigned char, double, double>
-            },
-            {
-                mulMat<signed char, float, unsigned char>,
-                mulMat<signed char, float, signed char>,
-                mulMat<signed char, float, unsigned short>,
-                mulMat<signed char, float, short>,
-                mulMat<signed char, float, int>,
-                mulMat<signed char, float, float>,
-                mulMat<signed char, double, double>
-            },
-            {
-                0 /*mulMat<unsigned short, float, unsigned char>*/,
-                0 /*mulMat<unsigned short, float, signed char>*/,
-                mulMat<unsigned short, float, unsigned short>,
-                mulMat<unsigned short, float, short>,
-                mulMat<unsigned short, float, int>,
-                mulMat<unsigned short, float, float>,
-                mulMat<unsigned short, double, double>
-            },
-            {
-                0 /*mulMat<short, float, unsigned char>*/,
-                0 /*mulMat<short, float, signed char>*/,
-                mulMat<short, float, unsigned short>,
-                mulMat<short, float, short>,
-                mulMat<short, float, int>,
-                mulMat<short, float, float>,
-                mulMat<short, double, double>
-            },
-            {
-                0 /*mulMat<int, float, unsigned char>*/,
-                0 /*mulMat<int, float, signed char>*/,
-                0 /*mulMat<int, float, unsigned short>*/,
-                0 /*mulMat<int, float, short>*/,
-                mulMat<int, float, int>,
-                mulMat<int, float, float>,
-                mulMat<int, double, double>
-            },
-            {
-                0 /*mulMat<float, float, unsigned char>*/,
-                0 /*mulMat<float, float, signed char>*/,
-                0 /*mulMat<float, float, unsigned short>*/,
-                0 /*mulMat<float, float, short>*/,
-                0 /*mulMat<float, float, int>*/,
-                mulMat<float, float, float>,
-                mulMat<float, double, double>
-            },
-            {
-                0 /*mulMat<double, double, unsigned char>*/,
-                0 /*mulMat<double, double, signed char>*/,
-                0 /*mulMat<double, double, unsigned short>*/,
-                0 /*mulMat<double, double, short>*/,
-                0 /*mulMat<double, double, int>*/,
-                0 /*mulMat<double, double, float>*/,
-                mulMat<double, double, double>
-            }
-        };
-
-        if (dtype < 0)
-            dtype = src1.depth();
-
-        const int sdepth = src1.depth();
-        const int ddepth = CV_MAT_DEPTH(dtype);
-        const int cn = src1.channels();
-
-        CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-        CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-
-        if (sdepth == CV_64F || ddepth == CV_64F)
+            arithm::mulMat<unsigned char, float, unsigned char>,
+            arithm::mulMat<unsigned char, float, signed char>,
+            arithm::mulMat<unsigned char, float, unsigned short>,
+            arithm::mulMat<unsigned char, float, short>,
+            arithm::mulMat<unsigned char, float, int>,
+            arithm::mulMat<unsigned char, float, float>,
+            arithm::mulMat<unsigned char, double, double>
+        },
         {
-            if (!deviceSupports(NATIVE_DOUBLE))
-                CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            arithm::mulMat<signed char, float, unsigned char>,
+            arithm::mulMat<signed char, float, signed char>,
+            arithm::mulMat<signed char, float, unsigned short>,
+            arithm::mulMat<signed char, float, short>,
+            arithm::mulMat<signed char, float, int>,
+            arithm::mulMat<signed char, float, float>,
+            arithm::mulMat<signed char, double, double>
+        },
+        {
+            0 /*arithm::mulMat<unsigned short, float, unsigned char>*/,
+            0 /*arithm::mulMat<unsigned short, float, signed char>*/,
+            arithm::mulMat<unsigned short, float, unsigned short>,
+            arithm::mulMat<unsigned short, float, short>,
+            arithm::mulMat<unsigned short, float, int>,
+            arithm::mulMat<unsigned short, float, float>,
+            arithm::mulMat<unsigned short, double, double>
+        },
+        {
+            0 /*arithm::mulMat<short, float, unsigned char>*/,
+            0 /*arithm::mulMat<short, float, signed char>*/,
+            arithm::mulMat<short, float, unsigned short>,
+            arithm::mulMat<short, float, short>,
+            arithm::mulMat<short, float, int>,
+            arithm::mulMat<short, float, float>,
+            arithm::mulMat<short, double, double>
+        },
+        {
+            0 /*arithm::mulMat<int, float, unsigned char>*/,
+            0 /*arithm::mulMat<int, float, signed char>*/,
+            0 /*arithm::mulMat<int, float, unsigned short>*/,
+            0 /*arithm::mulMat<int, float, short>*/,
+            arithm::mulMat<int, float, int>,
+            arithm::mulMat<int, float, float>,
+            arithm::mulMat<int, double, double>
+        },
+        {
+            0 /*arithm::mulMat<float, float, unsigned char>*/,
+            0 /*arithm::mulMat<float, float, signed char>*/,
+            0 /*arithm::mulMat<float, float, unsigned short>*/,
+            0 /*arithm::mulMat<float, float, short>*/,
+            0 /*arithm::mulMat<float, float, int>*/,
+            arithm::mulMat<float, float, float>,
+            arithm::mulMat<float, double, double>
+        },
+        {
+            0 /*arithm::mulMat<double, double, unsigned char>*/,
+            0 /*arithm::mulMat<double, double, signed char>*/,
+            0 /*arithm::mulMat<double, double, unsigned short>*/,
+            0 /*arithm::mulMat<double, double, short>*/,
+            0 /*arithm::mulMat<double, double, int>*/,
+            0 /*arithm::mulMat<double, double, float>*/,
+            arithm::mulMat<double, double, double>
         }
+    };
 
-        dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
+    const int sdepth = src1.depth();
+    const int ddepth = dst.depth();
+    const int cn = src1.channels();
 
-        PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
-        PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
-        PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
-        const func_t func = funcs[sdepth][ddepth];
+    PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
+    PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
+    PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-        if (!func)
-            CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+    const func_t func = funcs[sdepth][ddepth];
 
-        func(src1_, src2_, dst_, scale, stream);
-    }
+    if (!func)
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
+
+    func(src1_, src2_, dst_, scale, stream);
 }
 
 namespace arithm
@@ -1086,75 +942,73 @@ namespace arithm
     void mulScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::multiply(const GpuMat& src, const Scalar& sc, GpuMat& dst, double scale, int dtype, Stream& s)
+static void mulScalar(const GpuMat& src, Scalar val, bool, GpuMat& dst, const GpuMat&, double scale, Stream& _stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            mulScalar<unsigned char, float, unsigned char>,
-            mulScalar<unsigned char, float, signed char>,
-            mulScalar<unsigned char, float, unsigned short>,
-            mulScalar<unsigned char, float, short>,
-            mulScalar<unsigned char, float, int>,
-            mulScalar<unsigned char, float, float>,
-            mulScalar<unsigned char, double, double>
+            arithm::mulScalar<unsigned char, float, unsigned char>,
+            arithm::mulScalar<unsigned char, float, signed char>,
+            arithm::mulScalar<unsigned char, float, unsigned short>,
+            arithm::mulScalar<unsigned char, float, short>,
+            arithm::mulScalar<unsigned char, float, int>,
+            arithm::mulScalar<unsigned char, float, float>,
+            arithm::mulScalar<unsigned char, double, double>
         },
         {
-            mulScalar<signed char, float, unsigned char>,
-            mulScalar<signed char, float, signed char>,
-            mulScalar<signed char, float, unsigned short>,
-            mulScalar<signed char, float, short>,
-            mulScalar<signed char, float, int>,
-            mulScalar<signed char, float, float>,
-            mulScalar<signed char, double, double>
+            arithm::mulScalar<signed char, float, unsigned char>,
+            arithm::mulScalar<signed char, float, signed char>,
+            arithm::mulScalar<signed char, float, unsigned short>,
+            arithm::mulScalar<signed char, float, short>,
+            arithm::mulScalar<signed char, float, int>,
+            arithm::mulScalar<signed char, float, float>,
+            arithm::mulScalar<signed char, double, double>
         },
         {
-            0 /*mulScalar<unsigned short, float, unsigned char>*/,
-            0 /*mulScalar<unsigned short, float, signed char>*/,
-            mulScalar<unsigned short, float, unsigned short>,
-            mulScalar<unsigned short, float, short>,
-            mulScalar<unsigned short, float, int>,
-            mulScalar<unsigned short, float, float>,
-            mulScalar<unsigned short, double, double>
+            0 /*arithm::mulScalar<unsigned short, float, unsigned char>*/,
+            0 /*arithm::mulScalar<unsigned short, float, signed char>*/,
+            arithm::mulScalar<unsigned short, float, unsigned short>,
+            arithm::mulScalar<unsigned short, float, short>,
+            arithm::mulScalar<unsigned short, float, int>,
+            arithm::mulScalar<unsigned short, float, float>,
+            arithm::mulScalar<unsigned short, double, double>
         },
         {
-            0 /*mulScalar<short, float, unsigned char>*/,
-            0 /*mulScalar<short, float, signed char>*/,
-            mulScalar<short, float, unsigned short>,
-            mulScalar<short, float, short>,
-            mulScalar<short, float, int>,
-            mulScalar<short, float, float>,
-            mulScalar<short, double, double>
+            0 /*arithm::mulScalar<short, float, unsigned char>*/,
+            0 /*arithm::mulScalar<short, float, signed char>*/,
+            arithm::mulScalar<short, float, unsigned short>,
+            arithm::mulScalar<short, float, short>,
+            arithm::mulScalar<short, float, int>,
+            arithm::mulScalar<short, float, float>,
+            arithm::mulScalar<short, double, double>
         },
         {
-            0 /*mulScalar<int, float, unsigned char>*/,
-            0 /*mulScalar<int, float, signed char>*/,
-            0 /*mulScalar<int, float, unsigned short>*/,
-            0 /*mulScalar<int, float, short>*/,
-            mulScalar<int, float, int>,
-            mulScalar<int, float, float>,
-            mulScalar<int, double, double>
+            0 /*arithm::mulScalar<int, float, unsigned char>*/,
+            0 /*arithm::mulScalar<int, float, signed char>*/,
+            0 /*arithm::mulScalar<int, float, unsigned short>*/,
+            0 /*arithm::mulScalar<int, float, short>*/,
+            arithm::mulScalar<int, float, int>,
+            arithm::mulScalar<int, float, float>,
+            arithm::mulScalar<int, double, double>
         },
         {
-            0 /*mulScalar<float, float, unsigned char>*/,
-            0 /*mulScalar<float, float, signed char>*/,
-            0 /*mulScalar<float, float, unsigned short>*/,
-            0 /*mulScalar<float, float, short>*/,
-            0 /*mulScalar<float, float, int>*/,
-            mulScalar<float, float, float>,
-            mulScalar<float, double, double>
+            0 /*arithm::mulScalar<float, float, unsigned char>*/,
+            0 /*arithm::mulScalar<float, float, signed char>*/,
+            0 /*arithm::mulScalar<float, float, unsigned short>*/,
+            0 /*arithm::mulScalar<float, float, short>*/,
+            0 /*arithm::mulScalar<float, float, int>*/,
+            arithm::mulScalar<float, float, float>,
+            arithm::mulScalar<float, double, double>
         },
         {
-            0 /*mulScalar<double, double, unsigned char>*/,
-            0 /*mulScalar<double, double, signed char>*/,
-            0 /*mulScalar<double, double, unsigned short>*/,
-            0 /*mulScalar<double, double, short>*/,
-            0 /*mulScalar<double, double, int>*/,
-            0 /*mulScalar<double, double, float>*/,
-            mulScalar<double, double, double>
+            0 /*arithm::mulScalar<double, double, unsigned char>*/,
+            0 /*arithm::mulScalar<double, double, signed char>*/,
+            0 /*arithm::mulScalar<double, double, unsigned short>*/,
+            0 /*arithm::mulScalar<double, double, short>*/,
+            0 /*arithm::mulScalar<double, double, int>*/,
+            0 /*arithm::mulScalar<double, double, float>*/,
+            arithm::mulScalar<double, double, double>
         }
     };
 
@@ -1170,32 +1024,21 @@ void cv::gpu::multiply(const GpuMat& src, const Scalar& sc, GpuMat& dst, double
         {0                                                    , 0, 0                                                    , 0                                                    }
     };
 
-    if (dtype < 0)
-        dtype = src.depth();
-
     const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( cn <= 4 );
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    const Scalar nsc(sc.val[0] * scale, sc.val[1] * scale, sc.val[2] * scale, sc.val[3] * scale);
+    val[0] *= scale;
+    val[1] *= scale;
+    val[2] *= scale;
+    val[3] *= scale;
 
     const npp_func_t npp_func = npp_funcs[sdepth][cn - 1];
     if (ddepth == sdepth && cn > 1 && npp_func != 0)
     {
-        npp_func(src, nsc, dst, stream);
+        npp_func(src, val, dst, stream);
         return;
     }
 
@@ -1204,9 +1047,41 @@ void cv::gpu::multiply(const GpuMat& src, const Scalar& sc, GpuMat& dst, double
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
-    func(src, nsc.val[0], dst, stream);
+    func(src, val[0], dst, stream);
+}
+
+void cv::gpu::multiply(InputArray _src1, InputArray _src2, OutputArray _dst, double scale, int dtype, Stream& stream)
+{
+    if (_src1.type() == CV_8UC4 && _src2.type() == CV_32FC1)
+    {
+        GpuMat src1 = _src1.getGpuMat();
+        GpuMat src2 = _src2.getGpuMat();
+
+        CV_Assert( src1.size() == src2.size() );
+
+        _dst.create(src1.size(), src1.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        arithm::mulMat_8uc4_32f(src1, src2, dst, StreamAccessor::getStream(stream));
+    }
+    else if (_src1.type() == CV_16SC4 && _src2.type() == CV_32FC1)
+    {
+        GpuMat src1 = _src1.getGpuMat();
+        GpuMat src2 = _src2.getGpuMat();
+
+        CV_Assert( src1.size() == src2.size() );
+
+        _dst.create(src1.size(), src1.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        arithm::mulMat_16sc4_32f(src1, src2, dst, StreamAccessor::getStream(stream));
+    }
+    else
+    {
+        arithm_op(_src1, _src2, _dst, GpuMat(), scale, dtype, stream, mulMat, mulScalar);
+    }
 }
 
 ////////////////////////////////////////////////////////////////////////
@@ -1222,204 +1097,167 @@ namespace arithm
     void divMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
 }
 
-void cv::gpu::divide(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, double scale, int dtype, Stream& s)
+static void divMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat&, double scale, Stream& _stream, int)
 {
-    using namespace arithm;
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    if (src1.type() == CV_8UC4 && src2.type() == CV_32FC1)
+    typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
+    static const func_t funcs[7][7] =
     {
-        CV_Assert( src1.size() == src2.size() );
-
-        dst.create(src1.size(), src1.type());
-
-        divMat_8uc4_32f(src1, src2, dst, stream);
-    }
-    else if (src1.type() == CV_16SC4 && src2.type() == CV_32FC1)
-    {
-        CV_Assert( src1.size() == src2.size() );
-
-        dst.create(src1.size(), src1.type());
-
-        divMat_16sc4_32f(src1, src2, dst, stream);
-    }
-    else
-    {
-        typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, double scale, cudaStream_t stream);
-        static const func_t funcs[7][7] =
         {
-            {
-                divMat<unsigned char, float, unsigned char>,
-                divMat<unsigned char, float, signed char>,
-                divMat<unsigned char, float, unsigned short>,
-                divMat<unsigned char, float, short>,
-                divMat<unsigned char, float, int>,
-                divMat<unsigned char, float, float>,
-                divMat<unsigned char, double, double>
-            },
-            {
-                divMat<signed char, float, unsigned char>,
-                divMat<signed char, float, signed char>,
-                divMat<signed char, float, unsigned short>,
-                divMat<signed char, float, short>,
-                divMat<signed char, float, int>,
-                divMat<signed char, float, float>,
-                divMat<signed char, double, double>
-            },
-            {
-                0 /*divMat<unsigned short, float, unsigned char>*/,
-                0 /*divMat<unsigned short, float, signed char>*/,
-                divMat<unsigned short, float, unsigned short>,
-                divMat<unsigned short, float, short>,
-                divMat<unsigned short, float, int>,
-                divMat<unsigned short, float, float>,
-                divMat<unsigned short, double, double>
-            },
-            {
-                0 /*divMat<short, float, unsigned char>*/,
-                0 /*divMat<short, float, signed char>*/,
-                divMat<short, float, unsigned short>,
-                divMat<short, float, short>,
-                divMat<short, float, int>,
-                divMat<short, float, float>,
-                divMat<short, double, double>
-            },
-            {
-                0 /*divMat<int, float, unsigned char>*/,
-                0 /*divMat<int, float, signed char>*/,
-                0 /*divMat<int, float, unsigned short>*/,
-                0 /*divMat<int, float, short>*/,
-                divMat<int, float, int>,
-                divMat<int, float, float>,
-                divMat<int, double, double>
-            },
-            {
-                0 /*divMat<float, float, unsigned char>*/,
-                0 /*divMat<float, float, signed char>*/,
-                0 /*divMat<float, float, unsigned short>*/,
-                0 /*divMat<float, float, short>*/,
-                0 /*divMat<float, float, int>*/,
-                divMat<float, float, float>,
-                divMat<float, double, double>
-            },
-            {
-                0 /*divMat<double, double, unsigned char>*/,
-                0 /*divMat<double, double, signed char>*/,
-                0 /*divMat<double, double, unsigned short>*/,
-                0 /*divMat<double, double, short>*/,
-                0 /*divMat<double, double, int>*/,
-                0 /*divMat<double, double, float>*/,
-                divMat<double, double, double>
-            }
-        };
-
-        if (dtype < 0)
-            dtype = src1.depth();
-
-        const int sdepth = src1.depth();
-        const int ddepth = CV_MAT_DEPTH(dtype);
-        const int cn = src1.channels();
-
-        CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-        CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-
-        if (sdepth == CV_64F || ddepth == CV_64F)
+            arithm::divMat<unsigned char, float, unsigned char>,
+            arithm::divMat<unsigned char, float, signed char>,
+            arithm::divMat<unsigned char, float, unsigned short>,
+            arithm::divMat<unsigned char, float, short>,
+            arithm::divMat<unsigned char, float, int>,
+            arithm::divMat<unsigned char, float, float>,
+            arithm::divMat<unsigned char, double, double>
+        },
         {
-            if (!deviceSupports(NATIVE_DOUBLE))
-                CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            arithm::divMat<signed char, float, unsigned char>,
+            arithm::divMat<signed char, float, signed char>,
+            arithm::divMat<signed char, float, unsigned short>,
+            arithm::divMat<signed char, float, short>,
+            arithm::divMat<signed char, float, int>,
+            arithm::divMat<signed char, float, float>,
+            arithm::divMat<signed char, double, double>
+        },
+        {
+            0 /*arithm::divMat<unsigned short, float, unsigned char>*/,
+            0 /*arithm::divMat<unsigned short, float, signed char>*/,
+            arithm::divMat<unsigned short, float, unsigned short>,
+            arithm::divMat<unsigned short, float, short>,
+            arithm::divMat<unsigned short, float, int>,
+            arithm::divMat<unsigned short, float, float>,
+            arithm::divMat<unsigned short, double, double>
+        },
+        {
+            0 /*arithm::divMat<short, float, unsigned char>*/,
+            0 /*arithm::divMat<short, float, signed char>*/,
+            arithm::divMat<short, float, unsigned short>,
+            arithm::divMat<short, float, short>,
+            arithm::divMat<short, float, int>,
+            arithm::divMat<short, float, float>,
+            arithm::divMat<short, double, double>
+        },
+        {
+            0 /*arithm::divMat<int, float, unsigned char>*/,
+            0 /*arithm::divMat<int, float, signed char>*/,
+            0 /*arithm::divMat<int, float, unsigned short>*/,
+            0 /*arithm::divMat<int, float, short>*/,
+            arithm::divMat<int, float, int>,
+            arithm::divMat<int, float, float>,
+            arithm::divMat<int, double, double>
+        },
+        {
+            0 /*arithm::divMat<float, float, unsigned char>*/,
+            0 /*arithm::divMat<float, float, signed char>*/,
+            0 /*arithm::divMat<float, float, unsigned short>*/,
+            0 /*arithm::divMat<float, float, short>*/,
+            0 /*arithm::divMat<float, float, int>*/,
+            arithm::divMat<float, float, float>,
+            arithm::divMat<float, double, double>
+        },
+        {
+            0 /*arithm::divMat<double, double, unsigned char>*/,
+            0 /*arithm::divMat<double, double, signed char>*/,
+            0 /*arithm::divMat<double, double, unsigned short>*/,
+            0 /*arithm::divMat<double, double, short>*/,
+            0 /*arithm::divMat<double, double, int>*/,
+            0 /*arithm::divMat<double, double, float>*/,
+            arithm::divMat<double, double, double>
         }
+    };
 
-        dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
+    const int sdepth = src1.depth();
+    const int ddepth = dst.depth();
+    const int cn = src1.channels();
 
-        PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
-        PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
-        PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
-        const func_t func = funcs[sdepth][ddepth];
+    PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
+    PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
+    PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-        if (!func)
-            CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+    const func_t func = funcs[sdepth][ddepth];
 
-        func(src1_, src2_, dst_, scale, stream);
-    }
+    if (!func)
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
+
+    func(src1_, src2_, dst_, scale, stream);
 }
 
 namespace arithm
 {
     template <typename T, typename S, typename D>
-    void divScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    void divScalar(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::divide(const GpuMat& src, const Scalar& sc, GpuMat& dst, double scale, int dtype, Stream& s)
+static void divScalar(const GpuMat& src, Scalar val, bool inv, GpuMat& dst, const GpuMat&, double scale, Stream& _stream, int)
 {
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
+    typedef void (*func_t)(PtrStepSzb src1, double val, bool inv, PtrStepSzb dst, cudaStream_t stream);
     static const func_t funcs[7][7] =
     {
         {
-            divScalar<unsigned char, float, unsigned char>,
-            divScalar<unsigned char, float, signed char>,
-            divScalar<unsigned char, float, unsigned short>,
-            divScalar<unsigned char, float, short>,
-            divScalar<unsigned char, float, int>,
-            divScalar<unsigned char, float, float>,
-            divScalar<unsigned char, double, double>
+            arithm::divScalar<unsigned char, float, unsigned char>,
+            arithm::divScalar<unsigned char, float, signed char>,
+            arithm::divScalar<unsigned char, float, unsigned short>,
+            arithm::divScalar<unsigned char, float, short>,
+            arithm::divScalar<unsigned char, float, int>,
+            arithm::divScalar<unsigned char, float, float>,
+            arithm::divScalar<unsigned char, double, double>
         },
         {
-            divScalar<signed char, float, unsigned char>,
-            divScalar<signed char, float, signed char>,
-            divScalar<signed char, float, unsigned short>,
-            divScalar<signed char, float, short>,
-            divScalar<signed char, float, int>,
-            divScalar<signed char, float, float>,
-            divScalar<signed char, double, double>
+            arithm::divScalar<signed char, float, unsigned char>,
+            arithm::divScalar<signed char, float, signed char>,
+            arithm::divScalar<signed char, float, unsigned short>,
+            arithm::divScalar<signed char, float, short>,
+            arithm::divScalar<signed char, float, int>,
+            arithm::divScalar<signed char, float, float>,
+            arithm::divScalar<signed char, double, double>
         },
         {
-            0 /*divScalar<unsigned short, float, unsigned char>*/,
-            0 /*divScalar<unsigned short, float, signed char>*/,
-            divScalar<unsigned short, float, unsigned short>,
-            divScalar<unsigned short, float, short>,
-            divScalar<unsigned short, float, int>,
-            divScalar<unsigned short, float, float>,
-            divScalar<unsigned short, double, double>
+            0 /*arithm::divScalar<unsigned short, float, unsigned char>*/,
+            0 /*arithm::divScalar<unsigned short, float, signed char>*/,
+            arithm::divScalar<unsigned short, float, unsigned short>,
+            arithm::divScalar<unsigned short, float, short>,
+            arithm::divScalar<unsigned short, float, int>,
+            arithm::divScalar<unsigned short, float, float>,
+            arithm::divScalar<unsigned short, double, double>
         },
         {
-            0 /*divScalar<short, float, unsigned char>*/,
-            0 /*divScalar<short, float, signed char>*/,
-            divScalar<short, float, unsigned short>,
-            divScalar<short, float, short>,
-            divScalar<short, float, int>,
-            divScalar<short, float, float>,
-            divScalar<short, double, double>
+            0 /*arithm::divScalar<short, float, unsigned char>*/,
+            0 /*arithm::divScalar<short, float, signed char>*/,
+            arithm::divScalar<short, float, unsigned short>,
+            arithm::divScalar<short, float, short>,
+            arithm::divScalar<short, float, int>,
+            arithm::divScalar<short, float, float>,
+            arithm::divScalar<short, double, double>
         },
         {
-            0 /*divScalar<int, float, unsigned char>*/,
-            0 /*divScalar<int, float, signed char>*/,
-            0 /*divScalar<int, float, unsigned short>*/,
-            0 /*divScalar<int, float, short>*/,
-            divScalar<int, float, int>,
-            divScalar<int, float, float>,
-            divScalar<int, double, double>
+            0 /*arithm::divScalar<int, float, unsigned char>*/,
+            0 /*arithm::divScalar<int, float, signed char>*/,
+            0 /*arithm::divScalar<int, float, unsigned short>*/,
+            0 /*arithm::divScalar<int, float, short>*/,
+            arithm::divScalar<int, float, int>,
+            arithm::divScalar<int, float, float>,
+            arithm::divScalar<int, double, double>
         },
         {
-            0 /*divScalar<float, float, unsigned char>*/,
-            0 /*divScalar<float, float, signed char>*/,
-            0 /*divScalar<float, float, unsigned short>*/,
-            0 /*divScalar<float, float, short>*/,
-            0 /*divScalar<float, float, int>*/,
-            divScalar<float, float, float>,
-            divScalar<float, double, double>
+            0 /*arithm::divScalar<float, float, unsigned char>*/,
+            0 /*arithm::divScalar<float, float, signed char>*/,
+            0 /*arithm::divScalar<float, float, unsigned short>*/,
+            0 /*arithm::divScalar<float, float, short>*/,
+            0 /*arithm::divScalar<float, float, int>*/,
+            arithm::divScalar<float, float, float>,
+            arithm::divScalar<float, double, double>
         },
         {
-            0 /*divScalar<double, double, unsigned char>*/,
-            0 /*divScalar<double, double, signed char>*/,
-            0 /*divScalar<double, double, unsigned short>*/,
-            0 /*divScalar<double, double, short>*/,
-            0 /*divScalar<double, double, int>*/,
-            0 /*divScalar<double, double, float>*/,
-            divScalar<double, double, double>
+            0 /*arithm::divScalar<double, double, unsigned char>*/,
+            0 /*arithm::divScalar<double, double, signed char>*/,
+            0 /*arithm::divScalar<double, double, unsigned short>*/,
+            0 /*arithm::divScalar<double, double, short>*/,
+            0 /*arithm::divScalar<double, double, int>*/,
+            0 /*arithm::divScalar<double, double, float>*/,
+            arithm::divScalar<double, double, double>
         }
     };
 
@@ -1435,32 +1273,31 @@ void cv::gpu::divide(const GpuMat& src, const Scalar& sc, GpuMat& dst, double sc
         {0                                                    , 0, 0                                                    , 0                                                    }
     };
 
-    if (dtype < 0)
-        dtype = src.depth();
-
     const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
+    const int ddepth = dst.depth();
     const int cn = src.channels();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( cn <= 4 );
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
-    if (sdepth == CV_64F || ddepth == CV_64F)
+    if (inv)
     {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+        val[0] *= scale;
+        val[1] *= scale;
+        val[2] *= scale;
+        val[3] *= scale;
+    }
+    else
+    {
+        val[0] /= scale;
+        val[1] /= scale;
+        val[2] /= scale;
+        val[3] /= scale;
     }
 
-    dst.create(src.size(), CV_MAKE_TYPE(ddepth, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    const Scalar nsc(sc.val[0] / scale, sc.val[1] / scale, sc.val[2] / scale, sc.val[3] / scale);
-
     const npp_func_t npp_func = npp_funcs[sdepth][cn - 1];
-    if (ddepth == sdepth && cn > 1 && npp_func != 0)
+    if (ddepth == sdepth && cn > 1 && npp_func != 0 && !inv)
     {
-        npp_func(src, nsc, dst, stream);
+        npp_func(src, val, dst, stream);
         return;
     }
 
@@ -1469,115 +1306,41 @@ void cv::gpu::divide(const GpuMat& src, const Scalar& sc, GpuMat& dst, double sc
     const func_t func = funcs[sdepth][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
-    func(src, nsc.val[0], dst, stream);
+    func(src, val[0], inv, dst, stream);
 }
 
-namespace arithm
+void cv::gpu::divide(InputArray _src1, InputArray _src2, OutputArray _dst, double scale, int dtype, Stream& stream)
 {
-    template <typename T, typename S, typename D>
-    void divInv(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-}
-
-void cv::gpu::divide(double scale, const GpuMat& src, GpuMat& dst, int dtype, Stream& s)
-{
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[7][7] =
+    if (_src1.type() == CV_8UC4 && _src2.type() == CV_32FC1)
     {
-        {
-            divInv<unsigned char, float, unsigned char>,
-            divInv<unsigned char, float, signed char>,
-            divInv<unsigned char, float, unsigned short>,
-            divInv<unsigned char, float, short>,
-            divInv<unsigned char, float, int>,
-            divInv<unsigned char, float, float>,
-            divInv<unsigned char, double, double>
-        },
-        {
-            divInv<signed char, float, unsigned char>,
-            divInv<signed char, float, signed char>,
-            divInv<signed char, float, unsigned short>,
-            divInv<signed char, float, short>,
-            divInv<signed char, float, int>,
-            divInv<signed char, float, float>,
-            divInv<signed char, double, double>
-        },
-        {
-            0 /*divInv<unsigned short, float, unsigned char>*/,
-            0 /*divInv<unsigned short, float, signed char>*/,
-            divInv<unsigned short, float, unsigned short>,
-            divInv<unsigned short, float, short>,
-            divInv<unsigned short, float, int>,
-            divInv<unsigned short, float, float>,
-            divInv<unsigned short, double, double>
-        },
-        {
-            0 /*divInv<short, float, unsigned char>*/,
-            0 /*divInv<short, float, signed char>*/,
-            divInv<short, float, unsigned short>,
-            divInv<short, float, short>,
-            divInv<short, float, int>,
-            divInv<short, float, float>,
-            divInv<short, double, double>
-        },
-        {
-            0 /*divInv<int, float, unsigned char>*/,
-            0 /*divInv<int, float, signed char>*/,
-            0 /*divInv<int, float, unsigned short>*/,
-            0 /*divInv<int, float, short>*/,
-            divInv<int, float, int>,
-            divInv<int, float, float>,
-            divInv<int, double, double>
-        },
-        {
-            0 /*divInv<float, float, unsigned char>*/,
-            0 /*divInv<float, float, signed char>*/,
-            0 /*divInv<float, float, unsigned short>*/,
-            0 /*divInv<float, float, short>*/,
-            0 /*divInv<float, float, int>*/,
-            divInv<float, float, float>,
-            divInv<float, double, double>
-        },
-        {
-            0 /*divInv<double, double, unsigned char>*/,
-            0 /*divInv<double, double, signed char>*/,
-            0 /*divInv<double, double, unsigned short>*/,
-            0 /*divInv<double, double, short>*/,
-            0 /*divInv<double, double, int>*/,
-            0 /*divInv<double, double, float>*/,
-            divInv<double, double, double>
-        }
-    };
+        GpuMat src1 = _src1.getGpuMat();
+        GpuMat src2 = _src2.getGpuMat();
 
-    if (dtype < 0)
-        dtype = src.depth();
+        CV_Assert( src1.size() == src2.size() );
 
-    const int sdepth = src.depth();
-    const int ddepth = CV_MAT_DEPTH(dtype);
-    const int cn = src.channels();
+        _dst.create(src1.size(), src1.type());
+        GpuMat dst = _dst.getGpuMat();
 
-    CV_Assert( sdepth <= CV_64F && ddepth <= CV_64F );
-    CV_Assert( cn == 1 );
-
-    if (sdepth == CV_64F || ddepth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+        arithm::divMat_8uc4_32f(src1, src2, dst, StreamAccessor::getStream(stream));
     }
+    else if (_src1.type() == CV_16SC4 && _src2.type() == CV_32FC1)
+    {
+        GpuMat src1 = _src1.getGpuMat();
+        GpuMat src2 = _src2.getGpuMat();
 
-    dst.create(src.size(), CV_MAKE_TYPE(ddepth, cn));
+        CV_Assert( src1.size() == src2.size() );
 
-    cudaStream_t stream = StreamAccessor::getStream(s);
+        _dst.create(src1.size(), src1.type());
+        GpuMat dst = _dst.getGpuMat();
 
-    const func_t func = funcs[sdepth][ddepth];
-
-    if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
-
-    func(src, scale, dst, stream);
+        arithm::divMat_16sc4_32f(src1, src2, dst, StreamAccessor::getStream(stream));
+    }
+    else
+    {
+        arithm_op(_src1, _src2, _dst, GpuMat(), scale, dtype, stream, divMat, divScalar);
+    }
 }
 
 //////////////////////////////////////////////////////////////////////////////
@@ -1585,67 +1348,37 @@ void cv::gpu::divide(double scale, const GpuMat& src, GpuMat& dst, int dtype, St
 
 namespace arithm
 {
-    template <typename T>
-    void vabsDiff4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-
-    template <typename T>
-    void vabsDiff2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    void absDiffMat_v4(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    void absDiffMat_v2(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
 
     template <typename T>
     void absDiffMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::absdiff(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& s)
+static void absDiffMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat&, double, Stream& _stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     static const func_t funcs[] =
     {
-        absDiffMat<unsigned char>,
-        absDiffMat<signed char>,
-        absDiffMat<unsigned short>,
-        absDiffMat<short>,
-        absDiffMat<int>,
-        absDiffMat<float>,
-        absDiffMat<double>
-    };
-    static const func_t vfuncs4[] =
-    {
-        vabsDiff4<unsigned int>,
-        vabsDiff4<int>,
-        0,
-        0
-    };
-    static const func_t vfuncs2[] =
-    {
-        0,
-        0,
-        vabsDiff2<unsigned int>,
-        vabsDiff2<int>
+        arithm::absDiffMat<unsigned char>,
+        arithm::absDiffMat<signed char>,
+        arithm::absDiffMat<unsigned short>,
+        arithm::absDiffMat<short>,
+        arithm::absDiffMat<int>,
+        arithm::absDiffMat<float>,
+        arithm::absDiffMat<double>
     };
 
     const int depth = src1.depth();
     const int cn = src1.channels();
 
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
     PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
     PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-    if (depth < CV_32S)
+    if (depth == CV_8U || depth == CV_16U)
     {
         const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
         const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
@@ -1653,31 +1386,27 @@ void cv::gpu::absdiff(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Strea
 
         const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
 
-        if (deviceSupports(FEATURE_SET_COMPUTE_20) && isAllAligned)
+        if (isAllAligned)
         {
-            const func_t vfunc4 = vfuncs4[depth];
-            const func_t vfunc2 = vfuncs2[depth];
-
-            if (vfunc4 != 0 && (src1_.cols & 3) == 0)
+            if (depth == CV_8U && (src1_.cols & 3) == 0)
             {
                 const int vcols = src1_.cols >> 2;
 
-                vfunc4(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::absDiffMat_v4(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                      PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                      PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                      stream);
 
                 return;
             }
-
-            if (vfunc2 != 0 && (src1_.cols & 1) == 0)
+            else if (depth == CV_16U && (src1_.cols & 1) == 0)
             {
                 const int vcols = src1_.cols >> 1;
 
-                vfunc2(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                arithm::absDiffMat_v2(PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                                      PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                                      PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                                      stream);
 
                 return;
             }
@@ -1687,7 +1416,7 @@ void cv::gpu::absdiff(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Strea
     const func_t func = funcs[depth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
     func(src1_, src2_, dst_, stream);
 }
@@ -1698,36 +1427,28 @@ namespace arithm
     void absDiffScalar(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::absdiff(const GpuMat& src1, const Scalar& src2, GpuMat& dst, Stream& stream)
+static void absDiffScalar(const GpuMat& src, Scalar val, bool, GpuMat& dst, const GpuMat&, double, Stream& stream, int)
 {
-    using namespace arithm;
-
     typedef void (*func_t)(PtrStepSzb src1, double val, PtrStepSzb dst, cudaStream_t stream);
     static const func_t funcs[] =
     {
-        absDiffScalar<unsigned char, float>,
-        absDiffScalar<signed char, float>,
-        absDiffScalar<unsigned short, float>,
-        absDiffScalar<short, float>,
-        absDiffScalar<int, float>,
-        absDiffScalar<float, float>,
-        absDiffScalar<double, double>
+        arithm::absDiffScalar<unsigned char, float>,
+        arithm::absDiffScalar<signed char, float>,
+        arithm::absDiffScalar<unsigned short, float>,
+        arithm::absDiffScalar<short, float>,
+        arithm::absDiffScalar<int, float>,
+        arithm::absDiffScalar<float, float>,
+        arithm::absDiffScalar<double, double>
     };
 
-    const int depth = src1.depth();
+    const int depth = src.depth();
 
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src1.channels() == 1 );
+    funcs[depth](src, val[0], dst, StreamAccessor::getStream(stream));
+}
 
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), src1.type());
-
-    funcs[depth](src1, src2.val[0], dst, StreamAccessor::getStream(stream));
+void cv::gpu::absdiff(InputArray src1, InputArray src2, OutputArray dst, Stream& stream)
+{
+    arithm_op(src1, src2, dst, noArray(), 1.0, -1, stream, absDiffMat, absDiffScalar);
 }
 
 //////////////////////////////////////////////////////////////////////////////
@@ -1739,7 +1460,7 @@ namespace arithm
     void absMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::abs(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::abs(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace arithm;
 
@@ -1755,6 +1476,8 @@ void cv::gpu::abs(const GpuMat& src, GpuMat& dst, Stream& stream)
         absMat<double>
     };
 
+    GpuMat src = _src.getGpuMat();
+
     const int depth = src.depth();
 
     CV_Assert( depth <= CV_64F );
@@ -1763,10 +1486,11 @@ void cv::gpu::abs(const GpuMat& src, GpuMat& dst, Stream& stream)
     if (depth == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     funcs[depth](src, dst, StreamAccessor::getStream(stream));
 }
@@ -1780,7 +1504,7 @@ namespace arithm
     void sqrMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::sqr(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::sqr(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace arithm;
 
@@ -1796,6 +1520,8 @@ void cv::gpu::sqr(const GpuMat& src, GpuMat& dst, Stream& stream)
         sqrMat<double>
     };
 
+    GpuMat src = _src.getGpuMat();
+
     const int depth = src.depth();
 
     CV_Assert( depth <= CV_64F );
@@ -1804,10 +1530,11 @@ void cv::gpu::sqr(const GpuMat& src, GpuMat& dst, Stream& stream)
     if (depth == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     funcs[depth](src, dst, StreamAccessor::getStream(stream));
 }
@@ -1821,7 +1548,7 @@ namespace arithm
     void sqrtMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::sqrt(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::sqrt(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace arithm;
 
@@ -1837,46 +1564,7 @@ void cv::gpu::sqrt(const GpuMat& src, GpuMat& dst, Stream& stream)
         sqrtMat<double>
     };
 
-    const int depth = src.depth();
-
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src.channels() == 1 );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), src.type());
-
-    funcs[depth](src, dst, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// log
-
-namespace arithm
-{
-    template <typename T>
-    void logMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-}
-
-void cv::gpu::log(const GpuMat& src, GpuMat& dst, Stream& stream)
-{
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
-    {
-        logMat<unsigned char>,
-        logMat<signed char>,
-        logMat<unsigned short>,
-        logMat<short>,
-        logMat<int>,
-        logMat<float>,
-        logMat<double>
-    };
+    GpuMat src = _src.getGpuMat();
 
     const int depth = src.depth();
 
@@ -1886,10 +1574,11 @@ void cv::gpu::log(const GpuMat& src, GpuMat& dst, Stream& stream)
     if (depth == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     funcs[depth](src, dst, StreamAccessor::getStream(stream));
 }
@@ -1903,7 +1592,7 @@ namespace arithm
     void expMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::exp(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::exp(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace arithm;
 
@@ -1919,6 +1608,8 @@ void cv::gpu::exp(const GpuMat& src, GpuMat& dst, Stream& stream)
         expMat<double>
     };
 
+    GpuMat src = _src.getGpuMat();
+
     const int depth = src.depth();
 
     CV_Assert( depth <= CV_64F );
@@ -1927,26 +1618,120 @@ void cv::gpu::exp(const GpuMat& src, GpuMat& dst, Stream& stream)
     if (depth == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     funcs[depth](src, dst, StreamAccessor::getStream(stream));
 }
 
+////////////////////////////////////////////////////////////////////////
+// log
+
+namespace arithm
+{
+    template <typename T>
+    void logMat(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+}
+
+void cv::gpu::log(InputArray _src, OutputArray _dst, Stream& stream)
+{
+    using namespace arithm;
+
+    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    static const func_t funcs[] =
+    {
+        logMat<unsigned char>,
+        logMat<signed char>,
+        logMat<unsigned short>,
+        logMat<short>,
+        logMat<int>,
+        logMat<float>,
+        logMat<double>
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    const int depth = src.depth();
+
+    CV_Assert( depth <= CV_64F );
+    CV_Assert( src.channels() == 1 );
+
+    if (depth == CV_64F)
+    {
+        if (!deviceSupports(NATIVE_DOUBLE))
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+    }
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    funcs[depth](src, dst, StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// pow
+
+namespace arithm
+{
+    template<typename T> void pow(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+}
+
+void cv::gpu::pow(InputArray _src, double power, OutputArray _dst, Stream& stream)
+{
+    typedef void (*func_t)(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
+    static const func_t funcs[] =
+    {
+        arithm::pow<unsigned char>,
+        arithm::pow<signed char>,
+        arithm::pow<unsigned short>,
+        arithm::pow<short>,
+        arithm::pow<int>,
+        arithm::pow<float>,
+        arithm::pow<double>
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    const int depth = src.depth();
+    const int cn = src.channels();
+
+    CV_Assert(depth <= CV_64F);
+
+    if (depth == CV_64F)
+    {
+        if (!deviceSupports(NATIVE_DOUBLE))
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+    }
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    PtrStepSzb src_(src.rows, src.cols * cn, src.data, src.step);
+    PtrStepSzb dst_(src.rows, src.cols * cn, dst.data, dst.step);
+
+    funcs[depth](src_, power, dst_, StreamAccessor::getStream(stream));
+}
+
 //////////////////////////////////////////////////////////////////////////////
 // compare
 
 namespace arithm
 {
+    void cmpMatEq_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream);
+    void cmpMatNe_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream);
+    void cmpMatLt_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream);
+    void cmpMatLe_v4(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream);
+
     template <typename T> void cmpMatEq(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     template <typename T> void cmpMatNe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     template <typename T> void cmpMatLt(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     template <typename T> void cmpMatLe(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::compare(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, int cmpop, Stream& s)
+static void cmpMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat&, double, Stream& _stream, int cmpop)
 {
     using namespace arithm;
 
@@ -1962,22 +1747,16 @@ void cv::gpu::compare(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, int c
         {cmpMatEq<double>        , cmpMatNe<double>        , cmpMatLt<double>        , cmpMatLe<double>        }
     };
 
+    typedef void (*func_v4_t)(PtrStepSz<uint> src1, PtrStepSz<uint> src2, PtrStepSz<uint> dst, cudaStream_t stream);
+    static const func_v4_t funcs_v4[] =
+    {
+        cmpMatEq_v4, cmpMatNe_v4, cmpMatLt_v4, cmpMatLe_v4
+    };
+
     const int depth = src1.depth();
     const int cn = src1.channels();
 
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.size() == src1.size() && src2.type() == src1.type() );
-    CV_Assert( cmpop >= CMP_EQ && cmpop <= CMP_NE );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), CV_MAKE_TYPE(CV_8U, cn));
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     static const int codes[] =
     {
@@ -1997,6 +1776,27 @@ void cv::gpu::compare(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, int c
     PtrStepSzb src2_(src1.rows, src1.cols * cn, psrc2[cmpop]->data, psrc2[cmpop]->step);
     PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
+    if (depth == CV_8U && (src1_.cols & 3) == 0)
+    {
+        const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
+        const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
+        const intptr_t dstptr = reinterpret_cast<intptr_t>(dst_.data);
+
+        const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
+
+        if (isAllAligned)
+        {
+            const int vcols = src1_.cols >> 2;
+
+            funcs_v4[code](PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                           PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                           PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                           stream);
+
+            return;
+        }
+    }
+
     const func_t func = funcs[depth][code];
 
     func(src1_, src2_, dst_, stream);
@@ -2023,7 +1823,7 @@ namespace
     }
 }
 
-void cv::gpu::compare(const GpuMat& src, Scalar sc, GpuMat& dst, int cmpop, Stream& stream)
+static void cmpScalar(const GpuMat& src, Scalar val, bool inv, GpuMat& dst, const GpuMat&, double, Stream& stream, int cmpop)
 {
     using namespace arithm;
 
@@ -2045,46 +1845,50 @@ void cv::gpu::compare(const GpuMat& src, Scalar sc, GpuMat& dst, int cmpop, Stre
         castScalar<unsigned char>, castScalar<signed char>, castScalar<unsigned short>, castScalar<short>, castScalar<int>, castScalar<float>, castScalar<double>
     };
 
+    if (inv)
+    {
+        // src1 is a scalar; swap it with src2
+        cmpop = cmpop == CMP_LT ? CMP_GT : cmpop == CMP_LE ? CMP_GE :
+            cmpop == CMP_GE ? CMP_LE : cmpop == CMP_GT ? CMP_LT : cmpop;
+    }
+
     const int depth = src.depth();
     const int cn = src.channels();
 
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( cn <= 4 );
-    CV_Assert( cmpop >= CMP_EQ && cmpop <= CMP_NE );
+    cast_func[depth](val);
 
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
+    funcs[depth][cmpop](src, cn, val.val, dst, StreamAccessor::getStream(stream));
+}
 
-    dst.create(src.size(), CV_MAKE_TYPE(CV_8U, cn));
-
-    cast_func[depth](sc);
-
-    funcs[depth][cmpop](src, cn, sc.val, dst, StreamAccessor::getStream(stream));
+void cv::gpu::compare(InputArray src1, InputArray src2, OutputArray dst, int cmpop, Stream& stream)
+{
+    arithm_op(src1, src2, dst, noArray(), 1.0, CV_8U, stream, cmpMat, cmpScalar, cmpop);
 }
 
 //////////////////////////////////////////////////////////////////////////////
-// Unary bitwise logical operations
+// bitwise_not
 
 namespace arithm
 {
     template <typename T> void bitMatNot(PtrStepSzb src, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask, Stream& s)
+void cv::gpu::bitwise_not(InputArray _src, OutputArray _dst, InputArray _mask, Stream& _stream)
 {
     using namespace arithm;
 
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
+
     const int depth = src.depth();
 
     CV_Assert( depth <= CV_64F );
     CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src.size()) );
 
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     const int bcols = (int) (src.cols * src.elemSize());
 
@@ -2108,7 +1912,7 @@ void cv::gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask, St
     }
     else
     {
-        bitMatNot<unsigned short>(
+        bitMatNot<unsigned char>(
                     PtrStepSzb(src.rows, bcols, src.data, src.step),
                     PtrStepSzb(src.rows, bcols, dst.data, dst.step),
                     mask, stream);
@@ -2118,6 +1922,16 @@ void cv::gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask, St
 //////////////////////////////////////////////////////////////////////////////
 // Binary bitwise logical operations
 
+namespace
+{
+    enum
+    {
+        BIT_OP_AND,
+        BIT_OP_OR,
+        BIT_OP_XOR
+    };
+}
+
 namespace arithm
 {
     template <typename T> void bitMatAnd(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
@@ -2125,19 +1939,31 @@ namespace arithm
     template <typename T> void bitMatXor(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
 }
 
-void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, Stream& s)
+static void bitMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, double, Stream& _stream, int op)
 {
     using namespace arithm;
 
-    const int depth = src1.depth();
+    typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, PtrStepb mask, cudaStream_t stream);
+    static const func_t funcs32[] =
+    {
+        bitMatAnd<uint>,
+        bitMatOr<uint>,
+        bitMatXor<uint>
+    };
+    static const func_t funcs16[] =
+    {
+        bitMatAnd<ushort>,
+        bitMatOr<ushort>,
+        bitMatXor<ushort>
+    };
+    static const func_t funcs8[] =
+    {
+        bitMatAnd<uchar>,
+        bitMatOr<uchar>,
+        bitMatXor<uchar>
+    };
 
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.size() == src1.size() && src2.type() == src1.type() );
-    CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src1.size()) );
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     const int bcols = (int) (src1.cols * src1.elemSize());
 
@@ -2145,8 +1971,7 @@ void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, c
     {
         const int vcols = bcols >> 2;
 
-        bitMatAnd<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
+        funcs32[op](PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
                     PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
                     PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
                     mask, stream);
@@ -2155,8 +1980,7 @@ void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, c
     {
         const int vcols = bcols >> 1;
 
-        bitMatAnd<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
+        funcs16[op](PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
                     PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
                     PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
                     mask, stream);
@@ -2164,111 +1988,13 @@ void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, c
     else
     {
 
-        bitMatAnd<unsigned int>(
-                    PtrStepSzb(src1.rows, bcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, bcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, bcols, dst.data, dst.step),
-                    mask, stream);
+        funcs8[op](PtrStepSzb(src1.rows, bcols, src1.data, src1.step),
+                   PtrStepSzb(src1.rows, bcols, src2.data, src2.step),
+                   PtrStepSzb(src1.rows, bcols, dst.data, dst.step),
+                   mask, stream);
     }
 }
 
-void cv::gpu::bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, Stream& s)
-{
-    using namespace arithm;
-
-    const int depth = src1.depth();
-
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.size() == src1.size() && src2.type() == src1.type() );
-    CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src1.size()) );
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    const int bcols = (int) (src1.cols * src1.elemSize());
-
-    if ((bcols & 3) == 0)
-    {
-        const int vcols = bcols >> 2;
-
-        bitMatOr<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
-                    mask, stream);
-    }
-    else if ((bcols & 1) == 0)
-    {
-        const int vcols = bcols >> 1;
-
-        bitMatOr<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
-                    mask, stream);
-    }
-    else
-    {
-
-        bitMatOr<unsigned int>(
-                    PtrStepSzb(src1.rows, bcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, bcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, bcols, dst.data, dst.step),
-                    mask, stream);
-    }
-}
-
-void cv::gpu::bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, Stream& s)
-{
-    using namespace arithm;
-
-    const int depth = src1.depth();
-
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.size() == src1.size() && src2.type() == src1.type() );
-    CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src1.size()) );
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    const int bcols = (int) (src1.cols * src1.elemSize());
-
-    if ((bcols & 3) == 0)
-    {
-        const int vcols = bcols >> 2;
-
-        bitMatXor<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
-                    mask, stream);
-    }
-    else if ((bcols & 1) == 0)
-    {
-        const int vcols = bcols >> 1;
-
-        bitMatXor<unsigned int>(
-                    PtrStepSzb(src1.rows, vcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, vcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, vcols, dst.data, dst.step),
-                    mask, stream);
-    }
-    else
-    {
-
-        bitMatXor<unsigned int>(
-                    PtrStepSzb(src1.rows, bcols, src1.data, src1.step),
-                    PtrStepSzb(src1.rows, bcols, src2.data, src2.step),
-                    PtrStepSzb(src1.rows, bcols, dst.data, dst.step),
-                    mask, stream);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////////////
-// Binary bitwise logical operations with scalars
-
 namespace arithm
 {
     template <typename T> void bitScalarAnd(PtrStepSzb src1, unsigned int src2, PtrStepSzb dst, cudaStream_t stream);
@@ -2356,18 +2082,34 @@ namespace
     };
 }
 
-void cv::gpu::bitwise_and(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stream& stream)
+static void bitScalar(const GpuMat& src, Scalar val, bool, GpuMat& dst, const GpuMat& mask, double, Stream& stream, int op)
 {
     using namespace arithm;
 
     typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
-    static const func_t funcs[5][4] =
+    static const func_t funcs[3][5][4] =
     {
-        {BitScalar<unsigned char, bitScalarAnd<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiAndC_8u_C3R >::call, BitScalar4< bitScalarAnd<unsigned int> >::call},
-        {0,0,0,0},
-        {BitScalar<unsigned short, bitScalarAnd<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiAndC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiAndC_16u_C4R>::call},
-        {0,0,0,0},
-        {BitScalar<int, bitScalarAnd<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiAndC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiAndC_32s_C4R>::call}
+        {
+            {BitScalar<unsigned char, bitScalarAnd<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiAndC_8u_C3R >::call, BitScalar4< bitScalarAnd<unsigned int> >::call},
+            {0,0,0,0},
+            {BitScalar<unsigned short, bitScalarAnd<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiAndC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiAndC_16u_C4R>::call},
+            {0,0,0,0},
+            {BitScalar<int, bitScalarAnd<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiAndC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiAndC_32s_C4R>::call}
+        },
+        {
+            {BitScalar<unsigned char, bitScalarOr<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiOrC_8u_C3R >::call, BitScalar4< bitScalarOr<unsigned int> >::call},
+            {0,0,0,0},
+            {BitScalar<unsigned short, bitScalarOr<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiOrC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiOrC_16u_C4R>::call},
+            {0,0,0,0},
+            {BitScalar<int, bitScalarOr<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiOrC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiOrC_32s_C4R>::call}
+        },
+        {
+            {BitScalar<unsigned char, bitScalarXor<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiXorC_8u_C3R >::call, BitScalar4< bitScalarXor<unsigned int> >::call},
+            {0,0,0,0},
+            {BitScalar<unsigned short, bitScalarXor<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiXorC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiXorC_16u_C4R>::call},
+            {0,0,0,0},
+            {BitScalar<int, bitScalarXor<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiXorC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiXorC_32s_C4R>::call}
+        }
     };
 
     const int depth = src.depth();
@@ -2375,60 +2117,24 @@ void cv::gpu::bitwise_and(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stre
 
     CV_Assert( depth == CV_8U || depth == CV_16U || depth == CV_32S );
     CV_Assert( cn == 1 || cn == 3 || cn == 4 );
+    CV_Assert( mask.empty() );
 
-    dst.create(src.size(), src.type());
-
-    funcs[depth][cn - 1](src, sc, dst, StreamAccessor::getStream(stream));
+    funcs[op][depth][cn - 1](src, val, dst, StreamAccessor::getStream(stream));
 }
 
-void cv::gpu::bitwise_or(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stream& stream)
+void cv::gpu::bitwise_or(InputArray src1, InputArray src2, OutputArray dst, InputArray mask, Stream& stream)
 {
-    using namespace arithm;
-
-    typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
-    static const func_t funcs[5][4] =
-    {
-        {BitScalar<unsigned char, bitScalarOr<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiOrC_8u_C3R >::call, BitScalar4< bitScalarOr<unsigned int> >::call},
-        {0,0,0,0},
-        {BitScalar<unsigned short, bitScalarOr<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiOrC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiOrC_16u_C4R>::call},
-        {0,0,0,0},
-        {BitScalar<int, bitScalarOr<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiOrC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiOrC_32s_C4R>::call}
-    };
-
-    const int depth = src.depth();
-    const int cn = src.channels();
-
-    CV_Assert( depth == CV_8U || depth == CV_16U || depth == CV_32S );
-    CV_Assert( cn == 1 || cn == 3 || cn == 4 );
-
-    dst.create(src.size(), src.type());
-
-    funcs[depth][cn - 1](src, sc, dst, StreamAccessor::getStream(stream));
+    arithm_op(src1, src2, dst, mask, 1.0, -1, stream, bitMat, bitScalar, BIT_OP_OR);
 }
 
-void cv::gpu::bitwise_xor(const GpuMat& src, const Scalar& sc, GpuMat& dst, Stream& stream)
+void cv::gpu::bitwise_and(InputArray src1, InputArray src2, OutputArray dst, InputArray mask, Stream& stream)
 {
-    using namespace arithm;
+    arithm_op(src1, src2, dst, mask, 1.0, -1, stream, bitMat, bitScalar, BIT_OP_AND);
+}
 
-    typedef void (*func_t)(const GpuMat& src, Scalar sc, GpuMat& dst, cudaStream_t stream);
-    static const func_t funcs[5][4] =
-    {
-        {BitScalar<unsigned char, bitScalarXor<unsigned char> >::call  , 0, NppBitwiseC<CV_8U , 3, nppiXorC_8u_C3R >::call, BitScalar4< bitScalarXor<unsigned int> >::call},
-        {0,0,0,0},
-        {BitScalar<unsigned short, bitScalarXor<unsigned short> >::call, 0, NppBitwiseC<CV_16U, 3, nppiXorC_16u_C3R>::call, NppBitwiseC<CV_16U, 4, nppiXorC_16u_C4R>::call},
-        {0,0,0,0},
-        {BitScalar<int, bitScalarXor<int> >::call                      , 0, NppBitwiseC<CV_32S, 3, nppiXorC_32s_C3R>::call, NppBitwiseC<CV_32S, 4, nppiXorC_32s_C4R>::call}
-    };
-
-    const int depth = src.depth();
-    const int cn = src.channels();
-
-    CV_Assert( depth == CV_8U || depth == CV_16U || depth == CV_32S );
-    CV_Assert( cn == 1 || cn == 3 || cn == 4 );
-
-    dst.create(src.size(), src.type());
-
-    funcs[depth][cn - 1](src, sc, dst, StreamAccessor::getStream(stream));
+void cv::gpu::bitwise_xor(InputArray src1, InputArray src2, OutputArray dst, InputArray mask, Stream& stream)
+{
+    arithm_op(src1, src2, dst, mask, 1.0, -1, stream, bitMat, bitScalar, BIT_OP_XOR);
 }
 
 //////////////////////////////////////////////////////////////////////////////
@@ -2487,7 +2193,7 @@ namespace
     };
 }
 
-void cv::gpu::rshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream)
+void cv::gpu::rshift(InputArray _src, Scalar_<int> val, OutputArray _dst, Stream& stream)
 {
     typedef void (*func_t)(const GpuMat& src, Scalar_<Npp32u> sc, GpuMat& dst, cudaStream_t stream);
     static const func_t funcs[5][4] =
@@ -2499,15 +2205,18 @@ void cv::gpu::rshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& st
         {NppShift<CV_32S, 1, nppiRShiftC_32s_C1R>::call, 0, NppShift<CV_32S, 3, nppiRShiftC_32s_C3R>::call, NppShift<CV_32S, 4, nppiRShiftC_32s_C4R>::call},
     };
 
-    CV_Assert(src.depth() < CV_32F);
-    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
+    GpuMat src = _src.getGpuMat();
 
-    dst.create(src.size(), src.type());
+    CV_Assert( src.depth() < CV_32F );
+    CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
 
-    funcs[src.depth()][src.channels() - 1](src, sc, dst, StreamAccessor::getStream(stream));
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    funcs[src.depth()][src.channels() - 1](src, val, dst, StreamAccessor::getStream(stream));
 }
 
-void cv::gpu::lshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& stream)
+void cv::gpu::lshift(InputArray _src, Scalar_<int> val, OutputArray _dst, Stream& stream)
 {
     typedef void (*func_t)(const GpuMat& src, Scalar_<Npp32u> sc, GpuMat& dst, cudaStream_t stream);
     static const func_t funcs[5][4] =
@@ -2519,81 +2228,91 @@ void cv::gpu::lshift(const GpuMat& src, Scalar_<int> sc, GpuMat& dst, Stream& st
         {NppShift<CV_32S, 1, nppiLShiftC_32s_C1R>::call, 0, NppShift<CV_32S, 3, nppiLShiftC_32s_C3R>::call, NppShift<CV_32S, 4, nppiLShiftC_32s_C4R>::call},
     };
 
-    CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S);
-    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
+    GpuMat src = _src.getGpuMat();
 
-    dst.create(src.size(), src.type());
+    CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S );
+    CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
 
-    funcs[src.depth()][src.channels() - 1](src, sc, dst, StreamAccessor::getStream(stream));
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    funcs[src.depth()][src.channels() - 1](src, val, dst, StreamAccessor::getStream(stream));
 }
 
 //////////////////////////////////////////////////////////////////////////////
 // Minimum and maximum operations
 
+namespace
+{
+    enum
+    {
+        MIN_OP,
+        MAX_OP
+    };
+}
+
 namespace arithm
 {
-    template <typename T> void vmin4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template <typename T> void vmin2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    void minMat_v4(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    void minMat_v2(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
     template <typename T> void minMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     template <typename T> void minScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
 
-    template <typename T> void vmax4(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    template <typename T> void vmax2(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
+    void maxMat_v4(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    void maxMat_v2(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
     template <typename T> void maxMat(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
     template <typename T> void maxScalar(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& s)
+void minMaxMat(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat&, double, Stream& _stream, int op)
 {
     using namespace arithm;
 
     typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
+    static const func_t funcs[2][7] =
     {
-        minMat<unsigned char>,
-        minMat<signed char>,
-        minMat<unsigned short>,
-        minMat<short>,
-        minMat<int>,
-        minMat<float>,
-        minMat<double>
+        {
+            minMat<unsigned char>,
+            minMat<signed char>,
+            minMat<unsigned short>,
+            minMat<short>,
+            minMat<int>,
+            minMat<float>,
+            minMat<double>
+        },
+        {
+            maxMat<unsigned char>,
+            maxMat<signed char>,
+            maxMat<unsigned short>,
+            maxMat<short>,
+            maxMat<int>,
+            maxMat<float>,
+            maxMat<double>
+        }
     };
-    static const func_t vfuncs4[] =
+
+    typedef void (*opt_func_t)(PtrStepSz<unsigned int> src1, PtrStepSz<unsigned int> src2, PtrStepSz<unsigned int> dst, cudaStream_t stream);
+    static const opt_func_t funcs_v4[2] =
     {
-        vmin4<unsigned int>,
-        vmin4<int>,
-        0,
-        0
+        minMat_v4, maxMat_v4
     };
-    static const func_t vfuncs2[] =
+    static const opt_func_t funcs_v2[2] =
     {
-        0,
-        0,
-        vmin2<unsigned int>,
-        vmin2<int>
+        minMat_v2, maxMat_v2
     };
 
     const int depth = src1.depth();
     const int cn = src1.channels();
 
     CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
 
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
     PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
     PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
 
-    if (depth < CV_32S)
+    if (depth == CV_8U || depth == CV_16U)
     {
         const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
         const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
@@ -2601,138 +2320,37 @@ void cv::gpu::min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& s
 
         const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
 
-        if (deviceSupports(FEATURE_SET_COMPUTE_20) && isAllAligned)
+        if (isAllAligned)
         {
-            const func_t vfunc4 = vfuncs4[depth];
-            const func_t vfunc2 = vfuncs2[depth];
-
-            if (vfunc4 != 0 && (src1_.cols & 3) == 0)
+            if (depth == CV_8U && (src1_.cols & 3) == 0)
             {
                 const int vcols = src1_.cols >> 2;
 
-                vfunc4(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                funcs_v4[op](PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                             PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                             PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                             stream);
 
                 return;
             }
-
-            if (vfunc2 != 0 && (src1_.cols & 1) == 0)
+            else if (depth == CV_16U && (src1_.cols & 1) == 0)
             {
                 const int vcols = src1_.cols >> 1;
 
-                vfunc2(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
+                funcs_v2[op](PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src1_.data, src1_.step),
+                             PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) src2_.data, src2_.step),
+                             PtrStepSz<unsigned int>(src1_.rows, vcols, (unsigned int*) dst_.data, dst_.step),
+                             stream);
 
                 return;
             }
         }
     }
 
-    const func_t func = funcs[depth];
+    const func_t func = funcs[op][depth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
-
-    func(src1_, src2_, dst_, stream);
-}
-
-void cv::gpu::max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& s)
-{
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src1, PtrStepSzb src2, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
-    {
-        maxMat<unsigned char>,
-        maxMat<signed char>,
-        maxMat<unsigned short>,
-        maxMat<short>,
-        maxMat<int>,
-        maxMat<float>,
-        maxMat<double>
-    };
-    static const func_t vfuncs4[] =
-    {
-        vmax4<unsigned int>,
-        vmax4<int>,
-        0,
-        0
-    };
-    static const func_t vfuncs2[] =
-    {
-        0,
-        0,
-        vmax2<unsigned int>,
-        vmax2<int>
-    };
-
-    const int depth = src1.depth();
-    const int cn = src1.channels();
-
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src2.type() == src1.type() && src2.size() == src1.size() );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src1.size(), src1.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
-    PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
-    PtrStepSzb dst_(src1.rows, src1.cols * cn, dst.data, dst.step);
-
-    if (depth < CV_32S)
-    {
-        const intptr_t src1ptr = reinterpret_cast<intptr_t>(src1_.data);
-        const intptr_t src2ptr = reinterpret_cast<intptr_t>(src2_.data);
-        const intptr_t dstptr = reinterpret_cast<intptr_t>(dst_.data);
-
-        const bool isAllAligned = (src1ptr & 31) == 0 && (src2ptr & 31) == 0 && (dstptr & 31) == 0;
-
-        if (deviceSupports(FEATURE_SET_COMPUTE_20) && isAllAligned)
-        {
-            const func_t vfunc4 = vfuncs4[depth];
-            const func_t vfunc2 = vfuncs2[depth];
-
-            if (vfunc4 != 0 && (src1_.cols & 3) == 0)
-            {
-                const int vcols = src1_.cols >> 2;
-
-                vfunc4(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
-
-                return;
-            }
-
-            if (vfunc2 != 0 && (src1_.cols & 1) == 0)
-            {
-                const int vcols = src1_.cols >> 1;
-
-                vfunc2(PtrStepSzb(src1_.rows, vcols, src1_.data, src1_.step),
-                       PtrStepSzb(src1_.rows, vcols, src2_.data, src2_.step),
-                       PtrStepSzb(src1_.rows, vcols, dst_.data, dst_.step),
-                       stream);
-
-                return;
-            }
-        }
-    }
-
-    const func_t func = funcs[depth];
-
-    if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
     func(src1_, src2_, dst_, stream);
 }
@@ -2745,20 +2363,31 @@ namespace
     }
 }
 
-void cv::gpu::min(const GpuMat& src, double val, GpuMat& dst, Stream& stream)
+void minMaxScalar(const GpuMat& src, Scalar val, bool, GpuMat& dst, const GpuMat&, double, Stream& stream, int op)
 {
     using namespace arithm;
 
     typedef void (*func_t)(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
+    static const func_t funcs[2][7] =
     {
-        minScalar<unsigned char>,
-        minScalar<signed char>,
-        minScalar<unsigned short>,
-        minScalar<short>,
-        minScalar<int>,
-        minScalar<float>,
-        minScalar<double>
+        {
+            minScalar<unsigned char>,
+            minScalar<signed char>,
+            minScalar<unsigned short>,
+            minScalar<short>,
+            minScalar<int>,
+            minScalar<float>,
+            minScalar<double>
+        },
+        {
+            maxScalar<unsigned char>,
+            maxScalar<signed char>,
+            maxScalar<unsigned short>,
+            maxScalar<short>,
+            maxScalar<int>,
+            maxScalar<float>,
+            maxScalar<double>
+        }
     };
 
     typedef double (*cast_func_t)(double sc);
@@ -2772,233 +2401,17 @@ void cv::gpu::min(const GpuMat& src, double val, GpuMat& dst, Stream& stream)
     CV_Assert( depth <= CV_64F );
     CV_Assert( src.channels() == 1 );
 
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), src.type());
-
-    funcs[depth](src, cast_func[depth](val), dst, StreamAccessor::getStream(stream));
+    funcs[op][depth](src, cast_func[depth](val[0]), dst, StreamAccessor::getStream(stream));
 }
 
-void cv::gpu::max(const GpuMat& src, double val, GpuMat& dst, Stream& stream)
+void cv::gpu::min(InputArray src1, InputArray src2, OutputArray dst, Stream& stream)
 {
-    using namespace arithm;
-
-    typedef void (*func_t)(PtrStepSzb src1, double src2, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
-    {
-        maxScalar<unsigned char>,
-        maxScalar<signed char>,
-        maxScalar<unsigned short>,
-        maxScalar<short>,
-        maxScalar<int>,
-        maxScalar<float>,
-        maxScalar<double>
-    };
-
-    typedef double (*cast_func_t)(double sc);
-    static const cast_func_t cast_func[] =
-    {
-        castScalar<unsigned char>, castScalar<signed char>, castScalar<unsigned short>, castScalar<short>, castScalar<int>, castScalar<float>, castScalar<double>
-    };
-
-    const int depth = src.depth();
-
-    CV_Assert( depth <= CV_64F );
-    CV_Assert( src.channels() == 1 );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), src.type());
-
-    funcs[depth](src, cast_func[depth](val), dst, StreamAccessor::getStream(stream));
+    arithm_op(src1, src2, dst, noArray(), 1.0, -1, stream, minMaxMat, minMaxScalar, MIN_OP);
 }
 
-////////////////////////////////////////////////////////////////////////
-// threshold
-
-namespace arithm
+void cv::gpu::max(InputArray src1, InputArray src2, OutputArray dst, Stream& stream)
 {
-    template <typename T>
-    void threshold(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-}
-
-double cv::gpu::threshold(const GpuMat& src, GpuMat& dst, double thresh, double maxVal, int type, Stream& s)
-{
-    const int depth = src.depth();
-
-    CV_Assert( src.channels() == 1 && depth <= CV_64F );
-    CV_Assert( type <= THRESH_TOZERO_INV );
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), src.type());
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-
-    if (src.type() == CV_32FC1 && type == THRESH_TRUNC)
-    {
-        NppStreamHandler h(stream);
-
-        NppiSize sz;
-        sz.width  = src.cols;
-        sz.height = src.rows;
-
-        nppSafeCall( nppiThreshold_32f_C1R(src.ptr<Npp32f>(), static_cast<int>(src.step),
-            dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, static_cast<Npp32f>(thresh), NPP_CMP_GREATER) );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-    else
-    {
-        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
-        static const func_t funcs[] =
-        {
-            arithm::threshold<unsigned char>,
-            arithm::threshold<signed char>,
-            arithm::threshold<unsigned short>,
-            arithm::threshold<short>,
-            arithm::threshold<int>,
-            arithm::threshold<float>,
-            arithm::threshold<double>
-        };
-
-        if (depth != CV_32F && depth != CV_64F)
-        {
-            thresh = cvFloor(thresh);
-            maxVal = cvRound(maxVal);
-        }
-
-        funcs[depth](src, dst, thresh, maxVal, type, stream);
-    }
-
-    return thresh;
-}
-
-////////////////////////////////////////////////////////////////////////
-// pow
-
-namespace arithm
-{
-    template<typename T> void pow(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-}
-
-void cv::gpu::pow(const GpuMat& src, double power, GpuMat& dst, Stream& stream)
-{
-    typedef void (*func_t)(PtrStepSzb src, double power, PtrStepSzb dst, cudaStream_t stream);
-    static const func_t funcs[] =
-    {
-        arithm::pow<unsigned char>,
-        arithm::pow<signed char>,
-        arithm::pow<unsigned short>,
-        arithm::pow<short>,
-        arithm::pow<int>,
-        arithm::pow<float>,
-        arithm::pow<double>
-    };
-
-    const int depth = src.depth();
-    const int cn = src.channels();
-
-    CV_Assert(depth <= CV_64F);
-
-    if (depth == CV_64F)
-    {
-        if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
-    }
-
-    dst.create(src.size(), src.type());
-
-    PtrStepSzb src_(src.rows, src.cols * cn, src.data, src.step);
-    PtrStepSzb dst_(src.rows, src.cols * cn, dst.data, dst.step);
-
-    funcs[depth](src_, power, dst_, StreamAccessor::getStream(stream));
-}
-
-////////////////////////////////////////////////////////////////////////
-// alphaComp
-
-namespace
-{
-    template <int DEPTH> struct NppAlphaCompFunc
-    {
-        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
-
-        typedef NppStatus (*func_t)(const npp_t* pSrc1, int nSrc1Step, const npp_t* pSrc2, int nSrc2Step, npp_t* pDst, int nDstStep, NppiSize oSizeROI, NppiAlphaOp eAlphaOp);
-    };
-
-    template <int DEPTH, typename NppAlphaCompFunc<DEPTH>::func_t func> struct NppAlphaComp
-    {
-        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
-
-        static void call(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, NppiAlphaOp eAlphaOp, cudaStream_t stream)
-        {
-            NppStreamHandler h(stream);
-
-            NppiSize oSizeROI;
-            oSizeROI.width = img1.cols;
-            oSizeROI.height = img2.rows;
-
-            nppSafeCall( func(img1.ptr<npp_t>(), static_cast<int>(img1.step), img2.ptr<npp_t>(), static_cast<int>(img2.step),
-                              dst.ptr<npp_t>(), static_cast<int>(dst.step), oSizeROI, eAlphaOp) );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-    };
-}
-
-void cv::gpu::alphaComp(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, int alpha_op, Stream& stream)
-{
-    static const NppiAlphaOp npp_alpha_ops[] = {
-        NPPI_OP_ALPHA_OVER,
-        NPPI_OP_ALPHA_IN,
-        NPPI_OP_ALPHA_OUT,
-        NPPI_OP_ALPHA_ATOP,
-        NPPI_OP_ALPHA_XOR,
-        NPPI_OP_ALPHA_PLUS,
-        NPPI_OP_ALPHA_OVER_PREMUL,
-        NPPI_OP_ALPHA_IN_PREMUL,
-        NPPI_OP_ALPHA_OUT_PREMUL,
-        NPPI_OP_ALPHA_ATOP_PREMUL,
-        NPPI_OP_ALPHA_XOR_PREMUL,
-        NPPI_OP_ALPHA_PLUS_PREMUL,
-        NPPI_OP_ALPHA_PREMUL
-    };
-
-    typedef void (*func_t)(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, NppiAlphaOp eAlphaOp, cudaStream_t stream);
-
-    static const func_t funcs[] =
-    {
-        NppAlphaComp<CV_8U, nppiAlphaComp_8u_AC4R>::call,
-        0,
-        NppAlphaComp<CV_16U, nppiAlphaComp_16u_AC4R>::call,
-        0,
-        NppAlphaComp<CV_32S, nppiAlphaComp_32s_AC4R>::call,
-        NppAlphaComp<CV_32F, nppiAlphaComp_32f_AC4R>::call
-    };
-
-    CV_Assert( img1.type() == CV_8UC4 || img1.type() == CV_16UC4 || img1.type() == CV_32SC4 || img1.type() == CV_32FC4 );
-    CV_Assert( img1.size() == img2.size() && img1.type() == img2.type() );
-
-    dst.create(img1.size(), img1.type());
-
-    const func_t func = funcs[img1.depth()];
-
-    func(img1, img2, dst, npp_alpha_ops[alpha_op], StreamAccessor::getStream(stream));
+    arithm_op(src1, src2, dst, noArray(), 1.0, -1, stream, minMaxMat, minMaxScalar, MAX_OP);
 }
 
 ////////////////////////////////////////////////////////////////////////
@@ -3010,7 +2423,7 @@ namespace arithm
     void addWeighted(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
 }
 
-void cv::gpu::addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2, double beta, double gamma, GpuMat& dst, int ddepth, Stream& stream)
+void cv::gpu::addWeighted(InputArray _src1, double alpha, InputArray _src2, double beta, double gamma, OutputArray _dst, int ddepth, Stream& stream)
 {
     typedef void (*func_t)(PtrStepSzb src1, double alpha, PtrStepSzb src2, double beta, double gamma, PtrStepSzb dst, cudaStream_t stream);
     static const func_t funcs[7][7][7] =
@@ -3472,6 +2885,9 @@ void cv::gpu::addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2,
         }
     };
 
+    GpuMat src1 = _src1.getGpuMat();
+    GpuMat src2 = _src2.getGpuMat();
+
     int sdepth1 = src1.depth();
     int sdepth2 = src2.depth();
     ddepth = ddepth >= 0 ? CV_MAT_DEPTH(ddepth) : std::max(sdepth1, sdepth2);
@@ -3483,10 +2899,11 @@ void cv::gpu::addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2,
     if (sdepth1 == CV_64F || sdepth2 == CV_64F || ddepth == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
-    dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
+    _dst.create(src1.size(), CV_MAKE_TYPE(ddepth, cn));
+    GpuMat dst = _dst.getGpuMat();
 
     PtrStepSzb src1_(src1.rows, src1.cols * cn, src1.data, src1.step);
     PtrStepSzb src2_(src1.rows, src1.cols * cn, src2.data, src2.step);
@@ -3503,9 +2920,228 @@ void cv::gpu::addWeighted(const GpuMat& src1, double alpha, const GpuMat& src2,
     const func_t func = funcs[sdepth1][sdepth2][ddepth];
 
     if (!func)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of source and destination types");
+        CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
 
     func(src1_, alpha, src2_, beta, gamma, dst_, StreamAccessor::getStream(stream));
 }
 
+////////////////////////////////////////////////////////////////////////
+// threshold
+
+namespace arithm
+{
+    template <typename T>
+    void threshold(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+}
+
+double cv::gpu::threshold(InputArray _src, OutputArray _dst, double thresh, double maxVal, int type, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    const int depth = src.depth();
+
+    CV_Assert( src.channels() == 1 && depth <= CV_64F );
+    CV_Assert( type <= 4/*THRESH_TOZERO_INV*/ );
+
+    if (depth == CV_64F)
+    {
+        if (!deviceSupports(NATIVE_DOUBLE))
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
+    }
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    if (src.type() == CV_32FC1 && type == 2/*THRESH_TRUNC*/)
+    {
+        NppStreamHandler h(stream);
+
+        NppiSize sz;
+        sz.width  = src.cols;
+        sz.height = src.rows;
+
+        nppSafeCall( nppiThreshold_32f_C1R(src.ptr<Npp32f>(), static_cast<int>(src.step),
+            dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, static_cast<Npp32f>(thresh), NPP_CMP_GREATER) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+    else
+    {
+        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, double thresh, double maxVal, int type, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            arithm::threshold<unsigned char>,
+            arithm::threshold<signed char>,
+            arithm::threshold<unsigned short>,
+            arithm::threshold<short>,
+            arithm::threshold<int>,
+            arithm::threshold<float>,
+            arithm::threshold<double>
+        };
+
+        if (depth != CV_32F && depth != CV_64F)
+        {
+            thresh = cvFloor(thresh);
+            maxVal = cvRound(maxVal);
+        }
+
+        funcs[depth](src, dst, thresh, maxVal, type, stream);
+    }
+
+    return thresh;
+}
+
+////////////////////////////////////////////////////////////////////////
+// NPP magnitide
+
+namespace
+{
+    typedef NppStatus (*nppMagnitude_t)(const Npp32fc* pSrc, int nSrcStep, Npp32f* pDst, int nDstStep, NppiSize oSizeROI);
+
+    void npp_magnitude(const GpuMat& src, GpuMat& dst, nppMagnitude_t func, cudaStream_t stream)
+    {
+        CV_Assert(src.type() == CV_32FC2);
+
+        NppiSize sz;
+        sz.width = src.cols;
+        sz.height = src.rows;
+
+        NppStreamHandler h(stream);
+
+        nppSafeCall( func(src.ptr<Npp32fc>(), static_cast<int>(src.step), dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+void cv::gpu::magnitude(InputArray _src, OutputArray _dst, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    _dst.create(src.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    npp_magnitude(src, dst, nppiMagnitude_32fc32f_C1R, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::magnitudeSqr(InputArray _src, OutputArray _dst, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    _dst.create(src.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    npp_magnitude(src, dst, nppiMagnitudeSqr_32fc32f_C1R, StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// Polar <-> Cart
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace mathfunc
+    {
+        void cartToPolar_gpu(PtrStepSzf x, PtrStepSzf y, PtrStepSzf mag, bool magSqr, PtrStepSzf angle, bool angleInDegrees, cudaStream_t stream);
+        void polarToCart_gpu(PtrStepSzf mag, PtrStepSzf angle, PtrStepSzf x, PtrStepSzf y, bool angleInDegrees, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    void cartToPolar_caller(const GpuMat& x, const GpuMat& y, GpuMat* mag, bool magSqr, GpuMat* angle, bool angleInDegrees, cudaStream_t stream)
+    {
+        using namespace ::cv::gpu::cudev::mathfunc;
+
+        CV_Assert(x.size() == y.size() && x.type() == y.type());
+        CV_Assert(x.depth() == CV_32F);
+
+        GpuMat x1cn = x.reshape(1);
+        GpuMat y1cn = y.reshape(1);
+        GpuMat mag1cn = mag ? mag->reshape(1) : GpuMat();
+        GpuMat angle1cn = angle ? angle->reshape(1) : GpuMat();
+
+        cartToPolar_gpu(x1cn, y1cn, mag1cn, magSqr, angle1cn, angleInDegrees, stream);
+    }
+
+    void polarToCart_caller(const GpuMat& mag, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees, cudaStream_t stream)
+    {
+        using namespace ::cv::gpu::cudev::mathfunc;
+
+        CV_Assert((mag.empty() || mag.size() == angle.size()) && mag.type() == angle.type());
+        CV_Assert(mag.depth() == CV_32F);
+
+        GpuMat mag1cn = mag.reshape(1);
+        GpuMat angle1cn = angle.reshape(1);
+        GpuMat x1cn = x.reshape(1);
+        GpuMat y1cn = y.reshape(1);
+
+        polarToCart_gpu(mag1cn, angle1cn, x1cn, y1cn, angleInDegrees, stream);
+    }
+}
+
+void cv::gpu::magnitude(InputArray _x, InputArray _y, OutputArray _dst, Stream& stream)
+{
+    GpuMat x = _x.getGpuMat();
+    GpuMat y = _y.getGpuMat();
+
+    _dst.create(x.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    cartToPolar_caller(x, y, &dst, false, 0, false, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::magnitudeSqr(InputArray _x, InputArray _y, OutputArray _dst, Stream& stream)
+{
+    GpuMat x = _x.getGpuMat();
+    GpuMat y = _y.getGpuMat();
+
+    _dst.create(x.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    cartToPolar_caller(x, y, &dst, true, 0, false, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::phase(InputArray _x, InputArray _y, OutputArray _dst, bool angleInDegrees, Stream& stream)
+{
+    GpuMat x = _x.getGpuMat();
+    GpuMat y = _y.getGpuMat();
+
+    _dst.create(x.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    cartToPolar_caller(x, y, 0, false, &dst, angleInDegrees, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::cartToPolar(InputArray _x, InputArray _y, OutputArray _mag, OutputArray _angle, bool angleInDegrees, Stream& stream)
+{
+    GpuMat x = _x.getGpuMat();
+    GpuMat y = _y.getGpuMat();
+
+    _mag.create(x.size(), CV_32FC1);
+    GpuMat mag = _mag.getGpuMat();
+
+    _angle.create(x.size(), CV_32FC1);
+    GpuMat angle = _angle.getGpuMat();
+
+    cartToPolar_caller(x, y, &mag, false, &angle, angleInDegrees, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::polarToCart(InputArray _mag, InputArray _angle, OutputArray _x, OutputArray _y, bool angleInDegrees, Stream& stream)
+{
+    GpuMat mag = _mag.getGpuMat();
+    GpuMat angle = _angle.getGpuMat();
+
+    _x.create(mag.size(), CV_32FC1);
+    GpuMat x = _x.getGpuMat();
+
+    _y.create(mag.size(), CV_32FC1);
+    GpuMat y = _y.getGpuMat();
+
+    polarToCart_caller(mag, angle, x, y, angleInDegrees, StreamAccessor::getStream(stream));
+}
+
 #endif
diff --git a/modules/gpuarithm/src/precomp.hpp b/modules/gpuarithm/src/precomp.hpp
new file mode 100644
index 0000000000..6aab36a711
--- /dev/null
+++ b/modules/gpuarithm/src/precomp.hpp
@@ -0,0 +1,70 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+
+#include "cvconfig.h"
+
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPULEGACY
+#  include "opencv2/gpulegacy.hpp"
+#  include "opencv2/gpulegacy/private.hpp"
+#endif
+
+#ifdef HAVE_CUBLAS
+#  include <cublas.h>
+#endif
+
+#ifdef HAVE_CUFFT
+#  include <cufft.h>
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/src/matrix_reductions.cpp b/modules/gpuarithm/src/reductions.cpp
similarity index 64%
rename from modules/gpu/src/matrix_reductions.cpp
rename to modules/gpuarithm/src/reductions.cpp
index 67e65fc585..248fa9a4e7 100644
--- a/modules/gpu/src/matrix_reductions.cpp
+++ b/modules/gpuarithm/src/reductions.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,28 +47,28 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::gpu::meanStdDev(const GpuMat&, Scalar&, Scalar&) { throw_nogpu(); }
-void cv::gpu::meanStdDev(const GpuMat&, Scalar&, Scalar&, GpuMat&) { throw_nogpu(); }
-double cv::gpu::norm(const GpuMat&, int) { throw_nogpu(); return 0.0; }
-double cv::gpu::norm(const GpuMat&, int, GpuMat&) { throw_nogpu(); return 0.0; }
-double cv::gpu::norm(const GpuMat&, int, const GpuMat&, GpuMat&) { throw_nogpu(); return 0.0; }
-double cv::gpu::norm(const GpuMat&, const GpuMat&, int) { throw_nogpu(); return 0.0; }
-Scalar cv::gpu::sum(const GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::sum(const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::sum(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::absSum(const GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::absSum(const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::absSum(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::sqrSum(const GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::sqrSum(const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-Scalar cv::gpu::sqrSum(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); return Scalar(); }
-void cv::gpu::minMax(const GpuMat&, double*, double*, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::minMax(const GpuMat&, double*, double*, const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::minMaxLoc(const GpuMat&, double*, double*, Point*, Point*, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::minMaxLoc(const GpuMat&, double*, double*, Point*, Point*, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
-int cv::gpu::countNonZero(const GpuMat&) { throw_nogpu(); return 0; }
-int cv::gpu::countNonZero(const GpuMat&, GpuMat&) { throw_nogpu(); return 0; }
-void cv::gpu::reduce(const GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
+double cv::gpu::norm(InputArray, int, InputArray, GpuMat&) { throw_no_cuda(); return 0.0; }
+double cv::gpu::norm(InputArray, InputArray, GpuMat&, int) { throw_no_cuda(); return 0.0; }
+
+Scalar cv::gpu::sum(InputArray, InputArray, GpuMat&) { throw_no_cuda(); return Scalar(); }
+Scalar cv::gpu::absSum(InputArray, InputArray, GpuMat&) { throw_no_cuda(); return Scalar(); }
+Scalar cv::gpu::sqrSum(InputArray, InputArray, GpuMat&) { throw_no_cuda(); return Scalar(); }
+
+void cv::gpu::minMax(InputArray, double*, double*, InputArray, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::minMaxLoc(InputArray, double*, double*, Point*, Point*, InputArray, GpuMat&, GpuMat&) { throw_no_cuda(); }
+
+int cv::gpu::countNonZero(InputArray, GpuMat&) { throw_no_cuda(); return 0; }
+
+void cv::gpu::reduce(InputArray, OutputArray, int, int, int, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::meanStdDev(InputArray, Scalar&, Scalar&, GpuMat&) { throw_no_cuda(); }
+
+void cv::gpu::rectStdDev(InputArray, InputArray, OutputArray, Rect, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::normalize(InputArray, OutputArray, double, double, int, int, InputArray, GpuMat&, GpuMat&) { throw_no_cuda(); }
+
+void cv::gpu::integral(InputArray, OutputArray, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::sqrIntegral(InputArray, OutputArray, GpuMat&, Stream&) { throw_no_cuda(); }
 
 #else
 
@@ -108,104 +108,79 @@ namespace
     };
 }
 
-
-////////////////////////////////////////////////////////////////////////
-// meanStdDev
-
-void cv::gpu::meanStdDev(const GpuMat& src, Scalar& mean, Scalar& stddev)
-{
-    GpuMat buf;
-    meanStdDev(src, mean, stddev, buf);
-}
-
-void cv::gpu::meanStdDev(const GpuMat& src, Scalar& mean, Scalar& stddev, GpuMat& buf)
-{
-    CV_Assert(src.type() == CV_8UC1);
-
-    if (!deviceSupports(FEATURE_SET_COMPUTE_13))
-        CV_Error(CV_StsNotImplemented, "Not sufficient compute capebility");
-
-    NppiSize sz;
-    sz.width  = src.cols;
-    sz.height = src.rows;
-
-    DeviceBuffer dbuf(2);
-
-    int bufSize;
-#if (CUDA_VERSION <= 4020)
-    nppSafeCall( nppiMeanStdDev8uC1RGetBufferHostSize(sz, &bufSize) );
-#else
-    nppSafeCall( nppiMeanStdDevGetBufferHostSize_8u_C1R(sz, &bufSize) );
-#endif
-
-    ensureSizeIsEnough(1, bufSize, CV_8UC1, buf);
-
-    nppSafeCall( nppiMean_StdDev_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), sz, buf.ptr<Npp8u>(), dbuf, (double*)dbuf + 1) );
-
-    cudaSafeCall( cudaDeviceSynchronize() );
-
-    double* ptrs[2] = {mean.val, stddev.val};
-    dbuf.download(ptrs);
-}
-
 ////////////////////////////////////////////////////////////////////////
 // norm
 
-double cv::gpu::norm(const GpuMat& src, int normType)
+double cv::gpu::norm(InputArray _src, int normType, InputArray _mask, GpuMat& buf)
 {
-    GpuMat buf;
-    return norm(src, normType, GpuMat(), buf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-double cv::gpu::norm(const GpuMat& src, int normType, GpuMat& buf)
-{
-    return norm(src, normType, GpuMat(), buf);
-}
-
-double cv::gpu::norm(const GpuMat& src, int normType, const GpuMat& mask, GpuMat& buf)
-{
-    CV_Assert(normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2);
-    CV_Assert(mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src.size() && src.channels() == 1));
+    CV_Assert( normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 );
+    CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == src.size() && src.channels() == 1) );
 
     GpuMat src_single_channel = src.reshape(1);
 
     if (normType == NORM_L1)
-        return absSum(src_single_channel, mask, buf)[0];
+        return gpu::absSum(src_single_channel, mask, buf)[0];
 
     if (normType == NORM_L2)
-        return std::sqrt(sqrSum(src_single_channel, mask, buf)[0]);
+        return std::sqrt(gpu::sqrSum(src_single_channel, mask, buf)[0]);
 
     // NORM_INF
     double min_val, max_val;
-    minMax(src_single_channel, &min_val, &max_val, mask, buf);
+    gpu::minMax(src_single_channel, &min_val, &max_val, mask, buf);
     return std::max(std::abs(min_val), std::abs(max_val));
 }
 
-double cv::gpu::norm(const GpuMat& src1, const GpuMat& src2, int normType)
+double cv::gpu::norm(InputArray _src1, InputArray _src2, GpuMat& buf, int normType)
 {
-    CV_Assert(src1.type() == CV_8UC1);
-    CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
-    CV_Assert(normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2);
+#if CUDA_VERSION < 5050
+    (void) buf;
 
-    typedef NppStatus (*npp_norm_diff_func_t)(const Npp8u* pSrc1, int nSrcStep1, const Npp8u* pSrc2, int nSrcStep2,
-        NppiSize oSizeROI, Npp64f* pRetVal);
+    typedef NppStatus (*func_t)(const Npp8u* pSrc1, int nSrcStep1, const Npp8u* pSrc2, int nSrcStep2, NppiSize oSizeROI, Npp64f* pRetVal);
 
-    static const npp_norm_diff_func_t npp_norm_diff_func[] = {nppiNormDiff_Inf_8u_C1R, nppiNormDiff_L1_8u_C1R, nppiNormDiff_L2_8u_C1R};
+    static const func_t funcs[] = {nppiNormDiff_Inf_8u_C1R, nppiNormDiff_L1_8u_C1R, nppiNormDiff_L2_8u_C1R};
+#else
+    typedef NppStatus (*func_t)(const Npp8u* pSrc1, int nSrcStep1, const Npp8u* pSrc2, int nSrcStep2,
+        NppiSize oSizeROI, Npp64f* pRetVal, Npp8u * pDeviceBuffer);
+
+    typedef NppStatus (*buf_size_func_t)(NppiSize oSizeROI, int* hpBufferSize);
+
+    static const func_t funcs[] = {nppiNormDiff_Inf_8u_C1R, nppiNormDiff_L1_8u_C1R, nppiNormDiff_L2_8u_C1R};
+
+    static const buf_size_func_t buf_size_funcs[] = {nppiNormDiffInfGetBufferHostSize_8u_C1R, nppiNormDiffL1GetBufferHostSize_8u_C1R, nppiNormDiffL2GetBufferHostSize_8u_C1R};
+#endif
+
+    GpuMat src1 = _src1.getGpuMat();
+    GpuMat src2 = _src2.getGpuMat();
+
+    CV_Assert( src1.type() == CV_8UC1 );
+    CV_Assert( src1.size() == src2.size() && src1.type() == src2.type() );
+    CV_Assert( normType == NORM_INF || normType == NORM_L1 || normType == NORM_L2 );
 
     NppiSize sz;
     sz.width  = src1.cols;
     sz.height = src1.rows;
 
-    int funcIdx = normType >> 1;
-
-    double retVal;
+    const int funcIdx = normType >> 1;
 
     DeviceBuffer dbuf;
 
-    nppSafeCall( npp_norm_diff_func[funcIdx](src1.ptr<Npp8u>(), static_cast<int>(src1.step), src2.ptr<Npp8u>(), static_cast<int>(src2.step), sz, dbuf) );
+#if CUDA_VERSION < 5050
+    nppSafeCall( funcs[funcIdx](src1.ptr<Npp8u>(), static_cast<int>(src1.step), src2.ptr<Npp8u>(), static_cast<int>(src2.step), sz, dbuf) );
+#else
+    int bufSize;
+    buf_size_funcs[funcIdx](sz, &bufSize);
+
+    ensureSizeIsEnough(1, bufSize, CV_8UC1, buf);
+
+    nppSafeCall( funcs[funcIdx](src1.ptr<Npp8u>(), static_cast<int>(src1.step), src2.ptr<Npp8u>(), static_cast<int>(src2.step), sz, dbuf, buf.data) );
+#endif
 
     cudaSafeCall( cudaDeviceSynchronize() );
 
+    double retVal;
     dbuf.download(&retVal);
 
     return retVal;
@@ -228,19 +203,11 @@ namespace sum
     void runSqr(PtrStepSzb src, void* buf, double* sum, PtrStepSzb mask);
 }
 
-Scalar cv::gpu::sum(const GpuMat& src)
+Scalar cv::gpu::sum(InputArray _src, InputArray _mask, GpuMat& buf)
 {
-    GpuMat buf;
-    return sum(src, GpuMat(), buf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-Scalar cv::gpu::sum(const GpuMat& src, GpuMat& buf)
-{
-    return sum(src, GpuMat(), buf);
-}
-
-Scalar cv::gpu::sum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
-{
     typedef void (*func_t)(PtrStepSzb src, void* buf, double* sum, PtrStepSzb mask);
     static const func_t funcs[7][5] =
     {
@@ -258,7 +225,7 @@ Scalar cv::gpu::sum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size buf_size;
@@ -274,19 +241,11 @@ Scalar cv::gpu::sum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
     return Scalar(result[0], result[1], result[2], result[3]);
 }
 
-Scalar cv::gpu::absSum(const GpuMat& src)
+Scalar cv::gpu::absSum(InputArray _src, InputArray _mask, GpuMat& buf)
 {
-    GpuMat buf;
-    return absSum(src, GpuMat(), buf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-Scalar cv::gpu::absSum(const GpuMat& src, GpuMat& buf)
-{
-    return absSum(src, GpuMat(), buf);
-}
-
-Scalar cv::gpu::absSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
-{
     typedef void (*func_t)(PtrStepSzb src, void* buf, double* sum, PtrStepSzb mask);
     static const func_t funcs[7][5] =
     {
@@ -304,7 +263,7 @@ Scalar cv::gpu::absSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size buf_size;
@@ -320,19 +279,11 @@ Scalar cv::gpu::absSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
     return Scalar(result[0], result[1], result[2], result[3]);
 }
 
-Scalar cv::gpu::sqrSum(const GpuMat& src)
+Scalar cv::gpu::sqrSum(InputArray _src, InputArray _mask, GpuMat& buf)
 {
-    GpuMat buf;
-    return sqrSum(src, GpuMat(), buf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-Scalar cv::gpu::sqrSum(const GpuMat& src, GpuMat& buf)
-{
-    return sqrSum(src, GpuMat(), buf);
-}
-
-Scalar cv::gpu::sqrSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
-{
     typedef void (*func_t)(PtrStepSzb src, void* buf, double* sum, PtrStepSzb mask);
     static const func_t funcs[7][5] =
     {
@@ -350,7 +301,7 @@ Scalar cv::gpu::sqrSum(const GpuMat& src, const GpuMat& mask, GpuMat& buf)
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size buf_size;
@@ -377,14 +328,11 @@ namespace minMax
     void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
 }
 
-void cv::gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const GpuMat& mask)
+void cv::gpu::minMax(InputArray _src, double* minVal, double* maxVal, InputArray _mask, GpuMat& buf)
 {
-    GpuMat buf;
-    minMax(src, minVal, maxVal, mask, buf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-void cv::gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const GpuMat& mask, GpuMat& buf)
-{
     typedef void (*func_t)(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, PtrStepb buf);
     static const func_t funcs[] =
     {
@@ -403,7 +351,7 @@ void cv::gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const Gp
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size buf_size;
@@ -427,15 +375,12 @@ namespace minMaxLoc
     void run(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
 }
 
-void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc, const GpuMat& mask)
+void cv::gpu::minMaxLoc(InputArray _src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc,
+                        InputArray _mask, GpuMat& valBuf, GpuMat& locBuf)
 {
-    GpuMat valBuf, locBuf;
-    minMaxLoc(src, minVal, maxVal, minLoc, maxLoc, mask, valBuf, locBuf);
-}
+    GpuMat src = _src.getGpuMat();
+    GpuMat mask = _mask.getGpuMat();
 
-void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc,
-                        const GpuMat& mask, GpuMat& valBuf, GpuMat& locBuf)
-{
     typedef void (*func_t)(const PtrStepSzb src, const PtrStepb mask, double* minval, double* maxval, int* minloc, int* maxloc, PtrStepb valbuf, PtrStep<unsigned int> locbuf);
     static const func_t funcs[] =
     {
@@ -454,7 +399,7 @@ void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size valbuf_size, locbuf_size;
@@ -480,14 +425,10 @@ namespace countNonZero
     int run(const PtrStepSzb src, PtrStep<unsigned int> buf);
 }
 
-int cv::gpu::countNonZero(const GpuMat& src)
+int cv::gpu::countNonZero(InputArray _src, GpuMat& buf)
 {
-    GpuMat buf;
-    return countNonZero(src, buf);
-}
+    GpuMat src = _src.getGpuMat();
 
-int cv::gpu::countNonZero(const GpuMat& src, GpuMat& buf)
-{
     typedef int (*func_t)(const PtrStepSzb src, PtrStep<unsigned int> buf);
     static const func_t funcs[] =
     {
@@ -505,7 +446,7 @@ int cv::gpu::countNonZero(const GpuMat& src, GpuMat& buf)
     if (src.depth() == CV_64F)
     {
         if (!deviceSupports(NATIVE_DOUBLE))
-            CV_Error(CV_StsUnsupportedFormat, "The device doesn't support double");
+            CV_Error(cv::Error::StsUnsupportedFormat, "The device doesn't support double");
     }
 
     Size buf_size;
@@ -529,16 +470,19 @@ namespace reduce
     void cols(PtrStepSzb src, void* dst, int cn, int op, cudaStream_t stream);
 }
 
-void cv::gpu::reduce(const GpuMat& src, GpuMat& dst, int dim, int reduceOp, int dtype, Stream& stream)
+void cv::gpu::reduce(InputArray _src, OutputArray _dst, int dim, int reduceOp, int dtype, Stream& stream)
 {
+    GpuMat src = _src.getGpuMat();
+
     CV_Assert( src.channels() <= 4 );
     CV_Assert( dim == 0 || dim == 1 );
-    CV_Assert( reduceOp == CV_REDUCE_SUM || reduceOp == CV_REDUCE_AVG || reduceOp == CV_REDUCE_MAX || reduceOp == CV_REDUCE_MIN );
+    CV_Assert( reduceOp == REDUCE_SUM || reduceOp == REDUCE_AVG || reduceOp == REDUCE_MAX || reduceOp == REDUCE_MIN );
 
     if (dtype < 0)
         dtype = src.depth();
 
-    dst.create(1, dim == 0 ? src.cols : src.rows, CV_MAKE_TYPE(CV_MAT_DEPTH(dtype), src.channels()));
+    _dst.create(1, dim == 0 ? src.cols : src.rows, CV_MAKE_TYPE(CV_MAT_DEPTH(dtype), src.channels()));
+    GpuMat dst = _dst.getGpuMat();
 
     if (dim == 0)
     {
@@ -613,7 +557,7 @@ void cv::gpu::reduce(const GpuMat& src, GpuMat& dst, int dim, int reduceOp, int
         const func_t func = funcs[src.depth()][dst.depth()];
 
         if (!func)
-            CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of input and output array formats");
+            CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of input and output array formats");
 
         func(src.reshape(1), dst.data, reduceOp, StreamAccessor::getStream(stream));
     }
@@ -690,10 +634,230 @@ void cv::gpu::reduce(const GpuMat& src, GpuMat& dst, int dim, int reduceOp, int
         const func_t func = funcs[src.depth()][dst.depth()];
 
         if (!func)
-            CV_Error(CV_StsUnsupportedFormat, "Unsupported combination of input and output array formats");
+            CV_Error(cv::Error::StsUnsupportedFormat, "Unsupported combination of input and output array formats");
 
         func(src, dst.data, src.channels(), reduceOp, StreamAccessor::getStream(stream));
     }
 }
 
+////////////////////////////////////////////////////////////////////////
+// meanStdDev
+
+void cv::gpu::meanStdDev(InputArray _src, Scalar& mean, Scalar& stddev, GpuMat& buf)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC1 );
+
+    if (!deviceSupports(FEATURE_SET_COMPUTE_13))
+        CV_Error(cv::Error::StsNotImplemented, "Not sufficient compute capebility");
+
+    NppiSize sz;
+    sz.width  = src.cols;
+    sz.height = src.rows;
+
+    DeviceBuffer dbuf(2);
+
+    int bufSize;
+#if (CUDA_VERSION <= 4020)
+    nppSafeCall( nppiMeanStdDev8uC1RGetBufferHostSize(sz, &bufSize) );
+#else
+    nppSafeCall( nppiMeanStdDevGetBufferHostSize_8u_C1R(sz, &bufSize) );
+#endif
+
+    ensureSizeIsEnough(1, bufSize, CV_8UC1, buf);
+
+    nppSafeCall( nppiMean_StdDev_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), sz, buf.ptr<Npp8u>(), dbuf, (double*)dbuf + 1) );
+
+    cudaSafeCall( cudaDeviceSynchronize() );
+
+    double* ptrs[2] = {mean.val, stddev.val};
+    dbuf.download(ptrs);
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// rectStdDev
+
+void cv::gpu::rectStdDev(InputArray _src, InputArray _sqr, OutputArray _dst, Rect rect, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+    GpuMat sqr = _sqr.getGpuMat();
+
+    CV_Assert( src.type() == CV_32SC1 && sqr.type() == CV_64FC1 );
+
+    _dst.create(src.size(), CV_32FC1);
+    GpuMat dst = _dst.getGpuMat();
+
+    NppiSize sz;
+    sz.width = src.cols;
+    sz.height = src.rows;
+
+    NppiRect nppRect;
+    nppRect.height = rect.height;
+    nppRect.width = rect.width;
+    nppRect.x = rect.x;
+    nppRect.y = rect.y;
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    NppStreamHandler h(stream);
+
+    nppSafeCall( nppiRectStdDev_32s32f_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), sqr.ptr<Npp64f>(), static_cast<int>(sqr.step),
+                dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, nppRect) );
+
+    if (stream == 0)
+        cudaSafeCall( cudaDeviceSynchronize() );
+}
+
+////////////////////////////////////////////////////////////////////////
+// normalize
+
+void cv::gpu::normalize(InputArray _src, OutputArray dst, double a, double b, int norm_type, int dtype, InputArray mask, GpuMat& norm_buf, GpuMat& cvt_buf)
+{
+    GpuMat src = _src.getGpuMat();
+
+    double scale = 1, shift = 0;
+
+    if (norm_type == NORM_MINMAX)
+    {
+        double smin = 0, smax = 0;
+        double dmin = std::min(a, b), dmax = std::max(a, b);
+        gpu::minMax(src, &smin, &smax, mask, norm_buf);
+        scale = (dmax - dmin) * (smax - smin > std::numeric_limits<double>::epsilon() ? 1.0 / (smax - smin) : 0.0);
+        shift = dmin - smin * scale;
+    }
+    else if (norm_type == NORM_L2 || norm_type == NORM_L1 || norm_type == NORM_INF)
+    {
+        scale = gpu::norm(src, norm_type, mask, norm_buf);
+        scale = scale > std::numeric_limits<double>::epsilon() ? a / scale : 0.0;
+        shift = 0;
+    }
+    else
+    {
+        CV_Error(cv::Error::StsBadArg, "Unknown/unsupported norm type");
+    }
+
+    if (mask.empty())
+    {
+        src.convertTo(dst, dtype, scale, shift);
+    }
+    else
+    {
+        src.convertTo(cvt_buf, dtype, scale, shift);
+        cvt_buf.copyTo(dst, mask);
+    }
+}
+
+////////////////////////////////////////////////////////////////////////
+// integral
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void shfl_integral_gpu(const PtrStepSzb& img, PtrStepSz<unsigned int> integral, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::integral(InputArray _src, OutputArray _dst, GpuMat& buffer, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC1 );
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    cv::Size whole;
+    cv::Point offset;
+    src.locateROI(whole, offset);
+
+    if (deviceSupports(WARP_SHUFFLE_FUNCTIONS) && src.cols <= 2048
+        && offset.x % 16 == 0 && ((src.cols + 63) / 64) * 64 <= (static_cast<int>(src.step) - offset.x))
+    {
+        ensureSizeIsEnough(((src.rows + 7) / 8) * 8, ((src.cols + 63) / 64) * 64, CV_32SC1, buffer);
+
+        cv::gpu::cudev::imgproc::shfl_integral_gpu(src, buffer, stream);
+
+        _dst.create(src.rows + 1, src.cols + 1, CV_32SC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        dst.setTo(Scalar::all(0), _stream);
+
+        GpuMat inner = dst(Rect(1, 1, src.cols, src.rows));
+        GpuMat res = buffer(Rect(0, 0, src.cols, src.rows));
+
+        res.copyTo(inner, _stream);
+    }
+    else
+    {
+    #ifndef HAVE_OPENCV_GPULEGACY
+        throw_no_cuda();
+    #else
+        _dst.create(src.rows + 1, src.cols + 1, CV_32SC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        NcvSize32u roiSize;
+        roiSize.width = src.cols;
+        roiSize.height = src.rows;
+
+        cudaDeviceProp prop;
+        cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
+
+        Ncv32u bufSize;
+        ncvSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize, prop) );
+        ensureSizeIsEnough(1, bufSize, CV_8UC1, buffer);
+
+        NppStStreamHandler h(stream);
+
+        ncvSafeCall( nppiStIntegral_8u32u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>()), static_cast<int>(src.step),
+            dst.ptr<Ncv32u>(), static_cast<int>(dst.step), roiSize, buffer.ptr<Ncv8u>(), bufSize, prop) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    #endif
+    }
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// sqrIntegral
+
+void cv::gpu::sqrIntegral(InputArray _src, OutputArray _dst, GpuMat& buf, Stream& _stream)
+{
+#ifndef HAVE_OPENCV_GPULEGACY
+    (void) _src;
+    (void) _dst;
+    (void) _stream;
+    throw_no_cuda();
+#else
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8U );
+
+    NcvSize32u roiSize;
+    roiSize.width = src.cols;
+    roiSize.height = src.rows;
+
+    cudaDeviceProp prop;
+    cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
+
+    Ncv32u bufSize;
+    ncvSafeCall(nppiStSqrIntegralGetSize_8u64u(roiSize, &bufSize, prop));
+
+    ensureSizeIsEnough(1, bufSize, CV_8U, buf);
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+    NppStStreamHandler h(stream);
+
+    _dst.create(src.rows + 1, src.cols + 1, CV_64F);
+    GpuMat dst = _dst.getGpuMat();
+
+    ncvSafeCall(nppiStSqrIntegral_8u64u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>(0)), static_cast<int>(src.step),
+            dst.ptr<Ncv64u>(0), static_cast<int>(dst.step), roiSize, buf.ptr<Ncv8u>(0), bufSize, prop));
+
+    if (stream == 0)
+        cudaSafeCall( cudaDeviceSynchronize() );
+#endif
+}
+
 #endif
diff --git a/modules/gpuarithm/test/test_arithm.cpp b/modules/gpuarithm/test/test_arithm.cpp
new file mode 100644
index 0000000000..0534e219d8
--- /dev/null
+++ b/modules/gpuarithm/test/test_arithm.cpp
@@ -0,0 +1,441 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+//////////////////////////////////////////////////////////////////////////////
+// GEMM
+
+#ifdef HAVE_CUBLAS
+
+CV_FLAGS(GemmFlags, 0, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T);
+#define ALL_GEMM_FLAGS testing::Values(GemmFlags(0), GemmFlags(cv::GEMM_1_T), GemmFlags(cv::GEMM_2_T), GemmFlags(cv::GEMM_3_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_3_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T | cv::GEMM_3_T))
+
+PARAM_TEST_CASE(GEMM, cv::gpu::DeviceInfo, cv::Size, MatType, GemmFlags, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int flags;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        flags = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(GEMM, Accuracy)
+{
+    cv::Mat src1 = randomMat(size, type, -10.0, 10.0);
+    cv::Mat src2 = randomMat(size, type, -10.0, 10.0);
+    cv::Mat src3 = randomMat(size, type, -10.0, 10.0);
+    double alpha = randomDouble(-10.0, 10.0);
+    double beta = randomDouble(-10.0, 10.0);
+
+    if (CV_MAT_DEPTH(type) == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::gemm(loadMat(src1), loadMat(src2), alpha, loadMat(src3), beta, dst, flags);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else if (type == CV_64FC2 && flags != 0)
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::gemm(loadMat(src1), loadMat(src2), alpha, loadMat(src3), beta, dst, flags);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+        cv::gpu::gemm(loadMat(src1, useRoi), loadMat(src2, useRoi), alpha, loadMat(src3, useRoi), beta, dst, flags);
+
+        cv::Mat dst_gold;
+        cv::gemm(src1, src2, alpha, src3, beta, dst_gold, flags);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, CV_MAT_DEPTH(type) == CV_32F ? 1e-1 : 1e-10);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, GEMM, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_32FC1), MatType(CV_32FC2), MatType(CV_64FC1), MatType(CV_64FC2)),
+    ALL_GEMM_FLAGS,
+    WHOLE_SUBMAT));
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// Integral
+
+PARAM_TEST_CASE(Integral, cv::gpu::DeviceInfo, cv::Size, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        useRoi = GET_PARAM(2);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Integral, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    cv::gpu::GpuMat dst = createMat(cv::Size(src.cols + 1, src.rows + 1), CV_32SC1, useRoi);
+    cv::gpu::integral(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::integral(src, dst_gold, CV_32S);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Integral, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////
+// MulSpectrums
+
+CV_FLAGS(DftFlags, 0, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
+
+PARAM_TEST_CASE(MulSpectrums, cv::gpu::DeviceInfo, cv::Size, DftFlags)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int flag;
+
+    cv::Mat a, b;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        flag = GET_PARAM(2);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        a = randomMat(size, CV_32FC2);
+        b = randomMat(size, CV_32FC2);
+    }
+};
+
+GPU_TEST_P(MulSpectrums, Simple)
+{
+    cv::gpu::GpuMat c;
+    cv::gpu::mulSpectrums(loadMat(a), loadMat(b), c, flag, false);
+
+    cv::Mat c_gold;
+    cv::mulSpectrums(a, b, c_gold, flag, false);
+
+    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
+}
+
+GPU_TEST_P(MulSpectrums, Scaled)
+{
+    float scale = 1.f / size.area();
+
+    cv::gpu::GpuMat c;
+    cv::gpu::mulAndScaleSpectrums(loadMat(a), loadMat(b), c, flag, scale, false);
+
+    cv::Mat c_gold;
+    cv::mulSpectrums(a, b, c_gold, flag, false);
+    c_gold.convertTo(c_gold, c_gold.type(), scale);
+
+    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, MulSpectrums, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(DftFlags(0), DftFlags(cv::DFT_ROWS))));
+
+////////////////////////////////////////////////////////////////////////////
+// Dft
+
+struct Dft : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+namespace
+{
+    void testC2C(const std::string& hint, int cols, int rows, int flags, bool inplace)
+    {
+        SCOPED_TRACE(hint);
+
+        cv::Mat a = randomMat(cv::Size(cols, rows), CV_32FC2, 0.0, 10.0);
+
+        cv::Mat b_gold;
+        cv::dft(a, b_gold, flags);
+
+        cv::gpu::GpuMat d_b;
+        cv::gpu::GpuMat d_b_data;
+        if (inplace)
+        {
+            d_b_data.create(1, a.size().area(), CV_32FC2);
+            d_b = cv::gpu::GpuMat(a.rows, a.cols, CV_32FC2, d_b_data.ptr(), a.cols * d_b_data.elemSize());
+        }
+        cv::gpu::dft(loadMat(a), d_b, cv::Size(cols, rows), flags);
+
+        EXPECT_TRUE(!inplace || d_b.ptr() == d_b_data.ptr());
+        ASSERT_EQ(CV_32F, d_b.depth());
+        ASSERT_EQ(2, d_b.channels());
+        EXPECT_MAT_NEAR(b_gold, cv::Mat(d_b), rows * cols * 1e-4);
+    }
+}
+
+GPU_TEST_P(Dft, C2C)
+{
+    int cols = randomInt(2, 100);
+    int rows = randomInt(2, 100);
+
+    for (int i = 0; i < 2; ++i)
+    {
+        bool inplace = i != 0;
+
+        testC2C("no flags", cols, rows, 0, inplace);
+        testC2C("no flags 0 1", cols, rows + 1, 0, inplace);
+        testC2C("no flags 1 0", cols, rows + 1, 0, inplace);
+        testC2C("no flags 1 1", cols + 1, rows, 0, inplace);
+        testC2C("DFT_INVERSE", cols, rows, cv::DFT_INVERSE, inplace);
+        testC2C("DFT_ROWS", cols, rows, cv::DFT_ROWS, inplace);
+        testC2C("single col", 1, rows, 0, inplace);
+        testC2C("single row", cols, 1, 0, inplace);
+        testC2C("single col inversed", 1, rows, cv::DFT_INVERSE, inplace);
+        testC2C("single row inversed", cols, 1, cv::DFT_INVERSE, inplace);
+        testC2C("single row DFT_ROWS", cols, 1, cv::DFT_ROWS, inplace);
+        testC2C("size 1 2", 1, 2, 0, inplace);
+        testC2C("size 2 1", 2, 1, 0, inplace);
+    }
+}
+
+namespace
+{
+    void testR2CThenC2R(const std::string& hint, int cols, int rows, bool inplace)
+    {
+        SCOPED_TRACE(hint);
+
+        cv::Mat a = randomMat(cv::Size(cols, rows), CV_32FC1, 0.0, 10.0);
+
+        cv::gpu::GpuMat d_b, d_c;
+        cv::gpu::GpuMat d_b_data, d_c_data;
+        if (inplace)
+        {
+            if (a.cols == 1)
+            {
+                d_b_data.create(1, (a.rows / 2 + 1) * a.cols, CV_32FC2);
+                d_b = cv::gpu::GpuMat(a.rows / 2 + 1, a.cols, CV_32FC2, d_b_data.ptr(), a.cols * d_b_data.elemSize());
+            }
+            else
+            {
+                d_b_data.create(1, a.rows * (a.cols / 2 + 1), CV_32FC2);
+                d_b = cv::gpu::GpuMat(a.rows, a.cols / 2 + 1, CV_32FC2, d_b_data.ptr(), (a.cols / 2 + 1) * d_b_data.elemSize());
+            }
+            d_c_data.create(1, a.size().area(), CV_32F);
+            d_c = cv::gpu::GpuMat(a.rows, a.cols, CV_32F, d_c_data.ptr(), a.cols * d_c_data.elemSize());
+        }
+
+        cv::gpu::dft(loadMat(a), d_b, cv::Size(cols, rows), 0);
+        cv::gpu::dft(d_b, d_c, cv::Size(cols, rows), cv::DFT_REAL_OUTPUT | cv::DFT_SCALE);
+
+        EXPECT_TRUE(!inplace || d_b.ptr() == d_b_data.ptr());
+        EXPECT_TRUE(!inplace || d_c.ptr() == d_c_data.ptr());
+        ASSERT_EQ(CV_32F, d_c.depth());
+        ASSERT_EQ(1, d_c.channels());
+
+        cv::Mat c(d_c);
+        EXPECT_MAT_NEAR(a, c, rows * cols * 1e-5);
+    }
+}
+
+GPU_TEST_P(Dft, R2CThenC2R)
+{
+    int cols = randomInt(2, 100);
+    int rows = randomInt(2, 100);
+
+    testR2CThenC2R("sanity", cols, rows, false);
+    testR2CThenC2R("sanity 0 1", cols, rows + 1, false);
+    testR2CThenC2R("sanity 1 0", cols + 1, rows, false);
+    testR2CThenC2R("sanity 1 1", cols + 1, rows + 1, false);
+    testR2CThenC2R("single col", 1, rows, false);
+    testR2CThenC2R("single col 1", 1, rows + 1, false);
+    testR2CThenC2R("single row", cols, 1, false);
+    testR2CThenC2R("single row 1", cols + 1, 1, false);
+
+    testR2CThenC2R("sanity", cols, rows, true);
+    testR2CThenC2R("sanity 0 1", cols, rows + 1, true);
+    testR2CThenC2R("sanity 1 0", cols + 1, rows, true);
+    testR2CThenC2R("sanity 1 1", cols + 1, rows + 1, true);
+    testR2CThenC2R("single row", cols, 1, true);
+    testR2CThenC2R("single row 1", cols + 1, 1, true);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Dft, ALL_DEVICES);
+
+////////////////////////////////////////////////////////
+// Convolve
+
+namespace
+{
+    void convolveDFT(const cv::Mat& A, const cv::Mat& B, cv::Mat& C, bool ccorr = false)
+    {
+        // reallocate the output array if needed
+        C.create(std::abs(A.rows - B.rows) + 1, std::abs(A.cols - B.cols) + 1, A.type());
+        cv::Size dftSize;
+
+        // compute the size of DFT transform
+        dftSize.width = cv::getOptimalDFTSize(A.cols + B.cols - 1);
+        dftSize.height = cv::getOptimalDFTSize(A.rows + B.rows - 1);
+
+        // allocate temporary buffers and initialize them with 0s
+        cv::Mat tempA(dftSize, A.type(), cv::Scalar::all(0));
+        cv::Mat tempB(dftSize, B.type(), cv::Scalar::all(0));
+
+        // copy A and B to the top-left corners of tempA and tempB, respectively
+        cv::Mat roiA(tempA, cv::Rect(0, 0, A.cols, A.rows));
+        A.copyTo(roiA);
+        cv::Mat roiB(tempB, cv::Rect(0, 0, B.cols, B.rows));
+        B.copyTo(roiB);
+
+        // now transform the padded A & B in-place;
+        // use "nonzeroRows" hint for faster processing
+        cv::dft(tempA, tempA, 0, A.rows);
+        cv::dft(tempB, tempB, 0, B.rows);
+
+        // multiply the spectrums;
+        // the function handles packed spectrum representations well
+        cv::mulSpectrums(tempA, tempB, tempA, 0, ccorr);
+
+        // transform the product back from the frequency domain.
+        // Even though all the result rows will be non-zero,
+        // you need only the first C.rows of them, and thus you
+        // pass nonzeroRows == C.rows
+        cv::dft(tempA, tempA, cv::DFT_INVERSE + cv::DFT_SCALE, C.rows);
+
+        // now copy the result back to C.
+        tempA(cv::Rect(0, 0, C.cols, C.rows)).copyTo(C);
+    }
+
+    IMPLEMENT_PARAM_CLASS(KSize, int)
+    IMPLEMENT_PARAM_CLASS(Ccorr, bool)
+}
+
+PARAM_TEST_CASE(Convolve, cv::gpu::DeviceInfo, cv::Size, KSize, Ccorr)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int ksize;
+    bool ccorr;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        ksize = GET_PARAM(2);
+        ccorr = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Convolve, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_32FC1, 0.0, 100.0);
+    cv::Mat kernel = randomMat(cv::Size(ksize, ksize), CV_32FC1, 0.0, 1.0);
+
+    cv::Ptr<cv::gpu::Convolution> conv = cv::gpu::createConvolution();
+
+    cv::gpu::GpuMat dst;
+    conv->convolve(loadMat(src), loadMat(kernel), dst, ccorr);
+
+    cv::Mat dst_gold;
+    convolveDFT(src, kernel, dst_gold, ccorr);
+
+    EXPECT_MAT_NEAR(dst, dst_gold, 1e-1);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Convolve, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(KSize(3), KSize(7), KSize(11), KSize(17), KSize(19), KSize(23), KSize(45)),
+    testing::Values(Ccorr(false), Ccorr(true))));
+
+#endif // HAVE_CUBLAS
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuarithm/test/test_core.cpp b/modules/gpuarithm/test/test_core.cpp
new file mode 100644
index 0000000000..d465aa4634
--- /dev/null
+++ b/modules/gpuarithm/test/test_core.cpp
@@ -0,0 +1,419 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// Merge
+
+PARAM_TEST_CASE(Merge, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    int channels;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        channels = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Merge, Accuracy)
+{
+    std::vector<cv::Mat> src;
+    src.reserve(channels);
+    for (int i = 0; i < channels; ++i)
+        src.push_back(cv::Mat(size, depth, cv::Scalar::all(i)));
+
+    std::vector<cv::gpu::GpuMat> d_src;
+    for (int i = 0; i < channels; ++i)
+        d_src.push_back(loadMat(src[i], useRoi));
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::merge(d_src, dst);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst;
+        cv::gpu::merge(d_src, dst);
+
+        cv::Mat dst_gold;
+        cv::merge(src, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Merge, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    testing::Values(1, 2, 3, 4),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Split
+
+PARAM_TEST_CASE(Split, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    int channels;
+    bool useRoi;
+
+    int type;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        channels = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        type = CV_MAKE_TYPE(depth, channels);
+    }
+};
+
+GPU_TEST_P(Split, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            std::vector<cv::gpu::GpuMat> dst;
+            cv::gpu::split(loadMat(src), dst);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        std::vector<cv::gpu::GpuMat> dst;
+        cv::gpu::split(loadMat(src, useRoi), dst);
+
+        std::vector<cv::Mat> dst_gold;
+        cv::split(src, dst_gold);
+
+        ASSERT_EQ(dst_gold.size(), dst.size());
+
+        for (size_t i = 0; i < dst_gold.size(); ++i)
+        {
+            EXPECT_MAT_NEAR(dst_gold[i], dst[i], 0.0);
+        }
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Split, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    testing::Values(1, 2, 3, 4),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Transpose
+
+PARAM_TEST_CASE(Transpose, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Transpose, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+
+    if (CV_MAT_DEPTH(type) == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::transpose(loadMat(src), dst);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(cv::Size(size.height, size.width), type, useRoi);
+        cv::gpu::transpose(loadMat(src, useRoi), dst);
+
+        cv::Mat dst_gold;
+        cv::transpose(src, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Transpose, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1),
+                    MatType(CV_8UC4),
+                    MatType(CV_16UC2),
+                    MatType(CV_16SC2),
+                    MatType(CV_32SC1),
+                    MatType(CV_32SC2),
+                    MatType(CV_64FC1)),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Flip
+
+enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
+CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
+#define ALL_FLIP_CODES testing::Values(FlipCode(FLIP_BOTH), FlipCode(FLIP_X), FlipCode(FLIP_Y))
+
+PARAM_TEST_CASE(Flip, cv::gpu::DeviceInfo, cv::Size, MatType, FlipCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int flip_code;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        flip_code = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Flip, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    cv::gpu::flip(loadMat(src, useRoi), dst, flip_code);
+
+    cv::Mat dst_gold;
+    cv::flip(src, dst_gold, flip_code);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Flip, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1),
+                    MatType(CV_8UC3),
+                    MatType(CV_8UC4),
+                    MatType(CV_16UC1),
+                    MatType(CV_16UC3),
+                    MatType(CV_16UC4),
+                    MatType(CV_32SC1),
+                    MatType(CV_32SC3),
+                    MatType(CV_32SC4),
+                    MatType(CV_32FC1),
+                    MatType(CV_32FC3),
+                    MatType(CV_32FC4)),
+    ALL_FLIP_CODES,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// LUT
+
+PARAM_TEST_CASE(LUT, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(LUT, OneChannel)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat lut = randomMat(cv::Size(256, 1), CV_8UC1);
+
+    cv::Ptr<cv::gpu::LookUpTable> lutAlg = cv::gpu::createLookUpTable(lut);
+
+    cv::gpu::GpuMat dst = createMat(size, CV_MAKE_TYPE(lut.depth(), src.channels()));
+    lutAlg->transform(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::LUT(src, lut, dst_gold);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+GPU_TEST_P(LUT, MultiChannel)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat lut = randomMat(cv::Size(256, 1), CV_MAKE_TYPE(CV_8U, src.channels()));
+
+    cv::Ptr<cv::gpu::LookUpTable> lutAlg = cv::gpu::createLookUpTable(lut);
+
+    cv::gpu::GpuMat dst = createMat(size, CV_MAKE_TYPE(lut.depth(), src.channels()), useRoi);
+    lutAlg->transform(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::LUT(src, lut, dst_gold);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, LUT, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3)),
+    WHOLE_SUBMAT));
+
+//////////////////////////////////////////////////////////////////////////////
+// CopyMakeBorder
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(Border, int)
+}
+
+PARAM_TEST_CASE(CopyMakeBorder, cv::gpu::DeviceInfo, cv::Size, MatType, Border, BorderType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int border;
+    int borderType;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        border = GET_PARAM(3);
+        borderType = GET_PARAM(4);
+        useRoi = GET_PARAM(5);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CopyMakeBorder, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Scalar val = randomScalar(0, 255);
+
+    cv::gpu::GpuMat dst = createMat(cv::Size(size.width + 2 * border, size.height + 2 * border), type, useRoi);
+    cv::gpu::copyMakeBorder(loadMat(src, useRoi), dst, border, border, border, border, borderType, val);
+
+    cv::Mat dst_gold;
+    cv::copyMakeBorder(src, dst_gold, border, border, border, border, borderType, val);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, CopyMakeBorder, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1),
+                    MatType(CV_8UC3),
+                    MatType(CV_8UC4),
+                    MatType(CV_16UC1),
+                    MatType(CV_16UC3),
+                    MatType(CV_16UC4),
+                    MatType(CV_32FC1),
+                    MatType(CV_32FC3),
+                    MatType(CV_32FC4)),
+    testing::Values(Border(1), Border(10), Border(50)),
+    ALL_BORDER_TYPES,
+    WHOLE_SUBMAT));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_core.cpp b/modules/gpuarithm/test/test_element_operations.cpp
similarity index 68%
rename from modules/gpu/test/test_core.cpp
rename to modules/gpuarithm/test/test_element_operations.cpp
index affc30610a..61ea454ead 100644
--- a/modules/gpu/test/test_core.cpp
+++ b/modules/gpuarithm/test/test_element_operations.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,137 +44,7 @@
 
 #ifdef HAVE_CUDA
 
-////////////////////////////////////////////////////////////////////////////////
-// Merge
-
-PARAM_TEST_CASE(Merge, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    int channels;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        channels = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Merge, Accuracy)
-{
-    std::vector<cv::Mat> src;
-    src.reserve(channels);
-    for (int i = 0; i < channels; ++i)
-        src.push_back(cv::Mat(size, depth, cv::Scalar::all(i)));
-
-    std::vector<cv::gpu::GpuMat> d_src;
-    for (int i = 0; i < channels; ++i)
-        d_src.push_back(loadMat(src[i], useRoi));
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::merge(d_src, dst);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat dst;
-        cv::gpu::merge(d_src, dst);
-
-        cv::Mat dst_gold;
-        cv::merge(src, dst_gold);
-
-        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Merge, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    testing::Values(1, 2, 3, 4),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// Split
-
-PARAM_TEST_CASE(Split, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    int channels;
-    bool useRoi;
-
-    int type;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        channels = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        type = CV_MAKE_TYPE(depth, channels);
-    }
-};
-
-GPU_TEST_P(Split, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            std::vector<cv::gpu::GpuMat> dst;
-            cv::gpu::split(loadMat(src), dst);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        std::vector<cv::gpu::GpuMat> dst;
-        cv::gpu::split(loadMat(src, useRoi), dst);
-
-        std::vector<cv::Mat> dst_gold;
-        cv::split(src, dst_gold);
-
-        ASSERT_EQ(dst_gold.size(), dst.size());
-
-        for (size_t i = 0; i < dst_gold.size(); ++i)
-        {
-            EXPECT_MAT_NEAR(dst_gold[i], dst[i], 0.0);
-        }
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Split, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    testing::Values(1, 2, 3, 4),
-    WHOLE_SUBMAT));
+using namespace cvtest;
 
 ////////////////////////////////////////////////////////////////////////////////
 // Add_Array
@@ -218,7 +89,7 @@ GPU_TEST_P(Add_Array, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -234,7 +105,7 @@ GPU_TEST_P(Add_Array, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Add_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Add_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -280,7 +151,7 @@ GPU_TEST_P(Add_Array_Mask, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -296,7 +167,7 @@ GPU_TEST_P(Add_Array_Mask, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Add_Array_Mask, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Add_Array_Mask, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -337,7 +208,7 @@ GPU_TEST_P(Add_Scalar, WithOutMask)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -349,7 +220,7 @@ GPU_TEST_P(Add_Scalar, WithOutMask)
         cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
         cv::add(mat, val, dst_gold, cv::noArray(), depth.second);
 
-        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 1.0);
     }
 }
 
@@ -368,7 +239,7 @@ GPU_TEST_P(Add_Scalar, WithMask)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -380,11 +251,99 @@ GPU_TEST_P(Add_Scalar, WithMask)
         cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
         cv::add(mat, val, dst_gold, mask, depth.second);
 
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 1.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Add_Scalar, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    DEPTH_PAIRS,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Add_Scalar_First
+
+PARAM_TEST_CASE(Add_Scalar_First, cv::gpu::DeviceInfo, cv::Size, std::pair<MatDepth, MatDepth>, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    std::pair<MatDepth, MatDepth> depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Add_Scalar_First, WithOutMask)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::add(val, loadMat(mat), dst, cv::gpu::GpuMat(), depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        dst.setTo(cv::Scalar::all(0));
+        cv::gpu::add(val, loadMat(mat, useRoi), dst, cv::gpu::GpuMat(), depth.second);
+
+        cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
+        cv::add(val, mat, dst_gold, cv::noArray(), depth.second);
+
         EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Add_Scalar, testing::Combine(
+GPU_TEST_P(Add_Scalar_First, WithMask)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::add(val, loadMat(mat), dst, cv::gpu::GpuMat(), depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        dst.setTo(cv::Scalar::all(0));
+        cv::gpu::add(val, loadMat(mat, useRoi), dst, loadMat(mask, useRoi), depth.second);
+
+        cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
+        cv::add(val, mat, dst_gold, mask, depth.second);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Add_Scalar_First, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -433,7 +392,7 @@ GPU_TEST_P(Subtract_Array, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -449,7 +408,7 @@ GPU_TEST_P(Subtract_Array, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Subtract_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Subtract_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -495,7 +454,7 @@ GPU_TEST_P(Subtract_Array_Mask, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -511,7 +470,7 @@ GPU_TEST_P(Subtract_Array_Mask, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Subtract_Array_Mask, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Subtract_Array_Mask, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -552,7 +511,7 @@ GPU_TEST_P(Subtract_Scalar, WithOutMask)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -564,7 +523,7 @@ GPU_TEST_P(Subtract_Scalar, WithOutMask)
         cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
         cv::subtract(mat, val, dst_gold, cv::noArray(), depth.second);
 
-        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 1.0);
     }
 }
 
@@ -583,7 +542,7 @@ GPU_TEST_P(Subtract_Scalar, WithMask)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -595,11 +554,99 @@ GPU_TEST_P(Subtract_Scalar, WithMask)
         cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
         cv::subtract(mat, val, dst_gold, mask, depth.second);
 
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 1.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Subtract_Scalar, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    DEPTH_PAIRS,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Subtract_Scalar_First
+
+PARAM_TEST_CASE(Subtract_Scalar_First, cv::gpu::DeviceInfo, cv::Size, std::pair<MatDepth, MatDepth>, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    std::pair<MatDepth, MatDepth> depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Subtract_Scalar_First, WithOutMask)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::subtract(val, loadMat(mat), dst, cv::gpu::GpuMat(), depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        dst.setTo(cv::Scalar::all(0));
+        cv::gpu::subtract(val, loadMat(mat, useRoi), dst, cv::gpu::GpuMat(), depth.second);
+
+        cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
+        cv::subtract(val, mat, dst_gold, cv::noArray(), depth.second);
+
         EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Subtract_Scalar, testing::Combine(
+GPU_TEST_P(Subtract_Scalar_First, WithMask)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::subtract(val, loadMat(mat), dst, cv::gpu::GpuMat(), depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        dst.setTo(cv::Scalar::all(0));
+        cv::gpu::subtract(val, loadMat(mat, useRoi), dst, loadMat(mask, useRoi), depth.second);
+
+        cv::Mat dst_gold(size, depth.second, cv::Scalar::all(0));
+        cv::subtract(val, mat, dst_gold, mask, depth.second);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, depth.first >= CV_32F || depth.second >= CV_32F ? 1e-4 : 0.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Subtract_Scalar_First, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -648,7 +695,7 @@ GPU_TEST_P(Multiply_Array, WithOutScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -678,7 +725,7 @@ GPU_TEST_P(Multiply_Array, WithScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -693,7 +740,7 @@ GPU_TEST_P(Multiply_Array, WithScale)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Multiply_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Multiply_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -793,7 +840,7 @@ GPU_TEST_P(Multiply_Array_Special, Case_16SC4x_32FC1)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Multiply_Array_Special, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Multiply_Array_Special, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     WHOLE_SUBMAT));
@@ -833,7 +880,7 @@ GPU_TEST_P(Multiply_Scalar, WithOutScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -864,7 +911,7 @@ GPU_TEST_P(Multiply_Scalar, WithScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -879,7 +926,94 @@ GPU_TEST_P(Multiply_Scalar, WithScale)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Multiply_Scalar, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Multiply_Scalar, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    DEPTH_PAIRS,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// Multiply_Scalar_First
+
+PARAM_TEST_CASE(Multiply_Scalar_First, cv::gpu::DeviceInfo, cv::Size, std::pair<MatDepth, MatDepth>, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    std::pair<MatDepth, MatDepth> depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Multiply_Scalar_First, WithOutScale)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::multiply(val, loadMat(mat), dst, 1, depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        cv::gpu::multiply(val, loadMat(mat, useRoi), dst, 1, depth.second);
+
+        cv::Mat dst_gold;
+        cv::multiply(val, mat, dst_gold, 1, depth.second);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 1.0);
+    }
+}
+
+
+GPU_TEST_P(Multiply_Scalar_First, WithScale)
+{
+    cv::Mat mat = randomMat(size, depth.first);
+    cv::Scalar val = randomScalar(0, 255);
+    double scale = randomDouble(0.0, 255.0);
+
+    if ((depth.first == CV_64F || depth.second == CV_64F) && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::multiply(val, loadMat(mat), dst, scale, depth.second);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth.second, useRoi);
+        cv::gpu::multiply(val, loadMat(mat, useRoi), dst, scale, depth.second);
+
+        cv::Mat dst_gold;
+        cv::multiply(val, mat, dst_gold, scale, depth.second);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, 1.0);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Multiply_Scalar_First, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -928,7 +1062,7 @@ GPU_TEST_P(Divide_Array, WithOutScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -958,7 +1092,7 @@ GPU_TEST_P(Divide_Array, WithScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -973,7 +1107,7 @@ GPU_TEST_P(Divide_Array, WithScale)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Divide_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Divide_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -1073,7 +1207,7 @@ GPU_TEST_P(Divide_Array_Special, Case_16SC4x_32FC1)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Divide_Array_Special, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Divide_Array_Special, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     WHOLE_SUBMAT));
@@ -1113,7 +1247,7 @@ GPU_TEST_P(Divide_Scalar, WithOutScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1143,7 +1277,7 @@ GPU_TEST_P(Divide_Scalar, WithScale)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1158,16 +1292,16 @@ GPU_TEST_P(Divide_Scalar, WithScale)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Divide_Scalar, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Divide_Scalar, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
     WHOLE_SUBMAT));
 
 ////////////////////////////////////////////////////////////////////////////////
-// Divide_Scalar_Inv
+// Divide_Scalar_First
 
-PARAM_TEST_CASE(Divide_Scalar_Inv, cv::gpu::DeviceInfo, cv::Size, std::pair<MatDepth, MatDepth>, UseRoi)
+PARAM_TEST_CASE(Divide_Scalar_First, cv::gpu::DeviceInfo, cv::Size, std::pair<MatDepth, MatDepth>, UseRoi)
 {
     cv::gpu::DeviceInfo devInfo;
     cv::Size size;
@@ -1185,7 +1319,7 @@ PARAM_TEST_CASE(Divide_Scalar_Inv, cv::gpu::DeviceInfo, cv::Size, std::pair<MatD
     }
 };
 
-GPU_TEST_P(Divide_Scalar_Inv, Accuracy)
+GPU_TEST_P(Divide_Scalar_First, Accuracy)
 {
     double scale = randomDouble(0.0, 255.0);
     cv::Mat mat = randomMat(size, depth.first, 1.0, 255.0);
@@ -1199,7 +1333,7 @@ GPU_TEST_P(Divide_Scalar_Inv, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1214,7 +1348,7 @@ GPU_TEST_P(Divide_Scalar_Inv, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Divide_Scalar_Inv, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Divide_Scalar_First, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DEPTH_PAIRS,
@@ -1255,7 +1389,7 @@ GPU_TEST_P(AbsDiff, Array)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1284,7 +1418,7 @@ GPU_TEST_P(AbsDiff, Scalar)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1299,7 +1433,36 @@ GPU_TEST_P(AbsDiff, Scalar)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, AbsDiff, testing::Combine(
+GPU_TEST_P(AbsDiff, Scalar_First)
+{
+    cv::Mat src = randomMat(size, depth);
+    cv::Scalar val = randomScalar(0.0, 255.0);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::absdiff(val, loadMat(src), dst);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth, useRoi);
+        cv::gpu::absdiff(val, loadMat(src, useRoi), dst);
+
+        cv::Mat dst_gold;
+        cv::absdiff(val, src, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, depth <= CV_32F ? 1.0 : 1e-5);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, AbsDiff, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     ALL_DEPTH,
@@ -1338,7 +1501,7 @@ GPU_TEST_P(Abs, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Abs, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Abs, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_16S), MatDepth(CV_32F)),
@@ -1378,7 +1541,7 @@ GPU_TEST_P(Sqr, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Sqr, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Sqr, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U),
@@ -1448,7 +1611,7 @@ GPU_TEST_P(Sqrt, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 1.0 : 1e-5);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Sqrt, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Sqrt, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U),
@@ -1518,7 +1681,7 @@ GPU_TEST_P(Log, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 1.0 : 1e-6);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Log, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Log, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U),
@@ -1598,7 +1761,7 @@ GPU_TEST_P(Exp, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 1.0 : 1e-2);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Exp, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Exp, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U),
@@ -1607,6 +1770,65 @@ INSTANTIATE_TEST_CASE_P(GPU_Core, Exp, testing::Combine(
                     MatDepth(CV_32F)),
     WHOLE_SUBMAT));
 
+////////////////////////////////////////////////////////////////////////////////
+// Pow
+
+PARAM_TEST_CASE(Pow, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Pow, Accuracy)
+{
+    cv::Mat src = randomMat(size, depth, 0.0, 10.0);
+    double power = randomDouble(2.0, 4.0);
+
+    if (src.depth() < CV_32F)
+        power = static_cast<int>(power);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::GpuMat dst;
+            cv::gpu::pow(loadMat(src), power, dst);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        cv::gpu::GpuMat dst = createMat(size, depth, useRoi);
+        cv::gpu::pow(loadMat(src, useRoi), power, dst);
+
+        cv::Mat dst_gold;
+        cv::pow(src, power, dst_gold);
+
+        EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 0.0 : 1e-1);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Pow, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    WHOLE_SUBMAT));
+
 ////////////////////////////////////////////////////////////////////////////////
 // Compare_Array
 
@@ -1647,7 +1869,7 @@ GPU_TEST_P(Compare_Array, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1662,7 +1884,7 @@ GPU_TEST_P(Compare_Array, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Compare_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Compare_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     ALL_DEPTH,
@@ -1757,7 +1979,7 @@ GPU_TEST_P(Compare_Scalar, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -1773,7 +1995,7 @@ GPU_TEST_P(Compare_Scalar, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Compare_Scalar, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Compare_Scalar, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     TYPES(CV_8U, CV_64F, 1, 4),
@@ -1845,7 +2067,7 @@ GPU_TEST_P(Bitwise_Array, Xor)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Bitwise_Array, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Bitwise_Array, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     TYPES(CV_8U, CV_32S, 1, 4)));
@@ -1911,7 +2133,7 @@ GPU_TEST_P(Bitwise_Scalar, Xor)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Bitwise_Scalar, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Bitwise_Scalar, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_32S)),
@@ -1986,7 +2208,7 @@ GPU_TEST_P(RShift, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, RShift, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, RShift, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U),
@@ -2066,7 +2288,7 @@ GPU_TEST_P(LShift, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, LShift, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, LShift, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_32S)),
@@ -2108,7 +2330,7 @@ GPU_TEST_P(Min, Array)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -2136,7 +2358,7 @@ GPU_TEST_P(Min, Scalar)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -2146,11 +2368,11 @@ GPU_TEST_P(Min, Scalar)
 
         cv::Mat dst_gold = cv::min(src, val);
 
-        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+        EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 1.0 : 1e-5);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Min, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Min, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     ALL_DEPTH,
@@ -2191,7 +2413,7 @@ GPU_TEST_P(Max, Array)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -2219,7 +2441,7 @@ GPU_TEST_P(Max, Scalar)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -2229,70 +2451,11 @@ GPU_TEST_P(Max, Scalar)
 
         cv::Mat dst_gold = cv::max(src, val);
 
-        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+        EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 1.0 : 1e-5);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Max, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// Pow
-
-PARAM_TEST_CASE(Pow, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Pow, Accuracy)
-{
-    cv::Mat src = randomMat(size, depth, 0.0, 10.0);
-    double power = randomDouble(2.0, 4.0);
-
-    if (src.depth() < CV_32F)
-        power = static_cast<int>(power);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::pow(loadMat(src), power, dst);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat dst = createMat(size, depth, useRoi);
-        cv::gpu::pow(loadMat(src, useRoi), power, dst);
-
-        cv::Mat dst_gold;
-        cv::pow(src, power, dst_gold);
-
-        EXPECT_MAT_NEAR(dst_gold, dst, depth < CV_32F ? 0.0 : 1e-1);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Pow, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Max, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     ALL_DEPTH,
@@ -2340,7 +2503,7 @@ GPU_TEST_P(AddWeighted, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
         }
     }
     else
@@ -2355,7 +2518,7 @@ GPU_TEST_P(AddWeighted, Accuracy)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, AddWeighted, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, AddWeighted, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     ALL_DEPTH,
@@ -2363,20 +2526,18 @@ INSTANTIATE_TEST_CASE_P(GPU_Core, AddWeighted, testing::Combine(
     ALL_DEPTH,
     WHOLE_SUBMAT));
 
-//////////////////////////////////////////////////////////////////////////////
-// GEMM
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// Threshold
 
-#ifdef HAVE_CUBLAS
+CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
+#define ALL_THRESH_OPS testing::Values(ThreshOp(cv::THRESH_BINARY), ThreshOp(cv::THRESH_BINARY_INV), ThreshOp(cv::THRESH_TRUNC), ThreshOp(cv::THRESH_TOZERO), ThreshOp(cv::THRESH_TOZERO_INV))
 
-CV_FLAGS(GemmFlags, 0, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T);
-#define ALL_GEMM_FLAGS testing::Values(GemmFlags(0), GemmFlags(cv::GEMM_1_T), GemmFlags(cv::GEMM_2_T), GemmFlags(cv::GEMM_3_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_3_T), GemmFlags(cv::GEMM_1_T | cv::GEMM_2_T | cv::GEMM_3_T))
-
-PARAM_TEST_CASE(GEMM, cv::gpu::DeviceInfo, cv::Size, MatType, GemmFlags, UseRoi)
+PARAM_TEST_CASE(Threshold, cv::gpu::DeviceInfo, cv::Size, MatType, ThreshOp, UseRoi)
 {
     cv::gpu::DeviceInfo devInfo;
     cv::Size size;
     int type;
-    int flags;
+    int threshOp;
     bool useRoi;
 
     virtual void SetUp()
@@ -2384,238 +2545,33 @@ PARAM_TEST_CASE(GEMM, cv::gpu::DeviceInfo, cv::Size, MatType, GemmFlags, UseRoi)
         devInfo = GET_PARAM(0);
         size = GET_PARAM(1);
         type = GET_PARAM(2);
-        flags = GET_PARAM(3);
+        threshOp = GET_PARAM(3);
         useRoi = GET_PARAM(4);
 
         cv::gpu::setDevice(devInfo.deviceID());
     }
 };
 
-GPU_TEST_P(GEMM, Accuracy)
-{
-    cv::Mat src1 = randomMat(size, type, -10.0, 10.0);
-    cv::Mat src2 = randomMat(size, type, -10.0, 10.0);
-    cv::Mat src3 = randomMat(size, type, -10.0, 10.0);
-    double alpha = randomDouble(-10.0, 10.0);
-    double beta = randomDouble(-10.0, 10.0);
-
-    if (CV_MAT_DEPTH(type) == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::gemm(loadMat(src1), loadMat(src2), alpha, loadMat(src3), beta, dst, flags);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else if (type == CV_64FC2 && flags != 0)
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::gemm(loadMat(src1), loadMat(src2), alpha, loadMat(src3), beta, dst, flags);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-        cv::gpu::gemm(loadMat(src1, useRoi), loadMat(src2, useRoi), alpha, loadMat(src3, useRoi), beta, dst, flags);
-
-        cv::Mat dst_gold;
-        cv::gemm(src1, src2, alpha, src3, beta, dst_gold, flags);
-
-        EXPECT_MAT_NEAR(dst_gold, dst, CV_MAT_DEPTH(type) == CV_32F ? 1e-1 : 1e-10);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, GEMM, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_32FC1), MatType(CV_32FC2), MatType(CV_64FC1), MatType(CV_64FC2)),
-    ALL_GEMM_FLAGS,
-    WHOLE_SUBMAT));
-
-#endif // HAVE_CUBLAS
-
-////////////////////////////////////////////////////////////////////////////////
-// Transpose
-
-PARAM_TEST_CASE(Transpose, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Transpose, Accuracy)
+GPU_TEST_P(Threshold, Accuracy)
 {
     cv::Mat src = randomMat(size, type);
+    double maxVal = randomDouble(20.0, 127.0);
+    double thresh = randomDouble(0.0, maxVal);
 
-    if (CV_MAT_DEPTH(type) == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::transpose(loadMat(src), dst);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat dst = createMat(cv::Size(size.height, size.width), type, useRoi);
-        cv::gpu::transpose(loadMat(src, useRoi), dst);
-
-        cv::Mat dst_gold;
-        cv::transpose(src, dst_gold);
-
-        EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Transpose, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1),
-                    MatType(CV_8UC4),
-                    MatType(CV_16UC2),
-                    MatType(CV_16SC2),
-                    MatType(CV_32SC1),
-                    MatType(CV_32SC2),
-                    MatType(CV_64FC1)),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// Flip
-
-enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
-CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
-#define ALL_FLIP_CODES testing::Values(FlipCode(FLIP_BOTH), FlipCode(FLIP_X), FlipCode(FLIP_Y))
-
-PARAM_TEST_CASE(Flip, cv::gpu::DeviceInfo, cv::Size, MatType, FlipCode, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    int flip_code;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        flip_code = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Flip, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-
-    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::flip(loadMat(src, useRoi), dst, flip_code);
+    cv::gpu::GpuMat dst = createMat(src.size(), src.type(), useRoi);
+    cv::gpu::threshold(loadMat(src, useRoi), dst, thresh, maxVal, threshOp);
 
     cv::Mat dst_gold;
-    cv::flip(src, dst_gold, flip_code);
+    cv::threshold(src, dst_gold, thresh, maxVal, threshOp);
 
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Flip, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Threshold, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1),
-                    MatType(CV_8UC3),
-                    MatType(CV_8UC4),
-                    MatType(CV_16UC1),
-                    MatType(CV_16UC3),
-                    MatType(CV_16UC4),
-                    MatType(CV_32SC1),
-                    MatType(CV_32SC3),
-                    MatType(CV_32SC4),
-                    MatType(CV_32FC1),
-                    MatType(CV_32FC3),
-                    MatType(CV_32FC4)),
-    ALL_FLIP_CODES,
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// LUT
-
-PARAM_TEST_CASE(LUT, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(LUT, OneChannel)
-{
-    cv::Mat src = randomMat(size, type);
-    cv::Mat lut = randomMat(cv::Size(256, 1), CV_8UC1);
-
-    cv::gpu::GpuMat dst = createMat(size, CV_MAKE_TYPE(lut.depth(), src.channels()));
-    cv::gpu::LUT(loadMat(src, useRoi), lut, dst);
-
-    cv::Mat dst_gold;
-    cv::LUT(src, lut, dst_gold);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-}
-
-GPU_TEST_P(LUT, MultiChannel)
-{
-    cv::Mat src = randomMat(size, type);
-    cv::Mat lut = randomMat(cv::Size(256, 1), CV_MAKE_TYPE(CV_8U, src.channels()));
-
-    cv::gpu::GpuMat dst = createMat(size, CV_MAKE_TYPE(lut.depth(), src.channels()), useRoi);
-    cv::gpu::LUT(loadMat(src, useRoi), lut, dst);
-
-    cv::Mat dst_gold;
-    cv::LUT(src, lut, dst_gold);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, LUT, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3)),
+    testing::Values(MatType(CV_8UC1), MatType(CV_16SC1), MatType(CV_32FC1)),
+    ALL_THRESH_OPS,
     WHOLE_SUBMAT));
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -2697,7 +2653,7 @@ GPU_TEST_P(Magnitude, Sqr_Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 1e-1);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Magnitude, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Magnitude, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     WHOLE_SUBMAT));
@@ -2742,7 +2698,7 @@ GPU_TEST_P(Phase, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, angleInDegrees ? 1e-2 : 1e-3);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, Phase, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Phase, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(AngleInDegrees(false), AngleInDegrees(true)),
@@ -2786,7 +2742,7 @@ GPU_TEST_P(CartToPolar, Accuracy)
     EXPECT_MAT_NEAR(angle_gold, angle, angleInDegrees ? 1e-2 : 1e-3);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, CartToPolar, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, CartToPolar, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(AngleInDegrees(false), AngleInDegrees(true)),
@@ -2830,780 +2786,10 @@ GPU_TEST_P(PolarToCart, Accuracy)
     EXPECT_MAT_NEAR(y_gold, y, 1e-4);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Core, PolarToCart, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, PolarToCart, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(AngleInDegrees(false), AngleInDegrees(true)),
     WHOLE_SUBMAT));
 
-////////////////////////////////////////////////////////////////////////////////
-// MeanStdDev
-
-PARAM_TEST_CASE(MeanStdDev, cv::gpu::DeviceInfo, cv::Size, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        useRoi = GET_PARAM(2);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MeanStdDev, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_8UC1);
-
-    if (!supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_13))
-    {
-        try
-        {
-            cv::Scalar mean;
-            cv::Scalar stddev;
-            cv::gpu::meanStdDev(loadMat(src, useRoi), mean, stddev);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        cv::Scalar mean;
-        cv::Scalar stddev;
-        cv::gpu::meanStdDev(loadMat(src, useRoi), mean, stddev);
-
-        cv::Scalar mean_gold;
-        cv::Scalar stddev_gold;
-        cv::meanStdDev(src, mean_gold, stddev_gold);
-
-        EXPECT_SCALAR_NEAR(mean_gold, mean, 1e-5);
-        EXPECT_SCALAR_NEAR(stddev_gold, stddev, 1e-5);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, MeanStdDev, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// Norm
-
-PARAM_TEST_CASE(Norm, cv::gpu::DeviceInfo, cv::Size, MatDepth, NormCode, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    int normCode;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        normCode = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Norm, Accuracy)
-{
-    cv::Mat src = randomMat(size, depth);
-    cv::Mat mask = randomMat(size, CV_8UC1, 0, 2);
-
-    cv::gpu::GpuMat d_buf;
-    double val = cv::gpu::norm(loadMat(src, useRoi), normCode, loadMat(mask, useRoi), d_buf);
-
-    double val_gold = cv::norm(src, normCode, mask);
-
-    EXPECT_NEAR(val_gold, val, depth < CV_32F ? 0.0 : 1.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Norm, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatDepth(CV_8U),
-                    MatDepth(CV_8S),
-                    MatDepth(CV_16U),
-                    MatDepth(CV_16S),
-                    MatDepth(CV_32S),
-                    MatDepth(CV_32F)),
-    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF)),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// normDiff
-
-PARAM_TEST_CASE(NormDiff, cv::gpu::DeviceInfo, cv::Size, NormCode, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int normCode;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        normCode = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(NormDiff, Accuracy)
-{
-    cv::Mat src1 = randomMat(size, CV_8UC1);
-    cv::Mat src2 = randomMat(size, CV_8UC1);
-
-    double val = cv::gpu::norm(loadMat(src1, useRoi), loadMat(src2, useRoi), normCode);
-
-    double val_gold = cv::norm(src1, src2, normCode);
-
-    EXPECT_NEAR(val_gold, val, 0.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, NormDiff, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF)),
-    WHOLE_SUBMAT));
-
-//////////////////////////////////////////////////////////////////////////////
-// Sum
-
-namespace
-{
-    template <typename T>
-    cv::Scalar absSumImpl(const cv::Mat& src)
-    {
-        const int cn = src.channels();
-
-        cv::Scalar sum = cv::Scalar::all(0);
-
-        for (int y = 0; y < src.rows; ++y)
-        {
-            for (int x = 0; x < src.cols; ++x)
-            {
-                for (int c = 0; c < cn; ++c)
-                    sum[c] += std::abs(src.at<T>(y, x * cn + c));
-            }
-        }
-
-        return sum;
-    }
-
-    cv::Scalar absSumGold(const cv::Mat& src)
-    {
-        typedef cv::Scalar (*func_t)(const cv::Mat& src);
-
-        static const func_t funcs[] =
-        {
-            absSumImpl<uchar>,
-            absSumImpl<schar>,
-            absSumImpl<ushort>,
-            absSumImpl<short>,
-            absSumImpl<int>,
-            absSumImpl<float>,
-            absSumImpl<double>
-        };
-
-        return funcs[src.depth()](src);
-    }
-
-    template <typename T>
-    cv::Scalar sqrSumImpl(const cv::Mat& src)
-    {
-        const int cn = src.channels();
-
-        cv::Scalar sum = cv::Scalar::all(0);
-
-        for (int y = 0; y < src.rows; ++y)
-        {
-            for (int x = 0; x < src.cols; ++x)
-            {
-                for (int c = 0; c < cn; ++c)
-                {
-                    const T val = src.at<T>(y, x * cn + c);
-                    sum[c] += val * val;
-                }
-            }
-        }
-
-        return sum;
-    }
-
-    cv::Scalar sqrSumGold(const cv::Mat& src)
-    {
-        typedef cv::Scalar (*func_t)(const cv::Mat& src);
-
-        static const func_t funcs[] =
-        {
-            sqrSumImpl<uchar>,
-            sqrSumImpl<schar>,
-            sqrSumImpl<ushort>,
-            sqrSumImpl<short>,
-            sqrSumImpl<int>,
-            sqrSumImpl<float>,
-            sqrSumImpl<double>
-        };
-
-        return funcs[src.depth()](src);
-    }
-}
-
-PARAM_TEST_CASE(Sum, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    bool useRoi;
-
-    cv::Mat src;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        src = randomMat(size, type, -128.0, 128.0);
-    }
-};
-
-GPU_TEST_P(Sum, Simple)
-{
-    cv::Scalar val = cv::gpu::sum(loadMat(src, useRoi));
-
-    cv::Scalar val_gold = cv::sum(src);
-
-    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
-}
-
-GPU_TEST_P(Sum, Abs)
-{
-    cv::Scalar val = cv::gpu::absSum(loadMat(src, useRoi));
-
-    cv::Scalar val_gold = absSumGold(src);
-
-    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
-}
-
-GPU_TEST_P(Sum, Sqr)
-{
-    cv::Scalar val = cv::gpu::sqrSum(loadMat(src, useRoi));
-
-    cv::Scalar val_gold = sqrSumGold(src);
-
-    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Sum, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    TYPES(CV_8U, CV_64F, 1, 4),
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// MinMax
-
-PARAM_TEST_CASE(MinMax, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MinMax, WithoutMask)
-{
-    cv::Mat src = randomMat(size, depth);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::gpu::minMax(loadMat(src), &minVal, &maxVal);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::gpu::minMax(loadMat(src, useRoi), &minVal, &maxVal);
-
-        double minVal_gold, maxVal_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-    }
-}
-
-GPU_TEST_P(MinMax, WithMask)
-{
-    cv::Mat src = randomMat(size, depth);
-    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::gpu::minMax(loadMat(src), &minVal, &maxVal, loadMat(mask));
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::gpu::minMax(loadMat(src, useRoi), &minVal, &maxVal, loadMat(mask, useRoi));
-
-        double minVal_gold, maxVal_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold, 0, 0, mask);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-    }
-}
-
-GPU_TEST_P(MinMax, NullPtr)
-{
-    cv::Mat src = randomMat(size, depth);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::gpu::minMax(loadMat(src), &minVal, 0);
-            cv::gpu::minMax(loadMat(src), 0, &maxVal);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::gpu::minMax(loadMat(src, useRoi), &minVal, 0);
-        cv::gpu::minMax(loadMat(src, useRoi), 0, &maxVal);
-
-        double minVal_gold, maxVal_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold, 0, 0);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, MinMax, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////////
-// MinMaxLoc
-
-namespace
-{
-    template <typename T>
-    void expectEqualImpl(const cv::Mat& src, cv::Point loc_gold, cv::Point loc)
-    {
-        EXPECT_EQ(src.at<T>(loc_gold.y, loc_gold.x), src.at<T>(loc.y, loc.x));
-    }
-
-    void expectEqual(const cv::Mat& src, cv::Point loc_gold, cv::Point loc)
-    {
-        typedef void (*func_t)(const cv::Mat& src, cv::Point loc_gold, cv::Point loc);
-
-        static const func_t funcs[] =
-        {
-            expectEqualImpl<uchar>,
-            expectEqualImpl<schar>,
-            expectEqualImpl<ushort>,
-            expectEqualImpl<short>,
-            expectEqualImpl<int>,
-            expectEqualImpl<float>,
-            expectEqualImpl<double>
-        };
-
-        funcs[src.depth()](src, loc_gold, loc);
-    }
-}
-
-PARAM_TEST_CASE(MinMaxLoc, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(MinMaxLoc, WithoutMask)
-{
-    cv::Mat src = randomMat(size, depth);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            cv::gpu::minMaxLoc(loadMat(src), &minVal, &maxVal, &minLoc, &maxLoc);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::Point minLoc, maxLoc;
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, &maxVal, &minLoc, &maxLoc);
-
-        double minVal_gold, maxVal_gold;
-        cv::Point minLoc_gold, maxLoc_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-
-        expectEqual(src, minLoc_gold, minLoc);
-        expectEqual(src, maxLoc_gold, maxLoc);
-    }
-}
-
-GPU_TEST_P(MinMaxLoc, WithMask)
-{
-    cv::Mat src = randomMat(size, depth);
-    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            cv::gpu::minMaxLoc(loadMat(src), &minVal, &maxVal, &minLoc, &maxLoc, loadMat(mask));
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::Point minLoc, maxLoc;
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, &maxVal, &minLoc, &maxLoc, loadMat(mask, useRoi));
-
-        double minVal_gold, maxVal_gold;
-        cv::Point minLoc_gold, maxLoc_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold, mask);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-
-        expectEqual(src, minLoc_gold, minLoc);
-        expectEqual(src, maxLoc_gold, maxLoc);
-    }
-}
-
-GPU_TEST_P(MinMaxLoc, NullPtr)
-{
-    cv::Mat src = randomMat(size, depth);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, 0, 0, 0);
-            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, &maxVal, 0, 0);
-            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, &minLoc, 0);
-            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, 0, &maxLoc);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        double minVal, maxVal;
-        cv::Point minLoc, maxLoc;
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, 0, 0, 0);
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, &maxVal, 0, 0);
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, &minLoc, 0);
-        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, 0, &maxLoc);
-
-        double minVal_gold, maxVal_gold;
-        cv::Point minLoc_gold, maxLoc_gold;
-        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold);
-
-        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
-        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
-
-        expectEqual(src, minLoc_gold, minLoc);
-        expectEqual(src, maxLoc_gold, maxLoc);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, MinMaxLoc, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    WHOLE_SUBMAT));
-
-////////////////////////////////////////////////////////////////////////////
-// CountNonZero
-
-PARAM_TEST_CASE(CountNonZero, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    bool useRoi;
-
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        useRoi = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(CountNonZero, Accuracy)
-{
-    cv::Mat srcBase = randomMat(size, CV_8U, 0.0, 1.5);
-    cv::Mat src;
-    srcBase.convertTo(src, depth);
-
-    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
-    {
-        try
-        {
-            cv::gpu::countNonZero(loadMat(src));
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsUnsupportedFormat, e.code);
-        }
-    }
-    else
-    {
-        int val = cv::gpu::countNonZero(loadMat(src, useRoi));
-
-        int val_gold = cv::countNonZero(src);
-
-        ASSERT_EQ(val_gold, val);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, CountNonZero, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    WHOLE_SUBMAT));
-
-//////////////////////////////////////////////////////////////////////////////
-// Reduce
-
-CV_ENUM(ReduceCode, CV_REDUCE_SUM, CV_REDUCE_AVG, CV_REDUCE_MAX, CV_REDUCE_MIN)
-#define ALL_REDUCE_CODES testing::Values(ReduceCode(CV_REDUCE_SUM), ReduceCode(CV_REDUCE_AVG), ReduceCode(CV_REDUCE_MAX), ReduceCode(CV_REDUCE_MIN))
-
-PARAM_TEST_CASE(Reduce, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, ReduceCode, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int depth;
-    int channels;
-    int reduceOp;
-    bool useRoi;
-
-    int type;
-    int dst_depth;
-    int dst_type;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        depth = GET_PARAM(2);
-        channels = GET_PARAM(3);
-        reduceOp = GET_PARAM(4);
-        useRoi = GET_PARAM(5);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        type = CV_MAKE_TYPE(depth, channels);
-
-        if (reduceOp == CV_REDUCE_MAX || reduceOp == CV_REDUCE_MIN)
-            dst_depth = depth;
-        else if (reduceOp == CV_REDUCE_SUM)
-            dst_depth = depth == CV_8U ? CV_32S : depth < CV_64F ? CV_32F : depth;
-        else
-            dst_depth = depth < CV_32F ? CV_32F : depth;
-
-        dst_type = CV_MAKE_TYPE(dst_depth, channels);
-    }
-
-};
-
-GPU_TEST_P(Reduce, Rows)
-{
-    cv::Mat src = randomMat(size, type);
-
-    cv::gpu::GpuMat dst = createMat(cv::Size(src.cols, 1), dst_type, useRoi);
-    cv::gpu::reduce(loadMat(src, useRoi), dst, 0, reduceOp, dst_depth);
-
-    cv::Mat dst_gold;
-    cv::reduce(src, dst_gold, 0, reduceOp, dst_depth);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, dst_depth < CV_32F ? 0.0 : 0.02);
-}
-
-GPU_TEST_P(Reduce, Cols)
-{
-    cv::Mat src = randomMat(size, type);
-
-    cv::gpu::GpuMat dst = createMat(cv::Size(src.rows, 1), dst_type, useRoi);
-    cv::gpu::reduce(loadMat(src, useRoi), dst, 1, reduceOp, dst_depth);
-
-    cv::Mat dst_gold;
-    cv::reduce(src, dst_gold, 1, reduceOp, dst_depth);
-    dst_gold.cols = dst_gold.rows;
-    dst_gold.rows = 1;
-    dst_gold.step = dst_gold.cols * dst_gold.elemSize();
-
-    EXPECT_MAT_NEAR(dst_gold, dst, dst_depth < CV_32F ? 0.0 : 0.02);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Reduce, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatDepth(CV_8U),
-                    MatDepth(CV_16U),
-                    MatDepth(CV_16S),
-                    MatDepth(CV_32F),
-                    MatDepth(CV_64F)),
-    ALL_CHANNELS,
-    ALL_REDUCE_CODES,
-    WHOLE_SUBMAT));
-
-//////////////////////////////////////////////////////////////////////////////
-// Normalize
-
-PARAM_TEST_CASE(Normalize, cv::gpu::DeviceInfo, cv::Size, MatDepth, NormCode, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    int norm_type;
-    bool useRoi;
-
-    double alpha;
-    double beta;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        norm_type = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        alpha = 1;
-        beta = 0;
-    }
-
-};
-
-GPU_TEST_P(Normalize, WithOutMask)
-{
-    cv::Mat src = randomMat(size, type);
-
-    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type);
-
-    cv::Mat dst_gold;
-    cv::normalize(src, dst_gold, alpha, beta, norm_type, type);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
-}
-
-GPU_TEST_P(Normalize, WithMask)
-{
-    cv::Mat src = randomMat(size, type);
-    cv::Mat mask = randomMat(size, CV_8UC1, 0, 2);
-
-    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    dst.setTo(cv::Scalar::all(0));
-    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type, loadMat(mask, useRoi));
-
-    cv::Mat dst_gold(size, type);
-    dst_gold.setTo(cv::Scalar::all(0));
-    cv::normalize(src, dst_gold, alpha, beta, norm_type, type, mask);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Core, Normalize, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    ALL_DEPTH,
-    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF), NormCode(cv::NORM_MINMAX)),
-    WHOLE_SUBMAT));
-
 #endif // HAVE_CUDA
diff --git a/modules/ml/src/precomp.cpp b/modules/gpuarithm/test/test_main.cpp
similarity index 97%
rename from modules/ml/src/precomp.cpp
rename to modules/gpuarithm/test/test_main.cpp
index e540cc5e8a..eea3d7c008 100644
--- a/modules/ml/src/precomp.cpp
+++ b/modules/gpuarithm/test/test_main.cpp
@@ -40,6 +40,6 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
-/* End of file. */
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpuarithm/test/test_precomp.hpp b/modules/gpuarithm/test/test_precomp.hpp
new file mode 100644
index 0000000000..faa0f5adbd
--- /dev/null
+++ b/modules/gpuarithm/test/test_precomp.hpp
@@ -0,0 +1,65 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include <functional>
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpuarithm/test/test_reductions.cpp b/modules/gpuarithm/test/test_reductions.cpp
new file mode 100644
index 0000000000..cd25ba72f2
--- /dev/null
+++ b/modules/gpuarithm/test/test_reductions.cpp
@@ -0,0 +1,819 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// Norm
+
+PARAM_TEST_CASE(Norm, cv::gpu::DeviceInfo, cv::Size, MatDepth, NormCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    int normCode;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        normCode = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Norm, Accuracy)
+{
+    cv::Mat src = randomMat(size, depth);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0, 2);
+
+    cv::gpu::GpuMat d_buf;
+    double val = cv::gpu::norm(loadMat(src, useRoi), normCode, loadMat(mask, useRoi), d_buf);
+
+    double val_gold = cv::norm(src, normCode, mask);
+
+    EXPECT_NEAR(val_gold, val, depth < CV_32F ? 0.0 : 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Norm, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatDepth(CV_8U),
+                    MatDepth(CV_8S),
+                    MatDepth(CV_16U),
+                    MatDepth(CV_16S),
+                    MatDepth(CV_32S),
+                    MatDepth(CV_32F)),
+    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF)),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// normDiff
+
+PARAM_TEST_CASE(NormDiff, cv::gpu::DeviceInfo, cv::Size, NormCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int normCode;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        normCode = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(NormDiff, Accuracy)
+{
+    cv::Mat src1 = randomMat(size, CV_8UC1);
+    cv::Mat src2 = randomMat(size, CV_8UC1);
+
+    double val = cv::gpu::norm(loadMat(src1, useRoi), loadMat(src2, useRoi), normCode);
+
+    double val_gold = cv::norm(src1, src2, normCode);
+
+    EXPECT_NEAR(val_gold, val, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, NormDiff, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF)),
+    WHOLE_SUBMAT));
+
+//////////////////////////////////////////////////////////////////////////////
+// Sum
+
+namespace
+{
+    template <typename T>
+    cv::Scalar absSumImpl(const cv::Mat& src)
+    {
+        const int cn = src.channels();
+
+        cv::Scalar sum = cv::Scalar::all(0);
+
+        for (int y = 0; y < src.rows; ++y)
+        {
+            for (int x = 0; x < src.cols; ++x)
+            {
+                for (int c = 0; c < cn; ++c)
+                    sum[c] += std::abs(src.at<T>(y, x * cn + c));
+            }
+        }
+
+        return sum;
+    }
+
+    cv::Scalar absSumGold(const cv::Mat& src)
+    {
+        typedef cv::Scalar (*func_t)(const cv::Mat& src);
+
+        static const func_t funcs[] =
+        {
+            absSumImpl<uchar>,
+            absSumImpl<schar>,
+            absSumImpl<ushort>,
+            absSumImpl<short>,
+            absSumImpl<int>,
+            absSumImpl<float>,
+            absSumImpl<double>
+        };
+
+        return funcs[src.depth()](src);
+    }
+
+    template <typename T>
+    cv::Scalar sqrSumImpl(const cv::Mat& src)
+    {
+        const int cn = src.channels();
+
+        cv::Scalar sum = cv::Scalar::all(0);
+
+        for (int y = 0; y < src.rows; ++y)
+        {
+            for (int x = 0; x < src.cols; ++x)
+            {
+                for (int c = 0; c < cn; ++c)
+                {
+                    const T val = src.at<T>(y, x * cn + c);
+                    sum[c] += val * val;
+                }
+            }
+        }
+
+        return sum;
+    }
+
+    cv::Scalar sqrSumGold(const cv::Mat& src)
+    {
+        typedef cv::Scalar (*func_t)(const cv::Mat& src);
+
+        static const func_t funcs[] =
+        {
+            sqrSumImpl<uchar>,
+            sqrSumImpl<schar>,
+            sqrSumImpl<ushort>,
+            sqrSumImpl<short>,
+            sqrSumImpl<int>,
+            sqrSumImpl<float>,
+            sqrSumImpl<double>
+        };
+
+        return funcs[src.depth()](src);
+    }
+}
+
+PARAM_TEST_CASE(Sum, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    bool useRoi;
+
+    cv::Mat src;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        src = randomMat(size, type, -128.0, 128.0);
+    }
+};
+
+GPU_TEST_P(Sum, Simple)
+{
+    cv::Scalar val = cv::gpu::sum(loadMat(src, useRoi));
+
+    cv::Scalar val_gold = cv::sum(src);
+
+    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
+}
+
+GPU_TEST_P(Sum, Abs)
+{
+    cv::Scalar val = cv::gpu::absSum(loadMat(src, useRoi));
+
+    cv::Scalar val_gold = absSumGold(src);
+
+    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
+}
+
+GPU_TEST_P(Sum, Sqr)
+{
+    cv::Scalar val = cv::gpu::sqrSum(loadMat(src, useRoi));
+
+    cv::Scalar val_gold = sqrSumGold(src);
+
+    EXPECT_SCALAR_NEAR(val_gold, val, CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.5);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Sum, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    TYPES(CV_8U, CV_64F, 1, 4),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// MinMax
+
+PARAM_TEST_CASE(MinMax, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MinMax, WithoutMask)
+{
+    cv::Mat src = randomMat(size, depth);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::gpu::minMax(loadMat(src), &minVal, &maxVal);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::gpu::minMax(loadMat(src, useRoi), &minVal, &maxVal);
+
+        double minVal_gold, maxVal_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+    }
+}
+
+GPU_TEST_P(MinMax, WithMask)
+{
+    cv::Mat src = randomMat(size, depth);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::gpu::minMax(loadMat(src), &minVal, &maxVal, loadMat(mask));
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::gpu::minMax(loadMat(src, useRoi), &minVal, &maxVal, loadMat(mask, useRoi));
+
+        double minVal_gold, maxVal_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold, 0, 0, mask);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+    }
+}
+
+GPU_TEST_P(MinMax, NullPtr)
+{
+    cv::Mat src = randomMat(size, depth);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::gpu::minMax(loadMat(src), &minVal, 0);
+            cv::gpu::minMax(loadMat(src), 0, &maxVal);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::gpu::minMax(loadMat(src, useRoi), &minVal, 0);
+        cv::gpu::minMax(loadMat(src, useRoi), 0, &maxVal);
+
+        double minVal_gold, maxVal_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold, 0, 0);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, MinMax, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// MinMaxLoc
+
+namespace
+{
+    template <typename T>
+    void expectEqualImpl(const cv::Mat& src, cv::Point loc_gold, cv::Point loc)
+    {
+        EXPECT_EQ(src.at<T>(loc_gold.y, loc_gold.x), src.at<T>(loc.y, loc.x));
+    }
+
+    void expectEqual(const cv::Mat& src, cv::Point loc_gold, cv::Point loc)
+    {
+        typedef void (*func_t)(const cv::Mat& src, cv::Point loc_gold, cv::Point loc);
+
+        static const func_t funcs[] =
+        {
+            expectEqualImpl<uchar>,
+            expectEqualImpl<schar>,
+            expectEqualImpl<ushort>,
+            expectEqualImpl<short>,
+            expectEqualImpl<int>,
+            expectEqualImpl<float>,
+            expectEqualImpl<double>
+        };
+
+        funcs[src.depth()](src, loc_gold, loc);
+    }
+}
+
+PARAM_TEST_CASE(MinMaxLoc, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MinMaxLoc, WithoutMask)
+{
+    cv::Mat src = randomMat(size, depth);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::Point minLoc, maxLoc;
+            cv::gpu::minMaxLoc(loadMat(src), &minVal, &maxVal, &minLoc, &maxLoc);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::Point minLoc, maxLoc;
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, &maxVal, &minLoc, &maxLoc);
+
+        double minVal_gold, maxVal_gold;
+        cv::Point minLoc_gold, maxLoc_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+
+        expectEqual(src, minLoc_gold, minLoc);
+        expectEqual(src, maxLoc_gold, maxLoc);
+    }
+}
+
+GPU_TEST_P(MinMaxLoc, WithMask)
+{
+    cv::Mat src = randomMat(size, depth);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0.0, 2.0);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::Point minLoc, maxLoc;
+            cv::gpu::minMaxLoc(loadMat(src), &minVal, &maxVal, &minLoc, &maxLoc, loadMat(mask));
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::Point minLoc, maxLoc;
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, &maxVal, &minLoc, &maxLoc, loadMat(mask, useRoi));
+
+        double minVal_gold, maxVal_gold;
+        cv::Point minLoc_gold, maxLoc_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold, mask);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+
+        expectEqual(src, minLoc_gold, minLoc);
+        expectEqual(src, maxLoc_gold, maxLoc);
+    }
+}
+
+GPU_TEST_P(MinMaxLoc, NullPtr)
+{
+    cv::Mat src = randomMat(size, depth);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            double minVal, maxVal;
+            cv::Point minLoc, maxLoc;
+            cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, 0, 0, 0);
+            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, &maxVal, 0, 0);
+            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, &minLoc, 0);
+            cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, 0, &maxLoc);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        double minVal, maxVal;
+        cv::Point minLoc, maxLoc;
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), &minVal, 0, 0, 0);
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, &maxVal, 0, 0);
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, &minLoc, 0);
+        cv::gpu::minMaxLoc(loadMat(src, useRoi), 0, 0, 0, &maxLoc);
+
+        double minVal_gold, maxVal_gold;
+        cv::Point minLoc_gold, maxLoc_gold;
+        minMaxLocGold(src, &minVal_gold, &maxVal_gold, &minLoc_gold, &maxLoc_gold);
+
+        EXPECT_DOUBLE_EQ(minVal_gold, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_gold, maxVal);
+
+        expectEqual(src, minLoc_gold, minLoc);
+        expectEqual(src, maxLoc_gold, maxLoc);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, MinMaxLoc, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////
+// CountNonZero
+
+PARAM_TEST_CASE(CountNonZero, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    bool useRoi;
+
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CountNonZero, Accuracy)
+{
+    cv::Mat srcBase = randomMat(size, CV_8U, 0.0, 1.5);
+    cv::Mat src;
+    srcBase.convertTo(src, depth);
+
+    if (depth == CV_64F && !supportFeature(devInfo, cv::gpu::NATIVE_DOUBLE))
+    {
+        try
+        {
+            cv::gpu::countNonZero(loadMat(src));
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsUnsupportedFormat, e.code);
+        }
+    }
+    else
+    {
+        int val = cv::gpu::countNonZero(loadMat(src, useRoi));
+
+        int val_gold = cv::countNonZero(src);
+
+        ASSERT_EQ(val_gold, val);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, CountNonZero, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    WHOLE_SUBMAT));
+
+//////////////////////////////////////////////////////////////////////////////
+// Reduce
+
+CV_ENUM(ReduceCode, cv::REDUCE_SUM, cv::REDUCE_AVG, cv::REDUCE_MAX, cv::REDUCE_MIN)
+#define ALL_REDUCE_CODES testing::Values(ReduceCode(cv::REDUCE_SUM), ReduceCode(cv::REDUCE_AVG), ReduceCode(cv::REDUCE_MAX), ReduceCode(cv::REDUCE_MIN))
+
+PARAM_TEST_CASE(Reduce, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, ReduceCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    int channels;
+    int reduceOp;
+    bool useRoi;
+
+    int type;
+    int dst_depth;
+    int dst_type;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        channels = GET_PARAM(3);
+        reduceOp = GET_PARAM(4);
+        useRoi = GET_PARAM(5);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        type = CV_MAKE_TYPE(depth, channels);
+
+        if (reduceOp == cv::REDUCE_MAX || reduceOp == cv::REDUCE_MIN)
+            dst_depth = depth;
+        else if (reduceOp == cv::REDUCE_SUM)
+            dst_depth = depth == CV_8U ? CV_32S : depth < CV_64F ? CV_32F : depth;
+        else
+            dst_depth = depth < CV_32F ? CV_32F : depth;
+
+        dst_type = CV_MAKE_TYPE(dst_depth, channels);
+    }
+
+};
+
+GPU_TEST_P(Reduce, Rows)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::gpu::GpuMat dst = createMat(cv::Size(src.cols, 1), dst_type, useRoi);
+    cv::gpu::reduce(loadMat(src, useRoi), dst, 0, reduceOp, dst_depth);
+
+    cv::Mat dst_gold;
+    cv::reduce(src, dst_gold, 0, reduceOp, dst_depth);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, dst_depth < CV_32F ? 0.0 : 0.02);
+}
+
+GPU_TEST_P(Reduce, Cols)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::gpu::GpuMat dst = createMat(cv::Size(src.rows, 1), dst_type, useRoi);
+    cv::gpu::reduce(loadMat(src, useRoi), dst, 1, reduceOp, dst_depth);
+
+    cv::Mat dst_gold;
+    cv::reduce(src, dst_gold, 1, reduceOp, dst_depth);
+    dst_gold.cols = dst_gold.rows;
+    dst_gold.rows = 1;
+    dst_gold.step = dst_gold.cols * dst_gold.elemSize();
+
+    EXPECT_MAT_NEAR(dst_gold, dst, dst_depth < CV_32F ? 0.0 : 0.02);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Reduce, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatDepth(CV_8U),
+                    MatDepth(CV_16U),
+                    MatDepth(CV_16S),
+                    MatDepth(CV_32F),
+                    MatDepth(CV_64F)),
+    ALL_CHANNELS,
+    ALL_REDUCE_CODES,
+    WHOLE_SUBMAT));
+
+//////////////////////////////////////////////////////////////////////////////
+// Normalize
+
+PARAM_TEST_CASE(Normalize, cv::gpu::DeviceInfo, cv::Size, MatDepth, NormCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int norm_type;
+    bool useRoi;
+
+    double alpha;
+    double beta;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        norm_type = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        alpha = 1;
+        beta = 0;
+    }
+
+};
+
+GPU_TEST_P(Normalize, WithOutMask)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type);
+
+    cv::Mat dst_gold;
+    cv::normalize(src, dst_gold, alpha, beta, norm_type, type);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
+}
+
+GPU_TEST_P(Normalize, WithMask)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0, 2);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    dst.setTo(cv::Scalar::all(0));
+    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type, loadMat(mask, useRoi));
+
+    cv::Mat dst_gold(size, type);
+    dst_gold.setTo(cv::Scalar::all(0));
+    cv::normalize(src, dst_gold, alpha, beta, norm_type, type, mask);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, Normalize, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF), NormCode(cv::NORM_MINMAX)),
+    WHOLE_SUBMAT));
+
+////////////////////////////////////////////////////////////////////////////////
+// MeanStdDev
+
+PARAM_TEST_CASE(MeanStdDev, cv::gpu::DeviceInfo, cv::Size, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        useRoi = GET_PARAM(2);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MeanStdDev, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    if (!supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_13))
+    {
+        try
+        {
+            cv::Scalar mean;
+            cv::Scalar stddev;
+            cv::gpu::meanStdDev(loadMat(src, useRoi), mean, stddev);
+        }
+        catch (const cv::Exception& e)
+        {
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
+        }
+    }
+    else
+    {
+        cv::Scalar mean;
+        cv::Scalar stddev;
+        cv::gpu::meanStdDev(loadMat(src, useRoi), mean, stddev);
+
+        cv::Scalar mean_gold;
+        cv::Scalar stddev_gold;
+        cv::meanStdDev(src, mean_gold, stddev_gold);
+
+        EXPECT_SCALAR_NEAR(mean_gold, mean, 1e-5);
+        EXPECT_SCALAR_NEAR(stddev_gold, stddev, 1e-5);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Arithm, MeanStdDev, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    WHOLE_SUBMAT));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpubgsegm/CMakeLists.txt b/modules/gpubgsegm/CMakeLists.txt
new file mode 100644
index 0000000000..f87f3cb88c
--- /dev/null
+++ b/modules/gpubgsegm/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpubgsegm)
+endif()
+
+set(the_description "GPU-accelerated Background Segmentation")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_define_module(gpubgsegm opencv_video OPTIONAL opencv_legacy opencv_imgproc opencv_gpuarithm opencv_gpufilters opencv_gpuimgproc)
diff --git a/modules/gpubgsegm/doc/background_segmentation.rst b/modules/gpubgsegm/doc/background_segmentation.rst
new file mode 100644
index 0000000000..53e67ead60
--- /dev/null
+++ b/modules/gpubgsegm/doc/background_segmentation.rst
@@ -0,0 +1,127 @@
+Background Segmentation
+=======================
+
+.. highlight:: cpp
+
+
+
+gpu::BackgroundSubtractorMOG
+----------------------------
+Gaussian Mixture-based Background/Foreground Segmentation Algorithm.
+
+.. ocv:class:: gpu::BackgroundSubtractorMOG : public cv::BackgroundSubtractorMOG
+
+The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [MOG2001]_.
+
+.. seealso:: :ocv:class:`BackgroundSubtractorMOG`
+
+.. note::
+
+   * An example on gaussian mixture based background/foreground segmantation can be found at opencv_source_code/samples/gpu/bgfg_segm.cpp
+
+
+gpu::createBackgroundSubtractorMOG
+----------------------------------
+Creates mixture-of-gaussian background subtractor
+
+.. ocv:function:: Ptr<gpu::BackgroundSubtractorMOG> gpu::createBackgroundSubtractorMOG(int history=200, int nmixtures=5, double backgroundRatio=0.7, double noiseSigma=0)
+
+    :param history: Length of the history.
+
+    :param nmixtures: Number of Gaussian mixtures.
+
+    :param backgroundRatio: Background ratio.
+
+    :param noiseSigma: Noise strength (standard deviation of the brightness or each color channel). 0 means some automatic value.
+
+
+
+gpu::BackgroundSubtractorMOG2
+-----------------------------
+Gaussian Mixture-based Background/Foreground Segmentation Algorithm.
+
+.. ocv:class:: gpu::BackgroundSubtractorMOG2 : public cv::BackgroundSubtractorMOG2
+
+The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [MOG2004]_.
+
+.. seealso:: :ocv:class:`BackgroundSubtractorMOG2`
+
+
+
+gpu::createBackgroundSubtractorMOG2
+-----------------------------------
+Creates MOG2 Background Subtractor
+
+.. ocv:function:: Ptr<gpu::BackgroundSubtractorMOG2> gpu::createBackgroundSubtractorMOG2( int history=500, double varThreshold=16, bool detectShadows=true )
+
+  :param history: Length of the history.
+
+  :param varThreshold: Threshold on the squared Mahalanobis distance between the pixel and the model to decide whether a pixel is well described by the background model. This parameter does not affect the background update.
+
+  :param detectShadows: If true, the algorithm will detect shadows and mark them. It decreases the speed a bit, so if you do not need this feature, set the parameter to false.
+
+
+
+gpu::BackgroundSubtractorGMG
+----------------------------
+Background/Foreground Segmentation Algorithm.
+
+.. ocv:class:: gpu::BackgroundSubtractorGMG : public cv::BackgroundSubtractorGMG
+
+The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [GMG2012]_.
+
+
+
+gpu::createBackgroundSubtractorGMG
+----------------------------------
+Creates GMG Background Subtractor
+
+.. ocv:function:: Ptr<gpu::BackgroundSubtractorGMG> gpu::createBackgroundSubtractorGMG(int initializationFrames = 120, double decisionThreshold = 0.8)
+
+    :param initializationFrames: Number of frames of video to use to initialize histograms.
+
+    :param decisionThreshold: Value above which pixel is determined to be FG.
+
+
+
+gpu::BackgroundSubtractorFGD
+----------------------------
+
+.. ocv:class:: gpu::BackgroundSubtractorFGD : public cv::BackgroundSubtractor
+
+The class discriminates between foreground and background pixels by building and maintaining a model of the background. Any pixel which does not fit this model is then deemed to be foreground. The class implements algorithm described in [FGD2003]_. ::
+
+    class CV_EXPORTS BackgroundSubtractorFGD : public cv::BackgroundSubtractor
+    {
+    public:
+        virtual void getForegroundRegions(OutputArrayOfArrays foreground_regions) = 0;
+    };
+
+.. seealso:: :ocv:class:`BackgroundSubtractor`
+
+
+
+gpu::BackgroundSubtractorFGD::getForegroundRegions
+--------------------------------------------------
+Returns the output foreground regions calculated by :ocv:func:`findContours`.
+
+.. ocv:function:: void gpu::BackgroundSubtractorFGD::getForegroundRegions(OutputArrayOfArrays foreground_regions)
+
+    :params foreground_regions: Output array (CPU memory).
+
+
+
+gpu::createBackgroundSubtractorFGD
+----------------------------------
+Creates FGD Background Subtractor
+
+.. ocv:function:: Ptr<gpu::BackgroundSubtractorGMG> gpu::createBackgroundSubtractorFGD(const FGDParams& params = FGDParams())
+
+    :param params: Algorithm's parameters. See [FGD2003]_ for explanation.
+
+
+
+.. [FGD2003] Liyuan Li, Weimin Huang, Irene Y.H. Gu, and Qi Tian. *Foreground Object Detection from Videos Containing Complex Background*. ACM MM2003 9p, 2003.
+.. [MOG2001] P. KadewTraKuPong and R. Bowden. *An improved adaptive background mixture model for real-time tracking with shadow detection*. Proc. 2nd European Workshop on Advanced Video-Based Surveillance Systems, 2001
+.. [MOG2004] Z. Zivkovic. *Improved adaptive Gausian mixture model for background subtraction*. International Conference Pattern Recognition, UK, August, 2004
+.. [GMG2012] A. Godbehere, A. Matsukawa and K. Goldberg. *Visual Tracking of Human Visitors under Variable-Lighting Conditions for a Responsive Audio Art Installation*. American Control Conference, Montreal, June 2012
diff --git a/modules/gpubgsegm/doc/gpubgsegm.rst b/modules/gpubgsegm/doc/gpubgsegm.rst
new file mode 100644
index 0000000000..ffac8b39ca
--- /dev/null
+++ b/modules/gpubgsegm/doc/gpubgsegm.rst
@@ -0,0 +1,8 @@
+**************************************************
+gpubgsegm. GPU-accelerated Background Segmentation
+**************************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    background_segmentation
diff --git a/modules/gpubgsegm/include/opencv2/gpubgsegm.hpp b/modules/gpubgsegm/include/opencv2/gpubgsegm.hpp
new file mode 100644
index 0000000000..c6d9223ddd
--- /dev/null
+++ b/modules/gpubgsegm/include/opencv2/gpubgsegm.hpp
@@ -0,0 +1,152 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUBGSEGM_HPP__
+#define __OPENCV_GPUBGSEGM_HPP__
+
+#ifndef __cplusplus
+#  error gpubgsegm.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/video/background_segm.hpp"
+
+namespace cv { namespace gpu {
+
+////////////////////////////////////////////////////
+// MOG
+
+class CV_EXPORTS BackgroundSubtractorMOG : public cv::BackgroundSubtractorMOG
+{
+public:
+    using cv::BackgroundSubtractorMOG::apply;
+    using cv::BackgroundSubtractorMOG::getBackgroundImage;
+
+    virtual void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream) = 0;
+
+    virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+};
+
+CV_EXPORTS Ptr<gpu::BackgroundSubtractorMOG>
+    createBackgroundSubtractorMOG(int history = 200, int nmixtures = 5,
+                                  double backgroundRatio = 0.7, double noiseSigma = 0);
+
+////////////////////////////////////////////////////
+// MOG2
+
+class CV_EXPORTS BackgroundSubtractorMOG2 : public cv::BackgroundSubtractorMOG2
+{
+public:
+    using cv::BackgroundSubtractorMOG2::apply;
+    using cv::BackgroundSubtractorMOG2::getBackgroundImage;
+
+    virtual void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream) = 0;
+
+    virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+};
+
+CV_EXPORTS Ptr<gpu::BackgroundSubtractorMOG2>
+    createBackgroundSubtractorMOG2(int history = 500, double varThreshold = 16,
+                                   bool detectShadows = true);
+
+////////////////////////////////////////////////////
+// GMG
+
+class CV_EXPORTS BackgroundSubtractorGMG : public cv::BackgroundSubtractorGMG
+{
+public:
+    using cv::BackgroundSubtractorGMG::apply;
+
+    virtual void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream) = 0;
+};
+
+CV_EXPORTS Ptr<gpu::BackgroundSubtractorGMG>
+    createBackgroundSubtractorGMG(int initializationFrames = 120, double decisionThreshold = 0.8);
+
+////////////////////////////////////////////////////
+// FGD
+
+/**
+ * Foreground Object Detection from Videos Containing Complex Background.
+ * Liyuan Li, Weimin Huang, Irene Y.H. Gu, and Qi Tian.
+ * ACM MM2003 9p
+ */
+class CV_EXPORTS BackgroundSubtractorFGD : public cv::BackgroundSubtractor
+{
+public:
+    virtual void getForegroundRegions(OutputArrayOfArrays foreground_regions) = 0;
+};
+
+struct CV_EXPORTS FGDParams
+{
+    int Lc;  // Quantized levels per 'color' component. Power of two, typically 32, 64 or 128.
+    int N1c; // Number of color vectors used to model normal background color variation at a given pixel.
+    int N2c; // Number of color vectors retained at given pixel.  Must be > N1c, typically ~ 5/3 of N1c.
+    // Used to allow the first N1c vectors to adapt over time to changing background.
+
+    int Lcc;  // Quantized levels per 'color co-occurrence' component.  Power of two, typically 16, 32 or 64.
+    int N1cc; // Number of color co-occurrence vectors used to model normal background color variation at a given pixel.
+    int N2cc; // Number of color co-occurrence vectors retained at given pixel.  Must be > N1cc, typically ~ 5/3 of N1cc.
+    // Used to allow the first N1cc vectors to adapt over time to changing background.
+
+    bool is_obj_without_holes; // If TRUE we ignore holes within foreground blobs. Defaults to TRUE.
+    int perform_morphing;     // Number of erode-dilate-erode foreground-blob cleanup iterations.
+    // These erase one-pixel junk blobs and merge almost-touching blobs. Default value is 1.
+
+    float alpha1; // How quickly we forget old background pixel values seen. Typically set to 0.1.
+    float alpha2; // "Controls speed of feature learning". Depends on T. Typical value circa 0.005.
+    float alpha3; // Alternate to alpha2, used (e.g.) for quicker initial convergence. Typical value 0.1.
+
+    float delta;   // Affects color and color co-occurrence quantization, typically set to 2.
+    float T;       // A percentage value which determines when new features can be recognized as new background. (Typically 0.9).
+    float minArea; // Discard foreground blobs whose bounding box is smaller than this threshold.
+
+    // default Params
+    FGDParams();
+};
+
+CV_EXPORTS Ptr<gpu::BackgroundSubtractorFGD>
+    createBackgroundSubtractorFGD(const FGDParams& params = FGDParams());
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUBGSEGM_HPP__ */
diff --git a/modules/gpubgsegm/perf/perf_bgsegm.cpp b/modules/gpubgsegm/perf/perf_bgsegm.cpp
new file mode 100644
index 0000000000..34b65f6be1
--- /dev/null
+++ b/modules/gpubgsegm/perf/perf_bgsegm.cpp
@@ -0,0 +1,557 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+#ifdef HAVE_OPENCV_LEGACY
+#  include "opencv2/legacy.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUIMGPROC
+#  include "opencv2/gpuimgproc.hpp"
+#endif
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+#if defined(HAVE_XINE)         || \
+    defined(HAVE_GSTREAMER)    || \
+    defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
+    defined(HAVE_AVFOUNDATION) || \
+    defined(HAVE_FFMPEG)       || \
+    defined(WIN32) /* assume that we have ffmpeg */
+
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
+#else
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 0
+#endif
+
+//////////////////////////////////////////////////////
+// FGDStatModel
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+#ifdef HAVE_OPENCV_LEGACY
+
+namespace cv
+{
+    template<> void DefaultDeleter<CvBGStatModel>::operator ()(CvBGStatModel* obj) const
+    {
+        cvReleaseBGStatModel(&obj);
+    }
+}
+
+#endif
+
+DEF_PARAM_TEST_1(Video, string);
+
+PERF_TEST_P(Video, FGDStatModel,
+            Values(string("gpu/video/768x576.avi")))
+{
+    declare.time(60);
+
+    const string inputFile = perf::TestBase::getDataPath(GetParam());
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat d_frame(frame), foreground;
+
+        cv::Ptr<cv::gpu::BackgroundSubtractorFGD> d_fgd = cv::gpu::createBackgroundSubtractorFGD();
+        d_fgd->apply(d_frame, foreground);
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            d_frame.upload(frame);
+
+            startTimer(); next();
+            d_fgd->apply(d_frame, foreground);
+            stopTimer();
+        }
+
+        GPU_SANITY_CHECK(foreground, 1e-2, ERROR_RELATIVE);
+
+#ifdef HAVE_OPENCV_GPUIMGPROC
+        cv::gpu::GpuMat background3, background;
+        d_fgd->getBackgroundImage(background3);
+        cv::gpu::cvtColor(background3, background, cv::COLOR_BGR2BGRA);
+        GPU_SANITY_CHECK(background, 1e-2, ERROR_RELATIVE);
+#endif
+    }
+    else
+    {
+#ifdef HAVE_OPENCV_LEGACY
+        IplImage ipl_frame = frame;
+        cv::Ptr<CvBGStatModel> model(cvCreateFGDStatModel(&ipl_frame));
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            ipl_frame = frame;
+
+            startTimer(); next();
+            cvUpdateBGStatModel(&ipl_frame, model);
+            stopTimer();
+        }
+
+        const cv::Mat background = cv::cvarrToMat(model->background);
+        const cv::Mat foreground = cv::cvarrToMat(model->foreground);
+
+        CPU_SANITY_CHECK(background);
+        CPU_SANITY_CHECK(foreground);
+#else
+        FAIL_NO_CPU();
+#endif
+    }
+}
+
+#endif
+
+//////////////////////////////////////////////////////
+// MOG
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+DEF_PARAM_TEST(Video_Cn_LearningRate, string, MatCn, double);
+
+PERF_TEST_P(Video_Cn_LearningRate, MOG,
+            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+                    GPU_CHANNELS_1_3_4,
+                    Values(0.0, 0.01)))
+{
+    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
+    const int cn = GET_PARAM(1);
+    const float learningRate = static_cast<float>(GET_PARAM(2));
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    if (cn != 3)
+    {
+        cv::Mat temp;
+        if (cn == 1)
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+        else
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+        cv::swap(temp, frame);
+    }
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::BackgroundSubtractor> d_mog = cv::gpu::createBackgroundSubtractorMOG();
+
+        cv::gpu::GpuMat d_frame(frame);
+        cv::gpu::GpuMat foreground;
+
+        d_mog->apply(d_frame, foreground, learningRate);
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            d_frame.upload(frame);
+
+            startTimer(); next();
+            d_mog->apply(d_frame, foreground, learningRate);
+            stopTimer();
+        }
+
+        GPU_SANITY_CHECK(foreground);
+    }
+    else
+    {
+        cv::Ptr<cv::BackgroundSubtractor> mog = cv::createBackgroundSubtractorMOG();
+        cv::Mat foreground;
+
+        mog->apply(frame, foreground, learningRate);
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            startTimer(); next();
+            mog->apply(frame, foreground, learningRate);
+            stopTimer();
+        }
+
+        CPU_SANITY_CHECK(foreground);
+    }
+}
+
+#endif
+
+//////////////////////////////////////////////////////
+// MOG2
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+DEF_PARAM_TEST(Video_Cn, string, int);
+
+PERF_TEST_P(Video_Cn, MOG2,
+            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+                    GPU_CHANNELS_1_3_4))
+{
+    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
+    const int cn = GET_PARAM(1);
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    if (cn != 3)
+    {
+        cv::Mat temp;
+        if (cn == 1)
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+        else
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+        cv::swap(temp, frame);
+    }
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::BackgroundSubtractorMOG2> d_mog2 = cv::gpu::createBackgroundSubtractorMOG2();
+        d_mog2->setDetectShadows(false);
+
+        cv::gpu::GpuMat d_frame(frame);
+        cv::gpu::GpuMat foreground;
+
+        d_mog2->apply(d_frame, foreground);
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            d_frame.upload(frame);
+
+            startTimer(); next();
+            d_mog2->apply(d_frame, foreground);
+            stopTimer();
+        }
+
+        GPU_SANITY_CHECK(foreground);
+    }
+    else
+    {
+        cv::Ptr<cv::BackgroundSubtractorMOG2> mog2 = cv::createBackgroundSubtractorMOG2();
+        mog2->setDetectShadows(false);
+
+        cv::Mat foreground;
+
+        mog2->apply(frame, foreground);
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            startTimer(); next();
+            mog2->apply(frame, foreground);
+            stopTimer();
+        }
+
+        CPU_SANITY_CHECK(foreground);
+    }
+}
+
+#endif
+
+//////////////////////////////////////////////////////
+// MOG2GetBackgroundImage
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+PERF_TEST_P(Video_Cn, MOG2GetBackgroundImage,
+            Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+                    GPU_CHANNELS_1_3_4))
+{
+    const string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
+    const int cn = GET_PARAM(1);
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::BackgroundSubtractor> d_mog2 = cv::gpu::createBackgroundSubtractorMOG2();
+
+        cv::gpu::GpuMat d_frame;
+        cv::gpu::GpuMat d_foreground;
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            d_frame.upload(frame);
+
+            d_mog2->apply(d_frame, d_foreground);
+        }
+
+        cv::gpu::GpuMat background;
+
+        TEST_CYCLE() d_mog2->getBackgroundImage(background);
+
+        GPU_SANITY_CHECK(background, 1);
+    }
+    else
+    {
+        cv::Ptr<cv::BackgroundSubtractor> mog2 = cv::createBackgroundSubtractorMOG2();
+        cv::Mat foreground;
+
+        for (int i = 0; i < 10; ++i)
+        {
+            cap >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            mog2->apply(frame, foreground);
+        }
+
+        cv::Mat background;
+
+        TEST_CYCLE() mog2->getBackgroundImage(background);
+
+        CPU_SANITY_CHECK(background);
+    }
+}
+
+#endif
+
+//////////////////////////////////////////////////////
+// GMG
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+DEF_PARAM_TEST(Video_Cn_MaxFeatures, string, MatCn, int);
+
+PERF_TEST_P(Video_Cn_MaxFeatures, GMG,
+            Combine(Values(string("gpu/video/768x576.avi")),
+                    GPU_CHANNELS_1_3_4,
+                    Values(20, 40, 60)))
+{
+    const std::string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
+    const int cn = GET_PARAM(1);
+    const int maxFeatures = GET_PARAM(2);
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    if (cn != 3)
+    {
+        cv::Mat temp;
+        if (cn == 1)
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+        else
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+        cv::swap(temp, frame);
+    }
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat d_frame(frame);
+        cv::gpu::GpuMat foreground;
+
+        cv::Ptr<cv::BackgroundSubtractorGMG> d_gmg = cv::gpu::createBackgroundSubtractorGMG();
+        d_gmg->setMaxFeatures(maxFeatures);
+
+        d_gmg->apply(d_frame, foreground);
+
+        for (int i = 0; i < 150; ++i)
+        {
+            cap >> frame;
+            if (frame.empty())
+            {
+                cap.release();
+                cap.open(inputFile);
+                cap >> frame;
+            }
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            d_frame.upload(frame);
+
+            startTimer(); next();
+            d_gmg->apply(d_frame, foreground);
+            stopTimer();
+        }
+
+        GPU_SANITY_CHECK(foreground);
+    }
+    else
+    {
+        cv::Mat foreground;
+        cv::Mat zeros(frame.size(), CV_8UC1, cv::Scalar::all(0));
+
+        cv::Ptr<cv::BackgroundSubtractorGMG> gmg = cv::createBackgroundSubtractorGMG();
+        gmg->setMaxFeatures(maxFeatures);
+
+        gmg->apply(frame, foreground);
+
+        for (int i = 0; i < 150; ++i)
+        {
+            cap >> frame;
+            if (frame.empty())
+            {
+                cap.release();
+                cap.open(inputFile);
+                cap >> frame;
+            }
+
+            if (cn != 3)
+            {
+                cv::Mat temp;
+                if (cn == 1)
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+                else
+                    cv::cvtColor(frame, temp, cv::COLOR_BGR2BGRA);
+                cv::swap(temp, frame);
+            }
+
+            startTimer(); next();
+            gmg->apply(frame, foreground);
+            stopTimer();
+        }
+
+        CPU_SANITY_CHECK(foreground);
+    }
+}
+
+#endif
diff --git a/modules/gpubgsegm/perf/perf_main.cpp b/modules/gpubgsegm/perf/perf_main.cpp
new file mode 100644
index 0000000000..f69d0ea8fd
--- /dev/null
+++ b/modules/gpubgsegm/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpubgsegm)
diff --git a/modules/gpubgsegm/perf/perf_precomp.hpp b/modules/gpubgsegm/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..83d0fd4f0f
--- /dev/null
+++ b/modules/gpubgsegm/perf/perf_precomp.hpp
@@ -0,0 +1,66 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpubgsegm.hpp"
+#include "opencv2/video.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpu/src/cuda/fgd_bgfg.cu b/modules/gpubgsegm/src/cuda/fgd.cu
similarity index 98%
rename from modules/gpu/src/cuda/fgd_bgfg.cu
rename to modules/gpubgsegm/src/cuda/fgd.cu
index a14da0f854..7dd616c712 100644
--- a/modules/gpu/src/cuda/fgd_bgfg.cu
+++ b/modules/gpubgsegm/src/cuda/fgd.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,18 +42,18 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "fgd_bgfg_common.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "fgd.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
-namespace bgfg
+namespace fgd
 {
     ////////////////////////////////////////////////////////////////////////////
     // calcDiffHistogram
diff --git a/modules/gpu/src/cuda/fgd_bgfg_common.hpp b/modules/gpubgsegm/src/cuda/fgd.hpp
similarity index 70%
rename from modules/gpu/src/cuda/fgd_bgfg_common.hpp
rename to modules/gpubgsegm/src/cuda/fgd.hpp
index 986952e439..50b9838cdf 100644
--- a/modules/gpu/src/cuda/fgd_bgfg_common.hpp
+++ b/modules/gpubgsegm/src/cuda/fgd.hpp
@@ -1,9 +1,51 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifndef __FGD_BGFG_COMMON_HPP__
 #define __FGD_BGFG_COMMON_HPP__
 
-#include "opencv2/core/cuda_devptrs.hpp"
+#include "opencv2/core/gpu_types.hpp"
 
-namespace bgfg
+namespace fgd
 {
     struct BGPixelStat
     {
diff --git a/modules/gpu/src/cuda/bgfg_gmg.cu b/modules/gpubgsegm/src/cuda/gmg.cu
similarity index 97%
rename from modules/gpu/src/cuda/bgfg_gmg.cu
rename to modules/gpubgsegm/src/cuda/gmg.cu
index 8e2fd3b38b..235c1f0e2e 100644
--- a/modules/gpu/src/cuda/bgfg_gmg.cu
+++ b/modules/gpubgsegm/src/cuda/gmg.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/limits.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/limits.hpp"
 
-namespace cv { namespace gpu { namespace device {
-    namespace bgfg_gmg
+namespace cv { namespace gpu { namespace cudev {
+    namespace gmg
     {
         __constant__ int   c_width;
         __constant__ int   c_height;
@@ -255,4 +255,4 @@ namespace cv { namespace gpu { namespace device {
 }}}
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/bgfg_mog.cu b/modules/gpubgsegm/src/cuda/mog.cu
similarity index 56%
rename from modules/gpu/src/cuda/bgfg_mog.cu
rename to modules/gpubgsegm/src/cuda/mog.cu
index d99ffc498a..ed0720c30c 100644
--- a/modules/gpu/src/cuda/bgfg_mog.cu
+++ b/modules/gpubgsegm/src/cuda/mog.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/limits.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/limits.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace mog
     {
@@ -111,14 +111,6 @@ namespace cv { namespace gpu { namespace device
                                 0.0f);
         }
 
-        template <class Ptr2D>
-        __device__ __forceinline__ void swap(Ptr2D& ptr, int x, int y, int k, int rows)
-        {
-            typename Ptr2D::elem_type val = ptr(k * rows + y, x);
-            ptr(k * rows + y, x) = ptr((k + 1) * rows + y, x);
-            ptr((k + 1) * rows + y, x) = val;
-        }
-
         ///////////////////////////////////////////////////////////////
         // MOG without learning
 
@@ -426,337 +418,6 @@ namespace cv { namespace gpu { namespace device
 
             funcs[cn](weight, mean, dst, nmixtures, backgroundRatio, stream);
         }
-
-        ///////////////////////////////////////////////////////////////
-        // MOG2
-
-        __constant__ int           c_nmixtures;
-        __constant__ float         c_Tb;
-        __constant__ float         c_TB;
-        __constant__ float         c_Tg;
-        __constant__ float         c_varInit;
-        __constant__ float         c_varMin;
-        __constant__ float         c_varMax;
-        __constant__ float         c_tau;
-        __constant__ unsigned char c_shadowVal;
-
-        void loadConstants(int nmixtures, float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau, unsigned char shadowVal)
-        {
-            varMin = ::fminf(varMin, varMax);
-            varMax = ::fmaxf(varMin, varMax);
-
-            cudaSafeCall( cudaMemcpyToSymbol(c_nmixtures, &nmixtures, sizeof(int)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_Tb, &Tb, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_TB, &TB, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_Tg, &Tg, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_varInit, &varInit, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_varMin, &varMin, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_varMax, &varMax, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_tau, &tau, sizeof(float)) );
-            cudaSafeCall( cudaMemcpyToSymbol(c_shadowVal, &shadowVal, sizeof(unsigned char)) );
-        }
-
-        template <bool detectShadows, typename SrcT, typename WorkT>
-        __global__ void mog2(const PtrStepSz<SrcT> frame, PtrStepb fgmask, PtrStepb modesUsed,
-                             PtrStepf gmm_weight, PtrStepf gmm_variance, PtrStep<WorkT> gmm_mean,
-                             const float alphaT, const float alpha1, const float prune)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= frame.cols || y >= frame.rows)
-                return;
-
-            WorkT pix = cvt(frame(y, x));
-
-            //calculate distances to the modes (+ sort)
-            //here we need to go in descending order!!!
-
-            bool background = false; // true - the pixel classified as background
-
-            //internal:
-
-            bool fitsPDF = false; //if it remains zero a new GMM mode will be added
-
-            int nmodes = modesUsed(y, x);
-            int nNewModes = nmodes; //current number of modes in GMM
-
-            float totalWeight = 0.0f;
-
-            //go through all modes
-
-            for (int mode = 0; mode < nmodes; ++mode)
-            {
-                //need only weight if fit is found
-                float weight = alpha1 * gmm_weight(mode * frame.rows + y, x) + prune;
-
-                //fit not found yet
-                if (!fitsPDF)
-                {
-                    //check if it belongs to some of the remaining modes
-                    float var = gmm_variance(mode * frame.rows + y, x);
-
-                    WorkT mean = gmm_mean(mode * frame.rows + y, x);
-
-                    //calculate difference and distance
-                    WorkT diff = mean - pix;
-                    float dist2 = sqr(diff);
-
-                    //background? - Tb - usually larger than Tg
-                    if (totalWeight < c_TB && dist2 < c_Tb * var)
-                        background = true;
-
-                    //check fit
-                    if (dist2 < c_Tg * var)
-                    {
-                        //belongs to the mode
-                        fitsPDF = true;
-
-                        //update distribution
-
-                        //update weight
-                        weight += alphaT;
-                        float k = alphaT / weight;
-
-                        //update mean
-                        gmm_mean(mode * frame.rows + y, x) = mean - k * diff;
-
-                        //update variance
-                        float varnew = var + k * (dist2 - var);
-
-                        //limit the variance
-                        varnew = ::fmaxf(varnew, c_varMin);
-                        varnew = ::fminf(varnew, c_varMax);
-
-                        gmm_variance(mode * frame.rows + y, x) = varnew;
-
-                        //sort
-                        //all other weights are at the same place and
-                        //only the matched (iModes) is higher -> just find the new place for it
-
-                        for (int i = mode; i > 0; --i)
-                        {
-                            //check one up
-                            if (weight < gmm_weight((i - 1) * frame.rows + y, x))
-                                break;
-
-                            //swap one up
-                            swap(gmm_weight, x, y, i - 1, frame.rows);
-                            swap(gmm_variance, x, y, i - 1, frame.rows);
-                            swap(gmm_mean, x, y, i - 1, frame.rows);
-                        }
-
-                        //belongs to the mode - bFitsPDF becomes 1
-                    }
-                } // !fitsPDF
-
-                //check prune
-                if (weight < -prune)
-                {
-                    weight = 0.0;
-                    nmodes--;
-                }
-
-                gmm_weight(mode * frame.rows + y, x) = weight; //update weight by the calculated value
-                totalWeight += weight;
-            }
-
-            //renormalize weights
-
-            totalWeight = 1.f / totalWeight;
-            for (int mode = 0; mode < nmodes; ++mode)
-                gmm_weight(mode * frame.rows + y, x) *= totalWeight;
-
-            nmodes = nNewModes;
-
-            //make new mode if needed and exit
-
-            if (!fitsPDF)
-            {
-                // replace the weakest or add a new one
-                int mode = nmodes == c_nmixtures ? c_nmixtures - 1 : nmodes++;
-
-                if (nmodes == 1)
-                    gmm_weight(mode * frame.rows + y, x) = 1.f;
-                else
-                {
-                    gmm_weight(mode * frame.rows + y, x) = alphaT;
-
-                    // renormalize all other weights
-
-                    for (int i = 0; i < nmodes - 1; ++i)
-                        gmm_weight(i * frame.rows + y, x) *= alpha1;
-                }
-
-                // init
-
-                gmm_mean(mode * frame.rows + y, x) = pix;
-                gmm_variance(mode * frame.rows + y, x) = c_varInit;
-
-                //sort
-                //find the new place for it
-
-                for (int i = nmodes - 1; i > 0; --i)
-                {
-                    // check one up
-                    if (alphaT < gmm_weight((i - 1) * frame.rows + y, x))
-                        break;
-
-                    //swap one up
-                    swap(gmm_weight, x, y, i - 1, frame.rows);
-                    swap(gmm_variance, x, y, i - 1, frame.rows);
-                    swap(gmm_mean, x, y, i - 1, frame.rows);
-                }
-            }
-
-            //set the number of modes
-            modesUsed(y, x) = nmodes;
-
-            bool isShadow = false;
-            if (detectShadows && !background)
-            {
-                float tWeight = 0.0f;
-
-                // check all the components  marked as background:
-                for (int mode = 0; mode < nmodes; ++mode)
-                {
-                    WorkT mean = gmm_mean(mode * frame.rows + y, x);
-
-                    WorkT pix_mean = pix * mean;
-
-                    float numerator = sum(pix_mean);
-                    float denominator = sqr(mean);
-
-                    // no division by zero allowed
-                    if (denominator == 0)
-                        break;
-
-                    // if tau < a < 1 then also check the color distortion
-                    if (numerator <= denominator && numerator >= c_tau * denominator)
-                    {
-                        float a = numerator / denominator;
-
-                        WorkT dD = a * mean - pix;
-
-                        if (sqr(dD) < c_Tb * gmm_variance(mode * frame.rows + y, x) * a * a)
-                        {
-                            isShadow = true;
-                            break;
-                        }
-                    };
-
-                    tWeight += gmm_weight(mode * frame.rows + y, x);
-                    if (tWeight > c_TB)
-                        break;
-                }
-            }
-
-            fgmask(y, x) = background ? 0 : isShadow ? c_shadowVal : 255;
-        }
-
-        template <typename SrcT, typename WorkT>
-        void mog2_caller(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean,
-                         float alphaT, float prune, bool detectShadows, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(frame.cols, block.x), divUp(frame.rows, block.y));
-
-            const float alpha1 = 1.0f - alphaT;
-
-            if (detectShadows)
-            {
-                cudaSafeCall( cudaFuncSetCacheConfig(mog2<true, SrcT, WorkT>, cudaFuncCachePreferL1) );
-
-                mog2<true, SrcT, WorkT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, modesUsed,
-                                                                    weight, variance, (PtrStepSz<WorkT>) mean,
-                                                                    alphaT, alpha1, prune);
-            }
-            else
-            {
-                cudaSafeCall( cudaFuncSetCacheConfig(mog2<false, SrcT, WorkT>, cudaFuncCachePreferL1) );
-
-                mog2<false, SrcT, WorkT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, modesUsed,
-                                                                    weight, variance, (PtrStepSz<WorkT>) mean,
-                                                                    alphaT, alpha1, prune);
-            }
-
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void mog2_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean,
-                      float alphaT, float prune, bool detectShadows, cudaStream_t stream)
-        {
-            typedef void (*func_t)(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean, float alphaT, float prune, bool detectShadows, cudaStream_t stream);
-
-            static const func_t funcs[] =
-            {
-                0, mog2_caller<uchar, float>, 0, mog2_caller<uchar3, float3>, mog2_caller<uchar4, float4>
-            };
-
-            funcs[cn](frame, fgmask, modesUsed, weight, variance, mean, alphaT, prune, detectShadows, stream);
-        }
-
-        template <typename WorkT, typename OutT>
-        __global__ void getBackgroundImage2(const PtrStepSzb modesUsed, const PtrStepf gmm_weight, const PtrStep<WorkT> gmm_mean, PtrStep<OutT> dst)
-        {
-            const int x = blockIdx.x * blockDim.x + threadIdx.x;
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-
-            if (x >= modesUsed.cols || y >= modesUsed.rows)
-                return;
-
-            int nmodes = modesUsed(y, x);
-
-            WorkT meanVal = VecTraits<WorkT>::all(0.0f);
-            float totalWeight = 0.0f;
-
-            for (int mode = 0; mode < nmodes; ++mode)
-            {
-                float weight = gmm_weight(mode * modesUsed.rows + y, x);
-
-                WorkT mean = gmm_mean(mode * modesUsed.rows + y, x);
-                meanVal = meanVal + weight * mean;
-
-                totalWeight += weight;
-
-                if(totalWeight > c_TB)
-                    break;
-            }
-
-            meanVal = meanVal * (1.f / totalWeight);
-
-            dst(y, x) = saturate_cast<OutT>(meanVal);
-        }
-
-        template <typename WorkT, typename OutT>
-        void getBackgroundImage2_caller(PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream)
-        {
-            dim3 block(32, 8);
-            dim3 grid(divUp(modesUsed.cols, block.x), divUp(modesUsed.rows, block.y));
-
-            cudaSafeCall( cudaFuncSetCacheConfig(getBackgroundImage2<WorkT, OutT>, cudaFuncCachePreferL1) );
-
-            getBackgroundImage2<WorkT, OutT><<<grid, block, 0, stream>>>(modesUsed, weight, (PtrStepSz<WorkT>) mean, (PtrStepSz<OutT>) dst);
-            cudaSafeCall( cudaGetLastError() );
-
-            if (stream == 0)
-                cudaSafeCall( cudaDeviceSynchronize() );
-        }
-
-        void getBackgroundImage2_gpu(int cn, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream)
-        {
-            typedef void (*func_t)(PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream);
-
-            static const func_t funcs[] =
-            {
-                0, getBackgroundImage2_caller<float, uchar>, 0, getBackgroundImage2_caller<float3, uchar3>, getBackgroundImage2_caller<float4, uchar4>
-            };
-
-            funcs[cn](modesUsed, weight, mean, dst, stream);
-        }
     }
 }}}
 
diff --git a/modules/gpubgsegm/src/cuda/mog2.cu b/modules/gpubgsegm/src/cuda/mog2.cu
new file mode 100644
index 0000000000..50cb9fa56d
--- /dev/null
+++ b/modules/gpubgsegm/src/cuda/mog2.cu
@@ -0,0 +1,438 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace mog2
+    {
+        ///////////////////////////////////////////////////////////////
+        // Utility
+
+        __device__ __forceinline__ float cvt(uchar val)
+        {
+            return val;
+        }
+        __device__ __forceinline__ float3 cvt(const uchar3& val)
+        {
+            return make_float3(val.x, val.y, val.z);
+        }
+        __device__ __forceinline__ float4 cvt(const uchar4& val)
+        {
+            return make_float4(val.x, val.y, val.z, val.w);
+        }
+
+        __device__ __forceinline__ float sqr(float val)
+        {
+            return val * val;
+        }
+        __device__ __forceinline__ float sqr(const float3& val)
+        {
+            return val.x * val.x + val.y * val.y + val.z * val.z;
+        }
+        __device__ __forceinline__ float sqr(const float4& val)
+        {
+            return val.x * val.x + val.y * val.y + val.z * val.z;
+        }
+
+        __device__ __forceinline__ float sum(float val)
+        {
+            return val;
+        }
+        __device__ __forceinline__ float sum(const float3& val)
+        {
+            return val.x + val.y + val.z;
+        }
+        __device__ __forceinline__ float sum(const float4& val)
+        {
+            return val.x + val.y + val.z;
+        }
+
+        template <class Ptr2D>
+        __device__ __forceinline__ void swap(Ptr2D& ptr, int x, int y, int k, int rows)
+        {
+            typename Ptr2D::elem_type val = ptr(k * rows + y, x);
+            ptr(k * rows + y, x) = ptr((k + 1) * rows + y, x);
+            ptr((k + 1) * rows + y, x) = val;
+        }
+
+        ///////////////////////////////////////////////////////////////
+        // MOG2
+
+        __constant__ int           c_nmixtures;
+        __constant__ float         c_Tb;
+        __constant__ float         c_TB;
+        __constant__ float         c_Tg;
+        __constant__ float         c_varInit;
+        __constant__ float         c_varMin;
+        __constant__ float         c_varMax;
+        __constant__ float         c_tau;
+        __constant__ unsigned char c_shadowVal;
+
+        void loadConstants(int nmixtures, float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau, unsigned char shadowVal)
+        {
+            varMin = ::fminf(varMin, varMax);
+            varMax = ::fmaxf(varMin, varMax);
+
+            cudaSafeCall( cudaMemcpyToSymbol(c_nmixtures, &nmixtures, sizeof(int)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_Tb, &Tb, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_TB, &TB, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_Tg, &Tg, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_varInit, &varInit, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_varMin, &varMin, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_varMax, &varMax, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_tau, &tau, sizeof(float)) );
+            cudaSafeCall( cudaMemcpyToSymbol(c_shadowVal, &shadowVal, sizeof(unsigned char)) );
+        }
+
+        template <bool detectShadows, typename SrcT, typename WorkT>
+        __global__ void mog2(const PtrStepSz<SrcT> frame, PtrStepb fgmask, PtrStepb modesUsed,
+                             PtrStepf gmm_weight, PtrStepf gmm_variance, PtrStep<WorkT> gmm_mean,
+                             const float alphaT, const float alpha1, const float prune)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x >= frame.cols || y >= frame.rows)
+                return;
+
+            WorkT pix = cvt(frame(y, x));
+
+            //calculate distances to the modes (+ sort)
+            //here we need to go in descending order!!!
+
+            bool background = false; // true - the pixel classified as background
+
+            //internal:
+
+            bool fitsPDF = false; //if it remains zero a new GMM mode will be added
+
+            int nmodes = modesUsed(y, x);
+            int nNewModes = nmodes; //current number of modes in GMM
+
+            float totalWeight = 0.0f;
+
+            //go through all modes
+
+            for (int mode = 0; mode < nmodes; ++mode)
+            {
+                //need only weight if fit is found
+                float weight = alpha1 * gmm_weight(mode * frame.rows + y, x) + prune;
+
+                //fit not found yet
+                if (!fitsPDF)
+                {
+                    //check if it belongs to some of the remaining modes
+                    float var = gmm_variance(mode * frame.rows + y, x);
+
+                    WorkT mean = gmm_mean(mode * frame.rows + y, x);
+
+                    //calculate difference and distance
+                    WorkT diff = mean - pix;
+                    float dist2 = sqr(diff);
+
+                    //background? - Tb - usually larger than Tg
+                    if (totalWeight < c_TB && dist2 < c_Tb * var)
+                        background = true;
+
+                    //check fit
+                    if (dist2 < c_Tg * var)
+                    {
+                        //belongs to the mode
+                        fitsPDF = true;
+
+                        //update distribution
+
+                        //update weight
+                        weight += alphaT;
+                        float k = alphaT / weight;
+
+                        //update mean
+                        gmm_mean(mode * frame.rows + y, x) = mean - k * diff;
+
+                        //update variance
+                        float varnew = var + k * (dist2 - var);
+
+                        //limit the variance
+                        varnew = ::fmaxf(varnew, c_varMin);
+                        varnew = ::fminf(varnew, c_varMax);
+
+                        gmm_variance(mode * frame.rows + y, x) = varnew;
+
+                        //sort
+                        //all other weights are at the same place and
+                        //only the matched (iModes) is higher -> just find the new place for it
+
+                        for (int i = mode; i > 0; --i)
+                        {
+                            //check one up
+                            if (weight < gmm_weight((i - 1) * frame.rows + y, x))
+                                break;
+
+                            //swap one up
+                            swap(gmm_weight, x, y, i - 1, frame.rows);
+                            swap(gmm_variance, x, y, i - 1, frame.rows);
+                            swap(gmm_mean, x, y, i - 1, frame.rows);
+                        }
+
+                        //belongs to the mode - bFitsPDF becomes 1
+                    }
+                } // !fitsPDF
+
+                //check prune
+                if (weight < -prune)
+                {
+                    weight = 0.0f;
+                    nmodes--;
+                }
+
+                gmm_weight(mode * frame.rows + y, x) = weight; //update weight by the calculated value
+                totalWeight += weight;
+            }
+
+            //renormalize weights
+
+            totalWeight = 1.f / totalWeight;
+            for (int mode = 0; mode < nmodes; ++mode)
+                gmm_weight(mode * frame.rows + y, x) *= totalWeight;
+
+            nmodes = nNewModes;
+
+            //make new mode if needed and exit
+
+            if (!fitsPDF)
+            {
+                // replace the weakest or add a new one
+                int mode = nmodes == c_nmixtures ? c_nmixtures - 1 : nmodes++;
+
+                if (nmodes == 1)
+                    gmm_weight(mode * frame.rows + y, x) = 1.f;
+                else
+                {
+                    gmm_weight(mode * frame.rows + y, x) = alphaT;
+
+                    // renormalize all other weights
+
+                    for (int i = 0; i < nmodes - 1; ++i)
+                        gmm_weight(i * frame.rows + y, x) *= alpha1;
+                }
+
+                // init
+
+                gmm_mean(mode * frame.rows + y, x) = pix;
+                gmm_variance(mode * frame.rows + y, x) = c_varInit;
+
+                //sort
+                //find the new place for it
+
+                for (int i = nmodes - 1; i > 0; --i)
+                {
+                    // check one up
+                    if (alphaT < gmm_weight((i - 1) * frame.rows + y, x))
+                        break;
+
+                    //swap one up
+                    swap(gmm_weight, x, y, i - 1, frame.rows);
+                    swap(gmm_variance, x, y, i - 1, frame.rows);
+                    swap(gmm_mean, x, y, i - 1, frame.rows);
+                }
+            }
+
+            //set the number of modes
+            modesUsed(y, x) = nmodes;
+
+            bool isShadow = false;
+            if (detectShadows && !background)
+            {
+                float tWeight = 0.0f;
+
+                // check all the components  marked as background:
+                for (int mode = 0; mode < nmodes; ++mode)
+                {
+                    WorkT mean = gmm_mean(mode * frame.rows + y, x);
+
+                    WorkT pix_mean = pix * mean;
+
+                    float numerator = sum(pix_mean);
+                    float denominator = sqr(mean);
+
+                    // no division by zero allowed
+                    if (denominator == 0)
+                        break;
+
+                    // if tau < a < 1 then also check the color distortion
+                    if (numerator <= denominator && numerator >= c_tau * denominator)
+                    {
+                        float a = numerator / denominator;
+
+                        WorkT dD = a * mean - pix;
+
+                        if (sqr(dD) < c_Tb * gmm_variance(mode * frame.rows + y, x) * a * a)
+                        {
+                            isShadow = true;
+                            break;
+                        }
+                    };
+
+                    tWeight += gmm_weight(mode * frame.rows + y, x);
+                    if (tWeight > c_TB)
+                        break;
+                }
+            }
+
+            fgmask(y, x) = background ? 0 : isShadow ? c_shadowVal : 255;
+        }
+
+        template <typename SrcT, typename WorkT>
+        void mog2_caller(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean,
+                         float alphaT, float prune, bool detectShadows, cudaStream_t stream)
+        {
+            dim3 block(32, 8);
+            dim3 grid(divUp(frame.cols, block.x), divUp(frame.rows, block.y));
+
+            const float alpha1 = 1.0f - alphaT;
+
+            if (detectShadows)
+            {
+                cudaSafeCall( cudaFuncSetCacheConfig(mog2<true, SrcT, WorkT>, cudaFuncCachePreferL1) );
+
+                mog2<true, SrcT, WorkT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, modesUsed,
+                                                                    weight, variance, (PtrStepSz<WorkT>) mean,
+                                                                    alphaT, alpha1, prune);
+            }
+            else
+            {
+                cudaSafeCall( cudaFuncSetCacheConfig(mog2<false, SrcT, WorkT>, cudaFuncCachePreferL1) );
+
+                mog2<false, SrcT, WorkT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, modesUsed,
+                                                                    weight, variance, (PtrStepSz<WorkT>) mean,
+                                                                    alphaT, alpha1, prune);
+            }
+
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        void mog2_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean,
+                      float alphaT, float prune, bool detectShadows, cudaStream_t stream)
+        {
+            typedef void (*func_t)(PtrStepSzb frame, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean, float alphaT, float prune, bool detectShadows, cudaStream_t stream);
+
+            static const func_t funcs[] =
+            {
+                0, mog2_caller<uchar, float>, 0, mog2_caller<uchar3, float3>, mog2_caller<uchar4, float4>
+            };
+
+            funcs[cn](frame, fgmask, modesUsed, weight, variance, mean, alphaT, prune, detectShadows, stream);
+        }
+
+        template <typename WorkT, typename OutT>
+        __global__ void getBackgroundImage2(const PtrStepSzb modesUsed, const PtrStepf gmm_weight, const PtrStep<WorkT> gmm_mean, PtrStep<OutT> dst)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x >= modesUsed.cols || y >= modesUsed.rows)
+                return;
+
+            int nmodes = modesUsed(y, x);
+
+            WorkT meanVal = VecTraits<WorkT>::all(0.0f);
+            float totalWeight = 0.0f;
+
+            for (int mode = 0; mode < nmodes; ++mode)
+            {
+                float weight = gmm_weight(mode * modesUsed.rows + y, x);
+
+                WorkT mean = gmm_mean(mode * modesUsed.rows + y, x);
+                meanVal = meanVal + weight * mean;
+
+                totalWeight += weight;
+
+                if(totalWeight > c_TB)
+                    break;
+            }
+
+            meanVal = meanVal * (1.f / totalWeight);
+
+            dst(y, x) = saturate_cast<OutT>(meanVal);
+        }
+
+        template <typename WorkT, typename OutT>
+        void getBackgroundImage2_caller(PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream)
+        {
+            dim3 block(32, 8);
+            dim3 grid(divUp(modesUsed.cols, block.x), divUp(modesUsed.rows, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(getBackgroundImage2<WorkT, OutT>, cudaFuncCachePreferL1) );
+
+            getBackgroundImage2<WorkT, OutT><<<grid, block, 0, stream>>>(modesUsed, weight, (PtrStepSz<WorkT>) mean, (PtrStepSz<OutT>) dst);
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        void getBackgroundImage2_gpu(int cn, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream)
+        {
+            typedef void (*func_t)(PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream);
+
+            static const func_t funcs[] =
+            {
+                0, getBackgroundImage2_caller<float, uchar>, 0, getBackgroundImage2_caller<float3, uchar3>, getBackgroundImage2_caller<float4, uchar4>
+            };
+
+            funcs[cn](modesUsed, weight, mean, dst, stream);
+        }
+    }
+}}}
+
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpubgsegm/src/fgd.cpp b/modules/gpubgsegm/src/fgd.cpp
new file mode 100644
index 0000000000..31cf370c16
--- /dev/null
+++ b/modules/gpubgsegm/src/fgd.cpp
@@ -0,0 +1,731 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined(HAVE_CUDA) || defined(CUDA_DISABLER) || !defined(HAVE_OPENCV_IMGPROC) || !defined(HAVE_OPENCV_GPUARITHM) || !defined(HAVE_OPENCV_GPUIMGPROC)
+
+cv::gpu::FGDParams::FGDParams() { throw_no_cuda(); }
+
+Ptr<gpu::BackgroundSubtractorFGD> cv::gpu::createBackgroundSubtractorFGD(const FGDParams&) { throw_no_cuda(); return Ptr<gpu::BackgroundSubtractorFGD>(); }
+
+#else
+
+#include "cuda/fgd.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+
+/////////////////////////////////////////////////////////////////////////
+// FGDParams
+
+namespace
+{
+    // Default parameters of foreground detection algorithm:
+    const int BGFG_FGD_LC  = 128;
+    const int BGFG_FGD_N1C = 15;
+    const int BGFG_FGD_N2C = 25;
+
+    const int BGFG_FGD_LCC   = 64;
+    const int BGFG_FGD_N1CC = 25;
+    const int BGFG_FGD_N2CC = 40;
+
+    // Background reference image update parameter:
+    const float BGFG_FGD_ALPHA_1 = 0.1f;
+
+    // stat model update parameter
+    // 0.002f ~ 1K frame(~45sec), 0.005 ~ 18sec (if 25fps and absolutely static BG)
+    const float BGFG_FGD_ALPHA_2 = 0.005f;
+
+    // start value for alpha parameter (to fast initiate statistic model)
+    const float BGFG_FGD_ALPHA_3 = 0.1f;
+
+    const float BGFG_FGD_DELTA = 2.0f;
+
+    const float BGFG_FGD_T = 0.9f;
+
+    const float BGFG_FGD_MINAREA= 15.0f;
+}
+
+cv::gpu::FGDParams::FGDParams()
+{
+    Lc      = BGFG_FGD_LC;
+    N1c     = BGFG_FGD_N1C;
+    N2c     = BGFG_FGD_N2C;
+
+    Lcc     = BGFG_FGD_LCC;
+    N1cc    = BGFG_FGD_N1CC;
+    N2cc    = BGFG_FGD_N2CC;
+
+    delta   = BGFG_FGD_DELTA;
+
+    alpha1  = BGFG_FGD_ALPHA_1;
+    alpha2  = BGFG_FGD_ALPHA_2;
+    alpha3  = BGFG_FGD_ALPHA_3;
+
+    T       = BGFG_FGD_T;
+    minArea = BGFG_FGD_MINAREA;
+
+    is_obj_without_holes = true;
+    perform_morphing     = 1;
+}
+
+/////////////////////////////////////////////////////////////////////////
+// copyChannels
+
+namespace
+{
+    void copyChannels(const GpuMat& src, GpuMat& dst, int dst_cn = -1)
+    {
+        const int src_cn = src.channels();
+
+        if (dst_cn < 0)
+            dst_cn = src_cn;
+
+        gpu::ensureSizeIsEnough(src.size(), CV_MAKE_TYPE(src.depth(), dst_cn), dst);
+
+        if (src_cn == dst_cn)
+        {
+            src.copyTo(dst);
+        }
+        else
+        {
+            static const int cvt_codes[4][4] =
+            {
+                {-1, -1, COLOR_GRAY2BGR, COLOR_GRAY2BGRA},
+                {-1, -1, -1, -1},
+                {COLOR_BGR2GRAY, -1, -1, COLOR_BGR2BGRA},
+                {COLOR_BGRA2GRAY, -1, COLOR_BGRA2BGR, -1}
+            };
+
+            const int cvt_code = cvt_codes[src_cn - 1][dst_cn - 1];
+            CV_DbgAssert( cvt_code >= 0 );
+
+            gpu::cvtColor(src, dst, cvt_code, dst_cn);
+        }
+    }
+}
+
+/////////////////////////////////////////////////////////////////////////
+// changeDetection
+
+namespace
+{
+    void calcDiffHistogram(const GpuMat& prevFrame, const GpuMat& curFrame, GpuMat& hist, GpuMat& histBuf)
+    {
+        typedef void (*func_t)(PtrStepSzb prevFrame, PtrStepSzb curFrame,
+                               unsigned int* hist0, unsigned int* hist1, unsigned int* hist2,
+                               unsigned int* partialBuf0, unsigned int* partialBuf1, unsigned int* partialBuf2,
+                               bool cc20, cudaStream_t stream);
+        static const func_t funcs[4][4] =
+        {
+            {0,0,0,0},
+            {0,0,0,0},
+            {0,0,fgd::calcDiffHistogram_gpu<uchar3, uchar3>,fgd::calcDiffHistogram_gpu<uchar3, uchar4>},
+            {0,0,fgd::calcDiffHistogram_gpu<uchar4, uchar3>,fgd::calcDiffHistogram_gpu<uchar4, uchar4>}
+        };
+
+        hist.create(3, 256, CV_32SC1);
+        histBuf.create(3, fgd::PARTIAL_HISTOGRAM_COUNT * fgd::HISTOGRAM_BIN_COUNT, CV_32SC1);
+
+        funcs[prevFrame.channels() - 1][curFrame.channels() - 1](
+                    prevFrame, curFrame,
+                    hist.ptr<unsigned int>(0), hist.ptr<unsigned int>(1), hist.ptr<unsigned int>(2),
+                    histBuf.ptr<unsigned int>(0), histBuf.ptr<unsigned int>(1), histBuf.ptr<unsigned int>(2),
+                    deviceSupports(FEATURE_SET_COMPUTE_20), 0);
+    }
+
+    void calcRelativeVariance(unsigned int hist[3 * 256], double relativeVariance[3][fgd::HISTOGRAM_BIN_COUNT])
+    {
+        std::memset(relativeVariance, 0, 3 * fgd::HISTOGRAM_BIN_COUNT * sizeof(double));
+
+        for (int thres = fgd::HISTOGRAM_BIN_COUNT - 2; thres >= 0; --thres)
+        {
+            Vec3d sum(0.0, 0.0, 0.0);
+            Vec3d sqsum(0.0, 0.0, 0.0);
+            Vec3i count(0, 0, 0);
+
+            for (int j = thres; j < fgd::HISTOGRAM_BIN_COUNT; ++j)
+            {
+                sum[0]   += static_cast<double>(j) * hist[j];
+                sqsum[0] += static_cast<double>(j * j) * hist[j];
+                count[0] += hist[j];
+
+                sum[1]   += static_cast<double>(j) * hist[j + 256];
+                sqsum[1] += static_cast<double>(j * j) * hist[j + 256];
+                count[1] += hist[j + 256];
+
+                sum[2]   += static_cast<double>(j) * hist[j + 512];
+                sqsum[2] += static_cast<double>(j * j) * hist[j + 512];
+                count[2] += hist[j + 512];
+            }
+
+            count[0] = std::max(count[0], 1);
+            count[1] = std::max(count[1], 1);
+            count[2] = std::max(count[2], 1);
+
+            Vec3d my(
+                sum[0] / count[0],
+                sum[1] / count[1],
+                sum[2] / count[2]
+            );
+
+            relativeVariance[0][thres] = std::sqrt(sqsum[0] / count[0] - my[0] * my[0]);
+            relativeVariance[1][thres] = std::sqrt(sqsum[1] / count[1] - my[1] * my[1]);
+            relativeVariance[2][thres] = std::sqrt(sqsum[2] / count[2] - my[2] * my[2]);
+        }
+    }
+
+    void calcDiffThreshMask(const GpuMat& prevFrame, const GpuMat& curFrame, Vec3d bestThres, GpuMat& changeMask)
+    {
+        typedef void (*func_t)(PtrStepSzb prevFrame, PtrStepSzb curFrame, uchar3 bestThres, PtrStepSzb changeMask, cudaStream_t stream);
+        static const func_t funcs[4][4] =
+        {
+            {0,0,0,0},
+            {0,0,0,0},
+            {0,0,fgd::calcDiffThreshMask_gpu<uchar3, uchar3>,fgd::calcDiffThreshMask_gpu<uchar3, uchar4>},
+            {0,0,fgd::calcDiffThreshMask_gpu<uchar4, uchar3>,fgd::calcDiffThreshMask_gpu<uchar4, uchar4>}
+        };
+
+        changeMask.setTo(Scalar::all(0));
+
+        funcs[prevFrame.channels() - 1][curFrame.channels() - 1](prevFrame, curFrame,
+                                                                 make_uchar3((uchar)bestThres[0], (uchar)bestThres[1], (uchar)bestThres[2]),
+                                                                 changeMask, 0);
+    }
+
+    // performs change detection for Foreground detection algorithm
+    void changeDetection(const GpuMat& prevFrame, const GpuMat& curFrame, GpuMat& changeMask, GpuMat& hist, GpuMat& histBuf)
+    {
+        calcDiffHistogram(prevFrame, curFrame, hist, histBuf);
+
+        unsigned int histData[3 * 256];
+        Mat h_hist(3, 256, CV_32SC1, histData);
+        hist.download(h_hist);
+
+        double relativeVariance[3][fgd::HISTOGRAM_BIN_COUNT];
+        calcRelativeVariance(histData, relativeVariance);
+
+        // Find maximum:
+        Vec3d bestThres(10.0, 10.0, 10.0);
+        for (int i = 0; i < fgd::HISTOGRAM_BIN_COUNT; ++i)
+        {
+            bestThres[0] = std::max(bestThres[0], relativeVariance[0][i]);
+            bestThres[1] = std::max(bestThres[1], relativeVariance[1][i]);
+            bestThres[2] = std::max(bestThres[2], relativeVariance[2][i]);
+        }
+
+        calcDiffThreshMask(prevFrame, curFrame, bestThres, changeMask);
+    }
+}
+
+/////////////////////////////////////////////////////////////////////////
+// bgfgClassification
+
+namespace
+{
+    int bgfgClassification(const GpuMat& prevFrame, const GpuMat& curFrame,
+                           const GpuMat& Ftd, const GpuMat& Fbd,
+                           GpuMat& foreground, GpuMat& countBuf,
+                           const FGDParams& params, int out_cn)
+    {
+        typedef void (*func_t)(PtrStepSzb prevFrame, PtrStepSzb curFrame, PtrStepSzb Ftd, PtrStepSzb Fbd, PtrStepSzb foreground,
+                               int deltaC, int deltaCC, float alpha2, int N1c, int N1cc, cudaStream_t stream);
+        static const func_t funcs[4][4][4] =
+        {
+            {
+                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0},
+                {0,0,fgd::bgfgClassification_gpu<uchar3, uchar3, uchar3>,fgd::bgfgClassification_gpu<uchar3, uchar3, uchar4>},
+                {0,0,fgd::bgfgClassification_gpu<uchar3, uchar4, uchar3>,fgd::bgfgClassification_gpu<uchar3, uchar4, uchar4>}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0},
+                {0,0,fgd::bgfgClassification_gpu<uchar4, uchar3, uchar3>,fgd::bgfgClassification_gpu<uchar4, uchar3, uchar4>},
+                {0,0,fgd::bgfgClassification_gpu<uchar4, uchar4, uchar3>,fgd::bgfgClassification_gpu<uchar4, uchar4, uchar4>}
+            }
+        };
+
+        const int deltaC  = cvRound(params.delta * 256 / params.Lc);
+        const int deltaCC = cvRound(params.delta * 256 / params.Lcc);
+
+        funcs[prevFrame.channels() - 1][curFrame.channels() - 1][out_cn - 1](prevFrame, curFrame, Ftd, Fbd, foreground,
+                                                                             deltaC, deltaCC, params.alpha2,
+                                                                             params.N1c, params.N1cc, 0);
+
+        int count = gpu::countNonZero(foreground, countBuf);
+
+        gpu::multiply(foreground, Scalar::all(255), foreground);
+
+        return count;
+    }
+}
+
+/////////////////////////////////////////////////////////////////////////
+// smoothForeground
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+
+namespace
+{
+    void morphology(const GpuMat& src, GpuMat& dst, GpuMat& filterBrd, int brd, Ptr<gpu::Filter>& filter, Scalar brdVal)
+    {
+        gpu::copyMakeBorder(src, filterBrd, brd, brd, brd, brd, BORDER_CONSTANT, brdVal);
+        filter->apply(filterBrd(Rect(brd, brd, src.cols, src.rows)), dst);
+    }
+
+    void smoothForeground(GpuMat& foreground, GpuMat& filterBrd, GpuMat& buf,
+                          Ptr<gpu::Filter>& erodeFilter, Ptr<gpu::Filter>& dilateFilter,
+                          const FGDParams& params)
+    {
+        const int brd = params.perform_morphing;
+
+        const Scalar erodeBrdVal = Scalar::all(UCHAR_MAX);
+        const Scalar dilateBrdVal = Scalar::all(0);
+
+        // MORPH_OPEN
+        morphology(foreground, buf, filterBrd, brd, erodeFilter, erodeBrdVal);
+        morphology(buf, foreground, filterBrd, brd, dilateFilter, dilateBrdVal);
+
+        // MORPH_CLOSE
+        morphology(foreground, buf, filterBrd, brd, dilateFilter, dilateBrdVal);
+        morphology(buf, foreground, filterBrd, brd, erodeFilter, erodeBrdVal);
+    }
+}
+
+#endif
+
+/////////////////////////////////////////////////////////////////////////
+// findForegroundRegions
+
+namespace
+{
+    void seqToContours(CvSeq* _ccontours, CvMemStorage* storage, OutputArrayOfArrays _contours)
+    {
+        Seq<CvSeq*> all_contours(cvTreeToNodeSeq(_ccontours, sizeof(CvSeq), storage));
+
+        size_t total = all_contours.size();
+
+        _contours.create((int) total, 1, 0, -1, true);
+
+        SeqIterator<CvSeq*> it = all_contours.begin();
+        for (size_t i = 0; i < total; ++i, ++it)
+        {
+            CvSeq* c = *it;
+            ((CvContour*)c)->color = (int)i;
+            _contours.create((int)c->total, 1, CV_32SC2, (int)i, true);
+            Mat ci = _contours.getMat((int)i);
+            CV_Assert( ci.isContinuous() );
+            cvCvtSeqToArray(c, ci.data);
+        }
+    }
+
+    int findForegroundRegions(GpuMat& d_foreground, Mat& h_foreground, std::vector< std::vector<Point> >& foreground_regions,
+                              CvMemStorage* storage, const FGDParams& params)
+    {
+        int region_count = 0;
+
+        // Discard under-size foreground regions:
+
+        d_foreground.download(h_foreground);
+        IplImage ipl_foreground = h_foreground;
+        CvSeq* first_seq = 0;
+
+        cvFindContours(&ipl_foreground, storage, &first_seq, sizeof(CvContour), CV_RETR_LIST);
+
+        for (CvSeq* seq = first_seq; seq; seq = seq->h_next)
+        {
+            CvContour* cnt = reinterpret_cast<CvContour*>(seq);
+
+            if (cnt->rect.width * cnt->rect.height < params.minArea || (params.is_obj_without_holes && CV_IS_SEQ_HOLE(seq)))
+            {
+                // Delete under-size contour:
+                CvSeq* prev_seq = seq->h_prev;
+                if (prev_seq)
+                {
+                    prev_seq->h_next = seq->h_next;
+
+                    if (seq->h_next)
+                        seq->h_next->h_prev = prev_seq;
+                }
+                else
+                {
+                    first_seq = seq->h_next;
+
+                    if (seq->h_next)
+                        seq->h_next->h_prev = NULL;
+                }
+            }
+            else
+            {
+                region_count++;
+            }
+        }
+
+        seqToContours(first_seq, storage, foreground_regions);
+        h_foreground.setTo(0);
+
+        drawContours(h_foreground, foreground_regions, -1, Scalar::all(255), -1);
+
+        d_foreground.upload(h_foreground);
+
+        return region_count;
+    }
+}
+
+/////////////////////////////////////////////////////////////////////////
+// updateBackgroundModel
+
+namespace
+{
+    void updateBackgroundModel(const GpuMat& prevFrame, const GpuMat& curFrame, const GpuMat& Ftd, const GpuMat& Fbd,
+                               const GpuMat& foreground, GpuMat& background,
+                               const FGDParams& params)
+    {
+        typedef void (*func_t)(PtrStepSzb prevFrame, PtrStepSzb curFrame, PtrStepSzb Ftd, PtrStepSzb Fbd,
+                               PtrStepSzb foreground, PtrStepSzb background,
+                               int deltaC, int deltaCC, float alpha1, float alpha2, float alpha3, int N1c, int N1cc, int N2c, int N2cc, float T, cudaStream_t stream);
+        static const func_t funcs[4][4][4] =
+        {
+            {
+                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0}, {0,0,0,0}, {0,0,0,0}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0},
+                {0,0,fgd::updateBackgroundModel_gpu<uchar3, uchar3, uchar3>,fgd::updateBackgroundModel_gpu<uchar3, uchar3, uchar4>},
+                {0,0,fgd::updateBackgroundModel_gpu<uchar3, uchar4, uchar3>,fgd::updateBackgroundModel_gpu<uchar3, uchar4, uchar4>}
+            },
+            {
+                {0,0,0,0}, {0,0,0,0},
+                {0,0,fgd::updateBackgroundModel_gpu<uchar4, uchar3, uchar3>,fgd::updateBackgroundModel_gpu<uchar4, uchar3, uchar4>},
+                {0,0,fgd::updateBackgroundModel_gpu<uchar4, uchar4, uchar3>,fgd::updateBackgroundModel_gpu<uchar4, uchar4, uchar4>}
+            }
+        };
+
+        const int deltaC  = cvRound(params.delta * 256 / params.Lc);
+        const int deltaCC = cvRound(params.delta * 256 / params.Lcc);
+
+        funcs[prevFrame.channels() - 1][curFrame.channels() - 1][background.channels() - 1](
+                    prevFrame, curFrame, Ftd, Fbd, foreground, background,
+                    deltaC, deltaCC, params.alpha1, params.alpha2, params.alpha3,
+                    params.N1c, params.N1cc, params.N2c, params.N2cc, params.T,
+                    0);
+    }
+}
+
+
+namespace
+{
+    class BGPixelStat
+    {
+    public:
+        void create(Size size, const FGDParams& params);
+
+        void setTrained();
+
+        operator fgd::BGPixelStat();
+
+    private:
+        GpuMat Pbc_;
+        GpuMat Pbcc_;
+        GpuMat is_trained_st_model_;
+        GpuMat is_trained_dyn_model_;
+
+        GpuMat ctable_Pv_;
+        GpuMat ctable_Pvb_;
+        GpuMat ctable_v_;
+
+        GpuMat cctable_Pv_;
+        GpuMat cctable_Pvb_;
+        GpuMat cctable_v1_;
+        GpuMat cctable_v2_;
+    };
+
+    void BGPixelStat::create(Size size, const FGDParams& params)
+    {
+        gpu::ensureSizeIsEnough(size, CV_32FC1, Pbc_);
+        Pbc_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(size, CV_32FC1, Pbcc_);
+        Pbcc_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(size, CV_8UC1, is_trained_st_model_);
+        is_trained_st_model_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(size, CV_8UC1, is_trained_dyn_model_);
+        is_trained_dyn_model_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_32FC1, ctable_Pv_);
+        ctable_Pv_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_32FC1, ctable_Pvb_);
+        ctable_Pvb_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2c * size.height, size.width, CV_8UC4, ctable_v_);
+        ctable_v_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_32FC1, cctable_Pv_);
+        cctable_Pv_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_32FC1, cctable_Pvb_);
+        cctable_Pvb_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_8UC4, cctable_v1_);
+        cctable_v1_.setTo(Scalar::all(0));
+
+        gpu::ensureSizeIsEnough(params.N2cc * size.height, size.width, CV_8UC4, cctable_v2_);
+        cctable_v2_.setTo(Scalar::all(0));
+    }
+
+    void BGPixelStat::setTrained()
+    {
+        is_trained_st_model_.setTo(Scalar::all(1));
+        is_trained_dyn_model_.setTo(Scalar::all(1));
+    }
+
+    BGPixelStat::operator fgd::BGPixelStat()
+    {
+        fgd::BGPixelStat stat;
+
+        stat.rows_ = Pbc_.rows;
+
+        stat.Pbc_data_ = Pbc_.data;
+        stat.Pbc_step_ = Pbc_.step;
+
+        stat.Pbcc_data_ = Pbcc_.data;
+        stat.Pbcc_step_ = Pbcc_.step;
+
+        stat.is_trained_st_model_data_ = is_trained_st_model_.data;
+        stat.is_trained_st_model_step_ = is_trained_st_model_.step;
+
+        stat.is_trained_dyn_model_data_ = is_trained_dyn_model_.data;
+        stat.is_trained_dyn_model_step_ = is_trained_dyn_model_.step;
+
+        stat.ctable_Pv_data_ = ctable_Pv_.data;
+        stat.ctable_Pv_step_ = ctable_Pv_.step;
+
+        stat.ctable_Pvb_data_ = ctable_Pvb_.data;
+        stat.ctable_Pvb_step_ = ctable_Pvb_.step;
+
+        stat.ctable_v_data_ = ctable_v_.data;
+        stat.ctable_v_step_ = ctable_v_.step;
+
+        stat.cctable_Pv_data_ = cctable_Pv_.data;
+        stat.cctable_Pv_step_ = cctable_Pv_.step;
+
+        stat.cctable_Pvb_data_ = cctable_Pvb_.data;
+        stat.cctable_Pvb_step_ = cctable_Pvb_.step;
+
+        stat.cctable_v1_data_ = cctable_v1_.data;
+        stat.cctable_v1_step_ = cctable_v1_.step;
+
+        stat.cctable_v2_data_ = cctable_v2_.data;
+        stat.cctable_v2_step_ = cctable_v2_.step;
+
+        return stat;
+    }
+
+    class FGDImpl : public gpu::BackgroundSubtractorFGD
+    {
+    public:
+        explicit FGDImpl(const FGDParams& params);
+        ~FGDImpl();
+
+        void apply(InputArray image, OutputArray fgmask, double learningRate=-1);
+
+        void getBackgroundImage(OutputArray backgroundImage) const;
+
+        void getForegroundRegions(OutputArrayOfArrays foreground_regions);
+
+    private:
+        void initialize(const GpuMat& firstFrame);
+
+        FGDParams params_;
+        Size frameSize_;
+
+        GpuMat background_;
+        GpuMat foreground_;
+        std::vector< std::vector<Point> > foreground_regions_;
+
+        Mat h_foreground_;
+
+        GpuMat prevFrame_;
+        GpuMat Ftd_;
+        GpuMat Fbd_;
+        BGPixelStat stat_;
+
+        GpuMat hist_;
+        GpuMat histBuf_;
+
+        GpuMat countBuf_;
+
+        GpuMat buf_;
+        GpuMat filterBrd_;
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        Ptr<gpu::Filter> dilateFilter_;
+        Ptr<gpu::Filter> erodeFilter_;
+#endif
+
+        CvMemStorage* storage_;
+    };
+
+    FGDImpl::FGDImpl(const FGDParams& params) : params_(params), frameSize_(0, 0)
+    {
+        storage_ = cvCreateMemStorage();
+        CV_Assert( storage_ != 0 );
+    }
+
+    FGDImpl::~FGDImpl()
+    {
+        cvReleaseMemStorage(&storage_);
+    }
+
+    void FGDImpl::apply(InputArray _frame, OutputArray fgmask, double)
+    {
+        GpuMat curFrame = _frame.getGpuMat();
+
+        if (curFrame.size() != frameSize_)
+        {
+            initialize(curFrame);
+            return;
+        }
+
+        CV_Assert( curFrame.type() == CV_8UC3 || curFrame.type() == CV_8UC4 );
+        CV_Assert( curFrame.size() == prevFrame_.size() );
+
+        cvClearMemStorage(storage_);
+        foreground_regions_.clear();
+        foreground_.setTo(Scalar::all(0));
+
+        changeDetection(prevFrame_, curFrame, Ftd_, hist_, histBuf_);
+        changeDetection(background_, curFrame, Fbd_, hist_, histBuf_);
+
+        int FG_pixels_count = bgfgClassification(prevFrame_, curFrame, Ftd_, Fbd_, foreground_, countBuf_, params_, 4);
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        if (params_.perform_morphing > 0)
+            smoothForeground(foreground_, filterBrd_, buf_, erodeFilter_, dilateFilter_, params_);
+#endif
+
+        if (params_.minArea > 0 || params_.is_obj_without_holes)
+            findForegroundRegions(foreground_, h_foreground_, foreground_regions_, storage_, params_);
+
+        // Check ALL BG update condition:
+        const double BGFG_FGD_BG_UPDATE_TRESH = 0.5;
+        if (static_cast<double>(FG_pixels_count) / Ftd_.size().area() > BGFG_FGD_BG_UPDATE_TRESH)
+            stat_.setTrained();
+
+        updateBackgroundModel(prevFrame_, curFrame, Ftd_, Fbd_, foreground_, background_, params_);
+
+        copyChannels(curFrame, prevFrame_, 4);
+
+        foreground_.copyTo(fgmask);
+    }
+
+    void FGDImpl::getBackgroundImage(OutputArray backgroundImage) const
+    {
+        gpu::cvtColor(background_, backgroundImage, COLOR_BGRA2BGR);
+    }
+
+    void FGDImpl::getForegroundRegions(OutputArrayOfArrays dst)
+    {
+        size_t total = foreground_regions_.size();
+
+        dst.create((int) total, 1, 0, -1, true);
+
+        for (size_t i = 0; i < total; ++i)
+        {
+            std::vector<Point>& c = foreground_regions_[i];
+
+            dst.create((int) c.size(), 1, CV_32SC2, (int) i, true);
+            Mat ci = dst.getMat((int) i);
+
+            Mat(ci.size(), ci.type(), &c[0]).copyTo(ci);
+        }
+    }
+
+    void FGDImpl::initialize(const GpuMat& firstFrame)
+    {
+        CV_Assert( firstFrame.type() == CV_8UC3 || firstFrame.type() == CV_8UC4 );
+
+        frameSize_ = firstFrame.size();
+
+        gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, foreground_);
+
+        copyChannels(firstFrame, background_, 4);
+        copyChannels(firstFrame, prevFrame_, 4);
+
+        gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, Ftd_);
+        gpu::ensureSizeIsEnough(firstFrame.size(), CV_8UC1, Fbd_);
+
+        stat_.create(firstFrame.size(), params_);
+        fgd::setBGPixelStat(stat_);
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        if (params_.perform_morphing > 0)
+        {
+            Mat kernel = getStructuringElement(MORPH_RECT, Size(1 + params_.perform_morphing * 2, 1 + params_.perform_morphing * 2));
+            Point anchor(params_.perform_morphing, params_.perform_morphing);
+
+            dilateFilter_ = gpu::createMorphologyFilter(MORPH_DILATE, CV_8UC1, kernel, anchor);
+            erodeFilter_ = gpu::createMorphologyFilter(MORPH_ERODE, CV_8UC1, kernel, anchor);
+        }
+#endif
+    }
+}
+
+Ptr<gpu::BackgroundSubtractorFGD> cv::gpu::createBackgroundSubtractorFGD(const FGDParams& params)
+{
+    return makePtr<FGDImpl>(params);
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/gpubgsegm/src/gmg.cpp b/modules/gpubgsegm/src/gmg.cpp
new file mode 100644
index 0000000000..f8650b2420
--- /dev/null
+++ b/modules/gpubgsegm/src/gmg.cpp
@@ -0,0 +1,277 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+Ptr<gpu::BackgroundSubtractorGMG> cv::gpu::createBackgroundSubtractorGMG(int, double) { throw_no_cuda(); return Ptr<gpu::BackgroundSubtractorGMG>(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev {
+    namespace gmg
+    {
+        void loadConstants(int width, int height, float minVal, float maxVal, int quantizationLevels, float backgroundPrior,
+                           float decisionThreshold, int maxFeatures, int numInitializationFrames);
+
+        template <typename SrcT>
+        void update_gpu(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures,
+                        int frameNum,  float learningRate, bool updateBackgroundModel, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    class GMGImpl : public gpu::BackgroundSubtractorGMG
+    {
+    public:
+        GMGImpl(int initializationFrames, double decisionThreshold);
+
+        void apply(InputArray image, OutputArray fgmask, double learningRate=-1);
+        void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream);
+
+        void getBackgroundImage(OutputArray backgroundImage) const;
+
+        int getMaxFeatures() const { return maxFeatures_; }
+        void setMaxFeatures(int maxFeatures) { maxFeatures_ = maxFeatures; }
+
+        double getDefaultLearningRate() const { return learningRate_; }
+        void setDefaultLearningRate(double lr) { learningRate_ = (float) lr; }
+
+        int getNumFrames() const { return numInitializationFrames_; }
+        void setNumFrames(int nframes) { numInitializationFrames_ = nframes; }
+
+        int getQuantizationLevels() const { return quantizationLevels_; }
+        void setQuantizationLevels(int nlevels) { quantizationLevels_ = nlevels; }
+
+        double getBackgroundPrior() const { return backgroundPrior_; }
+        void setBackgroundPrior(double bgprior) { backgroundPrior_ = (float) bgprior; }
+
+        int getSmoothingRadius() const { return smoothingRadius_; }
+        void setSmoothingRadius(int radius) { smoothingRadius_ = radius; }
+
+        double getDecisionThreshold() const { return decisionThreshold_; }
+        void setDecisionThreshold(double thresh) { decisionThreshold_ = (float) thresh; }
+
+        bool getUpdateBackgroundModel() const { return updateBackgroundModel_; }
+        void setUpdateBackgroundModel(bool update) { updateBackgroundModel_ = update; }
+
+        double getMinVal() const { return minVal_; }
+        void setMinVal(double val) { minVal_ = (float) val; }
+
+        double getMaxVal() const { return maxVal_; }
+        void setMaxVal(double val) { maxVal_ = (float) val; }
+
+    private:
+        void initialize(Size frameSize, float min, float max);
+
+        //! Total number of distinct colors to maintain in histogram.
+        int maxFeatures_;
+
+        //! Set between 0.0 and 1.0, determines how quickly features are "forgotten" from histograms.
+        float learningRate_;
+
+        //! Number of frames of video to use to initialize histograms.
+        int numInitializationFrames_;
+
+        //! Number of discrete levels in each channel to be used in histograms.
+        int quantizationLevels_;
+
+        //! Prior probability that any given pixel is a background pixel. A sensitivity parameter.
+        float backgroundPrior_;
+
+        //! Smoothing radius, in pixels, for cleaning up FG image.
+        int smoothingRadius_;
+
+        //! Value above which pixel is determined to be FG.
+        float decisionThreshold_;
+
+        //! Perform background model update.
+        bool updateBackgroundModel_;
+
+        float minVal_, maxVal_;
+
+        Size frameSize_;
+        int frameNum_;
+
+        GpuMat nfeatures_;
+        GpuMat colors_;
+        GpuMat weights_;
+
+#if defined(HAVE_OPENCV_GPUFILTERS) && defined(HAVE_OPENCV_GPUARITHM)
+        Ptr<gpu::Filter> boxFilter_;
+        GpuMat buf_;
+#endif
+    };
+
+    GMGImpl::GMGImpl(int initializationFrames, double decisionThreshold)
+    {
+        maxFeatures_ = 64;
+        learningRate_ = 0.025f;
+        numInitializationFrames_ = initializationFrames;
+        quantizationLevels_ = 16;
+        backgroundPrior_ = 0.8f;
+        decisionThreshold_ = (float) decisionThreshold;
+        smoothingRadius_ = 7;
+        updateBackgroundModel_ = true;
+        minVal_ = maxVal_ = 0;
+    }
+
+    void GMGImpl::apply(InputArray image, OutputArray fgmask, double learningRate)
+    {
+        apply(image, fgmask, learningRate, Stream::Null());
+    }
+
+    void GMGImpl::apply(InputArray _frame, OutputArray _fgmask, double newLearningRate, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::gmg;
+
+        typedef void (*func_t)(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures,
+                               int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream);
+        static const func_t funcs[6][4] =
+        {
+            {update_gpu<uchar>, 0, update_gpu<uchar3>, update_gpu<uchar4>},
+            {0,0,0,0},
+            {update_gpu<ushort>, 0, update_gpu<ushort3>, update_gpu<ushort4>},
+            {0,0,0,0},
+            {0,0,0,0},
+            {update_gpu<float>, 0, update_gpu<float3>, update_gpu<float4>}
+        };
+
+        GpuMat frame = _frame.getGpuMat();
+
+        CV_Assert( frame.depth() == CV_8U || frame.depth() == CV_16U || frame.depth() == CV_32F );
+        CV_Assert( frame.channels() == 1 || frame.channels() == 3 || frame.channels() == 4 );
+
+        if (newLearningRate != -1.0)
+        {
+            CV_Assert( newLearningRate >= 0.0 && newLearningRate <= 1.0 );
+            learningRate_ = (float) newLearningRate;
+        }
+
+        if (frame.size() != frameSize_)
+        {
+            double minVal = minVal_;
+            double maxVal = maxVal_;
+
+            if (minVal_ == 0 && maxVal_ == 0)
+            {
+                minVal = 0;
+                maxVal = frame.depth() == CV_8U ? 255.0 : frame.depth() == CV_16U ? std::numeric_limits<ushort>::max() : 1.0;
+            }
+
+            initialize(frame.size(), (float) minVal, (float) maxVal);
+        }
+
+        _fgmask.create(frameSize_, CV_8UC1);
+        GpuMat fgmask = _fgmask.getGpuMat();
+
+        fgmask.setTo(Scalar::all(0), stream);
+
+        funcs[frame.depth()][frame.channels() - 1](frame, fgmask, colors_, weights_, nfeatures_, frameNum_,
+                                                   learningRate_, updateBackgroundModel_, StreamAccessor::getStream(stream));
+
+#if defined(HAVE_OPENCV_GPUFILTERS) && defined(HAVE_OPENCV_GPUARITHM)
+        // medianBlur
+        if (smoothingRadius_ > 0)
+        {
+            boxFilter_->apply(fgmask, buf_, stream);
+            const int minCount = (smoothingRadius_ * smoothingRadius_ + 1) / 2;
+            const double thresh = 255.0 * minCount / (smoothingRadius_ * smoothingRadius_);
+            gpu::threshold(buf_, fgmask, thresh, 255.0, THRESH_BINARY, stream);
+        }
+#endif
+
+        // keep track of how many frames we have processed
+        ++frameNum_;
+    }
+
+    void GMGImpl::getBackgroundImage(OutputArray backgroundImage) const
+    {
+        (void) backgroundImage;
+        CV_Error(Error::StsNotImplemented, "Not implemented");
+    }
+
+    void GMGImpl::initialize(Size frameSize, float min, float max)
+    {
+        using namespace cv::gpu::cudev::gmg;
+
+        CV_Assert( maxFeatures_ > 0 );
+        CV_Assert( learningRate_ >= 0.0f && learningRate_ <= 1.0f);
+        CV_Assert( numInitializationFrames_ >= 1);
+        CV_Assert( quantizationLevels_ >= 1 && quantizationLevels_ <= 255);
+        CV_Assert( backgroundPrior_ >= 0.0f && backgroundPrior_ <= 1.0f);
+
+        minVal_ = min;
+        maxVal_ = max;
+        CV_Assert( minVal_ < maxVal_ );
+
+        frameSize_ = frameSize;
+
+        frameNum_ = 0;
+
+        nfeatures_.create(frameSize_, CV_32SC1);
+        colors_.create(maxFeatures_ * frameSize_.height, frameSize_.width, CV_32SC1);
+        weights_.create(maxFeatures_ * frameSize_.height, frameSize_.width, CV_32FC1);
+
+        nfeatures_.setTo(Scalar::all(0));
+
+#if defined(HAVE_OPENCV_GPUFILTERS) && defined(HAVE_OPENCV_GPUARITHM)
+        if (smoothingRadius_ > 0)
+            boxFilter_ = gpu::createBoxFilter(CV_8UC1, -1, Size(smoothingRadius_, smoothingRadius_));
+#endif
+
+        loadConstants(frameSize_.width, frameSize_.height, minVal_, maxVal_,
+                      quantizationLevels_, backgroundPrior_, decisionThreshold_, maxFeatures_, numInitializationFrames_);
+    }
+}
+
+Ptr<gpu::BackgroundSubtractorGMG> cv::gpu::createBackgroundSubtractorGMG(int initializationFrames, double decisionThreshold)
+{
+    return makePtr<GMGImpl>(initializationFrames, decisionThreshold);
+}
+
+#endif
diff --git a/modules/gpubgsegm/src/mog.cpp b/modules/gpubgsegm/src/mog.cpp
new file mode 100644
index 0000000000..12e5b32304
--- /dev/null
+++ b/modules/gpubgsegm/src/mog.cpp
@@ -0,0 +1,209 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+Ptr<gpu::BackgroundSubtractorMOG> cv::gpu::createBackgroundSubtractorMOG(int, int, double, double)  { throw_no_cuda(); return Ptr<gpu::BackgroundSubtractorMOG>(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace mog
+    {
+        void mog_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzf weight, PtrStepSzf sortKey, PtrStepSzb mean, PtrStepSzb var,
+                     int nmixtures, float varThreshold, float learningRate, float backgroundRatio, float noiseSigma,
+                     cudaStream_t stream);
+        void getBackgroundImage_gpu(int cn, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, int nmixtures, float backgroundRatio, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    const int defaultNMixtures = 5;
+    const int defaultHistory = 200;
+    const float defaultBackgroundRatio = 0.7f;
+    const float defaultVarThreshold = 2.5f * 2.5f;
+    const float defaultNoiseSigma = 30.0f * 0.5f;
+    const float defaultInitialWeight = 0.05f;
+
+    class MOGImpl : public gpu::BackgroundSubtractorMOG
+    {
+    public:
+        MOGImpl(int history, int nmixtures, double backgroundRatio, double noiseSigma);
+
+        void apply(InputArray image, OutputArray fgmask, double learningRate=-1);
+        void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream);
+
+        void getBackgroundImage(OutputArray backgroundImage) const;
+        void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const;
+
+        int getHistory() const { return history_; }
+        void setHistory(int nframes) { history_ = nframes; }
+
+        int getNMixtures() const { return nmixtures_; }
+        void setNMixtures(int nmix) { nmixtures_ = nmix; }
+
+        double getBackgroundRatio() const { return backgroundRatio_; }
+        void setBackgroundRatio(double backgroundRatio) { backgroundRatio_ = (float) backgroundRatio; }
+
+        double getNoiseSigma() const { return noiseSigma_; }
+        void setNoiseSigma(double noiseSigma) { noiseSigma_ = (float) noiseSigma; }
+
+    private:
+        //! re-initiaization method
+        void initialize(Size frameSize, int frameType);
+
+        int history_;
+        int nmixtures_;
+        float backgroundRatio_;
+        float noiseSigma_;
+
+        float varThreshold_;
+
+        Size frameSize_;
+        int frameType_;
+        int nframes_;
+
+        GpuMat weight_;
+        GpuMat sortKey_;
+        GpuMat mean_;
+        GpuMat var_;
+    };
+
+    MOGImpl::MOGImpl(int history, int nmixtures, double backgroundRatio, double noiseSigma) :
+        frameSize_(0, 0), frameType_(0), nframes_(0)
+    {
+        history_ = history > 0 ? history : defaultHistory;
+        nmixtures_ = std::min(nmixtures > 0 ? nmixtures : defaultNMixtures, 8);
+        backgroundRatio_ = backgroundRatio > 0 ? (float) backgroundRatio : defaultBackgroundRatio;
+        noiseSigma_ = noiseSigma > 0 ? (float) noiseSigma : defaultNoiseSigma;
+
+        varThreshold_ = defaultVarThreshold;
+    }
+
+    void MOGImpl::apply(InputArray image, OutputArray fgmask, double learningRate)
+    {
+        apply(image, fgmask, learningRate, Stream::Null());
+    }
+
+    void MOGImpl::apply(InputArray _frame, OutputArray _fgmask, double learningRate, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::mog;
+
+        GpuMat frame = _frame.getGpuMat();
+
+        CV_Assert( frame.depth() == CV_8U );
+
+        int ch = frame.channels();
+        int work_ch = ch;
+
+        if (nframes_ == 0 || learningRate >= 1.0 || frame.size() != frameSize_ || work_ch != mean_.channels())
+            initialize(frame.size(), frame.type());
+
+        _fgmask.create(frameSize_, CV_8UC1);
+        GpuMat fgmask = _fgmask.getGpuMat();
+
+        ++nframes_;
+        learningRate = learningRate >= 0 && nframes_ > 1 ? learningRate : 1.0 / std::min(nframes_, history_);
+        CV_Assert( learningRate >= 0 );
+
+        mog_gpu(frame, ch, fgmask, weight_, sortKey_, mean_, var_, nmixtures_,
+                varThreshold_, (float) learningRate, backgroundRatio_, noiseSigma_,
+                StreamAccessor::getStream(stream));
+    }
+
+    void MOGImpl::getBackgroundImage(OutputArray backgroundImage) const
+    {
+        getBackgroundImage(backgroundImage, Stream::Null());
+    }
+
+    void MOGImpl::getBackgroundImage(OutputArray _backgroundImage, Stream& stream) const
+    {
+        using namespace cv::gpu::cudev::mog;
+
+        _backgroundImage.create(frameSize_, frameType_);
+        GpuMat backgroundImage = _backgroundImage.getGpuMat();
+
+        getBackgroundImage_gpu(backgroundImage.channels(), weight_, mean_, backgroundImage, nmixtures_, backgroundRatio_, StreamAccessor::getStream(stream));
+    }
+
+    void MOGImpl::initialize(Size frameSize, int frameType)
+    {
+        CV_Assert( frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4 );
+
+        frameSize_ = frameSize;
+        frameType_ = frameType;
+
+        int ch = CV_MAT_CN(frameType);
+        int work_ch = ch;
+
+        // for each gaussian mixture of each pixel bg model we store
+        // the mixture sort key (w/sum_of_variances), the mixture weight (w),
+        // the mean (nchannels values) and
+        // the diagonal covariance matrix (another nchannels values)
+
+        weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+        sortKey_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+        mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
+        var_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
+
+        weight_.setTo(cv::Scalar::all(0));
+        sortKey_.setTo(cv::Scalar::all(0));
+        mean_.setTo(cv::Scalar::all(0));
+        var_.setTo(cv::Scalar::all(0));
+
+        nframes_ = 0;
+    }
+}
+
+Ptr<gpu::BackgroundSubtractorMOG> cv::gpu::createBackgroundSubtractorMOG(int history, int nmixtures, double backgroundRatio, double noiseSigma)
+{
+    return makePtr<MOGImpl>(history, nmixtures, backgroundRatio, noiseSigma);
+}
+
+#endif
diff --git a/modules/gpubgsegm/src/mog2.cpp b/modules/gpubgsegm/src/mog2.cpp
new file mode 100644
index 0000000000..5bfb5a1c4f
--- /dev/null
+++ b/modules/gpubgsegm/src/mog2.cpp
@@ -0,0 +1,253 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+Ptr<gpu::BackgroundSubtractorMOG2> cv::gpu::createBackgroundSubtractorMOG2(int, double, bool) { throw_no_cuda(); return Ptr<gpu::BackgroundSubtractorMOG2>(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace mog2
+    {
+        void loadConstants(int nmixtures, float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau, unsigned char shadowVal);
+        void mog2_gpu(PtrStepSzb frame, int cn, PtrStepSzb fgmask, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzf variance, PtrStepSzb mean, float alphaT, float prune, bool detectShadows, cudaStream_t stream);
+        void getBackgroundImage2_gpu(int cn, PtrStepSzb modesUsed, PtrStepSzf weight, PtrStepSzb mean, PtrStepSzb dst, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    // default parameters of gaussian background detection algorithm
+    const int defaultHistory = 500; // Learning rate; alpha = 1/defaultHistory2
+    const float defaultVarThreshold = 4.0f * 4.0f;
+    const int defaultNMixtures = 5; // maximal number of Gaussians in mixture
+    const float defaultBackgroundRatio = 0.9f; // threshold sum of weights for background test
+    const float defaultVarThresholdGen = 3.0f * 3.0f;
+    const float defaultVarInit = 15.0f; // initial variance for new components
+    const float defaultVarMax = 5.0f * defaultVarInit;
+    const float defaultVarMin = 4.0f;
+
+    // additional parameters
+    const float defaultCT = 0.05f; // complexity reduction prior constant 0 - no reduction of number of components
+    const unsigned char defaultShadowValue = 127; // value to use in the segmentation mask for shadows, set 0 not to do shadow detection
+    const float defaultShadowThreshold = 0.5f; // Tau - shadow threshold, see the paper for explanation
+
+    class MOG2Impl : public gpu::BackgroundSubtractorMOG2
+    {
+    public:
+        MOG2Impl(int history, double varThreshold, bool detectShadows);
+
+        void apply(InputArray image, OutputArray fgmask, double learningRate=-1);
+        void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream);
+
+        void getBackgroundImage(OutputArray backgroundImage) const;
+        void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const;
+
+        int getHistory() const { return history_; }
+        void setHistory(int history) { history_ = history; }
+
+        int getNMixtures() const { return nmixtures_; }
+        void setNMixtures(int nmixtures) { nmixtures_ = nmixtures; }
+
+        double getBackgroundRatio() const { return backgroundRatio_; }
+        void setBackgroundRatio(double ratio) { backgroundRatio_ = (float) ratio; }
+
+        double getVarThreshold() const { return varThreshold_; }
+        void setVarThreshold(double varThreshold) { varThreshold_ = (float) varThreshold; }
+
+        double getVarThresholdGen() const { return varThresholdGen_; }
+        void setVarThresholdGen(double varThresholdGen) { varThresholdGen_ = (float) varThresholdGen; }
+
+        double getVarInit() const { return varInit_; }
+        void setVarInit(double varInit) { varInit_ = (float) varInit; }
+
+        double getVarMin() const { return varMin_; }
+        void setVarMin(double varMin) { varMin_ = (float) varMin; }
+
+        double getVarMax() const { return varMax_; }
+        void setVarMax(double varMax) { varMax_ = (float) varMax; }
+
+        double getComplexityReductionThreshold() const { return ct_; }
+        void setComplexityReductionThreshold(double ct) { ct_ = (float) ct; }
+
+        bool getDetectShadows() const { return detectShadows_; }
+        void setDetectShadows(bool detectShadows) { detectShadows_ = detectShadows; }
+
+        int getShadowValue() const { return shadowValue_; }
+        void setShadowValue(int value) { shadowValue_ = (uchar) value; }
+
+        double getShadowThreshold() const { return shadowThreshold_; }
+        void setShadowThreshold(double threshold) { shadowThreshold_ = (float) threshold; }
+
+    private:
+        void initialize(Size frameSize, int frameType);
+
+        int history_;
+        int nmixtures_;
+        float backgroundRatio_;
+        float varThreshold_;
+        float varThresholdGen_;
+        float varInit_;
+        float varMin_;
+        float varMax_;
+        float ct_;
+        bool detectShadows_;
+        uchar shadowValue_;
+        float shadowThreshold_;
+
+        Size frameSize_;
+        int frameType_;
+        int nframes_;
+
+        GpuMat weight_;
+        GpuMat variance_;
+        GpuMat mean_;
+
+        //keep track of number of modes per pixel
+        GpuMat bgmodelUsedModes_;
+    };
+
+    MOG2Impl::MOG2Impl(int history, double varThreshold, bool detectShadows) :
+        frameSize_(0, 0), frameType_(0), nframes_(0)
+    {
+        history_ = history > 0 ? history : defaultHistory;
+        varThreshold_ = varThreshold > 0 ? (float) varThreshold : defaultVarThreshold;
+        detectShadows_ = detectShadows;
+
+        nmixtures_ = defaultNMixtures;
+        backgroundRatio_ = defaultBackgroundRatio;
+        varInit_ = defaultVarInit;
+        varMax_ = defaultVarMax;
+        varMin_ = defaultVarMin;
+        varThresholdGen_ = defaultVarThresholdGen;
+        ct_ = defaultCT;
+        shadowValue_ =  defaultShadowValue;
+        shadowThreshold_ = defaultShadowThreshold;
+    }
+
+    void MOG2Impl::apply(InputArray image, OutputArray fgmask, double learningRate)
+    {
+        apply(image, fgmask, learningRate, Stream::Null());
+    }
+
+    void MOG2Impl::apply(InputArray _frame, OutputArray _fgmask, double learningRate, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::mog2;
+
+        GpuMat frame = _frame.getGpuMat();
+
+        int ch = frame.channels();
+        int work_ch = ch;
+
+        if (nframes_ == 0 || learningRate >= 1.0 || frame.size() != frameSize_ || work_ch != mean_.channels())
+            initialize(frame.size(), frame.type());
+
+        _fgmask.create(frameSize_, CV_8UC1);
+        GpuMat fgmask = _fgmask.getGpuMat();
+
+        fgmask.setTo(Scalar::all(0), stream);
+
+        ++nframes_;
+        learningRate = learningRate >= 0 && nframes_ > 1 ? learningRate : 1.0 / std::min(2 * nframes_, history_);
+        CV_Assert( learningRate >= 0 );
+
+        mog2_gpu(frame, frame.channels(), fgmask, bgmodelUsedModes_, weight_, variance_, mean_,
+                 (float) learningRate, static_cast<float>(-learningRate * ct_), detectShadows_, StreamAccessor::getStream(stream));
+    }
+
+    void MOG2Impl::getBackgroundImage(OutputArray backgroundImage) const
+    {
+        getBackgroundImage(backgroundImage, Stream::Null());
+    }
+
+    void MOG2Impl::getBackgroundImage(OutputArray _backgroundImage, Stream& stream) const
+    {
+        using namespace cv::gpu::cudev::mog2;
+
+        _backgroundImage.create(frameSize_, frameType_);
+        GpuMat backgroundImage = _backgroundImage.getGpuMat();
+
+        getBackgroundImage2_gpu(backgroundImage.channels(), bgmodelUsedModes_, weight_, mean_, backgroundImage, StreamAccessor::getStream(stream));
+    }
+
+    void MOG2Impl::initialize(cv::Size frameSize, int frameType)
+    {
+        using namespace cv::gpu::cudev::mog2;
+
+        CV_Assert( frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4 );
+
+        frameSize_ = frameSize;
+        frameType_ = frameType;
+        nframes_ = 0;
+
+        int ch = CV_MAT_CN(frameType);
+        int work_ch = ch;
+
+        // for each gaussian mixture of each pixel bg model we store ...
+        // the mixture weight (w),
+        // the mean (nchannels values) and
+        // the covariance
+        weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+        variance_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+        mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
+
+        //make the array for keeping track of the used modes per pixel - all zeros at start
+        bgmodelUsedModes_.create(frameSize_, CV_8UC1);
+        bgmodelUsedModes_.setTo(Scalar::all(0));
+
+        loadConstants(nmixtures_, varThreshold_, backgroundRatio_, varThresholdGen_, varInit_, varMin_, varMax_, shadowThreshold_, shadowValue_);
+    }
+}
+
+Ptr<gpu::BackgroundSubtractorMOG2> cv::gpu::createBackgroundSubtractorMOG2(int history, double varThreshold, bool detectShadows)
+{
+    return makePtr<MOG2Impl>(history, varThreshold, detectShadows);
+}
+
+#endif
diff --git a/modules/gpubgsegm/src/precomp.hpp b/modules/gpubgsegm/src/precomp.hpp
new file mode 100644
index 0000000000..c61cc08bbe
--- /dev/null
+++ b/modules/gpubgsegm/src/precomp.hpp
@@ -0,0 +1,66 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+
+#include "opencv2/gpubgsegm.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+#  include "opencv2/gpufilters.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUIMGPROC
+#  include "opencv2/gpuimgproc.hpp"
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/test/test_bgfg.cpp b/modules/gpubgsegm/test/test_bgsegm.cpp
similarity index 68%
rename from modules/gpu/test/test_bgfg.cpp
rename to modules/gpubgsegm/test/test_bgsegm.cpp
index ebf0a88afa..3a287d1bb8 100644
--- a/modules/gpu/test/test_bgfg.cpp
+++ b/modules/gpubgsegm/test/test_bgsegm.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -41,24 +42,44 @@
 
 #include "test_precomp.hpp"
 
+#ifdef HAVE_OPENCV_LEGACY
+#  include "opencv2/legacy.hpp"
+#endif
+
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
+#if defined(HAVE_XINE)         || \
+    defined(HAVE_GSTREAMER)    || \
+    defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
+    defined(HAVE_AVFOUNDATION) || \
+    defined(HAVE_FFMPEG)       || \
+    defined(WIN32) /* assume that we have ffmpeg */
+
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
+#else
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 0
+#endif
+
 //////////////////////////////////////////////////////
 // FGDStatModel
 
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT && defined(HAVE_OPENCV_LEGACY)
+
 namespace cv
 {
-    template<> void Ptr<CvBGStatModel>::delete_obj()
+    template<> void DefaultDeleter<CvBGStatModel>::operator ()(CvBGStatModel* obj) const
     {
         cvReleaseBGStatModel(&obj);
     }
 }
 
-PARAM_TEST_CASE(FGDStatModel, cv::gpu::DeviceInfo, std::string, Channels)
+PARAM_TEST_CASE(FGDStatModel, cv::gpu::DeviceInfo, std::string)
 {
     cv::gpu::DeviceInfo devInfo;
     std::string inputFile;
-    int out_cn;
 
     virtual void SetUp()
     {
@@ -66,8 +87,6 @@ PARAM_TEST_CASE(FGDStatModel, cv::gpu::DeviceInfo, std::string, Channels)
         cv::gpu::setDevice(devInfo.deviceID());
 
         inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + "video/" + GET_PARAM(1);
-
-        out_cn = GET_PARAM(2);
     }
 };
 
@@ -84,15 +103,10 @@ GPU_TEST_P(FGDStatModel, Update)
     cv::Ptr<CvBGStatModel> model(cvCreateFGDStatModel(&ipl_frame));
 
     cv::gpu::GpuMat d_frame(frame);
-    cv::gpu::FGDStatModel d_model(out_cn);
-    d_model.create(d_frame);
-
-    cv::Mat h_background;
-    cv::Mat h_foreground;
-    cv::Mat h_background3;
-
-    cv::Mat backgroundDiff;
-    cv::Mat foregroundDiff;
+    cv::Ptr<cv::gpu::BackgroundSubtractorFGD> d_fgd = cv::gpu::createBackgroundSubtractorFGD();
+    cv::gpu::GpuMat d_foreground, d_background;
+    std::vector< std::vector<cv::Point> > foreground_regions;
+    d_fgd->apply(d_frame, d_foreground);
 
     for (int i = 0; i < 5; ++i)
     {
@@ -103,36 +117,31 @@ GPU_TEST_P(FGDStatModel, Update)
         int gold_count = cvUpdateBGStatModel(&ipl_frame, model);
 
         d_frame.upload(frame);
+        d_fgd->apply(d_frame, d_foreground);
+        d_fgd->getBackgroundImage(d_background);
+        d_fgd->getForegroundRegions(foreground_regions);
+        int count = (int) foreground_regions.size();
 
-        int count = d_model.update(d_frame);
+        cv::Mat gold_background = cv::cvarrToMat(model->background);
+        cv::Mat gold_foreground = cv::cvarrToMat(model->foreground);
 
+        ASSERT_MAT_NEAR(gold_background, d_background, 1.0);
+        ASSERT_MAT_NEAR(gold_foreground, d_foreground, 0.0);
         ASSERT_EQ(gold_count, count);
-
-        cv::Mat gold_background(model->background);
-        cv::Mat gold_foreground(model->foreground);
-
-        if (out_cn == 3)
-            d_model.background.download(h_background3);
-        else
-        {
-            d_model.background.download(h_background);
-            cv::cvtColor(h_background, h_background3, cv::COLOR_BGRA2BGR);
-        }
-        d_model.foreground.download(h_foreground);
-
-        ASSERT_MAT_NEAR(gold_background, h_background3, 1.0);
-        ASSERT_MAT_NEAR(gold_foreground, h_foreground, 0.0);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, FGDStatModel, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_BgSegm, FGDStatModel, testing::Combine(
     ALL_DEVICES,
-    testing::Values(std::string("768x576.avi")),
-    testing::Values(Channels(3), Channels(4))));
+    testing::Values(std::string("768x576.avi"))));
+
+#endif
 
 //////////////////////////////////////////////////////
 // MOG
 
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
 namespace
 {
     IMPLEMENT_PARAM_CLASS(UseGray, bool)
@@ -171,10 +180,10 @@ GPU_TEST_P(MOG, Update)
     cap >> frame;
     ASSERT_FALSE(frame.empty());
 
-    cv::gpu::MOG_GPU mog;
+    cv::Ptr<cv::BackgroundSubtractorMOG> mog = cv::gpu::createBackgroundSubtractorMOG();
     cv::gpu::GpuMat foreground = createMat(frame.size(), CV_8UC1, useRoi);
 
-    cv::BackgroundSubtractorMOG mog_gold;
+    cv::Ptr<cv::BackgroundSubtractorMOG> mog_gold = cv::createBackgroundSubtractorMOG();
     cv::Mat foreground_gold;
 
     for (int i = 0; i < 10; ++i)
@@ -189,24 +198,28 @@ GPU_TEST_P(MOG, Update)
             cv::swap(temp, frame);
         }
 
-        mog(loadMat(frame, useRoi), foreground, (float)learningRate);
+        mog->apply(loadMat(frame, useRoi), foreground, learningRate);
 
-        mog_gold(frame, foreground_gold, learningRate);
+        mog_gold->apply(frame, foreground_gold, learningRate);
 
         ASSERT_MAT_NEAR(foreground_gold, foreground, 0.0);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, MOG, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_BgSegm, MOG, testing::Combine(
     ALL_DEVICES,
     testing::Values(std::string("768x576.avi")),
     testing::Values(UseGray(true), UseGray(false)),
     testing::Values(LearningRate(0.0), LearningRate(0.01)),
     WHOLE_SUBMAT));
 
+#endif
+
 //////////////////////////////////////////////////////
 // MOG2
 
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
 namespace
 {
     IMPLEMENT_PARAM_CLASS(DetectShadow, bool)
@@ -241,12 +254,12 @@ GPU_TEST_P(MOG2, Update)
     cap >> frame;
     ASSERT_FALSE(frame.empty());
 
-    cv::gpu::MOG2_GPU mog2;
-    mog2.bShadowDetection = detectShadow;
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2 = cv::gpu::createBackgroundSubtractorMOG2();
+    mog2->setDetectShadows(detectShadow);
     cv::gpu::GpuMat foreground = createMat(frame.size(), CV_8UC1, useRoi);
 
-    cv::BackgroundSubtractorMOG2 mog2_gold;
-    mog2_gold.set("detectShadows", detectShadow);
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2_gold = cv::createBackgroundSubtractorMOG2();
+    mog2_gold->setDetectShadows(detectShadow);
     cv::Mat foreground_gold;
 
     for (int i = 0; i < 10; ++i)
@@ -261,9 +274,9 @@ GPU_TEST_P(MOG2, Update)
             cv::swap(temp, frame);
         }
 
-        mog2(loadMat(frame, useRoi), foreground);
+        mog2->apply(loadMat(frame, useRoi), foreground);
 
-        mog2_gold(frame, foreground_gold);
+        mog2_gold->apply(frame, foreground_gold);
 
         if (detectShadow)
         {
@@ -286,12 +299,12 @@ GPU_TEST_P(MOG2, getBackgroundImage)
 
     cv::Mat frame;
 
-    cv::gpu::MOG2_GPU mog2;
-    mog2.bShadowDetection = detectShadow;
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2 = cv::gpu::createBackgroundSubtractorMOG2();
+    mog2->setDetectShadows(detectShadow);
     cv::gpu::GpuMat foreground;
 
-    cv::BackgroundSubtractorMOG2 mog2_gold;
-    mog2_gold.set("detectShadows", detectShadow);
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2_gold = cv::createBackgroundSubtractorMOG2();
+    mog2_gold->setDetectShadows(detectShadow);
     cv::Mat foreground_gold;
 
     for (int i = 0; i < 10; ++i)
@@ -299,67 +312,28 @@ GPU_TEST_P(MOG2, getBackgroundImage)
         cap >> frame;
         ASSERT_FALSE(frame.empty());
 
-        mog2(loadMat(frame, useRoi), foreground);
+        mog2->apply(loadMat(frame, useRoi), foreground);
 
-        mog2_gold(frame, foreground_gold);
+        mog2_gold->apply(frame, foreground_gold);
     }
 
     cv::gpu::GpuMat background = createMat(frame.size(), frame.type(), useRoi);
-    mog2.getBackgroundImage(background);
+    mog2->getBackgroundImage(background);
 
     cv::Mat background_gold;
-    mog2_gold.getBackgroundImage(background_gold);
+    mog2_gold->getBackgroundImage(background_gold);
 
     ASSERT_MAT_NEAR(background_gold, background, 0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, MOG2, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_BgSegm, MOG2, testing::Combine(
     ALL_DEVICES,
     testing::Values(std::string("768x576.avi")),
     testing::Values(UseGray(true), UseGray(false)),
     testing::Values(DetectShadow(true), DetectShadow(false)),
     WHOLE_SUBMAT));
 
-//////////////////////////////////////////////////////
-// VIBE
-
-PARAM_TEST_CASE(VIBE, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
-{
-};
-
-GPU_TEST_P(VIBE, Accuracy)
-{
-    const cv::gpu::DeviceInfo devInfo = GET_PARAM(0);
-    cv::gpu::setDevice(devInfo.deviceID());
-    const cv::Size size = GET_PARAM(1);
-    const int type = GET_PARAM(2);
-    const bool useRoi = GET_PARAM(3);
-
-    const cv::Mat fullfg(size, CV_8UC1, cv::Scalar::all(255));
-
-    cv::Mat frame = randomMat(size, type, 0.0, 100);
-    cv::gpu::GpuMat d_frame = loadMat(frame, useRoi);
-
-    cv::gpu::VIBE_GPU vibe;
-    cv::gpu::GpuMat d_fgmask = createMat(size, CV_8UC1, useRoi);
-    vibe.initialize(d_frame);
-
-    for (int i = 0; i < 20; ++i)
-        vibe(d_frame, d_fgmask);
-
-    frame = randomMat(size, type, 160, 255);
-    d_frame = loadMat(frame, useRoi);
-    vibe(d_frame, d_fgmask);
-
-    // now fgmask should be entirely foreground
-    ASSERT_MAT_NEAR(fullfg, d_fgmask, 0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Video, VIBE, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4)),
-    WHOLE_SUBMAT));
+#endif
 
 //////////////////////////////////////////////////////
 // GMG
@@ -385,16 +359,15 @@ GPU_TEST_P(GMG, Accuracy)
     cv::Mat frame = randomMat(size, type, 0, 100);
     cv::gpu::GpuMat d_frame = loadMat(frame, useRoi);
 
-    cv::gpu::GMG_GPU gmg;
-    gmg.numInitializationFrames = 5;
-    gmg.smoothingRadius = 0;
-    gmg.initialize(d_frame.size(), 0, 255);
+    cv::Ptr<cv::BackgroundSubtractorGMG> gmg = cv::gpu::createBackgroundSubtractorGMG();
+    gmg->setNumFrames(5);
+    gmg->setSmoothingRadius(0);
 
     cv::gpu::GpuMat d_fgmask = createMat(size, CV_8UC1, useRoi);
 
-    for (int i = 0; i < gmg.numInitializationFrames; ++i)
+    for (int i = 0; i < gmg->getNumFrames(); ++i)
     {
-        gmg(d_frame, d_fgmask);
+        gmg->apply(d_frame, d_fgmask);
 
         // fgmask should be entirely background during training
         ASSERT_MAT_NEAR(zeros, d_fgmask, 0);
@@ -402,13 +375,13 @@ GPU_TEST_P(GMG, Accuracy)
 
     frame = randomMat(size, type, 160, 255);
     d_frame = loadMat(frame, useRoi);
-    gmg(d_frame, d_fgmask);
+    gmg->apply(d_frame, d_fgmask);
 
     // now fgmask should be entirely foreground
     ASSERT_MAT_NEAR(fullfg, d_fgmask, 0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, GMG, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_BgSegm, GMG, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8U), MatType(CV_16U), MatType(CV_32F)),
diff --git a/modules/stitching/src/precomp.cpp b/modules/gpubgsegm/test/test_main.cpp
similarity index 95%
rename from modules/stitching/src/precomp.cpp
rename to modules/gpubgsegm/test/test_main.cpp
index 390dbfbc6b..eea3d7c008 100644
--- a/modules/stitching/src/precomp.cpp
+++ b/modules/gpubgsegm/test/test_main.cpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -40,4 +40,6 @@
 //
 //M*/
 
-#include "precomp.hpp"
\ No newline at end of file
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpubgsegm/test/test_precomp.hpp b/modules/gpubgsegm/test/test_precomp.hpp
new file mode 100644
index 0000000000..783e9e64cd
--- /dev/null
+++ b/modules/gpubgsegm/test/test_precomp.hpp
@@ -0,0 +1,65 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include <fstream>
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpubgsegm.hpp"
+#include "opencv2/video.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpucodec/CMakeLists.txt b/modules/gpucodec/CMakeLists.txt
new file mode 100644
index 0000000000..7c3be5c0e0
--- /dev/null
+++ b/modules/gpucodec/CMakeLists.txt
@@ -0,0 +1,29 @@
+if(ANDROID OR IOS OR APPLE)
+  ocv_module_disable(gpucodec)
+endif()
+
+set(the_description "GPU-accelerated Video Encoding/Decoding")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_add_module(gpucodec opencv_highgui)
+
+ocv_module_include_directories()
+ocv_glob_module_sources()
+
+set(extra_libs "")
+
+if(HAVE_NVCUVID)
+  list(APPEND extra_libs ${CUDA_CUDA_LIBRARY} ${CUDA_nvcuvid_LIBRARY})
+
+  if(WIN32)
+    list(APPEND extra_libs ${CUDA_nvcuvenc_LIBRARY})
+  endif()
+endif()
+
+ocv_create_module(${extra_libs})
+
+ocv_add_precompiled_headers(${the_module})
+
+ocv_add_accuracy_tests()
+ocv_add_perf_tests()
diff --git a/modules/gpucodec/doc/gpucodec.rst b/modules/gpucodec/doc/gpucodec.rst
new file mode 100644
index 0000000000..b9f763f84e
--- /dev/null
+++ b/modules/gpucodec/doc/gpucodec.rst
@@ -0,0 +1,9 @@
+*************************************************
+gpucodec. GPU-accelerated Video Encoding/Decoding
+*************************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    videodec
+    videoenc
diff --git a/modules/gpucodec/doc/videodec.rst b/modules/gpucodec/doc/videodec.rst
new file mode 100644
index 0000000000..dbf8951210
--- /dev/null
+++ b/modules/gpucodec/doc/videodec.rst
@@ -0,0 +1,156 @@
+Video Decoding
+==============
+
+.. highlight:: cpp
+
+
+
+gpucodec::VideoReader
+---------------------
+Video reader interface.
+
+.. ocv:class:: gpucodec::VideoReader
+
+.. note::
+
+   * An example on how to use the videoReader class can be found at opencv_source_code/samples/gpu/video_reader.cpp
+
+
+gpucodec::VideoReader::nextFrame
+--------------------------------
+Grabs, decodes and returns the next video frame.
+
+.. ocv:function:: bool gpucodec::VideoReader::nextFrame(OutputArray frame)
+
+If no frames has been grabbed (there are no more frames in video file), the methods return ``false`` . The method throws :ocv:class:`Exception` if error occurs.
+
+
+
+gpucodec::VideoReader::format
+-----------------------------
+Returns information about video file format.
+
+.. ocv:function:: FormatInfo gpucodec::VideoReader::format() const
+
+
+
+gpucodec::Codec
+---------------
+Video codecs supported by :ocv:class:`gpucodec::VideoReader` .
+
+.. ocv:enum:: gpucodec::Codec
+
+  .. ocv:emember:: MPEG1 = 0
+  .. ocv:emember:: MPEG2
+  .. ocv:emember:: MPEG4
+  .. ocv:emember:: VC1
+  .. ocv:emember:: H264
+  .. ocv:emember:: JPEG
+  .. ocv:emember:: H264_SVC
+  .. ocv:emember:: H264_MVC
+
+  .. ocv:emember:: Uncompressed_YUV420 = (('I'<<24)|('Y'<<16)|('U'<<8)|('V'))
+
+        Y,U,V (4:2:0)
+
+  .. ocv:emember:: Uncompressed_YV12   = (('Y'<<24)|('V'<<16)|('1'<<8)|('2'))
+
+        Y,V,U (4:2:0)
+
+  .. ocv:emember:: Uncompressed_NV12   = (('N'<<24)|('V'<<16)|('1'<<8)|('2'))
+
+        Y,UV  (4:2:0)
+
+  .. ocv:emember:: Uncompressed_YUYV   = (('Y'<<24)|('U'<<16)|('Y'<<8)|('V'))
+
+        YUYV/YUY2 (4:2:2)
+
+  .. ocv:emember:: Uncompressed_UYVY   = (('U'<<24)|('Y'<<16)|('V'<<8)|('Y'))
+
+        UYVY (4:2:2)
+
+
+
+gpucodec::ChromaFormat
+----------------------
+Chroma formats supported by :ocv:class:`gpucodec::VideoReader` .
+
+.. ocv:enum:: gpucodec::ChromaFormat
+
+  .. ocv:emember:: Monochrome = 0
+  .. ocv:emember:: YUV420
+  .. ocv:emember:: YUV422
+  .. ocv:emember:: YUV444
+
+
+
+gpucodec::FormatInfo
+--------------------
+.. ocv:struct:: gpucodec::FormatInfo
+
+Struct providing information about video file format. ::
+
+    struct FormatInfo
+    {
+        Codec codec;
+        ChromaFormat chromaFormat;
+        int width;
+        int height;
+    };
+
+
+
+gpucodec::createVideoReader
+---------------------------
+Creates video reader.
+
+.. ocv:function:: Ptr<VideoReader> gpucodec::createVideoReader(const String& filename)
+.. ocv:function:: Ptr<VideoReader> gpucodec::createVideoReader(const Ptr<RawVideoSource>& source)
+
+    :param filename: Name of the input video file.
+
+    :param source: RAW video source implemented by user.
+
+FFMPEG is used to read videos. User can implement own demultiplexing with :ocv:class:`gpucodec::RawVideoSource` .
+
+
+
+gpucodec::RawVideoSource
+------------------------
+.. ocv:class:: gpucodec::RawVideoSource
+
+Interface for video demultiplexing. ::
+
+    class RawVideoSource
+    {
+    public:
+        virtual ~RawVideoSource() {}
+
+        virtual bool getNextPacket(unsigned char** data, int* size, bool* endOfFile) = 0;
+
+        virtual FormatInfo format() const = 0;
+    };
+
+User can implement own demultiplexing by implementing this interface.
+
+
+
+gpucodec::RawVideoSource::getNextPacket
+---------------------------------------
+Returns next packet with RAW video frame.
+
+.. ocv:function:: bool gpucodec::VideoSource::getNextPacket(unsigned char** data, int* size, bool* endOfFile) = 0
+
+    :param data: Pointer to frame data.
+
+    :param size: Size in bytes of current frame.
+
+    :param endOfStream: Indicates that it is end of stream.
+
+
+
+gpucodec::RawVideoSource::format
+--------------------------------
+Returns information about video file format.
+
+.. ocv:function:: virtual FormatInfo gpucodec::RawVideoSource::format() const = 0
diff --git a/modules/gpucodec/doc/videoenc.rst b/modules/gpucodec/doc/videoenc.rst
new file mode 100644
index 0000000000..d4a6814a0f
--- /dev/null
+++ b/modules/gpucodec/doc/videoenc.rst
@@ -0,0 +1,192 @@
+Video Encoding
+==============
+
+.. highlight:: cpp
+
+
+
+gpucodec::VideoWriter
+---------------------
+Video writer interface.
+
+.. ocv:class:: gpucodec::VideoWriter
+
+The implementation uses H264 video codec.
+
+.. note:: Currently only Windows platform is supported.
+
+.. note::
+
+   * An example on how to use the videoWriter class can be found at opencv_source_code/samples/gpu/video_writer.cpp
+
+
+gpucodec::VideoWriter::write
+----------------------------
+Writes the next video frame.
+
+.. ocv:function:: void gpucodec::VideoWriter::write(InputArray frame, bool lastFrame = false) = 0
+
+    :param frame: The written frame.
+
+    :param lastFrame: Indicates that it is end of stream. The parameter can be ignored.
+
+The method write the specified image to video file. The image must have the same size and the same surface format as has been specified when opening the video writer.
+
+
+
+gpucodec::createVideoWriter
+---------------------------
+Creates video writer.
+
+.. ocv:function:: Ptr<gpucodec::VideoWriter> gpucodec::createVideoWriter(const String& fileName, Size frameSize, double fps, SurfaceFormat format = SF_BGR)
+.. ocv:function:: Ptr<gpucodec::VideoWriter> gpucodec::createVideoWriter(const String& fileName, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
+.. ocv:function:: Ptr<gpucodec::VideoWriter> gpucodec::createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, SurfaceFormat format = SF_BGR)
+.. ocv:function:: Ptr<gpucodec::VideoWriter> gpucodec::createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR)
+
+    :param fileName: Name of the output video file. Only AVI file format is supported.
+
+    :param frameSize: Size of the input video frames.
+
+    :param fps: Framerate of the created video stream.
+
+    :param params: Encoder parameters. See :ocv:struct:`gpucodec::EncoderParams` .
+
+    :param format: Surface format of input frames ( ``SF_UYVY`` , ``SF_YUY2`` , ``SF_YV12`` , ``SF_NV12`` , ``SF_IYUV`` , ``SF_BGR`` or ``SF_GRAY``). BGR or gray frames will be converted to YV12 format before encoding, frames with other formats will be used as is.
+
+    :param encoderCallback: Callbacks for video encoder. See :ocv:class:`gpucodec::EncoderCallBack` . Use it if you want to work with raw video stream.
+
+The constructors initialize video writer. FFMPEG is used to write videos. User can implement own multiplexing with :ocv:class:`gpucodec::EncoderCallBack` .
+
+
+
+gpucodec::EncoderParams
+-----------------------
+.. ocv:struct:: gpucodec::EncoderParams
+
+Different parameters for CUDA video encoder. ::
+
+    struct EncoderParams
+    {
+        int       P_Interval;      //    NVVE_P_INTERVAL,
+        int       IDR_Period;      //    NVVE_IDR_PERIOD,
+        int       DynamicGOP;      //    NVVE_DYNAMIC_GOP,
+        int       RCType;          //    NVVE_RC_TYPE,
+        int       AvgBitrate;      //    NVVE_AVG_BITRATE,
+        int       PeakBitrate;     //    NVVE_PEAK_BITRATE,
+        int       QP_Level_Intra;  //    NVVE_QP_LEVEL_INTRA,
+        int       QP_Level_InterP; //    NVVE_QP_LEVEL_INTER_P,
+        int       QP_Level_InterB; //    NVVE_QP_LEVEL_INTER_B,
+        int       DeblockMode;     //    NVVE_DEBLOCK_MODE,
+        int       ProfileLevel;    //    NVVE_PROFILE_LEVEL,
+        int       ForceIntra;      //    NVVE_FORCE_INTRA,
+        int       ForceIDR;        //    NVVE_FORCE_IDR,
+        int       ClearStat;       //    NVVE_CLEAR_STAT,
+        int       DIMode;          //    NVVE_SET_DEINTERLACE,
+        int       Presets;         //    NVVE_PRESETS,
+        int       DisableCabac;    //    NVVE_DISABLE_CABAC,
+        int       NaluFramingType; //    NVVE_CONFIGURE_NALU_FRAMING_TYPE
+        int       DisableSPSPPS;   //    NVVE_DISABLE_SPS_PPS
+
+        EncoderParams();
+        explicit EncoderParams(const String& configFile);
+
+        void load(const String& configFile);
+        void save(const String& configFile) const;
+    };
+
+
+
+gpucodec::EncoderParams::EncoderParams
+--------------------------------------
+Constructors.
+
+.. ocv:function:: gpucodec::EncoderParams::EncoderParams()
+.. ocv:function:: gpucodec::EncoderParams::EncoderParams(const String& configFile)
+
+    :param configFile: Config file name.
+
+Creates default parameters or reads parameters from config file.
+
+
+
+gpucodec::EncoderParams::load
+-----------------------------
+Reads parameters from config file.
+
+.. ocv:function:: void gpucodec::EncoderParams::load(const String& configFile)
+
+    :param configFile: Config file name.
+
+
+
+gpucodec::EncoderParams::save
+-----------------------------
+Saves parameters to config file.
+
+.. ocv:function:: void gpucodec::EncoderParams::save(const String& configFile) const
+
+    :param configFile: Config file name.
+
+
+
+gpucodec::EncoderCallBack
+-------------------------
+.. ocv:class:: gpucodec::EncoderCallBack
+
+Callbacks for CUDA video encoder. ::
+
+    class EncoderCallBack
+    {
+    public:
+        enum PicType
+        {
+            IFRAME = 1,
+            PFRAME = 2,
+            BFRAME = 3
+        };
+
+        virtual ~EncoderCallBack() {}
+
+        virtual unsigned char* acquireBitStream(int* bufferSize) = 0;
+        virtual void releaseBitStream(unsigned char* data, int size) = 0;
+        virtual void onBeginFrame(int frameNumber, PicType picType) = 0;
+        virtual void onEndFrame(int frameNumber, PicType picType) = 0;
+    };
+
+
+
+gpucodec::EncoderCallBack::acquireBitStream
+-------------------------------------------
+Callback function to signal the start of bitstream that is to be encoded.
+
+.. ocv:function:: virtual uchar* gpucodec::EncoderCallBack::acquireBitStream(int* bufferSize) = 0
+
+Callback must allocate buffer for CUDA encoder and return pointer to it and it's size.
+
+
+
+gpucodec::EncoderCallBack::releaseBitStream
+-------------------------------------------
+Callback function to signal that the encoded bitstream is ready to be written to file.
+
+.. ocv:function:: virtual void gpucodec::EncoderCallBack::releaseBitStream(unsigned char* data, int size) = 0
+
+
+
+gpucodec::EncoderCallBack::onBeginFrame
+---------------------------------------
+Callback function to signal that the encoding operation on the frame has started.
+
+.. ocv:function:: virtual void gpucodec::EncoderCallBack::onBeginFrame(int frameNumber, PicType picType) = 0
+
+    :param picType: Specify frame type (I-Frame, P-Frame or B-Frame).
+
+
+
+gpucodec::EncoderCallBack::onEndFrame
+-------------------------------------
+Callback function signals that the encoding operation on the frame has finished.
+
+.. ocv:function:: virtual void gpucodec::EncoderCallBack::onEndFrame(int frameNumber, PicType picType) = 0
+
+    :param picType: Specify frame type (I-Frame, P-Frame or B-Frame).
diff --git a/modules/gpucodec/include/opencv2/gpucodec.hpp b/modules/gpucodec/include/opencv2/gpucodec.hpp
new file mode 100644
index 0000000000..f2e298fd70
--- /dev/null
+++ b/modules/gpucodec/include/opencv2/gpucodec.hpp
@@ -0,0 +1,206 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUCODEC_HPP__
+#define __OPENCV_GPUCODEC_HPP__
+
+#ifndef __cplusplus
+#  error gpucodec.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpucodec {
+
+////////////////////////////////// Video Encoding //////////////////////////////////
+
+// Works only under Windows.
+// Supports olny H264 video codec and AVI files.
+
+enum SurfaceFormat
+{
+    SF_UYVY = 0,
+    SF_YUY2,
+    SF_YV12,
+    SF_NV12,
+    SF_IYUV,
+    SF_BGR,
+    SF_GRAY = SF_BGR
+};
+
+struct CV_EXPORTS EncoderParams
+{
+    int P_Interval;      // NVVE_P_INTERVAL,
+    int IDR_Period;      // NVVE_IDR_PERIOD,
+    int DynamicGOP;      // NVVE_DYNAMIC_GOP,
+    int RCType;          // NVVE_RC_TYPE,
+    int AvgBitrate;      // NVVE_AVG_BITRATE,
+    int PeakBitrate;     // NVVE_PEAK_BITRATE,
+    int QP_Level_Intra;  // NVVE_QP_LEVEL_INTRA,
+    int QP_Level_InterP; // NVVE_QP_LEVEL_INTER_P,
+    int QP_Level_InterB; // NVVE_QP_LEVEL_INTER_B,
+    int DeblockMode;     // NVVE_DEBLOCK_MODE,
+    int ProfileLevel;    // NVVE_PROFILE_LEVEL,
+    int ForceIntra;      // NVVE_FORCE_INTRA,
+    int ForceIDR;        // NVVE_FORCE_IDR,
+    int ClearStat;       // NVVE_CLEAR_STAT,
+    int DIMode;          // NVVE_SET_DEINTERLACE,
+    int Presets;         // NVVE_PRESETS,
+    int DisableCabac;    // NVVE_DISABLE_CABAC,
+    int NaluFramingType; // NVVE_CONFIGURE_NALU_FRAMING_TYPE
+    int DisableSPSPPS;   // NVVE_DISABLE_SPS_PPS
+
+    EncoderParams();
+    explicit EncoderParams(const String& configFile);
+
+    void load(const String& configFile);
+    void save(const String& configFile) const;
+};
+
+class CV_EXPORTS EncoderCallBack
+{
+public:
+    enum PicType
+    {
+        IFRAME = 1,
+        PFRAME = 2,
+        BFRAME = 3
+    };
+
+    virtual ~EncoderCallBack() {}
+
+    //! callback function to signal the start of bitstream that is to be encoded
+    //! callback must allocate host buffer for CUDA encoder and return pointer to it and it's size
+    virtual uchar* acquireBitStream(int* bufferSize) = 0;
+
+    //! callback function to signal that the encoded bitstream is ready to be written to file
+    virtual void releaseBitStream(unsigned char* data, int size) = 0;
+
+    //! callback function to signal that the encoding operation on the frame has started
+    virtual void onBeginFrame(int frameNumber, PicType picType) = 0;
+
+    //! callback function signals that the encoding operation on the frame has finished
+    virtual void onEndFrame(int frameNumber, PicType picType) = 0;
+};
+
+class CV_EXPORTS VideoWriter
+{
+public:
+    virtual ~VideoWriter() {}
+
+    //! writes the next frame from GPU memory
+    virtual void write(InputArray frame, bool lastFrame = false) = 0;
+
+    virtual EncoderParams getEncoderParams() const = 0;
+};
+
+//! create VideoWriter for specified output file (only AVI file format is supported)
+CV_EXPORTS Ptr<VideoWriter> createVideoWriter(const String& fileName, Size frameSize, double fps, SurfaceFormat format = SF_BGR);
+CV_EXPORTS Ptr<VideoWriter> createVideoWriter(const String& fileName, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
+
+//! create VideoWriter for user-defined callbacks
+CV_EXPORTS Ptr<VideoWriter> createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, SurfaceFormat format = SF_BGR);
+CV_EXPORTS Ptr<VideoWriter> createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format = SF_BGR);
+
+////////////////////////////////// Video Decoding //////////////////////////////////////////
+
+enum Codec
+{
+    MPEG1 = 0,
+    MPEG2,
+    MPEG4,
+    VC1,
+    H264,
+    JPEG,
+    H264_SVC,
+    H264_MVC,
+
+    Uncompressed_YUV420 = (('I'<<24)|('Y'<<16)|('U'<<8)|('V')),   // Y,U,V (4:2:0)
+    Uncompressed_YV12   = (('Y'<<24)|('V'<<16)|('1'<<8)|('2')),   // Y,V,U (4:2:0)
+    Uncompressed_NV12   = (('N'<<24)|('V'<<16)|('1'<<8)|('2')),   // Y,UV  (4:2:0)
+    Uncompressed_YUYV   = (('Y'<<24)|('U'<<16)|('Y'<<8)|('V')),   // YUYV/YUY2 (4:2:2)
+    Uncompressed_UYVY   = (('U'<<24)|('Y'<<16)|('V'<<8)|('Y'))    // UYVY (4:2:2)
+};
+
+enum ChromaFormat
+{
+    Monochrome = 0,
+    YUV420,
+    YUV422,
+    YUV444
+};
+
+struct FormatInfo
+{
+    Codec codec;
+    ChromaFormat chromaFormat;
+    int width;
+    int height;
+};
+
+class CV_EXPORTS VideoReader
+{
+public:
+    virtual ~VideoReader() {}
+
+    virtual bool nextFrame(OutputArray frame) = 0;
+
+    virtual FormatInfo format() const = 0;
+};
+
+class CV_EXPORTS RawVideoSource
+{
+public:
+    virtual ~RawVideoSource() {}
+
+    virtual bool getNextPacket(unsigned char** data, int* size, bool* endOfFile) = 0;
+
+    virtual FormatInfo format() const = 0;
+};
+
+CV_EXPORTS Ptr<VideoReader> createVideoReader(const String& filename);
+CV_EXPORTS Ptr<VideoReader> createVideoReader(const Ptr<RawVideoSource>& source);
+
+}} // namespace cv { namespace gpucodec {
+
+#endif /* __OPENCV_GPUCODEC_HPP__ */
diff --git a/modules/gpucodec/perf/perf_main.cpp b/modules/gpucodec/perf/perf_main.cpp
new file mode 100644
index 0000000000..7839653575
--- /dev/null
+++ b/modules/gpucodec/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpucodec)
diff --git a/modules/gpucodec/perf/perf_precomp.hpp b/modules/gpucodec/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..421fa57633
--- /dev/null
+++ b/modules/gpucodec/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpucodec.hpp"
+#include "opencv2/highgui.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpucodec/perf/perf_video.cpp b/modules/gpucodec/perf/perf_video.cpp
new file mode 100644
index 0000000000..c4ab227c84
--- /dev/null
+++ b/modules/gpucodec/perf/perf_video.cpp
@@ -0,0 +1,162 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+#include "opencv2/highgui/highgui_c.h"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+#if defined(HAVE_XINE)         || \
+    defined(HAVE_GSTREAMER)    || \
+    defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
+    defined(HAVE_AVFOUNDATION) || \
+    defined(HAVE_FFMPEG)       || \
+    defined(WIN32) /* assume that we have ffmpeg */
+
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
+#else
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 0
+#endif
+
+DEF_PARAM_TEST_1(FileName, string);
+
+//////////////////////////////////////////////////////
+// VideoReader
+
+#if defined(HAVE_NVCUVID) && BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+PERF_TEST_P(FileName, VideoReader, Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"))
+{
+    declare.time(20);
+
+    const string inputFile = perf::TestBase::getDataPath(GetParam());
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpucodec::VideoReader> d_reader = cv::gpucodec::createVideoReader(inputFile);
+
+        cv::gpu::GpuMat frame;
+
+        TEST_CYCLE_N(10) d_reader->nextFrame(frame);
+
+        GPU_SANITY_CHECK(frame);
+    }
+    else
+    {
+        cv::VideoCapture reader(inputFile);
+        ASSERT_TRUE( reader.isOpened() );
+
+        cv::Mat frame;
+
+        TEST_CYCLE_N(10) reader >> frame;
+
+        CPU_SANITY_CHECK(frame);
+    }
+}
+
+#endif
+
+//////////////////////////////////////////////////////
+// VideoWriter
+
+#if defined(HAVE_NVCUVID) && defined(WIN32)
+
+PERF_TEST_P(FileName, VideoWriter, Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"))
+{
+    declare.time(30);
+
+    const string inputFile = perf::TestBase::getDataPath(GetParam());
+    const string outputFile = cv::tempfile(".avi");
+
+    const double FPS = 25.0;
+
+    cv::VideoCapture reader(inputFile);
+    ASSERT_TRUE( reader.isOpened() );
+
+    cv::Mat frame;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpucodec::VideoWriter> d_writer;
+
+        cv::gpu::GpuMat d_frame;
+
+        for (int i = 0; i < 10; ++i)
+        {
+            reader >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            d_frame.upload(frame);
+
+            if (d_writer.empty())
+                d_writer = cv::gpucodec::createVideoWriter(outputFile, frame.size(), FPS);
+
+            startTimer(); next();
+            d_writer->write(d_frame);
+            stopTimer();
+        }
+    }
+    else
+    {
+        cv::VideoWriter writer;
+
+        for (int i = 0; i < 10; ++i)
+        {
+            reader >> frame;
+            ASSERT_FALSE(frame.empty());
+
+            if (!writer.isOpened())
+                writer.open(outputFile, CV_FOURCC('X', 'V', 'I', 'D'), FPS, frame.size());
+
+            startTimer(); next();
+            writer.write(frame);
+            stopTimer();
+        }
+    }
+
+    SANITY_CHECK(frame);
+}
+
+#endif
diff --git a/modules/gpucodec/src/cuda/nv12_to_rgb.cu b/modules/gpucodec/src/cuda/nv12_to_rgb.cu
new file mode 100644
index 0000000000..1de916e5a3
--- /dev/null
+++ b/modules/gpucodec/src/cuda/nv12_to_rgb.cu
@@ -0,0 +1,188 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/*
+ * NV12ToARGB color space conversion CUDA kernel
+ *
+ * This sample uses CUDA to perform a simple NV12 (YUV 4:2:0 planar)
+ * source and converts to output in ARGB format
+ */
+
+#include "opencv2/core/cuda/common.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    __constant__ float constHueColorSpaceMat[9] = {1.1644f, 0.0f, 1.596f, 1.1644f, -0.3918f, -0.813f, 1.1644f, 2.0172f, 0.0f};
+
+    __device__ void YUV2RGB(const uint* yuvi, float* red, float* green, float* blue)
+    {
+        float luma, chromaCb, chromaCr;
+
+        // Prepare for hue adjustment
+        luma     = (float)yuvi[0];
+        chromaCb = (float)((int)yuvi[1] - 512.0f);
+        chromaCr = (float)((int)yuvi[2] - 512.0f);
+
+       // Convert YUV To RGB with hue adjustment
+       *red   = (luma     * constHueColorSpaceMat[0]) +
+                (chromaCb * constHueColorSpaceMat[1]) +
+                (chromaCr * constHueColorSpaceMat[2]);
+
+       *green = (luma     * constHueColorSpaceMat[3]) +
+                (chromaCb * constHueColorSpaceMat[4]) +
+                (chromaCr * constHueColorSpaceMat[5]);
+
+       *blue  = (luma     * constHueColorSpaceMat[6]) +
+                (chromaCb * constHueColorSpaceMat[7]) +
+                (chromaCr * constHueColorSpaceMat[8]);
+    }
+
+    __device__ uint RGBA_pack_10bit(float red, float green, float blue, uint alpha)
+    {
+        uint ARGBpixel = 0;
+
+        // Clamp final 10 bit results
+        red   = ::fmin(::fmax(red,   0.0f), 1023.f);
+        green = ::fmin(::fmax(green, 0.0f), 1023.f);
+        blue  = ::fmin(::fmax(blue,  0.0f), 1023.f);
+
+        // Convert to 8 bit unsigned integers per color component
+        ARGBpixel = (((uint)blue  >> 2) |
+                    (((uint)green >> 2) << 8)  |
+                    (((uint)red   >> 2) << 16) |
+                    (uint)alpha);
+
+        return ARGBpixel;
+    }
+
+    // CUDA kernel for outputing the final ARGB output from NV12
+
+    #define COLOR_COMPONENT_BIT_SIZE 10
+    #define COLOR_COMPONENT_MASK     0x3FF
+
+    __global__ void NV12_to_RGB(uchar* srcImage, size_t nSourcePitch,
+                                uint* dstImage, size_t nDestPitch,
+                                uint width, uint height)
+    {
+        // Pad borders with duplicate pixels, and we multiply by 2 because we process 2 pixels per thread
+        const int x = blockIdx.x * (blockDim.x << 1) + (threadIdx.x << 1);
+        const int y = blockIdx.y *  blockDim.y       +  threadIdx.y;
+
+        if (x >= width || y >= height)
+            return;
+
+        // Read 2 Luma components at a time, so we don't waste processing since CbCr are decimated this way.
+        // if we move to texture we could read 4 luminance values
+
+        uint yuv101010Pel[2];
+
+        yuv101010Pel[0] = (srcImage[y * nSourcePitch + x    ]) << 2;
+        yuv101010Pel[1] = (srcImage[y * nSourcePitch + x + 1]) << 2;
+
+        const size_t chromaOffset = nSourcePitch * height;
+
+        const int y_chroma = y >> 1;
+
+        if (y & 1)  // odd scanline ?
+        {
+            uint chromaCb = srcImage[chromaOffset + y_chroma * nSourcePitch + x    ];
+            uint chromaCr = srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1];
+
+            if (y_chroma < ((height >> 1) - 1)) // interpolate chroma vertically
+            {
+                chromaCb = (chromaCb + srcImage[chromaOffset + (y_chroma + 1) * nSourcePitch + x    ] + 1) >> 1;
+                chromaCr = (chromaCr + srcImage[chromaOffset + (y_chroma + 1) * nSourcePitch + x + 1] + 1) >> 1;
+            }
+
+            yuv101010Pel[0] |= (chromaCb << ( COLOR_COMPONENT_BIT_SIZE       + 2));
+            yuv101010Pel[0] |= (chromaCr << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
+
+            yuv101010Pel[1] |= (chromaCb << ( COLOR_COMPONENT_BIT_SIZE       + 2));
+            yuv101010Pel[1] |= (chromaCr << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
+        }
+        else
+        {
+            yuv101010Pel[0] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x    ] << ( COLOR_COMPONENT_BIT_SIZE       + 2));
+            yuv101010Pel[0] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1] << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
+
+            yuv101010Pel[1] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x    ] << ( COLOR_COMPONENT_BIT_SIZE       + 2));
+            yuv101010Pel[1] |= ((uint)srcImage[chromaOffset + y_chroma * nSourcePitch + x + 1] << ((COLOR_COMPONENT_BIT_SIZE << 1) + 2));
+        }
+
+        // this steps performs the color conversion
+        uint yuvi[6];
+        float red[2], green[2], blue[2];
+
+        yuvi[0] =  (yuv101010Pel[0] &   COLOR_COMPONENT_MASK    );
+        yuvi[1] = ((yuv101010Pel[0] >>  COLOR_COMPONENT_BIT_SIZE)       & COLOR_COMPONENT_MASK);
+        yuvi[2] = ((yuv101010Pel[0] >> (COLOR_COMPONENT_BIT_SIZE << 1)) & COLOR_COMPONENT_MASK);
+
+        yuvi[3] =  (yuv101010Pel[1] &   COLOR_COMPONENT_MASK    );
+        yuvi[4] = ((yuv101010Pel[1] >>  COLOR_COMPONENT_BIT_SIZE)       & COLOR_COMPONENT_MASK);
+        yuvi[5] = ((yuv101010Pel[1] >> (COLOR_COMPONENT_BIT_SIZE << 1)) & COLOR_COMPONENT_MASK);
+
+        // YUV to RGB Transformation conversion
+        YUV2RGB(&yuvi[0], &red[0], &green[0], &blue[0]);
+        YUV2RGB(&yuvi[3], &red[1], &green[1], &blue[1]);
+
+        // Clamp the results to RGBA
+
+        const size_t dstImagePitch = nDestPitch >> 2;
+
+        dstImage[y * dstImagePitch + x     ] = RGBA_pack_10bit(red[0], green[0], blue[0], ((uint)0xff << 24));
+        dstImage[y * dstImagePitch + x + 1 ] = RGBA_pack_10bit(red[1], green[1], blue[1], ((uint)0xff << 24));
+    }
+
+    void NV12_to_RGB(const PtrStepb decodedFrame, PtrStepSz<uint> interopFrame, cudaStream_t stream)
+    {
+        dim3 block(32, 8);
+        dim3 grid(divUp(interopFrame.cols, 2 * block.x), divUp(interopFrame.rows, block.y));
+
+        NV12_to_RGB<<<grid, block, 0, stream>>>(decodedFrame.data, decodedFrame.step, interopFrame.data, interopFrame.step,
+            interopFrame.cols, interopFrame.rows);
+
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}}}
diff --git a/modules/gpucodec/src/cuda/rgb_to_yv12.cu b/modules/gpucodec/src/cuda/rgb_to_yv12.cu
new file mode 100644
index 0000000000..8787b1e685
--- /dev/null
+++ b/modules/gpucodec/src/cuda/rgb_to_yv12.cu
@@ -0,0 +1,170 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    __device__ __forceinline__ void rgb_to_y(const uchar b, const uchar g, const uchar r, uchar& y)
+    {
+        y = static_cast<uchar>(((int)(30 * r) + (int)(59 * g) + (int)(11 * b)) / 100);
+    }
+
+    __device__ __forceinline__ void rgb_to_yuv(const uchar b, const uchar g, const uchar r, uchar& y, uchar& u, uchar& v)
+    {
+        rgb_to_y(b, g, r, y);
+        u = static_cast<uchar>(((int)(-17 * r) - (int)(33 * g) + (int)(50 * b) + 12800) / 100);
+        v = static_cast<uchar>(((int)(50 * r) - (int)(42 * g) - (int)(8 * b) + 12800) / 100);
+    }
+
+    __global__ void Gray_to_YV12(const PtrStepSzb src, PtrStepb dst)
+    {
+        const int x = (blockIdx.x * blockDim.x + threadIdx.x) * 2;
+        const int y = (blockIdx.y * blockDim.y + threadIdx.y) * 2;
+
+        if (x + 1 >= src.cols || y + 1 >= src.rows)
+            return;
+
+        // get pointers to the data
+        const size_t planeSize = src.rows * dst.step;
+        PtrStepb y_plane(dst.data, dst.step);
+        PtrStepb u_plane(y_plane.data + planeSize, dst.step / 2);
+        PtrStepb v_plane(u_plane.data + (planeSize / 4), dst.step / 2);
+
+        uchar pix;
+        uchar y_val, u_val, v_val;
+
+        pix = src(y, x);
+        rgb_to_y(pix, pix, pix, y_val);
+        y_plane(y, x) = y_val;
+
+        pix = src(y, x + 1);
+        rgb_to_y(pix, pix, pix, y_val);
+        y_plane(y, x + 1) = y_val;
+
+        pix = src(y + 1, x);
+        rgb_to_y(pix, pix, pix, y_val);
+        y_plane(y + 1, x) = y_val;
+
+        pix = src(y + 1, x + 1);
+        rgb_to_yuv(pix, pix, pix, y_val, u_val, v_val);
+        y_plane(y + 1, x + 1) = y_val;
+        u_plane(y / 2, x / 2) = u_val;
+        v_plane(y / 2, x / 2) = v_val;
+    }
+
+    template <typename T>
+    __global__ void RGB_to_YV12(const PtrStepSz<T> src, PtrStepb dst)
+    {
+        const int x = (blockIdx.x * blockDim.x + threadIdx.x) * 2;
+        const int y = (blockIdx.y * blockDim.y + threadIdx.y) * 2;
+
+        if (x + 1 >= src.cols || y + 1 >= src.rows)
+            return;
+
+        // get pointers to the data
+        const size_t planeSize = src.rows * dst.step;
+        PtrStepb y_plane(dst.data, dst.step);
+        PtrStepb u_plane(y_plane.data + planeSize, dst.step / 2);
+        PtrStepb v_plane(u_plane.data + (planeSize / 4), dst.step / 2);
+
+        T pix;
+        uchar y_val, u_val, v_val;
+
+        pix = src(y, x);
+        rgb_to_y(pix.z, pix.y, pix.x, y_val);
+        y_plane(y, x) = y_val;
+
+        pix = src(y, x + 1);
+        rgb_to_y(pix.z, pix.y, pix.x, y_val);
+        y_plane(y, x + 1) = y_val;
+
+        pix = src(y + 1, x);
+        rgb_to_y(pix.z, pix.y, pix.x, y_val);
+        y_plane(y + 1, x) = y_val;
+
+        pix = src(y + 1, x + 1);
+        rgb_to_yuv(pix.z, pix.y, pix.x, y_val, u_val, v_val);
+        y_plane(y + 1, x + 1) = y_val;
+        u_plane(y / 2, x / 2) = u_val;
+        v_plane(y / 2, x / 2) = v_val;
+    }
+
+    void Gray_to_YV12_caller(const PtrStepSzb src, PtrStepb dst, cudaStream_t stream)
+    {
+        dim3 block(32, 8);
+        dim3 grid(divUp(src.cols, block.x * 2), divUp(src.rows, block.y * 2));
+
+        Gray_to_YV12<<<grid, block, 0, stream>>>(src, dst);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+    template <int cn>
+    void RGB_to_YV12_caller(const PtrStepSzb src, PtrStepb dst, cudaStream_t stream)
+    {
+        typedef typename TypeVec<uchar, cn>::vec_type src_t;
+
+        dim3 block(32, 8);
+        dim3 grid(divUp(src.cols, block.x * 2), divUp(src.rows, block.y * 2));
+
+        RGB_to_YV12<<<grid, block, 0, stream>>>(static_cast< PtrStepSz<src_t> >(src), dst);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    void RGB_to_YV12(const PtrStepSzb src, int cn, PtrStepSzb dst, cudaStream_t stream)
+    {
+        typedef void (*func_t)(const PtrStepSzb src, PtrStepb dst, cudaStream_t stream);
+
+        static const func_t funcs[] =
+        {
+            0, Gray_to_YV12_caller, 0, RGB_to_YV12_caller<3>, RGB_to_YV12_caller<4>
+        };
+
+        funcs[cn](src, dst, stream);
+    }
+}}}
diff --git a/modules/gpucodec/src/cuvid_video_source.cpp b/modules/gpucodec/src/cuvid_video_source.cpp
new file mode 100644
index 0000000000..477951e931
--- /dev/null
+++ b/modules/gpucodec/src/cuvid_video_source.cpp
@@ -0,0 +1,114 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+#ifdef HAVE_NVCUVID
+
+using namespace cv;
+using namespace cv::gpucodec;
+using namespace cv::gpucodec::detail;
+
+cv::gpucodec::detail::CuvidVideoSource::CuvidVideoSource(const String& fname)
+{
+    CUVIDSOURCEPARAMS params;
+    std::memset(&params, 0, sizeof(CUVIDSOURCEPARAMS));
+
+    // Fill parameter struct
+    params.pUserData = this;                        // will be passed to data handlers
+    params.pfnVideoDataHandler = HandleVideoData;   // our local video-handler callback
+    params.pfnAudioDataHandler = 0;
+
+    // now create the actual source
+    CUresult cuRes = cuvidCreateVideoSource(&videoSource_, fname.c_str(), &params);
+    if (cuRes == CUDA_ERROR_INVALID_SOURCE)
+        throw std::runtime_error("");
+    cuSafeCall( cuRes );
+
+    CUVIDEOFORMAT vidfmt;
+    cuSafeCall( cuvidGetSourceVideoFormat(videoSource_, &vidfmt, 0) );
+
+    format_.codec = static_cast<Codec>(vidfmt.codec);
+    format_.chromaFormat = static_cast<ChromaFormat>(vidfmt.chroma_format);
+    format_.width = vidfmt.coded_width;
+    format_.height = vidfmt.coded_height;
+}
+
+cv::gpucodec::detail::CuvidVideoSource::~CuvidVideoSource()
+{
+    cuvidDestroyVideoSource(videoSource_);
+}
+
+FormatInfo cv::gpucodec::detail::CuvidVideoSource::format() const
+{
+    return format_;
+}
+
+void cv::gpucodec::detail::CuvidVideoSource::start()
+{
+    cuSafeCall( cuvidSetVideoSourceState(videoSource_, cudaVideoState_Started) );
+}
+
+void cv::gpucodec::detail::CuvidVideoSource::stop()
+{
+    cuSafeCall( cuvidSetVideoSourceState(videoSource_, cudaVideoState_Stopped) );
+}
+
+bool cv::gpucodec::detail::CuvidVideoSource::isStarted() const
+{
+    return (cuvidGetVideoSourceState(videoSource_) == cudaVideoState_Started);
+}
+
+bool cv::gpucodec::detail::CuvidVideoSource::hasError() const
+{
+    return (cuvidGetVideoSourceState(videoSource_) == cudaVideoState_Error);
+}
+
+int CUDAAPI cv::gpucodec::detail::CuvidVideoSource::HandleVideoData(void* userData, CUVIDSOURCEDATAPACKET* packet)
+{
+    CuvidVideoSource* thiz = static_cast<CuvidVideoSource*>(userData);
+
+    return thiz->parseVideoData(packet->payload, packet->payload_size, (packet->flags & CUVID_PKT_ENDOFSTREAM) != 0);
+}
+
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/cuvid_video_source.hpp b/modules/gpucodec/src/cuvid_video_source.hpp
new file mode 100644
index 0000000000..c2f0e2f571
--- /dev/null
+++ b/modules/gpucodec/src/cuvid_video_source.hpp
@@ -0,0 +1,87 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __CUVID_VIDEO_SOURCE_HPP__
+#define __CUVID_VIDEO_SOURCE_HPP__
+
+#include <nvcuvid.h>
+
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/gpucodec.hpp"
+#include "video_source.hpp"
+
+namespace cv { namespace gpucodec { namespace detail
+{
+
+class CuvidVideoSource : public VideoSource
+{
+public:
+    explicit CuvidVideoSource(const String& fname);
+    ~CuvidVideoSource();
+
+    FormatInfo format() const;
+    void start();
+    void stop();
+    bool isStarted() const;
+    bool hasError() const;
+
+private:
+    // Callback for handling packages of demuxed video data.
+    //
+    // Parameters:
+    //      pUserData - Pointer to user data. We must pass a pointer to a
+    //          VideoSourceData struct here, that contains a valid CUvideoparser
+    //          and FrameQueue.
+    //      pPacket - video-source data packet.
+    //
+    // NOTE: called from a different thread that doesn't not have a cuda context
+    //
+    static int CUDAAPI HandleVideoData(void* pUserData, CUVIDSOURCEDATAPACKET* pPacket);
+
+    CUvideosource videoSource_;
+    FormatInfo format_;
+};
+
+}}}
+
+#endif // __CUVID_VIDEO_SOURCE_HPP__
diff --git a/modules/gpu/src/ffmpeg_video_source.cpp b/modules/gpucodec/src/ffmpeg_video_source.cpp
similarity index 71%
rename from modules/gpu/src/ffmpeg_video_source.cpp
rename to modules/gpucodec/src/ffmpeg_video_source.cpp
index bd3d70058a..86d1b59313 100644
--- a/modules/gpu/src/ffmpeg_video_source.cpp
+++ b/modules/gpucodec/src/ffmpeg_video_source.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,15 +41,13 @@
 //
 //M*/
 
-#include "ffmpeg_video_source.h"
+#include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#ifdef HAVE_NVCUVID
 
-#if defined(HAVE_FFMPEG) && defined(BUILD_SHARED_LIBS)
-    #include "../src/cap_ffmpeg_impl.hpp"
-#else
-    #include "../src/cap_ffmpeg_api.hpp"
-#endif
+using namespace cv;
+using namespace cv::gpucodec;
+using namespace cv::gpucodec::detail;
 
 namespace
 {
@@ -96,7 +95,7 @@ namespace
     }
 }
 
-cv::gpu::detail::FFmpegVideoSource::FFmpegVideoSource(const std::string& fname) :
+cv::gpucodec::detail::FFmpegVideoSource::FFmpegVideoSource(const String& fname) :
     stream_(0)
 {
     CV_Assert( init_MediaStream_FFMPEG() );
@@ -108,75 +107,33 @@ cv::gpu::detail::FFmpegVideoSource::FFmpegVideoSource(const std::string& fname)
 
     stream_ = create_InputMediaStream_FFMPEG_p(fname.c_str(), &codec, &chroma_format, &width, &height);
     if (!stream_)
-        CV_Error(CV_StsUnsupportedFormat, "Unsupported video source");
+        CV_Error(Error::StsUnsupportedFormat, "Unsupported video source");
 
-    format_.codec = static_cast<VideoReader_GPU::Codec>(codec);
-    format_.chromaFormat = static_cast<VideoReader_GPU::ChromaFormat>(chroma_format);
+    format_.codec = static_cast<Codec>(codec);
+    format_.chromaFormat = static_cast<ChromaFormat>(chroma_format);
     format_.width = width;
     format_.height = height;
 }
 
-cv::gpu::detail::FFmpegVideoSource::~FFmpegVideoSource()
+cv::gpucodec::detail::FFmpegVideoSource::~FFmpegVideoSource()
 {
-    release_InputMediaStream_FFMPEG_p(stream_);
+    if (stream_)
+        release_InputMediaStream_FFMPEG_p(stream_);
 }
 
-cv::gpu::VideoReader_GPU::FormatInfo cv::gpu::detail::FFmpegVideoSource::format() const
+FormatInfo cv::gpucodec::detail::FFmpegVideoSource::format() const
 {
     return format_;
 }
 
-void cv::gpu::detail::FFmpegVideoSource::start()
+bool cv::gpucodec::detail::FFmpegVideoSource::getNextPacket(unsigned char** data, int* size, bool* bEndOfFile)
 {
-    stop_ = false;
-    hasError_ = false;
-    thread_.reset(new Thread(readLoop, this));
-}
+    int endOfFile;
 
-void cv::gpu::detail::FFmpegVideoSource::stop()
-{
-    stop_ = true;
-    thread_->wait();
-    thread_.reset();
-}
+    int res = read_InputMediaStream_FFMPEG_p(stream_, data, size, &endOfFile);
 
-bool cv::gpu::detail::FFmpegVideoSource::isStarted() const
-{
-    return !stop_;
-}
-
-bool cv::gpu::detail::FFmpegVideoSource::hasError() const
-{
-    return hasError_;
-}
-
-void cv::gpu::detail::FFmpegVideoSource::readLoop(void* userData)
-{
-    FFmpegVideoSource* thiz = static_cast<FFmpegVideoSource*>(userData);
-
-    for (;;)
-    {
-        unsigned char* data;
-        int size;
-        int endOfFile;
-
-        if (!read_InputMediaStream_FFMPEG_p(thiz->stream_, &data, &size, &endOfFile))
-        {
-            thiz->hasError_ = !endOfFile;
-            break;
-        }
-
-        if (!thiz->parseVideoData(data, size))
-        {
-            thiz->hasError_ = true;
-            break;
-        }
-
-        if (thiz->stop_)
-            break;
-    }
-
-    thiz->parseVideoData(0, 0, true);
+    *bEndOfFile = (endOfFile != 0);
+    return res != 0;
 }
 
 #endif // HAVE_CUDA
diff --git a/modules/gpucodec/src/ffmpeg_video_source.hpp b/modules/gpucodec/src/ffmpeg_video_source.hpp
new file mode 100644
index 0000000000..6ea59ddac1
--- /dev/null
+++ b/modules/gpucodec/src/ffmpeg_video_source.hpp
@@ -0,0 +1,71 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __FFMPEG_VIDEO_SOURCE_HPP__
+#define __FFMPEG_VIDEO_SOURCE_HPP__
+
+#include "opencv2/gpucodec.hpp"
+
+struct InputMediaStream_FFMPEG;
+
+namespace cv { namespace gpucodec { namespace detail {
+
+class FFmpegVideoSource : public RawVideoSource
+{
+public:
+    FFmpegVideoSource(const String& fname);
+    ~FFmpegVideoSource();
+
+    bool getNextPacket(unsigned char** data, int* size, bool* endOfFile);
+
+    FormatInfo format() const;
+
+private:
+    FormatInfo format_;
+
+    InputMediaStream_FFMPEG* stream_;
+};
+
+}}}
+
+#endif // __FFMPEG_VIDEO_SOURCE_HPP__
diff --git a/modules/gpu/src/frame_queue.cpp b/modules/gpucodec/src/frame_queue.cpp
similarity index 86%
rename from modules/gpu/src/frame_queue.cpp
rename to modules/gpucodec/src/frame_queue.cpp
index 0d93918732..f9141d84f5 100644
--- a/modules/gpu/src/frame_queue.cpp
+++ b/modules/gpucodec/src/frame_queue.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,20 +41,20 @@
 //
 //M*/
 
-#include "frame_queue.h"
+#include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#ifdef HAVE_NVCUVID
 
-cv::gpu::detail::FrameQueue::FrameQueue() :
+cv::gpucodec::detail::FrameQueue::FrameQueue() :
     endOfDecode_(0),
     framesInQueue_(0),
     readPosition_(0)
 {
     std::memset(displayQueue_, 0, sizeof(displayQueue_));
-    std::memset((void*)isFrameInUse_, 0, sizeof(isFrameInUse_));
+    std::memset((void*) isFrameInUse_, 0, sizeof(isFrameInUse_));
 }
 
-bool cv::gpu::detail::FrameQueue::waitUntilFrameAvailable(int pictureIndex)
+bool cv::gpucodec::detail::FrameQueue::waitUntilFrameAvailable(int pictureIndex)
 {
     while (isInUse(pictureIndex))
     {
@@ -67,7 +68,7 @@ bool cv::gpu::detail::FrameQueue::waitUntilFrameAvailable(int pictureIndex)
     return true;
 }
 
-void cv::gpu::detail::FrameQueue::enqueue(const CUVIDPARSERDISPINFO* picParams)
+void cv::gpucodec::detail::FrameQueue::enqueue(const CUVIDPARSERDISPINFO* picParams)
 {
     // Mark the frame as 'in-use' so we don't re-use it for decoding until it is no longer needed
     // for display
@@ -79,7 +80,7 @@ void cv::gpu::detail::FrameQueue::enqueue(const CUVIDPARSERDISPINFO* picParams)
         bool isFramePlaced = false;
 
         {
-            CriticalSection::AutoLock autoLock(criticalSection_);
+            AutoLock autoLock(mtx_);
 
             if (framesInQueue_ < MaximumSize)
             {
@@ -98,9 +99,9 @@ void cv::gpu::detail::FrameQueue::enqueue(const CUVIDPARSERDISPINFO* picParams)
     } while (!isEndOfDecode());
 }
 
-bool cv::gpu::detail::FrameQueue::dequeue(CUVIDPARSERDISPINFO& displayInfo)
+bool cv::gpucodec::detail::FrameQueue::dequeue(CUVIDPARSERDISPINFO& displayInfo)
 {
-    CriticalSection::AutoLock autoLock(criticalSection_);
+    AutoLock autoLock(mtx_);
 
     if (framesInQueue_ > 0)
     {
@@ -114,4 +115,4 @@ bool cv::gpu::detail::FrameQueue::dequeue(CUVIDPARSERDISPINFO& displayInfo)
     return false;
 }
 
-#endif // HAVE_CUDA
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/frame_queue.hpp b/modules/gpucodec/src/frame_queue.hpp
new file mode 100644
index 0000000000..c3b427b74f
--- /dev/null
+++ b/modules/gpucodec/src/frame_queue.hpp
@@ -0,0 +1,98 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __FRAME_QUEUE_HPP__
+#define __FRAME_QUEUE_HPP__
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.gpu.hpp"
+
+#include <nvcuvid.h>
+
+namespace cv { namespace gpucodec { namespace detail
+{
+
+class FrameQueue
+{
+public:
+    static const int MaximumSize = 20; // MAX_FRM_CNT;
+
+    FrameQueue();
+
+    void endDecode() { endOfDecode_ = true; }
+    bool isEndOfDecode() const { return endOfDecode_ != 0;}
+
+    // Spins until frame becomes available or decoding gets canceled.
+    // If the requested frame is available the method returns true.
+    // If decoding was interupted before the requested frame becomes
+    // available, the method returns false.
+    bool waitUntilFrameAvailable(int pictureIndex);
+
+    void enqueue(const CUVIDPARSERDISPINFO* picParams);
+
+    // Deque the next frame.
+    // Parameters:
+    //      displayInfo - New frame info gets placed into this object.
+    // Returns:
+    //      true, if a new frame was returned,
+    //      false, if the queue was empty and no new frame could be returned.
+    bool dequeue(CUVIDPARSERDISPINFO& displayInfo);
+
+    void releaseFrame(const CUVIDPARSERDISPINFO& picParams) { isFrameInUse_[picParams.picture_index] = false; }
+
+private:
+    bool isInUse(int pictureIndex) const { return isFrameInUse_[pictureIndex] != 0; }
+
+    Mutex mtx_;
+
+    volatile int isFrameInUse_[MaximumSize];
+    volatile int endOfDecode_;
+
+    int framesInQueue_;
+    int readPosition_;
+    CUVIDPARSERDISPINFO displayQueue_[MaximumSize];
+};
+
+}}}
+
+#endif // __FRAME_QUEUE_HPP__
diff --git a/modules/gpucodec/src/precomp.hpp b/modules/gpucodec/src/precomp.hpp
new file mode 100644
index 0000000000..7cef1b7a94
--- /dev/null
+++ b/modules/gpucodec/src/precomp.hpp
@@ -0,0 +1,81 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <cstdlib>
+#include <cstring>
+#include <deque>
+#include <utility>
+#include <stdexcept>
+#include <iostream>
+
+#include "opencv2/gpucodec.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#ifdef HAVE_NVCUVID
+    #include <nvcuvid.h>
+
+    #ifdef WIN32
+        #define NOMINMAX
+        #include <windows.h>
+        #include <NVEncoderAPI.h>
+    #else
+        #include <pthread.h>
+        #include <unistd.h>
+    #endif
+
+    #include "thread.hpp"
+    #include "video_source.hpp"
+    #include "ffmpeg_video_source.hpp"
+    #include "cuvid_video_source.hpp"
+    #include "frame_queue.hpp"
+    #include "video_decoder.hpp"
+    #include "video_parser.hpp"
+
+    #include "../src/cap_ffmpeg_api.hpp"
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpucodec/src/thread.cpp b/modules/gpucodec/src/thread.cpp
new file mode 100644
index 0000000000..6784a1c8e4
--- /dev/null
+++ b/modules/gpucodec/src/thread.cpp
@@ -0,0 +1,170 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+#ifdef HAVE_NVCUVID
+
+using namespace cv::gpucodec::detail;
+
+#ifdef WIN32
+
+namespace
+{
+    struct UserData
+    {
+        Thread::Func func;
+        void* param;
+    };
+
+    DWORD WINAPI WinThreadFunction(LPVOID lpParam)
+    {
+        UserData* userData = static_cast<UserData*>(lpParam);
+
+        userData->func(userData->param);
+
+        return 0;
+    }
+}
+
+class cv::gpucodec::detail::Thread::Impl
+{
+public:
+    Impl(Thread::Func func, void* userData)
+    {
+        userData_.func = func;
+        userData_.param = userData;
+
+        thread_ = CreateThread(
+            NULL,                   // default security attributes
+            0,                      // use default stack size
+            WinThreadFunction,      // thread function name
+            &userData_,             // argument to thread function
+            0,                      // use default creation flags
+            &threadId_);            // returns the thread identifier
+    }
+
+    ~Impl()
+    {
+        CloseHandle(thread_);
+    }
+
+    void wait()
+    {
+        WaitForSingleObject(thread_, INFINITE);
+    }
+
+private:
+    UserData userData_;
+    HANDLE thread_;
+    DWORD threadId_;
+};
+
+#else
+
+namespace
+{
+    struct UserData
+    {
+        Thread::Func func;
+        void* param;
+    };
+
+    void* PThreadFunction(void* lpParam)
+    {
+        UserData* userData = static_cast<UserData*>(lpParam);
+
+        userData->func(userData->param);
+
+        return 0;
+    }
+}
+
+class cv::gpucodec::detail::Thread::Impl
+{
+public:
+    Impl(Thread::Func func, void* userData)
+    {
+        userData_.func = func;
+        userData_.param = userData;
+
+        pthread_create(&thread_, NULL, PThreadFunction, &userData_);
+    }
+
+    ~Impl()
+    {
+        pthread_detach(thread_);
+    }
+
+    void wait()
+    {
+        pthread_join(thread_, NULL);
+    }
+
+private:
+    pthread_t thread_;
+    UserData userData_;
+};
+
+#endif
+
+cv::gpucodec::detail::Thread::Thread(Func func, void* userData) :
+    impl_(new Impl(func, userData))
+{
+}
+
+void cv::gpucodec::detail::Thread::wait()
+{
+    impl_->wait();
+}
+
+void cv::gpucodec::detail::Thread::sleep(int ms)
+{
+#ifdef WIN32
+    ::Sleep(ms);
+#else
+    ::usleep(ms * 1000);
+#endif
+}
+
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/thread.hpp b/modules/gpucodec/src/thread.hpp
new file mode 100644
index 0000000000..c69367f96f
--- /dev/null
+++ b/modules/gpucodec/src/thread.hpp
@@ -0,0 +1,70 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __THREAD_WRAPPERS_HPP__
+#define __THREAD_WRAPPERS_HPP__
+
+#include "opencv2/core.hpp"
+
+namespace cv { namespace gpucodec { namespace detail {
+
+class Thread
+{
+public:
+    typedef void (*Func)(void* userData);
+
+    explicit Thread(Func func, void* userData = 0);
+
+    void wait();
+
+    static void sleep(int ms);
+
+    class Impl;
+
+private:
+    cv::Ptr<Impl> impl_;
+};
+
+}}}
+
+#endif // __THREAD_WRAPPERS_HPP__
diff --git a/modules/gpu/src/video_decoder.cpp b/modules/gpucodec/src/video_decoder.cpp
similarity index 94%
rename from modules/gpu/src/video_decoder.cpp
rename to modules/gpucodec/src/video_decoder.cpp
index c1e8f86a4c..d734ef363a 100644
--- a/modules/gpu/src/video_decoder.cpp
+++ b/modules/gpucodec/src/video_decoder.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,12 +41,11 @@
 //
 //M*/
 
-#include "video_decoder.h"
-#include "frame_queue.h"
+#include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#ifdef HAVE_NVCUVID
 
-void cv::gpu::detail::VideoDecoder::create(const VideoReader_GPU::FormatInfo& videoFormat)
+void cv::gpucodec::detail::VideoDecoder::create(const FormatInfo& videoFormat)
 {
     release();
 
@@ -104,7 +104,7 @@ void cv::gpu::detail::VideoDecoder::create(const VideoReader_GPU::FormatInfo& vi
     cuSafeCall( cuvidCreateDecoder(&decoder_, &createInfo_) );
 }
 
-void cv::gpu::detail::VideoDecoder::release()
+void cv::gpucodec::detail::VideoDecoder::release()
 {
     if (decoder_)
     {
@@ -113,4 +113,4 @@ void cv::gpu::detail::VideoDecoder::release()
     }
 }
 
-#endif // HAVE_CUDA
+#endif // HAVE_NVCUVID
diff --git a/modules/gpu/src/frame_queue.h b/modules/gpucodec/src/video_decoder.hpp
similarity index 53%
rename from modules/gpu/src/frame_queue.h
rename to modules/gpucodec/src/video_decoder.hpp
index 00c44a4015..05a92f2664 100644
--- a/modules/gpu/src/frame_queue.h
+++ b/modules/gpucodec/src/video_decoder.hpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,64 +41,72 @@
 //
 //M*/
 
-#ifndef __FRAME_QUEUE_H__
-#define __FRAME_QUEUE_H__
+#ifndef __VIDEO_DECODER_HPP__
+#define __VIDEO_DECODER_HPP__
 
-#include "precomp.hpp"
-#include "thread_wrappers.h"
+#include <nvcuvid.h>
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/gpucodec.hpp"
 
-namespace cv { namespace gpu
+namespace cv { namespace gpucodec { namespace detail
 {
-    namespace detail
+
+class VideoDecoder
+{
+public:
+    VideoDecoder(const FormatInfo& videoFormat, CUvideoctxlock lock) : lock_(lock), decoder_(0)
     {
-        class FrameQueue
-        {
-        public:
-            static const int MaximumSize = 20; // MAX_FRM_CNT;
-
-            FrameQueue();
-
-            void endDecode() { endOfDecode_ = true; }
-            bool isEndOfDecode() const { return endOfDecode_ != 0;}
-
-            // Spins until frame becomes available or decoding gets canceled.
-            // If the requested frame is available the method returns true.
-            // If decoding was interupted before the requested frame becomes
-            // available, the method returns false.
-            bool waitUntilFrameAvailable(int pictureIndex);
-
-            void enqueue(const CUVIDPARSERDISPINFO* picParams);
-
-            // Deque the next frame.
-            // Parameters:
-            //      displayInfo - New frame info gets placed into this object.
-            // Returns:
-            //      true, if a new frame was returned,
-            //      false, if the queue was empty and no new frame could be returned.
-            bool dequeue(CUVIDPARSERDISPINFO& displayInfo);
-
-            void releaseFrame(const CUVIDPARSERDISPINFO& picParams) { isFrameInUse_[picParams.picture_index] = false; }
-
-        private:
-            FrameQueue(const FrameQueue&);
-            FrameQueue& operator =(const FrameQueue&);
-
-            bool isInUse(int pictureIndex) const { return isFrameInUse_[pictureIndex] != 0; }
-
-            CriticalSection criticalSection_;
-
-            volatile int isFrameInUse_[MaximumSize];
-            volatile int endOfDecode_;
-
-            int framesInQueue_;
-            int readPosition_;
-            CUVIDPARSERDISPINFO displayQueue_[MaximumSize];
-        };
+        create(videoFormat);
     }
-}}
 
-#endif // HAVE_CUDA
+    ~VideoDecoder()
+    {
+        release();
+    }
 
-#endif // __FRAME_QUEUE_H__
+    void create(const FormatInfo& videoFormat);
+    void release();
+
+    // Get the code-type currently used.
+    cudaVideoCodec codec() const { return createInfo_.CodecType; }
+    unsigned long maxDecodeSurfaces() const { return createInfo_.ulNumDecodeSurfaces; }
+
+    unsigned long frameWidth() const { return createInfo_.ulWidth; }
+    unsigned long frameHeight() const { return createInfo_.ulHeight; }
+
+    unsigned long targetWidth() const { return createInfo_.ulTargetWidth; }
+    unsigned long targetHeight() const { return createInfo_.ulTargetHeight; }
+
+    cudaVideoChromaFormat chromaFormat() const { return createInfo_.ChromaFormat; }
+
+    bool decodePicture(CUVIDPICPARAMS* picParams)
+    {
+        return cuvidDecodePicture(decoder_, picParams) == CUDA_SUCCESS;
+    }
+
+    gpu::GpuMat mapFrame(int picIdx, CUVIDPROCPARAMS& videoProcParams)
+    {
+        CUdeviceptr ptr;
+        unsigned int pitch;
+
+        cuSafeCall( cuvidMapVideoFrame(decoder_, picIdx, &ptr, &pitch, &videoProcParams) );
+
+        return gpu::GpuMat(targetHeight() * 3 / 2, targetWidth(), CV_8UC1, (void*) ptr, pitch);
+    }
+
+    void unmapFrame(gpu::GpuMat& frame)
+    {
+        cuSafeCall( cuvidUnmapVideoFrame(decoder_, (CUdeviceptr) frame.data) );
+        frame.release();
+    }
+
+private:
+    CUvideoctxlock lock_;
+    CUVIDDECODECREATEINFO createInfo_;
+    CUvideodecoder        decoder_;
+};
+
+}}}
+
+#endif // __VIDEO_DECODER_HPP__
diff --git a/modules/gpu/src/video_parser.cpp b/modules/gpucodec/src/video_parser.cpp
similarity index 84%
rename from modules/gpu/src/video_parser.cpp
rename to modules/gpucodec/src/video_parser.cpp
index 70187fcb2f..66aab62ad1 100644
--- a/modules/gpu/src/video_parser.cpp
+++ b/modules/gpucodec/src/video_parser.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,16 +41,15 @@
 //
 //M*/
 
-#include "video_parser.h"
-#include "cu_safe_call.h"
+#include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#ifdef HAVE_NVCUVID
 
-cv::gpu::detail::VideoParser::VideoParser(VideoDecoder* videoDecoder, FrameQueue* frameQueue) :
+cv::gpucodec::detail::VideoParser::VideoParser(VideoDecoder* videoDecoder, FrameQueue* frameQueue) :
     videoDecoder_(videoDecoder), frameQueue_(frameQueue), unparsedPackets_(0), hasError_(false)
 {
     CUVIDPARSERPARAMS params;
-    memset(&params, 0, sizeof(CUVIDPARSERPARAMS));
+    std::memset(&params, 0, sizeof(CUVIDPARSERPARAMS));
 
     params.CodecType              = videoDecoder->codec();
     params.ulMaxNumDecodeSurfaces = videoDecoder->maxDecodeSurfaces();
@@ -62,7 +62,7 @@ cv::gpu::detail::VideoParser::VideoParser(VideoDecoder* videoDecoder, FrameQueue
     cuSafeCall( cuvidCreateVideoParser(&parser_, &params) );
 }
 
-bool cv::gpu::detail::VideoParser::parseVideoData(const unsigned char* data, size_t size, bool endOfStream)
+bool cv::gpucodec::detail::VideoParser::parseVideoData(const unsigned char* data, size_t size, bool endOfStream)
 {
     CUVIDSOURCEDATAPACKET packet;
     std::memset(&packet, 0, sizeof(CUVIDSOURCEDATAPACKET));
@@ -96,7 +96,7 @@ bool cv::gpu::detail::VideoParser::parseVideoData(const unsigned char* data, siz
     return !frameQueue_->isEndOfDecode();
 }
 
-int CUDAAPI cv::gpu::detail::VideoParser::HandleVideoSequence(void* userData, CUVIDEOFORMAT* format)
+int CUDAAPI cv::gpucodec::detail::VideoParser::HandleVideoSequence(void* userData, CUVIDEOFORMAT* format)
 {
     VideoParser* thiz = static_cast<VideoParser*>(userData);
 
@@ -107,10 +107,10 @@ int CUDAAPI cv::gpu::detail::VideoParser::HandleVideoSequence(void* userData, CU
         format->coded_height  != thiz->videoDecoder_->frameHeight() ||
         format->chroma_format != thiz->videoDecoder_->chromaFormat())
     {
-        VideoReader_GPU::FormatInfo newFormat;
+        FormatInfo newFormat;
 
-        newFormat.codec = static_cast<VideoReader_GPU::Codec>(format->codec);
-        newFormat.chromaFormat = static_cast<VideoReader_GPU::ChromaFormat>(format->chroma_format);
+        newFormat.codec = static_cast<Codec>(format->codec);
+        newFormat.chromaFormat = static_cast<ChromaFormat>(format->chroma_format);
         newFormat.width = format->coded_width;
         newFormat.height = format->coded_height;
 
@@ -128,7 +128,7 @@ int CUDAAPI cv::gpu::detail::VideoParser::HandleVideoSequence(void* userData, CU
     return true;
 }
 
-int CUDAAPI cv::gpu::detail::VideoParser::HandlePictureDecode(void* userData, CUVIDPICPARAMS* picParams)
+int CUDAAPI cv::gpucodec::detail::VideoParser::HandlePictureDecode(void* userData, CUVIDPICPARAMS* picParams)
 {
     VideoParser* thiz = static_cast<VideoParser*>(userData);
 
@@ -148,7 +148,7 @@ int CUDAAPI cv::gpu::detail::VideoParser::HandlePictureDecode(void* userData, CU
     return true;
 }
 
-int CUDAAPI cv::gpu::detail::VideoParser::HandlePictureDisplay(void* userData, CUVIDPARSERDISPINFO* picParams)
+int CUDAAPI cv::gpucodec::detail::VideoParser::HandlePictureDisplay(void* userData, CUVIDPARSERDISPINFO* picParams)
 {
     VideoParser* thiz = static_cast<VideoParser*>(userData);
 
@@ -159,4 +159,4 @@ int CUDAAPI cv::gpu::detail::VideoParser::HandlePictureDisplay(void* userData, C
     return true;
 }
 
-#endif // HAVE_CUDA
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/video_parser.hpp b/modules/gpucodec/src/video_parser.hpp
new file mode 100644
index 0000000000..b4dddb3895
--- /dev/null
+++ b/modules/gpucodec/src/video_parser.hpp
@@ -0,0 +1,95 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __VIDEO_PARSER_HPP__
+#define __VIDEO_PARSER_HPP__
+
+#include <nvcuvid.h>
+
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/gpucodec.hpp"
+#include "frame_queue.hpp"
+#include "video_decoder.hpp"
+
+namespace cv { namespace gpucodec { namespace detail
+{
+
+class VideoParser
+{
+public:
+    VideoParser(VideoDecoder* videoDecoder, FrameQueue* frameQueue);
+
+    ~VideoParser()
+    {
+        cuvidDestroyVideoParser(parser_);
+    }
+
+    bool parseVideoData(const unsigned char* data, size_t size, bool endOfStream);
+
+    bool hasError() const { return hasError_; }
+
+private:
+    VideoDecoder* videoDecoder_;
+    FrameQueue* frameQueue_;
+    CUvideoparser parser_;
+    int unparsedPackets_;
+    volatile bool hasError_;
+
+    // Called when the decoder encounters a video format change (or initial sequence header)
+    // This particular implementation of the callback returns 0 in case the video format changes
+    // to something different than the original format. Returning 0 causes a stop of the app.
+    static int CUDAAPI HandleVideoSequence(void* pUserData, CUVIDEOFORMAT* pFormat);
+
+    // Called by the video parser to decode a single picture
+    // Since the parser will deliver data as fast as it can, we need to make sure that the picture
+    // index we're attempting to use for decode is no longer used for display
+    static int CUDAAPI HandlePictureDecode(void* pUserData, CUVIDPICPARAMS* pPicParams);
+
+    // Called by the video parser to display a video frame (in the case of field pictures, there may be
+    // 2 decode calls per 1 display call, since two fields make up one frame)
+    static int CUDAAPI HandlePictureDisplay(void* pUserData, CUVIDPARSERDISPINFO* pPicParams);
+};
+
+}}}
+
+#endif // __VIDEO_PARSER_HPP__
diff --git a/modules/gpucodec/src/video_reader.cpp b/modules/gpucodec/src/video_reader.cpp
new file mode 100644
index 0000000000..3e758087a1
--- /dev/null
+++ b/modules/gpucodec/src/video_reader.cpp
@@ -0,0 +1,238 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::gpucodec;
+
+#ifndef HAVE_NVCUVID
+
+Ptr<VideoReader> cv::gpucodec::createVideoReader(const String&) { throw_no_cuda(); return Ptr<VideoReader>(); }
+Ptr<VideoReader> cv::gpucodec::createVideoReader(const Ptr<RawVideoSource>&) { throw_no_cuda(); return Ptr<VideoReader>(); }
+
+#else // HAVE_NVCUVID
+
+namespace cv { namespace gpu { namespace cudev
+{
+    void NV12_to_RGB(const PtrStepb decodedFrame, PtrStepSz<uint> interopFrame, cudaStream_t stream = 0);
+}}}
+
+using namespace cv::gpucodec::detail;
+
+namespace
+{
+    class VideoReaderImpl : public VideoReader
+    {
+    public:
+        explicit VideoReaderImpl(const Ptr<VideoSource>& source);
+        ~VideoReaderImpl();
+
+        bool nextFrame(OutputArray frame);
+
+        FormatInfo format() const;
+
+    private:
+        Ptr<VideoSource> videoSource_;
+
+        Ptr<FrameQueue> frameQueue_;
+        Ptr<VideoDecoder> videoDecoder_;
+        Ptr<VideoParser> videoParser_;
+
+        CUvideoctxlock lock_;
+
+        std::deque< std::pair<CUVIDPARSERDISPINFO, CUVIDPROCPARAMS> > frames_;
+    };
+
+    FormatInfo VideoReaderImpl::format() const
+    {
+        return videoSource_->format();
+    }
+
+    VideoReaderImpl::VideoReaderImpl(const Ptr<VideoSource>& source) :
+        videoSource_(source),
+        lock_(0)
+    {
+        // init context
+        GpuMat temp(1, 1, CV_8UC1);
+        temp.release();
+
+        CUcontext ctx;
+        cuSafeCall( cuCtxGetCurrent(&ctx) );
+        cuSafeCall( cuvidCtxLockCreate(&lock_, ctx) );
+
+        frameQueue_.reset(new FrameQueue);
+        videoDecoder_.reset(new VideoDecoder(videoSource_->format(), lock_));
+        videoParser_.reset(new VideoParser(videoDecoder_, frameQueue_));
+
+        videoSource_->setVideoParser(videoParser_);
+        videoSource_->start();
+    }
+
+    VideoReaderImpl::~VideoReaderImpl()
+    {
+        frameQueue_->endDecode();
+        videoSource_->stop();
+    }
+
+    class VideoCtxAutoLock
+    {
+    public:
+        VideoCtxAutoLock(CUvideoctxlock lock) : m_lock(lock) { cuSafeCall( cuvidCtxLock(m_lock, 0) ); }
+        ~VideoCtxAutoLock() { cuvidCtxUnlock(m_lock, 0); }
+
+    private:
+        CUvideoctxlock m_lock;
+    };
+
+    void cudaPostProcessFrame(const GpuMat& decodedFrame, OutputArray _outFrame, int width, int height)
+    {
+        using namespace cv::gpu::cudev;
+
+        // Final Stage: NV12toARGB color space conversion
+
+        _outFrame.create(height, width, CV_8UC4);
+        GpuMat outFrame = _outFrame.getGpuMat();
+
+        NV12_to_RGB(decodedFrame, outFrame);
+    }
+
+    bool VideoReaderImpl::nextFrame(OutputArray frame)
+    {
+        if (videoSource_->hasError() || videoParser_->hasError())
+            CV_Error(Error::StsUnsupportedFormat, "Unsupported video source");
+
+        if (!videoSource_->isStarted() || frameQueue_->isEndOfDecode())
+            return false;
+
+        if (frames_.empty())
+        {
+            CUVIDPARSERDISPINFO displayInfo;
+
+            for (;;)
+            {
+                if (frameQueue_->dequeue(displayInfo))
+                    break;
+
+                if (videoSource_->hasError() || videoParser_->hasError())
+                    CV_Error(Error::StsUnsupportedFormat, "Unsupported video source");
+
+                if (frameQueue_->isEndOfDecode())
+                    return false;
+
+                // Wait a bit
+                Thread::sleep(1);
+            }
+
+            bool isProgressive = displayInfo.progressive_frame != 0;
+            const int num_fields = isProgressive ? 1 : 2 + displayInfo.repeat_first_field;
+
+            for (int active_field = 0; active_field < num_fields; ++active_field)
+            {
+                CUVIDPROCPARAMS videoProcParams;
+                std::memset(&videoProcParams, 0, sizeof(CUVIDPROCPARAMS));
+
+                videoProcParams.progressive_frame = displayInfo.progressive_frame;
+                videoProcParams.second_field      = active_field;
+                videoProcParams.top_field_first   = displayInfo.top_field_first;
+                videoProcParams.unpaired_field    = (num_fields == 1);
+
+                frames_.push_back(std::make_pair(displayInfo, videoProcParams));
+            }
+        }
+
+        if (frames_.empty())
+            return false;
+
+        std::pair<CUVIDPARSERDISPINFO, CUVIDPROCPARAMS> frameInfo = frames_.front();
+        frames_.pop_front();
+
+        {
+            VideoCtxAutoLock autoLock(lock_);
+
+            // map decoded video frame to CUDA surface
+            GpuMat decodedFrame = videoDecoder_->mapFrame(frameInfo.first.picture_index, frameInfo.second);
+
+            // perform post processing on the CUDA surface (performs colors space conversion and post processing)
+            // comment this out if we inclue the line of code seen above
+            cudaPostProcessFrame(decodedFrame, frame, videoDecoder_->targetWidth(), videoDecoder_->targetHeight());
+
+            // unmap video frame
+            // unmapFrame() synchronizes with the VideoDecode API (ensures the frame has finished decoding)
+            videoDecoder_->unmapFrame(decodedFrame);
+        }
+
+        // release the frame, so it can be re-used in decoder
+        if (frames_.empty())
+            frameQueue_->releaseFrame(frameInfo.first);
+
+        return true;
+    }
+}
+
+Ptr<VideoReader> cv::gpucodec::createVideoReader(const String& filename)
+{
+    CV_Assert( !filename.empty() );
+
+    Ptr<VideoSource> videoSource;
+
+    try
+    {
+        videoSource.reset(new CuvidVideoSource(filename));
+    }
+    catch (...)
+    {
+        Ptr<RawVideoSource> source(new FFmpegVideoSource(filename));
+        videoSource.reset(new RawVideoSourceWrapper(source));
+    }
+
+    return makePtr<VideoReaderImpl>(videoSource);
+}
+
+Ptr<VideoReader> cv::gpucodec::createVideoReader(const Ptr<RawVideoSource>& source)
+{
+    Ptr<VideoSource> videoSource(new RawVideoSourceWrapper(source));
+    return makePtr<VideoReaderImpl>(videoSource);
+}
+
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/video_source.cpp b/modules/gpucodec/src/video_source.cpp
new file mode 100644
index 0000000000..e4241cf2d5
--- /dev/null
+++ b/modules/gpucodec/src/video_source.cpp
@@ -0,0 +1,121 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+#ifdef HAVE_NVCUVID
+
+using namespace cv;
+using namespace cv::gpucodec;
+using namespace cv::gpucodec::detail;
+
+bool cv::gpucodec::detail::VideoSource::parseVideoData(const unsigned char* data, size_t size, bool endOfStream)
+{
+    return videoParser_->parseVideoData(data, size, endOfStream);
+}
+
+cv::gpucodec::detail::RawVideoSourceWrapper::RawVideoSourceWrapper(const Ptr<RawVideoSource>& source) :
+    source_(source)
+{
+    CV_Assert( !source_.empty() );
+}
+
+cv::gpucodec::FormatInfo cv::gpucodec::detail::RawVideoSourceWrapper::format() const
+{
+    return source_->format();
+}
+
+void cv::gpucodec::detail::RawVideoSourceWrapper::start()
+{
+    stop_ = false;
+    hasError_ = false;
+    thread_.reset(new Thread(readLoop, this));
+}
+
+void cv::gpucodec::detail::RawVideoSourceWrapper::stop()
+{
+    stop_ = true;
+    thread_->wait();
+    thread_.release();
+}
+
+bool cv::gpucodec::detail::RawVideoSourceWrapper::isStarted() const
+{
+    return !stop_;
+}
+
+bool cv::gpucodec::detail::RawVideoSourceWrapper::hasError() const
+{
+    return hasError_;
+}
+
+void cv::gpucodec::detail::RawVideoSourceWrapper::readLoop(void* userData)
+{
+    RawVideoSourceWrapper* thiz = static_cast<RawVideoSourceWrapper*>(userData);
+
+    for (;;)
+    {
+        unsigned char* data;
+        int size;
+        bool endOfFile;
+
+        if (!thiz->source_->getNextPacket(&data, &size, &endOfFile))
+        {
+            thiz->hasError_ = !endOfFile;
+            break;
+        }
+
+        if (!thiz->parseVideoData(data, size))
+        {
+            thiz->hasError_ = true;
+            break;
+        }
+
+        if (thiz->stop_)
+            break;
+    }
+
+    thiz->parseVideoData(0, 0, true);
+}
+
+#endif // HAVE_NVCUVID
diff --git a/modules/gpucodec/src/video_source.hpp b/modules/gpucodec/src/video_source.hpp
new file mode 100644
index 0000000000..b4d930ee09
--- /dev/null
+++ b/modules/gpucodec/src/video_source.hpp
@@ -0,0 +1,99 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __GPUCODEC_VIDEO_SOURCE_H__
+#define __GPUCODEC_VIDEO_SOURCE_H__
+
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/gpucodec.hpp"
+#include "thread.hpp"
+
+namespace cv { namespace gpucodec { namespace detail
+{
+
+class VideoParser;
+
+class VideoSource
+{
+public:
+    virtual ~VideoSource() {}
+
+    virtual FormatInfo format() const = 0;
+    virtual void start() = 0;
+    virtual void stop() = 0;
+    virtual bool isStarted() const = 0;
+    virtual bool hasError() const = 0;
+
+    void setVideoParser(detail::VideoParser* videoParser) { videoParser_ = videoParser; }
+
+protected:
+    bool parseVideoData(const uchar* data, size_t size, bool endOfStream = false);
+
+private:
+    detail::VideoParser* videoParser_;
+};
+
+class RawVideoSourceWrapper : public VideoSource
+{
+public:
+    RawVideoSourceWrapper(const Ptr<RawVideoSource>& source);
+
+    FormatInfo format() const;
+    void start();
+    void stop();
+    bool isStarted() const;
+    bool hasError() const;
+
+private:
+    Ptr<RawVideoSource> source_;
+
+    Ptr<Thread> thread_;
+    volatile bool stop_;
+    volatile bool hasError_;
+
+    static void readLoop(void* userData);
+};
+
+}}}
+
+#endif // __GPUCODEC_VIDEO_SOURCE_H__
diff --git a/modules/gpucodec/src/video_writer.cpp b/modules/gpucodec/src/video_writer.cpp
new file mode 100644
index 0000000000..a1484c2b20
--- /dev/null
+++ b/modules/gpucodec/src/video_writer.cpp
@@ -0,0 +1,919 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::gpucodec;
+
+#if !defined(HAVE_NVCUVID) || !defined(WIN32)
+
+cv::gpucodec::EncoderParams::EncoderParams() { throw_no_cuda(); }
+cv::gpucodec::EncoderParams::EncoderParams(const String&) { throw_no_cuda(); }
+void cv::gpucodec::EncoderParams::load(const String&) { throw_no_cuda(); }
+void cv::gpucodec::EncoderParams::save(const String&) const { throw_no_cuda(); }
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const String&, Size, double, SurfaceFormat) { throw_no_cuda(); return Ptr<VideoWriter>(); }
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const String&, Size, double, const EncoderParams&, SurfaceFormat) { throw_no_cuda(); return Ptr<VideoWriter>(); }
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const Ptr<EncoderCallBack>&, Size, double, SurfaceFormat) { throw_no_cuda(); return Ptr<VideoWriter>(); }
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const Ptr<EncoderCallBack>&, Size, double, const EncoderParams&, SurfaceFormat) { throw_no_cuda(); return Ptr<VideoWriter>(); }
+
+#else // !defined HAVE_CUDA || !defined WIN32
+
+namespace cv { namespace gpu { namespace cudev
+{
+    void RGB_to_YV12(const PtrStepSzb src, int cn, PtrStepSzb dst, cudaStream_t stream = 0);
+}}}
+
+///////////////////////////////////////////////////////////////////////////
+// VideoWriterImpl
+
+namespace
+{
+    class NVEncoderWrapper
+    {
+    public:
+        NVEncoderWrapper() : encoder_(0)
+        {
+            int err;
+
+            err = NVGetHWEncodeCaps();
+            if (err)
+                CV_Error(Error::GpuNotSupported, "No CUDA capability present");
+
+            // Create the Encoder API Interface
+            err = NVCreateEncoder(&encoder_);
+            CV_Assert( err == 0 );
+        }
+
+        ~NVEncoderWrapper()
+        {
+            if (encoder_)
+                NVDestroyEncoder(encoder_);
+        }
+
+        operator NVEncoder() const
+        {
+            return encoder_;
+        }
+
+    private:
+        NVEncoder encoder_;
+    };
+
+    enum CodecType
+    {
+        MPEG1, // not supported yet
+        MPEG2, // not supported yet
+        MPEG4, // not supported yet
+        H264
+    };
+
+    class VideoWriterImpl : public VideoWriter
+    {
+    public:
+        VideoWriterImpl(const Ptr<EncoderCallBack>& callback, Size frameSize, double fps, SurfaceFormat format, CodecType codec = H264);
+        VideoWriterImpl(const Ptr<EncoderCallBack>& callback, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format, CodecType codec = H264);
+
+        void write(InputArray frame, bool lastFrame = false);
+
+        EncoderParams getEncoderParams() const;
+
+    private:
+        void initEncoder(double fps);
+        void setEncodeParams(const EncoderParams& params);
+        void initGpuMemory();
+        void initCallBacks();
+        void createHWEncoder();
+
+        Ptr<EncoderCallBack> callback_;
+        Size frameSize_;
+
+        CodecType codec_;
+        SurfaceFormat inputFormat_;
+        NVVE_SurfaceFormat surfaceFormat_;
+
+        NVEncoderWrapper encoder_;
+
+        GpuMat videoFrame_;
+        CUvideoctxlock cuCtxLock_;
+
+        // CallBacks
+
+        static unsigned char* NVENCAPI HandleAcquireBitStream(int* pBufferSize, void* pUserdata);
+        static void NVENCAPI HandleReleaseBitStream(int nBytesInBuffer, unsigned char* cb, void* pUserdata);
+        static void NVENCAPI HandleOnBeginFrame(const NVVE_BeginFrameInfo* pbfi, void* pUserdata);
+        static void NVENCAPI HandleOnEndFrame(const NVVE_EndFrameInfo* pefi, void* pUserdata);
+    };
+
+    VideoWriterImpl::VideoWriterImpl(const Ptr<EncoderCallBack>& callback, Size frameSize, double fps, SurfaceFormat format, CodecType codec) :
+        callback_(callback),
+        frameSize_(frameSize),
+        codec_(codec),
+        inputFormat_(format),
+        cuCtxLock_(0)
+    {
+        surfaceFormat_ = (inputFormat_ == SF_BGR ? YV12 : static_cast<NVVE_SurfaceFormat>(inputFormat_));
+
+        initEncoder(fps);
+
+        initGpuMemory();
+
+        initCallBacks();
+
+        createHWEncoder();
+    }
+
+    VideoWriterImpl::VideoWriterImpl(const Ptr<EncoderCallBack>& callback, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format, CodecType codec) :
+        callback_(callback),
+        frameSize_(frameSize),
+        codec_(codec),
+        inputFormat_(format),
+        cuCtxLock_(0)
+    {
+        surfaceFormat_ = (inputFormat_ == SF_BGR ? YV12 : static_cast<NVVE_SurfaceFormat>(inputFormat_));
+
+        initEncoder(fps);
+
+        setEncodeParams(params);
+
+        initGpuMemory();
+
+        initCallBacks();
+
+        createHWEncoder();
+    }
+
+    void VideoWriterImpl::initEncoder(double fps)
+    {
+        int err;
+
+        // Set codec
+
+        static const unsigned long codecs_id[] =
+        {
+            NV_CODEC_TYPE_MPEG1, NV_CODEC_TYPE_MPEG2, NV_CODEC_TYPE_MPEG4, NV_CODEC_TYPE_H264, NV_CODEC_TYPE_VC1
+        };
+        err = NVSetCodec(encoder_, codecs_id[codec_]);
+        if (err)
+            CV_Error(Error::StsNotImplemented, "Codec format is not supported");
+
+        // Set default params
+
+        err = NVSetDefaultParam(encoder_);
+        CV_Assert( err == 0 );
+
+        // Set some common params
+
+        int inputSize[] = { frameSize_.width, frameSize_.height };
+        err = NVSetParamValue(encoder_, NVVE_IN_SIZE, &inputSize);
+        CV_Assert( err == 0 );
+        err = NVSetParamValue(encoder_, NVVE_OUT_SIZE, &inputSize);
+        CV_Assert( err == 0 );
+
+        int aspectRatio[] = { frameSize_.width, frameSize_.height, ASPECT_RATIO_DAR };
+        err = NVSetParamValue(encoder_, NVVE_ASPECT_RATIO, &aspectRatio);
+        CV_Assert( err == 0 );
+
+        // FPS
+
+        int frame_rate = static_cast<int>(fps + 0.5);
+        int frame_rate_base = 1;
+        while (fabs(static_cast<double>(frame_rate) / frame_rate_base) - fps > 0.001)
+        {
+            frame_rate_base *= 10;
+            frame_rate = static_cast<int>(fps*frame_rate_base + 0.5);
+        }
+        int FrameRate[] = { frame_rate, frame_rate_base };
+        err = NVSetParamValue(encoder_, NVVE_FRAME_RATE, &FrameRate);
+        CV_Assert( err == 0 );
+
+        // Select device for encoding
+
+        int gpuID = getDevice();
+        err = NVSetParamValue(encoder_, NVVE_FORCE_GPU_SELECTION, &gpuID);
+        CV_Assert( err == 0 );
+    }
+
+    void VideoWriterImpl::setEncodeParams(const EncoderParams& params)
+    {
+        int err;
+
+        int P_Interval = params.P_Interval;
+        err = NVSetParamValue(encoder_, NVVE_P_INTERVAL, &P_Interval);
+        CV_Assert( err == 0 );
+
+        int IDR_Period = params.IDR_Period;
+        err = NVSetParamValue(encoder_, NVVE_IDR_PERIOD, &IDR_Period);
+        CV_Assert( err == 0 );
+
+        int DynamicGOP = params.DynamicGOP;
+        err = NVSetParamValue(encoder_, NVVE_DYNAMIC_GOP, &DynamicGOP);
+        CV_Assert( err == 0 );
+
+        NVVE_RateCtrlType RCType = static_cast<NVVE_RateCtrlType>(params.RCType);
+        err = NVSetParamValue(encoder_, NVVE_RC_TYPE, &RCType);
+        CV_Assert( err == 0 );
+
+        int AvgBitrate = params.AvgBitrate;
+        err = NVSetParamValue(encoder_, NVVE_AVG_BITRATE, &AvgBitrate);
+        CV_Assert( err == 0 );
+
+        int PeakBitrate = params.PeakBitrate;
+        err = NVSetParamValue(encoder_, NVVE_PEAK_BITRATE, &PeakBitrate);
+        CV_Assert( err == 0 );
+
+        int QP_Level_Intra = params.QP_Level_Intra;
+        err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTRA, &QP_Level_Intra);
+        CV_Assert( err == 0 );
+
+        int QP_Level_InterP = params.QP_Level_InterP;
+        err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTER_P, &QP_Level_InterP);
+        CV_Assert( err == 0 );
+
+        int QP_Level_InterB = params.QP_Level_InterB;
+        err = NVSetParamValue(encoder_, NVVE_QP_LEVEL_INTER_B, &QP_Level_InterB);
+        CV_Assert( err == 0 );
+
+        int DeblockMode = params.DeblockMode;
+        err = NVSetParamValue(encoder_, NVVE_DEBLOCK_MODE, &DeblockMode);
+        CV_Assert( err == 0 );
+
+        int ProfileLevel = params.ProfileLevel;
+        err = NVSetParamValue(encoder_, NVVE_PROFILE_LEVEL, &ProfileLevel);
+        CV_Assert( err == 0 );
+
+        int ForceIntra = params.ForceIntra;
+        err = NVSetParamValue(encoder_, NVVE_FORCE_INTRA, &ForceIntra);
+        CV_Assert( err == 0 );
+
+        int ForceIDR = params.ForceIDR;
+        err = NVSetParamValue(encoder_, NVVE_FORCE_IDR, &ForceIDR);
+        CV_Assert( err == 0 );
+
+        int ClearStat = params.ClearStat;
+        err = NVSetParamValue(encoder_, NVVE_CLEAR_STAT, &ClearStat);
+        CV_Assert( err == 0 );
+
+        NVVE_DI_MODE DIMode = static_cast<NVVE_DI_MODE>(params.DIMode);
+        err = NVSetParamValue(encoder_, NVVE_SET_DEINTERLACE, &DIMode);
+        CV_Assert( err == 0 );
+
+        if (params.Presets != -1)
+        {
+            NVVE_PRESETS_TARGET Presets = static_cast<NVVE_PRESETS_TARGET>(params.Presets);
+            err = NVSetParamValue(encoder_, NVVE_PRESETS, &Presets);
+            CV_Assert( err == 0 );
+        }
+
+        int DisableCabac = params.DisableCabac;
+        err = NVSetParamValue(encoder_, NVVE_DISABLE_CABAC, &DisableCabac);
+        CV_Assert( err == 0 );
+
+        int NaluFramingType = params.NaluFramingType;
+        err = NVSetParamValue(encoder_, NVVE_CONFIGURE_NALU_FRAMING_TYPE, &NaluFramingType);
+        CV_Assert( err == 0 );
+
+        int DisableSPSPPS = params.DisableSPSPPS;
+        err = NVSetParamValue(encoder_, NVVE_DISABLE_SPS_PPS, &DisableSPSPPS);
+        CV_Assert( err == 0 );
+    }
+
+    EncoderParams VideoWriterImpl::getEncoderParams() const
+    {
+        int err;
+
+        EncoderParams params;
+
+        int P_Interval;
+        err = NVGetParamValue(encoder_, NVVE_P_INTERVAL, &P_Interval);
+        CV_Assert( err == 0 );
+        params.P_Interval = P_Interval;
+
+        int IDR_Period;
+        err = NVGetParamValue(encoder_, NVVE_IDR_PERIOD, &IDR_Period);
+        CV_Assert( err == 0 );
+        params.IDR_Period = IDR_Period;
+
+        int DynamicGOP;
+        err = NVGetParamValue(encoder_, NVVE_DYNAMIC_GOP, &DynamicGOP);
+        CV_Assert( err == 0 );
+        params.DynamicGOP = DynamicGOP;
+
+        NVVE_RateCtrlType RCType;
+        err = NVGetParamValue(encoder_, NVVE_RC_TYPE, &RCType);
+        CV_Assert( err == 0 );
+        params.RCType = RCType;
+
+        int AvgBitrate;
+        err = NVGetParamValue(encoder_, NVVE_AVG_BITRATE, &AvgBitrate);
+        CV_Assert( err == 0 );
+        params.AvgBitrate = AvgBitrate;
+
+        int PeakBitrate;
+        err = NVGetParamValue(encoder_, NVVE_PEAK_BITRATE, &PeakBitrate);
+        CV_Assert( err == 0 );
+        params.PeakBitrate = PeakBitrate;
+
+        int QP_Level_Intra;
+        err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTRA, &QP_Level_Intra);
+        CV_Assert( err == 0 );
+        params.QP_Level_Intra = QP_Level_Intra;
+
+        int QP_Level_InterP;
+        err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTER_P, &QP_Level_InterP);
+        CV_Assert( err == 0 );
+        params.QP_Level_InterP = QP_Level_InterP;
+
+        int QP_Level_InterB;
+        err = NVGetParamValue(encoder_, NVVE_QP_LEVEL_INTER_B, &QP_Level_InterB);
+        CV_Assert( err == 0 );
+        params.QP_Level_InterB = QP_Level_InterB;
+
+        int DeblockMode;
+        err = NVGetParamValue(encoder_, NVVE_DEBLOCK_MODE, &DeblockMode);
+        CV_Assert( err == 0 );
+        params.DeblockMode = DeblockMode;
+
+        int ProfileLevel;
+        err = NVGetParamValue(encoder_, NVVE_PROFILE_LEVEL, &ProfileLevel);
+        CV_Assert( err == 0 );
+        params.ProfileLevel = ProfileLevel;
+
+        int ForceIntra;
+        err = NVGetParamValue(encoder_, NVVE_FORCE_INTRA, &ForceIntra);
+        CV_Assert( err == 0 );
+        params.ForceIntra = ForceIntra;
+
+        int ForceIDR;
+        err = NVGetParamValue(encoder_, NVVE_FORCE_IDR, &ForceIDR);
+        CV_Assert( err == 0 );
+        params.ForceIDR = ForceIDR;
+
+        int ClearStat;
+        err = NVGetParamValue(encoder_, NVVE_CLEAR_STAT, &ClearStat);
+        CV_Assert( err == 0 );
+        params.ClearStat = ClearStat;
+
+        NVVE_DI_MODE DIMode;
+        err = NVGetParamValue(encoder_, NVVE_SET_DEINTERLACE, &DIMode);
+        CV_Assert( err == 0 );
+        params.DIMode = DIMode;
+
+        params.Presets = -1;
+
+        int DisableCabac;
+        err = NVGetParamValue(encoder_, NVVE_DISABLE_CABAC, &DisableCabac);
+        CV_Assert( err == 0 );
+        params.DisableCabac = DisableCabac;
+
+        int NaluFramingType;
+        err = NVGetParamValue(encoder_, NVVE_CONFIGURE_NALU_FRAMING_TYPE, &NaluFramingType);
+        CV_Assert( err == 0 );
+        params.NaluFramingType = NaluFramingType;
+
+        int DisableSPSPPS;
+        err = NVGetParamValue(encoder_, NVVE_DISABLE_SPS_PPS, &DisableSPSPPS);
+        CV_Assert( err == 0 );
+        params.DisableSPSPPS = DisableSPSPPS;
+
+        return params;
+    }
+
+    void VideoWriterImpl::initGpuMemory()
+    {
+        int err;
+
+        // initialize context
+        GpuMat temp(1, 1, CV_8U);
+        temp.release();
+
+        static const int bpp[] =
+        {
+            16, // UYVY, 4:2:2
+            16, // YUY2, 4:2:2
+            12, // YV12, 4:2:0
+            12, // NV12, 4:2:0
+            12, // IYUV, 4:2:0
+        };
+
+        CUcontext cuContext;
+        cuSafeCall( cuCtxGetCurrent(&cuContext) );
+
+        // Allocate the CUDA memory Pitched Surface
+        if (surfaceFormat_ == UYVY || surfaceFormat_ == YUY2)
+            videoFrame_.create(frameSize_.height, (frameSize_.width * bpp[surfaceFormat_]) / 8, CV_8UC1);
+        else
+            videoFrame_.create((frameSize_.height * bpp[surfaceFormat_]) / 8, frameSize_.width, CV_8UC1);
+
+        // Create the Video Context Lock (used for synchronization)
+        cuSafeCall( cuvidCtxLockCreate(&cuCtxLock_, cuContext) );
+
+        // If we are using GPU Device Memory with NVCUVENC, it is necessary to create a
+        // CUDA Context with a Context Lock cuvidCtxLock.  The Context Lock needs to be passed to NVCUVENC
+
+        int iUseDeviceMem = 1;
+        err = NVSetParamValue(encoder_, NVVE_DEVICE_MEMORY_INPUT, &iUseDeviceMem);
+        CV_Assert( err == 0 );
+
+        err = NVSetParamValue(encoder_, NVVE_DEVICE_CTX_LOCK, &cuCtxLock_);
+        CV_Assert( err == 0 );
+    }
+
+    void VideoWriterImpl::initCallBacks()
+    {
+        NVVE_CallbackParams cb;
+        memset(&cb, 0, sizeof(NVVE_CallbackParams));
+
+        cb.pfnacquirebitstream = HandleAcquireBitStream;
+        cb.pfnonbeginframe     = HandleOnBeginFrame;
+        cb.pfnonendframe       = HandleOnEndFrame;
+        cb.pfnreleasebitstream = HandleReleaseBitStream;
+
+        NVRegisterCB(encoder_, cb, this);
+    }
+
+    void VideoWriterImpl::createHWEncoder()
+    {
+        int err;
+
+        // Create the NVIDIA HW resources for Encoding on NVIDIA hardware
+        err = NVCreateHWEncoder(encoder_);
+        CV_Assert( err == 0 );
+    }
+
+    // UYVY/YUY2 are both 4:2:2 formats (16bpc)
+    // Luma, U, V are interleaved, chroma is subsampled (w/2,h)
+    void copyUYVYorYUY2Frame(Size frameSize, const GpuMat& src, GpuMat& dst)
+    {
+        // Source is YUVY/YUY2 4:2:2, the YUV data in a packed and interleaved
+
+        // YUV Copy setup
+        CUDA_MEMCPY2D stCopyYUV422;
+        memset(&stCopyYUV422, 0, sizeof(CUDA_MEMCPY2D));
+
+        stCopyYUV422.srcXInBytes          = 0;
+        stCopyYUV422.srcY                 = 0;
+        stCopyYUV422.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyYUV422.srcHost              = 0;
+        stCopyYUV422.srcDevice            = (CUdeviceptr) src.data;
+        stCopyYUV422.srcArray             = 0;
+        stCopyYUV422.srcPitch             = src.step;
+
+        stCopyYUV422.dstXInBytes          = 0;
+        stCopyYUV422.dstY                 = 0;
+        stCopyYUV422.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyYUV422.dstHost              = 0;
+        stCopyYUV422.dstDevice            = (CUdeviceptr) dst.data;
+        stCopyYUV422.dstArray             = 0;
+        stCopyYUV422.dstPitch             = dst.step;
+
+        stCopyYUV422.WidthInBytes         = frameSize.width * 2;
+        stCopyYUV422.Height               = frameSize.height;
+
+        // DMA Luma/Chroma
+        cuSafeCall( cuMemcpy2D(&stCopyYUV422) );
+    }
+
+    // YV12/IYUV are both 4:2:0 planar formats (12bpc)
+    // Luma, U, V chroma planar (12bpc), chroma is subsampled (w/2,h/2)
+    void copyYV12orIYUVFrame(Size frameSize, const GpuMat& src, GpuMat& dst)
+    {
+        // Source is YV12/IYUV, this native format is converted to NV12 format by the video encoder
+
+        // (1) luma copy setup
+        CUDA_MEMCPY2D stCopyLuma;
+        memset(&stCopyLuma, 0, sizeof(CUDA_MEMCPY2D));
+
+        stCopyLuma.srcXInBytes          = 0;
+        stCopyLuma.srcY                 = 0;
+        stCopyLuma.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyLuma.srcHost              = 0;
+        stCopyLuma.srcDevice            = (CUdeviceptr) src.data;
+        stCopyLuma.srcArray             = 0;
+        stCopyLuma.srcPitch             = src.step;
+
+        stCopyLuma.dstXInBytes          = 0;
+        stCopyLuma.dstY                 = 0;
+        stCopyLuma.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyLuma.dstHost              = 0;
+        stCopyLuma.dstDevice            = (CUdeviceptr) dst.data;
+        stCopyLuma.dstArray             = 0;
+        stCopyLuma.dstPitch             = dst.step;
+
+        stCopyLuma.WidthInBytes         = frameSize.width;
+        stCopyLuma.Height               = frameSize.height;
+
+        // (2) chroma copy setup, U/V can be done together
+        CUDA_MEMCPY2D stCopyChroma;
+        memset(&stCopyChroma, 0, sizeof(CUDA_MEMCPY2D));
+
+        stCopyChroma.srcXInBytes        = 0;
+        stCopyChroma.srcY               = frameSize.height << 1; // U/V chroma offset
+        stCopyChroma.srcMemoryType      = CU_MEMORYTYPE_DEVICE;
+        stCopyChroma.srcHost            = 0;
+        stCopyChroma.srcDevice          = (CUdeviceptr) src.data;
+        stCopyChroma.srcArray           = 0;
+        stCopyChroma.srcPitch           = src.step >> 1; // chroma is subsampled by 2 (but it has U/V are next to each other)
+
+        stCopyChroma.dstXInBytes        = 0;
+        stCopyChroma.dstY               = frameSize.height << 1; // chroma offset (srcY*srcPitch now points to the chroma planes)
+        stCopyChroma.dstMemoryType      = CU_MEMORYTYPE_DEVICE;
+        stCopyChroma.dstHost            = 0;
+        stCopyChroma.dstDevice          = (CUdeviceptr) dst.data;
+        stCopyChroma.dstArray           = 0;
+        stCopyChroma.dstPitch           = dst.step >> 1;
+
+        stCopyChroma.WidthInBytes       = frameSize.width >> 1;
+        stCopyChroma.Height             = frameSize.height; // U/V are sent together
+
+        // DMA Luma
+        cuSafeCall( cuMemcpy2D(&stCopyLuma) );
+
+        // DMA Chroma channels (UV side by side)
+        cuSafeCall( cuMemcpy2D(&stCopyChroma) );
+    }
+
+    // NV12 is 4:2:0 format (12bpc)
+    // Luma followed by U/V chroma interleaved (12bpc), chroma is subsampled (w/2,h/2)
+    void copyNV12Frame(Size frameSize, const GpuMat& src, GpuMat& dst)
+    {
+        // Source is NV12 in pitch linear memory
+        // Because we are assume input is NV12 (if we take input in the native format), the encoder handles NV12 as a native format in pitch linear memory
+
+        // Luma/Chroma can be done in a single transfer
+        CUDA_MEMCPY2D stCopyNV12;
+        memset(&stCopyNV12, 0, sizeof(CUDA_MEMCPY2D));
+
+        stCopyNV12.srcXInBytes          = 0;
+        stCopyNV12.srcY                 = 0;
+        stCopyNV12.srcMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyNV12.srcHost              = 0;
+        stCopyNV12.srcDevice            = (CUdeviceptr) src.data;
+        stCopyNV12.srcArray             = 0;
+        stCopyNV12.srcPitch             = src.step;
+
+        stCopyNV12.dstXInBytes          = 0;
+        stCopyNV12.dstY                 = 0;
+        stCopyNV12.dstMemoryType        = CU_MEMORYTYPE_DEVICE;
+        stCopyNV12.dstHost              = 0;
+        stCopyNV12.dstDevice            = (CUdeviceptr) dst.data;
+        stCopyNV12.dstArray             = 0;
+        stCopyNV12.dstPitch             = dst.step;
+
+        stCopyNV12.WidthInBytes         = frameSize.width;
+        stCopyNV12.Height               = (frameSize.height * 3) >> 1;
+
+        // DMA Luma/Chroma
+        cuSafeCall( cuMemcpy2D(&stCopyNV12) );
+    }
+
+    void VideoWriterImpl::write(InputArray _frame, bool lastFrame)
+    {
+        GpuMat frame = _frame.getGpuMat();
+
+        if (inputFormat_ == SF_BGR)
+        {
+            CV_Assert( frame.size() == frameSize_ );
+            CV_Assert( frame.type() == CV_8UC1 || frame.type() == CV_8UC3 || frame.type() == CV_8UC4 );
+        }
+        else
+        {
+            CV_Assert( frame.size() == videoFrame_.size() );
+            CV_Assert( frame.type() == videoFrame_.type() );
+        }
+
+        NVVE_EncodeFrameParams efparams;
+        efparams.Width = frameSize_.width;
+        efparams.Height = frameSize_.height;
+        efparams.Pitch = static_cast<int>(videoFrame_.step);
+        efparams.SurfFmt = surfaceFormat_;
+        efparams.PictureStruc = FRAME_PICTURE;
+        efparams.topfieldfirst =  0;
+        efparams.repeatFirstField = 0;
+        efparams.progressiveFrame = (surfaceFormat_ == NV12) ? 1 : 0;
+        efparams.bLast = lastFrame;
+        efparams.picBuf = 0; // Must be set to NULL in order to support device memory input
+
+        // Don't forget we need to lock/unlock between memcopies
+        cuSafeCall( cuvidCtxLock(cuCtxLock_, 0) );
+
+        if (inputFormat_ == SF_BGR)
+        {
+            cudev::RGB_to_YV12(frame, frame.channels(), videoFrame_);
+        }
+        else
+        {
+            switch (surfaceFormat_)
+            {
+            case UYVY: // UYVY (4:2:2)
+            case YUY2: // YUY2 (4:2:2)
+                copyUYVYorYUY2Frame(frameSize_, frame, videoFrame_);
+                break;
+
+            case YV12: // YV12 (4:2:0), Y V U
+            case IYUV: // IYUV (4:2:0), Y U V
+                copyYV12orIYUVFrame(frameSize_, frame, videoFrame_);
+                break;
+
+            case NV12: // NV12 (4:2:0)
+                copyNV12Frame(frameSize_, frame, videoFrame_);
+                break;
+            }
+        }
+
+        cuSafeCall( cuvidCtxUnlock(cuCtxLock_, 0) );
+
+        int err = NVEncodeFrame(encoder_, &efparams, 0, videoFrame_.data);
+        CV_Assert( err == 0 );
+    }
+
+    unsigned char* NVENCAPI VideoWriterImpl::HandleAcquireBitStream(int* pBufferSize, void* pUserdata)
+    {
+        VideoWriterImpl* thiz = static_cast<VideoWriterImpl*>(pUserdata);
+
+        return thiz->callback_->acquireBitStream(pBufferSize);
+    }
+
+    void NVENCAPI VideoWriterImpl::HandleReleaseBitStream(int nBytesInBuffer, unsigned char* cb, void* pUserdata)
+    {
+        VideoWriterImpl* thiz = static_cast<VideoWriterImpl*>(pUserdata);
+
+        thiz->callback_->releaseBitStream(cb, nBytesInBuffer);
+    }
+
+    void NVENCAPI VideoWriterImpl::HandleOnBeginFrame(const NVVE_BeginFrameInfo* pbfi, void* pUserdata)
+    {
+        VideoWriterImpl* thiz = static_cast<VideoWriterImpl*>(pUserdata);
+
+        thiz->callback_->onBeginFrame(pbfi->nFrameNumber, static_cast<EncoderCallBack::PicType>(pbfi->nPicType));
+    }
+
+    void NVENCAPI VideoWriterImpl::HandleOnEndFrame(const NVVE_EndFrameInfo* pefi, void* pUserdata)
+    {
+        VideoWriterImpl* thiz = static_cast<VideoWriterImpl*>(pUserdata);
+
+        thiz->callback_->onEndFrame(pefi->nFrameNumber, static_cast<EncoderCallBack::PicType>(pefi->nPicType));
+    }
+
+    ///////////////////////////////////////////////////////////////////////////
+    // FFMPEG
+
+    class EncoderCallBackFFMPEG : public EncoderCallBack
+    {
+    public:
+        EncoderCallBackFFMPEG(const String& fileName, Size frameSize, double fps);
+        ~EncoderCallBackFFMPEG();
+
+        unsigned char* acquireBitStream(int* bufferSize);
+        void releaseBitStream(unsigned char* data, int size);
+        void onBeginFrame(int frameNumber, PicType picType);
+        void onEndFrame(int frameNumber, PicType picType);
+
+    private:
+        static bool init_MediaStream_FFMPEG();
+
+        struct OutputMediaStream_FFMPEG* stream_;
+        std::vector<uchar> buf_;
+        bool isKeyFrame_;
+
+        static Create_OutputMediaStream_FFMPEG_Plugin create_OutputMediaStream_FFMPEG_p;
+        static Release_OutputMediaStream_FFMPEG_Plugin release_OutputMediaStream_FFMPEG_p;
+        static Write_OutputMediaStream_FFMPEG_Plugin write_OutputMediaStream_FFMPEG_p;
+    };
+
+    Create_OutputMediaStream_FFMPEG_Plugin EncoderCallBackFFMPEG::create_OutputMediaStream_FFMPEG_p = 0;
+    Release_OutputMediaStream_FFMPEG_Plugin EncoderCallBackFFMPEG::release_OutputMediaStream_FFMPEG_p = 0;
+    Write_OutputMediaStream_FFMPEG_Plugin EncoderCallBackFFMPEG::write_OutputMediaStream_FFMPEG_p = 0;
+
+    bool EncoderCallBackFFMPEG::init_MediaStream_FFMPEG()
+    {
+        static bool initialized = false;
+
+        if (!initialized)
+        {
+            #if defined(WIN32) || defined(_WIN32)
+                const char* module_name = "opencv_ffmpeg"
+                    CVAUX_STR(CV_VERSION_EPOCH) CVAUX_STR(CV_VERSION_MAJOR) CVAUX_STR(CV_VERSION_MINOR)
+                #if (defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__)
+                    "_64"
+                #endif
+                    ".dll";
+
+                static HMODULE cvFFOpenCV = LoadLibrary(module_name);
+
+                if (cvFFOpenCV)
+                {
+                    create_OutputMediaStream_FFMPEG_p =
+                        (Create_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "create_OutputMediaStream_FFMPEG");
+                    release_OutputMediaStream_FFMPEG_p =
+                        (Release_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "release_OutputMediaStream_FFMPEG");
+                    write_OutputMediaStream_FFMPEG_p =
+                        (Write_OutputMediaStream_FFMPEG_Plugin)GetProcAddress(cvFFOpenCV, "write_OutputMediaStream_FFMPEG");
+
+                    initialized = create_OutputMediaStream_FFMPEG_p != 0 && release_OutputMediaStream_FFMPEG_p != 0 && write_OutputMediaStream_FFMPEG_p != 0;
+                }
+            #elif defined(HAVE_FFMPEG)
+                create_OutputMediaStream_FFMPEG_p = create_OutputMediaStream_FFMPEG;
+                release_OutputMediaStream_FFMPEG_p = release_OutputMediaStream_FFMPEG;
+                write_OutputMediaStream_FFMPEG_p = write_OutputMediaStream_FFMPEG;
+
+                initialized = true;
+            #endif
+        }
+
+        return initialized;
+    }
+
+    EncoderCallBackFFMPEG::EncoderCallBackFFMPEG(const String& fileName, Size frameSize, double fps) :
+        stream_(0), isKeyFrame_(false)
+    {
+        int buf_size = std::max(frameSize.area() * 4, 1024 * 1024);
+        buf_.resize(buf_size);
+
+        CV_Assert( init_MediaStream_FFMPEG() );
+
+        stream_ = create_OutputMediaStream_FFMPEG_p(fileName.c_str(), frameSize.width, frameSize.height, fps);
+        CV_Assert( stream_ != 0 );
+    }
+
+    EncoderCallBackFFMPEG::~EncoderCallBackFFMPEG()
+    {
+        release_OutputMediaStream_FFMPEG_p(stream_);
+    }
+
+    unsigned char* EncoderCallBackFFMPEG::acquireBitStream(int* bufferSize)
+    {
+        *bufferSize = static_cast<int>(buf_.size());
+        return &buf_[0];
+    }
+
+    void EncoderCallBackFFMPEG::releaseBitStream(unsigned char* data, int size)
+    {
+        write_OutputMediaStream_FFMPEG_p(stream_, data, size, isKeyFrame_);
+    }
+
+    void EncoderCallBackFFMPEG::onBeginFrame(int frameNumber, PicType picType)
+    {
+        (void) frameNumber;
+        isKeyFrame_ = (picType == IFRAME);
+    }
+
+    void EncoderCallBackFFMPEG::onEndFrame(int frameNumber, PicType picType)
+    {
+        (void) frameNumber;
+        (void) picType;
+    }
+}
+
+///////////////////////////////////////////////////////////////////////////
+// EncoderParams
+
+cv::gpucodec::EncoderParams::EncoderParams()
+{
+    P_Interval = 3;
+    IDR_Period = 15;
+    DynamicGOP = 0;
+    RCType = 1;
+    AvgBitrate = 4000000;
+    PeakBitrate = 10000000;
+    QP_Level_Intra = 25;
+    QP_Level_InterP = 28;
+    QP_Level_InterB = 31;
+    DeblockMode = 1;
+    ProfileLevel = 65357;
+    ForceIntra = 0;
+    ForceIDR = 0;
+    ClearStat = 0;
+    DIMode = 1;
+    Presets = 2;
+    DisableCabac = 0;
+    NaluFramingType = 0;
+    DisableSPSPPS = 0;
+}
+
+cv::gpucodec::EncoderParams::EncoderParams(const String& configFile)
+{
+    load(configFile);
+}
+
+void cv::gpucodec::EncoderParams::load(const String& configFile)
+{
+    FileStorage fs(configFile, FileStorage::READ);
+    CV_Assert( fs.isOpened() );
+
+    read(fs["P_Interval"     ], P_Interval, 3);
+    read(fs["IDR_Period"     ], IDR_Period, 15);
+    read(fs["DynamicGOP"     ], DynamicGOP, 0);
+    read(fs["RCType"         ], RCType, 1);
+    read(fs["AvgBitrate"     ], AvgBitrate, 4000000);
+    read(fs["PeakBitrate"    ], PeakBitrate, 10000000);
+    read(fs["QP_Level_Intra" ], QP_Level_Intra, 25);
+    read(fs["QP_Level_InterP"], QP_Level_InterP, 28);
+    read(fs["QP_Level_InterB"], QP_Level_InterB, 31);
+    read(fs["DeblockMode"    ], DeblockMode, 1);
+    read(fs["ProfileLevel"   ], ProfileLevel, 65357);
+    read(fs["ForceIntra"     ], ForceIntra, 0);
+    read(fs["ForceIDR"       ], ForceIDR, 0);
+    read(fs["ClearStat"      ], ClearStat, 0);
+    read(fs["DIMode"         ], DIMode, 1);
+    read(fs["Presets"        ], Presets, 2);
+    read(fs["DisableCabac"   ], DisableCabac, 0);
+    read(fs["NaluFramingType"], NaluFramingType, 0);
+    read(fs["DisableSPSPPS"  ], DisableSPSPPS, 0);
+}
+
+void cv::gpucodec::EncoderParams::save(const String& configFile) const
+{
+    FileStorage fs(configFile, FileStorage::WRITE);
+    CV_Assert( fs.isOpened() );
+
+    write(fs, "P_Interval"     , P_Interval);
+    write(fs, "IDR_Period"     , IDR_Period);
+    write(fs, "DynamicGOP"     , DynamicGOP);
+    write(fs, "RCType"         , RCType);
+    write(fs, "AvgBitrate"     , AvgBitrate);
+    write(fs, "PeakBitrate"    , PeakBitrate);
+    write(fs, "QP_Level_Intra" , QP_Level_Intra);
+    write(fs, "QP_Level_InterP", QP_Level_InterP);
+    write(fs, "QP_Level_InterB", QP_Level_InterB);
+    write(fs, "DeblockMode"    , DeblockMode);
+    write(fs, "ProfileLevel"   , ProfileLevel);
+    write(fs, "ForceIntra"     , ForceIntra);
+    write(fs, "ForceIDR"       , ForceIDR);
+    write(fs, "ClearStat"      , ClearStat);
+    write(fs, "DIMode"         , DIMode);
+    write(fs, "Presets"        , Presets);
+    write(fs, "DisableCabac"   , DisableCabac);
+    write(fs, "NaluFramingType", NaluFramingType);
+    write(fs, "DisableSPSPPS"  , DisableSPSPPS);
+}
+
+///////////////////////////////////////////////////////////////////////////
+// createVideoWriter
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const String& fileName, Size frameSize, double fps, SurfaceFormat format)
+{
+    Ptr<EncoderCallBack> encoderCallback(new EncoderCallBackFFMPEG(fileName, frameSize, fps));
+    return createVideoWriter(encoderCallback, frameSize, fps, format);
+}
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const String& fileName, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
+{
+    Ptr<EncoderCallBack> encoderCallback(new EncoderCallBackFFMPEG(fileName, frameSize, fps));
+    return createVideoWriter(encoderCallback, frameSize, fps, params, format);
+}
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, SurfaceFormat format)
+{
+    return makePtr<VideoWriterImpl>(encoderCallback, frameSize, fps, format);
+}
+
+Ptr<VideoWriter> cv::gpucodec::createVideoWriter(const Ptr<EncoderCallBack>& encoderCallback, Size frameSize, double fps, const EncoderParams& params, SurfaceFormat format)
+{
+    return makePtr<VideoWriterImpl>(encoderCallback, frameSize, fps, params, format);
+}
+
+#endif // !defined HAVE_CUDA || !defined WIN32
diff --git a/modules/gpucodec/test/test_main.cpp b/modules/gpucodec/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpucodec/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpucodec/test/test_precomp.hpp b/modules/gpucodec/test/test_precomp.hpp
new file mode 100644
index 0000000000..95e70d46df
--- /dev/null
+++ b/modules/gpucodec/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpucodec.hpp"
+#include "opencv2/highgui.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpu/test/test_video.cpp b/modules/gpucodec/test/test_video.cpp
similarity index 62%
rename from modules/gpu/test/test_video.cpp
rename to modules/gpucodec/test/test_video.cpp
index b9502814a1..26bcc02d58 100644
--- a/modules/gpu/test/test_video.cpp
+++ b/modules/gpucodec/test/test_video.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -41,78 +42,50 @@
 
 #include "test_precomp.hpp"
 
-#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#ifdef HAVE_NVCUVID
+
+PARAM_TEST_CASE(Video, cv::gpu::DeviceInfo, std::string)
+{
+};
 
 //////////////////////////////////////////////////////
 // VideoReader
 
-PARAM_TEST_CASE(VideoReader, cv::gpu::DeviceInfo, std::string)
+GPU_TEST_P(Video, Reader)
 {
-    cv::gpu::DeviceInfo devInfo;
-    std::string inputFile;
+    cv::gpu::setDevice(GET_PARAM(0).deviceID());
 
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        inputFile = GET_PARAM(1);
+    const std::string inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + "video/" + GET_PARAM(1);
 
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + "video/" + inputFile;
-    }
-};
-
-GPU_TEST_P(VideoReader, Regression)
-{
-    cv::gpu::VideoReader_GPU reader(inputFile);
-    ASSERT_TRUE(reader.isOpened());
+    cv::Ptr<cv::gpucodec::VideoReader> reader = cv::gpucodec::createVideoReader(inputFile);
 
     cv::gpu::GpuMat frame;
 
     for (int i = 0; i < 10; ++i)
     {
-        ASSERT_TRUE(reader.read(frame));
+        ASSERT_TRUE(reader->nextFrame(frame));
         ASSERT_FALSE(frame.empty());
     }
-
-    reader.close();
-    ASSERT_FALSE(reader.isOpened());
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, VideoReader, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(std::string("768x576.avi"), std::string("1920x1080.avi"))));
-
 //////////////////////////////////////////////////////
 // VideoWriter
 
 #ifdef WIN32
 
-PARAM_TEST_CASE(VideoWriter, cv::gpu::DeviceInfo, std::string)
+GPU_TEST_P(Video, Writer)
 {
-    cv::gpu::DeviceInfo devInfo;
-    std::string inputFile;
+    cv::gpu::setDevice(GET_PARAM(0).deviceID());
 
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        inputFile = GET_PARAM(1);
+    const std::string inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + "video/" + GET_PARAM(1);
 
-        cv::gpu::setDevice(devInfo.deviceID());
-
-        inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + std::string("video/") + inputFile;
-    }
-};
-
-GPU_TEST_P(VideoWriter, Regression)
-{
     std::string outputFile = cv::tempfile(".avi");
     const double FPS = 25.0;
 
     cv::VideoCapture reader(inputFile);
     ASSERT_TRUE(reader.isOpened());
 
-    cv::gpu::VideoWriter_GPU d_writer;
+    cv::Ptr<cv::gpucodec::VideoWriter> d_writer;
 
     cv::Mat frame;
     cv::gpu::GpuMat d_frame;
@@ -124,14 +97,14 @@ GPU_TEST_P(VideoWriter, Regression)
 
         d_frame.upload(frame);
 
-        if (!d_writer.isOpened())
-            d_writer.open(outputFile, frame.size(), FPS);
+        if (d_writer.empty())
+            d_writer = cv::gpucodec::createVideoWriter(outputFile, frame.size(), FPS);
 
-        d_writer.write(d_frame);
+        d_writer->write(d_frame);
     }
 
     reader.release();
-    d_writer.close();
+    d_writer.release();
 
     reader.open(outputFile);
     ASSERT_TRUE(reader.isOpened());
@@ -143,10 +116,10 @@ GPU_TEST_P(VideoWriter, Regression)
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, VideoWriter, testing::Combine(
+#endif // WIN32
+
+INSTANTIATE_TEST_CASE_P(GPU_Codec, Video, testing::Combine(
     ALL_DEVICES,
     testing::Values(std::string("768x576.avi"), std::string("1920x1080.avi"))));
 
-#endif // WIN32
-
-#endif //  defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
+#endif // HAVE_NVCUVID
diff --git a/modules/gpufeatures2d/CMakeLists.txt b/modules/gpufeatures2d/CMakeLists.txt
new file mode 100644
index 0000000000..641f864cd4
--- /dev/null
+++ b/modules/gpufeatures2d/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpufeatures2d)
+endif()
+
+set(the_description "GPU-accelerated Feature Detection and Description")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4100 /wd4324 /wd4512 /wd4515 -Wundef -Wmissing-declarations -Wshadow -Wunused-parameter)
+
+ocv_define_module(gpufeatures2d opencv_features2d opencv_gpufilters opencv_gpuwarping)
diff --git a/modules/gpu/doc/feature_detection_and_description.rst b/modules/gpufeatures2d/doc/feature_detection_and_description.rst
similarity index 85%
rename from modules/gpu/doc/feature_detection_and_description.rst
rename to modules/gpufeatures2d/doc/feature_detection_and_description.rst
index 0c4caf77c4..8a0288e152 100644
--- a/modules/gpu/doc/feature_detection_and_description.rst
+++ b/modules/gpufeatures2d/doc/feature_detection_and_description.rst
@@ -5,109 +5,6 @@ Feature Detection and Description
 
 
 
-gpu::SURF_GPU
--------------
-.. ocv:class:: gpu::SURF_GPU
-
-Class used for extracting Speeded Up Robust Features (SURF) from an image. ::
-
-    class SURF_GPU
-    {
-    public:
-        enum KeypointLayout
-        {
-            X_ROW = 0,
-            Y_ROW,
-            LAPLACIAN_ROW,
-            OCTAVE_ROW,
-            SIZE_ROW,
-            ANGLE_ROW,
-            HESSIAN_ROW,
-            ROWS_COUNT
-        };
-
-        //! the default constructor
-        SURF_GPU();
-        //! the full constructor taking all the necessary parameters
-        explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
-             int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f);
-
-        //! returns the descriptor size in float's (64 or 128)
-        int descriptorSize() const;
-
-        //! upload host keypoints to device memory
-        void uploadKeypoints(const vector<KeyPoint>& keypoints,
-            GpuMat& keypointsGPU);
-        //! download keypoints from device to host memory
-        void downloadKeypoints(const GpuMat& keypointsGPU,
-            vector<KeyPoint>& keypoints);
-
-        //! download descriptors from device to host memory
-        void downloadDescriptors(const GpuMat& descriptorsGPU,
-            vector<float>& descriptors);
-
-        void operator()(const GpuMat& img, const GpuMat& mask,
-            GpuMat& keypoints);
-
-        void operator()(const GpuMat& img, const GpuMat& mask,
-            GpuMat& keypoints, GpuMat& descriptors,
-            bool useProvidedKeypoints = false,
-            bool calcOrientation = true);
-
-        void operator()(const GpuMat& img, const GpuMat& mask,
-            std::vector<KeyPoint>& keypoints);
-
-        void operator()(const GpuMat& img, const GpuMat& mask,
-            std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
-            bool useProvidedKeypoints = false,
-            bool calcOrientation = true);
-
-        void operator()(const GpuMat& img, const GpuMat& mask,
-            std::vector<KeyPoint>& keypoints,
-            std::vector<float>& descriptors,
-            bool useProvidedKeypoints = false,
-            bool calcOrientation = true);
-
-        void releaseMemory();
-
-        // SURF parameters
-        double hessianThreshold;
-        int nOctaves;
-        int nOctaveLayers;
-        bool extended;
-        bool upright;
-
-        //! max keypoints = keypointsRatio * img.size().area()
-        float keypointsRatio;
-
-        GpuMat sum, mask1, maskSum, intBuffer;
-
-        GpuMat det, trace;
-
-        GpuMat maxPosBuffer;
-    };
-
-
-The class ``SURF_GPU`` implements Speeded Up Robust Features descriptor. There is a fast multi-scale Hessian keypoint detector that can be used to find the keypoints (which is the default option). But the descriptors can also be computed for the user-specified keypoints. Only 8-bit grayscale images are supported.
-
-The class ``SURF_GPU`` can store results in the GPU and CPU memory. It provides functions to convert results between CPU and GPU version ( ``uploadKeypoints``, ``downloadKeypoints``, ``downloadDescriptors`` ). The format of CPU results is the same as ``SURF`` results. GPU results are stored in ``GpuMat``. The ``keypoints`` matrix is :math:`\texttt{nFeatures} \times 7` matrix with the ``CV_32FC1`` type.
-
-* ``keypoints.ptr<float>(X_ROW)[i]`` contains x coordinate of the i-th feature.
-* ``keypoints.ptr<float>(Y_ROW)[i]`` contains y coordinate of the i-th feature.
-* ``keypoints.ptr<float>(LAPLACIAN_ROW)[i]``  contains the laplacian sign of the i-th feature.
-* ``keypoints.ptr<float>(OCTAVE_ROW)[i]`` contains the octave of the i-th feature.
-* ``keypoints.ptr<float>(SIZE_ROW)[i]`` contains the size of the i-th feature.
-* ``keypoints.ptr<float>(ANGLE_ROW)[i]`` contain orientation of the i-th feature.
-* ``keypoints.ptr<float>(HESSIAN_ROW)[i]`` contains the response of the i-th feature.
-
-The ``descriptors`` matrix is :math:`\texttt{nFeatures} \times \texttt{descriptorSize}` matrix with the ``CV_32FC1`` type.
-
-The class ``SURF_GPU`` uses some buffers and provides access to it. All buffers can be safely released between function calls.
-
-.. seealso:: :ocv:class:`SURF`
-
-
-
 gpu::FAST_GPU
 -------------
 .. ocv:class:: gpu::FAST_GPU
diff --git a/modules/gpufeatures2d/doc/gpufeatures2d.rst b/modules/gpufeatures2d/doc/gpufeatures2d.rst
new file mode 100644
index 0000000000..5679b1d8b8
--- /dev/null
+++ b/modules/gpufeatures2d/doc/gpufeatures2d.rst
@@ -0,0 +1,8 @@
+****************************************************************
+gpufeatures2d. GPU-accelerated Feature Detection and Description
+****************************************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    feature_detection_and_description
diff --git a/modules/gpufeatures2d/include/opencv2/gpufeatures2d.hpp b/modules/gpufeatures2d/include/opencv2/gpufeatures2d.hpp
new file mode 100644
index 0000000000..cc73da9d9e
--- /dev/null
+++ b/modules/gpufeatures2d/include/opencv2/gpufeatures2d.hpp
@@ -0,0 +1,361 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUFEATURES2D_HPP__
+#define __OPENCV_GPUFEATURES2D_HPP__
+
+#ifndef __cplusplus
+#  error gpufeatures2d.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/gpufilters.hpp"
+
+namespace cv { namespace gpu {
+
+class CV_EXPORTS BFMatcher_GPU
+{
+public:
+    explicit BFMatcher_GPU(int norm = cv::NORM_L2);
+
+    // Add descriptors to train descriptor collection
+    void add(const std::vector<GpuMat>& descCollection);
+
+    // Get train descriptors collection
+    const std::vector<GpuMat>& getTrainDescriptors() const;
+
+    // Clear train descriptors collection
+    void clear();
+
+    // Return true if there are not train descriptors in collection
+    bool empty() const;
+
+    // Return true if the matcher supports mask in match methods
+    bool isMaskSupported() const;
+
+    // Find one best match for each query descriptor
+    void matchSingle(const GpuMat& query, const GpuMat& train,
+        GpuMat& trainIdx, GpuMat& distance,
+        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
+
+    // Download trainIdx and distance and convert it to CPU vector with DMatch
+    static void matchDownload(const GpuMat& trainIdx, const GpuMat& distance, std::vector<DMatch>& matches);
+    // Convert trainIdx and distance to vector with DMatch
+    static void matchConvert(const Mat& trainIdx, const Mat& distance, std::vector<DMatch>& matches);
+
+    // Find one best match for each query descriptor
+    void match(const GpuMat& query, const GpuMat& train, std::vector<DMatch>& matches, const GpuMat& mask = GpuMat());
+
+    // Make gpu collection of trains and masks in suitable format for matchCollection function
+    void makeGpuCollection(GpuMat& trainCollection, GpuMat& maskCollection, const std::vector<GpuMat>& masks = std::vector<GpuMat>());
+
+    // Find one best match from train collection for each query descriptor
+    void matchCollection(const GpuMat& query, const GpuMat& trainCollection,
+        GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance,
+        const GpuMat& masks = GpuMat(), Stream& stream = Stream::Null());
+
+    // Download trainIdx, imgIdx and distance and convert it to vector with DMatch
+    static void matchDownload(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance, std::vector<DMatch>& matches);
+    // Convert trainIdx, imgIdx and distance to vector with DMatch
+    static void matchConvert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance, std::vector<DMatch>& matches);
+
+    // Find one best match from train collection for each query descriptor.
+    void match(const GpuMat& query, std::vector<DMatch>& matches, const std::vector<GpuMat>& masks = std::vector<GpuMat>());
+
+    // Find k best matches for each query descriptor (in increasing order of distances)
+    void knnMatchSingle(const GpuMat& query, const GpuMat& train,
+        GpuMat& trainIdx, GpuMat& distance, GpuMat& allDist, int k,
+        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
+
+    // Download trainIdx and distance and convert it to vector with DMatch
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    static void knnMatchDownload(const GpuMat& trainIdx, const GpuMat& distance,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+    // Convert trainIdx and distance to vector with DMatch
+    static void knnMatchConvert(const Mat& trainIdx, const Mat& distance,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+
+    // Find k best matches for each query descriptor (in increasing order of distances).
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    void knnMatch(const GpuMat& query, const GpuMat& train,
+        std::vector< std::vector<DMatch> >& matches, int k, const GpuMat& mask = GpuMat(),
+        bool compactResult = false);
+
+    // Find k best matches from train collection for each query descriptor (in increasing order of distances)
+    void knnMatch2Collection(const GpuMat& query, const GpuMat& trainCollection,
+        GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance,
+        const GpuMat& maskCollection = GpuMat(), Stream& stream = Stream::Null());
+
+    // Download trainIdx and distance and convert it to vector with DMatch
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    static void knnMatch2Download(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+    // Convert trainIdx and distance to vector with DMatch
+    static void knnMatch2Convert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+
+    // Find k best matches  for each query descriptor (in increasing order of distances).
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    void knnMatch(const GpuMat& query, std::vector< std::vector<DMatch> >& matches, int k,
+        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false);
+
+    // Find best matches for each query descriptor which have distance less than maxDistance.
+    // nMatches.at<int>(0, queryIdx) will contain matches count for queryIdx.
+    // carefully nMatches can be greater than trainIdx.cols - it means that matcher didn't find all matches,
+    // because it didn't have enough memory.
+    // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nTrain / 100), 10),
+    // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
+    // Matches doesn't sorted.
+    void radiusMatchSingle(const GpuMat& query, const GpuMat& train,
+        GpuMat& trainIdx, GpuMat& distance, GpuMat& nMatches, float maxDistance,
+        const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
+
+    // Download trainIdx, nMatches and distance and convert it to vector with DMatch.
+    // matches will be sorted in increasing order of distances.
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    static void radiusMatchDownload(const GpuMat& trainIdx, const GpuMat& distance, const GpuMat& nMatches,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+    // Convert trainIdx, nMatches and distance to vector with DMatch.
+    static void radiusMatchConvert(const Mat& trainIdx, const Mat& distance, const Mat& nMatches,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+
+    // Find best matches for each query descriptor which have distance less than maxDistance
+    // in increasing order of distances).
+    void radiusMatch(const GpuMat& query, const GpuMat& train,
+        std::vector< std::vector<DMatch> >& matches, float maxDistance,
+        const GpuMat& mask = GpuMat(), bool compactResult = false);
+
+    // Find best matches for each query descriptor which have distance less than maxDistance.
+    // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nQuery / 100), 10),
+    // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
+    // Matches doesn't sorted.
+    void radiusMatchCollection(const GpuMat& query, GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance, GpuMat& nMatches, float maxDistance,
+        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), Stream& stream = Stream::Null());
+
+    // Download trainIdx, imgIdx, nMatches and distance and convert it to vector with DMatch.
+    // matches will be sorted in increasing order of distances.
+    // compactResult is used when mask is not empty. If compactResult is false matches
+    // vector will have the same size as queryDescriptors rows. If compactResult is true
+    // matches vector will not contain matches for fully masked out query descriptors.
+    static void radiusMatchDownload(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance, const GpuMat& nMatches,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+    // Convert trainIdx, nMatches and distance to vector with DMatch.
+    static void radiusMatchConvert(const Mat& trainIdx, const Mat& imgIdx, const Mat& distance, const Mat& nMatches,
+        std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
+
+    // Find best matches from train collection for each query descriptor which have distance less than
+    // maxDistance (in increasing order of distances).
+    void radiusMatch(const GpuMat& query, std::vector< std::vector<DMatch> >& matches, float maxDistance,
+        const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false);
+
+    int norm;
+
+private:
+    std::vector<GpuMat> trainDescCollection;
+};
+
+class CV_EXPORTS FAST_GPU
+{
+public:
+    enum
+    {
+        LOCATION_ROW = 0,
+        RESPONSE_ROW,
+        ROWS_COUNT
+    };
+
+    // all features have same size
+    static const int FEATURE_SIZE = 7;
+
+    explicit FAST_GPU(int threshold, bool nonmaxSupression = true, double keypointsRatio = 0.05);
+
+    //! finds the keypoints using FAST detector
+    //! supports only CV_8UC1 images
+    void operator ()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints);
+    void operator ()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
+
+    //! download keypoints from device to host memory
+    static void downloadKeypoints(const GpuMat& d_keypoints, std::vector<KeyPoint>& keypoints);
+
+    //! convert keypoints to KeyPoint vector
+    static void convertKeypoints(const Mat& h_keypoints, std::vector<KeyPoint>& keypoints);
+
+    //! release temporary buffer's memory
+    void release();
+
+    bool nonmaxSupression;
+
+    int threshold;
+
+    //! max keypoints = keypointsRatio * img.size().area()
+    double keypointsRatio;
+
+    //! find keypoints and compute it's response if nonmaxSupression is true
+    //! return count of detected keypoints
+    int calcKeyPointsLocation(const GpuMat& image, const GpuMat& mask);
+
+    //! get final array of keypoints
+    //! performs nonmax supression if needed
+    //! return final count of keypoints
+    int getKeyPoints(GpuMat& keypoints);
+
+private:
+    GpuMat kpLoc_;
+    int count_;
+
+    GpuMat score_;
+
+    GpuMat d_keypoints_;
+};
+
+class CV_EXPORTS ORB_GPU
+{
+public:
+    enum
+    {
+        X_ROW = 0,
+        Y_ROW,
+        RESPONSE_ROW,
+        ANGLE_ROW,
+        OCTAVE_ROW,
+        SIZE_ROW,
+        ROWS_COUNT
+    };
+
+    enum
+    {
+        DEFAULT_FAST_THRESHOLD = 20
+    };
+
+    //! Constructor
+    explicit ORB_GPU(int nFeatures = 500, float scaleFactor = 1.2f, int nLevels = 8, int edgeThreshold = 31,
+                     int firstLevel = 0, int WTA_K = 2, int scoreType = 0, int patchSize = 31);
+
+    //! Compute the ORB features on an image
+    //! image - the image to compute the features (supports only CV_8UC1 images)
+    //! mask - the mask to apply
+    //! keypoints - the resulting keypoints
+    void operator()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
+    void operator()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints);
+
+    //! Compute the ORB features and descriptors on an image
+    //! image - the image to compute the features (supports only CV_8UC1 images)
+    //! mask - the mask to apply
+    //! keypoints - the resulting keypoints
+    //! descriptors - descriptors array
+    void operator()(const GpuMat& image, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors);
+    void operator()(const GpuMat& image, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors);
+
+    //! download keypoints from device to host memory
+    static void downloadKeyPoints(const GpuMat& d_keypoints, std::vector<KeyPoint>& keypoints);
+    //! convert keypoints to KeyPoint vector
+    static void convertKeyPoints(const Mat& d_keypoints, std::vector<KeyPoint>& keypoints);
+
+    //! returns the descriptor size in bytes
+    inline int descriptorSize() const { return kBytes; }
+
+    inline void setFastParams(int threshold, bool nonmaxSupression = true)
+    {
+        fastDetector_.threshold = threshold;
+        fastDetector_.nonmaxSupression = nonmaxSupression;
+    }
+
+    //! release temporary buffer's memory
+    void release();
+
+    //! if true, image will be blurred before descriptors calculation
+    bool blurForDescriptor;
+
+private:
+    enum { kBytes = 32 };
+
+    void buildScalePyramids(const GpuMat& image, const GpuMat& mask);
+
+    void computeKeyPointsPyramid();
+
+    void computeDescriptors(GpuMat& descriptors);
+
+    void mergeKeyPoints(GpuMat& keypoints);
+
+    int nFeatures_;
+    float scaleFactor_;
+    int nLevels_;
+    int edgeThreshold_;
+    int firstLevel_;
+    int WTA_K_;
+    int scoreType_;
+    int patchSize_;
+
+    // The number of desired features per scale
+    std::vector<size_t> n_features_per_level_;
+
+    // Points to compute BRIEF descriptors from
+    GpuMat pattern_;
+
+    std::vector<GpuMat> imagePyr_;
+    std::vector<GpuMat> maskPyr_;
+
+    GpuMat buf_;
+
+    std::vector<GpuMat> keyPointsPyr_;
+    std::vector<int> keyPointsCount_;
+
+    FAST_GPU fastDetector_;
+
+    Ptr<gpu::Filter> blurFilter;
+
+    GpuMat d_keypoints_;
+};
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUFEATURES2D_HPP__ */
diff --git a/modules/gpu/perf/perf_features2d.cpp b/modules/gpufeatures2d/perf/perf_features2d.cpp
similarity index 69%
rename from modules/gpu/perf/perf_features2d.cpp
rename to modules/gpufeatures2d/perf/perf_features2d.cpp
index 480f582383..fd28526335 100644
--- a/modules/gpu/perf/perf_features2d.cpp
+++ b/modules/gpufeatures2d/perf/perf_features2d.cpp
@@ -1,113 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
-
-struct KeypointIdxCompare
-{
-    std::vector<cv::KeyPoint>* keypoints;
-
-    explicit KeypointIdxCompare(std::vector<cv::KeyPoint>* _keypoints) : keypoints(_keypoints) {}
-
-    bool operator ()(size_t i1, size_t i2) const
-    {
-        cv::KeyPoint kp1 = (*keypoints)[i1];
-        cv::KeyPoint kp2 = (*keypoints)[i2];
-        if (kp1.pt.x != kp2.pt.x)
-            return kp1.pt.x < kp2.pt.x;
-        if (kp1.pt.y != kp2.pt.y)
-            return kp1.pt.y < kp2.pt.y;
-        if (kp1.response != kp2.response)
-            return kp1.response < kp2.response;
-        return kp1.octave < kp2.octave;
-    }
-};
-
-static void sortKeyPoints(std::vector<cv::KeyPoint>& keypoints, cv::InputOutputArray _descriptors = cv::noArray())
-{
-    std::vector<size_t> indexies(keypoints.size());
-    for (size_t i = 0; i < indexies.size(); ++i)
-        indexies[i] = i;
-
-    std::sort(indexies.begin(), indexies.end(), KeypointIdxCompare(&keypoints));
-
-    std::vector<cv::KeyPoint> new_keypoints;
-    cv::Mat new_descriptors;
-
-    new_keypoints.resize(keypoints.size());
-
-    cv::Mat descriptors;
-    if (_descriptors.needed())
-    {
-        descriptors = _descriptors.getMat();
-        new_descriptors.create(descriptors.size(), descriptors.type());
-    }
-
-    for (size_t i = 0; i < indexies.size(); ++i)
-    {
-        size_t new_idx = indexies[i];
-        new_keypoints[i] = keypoints[new_idx];
-        if (!new_descriptors.empty())
-            descriptors.row((int) new_idx).copyTo(new_descriptors.row((int) i));
-    }
-
-    keypoints.swap(new_keypoints);
-    if (_descriptors.needed())
-        new_descriptors.copyTo(_descriptors);
-}
-
-//////////////////////////////////////////////////////////////////////
-// SURF
-
-DEF_PARAM_TEST_1(Image, string);
-
-PERF_TEST_P(Image, Features2D_SURF,
-            Values<string>("gpu/perf/aloe.png"))
-{
-    declare.time(50.0);
-
-    const cv::Mat img = readImage(GetParam(), cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    if (PERF_RUN_GPU())
-    {
-        cv::gpu::SURF_GPU d_surf;
-
-        const cv::gpu::GpuMat d_img(img);
-        cv::gpu::GpuMat d_keypoints, d_descriptors;
-
-        TEST_CYCLE() d_surf(d_img, cv::gpu::GpuMat(), d_keypoints, d_descriptors);
-
-        std::vector<cv::KeyPoint> gpu_keypoints;
-        d_surf.downloadKeypoints(d_keypoints, gpu_keypoints);
-
-        cv::Mat gpu_descriptors(d_descriptors);
-
-        sortKeyPoints(gpu_keypoints, gpu_descriptors);
-
-        SANITY_CHECK_KEYPOINTS(gpu_keypoints);
-        SANITY_CHECK(gpu_descriptors, 1e-3);
-    }
-    else
-    {
-        cv::SURF surf;
-
-        std::vector<cv::KeyPoint> cpu_keypoints;
-        cv::Mat cpu_descriptors;
-
-        TEST_CYCLE() surf(img, cv::noArray(), cpu_keypoints, cpu_descriptors);
-
-        SANITY_CHECK_KEYPOINTS(cpu_keypoints);
-        SANITY_CHECK(cpu_descriptors);
-    }
-}
+using namespace perf;
 
 //////////////////////////////////////////////////////////////////////
 // FAST
 
 DEF_PARAM_TEST(Image_Threshold_NonMaxSupression, string, int, bool);
 
-PERF_TEST_P(Image_Threshold_NonMaxSupression, Features2D_FAST,
+PERF_TEST_P(Image_Threshold_NonMaxSupression, FAST,
             Combine(Values<string>("gpu/perf/aloe.png"),
                     Values(20),
                     Bool()))
@@ -149,10 +93,12 @@ PERF_TEST_P(Image_Threshold_NonMaxSupression, Features2D_FAST,
 
 DEF_PARAM_TEST(Image_NFeatures, string, int);
 
-PERF_TEST_P(Image_NFeatures, Features2D_ORB,
+PERF_TEST_P(Image_NFeatures, ORB,
             Combine(Values<string>("gpu/perf/aloe.png"),
                     Values(4000)))
 {
+    declare.time(300.0);
+
     const cv::Mat img = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
     ASSERT_FALSE(img.empty());
 
@@ -177,7 +123,7 @@ PERF_TEST_P(Image_NFeatures, Features2D_ORB,
 
         sortKeyPoints(gpu_keypoints, gpu_descriptors);
 
-        SANITY_CHECK_KEYPOINTS(gpu_keypoints);
+        SANITY_CHECK_KEYPOINTS(gpu_keypoints, 1e-4);
         SANITY_CHECK(gpu_descriptors);
     }
     else
@@ -199,7 +145,7 @@ PERF_TEST_P(Image_NFeatures, Features2D_ORB,
 
 DEF_PARAM_TEST(DescSize_Norm, int, NormType);
 
-PERF_TEST_P(DescSize_Norm, Features2D_BFMatch,
+PERF_TEST_P(DescSize_Norm, BFMatch,
             Combine(Values(64, 128, 256),
                     Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2), NormType(cv::NORM_HAMMING))))
 {
@@ -256,7 +202,7 @@ static void toOneRowMatches(const std::vector< std::vector<cv::DMatch> >& src, s
 
 DEF_PARAM_TEST(DescSize_K_Norm, int, int, NormType);
 
-PERF_TEST_P(DescSize_K_Norm, Features2D_BFKnnMatch,
+PERF_TEST_P(DescSize_K_Norm, BFKnnMatch,
             Combine(Values(64, 128, 256),
                     Values(2, 3),
                     Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
@@ -311,7 +257,7 @@ PERF_TEST_P(DescSize_K_Norm, Features2D_BFKnnMatch,
 //////////////////////////////////////////////////////////////////////
 // BFRadiusMatch
 
-PERF_TEST_P(DescSize_Norm, Features2D_BFRadiusMatch,
+PERF_TEST_P(DescSize_Norm, BFRadiusMatch,
             Combine(Values(64, 128, 256),
                     Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2))))
 {
diff --git a/modules/gpufeatures2d/perf/perf_main.cpp b/modules/gpufeatures2d/perf/perf_main.cpp
new file mode 100644
index 0000000000..5e7fb1b8be
--- /dev/null
+++ b/modules/gpufeatures2d/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpufeatures2d)
diff --git a/modules/gpufeatures2d/perf/perf_precomp.hpp b/modules/gpufeatures2d/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..4f767c4fb3
--- /dev/null
+++ b/modules/gpufeatures2d/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpufeatures2d.hpp"
+#include "opencv2/features2d.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpu/src/brute_force_matcher.cpp b/modules/gpufeatures2d/src/brute_force_matcher.cpp
similarity index 94%
rename from modules/gpu/src/brute_force_matcher.cpp
rename to modules/gpufeatures2d/src/brute_force_matcher.cpp
index 4276f75a9e..feb0cc6928 100644
--- a/modules/gpu/src/brute_force_matcher.cpp
+++ b/modules/gpufeatures2d/src/brute_force_matcher.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,41 +47,41 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::BFMatcher_GPU::BFMatcher_GPU(int) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::add(const std::vector<GpuMat>&) { throw_nogpu(); }
-const std::vector<GpuMat>& cv::gpu::BFMatcher_GPU::getTrainDescriptors() const { throw_nogpu(); return trainDescCollection; }
-void cv::gpu::BFMatcher_GPU::clear() { throw_nogpu(); }
-bool cv::gpu::BFMatcher_GPU::empty() const { throw_nogpu(); return true; }
-bool cv::gpu::BFMatcher_GPU::isMaskSupported() const { throw_nogpu(); return true; }
-void cv::gpu::BFMatcher_GPU::matchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::matchDownload(const GpuMat&, const GpuMat&, std::vector<DMatch>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::matchConvert(const Mat&, const Mat&, std::vector<DMatch>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::match(const GpuMat&, const GpuMat&, std::vector<DMatch>&, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::makeGpuCollection(GpuMat&, GpuMat&, const std::vector<GpuMat>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::matchCollection(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::matchDownload(const GpuMat&, const GpuMat&, const GpuMat&, std::vector<DMatch>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::matchConvert(const Mat&, const Mat&, const Mat&, std::vector<DMatch>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::match(const GpuMat&, std::vector<DMatch>&, const std::vector<GpuMat>&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatchDownload(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatchConvert(const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatch(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, int, const GpuMat&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatch2Collection(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatch2Download(const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatch2Convert(const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::knnMatch(const GpuMat&, std::vector< std::vector<DMatch> >&, int, const std::vector<GpuMat>&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, float, const GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchDownload(const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchConvert(const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatch(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, float, const GpuMat&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchCollection(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, float, const std::vector<GpuMat>&, Stream&) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchDownload(const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatchConvert(const Mat&, const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_nogpu(); }
-void cv::gpu::BFMatcher_GPU::radiusMatch(const GpuMat&, std::vector< std::vector<DMatch> >&, float, const std::vector<GpuMat>&, bool) { throw_nogpu(); }
+cv::gpu::BFMatcher_GPU::BFMatcher_GPU(int) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::add(const std::vector<GpuMat>&) { throw_no_cuda(); }
+const std::vector<GpuMat>& cv::gpu::BFMatcher_GPU::getTrainDescriptors() const { throw_no_cuda(); return trainDescCollection; }
+void cv::gpu::BFMatcher_GPU::clear() { throw_no_cuda(); }
+bool cv::gpu::BFMatcher_GPU::empty() const { throw_no_cuda(); return true; }
+bool cv::gpu::BFMatcher_GPU::isMaskSupported() const { throw_no_cuda(); return true; }
+void cv::gpu::BFMatcher_GPU::matchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::matchDownload(const GpuMat&, const GpuMat&, std::vector<DMatch>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::matchConvert(const Mat&, const Mat&, std::vector<DMatch>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::match(const GpuMat&, const GpuMat&, std::vector<DMatch>&, const GpuMat&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::makeGpuCollection(GpuMat&, GpuMat&, const std::vector<GpuMat>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::matchCollection(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::matchDownload(const GpuMat&, const GpuMat&, const GpuMat&, std::vector<DMatch>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::matchConvert(const Mat&, const Mat&, const Mat&, std::vector<DMatch>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::match(const GpuMat&, std::vector<DMatch>&, const std::vector<GpuMat>&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, const GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatchDownload(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatchConvert(const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatch(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, int, const GpuMat&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatch2Collection(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatch2Download(const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatch2Convert(const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::knnMatch(const GpuMat&, std::vector< std::vector<DMatch> >&, int, const std::vector<GpuMat>&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchSingle(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat&, float, const GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchDownload(const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchConvert(const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatch(const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, float, const GpuMat&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchCollection(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, float, const std::vector<GpuMat>&, Stream&) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchDownload(const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatchConvert(const Mat&, const Mat&, const Mat&, const Mat&, std::vector< std::vector<DMatch> >&, bool) { throw_no_cuda(); }
+void cv::gpu::BFMatcher_GPU::radiusMatch(const GpuMat&, std::vector< std::vector<DMatch> >&, float, const std::vector<GpuMat>&, bool) { throw_no_cuda(); }
 
 #else /* !defined (HAVE_CUDA) */
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace bf_match
     {
@@ -197,7 +197,7 @@ void cv::gpu::BFMatcher_GPU::matchSingle(const GpuMat& query, const GpuMat& trai
     if (query.empty() || train.empty())
         return;
 
-    using namespace cv::gpu::device::bf_match;
+    using namespace cv::gpu::cudev::bf_match;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb& train, const PtrStepSzb& mask,
                              const PtrStepSzi& trainIdx, const PtrStepSzf& distance,
@@ -340,7 +340,7 @@ void cv::gpu::BFMatcher_GPU::matchCollection(const GpuMat& query, const GpuMat&
     if (query.empty() || trainCollection.empty())
         return;
 
-    using namespace cv::gpu::device::bf_match;
+    using namespace cv::gpu::cudev::bf_match;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks,
                              const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance,
@@ -451,7 +451,7 @@ void cv::gpu::BFMatcher_GPU::knnMatchSingle(const GpuMat& query, const GpuMat& t
     if (query.empty() || train.empty())
         return;
 
-    using namespace cv::gpu::device::bf_knnmatch;
+    using namespace cv::gpu::cudev::bf_knnmatch;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb& train, int k, const PtrStepSzb& mask,
                              const PtrStepSzb& trainIdx, const PtrStepSzb& distance, const PtrStepSzf& allDist,
@@ -497,10 +497,7 @@ void cv::gpu::BFMatcher_GPU::knnMatchSingle(const GpuMat& query, const GpuMat& t
         ensureSizeIsEnough(nQuery, nTrain, CV_32FC1, allDist);
     }
 
-    if (stream)
-        stream.enqueueMemSet(trainIdx, Scalar::all(-1));
-    else
-        trainIdx.setTo(Scalar::all(-1));
+    trainIdx.setTo(Scalar::all(-1), stream);
 
     caller_t func = callers[query.depth()];
     CV_Assert(func != 0);
@@ -580,7 +577,7 @@ void cv::gpu::BFMatcher_GPU::knnMatch2Collection(const GpuMat& query, const GpuM
     if (query.empty() || trainCollection.empty())
         return;
 
-    using namespace cv::gpu::device::bf_knnmatch;
+    using namespace cv::gpu::cudev::bf_knnmatch;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks,
                              const PtrStepSzb& trainIdx, const PtrStepSzb& imgIdx, const PtrStepSzb& distance,
@@ -616,10 +613,7 @@ void cv::gpu::BFMatcher_GPU::knnMatch2Collection(const GpuMat& query, const GpuM
     ensureSizeIsEnough(1, nQuery, CV_32SC2, imgIdx);
     ensureSizeIsEnough(1, nQuery, CV_32FC2, distance);
 
-    if (stream)
-        stream.enqueueMemSet(trainIdx, Scalar::all(-1));
-    else
-        trainIdx.setTo(Scalar::all(-1));
+    trainIdx.setTo(Scalar::all(-1), stream);
 
     caller_t func = callers[query.depth()];
     CV_Assert(func != 0);
@@ -761,7 +755,7 @@ void cv::gpu::BFMatcher_GPU::radiusMatchSingle(const GpuMat& query, const GpuMat
     if (query.empty() || train.empty())
         return;
 
-    using namespace cv::gpu::device::bf_radius_match;
+    using namespace cv::gpu::cudev::bf_radius_match;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb& train, float maxDistance, const PtrStepSzb& mask,
                              const PtrStepSzi& trainIdx, const PtrStepSzf& distance, const PtrStepSz<unsigned int>& nMatches,
@@ -803,10 +797,7 @@ void cv::gpu::BFMatcher_GPU::radiusMatchSingle(const GpuMat& query, const GpuMat
         ensureSizeIsEnough(nQuery, std::max((nTrain / 100), 10), CV_32FC1, distance);
     }
 
-    if (stream)
-        stream.enqueueMemSet(nMatches, Scalar::all(0));
-    else
-        nMatches.setTo(Scalar::all(0));
+    nMatches.setTo(Scalar::all(0), stream);
 
     caller_t func = callers[query.depth()];
     CV_Assert(func != 0);
@@ -890,7 +881,7 @@ void cv::gpu::BFMatcher_GPU::radiusMatchCollection(const GpuMat& query, GpuMat&
     if (query.empty() || empty())
         return;
 
-    using namespace cv::gpu::device::bf_radius_match;
+    using namespace cv::gpu::cudev::bf_radius_match;
 
     typedef void (*caller_t)(const PtrStepSzb& query, const PtrStepSzb* trains, int n, float maxDistance, const PtrStepSzb* masks,
                              const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance, const PtrStepSz<unsigned int>& nMatches,
@@ -931,10 +922,7 @@ void cv::gpu::BFMatcher_GPU::radiusMatchCollection(const GpuMat& query, GpuMat&
         ensureSizeIsEnough(nQuery, std::max((nQuery / 100), 10), CV_32FC1, distance);
     }
 
-    if (stream)
-        stream.enqueueMemSet(nMatches, Scalar::all(0));
-    else
-        nMatches.setTo(Scalar::all(0));
+    nMatches.setTo(Scalar::all(0), stream);
 
     caller_t func = callers[query.depth()];
     CV_Assert(func != 0);
diff --git a/modules/gpu/src/cuda/bf_knnmatch.cu b/modules/gpufeatures2d/src/cuda/bf_knnmatch.cu
similarity index 99%
rename from modules/gpu/src/cuda/bf_knnmatch.cu
rename to modules/gpufeatures2d/src/cuda/bf_knnmatch.cu
index 49bc1dfcd2..d5d17bb8a0 100644
--- a/modules/gpu/src/cuda/bf_knnmatch.cu
+++ b/modules/gpufeatures2d/src/cuda/bf_knnmatch.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,15 +42,15 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/vec_distance.hpp"
-#include "opencv2/gpu/device/datamov_utils.hpp"
-#include "opencv2/gpu/device/warp_shuffle.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/vec_distance.hpp"
+#include "opencv2/core/cuda/datamov_utils.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace bf_knnmatch
     {
@@ -1249,7 +1249,7 @@ namespace cv { namespace gpu { namespace device
         //template void match2Hamming_gpu<short >(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks, const PtrStepSzb& trainIdx, const PtrStepSzb& imgIdx, const PtrStepSzb& distance, cudaStream_t stream);
         template void match2Hamming_gpu<int   >(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks, const PtrStepSzb& trainIdx, const PtrStepSzb& imgIdx, const PtrStepSzb& distance, cudaStream_t stream);
     } // namespace bf_knnmatch
-}}} // namespace cv { namespace gpu { namespace device {
+}}} // namespace cv { namespace gpu { namespace cudev {
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/bf_match.cu b/modules/gpufeatures2d/src/cuda/bf_match.cu
similarity index 98%
rename from modules/gpu/src/cuda/bf_match.cu
rename to modules/gpufeatures2d/src/cuda/bf_match.cu
index 5e64e31bd9..338fefcb69 100644
--- a/modules/gpu/src/cuda/bf_match.cu
+++ b/modules/gpufeatures2d/src/cuda/bf_match.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,14 +42,14 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/vec_distance.hpp"
-#include "opencv2/gpu/device/datamov_utils.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/vec_distance.hpp"
+#include "opencv2/core/cuda/datamov_utils.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace bf_match
     {
@@ -768,7 +768,7 @@ namespace cv { namespace gpu { namespace device
         //template void matchHamming_gpu<short >(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks, const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance, cudaStream_t stream);
         template void matchHamming_gpu<int   >(const PtrStepSzb& query, const PtrStepSzb& trains, const PtrStepSz<PtrStepb>& masks, const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance, cudaStream_t stream);
     } // namespace bf_match
-}}} // namespace cv { namespace gpu { namespace device {
+}}} // namespace cv { namespace gpu { namespace cudev {
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/bf_radius_match.cu b/modules/gpufeatures2d/src/cuda/bf_radius_match.cu
similarity index 98%
rename from modules/gpu/src/cuda/bf_radius_match.cu
rename to modules/gpufeatures2d/src/cuda/bf_radius_match.cu
index 19ee94e331..3c714d63f9 100644
--- a/modules/gpu/src/cuda/bf_radius_match.cu
+++ b/modules/gpufeatures2d/src/cuda/bf_radius_match.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,13 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/vec_distance.hpp"
-#include "opencv2/gpu/device/datamov_utils.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/vec_distance.hpp"
+#include "opencv2/core/cuda/datamov_utils.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace bf_radius_match
     {
@@ -457,7 +457,7 @@ namespace cv { namespace gpu { namespace device
         //template void matchHamming_gpu<short >(const PtrStepSzb& query, const PtrStepSzb* trains, int n, float maxDistance, const PtrStepSzb* masks, const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance, const PtrStepSz<unsigned int>& nMatches, cudaStream_t stream);
         template void matchHamming_gpu<int   >(const PtrStepSzb& query, const PtrStepSzb* trains, int n, float maxDistance, const PtrStepSzb* masks, const PtrStepSzi& trainIdx, const PtrStepSzi& imgIdx, const PtrStepSzf& distance, const PtrStepSz<unsigned int>& nMatches, cudaStream_t stream);
     } // namespace bf_radius_match
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/fast.cu b/modules/gpufeatures2d/src/cuda/fast.cu
similarity index 99%
rename from modules/gpu/src/cuda/fast.cu
rename to modules/gpufeatures2d/src/cuda/fast.cu
index 831cb92b60..39d66d0eda 100644
--- a/modules/gpu/src/cuda/fast.cu
+++ b/modules/gpufeatures2d/src/cuda/fast.cu
@@ -38,19 +38,14 @@
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
-// Copyright (c) 2010, Paul Furgale, Chi Hay Tong
-//
-// The original code was written by Paul Furgale and Chi Hay Tong
-// and later optimized and prepared for integration into OpenCV by Itseez.
-//
 //M*/
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace fast
     {
@@ -387,4 +382,4 @@ namespace cv { namespace gpu { namespace device
 }}}
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/orb.cu b/modules/gpufeatures2d/src/cuda/orb.cu
similarity index 96%
rename from modules/gpu/src/cuda/orb.cu
rename to modules/gpufeatures2d/src/cuda/orb.cu
index 95706dfa39..571ca12bde 100644
--- a/modules/gpu/src/cuda/orb.cu
+++ b/modules/gpufeatures2d/src/cuda/orb.cu
@@ -38,11 +38,6 @@
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
-// Copyright (c) 2010, Paul Furgale, Chi Hay Tong
-//
-// The original code was written by Paul Furgale and Chi Hay Tong
-// and later optimized and prepared for integration into OpenCV by Itseez.
-//
 //M*/
 
 #if !defined CUDA_DISABLER
@@ -50,11 +45,11 @@
 #include <thrust/device_ptr.h>
 #include <thrust/sort.h>
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/functional.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/functional.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace orb
     {
@@ -202,8 +197,8 @@ namespace cv { namespace gpu { namespace device
                 if (threadIdx.x == 0)
                 {
                     float kp_dir = ::atan2f((float)m_01, (float)m_10);
-                    kp_dir += (kp_dir < 0) * (2.0f * CV_PI);
-                    kp_dir *= 180.0f / CV_PI;
+                    kp_dir += (kp_dir < 0) * (2.0f * CV_PI_F);
+                    kp_dir *= 180.0f / CV_PI_F;
 
                     angle[ptidx] = kp_dir;
                 }
@@ -354,7 +349,7 @@ namespace cv { namespace gpu { namespace device
             if (ptidx < npoints && descidx < dsize)
             {
                 float angle = angle_[ptidx];
-                angle *= (float)(CV_PI / 180.f);
+                angle *= (float)(CV_PI_F / 180.f);
 
                 float sina, cosa;
                 ::sincosf(angle, &sina, &cosa);
diff --git a/modules/gpu/src/fast.cpp b/modules/gpufeatures2d/src/fast.cpp
similarity index 89%
rename from modules/gpu/src/fast.cpp
rename to modules/gpufeatures2d/src/fast.cpp
index e44b9e1644..539423e75b 100644
--- a/modules/gpu/src/fast.cpp
+++ b/modules/gpufeatures2d/src/fast.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,14 +47,14 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::FAST_GPU::FAST_GPU(int, bool, double) { throw_nogpu(); }
-void cv::gpu::FAST_GPU::operator ()(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::FAST_GPU::operator ()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::FAST_GPU::downloadKeypoints(const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::FAST_GPU::convertKeypoints(const Mat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::FAST_GPU::release() { throw_nogpu(); }
-int cv::gpu::FAST_GPU::calcKeyPointsLocation(const GpuMat&, const GpuMat&) { throw_nogpu(); return 0; }
-int cv::gpu::FAST_GPU::getKeyPoints(GpuMat&) { throw_nogpu(); return 0; }
+cv::gpu::FAST_GPU::FAST_GPU(int, bool, double) { throw_no_cuda(); }
+void cv::gpu::FAST_GPU::operator ()(const GpuMat&, const GpuMat&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::FAST_GPU::operator ()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::FAST_GPU::downloadKeypoints(const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::FAST_GPU::convertKeypoints(const Mat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::FAST_GPU::release() { throw_no_cuda(); }
+int cv::gpu::FAST_GPU::calcKeyPointsLocation(const GpuMat&, const GpuMat&) { throw_no_cuda(); return 0; }
+int cv::gpu::FAST_GPU::getKeyPoints(GpuMat&) { throw_no_cuda(); return 0; }
 
 #else /* !defined (HAVE_CUDA) */
 
@@ -108,7 +108,7 @@ void cv::gpu::FAST_GPU::operator ()(const GpuMat& img, const GpuMat& mask, GpuMa
     keypoints.cols = getKeyPoints(keypoints);
 }
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace fast
     {
@@ -119,7 +119,7 @@ namespace cv { namespace gpu { namespace device
 
 int cv::gpu::FAST_GPU::calcKeyPointsLocation(const GpuMat& img, const GpuMat& mask)
 {
-    using namespace cv::gpu::device::fast;
+    using namespace cv::gpu::cudev::fast;
 
     CV_Assert(img.type() == CV_8UC1);
     CV_Assert(mask.empty() || (mask.type() == CV_8UC1 && mask.size() == img.size()));
@@ -142,7 +142,7 @@ int cv::gpu::FAST_GPU::calcKeyPointsLocation(const GpuMat& img, const GpuMat& ma
 
 int cv::gpu::FAST_GPU::getKeyPoints(GpuMat& keypoints)
 {
-    using namespace cv::gpu::device::fast;
+    using namespace cv::gpu::cudev::fast;
 
     if (count_ == 0)
         return 0;
diff --git a/modules/gpu/src/orb.cpp b/modules/gpufeatures2d/src/orb.cpp
similarity index 95%
rename from modules/gpu/src/orb.cpp
rename to modules/gpufeatures2d/src/orb.cpp
index 95ac048833..7cb1cbecc1 100644
--- a/modules/gpu/src/orb.cpp
+++ b/modules/gpufeatures2d/src/orb.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,22 +47,22 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::ORB_GPU::ORB_GPU(int, float, int, int, int, int, int, int) : fastDetector_(20) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::downloadKeyPoints(const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::convertKeyPoints(const Mat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::release() { throw_nogpu(); }
-void cv::gpu::ORB_GPU::buildScalePyramids(const GpuMat&, const GpuMat&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::computeKeyPointsPyramid() { throw_nogpu(); }
-void cv::gpu::ORB_GPU::computeDescriptors(GpuMat&) { throw_nogpu(); }
-void cv::gpu::ORB_GPU::mergeKeyPoints(GpuMat&) { throw_nogpu(); }
+cv::gpu::ORB_GPU::ORB_GPU(int, float, int, int, int, int, int, int) : fastDetector_(20) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::downloadKeyPoints(const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::convertKeyPoints(const Mat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::release() { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::buildScalePyramids(const GpuMat&, const GpuMat&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::computeKeyPointsPyramid() { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::computeDescriptors(GpuMat&) { throw_no_cuda(); }
+void cv::gpu::ORB_GPU::mergeKeyPoints(GpuMat&) { throw_no_cuda(); }
 
 #else /* !defined (HAVE_CUDA) */
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace orb
     {
@@ -431,7 +431,7 @@ cv::gpu::ORB_GPU::ORB_GPU(int nFeatures, float scaleFactor, int nLevels, int edg
         ++v_0;
     }
     CV_Assert(u_max.size() < 32);
-    cv::gpu::device::orb::loadUMax(&u_max[0], static_cast<int>(u_max.size()));
+    cv::gpu::cudev::orb::loadUMax(&u_max[0], static_cast<int>(u_max.size()));
 
     // Calc pattern
     const int npoints = 512;
@@ -468,7 +468,7 @@ cv::gpu::ORB_GPU::ORB_GPU(int nFeatures, float scaleFactor, int nLevels, int edg
 
     pattern_.upload(h_pattern);
 
-    blurFilter = createGaussianFilter_GPU(CV_8UC1, Size(7, 7), 2, 2, BORDER_REFLECT_101);
+    blurFilter = gpu::createGaussianFilter(CV_8UC1, -1, Size(7, 7), 2, 2, BORDER_REFLECT_101);
 
     blurForDescriptor = false;
 }
@@ -504,19 +504,19 @@ void cv::gpu::ORB_GPU::buildScalePyramids(const GpuMat& image, const GpuMat& mas
         {
             if (level < firstLevel_)
             {
-                resize(image, imagePyr_[level], sz, 0, 0, INTER_LINEAR);
+                gpu::resize(image, imagePyr_[level], sz, 0, 0, INTER_LINEAR);
 
                 if (!mask.empty())
-                    resize(mask, maskPyr_[level], sz, 0, 0, INTER_LINEAR);
+                    gpu::resize(mask, maskPyr_[level], sz, 0, 0, INTER_LINEAR);
             }
             else
             {
-                resize(imagePyr_[level - 1], imagePyr_[level], sz, 0, 0, INTER_LINEAR);
+                gpu::resize(imagePyr_[level - 1], imagePyr_[level], sz, 0, 0, INTER_LINEAR);
 
                 if (!mask.empty())
                 {
-                    resize(maskPyr_[level - 1], maskPyr_[level], sz, 0, 0, INTER_LINEAR);
-                    threshold(maskPyr_[level], maskPyr_[level], 254, 0, THRESH_TOZERO);
+                    gpu::resize(maskPyr_[level - 1], maskPyr_[level], sz, 0, 0, INTER_LINEAR);
+                    gpu::threshold(maskPyr_[level], maskPyr_[level], 254, 0, THRESH_TOZERO);
                 }
             }
         }
@@ -534,7 +534,7 @@ void cv::gpu::ORB_GPU::buildScalePyramids(const GpuMat& image, const GpuMat& mas
         Rect inner(edgeThreshold_, edgeThreshold_, sz.width - 2 * edgeThreshold_, sz.height - 2 * edgeThreshold_);
         buf_(inner).setTo(Scalar::all(255));
 
-        bitwise_and(maskPyr_[level], buf_, maskPyr_[level]);
+        gpu::bitwise_and(maskPyr_[level], buf_, maskPyr_[level]);
     }
 }
 
@@ -543,7 +543,7 @@ namespace
     //takes keypoints and culls them by the response
     void cull(GpuMat& keypoints, int& count, int n_points)
     {
-        using namespace cv::gpu::device::orb;
+        using namespace cv::gpu::cudev::orb;
 
         //this is only necessary if the keypoints size is greater than the number of desired points.
         if (count > n_points)
@@ -561,7 +561,7 @@ namespace
 
 void cv::gpu::ORB_GPU::computeKeyPointsPyramid()
 {
-    using namespace cv::gpu::device::orb;
+    using namespace cv::gpu::cudev::orb;
 
     int half_patch_size = patchSize_ / 2;
 
@@ -604,7 +604,7 @@ void cv::gpu::ORB_GPU::computeKeyPointsPyramid()
 
 void cv::gpu::ORB_GPU::computeDescriptors(GpuMat& descriptors)
 {
-    using namespace cv::gpu::device::orb;
+    using namespace cv::gpu::cudev::orb;
 
     int nAllkeypoints = 0;
 
@@ -632,7 +632,7 @@ void cv::gpu::ORB_GPU::computeDescriptors(GpuMat& descriptors)
         {
             // preprocess the resized image
             ensureSizeIsEnough(imagePyr_[level].size(), imagePyr_[level].type(), buf_);
-            blurFilter->apply(imagePyr_[level], buf_, Rect(0, 0, imagePyr_[level].cols, imagePyr_[level].rows));
+            blurFilter->apply(imagePyr_[level], buf_);
         }
 
         computeOrbDescriptor_gpu(blurForDescriptor ? buf_ : imagePyr_[level], keyPointsPyr_[level].ptr<short2>(0), keyPointsPyr_[level].ptr<float>(2),
@@ -644,7 +644,7 @@ void cv::gpu::ORB_GPU::computeDescriptors(GpuMat& descriptors)
 
 void cv::gpu::ORB_GPU::mergeKeyPoints(GpuMat& keypoints)
 {
-    using namespace cv::gpu::device::orb;
+    using namespace cv::gpu::cudev::orb;
 
     int nAllkeypoints = 0;
 
diff --git a/modules/gpufeatures2d/src/precomp.hpp b/modules/gpufeatures2d/src/precomp.hpp
new file mode 100644
index 0000000000..33c5aa7d3e
--- /dev/null
+++ b/modules/gpufeatures2d/src/precomp.hpp
@@ -0,0 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <algorithm>
+#include <functional>
+#include <iterator>
+
+#include "opencv2/gpufeatures2d.hpp"
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/gpuwarping.hpp"
+#include "opencv2/features2d.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/test/test_features2d.cpp b/modules/gpufeatures2d/test/test_features2d.cpp
similarity index 68%
rename from modules/gpu/test/test_features2d.cpp
rename to modules/gpufeatures2d/test/test_features2d.cpp
index 3879ac0534..29dcc20407 100644
--- a/modules/gpu/test/test_features2d.cpp
+++ b/modules/gpufeatures2d/test/test_features2d.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,306 +44,7 @@
 
 #ifdef HAVE_CUDA
 
-namespace
-{
-    bool keyPointsEquals(const cv::KeyPoint& p1, const cv::KeyPoint& p2)
-    {
-        const double maxPtDif = 1.0;
-        const double maxSizeDif = 1.0;
-        const double maxAngleDif = 2.0;
-        const double maxResponseDif = 0.1;
-
-        double dist = cv::norm(p1.pt - p2.pt);
-
-        if (dist < maxPtDif &&
-            fabs(p1.size - p2.size) < maxSizeDif &&
-            abs(p1.angle - p2.angle) < maxAngleDif &&
-            abs(p1.response - p2.response) < maxResponseDif &&
-            p1.octave == p2.octave &&
-            p1.class_id == p2.class_id)
-        {
-            return true;
-        }
-
-        return false;
-    }
-
-    struct KeyPointLess : std::binary_function<cv::KeyPoint, cv::KeyPoint, bool>
-    {
-        bool operator()(const cv::KeyPoint& kp1, const cv::KeyPoint& kp2) const
-        {
-            return kp1.pt.y < kp2.pt.y || (kp1.pt.y == kp2.pt.y && kp1.pt.x < kp2.pt.x);
-        }
-    };
-
-    testing::AssertionResult assertKeyPointsEquals(const char* gold_expr, const char* actual_expr, std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual)
-    {
-        if (gold.size() != actual.size())
-        {
-            return testing::AssertionFailure() << "KeyPoints size mistmach\n"
-                                               << "\"" << gold_expr << "\" : " << gold.size() << "\n"
-                                               << "\"" << actual_expr << "\" : " << actual.size();
-        }
-
-        std::sort(actual.begin(), actual.end(), KeyPointLess());
-        std::sort(gold.begin(), gold.end(), KeyPointLess());
-
-        for (size_t i = 0; i < gold.size(); ++i)
-        {
-            const cv::KeyPoint& p1 = gold[i];
-            const cv::KeyPoint& p2 = actual[i];
-
-            if (!keyPointsEquals(p1, p2))
-            {
-                return testing::AssertionFailure() << "KeyPoints differ at " << i << "\n"
-                                                   << "\"" << gold_expr << "\" vs \"" << actual_expr << "\" : \n"
-                                                   << "pt : " << testing::PrintToString(p1.pt) << " vs " << testing::PrintToString(p2.pt) << "\n"
-                                                   << "size : " << p1.size << " vs " << p2.size << "\n"
-                                                   << "angle : " << p1.angle << " vs " << p2.angle << "\n"
-                                                   << "response : " << p1.response << " vs " << p2.response << "\n"
-                                                   << "octave : " << p1.octave << " vs " << p2.octave << "\n"
-                                                   << "class_id : " << p1.class_id << " vs " << p2.class_id;
-            }
-        }
-
-        return ::testing::AssertionSuccess();
-    }
-
-    #define ASSERT_KEYPOINTS_EQ(gold, actual) EXPECT_PRED_FORMAT2(assertKeyPointsEquals, gold, actual);
-
-    int getMatchedPointsCount(std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual)
-    {
-        std::sort(actual.begin(), actual.end(), KeyPointLess());
-        std::sort(gold.begin(), gold.end(), KeyPointLess());
-
-        int validCount = 0;
-
-        for (size_t i = 0; i < gold.size(); ++i)
-        {
-            const cv::KeyPoint& p1 = gold[i];
-            const cv::KeyPoint& p2 = actual[i];
-
-            if (keyPointsEquals(p1, p2))
-                ++validCount;
-        }
-
-        return validCount;
-    }
-
-    int getMatchedPointsCount(const std::vector<cv::KeyPoint>& keypoints1, const std::vector<cv::KeyPoint>& keypoints2, const std::vector<cv::DMatch>& matches)
-    {
-        int validCount = 0;
-
-        for (size_t i = 0; i < matches.size(); ++i)
-        {
-            const cv::DMatch& m = matches[i];
-
-            const cv::KeyPoint& p1 = keypoints1[m.queryIdx];
-            const cv::KeyPoint& p2 = keypoints2[m.trainIdx];
-
-            if (keyPointsEquals(p1, p2))
-                ++validCount;
-        }
-
-        return validCount;
-    }
-}
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// SURF
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(SURF_HessianThreshold, double)
-    IMPLEMENT_PARAM_CLASS(SURF_Octaves, int)
-    IMPLEMENT_PARAM_CLASS(SURF_OctaveLayers, int)
-    IMPLEMENT_PARAM_CLASS(SURF_Extended, bool)
-    IMPLEMENT_PARAM_CLASS(SURF_Upright, bool)
-}
-
-PARAM_TEST_CASE(SURF, cv::gpu::DeviceInfo, SURF_HessianThreshold, SURF_Octaves, SURF_OctaveLayers, SURF_Extended, SURF_Upright)
-{
-    cv::gpu::DeviceInfo devInfo;
-    double hessianThreshold;
-    int nOctaves;
-    int nOctaveLayers;
-    bool extended;
-    bool upright;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        hessianThreshold = GET_PARAM(1);
-        nOctaves = GET_PARAM(2);
-        nOctaveLayers = GET_PARAM(3);
-        extended = GET_PARAM(4);
-        upright = GET_PARAM(5);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(SURF, Detector)
-{
-    cv::Mat image = readImage("features2d/aloe.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    cv::gpu::SURF_GPU surf;
-    surf.hessianThreshold = hessianThreshold;
-    surf.nOctaves = nOctaves;
-    surf.nOctaveLayers = nOctaveLayers;
-    surf.extended = extended;
-    surf.upright = upright;
-    surf.keypointsRatio = 0.05f;
-
-    if (!supportFeature(devInfo, cv::gpu::GLOBAL_ATOMICS))
-    {
-        try
-        {
-            std::vector<cv::KeyPoint> keypoints;
-            surf(loadMat(image), cv::gpu::GpuMat(), keypoints);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        std::vector<cv::KeyPoint> keypoints;
-        surf(loadMat(image), cv::gpu::GpuMat(), keypoints);
-
-        cv::SURF surf_gold;
-        surf_gold.hessianThreshold = hessianThreshold;
-        surf_gold.nOctaves = nOctaves;
-        surf_gold.nOctaveLayers = nOctaveLayers;
-        surf_gold.extended = extended;
-        surf_gold.upright = upright;
-
-        std::vector<cv::KeyPoint> keypoints_gold;
-        surf_gold(image, cv::noArray(), keypoints_gold);
-
-        ASSERT_EQ(keypoints_gold.size(), keypoints.size());
-        int matchedCount = getMatchedPointsCount(keypoints_gold, keypoints);
-        double matchedRatio = static_cast<double>(matchedCount) / keypoints_gold.size();
-
-        EXPECT_GT(matchedRatio, 0.95);
-    }
-}
-
-GPU_TEST_P(SURF, Detector_Masked)
-{
-    cv::Mat image = readImage("features2d/aloe.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    cv::Mat mask(image.size(), CV_8UC1, cv::Scalar::all(1));
-    mask(cv::Range(0, image.rows / 2), cv::Range(0, image.cols / 2)).setTo(cv::Scalar::all(0));
-
-    cv::gpu::SURF_GPU surf;
-    surf.hessianThreshold = hessianThreshold;
-    surf.nOctaves = nOctaves;
-    surf.nOctaveLayers = nOctaveLayers;
-    surf.extended = extended;
-    surf.upright = upright;
-    surf.keypointsRatio = 0.05f;
-
-    if (!supportFeature(devInfo, cv::gpu::GLOBAL_ATOMICS))
-    {
-        try
-        {
-            std::vector<cv::KeyPoint> keypoints;
-            surf(loadMat(image), loadMat(mask), keypoints);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        std::vector<cv::KeyPoint> keypoints;
-        surf(loadMat(image), loadMat(mask), keypoints);
-
-        cv::SURF surf_gold;
-        surf_gold.hessianThreshold = hessianThreshold;
-        surf_gold.nOctaves = nOctaves;
-        surf_gold.nOctaveLayers = nOctaveLayers;
-        surf_gold.extended = extended;
-        surf_gold.upright = upright;
-
-        std::vector<cv::KeyPoint> keypoints_gold;
-        surf_gold(image, mask, keypoints_gold);
-
-        ASSERT_EQ(keypoints_gold.size(), keypoints.size());
-        int matchedCount = getMatchedPointsCount(keypoints_gold, keypoints);
-        double matchedRatio = static_cast<double>(matchedCount) / keypoints_gold.size();
-
-        EXPECT_GT(matchedRatio, 0.95);
-    }
-}
-
-GPU_TEST_P(SURF, Descriptor)
-{
-    cv::Mat image = readImage("features2d/aloe.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    cv::gpu::SURF_GPU surf;
-    surf.hessianThreshold = hessianThreshold;
-    surf.nOctaves = nOctaves;
-    surf.nOctaveLayers = nOctaveLayers;
-    surf.extended = extended;
-    surf.upright = upright;
-    surf.keypointsRatio = 0.05f;
-
-    cv::SURF surf_gold;
-    surf_gold.hessianThreshold = hessianThreshold;
-    surf_gold.nOctaves = nOctaves;
-    surf_gold.nOctaveLayers = nOctaveLayers;
-    surf_gold.extended = extended;
-    surf_gold.upright = upright;
-
-    if (!supportFeature(devInfo, cv::gpu::GLOBAL_ATOMICS))
-    {
-        try
-        {
-            std::vector<cv::KeyPoint> keypoints;
-            cv::gpu::GpuMat descriptors;
-            surf(loadMat(image), cv::gpu::GpuMat(), keypoints, descriptors);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        std::vector<cv::KeyPoint> keypoints;
-        surf_gold(image, cv::noArray(), keypoints);
-
-        cv::gpu::GpuMat descriptors;
-        surf(loadMat(image), cv::gpu::GpuMat(), keypoints, descriptors, true);
-
-        cv::Mat descriptors_gold;
-        surf_gold(image, cv::noArray(), keypoints, descriptors_gold, true);
-
-        cv::BFMatcher matcher(cv::NORM_L2);
-        std::vector<cv::DMatch> matches;
-        matcher.match(descriptors_gold, cv::Mat(descriptors), matches);
-
-        int matchedCount = getMatchedPointsCount(keypoints, keypoints, matches);
-        double matchedRatio = static_cast<double>(matchedCount) / keypoints.size();
-
-        EXPECT_GT(matchedRatio, 0.6);
-    }
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Features2D, SURF, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(SURF_HessianThreshold(100.0), SURF_HessianThreshold(500.0), SURF_HessianThreshold(1000.0)),
-    testing::Values(SURF_Octaves(3), SURF_Octaves(4)),
-    testing::Values(SURF_OctaveLayers(2), SURF_OctaveLayers(3)),
-    testing::Values(SURF_Extended(false), SURF_Extended(true)),
-    testing::Values(SURF_Upright(false), SURF_Upright(true))));
+using namespace cvtest;
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // FAST
@@ -386,7 +88,7 @@ GPU_TEST_P(FAST, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
         }
     }
     else
@@ -421,7 +123,7 @@ namespace
     IMPLEMENT_PARAM_CLASS(ORB_BlurForDescriptor, bool)
 }
 
-CV_ENUM(ORB_ScoreType, cv::ORB::HARRIS_SCORE, cv::ORB::FAST_SCORE)
+CV_ENUM(ORB_ScoreType, ORB::HARRIS_SCORE, ORB::FAST_SCORE)
 
 PARAM_TEST_CASE(ORB, cv::gpu::DeviceInfo, ORB_FeaturesCount, ORB_ScaleFactor, ORB_LevelsCount, ORB_EdgeThreshold, ORB_firstLevel, ORB_WTA_K, ORB_ScoreType, ORB_PatchSize, ORB_BlurForDescriptor)
 {
@@ -474,7 +176,7 @@ GPU_TEST_P(ORB, Accuracy)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
         }
     }
     else
@@ -879,7 +581,7 @@ GPU_TEST_P(BruteForceMatcher, RadiusMatch_Single)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
         }
     }
     else
@@ -944,7 +646,7 @@ GPU_TEST_P(BruteForceMatcher, RadiusMatch_Collection)
         }
         catch (const cv::Exception& e)
         {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
+            ASSERT_EQ(cv::Error::StsNotImplemented, e.code);
         }
     }
     else
diff --git a/modules/gpufeatures2d/test/test_main.cpp b/modules/gpufeatures2d/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpufeatures2d/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpufeatures2d/test/test_precomp.hpp b/modules/gpufeatures2d/test/test_precomp.hpp
new file mode 100644
index 0000000000..4ce8f78496
--- /dev/null
+++ b/modules/gpufeatures2d/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpufeatures2d.hpp"
+#include "opencv2/features2d.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpufilters/CMakeLists.txt b/modules/gpufilters/CMakeLists.txt
new file mode 100644
index 0000000000..640de8c115
--- /dev/null
+++ b/modules/gpufilters/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpufilters)
+endif()
+
+set(the_description "GPU-accelerated Image Filtering")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_define_module(gpufilters opencv_imgproc opencv_gpuarithm)
diff --git a/modules/gpufilters/doc/filtering.rst b/modules/gpufilters/doc/filtering.rst
new file mode 100644
index 0000000000..cb824bf2a4
--- /dev/null
+++ b/modules/gpufilters/doc/filtering.rst
@@ -0,0 +1,336 @@
+Image Filtering
+===============
+
+.. highlight:: cpp
+
+Functions and classes described in this section are used to perform various linear or non-linear filtering operations on 2D images.
+
+.. note::
+
+   * An example containing all basic morphology operators like erode and dilate can be found at opencv_source_code/samples/gpu/morphology.cpp
+
+gpu::Filter
+-----------
+.. ocv:class:: gpu::Filter
+
+Common interface for all GPU filters ::
+
+    class CV_EXPORTS Filter : public Algorithm
+    {
+    public:
+        virtual void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+    };
+
+
+
+gpu::Filter::apply
+------------------
+Applies the specified filter to the image.
+
+.. ocv:function:: void gpu::Filter::apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0
+
+    :param src: Input image.
+
+    :param dst: Output image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createBoxFilter
+--------------------
+Creates a normalized 2D box filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createBoxFilter(int srcType, int dstType, Size ksize, Point anchor = Point(-1,-1), int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input image type. Only ``CV_8UC1`` and ``CV_8UC4`` are supported for now.
+
+    :param dstType: Output image type. Only the same type as ``src`` is supported for now.
+
+    :param ksize: Kernel size.
+
+    :param anchor: Anchor point. The default value ``Point(-1, -1)`` means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+.. seealso:: :ocv:func:`boxFilter`
+
+
+
+gpu::createLinearFilter
+-----------------------
+Creates a non-separable linear 2D filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createLinearFilter(int srcType, int dstType, InputArray kernel, Point anchor = Point(-1,-1), int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input image type. Supports  ``CV_8U``  ,  ``CV_16U``  and  ``CV_32F``  one and four channel image.
+
+    :param dstType: Output image type. Only the same type as ``src`` is supported for now.
+
+    :param kernel: 2D array of filter coefficients.
+
+    :param anchor: Anchor point. The default value Point(-1, -1) means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+.. seealso:: :ocv:func:`filter2D`
+
+
+
+gpu::createLaplacianFilter
+--------------------------
+Creates a Laplacian operator.
+
+.. ocv:function:: Ptr<Filter> gpu::createLaplacianFilter(int srcType, int dstType, int ksize = 1, double scale = 1, int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input image type. Supports  ``CV_8U``  ,  ``CV_16U``  and  ``CV_32F``  one and four channel image.
+
+    :param dstType: Output image type. Only the same type as ``src`` is supported for now.
+
+    :param ksize: Aperture size used to compute the second-derivative filters (see :ocv:func:`getDerivKernels`). It must be positive and odd. Only  ``ksize``  = 1 and  ``ksize``  = 3 are supported.
+
+    :param scale: Optional scale factor for the computed Laplacian values. By default, no scaling is applied (see  :ocv:func:`getDerivKernels` ).
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+.. seealso:: :ocv:func:`Laplacian`
+
+
+
+gpu::createSeparableLinearFilter
+--------------------------------
+Creates a separable linear filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createSeparableLinearFilter(int srcType, int dstType, InputArray rowKernel, InputArray columnKernel, Point anchor = Point(-1,-1), int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1)
+
+    :param srcType: Source array type.
+
+    :param dstType: Destination array type.
+
+    :param rowKernel: Horizontal filter coefficients. Support kernels with ``size <= 32`` .
+
+    :param columnKernel: Vertical filter coefficients. Support kernels with ``size <= 32`` .
+
+    :param anchor: Anchor position within the kernel. Negative values mean that anchor is positioned at the aperture center.
+
+    :param rowBorderMode: Pixel extrapolation method in the vertical direction For details, see  :ocv:func:`borderInterpolate`.
+
+    :param columnBorderMode: Pixel extrapolation method in the horizontal direction.
+
+.. seealso:: :ocv:func:`sepFilter2D`
+
+
+
+gpu::createDerivFilter
+----------------------
+Creates a generalized Deriv operator.
+
+.. ocv:function:: Ptr<Filter> gpu::createDerivFilter(int srcType, int dstType, int dx, int dy, int ksize, bool normalize = false, double scale = 1, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1)
+
+    :param srcType: Source image type.
+
+    :param dstType: Destination array type.
+
+    :param dx: Derivative order in respect of x.
+
+    :param dy: Derivative order in respect of y.
+
+    :param ksize: Aperture size. See  :ocv:func:`getDerivKernels` for details.
+
+    :param normalize: Flag indicating whether to normalize (scale down) the filter coefficients or not. See  :ocv:func:`getDerivKernels` for details.
+
+    :param scale: Optional scale factor for the computed derivative values. By default, no scaling is applied. For details, see  :ocv:func:`getDerivKernels` .
+
+    :param rowBorderMode: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
+
+    :param columnBorderMode: Pixel extrapolation method in the horizontal direction.
+
+
+
+gpu::createSobelFilter
+----------------------
+Creates a Sobel operator.
+
+.. ocv:function:: Ptr<Filter> gpu::createSobelFilter(int srcType, int dstType, int dx, int dy, int ksize = 3, double scale = 1, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1)
+
+    :param srcType: Source image type.
+
+    :param dstType: Destination array type.
+
+    :param dx: Derivative order in respect of x.
+
+    :param dy: Derivative order in respect of y.
+
+    :param ksize: Size of the extended Sobel kernel. Possible values are 1, 3, 5 or 7.
+
+    :param scale: Optional scale factor for the computed derivative values. By default, no scaling is applied. For details, see  :ocv:func:`getDerivKernels` .
+
+    :param rowBorderMode: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
+
+    :param columnBorderMode: Pixel extrapolation method in the horizontal direction.
+
+.. seealso:: :ocv:func:`Sobel`
+
+
+
+gpu::createScharrFilter
+-----------------------
+Creates a vertical or horizontal Scharr operator.
+
+.. ocv:function:: Ptr<Filter> gpu::createScharrFilter(int srcType, int dstType, int dx, int dy, double scale = 1, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1)
+
+    :param srcType: Source image type.
+
+    :param dstType: Destination array type.
+
+    :param dx: Order of the derivative x.
+
+    :param dy: Order of the derivative y.
+
+    :param scale: Optional scale factor for the computed derivative values. By default, no scaling is applied. See  :ocv:func:`getDerivKernels`  for details.
+
+    :param rowBorderMode: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
+
+    :param columnBorderMode: Pixel extrapolation method in the horizontal direction.
+
+.. seealso:: :ocv:func:`Scharr`
+
+
+
+gpu::createGaussianFilter
+-------------------------
+Creates a Gaussian filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createGaussianFilter(int srcType, int dstType, Size ksize, double sigma1, double sigma2 = 0, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1)
+
+    :param srcType: Source image type.
+
+    :param dstType: Destination array type.
+
+    :param ksize: Aperture size. See  :ocv:func:`getGaussianKernel` for details.
+
+    :param sigma1: Gaussian sigma in the horizontal direction. See  :ocv:func:`getGaussianKernel` for details.
+
+    :param sigma2: Gaussian sigma in the vertical direction. If 0, then  :math:`\texttt{sigma2}\leftarrow\texttt{sigma1}` .
+
+    :param rowBorderMode: Pixel extrapolation method in the vertical direction. For details, see  :ocv:func:`borderInterpolate`.
+
+    :param columnBorderMode: Pixel extrapolation method in the horizontal direction.
+
+.. seealso:: :ocv:func:`GaussianBlur`
+
+
+
+gpu::createMorphologyFilter
+---------------------------
+Creates a 2D morphological filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createMorphologyFilter(int op, int srcType, InputArray kernel, Point anchor = Point(-1, -1), int iterations = 1)
+
+    :param op: Type of morphological operation. The following types are possible:
+
+        * **MORPH_ERODE** erode
+
+        * **MORPH_DILATE** dilate
+
+        * **MORPH_OPEN** opening
+
+        * **MORPH_CLOSE** closing
+
+        * **MORPH_GRADIENT** morphological gradient
+
+        * **MORPH_TOPHAT** "top hat"
+
+        * **MORPH_BLACKHAT** "black hat"
+
+    :param srcType: Input/output image type. Only  ``CV_8UC1``  and  ``CV_8UC4``  are supported.
+
+    :param kernel: 2D 8-bit structuring element for the morphological operation.
+
+    :param anchor: Anchor position within the structuring element. Negative values mean that the anchor is at the center.
+
+    :param iterations: Number of times erosion and dilation to be applied.
+
+.. seealso:: :ocv:func:`morphologyEx`
+
+
+
+gpu::createBoxMaxFilter
+-----------------------
+Creates the maximum filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createBoxMaxFilter(int srcType, Size ksize, Point anchor = Point(-1, -1), int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input/output image type. Only  ``CV_8UC1``  and  ``CV_8UC4`` are supported.
+
+    :param ksize: Kernel size.
+
+    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+
+
+gpu::createBoxMinFilter
+-----------------------
+Creates the minimum filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createBoxMinFilter(int srcType, Size ksize, Point anchor = Point(-1, -1), int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input/output image type. Only  ``CV_8UC1``  and  ``CV_8UC4`` are supported.
+
+    :param ksize: Kernel size.
+
+    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+
+
+gpu::createRowSumFilter
+-----------------------
+Creates a horizontal 1D box filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createRowSumFilter(int srcType, int dstType, int ksize, int anchor = -1, int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input image type. Only ``CV_8UC1`` type is supported for now.
+
+    :param sumType: Output image type. Only ``CV_32FC1`` type is supported for now.
+
+    :param ksize: Kernel size.
+
+    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
+
+
+
+gpu::createColumnSumFilter
+--------------------------
+Creates a vertical 1D box filter.
+
+.. ocv:function:: Ptr<Filter> gpu::createColumnSumFilter(int srcType, int dstType, int ksize, int anchor = -1, int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0))
+
+    :param srcType: Input image type. Only ``CV_8UC1`` type is supported for now.
+
+    :param sumType: Output image type. Only ``CV_32FC1`` type is supported for now.
+
+    :param ksize: Kernel size.
+
+    :param anchor: Anchor point. The default value (-1) means that the anchor is at the kernel center.
+
+    :param borderMode: Pixel extrapolation method. For details, see :ocv:func:`borderInterpolate` .
+
+    :param borderVal: Default border value.
diff --git a/modules/gpufilters/doc/gpufilters.rst b/modules/gpufilters/doc/gpufilters.rst
new file mode 100644
index 0000000000..778554b13b
--- /dev/null
+++ b/modules/gpufilters/doc/gpufilters.rst
@@ -0,0 +1,8 @@
+*******************************************
+gpufilters. GPU-accelerated Image Filtering
+*******************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    filtering
diff --git a/modules/gpufilters/include/opencv2/gpufilters.hpp b/modules/gpufilters/include/opencv2/gpufilters.hpp
new file mode 100644
index 0000000000..b0ebfd73c5
--- /dev/null
+++ b/modules/gpufilters/include/opencv2/gpufilters.hpp
@@ -0,0 +1,149 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUFILTERS_HPP__
+#define __OPENCV_GPUFILTERS_HPP__
+
+#ifndef __cplusplus
+#  error gpufilters.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/imgproc.hpp"
+
+namespace cv { namespace gpu {
+
+class CV_EXPORTS Filter : public Algorithm
+{
+public:
+    virtual void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Box Filter
+
+//! creates a normalized 2D box filter
+//! supports CV_8UC1, CV_8UC4 types
+CV_EXPORTS Ptr<Filter> createBoxFilter(int srcType, int dstType, Size ksize, Point anchor = Point(-1,-1),
+                                       int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Linear Filter
+
+//! Creates a non-separable linear 2D filter
+//! supports 1 and 4 channel CV_8U, CV_16U and CV_32F input
+CV_EXPORTS Ptr<Filter> createLinearFilter(int srcType, int dstType, InputArray kernel, Point anchor = Point(-1,-1),
+                                          int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Laplacian Filter
+
+//! creates a Laplacian operator
+//! supports only ksize = 1 and ksize = 3
+CV_EXPORTS Ptr<Filter> createLaplacianFilter(int srcType, int dstType, int ksize = 1, double scale = 1,
+                                             int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Separable Linear Filter
+
+//! creates a separable linear filter
+CV_EXPORTS Ptr<Filter> createSeparableLinearFilter(int srcType, int dstType, InputArray rowKernel, InputArray columnKernel,
+                                                   Point anchor = Point(-1,-1), int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1);
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Deriv Filter
+
+//! creates a generalized Deriv operator
+CV_EXPORTS Ptr<Filter> createDerivFilter(int srcType, int dstType, int dx, int dy,
+                                         int ksize, bool normalize = false, double scale = 1,
+                                         int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1);
+
+//! creates a Sobel operator
+CV_EXPORTS Ptr<Filter> createSobelFilter(int srcType, int dstType, int dx, int dy, int ksize = 3,
+                                         double scale = 1, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1);
+
+//! creates a vertical or horizontal Scharr operator
+CV_EXPORTS Ptr<Filter> createScharrFilter(int srcType, int dstType, int dx, int dy,
+                                          double scale = 1, int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1);
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Gaussian Filter
+
+//! creates a Gaussian filter
+CV_EXPORTS Ptr<Filter> createGaussianFilter(int srcType, int dstType, Size ksize,
+                                            double sigma1, double sigma2 = 0,
+                                            int rowBorderMode = BORDER_DEFAULT, int columnBorderMode = -1);
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Morphology Filter
+
+//! creates a 2D morphological filter
+//! supports CV_8UC1 and CV_8UC4 types
+CV_EXPORTS Ptr<Filter> createMorphologyFilter(int op, int srcType, InputArray kernel, Point anchor = Point(-1, -1), int iterations = 1);
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Image Rank Filter
+
+//! result pixel value is the maximum of pixel values under the rectangular mask region
+CV_EXPORTS Ptr<Filter> createBoxMaxFilter(int srcType, Size ksize,
+                                          Point anchor = Point(-1, -1),
+                                          int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+//! result pixel value is the maximum of pixel values under the rectangular mask region
+CV_EXPORTS Ptr<Filter> createBoxMinFilter(int srcType, Size ksize,
+                                          Point anchor = Point(-1, -1),
+                                          int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// 1D Sum Filter
+
+//! creates a horizontal 1D box filter
+//! supports only CV_8UC1 source type and CV_32FC1 sum type
+CV_EXPORTS Ptr<Filter> createRowSumFilter(int srcType, int dstType, int ksize, int anchor = -1, int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+//! creates a vertical 1D box filter
+//! supports only CV_8UC1 sum type and CV_32FC1 dst type
+CV_EXPORTS Ptr<Filter> createColumnSumFilter(int srcType, int dstType, int ksize, int anchor = -1, int borderMode = BORDER_DEFAULT, Scalar borderVal = Scalar::all(0));
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUFILTERS_HPP__ */
diff --git a/modules/gpu/perf/perf_filters.cpp b/modules/gpufilters/perf/perf_filters.cpp
similarity index 54%
rename from modules/gpu/perf/perf_filters.cpp
rename to modules/gpufilters/perf/perf_filters.cpp
index 3516954a6c..6ad0998a5b 100644
--- a/modules/gpu/perf/perf_filters.cpp
+++ b/modules/gpufilters/perf/perf_filters.cpp
@@ -1,14 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
 using namespace std;
 using namespace testing;
+using namespace perf;
 
 //////////////////////////////////////////////////////////////////////
 // Blur
 
 DEF_PARAM_TEST(Sz_Type_KernelSz, cv::Size, MatType, int);
 
-PERF_TEST_P(Sz_Type_KernelSz, Filters_Blur,
+PERF_TEST_P(Sz_Type_KernelSz, Blur,
             Combine(GPU_TYPICAL_MAT_SIZES,
                     Values(CV_8UC1, CV_8UC4),
                     Values(3, 5, 7)))
@@ -27,9 +70,11 @@ PERF_TEST_P(Sz_Type_KernelSz, Filters_Blur,
         const cv::gpu::GpuMat d_src(src);
         cv::gpu::GpuMat dst;
 
-        TEST_CYCLE() cv::gpu::blur(d_src, dst, cv::Size(ksize, ksize));
+        cv::Ptr<cv::gpu::Filter> blurFilter = cv::gpu::createBoxFilter(d_src.type(), -1, cv::Size(ksize, ksize));
 
-        GPU_SANITY_CHECK(dst);
+        TEST_CYCLE() blurFilter->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst, 1);
     }
     else
     {
@@ -41,256 +86,10 @@ PERF_TEST_P(Sz_Type_KernelSz, Filters_Blur,
     }
 }
 
-//////////////////////////////////////////////////////////////////////
-// Sobel
-
-PERF_TEST_P(Sz_Type_KernelSz, Filters_Sobel, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1), Values(3, 5, 7, 9, 11, 13, 15)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int ksize = GET_PARAM(2);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::Sobel(d_src, dst, -1, 1, 1, d_buf, ksize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::Sobel(src, dst, -1, 1, 1, ksize);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Scharr
-
-PERF_TEST_P(Sz_Type, Filters_Scharr, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::Scharr(d_src, dst, -1, 1, 0, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::Scharr(src, dst, -1, 1, 0);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// GaussianBlur
-
-PERF_TEST_P(Sz_Type_KernelSz, Filters_GaussianBlur, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1), Values(3, 5, 7, 9, 11, 13, 15)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int ksize = GET_PARAM(2);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::GaussianBlur(d_src, dst, cv::Size(ksize, ksize), d_buf, 0.5);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::GaussianBlur(src, dst, cv::Size(ksize, ksize), 0.5);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Laplacian
-
-PERF_TEST_P(Sz_Type_KernelSz, Filters_Laplacian, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1, 3)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int ksize = GET_PARAM(2);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::Laplacian(d_src, dst, -1, ksize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::Laplacian(src, dst, -1, ksize);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Erode
-
-PERF_TEST_P(Sz_Type, Filters_Erode, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::erode(d_src, dst, ker, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::erode(src, dst, ker);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// Dilate
-
-PERF_TEST_P(Sz_Type, Filters_Dilate, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4)))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf;
-
-        TEST_CYCLE() cv::gpu::dilate(d_src, dst, ker, d_buf);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::dilate(src, dst, ker);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
-//////////////////////////////////////////////////////////////////////
-// MorphologyEx
-
-CV_ENUM(MorphOp, cv::MORPH_OPEN, cv::MORPH_CLOSE, cv::MORPH_GRADIENT, cv::MORPH_TOPHAT, cv::MORPH_BLACKHAT)
-#define ALL_MORPH_OPS ValuesIn(MorphOp::all())
-
-DEF_PARAM_TEST(Sz_Type_Op, cv::Size, MatType, MorphOp);
-
-PERF_TEST_P(Sz_Type_Op, Filters_MorphologyEx, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4), ALL_MORPH_OPS))
-{
-    declare.time(20.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int type = GET_PARAM(1);
-    const int morphOp = GET_PARAM(2);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-        cv::gpu::GpuMat d_buf1;
-        cv::gpu::GpuMat d_buf2;
-
-        TEST_CYCLE() cv::gpu::morphologyEx(d_src, dst, morphOp, ker, d_buf1, d_buf2);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::morphologyEx(src, dst, morphOp, ker);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // Filter2D
 
-PERF_TEST_P(Sz_Type_KernelSz, Filters_Filter2D, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(3, 5, 7, 9, 11, 13, 15)))
+PERF_TEST_P(Sz_Type_KernelSz, Filter2D, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(3, 5, 7, 9, 11, 13, 15)))
 {
     declare.time(20.0);
 
@@ -309,7 +108,9 @@ PERF_TEST_P(Sz_Type_KernelSz, Filters_Filter2D, Combine(GPU_TYPICAL_MAT_SIZES, V
         const cv::gpu::GpuMat d_src(src);
         cv::gpu::GpuMat dst;
 
-        TEST_CYCLE() cv::gpu::filter2D(d_src, dst, -1, kernel);
+        cv::Ptr<cv::gpu::Filter> filter2D = cv::gpu::createLinearFilter(d_src.type(), -1, kernel);
+
+        TEST_CYCLE() filter2D->apply(d_src, dst);
 
         GPU_SANITY_CHECK(dst);
     }
@@ -322,3 +123,255 @@ PERF_TEST_P(Sz_Type_KernelSz, Filters_Filter2D, Combine(GPU_TYPICAL_MAT_SIZES, V
         CPU_SANITY_CHECK(dst);
     }
 }
+
+//////////////////////////////////////////////////////////////////////
+// Laplacian
+
+PERF_TEST_P(Sz_Type_KernelSz, Laplacian, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1, 3)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int ksize = GET_PARAM(2);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> laplacian = cv::gpu::createLaplacianFilter(d_src.type(), -1, ksize);
+
+        TEST_CYCLE() laplacian->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::Laplacian(src, dst, -1, ksize);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Sobel
+
+PERF_TEST_P(Sz_Type_KernelSz, Sobel, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1), Values(3, 5, 7, 9, 11, 13, 15)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int ksize = GET_PARAM(2);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> sobel = cv::gpu::createSobelFilter(d_src.type(), -1, 1, 1, ksize);
+
+        TEST_CYCLE() sobel->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::Sobel(src, dst, -1, 1, 1, ksize);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Scharr
+
+PERF_TEST_P(Sz_Type, Scharr, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> scharr = cv::gpu::createScharrFilter(d_src.type(), -1, 1, 0);
+
+        TEST_CYCLE() scharr->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::Scharr(src, dst, -1, 1, 0);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// GaussianBlur
+
+PERF_TEST_P(Sz_Type_KernelSz, GaussianBlur, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4, CV_32FC1), Values(3, 5, 7, 9, 11, 13, 15)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int ksize = GET_PARAM(2);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> gauss = cv::gpu::createGaussianFilter(d_src.type(), -1, cv::Size(ksize, ksize), 0.5);
+
+        TEST_CYCLE() gauss->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::GaussianBlur(src, dst, cv::Size(ksize, ksize), 0.5);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Erode
+
+PERF_TEST_P(Sz_Type, Erode, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> erode = cv::gpu::createMorphologyFilter(cv::MORPH_ERODE, src.type(), ker);
+
+        TEST_CYCLE() erode->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::erode(src, dst, ker);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Dilate
+
+PERF_TEST_P(Sz_Type, Dilate, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> dilate = cv::gpu::createMorphologyFilter(cv::MORPH_DILATE, src.type(), ker);
+
+        TEST_CYCLE() dilate->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::dilate(src, dst, ker);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MorphologyEx
+
+CV_ENUM(MorphOp, MORPH_OPEN, MORPH_CLOSE, MORPH_GRADIENT, MORPH_TOPHAT, MORPH_BLACKHAT)
+
+DEF_PARAM_TEST(Sz_Type_Op, cv::Size, MatType, MorphOp);
+
+PERF_TEST_P(Sz_Type_Op, MorphologyEx, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8UC1, CV_8UC4), MorphOp::all()))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int morphOp = GET_PARAM(2);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const cv::Mat ker = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::Filter> morph = cv::gpu::createMorphologyFilter(morphOp, src.type(), ker);
+
+        TEST_CYCLE() morph->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::morphologyEx(src, dst, morphOp, ker);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpufilters/perf/perf_main.cpp b/modules/gpufilters/perf/perf_main.cpp
new file mode 100644
index 0000000000..44b3129e1f
--- /dev/null
+++ b/modules/gpufilters/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpufilters)
diff --git a/modules/gpufilters/perf/perf_precomp.hpp b/modules/gpufilters/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..02ca5ceacb
--- /dev/null
+++ b/modules/gpufilters/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpufilters.hpp"
+#include "opencv2/imgproc.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpu/src/cuda/column_filter.8.cu b/modules/gpufilters/src/cuda/column_filter.16sc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.8.cu
rename to modules/gpufilters/src/cuda/column_filter.16sc1.cu
index 38e70e772e..d4c6d19ab8 100644
--- a/modules/gpu/src/cuda/column_filter.8.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16sc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.3.cu b/modules/gpufilters/src/cuda/column_filter.16sc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.3.cu
rename to modules/gpufilters/src/cuda/column_filter.16sc3.cu
index 1bf49219f9..419fdea652 100644
--- a/modules/gpu/src/cuda/column_filter.3.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16sc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.9.cu b/modules/gpufilters/src/cuda/column_filter.16sc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.9.cu
rename to modules/gpufilters/src/cuda/column_filter.16sc4.cu
index 5b58345820..1caeb87758 100644
--- a/modules/gpu/src/cuda/column_filter.9.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16sc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.10.cu b/modules/gpufilters/src/cuda/column_filter.16uc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.10.cu
rename to modules/gpufilters/src/cuda/column_filter.16uc1.cu
index 41e35bc1c6..dc68b710f5 100644
--- a/modules/gpu/src/cuda/column_filter.10.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16uc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.11.cu b/modules/gpufilters/src/cuda/column_filter.16uc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.11.cu
rename to modules/gpufilters/src/cuda/column_filter.16uc3.cu
index 981208a68b..f0a07d6ddd 100644
--- a/modules/gpu/src/cuda/column_filter.11.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16uc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.12.cu b/modules/gpufilters/src/cuda/column_filter.16uc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.12.cu
rename to modules/gpufilters/src/cuda/column_filter.16uc4.cu
index 13d2e60023..638ef794ad 100644
--- a/modules/gpu/src/cuda/column_filter.12.cu
+++ b/modules/gpufilters/src/cuda/column_filter.16uc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.5.cu b/modules/gpufilters/src/cuda/column_filter.32fc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.5.cu
rename to modules/gpufilters/src/cuda/column_filter.32fc1.cu
index 8194ee39aa..aa30933e69 100644
--- a/modules/gpu/src/cuda/column_filter.5.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32fc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.6.cu b/modules/gpufilters/src/cuda/column_filter.32fc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.6.cu
rename to modules/gpufilters/src/cuda/column_filter.32fc3.cu
index d8fc49be68..c0ed3ac3cd 100644
--- a/modules/gpu/src/cuda/column_filter.6.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32fc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.7.cu b/modules/gpufilters/src/cuda/column_filter.32fc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.7.cu
rename to modules/gpufilters/src/cuda/column_filter.32fc4.cu
index 534bd821ef..f37f71792b 100644
--- a/modules/gpu/src/cuda/column_filter.7.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32fc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.4.cu b/modules/gpufilters/src/cuda/column_filter.32sc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.4.cu
rename to modules/gpufilters/src/cuda/column_filter.32sc1.cu
index bec7a085a0..ee052050d6 100644
--- a/modules/gpu/src/cuda/column_filter.4.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32sc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.13.cu b/modules/gpufilters/src/cuda/column_filter.32sc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.13.cu
rename to modules/gpufilters/src/cuda/column_filter.32sc3.cu
index 09f6484af4..b921d96103 100644
--- a/modules/gpu/src/cuda/column_filter.13.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32sc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.14.cu b/modules/gpufilters/src/cuda/column_filter.32sc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.14.cu
rename to modules/gpufilters/src/cuda/column_filter.32sc4.cu
index 901ab03011..dd21524c5d 100644
--- a/modules/gpu/src/cuda/column_filter.14.cu
+++ b/modules/gpufilters/src/cuda/column_filter.32sc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.0.cu b/modules/gpufilters/src/cuda/column_filter.8uc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.0.cu
rename to modules/gpufilters/src/cuda/column_filter.8uc1.cu
index c35c6ee64d..470f3ee8e6 100644
--- a/modules/gpu/src/cuda/column_filter.0.cu
+++ b/modules/gpufilters/src/cuda/column_filter.8uc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.1.cu b/modules/gpufilters/src/cuda/column_filter.8uc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.1.cu
rename to modules/gpufilters/src/cuda/column_filter.8uc3.cu
index 9a2d6a0427..5d5be58310 100644
--- a/modules/gpu/src/cuda/column_filter.1.cu
+++ b/modules/gpufilters/src/cuda/column_filter.8uc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.2.cu b/modules/gpufilters/src/cuda/column_filter.8uc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/column_filter.2.cu
rename to modules/gpufilters/src/cuda/column_filter.8uc4.cu
index 05ee01c763..8a322f2995 100644
--- a/modules/gpu/src/cuda/column_filter.2.cu
+++ b/modules/gpufilters/src/cuda/column_filter.8uc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "column_filter.h"
+#include "column_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/column_filter.h b/modules/gpufilters/src/cuda/column_filter.hpp
similarity index 98%
rename from modules/gpu/src/cuda/column_filter.h
rename to modules/gpufilters/src/cuda/column_filter.hpp
index 52b9103393..6f10c36f5f 100644
--- a/modules/gpu/src/cuda/column_filter.h
+++ b/modules/gpufilters/src/cuda/column_filter.hpp
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -41,13 +40,13 @@
 //
 //M*/
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 namespace column_filter
 {
@@ -188,38 +187,38 @@ namespace filter
         {
             {
                 0,
-                column_filter::caller< 1, T, D, BrdColReflect101>,
-                column_filter::caller< 2, T, D, BrdColReflect101>,
-                column_filter::caller< 3, T, D, BrdColReflect101>,
-                column_filter::caller< 4, T, D, BrdColReflect101>,
-                column_filter::caller< 5, T, D, BrdColReflect101>,
-                column_filter::caller< 6, T, D, BrdColReflect101>,
-                column_filter::caller< 7, T, D, BrdColReflect101>,
-                column_filter::caller< 8, T, D, BrdColReflect101>,
-                column_filter::caller< 9, T, D, BrdColReflect101>,
-                column_filter::caller<10, T, D, BrdColReflect101>,
-                column_filter::caller<11, T, D, BrdColReflect101>,
-                column_filter::caller<12, T, D, BrdColReflect101>,
-                column_filter::caller<13, T, D, BrdColReflect101>,
-                column_filter::caller<14, T, D, BrdColReflect101>,
-                column_filter::caller<15, T, D, BrdColReflect101>,
-                column_filter::caller<16, T, D, BrdColReflect101>,
-                column_filter::caller<17, T, D, BrdColReflect101>,
-                column_filter::caller<18, T, D, BrdColReflect101>,
-                column_filter::caller<19, T, D, BrdColReflect101>,
-                column_filter::caller<20, T, D, BrdColReflect101>,
-                column_filter::caller<21, T, D, BrdColReflect101>,
-                column_filter::caller<22, T, D, BrdColReflect101>,
-                column_filter::caller<23, T, D, BrdColReflect101>,
-                column_filter::caller<24, T, D, BrdColReflect101>,
-                column_filter::caller<25, T, D, BrdColReflect101>,
-                column_filter::caller<26, T, D, BrdColReflect101>,
-                column_filter::caller<27, T, D, BrdColReflect101>,
-                column_filter::caller<28, T, D, BrdColReflect101>,
-                column_filter::caller<29, T, D, BrdColReflect101>,
-                column_filter::caller<30, T, D, BrdColReflect101>,
-                column_filter::caller<31, T, D, BrdColReflect101>,
-                column_filter::caller<32, T, D, BrdColReflect101>
+                column_filter::caller< 1, T, D, BrdColConstant>,
+                column_filter::caller< 2, T, D, BrdColConstant>,
+                column_filter::caller< 3, T, D, BrdColConstant>,
+                column_filter::caller< 4, T, D, BrdColConstant>,
+                column_filter::caller< 5, T, D, BrdColConstant>,
+                column_filter::caller< 6, T, D, BrdColConstant>,
+                column_filter::caller< 7, T, D, BrdColConstant>,
+                column_filter::caller< 8, T, D, BrdColConstant>,
+                column_filter::caller< 9, T, D, BrdColConstant>,
+                column_filter::caller<10, T, D, BrdColConstant>,
+                column_filter::caller<11, T, D, BrdColConstant>,
+                column_filter::caller<12, T, D, BrdColConstant>,
+                column_filter::caller<13, T, D, BrdColConstant>,
+                column_filter::caller<14, T, D, BrdColConstant>,
+                column_filter::caller<15, T, D, BrdColConstant>,
+                column_filter::caller<16, T, D, BrdColConstant>,
+                column_filter::caller<17, T, D, BrdColConstant>,
+                column_filter::caller<18, T, D, BrdColConstant>,
+                column_filter::caller<19, T, D, BrdColConstant>,
+                column_filter::caller<20, T, D, BrdColConstant>,
+                column_filter::caller<21, T, D, BrdColConstant>,
+                column_filter::caller<22, T, D, BrdColConstant>,
+                column_filter::caller<23, T, D, BrdColConstant>,
+                column_filter::caller<24, T, D, BrdColConstant>,
+                column_filter::caller<25, T, D, BrdColConstant>,
+                column_filter::caller<26, T, D, BrdColConstant>,
+                column_filter::caller<27, T, D, BrdColConstant>,
+                column_filter::caller<28, T, D, BrdColConstant>,
+                column_filter::caller<29, T, D, BrdColConstant>,
+                column_filter::caller<30, T, D, BrdColConstant>,
+                column_filter::caller<31, T, D, BrdColConstant>,
+                column_filter::caller<32, T, D, BrdColConstant>
             },
             {
                 0,
@@ -256,41 +255,6 @@ namespace filter
                 column_filter::caller<31, T, D, BrdColReplicate>,
                 column_filter::caller<32, T, D, BrdColReplicate>
             },
-            {
-                0,
-                column_filter::caller< 1, T, D, BrdColConstant>,
-                column_filter::caller< 2, T, D, BrdColConstant>,
-                column_filter::caller< 3, T, D, BrdColConstant>,
-                column_filter::caller< 4, T, D, BrdColConstant>,
-                column_filter::caller< 5, T, D, BrdColConstant>,
-                column_filter::caller< 6, T, D, BrdColConstant>,
-                column_filter::caller< 7, T, D, BrdColConstant>,
-                column_filter::caller< 8, T, D, BrdColConstant>,
-                column_filter::caller< 9, T, D, BrdColConstant>,
-                column_filter::caller<10, T, D, BrdColConstant>,
-                column_filter::caller<11, T, D, BrdColConstant>,
-                column_filter::caller<12, T, D, BrdColConstant>,
-                column_filter::caller<13, T, D, BrdColConstant>,
-                column_filter::caller<14, T, D, BrdColConstant>,
-                column_filter::caller<15, T, D, BrdColConstant>,
-                column_filter::caller<16, T, D, BrdColConstant>,
-                column_filter::caller<17, T, D, BrdColConstant>,
-                column_filter::caller<18, T, D, BrdColConstant>,
-                column_filter::caller<19, T, D, BrdColConstant>,
-                column_filter::caller<20, T, D, BrdColConstant>,
-                column_filter::caller<21, T, D, BrdColConstant>,
-                column_filter::caller<22, T, D, BrdColConstant>,
-                column_filter::caller<23, T, D, BrdColConstant>,
-                column_filter::caller<24, T, D, BrdColConstant>,
-                column_filter::caller<25, T, D, BrdColConstant>,
-                column_filter::caller<26, T, D, BrdColConstant>,
-                column_filter::caller<27, T, D, BrdColConstant>,
-                column_filter::caller<28, T, D, BrdColConstant>,
-                column_filter::caller<29, T, D, BrdColConstant>,
-                column_filter::caller<30, T, D, BrdColConstant>,
-                column_filter::caller<31, T, D, BrdColConstant>,
-                column_filter::caller<32, T, D, BrdColConstant>
-            },
             {
                 0,
                 column_filter::caller< 1, T, D, BrdColReflect>,
@@ -360,6 +324,41 @@ namespace filter
                 column_filter::caller<30, T, D, BrdColWrap>,
                 column_filter::caller<31, T, D, BrdColWrap>,
                 column_filter::caller<32, T, D, BrdColWrap>
+            },
+            {
+                0,
+                column_filter::caller< 1, T, D, BrdColReflect101>,
+                column_filter::caller< 2, T, D, BrdColReflect101>,
+                column_filter::caller< 3, T, D, BrdColReflect101>,
+                column_filter::caller< 4, T, D, BrdColReflect101>,
+                column_filter::caller< 5, T, D, BrdColReflect101>,
+                column_filter::caller< 6, T, D, BrdColReflect101>,
+                column_filter::caller< 7, T, D, BrdColReflect101>,
+                column_filter::caller< 8, T, D, BrdColReflect101>,
+                column_filter::caller< 9, T, D, BrdColReflect101>,
+                column_filter::caller<10, T, D, BrdColReflect101>,
+                column_filter::caller<11, T, D, BrdColReflect101>,
+                column_filter::caller<12, T, D, BrdColReflect101>,
+                column_filter::caller<13, T, D, BrdColReflect101>,
+                column_filter::caller<14, T, D, BrdColReflect101>,
+                column_filter::caller<15, T, D, BrdColReflect101>,
+                column_filter::caller<16, T, D, BrdColReflect101>,
+                column_filter::caller<17, T, D, BrdColReflect101>,
+                column_filter::caller<18, T, D, BrdColReflect101>,
+                column_filter::caller<19, T, D, BrdColReflect101>,
+                column_filter::caller<20, T, D, BrdColReflect101>,
+                column_filter::caller<21, T, D, BrdColReflect101>,
+                column_filter::caller<22, T, D, BrdColReflect101>,
+                column_filter::caller<23, T, D, BrdColReflect101>,
+                column_filter::caller<24, T, D, BrdColReflect101>,
+                column_filter::caller<25, T, D, BrdColReflect101>,
+                column_filter::caller<26, T, D, BrdColReflect101>,
+                column_filter::caller<27, T, D, BrdColReflect101>,
+                column_filter::caller<28, T, D, BrdColReflect101>,
+                column_filter::caller<29, T, D, BrdColReflect101>,
+                column_filter::caller<30, T, D, BrdColReflect101>,
+                column_filter::caller<31, T, D, BrdColReflect101>,
+                column_filter::caller<32, T, D, BrdColReflect101>
             }
         };
 
diff --git a/modules/gpufilters/src/cuda/filter2d.cu b/modules/gpufilters/src/cuda/filter2d.cu
new file mode 100644
index 0000000000..4e913124df
--- /dev/null
+++ b/modules/gpufilters/src/cuda/filter2d.cu
@@ -0,0 +1,151 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <class SrcPtr, typename D>
+    __global__ void filter2D(const SrcPtr src, PtrStepSz<D> dst,
+                             const float* __restrict__ kernel,
+                             const int kWidth, const int kHeight,
+                             const int anchorX, const int anchorY)
+    {
+        typedef typename TypeVec<float, VecTraits<D>::cn>::vec_type sum_t;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= dst.cols || y >= dst.rows)
+            return;
+
+        sum_t res = VecTraits<sum_t>::all(0);
+        int kInd = 0;
+
+        for (int i = 0; i < kHeight; ++i)
+        {
+            for (int j = 0; j < kWidth; ++j)
+                res = res + src(y - anchorY + i, x - anchorX + j) * kernel[kInd++];
+        }
+
+        dst(y, x) = saturate_cast<D>(res);
+    }
+
+    template <typename T, typename D, template <typename> class Brd> struct Filter2DCaller;
+
+    #define IMPLEMENT_FILTER2D_TEX_READER(type) \
+        texture< type , cudaTextureType2D, cudaReadModeElementType> tex_filter2D_ ## type (0, cudaFilterModePoint, cudaAddressModeClamp); \
+        struct tex_filter2D_ ## type ## _reader \
+        { \
+            typedef type elem_type; \
+            typedef int index_type; \
+            const int xoff; \
+            const int yoff; \
+            tex_filter2D_ ## type ## _reader (int xoff_, int yoff_) : xoff(xoff_), yoff(yoff_) {} \
+            __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const \
+            { \
+                return tex2D(tex_filter2D_ ## type , x + xoff, y + yoff); \
+            } \
+        }; \
+        template <typename D, template <typename> class Brd> struct Filter2DCaller< type , D, Brd> \
+        { \
+            static void call(const PtrStepSz< type > srcWhole, int xoff, int yoff, PtrStepSz<D> dst, const float* kernel, \
+                int kWidth, int kHeight, int anchorX, int anchorY, const float* borderValue, cudaStream_t stream) \
+            { \
+                typedef typename TypeVec<float, VecTraits< type >::cn>::vec_type work_type; \
+                dim3 block(16, 16); \
+                dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \
+                bindTexture(&tex_filter2D_ ## type , srcWhole); \
+                tex_filter2D_ ## type ##_reader texSrc(xoff, yoff); \
+                Brd<work_type> brd(dst.rows, dst.cols, VecTraits<work_type>::make(borderValue)); \
+                BorderReader< tex_filter2D_ ## type ##_reader, Brd<work_type> > brdSrc(texSrc, brd); \
+                filter2D<<<grid, block, 0, stream>>>(brdSrc, dst, kernel, kWidth, kHeight, anchorX, anchorY); \
+                cudaSafeCall( cudaGetLastError() ); \
+                if (stream == 0) \
+                    cudaSafeCall( cudaDeviceSynchronize() ); \
+            } \
+        };
+
+    IMPLEMENT_FILTER2D_TEX_READER(uchar);
+    IMPLEMENT_FILTER2D_TEX_READER(uchar4);
+
+    IMPLEMENT_FILTER2D_TEX_READER(ushort);
+    IMPLEMENT_FILTER2D_TEX_READER(ushort4);
+
+    IMPLEMENT_FILTER2D_TEX_READER(float);
+    IMPLEMENT_FILTER2D_TEX_READER(float4);
+
+    #undef IMPLEMENT_FILTER2D_TEX_READER
+
+    template <typename T, typename D>
+    void filter2D(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel,
+                  int kWidth, int kHeight, int anchorX, int anchorY,
+                  int borderMode, const float* borderValue, cudaStream_t stream)
+    {
+        typedef void (*func_t)(const PtrStepSz<T> srcWhole, int xoff, int yoff, PtrStepSz<D> dst, const float* kernel,
+                               int kWidth, int kHeight, int anchorX, int anchorY, const float* borderValue, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            Filter2DCaller<T, D, BrdConstant>::call,
+            Filter2DCaller<T, D, BrdReplicate>::call,
+            Filter2DCaller<T, D, BrdReflect>::call,
+            Filter2DCaller<T, D, BrdWrap>::call,
+            Filter2DCaller<T, D, BrdReflect101>::call
+        };
+
+        funcs[borderMode]((PtrStepSz<T>) srcWhole, ofsX, ofsY, (PtrStepSz<D>) dst, kernel,
+                          kWidth, kHeight, anchorX, anchorY, borderValue, stream);
+    }
+
+    template void filter2D<uchar  , uchar  >(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void filter2D<uchar4 , uchar4 >(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void filter2D<ushort , ushort >(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void filter2D<ushort4, ushort4>(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void filter2D<float  , float  >(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void filter2D<float4 , float4 >(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel, int kWidth, int kHeight, int anchorX, int anchorY, int borderMode, const float* borderValue, cudaStream_t stream);
+}}}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpu/src/cuda/row_filter.8.cu b/modules/gpufilters/src/cuda/row_filter.16sc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.8.cu
rename to modules/gpufilters/src/cuda/row_filter.16sc1.cu
index 77e4ed63f6..59ebb9f5f7 100644
--- a/modules/gpu/src/cuda/row_filter.8.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16sc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.3.cu b/modules/gpufilters/src/cuda/row_filter.16sc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.3.cu
rename to modules/gpufilters/src/cuda/row_filter.16sc3.cu
index 9d131a959d..fcf40d81e3 100644
--- a/modules/gpu/src/cuda/row_filter.3.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16sc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.9.cu b/modules/gpufilters/src/cuda/row_filter.16sc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.9.cu
rename to modules/gpufilters/src/cuda/row_filter.16sc4.cu
index 954944b007..c5d472692b 100644
--- a/modules/gpu/src/cuda/row_filter.9.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16sc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.10.cu b/modules/gpufilters/src/cuda/row_filter.16uc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.10.cu
rename to modules/gpufilters/src/cuda/row_filter.16uc1.cu
index 4509fe4a72..02e125abc0 100644
--- a/modules/gpu/src/cuda/row_filter.10.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16uc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.11.cu b/modules/gpufilters/src/cuda/row_filter.16uc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.11.cu
rename to modules/gpufilters/src/cuda/row_filter.16uc3.cu
index 839b56c7c0..494c604b4b 100644
--- a/modules/gpu/src/cuda/row_filter.11.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16uc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.12.cu b/modules/gpufilters/src/cuda/row_filter.16uc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.12.cu
rename to modules/gpufilters/src/cuda/row_filter.16uc4.cu
index c4879dbe4c..1eb1ac25a6 100644
--- a/modules/gpu/src/cuda/row_filter.12.cu
+++ b/modules/gpufilters/src/cuda/row_filter.16uc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.5.cu b/modules/gpufilters/src/cuda/row_filter.32fc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.5.cu
rename to modules/gpufilters/src/cuda/row_filter.32fc1.cu
index dd1f2be135..bf577c6b71 100644
--- a/modules/gpu/src/cuda/row_filter.5.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32fc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.6.cu b/modules/gpufilters/src/cuda/row_filter.32fc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.6.cu
rename to modules/gpufilters/src/cuda/row_filter.32fc3.cu
index 548069d363..594fc04b5c 100644
--- a/modules/gpu/src/cuda/row_filter.6.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32fc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.7.cu b/modules/gpufilters/src/cuda/row_filter.32fc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.7.cu
rename to modules/gpufilters/src/cuda/row_filter.32fc4.cu
index 8c5c09ed93..5f2812bb31 100644
--- a/modules/gpu/src/cuda/row_filter.7.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32fc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.4.cu b/modules/gpufilters/src/cuda/row_filter.32sc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.4.cu
rename to modules/gpufilters/src/cuda/row_filter.32sc1.cu
index 0aae534ce7..67f3fb04c5 100644
--- a/modules/gpu/src/cuda/row_filter.4.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32sc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.13.cu b/modules/gpufilters/src/cuda/row_filter.32sc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.13.cu
rename to modules/gpufilters/src/cuda/row_filter.32sc3.cu
index c5da957a58..8e881a22ad 100644
--- a/modules/gpu/src/cuda/row_filter.13.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32sc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.14.cu b/modules/gpufilters/src/cuda/row_filter.32sc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.14.cu
rename to modules/gpufilters/src/cuda/row_filter.32sc4.cu
index d8ccb9a2ed..66f00cf06b 100644
--- a/modules/gpu/src/cuda/row_filter.14.cu
+++ b/modules/gpufilters/src/cuda/row_filter.32sc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.0.cu b/modules/gpufilters/src/cuda/row_filter.8uc1.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.0.cu
rename to modules/gpufilters/src/cuda/row_filter.8uc1.cu
index a1a8f36cad..c94b39f1b4 100644
--- a/modules/gpu/src/cuda/row_filter.0.cu
+++ b/modules/gpufilters/src/cuda/row_filter.8uc1.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.1.cu b/modules/gpufilters/src/cuda/row_filter.8uc3.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.1.cu
rename to modules/gpufilters/src/cuda/row_filter.8uc3.cu
index ab2248e1b2..1c924c10b2 100644
--- a/modules/gpu/src/cuda/row_filter.1.cu
+++ b/modules/gpufilters/src/cuda/row_filter.8uc3.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.2.cu b/modules/gpufilters/src/cuda/row_filter.8uc4.cu
similarity index 96%
rename from modules/gpu/src/cuda/row_filter.2.cu
rename to modules/gpufilters/src/cuda/row_filter.8uc4.cu
index 5aa2e2b80a..1ae9651a80 100644
--- a/modules/gpu/src/cuda/row_filter.2.cu
+++ b/modules/gpufilters/src/cuda/row_filter.8uc4.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,7 +42,7 @@
 
 #if !defined CUDA_DISABLER
 
-#include "row_filter.h"
+#include "row_filter.hpp"
 
 namespace filter
 {
diff --git a/modules/gpu/src/cuda/row_filter.h b/modules/gpufilters/src/cuda/row_filter.hpp
similarity index 98%
rename from modules/gpu/src/cuda/row_filter.h
rename to modules/gpufilters/src/cuda/row_filter.hpp
index f2da684cc1..3199a02e6a 100644
--- a/modules/gpu/src/cuda/row_filter.h
+++ b/modules/gpufilters/src/cuda/row_filter.hpp
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -41,13 +40,13 @@
 //
 //M*/
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 namespace row_filter
 {
@@ -187,38 +186,38 @@ namespace filter
         {
             {
                 0,
-                row_filter::caller< 1, T, D, BrdRowReflect101>,
-                row_filter::caller< 2, T, D, BrdRowReflect101>,
-                row_filter::caller< 3, T, D, BrdRowReflect101>,
-                row_filter::caller< 4, T, D, BrdRowReflect101>,
-                row_filter::caller< 5, T, D, BrdRowReflect101>,
-                row_filter::caller< 6, T, D, BrdRowReflect101>,
-                row_filter::caller< 7, T, D, BrdRowReflect101>,
-                row_filter::caller< 8, T, D, BrdRowReflect101>,
-                row_filter::caller< 9, T, D, BrdRowReflect101>,
-                row_filter::caller<10, T, D, BrdRowReflect101>,
-                row_filter::caller<11, T, D, BrdRowReflect101>,
-                row_filter::caller<12, T, D, BrdRowReflect101>,
-                row_filter::caller<13, T, D, BrdRowReflect101>,
-                row_filter::caller<14, T, D, BrdRowReflect101>,
-                row_filter::caller<15, T, D, BrdRowReflect101>,
-                row_filter::caller<16, T, D, BrdRowReflect101>,
-                row_filter::caller<17, T, D, BrdRowReflect101>,
-                row_filter::caller<18, T, D, BrdRowReflect101>,
-                row_filter::caller<19, T, D, BrdRowReflect101>,
-                row_filter::caller<20, T, D, BrdRowReflect101>,
-                row_filter::caller<21, T, D, BrdRowReflect101>,
-                row_filter::caller<22, T, D, BrdRowReflect101>,
-                row_filter::caller<23, T, D, BrdRowReflect101>,
-                row_filter::caller<24, T, D, BrdRowReflect101>,
-                row_filter::caller<25, T, D, BrdRowReflect101>,
-                row_filter::caller<26, T, D, BrdRowReflect101>,
-                row_filter::caller<27, T, D, BrdRowReflect101>,
-                row_filter::caller<28, T, D, BrdRowReflect101>,
-                row_filter::caller<29, T, D, BrdRowReflect101>,
-                row_filter::caller<30, T, D, BrdRowReflect101>,
-                row_filter::caller<31, T, D, BrdRowReflect101>,
-                row_filter::caller<32, T, D, BrdRowReflect101>
+                row_filter::caller< 1, T, D, BrdRowConstant>,
+                row_filter::caller< 2, T, D, BrdRowConstant>,
+                row_filter::caller< 3, T, D, BrdRowConstant>,
+                row_filter::caller< 4, T, D, BrdRowConstant>,
+                row_filter::caller< 5, T, D, BrdRowConstant>,
+                row_filter::caller< 6, T, D, BrdRowConstant>,
+                row_filter::caller< 7, T, D, BrdRowConstant>,
+                row_filter::caller< 8, T, D, BrdRowConstant>,
+                row_filter::caller< 9, T, D, BrdRowConstant>,
+                row_filter::caller<10, T, D, BrdRowConstant>,
+                row_filter::caller<11, T, D, BrdRowConstant>,
+                row_filter::caller<12, T, D, BrdRowConstant>,
+                row_filter::caller<13, T, D, BrdRowConstant>,
+                row_filter::caller<14, T, D, BrdRowConstant>,
+                row_filter::caller<15, T, D, BrdRowConstant>,
+                row_filter::caller<16, T, D, BrdRowConstant>,
+                row_filter::caller<17, T, D, BrdRowConstant>,
+                row_filter::caller<18, T, D, BrdRowConstant>,
+                row_filter::caller<19, T, D, BrdRowConstant>,
+                row_filter::caller<20, T, D, BrdRowConstant>,
+                row_filter::caller<21, T, D, BrdRowConstant>,
+                row_filter::caller<22, T, D, BrdRowConstant>,
+                row_filter::caller<23, T, D, BrdRowConstant>,
+                row_filter::caller<24, T, D, BrdRowConstant>,
+                row_filter::caller<25, T, D, BrdRowConstant>,
+                row_filter::caller<26, T, D, BrdRowConstant>,
+                row_filter::caller<27, T, D, BrdRowConstant>,
+                row_filter::caller<28, T, D, BrdRowConstant>,
+                row_filter::caller<29, T, D, BrdRowConstant>,
+                row_filter::caller<30, T, D, BrdRowConstant>,
+                row_filter::caller<31, T, D, BrdRowConstant>,
+                row_filter::caller<32, T, D, BrdRowConstant>
             },
             {
                 0,
@@ -255,41 +254,6 @@ namespace filter
                 row_filter::caller<31, T, D, BrdRowReplicate>,
                 row_filter::caller<32, T, D, BrdRowReplicate>
             },
-            {
-                0,
-                row_filter::caller< 1, T, D, BrdRowConstant>,
-                row_filter::caller< 2, T, D, BrdRowConstant>,
-                row_filter::caller< 3, T, D, BrdRowConstant>,
-                row_filter::caller< 4, T, D, BrdRowConstant>,
-                row_filter::caller< 5, T, D, BrdRowConstant>,
-                row_filter::caller< 6, T, D, BrdRowConstant>,
-                row_filter::caller< 7, T, D, BrdRowConstant>,
-                row_filter::caller< 8, T, D, BrdRowConstant>,
-                row_filter::caller< 9, T, D, BrdRowConstant>,
-                row_filter::caller<10, T, D, BrdRowConstant>,
-                row_filter::caller<11, T, D, BrdRowConstant>,
-                row_filter::caller<12, T, D, BrdRowConstant>,
-                row_filter::caller<13, T, D, BrdRowConstant>,
-                row_filter::caller<14, T, D, BrdRowConstant>,
-                row_filter::caller<15, T, D, BrdRowConstant>,
-                row_filter::caller<16, T, D, BrdRowConstant>,
-                row_filter::caller<17, T, D, BrdRowConstant>,
-                row_filter::caller<18, T, D, BrdRowConstant>,
-                row_filter::caller<19, T, D, BrdRowConstant>,
-                row_filter::caller<20, T, D, BrdRowConstant>,
-                row_filter::caller<21, T, D, BrdRowConstant>,
-                row_filter::caller<22, T, D, BrdRowConstant>,
-                row_filter::caller<23, T, D, BrdRowConstant>,
-                row_filter::caller<24, T, D, BrdRowConstant>,
-                row_filter::caller<25, T, D, BrdRowConstant>,
-                row_filter::caller<26, T, D, BrdRowConstant>,
-                row_filter::caller<27, T, D, BrdRowConstant>,
-                row_filter::caller<28, T, D, BrdRowConstant>,
-                row_filter::caller<29, T, D, BrdRowConstant>,
-                row_filter::caller<30, T, D, BrdRowConstant>,
-                row_filter::caller<31, T, D, BrdRowConstant>,
-                row_filter::caller<32, T, D, BrdRowConstant>
-            },
             {
                 0,
                 row_filter::caller< 1, T, D, BrdRowReflect>,
@@ -359,6 +323,41 @@ namespace filter
                 row_filter::caller<30, T, D, BrdRowWrap>,
                 row_filter::caller<31, T, D, BrdRowWrap>,
                 row_filter::caller<32, T, D, BrdRowWrap>
+            },
+            {
+                0,
+                row_filter::caller< 1, T, D, BrdRowReflect101>,
+                row_filter::caller< 2, T, D, BrdRowReflect101>,
+                row_filter::caller< 3, T, D, BrdRowReflect101>,
+                row_filter::caller< 4, T, D, BrdRowReflect101>,
+                row_filter::caller< 5, T, D, BrdRowReflect101>,
+                row_filter::caller< 6, T, D, BrdRowReflect101>,
+                row_filter::caller< 7, T, D, BrdRowReflect101>,
+                row_filter::caller< 8, T, D, BrdRowReflect101>,
+                row_filter::caller< 9, T, D, BrdRowReflect101>,
+                row_filter::caller<10, T, D, BrdRowReflect101>,
+                row_filter::caller<11, T, D, BrdRowReflect101>,
+                row_filter::caller<12, T, D, BrdRowReflect101>,
+                row_filter::caller<13, T, D, BrdRowReflect101>,
+                row_filter::caller<14, T, D, BrdRowReflect101>,
+                row_filter::caller<15, T, D, BrdRowReflect101>,
+                row_filter::caller<16, T, D, BrdRowReflect101>,
+                row_filter::caller<17, T, D, BrdRowReflect101>,
+                row_filter::caller<18, T, D, BrdRowReflect101>,
+                row_filter::caller<19, T, D, BrdRowReflect101>,
+                row_filter::caller<20, T, D, BrdRowReflect101>,
+                row_filter::caller<21, T, D, BrdRowReflect101>,
+                row_filter::caller<22, T, D, BrdRowReflect101>,
+                row_filter::caller<23, T, D, BrdRowReflect101>,
+                row_filter::caller<24, T, D, BrdRowReflect101>,
+                row_filter::caller<25, T, D, BrdRowReflect101>,
+                row_filter::caller<26, T, D, BrdRowReflect101>,
+                row_filter::caller<27, T, D, BrdRowReflect101>,
+                row_filter::caller<28, T, D, BrdRowReflect101>,
+                row_filter::caller<29, T, D, BrdRowReflect101>,
+                row_filter::caller<30, T, D, BrdRowReflect101>,
+                row_filter::caller<31, T, D, BrdRowReflect101>,
+                row_filter::caller<32, T, D, BrdRowReflect101>
             }
         };
 
diff --git a/modules/gpufilters/src/filtering.cpp b/modules/gpufilters/src/filtering.cpp
new file mode 100644
index 0000000000..0043174cba
--- /dev/null
+++ b/modules/gpufilters/src/filtering.cpp
@@ -0,0 +1,998 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<Filter> cv::gpu::createBoxFilter(int, int, Size, Point, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createLinearFilter(int, int, InputArray, Point, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createLaplacianFilter(int, int, int, double, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createSeparableLinearFilter(int, int, InputArray, InputArray, Point, int, int) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createDerivFilter(int, int, int, int, int, bool, double, int, int) { throw_no_cuda(); return Ptr<Filter>(); }
+Ptr<Filter> cv::gpu::createSobelFilter(int, int, int, int, int, double, int, int) { throw_no_cuda(); return Ptr<Filter>(); }
+Ptr<Filter> cv::gpu::createScharrFilter(int, int, int, int, double, int, int) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createGaussianFilter(int, int, Size, double, double, int, int) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createMorphologyFilter(int, int, InputArray, Point, int) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createBoxMaxFilter(int, Size, Point, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+Ptr<Filter> cv::gpu::createBoxMinFilter(int, Size, Point, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+
+Ptr<Filter> cv::gpu::createRowSumFilter(int, int, int, int, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+Ptr<Filter> cv::gpu::createColumnSumFilter(int, int, int, int, int, Scalar) { throw_no_cuda(); return Ptr<Filter>(); }
+
+#else
+
+namespace
+{
+    void normalizeAnchor(int& anchor, int ksize)
+    {
+        if (anchor < 0)
+            anchor = ksize >> 1;
+
+        CV_Assert( 0 <= anchor && anchor < ksize );
+    }
+
+    void normalizeAnchor(Point& anchor, Size ksize)
+    {
+        normalizeAnchor(anchor.x, ksize.width);
+        normalizeAnchor(anchor.y, ksize.height);
+    }
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Box Filter
+
+namespace
+{
+    class NPPBoxFilter : public Filter
+    {
+    public:
+        NPPBoxFilter(int srcType, int dstType, Size ksize, Point anchor, int borderMode, Scalar borderVal);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        typedef NppStatus (*nppFilterBox_t)(const Npp8u* pSrc, Npp32s nSrcStep, Npp8u* pDst, Npp32s nDstStep,
+                                            NppiSize oSizeROI, NppiSize oMaskSize, NppiPoint oAnchor);
+
+        Size ksize_;
+        Point anchor_;
+        int type_;
+        nppFilterBox_t func_;
+        int borderMode_;
+        Scalar borderVal_;
+        GpuMat srcBorder_;
+    };
+
+    NPPBoxFilter::NPPBoxFilter(int srcType, int dstType, Size ksize, Point anchor, int borderMode, Scalar borderVal) :
+        ksize_(ksize), anchor_(anchor), type_(srcType), borderMode_(borderMode), borderVal_(borderVal)
+    {
+        static const nppFilterBox_t funcs[] = {0, nppiFilterBox_8u_C1R, 0, 0, nppiFilterBox_8u_C4R};
+
+        CV_Assert( srcType == CV_8UC1 || srcType == CV_8UC4 );
+        CV_Assert( dstType == srcType );
+
+        normalizeAnchor(anchor_, ksize);
+
+        func_ = funcs[CV_MAT_CN(srcType)];
+    }
+
+    void NPPBoxFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == type_ );
+
+        gpu::copyMakeBorder(src, srcBorder_, ksize_.height, ksize_.height, ksize_.width, ksize_.width, borderMode_, borderVal_, _stream);
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        GpuMat srcRoi = srcBorder_(Rect(ksize_.width, ksize_.height, src.cols, src.rows));
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        NppiSize oMaskSize;
+        oMaskSize.height = ksize_.height;
+        oMaskSize.width = ksize_.width;
+
+        NppiPoint oAnchor;
+        oAnchor.x = anchor_.x;
+        oAnchor.y = anchor_.y;
+
+        nppSafeCall( func_(srcRoi.ptr<Npp8u>(), static_cast<int>(srcRoi.step),
+                           dst.ptr<Npp8u>(), static_cast<int>(dst.step),
+                           oSizeROI, oMaskSize, oAnchor) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+Ptr<Filter> cv::gpu::createBoxFilter(int srcType, int dstType, Size ksize, Point anchor, int borderMode, Scalar borderVal)
+{
+    if (dstType < 0)
+        dstType = srcType;
+
+    dstType = CV_MAKE_TYPE(CV_MAT_DEPTH(dstType), CV_MAT_CN(srcType));
+
+    return makePtr<NPPBoxFilter>(srcType, dstType, ksize, anchor, borderMode, borderVal);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Linear Filter
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename D>
+    void filter2D(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel,
+                  int kWidth, int kHeight, int anchorX, int anchorY,
+                  int borderMode, const float* borderValue, cudaStream_t stream);
+}}}
+
+namespace
+{
+    class LinearFilter : public Filter
+    {
+    public:
+        LinearFilter(int srcType, int dstType, InputArray kernel, Point anchor, int borderMode, Scalar borderVal);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        typedef void (*filter2D_t)(PtrStepSzb srcWhole, int ofsX, int ofsY, PtrStepSzb dst, const float* kernel,
+                                   int kWidth, int kHeight, int anchorX, int anchorY,
+                                   int borderMode, const float* borderValue, cudaStream_t stream);
+
+        GpuMat kernel_;
+        Point anchor_;
+        int type_;
+        filter2D_t func_;
+        int borderMode_;
+        Scalar_<float> borderVal_;
+    };
+
+    LinearFilter::LinearFilter(int srcType, int dstType, InputArray _kernel, Point anchor, int borderMode, Scalar borderVal) :
+        anchor_(anchor), type_(srcType), borderMode_(borderMode), borderVal_(borderVal)
+    {
+        const int sdepth = CV_MAT_DEPTH(srcType);
+        const int scn = CV_MAT_CN(srcType);
+
+        Mat kernel = _kernel.getMat();
+
+        CV_Assert( sdepth == CV_8U || sdepth == CV_16U || sdepth == CV_32F );
+        CV_Assert( scn == 1 || scn == 4 );
+        CV_Assert( dstType == srcType );
+        CV_Assert( kernel.channels() == 1 );
+        CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
+
+        Mat kernel32F;
+        kernel.convertTo(kernel32F, CV_32F);
+
+        kernel_ = gpu::createContinuous(kernel.size(), CV_32FC1);
+        kernel_.upload(kernel32F);
+
+        normalizeAnchor(anchor_, kernel.size());
+
+        switch (srcType)
+        {
+        case CV_8UC1:
+            func_ = cv::gpu::cudev::filter2D<uchar, uchar>;
+            break;
+        case CV_8UC4:
+            func_ = cv::gpu::cudev::filter2D<uchar4, uchar4>;
+            break;
+        case CV_16UC1:
+            func_ = cv::gpu::cudev::filter2D<ushort, ushort>;
+            break;
+        case CV_16UC4:
+            func_ = cv::gpu::cudev::filter2D<ushort4, ushort4>;
+            break;
+        case CV_32FC1:
+            func_ = cv::gpu::cudev::filter2D<float, float>;
+            break;
+        case CV_32FC4:
+            func_ = cv::gpu::cudev::filter2D<float4, float4>;
+            break;
+        }
+    }
+
+    void LinearFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == type_ );
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        Point ofs;
+        Size wholeSize;
+        src.locateROI(wholeSize, ofs);
+
+        GpuMat srcWhole(wholeSize, src.type(), src.datastart);
+
+        func_(srcWhole, ofs.x, ofs.y, dst, kernel_.ptr<float>(),
+              kernel_.cols, kernel_.rows, anchor_.x, anchor_.y,
+              borderMode_, borderVal_.val, StreamAccessor::getStream(_stream));
+    }
+}
+
+Ptr<Filter> cv::gpu::createLinearFilter(int srcType, int dstType, InputArray kernel, Point anchor, int borderMode, Scalar borderVal)
+{
+    if (dstType < 0)
+        dstType = srcType;
+
+    dstType = CV_MAKE_TYPE(CV_MAT_DEPTH(dstType), CV_MAT_CN(srcType));
+
+    return makePtr<LinearFilter>(srcType, dstType, kernel, anchor, borderMode, borderVal);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Laplacian Filter
+
+Ptr<Filter> cv::gpu::createLaplacianFilter(int srcType, int dstType, int ksize, double scale, int borderMode, Scalar borderVal)
+{
+    CV_Assert( ksize == 1 || ksize == 3 );
+
+    static const float K[2][9] =
+    {
+        {0.0f, 1.0f, 0.0f, 1.0f, -4.0f, 1.0f, 0.0f, 1.0f, 0.0f},
+        {2.0f, 0.0f, 2.0f, 0.0f, -8.0f, 0.0f, 2.0f, 0.0f, 2.0f}
+    };
+
+    Mat kernel(3, 3, CV_32FC1, (void*)K[ksize == 3]);
+    if (scale != 1)
+        kernel *= scale;
+
+    return gpu::createLinearFilter(srcType, dstType, kernel, Point(-1,-1), borderMode, borderVal);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Separable Linear Filter
+
+namespace filter
+{
+    template <typename T, typename D>
+    void linearRow(PtrStepSzb src, PtrStepSzb dst, const float* kernel, int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
+
+    template <typename T, typename D>
+    void linearColumn(PtrStepSzb src, PtrStepSzb dst, const float* kernel, int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
+}
+
+namespace
+{
+    class SeparableLinearFilter : public Filter
+    {
+    public:
+        SeparableLinearFilter(int srcType, int dstType,
+                              InputArray rowKernel, InputArray columnKernel,
+                              Point anchor, int rowBorderMode, int columnBorderMode);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, const float* kernel, int ksize, int anchor, int brd_type, int cc, cudaStream_t stream);
+
+        int srcType_, bufType_, dstType_;
+        GpuMat rowKernel_, columnKernel_;
+        func_t rowFilter_, columnFilter_;
+        Point anchor_;
+        int rowBorderMode_, columnBorderMode_;
+
+        GpuMat buf_;
+    };
+
+    SeparableLinearFilter::SeparableLinearFilter(int srcType, int dstType,
+                                                 InputArray _rowKernel, InputArray _columnKernel,
+                                                 Point anchor, int rowBorderMode, int columnBorderMode) :
+        srcType_(srcType), dstType_(dstType), anchor_(anchor), rowBorderMode_(rowBorderMode), columnBorderMode_(columnBorderMode)
+    {
+        static const func_t rowFilterFuncs[7][4] =
+        {
+            {filter::linearRow<uchar, float>, 0, filter::linearRow<uchar3, float3>, filter::linearRow<uchar4, float4>},
+            {0, 0, 0, 0},
+            {filter::linearRow<ushort, float>, 0, filter::linearRow<ushort3, float3>, filter::linearRow<ushort4, float4>},
+            {filter::linearRow<short, float>, 0, filter::linearRow<short3, float3>, filter::linearRow<short4, float4>},
+            {filter::linearRow<int, float>, 0, filter::linearRow<int3, float3>, filter::linearRow<int4, float4>},
+            {filter::linearRow<float, float>, 0, filter::linearRow<float3, float3>, filter::linearRow<float4, float4>},
+            {0, 0, 0, 0}
+        };
+
+        static const func_t columnFilterFuncs[7][4] =
+        {
+            {filter::linearColumn<float, uchar>, 0, filter::linearColumn<float3, uchar3>, filter::linearColumn<float4, uchar4>},
+            {0, 0, 0, 0},
+            {filter::linearColumn<float, ushort>, 0, filter::linearColumn<float3, ushort3>, filter::linearColumn<float4, ushort4>},
+            {filter::linearColumn<float, short>, 0, filter::linearColumn<float3, short3>, filter::linearColumn<float4, short4>},
+            {filter::linearColumn<float, int>, 0, filter::linearColumn<float3, int3>, filter::linearColumn<float4, int4>},
+            {filter::linearColumn<float, float>, 0, filter::linearColumn<float3, float3>, filter::linearColumn<float4, float4>},
+            {0, 0, 0, 0}
+        };
+
+        const int sdepth = CV_MAT_DEPTH(srcType);
+        const int cn = CV_MAT_CN(srcType);
+        const int ddepth = CV_MAT_DEPTH(dstType);
+
+        Mat rowKernel = _rowKernel.getMat();
+        Mat columnKernel = _columnKernel.getMat();
+
+        CV_Assert( sdepth <= CV_64F && cn <= 4 );
+        CV_Assert( rowKernel.channels() == 1 );
+        CV_Assert( columnKernel.channels() == 1 );
+        CV_Assert( rowBorderMode == BORDER_REFLECT101 || rowBorderMode == BORDER_REPLICATE || rowBorderMode == BORDER_CONSTANT || rowBorderMode == BORDER_REFLECT || rowBorderMode == BORDER_WRAP );
+        CV_Assert( columnBorderMode == BORDER_REFLECT101 || columnBorderMode == BORDER_REPLICATE || columnBorderMode == BORDER_CONSTANT || columnBorderMode == BORDER_REFLECT || columnBorderMode == BORDER_WRAP );
+
+        Mat kernel32F;
+
+        rowKernel.convertTo(kernel32F, CV_32F);
+        rowKernel_.upload(kernel32F.reshape(1, 1));
+
+        columnKernel.convertTo(kernel32F, CV_32F);
+        columnKernel_.upload(kernel32F.reshape(1, 1));
+
+        CV_Assert( rowKernel_.cols > 0 && rowKernel_.cols <= 32 );
+        CV_Assert( columnKernel_.cols > 0 && columnKernel_.cols <= 32 );
+
+        normalizeAnchor(anchor_.x, rowKernel_.cols);
+        normalizeAnchor(anchor_.y, columnKernel_.cols);
+
+        bufType_ = CV_MAKE_TYPE(CV_32F, cn);
+
+        rowFilter_ = rowFilterFuncs[sdepth][cn - 1];
+        CV_Assert( rowFilter_ != 0 );
+
+        columnFilter_ = columnFilterFuncs[ddepth][cn - 1];
+        CV_Assert( columnFilter_ != 0 );
+    }
+
+    void SeparableLinearFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == srcType_ );
+
+        _dst.create(src.size(), dstType_);
+        GpuMat dst = _dst.getGpuMat();
+
+        ensureSizeIsEnough(src.size(), bufType_, buf_);
+
+        DeviceInfo devInfo;
+        const int cc = devInfo.majorVersion() * 10 + devInfo.minorVersion();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        rowFilter_(src, buf_, rowKernel_.ptr<float>(), rowKernel_.cols, anchor_.x, rowBorderMode_, cc, stream);
+        columnFilter_(buf_, dst, columnKernel_.ptr<float>(), columnKernel_.cols, anchor_.y, columnBorderMode_, cc, stream);
+    }
+}
+
+Ptr<Filter> cv::gpu::createSeparableLinearFilter(int srcType, int dstType, InputArray rowKernel, InputArray columnKernel, Point anchor, int rowBorderMode, int columnBorderMode)
+{
+    if (dstType < 0)
+        dstType = srcType;
+
+    dstType = CV_MAKE_TYPE(CV_MAT_DEPTH(dstType), CV_MAT_CN(srcType));
+
+    if (columnBorderMode < 0)
+        columnBorderMode = rowBorderMode;
+
+    return makePtr<SeparableLinearFilter>(srcType, dstType, rowKernel, columnKernel, anchor, rowBorderMode, columnBorderMode);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Deriv Filter
+
+Ptr<Filter> cv::gpu::createDerivFilter(int srcType, int dstType, int dx, int dy, int ksize, bool normalize, double scale, int rowBorderMode, int columnBorderMode)
+{
+    Mat kx, ky;
+    getDerivKernels(kx, ky, dx, dy, ksize, normalize, CV_32F);
+
+    if (scale != 1)
+    {
+        // usually the smoothing part is the slowest to compute,
+        // so try to scale it instead of the faster differenciating part
+        if (dx == 0)
+            kx *= scale;
+        else
+            ky *= scale;
+    }
+
+    return gpu::createSeparableLinearFilter(srcType, dstType, kx, ky, Point(-1, -1), rowBorderMode, columnBorderMode);
+}
+
+Ptr<Filter> cv::gpu::createSobelFilter(int srcType, int dstType, int dx, int dy, int ksize, double scale, int rowBorderMode, int columnBorderMode)
+{
+    return gpu::createDerivFilter(srcType, dstType, dx, dy, ksize, false, scale, rowBorderMode, columnBorderMode);
+}
+
+Ptr<Filter> cv::gpu::createScharrFilter(int srcType, int dstType, int dx, int dy, double scale, int rowBorderMode, int columnBorderMode)
+{
+    return gpu::createDerivFilter(srcType, dstType, dx, dy, -1, false, scale, rowBorderMode, columnBorderMode);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Gaussian Filter
+
+Ptr<Filter> cv::gpu::createGaussianFilter(int srcType, int dstType, Size ksize, double sigma1, double sigma2, int rowBorderMode, int columnBorderMode)
+{
+    const int depth = CV_MAT_DEPTH(srcType);
+
+    if (sigma2 <= 0)
+        sigma2 = sigma1;
+
+    // automatic detection of kernel size from sigma
+    if (ksize.width <= 0 && sigma1 > 0)
+        ksize.width = cvRound(sigma1 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
+    if (ksize.height <= 0 && sigma2 > 0)
+        ksize.height = cvRound(sigma2 * (depth == CV_8U ? 3 : 4)*2 + 1) | 1;
+
+    CV_Assert( ksize.width > 0 && ksize.width % 2 == 1 && ksize.height > 0 && ksize.height % 2 == 1 );
+
+    sigma1 = std::max(sigma1, 0.0);
+    sigma2 = std::max(sigma2, 0.0);
+
+    Mat kx = getGaussianKernel(ksize.width, sigma1, CV_32F);
+    Mat ky;
+    if (ksize.height == ksize.width && std::abs(sigma1 - sigma2) < DBL_EPSILON)
+        ky = kx;
+    else
+        ky = getGaussianKernel(ksize.height, sigma2, CV_32F);
+
+    return createSeparableLinearFilter(srcType, dstType, kx, ky, Point(-1,-1), rowBorderMode, columnBorderMode);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Morphology Filter
+
+namespace
+{
+    class MorphologyFilter : public Filter
+    {
+    public:
+        MorphologyFilter(int op, int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        typedef NppStatus (*nppMorfFilter_t)(const Npp8u* pSrc, Npp32s nSrcStep, Npp8u* pDst, Npp32s nDstStep, NppiSize oSizeROI,
+                                             const Npp8u* pMask, NppiSize oMaskSize, NppiPoint oAnchor);
+
+        int type_;
+        GpuMat kernel_;
+        Point anchor_;
+        int iters_;
+        nppMorfFilter_t func_;
+
+        GpuMat srcBorder_;
+        GpuMat buf_;
+    };
+
+    MorphologyFilter::MorphologyFilter(int op, int srcType, InputArray _kernel, Point anchor, int iterations) :
+        type_(srcType), anchor_(anchor), iters_(iterations)
+    {
+        static const nppMorfFilter_t funcs[2][5] =
+        {
+            {0, nppiErode_8u_C1R, 0, 0, nppiErode_8u_C4R },
+            {0, nppiDilate_8u_C1R, 0, 0, nppiDilate_8u_C4R }
+        };
+
+        CV_Assert( op == MORPH_ERODE || op == MORPH_DILATE );
+        CV_Assert( srcType == CV_8UC1 || srcType == CV_8UC4 );
+
+        Mat kernel = _kernel.getMat();
+        Size ksize = !kernel.empty() ? _kernel.size() : Size(3, 3);
+
+        normalizeAnchor(anchor_, ksize);
+
+        if (kernel.empty())
+        {
+            kernel = getStructuringElement(MORPH_RECT, Size(1 + iters_ * 2, 1 + iters_ * 2));
+            anchor_ = Point(iters_, iters_);
+            iters_ = 1;
+        }
+        else if (iters_ > 1 && countNonZero(kernel) == (int) kernel.total())
+        {
+            anchor_ = Point(anchor_.x * iters_, anchor_.y * iters_);
+            kernel = getStructuringElement(MORPH_RECT,
+                                           Size(ksize.width + (iters_ - 1) * (ksize.width - 1),
+                                                ksize.height + (iters_ - 1) * (ksize.height - 1)),
+                                           anchor_);
+            iters_ = 1;
+        }
+
+        CV_Assert( kernel.channels() == 1 );
+
+        Mat kernel8U;
+        kernel.convertTo(kernel8U, CV_8U);
+
+        kernel_ = gpu::createContinuous(kernel.size(), CV_8UC1);
+        kernel_.upload(kernel8U);
+
+        func_ = funcs[op][CV_MAT_CN(srcType)];
+    }
+
+    void MorphologyFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == type_ );
+
+        Size ksize = kernel_.size();
+        gpu::copyMakeBorder(src, srcBorder_, ksize.height, ksize.height, ksize.width, ksize.width, BORDER_DEFAULT, Scalar(), _stream);
+
+        GpuMat srcRoi = srcBorder_(Rect(ksize.width, ksize.height, src.cols, src.rows));
+
+        GpuMat bufRoi;
+        if (iters_ > 1)
+        {
+            ensureSizeIsEnough(srcBorder_.size(), type_, buf_);
+            buf_.setTo(Scalar::all(0), _stream);
+            bufRoi = buf_(Rect(ksize.width, ksize.height, src.cols, src.rows));
+        }
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        NppiSize oMaskSize;
+        oMaskSize.height = ksize.height;
+        oMaskSize.width = ksize.width;
+
+        NppiPoint oAnchor;
+        oAnchor.x = anchor_.x;
+        oAnchor.y = anchor_.y;
+
+        nppSafeCall( func_(srcRoi.ptr<Npp8u>(), static_cast<int>(srcRoi.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step),
+                           oSizeROI, kernel_.ptr<Npp8u>(), oMaskSize, oAnchor) );
+
+        for(int i = 1; i < iters_; ++i)
+        {
+            dst.copyTo(bufRoi, _stream);
+
+            nppSafeCall( func_(bufRoi.ptr<Npp8u>(), static_cast<int>(bufRoi.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step),
+                               oSizeROI, kernel_.ptr<Npp8u>(), oMaskSize, oAnchor) );
+        }
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+namespace
+{
+    class MorphologyExFilter : public Filter
+    {
+    public:
+        MorphologyExFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+    protected:
+        Ptr<gpu::Filter> erodeFilter_, dilateFilter_;
+        GpuMat buf_;
+    };
+
+    MorphologyExFilter::MorphologyExFilter(int srcType, InputArray kernel, Point anchor, int iterations)
+    {
+        erodeFilter_ = gpu::createMorphologyFilter(MORPH_ERODE, srcType, kernel, anchor, iterations);
+        dilateFilter_ = gpu::createMorphologyFilter(MORPH_DILATE, srcType, kernel, anchor, iterations);
+    }
+
+    // MORPH_OPEN
+
+    class MorphologyOpenFilter : public MorphologyExFilter
+    {
+    public:
+        MorphologyOpenFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+    };
+
+    MorphologyOpenFilter::MorphologyOpenFilter(int srcType, InputArray kernel, Point anchor, int iterations) :
+        MorphologyExFilter(srcType, kernel, anchor, iterations)
+    {
+    }
+
+    void MorphologyOpenFilter::apply(InputArray src, OutputArray dst, Stream& stream)
+    {
+        erodeFilter_->apply(src, buf_, stream);
+        dilateFilter_->apply(buf_, dst, stream);
+    }
+
+    // MORPH_CLOSE
+
+    class MorphologyCloseFilter : public MorphologyExFilter
+    {
+    public:
+        MorphologyCloseFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+    };
+
+    MorphologyCloseFilter::MorphologyCloseFilter(int srcType, InputArray kernel, Point anchor, int iterations) :
+        MorphologyExFilter(srcType, kernel, anchor, iterations)
+    {
+    }
+
+    void MorphologyCloseFilter::apply(InputArray src, OutputArray dst, Stream& stream)
+    {
+        dilateFilter_->apply(src, buf_, stream);
+        erodeFilter_->apply(buf_, dst, stream);
+    }
+
+    // MORPH_GRADIENT
+
+    class MorphologyGradientFilter : public MorphologyExFilter
+    {
+    public:
+        MorphologyGradientFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+    };
+
+    MorphologyGradientFilter::MorphologyGradientFilter(int srcType, InputArray kernel, Point anchor, int iterations) :
+        MorphologyExFilter(srcType, kernel, anchor, iterations)
+    {
+    }
+
+    void MorphologyGradientFilter::apply(InputArray src, OutputArray dst, Stream& stream)
+    {
+        erodeFilter_->apply(src, buf_, stream);
+        dilateFilter_->apply(src, dst, stream);
+        gpu::subtract(dst, buf_, dst, noArray(), -1, stream);
+    }
+
+    // MORPH_TOPHAT
+
+    class MorphologyTophatFilter : public MorphologyExFilter
+    {
+    public:
+        MorphologyTophatFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+    };
+
+    MorphologyTophatFilter::MorphologyTophatFilter(int srcType, InputArray kernel, Point anchor, int iterations) :
+        MorphologyExFilter(srcType, kernel, anchor, iterations)
+    {
+    }
+
+    void MorphologyTophatFilter::apply(InputArray src, OutputArray dst, Stream& stream)
+    {
+        erodeFilter_->apply(src, dst, stream);
+        dilateFilter_->apply(dst, buf_, stream);
+        gpu::subtract(src, buf_, dst, noArray(), -1, stream);
+    }
+
+    // MORPH_BLACKHAT
+
+    class MorphologyBlackhatFilter : public MorphologyExFilter
+    {
+    public:
+        MorphologyBlackhatFilter(int srcType, InputArray kernel, Point anchor, int iterations);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+    };
+
+    MorphologyBlackhatFilter::MorphologyBlackhatFilter(int srcType, InputArray kernel, Point anchor, int iterations) :
+        MorphologyExFilter(srcType, kernel, anchor, iterations)
+    {
+    }
+
+    void MorphologyBlackhatFilter::apply(InputArray src, OutputArray dst, Stream& stream)
+    {
+        dilateFilter_->apply(src, dst, stream);
+        erodeFilter_->apply(dst, buf_, stream);
+        gpu::subtract(buf_, src, dst, noArray(), -1, stream);
+    }
+}
+
+Ptr<Filter> cv::gpu::createMorphologyFilter(int op, int srcType, InputArray kernel, Point anchor, int iterations)
+{
+    switch( op )
+    {
+    case MORPH_ERODE:
+    case MORPH_DILATE:
+        return makePtr<MorphologyFilter>(op, srcType, kernel, anchor, iterations);
+        break;
+
+    case MORPH_OPEN:
+        return makePtr<MorphologyOpenFilter>(srcType, kernel, anchor, iterations);
+        break;
+
+    case MORPH_CLOSE:
+        return makePtr<MorphologyCloseFilter>(srcType, kernel, anchor, iterations);
+        break;
+
+    case MORPH_GRADIENT:
+        return makePtr<MorphologyGradientFilter>(srcType, kernel, anchor, iterations);
+        break;
+
+    case MORPH_TOPHAT:
+        return makePtr<MorphologyTophatFilter>(srcType, kernel, anchor, iterations);
+        break;
+
+    case MORPH_BLACKHAT:
+        return makePtr<MorphologyBlackhatFilter>(srcType, kernel, anchor, iterations);
+        break;
+
+    default:
+        CV_Error(Error::StsBadArg, "Unknown morphological operation");
+        return Ptr<Filter>();
+    }
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Image Rank Filter
+
+namespace
+{
+    enum RankType
+    {
+        RANK_MAX,
+        RANK_MIN
+    };
+
+    class NPPRankFilter : public Filter
+    {
+    public:
+        NPPRankFilter(int op, int srcType, Size ksize, Point anchor, int borderMode, Scalar borderVal);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        typedef NppStatus (*nppFilterRank_t)(const Npp8u* pSrc, Npp32s nSrcStep, Npp8u* pDst, Npp32s nDstStep, NppiSize oSizeROI,
+                                             NppiSize oMaskSize, NppiPoint oAnchor);
+
+        int type_;
+        Size ksize_;
+        Point anchor_;
+        int borderMode_;
+        Scalar borderVal_;
+        nppFilterRank_t func_;
+
+        GpuMat srcBorder_;
+    };
+
+    NPPRankFilter::NPPRankFilter(int op, int srcType, Size ksize, Point anchor, int borderMode, Scalar borderVal) :
+        type_(srcType), ksize_(ksize), anchor_(anchor), borderMode_(borderMode), borderVal_(borderVal)
+    {
+        static const nppFilterRank_t maxFuncs[] = {0, nppiFilterMax_8u_C1R, 0, 0, nppiFilterMax_8u_C4R};
+        static const nppFilterRank_t minFuncs[] = {0, nppiFilterMin_8u_C1R, 0, 0, nppiFilterMin_8u_C4R};
+
+        CV_Assert( srcType == CV_8UC1 || srcType == CV_8UC4 );
+
+        normalizeAnchor(anchor_, ksize_);
+
+        if (op == RANK_MAX)
+            func_ = maxFuncs[CV_MAT_CN(srcType)];
+        else
+            func_ = minFuncs[CV_MAT_CN(srcType)];
+    }
+
+    void NPPRankFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == type_ );
+
+        gpu::copyMakeBorder(src, srcBorder_, ksize_.height, ksize_.height, ksize_.width, ksize_.width, borderMode_, borderVal_, _stream);
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        GpuMat srcRoi = srcBorder_(Rect(ksize_.width, ksize_.height, src.cols, src.rows));
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        NppiSize oMaskSize;
+        oMaskSize.height = ksize_.height;
+        oMaskSize.width = ksize_.width;
+
+        NppiPoint oAnchor;
+        oAnchor.x = anchor_.x;
+        oAnchor.y = anchor_.y;
+
+        nppSafeCall( func_(srcRoi.ptr<Npp8u>(), static_cast<int>(srcRoi.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step),
+                           oSizeROI, oMaskSize, oAnchor) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+Ptr<Filter> cv::gpu::createBoxMaxFilter(int srcType, Size ksize, Point anchor, int borderMode, Scalar borderVal)
+{
+    return makePtr<NPPRankFilter>(RANK_MAX, srcType, ksize, anchor, borderMode, borderVal);
+}
+
+Ptr<Filter> cv::gpu::createBoxMinFilter(int srcType, Size ksize, Point anchor, int borderMode, Scalar borderVal)
+{
+    return makePtr<NPPRankFilter>(RANK_MIN, srcType, ksize, anchor, borderMode, borderVal);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// 1D Sum Filter
+
+namespace
+{
+    class NppRowSumFilter : public Filter
+    {
+    public:
+        NppRowSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        int srcType_, dstType_;
+        int ksize_;
+        int anchor_;
+        int borderMode_;
+        Scalar borderVal_;
+
+        GpuMat srcBorder_;
+    };
+
+    NppRowSumFilter::NppRowSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal) :
+        srcType_(srcType), dstType_(dstType), ksize_(ksize), anchor_(anchor), borderMode_(borderMode), borderVal_(borderVal)
+    {
+        CV_Assert( srcType_ == CV_8UC1 );
+        CV_Assert( dstType_ == CV_32FC1 );
+
+        normalizeAnchor(anchor_, ksize_);
+    }
+
+    void NppRowSumFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == srcType_ );
+
+        gpu::copyMakeBorder(src, srcBorder_, 0, 0, ksize_, ksize_, borderMode_, borderVal_, _stream);
+
+        _dst.create(src.size(), dstType_);
+        GpuMat dst = _dst.getGpuMat();
+
+        GpuMat srcRoi = srcBorder_(Rect(ksize_, 0, src.cols, src.rows));
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        nppSafeCall( nppiSumWindowRow_8u32f_C1R(srcRoi.ptr<Npp8u>(), static_cast<int>(srcRoi.step),
+                                                dst.ptr<Npp32f>(), static_cast<int>(dst.step),
+                                                oSizeROI, ksize_, anchor_) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+Ptr<Filter> cv::gpu::createRowSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal)
+{
+    return makePtr<NppRowSumFilter>(srcType, dstType, ksize, anchor, borderMode, borderVal);
+}
+
+namespace
+{
+    class NppColumnSumFilter : public Filter
+    {
+    public:
+        NppColumnSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal);
+
+        void apply(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        int srcType_, dstType_;
+        int ksize_;
+        int anchor_;
+        int borderMode_;
+        Scalar borderVal_;
+
+        GpuMat srcBorder_;
+    };
+
+    NppColumnSumFilter::NppColumnSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal) :
+        srcType_(srcType), dstType_(dstType), ksize_(ksize), anchor_(anchor), borderMode_(borderMode), borderVal_(borderVal)
+    {
+        CV_Assert( srcType_ == CV_8UC1 );
+        CV_Assert( dstType_ == CV_32FC1 );
+
+        normalizeAnchor(anchor_, ksize_);
+    }
+
+    void NppColumnSumFilter::apply(InputArray _src, OutputArray _dst, Stream& _stream)
+    {
+        GpuMat src = _src.getGpuMat();
+        CV_Assert( src.type() == srcType_ );
+
+        gpu::copyMakeBorder(src, srcBorder_, ksize_, ksize_, 0, 0, borderMode_, borderVal_, _stream);
+
+        _dst.create(src.size(), dstType_);
+        GpuMat dst = _dst.getGpuMat();
+
+        GpuMat srcRoi = srcBorder_(Rect(0, ksize_, src.cols, src.rows));
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        nppSafeCall( nppiSumWindowColumn_8u32f_C1R(srcRoi.ptr<Npp8u>(), static_cast<int>(srcRoi.step),
+                                                   dst.ptr<Npp32f>(), static_cast<int>(dst.step),
+                                                   oSizeROI, ksize_, anchor_) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+Ptr<Filter> cv::gpu::createColumnSumFilter(int srcType, int dstType, int ksize, int anchor, int borderMode, Scalar borderVal)
+{
+    return makePtr<NppColumnSumFilter>(srcType, dstType, ksize, anchor, borderMode, borderVal);
+}
+
+#endif
diff --git a/modules/gpufilters/src/precomp.hpp b/modules/gpufilters/src/precomp.hpp
new file mode 100644
index 0000000000..c3d5e020d2
--- /dev/null
+++ b/modules/gpufilters/src/precomp.hpp
@@ -0,0 +1,54 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+
+#include "opencv2/gpufilters.hpp"
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/test/test_filters.cpp b/modules/gpufilters/test/test_filters.cpp
similarity index 73%
rename from modules/gpu/test/test_filters.cpp
rename to modules/gpufilters/test/test_filters.cpp
index f6ae4067aa..03bea05e6d 100644
--- a/modules/gpu/test/test_filters.cpp
+++ b/modules/gpufilters/test/test_filters.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 namespace
 {
     IMPLEMENT_PARAM_CLASS(KSize, cv::Size)
@@ -67,13 +70,14 @@ namespace
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // Blur
 
-PARAM_TEST_CASE(Blur, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, Anchor, UseRoi)
+PARAM_TEST_CASE(Blur, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, Anchor, BorderType, UseRoi)
 {
     cv::gpu::DeviceInfo devInfo;
     cv::Size size;
     int type;
     cv::Size ksize;
     cv::Point anchor;
+    int borderType;
     bool useRoi;
 
     virtual void SetUp()
@@ -83,7 +87,8 @@ PARAM_TEST_CASE(Blur, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, Anchor, Use
         type = GET_PARAM(2);
         ksize = GET_PARAM(3);
         anchor = GET_PARAM(4);
-        useRoi = GET_PARAM(5);
+        borderType = GET_PARAM(5);
+        useRoi = GET_PARAM(6);
 
         cv::gpu::setDevice(devInfo.deviceID());
     }
@@ -93,21 +98,190 @@ GPU_TEST_P(Blur, Accuracy)
 {
     cv::Mat src = randomMat(size, type);
 
+    cv::Ptr<cv::gpu::Filter> blurFilter = cv::gpu::createBoxFilter(src.type(), -1, ksize, anchor, borderType);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::blur(loadMat(src, useRoi), dst, ksize, anchor);
+    blurFilter->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
-    cv::blur(src, dst_gold, ksize, anchor);
+    cv::blur(src, dst_gold, ksize, anchor, borderType);
 
-    EXPECT_MAT_NEAR(getInnerROI(dst_gold, ksize), getInnerROI(dst, ksize), 1.0);
+    EXPECT_MAT_NEAR(dst_gold, dst, 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Blur, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Blur, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC4)),
     testing::Values(KSize(cv::Size(3, 3)), KSize(cv::Size(5, 5)), KSize(cv::Size(7, 7))),
     testing::Values(Anchor(cv::Point(-1, -1)), Anchor(cv::Point(0, 0)), Anchor(cv::Point(2, 2))),
+    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_CONSTANT), BorderType(cv::BORDER_REFLECT)),
+    WHOLE_SUBMAT));
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// Filter2D
+
+PARAM_TEST_CASE(Filter2D, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, Anchor, BorderType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    cv::Size ksize;
+    cv::Point anchor;
+    int borderType;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        ksize = GET_PARAM(3);
+        anchor = GET_PARAM(4);
+        borderType = GET_PARAM(5);
+        useRoi = GET_PARAM(6);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Filter2D, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat kernel = randomMat(cv::Size(ksize.width, ksize.height), CV_32FC1, 0.0, 1.0);
+
+    cv::Ptr<cv::gpu::Filter> filter2D = cv::gpu::createLinearFilter(src.type(), -1, kernel, anchor, borderType);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    filter2D->apply(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::filter2D(src, dst_gold, -1, kernel, anchor, 0, borderType);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, CV_MAT_DEPTH(type) == CV_32F ? 1e-1 : 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Filter2D, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC4)),
+    testing::Values(KSize(cv::Size(3, 3)), KSize(cv::Size(5, 5)), KSize(cv::Size(7, 7)), KSize(cv::Size(11, 11)), KSize(cv::Size(13, 13)), KSize(cv::Size(15, 15))),
+    testing::Values(Anchor(cv::Point(-1, -1)), Anchor(cv::Point(0, 0)), Anchor(cv::Point(2, 2))),
+    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_CONSTANT), BorderType(cv::BORDER_REFLECT)),
+    WHOLE_SUBMAT));
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// Laplacian
+
+PARAM_TEST_CASE(Laplacian, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    cv::Size ksize;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        ksize = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Laplacian, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::Ptr<cv::gpu::Filter> laplacian = cv::gpu::createLaplacianFilter(src.type(), -1, ksize.width);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    laplacian->apply(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::Laplacian(src, dst_gold, -1, ksize.width);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 0.0 : 1e-3);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Laplacian, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC4), MatType(CV_32FC1)),
+    testing::Values(KSize(cv::Size(1, 1)), KSize(cv::Size(3, 3))),
+    WHOLE_SUBMAT));
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// SeparableLinearFilter
+
+PARAM_TEST_CASE(SeparableLinearFilter, cv::gpu::DeviceInfo, cv::Size, MatDepth, Channels, KSize, Anchor, BorderType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    int cn;
+    cv::Size ksize;
+    cv::Point anchor;
+    int borderType;
+    bool useRoi;
+
+    int type;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        cn = GET_PARAM(3);
+        ksize = GET_PARAM(4);
+        anchor = GET_PARAM(5);
+        borderType = GET_PARAM(6);
+        useRoi = GET_PARAM(7);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        type = CV_MAKE_TYPE(depth, cn);
+    }
+};
+
+GPU_TEST_P(SeparableLinearFilter, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat rowKernel = randomMat(Size(ksize.width, 1), CV_32FC1, 0.0, 1.0);
+    cv::Mat columnKernel = randomMat(Size(ksize.height, 1), CV_32FC1, 0.0, 1.0);
+
+    cv::Ptr<cv::gpu::Filter> filter = cv::gpu::createSeparableLinearFilter(src.type(), -1, rowKernel, columnKernel, anchor, borderType);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    filter->apply(loadMat(src, useRoi), dst);
+
+    cv::Mat dst_gold;
+    cv::sepFilter2D(src, dst_gold, -1, rowKernel, columnKernel, anchor, 0, borderType);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 1.0 : 1e-2);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Filters, SeparableLinearFilter, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32F)),
+    IMAGE_CHANNELS,
+    testing::Values(KSize(cv::Size(3, 3)),
+                    KSize(cv::Size(7, 7)),
+                    KSize(cv::Size(13, 13)),
+                    KSize(cv::Size(15, 15)),
+                    KSize(cv::Size(17, 17)),
+                    KSize(cv::Size(23, 15)),
+                    KSize(cv::Size(31, 3))),
+    testing::Values(Anchor(cv::Point(-1, -1)), Anchor(cv::Point(0, 0)), Anchor(cv::Point(2, 2))),
+    testing::Values(BorderType(cv::BORDER_REFLECT101),
+                    BorderType(cv::BORDER_REPLICATE),
+                    BorderType(cv::BORDER_CONSTANT),
+                    BorderType(cv::BORDER_REFLECT)),
     WHOLE_SUBMAT));
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
@@ -152,16 +326,18 @@ GPU_TEST_P(Sobel, Accuracy)
 
     cv::Mat src = randomMat(size, type);
 
+    cv::Ptr<cv::gpu::Filter> sobel = cv::gpu::createSobelFilter(src.type(), -1, dx, dy, ksize.width, 1.0, borderType);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::Sobel(loadMat(src, useRoi), dst, -1, dx, dy, ksize.width, 1.0, borderType);
+    sobel->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
     cv::Sobel(src, dst_gold, -1, dx, dy, ksize.width, 1.0, 0.0, borderType);
 
-    EXPECT_MAT_NEAR(getInnerROI(dst_gold, ksize), getInnerROI(dst, ksize), CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.1);
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 0.0 : 0.1);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Sobel, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Sobel, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32F)),
@@ -215,16 +391,18 @@ GPU_TEST_P(Scharr, Accuracy)
 
     cv::Mat src = randomMat(size, type);
 
+    cv::Ptr<cv::gpu::Filter> scharr = cv::gpu::createScharrFilter(src.type(), -1, dx, dy, 1.0, borderType);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::Scharr(loadMat(src, useRoi), dst, -1, dx, dy, 1.0, borderType);
+    scharr->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
     cv::Scharr(src, dst_gold, -1, dx, dy, 1.0, 0.0, borderType);
 
-    EXPECT_MAT_NEAR(getInnerROI(dst_gold, cv::Size(3, 3)), getInnerROI(dst, cv::Size(3, 3)), CV_MAT_DEPTH(type) < CV_32F ? 0.0 : 0.1);
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 0.0 : 0.1);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Scharr, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Scharr, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32F)),
@@ -274,31 +452,18 @@ GPU_TEST_P(GaussianBlur, Accuracy)
     double sigma1 = randomDouble(0.1, 1.0);
     double sigma2 = randomDouble(0.1, 1.0);
 
-    if (ksize.height > 16 && !supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
-    {
-        try
-        {
-            cv::gpu::GpuMat dst;
-            cv::gpu::GaussianBlur(loadMat(src), dst, ksize, sigma1, sigma2, borderType);
-        }
-        catch (const cv::Exception& e)
-        {
-            ASSERT_EQ(CV_StsNotImplemented, e.code);
-        }
-    }
-    else
-    {
-        cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-        cv::gpu::GaussianBlur(loadMat(src, useRoi), dst, ksize, sigma1, sigma2, borderType);
+    cv::Ptr<cv::gpu::Filter> gauss = cv::gpu::createGaussianFilter(src.type(), -1, ksize, sigma1, sigma2, borderType);
 
-        cv::Mat dst_gold;
-        cv::GaussianBlur(src, dst_gold, ksize, sigma1, sigma2, borderType);
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    gauss->apply(loadMat(src, useRoi), dst);
 
-        EXPECT_MAT_NEAR(dst_gold, dst, 4.0);
-    }
+    cv::Mat dst_gold;
+    cv::GaussianBlur(src, dst_gold, ksize, sigma1, sigma2, borderType);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 4.0 : 1e-4);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, GaussianBlur, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, GaussianBlur, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32F)),
@@ -324,49 +489,6 @@ INSTANTIATE_TEST_CASE_P(GPU_Filter, GaussianBlur, testing::Combine(
                     BorderType(cv::BORDER_REFLECT)),
     WHOLE_SUBMAT));
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// Laplacian
-
-PARAM_TEST_CASE(Laplacian, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    cv::Size ksize;
-    bool useRoi;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        ksize = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Laplacian, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-
-    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::Laplacian(loadMat(src, useRoi), dst, -1, ksize.width);
-
-    cv::Mat dst_gold;
-    cv::Laplacian(src, dst_gold, -1, ksize.width);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() < CV_32F ? 0.0 : 1e-3);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Laplacian, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC4), MatType(CV_32FC1)),
-    testing::Values(KSize(cv::Size(1, 1)), KSize(cv::Size(3, 3))),
-    WHOLE_SUBMAT));
-
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // Erode
 
@@ -397,8 +519,10 @@ GPU_TEST_P(Erode, Accuracy)
     cv::Mat src = randomMat(size, type);
     cv::Mat kernel = cv::Mat::ones(3, 3, CV_8U);
 
+    cv::Ptr<cv::gpu::Filter> erode = cv::gpu::createMorphologyFilter(cv::MORPH_ERODE, src.type(), kernel, anchor, iterations);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::erode(loadMat(src, useRoi), dst, kernel, anchor, iterations);
+    erode->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
     cv::erode(src, dst_gold, kernel, anchor, iterations);
@@ -408,7 +532,7 @@ GPU_TEST_P(Erode, Accuracy)
     EXPECT_MAT_NEAR(getInnerROI(dst_gold, ksize), getInnerROI(dst, ksize), 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Erode, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Erode, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC4)),
@@ -446,8 +570,10 @@ GPU_TEST_P(Dilate, Accuracy)
     cv::Mat src = randomMat(size, type);
     cv::Mat kernel = cv::Mat::ones(3, 3, CV_8U);
 
+    cv::Ptr<cv::gpu::Filter> dilate = cv::gpu::createMorphologyFilter(cv::MORPH_DILATE, src.type(), kernel, anchor, iterations);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::dilate(loadMat(src, useRoi), dst, kernel, anchor, iterations);
+    dilate->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
     cv::dilate(src, dst_gold, kernel, anchor, iterations);
@@ -457,7 +583,7 @@ GPU_TEST_P(Dilate, Accuracy)
     EXPECT_MAT_NEAR(getInnerROI(dst_gold, ksize), getInnerROI(dst, ksize), 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Dilate, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, Dilate, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC4)),
@@ -468,8 +594,7 @@ INSTANTIATE_TEST_CASE_P(GPU_Filter, Dilate, testing::Combine(
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // MorphEx
 
-CV_ENUM(MorphOp, cv::MORPH_OPEN, cv::MORPH_CLOSE, cv::MORPH_GRADIENT, cv::MORPH_TOPHAT, cv::MORPH_BLACKHAT)
-#define ALL_MORPH_OPS testing::Values(MorphOp(cv::MORPH_OPEN), MorphOp(cv::MORPH_CLOSE), MorphOp(cv::MORPH_GRADIENT), MorphOp(cv::MORPH_TOPHAT), MorphOp(cv::MORPH_BLACKHAT))
+CV_ENUM(MorphOp, MORPH_OPEN, MORPH_CLOSE, MORPH_GRADIENT, MORPH_TOPHAT, MORPH_BLACKHAT)
 
 PARAM_TEST_CASE(MorphEx, cv::gpu::DeviceInfo, cv::Size, MatType, MorphOp, Anchor, Iterations, UseRoi)
 {
@@ -500,8 +625,10 @@ GPU_TEST_P(MorphEx, Accuracy)
     cv::Mat src = randomMat(size, type);
     cv::Mat kernel = cv::Mat::ones(3, 3, CV_8U);
 
+    cv::Ptr<cv::gpu::Filter> morph = cv::gpu::createMorphologyFilter(morphOp, src.type(), kernel, anchor, iterations);
+
     cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::morphologyEx(loadMat(src, useRoi), dst, morphOp, kernel, anchor, iterations);
+    morph->apply(loadMat(src, useRoi), dst);
 
     cv::Mat dst_gold;
     cv::morphologyEx(src, dst_gold, morphOp, kernel, anchor, iterations);
@@ -511,65 +638,13 @@ GPU_TEST_P(MorphEx, Accuracy)
     EXPECT_MAT_NEAR(getInnerROI(dst_gold, border), getInnerROI(dst, border), 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Filter, MorphEx, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Filters, MorphEx, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC4)),
-    ALL_MORPH_OPS,
+    MorphOp::all(),
     testing::Values(Anchor(cv::Point(-1, -1)), Anchor(cv::Point(0, 0)), Anchor(cv::Point(2, 2))),
     testing::Values(Iterations(1), Iterations(2), Iterations(3)),
     WHOLE_SUBMAT));
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// Filter2D
-
-PARAM_TEST_CASE(Filter2D, cv::gpu::DeviceInfo, cv::Size, MatType, KSize, Anchor, BorderType, UseRoi)
-{
-    cv::gpu::DeviceInfo devInfo;
-    cv::Size size;
-    int type;
-    cv::Size ksize;
-    cv::Point anchor;
-    int borderType;
-    bool useRoi;
-
-    cv::Mat img;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        size = GET_PARAM(1);
-        type = GET_PARAM(2);
-        ksize = GET_PARAM(3);
-        anchor = GET_PARAM(4);
-        borderType = GET_PARAM(5);
-        useRoi = GET_PARAM(6);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(Filter2D, Accuracy)
-{
-    cv::Mat src = randomMat(size, type);
-    cv::Mat kernel = randomMat(cv::Size(ksize.width, ksize.height), CV_32FC1, 0.0, 1.0);
-
-    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
-    cv::gpu::filter2D(loadMat(src, useRoi), dst, -1, kernel, anchor, borderType);
-
-    cv::Mat dst_gold;
-    cv::filter2D(src, dst_gold, -1, kernel, anchor, 0, borderType);
-
-    EXPECT_MAT_NEAR(dst_gold, dst, CV_MAT_DEPTH(type) == CV_32F ? 1e-1 : 1.0);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Filter, Filter2D, testing::Combine(
-    ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC4)),
-    testing::Values(KSize(cv::Size(3, 3)), KSize(cv::Size(5, 5)), KSize(cv::Size(7, 7)), KSize(cv::Size(11, 11)), KSize(cv::Size(13, 13)), KSize(cv::Size(15, 15))),
-    testing::Values(Anchor(cv::Point(-1, -1)), Anchor(cv::Point(0, 0)), Anchor(cv::Point(2, 2))),
-    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_CONSTANT), BorderType(cv::BORDER_REFLECT)),
-    WHOLE_SUBMAT));
-
 #endif // HAVE_CUDA
diff --git a/modules/gpufilters/test/test_main.cpp b/modules/gpufilters/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpufilters/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpufilters/test/test_precomp.hpp b/modules/gpufilters/test/test_precomp.hpp
new file mode 100644
index 0000000000..7dfe713d3a
--- /dev/null
+++ b/modules/gpufilters/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpufilters.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpuimgproc/CMakeLists.txt b/modules/gpuimgproc/CMakeLists.txt
new file mode 100644
index 0000000000..dfecf1d219
--- /dev/null
+++ b/modules/gpuimgproc/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpuimgproc)
+endif()
+
+set(the_description "GPU-accelerated Image Processing")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4100 /wd4324 /wd4512 /wd4515 -Wundef -Wmissing-declarations -Wshadow -Wunused-parameter)
+
+ocv_define_module(gpuimgproc opencv_imgproc OPTIONAL opencv_gpuarithm opencv_gpufilters)
diff --git a/modules/gpuimgproc/doc/color.rst b/modules/gpuimgproc/doc/color.rst
new file mode 100644
index 0000000000..b1e8e03bee
--- /dev/null
+++ b/modules/gpuimgproc/doc/color.rst
@@ -0,0 +1,131 @@
+Color space processing
+======================
+
+.. highlight:: cpp
+
+
+
+gpu::cvtColor
+-------------
+Converts an image from one color space to another.
+
+.. ocv:function:: void gpu::cvtColor(InputArray src, OutputArray dst, int code, int dcn = 0, Stream& stream = Stream::Null())
+
+    :param src: Source image with  ``CV_8U`` , ``CV_16U`` , or  ``CV_32F`` depth and 1, 3, or 4 channels.
+
+    :param dst: Destination image.
+
+    :param code: Color space conversion code. For details, see  :ocv:func:`cvtColor` .
+
+    :param dcn: Number of channels in the destination image. If the parameter is 0, the number of the channels is derived automatically from  ``src`` and the  ``code`` .
+
+    :param stream: Stream for the asynchronous version.
+
+3-channel color spaces (like ``HSV``, ``XYZ``, and so on) can be stored in a 4-channel image for better performance.
+
+.. seealso:: :ocv:func:`cvtColor`
+
+
+
+gpu::demosaicing
+----------------
+Converts an image from Bayer pattern to RGB or grayscale.
+
+.. ocv:function:: void gpu::demosaicing(InputArray src, OutputArray dst, int code, int dcn = -1, Stream& stream = Stream::Null())
+
+    :param src: Source image (8-bit or 16-bit single channel).
+
+    :param dst: Destination image.
+
+    :param code: Color space conversion code (see the description below).
+
+    :param dcn: Number of channels in the destination image. If the parameter is 0, the number of the channels is derived automatically from  ``src`` and the  ``code`` .
+
+    :param stream: Stream for the asynchronous version.
+
+The function can do the following transformations:
+
+* Demosaicing using bilinear interpolation
+
+    * ``COLOR_BayerBG2GRAY`` , ``COLOR_BayerGB2GRAY`` , ``COLOR_BayerRG2GRAY`` , ``COLOR_BayerGR2GRAY``
+
+    * ``COLOR_BayerBG2BGR`` , ``COLOR_BayerGB2BGR`` , ``COLOR_BayerRG2BGR`` , ``COLOR_BayerGR2BGR``
+
+* Demosaicing using Malvar-He-Cutler algorithm ([MHT2011]_)
+
+    * ``COLOR_BayerBG2GRAY_MHT`` , ``COLOR_BayerGB2GRAY_MHT`` , ``COLOR_BayerRG2GRAY_MHT`` , ``COLOR_BayerGR2GRAY_MHT``
+
+    * ``COLOR_BayerBG2BGR_MHT`` , ``COLOR_BayerGB2BGR_MHT`` , ``COLOR_BayerRG2BGR_MHT`` , ``COLOR_BayerGR2BGR_MHT``
+
+.. seealso:: :ocv:func:`cvtColor`
+
+
+
+gpu::swapChannels
+-----------------
+Exchanges the color channels of an image in-place.
+
+.. ocv:function:: void gpu::swapChannels(InputOutputArray image, const int dstOrder[4], Stream& stream = Stream::Null())
+
+    :param image: Source image. Supports only ``CV_8UC4`` type.
+
+    :param dstOrder: Integer array describing how channel values are permutated. The n-th entry of the array contains the number of the channel that is stored in the n-th channel of the output image. E.g. Given an RGBA image, aDstOrder = [3,2,1,0] converts this to ABGR channel order.
+
+    :param stream: Stream for the asynchronous version.
+
+The methods support arbitrary permutations of the original channels, including replication.
+
+
+
+gpu::gammaCorrection
+--------------------
+Routines for correcting image color gamma.
+
+.. ocv:function:: void gpu::gammaCorrection(InputArray src, OutputArray dst, bool forward = true, Stream& stream = Stream::Null())
+
+    :param src: Source image (3- or 4-channel 8 bit).
+
+    :param dst: Destination image.
+
+    :param forward: ``true`` for forward gamma correction or ``false`` for inverse gamma correction.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::alphaComp
+--------------
+Composites two images using alpha opacity values contained in each image.
+
+.. ocv:function:: void gpu::alphaComp(InputArray img1, InputArray img2, OutputArray dst, int alpha_op, Stream& stream = Stream::Null())
+
+    :param img1: First image. Supports ``CV_8UC4`` , ``CV_16UC4`` , ``CV_32SC4`` and ``CV_32FC4`` types.
+
+    :param img2: Second image. Must have the same size and the same type as ``img1`` .
+
+    :param dst: Destination image.
+
+    :param alpha_op: Flag specifying the alpha-blending operation:
+
+            * **ALPHA_OVER**
+            * **ALPHA_IN**
+            * **ALPHA_OUT**
+            * **ALPHA_ATOP**
+            * **ALPHA_XOR**
+            * **ALPHA_PLUS**
+            * **ALPHA_OVER_PREMUL**
+            * **ALPHA_IN_PREMUL**
+            * **ALPHA_OUT_PREMUL**
+            * **ALPHA_ATOP_PREMUL**
+            * **ALPHA_XOR_PREMUL**
+            * **ALPHA_PLUS_PREMUL**
+            * **ALPHA_PREMUL**
+
+    :param stream: Stream for the asynchronous version.
+
+.. note::
+
+   * An example demonstrating the use of alphaComp can be found at opencv_source_code/samples/gpu/alpha_comp.cpp
+
+
+.. [MHT2011] Pascal Getreuer, Malvar-He-Cutler Linear Image Demosaicking, Image Processing On Line, 2011
diff --git a/modules/gpuimgproc/doc/feature_detection.rst b/modules/gpuimgproc/doc/feature_detection.rst
new file mode 100644
index 0000000000..fc5e592cf8
--- /dev/null
+++ b/modules/gpuimgproc/doc/feature_detection.rst
@@ -0,0 +1,122 @@
+Feature Detection
+=================
+
+.. highlight:: cpp
+
+
+
+gpu::CornernessCriteria
+-----------------------
+.. ocv:class:: gpu::CornernessCriteria : public Algorithm
+
+Base class for Cornerness Criteria computation. ::
+
+    class CV_EXPORTS CornernessCriteria : public Algorithm
+    {
+    public:
+        virtual void compute(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+    };
+
+
+
+gpu::CornernessCriteria::compute
+--------------------------------
+Computes the cornerness criteria at each image pixel.
+
+.. ocv:function:: void gpu::CornernessCriteria::compute(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source image.
+
+    :param dst: Destination image containing cornerness values. It will have the same size as ``src`` and ``CV_32FC1`` type.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createHarrisCorner
+-----------------------
+Creates implementation for Harris cornerness criteria.
+
+.. ocv:function:: Ptr<CornernessCriteria> gpu::createHarrisCorner(int srcType, int blockSize, int ksize, double k, int borderType = BORDER_REFLECT101)
+
+    :param srcType: Input source type. Only  ``CV_8UC1`` and  ``CV_32FC1`` are supported for now.
+
+    :param blockSize: Neighborhood size.
+
+    :param ksize: Aperture parameter for the Sobel operator.
+
+    :param k: Harris detector free parameter.
+
+    :param borderType: Pixel extrapolation method. Only  ``BORDER_REFLECT101`` and  ``BORDER_REPLICATE`` are supported for now.
+
+.. seealso:: :ocv:func:`cornerHarris`
+
+
+
+gpu::createMinEigenValCorner
+----------------------------
+Creates implementation for the minimum eigen value of a 2x2 derivative covariation matrix (the cornerness criteria).
+
+.. ocv:function:: Ptr<CornernessCriteria> gpu::createMinEigenValCorner(int srcType, int blockSize, int ksize, int borderType = BORDER_REFLECT101)
+
+    :param srcType: Input source type. Only  ``CV_8UC1`` and  ``CV_32FC1`` are supported for now.
+
+    :param blockSize: Neighborhood size.
+
+    :param ksize: Aperture parameter for the Sobel operator.
+
+    :param borderType: Pixel extrapolation method. Only  ``BORDER_REFLECT101`` and  ``BORDER_REPLICATE`` are supported for now.
+
+.. seealso:: :ocv:func:`cornerMinEigenVal`
+
+
+
+gpu::CornersDetector
+--------------------
+.. ocv:class:: gpu::CornersDetector : public Algorithm
+
+Base class for Corners Detector. ::
+
+    class CV_EXPORTS CornersDetector : public Algorithm
+    {
+    public:
+        virtual void detect(InputArray image, OutputArray corners, InputArray mask = noArray()) = 0;
+    };
+
+
+
+gpu::CornersDetector::detect
+----------------------------
+Determines strong corners on an image.
+
+.. ocv:function:: void gpu::CornersDetector::detect(InputArray image, OutputArray corners, InputArray mask = noArray())
+
+    :param image: Input 8-bit or floating-point 32-bit, single-channel image.
+
+    :param corners: Output vector of detected corners (1-row matrix with CV_32FC2 type with corners positions).
+
+    :param mask: Optional region of interest. If the image is not empty (it needs to have the type  ``CV_8UC1``  and the same size as  ``image`` ), it  specifies the region in which the corners are detected.
+
+
+
+gpu::createGoodFeaturesToTrackDetector
+--------------------------------------
+Creates implementation for :ocv:class:`gpu::CornersDetector` .
+
+.. ocv:function:: Ptr<CornersDetector> gpu::createGoodFeaturesToTrackDetector(int srcType, int maxCorners = 1000, double qualityLevel = 0.01, double minDistance = 0.0, int blockSize = 3, bool useHarrisDetector = false, double harrisK = 0.04)
+
+    :param srcType: Input source type. Only  ``CV_8UC1`` and  ``CV_32FC1`` are supported for now.
+
+    :param maxCorners: Maximum number of corners to return. If there are more corners than are found, the strongest of them is returned.
+
+    :param qualityLevel: Parameter characterizing the minimal accepted quality of image corners. The parameter value is multiplied by the best corner quality measure, which is the minimal eigenvalue (see  :ocv:func:`cornerMinEigenVal` ) or the Harris function response (see  :ocv:func:`cornerHarris` ). The corners with the quality measure less than the product are rejected. For example, if the best corner has the quality measure = 1500, and the  ``qualityLevel=0.01`` , then all the corners with the quality measure less than 15 are rejected.
+
+    :param minDistance: Minimum possible Euclidean distance between the returned corners.
+
+    :param blockSize: Size of an average block for computing a derivative covariation matrix over each pixel neighborhood. See  :ocv:func:`cornerEigenValsAndVecs` .
+
+    :param useHarrisDetector: Parameter indicating whether to use a Harris detector (see :ocv:func:`cornerHarris`) or :ocv:func:`cornerMinEigenVal`.
+
+    :param harrisK: Free parameter of the Harris detector.
+
+.. seealso:: :ocv:func:`goodFeaturesToTrack`
diff --git a/modules/gpuimgproc/doc/gpuimgproc.rst b/modules/gpuimgproc/doc/gpuimgproc.rst
new file mode 100644
index 0000000000..827b735f52
--- /dev/null
+++ b/modules/gpuimgproc/doc/gpuimgproc.rst
@@ -0,0 +1,12 @@
+********************************************
+gpuimgproc. GPU-accelerated Image Processing
+********************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    color
+    histogram
+    hough
+    feature_detection
+    imgproc
diff --git a/modules/gpuimgproc/doc/histogram.rst b/modules/gpuimgproc/doc/histogram.rst
new file mode 100644
index 0000000000..dfdf322864
--- /dev/null
+++ b/modules/gpuimgproc/doc/histogram.rst
@@ -0,0 +1,150 @@
+Histogram Calculation
+=====================
+
+.. highlight:: cpp
+
+
+
+gpu::calcHist
+-------------
+Calculates histogram for one channel 8-bit image.
+
+.. ocv:function:: void gpu::calcHist(InputArray src, OutputArray hist, Stream& stream = Stream::Null())
+
+    :param src: Source image with ``CV_8UC1`` type.
+
+    :param hist: Destination histogram with one row, 256 columns, and the  ``CV_32SC1`` type.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::equalizeHist
+-----------------
+Equalizes the histogram of a grayscale image.
+
+.. ocv:function:: void gpu::equalizeHist(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::equalizeHist(InputArray src, OutputArray dst, InputOutputArray buf, Stream& stream = Stream::Null())
+
+    :param src: Source image with ``CV_8UC1`` type.
+
+    :param dst: Destination image.
+
+    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`equalizeHist`
+
+
+
+gpu::CLAHE
+----------
+.. ocv:class:: gpu::CLAHE : public cv::CLAHE
+
+Base class for Contrast Limited Adaptive Histogram Equalization. ::
+
+    class CV_EXPORTS CLAHE : public cv::CLAHE
+    {
+    public:
+        using cv::CLAHE::apply;
+        virtual void apply(InputArray src, OutputArray dst, Stream& stream) = 0;
+    };
+
+
+
+gpu::CLAHE::apply
+-----------------
+Equalizes the histogram of a grayscale image using Contrast Limited Adaptive Histogram Equalization.
+
+.. ocv:function:: void gpu::CLAHE::apply(InputArray src, OutputArray dst)
+
+.. ocv:function:: void gpu::CLAHE::apply(InputArray src, OutputArray dst, Stream& stream)
+
+    :param src: Source image with ``CV_8UC1`` type.
+
+    :param dst: Destination image.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createCLAHE
+----------------
+Creates implementation for :ocv:class:`gpu::CLAHE` .
+
+.. ocv:function:: Ptr<gpu::CLAHE> createCLAHE(double clipLimit = 40.0, Size tileGridSize = Size(8, 8))
+
+    :param clipLimit: Threshold for contrast limiting.
+
+    :param tileGridSize: Size of grid for histogram equalization. Input image will be divided into equally sized rectangular tiles. ``tileGridSize`` defines the number of tiles in row and column.
+
+
+
+
+gpu::evenLevels
+---------------
+Computes levels with even distribution.
+
+.. ocv:function:: void gpu::evenLevels(OutputArray levels, int nLevels, int lowerLevel, int upperLevel)
+
+    :param levels: Destination array.  ``levels`` has 1 row, ``nLevels`` columns, and the ``CV_32SC1`` type.
+
+    :param nLevels: Number of computed levels.  ``nLevels`` must be at least 2.
+
+    :param lowerLevel: Lower boundary value of the lowest level.
+
+    :param upperLevel: Upper boundary value of the greatest level.
+
+
+
+gpu::histEven
+-------------
+Calculates a histogram with evenly distributed bins.
+
+.. ocv:function:: void gpu::histEven(InputArray src, OutputArray hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histEven(InputArray src, OutputArray hist, InputOutputArray buf, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histEven(InputArray src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histEven(InputArray src, GpuMat hist[4], InputOutputArray buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null())
+
+    :param src: Source image. ``CV_8U``, ``CV_16U``, or ``CV_16S`` depth and 1 or 4 channels are supported. For a four-channel image, all channels are processed separately.
+
+    :param hist: Destination histogram with one row, ``histSize`` columns, and the ``CV_32S`` type.
+
+    :param histSize: Size of the histogram.
+
+    :param lowerLevel: Lower boundary of lowest-level bin.
+
+    :param upperLevel: Upper boundary of highest-level bin.
+
+    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::histRange
+--------------
+Calculates a histogram with bins determined by the ``levels`` array.
+
+.. ocv:function:: void gpu::histRange(InputArray src, OutputArray hist, InputArray levels, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histRange(InputArray src, OutputArray hist, InputArray levels, InputOutputArray buf, Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histRange(InputArray src, GpuMat hist[4], const GpuMat levels[4], Stream& stream = Stream::Null())
+
+.. ocv:function:: void gpu::histRange(InputArray src, GpuMat hist[4], const GpuMat levels[4], InputOutputArray buf, Stream& stream = Stream::Null())
+
+    :param src: Source image. ``CV_8U`` , ``CV_16U`` , or  ``CV_16S`` depth and 1 or 4 channels are supported. For a four-channel image, all channels are processed separately.
+
+    :param hist: Destination histogram with one row, ``(levels.cols-1)`` columns, and the  ``CV_32SC1`` type.
+
+    :param levels: Number of levels in the histogram.
+
+    :param buf: Optional buffer to avoid extra memory allocations (for many calls with the same sizes).
+
+    :param stream: Stream for the asynchronous version.
diff --git a/modules/gpuimgproc/doc/hough.rst b/modules/gpuimgproc/doc/hough.rst
new file mode 100644
index 0000000000..ec7117198f
--- /dev/null
+++ b/modules/gpuimgproc/doc/hough.rst
@@ -0,0 +1,236 @@
+Hough Transform
+===============
+
+.. highlight:: cpp
+
+
+
+gpu::HoughLinesDetector
+-----------------------
+.. ocv:class:: gpu::HoughLinesDetector : public Algorithm
+
+Base class for lines detector algorithm. ::
+
+    class CV_EXPORTS HoughLinesDetector : public Algorithm
+    {
+    public:
+        virtual void detect(InputArray src, OutputArray lines) = 0;
+        virtual void downloadResults(InputArray d_lines, OutputArray h_lines, OutputArray h_votes = noArray()) = 0;
+
+        virtual void setRho(float rho) = 0;
+        virtual float getRho() const = 0;
+
+        virtual void setTheta(float theta) = 0;
+        virtual float getTheta() const = 0;
+
+        virtual void setThreshold(int threshold) = 0;
+        virtual int getThreshold() const = 0;
+
+        virtual void setDoSort(bool doSort) = 0;
+        virtual bool getDoSort() const = 0;
+
+        virtual void setMaxLines(int maxLines) = 0;
+        virtual int getMaxLines() const = 0;
+    };
+
+
+
+gpu::HoughLinesDetector::detect
+-------------------------------
+Finds lines in a binary image using the classical Hough transform.
+
+.. ocv:function:: void gpu::HoughLinesDetector::detect(InputArray src, OutputArray lines)
+
+    :param src: 8-bit, single-channel binary source image.
+
+    :param lines: Output vector of lines. Each line is represented by a two-element vector  :math:`(\rho, \theta)` .  :math:`\rho`  is the distance from the coordinate origin  :math:`(0,0)`  (top-left corner of the image).  :math:`\theta`  is the line rotation angle in radians ( :math:`0 \sim \textrm{vertical line}, \pi/2 \sim \textrm{horizontal line}` ).
+
+.. seealso:: :ocv:func:`HoughLines`
+
+
+
+gpu::HoughLinesDetector::downloadResults
+----------------------------------------
+Downloads results from :ocv:func:`gpu::HoughLinesDetector::detect` to host memory.
+
+.. ocv:function:: void gpu::HoughLinesDetector::downloadResults(InputArray d_lines, OutputArray h_lines, OutputArray h_votes = noArray())
+
+    :param d_lines: Result of :ocv:func:`gpu::HoughLinesDetector::detect` .
+
+    :param h_lines: Output host array.
+
+    :param h_votes: Optional output array for line's votes.
+
+
+
+gpu::createHoughLinesDetector
+-----------------------------
+Creates implementation for :ocv:class:`gpu::HoughLinesDetector` .
+
+.. ocv:function:: Ptr<HoughLinesDetector> gpu::createHoughLinesDetector(float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096)
+
+    :param rho: Distance resolution of the accumulator in pixels.
+
+    :param theta: Angle resolution of the accumulator in radians.
+
+    :param threshold: Accumulator threshold parameter. Only those lines are returned that get enough votes ( :math:`>\texttt{threshold}` ).
+
+    :param doSort: Performs lines sort by votes.
+
+    :param maxLines: Maximum number of output lines.
+
+
+
+gpu::HoughSegmentDetector
+-------------------------
+.. ocv:class:: gpu::HoughSegmentDetector : public Algorithm
+
+Base class for line segments detector algorithm. ::
+
+    class CV_EXPORTS HoughSegmentDetector : public Algorithm
+    {
+    public:
+        virtual void detect(InputArray src, OutputArray lines) = 0;
+
+        virtual void setRho(float rho) = 0;
+        virtual float getRho() const = 0;
+
+        virtual void setTheta(float theta) = 0;
+        virtual float getTheta() const = 0;
+
+        virtual void setMinLineLength(int minLineLength) = 0;
+        virtual int getMinLineLength() const = 0;
+
+        virtual void setMaxLineGap(int maxLineGap) = 0;
+        virtual int getMaxLineGap() const = 0;
+
+        virtual void setMaxLines(int maxLines) = 0;
+        virtual int getMaxLines() const = 0;
+    };
+
+.. note::
+
+   * An example using the Hough segment detector can be found at opencv_source_code/samples/gpu/houghlines.cpp
+
+
+gpu::HoughSegmentDetector::detect
+---------------------------------
+Finds line segments in a binary image using the probabilistic Hough transform.
+
+.. ocv:function:: void gpu::HoughSegmentDetector::detect(InputArray src, OutputArray lines)
+
+    :param src: 8-bit, single-channel binary source image.
+
+    :param lines: Output vector of lines. Each line is represented by a 4-element vector  :math:`(x_1, y_1, x_2, y_2)` , where  :math:`(x_1,y_1)`  and  :math:`(x_2, y_2)`  are the ending points of each detected line segment.
+
+.. seealso:: :ocv:func:`HoughLinesP`
+
+
+
+gpu::createHoughSegmentDetector
+-------------------------------
+Creates implementation for :ocv:class:`gpu::HoughSegmentDetector` .
+
+.. ocv:function:: Ptr<HoughSegmentDetector> gpu::createHoughSegmentDetector(float rho, float theta, int minLineLength, int maxLineGap, int maxLines = 4096)
+
+    :param rho: Distance resolution of the accumulator in pixels.
+
+    :param theta: Angle resolution of the accumulator in radians.
+
+    :param minLineLength: Minimum line length. Line segments shorter than that are rejected.
+
+    :param maxLineGap: Maximum allowed gap between points on the same line to link them.
+
+    :param maxLines: Maximum number of output lines.
+
+
+
+gpu::HoughCirclesDetector
+-------------------------
+.. ocv:class:: gpu::HoughCirclesDetector : public Algorithm
+
+Base class for circles detector algorithm. ::
+
+    class CV_EXPORTS HoughCirclesDetector : public Algorithm
+    {
+    public:
+        virtual void detect(InputArray src, OutputArray circles) = 0;
+
+        virtual void setDp(float dp) = 0;
+        virtual float getDp() const = 0;
+
+        virtual void setMinDist(float minDist) = 0;
+        virtual float getMinDist() const = 0;
+
+        virtual void setCannyThreshold(int cannyThreshold) = 0;
+        virtual int getCannyThreshold() const = 0;
+
+        virtual void setVotesThreshold(int votesThreshold) = 0;
+        virtual int getVotesThreshold() const = 0;
+
+        virtual void setMinRadius(int minRadius) = 0;
+        virtual int getMinRadius() const = 0;
+
+        virtual void setMaxRadius(int maxRadius) = 0;
+        virtual int getMaxRadius() const = 0;
+
+        virtual void setMaxCircles(int maxCircles) = 0;
+        virtual int getMaxCircles() const = 0;
+    };
+
+
+
+gpu::HoughCirclesDetector::detect
+---------------------------------
+Finds circles in a grayscale image using the Hough transform.
+
+.. ocv:function:: void gpu::HoughCirclesDetector::detect(InputArray src, OutputArray circles)
+
+    :param src: 8-bit, single-channel grayscale input image.
+
+    :param circles: Output vector of found circles. Each vector is encoded as a 3-element floating-point vector  :math:`(x, y, radius)` .
+
+.. seealso:: :ocv:func:`HoughCircles`
+
+
+
+gpu::createHoughCirclesDetector
+-------------------------------
+Creates implementation for :ocv:class:`gpu::HoughCirclesDetector` .
+
+.. ocv:function:: Ptr<HoughCirclesDetector> gpu::createHoughCirclesDetector(float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096)
+
+    :param dp: Inverse ratio of the accumulator resolution to the image resolution. For example, if  ``dp=1`` , the accumulator has the same resolution as the input image. If  ``dp=2`` , the accumulator has half as big width and height.
+
+    :param minDist: Minimum distance between the centers of the detected circles. If the parameter is too small, multiple neighbor circles may be falsely detected in addition to a true one. If it is too large, some circles may be missed.
+
+    :param cannyThreshold: The higher threshold of the two passed to Canny edge detector (the lower one is twice smaller).
+
+    :param votesThreshold: The accumulator threshold for the circle centers at the detection stage. The smaller it is, the more false circles may be detected.
+
+    :param minRadius: Minimum circle radius.
+
+    :param maxRadius: Maximum circle radius.
+
+    :param maxCircles: Maximum number of output circles.
+
+
+
+gpu::createGeneralizedHoughBallard
+----------------------------------
+Creates implementation for generalized hough transform from [Ballard1981]_ .
+
+.. ocv:function:: Ptr<GeneralizedHoughBallard> gpu::createGeneralizedHoughBallard()
+
+
+
+gpu::createGeneralizedHoughGuil
+-------------------------------
+Creates implementation for generalized hough transform from [Guil1999]_ .
+
+.. ocv:function:: Ptr<GeneralizedHoughGuil> gpu::createGeneralizedHoughGuil()
+
+
+
+.. [Ballard1981] Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
+.. [Guil1999] Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
diff --git a/modules/gpuimgproc/doc/imgproc.rst b/modules/gpuimgproc/doc/imgproc.rst
new file mode 100644
index 0000000000..776c339a53
--- /dev/null
+++ b/modules/gpuimgproc/doc/imgproc.rst
@@ -0,0 +1,231 @@
+Image Processing
+================
+
+.. highlight:: cpp
+
+
+
+gpu::CannyEdgeDetector
+----------------------
+.. ocv:class:: gpu::CannyEdgeDetector : public Algorithm
+
+Base class for Canny Edge Detector. ::
+
+    class CV_EXPORTS CannyEdgeDetector : public Algorithm
+    {
+    public:
+        virtual void detect(InputArray image, OutputArray edges) = 0;
+        virtual void detect(InputArray dx, InputArray dy, OutputArray edges) = 0;
+
+        virtual void setLowThreshold(double low_thresh) = 0;
+        virtual double getLowThreshold() const = 0;
+
+        virtual void setHighThreshold(double high_thresh) = 0;
+        virtual double getHighThreshold() const = 0;
+
+        virtual void setAppertureSize(int apperture_size) = 0;
+        virtual int getAppertureSize() const = 0;
+
+        virtual void setL2Gradient(bool L2gradient) = 0;
+        virtual bool getL2Gradient() const = 0;
+    };
+
+
+
+gpu::CannyEdgeDetector::detect
+------------------------------
+Finds edges in an image using the [Canny86]_ algorithm.
+
+.. ocv:function:: void gpu::CannyEdgeDetector::detect(InputArray image, OutputArray edges)
+
+.. ocv:function:: void gpu::CannyEdgeDetector::detect(InputArray dx, InputArray dy, OutputArray edges)
+
+    :param image: Single-channel 8-bit input image.
+
+    :param dx: First derivative of image in the vertical direction. Support only ``CV_32S`` type.
+
+    :param dy: First derivative of image in the horizontal direction. Support only ``CV_32S`` type.
+
+    :param edges: Output edge map. It has the same size and type as  ``image`` .
+
+
+
+gpu::createCannyEdgeDetector
+----------------------------
+Creates implementation for :ocv:class:`gpu::CannyEdgeDetector` .
+
+.. ocv:function:: Ptr<CannyEdgeDetector> gpu::createCannyEdgeDetector(double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false)
+
+    :param low_thresh: First threshold for the hysteresis procedure.
+
+    :param high_thresh: Second threshold for the hysteresis procedure.
+
+    :param apperture_size: Aperture size for the  :ocv:func:`Sobel`  operator.
+
+    :param L2gradient: Flag indicating whether a more accurate  :math:`L_2`  norm  :math:`=\sqrt{(dI/dx)^2 + (dI/dy)^2}`  should be used to compute the image gradient magnitude ( ``L2gradient=true`` ), or a faster default  :math:`L_1`  norm  :math:`=|dI/dx|+|dI/dy|`  is enough ( ``L2gradient=false`` ).
+
+
+
+gpu::meanShiftFiltering
+-----------------------
+Performs mean-shift filtering for each point of the source image.
+
+.. ocv:function:: void gpu::meanShiftFiltering(InputArray src, OutputArray dst, int sp, int sr, TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1), Stream& stream = Stream::Null())
+
+    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
+
+    :param dst: Destination image containing the color of mapped points. It has the same size and type as  ``src`` .
+
+    :param sp: Spatial window radius.
+
+    :param sr: Color window radius.
+
+    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
+
+It maps each point of the source image into another point. As a result, you have a new color and new position of each point.
+
+
+
+gpu::meanShiftProc
+------------------
+Performs a mean-shift procedure and stores information about processed points (their colors and positions) in two images.
+
+.. ocv:function:: void gpu::meanShiftProc(InputArray src, OutputArray dstr, OutputArray dstsp, int sp, int sr, TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1), Stream& stream = Stream::Null())
+
+    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
+
+    :param dstr: Destination image containing the color of mapped points. The size and type is the same as  ``src`` .
+
+    :param dstsp: Destination image containing the position of mapped points. The size is the same as  ``src`` size. The type is  ``CV_16SC2`` .
+
+    :param sp: Spatial window radius.
+
+    :param sr: Color window radius.
+
+    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
+
+.. seealso:: :ocv:func:`gpu::meanShiftFiltering`
+
+
+
+gpu::meanShiftSegmentation
+--------------------------
+Performs a mean-shift segmentation of the source image and eliminates small segments.
+
+.. ocv:function:: void gpu::meanShiftSegmentation(InputArray src, OutputArray dst, int sp, int sr, int minsize, TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1))
+
+    :param src: Source image. Only  ``CV_8UC4`` images are supported for now.
+
+    :param dst: Segmented image with the same size and type as  ``src`` (host memory).
+
+    :param sp: Spatial window radius.
+
+    :param sr: Color window radius.
+
+    :param minsize: Minimum segment size. Smaller segments are merged.
+
+    :param criteria: Termination criteria. See :ocv:class:`TermCriteria`.
+
+
+
+gpu::TemplateMatching
+---------------------
+.. ocv:class:: gpu::TemplateMatching : public Algorithm
+
+Base class for Template Matching. ::
+
+    class CV_EXPORTS TemplateMatching : public Algorithm
+    {
+    public:
+        virtual void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null()) = 0;
+    };
+
+
+
+gpu::TemplateMatching::match
+----------------------------
+Computes a proximity map for a raster template and an image where the template is searched for.
+
+.. ocv:function:: void gpu::TemplateMatching::match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null())
+
+    :param image: Source image.
+
+    :param templ: Template image with the size and type the same as  ``image`` .
+
+    :param result: Map containing comparison results ( ``CV_32FC1`` ). If  ``image`` is  *W x H*  and ``templ`` is  *w x h*, then  ``result`` must be *W-w+1 x H-h+1*.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::createTemplateMatching
+---------------------------
+Creates implementation for :ocv:class:`gpu::TemplateMatching` .
+
+.. ocv:function:: Ptr<TemplateMatching> gpu::createTemplateMatching(int srcType, int method, Size user_block_size = Size())
+
+    :param srcType: Input source type. ``CV_32F`` and  ``CV_8U`` depth images (1..4 channels) are supported for now.
+
+    :param method: Specifies the way to compare the template with the image.
+
+    :param user_block_size: You can use field `user_block_size` to set specific block size. If you leave its default value `Size(0,0)` then automatic estimation of block size will be used (which is optimized for speed). By varying `user_block_size` you can reduce memory requirements at the cost of speed.
+
+The following methods are supported for the ``CV_8U`` depth images for now:
+
+    * ``CV_TM_SQDIFF``
+    * ``CV_TM_SQDIFF_NORMED``
+    * ``CV_TM_CCORR``
+    * ``CV_TM_CCORR_NORMED``
+    * ``CV_TM_CCOEFF``
+    * ``CV_TM_CCOEFF_NORMED``
+
+The following methods are supported for the ``CV_32F`` images for now:
+
+    * ``CV_TM_SQDIFF``
+    * ``CV_TM_CCORR``
+
+.. seealso:: :ocv:func:`matchTemplate`
+
+
+
+gpu::bilateralFilter
+--------------------
+Performs bilateral filtering of passed image
+
+.. ocv:function:: void gpu::bilateralFilter(InputArray src, OutputArray dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode=BORDER_DEFAULT, Stream& stream=Stream::Null())
+
+    :param src: Source image. Supports only (channles != 2 && depth() != CV_8S && depth() != CV_32S && depth() != CV_64F).
+
+    :param dst: Destination imagwe.
+
+    :param kernel_size: Kernel window size.
+
+    :param sigma_color: Filter sigma in the color space.
+
+    :param sigma_spatial:  Filter sigma in the coordinate space.
+
+    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`bilateralFilter`
+
+
+
+gpu::blendLinear
+-------------------
+Performs linear blending of two images.
+
+.. ocv:function:: void gpu::blendLinear(InputArray img1, InputArray img2, InputArray weights1, InputArray weights2, OutputArray result, Stream& stream = Stream::Null())
+
+    :param img1: First image. Supports only ``CV_8U`` and ``CV_32F`` depth.
+
+    :param img2: Second image. Must have the same size and the same type as ``img1`` .
+
+    :param weights1: Weights for first image. Must have tha same size as ``img1`` . Supports only ``CV_32F`` type.
+
+    :param weights2: Weights for second image. Must have tha same size as ``img2`` . Supports only ``CV_32F`` type.
+
+    :param result: Destination image.
+
+    :param stream: Stream for the asynchronous version.
diff --git a/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp b/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp
new file mode 100644
index 0000000000..f0a0f1260a
--- /dev/null
+++ b/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp
@@ -0,0 +1,362 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUIMGPROC_HPP__
+#define __OPENCV_GPUIMGPROC_HPP__
+
+#ifndef __cplusplus
+#  error gpuimgproc.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/imgproc.hpp"
+
+namespace cv { namespace gpu {
+
+/////////////////////////// Color Processing ///////////////////////////
+
+//! converts image from one color space to another
+CV_EXPORTS void cvtColor(InputArray src, OutputArray dst, int code, int dcn = 0, Stream& stream = Stream::Null());
+
+enum
+{
+    // Bayer Demosaicing (Malvar, He, and Cutler)
+    COLOR_BayerBG2BGR_MHT = 256,
+    COLOR_BayerGB2BGR_MHT = 257,
+    COLOR_BayerRG2BGR_MHT = 258,
+    COLOR_BayerGR2BGR_MHT = 259,
+
+    COLOR_BayerBG2RGB_MHT = COLOR_BayerRG2BGR_MHT,
+    COLOR_BayerGB2RGB_MHT = COLOR_BayerGR2BGR_MHT,
+    COLOR_BayerRG2RGB_MHT = COLOR_BayerBG2BGR_MHT,
+    COLOR_BayerGR2RGB_MHT = COLOR_BayerGB2BGR_MHT,
+
+    COLOR_BayerBG2GRAY_MHT = 260,
+    COLOR_BayerGB2GRAY_MHT = 261,
+    COLOR_BayerRG2GRAY_MHT = 262,
+    COLOR_BayerGR2GRAY_MHT = 263
+};
+CV_EXPORTS void demosaicing(InputArray src, OutputArray dst, int code, int dcn = -1, Stream& stream = Stream::Null());
+
+//! swap channels
+//! dstOrder - Integer array describing how channel values are permutated. The n-th entry
+//!            of the array contains the number of the channel that is stored in the n-th channel of
+//!            the output image. E.g. Given an RGBA image, aDstOrder = [3,2,1,0] converts this to ABGR
+//!            channel order.
+CV_EXPORTS void swapChannels(InputOutputArray image, const int dstOrder[4], Stream& stream = Stream::Null());
+
+//! Routines for correcting image color gamma
+CV_EXPORTS void gammaCorrection(InputArray src, OutputArray dst, bool forward = true, Stream& stream = Stream::Null());
+
+enum { ALPHA_OVER, ALPHA_IN, ALPHA_OUT, ALPHA_ATOP, ALPHA_XOR, ALPHA_PLUS, ALPHA_OVER_PREMUL, ALPHA_IN_PREMUL, ALPHA_OUT_PREMUL,
+       ALPHA_ATOP_PREMUL, ALPHA_XOR_PREMUL, ALPHA_PLUS_PREMUL, ALPHA_PREMUL};
+
+//! Composite two images using alpha opacity values contained in each image
+//! Supports CV_8UC4, CV_16UC4, CV_32SC4 and CV_32FC4 types
+CV_EXPORTS void alphaComp(InputArray img1, InputArray img2, OutputArray dst, int alpha_op, Stream& stream = Stream::Null());
+
+////////////////////////////// Histogram ///////////////////////////////
+
+//! Calculates histogram for 8u one channel image
+//! Output hist will have one row, 256 cols and CV32SC1 type.
+CV_EXPORTS void calcHist(InputArray src, OutputArray hist, Stream& stream = Stream::Null());
+
+//! normalizes the grayscale image brightness and contrast by normalizing its histogram
+CV_EXPORTS void equalizeHist(InputArray src, OutputArray dst, InputOutputArray buf, Stream& stream = Stream::Null());
+
+static inline void equalizeHist(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+{
+    GpuMat buf;
+    gpu::equalizeHist(src, dst, buf, stream);
+}
+
+class CV_EXPORTS CLAHE : public cv::CLAHE
+{
+public:
+    using cv::CLAHE::apply;
+    virtual void apply(InputArray src, OutputArray dst, Stream& stream) = 0;
+};
+CV_EXPORTS Ptr<gpu::CLAHE> createCLAHE(double clipLimit = 40.0, Size tileGridSize = Size(8, 8));
+
+//! Compute levels with even distribution. levels will have 1 row and nLevels cols and CV_32SC1 type.
+CV_EXPORTS void evenLevels(OutputArray levels, int nLevels, int lowerLevel, int upperLevel);
+
+//! Calculates histogram with evenly distributed bins for signle channel source.
+//! Supports CV_8UC1, CV_16UC1 and CV_16SC1 source types.
+//! Output hist will have one row and histSize cols and CV_32SC1 type.
+CV_EXPORTS void histEven(InputArray src, OutputArray hist, InputOutputArray buf, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());
+
+static inline void histEven(InputArray src, OutputArray hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null())
+{
+    GpuMat buf;
+    gpu::histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
+}
+
+//! Calculates histogram with evenly distributed bins for four-channel source.
+//! All channels of source are processed separately.
+//! Supports CV_8UC4, CV_16UC4 and CV_16SC4 source types.
+//! Output hist[i] will have one row and histSize[i] cols and CV_32SC1 type.
+CV_EXPORTS void histEven(InputArray src, GpuMat hist[4], InputOutputArray buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());
+
+static inline void histEven(InputArray src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null())
+{
+    GpuMat buf;
+    gpu::histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
+}
+
+//! Calculates histogram with bins determined by levels array.
+//! levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
+//! Supports CV_8UC1, CV_16UC1, CV_16SC1 and CV_32FC1 source types.
+//! Output hist will have one row and (levels.cols-1) cols and CV_32SC1 type.
+CV_EXPORTS void histRange(InputArray src, OutputArray hist, InputArray levels, InputOutputArray buf, Stream& stream = Stream::Null());
+
+static inline void histRange(InputArray src, OutputArray hist, InputArray levels, Stream& stream = Stream::Null())
+{
+    GpuMat buf;
+    gpu::histRange(src, hist, levels, buf, stream);
+}
+
+//! Calculates histogram with bins determined by levels array.
+//! All levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
+//! All channels of source are processed separately.
+//! Supports CV_8UC4, CV_16UC4, CV_16SC4 and CV_32FC4 source types.
+//! Output hist[i] will have one row and (levels[i].cols-1) cols and CV_32SC1 type.
+CV_EXPORTS void histRange(InputArray src, GpuMat hist[4], const GpuMat levels[4], InputOutputArray buf, Stream& stream = Stream::Null());
+
+static inline void histRange(InputArray src, GpuMat hist[4], const GpuMat levels[4], Stream& stream = Stream::Null())
+{
+    GpuMat buf;
+    gpu::histRange(src, hist, levels, buf, stream);
+}
+
+//////////////////////////////// Canny ////////////////////////////////
+
+class CV_EXPORTS CannyEdgeDetector : public Algorithm
+{
+public:
+    virtual void detect(InputArray image, OutputArray edges) = 0;
+    virtual void detect(InputArray dx, InputArray dy, OutputArray edges) = 0;
+
+    virtual void setLowThreshold(double low_thresh) = 0;
+    virtual double getLowThreshold() const = 0;
+
+    virtual void setHighThreshold(double high_thresh) = 0;
+    virtual double getHighThreshold() const = 0;
+
+    virtual void setAppertureSize(int apperture_size) = 0;
+    virtual int getAppertureSize() const = 0;
+
+    virtual void setL2Gradient(bool L2gradient) = 0;
+    virtual bool getL2Gradient() const = 0;
+};
+
+CV_EXPORTS Ptr<CannyEdgeDetector> createCannyEdgeDetector(double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false);
+
+/////////////////////////// Hough Transform ////////////////////////////
+
+//////////////////////////////////////
+// HoughLines
+
+class CV_EXPORTS HoughLinesDetector : public Algorithm
+{
+public:
+    virtual void detect(InputArray src, OutputArray lines) = 0;
+    virtual void downloadResults(InputArray d_lines, OutputArray h_lines, OutputArray h_votes = noArray()) = 0;
+
+    virtual void setRho(float rho) = 0;
+    virtual float getRho() const = 0;
+
+    virtual void setTheta(float theta) = 0;
+    virtual float getTheta() const = 0;
+
+    virtual void setThreshold(int threshold) = 0;
+    virtual int getThreshold() const = 0;
+
+    virtual void setDoSort(bool doSort) = 0;
+    virtual bool getDoSort() const = 0;
+
+    virtual void setMaxLines(int maxLines) = 0;
+    virtual int getMaxLines() const = 0;
+};
+
+CV_EXPORTS Ptr<HoughLinesDetector> createHoughLinesDetector(float rho, float theta, int threshold, bool doSort = false, int maxLines = 4096);
+
+
+//////////////////////////////////////
+// HoughLinesP
+
+//! finds line segments in the black-n-white image using probabalistic Hough transform
+class CV_EXPORTS HoughSegmentDetector : public Algorithm
+{
+public:
+    virtual void detect(InputArray src, OutputArray lines) = 0;
+
+    virtual void setRho(float rho) = 0;
+    virtual float getRho() const = 0;
+
+    virtual void setTheta(float theta) = 0;
+    virtual float getTheta() const = 0;
+
+    virtual void setMinLineLength(int minLineLength) = 0;
+    virtual int getMinLineLength() const = 0;
+
+    virtual void setMaxLineGap(int maxLineGap) = 0;
+    virtual int getMaxLineGap() const = 0;
+
+    virtual void setMaxLines(int maxLines) = 0;
+    virtual int getMaxLines() const = 0;
+};
+
+CV_EXPORTS Ptr<HoughSegmentDetector> createHoughSegmentDetector(float rho, float theta, int minLineLength, int maxLineGap, int maxLines = 4096);
+
+//////////////////////////////////////
+// HoughCircles
+
+class CV_EXPORTS HoughCirclesDetector : public Algorithm
+{
+public:
+    virtual void detect(InputArray src, OutputArray circles) = 0;
+
+    virtual void setDp(float dp) = 0;
+    virtual float getDp() const = 0;
+
+    virtual void setMinDist(float minDist) = 0;
+    virtual float getMinDist() const = 0;
+
+    virtual void setCannyThreshold(int cannyThreshold) = 0;
+    virtual int getCannyThreshold() const = 0;
+
+    virtual void setVotesThreshold(int votesThreshold) = 0;
+    virtual int getVotesThreshold() const = 0;
+
+    virtual void setMinRadius(int minRadius) = 0;
+    virtual int getMinRadius() const = 0;
+
+    virtual void setMaxRadius(int maxRadius) = 0;
+    virtual int getMaxRadius() const = 0;
+
+    virtual void setMaxCircles(int maxCircles) = 0;
+    virtual int getMaxCircles() const = 0;
+};
+
+CV_EXPORTS Ptr<HoughCirclesDetector> createHoughCirclesDetector(float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles = 4096);
+
+//////////////////////////////////////
+// GeneralizedHough
+
+//! Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
+//! Detects position only without traslation and rotation
+CV_EXPORTS Ptr<GeneralizedHoughBallard> createGeneralizedHoughBallard();
+
+//! Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
+//! Detects position, traslation and rotation
+CV_EXPORTS Ptr<GeneralizedHoughGuil> createGeneralizedHoughGuil();
+
+////////////////////////// Corners Detection ///////////////////////////
+
+class CV_EXPORTS CornernessCriteria : public Algorithm
+{
+public:
+    virtual void compute(InputArray src, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+};
+
+//! computes Harris cornerness criteria at each image pixel
+CV_EXPORTS Ptr<CornernessCriteria> createHarrisCorner(int srcType, int blockSize, int ksize, double k, int borderType = BORDER_REFLECT101);
+
+//! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria
+CV_EXPORTS Ptr<CornernessCriteria> createMinEigenValCorner(int srcType, int blockSize, int ksize, int borderType = BORDER_REFLECT101);
+
+////////////////////////// Corners Detection ///////////////////////////
+
+class CV_EXPORTS CornersDetector : public Algorithm
+{
+public:
+    //! return 1 rows matrix with CV_32FC2 type
+    virtual void detect(InputArray image, OutputArray corners, InputArray mask = noArray()) = 0;
+};
+
+CV_EXPORTS Ptr<CornersDetector> createGoodFeaturesToTrackDetector(int srcType, int maxCorners = 1000, double qualityLevel = 0.01, double minDistance = 0.0,
+                                                                  int blockSize = 3, bool useHarrisDetector = false, double harrisK = 0.04);
+
+///////////////////////////// Mean Shift //////////////////////////////
+
+//! Does mean shift filtering on GPU.
+CV_EXPORTS void meanShiftFiltering(InputArray src, OutputArray dst, int sp, int sr,
+                                   TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
+                                   Stream& stream = Stream::Null());
+
+//! Does mean shift procedure on GPU.
+CV_EXPORTS void meanShiftProc(InputArray src, OutputArray dstr, OutputArray dstsp, int sp, int sr,
+                              TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
+                              Stream& stream = Stream::Null());
+
+//! Does mean shift segmentation with elimination of small regions.
+CV_EXPORTS void meanShiftSegmentation(InputArray src, OutputArray dst, int sp, int sr, int minsize,
+                                      TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
+
+/////////////////////////// Match Template ////////////////////////////
+
+//! computes the proximity map for the raster template and the image where the template is searched for
+class CV_EXPORTS TemplateMatching : public Algorithm
+{
+public:
+    virtual void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null()) = 0;
+};
+
+CV_EXPORTS Ptr<TemplateMatching> createTemplateMatching(int srcType, int method, Size user_block_size = Size());
+
+////////////////////////// Bilateral Filter ///////////////////////////
+
+//! Performa bilateral filtering of passsed image
+CV_EXPORTS void bilateralFilter(InputArray src, OutputArray dst, int kernel_size, float sigma_color, float sigma_spatial,
+                                int borderMode = BORDER_DEFAULT, Stream& stream = Stream::Null());
+
+///////////////////////////// Blending ////////////////////////////////
+
+//! performs linear blending of two images
+//! to avoid accuracy errors sum of weigths shouldn't be very close to zero
+CV_EXPORTS void blendLinear(InputArray img1, InputArray img2, InputArray weights1, InputArray weights2,
+                            OutputArray result, Stream& stream = Stream::Null());
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUIMGPROC_HPP__ */
diff --git a/modules/gpuimgproc/perf/perf_bilateral_filter.cpp b/modules/gpuimgproc/perf/perf_bilateral_filter.cpp
new file mode 100644
index 0000000000..1787fdc096
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_bilateral_filter.cpp
@@ -0,0 +1,93 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// BilateralFilter
+
+DEF_PARAM_TEST(Sz_Depth_Cn_KernelSz, cv::Size, MatDepth, MatCn, int);
+
+PERF_TEST_P(Sz_Depth_Cn_KernelSz, BilateralFilter,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_32F),
+                    GPU_CHANNELS_1_3,
+                    Values(3, 5, 9)))
+{
+    declare.time(60.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int kernel_size = GET_PARAM(3);
+
+    const float sigma_color = 7;
+    const float sigma_spatial = 5;
+    const int borderMode = cv::BORDER_REFLECT101;
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::bilateralFilter(d_src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::bilateralFilter(src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_blend.cpp b/modules/gpuimgproc/perf/perf_blend.cpp
new file mode 100644
index 0000000000..5d43817684
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_blend.cpp
@@ -0,0 +1,86 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// BlendLinear
+
+PERF_TEST_P(Sz_Depth_Cn, BlendLinear,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat img1(size, type);
+    cv::Mat img2(size, type);
+    declare.in(img1, img2, WARMUP_RNG);
+
+    const cv::Mat weights1(size, CV_32FC1, cv::Scalar::all(0.5));
+    const cv::Mat weights2(size, CV_32FC1, cv::Scalar::all(0.5));
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_img1(img1);
+        const cv::gpu::GpuMat d_img2(img2);
+        const cv::gpu::GpuMat d_weights1(weights1);
+        const cv::gpu::GpuMat d_weights2(weights2);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::blendLinear(d_img1, d_img2, d_weights1, d_weights2, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_canny.cpp b/modules/gpuimgproc/perf/perf_canny.cpp
new file mode 100644
index 0000000000..2bbf70a496
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_canny.cpp
@@ -0,0 +1,88 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// Canny
+
+DEF_PARAM_TEST(Image_AppertureSz_L2gradient, string, int, bool);
+
+PERF_TEST_P(Image_AppertureSz_L2gradient, Canny,
+            Combine(Values("perf/800x600.png", "perf/1280x1024.png", "perf/1680x1050.png"),
+                    Values(3, 5),
+                    Bool()))
+{
+    const string fileName = GET_PARAM(0);
+    const int apperture_size = GET_PARAM(1);
+    const bool useL2gradient = GET_PARAM(2);
+
+    const cv::Mat image = readImage(fileName, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    const double low_thresh = 50.0;
+    const double high_thresh = 100.0;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_image(image);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::CannyEdgeDetector> canny = cv::gpu::createCannyEdgeDetector(low_thresh, high_thresh, apperture_size, useL2gradient);
+
+        TEST_CYCLE() canny->detect(d_image, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::Canny(image, dst, low_thresh, high_thresh, apperture_size, useL2gradient);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_color.cpp b/modules/gpuimgproc/perf/perf_color.cpp
new file mode 100644
index 0000000000..1df324816e
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_color.cpp
@@ -0,0 +1,252 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// CvtColor
+
+DEF_PARAM_TEST(Sz_Depth_Code, cv::Size, MatDepth, CvtColorInfo);
+
+PERF_TEST_P(Sz_Depth_Code, CvtColor,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_32F),
+                    Values(CvtColorInfo(4, 4, cv::COLOR_RGBA2BGRA),
+                           CvtColorInfo(4, 1, cv::COLOR_BGRA2GRAY),
+                           CvtColorInfo(1, 4, cv::COLOR_GRAY2BGRA),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2XYZ),
+                           CvtColorInfo(3, 3, cv::COLOR_XYZ2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2YCrCb),
+                           CvtColorInfo(3, 3, cv::COLOR_YCrCb2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2YUV),
+                           CvtColorInfo(3, 3, cv::COLOR_YUV2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2HSV),
+                           CvtColorInfo(3, 3, cv::COLOR_HSV2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2HLS),
+                           CvtColorInfo(3, 3, cv::COLOR_HLS2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2Lab),
+                           CvtColorInfo(3, 3, cv::COLOR_LBGR2Lab),
+                           CvtColorInfo(3, 3, cv::COLOR_BGR2Luv),
+                           CvtColorInfo(3, 3, cv::COLOR_LBGR2Luv),
+                           CvtColorInfo(3, 3, cv::COLOR_Lab2BGR),
+                           CvtColorInfo(3, 3, cv::COLOR_Lab2LBGR),
+                           CvtColorInfo(3, 3, cv::COLOR_Luv2RGB),
+                           CvtColorInfo(3, 3, cv::COLOR_Luv2LRGB))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const CvtColorInfo info = GET_PARAM(2);
+
+    cv::Mat src(size, CV_MAKETYPE(depth, info.scn));
+    cv::randu(src, 0, depth == CV_8U ? 255.0 : 1.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::cvtColor(d_src, dst, info.code, info.dcn);
+
+        GPU_SANITY_CHECK(dst, 1e-4);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::cvtColor(src, dst, info.code, info.dcn);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+PERF_TEST_P(Sz_Depth_Code, CvtColorBayer,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U),
+                    Values(CvtColorInfo(1, 3, cv::COLOR_BayerBG2BGR),
+                           CvtColorInfo(1, 3, cv::COLOR_BayerGB2BGR),
+                           CvtColorInfo(1, 3, cv::COLOR_BayerRG2BGR),
+                           CvtColorInfo(1, 3, cv::COLOR_BayerGR2BGR),
+
+                           CvtColorInfo(1, 1, cv::COLOR_BayerBG2GRAY),
+                           CvtColorInfo(1, 1, cv::COLOR_BayerGB2GRAY),
+                           CvtColorInfo(1, 1, cv::COLOR_BayerRG2GRAY),
+                           CvtColorInfo(1, 1, cv::COLOR_BayerGR2GRAY))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const CvtColorInfo info = GET_PARAM(2);
+
+    cv::Mat src(size, CV_MAKETYPE(depth, info.scn));
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::cvtColor(d_src, dst, info.code, info.dcn);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::cvtColor(src, dst, info.code, info.dcn);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Demosaicing
+
+CV_ENUM(DemosaicingCode,
+        cv::COLOR_BayerBG2BGR, cv::COLOR_BayerGB2BGR, cv::COLOR_BayerRG2BGR, cv::COLOR_BayerGR2BGR,
+        cv::COLOR_BayerBG2GRAY, cv::COLOR_BayerGB2GRAY, cv::COLOR_BayerRG2GRAY, cv::COLOR_BayerGR2GRAY,
+        cv::gpu::COLOR_BayerBG2BGR_MHT, cv::gpu::COLOR_BayerGB2BGR_MHT, cv::gpu::COLOR_BayerRG2BGR_MHT, cv::gpu::COLOR_BayerGR2BGR_MHT,
+        cv::gpu::COLOR_BayerBG2GRAY_MHT, cv::gpu::COLOR_BayerGB2GRAY_MHT, cv::gpu::COLOR_BayerRG2GRAY_MHT, cv::gpu::COLOR_BayerGR2GRAY_MHT)
+
+DEF_PARAM_TEST(Sz_Code, cv::Size, DemosaicingCode);
+
+PERF_TEST_P(Sz_Code, Demosaicing,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    DemosaicingCode::all()))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int code = GET_PARAM(1);
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::demosaicing(d_src, dst, code);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        if (code >= cv::COLOR_COLORCVT_MAX)
+        {
+            FAIL_NO_CPU();
+        }
+        else
+        {
+            cv::Mat dst;
+
+            TEST_CYCLE() cv::cvtColor(src, dst, code);
+
+            CPU_SANITY_CHECK(dst);
+        }
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// SwapChannels
+
+PERF_TEST_P(Sz, SwapChannels,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC4);
+    declare.in(src, WARMUP_RNG);
+
+    const int dstOrder[] = {2, 1, 0, 3};
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat dst(src);
+
+        TEST_CYCLE() cv::gpu::swapChannels(dst, dstOrder);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// AlphaComp
+
+CV_ENUM(AlphaOp, cv::gpu::ALPHA_OVER, cv::gpu::ALPHA_IN, cv::gpu::ALPHA_OUT, cv::gpu::ALPHA_ATOP, cv::gpu::ALPHA_XOR, cv::gpu::ALPHA_PLUS, cv::gpu::ALPHA_OVER_PREMUL, cv::gpu::ALPHA_IN_PREMUL, cv::gpu::ALPHA_OUT_PREMUL, cv::gpu::ALPHA_ATOP_PREMUL, cv::gpu::ALPHA_XOR_PREMUL, cv::gpu::ALPHA_PLUS_PREMUL, cv::gpu::ALPHA_PREMUL)
+
+DEF_PARAM_TEST(Sz_Type_Op, cv::Size, MatType, AlphaOp);
+
+PERF_TEST_P(Sz_Type_Op, AlphaComp,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8UC4, CV_16UC4, CV_32SC4, CV_32FC4),
+                    AlphaOp::all()))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int alpha_op = GET_PARAM(2);
+
+    cv::Mat img1(size, type);
+    cv::Mat img2(size, type);
+    declare.in(img1, img2, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_img1(img1);
+        const cv::gpu::GpuMat d_img2(img2);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::alphaComp(d_img1, d_img2, dst, alpha_op);
+
+        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_corners.cpp b/modules/gpuimgproc/perf/perf_corners.cpp
new file mode 100644
index 0000000000..a0c1f8d30f
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_corners.cpp
@@ -0,0 +1,135 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// CornerHarris
+
+DEF_PARAM_TEST(Image_Type_Border_BlockSz_ApertureSz, string, MatType, BorderMode, int, int);
+
+PERF_TEST_P(Image_Type_Border_BlockSz_ApertureSz, CornerHarris,
+            Combine(Values<string>("gpu/stereobm/aloe-L.png"),
+                    Values(CV_8UC1, CV_32FC1),
+                    Values(BorderMode(cv::BORDER_REFLECT101), BorderMode(cv::BORDER_REPLICATE), BorderMode(cv::BORDER_REFLECT)),
+                    Values(3, 5, 7),
+                    Values(0, 3, 5, 7)))
+{
+    const string fileName = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int borderMode = GET_PARAM(2);
+    const int blockSize = GET_PARAM(3);
+    const int apertureSize = GET_PARAM(4);
+
+    cv::Mat img = readImage(fileName, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    img.convertTo(img, type, type == CV_32F ? 1.0 / 255.0 : 1.0);
+
+    const double k = 0.5;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_img(img);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::CornernessCriteria> harris = cv::gpu::createHarrisCorner(img.type(), blockSize, apertureSize, k, borderMode);
+
+        TEST_CYCLE() harris->compute(d_img, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-4);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::cornerHarris(img, dst, blockSize, apertureSize, k, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// CornerMinEigenVal
+
+PERF_TEST_P(Image_Type_Border_BlockSz_ApertureSz, CornerMinEigenVal,
+            Combine(Values<string>("gpu/stereobm/aloe-L.png"),
+                    Values(CV_8UC1, CV_32FC1),
+                    Values(BorderMode(cv::BORDER_REFLECT101), BorderMode(cv::BORDER_REPLICATE), BorderMode(cv::BORDER_REFLECT)),
+                    Values(3, 5, 7),
+                    Values(0, 3, 5, 7)))
+{
+    const string fileName = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+    const int borderMode = GET_PARAM(2);
+    const int blockSize = GET_PARAM(3);
+    const int apertureSize = GET_PARAM(4);
+
+    cv::Mat img = readImage(fileName, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    img.convertTo(img, type, type == CV_32F ? 1.0 / 255.0 : 1.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_img(img);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::CornernessCriteria> minEigenVal = cv::gpu::createMinEigenValCorner(img.type(), blockSize, apertureSize, borderMode);
+
+        TEST_CYCLE() minEigenVal->compute(d_img, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-4);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::cornerMinEigenVal(img, dst, blockSize, apertureSize, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_gftt.cpp b/modules/gpuimgproc/perf/perf_gftt.cpp
new file mode 100644
index 0000000000..ed8d6ac16d
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_gftt.cpp
@@ -0,0 +1,86 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////
+// GoodFeaturesToTrack
+
+DEF_PARAM_TEST(Image_MinDistance, string, double);
+
+PERF_TEST_P(Image_MinDistance, GoodFeaturesToTrack,
+            Combine(Values<string>("gpu/perf/aloe.png"),
+                    Values(0.0, 3.0)))
+{
+    const string fileName = GET_PARAM(0);
+    const double minDistance = GET_PARAM(1);
+
+    const cv::Mat image = readImage(fileName, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    const int maxCorners = 8000;
+    const double qualityLevel = 0.01;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::CornersDetector> d_detector = cv::gpu::createGoodFeaturesToTrackDetector(image.type(), maxCorners, qualityLevel, minDistance);
+
+        const cv::gpu::GpuMat d_image(image);
+        cv::gpu::GpuMat pts;
+
+        TEST_CYCLE() d_detector->detect(d_image, pts);
+
+        GPU_SANITY_CHECK(pts);
+    }
+    else
+    {
+        cv::Mat pts;
+
+        TEST_CYCLE() cv::goodFeaturesToTrack(image, pts, maxCorners, qualityLevel, minDistance);
+
+        CPU_SANITY_CHECK(pts);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_histogram.cpp b/modules/gpuimgproc/perf/perf_histogram.cpp
new file mode 100644
index 0000000000..d8def54ff1
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_histogram.cpp
@@ -0,0 +1,220 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// HistEvenC1
+
+PERF_TEST_P(Sz_Depth, HistEvenC1,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_16S)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, depth);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+        cv::gpu::GpuMat d_buf;
+
+        TEST_CYCLE() cv::gpu::histEven(d_src, dst, d_buf, 30, 0, 180);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        const int hbins = 30;
+        const float hranges[] = {0.0f, 180.0f};
+        const int histSize[] = {hbins};
+        const float* ranges[] = {hranges};
+        const int channels[] = {0};
+
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::calcHist(&src, 1, channels, cv::Mat(), dst, 1, histSize, ranges);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// HistEvenC4
+
+PERF_TEST_P(Sz_Depth, HistEvenC4,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_16S)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, CV_MAKE_TYPE(depth, 4));
+    declare.in(src, WARMUP_RNG);
+
+    int histSize[] = {30, 30, 30, 30};
+    int lowerLevel[] = {0, 0, 0, 0};
+    int upperLevel[] = {180, 180, 180, 180};
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_hist[4];
+        cv::gpu::GpuMat d_buf;
+
+        TEST_CYCLE() cv::gpu::histEven(d_src, d_hist, d_buf, histSize, lowerLevel, upperLevel);
+
+        cv::Mat cpu_hist0, cpu_hist1, cpu_hist2, cpu_hist3;
+        d_hist[0].download(cpu_hist0);
+        d_hist[1].download(cpu_hist1);
+        d_hist[2].download(cpu_hist2);
+        d_hist[3].download(cpu_hist3);
+        SANITY_CHECK(cpu_hist0);
+        SANITY_CHECK(cpu_hist1);
+        SANITY_CHECK(cpu_hist2);
+        SANITY_CHECK(cpu_hist3);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// CalcHist
+
+PERF_TEST_P(Sz, CalcHist,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::calcHist(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// EqualizeHist
+
+PERF_TEST_P(Sz, EqualizeHist,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+        cv::gpu::GpuMat d_buf;
+
+        TEST_CYCLE() cv::gpu::equalizeHist(d_src, dst, d_buf);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::equalizeHist(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// CLAHE
+
+DEF_PARAM_TEST(Sz_ClipLimit, cv::Size, double);
+
+PERF_TEST_P(Sz_ClipLimit, CLAHE,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(0.0, 40.0)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const double clipLimit = GET_PARAM(1);
+
+    cv::Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::CLAHE> clahe = cv::gpu::createCLAHE(clipLimit);
+        cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() clahe->apply(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Ptr<cv::CLAHE> clahe = cv::createCLAHE(clipLimit);
+        cv::Mat dst;
+
+        TEST_CYCLE() clahe->apply(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_hough.cpp b/modules/gpuimgproc/perf/perf_hough.cpp
new file mode 100644
index 0000000000..cce8e7432e
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_hough.cpp
@@ -0,0 +1,349 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// HoughLines
+
+namespace
+{
+    struct Vec4iComparator
+    {
+        bool operator()(const cv::Vec4i& a, const cv::Vec4i b) const
+        {
+            if (a[0] != b[0]) return a[0] < b[0];
+            else if(a[1] != b[1]) return a[1] < b[1];
+            else if(a[2] != b[2]) return a[2] < b[2];
+            else return a[3] < b[3];
+        }
+    };
+    struct Vec3fComparator
+    {
+        bool operator()(const cv::Vec3f& a, const cv::Vec3f b) const
+        {
+            if(a[0] != b[0]) return a[0] < b[0];
+            else if(a[1] != b[1]) return a[1] < b[1];
+            else return a[2] < b[2];
+        }
+    };
+    struct Vec2fComparator
+    {
+        bool operator()(const cv::Vec2f& a, const cv::Vec2f b) const
+        {
+            if(a[0] != b[0]) return a[0] < b[0];
+            else return a[1] < b[1];
+        }
+    };
+}
+
+PERF_TEST_P(Sz, HoughLines,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    declare.time(30.0);
+
+    const cv::Size size = GetParam();
+
+    const float rho = 1.0f;
+    const float theta = static_cast<float>(CV_PI / 180.0);
+    const int threshold = 300;
+
+    cv::Mat src(size, CV_8UC1, cv::Scalar::all(0));
+    cv::line(src, cv::Point(0, 100), cv::Point(src.cols, 100), cv::Scalar::all(255), 1);
+    cv::line(src, cv::Point(0, 200), cv::Point(src.cols, 200), cv::Scalar::all(255), 1);
+    cv::line(src, cv::Point(0, 400), cv::Point(src.cols, 400), cv::Scalar::all(255), 1);
+    cv::line(src, cv::Point(100, 0), cv::Point(100, src.rows), cv::Scalar::all(255), 1);
+    cv::line(src, cv::Point(200, 0), cv::Point(200, src.rows), cv::Scalar::all(255), 1);
+    cv::line(src, cv::Point(400, 0), cv::Point(400, src.rows), cv::Scalar::all(255), 1);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_lines;
+
+        cv::Ptr<cv::gpu::HoughLinesDetector> hough = cv::gpu::createHoughLinesDetector(rho, theta, threshold);
+
+        TEST_CYCLE() hough->detect(d_src, d_lines);
+
+        cv::Mat gpu_lines(d_lines.row(0));
+        cv::Vec2f* begin = gpu_lines.ptr<cv::Vec2f>(0);
+        cv::Vec2f* end = begin + gpu_lines.cols;
+        std::sort(begin, end, Vec2fComparator());
+        SANITY_CHECK(gpu_lines);
+    }
+    else
+    {
+        std::vector<cv::Vec2f> cpu_lines;
+
+        TEST_CYCLE() cv::HoughLines(src, cpu_lines, rho, theta, threshold);
+
+        SANITY_CHECK(cpu_lines);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// HoughLinesP
+
+DEF_PARAM_TEST_1(Image, std::string);
+
+PERF_TEST_P(Image, HoughLinesP,
+            testing::Values("cv/shared/pic5.png", "stitching/a1.png"))
+{
+    declare.time(30.0);
+
+    const std::string fileName = getDataPath(GetParam());
+
+    const float rho = 1.0f;
+    const float theta = static_cast<float>(CV_PI / 180.0);
+    const int threshold = 100;
+    const int minLineLenght = 50;
+    const int maxLineGap = 5;
+
+    const cv::Mat image = cv::imread(fileName, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::Mat mask;
+    cv::Canny(image, mask, 50, 100);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_mask(mask);
+        cv::gpu::GpuMat d_lines;
+
+        cv::Ptr<cv::gpu::HoughSegmentDetector> hough = cv::gpu::createHoughSegmentDetector(rho, theta, minLineLenght, maxLineGap);
+
+        TEST_CYCLE() hough->detect(d_mask, d_lines);
+
+        cv::Mat gpu_lines(d_lines);
+        cv::Vec4i* begin = gpu_lines.ptr<cv::Vec4i>();
+        cv::Vec4i* end = begin + gpu_lines.cols;
+        std::sort(begin, end, Vec4iComparator());
+        SANITY_CHECK(gpu_lines);
+    }
+    else
+    {
+        std::vector<cv::Vec4i> cpu_lines;
+
+        TEST_CYCLE() cv::HoughLinesP(mask, cpu_lines, rho, theta, threshold, minLineLenght, maxLineGap);
+
+        SANITY_CHECK(cpu_lines);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// HoughCircles
+
+DEF_PARAM_TEST(Sz_Dp_MinDist, cv::Size, float, float);
+
+PERF_TEST_P(Sz_Dp_MinDist, HoughCircles,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(1.0f, 2.0f, 4.0f),
+                    Values(1.0f)))
+{
+    declare.time(30.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const float dp = GET_PARAM(1);
+    const float minDist = GET_PARAM(2);
+
+    const int minRadius = 10;
+    const int maxRadius = 30;
+    const int cannyThreshold = 100;
+    const int votesThreshold = 15;
+
+    cv::Mat src(size, CV_8UC1, cv::Scalar::all(0));
+    cv::circle(src, cv::Point(100, 100), 20, cv::Scalar::all(255), -1);
+    cv::circle(src, cv::Point(200, 200), 25, cv::Scalar::all(255), -1);
+    cv::circle(src, cv::Point(200, 100), 25, cv::Scalar::all(255), -1);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_circles;
+
+        cv::Ptr<cv::gpu::HoughCirclesDetector> houghCircles = cv::gpu::createHoughCirclesDetector(dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+
+        TEST_CYCLE() houghCircles->detect(d_src, d_circles);
+
+        cv::Mat gpu_circles(d_circles);
+        cv::Vec3f* begin = gpu_circles.ptr<cv::Vec3f>(0);
+        cv::Vec3f* end = begin + gpu_circles.cols;
+        std::sort(begin, end, Vec3fComparator());
+        SANITY_CHECK(gpu_circles);
+    }
+    else
+    {
+        std::vector<cv::Vec3f> cpu_circles;
+
+        TEST_CYCLE() cv::HoughCircles(src, cpu_circles, cv::HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+
+        SANITY_CHECK(cpu_circles);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// GeneralizedHough
+
+PERF_TEST_P(Sz, GeneralizedHoughBallard, GPU_TYPICAL_MAT_SIZES)
+{
+    declare.time(10);
+
+    const cv::Size imageSize = GetParam();
+
+    const cv::Mat templ = readImage("cv/shared/templ.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(templ.empty());
+
+    cv::Mat image(imageSize, CV_8UC1, cv::Scalar::all(0));
+    templ.copyTo(image(cv::Rect(50, 50, templ.cols, templ.rows)));
+
+    cv::Mat edges;
+    cv::Canny(image, edges, 50, 100);
+
+    cv::Mat dx, dy;
+    cv::Sobel(image, dx, CV_32F, 1, 0);
+    cv::Sobel(image, dy, CV_32F, 0, 1);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::GeneralizedHoughBallard> alg = cv::gpu::createGeneralizedHoughBallard();
+
+        const cv::gpu::GpuMat d_edges(edges);
+        const cv::gpu::GpuMat d_dx(dx);
+        const cv::gpu::GpuMat d_dy(dy);
+        cv::gpu::GpuMat positions;
+
+        alg->setTemplate(cv::gpu::GpuMat(templ));
+
+        TEST_CYCLE() alg->detect(d_edges, d_dx, d_dy, positions);
+
+        GPU_SANITY_CHECK(positions);
+    }
+    else
+    {
+        cv::Ptr<cv::GeneralizedHoughBallard> alg = cv::createGeneralizedHoughBallard();
+
+        cv::Mat positions;
+
+        alg->setTemplate(templ);
+
+        TEST_CYCLE() alg->detect(edges, dx, dy, positions);
+
+        CPU_SANITY_CHECK(positions);
+    }
+}
+
+PERF_TEST_P(Sz, GeneralizedHoughGuil, GPU_TYPICAL_MAT_SIZES)
+{
+    declare.time(10);
+
+    const cv::Size imageSize = GetParam();
+
+    const cv::Mat templ = readImage("cv/shared/templ.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(templ.empty());
+
+    cv::Mat image(imageSize, CV_8UC1, cv::Scalar::all(0));
+    templ.copyTo(image(cv::Rect(50, 50, templ.cols, templ.rows)));
+
+    cv::RNG rng(123456789);
+    const int objCount = rng.uniform(5, 15);
+    for (int i = 0; i < objCount; ++i)
+    {
+        double scale = rng.uniform(0.7, 1.3);
+        bool rotate = 1 == rng.uniform(0, 2);
+
+        cv::Mat obj;
+        cv::resize(templ, obj, cv::Size(), scale, scale);
+        if (rotate)
+            obj = obj.t();
+
+        cv::Point pos;
+
+        pos.x = rng.uniform(0, image.cols - obj.cols);
+        pos.y = rng.uniform(0, image.rows - obj.rows);
+
+        cv::Mat roi = image(cv::Rect(pos, obj.size()));
+        cv::add(roi, obj, roi);
+    }
+
+    cv::Mat edges;
+    cv::Canny(image, edges, 50, 100);
+
+    cv::Mat dx, dy;
+    cv::Sobel(image, dx, CV_32F, 1, 0);
+    cv::Sobel(image, dy, CV_32F, 0, 1);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::GeneralizedHoughGuil> alg = cv::gpu::createGeneralizedHoughGuil();
+        alg->setMaxAngle(90.0);
+        alg->setAngleStep(2.0);
+
+        const cv::gpu::GpuMat d_edges(edges);
+        const cv::gpu::GpuMat d_dx(dx);
+        const cv::gpu::GpuMat d_dy(dy);
+        cv::gpu::GpuMat positions;
+
+        alg->setTemplate(cv::gpu::GpuMat(templ));
+
+        TEST_CYCLE() alg->detect(d_edges, d_dx, d_dy, positions);
+
+        GPU_SANITY_CHECK(positions);
+    }
+    else
+    {
+        cv::Ptr<cv::GeneralizedHoughGuil> alg = cv::createGeneralizedHoughGuil();
+        alg->setMaxAngle(90.0);
+        alg->setAngleStep(2.0);
+
+        cv::Mat positions;
+
+        alg->setTemplate(templ);
+
+        TEST_CYCLE() alg->detect(edges, dx, dy, positions);
+
+        CPU_SANITY_CHECK(positions);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_main.cpp b/modules/gpuimgproc/perf/perf_main.cpp
new file mode 100644
index 0000000000..8eb27a27b9
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpuimgproc)
diff --git a/modules/gpuimgproc/perf/perf_match_template.cpp b/modules/gpuimgproc/perf/perf_match_template.cpp
new file mode 100644
index 0000000000..35f36596c6
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_match_template.cpp
@@ -0,0 +1,135 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate8U
+
+CV_ENUM(TemplateMethod, TM_SQDIFF, TM_SQDIFF_NORMED, TM_CCORR, TM_CCORR_NORMED, TM_CCOEFF, TM_CCOEFF_NORMED)
+
+DEF_PARAM_TEST(Sz_TemplateSz_Cn_Method, cv::Size, cv::Size, MatCn, TemplateMethod);
+
+PERF_TEST_P(Sz_TemplateSz_Cn_Method, MatchTemplate8U,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
+                    GPU_CHANNELS_1_3_4,
+                    TemplateMethod::all()))
+{
+    declare.time(300.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const cv::Size templ_size = GET_PARAM(1);
+    const int cn = GET_PARAM(2);
+    const int method = GET_PARAM(3);
+
+    cv::Mat image(size, CV_MAKE_TYPE(CV_8U, cn));
+    cv::Mat templ(templ_size, CV_MAKE_TYPE(CV_8U, cn));
+    declare.in(image, templ, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_image(image);
+        const cv::gpu::GpuMat d_templ(templ);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), method);
+
+        TEST_CYCLE() alg->match(d_image, d_templ, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-5, ERROR_RELATIVE);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::matchTemplate(image, templ, dst, method);
+
+        CPU_SANITY_CHECK(dst);
+    }
+};
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate32F
+
+PERF_TEST_P(Sz_TemplateSz_Cn_Method, MatchTemplate32F,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
+                    GPU_CHANNELS_1_3_4,
+                    Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))))
+{
+    declare.time(300.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const cv::Size templ_size = GET_PARAM(1);
+    const int cn = GET_PARAM(2);
+    int method = GET_PARAM(3);
+
+    cv::Mat image(size, CV_MAKE_TYPE(CV_32F, cn));
+    cv::Mat templ(templ_size, CV_MAKE_TYPE(CV_32F, cn));
+    declare.in(image, templ, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_image(image);
+        const cv::gpu::GpuMat d_templ(templ);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), method);
+
+        TEST_CYCLE() alg->match(d_image, d_templ, dst);
+
+        GPU_SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::matchTemplate(image, templ, dst, method);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_mean_shift.cpp b/modules/gpuimgproc/perf/perf_mean_shift.cpp
new file mode 100644
index 0000000000..0ac0b71c44
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_mean_shift.cpp
@@ -0,0 +1,152 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// MeanShiftFiltering
+
+DEF_PARAM_TEST_1(Image, string);
+
+PERF_TEST_P(Image, MeanShiftFiltering,
+            Values<string>("gpu/meanshift/cones.png"))
+{
+    declare.time(300.0);
+
+    const cv::Mat img = readImage(GetParam());
+    ASSERT_FALSE(img.empty());
+
+    cv::Mat rgba;
+    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
+
+    const int sp = 50;
+    const int sr = 50;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(rgba);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::meanShiftFiltering(d_src, dst, sp, sr);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::pyrMeanShiftFiltering(img, dst, sp, sr);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MeanShiftProc
+
+PERF_TEST_P(Image, MeanShiftProc,
+            Values<string>("gpu/meanshift/cones.png"))
+{
+    declare.time(300.0);
+
+    const cv::Mat img = readImage(GetParam());
+    ASSERT_FALSE(img.empty());
+
+    cv::Mat rgba;
+    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
+
+    const int sp = 50;
+    const int sr = 50;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(rgba);
+        cv::gpu::GpuMat dstr;
+        cv::gpu::GpuMat dstsp;
+
+        TEST_CYCLE() cv::gpu::meanShiftProc(d_src, dstr, dstsp, sp, sr);
+
+        GPU_SANITY_CHECK(dstr);
+        GPU_SANITY_CHECK(dstsp);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// MeanShiftSegmentation
+
+PERF_TEST_P(Image, MeanShiftSegmentation,
+            Values<string>("gpu/meanshift/cones.png"))
+{
+    declare.time(300.0);
+
+    const cv::Mat img = readImage(GetParam());
+    ASSERT_FALSE(img.empty());
+
+    cv::Mat rgba;
+    cv::cvtColor(img, rgba, cv::COLOR_BGR2BGRA);
+
+    const int sp = 10;
+    const int sr = 10;
+    const int minsize = 20;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(rgba);
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::gpu::meanShiftSegmentation(d_src, dst, sp, sr, minsize);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuimgproc/perf/perf_precomp.hpp b/modules/gpuimgproc/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..adaa4ebb7e
--- /dev/null
+++ b/modules/gpuimgproc/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpuimgproc.hpp"
+#include "opencv2/imgproc.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpuimgproc/src/bilateral_filter.cpp b/modules/gpuimgproc/src/bilateral_filter.cpp
new file mode 100644
index 0000000000..b9d0b811e4
--- /dev/null
+++ b/modules/gpuimgproc/src/bilateral_filter.cpp
@@ -0,0 +1,99 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::bilateralFilter(InputArray, OutputArray, int, float, float, int, Stream&) { throw_no_cuda(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        template<typename T>
+        void bilateral_filter_gpu(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::bilateralFilter(InputArray _src, OutputArray _dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode, Stream& stream)
+{
+    using cv::gpu::cudev::imgproc::bilateral_filter_gpu;
+
+    typedef void (*func_t)(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t s);
+
+    static const func_t funcs[6][4] =
+    {
+        {bilateral_filter_gpu<uchar>      , 0 /*bilateral_filter_gpu<uchar2>*/ , bilateral_filter_gpu<uchar3>      , bilateral_filter_gpu<uchar4>      },
+        {0 /*bilateral_filter_gpu<schar>*/, 0 /*bilateral_filter_gpu<schar2>*/ , 0 /*bilateral_filter_gpu<schar3>*/, 0 /*bilateral_filter_gpu<schar4>*/},
+        {bilateral_filter_gpu<ushort>     , 0 /*bilateral_filter_gpu<ushort2>*/, bilateral_filter_gpu<ushort3>     , bilateral_filter_gpu<ushort4>     },
+        {bilateral_filter_gpu<short>      , 0 /*bilateral_filter_gpu<short2>*/ , bilateral_filter_gpu<short3>      , bilateral_filter_gpu<short4>      },
+        {0 /*bilateral_filter_gpu<int>*/  , 0 /*bilateral_filter_gpu<int2>*/   , 0 /*bilateral_filter_gpu<int3>*/  , 0 /*bilateral_filter_gpu<int4>*/  },
+        {bilateral_filter_gpu<float>      , 0 /*bilateral_filter_gpu<float2>*/ , bilateral_filter_gpu<float3>      , bilateral_filter_gpu<float4>      }
+    };
+
+    sigma_color = (sigma_color <= 0 ) ? 1 : sigma_color;
+    sigma_spatial = (sigma_spatial <= 0 ) ? 1 : sigma_spatial;
+
+    int radius = (kernel_size <= 0) ? cvRound(sigma_spatial*1.5) : kernel_size/2;
+    kernel_size = std::max(radius, 1)*2 + 1;
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
+
+    const func_t func = funcs[src.depth()][src.channels() - 1];
+    CV_Assert( func != 0 );
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    func(src, dst, kernel_size, sigma_spatial, sigma_color, borderMode, StreamAccessor::getStream(stream));
+}
+
+#endif
diff --git a/modules/gpu/src/blend.cpp b/modules/gpuimgproc/src/blend.cpp
similarity index 72%
rename from modules/gpu/src/blend.cpp
rename to modules/gpuimgproc/src/blend.cpp
index df73afad69..71c72a7153 100644
--- a/modules/gpu/src/blend.cpp
+++ b/modules/gpuimgproc/src/blend.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,11 +47,14 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::gpu::blendLinear(const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::blendLinear(InputArray, InputArray, InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
 #else
 
-namespace cv { namespace gpu { namespace device
+////////////////////////////////////////////////////////////////////////
+// blendLinear
+
+namespace cv { namespace gpu { namespace cudev
 {
     namespace blend
     {
@@ -62,23 +65,30 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-using namespace ::cv::gpu::device::blend;
+using namespace ::cv::gpu::cudev::blend;
 
-void cv::gpu::blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat& weights1, const GpuMat& weights2,
-                          GpuMat& result, Stream& stream)
+void cv::gpu::blendLinear(InputArray _img1, InputArray _img2, InputArray _weights1, InputArray _weights2,
+                          OutputArray _result, Stream& stream)
 {
-    CV_Assert(img1.size() == img2.size());
-    CV_Assert(img1.type() == img2.type());
-    CV_Assert(weights1.size() == img1.size());
-    CV_Assert(weights2.size() == img2.size());
-    CV_Assert(weights1.type() == CV_32F);
-    CV_Assert(weights2.type() == CV_32F);
+    GpuMat img1 = _img1.getGpuMat();
+    GpuMat img2 = _img2.getGpuMat();
+
+    GpuMat weights1 = _weights1.getGpuMat();
+    GpuMat weights2 = _weights2.getGpuMat();
+
+    CV_Assert( img1.size() == img2.size() );
+    CV_Assert( img1.type() == img2.type() );
+    CV_Assert( weights1.size() == img1.size() );
+    CV_Assert( weights2.size() == img2.size() );
+    CV_Assert( weights1.type() == CV_32FC1 );
+    CV_Assert( weights2.type() == CV_32FC1 );
 
     const Size size = img1.size();
     const int depth = img1.depth();
     const int cn = img1.channels();
 
-    result.create(size, CV_MAKE_TYPE(depth, cn));
+    _result.create(size, CV_MAKE_TYPE(depth, cn));
+    GpuMat result = _result.getGpuMat();
 
     switch (depth)
     {
@@ -92,7 +102,7 @@ void cv::gpu::blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat&
         blendLinearCaller<float>(size.height, size.width, cn, img1, img2, weights1, weights2, result, StreamAccessor::getStream(stream));
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "bad image depth in linear blending function");
+        CV_Error(cv::Error::StsUnsupportedFormat, "bad image depth in linear blending function");
     }
 }
 
diff --git a/modules/gpuimgproc/src/canny.cpp b/modules/gpuimgproc/src/canny.cpp
new file mode 100644
index 0000000000..17f03950f8
--- /dev/null
+++ b/modules/gpuimgproc/src/canny.cpp
@@ -0,0 +1,234 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<CannyEdgeDetector> cv::gpu::createCannyEdgeDetector(double, double, int, bool) { throw_no_cuda(); return Ptr<CannyEdgeDetector>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace canny
+{
+    void calcMagnitude(PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, bool L2Grad);
+    void calcMagnitude(PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, bool L2Grad);
+
+    void calcMap(PtrStepSzi dx, PtrStepSzi dy, PtrStepSzf mag, PtrStepSzi map, float low_thresh, float high_thresh);
+
+    void edgesHysteresisLocal(PtrStepSzi map, ushort2* st1);
+
+    void edgesHysteresisGlobal(PtrStepSzi map, ushort2* st1, ushort2* st2);
+
+    void getEdges(PtrStepSzi map, PtrStepSzb dst);
+}
+
+namespace
+{
+    class CannyImpl : public CannyEdgeDetector
+    {
+    public:
+        CannyImpl(double low_thresh, double high_thresh, int apperture_size, bool L2gradient) :
+            low_thresh_(low_thresh), high_thresh_(high_thresh), apperture_size_(apperture_size), L2gradient_(L2gradient)
+        {
+            old_apperture_size_ = -1;
+        }
+
+        void detect(InputArray image, OutputArray edges);
+        void detect(InputArray dx, InputArray dy, OutputArray edges);
+
+        void setLowThreshold(double low_thresh) { low_thresh_ = low_thresh; }
+        double getLowThreshold() const { return low_thresh_; }
+
+        void setHighThreshold(double high_thresh) { high_thresh_ = high_thresh; }
+        double getHighThreshold() const { return high_thresh_; }
+
+        void setAppertureSize(int apperture_size) { apperture_size_ = apperture_size; }
+        int getAppertureSize() const { return apperture_size_; }
+
+        void setL2Gradient(bool L2gradient) { L2gradient_ = L2gradient; }
+        bool getL2Gradient() const { return L2gradient_; }
+
+        void write(FileStorage& fs) const
+        {
+            fs << "name" << "Canny_GPU"
+            << "low_thresh" << low_thresh_
+            << "high_thresh" << high_thresh_
+            << "apperture_size" << apperture_size_
+            << "L2gradient" << L2gradient_;
+        }
+
+        void read(const FileNode& fn)
+        {
+            CV_Assert( String(fn["name"]) == "Canny_GPU" );
+            low_thresh_ = (double)fn["low_thresh"];
+            high_thresh_ = (double)fn["high_thresh"];
+            apperture_size_ = (int)fn["apperture_size"];
+            L2gradient_ = (int)fn["L2gradient"] != 0;
+        }
+
+    private:
+        void createBuf(Size image_size);
+        void CannyCaller(GpuMat& edges);
+
+        double low_thresh_;
+        double high_thresh_;
+        int apperture_size_;
+        bool L2gradient_;
+
+        GpuMat dx_, dy_;
+        GpuMat mag_;
+        GpuMat map_;
+        GpuMat st1_, st2_;
+#ifdef HAVE_OPENCV_GPUFILTERS
+        Ptr<Filter> filterDX_, filterDY_;
+#endif
+        int old_apperture_size_;
+    };
+
+    void CannyImpl::detect(InputArray _image, OutputArray _edges)
+    {
+        GpuMat image = _image.getGpuMat();
+
+        CV_Assert( image.type() == CV_8UC1 );
+        CV_Assert( deviceSupports(SHARED_ATOMICS) );
+
+        if (low_thresh_ > high_thresh_)
+            std::swap(low_thresh_, high_thresh_);
+
+        createBuf(image.size());
+
+        _edges.create(image.size(), CV_8UC1);
+        GpuMat edges = _edges.getGpuMat();
+
+        if (apperture_size_ == 3)
+        {
+            Size wholeSize;
+            Point ofs;
+            image.locateROI(wholeSize, ofs);
+            GpuMat srcWhole(wholeSize, image.type(), image.datastart, image.step);
+
+            canny::calcMagnitude(srcWhole, ofs.x, ofs.y, dx_, dy_, mag_, L2gradient_);
+        }
+        else
+        {
+#ifndef HAVE_OPENCV_GPUFILTERS
+            throw_no_cuda();
+#else
+            filterDX_->apply(image, dx_);
+            filterDY_->apply(image, dy_);
+
+            canny::calcMagnitude(dx_, dy_, mag_, L2gradient_);
+#endif
+        }
+
+        CannyCaller(edges);
+    }
+
+    void CannyImpl::detect(InputArray _dx, InputArray _dy, OutputArray _edges)
+    {
+        GpuMat dx = _dx.getGpuMat();
+        GpuMat dy = _dy.getGpuMat();
+
+        CV_Assert( dx.type() == CV_32SC1 );
+        CV_Assert( dy.type() == dx.type() && dy.size() == dx.size() );
+        CV_Assert( deviceSupports(SHARED_ATOMICS) );
+
+        dx.copyTo(dx_);
+        dy.copyTo(dy_);
+
+        if (low_thresh_ > high_thresh_)
+            std::swap(low_thresh_, high_thresh_);
+
+        createBuf(dx.size());
+
+        _edges.create(dx.size(), CV_8UC1);
+        GpuMat edges = _edges.getGpuMat();
+
+        canny::calcMagnitude(dx_, dy_, mag_, L2gradient_);
+
+        CannyCaller(edges);
+    }
+
+    void CannyImpl::createBuf(Size image_size)
+    {
+        ensureSizeIsEnough(image_size, CV_32SC1, dx_);
+        ensureSizeIsEnough(image_size, CV_32SC1, dy_);
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        if (apperture_size_ != 3 && apperture_size_ != old_apperture_size_)
+        {
+            filterDX_ = gpu::createDerivFilter(CV_8UC1, CV_32S, 1, 0, apperture_size_, false, 1, BORDER_REPLICATE);
+            filterDY_ = gpu::createDerivFilter(CV_8UC1, CV_32S, 0, 1, apperture_size_, false, 1, BORDER_REPLICATE);
+            old_apperture_size_ = apperture_size_;
+        }
+#endif
+
+        ensureSizeIsEnough(image_size, CV_32FC1, mag_);
+        ensureSizeIsEnough(image_size, CV_32SC1, map_);
+
+        ensureSizeIsEnough(1, image_size.area(), CV_16UC2, st1_);
+        ensureSizeIsEnough(1, image_size.area(), CV_16UC2, st2_);
+    }
+
+    void CannyImpl::CannyCaller(GpuMat& edges)
+    {
+        map_.setTo(Scalar::all(0));
+        canny::calcMap(dx_, dy_, mag_, map_, static_cast<float>(low_thresh_), static_cast<float>(high_thresh_));
+
+        canny::edgesHysteresisLocal(map_, st1_.ptr<ushort2>());
+
+        canny::edgesHysteresisGlobal(map_, st1_.ptr<ushort2>(), st2_.ptr<ushort2>());
+
+        canny::getEdges(map_, edges);
+    }
+}
+
+Ptr<CannyEdgeDetector> cv::gpu::createCannyEdgeDetector(double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
+{
+    return makePtr<CannyImpl>(low_thresh, high_thresh, apperture_size, L2gradient);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/color.cpp b/modules/gpuimgproc/src/color.cpp
new file mode 100644
index 0000000000..3d714b6287
--- /dev/null
+++ b/modules/gpuimgproc/src/color.cpp
@@ -0,0 +1,2330 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::cvtColor(InputArray, OutputArray, int, int, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::demosaicing(InputArray, OutputArray, int, int, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::swapChannels(InputOutputArray, const int[], Stream&) { throw_no_cuda(); }
+
+void cv::gpu::gammaCorrection(InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::alphaComp(InputArray, InputArray, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+
+#else /* !defined (HAVE_CUDA) */
+
+#include "cvt_color_internal.h"
+
+namespace cv { namespace gpu {
+    namespace cudev
+    {
+        template <int cn>
+        void Bayer2BGR_8u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+        template <int cn>
+        void Bayer2BGR_16u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+
+        template <int cn>
+        void MHCdemosaic(PtrStepSzb src, int2 sourceOffset, PtrStepSzb dst, int2 firstRed, cudaStream_t stream);
+    }
+}}
+
+using namespace ::cv::gpu::cudev;
+
+namespace
+{
+    typedef void (*gpu_func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+
+    void bgr_to_rgb(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgr_to_rgb_8u, 0, bgr_to_rgb_16u, 0, 0, bgr_to_rgb_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 3));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_bgra(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgr_to_bgra_8u, 0, bgr_to_bgra_16u, 0, 0, bgr_to_bgra_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 4));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_rgba(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgr_to_rgba_8u, 0, bgr_to_rgba_16u, 0, 0, bgr_to_rgba_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 4));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_bgr(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgra_to_bgr_8u, 0, bgra_to_bgr_16u, 0, 0, bgra_to_bgr_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 3));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_rgb(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgra_to_rgb_8u, 0, bgra_to_rgb_16u, 0, 0, bgra_to_rgb_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 3));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_rgba(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgra_to_rgba_8u, 0, bgra_to_rgba_16u, 0, 0, bgra_to_rgba_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 4));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_bgr555(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr_to_bgr555(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_bgr565(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr_to_bgr565(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_bgr555(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::rgb_to_bgr555(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_bgr565(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::rgb_to_bgr565(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_bgr555(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgra_to_bgr555(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_bgr565(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgra_to_bgr565(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgba_to_bgr555(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::rgba_to_bgr555(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgba_to_bgr565(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::rgba_to_bgr565(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr555_to_rgb(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC3);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr555_to_rgb(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr565_to_rgb(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC3);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr565_to_rgb(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr555_to_bgr(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC3);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr555_to_bgr(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr565_to_bgr(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC3);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr565_to_bgr(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr555_to_rgba(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC4);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr555_to_rgba(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr565_to_rgba(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC4);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr565_to_rgba(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr555_to_bgra(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC4);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr555_to_bgra(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr565_to_bgra(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC4);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr565_to_bgra(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void gray_to_bgr(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {gray_to_bgr_8u, 0, gray_to_bgr_16u, 0, 0, gray_to_bgr_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 1 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 3));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void gray_to_bgra(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {gray_to_bgra_8u, 0, gray_to_bgra_16u, 0, 0, gray_to_bgra_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 1 );
+
+        _dst.create(src.size(), CV_MAKETYPE(src.depth(), 4));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void gray_to_bgr555(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 1 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::gray_to_bgr555(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void gray_to_bgr565(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 1 );
+
+        _dst.create(src.size(), CV_8UC2);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::gray_to_bgr565(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr555_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr555_to_gray(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr565_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U );
+        CV_Assert( src.channels() == 2 );
+
+        _dst.create(src.size(), CV_8UC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        cv::gpu::cudev::bgr565_to_gray(src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {rgb_to_gray_8u, 0, rgb_to_gray_16u, 0, 0, rgb_to_gray_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgr_to_gray_8u, 0, bgr_to_gray_16u, 0, 0, bgr_to_gray_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgba_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {rgba_to_gray_8u, 0, rgba_to_gray_16u, 0, 0, rgba_to_gray_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgra_to_gray(InputArray _src, OutputArray _dst, int, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[] = {bgra_to_gray_8u, 0, bgra_to_gray_16u, 0, 0, bgra_to_gray_32f};
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_yuv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_yuv_8u, 0, rgb_to_yuv_16u, 0, 0, rgb_to_yuv_32f},
+                {rgba_to_yuv_8u, 0, rgba_to_yuv_16u, 0, 0, rgba_to_yuv_32f}
+            },
+            {
+                {rgb_to_yuv4_8u, 0, rgb_to_yuv4_16u, 0, 0, rgb_to_yuv4_32f},
+                {rgba_to_yuv4_8u, 0, rgba_to_yuv4_16u, 0, 0, rgba_to_yuv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_yuv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_yuv_8u, 0, bgr_to_yuv_16u, 0, 0, bgr_to_yuv_32f},
+                {bgra_to_yuv_8u, 0, bgra_to_yuv_16u, 0, 0, bgra_to_yuv_32f}
+            },
+            {
+                {bgr_to_yuv4_8u, 0, bgr_to_yuv4_16u, 0, 0, bgr_to_yuv4_32f},
+                {bgra_to_yuv4_8u, 0, bgra_to_yuv4_16u, 0, 0, bgra_to_yuv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void yuv_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {yuv_to_rgb_8u, 0, yuv_to_rgb_16u, 0, 0, yuv_to_rgb_32f},
+                {yuv4_to_rgb_8u, 0, yuv4_to_rgb_16u, 0, 0, yuv4_to_rgb_32f}
+            },
+            {
+                {yuv_to_rgba_8u, 0, yuv_to_rgba_16u, 0, 0, yuv_to_rgba_32f},
+                {yuv4_to_rgba_8u, 0, yuv4_to_rgba_16u, 0, 0, yuv4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void yuv_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {yuv_to_bgr_8u, 0, yuv_to_bgr_16u, 0, 0, yuv_to_bgr_32f},
+                {yuv4_to_bgr_8u, 0, yuv4_to_bgr_16u, 0, 0, yuv4_to_bgr_32f}
+            },
+            {
+                {yuv_to_bgra_8u, 0, yuv_to_bgra_16u, 0, 0, yuv_to_bgra_32f},
+                {yuv4_to_bgra_8u, 0, yuv4_to_bgra_16u, 0, 0, yuv4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_YCrCb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_YCrCb_8u, 0, rgb_to_YCrCb_16u, 0, 0, rgb_to_YCrCb_32f},
+                {rgba_to_YCrCb_8u, 0, rgba_to_YCrCb_16u, 0, 0, rgba_to_YCrCb_32f}
+            },
+            {
+                {rgb_to_YCrCb4_8u, 0, rgb_to_YCrCb4_16u, 0, 0, rgb_to_YCrCb4_32f},
+                {rgba_to_YCrCb4_8u, 0, rgba_to_YCrCb4_16u, 0, 0, rgba_to_YCrCb4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_YCrCb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_YCrCb_8u, 0, bgr_to_YCrCb_16u, 0, 0, bgr_to_YCrCb_32f},
+                {bgra_to_YCrCb_8u, 0, bgra_to_YCrCb_16u, 0, 0, bgra_to_YCrCb_32f}
+            },
+            {
+                {bgr_to_YCrCb4_8u, 0, bgr_to_YCrCb4_16u, 0, 0, bgr_to_YCrCb4_32f},
+                {bgra_to_YCrCb4_8u, 0, bgra_to_YCrCb4_16u, 0, 0, bgra_to_YCrCb4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void YCrCb_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {YCrCb_to_rgb_8u, 0, YCrCb_to_rgb_16u, 0, 0, YCrCb_to_rgb_32f},
+                {YCrCb4_to_rgb_8u, 0, YCrCb4_to_rgb_16u, 0, 0, YCrCb4_to_rgb_32f}
+            },
+            {
+                {YCrCb_to_rgba_8u, 0, YCrCb_to_rgba_16u, 0, 0, YCrCb_to_rgba_32f},
+                {YCrCb4_to_rgba_8u, 0, YCrCb4_to_rgba_16u, 0, 0, YCrCb4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void YCrCb_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {YCrCb_to_bgr_8u, 0, YCrCb_to_bgr_16u, 0, 0, YCrCb_to_bgr_32f},
+                {YCrCb4_to_bgr_8u, 0, YCrCb4_to_bgr_16u, 0, 0, YCrCb4_to_bgr_32f}
+            },
+            {
+                {YCrCb_to_bgra_8u, 0, YCrCb_to_bgra_16u, 0, 0, YCrCb_to_bgra_32f},
+                {YCrCb4_to_bgra_8u, 0, YCrCb4_to_bgra_16u, 0, 0, YCrCb4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_xyz(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_xyz_8u, 0, rgb_to_xyz_16u, 0, 0, rgb_to_xyz_32f},
+                {rgba_to_xyz_8u, 0, rgba_to_xyz_16u, 0, 0, rgba_to_xyz_32f}
+            },
+            {
+                {rgb_to_xyz4_8u, 0, rgb_to_xyz4_16u, 0, 0, rgb_to_xyz4_32f},
+                {rgba_to_xyz4_8u, 0, rgba_to_xyz4_16u, 0, 0, rgba_to_xyz4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_xyz(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_xyz_8u, 0, bgr_to_xyz_16u, 0, 0, bgr_to_xyz_32f},
+                {bgra_to_xyz_8u, 0, bgra_to_xyz_16u, 0, 0, bgra_to_xyz_32f}
+            },
+            {
+                {bgr_to_xyz4_8u, 0, bgr_to_xyz4_16u, 0, 0, bgr_to_xyz4_32f},
+                {bgra_to_xyz4_8u, 0, bgra_to_xyz4_16u, 0, 0, bgra_to_xyz4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void xyz_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {xyz_to_rgb_8u, 0, xyz_to_rgb_16u, 0, 0, xyz_to_rgb_32f},
+                {xyz4_to_rgb_8u, 0, xyz4_to_rgb_16u, 0, 0, xyz4_to_rgb_32f}
+            },
+            {
+                {xyz_to_rgba_8u, 0, xyz_to_rgba_16u, 0, 0, xyz_to_rgba_32f},
+                {xyz4_to_rgba_8u, 0, xyz4_to_rgba_16u, 0, 0, xyz4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void xyz_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {xyz_to_bgr_8u, 0, xyz_to_bgr_16u, 0, 0, xyz_to_bgr_32f},
+                {xyz4_to_bgr_8u, 0, xyz4_to_bgr_16u, 0, 0, xyz4_to_bgr_32f}
+            },
+            {
+                {xyz_to_bgra_8u, 0, xyz_to_bgra_16u, 0, 0, xyz_to_bgra_32f},
+                {xyz4_to_bgra_8u, 0, xyz4_to_bgra_16u, 0, 0, xyz4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_hsv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_hsv_8u, 0, 0, 0, 0, rgb_to_hsv_32f},
+                {rgba_to_hsv_8u, 0, 0, 0, 0, rgba_to_hsv_32f},
+            },
+            {
+                {rgb_to_hsv4_8u, 0, 0, 0, 0, rgb_to_hsv4_32f},
+                {rgba_to_hsv4_8u, 0, 0, 0, 0, rgba_to_hsv4_32f},
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_hsv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_hsv_8u, 0, 0, 0, 0, bgr_to_hsv_32f},
+                {bgra_to_hsv_8u, 0, 0, 0, 0, bgra_to_hsv_32f}
+            },
+            {
+                {bgr_to_hsv4_8u, 0, 0, 0, 0, bgr_to_hsv4_32f},
+                {bgra_to_hsv4_8u, 0, 0, 0, 0, bgra_to_hsv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hsv_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hsv_to_rgb_8u, 0, 0, 0, 0, hsv_to_rgb_32f},
+                {hsv4_to_rgb_8u, 0, 0, 0, 0, hsv4_to_rgb_32f}
+            },
+            {
+                {hsv_to_rgba_8u, 0, 0, 0, 0, hsv_to_rgba_32f},
+                {hsv4_to_rgba_8u, 0, 0, 0, 0, hsv4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hsv_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hsv_to_bgr_8u, 0, 0, 0, 0, hsv_to_bgr_32f},
+                {hsv4_to_bgr_8u, 0, 0, 0, 0, hsv4_to_bgr_32f}
+            },
+            {
+                {hsv_to_bgra_8u, 0, 0, 0, 0, hsv_to_bgra_32f},
+                {hsv4_to_bgra_8u, 0, 0, 0, 0, hsv4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_hls(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_hls_8u, 0, 0, 0, 0, rgb_to_hls_32f},
+                {rgba_to_hls_8u, 0, 0, 0, 0, rgba_to_hls_32f},
+            },
+            {
+                {rgb_to_hls4_8u, 0, 0, 0, 0, rgb_to_hls4_32f},
+                {rgba_to_hls4_8u, 0, 0, 0, 0, rgba_to_hls4_32f},
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_hls(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_hls_8u, 0, 0, 0, 0, bgr_to_hls_32f},
+                {bgra_to_hls_8u, 0, 0, 0, 0, bgra_to_hls_32f}
+            },
+            {
+                {bgr_to_hls4_8u, 0, 0, 0, 0, bgr_to_hls4_32f},
+                {bgra_to_hls4_8u, 0, 0, 0, 0, bgra_to_hls4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hls_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hls_to_rgb_8u, 0, 0, 0, 0, hls_to_rgb_32f},
+                {hls4_to_rgb_8u, 0, 0, 0, 0, hls4_to_rgb_32f}
+            },
+            {
+                {hls_to_rgba_8u, 0, 0, 0, 0, hls_to_rgba_32f},
+                {hls4_to_rgba_8u, 0, 0, 0, 0, hls4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hls_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hls_to_bgr_8u, 0, 0, 0, 0, hls_to_bgr_32f},
+                {hls4_to_bgr_8u, 0, 0, 0, 0, hls4_to_bgr_32f}
+            },
+            {
+                {hls_to_bgra_8u, 0, 0, 0, 0, hls_to_bgra_32f},
+                {hls4_to_bgra_8u, 0, 0, 0, 0, hls4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_hsv_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_hsv_full_8u, 0, 0, 0, 0, rgb_to_hsv_full_32f},
+                {rgba_to_hsv_full_8u, 0, 0, 0, 0, rgba_to_hsv_full_32f},
+            },
+            {
+                {rgb_to_hsv4_full_8u, 0, 0, 0, 0, rgb_to_hsv4_full_32f},
+                {rgba_to_hsv4_full_8u, 0, 0, 0, 0, rgba_to_hsv4_full_32f},
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_hsv_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_hsv_full_8u, 0, 0, 0, 0, bgr_to_hsv_full_32f},
+                {bgra_to_hsv_full_8u, 0, 0, 0, 0, bgra_to_hsv_full_32f}
+            },
+            {
+                {bgr_to_hsv4_full_8u, 0, 0, 0, 0, bgr_to_hsv4_full_32f},
+                {bgra_to_hsv4_full_8u, 0, 0, 0, 0, bgra_to_hsv4_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hsv_to_rgb_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hsv_to_rgb_full_8u, 0, 0, 0, 0, hsv_to_rgb_full_32f},
+                {hsv4_to_rgb_full_8u, 0, 0, 0, 0, hsv4_to_rgb_full_32f}
+            },
+            {
+                {hsv_to_rgba_full_8u, 0, 0, 0, 0, hsv_to_rgba_full_32f},
+                {hsv4_to_rgba_full_8u, 0, 0, 0, 0, hsv4_to_rgba_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hsv_to_bgr_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hsv_to_bgr_full_8u, 0, 0, 0, 0, hsv_to_bgr_full_32f},
+                {hsv4_to_bgr_full_8u, 0, 0, 0, 0, hsv4_to_bgr_full_32f}
+            },
+            {
+                {hsv_to_bgra_full_8u, 0, 0, 0, 0, hsv_to_bgra_full_32f},
+                {hsv4_to_bgra_full_8u, 0, 0, 0, 0, hsv4_to_bgra_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_hls_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {rgb_to_hls_full_8u, 0, 0, 0, 0, rgb_to_hls_full_32f},
+                {rgba_to_hls_full_8u, 0, 0, 0, 0, rgba_to_hls_full_32f},
+            },
+            {
+                {rgb_to_hls4_full_8u, 0, 0, 0, 0, rgb_to_hls4_full_32f},
+                {rgba_to_hls4_full_8u, 0, 0, 0, 0, rgba_to_hls4_full_32f},
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_hls_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {bgr_to_hls_full_8u, 0, 0, 0, 0, bgr_to_hls_full_32f},
+                {bgra_to_hls_full_8u, 0, 0, 0, 0, bgra_to_hls_full_32f}
+            },
+            {
+                {bgr_to_hls4_full_8u, 0, 0, 0, 0, bgr_to_hls4_full_32f},
+                {bgra_to_hls4_full_8u, 0, 0, 0, 0, bgra_to_hls4_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hls_to_rgb_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hls_to_rgb_full_8u, 0, 0, 0, 0, hls_to_rgb_full_32f},
+                {hls4_to_rgb_full_8u, 0, 0, 0, 0, hls4_to_rgb_full_32f}
+            },
+            {
+                {hls_to_rgba_full_8u, 0, 0, 0, 0, hls_to_rgba_full_32f},
+                {hls4_to_rgba_full_8u, 0, 0, 0, 0, hls4_to_rgba_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void hls_to_bgr_full(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][6] =
+        {
+            {
+                {hls_to_bgr_full_8u, 0, 0, 0, 0, hls_to_bgr_full_32f},
+                {hls4_to_bgr_full_8u, 0, 0, 0, 0, hls4_to_bgr_full_32f}
+            },
+            {
+                {hls_to_bgra_full_8u, 0, 0, 0, 0, hls_to_bgra_full_32f},
+                {hls4_to_bgra_full_8u, 0, 0, 0, 0, hls4_to_bgra_full_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth()](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_lab(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {bgr_to_lab_8u, bgr_to_lab_32f},
+                {bgra_to_lab_8u, bgra_to_lab_32f}
+            },
+            {
+                {bgr_to_lab4_8u, bgr_to_lab4_32f},
+                {bgra_to_lab4_8u, bgra_to_lab4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_lab(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {rgb_to_lab_8u, rgb_to_lab_32f},
+                {rgba_to_lab_8u, rgba_to_lab_32f}
+            },
+            {
+                {rgb_to_lab4_8u, rgb_to_lab4_32f},
+                {rgba_to_lab4_8u, rgba_to_lab4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lbgr_to_lab(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lbgr_to_lab_8u, lbgr_to_lab_32f},
+                {lbgra_to_lab_8u, lbgra_to_lab_32f}
+            },
+            {
+                {lbgr_to_lab4_8u, lbgr_to_lab4_32f},
+                {lbgra_to_lab4_8u, lbgra_to_lab4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lrgb_to_lab(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lrgb_to_lab_8u, lrgb_to_lab_32f},
+                {lrgba_to_lab_8u, lrgba_to_lab_32f}
+            },
+            {
+                {lrgb_to_lab4_8u, lrgb_to_lab4_32f},
+                {lrgba_to_lab4_8u, lrgba_to_lab4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lab_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lab_to_bgr_8u, lab_to_bgr_32f},
+                {lab4_to_bgr_8u, lab4_to_bgr_32f}
+            },
+            {
+                {lab_to_bgra_8u, lab_to_bgra_32f},
+                {lab4_to_bgra_8u, lab4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lab_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lab_to_rgb_8u, lab_to_rgb_32f},
+                {lab4_to_rgb_8u, lab4_to_rgb_32f}
+            },
+            {
+                {lab_to_rgba_8u, lab_to_rgba_32f},
+                {lab4_to_rgba_8u, lab4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lab_to_lbgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lab_to_lbgr_8u, lab_to_lbgr_32f},
+                {lab4_to_lbgr_8u, lab4_to_lbgr_32f}
+            },
+            {
+                {lab_to_lbgra_8u, lab_to_lbgra_32f},
+                {lab4_to_lbgra_8u, lab4_to_lbgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lab_to_lrgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lab_to_lrgb_8u, lab_to_lrgb_32f},
+                {lab4_to_lrgb_8u, lab4_to_lrgb_32f}
+            },
+            {
+                {lab_to_lrgba_8u, lab_to_lrgba_32f},
+                {lab4_to_lrgba_8u, lab4_to_lrgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void bgr_to_luv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {bgr_to_luv_8u, bgr_to_luv_32f},
+                {bgra_to_luv_8u, bgra_to_luv_32f}
+            },
+            {
+                {bgr_to_luv4_8u, bgr_to_luv4_32f},
+                {bgra_to_luv4_8u, bgra_to_luv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgb_to_luv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {rgb_to_luv_8u, rgb_to_luv_32f},
+                {rgba_to_luv_8u, rgba_to_luv_32f}
+            },
+            {
+                {rgb_to_luv4_8u, rgb_to_luv4_32f},
+                {rgba_to_luv4_8u, rgba_to_luv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lbgr_to_luv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lbgr_to_luv_8u, lbgr_to_luv_32f},
+                {lbgra_to_luv_8u, lbgra_to_luv_32f}
+            },
+            {
+                {lbgr_to_luv4_8u, lbgr_to_luv4_32f},
+                {lbgra_to_luv4_8u, lbgra_to_luv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void lrgb_to_luv(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {lrgb_to_luv_8u, lrgb_to_luv_32f},
+                {lrgba_to_luv_8u, lrgba_to_luv_32f}
+            },
+            {
+                {lrgb_to_luv4_8u, lrgb_to_luv4_32f},
+                {lrgba_to_luv4_8u, lrgba_to_luv4_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void luv_to_bgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {luv_to_bgr_8u, luv_to_bgr_32f},
+                {luv4_to_bgr_8u, luv4_to_bgr_32f}
+            },
+            {
+                {luv_to_bgra_8u, luv_to_bgra_32f},
+                {luv4_to_bgra_8u, luv4_to_bgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void luv_to_rgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {luv_to_rgb_8u, luv_to_rgb_32f},
+                {luv4_to_rgb_8u, luv4_to_rgb_32f}
+            },
+            {
+                {luv_to_rgba_8u, luv_to_rgba_32f},
+                {luv4_to_rgba_8u, luv4_to_rgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void luv_to_lbgr(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {luv_to_lbgr_8u, luv_to_lbgr_32f},
+                {luv4_to_lbgr_8u, luv4_to_lbgr_32f}
+            },
+            {
+                {luv_to_lbgra_8u, luv_to_lbgra_32f},
+                {luv4_to_lbgra_8u, luv4_to_lbgra_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void luv_to_lrgb(InputArray _src, OutputArray _dst, int dcn, Stream& stream)
+    {
+        using namespace cv::gpu::cudev;
+        static const gpu_func_t funcs[2][2][2] =
+        {
+            {
+                {luv_to_lrgb_8u, luv_to_lrgb_32f},
+                {luv4_to_lrgb_8u, luv4_to_lrgb_32f}
+            },
+            {
+                {luv_to_lrgba_8u, luv_to_lrgba_32f},
+                {luv4_to_lrgba_8u, luv4_to_lrgba_32f}
+            }
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.depth() == CV_8U || src.depth() == CV_32F );
+        CV_Assert( src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[dcn == 4][src.channels() == 4][src.depth() == CV_32F](src, dst, StreamAccessor::getStream(stream));
+    }
+
+    void rgba_to_mbgra(InputArray _src, OutputArray _dst, int, Stream& _stream)
+    {
+    #if (CUDA_VERSION < 5000)
+        (void) _src;
+        (void) _dst;
+        (void) _stream;
+        CV_Error( Error::StsBadFlag, "Unknown/unsupported color conversion code" );
+    #else
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC4 || src.type() == CV_16UC4 );
+
+        _dst.create(src.size(), src.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+        NppStreamHandler h(stream);
+
+        NppiSize oSizeROI;
+        oSizeROI.width = src.cols;
+        oSizeROI.height = src.rows;
+
+        if (src.depth() == CV_8U)
+            nppSafeCall( nppiAlphaPremul_8u_AC4R(src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), oSizeROI) );
+        else
+            nppSafeCall( nppiAlphaPremul_16u_AC4R(src.ptr<Npp16u>(), static_cast<int>(src.step), dst.ptr<Npp16u>(), static_cast<int>(dst.step), oSizeROI) );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    #endif
+    }
+
+    void bayer_to_bgr(InputArray _src, OutputArray _dst, int dcn, bool blue_last, bool start_with_green, Stream& stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+        static const func_t funcs[3][4] =
+        {
+            {0,0,Bayer2BGR_8u_gpu<3>, Bayer2BGR_8u_gpu<4>},
+            {0,0,0,0},
+            {0,0,Bayer2BGR_16u_gpu<3>, Bayer2BGR_16u_gpu<4>}
+        };
+
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 || src.type() == CV_16UC1 );
+        CV_Assert( src.rows > 2 && src.cols > 2 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()][dcn - 1](src, dst, blue_last, start_with_green, StreamAccessor::getStream(stream));
+    }
+    void bayerBG_to_bgr(InputArray src, OutputArray dst, int dcn, Stream& stream)
+    {
+        bayer_to_bgr(src, dst, dcn, false, false, stream);
+    }
+    void bayerGB_to_bgr(InputArray src, OutputArray dst, int dcn, Stream& stream)
+    {
+        bayer_to_bgr(src, dst, dcn, false, true, stream);
+    }
+    void bayerRG_to_bgr(InputArray src, OutputArray dst, int dcn, Stream& stream)
+    {
+        bayer_to_bgr(src, dst, dcn, true, false, stream);
+    }
+    void bayerGR_to_bgr(InputArray src, OutputArray dst, int dcn, Stream& stream)
+    {
+        bayer_to_bgr(src, dst, dcn, true, true, stream);
+    }
+
+    void bayer_to_gray(InputArray _src, OutputArray _dst, bool blue_last, bool start_with_green, Stream& stream)
+    {
+        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+        static const func_t funcs[3] =
+        {
+            Bayer2BGR_8u_gpu<1>,
+            0,
+            Bayer2BGR_16u_gpu<1>,
+        };
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 || src.type() == CV_16UC1 );
+        CV_Assert( src.rows > 2 && src.cols > 2 );
+
+        _dst.create(src.size(), CV_MAKE_TYPE(src.depth(), 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        funcs[src.depth()](src, dst, blue_last, start_with_green, StreamAccessor::getStream(stream));
+    }
+    void bayerBG_to_gray(InputArray src, OutputArray dst, int /*dcn*/, Stream& stream)
+    {
+        bayer_to_gray(src, dst, false, false, stream);
+    }
+    void bayerGB_to_gray(InputArray src, OutputArray dst, int /*dcn*/, Stream& stream)
+    {
+        bayer_to_gray(src, dst, false, true, stream);
+    }
+    void bayerRG_to_gray(InputArray src, OutputArray dst, int /*dcn*/, Stream& stream)
+    {
+        bayer_to_gray(src, dst, true, false, stream);
+    }
+    void bayerGR_to_gray(InputArray src, OutputArray dst, int /*dcn*/, Stream& stream)
+    {
+        bayer_to_gray(src, dst, true, true, stream);
+    }
+}
+
+////////////////////////////////////////////////////////////////////////
+// cvtColor
+
+void cv::gpu::cvtColor(InputArray src, OutputArray dst, int code, int dcn, Stream& stream)
+{
+    typedef void (*func_t)(InputArray src, OutputArray dst, int dcn, Stream& stream);
+    static const func_t funcs[] =
+    {
+        bgr_to_bgra,            // CV_BGR2BGRA    =0
+        bgra_to_bgr,            // CV_BGRA2BGR    =1
+        bgr_to_rgba,            // CV_BGR2RGBA    =2
+        bgra_to_rgb,            // CV_RGBA2BGR    =3
+        bgr_to_rgb,             // CV_BGR2RGB     =4
+        bgra_to_rgba,           // CV_BGRA2RGBA   =5
+
+        bgr_to_gray,            // CV_BGR2GRAY    =6
+        rgb_to_gray,            // CV_RGB2GRAY    =7
+        gray_to_bgr,            // CV_GRAY2BGR    =8
+        gray_to_bgra,           // CV_GRAY2BGRA   =9
+        bgra_to_gray,           // CV_BGRA2GRAY   =10
+        rgba_to_gray,           // CV_RGBA2GRAY   =11
+
+        bgr_to_bgr565,          // CV_BGR2BGR565  =12
+        rgb_to_bgr565,          // CV_RGB2BGR565  =13
+        bgr565_to_bgr,          // CV_BGR5652BGR  =14
+        bgr565_to_rgb,          // CV_BGR5652RGB  =15
+        bgra_to_bgr565,         // CV_BGRA2BGR565 =16
+        rgba_to_bgr565,         // CV_RGBA2BGR565 =17
+        bgr565_to_bgra,         // CV_BGR5652BGRA =18
+        bgr565_to_rgba,         // CV_BGR5652RGBA =19
+
+        gray_to_bgr565,         // CV_GRAY2BGR565 =20
+        bgr565_to_gray,         // CV_BGR5652GRAY =21
+
+        bgr_to_bgr555,          // CV_BGR2BGR555  =22
+        rgb_to_bgr555,          // CV_RGB2BGR555  =23
+        bgr555_to_bgr,          // CV_BGR5552BGR  =24
+        bgr555_to_rgb,          // CV_BGR5552RGB  =25
+        bgra_to_bgr555,         // CV_BGRA2BGR555 =26
+        rgba_to_bgr555,         // CV_RGBA2BGR555 =27
+        bgr555_to_bgra,         // CV_BGR5552BGRA =28
+        bgr555_to_rgba,         // CV_BGR5552RGBA =29
+
+        gray_to_bgr555,         // CV_GRAY2BGR555 =30
+        bgr555_to_gray,         // CV_BGR5552GRAY =31
+
+        bgr_to_xyz,             // CV_BGR2XYZ     =32
+        rgb_to_xyz,             // CV_RGB2XYZ     =33
+        xyz_to_bgr,             // CV_XYZ2BGR     =34
+        xyz_to_rgb,             // CV_XYZ2RGB     =35
+
+        bgr_to_YCrCb,           // CV_BGR2YCrCb   =36
+        rgb_to_YCrCb,           // CV_RGB2YCrCb   =37
+        YCrCb_to_bgr,           // CV_YCrCb2BGR   =38
+        YCrCb_to_rgb,           // CV_YCrCb2RGB   =39
+
+        bgr_to_hsv,             // CV_BGR2HSV     =40
+        rgb_to_hsv,             // CV_RGB2HSV     =41
+
+        0,                      //                =42
+        0,                      //                =43
+
+        bgr_to_lab,             // CV_BGR2Lab     =44
+        rgb_to_lab,             // CV_RGB2Lab     =45
+
+        bayerBG_to_bgr,         // CV_BayerBG2BGR =46
+        bayerGB_to_bgr,         // CV_BayerGB2BGR =47
+        bayerRG_to_bgr,         // CV_BayerRG2BGR =48
+        bayerGR_to_bgr,         // CV_BayerGR2BGR =49
+
+        bgr_to_luv,             // CV_BGR2Luv     =50
+        rgb_to_luv,             // CV_RGB2Luv     =51
+
+        bgr_to_hls,             // CV_BGR2HLS     =52
+        rgb_to_hls,             // CV_RGB2HLS     =53
+
+        hsv_to_bgr,             // CV_HSV2BGR     =54
+        hsv_to_rgb,             // CV_HSV2RGB     =55
+
+        lab_to_bgr,             // CV_Lab2BGR     =56
+        lab_to_rgb,             // CV_Lab2RGB     =57
+        luv_to_bgr,             // CV_Luv2BGR     =58
+        luv_to_rgb,             // CV_Luv2RGB     =59
+
+        hls_to_bgr,             // CV_HLS2BGR     =60
+        hls_to_rgb,             // CV_HLS2RGB     =61
+
+        0,                      // CV_BayerBG2BGR_VNG =62
+        0,                      // CV_BayerGB2BGR_VNG =63
+        0,                      // CV_BayerRG2BGR_VNG =64
+        0,                      // CV_BayerGR2BGR_VNG =65
+
+        bgr_to_hsv_full,        // CV_BGR2HSV_FULL = 66
+        rgb_to_hsv_full,        // CV_RGB2HSV_FULL = 67
+        bgr_to_hls_full,        // CV_BGR2HLS_FULL = 68
+        rgb_to_hls_full,        // CV_RGB2HLS_FULL = 69
+
+        hsv_to_bgr_full,        // CV_HSV2BGR_FULL = 70
+        hsv_to_rgb_full,        // CV_HSV2RGB_FULL = 71
+        hls_to_bgr_full,        // CV_HLS2BGR_FULL = 72
+        hls_to_rgb_full,        // CV_HLS2RGB_FULL = 73
+
+        lbgr_to_lab,            // CV_LBGR2Lab     = 74
+        lrgb_to_lab,            // CV_LRGB2Lab     = 75
+        lbgr_to_luv,            // CV_LBGR2Luv     = 76
+        lrgb_to_luv,            // CV_LRGB2Luv     = 77
+
+        lab_to_lbgr,            // CV_Lab2LBGR     = 78
+        lab_to_lrgb,            // CV_Lab2LRGB     = 79
+        luv_to_lbgr,            // CV_Luv2LBGR     = 80
+        luv_to_lrgb,            // CV_Luv2LRGB     = 81
+
+        bgr_to_yuv,             // CV_BGR2YUV      = 82
+        rgb_to_yuv,             // CV_RGB2YUV      = 83
+        yuv_to_bgr,             // CV_YUV2BGR      = 84
+        yuv_to_rgb,             // CV_YUV2RGB      = 85
+
+        bayerBG_to_gray,        // CV_BayerBG2GRAY = 86
+        bayerGB_to_gray,        // CV_BayerGB2GRAY = 87
+        bayerRG_to_gray,        // CV_BayerRG2GRAY = 88
+        bayerGR_to_gray,        // CV_BayerGR2GRAY = 89
+
+        //YUV 4:2:0 formats family
+        0,                      // CV_YUV2RGB_NV12 = 90,
+        0,                      // CV_YUV2BGR_NV12 = 91,
+        0,                      // CV_YUV2RGB_NV21 = 92,
+        0,                      // CV_YUV2BGR_NV21 = 93,
+
+        0,                      // CV_YUV2RGBA_NV12 = 94,
+        0,                      // CV_YUV2BGRA_NV12 = 95,
+        0,                      // CV_YUV2RGBA_NV21 = 96,
+        0,                      // CV_YUV2BGRA_NV21 = 97,
+
+        0,                      // CV_YUV2RGB_YV12 = 98,
+        0,                      // CV_YUV2BGR_YV12 = 99,
+        0,                      // CV_YUV2RGB_IYUV = 100,
+        0,                      // CV_YUV2BGR_IYUV = 101,
+
+        0,                      // CV_YUV2RGBA_YV12 = 102,
+        0,                      // CV_YUV2BGRA_YV12 = 103,
+        0,                      // CV_YUV2RGBA_IYUV = 104,
+        0,                      // CV_YUV2BGRA_IYUV = 105,
+
+        0,                      // CV_YUV2GRAY_420 = 106,
+
+        //YUV 4:2:2 formats family
+        0,                      // CV_YUV2RGB_UYVY = 107,
+        0,                      // CV_YUV2BGR_UYVY = 108,
+        0,                      // //CV_YUV2RGB_VYUY = 109,
+        0,                      // //CV_YUV2BGR_VYUY = 110,
+
+        0,                      // CV_YUV2RGBA_UYVY = 111,
+        0,                      // CV_YUV2BGRA_UYVY = 112,
+        0,                      // //CV_YUV2RGBA_VYUY = 113,
+        0,                      // //CV_YUV2BGRA_VYUY = 114,
+
+        0,                      // CV_YUV2RGB_YUY2 = 115,
+        0,                      // CV_YUV2BGR_YUY2 = 116,
+        0,                      // CV_YUV2RGB_YVYU = 117,
+        0,                      // CV_YUV2BGR_YVYU = 118,
+
+        0,                      // CV_YUV2RGBA_YUY2 = 119,
+        0,                      // CV_YUV2BGRA_YUY2 = 120,
+        0,                      // CV_YUV2RGBA_YVYU = 121,
+        0,                      // CV_YUV2BGRA_YVYU = 122,
+
+        0,                      // CV_YUV2GRAY_UYVY = 123,
+        0,                      // CV_YUV2GRAY_YUY2 = 124,
+
+        // alpha premultiplication
+        rgba_to_mbgra,          // CV_RGBA2mRGBA = 125,
+        0,                      // CV_mRGBA2RGBA = 126,
+
+        0,                      // CV_COLORCVT_MAX  = 127
+    };
+
+    CV_Assert( code < 128 );
+
+    func_t func = funcs[code];
+
+    if (func == 0)
+        CV_Error(Error::StsBadFlag, "Unknown/unsupported color conversion code");
+
+    func(src, dst, dcn, stream);
+}
+
+////////////////////////////////////////////////////////////////////////
+// demosaicing
+
+void cv::gpu::demosaicing(InputArray _src, OutputArray _dst, int code, int dcn, Stream& stream)
+{
+    switch (code)
+    {
+    case cv::COLOR_BayerBG2GRAY: case cv::COLOR_BayerGB2GRAY: case cv::COLOR_BayerRG2GRAY: case cv::COLOR_BayerGR2GRAY:
+        bayer_to_gray(_src, _dst, code == cv::COLOR_BayerBG2GRAY || code == cv::COLOR_BayerGB2GRAY, code == cv::COLOR_BayerGB2GRAY || code == cv::COLOR_BayerGR2GRAY, stream);
+        break;
+
+    case cv::COLOR_BayerBG2BGR: case cv::COLOR_BayerGB2BGR: case cv::COLOR_BayerRG2BGR: case cv::COLOR_BayerGR2BGR:
+        bayer_to_bgr(_src, _dst, dcn, code == cv::COLOR_BayerBG2BGR || code == cv::COLOR_BayerGB2BGR, code == cv::COLOR_BayerGB2BGR || code == cv::COLOR_BayerGR2BGR, stream);
+        break;
+
+    case COLOR_BayerBG2BGR_MHT: case COLOR_BayerGB2BGR_MHT: case COLOR_BayerRG2BGR_MHT: case COLOR_BayerGR2BGR_MHT:
+    {
+        if (dcn <= 0) dcn = 3;
+
+        GpuMat src = _src.getGpuMat();
+        const int depth = _src.depth();
+
+        CV_Assert( depth == CV_8U );
+        CV_Assert( src.channels() == 1 );
+        CV_Assert( dcn == 3 || dcn == 4 );
+
+        _dst.create(_src.size(), CV_MAKE_TYPE(depth, dcn));
+        GpuMat dst = _dst.getGpuMat();
+
+        dst.setTo(Scalar::all(0), stream);
+
+        Size wholeSize;
+        Point ofs;
+        src.locateROI(wholeSize, ofs);
+        PtrStepSzb srcWhole(wholeSize.height, wholeSize.width, src.datastart, src.step);
+
+        const int2 firstRed = make_int2(code == COLOR_BayerRG2BGR_MHT || code == COLOR_BayerGB2BGR_MHT ? 0 : 1,
+                                        code == COLOR_BayerRG2BGR_MHT || code == COLOR_BayerGR2BGR_MHT ? 0 : 1);
+
+        if (dcn == 3)
+            cv::gpu::cudev::MHCdemosaic<3>(srcWhole, make_int2(ofs.x, ofs.y), dst, firstRed, StreamAccessor::getStream(stream));
+        else
+            cv::gpu::cudev::MHCdemosaic<4>(srcWhole, make_int2(ofs.x, ofs.y), dst, firstRed, StreamAccessor::getStream(stream));
+
+        break;
+    }
+
+    case COLOR_BayerBG2GRAY_MHT: case COLOR_BayerGB2GRAY_MHT: case COLOR_BayerRG2GRAY_MHT: case COLOR_BayerGR2GRAY_MHT:
+    {
+        GpuMat src = _src.getGpuMat();
+        const int depth = _src.depth();
+
+        CV_Assert( depth == CV_8U );
+
+        _dst.create(_src.size(), CV_MAKE_TYPE(depth, 1));
+        GpuMat dst = _dst.getGpuMat();
+
+        dst.setTo(Scalar::all(0), stream);
+
+        Size wholeSize;
+        Point ofs;
+        src.locateROI(wholeSize, ofs);
+        PtrStepSzb srcWhole(wholeSize.height, wholeSize.width, src.datastart, src.step);
+
+        const int2 firstRed = make_int2(code == COLOR_BayerRG2BGR_MHT || code == COLOR_BayerGB2BGR_MHT ? 0 : 1,
+                                        code == COLOR_BayerRG2BGR_MHT || code == COLOR_BayerGR2BGR_MHT ? 0 : 1);
+
+        cv::gpu::cudev::MHCdemosaic<1>(srcWhole, make_int2(ofs.x, ofs.y), dst, firstRed, StreamAccessor::getStream(stream));
+
+        break;
+    }
+
+    default:
+        CV_Error(Error::StsBadFlag, "Unknown / unsupported color conversion code");
+    }
+}
+
+////////////////////////////////////////////////////////////////////////
+// swapChannels
+
+void cv::gpu::swapChannels(InputOutputArray _image, const int dstOrder[4], Stream& _stream)
+{
+    GpuMat image = _image.getGpuMat();
+
+    CV_Assert( image.type() == CV_8UC4 );
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+    NppStreamHandler h(stream);
+
+    NppiSize sz;
+    sz.width  = image.cols;
+    sz.height = image.rows;
+
+    nppSafeCall( nppiSwapChannels_8u_C4IR(image.ptr<Npp8u>(), static_cast<int>(image.step), sz, dstOrder) );
+
+    if (stream == 0)
+        cudaSafeCall( cudaDeviceSynchronize() );
+}
+
+////////////////////////////////////////////////////////////////////////
+// gammaCorrection
+
+void cv::gpu::gammaCorrection(InputArray _src, OutputArray _dst, bool forward, Stream& stream)
+{
+#if (CUDA_VERSION < 5000)
+    (void) _src;
+    (void) _dst;
+    (void) forward;
+    (void) stream;
+    CV_Error(Error::StsNotImplemented, "This function works only with CUDA 5.0 or higher");
+#else
+    typedef NppStatus (*func_t)(const Npp8u* pSrc, int nSrcStep, Npp8u* pDst, int nDstStep, NppiSize oSizeROI);
+    typedef NppStatus (*func_inplace_t)(Npp8u* pSrcDst, int nSrcDstStep, NppiSize oSizeROI);
+
+    static const func_t funcs[2][5] =
+    {
+        {0, 0, 0, nppiGammaInv_8u_C3R, nppiGammaInv_8u_AC4R},
+        {0, 0, 0, nppiGammaFwd_8u_C3R, nppiGammaFwd_8u_AC4R}
+    };
+    static const func_inplace_t funcs_inplace[2][5] =
+    {
+        {0, 0, 0, nppiGammaInv_8u_C3IR, nppiGammaInv_8u_AC4IR},
+        {0, 0, 0, nppiGammaFwd_8u_C3IR, nppiGammaFwd_8u_AC4IR}
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC3 || src.type() == CV_8UC4 );
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    NppStreamHandler h(StreamAccessor::getStream(stream));
+
+    NppiSize oSizeROI;
+    oSizeROI.width = src.cols;
+    oSizeROI.height = src.rows;
+
+    if (dst.data == src.data)
+        funcs_inplace[forward][src.channels()](dst.ptr<Npp8u>(), static_cast<int>(src.step), oSizeROI);
+    else
+        funcs[forward][src.channels()](src.ptr<Npp8u>(), static_cast<int>(src.step), dst.ptr<Npp8u>(), static_cast<int>(dst.step), oSizeROI);
+
+#endif
+}
+
+////////////////////////////////////////////////////////////////////////
+// alphaComp
+
+namespace
+{
+    template <int DEPTH> struct NppAlphaCompFunc
+    {
+        typedef typename NPPTypeTraits<DEPTH>::npp_type npp_t;
+
+        typedef NppStatus (*func_t)(const npp_t* pSrc1, int nSrc1Step, const npp_t* pSrc2, int nSrc2Step, npp_t* pDst, int nDstStep, NppiSize oSizeROI, NppiAlphaOp eAlphaOp);
+    };
+
+    template <int DEPTH, typename NppAlphaCompFunc<DEPTH>::func_t func> struct NppAlphaComp
+    {
+        typedef typename NPPTypeTraits<DEPTH>::npp_type npp_t;
+
+        static void call(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, NppiAlphaOp eAlphaOp, cudaStream_t stream)
+        {
+            NppStreamHandler h(stream);
+
+            NppiSize oSizeROI;
+            oSizeROI.width = img1.cols;
+            oSizeROI.height = img2.rows;
+
+            nppSafeCall( func(img1.ptr<npp_t>(), static_cast<int>(img1.step), img2.ptr<npp_t>(), static_cast<int>(img2.step),
+                              dst.ptr<npp_t>(), static_cast<int>(dst.step), oSizeROI, eAlphaOp) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+}
+
+void cv::gpu::alphaComp(InputArray _img1, InputArray _img2, OutputArray _dst, int alpha_op, Stream& stream)
+{
+    static const NppiAlphaOp npp_alpha_ops[] = {
+        NPPI_OP_ALPHA_OVER,
+        NPPI_OP_ALPHA_IN,
+        NPPI_OP_ALPHA_OUT,
+        NPPI_OP_ALPHA_ATOP,
+        NPPI_OP_ALPHA_XOR,
+        NPPI_OP_ALPHA_PLUS,
+        NPPI_OP_ALPHA_OVER_PREMUL,
+        NPPI_OP_ALPHA_IN_PREMUL,
+        NPPI_OP_ALPHA_OUT_PREMUL,
+        NPPI_OP_ALPHA_ATOP_PREMUL,
+        NPPI_OP_ALPHA_XOR_PREMUL,
+        NPPI_OP_ALPHA_PLUS_PREMUL,
+        NPPI_OP_ALPHA_PREMUL
+    };
+
+    typedef void (*func_t)(const GpuMat& img1, const GpuMat& img2, GpuMat& dst, NppiAlphaOp eAlphaOp, cudaStream_t stream);
+    static const func_t funcs[] =
+    {
+        NppAlphaComp<CV_8U, nppiAlphaComp_8u_AC4R>::call,
+        0,
+        NppAlphaComp<CV_16U, nppiAlphaComp_16u_AC4R>::call,
+        0,
+        NppAlphaComp<CV_32S, nppiAlphaComp_32s_AC4R>::call,
+        NppAlphaComp<CV_32F, nppiAlphaComp_32f_AC4R>::call
+    };
+
+    GpuMat img1 = _img1.getGpuMat();
+    GpuMat img2 = _img2.getGpuMat();
+
+    CV_Assert( img1.type() == CV_8UC4 || img1.type() == CV_16UC4 || img1.type() == CV_32SC4 || img1.type() == CV_32FC4 );
+    CV_Assert( img1.size() == img2.size() && img1.type() == img2.type() );
+
+    _dst.create(img1.size(), img1.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    const func_t func = funcs[img1.depth()];
+
+    func(img1, img2, dst, npp_alpha_ops[alpha_op], StreamAccessor::getStream(stream));
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/corners.cpp b/modules/gpuimgproc/src/corners.cpp
new file mode 100644
index 0000000000..6b53e6f85c
--- /dev/null
+++ b/modules/gpuimgproc/src/corners.cpp
@@ -0,0 +1,189 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) || !defined(HAVE_OPENCV_GPUFILTERS)
+
+Ptr<gpu::CornernessCriteria> cv::gpu::createHarrisCorner(int, int, int, double, int) { throw_no_cuda(); return Ptr<gpu::CornernessCriteria>(); }
+Ptr<gpu::CornernessCriteria> cv::gpu::createMinEigenValCorner(int, int, int, int) { throw_no_cuda(); return Ptr<gpu::CornernessCriteria>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void cornerHarris_gpu(int block_size, float k, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream);
+        void cornerMinEigenVal_gpu(int block_size, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    class CornerBase : public CornernessCriteria
+    {
+    protected:
+        CornerBase(int srcType, int blockSize, int ksize, int borderType);
+
+        void extractCovData(const GpuMat& src, Stream& stream);
+
+        int srcType_;
+        int blockSize_;
+        int ksize_;
+        int borderType_;
+        GpuMat Dx_, Dy_;
+
+    private:
+        Ptr<gpu::Filter> filterDx_, filterDy_;
+    };
+
+    CornerBase::CornerBase(int srcType, int blockSize, int ksize, int borderType) :
+        srcType_(srcType), blockSize_(blockSize), ksize_(ksize), borderType_(borderType)
+    {
+        CV_Assert( borderType_ == BORDER_REFLECT101 || borderType_ == BORDER_REPLICATE || borderType_ == BORDER_REFLECT );
+
+        const int sdepth = CV_MAT_DEPTH(srcType_);
+        const int cn = CV_MAT_CN(srcType_);
+
+        CV_Assert( cn == 1 );
+
+        double scale = static_cast<double>(1 << ((ksize_ > 0 ? ksize_ : 3) - 1)) * blockSize_;
+
+        if (ksize_ < 0)
+            scale *= 2.;
+
+        if (sdepth == CV_8U)
+            scale *= 255.;
+
+        scale = 1./scale;
+
+        if (ksize_ > 0)
+        {
+            filterDx_ = gpu::createSobelFilter(srcType, CV_32F, 1, 0, ksize_, scale, borderType_);
+            filterDy_ = gpu::createSobelFilter(srcType, CV_32F, 0, 1, ksize_, scale, borderType_);
+        }
+        else
+        {
+            filterDx_ = gpu::createScharrFilter(srcType, CV_32F, 1, 0, scale, borderType_);
+            filterDy_ = gpu::createScharrFilter(srcType, CV_32F, 0, 1, scale, borderType_);
+        }
+    }
+
+    void CornerBase::extractCovData(const GpuMat& src, Stream& stream)
+    {
+        CV_Assert( src.type() == srcType_ );
+        filterDx_->apply(src, Dx_, stream);
+        filterDy_->apply(src, Dy_, stream);
+    }
+
+    class Harris : public CornerBase
+    {
+    public:
+        Harris(int srcType, int blockSize, int ksize, double k, int borderType) :
+            CornerBase(srcType, blockSize, ksize, borderType), k_(static_cast<float>(k))
+        {
+        }
+
+        void compute(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        float k_;
+    };
+
+    void Harris::compute(InputArray _src, OutputArray _dst, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::imgproc;
+
+        GpuMat src = _src.getGpuMat();
+
+        extractCovData(src, stream);
+
+        _dst.create(src.size(), CV_32FC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        cornerHarris_gpu(blockSize_, k_, Dx_, Dy_, dst, borderType_, StreamAccessor::getStream(stream));
+    }
+
+    class MinEigenVal : public CornerBase
+    {
+    public:
+        MinEigenVal(int srcType, int blockSize, int ksize, int borderType) :
+            CornerBase(srcType, blockSize, ksize, borderType)
+        {
+        }
+
+        void compute(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+    private:
+        float k_;
+    };
+
+    void MinEigenVal::compute(InputArray _src, OutputArray _dst, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::imgproc;
+
+        GpuMat src = _src.getGpuMat();
+
+        extractCovData(src, stream);
+
+        _dst.create(src.size(), CV_32FC1);
+        GpuMat dst = _dst.getGpuMat();
+
+        cornerMinEigenVal_gpu(blockSize_, Dx_, Dy_, dst, borderType_, StreamAccessor::getStream(stream));
+    }
+}
+
+Ptr<gpu::CornernessCriteria> cv::gpu::createHarrisCorner(int srcType, int blockSize, int ksize, double k, int borderType)
+{
+    return makePtr<Harris>(srcType, blockSize, ksize, k, borderType);
+}
+
+Ptr<gpu::CornernessCriteria> cv::gpu::createMinEigenValCorner(int srcType, int blockSize, int ksize, int borderType)
+{
+    return makePtr<MinEigenVal>(srcType, blockSize, ksize, borderType);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/cuda/bilateral_filter.cu b/modules/gpuimgproc/src/cuda/bilateral_filter.cu
similarity index 92%
rename from modules/gpu/src/cuda/bilateral_filter.cu
rename to modules/gpuimgproc/src/cuda/bilateral_filter.cu
index 0f1d8537ef..3192f649b7 100644
--- a/modules/gpu/src/cuda/bilateral_filter.cu
+++ b/modules/gpuimgproc/src/cuda/bilateral_filter.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -29,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -43,11 +42,10 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
 using namespace cv::gpu;
 
@@ -57,7 +55,7 @@ typedef unsigned short ushort;
 //////////////////////////////////////////////////////////////////////////////////
 /// Bilateral filtering
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -135,7 +133,7 @@ namespace cv { namespace gpu { namespace device
             B<T> b(src.rows, src.cols);
 
             float sigma_spatial2_inv_half = -0.5f/(sigma_spatial * sigma_spatial);
-             float sigma_color2_inv_half = -0.5f/(sigma_color * sigma_color);
+            float sigma_color2_inv_half = -0.5f/(sigma_color * sigma_color);
 
             cudaSafeCall( cudaFuncSetCacheConfig (bilateral_kernel<T, B<T> >, cudaFuncCachePreferL1) );
             bilateral_kernel<<<grid, block>>>((PtrStepSz<T>)src, (PtrStepSz<T>)dst, b, kernel_size, sigma_spatial2_inv_half, sigma_color2_inv_half);
@@ -152,11 +150,11 @@ namespace cv { namespace gpu { namespace device
 
             static caller_t funcs[] =
             {
-                bilateral_caller<T, BrdReflect101>,
-                bilateral_caller<T, BrdReplicate>,
                 bilateral_caller<T, BrdConstant>,
+                bilateral_caller<T, BrdReplicate>,
                 bilateral_caller<T, BrdReflect>,
                 bilateral_caller<T, BrdWrap>,
+                bilateral_caller<T, BrdReflect101>
             };
             funcs[borderMode](src, dst, kernel_size, gauss_spatial_coeff, gauss_color_coeff, stream);
         }
@@ -165,7 +163,7 @@ namespace cv { namespace gpu { namespace device
 
 
 #define OCV_INSTANTIATE_BILATERAL_FILTER(T) \
-    template void cv::gpu::device::imgproc::bilateral_filter_gpu<T>(const PtrStepSzb&, PtrStepSzb, int, float, float, int, cudaStream_t);
+    template void cv::gpu::cudev::imgproc::bilateral_filter_gpu<T>(const PtrStepSzb&, PtrStepSzb, int, float, float, int, cudaStream_t);
 
 OCV_INSTANTIATE_BILATERAL_FILTER(uchar)
 //OCV_INSTANTIATE_BILATERAL_FILTER(uchar2)
@@ -198,4 +196,4 @@ OCV_INSTANTIATE_BILATERAL_FILTER(float3)
 OCV_INSTANTIATE_BILATERAL_FILTER(float4)
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/blend.cu b/modules/gpuimgproc/src/cuda/blend.cu
similarity index 95%
rename from modules/gpu/src/cuda/blend.cu
rename to modules/gpuimgproc/src/cuda/blend.cu
index 614ccd2964..be8c0b2f35 100644
--- a/modules/gpu/src/cuda/blend.cu
+++ b/modules/gpuimgproc/src/cuda/blend.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace blend
     {
@@ -115,7 +115,7 @@ namespace cv { namespace gpu { namespace device
                 cudaSafeCall(cudaDeviceSynchronize());
         }
     } // namespace blend
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/build_point_list.cu b/modules/gpuimgproc/src/cuda/build_point_list.cu
new file mode 100644
index 0000000000..c5f2b23f6f
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/build_point_list.cu
@@ -0,0 +1,138 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough
+    {
+        __device__ int g_counter;
+
+        template <int PIXELS_PER_THREAD>
+        __global__ void buildPointList(const PtrStepSzb src, unsigned int* list)
+        {
+            __shared__ unsigned int s_queues[4][32 * PIXELS_PER_THREAD];
+            __shared__ int s_qsize[4];
+            __shared__ int s_globStart[4];
+
+            const int x = blockIdx.x * blockDim.x * PIXELS_PER_THREAD + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (threadIdx.x == 0)
+                s_qsize[threadIdx.y] = 0;
+            __syncthreads();
+
+            if (y < src.rows)
+            {
+                // fill the queue
+                const uchar* srcRow = src.ptr(y);
+                for (int i = 0, xx = x; i < PIXELS_PER_THREAD && xx < src.cols; ++i, xx += blockDim.x)
+                {
+                    if (srcRow[xx])
+                    {
+                        const unsigned int val = (y << 16) | xx;
+                        const int qidx = Emulation::smem::atomicAdd(&s_qsize[threadIdx.y], 1);
+                        s_queues[threadIdx.y][qidx] = val;
+                    }
+                }
+            }
+
+            __syncthreads();
+
+            // let one thread reserve the space required in the global list
+            if (threadIdx.x == 0 && threadIdx.y == 0)
+            {
+                // find how many items are stored in each list
+                int totalSize = 0;
+                for (int i = 0; i < blockDim.y; ++i)
+                {
+                    s_globStart[i] = totalSize;
+                    totalSize += s_qsize[i];
+                }
+
+                // calculate the offset in the global list
+                const int globalOffset = atomicAdd(&g_counter, totalSize);
+                for (int i = 0; i < blockDim.y; ++i)
+                    s_globStart[i] += globalOffset;
+            }
+
+            __syncthreads();
+
+            // copy local queues to global queue
+            const int qsize = s_qsize[threadIdx.y];
+            int gidx = s_globStart[threadIdx.y] + threadIdx.x;
+            for(int i = threadIdx.x; i < qsize; i += blockDim.x, gidx += blockDim.x)
+                list[gidx] = s_queues[threadIdx.y][i];
+        }
+
+        int buildPointList_gpu(PtrStepSzb src, unsigned int* list)
+        {
+            const int PIXELS_PER_THREAD = 16;
+
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 4);
+            const dim3 grid(divUp(src.cols, block.x * PIXELS_PER_THREAD), divUp(src.rows, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(buildPointList<PIXELS_PER_THREAD>, cudaFuncCachePreferShared) );
+
+            buildPointList<PIXELS_PER_THREAD><<<grid, block>>>(src, list);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            return totalCount;
+        }
+    }
+}}}
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/canny.cu b/modules/gpuimgproc/src/cuda/canny.cu
similarity index 95%
rename from modules/gpu/src/cuda/canny.cu
rename to modules/gpuimgproc/src/cuda/canny.cu
index 1afcddc9c9..271fffbc7d 100644
--- a/modules/gpu/src/cuda/canny.cu
+++ b/modules/gpuimgproc/src/cuda/canny.cu
@@ -43,15 +43,15 @@
 #if !defined CUDA_DISABLER
 
 #include <utility>
-#include <algorithm>//std::swap
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/emulation.hpp"
-#include "opencv2/gpu/device/transform.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/utility.hpp"
+#include <algorithm>
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/utility.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 namespace canny
 {
@@ -62,8 +62,8 @@ namespace canny
             return ::abs(x) + ::abs(y);
         }
 
-        __device__ __forceinline__ L1() {}
-        __device__ __forceinline__ L1(const L1&) {}
+        __host__ __device__ __forceinline__ L1() {}
+        __host__ __device__ __forceinline__ L1(const L1&) {}
     };
     struct L2 : binary_function<int, int, float>
     {
@@ -72,12 +72,12 @@ namespace canny
             return ::sqrtf(x * x + y * y);
         }
 
-        __device__ __forceinline__ L2() {}
-        __device__ __forceinline__ L2(const L2&) {}
+        __host__ __device__ __forceinline__ L2() {}
+        __host__ __device__ __forceinline__ L2(const L2&) {}
     };
 }
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <> struct TransformFunctorTraits<canny::L1> : DefaultTransformFunctorTraits<canny::L1>
     {
@@ -94,8 +94,8 @@ namespace canny
     texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_src(false, cudaFilterModePoint, cudaAddressModeClamp);
     struct SrcTex
     {
-        const int xoff;
-        const int yoff;
+        int xoff;
+        int yoff;
         __host__ SrcTex(int _xoff, int _yoff) : xoff(_xoff), yoff(_yoff) {}
 
         __device__ __forceinline__ int operator ()(int y, int x) const
@@ -470,12 +470,12 @@ namespace canny
             return (uchar)(-(e >> 1));
         }
 
-        __device__ __forceinline__ GetEdges() {}
-        __device__ __forceinline__ GetEdges(const GetEdges&) {}
+        __host__ __device__ __forceinline__ GetEdges() {}
+        __host__ __device__ __forceinline__ GetEdges(const GetEdges&) {}
     };
 }
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <> struct TransformFunctorTraits<canny::GetEdges> : DefaultTransformFunctorTraits<canny::GetEdges>
     {
diff --git a/modules/gpuimgproc/src/cuda/clahe.cu b/modules/gpuimgproc/src/cuda/clahe.cu
new file mode 100644
index 0000000000..7c6645749b
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/clahe.cu
@@ -0,0 +1,186 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/scan.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace clahe
+{
+    __global__ void calcLutKernel(const PtrStepb src, PtrStepb lut,
+                                  const int2 tileSize, const int tilesX,
+                                  const int clipLimit, const float lutScale)
+    {
+        __shared__ int smem[512];
+
+        const int tx = blockIdx.x;
+        const int ty = blockIdx.y;
+        const unsigned int tid = threadIdx.y * blockDim.x + threadIdx.x;
+
+        smem[tid] = 0;
+        __syncthreads();
+
+        for (int i = threadIdx.y; i < tileSize.y; i += blockDim.y)
+        {
+            const uchar* srcPtr = src.ptr(ty * tileSize.y + i) + tx * tileSize.x;
+            for (int j = threadIdx.x; j < tileSize.x; j += blockDim.x)
+            {
+                const int data = srcPtr[j];
+                Emulation::smem::atomicAdd(&smem[data], 1);
+            }
+        }
+
+        __syncthreads();
+
+        int tHistVal = smem[tid];
+
+        __syncthreads();
+
+        if (clipLimit > 0)
+        {
+            // clip histogram bar
+
+            int clipped = 0;
+            if (tHistVal > clipLimit)
+            {
+                clipped = tHistVal - clipLimit;
+                tHistVal = clipLimit;
+            }
+
+            // find number of overall clipped samples
+
+            reduce<256>(smem, clipped, tid, plus<int>());
+
+            // broadcast evaluated value
+
+            __shared__ int totalClipped;
+
+            if (tid == 0)
+                totalClipped = clipped;
+            __syncthreads();
+
+            // redistribute clipped samples evenly
+
+            int redistBatch = totalClipped / 256;
+            tHistVal += redistBatch;
+
+            int residual = totalClipped - redistBatch * 256;
+            if (tid < residual)
+                ++tHistVal;
+        }
+
+        const int lutVal = blockScanInclusive<256>(tHistVal, smem, tid);
+
+        lut(ty * tilesX + tx, tid) = saturate_cast<uchar>(__float2int_rn(lutScale * lutVal));
+    }
+
+    void calcLut(PtrStepSzb src, PtrStepb lut, int tilesX, int tilesY, int2 tileSize, int clipLimit, float lutScale, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(tilesX, tilesY);
+
+        calcLutKernel<<<grid, block, 0, stream>>>(src, lut, tileSize, tilesX, clipLimit, lutScale);
+
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    __global__ void tranformKernel(const PtrStepSzb src, PtrStepb dst, const PtrStepb lut, const int2 tileSize, const int tilesX, const int tilesY)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= src.cols || y >= src.rows)
+            return;
+
+        const float tyf = (static_cast<float>(y) / tileSize.y) - 0.5f;
+        int ty1 = __float2int_rd(tyf);
+        int ty2 = ty1 + 1;
+        const float ya = tyf - ty1;
+        ty1 = ::max(ty1, 0);
+        ty2 = ::min(ty2, tilesY - 1);
+
+        const float txf = (static_cast<float>(x) / tileSize.x) - 0.5f;
+        int tx1 = __float2int_rd(txf);
+        int tx2 = tx1 + 1;
+        const float xa = txf - tx1;
+        tx1 = ::max(tx1, 0);
+        tx2 = ::min(tx2, tilesX - 1);
+
+        const int srcVal = src(y, x);
+
+        float res = 0;
+
+        res += lut(ty1 * tilesX + tx1, srcVal) * ((1.0f - xa) * (1.0f - ya));
+        res += lut(ty1 * tilesX + tx2, srcVal) * ((xa) * (1.0f - ya));
+        res += lut(ty2 * tilesX + tx1, srcVal) * ((1.0f - xa) * (ya));
+        res += lut(ty2 * tilesX + tx2, srcVal) * ((xa) * (ya));
+
+        dst(y, x) = saturate_cast<uchar>(res);
+    }
+
+    void transform(PtrStepSzb src, PtrStepSzb dst, PtrStepb lut, int tilesX, int tilesY, int2 tileSize, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, block.x), divUp(src.rows, block.y));
+
+        cudaSafeCall( cudaFuncSetCacheConfig(tranformKernel, cudaFuncCachePreferL1) );
+
+        tranformKernel<<<grid, block, 0, stream>>>(src, dst, lut, tileSize, tilesX, tilesY);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpu/src/cuda/color.cu b/modules/gpuimgproc/src/cuda/color.cu
similarity index 98%
rename from modules/gpu/src/cuda/color.cu
rename to modules/gpuimgproc/src/cuda/color.cu
index 5d8f6cbbb5..1a5d4865ed 100644
--- a/modules/gpu/src/cuda/color.cu
+++ b/modules/gpuimgproc/src/cuda/color.cu
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/transform.hpp"
-#include "opencv2/gpu/device/color.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/color.hpp"
 #include "cvt_color_internal.h"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     OPENCV_GPU_TRANSFORM_FUNCTOR_TRAITS(bgra_to_rgba_traits<uchar>::functor_type)
     {
@@ -229,7 +229,7 @@ namespace cv { namespace gpu { namespace device
         traits::functor_type functor = traits::create_functor(); \
         typedef typename traits::functor_type::argument_type src_t; \
         typedef typename traits::functor_type::result_type   dst_t; \
-        cv::gpu::device::transform((PtrStepSz<src_t>)src, (PtrStepSz<dst_t>)dst, functor, WithOutMask(), stream); \
+        cv::gpu::cudev::transform((PtrStepSz<src_t>)src, (PtrStepSz<dst_t>)dst, functor, WithOutMask(), stream); \
     }
 
 #define OPENCV_GPU_IMPLEMENT_CVTCOLOR_ONE(name) \
@@ -456,6 +456,6 @@ namespace cv { namespace gpu { namespace device
     #undef OPENCV_GPU_IMPLEMENT_CVTCOLOR_ALL
     #undef OPENCV_GPU_IMPLEMENT_CVTCOLOR_8U32F
     #undef OPENCV_GPU_IMPLEMENT_CVTCOLOR_8U32F_FULL
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/corners.cu b/modules/gpuimgproc/src/cuda/corners.cu
new file mode 100644
index 0000000000..aa65ac8f8b
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/corners.cu
@@ -0,0 +1,280 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        /////////////////////////////////////////// Corner Harris /////////////////////////////////////////////////
+
+        texture<float, cudaTextureType2D, cudaReadModeElementType> harrisDxTex(0, cudaFilterModePoint, cudaAddressModeClamp);
+        texture<float, cudaTextureType2D, cudaReadModeElementType> harrisDyTex(0, cudaFilterModePoint, cudaAddressModeClamp);
+
+        __global__ void cornerHarris_kernel(const int block_size, const float k, PtrStepSzf dst)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x < dst.cols && y < dst.rows)
+            {
+                float a = 0.f;
+                float b = 0.f;
+                float c = 0.f;
+
+                const int ibegin = y - (block_size / 2);
+                const int jbegin = x - (block_size / 2);
+                const int iend = ibegin + block_size;
+                const int jend = jbegin + block_size;
+
+                for (int i = ibegin; i < iend; ++i)
+                {
+                    for (int j = jbegin; j < jend; ++j)
+                    {
+                        float dx = tex2D(harrisDxTex, j, i);
+                        float dy = tex2D(harrisDyTex, j, i);
+
+                        a += dx * dx;
+                        b += dx * dy;
+                        c += dy * dy;
+                    }
+                }
+
+                dst(y, x) = a * c - b * b - k * (a + c) * (a + c);
+            }
+        }
+
+        template <typename BR, typename BC>
+        __global__ void cornerHarris_kernel(const int block_size, const float k, PtrStepSzf dst, const BR border_row, const BC border_col)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x < dst.cols && y < dst.rows)
+            {
+                float a = 0.f;
+                float b = 0.f;
+                float c = 0.f;
+
+                const int ibegin = y - (block_size / 2);
+                const int jbegin = x - (block_size / 2);
+                const int iend = ibegin + block_size;
+                const int jend = jbegin + block_size;
+
+                for (int i = ibegin; i < iend; ++i)
+                {
+                    const int y = border_col.idx_row(i);
+
+                    for (int j = jbegin; j < jend; ++j)
+                    {
+                        const int x = border_row.idx_col(j);
+
+                        float dx = tex2D(harrisDxTex, x, y);
+                        float dy = tex2D(harrisDyTex, x, y);
+
+                        a += dx * dx;
+                        b += dx * dy;
+                        c += dy * dy;
+                    }
+                }
+
+                dst(y, x) = a * c - b * b - k * (a + c) * (a + c);
+            }
+        }
+
+        void cornerHarris_gpu(int block_size, float k, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream)
+        {
+            dim3 block(32, 8);
+            dim3 grid(divUp(Dx.cols, block.x), divUp(Dx.rows, block.y));
+
+            bindTexture(&harrisDxTex, Dx);
+            bindTexture(&harrisDyTex, Dy);
+
+            switch (border_type)
+            {
+            case BORDER_REFLECT101:
+                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst, BrdRowReflect101<void>(Dx.cols), BrdColReflect101<void>(Dx.rows));
+                break;
+
+            case BORDER_REFLECT:
+                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst, BrdRowReflect<void>(Dx.cols), BrdColReflect<void>(Dx.rows));
+                break;
+
+            case BORDER_REPLICATE:
+                cornerHarris_kernel<<<grid, block, 0, stream>>>(block_size, k, dst);
+                break;
+            }
+
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        /////////////////////////////////////////// Corner Min Eigen Val /////////////////////////////////////////////////
+
+        texture<float, cudaTextureType2D, cudaReadModeElementType> minEigenValDxTex(0, cudaFilterModePoint, cudaAddressModeClamp);
+        texture<float, cudaTextureType2D, cudaReadModeElementType> minEigenValDyTex(0, cudaFilterModePoint, cudaAddressModeClamp);
+
+        __global__ void cornerMinEigenVal_kernel(const int block_size, PtrStepSzf dst)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x < dst.cols && y < dst.rows)
+            {
+                float a = 0.f;
+                float b = 0.f;
+                float c = 0.f;
+
+                const int ibegin = y - (block_size / 2);
+                const int jbegin = x - (block_size / 2);
+                const int iend = ibegin + block_size;
+                const int jend = jbegin + block_size;
+
+                for (int i = ibegin; i < iend; ++i)
+                {
+                    for (int j = jbegin; j < jend; ++j)
+                    {
+                        float dx = tex2D(minEigenValDxTex, j, i);
+                        float dy = tex2D(minEigenValDyTex, j, i);
+
+                        a += dx * dx;
+                        b += dx * dy;
+                        c += dy * dy;
+                    }
+                }
+
+                a *= 0.5f;
+                c *= 0.5f;
+
+                dst(y, x) = (a + c) - sqrtf((a - c) * (a - c) + b * b);
+            }
+        }
+
+
+        template <typename BR, typename BC>
+        __global__ void cornerMinEigenVal_kernel(const int block_size, PtrStepSzf dst, const BR border_row, const BC border_col)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x < dst.cols && y < dst.rows)
+            {
+                float a = 0.f;
+                float b = 0.f;
+                float c = 0.f;
+
+                const int ibegin = y - (block_size / 2);
+                const int jbegin = x - (block_size / 2);
+                const int iend = ibegin + block_size;
+                const int jend = jbegin + block_size;
+
+                for (int i = ibegin; i < iend; ++i)
+                {
+                    int y = border_col.idx_row(i);
+
+                    for (int j = jbegin; j < jend; ++j)
+                    {
+                        int x = border_row.idx_col(j);
+
+                        float dx = tex2D(minEigenValDxTex, x, y);
+                        float dy = tex2D(minEigenValDyTex, x, y);
+
+                        a += dx * dx;
+                        b += dx * dy;
+                        c += dy * dy;
+                    }
+                }
+
+                a *= 0.5f;
+                c *= 0.5f;
+
+                dst(y, x) = (a + c) - sqrtf((a - c) * (a - c) + b * b);
+            }
+        }
+
+        void cornerMinEigenVal_gpu(int block_size, PtrStepSzf Dx, PtrStepSzf Dy, PtrStepSzf dst, int border_type, cudaStream_t stream)
+        {
+            dim3 block(32, 8);
+            dim3 grid(divUp(Dx.cols, block.x), divUp(Dx.rows, block.y));
+
+            bindTexture(&minEigenValDxTex, Dx);
+            bindTexture(&minEigenValDyTex, Dy);
+
+            switch (border_type)
+            {
+            case BORDER_REFLECT101:
+                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst, BrdRowReflect101<void>(Dx.cols), BrdColReflect101<void>(Dx.rows));
+                break;
+
+            case BORDER_REFLECT:
+                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst, BrdRowReflect<void>(Dx.cols), BrdColReflect<void>(Dx.rows));
+                break;
+
+            case BORDER_REPLICATE:
+                cornerMinEigenVal_kernel<<<grid, block, 0, stream>>>(block_size, dst);
+                break;
+            }
+
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall(cudaDeviceSynchronize());
+        }
+    }
+}}}
+
+#endif // HAVE_OPENCV_GPUFILTERS
+
+#endif // CUDA_DISABLER
diff --git a/modules/gpuimgproc/src/cuda/debayer.cu b/modules/gpuimgproc/src/cuda/debayer.cu
new file mode 100644
index 0000000000..46a1c14ef4
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/debayer.cu
@@ -0,0 +1,544 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/color.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T> struct Bayer2BGR;
+
+    template <> struct Bayer2BGR<uchar>
+    {
+        uchar3 res0;
+        uchar3 res1;
+        uchar3 res2;
+        uchar3 res3;
+
+        __device__ void apply(const PtrStepSzb& src, int s_x, int s_y, bool blue_last, bool start_with_green)
+        {
+            uchar4 patch[3][3];
+            patch[0][1] = ((const uchar4*) src.ptr(s_y - 1))[s_x];
+            patch[0][0] = ((const uchar4*) src.ptr(s_y - 1))[::max(s_x - 1, 0)];
+            patch[0][2] = ((const uchar4*) src.ptr(s_y - 1))[::min(s_x + 1, ((src.cols + 3) >> 2) - 1)];
+
+            patch[1][1] = ((const uchar4*) src.ptr(s_y))[s_x];
+            patch[1][0] = ((const uchar4*) src.ptr(s_y))[::max(s_x - 1, 0)];
+            patch[1][2] = ((const uchar4*) src.ptr(s_y))[::min(s_x + 1, ((src.cols + 3) >> 2) - 1)];
+
+            patch[2][1] = ((const uchar4*) src.ptr(s_y + 1))[s_x];
+            patch[2][0] = ((const uchar4*) src.ptr(s_y + 1))[::max(s_x - 1, 0)];
+            patch[2][2] = ((const uchar4*) src.ptr(s_y + 1))[::min(s_x + 1, ((src.cols + 3) >> 2) - 1)];
+
+            if ((s_y & 1) ^ start_with_green)
+            {
+                const int t0 = (patch[0][1].x + patch[2][1].x + 1) >> 1;
+                const int t1 = (patch[1][0].w + patch[1][1].y + 1) >> 1;
+
+                const int t2 = (patch[0][1].x + patch[0][1].z + patch[2][1].x + patch[2][1].z + 2) >> 2;
+                const int t3 = (patch[0][1].y + patch[1][1].x + patch[1][1].z + patch[2][1].y + 2) >> 2;
+
+                const int t4 = (patch[0][1].z + patch[2][1].z + 1) >> 1;
+                const int t5 = (patch[1][1].y + patch[1][1].w + 1) >> 1;
+
+                const int t6 = (patch[0][1].z + patch[0][2].x + patch[2][1].z + patch[2][2].x + 2) >> 2;
+                const int t7 = (patch[0][1].w + patch[1][1].z + patch[1][2].x + patch[2][1].w + 2) >> 2;
+
+                if ((s_y & 1) ^ blue_last)
+                {
+                    res0.x = t1;
+                    res0.y = patch[1][1].x;
+                    res0.z = t0;
+
+                    res1.x = patch[1][1].y;
+                    res1.y = t3;
+                    res1.z = t2;
+
+                    res2.x = t5;
+                    res2.y = patch[1][1].z;
+                    res2.z = t4;
+
+                    res3.x = patch[1][1].w;
+                    res3.y = t7;
+                    res3.z = t6;
+                }
+                else
+                {
+                    res0.x = t0;
+                    res0.y = patch[1][1].x;
+                    res0.z = t1;
+
+                    res1.x = t2;
+                    res1.y = t3;
+                    res1.z = patch[1][1].y;
+
+                    res2.x = t4;
+                    res2.y = patch[1][1].z;
+                    res2.z = t5;
+
+                    res3.x = t6;
+                    res3.y = t7;
+                    res3.z = patch[1][1].w;
+                }
+            }
+            else
+            {
+                const int t0 = (patch[0][0].w + patch[0][1].y + patch[2][0].w + patch[2][1].y + 2) >> 2;
+                const int t1 = (patch[0][1].x + patch[1][0].w + patch[1][1].y + patch[2][1].x + 2) >> 2;
+
+                const int t2 = (patch[0][1].y + patch[2][1].y + 1) >> 1;
+                const int t3 = (patch[1][1].x + patch[1][1].z + 1) >> 1;
+
+                const int t4 = (patch[0][1].y + patch[0][1].w + patch[2][1].y + patch[2][1].w + 2) >> 2;
+                const int t5 = (patch[0][1].z + patch[1][1].y + patch[1][1].w + patch[2][1].z + 2) >> 2;
+
+                const int t6 = (patch[0][1].w + patch[2][1].w + 1) >> 1;
+                const int t7 = (patch[1][1].z + patch[1][2].x + 1) >> 1;
+
+                if ((s_y & 1) ^ blue_last)
+                {
+                    res0.x = patch[1][1].x;
+                    res0.y = t1;
+                    res0.z = t0;
+
+                    res1.x = t3;
+                    res1.y = patch[1][1].y;
+                    res1.z = t2;
+
+                    res2.x = patch[1][1].z;
+                    res2.y = t5;
+                    res2.z = t4;
+
+                    res3.x = t7;
+                    res3.y = patch[1][1].w;
+                    res3.z = t6;
+                }
+                else
+                {
+                    res0.x = t0;
+                    res0.y = t1;
+                    res0.z = patch[1][1].x;
+
+                    res1.x = t2;
+                    res1.y = patch[1][1].y;
+                    res1.z = t3;
+
+                    res2.x = t4;
+                    res2.y = t5;
+                    res2.z = patch[1][1].z;
+
+                    res3.x = t6;
+                    res3.y = patch[1][1].w;
+                    res3.z = t7;
+                }
+            }
+        }
+    };
+
+    template <typename D> __device__ __forceinline__ D toDst(const uchar3& pix);
+    template <> __device__ __forceinline__ uchar toDst<uchar>(const uchar3& pix)
+    {
+        typename bgr_to_gray_traits<uchar>::functor_type f = bgr_to_gray_traits<uchar>::create_functor();
+        return f(pix);
+    }
+    template <> __device__ __forceinline__ uchar3 toDst<uchar3>(const uchar3& pix)
+    {
+        return pix;
+    }
+    template <> __device__ __forceinline__ uchar4 toDst<uchar4>(const uchar3& pix)
+    {
+        return make_uchar4(pix.x, pix.y, pix.z, 255);
+    }
+
+    template <typename D>
+    __global__ void Bayer2BGR_8u(const PtrStepSzb src, PtrStep<D> dst, const bool blue_last, const bool start_with_green)
+    {
+        const int s_x = blockIdx.x * blockDim.x + threadIdx.x;
+        int s_y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (s_y >= src.rows || (s_x << 2) >= src.cols)
+            return;
+
+        s_y = ::min(::max(s_y, 1), src.rows - 2);
+
+        Bayer2BGR<uchar> bayer;
+        bayer.apply(src, s_x, s_y, blue_last, start_with_green);
+
+        const int d_x = (blockIdx.x * blockDim.x + threadIdx.x) << 2;
+        const int d_y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        dst(d_y, d_x) = toDst<D>(bayer.res0);
+        if (d_x + 1 < src.cols)
+            dst(d_y, d_x + 1) = toDst<D>(bayer.res1);
+        if (d_x + 2 < src.cols)
+            dst(d_y, d_x + 2) = toDst<D>(bayer.res2);
+        if (d_x + 3 < src.cols)
+            dst(d_y, d_x + 3) = toDst<D>(bayer.res3);
+    }
+
+    template <> struct Bayer2BGR<ushort>
+    {
+        ushort3 res0;
+        ushort3 res1;
+
+        __device__ void apply(const PtrStepSzb& src, int s_x, int s_y, bool blue_last, bool start_with_green)
+        {
+            ushort2 patch[3][3];
+            patch[0][1] = ((const ushort2*) src.ptr(s_y - 1))[s_x];
+            patch[0][0] = ((const ushort2*) src.ptr(s_y - 1))[::max(s_x - 1, 0)];
+            patch[0][2] = ((const ushort2*) src.ptr(s_y - 1))[::min(s_x + 1, ((src.cols + 1) >> 1) - 1)];
+
+            patch[1][1] = ((const ushort2*) src.ptr(s_y))[s_x];
+            patch[1][0] = ((const ushort2*) src.ptr(s_y))[::max(s_x - 1, 0)];
+            patch[1][2] = ((const ushort2*) src.ptr(s_y))[::min(s_x + 1, ((src.cols + 1) >> 1) - 1)];
+
+            patch[2][1] = ((const ushort2*) src.ptr(s_y + 1))[s_x];
+            patch[2][0] = ((const ushort2*) src.ptr(s_y + 1))[::max(s_x - 1, 0)];
+            patch[2][2] = ((const ushort2*) src.ptr(s_y + 1))[::min(s_x + 1, ((src.cols + 1) >> 1) - 1)];
+
+            if ((s_y & 1) ^ start_with_green)
+            {
+                const int t0 = (patch[0][1].x + patch[2][1].x + 1) >> 1;
+                const int t1 = (patch[1][0].y + patch[1][1].y + 1) >> 1;
+
+                const int t2 = (patch[0][1].x + patch[0][2].x + patch[2][1].x + patch[2][2].x + 2) >> 2;
+                const int t3 = (patch[0][1].y + patch[1][1].x + patch[1][2].x + patch[2][1].y + 2) >> 2;
+
+                if ((s_y & 1) ^ blue_last)
+                {
+                    res0.x = t1;
+                    res0.y = patch[1][1].x;
+                    res0.z = t0;
+
+                    res1.x = patch[1][1].y;
+                    res1.y = t3;
+                    res1.z = t2;
+                }
+                else
+                {
+                    res0.x = t0;
+                    res0.y = patch[1][1].x;
+                    res0.z = t1;
+
+                    res1.x = t2;
+                    res1.y = t3;
+                    res1.z = patch[1][1].y;
+                }
+            }
+            else
+            {
+                const int t0 = (patch[0][0].y + patch[0][1].y + patch[2][0].y + patch[2][1].y + 2) >> 2;
+                const int t1 = (patch[0][1].x + patch[1][0].y + patch[1][1].y + patch[2][1].x + 2) >> 2;
+
+                const int t2 = (patch[0][1].y + patch[2][1].y + 1) >> 1;
+                const int t3 = (patch[1][1].x + patch[1][2].x + 1) >> 1;
+
+                if ((s_y & 1) ^ blue_last)
+                {
+                    res0.x = patch[1][1].x;
+                    res0.y = t1;
+                    res0.z = t0;
+
+                    res1.x = t3;
+                    res1.y = patch[1][1].y;
+                    res1.z = t2;
+                }
+                else
+                {
+                    res0.x = t0;
+                    res0.y = t1;
+                    res0.z = patch[1][1].x;
+
+                    res1.x = t2;
+                    res1.y = patch[1][1].y;
+                    res1.z = t3;
+                }
+            }
+        }
+    };
+
+    template <typename D> __device__ __forceinline__ D toDst(const ushort3& pix);
+    template <> __device__ __forceinline__ ushort toDst<ushort>(const ushort3& pix)
+    {
+        typename bgr_to_gray_traits<ushort>::functor_type f = bgr_to_gray_traits<ushort>::create_functor();
+        return f(pix);
+    }
+    template <> __device__ __forceinline__ ushort3 toDst<ushort3>(const ushort3& pix)
+    {
+        return pix;
+    }
+    template <> __device__ __forceinline__ ushort4 toDst<ushort4>(const ushort3& pix)
+    {
+        return make_ushort4(pix.x, pix.y, pix.z, numeric_limits<ushort>::max());
+    }
+
+    template <typename D>
+    __global__ void Bayer2BGR_16u(const PtrStepSzb src, PtrStep<D> dst, const bool blue_last, const bool start_with_green)
+    {
+        const int s_x = blockIdx.x * blockDim.x + threadIdx.x;
+        int s_y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (s_y >= src.rows || (s_x << 1) >= src.cols)
+            return;
+
+        s_y = ::min(::max(s_y, 1), src.rows - 2);
+
+        Bayer2BGR<ushort> bayer;
+        bayer.apply(src, s_x, s_y, blue_last, start_with_green);
+
+        const int d_x = (blockIdx.x * blockDim.x + threadIdx.x) << 1;
+        const int d_y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        dst(d_y, d_x) = toDst<D>(bayer.res0);
+        if (d_x + 1 < src.cols)
+            dst(d_y, d_x + 1) = toDst<D>(bayer.res1);
+    }
+
+    template <int cn>
+    void Bayer2BGR_8u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream)
+    {
+        typedef typename TypeVec<uchar, cn>::vec_type dst_t;
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, 4 * block.x), divUp(src.rows, block.y));
+
+        cudaSafeCall( cudaFuncSetCacheConfig(Bayer2BGR_8u<dst_t>, cudaFuncCachePreferL1) );
+
+        Bayer2BGR_8u<dst_t><<<grid, block, 0, stream>>>(src, (PtrStepSz<dst_t>)dst, blue_last, start_with_green);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <int cn>
+    void Bayer2BGR_16u_gpu(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream)
+    {
+        typedef typename TypeVec<ushort, cn>::vec_type dst_t;
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, 2 * block.x), divUp(src.rows, block.y));
+
+        cudaSafeCall( cudaFuncSetCacheConfig(Bayer2BGR_16u<dst_t>, cudaFuncCachePreferL1) );
+
+        Bayer2BGR_16u<dst_t><<<grid, block, 0, stream>>>(src, (PtrStepSz<dst_t>)dst, blue_last, start_with_green);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void Bayer2BGR_8u_gpu<1>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+    template void Bayer2BGR_8u_gpu<3>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+    template void Bayer2BGR_8u_gpu<4>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+
+    template void Bayer2BGR_16u_gpu<1>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+    template void Bayer2BGR_16u_gpu<3>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+    template void Bayer2BGR_16u_gpu<4>(PtrStepSzb src, PtrStepSzb dst, bool blue_last, bool start_with_green, cudaStream_t stream);
+
+    //////////////////////////////////////////////////////////////
+    // Bayer Demosaicing (Malvar, He, and Cutler)
+    //
+    // by Morgan McGuire, Williams College
+    // http://graphics.cs.williams.edu/papers/BayerJGT09/#shaders
+    //
+    // ported to CUDA
+
+    texture<uchar, cudaTextureType2D, cudaReadModeElementType> sourceTex(false, cudaFilterModePoint, cudaAddressModeClamp);
+
+    template <typename DstType>
+    __global__ void MHCdemosaic(PtrStepSz<DstType> dst, const int2 sourceOffset, const int2 firstRed)
+    {
+        const float   kAx = -1.0f / 8.0f,     kAy = -1.5f / 8.0f,     kAz =  0.5f / 8.0f    /*kAw = -1.0f / 8.0f*/;
+        const float   kBx =  2.0f / 8.0f,   /*kBy =  0.0f / 8.0f,*/ /*kBz =  0.0f / 8.0f,*/   kBw =  4.0f / 8.0f  ;
+        const float   kCx =  4.0f / 8.0f,     kCy =  6.0f / 8.0f,     kCz =  5.0f / 8.0f    /*kCw =  5.0f / 8.0f*/;
+        const float /*kDx =  0.0f / 8.0f,*/   kDy =  2.0f / 8.0f,     kDz = -1.0f / 8.0f    /*kDw = -1.0f / 8.0f*/;
+        const float   kEx = -1.0f / 8.0f,     kEy = -1.5f / 8.0f,   /*kEz = -1.0f / 8.0f,*/   kEw =  0.5f / 8.0f  ;
+        const float   kFx =  2.0f / 8.0f,   /*kFy =  0.0f / 8.0f,*/   kFz =  4.0f / 8.0f    /*kFw =  0.0f / 8.0f*/;
+
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x == 0 || x >= dst.cols - 1 || y == 0 || y >= dst.rows - 1)
+            return;
+
+        int2 center;
+        center.x = x + sourceOffset.x;
+        center.y = y + sourceOffset.y;
+
+        int4 xCoord;
+        xCoord.x = center.x - 2;
+        xCoord.y = center.x - 1;
+        xCoord.z = center.x + 1;
+        xCoord.w = center.x + 2;
+
+        int4 yCoord;
+        yCoord.x = center.y - 2;
+        yCoord.y = center.y - 1;
+        yCoord.z = center.y + 1;
+        yCoord.w = center.y + 2;
+
+        float C = tex2D(sourceTex, center.x, center.y); // ( 0, 0)
+
+        float4 Dvec;
+        Dvec.x = tex2D(sourceTex, xCoord.y, yCoord.y); // (-1,-1)
+        Dvec.y = tex2D(sourceTex, xCoord.y, yCoord.z); // (-1, 1)
+        Dvec.z = tex2D(sourceTex, xCoord.z, yCoord.y); // ( 1,-1)
+        Dvec.w = tex2D(sourceTex, xCoord.z, yCoord.z); // ( 1, 1)
+
+        float4 value;
+        value.x = tex2D(sourceTex, center.x, yCoord.x); // ( 0,-2) A0
+        value.y = tex2D(sourceTex, center.x, yCoord.y); // ( 0,-1) B0
+        value.z = tex2D(sourceTex, xCoord.x, center.y); // (-2, 0) E0
+        value.w = tex2D(sourceTex, xCoord.y, center.y); // (-1, 0) F0
+
+        // (A0 + A1), (B0 + B1), (E0 + E1), (F0 + F1)
+        value.x += tex2D(sourceTex, center.x, yCoord.w); // ( 0, 2) A1
+        value.y += tex2D(sourceTex, center.x, yCoord.z); // ( 0, 1) B1
+        value.z += tex2D(sourceTex, xCoord.w, center.y); // ( 2, 0) E1
+        value.w += tex2D(sourceTex, xCoord.z, center.y); // ( 1, 0) F1
+
+        float4 PATTERN;
+        PATTERN.x = kCx * C;
+        PATTERN.y = kCy * C;
+        PATTERN.z = kCz * C;
+        PATTERN.w = PATTERN.z;
+
+        float D = Dvec.x + Dvec.y + Dvec.z + Dvec.w;
+
+        // There are five filter patterns (identity, cross, checker,
+        // theta, phi). Precompute the terms from all of them and then
+        // use swizzles to assign to color channels.
+        //
+        // Channel Matches
+        // x cross (e.g., EE G)
+        // y checker (e.g., EE B)
+        // z theta (e.g., EO R)
+        // w phi (e.g., EO B)
+
+        #define A value.x  // A0 + A1
+        #define B value.y  // B0 + B1
+        #define E value.z  // E0 + E1
+        #define F value.w  // F0 + F1
+
+        float3 temp;
+
+        // PATTERN.yzw += (kD.yz * D).xyy;
+        temp.x = kDy * D;
+        temp.y = kDz * D;
+        PATTERN.y += temp.x;
+        PATTERN.z += temp.y;
+        PATTERN.w += temp.y;
+
+        // PATTERN += (kA.xyz * A).xyzx;
+        temp.x = kAx * A;
+        temp.y = kAy * A;
+        temp.z = kAz * A;
+        PATTERN.x += temp.x;
+        PATTERN.y += temp.y;
+        PATTERN.z += temp.z;
+        PATTERN.w += temp.x;
+
+        // PATTERN += (kE.xyw * E).xyxz;
+        temp.x = kEx * E;
+        temp.y = kEy * E;
+        temp.z = kEw * E;
+        PATTERN.x += temp.x;
+        PATTERN.y += temp.y;
+        PATTERN.z += temp.x;
+        PATTERN.w += temp.z;
+
+        // PATTERN.xw += kB.xw * B;
+        PATTERN.x += kBx * B;
+        PATTERN.w += kBw * B;
+
+        // PATTERN.xz += kF.xz * F;
+        PATTERN.x += kFx * F;
+        PATTERN.z += kFz * F;
+
+        // Determine which of four types of pixels we are on.
+        int2 alternate;
+        alternate.x = (x + firstRed.x) % 2;
+        alternate.y = (y + firstRed.y) % 2;
+
+        // in BGR sequence;
+        uchar3 pixelColor =
+            (alternate.y == 0) ?
+                ((alternate.x == 0) ?
+                    make_uchar3(saturate_cast<uchar>(PATTERN.y), saturate_cast<uchar>(PATTERN.x), saturate_cast<uchar>(C)) :
+                    make_uchar3(saturate_cast<uchar>(PATTERN.w), saturate_cast<uchar>(C), saturate_cast<uchar>(PATTERN.z))) :
+                ((alternate.x == 0) ?
+                    make_uchar3(saturate_cast<uchar>(PATTERN.z), saturate_cast<uchar>(C), saturate_cast<uchar>(PATTERN.w)) :
+                    make_uchar3(saturate_cast<uchar>(C), saturate_cast<uchar>(PATTERN.x), saturate_cast<uchar>(PATTERN.y)));
+
+        dst(y, x) = toDst<DstType>(pixelColor);
+    }
+
+    template <int cn>
+    void MHCdemosaic(PtrStepSzb src, int2 sourceOffset, PtrStepSzb dst, int2 firstRed, cudaStream_t stream)
+    {
+        typedef typename TypeVec<uchar, cn>::vec_type dst_t;
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, block.x), divUp(src.rows, block.y));
+
+        bindTexture(&sourceTex, src);
+
+        MHCdemosaic<dst_t><<<grid, block, 0, stream>>>((PtrStepSz<dst_t>)dst, sourceOffset, firstRed);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void MHCdemosaic<1>(PtrStepSzb src, int2 sourceOffset, PtrStepSzb dst, int2 firstRed, cudaStream_t stream);
+    template void MHCdemosaic<3>(PtrStepSzb src, int2 sourceOffset, PtrStepSzb dst, int2 firstRed, cudaStream_t stream);
+    template void MHCdemosaic<4>(PtrStepSzb src, int2 sourceOffset, PtrStepSzb dst, int2 firstRed, cudaStream_t stream);
+}}}
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/generalized_hough.cu b/modules/gpuimgproc/src/cuda/generalized_hough.cu
new file mode 100644
index 0000000000..fdf691ff4a
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/generalized_hough.cu
@@ -0,0 +1,824 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include <thrust/device_ptr.h>
+#include <thrust/transform.h>
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace ght
+    {
+        __device__ int g_counter;
+
+        template <typename T, int PIXELS_PER_THREAD>
+        __global__ void buildEdgePointList(const PtrStepSzb edges, const PtrStep<T> dx, const PtrStep<T> dy,
+                                           unsigned int* coordList, float* thetaList)
+        {
+            __shared__ unsigned int s_coordLists[4][32 * PIXELS_PER_THREAD];
+            __shared__ float s_thetaLists[4][32 * PIXELS_PER_THREAD];
+            __shared__ int s_sizes[4];
+            __shared__ int s_globStart[4];
+
+            const int x = blockIdx.x * blockDim.x * PIXELS_PER_THREAD + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (threadIdx.x == 0)
+                s_sizes[threadIdx.y] = 0;
+            __syncthreads();
+
+            if (y < edges.rows)
+            {
+                // fill the queue
+                const uchar* edgesRow = edges.ptr(y);
+                const T* dxRow = dx.ptr(y);
+                const T* dyRow = dy.ptr(y);
+
+                for (int i = 0, xx = x; i < PIXELS_PER_THREAD && xx < edges.cols; ++i, xx += blockDim.x)
+                {
+                    const T dxVal = dxRow[xx];
+                    const T dyVal = dyRow[xx];
+
+                    if (edgesRow[xx] && (dxVal != 0 || dyVal != 0))
+                    {
+                        const unsigned int coord = (y << 16) | xx;
+
+                        float theta = ::atan2f(dyVal, dxVal);
+                        if (theta < 0)
+                            theta += 2.0f * CV_PI_F;
+
+                        const int qidx = Emulation::smem::atomicAdd(&s_sizes[threadIdx.y], 1);
+
+                        s_coordLists[threadIdx.y][qidx] = coord;
+                        s_thetaLists[threadIdx.y][qidx] = theta;
+                    }
+                }
+            }
+
+            __syncthreads();
+
+            // let one thread reserve the space required in the global list
+            if (threadIdx.x == 0 && threadIdx.y == 0)
+            {
+                // find how many items are stored in each list
+                int totalSize = 0;
+                for (int i = 0; i < blockDim.y; ++i)
+                {
+                    s_globStart[i] = totalSize;
+                    totalSize += s_sizes[i];
+                }
+
+                // calculate the offset in the global list
+                const int globalOffset = atomicAdd(&g_counter, totalSize);
+                for (int i = 0; i < blockDim.y; ++i)
+                    s_globStart[i] += globalOffset;
+            }
+
+            __syncthreads();
+
+            // copy local queues to global queue
+            const int qsize = s_sizes[threadIdx.y];
+            int gidx = s_globStart[threadIdx.y] + threadIdx.x;
+            for(int i = threadIdx.x; i < qsize; i += blockDim.x, gidx += blockDim.x)
+            {
+                coordList[gidx] = s_coordLists[threadIdx.y][i];
+                thetaList[gidx] = s_thetaLists[threadIdx.y][i];
+            }
+        }
+
+        template <typename T>
+        int buildEdgePointList_gpu(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList)
+        {
+            const int PIXELS_PER_THREAD = 8;
+
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 4);
+            const dim3 grid(divUp(edges.cols, block.x * PIXELS_PER_THREAD), divUp(edges.rows, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(buildEdgePointList<T, PIXELS_PER_THREAD>, cudaFuncCachePreferShared) );
+
+            buildEdgePointList<T, PIXELS_PER_THREAD><<<grid, block>>>(edges, (PtrStepSz<T>) dx, (PtrStepSz<T>) dy, coordList, thetaList);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            return totalCount;
+        }
+
+        template int buildEdgePointList_gpu<short>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
+        template int buildEdgePointList_gpu<int>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
+        template int buildEdgePointList_gpu<float>(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
+
+        __global__ void buildRTable(const unsigned int* coordList, const float* thetaList, const int pointsCount,
+                                    PtrStep<short2> r_table, int* r_sizes, int maxSize,
+                                    const short2 templCenter, const float thetaScale)
+        {
+            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
+
+            if (tid >= pointsCount)
+                return;
+
+            const unsigned int coord = coordList[tid];
+            short2 p;
+            p.x = (coord & 0xFFFF);
+            p.y = (coord >> 16) & 0xFFFF;
+
+            const float theta = thetaList[tid];
+            const int n = __float2int_rn(theta * thetaScale);
+
+            const int ind = ::atomicAdd(r_sizes + n, 1);
+            if (ind < maxSize)
+                r_table(n, ind) = saturate_cast<short2>(p - templCenter);
+        }
+
+        void buildRTable_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                             PtrStepSz<short2> r_table, int* r_sizes,
+                             short2 templCenter, int levels)
+        {
+            const dim3 block(256);
+            const dim3 grid(divUp(pointsCount, block.x));
+
+            const float thetaScale = levels / (2.0f * CV_PI_F);
+
+            buildRTable<<<grid, block>>>(coordList, thetaList, pointsCount, r_table, r_sizes, r_table.cols, templCenter, thetaScale);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        ////////////////////////////////////////////////////////////////////////
+        // Ballard_Pos
+
+        __global__ void Ballard_Pos_calcHist(const unsigned int* coordList, const float* thetaList, const int pointsCount,
+                                             const PtrStep<short2> r_table, const int* r_sizes,
+                                             PtrStepSzi hist,
+                                             const float idp, const float thetaScale)
+        {
+            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
+
+            if (tid >= pointsCount)
+                return;
+
+            const unsigned int coord = coordList[tid];
+            short2 p;
+            p.x = (coord & 0xFFFF);
+            p.y = (coord >> 16) & 0xFFFF;
+
+            const float theta = thetaList[tid];
+            const int n = __float2int_rn(theta * thetaScale);
+
+            const short2* r_row = r_table.ptr(n);
+            const int r_row_size = r_sizes[n];
+
+            for (int j = 0; j < r_row_size; ++j)
+            {
+                short2 c = saturate_cast<short2>(p - r_row[j]);
+
+                c.x = __float2int_rn(c.x * idp);
+                c.y = __float2int_rn(c.y * idp);
+
+                if (c.x >= 0 && c.x < hist.cols - 2 && c.y >= 0 && c.y < hist.rows - 2)
+                    ::atomicAdd(hist.ptr(c.y + 1) + c.x + 1, 1);
+            }
+        }
+
+        void Ballard_Pos_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                      PtrStepSz<short2> r_table, const int* r_sizes,
+                                      PtrStepSzi hist,
+                                      float dp, int levels)
+        {
+            const dim3 block(256);
+            const dim3 grid(divUp(pointsCount, block.x));
+
+            const float idp = 1.0f / dp;
+            const float thetaScale = levels / (2.0f * CV_PI_F);
+
+            Ballard_Pos_calcHist<<<grid, block>>>(coordList, thetaList, pointsCount, r_table, r_sizes, hist, idp, thetaScale);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        __global__ void Ballard_Pos_findPosInHist(const PtrStepSzi hist, float4* out, int3* votes,
+                                                  const int maxSize, const float dp, const int threshold)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x >= hist.cols - 2 || y >= hist.rows - 2)
+                return;
+
+            const int curVotes = hist(y + 1, x + 1);
+
+            if (curVotes > threshold &&
+                curVotes >  hist(y + 1, x) &&
+                curVotes >= hist(y + 1, x + 2) &&
+                curVotes >  hist(y, x + 1) &&
+                curVotes >= hist(y + 2, x + 1))
+            {
+                const int ind = ::atomicAdd(&g_counter, 1);
+
+                if (ind < maxSize)
+                {
+                    out[ind] = make_float4(x * dp, y * dp, 1.0f, 0.0f);
+                    votes[ind] = make_int3(curVotes, 0, 0);
+                }
+            }
+        }
+
+        int Ballard_Pos_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int maxSize, float dp, int threshold)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(hist.cols - 2, block.x), divUp(hist.rows - 2, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(Ballard_Pos_findPosInHist, cudaFuncCachePreferL1) );
+
+            Ballard_Pos_findPosInHist<<<grid, block>>>(hist, out, votes, maxSize, dp, threshold);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            totalCount = ::min(totalCount, maxSize);
+
+            return totalCount;
+        }
+
+        ////////////////////////////////////////////////////////////////////////
+        // Guil_Full
+
+        struct FeatureTable
+        {
+            uchar* p1_pos_data;
+            size_t p1_pos_step;
+
+            uchar* p1_theta_data;
+            size_t p1_theta_step;
+
+            uchar* p2_pos_data;
+            size_t p2_pos_step;
+
+            uchar* d12_data;
+            size_t d12_step;
+
+            uchar* r1_data;
+            size_t r1_step;
+
+            uchar* r2_data;
+            size_t r2_step;
+        };
+
+        __constant__ FeatureTable c_templFeatures;
+        __constant__ FeatureTable c_imageFeatures;
+
+        void Guil_Full_setTemplFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2)
+        {
+            FeatureTable tbl;
+
+            tbl.p1_pos_data = p1_pos.data;
+            tbl.p1_pos_step = p1_pos.step;
+
+            tbl.p1_theta_data = p1_theta.data;
+            tbl.p1_theta_step = p1_theta.step;
+
+            tbl.p2_pos_data = p2_pos.data;
+            tbl.p2_pos_step = p2_pos.step;
+
+            tbl.d12_data = d12.data;
+            tbl.d12_step = d12.step;
+
+            tbl.r1_data = r1.data;
+            tbl.r1_step = r1.step;
+
+            tbl.r2_data = r2.data;
+            tbl.r2_step = r2.step;
+
+            cudaSafeCall( cudaMemcpyToSymbol(c_templFeatures, &tbl, sizeof(FeatureTable)) );
+        }
+        void Guil_Full_setImageFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2)
+        {
+            FeatureTable tbl;
+
+            tbl.p1_pos_data = p1_pos.data;
+            tbl.p1_pos_step = p1_pos.step;
+
+            tbl.p1_theta_data = p1_theta.data;
+            tbl.p1_theta_step = p1_theta.step;
+
+            tbl.p2_pos_data = p2_pos.data;
+            tbl.p2_pos_step = p2_pos.step;
+
+            tbl.d12_data = d12.data;
+            tbl.d12_step = d12.step;
+
+            tbl.r1_data = r1.data;
+            tbl.r1_step = r1.step;
+
+            tbl.r2_data = r2.data;
+            tbl.r2_step = r2.step;
+
+            cudaSafeCall( cudaMemcpyToSymbol(c_imageFeatures, &tbl, sizeof(FeatureTable)) );
+        }
+
+        struct TemplFeatureTable
+        {
+            static __device__ float2* p1_pos(int n)
+            {
+                return (float2*)(c_templFeatures.p1_pos_data + n * c_templFeatures.p1_pos_step);
+            }
+            static __device__ float* p1_theta(int n)
+            {
+                return (float*)(c_templFeatures.p1_theta_data + n * c_templFeatures.p1_theta_step);
+            }
+            static __device__ float2* p2_pos(int n)
+            {
+                return (float2*)(c_templFeatures.p2_pos_data + n * c_templFeatures.p2_pos_step);
+            }
+
+            static __device__ float* d12(int n)
+            {
+                return (float*)(c_templFeatures.d12_data + n * c_templFeatures.d12_step);
+            }
+
+            static __device__ float2* r1(int n)
+            {
+                return (float2*)(c_templFeatures.r1_data + n * c_templFeatures.r1_step);
+            }
+            static __device__ float2* r2(int n)
+            {
+                return (float2*)(c_templFeatures.r2_data + n * c_templFeatures.r2_step);
+            }
+        };
+        struct ImageFeatureTable
+        {
+            static __device__ float2* p1_pos(int n)
+            {
+                return (float2*)(c_imageFeatures.p1_pos_data + n * c_imageFeatures.p1_pos_step);
+            }
+            static __device__ float* p1_theta(int n)
+            {
+                return (float*)(c_imageFeatures.p1_theta_data + n * c_imageFeatures.p1_theta_step);
+            }
+            static __device__ float2* p2_pos(int n)
+            {
+                return (float2*)(c_imageFeatures.p2_pos_data + n * c_imageFeatures.p2_pos_step);
+            }
+
+            static __device__ float* d12(int n)
+            {
+                return (float*)(c_imageFeatures.d12_data + n * c_imageFeatures.d12_step);
+            }
+
+            static __device__ float2* r1(int n)
+            {
+                return (float2*)(c_imageFeatures.r1_data + n * c_imageFeatures.r1_step);
+            }
+            static __device__ float2* r2(int n)
+            {
+                return (float2*)(c_imageFeatures.r2_data + n * c_imageFeatures.r2_step);
+            }
+        };
+
+        __device__ float clampAngle(float a)
+        {
+            float res = a;
+
+            while (res > 2.0f * CV_PI_F)
+                res -= 2.0f * CV_PI_F;
+            while (res < 0.0f)
+                res += 2.0f * CV_PI_F;
+
+            return res;
+        }
+
+        __device__ bool angleEq(float a, float b, float eps)
+        {
+            return (::fabs(clampAngle(a - b)) <= eps);
+        }
+
+        template <class FT, bool isTempl>
+        __global__ void Guil_Full_buildFeatureList(const unsigned int* coordList, const float* thetaList, const int pointsCount,
+                                                   int* sizes, const int maxSize,
+                                                   const float xi, const float angleEpsilon, const float alphaScale,
+                                                   const float2 center, const float maxDist)
+        {
+            const float p1_theta = thetaList[blockIdx.x];
+            const unsigned int coord1 = coordList[blockIdx.x];
+            float2 p1_pos;
+            p1_pos.x = (coord1 & 0xFFFF);
+            p1_pos.y = (coord1 >> 16) & 0xFFFF;
+
+            for (int i = threadIdx.x; i < pointsCount; i += blockDim.x)
+            {
+                const float p2_theta = thetaList[i];
+                const unsigned int coord2 = coordList[i];
+                float2 p2_pos;
+                p2_pos.x = (coord2 & 0xFFFF);
+                p2_pos.y = (coord2 >> 16) & 0xFFFF;
+
+                if (angleEq(p1_theta - p2_theta, xi, angleEpsilon))
+                {
+                    const float2 d = p1_pos - p2_pos;
+
+                    float alpha12 = clampAngle(::atan2(d.y, d.x) - p1_theta);
+                    float d12 = ::sqrtf(d.x * d.x + d.y * d.y);
+
+                    if (d12 > maxDist)
+                        continue;
+
+                    float2 r1 = p1_pos - center;
+                    float2 r2 = p2_pos - center;
+
+                    const int n = __float2int_rn(alpha12 * alphaScale);
+
+                    const int ind = ::atomicAdd(sizes + n, 1);
+
+                    if (ind < maxSize)
+                    {
+                        if (!isTempl)
+                        {
+                            FT::p1_pos(n)[ind] = p1_pos;
+                            FT::p2_pos(n)[ind] = p2_pos;
+                        }
+
+                        FT::p1_theta(n)[ind] = p1_theta;
+
+                        FT::d12(n)[ind] = d12;
+
+                        if (isTempl)
+                        {
+                            FT::r1(n)[ind] = r1;
+                            FT::r2(n)[ind] = r2;
+                        }
+                    }
+                }
+            }
+        }
+
+        template <class FT, bool isTempl>
+        void Guil_Full_buildFeatureList_caller(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                               int* sizes, int maxSize,
+                                               float xi, float angleEpsilon, int levels,
+                                               float2 center, float maxDist)
+        {
+            const dim3 block(256);
+            const dim3 grid(pointsCount);
+
+            const float alphaScale = levels / (2.0f * CV_PI_F);
+
+            Guil_Full_buildFeatureList<FT, isTempl><<<grid, block>>>(coordList, thetaList, pointsCount,
+                                                                     sizes, maxSize,
+                                                                     xi * (CV_PI_F / 180.0f), angleEpsilon * (CV_PI_F / 180.0f), alphaScale,
+                                                                     center, maxDist);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            thrust::device_ptr<int> sizesPtr(sizes);
+            thrust::transform(sizesPtr, sizesPtr + levels + 1, sizesPtr, cudev::bind2nd(cudev::minimum<int>(), maxSize));
+        }
+
+        void Guil_Full_buildTemplFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                                 int* sizes, int maxSize,
+                                                 float xi, float angleEpsilon, int levels,
+                                                 float2 center, float maxDist)
+        {
+            Guil_Full_buildFeatureList_caller<TemplFeatureTable, true>(coordList, thetaList, pointsCount,
+                                                                       sizes, maxSize,
+                                                                       xi, angleEpsilon, levels,
+                                                                       center, maxDist);
+        }
+        void Guil_Full_buildImageFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                                 int* sizes, int maxSize,
+                                                 float xi, float angleEpsilon, int levels,
+                                                 float2 center, float maxDist)
+        {
+            Guil_Full_buildFeatureList_caller<ImageFeatureTable, false>(coordList, thetaList, pointsCount,
+                                                                        sizes, maxSize,
+                                                                        xi, angleEpsilon, levels,
+                                                                        center, maxDist);
+        }
+
+        __global__ void Guil_Full_calcOHist(const int* templSizes, const int* imageSizes, int* OHist,
+                                            const float minAngle, const float maxAngle, const float iAngleStep, const int angleRange)
+        {
+            extern __shared__ int s_OHist[];
+            for (int i = threadIdx.x; i <= angleRange; i += blockDim.x)
+                s_OHist[i] = 0;
+            __syncthreads();
+
+            const int tIdx = blockIdx.x;
+            const int level = blockIdx.y;
+
+            const int tSize = templSizes[level];
+
+            if (tIdx < tSize)
+            {
+                const int imSize = imageSizes[level];
+
+                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx];
+
+                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
+                {
+                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
+
+                    const float angle = clampAngle(im_p1_theta - t_p1_theta);
+
+                    if (angle >= minAngle && angle <= maxAngle)
+                    {
+                        const int n = __float2int_rn((angle - minAngle) * iAngleStep);
+                        Emulation::smem::atomicAdd(&s_OHist[n], 1);
+                    }
+                }
+            }
+            __syncthreads();
+
+            for (int i = threadIdx.x; i <= angleRange; i += blockDim.x)
+                ::atomicAdd(OHist + i, s_OHist[i]);
+        }
+
+        void Guil_Full_calcOHist_gpu(const int* templSizes, const int* imageSizes, int* OHist,
+                                     float minAngle, float maxAngle, float angleStep, int angleRange,
+                                     int levels, int tMaxSize)
+        {
+            const dim3 block(256);
+            const dim3 grid(tMaxSize, levels + 1);
+
+            minAngle *= (CV_PI_F / 180.0f);
+            maxAngle *= (CV_PI_F / 180.0f);
+            angleStep *= (CV_PI_F / 180.0f);
+
+            const size_t smemSize = (angleRange + 1) * sizeof(float);
+
+            Guil_Full_calcOHist<<<grid, block, smemSize>>>(templSizes, imageSizes, OHist,
+                                                           minAngle, maxAngle, 1.0f / angleStep, angleRange);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        __global__ void Guil_Full_calcSHist(const int* templSizes, const int* imageSizes, int* SHist,
+                                            const float angle, const float angleEpsilon,
+                                            const float minScale, const float maxScale, const float iScaleStep, const int scaleRange)
+        {
+            extern __shared__ int s_SHist[];
+            for (int i = threadIdx.x; i <= scaleRange; i += blockDim.x)
+                s_SHist[i] = 0;
+            __syncthreads();
+
+            const int tIdx = blockIdx.x;
+            const int level = blockIdx.y;
+
+            const int tSize = templSizes[level];
+
+            if (tIdx < tSize)
+            {
+                const int imSize = imageSizes[level];
+
+                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx] + angle;
+                const float t_d12 = TemplFeatureTable::d12(level)[tIdx] + angle;
+
+                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
+                {
+                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
+                    const float im_d12 = ImageFeatureTable::d12(level)[i];
+
+                    if (angleEq(im_p1_theta, t_p1_theta, angleEpsilon))
+                    {
+                        const float scale = im_d12 / t_d12;
+
+                        if (scale >= minScale && scale <= maxScale)
+                        {
+                            const int s = __float2int_rn((scale - minScale) * iScaleStep);
+                            Emulation::smem::atomicAdd(&s_SHist[s], 1);
+                        }
+                    }
+                }
+            }
+            __syncthreads();
+
+            for (int i = threadIdx.x; i <= scaleRange; i += blockDim.x)
+                ::atomicAdd(SHist + i, s_SHist[i]);
+        }
+
+        void Guil_Full_calcSHist_gpu(const int* templSizes, const int* imageSizes, int* SHist,
+                                     float angle, float angleEpsilon,
+                                     float minScale, float maxScale, float iScaleStep, int scaleRange,
+                                     int levels, int tMaxSize)
+        {
+            const dim3 block(256);
+            const dim3 grid(tMaxSize, levels + 1);
+
+            angle *= (CV_PI_F / 180.0f);
+            angleEpsilon *= (CV_PI_F / 180.0f);
+
+            const size_t smemSize = (scaleRange + 1) * sizeof(float);
+
+            Guil_Full_calcSHist<<<grid, block, smemSize>>>(templSizes, imageSizes, SHist,
+                                                           angle, angleEpsilon,
+                                                           minScale, maxScale, iScaleStep, scaleRange);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        __global__ void Guil_Full_calcPHist(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
+                                            const float angle, const float sinVal, const float cosVal, const float angleEpsilon, const float scale,
+                                            const float idp)
+        {
+            const int tIdx = blockIdx.x;
+            const int level = blockIdx.y;
+
+            const int tSize = templSizes[level];
+
+            if (tIdx < tSize)
+            {
+                const int imSize = imageSizes[level];
+
+                const float t_p1_theta = TemplFeatureTable::p1_theta(level)[tIdx] + angle;
+
+                float2 r1 = TemplFeatureTable::r1(level)[tIdx];
+                float2 r2 = TemplFeatureTable::r2(level)[tIdx];
+
+                r1 = r1 * scale;
+                r2 = r2 * scale;
+
+                r1 = make_float2(cosVal * r1.x - sinVal * r1.y, sinVal * r1.x + cosVal * r1.y);
+                r2 = make_float2(cosVal * r2.x - sinVal * r2.y, sinVal * r2.x + cosVal * r2.y);
+
+                for (int i = threadIdx.x; i < imSize; i += blockDim.x)
+                {
+                    const float im_p1_theta = ImageFeatureTable::p1_theta(level)[i];
+
+                    const float2 im_p1_pos = ImageFeatureTable::p1_pos(level)[i];
+                    const float2 im_p2_pos = ImageFeatureTable::p2_pos(level)[i];
+
+                    if (angleEq(im_p1_theta, t_p1_theta, angleEpsilon))
+                    {
+                        float2 c1, c2;
+
+                        c1 = im_p1_pos - r1;
+                        c1 = c1 * idp;
+
+                        c2 = im_p2_pos - r2;
+                        c2 = c2 * idp;
+
+                        if (::fabs(c1.x - c2.x) > 1 || ::fabs(c1.y - c2.y) > 1)
+                            continue;
+
+                        if (c1.y >= 0 && c1.y < PHist.rows - 2 && c1.x >= 0 && c1.x < PHist.cols - 2)
+                            ::atomicAdd(PHist.ptr(__float2int_rn(c1.y) + 1) + __float2int_rn(c1.x) + 1, 1);
+                    }
+                }
+            }
+        }
+
+        void Guil_Full_calcPHist_gpu(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
+                                     float angle, float angleEpsilon, float scale,
+                                     float dp,
+                                     int levels, int tMaxSize)
+        {
+            const dim3 block(256);
+            const dim3 grid(tMaxSize, levels + 1);
+
+            angle *= (CV_PI_F / 180.0f);
+            angleEpsilon *= (CV_PI_F / 180.0f);
+
+            const float sinVal = ::sinf(angle);
+            const float cosVal = ::cosf(angle);
+
+            cudaSafeCall( cudaFuncSetCacheConfig(Guil_Full_calcPHist, cudaFuncCachePreferL1) );
+
+            Guil_Full_calcPHist<<<grid, block>>>(templSizes, imageSizes, PHist,
+                                                 angle, sinVal, cosVal, angleEpsilon, scale,
+                                                 1.0f / dp);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        __global__ void Guil_Full_findPosInHist(const PtrStepSzi hist, float4* out, int3* votes, const int maxSize,
+                                                const float angle, const int angleVotes, const float scale, const int scaleVotes,
+                                                const float dp, const int threshold)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x >= hist.cols - 2 || y >= hist.rows - 2)
+                return;
+
+            const int curVotes = hist(y + 1, x + 1);
+
+            if (curVotes > threshold &&
+                curVotes >  hist(y + 1, x) &&
+                curVotes >= hist(y + 1, x + 2) &&
+                curVotes >  hist(y, x + 1) &&
+                curVotes >= hist(y + 2, x + 1))
+            {
+                const int ind = ::atomicAdd(&g_counter, 1);
+
+                if (ind < maxSize)
+                {
+                    out[ind] = make_float4(x * dp, y * dp, scale, angle);
+                    votes[ind] = make_int3(curVotes, scaleVotes, angleVotes);
+                }
+            }
+        }
+
+        int Guil_Full_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int curSize, int maxSize,
+                                        float angle, int angleVotes, float scale, int scaleVotes,
+                                        float dp, int threshold)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemcpy(counterPtr, &curSize, sizeof(int), cudaMemcpyHostToDevice) );
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(hist.cols - 2, block.x), divUp(hist.rows - 2, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(Guil_Full_findPosInHist, cudaFuncCachePreferL1) );
+
+            Guil_Full_findPosInHist<<<grid, block>>>(hist, out, votes, maxSize,
+                                                     angle, angleVotes, scale, scaleVotes,
+                                                     dp, threshold);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            totalCount = ::min(totalCount, maxSize);
+
+            return totalCount;
+        }
+    }
+}}}
+
+#endif // HAVE_OPENCV_GPUARITHM
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/gftt.cu b/modules/gpuimgproc/src/cuda/gftt.cu
similarity index 86%
rename from modules/gpu/src/cuda/gftt.cu
rename to modules/gpuimgproc/src/cuda/gftt.cu
index cae217e394..b4af9e5dbc 100644
--- a/modules/gpu/src/cuda/gftt.cu
+++ b/modules/gpuimgproc/src/cuda/gftt.cu
@@ -38,11 +38,6 @@
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
-// Copyright (c) 2010, Paul Furgale, Chi Hay Tong
-//
-// The original code was written by Paul Furgale and Chi Hay Tong
-// and later optimized and prepared for integration into OpenCV by Itseez.
-//
 //M*/
 
 #if !defined CUDA_DISABLER
@@ -50,21 +45,19 @@
 #include <thrust/device_ptr.h>
 #include <thrust/sort.h>
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace gfft
     {
         texture<float, cudaTextureType2D, cudaReadModeElementType> eigTex(0, cudaFilterModePoint, cudaAddressModeClamp);
 
-        __device__ uint g_counter = 0;
+        __device__ int g_counter = 0;
 
-        template <class Mask> __global__ void findCorners(float threshold, const Mask mask, float2* corners, uint max_count, int rows, int cols)
+        template <class Mask> __global__ void findCorners(float threshold, const Mask mask, float2* corners, int max_count, int rows, int cols)
         {
-            #if __CUDA_ARCH__ >= 110
-
             const int j = blockIdx.x * blockDim.x + threadIdx.x;
             const int i = blockIdx.y * blockDim.y + threadIdx.y;
 
@@ -89,15 +82,13 @@ namespace cv { namespace gpu { namespace device
 
                     if (val == maxVal)
                     {
-                        const uint ind = atomicInc(&g_counter, (uint)(-1));
+                        const int ind = ::atomicAdd(&g_counter, 1);
 
                         if (ind < max_count)
                             corners[ind] = make_float2(j, i);
                     }
                 }
             }
-
-            #endif // __CUDA_ARCH__ >= 110
         }
 
         int findCorners_gpu(PtrStepSzf eig, float threshold, PtrStepSzb mask, float2* corners, int max_count)
@@ -105,7 +96,7 @@ namespace cv { namespace gpu { namespace device
             void* counter_ptr;
             cudaSafeCall( cudaGetSymbolAddress(&counter_ptr, g_counter) );
 
-            cudaSafeCall( cudaMemset(counter_ptr, 0, sizeof(uint)) );
+            cudaSafeCall( cudaMemset(counter_ptr, 0, sizeof(int)) );
 
             bindTexture(&eigTex, eig);
 
@@ -121,10 +112,10 @@ namespace cv { namespace gpu { namespace device
 
             cudaSafeCall( cudaDeviceSynchronize() );
 
-            uint count;
-            cudaSafeCall( cudaMemcpy(&count, counter_ptr, sizeof(uint), cudaMemcpyDeviceToHost) );
+            int count;
+            cudaSafeCall( cudaMemcpy(&count, counter_ptr, sizeof(int), cudaMemcpyDeviceToHost) );
 
-            return min(count, max_count);
+            return std::min(count, max_count);
         }
 
         class EigGreater
diff --git a/modules/gpu/src/cuda/hist.cu b/modules/gpuimgproc/src/cuda/hist.cu
similarity index 61%
rename from modules/gpu/src/cuda/hist.cu
rename to modules/gpuimgproc/src/cuda/hist.cu
index 9dd14182be..51931d7ce5 100644
--- a/modules/gpu/src/cuda/hist.cu
+++ b/modules/gpuimgproc/src/cuda/hist.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -29,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -43,13 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/emulation.hpp"
-#include "opencv2/gpu/device/transform.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/transform.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 namespace hist
 {
@@ -110,6 +109,86 @@ namespace hist
 
 /////////////////////////////////////////////////////////////////////////
 
+namespace hist
+{
+    __device__ __forceinline__ void histEvenInc(int* shist, uint data, int binSize, int lowerLevel, int upperLevel)
+    {
+        if (data >= lowerLevel && data <= upperLevel)
+        {
+            const uint ind = (data - lowerLevel) / binSize;
+            Emulation::smem::atomicAdd(shist + ind, 1);
+        }
+    }
+
+    __global__ void histEven8u(const uchar* src, const size_t step, const int rows, const int cols,
+                               int* hist, const int binCount, const int binSize, const int lowerLevel, const int upperLevel)
+    {
+        extern __shared__ int shist[];
+
+        const int y = blockIdx.x * blockDim.y + threadIdx.y;
+        const int tid = threadIdx.y * blockDim.x + threadIdx.x;
+
+        if (tid < binCount)
+            shist[tid] = 0;
+
+        __syncthreads();
+
+        if (y < rows)
+        {
+            const uchar* rowPtr = src + y * step;
+            const uint* rowPtr4 = (uint*) rowPtr;
+
+            const int cols_4 = cols / 4;
+            for (int x = threadIdx.x; x < cols_4; x += blockDim.x)
+            {
+                const uint data = rowPtr4[x];
+
+                histEvenInc(shist, (data >>  0) & 0xFFU, binSize, lowerLevel, upperLevel);
+                histEvenInc(shist, (data >>  8) & 0xFFU, binSize, lowerLevel, upperLevel);
+                histEvenInc(shist, (data >> 16) & 0xFFU, binSize, lowerLevel, upperLevel);
+                histEvenInc(shist, (data >> 24) & 0xFFU, binSize, lowerLevel, upperLevel);
+            }
+
+            if (cols % 4 != 0 && threadIdx.x == 0)
+            {
+                for (int x = cols_4 * 4; x < cols; ++x)
+                {
+                    const uchar data = rowPtr[x];
+                    histEvenInc(shist, data, binSize, lowerLevel, upperLevel);
+                }
+            }
+        }
+
+        __syncthreads();
+
+        if (tid < binCount)
+        {
+            const int histVal = shist[tid];
+
+            if (histVal > 0)
+                ::atomicAdd(hist + tid, histVal);
+        }
+    }
+
+    void histEven8u(PtrStepSzb src, int* hist, int binCount, int lowerLevel, int upperLevel, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.rows, block.y));
+
+        const int binSize = divUp(upperLevel - lowerLevel, binCount);
+
+        const size_t smem_size = binCount * sizeof(int);
+
+        histEven8u<<<grid, block, smem_size, stream>>>(src.data, src.step, src.rows, src.cols, hist, binCount, binSize, lowerLevel, upperLevel);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+/////////////////////////////////////////////////////////////////////////
+
 namespace hist
 {
     __constant__ int c_lut[256];
@@ -128,7 +207,7 @@ namespace hist
     };
 }
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     template <> struct TransformFunctorTraits<hist::EqualizeHist> : DefaultTransformFunctorTraits<hist::EqualizeHist>
     {
@@ -147,7 +226,7 @@ namespace hist
 
         const float scale = 255.0f / (src.cols * src.rows);
 
-        transform(src, dst, EqualizeHist(scale), WithOutMask(), stream);
+        cudev::transform(src, dst, EqualizeHist(scale), WithOutMask(), stream);
     }
 }
 
diff --git a/modules/gpuimgproc/src/cuda/hough_circles.cu b/modules/gpuimgproc/src/cuda/hough_circles.cu
new file mode 100644
index 0000000000..6757e430b6
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/hough_circles.cu
@@ -0,0 +1,260 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/dynamic_smem.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough_circles
+    {
+        __device__ int g_counter;
+
+        ////////////////////////////////////////////////////////////////////////
+        // circlesAccumCenters
+
+        __global__ void circlesAccumCenters(const unsigned int* list, const int count, const PtrStepi dx, const PtrStepi dy,
+                                            PtrStepi accum, const int width, const int height, const int minRadius, const int maxRadius, const float idp)
+        {
+            const int SHIFT = 10;
+            const int ONE = 1 << SHIFT;
+
+            const int tid = blockIdx.x * blockDim.x + threadIdx.x;
+
+            if (tid >= count)
+                return;
+
+            const unsigned int val = list[tid];
+
+            const int x = (val & 0xFFFF);
+            const int y = (val >> 16) & 0xFFFF;
+
+            const int vx = dx(y, x);
+            const int vy = dy(y, x);
+
+            if (vx == 0 && vy == 0)
+                return;
+
+            const float mag = ::sqrtf(vx * vx + vy * vy);
+
+            const int x0 = __float2int_rn((x * idp) * ONE);
+            const int y0 = __float2int_rn((y * idp) * ONE);
+
+            int sx = __float2int_rn((vx * idp) * ONE / mag);
+            int sy = __float2int_rn((vy * idp) * ONE / mag);
+
+            // Step from minRadius to maxRadius in both directions of the gradient
+            for (int k1 = 0; k1 < 2; ++k1)
+            {
+                int x1 = x0 + minRadius * sx;
+                int y1 = y0 + minRadius * sy;
+
+                for (int r = minRadius; r <= maxRadius; x1 += sx, y1 += sy, ++r)
+                {
+                    const int x2 = x1 >> SHIFT;
+                    const int y2 = y1 >> SHIFT;
+
+                    if (x2 < 0 || x2 >= width || y2 < 0 || y2 >= height)
+                        break;
+
+                    ::atomicAdd(accum.ptr(y2 + 1) + x2 + 1, 1);
+                }
+
+                sx = -sx;
+                sy = -sy;
+            }
+        }
+
+        void circlesAccumCenters_gpu(const unsigned int* list, int count, PtrStepi dx, PtrStepi dy, PtrStepSzi accum, int minRadius, int maxRadius, float idp)
+        {
+            const dim3 block(256);
+            const dim3 grid(divUp(count, block.x));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(circlesAccumCenters, cudaFuncCachePreferL1) );
+
+            circlesAccumCenters<<<grid, block>>>(list, count, dx, dy, accum, accum.cols - 2, accum.rows - 2, minRadius, maxRadius, idp);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        ////////////////////////////////////////////////////////////////////////
+        // buildCentersList
+
+        __global__ void buildCentersList(const PtrStepSzi accum, unsigned int* centers, const int threshold)
+        {
+            const int x = blockIdx.x * blockDim.x + threadIdx.x;
+            const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (x < accum.cols - 2 && y < accum.rows - 2)
+            {
+                const int top = accum(y, x + 1);
+
+                const int left = accum(y + 1, x);
+                const int cur = accum(y + 1, x + 1);
+                const int right = accum(y + 1, x + 2);
+
+                const int bottom = accum(y + 2, x + 1);
+
+                if (cur > threshold && cur > top && cur >= bottom && cur >  left && cur >= right)
+                {
+                    const unsigned int val = (y << 16) | x;
+                    const int idx = ::atomicAdd(&g_counter, 1);
+                    centers[idx] = val;
+                }
+            }
+        }
+
+        int buildCentersList_gpu(PtrStepSzi accum, unsigned int* centers, int threshold)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(buildCentersList, cudaFuncCachePreferL1) );
+
+            buildCentersList<<<grid, block>>>(accum, centers, threshold);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            return totalCount;
+        }
+
+        ////////////////////////////////////////////////////////////////////////
+        // circlesAccumRadius
+
+        __global__ void circlesAccumRadius(const unsigned int* centers, const unsigned int* list, const int count,
+                                           float3* circles, const int maxCircles, const float dp,
+                                           const int minRadius, const int maxRadius, const int histSize, const int threshold)
+        {
+            int* smem = DynamicSharedMem<int>();
+
+            for (int i = threadIdx.x; i < histSize + 2; i += blockDim.x)
+                smem[i] = 0;
+            __syncthreads();
+
+            unsigned int val = centers[blockIdx.x];
+
+            float cx = (val & 0xFFFF);
+            float cy = (val >> 16) & 0xFFFF;
+
+            cx = (cx + 0.5f) * dp;
+            cy = (cy + 0.5f) * dp;
+
+            for (int i = threadIdx.x; i < count; i += blockDim.x)
+            {
+                val = list[i];
+
+                const int x = (val & 0xFFFF);
+                const int y = (val >> 16) & 0xFFFF;
+
+                const float rad = ::sqrtf((cx - x) * (cx - x) + (cy - y) * (cy - y));
+                if (rad >= minRadius && rad <= maxRadius)
+                {
+                    const int r = __float2int_rn(rad - minRadius);
+
+                    Emulation::smem::atomicAdd(&smem[r + 1], 1);
+                }
+            }
+
+            __syncthreads();
+
+            for (int i = threadIdx.x; i < histSize; i += blockDim.x)
+            {
+                const int curVotes = smem[i + 1];
+
+                if (curVotes >= threshold && curVotes > smem[i] && curVotes >= smem[i + 2])
+                {
+                    const int ind = ::atomicAdd(&g_counter, 1);
+                    if (ind < maxCircles)
+                        circles[ind] = make_float3(cx, cy, i + minRadius);
+                }
+            }
+        }
+
+        int circlesAccumRadius_gpu(const unsigned int* centers, int centersCount, const unsigned int* list, int count,
+                                   float3* circles, int maxCircles, float dp, int minRadius, int maxRadius, int threshold, bool has20)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(has20 ? 1024 : 512);
+            const dim3 grid(centersCount);
+
+            const int histSize = maxRadius - minRadius + 1;
+            size_t smemSize = (histSize + 2) * sizeof(int);
+
+            circlesAccumRadius<<<grid, block, smemSize>>>(centers, list, count, circles, maxCircles, dp, minRadius, maxRadius, histSize, threshold);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            totalCount = ::min(totalCount, maxCircles);
+
+            return totalCount;
+        }
+    }
+}}}
+
+#endif // HAVE_OPENCV_GPUFILTERS
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/hough_lines.cu b/modules/gpuimgproc/src/cuda/hough_lines.cu
new file mode 100644
index 0000000000..0cee0a43d2
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/hough_lines.cu
@@ -0,0 +1,212 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include <thrust/device_ptr.h>
+#include <thrust/sort.h>
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/emulation.hpp"
+#include "opencv2/core/cuda/dynamic_smem.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough_lines
+    {
+        __device__ int g_counter;
+
+        ////////////////////////////////////////////////////////////////////////
+        // linesAccum
+
+        __global__ void linesAccumGlobal(const unsigned int* list, const int count, PtrStepi accum, const float irho, const float theta, const int numrho)
+        {
+            const int n = blockIdx.x;
+            const float ang = n * theta;
+
+            float sinVal;
+            float cosVal;
+            sincosf(ang, &sinVal, &cosVal);
+            sinVal *= irho;
+            cosVal *= irho;
+
+            const int shift = (numrho - 1) / 2;
+
+            int* accumRow = accum.ptr(n + 1);
+            for (int i = threadIdx.x; i < count; i += blockDim.x)
+            {
+                const unsigned int val = list[i];
+
+                const int x = (val & 0xFFFF);
+                const int y = (val >> 16) & 0xFFFF;
+
+                int r = __float2int_rn(x * cosVal + y * sinVal);
+                r += shift;
+
+                ::atomicAdd(accumRow + r + 1, 1);
+            }
+        }
+
+        __global__ void linesAccumShared(const unsigned int* list, const int count, PtrStepi accum, const float irho, const float theta, const int numrho)
+        {
+            int* smem = DynamicSharedMem<int>();
+
+            for (int i = threadIdx.x; i < numrho + 1; i += blockDim.x)
+                smem[i] = 0;
+
+            __syncthreads();
+
+            const int n = blockIdx.x;
+            const float ang = n * theta;
+
+            float sinVal;
+            float cosVal;
+            sincosf(ang, &sinVal, &cosVal);
+            sinVal *= irho;
+            cosVal *= irho;
+
+            const int shift = (numrho - 1) / 2;
+
+            for (int i = threadIdx.x; i < count; i += blockDim.x)
+            {
+                const unsigned int val = list[i];
+
+                const int x = (val & 0xFFFF);
+                const int y = (val >> 16) & 0xFFFF;
+
+                int r = __float2int_rn(x * cosVal + y * sinVal);
+                r += shift;
+
+                Emulation::smem::atomicAdd(&smem[r + 1], 1);
+            }
+
+            __syncthreads();
+
+            int* accumRow = accum.ptr(n + 1);
+            for (int i = threadIdx.x; i < numrho + 1; i += blockDim.x)
+                accumRow[i] = smem[i];
+        }
+
+        void linesAccum_gpu(const unsigned int* list, int count, PtrStepSzi accum, float rho, float theta, size_t sharedMemPerBlock, bool has20)
+        {
+            const dim3 block(has20 ? 1024 : 512);
+            const dim3 grid(accum.rows - 2);
+
+            size_t smemSize = (accum.cols - 1) * sizeof(int);
+
+            if (smemSize < sharedMemPerBlock - 1000)
+                linesAccumShared<<<grid, block, smemSize>>>(list, count, accum, 1.0f / rho, theta, accum.cols - 2);
+            else
+                linesAccumGlobal<<<grid, block>>>(list, count, accum, 1.0f / rho, theta, accum.cols - 2);
+
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        ////////////////////////////////////////////////////////////////////////
+        // linesGetResult
+
+        __global__ void linesGetResult(const PtrStepSzi accum, float2* out, int* votes, const int maxSize, const float rho, const float theta, const int threshold, const int numrho)
+        {
+            const int r = blockIdx.x * blockDim.x + threadIdx.x;
+            const int n = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (r >= accum.cols - 2 || n >= accum.rows - 2)
+                return;
+
+            const int curVotes = accum(n + 1, r + 1);
+
+            if (curVotes > threshold &&
+                curVotes >  accum(n + 1, r) &&
+                curVotes >= accum(n + 1, r + 2) &&
+                curVotes >  accum(n, r + 1) &&
+                curVotes >= accum(n + 2, r + 1))
+            {
+                const float radius = (r - (numrho - 1) * 0.5f) * rho;
+                const float angle = n * theta;
+
+                const int ind = ::atomicAdd(&g_counter, 1);
+                if (ind < maxSize)
+                {
+                    out[ind] = make_float2(radius, angle);
+                    votes[ind] = curVotes;
+                }
+            }
+        }
+
+        int linesGetResult_gpu(PtrStepSzi accum, float2* out, int* votes, int maxSize, float rho, float theta, int threshold, bool doSort)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
+
+            cudaSafeCall( cudaFuncSetCacheConfig(linesGetResult, cudaFuncCachePreferL1) );
+
+            linesGetResult<<<grid, block>>>(accum, out, votes, maxSize, rho, theta, threshold, accum.cols - 2);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            totalCount = ::min(totalCount, maxSize);
+
+            if (doSort && totalCount > 0)
+            {
+                thrust::device_ptr<float2> outPtr(out);
+                thrust::device_ptr<int> votesPtr(votes);
+                thrust::sort_by_key(votesPtr, votesPtr + totalCount, outPtr, thrust::greater<int>());
+            }
+
+            return totalCount;
+        }
+    }
+}}}
+
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/hough_segments.cu b/modules/gpuimgproc/src/cuda/hough_segments.cu
new file mode 100644
index 0000000000..e420449fae
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/hough_segments.cu
@@ -0,0 +1,249 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough_segments
+    {
+        __device__ int g_counter;
+
+        texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_mask(false, cudaFilterModePoint, cudaAddressModeClamp);
+
+        __global__ void houghLinesProbabilistic(const PtrStepSzi accum,
+                                                int4* out, const int maxSize,
+                                                const float rho, const float theta,
+                                                const int lineGap, const int lineLength,
+                                                const int rows, const int cols)
+        {
+            const int r = blockIdx.x * blockDim.x + threadIdx.x;
+            const int n = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if (r >= accum.cols - 2 || n >= accum.rows - 2)
+                return;
+
+            const int curVotes = accum(n + 1, r + 1);
+
+            if (curVotes >= lineLength &&
+                curVotes > accum(n, r) &&
+                curVotes > accum(n, r + 1) &&
+                curVotes > accum(n, r + 2) &&
+                curVotes > accum(n + 1, r) &&
+                curVotes > accum(n + 1, r + 2) &&
+                curVotes > accum(n + 2, r) &&
+                curVotes > accum(n + 2, r + 1) &&
+                curVotes > accum(n + 2, r + 2))
+            {
+                const float radius = (r - (accum.cols - 2 - 1) * 0.5f) * rho;
+                const float angle = n * theta;
+
+                float cosa;
+                float sina;
+                sincosf(angle, &sina, &cosa);
+
+                float2 p0 = make_float2(cosa * radius, sina * radius);
+                float2 dir = make_float2(-sina, cosa);
+
+                float2 pb[4] = {make_float2(-1, -1), make_float2(-1, -1), make_float2(-1, -1), make_float2(-1, -1)};
+                float a;
+
+                if (dir.x != 0)
+                {
+                    a = -p0.x / dir.x;
+                    pb[0].x = 0;
+                    pb[0].y = p0.y + a * dir.y;
+
+                    a = (cols - 1 - p0.x) / dir.x;
+                    pb[1].x = cols - 1;
+                    pb[1].y = p0.y + a * dir.y;
+                }
+                if (dir.y != 0)
+                {
+                    a = -p0.y / dir.y;
+                    pb[2].x = p0.x + a * dir.x;
+                    pb[2].y = 0;
+
+                    a = (rows - 1 - p0.y) / dir.y;
+                    pb[3].x = p0.x + a * dir.x;
+                    pb[3].y = rows - 1;
+                }
+
+                if (pb[0].x == 0 && (pb[0].y >= 0 && pb[0].y < rows))
+                {
+                    p0 = pb[0];
+                    if (dir.x < 0)
+                        dir = -dir;
+                }
+                else if (pb[1].x == cols - 1 && (pb[0].y >= 0 && pb[0].y < rows))
+                {
+                    p0 = pb[1];
+                    if (dir.x > 0)
+                        dir = -dir;
+                }
+                else if (pb[2].y == 0 && (pb[2].x >= 0 && pb[2].x < cols))
+                {
+                    p0 = pb[2];
+                    if (dir.y < 0)
+                        dir = -dir;
+                }
+                else if (pb[3].y == rows - 1 && (pb[3].x >= 0 && pb[3].x < cols))
+                {
+                    p0 = pb[3];
+                    if (dir.y > 0)
+                        dir = -dir;
+                }
+
+                float2 d;
+                if (::fabsf(dir.x) > ::fabsf(dir.y))
+                {
+                    d.x = dir.x > 0 ? 1 : -1;
+                    d.y = dir.y / ::fabsf(dir.x);
+                }
+                else
+                {
+                    d.x = dir.x / ::fabsf(dir.y);
+                    d.y = dir.y > 0 ? 1 : -1;
+                }
+
+                float2 line_end[2];
+                int gap;
+                bool inLine = false;
+
+                float2 p1 = p0;
+                if (p1.x < 0 || p1.x >= cols || p1.y < 0 || p1.y >= rows)
+                    return;
+
+                for (;;)
+                {
+                    if (tex2D(tex_mask, p1.x, p1.y))
+                    {
+                        gap = 0;
+
+                        if (!inLine)
+                        {
+                            line_end[0] = p1;
+                            line_end[1] = p1;
+                            inLine = true;
+                        }
+                        else
+                        {
+                            line_end[1] = p1;
+                        }
+                    }
+                    else if (inLine)
+                    {
+                        if (++gap > lineGap)
+                        {
+                            bool good_line = ::abs(line_end[1].x - line_end[0].x) >= lineLength ||
+                                             ::abs(line_end[1].y - line_end[0].y) >= lineLength;
+
+                            if (good_line)
+                            {
+                                const int ind = ::atomicAdd(&g_counter, 1);
+                                if (ind < maxSize)
+                                    out[ind] = make_int4(line_end[0].x, line_end[0].y, line_end[1].x, line_end[1].y);
+                            }
+
+                            gap = 0;
+                            inLine = false;
+                        }
+                    }
+
+                    p1 = p1 + d;
+                    if (p1.x < 0 || p1.x >= cols || p1.y < 0 || p1.y >= rows)
+                    {
+                        if (inLine)
+                        {
+                            bool good_line = ::abs(line_end[1].x - line_end[0].x) >= lineLength ||
+                                             ::abs(line_end[1].y - line_end[0].y) >= lineLength;
+
+                            if (good_line)
+                            {
+                                const int ind = ::atomicAdd(&g_counter, 1);
+                                if (ind < maxSize)
+                                    out[ind] = make_int4(line_end[0].x, line_end[0].y, line_end[1].x, line_end[1].y);
+                            }
+
+                        }
+                        break;
+                    }
+                }
+            }
+        }
+
+        int houghLinesProbabilistic_gpu(PtrStepSzb mask, PtrStepSzi accum, int4* out, int maxSize, float rho, float theta, int lineGap, int lineLength)
+        {
+            void* counterPtr;
+            cudaSafeCall( cudaGetSymbolAddress(&counterPtr, g_counter) );
+
+            cudaSafeCall( cudaMemset(counterPtr, 0, sizeof(int)) );
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(accum.cols - 2, block.x), divUp(accum.rows - 2, block.y));
+
+            bindTexture(&tex_mask, mask);
+
+            houghLinesProbabilistic<<<grid, block>>>(accum,
+                                                     out, maxSize,
+                                                     rho, theta,
+                                                     lineGap, lineLength,
+                                                     mask.rows, mask.cols);
+            cudaSafeCall( cudaGetLastError() );
+
+            cudaSafeCall( cudaDeviceSynchronize() );
+
+            int totalCount;
+            cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
+
+            totalCount = ::min(totalCount, maxSize);
+
+            return totalCount;
+        }
+    }
+}}}
+
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/match_template.cu b/modules/gpuimgproc/src/cuda/match_template.cu
similarity index 99%
rename from modules/gpu/src/cuda/match_template.cu
rename to modules/gpuimgproc/src/cuda/match_template.cu
index 71afea2126..6670639290 100644
--- a/modules/gpu/src/cuda/match_template.cu
+++ b/modules/gpuimgproc/src/cuda/match_template.cu
@@ -42,10 +42,10 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace match_template
     {
@@ -910,7 +910,7 @@ namespace cv { namespace gpu { namespace device
                 cudaSafeCall( cudaDeviceSynchronize() );
         }
     } //namespace match_template
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpuimgproc/src/cuda/mean_shift.cu b/modules/gpuimgproc/src/cuda/mean_shift.cu
new file mode 100644
index 0000000000..aa82f295e4
--- /dev/null
+++ b/modules/gpuimgproc/src/cuda/mean_shift.cu
@@ -0,0 +1,182 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        texture<uchar4, 2> tex_meanshift;
+
+        __device__ short2 do_mean_shift(int x0, int y0, unsigned char* out,
+                                        size_t out_step, int cols, int rows,
+                                        int sp, int sr, int maxIter, float eps)
+        {
+            int isr2 = sr*sr;
+            uchar4 c = tex2D(tex_meanshift, x0, y0 );
+
+            // iterate meanshift procedure
+            for( int iter = 0; iter < maxIter; iter++ )
+            {
+                int count = 0;
+                int s0 = 0, s1 = 0, s2 = 0, sx = 0, sy = 0;
+                float icount;
+
+                //mean shift: process pixels in window (p-sigmaSp)x(p+sigmaSp)
+                int minx = x0-sp;
+                int miny = y0-sp;
+                int maxx = x0+sp;
+                int maxy = y0+sp;
+
+                for( int y = miny; y <= maxy; y++)
+                {
+                    int rowCount = 0;
+                    for( int x = minx; x <= maxx; x++ )
+                    {
+                        uchar4 t = tex2D( tex_meanshift, x, y );
+
+                        int norm2 = (t.x - c.x) * (t.x - c.x) + (t.y - c.y) * (t.y - c.y) + (t.z - c.z) * (t.z - c.z);
+                        if( norm2 <= isr2 )
+                        {
+                            s0 += t.x; s1 += t.y; s2 += t.z;
+                            sx += x; rowCount++;
+                        }
+                    }
+                    count += rowCount;
+                    sy += y*rowCount;
+                }
+
+                if( count == 0 )
+                    break;
+
+                icount = 1.f/count;
+                int x1 = __float2int_rz(sx*icount);
+                int y1 = __float2int_rz(sy*icount);
+                s0 = __float2int_rz(s0*icount);
+                s1 = __float2int_rz(s1*icount);
+                s2 = __float2int_rz(s2*icount);
+
+                int norm2 = (s0 - c.x) * (s0 - c.x) + (s1 - c.y) * (s1 - c.y) + (s2 - c.z) * (s2 - c.z);
+
+                bool stopFlag = (x0 == x1 && y0 == y1) || (::abs(x1-x0) + ::abs(y1-y0) + norm2 <= eps);
+
+                x0 = x1; y0 = y1;
+                c.x = s0; c.y = s1; c.z = s2;
+
+                if( stopFlag )
+                    break;
+            }
+
+            int base = (blockIdx.y * blockDim.y + threadIdx.y) * out_step + (blockIdx.x * blockDim.x + threadIdx.x) * 4 * sizeof(uchar);
+            *(uchar4*)(out + base) = c;
+
+            return make_short2((short)x0, (short)y0);
+        }
+
+        __global__ void meanshift_kernel(unsigned char* out, size_t out_step, int cols, int rows, int sp, int sr, int maxIter, float eps )
+        {
+            int x0 = blockIdx.x * blockDim.x + threadIdx.x;
+            int y0 = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if( x0 < cols && y0 < rows )
+                do_mean_shift(x0, y0, out, out_step, cols, rows, sp, sr, maxIter, eps);
+        }
+
+        void meanShiftFiltering_gpu(const PtrStepSzb& src, PtrStepSzb dst, int sp, int sr, int maxIter, float eps, cudaStream_t stream)
+        {
+            dim3 grid(1, 1, 1);
+            dim3 threads(32, 8, 1);
+            grid.x = divUp(src.cols, threads.x);
+            grid.y = divUp(src.rows, threads.y);
+
+            cudaChannelFormatDesc desc = cudaCreateChannelDesc<uchar4>();
+            cudaSafeCall( cudaBindTexture2D( 0, tex_meanshift, src.data, desc, src.cols, src.rows, src.step ) );
+
+            meanshift_kernel<<< grid, threads, 0, stream >>>( dst.data, dst.step, dst.cols, dst.rows, sp, sr, maxIter, eps );
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        __global__ void meanshiftproc_kernel(unsigned char* outr, size_t outrstep,
+                                             unsigned char* outsp, size_t outspstep,
+                                             int cols, int rows,
+                                             int sp, int sr, int maxIter, float eps)
+        {
+            int x0 = blockIdx.x * blockDim.x + threadIdx.x;
+            int y0 = blockIdx.y * blockDim.y + threadIdx.y;
+
+            if( x0 < cols && y0 < rows )
+            {
+                int basesp = (blockIdx.y * blockDim.y + threadIdx.y) * outspstep + (blockIdx.x * blockDim.x + threadIdx.x) * 2 * sizeof(short);
+                *(short2*)(outsp + basesp) = do_mean_shift(x0, y0, outr, outrstep, cols, rows, sp, sr, maxIter, eps);
+            }
+        }
+
+        void meanShiftProc_gpu(const PtrStepSzb& src, PtrStepSzb dstr, PtrStepSzb dstsp, int sp, int sr, int maxIter, float eps, cudaStream_t stream)
+        {
+            dim3 grid(1, 1, 1);
+            dim3 threads(32, 8, 1);
+            grid.x = divUp(src.cols, threads.x);
+            grid.y = divUp(src.rows, threads.y);
+
+            cudaChannelFormatDesc desc = cudaCreateChannelDesc<uchar4>();
+            cudaSafeCall( cudaBindTexture2D( 0, tex_meanshift, src.data, desc, src.cols, src.rows, src.step ) );
+
+            meanshiftproc_kernel<<< grid, threads, 0, stream >>>( dstr.data, dstr.step, dstsp.data, dstsp.step, dstr.cols, dstr.rows, sp, sr, maxIter, eps );
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    }
+}}}
+
+#endif
diff --git a/modules/gpu/src/cvt_color_internal.h b/modules/gpuimgproc/src/cvt_color_internal.h
similarity index 99%
rename from modules/gpu/src/cvt_color_internal.h
rename to modules/gpuimgproc/src/cvt_color_internal.h
index 1b7c68f35f..010d832a25 100644
--- a/modules/gpu/src/cvt_color_internal.h
+++ b/modules/gpuimgproc/src/cvt_color_internal.h
@@ -43,7 +43,7 @@
 #ifndef __cvt_color_internal_h__
 #define __cvt_color_internal_h__
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
 #define OPENCV_GPU_DECLARE_CVTCOLOR_ONE(name) \
     void name(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
diff --git a/modules/gpuimgproc/src/generalized_hough.cpp b/modules/gpuimgproc/src/generalized_hough.cpp
new file mode 100644
index 0000000000..abcd3ef496
--- /dev/null
+++ b/modules/gpuimgproc/src/generalized_hough.cpp
@@ -0,0 +1,906 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) || !defined(HAVE_OPENCV_GPUARITHM)
+
+Ptr<GeneralizedHoughBallard> cv::gpu::createGeneralizedHoughBallard() { throw_no_cuda(); return Ptr<GeneralizedHoughBallard>(); }
+
+Ptr<GeneralizedHoughGuil> cv::gpu::createGeneralizedHoughGuil() { throw_no_cuda(); return Ptr<GeneralizedHoughGuil>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace ght
+    {
+        template <typename T>
+        int buildEdgePointList_gpu(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
+        void buildRTable_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                             PtrStepSz<short2> r_table, int* r_sizes,
+                             short2 templCenter, int levels);
+
+        void Ballard_Pos_calcHist_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                      PtrStepSz<short2> r_table, const int* r_sizes,
+                                      PtrStepSzi hist,
+                                      float dp, int levels);
+        int Ballard_Pos_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int maxSize, float dp, int threshold);
+
+        void Guil_Full_setTemplFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
+        void Guil_Full_setImageFeatures(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
+        void Guil_Full_buildTemplFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                                 int* sizes, int maxSize,
+                                                 float xi, float angleEpsilon, int levels,
+                                                 float2 center, float maxDist);
+        void Guil_Full_buildImageFeatureList_gpu(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                                 int* sizes, int maxSize,
+                                                 float xi, float angleEpsilon, int levels,
+                                                 float2 center, float maxDist);
+        void Guil_Full_calcOHist_gpu(const int* templSizes, const int* imageSizes, int* OHist,
+                                     float minAngle, float maxAngle, float angleStep, int angleRange,
+                                     int levels, int tMaxSize);
+        void Guil_Full_calcSHist_gpu(const int* templSizes, const int* imageSizes, int* SHist,
+                                     float angle, float angleEpsilon,
+                                     float minScale, float maxScale, float iScaleStep, int scaleRange,
+                                     int levels, int tMaxSize);
+        void Guil_Full_calcPHist_gpu(const int* templSizes, const int* imageSizes, PtrStepSzi PHist,
+                                     float angle, float angleEpsilon, float scale,
+                                     float dp,
+                                     int levels, int tMaxSize);
+        int Guil_Full_findPosInHist_gpu(PtrStepSzi hist, float4* out, int3* votes, int curSize, int maxSize,
+                                        float angle, int angleVotes, float scale, int scaleVotes,
+                                        float dp, int threshold);
+    }
+}}}
+
+// common
+
+namespace
+{
+    class GeneralizedHoughBase
+    {
+    protected:
+        GeneralizedHoughBase();
+        virtual ~GeneralizedHoughBase() {}
+
+        void setTemplateImpl(InputArray templ, Point templCenter);
+        void setTemplateImpl(InputArray edges, InputArray dx, InputArray dy, Point templCenter);
+
+        void detectImpl(InputArray image, OutputArray positions, OutputArray votes);
+        void detectImpl(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes);
+
+        void buildEdgePointList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy);
+
+        virtual void processTempl() = 0;
+        virtual void processImage() = 0;
+
+        int cannyLowThresh_;
+        int cannyHighThresh_;
+        double minDist_;
+        double dp_;
+        int maxBufferSize_;
+
+        Size templSize_;
+        Point templCenter_;
+        GpuMat templEdges_;
+        GpuMat templDx_;
+        GpuMat templDy_;
+
+        Size imageSize_;
+        GpuMat imageEdges_;
+        GpuMat imageDx_;
+        GpuMat imageDy_;
+
+        GpuMat edgePointList_;
+
+        GpuMat outBuf_;
+        int posCount_;
+
+    private:
+#ifdef HAVE_OPENCV_GPUFILTERS
+        void calcEdges(InputArray src, GpuMat& edges, GpuMat& dx, GpuMat& dy);
+#endif
+
+        void filterMinDist();
+        void convertTo(OutputArray positions, OutputArray votes);
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        Ptr<gpu::CannyEdgeDetector> canny_;
+        Ptr<gpu::Filter> filterDx_;
+        Ptr<gpu::Filter> filterDy_;
+#endif
+
+        std::vector<float4> oldPosBuf_;
+        std::vector<int3> oldVoteBuf_;
+        std::vector<float4> newPosBuf_;
+        std::vector<int3> newVoteBuf_;
+        std::vector<int> indexies_;
+    };
+
+    GeneralizedHoughBase::GeneralizedHoughBase()
+    {
+        cannyLowThresh_ = 50;
+        cannyHighThresh_ = 100;
+        minDist_ = 1.0;
+        dp_ = 1.0;
+
+        maxBufferSize_ = 10000;
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+        canny_ = gpu::createCannyEdgeDetector(cannyLowThresh_, cannyHighThresh_);
+        filterDx_ = gpu::createSobelFilter(CV_8UC1, CV_32S, 1, 0);
+        filterDy_ = gpu::createSobelFilter(CV_8UC1, CV_32S, 0, 1);
+#endif
+    }
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+    void GeneralizedHoughBase::calcEdges(InputArray _src, GpuMat& edges, GpuMat& dx, GpuMat& dy)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+        CV_Assert( cannyLowThresh_ > 0 && cannyLowThresh_ < cannyHighThresh_ );
+
+        ensureSizeIsEnough(src.size(), CV_32SC1, dx);
+        ensureSizeIsEnough(src.size(), CV_32SC1, dy);
+
+        filterDx_->apply(src, dx);
+        filterDy_->apply(src, dy);
+
+        ensureSizeIsEnough(src.size(), CV_8UC1, edges);
+
+        canny_->setLowThreshold(cannyLowThresh_);
+        canny_->setHighThreshold(cannyHighThresh_);
+        canny_->detect(dx, dy, edges);
+    }
+#endif
+
+    void GeneralizedHoughBase::setTemplateImpl(InputArray templ, Point templCenter)
+    {
+#ifndef HAVE_OPENCV_GPUFILTERS
+        (void) templ;
+        (void) templCenter;
+        throw_no_cuda();
+#else
+        calcEdges(templ, templEdges_, templDx_, templDy_);
+
+        if (templCenter == Point(-1, -1))
+            templCenter = Point(templEdges_.cols / 2, templEdges_.rows / 2);
+
+        templSize_ = templEdges_.size();
+        templCenter_ = templCenter;
+
+        processTempl();
+#endif
+    }
+
+    void GeneralizedHoughBase::setTemplateImpl(InputArray edges, InputArray dx, InputArray dy, Point templCenter)
+    {
+        edges.getGpuMat().copyTo(templEdges_);
+        dx.getGpuMat().copyTo(templDx_);
+        dy.getGpuMat().copyTo(templDy_);
+
+        CV_Assert( templEdges_.type() == CV_8UC1 );
+        CV_Assert( templDx_.type() == CV_32FC1 && templDx_.size() == templEdges_.size() );
+        CV_Assert( templDy_.type() == templDx_.type() && templDy_.size() == templEdges_.size() );
+
+        if (templCenter == Point(-1, -1))
+            templCenter = Point(templEdges_.cols / 2, templEdges_.rows / 2);
+
+        templSize_ = templEdges_.size();
+        templCenter_ = templCenter;
+
+        processTempl();
+    }
+
+    void GeneralizedHoughBase::detectImpl(InputArray image, OutputArray positions, OutputArray votes)
+    {
+#ifndef HAVE_OPENCV_GPUFILTERS
+        (void) templ;
+        (void) templCenter;
+        throw_no_cuda();
+#else
+        calcEdges(image, imageEdges_, imageDx_, imageDy_);
+
+        imageSize_ = imageEdges_.size();
+
+        posCount_ = 0;
+
+        processImage();
+
+        if (posCount_ == 0)
+        {
+            positions.release();
+            if (votes.needed())
+                votes.release();
+        }
+        else
+        {
+            if (minDist_ > 1)
+                filterMinDist();
+            convertTo(positions, votes);
+        }
+#endif
+    }
+
+    void GeneralizedHoughBase::detectImpl(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes)
+    {
+        edges.getGpuMat().copyTo(imageEdges_);
+        dx.getGpuMat().copyTo(imageDx_);
+        dy.getGpuMat().copyTo(imageDy_);
+
+        CV_Assert( imageEdges_.type() == CV_8UC1 );
+        CV_Assert( imageDx_.type() == CV_32FC1 && imageDx_.size() == imageEdges_.size() );
+        CV_Assert( imageDy_.type() == imageDx_.type() && imageDy_.size() == imageEdges_.size() );
+
+        imageSize_ = imageEdges_.size();
+
+        posCount_ = 0;
+
+        processImage();
+
+        if (posCount_ == 0)
+        {
+            positions.release();
+            if (votes.needed())
+                votes.release();
+        }
+        else
+        {
+            if (minDist_ > 1)
+                filterMinDist();
+            convertTo(positions, votes);
+        }
+    }
+
+    void GeneralizedHoughBase::buildEdgePointList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy)
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        typedef int (*func_t)(PtrStepSzb edges, PtrStepSzb dx, PtrStepSzb dy, unsigned int* coordList, float* thetaList);
+        static const func_t funcs[] =
+        {
+            0,
+            0,
+            0,
+            buildEdgePointList_gpu<short>,
+            buildEdgePointList_gpu<int>,
+            buildEdgePointList_gpu<float>,
+            0
+        };
+
+        CV_Assert( edges.type() == CV_8UC1 );
+        CV_Assert( dx.size() == edges.size() );
+        CV_Assert( dy.type() == dx.type() && dy.size() == edges.size() );
+
+        const func_t func = funcs[dx.depth()];
+        CV_Assert( func != 0 );
+
+        edgePointList_.cols = (int) (edgePointList_.step / sizeof(int));
+        ensureSizeIsEnough(2, edges.size().area(), CV_32SC1, edgePointList_);
+
+        edgePointList_.cols = func(edges, dx, dy, edgePointList_.ptr<unsigned int>(0), edgePointList_.ptr<float>(1));
+    }
+
+    struct IndexCmp
+    {
+        const int3* aux;
+
+        explicit IndexCmp(const int3* _aux) : aux(_aux) {}
+
+        bool operator ()(int l1, int l2) const
+        {
+            return aux[l1].x > aux[l2].x;
+        }
+    };
+
+    void GeneralizedHoughBase::filterMinDist()
+    {
+        oldPosBuf_.resize(posCount_);
+        oldVoteBuf_.resize(posCount_);
+
+        cudaSafeCall( cudaMemcpy(&oldPosBuf_[0], outBuf_.ptr(0), posCount_ * sizeof(float4), cudaMemcpyDeviceToHost) );
+        cudaSafeCall( cudaMemcpy(&oldVoteBuf_[0], outBuf_.ptr(1), posCount_ * sizeof(int3), cudaMemcpyDeviceToHost) );
+
+        indexies_.resize(posCount_);
+        for (int i = 0; i < posCount_; ++i)
+            indexies_[i] = i;
+        std::sort(indexies_.begin(), indexies_.end(), IndexCmp(&oldVoteBuf_[0]));
+
+        newPosBuf_.clear();
+        newVoteBuf_.clear();
+        newPosBuf_.reserve(posCount_);
+        newVoteBuf_.reserve(posCount_);
+
+        const int cellSize = cvRound(minDist_);
+        const int gridWidth = (imageSize_.width + cellSize - 1) / cellSize;
+        const int gridHeight = (imageSize_.height + cellSize - 1) / cellSize;
+
+        std::vector< std::vector<Point2f> > grid(gridWidth * gridHeight);
+
+        const double minDist2 = minDist_ * minDist_;
+
+        for (int i = 0; i < posCount_; ++i)
+        {
+            const int ind = indexies_[i];
+
+            Point2f p(oldPosBuf_[ind].x, oldPosBuf_[ind].y);
+
+            bool good = true;
+
+            const int xCell = static_cast<int>(p.x / cellSize);
+            const int yCell = static_cast<int>(p.y / cellSize);
+
+            int x1 = xCell - 1;
+            int y1 = yCell - 1;
+            int x2 = xCell + 1;
+            int y2 = yCell + 1;
+
+            // boundary check
+            x1 = std::max(0, x1);
+            y1 = std::max(0, y1);
+            x2 = std::min(gridWidth - 1, x2);
+            y2 = std::min(gridHeight - 1, y2);
+
+            for (int yy = y1; yy <= y2; ++yy)
+            {
+                for (int xx = x1; xx <= x2; ++xx)
+                {
+                    const std::vector<Point2f>& m = grid[yy * gridWidth + xx];
+
+                    for(size_t j = 0; j < m.size(); ++j)
+                    {
+                        const Point2f d = p - m[j];
+
+                        if (d.ddot(d) < minDist2)
+                        {
+                            good = false;
+                            goto break_out;
+                        }
+                    }
+                }
+            }
+
+            break_out:
+
+            if(good)
+            {
+                grid[yCell * gridWidth + xCell].push_back(p);
+
+                newPosBuf_.push_back(oldPosBuf_[ind]);
+                newVoteBuf_.push_back(oldVoteBuf_[ind]);
+            }
+        }
+
+        posCount_ = static_cast<int>(newPosBuf_.size());
+        cudaSafeCall( cudaMemcpy(outBuf_.ptr(0), &newPosBuf_[0], posCount_ * sizeof(float4), cudaMemcpyHostToDevice) );
+        cudaSafeCall( cudaMemcpy(outBuf_.ptr(1), &newVoteBuf_[0], posCount_ * sizeof(int3), cudaMemcpyHostToDevice) );
+    }
+
+    void GeneralizedHoughBase::convertTo(OutputArray positions, OutputArray votes)
+    {
+        ensureSizeIsEnough(1, posCount_, CV_32FC4, positions);
+        GpuMat(1, posCount_, CV_32FC4, outBuf_.ptr(0), outBuf_.step).copyTo(positions);
+
+        if (votes.needed())
+        {
+            ensureSizeIsEnough(1, posCount_, CV_32FC3, votes);
+            GpuMat(1, posCount_, CV_32FC4, outBuf_.ptr(1), outBuf_.step).copyTo(votes);
+        }
+    }
+}
+
+// GeneralizedHoughBallard
+
+namespace
+{
+    class GeneralizedHoughBallardImpl : public GeneralizedHoughBallard, private GeneralizedHoughBase
+    {
+    public:
+        GeneralizedHoughBallardImpl();
+
+        void setTemplate(InputArray templ, Point templCenter) { setTemplateImpl(templ, templCenter); }
+        void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter) { setTemplateImpl(edges, dx, dy, templCenter); }
+
+        void detect(InputArray image, OutputArray positions, OutputArray votes) { detectImpl(image, positions, votes); }
+        void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes) { detectImpl(edges, dx, dy, positions, votes); }
+
+        void setCannyLowThresh(int cannyLowThresh) { cannyLowThresh_ = cannyLowThresh; }
+        int getCannyLowThresh() const { return cannyLowThresh_; }
+
+        void setCannyHighThresh(int cannyHighThresh) { cannyHighThresh_ = cannyHighThresh; }
+        int getCannyHighThresh() const { return cannyHighThresh_; }
+
+        void setMinDist(double minDist) { minDist_ = minDist; }
+        double getMinDist() const { return minDist_; }
+
+        void setDp(double dp) { dp_ = dp; }
+        double getDp() const { return dp_; }
+
+        void setMaxBufferSize(int maxBufferSize) { maxBufferSize_ = maxBufferSize; }
+        int getMaxBufferSize() const { return maxBufferSize_; }
+
+        void setLevels(int levels) { levels_ = levels; }
+        int getLevels() const { return levels_; }
+
+        void setVotesThreshold(int votesThreshold) { votesThreshold_ = votesThreshold; }
+        int getVotesThreshold() const { return votesThreshold_; }
+
+    private:
+        void processTempl();
+        void processImage();
+
+        void calcHist();
+        void findPosInHist();
+
+        int levels_;
+        int votesThreshold_;
+
+        GpuMat r_table_;
+        GpuMat r_sizes_;
+
+        GpuMat hist_;
+    };
+
+    GeneralizedHoughBallardImpl::GeneralizedHoughBallardImpl()
+    {
+        levels_ = 360;
+        votesThreshold_ = 100;
+    }
+
+    void GeneralizedHoughBallardImpl::processTempl()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        CV_Assert( levels_ > 0 );
+
+        buildEdgePointList(templEdges_, templDx_, templDy_);
+
+        ensureSizeIsEnough(levels_ + 1, maxBufferSize_, CV_16SC2, r_table_);
+        ensureSizeIsEnough(1, levels_ + 1, CV_32SC1, r_sizes_);
+        r_sizes_.setTo(Scalar::all(0));
+
+        if (edgePointList_.cols > 0)
+        {
+            buildRTable_gpu(edgePointList_.ptr<unsigned int>(0), edgePointList_.ptr<float>(1), edgePointList_.cols,
+                            r_table_, r_sizes_.ptr<int>(), make_short2(templCenter_.x, templCenter_.y), levels_);
+            gpu::min(r_sizes_, maxBufferSize_, r_sizes_);
+        }
+    }
+
+    void GeneralizedHoughBallardImpl::processImage()
+    {
+        calcHist();
+        findPosInHist();
+    }
+
+    void GeneralizedHoughBallardImpl::calcHist()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        CV_Assert( levels_ > 0 && r_table_.rows == (levels_ + 1) && r_sizes_.cols == (levels_ + 1) );
+        CV_Assert( dp_ > 0.0);
+
+        const double idp = 1.0 / dp_;
+
+        buildEdgePointList(imageEdges_, imageDx_, imageDy_);
+
+        ensureSizeIsEnough(cvCeil(imageSize_.height * idp) + 2, cvCeil(imageSize_.width * idp) + 2, CV_32SC1, hist_);
+        hist_.setTo(Scalar::all(0));
+
+        if (edgePointList_.cols > 0)
+        {
+            Ballard_Pos_calcHist_gpu(edgePointList_.ptr<unsigned int>(0), edgePointList_.ptr<float>(1), edgePointList_.cols,
+                                     r_table_, r_sizes_.ptr<int>(),
+                                     hist_,
+                                     (float)dp_, levels_);
+        }
+    }
+
+    void GeneralizedHoughBallardImpl::findPosInHist()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        CV_Assert( votesThreshold_ > 0 );
+
+        ensureSizeIsEnough(2, maxBufferSize_, CV_32FC4, outBuf_);
+
+        posCount_ = Ballard_Pos_findPosInHist_gpu(hist_, outBuf_.ptr<float4>(0), outBuf_.ptr<int3>(1), maxBufferSize_, (float)dp_, votesThreshold_);
+    }
+}
+
+Ptr<GeneralizedHoughBallard> cv::gpu::createGeneralizedHoughBallard()
+{
+    return makePtr<GeneralizedHoughBallardImpl>();
+}
+
+// GeneralizedHoughGuil
+
+namespace
+{
+    class GeneralizedHoughGuilImpl : public GeneralizedHoughGuil, private GeneralizedHoughBase
+    {
+    public:
+        GeneralizedHoughGuilImpl();
+
+        void setTemplate(InputArray templ, Point templCenter) { setTemplateImpl(templ, templCenter); }
+        void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter) { setTemplateImpl(edges, dx, dy, templCenter); }
+
+        void detect(InputArray image, OutputArray positions, OutputArray votes) { detectImpl(image, positions, votes); }
+        void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes) { detectImpl(edges, dx, dy, positions, votes); }
+
+        void setCannyLowThresh(int cannyLowThresh) { cannyLowThresh_ = cannyLowThresh; }
+        int getCannyLowThresh() const { return cannyLowThresh_; }
+
+        void setCannyHighThresh(int cannyHighThresh) { cannyHighThresh_ = cannyHighThresh; }
+        int getCannyHighThresh() const { return cannyHighThresh_; }
+
+        void setMinDist(double minDist) { minDist_ = minDist; }
+        double getMinDist() const { return minDist_; }
+
+        void setDp(double dp) { dp_ = dp; }
+        double getDp() const { return dp_; }
+
+        void setMaxBufferSize(int maxBufferSize) { maxBufferSize_ = maxBufferSize; }
+        int getMaxBufferSize() const { return maxBufferSize_; }
+
+        void setXi(double xi) { xi_ = xi; }
+        double getXi() const { return xi_; }
+
+        void setLevels(int levels) { levels_ = levels; }
+        int getLevels() const { return levels_; }
+
+        void setAngleEpsilon(double angleEpsilon) { angleEpsilon_ = angleEpsilon; }
+        double getAngleEpsilon() const { return angleEpsilon_; }
+
+        void setMinAngle(double minAngle) { minAngle_ = minAngle; }
+        double getMinAngle() const { return minAngle_; }
+
+        void setMaxAngle(double maxAngle) { maxAngle_ = maxAngle; }
+        double getMaxAngle() const { return maxAngle_; }
+
+        void setAngleStep(double angleStep) { angleStep_ = angleStep; }
+        double getAngleStep() const { return angleStep_; }
+
+        void setAngleThresh(int angleThresh) { angleThresh_ = angleThresh; }
+        int getAngleThresh() const { return angleThresh_; }
+
+        void setMinScale(double minScale) { minScale_ = minScale; }
+        double getMinScale() const { return minScale_; }
+
+        void setMaxScale(double maxScale) { maxScale_ = maxScale; }
+        double getMaxScale() const { return maxScale_; }
+
+        void setScaleStep(double scaleStep) { scaleStep_ = scaleStep; }
+        double getScaleStep() const { return scaleStep_; }
+
+        void setScaleThresh(int scaleThresh) { scaleThresh_ = scaleThresh; }
+        int getScaleThresh() const { return scaleThresh_; }
+
+        void setPosThresh(int posThresh) { posThresh_ = posThresh; }
+        int getPosThresh() const { return posThresh_; }
+
+    private:
+        void processTempl();
+        void processImage();
+
+        double xi_;
+        int levels_;
+        double angleEpsilon_;
+
+        double minAngle_;
+        double maxAngle_;
+        double angleStep_;
+        int angleThresh_;
+
+        double minScale_;
+        double maxScale_;
+        double scaleStep_;
+        int scaleThresh_;
+
+        int posThresh_;
+
+        struct Feature
+        {
+            GpuMat p1_pos;
+            GpuMat p1_theta;
+            GpuMat p2_pos;
+
+            GpuMat d12;
+
+            GpuMat r1;
+            GpuMat r2;
+
+            GpuMat sizes;
+            int maxSize;
+
+            void create(int levels, int maxCapacity, bool isTempl);
+        };
+
+        typedef void (*set_func_t)(PtrStepb p1_pos, PtrStepb p1_theta, PtrStepb p2_pos, PtrStepb d12, PtrStepb r1, PtrStepb r2);
+        typedef void (*build_func_t)(const unsigned int* coordList, const float* thetaList, int pointsCount,
+                                     int* sizes, int maxSize,
+                                     float xi, float angleEpsilon, int levels,
+                                     float2 center, float maxDist);
+
+        void buildFeatureList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Feature& features,
+                              set_func_t set_func, build_func_t build_func, bool isTempl, Point2d center = Point2d());
+
+        void calcOrientation();
+        void calcScale(double angle);
+        void calcPosition(double angle, int angleVotes, double scale, int scaleVotes);
+
+        Feature templFeatures_;
+        Feature imageFeatures_;
+
+        std::vector< std::pair<double, int> > angles_;
+        std::vector< std::pair<double, int> > scales_;
+
+        GpuMat hist_;
+        std::vector<int> h_buf_;
+    };
+
+    double toRad(double a)
+    {
+        return a * CV_PI / 180.0;
+    }
+
+    double clampAngle(double a)
+    {
+        double res = a;
+
+        while (res > 360.0)
+            res -= 360.0;
+        while (res < 0)
+            res += 360.0;
+
+        return res;
+    }
+
+    bool angleEq(double a, double b, double eps = 1.0)
+    {
+        return (fabs(clampAngle(a - b)) <= eps);
+    }
+
+    GeneralizedHoughGuilImpl::GeneralizedHoughGuilImpl()
+    {
+        maxBufferSize_ = 1000;
+
+        xi_ = 90.0;
+        levels_ = 360;
+        angleEpsilon_ = 1.0;
+
+        minAngle_ = 0.0;
+        maxAngle_ = 360.0;
+        angleStep_ = 1.0;
+        angleThresh_ = 15000;
+
+        minScale_ = 0.5;
+        maxScale_ = 2.0;
+        scaleStep_ = 0.05;
+        scaleThresh_ = 1000;
+
+        posThresh_ = 100;
+    }
+
+    void GeneralizedHoughGuilImpl::processTempl()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        buildFeatureList(templEdges_, templDx_, templDy_, templFeatures_,
+            Guil_Full_setTemplFeatures, Guil_Full_buildTemplFeatureList_gpu,
+            true, templCenter_);
+
+        h_buf_.resize(templFeatures_.sizes.cols);
+        cudaSafeCall( cudaMemcpy(&h_buf_[0], templFeatures_.sizes.data, h_buf_.size() * sizeof(int), cudaMemcpyDeviceToHost) );
+        templFeatures_.maxSize = *std::max_element(h_buf_.begin(), h_buf_.end());
+    }
+
+    void GeneralizedHoughGuilImpl::processImage()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        CV_Assert( levels_ > 0 );
+        CV_Assert( templFeatures_.sizes.cols == levels_ + 1 );
+        CV_Assert( minAngle_ >= 0.0 && minAngle_ < maxAngle_ && maxAngle_ <= 360.0 );
+        CV_Assert( angleStep_ > 0.0 && angleStep_ < 360.0 );
+        CV_Assert( angleThresh_ > 0 );
+        CV_Assert( minScale_ > 0.0 && minScale_ < maxScale_ );
+        CV_Assert( scaleStep_ > 0.0 );
+        CV_Assert( scaleThresh_ > 0 );
+        CV_Assert( dp_ > 0.0 );
+        CV_Assert( posThresh_ > 0 );
+
+        const double iAngleStep = 1.0 / angleStep_;
+        const int angleRange = cvCeil((maxAngle_ - minAngle_) * iAngleStep);
+
+        const double iScaleStep = 1.0 / scaleStep_;
+        const int scaleRange = cvCeil((maxScale_ - minScale_) * iScaleStep);
+
+        const double idp = 1.0 / dp_;
+        const int histRows = cvCeil(imageSize_.height * idp);
+        const int histCols = cvCeil(imageSize_.width * idp);
+
+        ensureSizeIsEnough(histRows + 2, std::max(angleRange + 1, std::max(scaleRange + 1, histCols + 2)), CV_32SC1, hist_);
+        h_buf_.resize(std::max(angleRange + 1, scaleRange + 1));
+
+        ensureSizeIsEnough(2, maxBufferSize_, CV_32FC4, outBuf_);
+
+        buildFeatureList(imageEdges_, imageDx_, imageDy_, imageFeatures_,
+            Guil_Full_setImageFeatures, Guil_Full_buildImageFeatureList_gpu,
+            false);
+
+        calcOrientation();
+
+        for (size_t i = 0; i < angles_.size(); ++i)
+        {
+            const double angle = angles_[i].first;
+            const int angleVotes = angles_[i].second;
+
+            calcScale(angle);
+
+            for (size_t j = 0; j < scales_.size(); ++j)
+            {
+                const double scale = scales_[j].first;
+                const int scaleVotes = scales_[j].second;
+
+                calcPosition(angle, angleVotes, scale, scaleVotes);
+            }
+        }
+    }
+
+    void GeneralizedHoughGuilImpl::Feature::create(int levels, int maxCapacity, bool isTempl)
+    {
+        if (!isTempl)
+        {
+            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, p1_pos);
+            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, p2_pos);
+        }
+
+        ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC1, p1_theta);
+
+        ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC1, d12);
+
+        if (isTempl)
+        {
+            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, r1);
+            ensureSizeIsEnough(levels + 1, maxCapacity, CV_32FC2, r2);
+        }
+
+        ensureSizeIsEnough(1, levels + 1, CV_32SC1, sizes);
+        sizes.setTo(Scalar::all(0));
+
+        maxSize = 0;
+    }
+
+    void GeneralizedHoughGuilImpl::buildFeatureList(const GpuMat& edges, const GpuMat& dx, const GpuMat& dy, Feature& features,
+                                                    set_func_t set_func, build_func_t build_func, bool isTempl, Point2d center)
+    {
+        CV_Assert( levels_ > 0 );
+
+        const double maxDist = sqrt((double) templSize_.width * templSize_.width + templSize_.height * templSize_.height) * maxScale_;
+
+        features.create(levels_, maxBufferSize_, isTempl);
+        set_func(features.p1_pos, features.p1_theta, features.p2_pos, features.d12, features.r1, features.r2);
+
+        buildEdgePointList(edges, dx, dy);
+
+        if (edgePointList_.cols > 0)
+        {
+            build_func(edgePointList_.ptr<unsigned int>(0), edgePointList_.ptr<float>(1), edgePointList_.cols,
+                features.sizes.ptr<int>(), maxBufferSize_, (float)xi_, (float)angleEpsilon_, levels_, make_float2((float)center.x, (float)center.y), (float)maxDist);
+        }
+    }
+
+    void GeneralizedHoughGuilImpl::calcOrientation()
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        const double iAngleStep = 1.0 / angleStep_;
+        const int angleRange = cvCeil((maxAngle_ - minAngle_) * iAngleStep);
+
+        hist_.setTo(Scalar::all(0));
+        Guil_Full_calcOHist_gpu(templFeatures_.sizes.ptr<int>(), imageFeatures_.sizes.ptr<int>(0), hist_.ptr<int>(),
+                                (float)minAngle_, (float)maxAngle_, (float)angleStep_, angleRange, levels_, templFeatures_.maxSize);
+        cudaSafeCall( cudaMemcpy(&h_buf_[0], hist_.data, h_buf_.size() * sizeof(int), cudaMemcpyDeviceToHost) );
+
+        angles_.clear();
+
+        for (int n = 0; n < angleRange; ++n)
+        {
+            if (h_buf_[n] >= angleThresh_)
+            {
+                const double angle = minAngle_ + n * angleStep_;
+                angles_.push_back(std::make_pair(angle, h_buf_[n]));
+            }
+        }
+    }
+
+    void GeneralizedHoughGuilImpl::calcScale(double angle)
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        const double iScaleStep = 1.0 / scaleStep_;
+        const int scaleRange = cvCeil((maxScale_ - minScale_) * iScaleStep);
+
+        hist_.setTo(Scalar::all(0));
+        Guil_Full_calcSHist_gpu(templFeatures_.sizes.ptr<int>(), imageFeatures_.sizes.ptr<int>(0), hist_.ptr<int>(),
+                                (float)angle, (float)angleEpsilon_, (float)minScale_, (float)maxScale_,
+                                (float)iScaleStep, scaleRange, levels_, templFeatures_.maxSize);
+        cudaSafeCall( cudaMemcpy(&h_buf_[0], hist_.data, h_buf_.size() * sizeof(int), cudaMemcpyDeviceToHost) );
+
+        scales_.clear();
+
+        for (int s = 0; s < scaleRange; ++s)
+        {
+            if (h_buf_[s] >= scaleThresh_)
+            {
+                const double scale = minScale_ + s * scaleStep_;
+                scales_.push_back(std::make_pair(scale, h_buf_[s]));
+            }
+        }
+    }
+
+    void GeneralizedHoughGuilImpl::calcPosition(double angle, int angleVotes, double scale, int scaleVotes)
+    {
+        using namespace cv::gpu::cudev::ght;
+
+        hist_.setTo(Scalar::all(0));
+        Guil_Full_calcPHist_gpu(templFeatures_.sizes.ptr<int>(), imageFeatures_.sizes.ptr<int>(0), hist_,
+                                (float)angle, (float)angleEpsilon_, (float)scale, (float)dp_, levels_, templFeatures_.maxSize);
+
+        posCount_ = Guil_Full_findPosInHist_gpu(hist_, outBuf_.ptr<float4>(0), outBuf_.ptr<int3>(1),
+                                                posCount_, maxBufferSize_, (float)angle, angleVotes,
+                                                (float)scale, scaleVotes, (float)dp_, posThresh_);
+    }
+}
+
+Ptr<GeneralizedHoughGuil> cv::gpu::createGeneralizedHoughGuil()
+{
+    return makePtr<GeneralizedHoughGuilImpl>();
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/gftt.cpp b/modules/gpuimgproc/src/gftt.cpp
new file mode 100644
index 0000000000..0446b6cdbb
--- /dev/null
+++ b/modules/gpuimgproc/src/gftt.cpp
@@ -0,0 +1,216 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) || !defined(HAVE_OPENCV_GPUARITHM)
+
+Ptr<gpu::CornersDetector> cv::gpu::createGoodFeaturesToTrackDetector(int, int, double, double, int, bool, double) { throw_no_cuda(); return Ptr<gpu::CornersDetector>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace gfft
+    {
+        int findCorners_gpu(PtrStepSzf eig, float threshold, PtrStepSzb mask, float2* corners, int max_count);
+        void sortCorners_gpu(PtrStepSzf eig, float2* corners, int count);
+    }
+}}}
+
+namespace
+{
+    class GoodFeaturesToTrackDetector : public CornersDetector
+    {
+    public:
+        GoodFeaturesToTrackDetector(int srcType, int maxCorners, double qualityLevel, double minDistance,
+                                    int blockSize, bool useHarrisDetector, double harrisK);
+
+        void detect(InputArray image, OutputArray corners, InputArray mask = noArray());
+
+    private:
+        int maxCorners_;
+        double qualityLevel_;
+        double minDistance_;
+
+        Ptr<gpu::CornernessCriteria> cornerCriteria_;
+
+        GpuMat Dx_;
+        GpuMat Dy_;
+        GpuMat buf_;
+        GpuMat eig_;
+        GpuMat minMaxbuf_;
+        GpuMat tmpCorners_;
+    };
+
+    GoodFeaturesToTrackDetector::GoodFeaturesToTrackDetector(int srcType, int maxCorners, double qualityLevel, double minDistance,
+                                                             int blockSize, bool useHarrisDetector, double harrisK) :
+        maxCorners_(maxCorners), qualityLevel_(qualityLevel), minDistance_(minDistance)
+    {
+        CV_Assert( qualityLevel_ > 0 && minDistance_ >= 0 && maxCorners_ >= 0 );
+
+        cornerCriteria_ = useHarrisDetector ?
+                    gpu::createHarrisCorner(srcType, blockSize, 3, harrisK) :
+                    gpu::createMinEigenValCorner(srcType, blockSize, 3);
+    }
+
+    void GoodFeaturesToTrackDetector::detect(InputArray _image, OutputArray _corners, InputArray _mask)
+    {
+        using namespace cv::gpu::cudev::gfft;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat mask = _mask.getGpuMat();
+
+        CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == image.size()) );
+
+        ensureSizeIsEnough(image.size(), CV_32FC1, eig_);
+        cornerCriteria_->compute(image, eig_);
+
+        double maxVal = 0;
+        gpu::minMax(eig_, 0, &maxVal, noArray(), minMaxbuf_);
+
+        ensureSizeIsEnough(1, std::max(1000, static_cast<int>(image.size().area() * 0.05)), CV_32FC2, tmpCorners_);
+
+        int total = findCorners_gpu(eig_, static_cast<float>(maxVal * qualityLevel_), mask, tmpCorners_.ptr<float2>(), tmpCorners_.cols);
+
+        if (total == 0)
+        {
+            _corners.release();
+            return;
+        }
+
+        sortCorners_gpu(eig_, tmpCorners_.ptr<float2>(), total);
+
+        if (minDistance_ < 1)
+        {
+            tmpCorners_.colRange(0, maxCorners_ > 0 ? std::min(maxCorners_, total) : total).copyTo(_corners);
+        }
+        else
+        {
+            std::vector<Point2f> tmp(total);
+            Mat tmpMat(1, total, CV_32FC2, (void*)&tmp[0]);
+            tmpCorners_.colRange(0, total).download(tmpMat);
+
+            std::vector<Point2f> tmp2;
+            tmp2.reserve(total);
+
+            const int cell_size = cvRound(minDistance_);
+            const int grid_width = (image.cols + cell_size - 1) / cell_size;
+            const int grid_height = (image.rows + cell_size - 1) / cell_size;
+
+            std::vector< std::vector<Point2f> > grid(grid_width * grid_height);
+
+            for (int i = 0; i < total; ++i)
+            {
+                Point2f p = tmp[i];
+
+                bool good = true;
+
+                int x_cell = static_cast<int>(p.x / cell_size);
+                int y_cell = static_cast<int>(p.y / cell_size);
+
+                int x1 = x_cell - 1;
+                int y1 = y_cell - 1;
+                int x2 = x_cell + 1;
+                int y2 = y_cell + 1;
+
+                // boundary check
+                x1 = std::max(0, x1);
+                y1 = std::max(0, y1);
+                x2 = std::min(grid_width - 1, x2);
+                y2 = std::min(grid_height - 1, y2);
+
+                for (int yy = y1; yy <= y2; yy++)
+                {
+                    for (int xx = x1; xx <= x2; xx++)
+                    {
+                        std::vector<Point2f>& m = grid[yy * grid_width + xx];
+
+                        if (!m.empty())
+                        {
+                            for(size_t j = 0; j < m.size(); j++)
+                            {
+                                float dx = p.x - m[j].x;
+                                float dy = p.y - m[j].y;
+
+                                if (dx * dx + dy * dy < minDistance_ * minDistance_)
+                                {
+                                    good = false;
+                                    goto break_out;
+                                }
+                            }
+                        }
+                    }
+                }
+
+                break_out:
+
+                if(good)
+                {
+                    grid[y_cell * grid_width + x_cell].push_back(p);
+
+                    tmp2.push_back(p);
+
+                    if (maxCorners_ > 0 && tmp2.size() == static_cast<size_t>(maxCorners_))
+                        break;
+                }
+            }
+
+            _corners.create(1, static_cast<int>(tmp2.size()), CV_32FC2);
+            GpuMat corners = _corners.getGpuMat();
+
+            corners.upload(Mat(1, static_cast<int>(tmp2.size()), CV_32FC2, &tmp2[0]));
+        }
+    }
+}
+
+Ptr<gpu::CornersDetector> cv::gpu::createGoodFeaturesToTrackDetector(int srcType, int maxCorners, double qualityLevel, double minDistance,
+                                                                     int blockSize, bool useHarrisDetector, double harrisK)
+{
+    return Ptr<gpu::CornersDetector>(
+        new GoodFeaturesToTrackDetector(srcType, maxCorners, qualityLevel, minDistance, blockSize, useHarrisDetector, harrisK));
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/histogram.cpp b/modules/gpuimgproc/src/histogram.cpp
new file mode 100644
index 0000000000..eb78ee0deb
--- /dev/null
+++ b/modules/gpuimgproc/src/histogram.cpp
@@ -0,0 +1,579 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::calcHist(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::equalizeHist(InputArray, OutputArray, InputOutputArray, Stream&) { throw_no_cuda(); }
+
+cv::Ptr<cv::gpu::CLAHE> cv::gpu::createCLAHE(double, cv::Size) { throw_no_cuda(); return cv::Ptr<cv::gpu::CLAHE>(); }
+
+void cv::gpu::evenLevels(OutputArray, int, int, int) { throw_no_cuda(); }
+
+void cv::gpu::histEven(InputArray, OutputArray, InputOutputArray, int, int, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::histEven(InputArray, GpuMat*, InputOutputArray, int*, int*, int*, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::histRange(InputArray, OutputArray, InputArray, InputOutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::histRange(InputArray, GpuMat*, const GpuMat*, InputOutputArray, Stream&) { throw_no_cuda(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+////////////////////////////////////////////////////////////////////////
+// calcHist
+
+namespace hist
+{
+    void histogram256(PtrStepSzb src, int* hist, cudaStream_t stream);
+}
+
+void cv::gpu::calcHist(InputArray _src, OutputArray _hist, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC1 );
+
+    _hist.create(1, 256, CV_32SC1);
+    GpuMat hist = _hist.getGpuMat();
+
+    hist.setTo(Scalar::all(0), stream);
+
+    hist::histogram256(src, hist.ptr<int>(), StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// equalizeHist
+
+namespace hist
+{
+    void equalizeHist(PtrStepSzb src, PtrStepSzb dst, const int* lut, cudaStream_t stream);
+}
+
+void cv::gpu::equalizeHist(InputArray _src, OutputArray _dst, InputOutputArray _buf, Stream& _stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC1 );
+
+    _dst.create(src.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    int intBufSize;
+    nppSafeCall( nppsIntegralGetBufferSize_32s(256, &intBufSize) );
+
+    size_t bufSize = intBufSize + 2 * 256 * sizeof(int);
+
+    ensureSizeIsEnough(1, static_cast<int>(bufSize), CV_8UC1, _buf);
+    GpuMat buf = _buf.getGpuMat();
+
+    GpuMat hist(1, 256, CV_32SC1, buf.data);
+    GpuMat lut(1, 256, CV_32SC1, buf.data + 256 * sizeof(int));
+    GpuMat intBuf(1, intBufSize, CV_8UC1, buf.data + 2 * 256 * sizeof(int));
+
+    gpu::calcHist(src, hist, _stream);
+
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
+    NppStreamHandler h(stream);
+
+    nppSafeCall( nppsIntegral_32s(hist.ptr<Npp32s>(), lut.ptr<Npp32s>(), 256, intBuf.ptr<Npp8u>()) );
+
+    hist::equalizeHist(src, dst, lut.ptr<int>(), stream);
+}
+
+////////////////////////////////////////////////////////////////////////
+// CLAHE
+
+namespace clahe
+{
+    void calcLut(PtrStepSzb src, PtrStepb lut, int tilesX, int tilesY, int2 tileSize, int clipLimit, float lutScale, cudaStream_t stream);
+    void transform(PtrStepSzb src, PtrStepSzb dst, PtrStepb lut, int tilesX, int tilesY, int2 tileSize, cudaStream_t stream);
+}
+
+namespace
+{
+    class CLAHE_Impl : public cv::gpu::CLAHE
+    {
+    public:
+        CLAHE_Impl(double clipLimit = 40.0, int tilesX = 8, int tilesY = 8);
+
+        cv::AlgorithmInfo* info() const;
+
+        void apply(cv::InputArray src, cv::OutputArray dst);
+        void apply(InputArray src, OutputArray dst, Stream& stream);
+
+        void setClipLimit(double clipLimit);
+        double getClipLimit() const;
+
+        void setTilesGridSize(cv::Size tileGridSize);
+        cv::Size getTilesGridSize() const;
+
+        void collectGarbage();
+
+    private:
+        double clipLimit_;
+        int tilesX_;
+        int tilesY_;
+
+        GpuMat srcExt_;
+        GpuMat lut_;
+    };
+
+    CLAHE_Impl::CLAHE_Impl(double clipLimit, int tilesX, int tilesY) :
+        clipLimit_(clipLimit), tilesX_(tilesX), tilesY_(tilesY)
+    {
+    }
+
+    CV_INIT_ALGORITHM(CLAHE_Impl, "CLAHE_GPU",
+        obj.info()->addParam(obj, "clipLimit", obj.clipLimit_);
+        obj.info()->addParam(obj, "tilesX", obj.tilesX_);
+        obj.info()->addParam(obj, "tilesY", obj.tilesY_))
+
+    void CLAHE_Impl::apply(cv::InputArray _src, cv::OutputArray _dst)
+    {
+        apply(_src, _dst, Stream::Null());
+    }
+
+    void CLAHE_Impl::apply(InputArray _src, OutputArray _dst, Stream& s)
+    {
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+
+        _dst.create( src.size(), src.type() );
+        GpuMat dst = _dst.getGpuMat();
+
+        const int histSize = 256;
+
+        ensureSizeIsEnough(tilesX_ * tilesY_, histSize, CV_8UC1, lut_);
+
+        cudaStream_t stream = StreamAccessor::getStream(s);
+
+        cv::Size tileSize;
+        GpuMat srcForLut;
+
+        if (src.cols % tilesX_ == 0 && src.rows % tilesY_ == 0)
+        {
+            tileSize = cv::Size(src.cols / tilesX_, src.rows / tilesY_);
+            srcForLut = src;
+        }
+        else
+        {
+#ifndef HAVE_OPENCV_GPUARITHM
+            throw_no_cuda();
+#else
+            cv::gpu::copyMakeBorder(src, srcExt_, 0, tilesY_ - (src.rows % tilesY_), 0, tilesX_ - (src.cols % tilesX_), cv::BORDER_REFLECT_101, cv::Scalar(), s);
+#endif
+
+            tileSize = cv::Size(srcExt_.cols / tilesX_, srcExt_.rows / tilesY_);
+            srcForLut = srcExt_;
+        }
+
+        const int tileSizeTotal = tileSize.area();
+        const float lutScale = static_cast<float>(histSize - 1) / tileSizeTotal;
+
+        int clipLimit = 0;
+        if (clipLimit_ > 0.0)
+        {
+            clipLimit = static_cast<int>(clipLimit_ * tileSizeTotal / histSize);
+            clipLimit = std::max(clipLimit, 1);
+        }
+
+        clahe::calcLut(srcForLut, lut_, tilesX_, tilesY_, make_int2(tileSize.width, tileSize.height), clipLimit, lutScale, stream);
+
+        clahe::transform(src, dst, lut_, tilesX_, tilesY_, make_int2(tileSize.width, tileSize.height), stream);
+    }
+
+    void CLAHE_Impl::setClipLimit(double clipLimit)
+    {
+        clipLimit_ = clipLimit;
+    }
+
+    double CLAHE_Impl::getClipLimit() const
+    {
+        return clipLimit_;
+    }
+
+    void CLAHE_Impl::setTilesGridSize(cv::Size tileGridSize)
+    {
+        tilesX_ = tileGridSize.width;
+        tilesY_ = tileGridSize.height;
+    }
+
+    cv::Size CLAHE_Impl::getTilesGridSize() const
+    {
+        return cv::Size(tilesX_, tilesY_);
+    }
+
+    void CLAHE_Impl::collectGarbage()
+    {
+        srcExt_.release();
+        lut_.release();
+    }
+}
+
+cv::Ptr<cv::gpu::CLAHE> cv::gpu::createCLAHE(double clipLimit, cv::Size tileGridSize)
+{
+    return makePtr<CLAHE_Impl>(clipLimit, tileGridSize.width, tileGridSize.height);
+}
+
+////////////////////////////////////////////////////////////////////////
+// NPP Histogram
+
+namespace
+{
+    typedef NppStatus (*get_buf_size_c1_t)(NppiSize oSizeROI, int nLevels, int* hpBufferSize);
+    typedef NppStatus (*get_buf_size_c4_t)(NppiSize oSizeROI, int nLevels[], int* hpBufferSize);
+
+    template<int SDEPTH> struct NppHistogramEvenFuncC1
+    {
+        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
+
+    typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s * pHist,
+            int nLevels, Npp32s nLowerLevel, Npp32s nUpperLevel, Npp8u * pBuffer);
+    };
+    template<int SDEPTH> struct NppHistogramEvenFuncC4
+    {
+        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
+
+        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI,
+            Npp32s * pHist[4], int nLevels[4], Npp32s nLowerLevel[4], Npp32s nUpperLevel[4], Npp8u * pBuffer);
+    };
+
+    template<int SDEPTH, typename NppHistogramEvenFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
+    struct NppHistogramEvenC1
+    {
+        typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;
+
+        static void hist(const GpuMat& src, OutputArray _hist, InputOutputArray _buf, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)
+        {
+            const int levels = histSize + 1;
+
+            _hist.create(1, histSize, CV_32S);
+            GpuMat hist = _hist.getGpuMat();
+
+            NppiSize sz;
+            sz.width = src.cols;
+            sz.height = src.rows;
+
+            int buf_size;
+            get_buf_size(sz, levels, &buf_size);
+
+            ensureSizeIsEnough(1, buf_size, CV_8UC1, _buf);
+            GpuMat buf = _buf.getGpuMat();
+
+            NppStreamHandler h(stream);
+
+            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels,
+                lowerLevel, upperLevel, buf.ptr<Npp8u>()) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+    template<int SDEPTH, typename NppHistogramEvenFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
+    struct NppHistogramEvenC4
+    {
+        typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;
+
+        static void hist(const GpuMat& src, GpuMat hist[4],InputOutputArray _buf, int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)
+        {
+            int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};
+            hist[0].create(1, histSize[0], CV_32S);
+            hist[1].create(1, histSize[1], CV_32S);
+            hist[2].create(1, histSize[2], CV_32S);
+            hist[3].create(1, histSize[3], CV_32S);
+
+            NppiSize sz;
+            sz.width = src.cols;
+            sz.height = src.rows;
+
+            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
+
+            int buf_size;
+            get_buf_size(sz, levels, &buf_size);
+
+            ensureSizeIsEnough(1, buf_size, CV_8U, _buf);
+            GpuMat buf = _buf.getGpuMat();
+
+            NppStreamHandler h(stream);
+
+            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, levels, lowerLevel, upperLevel, buf.ptr<Npp8u>()) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+
+    template<int SDEPTH> struct NppHistogramRangeFuncC1
+    {
+        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
+        typedef Npp32s level_t;
+        enum {LEVEL_TYPE_CODE=CV_32SC1};
+
+        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
+            const Npp32s* pLevels, int nLevels, Npp8u* pBuffer);
+    };
+    template<> struct NppHistogramRangeFuncC1<CV_32F>
+    {
+        typedef Npp32f src_t;
+        typedef Npp32f level_t;
+        enum {LEVEL_TYPE_CODE=CV_32FC1};
+
+        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
+            const Npp32f* pLevels, int nLevels, Npp8u* pBuffer);
+    };
+    template<int SDEPTH> struct NppHistogramRangeFuncC4
+    {
+        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
+        typedef Npp32s level_t;
+        enum {LEVEL_TYPE_CODE=CV_32SC1};
+
+        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
+            const Npp32s* pLevels[4], int nLevels[4], Npp8u* pBuffer);
+    };
+    template<> struct NppHistogramRangeFuncC4<CV_32F>
+    {
+        typedef Npp32f src_t;
+        typedef Npp32f level_t;
+        enum {LEVEL_TYPE_CODE=CV_32FC1};
+
+        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
+            const Npp32f* pLevels[4], int nLevels[4], Npp8u* pBuffer);
+    };
+
+    template<int SDEPTH, typename NppHistogramRangeFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
+    struct NppHistogramRangeC1
+    {
+        typedef typename NppHistogramRangeFuncC1<SDEPTH>::src_t src_t;
+        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
+        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
+
+        static void hist(const GpuMat& src, OutputArray _hist, const GpuMat& levels, InputOutputArray _buf, cudaStream_t stream)
+        {
+            CV_Assert( levels.type() == LEVEL_TYPE_CODE && levels.rows == 1 );
+
+            _hist.create(1, levels.cols - 1, CV_32S);
+            GpuMat hist = _hist.getGpuMat();
+
+            NppiSize sz;
+            sz.width = src.cols;
+            sz.height = src.rows;
+
+            int buf_size;
+            get_buf_size(sz, levels.cols, &buf_size);
+
+            ensureSizeIsEnough(1, buf_size, CV_8U, _buf);
+            GpuMat buf = _buf.getGpuMat();
+
+            NppStreamHandler h(stream);
+
+            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels.ptr<level_t>(), levels.cols, buf.ptr<Npp8u>()) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+    template<int SDEPTH, typename NppHistogramRangeFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
+    struct NppHistogramRangeC4
+    {
+        typedef typename NppHistogramRangeFuncC4<SDEPTH>::src_t src_t;
+        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
+        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
+
+        static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4],InputOutputArray _buf, cudaStream_t stream)
+        {
+            CV_Assert( levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1 );
+            CV_Assert( levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1 );
+            CV_Assert( levels[2].type() == LEVEL_TYPE_CODE && levels[2].rows == 1 );
+            CV_Assert( levels[3].type() == LEVEL_TYPE_CODE && levels[3].rows == 1 );
+
+            hist[0].create(1, levels[0].cols - 1, CV_32S);
+            hist[1].create(1, levels[1].cols - 1, CV_32S);
+            hist[2].create(1, levels[2].cols - 1, CV_32S);
+            hist[3].create(1, levels[3].cols - 1, CV_32S);
+
+            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
+            int nLevels[] = {levels[0].cols, levels[1].cols, levels[2].cols, levels[3].cols};
+            const level_t* pLevels[] = {levels[0].ptr<level_t>(), levels[1].ptr<level_t>(), levels[2].ptr<level_t>(), levels[3].ptr<level_t>()};
+
+            NppiSize sz;
+            sz.width = src.cols;
+            sz.height = src.rows;
+
+            int buf_size;
+            get_buf_size(sz, nLevels, &buf_size);
+
+            ensureSizeIsEnough(1, buf_size, CV_8U, _buf);
+            GpuMat buf = _buf.getGpuMat();
+
+            NppStreamHandler h(stream);
+
+            nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, pLevels, nLevels, buf.ptr<Npp8u>()) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+}
+
+void cv::gpu::evenLevels(OutputArray _levels, int nLevels, int lowerLevel, int upperLevel)
+{
+    const int kind = _levels.kind();
+
+    _levels.create(1, nLevels, CV_32SC1);
+
+    Mat host_levels;
+    if (kind == _InputArray::GPU_MAT)
+        host_levels.create(1, nLevels, CV_32SC1);
+    else
+        host_levels = _levels.getMat();
+
+    nppSafeCall( nppiEvenLevelsHost_32s(host_levels.ptr<Npp32s>(), nLevels, lowerLevel, upperLevel) );
+
+    if (kind == _InputArray::GPU_MAT)
+        _levels.getGpuMatRef().upload(host_levels);
+}
+
+namespace hist
+{
+    void histEven8u(PtrStepSzb src, int* hist, int binCount, int lowerLevel, int upperLevel, cudaStream_t stream);
+}
+
+namespace
+{
+    void histEven8u(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)
+    {
+        hist.create(1, histSize, CV_32S);
+        cudaSafeCall( cudaMemsetAsync(hist.data, 0, histSize * sizeof(int), stream) );
+        hist::histEven8u(src, hist.ptr<int>(), histSize, lowerLevel, upperLevel, stream);
+    }
+}
+
+void cv::gpu::histEven(InputArray _src, OutputArray hist, InputOutputArray buf, int histSize, int lowerLevel, int upperLevel, Stream& stream)
+{
+    typedef void (*hist_t)(const GpuMat& src, OutputArray hist, InputOutputArray buf, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);
+    static const hist_t hist_callers[] =
+    {
+        NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,
+        0,
+        NppHistogramEvenC1<CV_16U, nppiHistogramEven_16u_C1R, nppiHistogramEvenGetBufferSize_16u_C1R>::hist,
+        NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    if (src.depth() == CV_8U && deviceSupports(FEATURE_SET_COMPUTE_30))
+    {
+        histEven8u(src, hist.getGpuMatRef(), histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
+        return;
+    }
+
+    CV_Assert( src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );
+
+    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::histEven(InputArray _src, GpuMat hist[4], InputOutputArray buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
+{
+    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], InputOutputArray buf, int levels[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream);
+    static const hist_t hist_callers[] =
+    {
+        NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,
+        0,
+        NppHistogramEvenC4<CV_16U, nppiHistogramEven_16u_C4R, nppiHistogramEvenGetBufferSize_16u_C4R>::hist,
+        NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );
+
+    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::histRange(InputArray _src, OutputArray hist, InputArray _levels, InputOutputArray buf, Stream& stream)
+{
+    typedef void (*hist_t)(const GpuMat& src, OutputArray hist, const GpuMat& levels, InputOutputArray buf, cudaStream_t stream);
+    static const hist_t hist_callers[] =
+    {
+        NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,
+        0,
+        NppHistogramRangeC1<CV_16U, nppiHistogramRange_16u_C1R, nppiHistogramRangeGetBufferSize_16u_C1R>::hist,
+        NppHistogramRangeC1<CV_16S, nppiHistogramRange_16s_C1R, nppiHistogramRangeGetBufferSize_16s_C1R>::hist,
+        0,
+        NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist
+    };
+
+    GpuMat src = _src.getGpuMat();
+    GpuMat levels = _levels.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1 );
+
+    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
+}
+
+void cv::gpu::histRange(InputArray _src, GpuMat hist[4], const GpuMat levels[4], InputOutputArray buf, Stream& stream)
+{
+    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], InputOutputArray buf, cudaStream_t stream);
+    static const hist_t hist_callers[] =
+    {
+        NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,
+        0,
+        NppHistogramRangeC4<CV_16U, nppiHistogramRange_16u_C4R, nppiHistogramRangeGetBufferSize_16u_C4R>::hist,
+        NppHistogramRangeC4<CV_16S, nppiHistogramRange_16s_C4R, nppiHistogramRangeGetBufferSize_16s_C4R>::hist,
+        0,
+        NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4 );
+
+    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/hough_circles.cpp b/modules/gpuimgproc/src/hough_circles.cpp
new file mode 100644
index 0000000000..f123cf5795
--- /dev/null
+++ b/modules/gpuimgproc/src/hough_circles.cpp
@@ -0,0 +1,297 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) || !defined(HAVE_OPENCV_GPUFILTERS)
+
+Ptr<gpu::HoughCirclesDetector> cv::gpu::createHoughCirclesDetector(float, float, int, int, int, int, int) { throw_no_cuda(); return Ptr<HoughCirclesDetector>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough
+    {
+        int buildPointList_gpu(PtrStepSzb src, unsigned int* list);
+    }
+
+    namespace hough_circles
+    {
+        void circlesAccumCenters_gpu(const unsigned int* list, int count, PtrStepi dx, PtrStepi dy, PtrStepSzi accum, int minRadius, int maxRadius, float idp);
+        int buildCentersList_gpu(PtrStepSzi accum, unsigned int* centers, int threshold);
+        int circlesAccumRadius_gpu(const unsigned int* centers, int centersCount, const unsigned int* list, int count,
+                                   float3* circles, int maxCircles, float dp, int minRadius, int maxRadius, int threshold, bool has20);
+    }
+}}}
+
+namespace
+{
+    class HoughCirclesDetectorImpl : public HoughCirclesDetector
+    {
+    public:
+        HoughCirclesDetectorImpl(float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles);
+
+        void detect(InputArray src, OutputArray circles);
+
+        void setDp(float dp) { dp_ = dp; }
+        float getDp() const { return dp_; }
+
+        void setMinDist(float minDist) { minDist_ = minDist; }
+        float getMinDist() const { return minDist_; }
+
+        void setCannyThreshold(int cannyThreshold) { cannyThreshold_ = cannyThreshold; }
+        int getCannyThreshold() const { return cannyThreshold_; }
+
+        void setVotesThreshold(int votesThreshold) { votesThreshold_ = votesThreshold; }
+        int getVotesThreshold() const { return votesThreshold_; }
+
+        void setMinRadius(int minRadius) { minRadius_ = minRadius; }
+        int getMinRadius() const { return minRadius_; }
+
+        void setMaxRadius(int maxRadius) { maxRadius_ = maxRadius; }
+        int getMaxRadius() const { return maxRadius_; }
+
+        void setMaxCircles(int maxCircles) { maxCircles_ = maxCircles; }
+        int getMaxCircles() const { return maxCircles_; }
+
+        void write(FileStorage& fs) const
+        {
+            fs << "name" << "HoughCirclesDetector_GPU"
+            << "dp" << dp_
+            << "minDist" << minDist_
+            << "cannyThreshold" << cannyThreshold_
+            << "votesThreshold" << votesThreshold_
+            << "minRadius" << minRadius_
+            << "maxRadius" << maxRadius_
+            << "maxCircles" << maxCircles_;
+        }
+
+        void read(const FileNode& fn)
+        {
+            CV_Assert( String(fn["name"]) == "HoughCirclesDetector_GPU" );
+            dp_ = (float)fn["dp"];
+            minDist_ = (float)fn["minDist"];
+            cannyThreshold_ = (int)fn["cannyThreshold"];
+            votesThreshold_ = (int)fn["votesThreshold"];
+            minRadius_ = (int)fn["minRadius"];
+            maxRadius_ = (int)fn["maxRadius"];
+            maxCircles_ = (int)fn["maxCircles"];
+        }
+
+    private:
+        float dp_;
+        float minDist_;
+        int cannyThreshold_;
+        int votesThreshold_;
+        int minRadius_;
+        int maxRadius_;
+        int maxCircles_;
+
+        GpuMat dx_, dy_;
+        GpuMat edges_;
+        GpuMat accum_;
+        GpuMat list_;
+        GpuMat result_;
+        Ptr<gpu::Filter> filterDx_;
+        Ptr<gpu::Filter> filterDy_;
+        Ptr<gpu::CannyEdgeDetector> canny_;
+    };
+
+    HoughCirclesDetectorImpl::HoughCirclesDetectorImpl(float dp, float minDist, int cannyThreshold, int votesThreshold,
+                                                       int minRadius, int maxRadius, int maxCircles) :
+        dp_(dp), minDist_(minDist), cannyThreshold_(cannyThreshold), votesThreshold_(votesThreshold),
+        minRadius_(minRadius), maxRadius_(maxRadius), maxCircles_(maxCircles)
+    {
+        canny_ = gpu::createCannyEdgeDetector(std::max(cannyThreshold_ / 2, 1), cannyThreshold_);
+
+        filterDx_ = gpu::createSobelFilter(CV_8UC1, CV_32S, 1, 0);
+        filterDy_ = gpu::createSobelFilter(CV_8UC1, CV_32S, 0, 1);
+    }
+
+    void HoughCirclesDetectorImpl::detect(InputArray _src, OutputArray circles)
+    {
+        using namespace cv::gpu::cudev::hough;
+        using namespace cv::gpu::cudev::hough_circles;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+        CV_Assert( src.cols < std::numeric_limits<unsigned short>::max() );
+        CV_Assert( src.rows < std::numeric_limits<unsigned short>::max() );
+        CV_Assert( dp_ > 0 );
+        CV_Assert( minRadius_ > 0 && maxRadius_ > minRadius_ );
+        CV_Assert( cannyThreshold_ > 0 );
+        CV_Assert( votesThreshold_ > 0 );
+        CV_Assert( maxCircles_ > 0 );
+
+        const float idp = 1.0f / dp_;
+
+        filterDx_->apply(src, dx_);
+        filterDy_->apply(src, dy_);
+
+        canny_->setLowThreshold(std::max(cannyThreshold_ / 2, 1));
+        canny_->setHighThreshold(cannyThreshold_);
+
+        canny_->detect(dx_, dy_, edges_);
+
+        ensureSizeIsEnough(2, src.size().area(), CV_32SC1, list_);
+        unsigned int* srcPoints = list_.ptr<unsigned int>(0);
+        unsigned int* centers = list_.ptr<unsigned int>(1);
+
+        const int pointsCount = buildPointList_gpu(edges_, srcPoints);
+        if (pointsCount == 0)
+        {
+            circles.release();
+            return;
+        }
+
+        ensureSizeIsEnough(cvCeil(src.rows * idp) + 2, cvCeil(src.cols * idp) + 2, CV_32SC1, accum_);
+        accum_.setTo(Scalar::all(0));
+
+        circlesAccumCenters_gpu(srcPoints, pointsCount, dx_, dy_, accum_, minRadius_, maxRadius_, idp);
+
+        int centersCount = buildCentersList_gpu(accum_, centers, votesThreshold_);
+        if (centersCount == 0)
+        {
+            circles.release();
+            return;
+        }
+
+        if (minDist_ > 1)
+        {
+            AutoBuffer<ushort2> oldBuf_(centersCount);
+            AutoBuffer<ushort2> newBuf_(centersCount);
+            int newCount = 0;
+
+            ushort2* oldBuf = oldBuf_;
+            ushort2* newBuf = newBuf_;
+
+            cudaSafeCall( cudaMemcpy(oldBuf, centers, centersCount * sizeof(ushort2), cudaMemcpyDeviceToHost) );
+
+            const int cellSize = cvRound(minDist_);
+            const int gridWidth = (src.cols + cellSize - 1) / cellSize;
+            const int gridHeight = (src.rows + cellSize - 1) / cellSize;
+
+            std::vector< std::vector<ushort2> > grid(gridWidth * gridHeight);
+
+            const float minDist2 = minDist_ * minDist_;
+
+            for (int i = 0; i < centersCount; ++i)
+            {
+                ushort2 p = oldBuf[i];
+
+                bool good = true;
+
+                int xCell = static_cast<int>(p.x / cellSize);
+                int yCell = static_cast<int>(p.y / cellSize);
+
+                int x1 = xCell - 1;
+                int y1 = yCell - 1;
+                int x2 = xCell + 1;
+                int y2 = yCell + 1;
+
+                // boundary check
+                x1 = std::max(0, x1);
+                y1 = std::max(0, y1);
+                x2 = std::min(gridWidth - 1, x2);
+                y2 = std::min(gridHeight - 1, y2);
+
+                for (int yy = y1; yy <= y2; ++yy)
+                {
+                    for (int xx = x1; xx <= x2; ++xx)
+                    {
+                        std::vector<ushort2>& m = grid[yy * gridWidth + xx];
+
+                        for(size_t j = 0; j < m.size(); ++j)
+                        {
+                            float dx = (float)(p.x - m[j].x);
+                            float dy = (float)(p.y - m[j].y);
+
+                            if (dx * dx + dy * dy < minDist2)
+                            {
+                                good = false;
+                                goto break_out;
+                            }
+                        }
+                    }
+                }
+
+                break_out:
+
+                if(good)
+                {
+                    grid[yCell * gridWidth + xCell].push_back(p);
+
+                    newBuf[newCount++] = p;
+                }
+            }
+
+            cudaSafeCall( cudaMemcpy(centers, newBuf, newCount * sizeof(unsigned int), cudaMemcpyHostToDevice) );
+            centersCount = newCount;
+        }
+
+        ensureSizeIsEnough(1, maxCircles_, CV_32FC3, result_);
+
+        int circlesCount = circlesAccumRadius_gpu(centers, centersCount, srcPoints, pointsCount, result_.ptr<float3>(), maxCircles_,
+                                                  dp_, minRadius_, maxRadius_, votesThreshold_, deviceSupports(FEATURE_SET_COMPUTE_20));
+
+        if (circlesCount == 0)
+        {
+            circles.release();
+            return;
+        }
+
+        result_.cols = circlesCount;
+        result_.copyTo(circles);
+    }
+}
+
+Ptr<HoughCirclesDetector> cv::gpu::createHoughCirclesDetector(float dp, float minDist, int cannyThreshold, int votesThreshold, int minRadius, int maxRadius, int maxCircles)
+{
+    return makePtr<HoughCirclesDetectorImpl>(dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius, maxCircles);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/hough_lines.cpp b/modules/gpuimgproc/src/hough_lines.cpp
new file mode 100644
index 0000000000..e3e03cba73
--- /dev/null
+++ b/modules/gpuimgproc/src/hough_lines.cpp
@@ -0,0 +1,202 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<gpu::HoughLinesDetector> cv::gpu::createHoughLinesDetector(float, float, int, bool, int) { throw_no_cuda(); return Ptr<HoughLinesDetector>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough
+    {
+        int buildPointList_gpu(PtrStepSzb src, unsigned int* list);
+    }
+
+    namespace hough_lines
+    {
+        void linesAccum_gpu(const unsigned int* list, int count, PtrStepSzi accum, float rho, float theta, size_t sharedMemPerBlock, bool has20);
+        int linesGetResult_gpu(PtrStepSzi accum, float2* out, int* votes, int maxSize, float rho, float theta, int threshold, bool doSort);
+    }
+}}}
+
+namespace
+{
+    class HoughLinesDetectorImpl : public HoughLinesDetector
+    {
+    public:
+        HoughLinesDetectorImpl(float rho, float theta, int threshold, bool doSort, int maxLines) :
+            rho_(rho), theta_(theta), threshold_(threshold), doSort_(doSort), maxLines_(maxLines)
+        {
+        }
+
+        void detect(InputArray src, OutputArray lines);
+        void downloadResults(InputArray d_lines, OutputArray h_lines, OutputArray h_votes = noArray());
+
+        void setRho(float rho) { rho_ = rho; }
+        float getRho() const { return rho_; }
+
+        void setTheta(float theta) { theta_ = theta; }
+        float getTheta() const { return theta_; }
+
+        void setThreshold(int threshold) { threshold_ = threshold; }
+        int getThreshold() const { return threshold_; }
+
+        void setDoSort(bool doSort) { doSort_ = doSort; }
+        bool getDoSort() const { return doSort_; }
+
+        void setMaxLines(int maxLines) { maxLines_ = maxLines; }
+        int getMaxLines() const { return maxLines_; }
+
+        void write(FileStorage& fs) const
+        {
+            fs << "name" << "HoughLinesDetector_GPU"
+            << "rho" << rho_
+            << "theta" << theta_
+            << "threshold" << threshold_
+            << "doSort" << doSort_
+            << "maxLines" << maxLines_;
+        }
+
+        void read(const FileNode& fn)
+        {
+            CV_Assert( String(fn["name"]) == "HoughLinesDetector_GPU" );
+            rho_ = (float)fn["rho"];
+            theta_ = (float)fn["theta"];
+            threshold_ = (int)fn["threshold"];
+            doSort_ = (int)fn["doSort"] != 0;
+            maxLines_ = (int)fn["maxLines"];
+        }
+
+    private:
+        float rho_;
+        float theta_;
+        int threshold_;
+        bool doSort_;
+        int maxLines_;
+
+        GpuMat accum_;
+        GpuMat list_;
+        GpuMat result_;
+    };
+
+    void HoughLinesDetectorImpl::detect(InputArray _src, OutputArray lines)
+    {
+        using namespace cv::gpu::cudev::hough;
+        using namespace cv::gpu::cudev::hough_lines;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+        CV_Assert( src.cols < std::numeric_limits<unsigned short>::max() );
+        CV_Assert( src.rows < std::numeric_limits<unsigned short>::max() );
+
+        ensureSizeIsEnough(1, src.size().area(), CV_32SC1, list_);
+        unsigned int* srcPoints = list_.ptr<unsigned int>();
+
+        const int pointsCount = buildPointList_gpu(src, srcPoints);
+        if (pointsCount == 0)
+        {
+            lines.release();
+            return;
+        }
+
+        const int numangle = cvRound(CV_PI / theta_);
+        const int numrho = cvRound(((src.cols + src.rows) * 2 + 1) / rho_);
+        CV_Assert( numangle > 0 && numrho > 0 );
+
+        ensureSizeIsEnough(numangle + 2, numrho + 2, CV_32SC1, accum_);
+        accum_.setTo(Scalar::all(0));
+
+        DeviceInfo devInfo;
+        linesAccum_gpu(srcPoints, pointsCount, accum_, rho_, theta_, devInfo.sharedMemPerBlock(), devInfo.supports(FEATURE_SET_COMPUTE_20));
+
+        ensureSizeIsEnough(2, maxLines_, CV_32FC2, result_);
+
+        int linesCount = linesGetResult_gpu(accum_, result_.ptr<float2>(0), result_.ptr<int>(1), maxLines_, rho_, theta_, threshold_, doSort_);
+
+        if (linesCount == 0)
+        {
+            lines.release();
+            return;
+        }
+
+        result_.cols = linesCount;
+        result_.copyTo(lines);
+    }
+
+    void HoughLinesDetectorImpl::downloadResults(InputArray _d_lines, OutputArray h_lines, OutputArray h_votes)
+    {
+        GpuMat d_lines = _d_lines.getGpuMat();
+
+        if (d_lines.empty())
+        {
+            h_lines.release();
+            if (h_votes.needed())
+                h_votes.release();
+            return;
+        }
+
+        CV_Assert( d_lines.rows == 2 && d_lines.type() == CV_32FC2 );
+
+        d_lines.row(0).download(h_lines);
+
+        if (h_votes.needed())
+        {
+            GpuMat d_votes(1, d_lines.cols, CV_32SC1, d_lines.ptr<int>(1));
+            d_votes.download(h_votes);
+        }
+    }
+}
+
+Ptr<HoughLinesDetector> cv::gpu::createHoughLinesDetector(float rho, float theta, int threshold, bool doSort, int maxLines)
+{
+    return makePtr<HoughLinesDetectorImpl>(rho, theta, threshold, doSort, maxLines);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/hough_segments.cpp b/modules/gpuimgproc/src/hough_segments.cpp
new file mode 100644
index 0000000000..6f888a2c06
--- /dev/null
+++ b/modules/gpuimgproc/src/hough_segments.cpp
@@ -0,0 +1,183 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<gpu::HoughSegmentDetector> cv::gpu::createHoughSegmentDetector(float, float, int, int, int) { throw_no_cuda(); return Ptr<HoughSegmentDetector>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace hough
+    {
+        int buildPointList_gpu(PtrStepSzb src, unsigned int* list);
+    }
+
+    namespace hough_lines
+    {
+        void linesAccum_gpu(const unsigned int* list, int count, PtrStepSzi accum, float rho, float theta, size_t sharedMemPerBlock, bool has20);
+    }
+
+    namespace hough_segments
+    {
+        int houghLinesProbabilistic_gpu(PtrStepSzb mask, PtrStepSzi accum, int4* out, int maxSize, float rho, float theta, int lineGap, int lineLength);
+    }
+}}}
+
+namespace
+{
+    class HoughSegmentDetectorImpl : public HoughSegmentDetector
+    {
+    public:
+        HoughSegmentDetectorImpl(float rho, float theta, int minLineLength, int maxLineGap, int maxLines) :
+            rho_(rho), theta_(theta), minLineLength_(minLineLength), maxLineGap_(maxLineGap), maxLines_(maxLines)
+        {
+        }
+
+        void detect(InputArray src, OutputArray lines);
+
+        void setRho(float rho) { rho_ = rho; }
+        float getRho() const { return rho_; }
+
+        void setTheta(float theta) { theta_ = theta; }
+        float getTheta() const { return theta_; }
+
+        void setMinLineLength(int minLineLength) { minLineLength_ = minLineLength; }
+        int getMinLineLength() const { return minLineLength_; }
+
+        void setMaxLineGap(int maxLineGap) { maxLineGap_ = maxLineGap; }
+        int getMaxLineGap() const { return maxLineGap_; }
+
+        void setMaxLines(int maxLines) { maxLines_ = maxLines; }
+        int getMaxLines() const { return maxLines_; }
+
+        void write(FileStorage& fs) const
+        {
+            fs << "name" << "PHoughLinesDetector_GPU"
+            << "rho" << rho_
+            << "theta" << theta_
+            << "minLineLength" << minLineLength_
+            << "maxLineGap" << maxLineGap_
+            << "maxLines" << maxLines_;
+        }
+
+        void read(const FileNode& fn)
+        {
+            CV_Assert( String(fn["name"]) == "PHoughLinesDetector_GPU" );
+            rho_ = (float)fn["rho"];
+            theta_ = (float)fn["theta"];
+            minLineLength_ = (int)fn["minLineLength"];
+            maxLineGap_ = (int)fn["maxLineGap"];
+            maxLines_ = (int)fn["maxLines"];
+        }
+
+    private:
+        float rho_;
+        float theta_;
+        int minLineLength_;
+        int maxLineGap_;
+        int maxLines_;
+
+        GpuMat accum_;
+        GpuMat list_;
+        GpuMat result_;
+    };
+
+    void HoughSegmentDetectorImpl::detect(InputArray _src, OutputArray lines)
+    {
+        using namespace cv::gpu::cudev::hough;
+        using namespace cv::gpu::cudev::hough_lines;
+        using namespace cv::gpu::cudev::hough_segments;
+
+        GpuMat src = _src.getGpuMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+        CV_Assert( src.cols < std::numeric_limits<unsigned short>::max() );
+        CV_Assert( src.rows < std::numeric_limits<unsigned short>::max() );
+
+        ensureSizeIsEnough(1, src.size().area(), CV_32SC1, list_);
+        unsigned int* srcPoints = list_.ptr<unsigned int>();
+
+        const int pointsCount = buildPointList_gpu(src, srcPoints);
+        if (pointsCount == 0)
+        {
+            lines.release();
+            return;
+        }
+
+        const int numangle = cvRound(CV_PI / theta_);
+        const int numrho = cvRound(((src.cols + src.rows) * 2 + 1) / rho_);
+        CV_Assert( numangle > 0 && numrho > 0 );
+
+        ensureSizeIsEnough(numangle + 2, numrho + 2, CV_32SC1, accum_);
+        accum_.setTo(Scalar::all(0));
+
+        DeviceInfo devInfo;
+        linesAccum_gpu(srcPoints, pointsCount, accum_, rho_, theta_, devInfo.sharedMemPerBlock(), devInfo.supports(FEATURE_SET_COMPUTE_20));
+
+        ensureSizeIsEnough(1, maxLines_, CV_32SC4, result_);
+
+        int linesCount = houghLinesProbabilistic_gpu(src, accum_, result_.ptr<int4>(), maxLines_, rho_, theta_, maxLineGap_, minLineLength_);
+
+        if (linesCount == 0)
+        {
+            lines.release();
+            return;
+        }
+
+        result_.cols = linesCount;
+        result_.copyTo(lines);
+    }
+}
+
+Ptr<HoughSegmentDetector> cv::gpu::createHoughSegmentDetector(float rho, float theta, int minLineLength, int maxLineGap, int maxLines)
+{
+    return makePtr<HoughSegmentDetectorImpl>(rho, theta, minLineLength, maxLineGap, maxLines);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpuimgproc/src/match_template.cpp b/modules/gpuimgproc/src/match_template.cpp
new file mode 100644
index 0000000000..aeebd01d94
--- /dev/null
+++ b/modules/gpuimgproc/src/match_template.cpp
@@ -0,0 +1,649 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || !defined (HAVE_OPENCV_GPUARITHM) || defined (CUDA_DISABLER)
+
+Ptr<gpu::TemplateMatching> cv::gpu::createTemplateMatching(int, int, Size) { throw_no_cuda(); return Ptr<gpu::TemplateMatching>(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace match_template
+    {
+        void matchTemplateNaive_CCORR_8U(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
+        void matchTemplateNaive_CCORR_32F(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
+
+        void matchTemplateNaive_SQDIFF_8U(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
+        void matchTemplateNaive_SQDIFF_32F(const PtrStepSzb image, const PtrStepSzb templ, PtrStepSzf result, int cn, cudaStream_t stream);
+
+        void matchTemplatePrepared_SQDIFF_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum, unsigned long long templ_sqsum, PtrStepSzf result,
+            int cn, cudaStream_t stream);
+
+        void matchTemplatePrepared_SQDIFF_NORMED_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum, unsigned long long templ_sqsum, PtrStepSzf result,
+            int cn, cudaStream_t stream);
+
+        void matchTemplatePrepared_CCOFF_8U(int w, int h, const PtrStepSz<unsigned int> image_sum, unsigned int templ_sum, PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_8UC2(
+            int w, int h,
+            const PtrStepSz<unsigned int> image_sum_r,
+            const PtrStepSz<unsigned int> image_sum_g,
+            unsigned int templ_sum_r,
+            unsigned int templ_sum_g,
+            PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_8UC3(
+                int w, int h,
+                const PtrStepSz<unsigned int> image_sum_r,
+                const PtrStepSz<unsigned int> image_sum_g,
+                const PtrStepSz<unsigned int> image_sum_b,
+                unsigned int templ_sum_r,
+                unsigned int templ_sum_g,
+                unsigned int templ_sum_b,
+                PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_8UC4(
+                int w, int h,
+                const PtrStepSz<unsigned int> image_sum_r,
+                const PtrStepSz<unsigned int> image_sum_g,
+                const PtrStepSz<unsigned int> image_sum_b,
+                const PtrStepSz<unsigned int> image_sum_a,
+                unsigned int templ_sum_r,
+                unsigned int templ_sum_g,
+                unsigned int templ_sum_b,
+                unsigned int templ_sum_a,
+                PtrStepSzf result, cudaStream_t stream);
+
+
+        void matchTemplatePrepared_CCOFF_NORMED_8U(
+                int w, int h, const PtrStepSz<unsigned int> image_sum,
+                const PtrStepSz<unsigned long long> image_sqsum,
+                unsigned int templ_sum, unsigned long long templ_sqsum,
+                PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_NORMED_8UC2(
+                int w, int h,
+                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
+                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
+                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
+                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
+                PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_NORMED_8UC3(
+                int w, int h,
+                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
+                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
+                const PtrStepSz<unsigned int> image_sum_b, const PtrStepSz<unsigned long long> image_sqsum_b,
+                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
+                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
+                unsigned int templ_sum_b, unsigned long long templ_sqsum_b,
+                PtrStepSzf result, cudaStream_t stream);
+        void matchTemplatePrepared_CCOFF_NORMED_8UC4(
+                int w, int h,
+                const PtrStepSz<unsigned int> image_sum_r, const PtrStepSz<unsigned long long> image_sqsum_r,
+                const PtrStepSz<unsigned int> image_sum_g, const PtrStepSz<unsigned long long> image_sqsum_g,
+                const PtrStepSz<unsigned int> image_sum_b, const PtrStepSz<unsigned long long> image_sqsum_b,
+                const PtrStepSz<unsigned int> image_sum_a, const PtrStepSz<unsigned long long> image_sqsum_a,
+                unsigned int templ_sum_r, unsigned long long templ_sqsum_r,
+                unsigned int templ_sum_g, unsigned long long templ_sqsum_g,
+                unsigned int templ_sum_b, unsigned long long templ_sqsum_b,
+                unsigned int templ_sum_a, unsigned long long templ_sqsum_a,
+                PtrStepSzf result, cudaStream_t stream);
+
+        void normalize_8U(int w, int h, const PtrStepSz<unsigned long long> image_sqsum,
+                          unsigned long long templ_sqsum, PtrStepSzf result, int cn, cudaStream_t stream);
+
+        void extractFirstChannel_32F(const PtrStepSzb image, PtrStepSzf result, int cn, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    // Evaluates optimal template's area threshold. If
+    // template's area is less  than the threshold, we use naive match
+    // template version, otherwise FFT-based (if available)
+    int getTemplateThreshold(int method, int depth)
+    {
+        switch (method)
+        {
+        case TM_CCORR:
+            if (depth == CV_32F) return 250;
+            if (depth == CV_8U) return 300;
+            break;
+
+        case TM_SQDIFF:
+            if (depth == CV_8U) return 300;
+            break;
+        }
+
+        CV_Error(Error::StsBadArg, "unsupported match template mode");
+        return 0;
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // CCORR_32F
+
+    class Match_CCORR_32F : public TemplateMatching
+    {
+    public:
+        explicit Match_CCORR_32F(Size user_block_size);
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        Ptr<gpu::Convolution> conv_;
+        GpuMat result_;
+    };
+
+    Match_CCORR_32F::Match_CCORR_32F(Size user_block_size)
+    {
+        conv_ = gpu::createConvolution(user_block_size);
+    }
+
+    void Match_CCORR_32F::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& _stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_32F );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        _result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32FC1);
+        GpuMat result = _result.getGpuMat();
+
+        if (templ.size().area() < getTemplateThreshold(TM_CCORR, CV_32F))
+        {
+            matchTemplateNaive_CCORR_32F(image, templ, result, image.channels(), stream);
+            return;
+        }
+
+        if (image.channels() == 1)
+        {
+            conv_->convolve(image.reshape(1), templ.reshape(1), result, true, _stream);
+        }
+        else
+        {
+            conv_->convolve(image.reshape(1), templ.reshape(1), result_, true, _stream);
+            extractFirstChannel_32F(result_, result, image.channels(), stream);
+        }
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // CCORR_8U
+
+    class Match_CCORR_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_CCORR_8U(Size user_block_size) : match32F_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        GpuMat imagef_, templf_;
+        Match_CCORR_32F match32F_;
+    };
+
+    void Match_CCORR_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        if (templ.size().area() < getTemplateThreshold(TM_CCORR, CV_8U))
+        {
+            _result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32FC1);
+            GpuMat result = _result.getGpuMat();
+
+            matchTemplateNaive_CCORR_8U(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
+            return;
+        }
+
+        image.convertTo(imagef_, CV_32F, stream);
+        templ.convertTo(templf_, CV_32F, stream);
+
+        match32F_.match(imagef_, templf_, _result, stream);
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // CCORR_NORMED_8U
+
+    class Match_CCORR_NORMED_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_CCORR_NORMED_8U(Size user_block_size) : match_CCORR_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        Match_CCORR_8U match_CCORR_;
+        GpuMat image_sqsums_;
+        GpuMat intBuffer_;
+    };
+
+    void Match_CCORR_NORMED_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        match_CCORR_.match(image, templ, _result, stream);
+        GpuMat result = _result.getGpuMat();
+
+        gpu::sqrIntegral(image.reshape(1), image_sqsums_, intBuffer_, stream);
+
+        unsigned long long templ_sqsum = (unsigned long long) gpu::sqrSum(templ.reshape(1))[0];
+
+        normalize_8U(templ.cols, templ.rows, image_sqsums_, templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // SQDIFF_32F
+
+    class Match_SQDIFF_32F : public TemplateMatching
+    {
+    public:
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+    };
+
+    void Match_SQDIFF_32F::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_32F );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        _result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32FC1);
+        GpuMat result = _result.getGpuMat();
+
+        matchTemplateNaive_SQDIFF_32F(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // SQDIFF_8U
+
+    class Match_SQDIFF_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_SQDIFF_8U(Size user_block_size) : match_CCORR_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        GpuMat image_sqsums_;
+        GpuMat intBuffer_;
+        Match_CCORR_8U match_CCORR_;
+    };
+
+    void Match_SQDIFF_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        if (templ.size().area() < getTemplateThreshold(TM_SQDIFF, CV_8U))
+        {
+            _result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32FC1);
+            GpuMat result = _result.getGpuMat();
+
+            matchTemplateNaive_SQDIFF_8U(image, templ, result, image.channels(), StreamAccessor::getStream(stream));
+            return;
+        }
+
+        gpu::sqrIntegral(image.reshape(1), image_sqsums_, intBuffer_, stream);
+
+        unsigned long long templ_sqsum = (unsigned long long) gpu::sqrSum(templ.reshape(1))[0];
+
+        match_CCORR_.match(image, templ, _result, stream);
+        GpuMat result = _result.getGpuMat();
+
+        matchTemplatePrepared_SQDIFF_8U(templ.cols, templ.rows, image_sqsums_, templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // SQDIFF_NORMED_8U
+
+    class Match_SQDIFF_NORMED_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_SQDIFF_NORMED_8U(Size user_block_size) : match_CCORR_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        GpuMat image_sqsums_;
+        GpuMat intBuffer_;
+        Match_CCORR_8U match_CCORR_;
+    };
+
+    void Match_SQDIFF_NORMED_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        gpu::sqrIntegral(image.reshape(1), image_sqsums_, intBuffer_, stream);
+
+        unsigned long long templ_sqsum = (unsigned long long) gpu::sqrSum(templ.reshape(1))[0];
+
+        match_CCORR_.match(image, templ, _result, stream);
+        GpuMat result = _result.getGpuMat();
+
+        matchTemplatePrepared_SQDIFF_NORMED_8U(templ.cols, templ.rows, image_sqsums_, templ_sqsum, result, image.channels(), StreamAccessor::getStream(stream));
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // CCOFF_8U
+
+    class Match_CCOEFF_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_CCOEFF_8U(Size user_block_size) : match_CCORR_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        GpuMat intBuffer_;
+        std::vector<GpuMat> images_;
+        std::vector<GpuMat> image_sums_;
+        Match_CCORR_8U match_CCORR_;
+    };
+
+    void Match_CCOEFF_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        match_CCORR_.match(image, templ, _result, stream);
+        GpuMat result = _result.getGpuMat();
+
+        if (image.channels() == 1)
+        {
+            image_sums_.resize(1);
+            gpu::integral(image, image_sums_[0], intBuffer_, stream);
+
+            unsigned int templ_sum = (unsigned int) gpu::sum(templ)[0];
+
+            matchTemplatePrepared_CCOFF_8U(templ.cols, templ.rows, image_sums_[0], templ_sum, result, StreamAccessor::getStream(stream));
+        }
+        else
+        {
+            gpu::split(image, images_);
+
+            image_sums_.resize(images_.size());
+            for (int i = 0; i < image.channels(); ++i)
+                gpu::integral(images_[i], image_sums_[i], intBuffer_, stream);
+
+            Scalar templ_sum = gpu::sum(templ);
+
+            switch (image.channels())
+            {
+            case 2:
+                matchTemplatePrepared_CCOFF_8UC2(
+                        templ.cols, templ.rows, image_sums_[0], image_sums_[1],
+                        (unsigned int) templ_sum[0], (unsigned int) templ_sum[1],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            case 3:
+                matchTemplatePrepared_CCOFF_8UC3(
+                        templ.cols, templ.rows, image_sums_[0], image_sums_[1], image_sums_[2],
+                        (unsigned int) templ_sum[0], (unsigned int) templ_sum[1], (unsigned int) templ_sum[2],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            case 4:
+                matchTemplatePrepared_CCOFF_8UC4(
+                        templ.cols, templ.rows, image_sums_[0], image_sums_[1], image_sums_[2], image_sums_[3],
+                        (unsigned int) templ_sum[0], (unsigned int) templ_sum[1], (unsigned int) templ_sum[2], (unsigned int) templ_sum[3],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            default:
+                CV_Error(Error::StsBadArg, "unsupported number of channels");
+            }
+        }
+    }
+
+    ///////////////////////////////////////////////////////////////
+    // CCOFF_NORMED_8U
+
+    class Match_CCOEFF_NORMED_8U : public TemplateMatching
+    {
+    public:
+        explicit Match_CCOEFF_NORMED_8U(Size user_block_size) : match_CCORR_32F_(user_block_size)
+        {
+        }
+
+        void match(InputArray image, InputArray templ, OutputArray result, Stream& stream = Stream::Null());
+
+    private:
+        GpuMat imagef_, templf_;
+        Match_CCORR_32F match_CCORR_32F_;
+        GpuMat intBuffer_;
+        std::vector<GpuMat> images_;
+        std::vector<GpuMat> image_sums_;
+        std::vector<GpuMat> image_sqsums_;
+    };
+
+    void Match_CCOEFF_NORMED_8U::match(InputArray _image, InputArray _templ, OutputArray _result, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::match_template;
+
+        GpuMat image = _image.getGpuMat();
+        GpuMat templ = _templ.getGpuMat();
+
+        CV_Assert( image.depth() == CV_8U );
+        CV_Assert( image.type() == templ.type() );
+        CV_Assert( image.cols >= templ.cols && image.rows >= templ.rows );
+
+        image.convertTo(imagef_, CV_32F, stream);
+        templ.convertTo(templf_, CV_32F, stream);
+
+        match_CCORR_32F_.match(imagef_, templf_, _result, stream);
+        GpuMat result = _result.getGpuMat();
+
+        if (image.channels() == 1)
+        {
+            image_sums_.resize(1);
+            gpu::integral(image, image_sums_[0], intBuffer_, stream);
+
+            image_sqsums_.resize(1);
+            gpu::sqrIntegral(image, image_sqsums_[0], intBuffer_, stream);
+
+            unsigned int templ_sum = (unsigned int) gpu::sum(templ)[0];
+            unsigned long long templ_sqsum = (unsigned long long) gpu::sqrSum(templ)[0];
+
+            matchTemplatePrepared_CCOFF_NORMED_8U(
+                    templ.cols, templ.rows, image_sums_[0], image_sqsums_[0],
+                    templ_sum, templ_sqsum, result, StreamAccessor::getStream(stream));
+        }
+        else
+        {
+            gpu::split(image, images_);
+
+            image_sums_.resize(images_.size());
+            image_sqsums_.resize(images_.size());
+            for (int i = 0; i < image.channels(); ++i)
+            {
+                gpu::integral(images_[i], image_sums_[i], intBuffer_, stream);
+                gpu::sqrIntegral(images_[i], image_sqsums_[i], intBuffer_, stream);
+            }
+
+            Scalar templ_sum = gpu::sum(templ);
+            Scalar templ_sqsum = gpu::sqrSum(templ);
+
+            switch (image.channels())
+            {
+            case 2:
+                matchTemplatePrepared_CCOFF_NORMED_8UC2(
+                        templ.cols, templ.rows,
+                        image_sums_[0], image_sqsums_[0],
+                        image_sums_[1], image_sqsums_[1],
+                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
+                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            case 3:
+                matchTemplatePrepared_CCOFF_NORMED_8UC3(
+                        templ.cols, templ.rows,
+                        image_sums_[0], image_sqsums_[0],
+                        image_sums_[1], image_sqsums_[1],
+                        image_sums_[2], image_sqsums_[2],
+                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
+                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
+                        (unsigned int)templ_sum[2], (unsigned long long)templ_sqsum[2],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            case 4:
+                matchTemplatePrepared_CCOFF_NORMED_8UC4(
+                        templ.cols, templ.rows,
+                        image_sums_[0], image_sqsums_[0],
+                        image_sums_[1], image_sqsums_[1],
+                        image_sums_[2], image_sqsums_[2],
+                        image_sums_[3], image_sqsums_[3],
+                        (unsigned int)templ_sum[0], (unsigned long long)templ_sqsum[0],
+                        (unsigned int)templ_sum[1], (unsigned long long)templ_sqsum[1],
+                        (unsigned int)templ_sum[2], (unsigned long long)templ_sqsum[2],
+                        (unsigned int)templ_sum[3], (unsigned long long)templ_sqsum[3],
+                        result, StreamAccessor::getStream(stream));
+                break;
+            default:
+                CV_Error(Error::StsBadArg, "unsupported number of channels");
+            }
+        }
+    }
+}
+
+Ptr<gpu::TemplateMatching> cv::gpu::createTemplateMatching(int srcType, int method, Size user_block_size)
+{
+    const int sdepth = CV_MAT_DEPTH(srcType);
+
+    CV_Assert( sdepth == CV_8U || sdepth == CV_32F );
+
+    if (sdepth == CV_32F)
+    {
+        switch (method)
+        {
+        case TM_SQDIFF:
+            return makePtr<Match_SQDIFF_32F>();
+
+        case TM_CCORR:
+            return makePtr<Match_CCORR_32F>(user_block_size);
+
+        default:
+            CV_Error( Error::StsBadFlag, "Unsopported method" );
+            return Ptr<gpu::TemplateMatching>();
+        }
+    }
+    else
+    {
+        switch (method)
+        {
+        case TM_SQDIFF:
+            return makePtr<Match_SQDIFF_8U>(user_block_size);
+
+        case TM_SQDIFF_NORMED:
+            return makePtr<Match_SQDIFF_NORMED_8U>(user_block_size);
+
+        case TM_CCORR:
+            return makePtr<Match_CCORR_8U>(user_block_size);
+
+        case TM_CCORR_NORMED:
+            return makePtr<Match_CCORR_NORMED_8U>(user_block_size);
+
+        case TM_CCOEFF:
+            return makePtr<Match_CCOEFF_8U>(user_block_size);
+
+        case TM_CCOEFF_NORMED:
+            return makePtr<Match_CCOEFF_NORMED_8U>(user_block_size);
+
+        default:
+            CV_Error( Error::StsBadFlag, "Unsopported method" );
+            return Ptr<gpu::TemplateMatching>();
+        }
+    }
+}
+
+#endif
diff --git a/modules/gpuimgproc/src/mean_shift.cpp b/modules/gpuimgproc/src/mean_shift.cpp
new file mode 100644
index 0000000000..26368ca5ac
--- /dev/null
+++ b/modules/gpuimgproc/src/mean_shift.cpp
@@ -0,0 +1,128 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::meanShiftFiltering(InputArray, OutputArray, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
+void cv::gpu::meanShiftProc(InputArray, OutputArray, OutputArray, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+////////////////////////////////////////////////////////////////////////
+// meanShiftFiltering
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void meanShiftFiltering_gpu(const PtrStepSzb& src, PtrStepSzb dst, int sp, int sr, int maxIter, float eps, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::meanShiftFiltering(InputArray _src, OutputArray _dst, int sp, int sr, TermCriteria criteria, Stream& stream)
+{
+    using namespace ::cv::gpu::cudev::imgproc;
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 );
+
+    _dst.create(src.size(), CV_8UC4);
+    GpuMat dst = _dst.getGpuMat();
+
+    if (!(criteria.type & TermCriteria::MAX_ITER))
+        criteria.maxCount = 5;
+
+    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
+
+    if (!(criteria.type & TermCriteria::EPS))
+        criteria.epsilon = 1.f;
+
+    float eps = (float) std::max(criteria.epsilon, 0.0);
+
+    meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// meanShiftProc_GPU
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void meanShiftProc_gpu(const PtrStepSzb& src, PtrStepSzb dstr, PtrStepSzb dstsp, int sp, int sr, int maxIter, float eps, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::meanShiftProc(InputArray _src, OutputArray _dstr, OutputArray _dstsp, int sp, int sr, TermCriteria criteria, Stream& stream)
+{
+    using namespace ::cv::gpu::cudev::imgproc;
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 );
+
+    _dstr.create(src.size(), CV_8UC4);
+    _dstsp.create(src.size(), CV_16SC2);
+
+    GpuMat dstr = _dstr.getGpuMat();
+    GpuMat dstsp = _dstsp.getGpuMat();
+
+    if (!(criteria.type & TermCriteria::MAX_ITER))
+        criteria.maxCount = 5;
+
+    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
+
+    if (!(criteria.type & TermCriteria::EPS))
+        criteria.epsilon = 1.f;
+
+    float eps = (float) std::max(criteria.epsilon, 0.0);
+
+    meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/src/mssegmentation.cpp b/modules/gpuimgproc/src/mssegmentation.cpp
similarity index 95%
rename from modules/gpu/src/mssegmentation.cpp
rename to modules/gpuimgproc/src/mssegmentation.cpp
index a0bce6ad07..ec1c5feb4b 100644
--- a/modules/gpu/src/mssegmentation.cpp
+++ b/modules/gpuimgproc/src/mssegmentation.cpp
@@ -39,12 +39,11 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
-
 #include "precomp.hpp"
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::meanShiftSegmentation(const GpuMat&, Mat&, int, int, int, TermCriteria) { throw_nogpu(); }
+void cv::gpu::meanShiftSegmentation(InputArray, OutputArray, int, int, int, TermCriteria) { throw_no_cuda(); }
 
 #else
 
@@ -223,9 +222,12 @@ inline int dist2(const cv::Vec2s& lhs, const cv::Vec2s& rhs)
 } // anonymous namespace
 
 
-void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize, TermCriteria criteria)
+void cv::gpu::meanShiftSegmentation(InputArray _src, OutputArray _dst, int sp, int sr, int minsize, TermCriteria criteria)
 {
-    CV_Assert(src.type() == CV_8UC4);
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 );
+
     const int nrows = src.rows;
     const int ncols = src.cols;
     const int hr = sr;
@@ -233,7 +235,7 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
 
     // Perform mean shift procedure and obtain region and spatial maps
     GpuMat d_rmap, d_spmap;
-    meanShiftProc(src, d_rmap, d_spmap, sp, sr, criteria);
+    gpu::meanShiftProc(src, d_rmap, d_spmap, sp, sr, criteria);
     Mat rmap(d_rmap);
     Mat spmap(d_spmap);
 
@@ -338,7 +340,7 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
     }
 
     // Sort all graph's edges connecting differnet components (in asceding order)
-    sort(edges.begin(), edges.end());
+    std::sort(edges.begin(), edges.end());
 
     // Exclude small components (starting from the nearest couple)
     for (size_t i = 0; i < edges.size(); ++i)
@@ -367,7 +369,8 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
     }
 
     // Create final image, color of each segment is the average color of its pixels
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    Mat dst = _dst.getMat();
 
     for (int y = 0; y < nrows; ++y)
     {
diff --git a/modules/gpuimgproc/src/precomp.hpp b/modules/gpuimgproc/src/precomp.hpp
new file mode 100644
index 0000000000..b1edbf1239
--- /dev/null
+++ b/modules/gpuimgproc/src/precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include "opencv2/gpuimgproc.hpp"
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+#  include "opencv2/gpufilters.hpp"
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/test/test_threshold.cpp b/modules/gpuimgproc/test/test_bilateral_filter.cpp
similarity index 64%
rename from modules/gpu/test/test_threshold.cpp
rename to modules/gpuimgproc/test/test_bilateral_filter.cpp
index 43e651ad81..23dd3b85d6 100644
--- a/modules/gpu/test/test_threshold.cpp
+++ b/modules/gpuimgproc/test/test_bilateral_filter.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,49 +44,54 @@
 
 #ifdef HAVE_CUDA
 
-CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
-#define ALL_THRESH_OPS testing::Values(ThreshOp(cv::THRESH_BINARY), ThreshOp(cv::THRESH_BINARY_INV), ThreshOp(cv::THRESH_TRUNC), ThreshOp(cv::THRESH_TOZERO), ThreshOp(cv::THRESH_TOZERO_INV))
+using namespace cvtest;
 
-PARAM_TEST_CASE(Threshold, cv::gpu::DeviceInfo, cv::Size, MatType, ThreshOp, UseRoi)
+////////////////////////////////////////////////////////
+// BilateralFilter
+
+PARAM_TEST_CASE(BilateralFilter, cv::gpu::DeviceInfo, cv::Size, MatType)
 {
     cv::gpu::DeviceInfo devInfo;
     cv::Size size;
     int type;
-    int threshOp;
-    bool useRoi;
+    int kernel_size;
+    float sigma_color;
+    float sigma_spatial;
 
     virtual void SetUp()
     {
         devInfo = GET_PARAM(0);
         size = GET_PARAM(1);
         type = GET_PARAM(2);
-        threshOp = GET_PARAM(3);
-        useRoi = GET_PARAM(4);
+
+        kernel_size = 5;
+        sigma_color = 10.f;
+        sigma_spatial = 3.5f;
 
         cv::gpu::setDevice(devInfo.deviceID());
     }
 };
 
-GPU_TEST_P(Threshold, Accuracy)
+GPU_TEST_P(BilateralFilter, Accuracy)
 {
     cv::Mat src = randomMat(size, type);
-    double maxVal = randomDouble(20.0, 127.0);
-    double thresh = randomDouble(0.0, maxVal);
 
-    cv::gpu::GpuMat dst = createMat(src.size(), src.type(), useRoi);
-    cv::gpu::threshold(loadMat(src, useRoi), dst, thresh, maxVal, threshOp);
+    src.convertTo(src, type);
+    cv::gpu::GpuMat dst;
+
+    cv::gpu::bilateralFilter(loadMat(src), dst, kernel_size, sigma_color, sigma_spatial);
 
     cv::Mat dst_gold;
-    cv::threshold(src, dst_gold, thresh, maxVal, threshOp);
+    cv::bilateralFilter(src, dst_gold, kernel_size, sigma_color, sigma_spatial);
 
-    EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-3 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Threshold, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, BilateralFilter, testing::Combine(
     ALL_DEVICES,
-    DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_16SC1), MatType(CV_32FC1)),
-    ALL_THRESH_OPS,
-    WHOLE_SUBMAT));
+    testing::Values(cv::Size(128, 128), cv::Size(113, 113), cv::Size(639, 481)),
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_32FC1), MatType(CV_32FC3))
+    ));
+
 
 #endif // HAVE_CUDA
diff --git a/modules/gpuimgproc/test/test_blend.cpp b/modules/gpuimgproc/test/test_blend.cpp
new file mode 100644
index 0000000000..87359b500c
--- /dev/null
+++ b/modules/gpuimgproc/test/test_blend.cpp
@@ -0,0 +1,124 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////
+// Blend
+
+namespace
+{
+    template <typename T>
+    void blendLinearGold(const cv::Mat& img1, const cv::Mat& img2, const cv::Mat& weights1, const cv::Mat& weights2, cv::Mat& result_gold)
+    {
+        result_gold.create(img1.size(), img1.type());
+
+        int cn = img1.channels();
+
+        for (int y = 0; y < img1.rows; ++y)
+        {
+            const float* weights1_row = weights1.ptr<float>(y);
+            const float* weights2_row = weights2.ptr<float>(y);
+            const T* img1_row = img1.ptr<T>(y);
+            const T* img2_row = img2.ptr<T>(y);
+            T* result_gold_row = result_gold.ptr<T>(y);
+
+            for (int x = 0; x < img1.cols * cn; ++x)
+            {
+                float w1 = weights1_row[x / cn];
+                float w2 = weights2_row[x / cn];
+                result_gold_row[x] = static_cast<T>((img1_row[x] * w1 + img2_row[x] * w2) / (w1 + w2 + 1e-5f));
+            }
+        }
+    }
+}
+
+PARAM_TEST_CASE(Blend, cv::gpu::DeviceInfo, cv::Size, MatType, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Blend, Accuracy)
+{
+    int depth = CV_MAT_DEPTH(type);
+
+    cv::Mat img1 = randomMat(size, type, 0.0, depth == CV_8U ? 255.0 : 1.0);
+    cv::Mat img2 = randomMat(size, type, 0.0, depth == CV_8U ? 255.0 : 1.0);
+    cv::Mat weights1 = randomMat(size, CV_32F, 0, 1);
+    cv::Mat weights2 = randomMat(size, CV_32F, 0, 1);
+
+    cv::gpu::GpuMat result;
+    cv::gpu::blendLinear(loadMat(img1, useRoi), loadMat(img2, useRoi), loadMat(weights1, useRoi), loadMat(weights2, useRoi), result);
+
+    cv::Mat result_gold;
+    if (depth == CV_8U)
+        blendLinearGold<uchar>(img1, img2, weights1, weights2, result_gold);
+    else
+        blendLinearGold<float>(img1, img2, weights1, weights2, result_gold);
+
+    EXPECT_MAT_NEAR(result_gold, result, CV_MAT_DEPTH(type) == CV_8U ? 1.0 : 1e-5);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Blend, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
+    WHOLE_SUBMAT));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/src/cuda/texture_binder.hpp b/modules/gpuimgproc/test/test_canny.cpp
similarity index 60%
rename from modules/gpu/src/cuda/texture_binder.hpp
rename to modules/gpuimgproc/test/test_canny.cpp
index 391eb9a190..3d9d350165 100644
--- a/modules/gpu/src/cuda/texture_binder.hpp
+++ b/modules/gpuimgproc/test/test_canny.cpp
@@ -40,53 +40,62 @@
 //
 //M*/
 
-#ifndef OPENCV_GPU_TEXTURE_BINDER_HPP_
-#define OPENCV_GPU_TEXTURE_BINDER_HPP_
+#include "test_precomp.hpp"
 
-#include "opencv2/gpu/devmem2d.hpp"
-#include <safe_call.hpp>
+#ifdef HAVE_CUDA
 
-namespace cv
+using namespace cvtest;
+
+////////////////////////////////////////////////////////
+// Canny
+
+namespace
 {
-  namespace gpu
-  {
-    class TextureBinder
-    {
-    public:
-      template<class T, enum cudaTextureReadMode readMode>
-      TextureBinder(const PtrStepSz<T>& arr, const struct texture<T, 2, readMode>& tex) : texref(&tex)
-      {
-        cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
-        cudaSafeCall( cudaBindTexture2D(0, tex, arr.data, desc, arr.cols, arr.rows, arr.step) );
-      }
-
-      template<class T, enum cudaTextureReadMode readMode>
-      TextureBinder(const PtrSz<T>& arr, const struct texture<T, 1, readMode> &tex) : texref(&tex)
-      {
-        cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
-        cudaSafeCall( cudaBindTexture(0, tex, arr.data, desc, arr.size * arr.elemSize()) );
-      }
-
-      template<class A, class T, enum cudaTextureReadMode readMode>
-      TextureBinder(const A& arr, const struct texture<T, 2, readMode>& tex, const cudaChannelFormatDesc& desc) : texref(&tex)
-      {
-        cudaSafeCall( cudaBindTexture2D(0, tex, arr.data, desc, arr.cols, arr.rows, arr.step) );
-      }
-
-
-      ~TextureBinder()
-      {
-        cudaSafeCall( cudaUnbindTexture(texref) );
-      }
-    private:
-      const struct textureReference *texref;
-    };
-  }
-
-  namespace device
-  {
-      using cv::gpu::TextureBinder;
-  }
+    IMPLEMENT_PARAM_CLASS(AppertureSize, int)
+    IMPLEMENT_PARAM_CLASS(L2gradient, bool)
 }
 
-#endif /* OPENCV_GPU_TEXTURE_BINDER_HPP_*/
\ No newline at end of file
+PARAM_TEST_CASE(Canny, cv::gpu::DeviceInfo, AppertureSize, L2gradient, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    int apperture_size;
+    bool useL2gradient;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        apperture_size = GET_PARAM(1);
+        useL2gradient = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(Canny, Accuracy)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    double low_thresh = 50.0;
+    double high_thresh = 100.0;
+
+    cv::Ptr<cv::gpu::CannyEdgeDetector> canny = cv::gpu::createCannyEdgeDetector(low_thresh, high_thresh, apperture_size, useL2gradient);
+
+    cv::gpu::GpuMat edges;
+    canny->detect(loadMat(img, useRoi), edges);
+
+    cv::Mat edges_gold;
+    cv::Canny(img, edges_gold, low_thresh, high_thresh, apperture_size, useL2gradient);
+
+    EXPECT_MAT_SIMILAR(edges_gold, edges, 2e-2);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Canny, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(AppertureSize(3), AppertureSize(5)),
+    testing::Values(L2gradient(false), L2gradient(true)),
+    WHOLE_SUBMAT));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_color.cpp b/modules/gpuimgproc/test/test_color.cpp
similarity index 89%
rename from modules/gpu/test/test_color.cpp
rename to modules/gpuimgproc/test/test_color.cpp
index 5aee14d495..4bd53c9194 100644
--- a/modules/gpu/test/test_color.cpp
+++ b/modules/gpuimgproc/test/test_color.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 // cvtColor
 
@@ -2218,12 +2221,245 @@ GPU_TEST_P(CvtColor, BayerGR2BGR4)
     EXPECT_MAT_NEAR(dst_gold(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), dst3(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), 0);
 }
 
+GPU_TEST_P(CvtColor, BayerBG2Gray)
+{
+    if ((depth != CV_8U && depth != CV_16U) || useRoi)
+        return;
+
+    cv::Mat src = randomMat(size, depth);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::cvtColor(loadMat(src, useRoi), dst, cv::COLOR_BayerBG2GRAY);
+
+    cv::Mat dst_gold;
+    cv::cvtColor(src, dst_gold, cv::COLOR_BayerBG2GRAY);
+
+    EXPECT_MAT_NEAR(dst_gold(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), dst(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), 2);
+}
+
+GPU_TEST_P(CvtColor, BayerGB2Gray)
+{
+    if ((depth != CV_8U && depth != CV_16U) || useRoi)
+        return;
+
+    cv::Mat src = randomMat(size, depth);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::cvtColor(loadMat(src, useRoi), dst, cv::COLOR_BayerGB2GRAY);
+
+    cv::Mat dst_gold;
+    cv::cvtColor(src, dst_gold, cv::COLOR_BayerGB2GRAY);
+
+    EXPECT_MAT_NEAR(dst_gold(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), dst(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), 2);
+}
+
+GPU_TEST_P(CvtColor, BayerRG2Gray)
+{
+    if ((depth != CV_8U && depth != CV_16U) || useRoi)
+        return;
+
+    cv::Mat src = randomMat(size, depth);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::cvtColor(loadMat(src, useRoi), dst, cv::COLOR_BayerRG2GRAY);
+
+    cv::Mat dst_gold;
+    cv::cvtColor(src, dst_gold, cv::COLOR_BayerRG2GRAY);
+
+    EXPECT_MAT_NEAR(dst_gold(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), dst(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), 2);
+}
+
+GPU_TEST_P(CvtColor, BayerGR2Gray)
+{
+    if ((depth != CV_8U && depth != CV_16U) || useRoi)
+        return;
+
+    cv::Mat src = randomMat(size, depth);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::cvtColor(loadMat(src, useRoi), dst, cv::COLOR_BayerGR2GRAY);
+
+    cv::Mat dst_gold;
+    cv::cvtColor(src, dst_gold, cv::COLOR_BayerGR2GRAY);
+
+    EXPECT_MAT_NEAR(dst_gold(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), dst(cv::Rect(1, 1, dst.cols - 2, dst.rows - 2)), 2);
+}
+
 INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CvtColor, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_32F)),
     WHOLE_SUBMAT));
 
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// Demosaicing
+
+struct Demosaicing : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+
+    static void mosaic(const cv::Mat_<cv::Vec3b>& src, cv::Mat_<uchar>& dst, cv::Point firstRed)
+    {
+        dst.create(src.size());
+
+        for (int y = 0; y < src.rows; ++y)
+        {
+            for (int x = 0; x < src.cols; ++x)
+            {
+                cv::Vec3b pix = src(y, x);
+
+                cv::Point alternate;
+                alternate.x = (x + firstRed.x) % 2;
+                alternate.y = (y + firstRed.y) % 2;
+
+                if (alternate.y == 0)
+                {
+                    if (alternate.x == 0)
+                    {
+                        // RG
+                        // GB
+                        dst(y, x) = pix[2];
+                    }
+                    else
+                    {
+                        // GR
+                        // BG
+                        dst(y, x) = pix[1];
+                    }
+                }
+                else
+                {
+                    if (alternate.x == 0)
+                    {
+                        // GB
+                        // RG
+                        dst(y, x) = pix[1];
+                    }
+                    else
+                    {
+                        // BG
+                        // GR
+                        dst(y, x) = pix[0];
+                    }
+                }
+            }
+        }
+    }
+};
+
+GPU_TEST_P(Demosaicing, BayerBG2BGR)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(1, 1));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::COLOR_BayerBG2BGR);
+
+    EXPECT_MAT_SIMILAR(img, dst, 2e-2);
+}
+
+GPU_TEST_P(Demosaicing, BayerGB2BGR)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(0, 1));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::COLOR_BayerGB2BGR);
+
+    EXPECT_MAT_SIMILAR(img, dst, 2e-2);
+}
+
+GPU_TEST_P(Demosaicing, BayerRG2BGR)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(0, 0));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::COLOR_BayerRG2BGR);
+
+    EXPECT_MAT_SIMILAR(img, dst, 2e-2);
+}
+
+GPU_TEST_P(Demosaicing, BayerGR2BGR)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(1, 0));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::COLOR_BayerGR2BGR);
+
+    EXPECT_MAT_SIMILAR(img, dst, 2e-2);
+}
+
+GPU_TEST_P(Demosaicing, BayerBG2BGR_MHT)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(1, 1));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::gpu::COLOR_BayerBG2BGR_MHT);
+
+    EXPECT_MAT_SIMILAR(img, dst, 5e-3);
+}
+
+GPU_TEST_P(Demosaicing, BayerGB2BGR_MHT)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(0, 1));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::gpu::COLOR_BayerGB2BGR_MHT);
+
+    EXPECT_MAT_SIMILAR(img, dst, 5e-3);
+}
+
+GPU_TEST_P(Demosaicing, BayerRG2BGR_MHT)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(0, 0));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::gpu::COLOR_BayerRG2BGR_MHT);
+
+    EXPECT_MAT_SIMILAR(img, dst, 5e-3);
+}
+
+GPU_TEST_P(Demosaicing, BayerGR2BGR_MHT)
+{
+    cv::Mat img = readImage("stereobm/aloe-L.png");
+
+    cv::Mat_<uchar> src;
+    mosaic(img, src, cv::Point(1, 0));
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::demosaicing(loadMat(src), dst, cv::gpu::COLOR_BayerGR2BGR_MHT);
+
+    EXPECT_MAT_SIMILAR(img, dst, 5e-3);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Demosaicing, ALL_DEVICES);
+
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 // swapChannels
 
diff --git a/modules/gpuimgproc/test/test_corners.cpp b/modules/gpuimgproc/test/test_corners.cpp
new file mode 100644
index 0000000000..2625480565
--- /dev/null
+++ b/modules/gpuimgproc/test/test_corners.cpp
@@ -0,0 +1,149 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// CornerHarris
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(BlockSize, int);
+    IMPLEMENT_PARAM_CLASS(ApertureSize, int);
+}
+
+PARAM_TEST_CASE(CornerHarris, cv::gpu::DeviceInfo, MatType, BorderType, BlockSize, ApertureSize)
+{
+    cv::gpu::DeviceInfo devInfo;
+    int type;
+    int borderType;
+    int blockSize;
+    int apertureSize;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        type = GET_PARAM(1);
+        borderType = GET_PARAM(2);
+        blockSize = GET_PARAM(3);
+        apertureSize = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CornerHarris, Accuracy)
+{
+    cv::Mat src = readImageType("stereobm/aloe-L.png", type);
+    ASSERT_FALSE(src.empty());
+
+    double k = randomDouble(0.1, 0.9);
+
+    cv::Ptr<cv::gpu::CornernessCriteria> harris = cv::gpu::createHarrisCorner(src.type(), blockSize, apertureSize, k, borderType);
+
+    cv::gpu::GpuMat dst;
+    harris->compute(loadMat(src), dst);
+
+    cv::Mat dst_gold;
+    cv::cornerHarris(src, dst_gold, blockSize, apertureSize, k, borderType);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.02);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CornerHarris, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_32FC1)),
+    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_REFLECT)),
+    testing::Values(BlockSize(3), BlockSize(5), BlockSize(7)),
+    testing::Values(ApertureSize(0), ApertureSize(3), ApertureSize(5), ApertureSize(7))));
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// cornerMinEigen
+
+PARAM_TEST_CASE(CornerMinEigen, cv::gpu::DeviceInfo, MatType, BorderType, BlockSize, ApertureSize)
+{
+    cv::gpu::DeviceInfo devInfo;
+    int type;
+    int borderType;
+    int blockSize;
+    int apertureSize;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        type = GET_PARAM(1);
+        borderType = GET_PARAM(2);
+        blockSize = GET_PARAM(3);
+        apertureSize = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CornerMinEigen, Accuracy)
+{
+    cv::Mat src = readImageType("stereobm/aloe-L.png", type);
+    ASSERT_FALSE(src.empty());
+
+    cv::Ptr<cv::gpu::CornernessCriteria> minEigenVal = cv::gpu::createMinEigenValCorner(src.type(), blockSize, apertureSize, borderType);
+
+    cv::gpu::GpuMat dst;
+    minEigenVal->compute(loadMat(src), dst);
+
+    cv::Mat dst_gold;
+    cv::cornerMinEigenVal(src, dst_gold, blockSize, apertureSize, borderType);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 0.02);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CornerMinEigen, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(MatType(CV_8UC1), MatType(CV_32FC1)),
+    testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_REFLECT)),
+    testing::Values(BlockSize(3), BlockSize(5), BlockSize(7)),
+    testing::Values(ApertureSize(0), ApertureSize(3), ApertureSize(5), ApertureSize(7))));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuimgproc/test/test_gftt.cpp b/modules/gpuimgproc/test/test_gftt.cpp
new file mode 100644
index 0000000000..6ba6e0cfff
--- /dev/null
+++ b/modules/gpuimgproc/test/test_gftt.cpp
@@ -0,0 +1,131 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+//////////////////////////////////////////////////////
+// GoodFeaturesToTrack
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(MinDistance, double)
+}
+
+PARAM_TEST_CASE(GoodFeaturesToTrack, cv::gpu::DeviceInfo, MinDistance)
+{
+    cv::gpu::DeviceInfo devInfo;
+    double minDistance;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        minDistance = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(GoodFeaturesToTrack, Accuracy)
+{
+    cv::Mat image = readImage("opticalflow/frame0.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    int maxCorners = 1000;
+    double qualityLevel = 0.01;
+
+    cv::Ptr<cv::gpu::CornersDetector> detector = cv::gpu::createGoodFeaturesToTrackDetector(image.type(), maxCorners, qualityLevel, minDistance);
+
+    cv::gpu::GpuMat d_pts;
+    detector->detect(loadMat(image), d_pts);
+
+    ASSERT_FALSE(d_pts.empty());
+
+    std::vector<cv::Point2f> pts(d_pts.cols);
+    cv::Mat pts_mat(1, d_pts.cols, CV_32FC2, (void*) &pts[0]);
+    d_pts.download(pts_mat);
+
+    std::vector<cv::Point2f> pts_gold;
+    cv::goodFeaturesToTrack(image, pts_gold, maxCorners, qualityLevel, minDistance);
+
+    ASSERT_EQ(pts_gold.size(), pts.size());
+
+    size_t mistmatch = 0;
+    for (size_t i = 0; i < pts.size(); ++i)
+    {
+        cv::Point2i a = pts_gold[i];
+        cv::Point2i b = pts[i];
+
+        bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
+
+        if (!eq)
+            ++mistmatch;
+    }
+
+    double bad_ratio = static_cast<double>(mistmatch) / pts.size();
+
+    ASSERT_LE(bad_ratio, 0.01);
+}
+
+GPU_TEST_P(GoodFeaturesToTrack, EmptyCorners)
+{
+    int maxCorners = 1000;
+    double qualityLevel = 0.01;
+
+    cv::gpu::GpuMat src(100, 100, CV_8UC1, cv::Scalar::all(0));
+    cv::gpu::GpuMat corners(1, maxCorners, CV_32FC2);
+
+    cv::Ptr<cv::gpu::CornersDetector> detector = cv::gpu::createGoodFeaturesToTrackDetector(src.type(), maxCorners, qualityLevel, minDistance);
+
+    detector->detect(src, corners);
+
+    ASSERT_TRUE(corners.empty());
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, GoodFeaturesToTrack, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(MinDistance(0.0), MinDistance(3.0))));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuimgproc/test/test_histogram.cpp b/modules/gpuimgproc/test/test_histogram.cpp
new file mode 100644
index 0000000000..556211729b
--- /dev/null
+++ b/modules/gpuimgproc/test/test_histogram.cpp
@@ -0,0 +1,216 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// HistEven
+
+PARAM_TEST_CASE(HistEven, cv::gpu::DeviceInfo, cv::Size)
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    cv::Size size;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(HistEven, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    int hbins = 30;
+    float hranges[] = {50.0f, 200.0f};
+
+    cv::gpu::GpuMat hist;
+    cv::gpu::histEven(loadMat(src), hist, hbins, (int) hranges[0], (int) hranges[1]);
+
+    cv::Mat hist_gold;
+
+    int histSize[] = {hbins};
+    const float* ranges[] = {hranges};
+    int channels[] = {0};
+    cv::calcHist(&src, 1, channels, cv::Mat(), hist_gold, 1, histSize, ranges);
+
+    hist_gold = hist_gold.t();
+    hist_gold.convertTo(hist_gold, CV_32S);
+
+    EXPECT_MAT_NEAR(hist_gold, hist, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, HistEven, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES));
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// CalcHist
+
+PARAM_TEST_CASE(CalcHist, cv::gpu::DeviceInfo, cv::Size)
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    cv::Size size;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CalcHist, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    cv::gpu::GpuMat hist;
+    cv::gpu::calcHist(loadMat(src), hist);
+
+    cv::Mat hist_gold;
+
+    const int hbins = 256;
+    const float hranges[] = {0.0f, 256.0f};
+    const int histSize[] = {hbins};
+    const float* ranges[] = {hranges};
+    const int channels[] = {0};
+
+    cv::calcHist(&src, 1, channels, cv::Mat(), hist_gold, 1, histSize, ranges);
+    hist_gold = hist_gold.reshape(1, 1);
+    hist_gold.convertTo(hist_gold, CV_32S);
+
+    EXPECT_MAT_NEAR(hist_gold, hist, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CalcHist, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES));
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// EqualizeHist
+
+PARAM_TEST_CASE(EqualizeHist, cv::gpu::DeviceInfo, cv::Size)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(EqualizeHist, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::equalizeHist(loadMat(src), dst);
+
+    cv::Mat dst_gold;
+    cv::equalizeHist(src, dst_gold);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 3.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, EqualizeHist, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES));
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// CLAHE
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(ClipLimit, double)
+}
+
+PARAM_TEST_CASE(CLAHE, cv::gpu::DeviceInfo, cv::Size, ClipLimit)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    double clipLimit;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        clipLimit = GET_PARAM(2);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(CLAHE, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_8UC1);
+
+    cv::Ptr<cv::gpu::CLAHE> clahe = cv::gpu::createCLAHE(clipLimit);
+    cv::gpu::GpuMat dst;
+    clahe->apply(loadMat(src), dst);
+
+    cv::Ptr<cv::CLAHE> clahe_gold = cv::createCLAHE(clipLimit);
+    cv::Mat dst_gold;
+    clahe_gold->apply(src, dst_gold);
+
+    ASSERT_MAT_NEAR(dst_gold, dst, 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, CLAHE, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(0.0, 40.0)));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_hough.cpp b/modules/gpuimgproc/test/test_hough.cpp
similarity index 88%
rename from modules/gpu/test/test_hough.cpp
rename to modules/gpuimgproc/test/test_hough.cpp
index b348331a10..969899d8b6 100644
--- a/modules/gpu/test/test_hough.cpp
+++ b/modules/gpuimgproc/test/test_hough.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 // HoughLines
 
@@ -91,11 +94,13 @@ GPU_TEST_P(HoughLines, Accuracy)
     cv::Mat src(size, CV_8UC1);
     generateLines(src);
 
+    cv::Ptr<cv::gpu::HoughLinesDetector> hough = cv::gpu::createHoughLinesDetector(rho, theta, threshold);
+
     cv::gpu::GpuMat d_lines;
-    cv::gpu::HoughLines(loadMat(src, useRoi), d_lines, rho, theta, threshold);
+    hough->detect(loadMat(src, useRoi), d_lines);
 
     std::vector<cv::Vec2f> lines;
-    cv::gpu::HoughLinesDownload(d_lines, lines);
+    hough->downloadResults(d_lines, lines);
 
     cv::Mat dst(size, CV_8UC1);
     drawLines(dst, lines);
@@ -145,11 +150,13 @@ GPU_TEST_P(HoughCircles, Accuracy)
     cv::Mat src(size, CV_8UC1);
     drawCircles(src, circles_gold, true);
 
+    cv::Ptr<cv::gpu::HoughCirclesDetector> houghCircles = cv::gpu::createHoughCirclesDetector(dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+
     cv::gpu::GpuMat d_circles;
-    cv::gpu::HoughCircles(loadMat(src, useRoi), d_circles, CV_HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+    houghCircles->detect(loadMat(src, useRoi), d_circles);
 
     std::vector<cv::Vec3f> circles;
-    cv::gpu::HoughCirclesDownload(d_circles, circles);
+    d_circles.download(circles);
 
     ASSERT_FALSE(circles.empty());
 
@@ -186,7 +193,7 @@ PARAM_TEST_CASE(GeneralizedHough, cv::gpu::DeviceInfo, UseRoi)
 {
 };
 
-GPU_TEST_P(GeneralizedHough, POSITION)
+GPU_TEST_P(GeneralizedHough, Ballard)
 {
     const cv::gpu::DeviceInfo devInfo = GET_PARAM(0);
     cv::gpu::setDevice(devInfo.deviceID());
@@ -211,16 +218,16 @@ GPU_TEST_P(GeneralizedHough, POSITION)
         templ.copyTo(imageROI);
     }
 
-    cv::Ptr<cv::gpu::GeneralizedHough_GPU> hough = cv::gpu::GeneralizedHough_GPU::create(cv::GHT_POSITION);
-    hough->set("votesThreshold", 200);
+    cv::Ptr<cv::GeneralizedHoughBallard> alg = cv::gpu::createGeneralizedHoughBallard();
+    alg->setVotesThreshold(200);
 
-    hough->setTemplate(loadMat(templ, useRoi));
+    alg->setTemplate(loadMat(templ, useRoi));
 
     cv::gpu::GpuMat d_pos;
-    hough->detect(loadMat(image, useRoi), d_pos);
+    alg->detect(loadMat(image, useRoi), d_pos);
 
     std::vector<cv::Vec4f> pos;
-    hough->download(d_pos, pos);
+    d_pos.download(pos);
 
     ASSERT_EQ(gold_count, pos.size());
 
diff --git a/modules/gpuimgproc/test/test_main.cpp b/modules/gpuimgproc/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpuimgproc/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpuimgproc/test/test_match_template.cpp b/modules/gpuimgproc/test/test_match_template.cpp
new file mode 100644
index 0000000000..b6fd161408
--- /dev/null
+++ b/modules/gpuimgproc/test/test_match_template.cpp
@@ -0,0 +1,317 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate8U
+
+CV_ENUM(TemplateMethod, cv::TM_SQDIFF, cv::TM_SQDIFF_NORMED, cv::TM_CCORR, cv::TM_CCORR_NORMED, cv::TM_CCOEFF, cv::TM_CCOEFF_NORMED)
+#define ALL_TEMPLATE_METHODS testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_SQDIFF_NORMED), TemplateMethod(cv::TM_CCORR), TemplateMethod(cv::TM_CCORR_NORMED), TemplateMethod(cv::TM_CCOEFF), TemplateMethod(cv::TM_CCOEFF_NORMED))
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(TemplateSize, cv::Size);
+}
+
+PARAM_TEST_CASE(MatchTemplate8U, cv::gpu::DeviceInfo, cv::Size, TemplateSize, Channels, TemplateMethod)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    cv::Size templ_size;
+    int cn;
+    int method;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        templ_size = GET_PARAM(2);
+        cn = GET_PARAM(3);
+        method = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MatchTemplate8U, Accuracy)
+{
+    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_8U, cn));
+    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_8U, cn));
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), method);
+
+    cv::gpu::GpuMat dst;
+    alg->match(loadMat(image), loadMat(templ), dst);
+
+    cv::Mat dst_gold;
+    cv::matchTemplate(image, templ, dst_gold, method);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, templ_size.area() * 1e-1);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate8U, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
+    testing::Values(Channels(1), Channels(3), Channels(4)),
+    ALL_TEMPLATE_METHODS));
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate32F
+
+PARAM_TEST_CASE(MatchTemplate32F, cv::gpu::DeviceInfo, cv::Size, TemplateSize, Channels, TemplateMethod)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    cv::Size templ_size;
+    int cn;
+    int method;
+
+    int n, m, h, w;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        templ_size = GET_PARAM(2);
+        cn = GET_PARAM(3);
+        method = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MatchTemplate32F, Regression)
+{
+    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_32F, cn));
+    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_32F, cn));
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), method);
+
+    cv::gpu::GpuMat dst;
+    alg->match(loadMat(image), loadMat(templ), dst);
+
+    cv::Mat dst_gold;
+    cv::matchTemplate(image, templ, dst_gold, method);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, templ_size.area() * 1e-1);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate32F, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
+    testing::Values(Channels(1), Channels(3), Channels(4)),
+    testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))));
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplateBlackSource
+
+PARAM_TEST_CASE(MatchTemplateBlackSource, cv::gpu::DeviceInfo, TemplateMethod)
+{
+    cv::gpu::DeviceInfo devInfo;
+    int method;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        method = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MatchTemplateBlackSource, Accuracy)
+{
+    cv::Mat image = readImage("matchtemplate/black.png");
+    ASSERT_FALSE(image.empty());
+
+    cv::Mat pattern = readImage("matchtemplate/cat.png");
+    ASSERT_FALSE(pattern.empty());
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), method);
+
+    cv::gpu::GpuMat d_dst;
+    alg->match(loadMat(image), loadMat(pattern), d_dst);
+
+    cv::Mat dst(d_dst);
+
+    double maxValue;
+    cv::Point maxLoc;
+    cv::minMaxLoc(dst, NULL, &maxValue, NULL, &maxLoc);
+
+    cv::Point maxLocGold = cv::Point(284, 12);
+
+    ASSERT_EQ(maxLocGold, maxLoc);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplateBlackSource, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(TemplateMethod(cv::TM_CCOEFF_NORMED), TemplateMethod(cv::TM_CCORR_NORMED))));
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate_CCOEF_NORMED
+
+PARAM_TEST_CASE(MatchTemplate_CCOEF_NORMED, cv::gpu::DeviceInfo, std::pair<std::string, std::string>)
+{
+    cv::gpu::DeviceInfo devInfo;
+    std::string imageName;
+    std::string patternName;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        imageName = GET_PARAM(1).first;
+        patternName = GET_PARAM(1).second;
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MatchTemplate_CCOEF_NORMED, Accuracy)
+{
+    cv::Mat image = readImage(imageName);
+    ASSERT_FALSE(image.empty());
+
+    cv::Mat pattern = readImage(patternName);
+    ASSERT_FALSE(pattern.empty());
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(image.type(), cv::TM_CCOEFF_NORMED);
+
+    cv::gpu::GpuMat d_dst;
+    alg->match(loadMat(image), loadMat(pattern), d_dst);
+
+    cv::Mat dst(d_dst);
+
+    cv::Point minLoc, maxLoc;
+    double minVal, maxVal;
+    cv::minMaxLoc(dst, &minVal, &maxVal, &minLoc, &maxLoc);
+
+    cv::Mat dstGold;
+    cv::matchTemplate(image, pattern, dstGold, cv::TM_CCOEFF_NORMED);
+
+    double minValGold, maxValGold;
+    cv::Point minLocGold, maxLocGold;
+    cv::minMaxLoc(dstGold, &minValGold, &maxValGold, &minLocGold, &maxLocGold);
+
+    ASSERT_EQ(minLocGold, minLoc);
+    ASSERT_EQ(maxLocGold, maxLoc);
+    ASSERT_LE(maxVal, 1.0);
+    ASSERT_GE(minVal, -1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate_CCOEF_NORMED, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(std::make_pair(std::string("matchtemplate/source-0.png"), std::string("matchtemplate/target-0.png")))));
+
+////////////////////////////////////////////////////////////////////////////////
+// MatchTemplate_CanFindBigTemplate
+
+struct MatchTemplate_CanFindBigTemplate : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MatchTemplate_CanFindBigTemplate, SQDIFF_NORMED)
+{
+    cv::Mat scene = readImage("matchtemplate/scene.png");
+    ASSERT_FALSE(scene.empty());
+
+    cv::Mat templ = readImage("matchtemplate/template.png");
+    ASSERT_FALSE(templ.empty());
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(scene.type(), cv::TM_SQDIFF_NORMED);
+
+    cv::gpu::GpuMat d_result;
+    alg->match(loadMat(scene), loadMat(templ), d_result);
+
+    cv::Mat result(d_result);
+
+    double minVal;
+    cv::Point minLoc;
+    cv::minMaxLoc(result, &minVal, 0, &minLoc, 0);
+
+    ASSERT_GE(minVal, 0);
+    ASSERT_LT(minVal, 1e-3);
+    ASSERT_EQ(344, minLoc.x);
+    ASSERT_EQ(0, minLoc.y);
+}
+
+GPU_TEST_P(MatchTemplate_CanFindBigTemplate, SQDIFF)
+{
+    cv::Mat scene = readImage("matchtemplate/scene.png");
+    ASSERT_FALSE(scene.empty());
+
+    cv::Mat templ = readImage("matchtemplate/template.png");
+    ASSERT_FALSE(templ.empty());
+
+    cv::Ptr<cv::gpu::TemplateMatching> alg = cv::gpu::createTemplateMatching(scene.type(), cv::TM_SQDIFF);
+
+    cv::gpu::GpuMat d_result;
+    alg->match(loadMat(scene), loadMat(templ), d_result);
+
+    cv::Mat result(d_result);
+
+    double minVal;
+    cv::Point minLoc;
+    cv::minMaxLoc(result, &minVal, 0, &minLoc, 0);
+
+    ASSERT_GE(minVal, 0);
+    ASSERT_EQ(344, minLoc.x);
+    ASSERT_EQ(0, minLoc.y);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate_CanFindBigTemplate, ALL_DEVICES);
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuimgproc/test/test_mean_shift.cpp b/modules/gpuimgproc/test/test_mean_shift.cpp
new file mode 100644
index 0000000000..e9101802b4
--- /dev/null
+++ b/modules/gpuimgproc/test/test_mean_shift.cpp
@@ -0,0 +1,174 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////////////////////////////
+// MeanShift
+
+struct MeanShift : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    cv::Mat img;
+
+    int spatialRad;
+    int colorRad;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        img = readImageType("meanshift/cones.png", CV_8UC4);
+        ASSERT_FALSE(img.empty());
+
+        spatialRad = 30;
+        colorRad = 30;
+    }
+};
+
+GPU_TEST_P(MeanShift, Filtering)
+{
+    cv::Mat img_template;
+    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
+        img_template = readImage("meanshift/con_result.png");
+    else
+        img_template = readImage("meanshift/con_result_CC1X.png");
+    ASSERT_FALSE(img_template.empty());
+
+    cv::gpu::GpuMat d_dst;
+    cv::gpu::meanShiftFiltering(loadMat(img), d_dst, spatialRad, colorRad);
+
+    ASSERT_EQ(CV_8UC4, d_dst.type());
+
+    cv::Mat dst(d_dst);
+
+    cv::Mat result;
+    cv::cvtColor(dst, result, cv::COLOR_BGRA2BGR);
+
+    EXPECT_MAT_NEAR(img_template, result, 0.0);
+}
+
+GPU_TEST_P(MeanShift, Proc)
+{
+    cv::FileStorage fs;
+    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
+        fs.open(std::string(cvtest::TS::ptr()->get_data_path()) + "meanshift/spmap.yaml", cv::FileStorage::READ);
+    else
+        fs.open(std::string(cvtest::TS::ptr()->get_data_path()) + "meanshift/spmap_CC1X.yaml", cv::FileStorage::READ);
+    ASSERT_TRUE(fs.isOpened());
+
+    cv::Mat spmap_template;
+    fs["spmap"] >> spmap_template;
+    ASSERT_FALSE(spmap_template.empty());
+
+    cv::gpu::GpuMat rmap_filtered;
+    cv::gpu::meanShiftFiltering(loadMat(img), rmap_filtered, spatialRad, colorRad);
+
+    cv::gpu::GpuMat rmap;
+    cv::gpu::GpuMat spmap;
+    cv::gpu::meanShiftProc(loadMat(img), rmap, spmap, spatialRad, colorRad);
+
+    ASSERT_EQ(CV_8UC4, rmap.type());
+
+    EXPECT_MAT_NEAR(rmap_filtered, rmap, 0.0);
+    EXPECT_MAT_NEAR(spmap_template, spmap, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MeanShift, ALL_DEVICES);
+
+////////////////////////////////////////////////////////////////////////////////
+// MeanShiftSegmentation
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(MinSize, int);
+}
+
+PARAM_TEST_CASE(MeanShiftSegmentation, cv::gpu::DeviceInfo, MinSize)
+{
+    cv::gpu::DeviceInfo devInfo;
+    int minsize;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        minsize = GET_PARAM(1);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(MeanShiftSegmentation, Regression)
+{
+    cv::Mat img = readImageType("meanshift/cones.png", CV_8UC4);
+    ASSERT_FALSE(img.empty());
+
+    std::ostringstream path;
+    path << "meanshift/cones_segmented_sp10_sr10_minsize" << minsize;
+    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
+        path << ".png";
+    else
+        path << "_CC1X.png";
+    cv::Mat dst_gold = readImage(path.str());
+    ASSERT_FALSE(dst_gold.empty());
+
+    cv::Mat dst;
+    cv::gpu::meanShiftSegmentation(loadMat(img), dst, 10, 10, minsize);
+
+    cv::Mat dst_rgb;
+    cv::cvtColor(dst, dst_rgb, cv::COLOR_BGRA2BGR);
+
+    EXPECT_MAT_SIMILAR(dst_gold, dst_rgb, 1e-3);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MeanShiftSegmentation, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(MinSize(0), MinSize(4), MinSize(20), MinSize(84), MinSize(340), MinSize(1364))));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuimgproc/test/test_precomp.hpp b/modules/gpuimgproc/test/test_precomp.hpp
new file mode 100644
index 0000000000..7a8b1cbcd1
--- /dev/null
+++ b/modules/gpuimgproc/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpuimgproc.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpulegacy/CMakeLists.txt b/modules/gpulegacy/CMakeLists.txt
new file mode 100644
index 0000000000..1a679d821c
--- /dev/null
+++ b/modules/gpulegacy/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(NOT HAVE_CUDA)
+  ocv_module_disable(gpulegacy)
+endif()
+
+set(the_description "GPU-accelerated Computer Vision (legacy)")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4130 /wd4324 /wd4512 /wd4310 -Wundef -Wmissing-declarations -Wuninitialized)
+
+ocv_define_module(gpulegacy opencv_core OPTIONAL opencv_objdetect)
diff --git a/modules/gpulegacy/include/opencv2/gpulegacy.hpp b/modules/gpulegacy/include/opencv2/gpulegacy.hpp
new file mode 100644
index 0000000000..fb88481a8f
--- /dev/null
+++ b/modules/gpulegacy/include/opencv2/gpulegacy.hpp
@@ -0,0 +1,52 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPULEGACY_HPP__
+#define __OPENCV_GPULEGACY_HPP__
+
+#include "opencv2/gpulegacy/NCV.hpp"
+#include "opencv2/gpulegacy/NPP_staging.hpp"
+#include "opencv2/gpulegacy/NCVPyramid.hpp"
+#include "opencv2/gpulegacy/NCVHaarObjectDetection.hpp"
+#include "opencv2/gpulegacy/NCVBroxOpticalFlow.hpp"
+
+#endif /* __OPENCV_GPULEGACY_HPP__ */
diff --git a/modules/gpu/src/nvidia/core/NCV.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/NCV.hpp
similarity index 91%
rename from modules/gpu/src/nvidia/core/NCV.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/NCV.hpp
index 703cb827b9..cb84c23ad5 100644
--- a/modules/gpu/src/nvidia/core/NCV.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/NCV.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -42,19 +43,15 @@
 #ifndef _ncv_hpp_
 #define _ncv_hpp_
 
-#if (defined WIN32 || defined _WIN32 || defined WINCE) && defined CVAPI_EXPORTS
-    #define NCV_EXPORTS __declspec(dllexport)
-#else
-    #define NCV_EXPORTS
-#endif
+#include "opencv2/core/cvdef.h"
 
 #ifdef _WIN32
     #define WIN32_LEAN_AND_MEAN
 #endif
 
 #include <cuda_runtime.h>
-#include <sstream>
-#include <iostream>
+#include "opencv2/core/cvstd.hpp"
+#include "opencv2/core/utility.hpp"
 
 
 //==============================================================================
@@ -129,7 +126,7 @@ typedef                int Ncv32s;
 typedef       unsigned int Ncv32u;
 typedef              short Ncv16s;
 typedef     unsigned short Ncv16u;
-typedef               char Ncv8s;
+typedef        signed char Ncv8s;
 typedef      unsigned char Ncv8u;
 typedef              float Ncv32f;
 typedef             double Ncv64f;
@@ -243,13 +240,13 @@ const Ncv32u K_LOG2_WARP_SIZE = 5;
 //==============================================================================
 
 
-NCV_EXPORTS void ncvDebugOutput(const std::string &msg);
+CV_EXPORTS void ncvDebugOutput(const cv::String &msg);
 
 
-typedef void NCVDebugOutputHandler(const std::string &msg);
+typedef void NCVDebugOutputHandler(const cv::String &msg);
 
 
-NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
+CV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
 
 
 #define ncvAssertPrintCheck(pred, msg) \
@@ -257,9 +254,8 @@ NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
     { \
         if (!(pred)) \
         { \
-            std::ostringstream oss; \
-            oss << "NCV Assertion Failed: " << msg << ", file=" << __FILE__ << ", line=" << __LINE__ << std::endl; \
-            ncvDebugOutput(oss.str()); \
+            cv::String str = cv::format("NCV Assertion Failed: %s, file=%s, line=%d", msg, __FILE__, __LINE__); \
+            ncvDebugOutput(str); \
         } \
     } while (0)
 
@@ -273,14 +269,19 @@ NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
 
 
 #define ncvAssertReturn(pred, err) \
-    ncvAssertPrintReturn(pred, "retcode=" << (int)err, err)
+    do \
+    { \
+        cv::String msg = cv::format("retcode=%d", (int)err); \
+        ncvAssertPrintReturn(pred, msg.c_str(), err); \
+    } while (0)
 
 
 #define ncvAssertReturnNcvStat(ncvOp) \
     do \
     { \
         NCVStatus _ncvStat = ncvOp; \
-        ncvAssertPrintReturn(NCV_SUCCESS==_ncvStat, "NcvStat=" << (int)_ncvStat, _ncvStat); \
+        cv::String msg = cv::format("NcvStat=%d", (int)_ncvStat); \
+        ncvAssertPrintReturn(NCV_SUCCESS==_ncvStat, msg.c_str(), _ncvStat); \
     } while (0)
 
 
@@ -288,7 +289,8 @@ NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
     do \
     { \
         cudaError_t res = cudacall; \
-        ncvAssertPrintReturn(cudaSuccess==res, "cudaError_t=" << (int)res, errCode); \
+        cv::String msg = cv::format("cudaError_t=%d", (int)res); \
+        ncvAssertPrintReturn(cudaSuccess==res, msg.c_str(), errCode); \
     } while (0)
 
 
@@ -296,7 +298,8 @@ NCV_EXPORTS void ncvSetDebugOutputHandler(NCVDebugOutputHandler* func);
     do \
     { \
         cudaError_t res = cudaGetLastError(); \
-        ncvAssertPrintReturn(cudaSuccess==res, "cudaError_t=" << (int)res, errCode); \
+        cv::String msg = cv::format("cudaError_t=%d", (int)res); \
+        ncvAssertPrintReturn(cudaSuccess==res, msg.c_str(), errCode); \
     } while (0)
 
 
@@ -394,11 +397,11 @@ typedef Ncv32u NCVStatus;
 
 typedef struct _NcvTimer *NcvTimer;
 
-NCV_EXPORTS NcvTimer ncvStartTimer(void);
+CV_EXPORTS NcvTimer ncvStartTimer(void);
 
-NCV_EXPORTS double ncvEndQueryTimerUs(NcvTimer t);
+CV_EXPORTS double ncvEndQueryTimerUs(NcvTimer t);
 
-NCV_EXPORTS double ncvEndQueryTimerMs(NcvTimer t);
+CV_EXPORTS double ncvEndQueryTimerMs(NcvTimer t);
 
 
 //==============================================================================
@@ -411,7 +414,7 @@ NCV_EXPORTS double ncvEndQueryTimerMs(NcvTimer t);
 /**
 * Calculates the aligned top bound value
 */
-NCV_EXPORTS Ncv32u alignUp(Ncv32u what, Ncv32u alignment);
+CV_EXPORTS Ncv32u alignUp(Ncv32u what, Ncv32u alignment);
 
 
 /**
@@ -429,7 +432,7 @@ enum NCVMemoryType
 /**
 * NCVMemPtr
 */
-struct NCV_EXPORTS NCVMemPtr
+struct CV_EXPORTS NCVMemPtr
 {
     void *ptr;
     NCVMemoryType memtype;
@@ -440,7 +443,7 @@ struct NCV_EXPORTS NCVMemPtr
 /**
 * NCVMemSegment
 */
-struct NCV_EXPORTS NCVMemSegment
+struct CV_EXPORTS NCVMemSegment
 {
     NCVMemPtr begin;
     size_t size;
@@ -451,7 +454,7 @@ struct NCV_EXPORTS NCVMemSegment
 /**
 * INCVMemAllocator (Interface)
 */
-class NCV_EXPORTS INCVMemAllocator
+class CV_EXPORTS INCVMemAllocator
 {
 public:
     virtual ~INCVMemAllocator() = 0;
@@ -473,7 +476,7 @@ inline INCVMemAllocator::~INCVMemAllocator() {}
 /**
 * NCVMemStackAllocator
 */
-class NCV_EXPORTS NCVMemStackAllocator : public INCVMemAllocator
+class CV_EXPORTS NCVMemStackAllocator : public INCVMemAllocator
 {
     NCVMemStackAllocator();
     NCVMemStackAllocator(const NCVMemStackAllocator &);
@@ -510,7 +513,7 @@ private:
 /**
 * NCVMemNativeAllocator
 */
-class NCV_EXPORTS NCVMemNativeAllocator : public INCVMemAllocator
+class CV_EXPORTS NCVMemNativeAllocator : public INCVMemAllocator
 {
 public:
 
@@ -542,12 +545,12 @@ private:
 /**
 * Copy dispatchers
 */
-NCV_EXPORTS NCVStatus memSegCopyHelper(void *dst, NCVMemoryType dstType,
+CV_EXPORTS NCVStatus memSegCopyHelper(void *dst, NCVMemoryType dstType,
                                        const void *src, NCVMemoryType srcType,
                                        size_t sz, cudaStream_t cuStream);
 
 
-NCV_EXPORTS NCVStatus memSegCopyHelper2D(void *dst, Ncv32u dstPitch, NCVMemoryType dstType,
+CV_EXPORTS NCVStatus memSegCopyHelper2D(void *dst, Ncv32u dstPitch, NCVMemoryType dstType,
                                          const void *src, Ncv32u srcPitch, NCVMemoryType srcType,
                                          Ncv32u widthbytes, Ncv32u height, cudaStream_t cuStream);
 
@@ -805,7 +808,7 @@ public:
     T& at(Ncv32u x, Ncv32u y) const
     {
         NcvBool bOutRange = (x >= this->_width || y >= this->_height);
-        ncvAssertPrintCheck(!bOutRange, "Error addressing matrix at [" << x << ", " << y << "]");
+        ncvAssertPrintCheck(!bOutRange, "Error addressing matrix");
         if (bOutRange)
         {
             return *this->_ptr;
@@ -983,23 +986,23 @@ private:
 /**
 * Operations with rectangles
 */
-NCV_EXPORTS NCVStatus ncvGroupRectangles_host(NCVVector<NcvRect32u> &hypotheses, Ncv32u &numHypotheses,
+CV_EXPORTS NCVStatus ncvGroupRectangles_host(NCVVector<NcvRect32u> &hypotheses, Ncv32u &numHypotheses,
                                               Ncv32u minNeighbors, Ncv32f intersectEps, NCVVector<Ncv32u> *hypothesesWeights);
 
 
-NCV_EXPORTS NCVStatus ncvDrawRects_8u_host(Ncv8u *h_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
+CV_EXPORTS NCVStatus ncvDrawRects_8u_host(Ncv8u *h_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
                                            NcvRect32u *h_rects, Ncv32u numRects, Ncv8u color);
 
 
-NCV_EXPORTS NCVStatus ncvDrawRects_32u_host(Ncv32u *h_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
+CV_EXPORTS NCVStatus ncvDrawRects_32u_host(Ncv32u *h_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
                                             NcvRect32u *h_rects, Ncv32u numRects, Ncv32u color);
 
 
-NCV_EXPORTS NCVStatus ncvDrawRects_8u_device(Ncv8u *d_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
+CV_EXPORTS NCVStatus ncvDrawRects_8u_device(Ncv8u *d_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
                                              NcvRect32u *d_rects, Ncv32u numRects, Ncv8u color, cudaStream_t cuStream);
 
 
-NCV_EXPORTS NCVStatus ncvDrawRects_32u_device(Ncv32u *d_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
+CV_EXPORTS NCVStatus ncvDrawRects_32u_device(Ncv32u *d_dst, Ncv32u dstStride, Ncv32u dstWidth, Ncv32u dstHeight,
                                               NcvRect32u *d_rects, Ncv32u numRects, Ncv32u color, cudaStream_t cuStream);
 
 
diff --git a/modules/gpu/src/nvidia/NCVBroxOpticalFlow.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/NCVBroxOpticalFlow.hpp
similarity index 94%
rename from modules/gpu/src/nvidia/NCVBroxOpticalFlow.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/NCVBroxOpticalFlow.hpp
index dde2c41b52..689347ac48 100644
--- a/modules/gpu/src/nvidia/NCVBroxOpticalFlow.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/NCVBroxOpticalFlow.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -59,7 +60,7 @@
 #ifndef _ncv_optical_flow_h_
 #define _ncv_optical_flow_h_
 
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
 
 /// \brief Model and solver parameters
 struct NCVBroxOpticalFlowDescriptor
@@ -91,7 +92,7 @@ struct NCVBroxOpticalFlowDescriptor
 /// \return                        computation status
 /////////////////////////////////////////////////////////////////////////////////////////
 
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus NCVBroxOpticalFlow(const NCVBroxOpticalFlowDescriptor desc,
                              INCVMemAllocator &gpu_mem_allocator,
                              const NCVMatrix<Ncv32f> &frame0,
diff --git a/modules/gpu/src/nvidia/NCVHaarObjectDetection.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/NCVHaarObjectDetection.hpp
similarity index 94%
rename from modules/gpu/src/nvidia/NCVHaarObjectDetection.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/NCVHaarObjectDetection.hpp
index 9f29206f06..78ecf3b7e5 100644
--- a/modules/gpu/src/nvidia/NCVHaarObjectDetection.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/NCVHaarObjectDetection.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -58,8 +59,7 @@
 #ifndef _ncvhaarobjectdetection_hpp_
 #define _ncvhaarobjectdetection_hpp_
 
-#include <string>
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
 
 
 //==============================================================================
@@ -355,7 +355,7 @@ enum
 };
 
 
-NCV_EXPORTS NCVStatus ncvDetectObjectsMultiScale_device(NCVMatrix<Ncv8u> &d_srcImg,
+CV_EXPORTS NCVStatus ncvDetectObjectsMultiScale_device(NCVMatrix<Ncv8u> &d_srcImg,
                                                         NcvSize32u srcRoi,
                                                         NCVVector<NcvRect32u> &d_dstRects,
                                                         Ncv32u &dstNumRects,
@@ -382,7 +382,7 @@ NCV_EXPORTS NCVStatus ncvDetectObjectsMultiScale_device(NCVMatrix<Ncv8u> &d_srcI
 #define HAAR_STDDEV_BORDER                  1
 
 
-NCV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_device(NCVMatrix<Ncv32u> &d_integralImage,
+CV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_device(NCVMatrix<Ncv32u> &d_integralImage,
                                                            NCVMatrix<Ncv32f> &d_weights,
                                                            NCVMatrixAlloc<Ncv32u> &d_pixelMask,
                                                            Ncv32u &numDetections,
@@ -401,7 +401,7 @@ NCV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_device(NCVMatrix<Ncv32u> &d_
                                                            cudaStream_t cuStream);
 
 
-NCV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_host(NCVMatrix<Ncv32u> &h_integralImage,
+CV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_host(NCVMatrix<Ncv32u> &h_integralImage,
                                                          NCVMatrix<Ncv32f> &h_weights,
                                                          NCVMatrixAlloc<Ncv32u> &h_pixelMask,
                                                          Ncv32u &numDetections,
@@ -418,7 +418,7 @@ NCV_EXPORTS NCVStatus ncvApplyHaarClassifierCascade_host(NCVMatrix<Ncv32u> &h_in
 #define RECT_SIMILARITY_PROPORTION      0.2f
 
 
-NCV_EXPORTS NCVStatus ncvGrowDetectionsVector_device(NCVVector<Ncv32u> &pixelMask,
+CV_EXPORTS NCVStatus ncvGrowDetectionsVector_device(NCVVector<Ncv32u> &pixelMask,
                                                      Ncv32u numPixelMaskDetections,
                                                      NCVVector<NcvRect32u> &hypotheses,
                                                      Ncv32u &totalDetections,
@@ -429,7 +429,7 @@ NCV_EXPORTS NCVStatus ncvGrowDetectionsVector_device(NCVVector<Ncv32u> &pixelMas
                                                      cudaStream_t cuStream);
 
 
-NCV_EXPORTS NCVStatus ncvGrowDetectionsVector_host(NCVVector<Ncv32u> &pixelMask,
+CV_EXPORTS NCVStatus ncvGrowDetectionsVector_host(NCVVector<Ncv32u> &pixelMask,
                                                    Ncv32u numPixelMaskDetections,
                                                    NCVVector<NcvRect32u> &hypotheses,
                                                    Ncv32u &totalDetections,
@@ -439,18 +439,18 @@ NCV_EXPORTS NCVStatus ncvGrowDetectionsVector_host(NCVVector<Ncv32u> &pixelMask,
                                                    Ncv32f curScale);
 
 
-NCV_EXPORTS NCVStatus ncvHaarGetClassifierSize(const std::string &filename, Ncv32u &numStages,
+CV_EXPORTS NCVStatus ncvHaarGetClassifierSize(const cv::String &filename, Ncv32u &numStages,
                                                Ncv32u &numNodes, Ncv32u &numFeatures);
 
 
-NCV_EXPORTS NCVStatus ncvHaarLoadFromFile_host(const std::string &filename,
+CV_EXPORTS NCVStatus ncvHaarLoadFromFile_host(const cv::String &filename,
                                                HaarClassifierCascadeDescriptor &haar,
                                                NCVVector<HaarStage64> &h_HaarStages,
                                                NCVVector<HaarClassifierNode128> &h_HaarNodes,
                                                NCVVector<HaarFeature64> &h_HaarFeatures);
 
 
-NCV_EXPORTS NCVStatus ncvHaarStoreNVBIN_host(const std::string &filename,
+CV_EXPORTS NCVStatus ncvHaarStoreNVBIN_host(const cv::String &filename,
                                              HaarClassifierCascadeDescriptor haar,
                                              NCVVector<HaarStage64> &h_HaarStages,
                                              NCVVector<HaarClassifierNode128> &h_HaarNodes,
@@ -458,4 +458,4 @@ NCV_EXPORTS NCVStatus ncvHaarStoreNVBIN_host(const std::string &filename,
 
 
 
-#endif // _ncvhaarobjectdetection_hpp_
\ No newline at end of file
+#endif // _ncvhaarobjectdetection_hpp_
diff --git a/modules/gpu/src/nvidia/core/NCVPyramid.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/NCVPyramid.hpp
similarity index 83%
rename from modules/gpu/src/nvidia/core/NCVPyramid.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/NCVPyramid.hpp
index 8660ab24ea..88e2296bea 100644
--- a/modules/gpu/src/nvidia/core/NCVPyramid.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/NCVPyramid.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -39,18 +40,27 @@
 //
 //M*/
 
-
 #ifndef _ncvpyramid_hpp_
 #define _ncvpyramid_hpp_
 
 #include <memory>
 #include <vector>
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
+#include "opencv2/core/cuda/common.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace pyramid
+    {
+        CV_EXPORTS void downsampleX2(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream);
+        CV_EXPORTS void interpolateFrom1(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream);
+    }
+}}}
 
 #if 0 //def _WIN32
 
 template <class T>
-class NCV_EXPORTS NCVMatrixStack
+class CV_EXPORTS NCVMatrixStack
 {
 public:
     NCVMatrixStack() {this->_arr.clear();}
@@ -71,7 +81,7 @@ private:
 
 
 template <class T>
-class NCV_EXPORTS NCVImagePyramid
+class CV_EXPORTS NCVImagePyramid
 {
 public:
 
diff --git a/modules/gpu/src/nvidia/NPP_staging/NPP_staging.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/NPP_staging.hpp
similarity index 97%
rename from modules/gpu/src/nvidia/NPP_staging/NPP_staging.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/NPP_staging.hpp
index 54c9ff5b92..ce64ae8103 100644
--- a/modules/gpu/src/nvidia/NPP_staging/NPP_staging.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/NPP_staging.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -42,7 +43,7 @@
 #ifndef _npp_staging_hpp_
 #define _npp_staging_hpp_
 
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
 
 
 /**
@@ -62,7 +63,7 @@
  * NOT THREAD SAFE
  * \return Current CUDA stream
  */
-NCV_EXPORTS
+CV_EXPORTS
 cudaStream_t nppStGetActiveCUDAstream();
 
 
@@ -72,7 +73,7 @@ cudaStream_t nppStGetActiveCUDAstream();
  * \param cudaStream        [IN] cudaStream CUDA stream to become current
  * \return CUDA stream used before
  */
-NCV_EXPORTS
+CV_EXPORTS
 cudaStream_t nppStSetActiveCUDAstream(cudaStream_t cudaStream);
 
 
@@ -141,7 +142,7 @@ enum NppStInterpMode
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStGetInterpolationBufferSize(NcvSize32u srcSize,
                                            Ncv32u nStep,
                                            Ncv32u *hpSize);
@@ -154,7 +155,7 @@ NCVStatus nppiStGetInterpolationBufferSize(NcvSize32u srcSize,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStInterpolateFrames(const NppStInterpolationState *pState);
 
 
@@ -176,7 +177,7 @@ NCVStatus nppiStInterpolateFrames(const NppStInterpolationState *pState);
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStFilterRowBorder_32f_C1R(const Ncv32f *pSrc,
                                         NcvSize32u srcSize,
                                         Ncv32u nSrcStep,
@@ -209,7 +210,7 @@ NCVStatus nppiStFilterRowBorder_32f_C1R(const Ncv32f *pSrc,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStFilterColumnBorder_32f_C1R(const Ncv32f *pSrc,
                                            NcvSize32u srcSize,
                                            Ncv32u nSrcStep,
@@ -232,7 +233,7 @@ NCVStatus nppiStFilterColumnBorder_32f_C1R(const Ncv32f *pSrc,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStVectorWarpGetBufferSize(NcvSize32u srcSize,
                                         Ncv32u nSrcStep,
                                         Ncv32u *hpSize);
@@ -258,7 +259,7 @@ NCVStatus nppiStVectorWarpGetBufferSize(NcvSize32u srcSize,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStVectorWarp_PSF1x1_32f_C1(const Ncv32f *pSrc,
                                          NcvSize32u srcSize,
                                          Ncv32u nSrcStep,
@@ -289,7 +290,7 @@ NCVStatus nppiStVectorWarp_PSF1x1_32f_C1(const Ncv32f *pSrc,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStVectorWarp_PSF2x2_32f_C1(const Ncv32f *pSrc,
                                          NcvSize32u srcSize,
                                          Ncv32u nSrcStep,
@@ -319,7 +320,7 @@ NCVStatus nppiStVectorWarp_PSF2x2_32f_C1(const Ncv32f *pSrc,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStResize_32f_C1R(const Ncv32f *pSrc,
                                NcvSize32u srcSize,
                                Ncv32u nSrcStep,
@@ -346,7 +347,7 @@ NCVStatus nppiStResize_32f_C1R(const Ncv32f *pSrc,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32u_C1R(Ncv32u *d_src, Ncv32u srcStep,
                                  Ncv32u *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -357,7 +358,7 @@ NCVStatus nppiStDecimate_32u_C1R(Ncv32u *d_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 32-bit signed pixels, single channel.
  * \see nppiStDecimate_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32s_C1R(Ncv32s *d_src, Ncv32u srcStep,
                                  Ncv32s *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -368,7 +369,7 @@ NCVStatus nppiStDecimate_32s_C1R(Ncv32s *d_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 32-bit float pixels, single channel.
  * \see nppiStDecimate_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32f_C1R(Ncv32f *d_src, Ncv32u srcStep,
                                  Ncv32f *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -379,7 +380,7 @@ NCVStatus nppiStDecimate_32f_C1R(Ncv32f *d_src, Ncv32u srcStep,
 * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit unsigned pixels, single channel.
 * \see nppiStDecimate_32u_C1R
 */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64u_C1R(Ncv64u *d_src, Ncv32u srcStep,
                                  Ncv64u *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -390,7 +391,7 @@ NCVStatus nppiStDecimate_64u_C1R(Ncv64u *d_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit signed pixels, single channel.
  * \see nppiStDecimate_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64s_C1R(Ncv64s *d_src, Ncv32u srcStep,
                                  Ncv64s *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -401,7 +402,7 @@ NCVStatus nppiStDecimate_64s_C1R(Ncv64s *d_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit float pixels, single channel.
  * \see nppiStDecimate_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64f_C1R(Ncv64f *d_src, Ncv32u srcStep,
                                  Ncv64f *d_dst, Ncv32u dstStep,
                                  NcvSize32u srcRoi, Ncv32u scale,
@@ -420,7 +421,7 @@ NCVStatus nppiStDecimate_64f_C1R(Ncv64f *d_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32u_C1R_host(Ncv32u *h_src, Ncv32u srcStep,
                                       Ncv32u *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -430,7 +431,7 @@ NCVStatus nppiStDecimate_32u_C1R_host(Ncv32u *h_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 32-bit signed pixels, single channel. Host implementation.
  * \see nppiStDecimate_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32s_C1R_host(Ncv32s *h_src, Ncv32u srcStep,
                                       Ncv32s *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -440,7 +441,7 @@ NCVStatus nppiStDecimate_32s_C1R_host(Ncv32s *h_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 32-bit float pixels, single channel. Host implementation.
  * \see nppiStDecimate_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_32f_C1R_host(Ncv32f *h_src, Ncv32u srcStep,
                                       Ncv32f *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -450,7 +451,7 @@ NCVStatus nppiStDecimate_32f_C1R_host(Ncv32f *h_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit unsigned pixels, single channel. Host implementation.
  * \see nppiStDecimate_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64u_C1R_host(Ncv64u *h_src, Ncv32u srcStep,
                                       Ncv64u *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -460,7 +461,7 @@ NCVStatus nppiStDecimate_64u_C1R_host(Ncv64u *h_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit signed pixels, single channel. Host implementation.
  * \see nppiStDecimate_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64s_C1R_host(Ncv64s *h_src, Ncv32u srcStep,
                                       Ncv64s *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -470,7 +471,7 @@ NCVStatus nppiStDecimate_64s_C1R_host(Ncv64s *h_src, Ncv32u srcStep,
  * Downsamples (decimates) an image using the nearest neighbor algorithm. 64-bit float pixels, single channel. Host implementation.
  * \see nppiStDecimate_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStDecimate_64f_C1R_host(Ncv64f *h_src, Ncv32u srcStep,
                                       Ncv64f *h_dst, Ncv32u dstStep,
                                       NcvSize32u srcRoi, Ncv32u scale);
@@ -492,7 +493,7 @@ NCVStatus nppiStDecimate_64f_C1R_host(Ncv64f *h_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStRectStdDev_32f_C1R(Ncv32u *d_sum, Ncv32u sumStep,
                                    Ncv64u *d_sqsum, Ncv32u sqsumStep,
                                    Ncv32f *d_norm, Ncv32u normStep,
@@ -515,7 +516,7 @@ NCVStatus nppiStRectStdDev_32f_C1R(Ncv32u *d_sum, Ncv32u sumStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStRectStdDev_32f_C1R_host(Ncv32u *h_sum, Ncv32u sumStep,
                                         Ncv64u *h_sqsum, Ncv32u sqsumStep,
                                         Ncv32f *h_norm, Ncv32u normStep,
@@ -534,7 +535,7 @@ NCVStatus nppiStRectStdDev_32f_C1R_host(Ncv32u *h_sum, Ncv32u sumStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32u_C1R(Ncv32u *d_src, Ncv32u srcStride,
                                   Ncv32u *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -543,7 +544,7 @@ NCVStatus nppiStTranspose_32u_C1R(Ncv32u *d_src, Ncv32u srcStride,
  * Transposes an image. 32-bit signed pixels, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32s_C1R(Ncv32s *d_src, Ncv32u srcStride,
                                   Ncv32s *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -552,7 +553,7 @@ NCVStatus nppiStTranspose_32s_C1R(Ncv32s *d_src, Ncv32u srcStride,
  * Transposes an image. 32-bit float pixels, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32f_C1R(Ncv32f *d_src, Ncv32u srcStride,
                                   Ncv32f *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -561,7 +562,7 @@ NCVStatus nppiStTranspose_32f_C1R(Ncv32f *d_src, Ncv32u srcStride,
  * Transposes an image. 64-bit unsigned pixels, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64u_C1R(Ncv64u *d_src, Ncv32u srcStride,
                                   Ncv64u *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -570,7 +571,7 @@ NCVStatus nppiStTranspose_64u_C1R(Ncv64u *d_src, Ncv32u srcStride,
  * Transposes an image. 64-bit signed pixels, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64s_C1R(Ncv64s *d_src, Ncv32u srcStride,
                                   Ncv64s *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -579,7 +580,7 @@ NCVStatus nppiStTranspose_64s_C1R(Ncv64s *d_src, Ncv32u srcStride,
  * Transposes an image. 64-bit float pixels, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64f_C1R(Ncv64f *d_src, Ncv32u srcStride,
                                   Ncv64f *d_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -588,7 +589,7 @@ NCVStatus nppiStTranspose_64f_C1R(Ncv64f *d_src, Ncv32u srcStride,
  * Transposes an image. 128-bit pixels of any type, single channel
  * \see nppiStTranspose_32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_128_C1R(void *d_src, Ncv32u srcStep,
                                   void *d_dst, Ncv32u dstStep, NcvSize32u srcRoi);
 
@@ -604,7 +605,7 @@ NCVStatus nppiStTranspose_128_C1R(void *d_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32u_C1R_host(Ncv32u *h_src, Ncv32u srcStride,
                                        Ncv32u *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -613,7 +614,7 @@ NCVStatus nppiStTranspose_32u_C1R_host(Ncv32u *h_src, Ncv32u srcStride,
  * Transposes an image. 32-bit signed pixels, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32s_C1R_host(Ncv32s *h_src, Ncv32u srcStride,
                                        Ncv32s *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -622,7 +623,7 @@ NCVStatus nppiStTranspose_32s_C1R_host(Ncv32s *h_src, Ncv32u srcStride,
  * Transposes an image. 32-bit float pixels, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_32f_C1R_host(Ncv32f *h_src, Ncv32u srcStride,
                                        Ncv32f *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -631,7 +632,7 @@ NCVStatus nppiStTranspose_32f_C1R_host(Ncv32f *h_src, Ncv32u srcStride,
  * Transposes an image. 64-bit unsigned pixels, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64u_C1R_host(Ncv64u *h_src, Ncv32u srcStride,
                                        Ncv64u *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -640,7 +641,7 @@ NCVStatus nppiStTranspose_64u_C1R_host(Ncv64u *h_src, Ncv32u srcStride,
  * Transposes an image. 64-bit signed pixels, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64s_C1R_host(Ncv64s *h_src, Ncv32u srcStride,
                                        Ncv64s *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -649,7 +650,7 @@ NCVStatus nppiStTranspose_64s_C1R_host(Ncv64s *h_src, Ncv32u srcStride,
  * Transposes an image. 64-bit float pixels, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_64f_C1R_host(Ncv64f *h_src, Ncv32u srcStride,
                                        Ncv64f *h_dst, Ncv32u dstStride, NcvSize32u srcRoi);
 
@@ -658,7 +659,7 @@ NCVStatus nppiStTranspose_64f_C1R_host(Ncv64f *h_src, Ncv32u srcStride,
  * Transposes an image. 128-bit pixels of any type, single channel. Host implementation
  * \see nppiStTranspose_32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStTranspose_128_C1R_host(void *d_src, Ncv32u srcStep,
                                        void *d_dst, Ncv32u dstStep, NcvSize32u srcRoi);
 
@@ -672,7 +673,7 @@ NCVStatus nppiStTranspose_128_C1R_host(void *d_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegralGetSize_8u32u(NcvSize32u roiSize, Ncv32u *pBufsize, cudaDeviceProp &devProp);
 
 
@@ -680,7 +681,7 @@ NCVStatus nppiStIntegralGetSize_8u32u(NcvSize32u roiSize, Ncv32u *pBufsize, cuda
  * Calculates the size of the temporary buffer for integral image creation
  * \see nppiStIntegralGetSize_8u32u
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegralGetSize_32f32f(NcvSize32u roiSize, Ncv32u *pBufsize, cudaDeviceProp &devProp);
 
 
@@ -698,7 +699,7 @@ NCVStatus nppiStIntegralGetSize_32f32f(NcvSize32u roiSize, Ncv32u *pBufsize, cud
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegral_8u32u_C1R(Ncv8u *d_src, Ncv32u srcStep,
                                    Ncv32u *d_dst, Ncv32u dstStep, NcvSize32u roiSize,
                                    Ncv8u *pBuffer, Ncv32u bufSize, cudaDeviceProp &devProp);
@@ -708,7 +709,7 @@ NCVStatus nppiStIntegral_8u32u_C1R(Ncv8u *d_src, Ncv32u srcStep,
  * Creates an integral image representation for the input image
  * \see nppiStIntegral_8u32u_C1R
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegral_32f32f_C1R(Ncv32f *d_src, Ncv32u srcStep,
                                     Ncv32f *d_dst, Ncv32u dstStep, NcvSize32u roiSize,
                                     Ncv8u *pBuffer, Ncv32u bufSize, cudaDeviceProp &devProp);
@@ -725,7 +726,7 @@ NCVStatus nppiStIntegral_32f32f_C1R(Ncv32f *d_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegral_8u32u_C1R_host(Ncv8u *h_src, Ncv32u srcStep,
                                         Ncv32u *h_dst, Ncv32u dstStep, NcvSize32u roiSize);
 
@@ -734,7 +735,7 @@ NCVStatus nppiStIntegral_8u32u_C1R_host(Ncv8u *h_src, Ncv32u srcStep,
  * Creates an integral image representation for the input image. Host implementation
  * \see nppiStIntegral_8u32u_C1R_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStIntegral_32f32f_C1R_host(Ncv32f *h_src, Ncv32u srcStep,
                                          Ncv32f *h_dst, Ncv32u dstStep, NcvSize32u roiSize);
 
@@ -748,7 +749,7 @@ NCVStatus nppiStIntegral_32f32f_C1R_host(Ncv32f *h_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStSqrIntegralGetSize_8u64u(NcvSize32u roiSize, Ncv32u *pBufsize, cudaDeviceProp &devProp);
 
 
@@ -766,7 +767,7 @@ NCVStatus nppiStSqrIntegralGetSize_8u64u(NcvSize32u roiSize, Ncv32u *pBufsize, c
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStSqrIntegral_8u64u_C1R(Ncv8u *d_src, Ncv32u srcStep,
                                       Ncv64u *d_dst, Ncv32u dstStep, NcvSize32u roiSize,
                                       Ncv8u *pBuffer, Ncv32u bufSize, cudaDeviceProp &devProp);
@@ -783,7 +784,7 @@ NCVStatus nppiStSqrIntegral_8u64u_C1R(Ncv8u *d_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppiStSqrIntegral_8u64u_C1R_host(Ncv8u *h_src, Ncv32u srcStep,
                                            Ncv64u *h_dst, Ncv32u dstStep, NcvSize32u roiSize);
 
@@ -805,7 +806,7 @@ NCVStatus nppiStSqrIntegral_8u64u_C1R_host(Ncv8u *h_src, Ncv32u srcStep,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompactGetSize_32u(Ncv32u srcLen, Ncv32u *pBufsize, cudaDeviceProp &devProp);
 
 
@@ -837,7 +838,7 @@ NCVStatus nppsStCompactGetSize_32f(Ncv32u srcLen, Ncv32u *pBufsize, cudaDevicePr
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32u(Ncv32u *d_src, Ncv32u srcLen,
                             Ncv32u *d_dst, Ncv32u *p_dstLen,
                             Ncv32u elemRemove, Ncv8u *pBuffer,
@@ -848,7 +849,7 @@ NCVStatus nppsStCompact_32u(Ncv32u *d_src, Ncv32u srcLen,
  * Compacts the input vector by removing elements of specified value. 32-bit signed values
  * \see nppsStCompact_32u
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32s(Ncv32s *d_src, Ncv32u srcLen,
                             Ncv32s *d_dst, Ncv32u *p_dstLen,
                             Ncv32s elemRemove, Ncv8u *pBuffer,
@@ -859,7 +860,7 @@ NCVStatus nppsStCompact_32s(Ncv32s *d_src, Ncv32u srcLen,
  * Compacts the input vector by removing elements of specified value. 32-bit float values
  * \see nppsStCompact_32u
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32f(Ncv32f *d_src, Ncv32u srcLen,
                             Ncv32f *d_dst, Ncv32u *p_dstLen,
                             Ncv32f elemRemove, Ncv8u *pBuffer,
@@ -877,7 +878,7 @@ NCVStatus nppsStCompact_32f(Ncv32f *d_src, Ncv32u srcLen,
  *
  * \return NCV status code
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32u_host(Ncv32u *h_src, Ncv32u srcLen,
                                  Ncv32u *h_dst, Ncv32u *dstLen, Ncv32u elemRemove);
 
@@ -886,7 +887,7 @@ NCVStatus nppsStCompact_32u_host(Ncv32u *h_src, Ncv32u srcLen,
  * Compacts the input vector by removing elements of specified value. 32-bit signed values. Host implementation
  * \see nppsStCompact_32u_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32s_host(Ncv32s *h_src, Ncv32u srcLen,
                                  Ncv32s *h_dst, Ncv32u *dstLen, Ncv32s elemRemove);
 
@@ -895,7 +896,7 @@ NCVStatus nppsStCompact_32s_host(Ncv32s *h_src, Ncv32u srcLen,
  * Compacts the input vector by removing elements of specified value. 32-bit float values. Host implementation
  * \see nppsStCompact_32u_host
  */
-NCV_EXPORTS
+CV_EXPORTS
 NCVStatus nppsStCompact_32f_host(Ncv32f *h_src, Ncv32u srcLen,
                                  Ncv32f *h_dst, Ncv32u *dstLen, Ncv32f elemRemove);
 
diff --git a/modules/gpu/src/cuda/internal_shared.hpp b/modules/gpulegacy/include/opencv2/gpulegacy/private.hpp
similarity index 69%
rename from modules/gpu/src/cuda/internal_shared.hpp
rename to modules/gpulegacy/include/opencv2/gpulegacy/private.hpp
index a0850d0081..26c76d4dde 100644
--- a/modules/gpu/src/cuda/internal_shared.hpp
+++ b/modules/gpulegacy/include/opencv2/gpulegacy/private.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,45 +41,23 @@
 //
 //M*/
 
-#ifndef __OPENCV_internal_shared_HPP__
-#define __OPENCV_internal_shared_HPP__
+#ifndef __OPENCV_CORE_GPULEGACY_PRIVATE_HPP__
+#define __OPENCV_CORE_GPULEGACY_PRIVATE_HPP__
 
-#include <cuda_runtime.h>
-#include <npp.h>
-#include "NPP_staging.hpp"
-#include "opencv2/gpu/devmem2d.hpp"
-#include "safe_call.hpp"
-#include "opencv2/gpu/device/common.hpp"
+#ifndef __OPENCV_BUILD
+#  error this is a private header which should not be used from outside of the OpenCV library
+#endif
+
+#include "opencv2/core/private.gpu.hpp"
+
+#ifndef HAVE_CUDA
+#  error gpulegacy module requires CUDA
+#endif
+
+#include "opencv2/gpulegacy.hpp"
 
 namespace cv { namespace gpu
 {
-    enum
-    {
-        BORDER_REFLECT101_GPU = 0,
-        BORDER_REPLICATE_GPU,
-        BORDER_CONSTANT_GPU,
-        BORDER_REFLECT_GPU,
-        BORDER_WRAP_GPU
-    };
-
-    class NppStreamHandler
-    {
-    public:
-        inline explicit NppStreamHandler(cudaStream_t newStream = 0)
-        {
-            oldStream = nppGetStream();
-            nppSetStream(newStream);
-        }
-
-        inline ~NppStreamHandler()
-        {
-            nppSetStream(oldStream);
-        }
-
-    private:
-        cudaStream_t oldStream;
-    };
-
     class NppStStreamHandler
     {
     public:
@@ -95,6 +74,23 @@ namespace cv { namespace gpu
     private:
         cudaStream_t oldStream;
     };
+
+    CV_EXPORTS cv::String getNcvErrorMessage(int code);
+
+    static inline void checkNcvError(int err, const char* file, const int line, const char* func)
+    {
+        if (NCV_SUCCESS != err)
+        {
+            cv::String msg = getNcvErrorMessage(err);
+            cv::error(cv::Error::GpuApiCallError, msg, func, file, line);
+        }
+    }
 }}
 
-#endif /* __OPENCV_internal_shared_HPP__ */
+#if defined(__GNUC__)
+    #define ncvSafeCall(expr)  cv::gpu::checkNcvError(expr, __FILE__, __LINE__, __func__)
+#else /* defined(__CUDACC__) || defined(__MSVC__) */
+    #define ncvSafeCall(expr)  cv::gpu::checkNcvError(expr, __FILE__, __LINE__, "")
+#endif
+
+#endif // __OPENCV_CORE_GPULEGACY_PRIVATE_HPP__
diff --git a/modules/gpu/src/nvidia/core/NCV.cu b/modules/gpulegacy/src/NCV.cpp
similarity index 81%
rename from modules/gpu/src/nvidia/core/NCV.cu
rename to modules/gpulegacy/src/NCV.cpp
index 77e59cc5c1..be82423d08 100644
--- a/modules/gpu/src/nvidia/core/NCV.cu
+++ b/modules/gpulegacy/src/NCV.cpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -39,12 +40,7 @@
 //
 //M*/
 
-
-#include <iostream>
-#include <string>
-#include <vector>
-#include "NCV.hpp"
-
+#include "precomp.hpp"
 
 //==============================================================================
 //
@@ -52,17 +48,97 @@
 //
 //==============================================================================
 
-
-static void stdDebugOutput(const std::string &msg)
+namespace
 {
-    std::cout << msg;
+    #define error_entry(entry)  { entry, #entry }
+
+    struct ErrorEntry
+    {
+        int code;
+        const char* str;
+    };
+
+    struct ErrorEntryComparer
+    {
+        int code;
+        ErrorEntryComparer(int code_) : code(code_) {}
+        bool operator()(const ErrorEntry& e) const { return e.code == code; }
+    };
+
+    //////////////////////////////////////////////////////////////////////////
+    // NCV errors
+
+    const ErrorEntry ncv_errors [] =
+    {
+        error_entry( NCV_SUCCESS ),
+        error_entry( NCV_UNKNOWN_ERROR ),
+        error_entry( NCV_CUDA_ERROR ),
+        error_entry( NCV_NPP_ERROR ),
+        error_entry( NCV_FILE_ERROR ),
+        error_entry( NCV_NULL_PTR ),
+        error_entry( NCV_INCONSISTENT_INPUT ),
+        error_entry( NCV_TEXTURE_BIND_ERROR ),
+        error_entry( NCV_DIMENSIONS_INVALID ),
+        error_entry( NCV_INVALID_ROI ),
+        error_entry( NCV_INVALID_STEP ),
+        error_entry( NCV_INVALID_SCALE ),
+        error_entry( NCV_INVALID_SCALE ),
+        error_entry( NCV_ALLOCATOR_NOT_INITIALIZED ),
+        error_entry( NCV_ALLOCATOR_BAD_ALLOC ),
+        error_entry( NCV_ALLOCATOR_BAD_DEALLOC ),
+        error_entry( NCV_ALLOCATOR_INSUFFICIENT_CAPACITY ),
+        error_entry( NCV_ALLOCATOR_DEALLOC_ORDER ),
+        error_entry( NCV_ALLOCATOR_BAD_REUSE ),
+        error_entry( NCV_MEM_COPY_ERROR ),
+        error_entry( NCV_MEM_RESIDENCE_ERROR ),
+        error_entry( NCV_MEM_INSUFFICIENT_CAPACITY ),
+        error_entry( NCV_HAAR_INVALID_PIXEL_STEP ),
+        error_entry( NCV_HAAR_TOO_MANY_FEATURES_IN_CLASSIFIER ),
+        error_entry( NCV_HAAR_TOO_MANY_FEATURES_IN_CASCADE ),
+        error_entry( NCV_HAAR_TOO_LARGE_FEATURES ),
+        error_entry( NCV_HAAR_XML_LOADING_EXCEPTION ),
+        error_entry( NCV_NOIMPL_HAAR_TILTED_FEATURES ),
+        error_entry( NCV_WARNING_HAAR_DETECTIONS_VECTOR_OVERFLOW ),
+        error_entry( NPPST_SUCCESS ),
+        error_entry( NPPST_ERROR ),
+        error_entry( NPPST_CUDA_KERNEL_EXECUTION_ERROR ),
+        error_entry( NPPST_NULL_POINTER_ERROR ),
+        error_entry( NPPST_TEXTURE_BIND_ERROR ),
+        error_entry( NPPST_MEMCPY_ERROR ),
+        error_entry( NPPST_MEM_ALLOC_ERR ),
+        error_entry( NPPST_MEMFREE_ERR ),
+        error_entry( NPPST_INVALID_ROI ),
+        error_entry( NPPST_INVALID_STEP ),
+        error_entry( NPPST_INVALID_SCALE ),
+        error_entry( NPPST_MEM_INSUFFICIENT_BUFFER ),
+        error_entry( NPPST_MEM_RESIDENCE_ERROR ),
+        error_entry( NPPST_MEM_INTERNAL_ERROR )
+    };
+
+    const size_t ncv_error_num = sizeof(ncv_errors) / sizeof(ncv_errors[0]);
+}
+
+cv::String cv::gpu::getNcvErrorMessage(int code)
+{
+    size_t idx = std::find_if(ncv_errors, ncv_errors + ncv_error_num, ErrorEntryComparer(code)) - ncv_errors;
+
+    const char* msg = (idx != ncv_error_num) ? ncv_errors[idx].str : "Unknown error code";
+    String str = cv::format("%s [Code = %d]", msg, code);
+
+    return str;
+}
+
+
+static void stdDebugOutput(const cv::String &msg)
+{
+    std::cout << msg.c_str() << std::endl;
 }
 
 
 static NCVDebugOutputHandler *debugOutputHandler = stdDebugOutput;
 
 
-void ncvDebugOutput(const std::string &msg)
+void ncvDebugOutput(const cv::String &msg)
 {
     debugOutputHandler(msg);
 }
@@ -73,8 +149,6 @@ void ncvSetDebugOutputHandler(NCVDebugOutputHandler *func)
     debugOutputHandler = func;
 }
 
-#if !defined CUDA_DISABLER
-
 
 //==============================================================================
 //
@@ -252,15 +326,14 @@ NCVStatus memSegCopyHelper2D(void *dst, Ncv32u dstPitch, NCVMemoryType dstType,
 //===================================================================
 
 
-NCVMemStackAllocator::NCVMemStackAllocator(Ncv32u alignment_)
-    :
-    currentSize(0),
-    _maxSize(0),
+NCVMemStackAllocator::NCVMemStackAllocator(Ncv32u alignment_) :
+    _memType(NCVMemoryTypeNone),
+    _alignment(alignment_),
     allocBegin(NULL),
     begin(NULL),
     end(NULL),
-    _memType(NCVMemoryTypeNone),
-    _alignment(alignment_),
+    currentSize(0),
+    _maxSize(0),
     bReusesMemory(false)
 {
     NcvBool bProperAlignment = (alignment_ & (alignment_ - 1)) == 0;
@@ -268,13 +341,12 @@ NCVMemStackAllocator::NCVMemStackAllocator(Ncv32u alignment_)
 }
 
 
-NCVMemStackAllocator::NCVMemStackAllocator(NCVMemoryType memT, size_t capacity, Ncv32u alignment_, void *reusePtr)
-    :
-    currentSize(0),
-    _maxSize(0),
-    allocBegin(NULL),
+NCVMemStackAllocator::NCVMemStackAllocator(NCVMemoryType memT, size_t capacity, Ncv32u alignment_, void *reusePtr) :
     _memType(memT),
-    _alignment(alignment_)
+    _alignment(alignment_),
+    allocBegin(NULL),
+    currentSize(0),
+    _maxSize(0)
 {
     NcvBool bProperAlignment = (alignment_ & (alignment_ - 1)) == 0;
     ncvAssertPrintCheck(bProperAlignment, "NCVMemStackAllocator ctor:: _alignment not power of 2");
@@ -390,7 +462,7 @@ NCVStatus NCVMemStackAllocator::dealloc(NCVMemSegment &seg)
 
 NcvBool NCVMemStackAllocator::isInitialized(void) const
 {
-    return ((this->_alignment & (this->_alignment-1)) == 0) && isCounting() || this->allocBegin != NULL;
+    return (((this->_alignment & (this->_alignment-1)) == 0) && isCounting()) || this->allocBegin != NULL;
 }
 
 
@@ -425,12 +497,11 @@ size_t NCVMemStackAllocator::maxSize(void) const
 //===================================================================
 
 
-NCVMemNativeAllocator::NCVMemNativeAllocator(NCVMemoryType memT, Ncv32u alignment_)
-    :
-    currentSize(0),
-    _maxSize(0),
+NCVMemNativeAllocator::NCVMemNativeAllocator(NCVMemoryType memT, Ncv32u alignment_) :
     _memType(memT),
-    _alignment(alignment_)
+    _alignment(alignment_),
+    currentSize(0),
+    _maxSize(0)
 {
     ncvAssertPrintReturn(memT != NCVMemoryTypeNone, "NCVMemNativeAllocator ctor:: counting not permitted for this allocator type", );
 }
@@ -650,9 +721,47 @@ double ncvEndQueryTimerMs(NcvTimer t)
 //
 //===================================================================
 
+struct RectConvert
+{
+    cv::Rect operator()(const NcvRect32u& nr) const { return cv::Rect(nr.x, nr.y, nr.width, nr.height); }
+    NcvRect32u operator()(const cv::Rect& nr) const
+    {
+        NcvRect32u rect;
+        rect.x = nr.x;
+        rect.y = nr.y;
+        rect.width = nr.width;
+        rect.height = nr.height;
+        return rect;
+    }
+};
+
+static void groupRectangles(std::vector<NcvRect32u> &hypotheses, int groupThreshold, double eps, std::vector<Ncv32u> *weights)
+{
+#ifndef HAVE_OPENCV_OBJDETECT
+    (void) hypotheses;
+    (void) groupThreshold;
+    (void) eps;
+    (void) weights;
+    CV_Error(cv::Error::StsNotImplemented, "This functionality requires objdetect module");
+#else
+    std::vector<cv::Rect> rects(hypotheses.size());
+    std::transform(hypotheses.begin(), hypotheses.end(), rects.begin(), RectConvert());
+
+    if (weights)
+    {
+        std::vector<int> weights_int;
+        weights_int.assign(weights->begin(), weights->end());
+        cv::groupRectangles(rects, weights_int, groupThreshold, eps);
+    }
+    else
+    {
+        cv::groupRectangles(rects, groupThreshold, eps);
+    }
+    std::transform(rects.begin(), rects.end(), hypotheses.begin(), RectConvert());
+    hypotheses.resize(rects.size());
+#endif
+}
 
-//from OpenCV
-void groupRectangles(std::vector<NcvRect32u> &hypotheses, int groupThreshold, double eps, std::vector<Ncv32u> *weights);
 
 
 NCVStatus ncvGroupRectangles_host(NCVVector<NcvRect32u> &hypotheses,
@@ -777,133 +886,3 @@ NCVStatus ncvDrawRects_32u_host(Ncv32u *h_dst,
 {
     return drawRectsWrapperHost(h_dst, dstStride, dstWidth, dstHeight, h_rects, numRects, color);
 }
-
-
-const Ncv32u NUMTHREADS_DRAWRECTS = 32;
-const Ncv32u NUMTHREADS_DRAWRECTS_LOG2 = 5;
-
-
-template <class T>
-__global__ void drawRects(T *d_dst,
-                          Ncv32u dstStride,
-                          Ncv32u dstWidth,
-                          Ncv32u dstHeight,
-                          NcvRect32u *d_rects,
-                          Ncv32u numRects,
-                          T color)
-{
-    Ncv32u blockId = blockIdx.y * 65535 + blockIdx.x;
-    if (blockId > numRects * 4)
-    {
-        return;
-    }
-
-    NcvRect32u curRect = d_rects[blockId >> 2];
-    NcvBool bVertical = blockId & 0x1;
-    NcvBool bTopLeft = blockId & 0x2;
-
-    Ncv32u pt0x, pt0y;
-    if (bVertical)
-    {
-        Ncv32u numChunks = (curRect.height + NUMTHREADS_DRAWRECTS - 1) >> NUMTHREADS_DRAWRECTS_LOG2;
-
-        pt0x = bTopLeft ? curRect.x : curRect.x + curRect.width - 1;
-        pt0y = curRect.y;
-
-        if (pt0x < dstWidth)
-        {
-            for (Ncv32u chunkId = 0; chunkId < numChunks; chunkId++)
-            {
-                Ncv32u ptY = pt0y + chunkId * NUMTHREADS_DRAWRECTS + threadIdx.x;
-                if (ptY < pt0y + curRect.height && ptY < dstHeight)
-                {
-                    d_dst[ptY * dstStride + pt0x] = color;
-                }
-            }
-        }
-    }
-    else
-    {
-        Ncv32u numChunks = (curRect.width + NUMTHREADS_DRAWRECTS - 1) >> NUMTHREADS_DRAWRECTS_LOG2;
-
-        pt0x = curRect.x;
-        pt0y = bTopLeft ? curRect.y : curRect.y + curRect.height - 1;
-
-        if (pt0y < dstHeight)
-        {
-            for (Ncv32u chunkId = 0; chunkId < numChunks; chunkId++)
-            {
-                Ncv32u ptX = pt0x + chunkId * NUMTHREADS_DRAWRECTS + threadIdx.x;
-                if (ptX < pt0x + curRect.width && ptX < dstWidth)
-                {
-                    d_dst[pt0y * dstStride + ptX] = color;
-                }
-            }
-        }
-    }
-}
-
-
-template <class T>
-static NCVStatus drawRectsWrapperDevice(T *d_dst,
-                                        Ncv32u dstStride,
-                                        Ncv32u dstWidth,
-                                        Ncv32u dstHeight,
-                                        NcvRect32u *d_rects,
-                                        Ncv32u numRects,
-                                        T color,
-                                        cudaStream_t cuStream)
-{
-    (void)cuStream;
-    ncvAssertReturn(d_dst != NULL && d_rects != NULL, NCV_NULL_PTR);
-    ncvAssertReturn(dstWidth > 0 && dstHeight > 0, NCV_DIMENSIONS_INVALID);
-    ncvAssertReturn(dstStride >= dstWidth, NCV_INVALID_STEP);
-    ncvAssertReturn(numRects <= dstWidth * dstHeight, NCV_DIMENSIONS_INVALID);
-
-    if (numRects == 0)
-    {
-        return NCV_SUCCESS;
-    }
-
-    dim3 grid(numRects * 4);
-    dim3 block(NUMTHREADS_DRAWRECTS);
-    if (grid.x > 65535)
-    {
-        grid.y = (grid.x + 65534) / 65535;
-        grid.x = 65535;
-    }
-
-    drawRects<T><<<grid, block>>>(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color);
-
-    ncvAssertCUDALastErrorReturn(NCV_CUDA_ERROR);
-
-    return NCV_SUCCESS;
-}
-
-
-NCVStatus ncvDrawRects_8u_device(Ncv8u *d_dst,
-                                 Ncv32u dstStride,
-                                 Ncv32u dstWidth,
-                                 Ncv32u dstHeight,
-                                 NcvRect32u *d_rects,
-                                 Ncv32u numRects,
-                                 Ncv8u color,
-                                 cudaStream_t cuStream)
-{
-    return drawRectsWrapperDevice(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color, cuStream);
-}
-
-
-NCVStatus ncvDrawRects_32u_device(Ncv32u *d_dst,
-                                  Ncv32u dstStride,
-                                  Ncv32u dstWidth,
-                                  Ncv32u dstHeight,
-                                  NcvRect32u *d_rects,
-                                  Ncv32u numRects,
-                                  Ncv32u color,
-                                  cudaStream_t cuStream)
-{
-    return drawRectsWrapperDevice(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color, cuStream);
-}
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/src/cuda/NCV.cu b/modules/gpulegacy/src/cuda/NCV.cu
new file mode 100644
index 0000000000..8774af5c95
--- /dev/null
+++ b/modules/gpulegacy/src/cuda/NCV.cu
@@ -0,0 +1,180 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include <iostream>
+#include <vector>
+
+#include "opencv2/gpulegacy/NCV.hpp"
+
+//===================================================================
+//
+// Operations with rectangles
+//
+//===================================================================
+
+
+const Ncv32u NUMTHREADS_DRAWRECTS = 32;
+const Ncv32u NUMTHREADS_DRAWRECTS_LOG2 = 5;
+
+
+template <class T>
+__global__ void drawRects(T *d_dst,
+                          Ncv32u dstStride,
+                          Ncv32u dstWidth,
+                          Ncv32u dstHeight,
+                          NcvRect32u *d_rects,
+                          Ncv32u numRects,
+                          T color)
+{
+    Ncv32u blockId = blockIdx.y * 65535 + blockIdx.x;
+    if (blockId > numRects * 4)
+    {
+        return;
+    }
+
+    NcvRect32u curRect = d_rects[blockId >> 2];
+    NcvBool bVertical = blockId & 0x1;
+    NcvBool bTopLeft = blockId & 0x2;
+
+    Ncv32u pt0x, pt0y;
+    if (bVertical)
+    {
+        Ncv32u numChunks = (curRect.height + NUMTHREADS_DRAWRECTS - 1) >> NUMTHREADS_DRAWRECTS_LOG2;
+
+        pt0x = bTopLeft ? curRect.x : curRect.x + curRect.width - 1;
+        pt0y = curRect.y;
+
+        if (pt0x < dstWidth)
+        {
+            for (Ncv32u chunkId = 0; chunkId < numChunks; chunkId++)
+            {
+                Ncv32u ptY = pt0y + chunkId * NUMTHREADS_DRAWRECTS + threadIdx.x;
+                if (ptY < pt0y + curRect.height && ptY < dstHeight)
+                {
+                    d_dst[ptY * dstStride + pt0x] = color;
+                }
+            }
+        }
+    }
+    else
+    {
+        Ncv32u numChunks = (curRect.width + NUMTHREADS_DRAWRECTS - 1) >> NUMTHREADS_DRAWRECTS_LOG2;
+
+        pt0x = curRect.x;
+        pt0y = bTopLeft ? curRect.y : curRect.y + curRect.height - 1;
+
+        if (pt0y < dstHeight)
+        {
+            for (Ncv32u chunkId = 0; chunkId < numChunks; chunkId++)
+            {
+                Ncv32u ptX = pt0x + chunkId * NUMTHREADS_DRAWRECTS + threadIdx.x;
+                if (ptX < pt0x + curRect.width && ptX < dstWidth)
+                {
+                    d_dst[pt0y * dstStride + ptX] = color;
+                }
+            }
+        }
+    }
+}
+
+
+template <class T>
+static NCVStatus drawRectsWrapperDevice(T *d_dst,
+                                        Ncv32u dstStride,
+                                        Ncv32u dstWidth,
+                                        Ncv32u dstHeight,
+                                        NcvRect32u *d_rects,
+                                        Ncv32u numRects,
+                                        T color,
+                                        cudaStream_t cuStream)
+{
+    (void)cuStream;
+    ncvAssertReturn(d_dst != NULL && d_rects != NULL, NCV_NULL_PTR);
+    ncvAssertReturn(dstWidth > 0 && dstHeight > 0, NCV_DIMENSIONS_INVALID);
+    ncvAssertReturn(dstStride >= dstWidth, NCV_INVALID_STEP);
+    ncvAssertReturn(numRects <= dstWidth * dstHeight, NCV_DIMENSIONS_INVALID);
+
+    if (numRects == 0)
+    {
+        return NCV_SUCCESS;
+    }
+
+    dim3 grid(numRects * 4);
+    dim3 block(NUMTHREADS_DRAWRECTS);
+    if (grid.x > 65535)
+    {
+        grid.y = (grid.x + 65534) / 65535;
+        grid.x = 65535;
+    }
+
+    drawRects<T><<<grid, block>>>(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color);
+
+    ncvAssertCUDALastErrorReturn(NCV_CUDA_ERROR);
+
+    return NCV_SUCCESS;
+}
+
+
+NCVStatus ncvDrawRects_8u_device(Ncv8u *d_dst,
+                                 Ncv32u dstStride,
+                                 Ncv32u dstWidth,
+                                 Ncv32u dstHeight,
+                                 NcvRect32u *d_rects,
+                                 Ncv32u numRects,
+                                 Ncv8u color,
+                                 cudaStream_t cuStream)
+{
+    return drawRectsWrapperDevice(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color, cuStream);
+}
+
+
+NCVStatus ncvDrawRects_32u_device(Ncv32u *d_dst,
+                                  Ncv32u dstStride,
+                                  Ncv32u dstWidth,
+                                  Ncv32u dstHeight,
+                                  NcvRect32u *d_rects,
+                                  Ncv32u numRects,
+                                  Ncv32u color,
+                                  cudaStream_t cuStream)
+{
+    return drawRectsWrapperDevice(d_dst, dstStride, dstWidth, dstHeight, d_rects, numRects, color, cuStream);
+}
diff --git a/modules/gpu/src/nvidia/core/NCVAlg.hpp b/modules/gpulegacy/src/cuda/NCVAlg.hpp
similarity index 95%
rename from modules/gpu/src/nvidia/core/NCVAlg.hpp
rename to modules/gpulegacy/src/cuda/NCVAlg.hpp
index a1614720cc..dc086d944c 100644
--- a/modules/gpu/src/nvidia/core/NCVAlg.hpp
+++ b/modules/gpulegacy/src/cuda/NCVAlg.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -42,7 +43,7 @@
 #ifndef _ncv_alg_hpp_
 #define _ncv_alg_hpp_
 
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
 
 
 template <class T>
diff --git a/modules/gpu/src/nvidia/NCVBroxOpticalFlow.cu b/modules/gpulegacy/src/cuda/NCVBroxOpticalFlow.cu
similarity index 99%
rename from modules/gpu/src/nvidia/NCVBroxOpticalFlow.cu
rename to modules/gpulegacy/src/cuda/NCVBroxOpticalFlow.cu
index 5d1b46c448..d9848ad4b9 100644
--- a/modules/gpu/src/nvidia/NCVBroxOpticalFlow.cu
+++ b/modules/gpulegacy/src/cuda/NCVBroxOpticalFlow.cu
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -56,15 +57,14 @@
 //
 ////////////////////////////////////////////////////////////////////////////////
 
-#if !defined CUDA_DISABLER
-
 #include <iostream>
 #include <vector>
 #include <memory>
 
-#include "NPP_staging/NPP_staging.hpp"
-#include "NCVBroxOpticalFlow.hpp"
-#include "opencv2/gpu/device/utility.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+
+#include "opencv2/gpulegacy/NPP_staging.hpp"
+#include "opencv2/gpulegacy/NCVBroxOpticalFlow.hpp"
 
 
 typedef NCVVectorAlloc<Ncv32f> FloatVector;
@@ -1140,8 +1140,8 @@ NCVStatus NCVBroxOpticalFlow(const NCVBroxOpticalFlowDescriptor desc,
                 ScaleVector(ptrVNew->ptr(), ptrVNew->ptr(), 1.0f/scale_factor, ns * nh, stream);
                 ncvAssertCUDALastErrorReturn((int)NCV_CUDA_ERROR);
 
-                cv::gpu::device::swap<FloatVector*>(ptrU, ptrUNew);
-                cv::gpu::device::swap<FloatVector*>(ptrV, ptrVNew);
+                cv::gpu::cudev::swap<FloatVector*>(ptrU, ptrUNew);
+                cv::gpu::cudev::swap<FloatVector*>(ptrV, ptrVNew);
             }
             scale /= scale_factor;
         }
@@ -1162,5 +1162,3 @@ NCVStatus NCVBroxOpticalFlow(const NCVBroxOpticalFlowDescriptor desc,
 
     return NCV_SUCCESS;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/nvidia/core/NCVColorConversion.hpp b/modules/gpulegacy/src/cuda/NCVColorConversion.hpp
similarity index 95%
rename from modules/gpu/src/nvidia/core/NCVColorConversion.hpp
rename to modules/gpulegacy/src/cuda/NCVColorConversion.hpp
index f208cf06f6..c1293f2b34 100644
--- a/modules/gpu/src/nvidia/core/NCVColorConversion.hpp
+++ b/modules/gpulegacy/src/cuda/NCVColorConversion.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
diff --git a/modules/gpu/src/nvidia/NCVHaarObjectDetection.cu b/modules/gpulegacy/src/cuda/NCVHaarObjectDetection.cu
similarity index 91%
rename from modules/gpu/src/nvidia/NCVHaarObjectDetection.cu
rename to modules/gpulegacy/src/cuda/NCVHaarObjectDetection.cu
index fb057ae79d..3ac1e9491d 100644
--- a/modules/gpu/src/nvidia/NCVHaarObjectDetection.cu
+++ b/modules/gpulegacy/src/cuda/NCVHaarObjectDetection.cu
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -55,18 +56,25 @@
 //
 ////////////////////////////////////////////////////////////////////////////////
 
-#if !defined CUDA_DISABLER
-
 #include <algorithm>
 #include <cstdio>
 
-#include "NCV.hpp"
-#include "NCVAlg.hpp"
-#include "NPP_staging/NPP_staging.hpp"
+#include "opencv2/core/cuda/warp.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_OBJDETECT
+#  include "opencv2/objdetect.hpp"
+#  include "opencv2/objdetect/objdetect_c.h"
+#endif
+
+#include "opencv2/gpulegacy/NCV.hpp"
+#include "opencv2/gpulegacy/NPP_staging.hpp"
+#include "opencv2/gpulegacy/NCVHaarObjectDetection.hpp"
+
 #include "NCVRuntimeTemplates.hpp"
-#include "NCVHaarObjectDetection.hpp"
-#include "opencv2/gpu/device/warp.hpp"
-#include "opencv2/gpu/device/warp_shuffle.hpp"
+#include "NCVAlg.hpp"
 
 
 //==============================================================================
@@ -84,13 +92,13 @@ NCV_CT_ASSERT(K_WARP_SIZE == 32); //this is required for the manual unroll of th
 __device__ Ncv32u warpScanInclusive(Ncv32u idata, volatile Ncv32u *s_Data)
 {
 #if __CUDA_ARCH__ >= 300
-    const unsigned int laneId = cv::gpu::device::Warp::laneId();
+    const unsigned int laneId = cv::gpu::cudev::Warp::laneId();
 
     // scan on shuffl functions
     #pragma unroll
     for (int i = 1; i <= (K_WARP_SIZE / 2); i *= 2)
     {
-        const Ncv32u n = cv::gpu::device::shfl_up(idata, i);
+        const Ncv32u n = cv::gpu::cudev::shfl_up(idata, i);
         if (laneId >= i)
               idata += n;
     }
@@ -2098,19 +2106,218 @@ NCVStatus ncvGrowDetectionsVector_host(NCVVector<Ncv32u> &pixelMask,
     return ncvStat;
 }
 
-
-NCVStatus loadFromXML(const std::string &filename,
+static NCVStatus loadFromXML(const cv::String &filename,
                       HaarClassifierCascadeDescriptor &haar,
                       std::vector<HaarStage64> &haarStages,
                       std::vector<HaarClassifierNode128> &haarClassifierNodes,
-                      std::vector<HaarFeature64> &haarFeatures);
+                      std::vector<HaarFeature64> &haarFeatures)
+{
+#ifndef HAVE_OPENCV_OBJDETECT
+    (void) filename;
+    (void) haar;
+    (void) haarStages;
+    (void) haarClassifierNodes;
+    (void) haarFeatures;
+    CV_Error(cv::Error::StsNotImplemented, "This functionality requires objdetect module");
+    return NCV_HAAR_XML_LOADING_EXCEPTION;
+#else
+    NCVStatus ncvStat;
+
+    haar.NumStages = 0;
+    haar.NumClassifierRootNodes = 0;
+    haar.NumClassifierTotalNodes = 0;
+    haar.NumFeatures = 0;
+    haar.ClassifierSize.width = 0;
+    haar.ClassifierSize.height = 0;
+    haar.bHasStumpsOnly = true;
+    haar.bNeedsTiltedII = false;
+    Ncv32u curMaxTreeDepth = 0;
+
+    std::vector<HaarClassifierNode128> h_TmpClassifierNotRootNodes;
+    haarStages.resize(0);
+    haarClassifierNodes.resize(0);
+    haarFeatures.resize(0);
+
+    cv::Ptr<CvHaarClassifierCascade> oldCascade((CvHaarClassifierCascade*)cvLoad(filename.c_str(), 0, 0, 0));
+    if (!oldCascade)
+    {
+        return NCV_HAAR_XML_LOADING_EXCEPTION;
+    }
+
+    haar.ClassifierSize.width = oldCascade->orig_window_size.width;
+    haar.ClassifierSize.height = oldCascade->orig_window_size.height;
+
+    int stagesCound = oldCascade->count;
+    for(int s = 0; s < stagesCound; ++s) // by stages
+    {
+        HaarStage64 curStage;
+        curStage.setStartClassifierRootNodeOffset(static_cast<Ncv32u>(haarClassifierNodes.size()));
+
+        curStage.setStageThreshold(oldCascade->stage_classifier[s].threshold);
+
+        int treesCount = oldCascade->stage_classifier[s].count;
+        for(int t = 0; t < treesCount; ++t) // by trees
+        {
+            Ncv32u nodeId = 0;
+            CvHaarClassifier* tree = &oldCascade->stage_classifier[s].classifier[t];
+
+            int nodesCount = tree->count;
+            for(int n = 0; n < nodesCount; ++n)  //by features
+            {
+                CvHaarFeature* feature = &tree->haar_feature[n];
+
+                HaarClassifierNode128 curNode;
+                curNode.setThreshold(tree->threshold[n]);
+
+                NcvBool bIsLeftNodeLeaf = false;
+                NcvBool bIsRightNodeLeaf = false;
+
+                HaarClassifierNodeDescriptor32 nodeLeft;
+                if ( tree->left[n] <= 0 )
+                {
+                    Ncv32f leftVal = tree->alpha[-tree->left[n]];
+                    ncvStat = nodeLeft.create(leftVal);
+                    ncvAssertReturn(ncvStat == NCV_SUCCESS, ncvStat);
+                    bIsLeftNodeLeaf = true;
+                }
+                else
+                {
+                    Ncv32u leftNodeOffset = tree->left[n];
+                    nodeLeft.create((Ncv32u)(h_TmpClassifierNotRootNodes.size() + leftNodeOffset - 1));
+                    haar.bHasStumpsOnly = false;
+                }
+                curNode.setLeftNodeDesc(nodeLeft);
+
+                HaarClassifierNodeDescriptor32 nodeRight;
+                if ( tree->right[n] <= 0 )
+                {
+                    Ncv32f rightVal = tree->alpha[-tree->right[n]];
+                    ncvStat = nodeRight.create(rightVal);
+                    ncvAssertReturn(ncvStat == NCV_SUCCESS, ncvStat);
+                    bIsRightNodeLeaf = true;
+                }
+                else
+                {
+                    Ncv32u rightNodeOffset = tree->right[n];
+                    nodeRight.create((Ncv32u)(h_TmpClassifierNotRootNodes.size() + rightNodeOffset - 1));
+                    haar.bHasStumpsOnly = false;
+                }
+                curNode.setRightNodeDesc(nodeRight);
+
+                Ncv32u tiltedVal = feature->tilted;
+                haar.bNeedsTiltedII = (tiltedVal != 0);
+
+                Ncv32u featureId = 0;
+                for(int l = 0; l < CV_HAAR_FEATURE_MAX; ++l) //by rects
+                {
+                    Ncv32u rectX = feature->rect[l].r.x;
+                    Ncv32u rectY = feature->rect[l].r.y;
+                    Ncv32u rectWidth = feature->rect[l].r.width;
+                    Ncv32u rectHeight = feature->rect[l].r.height;
+
+                    Ncv32f rectWeight = feature->rect[l].weight;
+
+                    if (rectWeight == 0/* && rectX == 0 &&rectY == 0 && rectWidth == 0 && rectHeight == 0*/)
+                        break;
+
+                    HaarFeature64 curFeature;
+                    ncvStat = curFeature.setRect(rectX, rectY, rectWidth, rectHeight, haar.ClassifierSize.width, haar.ClassifierSize.height);
+                    curFeature.setWeight(rectWeight);
+                    ncvAssertReturn(NCV_SUCCESS == ncvStat, ncvStat);
+                    haarFeatures.push_back(curFeature);
+
+                    featureId++;
+                }
+
+                HaarFeatureDescriptor32 tmpFeatureDesc;
+                ncvStat = tmpFeatureDesc.create(haar.bNeedsTiltedII, bIsLeftNodeLeaf, bIsRightNodeLeaf,
+                    featureId, static_cast<Ncv32u>(haarFeatures.size()) - featureId);
+                ncvAssertReturn(NCV_SUCCESS == ncvStat, ncvStat);
+                curNode.setFeatureDesc(tmpFeatureDesc);
+
+                if (!nodeId)
+                {
+                    //root node
+                    haarClassifierNodes.push_back(curNode);
+                    curMaxTreeDepth = 1;
+                }
+                else
+                {
+                    //other node
+                    h_TmpClassifierNotRootNodes.push_back(curNode);
+                    curMaxTreeDepth++;
+                }
+
+                nodeId++;
+            }
+        }
+
+        curStage.setNumClassifierRootNodes(treesCount);
+        haarStages.push_back(curStage);
+    }
+
+    //fill in cascade stats
+    haar.NumStages = static_cast<Ncv32u>(haarStages.size());
+    haar.NumClassifierRootNodes = static_cast<Ncv32u>(haarClassifierNodes.size());
+    haar.NumClassifierTotalNodes = static_cast<Ncv32u>(haar.NumClassifierRootNodes + h_TmpClassifierNotRootNodes.size());
+    haar.NumFeatures = static_cast<Ncv32u>(haarFeatures.size());
+
+    //merge root and leaf nodes in one classifiers array
+    Ncv32u offsetRoot = static_cast<Ncv32u>(haarClassifierNodes.size());
+    for (Ncv32u i=0; i<haarClassifierNodes.size(); i++)
+    {
+        HaarFeatureDescriptor32 featureDesc = haarClassifierNodes[i].getFeatureDesc();
+
+        HaarClassifierNodeDescriptor32 nodeLeft = haarClassifierNodes[i].getLeftNodeDesc();
+        if (!featureDesc.isLeftNodeLeaf())
+        {
+            Ncv32u newOffset = nodeLeft.getNextNodeOffset() + offsetRoot;
+            nodeLeft.create(newOffset);
+        }
+        haarClassifierNodes[i].setLeftNodeDesc(nodeLeft);
+
+        HaarClassifierNodeDescriptor32 nodeRight = haarClassifierNodes[i].getRightNodeDesc();
+        if (!featureDesc.isRightNodeLeaf())
+        {
+            Ncv32u newOffset = nodeRight.getNextNodeOffset() + offsetRoot;
+            nodeRight.create(newOffset);
+        }
+        haarClassifierNodes[i].setRightNodeDesc(nodeRight);
+    }
+
+    for (Ncv32u i=0; i<h_TmpClassifierNotRootNodes.size(); i++)
+    {
+        HaarFeatureDescriptor32 featureDesc = h_TmpClassifierNotRootNodes[i].getFeatureDesc();
+
+        HaarClassifierNodeDescriptor32 nodeLeft = h_TmpClassifierNotRootNodes[i].getLeftNodeDesc();
+        if (!featureDesc.isLeftNodeLeaf())
+        {
+            Ncv32u newOffset = nodeLeft.getNextNodeOffset() + offsetRoot;
+            nodeLeft.create(newOffset);
+        }
+        h_TmpClassifierNotRootNodes[i].setLeftNodeDesc(nodeLeft);
+
+        HaarClassifierNodeDescriptor32 nodeRight = h_TmpClassifierNotRootNodes[i].getRightNodeDesc();
+        if (!featureDesc.isRightNodeLeaf())
+        {
+            Ncv32u newOffset = nodeRight.getNextNodeOffset() + offsetRoot;
+            nodeRight.create(newOffset);
+        }
+        h_TmpClassifierNotRootNodes[i].setRightNodeDesc(nodeRight);
+
+        haarClassifierNodes.push_back(h_TmpClassifierNotRootNodes[i]);
+    }
+
+    return NCV_SUCCESS;
+#endif
+}
 
 
 #define NVBIN_HAAR_SIZERESERVED     16
 #define NVBIN_HAAR_VERSION          0x1
 
 
-static NCVStatus loadFromNVBIN(const std::string &filename,
+static NCVStatus loadFromNVBIN(const cv::String &filename,
                                HaarClassifierCascadeDescriptor &haar,
                                std::vector<HaarStage64> &haarStages,
                                std::vector<HaarClassifierNode128> &haarClassifierNodes,
@@ -2174,14 +2381,14 @@ static NCVStatus loadFromNVBIN(const std::string &filename,
 }
 
 
-NCVStatus ncvHaarGetClassifierSize(const std::string &filename, Ncv32u &numStages,
+NCVStatus ncvHaarGetClassifierSize(const cv::String &filename, Ncv32u &numStages,
                                    Ncv32u &numNodes, Ncv32u &numFeatures)
 {
     size_t readCount;
     NCVStatus ncvStat;
 
-    std::string fext = filename.substr(filename.find_last_of(".") + 1);
-    std::transform(fext.begin(), fext.end(), fext.begin(), ::tolower);
+    cv::String fext = filename.substr(filename.find_last_of(".") + 1);
+    fext = fext.toLowerCase();
 
     if (fext == "nvbin")
     {
@@ -2226,7 +2433,7 @@ NCVStatus ncvHaarGetClassifierSize(const std::string &filename, Ncv32u &numStage
 }
 
 
-NCVStatus ncvHaarLoadFromFile_host(const std::string &filename,
+NCVStatus ncvHaarLoadFromFile_host(const cv::String &filename,
                                    HaarClassifierCascadeDescriptor &haar,
                                    NCVVector<HaarStage64> &h_HaarStages,
                                    NCVVector<HaarClassifierNode128> &h_HaarNodes,
@@ -2238,8 +2445,8 @@ NCVStatus ncvHaarLoadFromFile_host(const std::string &filename,
 
     NCVStatus ncvStat;
 
-    std::string fext = filename.substr(filename.find_last_of(".") + 1);
-    std::transform(fext.begin(), fext.end(), fext.begin(), ::tolower);
+    cv::String fext = filename.substr(filename.find_last_of(".") + 1);
+    fext = fext.toLowerCase();
 
     std::vector<HaarStage64> haarStages;
     std::vector<HaarClassifierNode128> haarNodes;
@@ -2272,7 +2479,7 @@ NCVStatus ncvHaarLoadFromFile_host(const std::string &filename,
 }
 
 
-NCVStatus ncvHaarStoreNVBIN_host(const std::string &filename,
+NCVStatus ncvHaarStoreNVBIN_host(const cv::String &filename,
                                  HaarClassifierCascadeDescriptor haar,
                                  NCVVector<HaarStage64> &h_HaarStages,
                                  NCVVector<HaarClassifierNode128> &h_HaarNodes,
@@ -2333,5 +2540,3 @@ NCVStatus ncvHaarStoreNVBIN_host(const std::string &filename,
     fclose(fp);
     return NCV_SUCCESS;
 }
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/nvidia/core/NCVPixelOperations.hpp b/modules/gpulegacy/src/cuda/NCVPixelOperations.hpp
similarity index 98%
rename from modules/gpu/src/nvidia/core/NCVPixelOperations.hpp
rename to modules/gpulegacy/src/cuda/NCVPixelOperations.hpp
index de78cf9deb..5cf902a9f4 100644
--- a/modules/gpu/src/nvidia/core/NCVPixelOperations.hpp
+++ b/modules/gpulegacy/src/cuda/NCVPixelOperations.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -44,13 +45,13 @@
 
 #include <limits.h>
 #include <float.h>
-#include "NCV.hpp"
+#include "opencv2/gpulegacy/NCV.hpp"
 
 template<typename TBase> inline __host__ __device__ TBase _pixMaxVal();
 template<> static inline __host__ __device__ Ncv8u  _pixMaxVal<Ncv8u>()  {return UCHAR_MAX;}
 template<> static inline __host__ __device__ Ncv16u _pixMaxVal<Ncv16u>() {return USHRT_MAX;}
 template<> static inline __host__ __device__ Ncv32u _pixMaxVal<Ncv32u>() {return  UINT_MAX;}
-template<> static inline __host__ __device__ Ncv8s  _pixMaxVal<Ncv8s>()  {return  CHAR_MAX;}
+template<> static inline __host__ __device__ Ncv8s  _pixMaxVal<Ncv8s>()  {return  SCHAR_MAX;}
 template<> static inline __host__ __device__ Ncv16s _pixMaxVal<Ncv16s>() {return  SHRT_MAX;}
 template<> static inline __host__ __device__ Ncv32s _pixMaxVal<Ncv32s>() {return   INT_MAX;}
 template<> static inline __host__ __device__ Ncv32f _pixMaxVal<Ncv32f>() {return   FLT_MAX;}
@@ -60,7 +61,7 @@ template<typename TBase> inline __host__ __device__ TBase _pixMinVal();
 template<> static inline __host__ __device__ Ncv8u  _pixMinVal<Ncv8u>()  {return 0;}
 template<> static inline __host__ __device__ Ncv16u _pixMinVal<Ncv16u>() {return 0;}
 template<> static inline __host__ __device__ Ncv32u _pixMinVal<Ncv32u>() {return 0;}
-template<> static inline __host__ __device__ Ncv8s  _pixMinVal<Ncv8s>()  {return CHAR_MIN;}
+template<> static inline __host__ __device__ Ncv8s  _pixMinVal<Ncv8s>()  {return SCHAR_MIN;}
 template<> static inline __host__ __device__ Ncv16s _pixMinVal<Ncv16s>() {return SHRT_MIN;}
 template<> static inline __host__ __device__ Ncv32s _pixMinVal<Ncv32s>() {return INT_MIN;}
 template<> static inline __host__ __device__ Ncv32f _pixMinVal<Ncv32f>() {return FLT_MIN;}
diff --git a/modules/gpu/src/nvidia/core/NCVPyramid.cu b/modules/gpulegacy/src/cuda/NCVPyramid.cu
similarity index 86%
rename from modules/gpu/src/nvidia/core/NCVPyramid.cu
rename to modules/gpulegacy/src/cuda/NCVPyramid.cu
index a3d8954786..d42b46bcb5 100644
--- a/modules/gpu/src/nvidia/core/NCVPyramid.cu
+++ b/modules/gpulegacy/src/cuda/NCVPyramid.cu
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -39,15 +40,16 @@
 //
 //M*/
 
-#if !defined CUDA_DISABLER
-
-#include <cuda_runtime.h>
 #include <stdio.h>
-#include "NCV.hpp"
+#include <cuda_runtime.h>
+
+#include "opencv2/core/cuda/common.hpp"
+
+#include "opencv2/gpulegacy/NCV.hpp"
+#include "opencv2/gpulegacy/NCVPyramid.hpp"
+
 #include "NCVAlg.hpp"
-#include "NCVPyramid.hpp"
 #include "NCVPixelOperations.hpp"
-#include "opencv2/gpu/device/common.hpp"
 
 template<typename T, Ncv32u CN> struct __average4_CN {static __host__ __device__ T _average4_CN(const T &p00, const T &p01, const T &p10, const T &p11);};
 
@@ -203,7 +205,7 @@ __global__ void kernelDownsampleX2(T *d_src,
     }
 }
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace pyramid
     {
@@ -221,17 +223,25 @@ namespace cv { namespace gpu { namespace device
                 cudaSafeCall( cudaDeviceSynchronize() );
         }
 
-        template void kernelDownsampleX2_gpu<uchar1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<uchar3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<uchar4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+        void downsampleX2(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream)
+        {
+            typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 
-        template void kernelDownsampleX2_gpu<ushort1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<ushort3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<ushort4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+            static const func_t funcs[6][4] =
+            {
+                {kernelDownsampleX2_gpu<uchar1>       , 0 /*kernelDownsampleX2_gpu<uchar2>*/ , kernelDownsampleX2_gpu<uchar3>      , kernelDownsampleX2_gpu<uchar4>      },
+                {0 /*kernelDownsampleX2_gpu<char1>*/  , 0 /*kernelDownsampleX2_gpu<char2>*/  , 0 /*kernelDownsampleX2_gpu<char3>*/ , 0 /*kernelDownsampleX2_gpu<char4>*/ },
+                {kernelDownsampleX2_gpu<ushort1>      , 0 /*kernelDownsampleX2_gpu<ushort2>*/, kernelDownsampleX2_gpu<ushort3>     , kernelDownsampleX2_gpu<ushort4>     },
+                {0 /*kernelDownsampleX2_gpu<short1>*/ , 0 /*kernelDownsampleX2_gpu<short2>*/ , 0 /*kernelDownsampleX2_gpu<short3>*/, 0 /*kernelDownsampleX2_gpu<short4>*/},
+                {0 /*kernelDownsampleX2_gpu<int1>*/   , 0 /*kernelDownsampleX2_gpu<int2>*/   , 0 /*kernelDownsampleX2_gpu<int3>*/  , 0 /*kernelDownsampleX2_gpu<int4>*/  },
+                {kernelDownsampleX2_gpu<float1>       , 0 /*kernelDownsampleX2_gpu<float2>*/ , kernelDownsampleX2_gpu<float3>      , kernelDownsampleX2_gpu<float4>      }
+            };
 
-        template void kernelDownsampleX2_gpu<float1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelDownsampleX2_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+            const func_t func = funcs[depth][cn - 1];
+            CV_Assert(func != 0);
+
+            func(src, dst, stream);
+        }
     }
 }}}
 
@@ -278,7 +288,7 @@ __global__ void kernelInterpolateFrom1(T *d_srcTop,
         d_dst_line[j] = outPix;
     }
 }
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace pyramid
     {
@@ -296,17 +306,25 @@ namespace cv { namespace gpu { namespace device
                 cudaSafeCall( cudaDeviceSynchronize() );
         }
 
-        template void kernelInterpolateFrom1_gpu<uchar1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<uchar3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<uchar4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+        void interpolateFrom1(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream)
+        {
+            typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
 
-        template void kernelInterpolateFrom1_gpu<ushort1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<ushort3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<ushort4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+            static const func_t funcs[6][4] =
+            {
+                {kernelInterpolateFrom1_gpu<uchar1>      , 0 /*kernelInterpolateFrom1_gpu<uchar2>*/ , kernelInterpolateFrom1_gpu<uchar3>      , kernelInterpolateFrom1_gpu<uchar4>      },
+                {0 /*kernelInterpolateFrom1_gpu<char1>*/ , 0 /*kernelInterpolateFrom1_gpu<char2>*/  , 0 /*kernelInterpolateFrom1_gpu<char3>*/ , 0 /*kernelInterpolateFrom1_gpu<char4>*/ },
+                {kernelInterpolateFrom1_gpu<ushort1>     , 0 /*kernelInterpolateFrom1_gpu<ushort2>*/, kernelInterpolateFrom1_gpu<ushort3>     , kernelInterpolateFrom1_gpu<ushort4>     },
+                {0 /*kernelInterpolateFrom1_gpu<short1>*/, 0 /*kernelInterpolateFrom1_gpu<short2>*/ , 0 /*kernelInterpolateFrom1_gpu<short3>*/, 0 /*kernelInterpolateFrom1_gpu<short4>*/},
+                {0 /*kernelInterpolateFrom1_gpu<int1>*/  , 0 /*kernelInterpolateFrom1_gpu<int2>*/   , 0 /*kernelInterpolateFrom1_gpu<int3>*/  , 0 /*kernelInterpolateFrom1_gpu<int4>*/  },
+                {kernelInterpolateFrom1_gpu<float1>      , 0 /*kernelInterpolateFrom1_gpu<float2>*/ , kernelInterpolateFrom1_gpu<float3>      , kernelInterpolateFrom1_gpu<float4>      }
+            };
 
-        template void kernelInterpolateFrom1_gpu<float1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-        template void kernelInterpolateFrom1_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+            const func_t func = funcs[depth][cn - 1];
+            CV_Assert(func != 0);
+
+            func(src, dst, stream);
+        }
     }
 }}}
 
@@ -601,5 +619,3 @@ template class NCVImagePyramid<float3>;
 template class NCVImagePyramid<float4>;
 
 #endif //_WIN32
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpu/src/nvidia/core/NCVRuntimeTemplates.hpp b/modules/gpulegacy/src/cuda/NCVRuntimeTemplates.hpp
similarity index 97%
rename from modules/gpu/src/nvidia/core/NCVRuntimeTemplates.hpp
rename to modules/gpulegacy/src/cuda/NCVRuntimeTemplates.hpp
index 7b248a8324..33f001991b 100644
--- a/modules/gpu/src/nvidia/core/NCVRuntimeTemplates.hpp
+++ b/modules/gpulegacy/src/cuda/NCVRuntimeTemplates.hpp
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
diff --git a/modules/gpu/src/nvidia/NPP_staging/NPP_staging.cu b/modules/gpulegacy/src/cuda/NPP_staging.cu
similarity index 99%
rename from modules/gpu/src/nvidia/NPP_staging/NPP_staging.cu
rename to modules/gpulegacy/src/cuda/NPP_staging.cu
index f4ec9aace6..9234e1795b 100644
--- a/modules/gpu/src/nvidia/NPP_staging/NPP_staging.cu
+++ b/modules/gpulegacy/src/cuda/NPP_staging.cu
@@ -1,6 +1,6 @@
 /*M///////////////////////////////////////////////////////////////////////////////////////
 //
-// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
 //  By downloading, copying, installing or using the software you agree to this license.
 //  If you do not agree to this license, do not download, install,
@@ -10,7 +10,8 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2009-2010, NVIDIA Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -43,9 +44,11 @@
 
 #include <vector>
 #include <cuda_runtime.h>
-#include "NPP_staging.hpp"
-#include "opencv2/gpu/device/warp.hpp"
-#include "opencv2/gpu/device/warp_shuffle.hpp"
+
+#include "opencv2/core/cuda/warp.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
+
+#include "opencv2/gpulegacy/NPP_staging.hpp"
 
 
 texture<Ncv8u,  1, cudaReadModeElementType> tex8u;
@@ -94,13 +97,13 @@ template <class T>
 inline __device__ T warpScanInclusive(T idata, volatile T *s_Data)
 {
 #if __CUDA_ARCH__ >= 300
-    const unsigned int laneId = cv::gpu::device::Warp::laneId();
+    const unsigned int laneId = cv::gpu::cudev::Warp::laneId();
 
     // scan on shuffl functions
     #pragma unroll
     for (int i = 1; i <= (K_WARP_SIZE / 2); i *= 2)
     {
-        const T n = cv::gpu::device::shfl_up(idata, i);
+        const T n = cv::gpu::cudev::shfl_up(idata, i);
         if (laneId >= i)
               idata += n;
     }
diff --git a/modules/gpulegacy/src/precomp.hpp b/modules/gpulegacy/src/precomp.hpp
new file mode 100644
index 0000000000..85e6f88e73
--- /dev/null
+++ b/modules/gpulegacy/src/precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+#include <iostream>
+#include <algorithm>
+
+#include "opencv2/gpulegacy.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_OBJDETECT
+#  include "opencv2/objdetect.hpp"
+#endif
+
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/gpulegacy/private.hpp"
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/test/nvidia/NCVAutoTestLister.hpp b/modules/gpulegacy/test/NCVAutoTestLister.hpp
similarity index 58%
rename from modules/gpu/test/nvidia/NCVAutoTestLister.hpp
rename to modules/gpulegacy/test/NCVAutoTestLister.hpp
index 48ad754940..8730eeea7e 100644
--- a/modules/gpu/test/nvidia/NCVAutoTestLister.hpp
+++ b/modules/gpulegacy/test/NCVAutoTestLister.hpp
@@ -1,26 +1,52 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifndef _ncvautotestlister_hpp_
 #define _ncvautotestlister_hpp_
 
 #include <vector>
 
 #include "NCVTest.hpp"
-#include <main_test_nvidia.h>
-//enum OutputLevel
-//{
-//    OutputLevelNone,
-//    OutputLevelCompact,
-//    OutputLevelFull
-//};
+#include "main_test_nvidia.h"
 
 class NCVAutoTestLister
 {
diff --git a/modules/gpu/test/nvidia/NCVTest.hpp b/modules/gpulegacy/test/NCVTest.hpp
similarity index 70%
rename from modules/gpu/test/nvidia/NCVTest.hpp
rename to modules/gpulegacy/test/NCVTest.hpp
index d2289eea02..b79c55e085 100644
--- a/modules/gpu/test/nvidia/NCVTest.hpp
+++ b/modules/gpulegacy/test/NCVTest.hpp
@@ -1,13 +1,45 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifndef _ncvtest_hpp_
 #define _ncvtest_hpp_
 
@@ -23,7 +55,8 @@
 #include <fstream>
 
 #include <cuda_runtime.h>
-#include "NPP_staging.hpp"
+
+#include "opencv2/gpulegacy.hpp"
 
 
 struct NCVTestReport
diff --git a/modules/gpu/test/nvidia/NCVTestSourceProvider.hpp b/modules/gpulegacy/test/NCVTestSourceProvider.hpp
similarity index 68%
rename from modules/gpu/test/nvidia/NCVTestSourceProvider.hpp
rename to modules/gpulegacy/test/NCVTestSourceProvider.hpp
index 0ae104c96c..f2ffae56b2 100644
--- a/modules/gpu/test/nvidia/NCVTestSourceProvider.hpp
+++ b/modules/gpulegacy/test/NCVTestSourceProvider.hpp
@@ -1,20 +1,52 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #ifndef _ncvtestsourceprovider_hpp_
 #define _ncvtestsourceprovider_hpp_
 
 #include <memory>
 
-#include "NCV.hpp"
-#include <opencv2/highgui.hpp>
+#include "opencv2/highgui.hpp"
+#include "opencv2/gpulegacy.hpp"
 
 
 template <class T>
diff --git a/modules/gpu/test/nvidia/TestCompact.cpp b/modules/gpulegacy/test/TestCompact.cpp
similarity index 61%
rename from modules/gpu/test/nvidia/TestCompact.cpp
rename to modules/gpulegacy/test/TestCompact.cpp
index 273b56b501..70640f37db 100644
--- a/modules/gpu/test/nvidia/TestCompact.cpp
+++ b/modules/gpulegacy/test/TestCompact.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
-
-#if !defined CUDA_DISABLER
-
-#include "TestCompact.h"
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
+#include "test_precomp.hpp"
 
 TestCompact::TestCompact(std::string testName_, NCVTestSourceProvider<Ncv32u> &src_,
                                              Ncv32u length_, Ncv32u badElem_, Ncv32u badElemPercentage_)
@@ -129,5 +157,3 @@ bool TestCompact::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestCompact.h b/modules/gpulegacy/test/TestCompact.h
new file mode 100644
index 0000000000..256cdf4cf9
--- /dev/null
+++ b/modules/gpulegacy/test/TestCompact.h
@@ -0,0 +1,73 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testhypothesescompact_h_
+#define _testhypothesescompact_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestCompact : public NCVTestProvider
+{
+public:
+
+    TestCompact(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
+                          Ncv32u length, Ncv32u badElem, Ncv32u badElemPercentage);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestCompact(const TestCompact&);
+    TestCompact& operator=(const TestCompact&);
+
+
+    NCVTestSourceProvider<Ncv32u> &src;
+    Ncv32u length;
+    Ncv32u badElem;
+    Ncv32u badElemPercentage;
+};
+
+#endif // _testhypothesescompact_h_
diff --git a/modules/gpu/test/nvidia/TestDrawRects.cpp b/modules/gpulegacy/test/TestDrawRects.cpp
similarity index 70%
rename from modules/gpu/test/nvidia/TestDrawRects.cpp
rename to modules/gpulegacy/test/TestDrawRects.cpp
index 1525772fcd..40d8e21b49 100644
--- a/modules/gpu/test/nvidia/TestDrawRects.cpp
+++ b/modules/gpulegacy/test/TestDrawRects.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include "TestDrawRects.h"
-#include "NCVHaarObjectDetection.hpp"
+#include "test_precomp.hpp"
 
 
 template <class T>
@@ -164,5 +192,3 @@ bool TestDrawRects<T>::deinit()
 
 template class TestDrawRects<Ncv8u>;
 template class TestDrawRects<Ncv32u>;
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestDrawRects.h b/modules/gpulegacy/test/TestDrawRects.h
new file mode 100644
index 0000000000..b64c133d50
--- /dev/null
+++ b/modules/gpulegacy/test/TestDrawRects.h
@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testdrawrects_h_
+#define _testdrawrects_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+template <class T>
+class TestDrawRects : public NCVTestProvider
+{
+public:
+
+    TestDrawRects(std::string testName, NCVTestSourceProvider<T> &src, NCVTestSourceProvider<Ncv32u> &src32u,
+                  Ncv32u width, Ncv32u height, Ncv32u numRects, T color);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+
+    TestDrawRects(const TestDrawRects&);
+    TestDrawRects& operator=(const TestDrawRects&);
+
+    NCVTestSourceProvider<T> &src;
+    NCVTestSourceProvider<Ncv32u> &src32u;
+    Ncv32u width;
+    Ncv32u height;
+    Ncv32u numRects;
+    T color;
+};
+
+#endif // _testdrawrects_h_
diff --git a/modules/gpu/test/nvidia/TestHaarCascadeApplication.cpp b/modules/gpulegacy/test/TestHaarCascadeApplication.cpp
similarity index 83%
rename from modules/gpu/test/nvidia/TestHaarCascadeApplication.cpp
rename to modules/gpulegacy/test/TestHaarCascadeApplication.cpp
index 4867189606..121f31e434 100644
--- a/modules/gpu/test/nvidia/TestHaarCascadeApplication.cpp
+++ b/modules/gpulegacy/test/TestHaarCascadeApplication.cpp
@@ -1,21 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include <float.h>
-
-#if defined(__GNUC__) && !defined(__APPLE__) && !defined(__arm__)
-    #include <fpu_control.h>
-#endif
+#include "test_precomp.hpp"
 
 namespace
 {
@@ -57,10 +82,6 @@ namespace
     }
 }
 
-#include "TestHaarCascadeApplication.h"
-#include "NCVHaarObjectDetection.hpp"
-
-
 TestHaarCascadeApplication::TestHaarCascadeApplication(std::string testName_, NCVTestSourceProvider<Ncv8u> &src_,
                                                        std::string cascadeName_, Ncv32u width_, Ncv32u height_)
     :
@@ -312,5 +333,3 @@ bool TestHaarCascadeApplication::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpulegacy/test/TestHaarCascadeApplication.h b/modules/gpulegacy/test/TestHaarCascadeApplication.h
new file mode 100644
index 0000000000..eaa4371745
--- /dev/null
+++ b/modules/gpulegacy/test/TestHaarCascadeApplication.h
@@ -0,0 +1,73 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testhaarcascadeapplication_h_
+#define _testhaarcascadeapplication_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestHaarCascadeApplication : public NCVTestProvider
+{
+public:
+
+    TestHaarCascadeApplication(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
+                               std::string cascadeName, Ncv32u width, Ncv32u height);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestHaarCascadeApplication(const TestHaarCascadeApplication&);
+    TestHaarCascadeApplication& operator=(const TestHaarCascadeApplication&);
+
+
+    NCVTestSourceProvider<Ncv8u> &src;
+    std::string cascadeName;
+    Ncv32u width;
+    Ncv32u height;
+};
+
+#endif // _testhaarcascadeapplication_h_
diff --git a/modules/gpu/test/nvidia/TestHaarCascadeLoader.cpp b/modules/gpulegacy/test/TestHaarCascadeLoader.cpp
similarity index 64%
rename from modules/gpu/test/nvidia/TestHaarCascadeLoader.cpp
rename to modules/gpulegacy/test/TestHaarCascadeLoader.cpp
index 099596de0c..b1e840a54c 100644
--- a/modules/gpu/test/nvidia/TestHaarCascadeLoader.cpp
+++ b/modules/gpulegacy/test/TestHaarCascadeLoader.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include "TestHaarCascadeLoader.h"
-#include "NCVHaarObjectDetection.hpp"
+#include "test_precomp.hpp"
 
 
 TestHaarCascadeLoader::TestHaarCascadeLoader(std::string testName_, std::string cascadeName_)
@@ -123,5 +151,3 @@ bool TestHaarCascadeLoader::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestHaarCascadeLoader.h b/modules/gpulegacy/test/TestHaarCascadeLoader.h
new file mode 100644
index 0000000000..86c6324635
--- /dev/null
+++ b/modules/gpulegacy/test/TestHaarCascadeLoader.h
@@ -0,0 +1,66 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testhaarcascadeloader_h_
+#define _testhaarcascadeloader_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestHaarCascadeLoader : public NCVTestProvider
+{
+public:
+
+    TestHaarCascadeLoader(std::string testName, std::string cascadeName);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+
+    std::string cascadeName;
+};
+
+#endif // _testhaarcascadeloader_h_
diff --git a/modules/gpu/test/nvidia/TestHypothesesFilter.cpp b/modules/gpulegacy/test/TestHypothesesFilter.cpp
similarity index 74%
rename from modules/gpu/test/nvidia/TestHypothesesFilter.cpp
rename to modules/gpulegacy/test/TestHypothesesFilter.cpp
index 7899cd98c1..39d6556616 100644
--- a/modules/gpu/test/nvidia/TestHypothesesFilter.cpp
+++ b/modules/gpulegacy/test/TestHypothesesFilter.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include "TestHypothesesFilter.h"
-#include "NCVHaarObjectDetection.hpp"
+#include "test_precomp.hpp"
 
 
 TestHypothesesFilter::TestHypothesesFilter(std::string testName_, NCVTestSourceProvider<Ncv32u> &src_,
@@ -176,5 +204,3 @@ bool TestHypothesesFilter::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestHypothesesFilter.h b/modules/gpulegacy/test/TestHypothesesFilter.h
new file mode 100644
index 0000000000..f190785fe7
--- /dev/null
+++ b/modules/gpulegacy/test/TestHypothesesFilter.h
@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testhypothesesfilter_h_
+#define _testhypothesesfilter_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestHypothesesFilter : public NCVTestProvider
+{
+public:
+
+    TestHypothesesFilter(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
+                         Ncv32u numDstRects, Ncv32u minNeighbors, Ncv32f eps);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+
+    TestHypothesesFilter(const TestHypothesesFilter&);
+    TestHypothesesFilter& operator=(const TestHypothesesFilter&);
+
+    NCVTestSourceProvider<Ncv32u> &src;
+    Ncv32u numDstRects;
+    Ncv32u minNeighbors;
+    Ncv32f eps;
+
+    Ncv32u canvasWidth;
+    Ncv32u canvasHeight;
+};
+
+#endif // _testhypothesesfilter_h_
diff --git a/modules/gpu/test/nvidia/TestHypothesesGrow.cpp b/modules/gpulegacy/test/TestHypothesesGrow.cpp
similarity index 65%
rename from modules/gpu/test/nvidia/TestHypothesesGrow.cpp
rename to modules/gpulegacy/test/TestHypothesesGrow.cpp
index 685c50056f..e7fe4d939d 100644
--- a/modules/gpu/test/nvidia/TestHypothesesGrow.cpp
+++ b/modules/gpulegacy/test/TestHypothesesGrow.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include "TestHypothesesGrow.h"
-#include "NCVHaarObjectDetection.hpp"
+#include "test_precomp.hpp"
 
 
 TestHypothesesGrow::TestHypothesesGrow(std::string testName_, NCVTestSourceProvider<Ncv32u> &src_,
@@ -134,5 +162,3 @@ bool TestHypothesesGrow::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestHypothesesGrow.h b/modules/gpulegacy/test/TestHypothesesGrow.h
new file mode 100644
index 0000000000..51e879f697
--- /dev/null
+++ b/modules/gpulegacy/test/TestHypothesesGrow.h
@@ -0,0 +1,78 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testhypothesesgrow_h_
+#define _testhypothesesgrow_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestHypothesesGrow : public NCVTestProvider
+{
+public:
+
+    TestHypothesesGrow(std::string testName, NCVTestSourceProvider<Ncv32u> &src,
+                       Ncv32u rectWidth, Ncv32u rectHeight, Ncv32f rectScale,
+                       Ncv32u maxLenSrc, Ncv32u lenSrc, Ncv32u maxLenDst, Ncv32u lenDst);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestHypothesesGrow(const TestHypothesesGrow&);
+    TestHypothesesGrow& operator=(const TestHypothesesGrow&);
+
+
+    NCVTestSourceProvider<Ncv32u> &src;
+    Ncv32u rectWidth;
+    Ncv32u rectHeight;
+    Ncv32f rectScale;
+    Ncv32u maxLenSrc;
+    Ncv32u lenSrc;
+    Ncv32u maxLenDst;
+    Ncv32u lenDst;
+};
+
+#endif // _testhypothesesgrow_h_
diff --git a/modules/gpu/test/nvidia/TestIntegralImage.cpp b/modules/gpulegacy/test/TestIntegralImage.cpp
similarity index 72%
rename from modules/gpu/test/nvidia/TestIntegralImage.cpp
rename to modules/gpulegacy/test/TestIntegralImage.cpp
index 4944e14f99..c04edff7c9 100644
--- a/modules/gpu/test/nvidia/TestIntegralImage.cpp
+++ b/modules/gpulegacy/test/TestIntegralImage.cpp
@@ -1,18 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include <math.h>
-#include "TestIntegralImage.h"
+#include "test_precomp.hpp"
 
 
 template <class T_in, class T_out>
@@ -185,5 +213,3 @@ bool TestIntegralImage<T_in, T_out>::deinit()
 
 template class TestIntegralImage<Ncv8u, Ncv32u>;
 template class TestIntegralImage<Ncv32f, Ncv32f>;
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestIntegralImage.h b/modules/gpulegacy/test/TestIntegralImage.h
new file mode 100644
index 0000000000..c4f58ba146
--- /dev/null
+++ b/modules/gpulegacy/test/TestIntegralImage.h
@@ -0,0 +1,72 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testintegralimage_h_
+#define _testintegralimage_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+template <class T_in, class T_out>
+class TestIntegralImage : public NCVTestProvider
+{
+public:
+
+    TestIntegralImage(std::string testName, NCVTestSourceProvider<T_in> &src,
+                      Ncv32u width, Ncv32u height);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestIntegralImage(const TestIntegralImage&);
+    TestIntegralImage& operator=(const TestIntegralImage&);
+
+    NCVTestSourceProvider<T_in> &src;
+    Ncv32u width;
+    Ncv32u height;
+};
+
+#endif // _testintegralimage_h_
diff --git a/modules/gpu/test/nvidia/TestIntegralImageSquared.cpp b/modules/gpulegacy/test/TestIntegralImageSquared.cpp
similarity index 60%
rename from modules/gpu/test/nvidia/TestIntegralImageSquared.cpp
rename to modules/gpulegacy/test/TestIntegralImageSquared.cpp
index 6366fc4ce2..5481fa2e3a 100644
--- a/modules/gpu/test/nvidia/TestIntegralImageSquared.cpp
+++ b/modules/gpulegacy/test/TestIntegralImageSquared.cpp
@@ -1,17 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include "TestIntegralImageSquared.h"
+#include "test_precomp.hpp"
 
 
 TestIntegralImageSquared::TestIntegralImageSquared(std::string testName_, NCVTestSourceProvider<Ncv8u> &src_,
@@ -117,5 +146,3 @@ bool TestIntegralImageSquared::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestIntegralImageSquared.h b/modules/gpulegacy/test/TestIntegralImageSquared.h
new file mode 100644
index 0000000000..20e5ca8df3
--- /dev/null
+++ b/modules/gpulegacy/test/TestIntegralImageSquared.h
@@ -0,0 +1,71 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testintegralimagesquared_h_
+#define _testintegralimagesquared_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestIntegralImageSquared : public NCVTestProvider
+{
+public:
+
+    TestIntegralImageSquared(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
+                             Ncv32u width, Ncv32u height);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestIntegralImageSquared(const TestIntegralImageSquared&);
+    TestIntegralImageSquared& operator=(const TestIntegralImageSquared&);
+
+    NCVTestSourceProvider<Ncv8u> &src;
+    Ncv32u width;
+    Ncv32u height;
+};
+
+#endif // _testintegralimagesquared_h_
diff --git a/modules/gpu/test/nvidia/TestRectStdDev.cpp b/modules/gpulegacy/test/TestRectStdDev.cpp
similarity index 74%
rename from modules/gpu/test/nvidia/TestRectStdDev.cpp
rename to modules/gpulegacy/test/TestRectStdDev.cpp
index 3a31b0d0c0..86bb9ed23b 100644
--- a/modules/gpu/test/nvidia/TestRectStdDev.cpp
+++ b/modules/gpulegacy/test/TestRectStdDev.cpp
@@ -1,19 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include <math.h>
-
-#include "TestRectStdDev.h"
+#include "test_precomp.hpp"
 
 
 TestRectStdDev::TestRectStdDev(std::string testName_, NCVTestSourceProvider<Ncv8u> &src_,
@@ -180,5 +207,3 @@ bool TestRectStdDev::deinit()
 {
     return true;
 }
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestRectStdDev.h b/modules/gpulegacy/test/TestRectStdDev.h
new file mode 100644
index 0000000000..e22fbe87c7
--- /dev/null
+++ b/modules/gpulegacy/test/TestRectStdDev.h
@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testrectstddev_h_
+#define _testrectstddev_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+class TestRectStdDev : public NCVTestProvider
+{
+public:
+
+    TestRectStdDev(std::string testName, NCVTestSourceProvider<Ncv8u> &src,
+                   Ncv32u width, Ncv32u height, NcvRect32u rect, Ncv32f scaleFactor,
+                   NcvBool bTextureCache);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestRectStdDev(const TestRectStdDev&);
+    TestRectStdDev& operator=(const TestRectStdDev&);
+
+    NCVTestSourceProvider<Ncv8u> &src;
+    Ncv32u width;
+    Ncv32u height;
+    NcvRect32u rect;
+    Ncv32f scaleFactor;
+
+    NcvBool bTextureCache;
+};
+
+#endif // _testrectstddev_h_
diff --git a/modules/gpu/test/nvidia/TestResize.cpp b/modules/gpulegacy/test/TestResize.cpp
similarity index 67%
rename from modules/gpu/test/nvidia/TestResize.cpp
rename to modules/gpulegacy/test/TestResize.cpp
index cb68a18e68..d2080f06de 100644
--- a/modules/gpu/test/nvidia/TestResize.cpp
+++ b/modules/gpulegacy/test/TestResize.cpp
@@ -1,19 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include <math.h>
-
-#include "TestResize.h"
+#include "test_precomp.hpp"
 
 
 template <class T>
@@ -161,5 +188,3 @@ bool TestResize<T>::deinit()
 
 template class TestResize<Ncv32u>;
 template class TestResize<Ncv64u>;
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestResize.h b/modules/gpulegacy/test/TestResize.h
new file mode 100644
index 0000000000..b2b28a83ff
--- /dev/null
+++ b/modules/gpulegacy/test/TestResize.h
@@ -0,0 +1,74 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testresize_h_
+#define _testresize_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+template <class T>
+class TestResize : public NCVTestProvider
+{
+public:
+
+    TestResize(std::string testName, NCVTestSourceProvider<T> &src,
+               Ncv32u width, Ncv32u height, Ncv32u scaleFactor, NcvBool bTextureCache);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+    TestResize(const TestResize&);
+    TestResize& operator=(const TestResize&);
+
+    NCVTestSourceProvider<T> &src;
+    Ncv32u width;
+    Ncv32u height;
+    Ncv32u scaleFactor;
+
+    NcvBool bTextureCache;
+};
+
+#endif // _testresize_h_
diff --git a/modules/gpu/test/nvidia/TestTranspose.cpp b/modules/gpulegacy/test/TestTranspose.cpp
similarity index 64%
rename from modules/gpu/test/nvidia/TestTranspose.cpp
rename to modules/gpulegacy/test/TestTranspose.cpp
index 74e96b1338..3322a0758b 100644
--- a/modules/gpu/test/nvidia/TestTranspose.cpp
+++ b/modules/gpulegacy/test/TestTranspose.cpp
@@ -1,19 +1,46 @@
-/*
- * Copyright 1993-2010 NVIDIA Corporation.  All rights reserved.
- *
- * NVIDIA Corporation and its licensors retain all intellectual
- * property and proprietary rights in and to this software and
- * related documentation and any modifications thereto.
- * Any use, reproduction, disclosure, or distribution of this
- * software and related documentation without an express license
- * agreement from NVIDIA Corporation is strictly prohibited.
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
 
-#if !defined CUDA_DISABLER
-
-#include <math.h>
-
-#include "TestTranspose.h"
+#include "test_precomp.hpp"
 
 
 template <class T>
@@ -148,5 +175,3 @@ bool TestTranspose<T>::deinit()
 
 template class TestTranspose<Ncv32u>;
 template class TestTranspose<Ncv64u>;
-
-#endif /* CUDA_DISABLER */
\ No newline at end of file
diff --git a/modules/gpulegacy/test/TestTranspose.h b/modules/gpulegacy/test/TestTranspose.h
new file mode 100644
index 0000000000..c83306fd1c
--- /dev/null
+++ b/modules/gpulegacy/test/TestTranspose.h
@@ -0,0 +1,73 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef _testtranspose_h_
+#define _testtranspose_h_
+
+#include "NCVTest.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+
+template <class T>
+class TestTranspose : public NCVTestProvider
+{
+public:
+
+    TestTranspose(std::string testName, NCVTestSourceProvider<T> &src,
+                  Ncv32u width, Ncv32u height);
+
+    virtual bool init();
+    virtual bool process();
+    virtual bool deinit();
+    virtual bool toString(std::ofstream &strOut);
+
+private:
+
+    TestTranspose(const TestTranspose&);
+    TestTranspose& operator=(const TestTranspose&);
+
+    NCVTestSourceProvider<T> &src;
+    Ncv32u width;
+    Ncv32u height;
+};
+
+#endif // _testtranspose_h_
diff --git a/modules/gpu/test/nvidia/main_nvidia.cpp b/modules/gpulegacy/test/main_nvidia.cpp
similarity index 85%
rename from modules/gpu/test/nvidia/main_nvidia.cpp
rename to modules/gpulegacy/test/main_nvidia.cpp
index 86839a4421..0c82a1abf4 100644
--- a/modules/gpu/test/nvidia/main_nvidia.cpp
+++ b/modules/gpulegacy/test/main_nvidia.cpp
@@ -1,32 +1,51 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
 #if defined _MSC_VER && _MSC_VER >= 1200
 # pragma warning (disable : 4408 4201 4100)
 #endif
 
-#if !defined CUDA_DISABLER
-
-#include <cstdio>
-
-#include "NCV.hpp"
-#include "NCVHaarObjectDetection.hpp"
-
-#include "TestIntegralImage.h"
-#include "TestIntegralImageSquared.h"
-#include "TestRectStdDev.h"
-#include "TestResize.h"
-#include "TestCompact.h"
-#include "TestTranspose.h"
-
-#include "TestDrawRects.h"
-#include "TestHypothesesGrow.h"
-#include "TestHypothesesFilter.h"
-#include "TestHaarCascadeLoader.h"
-#include "TestHaarCascadeApplication.h"
-
-#include "NCVAutoTestLister.hpp"
-#include "NCVTestSourceProvider.hpp"
-
-#include "main_test_nvidia.h"
-
 static std::string path;
 
 namespace {
@@ -271,7 +290,7 @@ void generateHaarApplicationTests(NCVAutoTestLister &testLister, NCVTestSourcePr
     }
 }
 
-static void devNullOutput(const std::string& msg)
+static void devNullOutput(const cv::String& msg)
 {
     (void)msg;
 }
@@ -330,7 +349,7 @@ bool nvidia_NPPST_Resize(const std::string& test_data_path, OutputLevel outputLe
     NCVAutoTestLister testListerResize("NPPST Resize", outputLevel);
 
     NCVTestSourceProvider<Ncv32u> testSrcRandom_32u(2010, 0, 0xFFFFFFFF, 2048, 2048);
-    NCVTestSourceProvider<Ncv64u> testSrcRandom_64u(2010, 0, -1, 2048, 2048);
+    NCVTestSourceProvider<Ncv64u> testSrcRandom_64u(2010, 0, (Ncv64u) -1, 2048, 2048);
 
     generateResizeTests(testListerResize, testSrcRandom_32u);
     generateResizeTests(testListerResize, testSrcRandom_64u);
@@ -360,7 +379,7 @@ bool nvidia_NPPST_Transpose(const std::string& test_data_path, OutputLevel outpu
     NCVAutoTestLister testListerTranspose("NPPST Transpose", outputLevel);
 
     NCVTestSourceProvider<Ncv32u> testSrcRandom_32u(2010, 0, 0xFFFFFFFF, 2048, 2048);
-    NCVTestSourceProvider<Ncv64u> testSrcRandom_64u(2010, 0, -1, 2048, 2048);
+    NCVTestSourceProvider<Ncv64u> testSrcRandom_64u(2010, 0, (Ncv64u) -1, 2048, 2048);
 
     generateTransposeTests(testListerTranspose, testSrcRandom_32u);
     generateTransposeTests(testListerTranspose, testSrcRandom_64u);
@@ -438,5 +457,3 @@ bool nvidia_NCV_Visualization(const std::string& test_data_path, OutputLevel out
 
     return testListerVisualize.invoke();
 }
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/gpulegacy/test/main_test_nvidia.h b/modules/gpulegacy/test/main_test_nvidia.h
new file mode 100644
index 0000000000..9a3ae1e06a
--- /dev/null
+++ b/modules/gpulegacy/test/main_test_nvidia.h
@@ -0,0 +1,67 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __main_test_nvidia_h__
+#define __main_test_nvidia_h__
+
+enum OutputLevel
+{
+    OutputLevelNone,
+    OutputLevelCompact,
+    OutputLevelFull
+};
+
+extern OutputLevel nvidiaTestOutputLevel;
+
+bool nvidia_NPPST_Integral_Image(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NPPST_Squared_Integral_Image(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NPPST_RectStdDev(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NPPST_Resize(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NPPST_Vector_Operations(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NPPST_Transpose(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NCV_Vector_Operations(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NCV_Haar_Cascade_Loader(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NCV_Haar_Cascade_Application(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NCV_Hypotheses_Filtration(const std::string& test_data_path, OutputLevel outputLevel);
+bool nvidia_NCV_Visualization(const std::string& test_data_path, OutputLevel outputLevel);
+
+#endif
diff --git a/modules/gpu/test/main.cpp b/modules/gpulegacy/test/test_main.cpp
similarity index 65%
rename from modules/gpu/test/main.cpp
rename to modules/gpulegacy/test/test_main.cpp
index 6c7c050b06..cbb66d2963 100644
--- a/modules/gpu/test/main.cpp
+++ b/modules/gpulegacy/test/test_main.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -49,71 +50,6 @@ using namespace cv::gpu;
 using namespace cvtest;
 using namespace testing;
 
-void printOsInfo()
-{
-#if defined _WIN32
-#   if defined _WIN64
-        cout << "OS: Windows x64 \n" << endl;
-#   else
-        cout << "OS: Windows x32 \n" << endl;
-#   endif
-#elif defined linux
-#   if defined _LP64
-        cout << "OS: Linux x64 \n" << endl;
-#   else
-        cout << "OS: Linux x32 \n" << endl;
-#   endif
-#elif defined __APPLE__
-#   if defined _LP64
-        cout << "OS: Apple x64 \n" << endl;
-#   else
-        cout << "OS: Apple x32 \n" << endl;
-#   endif
-#endif
-}
-
-void printCudaInfo()
-{
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
-    cout << "OpenCV was built without CUDA support \n" << endl;
-#else
-    int driver;
-    cudaDriverGetVersion(&driver);
-
-    cout << "CUDA Driver  version: " << driver << '\n';
-    cout << "CUDA Runtime version: " << CUDART_VERSION << '\n';
-
-    cout << endl;
-
-    cout << "GPU module was compiled for the following GPU archs:" << endl;
-    cout << "    BIN: " << CUDA_ARCH_BIN << '\n';
-    cout << "    PTX: " << CUDA_ARCH_PTX << '\n';
-
-    cout << endl;
-
-    int deviceCount = getCudaEnabledDeviceCount();
-    cout << "CUDA device count: " << deviceCount << '\n';
-
-    cout << endl;
-
-    for (int i = 0; i < deviceCount; ++i)
-    {
-        DeviceInfo info(i);
-
-        cout << "Device [" << i << "] \n";
-        cout << "\t Name: " << info.name() << '\n';
-        cout << "\t Compute capability: " << info.majorVersion() << '.' << info.minorVersion()<< '\n';
-        cout << "\t Multi Processor Count: " << info.multiProcessorCount() << '\n';
-        cout << "\t Total memory: " << static_cast<int>(static_cast<int>(info.totalMemory() / 1024.0) / 1024.0) << " Mb \n";
-        cout << "\t Free  memory: " << static_cast<int>(static_cast<int>(info.freeMemory() / 1024.0) / 1024.0) << " Mb \n";
-        if (!info.isCompatible())
-            cout << "\t !!! This device is NOT compatible with current GPU module build \n";
-
-        cout << endl;
-    }
-#endif
-}
-
 int main(int argc, char** argv)
 {
     try
@@ -133,7 +69,6 @@ int main(int argc, char** argv)
             return 0;
         }
 
-        printOsInfo();
         printCudaInfo();
 
         if (cmd.has("info"))
diff --git a/modules/gpu/test/test_nvidia.cpp b/modules/gpulegacy/test/test_nvidia.cpp
similarity index 91%
rename from modules/gpu/test/test_nvidia.cpp
rename to modules/gpulegacy/test/test_nvidia.cpp
index 53ed8c2235..1cda187bce 100644
--- a/modules/gpu/test/test_nvidia.cpp
+++ b/modules/gpulegacy/test/test_nvidia.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -139,13 +140,12 @@ GPU_TEST_P(NCV, HypothesesFiltration)
 
 GPU_TEST_P(NCV, Visualization)
 {
-    // this functionality doesn't used in gpu module
     bool res = nvidia_NCV_Visualization(_path, nvidiaTestOutputLevel);
 
     ASSERT_TRUE(res);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_NVidia, NPPST, ALL_DEVICES);
-INSTANTIATE_TEST_CASE_P(GPU_NVidia, NCV, ALL_DEVICES);
+INSTANTIATE_TEST_CASE_P(GPU_Legacy, NPPST, ALL_DEVICES);
+INSTANTIATE_TEST_CASE_P(GPU_Legacy, NCV, ALL_DEVICES);
 
 #endif // HAVE_CUDA
diff --git a/modules/ocl/perf/precomp.hpp b/modules/gpulegacy/test/test_precomp.hpp
similarity index 69%
rename from modules/ocl/perf/precomp.hpp
rename to modules/gpulegacy/test/test_precomp.hpp
index ac7bd8be5d..e1f2d5def4 100644
--- a/modules/ocl/perf/precomp.hpp
+++ b/modules/gpulegacy/test/test_precomp.hpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -50,32 +51,47 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
-#include <cmath>
-#include <cstdio>
-#include <iostream>
-#include <fstream>
-#include <sstream>
-#include <string>
-#include <limits>
-#include <algorithm>
-#include <iterator>
-#include <string>
-#include <cstdarg>
-#include "cvconfig.h"
-#include "opencv2/core.hpp"
-#include "opencv2/highgui.hpp"
-//#include "opencv2/calib3d.hpp"
-#include "opencv2/imgproc.hpp"
-#include "opencv2/video.hpp"
-#include "opencv2/ts.hpp"
-#include "opencv2/ts/ts_perf.hpp"
-#include "opencv2/ocl.hpp"
-//#include "opencv2/nonfree.hpp"
-
-#include "utility.hpp"
-#include "interpolation.hpp"
-//#include "add_test_info.h"
-//#define  PERF_TEST_OCL 1
-
+#if defined(__GNUC__) && !defined(__APPLE__) && !defined(__arm__)
+    #include <fpu_control.h>
 #endif
 
+#include <cfloat>
+#include <cstdio>
+#include <cmath>
+#include <vector>
+#include <string>
+#include <map>
+#include <memory>
+#include <algorithm>
+#include <fstream>
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/gpulegacy.hpp"
+#include "opencv2/highgui.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#include "cvconfig.h"
+
+#include "NCVTest.hpp"
+#include "NCVAutoTestLister.hpp"
+#include "NCVTestSourceProvider.hpp"
+
+#include "TestIntegralImage.h"
+#include "TestIntegralImageSquared.h"
+#include "TestRectStdDev.h"
+#include "TestResize.h"
+#include "TestCompact.h"
+#include "TestTranspose.h"
+#include "TestDrawRects.h"
+#include "TestHypothesesGrow.h"
+#include "TestHypothesesFilter.h"
+#include "TestHaarCascadeLoader.h"
+#include "TestHaarCascadeApplication.h"
+
+#include "main_test_nvidia.h"
+
+#endif
diff --git a/modules/gpuoptflow/CMakeLists.txt b/modules/gpuoptflow/CMakeLists.txt
new file mode 100644
index 0000000000..283891bb0f
--- /dev/null
+++ b/modules/gpuoptflow/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpuoptflow)
+endif()
+
+set(the_description "GPU-accelerated Optical Flow")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_define_module(gpuoptflow opencv_video opencv_legacy opencv_gpuarithm opencv_gpuwarping opencv_gpuimgproc OPTIONAL opencv_gpulegacy)
diff --git a/modules/gpuoptflow/doc/gpuoptflow.rst b/modules/gpuoptflow/doc/gpuoptflow.rst
new file mode 100644
index 0000000000..925f393388
--- /dev/null
+++ b/modules/gpuoptflow/doc/gpuoptflow.rst
@@ -0,0 +1,8 @@
+****************************************
+gpuoptflow. GPU-accelerated Optical Flow
+****************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    optflow
diff --git a/modules/gpuoptflow/doc/optflow.rst b/modules/gpuoptflow/doc/optflow.rst
new file mode 100644
index 0000000000..937b209ae1
--- /dev/null
+++ b/modules/gpuoptflow/doc/optflow.rst
@@ -0,0 +1,229 @@
+Optical Flow
+============
+
+.. highlight:: cpp
+
+.. note::
+
+   * A general optical flow example can be found at opencv_source_code/samples/gpu/optical_flow.cpp
+   * A general optical flow example using the Nvidia API can be found at opencv_source_code/samples/gpu/opticalflow_nvidia_api.cpp
+
+
+gpu::BroxOpticalFlow
+--------------------
+.. ocv:class:: gpu::BroxOpticalFlow
+
+Class computing the optical flow for two images using Brox et al Optical Flow algorithm ([Brox2004]_). ::
+
+    class BroxOpticalFlow
+    {
+    public:
+        BroxOpticalFlow(float alpha_, float gamma_, float scale_factor_, int inner_iterations_, int outer_iterations_, int solver_iterations_);
+
+        //! Compute optical flow
+        //! frame0 - source frame (supports only CV_32FC1 type)
+        //! frame1 - frame to track (with the same size and type as frame0)
+        //! u      - flow horizontal component (along x axis)
+        //! v      - flow vertical component (along y axis)
+        void operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& stream = Stream::Null());
+
+        //! flow smoothness
+        float alpha;
+
+        //! gradient constancy importance
+        float gamma;
+
+        //! pyramid scale factor
+        float scale_factor;
+
+        //! number of lagged non-linearity iterations (inner loop)
+        int inner_iterations;
+
+        //! number of warping iterations (number of pyramid levels)
+        int outer_iterations;
+
+        //! number of linear system solver iterations
+        int solver_iterations;
+
+        GpuMat buf;
+    };
+
+.. note::
+
+   * An example illustrating the Brox et al optical flow algorithm can be found at opencv_source_code/samples/gpu/brox_optical_flow.cpp
+
+
+gpu::FarnebackOpticalFlow
+-------------------------
+.. ocv:class:: gpu::FarnebackOpticalFlow
+
+Class computing a dense optical flow using the Gunnar Farneback’s algorithm. ::
+
+    class CV_EXPORTS FarnebackOpticalFlow
+    {
+    public:
+        FarnebackOpticalFlow()
+        {
+            numLevels = 5;
+            pyrScale = 0.5;
+            fastPyramids = false;
+            winSize = 13;
+            numIters = 10;
+            polyN = 5;
+            polySigma = 1.1;
+            flags = 0;
+        }
+
+        int numLevels;
+        double pyrScale;
+        bool fastPyramids;
+        int winSize;
+        int numIters;
+        int polyN;
+        double polySigma;
+        int flags;
+
+        void operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null());
+
+        void releaseMemory();
+
+    private:
+        /* hidden */
+    };
+
+
+
+gpu::FarnebackOpticalFlow::operator ()
+--------------------------------------
+Computes a dense optical flow using the Gunnar Farneback’s algorithm.
+
+.. ocv:function:: void gpu::FarnebackOpticalFlow::operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null())
+
+    :param frame0: First 8-bit gray-scale input image
+    :param frame1: Second 8-bit gray-scale input image
+    :param flowx: Flow horizontal component
+    :param flowy: Flow vertical component
+    :param s: Stream
+
+.. seealso:: :ocv:func:`calcOpticalFlowFarneback`
+
+
+
+gpu::FarnebackOpticalFlow::releaseMemory
+----------------------------------------
+Releases unused auxiliary memory buffers.
+
+.. ocv:function:: void gpu::FarnebackOpticalFlow::releaseMemory()
+
+
+
+gpu::PyrLKOpticalFlow
+---------------------
+.. ocv:class:: gpu::PyrLKOpticalFlow
+
+Class used for calculating an optical flow. ::
+
+    class PyrLKOpticalFlow
+    {
+    public:
+        PyrLKOpticalFlow();
+
+        void sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts,
+            GpuMat& status, GpuMat* err = 0);
+
+        void dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0);
+
+        Size winSize;
+        int maxLevel;
+        int iters;
+        bool useInitialFlow;
+
+        void releaseMemory();
+    };
+
+The class can calculate an optical flow for a sparse feature set or dense optical flow using the iterative Lucas-Kanade method with pyramids.
+
+.. seealso:: :ocv:func:`calcOpticalFlowPyrLK`
+
+.. note::
+
+   * An example of the Lucas Kanade optical flow algorithm can be found at opencv_source_code/samples/gpu/pyrlk_optical_flow.cpp
+
+
+gpu::PyrLKOpticalFlow::sparse
+-----------------------------
+Calculate an optical flow for a sparse feature set.
+
+.. ocv:function:: void gpu::PyrLKOpticalFlow::sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts, GpuMat& status, GpuMat* err = 0)
+
+    :param prevImg: First 8-bit input image (supports both grayscale and color images).
+
+    :param nextImg: Second input image of the same size and the same type as  ``prevImg`` .
+
+    :param prevPts: Vector of 2D points for which the flow needs to be found. It must be one row matrix with CV_32FC2 type.
+
+    :param nextPts: Output vector of 2D points (with single-precision floating-point coordinates) containing the calculated new positions of input features in the second image. When ``useInitialFlow`` is true, the vector must have the same size as in the input.
+
+    :param status: Output status vector (CV_8UC1 type). Each element of the vector is set to 1 if the flow for the corresponding features has been found. Otherwise, it is set to 0.
+
+    :param err: Output vector (CV_32FC1 type) that contains the difference between patches around the original and moved points or min eigen value if ``getMinEigenVals`` is checked. It can be NULL, if not needed.
+
+.. seealso:: :ocv:func:`calcOpticalFlowPyrLK`
+
+
+
+gpu::PyrLKOpticalFlow::dense
+-----------------------------
+Calculate dense optical flow.
+
+.. ocv:function:: void gpu::PyrLKOpticalFlow::dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0)
+
+    :param prevImg: First 8-bit grayscale input image.
+
+    :param nextImg: Second input image of the same size and the same type as  ``prevImg`` .
+
+    :param u: Horizontal component of the optical flow of the same size as input images, 32-bit floating-point, single-channel
+
+    :param v: Vertical component of the optical flow of the same size as input images, 32-bit floating-point, single-channel
+
+    :param err: Output vector (CV_32FC1 type) that contains the difference between patches around the original and moved points or min eigen value if ``getMinEigenVals`` is checked. It can be NULL, if not needed.
+
+
+
+gpu::PyrLKOpticalFlow::releaseMemory
+------------------------------------
+Releases inner buffers memory.
+
+.. ocv:function:: void gpu::PyrLKOpticalFlow::releaseMemory()
+
+
+
+gpu::interpolateFrames
+----------------------
+Interpolates frames (images) using provided optical flow (displacement field).
+
+.. ocv:function:: void gpu::interpolateFrames(const GpuMat& frame0, const GpuMat& frame1, const GpuMat& fu, const GpuMat& fv, const GpuMat& bu, const GpuMat& bv, float pos, GpuMat& newFrame, GpuMat& buf, Stream& stream = Stream::Null())
+
+    :param frame0: First frame (32-bit floating point images, single channel).
+
+    :param frame1: Second frame. Must have the same type and size as ``frame0`` .
+
+    :param fu: Forward horizontal displacement.
+
+    :param fv: Forward vertical displacement.
+
+    :param bu: Backward horizontal displacement.
+
+    :param bv: Backward vertical displacement.
+
+    :param pos: New frame position.
+
+    :param newFrame: Output image.
+
+    :param buf: Temporary buffer, will have width x 6*height size, CV_32FC1 type and contain 6 GpuMat: occlusion masks for first frame, occlusion masks for second, interpolated forward horizontal flow, interpolated forward vertical flow, interpolated backward horizontal flow, interpolated backward vertical flow.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+.. [Brox2004] T. Brox, A. Bruhn, N. Papenberg, J. Weickert. *High accuracy optical flow estimation based on a theory for warping*. ECCV 2004.
diff --git a/modules/gpuoptflow/include/opencv2/gpuoptflow.hpp b/modules/gpuoptflow/include/opencv2/gpuoptflow.hpp
new file mode 100644
index 0000000000..e9e0719879
--- /dev/null
+++ b/modules/gpuoptflow/include/opencv2/gpuoptflow.hpp
@@ -0,0 +1,311 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUOPTFLOW_HPP__
+#define __OPENCV_GPUOPTFLOW_HPP__
+
+#ifndef __cplusplus
+#  error gpuoptflow.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpu {
+
+class CV_EXPORTS BroxOpticalFlow
+{
+public:
+    BroxOpticalFlow(float alpha_, float gamma_, float scale_factor_, int inner_iterations_, int outer_iterations_, int solver_iterations_) :
+        alpha(alpha_), gamma(gamma_), scale_factor(scale_factor_),
+        inner_iterations(inner_iterations_), outer_iterations(outer_iterations_), solver_iterations(solver_iterations_)
+    {
+    }
+
+    //! Compute optical flow
+    //! frame0 - source frame (supports only CV_32FC1 type)
+    //! frame1 - frame to track (with the same size and type as frame0)
+    //! u      - flow horizontal component (along x axis)
+    //! v      - flow vertical component (along y axis)
+    void operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& stream = Stream::Null());
+
+    //! flow smoothness
+    float alpha;
+
+    //! gradient constancy importance
+    float gamma;
+
+    //! pyramid scale factor
+    float scale_factor;
+
+    //! number of lagged non-linearity iterations (inner loop)
+    int inner_iterations;
+
+    //! number of warping iterations (number of pyramid levels)
+    int outer_iterations;
+
+    //! number of linear system solver iterations
+    int solver_iterations;
+
+    GpuMat buf;
+};
+
+class CV_EXPORTS PyrLKOpticalFlow
+{
+public:
+    PyrLKOpticalFlow();
+
+    void sparse(const GpuMat& prevImg, const GpuMat& nextImg, const GpuMat& prevPts, GpuMat& nextPts,
+        GpuMat& status, GpuMat* err = 0);
+
+    void dense(const GpuMat& prevImg, const GpuMat& nextImg, GpuMat& u, GpuMat& v, GpuMat* err = 0);
+
+    void releaseMemory();
+
+    Size winSize;
+    int maxLevel;
+    int iters;
+    bool useInitialFlow;
+
+private:
+    std::vector<GpuMat> prevPyr_;
+    std::vector<GpuMat> nextPyr_;
+
+    GpuMat buf_;
+
+    GpuMat uPyr_[2];
+    GpuMat vPyr_[2];
+};
+
+class CV_EXPORTS FarnebackOpticalFlow
+{
+public:
+    FarnebackOpticalFlow()
+    {
+        numLevels = 5;
+        pyrScale = 0.5;
+        fastPyramids = false;
+        winSize = 13;
+        numIters = 10;
+        polyN = 5;
+        polySigma = 1.1;
+        flags = 0;
+    }
+
+    int numLevels;
+    double pyrScale;
+    bool fastPyramids;
+    int winSize;
+    int numIters;
+    int polyN;
+    double polySigma;
+    int flags;
+
+    void operator ()(const GpuMat &frame0, const GpuMat &frame1, GpuMat &flowx, GpuMat &flowy, Stream &s = Stream::Null());
+
+    void releaseMemory()
+    {
+        frames_[0].release();
+        frames_[1].release();
+        pyrLevel_[0].release();
+        pyrLevel_[1].release();
+        M_.release();
+        bufM_.release();
+        R_[0].release();
+        R_[1].release();
+        blurredFrame_[0].release();
+        blurredFrame_[1].release();
+        pyramid0_.clear();
+        pyramid1_.clear();
+    }
+
+private:
+    void prepareGaussian(
+            int n, double sigma, float *g, float *xg, float *xxg,
+            double &ig11, double &ig03, double &ig33, double &ig55);
+
+    void setPolynomialExpansionConsts(int n, double sigma);
+
+    void updateFlow_boxFilter(
+            const GpuMat& R0, const GpuMat& R1, GpuMat& flowx, GpuMat &flowy,
+            GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[]);
+
+    void updateFlow_gaussianBlur(
+            const GpuMat& R0, const GpuMat& R1, GpuMat& flowx, GpuMat& flowy,
+            GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[]);
+
+    GpuMat frames_[2];
+    GpuMat pyrLevel_[2], M_, bufM_, R_[2], blurredFrame_[2];
+    std::vector<GpuMat> pyramid0_, pyramid1_;
+};
+
+// Implementation of the Zach, Pock and Bischof Dual TV-L1 Optical Flow method
+//
+// see reference:
+//   [1] C. Zach, T. Pock and H. Bischof, "A Duality Based Approach for Realtime TV-L1 Optical Flow".
+//   [2] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
+class CV_EXPORTS OpticalFlowDual_TVL1_GPU
+{
+public:
+    OpticalFlowDual_TVL1_GPU();
+
+    void operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy);
+
+    void collectGarbage();
+
+    /**
+     * Time step of the numerical scheme.
+     */
+    double tau;
+
+    /**
+     * Weight parameter for the data term, attachment parameter.
+     * This is the most relevant parameter, which determines the smoothness of the output.
+     * The smaller this parameter is, the smoother the solutions we obtain.
+     * It depends on the range of motions of the images, so its value should be adapted to each image sequence.
+     */
+    double lambda;
+
+    /**
+     * Weight parameter for (u - v)^2, tightness parameter.
+     * It serves as a link between the attachment and the regularization terms.
+     * In theory, it should have a small value in order to maintain both parts in correspondence.
+     * The method is stable for a large range of values of this parameter.
+     */
+    double theta;
+
+    /**
+     * Number of scales used to create the pyramid of images.
+     */
+    int nscales;
+
+    /**
+     * Number of warpings per scale.
+     * Represents the number of times that I1(x+u0) and grad( I1(x+u0) ) are computed per scale.
+     * This is a parameter that assures the stability of the method.
+     * It also affects the running time, so it is a compromise between speed and accuracy.
+     */
+    int warps;
+
+    /**
+     * Stopping criterion threshold used in the numerical scheme, which is a trade-off between precision and running time.
+     * A small value will yield more accurate solutions at the expense of a slower convergence.
+     */
+    double epsilon;
+
+    /**
+     * Stopping criterion iterations number used in the numerical scheme.
+     */
+    int iterations;
+
+    double scaleStep;
+
+    bool useInitialFlow;
+
+private:
+    void procOneScale(const GpuMat& I0, const GpuMat& I1, GpuMat& u1, GpuMat& u2);
+
+    std::vector<GpuMat> I0s;
+    std::vector<GpuMat> I1s;
+    std::vector<GpuMat> u1s;
+    std::vector<GpuMat> u2s;
+
+    GpuMat I1x_buf;
+    GpuMat I1y_buf;
+
+    GpuMat I1w_buf;
+    GpuMat I1wx_buf;
+    GpuMat I1wy_buf;
+
+    GpuMat grad_buf;
+    GpuMat rho_c_buf;
+
+    GpuMat p11_buf;
+    GpuMat p12_buf;
+    GpuMat p21_buf;
+    GpuMat p22_buf;
+
+    GpuMat diff_buf;
+    GpuMat norm_buf;
+};
+
+//! Calculates optical flow for 2 images using block matching algorithm */
+CV_EXPORTS void calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr,
+                                  Size block_size, Size shift_size, Size max_range, bool use_previous,
+                                  GpuMat& velx, GpuMat& vely, GpuMat& buf,
+                                  Stream& stream = Stream::Null());
+
+class CV_EXPORTS FastOpticalFlowBM
+{
+public:
+    void operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy, int search_window = 21, int block_window = 7, Stream& s = Stream::Null());
+
+private:
+    GpuMat buffer;
+    GpuMat extended_I0;
+    GpuMat extended_I1;
+};
+
+//! Interpolate frames (images) using provided optical flow (displacement field).
+//! frame0   - frame 0 (32-bit floating point images, single channel)
+//! frame1   - frame 1 (the same type and size)
+//! fu       - forward horizontal displacement
+//! fv       - forward vertical displacement
+//! bu       - backward horizontal displacement
+//! bv       - backward vertical displacement
+//! pos      - new frame position
+//! newFrame - new frame
+//! buf      - temporary buffer, will have width x 6*height size, CV_32FC1 type and contain 6 GpuMat;
+//!            occlusion masks            0, occlusion masks            1,
+//!            interpolated forward flow  0, interpolated forward flow  1,
+//!            interpolated backward flow 0, interpolated backward flow 1
+//!
+CV_EXPORTS void interpolateFrames(const GpuMat& frame0, const GpuMat& frame1,
+                                  const GpuMat& fu, const GpuMat& fv,
+                                  const GpuMat& bu, const GpuMat& bv,
+                                  float pos, GpuMat& newFrame, GpuMat& buf,
+                                  Stream& stream = Stream::Null());
+
+CV_EXPORTS void createOpticalFlowNeedleMap(const GpuMat& u, const GpuMat& v, GpuMat& vertex, GpuMat& colors);
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUOPTFLOW_HPP__ */
diff --git a/modules/gpuoptflow/perf/perf_main.cpp b/modules/gpuoptflow/perf/perf_main.cpp
new file mode 100644
index 0000000000..e261932f8e
--- /dev/null
+++ b/modules/gpuoptflow/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpuoptflow)
diff --git a/modules/gpuoptflow/perf/perf_optflow.cpp b/modules/gpuoptflow/perf/perf_optflow.cpp
new file mode 100644
index 0000000000..562eb4b208
--- /dev/null
+++ b/modules/gpuoptflow/perf/perf_optflow.cpp
@@ -0,0 +1,479 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+#include "opencv2/legacy.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////
+// InterpolateFrames
+
+typedef pair<string, string> pair_string;
+
+DEF_PARAM_TEST_1(ImagePair, pair_string);
+
+PERF_TEST_P(ImagePair, InterpolateFrames,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
+    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat d_fu, d_fv;
+        cv::gpu::GpuMat d_bu, d_bv;
+
+        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
+                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
+
+        d_flow(d_frame0, d_frame1, d_fu, d_fv);
+        d_flow(d_frame1, d_frame0, d_bu, d_bv);
+
+        cv::gpu::GpuMat newFrame;
+        cv::gpu::GpuMat d_buf;
+
+        TEST_CYCLE() cv::gpu::interpolateFrames(d_frame0, d_frame1, d_fu, d_fv, d_bu, d_bv, 0.5f, newFrame, d_buf);
+
+        GPU_SANITY_CHECK(newFrame, 1e-4);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////
+// CreateOpticalFlowNeedleMap
+
+PERF_TEST_P(ImagePair, CreateOpticalFlowNeedleMap,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
+    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u;
+        cv::gpu::GpuMat v;
+
+        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
+                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
+
+        d_flow(d_frame0, d_frame1, u, v);
+
+        cv::gpu::GpuMat vertex, colors;
+
+        TEST_CYCLE() cv::gpu::createOpticalFlowNeedleMap(u, v, vertex, colors);
+
+        GPU_SANITY_CHECK(vertex, 1e-6);
+        GPU_SANITY_CHECK(colors);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////
+// BroxOpticalFlow
+
+PERF_TEST_P(ImagePair, BroxOpticalFlow,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    declare.time(300);
+
+    cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    frame0.convertTo(frame0, CV_32FC1, 1.0 / 255.0);
+    frame1.convertTo(frame1, CV_32FC1, 1.0 / 255.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u;
+        cv::gpu::GpuMat v;
+
+        cv::gpu::BroxOpticalFlow d_flow(0.197f /*alpha*/, 50.0f /*gamma*/, 0.8f /*scale_factor*/,
+                                        10 /*inner_iterations*/, 77 /*outer_iterations*/, 10 /*solver_iterations*/);
+
+        TEST_CYCLE() d_flow(d_frame0, d_frame1, u, v);
+
+        GPU_SANITY_CHECK(u, 1e-1);
+        GPU_SANITY_CHECK(v, 1e-1);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////
+// PyrLKOpticalFlowSparse
+
+DEF_PARAM_TEST(ImagePair_Gray_NPts_WinSz_Levels_Iters, pair_string, bool, int, int, int, int);
+
+PERF_TEST_P(ImagePair_Gray_NPts_WinSz_Levels_Iters, PyrLKOpticalFlowSparse,
+            Combine(Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")),
+                    Bool(),
+                    Values(8000),
+                    Values(21),
+                    Values(1, 3),
+                    Values(1, 30)))
+{
+    declare.time(20.0);
+
+    const pair_string imagePair = GET_PARAM(0);
+    const bool useGray = GET_PARAM(1);
+    const int points = GET_PARAM(2);
+    const int winSize = GET_PARAM(3);
+    const int levels = GET_PARAM(4);
+    const int iters = GET_PARAM(5);
+
+    const cv::Mat frame0 = readImage(imagePair.first, useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(imagePair.second, useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
+    ASSERT_FALSE(frame1.empty());
+
+    cv::Mat gray_frame;
+    if (useGray)
+        gray_frame = frame0;
+    else
+        cv::cvtColor(frame0, gray_frame, cv::COLOR_BGR2GRAY);
+
+    cv::Mat pts;
+    cv::goodFeaturesToTrack(gray_frame, pts, points, 0.01, 0.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_pts(pts.reshape(2, 1));
+
+        cv::gpu::PyrLKOpticalFlow d_pyrLK;
+        d_pyrLK.winSize = cv::Size(winSize, winSize);
+        d_pyrLK.maxLevel = levels - 1;
+        d_pyrLK.iters = iters;
+
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat nextPts;
+        cv::gpu::GpuMat status;
+
+        TEST_CYCLE() d_pyrLK.sparse(d_frame0, d_frame1, d_pts, nextPts, status);
+
+        GPU_SANITY_CHECK(nextPts);
+        GPU_SANITY_CHECK(status);
+    }
+    else
+    {
+        cv::Mat nextPts;
+        cv::Mat status;
+
+        TEST_CYCLE()
+        {
+            cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, cv::noArray(),
+                                     cv::Size(winSize, winSize), levels - 1,
+                                     cv::TermCriteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, iters, 0.01));
+        }
+
+        CPU_SANITY_CHECK(nextPts);
+        CPU_SANITY_CHECK(status);
+    }
+}
+
+//////////////////////////////////////////////////////
+// PyrLKOpticalFlowDense
+
+DEF_PARAM_TEST(ImagePair_WinSz_Levels_Iters, pair_string, int, int, int);
+
+PERF_TEST_P(ImagePair_WinSz_Levels_Iters, PyrLKOpticalFlowDense,
+            Combine(Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")),
+                    Values(3, 5, 7, 9, 13, 17, 21),
+                    Values(1, 3),
+                    Values(1, 10)))
+{
+    declare.time(30);
+
+    const pair_string imagePair = GET_PARAM(0);
+    const int winSize = GET_PARAM(1);
+    const int levels = GET_PARAM(2);
+    const int iters = GET_PARAM(3);
+
+    const cv::Mat frame0 = readImage(imagePair.first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(imagePair.second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u;
+        cv::gpu::GpuMat v;
+
+        cv::gpu::PyrLKOpticalFlow d_pyrLK;
+        d_pyrLK.winSize = cv::Size(winSize, winSize);
+        d_pyrLK.maxLevel = levels - 1;
+        d_pyrLK.iters = iters;
+
+        TEST_CYCLE() d_pyrLK.dense(d_frame0, d_frame1, u, v);
+
+        GPU_SANITY_CHECK(u);
+        GPU_SANITY_CHECK(v);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////
+// FarnebackOpticalFlow
+
+PERF_TEST_P(ImagePair, FarnebackOpticalFlow,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    declare.time(10);
+
+    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    const int numLevels = 5;
+    const double pyrScale = 0.5;
+    const int winSize = 13;
+    const int numIters = 10;
+    const int polyN = 5;
+    const double polySigma = 1.1;
+    const int flags = 0;
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u;
+        cv::gpu::GpuMat v;
+
+        cv::gpu::FarnebackOpticalFlow d_farneback;
+        d_farneback.numLevels = numLevels;
+        d_farneback.pyrScale = pyrScale;
+        d_farneback.winSize = winSize;
+        d_farneback.numIters = numIters;
+        d_farneback.polyN = polyN;
+        d_farneback.polySigma = polySigma;
+        d_farneback.flags = flags;
+
+        TEST_CYCLE() d_farneback(d_frame0, d_frame1, u, v);
+
+        GPU_SANITY_CHECK(u, 1e-4);
+        GPU_SANITY_CHECK(v, 1e-4);
+    }
+    else
+    {
+        cv::Mat flow;
+
+        TEST_CYCLE() cv::calcOpticalFlowFarneback(frame0, frame1, flow, pyrScale, numLevels, winSize, numIters, polyN, polySigma, flags);
+
+        CPU_SANITY_CHECK(flow);
+    }
+}
+
+//////////////////////////////////////////////////////
+// OpticalFlowDual_TVL1
+
+PERF_TEST_P(ImagePair, OpticalFlowDual_TVL1,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    declare.time(20);
+
+    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u;
+        cv::gpu::GpuMat v;
+
+        cv::gpu::OpticalFlowDual_TVL1_GPU d_alg;
+
+        TEST_CYCLE() d_alg(d_frame0, d_frame1, u, v);
+
+        GPU_SANITY_CHECK(u, 1e-1);
+        GPU_SANITY_CHECK(v, 1e-1);
+    }
+    else
+    {
+        cv::Mat flow;
+
+        cv::Ptr<cv::DenseOpticalFlow> alg = cv::createOptFlow_DualTVL1();
+        alg->set("medianFiltering", 1);
+        alg->set("innerIterations", 1);
+        alg->set("outerIterations", 300);
+
+        TEST_CYCLE() alg->calc(frame0, frame1, flow);
+
+        CPU_SANITY_CHECK(flow);
+    }
+}
+
+//////////////////////////////////////////////////////
+// OpticalFlowBM
+
+void calcOpticalFlowBM(const cv::Mat& prev, const cv::Mat& curr,
+                       cv::Size bSize, cv::Size shiftSize, cv::Size maxRange, int usePrevious,
+                       cv::Mat& velx, cv::Mat& vely)
+{
+    cv::Size sz((curr.cols - bSize.width + shiftSize.width)/shiftSize.width, (curr.rows - bSize.height + shiftSize.height)/shiftSize.height);
+
+    velx.create(sz, CV_32FC1);
+    vely.create(sz, CV_32FC1);
+
+    CvMat cvprev = prev;
+    CvMat cvcurr = curr;
+
+    CvMat cvvelx = velx;
+    CvMat cvvely = vely;
+
+    cvCalcOpticalFlowBM(&cvprev, &cvcurr, bSize, shiftSize, maxRange, usePrevious, &cvvelx, &cvvely);
+}
+
+PERF_TEST_P(ImagePair, OpticalFlowBM,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    declare.time(400);
+
+    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    const cv::Size block_size(16, 16);
+    const cv::Size shift_size(1, 1);
+    const cv::Size max_range(16, 16);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u, v, buf;
+
+        TEST_CYCLE() cv::gpu::calcOpticalFlowBM(d_frame0, d_frame1, block_size, shift_size, max_range, false, u, v, buf);
+
+        GPU_SANITY_CHECK(u);
+        GPU_SANITY_CHECK(v);
+    }
+    else
+    {
+        cv::Mat u, v;
+
+        TEST_CYCLE() calcOpticalFlowBM(frame0, frame1, block_size, shift_size, max_range, false, u, v);
+
+        CPU_SANITY_CHECK(u);
+        CPU_SANITY_CHECK(v);
+    }
+}
+
+PERF_TEST_P(ImagePair, FastOpticalFlowBM,
+            Values<pair_string>(make_pair("gpu/opticalflow/frame0.png", "gpu/opticalflow/frame1.png")))
+{
+    declare.time(400);
+
+    const cv::Mat frame0 = readImage(GetParam().first, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    const cv::Mat frame1 = readImage(GetParam().second, cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    const cv::Size block_size(16, 16);
+    const cv::Size shift_size(1, 1);
+    const cv::Size max_range(16, 16);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_frame0(frame0);
+        const cv::gpu::GpuMat d_frame1(frame1);
+        cv::gpu::GpuMat u, v;
+
+        cv::gpu::FastOpticalFlowBM fastBM;
+
+        TEST_CYCLE() fastBM(d_frame0, d_frame1, u, v, max_range.width, block_size.width);
+
+        GPU_SANITY_CHECK(u, 2);
+        GPU_SANITY_CHECK(v, 2);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuoptflow/perf/perf_precomp.hpp b/modules/gpuoptflow/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..9235b62dcd
--- /dev/null
+++ b/modules/gpuoptflow/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpuoptflow.hpp"
+#include "opencv2/video.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpu/src/optflowbm.cpp b/modules/gpuoptflow/src/bm.cpp
similarity index 73%
rename from modules/gpu/src/optflowbm.cpp
rename to modules/gpuoptflow/src/bm.cpp
index ea143b6547..b8daa96b09 100644
--- a/modules/gpu/src/optflowbm.cpp
+++ b/modules/gpuoptflow/src/bm.cpp
@@ -47,9 +47,7 @@ using namespace cv::gpu;
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::calcOpticalFlowBM(const GpuMat&, const GpuMat&, Size, Size, Size, bool, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-
-void cv::gpu::FastOpticalFlowBM::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, int, int, Stream&) { throw_nogpu(); }
+void cv::gpu::calcOpticalFlowBM(const GpuMat&, const GpuMat&, Size, Size, Size, bool, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
@@ -94,29 +92,29 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
         // upper side
         for (int j = -i; j <= i + 1; ++j, ++ssCount)
         {
-            ss[ssCount].x = ++x;
-            ss[ssCount].y = y;
+            ss[ssCount].x = (short) ++x;
+            ss[ssCount].y = (short) y;
         }
 
         // right side
         for (int j = -i; j <= i + 1; ++j, ++ssCount)
         {
-            ss[ssCount].x = x;
-            ss[ssCount].y = ++y;
+            ss[ssCount].x = (short) x;
+            ss[ssCount].y = (short) ++y;
         }
 
         // bottom side
         for (int j = -i; j <= i + 1; ++j, ++ssCount)
         {
-            ss[ssCount].x = --x;
-            ss[ssCount].y = y;
+            ss[ssCount].x = (short) --x;
+            ss[ssCount].y = (short) y;
         }
 
         // left side
         for (int j = -i; j <= i + 1; ++j, ++ssCount)
         {
-            ss[ssCount].x = x;
-            ss[ssCount].y = --y;
+            ss[ssCount].x = (short) x;
+            ss[ssCount].y = (short) --y;
         }
     }
 
@@ -135,8 +133,8 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
             // upper side
             for (int j = -maxRange.width; j <= maxRange.width; ++j, ++ssCount, ++x)
             {
-                ss[ssCount].x = x;
-                ss[ssCount].y = y;
+                ss[ssCount].x = (short) x;
+                ss[ssCount].y = (short) y;
             }
 
             x = xleft;
@@ -145,8 +143,8 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
             // bottom side
             for (int j = -maxRange.width; j <= maxRange.width; ++j, ++ssCount, ++x)
             {
-                ss[ssCount].x = x;
-                ss[ssCount].y = y;
+                ss[ssCount].x = (short) x;
+                ss[ssCount].y = (short) y;
             }
         }
     }
@@ -164,8 +162,8 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
             // left side
             for (int j = -maxRange.height; j <= maxRange.height; ++j, ++ssCount, ++y)
             {
-                ss[ssCount].x = x;
-                ss[ssCount].y = y;
+                ss[ssCount].x = (short) x;
+                ss[ssCount].y = (short) y;
             }
 
             y = yupper;
@@ -174,8 +172,8 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
             // right side
             for (int j = -maxRange.height; j <= maxRange.height; ++j, ++ssCount, ++y)
             {
-                ss[ssCount].x = x;
-                ss[ssCount].y = y;
+                ss[ssCount].x = (short) x;
+                ss[ssCount].y = (short) y;
             }
         }
     }
@@ -203,40 +201,4 @@ void cv::gpu::calcOpticalFlowBM(const GpuMat& prev, const GpuMat& curr, Size blo
                     maxX, maxY, acceptLevel, escapeLevel, buf.ptr<short2>(), ssCount, stream);
 }
 
-namespace optflowbm_fast
-{
-    void get_buffer_size(int src_cols, int src_rows, int search_window, int block_window, int& buffer_cols, int& buffer_rows);
-
-    template <typename T>
-    void calc(PtrStepSzb I0, PtrStepSzb I1, PtrStepSzf velx, PtrStepSzf vely, PtrStepi buffer, int search_window, int block_window, cudaStream_t stream);
-}
-
-void cv::gpu::FastOpticalFlowBM::operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy, int search_window, int block_window, Stream& stream)
-{
-    CV_Assert( I0.type() == CV_8UC1 );
-    CV_Assert( I1.size() == I0.size() && I1.type() == I0.type() );
-
-    int border_size = search_window / 2 + block_window / 2;
-    Size esize = I0.size() + Size(border_size, border_size) * 2;
-
-    ensureSizeIsEnough(esize, I0.type(), extended_I0);
-    ensureSizeIsEnough(esize, I0.type(), extended_I1);
-
-    copyMakeBorder(I0, extended_I0, border_size, border_size, border_size, border_size, cv::BORDER_DEFAULT, Scalar(), stream);
-    copyMakeBorder(I1, extended_I1, border_size, border_size, border_size, border_size, cv::BORDER_DEFAULT, Scalar(), stream);
-
-    GpuMat I0_hdr = extended_I0(Rect(Point2i(border_size, border_size), I0.size()));
-    GpuMat I1_hdr = extended_I1(Rect(Point2i(border_size, border_size), I0.size()));
-
-    int bcols, brows;
-    optflowbm_fast::get_buffer_size(I0.cols, I0.rows, search_window, block_window, bcols, brows);
-
-    ensureSizeIsEnough(brows, bcols, CV_32SC1, buffer);
-
-    flowx.create(I0.size(), CV_32FC1);
-    flowy.create(I0.size(), CV_32FC1);
-
-    optflowbm_fast::calc<uchar>(I0_hdr, I1_hdr, flowx, flowy, buffer, search_window, block_window, StreamAccessor::getStream(stream));
-}
-
 #endif // HAVE_CUDA
diff --git a/modules/gpuoptflow/src/bm_fast.cpp b/modules/gpuoptflow/src/bm_fast.cpp
new file mode 100644
index 0000000000..edab653861
--- /dev/null
+++ b/modules/gpuoptflow/src/bm_fast.cpp
@@ -0,0 +1,90 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+void cv::gpu::FastOpticalFlowBM::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, int, int, Stream&) { throw_no_cuda(); }
+
+#else // HAVE_CUDA
+
+namespace optflowbm_fast
+{
+    void get_buffer_size(int src_cols, int src_rows, int search_window, int block_window, int& buffer_cols, int& buffer_rows);
+
+    template <typename T>
+    void calc(PtrStepSzb I0, PtrStepSzb I1, PtrStepSzf velx, PtrStepSzf vely, PtrStepi buffer, int search_window, int block_window, cudaStream_t stream);
+}
+
+void cv::gpu::FastOpticalFlowBM::operator ()(const GpuMat& I0, const GpuMat& I1, GpuMat& flowx, GpuMat& flowy, int search_window, int block_window, Stream& stream)
+{
+    CV_Assert( I0.type() == CV_8UC1 );
+    CV_Assert( I1.size() == I0.size() && I1.type() == I0.type() );
+
+    int border_size = search_window / 2 + block_window / 2;
+    Size esize = I0.size() + Size(border_size, border_size) * 2;
+
+    ensureSizeIsEnough(esize, I0.type(), extended_I0);
+    ensureSizeIsEnough(esize, I0.type(), extended_I1);
+
+    gpu::copyMakeBorder(I0, extended_I0, border_size, border_size, border_size, border_size, cv::BORDER_DEFAULT, Scalar(), stream);
+    gpu::copyMakeBorder(I1, extended_I1, border_size, border_size, border_size, border_size, cv::BORDER_DEFAULT, Scalar(), stream);
+
+    GpuMat I0_hdr = extended_I0(Rect(Point2i(border_size, border_size), I0.size()));
+    GpuMat I1_hdr = extended_I1(Rect(Point2i(border_size, border_size), I0.size()));
+
+    int bcols, brows;
+    optflowbm_fast::get_buffer_size(I0.cols, I0.rows, search_window, block_window, bcols, brows);
+
+    ensureSizeIsEnough(brows, bcols, CV_32SC1, buffer);
+
+    flowx.create(I0.size(), CV_32FC1);
+    flowy.create(I0.size(), CV_32FC1);
+
+    optflowbm_fast::calc<uchar>(I0_hdr, I1_hdr, flowx, flowy, buffer, search_window, block_window, StreamAccessor::getStream(stream));
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/src/optical_flow.cpp b/modules/gpuoptflow/src/brox.cpp
similarity index 52%
rename from modules/gpu/src/optical_flow.cpp
rename to modules/gpuoptflow/src/brox.cpp
index 3d8fc05ec3..b5db69e2b1 100644
--- a/modules/gpu/src/optical_flow.cpp
+++ b/modules/gpuoptflow/src/brox.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -45,11 +45,9 @@
 using namespace cv;
 using namespace cv::gpu;
 
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+#if !defined (HAVE_CUDA) || !defined (HAVE_OPENCV_GPULEGACY) || defined (CUDA_DISABLER)
 
-void cv::gpu::BroxOpticalFlow::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::interpolateFrames(const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, float, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
-void cv::gpu::createOpticalFlowNeedleMap(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
+void cv::gpu::BroxOpticalFlow::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
 #else
 
@@ -68,7 +66,7 @@ namespace
 
 namespace
 {
-    static void outputHandler(const std::string &msg) { CV_Error(CV_GpuApiCallError, msg.c_str()); }
+    static void outputHandler(const String &msg) { CV_Error(cv::Error::GpuApiCallError, msg.c_str()); }
 }
 
 void cv::gpu::BroxOpticalFlow::operator ()(const GpuMat& frame0, const GpuMat& frame1, GpuMat& u, GpuMat& v, Stream& s)
@@ -129,109 +127,4 @@ void cv::gpu::BroxOpticalFlow::operator ()(const GpuMat& frame0, const GpuMat& f
     ncvSafeCall( NCVBroxOpticalFlow(desc, gpuAllocator, frame0Mat, frame1Mat, uMat, vMat, stream) );
 }
 
-void cv::gpu::interpolateFrames(const GpuMat& frame0, const GpuMat& frame1, const GpuMat& fu, const GpuMat& fv, const GpuMat& bu, const GpuMat& bv,
-                                float pos, GpuMat& newFrame, GpuMat& buf, Stream& s)
-{
-    CV_Assert(frame0.type() == CV_32FC1);
-    CV_Assert(frame1.size() == frame0.size() && frame1.type() == frame0.type());
-    CV_Assert(fu.size() == frame0.size() && fu.type() == frame0.type());
-    CV_Assert(fv.size() == frame0.size() && fv.type() == frame0.type());
-    CV_Assert(bu.size() == frame0.size() && bu.type() == frame0.type());
-    CV_Assert(bv.size() == frame0.size() && bv.type() == frame0.type());
-
-    newFrame.create(frame0.size(), frame0.type());
-
-    buf.create(6 * frame0.rows, frame0.cols, CV_32FC1);
-    buf.setTo(Scalar::all(0));
-
-    // occlusion masks
-    GpuMat occ0 = buf.rowRange(0 * frame0.rows, 1 * frame0.rows);
-    GpuMat occ1 = buf.rowRange(1 * frame0.rows, 2 * frame0.rows);
-
-    // interpolated forward flow
-    GpuMat fui = buf.rowRange(2 * frame0.rows, 3 * frame0.rows);
-    GpuMat fvi = buf.rowRange(3 * frame0.rows, 4 * frame0.rows);
-
-    // interpolated backward flow
-    GpuMat bui = buf.rowRange(4 * frame0.rows, 5 * frame0.rows);
-    GpuMat bvi = buf.rowRange(5 * frame0.rows, 6 * frame0.rows);
-
-    size_t step = frame0.step;
-
-    CV_Assert(frame1.step == step && fu.step == step && fv.step == step && bu.step == step && bv.step == step && newFrame.step == step && buf.step == step);
-
-    cudaStream_t stream = StreamAccessor::getStream(s);
-    NppStStreamHandler h(stream);
-
-    NppStInterpolationState state;
-
-    state.size         = NcvSize32u(frame0.cols, frame0.rows);
-    state.nStep        = static_cast<Ncv32u>(step);
-    state.pSrcFrame0   = const_cast<Ncv32f*>(frame0.ptr<Ncv32f>());
-    state.pSrcFrame1   = const_cast<Ncv32f*>(frame1.ptr<Ncv32f>());
-    state.pFU          = const_cast<Ncv32f*>(fu.ptr<Ncv32f>());
-    state.pFV          = const_cast<Ncv32f*>(fv.ptr<Ncv32f>());
-    state.pBU          = const_cast<Ncv32f*>(bu.ptr<Ncv32f>());
-    state.pBV          = const_cast<Ncv32f*>(bv.ptr<Ncv32f>());
-    state.pos          = pos;
-    state.pNewFrame    = newFrame.ptr<Ncv32f>();
-    state.ppBuffers[0] = occ0.ptr<Ncv32f>();
-    state.ppBuffers[1] = occ1.ptr<Ncv32f>();
-    state.ppBuffers[2] = fui.ptr<Ncv32f>();
-    state.ppBuffers[3] = fvi.ptr<Ncv32f>();
-    state.ppBuffers[4] = bui.ptr<Ncv32f>();
-    state.ppBuffers[5] = bvi.ptr<Ncv32f>();
-
-    ncvSafeCall( nppiStInterpolateFrames(&state) );
-
-    if (stream == 0)
-        cudaSafeCall( cudaDeviceSynchronize() );
-}
-
-namespace cv { namespace gpu { namespace device
-{
-    namespace optical_flow
-    {
-        void NeedleMapAverage_gpu(PtrStepSzf u, PtrStepSzf v, PtrStepSzf u_avg, PtrStepSzf v_avg);
-        void CreateOpticalFlowNeedleMap_gpu(PtrStepSzf u_avg, PtrStepSzf v_avg, float* vertex_buffer, float* color_data, float max_flow, float xscale, float yscale);
-    }
-}}}
-
-void cv::gpu::createOpticalFlowNeedleMap(const GpuMat& u, const GpuMat& v, GpuMat& vertex, GpuMat& colors)
-{
-    using namespace cv::gpu::device::optical_flow;
-
-    CV_Assert(u.type() == CV_32FC1);
-    CV_Assert(v.type() == u.type() && v.size() == u.size());
-
-    const int NEEDLE_MAP_SCALE = 16;
-
-    const int x_needles = u.cols / NEEDLE_MAP_SCALE;
-    const int y_needles = u.rows / NEEDLE_MAP_SCALE;
-
-    GpuMat u_avg(y_needles, x_needles, CV_32FC1);
-    GpuMat v_avg(y_needles, x_needles, CV_32FC1);
-
-    NeedleMapAverage_gpu(u, v, u_avg, v_avg);
-
-    const int NUM_VERTS_PER_ARROW = 6;
-
-    const int num_arrows = x_needles * y_needles * NUM_VERTS_PER_ARROW;
-
-    vertex.create(1, num_arrows, CV_32FC3);
-    colors.create(1, num_arrows, CV_32FC3);
-
-    colors.setTo(Scalar::all(1.0));
-
-    double uMax, vMax;
-    minMax(u_avg, 0, &uMax);
-    minMax(v_avg, 0, &vMax);
-
-    float max_flow = static_cast<float>(std::sqrt(uMax * uMax + vMax * vMax));
-
-    CreateOpticalFlowNeedleMap_gpu(u_avg, v_avg, vertex.ptr<float>(), colors.ptr<float>(), max_flow, 1.0f / u.cols, 1.0f / u.rows);
-
-    cvtColor(colors, colors, COLOR_HSV2RGB);
-}
-
 #endif /* HAVE_CUDA */
diff --git a/modules/gpuoptflow/src/cuda/bm.cu b/modules/gpuoptflow/src/cuda/bm.cu
new file mode 100644
index 0000000000..9150d29a14
--- /dev/null
+++ b/modules/gpuoptflow/src/cuda/bm.cu
@@ -0,0 +1,169 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace optflowbm
+{
+    texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_prev(false, cudaFilterModePoint, cudaAddressModeClamp);
+    texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_curr(false, cudaFilterModePoint, cudaAddressModeClamp);
+
+    __device__ int cmpBlocks(int X1, int Y1, int X2, int Y2, int2 blockSize)
+    {
+        int s = 0;
+
+        for (int y = 0; y < blockSize.y; ++y)
+        {
+            for (int x = 0; x < blockSize.x; ++x)
+                s += ::abs(tex2D(tex_prev, X1 + x, Y1 + y) - tex2D(tex_curr, X2 + x, Y2 + y));
+        }
+
+        return s;
+    }
+
+    __global__ void calcOptFlowBM(PtrStepSzf velx, PtrStepf vely, const int2 blockSize, const int2 shiftSize, const bool usePrevious,
+                                  const int maxX, const int maxY, const int acceptLevel, const int escapeLevel,
+                                  const short2* ss, const int ssCount)
+    {
+        const int j = blockIdx.x * blockDim.x + threadIdx.x;
+        const int i = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (i >= velx.rows || j >= velx.cols)
+            return;
+
+        const int X1 = j * shiftSize.x;
+        const int Y1 = i * shiftSize.y;
+
+        const int offX = usePrevious ? __float2int_rn(velx(i, j)) : 0;
+        const int offY = usePrevious ? __float2int_rn(vely(i, j)) : 0;
+
+        int X2 = X1 + offX;
+        int Y2 = Y1 + offY;
+
+        int dist = numeric_limits<int>::max();
+
+        if (0 <= X2 && X2 <= maxX && 0 <= Y2 && Y2 <= maxY)
+            dist = cmpBlocks(X1, Y1, X2, Y2, blockSize);
+
+        int countMin = 1;
+        int sumx = offX;
+        int sumy = offY;
+
+        if (dist > acceptLevel)
+        {
+            // do brute-force search
+            for (int k = 0; k < ssCount; ++k)
+            {
+                const short2 ssVal = ss[k];
+
+                const int dx = offX + ssVal.x;
+                const int dy = offY + ssVal.y;
+
+                X2 = X1 + dx;
+                Y2 = Y1 + dy;
+
+                if (0 <= X2 && X2 <= maxX && 0 <= Y2 && Y2 <= maxY)
+                {
+                    const int tmpDist = cmpBlocks(X1, Y1, X2, Y2, blockSize);
+                    if (tmpDist < acceptLevel)
+                    {
+                        sumx = dx;
+                        sumy = dy;
+                        countMin = 1;
+                        break;
+                    }
+
+                    if (tmpDist < dist)
+                    {
+                        dist = tmpDist;
+                        sumx = dx;
+                        sumy = dy;
+                        countMin = 1;
+                    }
+                    else if (tmpDist == dist)
+                    {
+                        sumx += dx;
+                        sumy += dy;
+                        countMin++;
+                    }
+                }
+            }
+
+            if (dist > escapeLevel)
+            {
+                sumx = offX;
+                sumy = offY;
+                countMin = 1;
+            }
+        }
+
+        velx(i, j) = static_cast<float>(sumx) / countMin;
+        vely(i, j) = static_cast<float>(sumy) / countMin;
+    }
+
+    void calc(PtrStepSzb prev, PtrStepSzb curr, PtrStepSzf velx, PtrStepSzf vely, int2 blockSize, int2 shiftSize, bool usePrevious,
+              int maxX, int maxY, int acceptLevel, int escapeLevel, const short2* ss, int ssCount, cudaStream_t stream)
+    {
+        bindTexture(&tex_prev, prev);
+        bindTexture(&tex_curr, curr);
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(velx.cols, block.x), divUp(vely.rows, block.y));
+
+        calcOptFlowBM<<<grid, block, 0, stream>>>(velx, vely, blockSize, shiftSize, usePrevious,
+                                                  maxX, maxY, acceptLevel,  escapeLevel, ss, ssCount);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}
+
+#endif // !defined CUDA_DISABLER
diff --git a/modules/gpu/src/cuda/optflowbm.cu b/modules/gpuoptflow/src/cuda/bm_fast.cu
similarity index 70%
rename from modules/gpu/src/cuda/optflowbm.cu
rename to modules/gpuoptflow/src/cuda/bm_fast.cu
index baf8dfb362..46f78a9f90 100644
--- a/modules/gpu/src/cuda/optflowbm.cu
+++ b/modules/gpuoptflow/src/cuda/bm_fast.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,132 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
-
-namespace optflowbm
-{
-    texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_prev(false, cudaFilterModePoint, cudaAddressModeClamp);
-    texture<uchar, cudaTextureType2D, cudaReadModeElementType> tex_curr(false, cudaFilterModePoint, cudaAddressModeClamp);
-
-    __device__ int cmpBlocks(int X1, int Y1, int X2, int Y2, int2 blockSize)
-    {
-        int s = 0;
-
-        for (int y = 0; y < blockSize.y; ++y)
-        {
-            for (int x = 0; x < blockSize.x; ++x)
-                s += ::abs(tex2D(tex_prev, X1 + x, Y1 + y) - tex2D(tex_curr, X2 + x, Y2 + y));
-        }
-
-        return s;
-    }
-
-    __global__ void calcOptFlowBM(PtrStepSzf velx, PtrStepf vely, const int2 blockSize, const int2 shiftSize, const bool usePrevious,
-                                  const int maxX, const int maxY, const int acceptLevel, const int escapeLevel,
-                                  const short2* ss, const int ssCount)
-    {
-        const int j = blockIdx.x * blockDim.x + threadIdx.x;
-        const int i = blockIdx.y * blockDim.y + threadIdx.y;
-
-        if (i >= velx.rows || j >= velx.cols)
-            return;
-
-        const int X1 = j * shiftSize.x;
-        const int Y1 = i * shiftSize.y;
-
-        const int offX = usePrevious ? __float2int_rn(velx(i, j)) : 0;
-        const int offY = usePrevious ? __float2int_rn(vely(i, j)) : 0;
-
-        int X2 = X1 + offX;
-        int Y2 = Y1 + offY;
-
-        int dist = numeric_limits<int>::max();
-
-        if (0 <= X2 && X2 <= maxX && 0 <= Y2 && Y2 <= maxY)
-            dist = cmpBlocks(X1, Y1, X2, Y2, blockSize);
-
-        int countMin = 1;
-        int sumx = offX;
-        int sumy = offY;
-
-        if (dist > acceptLevel)
-        {
-            // do brute-force search
-            for (int k = 0; k < ssCount; ++k)
-            {
-                const short2 ssVal = ss[k];
-
-                const int dx = offX + ssVal.x;
-                const int dy = offY + ssVal.y;
-
-                X2 = X1 + dx;
-                Y2 = Y1 + dy;
-
-                if (0 <= X2 && X2 <= maxX && 0 <= Y2 && Y2 <= maxY)
-                {
-                    const int tmpDist = cmpBlocks(X1, Y1, X2, Y2, blockSize);
-                    if (tmpDist < acceptLevel)
-                    {
-                        sumx = dx;
-                        sumy = dy;
-                        countMin = 1;
-                        break;
-                    }
-
-                    if (tmpDist < dist)
-                    {
-                        dist = tmpDist;
-                        sumx = dx;
-                        sumy = dy;
-                        countMin = 1;
-                    }
-                    else if (tmpDist == dist)
-                    {
-                        sumx += dx;
-                        sumy += dy;
-                        countMin++;
-                    }
-                }
-            }
-
-            if (dist > escapeLevel)
-            {
-                sumx = offX;
-                sumy = offY;
-                countMin = 1;
-            }
-        }
-
-        velx(i, j) = static_cast<float>(sumx) / countMin;
-        vely(i, j) = static_cast<float>(sumy) / countMin;
-    }
-
-    void calc(PtrStepSzb prev, PtrStepSzb curr, PtrStepSzf velx, PtrStepSzf vely, int2 blockSize, int2 shiftSize, bool usePrevious,
-              int maxX, int maxY, int acceptLevel, int escapeLevel, const short2* ss, int ssCount, cudaStream_t stream)
-    {
-        bindTexture(&tex_prev, prev);
-        bindTexture(&tex_curr, curr);
-
-        const dim3 block(32, 8);
-        const dim3 grid(divUp(velx.cols, block.x), divUp(vely.rows, block.y));
-
-        calcOptFlowBM<<<grid, block, 0, stream>>>(velx, vely, blockSize, shiftSize, usePrevious,
-                                                  maxX, maxY, acceptLevel,  escapeLevel, ss, ssCount);
-        cudaSafeCall( cudaGetLastError() );
-
-        if (stream == 0)
-            cudaSafeCall( cudaDeviceSynchronize() );
-    }
-}
-
-/////////////////////////////////////////////////////////
-// Fast approximate version
+using namespace cv::gpu::cudev;
 
 namespace optflowbm_fast
 {
diff --git a/modules/gpu/src/cuda/optical_flow_farneback.cu b/modules/gpuoptflow/src/cuda/farneback.cu
similarity index 95%
rename from modules/gpu/src/cuda/optical_flow_farneback.cu
rename to modules/gpuoptflow/src/cuda/farneback.cu
index 5bbca34f18..68a58c16d8 100644
--- a/modules/gpu/src/cuda/optical_flow_farneback.cu
+++ b/modules/gpuoptflow/src/cuda/farneback.cu
@@ -42,9 +42,8 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
 #define tx threadIdx.x
 #define ty threadIdx.y
@@ -56,7 +55,7 @@
 #define BORDER_SIZE 5
 #define MAX_KSIZE_HALF 100
 
-namespace cv { namespace gpu { namespace device { namespace optflow_farneback
+namespace cv { namespace gpu { namespace cudev { namespace optflow_farneback
 {
     __constant__ float c_g[8];
     __constant__ float c_xg[8];
@@ -526,8 +525,11 @@ namespace cv { namespace gpu { namespace device { namespace optflow_farneback
 
         static const caller_t callers[] =
         {
-            gaussianBlurCaller<BrdReflect101<float> >,
+            0 /*gaussianBlurCaller<BrdConstant<float> >*/,
             gaussianBlurCaller<BrdReplicate<float> >,
+            0 /*gaussianBlurCaller<BrdReflect<float> >*/,
+            0 /*gaussianBlurCaller<BrdWrap<float> >*/,
+            gaussianBlurCaller<BrdReflect101<float> >
         };
 
         callers[borderMode](src, ksizeHalf, dst, stream);
@@ -621,8 +623,11 @@ namespace cv { namespace gpu { namespace device { namespace optflow_farneback
 
         static const caller_t callers[] =
         {
-            gaussianBlur5Caller<BrdReflect101<float>,256>,
+            0 /*gaussianBlur5Caller<BrdConstant<float>,256>*/,
             gaussianBlur5Caller<BrdReplicate<float>,256>,
+            0 /*gaussianBlur5Caller<BrdReflect<float>,256>*/,
+            0 /*gaussianBlur5Caller<BrdWrap<float>,256>*/,
+            gaussianBlur5Caller<BrdReflect101<float>,256>
         };
 
         callers[borderMode](src, ksizeHalf, dst, stream);
@@ -635,14 +640,17 @@ namespace cv { namespace gpu { namespace device { namespace optflow_farneback
 
         static const caller_t callers[] =
         {
-            gaussianBlur5Caller<BrdReflect101<float>,128>,
+            0 /*gaussianBlur5Caller<BrdConstant<float>,128>*/,
             gaussianBlur5Caller<BrdReplicate<float>,128>,
+            0 /*gaussianBlur5Caller<BrdReflect<float>,128>*/,
+            0 /*gaussianBlur5Caller<BrdWrap<float>,128>*/,
+            gaussianBlur5Caller<BrdReflect101<float>,128>
         };
 
         callers[borderMode](src, ksizeHalf, dst, stream);
     }
 
-}}}} // namespace cv { namespace gpu { namespace device { namespace optflow_farneback
+}}}} // namespace cv { namespace gpu { namespace cudev { namespace optflow_farneback
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/optical_flow.cu b/modules/gpuoptflow/src/cuda/needle_map.cu
similarity index 98%
rename from modules/gpu/src/cuda/optical_flow.cu
rename to modules/gpuoptflow/src/cuda/needle_map.cu
index d2c68a79fa..e0b1ef6b78 100644
--- a/modules/gpu/src/cuda/optical_flow.cu
+++ b/modules/gpuoptflow/src/cuda/needle_map.cu
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace optical_flow
     {
@@ -140,7 +140,7 @@ namespace cv { namespace gpu { namespace device
                 const float u_avg_val = u_avg(y, x);
                 const float v_avg_val = v_avg(y, x);
 
-                const float theta = ::atan2f(v_avg_val, u_avg_val);// + CV_PI;
+                const float theta = ::atan2f(v_avg_val, u_avg_val);
 
                 float r = ::sqrtf(v_avg_val * v_avg_val + u_avg_val * u_avg_val);
                 r = fmin(14.0f * (r / max_flow), 14.0f);
diff --git a/modules/gpu/src/cuda/pyrlk.cu b/modules/gpuoptflow/src/cuda/pyrlk.cu
similarity index 97%
rename from modules/gpu/src/cuda/pyrlk.cu
rename to modules/gpuoptflow/src/cuda/pyrlk.cu
index 8d746143c8..410666fcfb 100644
--- a/modules/gpu/src/cuda/pyrlk.cu
+++ b/modules/gpuoptflow/src/cuda/pyrlk.cu
@@ -38,24 +38,19 @@
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
-// Copyright (c) 2010, Paul Furgale, Chi Hay Tong
-//
-// The original code was written by Paul Furgale and Chi Hay Tong
-// and later optimized and prepared for integration into OpenCV by Itseez.
-//
 //M*/
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 namespace pyrlk
 {
diff --git a/modules/gpu/src/cuda/tvl1flow.cu b/modules/gpuoptflow/src/cuda/tvl1flow.cu
similarity index 94%
rename from modules/gpu/src/cuda/tvl1flow.cu
rename to modules/gpuoptflow/src/cuda/tvl1flow.cu
index 27694ad269..75d30a624d 100644
--- a/modules/gpu/src/cuda/tvl1flow.cu
+++ b/modules/gpuoptflow/src/cuda/tvl1flow.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,12 +42,12 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/limits.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/limits.hpp"
 
 using namespace cv::gpu;
-using namespace cv::gpu::device;
+using namespace cv::gpu::cudev;
 
 ////////////////////////////////////////////////////////////
 // centeredGradient
@@ -211,7 +211,7 @@ namespace tvl1flow
                               const PtrStepf grad, const PtrStepf rho_c,
                               const PtrStepf p11, const PtrStepf p12, const PtrStepf p21, const PtrStepf p22,
                               PtrStepf u1, PtrStepf u2, PtrStepf error,
-                              const float l_t, const float theta)
+                              const float l_t, const float theta, const bool calcError)
     {
         const int x = blockIdx.x * blockDim.x + threadIdx.x;
         const int y = blockIdx.y * blockDim.y + threadIdx.y;
@@ -265,21 +265,24 @@ namespace tvl1flow
         u1(y, x) = u1NewVal;
         u2(y, x) = u2NewVal;
 
-        const float n1 = (u1OldVal - u1NewVal) * (u1OldVal - u1NewVal);
-        const float n2 = (u2OldVal - u2NewVal) * (u2OldVal - u2NewVal);
-        error(y, x) = n1 + n2;
+        if (calcError)
+        {
+            const float n1 = (u1OldVal - u1NewVal) * (u1OldVal - u1NewVal);
+            const float n2 = (u2OldVal - u2NewVal) * (u2OldVal - u2NewVal);
+            error(y, x) = n1 + n2;
+        }
     }
 
     void estimateU(PtrStepSzf I1wx, PtrStepSzf I1wy,
                    PtrStepSzf grad, PtrStepSzf rho_c,
                    PtrStepSzf p11, PtrStepSzf p12, PtrStepSzf p21, PtrStepSzf p22,
                    PtrStepSzf u1, PtrStepSzf u2, PtrStepSzf error,
-                   float l_t, float theta)
+                   float l_t, float theta, bool calcError)
     {
         const dim3 block(32, 8);
         const dim3 grid(divUp(I1wx.cols, block.x), divUp(I1wx.rows, block.y));
 
-        estimateUKernel<<<grid, block>>>(I1wx, I1wy, grad, rho_c, p11, p12, p21, p22, u1, u2, error, l_t, theta);
+        estimateUKernel<<<grid, block>>>(I1wx, I1wy, grad, rho_c, p11, p12, p21, p22, u1, u2, error, l_t, theta, calcError);
         cudaSafeCall( cudaGetLastError() );
 
         cudaSafeCall( cudaDeviceSynchronize() );
diff --git a/modules/gpu/src/optical_flow_farneback.cpp b/modules/gpuoptflow/src/farneback.cpp
similarity index 69%
rename from modules/gpu/src/optical_flow_farneback.cpp
rename to modules/gpuoptflow/src/farneback.cpp
index bc36a75e5e..9ed6403eea 100644
--- a/modules/gpu/src/optical_flow_farneback.cpp
+++ b/modules/gpuoptflow/src/farneback.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -55,11 +55,11 @@ using namespace cv::gpu;
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::FarnebackOpticalFlow::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::FarnebackOpticalFlow::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
 
 #else
 
-namespace cv { namespace gpu { namespace device { namespace optflow_farneback
+namespace cv { namespace gpu { namespace cudev { namespace optflow_farneback
 {
     void setPolynomialExpansionConsts(
             int polyN, const float *g, const float *xg, const float *xxg,
@@ -93,7 +93,7 @@ namespace cv { namespace gpu { namespace device { namespace optflow_farneback
     void gaussianBlur5Gpu_CC11(
             const PtrStepSzf src, int ksizeHalf, PtrStepSzf dst, int borderType, cudaStream_t stream);
 
-}}}} // namespace cv { namespace gpu { namespace device { namespace optflow_farneback
+}}}} // namespace cv { namespace gpu { namespace cudev { namespace optflow_farneback
 
 
 void cv::gpu::FarnebackOpticalFlow::prepareGaussian(
@@ -163,7 +163,7 @@ void cv::gpu::FarnebackOpticalFlow::setPolynomialExpansionConsts(int n, double s
     double ig11, ig03, ig33, ig55;
     prepareGaussian(n, sigma, g, xg, xxg, ig11, ig03, ig33, ig55);
 
-    device::optflow_farneback::setPolynomialExpansionConsts(n, g, xg, xxg, static_cast<float>(ig11), static_cast<float>(ig03), static_cast<float>(ig33), static_cast<float>(ig55));
+    cudev::optflow_farneback::setPolynomialExpansionConsts(n, g, xg, xxg, static_cast<float>(ig11), static_cast<float>(ig03), static_cast<float>(ig33), static_cast<float>(ig55));
 }
 
 
@@ -172,17 +172,17 @@ void cv::gpu::FarnebackOpticalFlow::updateFlow_boxFilter(
         GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[])
 {
     if (deviceSupports(FEATURE_SET_COMPUTE_12))
-        device::optflow_farneback::boxFilter5Gpu(M, blockSize/2, bufM, S(streams[0]));
+        cudev::optflow_farneback::boxFilter5Gpu(M, blockSize/2, bufM, S(streams[0]));
     else
-        device::optflow_farneback::boxFilter5Gpu_CC11(M, blockSize/2, bufM, S(streams[0]));
+        cudev::optflow_farneback::boxFilter5Gpu_CC11(M, blockSize/2, bufM, S(streams[0]));
     swap(M, bufM);
 
     for (int i = 1; i < 5; ++i)
         streams[i].waitForCompletion();
-    device::optflow_farneback::updateFlowGpu(M, flowx, flowy, S(streams[0]));
+    cudev::optflow_farneback::updateFlowGpu(M, flowx, flowy, S(streams[0]));
 
     if (updateMatrices)
-        device::optflow_farneback::updateMatricesGpu(flowx, flowy, R0, R1, M, S(streams[0]));
+        cudev::optflow_farneback::updateMatricesGpu(flowx, flowy, R0, R1, M, S(streams[0]));
 }
 
 
@@ -191,17 +191,17 @@ void cv::gpu::FarnebackOpticalFlow::updateFlow_gaussianBlur(
         GpuMat& M, GpuMat &bufM, int blockSize, bool updateMatrices, Stream streams[])
 {
     if (deviceSupports(FEATURE_SET_COMPUTE_12))
-        device::optflow_farneback::gaussianBlur5Gpu(
-                    M, blockSize/2, bufM, BORDER_REPLICATE_GPU, S(streams[0]));
+        cudev::optflow_farneback::gaussianBlur5Gpu(
+                    M, blockSize/2, bufM, BORDER_REPLICATE, S(streams[0]));
     else
-        device::optflow_farneback::gaussianBlur5Gpu_CC11(
-                    M, blockSize/2, bufM, BORDER_REPLICATE_GPU, S(streams[0]));
+        cudev::optflow_farneback::gaussianBlur5Gpu_CC11(
+                    M, blockSize/2, bufM, BORDER_REPLICATE, S(streams[0]));
     swap(M, bufM);
 
-    device::optflow_farneback::updateFlowGpu(M, flowx, flowy, S(streams[0]));
+    cudev::optflow_farneback::updateFlowGpu(M, flowx, flowy, S(streams[0]));
 
     if (updateMatrices)
-        device::optflow_farneback::updateMatricesGpu(flowx, flowy, R0, R1, M, S(streams[0]));
+        cudev::optflow_farneback::updateMatricesGpu(flowx, flowy, R0, R1, M, S(streams[0]));
 }
 
 
@@ -235,8 +235,8 @@ void cv::gpu::FarnebackOpticalFlow::operator ()(
             break;
     }
 
-    streams[0].enqueueConvert(frame0, frames_[0], CV_32F);
-    streams[1].enqueueConvert(frame1, frames_[1], CV_32F);
+    frame0.convertTo(frames_[0], CV_32F, streams[0]);
+    frame1.convertTo(frames_[1], CV_32F, streams[1]);
 
     if (fastPyramids)
     {
@@ -247,13 +247,13 @@ void cv::gpu::FarnebackOpticalFlow::operator ()(
         pyramid1_[0] = frames_[1];
         for (int i = 1; i <= numLevelsCropped; ++i)
         {
-            pyrDown(pyramid0_[i - 1], pyramid0_[i], streams[0]);
-            pyrDown(pyramid1_[i - 1], pyramid1_[i], streams[1]);
+            gpu::pyrDown(pyramid0_[i - 1], pyramid0_[i], streams[0]);
+            gpu::pyrDown(pyramid1_[i - 1], pyramid1_[i], streams[1]);
         }
     }
 
     setPolynomialExpansionConsts(polyN, polySigma);
-    device::optflow_farneback::setUpdateMatricesConsts();
+    cudev::optflow_farneback::setUpdateMatricesConsts();
 
     for (int k = numLevelsCropped; k >= 0; k--)
     {
@@ -291,47 +291,23 @@ void cv::gpu::FarnebackOpticalFlow::operator ()(
         {
             if (flags & OPTFLOW_USE_INITIAL_FLOW)
             {
-#if ENABLE_GPU_RESIZE
-                resize(flowx0, curFlowX, Size(width, height), 0, 0, INTER_LINEAR, streams[0]);
-                resize(flowy0, curFlowY, Size(width, height), 0, 0, INTER_LINEAR, streams[1]);
-                streams[0].enqueueConvert(curFlowX, curFlowX, curFlowX.depth(), scale);
-                streams[1].enqueueConvert(curFlowY, curFlowY, curFlowY.depth(), scale);
-#else
-                Mat tmp1, tmp2;
-                flowx0.download(tmp1);
-                resize(tmp1, tmp2, Size(width, height), 0, 0, INTER_AREA);
-                tmp2 *= scale;
-                curFlowX.upload(tmp2);
-                flowy0.download(tmp1);
-                resize(tmp1, tmp2, Size(width, height), 0, 0, INTER_AREA);
-                tmp2 *= scale;
-                curFlowY.upload(tmp2);
-#endif
+                gpu::resize(flowx0, curFlowX, Size(width, height), 0, 0, INTER_LINEAR, streams[0]);
+                gpu::resize(flowy0, curFlowY, Size(width, height), 0, 0, INTER_LINEAR, streams[1]);
+                curFlowX.convertTo(curFlowX, curFlowX.depth(), scale, streams[0]);
+                curFlowY.convertTo(curFlowY, curFlowY.depth(), scale, streams[1]);
             }
             else
             {
-                streams[0].enqueueMemSet(curFlowX, 0);
-                streams[1].enqueueMemSet(curFlowY, 0);
+                curFlowX.setTo(0, streams[0]);
+                curFlowY.setTo(0, streams[1]);
             }
         }
         else
         {
-#if ENABLE_GPU_RESIZE
-            resize(prevFlowX, curFlowX, Size(width, height), 0, 0, INTER_LINEAR, streams[0]);
-            resize(prevFlowY, curFlowY, Size(width, height), 0, 0, INTER_LINEAR, streams[1]);
-            streams[0].enqueueConvert(curFlowX, curFlowX, curFlowX.depth(), 1./pyrScale);
-            streams[1].enqueueConvert(curFlowY, curFlowY, curFlowY.depth(), 1./pyrScale);
-#else
-            Mat tmp1, tmp2;
-            prevFlowX.download(tmp1);
-            resize(tmp1, tmp2, Size(width, height), 0, 0, INTER_LINEAR);
-            tmp2 *= 1./pyrScale;
-            curFlowX.upload(tmp2);
-            prevFlowY.download(tmp1);
-            resize(tmp1, tmp2, Size(width, height), 0, 0, INTER_LINEAR);
-            tmp2 *= 1./pyrScale;
-            curFlowY.upload(tmp2);
-#endif
+            gpu::resize(prevFlowX, curFlowX, Size(width, height), 0, 0, INTER_LINEAR, streams[0]);
+            gpu::resize(prevFlowY, curFlowY, Size(width, height), 0, 0, INTER_LINEAR, streams[1]);
+            curFlowX.convertTo(curFlowX, curFlowX.depth(), 1./pyrScale, streams[0]);
+            curFlowY.convertTo(curFlowY, curFlowY.depth(), 1./pyrScale, streams[1]);
         }
 
         GpuMat M = allocMatFromBuf(5*height, width, CV_32F, M_);
@@ -344,8 +320,8 @@ void cv::gpu::FarnebackOpticalFlow::operator ()(
 
         if (fastPyramids)
         {
-            device::optflow_farneback::polynomialExpansionGpu(pyramid0_[k], polyN, R[0], S(streams[0]));
-            device::optflow_farneback::polynomialExpansionGpu(pyramid1_[k], polyN, R[1], S(streams[1]));
+            cudev::optflow_farneback::polynomialExpansionGpu(pyramid0_[k], polyN, R[0], S(streams[0]));
+            cudev::optflow_farneback::polynomialExpansionGpu(pyramid1_[k], polyN, R[1], S(streams[1]));
         }
         else
         {
@@ -361,31 +337,24 @@ void cv::gpu::FarnebackOpticalFlow::operator ()(
             };
 
             Mat g = getGaussianKernel(smoothSize, sigma, CV_32F);
-            device::optflow_farneback::setGaussianBlurKernel(g.ptr<float>(smoothSize/2), smoothSize/2);
+            cudev::optflow_farneback::setGaussianBlurKernel(g.ptr<float>(smoothSize/2), smoothSize/2);
 
             for (int i = 0; i < 2; i++)
             {
-                device::optflow_farneback::gaussianBlurGpu(
-                        frames_[i], smoothSize/2, blurredFrame[i], BORDER_REFLECT101_GPU, S(streams[i]));
-#if ENABLE_GPU_RESIZE
-                resize(blurredFrame[i], pyrLevel[i], Size(width, height), INTER_LINEAR, streams[i]);
-#else
-                Mat tmp1, tmp2;
-                tmp[i].download(tmp1);
-                resize(tmp1, tmp2, Size(width, height), INTER_LINEAR);
-                I[i].upload(tmp2);
-#endif
-                device::optflow_farneback::polynomialExpansionGpu(pyrLevel[i], polyN, R[i], S(streams[i]));
+                cudev::optflow_farneback::gaussianBlurGpu(
+                        frames_[i], smoothSize/2, blurredFrame[i], BORDER_REFLECT101, S(streams[i]));
+                gpu::resize(blurredFrame[i], pyrLevel[i], Size(width, height), 0.0, 0.0, INTER_LINEAR, streams[i]);
+                cudev::optflow_farneback::polynomialExpansionGpu(pyrLevel[i], polyN, R[i], S(streams[i]));
             }
         }
 
         streams[1].waitForCompletion();
-        device::optflow_farneback::updateMatricesGpu(curFlowX, curFlowY, R[0], R[1], M, S(streams[0]));
+        cudev::optflow_farneback::updateMatricesGpu(curFlowX, curFlowY, R[0], R[1], M, S(streams[0]));
 
         if (flags & OPTFLOW_FARNEBACK_GAUSSIAN)
         {
             Mat g = getGaussianKernel(winSize, winSize/2*0.3f, CV_32F);
-            device::optflow_farneback::setGaussianBlurKernel(g.ptr<float>(winSize/2), winSize/2);
+            cudev::optflow_farneback::setGaussianBlurKernel(g.ptr<float>(winSize/2), winSize/2);
         }
         for (int i = 0; i < numIters; i++)
         {
diff --git a/modules/gpuoptflow/src/interpolate_frames.cpp b/modules/gpuoptflow/src/interpolate_frames.cpp
new file mode 100644
index 0000000000..f6fe9c510c
--- /dev/null
+++ b/modules/gpuoptflow/src/interpolate_frames.cpp
@@ -0,0 +1,113 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || !defined (HAVE_OPENCV_GPULEGACY) || defined (CUDA_DISABLER)
+
+void cv::gpu::interpolateFrames(const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, const GpuMat&, float, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+
+#else
+
+void cv::gpu::interpolateFrames(const GpuMat& frame0, const GpuMat& frame1, const GpuMat& fu, const GpuMat& fv, const GpuMat& bu, const GpuMat& bv,
+                                float pos, GpuMat& newFrame, GpuMat& buf, Stream& s)
+{
+    CV_Assert(frame0.type() == CV_32FC1);
+    CV_Assert(frame1.size() == frame0.size() && frame1.type() == frame0.type());
+    CV_Assert(fu.size() == frame0.size() && fu.type() == frame0.type());
+    CV_Assert(fv.size() == frame0.size() && fv.type() == frame0.type());
+    CV_Assert(bu.size() == frame0.size() && bu.type() == frame0.type());
+    CV_Assert(bv.size() == frame0.size() && bv.type() == frame0.type());
+
+    newFrame.create(frame0.size(), frame0.type());
+
+    buf.create(6 * frame0.rows, frame0.cols, CV_32FC1);
+    buf.setTo(Scalar::all(0));
+
+    // occlusion masks
+    GpuMat occ0 = buf.rowRange(0 * frame0.rows, 1 * frame0.rows);
+    GpuMat occ1 = buf.rowRange(1 * frame0.rows, 2 * frame0.rows);
+
+    // interpolated forward flow
+    GpuMat fui = buf.rowRange(2 * frame0.rows, 3 * frame0.rows);
+    GpuMat fvi = buf.rowRange(3 * frame0.rows, 4 * frame0.rows);
+
+    // interpolated backward flow
+    GpuMat bui = buf.rowRange(4 * frame0.rows, 5 * frame0.rows);
+    GpuMat bvi = buf.rowRange(5 * frame0.rows, 6 * frame0.rows);
+
+    size_t step = frame0.step;
+
+    CV_Assert(frame1.step == step && fu.step == step && fv.step == step && bu.step == step && bv.step == step && newFrame.step == step && buf.step == step);
+
+    cudaStream_t stream = StreamAccessor::getStream(s);
+    NppStStreamHandler h(stream);
+
+    NppStInterpolationState state;
+
+    state.size         = NcvSize32u(frame0.cols, frame0.rows);
+    state.nStep        = static_cast<Ncv32u>(step);
+    state.pSrcFrame0   = const_cast<Ncv32f*>(frame0.ptr<Ncv32f>());
+    state.pSrcFrame1   = const_cast<Ncv32f*>(frame1.ptr<Ncv32f>());
+    state.pFU          = const_cast<Ncv32f*>(fu.ptr<Ncv32f>());
+    state.pFV          = const_cast<Ncv32f*>(fv.ptr<Ncv32f>());
+    state.pBU          = const_cast<Ncv32f*>(bu.ptr<Ncv32f>());
+    state.pBV          = const_cast<Ncv32f*>(bv.ptr<Ncv32f>());
+    state.pos          = pos;
+    state.pNewFrame    = newFrame.ptr<Ncv32f>();
+    state.ppBuffers[0] = occ0.ptr<Ncv32f>();
+    state.ppBuffers[1] = occ1.ptr<Ncv32f>();
+    state.ppBuffers[2] = fui.ptr<Ncv32f>();
+    state.ppBuffers[3] = fvi.ptr<Ncv32f>();
+    state.ppBuffers[4] = bui.ptr<Ncv32f>();
+    state.ppBuffers[5] = bvi.ptr<Ncv32f>();
+
+    ncvSafeCall( nppiStInterpolateFrames(&state) );
+
+    if (stream == 0)
+        cudaSafeCall( cudaDeviceSynchronize() );
+}
+
+#endif /* HAVE_CUDA */
diff --git a/modules/gpuoptflow/src/needle_map.cpp b/modules/gpuoptflow/src/needle_map.cpp
new file mode 100644
index 0000000000..9ca8fe5e44
--- /dev/null
+++ b/modules/gpuoptflow/src/needle_map.cpp
@@ -0,0 +1,100 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::createOpticalFlowNeedleMap(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_no_cuda(); }
+
+#else
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace optical_flow
+    {
+        void NeedleMapAverage_gpu(PtrStepSzf u, PtrStepSzf v, PtrStepSzf u_avg, PtrStepSzf v_avg);
+        void CreateOpticalFlowNeedleMap_gpu(PtrStepSzf u_avg, PtrStepSzf v_avg, float* vertex_buffer, float* color_data, float max_flow, float xscale, float yscale);
+    }
+}}}
+
+void cv::gpu::createOpticalFlowNeedleMap(const GpuMat& u, const GpuMat& v, GpuMat& vertex, GpuMat& colors)
+{
+    using namespace cv::gpu::cudev::optical_flow;
+
+    CV_Assert(u.type() == CV_32FC1);
+    CV_Assert(v.type() == u.type() && v.size() == u.size());
+
+    const int NEEDLE_MAP_SCALE = 16;
+
+    const int x_needles = u.cols / NEEDLE_MAP_SCALE;
+    const int y_needles = u.rows / NEEDLE_MAP_SCALE;
+
+    GpuMat u_avg(y_needles, x_needles, CV_32FC1);
+    GpuMat v_avg(y_needles, x_needles, CV_32FC1);
+
+    NeedleMapAverage_gpu(u, v, u_avg, v_avg);
+
+    const int NUM_VERTS_PER_ARROW = 6;
+
+    const int num_arrows = x_needles * y_needles * NUM_VERTS_PER_ARROW;
+
+    vertex.create(1, num_arrows, CV_32FC3);
+    colors.create(1, num_arrows, CV_32FC3);
+
+    colors.setTo(Scalar::all(1.0));
+
+    double uMax, vMax;
+    gpu::minMax(u_avg, 0, &uMax);
+    gpu::minMax(v_avg, 0, &vMax);
+
+    float max_flow = static_cast<float>(std::sqrt(uMax * uMax + vMax * vMax));
+
+    CreateOpticalFlowNeedleMap_gpu(u_avg, v_avg, vertex.ptr<float>(), colors.ptr<float>(), max_flow, 1.0f / u.cols, 1.0f / u.rows);
+
+    gpu::cvtColor(colors, colors, COLOR_HSV2RGB);
+}
+
+#endif /* HAVE_CUDA */
diff --git a/modules/gpuoptflow/src/precomp.hpp b/modules/gpuoptflow/src/precomp.hpp
new file mode 100644
index 0000000000..8a52a9ae9e
--- /dev/null
+++ b/modules/gpuoptflow/src/precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+
+#include "opencv2/gpuoptflow.hpp"
+#include "opencv2/gpuarithm.hpp"
+#include "opencv2/gpuwarping.hpp"
+#include "opencv2/gpuimgproc.hpp"
+#include "opencv2/video.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPULEGACY
+#  include "opencv2/gpulegacy/private.hpp"
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpu/src/pyrlk.cpp b/modules/gpuoptflow/src/pyrlk.cpp
similarity index 91%
rename from modules/gpu/src/pyrlk.cpp
rename to modules/gpuoptflow/src/pyrlk.cpp
index dac6f0bd9b..1992bf9038 100644
--- a/modules/gpu/src/pyrlk.cpp
+++ b/modules/gpuoptflow/src/pyrlk.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -47,9 +47,9 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-cv::gpu::PyrLKOpticalFlow::PyrLKOpticalFlow() { throw_nogpu(); }
-void cv::gpu::PyrLKOpticalFlow::sparse(const GpuMat&, const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat*) { throw_nogpu(); }
-void cv::gpu::PyrLKOpticalFlow::dense(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat*) { throw_nogpu(); }
+cv::gpu::PyrLKOpticalFlow::PyrLKOpticalFlow() { throw_no_cuda(); }
+void cv::gpu::PyrLKOpticalFlow::sparse(const GpuMat&, const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat*) { throw_no_cuda(); }
+void cv::gpu::PyrLKOpticalFlow::dense(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, GpuMat*) { throw_no_cuda(); }
 void cv::gpu::PyrLKOpticalFlow::releaseMemory() {}
 
 #else /* !defined (HAVE_CUDA) */
@@ -124,7 +124,7 @@ void cv::gpu::PyrLKOpticalFlow::sparse(const GpuMat& prevImg, const GpuMat& next
 
     GpuMat temp1 = (useInitialFlow ? nextPts : prevPts).reshape(1);
     GpuMat temp2 = nextPts.reshape(1);
-    multiply(temp1, Scalar::all(1.0 / (1 << maxLevel) / 2.0), temp2);
+    gpu::multiply(temp1, Scalar::all(1.0 / (1 << maxLevel) / 2.0), temp2);
 
     ensureSizeIsEnough(1, prevPts.cols, CV_8UC1, status);
     status.setTo(Scalar::all(1));
@@ -146,17 +146,17 @@ void cv::gpu::PyrLKOpticalFlow::sparse(const GpuMat& prevImg, const GpuMat& next
     }
     else
     {
-        cvtColor(prevImg, buf_, COLOR_BGR2BGRA);
+        gpu::cvtColor(prevImg, buf_, COLOR_BGR2BGRA);
         buf_.convertTo(prevPyr_[0], CV_32F);
 
-        cvtColor(nextImg, buf_, COLOR_BGR2BGRA);
+        gpu::cvtColor(nextImg, buf_, COLOR_BGR2BGRA);
         buf_.convertTo(nextPyr_[0], CV_32F);
     }
 
     for (int level = 1; level <= maxLevel; ++level)
     {
-        pyrDown(prevPyr_[level - 1], prevPyr_[level]);
-        pyrDown(nextPyr_[level - 1], nextPyr_[level]);
+        gpu::pyrDown(prevPyr_[level - 1], prevPyr_[level]);
+        gpu::pyrDown(nextPyr_[level - 1], nextPyr_[level]);
     }
 
     pyrlk::loadConstants(make_int2(winSize.width, winSize.height), iters);
@@ -198,8 +198,8 @@ void cv::gpu::PyrLKOpticalFlow::dense(const GpuMat& prevImg, const GpuMat& nextI
 
     for (int level = 1; level <= maxLevel; ++level)
     {
-        pyrDown(prevPyr_[level - 1], prevPyr_[level]);
-        pyrDown(nextPyr_[level - 1], nextPyr_[level]);
+        gpu::pyrDown(prevPyr_[level - 1], prevPyr_[level]);
+        gpu::pyrDown(nextPyr_[level - 1], nextPyr_[level]);
     }
 
     ensureSizeIsEnough(prevImg.size(), CV_32FC1, uPyr_[0]);
diff --git a/modules/gpu/src/tvl1flow.cpp b/modules/gpuoptflow/src/tvl1flow.cpp
similarity index 82%
rename from modules/gpu/src/tvl1flow.cpp
rename to modules/gpuoptflow/src/tvl1flow.cpp
index c6aa4be865..50b280157f 100644
--- a/modules/gpu/src/tvl1flow.cpp
+++ b/modules/gpuoptflow/src/tvl1flow.cpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -44,10 +44,10 @@
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-cv::gpu::OpticalFlowDual_TVL1_GPU::OpticalFlowDual_TVL1_GPU() { throw_nogpu(); }
-void cv::gpu::OpticalFlowDual_TVL1_GPU::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
+cv::gpu::OpticalFlowDual_TVL1_GPU::OpticalFlowDual_TVL1_GPU() { throw_no_cuda(); }
+void cv::gpu::OpticalFlowDual_TVL1_GPU::operator ()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_no_cuda(); }
 void cv::gpu::OpticalFlowDual_TVL1_GPU::collectGarbage() {}
-void cv::gpu::OpticalFlowDual_TVL1_GPU::procOneScale(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
+void cv::gpu::OpticalFlowDual_TVL1_GPU::procOneScale(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&) { throw_no_cuda(); }
 
 #else
 
@@ -63,6 +63,7 @@ cv::gpu::OpticalFlowDual_TVL1_GPU::OpticalFlowDual_TVL1_GPU()
     warps          = 5;
     epsilon        = 0.01;
     iterations     = 300;
+    scaleStep      = 0.8;
     useInitialFlow = false;
 }
 
@@ -112,8 +113,8 @@ void cv::gpu::OpticalFlowDual_TVL1_GPU::operator ()(const GpuMat& I0, const GpuM
     // create the scales
     for (int s = 1; s < nscales; ++s)
     {
-        gpu::pyrDown(I0s[s - 1], I0s[s]);
-        gpu::pyrDown(I1s[s - 1], I1s[s]);
+        gpu::resize(I0s[s-1], I0s[s], Size(), scaleStep, scaleStep);
+        gpu::resize(I1s[s-1], I1s[s], Size(), scaleStep, scaleStep);
 
         if (I0s[s].cols < 16 || I0s[s].rows < 16)
         {
@@ -123,12 +124,23 @@ void cv::gpu::OpticalFlowDual_TVL1_GPU::operator ()(const GpuMat& I0, const GpuM
 
         if (useInitialFlow)
         {
-            gpu::pyrDown(u1s[s - 1], u1s[s]);
-            gpu::pyrDown(u2s[s - 1], u2s[s]);
+            gpu::resize(u1s[s-1], u1s[s], Size(), scaleStep, scaleStep);
+            gpu::resize(u2s[s-1], u2s[s], Size(), scaleStep, scaleStep);
 
-            gpu::multiply(u1s[s], Scalar::all(0.5), u1s[s]);
-            gpu::multiply(u2s[s], Scalar::all(0.5), u2s[s]);
+            gpu::multiply(u1s[s], Scalar::all(scaleStep), u1s[s]);
+            gpu::multiply(u2s[s], Scalar::all(scaleStep), u2s[s]);
         }
+        else
+        {
+            u1s[s].create(I0s[s].size(), CV_32FC1);
+            u2s[s].create(I0s[s].size(), CV_32FC1);
+        }
+    }
+
+    if (!useInitialFlow)
+    {
+        u1s[nscales-1].setTo(Scalar::all(0));
+        u2s[nscales-1].setTo(Scalar::all(0));
     }
 
     // pyramidal structure for computing the optical flow
@@ -148,8 +160,8 @@ void cv::gpu::OpticalFlowDual_TVL1_GPU::operator ()(const GpuMat& I0, const GpuM
         gpu::resize(u2s[s], u2s[s - 1], I0s[s - 1].size());
 
         // scale the optical flow with the appropriate zoom factor
-        gpu::multiply(u1s[s - 1], Scalar::all(2), u1s[s - 1]);
-        gpu::multiply(u2s[s - 1], Scalar::all(2), u2s[s - 1]);
+        gpu::multiply(u1s[s - 1], Scalar::all(1/scaleStep), u1s[s - 1]);
+        gpu::multiply(u2s[s - 1], Scalar::all(1/scaleStep), u2s[s - 1]);
     }
 }
 
@@ -161,7 +173,7 @@ namespace tvl1flow
                    PtrStepSzf grad, PtrStepSzf rho_c,
                    PtrStepSzf p11, PtrStepSzf p12, PtrStepSzf p21, PtrStepSzf p22,
                    PtrStepSzf u1, PtrStepSzf u2, PtrStepSzf error,
-                   float l_t, float theta);
+                   float l_t, float theta, bool calcError);
     void estimateDualVariables(PtrStepSzf u1, PtrStepSzf u2, PtrStepSzf p11, PtrStepSzf p12, PtrStepSzf p21, PtrStepSzf p22, float taut);
 }
 
@@ -173,18 +185,9 @@ void cv::gpu::OpticalFlowDual_TVL1_GPU::procOneScale(const GpuMat& I0, const Gpu
 
     CV_DbgAssert( I1.size() == I0.size() );
     CV_DbgAssert( I1.type() == I0.type() );
-    CV_DbgAssert( u1.empty() || u1.size() == I0.size() );
+    CV_DbgAssert( u1.size() == I0.size() );
     CV_DbgAssert( u2.size() == u1.size() );
 
-    if (u1.empty())
-    {
-        u1.create(I0.size(), CV_32FC1);
-        u1.setTo(Scalar::all(0));
-
-        u2.create(I0.size(), CV_32FC1);
-        u2.setTo(Scalar::all(0));
-    }
-
     GpuMat I1x = I1x_buf(Rect(0, 0, I0.cols, I0.rows));
     GpuMat I1y = I1y_buf(Rect(0, 0, I0.cols, I0.rows));
     centeredGradient(I1, I1x, I1y);
@@ -215,11 +218,24 @@ void cv::gpu::OpticalFlowDual_TVL1_GPU::procOneScale(const GpuMat& I0, const Gpu
         warpBackward(I0, I1, I1x, I1y, u1, u2, I1w, I1wx, I1wy, grad, rho_c);
 
         double error = std::numeric_limits<double>::max();
+        double prevError = 0.0;
         for (int n = 0; error > scaledEpsilon && n < iterations; ++n)
         {
-            estimateU(I1wx, I1wy, grad, rho_c, p11, p12, p21, p22, u1, u2, diff, l_t, static_cast<float>(theta));
+            // some tweaks to make sum operation less frequently
+            bool calcError = (epsilon > 0) && (n & 0x1) && (prevError < scaledEpsilon);
 
-            error = gpu::sum(diff, norm_buf)[0];
+            estimateU(I1wx, I1wy, grad, rho_c, p11, p12, p21, p22, u1, u2, diff, l_t, static_cast<float>(theta), calcError);
+
+            if (calcError)
+            {
+                error = gpu::sum(diff, norm_buf)[0];
+                prevError = error;
+            }
+            else
+            {
+                error = std::numeric_limits<double>::max();
+                prevError -= scaledEpsilon;
+            }
 
             estimateDualVariables(u1, u2, p11, p12, p21, p22, taut);
         }
diff --git a/modules/gpuoptflow/test/test_main.cpp b/modules/gpuoptflow/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpuoptflow/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpu/test/test_optflow.cpp b/modules/gpuoptflow/test/test_optflow.cpp
similarity index 84%
rename from modules/gpu/test/test_optflow.cpp
rename to modules/gpuoptflow/test/test_optflow.cpp
index a97516f6d1..fce07551dc 100644
--- a/modules/gpu/test/test_optflow.cpp
+++ b/modules/gpuoptflow/test/test_optflow.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -40,9 +41,12 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/legacy.hpp"
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 //////////////////////////////////////////////////////
 // BroxOpticalFlow
 
@@ -99,8 +103,8 @@ GPU_TEST_P(BroxOpticalFlow, Regression)
     for (int i = 0; i < v_gold.rows; ++i)
         f.read(v_gold.ptr<char>(i), v_gold.cols * sizeof(float));
 
-    EXPECT_MAT_NEAR(u_gold, u, 0);
-    EXPECT_MAT_NEAR(v_gold, v, 0);
+    EXPECT_MAT_SIMILAR(u_gold, u, 1e-3);
+    EXPECT_MAT_SIMILAR(v_gold, v, 1e-3);
 #else
     std::ofstream f(fname.c_str(), std::ios_base::binary);
 
@@ -145,89 +149,7 @@ GPU_TEST_P(BroxOpticalFlow, OpticalFlowNan)
     EXPECT_TRUE(cv::checkRange(h_v));
 };
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, BroxOpticalFlow, ALL_DEVICES);
-
-//////////////////////////////////////////////////////
-// GoodFeaturesToTrack
-
-namespace
-{
-    IMPLEMENT_PARAM_CLASS(MinDistance, double)
-}
-
-PARAM_TEST_CASE(GoodFeaturesToTrack, cv::gpu::DeviceInfo, MinDistance)
-{
-    cv::gpu::DeviceInfo devInfo;
-    double minDistance;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        minDistance = GET_PARAM(1);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(GoodFeaturesToTrack, Accuracy)
-{
-    cv::Mat image = readImage("opticalflow/frame0.png", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(image.empty());
-
-    int maxCorners = 1000;
-    double qualityLevel = 0.01;
-
-    cv::gpu::GoodFeaturesToTrackDetector_GPU detector(maxCorners, qualityLevel, minDistance);
-
-    cv::gpu::GpuMat d_pts;
-    detector(loadMat(image), d_pts);
-
-    ASSERT_FALSE(d_pts.empty());
-
-    std::vector<cv::Point2f> pts(d_pts.cols);
-    cv::Mat pts_mat(1, d_pts.cols, CV_32FC2, (void*) &pts[0]);
-    d_pts.download(pts_mat);
-
-    std::vector<cv::Point2f> pts_gold;
-    cv::goodFeaturesToTrack(image, pts_gold, maxCorners, qualityLevel, minDistance);
-
-    ASSERT_EQ(pts_gold.size(), pts.size());
-
-    size_t mistmatch = 0;
-    for (size_t i = 0; i < pts.size(); ++i)
-    {
-        cv::Point2i a = pts_gold[i];
-        cv::Point2i b = pts[i];
-
-        bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
-
-        if (!eq)
-            ++mistmatch;
-    }
-
-    double bad_ratio = static_cast<double>(mistmatch) / pts.size();
-
-    ASSERT_LE(bad_ratio, 0.01);
-}
-
-GPU_TEST_P(GoodFeaturesToTrack, EmptyCorners)
-{
-    int maxCorners = 1000;
-    double qualityLevel = 0.01;
-
-    cv::gpu::GoodFeaturesToTrackDetector_GPU detector(maxCorners, qualityLevel, minDistance);
-
-    cv::gpu::GpuMat src(100, 100, CV_8UC1, cv::Scalar::all(0));
-    cv::gpu::GpuMat corners(1, maxCorners, CV_32FC2);
-
-    detector(src, corners);
-
-    ASSERT_TRUE(corners.empty());
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_Video, GoodFeaturesToTrack, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(MinDistance(0.0), MinDistance(3.0))));
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, BroxOpticalFlow, ALL_DEVICES);
 
 //////////////////////////////////////////////////////
 // PyrLKOpticalFlow
@@ -319,7 +241,7 @@ GPU_TEST_P(PyrLKOpticalFlow, Sparse)
     ASSERT_LE(bad_ratio, 0.01);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, PyrLKOpticalFlow, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, PyrLKOpticalFlow, testing::Combine(
     ALL_DEVICES,
     testing::Values(UseGray(true), UseGray(false))));
 
@@ -330,7 +252,7 @@ namespace
 {
     IMPLEMENT_PARAM_CLASS(PyrScale, double)
     IMPLEMENT_PARAM_CLASS(PolyN, int)
-    CV_FLAGS(FarnebackOptFlowFlags, 0, cv::OPTFLOW_FARNEBACK_GAUSSIAN)
+    CV_FLAGS(FarnebackOptFlowFlags, 0, OPTFLOW_FARNEBACK_GAUSSIAN)
     IMPLEMENT_PARAM_CLASS(UseInitFlow, bool)
 }
 
@@ -394,7 +316,7 @@ GPU_TEST_P(FarnebackOpticalFlow, Accuracy)
     EXPECT_MAT_SIMILAR(flowxy[1], d_flowy, 0.1);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, FarnebackOpticalFlow, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, FarnebackOpticalFlow, testing::Combine(
     ALL_DEVICES,
     testing::Values(PyrScale(0.3), PyrScale(0.5), PyrScale(0.8)),
     testing::Values(PolyN(5), PolyN(7)),
@@ -432,16 +354,19 @@ GPU_TEST_P(OpticalFlowDual_TVL1, Accuracy)
     d_alg(loadMat(frame0, useRoi), loadMat(frame1, useRoi), d_flowx, d_flowy);
 
     cv::Ptr<cv::DenseOpticalFlow> alg = cv::createOptFlow_DualTVL1();
+    alg->set("medianFiltering", 1);
+    alg->set("innerIterations", 1);
+    alg->set("outerIterations", d_alg.iterations);
     cv::Mat flow;
     alg->calc(frame0, frame1, flow);
     cv::Mat gold[2];
     cv::split(flow, gold);
 
-    EXPECT_MAT_SIMILAR(gold[0], d_flowx, 3e-3);
-    EXPECT_MAT_SIMILAR(gold[1], d_flowy, 3e-3);
+    EXPECT_MAT_SIMILAR(gold[0], d_flowx, 4e-3);
+    EXPECT_MAT_SIMILAR(gold[1], d_flowy, 4e-3);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, OpticalFlowDual_TVL1, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, OpticalFlowDual_TVL1, testing::Combine(
     ALL_DEVICES,
     WHOLE_SUBMAT));
 
@@ -500,7 +425,7 @@ GPU_TEST_P(OpticalFlowBM, Accuracy)
     EXPECT_MAT_NEAR(vely, d_vely, 0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, OpticalFlowBM, ALL_DEVICES);
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, OpticalFlowBM, ALL_DEVICES);
 
 //////////////////////////////////////////////////////
 // FastOpticalFlowBM
@@ -618,6 +543,6 @@ GPU_TEST_P(FastOpticalFlowBM, Accuracy)
     EXPECT_LE(err, MAX_RMSE);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_Video, FastOpticalFlowBM, ALL_DEVICES);
+INSTANTIATE_TEST_CASE_P(GPU_OptFlow, FastOpticalFlowBM, ALL_DEVICES);
 
 #endif // HAVE_CUDA
diff --git a/modules/gpuoptflow/test/test_precomp.hpp b/modules/gpuoptflow/test/test_precomp.hpp
new file mode 100644
index 0000000000..afac60d876
--- /dev/null
+++ b/modules/gpuoptflow/test/test_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include <fstream>
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpuoptflow.hpp"
+#include "opencv2/video.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpustereo/CMakeLists.txt b/modules/gpustereo/CMakeLists.txt
new file mode 100644
index 0000000000..c2a302c8a2
--- /dev/null
+++ b/modules/gpustereo/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpustereo)
+endif()
+
+set(the_description "GPU-accelerated Stereo Correspondence")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_define_module(gpustereo opencv_calib3d)
diff --git a/modules/gpustereo/doc/gpustereo.rst b/modules/gpustereo/doc/gpustereo.rst
new file mode 100644
index 0000000000..d3076794c1
--- /dev/null
+++ b/modules/gpustereo/doc/gpustereo.rst
@@ -0,0 +1,8 @@
+************************************************
+gpustereo. GPU-accelerated Stereo Correspondence
+************************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    stereo
diff --git a/modules/gpustereo/doc/stereo.rst b/modules/gpustereo/doc/stereo.rst
new file mode 100644
index 0000000000..f87d43ff87
--- /dev/null
+++ b/modules/gpustereo/doc/stereo.rst
@@ -0,0 +1,336 @@
+Stereo Correspondence
+=====================
+
+.. highlight:: cpp
+
+.. note::
+
+   * A basic stereo matching example can be found at opencv_source_code/samples/gpu/stereo_match.cpp
+   * A stereo matching example using several GPU's can be found at opencv_source_code/samples/gpu/stereo_multi.cpp
+   * A stereo matching example using several GPU's and driver API can be found at opencv_source_code/samples/gpu/driver_api_stereo_multi.cpp
+
+gpu::StereoBM
+-------------
+.. ocv:class:: gpu::StereoBM : public cv::StereoBM
+
+Class computing stereo correspondence (disparity map) using the block matching algorithm. ::
+
+.. seealso:: :ocv:class:`StereoBM`
+
+
+
+gpu::createStereoBM
+-------------------
+Creates StereoBM object.
+
+.. ocv:function:: Ptr<gpu::StereoBM> gpu::createStereoBM(int numDisparities = 64, int blockSize = 19)
+
+    :param numDisparities: the disparity search range. For each pixel algorithm will find the best disparity from 0 (default minimum disparity) to ``numDisparities``. The search range can then be shifted by changing the minimum disparity.
+
+    :param blockSize: the linear size of the blocks compared by the algorithm. The size should be odd (as the block is centered at the current pixel). Larger block size implies smoother, though less accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher chance for algorithm to find a wrong correspondence.
+
+
+
+gpu::StereoBeliefPropagation
+----------------------------
+.. ocv:class:: gpu::StereoBeliefPropagation : public cv::StereoMatcher
+
+Class computing stereo correspondence using the belief propagation algorithm. ::
+
+    class CV_EXPORTS StereoBeliefPropagation : public cv::StereoMatcher
+    {
+    public:
+        using cv::StereoMatcher::compute;
+
+        virtual void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream) = 0;
+
+        //! version for user specified data term
+        virtual void compute(InputArray data, OutputArray disparity, Stream& stream = Stream::Null()) = 0;
+
+        //! number of BP iterations on each level
+        virtual int getNumIters() const = 0;
+        virtual void setNumIters(int iters) = 0;
+
+        //! number of levels
+        virtual int getNumLevels() const = 0;
+        virtual void setNumLevels(int levels) = 0;
+
+        //! truncation of data cost
+        virtual double getMaxDataTerm() const = 0;
+        virtual void setMaxDataTerm(double max_data_term) = 0;
+
+        //! data weight
+        virtual double getDataWeight() const = 0;
+        virtual void setDataWeight(double data_weight) = 0;
+
+        //! truncation of discontinuity cost
+        virtual double getMaxDiscTerm() const = 0;
+        virtual void setMaxDiscTerm(double max_disc_term) = 0;
+
+        //! discontinuity single jump
+        virtual double getDiscSingleJump() const = 0;
+        virtual void setDiscSingleJump(double disc_single_jump) = 0;
+
+        virtual int getMsgType() const = 0;
+        virtual void setMsgType(int msg_type) = 0;
+
+        static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels);
+    };
+
+
+The class implements algorithm described in [Felzenszwalb2006]_ . It can compute own data cost (using a truncated linear model) or use a user-provided data cost.
+
+.. note::
+
+    ``StereoBeliefPropagation`` requires a lot of memory for message storage:
+
+    .. math::
+
+        width \_ step  \cdot height  \cdot ndisp  \cdot 4  \cdot (1 + 0.25)
+
+    and for data cost storage:
+
+    .. math::
+
+        width\_step \cdot height \cdot ndisp \cdot (1 + 0.25 + 0.0625 +  \dotsm + \frac{1}{4^{levels}})
+
+    ``width_step`` is the number of bytes in a line including padding.
+
+``StereoBeliefPropagation`` uses a truncated linear model for the data cost and discontinuity terms:
+
+.. math::
+
+    DataCost = data \_ weight  \cdot \min ( \lvert Img_Left(x,y)-Img_Right(x-d,y)  \rvert , max \_ data \_ term)
+
+.. math::
+
+    DiscTerm =  \min (disc \_ single \_ jump  \cdot \lvert f_1-f_2  \rvert , max \_ disc \_ term)
+
+For more details, see [Felzenszwalb2006]_.
+
+By default, ``StereoBeliefPropagation`` uses floating-point arithmetics and the ``CV_32FC1`` type for messages. But it can also use fixed-point arithmetics and the ``CV_16SC1`` message type for better performance. To avoid an overflow in this case, the parameters must satisfy the following requirement:
+
+.. math::
+
+    10  \cdot 2^{levels-1}  \cdot max \_ data \_ term < SHRT \_ MAX
+
+.. seealso:: :ocv:class:`StereoMatcher`
+
+
+
+gpu::createStereoBeliefPropagation
+----------------------------------
+Creates StereoBeliefPropagation object.
+
+.. ocv:function:: Ptr<gpu::StereoBeliefPropagation> gpu::createStereoBeliefPropagation(int ndisp = 64, int iters = 5, int levels = 5, int msg_type = CV_32F)
+
+    :param ndisp: Number of disparities.
+
+    :param iters: Number of BP iterations on each level.
+
+    :param levels: Number of levels.
+
+    :param msg_type: Type for messages.  ``CV_16SC1``  and  ``CV_32FC1`` types are supported.
+
+
+
+gpu::StereoBeliefPropagation::estimateRecommendedParams
+-------------------------------------------------------
+Uses a heuristic method to compute the recommended parameters ( ``ndisp``, ``iters`` and ``levels`` ) for the specified image size ( ``width`` and ``height`` ).
+
+.. ocv:function:: void gpu::StereoBeliefPropagation::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels)
+
+
+
+gpu::StereoBeliefPropagation::compute
+-------------------------------------
+Enables the stereo correspondence operator that finds the disparity for the specified data cost.
+
+.. ocv:function:: void gpu::StereoBeliefPropagation::compute(InputArray data, OutputArray disparity, Stream& stream = Stream::Null())
+
+    :param data: User-specified data cost, a matrix of ``msg_type`` type and ``Size(<image columns>*ndisp, <image rows>)`` size.
+
+    :param disparity: Output disparity map. If  ``disparity``  is empty, the output type is  ``CV_16SC1`` . Otherwise, the type is retained.
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::StereoConstantSpaceBP
+--------------------------
+.. ocv:class:: gpu::StereoConstantSpaceBP : public gpu::StereoBeliefPropagation
+
+Class computing stereo correspondence using the constant space belief propagation algorithm. ::
+
+    class CV_EXPORTS StereoConstantSpaceBP : public gpu::StereoBeliefPropagation
+    {
+    public:
+        //! number of active disparity on the first level
+        virtual int getNrPlane() const = 0;
+        virtual void setNrPlane(int nr_plane) = 0;
+
+        virtual bool getUseLocalInitDataCost() const = 0;
+        virtual void setUseLocalInitDataCost(bool use_local_init_data_cost) = 0;
+
+        static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane);
+    };
+
+
+The class implements algorithm described in [Yang2010]_. ``StereoConstantSpaceBP`` supports both local minimum and global minimum data cost initialization algorithms. For more details, see the paper mentioned above. By default, a local algorithm is used. To enable a global algorithm, set ``use_local_init_data_cost`` to ``false`` .
+
+``StereoConstantSpaceBP`` uses a truncated linear model for the data cost and discontinuity terms:
+
+.. math::
+
+    DataCost = data \_ weight  \cdot \min ( \lvert I_2-I_1  \rvert , max \_ data \_ term)
+
+.. math::
+
+    DiscTerm =  \min (disc \_ single \_ jump  \cdot \lvert f_1-f_2  \rvert , max \_ disc \_ term)
+
+For more details, see [Yang2010]_.
+
+By default, ``StereoConstantSpaceBP`` uses floating-point arithmetics and the ``CV_32FC1`` type for messages. But it can also use fixed-point arithmetics and the ``CV_16SC1`` message type for better performance. To avoid an overflow in this case, the parameters must satisfy the following requirement:
+
+.. math::
+
+    10  \cdot 2^{levels-1}  \cdot max \_ data \_ term < SHRT \_ MAX
+
+
+
+gpu::createStereoConstantSpaceBP
+--------------------------------
+Creates StereoConstantSpaceBP object.
+
+.. ocv:function:: Ptr<gpu::StereoConstantSpaceBP> gpu::createStereoConstantSpaceBP(int ndisp = 128, int iters = 8, int levels = 4, int nr_plane = 4, int msg_type = CV_32F)
+
+    :param ndisp: Number of disparities.
+
+    :param iters: Number of BP iterations on each level.
+
+    :param levels: Number of levels.
+
+    :param nr_plane: Number of disparity levels on the first level.
+
+    :param msg_type: Type for messages.  ``CV_16SC1``  and  ``CV_32FC1`` types are supported.
+
+
+
+gpu::StereoConstantSpaceBP::estimateRecommendedParams
+-----------------------------------------------------
+Uses a heuristic method to compute parameters (ndisp, iters, levelsand nrplane) for the specified image size (widthand height).
+
+.. ocv:function:: void gpu::StereoConstantSpaceBP::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane)
+
+
+
+gpu::DisparityBilateralFilter
+-----------------------------
+.. ocv:class:: gpu::DisparityBilateralFilter : public cv::Algorithm
+
+Class refining a disparity map using joint bilateral filtering. ::
+
+    class CV_EXPORTS DisparityBilateralFilter : public cv::Algorithm
+    {
+    public:
+        //! the disparity map refinement operator. Refine disparity map using joint bilateral filtering given a single color image.
+        //! disparity must have CV_8U or CV_16S type, image must have CV_8UC1 or CV_8UC3 type.
+        virtual void apply(InputArray disparity, InputArray image, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+
+        virtual int getNumDisparities() const = 0;
+        virtual void setNumDisparities(int numDisparities) = 0;
+
+        virtual int getRadius() const = 0;
+        virtual void setRadius(int radius) = 0;
+
+        virtual int getNumIters() const = 0;
+        virtual void setNumIters(int iters) = 0;
+
+        //! truncation of data continuity
+        virtual double getEdgeThreshold() const = 0;
+        virtual void setEdgeThreshold(double edge_threshold) = 0;
+
+        //! truncation of disparity continuity
+        virtual double getMaxDiscThreshold() const = 0;
+        virtual void setMaxDiscThreshold(double max_disc_threshold) = 0;
+
+        //! filter range sigma
+        virtual double getSigmaRange() const = 0;
+        virtual void setSigmaRange(double sigma_range) = 0;
+    };
+
+
+The class implements [Yang2010]_ algorithm.
+
+
+
+gpu::createDisparityBilateralFilter
+-----------------------------------
+Creates DisparityBilateralFilter object.
+
+.. ocv:function:: Ptr<gpu::DisparityBilateralFilter> gpu::createDisparityBilateralFilter(int ndisp = 64, int radius = 3, int iters = 1)
+
+    :param ndisp: Number of disparities.
+
+    :param radius: Filter radius.
+
+    :param iters: Number of iterations.
+
+
+
+gpu::DisparityBilateralFilter::apply
+------------------------------------
+Refines a disparity map using joint bilateral filtering.
+
+.. ocv:function:: void gpu::DisparityBilateralFilter::apply(InputArray disparity, InputArray image, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param disparity: Input disparity map.  ``CV_8UC1``  and  ``CV_16SC1``  types are supported.
+
+    :param image: Input image. ``CV_8UC1``  and  ``CV_8UC3``  types are supported.
+
+    :param dst: Destination disparity map. It has the same size and type as  ``disparity`` .
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::reprojectImageTo3D
+-----------------------
+Reprojects a disparity image to 3D space.
+
+.. ocv:function:: void gpu::reprojectImageTo3D(InputArray disp, OutputArray xyzw, InputArray Q, int dst_cn = 4, Stream& stream = Stream::Null())
+
+    :param disp: Input disparity image.  ``CV_8U``  and  ``CV_16S``  types are supported.
+
+    :param xyzw: Output 3- or 4-channel floating-point image of the same size as  ``disp`` . Each element of  ``xyzw(x,y)``  contains 3D coordinates ``(x,y,z)`` or ``(x,y,z,1)``  of the point  ``(x,y)`` , computed from the disparity map.
+
+    :param Q: :math:`4 \times 4`  perspective transformation matrix that can be obtained via  :ocv:func:`stereoRectify` .
+
+    :param dst_cn: The number of channels for output image. Can be 3 or 4.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`reprojectImageTo3D`
+
+
+
+gpu::drawColorDisp
+------------------
+Colors a disparity image.
+
+.. ocv:function:: void gpu::drawColorDisp(InputArray src_disp, OutputArray dst_disp, int ndisp, Stream& stream = Stream::Null())
+
+    :param src_disp: Source disparity image.  ``CV_8UC1``  and  ``CV_16SC1``  types are supported.
+
+    :param dst_disp: Output disparity image. It has the same size as  ``src_disp`` . The  type is ``CV_8UC4``  in  ``BGRA``  format (alpha = 255).
+
+    :param ndisp: Number of disparities.
+
+    :param stream: Stream for the asynchronous version.
+
+This function draws a colored disparity map by converting disparity values from ``[0..ndisp)`` interval first to ``HSV`` color space (where different disparity values correspond to different hues) and then converting the pixels to ``RGB`` for visualization.
+
+
+
+.. [Felzenszwalb2006] Pedro F. Felzenszwalb algorithm [Pedro F. Felzenszwalb and Daniel P. Huttenlocher. *Efficient belief propagation for early vision*. International Journal of Computer Vision, 70(1), October 2006
+.. [Yang2010] Q. Yang, L. Wang, and N. Ahuja. *A constant-space belief propagation algorithm for stereo matching*. In CVPR, 2010.
diff --git a/modules/gpustereo/include/opencv2/gpustereo.hpp b/modules/gpustereo/include/opencv2/gpustereo.hpp
new file mode 100644
index 0000000000..250e89b85c
--- /dev/null
+++ b/modules/gpustereo/include/opencv2/gpustereo.hpp
@@ -0,0 +1,193 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUSTEREO_HPP__
+#define __OPENCV_GPUSTEREO_HPP__
+
+#ifndef __cplusplus
+#  error gpustereo.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/calib3d.hpp"
+
+namespace cv { namespace gpu {
+
+/////////////////////////////////////////
+// StereoBM
+
+class CV_EXPORTS StereoBM : public cv::StereoBM
+{
+public:
+    using cv::StereoBM::compute;
+
+    virtual void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream) = 0;
+};
+
+CV_EXPORTS Ptr<gpu::StereoBM> createStereoBM(int numDisparities = 64, int blockSize = 19);
+
+/////////////////////////////////////////
+// StereoBeliefPropagation
+
+//! "Efficient Belief Propagation for Early Vision" P.Felzenszwalb
+class CV_EXPORTS StereoBeliefPropagation : public cv::StereoMatcher
+{
+public:
+    using cv::StereoMatcher::compute;
+
+    virtual void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream) = 0;
+
+    //! version for user specified data term
+    virtual void compute(InputArray data, OutputArray disparity, Stream& stream = Stream::Null()) = 0;
+
+    //! number of BP iterations on each level
+    virtual int getNumIters() const = 0;
+    virtual void setNumIters(int iters) = 0;
+
+    //! number of levels
+    virtual int getNumLevels() const = 0;
+    virtual void setNumLevels(int levels) = 0;
+
+    //! truncation of data cost
+    virtual double getMaxDataTerm() const = 0;
+    virtual void setMaxDataTerm(double max_data_term) = 0;
+
+    //! data weight
+    virtual double getDataWeight() const = 0;
+    virtual void setDataWeight(double data_weight) = 0;
+
+    //! truncation of discontinuity cost
+    virtual double getMaxDiscTerm() const = 0;
+    virtual void setMaxDiscTerm(double max_disc_term) = 0;
+
+    //! discontinuity single jump
+    virtual double getDiscSingleJump() const = 0;
+    virtual void setDiscSingleJump(double disc_single_jump) = 0;
+
+    //! type for messages (CV_16SC1 or CV_32FC1)
+    virtual int getMsgType() const = 0;
+    virtual void setMsgType(int msg_type) = 0;
+
+    static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels);
+};
+
+CV_EXPORTS Ptr<gpu::StereoBeliefPropagation>
+    createStereoBeliefPropagation(int ndisp = 64, int iters = 5, int levels = 5, int msg_type = CV_32F);
+
+/////////////////////////////////////////
+// StereoConstantSpaceBP
+
+//! "A Constant-Space Belief Propagation Algorithm for Stereo Matching"
+//! Qingxiong Yang, Liang Wang, Narendra Ahuja
+//! http://vision.ai.uiuc.edu/~qyang6/
+class CV_EXPORTS StereoConstantSpaceBP : public gpu::StereoBeliefPropagation
+{
+public:
+    //! number of active disparity on the first level
+    virtual int getNrPlane() const = 0;
+    virtual void setNrPlane(int nr_plane) = 0;
+
+    virtual bool getUseLocalInitDataCost() const = 0;
+    virtual void setUseLocalInitDataCost(bool use_local_init_data_cost) = 0;
+
+    static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane);
+};
+
+CV_EXPORTS Ptr<gpu::StereoConstantSpaceBP>
+    createStereoConstantSpaceBP(int ndisp = 128, int iters = 8, int levels = 4, int nr_plane = 4, int msg_type = CV_32F);
+
+/////////////////////////////////////////
+// DisparityBilateralFilter
+
+//! Disparity map refinement using joint bilateral filtering given a single color image.
+//! Qingxiong Yang, Liang Wang, Narendra Ahuja
+//! http://vision.ai.uiuc.edu/~qyang6/
+class CV_EXPORTS DisparityBilateralFilter : public cv::Algorithm
+{
+public:
+    //! the disparity map refinement operator. Refine disparity map using joint bilateral filtering given a single color image.
+    //! disparity must have CV_8U or CV_16S type, image must have CV_8UC1 or CV_8UC3 type.
+    virtual void apply(InputArray disparity, InputArray image, OutputArray dst, Stream& stream = Stream::Null()) = 0;
+
+    virtual int getNumDisparities() const = 0;
+    virtual void setNumDisparities(int numDisparities) = 0;
+
+    virtual int getRadius() const = 0;
+    virtual void setRadius(int radius) = 0;
+
+    virtual int getNumIters() const = 0;
+    virtual void setNumIters(int iters) = 0;
+
+    //! truncation of data continuity
+    virtual double getEdgeThreshold() const = 0;
+    virtual void setEdgeThreshold(double edge_threshold) = 0;
+
+    //! truncation of disparity continuity
+    virtual double getMaxDiscThreshold() const = 0;
+    virtual void setMaxDiscThreshold(double max_disc_threshold) = 0;
+
+    //! filter range sigma
+    virtual double getSigmaRange() const = 0;
+    virtual void setSigmaRange(double sigma_range) = 0;
+};
+
+CV_EXPORTS Ptr<gpu::DisparityBilateralFilter>
+    createDisparityBilateralFilter(int ndisp = 64, int radius = 3, int iters = 1);
+
+/////////////////////////////////////////
+// Utility
+
+//! Reprojects disparity image to 3D space.
+//! Supports CV_8U and CV_16S types of input disparity.
+//! The output is a 3- or 4-channel floating-point matrix.
+//! Each element of this matrix will contain the 3D coordinates of the point (x,y,z,1), computed from the disparity map.
+//! Q is the 4x4 perspective transformation matrix that can be obtained with cvStereoRectify.
+CV_EXPORTS void reprojectImageTo3D(InputArray disp, OutputArray xyzw, InputArray Q, int dst_cn = 4, Stream& stream = Stream::Null());
+
+//! Does coloring of disparity image: [0..ndisp) -> [0..240, 1, 1] in HSV.
+//! Supported types of input disparity: CV_8U, CV_16S.
+//! Output disparity has CV_8UC4 type in BGRA format (alpha = 255).
+CV_EXPORTS void drawColorDisp(InputArray src_disp, OutputArray dst_disp, int ndisp, Stream& stream = Stream::Null());
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUSTEREO_HPP__ */
diff --git a/modules/gpustereo/perf/perf_main.cpp b/modules/gpustereo/perf/perf_main.cpp
new file mode 100644
index 0000000000..0766ccf05d
--- /dev/null
+++ b/modules/gpustereo/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpustereo)
diff --git a/modules/gpustereo/perf/perf_precomp.hpp b/modules/gpustereo/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..f61de9fae5
--- /dev/null
+++ b/modules/gpustereo/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpustereo.hpp"
+#include "opencv2/calib3d.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpustereo/perf/perf_stereo.cpp b/modules/gpustereo/perf/perf_stereo.cpp
new file mode 100644
index 0000000000..476a591a1c
--- /dev/null
+++ b/modules/gpustereo/perf/perf_stereo.cpp
@@ -0,0 +1,253 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// StereoBM
+
+typedef std::tr1::tuple<string, string> pair_string;
+DEF_PARAM_TEST_1(ImagePair, pair_string);
+
+PERF_TEST_P(ImagePair, StereoBM,
+            Values(pair_string("gpu/perf/aloe.png", "gpu/perf/aloeR.png")))
+{
+    declare.time(300.0);
+
+    const cv::Mat imgLeft = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(imgLeft.empty());
+
+    const cv::Mat imgRight = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(imgRight.empty());
+
+    const int ndisp = 256;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::StereoBM> d_bm = cv::gpu::createStereoBM(ndisp);
+
+        const cv::gpu::GpuMat d_imgLeft(imgLeft);
+        const cv::gpu::GpuMat d_imgRight(imgRight);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() d_bm->compute(d_imgLeft, d_imgRight, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Ptr<cv::StereoBM> bm = cv::createStereoBM(ndisp);
+
+        cv::Mat dst;
+
+        TEST_CYCLE() bm->compute(imgLeft, imgRight, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// StereoBeliefPropagation
+
+PERF_TEST_P(ImagePair, StereoBeliefPropagation,
+            Values(pair_string("gpu/stereobp/aloe-L.png", "gpu/stereobp/aloe-R.png")))
+{
+    declare.time(300.0);
+
+    const cv::Mat imgLeft = readImage(GET_PARAM(0));
+    ASSERT_FALSE(imgLeft.empty());
+
+    const cv::Mat imgRight = readImage(GET_PARAM(1));
+    ASSERT_FALSE(imgRight.empty());
+
+    const int ndisp = 64;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::StereoBeliefPropagation> d_bp = cv::gpu::createStereoBeliefPropagation(ndisp);
+
+        const cv::gpu::GpuMat d_imgLeft(imgLeft);
+        const cv::gpu::GpuMat d_imgRight(imgRight);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() d_bp->compute(d_imgLeft, d_imgRight, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// StereoConstantSpaceBP
+
+PERF_TEST_P(ImagePair, StereoConstantSpaceBP,
+            Values(pair_string("gpu/stereobm/aloe-L.png", "gpu/stereobm/aloe-R.png")))
+{
+    declare.time(300.0);
+
+    const cv::Mat imgLeft = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(imgLeft.empty());
+
+    const cv::Mat imgRight = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(imgRight.empty());
+
+    const int ndisp = 128;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::StereoConstantSpaceBP> d_csbp = cv::gpu::createStereoConstantSpaceBP(ndisp);
+
+        const cv::gpu::GpuMat d_imgLeft(imgLeft);
+        const cv::gpu::GpuMat d_imgRight(imgRight);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() d_csbp->compute(d_imgLeft, d_imgRight, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// DisparityBilateralFilter
+
+PERF_TEST_P(ImagePair, DisparityBilateralFilter,
+            Values(pair_string("gpu/stereobm/aloe-L.png", "gpu/stereobm/aloe-disp.png")))
+{
+    const cv::Mat img = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    const cv::Mat disp = readImage(GET_PARAM(1), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(disp.empty());
+
+    const int ndisp = 128;
+
+    if (PERF_RUN_GPU())
+    {
+        cv::Ptr<cv::gpu::DisparityBilateralFilter> d_filter = cv::gpu::createDisparityBilateralFilter(ndisp);
+
+        const cv::gpu::GpuMat d_img(img);
+        const cv::gpu::GpuMat d_disp(disp);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() d_filter->apply(d_disp, d_img, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// ReprojectImageTo3D
+
+PERF_TEST_P(Sz_Depth, ReprojectImageTo3D,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16S)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+
+    cv::Mat src(size, depth);
+    declare.in(src, WARMUP_RNG);
+
+    cv::Mat Q(4, 4, CV_32FC1);
+    cv::randu(Q, 0.1, 1.0);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::reprojectImageTo3D(d_src, dst, Q);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::reprojectImageTo3D(src, dst, Q);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// DrawColorDisp
+
+PERF_TEST_P(Sz_Depth, DrawColorDisp,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16S)))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int type = GET_PARAM(1);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::drawColorDisp(d_src, dst, 255);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpu/src/cuda/disp_bilateral_filter.cu b/modules/gpustereo/src/cuda/disparity_bilateral_filter.cu
similarity index 96%
rename from modules/gpu/src/cuda/disp_bilateral_filter.cu
rename to modules/gpustereo/src/cuda/disparity_bilateral_filter.cu
index 56b39eaa7d..cfea880ecd 100644
--- a/modules/gpu/src/cuda/disp_bilateral_filter.cu
+++ b/modules/gpustereo/src/cuda/disparity_bilateral_filter.cu
@@ -42,10 +42,10 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/limits.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/limits.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace disp_bilateral_filter
     {
@@ -208,7 +208,7 @@ namespace cv { namespace gpu { namespace device
                 }
                 break;
             default:
-                cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "disp_bilateral_filter");
+                CV_Error(cv::Error::BadNumChannels, "Unsupported channels count");
             }
 
             if (stream == 0)
@@ -218,6 +218,6 @@ namespace cv { namespace gpu { namespace device
         template void disp_bilateral_filter<uchar>(PtrStepSz<uchar> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream);
         template void disp_bilateral_filter<short>(PtrStepSz<short> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream);
     } // namespace bilateral_filter
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/stereobm.cu b/modules/gpustereo/src/cuda/stereobm.cu
similarity index 98%
rename from modules/gpu/src/cuda/stereobm.cu
rename to modules/gpustereo/src/cuda/stereobm.cu
index 44ad0abd88..920616b72e 100644
--- a/modules/gpu/src/cuda/stereobm.cu
+++ b/modules/gpustereo/src/cuda/stereobm.cu
@@ -42,9 +42,9 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace stereobm
     {
@@ -348,7 +348,7 @@ namespace cv { namespace gpu { namespace device
             int winsz2 = winsz >> 1;
 
             if (winsz2 == 0 || winsz2 >= calles_num)
-                cv::gpu::error("Unsupported window size", __FILE__, __LINE__, "stereoBM_GPU");
+                CV_Error(cv::Error::StsBadArg, "Unsupported window size");
 
             //cudaSafeCall( cudaFuncSetCacheConfig(&stereoKernel, cudaFuncCachePreferL1) );
             //cudaSafeCall( cudaFuncSetCacheConfig(&stereoKernel, cudaFuncCachePreferShared) );
@@ -534,7 +534,7 @@ namespace cv { namespace gpu { namespace device
             cudaSafeCall( cudaUnbindTexture (texForTF) );
         }
     } // namespace stereobm
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/stereobp.cu b/modules/gpustereo/src/cuda/stereobp.cu
similarity index 98%
rename from modules/gpu/src/cuda/stereobp.cu
rename to modules/gpustereo/src/cuda/stereobp.cu
index 18d3ae7979..ff437edc27 100644
--- a/modules/gpu/src/cuda/stereobp.cu
+++ b/modules/gpustereo/src/cuda/stereobp.cu
@@ -28,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,11 +42,11 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/limits.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/limits.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace stereobp
     {
@@ -384,7 +384,7 @@ namespace cv { namespace gpu { namespace device
         template <typename T>
         __device__ void message(const T* msg1, const T* msg2, const T* msg3, const T* data, T* dst, size_t msg_disp_step, size_t data_disp_step)
         {
-            float minimum = device::numeric_limits<float>::max();
+            float minimum = cudev::numeric_limits<float>::max();
 
             for(int i = 0; i < cndisp; ++i)
             {
@@ -533,6 +533,6 @@ namespace cv { namespace gpu { namespace device
         template void output_gpu<short>(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, const PtrStepSz<short>& disp, cudaStream_t stream);
         template void output_gpu<float>(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, const PtrStepSz<short>& disp, cudaStream_t stream);
     } // namespace stereobp
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/stereocsbp.cu b/modules/gpustereo/src/cuda/stereocsbp.cu
similarity index 97%
rename from modules/gpu/src/cuda/stereocsbp.cu
rename to modules/gpustereo/src/cuda/stereocsbp.cu
index 7b76f478b4..cb1340a86d 100644
--- a/modules/gpu/src/cuda/stereocsbp.cu
+++ b/modules/gpustereo/src/cuda/stereocsbp.cu
@@ -42,13 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/functional.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/functional.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace stereocsbp
     {
@@ -146,7 +146,7 @@ namespace cv { namespace gpu { namespace device
 
                 for(int i = 0; i < nr_plane; i++)
                 {
-                    T minimum = device::numeric_limits<T>::max();
+                    T minimum = cudev::numeric_limits<T>::max();
                     int id = 0;
                     for(int d = 0; d < cndisp; d++)
                     {
@@ -324,7 +324,7 @@ namespace cv { namespace gpu { namespace device
             case 1: init_data_cost<T, 1><<<grid, threads, 0, stream>>>(h, w, level); break;
             case 3: init_data_cost<T, 3><<<grid, threads, 0, stream>>>(h, w, level); break;
             case 4: init_data_cost<T, 4><<<grid, threads, 0, stream>>>(h, w, level); break;
-            default: cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "init_data_cost_caller_");
+            default: CV_Error(cv::Error::BadNumChannels, "Unsupported channels count");
             }
         }
 
@@ -343,7 +343,7 @@ namespace cv { namespace gpu { namespace device
             case 1: init_data_cost_reduce<T, winsz, 1><<<grid, threads, smem_size, stream>>>(level, rows, cols, h); break;
             case 3: init_data_cost_reduce<T, winsz, 3><<<grid, threads, smem_size, stream>>>(level, rows, cols, h); break;
             case 4: init_data_cost_reduce<T, winsz, 4><<<grid, threads, smem_size, stream>>>(level, rows, cols, h); break;
-            default: cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "init_data_cost_reduce_caller_");
+            default: CV_Error(cv::Error::BadNumChannels, "Unsupported channels count");
             }
         }
 
@@ -506,7 +506,7 @@ namespace cv { namespace gpu { namespace device
             case 1: compute_data_cost<T, 1><<<grid, threads, 0, stream>>>(disp_selected_pyr, data_cost, h, w, level, nr_plane); break;
             case 3: compute_data_cost<T, 3><<<grid, threads, 0, stream>>>(disp_selected_pyr, data_cost, h, w, level, nr_plane); break;
             case 4: compute_data_cost<T, 4><<<grid, threads, 0, stream>>>(disp_selected_pyr, data_cost, h, w, level, nr_plane); break;
-            default: cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "compute_data_cost_caller_");
+            default: CV_Error(cv::Error::BadNumChannels, "Unsupported channels count");
             }
         }
 
@@ -526,7 +526,7 @@ namespace cv { namespace gpu { namespace device
             case 1: compute_data_cost_reduce<T, winsz, 1><<<grid, threads, smem_size, stream>>>(disp_selected_pyr, data_cost, level, rows, cols, h, nr_plane); break;
             case 3: compute_data_cost_reduce<T, winsz, 3><<<grid, threads, smem_size, stream>>>(disp_selected_pyr, data_cost, level, rows, cols, h, nr_plane); break;
             case 4: compute_data_cost_reduce<T, winsz, 4><<<grid, threads, smem_size, stream>>>(disp_selected_pyr, data_cost, level, rows, cols, h, nr_plane); break;
-            default: cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "compute_data_cost_reduce_caller_");
+            default: CV_Error(cv::Error::BadNumChannels, "Unsupported channels count");
             }
         }
 
@@ -859,6 +859,6 @@ namespace cv { namespace gpu { namespace device
         template void compute_disp(const float* u, const float* d, const float* l, const float* r, const float* data_cost_selected, const float* disp_selected, size_t msg_step,
             const PtrStepSz<short>& disp, int nr_plane, cudaStream_t stream);
     } // namespace stereocsbp
-}}} // namespace cv { namespace gpu { namespace device {
+}}} // namespace cv { namespace gpu { namespace cudev {
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpustereo/src/cuda/util.cu b/modules/gpustereo/src/cuda/util.cu
new file mode 100644
index 0000000000..1945d2463a
--- /dev/null
+++ b/modules/gpustereo/src/cuda/util.cu
@@ -0,0 +1,235 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    /////////////////////////////////// reprojectImageTo3D ///////////////////////////////////////////////
+
+    __constant__ float cq[16];
+
+    template <typename T, typename D>
+    __global__ void reprojectImageTo3D(const PtrStepSz<T> disp, PtrStep<D> xyz)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (y >= disp.rows || x >= disp.cols)
+            return;
+
+        const float qx = x * cq[ 0] + y * cq[ 1] + cq[ 3];
+        const float qy = x * cq[ 4] + y * cq[ 5] + cq[ 7];
+        const float qz = x * cq[ 8] + y * cq[ 9] + cq[11];
+        const float qw = x * cq[12] + y * cq[13] + cq[15];
+
+        const T d = disp(y, x);
+
+        const float iW = 1.f / (qw + cq[14] * d);
+
+        D v = VecTraits<D>::all(1.0f);
+        v.x = (qx + cq[2] * d) * iW;
+        v.y = (qy + cq[6] * d) * iW;
+        v.z = (qz + cq[10] * d) * iW;
+
+        xyz(y, x) = v;
+    }
+
+    template <typename T, typename D>
+    void reprojectImageTo3D_gpu(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream)
+    {
+        dim3 block(32, 8);
+        dim3 grid(divUp(disp.cols, block.x), divUp(disp.rows, block.y));
+
+        cudaSafeCall( cudaMemcpyToSymbol(cq, q, 16 * sizeof(float)) );
+
+        reprojectImageTo3D<T, D><<<grid, block, 0, stream>>>((PtrStepSz<T>)disp, (PtrStepSz<D>)xyz);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void reprojectImageTo3D_gpu<uchar, float3>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+    template void reprojectImageTo3D_gpu<uchar, float4>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+    template void reprojectImageTo3D_gpu<short, float3>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+    template void reprojectImageTo3D_gpu<short, float4>(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+
+    /////////////////////////////////// drawColorDisp ///////////////////////////////////////////////
+
+    template <typename T>
+    __device__ unsigned int cvtPixel(T d, int ndisp, float S = 1, float V = 1)
+    {
+        unsigned int H = ((ndisp-d) * 240)/ndisp;
+
+        unsigned int hi = (H/60) % 6;
+        float f = H/60.f - H/60;
+        float p = V * (1 - S);
+        float q = V * (1 - f * S);
+        float t = V * (1 - (1 - f) * S);
+
+        float3 res;
+
+        if (hi == 0) //R = V,	G = t,	B = p
+        {
+            res.x = p;
+            res.y = t;
+            res.z = V;
+        }
+
+        if (hi == 1) // R = q,	G = V,	B = p
+        {
+            res.x = p;
+            res.y = V;
+            res.z = q;
+        }
+
+        if (hi == 2) // R = p,	G = V,	B = t
+        {
+            res.x = t;
+            res.y = V;
+            res.z = p;
+        }
+
+        if (hi == 3) // R = p,	G = q,	B = V
+        {
+            res.x = V;
+            res.y = q;
+            res.z = p;
+        }
+
+        if (hi == 4) // R = t,	G = p,	B = V
+        {
+            res.x = V;
+            res.y = p;
+            res.z = t;
+        }
+
+        if (hi == 5) // R = V,	G = p,	B = q
+        {
+            res.x = q;
+            res.y = p;
+            res.z = V;
+        }
+        const unsigned int b = (unsigned int)(::max(0.f, ::min(res.x, 1.f)) * 255.f);
+        const unsigned int g = (unsigned int)(::max(0.f, ::min(res.y, 1.f)) * 255.f);
+        const unsigned int r = (unsigned int)(::max(0.f, ::min(res.z, 1.f)) * 255.f);
+        const unsigned int a = 255U;
+
+        return (a << 24) + (r << 16) + (g << 8) + b;
+    }
+
+    __global__ void drawColorDisp(uchar* disp, size_t disp_step, uchar* out_image, size_t out_step, int width, int height, int ndisp)
+    {
+        const int x = (blockIdx.x * blockDim.x + threadIdx.x) << 2;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if(x < width && y < height)
+        {
+            uchar4 d4 = *(uchar4*)(disp + y * disp_step + x);
+
+            uint4 res;
+            res.x = cvtPixel(d4.x, ndisp);
+            res.y = cvtPixel(d4.y, ndisp);
+            res.z = cvtPixel(d4.z, ndisp);
+            res.w = cvtPixel(d4.w, ndisp);
+
+            uint4* line = (uint4*)(out_image + y * out_step);
+            line[x >> 2] = res;
+        }
+    }
+
+    __global__ void drawColorDisp(short* disp, size_t disp_step, uchar* out_image, size_t out_step, int width, int height, int ndisp)
+    {
+        const int x = (blockIdx.x * blockDim.x + threadIdx.x) << 1;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if(x < width && y < height)
+        {
+            short2 d2 = *(short2*)(disp + y * disp_step + x);
+
+            uint2 res;
+            res.x = cvtPixel(d2.x, ndisp);
+            res.y = cvtPixel(d2.y, ndisp);
+
+            uint2* line = (uint2*)(out_image + y * out_step);
+            line[x >> 1] = res;
+        }
+    }
+
+
+    void drawColorDisp_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream)
+    {
+        dim3 threads(16, 16, 1);
+        dim3 grid(1, 1, 1);
+        grid.x = divUp(src.cols, threads.x << 2);
+        grid.y = divUp(src.rows, threads.y);
+
+        drawColorDisp<<<grid, threads, 0, stream>>>(src.data, src.step, dst.data, dst.step, src.cols, src.rows, ndisp);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    void drawColorDisp_gpu(const PtrStepSz<short>& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream)
+    {
+        dim3 threads(32, 8, 1);
+        dim3 grid(1, 1, 1);
+        grid.x = divUp(src.cols, threads.x << 1);
+        grid.y = divUp(src.rows, threads.y);
+
+        drawColorDisp<<<grid, threads, 0, stream>>>(src.data, src.step / sizeof(short), dst.data, dst.step, src.cols, src.rows, ndisp);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+}}} // namespace cv { namespace gpu { namespace cudev
+
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpustereo/src/disparity_bilateral_filter.cpp b/modules/gpustereo/src/disparity_bilateral_filter.cpp
new file mode 100644
index 0000000000..9808b7f66e
--- /dev/null
+++ b/modules/gpustereo/src/disparity_bilateral_filter.cpp
@@ -0,0 +1,206 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<gpu::DisparityBilateralFilter> cv::gpu::createDisparityBilateralFilter(int, int, int) { throw_no_cuda(); return Ptr<gpu::DisparityBilateralFilter>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace disp_bilateral_filter
+    {
+        void disp_load_constants(float* table_color, PtrStepSzf table_space, int ndisp, int radius, short edge_disc, short max_disc);
+
+        template<typename T>
+        void disp_bilateral_filter(PtrStepSz<T> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    class DispBilateralFilterImpl : public gpu::DisparityBilateralFilter
+    {
+    public:
+        DispBilateralFilterImpl(int ndisp, int radius, int iters);
+
+        void apply(InputArray disparity, InputArray image, OutputArray dst, Stream& stream);
+
+        int getNumDisparities() const { return ndisp_; }
+        void setNumDisparities(int numDisparities) { ndisp_ = numDisparities; }
+
+        int getRadius() const { return radius_; }
+        void setRadius(int radius);
+
+        int getNumIters() const { return iters_; }
+        void setNumIters(int iters) { iters_ = iters; }
+
+        double getEdgeThreshold() const { return edge_threshold_; }
+        void setEdgeThreshold(double edge_threshold) { edge_threshold_ = (float) edge_threshold; }
+
+        double getMaxDiscThreshold() const { return max_disc_threshold_; }
+        void setMaxDiscThreshold(double max_disc_threshold) { max_disc_threshold_ = (float) max_disc_threshold; }
+
+        double getSigmaRange() const { return sigma_range_; }
+        void setSigmaRange(double sigma_range);
+
+    private:
+        int ndisp_;
+        int radius_;
+        int iters_;
+        float edge_threshold_;
+        float max_disc_threshold_;
+        float sigma_range_;
+
+        GpuMat table_color_;
+        GpuMat table_space_;
+    };
+
+    void calc_color_weighted_table(GpuMat& table_color, float sigma_range, int len)
+    {
+        Mat cpu_table_color(1, len, CV_32F);
+
+        float* line = cpu_table_color.ptr<float>();
+
+        for(int i = 0; i < len; i++)
+            line[i] = static_cast<float>(std::exp(-double(i * i) / (2 * sigma_range * sigma_range)));
+
+        table_color.upload(cpu_table_color);
+    }
+
+    void calc_space_weighted_filter(GpuMat& table_space, int win_size, float dist_space)
+    {
+        int half = (win_size >> 1);
+
+        Mat cpu_table_space(half + 1, half + 1, CV_32F);
+
+        for (int y = 0; y <= half; ++y)
+        {
+            float* row = cpu_table_space.ptr<float>(y);
+            for (int x = 0; x <= half; ++x)
+                row[x] = exp(-sqrt(float(y * y) + float(x * x)) / dist_space);
+        }
+
+        table_space.upload(cpu_table_space);
+    }
+
+    const float DEFAULT_EDGE_THRESHOLD = 0.1f;
+    const float DEFAULT_MAX_DISC_THRESHOLD = 0.2f;
+    const float DEFAULT_SIGMA_RANGE = 10.0f;
+
+    DispBilateralFilterImpl::DispBilateralFilterImpl(int ndisp, int radius, int iters) :
+        ndisp_(ndisp), radius_(radius), iters_(iters),
+        edge_threshold_(DEFAULT_EDGE_THRESHOLD), max_disc_threshold_(DEFAULT_MAX_DISC_THRESHOLD),
+        sigma_range_(DEFAULT_SIGMA_RANGE)
+    {
+        calc_color_weighted_table(table_color_, sigma_range_, 255);
+        calc_space_weighted_filter(table_space_, radius_ * 2 + 1, radius_ + 1.0f);
+    }
+
+    void DispBilateralFilterImpl::setRadius(int radius)
+    {
+        radius_ = radius;
+        calc_space_weighted_filter(table_space_, radius_ * 2 + 1, radius_ + 1.0f);
+    }
+
+    void DispBilateralFilterImpl::setSigmaRange(double sigma_range)
+    {
+        sigma_range_ = (float) sigma_range;
+        calc_color_weighted_table(table_color_, sigma_range_, 255);
+    }
+
+    template <typename T>
+    void disp_bilateral_filter_operator(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold,
+                                        GpuMat& table_color, GpuMat& table_space,
+                                        const GpuMat& disp, const GpuMat& img,
+                                        OutputArray _dst, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::disp_bilateral_filter;
+
+        const short edge_disc = std::max<short>(short(1), short(ndisp * edge_threshold + 0.5));
+        const short max_disc = short(ndisp * max_disc_threshold + 0.5);
+
+        disp_load_constants(table_color.ptr<float>(), table_space, ndisp, radius, edge_disc, max_disc);
+
+        _dst.create(disp.size(), disp.type());
+        GpuMat dst = _dst.getGpuMat();
+
+        if (dst.data != disp.data)
+            disp.copyTo(dst, stream);
+
+        disp_bilateral_filter<T>(dst, img, img.channels(), iters, StreamAccessor::getStream(stream));
+    }
+
+    void DispBilateralFilterImpl::apply(InputArray _disp, InputArray _image, OutputArray dst, Stream& stream)
+    {
+        typedef void (*bilateral_filter_operator_t)(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold,
+                                                    GpuMat& table_color, GpuMat& table_space,
+                                                    const GpuMat& disp, const GpuMat& img, OutputArray dst, Stream& stream);
+        const bilateral_filter_operator_t operators[] =
+            {disp_bilateral_filter_operator<unsigned char>, 0, 0, disp_bilateral_filter_operator<short>, 0, 0, 0, 0};
+
+        CV_Assert( 0 < ndisp_ && 0 < radius_ && 0 < iters_ );
+
+        GpuMat disp = _disp.getGpuMat();
+        GpuMat img = _image.getGpuMat();
+
+        CV_Assert( disp.type() == CV_8U || disp.type() == CV_16S );
+        CV_Assert( img.type() == CV_8UC1 || img.type() == CV_8UC3 );
+        CV_Assert( disp.size() == img.size() );
+
+        operators[disp.type()](ndisp_, radius_, iters_, edge_threshold_, max_disc_threshold_,
+                               table_color_, table_space_, disp, img, dst, stream);
+    }
+}
+
+Ptr<gpu::DisparityBilateralFilter> cv::gpu::createDisparityBilateralFilter(int ndisp, int radius, int iters)
+{
+    return makePtr<DispBilateralFilterImpl>(ndisp, radius, iters);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpustereo/src/precomp.hpp b/modules/gpustereo/src/precomp.hpp
new file mode 100644
index 0000000000..934da9fd23
--- /dev/null
+++ b/modules/gpustereo/src/precomp.hpp
@@ -0,0 +1,53 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <limits>
+
+#include "opencv2/gpustereo.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+#include "opencv2/core/utility.hpp"
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpustereo/src/stereobm.cpp b/modules/gpustereo/src/stereobm.cpp
new file mode 100644
index 0000000000..2ef896d574
--- /dev/null
+++ b/modules/gpustereo/src/stereobm.cpp
@@ -0,0 +1,185 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+Ptr<gpu::StereoBM> cv::gpu::createStereoBM(int, int) { throw_no_cuda(); return Ptr<gpu::StereoBM>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace stereobm
+    {
+        void stereoBM_GPU(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& disp, int ndisp, int winsz, const PtrStepSz<unsigned int>& minSSD_buf, cudaStream_t & stream);
+        void prefilter_xsobel(const PtrStepSzb& input, const PtrStepSzb& output, int prefilterCap /*= 31*/, cudaStream_t & stream);
+        void postfilter_textureness(const PtrStepSzb& input, int winsz, float avgTexturenessThreshold, const PtrStepSzb& disp, cudaStream_t & stream);
+    }
+}}}
+
+namespace
+{
+    class StereoBMImpl : public gpu::StereoBM
+    {
+    public:
+        StereoBMImpl(int numDisparities, int blockSize);
+
+        void compute(InputArray left, InputArray right, OutputArray disparity);
+        void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream);
+
+        int getMinDisparity() const { return 0; }
+        void setMinDisparity(int /*minDisparity*/) {}
+
+        int getNumDisparities() const { return ndisp_; }
+        void setNumDisparities(int numDisparities) { ndisp_ = numDisparities; }
+
+        int getBlockSize() const { return winSize_; }
+        void setBlockSize(int blockSize) { winSize_ = blockSize; }
+
+        int getSpeckleWindowSize() const { return 0; }
+        void setSpeckleWindowSize(int /*speckleWindowSize*/) {}
+
+        int getSpeckleRange() const { return 0; }
+        void setSpeckleRange(int /*speckleRange*/) {}
+
+        int getDisp12MaxDiff() const { return 0; }
+        void setDisp12MaxDiff(int /*disp12MaxDiff*/) {}
+
+        int getPreFilterType() const { return preset_; }
+        void setPreFilterType(int preFilterType) { preset_ = preFilterType; }
+
+        int getPreFilterSize() const { return 0; }
+        void setPreFilterSize(int /*preFilterSize*/) {}
+
+        int getPreFilterCap() const { return preFilterCap_; }
+        void setPreFilterCap(int preFilterCap) { preFilterCap_ = preFilterCap; }
+
+        int getTextureThreshold() const { return static_cast<int>(avergeTexThreshold_); }
+        void setTextureThreshold(int textureThreshold) { avergeTexThreshold_ = static_cast<float>(textureThreshold); }
+
+        int getUniquenessRatio() const { return 0; }
+        void setUniquenessRatio(int /*uniquenessRatio*/) {}
+
+        int getSmallerBlockSize() const { return 0; }
+        void setSmallerBlockSize(int /*blockSize*/){}
+
+        Rect getROI1() const { return Rect(); }
+        void setROI1(Rect /*roi1*/) {}
+
+        Rect getROI2() const { return Rect(); }
+        void setROI2(Rect /*roi2*/) {}
+
+    private:
+        int preset_;
+        int ndisp_;
+        int winSize_;
+        int preFilterCap_;
+        float avergeTexThreshold_;
+
+        GpuMat minSSD_, leBuf_, riBuf_;
+    };
+
+    StereoBMImpl::StereoBMImpl(int numDisparities, int blockSize)
+        : preset_(0), ndisp_(numDisparities), winSize_(blockSize), preFilterCap_(31), avergeTexThreshold_(3)
+    {
+    }
+
+    void StereoBMImpl::compute(InputArray left, InputArray right, OutputArray disparity)
+    {
+        compute(left, right, disparity, Stream::Null());
+    }
+
+    void StereoBMImpl::compute(InputArray _left, InputArray _right, OutputArray _disparity, Stream& _stream)
+    {
+        using namespace ::cv::gpu::cudev::stereobm;
+
+        const int max_supported_ndisp = 1 << (sizeof(unsigned char) * 8);
+        CV_Assert( 0 < ndisp_ && ndisp_ <= max_supported_ndisp );
+        CV_Assert( ndisp_ % 8 == 0 );
+        CV_Assert( winSize_ % 2 == 1 );
+
+        GpuMat left = _left.getGpuMat();
+        GpuMat right = _right.getGpuMat();
+
+        CV_Assert( left.type() == CV_8UC1 );
+        CV_Assert( left.size() == right.size() && left.type() == right.type() );
+
+        _disparity.create(left.size(), CV_8UC1);
+        GpuMat disparity = _disparity.getGpuMat();
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        gpu::ensureSizeIsEnough(left.size(), CV_32SC1, minSSD_);
+
+        PtrStepSzb le_for_bm =  left;
+        PtrStepSzb ri_for_bm = right;
+
+        if (preset_ == cv::StereoBM::PREFILTER_XSOBEL)
+        {
+            gpu::ensureSizeIsEnough(left.size(), left.type(), leBuf_);
+            gpu::ensureSizeIsEnough(right.size(), right.type(), riBuf_);
+
+            prefilter_xsobel( left, leBuf_, preFilterCap_, stream);
+            prefilter_xsobel(right, riBuf_, preFilterCap_, stream);
+
+            le_for_bm = leBuf_;
+            ri_for_bm = riBuf_;
+        }
+
+        stereoBM_GPU(le_for_bm, ri_for_bm, disparity, ndisp_, winSize_, minSSD_, stream);
+
+        if (avergeTexThreshold_ > 0)
+            postfilter_textureness(le_for_bm, winSize_, avergeTexThreshold_, disparity, stream);
+    }
+}
+
+Ptr<gpu::StereoBM> cv::gpu::createStereoBM(int numDisparities, int blockSize)
+{
+    return makePtr<StereoBMImpl>(numDisparities, blockSize);
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpustereo/src/stereobp.cpp b/modules/gpustereo/src/stereobp.cpp
new file mode 100644
index 0000000000..f1ea9adf41
--- /dev/null
+++ b/modules/gpustereo/src/stereobp.cpp
@@ -0,0 +1,380 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::StereoBeliefPropagation::estimateRecommendedParams(int, int, int&, int&, int&) { throw_no_cuda(); }
+
+Ptr<gpu::StereoBeliefPropagation> cv::gpu::createStereoBeliefPropagation(int, int, int, int) { throw_no_cuda(); return Ptr<gpu::StereoBeliefPropagation>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace stereobp
+    {
+        void load_constants(int ndisp, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump);
+        template<typename T, typename D>
+        void comp_data_gpu(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& data, cudaStream_t stream);
+        template<typename T>
+        void data_step_down_gpu(int dst_cols, int dst_rows, int src_rows, const PtrStepSzb& src, const PtrStepSzb& dst, cudaStream_t stream);
+        template <typename T>
+        void level_up_messages_gpu(int dst_idx, int dst_cols, int dst_rows, int src_rows, PtrStepSzb* mus, PtrStepSzb* mds, PtrStepSzb* mls, PtrStepSzb* mrs, cudaStream_t stream);
+        template <typename T>
+        void calc_all_iterations_gpu(int cols, int rows, int iters, const PtrStepSzb& u, const PtrStepSzb& d,
+            const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, cudaStream_t stream);
+        template <typename T>
+        void output_gpu(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data,
+            const PtrStepSz<short>& disp, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    class StereoBPImpl : public gpu::StereoBeliefPropagation
+    {
+    public:
+        StereoBPImpl(int ndisp, int iters, int levels, int msg_type);
+
+        void compute(InputArray left, InputArray right, OutputArray disparity);
+        void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream);
+        void compute(InputArray data, OutputArray disparity, Stream& stream);
+
+        int getMinDisparity() const { return 0; }
+        void setMinDisparity(int /*minDisparity*/) {}
+
+        int getNumDisparities() const { return ndisp_; }
+        void setNumDisparities(int numDisparities) { ndisp_ = numDisparities; }
+
+        int getBlockSize() const { return 0; }
+        void setBlockSize(int /*blockSize*/) {}
+
+        int getSpeckleWindowSize() const { return 0; }
+        void setSpeckleWindowSize(int /*speckleWindowSize*/) {}
+
+        int getSpeckleRange() const { return 0; }
+        void setSpeckleRange(int /*speckleRange*/) {}
+
+        int getDisp12MaxDiff() const { return 0; }
+        void setDisp12MaxDiff(int /*disp12MaxDiff*/) {}
+
+        int getNumIters() const { return iters_; }
+        void setNumIters(int iters) { iters_ = iters; }
+
+        int getNumLevels() const { return levels_; }
+        void setNumLevels(int levels) { levels_ = levels; }
+
+        double getMaxDataTerm() const { return max_data_term_; }
+        void setMaxDataTerm(double max_data_term) { max_data_term_ = (float) max_data_term; }
+
+        double getDataWeight() const { return data_weight_; }
+        void setDataWeight(double data_weight) { data_weight_ = (float) data_weight; }
+
+        double getMaxDiscTerm() const { return max_disc_term_; }
+        void setMaxDiscTerm(double max_disc_term) { max_disc_term_ = (float) max_disc_term; }
+
+        double getDiscSingleJump() const { return disc_single_jump_; }
+        void setDiscSingleJump(double disc_single_jump) { disc_single_jump_ = (float) disc_single_jump; }
+
+        int getMsgType() const { return msg_type_; }
+        void setMsgType(int msg_type) { msg_type_ = msg_type; }
+
+    private:
+        void init(Stream& stream);
+        void calcBP(OutputArray disp, Stream& stream);
+
+        int ndisp_;
+        int iters_;
+        int levels_;
+        float max_data_term_;
+        float data_weight_;
+        float max_disc_term_;
+        float disc_single_jump_;
+        int msg_type_;
+
+        float scale_;
+        int rows_, cols_;
+        std::vector<int> cols_all_, rows_all_;
+        GpuMat u_, d_, l_, r_, u2_, d2_, l2_, r2_;
+        std::vector<GpuMat> datas_;
+        GpuMat outBuf_;
+    };
+
+    const float DEFAULT_MAX_DATA_TERM = 10.0f;
+    const float DEFAULT_DATA_WEIGHT = 0.07f;
+    const float DEFAULT_MAX_DISC_TERM = 1.7f;
+    const float DEFAULT_DISC_SINGLE_JUMP = 1.0f;
+
+    StereoBPImpl::StereoBPImpl(int ndisp, int iters, int levels, int msg_type) :
+        ndisp_(ndisp), iters_(iters), levels_(levels),
+        max_data_term_(DEFAULT_MAX_DATA_TERM), data_weight_(DEFAULT_DATA_WEIGHT),
+        max_disc_term_(DEFAULT_MAX_DISC_TERM), disc_single_jump_(DEFAULT_DISC_SINGLE_JUMP),
+        msg_type_(msg_type)
+    {
+    }
+
+    void StereoBPImpl::compute(InputArray left, InputArray right, OutputArray disparity)
+    {
+        compute(left, right, disparity, Stream::Null());
+    }
+
+    void StereoBPImpl::compute(InputArray _left, InputArray _right, OutputArray disparity, Stream& stream)
+    {
+        using namespace cv::gpu::cudev::stereobp;
+
+        typedef void (*comp_data_t)(const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& data, cudaStream_t stream);
+        static const comp_data_t comp_data_callers[2][5] =
+        {
+            {0, comp_data_gpu<unsigned char, short>, 0, comp_data_gpu<uchar3, short>, comp_data_gpu<uchar4, short>},
+            {0, comp_data_gpu<unsigned char, float>, 0, comp_data_gpu<uchar3, float>, comp_data_gpu<uchar4, float>}
+        };
+
+        scale_ = msg_type_ == CV_32F ? 1.0f : 10.0f;
+
+        CV_Assert( 0 < ndisp_ && 0 < iters_ && 0 < levels_ );
+        CV_Assert( msg_type_ == CV_32F || msg_type_ == CV_16S );
+        CV_Assert( msg_type_ == CV_32F || (1 << (levels_ - 1)) * scale_ * max_data_term_ < std::numeric_limits<short>::max() );
+
+        GpuMat left = _left.getGpuMat();
+        GpuMat right = _right.getGpuMat();
+
+        CV_Assert( left.type() == CV_8UC1 || left.type() == CV_8UC3 || left.type() == CV_8UC4 );
+        CV_Assert( left.size() == right.size() && left.type() == right.type() );
+
+        rows_ = left.rows;
+        cols_ = left.cols;
+
+        const int divisor = (int) pow(2.f, levels_ - 1.0f);
+        const int lowest_cols = cols_ / divisor;
+        const int lowest_rows = rows_ / divisor;
+        const int min_image_dim_size = 2;
+        CV_Assert( std::min(lowest_cols, lowest_rows) > min_image_dim_size );
+
+        init(stream);
+
+        datas_[0].create(rows_ * ndisp_, cols_, msg_type_);
+
+        comp_data_callers[msg_type_ == CV_32F][left.channels()](left, right, datas_[0], StreamAccessor::getStream(stream));
+
+        calcBP(disparity, stream);
+    }
+
+    void StereoBPImpl::compute(InputArray _data, OutputArray disparity, Stream& stream)
+    {
+        scale_ = msg_type_ == CV_32F ? 1.0f : 10.0f;
+
+        CV_Assert( 0 < ndisp_ && 0 < iters_ && 0 < levels_ );
+        CV_Assert( msg_type_ == CV_32F || msg_type_ == CV_16S );
+        CV_Assert( msg_type_ == CV_32F || (1 << (levels_ - 1)) * scale_ * max_data_term_ < std::numeric_limits<short>::max() );
+
+        GpuMat data = _data.getGpuMat();
+
+        CV_Assert( (data.type() == msg_type_) && (data.rows % ndisp_ == 0) );
+
+        rows_ = data.rows / ndisp_;
+        cols_ = data.cols;
+
+        const int divisor = (int) pow(2.f, levels_ - 1.0f);
+        const int lowest_cols = cols_ / divisor;
+        const int lowest_rows = rows_ / divisor;
+        const int min_image_dim_size = 2;
+        CV_Assert( std::min(lowest_cols, lowest_rows) > min_image_dim_size );
+
+        init(stream);
+
+        data.copyTo(datas_[0], stream);
+
+        calcBP(disparity, stream);
+    }
+
+    void StereoBPImpl::init(Stream& stream)
+    {
+        using namespace cv::gpu::cudev::stereobp;
+
+        u_.create(rows_ * ndisp_, cols_, msg_type_);
+        d_.create(rows_ * ndisp_, cols_, msg_type_);
+        l_.create(rows_ * ndisp_, cols_, msg_type_);
+        r_.create(rows_ * ndisp_, cols_, msg_type_);
+
+        if (levels_ & 1)
+        {
+            //can clear less area
+            u_.setTo(0, stream);
+            d_.setTo(0, stream);
+            l_.setTo(0, stream);
+            r_.setTo(0, stream);
+        }
+
+        if (levels_ > 1)
+        {
+            int less_rows = (rows_ + 1) / 2;
+            int less_cols = (cols_ + 1) / 2;
+
+            u2_.create(less_rows * ndisp_, less_cols, msg_type_);
+            d2_.create(less_rows * ndisp_, less_cols, msg_type_);
+            l2_.create(less_rows * ndisp_, less_cols, msg_type_);
+            r2_.create(less_rows * ndisp_, less_cols, msg_type_);
+
+            if ((levels_ & 1) == 0)
+            {
+                u2_.setTo(0, stream);
+                d2_.setTo(0, stream);
+                l2_.setTo(0, stream);
+                r2_.setTo(0, stream);
+            }
+        }
+
+        load_constants(ndisp_, max_data_term_, scale_ * data_weight_, scale_ * max_disc_term_, scale_ * disc_single_jump_);
+
+        datas_.resize(levels_);
+
+        cols_all_.resize(levels_);
+        rows_all_.resize(levels_);
+
+        cols_all_[0] = cols_;
+        rows_all_[0] = rows_;
+    }
+
+    void StereoBPImpl::calcBP(OutputArray disp, Stream& _stream)
+    {
+        using namespace cv::gpu::cudev::stereobp;
+
+        typedef void (*data_step_down_t)(int dst_cols, int dst_rows, int src_rows, const PtrStepSzb& src, const PtrStepSzb& dst, cudaStream_t stream);
+        static const data_step_down_t data_step_down_callers[2] =
+        {
+            data_step_down_gpu<short>, data_step_down_gpu<float>
+        };
+
+        typedef void (*level_up_messages_t)(int dst_idx, int dst_cols, int dst_rows, int src_rows, PtrStepSzb* mus, PtrStepSzb* mds, PtrStepSzb* mls, PtrStepSzb* mrs, cudaStream_t stream);
+        static const level_up_messages_t level_up_messages_callers[2] =
+        {
+            level_up_messages_gpu<short>, level_up_messages_gpu<float>
+        };
+
+        typedef void (*calc_all_iterations_t)(int cols, int rows, int iters, const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, cudaStream_t stream);
+        static const calc_all_iterations_t calc_all_iterations_callers[2] =
+        {
+            calc_all_iterations_gpu<short>, calc_all_iterations_gpu<float>
+        };
+
+        typedef void (*output_t)(const PtrStepSzb& u, const PtrStepSzb& d, const PtrStepSzb& l, const PtrStepSzb& r, const PtrStepSzb& data, const PtrStepSz<short>& disp, cudaStream_t stream);
+        static const output_t output_callers[2] =
+        {
+            output_gpu<short>, output_gpu<float>
+        };
+
+        const int funcIdx = msg_type_ == CV_32F;
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        for (int i = 1; i < levels_; ++i)
+        {
+            cols_all_[i] = (cols_all_[i-1] + 1) / 2;
+            rows_all_[i] = (rows_all_[i-1] + 1) / 2;
+
+            datas_[i].create(rows_all_[i] * ndisp_, cols_all_[i], msg_type_);
+
+            data_step_down_callers[funcIdx](cols_all_[i], rows_all_[i], rows_all_[i-1], datas_[i-1], datas_[i], stream);
+        }
+
+        PtrStepSzb mus[] = {u_, u2_};
+        PtrStepSzb mds[] = {d_, d2_};
+        PtrStepSzb mrs[] = {r_, r2_};
+        PtrStepSzb mls[] = {l_, l2_};
+
+        int mem_idx = (levels_ & 1) ? 0 : 1;
+
+        for (int i = levels_ - 1; i >= 0; --i)
+        {
+            // for lower level we have already computed messages by setting to zero
+            if (i != levels_ - 1)
+                level_up_messages_callers[funcIdx](mem_idx, cols_all_[i], rows_all_[i], rows_all_[i+1], mus, mds, mls, mrs, stream);
+
+            calc_all_iterations_callers[funcIdx](cols_all_[i], rows_all_[i], iters_, mus[mem_idx], mds[mem_idx], mls[mem_idx], mrs[mem_idx], datas_[i], stream);
+
+            mem_idx = (mem_idx + 1) & 1;
+        }
+
+        const int dtype = disp.fixedType() ? disp.type() : CV_16SC1;
+
+        disp.create(rows_, cols_, dtype);
+        GpuMat out = disp.getGpuMat();
+
+        if (dtype != CV_16SC1)
+        {
+            outBuf_.create(rows_, cols_, CV_16SC1);
+            out = outBuf_;
+        }
+
+        out.setTo(0, _stream);
+
+        output_callers[funcIdx](u_, d_, l_, r_, datas_.front(), out, stream);
+
+        if (dtype != CV_16SC1)
+            out.convertTo(disp, dtype, _stream);
+    }
+}
+
+Ptr<gpu::StereoBeliefPropagation> cv::gpu::createStereoBeliefPropagation(int ndisp, int iters, int levels, int msg_type)
+{
+    return makePtr<StereoBPImpl>(ndisp, iters, levels, msg_type);
+}
+
+void cv::gpu::StereoBeliefPropagation::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels)
+{
+    ndisp = width / 4;
+    if ((ndisp & 1) != 0)
+        ndisp++;
+
+    int mm = std::max(width, height);
+    iters = mm / 100 + 2;
+
+    levels = (int)(::log(static_cast<double>(mm)) + 1) * 4 / 5;
+    if (levels == 0) levels++;
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpustereo/src/stereocsbp.cpp b/modules/gpustereo/src/stereocsbp.cpp
new file mode 100644
index 0000000000..b1135b2765
--- /dev/null
+++ b/modules/gpustereo/src/stereocsbp.cpp
@@ -0,0 +1,387 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+
+void cv::gpu::StereoConstantSpaceBP::estimateRecommendedParams(int, int, int&, int&, int&, int&) { throw_no_cuda(); }
+
+Ptr<gpu::StereoConstantSpaceBP> cv::gpu::createStereoConstantSpaceBP(int, int, int, int, int) { throw_no_cuda(); return Ptr<gpu::StereoConstantSpaceBP>(); }
+
+#else /* !defined (HAVE_CUDA) */
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace stereocsbp
+    {
+        void load_constants(int ndisp, float max_data_term, float data_weight, float max_disc_term, float disc_single_jump, int min_disp_th,
+            const PtrStepSzb& left, const PtrStepSzb& right, const PtrStepSzb& temp);
+
+        template<class T>
+        void init_data_cost(int rows, int cols, T* disp_selected_pyr, T* data_cost_selected, size_t msg_step,
+                    int h, int w, int level, int nr_plane, int ndisp, int channels, bool use_local_init_data_cost, cudaStream_t stream);
+
+        template<class T>
+        void compute_data_cost(const T* disp_selected_pyr, T* data_cost, size_t msg_step,
+                               int rows, int cols, int h, int w, int h2, int level, int nr_plane, int channels, cudaStream_t stream);
+
+        template<class T>
+        void init_message(T* u_new, T* d_new, T* l_new, T* r_new,
+                          const T* u_cur, const T* d_cur, const T* l_cur, const T* r_cur,
+                          T* selected_disp_pyr_new, const T* selected_disp_pyr_cur,
+                          T* data_cost_selected, const T* data_cost, size_t msg_step,
+                          int h, int w, int nr_plane, int h2, int w2, int nr_plane2, cudaStream_t stream);
+
+        template<class T>
+        void calc_all_iterations(T* u, T* d, T* l, T* r, const T* data_cost_selected,
+            const T* selected_disp_pyr_cur, size_t msg_step, int h, int w, int nr_plane, int iters, cudaStream_t stream);
+
+        template<class T>
+        void compute_disp(const T* u, const T* d, const T* l, const T* r, const T* data_cost_selected, const T* disp_selected, size_t msg_step,
+            const PtrStepSz<short>& disp, int nr_plane, cudaStream_t stream);
+    }
+}}}
+
+namespace
+{
+    class StereoCSBPImpl : public gpu::StereoConstantSpaceBP
+    {
+    public:
+        StereoCSBPImpl(int ndisp, int iters, int levels, int nr_plane, int msg_type);
+
+        void compute(InputArray left, InputArray right, OutputArray disparity);
+        void compute(InputArray left, InputArray right, OutputArray disparity, Stream& stream);
+        void compute(InputArray data, OutputArray disparity, Stream& stream);
+
+        int getMinDisparity() const { return min_disp_th_; }
+        void setMinDisparity(int minDisparity) { min_disp_th_ = minDisparity; }
+
+        int getNumDisparities() const { return ndisp_; }
+        void setNumDisparities(int numDisparities) { ndisp_ = numDisparities; }
+
+        int getBlockSize() const { return 0; }
+        void setBlockSize(int /*blockSize*/) {}
+
+        int getSpeckleWindowSize() const { return 0; }
+        void setSpeckleWindowSize(int /*speckleWindowSize*/) {}
+
+        int getSpeckleRange() const { return 0; }
+        void setSpeckleRange(int /*speckleRange*/) {}
+
+        int getDisp12MaxDiff() const { return 0; }
+        void setDisp12MaxDiff(int /*disp12MaxDiff*/) {}
+
+        int getNumIters() const { return iters_; }
+        void setNumIters(int iters) { iters_ = iters; }
+
+        int getNumLevels() const { return levels_; }
+        void setNumLevels(int levels) { levels_ = levels; }
+
+        double getMaxDataTerm() const { return max_data_term_; }
+        void setMaxDataTerm(double max_data_term) { max_data_term_ = (float) max_data_term; }
+
+        double getDataWeight() const { return data_weight_; }
+        void setDataWeight(double data_weight) { data_weight_ = (float) data_weight; }
+
+        double getMaxDiscTerm() const { return max_disc_term_; }
+        void setMaxDiscTerm(double max_disc_term) { max_disc_term_ = (float) max_disc_term; }
+
+        double getDiscSingleJump() const { return disc_single_jump_; }
+        void setDiscSingleJump(double disc_single_jump) { disc_single_jump_ = (float) disc_single_jump; }
+
+        int getMsgType() const { return msg_type_; }
+        void setMsgType(int msg_type) { msg_type_ = msg_type; }
+
+        int getNrPlane() const { return nr_plane_; }
+        void setNrPlane(int nr_plane) { nr_plane_ = nr_plane; }
+
+        bool getUseLocalInitDataCost() const { return use_local_init_data_cost_; }
+        void setUseLocalInitDataCost(bool use_local_init_data_cost) { use_local_init_data_cost_ = use_local_init_data_cost; }
+
+    private:
+        int min_disp_th_;
+        int ndisp_;
+        int iters_;
+        int levels_;
+        float max_data_term_;
+        float data_weight_;
+        float max_disc_term_;
+        float disc_single_jump_;
+        int msg_type_;
+        int nr_plane_;
+        bool use_local_init_data_cost_;
+
+        GpuMat mbuf_;
+        GpuMat temp_;
+        GpuMat outBuf_;
+    };
+
+    const float DEFAULT_MAX_DATA_TERM = 30.0f;
+    const float DEFAULT_DATA_WEIGHT = 1.0f;
+    const float DEFAULT_MAX_DISC_TERM = 160.0f;
+    const float DEFAULT_DISC_SINGLE_JUMP = 10.0f;
+
+    StereoCSBPImpl::StereoCSBPImpl(int ndisp, int iters, int levels, int nr_plane, int msg_type) :
+        min_disp_th_(0), ndisp_(ndisp), iters_(iters), levels_(levels),
+        max_data_term_(DEFAULT_MAX_DATA_TERM), data_weight_(DEFAULT_DATA_WEIGHT),
+        max_disc_term_(DEFAULT_MAX_DISC_TERM), disc_single_jump_(DEFAULT_DISC_SINGLE_JUMP),
+        msg_type_(msg_type), nr_plane_(nr_plane), use_local_init_data_cost_(true)
+    {
+    }
+
+    void StereoCSBPImpl::compute(InputArray left, InputArray right, OutputArray disparity)
+    {
+        compute(left, right, disparity, Stream::Null());
+    }
+
+    void StereoCSBPImpl::compute(InputArray _left, InputArray _right, OutputArray disp, Stream& _stream)
+    {
+        using namespace cv::gpu::cudev::stereocsbp;
+
+        CV_Assert( msg_type_ == CV_32F || msg_type_ == CV_16S );
+        CV_Assert( 0 < ndisp_ && 0 < iters_ && 0 < levels_ && 0 < nr_plane_ && levels_ <= 8 );
+
+        GpuMat left = _left.getGpuMat();
+        GpuMat right = _right.getGpuMat();
+
+        CV_Assert( left.type() == CV_8UC1 || left.type() == CV_8UC3 || left.type() == CV_8UC4 );
+        CV_Assert( left.size() == right.size() && left.type() == right.type() );
+
+        cudaStream_t stream = StreamAccessor::getStream(_stream);
+
+        ////////////////////////////////////////////////////////////////////////////////////////////
+        // Init
+
+        int rows = left.rows;
+        int cols = left.cols;
+
+        levels_ = std::min(levels_, int(log((double)ndisp_) / log(2.0)));
+
+        // compute sizes
+        AutoBuffer<int> buf(levels_ * 3);
+        int* cols_pyr = buf;
+        int* rows_pyr = cols_pyr + levels_;
+        int* nr_plane_pyr = rows_pyr + levels_;
+
+        cols_pyr[0]     = cols;
+        rows_pyr[0]     = rows;
+        nr_plane_pyr[0] = nr_plane_;
+
+        for (int i = 1; i < levels_; i++)
+        {
+            cols_pyr[i]     = cols_pyr[i-1] / 2;
+            rows_pyr[i]     = rows_pyr[i-1] / 2;
+            nr_plane_pyr[i] = nr_plane_pyr[i-1] * 2;
+        }
+
+        GpuMat u[2], d[2], l[2], r[2], disp_selected_pyr[2], data_cost, data_cost_selected;
+
+        //allocate buffers
+        int buffers_count = 10; // (up + down + left + right + disp_selected_pyr) * 2
+        buffers_count += 2; //  data_cost has twice more rows than other buffers, what's why +2, not +1;
+        buffers_count += 1; //  data_cost_selected
+        mbuf_.create(rows * nr_plane_ * buffers_count, cols, msg_type_);
+
+        data_cost          = mbuf_.rowRange(0, rows * nr_plane_ * 2);
+        data_cost_selected = mbuf_.rowRange(data_cost.rows, data_cost.rows + rows * nr_plane_);
+
+        for(int k = 0; k < 2; ++k) // in/out
+        {
+            GpuMat sub1 = mbuf_.rowRange(data_cost.rows + data_cost_selected.rows, mbuf_.rows);
+            GpuMat sub2 = sub1.rowRange((k+0)*sub1.rows/2, (k+1)*sub1.rows/2);
+
+            GpuMat *buf_ptrs[] = { &u[k], &d[k], &l[k], &r[k], &disp_selected_pyr[k] };
+            for(int _r = 0; _r < 5; ++_r)
+            {
+                *buf_ptrs[_r] = sub2.rowRange(_r * sub2.rows/5, (_r+1) * sub2.rows/5);
+                CV_DbgAssert( buf_ptrs[_r]->cols == cols && buf_ptrs[_r]->rows == rows * nr_plane_ );
+            }
+        };
+
+        size_t elem_step = mbuf_.step / mbuf_.elemSize();
+
+        Size temp_size = data_cost.size();
+        if ((size_t)temp_size.area() < elem_step * rows_pyr[levels_ - 1] * ndisp_)
+            temp_size = Size(static_cast<int>(elem_step), rows_pyr[levels_ - 1] * ndisp_);
+
+        temp_.create(temp_size, msg_type_);
+
+        ////////////////////////////////////////////////////////////////////////////
+        // Compute
+
+        load_constants(ndisp_, max_data_term_, data_weight_, max_disc_term_, disc_single_jump_, min_disp_th_, left, right, temp_);
+
+        l[0].setTo(0, _stream);
+        d[0].setTo(0, _stream);
+        r[0].setTo(0, _stream);
+        u[0].setTo(0, _stream);
+
+        l[1].setTo(0, _stream);
+        d[1].setTo(0, _stream);
+        r[1].setTo(0, _stream);
+        u[1].setTo(0, _stream);
+
+        data_cost.setTo(0, _stream);
+        data_cost_selected.setTo(0, _stream);
+
+        int cur_idx = 0;
+
+        if (msg_type_ == CV_32F)
+        {
+            for (int i = levels_ - 1; i >= 0; i--)
+            {
+                if (i == levels_ - 1)
+                {
+                    init_data_cost(left.rows, left.cols, disp_selected_pyr[cur_idx].ptr<float>(), data_cost_selected.ptr<float>(),
+                        elem_step, rows_pyr[i], cols_pyr[i], i, nr_plane_pyr[i], ndisp_, left.channels(), use_local_init_data_cost_, stream);
+                }
+                else
+                {
+                    compute_data_cost(disp_selected_pyr[cur_idx].ptr<float>(), data_cost.ptr<float>(), elem_step,
+                        left.rows, left.cols, rows_pyr[i], cols_pyr[i], rows_pyr[i+1], i, nr_plane_pyr[i+1], left.channels(), stream);
+
+                    int new_idx = (cur_idx + 1) & 1;
+
+                    init_message(u[new_idx].ptr<float>(), d[new_idx].ptr<float>(), l[new_idx].ptr<float>(), r[new_idx].ptr<float>(),
+                                 u[cur_idx].ptr<float>(), d[cur_idx].ptr<float>(), l[cur_idx].ptr<float>(), r[cur_idx].ptr<float>(),
+                                 disp_selected_pyr[new_idx].ptr<float>(), disp_selected_pyr[cur_idx].ptr<float>(),
+                                 data_cost_selected.ptr<float>(), data_cost.ptr<float>(), elem_step, rows_pyr[i],
+                                 cols_pyr[i], nr_plane_pyr[i], rows_pyr[i+1], cols_pyr[i+1], nr_plane_pyr[i+1], stream);
+
+                    cur_idx = new_idx;
+                }
+
+                calc_all_iterations(u[cur_idx].ptr<float>(), d[cur_idx].ptr<float>(), l[cur_idx].ptr<float>(), r[cur_idx].ptr<float>(),
+                                    data_cost_selected.ptr<float>(), disp_selected_pyr[cur_idx].ptr<float>(), elem_step,
+                                    rows_pyr[i], cols_pyr[i], nr_plane_pyr[i], iters_, stream);
+            }
+        }
+        else
+        {
+            for (int i = levels_ - 1; i >= 0; i--)
+            {
+                if (i == levels_ - 1)
+                {
+                    init_data_cost(left.rows, left.cols, disp_selected_pyr[cur_idx].ptr<short>(), data_cost_selected.ptr<short>(),
+                        elem_step, rows_pyr[i], cols_pyr[i], i, nr_plane_pyr[i], ndisp_, left.channels(), use_local_init_data_cost_, stream);
+                }
+                else
+                {
+                    compute_data_cost(disp_selected_pyr[cur_idx].ptr<short>(), data_cost.ptr<short>(), elem_step,
+                        left.rows, left.cols, rows_pyr[i], cols_pyr[i], rows_pyr[i+1], i, nr_plane_pyr[i+1], left.channels(), stream);
+
+                    int new_idx = (cur_idx + 1) & 1;
+
+                    init_message(u[new_idx].ptr<short>(), d[new_idx].ptr<short>(), l[new_idx].ptr<short>(), r[new_idx].ptr<short>(),
+                                 u[cur_idx].ptr<short>(), d[cur_idx].ptr<short>(), l[cur_idx].ptr<short>(), r[cur_idx].ptr<short>(),
+                                 disp_selected_pyr[new_idx].ptr<short>(), disp_selected_pyr[cur_idx].ptr<short>(),
+                                 data_cost_selected.ptr<short>(), data_cost.ptr<short>(), elem_step, rows_pyr[i],
+                                 cols_pyr[i], nr_plane_pyr[i], rows_pyr[i+1], cols_pyr[i+1], nr_plane_pyr[i+1], stream);
+
+                    cur_idx = new_idx;
+                }
+
+                calc_all_iterations(u[cur_idx].ptr<short>(), d[cur_idx].ptr<short>(), l[cur_idx].ptr<short>(), r[cur_idx].ptr<short>(),
+                                    data_cost_selected.ptr<short>(), disp_selected_pyr[cur_idx].ptr<short>(), elem_step,
+                                    rows_pyr[i], cols_pyr[i], nr_plane_pyr[i], iters_, stream);
+            }
+        }
+
+        const int dtype = disp.fixedType() ? disp.type() : CV_16SC1;
+
+        disp.create(rows, cols, dtype);
+        GpuMat out = disp.getGpuMat();
+
+        if (dtype != CV_16SC1)
+        {
+            outBuf_.create(rows, cols, CV_16SC1);
+            out = outBuf_;
+        }
+
+        out.setTo(0, _stream);
+
+        if (msg_type_ == CV_32F)
+        {
+            compute_disp(u[cur_idx].ptr<float>(), d[cur_idx].ptr<float>(), l[cur_idx].ptr<float>(), r[cur_idx].ptr<float>(),
+                         data_cost_selected.ptr<float>(), disp_selected_pyr[cur_idx].ptr<float>(), elem_step, out, nr_plane_pyr[0], stream);
+        }
+        else
+        {
+            compute_disp(u[cur_idx].ptr<short>(), d[cur_idx].ptr<short>(), l[cur_idx].ptr<short>(), r[cur_idx].ptr<short>(),
+                         data_cost_selected.ptr<short>(), disp_selected_pyr[cur_idx].ptr<short>(), elem_step, out, nr_plane_pyr[0], stream);
+        }
+
+        if (dtype != CV_16SC1)
+            out.convertTo(disp, dtype, _stream);
+    }
+
+    void StereoCSBPImpl::compute(InputArray /*data*/, OutputArray /*disparity*/, Stream& /*stream*/)
+    {
+        CV_Error(Error::StsNotImplemented, "Not implemented");
+    }
+}
+
+Ptr<gpu::StereoConstantSpaceBP> cv::gpu::createStereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane, int msg_type)
+{
+    return makePtr<StereoCSBPImpl>(ndisp, iters, levels, nr_plane, msg_type);
+}
+
+void cv::gpu::StereoConstantSpaceBP::estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane)
+{
+    ndisp = (int) ((float) width / 3.14f);
+    if ((ndisp & 1) != 0)
+        ndisp++;
+
+    int mm = std::max(width, height);
+    iters = mm / 100 + ((mm > 1200)? - 4 : 4);
+
+    levels = (int)::log(static_cast<double>(mm)) * 2 / 3;
+    if (levels == 0) levels++;
+
+    nr_plane = (int) ((float) ndisp / std::pow(2.0, levels + 1));
+}
+
+#endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpustereo/src/util.cpp b/modules/gpustereo/src/util.cpp
new file mode 100644
index 0000000000..e58b5a18e0
--- /dev/null
+++ b/modules/gpustereo/src/util.cpp
@@ -0,0 +1,123 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+void cv::gpu::reprojectImageTo3D(InputArray, OutputArray, InputArray, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::drawColorDisp(InputArray, OutputArray, int, Stream&) { throw_no_cuda(); }
+
+#else
+
+////////////////////////////////////////////////////////////////////////
+// reprojectImageTo3D
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T, typename D>
+    void reprojectImageTo3D_gpu(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+}}}
+
+void cv::gpu::reprojectImageTo3D(InputArray _disp, OutputArray _xyz, InputArray _Q, int dst_cn, Stream& stream)
+{
+    using namespace cv::gpu::cudev;
+
+    typedef void (*func_t)(const PtrStepSzb disp, PtrStepSzb xyz, const float* q, cudaStream_t stream);
+    static const func_t funcs[2][4] =
+    {
+        {reprojectImageTo3D_gpu<uchar, float3>, 0, 0, reprojectImageTo3D_gpu<short, float3>},
+        {reprojectImageTo3D_gpu<uchar, float4>, 0, 0, reprojectImageTo3D_gpu<short, float4>}
+    };
+
+    GpuMat disp = _disp.getGpuMat();
+    Mat Q = _Q.getMat();
+
+    CV_Assert( disp.type() == CV_8U || disp.type() == CV_16S );
+    CV_Assert( Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4 && Q.isContinuous() );
+    CV_Assert( dst_cn == 3 || dst_cn == 4 );
+
+    _xyz.create(disp.size(), CV_MAKE_TYPE(CV_32F, dst_cn));
+    GpuMat xyz = _xyz.getGpuMat();
+
+    funcs[dst_cn == 4][disp.type()](disp, xyz, Q.ptr<float>(), StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// drawColorDisp
+
+namespace cv { namespace gpu { namespace cudev
+{
+    void drawColorDisp_gpu(const PtrStepSzb& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream);
+    void drawColorDisp_gpu(const PtrStepSz<short>& src, const PtrStepSzb& dst, int ndisp, const cudaStream_t& stream);
+}}}
+
+namespace
+{
+    template <typename T>
+    void drawColorDisp_caller(const GpuMat& src, OutputArray _dst, int ndisp, const cudaStream_t& stream)
+    {
+        using namespace ::cv::gpu::cudev;
+
+        _dst.create(src.size(), CV_8UC4);
+        GpuMat dst = _dst.getGpuMat();
+
+        drawColorDisp_gpu((PtrStepSz<T>)src, dst, ndisp, stream);
+    }
+}
+
+void cv::gpu::drawColorDisp(InputArray _src, OutputArray dst, int ndisp, Stream& stream)
+{
+    typedef void (*drawColorDisp_caller_t)(const GpuMat& src, OutputArray dst, int ndisp, const cudaStream_t& stream);
+    const drawColorDisp_caller_t drawColorDisp_callers[] = {drawColorDisp_caller<unsigned char>, 0, 0, drawColorDisp_caller<short>, 0, 0, 0, 0};
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8U || src.type() == CV_16S );
+
+    drawColorDisp_callers[src.type()](src, dst, ndisp, StreamAccessor::getStream(stream));
+}
+
+#endif
diff --git a/modules/gpustereo/test/test_main.cpp b/modules/gpustereo/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpustereo/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpustereo/test/test_precomp.hpp b/modules/gpustereo/test/test_precomp.hpp
new file mode 100644
index 0000000000..eb34fcb568
--- /dev/null
+++ b/modules/gpustereo/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpustereo.hpp"
+#include "opencv2/calib3d.hpp"
+
+#include "cvconfig.h"
+
+#endif
diff --git a/modules/gpustereo/test/test_stereo.cpp b/modules/gpustereo/test/test_stereo.cpp
new file mode 100644
index 0000000000..9a3d94627e
--- /dev/null
+++ b/modules/gpustereo/test/test_stereo.cpp
@@ -0,0 +1,212 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+//////////////////////////////////////////////////////////////////////////
+// StereoBM
+
+struct StereoBM : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(StereoBM, Regression)
+{
+    cv::Mat left_image  = readImage("stereobm/aloe-L.png", cv::IMREAD_GRAYSCALE);
+    cv::Mat right_image = readImage("stereobm/aloe-R.png", cv::IMREAD_GRAYSCALE);
+    cv::Mat disp_gold   = readImage("stereobm/aloe-disp.png", cv::IMREAD_GRAYSCALE);
+
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+
+    cv::Ptr<cv::StereoBM> bm = cv::gpu::createStereoBM(128, 19);
+    cv::gpu::GpuMat disp;
+
+    bm->compute(loadMat(left_image), loadMat(right_image), disp);
+
+    EXPECT_MAT_NEAR(disp_gold, disp, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Stereo, StereoBM, ALL_DEVICES);
+
+//////////////////////////////////////////////////////////////////////////
+// StereoBeliefPropagation
+
+struct StereoBeliefPropagation : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(StereoBeliefPropagation, Regression)
+{
+    cv::Mat left_image  = readImage("stereobp/aloe-L.png");
+    cv::Mat right_image = readImage("stereobp/aloe-R.png");
+    cv::Mat disp_gold   = readImage("stereobp/aloe-disp.png", cv::IMREAD_GRAYSCALE);
+
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+
+    cv::Ptr<cv::gpu::StereoBeliefPropagation> bp = cv::gpu::createStereoBeliefPropagation(64, 8, 2, CV_16S);
+    bp->setMaxDataTerm(25.0);
+    bp->setDataWeight(0.1);
+    bp->setMaxDiscTerm(15.0);
+    bp->setDiscSingleJump(1.0);
+
+    cv::gpu::GpuMat disp;
+
+    bp->compute(loadMat(left_image), loadMat(right_image), disp);
+
+    cv::Mat h_disp(disp);
+    h_disp.convertTo(h_disp, disp_gold.depth());
+
+    EXPECT_MAT_NEAR(disp_gold, h_disp, 0.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Stereo, StereoBeliefPropagation, ALL_DEVICES);
+
+//////////////////////////////////////////////////////////////////////////
+// StereoConstantSpaceBP
+
+struct StereoConstantSpaceBP : testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(StereoConstantSpaceBP, Regression)
+{
+    cv::Mat left_image  = readImage("csstereobp/aloe-L.png");
+    cv::Mat right_image = readImage("csstereobp/aloe-R.png");
+
+    cv::Mat disp_gold;
+
+    if (supportFeature(devInfo, cv::gpu::FEATURE_SET_COMPUTE_20))
+        disp_gold = readImage("csstereobp/aloe-disp.png", cv::IMREAD_GRAYSCALE);
+    else
+        disp_gold = readImage("csstereobp/aloe-disp_CC1X.png", cv::IMREAD_GRAYSCALE);
+
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+
+    cv::Ptr<cv::gpu::StereoConstantSpaceBP> csbp = cv::gpu::createStereoConstantSpaceBP(128, 16, 4, 4);
+    cv::gpu::GpuMat disp;
+
+    csbp->compute(loadMat(left_image), loadMat(right_image), disp);
+
+    cv::Mat h_disp(disp);
+    h_disp.convertTo(h_disp, disp_gold.depth());
+
+    EXPECT_MAT_NEAR(disp_gold, h_disp, 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Stereo, StereoConstantSpaceBP, ALL_DEVICES);
+
+////////////////////////////////////////////////////////////////////////////////
+// reprojectImageTo3D
+
+PARAM_TEST_CASE(ReprojectImageTo3D, cv::gpu::DeviceInfo, cv::Size, MatDepth, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int depth;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        depth = GET_PARAM(2);
+        useRoi = GET_PARAM(3);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+GPU_TEST_P(ReprojectImageTo3D, Accuracy)
+{
+    cv::Mat disp = randomMat(size, depth, 5.0, 30.0);
+    cv::Mat Q = randomMat(cv::Size(4, 4), CV_32FC1, 0.1, 1.0);
+
+    cv::gpu::GpuMat dst;
+    cv::gpu::reprojectImageTo3D(loadMat(disp, useRoi), dst, Q, 3);
+
+    cv::Mat dst_gold;
+    cv::reprojectImageTo3D(disp, dst_gold, Q, false);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-5);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Stereo, ReprojectImageTo3D, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    testing::Values(MatDepth(CV_8U), MatDepth(CV_16S)),
+    WHOLE_SUBMAT));
+
+#endif // HAVE_CUDA
diff --git a/modules/gpuwarping/CMakeLists.txt b/modules/gpuwarping/CMakeLists.txt
new file mode 100644
index 0000000000..0c4ca7b12e
--- /dev/null
+++ b/modules/gpuwarping/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(gpuwarping)
+endif()
+
+set(the_description "GPU-accelerated Image Warping")
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 /wd4324 /wd4512 -Wundef -Wmissing-declarations)
+
+ocv_define_module(gpuwarping opencv_imgproc OPTIONAL opencv_gpulegacy)
diff --git a/modules/gpuwarping/doc/gpuwarping.rst b/modules/gpuwarping/doc/gpuwarping.rst
new file mode 100644
index 0000000000..4bdaa8d4eb
--- /dev/null
+++ b/modules/gpuwarping/doc/gpuwarping.rst
@@ -0,0 +1,8 @@
+*****************************************
+gpuwarping. GPU-accelerated Image Warping
+*****************************************
+
+.. toctree::
+    :maxdepth: 1
+
+    warping
diff --git a/modules/gpuwarping/doc/warping.rst b/modules/gpuwarping/doc/warping.rst
new file mode 100644
index 0000000000..5c88eb772b
--- /dev/null
+++ b/modules/gpuwarping/doc/warping.rst
@@ -0,0 +1,251 @@
+Image Warping
+=============
+
+.. highlight:: cpp
+
+
+
+gpu::remap
+--------------
+Applies a generic geometrical transformation to an image.
+
+.. ocv:function:: void gpu::remap(InputArray src, OutputArray dst, InputArray xmap, InputArray ymap, int interpolation, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
+
+    :param src: Source image.
+
+    :param dst: Destination image with the size the same as  ``xmap`` and the type the same as  ``src`` .
+
+    :param xmap: X values. Only  ``CV_32FC1`` type is supported.
+
+    :param ymap: Y values. Only  ``CV_32FC1`` type is supported.
+
+    :param interpolation: Interpolation method (see  :ocv:func:`resize` ). ``INTER_NEAREST`` , ``INTER_LINEAR`` and ``INTER_CUBIC`` are supported for now.
+
+    :param borderMode: Pixel extrapolation method (see  :ocv:func:`borderInterpolate` ). ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param borderValue: Value used in case of a constant border. By default, it is 0.
+
+    :param stream: Stream for the asynchronous version.
+
+The function transforms the source image using the specified map:
+
+.. math::
+
+    \texttt{dst} (x,y) =  \texttt{src} (xmap(x,y), ymap(x,y))
+
+Values of pixels with non-integer coordinates are computed using the bilinear interpolation.
+
+.. seealso:: :ocv:func:`remap`
+
+
+
+gpu::resize
+---------------
+Resizes an image.
+
+.. ocv:function:: void gpu::resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
+
+    :param src: Source image.
+
+    :param dst: Destination image  with the same type as  ``src`` . The size is ``dsize`` (when it is non-zero) or the size is computed from  ``src.size()`` , ``fx`` , and  ``fy`` .
+
+    :param dsize: Destination image size. If it is zero, it is computed as:
+
+        .. math::
+            \texttt{dsize = Size(round(fx*src.cols), round(fy*src.rows))}
+
+        Either  ``dsize`` or both  ``fx`` and  ``fy`` must be non-zero.
+
+    :param fx: Scale factor along the horizontal axis. If it is zero, it is computed as:
+
+        .. math::
+
+            \texttt{(double)dsize.width/src.cols}
+
+    :param fy: Scale factor along the vertical axis. If it is zero, it is computed as:
+
+        .. math::
+
+            \texttt{(double)dsize.height/src.rows}
+
+    :param interpolation: Interpolation method. ``INTER_NEAREST`` , ``INTER_LINEAR`` and ``INTER_CUBIC`` are supported for now.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`resize`
+
+
+
+gpu::warpAffine
+-------------------
+Applies an affine transformation to an image.
+
+.. ocv:function:: void gpu::warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
+
+    :param src: Source image.  ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
+
+    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
+
+    :param M: *2x3*  transformation matrix.
+
+    :param dsize: Size of the destination image.
+
+    :param flags: Combination of interpolation methods (see  :ocv:func:`resize`) and the optional flag  ``WARP_INVERSE_MAP`` specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ). Only ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` interpolation methods are supported.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`warpAffine`
+
+
+
+gpu::buildWarpAffineMaps
+------------------------
+Builds transformation maps for affine transformation.
+
+.. ocv:function:: void gpu::buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null())
+
+    :param M: *2x3*  transformation matrix.
+
+    :param inverse: Flag  specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ).
+
+    :param dsize: Size of the destination image.
+
+    :param xmap: X values with  ``CV_32FC1`` type.
+
+    :param ymap: Y values with  ``CV_32FC1`` type.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`gpu::warpAffine` , :ocv:func:`gpu::remap`
+
+
+
+gpu::warpPerspective
+------------------------
+Applies a perspective transformation to an image.
+
+.. ocv:function:: void gpu::warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
+
+    :param src: Source image. ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
+
+    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
+
+    :param M: *3x3* transformation matrix.
+
+    :param dsize: Size of the destination image.
+
+    :param flags: Combination of interpolation methods (see  :ocv:func:`resize` ) and the optional flag  ``WARP_INVERSE_MAP`` specifying that  ``M`` is the inverse transformation ( ``dst => src`` ). Only  ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` interpolation methods are supported.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`warpPerspective`
+
+
+
+gpu::buildWarpPerspectiveMaps
+-----------------------------
+Builds transformation maps for perspective transformation.
+
+.. ocv:function:: void gpu::buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null())
+
+    :param M: *3x3*  transformation matrix.
+
+    :param inverse: Flag  specifying that  ``M`` is an inverse transformation ( ``dst=>src`` ).
+
+    :param dsize: Size of the destination image.
+
+    :param xmap: X values with  ``CV_32FC1`` type.
+
+    :param ymap: Y values with  ``CV_32FC1`` type.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`gpu::warpPerspective` , :ocv:func:`gpu::remap`
+
+
+
+gpu::buildWarpPlaneMaps
+-----------------------
+Builds plane warping maps.
+
+.. ocv:function:: void gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, InputArray T, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::buildWarpCylindricalMaps
+-----------------------------
+Builds cylindrical warping maps.
+
+.. ocv:function:: void gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::buildWarpSphericalMaps
+---------------------------
+Builds spherical warping maps.
+
+.. ocv:function:: void gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::rotate
+---------------
+Rotates an image around the origin (0,0) and then shifts it.
+
+.. ocv:function:: void gpu::rotate(InputArray src, OutputArray dst, Size dsize, double angle, double xShift = 0, double yShift = 0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
+
+    :param src: Source image. Supports 1, 3 or 4 channels images with ``CV_8U`` , ``CV_16U`` or ``CV_32F`` depth.
+
+    :param dst: Destination image with the same type as  ``src`` . The size is  ``dsize`` .
+
+    :param dsize: Size of the destination image.
+
+    :param angle: Angle of rotation in degrees.
+
+    :param xShift: Shift along the horizontal axis.
+
+    :param yShift: Shift along the vertical axis.
+
+    :param interpolation: Interpolation method. Only  ``INTER_NEAREST`` , ``INTER_LINEAR`` , and  ``INTER_CUBIC`` are supported.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`gpu::warpAffine`
+
+
+
+gpu::pyrDown
+-------------------
+Smoothes an image and downsamples it.
+
+.. ocv:function:: void gpu::pyrDown(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source image.
+
+    :param dst: Destination image. Will have ``Size((src.cols+1)/2, (src.rows+1)/2)`` size and the same type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`pyrDown`
+
+
+
+gpu::pyrUp
+-------------------
+Upsamples an image and then smoothes it.
+
+.. ocv:function:: void gpu::pyrUp(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
+
+    :param src: Source image.
+
+    :param dst: Destination image. Will have ``Size(src.cols*2, src.rows*2)`` size and the same type as ``src`` .
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso:: :ocv:func:`pyrUp`
diff --git a/modules/gpuwarping/include/opencv2/gpuwarping.hpp b/modules/gpuwarping/include/opencv2/gpuwarping.hpp
new file mode 100644
index 0000000000..ed17464ec4
--- /dev/null
+++ b/modules/gpuwarping/include/opencv2/gpuwarping.hpp
@@ -0,0 +1,113 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPUWARPING_HPP__
+#define __OPENCV_GPUWARPING_HPP__
+
+#ifndef __cplusplus
+#  error gpuwarping.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/imgproc.hpp"
+
+namespace cv { namespace gpu {
+
+//! DST[x,y] = SRC[xmap[x,y],ymap[x,y]]
+//! supports only CV_32FC1 map type
+CV_EXPORTS void remap(InputArray src, OutputArray dst, InputArray xmap, InputArray ymap,
+                      int interpolation, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(),
+                      Stream& stream = Stream::Null());
+
+//! resizes the image
+//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_AREA
+CV_EXPORTS void resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
+
+//! warps the image using affine transformation
+//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
+CV_EXPORTS void warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
+
+CV_EXPORTS void buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+
+//! warps the image using perspective transformation
+//! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
+CV_EXPORTS void warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
+
+CV_EXPORTS void buildWarpPerspectiveMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+
+//! builds plane warping maps
+CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, InputArray T, float scale,
+                                   OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
+
+//! builds cylindrical warping maps
+CV_EXPORTS void buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale,
+                                         OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
+
+//! builds spherical warping maps
+CV_EXPORTS void buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale,
+                                       OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
+
+//! rotates an image around the origin (0,0) and then shifts it
+//! supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
+//! supports 1, 3 or 4 channels images with CV_8U, CV_16U or CV_32F depth
+CV_EXPORTS void rotate(InputArray src, OutputArray dst, Size dsize, double angle, double xShift = 0, double yShift = 0,
+                       int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
+
+//! smoothes the source image and downsamples it
+CV_EXPORTS void pyrDown(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+//! upsamples the source image and then smoothes it
+CV_EXPORTS void pyrUp(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
+
+class CV_EXPORTS ImagePyramid : public Algorithm
+{
+public:
+    virtual void getLayer(OutputArray outImg, Size outRoi, Stream& stream = Stream::Null()) const = 0;
+};
+
+CV_EXPORTS Ptr<ImagePyramid> createImagePyramid(InputArray img, int nLayers = -1, Stream& stream = Stream::Null());
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_GPUWARPING_HPP__ */
diff --git a/modules/gpuwarping/perf/perf_main.cpp b/modules/gpuwarping/perf/perf_main.cpp
new file mode 100644
index 0000000000..8566b097d4
--- /dev/null
+++ b/modules/gpuwarping/perf/perf_main.cpp
@@ -0,0 +1,47 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+CV_PERF_TEST_CUDA_MAIN(gpuwarping)
diff --git a/modules/gpuwarping/perf/perf_precomp.hpp b/modules/gpuwarping/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..fc46b26474
--- /dev/null
+++ b/modules/gpuwarping/perf/perf_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/gpuwarping.hpp"
+#include "opencv2/imgproc.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/gpuwarping/perf/perf_warping.cpp b/modules/gpuwarping/perf/perf_warping.cpp
new file mode 100644
index 0000000000..266475bb0c
--- /dev/null
+++ b/modules/gpuwarping/perf/perf_warping.cpp
@@ -0,0 +1,553 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// Remap
+
+enum { HALF_SIZE=0, UPSIDE_DOWN, REFLECTION_X, REFLECTION_BOTH };
+CV_ENUM(RemapMode, HALF_SIZE, UPSIDE_DOWN, REFLECTION_X, REFLECTION_BOTH)
+
+void generateMap(cv::Mat& map_x, cv::Mat& map_y, int remapMode)
+{
+    for (int j = 0; j < map_x.rows; ++j)
+    {
+        for (int i = 0; i < map_x.cols; ++i)
+        {
+            switch (remapMode)
+            {
+            case HALF_SIZE:
+                if (i > map_x.cols*0.25 && i < map_x.cols*0.75 && j > map_x.rows*0.25 && j < map_x.rows*0.75)
+                {
+                    map_x.at<float>(j,i) = 2.f * (i - map_x.cols * 0.25f) + 0.5f;
+                    map_y.at<float>(j,i) = 2.f * (j - map_x.rows * 0.25f) + 0.5f;
+                }
+                else
+                {
+                    map_x.at<float>(j,i) = 0.f;
+                    map_y.at<float>(j,i) = 0.f;
+                }
+                break;
+            case UPSIDE_DOWN:
+                map_x.at<float>(j,i) = static_cast<float>(i);
+                map_y.at<float>(j,i) = static_cast<float>(map_x.rows - j);
+                break;
+            case REFLECTION_X:
+                map_x.at<float>(j,i) = static_cast<float>(map_x.cols - i);
+                map_y.at<float>(j,i) = static_cast<float>(j);
+                break;
+            case REFLECTION_BOTH:
+                map_x.at<float>(j,i) = static_cast<float>(map_x.cols - i);
+                map_y.at<float>(j,i) = static_cast<float>(map_x.rows - j);
+                break;
+            } // end of switch
+        }
+    }
+}
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border_Mode, cv::Size, MatDepth, MatCn, Interpolation, BorderMode, RemapMode);
+
+PERF_TEST_P(Sz_Depth_Cn_Inter_Border_Mode, Remap,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
+                    ALL_BORDER_MODES,
+                    RemapMode::all()))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = GET_PARAM(3);
+    const int borderMode = GET_PARAM(4);
+    const int remapMode = GET_PARAM(5);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    cv::Mat xmap(size, CV_32FC1);
+    cv::Mat ymap(size, CV_32FC1);
+    generateMap(xmap, ymap, remapMode);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        const cv::gpu::GpuMat d_xmap(xmap);
+        const cv::gpu::GpuMat d_ymap(ymap);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::remap(d_src, dst, d_xmap, d_ymap, interpolation, borderMode);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::remap(src, dst, xmap, ymap, interpolation, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Resize
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Scale, cv::Size, MatDepth, MatCn, Interpolation, double);
+
+PERF_TEST_P(Sz_Depth_Cn_Inter_Scale, Resize,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
+                    Values(0.5, 0.3, 2.0)))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = GET_PARAM(3);
+    const double f = GET_PARAM(4);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::resize(d_src, dst, cv::Size(), f, f, interpolation);
+
+        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::resize(src, dst, cv::Size(), f, f, interpolation);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// ResizeArea
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Scale, cv::Size, MatDepth, MatCn, double);
+
+PERF_TEST_P(Sz_Depth_Cn_Scale, ResizeArea,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(0.2, 0.1, 0.05)))
+{
+    declare.time(1.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = cv::INTER_AREA;
+    const double f = GET_PARAM(3);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::resize(d_src, dst, cv::Size(), f, f, interpolation);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::resize(src, dst, cv::Size(), f, f, interpolation);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// WarpAffine
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border, cv::Size, MatDepth, MatCn, Interpolation, BorderMode);
+
+PERF_TEST_P(Sz_Depth_Cn_Inter_Border, WarpAffine,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
+                    ALL_BORDER_MODES))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = GET_PARAM(3);
+    const int borderMode = GET_PARAM(4);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const double aplha = CV_PI / 4;
+    const double mat[2 * 3] =
+    {
+        std::cos(aplha), -std::sin(aplha), src.cols / 2,
+        std::sin(aplha),  std::cos(aplha), 0
+    };
+    const cv::Mat M(2, 3, CV_64F, (void*) mat);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::warpAffine(d_src, dst, M, size, interpolation, borderMode);
+
+        GPU_SANITY_CHECK(dst, 1);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::warpAffine(src, dst, M, size, interpolation, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// WarpPerspective
+
+PERF_TEST_P(Sz_Depth_Cn_Inter_Border, WarpPerspective,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
+                    ALL_BORDER_MODES))
+{
+    declare.time(20.0);
+
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = GET_PARAM(3);
+    const int borderMode = GET_PARAM(4);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const double aplha = CV_PI / 4;
+    double mat[3][3] = { {std::cos(aplha), -std::sin(aplha), src.cols / 2},
+                         {std::sin(aplha),  std::cos(aplha), 0},
+                         {0.0,              0.0,             1.0}};
+    const cv::Mat M(3, 3, CV_64F, (void*) mat);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::warpPerspective(d_src, dst, M, size, interpolation, borderMode);
+
+        GPU_SANITY_CHECK(dst, 1);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::warpPerspective(src, dst, M, size, interpolation, borderMode);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// BuildWarpPlaneMaps
+
+PERF_TEST_P(Sz, BuildWarpPlaneMaps,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
+    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
+    const cv::Mat T = cv::Mat::zeros(1, 3, CV_32F);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat map_x;
+        cv::gpu::GpuMat map_y;
+
+        TEST_CYCLE() cv::gpu::buildWarpPlaneMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, T, 1.0, map_x, map_y);
+
+        GPU_SANITY_CHECK(map_x);
+        GPU_SANITY_CHECK(map_y);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// BuildWarpCylindricalMaps
+
+PERF_TEST_P(Sz, BuildWarpCylindricalMaps,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
+    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat map_x;
+        cv::gpu::GpuMat map_y;
+
+        TEST_CYCLE() cv::gpu::buildWarpCylindricalMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, 1.0, map_x, map_y);
+
+        GPU_SANITY_CHECK(map_x);
+        GPU_SANITY_CHECK(map_y);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// BuildWarpSphericalMaps
+
+PERF_TEST_P(Sz, BuildWarpSphericalMaps,
+            GPU_TYPICAL_MAT_SIZES)
+{
+    const cv::Size size = GetParam();
+
+    const cv::Mat K = cv::Mat::eye(3, 3, CV_32FC1);
+    const cv::Mat R = cv::Mat::ones(3, 3, CV_32FC1);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat map_x;
+        cv::gpu::GpuMat map_y;
+
+        TEST_CYCLE() cv::gpu::buildWarpSphericalMaps(size, cv::Rect(0, 0, size.width, size.height), K, R, 1.0, map_x, map_y);
+
+        GPU_SANITY_CHECK(map_x);
+        GPU_SANITY_CHECK(map_y);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// Rotate
+
+DEF_PARAM_TEST(Sz_Depth_Cn_Inter, cv::Size, MatDepth, MatCn, Interpolation);
+
+PERF_TEST_P(Sz_Depth_Cn_Inter, Rotate,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4,
+                    Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC))))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+    const int interpolation = GET_PARAM(3);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::rotate(d_src, dst, size, 30.0, 0, 0, interpolation);
+
+        GPU_SANITY_CHECK(dst, 1e-3, ERROR_RELATIVE);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// PyrDown
+
+PERF_TEST_P(Sz_Depth_Cn, PyrDown,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::pyrDown(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::pyrDown(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// PyrUp
+
+PERF_TEST_P(Sz_Depth_Cn, PyrUp,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        TEST_CYCLE() cv::gpu::pyrUp(d_src, dst);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::pyrUp(src, dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
+
+//////////////////////////////////////////////////////////////////////
+// ImagePyramidGetLayer
+
+PERF_TEST_P(Sz_Depth_Cn, ImagePyramidGetLayer,
+            Combine(GPU_TYPICAL_MAT_SIZES,
+                    Values(CV_8U, CV_16U, CV_32F),
+                    GPU_CHANNELS_1_3_4))
+{
+    const cv::Size size = GET_PARAM(0);
+    const int depth = GET_PARAM(1);
+    const int channels = GET_PARAM(2);
+
+    const int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    declare.in(src, WARMUP_RNG);
+
+    const int nLayers = 3;
+    const cv::Size dstSize(size.width / 2 + 10, size.height / 2 + 10);
+
+    if (PERF_RUN_GPU())
+    {
+        const cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat dst;
+
+        cv::Ptr<cv::gpu::ImagePyramid> d_pyr = cv::gpu::createImagePyramid(d_src, nLayers);
+
+        TEST_CYCLE() d_pyr->getLayer(dst, dstSize);
+
+        GPU_SANITY_CHECK(dst);
+    }
+    else
+    {
+        FAIL_NO_CPU();
+    }
+}
diff --git a/modules/gpuwarping/src/cuda/build_warp_maps.cu b/modules/gpuwarping/src/cuda/build_warp_maps.cu
new file mode 100644
index 0000000000..6bd4e335bb
--- /dev/null
+++ b/modules/gpuwarping/src/cuda/build_warp_maps.cu
@@ -0,0 +1,221 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        // TODO use intrinsics like __sinf and so on
+
+        namespace build_warp_maps
+        {
+
+            __constant__ float ck_rinv[9];
+            __constant__ float cr_kinv[9];
+            __constant__ float ct[3];
+            __constant__ float cscale;
+        }
+
+
+        class PlaneMapper
+        {
+        public:
+            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
+            {
+                using namespace build_warp_maps;
+
+                float x_ = u / cscale - ct[0];
+                float y_ = v / cscale - ct[1];
+
+                float z;
+                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * (1 - ct[2]);
+                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * (1 - ct[2]);
+                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * (1 - ct[2]);
+
+                x /= z;
+                y /= z;
+            }
+        };
+
+
+        class CylindricalMapper
+        {
+        public:
+            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
+            {
+                using namespace build_warp_maps;
+
+                u /= cscale;
+                float x_ = ::sinf(u);
+                float y_ = v / cscale;
+                float z_ = ::cosf(u);
+
+                float z;
+                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * z_;
+                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * z_;
+                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * z_;
+
+                if (z > 0) { x /= z; y /= z; }
+                else x = y = -1;
+            }
+        };
+
+
+        class SphericalMapper
+        {
+        public:
+            static __device__ __forceinline__ void mapBackward(float u, float v, float &x, float &y)
+            {
+                using namespace build_warp_maps;
+
+                v /= cscale;
+                u /= cscale;
+
+                float sinv = ::sinf(v);
+                float x_ = sinv * ::sinf(u);
+                float y_ = -::cosf(v);
+                float z_ = sinv * ::cosf(u);
+
+                float z;
+                x = ck_rinv[0] * x_ + ck_rinv[1] * y_ + ck_rinv[2] * z_;
+                y = ck_rinv[3] * x_ + ck_rinv[4] * y_ + ck_rinv[5] * z_;
+                z = ck_rinv[6] * x_ + ck_rinv[7] * y_ + ck_rinv[8] * z_;
+
+                if (z > 0) { x /= z; y /= z; }
+                else x = y = -1;
+            }
+        };
+
+
+        template <typename Mapper>
+        __global__ void buildWarpMapsKernel(int tl_u, int tl_v, int cols, int rows,
+                                            PtrStepf map_x, PtrStepf map_y)
+        {
+            int du = blockIdx.x * blockDim.x + threadIdx.x;
+            int dv = blockIdx.y * blockDim.y + threadIdx.y;
+            if (du < cols && dv < rows)
+            {
+                float u = tl_u + du;
+                float v = tl_v + dv;
+                float x, y;
+                Mapper::mapBackward(u, v, x, y);
+                map_x.ptr(dv)[du] = x;
+                map_y.ptr(dv)[du] = y;
+            }
+        }
+
+
+        void buildWarpPlaneMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                const float k_rinv[9], const float r_kinv[9], const float t[3],
+                                float scale, cudaStream_t stream)
+        {
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ct, t, 3*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
+
+            int cols = map_x.cols;
+            int rows = map_x.rows;
+
+            dim3 threads(32, 8);
+            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
+
+            buildWarpMapsKernel<PlaneMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
+            cudaSafeCall(cudaGetLastError());
+            if (stream == 0)
+                cudaSafeCall(cudaDeviceSynchronize());
+        }
+
+
+        void buildWarpCylindricalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                      const float k_rinv[9], const float r_kinv[9], float scale,
+                                      cudaStream_t stream)
+        {
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
+
+            int cols = map_x.cols;
+            int rows = map_x.rows;
+
+            dim3 threads(32, 8);
+            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
+
+            buildWarpMapsKernel<CylindricalMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
+            cudaSafeCall(cudaGetLastError());
+            if (stream == 0)
+                cudaSafeCall(cudaDeviceSynchronize());
+        }
+
+
+        void buildWarpSphericalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                    const float k_rinv[9], const float r_kinv[9], float scale,
+                                    cudaStream_t stream)
+        {
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::ck_rinv, k_rinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cr_kinv, r_kinv, 9*sizeof(float)));
+            cudaSafeCall(cudaMemcpyToSymbol(build_warp_maps::cscale, &scale, sizeof(float)));
+
+            int cols = map_x.cols;
+            int rows = map_x.rows;
+
+            dim3 threads(32, 8);
+            dim3 grid(divUp(cols, threads.x), divUp(rows, threads.y));
+
+            buildWarpMapsKernel<SphericalMapper><<<grid,threads>>>(tl_u, tl_v, cols, rows, map_x, map_y);
+            cudaSafeCall(cudaGetLastError());
+            if (stream == 0)
+                cudaSafeCall(cudaDeviceSynchronize());
+        }
+    } // namespace imgproc
+}}} // namespace cv { namespace gpu { namespace cudev {
+
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/pyr_down.cu b/modules/gpuwarping/src/cuda/pyr_down.cu
similarity index 96%
rename from modules/gpu/src/cuda/pyr_down.cu
rename to modules/gpuwarping/src/cuda/pyr_down.cu
index 35c24e4b8a..904f549bad 100644
--- a/modules/gpu/src/cuda/pyr_down.cu
+++ b/modules/gpuwarping/src/cuda/pyr_down.cu
@@ -42,13 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -222,7 +222,7 @@ namespace cv { namespace gpu { namespace device
         template void pyrDown_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
         template void pyrDown_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
     } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/pyr_up.cu b/modules/gpuwarping/src/cuda/pyr_up.cu
similarity index 96%
rename from modules/gpu/src/cuda/pyr_up.cu
rename to modules/gpuwarping/src/cuda/pyr_up.cu
index 5966b03642..36a72274cf 100644
--- a/modules/gpu/src/cuda/pyr_up.cu
+++ b/modules/gpuwarping/src/cuda/pyr_up.cu
@@ -42,13 +42,13 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -191,6 +191,6 @@ namespace cv { namespace gpu { namespace device
         template void pyrUp_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
         template void pyrUp_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
     } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-#endif /* CUDA_DISABLER */
\ No newline at end of file
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/remap.cu b/modules/gpuwarping/src/cuda/remap.cu
similarity index 97%
rename from modules/gpu/src/cuda/remap.cu
rename to modules/gpuwarping/src/cuda/remap.cu
index f40ada0302..c4ea317fbd 100644
--- a/modules/gpu/src/cuda/remap.cu
+++ b/modules/gpuwarping/src/cuda/remap.cu
@@ -42,14 +42,14 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/filters.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/filters.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -212,25 +212,25 @@ namespace cv { namespace gpu { namespace device
             static const caller_t callers[3][5] =
             {
                 {
-                    RemapDispatcher<PointFilter, BrdReflect101, T>::call,
-                    RemapDispatcher<PointFilter, BrdReplicate, T>::call,
                     RemapDispatcher<PointFilter, BrdConstant, T>::call,
+                    RemapDispatcher<PointFilter, BrdReplicate, T>::call,
                     RemapDispatcher<PointFilter, BrdReflect, T>::call,
-                    RemapDispatcher<PointFilter, BrdWrap, T>::call
+                    RemapDispatcher<PointFilter, BrdWrap, T>::call,
+                    RemapDispatcher<PointFilter, BrdReflect101, T>::call
                 },
                 {
-                    RemapDispatcher<LinearFilter, BrdReflect101, T>::call,
-                    RemapDispatcher<LinearFilter, BrdReplicate, T>::call,
                     RemapDispatcher<LinearFilter, BrdConstant, T>::call,
+                    RemapDispatcher<LinearFilter, BrdReplicate, T>::call,
                     RemapDispatcher<LinearFilter, BrdReflect, T>::call,
-                    RemapDispatcher<LinearFilter, BrdWrap, T>::call
+                    RemapDispatcher<LinearFilter, BrdWrap, T>::call,
+                    RemapDispatcher<LinearFilter, BrdReflect101, T>::call
                 },
                 {
-                    RemapDispatcher<CubicFilter, BrdReflect101, T>::call,
-                    RemapDispatcher<CubicFilter, BrdReplicate, T>::call,
                     RemapDispatcher<CubicFilter, BrdConstant, T>::call,
+                    RemapDispatcher<CubicFilter, BrdReplicate, T>::call,
                     RemapDispatcher<CubicFilter, BrdReflect, T>::call,
-                    RemapDispatcher<CubicFilter, BrdWrap, T>::call
+                    RemapDispatcher<CubicFilter, BrdWrap, T>::call,
+                    RemapDispatcher<CubicFilter, BrdReflect101, T>::call
                 }
             };
 
@@ -268,7 +268,7 @@ namespace cv { namespace gpu { namespace device
         template void remap_gpu<float3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
         template void remap_gpu<float4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
     } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpuwarping/src/cuda/resize.cu b/modules/gpuwarping/src/cuda/resize.cu
new file mode 100644
index 0000000000..94422db9dc
--- /dev/null
+++ b/modules/gpuwarping/src/cuda/resize.cu
@@ -0,0 +1,482 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if !defined CUDA_DISABLER
+
+#include <cfloat>
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/filters.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    // kernels
+
+    template <typename T> __global__ void resize_nearest(const PtrStep<T> src, PtrStepSz<T> dst, const float fy, const float fx)
+    {
+        const int dst_x = blockDim.x * blockIdx.x + threadIdx.x;
+        const int dst_y = blockDim.y * blockIdx.y + threadIdx.y;
+
+        if (dst_x < dst.cols && dst_y < dst.rows)
+        {
+            const float src_x = dst_x * fx;
+            const float src_y = dst_y * fy;
+
+            dst(dst_y, dst_x) = src(__float2int_rz(src_y), __float2int_rz(src_x));
+        }
+    }
+
+    template <typename T> __global__ void resize_linear(const PtrStepSz<T> src, PtrStepSz<T> dst, const float fy, const float fx)
+    {
+        typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type work_type;
+
+        const int dst_x = blockDim.x * blockIdx.x + threadIdx.x;
+        const int dst_y = blockDim.y * blockIdx.y + threadIdx.y;
+
+        if (dst_x < dst.cols && dst_y < dst.rows)
+        {
+            const float src_x = dst_x * fx;
+            const float src_y = dst_y * fy;
+
+            work_type out = VecTraits<work_type>::all(0);
+
+            const int x1 = __float2int_rd(src_x);
+            const int y1 = __float2int_rd(src_y);
+            const int x2 = x1 + 1;
+            const int y2 = y1 + 1;
+            const int x2_read = ::min(x2, src.cols - 1);
+            const int y2_read = ::min(y2, src.rows - 1);
+
+            T src_reg = src(y1, x1);
+            out = out + src_reg * ((x2 - src_x) * (y2 - src_y));
+
+            src_reg = src(y1, x2_read);
+            out = out + src_reg * ((src_x - x1) * (y2 - src_y));
+
+            src_reg = src(y2_read, x1);
+            out = out + src_reg * ((x2 - src_x) * (src_y - y1));
+
+            src_reg = src(y2_read, x2_read);
+            out = out + src_reg * ((src_x - x1) * (src_y - y1));
+
+            dst(dst_y, dst_x) = saturate_cast<T>(out);
+        }
+    }
+
+    template <class Ptr2D, typename T> __global__ void resize(const Ptr2D src, PtrStepSz<T> dst, const float fy, const float fx)
+    {
+        const int dst_x = blockDim.x * blockIdx.x + threadIdx.x;
+        const int dst_y = blockDim.y * blockIdx.y + threadIdx.y;
+
+        if (dst_x < dst.cols && dst_y < dst.rows)
+        {
+            const float src_x = dst_x * fx;
+            const float src_y = dst_y * fy;
+
+            dst(dst_y, dst_x) = src(src_y, src_x);
+        }
+    }
+
+    template <typename Ptr2D, typename T> __global__ void resize_area(const Ptr2D src, PtrStepSz<T> dst)
+    {
+        const int x = blockDim.x * blockIdx.x + threadIdx.x;
+        const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+        if (x < dst.cols && y < dst.rows)
+        {
+            dst(y, x) = src(y, x);
+        }
+    }
+
+    // textures
+
+    template <typename T> struct TextureAccessor;
+
+    #define OPENCV_GPU_IMPLEMENT_RESIZE_TEX(type) \
+        texture<type, cudaTextureType2D, cudaReadModeElementType> tex_resize_##type (0, cudaFilterModePoint, cudaAddressModeClamp); \
+        template <> struct TextureAccessor<type> \
+        { \
+            typedef type elem_type; \
+            typedef int index_type; \
+            int xoff; \
+            int yoff; \
+            __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const \
+            { \
+                return tex2D(tex_resize_##type, x + xoff, y + yoff); \
+            } \
+            __host__ static void bind(const PtrStepSz<type>& mat) \
+            { \
+                bindTexture(&tex_resize_##type, mat); \
+            } \
+        };
+
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(uchar)
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(uchar4)
+
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(ushort)
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(ushort4)
+
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(short)
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(short4)
+
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(float)
+    OPENCV_GPU_IMPLEMENT_RESIZE_TEX(float4)
+
+    #undef OPENCV_GPU_IMPLEMENT_RESIZE_TEX
+
+    template <typename T>
+    TextureAccessor<T> texAccessor(const PtrStepSz<T>& mat, int yoff, int xoff)
+    {
+        TextureAccessor<T>::bind(mat);
+
+        TextureAccessor<T> t;
+        t.xoff = xoff;
+        t.yoff = yoff;
+
+        return t;
+    }
+
+    // callers for nearest interpolation
+
+    template <typename T>
+    void call_resize_nearest_glob(const PtrStepSz<T>& src, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        resize_nearest<<<grid, block, 0, stream>>>(src, dst, fy, fx);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <typename T>
+    void call_resize_nearest_tex(const PtrStepSz<T>& /*src*/, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        resize<<<grid, block>>>(texAccessor(srcWhole, yoff, xoff), dst, fy, fx);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    // callers for linear interpolation
+
+    template <typename T>
+    void call_resize_linear_glob(const PtrStepSz<T>& src, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        resize_linear<<<grid, block>>>(src, dst, fy, fx);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <typename T>
+    void call_resize_linear_tex(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        if (srcWhole.data == src.data)
+        {
+            TextureAccessor<T> texSrc = texAccessor(src, 0, 0);
+            LinearFilter< TextureAccessor<T> > filteredSrc(texSrc);
+
+            resize<<<grid, block>>>(filteredSrc, dst, fy, fx);
+        }
+        else
+        {
+            TextureAccessor<T> texSrc = texAccessor(srcWhole, yoff, xoff);
+
+            BrdReplicate<T> brd(src.rows, src.cols);
+            BorderReader<TextureAccessor<T>, BrdReplicate<T> > brdSrc(texSrc, brd);
+            LinearFilter< BorderReader<TextureAccessor<T>, BrdReplicate<T> > > filteredSrc(brdSrc);
+
+            resize<<<grid, block>>>(filteredSrc, dst, fy, fx);
+        }
+
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    // callers for cubic interpolation
+
+    template <typename T>
+    void call_resize_cubic_glob(const PtrStepSz<T>& src, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        BrdReplicate<T> brd(src.rows, src.cols);
+        BorderReader< PtrStep<T>, BrdReplicate<T> > brdSrc(src, brd);
+        CubicFilter< BorderReader< PtrStep<T>, BrdReplicate<T> > > filteredSrc(brdSrc);
+
+        resize<<<grid, block, 0, stream>>>(filteredSrc, dst, fy, fx);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <typename T>
+    void call_resize_cubic_tex(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+        if (srcWhole.data == src.data)
+        {
+            TextureAccessor<T> texSrc = texAccessor(src, 0, 0);
+            CubicFilter< TextureAccessor<T> > filteredSrc(texSrc);
+
+            resize<<<grid, block>>>(filteredSrc, dst, fy, fx);
+        }
+        else
+        {
+            TextureAccessor<T> texSrc = texAccessor(srcWhole, yoff, xoff);
+
+            BrdReplicate<T> brd(src.rows, src.cols);
+            BorderReader<TextureAccessor<T>, BrdReplicate<T> > brdSrc(texSrc, brd);
+            CubicFilter< BorderReader<TextureAccessor<T>, BrdReplicate<T> > > filteredSrc(brdSrc);
+
+            resize<<<grid, block>>>(filteredSrc, dst, fy, fx);
+        }
+
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    // ResizeNearestDispatcher
+
+    template <typename T> struct ResizeNearestDispatcher
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& /*srcWhole*/, int /*yoff*/, int /*xoff*/, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            call_resize_nearest_glob(src, dst, fy, fx, stream);
+        }
+    };
+
+    template <typename T> struct SelectImplForNearest
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            if (stream)
+                call_resize_nearest_glob(src, dst, fy, fx, stream);
+            else
+            {
+                if (fx > 1 || fy > 1)
+                    call_resize_nearest_glob(src, dst, fy, fx, 0);
+                else
+                    call_resize_nearest_tex(src, srcWhole, yoff, xoff, dst, fy, fx);
+            }
+        }
+    };
+
+    template <> struct ResizeNearestDispatcher<uchar> : SelectImplForNearest<uchar> {};
+    template <> struct ResizeNearestDispatcher<uchar4> : SelectImplForNearest<uchar4> {};
+
+    template <> struct ResizeNearestDispatcher<ushort> : SelectImplForNearest<ushort> {};
+    template <> struct ResizeNearestDispatcher<ushort4> : SelectImplForNearest<ushort4> {};
+
+    template <> struct ResizeNearestDispatcher<short> : SelectImplForNearest<short> {};
+    template <> struct ResizeNearestDispatcher<short4> : SelectImplForNearest<short4> {};
+
+    template <> struct ResizeNearestDispatcher<float> : SelectImplForNearest<float> {};
+    template <> struct ResizeNearestDispatcher<float4> : SelectImplForNearest<float4> {};
+
+    // ResizeLinearDispatcher
+
+    template <typename T> struct ResizeLinearDispatcher
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            call_resize_linear_glob(src, dst, fy, fx, stream);
+        }
+    };
+
+    template <typename T> struct SelectImplForLinear
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            if (stream)
+                call_resize_linear_glob(src, dst, fy, fx, stream);
+            else
+            {
+                if (fx > 1 || fy > 1)
+                    call_resize_linear_glob(src, dst, fy, fx, 0);
+                else
+                    call_resize_linear_tex(src, srcWhole, yoff, xoff, dst, fy, fx);
+            }
+        }
+    };
+
+    template <> struct ResizeLinearDispatcher<uchar> : SelectImplForLinear<uchar> {};
+    template <> struct ResizeLinearDispatcher<uchar4> : SelectImplForLinear<uchar4> {};
+
+    template <> struct ResizeLinearDispatcher<ushort> : SelectImplForLinear<ushort> {};
+    template <> struct ResizeLinearDispatcher<ushort4> : SelectImplForLinear<ushort4> {};
+
+    template <> struct ResizeLinearDispatcher<short> : SelectImplForLinear<short> {};
+    template <> struct ResizeLinearDispatcher<short4> : SelectImplForLinear<short4> {};
+
+    template <> struct ResizeLinearDispatcher<float> : SelectImplForLinear<float> {};
+    template <> struct ResizeLinearDispatcher<float4> : SelectImplForLinear<float4> {};
+
+    // ResizeCubicDispatcher
+
+    template <typename T> struct ResizeCubicDispatcher
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            call_resize_cubic_glob(src, dst, fy, fx, stream);
+        }
+    };
+
+    template <typename T> struct SelectImplForCubic
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            if (stream)
+                call_resize_cubic_glob(src, dst, fy, fx, stream);
+            else
+                call_resize_cubic_tex(src, srcWhole, yoff, xoff, dst, fy, fx);
+        }
+    };
+
+    template <> struct ResizeCubicDispatcher<uchar> : SelectImplForCubic<uchar> {};
+    template <> struct ResizeCubicDispatcher<uchar4> : SelectImplForCubic<uchar4> {};
+
+    template <> struct ResizeCubicDispatcher<ushort> : SelectImplForCubic<ushort> {};
+    template <> struct ResizeCubicDispatcher<ushort4> : SelectImplForCubic<ushort4> {};
+
+    template <> struct ResizeCubicDispatcher<short> : SelectImplForCubic<short> {};
+    template <> struct ResizeCubicDispatcher<short4> : SelectImplForCubic<short4> {};
+
+    template <> struct ResizeCubicDispatcher<float> : SelectImplForCubic<float> {};
+    template <> struct ResizeCubicDispatcher<float4> : SelectImplForCubic<float4> {};
+
+    // ResizeAreaDispatcher
+
+    template <typename T> struct ResizeAreaDispatcher
+    {
+        static void call(const PtrStepSz<T>& src, const PtrStepSz<T>&, int, int, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream)
+        {
+            const int iscale_x = (int) round(fx);
+            const int iscale_y = (int) round(fy);
+
+            const dim3 block(32, 8);
+            const dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y));
+
+            if (std::abs(fx - iscale_x) < FLT_MIN && std::abs(fy - iscale_y) < FLT_MIN)
+            {
+                BrdConstant<T> brd(src.rows, src.cols);
+                BorderReader< PtrStep<T>, BrdConstant<T> > brdSrc(src, brd);
+                IntegerAreaFilter< BorderReader< PtrStep<T>, BrdConstant<T> > > filteredSrc(brdSrc, fx, fy);
+
+                resize_area<<<grid, block, 0, stream>>>(filteredSrc, dst);
+            }
+            else
+            {
+                BrdConstant<T> brd(src.rows, src.cols);
+                BorderReader< PtrStep<T>, BrdConstant<T> > brdSrc(src, brd);
+                AreaFilter< BorderReader< PtrStep<T>, BrdConstant<T> > > filteredSrc(brdSrc, fx, fy);
+
+                resize_area<<<grid, block, 0, stream>>>(filteredSrc, dst);
+            }
+
+            cudaSafeCall( cudaGetLastError() );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+
+    // resize
+
+    template <typename T> void resize(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream)
+    {
+        typedef void (*func_t)(const PtrStepSz<T>& src, const PtrStepSz<T>& srcWhole, int yoff, int xoff, const PtrStepSz<T>& dst, float fy, float fx, cudaStream_t stream);
+        static const func_t funcs[4] =
+        {
+            ResizeNearestDispatcher<T>::call,
+            ResizeLinearDispatcher<T>::call,
+            ResizeCubicDispatcher<T>::call,
+            ResizeAreaDispatcher<T>::call
+        };
+
+        // change to linear if area interpolation upscaling
+        if (interpolation == 3 && (fx <= 1.f || fy <= 1.f))
+            interpolation = 1;
+
+        funcs[interpolation](static_cast< PtrStepSz<T> >(src), static_cast< PtrStepSz<T> >(srcWhole), yoff, xoff, static_cast< PtrStepSz<T> >(dst), fy, fx, stream);
+    }
+
+    template void resize<uchar >(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<uchar3>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<uchar4>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+
+    template void resize<ushort >(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<ushort3>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<ushort4>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+
+    template void resize<short >(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<short3>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<short4>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+
+    template void resize<float >(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<float3>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    template void resize<float4>(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+}}}
+
+#endif /* CUDA_DISABLER */
diff --git a/modules/gpu/src/cuda/warp.cu b/modules/gpuwarping/src/cuda/warp.cu
similarity index 98%
rename from modules/gpu/src/cuda/warp.cu
rename to modules/gpuwarping/src/cuda/warp.cu
index 49130d9405..83db79ebf1 100644
--- a/modules/gpu/src/cuda/warp.cu
+++ b/modules/gpuwarping/src/cuda/warp.cu
@@ -42,14 +42,14 @@
 
 #if !defined CUDA_DISABLER
 
-#include "internal_shared.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/filters.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/filters.hpp"
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -281,25 +281,25 @@ namespace cv { namespace gpu { namespace device
             static const func_t funcs[3][5] =
             {
                 {
-                    WarpDispatcher<Transform, PointFilter, BrdReflect101, T>::call,
-                    WarpDispatcher<Transform, PointFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, PointFilter, BrdConstant, T>::call,
+                    WarpDispatcher<Transform, PointFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, PointFilter, BrdReflect, T>::call,
-                    WarpDispatcher<Transform, PointFilter, BrdWrap, T>::call
+                    WarpDispatcher<Transform, PointFilter, BrdWrap, T>::call,
+                    WarpDispatcher<Transform, PointFilter, BrdReflect101, T>::call
                 },
                 {
-                    WarpDispatcher<Transform, LinearFilter, BrdReflect101, T>::call,
-                    WarpDispatcher<Transform, LinearFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, LinearFilter, BrdConstant, T>::call,
+                    WarpDispatcher<Transform, LinearFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, LinearFilter, BrdReflect, T>::call,
-                    WarpDispatcher<Transform, LinearFilter, BrdWrap, T>::call
+                    WarpDispatcher<Transform, LinearFilter, BrdWrap, T>::call,
+                    WarpDispatcher<Transform, LinearFilter, BrdReflect101, T>::call
                 },
                 {
-                    WarpDispatcher<Transform, CubicFilter, BrdReflect101, T>::call,
-                    WarpDispatcher<Transform, CubicFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, CubicFilter, BrdConstant, T>::call,
+                    WarpDispatcher<Transform, CubicFilter, BrdReplicate, T>::call,
                     WarpDispatcher<Transform, CubicFilter, BrdReflect, T>::call,
-                    WarpDispatcher<Transform, CubicFilter, BrdWrap, T>::call
+                    WarpDispatcher<Transform, CubicFilter, BrdWrap, T>::call,
+                    WarpDispatcher<Transform, CubicFilter, BrdReflect101, T>::call
                 }
             };
 
@@ -383,7 +383,7 @@ namespace cv { namespace gpu { namespace device
         template void warpPerspective_gpu<float3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float coeffs[3 * 3], PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
         template void warpPerspective_gpu<float4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float coeffs[3 * 3], PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
     } // namespace imgproc
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
 
 #endif /* CUDA_DISABLER */
diff --git a/modules/gpuwarping/src/precomp.hpp b/modules/gpuwarping/src/precomp.hpp
new file mode 100644
index 0000000000..e19853712f
--- /dev/null
+++ b/modules/gpuwarping/src/precomp.hpp
@@ -0,0 +1,57 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include "opencv2/gpuwarping.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPULEGACY
+#  include "opencv2/gpulegacy.hpp"
+#  include "opencv2/gpulegacy/private.hpp"
+#endif
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/gpuwarping/src/pyramids.cpp b/modules/gpuwarping/src/pyramids.cpp
new file mode 100644
index 0000000000..aa86e3a40b
--- /dev/null
+++ b/modules/gpuwarping/src/pyramids.cpp
@@ -0,0 +1,244 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+void cv::gpu::pyrDown(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::pyrUp(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+Ptr<ImagePyramid> cv::gpu::createImagePyramid(InputArray, int, Stream&) { throw_no_cuda(); return Ptr<ImagePyramid>(); }
+
+#else // HAVE_CUDA
+
+//////////////////////////////////////////////////////////////////////////////
+// pyrDown
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        template <typename T> void pyrDown_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::pyrDown(InputArray _src, OutputArray _dst, Stream& stream)
+{
+    using namespace cv::gpu::cudev::imgproc;
+
+    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    static const func_t funcs[6][4] =
+    {
+        {pyrDown_gpu<uchar>      , 0 /*pyrDown_gpu<uchar2>*/ , pyrDown_gpu<uchar3>      , pyrDown_gpu<uchar4>      },
+        {0 /*pyrDown_gpu<schar>*/, 0 /*pyrDown_gpu<schar2>*/ , 0 /*pyrDown_gpu<schar3>*/, 0 /*pyrDown_gpu<schar4>*/},
+        {pyrDown_gpu<ushort>     , 0 /*pyrDown_gpu<ushort2>*/, pyrDown_gpu<ushort3>     , pyrDown_gpu<ushort4>     },
+        {pyrDown_gpu<short>      , 0 /*pyrDown_gpu<short2>*/ , pyrDown_gpu<short3>      , pyrDown_gpu<short4>      },
+        {0 /*pyrDown_gpu<int>*/  , 0 /*pyrDown_gpu<int2>*/   , 0 /*pyrDown_gpu<int3>*/  , 0 /*pyrDown_gpu<int4>*/  },
+        {pyrDown_gpu<float>      , 0 /*pyrDown_gpu<float2>*/ , pyrDown_gpu<float3>      , pyrDown_gpu<float4>      }
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+
+    const func_t func = funcs[src.depth()][src.channels() - 1];
+    CV_Assert( func != 0 );
+
+    _dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    func(src, dst, StreamAccessor::getStream(stream));
+}
+
+
+//////////////////////////////////////////////////////////////////////////////
+// pyrUp
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        template <typename T> void pyrUp_gpu(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::pyrUp(InputArray _src, OutputArray _dst, Stream& stream)
+{
+    using namespace cv::gpu::cudev::imgproc;
+
+    typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
+    static const func_t funcs[6][4] =
+    {
+        {pyrUp_gpu<uchar>      , 0 /*pyrUp_gpu<uchar2>*/ , pyrUp_gpu<uchar3>      , pyrUp_gpu<uchar4>      },
+        {0 /*pyrUp_gpu<schar>*/, 0 /*pyrUp_gpu<schar2>*/ , 0 /*pyrUp_gpu<schar3>*/, 0 /*pyrUp_gpu<schar4>*/},
+        {pyrUp_gpu<ushort>     , 0 /*pyrUp_gpu<ushort2>*/, pyrUp_gpu<ushort3>     , pyrUp_gpu<ushort4>     },
+        {pyrUp_gpu<short>      , 0 /*pyrUp_gpu<short2>*/ , pyrUp_gpu<short3>      , pyrUp_gpu<short4>      },
+        {0 /*pyrUp_gpu<int>*/  , 0 /*pyrUp_gpu<int2>*/   , 0 /*pyrUp_gpu<int3>*/  , 0 /*pyrUp_gpu<int4>*/  },
+        {pyrUp_gpu<float>      , 0 /*pyrUp_gpu<float2>*/ , pyrUp_gpu<float3>      , pyrUp_gpu<float4>      }
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+
+    const func_t func = funcs[src.depth()][src.channels() - 1];
+    CV_Assert( func != 0 );
+
+    _dst.create(src.rows * 2, src.cols * 2, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    func(src, dst, StreamAccessor::getStream(stream));
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// ImagePyramid
+
+#ifdef HAVE_OPENCV_GPULEGACY
+
+namespace
+{
+    class ImagePyramidImpl : public ImagePyramid
+    {
+    public:
+        ImagePyramidImpl(InputArray img, int nLayers, Stream& stream);
+
+        void getLayer(OutputArray outImg, Size outRoi, Stream& stream = Stream::Null()) const;
+
+    private:
+        GpuMat layer0_;
+        std::vector<GpuMat> pyramid_;
+        int nLayers_;
+    };
+
+    ImagePyramidImpl::ImagePyramidImpl(InputArray _img, int numLayers, Stream& stream)
+    {
+        GpuMat img = _img.getGpuMat();
+
+        CV_Assert( img.depth() <= CV_32F && img.channels() <= 4 );
+
+        img.copyTo(layer0_, stream);
+
+        Size szLastLayer = img.size();
+        nLayers_ = 1;
+
+        if (numLayers <= 0)
+            numLayers = 255; // it will cut-off when any of the dimensions goes 1
+
+        pyramid_.resize(numLayers);
+
+        for (int i = 0; i < numLayers - 1; ++i)
+        {
+            Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
+
+            if (szCurLayer.width == 0 || szCurLayer.height == 0)
+                break;
+
+            ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
+            nLayers_++;
+
+            const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
+
+            cv::gpu::cudev::pyramid::downsampleX2(prevLayer, pyramid_[i], img.depth(), img.channels(), StreamAccessor::getStream(stream));
+
+            szLastLayer = szCurLayer;
+        }
+    }
+
+    void ImagePyramidImpl::getLayer(OutputArray _outImg, Size outRoi, Stream& stream) const
+    {
+        CV_Assert( outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0 );
+
+        ensureSizeIsEnough(outRoi, layer0_.type(), _outImg);
+        GpuMat outImg = _outImg.getGpuMat();
+
+        if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
+        {
+            layer0_.copyTo(outImg, stream);
+            return;
+        }
+
+        float lastScale = 1.0f;
+        float curScale;
+        GpuMat lastLayer = layer0_;
+        GpuMat curLayer;
+
+        for (int i = 0; i < nLayers_ - 1; ++i)
+        {
+            curScale = lastScale * 0.5f;
+            curLayer = pyramid_[i];
+
+            if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
+            {
+                curLayer.copyTo(outImg, stream);
+            }
+
+            if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
+                break;
+
+            lastScale = curScale;
+            lastLayer = curLayer;
+        }
+
+        cv::gpu::cudev::pyramid::interpolateFrom1(lastLayer, outImg, outImg.depth(), outImg.channels(), StreamAccessor::getStream(stream));
+    }
+}
+
+#endif
+
+Ptr<ImagePyramid> cv::gpu::createImagePyramid(InputArray img, int nLayers, Stream& stream)
+{
+#ifndef HAVE_OPENCV_GPULEGACY
+    (void) img;
+    (void) numLayers;
+    (void) stream;
+    throw_no_cuda();
+    return Ptr<ImagePyramid>();
+#else
+    return Ptr<ImagePyramid>(new ImagePyramidImpl(img, nLayers, stream));
+#endif
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/src/remap.cpp b/modules/gpuwarping/src/remap.cpp
similarity index 75%
rename from modules/gpu/src/remap.cpp
rename to modules/gpuwarping/src/remap.cpp
index 4b87286331..c3d797783a 100644
--- a/modules/gpu/src/remap.cpp
+++ b/modules/gpuwarping/src/remap.cpp
@@ -44,11 +44,11 @@
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::remap(const GpuMat&, GpuMat&, const GpuMat&, const GpuMat&, int, int, Scalar, Stream&){ throw_nogpu(); }
+void cv::gpu::remap(InputArray, OutputArray, InputArray, InputArray, int, int, Scalar, Stream&){ throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -58,13 +58,12 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode, Scalar borderValue, Stream& stream)
+void cv::gpu::remap(InputArray _src, OutputArray _dst, InputArray _xmap, InputArray _ymap, int interpolation, int borderMode, Scalar borderValue, Stream& stream)
 {
-    using namespace cv::gpu::device::imgproc;
+    using namespace cv::gpu::cudev::imgproc;
 
     typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation,
         int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
-
     static const func_t funcs[6][4] =
     {
         {remap_gpu<uchar>      , 0 /*remap_gpu<uchar2>*/ , remap_gpu<uchar3>     , remap_gpu<uchar4>     },
@@ -75,18 +74,21 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
         {remap_gpu<float>      , 0 /*remap_gpu<float2>*/ , remap_gpu<float3>     , remap_gpu<float4>     }
     };
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    CV_Assert(xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size());
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
+    GpuMat src = _src.getGpuMat();
+    GpuMat xmap = _xmap.getGpuMat();
+    GpuMat ymap = _ymap.getGpuMat();
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size() );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
 
     const func_t func = funcs[src.depth()][src.channels() - 1];
-    CV_Assert(func != 0);
+    if (!func)
+        CV_Error(Error::StsUnsupportedFormat, "Unsupported input type");
 
-    int gpuBorderType;
-    CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
-
-    dst.create(xmap.size(), src.type());
+    _dst.create(xmap.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Scalar_<float> borderValueFloat;
     borderValueFloat = borderValue;
@@ -96,7 +98,7 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
     src.locateROI(wholeSize, ofs);
 
     func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y, xmap, ymap,
-        dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
+        dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
 }
 
 #endif // HAVE_CUDA
diff --git a/modules/gpuwarping/src/resize.cpp b/modules/gpuwarping/src/resize.cpp
new file mode 100644
index 0000000000..9c82318f9f
--- /dev/null
+++ b/modules/gpuwarping/src/resize.cpp
@@ -0,0 +1,108 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+#if !defined HAVE_CUDA || defined(CUDA_DISABLER)
+
+void cv::gpu::resize(InputArray, OutputArray, Size, double, double, int, Stream&) { throw_no_cuda(); }
+
+#else // HAVE_CUDA
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <typename T>
+    void resize(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+}}}
+
+void cv::gpu::resize(InputArray _src, OutputArray _dst, Size dsize, double fx, double fy, int interpolation, Stream& stream)
+{
+    GpuMat src = _src.getGpuMat();
+
+    typedef void (*func_t)(const PtrStepSzb& src, const PtrStepSzb& srcWhole, int yoff, int xoff, const PtrStepSzb& dst, float fy, float fx, int interpolation, cudaStream_t stream);
+    static const func_t funcs[6][4] =
+    {
+        {cudev::resize<uchar>      , 0 /*cudev::resize<uchar2>*/ , cudev::resize<uchar3>     , cudev::resize<uchar4>     },
+        {0 /*cudev::resize<schar>*/, 0 /*cudev::resize<char2>*/  , 0 /*cudev::resize<char3>*/, 0 /*cudev::resize<char4>*/},
+        {cudev::resize<ushort>     , 0 /*cudev::resize<ushort2>*/, cudev::resize<ushort3>    , cudev::resize<ushort4>    },
+        {cudev::resize<short>      , 0 /*cudev::resize<short2>*/ , cudev::resize<short3>     , cudev::resize<short4>     },
+        {0 /*cudev::resize<int>*/  , 0 /*cudev::resize<int2>*/   , 0 /*cudev::resize<int3>*/ , 0 /*cudev::resize<int4>*/ },
+        {cudev::resize<float>      , 0 /*cudev::resize<float2>*/ , cudev::resize<float3>     , cudev::resize<float4>     }
+    };
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC || interpolation == INTER_AREA );
+    CV_Assert( !(dsize == Size()) || (fx > 0 && fy > 0) );
+
+    if (dsize == Size())
+    {
+        dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));
+    }
+    else
+    {
+        fx = static_cast<double>(dsize.width) / src.cols;
+        fy = static_cast<double>(dsize.height) / src.rows;
+    }
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    if (dsize == src.size())
+    {
+        src.copyTo(dst, stream);
+        return;
+    }
+
+    const func_t func = funcs[src.depth()][src.channels() - 1];
+
+    if (!func)
+        CV_Error(Error::StsUnsupportedFormat, "Unsupported combination of source and destination types");
+
+    Size wholeSize;
+    Point ofs;
+    src.locateROI(wholeSize, ofs);
+    PtrStepSzb wholeSrc(wholeSize.height, wholeSize.width, src.datastart, src.step);
+
+    func(src, wholeSrc, ofs.y, ofs.x, dst, static_cast<float>(1.0 / fy), static_cast<float>(1.0 / fx), interpolation, StreamAccessor::getStream(stream));
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/gpu/src/warp.cpp b/modules/gpuwarping/src/warp.cpp
similarity index 57%
rename from modules/gpu/src/warp.cpp
rename to modules/gpuwarping/src/warp.cpp
index 827d5219f1..2775fc0847 100644
--- a/modules/gpu/src/warp.cpp
+++ b/modules/gpuwarping/src/warp.cpp
@@ -42,18 +42,26 @@
 
 #include "precomp.hpp"
 
+using namespace cv;
+using namespace cv::gpu;
+
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
+void cv::gpu::warpAffine(InputArray, OutputArray, InputArray, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpAffineMaps(InputArray, bool, Size, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int, int, Scalar, Stream&) { throw_nogpu(); }
-void cv::gpu::buildWarpAffineMaps(const Mat&, bool, Size, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::warpPerspective(InputArray, OutputArray, InputArray, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpPerspectiveMaps(InputArray, bool, Size, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int, int, Scalar, Stream&) { throw_nogpu(); }
-void cv::gpu::buildWarpPerspectiveMaps(const Mat&, bool, Size, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
+void cv::gpu::buildWarpPlaneMaps(Size, Rect, InputArray, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpCylindricalMaps(Size, Rect, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpSphericalMaps(Size, Rect, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
+
+void cv::gpu::rotate(InputArray, OutputArray, Size, double, double, double, int, Stream&) { throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -71,14 +79,19 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-void cv::gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream)
+void cv::gpu::buildWarpAffineMaps(InputArray _M, bool inverse, Size dsize, OutputArray _xmap, OutputArray _ymap, Stream& stream)
 {
-    using namespace cv::gpu::device::imgproc;
+    using namespace cv::gpu::cudev::imgproc;
 
-    CV_Assert(M.rows == 2 && M.cols == 3);
+    Mat M = _M.getMat();
 
-    xmap.create(dsize, CV_32FC1);
-    ymap.create(dsize, CV_32FC1);
+    CV_Assert( M.rows == 2 && M.cols == 3 );
+
+    _xmap.create(dsize, CV_32FC1);
+    _ymap.create(dsize, CV_32FC1);
+
+    GpuMat xmap = _xmap.getGpuMat();
+    GpuMat ymap = _ymap.getGpuMat();
 
     float coeffs[2 * 3];
     Mat coeffsMat(2, 3, CV_32F, (void*)coeffs);
@@ -95,14 +108,19 @@ void cv::gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat
     buildWarpAffineMaps_gpu(coeffs, xmap, ymap, StreamAccessor::getStream(stream));
 }
 
-void cv::gpu::buildWarpPerspectiveMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream)
+void cv::gpu::buildWarpPerspectiveMaps(InputArray _M, bool inverse, Size dsize, OutputArray _xmap, OutputArray _ymap, Stream& stream)
 {
-    using namespace cv::gpu::device::imgproc;
+    using namespace cv::gpu::cudev::imgproc;
 
-    CV_Assert(M.rows == 3 && M.cols == 3);
+    Mat M = _M.getMat();
 
-    xmap.create(dsize, CV_32FC1);
-    ymap.create(dsize, CV_32FC1);
+    CV_Assert( M.rows == 3 && M.cols == 3 );
+
+    _xmap.create(dsize, CV_32FC1);
+    _ymap.create(dsize, CV_32FC1);
+
+    GpuMat xmap = _xmap.getGpuMat();
+    GpuMat ymap = _ymap.getGpuMat();
 
     float coeffs[3 * 3];
     Mat coeffsMat(3, 3, CV_32F, (void*)coeffs);
@@ -121,27 +139,18 @@ void cv::gpu::buildWarpPerspectiveMaps(const Mat& M, bool inverse, Size dsize, G
 
 namespace
 {
-    template<int DEPTH> struct NppTypeTraits;
-    template<> struct NppTypeTraits<CV_8U>  { typedef Npp8u npp_t; };
-    template<> struct NppTypeTraits<CV_8S>  { typedef Npp8s npp_t; };
-    template<> struct NppTypeTraits<CV_16U> { typedef Npp16u npp_t; };
-    template<> struct NppTypeTraits<CV_16S> { typedef Npp16s npp_t; typedef Npp16sc npp_complex_type; };
-    template<> struct NppTypeTraits<CV_32S> { typedef Npp32s npp_t; typedef Npp32sc npp_complex_type; };
-    template<> struct NppTypeTraits<CV_32F> { typedef Npp32f npp_t; typedef Npp32fc npp_complex_type; };
-    template<> struct NppTypeTraits<CV_64F> { typedef Npp64f npp_t; typedef Npp64fc npp_complex_type; };
-
     template <int DEPTH> struct NppWarpFunc
     {
-        typedef typename NppTypeTraits<DEPTH>::npp_t npp_t;
+        typedef typename NPPTypeTraits<DEPTH>::npp_type npp_type;
 
-        typedef NppStatus (*func_t)(const npp_t* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, npp_t* pDst,
+        typedef NppStatus (*func_t)(const npp_type* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, npp_type* pDst,
                                     int dstStep, NppiRect dstRoi, const double coeffs[][3],
                                     int interpolation);
     };
 
     template <int DEPTH, typename NppWarpFunc<DEPTH>::func_t func> struct NppWarp
     {
-        typedef typename NppWarpFunc<DEPTH>::npp_t npp_t;
+        typedef typename NppWarpFunc<DEPTH>::npp_type npp_type;
 
         static void call(const cv::gpu::GpuMat& src, cv::gpu::GpuMat& dst, double coeffs[][3], int interpolation, cudaStream_t stream)
         {
@@ -165,8 +174,8 @@ namespace
 
             cv::gpu::NppStreamHandler h(stream);
 
-            nppSafeCall( func(src.ptr<npp_t>(), srcsz, static_cast<int>(src.step), srcroi,
-                              dst.ptr<npp_t>(), static_cast<int>(dst.step), dstroi,
+            nppSafeCall( func(src.ptr<npp_type>(), srcsz, static_cast<int>(src.step), srcroi,
+                              dst.ptr<npp_type>(), static_cast<int>(dst.step), dstroi,
                               coeffs, npp_inter[interpolation]) );
 
             if (stream == 0)
@@ -175,17 +184,21 @@ namespace
     };
 }
 
-void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& s)
+void cv::gpu::warpAffine(InputArray _src, OutputArray _dst, InputArray _M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& stream)
 {
-    CV_Assert(M.rows == 2 && M.cols == 3);
+    GpuMat src = _src.getGpuMat();
+    Mat M = _M.getMat();
 
-    int interpolation = flags & INTER_MAX;
+    CV_Assert( M.rows == 2 && M.cols == 3 );
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
+    const int interpolation = flags & INTER_MAX;
 
-    dst.create(dsize, src.type());
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Size wholeSize;
     Point ofs;
@@ -259,7 +272,7 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
             }
         };
 
-        dst.setTo(borderValue);
+        dst.setTo(borderValue, stream);
 
         double coeffs[2][3];
         Mat coeffsMat(2, 3, CV_64F, (void*)coeffs);
@@ -268,11 +281,11 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
         const func_t func = funcs[(flags & WARP_INVERSE_MAP) != 0][src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(s));
+        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(stream));
     }
     else
     {
-        using namespace cv::gpu::device::imgproc;
+        using namespace cv::gpu::cudev::imgproc;
 
         typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float coeffs[2 * 3], PtrStepSzb dst, int interpolation,
             int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
@@ -290,9 +303,6 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
         const func_t func = funcs[src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        int gpuBorderType;
-        CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
-
         float coeffs[2 * 3];
         Mat coeffsMat(2, 3, CV_32F, (void*)coeffs);
 
@@ -309,21 +319,25 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
         borderValueFloat = borderValue;
 
         func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y, coeffs,
-            dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(s), deviceSupports(FEATURE_SET_COMPUTE_20));
+            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
     }
 }
 
-void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& s)
+void cv::gpu::warpPerspective(InputArray _src, OutputArray _dst, InputArray _M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& stream)
 {
-    CV_Assert(M.rows == 3 && M.cols == 3);
+    GpuMat src = _src.getGpuMat();
+    Mat M = _M.getMat();
 
-    int interpolation = flags & INTER_MAX;
+    CV_Assert( M.rows == 3 && M.cols == 3 );
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
+    const int interpolation = flags & INTER_MAX;
 
-    dst.create(dsize, src.type());
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP) ;
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Size wholeSize;
     Point ofs;
@@ -397,7 +411,7 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
             }
         };
 
-        dst.setTo(borderValue);
+        dst.setTo(borderValue, stream);
 
         double coeffs[3][3];
         Mat coeffsMat(3, 3, CV_64F, (void*)coeffs);
@@ -406,11 +420,11 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
         const func_t func = funcs[(flags & WARP_INVERSE_MAP) != 0][src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(s));
+        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(stream));
     }
     else
     {
-        using namespace cv::gpu::device::imgproc;
+        using namespace cv::gpu::cudev::imgproc;
 
         typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, float coeffs[2 * 3], PtrStepSzb dst, int interpolation,
             int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
@@ -428,9 +442,6 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
         const func_t func = funcs[src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        int gpuBorderType;
-        CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
-
         float coeffs[3 * 3];
         Mat coeffsMat(3, 3, CV_32F, (void*)coeffs);
 
@@ -447,8 +458,199 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
         borderValueFloat = borderValue;
 
         func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y, coeffs,
-            dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(s), deviceSupports(FEATURE_SET_COMPUTE_20));
+            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
     }
 }
 
+//////////////////////////////////////////////////////////////////////////////
+// buildWarpPlaneMaps
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void buildWarpPlaneMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                const float k_rinv[9], const float r_kinv[9], const float t[3], float scale,
+                                cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, InputArray _T,
+                                 float scale, OutputArray _map_x, OutputArray _map_y, Stream& stream)
+{
+    (void) src_size;
+
+    Mat K = _K.getMat();
+    Mat R = _R.getMat();
+    Mat T = _T.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
+    CV_Assert( (T.size() == Size(3,1) || T.size() == Size(1,3)) && T.type() == CV_32FC1 && T.isContinuous() );
+
+    Mat K_Rinv = K * R.t();
+    Mat R_Kinv = R * K.inv();
+    CV_Assert( K_Rinv.isContinuous() );
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
+
+    cudev::imgproc::buildWarpPlaneMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(),
+                       T.ptr<float>(), scale, StreamAccessor::getStream(stream));
+}
+
+//////////////////////////////////////////////////////////////////////////////
+// buildWarpCylyndricalMaps
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void buildWarpCylindricalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                      const float k_rinv[9], const float r_kinv[9], float scale,
+                                      cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, float scale,
+                                       OutputArray _map_x, OutputArray _map_y, Stream& stream)
+{
+    (void) src_size;
+
+    Mat K = _K.getMat();
+    Mat R = _R.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
+
+    Mat K_Rinv = K * R.t();
+    Mat R_Kinv = R * K.inv();
+    CV_Assert( K_Rinv.isContinuous() );
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
+
+    cudev::imgproc::buildWarpCylindricalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
+}
+
+
+//////////////////////////////////////////////////////////////////////////////
+// buildWarpSphericalMaps
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace imgproc
+    {
+        void buildWarpSphericalMaps(int tl_u, int tl_v, PtrStepSzf map_x, PtrStepSzf map_y,
+                                    const float k_rinv[9], const float r_kinv[9], float scale,
+                                    cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, float scale,
+                                     OutputArray _map_x, OutputArray _map_y, Stream& stream)
+{
+    (void) src_size;
+
+    Mat K = _K.getMat();
+    Mat R = _R.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
+
+    Mat K_Rinv = K * R.t();
+    Mat R_Kinv = R * K.inv();
+    CV_Assert( K_Rinv.isContinuous() );
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
+
+    cudev::imgproc::buildWarpSphericalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
+}
+
+////////////////////////////////////////////////////////////////////////
+// rotate
+
+namespace
+{
+    template <int DEPTH> struct NppRotateFunc
+    {
+        typedef typename NPPTypeTraits<DEPTH>::npp_type npp_type;
+
+        typedef NppStatus (*func_t)(const npp_type* pSrc, NppiSize oSrcSize, int nSrcStep, NppiRect oSrcROI,
+                                    npp_type* pDst, int nDstStep, NppiRect oDstROI,
+                                    double nAngle, double nShiftX, double nShiftY, int eInterpolation);
+    };
+
+    template <int DEPTH, typename NppRotateFunc<DEPTH>::func_t func> struct NppRotate
+    {
+        typedef typename NppRotateFunc<DEPTH>::npp_type npp_type;
+
+        static void call(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, cudaStream_t stream)
+        {
+            (void)dsize;
+            static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
+
+            NppStreamHandler h(stream);
+
+            NppiSize srcsz;
+            srcsz.height = src.rows;
+            srcsz.width = src.cols;
+            NppiRect srcroi;
+            srcroi.x = srcroi.y = 0;
+            srcroi.height = src.rows;
+            srcroi.width = src.cols;
+            NppiRect dstroi;
+            dstroi.x = dstroi.y = 0;
+            dstroi.height = dst.rows;
+            dstroi.width = dst.cols;
+
+            nppSafeCall( func(src.ptr<npp_type>(), srcsz, static_cast<int>(src.step), srcroi,
+                dst.ptr<npp_type>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+    };
+}
+
+void cv::gpu::rotate(InputArray _src, OutputArray _dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& stream)
+{
+    typedef void (*func_t)(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, cudaStream_t stream);
+    static const func_t funcs[6][4] =
+    {
+        {NppRotate<CV_8U, nppiRotate_8u_C1R>::call, 0, NppRotate<CV_8U, nppiRotate_8u_C3R>::call, NppRotate<CV_8U, nppiRotate_8u_C4R>::call},
+        {0,0,0,0},
+        {NppRotate<CV_16U, nppiRotate_16u_C1R>::call, 0, NppRotate<CV_16U, nppiRotate_16u_C3R>::call, NppRotate<CV_16U, nppiRotate_16u_C4R>::call},
+        {0,0,0,0},
+        {0,0,0,0},
+        {NppRotate<CV_32F, nppiRotate_32f_C1R>::call, 0, NppRotate<CV_32F, nppiRotate_32f_C3R>::call, NppRotate<CV_32F, nppiRotate_32f_C4R>::call}
+    };
+
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
+    CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    dst.setTo(Scalar::all(0), stream);
+
+    funcs[src.depth()][src.channels() - 1](src, dst, dsize, angle, xShift, yShift, interpolation, StreamAccessor::getStream(stream));
+}
+
 #endif // HAVE_CUDA
diff --git a/modules/gpu/test/interpolation.hpp b/modules/gpuwarping/test/interpolation.hpp
similarity index 94%
rename from modules/gpu/test/interpolation.hpp
rename to modules/gpuwarping/test/interpolation.hpp
index 9ad08b1652..7a00143e1d 100644
--- a/modules/gpu/test/interpolation.hpp
+++ b/modules/gpuwarping/test/interpolation.hpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
diff --git a/modules/gpuwarping/test/test_main.cpp b/modules/gpuwarping/test/test_main.cpp
new file mode 100644
index 0000000000..eea3d7c008
--- /dev/null
+++ b/modules/gpuwarping/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_GPU_TEST_MAIN("gpu")
diff --git a/modules/gpuwarping/test/test_precomp.hpp b/modules/gpuwarping/test/test_precomp.hpp
new file mode 100644
index 0000000000..9ac7fc0b83
--- /dev/null
+++ b/modules/gpuwarping/test/test_precomp.hpp
@@ -0,0 +1,64 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/gpuwarping.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include "cvconfig.h"
+
+#include "interpolation.hpp"
+
+#endif
diff --git a/modules/gpu/test/test_pyramids.cpp b/modules/gpuwarping/test/test_pyramids.cpp
similarity index 90%
rename from modules/gpu/test/test_pyramids.cpp
rename to modules/gpuwarping/test/test_pyramids.cpp
index c3d56b63a0..f296b7d6f2 100644
--- a/modules/gpu/test/test_pyramids.cpp
+++ b/modules/gpuwarping/test/test_pyramids.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ////////////////////////////////////////////////////////
 // pyrDown
 
@@ -77,7 +80,7 @@ GPU_TEST_P(PyrDown, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-4 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrDown, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, PyrDown, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
@@ -117,7 +120,7 @@ GPU_TEST_P(PyrUp, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-4 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrUp, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, PyrUp, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
diff --git a/modules/gpu/test/test_remap.cpp b/modules/gpuwarping/test/test_remap.cpp
similarity index 94%
rename from modules/gpu/test/test_remap.cpp
rename to modules/gpuwarping/test/test_remap.cpp
index a815ed0e0d..c1899ff611 100644
--- a/modules/gpu/test/test_remap.cpp
+++ b/modules/gpuwarping/test/test_remap.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ///////////////////////////////////////////////////////////////////
 // Gold implementation
 
@@ -166,7 +169,7 @@ GPU_TEST_P(Remap, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-3 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Remap, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, Remap, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
diff --git a/modules/gpu/test/test_resize.cpp b/modules/gpuwarping/test/test_resize.cpp
similarity index 78%
rename from modules/gpu/test/test_resize.cpp
rename to modules/gpuwarping/test/test_resize.cpp
index a34da54fdb..13326c07bb 100644
--- a/modules/gpu/test/test_resize.cpp
+++ b/modules/gpuwarping/test/test_resize.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 ///////////////////////////////////////////////////////////////////
 // Gold implementation
 
@@ -149,10 +152,10 @@ GPU_TEST_P(Resize, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-2 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Resize, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, Resize, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC3), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
     testing::Values(0.3, 0.5, 1.5, 2.0),
     testing::Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
     WHOLE_SUBMAT));
@@ -195,53 +198,12 @@ GPU_TEST_P(ResizeSameAsHost, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-2 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, ResizeSameAsHost, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, ResizeSameAsHost, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
-    testing::Values(MatType(CV_8UC3), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
     testing::Values(0.3, 0.5),
-    testing::Values(Interpolation(cv::INTER_AREA), Interpolation(cv::INTER_NEAREST)),  //, Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)
+    testing::Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_AREA)),
     WHOLE_SUBMAT));
 
-///////////////////////////////////////////////////////////////////
-// Test NPP
-
-PARAM_TEST_CASE(ResizeNPP, cv::gpu::DeviceInfo, MatType, double, Interpolation)
-{
-    cv::gpu::DeviceInfo devInfo;
-    double coeff;
-    int interpolation;
-    int type;
-
-    virtual void SetUp()
-    {
-        devInfo = GET_PARAM(0);
-        type = GET_PARAM(1);
-        coeff = GET_PARAM(2);
-        interpolation = GET_PARAM(3);
-
-        cv::gpu::setDevice(devInfo.deviceID());
-    }
-};
-
-GPU_TEST_P(ResizeNPP, Accuracy)
-{
-    cv::Mat src = readImageType("stereobp/aloe-L.png", type);
-    ASSERT_FALSE(src.empty());
-
-    cv::gpu::GpuMat dst;
-    cv::gpu::resize(loadMat(src), dst, cv::Size(), coeff, coeff, interpolation);
-
-    cv::Mat dst_gold;
-    resizeGold(src, dst_gold, coeff, coeff, interpolation);
-
-    EXPECT_MAT_SIMILAR(dst_gold, dst, 1e-1);
-}
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, ResizeNPP, testing::Combine(
-    ALL_DEVICES,
-    testing::Values(MatType(CV_8UC1), MatType(CV_8UC4)),
-    testing::Values(0.3, 0.5, 1.5, 2.0),
-    testing::Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR))));
-
 #endif // HAVE_CUDA
diff --git a/modules/gpu/test/test_warp_affine.cpp b/modules/gpuwarping/test/test_warp_affine.cpp
similarity index 95%
rename from modules/gpu/test/test_warp_affine.cpp
rename to modules/gpuwarping/test/test_warp_affine.cpp
index fe3a1353b1..206446c4d0 100644
--- a/modules/gpu/test/test_warp_affine.cpp
+++ b/modules/gpuwarping/test/test_warp_affine.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 namespace
 {
     cv::Mat createTransfomMatrix(cv::Size srcSize, double angle)
@@ -98,7 +101,7 @@ GPU_TEST_P(BuildWarpAffineMaps, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, BuildWarpAffineMaps, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, BuildWarpAffineMaps, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DIRECT_INVERSE));
@@ -219,7 +222,7 @@ GPU_TEST_P(WarpAffine, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-1 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, WarpAffine, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, WarpAffine, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
@@ -268,7 +271,7 @@ GPU_TEST_P(WarpAffineNPP, Accuracy)
     EXPECT_MAT_SIMILAR(dst_gold, dst, 2e-2);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, WarpAffineNPP, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, WarpAffineNPP, testing::Combine(
     ALL_DEVICES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
     DIRECT_INVERSE,
diff --git a/modules/gpu/test/test_warp_perspective.cpp b/modules/gpuwarping/test/test_warp_perspective.cpp
similarity index 95%
rename from modules/gpu/test/test_warp_perspective.cpp
rename to modules/gpuwarping/test/test_warp_perspective.cpp
index dd2054a049..49f844c3f6 100644
--- a/modules/gpu/test/test_warp_perspective.cpp
+++ b/modules/gpuwarping/test/test_warp_perspective.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 
 #ifdef HAVE_CUDA
 
+using namespace cvtest;
+
 namespace
 {
     cv::Mat createTransfomMatrix(cv::Size srcSize, double angle)
@@ -99,7 +102,7 @@ GPU_TEST_P(BuildWarpPerspectiveMaps, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, 0.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, BuildWarpPerspectiveMaps, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, BuildWarpPerspectiveMaps, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     DIRECT_INVERSE));
@@ -222,7 +225,7 @@ GPU_TEST_P(WarpPerspective, Accuracy)
     EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-1 : 1.0);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, WarpPerspective, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, WarpPerspective, testing::Combine(
     ALL_DEVICES,
     DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
@@ -271,7 +274,7 @@ GPU_TEST_P(WarpPerspectiveNPP, Accuracy)
     EXPECT_MAT_SIMILAR(dst_gold, dst, 2e-2);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, WarpPerspectiveNPP, testing::Combine(
+INSTANTIATE_TEST_CASE_P(GPU_Warping, WarpPerspectiveNPP, testing::Combine(
     ALL_DEVICES,
     testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)),
     DIRECT_INVERSE,
diff --git a/modules/highgui/CMakeLists.txt b/modules/highgui/CMakeLists.txt
index 8d3056742a..011bb0dc33 100644
--- a/modules/highgui/CMakeLists.txt
+++ b/modules/highgui/CMakeLists.txt
@@ -9,13 +9,12 @@ ocv_add_module(highgui opencv_imgproc OPTIONAL opencv_androidcamera)
 
 ocv_clear_vars(GRFMT_LIBS)
 
-if(WITH_PNG OR WITH_TIFF OR WITH_OPENEXR)
-  ocv_include_directories(${ZLIB_INCLUDE_DIR})
+if(HAVE_PNG OR HAVE_TIFF OR HAVE_OPENEXR)
+  ocv_include_directories(${ZLIB_INCLUDE_DIRS})
   list(APPEND GRFMT_LIBS ${ZLIB_LIBRARIES})
 endif()
 
-if(WITH_JPEG)
-  add_definitions(-DHAVE_JPEG)
+if(HAVE_JPEG)
   ocv_include_directories(${JPEG_INCLUDE_DIR})
   list(APPEND GRFMT_LIBS ${JPEG_LIBRARIES})
 endif()
@@ -26,27 +25,23 @@ if(WITH_WEBP)
   list(APPEND GRFMT_LIBS ${WEBP_LIBRARIES})
 endif()
 
-if(WITH_PNG)
-  add_definitions(-DHAVE_PNG)
+if(HAVE_PNG)
   add_definitions(${PNG_DEFINITIONS})
   ocv_include_directories(${PNG_INCLUDE_DIR})
   list(APPEND GRFMT_LIBS ${PNG_LIBRARIES})
 endif()
 
-if(WITH_TIFF)
-  add_definitions(-DHAVE_TIFF)
+if(HAVE_TIFF)
   ocv_include_directories(${TIFF_INCLUDE_DIR})
   list(APPEND GRFMT_LIBS ${TIFF_LIBRARIES})
 endif()
 
-if(WITH_JASPER)
-  add_definitions(-DHAVE_JASPER)
+if(HAVE_JASPER)
   ocv_include_directories(${JASPER_INCLUDE_DIR})
   list(APPEND GRFMT_LIBS ${JASPER_LIBRARIES})
 endif()
 
-if(WITH_OPENEXR)
-  add_definitions(-DHAVE_OPENEXR)
+if(HAVE_OPENEXR)
   include_directories(SYSTEM ${OPENEXR_INCLUDE_PATHS})
   list(APPEND GRFMT_LIBS ${OPENEXR_LIBRARIES})
 endif()
@@ -69,51 +64,73 @@ set(highgui_srcs
     src/cap_images.cpp
     src/cap_ffmpeg.cpp
     src/loadsave.cpp
-    src/precomp.cpp
     src/utils.cpp
     src/window.cpp
     )
 
 file(GLOB highgui_ext_hdrs "include/opencv2/*.hpp" "include/opencv2/${name}/*.hpp" "include/opencv2/${name}/*.h")
 
-if(HAVE_QT)
+if(HAVE_QT5)
+  set(CMAKE_AUTOMOC ON)
+  set(CMAKE_INCLUDE_CURRENT_DIR ON)
+
+  QT5_ADD_RESOURCES(_RCC_OUTFILES src/window_QT.qrc)
+  list(APPEND highgui_srcs src/window_QT.cpp src/window_QT.h ${_RCC_OUTFILES})
+
+  foreach(dt5_dep Core Gui Widgets Test Concurrent)
+    add_definitions(${Qt5${dt5_dep}_DEFINITIONS})
+    include_directories(${Qt5${dt5_dep}_INCLUDE_DIRS})
+    list(APPEND HIGHGUI_LIBRARIES ${Qt5${dt5_dep}_LIBRARIES})
+  endforeach()
+
+  if(HAVE_QT_OPENGL)
+    add_definitions(${Qt5OpenGL_DEFINITIONS})
+    include_directories(${Qt5OpenGL_INCLUDE_DIRS})
+    list(APPEND HIGHGUI_LIBRARIES ${Qt5OpenGL_LIBRARIES})
+  endif()
+
+elseif(HAVE_QT)
   if (HAVE_QT_OPENGL)
     set(QT_USE_QTOPENGL TRUE)
   endif()
   include(${QT_USE_FILE})
 
-  if(QT_INCLUDE_DIR)
-    ocv_include_directories(${QT_INCLUDE_DIR})
-  endif()
-
   QT4_ADD_RESOURCES(_RCC_OUTFILES src/window_QT.qrc)
   QT4_WRAP_CPP(_MOC_OUTFILES src/window_QT.h)
 
-  list(APPEND HIGHGUI_LIBRARIES ${QT_LIBRARIES} ${QT_QTTEST_LIBRARY})
+  list(APPEND HIGHGUI_LIBRARIES ${QT_LIBRARIES})
   list(APPEND highgui_srcs src/window_QT.cpp ${_MOC_OUTFILES} ${_RCC_OUTFILES})
   ocv_check_flag_support(CXX -Wno-missing-declarations _have_flag)
   if(${_have_flag})
     set_source_files_properties(${_RCC_OUTFILES} PROPERTIES COMPILE_FLAGS -Wno-missing-declarations)
   endif()
-elseif(WIN32)
+elseif(HAVE_WIN32UI)
   list(APPEND highgui_srcs src/window_w32.cpp)
 elseif(HAVE_GTK)
   list(APPEND highgui_srcs src/window_gtk.cpp)
-elseif(APPLE)
-  if(WITH_CARBON)
-    add_definitions(-DHAVE_CARBON=1)
-    list(APPEND highgui_srcs src/window_carbon.cpp)
-    list(APPEND HIGHGUI_LIBRARIES "-framework Carbon" "-framework QuickTime")
-  elseif(NOT IOS)
-    add_definitions(-DHAVE_COCOA=1)
-    list(APPEND highgui_srcs src/window_cocoa.mm)
-    list(APPEND HIGHGUI_LIBRARIES "-framework Cocoa")
-  endif()
+elseif(HAVE_CARBON)
+  list(APPEND highgui_srcs src/window_carbon.cpp)
+  list(APPEND HIGHGUI_LIBRARIES "-framework Carbon" "-framework QuickTime")
+elseif(HAVE_COCOA)
+  list(APPEND highgui_srcs src/window_cocoa.mm)
+  list(APPEND HIGHGUI_LIBRARIES "-framework Cocoa")
 endif()
 
-if(WIN32)
-  list(APPEND highgui_srcs src/cap_vfw.cpp src/cap_cmu.cpp src/cap_dshow.cpp)
-endif(WIN32)
+if(WIN32 AND NOT ARM)
+  list(APPEND highgui_srcs src/cap_cmu.cpp)
+endif()
+
+if (WIN32 AND HAVE_DSHOW)
+  list(APPEND highgui_srcs src/cap_dshow.cpp)
+endif()
+
+if (WIN32 AND HAVE_MSMF)
+  list(APPEND highgui_srcs src/cap_msmf.cpp)
+endif()
+
+if (WIN32 AND HAVE_VFW)
+  list(APPEND highgui_srcs src/cap_vfw.cpp)
+endif()
 
 if(HAVE_XINE)
   list(APPEND highgui_srcs src/cap_xine.cpp)
@@ -158,7 +175,11 @@ if(HAVE_XIMEA)
   if(XIMEA_LIBRARY_DIR)
     link_directories(${XIMEA_LIBRARY_DIR})
   endif()
-  list(APPEND HIGHGUI_LIBRARIES m3api)
+  if(CMAKE_CL_64)
+    list(APPEND HIGHGUI_LIBRARIES m3apiX64)
+  else()
+    list(APPEND HIGHGUI_LIBRARIES m3api)
+  endif()
 endif(HAVE_XIMEA)
 
 if(HAVE_FFMPEG)
@@ -172,6 +193,7 @@ endif(HAVE_FFMPEG)
 
 if(HAVE_PVAPI)
   add_definitions(-DHAVE_PVAPI)
+  add_definitions(${PVAPI_DEFINITIONS})
   ocv_include_directories(${PVAPI_INCLUDE_PATH})
   set(highgui_srcs src/cap_pvapi.cpp ${highgui_srcs})
   list(APPEND HIGHGUI_LIBRARIES ${PVAPI_LIBRARY})
@@ -185,30 +207,29 @@ if(HAVE_GIGE_API)
   list(APPEND highgui_srcs src/cap_giganetix.cpp)
 endif(HAVE_GIGE_API)
 
-if(WITH_AVFOUNDATION)
-  add_definitions(-DHAVE_AVFOUNDATION=1)
+if(HAVE_AVFOUNDATION)
   list(APPEND highgui_srcs src/cap_avfoundation.mm)
   list(APPEND HIGHGUI_LIBRARIES "-framework AVFoundation" "-framework QuartzCore")
-elseif(APPLE)
-  add_definitions(-DHAVE_QUICKTIME=1)
-  if(WITH_QUICKTIME)
-    list(APPEND highgui_srcs src/cap_qt.cpp)
-    list(APPEND HIGHGUI_LIBRARIES "-framework Carbon" "-framework QuickTime" "-framework CoreFoundation" "-framework QuartzCore")
-  else()
-    list(APPEND highgui_srcs src/cap_qtkit.mm)
-    list(APPEND HIGHGUI_LIBRARIES "-framework QTKit" "-framework QuartzCore" "-framework AppKit")
-  endif()
+endif()
+
+if(HAVE_QUICKTIME)
+  list(APPEND highgui_srcs src/cap_qt.cpp)
+  list(APPEND HIGHGUI_LIBRARIES "-framework Carbon" "-framework QuickTime" "-framework CoreFoundation" "-framework QuartzCore")
+elseif(HAVE_QTKIT)
+  list(APPEND highgui_srcs src/cap_qtkit.mm)
+  list(APPEND HIGHGUI_LIBRARIES "-framework QTKit" "-framework QuartzCore" "-framework AppKit")
 endif()
 
 if(IOS)
   add_definitions(-DHAVE_IOS=1)
-  list(APPEND highgui_srcs src/cap_ios_abstract_camera.mm src/cap_ios_photo_camera.mm src/cap_ios_video_camera.mm)
+  list(APPEND highgui_srcs src/ios_conversions.mm src/cap_ios_abstract_camera.mm src/cap_ios_photo_camera.mm src/cap_ios_video_camera.mm)
   list(APPEND HIGHGUI_LIBRARIES "-framework Accelerate" "-framework AVFoundation" "-framework CoreGraphics" "-framework CoreImage" "-framework CoreMedia" "-framework CoreVideo" "-framework QuartzCore" "-framework AssetsLibrary")
 endif()
 
 if(WIN32)
   link_directories("${OpenCV_SOURCE_DIR}/3rdparty/lib") # for ffmpeg wrapper only
   include_directories(AFTER SYSTEM "${OpenCV_SOURCE_DIR}/3rdparty/include") # for directshow in VS2005 and multi-monitor support on MinGW
+  include_directories(AFTER SYSTEM "${OpenCV_SOURCE_DIR}/3rdparty/include/ffmpeg_") # for tests
 endif()
 
 if(UNIX)
@@ -271,11 +292,6 @@ if(WIN32 AND WITH_FFMPEG)
   set(ffmpeg_bare_name_ver "opencv_ffmpeg${OPENCV_DLLVERSION}${FFMPEG_SUFFIX}.dll")
   set(ffmpeg_path "${OpenCV_SOURCE_DIR}/3rdparty/ffmpeg/${ffmpeg_bare_name}")
 
-  #if(MSVC AND CMAKE_VERSION VERSION_GREATER "2.8.2")
-  #  add_custom_command(TARGET ${the_module} POST_BUILD
-  #                     COMMAND ${CMAKE_COMMAND} -E copy "${ffmpeg_path}" "${EXECUTABLE_OUTPUT_PATH}/$<CONFIGURATION>/${ffmpeg_bare_name_ver}"
-  #                     COMMENT "Copying ${ffmpeg_path} to the output directory")
-  #else
   if(MSVC_IDE)
     add_custom_command(TARGET ${the_module} POST_BUILD
                        COMMAND ${CMAKE_COMMAND} -E copy "${ffmpeg_path}" "${EXECUTABLE_OUTPUT_PATH}/Release/${ffmpeg_bare_name_ver}"
diff --git a/modules/highgui/doc/qt_new_functions.rst b/modules/highgui/doc/qt_new_functions.rst
index bd889db8a2..3d11445b1d 100644
--- a/modules/highgui/doc/qt_new_functions.rst
+++ b/modules/highgui/doc/qt_new_functions.rst
@@ -61,7 +61,7 @@ setWindowProperty
 ---------------------
 Changes parameters of a window dynamically.
 
-.. ocv:function:: void setWindowProperty( const string& winname, int prop_id, double prop_value )
+.. ocv:function:: void setWindowProperty( const String& winname, int prop_id, double prop_value )
 
 .. ocv:pyfunction:: cv2.setWindowProperty(winname, prop_id, prop_value) -> None
 
@@ -97,7 +97,7 @@ getWindowProperty
 ---------------------
 Provides parameters of a window.
 
-.. ocv:function:: double getWindowProperty( const string& winname, int prop_id )
+.. ocv:function:: double getWindowProperty( const String& winname, int prop_id )
 
 .. ocv:pyfunction:: cv2.getWindowProperty(winname, prop_id) -> retval
 
@@ -123,7 +123,7 @@ fontQt
 ----------
 Creates the font to draw a text on an image.
 
-.. ocv:function:: CvFont fontQt(const string& nameFont, int pointSize  = -1, Scalar color = Scalar::all(0), int weight = CV_FONT_NORMAL,  int style = CV_STYLE_NORMAL, int spacing = 0)
+.. ocv:function:: QtFont fontQt( const String& nameFont, int pointSize=-1, Scalar color=Scalar::all(0), int weight=QT_FONT_NORMAL, int style=QT_STYLE_NORMAL, int spacing=0 )
 
 .. ocv:cfunction:: CvFont cvFontQt(const char* nameFont, int pointSize=-1, CvScalar color=cvScalarAll(0), int weight=CV_FONT_NORMAL, int style=CV_STYLE_NORMAL, int spacing=0)
 
@@ -169,7 +169,7 @@ addText
 -----------
 Creates the font to draw a text on an image.
 
-.. ocv:function:: void addText( const Mat& img, const string& text, Point org, CvFont font )
+.. ocv:function:: void addText( const Mat& img, const String& text, Point org, const QtFont& font )
 
 .. ocv:cfunction:: void cvAddText( const CvArr* img, const char* text, CvPoint org, CvFont * arg2 )
 
@@ -195,7 +195,7 @@ displayOverlay
 ------------------
 Displays a  text on a window image as an overlay for a specified duration.
 
-.. ocv:function:: void displayOverlay( const string& winname, const string& text, int delayms=0 )
+.. ocv:function:: void displayOverlay( const String& winname, const String& text, int delayms=0 )
 
 .. ocv:cfunction:: void cvDisplayOverlay(const char* name, const char* text, int delayms = 0)
 
@@ -212,7 +212,7 @@ displayStatusBar
 --------------------
 Displays a text on the window statusbar during the specified period of time.
 
-.. ocv:function:: void displayStatusBar( const string& winname, const string& text, int delayms=0 )
+.. ocv:function:: void displayStatusBar( const String& winname, const String& text, int delayms=0 )
 
 .. ocv:cfunction:: void cvDisplayStatusBar(const char* name, const char* text, int delayms = 0)
 
@@ -230,7 +230,7 @@ setOpenGlDrawCallback
 ------------------------
 Sets a callback function to be called to draw on top of displayed image.
 
-.. ocv:function:: void setOpenGlDrawCallback( const string& winname, OpenGlDrawCallback onOpenGlDraw, void* userdata=0 )
+.. ocv:function:: void setOpenGlDrawCallback( const String& winname, OpenGlDrawCallback onOpenGlDraw, void* userdata=0 )
 
 .. ocv:cfunction:: void cvSetOpenGlDrawCallback( const char* window_name, CvOpenGlDrawCallback callback, void* userdata=NULL )
 
@@ -276,7 +276,7 @@ saveWindowParameters
 ------------------------
 Saves parameters of the specified window.
 
-.. ocv:function:: void saveWindowParameters( const string& windowName )
+.. ocv:function:: void saveWindowParameters( const String& windowName )
 
 .. ocv:cfunction:: void cvSaveWindowParameters(const char* name)
 
@@ -289,7 +289,7 @@ loadWindowParameters
 ------------------------
 Loads parameters of the specified window.
 
-.. ocv:function:: void loadWindowParameters( const string& windowName )
+.. ocv:function:: void loadWindowParameters( const String& windowName )
 
 .. ocv:cfunction:: void cvLoadWindowParameters(const char* name)
 
@@ -302,7 +302,7 @@ createButton
 ----------------
 Attaches a button to the control panel.
 
-.. ocv:function:: int createButton( const string& bar_name, ButtonCallback on_change, void* userdata=NULL, int type=CV_PUSH_BUTTON, bool initial_button_state=0 )
+.. ocv:function:: int createButton( const String& bar_name, ButtonCallback on_change, void* userdata=0, int type=QT_PUSH_BUTTON, bool initial_button_state=false )
 
 .. ocv:cfunction:: int cvCreateButton( const char* button_name=NULL, CvButtonCallback on_change=NULL, void* userdata=NULL, int button_type=CV_PUSH_BUTTON, int initial_button_state=0 )
 
@@ -334,5 +334,3 @@ See below various examples of the ``createButton`` function call: ::
     createButton("button6",callbackButton2,NULL,CV_PUSH_BUTTON,1);
 
 ..
-
-
diff --git a/modules/highgui/doc/reading_and_writing_images_and_video.rst b/modules/highgui/doc/reading_and_writing_images_and_video.rst
index d1829a01d0..6b956b40dd 100644
--- a/modules/highgui/doc/reading_and_writing_images_and_video.rst
+++ b/modules/highgui/doc/reading_and_writing_images_and_video.rst
@@ -35,7 +35,7 @@ imencode
 --------
 Encodes an image into a memory buffer.
 
-.. ocv:function:: bool imencode( const string& ext, InputArray img, vector<uchar>& buf, const vector<int>& params=vector<int>())
+.. ocv:function:: bool imencode( const String& ext, InputArray img, vector<uchar>& buf, const vector<int>& params=vector<int>())
 
 .. ocv:cfunction:: CvMat* cvEncodeImage( const char* ext, const CvArr* image, const int* params=0 )
 
@@ -59,7 +59,7 @@ imread
 ------
 Loads an image from a file.
 
-.. ocv:function:: Mat imread( const string& filename, int flags=1 )
+.. ocv:function:: Mat imread( const String& filename, int flags=IMREAD_COLOR )
 
 .. ocv:pyfunction:: cv2.imread(filename[, flags]) -> retval
 
@@ -67,10 +67,6 @@ Loads an image from a file.
 
 .. ocv:cfunction:: CvMat* cvLoadImageM( const char* filename, int iscolor=CV_LOAD_IMAGE_COLOR )
 
-.. ocv:pyoldfunction:: cv.LoadImage(filename, iscolor=CV_LOAD_IMAGE_COLOR) -> None
-
-.. ocv:pyoldfunction:: cv.LoadImageM(filename, iscolor=CV_LOAD_IMAGE_COLOR) -> None
-
     :param filename: Name of file to be loaded.
 
     :param flags: Flags specifying the color type of a loaded image:
@@ -120,14 +116,12 @@ imwrite
 -----------
 Saves an image to a specified file.
 
-.. ocv:function:: bool imwrite( const string& filename, InputArray img, const vector<int>& params=vector<int>() )
+.. ocv:function:: bool imwrite( const String& filename, InputArray img, const vector<int>& params=vector<int>() )
 
 .. ocv:pyfunction:: cv2.imwrite(filename, img[, params]) -> retval
 
 .. ocv:cfunction:: int cvSaveImage( const char* filename, const CvArr* image, const int* params=0 )
 
-.. ocv:pyoldfunction:: cv.SaveImage(filename, image)-> None
-
     :param filename: Name of the file.
 
     :param image: Image to be saved.
@@ -197,8 +191,8 @@ VideoCapture
 ------------
 .. ocv:class:: VideoCapture
 
-Class for video capturing from video files or cameras.
-The class provides C++ API for capturing video from cameras or for reading video files. Here is how the class can be used: ::
+Class for video capturing from video files, image sequences or cameras.
+The class provides C++ API for capturing video from cameras or for reading video files and image sequences. Here is how the class can be used: ::
 
     #include "opencv2/opencv.hpp"
 
@@ -216,7 +210,7 @@ The class provides C++ API for capturing video from cameras or for reading video
         {
             Mat frame;
             cap >> frame; // get a new frame from camera
-            cvtColor(frame, edges, CV_BGR2GRAY);
+            cvtColor(frame, edges, COLOR_BGR2GRAY);
             GaussianBlur(edges, edges, Size(7,7), 1.5, 1.5);
             Canny(edges, edges, 0, 30, 3);
             imshow("edges", edges);
@@ -229,6 +223,15 @@ The class provides C++ API for capturing video from cameras or for reading video
 
 .. note:: In C API the black-box structure ``CvCapture`` is used instead of ``VideoCapture``.
 
+.. note::
+
+   * A basic sample on using the VideoCapture interface can be found at opencv_source_code/samples/cpp/starter_video.cpp
+   * Another basic video processing sample can be found at opencv_source_code/samples/cpp/video_dmtx.cpp
+
+   * (Python) A basic sample on using the VideoCapture interface can be found at opencv_source_code/samples/python2/video.py
+   * (Python) Another basic video processing sample can be found at opencv_source_code/samples/python2/video_dmtx.py
+   * (Python) A multi threaded video processing sample can be found at opencv_source_code/samples/python2/video_threaded.py
+
 
 VideoCapture::VideoCapture
 ------------------------------
@@ -236,7 +239,7 @@ VideoCapture constructors.
 
 .. ocv:function:: VideoCapture::VideoCapture()
 
-.. ocv:function:: VideoCapture::VideoCapture(const string& filename)
+.. ocv:function:: VideoCapture::VideoCapture(const String& filename)
 
 .. ocv:function:: VideoCapture::VideoCapture(int device)
 
@@ -245,11 +248,9 @@ VideoCapture constructors.
 .. ocv:pyfunction:: cv2.VideoCapture(device) -> <VideoCapture object>
 
 .. ocv:cfunction:: CvCapture* cvCaptureFromCAM( int device )
-.. ocv:pyoldfunction:: cv.CaptureFromCAM(index) -> CvCapture
 .. ocv:cfunction:: CvCapture* cvCaptureFromFile( const char* filename )
-.. ocv:pyoldfunction:: cv.CaptureFromFile(filename) -> CvCapture
 
-    :param filename: name of the opened video file
+    :param filename: name of the opened video file (eg. video.avi) or image sequence (eg. img%02d.jpg)
 
     :param device: id of the opened video capturing device (i.e. a camera index). If there is a single camera connected, just pass 0.
 
@@ -260,13 +261,13 @@ VideoCapture::open
 ---------------------
 Open video file or a capturing device for video capturing
 
-.. ocv:function:: bool VideoCapture::open(const string& filename)
+.. ocv:function:: bool VideoCapture::open(const String& filename)
 .. ocv:function:: bool VideoCapture::open(int device)
 
 .. ocv:pyfunction:: cv2.VideoCapture.open(filename) -> retval
 .. ocv:pyfunction:: cv2.VideoCapture.open(device) -> retval
 
-    :param filename: name of the opened video file
+    :param filename: name of the opened video file (eg. video.avi) or image sequence (eg. img%02d.jpg)
 
     :param device: id of the opened video capturing device (i.e. a camera index).
 
@@ -308,8 +309,6 @@ Grabs the next frame from video file or capturing device.
 
 .. ocv:cfunction:: int cvGrabFrame(CvCapture* capture)
 
-.. ocv:pyoldfunction:: cv.GrabFrame(capture) -> int
-
 The methods/functions grab the next frame from video file or camera and return true (non-zero) in the case of success.
 
 The primary use of the function is in multi-camera environments, especially when the cameras do not have hardware synchronization. That is, you call ``VideoCapture::grab()`` for each camera and after that call the slower method ``VideoCapture::retrieve()`` to decode and get frame from each camera. This way the overhead on demosaicing or motion jpeg decompression etc. is eliminated and the retrieved frames from different cameras will be closer in time.
@@ -321,14 +320,12 @@ VideoCapture::retrieve
 ----------------------
 Decodes and returns the grabbed video frame.
 
-.. ocv:function:: bool VideoCapture::retrieve(Mat& image, int channel=0)
+.. ocv:function:: bool VideoCapture::retrieve( Mat& image, int flag=0 )
 
-.. ocv:pyfunction:: cv2.VideoCapture.retrieve([image[, channel]]) -> retval, image
+.. ocv:pyfunction:: cv2.VideoCapture.retrieve([image[, flag]]) -> retval, image
 
 .. ocv:cfunction:: IplImage* cvRetrieveFrame( CvCapture* capture, int streamIdx=0 )
 
-.. ocv:pyoldfunction:: cv.RetrieveFrame(capture) -> image
-
 The methods/functions decode and return the just grabbed frame. If no frames has been grabbed (camera has been disconnected, or there are no more frames in video file), the methods return false and the functions return NULL pointer.
 
 .. note:: OpenCV 1.x functions ``cvRetrieveFrame`` and ``cv.RetrieveFrame`` return image stored inside the video capturing structure. It is not allowed to modify or release the image! You can copy the frame using :ocv:cfunc:`cvCloneImage` and then do whatever you want with the copy.
@@ -346,8 +343,6 @@ Grabs, decodes and returns the next video frame.
 
 .. ocv:cfunction:: IplImage* cvQueryFrame(CvCapture* capture)
 
-.. ocv:pyoldfunction:: cv.QueryFrame(capture) -> image
-
 The methods/functions combine :ocv:func:`VideoCapture::grab` and :ocv:func:`VideoCapture::retrieve` in one call. This is the most convenient method for reading video files or capturing data from decode and return the just grabbed frame. If no frames has been grabbed (camera has been disconnected, or there are no more frames in video file), the methods return false and the functions return NULL pointer.
 
 .. note:: OpenCV 1.x functions ``cvRetrieveFrame`` and ``cv.RetrieveFrame`` return image stored inside the video capturing structure. It is not allowed to modify or release the image! You can copy the frame using :ocv:cfunc:`cvCloneImage` and then do whatever you want with the copy.
@@ -363,9 +358,6 @@ Returns the specified ``VideoCapture`` property
 
 .. ocv:cfunction:: double cvGetCaptureProperty( CvCapture* capture, int property_id )
 
-.. ocv:pyoldfunction:: cv.GetCaptureProperty(capture, property_id) -> float
-
-
     :param propId: Property identifier. It can be one of the following:
 
         * **CV_CAP_PROP_POS_MSEC** Current position of the video file in milliseconds or video capture timestamp.
@@ -419,8 +411,6 @@ Sets a property in the ``VideoCapture``.
 
 .. ocv:cfunction:: int cvSetCaptureProperty( CvCapture* capture, int property_id, double value )
 
-.. ocv:pyoldfunction:: cv.SetCaptureProperty(capture, property_id, value) -> retval
-
     :param propId: Property identifier. It can be one of the following:
 
         * **CV_CAP_PROP_POS_MSEC** Current position of the video file in milliseconds.
@@ -479,12 +469,11 @@ VideoWriter constructors
 
 .. ocv:function:: VideoWriter::VideoWriter()
 
-.. ocv:function:: VideoWriter::VideoWriter(const string& filename, int fourcc, double fps, Size frameSize, bool isColor=true)
+.. ocv:function:: VideoWriter::VideoWriter(const String& filename, int fourcc, double fps, Size frameSize, bool isColor=true)
 
 .. ocv:pyfunction:: cv2.VideoWriter([filename, fourcc, fps, frameSize[, isColor]]) -> <VideoWriter object>
 
 .. ocv:cfunction:: CvVideoWriter* cvCreateVideoWriter( const char* filename, int fourcc, double fps, CvSize frame_size, int is_color=1 )
-.. ocv:pyoldfunction:: cv.CreateVideoWriter(filename, fourcc, fps, frame_size, is_color=true) -> CvVideoWriter
 
 .. ocv:pyfunction:: cv2.VideoWriter.isOpened() -> retval
 .. ocv:pyfunction:: cv2.VideoWriter.open(filename, fourcc, fps, frameSize[, isColor]) -> retval
@@ -517,7 +506,7 @@ VideoWriter::open
 -----------------
 Initializes or reinitializes video writer.
 
-.. ocv:function:: bool VideoWriter::open(const string& filename, int fourcc, double fps, Size frameSize, bool isColor=true)
+.. ocv:function:: bool VideoWriter::open(const String& filename, int fourcc, double fps, Size frameSize, bool isColor=true)
 
 .. ocv:pyfunction:: cv2.VideoWriter.open(filename, fourcc, fps, frameSize[, isColor]) -> retval
 
@@ -544,11 +533,9 @@ Writes the next video frame
 .. ocv:pyfunction:: cv2.VideoWriter.write(image) -> None
 
 .. ocv:cfunction:: int cvWriteFrame( CvVideoWriter* writer, const IplImage* image )
-.. ocv:pyoldfunction:: cv.WriteFrame(writer, image)->int
 
     :param writer: Video writer structure (OpenCV 1.x API)
 
     :param image: The written frame
 
 The functions/methods write the specified image to video file. It must have the same size as has been specified when opening the video writer.
-
diff --git a/modules/highgui/doc/user_interface.rst b/modules/highgui/doc/user_interface.rst
index def8451a2c..d4055ef6f0 100644
--- a/modules/highgui/doc/user_interface.rst
+++ b/modules/highgui/doc/user_interface.rst
@@ -7,12 +7,12 @@ createTrackbar
 ------------------
 Creates a trackbar and attaches it to the specified window.
 
-.. ocv:function:: int createTrackbar( const string& trackbarname, const string& winname, int* value, int count, TrackbarCallback onChange=0, void* userdata=0)
+.. ocv:function:: int createTrackbar( const String& trackbarname, const String& winname, int* value, int count, TrackbarCallback onChange=0, void* userdata=0)
+
+.. ocv:pyfunction:: cv2.createTrackbar(trackbarName, windowName, value, count, onChange) -> None
 
 .. ocv:cfunction:: int cvCreateTrackbar( const char* trackbar_name, const char* window_name, int* value, int count, CvTrackbarCallback on_change=NULL )
 
-.. ocv:pyoldfunction:: cv.CreateTrackbar(trackbarName, windowName, value, count, onChange) -> None
-
     :param trackbarname: Name of the created trackbar.
 
     :param winname: Name of the window that will be used as a parent of the created trackbar.
@@ -33,18 +33,20 @@ The function ``createTrackbar`` creates a trackbar (a slider or range control) w
 
 Clicking the label of each trackbar enables editing the trackbar values manually.
 
+.. note::
+
+   * An example of using the trackbar functionality can be found at opencv_source_code/samples/cpp/connected_components.cpp
+
 getTrackbarPos
 ------------------
 Returns the trackbar position.
 
-.. ocv:function:: int getTrackbarPos( const string& trackbarname, const string& winname )
+.. ocv:function:: int getTrackbarPos( const String& trackbarname, const String& winname )
 
 .. ocv:pyfunction:: cv2.getTrackbarPos(trackbarname, winname) -> retval
 
 .. ocv:cfunction:: int cvGetTrackbarPos( const char* trackbar_name, const char* window_name )
 
-.. ocv:pyoldfunction:: cv.GetTrackbarPos(trackbarName, windowName) -> retval
-
     :param trackbarname: Name of the trackbar.
 
     :param winname: Name of the window that is the parent of the trackbar.
@@ -59,14 +61,12 @@ imshow
 ----------
 Displays an image in the specified window.
 
-.. ocv:function:: void imshow( const string& winname, InputArray mat )
+.. ocv:function:: void imshow( const String& winname, InputArray mat )
 
 .. ocv:pyfunction:: cv2.imshow(winname, mat) -> None
 
 .. ocv:cfunction:: void cvShowImage( const char* name, const CvArr* image )
 
-.. ocv:pyoldfunction:: cv.ShowImage(name, image) -> None
-
     :param winname: Name of the window.
 
     :param image: Image to be shown.
@@ -79,22 +79,30 @@ The function ``imshow`` displays an image in the specified window. If the window
 
     * If the image is 32-bit floating-point, the pixel values are multiplied by 255. That is, the value range [0,1] is mapped to [0,255].
 
+If window was created with OpenGL support, ``imshow`` also support :ocv:class:`ogl::Buffer` ,  :ocv:class:`ogl::Texture2D` and  :ocv:class:`gpu::GpuMat` as input.
+
+.. note:: This function should be followed by ``waitKey`` function which displays the image for specified milliseconds. Otherwise, it won't display the image. For example, ``waitKey(0)`` will display the window infinitely until any keypress (it is suitable for image display). ``waitKey(25)`` will display a frame for 25 ms, after which display will be automatically closed. (If you put it in a loop to read videos, it will display the video frame-by-frame)
+
 
 namedWindow
 ---------------
 Creates a window.
 
-.. ocv:function:: void namedWindow( const string& winname, int flags=WINDOW_AUTOSIZE )
+.. ocv:function:: void namedWindow( const String& winname, int flags=WINDOW_AUTOSIZE )
 
 .. ocv:pyfunction:: cv2.namedWindow(winname[, flags]) -> None
 
 .. ocv:cfunction:: int cvNamedWindow( const char* name, int flags=CV_WINDOW_AUTOSIZE )
 
-.. ocv:pyoldfunction:: cv.NamedWindow(name, flags=CV_WINDOW_AUTOSIZE)-> None
-
     :param name: Name of the window in the window caption that may be used as a window identifier.
 
-    :param flags: Flags of the window. Currently the only supported flag is  ``CV_WINDOW_AUTOSIZE`` . If this is set, the window size is automatically adjusted to fit the displayed image (see  :ocv:func:`imshow` ), and you cannot change the window size manually.
+    :param flags: Flags of the window. The supported flags are:
+
+        * **WINDOW_NORMAL** If this is set, the user can resize the window (no constraint).
+
+        * **WINDOW_AUTOSIZE** If this is set, the window size is automatically adjusted to fit the displayed image (see  :ocv:func:`imshow` ), and you cannot change the window size manually.
+
+        * **WINDOW_OPENGL** If this is set, the window will be created with OpenGL support.
 
 The function ``namedWindow`` creates a window that can be used as a placeholder for images and trackbars. Created windows are referred to by their names.
 
@@ -119,14 +127,12 @@ destroyWindow
 -------------
 Destroys a window.
 
-.. ocv:function:: void destroyWindow( const string& winname )
+.. ocv:function:: void destroyWindow( const String& winname )
 
 .. ocv:pyfunction:: cv2.destroyWindow(winname) -> None
 
 .. ocv:cfunction:: void cvDestroyWindow( const char* name )
 
-.. ocv:pyoldfunction:: cv.DestroyWindow(name)-> None
-
     :param winname: Name of the window to be destroyed.
 
 The function ``destroyWindow`` destroys the window with the given name.
@@ -142,8 +148,6 @@ Destroys all of the HighGUI windows.
 
 .. ocv:cfunction:: void cvDestroyAllWindows()
 
-.. ocv:pyoldfunction:: cv.DestroyAllWindows()-> None
-
 The function ``destroyAllWindows`` destroys all of the opened HighGUI windows.
 
 
@@ -151,14 +155,12 @@ MoveWindow
 ----------
 Moves window to the specified position
 
-.. ocv:function:: void moveWindow( const string& winname, int x, int y )
+.. ocv:function:: void moveWindow( const String& winname, int x, int y )
 
 .. ocv:pyfunction:: cv2.moveWindow(winname, x, y) -> None
 
 .. ocv:cfunction:: void cvMoveWindow( const char* name, int x, int y )
 
-.. ocv:pyoldfunction:: cv.MoveWindow(name, x, y)-> None
-
     :param winname: Window name
 
     :param x: The new x-coordinate of the window
@@ -170,14 +172,12 @@ ResizeWindow
 ------------
 Resizes window to the specified size
 
-.. ocv:function:: void resizeWindow( const string& winname, int width, int height )
+.. ocv:function:: void resizeWindow( const String& winname, int width, int height )
 
 .. ocv:pyfunction:: cv2.resizeWindow(winname, width, height) -> None
 
 .. ocv:cfunction:: void cvResizeWindow( const char* name, int width, int height )
 
-.. ocv:pyoldfunction:: cv.ResizeWindow(name, width, height)-> None
-
     :param winname: Window name
 
     :param width: The new window width
@@ -195,12 +195,12 @@ SetMouseCallback
 ----------------
 Sets mouse handler for the specified window
 
-.. ocv:function:: void setMouseCallback( const string& winname, MouseCallback onMouse, void* userdata=0 )
+.. ocv:function:: void setMouseCallback( const String& winname, MouseCallback onMouse, void* userdata=0 )
+
+.. ocv:pyfunction:: cv2.setMouseCallback(windowName, onMouse [, param]) -> None
 
 .. ocv:cfunction:: void cvSetMouseCallback( const char* window_name, CvMouseCallback on_mouse, void* param=NULL )
 
-.. ocv:pyoldfunction:: cv.SetMouseCallback(windowName, onMouse, param=None) -> None
-
     :param winname: Window name
 
     :param onMouse: Mouse callback. See OpenCV samples, such as  http://code.opencv.org/projects/opencv/repository/revisions/master/entry/samples/cpp/ffilldemo.cpp, on how to specify and use the callback.
@@ -212,14 +212,12 @@ setTrackbarPos
 ------------------
 Sets the trackbar position.
 
-.. ocv:function:: void setTrackbarPos( const string& trackbarname, const string& winname, int pos )
+.. ocv:function:: void setTrackbarPos( const String& trackbarname, const String& winname, int pos )
 
 .. ocv:pyfunction:: cv2.setTrackbarPos(trackbarname, winname, pos) -> None
 
 .. ocv:cfunction:: void cvSetTrackbarPos( const char* trackbar_name, const char* window_name, int pos )
 
-.. ocv:pyoldfunction:: cv.SetTrackbarPos(trackbarName, windowName, pos)-> None
-
     :param trackbarname: Name of the trackbar.
 
     :param winname: Name of the window that is the parent of trackbar.
@@ -242,8 +240,6 @@ Waits for a pressed key.
 
 .. ocv:cfunction:: int cvWaitKey( int delay=0 )
 
-.. ocv:pyoldfunction:: cv.WaitKey(delay=0)-> int
-
     :param delay: Delay in milliseconds. 0 is the special value that means "forever".
 
 The function ``waitKey`` waits for a key event infinitely (when
@@ -256,3 +252,31 @@ The function ``waitKey`` waits for a key event infinitely (when
 .. note::
 
     The function only works if there is at least one HighGUI window created and the window is active. If there are several HighGUI windows, any of them can be active.
+
+setOpenGlDrawCallback
+---------------------
+Set OpenGL render handler for the specified window.
+
+.. ocv:function:: void setOpenGlDrawCallback(const string& winname, OpenGlDrawCallback onOpenGlDraw, void* userdata = 0)
+
+    :param winname: Window name
+
+    :param onOpenGlDraw: Draw callback.
+
+    :param userdata: The optional parameter passed to the callback.
+
+setOpenGlContext
+----------------
+Sets the specified window as current OpenGL context.
+
+.. ocv:function:: void setOpenGlContext(const String& winname)
+
+    :param winname: Window name
+
+updateWindow
+------------
+Force window to redraw its context and call draw callback ( :ocv:func:`setOpenGlDrawCallback` ).
+
+.. ocv:function:: void updateWindow(const String& winname)
+
+    :param winname: Window name
diff --git a/modules/highgui/include/opencv2/highgui.hpp b/modules/highgui/include/opencv2/highgui.hpp
index 0b1c617cd3..c4fc73a81f 100644
--- a/modules/highgui/include/opencv2/highgui.hpp
+++ b/modules/highgui/include/opencv2/highgui.hpp
@@ -44,173 +44,474 @@
 #define __OPENCV_HIGHGUI_HPP__
 
 #include "opencv2/core.hpp"
-#include "opencv2/highgui/highgui_c.h"
 
-#ifdef __cplusplus
-
-struct CvCapture;
-struct CvVideoWriter;
 
+///////////////////////// graphical user interface //////////////////////////
 namespace cv
 {
 
-enum {
-    // Flags for namedWindow
-    WINDOW_NORMAL   = CV_WINDOW_NORMAL,   // the user can resize the window (no constraint) / also use to switch a fullscreen window to a normal size
-    WINDOW_AUTOSIZE = CV_WINDOW_AUTOSIZE, // the user cannot resize the window, the size is constrainted by the image displayed
-    WINDOW_OPENGL   = CV_WINDOW_OPENGL,   // window with opengl support
+// Flags for namedWindow
+enum { WINDOW_NORMAL     = 0x00000000, // the user can resize the window (no constraint) / also use to switch a fullscreen window to a normal size
+       WINDOW_AUTOSIZE   = 0x00000001, // the user cannot resize the window, the size is constrainted by the image displayed
+       WINDOW_OPENGL     = 0x00001000, // window with opengl support
 
-    // Flags for set / getWindowProperty
-    WND_PROP_FULLSCREEN   = CV_WND_PROP_FULLSCREEN,  // fullscreen property
-    WND_PROP_AUTOSIZE     = CV_WND_PROP_AUTOSIZE,    // autosize property
-    WND_PROP_ASPECT_RATIO = CV_WND_PROP_ASPECTRATIO, // window's aspect ration
-    WND_PROP_OPENGL       = CV_WND_PROP_OPENGL       // opengl support
-};
+       WINDOW_FULLSCREEN = 1,          // change the window to fullscreen
+       WINDOW_FREERATIO  = 0x00000100, // the image expends as much as it can (no ratio constraint)
+       WINDOW_KEEPRATIO  = 0x00000000  // the ratio of the image is respected
+     };
+
+// Flags for set / getWindowProperty
+enum { WND_PROP_FULLSCREEN   = 0, // fullscreen property    (can be WINDOW_NORMAL or WINDOW_FULLSCREEN)
+       WND_PROP_AUTOSIZE     = 1, // autosize property      (can be WINDOW_NORMAL or WINDOW_AUTOSIZE)
+       WND_PROP_ASPECT_RATIO = 2, // window's aspect ration (can be set to WINDOW_FREERATIO or WINDOW_KEEPRATIO);
+       WND_PROP_OPENGL       = 3  // opengl support
+     };
+
+enum { EVENT_MOUSEMOVE      = 0,
+       EVENT_LBUTTONDOWN    = 1,
+       EVENT_RBUTTONDOWN    = 2,
+       EVENT_MBUTTONDOWN    = 3,
+       EVENT_LBUTTONUP      = 4,
+       EVENT_RBUTTONUP      = 5,
+       EVENT_MBUTTONUP      = 6,
+       EVENT_LBUTTONDBLCLK  = 7,
+       EVENT_RBUTTONDBLCLK  = 8,
+       EVENT_MBUTTONDBLCLK  = 9
+     };
+
+enum { EVENT_FLAG_LBUTTON   = 1,
+       EVENT_FLAG_RBUTTON   = 2,
+       EVENT_FLAG_MBUTTON   = 4,
+       EVENT_FLAG_CTRLKEY   = 8,
+       EVENT_FLAG_SHIFTKEY  = 16,
+       EVENT_FLAG_ALTKEY    = 32
+     };
+
+// Qt font
+enum {  QT_FONT_LIGHT           = 25, //QFont::Light,
+        QT_FONT_NORMAL          = 50, //QFont::Normal,
+        QT_FONT_DEMIBOLD        = 63, //QFont::DemiBold,
+        QT_FONT_BOLD            = 75, //QFont::Bold,
+        QT_FONT_BLACK           = 87  //QFont::Black
+     };
+
+// Qt font style
+enum {  QT_STYLE_NORMAL         = 0, //QFont::StyleNormal,
+        QT_STYLE_ITALIC         = 1, //QFont::StyleItalic,
+        QT_STYLE_OBLIQUE        = 2  //QFont::StyleOblique
+     };
+
+// Qt "button" type
+enum { QT_PUSH_BUTTON = 0,
+       QT_CHECKBOX    = 1,
+       QT_RADIOBOX    = 2
+     };
+
+
+typedef void (*MouseCallback)(int event, int x, int y, int flags, void* userdata);
+typedef void (*TrackbarCallback)(int pos, void* userdata);
+typedef void (*OpenGlDrawCallback)(void* userdata);
+typedef void (*ButtonCallback)(int state, void* userdata);
+
+
+CV_EXPORTS_W void namedWindow(const String& winname, int flags = WINDOW_AUTOSIZE);
+
+CV_EXPORTS_W void destroyWindow(const String& winname);
 
-CV_EXPORTS_W void namedWindow(const std::string& winname, int flags = WINDOW_AUTOSIZE);
-CV_EXPORTS_W void destroyWindow(const std::string& winname);
 CV_EXPORTS_W void destroyAllWindows();
 
 CV_EXPORTS_W int startWindowThread();
 
 CV_EXPORTS_W int waitKey(int delay = 0);
 
-CV_EXPORTS_W void imshow(const std::string& winname, InputArray mat);
+CV_EXPORTS_W void imshow(const String& winname, InputArray mat);
 
-CV_EXPORTS_W void resizeWindow(const std::string& winname, int width, int height);
-CV_EXPORTS_W void moveWindow(const std::string& winname, int x, int y);
+CV_EXPORTS_W void resizeWindow(const String& winname, int width, int height);
 
-CV_EXPORTS_W void setWindowProperty(const std::string& winname, int prop_id, double prop_value);//YV
-CV_EXPORTS_W double getWindowProperty(const std::string& winname, int prop_id);//YV
+CV_EXPORTS_W void moveWindow(const String& winname, int x, int y);
 
-enum
-{
-    EVENT_MOUSEMOVE      =0,
-    EVENT_LBUTTONDOWN    =1,
-    EVENT_RBUTTONDOWN    =2,
-    EVENT_MBUTTONDOWN    =3,
-    EVENT_LBUTTONUP      =4,
-    EVENT_RBUTTONUP      =5,
-    EVENT_MBUTTONUP      =6,
-    EVENT_LBUTTONDBLCLK  =7,
-    EVENT_RBUTTONDBLCLK  =8,
-    EVENT_MBUTTONDBLCLK  =9
-};
+CV_EXPORTS_W void setWindowProperty(const String& winname, int prop_id, double prop_value);
 
-enum
-{
-    EVENT_FLAG_LBUTTON   =1,
-    EVENT_FLAG_RBUTTON   =2,
-    EVENT_FLAG_MBUTTON   =4,
-    EVENT_FLAG_CTRLKEY   =8,
-    EVENT_FLAG_SHIFTKEY  =16,
-    EVENT_FLAG_ALTKEY    =32
-};
-
-typedef void (*MouseCallback)(int event, int x, int y, int flags, void* userdata);
+CV_EXPORTS_W double getWindowProperty(const String& winname, int prop_id);
 
 //! assigns callback for mouse events
-CV_EXPORTS void setMouseCallback(const std::string& winname, MouseCallback onMouse, void* userdata = 0);
+CV_EXPORTS void setMouseCallback(const String& winname, MouseCallback onMouse, void* userdata = 0);
 
-
-typedef void (CV_CDECL *TrackbarCallback)(int pos, void* userdata);
-
-CV_EXPORTS int createTrackbar(const std::string& trackbarname, const std::string& winname,
+CV_EXPORTS int createTrackbar(const String& trackbarname, const String& winname,
                               int* value, int count,
                               TrackbarCallback onChange = 0,
                               void* userdata = 0);
 
-CV_EXPORTS_W int getTrackbarPos(const std::string& trackbarname, const std::string& winname);
-CV_EXPORTS_W void setTrackbarPos(const std::string& trackbarname, const std::string& winname, int pos);
+CV_EXPORTS_W int getTrackbarPos(const String& trackbarname, const String& winname);
+
+CV_EXPORTS_W void setTrackbarPos(const String& trackbarname, const String& winname, int pos);
+
 
 // OpenGL support
+CV_EXPORTS void imshow(const String& winname, const ogl::Texture2D& tex);
 
-typedef void (*OpenGlDrawCallback)(void* userdata);
-CV_EXPORTS void setOpenGlDrawCallback(const std::string& winname, OpenGlDrawCallback onOpenGlDraw, void* userdata = 0);
+CV_EXPORTS void setOpenGlDrawCallback(const String& winname, OpenGlDrawCallback onOpenGlDraw, void* userdata = 0);
 
-CV_EXPORTS void setOpenGlContext(const std::string& winname);
+CV_EXPORTS void setOpenGlContext(const String& winname);
 
-CV_EXPORTS void updateWindow(const std::string& winname);
+CV_EXPORTS void updateWindow(const String& winname);
 
-//Only for Qt
 
-CV_EXPORTS CvFont fontQt(const std::string& nameFont, int pointSize=-1,
-                         Scalar color=Scalar::all(0), int weight=CV_FONT_NORMAL,
-                         int style=CV_STYLE_NORMAL, int spacing=0);
-CV_EXPORTS void addText( const Mat& img, const std::string& text, Point org, CvFont font);
+// Only for Qt
 
-CV_EXPORTS void displayOverlay(const std::string& winname, const std::string& text, int delayms CV_DEFAULT(0));
-CV_EXPORTS void displayStatusBar(const std::string& winname, const std::string& text, int delayms CV_DEFAULT(0));
+struct QtFont
+{
+    const char* nameFont;  // Qt: nameFont
+    Scalar      color;     // Qt: ColorFont -> cvScalar(blue_component, green_component, red\_component[, alpha_component])
+    int         font_face; // Qt: bool italic
+    const int*  ascii;     // font data and metrics
+    const int*  greek;
+    const int*  cyrillic;
+    float       hscale, vscale;
+    float       shear;     // slope coefficient: 0 - normal, >0 - italic
+    int         thickness; // Qt: weight
+    float       dx;        // horizontal interval between letters
+    int         line_type; // Qt: PointSize
+};
+
+CV_EXPORTS QtFont fontQt(const String& nameFont, int pointSize = -1,
+                         Scalar color = Scalar::all(0), int weight = QT_FONT_NORMAL,
+                         int style = QT_STYLE_NORMAL, int spacing = 0);
+
+CV_EXPORTS void addText( const Mat& img, const String& text, Point org, const QtFont& font);
+
+CV_EXPORTS void displayOverlay(const String& winname, const String& text, int delayms = 0);
+
+CV_EXPORTS void displayStatusBar(const String& winname, const String& text, int delayms = 0);
+
+CV_EXPORTS void saveWindowParameters(const String& windowName);
+
+CV_EXPORTS void loadWindowParameters(const String& windowName);
 
-CV_EXPORTS void saveWindowParameters(const std::string& windowName);
-CV_EXPORTS void loadWindowParameters(const std::string& windowName);
 CV_EXPORTS  int startLoop(int (*pt2Func)(int argc, char *argv[]), int argc, char* argv[]);
+
 CV_EXPORTS  void stopLoop();
 
-typedef void (CV_CDECL *ButtonCallback)(int state, void* userdata);
-CV_EXPORTS int createButton( const std::string& bar_name, ButtonCallback on_change,
-                             void* userdata=NULL, int type=CV_PUSH_BUTTON,
-                             bool initial_button_state=0);
+CV_EXPORTS int createButton( const String& bar_name, ButtonCallback on_change,
+                             void* userdata = 0, int type = QT_PUSH_BUTTON,
+                             bool initial_button_state = false);
 
-//-------------------------
+} // cv
 
-enum
+
+
+//////////////////////////////// image codec ////////////////////////////////
+namespace cv
 {
-    // 8bit, color or not
-    IMREAD_UNCHANGED  =-1,
-    // 8bit, gray
-    IMREAD_GRAYSCALE  =0,
-    // ?, color
-    IMREAD_COLOR      =1,
-    // any depth, ?
-    IMREAD_ANYDEPTH   =2,
-    // ?, any color
-    IMREAD_ANYCOLOR   =4
-};
 
-enum
-{
-    IMWRITE_JPEG_QUALITY =1,
-    IMWRITE_PNG_COMPRESSION =16,
-    IMWRITE_PNG_STRATEGY =17,
-    IMWRITE_PNG_BILEVEL =18,
-    IMWRITE_PNG_STRATEGY_DEFAULT =0,
-    IMWRITE_PNG_STRATEGY_FILTERED =1,
-    IMWRITE_PNG_STRATEGY_HUFFMAN_ONLY =2,
-    IMWRITE_PNG_STRATEGY_RLE =3,
-    IMWRITE_PNG_STRATEGY_FIXED =4,
-    IMWRITE_PXM_BINARY =32
-};
+enum { IMREAD_UNCHANGED  = -1, // 8bit, color or not
+       IMREAD_GRAYSCALE  = 0,  // 8bit, gray
+       IMREAD_COLOR      = 1,  // ?, color
+       IMREAD_ANYDEPTH   = 2,  // any depth, ?
+       IMREAD_ANYCOLOR   = 4   // ?, any color
+     };
+
+enum { IMWRITE_JPEG_QUALITY    = 1,
+       IMWRITE_PNG_COMPRESSION = 16,
+       IMWRITE_PNG_STRATEGY    = 17,
+       IMWRITE_PNG_BILEVEL     = 18,
+       IMWRITE_PXM_BINARY      = 32,
+       IMWRITE_WEBP_QUALITY    = 64
+     };
+
+enum { IMWRITE_PNG_STRATEGY_DEFAULT      = 0,
+       IMWRITE_PNG_STRATEGY_FILTERED     = 1,
+       IMWRITE_PNG_STRATEGY_HUFFMAN_ONLY = 2,
+       IMWRITE_PNG_STRATEGY_RLE          = 3,
+       IMWRITE_PNG_STRATEGY_FIXED        = 4
+     };
+
+CV_EXPORTS_W Mat imread( const String& filename, int flags = IMREAD_COLOR );
+
+CV_EXPORTS_W bool imwrite( const String& filename, InputArray img,
+              const std::vector<int>& params = std::vector<int>());
 
-CV_EXPORTS_W Mat imread( const std::string& filename, int flags=1 );
-CV_EXPORTS_W bool imwrite( const std::string& filename, InputArray img,
-              const std::vector<int>& params=std::vector<int>());
 CV_EXPORTS_W Mat imdecode( InputArray buf, int flags );
-CV_EXPORTS Mat imdecode( InputArray buf, int flags, Mat* dst );
-CV_EXPORTS_W bool imencode( const std::string& ext, InputArray img,
+
+CV_EXPORTS Mat imdecode( InputArray buf, int flags, Mat* dst);
+
+CV_EXPORTS_W bool imencode( const String& ext, InputArray img,
                             CV_OUT std::vector<uchar>& buf,
-                            const std::vector<int>& params=std::vector<int>());
+                            const std::vector<int>& params = std::vector<int>());
 
-#ifndef CV_NO_VIDEO_CAPTURE_CPP_API
+} // cv
+
+
+
+////////////////////////////////// video io /////////////////////////////////
+
+typedef struct CvCapture CvCapture;
+typedef struct CvVideoWriter CvVideoWriter;
+
+namespace cv
+{
+
+// Camera API
+enum { CAP_ANY          = 0,     // autodetect
+       CAP_VFW          = 200,   // platform native
+       CAP_V4L          = 200,
+       CAP_V4L2         = CAP_V4L,
+       CAP_FIREWARE     = 300,   // IEEE 1394 drivers
+       CAP_FIREWIRE     = CAP_FIREWARE,
+       CAP_IEEE1394     = CAP_FIREWARE,
+       CAP_DC1394       = CAP_FIREWARE,
+       CAP_CMU1394      = CAP_FIREWARE,
+       CAP_QT           = 500,   // QuickTime
+       CAP_UNICAP       = 600,   // Unicap drivers
+       CAP_DSHOW        = 700,   // DirectShow (via videoInput)
+       CAP_PVAPI        = 800,   // PvAPI, Prosilica GigE SDK
+       CAP_OPENNI       = 900,   // OpenNI (for Kinect)
+       CAP_OPENNI_ASUS  = 910,   // OpenNI (for Asus Xtion)
+       CAP_ANDROID      = 1000,  // Android
+       CAP_XIAPI        = 1100,  // XIMEA Camera API
+       CAP_AVFOUNDATION = 1200,  // AVFoundation framework for iOS (OS X Lion will have the same API)
+       CAP_GIGANETIX    = 1300,  // Smartek Giganetix GigEVisionSDK
+       CAP_MSMF         = 1400   // Microsoft Media Foundation (via videoInput)
+     };
+
+// generic properties (based on DC1394 properties)
+enum { CAP_PROP_POS_MSEC       =0,
+       CAP_PROP_POS_FRAMES     =1,
+       CAP_PROP_POS_AVI_RATIO  =2,
+       CAP_PROP_FRAME_WIDTH    =3,
+       CAP_PROP_FRAME_HEIGHT   =4,
+       CAP_PROP_FPS            =5,
+       CAP_PROP_FOURCC         =6,
+       CAP_PROP_FRAME_COUNT    =7,
+       CAP_PROP_FORMAT         =8,
+       CAP_PROP_MODE           =9,
+       CAP_PROP_BRIGHTNESS    =10,
+       CAP_PROP_CONTRAST      =11,
+       CAP_PROP_SATURATION    =12,
+       CAP_PROP_HUE           =13,
+       CAP_PROP_GAIN          =14,
+       CAP_PROP_EXPOSURE      =15,
+       CAP_PROP_CONVERT_RGB   =16,
+       CAP_PROP_WHITE_BALANCE_BLUE_U =17,
+       CAP_PROP_RECTIFICATION =18,
+       CAP_PROP_MONOCROME     =19,
+       CAP_PROP_SHARPNESS     =20,
+       CAP_PROP_AUTO_EXPOSURE =21, // DC1394: exposure control done by camera, user can adjust refernce level using this feature
+       CAP_PROP_GAMMA         =22,
+       CAP_PROP_TEMPERATURE   =23,
+       CAP_PROP_TRIGGER       =24,
+       CAP_PROP_TRIGGER_DELAY =25,
+       CAP_PROP_WHITE_BALANCE_RED_V =26,
+       CAP_PROP_ZOOM          =27,
+       CAP_PROP_FOCUS         =28,
+       CAP_PROP_GUID          =29,
+       CAP_PROP_ISO_SPEED     =30,
+       CAP_PROP_BACKLIGHT     =32,
+       CAP_PROP_PAN           =33,
+       CAP_PROP_TILT          =34,
+       CAP_PROP_ROLL          =35,
+       CAP_PROP_IRIS          =36,
+       CAP_PROP_SETTINGS      =37
+     };
+
+
+// DC1394 only
+// modes of the controlling registers (can be: auto, manual, auto single push, absolute Latter allowed with any other mode)
+// every feature can have only one mode turned on at a time
+enum { CAP_PROP_DC1394_OFF                = -4, //turn the feature off (not controlled manually nor automatically)
+       CAP_PROP_DC1394_MODE_MANUAL        = -3, //set automatically when a value of the feature is set by the user
+       CAP_PROP_DC1394_MODE_AUTO          = -2,
+       CAP_PROP_DC1394_MODE_ONE_PUSH_AUTO = -1,
+       CAP_PROP_DC1394_MAX                = 31
+     };
+
+
+// OpenNI map generators
+enum { CAP_OPENNI_DEPTH_GENERATOR = 1 << 31,
+       CAP_OPENNI_IMAGE_GENERATOR = 1 << 30,
+       CAP_OPENNI_GENERATORS_MASK = CAP_OPENNI_DEPTH_GENERATOR + CAP_OPENNI_IMAGE_GENERATOR
+     };
+
+// Properties of cameras available through OpenNI interfaces
+enum { CAP_PROP_OPENNI_OUTPUT_MODE       = 100,
+       CAP_PROP_OPENNI_FRAME_MAX_DEPTH   = 101, // in mm
+       CAP_PROP_OPENNI_BASELINE          = 102, // in mm
+       CAP_PROP_OPENNI_FOCAL_LENGTH      = 103, // in pixels
+       CAP_PROP_OPENNI_REGISTRATION      = 104, // flag that synchronizes the remapping depth map to image map
+                                                // by changing depth generator's view point (if the flag is "on") or
+                                                // sets this view point to its normal one (if the flag is "off").
+       CAP_PROP_OPENNI_REGISTRATION_ON   = CAP_PROP_OPENNI_REGISTRATION,
+       CAP_PROP_OPENNI_APPROX_FRAME_SYNC = 105,
+       CAP_PROP_OPENNI_MAX_BUFFER_SIZE   = 106,
+       CAP_PROP_OPENNI_CIRCLE_BUFFER     = 107,
+       CAP_PROP_OPENNI_MAX_TIME_DURATION = 108,
+       CAP_PROP_OPENNI_GENERATOR_PRESENT = 109
+     };
+
+// OpenNI shortcats
+enum { CAP_OPENNI_IMAGE_GENERATOR_PRESENT         = CAP_OPENNI_IMAGE_GENERATOR + CAP_PROP_OPENNI_GENERATOR_PRESENT,
+       CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE     = CAP_OPENNI_IMAGE_GENERATOR + CAP_PROP_OPENNI_OUTPUT_MODE,
+       CAP_OPENNI_DEPTH_GENERATOR_BASELINE        = CAP_OPENNI_DEPTH_GENERATOR + CAP_PROP_OPENNI_BASELINE,
+       CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH    = CAP_OPENNI_DEPTH_GENERATOR + CAP_PROP_OPENNI_FOCAL_LENGTH,
+       CAP_OPENNI_DEPTH_GENERATOR_REGISTRATION    = CAP_OPENNI_DEPTH_GENERATOR + CAP_PROP_OPENNI_REGISTRATION,
+       CAP_OPENNI_DEPTH_GENERATOR_REGISTRATION_ON = CAP_OPENNI_DEPTH_GENERATOR_REGISTRATION
+     };
+
+// OpenNI data given from depth generator
+enum { CAP_OPENNI_DEPTH_MAP         = 0, // Depth values in mm (CV_16UC1)
+       CAP_OPENNI_POINT_CLOUD_MAP   = 1, // XYZ in meters (CV_32FC3)
+       CAP_OPENNI_DISPARITY_MAP     = 2, // Disparity in pixels (CV_8UC1)
+       CAP_OPENNI_DISPARITY_MAP_32F = 3, // Disparity in pixels (CV_32FC1)
+       CAP_OPENNI_VALID_DEPTH_MASK  = 4, // CV_8UC1
+
+       // Data given from RGB image generator
+       CAP_OPENNI_BGR_IMAGE         = 5,
+       CAP_OPENNI_GRAY_IMAGE        = 6
+     };
+
+// Supported output modes of OpenNI image generator
+enum { CAP_OPENNI_VGA_30HZ  = 0,
+       CAP_OPENNI_SXGA_15HZ = 1,
+       CAP_OPENNI_SXGA_30HZ = 2,
+       CAP_OPENNI_QVGA_30HZ = 3,
+       CAP_OPENNI_QVGA_60HZ = 4
+     };
+
+
+// GStreamer
+enum { CAP_PROP_GSTREAMER_QUEUE_LENGTH = 200 // default is 1
+     };
+
+
+// PVAPI
+enum { CAP_PROP_PVAPI_MULTICASTIP = 300 // ip for anable multicast master mode. 0 for disable multicast
+     };
+
+
+// Properties of cameras available through XIMEA SDK interface
+enum { CAP_PROP_XI_DOWNSAMPLING  = 400, // Change image resolution by binning or skipping.
+       CAP_PROP_XI_DATA_FORMAT   = 401, // Output data format.
+       CAP_PROP_XI_OFFSET_X      = 402, // Horizontal offset from the origin to the area of interest (in pixels).
+       CAP_PROP_XI_OFFSET_Y      = 403, // Vertical offset from the origin to the area of interest (in pixels).
+       CAP_PROP_XI_TRG_SOURCE    = 404, // Defines source of trigger.
+       CAP_PROP_XI_TRG_SOFTWARE  = 405, // Generates an internal trigger. PRM_TRG_SOURCE must be set to TRG_SOFTWARE.
+       CAP_PROP_XI_GPI_SELECTOR  = 406, // Selects general purpose input
+       CAP_PROP_XI_GPI_MODE      = 407, // Set general purpose input mode
+       CAP_PROP_XI_GPI_LEVEL     = 408, // Get general purpose level
+       CAP_PROP_XI_GPO_SELECTOR  = 409, // Selects general purpose output
+       CAP_PROP_XI_GPO_MODE      = 410, // Set general purpose output mode
+       CAP_PROP_XI_LED_SELECTOR  = 411, // Selects camera signalling LED
+       CAP_PROP_XI_LED_MODE      = 412, // Define camera signalling LED functionality
+       CAP_PROP_XI_MANUAL_WB     = 413, // Calculates White Balance(must be called during acquisition)
+       CAP_PROP_XI_AUTO_WB       = 414, // Automatic white balance
+       CAP_PROP_XI_AEAG          = 415, // Automatic exposure/gain
+       CAP_PROP_XI_EXP_PRIORITY  = 416, // Exposure priority (0.5 - exposure 50%, gain 50%).
+       CAP_PROP_XI_AE_MAX_LIMIT  = 417, // Maximum limit of exposure in AEAG procedure
+       CAP_PROP_XI_AG_MAX_LIMIT  = 418, // Maximum limit of gain in AEAG procedure
+       CAP_PROP_XI_AEAG_LEVEL    = 419, // Average intensity of output signal AEAG should achieve(in %)
+       CAP_PROP_XI_TIMEOUT       = 420  // Image capture timeout in milliseconds
+     };
+
+
+// Properties for Android cameras
+enum { CAP_PROP_ANDROID_AUTOGRAB               = 1024,
+       CAP_PROP_ANDROID_PREVIEW_SIZES_STRING   = 1025, // readonly, tricky property, returns const char* indeed
+       CAP_PROP_ANDROID_PREVIEW_FORMAT         = 1026, // readonly, tricky property, returns const char* indeed
+       CAP_PROP_ANDROID_FLASH_MODE             = 8001,
+       CAP_PROP_ANDROID_FOCUS_MODE             = 8002,
+       CAP_PROP_ANDROID_WHITE_BALANCE          = 8003,
+       CAP_PROP_ANDROID_ANTIBANDING            = 8004,
+       CAP_PROP_ANDROID_FOCAL_LENGTH           = 8005,
+       CAP_PROP_ANDROID_FOCUS_DISTANCE_NEAR    = 8006,
+       CAP_PROP_ANDROID_FOCUS_DISTANCE_OPTIMAL = 8007,
+       CAP_PROP_ANDROID_FOCUS_DISTANCE_FAR     = 8008
+     };
+
+
+// Android camera output formats
+enum { CAP_ANDROID_COLOR_FRAME_BGR  = 0, //BGR
+       CAP_ANDROID_COLOR_FRAME      = CAP_ANDROID_COLOR_FRAME_BGR,
+       CAP_ANDROID_GREY_FRAME       = 1,  //Y
+       CAP_ANDROID_COLOR_FRAME_RGB  = 2,
+       CAP_ANDROID_COLOR_FRAME_BGRA = 3,
+       CAP_ANDROID_COLOR_FRAME_RGBA = 4
+     };
+
+
+// Android camera flash modes
+enum { CAP_ANDROID_FLASH_MODE_AUTO     = 0,
+       CAP_ANDROID_FLASH_MODE_OFF      = 1,
+       CAP_ANDROID_FLASH_MODE_ON       = 2,
+       CAP_ANDROID_FLASH_MODE_RED_EYE  = 3,
+       CAP_ANDROID_FLASH_MODE_TORCH    = 4
+     };
+
+
+// Android camera focus modes
+enum { CAP_ANDROID_FOCUS_MODE_AUTO             = 0,
+       CAP_ANDROID_FOCUS_MODE_CONTINUOUS_VIDEO = 1,
+       CAP_ANDROID_FOCUS_MODE_EDOF             = 2,
+       CAP_ANDROID_FOCUS_MODE_FIXED            = 3,
+       CAP_ANDROID_FOCUS_MODE_INFINITY         = 4,
+       CAP_ANDROID_FOCUS_MODE_MACRO            = 5
+     };
+
+
+// Android camera white balance modes
+enum { CAP_ANDROID_WHITE_BALANCE_AUTO             = 0,
+       CAP_ANDROID_WHITE_BALANCE_CLOUDY_DAYLIGHT  = 1,
+       CAP_ANDROID_WHITE_BALANCE_DAYLIGHT         = 2,
+       CAP_ANDROID_WHITE_BALANCE_FLUORESCENT      = 3,
+       CAP_ANDROID_WHITE_BALANCE_INCANDESCENT     = 4,
+       CAP_ANDROID_WHITE_BALANCE_SHADE            = 5,
+       CAP_ANDROID_WHITE_BALANCE_TWILIGHT         = 6,
+       CAP_ANDROID_WHITE_BALANCE_WARM_FLUORESCENT = 7
+     };
+
+
+// Android camera antibanding modes
+enum { CAP_ANDROID_ANTIBANDING_50HZ = 0,
+       CAP_ANDROID_ANTIBANDING_60HZ = 1,
+       CAP_ANDROID_ANTIBANDING_AUTO = 2,
+       CAP_ANDROID_ANTIBANDING_OFF  = 3
+     };
+
+
+// Properties of cameras available through AVFOUNDATION interface
+enum { CAP_PROP_IOS_DEVICE_FOCUS        = 9001,
+       CAP_PROP_IOS_DEVICE_EXPOSURE     = 9002,
+       CAP_PROP_IOS_DEVICE_FLASH        = 9003,
+       CAP_PROP_IOS_DEVICE_WHITEBALANCE = 9004,
+       CAP_PROP_IOS_DEVICE_TORCH        = 9005
+     };
+
+
+// Properties of cameras available through Smartek Giganetix Ethernet Vision interface
+/* --- Vladimir Litvinenko (litvinenko.vladimir@gmail.com) --- */
+enum { CAP_PROP_GIGA_FRAME_OFFSET_X   = 10001,
+       CAP_PROP_GIGA_FRAME_OFFSET_Y   = 10002,
+       CAP_PROP_GIGA_FRAME_WIDTH_MAX  = 10003,
+       CAP_PROP_GIGA_FRAME_HEIGH_MAX  = 10004,
+       CAP_PROP_GIGA_FRAME_SENS_WIDTH = 10005,
+       CAP_PROP_GIGA_FRAME_SENS_HEIGH = 10006
+     };
 
-template<> void CV_EXPORTS Ptr<CvCapture>::delete_obj();
-template<> void CV_EXPORTS Ptr<CvVideoWriter>::delete_obj();
 
 class CV_EXPORTS_W VideoCapture
 {
 public:
     CV_WRAP VideoCapture();
-    CV_WRAP VideoCapture(const std::string& filename);
+    CV_WRAP VideoCapture(const String& filename);
     CV_WRAP VideoCapture(int device);
 
     virtual ~VideoCapture();
-    CV_WRAP virtual bool open(const std::string& filename);
+    CV_WRAP virtual bool open(const String& filename);
     CV_WRAP virtual bool open(int device);
     CV_WRAP virtual bool isOpened() const;
     CV_WRAP virtual void release();
 
     CV_WRAP virtual bool grab();
-    CV_WRAP virtual bool retrieve(CV_OUT Mat& image, int channel=0);
+    CV_WRAP virtual bool retrieve(CV_OUT Mat& image, int flag = 0);
     virtual VideoCapture& operator >> (CV_OUT Mat& image);
     CV_WRAP virtual bool read(CV_OUT Mat& image);
 
@@ -226,25 +527,26 @@ class CV_EXPORTS_W VideoWriter
 {
 public:
     CV_WRAP VideoWriter();
-    CV_WRAP VideoWriter(const std::string& filename, int fourcc, double fps,
-                Size frameSize, bool isColor=true);
+    CV_WRAP VideoWriter(const String& filename, int fourcc, double fps,
+                Size frameSize, bool isColor = true);
 
     virtual ~VideoWriter();
-    CV_WRAP virtual bool open(const std::string& filename, int fourcc, double fps,
-                      Size frameSize, bool isColor=true);
+    CV_WRAP virtual bool open(const String& filename, int fourcc, double fps,
+                      Size frameSize, bool isColor = true);
     CV_WRAP virtual bool isOpened() const;
     CV_WRAP virtual void release();
     virtual VideoWriter& operator << (const Mat& image);
     CV_WRAP virtual void write(const Mat& image);
 
+    CV_WRAP static int fourcc(char c1, char c2, char c3, char c4);
+
 protected:
     Ptr<CvVideoWriter> writer;
 };
 
-#endif
+template<> CV_EXPORTS void DefaultDeleter<CvCapture>::operator ()(CvCapture* obj) const;
+template<> CV_EXPORTS void DefaultDeleter<CvVideoWriter>::operator ()(CvVideoWriter* obj) const;
 
-}
-
-#endif
+} // cv
 
 #endif
diff --git a/modules/highgui/include/opencv2/highgui/cap_ios.h b/modules/highgui/include/opencv2/highgui/cap_ios.h
index 97d9a08af6..4d270aba9e 100644
--- a/modules/highgui/include/opencv2/highgui/cap_ios.h
+++ b/modules/highgui/include/opencv2/highgui/cap_ios.h
@@ -1,6 +1,4 @@
-/*
- *  cap_ios.h
- *  For iOS video I/O
+/*  For iOS video I/O
  *  by Eduard Feicho on 29/07/12
  *  Copyright 2012. All rights reserved.
  *
@@ -90,6 +88,12 @@
 - (void)createVideoPreviewLayer;
 - (void)updateOrientation;
 
+- (void)lockFocus;
+- (void)unlockFocus;
+- (void)lockExposure;
+- (void)unlockExposure;
+- (void)lockBalance;
+- (void)unlockBalance;
 
 @end
 
@@ -116,6 +120,7 @@
     BOOL grayscaleMode;
 
     BOOL recordVideo;
+    BOOL rotateVideo;
     AVAssetWriterInput* recordAssetWriterInput;
     AVAssetWriterInputPixelBufferAdaptor* recordPixelBufferAdaptor;
     AVAssetWriter* recordAssetWriter;
@@ -128,6 +133,7 @@
 @property (nonatomic, assign) BOOL grayscaleMode;
 
 @property (nonatomic, assign) BOOL recordVideo;
+@property (nonatomic, assign) BOOL rotateVideo;
 @property (nonatomic, retain) AVAssetWriterInput* recordAssetWriterInput;
 @property (nonatomic, retain) AVAssetWriterInputPixelBufferAdaptor* recordPixelBufferAdaptor;
 @property (nonatomic, retain) AVAssetWriter* recordAssetWriter;
diff --git a/modules/highgui/include/opencv2/highgui/highgui.hpp b/modules/highgui/include/opencv2/highgui/highgui.hpp
index cb84ef620b..160c9cf4af 100644
--- a/modules/highgui/include/opencv2/highgui/highgui.hpp
+++ b/modules/highgui/include/opencv2/highgui/highgui.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/highgui.hpp"
\ No newline at end of file
+#include "opencv2/highgui.hpp"
diff --git a/modules/highgui/include/opencv2/highgui/highgui_c.h b/modules/highgui/include/opencv2/highgui/highgui_c.h
index 116e8c1f45..2ebea0d30b 100644
--- a/modules/highgui/include/opencv2/highgui/highgui_c.h
+++ b/modules/highgui/include/opencv2/highgui/highgui_c.h
@@ -298,6 +298,7 @@ enum
     CV_CAP_UNICAP   =600,   // Unicap drivers
 
     CV_CAP_DSHOW    =700,   // DirectShow (via videoInput)
+    CV_CAP_MSMF     =1400,  // Microsoft Media Foundation (via videoInput)
 
     CV_CAP_PVAPI    =800,   // PvAPI, Prosilica GigE SDK
 
@@ -305,6 +306,8 @@ enum
     CV_CAP_OPENNI_ASUS =910,   // OpenNI (for Asus Xtion)
 
     CV_CAP_ANDROID  =1000,  // Android
+    CV_CAP_ANDROID_BACK =CV_CAP_ANDROID+99, // Android back camera
+    CV_CAP_ANDROID_FRONT =CV_CAP_ANDROID+98, // Android front camera
 
     CV_CAP_XIAPI    =1100,   // XIMEA Camera API
 
@@ -556,9 +559,11 @@ CVAPI(int)    cvGetCaptureDomain( CvCapture* capture);
 /* "black box" video file writer structure */
 typedef struct CvVideoWriter CvVideoWriter;
 
+#define CV_FOURCC_MACRO(c1, c2, c3, c4) (((c1) & 255) + (((c2) & 255) << 8) + (((c3) & 255) << 16) + (((c4) & 255) << 24))
+
 CV_INLINE int CV_FOURCC(char c1, char c2, char c3, char c4)
 {
-    return (c1 & 255) + ((c2 & 255) << 8) + ((c3 & 255) << 16) + ((c4 & 255) << 24);
+    return CV_FOURCC_MACRO(c1, c2, c3, c4);
 }
 
 #define CV_FOURCC_PROMPT -1  /* Open Codec Selection Dialog (Windows only) */
@@ -569,9 +574,6 @@ CVAPI(CvVideoWriter*) cvCreateVideoWriter( const char* filename, int fourcc,
                                            double fps, CvSize frame_size,
                                            int is_color CV_DEFAULT(1));
 
-//CVAPI(CvVideoWriter*) cvCreateImageSequenceWriter( const char* filename,
-//                                                   int is_color CV_DEFAULT(1));
-
 /* write frame to video file */
 CVAPI(int) cvWriteFrame( CvVideoWriter* writer, const IplImage* image );
 
diff --git a/modules/highgui/include/opencv2/highgui/ios.h b/modules/highgui/include/opencv2/highgui/ios.h
new file mode 100644
index 0000000000..a7f0395d71
--- /dev/null
+++ b/modules/highgui/include/opencv2/highgui/ios.h
@@ -0,0 +1,49 @@
+
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/core/core.hpp"
+#import "opencv2/highgui/cap_ios.h"
+
+UIImage* MatToUIImage(const cv::Mat& image);
+void UIImageToMat(const UIImage* image,
+                         cv::Mat& m, bool alphaExist = false);
diff --git a/modules/highgui/perf/perf_input.cpp b/modules/highgui/perf/perf_input.cpp
index b32f88da97..10ce30a83c 100644
--- a/modules/highgui/perf/perf_input.cpp
+++ b/modules/highgui/perf/perf_input.cpp
@@ -11,11 +11,21 @@ using std::tr1::get;
 
 typedef perf::TestBaseWithParam<std::string> VideoCapture_Reading;
 
+#if defined(HAVE_MSMF)
+// MPEG2 is not supported by Media Foundation yet
+// http://social.msdn.microsoft.com/Forums/en-US/mediafoundationdevelopment/thread/39a36231-8c01-40af-9af5-3c105d684429
+PERF_TEST_P(VideoCapture_Reading, ReadFile, testing::Values( "highgui/video/big_buck_bunny.avi",
+                                               "highgui/video/big_buck_bunny.mov",
+                                               "highgui/video/big_buck_bunny.mp4",
+                                               "highgui/video/big_buck_bunny.wmv" ) )
+
+#else
 PERF_TEST_P(VideoCapture_Reading, ReadFile, testing::Values( "highgui/video/big_buck_bunny.avi",
                                                "highgui/video/big_buck_bunny.mov",
                                                "highgui/video/big_buck_bunny.mp4",
                                                "highgui/video/big_buck_bunny.mpg",
                                                "highgui/video/big_buck_bunny.wmv" ) )
+#endif
 {
   string filename = getDataPath(GetParam());
 
@@ -27,4 +37,4 @@ PERF_TEST_P(VideoCapture_Reading, ReadFile, testing::Values( "highgui/video/big_
   SANITY_CHECK(dummy);
 }
 
-#endif // BUILD_WITH_VIDEO_INPUT_SUPPORT
\ No newline at end of file
+#endif // BUILD_WITH_VIDEO_INPUT_SUPPORT
diff --git a/modules/highgui/perf/perf_output.cpp b/modules/highgui/perf/perf_output.cpp
index 50d86adf78..8e7ec54a97 100644
--- a/modules/highgui/perf/perf_output.cpp
+++ b/modules/highgui/perf/perf_output.cpp
@@ -22,13 +22,19 @@ PERF_TEST_P(VideoWriter_Writing, WriteFrame,
 {
   string filename = getDataPath(get<0>(GetParam()));
   bool isColor = get<1>(GetParam());
+  Mat image = imread(filename, 1);
+#if defined(HAVE_MSMF) && !defined(HAVE_VFW) && !defined(HAVE_FFMPEG) // VFW has greater priority
+  VideoWriter writer(cv::tempfile(".wmv"), VideoWriter::fourcc('W', 'M', 'V', '3'),
+                            25, cv::Size(image.cols, image.rows), isColor);
+#else
+  VideoWriter writer(cv::tempfile(".avi"), VideoWriter::fourcc('X', 'V', 'I', 'D'),
+                            25, cv::Size(image.cols, image.rows), isColor);
+#endif
 
-  VideoWriter writer(cv::tempfile(".avi"), CV_FOURCC('X', 'V', 'I', 'D'), 25, cv::Size(640, 480), isColor);
-
-  TEST_CYCLE() { Mat image = imread(filename, 1); writer << image; }
+  TEST_CYCLE() { image = imread(filename, 1); writer << image; }
 
   bool dummy = writer.isOpened();
   SANITY_CHECK(dummy);
 }
 
-#endif // BUILD_WITH_VIDEO_OUTPUT_SUPPORT
\ No newline at end of file
+#endif // BUILD_WITH_VIDEO_OUTPUT_SUPPORT
diff --git a/modules/highgui/perf/perf_precomp.cpp b/modules/highgui/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/highgui/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/highgui/perf/perf_precomp.hpp b/modules/highgui/perf/perf_precomp.hpp
index 30e9b7ff6d..faf34617e3 100644
--- a/modules/highgui/perf/perf_precomp.hpp
+++ b/modules/highgui/perf/perf_precomp.hpp
@@ -19,10 +19,12 @@
 #if defined(HAVE_XINE)         || \
     defined(HAVE_GSTREAMER)    || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_AVFOUNDATION) || \
-    /*defined(HAVE_OPENNI)     || too specialized */ \
     defined(HAVE_FFMPEG)       || \
-    defined(WIN32) /* assume that we have ffmpeg */
+    defined(HAVE_MSMF)         || \
+    defined(HAVE_VFW)
+    /*defined(HAVE_OPENNI) too specialized */ \
 
 #  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
 #else
@@ -32,9 +34,11 @@
 #if /*defined(HAVE_XINE)       || */\
     defined(HAVE_GSTREAMER)    || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_AVFOUNDATION) || \
     defined(HAVE_FFMPEG)       || \
-    defined(WIN32) /* assume that we have ffmpeg */
+    defined(HAVE_MSMF)         || \
+    defined(HAVE_VFW)
 #  define BUILD_WITH_VIDEO_OUTPUT_SUPPORT 1
 #else
 #  define BUILD_WITH_VIDEO_OUTPUT_SUPPORT 0
diff --git a/modules/highgui/src/bitstrm.cpp b/modules/highgui/src/bitstrm.cpp
index 89f46fb59a..c47744bb30 100644
--- a/modules/highgui/src/bitstrm.cpp
+++ b/modules/highgui/src/bitstrm.cpp
@@ -110,7 +110,7 @@ void  RBaseStream::readBlock()
 }
 
 
-bool  RBaseStream::open( const std::string& filename )
+bool  RBaseStream::open( const String& filename )
 {
     close();
     allocate();
@@ -388,7 +388,7 @@ void  WBaseStream::writeBlock()
 }
 
 
-bool  WBaseStream::open( const std::string& filename )
+bool  WBaseStream::open( const String& filename )
 {
     close();
     allocate();
diff --git a/modules/highgui/src/bitstrm.hpp b/modules/highgui/src/bitstrm.hpp
index 221c2c709b..df72e73732 100644
--- a/modules/highgui/src/bitstrm.hpp
+++ b/modules/highgui/src/bitstrm.hpp
@@ -66,7 +66,7 @@ public:
     RBaseStream();
     virtual ~RBaseStream();
 
-    virtual bool  open( const std::string& filename );
+    virtual bool  open( const String& filename );
     virtual bool  open( const Mat& buf );
     virtual void  close();
     bool          isOpened();
@@ -123,7 +123,7 @@ public:
     WBaseStream();
     virtual ~WBaseStream();
 
-    virtual bool  open( const std::string& filename );
+    virtual bool  open( const String& filename );
     virtual bool  open( std::vector<uchar>& buf );
     virtual void  close();
     bool          isOpened();
diff --git a/modules/highgui/src/cap.cpp b/modules/highgui/src/cap.cpp
index 012f712dc7..be62ce9d38 100644
--- a/modules/highgui/src/cap.cpp
+++ b/modules/highgui/src/cap.cpp
@@ -49,10 +49,10 @@
 namespace cv
 {
 
-template<> void Ptr<CvCapture>::delete_obj()
+template<> void DefaultDeleter<CvCapture>::operator ()(CvCapture* obj) const
 { cvReleaseCapture(&obj); }
 
-template<> void Ptr<CvVideoWriter>::delete_obj()
+template<> void DefaultDeleter<CvVideoWriter>::operator ()(CvVideoWriter* obj) const
 { cvReleaseVideoWriter(&obj); }
 
 }
@@ -114,9 +114,12 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
 {
     int  domains[] =
     {
-#ifdef HAVE_VIDEOINPUT
+#ifdef HAVE_DSHOW
         CV_CAP_DSHOW,
 #endif
+#ifdef HAVE_MSMF
+        CV_CAP_MSMF,
+#endif
 #if 1
         CV_CAP_IEEE1394,   // identical to CV_CAP_DC1394
 #endif
@@ -132,7 +135,7 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
 #ifdef HAVE_MIL
         CV_CAP_MIL,
 #endif
-#ifdef HAVE_QUICKTIME
+#if defined(HAVE_QUICKTIME) || defined(HAVE_QTKIT)
         CV_CAP_QT,
 #endif
 #ifdef HAVE_UNICAP
@@ -168,7 +171,8 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
     // try every possibly installed camera API
     for (int i = 0; domains[i] >= 0; i++)
     {
-#if defined(HAVE_VIDEOINPUT)   || \
+#if defined(HAVE_DSHOW)        || \
+    defined(HAVE_MSMF)         || \
     defined(HAVE_TYZX)         || \
     defined(HAVE_VFW)          || \
     defined(HAVE_LIBV4L)       || \
@@ -181,6 +185,7 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
     defined(HAVE_CMU1394)      || \
     defined(HAVE_MIL)          || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_UNICAP)       || \
     defined(HAVE_PVAPI)        || \
     defined(HAVE_OPENNI)       || \
@@ -195,14 +200,20 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
 
         switch (domains[i])
         {
-#ifdef HAVE_VIDEOINPUT
+#ifdef HAVE_DSHOW
         case CV_CAP_DSHOW:
-            capture = cvCreateCameraCapture_DShow (index);
-            if (capture)
-                return capture;
+             capture = cvCreateCameraCapture_DShow (index);
+             if (capture)
+                 return capture;
+            break;
+#endif
+#ifdef HAVE_MSMF
+        case CV_CAP_MSMF:
+             capture = cvCreateCameraCapture_MSMF (index);
+             if (capture)
+                 return capture;
             break;
 #endif
-
 #ifdef HAVE_TYZX
         case CV_CAP_STEREO:
             capture = cvCreateCameraCapture_TYZX (index);
@@ -210,14 +221,12 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
                 return capture;
             break;
 #endif
-
         case CV_CAP_VFW:
 #ifdef HAVE_VFW
             capture = cvCreateCameraCapture_VFW (index);
             if (capture)
                 return capture;
 #endif
-
 #if defined HAVE_LIBV4L || defined HAVE_CAMV4L || defined HAVE_CAMV4L2 || defined HAVE_VIDEOIO
             capture = cvCreateCameraCapture_V4L (index);
             if (capture)
@@ -269,7 +278,7 @@ CV_IMPL CvCapture * cvCreateCameraCapture (int index)
             break;
 #endif
 
-#ifdef HAVE_QUICKTIME
+#if defined(HAVE_QUICKTIME) || defined(HAVE_QTKIT)
         case CV_CAP_QT:
             capture = cvCreateCameraCapture_QT (index);
             if (capture)
@@ -351,6 +360,16 @@ CV_IMPL CvCapture * cvCreateFileCapture (const char * filename)
     if (! result)
         result = cvCreateFileCapture_FFMPEG_proxy (filename);
 
+#ifdef HAVE_VFW
+    if (! result)
+        result = cvCreateFileCapture_VFW (filename);
+#endif
+
+#ifdef HAVE_MSMF
+    if (! result)
+        result = cvCreateFileCapture_MSMF (filename);
+#endif
+
 #ifdef HAVE_XINE
     if (! result)
         result = cvCreateFileCapture_XINE (filename);
@@ -361,7 +380,7 @@ CV_IMPL CvCapture * cvCreateFileCapture (const char * filename)
         result = cvCreateCapture_GStreamer (CV_CAP_GSTREAMER_FILE, filename);
 #endif
 
-#ifdef HAVE_QUICKTIME
+#if defined(HAVE_QUICKTIME) || defined(HAVE_QTKIT)
     if (! result)
         result = cvCreateFileCapture_QT (filename);
 #endif
@@ -399,6 +418,16 @@ CV_IMPL CvVideoWriter* cvCreateVideoWriter( const char* filename, int fourcc,
     if(!result)
         result = cvCreateVideoWriter_FFMPEG_proxy (filename, fourcc, fps, frameSize, is_color);
 
+#ifdef HAVE_VFW
+    if(!result)
+        result = cvCreateVideoWriter_VFW(filename, fourcc, fps, frameSize, is_color);
+#endif
+
+#ifdef HAVE_MSMF
+    if (!result)
+        result = cvCreateVideoWriter_MSMF(filename, fourcc, fps, frameSize, is_color);
+#endif
+
 /*  #ifdef HAVE_XINE
     if(!result)
         result = cvCreateVideoWriter_XINE(filename, fourcc, fps, frameSize, is_color);
@@ -409,7 +438,7 @@ CV_IMPL CvVideoWriter* cvCreateVideoWriter( const char* filename, int fourcc,
         result = cvCreateVideoWriter_AVFoundation(filename, fourcc, fps, frameSize, is_color);
 #endif
 
-#ifdef HAVE_QUICKTIME
+#if defined(HAVE_QUICKTIME) || defined(HAVE_QTKIT)
     if(!result)
         result = cvCreateVideoWriter_QT(filename, fourcc, fps, frameSize, is_color);
 #endif
@@ -445,7 +474,7 @@ namespace cv
 VideoCapture::VideoCapture()
 {}
 
-VideoCapture::VideoCapture(const std::string& filename)
+VideoCapture::VideoCapture(const String& filename)
 {
     open(filename);
 }
@@ -460,17 +489,17 @@ VideoCapture::~VideoCapture()
     cap.release();
 }
 
-bool VideoCapture::open(const std::string& filename)
+bool VideoCapture::open(const String& filename)
 {
-    if (!isOpened())
-    cap = cvCreateFileCapture(filename.c_str());
+    if (isOpened()) release();
+    cap.reset(cvCreateFileCapture(filename.c_str()));
     return isOpened();
 }
 
 bool VideoCapture::open(int device)
 {
-    if (!isOpened())
-    cap = cvCreateCameraCapture(device);
+    if (isOpened()) release();
+    cap.reset(cvCreateCameraCapture(device));
     return isOpened();
 }
 
@@ -495,10 +524,10 @@ bool VideoCapture::retrieve(Mat& image, int channel)
         return false;
     }
     if(_img->origin == IPL_ORIGIN_TL)
-        image = Mat(_img);
+        image = cv::cvarrToMat(_img);
     else
     {
-        Mat temp(_img);
+        Mat temp = cv::cvarrToMat(_img);
         flip(temp, image, 0);
     }
     return true;
@@ -532,9 +561,9 @@ double VideoCapture::get(int propId)
 VideoWriter::VideoWriter()
 {}
 
-VideoWriter::VideoWriter(const std::string& filename, int fourcc, double fps, Size frameSize, bool isColor)
+VideoWriter::VideoWriter(const String& filename, int _fourcc, double fps, Size frameSize, bool isColor)
 {
-    open(filename, fourcc, fps, frameSize, isColor);
+    open(filename, _fourcc, fps, frameSize, isColor);
 }
 
 void VideoWriter::release()
@@ -547,9 +576,9 @@ VideoWriter::~VideoWriter()
     release();
 }
 
-bool VideoWriter::open(const std::string& filename, int fourcc, double fps, Size frameSize, bool isColor)
+bool VideoWriter::open(const String& filename, int _fourcc, double fps, Size frameSize, bool isColor)
 {
-    writer = cvCreateVideoWriter(filename.c_str(), fourcc, fps, frameSize, isColor);
+    writer.reset(cvCreateVideoWriter(filename.c_str(), _fourcc, fps, frameSize, isColor));
     return isOpened();
 }
 
@@ -570,4 +599,9 @@ VideoWriter& VideoWriter::operator << (const Mat& image)
     return *this;
 }
 
+int VideoWriter::fourcc(char c1, char c2, char c3, char c4)
+{
+    return (c1 & 255) + ((c2 & 255) << 8) + ((c3 & 255) << 16) + ((c4 & 255) << 24);
+}
+
 }
diff --git a/modules/highgui/src/cap_avfoundation.mm b/modules/highgui/src/cap_avfoundation.mm
index c7471e2e86..71777f8750 100644
--- a/modules/highgui/src/cap_avfoundation.mm
+++ b/modules/highgui/src/cap_avfoundation.mm
@@ -1325,4 +1325,3 @@ bool CvVideoWriter_AVFoundation::writeFrame(const IplImage* iplimage) {
     }
 
 }
-
diff --git a/modules/highgui/src/cap_dc1394_v2.cpp b/modules/highgui/src/cap_dc1394_v2.cpp
index 2aa494fac7..0d5f898186 100644
--- a/modules/highgui/src/cap_dc1394_v2.cpp
+++ b/modules/highgui/src/cap_dc1394_v2.cpp
@@ -45,7 +45,16 @@
 
 #include <unistd.h>
 #include <stdint.h>
-#include <sys/select.h>
+#ifdef WIN32
+  // On Windows, we have no sys/select.h, but we need to pick up
+  // select() which is in winsock2.
+  #ifndef __SYS_SELECT_H__
+    #define __SYS_SELECT_H__ 1
+    #include <winsock2.h>
+  #endif
+#else
+  #include <sys/select.h>
+#endif /*WIN32*/
 #include <dc1394/dc1394.h>
 #include <stdlib.h>
 #include <string.h>
@@ -295,6 +304,11 @@ bool CvCaptureCAM_DC1394_v2_CPP::startCapture()
         return false;
     if (isoSpeed > 0)
     {
+        // if capable set operation mode to 1394b for iso speeds above 400
+        if (isoSpeed > 400 && dcCam->bmode_capable == DC1394_TRUE)
+        {
+            dc1394_video_set_operation_mode(dcCam, DC1394_OPERATION_MODE_1394B);
+        }
         code = dc1394_video_set_iso_speed(dcCam,
                                           isoSpeed <= 100 ? DC1394_ISO_SPEED_100 :
                                           isoSpeed <= 200 ? DC1394_ISO_SPEED_200 :
diff --git a/modules/highgui/src/cap_dshow.cpp b/modules/highgui/src/cap_dshow.cpp
index c2513d788d..90ffa00034 100644
--- a/modules/highgui/src/cap_dshow.cpp
+++ b/modules/highgui/src/cap_dshow.cpp
@@ -41,7 +41,7 @@
 
 #include "precomp.hpp"
 
-#if (defined WIN32 || defined _WIN32) && defined HAVE_VIDEOINPUT
+#if (defined WIN32 || defined _WIN32) && defined HAVE_DSHOW
 
 /*
    DirectShow-based Video Capturing module is based on
@@ -87,8 +87,6 @@ Thanks to:
 */
 /////////////////////////////////////////////////////////
 
-#include "precomp.hpp"
-
 #if defined _MSC_VER && _MSC_VER >= 100
 //'sprintf': name was marked as #pragma deprecated
 #pragma warning(disable: 4995)
@@ -3100,6 +3098,7 @@ HRESULT videoInput::routeCrossbar(ICaptureGraphBuilder2 **ppBuild, IBaseFilter *
     return hr;
 }
 
+
 /********************* Capturing video from camera via DirectShow *********************/
 
 class CvCaptureCAM_DShow : public CvCapture
@@ -3163,18 +3162,18 @@ void CvCaptureCAM_DShow::close()
 // Initialize camera input
 bool CvCaptureCAM_DShow::open( int _index )
 {
-    int try_index = _index;
     int devices = 0;
 
     close();
     devices = VI.listDevices(true);
     if (devices == 0)
         return false;
-    try_index = try_index < 0 ? 0 : (try_index > devices-1 ? devices-1 : try_index);
-    VI.setupDevice(try_index);
-    if( !VI.isDeviceSetup(try_index) )
+    if (_index < 0 || _index > devices-1)
         return false;
-    index = try_index;
+    VI.setupDevice(_index);
+    if( !VI.isDeviceSetup(_index) )
+        return false;
+    index = _index;
     return true;
 }
 
@@ -3194,8 +3193,10 @@ IplImage* CvCaptureCAM_DShow::retrieveFrame(int)
         frame = cvCreateImage( cvSize(w,h), 8, 3 );
     }
 
-    VI.getPixels( index, (uchar*)frame->imageData, false, true );
-    return frame;
+    if (VI.getPixels( index, (uchar*)frame->imageData, false, true ))
+        return frame;
+    else
+        return NULL;
 }
 
 double CvCaptureCAM_DShow::getProperty( int property_id )
diff --git a/modules/highgui/src/cap_ffmpeg.cpp b/modules/highgui/src/cap_ffmpeg.cpp
index 6029fa2e0e..192c0da694 100644
--- a/modules/highgui/src/cap_ffmpeg.cpp
+++ b/modules/highgui/src/cap_ffmpeg.cpp
@@ -41,7 +41,7 @@
 
 #include "precomp.hpp"
 
-#ifdef HAVE_FFMPEG
+#if defined HAVE_FFMPEG && !defined WIN32
 #include "cap_ffmpeg_impl.hpp"
 #else
 #include "cap_ffmpeg_api.hpp"
@@ -85,6 +85,16 @@ private:
     icvInitFFMPEG()
     {
     #if defined WIN32 || defined _WIN32
+    # ifdef HAVE_WINRT
+        const wchar_t* module_name = L"opencv_ffmpeg"
+            CVAUX_STRW(CV_MAJOR_VERSION) CVAUX_STRW(CV_MINOR_VERSION) CVAUX_STRW(CV_SUBMINOR_VERSION)
+        #if (defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__)
+            L"_64"
+        #endif
+            L".dll";
+
+        icvFFOpenCV = LoadPackagedLibrary( module_name, 0 );
+    # else
         const char* module_name = "opencv_ffmpeg"
             CVAUX_STR(CV_MAJOR_VERSION) CVAUX_STR(CV_MINOR_VERSION) CVAUX_STR(CV_SUBMINOR_VERSION)
         #if (defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__)
@@ -93,6 +103,8 @@ private:
             ".dll";
 
         icvFFOpenCV = LoadLibrary( module_name );
+    # endif
+
         if( icvFFOpenCV )
         {
             icvCreateFileCapture_FFMPEG_p =
@@ -209,11 +221,7 @@ CvCapture* cvCreateFileCapture_FFMPEG_proxy(const char * filename)
     if( result->open( filename ))
         return result;
     delete result;
-#if defined WIN32 || defined _WIN32
-    return cvCreateFileCapture_VFW(filename);
-#else
     return 0;
-#endif
 }
 
 class CvVideoWriter_FFMPEG_proxy :
@@ -263,9 +271,5 @@ CvVideoWriter* cvCreateVideoWriter_FFMPEG_proxy( const char* filename, int fourc
     if( result->open( filename, fourcc, fps, frameSize, isColor != 0 ))
         return result;
     delete result;
-#if defined WIN32 || defined _WIN32
-    return cvCreateVideoWriter_VFW(filename, fourcc, fps, frameSize, isColor);
-#else
     return 0;
-#endif
 }
diff --git a/modules/highgui/src/cap_ffmpeg_api.hpp b/modules/highgui/src/cap_ffmpeg_api.hpp
index 1fe488c049..e7a956063c 100644
--- a/modules/highgui/src/cap_ffmpeg_api.hpp
+++ b/modules/highgui/src/cap_ffmpeg_api.hpp
@@ -7,9 +7,11 @@ extern "C"
 #endif
 
 #if defined WIN32 || defined _WIN32
-#define OPENCV_FFMPEG_API __declspec(dllexport)
+#   define OPENCV_FFMPEG_API __declspec(dllexport)
+#elif defined __GNUC__ && __GNUC__ >= 4
+#   define OPENCV_FFMPEG_API __attribute__ ((visibility ("default")))
 #else
-#define OPENCV_FFMPEG_API
+#   define OPENCV_FFMPEG_API
 #endif
 
 enum
@@ -94,4 +96,3 @@ typedef int (*Read_InputMediaStream_FFMPEG_Plugin)(struct InputMediaStream_FFMPE
 #endif
 
 #endif
-
diff --git a/modules/highgui/src/cap_ffmpeg_impl.hpp b/modules/highgui/src/cap_ffmpeg_impl.hpp
index dd9c47c544..91fc8953dd 100644
--- a/modules/highgui/src/cap_ffmpeg_impl.hpp
+++ b/modules/highgui/src/cap_ffmpeg_impl.hpp
@@ -41,6 +41,9 @@
 //M*/
 
 #include "cap_ffmpeg_api.hpp"
+#if !(defined(WIN32) || defined(_WIN32) || defined(WINCE))
+# include <pthread.h>
+#endif
 #include <assert.h>
 #include <algorithm>
 #include <limits>
@@ -153,6 +156,14 @@ extern "C" {
 #define AVERROR_EOF (-MKTAG( 'E','O','F',' '))
 #endif
 
+#if LIBAVCODEC_BUILD >= CALC_FFMPEG_VERSION(54,25,0)
+#  define CV_CODEC_ID AVCodecID
+#  define CV_CODEC(name) AV_##name
+#else
+#  define CV_CODEC_ID CodecID
+#  define CV_CODEC(name) name
+#endif
+
 static int get_number_of_cpus(void)
 {
 #if LIBAVFORMAT_BUILD < CALC_FFMPEG_VERSION(52, 111, 0)
@@ -363,7 +374,15 @@ private:
 
 struct ImplMutex::Impl
 {
-    void init() { InitializeCriticalSection(&cs); refcount = 1; }
+    void init()
+    {
+#if (_WIN32_WINNT >= 0x0600)
+        ::InitializeCriticalSectionEx(&cs, 1000, 0);
+#else
+        ::InitializeCriticalSection(&cs);
+#endif
+        refcount = 1;
+    }
     void destroy() { DeleteCriticalSection(&cs); }
 
     void lock() { EnterCriticalSection(&cs); }
@@ -1026,7 +1045,7 @@ static const char * icvFFMPEGErrStr(int err)
 
 /* function internal to FFMPEG (libavformat/riff.c) to lookup codec id by fourcc tag*/
 extern "C" {
-    enum CodecID codec_get_bmp_id(unsigned int tag);
+    enum CV_CODEC_ID codec_get_bmp_id(unsigned int tag);
 }
 
 void CvVideoWriter_FFMPEG::init()
@@ -1078,7 +1097,7 @@ static AVFrame * icv_alloc_picture_FFMPEG(int pix_fmt, int width, int height, bo
 
 /* add a video output stream to the container */
 static AVStream *icv_add_video_stream_FFMPEG(AVFormatContext *oc,
-                                             CodecID codec_id,
+                                             CV_CODEC_ID codec_id,
                                              int w, int h, int bitrate,
                                              double fps, int pixel_format)
 {
@@ -1110,7 +1129,7 @@ static AVStream *icv_add_video_stream_FFMPEG(AVFormatContext *oc,
     c->codec_id = oc->oformat->video_codec;
 #endif
 
-    if(codec_id != CODEC_ID_NONE){
+    if(codec_id != CV_CODEC(CODEC_ID_NONE)){
         c->codec_id = codec_id;
     }
 
@@ -1179,10 +1198,10 @@ static AVStream *icv_add_video_stream_FFMPEG(AVFormatContext *oc,
     c->gop_size = 12; /* emit one intra frame every twelve frames at most */
     c->pix_fmt = (PixelFormat) pixel_format;
 
-    if (c->codec_id == CODEC_ID_MPEG2VIDEO) {
+    if (c->codec_id == CV_CODEC(CODEC_ID_MPEG2VIDEO)) {
         c->max_b_frames = 2;
     }
-    if (c->codec_id == CODEC_ID_MPEG1VIDEO || c->codec_id == CODEC_ID_MSMPEG4V3){
+    if (c->codec_id == CV_CODEC(CODEC_ID_MPEG1VIDEO) || c->codec_id == CV_CODEC(CODEC_ID_MSMPEG4V3)){
         /* needed to avoid using macroblocks in which some coeffs overflow
            this doesnt happen with normal video, it just happens here as the
            motion of the chroma plane doesnt match the luma plane */
@@ -1193,7 +1212,7 @@ static AVStream *icv_add_video_stream_FFMPEG(AVFormatContext *oc,
 #if LIBAVUTIL_BUILD > CALC_FFMPEG_VERSION(51,11,0)
     /* Some settings for libx264 encoding, restore dummy values for gop_size
      and qmin since they will be set to reasonable defaults by the libx264
-     preset system. Also, use a crf encode with the default quality rating, 
+     preset system. Also, use a crf encode with the default quality rating,
      this seems easier than finding an appropriate default bitrate. */
     if (c->codec_id == CODEC_ID_H264) {
       c->gop_size = -1;
@@ -1290,7 +1309,7 @@ bool CvVideoWriter_FFMPEG::writeFrame( const unsigned char* data, int step, int
 
 #if LIBAVFORMAT_BUILD < 5231
     // It is not needed in the latest versions of the ffmpeg
-    if( c->codec_id == CODEC_ID_RAWVIDEO && origin != 1 )
+    if( c->codec_id == CV_CODEC(CODEC_ID_RAWVIDEO) && origin != 1 )
     {
         if( !temp_image.data )
         {
@@ -1381,8 +1400,6 @@ bool CvVideoWriter_FFMPEG::writeFrame( const unsigned char* data, int step, int
 /// close video output stream and free associated memory
 void CvVideoWriter_FFMPEG::close()
 {
-    unsigned i;
-
     // nothing to do if already released
     if ( !picture )
         return;
@@ -1438,13 +1455,6 @@ void CvVideoWriter_FFMPEG::close()
 
     av_free(outbuf);
 
-    /* free the streams */
-    for(i = 0; i < oc->nb_streams; i++)
-    {
-        av_freep(&oc->streams[i]->codec);
-        av_freep(&oc->streams[i]);
-    }
-
     if (!(fmt->flags & AVFMT_NOFILE))
     {
         /* close the output file */
@@ -1462,7 +1472,7 @@ void CvVideoWriter_FFMPEG::close()
     }
 
     /* free the stream */
-    av_free(oc);
+    avformat_free_context(oc);
 
     if( temp_image.data )
     {
@@ -1477,7 +1487,7 @@ void CvVideoWriter_FFMPEG::close()
 bool CvVideoWriter_FFMPEG::open( const char * filename, int fourcc,
                                  double fps, int width, int height, bool is_color )
 {
-    CodecID codec_id = CODEC_ID_NONE;
+    CV_CODEC_ID codec_id = CV_CODEC(CODEC_ID_NONE);
     int err, codec_pix_fmt;
     double bitrate_scale = 1;
 
@@ -1518,11 +1528,11 @@ bool CvVideoWriter_FFMPEG::open( const char * filename, int fourcc,
 
     /* Lookup codec_id for given fourcc */
 #if LIBAVCODEC_VERSION_INT<((51<<16)+(49<<8)+0)
-    if( (codec_id = codec_get_bmp_id( fourcc )) == CODEC_ID_NONE )
+    if( (codec_id = codec_get_bmp_id( fourcc )) == CV_CODEC(CODEC_ID_NONE) )
         return false;
 #else
     const struct AVCodecTag * tags[] = { codec_bmp_tags, NULL};
-    if( (codec_id = av_codec_get_id(tags, fourcc)) == CODEC_ID_NONE )
+    if( (codec_id = av_codec_get_id(tags, fourcc)) == CV_CODEC(CODEC_ID_NONE) )
         return false;
 #endif
 
@@ -1544,20 +1554,20 @@ bool CvVideoWriter_FFMPEG::open( const char * filename, int fourcc,
     // set a few optimal pixel formats for lossless codecs of interest..
     switch (codec_id) {
 #if LIBAVCODEC_VERSION_INT>((50<<16)+(1<<8)+0)
-    case CODEC_ID_JPEGLS:
+    case CV_CODEC(CODEC_ID_JPEGLS):
         // BGR24 or GRAY8 depending on is_color...
         codec_pix_fmt = input_pix_fmt;
         break;
 #endif
-    case CODEC_ID_HUFFYUV:
+    case CV_CODEC(CODEC_ID_HUFFYUV):
         codec_pix_fmt = PIX_FMT_YUV422P;
         break;
-    case CODEC_ID_MJPEG:
-    case CODEC_ID_LJPEG:
+    case CV_CODEC(CODEC_ID_MJPEG):
+    case CV_CODEC(CODEC_ID_LJPEG):
         codec_pix_fmt = PIX_FMT_YUVJ420P;
         bitrate_scale = 3;
         break;
-    case CODEC_ID_RAWVIDEO:
+    case CV_CODEC(CODEC_ID_RAWVIDEO):
         codec_pix_fmt = input_pix_fmt == PIX_FMT_GRAY8 ||
                         input_pix_fmt == PIX_FMT_GRAY16LE ||
                         input_pix_fmt == PIX_FMT_GRAY16BE ? input_pix_fmt : PIX_FMT_YUV420P;
@@ -1788,7 +1798,7 @@ struct OutputMediaStream_FFMPEG
     void write(unsigned char* data, int size, int keyFrame);
 
     // add a video output stream to the container
-    static AVStream* addVideoStream(AVFormatContext *oc, CodecID codec_id, int w, int h, int bitrate, double fps, PixelFormat pixel_format);
+    static AVStream* addVideoStream(AVFormatContext *oc, CV_CODEC_ID codec_id, int w, int h, int bitrate, double fps, PixelFormat pixel_format);
 
     AVOutputFormat* fmt_;
     AVFormatContext* oc_;
@@ -1835,7 +1845,7 @@ void OutputMediaStream_FFMPEG::close()
     }
 }
 
-AVStream* OutputMediaStream_FFMPEG::addVideoStream(AVFormatContext *oc, CodecID codec_id, int w, int h, int bitrate, double fps, PixelFormat pixel_format)
+AVStream* OutputMediaStream_FFMPEG::addVideoStream(AVFormatContext *oc, CV_CODEC_ID codec_id, int w, int h, int bitrate, double fps, PixelFormat pixel_format)
 {
     #if LIBAVFORMAT_BUILD >= CALC_FFMPEG_VERSION(53, 10, 0)
         AVStream* st = avformat_new_stream(oc, 0);
@@ -1915,10 +1925,10 @@ AVStream* OutputMediaStream_FFMPEG::addVideoStream(AVFormatContext *oc, CodecID
     c->gop_size = 12; // emit one intra frame every twelve frames at most
     c->pix_fmt = pixel_format;
 
-    if (c->codec_id == CODEC_ID_MPEG2VIDEO)
+    if (c->codec_id == CV_CODEC(CODEC_ID_MPEG2VIDEO))
         c->max_b_frames = 2;
 
-    if (c->codec_id == CODEC_ID_MPEG1VIDEO || c->codec_id == CODEC_ID_MSMPEG4V3)
+    if (c->codec_id == CV_CODEC(CODEC_ID_MPEG1VIDEO) || c->codec_id == CV_CODEC(CODEC_ID_MSMPEG4V3))
     {
         // needed to avoid using macroblocks in which some coeffs overflow
         // this doesnt happen with normal video, it just happens here as the
@@ -1955,7 +1965,7 @@ bool OutputMediaStream_FFMPEG::open(const char* fileName, int width, int height,
     if (!fmt_)
         return false;
 
-    CodecID codec_id = CODEC_ID_H264;
+    CV_CODEC_ID codec_id = CV_CODEC(CODEC_ID_H264);
 
     // alloc memory for context
     #if LIBAVFORMAT_BUILD >= CALC_FFMPEG_VERSION(53, 2, 0)
@@ -2156,23 +2166,23 @@ bool InputMediaStream_FFMPEG::open(const char* fileName, int* codec, int* chroma
 
             switch (enc->codec_id)
             {
-            case CODEC_ID_MPEG1VIDEO:
+            case CV_CODEC(CODEC_ID_MPEG1VIDEO):
                 *codec = ::VideoCodec_MPEG1;
                 break;
 
-            case CODEC_ID_MPEG2VIDEO:
+            case CV_CODEC(CODEC_ID_MPEG2VIDEO):
                 *codec = ::VideoCodec_MPEG2;
                 break;
 
-            case CODEC_ID_MPEG4:
+            case CV_CODEC(CODEC_ID_MPEG4):
                 *codec = ::VideoCodec_MPEG4;
                 break;
 
-            case CODEC_ID_VC1:
+            case CV_CODEC(CODEC_ID_VC1):
                 *codec = ::VideoCodec_VC1;
                 break;
 
-            case CODEC_ID_H264:
+            case CV_CODEC(CODEC_ID_H264):
                 *codec = ::VideoCodec_H264;
                 break;
 
diff --git a/modules/highgui/src/cap_gstreamer.cpp b/modules/highgui/src/cap_gstreamer.cpp
index cafc803db4..4d4dc711bc 100644
--- a/modules/highgui/src/cap_gstreamer.cpp
+++ b/modules/highgui/src/cap_gstreamer.cpp
@@ -39,25 +39,33 @@
 //
 //M*/
 
-// Author: Nils Hasler <hasler@mpi-inf.mpg.de>
-//
-//         Max-Planck-Institut Informatik
-//
-// this implementation was inspired by gnash's gstreamer interface
-
-//
-// use GStreamer to read a video
-//
-
+/*!
+ * \file cap_gstreamer.cpp
+ * \author Nils Hasler <hasler@mpi-inf.mpg.de>
+ *         Max-Planck-Institut Informatik
+ * \author Dirk Van Haerenborgh <vhdirk@gmail.com>
+ *
+ * \brief Use GStreamer to read/write video
+ */
 #include "precomp.hpp"
 #include <unistd.h>
 #include <string.h>
-#include <map>
 #include <gst/gst.h>
+#include <gst/gstbuffer.h>
 #include <gst/video/video.h>
 #include <gst/app/gstappsink.h>
 #include <gst/app/gstappsrc.h>
 #include <gst/riff/riff-media.h>
+#include <gst/pbutils/missing-plugins.h>
+
+#define VERSION_NUM(major, minor, micro) (major * 1000000 + minor * 1000 + micro)
+#define FULL_GST_VERSION VERSION_NUM(GST_VERSION_MAJOR, GST_VERSION_MINOR, GST_VERSION_MICRO)
+
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,32)
+#include <gst/pbutils/encoding-profile.h>
+//#include <gst/base/gsttypefindhelper.h>
+#endif
+
 
 #ifdef NDEBUG
 #define CV_WARN(message)
@@ -65,8 +73,23 @@
 #define CV_WARN(message) fprintf(stderr, "warning: %s (%s:%d)\n", message, __FILE__, __LINE__)
 #endif
 
+#if GST_VERSION_MAJOR > 0
+#define COLOR_ELEM "videoconvert"
+#else
+#define COLOR_ELEM "ffmpegcolorspace"
+#endif
+
+
+void toFraction(double decimal, double &numerator, double &denominator);
+void handleMessage(GstElement * pipeline);
+
+
 static cv::Mutex gst_initializer_mutex;
 
+/*!
+ * \brief The gst_initializer class
+ * Initializes gstreamer once in the whole process
+ */
 class gst_initializer
 {
 public:
@@ -80,9 +103,16 @@ private:
     gst_initializer()
     {
         gst_init(NULL, NULL);
+//        gst_debug_set_active(1);
+//        gst_debug_set_colored(1);
+//        gst_debug_set_default_threshold(GST_LEVEL_INFO);
     }
 };
 
+/*!
+ * \brief The CvCapture_GStreamer class
+ * Use GStreamer to capture video
+ */
 class CvCapture_GStreamer : public CvCapture
 {
 public:
@@ -100,259 +130,486 @@ public:
 protected:
     void init();
     bool reopen();
-    void handleMessage();
+    bool isPipelinePlaying();
+    void startPipeline();
+    void stopPipeline();
     void restartPipeline();
-    void setFilter(const char*, int, int, int);
+    void setFilter(const char* prop, int type, int v1, int v2 = NULL);
     void removeFilter(const char *filter);
-    void static newPad(GstElement *myelement,
-                             GstPad     *pad,
-                             gpointer    data);
-    GstElement           *pipeline;
-    GstElement           *uridecodebin;
-    GstElement           *color;
-    GstElement           *sink;
-
-    GstBuffer           *buffer;
-    GstCaps            *caps;
-    IplImage           *frame;
+    static void newPad(GstElement *myelement,
+                       GstPad     *pad,
+                       gpointer    data);
+    GstElement*   pipeline;
+    GstElement*   uridecodebin;
+    GstElement*   color;
+    GstElement*   sink;
+#if GST_VERSION_MAJOR > 0
+    GstSample*    sample;
+    GstMapInfo*   info;
+#endif
+    GstBuffer*    buffer;
+    GstCaps*      caps;
+    GstCaps*      buffer_caps;
+    IplImage*     frame;
 };
 
+/*!
+ * \brief CvCapture_GStreamer::init
+ * inits the class
+ */
 void CvCapture_GStreamer::init()
 {
-    pipeline=0;
-    frame=0;
-    buffer=0;
-    frame=0;
-
+    pipeline = NULL;
+    frame = NULL;
+    buffer = NULL;
+    buffer_caps = NULL;
+#if GST_VERSION_MAJOR > 0
+    sample = NULL;
+    info = new GstMapInfo;
+#endif
 }
 
-void CvCapture_GStreamer::handleMessage()
+/*!
+ * \brief CvCapture_GStreamer::close
+ * Closes the pipeline and destroys all instances
+ */
+void CvCapture_GStreamer::close()
 {
-    GstBus* bus = gst_element_get_bus(pipeline);
+    if (isPipelinePlaying())
+        this->stopPipeline();
 
-    while(gst_bus_have_pending(bus)) {
-        GstMessage* msg = gst_bus_pop(bus);
-
-//        printf("Got %s message\n", GST_MESSAGE_TYPE_NAME(msg));
-
-        switch (GST_MESSAGE_TYPE (msg)) {
-        case GST_MESSAGE_STATE_CHANGED:
-            GstState oldstate, newstate, pendstate;
-            gst_message_parse_state_changed(msg, &oldstate, &newstate, &pendstate);
-//            printf("state changed from %d to %d (%d)\n", oldstate, newstate, pendstate);
-            break;
-        case GST_MESSAGE_ERROR: {
-            GError *err;
-            gchar *debug;
-            gst_message_parse_error(msg, &err, &debug);
-
-            fprintf(stderr, "GStreamer Plugin: Embedded video playback halted; module %s reported: %s\n",
-                  gst_element_get_name(GST_MESSAGE_SRC (msg)), err->message);
-
-            g_error_free(err);
-            g_free(debug);
-
-            gst_element_set_state(pipeline, GST_STATE_NULL);
-
-            break;
-            }
-        case GST_MESSAGE_EOS:
-//            CV_WARN("NetStream has reached the end of the stream.");
-
-            break;
-        default:
-//            CV_WARN("unhandled message\n");
-            break;
-        }
-
-        gst_message_unref(msg);
+    if(pipeline) {
+        gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_NULL);
+        gst_object_unref(GST_OBJECT(pipeline));
     }
+    if(uridecodebin){
+        gst_object_unref(GST_OBJECT(uridecodebin));
+    }
+    if(color){
+        gst_object_unref(GST_OBJECT(color));
+    }
+    if(sink){
+        gst_object_unref(GST_OBJECT(sink));
+    }
+    if(buffer)
+        gst_buffer_unref(buffer);
+    if(frame) {
+        frame->imageData = 0;
+        cvReleaseImage(&frame);
+    }
+    if(caps){
+        gst_caps_unref(caps);
+    }
+    if(buffer_caps){
+        gst_caps_unref(buffer_caps);
+    }
+#if GST_VERSION_MAJOR > 0
+    if(sample){
+        gst_sample_unref(sample);
+    }
+#endif
 
-    gst_object_unref(GST_OBJECT(bus));
 }
 
-//
-// start the pipeline, grab a buffer, and pause again
-//
+/*!
+ * \brief CvCapture_GStreamer::grabFrame
+ * \return
+ * Grabs a sample from the pipeline, awaiting consumation by retreiveFrame.
+ * The pipeline is started if it was not running yet
+ */
 bool CvCapture_GStreamer::grabFrame()
 {
-
     if(!pipeline)
         return false;
 
+    // start the pipeline if it was not in playing state yet
+    if(!this->isPipelinePlaying())
+        this->startPipeline();
+
+    // bail out if EOS
     if(gst_app_sink_is_eos(GST_APP_SINK(sink)))
         return false;
 
+#if GST_VERSION_MAJOR == 0
     if(buffer)
         gst_buffer_unref(buffer);
-    handleMessage();
 
     buffer = gst_app_sink_pull_buffer(GST_APP_SINK(sink));
+#else
+    if(sample)
+        gst_sample_unref(sample);
+
+    sample = gst_app_sink_pull_sample(GST_APP_SINK(sink));
+
+    if(!sample)
+        return false;
+
+    buffer = gst_sample_get_buffer(sample);
+#endif
+
     if(!buffer)
         return false;
 
     return true;
 }
 
-//
-// decode buffer
-//
+/*!
+ * \brief CvCapture_GStreamer::retrieveFrame
+ * \return IplImage pointer. [Transfer Full]
+ *  Retreive the previously grabbed buffer, and wrap it in an IPLImage structure
+ */
 IplImage * CvCapture_GStreamer::retrieveFrame(int)
 {
     if(!buffer)
         return 0;
 
-    if(!frame) {
+    //construct a frame header if we did not have any yet
+    if(!frame)
+    {
         gint height, width;
-        GstCaps *buff_caps = gst_buffer_get_caps(buffer);
-        assert(gst_caps_get_size(buff_caps) == 1);
-        GstStructure* structure = gst_caps_get_structure(buff_caps, 0);
 
+        //reuse the caps ptr
+        if (buffer_caps)
+            gst_caps_unref(buffer_caps);
+
+#if GST_VERSION_MAJOR == 0
+        buffer_caps = gst_buffer_get_caps(buffer);
+#else
+        buffer_caps = gst_sample_get_caps(sample);
+#endif
+        // bail out in no caps
+        assert(gst_caps_get_size(buffer_caps) == 1);
+        GstStructure* structure = gst_caps_get_structure(buffer_caps, 0);
+
+        // bail out if width or height are 0
         if(!gst_structure_get_int(structure, "width", &width) ||
-           !gst_structure_get_int(structure, "height", &height))
+                !gst_structure_get_int(structure, "height", &height))
+        {
+            return 0;
+        }
+
+
+        int depth = 3;
+#if GST_VERSION_MAJOR > 0
+        depth = 0;
+        const gchar* name = gst_structure_get_name(structure);
+        const gchar* format = gst_structure_get_string(structure, "format");
+
+        if (!name || !format)
             return 0;
 
-        frame = cvCreateImageHeader(cvSize(width, height), IPL_DEPTH_8U, 3);
-        gst_caps_unref(buff_caps);
+        // we support 3 types of data:
+        //     video/x-raw, format=BGR   -> 8bit, 3 channels
+        //     video/x-raw, format=GRAY8 -> 8bit, 1 channel
+        //     video/x-bayer             -> 8bit, 1 channel
+        // bayer data is never decoded, the user is responsible for that
+        // everything is 8 bit, so we just test the caps for bit depth
+
+        if (strcasecmp(name, "video/x-raw") == 0)
+        {
+            if (strcasecmp(format, "BGR") == 0) {
+                depth = 3;
+            }
+            else if(strcasecmp(format, "GRAY8") == 0){
+                depth = 1;
+            }
+        }
+        else if (strcasecmp(name, "video/x-bayer") == 0)
+        {
+            depth = 1;
+        }
+#endif
+        if (depth > 0) {
+            frame = cvCreateImageHeader(cvSize(width, height), IPL_DEPTH_8U, depth);
+        }else{
+            return 0;
+        }
     }
 
-    // no need to memcpy, just use gstreamer's buffer :-)
+    // gstreamer expects us to handle the memory at this point
+    // so we can just wrap the raw buffer and be done with it
+#if GST_VERSION_MAJOR == 0
     frame->imageData = (char *)GST_BUFFER_DATA(buffer);
-    //memcpy (frame->imageData, GST_BUFFER_DATA(buffer), GST_BUFFER_SIZE (buffer));
-    //gst_buffer_unref(buffer);
-    //buffer = 0;
+#else
+    // the data ptr in GstMapInfo is only valid throughout the mapifo objects life.
+    // TODO: check if reusing the mapinfo object is ok.
+
+    gboolean success = gst_buffer_map(buffer,info, (GstMapFlags)GST_MAP_READ);
+    if (!success){
+        //something weird went wrong here. abort. abort.
+        //fprintf(stderr,"GStreamer: unable to map buffer");
+        return 0;
+    }
+    frame->imageData = (char*)info->data;
+    gst_buffer_unmap(buffer,info);
+#endif
+
     return frame;
 }
 
-void CvCapture_GStreamer::restartPipeline()
+
+/*!
+ * \brief CvCapture_GStreamer::isPipelinePlaying
+ * \return if the pipeline is currently playing.
+ */
+bool CvCapture_GStreamer::isPipelinePlaying()
 {
-    CV_FUNCNAME("icvRestartPipeline");
+    GstState current, pending;
+    GstClockTime timeout = 5*GST_SECOND;
+    if(!GST_IS_ELEMENT(pipeline)){
+        return false;
+    }
+
+    GstStateChangeReturn ret = gst_element_get_state(GST_ELEMENT(pipeline),&current, &pending, timeout);
+    if (!ret){
+        //fprintf(stderr, "GStreamer: unable to query pipeline state\n");
+        return false;
+    }
+
+    return current == GST_STATE_PLAYING;
+}
+
+/*!
+ * \brief CvCapture_GStreamer::startPipeline
+ * Start the pipeline by setting it to the playing state
+ */
+void CvCapture_GStreamer::startPipeline()
+{
+    CV_FUNCNAME("icvStartPipeline");
 
     __BEGIN__;
 
-    printf("restarting pipeline, going to ready\n");
-
-    if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_READY) ==
-       GST_STATE_CHANGE_FAILURE) {
-        CV_ERROR(CV_StsError, "GStreamer: unable to start pipeline\n");
-        return;
-    }
-
-    printf("ready, relinking\n");
-
-    gst_element_unlink(uridecodebin, color);
-    printf("filtering with %s\n", gst_caps_to_string(caps));
-    gst_element_link_filtered(uridecodebin, color, caps);
-
-    printf("relinked, pausing\n");
-
+    //fprintf(stderr, "relinked, pausing\n");
     if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_PLAYING) ==
-       GST_STATE_CHANGE_FAILURE) {
+            GST_STATE_CHANGE_FAILURE) {
         CV_ERROR(CV_StsError, "GStreamer: unable to start pipeline\n");
+        gst_object_unref(pipeline);
         return;
     }
 
-    printf("state now paused\n");
-
-     __END__;
+    //printf("state now playing\n");
+    handleMessage(pipeline);
+    __END__;
 }
 
-void CvCapture_GStreamer::setFilter(const char *property, int type, int v1, int v2)
-{
 
-    if(!caps) {
+/*!
+ * \brief CvCapture_GStreamer::stopPipeline
+ * Stop the pipeline by setting it to NULL
+ */
+void CvCapture_GStreamer::stopPipeline()
+{
+    CV_FUNCNAME("icvStopPipeline");
+
+    __BEGIN__;
+
+    //fprintf(stderr, "restarting pipeline, going to ready\n");
+    if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_NULL) ==
+            GST_STATE_CHANGE_FAILURE) {
+        CV_ERROR(CV_StsError, "GStreamer: unable to stop pipeline\n");
+        gst_object_unref(pipeline);
+        return;
+    }
+    __END__;
+}
+
+/*!
+ * \brief CvCapture_GStreamer::restartPipeline
+ * Restart the pipeline
+ */
+void CvCapture_GStreamer::restartPipeline()
+{
+    handleMessage(pipeline);
+
+    this->stopPipeline();
+    this->startPipeline();
+}
+
+
+/*!
+ * \brief CvCapture_GStreamer::setFilter
+ * \param prop the property name
+ * \param type glib property type
+ * \param v1 the value
+ * \param v2 second value of property type requires it, else NULL
+ * Filter the output formats by setting appsink caps properties
+ */
+void CvCapture_GStreamer::setFilter(const char *prop, int type, int v1, int v2)
+{
+    //printf("GStreamer: setFilter \n");
+    if(!caps || !( GST_IS_CAPS (caps) ))
+    {
         if(type == G_TYPE_INT)
-            caps = gst_caps_new_simple("video/x-raw-rgb", property, type, v1, NULL);
+        {
+#if GST_VERSION_MAJOR == 0
+            caps = gst_caps_new_simple("video/x-raw-rgb", prop, type, v1, NULL);
+#else
+            caps = gst_caps_new_simple("video/x-raw","format",G_TYPE_STRING,"BGR", prop, type, v1, NULL);
+#endif
+        }
         else
-            caps = gst_caps_new_simple("video/x-raw-rgb", property, type, v1, v2, NULL);
-    } else {
-        //printf("caps before setting %s\n", gst_caps_to_string(caps));
-        if(type == G_TYPE_INT)
-            gst_caps_set_simple(caps, "video/x-raw-rgb", property, type, v1, NULL);
-        else
-            gst_caps_set_simple(caps, "video/x-raw-rgb", property, type, v1, v2, NULL);
+        {
+#if GST_VERSION_MAJOR == 0
+            caps = gst_caps_new_simple("video/x-raw-rgb", prop, type, v1, v2, NULL);
+#else
+            caps = gst_caps_new_simple("video/x-raw","format",G_TYPE_STRING,"BGR", prop, type, v1, v2, NULL);
+#endif
+        }
+    }
+    else
+    {
+#if GST_VERSION_MAJOR > 0
+        if (! gst_caps_is_writable(caps))
+            caps = gst_caps_make_writable (caps);
+#endif
+        if(type == G_TYPE_INT){
+            gst_caps_set_simple(caps, prop, type, v1, NULL);
+        }else{
+            gst_caps_set_simple(caps, prop, type, v1, v2, NULL);
+        }
     }
 
-    restartPipeline();
+#if GST_VERSION_MAJOR > 0
+    caps = gst_caps_fixate(caps);
+#endif
+
+    gst_app_sink_set_caps(GST_APP_SINK(sink), caps);
+    //printf("filtering with %s\n", gst_caps_to_string(caps));
 }
 
+
+/*!
+ * \brief CvCapture_GStreamer::removeFilter
+ * \param filter filter to remove
+ * remove the specified filter from the appsink template caps
+ */
 void CvCapture_GStreamer::removeFilter(const char *filter)
 {
     if(!caps)
         return;
 
+#if GST_VERSION_MAJOR > 0
+    if (! gst_caps_is_writable(caps))
+        caps = gst_caps_make_writable (caps);
+#endif
+
     GstStructure *s = gst_caps_get_structure(caps, 0);
     gst_structure_remove_field(s, filter);
 
-    restartPipeline();
+    gst_app_sink_set_caps(GST_APP_SINK(sink), caps);
 }
 
-
-//
-// connect uridecodebin dynamically created source pads to colourconverter
-//
-void CvCapture_GStreamer::newPad(GstElement * /*uridecodebin*/,
-                             GstPad     *pad,
-                             gpointer    data)
+/*!
+ * \brief CvCapture_GStreamer::newPad link dynamic padd
+ * \param pad
+ * \param data
+ * decodebin creates pads based on stream information, which is not known upfront
+ * on receiving the pad-added signal, we connect it to the colorspace conversion element
+ */
+void CvCapture_GStreamer::newPad(GstElement * /*elem*/,
+                                 GstPad     *pad,
+                                 gpointer    data)
 {
     GstPad *sinkpad;
     GstElement *color = (GstElement *) data;
 
-
     sinkpad = gst_element_get_static_pad (color, "sink");
-
-//  printf("linking dynamic pad to colourconverter %p %p\n", uridecodebin, pad);
+    if (!sinkpad){
+        //fprintf(stderr, "Gstreamer: no pad named sink\n");
+        return;
+    }
 
     gst_pad_link (pad, sinkpad);
-
     gst_object_unref (sinkpad);
 }
 
+/*!
+ * \brief CvCapture_GStreamer::open Open the given file with gstreamer
+ * \param type CvCapture type. One of CV_CAP_GSTREAMER_*
+ * \param filename Filename to open in case of CV_CAP_GSTREAMER_FILE
+ * \return boolean. Specifies if opening was succesful.
+ *
+ * In case of CV_CAP_GSTREAMER_V4L(2), a pipelin is constructed as follows:
+ *    v4l2src ! autoconvert ! appsink
+ *
+ *
+ * The 'filename' parameter is not limited to filesystem paths, and may be one of the following:
+ *
+ *  - a normal filesystem path:
+ *        e.g. video.avi or /path/to/video.avi or C:\\video.avi
+ *  - an uri:
+ *        e.g. file:///path/to/video.avi or rtsp:///path/to/stream.asf
+ *  - a gstreamer pipeline description:
+ *        e.g. videotestsrc ! videoconvert ! appsink
+ *        the appsink name should be either 'appsink0' (the default) or 'opencvsink'
+ *
+ *  When dealing with a file, CvCapture_GStreamer will not drop frames if the grabbing interval
+ *  larger than the framerate period. (Unlike the uri or manual pipeline description, which assume
+ *  a live source)
+ *
+ *  The pipeline will only be started whenever the first frame is grabbed. Setting pipeline properties
+ *  is really slow if we need to restart the pipeline over and over again.
+ *
+ *  TODO: the 'type' parameter is imo unneeded. for v4l2, filename 'v4l2:///dev/video0' can be used.
+ *  I expect this to be the same for CV_CAP_GSTREAMER_1394. Is anyone actually still using v4l (v1)?
+ *
+ */
 bool CvCapture_GStreamer::open( int type, const char* filename )
 {
-    close();
     CV_FUNCNAME("cvCaptureFromCAM_GStreamer");
 
     __BEGIN__;
 
     gst_initializer::init();
 
-//    if(!isInited) {
-//        printf("gst_init\n");
-//        gst_init (NULL, NULL);
-
-//        gst_debug_set_active(TRUE);
-//        gst_debug_set_colored(TRUE);
-//        gst_debug_set_default_threshold(GST_LEVEL_WARNING);
-
-//        isInited = true;
-//    }
     bool stream = false;
     bool manualpipeline = false;
     char *uri = NULL;
     uridecodebin = NULL;
-    if(type != CV_CAP_GSTREAMER_FILE) {
-        close();
-        return false;
+    GstElementFactory * testfac;
+
+    if (type == CV_CAP_GSTREAMER_V4L){
+        testfac = gst_element_factory_find("v4lsrc");
+        if (!testfac){
+            return false;
+        }
+        g_object_unref(G_OBJECT(testfac));
+        filename = "v4lsrc ! "COLOR_ELEM" ! appsink";
+    }
+    if (type == CV_CAP_GSTREAMER_V4L2){
+        testfac = gst_element_factory_find("v4l2src");
+        if (!testfac){
+            return false;
+        }
+        g_object_unref(G_OBJECT(testfac));
+        filename = "v4l2src ! "COLOR_ELEM" ! appsink";
     }
 
-    if(!gst_uri_is_valid(filename)) {
+
+    // test if we have a valid uri. If so, open it with an uridecodebin
+    // else, we might have a file or a manual pipeline.
+    // if gstreamer cannot parse the manual pipeline, we assume we were given and
+    // ordinary file path.
+    if(!gst_uri_is_valid(filename))
+    {
         uri = realpath(filename, NULL);
-        stream=false;
-        if(uri) {
+        stream = false;
+        if(uri)
+        {
             uri = g_filename_to_uri(uri, NULL, NULL);
             if(!uri) {
                 CV_WARN("GStreamer: Error opening file\n");
                 close();
                 return false;
             }
-        } else {
+        }
+        else
+        {
             GError *err = NULL;
-            //uridecodebin = gst_parse_bin_from_description(filename, FALSE, &err);
             uridecodebin = gst_parse_launch(filename, &err);
             if(!uridecodebin) {
-                CV_WARN("GStreamer: Error opening bin\n");
-                close();
+                //fprintf(stderr, "GStreamer: Error opening bin: %s\n", err->message);
+                //close();
                 return false;
             }
             stream = true;
@@ -363,32 +620,105 @@ bool CvCapture_GStreamer::open( int type, const char* filename )
         uri = g_strdup(filename);
     }
 
-    if(!uridecodebin) {
-        uridecodebin = gst_element_factory_make ("uridecodebin", NULL);
-        g_object_set(G_OBJECT(uridecodebin),"uri",uri, NULL);
+    bool element_from_uri = false;
+    if(!uridecodebin)
+    {
+        // At this writing, the v4l2 element (and maybe others too) does not support caps renegotiation.
+        // This means that we cannot use an uridecodebin when dealing with v4l2, since setting
+        // capture properties will not work.
+        // The solution (probably only until gstreamer 1.2) is to make an element from uri when dealing with v4l2.
+        gchar * protocol = gst_uri_get_protocol(uri);
+        if (!strcasecmp(protocol , "v4l2"))
+        {
+#if GST_VERSION_MAJOR == 0
+            uridecodebin = gst_element_make_from_uri(GST_URI_SRC, uri, "src");
+#else
+            uridecodebin = gst_element_make_from_uri(GST_URI_SRC, uri, "src", NULL);
+#endif
+            element_from_uri = true;
+        }else{
+            uridecodebin = gst_element_factory_make ("uridecodebin", NULL);
+            g_object_set(G_OBJECT(uridecodebin),"uri",uri, NULL);
+        }
+        g_free(protocol);
+
         if(!uridecodebin) {
-            CV_WARN("GStreamer: Failed to create uridecodebin\n");
+            //fprintf(stderr, "GStreamer: Error opening bin: %s\n", err->message);
             close();
             return false;
         }
     }
 
-    if(manualpipeline) {
-        GstIterator *it = gst_bin_iterate_sinks(GST_BIN(uridecodebin));
+    if(manualpipeline)
+    {
+        GstIterator *it = NULL;
+#if GST_VERSION_MAJOR == 0
+        it = gst_bin_iterate_sinks(GST_BIN(uridecodebin));
         if(gst_iterator_next(it, (gpointer *)&sink) != GST_ITERATOR_OK) {
-        CV_ERROR(CV_StsError, "GStreamer: cannot find appsink in manual pipeline\n");
-        return false;
+            CV_ERROR(CV_StsError, "GStreamer: cannot find appsink in manual pipeline\n");
+            return false;
         }
+#else
+        it = gst_bin_iterate_sinks (GST_BIN(uridecodebin));
 
-    pipeline = uridecodebin;
-    } else {
-    pipeline = gst_pipeline_new (NULL);
+        gboolean done = FALSE;
+        GstElement *element = NULL;
+        gchar* name = NULL;
+        GValue value = G_VALUE_INIT;
 
-        color = gst_element_factory_make("ffmpegcolorspace", NULL);
+        while (!done) {
+          switch (gst_iterator_next (it, &value)) {
+            case GST_ITERATOR_OK:
+              element = GST_ELEMENT (g_value_get_object (&value));
+              name = gst_element_get_name(element);
+              if (name){
+                if(strstr(name, "opencvsink") != NULL || strstr(name, "appsink") != NULL) {
+                  sink = GST_ELEMENT ( gst_object_ref (element) );
+                  done = TRUE;
+                }
+                g_free(name);
+              }
+              g_value_unset (&value);
+
+              break;
+            case GST_ITERATOR_RESYNC:
+              gst_iterator_resync (it);
+              break;
+            case GST_ITERATOR_ERROR:
+            case GST_ITERATOR_DONE:
+              done = TRUE;
+              break;
+          }
+        }
+        gst_iterator_free (it);
+
+
+        if (!sink){
+            CV_ERROR(CV_StsError, "GStreamer: cannot find appsink in manual pipeline\n");
+            return false;
+        }
+#endif
+        pipeline = uridecodebin;
+    }
+    else
+    {
+        pipeline = gst_pipeline_new (NULL);
+        // videoconvert (in 0.10: ffmpegcolorspace) automatically selects the correct colorspace
+        // conversion based on caps.
+        color = gst_element_factory_make(COLOR_ELEM, NULL);
         sink = gst_element_factory_make("appsink", NULL);
 
         gst_bin_add_many(GST_BIN(pipeline), uridecodebin, color, sink, NULL);
-        g_signal_connect(uridecodebin, "pad-added", G_CALLBACK(newPad), color);
+
+        if(element_from_uri) {
+            if(!gst_element_link(uridecodebin, color)) {
+                CV_ERROR(CV_StsError, "GStreamer: cannot link color -> sink\n");
+                gst_object_unref(pipeline);
+                return false;
+            }
+        }else{
+            g_signal_connect(uridecodebin, "pad-added", G_CALLBACK(newPad), color);
+        }
 
         if(!gst_element_link(color, sink)) {
             CV_ERROR(CV_StsError, "GStreamer: cannot link color -> sink\n");
@@ -397,230 +727,52 @@ bool CvCapture_GStreamer::open( int type, const char* filename )
         }
     }
 
+    //TODO: is 1 single buffer really high enough?
     gst_app_sink_set_max_buffers (GST_APP_SINK(sink), 1);
     gst_app_sink_set_drop (GST_APP_SINK(sink), stream);
+    //do not emit signals: all calls will be synchronous and blocking
+    gst_app_sink_set_emit_signals (GST_APP_SINK(sink), 0);
+
+#if GST_VERSION_MAJOR == 0
     caps = gst_caps_new_simple("video/x-raw-rgb",
                                "red_mask",   G_TYPE_INT, 0x0000FF,
                                "green_mask", G_TYPE_INT, 0x00FF00,
                                "blue_mask",  G_TYPE_INT, 0xFF0000,
                                NULL);
+#else
+    // support 1 and 3 channel 8 bit data, as well as bayer (also  1 channel, 8bit)
+    caps = gst_caps_from_string("video/x-raw, format=(string){BGR, GRAY8}; video/x-bayer,format=(string){rggb,bggr,grbg,gbrg}");
+#endif
     gst_app_sink_set_caps(GST_APP_SINK(sink), caps);
     gst_caps_unref(caps);
 
-    if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_READY) ==
-       GST_STATE_CHANGE_FAILURE) {
-        CV_WARN("GStreamer: unable to set pipeline to ready\n");
-        gst_object_unref(pipeline);
-        return false;
-    }
-
-    if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_PLAYING) ==
-       GST_STATE_CHANGE_FAILURE) {
-        gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_NULL);
-        CV_WARN("GStreamer: unable to set pipeline to playing\n");
-        gst_object_unref(pipeline);
-        return false;
-    }
-
-
-
-    handleMessage();
-
+    //we do not start recording here just yet.
+    // the user probably wants to set capture properties first, so start recording whenever the first frame is requested
     __END__;
 
     return true;
 }
 
-//
-//
-// gstreamer image sequence writer
-//
-//
-class CvVideoWriter_GStreamer : public CvVideoWriter
-{
-public:
-    CvVideoWriter_GStreamer() { init(); }
-    virtual ~CvVideoWriter_GStreamer() { close(); }
-
-    virtual bool open( const char* filename, int fourcc,
-                       double fps, CvSize frameSize, bool isColor );
-    virtual void close();
-    virtual bool writeFrame( const IplImage* image );
-protected:
-    void init();
-    std::map<int, char*> encs;
-    GstElement* source;
-    GstElement* file;
-    GstElement* enc;
-    GstElement* mux;
-    GstElement* color;
-    GstBuffer* buffer;
-    GstElement* pipeline;
-    int input_pix_fmt;
-};
-
-void CvVideoWriter_GStreamer::init()
-{
-    encs[CV_FOURCC('D','R','A','C')]=(char*)"diracenc";
-    encs[CV_FOURCC('H','F','Y','U')]=(char*)"ffenc_huffyuv";
-    encs[CV_FOURCC('J','P','E','G')]=(char*)"jpegenc";
-    encs[CV_FOURCC('M','J','P','G')]=(char*)"jpegenc";
-    encs[CV_FOURCC('M','P','1','V')]=(char*)"mpeg2enc";
-    encs[CV_FOURCC('M','P','2','V')]=(char*)"mpeg2enc";
-    encs[CV_FOURCC('T','H','E','O')]=(char*)"theoraenc";
-    encs[CV_FOURCC('V','P','8','0')]=(char*)"vp8enc";
-    encs[CV_FOURCC('H','2','6','4')]=(char*)"x264enc";
-    encs[CV_FOURCC('X','2','6','4')]=(char*)"x264enc";
-    encs[CV_FOURCC('X','V','I','D')]=(char*)"xvidenc";
-    encs[CV_FOURCC('F','F','Y','U')]=(char*)"y4menc";
-    //encs[CV_FOURCC('H','F','Y','U')]=(char*)"y4menc";
-    pipeline=0;
-    buffer=0;
-}
-void CvVideoWriter_GStreamer::close()
-{
-    if (pipeline) {
-        gst_app_src_end_of_stream(GST_APP_SRC(source));
-        gst_element_set_state (pipeline, GST_STATE_NULL);
-        gst_object_unref (GST_OBJECT (pipeline));
-    }
-}
-bool CvVideoWriter_GStreamer::open( const char * filename, int fourcc,
-        double fps, CvSize frameSize, bool is_color )
-{
-    CV_FUNCNAME("CvVideoWriter_GStreamer::open");
-
-    __BEGIN__;
-    //actually doesn't support fourcc parameter and encode an avi with jpegenc
-    //we need to find a common api between backend to support fourcc for avi
-    //but also to choose in a common way codec and container format (ogg,dirac,matroska)
-    // check arguments
-
-    assert (filename);
-    assert (fps > 0);
-    assert (frameSize.width > 0  &&  frameSize.height > 0);
-    std::map<int,char*>::iterator encit;
-    encit=encs.find(fourcc);
-    if (encit==encs.end())
-        CV_ERROR( CV_StsUnsupportedFormat,"Gstreamer Opencv backend doesn't support this codec acutally.");
-//    if(!isInited) {
-//        gst_init (NULL, NULL);
-//        isInited = true;
-//    }
-    gst_initializer::init();
-    close();
-    source=gst_element_factory_make("appsrc",NULL);
-    file=gst_element_factory_make("filesink", NULL);
-    enc=gst_element_factory_make(encit->second, NULL);
-    mux=gst_element_factory_make("avimux", NULL);
-    color = gst_element_factory_make("ffmpegcolorspace", NULL);
-    if (!enc)
-        CV_ERROR( CV_StsUnsupportedFormat, "Your version of Gstreamer doesn't support this codec acutally or needed plugin missing.");
-    g_object_set(G_OBJECT(file), "location", filename, NULL);
-    pipeline = gst_pipeline_new (NULL);
-    GstCaps* caps;
-    if (is_color) {
-        input_pix_fmt=1;
-        caps= gst_video_format_new_caps(GST_VIDEO_FORMAT_BGR,
-                                        frameSize.width,
-                                        frameSize.height,
-                                        (int) (fps * 1000),
-                                        1000,
-                                        1,
-                                        1);
-    }
-    else  {
-        input_pix_fmt=0;
-        caps= gst_caps_new_simple("video/x-raw-gray",
-                                  "width", G_TYPE_INT, frameSize.width,
-                                  "height", G_TYPE_INT, frameSize.height,
-                                  "framerate", GST_TYPE_FRACTION, int(fps),1,
-                                  "bpp",G_TYPE_INT,8,
-                                  "depth",G_TYPE_INT,8,
-                                  NULL);
-    }
-    gst_app_src_set_caps(GST_APP_SRC(source), caps);
-    if (fourcc==CV_FOURCC_DEFAULT) {
-        gst_bin_add_many(GST_BIN(pipeline), source, color,mux, file, NULL);
-        if(!gst_element_link_many(source,color,enc,mux,file,NULL)) {
-            CV_ERROR(CV_StsError, "GStreamer: cannot link elements\n");
-        }
-    }
-    else {
-        gst_bin_add_many(GST_BIN(pipeline), source, color,enc,mux, file, NULL);
-        if(!gst_element_link_many(source,color,enc,mux,file,NULL)) {
-            CV_ERROR(CV_StsError, "GStreamer: cannot link elements\n");
-        }
-    }
-
-
-    if(gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_PLAYING) ==
-        GST_STATE_CHANGE_FAILURE) {
-            CV_ERROR(CV_StsError, "GStreamer: cannot put pipeline to play\n");
-    }
-    __END__;
-    return true;
-}
-bool CvVideoWriter_GStreamer::writeFrame( const IplImage * image )
-{
-
-    CV_FUNCNAME("CvVideoWriter_GStreamer::writerFrame");
-
-    __BEGIN__;
-    if (input_pix_fmt == 1) {
-        if (image->nChannels != 3 || image->depth != IPL_DEPTH_8U) {
-            CV_ERROR(CV_StsUnsupportedFormat, "cvWriteFrame() needs images with depth = IPL_DEPTH_8U and nChannels = 3.");
-        }
-    }
-    else if (input_pix_fmt == 0) {
-        if (image->nChannels != 1 || image->depth != IPL_DEPTH_8U) {
-            CV_ERROR(CV_StsUnsupportedFormat, "cvWriteFrame() needs images with depth = IPL_DEPTH_8U and nChannels = 1.");
-        }
-    }
-    else {
-        assert(false);
-    }
-    int size;
-    size = image->imageSize;
-    buffer = gst_buffer_new_and_alloc (size);
-    //gst_buffer_set_data (buffer,(guint8*)image->imageData, size);
-    memcpy (GST_BUFFER_DATA(buffer),image->imageData, size);
-    gst_app_src_push_buffer(GST_APP_SRC(source),buffer);
-    //gst_buffer_unref(buffer);
-    //buffer = 0;
-    __END__;
-    return true;
-}
-CvVideoWriter* cvCreateVideoWriter_GStreamer(const char* filename, int fourcc, double fps,
-                                           CvSize frameSize, int isColor )
-{
-    CvVideoWriter_GStreamer* wrt = new CvVideoWriter_GStreamer;
-    if( wrt->open(filename, fourcc, fps,frameSize, isColor))
-        return wrt;
-
-    delete wrt;
-    return 0;
-}
-
-void CvCapture_GStreamer::close()
-{
-    if(pipeline) {
-        gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_NULL);
-        gst_object_unref(GST_OBJECT(pipeline));
-    }
-    if(buffer)
-        gst_buffer_unref(buffer);
-    if(frame) {
-        frame->imageData = 0;
-        cvReleaseImage(&frame);
-    }
-}
-
+/*!
+ * \brief CvCapture_GStreamer::getProperty retreive the requested property from the pipeline
+ * \param propId requested property
+ * \return property value
+ *
+ * There are two ways the properties can be retreived. For seek-based properties we can query the pipeline.
+ * For frame-based properties, we use the caps of the lasst receivef sample. This means that some properties
+ * are not available until a first frame was received
+ */
 double CvCapture_GStreamer::getProperty( int propId )
 {
     GstFormat format;
-    //GstQuery q;
     gint64 value;
+    gboolean status;
+
+#if GST_VERSION_MAJOR == 0
+#define FORMAT &format
+#else
+#define FORMAT format
+#endif
 
     if(!pipeline) {
         CV_WARN("GStreamer: no pipeline");
@@ -630,33 +782,76 @@ double CvCapture_GStreamer::getProperty( int propId )
     switch(propId) {
     case CV_CAP_PROP_POS_MSEC:
         format = GST_FORMAT_TIME;
-        if(!gst_element_query_position(sink, &format, &value)) {
+        status = gst_element_query_position(sink, FORMAT, &value);
+        if(!status) {
             CV_WARN("GStreamer: unable to query position of stream");
             return false;
         }
         return value * 1e-6; // nano seconds to milli seconds
     case CV_CAP_PROP_POS_FRAMES:
         format = GST_FORMAT_DEFAULT;
-        if(!gst_element_query_position(sink, &format, &value)) {
+        status = gst_element_query_position(sink, FORMAT, &value);
+        if(!status) {
             CV_WARN("GStreamer: unable to query position of stream");
             return false;
         }
         return value;
     case CV_CAP_PROP_POS_AVI_RATIO:
         format = GST_FORMAT_PERCENT;
-        if(!gst_element_query_position(pipeline, &format, &value)) {
+        status = gst_element_query_position(sink, FORMAT, &value);
+        if(!status) {
             CV_WARN("GStreamer: unable to query position of stream");
             return false;
         }
         return ((double) value) / GST_FORMAT_PERCENT_MAX;
-    case CV_CAP_PROP_FRAME_WIDTH:
-    case CV_CAP_PROP_FRAME_HEIGHT:
-    case CV_CAP_PROP_FPS:
+    case CV_CAP_PROP_FRAME_WIDTH: {
+        if (!buffer_caps){
+            CV_WARN("GStreamer: unable to query width of frame; no frame grabbed yet");
+            return 0;
+        }
+        GstStructure* structure = gst_caps_get_structure(buffer_caps, 0);
+        gint width = 0;
+        if(!gst_structure_get_int(structure, "width", &width)){
+            CV_WARN("GStreamer: unable to query width of frame");
+            return 0;
+        }
+        return width;
+        break;
+    }
+    case CV_CAP_PROP_FRAME_HEIGHT: {
+        if (!buffer_caps){
+            CV_WARN("GStreamer: unable to query height of frame; no frame grabbed yet");
+            return 0;
+        }
+        GstStructure* structure = gst_caps_get_structure(buffer_caps, 0);
+        gint height = 0;
+        if(!gst_structure_get_int(structure, "height", &height)){
+            CV_WARN("GStreamer: unable to query height of frame");
+            return 0;
+        }
+        return height;
+        break;
+    }
+    case CV_CAP_PROP_FPS: {
+        if (!buffer_caps){
+            CV_WARN("GStreamer: unable to query framerate of stream; no frame grabbed yet");
+            return 0;
+        }
+        GstStructure* structure = gst_caps_get_structure(buffer_caps, 0);
+        gint num = 0, denom=1;
+        if(!gst_structure_get_fraction(structure, "framerate", &num, &denom)){
+            CV_WARN("GStreamer: unable to query framerate of stream");
+            return 0;
+        }
+        return (double)num/(double)denom;
+        break;
+    }
     case CV_CAP_PROP_FOURCC:
         break;
     case CV_CAP_PROP_FRAME_COUNT:
         format = GST_FORMAT_DEFAULT;
-        if(!gst_element_query_duration(pipeline, &format, &value)) {
+        status = gst_element_query_position(sink, FORMAT, &value);
+        if(!status) {
             CV_WARN("GStreamer: unable to query position of stream");
             return false;
         }
@@ -672,20 +867,31 @@ double CvCapture_GStreamer::getProperty( int propId )
         break;
     case CV_CAP_GSTREAMER_QUEUE_LENGTH:
         if(!sink) {
-                CV_WARN("GStreamer: there is no sink yet");
-                return false;
+            CV_WARN("GStreamer: there is no sink yet");
+            return false;
         }
         return gst_app_sink_get_max_buffers(GST_APP_SINK(sink));
     default:
         CV_WARN("GStreamer: unhandled property");
         break;
     }
+
+#undef FORMAT
+
     return false;
 }
 
+/*!
+ * \brief CvCapture_GStreamer::setProperty
+ * \param propId
+ * \param value
+ * \return success
+ * Sets the desired property id with val. If the pipeline is running,
+ * it is briefly stopped and started again after the property was set
+ */
 bool CvCapture_GStreamer::setProperty( int propId, double value )
 {
-       GstFormat format;
+    GstFormat format;
     GstSeekFlags flags;
 
     if(!pipeline) {
@@ -693,12 +899,17 @@ bool CvCapture_GStreamer::setProperty( int propId, double value )
         return false;
     }
 
+    bool wasPlaying = this->isPipelinePlaying();
+    if (wasPlaying)
+        this->stopPipeline();
+
+
     switch(propId) {
     case CV_CAP_PROP_POS_MSEC:
         format = GST_FORMAT_TIME;
         flags = (GstSeekFlags) (GST_SEEK_FLAG_FLUSH|GST_SEEK_FLAG_ACCURATE);
         if(!gst_element_seek_simple(GST_ELEMENT(pipeline), format,
-                        flags, (gint64) (value * GST_MSECOND))) {
+                                    flags, (gint64) (value * GST_MSECOND))) {
             CV_WARN("GStreamer: unable to seek");
         }
         break;
@@ -706,7 +917,7 @@ bool CvCapture_GStreamer::setProperty( int propId, double value )
         format = GST_FORMAT_DEFAULT;
         flags = (GstSeekFlags) (GST_SEEK_FLAG_FLUSH|GST_SEEK_FLAG_ACCURATE);
         if(!gst_element_seek_simple(GST_ELEMENT(pipeline), format,
-                        flags, (gint64) value)) {
+                                    flags, (gint64) value)) {
             CV_WARN("GStreamer: unable to seek");
         }
         break;
@@ -714,7 +925,7 @@ bool CvCapture_GStreamer::setProperty( int propId, double value )
         format = GST_FORMAT_PERCENT;
         flags = (GstSeekFlags) (GST_SEEK_FLAG_FLUSH|GST_SEEK_FLAG_ACCURATE);
         if(!gst_element_seek_simple(GST_ELEMENT(pipeline), format,
-                        flags, (gint64) (value * GST_FORMAT_PERCENT_MAX))) {
+                                    flags, (gint64) (value * GST_FORMAT_PERCENT_MAX))) {
             CV_WARN("GStreamer: unable to seek");
         }
         break;
@@ -732,15 +943,9 @@ bool CvCapture_GStreamer::setProperty( int propId, double value )
         break;
     case CV_CAP_PROP_FPS:
         if(value > 0) {
-            int num, denom;
-            num = (int) value;
-            if(value != num) { // FIXME this supports only fractions x/1 and x/2
-                num = (int) (value * 2);
-                denom = 2;
-            } else
-                denom = 1;
-
-            setFilter("framerate", GST_TYPE_FRACTION, num, denom);
+            double num=0, denom = 1;
+            toFraction(value, num,  denom);
+            setFilter("framerate", GST_TYPE_FRACTION, value, denom);
         } else
             removeFilter("framerate");
         break;
@@ -763,8 +968,19 @@ bool CvCapture_GStreamer::setProperty( int propId, double value )
     default:
         CV_WARN("GStreamer: unhandled property");
     }
+
+    if (wasPlaying)
+        this->startPipeline();
+
     return false;
 }
+
+/*!
+ * \brief cvCreateCapture_GStreamer
+ * \param type
+ * \param filename
+ * \return
+ */
 CvCapture* cvCreateCapture_GStreamer(int type, const char* filename )
 {
     CvCapture_GStreamer* capture = new CvCapture_GStreamer;
@@ -775,3 +991,544 @@ CvCapture* cvCreateCapture_GStreamer(int type, const char* filename )
     delete capture;
     return 0;
 }
+
+
+/*!
+ * \brief The CvVideoWriter_GStreamer class
+ * Use Gstreamer to write video
+ */
+class CvVideoWriter_GStreamer : public CvVideoWriter
+{
+public:
+    CvVideoWriter_GStreamer() { init(); }
+    virtual ~CvVideoWriter_GStreamer() { close(); }
+
+    virtual bool open( const char* filename, int fourcc,
+                       double fps, CvSize frameSize, bool isColor );
+    virtual void close();
+    virtual bool writeFrame( const IplImage* image );
+protected:
+    void init();
+    const char* filenameToMimetype(const char* filename);
+    GstElement* pipeline;
+    GstElement* source;
+    GstElement* encodebin;
+    GstElement* file;
+
+    GstBuffer* buffer;
+    int input_pix_fmt;
+    int num_frames;
+    double framerate;
+};
+
+/*!
+ * \brief CvVideoWriter_GStreamer::init
+ * initialise all variables
+ */
+void CvVideoWriter_GStreamer::init()
+{
+    pipeline = NULL;
+    source = NULL;
+    encodebin = NULL;
+    file = NULL;
+    buffer = NULL;
+
+    num_frames = 0;
+    framerate = 0;
+}
+
+/*!
+ * \brief CvVideoWriter_GStreamer::close
+ * ends the pipeline by sending EOS and destroys the pipeline and all
+ * elements afterwards
+ */
+void CvVideoWriter_GStreamer::close()
+{
+    if (pipeline)
+    {
+        gst_app_src_end_of_stream(GST_APP_SRC(source));
+
+        //wait for EOS to trickle down the pipeline. This will let all elements finish properly
+        GstBus* bus = gst_element_get_bus(pipeline);
+        GstMessage *msg = gst_bus_timed_pop_filtered(bus, GST_CLOCK_TIME_NONE, (GstMessageType)(GST_MESSAGE_ERROR | GST_MESSAGE_EOS));
+        if(msg != NULL){
+            gst_message_unref(msg);
+            g_object_unref(G_OBJECT(bus));
+        }
+
+        gst_element_set_state (pipeline, GST_STATE_NULL);
+        handleMessage(pipeline);
+
+        gst_object_unref (GST_OBJECT (pipeline));
+
+        if (source)
+          gst_object_unref (GST_OBJECT (source));
+
+        if (file)
+          gst_object_unref (GST_OBJECT (file));
+    }
+}
+
+
+/*!
+ * \brief CvVideoWriter_GStreamer::filenameToMimetype
+ * \param filename
+ * \return mimetype
+ * Resturns a container mime type for a given filename by looking at it's extension
+ */
+const char* CvVideoWriter_GStreamer::filenameToMimetype(const char *filename)
+{
+    //get extension
+    const char *ext = strrchr(filename, '.');
+    if(!ext || ext == filename) return NULL;
+    ext += 1; //exclude the dot
+
+    // return a container mime based on the given extension.
+    // gstreamer's function returns too much possibilities, which is not useful to us
+
+    //return the appropriate mime
+    if (strncasecmp(ext,"avi", 3) == 0)
+        return (const char*)"video/x-msvideo";
+
+    if (strncasecmp(ext,"mkv", 3) == 0 || strncasecmp(ext,"mk3d",4) == 0  || strncasecmp(ext,"webm",4) == 0 )
+        return (const char*)"video/x-matroska";
+
+    if (strncasecmp(ext,"wmv", 3) == 0)
+        return (const char*)"video/x-ms-asf";
+
+    if (strncasecmp(ext,"mov", 3) == 0)
+        return (const char*)"video/x-quicktime";
+
+    if (strncasecmp(ext,"ogg", 3) == 0 || strncasecmp(ext,"ogv", 3) == 0)
+        return (const char*)"application/ogg";
+
+    if (strncasecmp(ext,"rm", 3) == 0)
+        return (const char*)"vnd.rn-realmedia";
+
+    if (strncasecmp(ext,"swf", 3) == 0)
+        return (const char*)"application/x-shockwave-flash";
+
+    if (strncasecmp(ext,"mp4", 3) == 0)
+        return (const char*)"video/x-quicktime, variant=(string)iso";
+
+    //default to avi
+    return (const char*)"video/x-msvideo";
+}
+
+
+/*!
+ * \brief CvVideoWriter_GStreamer::open
+ * \param filename filename to output to
+ * \param fourcc desired codec fourcc
+ * \param fps desired framerate
+ * \param frameSize the size of the expected frames
+ * \param is_color color or grayscale
+ * \return success
+ *
+ * We support 2 modes of operation. Either the user enters a filename and a fourcc
+ * code, or enters a manual pipeline description like in CvVideoCapture_Gstreamer.
+ * In the latter case, we just push frames on the appsink with appropriate caps.
+ * In the former case, we try to deduce the correct container from the filename,
+ * and the correct encoder from the fourcc profile.
+ *
+ * If the file extension did was not recognize, an avi container is used
+ *
+ */
+bool CvVideoWriter_GStreamer::open( const char * filename, int fourcc,
+                                    double fps, CvSize frameSize, bool is_color )
+{
+    CV_FUNCNAME("CvVideoWriter_GStreamer::open");
+
+    // check arguments
+    assert (filename);
+    assert (fps > 0);
+    assert (frameSize.width > 0  &&  frameSize.height > 0);
+
+    // init gstreamer
+    gst_initializer::init();
+
+    // init vars
+    bool manualpipeline = true;
+    int  bufsize = 0;
+    GError *err = NULL;
+    const char* mime = NULL;
+    GstStateChangeReturn stateret;
+
+    GstCaps* caps = NULL;
+    GstCaps* videocaps = NULL;
+
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,32)
+    GstCaps* containercaps = NULL;
+    GstEncodingContainerProfile* containerprofile = NULL;
+    GstEncodingVideoProfile* videoprofile = NULL;
+#endif
+
+    GstIterator *it = NULL;
+
+    // we first try to construct a pipeline from the given string.
+    // if that fails, we assume it is an ordinary filename
+
+    __BEGIN__;
+
+    encodebin = gst_parse_launch(filename, &err);
+    if(!encodebin) {
+        manualpipeline = false;
+    }
+
+    if(manualpipeline)
+    {
+#if GST_VERSION_MAJOR == 0
+        it = gst_bin_iterate_sources(GST_BIN(encodebin));
+        if(gst_iterator_next(it, (gpointer *)&source) != GST_ITERATOR_OK) {
+            CV_ERROR(CV_StsError, "GStreamer: cannot find appsink in manual pipeline\n");
+            return false;
+        }
+#else
+        it = gst_bin_iterate_sources (GST_BIN(encodebin));
+
+        gboolean done = FALSE;
+        GstElement *element = NULL;
+        gchar* name = NULL;
+        GValue value = G_VALUE_INIT;
+
+        while (!done) {
+          switch (gst_iterator_next (it, &value)) {
+            case GST_ITERATOR_OK:
+              element = GST_ELEMENT (g_value_get_object (&value));
+              name = gst_element_get_name(element);
+              if (name){
+                if(strstr(name, "opencvsrc") != NULL || strstr(name, "appsrc") != NULL) {
+                  source = GST_ELEMENT ( gst_object_ref (element) );
+                  done = TRUE;
+                }
+                g_free(name);
+              }
+              g_value_unset (&value);
+
+              break;
+            case GST_ITERATOR_RESYNC:
+              gst_iterator_resync (it);
+              break;
+            case GST_ITERATOR_ERROR:
+            case GST_ITERATOR_DONE:
+              done = TRUE;
+              break;
+          }
+        }
+        gst_iterator_free (it);
+
+        if (!source){
+            CV_ERROR(CV_StsError, "GStreamer: cannot find appsrc in manual pipeline\n");
+            return false;
+        }
+#endif
+        pipeline = encodebin;
+    }
+    else
+    {
+        pipeline = gst_pipeline_new (NULL);
+
+        // we just got a filename and a fourcc code.
+        // first, try to guess the container from the filename
+        //encodebin = gst_element_factory_make("encodebin", NULL);
+
+        //proxy old non existing fourcc ids. These were used in previous opencv versions,
+        //but do not even exist in gstreamer any more
+        if (fourcc == CV_FOURCC('M','P','1','V')) fourcc = CV_FOURCC('M', 'P', 'G' ,'1');
+        if (fourcc == CV_FOURCC('M','P','2','V')) fourcc = CV_FOURCC('M', 'P', 'G' ,'2');
+        if (fourcc == CV_FOURCC('D','R','A','C')) fourcc = CV_FOURCC('d', 'r', 'a' ,'c');
+
+        //create encoder caps from fourcc
+        videocaps = gst_riff_create_video_caps(fourcc, NULL, NULL, NULL, NULL, NULL);
+        if (!videocaps){
+            CV_ERROR( CV_StsUnsupportedFormat, "Gstreamer Opencv backend does not support this codec.");
+        }
+
+        //create container caps from file extension
+        mime = filenameToMimetype(filename);
+        if (!mime) {
+            CV_ERROR( CV_StsUnsupportedFormat, "Gstreamer Opencv backend does not support this file type.");
+        }
+
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,32)
+        containercaps = gst_caps_from_string(mime);
+
+        //create encodebin profile
+        containerprofile = gst_encoding_container_profile_new("container", "container", containercaps, NULL);
+        videoprofile = gst_encoding_video_profile_new(videocaps, NULL, NULL, 1);
+        gst_encoding_container_profile_add_profile(containerprofile, (GstEncodingProfile *) videoprofile);
+#endif
+
+        //create pipeline elements
+        encodebin = gst_element_factory_make("encodebin", NULL);
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,32)
+        g_object_set(G_OBJECT(encodebin), "profile", containerprofile, NULL);
+#endif
+        source = gst_element_factory_make("appsrc", NULL);
+        file = gst_element_factory_make("filesink", NULL);
+        g_object_set(G_OBJECT(file), "location", filename, NULL);
+    }
+
+    if (is_color)
+    {
+        input_pix_fmt = GST_VIDEO_FORMAT_BGR;
+        bufsize = frameSize.width * frameSize.height * 3;
+
+#if GST_VERSION_MAJOR == 0
+        caps = gst_video_format_new_caps(GST_VIDEO_FORMAT_BGR,
+                                         frameSize.width,
+                                         frameSize.height,
+                                         int(fps), 1,
+                                         1, 1);
+#else
+        caps = gst_caps_new_simple("video/x-raw",
+                                   "format", G_TYPE_STRING, "BGR",
+                                   "width", G_TYPE_INT, frameSize.width,
+                                   "height", G_TYPE_INT, frameSize.height,
+                                   "framerate", GST_TYPE_FRACTION, int(fps), 1,
+                                   NULL);
+        caps = gst_caps_fixate(caps);
+
+#endif
+
+    }
+    else
+    {
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,29)
+        input_pix_fmt = GST_VIDEO_FORMAT_GRAY8;
+        bufsize = frameSize.width * frameSize.height;
+
+#if GST_VERSION_MAJOR == 0
+        caps = gst_video_format_new_caps(GST_VIDEO_FORMAT_GRAY8,
+                                         frameSize.width,
+                                         frameSize.height,
+                                         int(fps), 1,
+                                         1, 1);
+#else
+        caps = gst_caps_new_simple("video/x-raw",
+                                   "format", G_TYPE_STRING, "GRAY8",
+                                   "width", G_TYPE_INT, frameSize.width,
+                                   "height", G_TYPE_INT, frameSize.height,
+                                   "framerate", GST_TYPE_FRACTION, int(fps), 1,
+                                   NULL);
+        caps = gst_caps_fixate(caps);
+#endif
+#else
+        CV_Assert(!"Gstreamer 0.10.29 or newer is required for grayscale input");
+#endif
+    }
+
+    gst_app_src_set_caps(GST_APP_SRC(source), caps);
+    gst_app_src_set_stream_type(GST_APP_SRC(source), GST_APP_STREAM_TYPE_STREAM);
+    gst_app_src_set_size (GST_APP_SRC(source), -1);
+
+    g_object_set(G_OBJECT(source), "format", GST_FORMAT_TIME, NULL);
+    g_object_set(G_OBJECT(source), "block", 1, NULL);
+    g_object_set(G_OBJECT(source), "is-live", 0, NULL);
+    g_object_set(G_OBJECT(source), "emit-signals", 1, NULL);
+
+    if(!manualpipeline)
+    {
+        g_object_set(G_OBJECT(file), "buffer-size", bufsize, NULL);
+        gst_bin_add_many(GST_BIN(pipeline), source, encodebin, file, NULL);
+        if(!gst_element_link_many(source, encodebin, file, NULL)) {
+            CV_ERROR(CV_StsError, "GStreamer: cannot link elements\n");
+        }
+    }
+
+    stateret = gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_PLAYING);
+    if(stateret  == GST_STATE_CHANGE_FAILURE) {
+        CV_ERROR(CV_StsError, "GStreamer: cannot put pipeline to play\n");
+    }
+    handleMessage(pipeline);
+
+    framerate = fps;
+    num_frames = 0;
+
+    __END__;
+
+    return true;
+}
+
+
+/*!
+ * \brief CvVideoWriter_GStreamer::writeFrame
+ * \param image
+ * \return
+ * Pushes the given frame on the pipeline.
+ * The timestamp for the buffer is generated from the framerate set in open
+ * and ensures a smooth video
+ */
+bool CvVideoWriter_GStreamer::writeFrame( const IplImage * image )
+{
+
+    CV_FUNCNAME("CvVideoWriter_GStreamer::writerFrame");
+
+    GstClockTime duration, timestamp;
+    GstFlowReturn ret;
+    int size;
+
+    __BEGIN__;
+    handleMessage(pipeline);
+
+    if (input_pix_fmt == GST_VIDEO_FORMAT_BGR) {
+        if (image->nChannels != 3 || image->depth != IPL_DEPTH_8U) {
+            CV_ERROR(CV_StsUnsupportedFormat, "cvWriteFrame() needs images with depth = IPL_DEPTH_8U and nChannels = 3.");
+        }
+    }
+#if FULL_GST_VERSION >= VERSION_NUM(0,10,29)
+    else if (input_pix_fmt == GST_VIDEO_FORMAT_GRAY8) {
+        if (image->nChannels != 1 || image->depth != IPL_DEPTH_8U) {
+            CV_ERROR(CV_StsUnsupportedFormat, "cvWriteFrame() needs images with depth = IPL_DEPTH_8U and nChannels = 1.");
+        }
+    }
+#endif
+    else {
+        assert(false);
+    }
+
+    size = image->imageSize;
+    duration = ((double)1/framerate) * GST_SECOND;
+    timestamp = num_frames * duration;
+
+    //gst_app_src_push_buffer takes ownership of the buffer, so we need to supply it a copy
+#if GST_VERSION_MAJOR == 0
+    buffer = gst_buffer_new_and_alloc (size);
+    memcpy(GST_BUFFER_DATA (buffer), (guint8*)image->imageData, size);
+    GST_BUFFER_DURATION(buffer) = duration;
+    GST_BUFFER_TIMESTAMP(buffer) = timestamp;
+#else
+    buffer = gst_buffer_new_allocate (NULL, size, NULL);
+    GstMapInfo info;
+    gst_buffer_map(buffer, &info, (GstMapFlags)GST_MAP_READ);
+    memcpy(info.data, (guint8*)image->imageData, size);
+    gst_buffer_unmap(buffer, &info);
+    GST_BUFFER_DURATION(buffer) = duration;
+    GST_BUFFER_PTS(buffer) = timestamp;
+    GST_BUFFER_DTS(buffer) = timestamp;
+#endif
+    //set the current number in the frame
+    GST_BUFFER_OFFSET(buffer) =  num_frames;
+
+    ret = gst_app_src_push_buffer(GST_APP_SRC(source), buffer);
+    if (ret != GST_FLOW_OK) {
+        /* something wrong, stop pushing */
+        assert(false);
+    }
+    //gst_debug_bin_to_dot_file (GST_BIN(pipeline), GST_DEBUG_GRAPH_SHOW_ALL, "pipeline");
+    ++num_frames;
+
+    __END__;
+    return true;
+}
+
+/*!
+ * \brief cvCreateVideoWriter_GStreamer
+ * \param filename
+ * \param fourcc
+ * \param fps
+ * \param frameSize
+ * \param isColor
+ * \return
+ * Constructor
+ */
+CvVideoWriter* cvCreateVideoWriter_GStreamer(const char* filename, int fourcc, double fps,
+                                             CvSize frameSize, int isColor )
+{
+    CvVideoWriter_GStreamer* wrt = new CvVideoWriter_GStreamer;
+    if( wrt->open(filename, fourcc, fps,frameSize, isColor))
+        return wrt;
+
+    delete wrt;
+    return 0;
+}
+
+// utility functions
+
+/*!
+ * \brief toFraction
+ * \param decimal
+ * \param numerator
+ * \param denominator
+ * Split a floating point value into numerator and denominator
+ */
+void toFraction(double decimal, double &numerator, double &denominator)
+{
+    double dummy;
+    double whole;
+    decimal = modf (decimal, &whole);
+    for (denominator = 1; denominator<=100; denominator++){
+        if (modf(denominator * decimal, &dummy) < 0.001f)
+            break;
+    }
+    numerator = denominator * decimal;
+}
+
+
+/*!
+ * \brief handleMessage
+ * Handles gstreamer bus messages. Mainly for debugging purposes and ensuring clean shutdown on error
+ */
+void handleMessage(GstElement * pipeline)
+{
+    CV_FUNCNAME("handlemessage");
+
+    GError *err = NULL;
+    gchar *debug = NULL;
+    GstBus* bus = NULL;
+    GstStreamStatusType tp;
+    GstElement * elem = NULL;
+    GstMessage* msg  = NULL;
+
+    __BEGIN__;
+    bus = gst_element_get_bus(pipeline);
+
+    while(gst_bus_have_pending(bus)) {
+        msg = gst_bus_pop(bus);
+
+        //printf("Got %s message\n", GST_MESSAGE_TYPE_NAME(msg));
+
+        if(gst_is_missing_plugin_message(msg))
+        {
+            CV_ERROR(CV_StsError, "GStreamer: your gstreamer installation is missing a required plugin\n");
+        }
+        else
+        {
+            switch (GST_MESSAGE_TYPE (msg)) {
+            case GST_MESSAGE_STATE_CHANGED:
+                GstState oldstate, newstate, pendstate;
+                gst_message_parse_state_changed(msg, &oldstate, &newstate, &pendstate);
+                //fprintf(stderr, "state changed from %s to %s (pending: %s)\n", gst_element_state_get_name(oldstate),
+                //                gst_element_state_get_name(newstate), gst_element_state_get_name(pendstate));
+                break;
+            case GST_MESSAGE_ERROR:
+                gst_message_parse_error(msg, &err, &debug);
+
+                //fprintf(stderr, "GStreamer Plugin: Embedded video playback halted; module %s reported: %s\n",
+                //                gst_element_get_name(GST_MESSAGE_SRC (msg)), err->message);
+
+                g_error_free(err);
+                g_free(debug);
+
+                gst_element_set_state(GST_ELEMENT(pipeline), GST_STATE_NULL);
+                break;
+            case GST_MESSAGE_EOS:
+                //fprintf(stderr, "reached the end of the stream.");
+                break;
+            case GST_MESSAGE_STREAM_STATUS:
+
+                gst_message_parse_stream_status(msg,&tp,&elem);
+                //fprintf(stderr, "stream status: elem %s, %i\n", GST_ELEMENT_NAME(elem), tp);
+                break;
+            default:
+                //fprintf(stderr, "unhandled message\n");
+                break;
+            }
+        }
+        gst_message_unref(msg);
+    }
+
+    gst_object_unref(GST_OBJECT(bus));
+
+    __END__
+}
diff --git a/modules/highgui/src/cap_ios_abstract_camera.mm b/modules/highgui/src/cap_ios_abstract_camera.mm
index b6a7d944fa..a77e200a82 100644
--- a/modules/highgui/src/cap_ios_abstract_camera.mm
+++ b/modules/highgui/src/cap_ios_abstract_camera.mm
@@ -2,6 +2,7 @@
  *  cap_ios_abstract_camera.mm
  *  For iOS video I/O
  *  by Eduard Feicho on 29/07/12
+ *  by Alexander Shishkov on 17/07/13
  *  Copyright 2012. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -405,4 +406,88 @@
     }
 }
 
+- (void)lockFocus;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isFocusModeSupported:AVCaptureFocusModeLocked]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.focusMode = AVCaptureFocusModeLocked;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for locked focus configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
+- (void) unlockFocus;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isFocusModeSupported:AVCaptureFocusModeContinuousAutoFocus]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.focusMode = AVCaptureFocusModeContinuousAutoFocus;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for autofocus configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
+- (void)lockExposure;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isExposureModeSupported:AVCaptureExposureModeLocked]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.exposureMode = AVCaptureExposureModeLocked;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for locked exposure configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
+- (void) unlockExposure;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isExposureModeSupported:AVCaptureExposureModeContinuousAutoExposure]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.exposureMode = AVCaptureExposureModeContinuousAutoExposure;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for autoexposure configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
+- (void)lockBalance;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isWhiteBalanceModeSupported:AVCaptureWhiteBalanceModeLocked]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.whiteBalanceMode = AVCaptureWhiteBalanceModeLocked;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for locked white balance configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
+- (void) unlockBalance;
+{
+    AVCaptureDevice *device = [AVCaptureDevice defaultDeviceWithMediaType:AVMediaTypeVideo];
+    if ([device isWhiteBalanceModeSupported:AVCaptureWhiteBalanceModeContinuousAutoWhiteBalance]) {
+        NSError *error = nil;
+        if ([device lockForConfiguration:&error]) {
+            device.whiteBalanceMode = AVCaptureWhiteBalanceModeContinuousAutoWhiteBalance;
+            [device unlockForConfiguration];
+        } else {
+            NSLog(@"unable to lock device for auto white balance configuration %@", [error localizedDescription]);
+        }
+    }
+}
+
 @end
diff --git a/modules/highgui/src/cap_ios_video_camera.mm b/modules/highgui/src/cap_ios_video_camera.mm
index 1f9ea14bf8..99f8a75ae6 100644
--- a/modules/highgui/src/cap_ios_video_camera.mm
+++ b/modules/highgui/src/cap_ios_video_camera.mm
@@ -2,6 +2,7 @@
  *  cap_ios_video_camera.mm
  *  For iOS video I/O
  *  by Eduard Feicho on 29/07/12
+ *  by Alexander Shishkov on 17/07/13
  *  Copyright 2012. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -30,7 +31,6 @@
 
 #import "opencv2/highgui/cap_ios.h"
 #include "precomp.hpp"
-
 #import <AssetsLibrary/AssetsLibrary.h>
 
 
@@ -70,6 +70,7 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
 @synthesize videoDataOutput;
 
 @synthesize recordVideo;
+@synthesize rotateVideo;
 //@synthesize videoFileOutput;
 @synthesize recordAssetWriterInput;
 @synthesize recordPixelBufferAdaptor;
@@ -85,6 +86,7 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
     if (self) {
         self.useAVCaptureVideoPreviewLayer = NO;
         self.recordVideo = NO;
+        self.rotateVideo = NO;
     }
     return self;
 }
@@ -269,13 +271,8 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
 
 }
 
-
-
-
 #pragma mark - Private Interface
 
-
-
 - (void)createVideoDataOutput;
 {
     // Make a video data output
@@ -389,6 +386,38 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
     [self.parentView.layer addSublayer:self.customPreviewLayer];
 }
 
+- (CVPixelBufferRef) pixelBufferFromCGImage: (CGImageRef) image
+{
+
+    CGSize frameSize = CGSizeMake(CGImageGetWidth(image), CGImageGetHeight(image));
+    NSDictionary *options = [NSDictionary dictionaryWithObjectsAndKeys:
+                             [NSNumber numberWithBool:NO], kCVPixelBufferCGImageCompatibilityKey,
+                             [NSNumber numberWithBool:NO], kCVPixelBufferCGBitmapContextCompatibilityKey,
+                             nil];
+    CVPixelBufferRef pxbuffer = NULL;
+    CVReturn status = CVPixelBufferCreate(kCFAllocatorDefault, frameSize.width,
+                                          frameSize.height,  kCVPixelFormatType_32ARGB, (CFDictionaryRef) CFBridgingRetain(options),
+                                          &pxbuffer);
+    NSParameterAssert(status == kCVReturnSuccess && pxbuffer != NULL);
+
+    CVPixelBufferLockBaseAddress(pxbuffer, 0);
+    void *pxdata = CVPixelBufferGetBaseAddress(pxbuffer);
+
+
+    CGColorSpaceRef rgbColorSpace = CGColorSpaceCreateDeviceRGB();
+    CGContextRef context = CGBitmapContextCreate(pxdata, frameSize.width,
+                                                 frameSize.height, 8, 4*frameSize.width, rgbColorSpace,
+                                                 kCGImageAlphaPremultipliedFirst);
+
+    CGContextDrawImage(context, CGRectMake(0, 0, CGImageGetWidth(image),
+                                           CGImageGetHeight(image)), image);
+    CGColorSpaceRelease(rgbColorSpace);
+    CGContextRelease(context);
+
+    CVPixelBufferUnlockBaseAddress(pxbuffer, 0);
+
+    return pxbuffer;
+}
 
 #pragma mark - Protocol AVCaptureVideoDataOutputSampleBufferDelegate
 
@@ -522,7 +551,8 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
             }
 
             if (self.recordAssetWriterInput.readyForMoreMediaData) {
-                if (! [self.recordPixelBufferAdaptor appendPixelBuffer:imageBuffer
+                CVImageBufferRef pixelBuffer = [self pixelBufferFromCGImage:dstImage];
+                if (! [self.recordPixelBufferAdaptor appendPixelBuffer:pixelBuffer
                                                   withPresentationTime:lastSampleTime] ) {
                     NSLog(@"Video Writing Error");
                 }
@@ -543,9 +573,12 @@ static CGFloat DegreesToRadians(CGFloat degrees) {return degrees * M_PI / 180;};
 
 - (void)updateOrientation;
 {
-    NSLog(@"rotate..");
-    self.customPreviewLayer.bounds = CGRectMake(0, 0, self.parentView.frame.size.width, self.parentView.frame.size.height);
-    [self layoutPreviewLayer];
+    if (self.rotateVideo == YES)
+    {
+        NSLog(@"rotate..");
+        self.customPreviewLayer.bounds = CGRectMake(0, 0, self.parentView.frame.size.width, self.parentView.frame.size.height);
+        [self layoutPreviewLayer];
+    }
 }
 
 
diff --git a/modules/highgui/src/cap_libv4l.cpp b/modules/highgui/src/cap_libv4l.cpp
index 63a2ff96b0..3fd6dd59ad 100644
--- a/modules/highgui/src/cap_libv4l.cpp
+++ b/modules/highgui/src/cap_libv4l.cpp
@@ -14,7 +14,9 @@ It has been tested with the motempl sample program
 First Patch:  August 24, 2004 Travis Wood   TravisOCV@tkwood.com
 For Release:  OpenCV-Linux Beta4  opencv-0.9.6
 Tested On:    LMLBT44 with 8 video inputs
-Problems?     Post problems/fixes to OpenCV group on groups.yahoo.com
+Problems?     Post your questions at answers.opencv.org,
+              Report bugs at code.opencv.org,
+              Submit your fixes at https://github.com/Itseez/opencv/
 Patched Comments:
 
 TW: The cv cam utils that came with the initial release of OpenCV for LINUX Beta4
@@ -1665,6 +1667,17 @@ static int icvSetPropertyCAM_V4L(CvCaptureCAM_V4L* capture, int property_id, dou
             width = height = 0;
         }
         break;
+    case CV_CAP_PROP_FPS:
+        struct v4l2_streamparm setfps;
+        memset (&setfps, 0, sizeof(struct v4l2_streamparm));
+        setfps.type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
+        setfps.parm.capture.timeperframe.numerator = 1;
+        setfps.parm.capture.timeperframe.denominator = value;
+        if (xioctl (capture->deviceHandle, VIDIOC_S_PARM, &setfps) < 0){
+            fprintf(stderr, "HIGHGUI ERROR: V4L: Unable to set camera FPS\n");
+            retval=0;
+        }
+        break;
     default:
         retval = icvSetControl(capture, property_id, value);
     }
@@ -1714,6 +1727,7 @@ static void icvCloseCAM_V4L( CvCaptureCAM_V4L* capture ){
 #endif
 
      free(capture->deviceName);
+     capture->deviceName = NULL;
      //v4l2_free_ranges(capture);
      //cvFree((void **)capture);
    }
diff --git a/modules/highgui/src/cap_msmf.cpp b/modules/highgui/src/cap_msmf.cpp
new file mode 100644
index 0000000000..df52f04bdf
--- /dev/null
+++ b/modules/highgui/src/cap_msmf.cpp
@@ -0,0 +1,3535 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "precomp.hpp"
+#if (defined WIN32 || defined _WIN32) && defined HAVE_MSMF
+/*
+   Media Foundation-based Video Capturing module is based on
+   videoInput library by Evgeny Pereguda:
+   http://www.codeproject.com/Articles/559437/Capturing-of-video-from-web-camera-on-Windows-7-an
+   Originaly licensed under The Code Project Open License (CPOL) 1.02:
+   http://www.codeproject.com/info/cpol10.aspx
+*/
+#include <windows.h>
+#include <guiddef.h>
+#include <mfidl.h>
+#include <Mfapi.h>
+#include <mfplay.h>
+#include <mfobjects.h>
+#include <strsafe.h>
+#include <Mfreadwrite.h>
+#include <new>
+#include <map>
+#include <vector>
+#include <string>
+#include <stdio.h>
+#include <stdarg.h>
+#include <string.h>
+
+#pragma warning(disable:4503)
+#pragma comment(lib, "mfplat")
+#pragma comment(lib, "mf")
+#pragma comment(lib, "mfuuid")
+#pragma comment(lib, "Strmiids")
+#pragma comment(lib, "Mfreadwrite")
+#pragma comment(lib, "MinCore_Downlevel")
+
+// for ComPtr usage
+#include <wrl/client.h>
+using namespace Microsoft::WRL;
+
+struct IMFMediaType;
+struct IMFActivate;
+struct IMFMediaSource;
+struct IMFAttributes;
+
+namespace
+{
+
+template <class T> void SafeRelease(T **ppT)
+{
+    if (*ppT)
+    {
+        (*ppT)->Release();
+        *ppT = NULL;
+    }
+}
+
+/// Class for printing info into consol
+class DebugPrintOut
+{
+public:
+    ~DebugPrintOut(void);
+    static DebugPrintOut& getInstance();
+    void printOut(const wchar_t *format, ...);
+    void setVerbose(bool state);
+    bool verbose;
+private:
+    DebugPrintOut(void);
+};
+
+// Structure for collecting info about types of video, which are supported by current video device
+struct MediaType
+{
+    unsigned int MF_MT_FRAME_SIZE;
+    unsigned int height;
+    unsigned int width;
+    unsigned int MF_MT_YUV_MATRIX;
+    unsigned int MF_MT_VIDEO_LIGHTING;
+    int MF_MT_DEFAULT_STRIDE; // stride is negative if image is bottom-up
+    unsigned int MF_MT_VIDEO_CHROMA_SITING;
+    GUID MF_MT_AM_FORMAT_TYPE;
+    wchar_t *pMF_MT_AM_FORMAT_TYPEName;
+    unsigned int MF_MT_FIXED_SIZE_SAMPLES;
+    unsigned int MF_MT_VIDEO_NOMINAL_RANGE;
+    unsigned int MF_MT_FRAME_RATE_NUMERATOR;
+    unsigned int MF_MT_FRAME_RATE_DENOMINATOR;
+    unsigned int MF_MT_PIXEL_ASPECT_RATIO;
+    unsigned int MF_MT_PIXEL_ASPECT_RATIO_low;
+    unsigned int MF_MT_ALL_SAMPLES_INDEPENDENT;
+    unsigned int MF_MT_FRAME_RATE_RANGE_MIN;
+    unsigned int MF_MT_FRAME_RATE_RANGE_MIN_low;
+    unsigned int MF_MT_SAMPLE_SIZE;
+    unsigned int MF_MT_VIDEO_PRIMARIES;
+    unsigned int MF_MT_INTERLACE_MODE;
+    unsigned int MF_MT_FRAME_RATE_RANGE_MAX;
+    unsigned int MF_MT_FRAME_RATE_RANGE_MAX_low;
+    GUID MF_MT_MAJOR_TYPE;
+    GUID MF_MT_SUBTYPE;
+    wchar_t *pMF_MT_MAJOR_TYPEName;
+    wchar_t *pMF_MT_SUBTYPEName;
+    MediaType();
+    ~MediaType();
+    void Clear();
+};
+
+/// Class for parsing info from IMFMediaType into the local MediaType
+class FormatReader
+{
+public:
+    static MediaType Read(IMFMediaType *pType);
+    ~FormatReader(void);
+private:
+    FormatReader(void);
+};
+
+DWORD WINAPI MainThreadFunction( LPVOID lpParam );
+typedef void(*emergensyStopEventCallback)(int, void *);
+
+class RawImage
+{
+public:
+    ~RawImage(void);
+    // Function of creation of the instance of the class
+    static long CreateInstance(RawImage **ppRImage,unsigned int size);
+    void setCopy(const BYTE * pSampleBuffer);
+    void fastCopy(const BYTE * pSampleBuffer);
+    unsigned char * getpPixels();
+    bool isNew();
+    unsigned int getSize();
+private:
+    bool ri_new;
+    unsigned int ri_size;
+    unsigned char *ri_pixels;
+    RawImage(unsigned int size);
+};
+
+// Class for grabbing image from video stream
+class ImageGrabber : public IMFSampleGrabberSinkCallback
+{
+public:
+    ~ImageGrabber(void);
+    HRESULT initImageGrabber(IMFMediaSource *pSource, GUID VideoFormat);
+    HRESULT startGrabbing(void);
+    void pauseGrabbing();
+    void resumeGrabbing();
+    void stopGrabbing();
+    RawImage *getRawImage();
+    // Function of creation of the instance of the class
+    static HRESULT CreateInstance(ImageGrabber **ppIG, unsigned int deviceID, bool synchronous = false);
+
+    const HANDLE ig_hFrameReady;
+    const HANDLE ig_hFrameGrabbed;
+    const HANDLE ig_hFinish;
+
+private:
+    bool ig_RIE;
+    bool ig_Close;
+    bool ig_Synchronous;
+    long m_cRef;
+    unsigned int ig_DeviceID;
+    IMFMediaSource *ig_pSource;
+    IMFMediaSession *ig_pSession;
+    IMFTopology *ig_pTopology;
+    RawImage *ig_RIFirst;
+    RawImage *ig_RISecond;
+    RawImage *ig_RIOut;
+    ImageGrabber(unsigned int deviceID, bool synchronous);
+    HRESULT CreateTopology(IMFMediaSource *pSource, IMFActivate *pSinkActivate, IMFTopology **ppTopo);
+    HRESULT AddSourceNode(IMFTopology *pTopology, IMFMediaSource *pSource,
+        IMFPresentationDescriptor *pPD, IMFStreamDescriptor *pSD, IMFTopologyNode **ppNode);
+    HRESULT AddOutputNode(IMFTopology *pTopology, IMFActivate *pActivate, DWORD dwId, IMFTopologyNode **ppNode);
+    // IUnknown methods
+    STDMETHODIMP QueryInterface(REFIID iid, void** ppv);
+    STDMETHODIMP_(ULONG) AddRef();
+    STDMETHODIMP_(ULONG) Release();
+    // IMFClockStateSink methods
+    STDMETHODIMP OnClockStart(MFTIME hnsSystemTime, LONGLONG llClockStartOffset);
+    STDMETHODIMP OnClockStop(MFTIME hnsSystemTime);
+    STDMETHODIMP OnClockPause(MFTIME hnsSystemTime);
+    STDMETHODIMP OnClockRestart(MFTIME hnsSystemTime);
+    STDMETHODIMP OnClockSetRate(MFTIME hnsSystemTime, float flRate);
+    // IMFSampleGrabberSinkCallback methods
+    STDMETHODIMP OnSetPresentationClock(IMFPresentationClock* pClock);
+    STDMETHODIMP OnProcessSample(REFGUID guidMajorMediaType, DWORD dwSampleFlags,
+        LONGLONG llSampleTime, LONGLONG llSampleDuration, const BYTE * pSampleBuffer,
+        DWORD dwSampleSize);
+    STDMETHODIMP OnShutdown();
+};
+
+/// Class for controlling of thread of the grabbing raw data from video device
+class ImageGrabberThread
+{
+    friend DWORD WINAPI MainThreadFunction( LPVOID lpParam );
+public:
+    ~ImageGrabberThread(void);
+    static HRESULT CreateInstance(ImageGrabberThread **ppIGT, IMFMediaSource *pSource, unsigned int deviceID, bool synchronious = false);
+    void start();
+    void stop();
+    void setEmergencyStopEvent(void *userData, void(*func)(int, void *));
+    ImageGrabber *getImageGrabber();
+protected:
+    virtual void run();
+private:
+    ImageGrabberThread(IMFMediaSource *pSource, unsigned int deviceID, bool synchronious);
+    HANDLE igt_Handle;
+    DWORD   igt_ThreadIdArray;
+    ImageGrabber *igt_pImageGrabber;
+    emergensyStopEventCallback igt_func;
+    void *igt_userData;
+    bool igt_stop;
+    unsigned int igt_DeviceID;
+};
+
+// Structure for collecting info about one parametr of current video device
+struct Parametr
+{
+    long CurrentValue;
+    long Min;
+    long Max;
+    long Step;
+    long Default;
+    long Flag;
+    Parametr();
+};
+
+// Structure for collecting info about 17 parametrs of current video device
+struct CamParametrs
+{
+        Parametr Brightness;
+        Parametr Contrast;
+        Parametr Hue;
+        Parametr Saturation;
+        Parametr Sharpness;
+        Parametr Gamma;
+        Parametr ColorEnable;
+        Parametr WhiteBalance;
+        Parametr BacklightCompensation;
+        Parametr Gain;
+        Parametr Pan;
+        Parametr Tilt;
+        Parametr Roll;
+        Parametr Zoom;
+        Parametr Exposure;
+        Parametr Iris;
+        Parametr Focus;
+};
+
+typedef std::wstring String;
+typedef std::vector<int> vectorNum;
+typedef std::map<String, vectorNum> SUBTYPEMap;
+typedef std::map<UINT64, SUBTYPEMap> FrameRateMap;
+typedef void(*emergensyStopEventCallback)(int, void *);
+
+/// Class for controlling of video device
+class videoDevice
+{
+public:
+    videoDevice(void);
+    ~videoDevice(void);
+    void closeDevice();
+    CamParametrs getParametrs();
+    void setParametrs(CamParametrs parametrs);
+    void setEmergencyStopEvent(void *userData, void(*func)(int, void *));
+    long readInfoOfDevice(IMFActivate *pActivate, unsigned int Num);
+    wchar_t *getName();
+    int getCountFormats();
+    unsigned int getWidth();
+    unsigned int getHeight();
+    MediaType getFormat(unsigned int id);
+    bool setupDevice(unsigned int w, unsigned int h, unsigned int idealFramerate = 0);
+    bool setupDevice(unsigned int id);
+    bool isDeviceSetup();
+    bool isDeviceMediaSource();
+    bool isDeviceRawDataSource();
+    bool isFrameNew();
+    IMFMediaSource *getMediaSource();
+    RawImage *getRawImageOut();
+private:
+    enum typeLock
+    {
+        MediaSourceLock,
+        RawDataLock,
+        OpenLock
+    } vd_LockOut;
+    wchar_t *vd_pFriendlyName;
+    ImageGrabberThread *vd_pImGrTh;
+    CamParametrs vd_PrevParametrs;
+    unsigned int vd_Width;
+    unsigned int vd_Height;
+    unsigned int vd_CurrentNumber;
+    bool vd_IsSetuped;
+    std::map<UINT64, FrameRateMap> vd_CaptureFormats;
+    std::vector<MediaType> vd_CurrentFormats;
+    IMFMediaSource *vd_pSource;
+    emergensyStopEventCallback vd_func;
+    void *vd_userData;
+    HRESULT enumerateCaptureFormats(IMFMediaSource *pSource);
+    long setDeviceFormat(IMFMediaSource *pSource, unsigned long dwFormatIndex);
+    void buildLibraryofTypes();
+    int findType(unsigned int size, unsigned int frameRate = 0);
+    long resetDevice(IMFActivate *pActivate);
+    long initDevice();
+    long checkDevice(IMFAttributes *pAttributes, IMFActivate **pDevice);
+};
+
+/// Class for managing of list of video devices
+class videoDevices
+{
+public:
+    ~videoDevices(void);
+    long initDevices(IMFAttributes *pAttributes);
+    static videoDevices& getInstance();
+    videoDevice *getDevice(unsigned int i);
+    unsigned int getCount();
+    void clearDevices();
+private:
+    UINT32 count;
+    std::vector<videoDevice *> vds_Devices;
+    videoDevices(void);
+};
+
+// Class for creating of Media Foundation context
+class Media_Foundation
+{
+public:
+    virtual ~Media_Foundation(void);
+    static Media_Foundation& getInstance();
+    bool buildListOfDevices();
+private:
+    Media_Foundation(void);
+};
+
+/// The only visiable class for controlling of video devices in format singelton
+class videoInput
+{
+public:
+    virtual ~videoInput(void);
+    // Getting of static instance of videoInput class
+    static videoInput& getInstance();
+    // Closing video device with deviceID
+    void closeDevice(int deviceID);
+    // Setting callback function for emergency events(for example: removing video device with deviceID) with userData
+    void setEmergencyStopEvent(int deviceID, void *userData, void(*func)(int, void *));
+    // Closing all devices
+    void closeAllDevices();
+    // Getting of parametrs of video device with deviceID
+    CamParametrs getParametrs(int deviceID);
+    // Setting of parametrs of video device with deviceID
+    void setParametrs(int deviceID, CamParametrs parametrs);
+    // Getting numbers of existence videodevices with listing in consol
+    unsigned int listDevices(bool silent = false);
+    // Getting numbers of formats, which are supported by videodevice with deviceID
+    unsigned int getCountFormats(int deviceID);
+    // Getting width of image, which is getting from videodevice with deviceID
+    unsigned int getWidth(int deviceID);
+    // Getting height of image, which is getting from videodevice with deviceID
+    unsigned int getHeight(int deviceID);
+    // Getting name of videodevice with deviceID
+    wchar_t *getNameVideoDevice(int deviceID);
+    // Getting interface MediaSource for Media Foundation from videodevice with deviceID
+    IMFMediaSource *getMediaSource(int deviceID);
+    // Getting format with id, which is supported by videodevice with deviceID
+    MediaType getFormat(int deviceID, int unsigned id);
+    // Checking of existence of the suitable video devices
+    bool isDevicesAcceable();
+    // Checking of using the videodevice with deviceID
+    bool isDeviceSetup(int deviceID);
+    // Checking of using MediaSource from videodevice with deviceID
+    bool isDeviceMediaSource(int deviceID);
+    // Checking of using Raw Data of pixels from videodevice with deviceID
+    bool isDeviceRawDataSource(int deviceID);
+    // Setting of the state of outprinting info in console
+    static void setVerbose(bool state);
+    // Initialization of video device with deviceID by media type with id
+    bool setupDevice(int deviceID, unsigned int id = 0);
+    // Initialization of video device with deviceID by wisth w, height h and fps idealFramerate
+    bool setupDevice(int deviceID, unsigned int w, unsigned int h, unsigned int idealFramerate = 30);
+    // Checking of recivig of new frame from video device with deviceID
+    bool isFrameNew(int deviceID);
+    // Writing of Raw Data pixels from video device with deviceID with correction of RedAndBlue flipping flipRedAndBlue and vertical flipping flipImage
+    bool getPixels(int deviceID, unsigned char * pixels, bool flipRedAndBlue = false, bool flipImage = false);
+    static void processPixels(unsigned char * src, unsigned char * dst, unsigned int width, unsigned int height, unsigned int bpp, bool bRGB, bool bFlip);
+private:
+    bool accessToDevices;
+    videoInput(void);
+    void updateListOfDevices();
+};
+
+DebugPrintOut::DebugPrintOut(void):verbose(true)
+{
+}
+
+DebugPrintOut::~DebugPrintOut(void)
+{
+}
+
+DebugPrintOut& DebugPrintOut::getInstance()
+{
+    static DebugPrintOut instance;
+    return instance;
+}
+
+void DebugPrintOut::printOut(const wchar_t *format, ...)
+{
+    if(verbose)
+    {
+        int i = 0;
+        wchar_t *p = NULL;
+        va_list args;
+        va_start(args, format);
+        if(wcscmp(format, L"%i"))
+        {
+            i = va_arg (args, int);
+        }
+        if(wcscmp(format, L"%s"))
+        {
+            p = va_arg (args, wchar_t *);
+        }
+        wprintf(format, i,p);
+        va_end (args);
+    }
+}
+
+void DebugPrintOut::setVerbose(bool state)
+{
+    verbose = state;
+}
+
+LPCWSTR GetGUIDNameConstNew(const GUID& guid);
+HRESULT GetGUIDNameNew(const GUID& guid, WCHAR **ppwsz);
+HRESULT LogAttributeValueByIndexNew(IMFAttributes *pAttr, DWORD index);
+HRESULT SpecialCaseAttributeValueNew(GUID guid, const PROPVARIANT& var, MediaType &out);
+
+unsigned int *GetParametr(GUID guid, MediaType &out)
+{
+    if(guid == MF_MT_YUV_MATRIX)
+        return &(out.MF_MT_YUV_MATRIX);
+    if(guid == MF_MT_VIDEO_LIGHTING)
+        return &(out.MF_MT_VIDEO_LIGHTING);
+    if(guid == MF_MT_DEFAULT_STRIDE)
+        return (unsigned int*)&(out.MF_MT_DEFAULT_STRIDE);
+    if(guid == MF_MT_VIDEO_CHROMA_SITING)
+        return &(out.MF_MT_VIDEO_CHROMA_SITING);
+    if(guid == MF_MT_VIDEO_NOMINAL_RANGE)
+        return &(out.MF_MT_VIDEO_NOMINAL_RANGE);
+    if(guid == MF_MT_ALL_SAMPLES_INDEPENDENT)
+        return &(out.MF_MT_ALL_SAMPLES_INDEPENDENT);
+    if(guid == MF_MT_FIXED_SIZE_SAMPLES)
+        return &(out.MF_MT_FIXED_SIZE_SAMPLES);
+    if(guid == MF_MT_SAMPLE_SIZE)
+        return &(out.MF_MT_SAMPLE_SIZE);
+    if(guid == MF_MT_VIDEO_PRIMARIES)
+        return &(out.MF_MT_VIDEO_PRIMARIES);
+    if(guid == MF_MT_INTERLACE_MODE)
+        return &(out.MF_MT_INTERLACE_MODE);
+    return NULL;
+}
+
+HRESULT LogAttributeValueByIndexNew(IMFAttributes *pAttr, DWORD index, MediaType &out)
+{
+    WCHAR *pGuidName = NULL;
+    WCHAR *pGuidValName = NULL;
+    GUID guid = { 0 };
+    PROPVARIANT var;
+    PropVariantInit(&var);
+    HRESULT hr = pAttr->GetItemByIndex(index, &guid, &var);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = GetGUIDNameNew(guid, &pGuidName);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = SpecialCaseAttributeValueNew(guid, var, out);
+    unsigned int *p;
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    if (hr == S_FALSE)
+    {
+        switch (var.vt)
+        {
+        case VT_UI4:
+            p = GetParametr(guid, out);
+            if(p)
+            {
+                *p = var.ulVal;
+            }
+            break;
+        case VT_UI8:
+            break;
+        case VT_R8:
+            break;
+        case VT_CLSID:
+            if(guid == MF_MT_AM_FORMAT_TYPE)
+            {
+                hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
+                if (SUCCEEDED(hr))
+                {
+                    out.MF_MT_AM_FORMAT_TYPE = MF_MT_AM_FORMAT_TYPE;
+                    out.pMF_MT_AM_FORMAT_TYPEName = pGuidValName;
+                    pGuidValName = NULL;
+                }
+            }
+            if(guid == MF_MT_MAJOR_TYPE)
+            {
+                hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
+                if (SUCCEEDED(hr))
+                {
+                    out.MF_MT_MAJOR_TYPE = MF_MT_MAJOR_TYPE;
+                    out.pMF_MT_MAJOR_TYPEName = pGuidValName;
+                    pGuidValName = NULL;
+                }
+            }
+            if(guid == MF_MT_SUBTYPE)
+            {
+                hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
+                if (SUCCEEDED(hr))
+                {
+                    out.MF_MT_SUBTYPE = MF_MT_SUBTYPE;
+                    out.pMF_MT_SUBTYPEName = pGuidValName;
+                    pGuidValName = NULL;
+                }
+            }
+            break;
+        case VT_LPWSTR:
+            break;
+        case VT_VECTOR | VT_UI1:
+            break;
+        case VT_UNKNOWN:
+            break;
+        default:
+            break;
+        }
+    }
+done:
+    CoTaskMemFree(pGuidName);
+    CoTaskMemFree(pGuidValName);
+    PropVariantClear(&var);
+    return hr;
+}
+
+HRESULT GetGUIDNameNew(const GUID& guid, WCHAR **ppwsz)
+{
+    HRESULT hr = S_OK;
+    WCHAR *pName = NULL;
+    LPCWSTR pcwsz = GetGUIDNameConstNew(guid);
+    if (pcwsz)
+    {
+        size_t cchLength = 0;
+        hr = StringCchLengthW(pcwsz, STRSAFE_MAX_CCH, &cchLength);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+        pName = (WCHAR*)CoTaskMemAlloc((cchLength + 1) * sizeof(WCHAR));
+        if (pName == NULL)
+        {
+            hr = E_OUTOFMEMORY;
+            goto done;
+        }
+        hr = StringCchCopyW(pName, cchLength + 1, pcwsz);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+    }
+    else
+    {
+        hr = StringFromCLSID(guid, &pName);
+    }
+done:
+    if (FAILED(hr))
+    {
+        *ppwsz = NULL;
+        CoTaskMemFree(pName);
+    }
+    else
+    {
+        *ppwsz = pName;
+    }
+    return hr;
+}
+
+void LogUINT32AsUINT64New(const PROPVARIANT& var, UINT32 &uHigh, UINT32 &uLow)
+{
+    Unpack2UINT32AsUINT64(var.uhVal.QuadPart, &uHigh, &uLow);
+}
+
+float OffsetToFloatNew(const MFOffset& offset)
+{
+    return offset.value + (static_cast<float>(offset.fract) / 65536.0f);
+}
+
+HRESULT LogVideoAreaNew(const PROPVARIANT& var)
+{
+    if (var.caub.cElems < sizeof(MFVideoArea))
+    {
+        return S_OK;
+    }
+    return S_OK;
+}
+
+HRESULT SpecialCaseAttributeValueNew(GUID guid, const PROPVARIANT& var, MediaType &out)
+{
+    if (guid == MF_MT_DEFAULT_STRIDE)
+    {
+        out.MF_MT_DEFAULT_STRIDE = var.intVal;
+    } else
+    if (guid == MF_MT_FRAME_SIZE)
+    {
+        UINT32 uHigh = 0, uLow = 0;
+        LogUINT32AsUINT64New(var, uHigh, uLow);
+        out.width = uHigh;
+        out.height = uLow;
+        out.MF_MT_FRAME_SIZE = out.width * out.height;
+    }
+    else
+    if (guid == MF_MT_FRAME_RATE)
+    {
+        UINT32 uHigh = 0, uLow = 0;
+        LogUINT32AsUINT64New(var, uHigh, uLow);
+        out.MF_MT_FRAME_RATE_NUMERATOR = uHigh;
+        out.MF_MT_FRAME_RATE_DENOMINATOR = uLow;
+    }
+    else
+    if (guid == MF_MT_FRAME_RATE_RANGE_MAX)
+    {
+        UINT32 uHigh = 0, uLow = 0;
+        LogUINT32AsUINT64New(var, uHigh, uLow);
+        out.MF_MT_FRAME_RATE_RANGE_MAX = uHigh;
+        out.MF_MT_FRAME_RATE_RANGE_MAX_low = uLow;
+    }
+    else
+    if (guid == MF_MT_FRAME_RATE_RANGE_MIN)
+    {
+        UINT32 uHigh = 0, uLow = 0;
+        LogUINT32AsUINT64New(var, uHigh, uLow);
+        out.MF_MT_FRAME_RATE_RANGE_MIN = uHigh;
+        out.MF_MT_FRAME_RATE_RANGE_MIN_low = uLow;
+    }
+    else
+    if (guid == MF_MT_PIXEL_ASPECT_RATIO)
+    {
+        UINT32 uHigh = 0, uLow = 0;
+        LogUINT32AsUINT64New(var, uHigh, uLow);
+        out.MF_MT_PIXEL_ASPECT_RATIO = uHigh;
+        out.MF_MT_PIXEL_ASPECT_RATIO_low = uLow;
+    }
+    else
+    {
+        return S_FALSE;
+    }
+    return S_OK;
+}
+
+#ifndef IF_EQUAL_RETURN
+#define IF_EQUAL_RETURN(param, val) if(val == param) return L#val
+#endif
+
+LPCWSTR GetGUIDNameConstNew(const GUID& guid)
+{
+    IF_EQUAL_RETURN(guid, MF_MT_MAJOR_TYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_MAJOR_TYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_SUBTYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_ALL_SAMPLES_INDEPENDENT);
+    IF_EQUAL_RETURN(guid, MF_MT_FIXED_SIZE_SAMPLES);
+    IF_EQUAL_RETURN(guid, MF_MT_COMPRESSED);
+    IF_EQUAL_RETURN(guid, MF_MT_SAMPLE_SIZE);
+    IF_EQUAL_RETURN(guid, MF_MT_WRAPPED_TYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_NUM_CHANNELS);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_SAMPLES_PER_SECOND);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_FLOAT_SAMPLES_PER_SECOND);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_AVG_BYTES_PER_SECOND);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_BLOCK_ALIGNMENT);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_BITS_PER_SAMPLE);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_VALID_BITS_PER_SAMPLE);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_SAMPLES_PER_BLOCK);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_CHANNEL_MASK);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_FOLDDOWN_MATRIX);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_PEAKREF);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_PEAKTARGET);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_AVGREF);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_AVGTARGET);
+    IF_EQUAL_RETURN(guid, MF_MT_AUDIO_PREFER_WAVEFORMATEX);
+    IF_EQUAL_RETURN(guid, MF_MT_AAC_PAYLOAD_TYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_AAC_AUDIO_PROFILE_LEVEL_INDICATION);
+    IF_EQUAL_RETURN(guid, MF_MT_FRAME_SIZE);
+    IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE);
+    IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE_RANGE_MAX);
+    IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE_RANGE_MIN);
+    IF_EQUAL_RETURN(guid, MF_MT_PIXEL_ASPECT_RATIO);
+    IF_EQUAL_RETURN(guid, MF_MT_DRM_FLAGS);
+    IF_EQUAL_RETURN(guid, MF_MT_PAD_CONTROL_FLAGS);
+    IF_EQUAL_RETURN(guid, MF_MT_SOURCE_CONTENT_HINT);
+    IF_EQUAL_RETURN(guid, MF_MT_VIDEO_CHROMA_SITING);
+    IF_EQUAL_RETURN(guid, MF_MT_INTERLACE_MODE);
+    IF_EQUAL_RETURN(guid, MF_MT_TRANSFER_FUNCTION);
+    IF_EQUAL_RETURN(guid, MF_MT_VIDEO_PRIMARIES);
+    IF_EQUAL_RETURN(guid, MF_MT_CUSTOM_VIDEO_PRIMARIES);
+    IF_EQUAL_RETURN(guid, MF_MT_YUV_MATRIX);
+    IF_EQUAL_RETURN(guid, MF_MT_VIDEO_LIGHTING);
+    IF_EQUAL_RETURN(guid, MF_MT_VIDEO_NOMINAL_RANGE);
+    IF_EQUAL_RETURN(guid, MF_MT_GEOMETRIC_APERTURE);
+    IF_EQUAL_RETURN(guid, MF_MT_MINIMUM_DISPLAY_APERTURE);
+    IF_EQUAL_RETURN(guid, MF_MT_PAN_SCAN_APERTURE);
+    IF_EQUAL_RETURN(guid, MF_MT_PAN_SCAN_ENABLED);
+    IF_EQUAL_RETURN(guid, MF_MT_AVG_BITRATE);
+    IF_EQUAL_RETURN(guid, MF_MT_AVG_BIT_ERROR_RATE);
+    IF_EQUAL_RETURN(guid, MF_MT_MAX_KEYFRAME_SPACING);
+    IF_EQUAL_RETURN(guid, MF_MT_DEFAULT_STRIDE);
+    IF_EQUAL_RETURN(guid, MF_MT_PALETTE);
+    IF_EQUAL_RETURN(guid, MF_MT_USER_DATA);
+    IF_EQUAL_RETURN(guid, MF_MT_AM_FORMAT_TYPE);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG_START_TIME_CODE);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG2_PROFILE);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG2_LEVEL);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG2_FLAGS);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG_SEQUENCE_HEADER);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_SRC_PACK_0);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_CTRL_PACK_0);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_SRC_PACK_1);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_CTRL_PACK_1);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_VAUX_SRC_PACK);
+    IF_EQUAL_RETURN(guid, MF_MT_DV_VAUX_CTRL_PACK);
+    IF_EQUAL_RETURN(guid, MF_MT_ARBITRARY_HEADER);
+    IF_EQUAL_RETURN(guid, MF_MT_ARBITRARY_FORMAT);
+    IF_EQUAL_RETURN(guid, MF_MT_IMAGE_LOSS_TOLERANT);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG4_SAMPLE_DESCRIPTION);
+    IF_EQUAL_RETURN(guid, MF_MT_MPEG4_CURRENT_SAMPLE_ENTRY);
+    IF_EQUAL_RETURN(guid, MF_MT_ORIGINAL_4CC);
+    IF_EQUAL_RETURN(guid, MF_MT_ORIGINAL_WAVE_FORMAT_TAG);
+    // Media types
+    IF_EQUAL_RETURN(guid, MFMediaType_Audio);
+    IF_EQUAL_RETURN(guid, MFMediaType_Video);
+    IF_EQUAL_RETURN(guid, MFMediaType_Protected);
+    IF_EQUAL_RETURN(guid, MFMediaType_SAMI);
+    IF_EQUAL_RETURN(guid, MFMediaType_Script);
+    IF_EQUAL_RETURN(guid, MFMediaType_Image);
+    IF_EQUAL_RETURN(guid, MFMediaType_HTML);
+    IF_EQUAL_RETURN(guid, MFMediaType_Binary);
+    IF_EQUAL_RETURN(guid, MFMediaType_FileTransfer);
+    IF_EQUAL_RETURN(guid, MFVideoFormat_AI44); //     FCC('AI44')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_ARGB32); //   D3DFMT_A8R8G8B8
+    IF_EQUAL_RETURN(guid, MFVideoFormat_AYUV); //     FCC('AYUV')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_DV25); //     FCC('dv25')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_DV50); //     FCC('dv50')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_DVH1); //     FCC('dvh1')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_DVSD); //     FCC('dvsd')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_DVSL); //     FCC('dvsl')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_H264); //     FCC('H264')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_I420); //     FCC('I420')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_IYUV); //     FCC('IYUV')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_M4S2); //     FCC('M4S2')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MJPG);
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MP43); //     FCC('MP43')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MP4S); //     FCC('MP4S')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MP4V); //     FCC('MP4V')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MPG1); //     FCC('MPG1')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MSS1); //     FCC('MSS1')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_MSS2); //     FCC('MSS2')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_NV11); //     FCC('NV11')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_NV12); //     FCC('NV12')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_P010); //     FCC('P010')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_P016); //     FCC('P016')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_P210); //     FCC('P210')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_P216); //     FCC('P216')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_RGB24); //    D3DFMT_R8G8B8
+    IF_EQUAL_RETURN(guid, MFVideoFormat_RGB32); //    D3DFMT_X8R8G8B8
+    IF_EQUAL_RETURN(guid, MFVideoFormat_RGB555); //   D3DFMT_X1R5G5B5
+    IF_EQUAL_RETURN(guid, MFVideoFormat_RGB565); //   D3DFMT_R5G6B5
+    IF_EQUAL_RETURN(guid, MFVideoFormat_RGB8);
+    IF_EQUAL_RETURN(guid, MFVideoFormat_UYVY); //     FCC('UYVY')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_v210); //     FCC('v210')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_v410); //     FCC('v410')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_WMV1); //     FCC('WMV1')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_WMV2); //     FCC('WMV2')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_WMV3); //     FCC('WMV3')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_WVC1); //     FCC('WVC1')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y210); //     FCC('Y210')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y216); //     FCC('Y216')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y410); //     FCC('Y410')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y416); //     FCC('Y416')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y41P);
+    IF_EQUAL_RETURN(guid, MFVideoFormat_Y41T);
+    IF_EQUAL_RETURN(guid, MFVideoFormat_YUY2); //     FCC('YUY2')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_YV12); //     FCC('YV12')
+    IF_EQUAL_RETURN(guid, MFVideoFormat_YVYU);
+    IF_EQUAL_RETURN(guid, MFAudioFormat_PCM); //              WAVE_FORMAT_PCM
+    IF_EQUAL_RETURN(guid, MFAudioFormat_Float); //            WAVE_FORMAT_IEEE_FLOAT
+    IF_EQUAL_RETURN(guid, MFAudioFormat_DTS); //              WAVE_FORMAT_DTS
+    IF_EQUAL_RETURN(guid, MFAudioFormat_Dolby_AC3_SPDIF); //  WAVE_FORMAT_DOLBY_AC3_SPDIF
+    IF_EQUAL_RETURN(guid, MFAudioFormat_DRM); //              WAVE_FORMAT_DRM
+    IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudioV8); //        WAVE_FORMAT_WMAUDIO2
+    IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudioV9); //        WAVE_FORMAT_WMAUDIO3
+    IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudio_Lossless); // WAVE_FORMAT_WMAUDIO_LOSSLESS
+    IF_EQUAL_RETURN(guid, MFAudioFormat_WMASPDIF); //         WAVE_FORMAT_WMASPDIF
+    IF_EQUAL_RETURN(guid, MFAudioFormat_MSP1); //             WAVE_FORMAT_WMAVOICE9
+    IF_EQUAL_RETURN(guid, MFAudioFormat_MP3); //              WAVE_FORMAT_MPEGLAYER3
+    IF_EQUAL_RETURN(guid, MFAudioFormat_MPEG); //             WAVE_FORMAT_MPEG
+    IF_EQUAL_RETURN(guid, MFAudioFormat_AAC); //              WAVE_FORMAT_MPEG_HEAAC
+    IF_EQUAL_RETURN(guid, MFAudioFormat_ADTS); //             WAVE_FORMAT_MPEG_ADTS_AAC
+    return NULL;
+}
+
+FormatReader::FormatReader(void)
+{
+}
+
+MediaType FormatReader::Read(IMFMediaType *pType)
+{
+    UINT32 count = 0;
+    HRESULT hr = S_OK;
+    MediaType out;
+    hr = pType->LockStore();
+    if (FAILED(hr))
+    {
+        return out;
+    }
+    hr = pType->GetCount(&count);
+    if (FAILED(hr))
+    {
+        return out;
+    }
+    for (UINT32 i = 0; i < count; i++)
+    {
+        hr = LogAttributeValueByIndexNew(pType, i, out);
+        if (FAILED(hr))
+        {
+            break;
+        }
+    }
+    hr = pType->UnlockStore();
+    if (FAILED(hr))
+    {
+        return out;
+    }
+    return out;
+}
+
+FormatReader::~FormatReader(void)
+{
+}
+
+#define CHECK_HR(x) if (FAILED(x)) { goto done; }
+
+ImageGrabber::ImageGrabber(unsigned int deviceID, bool synchronous):
+    m_cRef(1),
+    ig_DeviceID(deviceID),
+    ig_pSource(NULL),
+    ig_pSession(NULL),
+    ig_pTopology(NULL),
+    ig_RIE(true),
+    ig_Close(false),
+    ig_Synchronous(synchronous),
+    ig_hFrameReady(synchronous ? CreateEvent(NULL, FALSE, FALSE, NULL): 0),
+    ig_hFrameGrabbed(synchronous ? CreateEvent(NULL, FALSE, TRUE, NULL): 0),
+    ig_hFinish(CreateEvent(NULL, TRUE, FALSE, NULL))
+{}
+
+ImageGrabber::~ImageGrabber(void)
+{
+    if (ig_pSession)
+    {
+        ig_pSession->Shutdown();
+    }
+
+    CloseHandle(ig_hFinish);
+
+    if (ig_Synchronous)
+    {
+        CloseHandle(ig_hFrameReady);
+        CloseHandle(ig_hFrameGrabbed);
+    }
+
+    SafeRelease(&ig_pSession);
+    SafeRelease(&ig_pTopology);
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+
+    DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Destroing instance of the ImageGrabber class\n", ig_DeviceID);
+}
+
+HRESULT ImageGrabber::initImageGrabber(IMFMediaSource *pSource, GUID VideoFormat)
+{
+    ComPtr<IMFActivate> pSinkActivate = NULL;
+    ComPtr<IMFMediaType> pType = NULL;
+    ComPtr<IMFPresentationDescriptor> pPD = NULL;
+    ComPtr<IMFStreamDescriptor> pSD = NULL;
+    ComPtr<IMFMediaTypeHandler> pHandler = NULL;
+    ComPtr<IMFMediaType> pCurrentType = NULL;
+    HRESULT hr = S_OK;
+    MediaType MT;
+     // Clean up.
+    if (ig_pSession)
+    {
+        ig_pSession->Shutdown();
+    }
+    SafeRelease(&ig_pSession);
+    SafeRelease(&ig_pTopology);
+    ig_pSource = pSource;
+    hr = pSource->CreatePresentationDescriptor(&pPD);
+    if (FAILED(hr))
+    {
+        goto err;
+    }
+    BOOL fSelected;
+    hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, &pSD);
+    if (FAILED(hr)) {
+        goto err;
+    }
+    hr = pSD->GetMediaTypeHandler(&pHandler);
+    if (FAILED(hr)) {
+        goto err;
+    }
+    DWORD cTypes = 0;
+    hr = pHandler->GetMediaTypeCount(&cTypes);
+    if (FAILED(hr)) {
+        goto err;
+    }
+    if(cTypes > 0)
+    {
+        hr = pHandler->GetCurrentMediaType(&pCurrentType);
+        if (FAILED(hr)) {
+            goto err;
+        }
+        MT = FormatReader::Read(pCurrentType.Get());
+    }
+err:
+    unsigned int sizeRawImage = 0;
+    if(VideoFormat == MFVideoFormat_RGB24)
+    {
+        sizeRawImage = MT.MF_MT_FRAME_SIZE * 3;
+    }
+    else if(VideoFormat == MFVideoFormat_RGB32)
+    {
+        sizeRawImage = MT.MF_MT_FRAME_SIZE * 4;
+    }
+    CHECK_HR(hr = RawImage::CreateInstance(&ig_RIFirst, sizeRawImage));
+    CHECK_HR(hr = RawImage::CreateInstance(&ig_RISecond, sizeRawImage));
+    ig_RIOut = ig_RISecond;
+    // Configure the media type that the Sample Grabber will receive.
+    // Setting the major and subtype is usually enough for the topology loader
+    // to resolve the topology.
+    CHECK_HR(hr = MFCreateMediaType(pType.GetAddressOf()));
+    CHECK_HR(hr = pType->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video));
+    CHECK_HR(hr = pType->SetGUID(MF_MT_SUBTYPE, VideoFormat));
+    // Create the sample grabber sink.
+    CHECK_HR(hr = MFCreateSampleGrabberSinkActivate(pType.Get(), this, pSinkActivate.GetAddressOf()));
+    // To run as fast as possible, set this attribute (requires Windows 7):
+    CHECK_HR(hr = pSinkActivate->SetUINT32(MF_SAMPLEGRABBERSINK_IGNORE_CLOCK, TRUE));
+    // Create the Media Session.
+    CHECK_HR(hr = MFCreateMediaSession(NULL, &ig_pSession));
+    // Create the topology.
+    CHECK_HR(hr = CreateTopology(pSource, pSinkActivate.Get(), &ig_pTopology));
+done:
+    // Clean up.
+    if (FAILED(hr))
+    {
+        if (ig_pSession)
+        {
+            ig_pSession->Shutdown();
+        }
+        SafeRelease(&ig_pSession);
+        SafeRelease(&ig_pTopology);
+    }
+
+    return hr;
+}
+
+void ImageGrabber::stopGrabbing()
+{
+    if(ig_pSession)
+        ig_pSession->Stop();
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Stopping of of grabbing of images\n", ig_DeviceID);
+}
+
+HRESULT ImageGrabber::startGrabbing(void)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    ComPtr<IMFMediaEvent> pEvent = NULL;
+    PROPVARIANT var;
+    PropVariantInit(&var);
+    HRESULT hr = S_OK;
+    hr = ig_pSession->SetTopology(0, ig_pTopology);
+    DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Start Grabbing of the images\n", ig_DeviceID);
+    hr = ig_pSession->Start(&GUID_NULL, &var);
+    for(;;)
+    {
+        HRESULT hrStatus = S_OK;
+        MediaEventType met;
+        if(!ig_pSession) break;
+        hr = ig_pSession->GetEvent(0, &pEvent);
+        if(!SUCCEEDED(hr))
+        {
+            hr = S_OK;
+            goto done;
+        }
+        hr = pEvent->GetStatus(&hrStatus);
+        if(!SUCCEEDED(hr))
+        {
+            hr = S_OK;
+            goto done;
+        }
+        hr = pEvent->GetType(&met);
+        if(!SUCCEEDED(hr))
+        {
+            hr = S_OK;
+            goto done;
+        }
+        if (met == MESessionEnded)
+        {
+            DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MESessionEnded \n", ig_DeviceID);
+            ig_pSession->Stop();
+            break;
+        }
+        if (met == MESessionStopped)
+        {
+            DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MESessionStopped \n", ig_DeviceID);
+            break;
+        }
+        if (met == MEVideoCaptureDeviceRemoved)
+        {
+            DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MEVideoCaptureDeviceRemoved \n", ig_DeviceID);
+            break;
+        }
+        if ((met == MEError) || (met == MENonFatalError))
+        {
+            pEvent->GetStatus(&hrStatus);
+            DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MEError | MENonFatalError: %u\n", ig_DeviceID, hrStatus);
+            break;
+        }
+    }
+    DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Finish startGrabbing \n", ig_DeviceID);
+
+done:
+    SetEvent(ig_hFinish);
+
+    return hr;
+}
+
+void ImageGrabber::pauseGrabbing()
+{
+}
+
+void ImageGrabber::resumeGrabbing()
+{
+}
+
+HRESULT ImageGrabber::CreateTopology(IMFMediaSource *pSource, IMFActivate *pSinkActivate, IMFTopology **ppTopo)
+{
+    IMFTopology* pTopology = NULL;
+    ComPtr<IMFPresentationDescriptor> pPD = NULL;
+    ComPtr<IMFStreamDescriptor> pSD = NULL;
+    ComPtr<IMFMediaTypeHandler> pHandler = NULL;
+    ComPtr<IMFTopologyNode> pNode1 = NULL;
+    ComPtr<IMFTopologyNode> pNode2 = NULL;
+    HRESULT hr = S_OK;
+    DWORD cStreams = 0;
+    CHECK_HR(hr = MFCreateTopology(&pTopology));
+    CHECK_HR(hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf()));
+    CHECK_HR(hr = pPD->GetStreamDescriptorCount(&cStreams));
+    for (DWORD i = 0; i < cStreams; i++)
+    {
+        // In this example, we look for audio streams and connect them to the sink.
+        BOOL fSelected = FALSE;
+        GUID majorType;
+        CHECK_HR(hr = pPD->GetStreamDescriptorByIndex(i, &fSelected, &pSD));
+        CHECK_HR(hr = pSD->GetMediaTypeHandler(&pHandler));
+        CHECK_HR(hr = pHandler->GetMajorType(&majorType));
+        if (majorType == MFMediaType_Video && fSelected)
+        {
+            CHECK_HR(hr = AddSourceNode(pTopology, pSource, pPD.Get(), pSD.Get(), pNode1.GetAddressOf()));
+            CHECK_HR(hr = AddOutputNode(pTopology, pSinkActivate, 0, pNode2.GetAddressOf()));
+            CHECK_HR(hr = pNode1->ConnectOutput(0, pNode2.Get(), 0));
+            break;
+        }
+        else
+        {
+            CHECK_HR(hr = pPD->DeselectStream(i));
+        }
+    }
+    *ppTopo = pTopology;
+    (*ppTopo)->AddRef();
+
+done:
+    return hr;
+}
+
+HRESULT ImageGrabber::AddSourceNode(
+    IMFTopology *pTopology,           // Topology.
+    IMFMediaSource *pSource,          // Media source.
+    IMFPresentationDescriptor *pPD,   // Presentation descriptor.
+    IMFStreamDescriptor *pSD,         // Stream descriptor.
+    IMFTopologyNode **ppNode)         // Receives the node pointer.
+{
+    ComPtr<IMFTopologyNode> pNode = NULL;
+    HRESULT hr = S_OK;
+    CHECK_HR(hr = MFCreateTopologyNode(MF_TOPOLOGY_SOURCESTREAM_NODE, pNode.GetAddressOf()));
+    CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_SOURCE, pSource));
+    CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_PRESENTATION_DESCRIPTOR, pPD));
+    CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_STREAM_DESCRIPTOR, pSD));
+    CHECK_HR(hr = pTopology->AddNode(pNode.Get()));
+    // Return the pointer to the caller.
+    *ppNode = pNode.Get();
+    (*ppNode)->AddRef();
+
+done:
+    return hr;
+}
+
+HRESULT ImageGrabber::AddOutputNode(
+    IMFTopology *pTopology,     // Topology.
+    IMFActivate *pActivate,     // Media sink activation object.
+    DWORD dwId,                 // Identifier of the stream sink.
+    IMFTopologyNode **ppNode)   // Receives the node pointer.
+{
+    ComPtr<IMFTopologyNode> pNode = NULL;
+    HRESULT hr = S_OK;
+    CHECK_HR(hr = MFCreateTopologyNode(MF_TOPOLOGY_OUTPUT_NODE, pNode.GetAddressOf()));
+    CHECK_HR(hr = pNode->SetObject(pActivate));
+    CHECK_HR(hr = pNode->SetUINT32(MF_TOPONODE_STREAMID, dwId));
+    CHECK_HR(hr = pNode->SetUINT32(MF_TOPONODE_NOSHUTDOWN_ON_REMOVE, FALSE));
+    CHECK_HR(hr = pTopology->AddNode(pNode.Get()));
+    // Return the pointer to the caller.
+    *ppNode = pNode.Get();
+    (*ppNode)->AddRef();
+
+done:
+    return hr;
+}
+
+HRESULT ImageGrabber::CreateInstance(ImageGrabber **ppIG, unsigned int deviceID, bool synchronious)
+{
+    *ppIG = new (std::nothrow) ImageGrabber(deviceID, synchronious);
+    if (ppIG == NULL)
+    {
+        return E_OUTOFMEMORY;
+    }
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Creating instance of ImageGrabber\n", deviceID);
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::QueryInterface(REFIID riid, void** ppv)
+{
+    HRESULT hr = E_NOINTERFACE;
+    *ppv = NULL;
+    if(riid == IID_IUnknown || riid == IID_IMFSampleGrabberSinkCallback)
+    {
+        *ppv = static_cast<IMFSampleGrabberSinkCallback *>(this);
+        hr = S_OK;
+    }
+    if(riid == IID_IMFClockStateSink)
+    {
+        *ppv = static_cast<IMFClockStateSink *>(this);
+        hr = S_OK;
+    }
+    if(SUCCEEDED(hr))
+    {
+        reinterpret_cast<IUnknown *>(*ppv)->AddRef();
+    }
+    return hr;
+}
+
+STDMETHODIMP_(ULONG) ImageGrabber::AddRef()
+{
+    return InterlockedIncrement(&m_cRef);
+}
+
+STDMETHODIMP_(ULONG) ImageGrabber::Release()
+{
+    ULONG cRef = InterlockedDecrement(&m_cRef);
+    if (cRef == 0)
+    {
+        delete this;
+    }
+    return cRef;
+}
+
+STDMETHODIMP ImageGrabber::OnClockStart(MFTIME hnsSystemTime, LONGLONG llClockStartOffset)
+{
+    (void)hnsSystemTime;
+    (void)llClockStartOffset;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnClockStop(MFTIME hnsSystemTime)
+{
+    (void)hnsSystemTime;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnClockPause(MFTIME hnsSystemTime)
+{
+    (void)hnsSystemTime;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnClockRestart(MFTIME hnsSystemTime)
+{
+    (void)hnsSystemTime;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnClockSetRate(MFTIME hnsSystemTime, float flRate)
+{
+    (void)flRate;
+    (void)hnsSystemTime;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnSetPresentationClock(IMFPresentationClock* pClock)
+{
+    (void)pClock;
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnProcessSample(REFGUID guidMajorMediaType, DWORD dwSampleFlags,
+    LONGLONG llSampleTime, LONGLONG llSampleDuration, const BYTE * pSampleBuffer,
+    DWORD dwSampleSize)
+{
+    (void)guidMajorMediaType;
+    (void)llSampleTime;
+    (void)dwSampleFlags;
+    (void)llSampleDuration;
+    (void)dwSampleSize;
+
+    HANDLE tmp[] = {ig_hFinish, ig_hFrameGrabbed, NULL};
+
+    DWORD status = WaitForMultipleObjects(2, tmp, FALSE, INFINITE);
+    if (status == WAIT_OBJECT_0)
+    {
+        printf("OnProcessFrame called after ig_hFinish event\n");
+        return S_OK;
+    }
+
+    if(ig_RIE)
+    {
+        ig_RIFirst->fastCopy(pSampleBuffer);
+        ig_RIOut = ig_RIFirst;
+    }
+    else
+    {
+        ig_RISecond->fastCopy(pSampleBuffer);
+        ig_RIOut = ig_RISecond;
+    }
+
+    if (ig_Synchronous)
+    {
+        SetEvent(ig_hFrameReady);
+    }
+    else
+    {
+        ig_RIE = !ig_RIE;
+    }
+
+    return S_OK;
+}
+
+STDMETHODIMP ImageGrabber::OnShutdown()
+{
+    SetEvent(ig_hFinish);
+    return S_OK;
+}
+
+RawImage *ImageGrabber::getRawImage()
+{
+    return ig_RIOut;
+}
+
+DWORD WINAPI MainThreadFunction( LPVOID lpParam )
+{
+    ImageGrabberThread *pIGT = (ImageGrabberThread *)lpParam;
+    pIGT->run();
+    return 0;
+}
+
+HRESULT ImageGrabberThread::CreateInstance(ImageGrabberThread **ppIGT, IMFMediaSource *pSource, unsigned int deviceID, bool synchronious)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    *ppIGT = new (std::nothrow) ImageGrabberThread(pSource, deviceID, synchronious);
+    if (ppIGT == NULL)
+    {
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Memory cannot be allocated\n", deviceID);
+        return E_OUTOFMEMORY;
+    }
+    else
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Creating of the instance of ImageGrabberThread\n", deviceID);
+    return S_OK;
+}
+
+ImageGrabberThread::ImageGrabberThread(IMFMediaSource *pSource, unsigned int deviceID, bool synchronious):
+    igt_func(NULL),
+    igt_Handle(NULL),
+    igt_stop(false)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    HRESULT hr = ImageGrabber::CreateInstance(&igt_pImageGrabber, deviceID, synchronious);
+    igt_DeviceID = deviceID;
+    if(SUCCEEDED(hr))
+    {
+        hr = igt_pImageGrabber->initImageGrabber(pSource, MFVideoFormat_RGB24);
+        if(!SUCCEEDED(hr))
+        {
+            DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: There is a problem with initialization of the instance of the ImageGrabber class\n", deviceID);
+        }
+        else
+        {
+            DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Initialization of instance of the ImageGrabber class\n", deviceID);
+        }
+    }
+    else
+    {
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i There is a problem with creation of the instance of the ImageGrabber class\n", deviceID);
+    }
+}
+
+void ImageGrabberThread::setEmergencyStopEvent(void *userData, void(*func)(int, void *))
+{
+    if(func)
+    {
+        igt_func = func;
+        igt_userData = userData;
+    }
+}
+
+ImageGrabberThread::~ImageGrabberThread(void)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Destroing ImageGrabberThread\n", igt_DeviceID);
+    if (igt_Handle)
+        WaitForSingleObject(igt_Handle, INFINITE);
+    delete igt_pImageGrabber;
+}
+
+void ImageGrabberThread::stop()
+{
+    igt_stop = true;
+    if(igt_pImageGrabber)
+    {
+        igt_pImageGrabber->stopGrabbing();
+    }
+}
+
+void ImageGrabberThread::start()
+{
+    igt_Handle = CreateThread(
+            NULL,                  // default security attributes
+            0,                     // use default stack size
+            MainThreadFunction,    // thread function name
+            this,                  // argument to thread function
+            0,                     // use default creation flags
+            &igt_ThreadIdArray);   // returns the thread identifier
+}
+
+void ImageGrabberThread::run()
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if(igt_pImageGrabber)
+    {
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Thread for grabbing images is started\n", igt_DeviceID);
+        HRESULT hr = igt_pImageGrabber->startGrabbing();
+        if(!SUCCEEDED(hr))
+        {
+            DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: There is a problem with starting the process of grabbing\n", igt_DeviceID);
+        }
+    }
+    else
+    {
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i The thread is finished without execution of grabbing\n", igt_DeviceID);
+    }
+    if(!igt_stop)
+    {
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Emergency Stop thread\n", igt_DeviceID);
+        if(igt_func)
+        {
+            igt_func(igt_DeviceID, igt_userData);
+        }
+    }
+    else
+        DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Finish thread\n", igt_DeviceID);
+}
+
+ImageGrabber *ImageGrabberThread::getImageGrabber()
+{
+    return igt_pImageGrabber;
+}
+
+Media_Foundation::Media_Foundation(void)
+{
+    HRESULT hr = MFStartup(MF_VERSION);
+    if(!SUCCEEDED(hr))
+    {
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        DPO->printOut(L"MEDIA FOUNDATION: It cannot be created!!!\n");
+    }
+}
+
+Media_Foundation::~Media_Foundation(void)
+{
+    HRESULT hr = MFShutdown();
+    if(!SUCCEEDED(hr))
+    {
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        DPO->printOut(L"MEDIA FOUNDATION: Resources cannot be released\n");
+    }
+}
+
+bool Media_Foundation::buildListOfDevices()
+{
+    HRESULT hr = S_OK;
+    ComPtr<IMFAttributes> pAttributes = NULL;
+    CoInitialize(NULL);
+    hr = MFCreateAttributes(pAttributes.GetAddressOf(), 1);
+    if (SUCCEEDED(hr))
+    {
+        hr = pAttributes->SetGUID(
+            MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE,
+            MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE_VIDCAP_GUID
+            );
+    }
+    if (SUCCEEDED(hr))
+    {
+        videoDevices *vDs = &videoDevices::getInstance();
+        hr = vDs->initDevices(pAttributes.Get());
+    }
+    else
+    {
+       DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+       DPO->printOut(L"MEDIA FOUNDATION: The access to the video cameras denied\n");
+    }
+
+    return (SUCCEEDED(hr));
+}
+
+Media_Foundation& Media_Foundation::getInstance()
+{
+    static Media_Foundation instance;
+    return instance;
+}
+
+RawImage::RawImage(unsigned int size): ri_new(false), ri_pixels(NULL)
+{
+    ri_size = size;
+    ri_pixels = new unsigned char[size];
+    memset((void *)ri_pixels,0,ri_size);
+}
+
+bool RawImage::isNew()
+{
+    return ri_new;
+}
+
+unsigned int RawImage::getSize()
+{
+    return ri_size;
+}
+
+RawImage::~RawImage(void)
+{
+    delete []ri_pixels;
+    ri_pixels = NULL;
+}
+
+long RawImage::CreateInstance(RawImage **ppRImage,unsigned int size)
+{
+    *ppRImage = new (std::nothrow) RawImage(size);
+    if (ppRImage == NULL)
+    {
+        return E_OUTOFMEMORY;
+    }
+    return S_OK;
+}
+
+void RawImage::setCopy(const BYTE * pSampleBuffer)
+{
+    memcpy(ri_pixels, pSampleBuffer, ri_size);
+    ri_new = true;
+}
+
+void RawImage::fastCopy(const BYTE * pSampleBuffer)
+{
+    memcpy(ri_pixels, pSampleBuffer, ri_size);
+    ri_new = true;
+}
+
+unsigned char * RawImage::getpPixels()
+{
+    ri_new = false;
+    return ri_pixels;
+}
+
+videoDevice::videoDevice(void): vd_IsSetuped(false), vd_LockOut(OpenLock), vd_pFriendlyName(NULL),
+    vd_Width(0), vd_Height(0), vd_pSource(NULL), vd_func(NULL), vd_userData(NULL)
+{
+}
+
+void videoDevice::setParametrs(CamParametrs parametrs)
+{
+    if(vd_IsSetuped)
+    {
+        if(vd_pSource)
+        {
+            Parametr *pParametr = (Parametr *)(&parametrs);
+            Parametr *pPrevParametr = (Parametr *)(&vd_PrevParametrs);
+            IAMVideoProcAmp *pProcAmp = NULL;
+            HRESULT hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcAmp));
+            if (SUCCEEDED(hr))
+            {
+                for(unsigned int i = 0; i < 10; i++)
+                {
+                    if(pPrevParametr[i].CurrentValue != pParametr[i].CurrentValue || pPrevParametr[i].Flag != pParametr[i].Flag)
+                        hr = pProcAmp->Set(VideoProcAmp_Brightness + i, pParametr[i].CurrentValue, pParametr[i].Flag);
+                }
+                pProcAmp->Release();
+            }
+            IAMCameraControl *pProcControl = NULL;
+            hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcControl));
+            if (SUCCEEDED(hr))
+            {
+                for(unsigned int i = 0; i < 7; i++)
+                {
+                    if(pPrevParametr[10 + i].CurrentValue != pParametr[10 + i].CurrentValue || pPrevParametr[10 + i].Flag != pParametr[10 + i].Flag)
+                    hr = pProcControl->Set(CameraControl_Pan+i, pParametr[10 + i].CurrentValue, pParametr[10 + i].Flag);
+                }
+                pProcControl->Release();
+            }
+            vd_PrevParametrs = parametrs;
+        }
+    }
+}
+
+CamParametrs videoDevice::getParametrs()
+{
+    CamParametrs out;
+    if(vd_IsSetuped)
+    {
+        if(vd_pSource)
+        {
+            Parametr *pParametr = (Parametr *)(&out);
+            IAMVideoProcAmp *pProcAmp = NULL;
+            HRESULT hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcAmp));
+            if (SUCCEEDED(hr))
+            {
+                for(unsigned int i = 0; i < 10; i++)
+                {
+                    Parametr temp;
+                    hr = pProcAmp->GetRange(VideoProcAmp_Brightness+i, &temp.Min, &temp.Max, &temp.Step, &temp.Default, &temp.Flag);
+                    if (SUCCEEDED(hr))
+                    {
+                        temp.CurrentValue = temp.Default;
+                        pParametr[i] = temp;
+                    }
+                }
+                pProcAmp->Release();
+            }
+            IAMCameraControl *pProcControl = NULL;
+            hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcControl));
+            if (SUCCEEDED(hr))
+            {
+                for(unsigned int i = 0; i < 7; i++)
+                {
+                    Parametr temp;
+                    hr = pProcControl->GetRange(CameraControl_Pan+i, &temp.Min, &temp.Max, &temp.Step, &temp.Default, &temp.Flag);
+                    if (SUCCEEDED(hr))
+                    {
+                        temp.CurrentValue = temp.Default;
+                        pParametr[10 + i] = temp;
+                    }
+                }
+                pProcControl->Release();
+            }
+        }
+    }
+    return out;
+}
+
+long videoDevice::resetDevice(IMFActivate *pActivate)
+{
+    HRESULT hr = -1;
+    vd_CurrentFormats.clear();
+    if(vd_pFriendlyName)
+        CoTaskMemFree(vd_pFriendlyName);
+    vd_pFriendlyName = NULL;
+    if(pActivate)
+    {
+        IMFMediaSource *pSource = NULL;
+        hr = pActivate->GetAllocatedString(
+                MF_DEVSOURCE_ATTRIBUTE_FRIENDLY_NAME,
+                &vd_pFriendlyName,
+                NULL
+                );
+        hr = pActivate->ActivateObject(
+            __uuidof(IMFMediaSource),
+            (void**)&pSource
+            );
+        enumerateCaptureFormats(pSource);
+        buildLibraryofTypes();
+        SafeRelease(&pSource);
+        if(FAILED(hr))
+        {
+            vd_pFriendlyName = NULL;
+            DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+            DPO->printOut(L"VIDEODEVICE %i: IMFMediaSource interface cannot be created \n", vd_CurrentNumber);
+        }
+    }
+    return hr;
+}
+
+long videoDevice::readInfoOfDevice(IMFActivate *pActivate, unsigned int Num)
+{
+    HRESULT hr = -1;
+    vd_CurrentNumber = Num;
+    hr = resetDevice(pActivate);
+    return hr;
+}
+
+long videoDevice::checkDevice(IMFAttributes *pAttributes, IMFActivate **pDevice)
+{
+    HRESULT hr = S_OK;
+    IMFActivate **ppDevices = NULL;
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    UINT32 count;
+    wchar_t *newFriendlyName = NULL;
+    hr = MFEnumDeviceSources(pAttributes, &ppDevices, &count);
+    if (SUCCEEDED(hr))
+    {
+        if(count > 0)
+        {
+            if(count > vd_CurrentNumber)
+            {
+                hr = ppDevices[vd_CurrentNumber]->GetAllocatedString(
+                MF_DEVSOURCE_ATTRIBUTE_FRIENDLY_NAME,
+                &newFriendlyName,
+                NULL
+                );
+                if (SUCCEEDED(hr))
+                {
+                    if(wcscmp(newFriendlyName, vd_pFriendlyName) != 0)
+                    {
+                        DPO->printOut(L"VIDEODEVICE %i: Chosen device cannot be found \n", vd_CurrentNumber);
+                        hr = -1;
+                        pDevice = NULL;
+                    }
+                    else
+                    {
+                        *pDevice = ppDevices[vd_CurrentNumber];
+                        (*pDevice)->AddRef();
+                    }
+                }
+                else
+                {
+                    DPO->printOut(L"VIDEODEVICE %i: Name of device cannot be gotten \n", vd_CurrentNumber);
+                }
+            }
+            else
+            {
+                DPO->printOut(L"VIDEODEVICE %i: Number of devices more than corrent number of the device \n", vd_CurrentNumber);
+                hr = -1;
+            }
+            for(UINT32 i = 0; i < count; i++)
+            {
+                SafeRelease(&ppDevices[i]);
+            }
+            SafeRelease(ppDevices);
+        }
+        else
+            hr = -1;
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE %i: List of DeviceSources cannot be enumerated \n", vd_CurrentNumber);
+    }
+    return hr;
+}
+
+long videoDevice::initDevice()
+{
+    HRESULT hr = -1;
+    ComPtr<IMFAttributes> pAttributes = NULL;
+    IMFActivate *vd_pActivate = NULL;
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    CoInitialize(NULL);
+    hr = MFCreateAttributes(pAttributes.GetAddressOf(), 1);
+    if (SUCCEEDED(hr))
+    {
+        hr = pAttributes->SetGUID(
+            MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE,
+            MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE_VIDCAP_GUID
+            );
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = checkDevice(pAttributes.Get(), &vd_pActivate);
+        if (SUCCEEDED(hr) && vd_pActivate)
+        {
+            SafeRelease(&vd_pSource);
+            hr = vd_pActivate->ActivateObject(
+                __uuidof(IMFMediaSource),
+                (void**)&vd_pSource
+                );
+            if (SUCCEEDED(hr))
+            {
+            }
+            SafeRelease(&vd_pActivate);
+        }
+        else
+        {
+            DPO->printOut(L"VIDEODEVICE %i: Device there is not \n", vd_CurrentNumber);
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE %i: The attribute of video cameras cannot be getting \n", vd_CurrentNumber);
+    }
+
+    return hr;
+}
+
+MediaType videoDevice::getFormat(unsigned int id)
+{
+    if(id < vd_CurrentFormats.size())
+    {
+        return vd_CurrentFormats[id];
+    }
+    else return MediaType();
+}
+int videoDevice::getCountFormats()
+{
+    return vd_CurrentFormats.size();
+}
+void videoDevice::setEmergencyStopEvent(void *userData, void(*func)(int, void *))
+{
+    vd_func = func;
+    vd_userData = userData;
+}
+void videoDevice::closeDevice()
+{
+    if(vd_IsSetuped)
+    {
+        vd_IsSetuped = false;
+        vd_pSource->Stop();
+        SafeRelease(&vd_pSource);
+        if(vd_LockOut == RawDataLock)
+        {
+            vd_pImGrTh->stop();
+            Sleep(500);
+            delete vd_pImGrTh;
+        }
+        vd_pImGrTh = NULL;
+        vd_LockOut = OpenLock;
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        DPO->printOut(L"VIDEODEVICE %i: Device is stopped \n", vd_CurrentNumber);
+    }
+}
+unsigned int videoDevice::getWidth()
+{
+    if(vd_IsSetuped)
+        return vd_Width;
+    else
+        return 0;
+}
+unsigned int videoDevice::getHeight()
+{
+    if(vd_IsSetuped)
+        return vd_Height;
+    else
+        return 0;
+}
+IMFMediaSource *videoDevice::getMediaSource()
+{
+    IMFMediaSource *out = NULL;
+    if(vd_LockOut == OpenLock)
+    {
+        vd_LockOut = MediaSourceLock;
+        out = vd_pSource;
+    }
+    return out;
+}
+int videoDevice::findType(unsigned int size, unsigned int frameRate)
+{
+    if(vd_CaptureFormats.size() == 0)
+        return 0;
+    FrameRateMap FRM = vd_CaptureFormats[size];
+    if(FRM.size() == 0)
+        return 0;
+    UINT64 frameRateMax = 0;  SUBTYPEMap STMMax;
+    if(frameRate == 0)
+    {
+        std::map<UINT64, SUBTYPEMap>::iterator f = FRM.begin();
+        for(; f != FRM.end(); f++)
+        {
+             if((*f).first >= frameRateMax)
+             {
+                 frameRateMax = (*f).first;
+                 STMMax = (*f).second;
+             }
+        }
+    }
+    else
+    {
+        std::map<UINT64, SUBTYPEMap>::iterator f = FRM.begin();
+        for(; f != FRM.end(); f++)
+        {
+             if((*f).first >= frameRateMax)
+             {
+                 if(frameRate > (*f).first)
+                 {
+                     frameRateMax = (*f).first;
+                     STMMax = (*f).second;
+                 }
+             }
+        }
+    }
+    if(STMMax.size() == 0)
+        return 0;
+    std::map<String, vectorNum>::iterator S = STMMax.begin();
+    vectorNum VN = (*S).second;
+    if(VN.size() == 0)
+        return 0;
+    return VN[0];
+}
+
+void videoDevice::buildLibraryofTypes()
+{
+    unsigned int size;
+    unsigned int framerate;
+    std::vector<MediaType>::iterator i = vd_CurrentFormats.begin();
+    int count = 0;
+    for(; i != vd_CurrentFormats.end(); i++)
+    {
+        size = (*i).MF_MT_FRAME_SIZE;
+        framerate = (*i).MF_MT_FRAME_RATE_NUMERATOR;
+        FrameRateMap FRM = vd_CaptureFormats[size];
+        SUBTYPEMap STM = FRM[framerate];
+        String subType((*i).pMF_MT_SUBTYPEName);
+        vectorNum VN = STM[subType];
+        VN.push_back(count);
+        STM[subType] = VN;
+        FRM[framerate] = STM;
+        vd_CaptureFormats[size] = FRM;
+        count++;
+    }
+}
+
+long videoDevice::setDeviceFormat(IMFMediaSource *pSource, unsigned long  dwFormatIndex)
+{
+    ComPtr<IMFPresentationDescriptor> pPD = NULL;
+    ComPtr<IMFStreamDescriptor> pSD = NULL;
+    ComPtr<IMFMediaTypeHandler> pHandler = NULL;
+    ComPtr<IMFMediaType> pType = NULL;
+    HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    BOOL fSelected;
+    hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = pHandler->GetMediaTypeByIndex((DWORD)dwFormatIndex, pType.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = pHandler->SetCurrentMediaType(pType.Get());
+
+done:
+    return hr;
+}
+
+bool videoDevice::isDeviceSetup()
+{
+    return vd_IsSetuped;
+}
+
+RawImage * videoDevice::getRawImageOut()
+{
+    if(!vd_IsSetuped) return NULL;
+    if(vd_pImGrTh)
+            return vd_pImGrTh->getImageGrabber()->getRawImage();
+    else
+    {
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        DPO->printOut(L"VIDEODEVICE %i: The instance of ImageGrabberThread class does not exist  \n", vd_CurrentNumber);
+    }
+    return NULL;
+}
+
+bool videoDevice::isFrameNew()
+{
+    if(!vd_IsSetuped) return false;
+    if(vd_LockOut == RawDataLock || vd_LockOut == OpenLock)
+    {
+        if(vd_LockOut == OpenLock)
+        {
+            vd_LockOut = RawDataLock;
+            HRESULT hr = ImageGrabberThread::CreateInstance(&vd_pImGrTh, vd_pSource, vd_CurrentNumber);
+            if(FAILED(hr))
+            {
+                DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+                DPO->printOut(L"VIDEODEVICE %i: The instance of ImageGrabberThread class cannot be created.\n", vd_CurrentNumber);
+                return false;
+            }
+            vd_pImGrTh->setEmergencyStopEvent(vd_userData, vd_func);
+            vd_pImGrTh->start();
+            return true;
+        }
+        if(vd_pImGrTh)
+            return vd_pImGrTh->getImageGrabber()->getRawImage()->isNew();
+    }
+    return false;
+}
+
+bool videoDevice::isDeviceMediaSource()
+{
+    if(vd_LockOut == MediaSourceLock) return true;
+    return false;
+}
+
+bool videoDevice::isDeviceRawDataSource()
+{
+    if(vd_LockOut == RawDataLock) return true;
+    return false;
+}
+
+bool videoDevice::setupDevice(unsigned int id)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if(!vd_IsSetuped)
+    {
+        HRESULT hr = -1;
+        hr = initDevice();
+        if(SUCCEEDED(hr))
+        {
+            vd_Width = vd_CurrentFormats[id].width;
+            vd_Height = vd_CurrentFormats[id].height;
+            hr = setDeviceFormat(vd_pSource, (DWORD) id);
+            vd_IsSetuped = (SUCCEEDED(hr));
+            if(vd_IsSetuped)
+                DPO->printOut(L"\n\nVIDEODEVICE %i: Device is setuped \n", vd_CurrentNumber);
+            vd_PrevParametrs = getParametrs();
+            return vd_IsSetuped;
+        }
+        else
+        {
+            DPO->printOut(L"VIDEODEVICE %i: Interface IMFMediaSource cannot be got \n", vd_CurrentNumber);
+            return false;
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Device is setuped already \n", vd_CurrentNumber);
+        return false;
+    }
+}
+
+bool videoDevice::setupDevice(unsigned int w, unsigned int h, unsigned int idealFramerate)
+{
+    unsigned int id = findType(w * h, idealFramerate);
+    return setupDevice(id);
+}
+
+wchar_t *videoDevice::getName()
+{
+    return vd_pFriendlyName;
+}
+
+videoDevice::~videoDevice(void)
+{
+    closeDevice();
+    SafeRelease(&vd_pSource);
+    if(vd_pFriendlyName)
+        CoTaskMemFree(vd_pFriendlyName);
+}
+
+HRESULT videoDevice::enumerateCaptureFormats(IMFMediaSource *pSource)
+{
+    ComPtr<IMFPresentationDescriptor> pPD = NULL;
+    ComPtr<IMFStreamDescriptor> pSD = NULL;
+    ComPtr<IMFMediaTypeHandler> pHandler = NULL;
+    ComPtr<IMFMediaType> pType = NULL;
+    HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    BOOL fSelected;
+    hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    DWORD cTypes = 0;
+    hr = pHandler->GetMediaTypeCount(&cTypes);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    for (DWORD i = 0; i < cTypes; i++)
+    {
+        hr = pHandler->GetMediaTypeByIndex(i, pType.GetAddressOf());
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+        MediaType MT = FormatReader::Read(pType.Get());
+        vd_CurrentFormats.push_back(MT);
+    }
+
+done:
+    return hr;
+}
+
+videoDevices::videoDevices(void): count(0)
+{}
+
+void videoDevices::clearDevices()
+{
+    std::vector<videoDevice *>::iterator i = vds_Devices.begin();
+    for(; i != vds_Devices.end(); ++i)
+        delete (*i);
+    vds_Devices.clear();
+}
+
+videoDevices::~videoDevices(void)
+{
+    clearDevices();
+}
+
+videoDevice * videoDevices::getDevice(unsigned int i)
+{
+    if(i >= vds_Devices.size())
+    {
+        return NULL;
+    }
+    if(i < 0)
+    {
+        return NULL;
+    }
+    return vds_Devices[i];
+}
+
+long videoDevices::initDevices(IMFAttributes *pAttributes)
+{
+    HRESULT hr = S_OK;
+    IMFActivate **ppDevices = NULL;
+    clearDevices();
+    hr = MFEnumDeviceSources(pAttributes, &ppDevices, &count);
+    if (SUCCEEDED(hr))
+    {
+        if(count > 0)
+        {
+            for(UINT32 i = 0; i < count; i++)
+            {
+                videoDevice *vd = new videoDevice;
+                vd->readInfoOfDevice(ppDevices[i], i);
+                vds_Devices.push_back(vd);
+                SafeRelease(&ppDevices[i]);
+            }
+            SafeRelease(ppDevices);
+        }
+        else
+            hr = -1;
+    }
+    else
+    {
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        DPO->printOut(L"VIDEODEVICES: The instances of the videoDevice class cannot be created\n");
+    }
+    return hr;
+}
+
+unsigned int videoDevices::getCount()
+{
+    return vds_Devices.size();
+}
+
+videoDevices& videoDevices::getInstance()
+{
+    static videoDevices instance;
+    return instance;
+}
+
+Parametr::Parametr()
+{
+    CurrentValue = 0;
+    Min = 0;
+    Max = 0;
+    Step = 0;
+    Default = 0;
+    Flag = 0;
+}
+
+MediaType::MediaType()
+{
+    pMF_MT_AM_FORMAT_TYPEName = NULL;
+    pMF_MT_MAJOR_TYPEName = NULL;
+    pMF_MT_SUBTYPEName = NULL;
+    Clear();
+}
+
+MediaType::~MediaType()
+{
+    Clear();
+}
+
+void MediaType::Clear()
+{
+    MF_MT_FRAME_SIZE = 0;
+    height = 0;
+    width = 0;
+    MF_MT_YUV_MATRIX = 0;
+    MF_MT_VIDEO_LIGHTING = 0;
+    MF_MT_DEFAULT_STRIDE = 0;
+    MF_MT_VIDEO_CHROMA_SITING = 0;
+    MF_MT_FIXED_SIZE_SAMPLES = 0;
+    MF_MT_VIDEO_NOMINAL_RANGE = 0;
+    MF_MT_FRAME_RATE_NUMERATOR = 0;
+    MF_MT_FRAME_RATE_DENOMINATOR = 0;
+    MF_MT_PIXEL_ASPECT_RATIO = 0;
+    MF_MT_PIXEL_ASPECT_RATIO_low = 0;
+    MF_MT_ALL_SAMPLES_INDEPENDENT = 0;
+    MF_MT_FRAME_RATE_RANGE_MIN = 0;
+    MF_MT_FRAME_RATE_RANGE_MIN_low = 0;
+    MF_MT_SAMPLE_SIZE = 0;
+    MF_MT_VIDEO_PRIMARIES = 0;
+    MF_MT_INTERLACE_MODE = 0;
+    MF_MT_FRAME_RATE_RANGE_MAX = 0;
+    MF_MT_FRAME_RATE_RANGE_MAX_low = 0;
+    memset(&MF_MT_MAJOR_TYPE, 0, sizeof(GUID));
+    memset(&MF_MT_AM_FORMAT_TYPE, 0, sizeof(GUID));
+    memset(&MF_MT_SUBTYPE, 0, sizeof(GUID));
+}
+
+videoInput::videoInput(void): accessToDevices(false)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->printOut(L"\n***** VIDEOINPUT LIBRARY - 2013 (Author: Evgeny Pereguda) *****\n\n");
+    updateListOfDevices();
+    if(!accessToDevices)
+        DPO->printOut(L"INITIALIZATION: Ther is not any suitable video device\n");
+}
+
+void videoInput::updateListOfDevices()
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    Media_Foundation *MF = &Media_Foundation::getInstance();
+    accessToDevices = MF->buildListOfDevices();
+    if(!accessToDevices)
+        DPO->printOut(L"UPDATING: Ther is not any suitable video device\n");
+}
+
+videoInput::~videoInput(void)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->printOut(L"\n***** CLOSE VIDEOINPUT LIBRARY - 2013 *****\n\n");
+}
+
+IMFMediaSource *videoInput::getMediaSource(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+        {
+            IMFMediaSource *out = VD->getMediaSource();
+            if(!out)
+                DPO->printOut(L"VideoDevice %i: There is not any suitable IMFMediaSource interface\n", deviceID);
+            return out;
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return NULL;
+}
+
+bool videoInput::setupDevice(int deviceID, unsigned int id)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0 )
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+        {
+            bool out = VD->setupDevice(id);
+            if(!out)
+                DPO->printOut(L"VIDEODEVICE %i: This device cannot be started\n", deviceID);
+            return out;
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return false;
+}
+
+bool videoInput::setupDevice(int deviceID, unsigned int w, unsigned int h, unsigned int idealFramerate)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0 )
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+        {
+            bool out = VD->setupDevice(w, h, idealFramerate);
+            if(!out)
+                DPO->printOut(L"VIDEODEVICE %i: this device cannot be started\n", deviceID);
+            return out;
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n", deviceID);
+    }
+    return false;
+}
+
+MediaType videoInput::getFormat(int deviceID, unsigned int id)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return MediaType();
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->getFormat(id);
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return MediaType();
+}
+
+bool videoInput::isDeviceSetup(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->isDeviceSetup();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return false;
+}
+
+bool videoInput::isDeviceMediaSource(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->isDeviceMediaSource();
+    }
+    else
+    {
+        DPO->printOut(L"Device(s): There is not any suitable video device\n");
+    }
+    return false;
+}
+
+bool videoInput::isDeviceRawDataSource(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+        {
+            bool isRaw = VD->isDeviceRawDataSource();
+            return isRaw;
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return false;
+}
+
+bool videoInput::isFrameNew(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return false;
+    }
+    if(accessToDevices)
+    {
+        if(!isDeviceSetup(deviceID))
+        {
+            if(isDeviceMediaSource(deviceID))
+                return false;
+        }
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+        {
+            return VD->isFrameNew();
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return false;
+}
+
+unsigned int videoInput::getCountFormats(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return 0;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->getCountFormats();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return 0;
+}
+
+void videoInput::closeAllDevices()
+{
+    videoDevices *VDS = &videoDevices::getInstance();
+    for(unsigned int i = 0; i < VDS->getCount(); i++)
+        closeDevice(i);
+}
+
+void videoInput::setParametrs(int deviceID, CamParametrs parametrs)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice *VD = VDS->getDevice(deviceID);
+        if(VD)
+            VD->setParametrs(parametrs);
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+}
+
+CamParametrs videoInput::getParametrs(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    CamParametrs out;
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return out;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice *VD = VDS->getDevice(deviceID);
+        if(VD)
+            out = VD->getParametrs();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return out;
+}
+
+void videoInput::closeDevice(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice *VD = VDS->getDevice(deviceID);
+        if(VD)
+            VD->closeDevice();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+}
+
+unsigned int videoInput::getWidth(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return 0;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->getWidth();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return 0;
+}
+
+unsigned int videoInput::getHeight(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return 0;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->getHeight();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return 0;
+}
+
+wchar_t *videoInput::getNameVideoDevice(int deviceID)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return NULL;
+    }
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        videoDevice * VD = VDS->getDevice(deviceID);
+        if(VD)
+            return VD->getName();
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return L"Empty";
+}
+
+unsigned int videoInput::listDevices(bool silent)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    int out = 0;
+    if(accessToDevices)
+    {
+        videoDevices *VDS = &videoDevices::getInstance();
+        out = VDS->getCount();
+        DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+        if(!silent)DPO->printOut(L"\nVIDEOINPUT SPY MODE!\n\n");
+        if(!silent)DPO->printOut(L"SETUP: Looking For Capture Devices\n");
+        for(int i = 0; i < out; i++)
+        {
+            if(!silent)DPO->printOut(L"SETUP: %i) %s \n",i, getNameVideoDevice(i));
+        }
+        if(!silent)DPO->printOut(L"SETUP: %i Device(s) found\n\n", out);
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return out;
+}
+
+videoInput& videoInput::getInstance()
+{
+    static videoInput instance;
+    return instance;
+}
+
+bool videoInput::isDevicesAcceable()
+{
+    return accessToDevices;
+}
+
+void videoInput::setVerbose(bool state)
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    DPO->setVerbose(state);
+}
+
+void videoInput::setEmergencyStopEvent(int deviceID, void *userData, void(*func)(int, void *))
+{
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return;
+    }
+    if(accessToDevices)
+    {
+        if(func)
+        {
+            videoDevices *VDS = &videoDevices::getInstance();
+            videoDevice * VD = VDS->getDevice(deviceID);
+            if(VD)
+                VD->setEmergencyStopEvent(userData, func);
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+}
+
+bool videoInput::getPixels(int deviceID, unsigned char * dstBuffer, bool flipRedAndBlue, bool flipImage)
+{
+    bool success = false;
+    unsigned int bytes = 3;
+    DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+    if (deviceID < 0)
+    {
+        DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
+        return success;
+    }
+    if(accessToDevices)
+    {
+        bool isRaw = isDeviceRawDataSource(deviceID);
+        if(isRaw)
+        {
+            videoDevices *VDS = &videoDevices::getInstance();
+            DebugPrintOut *DPO = &DebugPrintOut::getInstance();
+            RawImage *RIOut = VDS->getDevice(deviceID)->getRawImageOut();
+            if(RIOut)
+            {
+                unsigned int height = VDS->getDevice(deviceID)->getHeight();
+                unsigned int width  = VDS->getDevice(deviceID)->getWidth();
+                unsigned int size = bytes * width * height;
+                if(size == RIOut->getSize())
+                {
+                    processPixels(RIOut->getpPixels(), dstBuffer, width, height, bytes, flipRedAndBlue, flipImage);
+                    success = true;
+                }
+                else
+                {
+                    DPO->printOut(L"ERROR: GetPixels() - bufferSizes do not match!\n");
+                }
+            }
+            else
+            {
+                DPO->printOut(L"ERROR: GetPixels() - Unable to grab frame for device %i\n", deviceID);
+            }
+        }
+        else
+        {
+            DPO->printOut(L"ERROR: GetPixels() - Not raw data source device %i\n", deviceID);
+        }
+    }
+    else
+    {
+        DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
+    }
+    return success;
+}
+
+void videoInput::processPixels(unsigned char * src, unsigned char * dst, unsigned int width,
+                                unsigned int height, unsigned int bpp, bool bRGB, bool bFlip)
+{
+    unsigned int widthInBytes = width * bpp;
+    unsigned int numBytes = widthInBytes * height;
+    int *dstInt, *srcInt;
+    if(!bRGB)
+    {
+        if(bFlip)
+        {
+            for(unsigned int y = 0; y < height; y++)
+            {
+                dstInt = (int *)(dst + (y * widthInBytes));
+                srcInt = (int *)(src + ( (height -y -1) * widthInBytes));
+                memcpy(dstInt, srcInt, widthInBytes);
+            }
+        }
+        else
+        {
+            memcpy(dst, src, numBytes);
+        }
+    }
+    else
+    {
+        if(bFlip)
+        {
+            unsigned int x = 0;
+            unsigned int y = (height - 1) * widthInBytes;
+            src += y;
+            for(unsigned int i = 0; i < numBytes; i+=3)
+            {
+                if(x >= width)
+                {
+                    x = 0;
+                    src -= widthInBytes*2;
+                }
+                *dst = *(src+2);
+                dst++;
+                *dst = *(src+1);
+                dst++;
+                *dst = *src;
+                dst++;
+                src+=3;
+                x++;
+            }
+        }
+        else
+        {
+            for(unsigned int i = 0; i < numBytes; i+=3)
+            {
+                *dst = *(src+2);
+                dst++;
+                *dst = *(src+1);
+                dst++;
+                *dst = *src;
+                dst++;
+                src+=3;
+            }
+        }
+    }
+}
+}
+
+/******* Capturing video from camera via Microsoft Media Foundation **********/
+class CvCaptureCAM_MSMF : public CvCapture
+{
+public:
+    CvCaptureCAM_MSMF();
+    virtual ~CvCaptureCAM_MSMF();
+    virtual bool open( int index );
+    virtual void close();
+    virtual double getProperty(int);
+    virtual bool setProperty(int, double);
+    virtual bool grabFrame();
+    virtual IplImage* retrieveFrame(int);
+    virtual int getCaptureDomain() { return CV_CAP_MSMF; } // Return the type of the capture object: CV_CAP_VFW, etc...
+protected:
+    void init();
+    int index, width, height, fourcc;
+    IplImage* frame;
+    videoInput VI;
+};
+
+struct SuppressVideoInputMessages
+{
+    SuppressVideoInputMessages() { videoInput::setVerbose(true); }
+};
+
+static SuppressVideoInputMessages do_it;
+
+CvCaptureCAM_MSMF::CvCaptureCAM_MSMF():
+    index(-1),
+    width(-1),
+    height(-1),
+    fourcc(-1),
+    frame(NULL),
+    VI(videoInput::getInstance())
+{
+    CoInitialize(0);
+}
+
+CvCaptureCAM_MSMF::~CvCaptureCAM_MSMF()
+{
+    close();
+    CoUninitialize();
+}
+
+void CvCaptureCAM_MSMF::close()
+{
+    if( index >= 0 )
+    {
+        VI.closeDevice(index);
+        index = -1;
+        cvReleaseImage(&frame);
+    }
+    width = height = -1;
+}
+
+// Initialize camera input
+bool CvCaptureCAM_MSMF::open( int _index )
+{
+    int try_index = _index;
+    int devices = 0;
+    close();
+    devices = VI.listDevices(true);
+    if (devices == 0)
+        return false;
+    try_index = try_index < 0 ? 0 : (try_index > devices-1 ? devices-1 : try_index);
+    VI.setupDevice(try_index);
+    if( !VI.isFrameNew(try_index) )
+        return false;
+    index = try_index;
+    return true;
+}
+
+bool CvCaptureCAM_MSMF::grabFrame()
+{
+    while (VI.isDeviceSetup(index) && !VI.isFrameNew(index))
+        Sleep(1);
+    return VI.isDeviceSetup(index);
+}
+
+IplImage* CvCaptureCAM_MSMF::retrieveFrame(int)
+{
+    if( !frame || (int)VI.getWidth(index) != frame->width || (int)VI.getHeight(index) != frame->height )
+    {
+        if (frame)
+            cvReleaseImage( &frame );
+        unsigned int w = VI.getWidth(index), h = VI.getHeight(index);
+        frame = cvCreateImage( cvSize(w,h), 8, 3 );
+    }
+    VI.getPixels( index, (uchar*)frame->imageData, false, true );
+    return frame;
+}
+
+double CvCaptureCAM_MSMF::getProperty( int property_id )
+{
+    // image format proprrties
+    switch( property_id )
+    {
+    case CV_CAP_PROP_FRAME_WIDTH:
+        return VI.getWidth(index);
+    case CV_CAP_PROP_FRAME_HEIGHT:
+        return VI.getHeight(index);
+    }
+    return -1;
+}
+bool CvCaptureCAM_MSMF::setProperty( int property_id, double value )
+{
+    // image capture properties
+    bool handled = false;
+    switch( property_id )
+    {
+    case CV_CAP_PROP_FRAME_WIDTH:
+        width = cvRound(value);
+        handled = true;
+        break;
+    case CV_CAP_PROP_FRAME_HEIGHT:
+        height = cvRound(value);
+        handled = true;
+        break;
+    }
+
+    if ( handled ) {
+        if( width > 0 && height > 0 )
+        {
+            if( width != (int)VI.getWidth(index) || height != (int)VI.getHeight(index)  && VI.isDeviceSetup(index))//|| fourcc != VI.getFourcc(index) )
+            {
+                VI.closeDevice(index);
+                VI.setupDevice(index, width, height);
+            }
+            return VI.isDeviceSetup(index);
+        }
+        return true;
+    }
+
+    return false;
+}
+
+class CvCaptureFile_MSMF : public CvCapture
+{
+public:
+    CvCaptureFile_MSMF();
+    virtual ~CvCaptureFile_MSMF();
+
+    virtual bool open( const char* filename );
+    virtual void close();
+
+    virtual double getProperty(int);
+    virtual bool setProperty(int, double);
+    virtual bool grabFrame();
+    virtual IplImage* retrieveFrame(int);
+    virtual int getCaptureDomain() { return CV_CAP_MSMF; }
+protected:
+    ImageGrabberThread* grabberThread;
+    IMFMediaSource* videoFileSource;
+    std::vector<MediaType> captureFormats;
+    int captureFormatIndex;
+    IplImage* frame;
+    bool isOpened;
+
+    HRESULT enumerateCaptureFormats(IMFMediaSource *pSource);
+    HRESULT getSourceDuration(IMFMediaSource *pSource, MFTIME *pDuration);
+};
+
+CvCaptureFile_MSMF::CvCaptureFile_MSMF():
+    grabberThread(NULL),
+    videoFileSource(NULL),
+    captureFormatIndex(0),
+    frame(NULL),
+    isOpened(false)
+{
+    MFStartup(MF_VERSION);
+}
+
+CvCaptureFile_MSMF::~CvCaptureFile_MSMF()
+{
+    close();
+    MFShutdown();
+}
+
+bool CvCaptureFile_MSMF::open(const char* filename)
+{
+    if (!filename)
+        return false;
+
+    wchar_t* unicodeFileName = new wchar_t[strlen(filename)+1];
+    MultiByteToWideChar(CP_ACP, 0, filename, -1, unicodeFileName, strlen(filename)+1);
+
+    HRESULT hr = S_OK;
+
+    MF_OBJECT_TYPE ObjectType = MF_OBJECT_INVALID;
+
+    ComPtr<IMFSourceResolver> pSourceResolver = NULL;
+    IUnknown* pUnkSource = NULL;
+
+    hr = MFCreateSourceResolver(pSourceResolver.GetAddressOf());
+
+    if (SUCCEEDED(hr))
+    {
+        hr = pSourceResolver->CreateObjectFromURL(
+            unicodeFileName,
+            MF_RESOLUTION_MEDIASOURCE,
+            NULL, // Optional property store.
+            &ObjectType,
+            &pUnkSource
+            );
+    }
+
+    // Get the IMFMediaSource from the IUnknown pointer.
+    if (SUCCEEDED(hr))
+    {
+        hr = pUnkSource->QueryInterface(IID_PPV_ARGS(&videoFileSource));
+    }
+
+    SafeRelease(&pUnkSource);
+
+    if (SUCCEEDED(hr))
+    {
+        hr = enumerateCaptureFormats(videoFileSource);
+    }
+
+    if (SUCCEEDED(hr))
+    {
+        hr = ImageGrabberThread::CreateInstance(&grabberThread, videoFileSource, (unsigned int)-2, true);
+    }
+
+    if (SUCCEEDED(hr))
+    {
+        grabberThread->start();
+    }
+
+    isOpened = SUCCEEDED(hr);
+
+    return isOpened;
+}
+
+void CvCaptureFile_MSMF::close()
+{
+    if (grabberThread)
+    {
+        isOpened = false;
+        SetEvent(grabberThread->getImageGrabber()->ig_hFinish);
+        grabberThread->stop();
+        delete grabberThread;
+    }
+
+    if (videoFileSource)
+    {
+        videoFileSource->Shutdown();
+    }
+}
+
+bool CvCaptureFile_MSMF::setProperty(int property_id, double value)
+{
+    // image capture properties
+    // FIXME: implement method in VideoInput back end
+    (void) property_id;
+    (void) value;
+    return false;
+}
+
+double CvCaptureFile_MSMF::getProperty(int property_id)
+{
+    // image format proprrties
+    switch( property_id )
+    {
+    case CV_CAP_PROP_FRAME_WIDTH:
+        return captureFormats[captureFormatIndex].width;
+    case CV_CAP_PROP_FRAME_HEIGHT:
+        return captureFormats[captureFormatIndex].height;
+    case CV_CAP_PROP_FRAME_COUNT:
+        {
+            MFTIME duration;
+            getSourceDuration(this->videoFileSource, &duration);
+            double fps = ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_NUMERATOR) /
+            ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_DENOMINATOR);
+            return (double)floor(((double)duration/1e7)*fps+0.5);
+        }
+    case CV_CAP_PROP_FOURCC:
+        return captureFormats[captureFormatIndex].MF_MT_SUBTYPE.Data1;
+    case CV_CAP_PROP_FPS:
+        return ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_NUMERATOR) /
+            ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_DENOMINATOR);
+    }
+
+    return -1;
+}
+
+bool CvCaptureFile_MSMF::grabFrame()
+{
+    DWORD waitResult = (DWORD)-1;
+    if (isOpened)
+    {
+        SetEvent(grabberThread->getImageGrabber()->ig_hFrameGrabbed);
+        HANDLE tmp[] = {grabberThread->getImageGrabber()->ig_hFrameReady, grabberThread->getImageGrabber()->ig_hFinish, 0};
+        waitResult = WaitForMultipleObjects(2, tmp, FALSE, INFINITE);
+    }
+
+    return isOpened && grabberThread->getImageGrabber()->getRawImage()->isNew() && (waitResult == WAIT_OBJECT_0);
+}
+
+IplImage* CvCaptureFile_MSMF::retrieveFrame(int)
+{
+    unsigned int width = captureFormats[captureFormatIndex].width;
+    unsigned int height = captureFormats[captureFormatIndex].height;
+    unsigned int bytes = 3;
+    if( !frame || (int)width != frame->width || (int)height != frame->height )
+    {
+        if (frame)
+            cvReleaseImage( &frame );
+        frame = cvCreateImage( cvSize(width,height), 8, 3 );
+    }
+
+    RawImage *RIOut = grabberThread->getImageGrabber()->getRawImage();
+    unsigned int size = bytes * width * height;
+
+    bool verticalFlip = captureFormats[captureFormatIndex].MF_MT_DEFAULT_STRIDE < 0;
+
+    if(RIOut && size == RIOut->getSize())
+    {
+         videoInput::processPixels(RIOut->getpPixels(), (unsigned char*)frame->imageData, width,
+             height, bytes, false, verticalFlip);
+    }
+
+    return frame;
+}
+
+HRESULT CvCaptureFile_MSMF::enumerateCaptureFormats(IMFMediaSource *pSource)
+{
+    ComPtr<IMFPresentationDescriptor> pPD = NULL;
+    ComPtr<IMFStreamDescriptor> pSD = NULL;
+    ComPtr<IMFMediaTypeHandler> pHandler = NULL;
+    ComPtr<IMFMediaType> pType = NULL;
+    HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    BOOL fSelected;
+    hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    DWORD cTypes = 0;
+    hr = pHandler->GetMediaTypeCount(&cTypes);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+    for (DWORD i = 0; i < cTypes; i++)
+    {
+        hr = pHandler->GetMediaTypeByIndex(i, pType.GetAddressOf());
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+        MediaType MT = FormatReader::Read(pType.Get());
+        captureFormats.push_back(MT);
+    }
+
+done:
+    return hr;
+}
+
+HRESULT CvCaptureFile_MSMF::getSourceDuration(IMFMediaSource *pSource, MFTIME *pDuration)
+{
+    *pDuration = 0;
+
+    IMFPresentationDescriptor *pPD = NULL;
+
+    HRESULT hr = pSource->CreatePresentationDescriptor(&pPD);
+    if (SUCCEEDED(hr))
+    {
+        hr = pPD->GetUINT64(MF_PD_DURATION, (UINT64*)pDuration);
+        pPD->Release();
+    }
+    return hr;
+}
+
+CvCapture* cvCreateCameraCapture_MSMF( int index )
+{
+    CvCaptureCAM_MSMF* capture = new CvCaptureCAM_MSMF;
+    try
+    {
+        if( capture->open( index ))
+            return capture;
+    }
+    catch(...)
+    {
+        delete capture;
+        throw;
+    }
+    delete capture;
+    return 0;
+}
+
+CvCapture* cvCreateFileCapture_MSMF (const char* filename)
+{
+    CvCaptureFile_MSMF* capture = new CvCaptureFile_MSMF;
+    try
+    {
+        if( capture->open(filename) )
+            return capture;
+        else
+        {
+            delete capture;
+            return NULL;
+        }
+    }
+    catch(...)
+    {
+        delete capture;
+        throw;
+    }
+}
+
+//
+//
+// Media Foundation-based Video Writer
+//
+//
+
+class CvVideoWriter_MSMF : public CvVideoWriter
+{
+public:
+    CvVideoWriter_MSMF();
+    virtual ~CvVideoWriter_MSMF();
+    virtual bool open(const char* filename, int fourcc,
+                       double fps, CvSize frameSize, bool isColor);
+    virtual void close();
+    virtual bool writeFrame(const IplImage* img);
+
+private:
+    UINT32 videoWidth;
+    UINT32 videoHeight;
+    double fps;
+    UINT32 bitRate;
+    UINT32 frameSize;
+    GUID   encodingFormat;
+    GUID   inputFormat;
+
+    DWORD  streamIndex;
+    ComPtr<IMFSinkWriter> sinkWriter;
+
+    bool   initiated;
+
+    LONGLONG rtStart;
+    UINT64 rtDuration;
+
+    HRESULT InitializeSinkWriter(const char* filename);
+    static const GUID FourCC2GUID(int fourcc);
+    HRESULT WriteFrame(DWORD *videoFrameBuffer, const LONGLONG& rtStart, const LONGLONG& rtDuration);
+};
+
+CvVideoWriter_MSMF::CvVideoWriter_MSMF():
+    initiated(false)
+{
+}
+
+CvVideoWriter_MSMF::~CvVideoWriter_MSMF()
+{
+    close();
+}
+
+const GUID CvVideoWriter_MSMF::FourCC2GUID(int fourcc)
+{
+    switch(fourcc)
+    {
+        case CV_FOURCC_MACRO('d', 'v', '2', '5'):
+            return MFVideoFormat_DV25; break;
+        case CV_FOURCC_MACRO('d', 'v', '5', '0'):
+            return MFVideoFormat_DV50; break;
+        case CV_FOURCC_MACRO('d', 'v', 'c', ' '):
+            return MFVideoFormat_DVC; break;
+        case CV_FOURCC_MACRO('d', 'v', 'h', '1'):
+            return MFVideoFormat_DVH1; break;
+        case CV_FOURCC_MACRO('d', 'v', 'h', 'd'):
+            return MFVideoFormat_DVHD; break;
+        case CV_FOURCC_MACRO('d', 'v', 's', 'd'):
+            return MFVideoFormat_DVSD; break;
+        case CV_FOURCC_MACRO('d', 'v', 's', 'l'):
+                return MFVideoFormat_DVSL; break;
+        case CV_FOURCC_MACRO('H', '2', '6', '3'):
+                return MFVideoFormat_H263; break;
+        case CV_FOURCC_MACRO('H', '2', '6', '4'):
+                return MFVideoFormat_H264; break;
+        case CV_FOURCC_MACRO('M', '4', 'S', '2'):
+                return MFVideoFormat_M4S2; break;
+        case CV_FOURCC_MACRO('M', 'J', 'P', 'G'):
+                return MFVideoFormat_MJPG; break;
+        case CV_FOURCC_MACRO('M', 'P', '4', '3'):
+                return MFVideoFormat_MP43; break;
+        case CV_FOURCC_MACRO('M', 'P', '4', 'S'):
+                return MFVideoFormat_MP4S; break;
+        case CV_FOURCC_MACRO('M', 'P', '4', 'V'):
+                return MFVideoFormat_MP4V; break;
+        case CV_FOURCC_MACRO('M', 'P', 'G', '1'):
+                return MFVideoFormat_MPG1; break;
+        case CV_FOURCC_MACRO('M', 'S', 'S', '1'):
+                return MFVideoFormat_MSS1; break;
+        case CV_FOURCC_MACRO('M', 'S', 'S', '2'):
+                return MFVideoFormat_MSS2; break;
+        case CV_FOURCC_MACRO('W', 'M', 'V', '1'):
+                return MFVideoFormat_WMV1; break;
+        case CV_FOURCC_MACRO('W', 'M', 'V', '2'):
+                return MFVideoFormat_WMV2; break;
+        case CV_FOURCC_MACRO('W', 'M', 'V', '3'):
+                return MFVideoFormat_WMV3; break;
+        case CV_FOURCC_MACRO('W', 'V', 'C', '1'):
+                return MFVideoFormat_WVC1; break;
+        default:
+            return MFVideoFormat_H264;
+    }
+}
+
+bool CvVideoWriter_MSMF::open( const char* filename, int fourcc,
+                       double _fps, CvSize frameSize, bool /*isColor*/ )
+{
+    videoWidth = frameSize.width;
+    videoHeight = frameSize.height;
+    fps = _fps;
+    bitRate = (UINT32)fps*videoWidth*videoHeight; // 1-bit per pixel
+    encodingFormat = FourCC2GUID(fourcc);
+    inputFormat = MFVideoFormat_RGB32;
+
+    HRESULT hr = CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
+    if (SUCCEEDED(hr))
+    {
+        hr = MFStartup(MF_VERSION);
+        if (SUCCEEDED(hr))
+        {
+            hr = InitializeSinkWriter(filename);
+            if (SUCCEEDED(hr))
+            {
+                initiated = true;
+                rtStart = 0;
+                MFFrameRateToAverageTimePerFrame((UINT32)fps, 1, &rtDuration);
+            }
+        }
+    }
+
+    return SUCCEEDED(hr);
+}
+
+void CvVideoWriter_MSMF::close()
+{
+    if (!initiated)
+    {
+        return;
+    }
+
+    initiated = false;
+    sinkWriter->Finalize();
+    MFShutdown();
+}
+
+bool CvVideoWriter_MSMF::writeFrame(const IplImage* img)
+{
+    if (!img)
+        return false;
+
+    int length = img->width * img->height * 4;
+    DWORD* target = new DWORD[length];
+
+    for (int rowIdx = 0; rowIdx < img->height; rowIdx++)
+    {
+        char* rowStart = img->imageData + rowIdx*img->widthStep;
+        for (int colIdx = 0; colIdx < img->width; colIdx++)
+        {
+            BYTE b = rowStart[colIdx * img->nChannels + 0];
+            BYTE g = rowStart[colIdx * img->nChannels + 1];
+            BYTE r = rowStart[colIdx * img->nChannels + 2];
+
+            target[rowIdx*img->width+colIdx] = (r << 16) + (g << 8) + b;
+        }
+    }
+
+    // Send frame to the sink writer.
+    HRESULT hr = WriteFrame(target, rtStart, rtDuration);
+    if (FAILED(hr))
+    {
+        delete[] target;
+        return false;
+    }
+    rtStart += rtDuration;
+
+    delete[] target;
+
+    return true;
+}
+
+HRESULT CvVideoWriter_MSMF::InitializeSinkWriter(const char* filename)
+{
+    ComPtr<IMFAttributes> spAttr;
+    ComPtr<IMFMediaType>  mediaTypeOut;
+    ComPtr<IMFMediaType>  mediaTypeIn;
+    ComPtr<IMFByteStream> spByteStream;
+
+    MFCreateAttributes(&spAttr, 10);
+    spAttr->SetUINT32(MF_READWRITE_ENABLE_HARDWARE_TRANSFORMS, true);
+
+    wchar_t* unicodeFileName = new wchar_t[strlen(filename)+1];
+    MultiByteToWideChar(CP_ACP, 0, filename, -1, unicodeFileName, strlen(filename)+1);
+
+    HRESULT hr = MFCreateSinkWriterFromURL(unicodeFileName, NULL, spAttr.Get(), &sinkWriter);
+
+    delete[] unicodeFileName;
+
+    // Set the output media type.
+    if (SUCCEEDED(hr))
+    {
+        hr = MFCreateMediaType(&mediaTypeOut);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeOut->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeOut->SetGUID(MF_MT_SUBTYPE, encodingFormat);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeOut->SetUINT32(MF_MT_AVG_BITRATE, bitRate);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeOut->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeSize(mediaTypeOut.Get(), MF_MT_FRAME_SIZE, videoWidth, videoHeight);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeRatio(mediaTypeOut.Get(), MF_MT_FRAME_RATE, (UINT32)fps, 1);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeRatio(mediaTypeOut.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
+    }
+
+    if (SUCCEEDED(hr))
+    {
+        hr = sinkWriter->AddStream(mediaTypeOut.Get(), &streamIndex);
+    }
+
+    // Set the input media type.
+    if (SUCCEEDED(hr))
+    {
+        hr = MFCreateMediaType(&mediaTypeIn);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeIn->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeIn->SetGUID(MF_MT_SUBTYPE, inputFormat);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = mediaTypeIn->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeSize(mediaTypeIn.Get(), MF_MT_FRAME_SIZE, videoWidth, videoHeight);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeRatio(mediaTypeIn.Get(), MF_MT_FRAME_RATE, (UINT32)fps, 1);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = MFSetAttributeRatio(mediaTypeIn.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
+    }
+
+    if (SUCCEEDED(hr))
+    {
+        hr = sinkWriter->SetInputMediaType(streamIndex, mediaTypeIn.Get(), NULL);
+    }
+
+    // Tell the sink writer to start accepting data.
+    if (SUCCEEDED(hr))
+    {
+        hr = sinkWriter->BeginWriting();
+    }
+
+    return hr;
+}
+
+HRESULT CvVideoWriter_MSMF::WriteFrame(DWORD *videoFrameBuffer, const LONGLONG& Start, const LONGLONG& Duration)
+{
+    ComPtr<IMFSample> sample;
+    ComPtr<IMFMediaBuffer> buffer;
+
+    const LONG cbWidth = 4 * videoWidth;
+    const DWORD cbBuffer = cbWidth * videoHeight;
+
+    BYTE *pData = NULL;
+
+    // Create a new memory buffer.
+    HRESULT hr = MFCreateMemoryBuffer(cbBuffer, &buffer);
+
+    // Lock the buffer and copy the video frame to the buffer.
+    if (SUCCEEDED(hr))
+    {
+        hr = buffer->Lock(&pData, NULL, NULL);
+    }
+
+    if (SUCCEEDED(hr))
+    {
+#if defined(_M_ARM)
+        hr = MFCopyImage(
+            pData,                      // Destination buffer.
+            -cbWidth,                   // Destination stride.
+            (BYTE*)videoFrameBuffer,    // First row in source image.
+            cbWidth,                    // Source stride.
+            cbWidth,                    // Image width in bytes.
+            videoHeight                 // Image height in pixels.
+            );
+#else
+        hr = MFCopyImage(
+            pData,                      // Destination buffer.
+            cbWidth,                    // Destination stride.
+            (BYTE*)videoFrameBuffer,    // First row in source image.
+            cbWidth,                    // Source stride.
+            cbWidth,                    // Image width in bytes.
+            videoHeight                 // Image height in pixels.
+            );
+#endif
+    }
+
+    if (buffer)
+    {
+        buffer->Unlock();
+    }
+
+    // Set the data length of the buffer.
+    if (SUCCEEDED(hr))
+    {
+        hr = buffer->SetCurrentLength(cbBuffer);
+    }
+
+    // Create a media sample and add the buffer to the sample.
+    if (SUCCEEDED(hr))
+    {
+        hr = MFCreateSample(&sample);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = sample->AddBuffer(buffer.Get());
+    }
+
+    // Set the time stamp and the duration.
+    if (SUCCEEDED(hr))
+    {
+        hr = sample->SetSampleTime(Start);
+    }
+    if (SUCCEEDED(hr))
+    {
+        hr = sample->SetSampleDuration(Duration);
+    }
+
+    // Send the sample to the Sink Writer.
+    if (SUCCEEDED(hr))
+    {
+        hr = sinkWriter->WriteSample(streamIndex, sample.Get());
+    }
+
+    return hr;
+}
+
+CvVideoWriter* cvCreateVideoWriter_MSMF( const char* filename, int fourcc,
+                                        double fps, CvSize frameSize, int isColor )
+{
+    CvVideoWriter_MSMF* writer = new CvVideoWriter_MSMF;
+    if( writer->open( filename, fourcc, fps, frameSize, isColor != 0 ))
+        return writer;
+    delete writer;
+    return NULL;
+}
+
+#endif
diff --git a/modules/highgui/src/cap_openni.cpp b/modules/highgui/src/cap_openni.cpp
index fed59de748..1ca140634a 100644
--- a/modules/highgui/src/cap_openni.cpp
+++ b/modules/highgui/src/cap_openni.cpp
@@ -48,7 +48,6 @@
 # undef HAVE_TBB
 #endif
 
-#include <iostream>
 #include <queue>
 
 #ifndef i386
@@ -66,7 +65,7 @@
 
 #include "XnCppWrapper.h"
 
-const std::string XMLConfig =
+const cv::String XMLConfig =
 "<OpenNI>"
         "<Licenses>"
         "<License vendor=\"PrimeSense\" key=\"0KOIk2JeIBYClPWVnMoRKn5cdY4=\"/>"
@@ -558,8 +557,7 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.Init();
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to initialize the context: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to initialize the context: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -568,17 +566,16 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.EnumerateProductionTrees( XN_NODE_TYPE_DEVICE, NULL, devicesList, 0 );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate production trees: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate production trees: %s\n", xnGetStatusString(status));
         return;
     }
 
     // Chose device according to index
     xn::NodeInfoList::Iterator it = devicesList.Begin();
-        for( int i = 0; i < index && it!=devicesList.End(); ++i ) it++;
+    for( int i = 0; i < index && it!=devicesList.End(); ++i ) it++;
         if ( it == devicesList.End() )
         {
-            std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed device with index " << index << std::endl;
+            fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed device with index %d\n", index);
             return;
         }
 
@@ -586,8 +583,7 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.CreateProductionTree( deviceNode, productionNode );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create production tree: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create production tree: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -595,8 +591,7 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.RunXmlScript( XMLConfig.c_str(), scriptNode );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to run xml script: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to run xml script: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -606,20 +601,18 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.EnumerateExistingNodes( depthList, XN_NODE_TYPE_DEPTH );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate depth generators: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate depth generators: %s\n", xnGetStatusString(status));
         return;
     }
     if( depthList.IsEmpty() )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : The device doesn't have depth generator. Such devices aren't supported now." << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : The device doesn't have depth generator. Such devices aren't supported now.\n");
         return;
     }
     status = depthGenerator.Create( context );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create depth generator: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create depth generator: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -628,8 +621,7 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.EnumerateExistingNodes( imageList, XN_NODE_TYPE_IMAGE );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate image generators: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to enumerate image generators: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -638,8 +630,7 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
         status = imageGenerator.Create( context );
         if( status != XN_STATUS_OK )
         {
-            std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create image generator: "
-                      <<  std::string(xnGetStatusString(status)) << std::endl;
+            fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to create image generator: %s\n", xnGetStatusString(status));
             return;
         }
     }
@@ -666,14 +657,13 @@ CvCapture_OpenNI::CvCapture_OpenNI( int index )
     status = context.StartGeneratingAll();
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to start generating OpenNI data: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to start generating OpenNI data: %s\n", xnGetStatusString(status));
         return;
     }
 
     if( !readCamerasParams() )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Could not read cameras parameters" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Could not read cameras parameters\n");
         return;
     }
 
@@ -697,8 +687,7 @@ CvCapture_OpenNI::CvCapture_OpenNI(const char * filename)
     status = context.Init();
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to initialize the context: "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to initialize the context: %s\n", xnGetStatusString(status));
         return;
     }
 
@@ -706,8 +695,7 @@ CvCapture_OpenNI::CvCapture_OpenNI(const char * filename)
     status = context.OpenFileRecording( filename, productionNode );
     if( status != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Failed to open input file (" << filename << "): "
-                  << std::string(xnGetStatusString(status)) << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Failed to open input file (%s): %s\n", filename, xnGetStatusString(status));
         return;
     }
 
@@ -716,7 +704,7 @@ CvCapture_OpenNI::CvCapture_OpenNI(const char * filename)
 
     if( !readCamerasParams() )
     {
-        std::cerr << "CvCapture_OpenNI::CvCapture_OpenNI : Could not read cameras parameters" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::CvCapture_OpenNI : Could not read cameras parameters\n");
         return;
     }
 
@@ -736,7 +724,7 @@ bool CvCapture_OpenNI::readCamerasParams()
     XnDouble pixelSize = 0;
     if( depthGenerator.GetRealProperty( "ZPPS", pixelSize ) != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::readCamerasParams : Could not read pixel size!" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::readCamerasParams : Could not read pixel size!\n");
         return false;
     }
 
@@ -747,13 +735,13 @@ bool CvCapture_OpenNI::readCamerasParams()
     XnUInt64 zeroPlanDistance; // in mm
     if( depthGenerator.GetIntProperty( "ZPD", zeroPlanDistance ) != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::readCamerasParams : Could not read virtual plane distance!" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::readCamerasParams : Could not read virtual plane distance!\n");
         return false;
     }
 
     if( depthGenerator.GetRealProperty( "LDDIS", baseline ) != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::readCamerasParams : Could not read base line!" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::readCamerasParams : Could not read base line!\n");
         return false;
     }
 
@@ -765,13 +753,13 @@ bool CvCapture_OpenNI::readCamerasParams()
 
     if( depthGenerator.GetIntProperty( "ShadowValue", shadowValue ) != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::readCamerasParams : Could not read property \"ShadowValue\"!" << std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::readCamerasParams : Could not read property \"ShadowValue\"!\n");
         return false;
     }
 
     if( depthGenerator.GetIntProperty("NoSampleValue", noSampleValue ) != XN_STATUS_OK )
     {
-        std::cerr << "CvCapture_OpenNI::readCamerasParams : Could not read property \"NoSampleValue\"!" <<std::endl;
+        fprintf(stderr, "CvCapture_OpenNI::readCamerasParams : Could not read property \"NoSampleValue\"!\n");
         return false;
     }
 
@@ -861,11 +849,7 @@ double CvCapture_OpenNI::getCommonProperty( int propIdx )
         propValue = maxTimeDuration;
         break;
     default :
-    {
-        std::stringstream ss;
-        ss << "Such parameter (propIdx=" << propIdx << ") isn't supported for getting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Such parameter (propIdx=%d) isn't supported for getting.\n", propIdx) );
     }
 
     return propValue;
@@ -922,11 +906,7 @@ bool CvCapture_OpenNI::setCommonProperty( int propIdx, double propValue )
             approxSyncGrabber->setMaxTimeDuration(maxTimeDuration);
         break;
     default:
-    {
-        std::stringstream ss;
-        ss << "Such parameter (propIdx=" << propIdx << ") isn't supported for setting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Such parameter (propIdx=%d) isn't supported for setting.\n", propIdx) );
     }
 
     return isSet;
@@ -977,11 +957,7 @@ double CvCapture_OpenNI::getDepthGeneratorProperty( int propIdx )
         propValue = depthGenerator.GetFrameID();
         break;
     default :
-    {
-        std::stringstream ss;
-        ss << "Depth generator does not support such parameter (propIdx=" << propIdx << ") for getting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Depth generator does not support such parameter (propIdx=%d) for getting.\n", propIdx) );
     }
 
     return propValue;
@@ -1009,12 +985,12 @@ bool CvCapture_OpenNI::setDepthGeneratorProperty( int propIdx, double propValue
                         {
                             XnStatus status = depthGenerator.GetAlternativeViewPointCap().SetViewPoint(imageGenerator);
                             if( status != XN_STATUS_OK )
-                                std::cerr << "CvCapture_OpenNI::setDepthGeneratorProperty : " << xnGetStatusString(status) << std::endl;
+                                fprintf(stderr, "CvCapture_OpenNI::setDepthGeneratorProperty : %s\n", xnGetStatusString(status));
                             else
                                 isSet = true;
                         }
                         else
-                            std::cerr << "CvCapture_OpenNI::setDepthGeneratorProperty : Unsupported viewpoint." << std::endl;
+                            fprintf(stderr, "CvCapture_OpenNI::setDepthGeneratorProperty : Unsupported viewpoint.\n");
                     }
                     else
                         isSet = true;
@@ -1024,18 +1000,14 @@ bool CvCapture_OpenNI::setDepthGeneratorProperty( int propIdx, double propValue
             {
                 XnStatus status = depthGenerator.GetAlternativeViewPointCap().ResetViewPoint();
                 if( status != XN_STATUS_OK )
-                    std::cerr << "CvCapture_OpenNI::setDepthGeneratorProperty : " << xnGetStatusString(status) << std::endl;
+                    fprintf(stderr, "CvCapture_OpenNI::setDepthGeneratorProperty : %s\n", xnGetStatusString(status));
                 else
                     isSet = true;
             }
         }
         break;
     default:
-    {
-        std::stringstream ss;
-        ss << "Depth generator does not support such parameter (propIdx=" << propIdx << ") for setting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Depth generator does not support such parameter (propIdx=%d) for setting.\n", propIdx) );
     }
 
     return isSet;
@@ -1073,11 +1045,7 @@ double CvCapture_OpenNI::getImageGeneratorProperty( int propIdx )
         propValue = imageGenerator.GetFrameID();
         break;
     default :
-    {
-        std::stringstream ss;
-        ss << "Image generator does not support such parameter (propIdx=" << propIdx << ") for getting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Image generator does not support such parameter (propIdx=%d) for getting.\n", propIdx) );
     }
 
     return propValue;
@@ -1128,17 +1096,13 @@ bool CvCapture_OpenNI::setImageGeneratorProperty( int propIdx, double propValue
 
         XnStatus status = imageGenerator.SetMapOutputMode( mode );
         if( status != XN_STATUS_OK )
-            std::cerr << "CvCapture_OpenNI::setImageGeneratorProperty : " << xnGetStatusString(status) << std::endl;
+            fprintf(stderr, "CvCapture_OpenNI::setImageGeneratorProperty : %s\n", xnGetStatusString(status));
         else
             isSet = true;
         break;
     }
     default:
-    {
-        std::stringstream ss;
-        ss << "Image generator does not support such parameter (propIdx=" << propIdx << ") for setting.\n";
-        CV_Error( CV_StsBadArg, ss.str().c_str() );
-    }
+        CV_Error( CV_StsBadArg, cv::format("Image generator does not support such parameter (propIdx=%d) for setting.\n", propIdx) );
     }
 
     return isSet;
diff --git a/modules/highgui/src/cap_pvapi.cpp b/modules/highgui/src/cap_pvapi.cpp
index 8d4a4a2b45..6ed3aea840 100644
--- a/modules/highgui/src/cap_pvapi.cpp
+++ b/modules/highgui/src/cap_pvapi.cpp
@@ -63,7 +63,6 @@
 #  include <unistd.h>
 #endif
 
-#include <string>
 //#include <arpa/inet.h>
 
 #define MAX_CAMERAS 10
@@ -356,7 +355,7 @@ bool CvCaptureCAM_PvAPI::setProperty( int property_id, double value )
         }
         else
         {
-            std::string ip=cv::format("%d.%d.%d.%d", ((int)value>>24)&255, ((int)value>>16)&255, ((int)value>>8)&255, (int)value&255);
+            cv::String ip=cv::format("%d.%d.%d.%d", ((int)value>>24)&255, ((int)value>>16)&255, ((int)value>>8)&255, (int)value&255);
             if ((PvAttrEnumSet(Camera.Handle,"MulticastEnable", "On")==ePvErrSuccess) &&
                 (PvAttrStringSet(Camera.Handle, "MulticastIPAddress", ip.c_str())==ePvErrSuccess))
                 break;
diff --git a/modules/highgui/src/cap_qt.cpp b/modules/highgui/src/cap_qt.cpp
index 1e941d06af..033b1f274e 100644
--- a/modules/highgui/src/cap_qt.cpp
+++ b/modules/highgui/src/cap_qt.cpp
@@ -1196,7 +1196,7 @@ static void icvReleaseVideoWriter_QT( CvVideoWriter_QT ** writer ) {
         // force compression session to complete encoding of outstanding source
         // frames
         ICMCompressionSessionCompleteFrames(
-            video_writer->compression_session_ref, TRUE, 0, 0
+            video_writer->compression_session_ref, 1, 0, 0
         );
 
         EndMediaEdits( video_writer->video );
diff --git a/modules/highgui/src/cap_qtkit.mm b/modules/highgui/src/cap_qtkit.mm
index d44b7636d7..8a7b3d84ec 100644
--- a/modules/highgui/src/cap_qtkit.mm
+++ b/modules/highgui/src/cap_qtkit.mm
@@ -1,32 +1,44 @@
-/*
- *  CvCapture.mm
- *
- *  Created by Nicholas Butko on 11/3/09.
- *  Copyright 2009. All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *
- * 1. Redistributions of source code must retain the above copyright notice,
- *    this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright notice,
- *    this list of conditions and the following disclaimer in the documentation
- *    and/or other materials provided with the distribution.
- * 3. The name of the author may not be used to endorse or promote products
- *    derived from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR "AS IS" AND ANY EXPRESS OR IMPLIED
- * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
- * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
- * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
- * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
- * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
- * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
- * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- */
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+// By downloading, copying, installing or using the software you agree to this license.
+// If you do not agree to this license, do not download, install,
+// copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+// * Redistribution's of source code must retain the above copyright notice,
+// this list of conditions and the following disclaimer.
+//
+// * Redistribution's in binary form must reproduce the above copyright notice,
+// this list of conditions and the following disclaimer in the documentation
+// and/or other materials provided with the distribution.
+//
+// * The name of the copyright holders may not be used to endorse or promote products
+// derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the contributor be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*////////////////////////////////////////////////////////////////////////////////////////
+
 
 #include "precomp.hpp"
 #include "opencv2/imgproc.hpp"
@@ -399,6 +411,9 @@ int CvCaptureCAM::startCaptureDevice(int cameraNum) {
 
 void CvCaptureCAM::setWidthHeight() {
     NSAutoreleasePool* localpool = [[NSAutoreleasePool alloc] init];
+
+    [mCaptureSession stopRunning];
+
     NSDictionary* pixelBufferOptions = [NSDictionary dictionaryWithObjectsAndKeys:
                           [NSNumber numberWithDouble:1.0*width], (id)kCVPixelBufferWidthKey,
                           [NSNumber numberWithDouble:1.0*height], (id)kCVPixelBufferHeightKey,
@@ -407,6 +422,9 @@ void CvCaptureCAM::setWidthHeight() {
                           nil];
 
     [mCaptureDecompressedVideoOutput setPixelBufferAttributes:pixelBufferOptions];
+
+    [mCaptureSession startRunning];
+
     grabFrame(60);
     [localpool drain];
 }
@@ -1026,4 +1044,3 @@ bool CvVideoWriter_QT::writeFrame(const IplImage* image) {
 
     return 1;
 }
-
diff --git a/modules/highgui/src/cap_v4l.cpp b/modules/highgui/src/cap_v4l.cpp
index 829d0ab638..045c6f889c 100644
--- a/modules/highgui/src/cap_v4l.cpp
+++ b/modules/highgui/src/cap_v4l.cpp
@@ -14,7 +14,9 @@ It has been tested with the motempl sample program
 First Patch:  August 24, 2004 Travis Wood   TravisOCV@tkwood.com
 For Release:  OpenCV-Linux Beta4  opencv-0.9.6
 Tested On:    LMLBT44 with 8 video inputs
-Problems?     Post problems/fixes to OpenCV group on groups.yahoo.com
+Problems?     Post your questions at answers.opencv.org,
+              Report bugs at code.opencv.org,
+              Submit your fixes at https://github.com/Itseez/opencv/
 Patched Comments:
 
 TW: The cv cam utils that came with the initial release of OpenCV for LINUX Beta4
@@ -154,6 +156,11 @@ the symptoms were damaged image and 'Corrupt JPEG data: premature end of data se
 - USE_TEMP_BUFFER fixes the main problem (improper buffer management) and
   prevents bad images in the first place
 
+11th patch: April 2, 2013, Forrest Reiling forrest.reiling@gmail.com
+Added v4l2 support for getting capture property CV_CAP_PROP_POS_MSEC.
+Returns the millisecond timestamp of the last frame grabbed or 0 if no frames have been grabbed
+Used to successfully synchonize 2 Logitech C310 USB webcams to within 16 ms of one another
+
 
 make & enjoy!
 
@@ -320,6 +327,8 @@ typedef struct CvCaptureCAM_V4L
    struct v4l2_queryctrl queryctrl;
    struct v4l2_querymenu querymenu;
 
+   struct timeval timestamp;
+
    /* V4L2 control variables */
    int v4l2_brightness, v4l2_brightness_min, v4l2_brightness_max;
    int v4l2_contrast, v4l2_contrast_min, v4l2_contrast_max;
@@ -836,6 +845,9 @@ static int _capture_V4L2 (CvCaptureCAM_V4L *capture, char *deviceName)
    capture->v4l2_gain_max = 0;
    capture->v4l2_exposure_max = 0;
 
+   capture->timestamp.tv_sec = 0;
+   capture->timestamp.tv_usec = 0;
+
    /* Scan V4L2 controls */
    v4l2_scan_controls(capture);
 
@@ -1221,6 +1233,9 @@ static int read_frame_v4l2(CvCaptureCAM_V4L* capture) {
    if (-1 == ioctl (capture->deviceHandle, VIDIOC_QBUF, &buf))
        perror ("VIDIOC_QBUF");
 
+   //set timestamp in capture struct to be timestamp of most recent frame
+   capture->timestamp = buf.timestamp;
+
    return 1;
 }
 
@@ -2308,6 +2323,13 @@ static double icvGetPropertyCAM_V4L (CvCaptureCAM_V4L* capture,
       /* initialize the control structure */
 
       switch (property_id) {
+      case CV_CAP_PROP_POS_MSEC:
+          if (capture->FirstCapture) {
+            return 0;
+          } else {
+            return 1000 * capture->timestamp.tv_sec + ((double) capture->timestamp.tv_usec) / 1000;
+          }
+          break;
       case CV_CAP_PROP_BRIGHTNESS:
           capture->control.id = V4L2_CID_BRIGHTNESS;
           break;
diff --git a/modules/highgui/src/cap_vfw.cpp b/modules/highgui/src/cap_vfw.cpp
index 4e6ff5e13a..d845953f8e 100644
--- a/modules/highgui/src/cap_vfw.cpp
+++ b/modules/highgui/src/cap_vfw.cpp
@@ -406,7 +406,7 @@ bool CvCaptureCAM_VFW::open( int wIndex )
         fourcc = (DWORD)-1;
 
         memset( &caps, 0, sizeof(caps));
-        capDriverGetCaps( hWndC, &caps, sizeof(&caps));
+        capDriverGetCaps( hWndC, &caps, sizeof(caps));
         ::MoveWindow( hWndC, 0, 0, 320, 240, TRUE );
         capSetUserData( hWndC, (size_t)this );
         capSetCallbackOnFrame( hWndC, frameCallback );
@@ -613,8 +613,10 @@ bool CvVideoWriter_VFW::open( const char* filename, int _fourcc, double _fps, Cv
             close();
             return false;
         }
+        return true;
     }
-    return true;
+    else
+        return false;
 }
 
 
diff --git a/modules/highgui/src/cap_ximea.cpp b/modules/highgui/src/cap_ximea.cpp
index dbb8f58683..7292727b76 100644
--- a/modules/highgui/src/cap_ximea.cpp
+++ b/modules/highgui/src/cap_ximea.cpp
@@ -20,25 +20,24 @@ public:
     virtual IplImage* retrieveFrame(int);
     virtual int getCaptureDomain() { return CV_CAP_XIAPI; } // Return the type of the capture object: CV_CAP_VFW, etc...
 
-protected:
+private:
     void init();
     void errMsg(const char* msg, int errNum);
+    void resetCvImage();
+    int  getBpp();
     IplImage* frame;
 
     HANDLE    hmv;
     DWORD     numDevices;
-    XI_IMG    image;
-    int       width;
-    int       height;
-    int       format;
     int       timeout;
+    XI_IMG    image;
 };
 
 /**********************************************************************************/
 
 CvCapture* cvCreateCameraCapture_XIMEA( int index )
 {
-     CvCaptureCAM_XIMEA* capture = new CvCaptureCAM_XIMEA;
+    CvCaptureCAM_XIMEA* capture = new CvCaptureCAM_XIMEA;
 
     if( capture->open( index ))
         return capture;
@@ -53,6 +52,8 @@ void CvCaptureCAM_XIMEA::init()
 {
     xiGetNumberDevices( &numDevices);
     hmv = NULL;
+    frame = NULL;
+    timeout = 0;
     memset(&image, 0, sizeof(XI_IMG));
 }
 
@@ -61,6 +62,8 @@ void CvCaptureCAM_XIMEA::init()
 // Initialize camera input
 bool CvCaptureCAM_XIMEA::open( int wIndex )
 {
+#define HandleXiResult(res) if (res!=XI_OK)  goto error;
+
     int mvret = XI_OK;
 
     if(numDevices == 0)
@@ -74,25 +77,42 @@ bool CvCaptureCAM_XIMEA::open( int wIndex )
 
     // always use auto exposure/gain
     mvret = xiSetParamInt( hmv, XI_PRM_AEAG, 1);
-    if(mvret != XI_OK) goto error;
-
-    // always use auto white ballance
-    mvret = xiSetParamInt( hmv, XI_PRM_AUTO_WB, 1);
-    if(mvret != XI_OK) goto error;
+    HandleXiResult(mvret);
 
+    int width = 0;
     mvret = xiGetParamInt( hmv, XI_PRM_WIDTH, &width);
-    if(mvret != XI_OK) goto error;
+    HandleXiResult(mvret);
 
+    int height = 0;
     mvret = xiGetParamInt( hmv, XI_PRM_HEIGHT, &height);
-    if(mvret != XI_OK) goto error;
+    HandleXiResult(mvret);
 
-    // default image format RGB24
-    format = XI_RGB24;
-    mvret = xiSetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, format);
-    if(mvret != XI_OK) goto error;
+    int isColor = 0;
+    mvret = xiGetParamInt(hmv, XI_PRM_IMAGE_IS_COLOR, &isColor);
+    HandleXiResult(mvret);
 
-    // allocate frame buffer for RGB24 image
-    frame = cvCreateImage(cvSize( width, height), IPL_DEPTH_8U, 3);
+    if(isColor)	// for color cameras
+    {
+        // default image format RGB24
+        mvret = xiSetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, XI_RGB24);
+        HandleXiResult(mvret);
+
+        // always use auto white ballance for color cameras
+        mvret = xiSetParamInt( hmv, XI_PRM_AUTO_WB, 1);
+        HandleXiResult(mvret);
+
+        // allocate frame buffer for RGB24 image
+        frame = cvCreateImage(cvSize( width, height), IPL_DEPTH_8U, 3);
+    }
+    else // for mono cameras
+    {
+        // default image format MONO8
+        mvret = xiSetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, XI_MONO8);
+        HandleXiResult(mvret);
+
+        // allocate frame buffer for MONO8 image
+        frame = cvCreateImage(cvSize( width, height), IPL_DEPTH_8U, 1);
+    }
 
     //default capture timeout 10s
     timeout = 10000;
@@ -103,10 +123,10 @@ bool CvCaptureCAM_XIMEA::open( int wIndex )
         errMsg("StartAcquisition XI_DEVICE failed", mvret);
         goto error;
     }
-
     return true;
 
 error:
+    errMsg("Open XI_DEVICE failed", mvret);
     xiCloseDevice(hmv);
     hmv = NULL;
     return false;
@@ -116,18 +136,22 @@ error:
 
 void CvCaptureCAM_XIMEA::close()
 {
+    if(frame)
+        cvReleaseImage(&frame);
+
     if(hmv)
     {
         xiStopAcquisition(hmv);
         xiCloseDevice(hmv);
-        hmv = NULL;
     }
+    hmv = NULL;
 }
 
 /**********************************************************************************/
 
 bool CvCaptureCAM_XIMEA::grabFrame()
 {
+    memset(&image, 0, sizeof(XI_IMG));
     image.size = sizeof(XI_IMG);
     int mvret = xiGetImage( hmv, timeout, &image);
 
@@ -151,31 +175,18 @@ bool CvCaptureCAM_XIMEA::grabFrame()
 IplImage* CvCaptureCAM_XIMEA::retrieveFrame(int)
 {
     // update cvImage after format has changed
-    if( (int)image.width != width || (int)image.height != height || image.frm != (XI_IMG_FORMAT)format)
-    {
-        cvReleaseImage(&frame);
-        switch( image.frm)
-        {
-        case XI_MONO8  : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 1); break;
-        case XI_MONO16 : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_16U, 1); break;
-        case XI_RGB24  : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 3); break;
-        case XI_RGB32  : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 4); break;
-        default :
-            return frame;
-        }
-        // update global image format
-        format = image.frm;
-        width = image.width;
-        height = image.height;
-    }
+    resetCvImage();
 
     // copy pixel data
     switch( image.frm)
     {
-    case XI_MONO8  : memcpy( frame->imageData, image.bp, image.width*image.height); break;
-    case XI_MONO16 : memcpy( frame->imageData, image.bp, image.width*image.height*sizeof(WORD)); break;
-    case XI_RGB24  : memcpy( frame->imageData, image.bp, image.width*image.height*3); break;
-    case XI_RGB32  : memcpy( frame->imageData, image.bp, image.width*image.height*sizeof(DWORD)); break;
+    case XI_MONO8       :
+    case XI_RAW8        : memcpy( frame->imageData, image.bp, image.width*image.height); break;
+    case XI_MONO16      :
+    case XI_RAW16       : memcpy( frame->imageData, image.bp, image.width*image.height*sizeof(WORD)); break;
+    case XI_RGB24       :
+    case XI_RGB_PLANAR  : memcpy( frame->imageData, image.bp, image.width*image.height*3); break;
+    case XI_RGB32       : memcpy( frame->imageData, image.bp, image.width*image.height*4); break;
     default: break;
     }
     return frame;
@@ -183,6 +194,35 @@ IplImage* CvCaptureCAM_XIMEA::retrieveFrame(int)
 
 /**********************************************************************************/
 
+void CvCaptureCAM_XIMEA::resetCvImage()
+{
+    int width = 0, height = 0, format = 0;
+    xiGetParamInt( hmv, XI_PRM_WIDTH, &width);
+    xiGetParamInt( hmv, XI_PRM_HEIGHT, &height);
+    xiGetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, &format);
+
+    if( (int)image.width != width || (int)image.height != height || image.frm != (XI_IMG_FORMAT)format)
+    {
+        if(frame) cvReleaseImage(&frame);
+        frame = NULL;
+
+        switch( image.frm)
+        {
+        case XI_MONO8       :
+        case XI_RAW8        : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 1); break;
+        case XI_MONO16      :
+        case XI_RAW16       : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_16U, 1); break;
+        case XI_RGB24       :
+        case XI_RGB_PLANAR  : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 3); break;
+        case XI_RGB32       : frame = cvCreateImage(cvSize( image.width, image.height), IPL_DEPTH_8U, 4); break;
+        default :
+            return;
+        }
+    }
+    cvZero(frame);
+}
+/**********************************************************************************/
+
 double CvCaptureCAM_XIMEA::getProperty( int property_id )
 {
     if(hmv == NULL)
@@ -238,20 +278,14 @@ bool CvCaptureCAM_XIMEA::setProperty( int property_id, double value )
     switch(property_id)
     {
     // OCV parameters
-    case CV_CAP_PROP_FRAME_WIDTH  : mvret = xiSetParamInt( hmv, XI_PRM_WIDTH, ival);
-        if(mvret == XI_OK) width = ival;
-        break;
-    case CV_CAP_PROP_FRAME_HEIGHT : mvret = xiSetParamInt( hmv, XI_PRM_HEIGHT, ival);
-        if(mvret == XI_OK) height = ival;
-        break;
+    case CV_CAP_PROP_FRAME_WIDTH  : mvret = xiSetParamInt( hmv, XI_PRM_WIDTH, ival); break;
+    case CV_CAP_PROP_FRAME_HEIGHT : mvret = xiSetParamInt( hmv, XI_PRM_HEIGHT, ival); break;
     case CV_CAP_PROP_FPS          : mvret = xiSetParamFloat( hmv, XI_PRM_FRAMERATE, fval); break;
     case CV_CAP_PROP_GAIN         : mvret = xiSetParamFloat( hmv, XI_PRM_GAIN, fval); break;
     case CV_CAP_PROP_EXPOSURE     : mvret = xiSetParamInt( hmv, XI_PRM_EXPOSURE, ival); break;
     // XIMEA camera properties
     case CV_CAP_PROP_XI_DOWNSAMPLING  : mvret = xiSetParamInt( hmv, XI_PRM_DOWNSAMPLING, ival); break;
-    case CV_CAP_PROP_XI_DATA_FORMAT   : mvret = xiSetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, ival);
-        if(mvret == XI_OK) format = ival;
-        break;
+    case CV_CAP_PROP_XI_DATA_FORMAT   : mvret = xiSetParamInt( hmv, XI_PRM_IMAGE_DATA_FORMAT, ival); break;
     case CV_CAP_PROP_XI_OFFSET_X      : mvret = xiSetParamInt( hmv, XI_PRM_OFFSET_X, ival); break;
     case CV_CAP_PROP_XI_OFFSET_Y      : mvret = xiSetParamInt( hmv, XI_PRM_OFFSET_Y, ival); break;
     case CV_CAP_PROP_XI_TRG_SOURCE    : mvret = xiSetParamInt( hmv, XI_PRM_TRG_SOURCE, ival); break;
@@ -288,7 +322,7 @@ bool CvCaptureCAM_XIMEA::setProperty( int property_id, double value )
 void CvCaptureCAM_XIMEA::errMsg(const char* msg, int errNum)
 {
 #if defined WIN32 || defined _WIN32
-    char buf[512];
+    char buf[512]="";
     sprintf( buf, "%s : %d\n", msg, errNum);
     OutputDebugString(buf);
 #else
@@ -296,4 +330,22 @@ void CvCaptureCAM_XIMEA::errMsg(const char* msg, int errNum)
 #endif
 }
 
-/**********************************************************************************/
\ No newline at end of file
+/**********************************************************************************/
+
+int  CvCaptureCAM_XIMEA::getBpp()
+{
+    switch( image.frm)
+    {
+    case XI_MONO8       :
+    case XI_RAW8        : return 1;
+    case XI_MONO16      :
+    case XI_RAW16       : return 2;
+    case XI_RGB24       :
+    case XI_RGB_PLANAR  : return 3;
+    case XI_RGB32       : return 4;
+    default :
+        return 0;
+    }
+}
+
+/**********************************************************************************/
diff --git a/modules/highgui/src/ffmpeg_codecs.hpp b/modules/highgui/src/ffmpeg_codecs.hpp
index ca7f55c656..c1e116b756 100644
--- a/modules/highgui/src/ffmpeg_codecs.hpp
+++ b/modules/highgui/src/ffmpeg_codecs.hpp
@@ -61,7 +61,11 @@ extern "C" {
 #endif
 
 #ifdef WIN32
-  #include <libavformat/avformat.h>
+#  ifdef __OPENCV_BUILD
+#    define AVUTIL_COMMON_H
+#    define MKBETAG(a,b,c,d) ((d) | ((c) << 8) | ((b) << 16) | ((unsigned)(a) << 24))
+#  endif
+#  include <libavformat/avformat.h>
 #else
 
 // if the header path is not specified explicitly, let's deduce it
diff --git a/modules/highgui/src/files_Qt/Milky/48/1.png b/modules/highgui/src/files_Qt/Milky/48/1.png
index af3dc132c6..69b4dee0a5 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/1.png and b/modules/highgui/src/files_Qt/Milky/48/1.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/10.png b/modules/highgui/src/files_Qt/Milky/48/10.png
index d016260505..34185e1fa1 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/10.png and b/modules/highgui/src/files_Qt/Milky/48/10.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/11.png b/modules/highgui/src/files_Qt/Milky/48/11.png
index ac236491b1..565ad498e9 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/11.png and b/modules/highgui/src/files_Qt/Milky/48/11.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/12.png b/modules/highgui/src/files_Qt/Milky/48/12.png
index e297afe22f..72712fe9e1 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/12.png and b/modules/highgui/src/files_Qt/Milky/48/12.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/13.png b/modules/highgui/src/files_Qt/Milky/48/13.png
index 691e73bf94..6cb5c5dca7 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/13.png and b/modules/highgui/src/files_Qt/Milky/48/13.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/14.png b/modules/highgui/src/files_Qt/Milky/48/14.png
index ede537cef4..8d217be464 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/14.png and b/modules/highgui/src/files_Qt/Milky/48/14.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/16.png b/modules/highgui/src/files_Qt/Milky/48/16.png
index 3e675d4b1b..2e011c7396 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/16.png and b/modules/highgui/src/files_Qt/Milky/48/16.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/17.png b/modules/highgui/src/files_Qt/Milky/48/17.png
index 32ddc3eba8..4a7e5de411 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/17.png and b/modules/highgui/src/files_Qt/Milky/48/17.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/18.png b/modules/highgui/src/files_Qt/Milky/48/18.png
index be8e3e7160..43f5405f5c 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/18.png and b/modules/highgui/src/files_Qt/Milky/48/18.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/19.png b/modules/highgui/src/files_Qt/Milky/48/19.png
index e5145d0523..203510ddd4 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/19.png and b/modules/highgui/src/files_Qt/Milky/48/19.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/2.png b/modules/highgui/src/files_Qt/Milky/48/2.png
index d2ad4a4c39..8f4903eeab 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/2.png and b/modules/highgui/src/files_Qt/Milky/48/2.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/20.png b/modules/highgui/src/files_Qt/Milky/48/20.png
index c53289aaef..1a591ca419 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/20.png and b/modules/highgui/src/files_Qt/Milky/48/20.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/21.png b/modules/highgui/src/files_Qt/Milky/48/21.png
index 2df56ef77e..e65e4acd36 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/21.png and b/modules/highgui/src/files_Qt/Milky/48/21.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/22.png b/modules/highgui/src/files_Qt/Milky/48/22.png
index 36d41cc634..a81aca1913 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/22.png and b/modules/highgui/src/files_Qt/Milky/48/22.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/23.png b/modules/highgui/src/files_Qt/Milky/48/23.png
index eb51b83855..ab9e60cfcb 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/23.png and b/modules/highgui/src/files_Qt/Milky/48/23.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/24.png b/modules/highgui/src/files_Qt/Milky/48/24.png
index b0033cf64a..4e5629cb45 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/24.png and b/modules/highgui/src/files_Qt/Milky/48/24.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/25.png b/modules/highgui/src/files_Qt/Milky/48/25.png
index d41d792950..da93a5962d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/25.png and b/modules/highgui/src/files_Qt/Milky/48/25.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/26.png b/modules/highgui/src/files_Qt/Milky/48/26.png
index 055c496a28..6ba5d6c106 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/26.png and b/modules/highgui/src/files_Qt/Milky/48/26.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/27.png b/modules/highgui/src/files_Qt/Milky/48/27.png
index 34f5f0c8cd..a14e20420b 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/27.png and b/modules/highgui/src/files_Qt/Milky/48/27.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/28.png b/modules/highgui/src/files_Qt/Milky/48/28.png
index 9c94db1aa5..f0df2d35e8 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/28.png and b/modules/highgui/src/files_Qt/Milky/48/28.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/29.png b/modules/highgui/src/files_Qt/Milky/48/29.png
index 9ca7137edf..6d79d929fc 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/29.png and b/modules/highgui/src/files_Qt/Milky/48/29.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/3.png b/modules/highgui/src/files_Qt/Milky/48/3.png
index 5144bbfc32..40d5946680 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/3.png and b/modules/highgui/src/files_Qt/Milky/48/3.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/30.png b/modules/highgui/src/files_Qt/Milky/48/30.png
index db76e78f81..44037a72f4 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/30.png and b/modules/highgui/src/files_Qt/Milky/48/30.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/31.png b/modules/highgui/src/files_Qt/Milky/48/31.png
index e79c0dfda9..b9d421337d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/31.png and b/modules/highgui/src/files_Qt/Milky/48/31.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/32.png b/modules/highgui/src/files_Qt/Milky/48/32.png
index 414eecfded..d72749c50d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/32.png and b/modules/highgui/src/files_Qt/Milky/48/32.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/33.png b/modules/highgui/src/files_Qt/Milky/48/33.png
index 2ec44d312d..85bb86ff02 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/33.png and b/modules/highgui/src/files_Qt/Milky/48/33.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/34.png b/modules/highgui/src/files_Qt/Milky/48/34.png
index 63f3b0465f..fd095ee77b 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/34.png and b/modules/highgui/src/files_Qt/Milky/48/34.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/35.png b/modules/highgui/src/files_Qt/Milky/48/35.png
index 41a041599e..abb64d6127 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/35.png and b/modules/highgui/src/files_Qt/Milky/48/35.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/37.png b/modules/highgui/src/files_Qt/Milky/48/37.png
index 70be99dace..fc9f361ec0 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/37.png and b/modules/highgui/src/files_Qt/Milky/48/37.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/38.png b/modules/highgui/src/files_Qt/Milky/48/38.png
index 1faaa2e59c..81cd7e139b 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/38.png and b/modules/highgui/src/files_Qt/Milky/48/38.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/39.png b/modules/highgui/src/files_Qt/Milky/48/39.png
index b5b7b980f7..d76effcd79 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/39.png and b/modules/highgui/src/files_Qt/Milky/48/39.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/4.png b/modules/highgui/src/files_Qt/Milky/48/4.png
index fe93c3503f..a6a8d07a74 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/4.png and b/modules/highgui/src/files_Qt/Milky/48/4.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/40.png b/modules/highgui/src/files_Qt/Milky/48/40.png
index 1035906078..f17ad6aa10 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/40.png and b/modules/highgui/src/files_Qt/Milky/48/40.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/41.png b/modules/highgui/src/files_Qt/Milky/48/41.png
index 622e1d8b27..4553c04588 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/41.png and b/modules/highgui/src/files_Qt/Milky/48/41.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/42.png b/modules/highgui/src/files_Qt/Milky/48/42.png
index 2c20bf60fd..fb5f9a2a6c 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/42.png and b/modules/highgui/src/files_Qt/Milky/48/42.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/43.png b/modules/highgui/src/files_Qt/Milky/48/43.png
index b849f939b3..3c958420b8 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/43.png and b/modules/highgui/src/files_Qt/Milky/48/43.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/44.png b/modules/highgui/src/files_Qt/Milky/48/44.png
index 3902ba1e3b..ef3c114d45 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/44.png and b/modules/highgui/src/files_Qt/Milky/48/44.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/45.png b/modules/highgui/src/files_Qt/Milky/48/45.png
index cd4d6deebc..a77fb9e14c 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/45.png and b/modules/highgui/src/files_Qt/Milky/48/45.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/5.png b/modules/highgui/src/files_Qt/Milky/48/5.png
index 60827ff7ed..46df26f5e4 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/5.png and b/modules/highgui/src/files_Qt/Milky/48/5.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/6.png b/modules/highgui/src/files_Qt/Milky/48/6.png
index ed04e555d5..72853e6856 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/6.png and b/modules/highgui/src/files_Qt/Milky/48/6.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/7.png b/modules/highgui/src/files_Qt/Milky/48/7.png
index 6eca1fadf5..832772b59d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/7.png and b/modules/highgui/src/files_Qt/Milky/48/7.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/48/9.png b/modules/highgui/src/files_Qt/Milky/48/9.png
index e6a9be1541..4ece823fe3 100644
Binary files a/modules/highgui/src/files_Qt/Milky/48/9.png and b/modules/highgui/src/files_Qt/Milky/48/9.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/1.png b/modules/highgui/src/files_Qt/Milky/64/1.png
index 9222f069f4..36a19f6e0e 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/1.png and b/modules/highgui/src/files_Qt/Milky/64/1.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/10.png b/modules/highgui/src/files_Qt/Milky/64/10.png
index e450e4ae30..28e0be1f5f 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/10.png and b/modules/highgui/src/files_Qt/Milky/64/10.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/100.png b/modules/highgui/src/files_Qt/Milky/64/100.png
index 2920ba3db5..fbcfbaabbb 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/100.png and b/modules/highgui/src/files_Qt/Milky/64/100.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/11.png b/modules/highgui/src/files_Qt/Milky/64/11.png
index 09e42528a1..1002f34573 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/11.png and b/modules/highgui/src/files_Qt/Milky/64/11.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/12.png b/modules/highgui/src/files_Qt/Milky/64/12.png
index 42ddcf2fa3..822bc42e23 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/12.png and b/modules/highgui/src/files_Qt/Milky/64/12.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/126.png b/modules/highgui/src/files_Qt/Milky/64/126.png
index 4d34cfeb15..d8715004ad 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/126.png and b/modules/highgui/src/files_Qt/Milky/64/126.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/13.png b/modules/highgui/src/files_Qt/Milky/64/13.png
index a5d1aff9cc..c4ae0c42fa 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/13.png and b/modules/highgui/src/files_Qt/Milky/64/13.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/14.png b/modules/highgui/src/files_Qt/Milky/64/14.png
index 091cf58478..ce96e04ceb 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/14.png and b/modules/highgui/src/files_Qt/Milky/64/14.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/15.png b/modules/highgui/src/files_Qt/Milky/64/15.png
index f55ea7dd45..eb9dca50a9 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/15.png and b/modules/highgui/src/files_Qt/Milky/64/15.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/16.png b/modules/highgui/src/files_Qt/Milky/64/16.png
index 008823301d..9acf8c20bd 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/16.png and b/modules/highgui/src/files_Qt/Milky/64/16.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/17.png b/modules/highgui/src/files_Qt/Milky/64/17.png
index 7ab1b957fb..18d286ab22 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/17.png and b/modules/highgui/src/files_Qt/Milky/64/17.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/18.png b/modules/highgui/src/files_Qt/Milky/64/18.png
index c0772630f6..3b76256bab 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/18.png and b/modules/highgui/src/files_Qt/Milky/64/18.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/19.png b/modules/highgui/src/files_Qt/Milky/64/19.png
index d167b9acee..b0de5da46b 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/19.png and b/modules/highgui/src/files_Qt/Milky/64/19.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/2.png b/modules/highgui/src/files_Qt/Milky/64/2.png
index f540987679..3523cc6706 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/2.png and b/modules/highgui/src/files_Qt/Milky/64/2.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/21.png b/modules/highgui/src/files_Qt/Milky/64/21.png
index d1ae8e6482..4c94f54422 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/21.png and b/modules/highgui/src/files_Qt/Milky/64/21.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/22.png b/modules/highgui/src/files_Qt/Milky/64/22.png
index ca026ecbcf..f7f0c3d9ec 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/22.png and b/modules/highgui/src/files_Qt/Milky/64/22.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/23.png b/modules/highgui/src/files_Qt/Milky/64/23.png
index a374ea029c..11b0899f5c 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/23.png and b/modules/highgui/src/files_Qt/Milky/64/23.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/24.png b/modules/highgui/src/files_Qt/Milky/64/24.png
index 5566ebd4d6..94a8e94969 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/24.png and b/modules/highgui/src/files_Qt/Milky/64/24.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/25.png b/modules/highgui/src/files_Qt/Milky/64/25.png
index 368f7a1c64..4e650dc817 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/25.png and b/modules/highgui/src/files_Qt/Milky/64/25.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/26.png b/modules/highgui/src/files_Qt/Milky/64/26.png
index 10ecc3be9e..6dec633836 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/26.png and b/modules/highgui/src/files_Qt/Milky/64/26.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/27.png b/modules/highgui/src/files_Qt/Milky/64/27.png
index 9946afb50a..1ab2410c77 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/27.png and b/modules/highgui/src/files_Qt/Milky/64/27.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/28.png b/modules/highgui/src/files_Qt/Milky/64/28.png
index 9094ba190a..7d4d62435f 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/28.png and b/modules/highgui/src/files_Qt/Milky/64/28.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/29.png b/modules/highgui/src/files_Qt/Milky/64/29.png
index dc593dfcdd..74a499650a 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/29.png and b/modules/highgui/src/files_Qt/Milky/64/29.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/30.png b/modules/highgui/src/files_Qt/Milky/64/30.png
index 2430966bea..d1fbb72085 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/30.png and b/modules/highgui/src/files_Qt/Milky/64/30.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/31.png b/modules/highgui/src/files_Qt/Milky/64/31.png
index 1748d71359..70d95908f7 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/31.png and b/modules/highgui/src/files_Qt/Milky/64/31.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/32.png b/modules/highgui/src/files_Qt/Milky/64/32.png
index d225fb3d99..11b66ad441 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/32.png and b/modules/highgui/src/files_Qt/Milky/64/32.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/33.png b/modules/highgui/src/files_Qt/Milky/64/33.png
index 3fb4ac5b15..c76151cc32 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/33.png and b/modules/highgui/src/files_Qt/Milky/64/33.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/34.png b/modules/highgui/src/files_Qt/Milky/64/34.png
index 0dd5f23ef9..1b009aac09 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/34.png and b/modules/highgui/src/files_Qt/Milky/64/34.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/35.png b/modules/highgui/src/files_Qt/Milky/64/35.png
index 5cdb35c97e..c9b408445f 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/35.png and b/modules/highgui/src/files_Qt/Milky/64/35.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/36.png b/modules/highgui/src/files_Qt/Milky/64/36.png
index d2b2444196..d5aad30b69 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/36.png and b/modules/highgui/src/files_Qt/Milky/64/36.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/37.png b/modules/highgui/src/files_Qt/Milky/64/37.png
index ef2b8a56d9..b0898c8cc9 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/37.png and b/modules/highgui/src/files_Qt/Milky/64/37.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/38.png b/modules/highgui/src/files_Qt/Milky/64/38.png
index f1a700c146..0aa9224a6b 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/38.png and b/modules/highgui/src/files_Qt/Milky/64/38.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/39.png b/modules/highgui/src/files_Qt/Milky/64/39.png
index 121a3f8a09..f67ae06b5e 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/39.png and b/modules/highgui/src/files_Qt/Milky/64/39.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/4.png b/modules/highgui/src/files_Qt/Milky/64/4.png
index 264f8bd6a7..eb1a5facd2 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/4.png and b/modules/highgui/src/files_Qt/Milky/64/4.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/40.png b/modules/highgui/src/files_Qt/Milky/64/40.png
index a04765de14..f39b08537f 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/40.png and b/modules/highgui/src/files_Qt/Milky/64/40.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/41.png b/modules/highgui/src/files_Qt/Milky/64/41.png
index 3bed0ebdaa..3061701f56 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/41.png and b/modules/highgui/src/files_Qt/Milky/64/41.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/42.png b/modules/highgui/src/files_Qt/Milky/64/42.png
index c176620993..1b979766a7 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/42.png and b/modules/highgui/src/files_Qt/Milky/64/42.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/43.png b/modules/highgui/src/files_Qt/Milky/64/43.png
index 295d2886d3..f3d9e5d435 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/43.png and b/modules/highgui/src/files_Qt/Milky/64/43.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/44.png b/modules/highgui/src/files_Qt/Milky/64/44.png
index a8fba75456..a549bfe0cb 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/44.png and b/modules/highgui/src/files_Qt/Milky/64/44.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/45.png b/modules/highgui/src/files_Qt/Milky/64/45.png
index d65e2bd935..123fa1a3dd 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/45.png and b/modules/highgui/src/files_Qt/Milky/64/45.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/5.png b/modules/highgui/src/files_Qt/Milky/64/5.png
index 185219d267..9d3c5a3683 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/5.png and b/modules/highgui/src/files_Qt/Milky/64/5.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/6.png b/modules/highgui/src/files_Qt/Milky/64/6.png
index fa55cbc04f..5ae8a8e67d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/6.png and b/modules/highgui/src/files_Qt/Milky/64/6.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/65.png b/modules/highgui/src/files_Qt/Milky/64/65.png
index 46ca28ee1f..f4e9a2880c 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/65.png and b/modules/highgui/src/files_Qt/Milky/64/65.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/7.png b/modules/highgui/src/files_Qt/Milky/64/7.png
index 98083e066c..e97ab37ce8 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/7.png and b/modules/highgui/src/files_Qt/Milky/64/7.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/8.png b/modules/highgui/src/files_Qt/Milky/64/8.png
index 2daaaeffca..639ac9a1f1 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/8.png and b/modules/highgui/src/files_Qt/Milky/64/8.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/9.png b/modules/highgui/src/files_Qt/Milky/64/9.png
index 5eab48e625..cce6c7ef88 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/9.png and b/modules/highgui/src/files_Qt/Milky/64/9.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/94.png b/modules/highgui/src/files_Qt/Milky/64/94.png
index 9be1178893..aca0ff936d 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/94.png and b/modules/highgui/src/files_Qt/Milky/64/94.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/98.png b/modules/highgui/src/files_Qt/Milky/64/98.png
index a2e58c7bb8..0a11c643bb 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/98.png and b/modules/highgui/src/files_Qt/Milky/64/98.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/64/99.png b/modules/highgui/src/files_Qt/Milky/64/99.png
index 5f952b8b54..122cf2de06 100644
Binary files a/modules/highgui/src/files_Qt/Milky/64/99.png and b/modules/highgui/src/files_Qt/Milky/64/99.png differ
diff --git a/modules/highgui/src/files_Qt/Milky/README.txt b/modules/highgui/src/files_Qt/Milky/README.txt
index df2a2e83f4..01ab431457 100644
--- a/modules/highgui/src/files_Qt/Milky/README.txt
+++ b/modules/highgui/src/files_Qt/Milky/README.txt
@@ -16,4 +16,4 @@ The license does not permit the following uses:
 
 You may not use, or allow anyone else to use the icons to create pornographic, libelous, obscene, or defamatory material.
 
-All icon files are provided "as is". You agree not to hold IconEden.com liable for any damages that may occur due to use, or inability to use, icons or image data from IconEden.com.
\ No newline at end of file
+All icon files are provided "as is". You agree not to hold IconEden.com liable for any damages that may occur due to use, or inability to use, icons or image data from IconEden.com.
diff --git a/modules/highgui/src/grfmt_base.cpp b/modules/highgui/src/grfmt_base.cpp
index 0fe798ea8b..267cb31b57 100644
--- a/modules/highgui/src/grfmt_base.cpp
+++ b/modules/highgui/src/grfmt_base.cpp
@@ -54,7 +54,7 @@ BaseImageDecoder::BaseImageDecoder()
     m_buf_supported = false;
 }
 
-bool BaseImageDecoder::setSource( const std::string& filename )
+bool BaseImageDecoder::setSource( const String& filename )
 {
     m_filename = filename;
     m_buf.release();
@@ -65,7 +65,7 @@ bool BaseImageDecoder::setSource( const Mat& buf )
 {
     if( !m_buf_supported )
         return false;
-    m_filename = std::string();
+    m_filename = String();
     m_buf = buf;
     return true;
 }
@@ -75,7 +75,7 @@ size_t BaseImageDecoder::signatureLength() const
     return m_signature.size();
 }
 
-bool BaseImageDecoder::checkSignature( const std::string& signature ) const
+bool BaseImageDecoder::checkSignature( const String& signature ) const
 {
     size_t len = signatureLength();
     return signature.size() >= len && memcmp( signature.c_str(), m_signature.c_str(), len ) == 0;
@@ -96,12 +96,12 @@ bool  BaseImageEncoder::isFormatSupported( int depth ) const
     return depth == CV_8U;
 }
 
-std::string BaseImageEncoder::getDescription() const
+String BaseImageEncoder::getDescription() const
 {
     return m_description;
 }
 
-bool BaseImageEncoder::setDestination( const std::string& filename )
+bool BaseImageEncoder::setDestination( const String& filename )
 {
     m_filename = filename;
     m_buf = 0;
@@ -114,7 +114,7 @@ bool BaseImageEncoder::setDestination( std::vector<uchar>& buf )
         return false;
     m_buf = &buf;
     m_buf->clear();
-    m_filename = std::string();
+    m_filename = String();
     return true;
 }
 
@@ -127,7 +127,7 @@ void BaseImageEncoder::throwOnEror() const
 {
     if(!m_last_error.empty())
     {
-        std::string msg = "Raw image encoder error: " + m_last_error;
+        String msg = "Raw image encoder error: " + m_last_error;
         CV_Error( CV_BadImageSize, msg.c_str() );
     }
 }
diff --git a/modules/highgui/src/grfmt_base.hpp b/modules/highgui/src/grfmt_base.hpp
index a97fa2919d..f7fde90cfe 100644
--- a/modules/highgui/src/grfmt_base.hpp
+++ b/modules/highgui/src/grfmt_base.hpp
@@ -65,21 +65,21 @@ public:
     int height() const { return m_height; };
     virtual int type() const { return m_type; };
 
-    virtual bool setSource( const std::string& filename );
+    virtual bool setSource( const String& filename );
     virtual bool setSource( const Mat& buf );
     virtual bool readHeader() = 0;
     virtual bool readData( Mat& img ) = 0;
 
     virtual size_t signatureLength() const;
-    virtual bool checkSignature( const std::string& signature ) const;
+    virtual bool checkSignature( const String& signature ) const;
     virtual ImageDecoder newDecoder() const;
 
 protected:
     int  m_width;  // width  of the image ( filled by readHeader )
     int  m_height; // height of the image ( filled by readHeader )
     int  m_type;
-    std::string m_filename;
-    std::string m_signature;
+    String m_filename;
+    String m_signature;
     Mat m_buf;
     bool m_buf_supported;
 };
@@ -93,23 +93,23 @@ public:
     virtual ~BaseImageEncoder() {};
     virtual bool isFormatSupported( int depth ) const;
 
-    virtual bool setDestination( const std::string& filename );
+    virtual bool setDestination( const String& filename );
     virtual bool setDestination( std::vector<uchar>& buf );
     virtual bool write( const Mat& img, const std::vector<int>& params ) = 0;
 
-    virtual std::string getDescription() const;
+    virtual String getDescription() const;
     virtual ImageEncoder newEncoder() const;
 
     virtual void throwOnEror() const;
 
 protected:
-    std::string m_description;
+    String m_description;
 
-    std::string m_filename;
+    String m_filename;
     std::vector<uchar>* m_buf;
     bool m_buf_supported;
 
-    std::string m_last_error;
+    String m_last_error;
 };
 
 }
diff --git a/modules/highgui/src/grfmt_bmp.cpp b/modules/highgui/src/grfmt_bmp.cpp
index 5f1083e435..f7147b5fc1 100644
--- a/modules/highgui/src/grfmt_bmp.cpp
+++ b/modules/highgui/src/grfmt_bmp.cpp
@@ -70,7 +70,7 @@ void  BmpDecoder::close()
 
 ImageDecoder BmpDecoder::newDecoder() const
 {
-    return new BmpDecoder;
+    return makePtr<BmpDecoder>();
 }
 
 bool  BmpDecoder::readHeader()
@@ -496,7 +496,7 @@ BmpEncoder::~BmpEncoder()
 
 ImageEncoder BmpEncoder::newEncoder() const
 {
-    return new BmpEncoder;
+    return makePtr<BmpEncoder>();
 }
 
 bool  BmpEncoder::write( const Mat& img, const std::vector<int>& )
@@ -563,4 +563,3 @@ bool  BmpEncoder::write( const Mat& img, const std::vector<int>& )
 }
 
 }
-
diff --git a/modules/highgui/src/grfmt_exr.cpp b/modules/highgui/src/grfmt_exr.cpp
index a537e66fad..079de58340 100644
--- a/modules/highgui/src/grfmt_exr.cpp
+++ b/modules/highgui/src/grfmt_exr.cpp
@@ -348,8 +348,7 @@ bool  ExrDecoder::readData( Mat& img )
                     float *fi = (float *)buffer;
                     for( x = 0; x < m_width * 3; x++)
                     {
-                        int t = cvRound(fi[x]*5);
-                        out[x] = CV_CAST_8U(t);
+                        out[x] = cv::saturate_cast<uchar>(fi[x]*5);
                     }
                 }
                 else
@@ -357,8 +356,7 @@ bool  ExrDecoder::readData( Mat& img )
                     unsigned *ui = (unsigned *)buffer;
                     for( x = 0; x < m_width * 3; x++)
                     {
-                        unsigned t = ui[x];
-                        out[x] = CV_CAST_8U(t);
+                        out[x] = cv::saturate_cast<uchar>(ui[x]);
                     }
                 }
             }
@@ -487,12 +485,9 @@ void  ExrDecoder::ChromaToBGR( float *data, int numlines, int step )
 
             if( !m_native_depth )
             {
-                int t = cvRound(b);
-                ((uchar *)data)[y * step + x * 3] = CV_CAST_8U(t);
-                t = cvRound(Y);
-                ((uchar *)data)[y * step + x * 3 + 1] = CV_CAST_8U(t);
-                t = cvRound(r);
-                ((uchar *)data)[y * step + x * 3 + 2] = CV_CAST_8U(t);
+                ((uchar *)data)[y * step + x * 3 + 0] = cv::saturate_cast<uchar>(b);
+                ((uchar *)data)[y * step + x * 3 + 1] = cv::saturate_cast<uchar>(Y);
+                ((uchar *)data)[y * step + x * 3 + 2] = cv::saturate_cast<uchar>(r);
             }
             else if( m_type == FLOAT )
             {
@@ -503,11 +498,11 @@ void  ExrDecoder::ChromaToBGR( float *data, int numlines, int step )
             else
             {
                 int t = cvRound(b);
-                ((unsigned *)data)[y * step + x * 3] = (unsigned)MAX(t,0);
+                ((unsigned *)data)[y * step + x * 3 + 0] = (unsigned)MAX(t, 0);
                 t = cvRound(Y);
-                ((unsigned *)data)[y * step + x * 3 + 1] = (unsigned)MAX(t,0);
+                ((unsigned *)data)[y * step + x * 3 + 1] = (unsigned)MAX(t, 0);
                 t = cvRound(r);
-                ((unsigned *)data)[y * step + x * 3 + 2] = (unsigned)MAX(t,0);
+                ((unsigned *)data)[y * step + x * 3 + 2] = (unsigned)MAX(t, 0);
             }
         }
     }
@@ -556,7 +551,7 @@ void  ExrDecoder::RGBToGray( float *in, float *out )
 
 ImageDecoder ExrDecoder::newDecoder() const
 {
-    return new ExrDecoder;
+    return makePtr<ExrDecoder>();
 }
 
 /////////////////////// ExrEncoder ///////////////////
@@ -731,7 +726,7 @@ bool  ExrEncoder::write( const Mat& img, const std::vector<int>& )
 
 ImageEncoder ExrEncoder::newEncoder() const
 {
-    return new ExrEncoder;
+    return makePtr<ExrEncoder>();
 }
 
 }
diff --git a/modules/highgui/src/grfmt_jpeg.cpp b/modules/highgui/src/grfmt_jpeg.cpp
index 7054d6dbc1..28c52e8598 100644
--- a/modules/highgui/src/grfmt_jpeg.cpp
+++ b/modules/highgui/src/grfmt_jpeg.cpp
@@ -52,6 +52,11 @@
 #include <stdio.h>
 #include <setjmp.h>
 
+// the following defines are a hack to avoid multiple problems with frame ponter handling and setjmp
+// see http://gcc.gnu.org/ml/gcc/2011-10/msg00324.html for some details
+#define mingw_getsp(...) 0
+#define __builtin_frame_address(...) 0
+
 #ifdef WIN32
 
 #define XMD_H // prevent redefinition of INT32
@@ -203,7 +208,7 @@ void  JpegDecoder::close()
 
 ImageDecoder JpegDecoder::newDecoder() const
 {
-    return new JpegDecoder;
+    return makePtr<JpegDecoder>();
 }
 
 bool  JpegDecoder::readHeader()
@@ -534,7 +539,7 @@ JpegEncoder::~JpegEncoder()
 
 ImageEncoder JpegEncoder::newEncoder() const
 {
-    return new JpegEncoder;
+    return makePtr<JpegEncoder>();
 }
 
 bool JpegEncoder::write( const Mat& img, const std::vector<int>& params )
diff --git a/modules/highgui/src/grfmt_jpeg2000.cpp b/modules/highgui/src/grfmt_jpeg2000.cpp
index e09eccc906..d711846ce8 100644
--- a/modules/highgui/src/grfmt_jpeg2000.cpp
+++ b/modules/highgui/src/grfmt_jpeg2000.cpp
@@ -53,12 +53,6 @@
 #endif
 #endif
 
-#undef PACKAGE
-#undef PACKAGE_BUGREPORT
-#undef PACKAGE_NAME
-#undef PACKAGE_STRING
-#undef PACKAGE_TARNAME
-#undef PACKAGE_VERSION
 #undef VERSION
 
 #include <jasper/jasper.h>
@@ -82,7 +76,7 @@ static JasperInitializer initialize_jasper;
 
 Jpeg2KDecoder::Jpeg2KDecoder()
 {
-    m_signature = '\0' + std::string() + '\0' + std::string() + '\0' + std::string("\x0cjP  \r\n\x87\n");
+    m_signature = '\0' + String() + '\0' + String() + '\0' + String("\x0cjP  \r\n\x87\n");
     m_stream = 0;
     m_image = 0;
 }
@@ -94,7 +88,7 @@ Jpeg2KDecoder::~Jpeg2KDecoder()
 
 ImageDecoder Jpeg2KDecoder::newDecoder() const
 {
-    return new Jpeg2KDecoder;
+    return makePtr<Jpeg2KDecoder>();
 }
 
 void  Jpeg2KDecoder::close()
@@ -297,26 +291,26 @@ bool  Jpeg2KDecoder::readComponent8u( uchar *data, void *_buffer,
                 for( x = 0; x < xend - xstart; x++ )
                 {
                     int pix = pix_row[x];
-                    dst[x*ncmpts] = CV_CAST_8U(pix);
+                    dst[x*ncmpts] = cv::saturate_cast<uchar>(pix);
                 }
             else
                 for( x = 0; x < xend - xstart; x++ )
                 {
                     int pix = ((pix_row[x] + delta) >> rshift) << lshift;
-                    dst[x*ncmpts] = CV_CAST_8U(pix);
+                    dst[x*ncmpts] = cv::saturate_cast<uchar>(pix);
                 }
         }
         else if( xstep == 2 && offset == 0 )
             for( x = 0, j = 0; x < xend - xstart; x += 2, j++ )
             {
                 int pix = ((pix_row[j] + delta) >> rshift) << lshift;
-                dst[x*ncmpts] = dst[(x+1)*ncmpts] = CV_CAST_8U(pix);
+                dst[x*ncmpts] = dst[(x+1)*ncmpts] = cv::saturate_cast<uchar>(pix);
             }
         else
             for( x = 0, j = 0; x < xend - xstart; j++ )
             {
                 int pix = ((pix_row[j] + delta) >> rshift) << lshift;
-                pix = CV_CAST_8U(pix);
+                pix = cv::saturate_cast<uchar>(pix);
                 for( x1 = x + xstep; x < x1; x++ )
                     dst[x*ncmpts] = (uchar)pix;
             }
@@ -361,26 +355,26 @@ bool  Jpeg2KDecoder::readComponent16u( unsigned short *data, void *_buffer,
                 for( x = 0; x < xend - xstart; x++ )
                 {
                     int pix = pix_row[x];
-                    dst[x*ncmpts] = CV_CAST_16U(pix);
+                    dst[x*ncmpts] = cv::saturate_cast<ushort>(pix);
                 }
             else
                 for( x = 0; x < xend - xstart; x++ )
                 {
                     int pix = ((pix_row[x] + delta) >> rshift) << lshift;
-                    dst[x*ncmpts] = CV_CAST_16U(pix);
+                    dst[x*ncmpts] = cv::saturate_cast<ushort>(pix);
                 }
         }
         else if( xstep == 2 && offset == 0 )
             for( x = 0, j = 0; x < xend - xstart; x += 2, j++ )
             {
                 int pix = ((pix_row[j] + delta) >> rshift) << lshift;
-                dst[x*ncmpts] = dst[(x+1)*ncmpts] = CV_CAST_16U(pix);
+                dst[x*ncmpts] = dst[(x+1)*ncmpts] = cv::saturate_cast<ushort>(pix);
             }
         else
             for( x = 0, j = 0; x < xend - xstart; j++ )
             {
                 int pix = ((pix_row[j] + delta) >> rshift) << lshift;
-                pix = CV_CAST_16U(pix);
+                pix = cv::saturate_cast<ushort>(pix);
                 for( x1 = x + xstep; x < x1; x++ )
                     dst[x*ncmpts] = (ushort)pix;
             }
@@ -409,7 +403,7 @@ Jpeg2KEncoder::~Jpeg2KEncoder()
 
 ImageEncoder Jpeg2KEncoder::newEncoder() const
 {
-    return new Jpeg2KEncoder;
+    return makePtr<Jpeg2KEncoder>();
 }
 
 bool  Jpeg2KEncoder::isFormatSupported( int depth ) const
diff --git a/modules/highgui/src/grfmt_png.cpp b/modules/highgui/src/grfmt_png.cpp
index 77b317f0f9..4d5c7793ea 100644
--- a/modules/highgui/src/grfmt_png.cpp
+++ b/modules/highgui/src/grfmt_png.cpp
@@ -51,7 +51,6 @@
     and png2bmp sample from libpng distribution (Copyright (C) 1999-2001 MIYASAKA Masaru)
 \****************************************************************************************/
 
-#undef HAVE_UNISTD_H //to avoid redefinition
 #ifndef _LFS64_LARGEFILE
 #  define _LFS64_LARGEFILE 0
 #endif
@@ -73,6 +72,11 @@
     #pragma warning( disable: 4611 )
 #endif
 
+// the following defines are a hack to avoid multiple problems with frame ponter handling and setjmp
+// see http://gcc.gnu.org/ml/gcc/2011-10/msg00324.html for some details
+#define mingw_getsp(...) 0
+#define __builtin_frame_address(...) 0
+
 namespace cv
 {
 
@@ -97,7 +101,7 @@ PngDecoder::~PngDecoder()
 
 ImageDecoder PngDecoder::newDecoder() const
 {
-    return new PngDecoder;
+    return makePtr<PngDecoder>();
 }
 
 void  PngDecoder::close()
@@ -167,7 +171,9 @@ bool  PngDecoder::readHeader()
                 if( !m_buf.empty() || m_f )
                 {
                     png_uint_32 wdth, hght;
-                    int bit_depth, color_type;
+                    int bit_depth, color_type, num_trans=0;
+                    png_bytep trans;
+                    png_color_16p trans_values;
 
                     png_read_info( png_ptr, info_ptr );
 
@@ -183,15 +189,22 @@ bool  PngDecoder::readHeader()
                     {
                         switch(color_type)
                         {
-                           case PNG_COLOR_TYPE_RGB:
-                           case PNG_COLOR_TYPE_PALETTE:
-                               m_type = CV_8UC3;
-                               break;
-                          case PNG_COLOR_TYPE_RGB_ALPHA:
-                               m_type = CV_8UC4;
-                               break;
-                          default:
-                               m_type = CV_8UC1;
+                            case PNG_COLOR_TYPE_RGB:
+                                m_type = CV_8UC3;
+                                break;
+                            case PNG_COLOR_TYPE_PALETTE:
+                                png_get_tRNS( png_ptr, info_ptr, &trans, &num_trans, &trans_values);
+                                //Check if there is a transparency value in the palette
+                                if ( num_trans > 0 )
+                                    m_type = CV_8UC4;
+                                else
+                                    m_type = CV_8UC3;
+                                break;
+                            case PNG_COLOR_TYPE_RGB_ALPHA:
+                                m_type = CV_8UC4;
+                                break;
+                            default:
+                                m_type = CV_8UC1;
                         }
                         if( bit_depth == 16 )
                             m_type = CV_MAKETYPE(CV_16U, CV_MAT_CN(m_type));
@@ -264,6 +277,7 @@ bool  PngDecoder::readData( Mat& img )
             else
                 png_set_rgb_to_gray( png_ptr, 1, 0.299, 0.587 ); // RGB->Gray
 
+            png_set_interlace_handling( png_ptr );
             png_read_update_info( png_ptr, info_ptr );
 
             for( y = 0; y < m_height; y++ )
@@ -303,7 +317,7 @@ bool  PngEncoder::isFormatSupported( int depth ) const
 
 ImageEncoder PngEncoder::newEncoder() const
 {
-    return new PngEncoder;
+    return makePtr<PngEncoder>();
 }
 
 
diff --git a/modules/highgui/src/grfmt_pxm.cpp b/modules/highgui/src/grfmt_pxm.cpp
index c2eb100d7b..425a296cb9 100644
--- a/modules/highgui/src/grfmt_pxm.cpp
+++ b/modules/highgui/src/grfmt_pxm.cpp
@@ -107,7 +107,7 @@ size_t PxMDecoder::signatureLength() const
     return 3;
 }
 
-bool PxMDecoder::checkSignature( const std::string& signature ) const
+bool PxMDecoder::checkSignature( const String& signature ) const
 {
     return signature.size() >= 3 && signature[0] == 'P' &&
            '1' <= signature[1] && signature[1] <= '6' &&
@@ -116,7 +116,7 @@ bool PxMDecoder::checkSignature( const std::string& signature ) const
 
 ImageDecoder PxMDecoder::newDecoder() const
 {
-    return new PxMDecoder;
+    return makePtr<PxMDecoder>();
 }
 
 void  PxMDecoder::close()
@@ -357,7 +357,7 @@ PxMEncoder::~PxMEncoder()
 
 ImageEncoder  PxMEncoder::newEncoder() const
 {
-    return new PxMEncoder;
+    return makePtr<PxMEncoder>();
 }
 
 
diff --git a/modules/highgui/src/grfmt_pxm.hpp b/modules/highgui/src/grfmt_pxm.hpp
index 59c6694149..5460e11002 100644
--- a/modules/highgui/src/grfmt_pxm.hpp
+++ b/modules/highgui/src/grfmt_pxm.hpp
@@ -61,7 +61,7 @@ public:
     void  close();
 
     size_t signatureLength() const;
-    bool checkSignature( const std::string& signature ) const;
+    bool checkSignature( const String& signature ) const;
     ImageDecoder newDecoder() const;
 
 protected:
diff --git a/modules/highgui/src/grfmt_sunras.cpp b/modules/highgui/src/grfmt_sunras.cpp
index a16e5d8004..b67400f337 100644
--- a/modules/highgui/src/grfmt_sunras.cpp
+++ b/modules/highgui/src/grfmt_sunras.cpp
@@ -63,7 +63,7 @@ SunRasterDecoder::~SunRasterDecoder()
 
 ImageDecoder SunRasterDecoder::newDecoder() const
 {
-    return new SunRasterDecoder;
+    return makePtr<SunRasterDecoder>();
 }
 
 void  SunRasterDecoder::close()
@@ -388,7 +388,7 @@ SunRasterEncoder::SunRasterEncoder()
 
 ImageEncoder SunRasterEncoder::newEncoder() const
 {
-    return new SunRasterEncoder;
+    return makePtr<SunRasterEncoder>();
 }
 
 SunRasterEncoder::~SunRasterEncoder()
diff --git a/modules/highgui/src/grfmt_tiff.cpp b/modules/highgui/src/grfmt_tiff.cpp
index 35ab2f81a9..d8937db251 100644
--- a/modules/highgui/src/grfmt_tiff.cpp
+++ b/modules/highgui/src/grfmt_tiff.cpp
@@ -94,7 +94,7 @@ size_t TiffDecoder::signatureLength() const
     return 4;
 }
 
-bool TiffDecoder::checkSignature( const std::string& signature ) const
+bool TiffDecoder::checkSignature( const String& signature ) const
 {
     return signature.size() >= 4 &&
         (memcmp(signature.c_str(), fmtSignTiffII, 4) == 0 ||
@@ -108,7 +108,7 @@ int TiffDecoder::normalizeChannelsNumber(int channels) const
 
 ImageDecoder TiffDecoder::newDecoder() const
 {
-    return new TiffDecoder;
+    return makePtr<TiffDecoder>();
 }
 
 bool TiffDecoder::readHeader()
@@ -400,7 +400,7 @@ TiffEncoder::~TiffEncoder()
 
 ImageEncoder TiffEncoder::newEncoder() const
 {
-    return new TiffEncoder;
+    return makePtr<TiffEncoder>();
 }
 
 bool TiffEncoder::isFormatSupported( int depth ) const
diff --git a/modules/highgui/src/grfmt_tiff.hpp b/modules/highgui/src/grfmt_tiff.hpp
index 3fa0ab4976..d3745a9f8c 100644
--- a/modules/highgui/src/grfmt_tiff.hpp
+++ b/modules/highgui/src/grfmt_tiff.hpp
@@ -102,7 +102,7 @@ public:
     void  close();
 
     size_t signatureLength() const;
-    bool checkSignature( const std::string& signature ) const;
+    bool checkSignature( const String& signature ) const;
     ImageDecoder newDecoder() const;
 
 protected:
diff --git a/modules/highgui/src/grfmt_webp.cpp b/modules/highgui/src/grfmt_webp.cpp
index 981e7a5eb6..1c64adec60 100644
--- a/modules/highgui/src/grfmt_webp.cpp
+++ b/modules/highgui/src/grfmt_webp.cpp
@@ -46,6 +46,7 @@
 
 #include <webp/decode.h>
 #include <webp/encode.h>
+
 #include <stdio.h>
 #include <limits.h>
 
@@ -53,36 +54,43 @@
 
 #include "opencv2/imgproc.hpp"
 
+const size_t WEBP_HEADER_SIZE = 32;
+
 namespace cv
 {
 
 WebPDecoder::WebPDecoder()
 {
-    m_signature = "RIFF....WEBPVP8 ";
     m_buf_supported = true;
 }
 
-WebPDecoder::~WebPDecoder()
+WebPDecoder::~WebPDecoder() {}
+
+size_t WebPDecoder::signatureLength() const
 {
+    return WEBP_HEADER_SIZE;
+}
+
+bool WebPDecoder::checkSignature(const String & signature) const
+{
+    bool ret = false;
+
+    if(signature.size() >= WEBP_HEADER_SIZE)
+    {
+        WebPBitstreamFeatures features;
+        if(VP8_STATUS_OK == WebPGetFeatures((uint8_t *)signature.c_str(),
+                                            WEBP_HEADER_SIZE, &features))
+        {
+            ret = true;
+        }
+    }
+
+    return ret;
 }
 
 ImageDecoder WebPDecoder::newDecoder() const
 {
-    return new WebPDecoder;
-}
-
-bool WebPDecoder::checkSignature( const std::string& signature ) const
-{
-    size_t len = signatureLength();
-    bool ret = false;
-
-    if(signature.size() >= len)
-    {
-        ret = ( (memcmp(signature.c_str(), m_signature.c_str(), 4) == 0) &&
-            (memcmp(signature.c_str() + 8, m_signature.c_str() + 8, 4) == 0) );
-    }
-
-    return ret;
+    return makePtr<WebPDecoder>();
 }
 
 bool WebPDecoder::readHeader()
@@ -99,16 +107,16 @@ bool WebPDecoder::readHeader()
         }
 
         fseek(wfile, 0, SEEK_END);
-        size_t wfile_size = ftell(wfile);
+        long int wfile_size = ftell(wfile);
         fseek(wfile, 0, SEEK_SET);
 
-        if(wfile_size > (size_t)INT_MAX)
+        if(wfile_size > static_cast<long int>(INT_MAX))
         {
             fclose(wfile);
             return false;
         }
 
-        data.create(1, (int)wfile_size, CV_8U);
+        data.create(1, wfile_size, CV_8U);
 
         size_t data_size = fread(data.data, 1, wfile_size, wfile);
 
@@ -117,7 +125,7 @@ bool WebPDecoder::readHeader()
             fclose(wfile);
         }
 
-        if( data_size < wfile_size )
+        if(static_cast<long int>(data_size) != wfile_size)
         {
             return false;
         }
@@ -127,9 +135,23 @@ bool WebPDecoder::readHeader()
         data = m_buf;
     }
 
-    if(WebPGetInfo(data.data, data.total(), &m_width, &m_height) == 1)
+    WebPBitstreamFeatures features;
+    if(VP8_STATUS_OK == WebPGetFeatures(data.data, WEBP_HEADER_SIZE, &features))
     {
-        m_type = CV_8UC3;
+        m_width  = features.width;
+        m_height = features.height;
+
+        if (features.has_alpha)
+        {
+            m_type = CV_8UC4;
+            channels = 4;
+        }
+        else
+        {
+            m_type = CV_8UC3;
+            channels = 3;
+        }
+
         return true;
     }
 
@@ -140,10 +162,25 @@ bool WebPDecoder::readData(Mat &img)
 {
     if( m_width > 0 && m_height > 0 )
     {
-        uchar* out_data = img.data;
-        unsigned int out_data_size = m_width * m_height * 3 * sizeof(uchar);
+        if (img.cols != m_width || img.rows != m_height || img.type() != m_type)
+        {
+            img.create(m_height, m_width, m_type);
+        }
 
-        uchar *res_ptr = WebPDecodeBGRInto(data.data, data.total(), out_data, out_data_size, m_width * 3);
+        uchar* out_data = img.data;
+        size_t out_data_size = img.cols * img.rows * img.elemSize();
+
+        uchar *res_ptr = 0;
+        if (channels == 3)
+        {
+            res_ptr = WebPDecodeBGRInto(data.data, data.total(), out_data,
+                                        out_data_size, img.step);
+        }
+        else if (channels == 4)
+        {
+            res_ptr = WebPDecodeBGRAInto(data.data, data.total(), out_data,
+                                         out_data_size, img.step);
+        }
 
         if(res_ptr == out_data)
         {
@@ -160,13 +197,11 @@ WebPEncoder::WebPEncoder()
     m_buf_supported = true;
 }
 
-WebPEncoder::~WebPEncoder()
-{
-}
+WebPEncoder::~WebPEncoder() { }
 
 ImageEncoder WebPEncoder::newEncoder() const
 {
-    return new WebPEncoder();
+    return makePtr<WebPEncoder>();
 }
 
 bool WebPEncoder::write(const Mat& img, const std::vector<int>& params)
@@ -179,19 +214,19 @@ bool WebPEncoder::write(const Mat& img, const std::vector<int>& params)
     size_t size = 0;
 
     bool comp_lossless = true;
-    int quality = 100;
+    float quality = 100.0f;
 
     if (params.size() > 1)
     {
         if (params[0] == CV_IMWRITE_WEBP_QUALITY)
         {
             comp_lossless = false;
-            quality = params[1];
-            if (quality < 1)
+            quality = static_cast<float>(params[1]);
+            if (quality < 1.0f)
             {
-                quality = 1;
+                quality = 1.0f;
             }
-            if (quality > 100)
+            if (quality > 100.0f)
             {
                 comp_lossless = true;
             }
@@ -211,14 +246,32 @@ bool WebPEncoder::write(const Mat& img, const std::vector<int>& params)
         image = &temp;
         channels = 3;
     }
+    else if (channels == 2)
+    {
+        return false;
+    }
 
     if (comp_lossless)
     {
-        size = WebPEncodeLosslessBGR(image->data, width, height, ((width * 3 + 3) & ~3), &out);
+        if(channels == 3)
+        {
+            size = WebPEncodeLosslessBGR(image->data, width, height, image->step, &out);
+        }
+        else if(channels == 4)
+        {
+            size = WebPEncodeLosslessBGRA(image->data, width, height, image->step, &out);
+        }
     }
     else
     {
-        size = WebPEncodeBGR(image->data, width, height, ((width * 3 + 3) & ~3), (float)quality, &out);
+        if(channels == 3)
+        {
+            size = WebPEncodeBGR(image->data, width, height, image->step, quality, &out);
+        }
+        else if(channels == 4)
+        {
+            size = WebPEncodeBGRA(image->data, width, height, image->step, quality, &out);
+        }
     }
 
     if(size > 0)
diff --git a/modules/highgui/src/grfmt_webp.hpp b/modules/highgui/src/grfmt_webp.hpp
index 15033d120a..ea692bf8d2 100644
--- a/modules/highgui/src/grfmt_webp.hpp
+++ b/modules/highgui/src/grfmt_webp.hpp
@@ -47,6 +47,8 @@
 
 #ifdef HAVE_WEBP
 
+
+
 namespace cv
 {
 
@@ -60,12 +62,15 @@ public:
     bool readData( Mat& img );
     bool readHeader();
     void close();
-    bool checkSignature( const std::string& signature ) const;
+
+    size_t signatureLength() const;
+    bool checkSignature( const String& signature) const;
 
     ImageDecoder newDecoder() const;
 
 protected:
     Mat data;
+    int channels;
 };
 
 class WebPEncoder : public BaseImageEncoder
diff --git a/modules/highgui/src/ios_conversions.mm b/modules/highgui/src/ios_conversions.mm
new file mode 100644
index 0000000000..fa6208a17f
--- /dev/null
+++ b/modules/highgui/src/ios_conversions.mm
@@ -0,0 +1,117 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#import "opencv2/highgui/cap_ios.h"
+#include "precomp.hpp"
+
+UIImage* MatToUIImage(const cv::Mat& image) {
+
+    NSData *data = [NSData dataWithBytes:image.data
+                                  length:image.elemSize()*image.total()];
+
+    CGColorSpaceRef colorSpace;
+
+    if (image.elemSize() == 1) {
+        colorSpace = CGColorSpaceCreateDeviceGray();
+    } else {
+        colorSpace = CGColorSpaceCreateDeviceRGB();
+    }
+
+    CGDataProviderRef provider =
+            CGDataProviderCreateWithCFData((__bridge CFDataRef)data);
+
+    // Creating CGImage from cv::Mat
+    CGImageRef imageRef = CGImageCreate(image.cols,
+                                        image.rows,
+                                        8,
+                                        8 * image.elemSize(),
+                                        image.step.p[0],
+                                        colorSpace,
+                                        kCGImageAlphaNone|
+                                        kCGBitmapByteOrderDefault,
+                                        provider,
+                                        NULL,
+                                        false,
+                                        kCGRenderingIntentDefault
+                                        );
+
+
+    // Getting UIImage from CGImage
+    UIImage *finalImage = [UIImage imageWithCGImage:imageRef];
+    CGImageRelease(imageRef);
+    CGDataProviderRelease(provider);
+    CGColorSpaceRelease(colorSpace);
+
+    return finalImage;
+}
+
+void UIImageToMat(const UIImage* image,
+                         cv::Mat& m, bool alphaExist) {
+    CGColorSpaceRef colorSpace = CGImageGetColorSpace(image.CGImage);
+    CGFloat cols = image.size.width, rows = image.size.height;
+    CGContextRef contextRef;
+    CGBitmapInfo bitmapInfo = kCGImageAlphaPremultipliedLast;
+    if (CGColorSpaceGetModel(colorSpace) == 0)
+    {
+        m.create(rows, cols, CV_8UC1); // 8 bits per component, 1 channel
+        bitmapInfo = kCGImageAlphaNone;
+        if (!alphaExist)
+            bitmapInfo = kCGImageAlphaNone;
+        contextRef = CGBitmapContextCreate(m.data, m.cols, m.rows, 8,
+                                           m.step[0], colorSpace,
+                                           bitmapInfo);
+    }
+    else
+    {
+        m.create(rows, cols, CV_8UC4); // 8 bits per component, 4 channels
+        if (!alphaExist)
+            bitmapInfo = kCGImageAlphaNoneSkipLast |
+                                kCGBitmapByteOrderDefault;
+        contextRef = CGBitmapContextCreate(m.data, m.cols, m.rows, 8,
+                                           m.step[0], colorSpace,
+                                           bitmapInfo);
+    }
+    CGContextDrawImage(contextRef, CGRectMake(0, 0, cols, rows),
+                       image.CGImage);
+    CGContextRelease(contextRef);
+    CGColorSpaceRelease(colorSpace);
+}
diff --git a/modules/highgui/src/loadsave.cpp b/modules/highgui/src/loadsave.cpp
index a27f3399d0..58a0a6fbf7 100644
--- a/modules/highgui/src/loadsave.cpp
+++ b/modules/highgui/src/loadsave.cpp
@@ -58,35 +58,35 @@ struct ImageCodecInitializer
 {
     ImageCodecInitializer()
     {
-        decoders.push_back( new BmpDecoder );
-        encoders.push_back( new BmpEncoder );
+        decoders.push_back( makePtr<BmpDecoder>() );
+        encoders.push_back( makePtr<BmpEncoder>() );
     #ifdef HAVE_JPEG
-        decoders.push_back( new JpegDecoder );
-        encoders.push_back( new JpegEncoder );
+        decoders.push_back( makePtr<JpegDecoder>() );
+        encoders.push_back( makePtr<JpegEncoder>() );
     #endif
     #ifdef HAVE_WEBP
-        decoders.push_back( new WebPDecoder );
-        encoders.push_back( new WebPEncoder );
+        decoders.push_back( makePtr<WebPDecoder>() );
+        encoders.push_back( makePtr<WebPEncoder>() );
     #endif
-        decoders.push_back( new SunRasterDecoder );
-        encoders.push_back( new SunRasterEncoder );
-        decoders.push_back( new PxMDecoder );
-        encoders.push_back( new PxMEncoder );
+        decoders.push_back( makePtr<SunRasterDecoder>() );
+        encoders.push_back( makePtr<SunRasterEncoder>() );
+        decoders.push_back( makePtr<PxMDecoder>() );
+        encoders.push_back( makePtr<PxMEncoder>() );
     #ifdef HAVE_TIFF
-        decoders.push_back( new TiffDecoder );
+        decoders.push_back( makePtr<TiffDecoder>() );
     #endif
-        encoders.push_back( new TiffEncoder );
+        encoders.push_back( makePtr<TiffEncoder>() );
     #ifdef HAVE_PNG
-        decoders.push_back( new PngDecoder );
-        encoders.push_back( new PngEncoder );
+        decoders.push_back( makePtr<PngDecoder>() );
+        encoders.push_back( makePtr<PngEncoder>() );
     #endif
     #ifdef HAVE_JASPER
-        decoders.push_back( new Jpeg2KDecoder );
-        encoders.push_back( new Jpeg2KEncoder );
+        decoders.push_back( makePtr<Jpeg2KDecoder>() );
+        encoders.push_back( makePtr<Jpeg2KEncoder>() );
     #endif
     #ifdef HAVE_OPENEXR
-        decoders.push_back( new ExrDecoder );
-        encoders.push_back( new ExrEncoder );
+        decoders.push_back( makePtr<ExrDecoder>() );
+        encoders.push_back( makePtr<ExrEncoder>() );
     #endif
     }
 
@@ -96,7 +96,7 @@ struct ImageCodecInitializer
 
 static ImageCodecInitializer codecs;
 
-static ImageDecoder findDecoder( const std::string& filename )
+static ImageDecoder findDecoder( const String& filename )
 {
     size_t i, maxlen = 0;
     for( i = 0; i < codecs.decoders.size(); i++ )
@@ -108,8 +108,8 @@ static ImageDecoder findDecoder( const std::string& filename )
     FILE* f= fopen( filename.c_str(), "rb" );
     if( !f )
         return ImageDecoder();
-    std::string signature(maxlen, ' ');
-    maxlen = fread( &signature[0], 1, maxlen, f );
+    String signature(maxlen, ' ');
+    maxlen = fread( (void*)signature.c_str(), 1, maxlen, f );
     fclose(f);
     signature = signature.substr(0, maxlen);
 
@@ -137,8 +137,8 @@ static ImageDecoder findDecoder( const Mat& buf )
 
     size_t bufSize = buf.rows*buf.cols*buf.elemSize();
     maxlen = std::min(maxlen, bufSize);
-    std::string signature(maxlen, ' ');
-    memcpy( &signature[0], buf.data, maxlen );
+    String signature(maxlen, ' ');
+    memcpy( (void*)signature.c_str(), buf.data, maxlen );
 
     for( i = 0; i < codecs.decoders.size(); i++ )
     {
@@ -149,7 +149,7 @@ static ImageDecoder findDecoder( const Mat& buf )
     return ImageDecoder();
 }
 
-static ImageEncoder findEncoder( const std::string& _ext )
+static ImageEncoder findEncoder( const String& _ext )
 {
     if( _ext.size() <= 1 )
         return ImageEncoder();
@@ -163,7 +163,7 @@ static ImageEncoder findEncoder( const std::string& _ext )
 
     for( size_t i = 0; i < codecs.encoders.size(); i++ )
     {
-        std::string description = codecs.encoders[i]->getDescription();
+        String description = codecs.encoders[i]->getDescription();
         const char* descr = strchr( description.c_str(), '(' );
 
         while( descr )
@@ -191,14 +191,14 @@ static ImageEncoder findEncoder( const std::string& _ext )
 enum { LOAD_CVMAT=0, LOAD_IMAGE=1, LOAD_MAT=2 };
 
 static void*
-imread_( const std::string& filename, int flags, int hdrtype, Mat* mat=0 )
+imread_( const String& filename, int flags, int hdrtype, Mat* mat=0 )
 {
     IplImage* image = 0;
     CvMat *matrix = 0;
     Mat temp, *data = &temp;
 
     ImageDecoder decoder = findDecoder(filename);
-    if( decoder.empty() )
+    if( !decoder )
         return 0;
     decoder->setSource(filename);
     if( !decoder->readHeader() )
@@ -253,14 +253,14 @@ imread_( const std::string& filename, int flags, int hdrtype, Mat* mat=0 )
         hdrtype == LOAD_IMAGE ? (void*)image : (void*)mat;
 }
 
-Mat imread( const std::string& filename, int flags )
+Mat imread( const String& filename, int flags )
 {
     Mat img;
     imread_( filename, flags, LOAD_MAT, &img );
     return img;
 }
 
-static bool imwrite_( const std::string& filename, const Mat& image,
+static bool imwrite_( const String& filename, const Mat& image,
                       const std::vector<int>& params, bool flipv )
 {
     Mat temp;
@@ -269,7 +269,7 @@ static bool imwrite_( const std::string& filename, const Mat& image,
     CV_Assert( image.channels() == 1 || image.channels() == 3 || image.channels() == 4 );
 
     ImageEncoder encoder = findEncoder( filename );
-    if( encoder.empty() )
+    if( !encoder )
         CV_Error( CV_StsError, "could not find a writer for the specified extension" );
 
     if( !encoder->isFormatSupported(image.depth()) )
@@ -292,7 +292,7 @@ static bool imwrite_( const std::string& filename, const Mat& image,
     return code;
 }
 
-bool imwrite( const std::string& filename, InputArray _img,
+bool imwrite( const String& filename, InputArray _img,
               const std::vector<int>& params )
 {
     Mat img = _img.getMat();
@@ -306,10 +306,10 @@ imdecode_( const Mat& buf, int flags, int hdrtype, Mat* mat=0 )
     IplImage* image = 0;
     CvMat *matrix = 0;
     Mat temp, *data = &temp;
-    std::string filename;
+    String filename;
 
     ImageDecoder decoder = findDecoder(buf);
-    if( decoder.empty() )
+    if( !decoder )
         return 0;
 
     if( !decoder->setSource(buf) )
@@ -400,7 +400,7 @@ Mat imdecode( InputArray _buf, int flags, Mat* dst )
     return *dst;
 }
 
-bool imencode( const std::string& ext, InputArray _image,
+bool imencode( const String& ext, InputArray _image,
                std::vector<uchar>& buf, const std::vector<int>& params )
 {
     Mat image = _image.getMat();
@@ -409,7 +409,7 @@ bool imencode( const std::string& ext, InputArray _image,
     CV_Assert( channels == 1 || channels == 3 || channels == 4 );
 
     ImageEncoder encoder = findEncoder( ext );
-    if( encoder.empty() )
+    if( !encoder )
         CV_Error( CV_StsError, "could not find encoder for the specified extension" );
 
     if( !encoder->isFormatSupported(image.depth()) )
@@ -429,7 +429,7 @@ bool imencode( const std::string& ext, InputArray _image,
     }
     else
     {
-        std::string filename = tempfile();
+        String filename = tempfile();
         code = encoder->setDestination(filename);
         CV_Assert( code );
 
diff --git a/modules/highgui/src/precomp.cpp b/modules/highgui/src/precomp.cpp
deleted file mode 100644
index d6f6e18f74..0000000000
--- a/modules/highgui/src/precomp.cpp
+++ /dev/null
@@ -1,43 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
diff --git a/modules/highgui/src/precomp.hpp b/modules/highgui/src/precomp.hpp
index d83401cd65..d225cb3146 100644
--- a/modules/highgui/src/precomp.hpp
+++ b/modules/highgui/src/precomp.hpp
@@ -42,12 +42,13 @@
 #ifndef __HIGHGUI_H_
 #define __HIGHGUI_H_
 
-#include "cvconfig.h"
-
 #include "opencv2/highgui.hpp"
-#include "opencv2/highgui/highgui_c.h"
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
+
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/highgui/highgui_c.h"
 
 #include <stdlib.h>
 #include <stdio.h>
@@ -103,14 +104,6 @@ struct CvVideoWriter
     virtual bool writeFrame(const IplImage*) { return false; }
 };
 
-#if defined WIN32 || defined _WIN32
-#define HAVE_VFW 1
-
-/* uncomment to enable CMUCamera1394 fireware camera module */
-//#define HAVE_CMU1394 1
-#endif
-
-
 CvCapture * cvCreateCameraCapture_V4L( int index );
 CvCapture * cvCreateCameraCapture_DC1394( int index );
 CvCapture * cvCreateCameraCapture_DC1394_2( int index );
@@ -126,6 +119,10 @@ CvVideoWriter* cvCreateVideoWriter_Win32( const char* filename, int fourcc,
 CvVideoWriter* cvCreateVideoWriter_VFW( const char* filename, int fourcc,
                                         double fps, CvSize frameSize, int is_color );
 CvCapture* cvCreateCameraCapture_DShow( int index );
+CvCapture* cvCreateCameraCapture_MSMF( int index );
+CvCapture* cvCreateFileCapture_MSMF (const char* filename);
+CvVideoWriter* cvCreateVideoWriter_MSMF( const char* filename, int fourcc,
+                                        double fps, CvSize frameSize, int is_color );
 CvCapture* cvCreateCameraCapture_OpenNI( int index );
 CvCapture* cvCreateFileCapture_OpenNI( const char* filename );
 CvCapture* cvCreateCameraCapture_Android( int index );
@@ -206,41 +203,4 @@ void cvSetRatioWindow_QT(const char* name,double prop_value);
 double cvGetOpenGlProp_QT(const char* name);
 #endif
 
-
-
-/*namespace cv
-{
-
-class CV_EXPORTS BaseWindow
-{
-public:
-    BaseWindow(const String& name, int flags=0);
-    virtual ~BaseWindow();
-    virtual void close();
-    virtual void show(const Mat& mat);
-    virtual void resize(Size size);
-    virtual void move(Point topleft);
-    virtual Size size() const;
-    virtual Point topLeft() const;
-    virtual void setGeometry(Point topLeft, Size size);
-    virtual void getGeometry(Point& topLeft, Size& size) const;
-    virtual String getTitle() const;
-    virtual void setTitle(const String& str);
-    virtual String getName() const;
-    virtual void setScaleMode(int mode);
-    virtual int getScaleMode();
-    virtual void setScrollPos(double pos);
-    virtual double getScrollPos() const;
-    virtual void setScale(double scale);
-    virtual double getScale() const;
-    virtual Point getImageCoords(Point pos) const;
-    virtual Scalar getPixelValue(Point pos, const String& colorspace=String()) const;
-
-    virtual void addTrackbar( const String& trackbar, int low, int high, int step );
-};
-
-typedef Ptr<BaseWindow> Window;
-
-}*/
-
 #endif /* __HIGHGUI_H_ */
diff --git a/modules/highgui/src/utils.cpp b/modules/highgui/src/utils.cpp
index 095a256ae2..2ee5bafc71 100644
--- a/modules/highgui/src/utils.cpp
+++ b/modules/highgui/src/utils.cpp
@@ -641,13 +641,13 @@ cvConvertImage( const CvArr* srcarr, CvArr* dstarr, int flags )
         uchar *s = src->data.ptr, *d = dst->data.ptr;
         int s_step = src->step, d_step = dst->step;
         int code = src_cn*10 + dst_cn;
-        CvSize size = { src->cols, src->rows };
+        CvSize size(src->cols, src->rows);
 
         if( CV_IS_MAT_CONT(src->type & dst->type) )
         {
             size.width *= size.height;
             size.height = 1;
-            s_step = d_step = CV_STUB_STEP;
+            s_step = d_step = /*CV_STUB_STEP*/ (1 << 30);
         }
 
         switch( code )
diff --git a/modules/highgui/src/window.cpp b/modules/highgui/src/window.cpp
index c89a0a514d..428ef51ef5 100644
--- a/modules/highgui/src/window.cpp
+++ b/modules/highgui/src/window.cpp
@@ -57,7 +57,7 @@ CV_IMPL void cvSetWindowProperty(const char* name, int prop_id, double prop_valu
 
         #if defined (HAVE_QT)
             cvSetModeWindow_QT(name,prop_value);
-        #elif defined WIN32 || defined _WIN32
+        #elif defined(HAVE_WIN32UI)
             cvSetModeWindow_W32(name,prop_value);
         #elif defined (HAVE_GTK)
             cvSetModeWindow_GTK(name,prop_value);
@@ -96,7 +96,7 @@ CV_IMPL double cvGetWindowProperty(const char* name, int prop_id)
 
         #if defined (HAVE_QT)
             return cvGetModeWindow_QT(name);
-        #elif defined WIN32 || defined _WIN32
+        #elif defined(HAVE_WIN32UI)
             return cvGetModeWindow_W32(name);
         #elif defined (HAVE_GTK)
             return cvGetModeWindow_GTK(name);
@@ -113,7 +113,7 @@ CV_IMPL double cvGetWindowProperty(const char* name, int prop_id)
 
         #if defined (HAVE_QT)
             return cvGetPropWindow_QT(name);
-        #elif defined WIN32 || defined _WIN32
+        #elif defined(HAVE_WIN32UI)
             return cvGetPropWindowAutoSize_W32(name);
         #elif defined (HAVE_GTK)
             return cvGetPropWindowAutoSize_GTK(name);
@@ -126,7 +126,7 @@ CV_IMPL double cvGetWindowProperty(const char* name, int prop_id)
 
         #if defined (HAVE_QT)
             return cvGetRatioWindow_QT(name);
-        #elif defined WIN32 || defined _WIN32
+        #elif defined(HAVE_WIN32UI)
             return cvGetRatioWindow_W32(name);
         #elif defined (HAVE_GTK)
             return cvGetRatioWindow_GTK(name);
@@ -139,7 +139,7 @@ CV_IMPL double cvGetWindowProperty(const char* name, int prop_id)
 
         #if defined (HAVE_QT)
             return cvGetOpenGlProp_QT(name);
-        #elif defined WIN32 || defined _WIN32
+        #elif defined(HAVE_WIN32UI)
             return cvGetOpenGlProp_W32(name);
         #elif defined (HAVE_GTK)
             return cvGetOpenGlProp_GTK(name);
@@ -153,12 +153,12 @@ CV_IMPL double cvGetWindowProperty(const char* name, int prop_id)
     }
 }
 
-void cv::namedWindow( const std::string& winname, int flags )
+void cv::namedWindow( const String& winname, int flags )
 {
     cvNamedWindow( winname.c_str(), flags );
 }
 
-void cv::destroyWindow( const std::string& winname )
+void cv::destroyWindow( const String& winname )
 {
     cvDestroyWindow( winname.c_str() );
 }
@@ -168,22 +168,22 @@ void cv::destroyAllWindows()
     cvDestroyAllWindows();
 }
 
-void cv::resizeWindow( const std::string& winname, int width, int height )
+void cv::resizeWindow( const String& winname, int width, int height )
 {
     cvResizeWindow( winname.c_str(), width, height );
 }
 
-void cv::moveWindow( const std::string& winname, int x, int y )
+void cv::moveWindow( const String& winname, int x, int y )
 {
     cvMoveWindow( winname.c_str(), x, y );
 }
 
-void cv::setWindowProperty(const std::string& winname, int prop_id, double prop_value)
+void cv::setWindowProperty(const String& winname, int prop_id, double prop_value)
 {
     cvSetWindowProperty( winname.c_str(), prop_id, prop_value);
 }
 
-double cv::getWindowProperty(const std::string& winname, int prop_id)
+double cv::getWindowProperty(const String& winname, int prop_id)
 {
     return cvGetWindowProperty(winname.c_str(), prop_id);
 }
@@ -193,7 +193,7 @@ int cv::waitKey(int delay)
     return cvWaitKey(delay);
 }
 
-int cv::createTrackbar(const std::string& trackbarName, const std::string& winName,
+int cv::createTrackbar(const String& trackbarName, const String& winName,
                    int* value, int count, TrackbarCallback callback,
                    void* userdata)
 {
@@ -201,17 +201,17 @@ int cv::createTrackbar(const std::string& trackbarName, const std::string& winNa
                              value, count, callback, userdata);
 }
 
-void cv::setTrackbarPos( const std::string& trackbarName, const std::string& winName, int value )
+void cv::setTrackbarPos( const String& trackbarName, const String& winName, int value )
 {
     cvSetTrackbarPos(trackbarName.c_str(), winName.c_str(), value );
 }
 
-int cv::getTrackbarPos( const std::string& trackbarName, const std::string& winName )
+int cv::getTrackbarPos( const String& trackbarName, const String& winName )
 {
     return cvGetTrackbarPos(trackbarName.c_str(), winName.c_str());
 }
 
-void cv::setMouseCallback( const std::string& windowName, MouseCallback onMouse, void* param)
+void cv::setMouseCallback( const String& windowName, MouseCallback onMouse, void* param)
 {
     cvSetMouseCallback(windowName.c_str(), onMouse, param);
 }
@@ -223,17 +223,17 @@ int cv::startWindowThread()
 
 // OpenGL support
 
-void cv::setOpenGlDrawCallback(const std::string& name, OpenGlDrawCallback callback, void* userdata)
+void cv::setOpenGlDrawCallback(const String& name, OpenGlDrawCallback callback, void* userdata)
 {
     cvSetOpenGlDrawCallback(name.c_str(), callback, userdata);
 }
 
-void cv::setOpenGlContext(const std::string& windowName)
+void cv::setOpenGlContext(const String& windowName)
 {
     cvSetOpenGlContext(windowName.c_str());
 }
 
-void cv::updateWindow(const std::string& windowName)
+void cv::updateWindow(const String& windowName)
 {
     cvUpdateWindow(windowName.c_str());
 }
@@ -241,9 +241,9 @@ void cv::updateWindow(const std::string& windowName)
 #ifdef HAVE_OPENGL
 namespace
 {
-    std::map<std::string, cv::ogl::Texture2D> wndTexs;
-    std::map<std::string, cv::ogl::Texture2D> ownWndTexs;
-    std::map<std::string, cv::ogl::Buffer> ownWndBufs;
+    std::map<cv::String, cv::ogl::Texture2D> wndTexs;
+    std::map<cv::String, cv::ogl::Texture2D> ownWndTexs;
+    std::map<cv::String, cv::ogl::Buffer> ownWndBufs;
 
     void glDrawTextureCallback(void* userdata)
     {
@@ -254,14 +254,19 @@ namespace
 }
 #endif // HAVE_OPENGL
 
-void cv::imshow( const std::string& winname, InputArray _img )
+void cv::imshow( const String& winname, InputArray _img )
 {
+    const Size size = _img.size();
 #ifndef HAVE_OPENGL
-    Mat img = _img.getMat();
-    CvMat c_img = img;
-    cvShowImage(winname.c_str(), &c_img);
+    CV_Assert(size.width>0 && size.height>0);
+    {
+        Mat img = _img.getMat();
+        CvMat c_img = img;
+        cvShowImage(winname.c_str(), &c_img);
+    }
 #else
     const double useGl = getWindowProperty(winname, WND_PROP_OPENGL);
+    CV_Assert(size.width>0 && size.height>0);
 
     if (useGl <= 0)
     {
@@ -275,45 +280,69 @@ void cv::imshow( const std::string& winname, InputArray _img )
 
         if (autoSize > 0)
         {
-            Size size = _img.size();
             resizeWindow(winname, size.width, size.height);
         }
 
         setOpenGlContext(winname);
 
-        if (_img.kind() == _InputArray::OPENGL_TEXTURE)
+        cv::ogl::Texture2D& tex = ownWndTexs[winname];
+
+        if (_img.kind() == _InputArray::GPU_MAT)
         {
-            cv::ogl::Texture2D& tex = wndTexs[winname];
-
-            tex = _img.getOGlTexture2D();
+            cv::ogl::Buffer& buf = ownWndBufs[winname];
+            buf.copyFrom(_img);
+            buf.setAutoRelease(false);
 
+            tex.copyFrom(buf);
             tex.setAutoRelease(false);
-
-            setOpenGlDrawCallback(winname, glDrawTextureCallback, &tex);
         }
         else
         {
-            cv::ogl::Texture2D& tex = ownWndTexs[winname];
-
-            if (_img.kind() == _InputArray::GPU_MAT)
-            {
-                cv::ogl::Buffer& buf = ownWndBufs[winname];
-                buf.copyFrom(_img);
-                buf.setAutoRelease(false);
-
-                tex.copyFrom(buf);
-                tex.setAutoRelease(false);
-            }
-            else
-            {
-                tex.copyFrom(_img);
-            }
-
-            tex.setAutoRelease(false);
-
-            setOpenGlDrawCallback(winname, glDrawTextureCallback, &tex);
+            tex.copyFrom(_img);
         }
 
+        tex.setAutoRelease(false);
+
+        setOpenGlDrawCallback(winname, glDrawTextureCallback, &tex);
+
+        updateWindow(winname);
+    }
+#endif
+}
+
+void cv::imshow(const String& winname, const ogl::Texture2D& _tex)
+{
+#ifndef HAVE_OPENGL
+    (void) winname;
+    (void) _tex;
+    CV_Error(cv::Error::OpenGlNotSupported, "The library is compiled without OpenGL support");
+#else
+    const double useGl = getWindowProperty(winname, WND_PROP_OPENGL);
+
+    if (useGl <= 0)
+    {
+        CV_Error(cv::Error::OpenGlNotSupported, "The window was created without OpenGL context");
+    }
+    else
+    {
+        const double autoSize = getWindowProperty(winname, WND_PROP_AUTOSIZE);
+
+        if (autoSize > 0)
+        {
+            Size size = _tex.size();
+            resizeWindow(winname, size.width, size.height);
+        }
+
+        setOpenGlContext(winname);
+
+        cv::ogl::Texture2D& tex = wndTexs[winname];
+
+        tex = _tex;
+
+        tex.setAutoRelease(false);
+
+        setOpenGlDrawCallback(winname, glDrawTextureCallback, &tex);
+
         updateWindow(winname);
     }
 #endif
@@ -342,23 +371,24 @@ CV_IMPL void cvUpdateWindow(const char*)
 
 #if defined (HAVE_QT)
 
-CvFont cv::fontQt(const std::string& nameFont, int pointSize, Scalar color, int weight,  int style, int /*spacing*/)
+cv::QtFont cv::fontQt(const String& nameFont, int pointSize, Scalar color, int weight,  int style, int /*spacing*/)
 {
-return cvFontQt(nameFont.c_str(), pointSize,color,weight, style);
+    CvFont f = cvFontQt(nameFont.c_str(), pointSize,color,weight, style);
+    return *(cv::QtFont*)(&f);
 }
 
-void cv::addText( const Mat& img, const std::string& text, Point org, CvFont font)
+void cv::addText( const Mat& img, const String& text, Point org, const QtFont& font)
 {
     CvMat _img = img;
-    cvAddText( &_img, text.c_str(), org,&font);
+    cvAddText( &_img, text.c_str(), org, (CvFont*)&font);
 }
 
-void cv::displayStatusBar(const std::string& name,  const std::string& text, int delayms)
+void cv::displayStatusBar(const String& name,  const String& text, int delayms)
 {
     cvDisplayStatusBar(name.c_str(),text.c_str(), delayms);
 }
 
-void cv::displayOverlay(const std::string& name,  const std::string& text, int delayms)
+void cv::displayOverlay(const String& name,  const String& text, int delayms)
 {
     cvDisplayOverlay(name.c_str(),text.c_str(), delayms);
 }
@@ -373,40 +403,40 @@ void cv::stopLoop()
     cvStopLoop();
 }
 
-void cv::saveWindowParameters(const std::string& windowName)
+void cv::saveWindowParameters(const String& windowName)
 {
     cvSaveWindowParameters(windowName.c_str());
 }
 
-void cv::loadWindowParameters(const std::string& windowName)
+void cv::loadWindowParameters(const String& windowName)
 {
     cvLoadWindowParameters(windowName.c_str());
 }
 
-int cv::createButton(const std::string& button_name, ButtonCallback on_change, void* userdata, int button_type , bool initial_button_state  )
+int cv::createButton(const String& button_name, ButtonCallback on_change, void* userdata, int button_type , bool initial_button_state  )
 {
     return cvCreateButton(button_name.c_str(), on_change, userdata, button_type , initial_button_state );
 }
 
 #else
 
-CvFont cv::fontQt(const std::string&, int, Scalar, int,  int, int)
+cv::QtFont cv::fontQt(const String&, int, Scalar, int,  int, int)
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
-    return CvFont();
+    return QtFont();
 }
 
-void cv::addText( const Mat&, const std::string&, Point, CvFont)
+void cv::addText( const Mat&, const String&, Point, const QtFont&)
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
 
-void cv::displayStatusBar(const std::string&,  const std::string&, int)
+void cv::displayStatusBar(const String&,  const String&, int)
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
 
-void cv::displayOverlay(const std::string&,  const std::string&, int )
+void cv::displayOverlay(const String&,  const String&, int )
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
@@ -422,17 +452,17 @@ void cv::stopLoop()
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
 
-void cv::saveWindowParameters(const std::string&)
+void cv::saveWindowParameters(const String&)
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
 
-void cv::loadWindowParameters(const std::string&)
+void cv::loadWindowParameters(const String&)
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
 }
 
-int cv::createButton(const std::string&, ButtonCallback, void*, int , bool )
+int cv::createButton(const String&, ButtonCallback, void*, int , bool )
 {
     CV_Error(CV_StsNotImplemented, "The library is compiled without QT support");
     return 0;
@@ -440,11 +470,11 @@ int cv::createButton(const std::string&, ButtonCallback, void*, int , bool )
 
 #endif
 
-#if   defined WIN32 || defined _WIN32         // see window_w32.cpp
+#if   defined(HAVE_WIN32UI)   // see window_w32.cpp
 #elif defined (HAVE_GTK)      // see window_gtk.cpp
-#elif defined (HAVE_COCOA)   // see window_carbon.cpp
+#elif defined (HAVE_COCOA)    // see window_carbon.cpp
 #elif defined (HAVE_CARBON)
-#elif defined (HAVE_QT) //YV see window_QT.cpp
+#elif defined (HAVE_QT)       //YV see window_QT.cpp
 
 #else
 
diff --git a/modules/highgui/src/window_QT.cpp b/modules/highgui/src/window_QT.cpp
index 50f2b9e787..2b7775bd82 100644
--- a/modules/highgui/src/window_QT.cpp
+++ b/modules/highgui/src/window_QT.cpp
@@ -38,6 +38,7 @@
 
 //--------------------Google Code 2010 -- Yannick Verdie--------------------//
 
+#include "precomp.hpp"
 
 #if defined(HAVE_QT)
 
@@ -948,8 +949,8 @@ void GuiReceiver::showImage(QString name, void* arr)
 
         mat = cvGetMat(arr, &stub);
 
-        cv::Mat im(mat);
-        cv::imshow(name.toStdString(), im);
+        cv::Mat im = cv::cvarrToMat(mat);
+        cv::imshow(name.toUtf8().data(), im);
     }
     else
     {
@@ -2474,34 +2475,32 @@ void DefaultViewPort::saveView()
     {
         QString extension = fileName.right(3);
 
-        //   (no need anymore) create the image resized to receive the 'screenshot'
-        //    image2Draw_qt_resized = QImage(viewport()->width(), viewport()->height(),QImage::Format_RGB888);
-
-        QPainter saveimage(&image2Draw_qt_resized);
-        this->render(&saveimage);
+        // Create a new pixmap to render the viewport into
+        QPixmap viewportPixmap(viewport()->size());
+        viewport()->render(&viewportPixmap);
 
         // Save it..
         if (QString::compare(extension, "png", Qt::CaseInsensitive) == 0)
         {
-            image2Draw_qt_resized.save(fileName, "PNG");
+            viewportPixmap.save(fileName, "PNG");
             return;
         }
 
         if (QString::compare(extension, "jpg", Qt::CaseInsensitive) == 0)
         {
-            image2Draw_qt_resized.save(fileName, "JPG");
+            viewportPixmap.save(fileName, "JPG");
             return;
         }
 
         if (QString::compare(extension, "bmp", Qt::CaseInsensitive) == 0)
         {
-            image2Draw_qt_resized.save(fileName, "BMP");
+            viewportPixmap.save(fileName, "BMP");
             return;
         }
 
         if (QString::compare(extension, "jpeg", Qt::CaseInsensitive) == 0)
         {
-            image2Draw_qt_resized.save(fileName, "JPEG");
+            viewportPixmap.save(fileName, "JPEG");
             return;
         }
 
@@ -2651,17 +2650,16 @@ void DefaultViewPort::paintEvent(QPaintEvent* evnt)
     //Now disable matrixWorld for overlay display
     myPainter.setWorldMatrixEnabled(false);
 
+    //overlay pixel values if zoomed in far enough
+    if (param_matrixWorld.m11()*ratioX >= threshold_zoom_img_region &&
+        param_matrixWorld.m11()*ratioY >= threshold_zoom_img_region)
+    {
+        drawImgRegion(&myPainter);
+    }
+
     //in mode zoom/panning
     if (param_matrixWorld.m11() > 1)
     {
-        if (param_matrixWorld.m11() >= threshold_zoom_img_region)
-        {
-            if (centralWidget->param_flags == CV_WINDOW_NORMAL)
-                startDisplayInfo("WARNING: The values displayed are the resized image's values. If you want the original image's values, use CV_WINDOW_AUTOSIZE", 1000);
-
-            drawImgRegion(&myPainter);
-        }
-
         drawViewOverview(&myPainter);
     }
 
@@ -2887,22 +2885,24 @@ void DefaultViewPort::drawStatusBar()
 //accept only CV_8UC1 and CV_8UC8 image for now
 void DefaultViewPort::drawImgRegion(QPainter *painter)
 {
-
     if (nbChannelOriginImage!=CV_8UC1 && nbChannelOriginImage!=CV_8UC3)
         return;
 
-    qreal offsetX = param_matrixWorld.dx()/param_matrixWorld.m11();
+    double pixel_width = param_matrixWorld.m11()*ratioX;
+    double pixel_height = param_matrixWorld.m11()*ratioY;
+
+    qreal offsetX = param_matrixWorld.dx()/pixel_width;
     offsetX = offsetX - floor(offsetX);
-    qreal offsetY = param_matrixWorld.dy()/param_matrixWorld.m11();
+    qreal offsetY = param_matrixWorld.dy()/pixel_height;
     offsetY = offsetY - floor(offsetY);
 
     QSize view = size();
     QVarLengthArray<QLineF, 30> linesX;
-    for (qreal _x = offsetX*param_matrixWorld.m11(); _x < view.width(); _x += param_matrixWorld.m11() )
+    for (qreal _x = offsetX*pixel_width; _x < view.width(); _x += pixel_width )
         linesX.append(QLineF(_x, 0, _x, view.height()));
 
     QVarLengthArray<QLineF, 30> linesY;
-    for (qreal _y = offsetY*param_matrixWorld.m11(); _y < view.height(); _y += param_matrixWorld.m11() )
+    for (qreal _y = offsetY*pixel_height; _y < view.height(); _y += pixel_height )
         linesY.append(QLineF(0, _y, view.width(), _y));
 
 
@@ -2910,27 +2910,25 @@ void DefaultViewPort::drawImgRegion(QPainter *painter)
     int original_font_size = f.pointSize();
     //change font size
     //f.setPointSize(4+(param_matrixWorld.m11()-threshold_zoom_img_region)/5);
-    f.setPixelSize(10+(param_matrixWorld.m11()-threshold_zoom_img_region)/5);
+    f.setPixelSize(10+(pixel_height-threshold_zoom_img_region)/5);
     painter->setFont(f);
-    QString val;
-    QRgb rgbValue;
 
-    QPointF point1;//sorry, I do not know how to name it
-    QPointF point2;//idem
 
-    for (int j=-1;j<height()/param_matrixWorld.m11();j++)//-1 because display the pixels top rows left colums
-        for (int i=-1;i<width()/param_matrixWorld.m11();i++)//-1
+    for (int j=-1;j<height()/pixel_height;j++)//-1 because display the pixels top rows left columns
+        for (int i=-1;i<width()/pixel_width;i++)//-1
         {
-            point1.setX((i+offsetX)*param_matrixWorld.m11());
-            point1.setY((j+offsetY)*param_matrixWorld.m11());
+            // Calculate top left of the pixel's position in the viewport (screen space)
+            QPointF pos_in_view((i+offsetX)*pixel_width, (j+offsetY)*pixel_height);
 
-            matrixWorld_inv.map(point1.x(),point1.y(),&point2.rx(),&point2.ry());
+            // Calculate top left of the pixel's position in the image (image space)
+            QPointF pos_in_image = matrixWorld_inv.map(pos_in_view);// Top left of pixel in view
+            pos_in_image.rx() = pos_in_image.x()/ratioX;
+            pos_in_image.ry() = pos_in_image.y()/ratioY;
+            QPoint point_in_image(pos_in_image.x() + 0.5f,pos_in_image.y() + 0.5f);// Add 0.5 for rounding
 
-            point2.rx()= (long) (point2.x() + 0.5);
-            point2.ry()= (long) (point2.y() + 0.5);
-
-            if (point2.x() >= 0 && point2.y() >= 0)
-                rgbValue = image2Draw_qt_resized.pixel(QPoint(point2.x(),point2.y()));
+            QRgb rgbValue;
+            if (image2Draw_qt.valid(point_in_image))
+                rgbValue = image2Draw_qt.pixel(point_in_image);
             else
                 rgbValue = qRgb(0,0,0);
 
@@ -2943,29 +2941,29 @@ void DefaultViewPort::drawImgRegion(QPainter *painter)
                 painter->drawText(QRect(point1.x(),point1.y(),param_matrixWorld.m11(),param_matrixWorld.m11()/2),
                     Qt::AlignCenter, val);
                 */
+                QString val;
 
                 val = tr("%1").arg(qRed(rgbValue));
                 painter->setPen(QPen(Qt::red, 1));
-                painter->drawText(QRect(point1.x(),point1.y(),param_matrixWorld.m11(),param_matrixWorld.m11()/3),
+                painter->drawText(QRect(pos_in_view.x(),pos_in_view.y(),pixel_width,pixel_height/3),
                     Qt::AlignCenter, val);
 
                 val = tr("%1").arg(qGreen(rgbValue));
                 painter->setPen(QPen(Qt::green, 1));
-                painter->drawText(QRect(point1.x(),point1.y()+param_matrixWorld.m11()/3,param_matrixWorld.m11(),param_matrixWorld.m11()/3),
+                painter->drawText(QRect(pos_in_view.x(),pos_in_view.y()+pixel_height/3,pixel_width,pixel_height/3),
                     Qt::AlignCenter, val);
 
                 val = tr("%1").arg(qBlue(rgbValue));
                 painter->setPen(QPen(Qt::blue, 1));
-                painter->drawText(QRect(point1.x(),point1.y()+2*param_matrixWorld.m11()/3,param_matrixWorld.m11(),param_matrixWorld.m11()/3),
+                painter->drawText(QRect(pos_in_view.x(),pos_in_view.y()+2*pixel_height/3,pixel_width,pixel_height/3),
                     Qt::AlignCenter, val);
 
             }
 
             if (nbChannelOriginImage==CV_8UC1)
             {
-
-                val = tr("%1").arg(qRed(rgbValue));
-                painter->drawText(QRect(point1.x(),point1.y(),param_matrixWorld.m11(),param_matrixWorld.m11()),
+                QString val = tr("%1").arg(qRed(rgbValue));
+                painter->drawText(QRect(pos_in_view.x(),pos_in_view.y(),pixel_width,pixel_height),
                     Qt::AlignCenter, val);
             }
         }
diff --git a/modules/highgui/src/window_QT.h b/modules/highgui/src/window_QT.h
index 1b07442dc9..a96a8c6e69 100644
--- a/modules/highgui/src/window_QT.h
+++ b/modules/highgui/src/window_QT.h
@@ -48,13 +48,13 @@
 #endif
 
 #include <QAbstractEventDispatcher>
-#include <QtGui/QApplication>
+#include <QApplication>
 #include <QFile>
 #include <QPushButton>
-#include <QtGui/QGraphicsView>
+#include <QGraphicsView>
 #include <QSizePolicy>
 #include <QInputDialog>
-#include <QtGui/QBoxLayout>
+#include <QBoxLayout>
 #include <QSettings>
 #include <qtimer.h>
 #include <QtConcurrentRun>
@@ -78,7 +78,7 @@
 #include <QRadioButton>
 #include <QButtonGroup>
 #include <QMenu>
-#include <QtTest/QTest>
+#include <QTest>
 
 //start private enum
 enum { CV_MODE_NORMAL = 0, CV_MODE_OPENGL = 1 };
@@ -522,7 +522,6 @@ private:
 
     CvMat* image2Draw_mat;
     QImage image2Draw_qt;
-    QImage image2Draw_qt_resized;
     int nbChannelOriginImage;
 
     //for mouse callback
diff --git a/modules/highgui/src/window_gtk.cpp b/modules/highgui/src/window_gtk.cpp
index 3ee3d0c480..82c49770c4 100644
--- a/modules/highgui/src/window_gtk.cpp
+++ b/modules/highgui/src/window_gtk.cpp
@@ -1096,10 +1096,7 @@ cvShowImage( const char* name, const CvArr* arr )
     #ifdef HAVE_OPENGL
         if (window->useGl)
         {
-            CvMat stub;
-            CvMat* mat = cvGetMat(arr, &stub);
-            cv::Mat im(mat);
-            cv::imshow(name, im);
+            cv::imshow(name, cv::cvarrToMat(arr));
             return;
         }
     #endif
@@ -1501,8 +1498,8 @@ static gboolean icvOnMouse( GtkWidget *widget, GdkEvent *event, gpointer user_da
 {
     // TODO move this logic to CvImageWidget
     CvWindow* window = (CvWindow*)user_data;
-    CvPoint2D32f pt32f = {-1., -1.};
-    CvPoint pt = {-1,-1};
+    CvPoint2D32f pt32f(-1., -1.);
+    CvPoint pt(-1,-1);
     int cv_event = -1, state = 0;
     CvImageWidget * image_widget = CV_IMAGE_WIDGET( widget );
 
diff --git a/modules/highgui/src/window_w32.cpp b/modules/highgui/src/window_w32.cpp
index 71445297a7..65e9966cee 100644
--- a/modules/highgui/src/window_w32.cpp
+++ b/modules/highgui/src/window_w32.cpp
@@ -1107,8 +1107,7 @@ cvShowImage( const char* name, const CvArr* arr )
 #ifdef HAVE_OPENGL
     if (window->useGl)
     {
-        cv::Mat im(image);
-        cv::imshow(name, im);
+        cv::imshow(name, cv::cvarrToMat(image));
         return;
     }
 #endif
diff --git a/modules/highgui/test/test_ffmpeg.cpp b/modules/highgui/test/test_ffmpeg.cpp
index ca22ca0ddb..20ba7c5cbb 100644
--- a/modules/highgui/test/test_ffmpeg.cpp
+++ b/modules/highgui/test/test_ffmpeg.cpp
@@ -95,15 +95,15 @@ public:
             double fps = fps0;
             Size frame_s = Size(img_c, img_r);
 
-            if( tag == CV_FOURCC('H', '2', '6', '1') )
+            if( tag == VideoWriter::fourcc('H', '2', '6', '1') )
                 frame_s = Size(352, 288);
-            else if( tag == CV_FOURCC('H', '2', '6', '3') )
+            else if( tag == VideoWriter::fourcc('H', '2', '6', '3') )
                 frame_s = Size(704, 576);
             /*else if( tag == CV_FOURCC('M', 'J', 'P', 'G') ||
                      tag == CV_FOURCC('j', 'p', 'e', 'g') )
                 frame_s = Size(1920, 1080);*/
 
-            if( tag == CV_FOURCC('M', 'P', 'E', 'G') )
+            if( tag == VideoWriter::fourcc('M', 'P', 'E', 'G') )
                 fps = 25;
 
             VideoWriter writer(filename, tag, fps, frame_s);
@@ -176,7 +176,7 @@ TEST(Highgui_Video, ffmpeg_image) { CV_FFmpegReadImageTest test; test.safe_run()
 
 #endif
 
-#if defined(HAVE_FFMPEG) || defined(WIN32) || defined(_WIN32)
+#if defined(HAVE_FFMPEG)
 
 //////////////////////////////// Parallel VideoWriters and VideoCaptures ////////////////////////////////////
 
@@ -201,7 +201,7 @@ public:
             std::string fileName = tempfile(stream.str().c_str());
 
             files->operator[](i) = fileName;
-            writers->operator[](i) = new VideoWriter(fileName, CV_FOURCC('X','V','I','D'), 25.0f, FrameSize);
+            writers->operator[](i) = new VideoWriter(fileName, VideoWriter::fourcc('X','V','I','D'), 25.0f, FrameSize);
 
             CV_Assert(writers->operator[](i)->isOpened());
         }
@@ -311,7 +311,7 @@ public:
             CV_Assert(capture->isOpened());
 
             const static double eps = 23.0;
-            unsigned int frameCount = static_cast<unsigned int>(capture->get(CV_CAP_PROP_FRAME_COUNT));
+            unsigned int frameCount = static_cast<unsigned int>(capture->get(CAP_PROP_FRAME_COUNT));
             CV_Assert(frameCount == WriteVideo_Invoker::FrameCount);
             Mat reference(CreateVideoWriterInvoker::FrameSize, CV_8UC3);
 
diff --git a/modules/highgui/test/test_fourcc.cpp b/modules/highgui/test/test_fourcc.cpp
index 584b8c5964..82f699b098 100644
--- a/modules/highgui/test/test_fourcc.cpp
+++ b/modules/highgui/test/test_fourcc.cpp
@@ -112,4 +112,4 @@ TEST(Highgui_dshow, fourcc_conversion)
 
         EXPECT_EQ(fourcc, (unsigned long)(unsigned)fourccFromParam);
     }
-}
\ No newline at end of file
+}
diff --git a/modules/highgui/test/test_framecount.cpp b/modules/highgui/test/test_framecount.cpp
index 69ebf5cf83..30f6e67ce6 100644
--- a/modules/highgui/test/test_framecount.cpp
+++ b/modules/highgui/test/test_framecount.cpp
@@ -41,7 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
-#include "opencv2/highgui.hpp"
+#include "opencv2/highgui/highgui_c.h"
 #include <stdio.h>
 
 using namespace cv;
@@ -71,8 +71,8 @@ void CV_FramecountTest::run(int)
     {
         string file_path = src_dir+"video/big_buck_bunny."+ext[i];
 
-        cap = cvCreateFileCapture(file_path.c_str());
-        if (cap.empty())
+        cap.reset(cvCreateFileCapture(file_path.c_str()));
+        if (!cap)
         {
             ts->printf(cvtest::TS::LOG, "\nFile information (video %d): \n\nName: big_buck_bunny.%s\nFAILED\n\n", i+1, ext[i].c_str());
             ts->printf(cvtest::TS::LOG, "Error: cannot read source video file.\n");
@@ -91,7 +91,7 @@ void CV_FramecountTest::run(int)
             FrameCount++;
         }
 
-        int framecount = (int)cvGetCaptureProperty(cap, CV_CAP_PROP_FRAME_COUNT);
+        int framecount = (int)cvGetCaptureProperty(cap, CAP_PROP_FRAME_COUNT);
 
         ts->printf(cvtest::TS::LOG, "\nFile information (video %d): \n"\
                    "\nName: big_buck_bunny.%s\nActual frame count: %d\n"\
diff --git a/modules/highgui/test/test_grfmt.cpp b/modules/highgui/test/test_grfmt.cpp
index 57b48a290e..8f3c12a8a1 100644
--- a/modules/highgui/test/test_grfmt.cpp
+++ b/modules/highgui/test/test_grfmt.cpp
@@ -113,7 +113,7 @@ public:
                         imwrite(img_path, img);
 
                         ts->printf(ts->LOG, "reading test image : %s\n", img_path.c_str());
-                        Mat img_test = imread(img_path, CV_LOAD_IMAGE_UNCHANGED);
+                        Mat img_test = imread(img_path, IMREAD_UNCHANGED);
 
                         if (img_test.empty()) ts->set_failed_test_info(ts->FAIL_MISMATCH);
 
@@ -140,11 +140,11 @@ public:
 
                     string filename = cv::tempfile(".jpg");
                     imwrite(filename, img);
-                    img = imread(filename, CV_LOAD_IMAGE_UNCHANGED);
+                    img = imread(filename, IMREAD_UNCHANGED);
 
                     filename = string(ts->get_data_path() + "readwrite/test_" + char(k + 48) + "_c" + char(num_channels + 48) + ".jpg");
                     ts->printf(ts->LOG, "reading test image : %s\n", filename.c_str());
-                    Mat img_test = imread(filename, CV_LOAD_IMAGE_UNCHANGED);
+                    Mat img_test = imread(filename, IMREAD_UNCHANGED);
 
                     if (img_test.empty()) ts->set_failed_test_info(ts->FAIL_MISMATCH);
 
@@ -171,7 +171,7 @@ public:
                     string filename = cv::tempfile(".tiff");
                     imwrite(filename, img);
                     ts->printf(ts->LOG, "reading test image : %s\n", filename.c_str());
-                    Mat img_test = imread(filename, CV_LOAD_IMAGE_UNCHANGED);
+                    Mat img_test = imread(filename, IMREAD_UNCHANGED);
 
                     if (img_test.empty()) ts->set_failed_test_info(ts->FAIL_MISMATCH);
 
@@ -242,12 +242,12 @@ public:
             Mat im = Mat::zeros(1000,1000, CV_8U);
             //randu(im, 0, 256);
             vector<int> param;
-            param.push_back(CV_IMWRITE_PNG_COMPRESSION);
+            param.push_back(IMWRITE_PNG_COMPRESSION);
             param.push_back(3); //default(3) 0-9.
             cv::imencode(".png" ,im ,buff, param);
 
             // hangs
-            Mat im2 = imdecode(buff,CV_LOAD_IMAGE_ANYDEPTH);
+            Mat im2 = imdecode(buff,IMREAD_ANYDEPTH);
         }
         catch(...)
         {
@@ -284,6 +284,98 @@ TEST(Highgui_ImreadVSCvtColor, regression)
     EXPECT_LT(actual_avg_diff, MAX_MEAN_DIFF);
     EXPECT_LT(actual_maxval, MAX_ABS_DIFF);
 }
+
+//Test OpenCV issue 3075 is solved
+class CV_GrfmtReadPNGColorPaletteWithAlphaTest : public cvtest::BaseTest
+{
+public:
+    void run(int)
+    {
+        try
+        {
+            // First Test : Read PNG with alpha, imread flag -1
+            Mat img = imread(string(ts->get_data_path()) + "readwrite/color_palette_alpha.png",-1);
+            if (img.empty()) ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
+
+            ASSERT_TRUE(img.channels() == 4);
+
+            unsigned char* img_data = (unsigned char*)img.data;
+
+            // Verification first pixel is red in BGRA
+            ASSERT_TRUE(img_data[0] == 0x00);
+            ASSERT_TRUE(img_data[1] == 0x00);
+            ASSERT_TRUE(img_data[2] == 0xFF);
+            ASSERT_TRUE(img_data[3] == 0xFF);
+
+            // Verification second pixel is red in BGRA
+            ASSERT_TRUE(img_data[4] == 0x00);
+            ASSERT_TRUE(img_data[5] == 0x00);
+            ASSERT_TRUE(img_data[6] == 0xFF);
+            ASSERT_TRUE(img_data[7] == 0xFF);
+
+            // Second Test : Read PNG without alpha, imread flag -1
+            img = imread(string(ts->get_data_path()) + "readwrite/color_palette_no_alpha.png",-1);
+            if (img.empty()) ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
+
+            ASSERT_TRUE(img.channels() == 3);
+
+            img_data = (unsigned char*)img.data;
+
+            // Verification first pixel is red in BGR
+            ASSERT_TRUE(img_data[0] == 0x00);
+            ASSERT_TRUE(img_data[1] == 0x00);
+            ASSERT_TRUE(img_data[2] == 0xFF);
+
+            // Verification second pixel is red in BGR
+            ASSERT_TRUE(img_data[3] == 0x00);
+            ASSERT_TRUE(img_data[4] == 0x00);
+            ASSERT_TRUE(img_data[5] == 0xFF);
+
+            // Third Test : Read PNG with alpha, imread flag 1
+            img = imread(string(ts->get_data_path()) + "readwrite/color_palette_alpha.png",1);
+            if (img.empty()) ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
+
+            ASSERT_TRUE(img.channels() == 3);
+
+            img_data = (unsigned char*)img.data;
+
+            // Verification first pixel is red in BGR
+            ASSERT_TRUE(img_data[0] == 0x00);
+            ASSERT_TRUE(img_data[1] == 0x00);
+            ASSERT_TRUE(img_data[2] == 0xFF);
+
+            // Verification second pixel is red in BGR
+            ASSERT_TRUE(img_data[3] == 0x00);
+            ASSERT_TRUE(img_data[4] == 0x00);
+            ASSERT_TRUE(img_data[5] == 0xFF);
+
+            // Fourth Test : Read PNG without alpha, imread flag 1
+            img = imread(string(ts->get_data_path()) + "readwrite/color_palette_no_alpha.png",1);
+            if (img.empty()) ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
+
+            ASSERT_TRUE(img.channels() == 3);
+
+            img_data = (unsigned char*)img.data;
+
+            // Verification first pixel is red in BGR
+            ASSERT_TRUE(img_data[0] == 0x00);
+            ASSERT_TRUE(img_data[1] == 0x00);
+            ASSERT_TRUE(img_data[2] == 0xFF);
+
+            // Verification second pixel is red in BGR
+            ASSERT_TRUE(img_data[3] == 0x00);
+            ASSERT_TRUE(img_data[4] == 0x00);
+            ASSERT_TRUE(img_data[5] == 0xFF);
+        }
+        catch(...)
+        {
+            ts->set_failed_test_info(cvtest::TS::FAIL_EXCEPTION);
+    }
+        ts->set_failed_test_info(cvtest::TS::OK);
+    }
+};
+
+TEST(Highgui_Image, read_png_color_palette_with_alpha) { CV_GrfmtReadPNGColorPaletteWithAlphaTest test; test.safe_run(); }
 #endif
 
 #ifdef HAVE_JPEG
@@ -373,7 +465,9 @@ TEST(Highgui_WebP, encode_decode_lossless_webp)
         }
     }
 
-    cv::Mat decode = cv::imdecode(buf, CV_LOAD_IMAGE_COLOR);
+    remove(output.c_str());
+
+    cv::Mat decode = cv::imdecode(buf, IMREAD_COLOR);
     ASSERT_FALSE(decode.empty());
     EXPECT_TRUE(cv::norm(decode, img_webp, NORM_INF) == 0);
 
@@ -385,41 +479,49 @@ TEST(Highgui_WebP, encode_decode_lossless_webp)
 TEST(Highgui_WebP, encode_decode_lossy_webp)
 {
     cvtest::TS& ts = *cvtest::TS::ptr();
-    std::string input = std::string(ts.get_data_path()) + "/../cv/shared/lena.png";
+    std::string input = std::string(ts.get_data_path()) + "../cv/shared/lena.png";
     cv::Mat img = cv::imread(input);
     ASSERT_FALSE(img.empty());
 
-    for(int q = 100; q>=0; q-=5)
+    for(int q = 100; q>=0; q-=20)
     {
         std::vector<int> params;
-        params.push_back(CV_IMWRITE_WEBP_QUALITY);
+        params.push_back(IMWRITE_WEBP_QUALITY);
         params.push_back(q);
         string output = cv::tempfile(".webp");
 
         EXPECT_NO_THROW(cv::imwrite(output, img, params));
         cv::Mat img_webp = cv::imread(output);
+        remove(output.c_str());
         EXPECT_FALSE(img_webp.empty());
+        EXPECT_EQ(3,   img_webp.channels());
+        EXPECT_EQ(512, img_webp.cols);
+        EXPECT_EQ(512, img_webp.rows);
     }
 }
 
-TEST(Highgui_WebP, encode_big_image_webp)
+TEST(Highgui_WebP, encode_decode_with_alpha_webp)
 {
     cvtest::TS& ts = *cvtest::TS::ptr();
-    std::string input = std::string(ts.get_data_path()) + "/../cv/shared/lena.png";
+    std::string input = std::string(ts.get_data_path()) + "../cv/shared/lena.png";
     cv::Mat img = cv::imread(input);
-    cv::resize(img, img, cv::Size(8192, 8192));
+    ASSERT_FALSE(img.empty());
 
-    std::vector<int> params;
-    params.push_back(CV_IMWRITE_WEBP_QUALITY);
-    params.push_back(95);
+    std::vector<cv::Mat> imgs;
+    cv::split(img, imgs);
+    imgs.push_back(cv::Mat(imgs[0]));
+    imgs[imgs.size() - 1] = cv::Scalar::all(128);
+    cv::merge(imgs, img);
 
-    std::string output = cv::tempfile(".webp");
-    EXPECT_NO_THROW(cv::imwrite(output, img, params));
+    string output = cv::tempfile(".webp");
 
-    img.create(10000, 10000, CV_8UC3);
-    cv::randu(img, 0, 256);
-    output = cv::tempfile(".webp");
     EXPECT_NO_THROW(cv::imwrite(output, img));
+    cv::Mat img_webp = cv::imread(output);
+    remove(output.c_str());
+    EXPECT_FALSE(img_webp.empty());
+    EXPECT_EQ(4,   img_webp.channels());
+    EXPECT_EQ(512, img_webp.cols);
+    EXPECT_EQ(512, img_webp.rows);
 }
 
 #endif
diff --git a/modules/highgui/test/test_gui.cpp b/modules/highgui/test/test_gui.cpp
index 9ca0aaf7cb..e53accfe71 100644
--- a/modules/highgui/test/test_gui.cpp
+++ b/modules/highgui/test/test_gui.cpp
@@ -43,7 +43,7 @@
 #include "test_precomp.hpp"
 #include "opencv2/highgui.hpp"
 
-#if defined HAVE_GTK  || defined HAVE_QT || defined WIN32 || defined _WIN32 || defined HAVE_CARBON || defined HAVE_COCOA
+#if defined HAVE_GTK || defined HAVE_QT || defined HAVE_WIN32UI || defined HAVE_CARBON || defined HAVE_COCOA
 
 using namespace cv;
 using namespace std;
@@ -59,7 +59,7 @@ void Foo(int /*k*/, void* /*z*/) {}
 void CV_HighGuiOnlyGuiTest::run( int /*start_from */)
 {
     ts->printf(ts->LOG, "GUI 0\n");
-    cvDestroyAllWindows();
+    destroyAllWindows();
 
     ts->printf(ts->LOG, "GUI 1\n");
     namedWindow("Win");
@@ -84,7 +84,7 @@ void CV_HighGuiOnlyGuiTest::run( int /*start_from */)
     waitKey(500);
 
     ts->printf(ts->LOG, "GUI 8\n");
-    cvDestroyAllWindows();
+    destroyAllWindows();
     ts->set_failed_test_info(cvtest::TS::OK);
 }
 
diff --git a/modules/highgui/test/test_main.cpp b/modules/highgui/test/test_main.cpp
index 363b54161e..4727b9565c 100644
--- a/modules/highgui/test/test_main.cpp
+++ b/modules/highgui/test/test_main.cpp
@@ -1,4 +1,3 @@
 #include "test_precomp.hpp"
 
 CV_TEST_MAIN("highgui")
-
diff --git a/modules/highgui/test/test_positioning.cpp b/modules/highgui/test/test_positioning.cpp
index a13aab0f54..edc8dcf42a 100644
--- a/modules/highgui/test/test_positioning.cpp
+++ b/modules/highgui/test/test_positioning.cpp
@@ -41,7 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
-#include "opencv2/highgui.hpp"
+#include "opencv2/highgui/highgui_c.h"
 #include <stdio.h>
 
 using namespace cv;
@@ -88,7 +88,7 @@ CV_VideoRandomPositioningTest::~CV_VideoRandomPositioningTest() {}
 void CV_VideoPositioningTest::generate_idx_seq(CvCapture* cap, int method)
 {
     idx.clear();
-    int N = (int)cvGetCaptureProperty(cap, CV_CAP_PROP_FRAME_COUNT);
+    int N = (int)cvGetCaptureProperty(cap, CAP_PROP_FRAME_COUNT);
     switch(method)
     {
     case PROGRESSIVE:
@@ -147,7 +147,7 @@ void CV_VideoPositioningTest::run_test(int method)
             failed_videos++; continue;
         }
 
-        cvSetCaptureProperty(cap, CV_CAP_PROP_POS_FRAMES, 0);
+        cvSetCaptureProperty(cap, CAP_PROP_POS_FRAMES, 0);
 
         generate_idx_seq(cap, method);
 
@@ -157,7 +157,7 @@ void CV_VideoPositioningTest::run_test(int method)
         {
             bool flag = false;
 
-            cvSetCaptureProperty(cap, CV_CAP_PROP_POS_FRAMES, idx.at(j));
+            cvSetCaptureProperty(cap, CAP_PROP_POS_FRAMES, idx.at(j));
 
             /* IplImage* frame = cvRetrieveFrame(cap);
 
@@ -173,7 +173,7 @@ void CV_VideoPositioningTest::run_test(int method)
                 flag = !flag;
             } */
 
-            int val = (int)cvGetCaptureProperty(cap, CV_CAP_PROP_POS_FRAMES);
+            int val = (int)cvGetCaptureProperty(cap, CAP_PROP_POS_FRAMES);
 
             if (idx.at(j) != val)
             {
@@ -220,4 +220,4 @@ void CV_VideoRandomPositioningTest::run(int)
 #if BUILD_WITH_VIDEO_INPUT_SUPPORT && defined HAVE_FFMPEG
 TEST (Highgui_Video, seek_progressive) { CV_VideoProgressivePositioningTest test; test.safe_run(); }
 TEST (Highgui_Video, seek_random) { CV_VideoRandomPositioningTest test; test.safe_run(); }
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/highgui/test/test_precomp.cpp b/modules/highgui/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/highgui/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/highgui/test/test_precomp.hpp b/modules/highgui/test/test_precomp.hpp
index ba4241e25f..8468e46181 100644
--- a/modules/highgui/test/test_precomp.hpp
+++ b/modules/highgui/test/test_precomp.hpp
@@ -9,16 +9,15 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
-#ifdef HAVE_CVCONFIG_H
-# include "cvconfig.h"
-#endif
-
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include <iostream>
 
-#if defined(HAVE_VIDEOINPUT)   || \
+#include "opencv2/core/private.hpp"
+
+#if defined(HAVE_DSHOW)        || \
     defined(HAVE_TYZX)         || \
     defined(HAVE_VFW)          || \
     defined(HAVE_LIBV4L)       || \
@@ -29,12 +28,13 @@
     defined(HAVE_CMU1394)      || \
     defined(HAVE_MIL)          || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_UNICAP)       || \
     defined(HAVE_PVAPI)        || \
     defined(HAVE_OPENNI)       || \
     defined(HAVE_XIMEA)        || \
     defined(HAVE_AVFOUNDATION) || \
-    defined(HAVE_GIGE_API) || \
+    defined(HAVE_GIGE_API)     || \
     (0)
     //defined(HAVE_ANDROID_NATIVE_CAMERA) ||   - enable after #1193
 #  define BUILD_WITH_CAMERA_SUPPORT 1
@@ -45,11 +45,11 @@
 #if defined(HAVE_XINE)         || \
     defined(HAVE_GSTREAMER)    || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_AVFOUNDATION) || \
     /*defined(HAVE_OPENNI)     || too specialized */ \
     defined(HAVE_FFMPEG)       || \
-    defined(WIN32) /* assume that we have ffmpeg */
-
+    defined(HAVE_MSMF)
 #  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
 #else
 #  define BUILD_WITH_VIDEO_INPUT_SUPPORT 0
@@ -58,9 +58,10 @@
 #if /*defined(HAVE_XINE)       || */\
     defined(HAVE_GSTREAMER)    || \
     defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_QTKIT)        || \
     defined(HAVE_AVFOUNDATION) || \
     defined(HAVE_FFMPEG)       || \
-    defined(WIN32) /* assume that we have ffmpeg */
+    defined(HAVE_MSMF)
 #  define BUILD_WITH_VIDEO_OUTPUT_SUPPORT 1
 #else
 #  define BUILD_WITH_VIDEO_OUTPUT_SUPPORT 0
diff --git a/modules/highgui/test/test_video_io.cpp b/modules/highgui/test/test_video_io.cpp
index b91acf74ab..059d1e4fbd 100644
--- a/modules/highgui/test/test_video_io.cpp
+++ b/modules/highgui/test/test_video_io.cpp
@@ -41,7 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
-#include "opencv2/highgui.hpp"
+#include "opencv2/highgui/highgui_c.h"
 
 using namespace cv;
 using namespace std;
@@ -54,26 +54,55 @@ string fourccToString(int fourcc)
     return format("%c%c%c%c", fourcc & 255, (fourcc >> 8) & 255, (fourcc >> 16) & 255, (fourcc >> 24) & 255);
 }
 
+#ifdef HAVE_MSMF
 const VideoFormat g_specific_fmt_list[] =
 {
-    VideoFormat("avi", CV_FOURCC('X', 'V', 'I', 'D')),
-    VideoFormat("avi", CV_FOURCC('M', 'P', 'E', 'G')),
-    VideoFormat("avi", CV_FOURCC('M', 'J', 'P', 'G')),
-    //VideoFormat("avi", CV_FOURCC('I', 'Y', 'U', 'V')),
-    VideoFormat("mkv", CV_FOURCC('X', 'V', 'I', 'D')),
-    VideoFormat("mkv", CV_FOURCC('M', 'P', 'E', 'G')),
-    VideoFormat("mkv", CV_FOURCC('M', 'J', 'P', 'G')),
-
-    VideoFormat("mov", CV_FOURCC('m', 'p', '4', 'v')),
+        /*VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', '2', '5')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', '5', '0')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', 'c', ' ')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', 'h', '1')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', 'h', 'd')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', 's', 'd')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('d', 'v', 's', 'l')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('H', '2', '6', '3')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('M', '4', 'S', '2')),
+        VideoFormat("avi", CV_FOURCC_MACRO('M', 'J', 'P', 'G')),
+        VideoFormat("mp4", CV_FOURCC_MACRO('M', 'P', '4', 'S')),
+        VideoFormat("mp4", CV_FOURCC_MACRO('M', 'P', '4', 'V')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('M', 'P', '4', '3')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('M', 'P', 'G', '1')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('M', 'S', 'S', '1')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('M', 'S', 'S', '2')),*/
+#if !defined(_M_ARM)
+        VideoFormat("wmv", CV_FOURCC_MACRO('W', 'M', 'V', '1')),
+        VideoFormat("wmv", CV_FOURCC_MACRO('W', 'M', 'V', '2')),
+#endif
+        VideoFormat("wmv", CV_FOURCC_MACRO('W', 'M', 'V', '3')),
+        VideoFormat("avi", CV_FOURCC_MACRO('H', '2', '6', '4')),
+        //VideoFormat("wmv", CV_FOURCC_MACRO('W', 'V', 'C', '1')),
+        VideoFormat()
+};
+#else
+const VideoFormat g_specific_fmt_list[] =
+{
+    VideoFormat("avi", VideoWriter::fourcc('X', 'V', 'I', 'D')),
+    VideoFormat("avi", VideoWriter::fourcc('M', 'P', 'E', 'G')),
+    VideoFormat("avi", VideoWriter::fourcc('M', 'J', 'P', 'G')),
+    //VideoFormat("avi", VideoWriter::fourcc('I', 'Y', 'U', 'V')),
+    VideoFormat("mkv", VideoWriter::fourcc('X', 'V', 'I', 'D')),
+    VideoFormat("mkv", VideoWriter::fourcc('M', 'P', 'E', 'G')),
+    VideoFormat("mkv", VideoWriter::fourcc('M', 'J', 'P', 'G')),
+    VideoFormat("mov", VideoWriter::fourcc('m', 'p', '4', 'v')),
     VideoFormat()
 };
+#endif
 
 }
 
 class CV_HighGuiTest : public cvtest::BaseTest
 {
 protected:
-    void ImageTest(const string& dir);
+    void ImageTest (const string& dir);
     void VideoTest (const string& dir, const cvtest::VideoFormat& fmt);
     void SpecificImageTest (const string& dir);
     void SpecificVideoTest (const string& dir, const cvtest::VideoFormat& fmt);
@@ -242,19 +271,19 @@ void CV_HighGuiTest::VideoTest(const string& dir, const cvtest::VideoFormat& fmt
 
     for(;;)
     {
-        IplImage * img = cvQueryFrame( cap );
+        IplImage* img = cvQueryFrame( cap );
 
         if (!img)
             break;
 
-        frames.push_back(Mat(img).clone());
+        frames.push_back(cv::cvarrToMat(img, true));
 
-        if (writer == 0)
+        if (writer == NULL)
         {
             writer = cvCreateVideoWriter(tmp_name.c_str(), fmt.fourcc, 24, cvGetSize(img));
-            if (writer == 0)
+            if (writer == NULL)
             {
-                ts->printf(ts->LOG, "can't create writer (with fourcc : %d)\n",
+                ts->printf(ts->LOG, "can't create writer (with fourcc : %s)\n",
                            cvtest::fourccToString(fmt.fourcc).c_str());
                 cvReleaseCapture( &cap );
                 ts->set_failed_test_info(ts->FAIL_MISMATCH);
@@ -285,20 +314,27 @@ void CV_HighGuiTest::VideoTest(const string& dir, const cvtest::VideoFormat& fmt
             break;
 
         Mat img = frames[i];
-        Mat img1(ipl1);
+        Mat img1 = cv::cvarrToMat(ipl1);
 
         double psnr = PSNR(img1, img);
         if (psnr < thresDbell)
         {
-            printf("Too low psnr = %gdb\n", psnr);
-            // imwrite("img.png", img);
-            // imwrite("img1.png", img1);
+            ts->printf(ts->LOG, "Too low frame %d psnr = %gdb\n", i, psnr);
             ts->set_failed_test_info(ts->FAIL_MISMATCH);
+
+            //imwrite("original.png", img);
+            //imwrite("after_test.png", img1);
+            //Mat diff;
+            //absdiff(img, img1, diff);
+            //imwrite("diff.png", diff);
+
             break;
         }
     }
 
+    printf("Before saved release for %s\n", tmp_name.c_str());
     cvReleaseCapture( &saved );
+    printf("After release\n");
 
     ts->printf(ts->LOG, "end test function : ImagesVideo \n");
 }
@@ -416,7 +452,7 @@ void CV_HighGuiTest::SpecificVideoTest(const string& dir, const cvtest::VideoFor
     for( size_t i = 0; i < IMAGE_COUNT; ++i )
     {
         string file_path = format("%s../python/images/QCIF_%02d.bmp", dir.c_str(), i);
-        Mat img = imread(file_path, CV_LOAD_IMAGE_COLOR);
+        Mat img = imread(file_path, IMREAD_COLOR);
 
         if (img.empty())
         {
@@ -442,7 +478,7 @@ void CV_HighGuiTest::SpecificVideoTest(const string& dir, const cvtest::VideoFor
     writer.release();
     VideoCapture cap(video_file);
 
-    size_t FRAME_COUNT = (size_t)cap.get(CV_CAP_PROP_FRAME_COUNT);
+    size_t FRAME_COUNT = (size_t)cap.get(CAP_PROP_FRAME_COUNT);
 
     if (FRAME_COUNT != IMAGE_COUNT )
     {
diff --git a/modules/highgui/test/test_video_pos.cpp b/modules/highgui/test/test_video_pos.cpp
index a9fa36c250..37cf8039e0 100644
--- a/modules/highgui/test/test_video_pos.cpp
+++ b/modules/highgui/test/test_video_pos.cpp
@@ -110,9 +110,9 @@ public:
                 return;
             }
 
-            int N0 = (int)cap.get(CV_CAP_PROP_FRAME_COUNT);
-            cap.set(CV_CAP_PROP_POS_FRAMES, 0);
-            int N = (int)cap.get(CV_CAP_PROP_FRAME_COUNT);
+            int N0 = (int)cap.get(CAP_PROP_FRAME_COUNT);
+            cap.set(CAP_PROP_POS_FRAMES, 0);
+            int N = (int)cap.get(CAP_PROP_FRAME_COUNT);
 
             if (N != n_frames || N != N0)
             {
@@ -125,14 +125,14 @@ public:
             {
                 int idx = theRNG().uniform(0, N);
 
-                if( !cap.set(CV_CAP_PROP_POS_FRAMES, idx) )
+                if( !cap.set(CAP_PROP_POS_FRAMES, idx) )
                 {
                     ts->printf(ts->LOG, "\nError: cannot seek to frame %d.\n", idx);
                     ts->set_failed_test_info(ts->FAIL_INVALID_OUTPUT);
                     return;
                 }
 
-                int idx1 = (int)cap.get(CV_CAP_PROP_POS_FRAMES);
+                int idx1 = (int)cap.get(CAP_PROP_POS_FRAMES);
 
                 Mat img; cap >> img;
                 Mat img0 = drawFrame(idx);
diff --git a/modules/imgproc/doc/feature_detection.rst b/modules/imgproc/doc/feature_detection.rst
index 5098ba1352..a6d5817dd7 100644
--- a/modules/imgproc/doc/feature_detection.rst
+++ b/modules/imgproc/doc/feature_detection.rst
@@ -15,8 +15,6 @@ Finds edges in an image using the [Canny86]_ algorithm.
 
 .. ocv:cfunction:: void cvCanny( const CvArr* image, CvArr* edges, double threshold1, double threshold2, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.Canny(image, edges, threshold1, threshold2, aperture_size=3) -> None
-
     :param image: single-channel 8-bit input image.
 
     :param edges: output edge map; it has the same size and type as  ``image`` .
@@ -32,7 +30,11 @@ Finds edges in an image using the [Canny86]_ algorithm.
 The function finds edges in the input image ``image`` and marks them in the output map ``edges`` using the Canny algorithm. The smallest value between ``threshold1`` and ``threshold2`` is used for edge linking. The largest value is used to find initial segments of strong edges. See
 http://en.wikipedia.org/wiki/Canny_edge_detector
 
+.. note::
 
+   * An example on using the canny edge detector can be found at opencv_source_code/samples/cpp/edge.cpp
+
+   * (Python) An example on using the canny edge detector can be found at opencv_source_code/samples/cpp/edge.py
 
 cornerEigenValsAndVecs
 ----------------------
@@ -44,8 +46,6 @@ Calculates eigenvalues and eigenvectors of image blocks for corner detection.
 
 .. ocv:cfunction:: void cvCornerEigenValsAndVecs( const CvArr* image, CvArr* eigenvv, int block_size, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.CornerEigenValsAndVecs(image, eigenvv, blockSize, aperture_size=3) -> None
-
     :param src: Input single-channel 8-bit or floating-point image.
 
     :param dst: Image to store the results. It has the same size as  ``src``  and the type  ``CV_32FC(6)`` .
@@ -85,11 +85,13 @@ The output of the function can be used for robust edge or corner detection.
     :ocv:func:`cornerHarris`,
     :ocv:func:`preCornerDetect`
 
+.. note::
 
+   * (Python) An example on how to use eigenvectors and eigenvalues to estimate image texture flow direction can be found at opencv_source_code/samples/python2/texture_flow.py
 
 cornerHarris
 ------------
-Harris edge detector.
+Harris corner detector.
 
 .. ocv:function:: void cornerHarris( InputArray src, OutputArray dst, int blockSize, int ksize, double k, int borderType=BORDER_DEFAULT )
 
@@ -97,8 +99,6 @@ Harris edge detector.
 
 .. ocv:cfunction:: void cvCornerHarris( const CvArr* image, CvArr* harris_responce, int block_size, int aperture_size=3, double k=0.04 )
 
-.. ocv:pyoldfunction:: cv.CornerHarris(image, harris_dst, blockSize, aperture_size=3, k=0.04) -> None
-
     :param src: Input single-channel 8-bit or floating-point image.
 
     :param dst: Image to store the Harris detector responses. It has the type  ``CV_32FC1``  and the same size as  ``src`` .
@@ -111,7 +111,7 @@ Harris edge detector.
 
     :param borderType: Pixel extrapolation method. See  :ocv:func:`borderInterpolate` .
 
-The function runs the Harris edge detector on the image. Similarly to
+The function runs the Harris corner detector on the image. Similarly to
 :ocv:func:`cornerMinEigenVal` and
 :ocv:func:`cornerEigenValsAndVecs` , for each pixel
 :math:`(x, y)` it calculates a
@@ -137,8 +137,6 @@ Calculates the minimal eigenvalue of gradient matrices for corner detection.
 
 .. ocv:cfunction:: void cvCornerMinEigenVal( const CvArr* image, CvArr* eigenval, int block_size, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.CornerMinEigenVal(image, eigenval, blockSize, aperture_size=3) -> None
-
     :param src: Input single-channel 8-bit or floating-point image.
 
     :param dst: Image to store the minimal eigenvalues. It has the type  ``CV_32FC1``  and the same size as  ``src`` .
@@ -166,8 +164,6 @@ Refines the corner locations.
 
 .. ocv:cfunction:: void cvFindCornerSubPix( const CvArr* image, CvPoint2D32f* corners, int count, CvSize win, CvSize zero_zone, CvTermCriteria criteria )
 
-.. ocv:pyoldfunction:: cv.FindCornerSubPix(image, corners, win, zero_zone, criteria) -> corners
-
     :param image: Input image.
 
     :param corners: Initial coordinates of the input corners and refined coordinates provided for output.
@@ -228,8 +224,6 @@ Determines strong corners on an image.
 
 .. ocv:cfunction:: void cvGoodFeaturesToTrack( const CvArr* image, CvArr* eig_image, CvArr* temp_image, CvPoint2D32f* corners, int* corner_count, double quality_level, double min_distance, const CvArr* mask=NULL, int block_size=3, int use_harris=0, double k=0.04 )
 
-.. ocv:pyoldfunction:: cv.GoodFeaturesToTrack(image, eigImage, tempImage, cornerCount, qualityLevel, minDistance, mask=None, blockSize=3, useHarris=0, k=0.04) -> cornerCount
-
     :param image: Input 8-bit or floating-point 32-bit, single-channel image.
 
     :param eig_image: The parameter is ignored.
@@ -318,8 +312,8 @@ The function finds circles in a grayscale image using a modification of the Houg
 
 Example: ::
 
-    #include <cv.h>
-    #include <highgui.h>
+    #include <opencv2/imgproc.hpp>
+    #include <opencv2/highgui.hpp>
     #include <math.h>
 
     using namespace cv;
@@ -329,11 +323,11 @@ Example: ::
         Mat img, gray;
         if( argc != 2 && !(img=imread(argv[1], 1)).data)
             return -1;
-        cvtColor(img, gray, CV_BGR2GRAY);
+        cvtColor(img, gray, COLOR_BGR2GRAY);
         // smooth it, otherwise a lot of false circles may be detected
         GaussianBlur( gray, gray, Size(9, 9), 2, 2 );
         vector<Vec3f> circles;
-        HoughCircles(gray, circles, CV_HOUGH_GRADIENT,
+        HoughCircles(gray, circles, HOUGH_GRADIENT,
                      2, gray->rows/4, 200, 100 );
         for( size_t i = 0; i < circles.size(); i++ )
         {
@@ -356,6 +350,9 @@ Example: ::
     :ocv:func:`fitEllipse`,
     :ocv:func:`minEnclosingCircle`
 
+.. note::
+
+   * An example using the Hough circle detector can be found at opencv_source_code/samples/cpp/houghcircles.cpp
 
 HoughLines
 ----------
@@ -367,8 +364,6 @@ Finds lines in a binary image using the standard Hough transform.
 
 .. ocv:cfunction:: CvSeq* cvHoughLines2( CvArr* image, void* line_storage, int method, double rho, double theta, int threshold, double param1=0, double param2=0 )
 
-.. ocv:pyoldfunction:: cv.HoughLines2(image, storage, method, rho, theta, threshold, param1=0, param2=0)-> lines
-
     :param image: 8-bit, single-channel binary source image. The image may be modified by the function.
 
     :param lines: Output vector of lines. Each line is represented by a two-element vector  :math:`(\rho, \theta)` .  :math:`\rho`  is the distance from the coordinate origin  :math:`(0,0)`  (top-left corner of the image).  :math:`\theta`  is the line rotation angle in radians ( :math:`0 \sim \textrm{vertical line}, \pi/2 \sim \textrm{horizontal line}` ).
@@ -412,6 +407,10 @@ Finds lines in a binary image using the standard Hough transform.
 The function implements the standard or standard multi-scale Hough transform algorithm for line detection.  See http://homepages.inf.ed.ac.uk/rbf/HIPR2/hough.htm for a good explanation of Hough transform.
 See also the example in :ocv:func:`HoughLinesP` description.
 
+.. note::
+
+   * An example using the Hough line detector can be found at opencv_source_code/samples/cpp/houghlines.cpp
+
 HoughLinesP
 -----------
 Finds line segments in a binary image using the probabilistic Hough transform.
@@ -440,9 +439,8 @@ The function implements the probabilistic Hough transform algorithm for line det
     /* This is a standalone program. Pass an image name as the first parameter
     of the program.  Switch between standard and probabilistic Hough transform
     by changing "#if 1" to "#if 0" and back */
-    #include <cv.h>
-    #include <highgui.h>
-    #include <math.h>
+    #include <opencv2/imgproc.hpp>
+    #include <opencv2/highgui.hpp>
 
     using namespace cv;
 
@@ -453,7 +451,7 @@ The function implements the probabilistic Hough transform algorithm for line det
             return -1;
 
         Canny( src, dst, 50, 200, 3 );
-        cvtColor( dst, color_dst, CV_GRAY2BGR );
+        cvtColor( dst, color_dst, COLOR_GRAY2BGR );
 
     #if 0
         vector<Vec2f> lines;
@@ -498,6 +496,110 @@ And this is the output of the above program in case of the probabilistic Hough t
 
 .. image:: pics/houghp.png
 
+.. seealso::
+
+    :ocv:class:`LineSegmentDetector`
+
+
+
+LineSegmentDetector
+-------------------
+Line segment detector class, following the algorithm described at [Rafael12]_.
+
+.. ocv:class:: LineSegmentDetector : public Algorithm
+
+
+createLineSegmentDetectorPtr
+----------------------------
+Creates a smart pointer to a LineSegmentDetector object and initializes it.
+
+.. ocv:function:: Ptr<LineSegmentDetector> createLineSegmentDetectorPtr(int _refine = LSD_REFINE_STD, double _scale = 0.8, double _sigma_scale = 0.6, double _quant = 2.0, double _ang_th = 22.5, double _log_eps = 0, double _density_th = 0.7, int _n_bins = 1024)
+
+    :param _refine: The way found lines will be refined:
+
+        * **LSD_REFINE_NONE** - No refinement applied.
+
+        * **LSD_REFINE_STD**  - Standard refinement is applied. E.g. breaking arches into smaller straighter line approximations.
+
+        * **LSD_REFINE_ADV**  - Advanced refinement. Number of false alarms is calculated, lines are refined through increase of precision, decrement in size, etc.
+
+    :param scale: The scale of the image that will be used to find the lines. Range (0..1].
+
+    :param sigma_scale: Sigma for Gaussian filter. It is computed as sigma = _sigma_scale/_scale.
+
+    :param quant: Bound to the quantization error on the gradient norm.
+
+    :param ang_th: Gradient angle tolerance in degrees.
+
+    :param log_eps: Detection threshold: -log10(NFA) > log_eps. Used only when advancent refinement is chosen.
+
+    :param density_th: Minimal density of aligned region points in the enclosing rectangle.
+
+    :param n_bins: Number of bins in pseudo-ordering of gradient modulus.
+
+The LineSegmentDetector algorithm is defined using the standard values. Only advanced users may want to edit those, as to tailor it for their own application.
+
+
+LineSegmentDetector::detect
+---------------------------
+Finds lines in the input image. See the lsd_lines.cpp sample for possible usage.
+
+.. ocv:function:: void LineSegmentDetector::detect(const InputArray _image, OutputArray _lines, OutputArray width = noArray(), OutputArray prec = noArray(), OutputArray nfa = noArray())
+
+    :param _image A grayscale (CV_8UC1) input image.
+        If only a roi needs to be selected, use ::
+        lsd_ptr->detect(image(roi), lines, ...);
+        lines += Scalar(roi.x, roi.y, roi.x, roi.y);
+
+    :param lines: A vector of Vec4i elements specifying the beginning and ending point of a line. Where Vec4i is (x1, y1, x2, y2), point 1 is the start, point 2 - end. Returned lines are strictly oriented depending on the gradient.
+
+    :param width: Vector of widths of the regions, where the lines are found. E.g. Width of line.
+
+    :param prec: Vector of precisions with which the lines are found.
+
+    :param nfa: Vector containing number of false alarms in the line region, with precision of 10%. The bigger the value, logarithmically better the detection.
+
+        * -1 corresponds to 10 mean false alarms
+
+        * 0 corresponds to 1 mean false alarm
+
+        * 1 corresponds to 0.1 mean false alarms
+
+    This vector will be calculated only when the objects type is LSD_REFINE_ADV.
+
+This is the output of the default parameters of the algorithm on the above shown image.
+
+.. image:: pics/building_lsd.png
+
+.. note::
+
+   * An example using the LineSegmentDetector can be found at opencv_source_code/samples/cpp/lsd_lines.cpp
+
+LineSegmentDetector::drawSegments
+---------------------------------
+Draws the line segments on a given image.
+
+.. ocv:function:: void LineSegmentDetector::drawSegments(InputOutputArray _image, InputArray lines)
+
+    :param image: The image, where the liens will be drawn. Should be bigger or equal to the image, where the lines were found.
+
+    :param lines: A vector of the lines that needed to be drawn.
+
+
+LineSegmentDetector::compareSegments
+------------------------------------
+Draws two groups of lines in blue and red, counting the non overlapping (mismatching) pixels.
+
+.. ocv:function:: int LineSegmentDetector::compareSegments(const Size& size, InputArray lines1, InputArray lines2, InputOutputArray _image = noArray())
+
+    :param size: The size of the image, where lines1 and lines2 were found.
+
+    :param lines1: The first group of lines that needs to be drawn. It is visualized in blue color.
+
+    :param lines2: The second group of lines. They visualized in red color.
+
+    :param image: Optional image, where the lines will be drawn. The image should be color in order for lines1 and lines2 to be drawn in the above mentioned colors.
+
 
 
 preCornerDetect
@@ -510,8 +612,6 @@ Calculates a feature map for corner detection.
 
 .. ocv:cfunction:: void cvPreCornerDetect( const CvArr* image, CvArr* corners, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.PreCornerDetect(image, corners, apertureSize=3)-> None
-
     :param src: Source single-channel 8-bit of floating-point image.
 
     :param dst: Output image that has the type  ``CV_32F``  and the same size as  ``src`` .
@@ -546,3 +646,5 @@ The corners can be found as local maximums of the functions, as shown below: ::
 .. [Shi94] J. Shi and C. Tomasi. *Good Features to Track*. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 593-600, June 1994.
 
 .. [Yuen90] Yuen, H. K. and Princen, J. and Illingworth, J. and Kittler, J., *Comparative study of Hough transform methods for circle finding*. Image Vision Comput. 8 1, pp 71–77 (1990)
+
+.. [Rafael12] Rafael Grompone von Gioi, Jérémie Jakubowicz, Jean-Michel Morel, and Gregory Randall, LSD: a Line Segment Detector, Image Processing On Line, vol. 2012. http://dx.doi.org/10.5201/ipol.2012.gjmr-lsd
diff --git a/modules/imgproc/doc/filtering.rst b/modules/imgproc/doc/filtering.rst
index 0b1f9a34e4..d019136fab 100755
--- a/modules/imgproc/doc/filtering.rst
+++ b/modules/imgproc/doc/filtering.rst
@@ -22,6 +22,10 @@ OpenCV enables you to specify the extrapolation method. For details, see the fun
     * BORDER_CONSTANT:      iiiiii|abcdefgh|iiiiiii  with some specified 'i'
     */
 
+.. note::
+
+   * (Python) A complete example illustrating different morphological operations like erode/dilate, open/close, blackhat/tophat ... can be found at opencv_source_code/samples/python2/morphology.py
+
 BaseColumnFilter
 ----------------
 .. ocv:class:: BaseColumnFilter
@@ -408,6 +412,28 @@ http://www.dai.ed.ac.uk/CVonline/LOCAL\_COPIES/MANDUCHI1/Bilateral\_Filtering.ht
 This filter does not work inplace.
 
 
+adaptiveBilateralFilter
+-----------------------
+Applies the adaptive bilateral filter to an image.
+
+.. ocv:function:: void adaptiveBilateralFilter( InputArray src, OutputArray dst, Size ksize, double sigmaSpace, Point anchor=Point(-1, -1), int borderType=BORDER_DEFAULT )
+
+.. ocv:pyfunction:: cv2.adaptiveBilateralFilter(src, ksize, sigmaSpace[, dst[, anchor[, borderType]]]) -> dst
+
+    :param src: Source 8-bit, 1-channel or 3-channel image.
+
+    :param dst: Destination image of the same size and type as  ``src`` .
+
+    :param ksize: filter kernel size.
+
+    :param sigmaSpace: Filter sigma in the coordinate space. It has similar meaning with ``sigmaSpace`` in ``bilateralFilter``.
+
+    :param anchor: anchor point; default value ``Point(-1,-1)`` means that the anchor is at the kernel center. Only default value is supported now.
+
+    :param borderType: border mode used to extrapolate pixels outside of the image.
+
+The function applies adaptive bilateral filtering to the input image. This filter is similar to ``bilateralFilter``, in that dissimilarity from and distance to the center pixel is punished. Instead of using ``sigmaColor``, we employ the variance of pixel values in the neighbourhood.
+
 
 
 blur
@@ -444,37 +470,6 @@ The call ``blur(src, dst, ksize, anchor, borderType)`` is equivalent to ``boxFil
    :ocv:func:`medianBlur`
 
 
-borderInterpolate
------------------
-Computes the source location of an extrapolated pixel.
-
-.. ocv:function:: int borderInterpolate( int p, int len, int borderType )
-
-.. ocv:pyfunction:: cv2.borderInterpolate(p, len, borderType) -> retval
-
-    :param p: 0-based coordinate of the extrapolated pixel along one of the axes, likely <0 or >= ``len`` .
-
-    :param len: Length of the array along the corresponding axis.
-
-    :param borderType: Border type, one of the  ``BORDER_*`` , except for  ``BORDER_TRANSPARENT``  and  ``BORDER_ISOLATED`` . When  ``borderType==BORDER_CONSTANT`` , the function always returns -1, regardless of  ``p``  and  ``len`` .
-
-The function computes and returns the coordinate of a donor pixel corresponding to the specified extrapolated pixel when using the specified extrapolation border mode. For example, if you use ``BORDER_WRAP`` mode in the horizontal direction, ``BORDER_REFLECT_101`` in the vertical direction and want to compute value of the "virtual" pixel ``Point(-5, 100)`` in a floating-point image ``img`` , it looks like: ::
-
-    float val = img.at<float>(borderInterpolate(100, img.rows, BORDER_REFLECT_101),
-                              borderInterpolate(-5, img.cols, BORDER_WRAP));
-
-
-Normally, the function is not called directly. It is used inside
-:ocv:class:`FilterEngine` and
-:ocv:func:`copyMakeBorder` to compute tables for quick extrapolation.
-
-.. seealso::
-
-    :ocv:class:`FilterEngine`,
-    :ocv:func:`copyMakeBorder`
-
-
-
 boxFilter
 ---------
 Blurs an image using the box filter.
@@ -537,63 +532,6 @@ The function constructs a vector of images and builds the Gaussian pyramid by re
 :ocv:func:`pyrDown` to the previously built pyramid layers, starting from ``dst[0]==src`` .
 
 
-
-copyMakeBorder
---------------
-Forms a border around an image.
-
-.. ocv:function:: void copyMakeBorder( InputArray src, OutputArray dst, int top, int bottom, int left, int right, int borderType, const Scalar& value=Scalar() )
-
-.. ocv:pyfunction:: cv2.copyMakeBorder(src, top, bottom, left, right, borderType[, dst[, value]]) -> dst
-
-.. ocv:cfunction:: void cvCopyMakeBorder( const CvArr* src, CvArr* dst, CvPoint offset, int bordertype, CvScalar value=cvScalarAll(0) )
-.. ocv:pyoldfunction:: cv.CopyMakeBorder(src, dst, offset, bordertype, value=(0, 0, 0, 0))-> None
-
-    :param src: Source image.
-
-    :param dst: Destination image of the same type as  ``src``  and the size  ``Size(src.cols+left+right, src.rows+top+bottom)`` .
-
-    :param top:
-
-    :param bottom:
-
-    :param left:
-
-    :param right: Parameter specifying how many pixels in each direction from the source image rectangle to extrapolate. For example,  ``top=1, bottom=1, left=1, right=1``  mean that 1 pixel-wide border needs to be built.
-
-    :param borderType: Border type. See  :ocv:func:`borderInterpolate` for details.
-
-    :param value: Border value if  ``borderType==BORDER_CONSTANT`` .
-
-The function copies the source image into the middle of the destination image. The areas to the left, to the right, above and below the copied source image will be filled with extrapolated pixels. This is not what
-:ocv:class:`FilterEngine` or filtering functions based on it do (they extrapolate pixels on-fly), but what other more complex functions, including your own, may do to simplify image boundary handling.
-
-The function supports the mode when ``src`` is already in the middle of ``dst`` . In this case, the function does not copy ``src`` itself but simply constructs the border, for example: ::
-
-    // let border be the same in all directions
-    int border=2;
-    // constructs a larger image to fit both the image and the border
-    Mat gray_buf(rgb.rows + border*2, rgb.cols + border*2, rgb.depth());
-    // select the middle part of it w/o copying data
-    Mat gray(gray_canvas, Rect(border, border, rgb.cols, rgb.rows));
-    // convert image from RGB to grayscale
-    cvtColor(rgb, gray, CV_RGB2GRAY);
-    // form a border in-place
-    copyMakeBorder(gray, gray_buf, border, border,
-                   border, border, BORDER_REPLICATE);
-    // now do some custom filtering ...
-    ...
-
-
-.. note::
-
-    When the source image is a part (ROI) of a bigger image, the function will try to use the pixels outside of the ROI to form a border. To disable this feature and always do extrapolation, as if ``src`` was not a ROI, use ``borderType | BORDER_ISOLATED``.
-
-.. seealso::
-
-    :ocv:func:`borderInterpolate`
-
-
 createBoxFilter
 -------------------
 Returns a box filter engine.
@@ -838,13 +776,12 @@ Dilates an image by using a specific structuring element.
 .. ocv:pyfunction:: cv2.dilate(src, kernel[, dst[, anchor[, iterations[, borderType[, borderValue]]]]]) -> dst
 
 .. ocv:cfunction:: void cvDilate( const CvArr* src, CvArr* dst, IplConvKernel* element=NULL, int iterations=1 )
-.. ocv:pyoldfunction:: cv.Dilate(src, dst, element=None, iterations=1)-> None
 
     :param src: input image; the number of channels can be arbitrary, but the depth should be one of ``CV_8U``, ``CV_16U``, ``CV_16S``,  ``CV_32F` or ``CV_64F``.
 
     :param dst: output image of the same size and type as ``src``.
 
-    :param element: structuring element used for dilation; if  ``element=Mat()`` , a  ``3 x 3`` rectangular structuring element is used.
+    :param kernel: structuring element used for dilation; if  ``element=Mat()`` , a  ``3 x 3`` rectangular structuring element is used. Kernel can be created using :ocv:func:`getStructuringElement`
 
     :param anchor: position of the anchor within the element; default value ``(-1, -1)`` means that the anchor is at the element center.
 
@@ -867,6 +804,14 @@ The function supports the in-place mode. Dilation can be applied several ( ``ite
     :ocv:func:`erode`,
     :ocv:func:`morphologyEx`,
     :ocv:func:`createMorphologyFilter`
+    :ocv:func:`getStructuringElement`
+
+
+.. note::
+
+   * An example using the morphological dilate operation can be found at opencv_source_code/samples/cpp/morphology2.cpp
+
+
 
 
 erode
@@ -878,13 +823,12 @@ Erodes an image by using a specific structuring element.
 .. ocv:pyfunction:: cv2.erode(src, kernel[, dst[, anchor[, iterations[, borderType[, borderValue]]]]]) -> dst
 
 .. ocv:cfunction:: void cvErode( const CvArr* src, CvArr* dst, IplConvKernel* element=NULL, int iterations=1)
-.. ocv:pyoldfunction:: cv.Erode(src, dst, element=None, iterations=1)-> None
 
     :param src: input image; the number of channels can be arbitrary, but the depth should be one of ``CV_8U``, ``CV_16U``, ``CV_16S``,  ``CV_32F` or ``CV_64F``.
 
     :param dst: output image of the same size and type as ``src``.
 
-    :param element: structuring element used for erosion; if  ``element=Mat()`` , a  ``3 x 3``  rectangular structuring element is used.
+    :param kernel: structuring element used for erosion; if  ``element=Mat()`` , a  ``3 x 3``  rectangular structuring element is used. Kernel can be created using :ocv:func:`getStructuringElement`.
 
     :param anchor: position of the anchor within the element; default value  ``(-1, -1)``  means that the anchor is at the element center.
 
@@ -906,9 +850,12 @@ The function supports the in-place mode. Erosion can be applied several ( ``iter
 
     :ocv:func:`dilate`,
     :ocv:func:`morphologyEx`,
-    :ocv:func:`createMorphologyFilter`
+    :ocv:func:`createMorphologyFilter`,
+    :ocv:func:`getStructuringElement`
 
+.. note::
 
+   * An example using the morphological erode operation can be found at opencv_source_code/samples/cpp/morphology2.cpp
 
 filter2D
 --------
@@ -920,8 +867,6 @@ Convolves an image with the kernel.
 
 .. ocv:cfunction:: void cvFilter2D( const CvArr* src, CvArr* dst, const CvMat* kernel, CvPoint anchor=cvPoint(-1,-1) )
 
-.. ocv:pyoldfunction:: cv.Filter2D(src, dst, kernel, anchor=(-1, -1))-> None
-
     :param src: input image.
 
     :param dst: output image of the same size and the same number of channels as ``src``.
@@ -1039,7 +984,7 @@ Returns Gaussian filter coefficients.
     :param ksize: Aperture size. It should be odd ( :math:`\texttt{ksize} \mod 2 = 1` ) and positive.
 
     :param sigma: Gaussian standard deviation. If it is non-positive, it is computed from  ``ksize``  as  \ ``sigma = 0.3*((ksize-1)*0.5 - 1) + 0.8`` .
-    :param ktype: Type of filter coefficients. It can be  ``CV_32f``  or  ``CV_64F`` .
+    :param ktype: Type of filter coefficients. It can be  ``CV_32F``  or  ``CV_64F`` .
 
 The function computes and returns the
 :math:`\texttt{ksize} \times 1` matrix of Gaussian filter coefficients:
@@ -1068,6 +1013,32 @@ Two of such generated kernels can be passed to
 
 
 
+getGaborKernel
+-----------------
+Returns Gabor filter coefficients.
+
+.. ocv:function:: Mat getGaborKernel( Size ksize, double sigma, double theta, double lambd, double gamma, double psi = CV_PI*0.5, int ktype = CV_64F )
+
+.. ocv:pyfunction:: cv2.getGaborKernel(ksize, sigma, theta, lambd, gamma[, psi[, ktype]]) -> retval
+
+    :param ksize: Size of the filter returned.
+
+    :param sigma: Standard deviation of the gaussian envelope.
+
+    :param theta: Orientation of the normal to the parallel stripes of a Gabor function.
+
+    :param lambd: Wavelength of the sinusoidal factor.
+
+    :param gamma: Spatial aspect ratio.
+
+    :param psi: Phase offset.
+
+    :param ktype: Type of filter coefficients. It can be  ``CV_32F``  or  ``CV_64F`` .
+
+For more details about gabor filter equations and parameters, see: `Gabor Filter <http://en.wikipedia.org/wiki/Gabor_filter>`_.
+
+
+
 getKernelType
 -------------
 Returns the kernel type.
@@ -1101,8 +1072,6 @@ Returns a structuring element of the specified size and shape for morphological
 
 .. ocv:cfunction:: IplConvKernel* cvCreateStructuringElementEx( int cols, int rows, int anchor_x, int anchor_y, int shape, int* values=NULL )
 
-.. ocv:pyoldfunction:: cv.CreateStructuringElementEx(cols, rows, anchorX, anchorY, shape, values=None)-> kernel
-
     :param shape: Element shape that could be one of the following:
 
       * **MORPH_RECT**         - a rectangular structuring element:
@@ -1179,13 +1148,14 @@ Performs advanced morphological transformations.
 .. ocv:pyfunction:: cv2.morphologyEx(src, op, kernel[, dst[, anchor[, iterations[, borderType[, borderValue]]]]]) -> dst
 
 .. ocv:cfunction:: void cvMorphologyEx( const CvArr* src, CvArr* dst, CvArr* temp, IplConvKernel* element, int operation, int iterations=1 )
-.. ocv:pyoldfunction:: cv.MorphologyEx(src, dst, temp, element, operation, iterations=1)-> None
 
     :param src: Source image. The number of channels can be arbitrary. The depth should be one of ``CV_8U``, ``CV_16U``, ``CV_16S``,  ``CV_32F` or ``CV_64F``.
 
     :param dst: Destination image of the same size and type as  ``src`` .
 
-    :param element: Structuring element.
+    :param kernel: Structuring element. It can be created using :ocv:func:`getStructuringElement`.
+
+    :param anchor: Anchor position with the kernel. Negative values mean that the anchor is at the kernel center.
 
     :param op: Type of a morphological operation that can be one of the following:
 
@@ -1243,8 +1213,12 @@ Any of the operations can be done in-place. In case of multi-channel images, eac
 
     :ocv:func:`dilate`,
     :ocv:func:`erode`,
-    :ocv:func:`createMorphologyFilter`
+    :ocv:func:`createMorphologyFilter`,
+    :ocv:func:`getStructuringElement`
 
+.. note::
+
+   * An example using the morphologyEx function for the morphological opening and closing operations can be found at opencv_source_code/samples/cpp/morphology2.cpp
 
 Laplacian
 ---------
@@ -1256,8 +1230,6 @@ Calculates the Laplacian of an image.
 
 .. ocv:cfunction:: void cvLaplace( const CvArr* src, CvArr* dst, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.Laplace(src, dst, apertureSize=3) -> None
-
     :param src: Source image.
 
     :param dst: Destination image of the same size and the same number of channels as  ``src`` .
@@ -1290,7 +1262,9 @@ This is done when ``ksize > 1`` . When ``ksize == 1`` , the Laplacian is compute
     :ocv:func:`Sobel`,
     :ocv:func:`Scharr`
 
+.. note::
 
+   * An example using the Laplace transformation for edge detection can be found at opencv_source_code/samples/cpp/laplace.cpp
 
 pyrDown
 -------
@@ -1302,8 +1276,6 @@ Blurs an image and downsamples it.
 
 .. ocv:cfunction:: void cvPyrDown( const CvArr* src, CvArr* dst, int filter=CV_GAUSSIAN_5x5 )
 
-.. ocv:pyoldfunction:: cv.PyrDown(src, dst, filter=CV_GAUSSIAN_5X5) -> None
-
     :param src: input image.
 
     :param dst: output image; it has the specified size and the same type as ``src``.
@@ -1335,8 +1307,6 @@ Upsamples an image and then blurs it.
 
 .. ocv:cfunction:: cvPyrUp( const CvArr* src, CvArr* dst, int filter=CV_GAUSSIAN_5x5 )
 
-.. ocv:pyoldfunction:: cv.PyrUp(src, dst, filter=CV_GAUSSIAN_5X5) -> None
-
     :param src: input image.
 
     :param dst: output image. It has the specified size and the same type as  ``src`` .
@@ -1351,19 +1321,21 @@ Upsamples an image and then blurs it.
 The function performs the upsampling step of the Gaussian pyramid construction, though it can actually be used to construct the Laplacian pyramid. First, it upsamples the source image by injecting even zero rows and columns and then convolves the result with the same kernel as in
 :ocv:func:`pyrDown`  multiplied by 4.
 
+.. note::
+
+   * (Python) An example of Laplacian Pyramid construction and merging can be found at opencv_source_code/samples/python2/lappyr.py
+
 
 pyrMeanShiftFiltering
 ---------------------
 Performs initial step of meanshift segmentation of an image.
 
-.. ocv:function:: void pyrMeanShiftFiltering( InputArray src, OutputArray dst, double sp, double sr, int maxLevel=1, TermCriteria termcrit=TermCriteria( TermCriteria::MAX_ITER+TermCriteria::EPS,5,1) )
+.. ocv:function:: void pyrMeanShiftFiltering( InputArray src, OutputArray dst, double sp, double sr, int maxLevel=1, TermCriteria termcrit=TermCriteria(TermCriteria::MAX_ITER+TermCriteria::EPS,5,1) )
 
 .. ocv:pyfunction:: cv2.pyrMeanShiftFiltering(src, sp, sr[, dst[, maxLevel[, termcrit]]]) -> dst
 
 .. ocv:cfunction:: void cvPyrMeanShiftFiltering( const CvArr* src, CvArr* dst, double sp,  double sr,  int max_level=1, CvTermCriteria termcrit= cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,5,1))
 
-.. ocv:pyoldfunction:: cv.PyrMeanShiftFiltering(src, dst, sp, sr, max_level=1, termcrit=(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 5, 1)) -> None
-
     :param src: The source 8-bit, 3-channel image.
 
     :param dst: The destination image of the same format and the same size as the source.
@@ -1400,6 +1372,9 @@ After the iterations over, the color components of the initial pixel (that is, t
 
 When ``maxLevel > 0``, the gaussian pyramid of ``maxLevel+1`` levels is built, and the above procedure is run on the smallest layer first. After that, the results are propagated to the larger layer and the iterations are run again only on those pixels where the layer colors differ by more than ``sr`` from the lower-resolution layer of the pyramid. That makes boundaries of color regions sharper. Note that the results will be actually different from the ones obtained by running the meanshift procedure on the whole original image (i.e. when ``maxLevel==0``).
 
+.. note::
+
+   * An example using mean-shift image segmentation can be found at opencv_source_code/samples/cpp/meanshift_segmentation.cpp
 
 sepFilter2D
 -----------
@@ -1425,7 +1400,7 @@ Applies a separable linear filter to an image.
 
     :param kernelY: Coefficients for filtering each column.
 
-    :param anchor: Anchor position within the kernel. The default value  :math:`(-1, 1)`  means that the anchor is at the kernel center.
+    :param anchor: Anchor position within the kernel. The default value  :math:`(-1,-1)`  means that the anchor is at the kernel center.
 
     :param delta: Value added to the filtered results before storing them.
 
@@ -1449,8 +1424,6 @@ Smooths the image in one of several ways.
 
 .. ocv:cfunction:: void cvSmooth( const CvArr* src, CvArr* dst, int smoothtype=CV_GAUSSIAN, int size1=3, int size2=0, double sigma1=0, double sigma2=0 )
 
-.. ocv:pyoldfunction:: cv.Smooth(src, dst, smoothtype=CV_GAUSSIAN, param1=3, param2=0, param3=0, param4=0)-> None
-
     :param src: The source image
 
     :param dst: The destination image
@@ -1505,8 +1478,6 @@ Calculates the first, second, third, or mixed image derivatives using an extende
 
 .. ocv:cfunction:: void cvSobel( const CvArr* src, CvArr* dst, int xorder, int yorder, int aperture_size=3 )
 
-.. ocv:pyoldfunction:: cv.Sobel(src, dst, xorder, yorder, apertureSize=3)-> None
-
     :param src: input image.
 
     :param dst: output image of the same size and the same number of channels as  ``src`` .
@@ -1621,4 +1592,3 @@ is equivalent to
 .. seealso::
 
     :ocv:func:`cartToPolar`
-
diff --git a/modules/imgproc/doc/geometric_transformations.rst b/modules/imgproc/doc/geometric_transformations.rst
index c48da01d7e..c94178f6cc 100644
--- a/modules/imgproc/doc/geometric_transformations.rst
+++ b/modules/imgproc/doc/geometric_transformations.rst
@@ -85,8 +85,6 @@ Calculates an affine transform from three pairs of the corresponding points.
 
 .. ocv:cfunction:: CvMat* cvGetAffineTransform( const CvPoint2D32f * src, const CvPoint2D32f * dst, CvMat * map_matrix )
 
-.. ocv:pyoldfunction:: cv.GetAffineTransform(src, dst, mapMatrix)-> None
-
     :param src: Coordinates of triangle vertices in the source image.
 
     :param dst: Coordinates of the corresponding triangle vertices in the destination image.
@@ -124,8 +122,6 @@ Calculates a perspective transform from four pairs of the corresponding points.
 
 .. ocv:cfunction:: CvMat* cvGetPerspectiveTransform( const CvPoint2D32f* src, const CvPoint2D32f* dst, CvMat* map_matrix )
 
-.. ocv:pyoldfunction:: cv.GetPerspectiveTransform(src, dst, mapMatrix)-> None
-
     :param src: Coordinates of quadrangle vertices in the source image.
 
     :param dst: Coordinates of the corresponding quadrangle vertices in the destination image.
@@ -160,7 +156,6 @@ Retrieves a pixel rectangle from an image with sub-pixel accuracy.
 .. ocv:pyfunction:: cv2.getRectSubPix(image, patchSize, center[, patch[, patchType]]) -> patch
 
 .. ocv:cfunction:: void cvGetRectSubPix( const CvArr* src, CvArr* dst, CvPoint2D32f center )
-.. ocv:pyoldfunction:: cv.GetRectSubPix(src, dst, center)-> None
 
     :param src: Source image.
 
@@ -202,8 +197,6 @@ Calculates an affine matrix of 2D rotation.
 
 .. ocv:cfunction:: CvMat* cv2DRotationMatrix( CvPoint2D32f center, double angle, double scale, CvMat* map_matrix )
 
-.. ocv:pyoldfunction:: cv.GetRotationMatrix2D(center, angle, scale, mapMatrix)-> None
-
     :param center: Center of the rotation in the source image.
 
     :param angle: Rotation angle in degrees. Positive values mean counter-clockwise rotation (the coordinate origin is assumed to be the top-left corner).
@@ -264,8 +257,6 @@ Remaps an image to log-polar space.
 
 .. ocv:cfunction:: void cvLogPolar( const CvArr* src, CvArr* dst, CvPoint2D32f center, double M, int flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS )
 
-.. ocv:pyoldfunction:: cv.LogPolar(src, dst, center, M, flags=CV_INNER_LINEAR+CV_WARP_FILL_OUTLIERS)-> None
-
     :param src: Source image
 
     :param dst: Destination image
@@ -307,6 +298,9 @@ where
 
 The function emulates the human "foveal" vision and can be used for fast scale and rotation-invariant template matching, for object tracking and so forth. The function can not operate in-place.
 
+.. note::
+
+   * An example using the geometric logpolar operation in 4 applications can be found at opencv_source_code/samples/cpp/logpolar_bsm.cpp
 
 remap
 -----
@@ -317,7 +311,6 @@ Applies a generic geometrical transformation to an image.
 .. ocv:pyfunction:: cv2.remap(src, map1, map2, interpolation[, dst[, borderMode[, borderValue]]]) -> dst
 
 .. ocv:cfunction:: void cvRemap( const CvArr* src, CvArr* dst, const CvArr* mapx, const CvArr* mapy, int flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, CvScalar fillval=cvScalarAll(0) )
-.. ocv:pyoldfunction:: cv.Remap(src, dst, mapx, mapy, flags=CV_INNER_LINEAR+CV_WARP_FILL_OUTLIERS, fillval=(0, 0, 0, 0))-> None
 
     :param src: Source image.
 
@@ -364,7 +357,6 @@ Resizes an image.
 .. ocv:pyfunction:: cv2.resize(src, dsize[, dst[, fx[, fy[, interpolation]]]]) -> dst
 
 .. ocv:cfunction:: void cvResize( const CvArr* src, CvArr* dst, int interpolation=CV_INTER_LINEAR )
-.. ocv:pyoldfunction:: cv.Resize(src, dst, interpolation=CV_INTER_LINEAR)-> None
 
     :param src: input image.
 
@@ -434,12 +426,8 @@ Applies an affine transformation to an image.
 
 .. ocv:cfunction:: void cvWarpAffine( const CvArr* src, CvArr* dst, const CvMat* map_matrix, int flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, CvScalar fillval=cvScalarAll(0) )
 
-.. ocv:pyoldfunction:: cv.WarpAffine(src, dst, mapMatrix, flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, fillval=(0, 0, 0, 0))-> None
-
 .. ocv:cfunction:: void cvGetQuadrangleSubPix( const CvArr* src, CvArr* dst, const CvMat* map_matrix )
 
-.. ocv:pyoldfunction:: cv.GetQuadrangleSubPix(src, dst, mapMatrix)-> None
-
     :param src: input image.
 
     :param dst: output image that has the size  ``dsize``  and the same type as  ``src`` .
@@ -485,8 +473,6 @@ Applies a perspective transformation to an image.
 
 .. ocv:cfunction:: void cvWarpPerspective( const CvArr* src, CvArr* dst, const CvMat* map_matrix, int flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, CvScalar fillval=cvScalarAll(0) )
 
-.. ocv:pyoldfunction:: cv.WarpPerspective(src, dst, mapMatrix, flags=CV_INNER_LINEAR+CV_WARP_FILL_OUTLIERS, fillval=(0, 0, 0, 0))-> None
-
     :param src: input image.
 
     :param dst: output image that has the size  ``dsize``  and the same type as  ``src`` .
@@ -534,9 +520,6 @@ Computes the undistortion and rectification transformation map.
 .. ocv:cfunction:: void cvInitUndistortRectifyMap( const CvMat* camera_matrix, const CvMat* dist_coeffs, const CvMat * R, const CvMat* new_camera_matrix, CvArr* mapx, CvArr* mapy )
 .. ocv:cfunction:: void cvInitUndistortMap( const CvMat* camera_matrix, const CvMat* distortion_coeffs, CvArr* mapx, CvArr* mapy )
 
-.. ocv:pyoldfunction:: cv.InitUndistortRectifyMap(cameraMatrix, distCoeffs, R, newCameraMatrix, map1, map2)-> None
-.. ocv:pyoldfunction:: cv.InitUndistortMap(cameraMatrix, distCoeffs, map1, map2)-> None
-
     :param cameraMatrix: Input camera matrix  :math:`A=\vecthreethree{f_x}{0}{c_x}{0}{f_y}{c_y}{0}{0}{1}` .
 
     :param distCoeffs: Input vector of distortion coefficients  :math:`(k_1, k_2, p_1, p_2[, k_3[, k_4, k_5, k_6]])`  of 4, 5, or 8 elements. If the vector is NULL/empty, the zero distortion coefficients are assumed.
@@ -630,8 +613,6 @@ Transforms an image to compensate for lens distortion.
 
 .. ocv:cfunction:: void cvUndistort2( const CvArr* src, CvArr* dst, const CvMat* camera_matrix, const CvMat* distortion_coeffs, const CvMat* new_camera_matrix=0 )
 
-.. ocv:pyoldfunction:: cv.Undistort2(src, dst, cameraMatrix, distCoeffs)-> None
-
     :param src: Input (distorted) image.
 
     :param dst: Output (corrected) image that has the same size and type as  ``src`` .
@@ -668,7 +649,6 @@ Computes the ideal point coordinates from the observed point coordinates.
 .. ocv:function:: void undistortPoints( InputArray src, OutputArray dst, InputArray cameraMatrix, InputArray distCoeffs, InputArray R=noArray(), InputArray P=noArray())
 
 .. ocv:cfunction:: void cvUndistortPoints( const CvMat* src, CvMat* dst, const CvMat* camera_matrix, const CvMat* dist_coeffs, const CvMat* R=0, const CvMat* P=0 )
-.. ocv:pyoldfunction:: cv.UndistortPoints(src, dst, cameraMatrix, distCoeffs, R=None, P=None)-> None
 
     :param src: Observed point coordinates, 1xN or Nx1 2-channel (CV_32FC2 or CV_64FC2).
 
@@ -702,5 +682,3 @@ The function is similar to
 where ``undistort()`` is an approximate iterative algorithm that estimates the normalized original point coordinates out of the normalized distorted point coordinates ("normalized" means that the coordinates do not depend on the camera matrix).
 
 The function can be used for both a stereo camera head or a monocular camera (when R is empty).
-
-
diff --git a/modules/imgproc/doc/histograms.rst b/modules/imgproc/doc/histograms.rst
index f478566f73..91199f3781 100644
--- a/modules/imgproc/doc/histograms.rst
+++ b/modules/imgproc/doc/histograms.rst
@@ -16,7 +16,6 @@ Calculates a histogram of a set of arrays.
 .. ocv:pyfunction:: cv2.calcHist(images, channels, mask, histSize, ranges[, hist[, accumulate]]) -> hist
 
 .. ocv:cfunction:: void cvCalcHist( IplImage** image, CvHistogram* hist, int accumulate=0, const CvArr* mask=NULL )
-.. ocv:pyoldfunction:: cv.CalcHist(image, hist, accumulate=0, mask=None)-> None
 
     :param images: Source arrays. They all should have the same depth,  ``CV_8U``  or  ``CV_32F`` , and the same size. Each of them can have an arbitrary number of channels.
 
@@ -43,8 +42,8 @@ arrays. The elements of a tuple used to increment
 a histogram bin are taken from the corresponding
 input arrays at the same location. The sample below shows how to compute a 2D Hue-Saturation histogram for a color image. ::
 
-    #include <cv.h>
-    #include <highgui.h>
+    #include <opencv2/imgproc.hpp>
+    #include <opencv2/highgui.hpp>
 
     using namespace cv;
 
@@ -54,7 +53,7 @@ input arrays at the same location. The sample below shows how to compute a 2D Hu
         if( argc != 2 || !(src=imread(argv[1], 1)).data )
             return -1;
 
-        cvtColor(src, hsv, CV_BGR2HSV);
+        cvtColor(src, hsv, COLOR_BGR2HSV);
 
         // Quantize the hue to 30 levels
         // and the saturation to 32 levels
@@ -99,7 +98,12 @@ input arrays at the same location. The sample below shows how to compute a 2D Hu
         waitKey();
     }
 
+.. note::
 
+   * An example for creating histograms of an image can be found at opencv_source_code/samples/cpp/demhist.cpp
+
+   * (Python) An example for creating color histograms can be found at opencv_source/samples/python2/color_histogram.py
+   * (Python) An example illustrating RGB and grayscale histogram plotting can be found at opencv_source/samples/python2/hist.py
 
 
 calcBackProject
@@ -113,7 +117,6 @@ Calculates the back projection of a histogram.
 .. ocv:pyfunction:: cv2.calcBackProject(images, channels, hist, ranges, scale[, dst]) -> dst
 
 .. ocv:cfunction:: void cvCalcBackProject( IplImage** image, CvArr* backProject, const CvHistogram* hist )
-.. ocv:pyoldfunction:: cv.CalcBackProject(image, back_project, hist) -> None
 
     :param images: Source arrays. They all should have the same depth,  ``CV_8U``  or  ``CV_32F`` , and the same size. Each of them can have an arbitrary number of channels.
 
@@ -159,7 +162,6 @@ Compares two histograms.
 .. ocv:pyfunction:: cv2.compareHist(H1, H2, method) -> retval
 
 .. ocv:cfunction:: double cvCompareHist( const CvHistogram* hist1, const CvHistogram* hist2, int method )
-.. ocv:pyoldfunction:: cv.CompareHist(hist1, hist2, method)->float
 
     :param H1: First compared histogram.
 
@@ -171,6 +173,8 @@ Compares two histograms.
 
             * **CV_COMP_CHISQR**     Chi-Square
 
+            * **CV_COMP_CHISQR_ALT**     Alternative Chi-Square
+
             * **CV_COMP_INTERSECT**     Intersection
 
             * **CV_COMP_BHATTACHARYYA**     Bhattacharyya distance
@@ -200,6 +204,14 @@ The functions ``compareHist`` compare two dense or two sparse histograms using t
 
         d(H_1,H_2) =  \sum _I  \frac{\left(H_1(I)-H_2(I)\right)^2}{H_1(I)}
 
+* Alternative Chi-Square (``method=CV_COMP_CHISQR_ALT``)
+
+    .. math::
+
+        d(H_1,H_2) =  2 * \sum _I  \frac{\left(H_1(I)-H_2(I)\right)^2}{H_1(I)+H_2(I)}
+
+    This alternative formula is regularly used for texture comparison. See e.g. [Puzicha1997]_.
+
 * Intersection (``method=CV_COMP_INTERSECT``)
 
     .. math::
@@ -229,8 +241,6 @@ Computes the "minimal work" distance between two weighted point configurations.
 
 .. ocv:cfunction:: float cvCalcEMD2( const CvArr* signature1, const CvArr* signature2, int distance_type, CvDistanceFunction distance_func=NULL, const CvArr* cost_matrix=NULL, CvArr* flow=NULL, float* lower_bound=NULL, void* userdata=NULL )
 
-.. ocv:pyoldfunction:: cv.CalcEMD2(signature1, signature2, distance_type, distance_func=None, cost_matrix=None, flow=None, lower_bound=None, userdata=None) -> float
-
     :param signature1: First signature, a  :math:`\texttt{size1}\times \texttt{dims}+1`  floating-point matrix. Each row stores the point weight followed by the point coordinates. The matrix is allowed to have a single column (weights only) if the user-defined cost matrix is used.
 
     :param signature2: Second signature of the same format as  ``signature1`` , though the number of rows may be different. The total weights may be different. In this case an extra "dummy" point is added to either  ``signature1``  or  ``signature2`` .
@@ -304,8 +314,6 @@ Locates a template within an image by using a histogram comparison.
 
 .. ocv:cfunction:: void cvCalcBackProjectPatch( IplImage** images, CvArr* dst, CvSize patch_size, CvHistogram* hist, int method, double factor )
 
-.. ocv:pyoldfunction:: cv.CalcBackProjectPatch(images, dst, patch_size, hist, method, factor)-> None
-
     :param images: Source images (though, you may pass CvMat** as well).
 
     :param dst: Destination image.
@@ -329,8 +337,6 @@ Divides one histogram by another.
 
 .. ocv:cfunction:: void cvCalcProbDensity( const CvHistogram* hist1, const CvHistogram* hist2, CvHistogram* dst_hist, double scale=255 )
 
-.. ocv:pyoldfunction:: cv.CalcProbDensity(hist1, hist2, dst_hist, scale=255) -> None
-
     :param hist1: First histogram (the divisor).
 
     :param hist2: Second histogram.
@@ -351,7 +357,6 @@ ClearHist
 Clears the histogram.
 
 .. ocv:cfunction:: void cvClearHist( CvHistogram* hist )
-.. ocv:pyoldfunction:: cv.ClearHist(hist)-> None
 
     :param hist: Histogram.
 
@@ -378,8 +383,6 @@ Creates a histogram.
 
 .. ocv:cfunction:: CvHistogram* cvCreateHist( int dims, int* sizes, int type, float** ranges=NULL, int uniform=1 )
 
-.. ocv:pyoldfunction:: cv.CreateHist(dims, type, ranges=None, uniform=1) -> hist
-
     :param dims: Number of histogram dimensions.
 
     :param sizes: Array of the histogram dimension sizes.
@@ -414,8 +417,6 @@ Finds the minimum and maximum histogram bins.
 
 .. ocv:cfunction:: void cvGetMinMaxHistValue(  const CvHistogram* hist, float* min_value, float* max_value, int* min_idx=NULL, int* max_idx=NULL )
 
-.. ocv:pyoldfunction:: cv.GetMinMaxHistValue(hist)-> (min_value, max_value, min_idx, max_idx)
-
     :param hist: Histogram.
 
     :param min_value: Pointer to the minimum value of the histogram.
@@ -454,7 +455,6 @@ NormalizeHist
 Normalizes the histogram.
 
 .. ocv:cfunction:: void cvNormalizeHist( CvHistogram* hist, double factor )
-.. ocv:pyoldfunction:: cv.NormalizeHist(hist, factor)-> None
 
     :param hist: Pointer to the histogram.
 
@@ -494,7 +494,6 @@ ThreshHist
 Thresholds the histogram.
 
 .. ocv:cfunction:: void cvThreshHist( CvHistogram* hist, double threshold )
-.. ocv:pyoldfunction:: cv.ThreshHist(hist, threshold) -> None
 
     :param hist: Pointer to the histogram.
 
@@ -504,3 +503,4 @@ The function clears histogram bins that are below the specified threshold.
 
 
 .. [RubnerSept98] Y. Rubner. C. Tomasi, L.J. Guibas. *The Earth Mover’s Distance as a Metric for Image Retrieval*. Technical Report STAN-CS-TN-98-86, Department of Computer Science, Stanford University, September 1998.
+.. [Puzicha1997] Puzicha, J., Hofmann, T., and Buhmann, J. *Non-parametric similarity measures for unsupervised texture segmentation and image retrieval.* In Proc. IEEE Conf. Computer Vision and Pattern Recognition, San Juan, Puerto Rico, pp. 267-272, 1997.
diff --git a/modules/imgproc/doc/miscellaneous_transformations.rst b/modules/imgproc/doc/miscellaneous_transformations.rst
index a82923fab6..47de0b442d 100644
--- a/modules/imgproc/doc/miscellaneous_transformations.rst
+++ b/modules/imgproc/doc/miscellaneous_transformations.rst
@@ -14,8 +14,6 @@ Applies an adaptive threshold to an array.
 
 .. ocv:cfunction:: void cvAdaptiveThreshold( const CvArr* src, CvArr* dst, double max_value, int adaptive_method=CV_ADAPTIVE_THRESH_MEAN_C, int threshold_type=CV_THRESH_BINARY, int block_size=3, double param1=5 )
 
-.. ocv:pyoldfunction:: cv.AdaptiveThreshold(src, dst, maxValue, adaptive_method=CV_ADAPTIVE_THRESH_MEAN_C, thresholdType=CV_THRESH_BINARY, blockSize=3, param1=5)-> None
-
     :param src: Source 8-bit single-channel image.
 
     :param dst: Destination image of the same size and the same type as  ``src`` .
@@ -79,7 +77,6 @@ Converts an image from one color space to another.
 .. ocv:pyfunction:: cv2.cvtColor(src, code[, dst[, dstCn]]) -> dst
 
 .. ocv:cfunction:: void cvCvtColor( const CvArr* src, CvArr* dst, int code )
-.. ocv:pyoldfunction:: cv.CvtColor(src, dst, code)-> None
 
     :param src: input image: 8-bit unsigned, 16-bit unsigned ( ``CV_16UC...`` ), or single-precision floating-point.
 
@@ -109,13 +106,14 @@ But in case of a non-linear transformation, an input RGB image should be normali
 :math:`\rightarrow` L*u*v* transformation. For example, if you have a 32-bit floating-point image directly converted from an 8-bit image without any scaling, then it will have the 0..255 value range instead of 0..1 assumed by the function. So, before calling ``cvtColor`` , you need first to scale the image down: ::
 
     img *= 1./255;
-    cvtColor(img, img, CV_BGR2Luv);
+    cvtColor(img, img, COLOR_BGR2Luv);
 
 If you use ``cvtColor`` with 8-bit images, the conversion will have some information lost. For many applications, this will not be noticeable but it is recommended to use 32-bit images in applications that need the full range of colors or that convert an image before an operation and then convert back.
 
 The function can do the following transformations:
 
 *
+    RGB :math:`\leftrightarrow` GRAY ( ``CV_BGR2GRAY, CV_RGB2GRAY, CV_GRAY2BGR, CV_GRAY2RGB``     )
     Transformations within RGB space like adding/removing the alpha channel, reversing the channel order, conversion to/from 16-bit RGB color (R5:G6:B5 or R5:G5:B5), as well as conversion to/from grayscale using:
 
     .. math::
@@ -132,7 +130,7 @@ The function can do the following transformations:
 
     ::
 
-        cvtColor(src, bwsrc, CV_RGB2GRAY);
+        cvtColor(src, bwsrc, COLOR_RGB2GRAY);
 
     ..
 
@@ -141,7 +139,7 @@ The function can do the following transformations:
 
 *
     RGB
-    :math:`\leftrightarrow`     CIE XYZ.Rec 709 with D65 white point ( ``CV_BGR2XYZ, CV_RGB2XYZ, CV_XYZ2BGR, CV_XYZ2RGB``     ):
+    :math:`\leftrightarrow`     CIE XYZ.Rec 709 with D65 white point ( ``COLOR_BGR2XYZ, COLOR_RGB2XYZ, COLOR_XYZ2BGR, COLOR_XYZ2RGB``     ):
 
     .. math::
 
@@ -163,7 +161,7 @@ The function can do the following transformations:
 
 *
     RGB
-    :math:`\leftrightarrow`     YCrCb JPEG (or YCC) ( ``CV_BGR2YCrCb, CV_RGB2YCrCb, CV_YCrCb2BGR, CV_YCrCb2RGB``     )
+    :math:`\leftrightarrow`     YCrCb JPEG (or YCC) ( ``COLOR_BGR2YCrCb, COLOR_RGB2YCrCb, COLOR_YCrCb2BGR, COLOR_YCrCb2RGB``     )
 
     .. math::
 
@@ -198,7 +196,7 @@ The function can do the following transformations:
     Y, Cr, and Cb cover the whole value range.
 
 *
-    RGB :math:`\leftrightarrow` HSV ( ``CV_BGR2HSV, CV_RGB2HSV, CV_HSV2BGR, CV_HSV2RGB``     )
+    RGB :math:`\leftrightarrow` HSV ( ``COLOR_BGR2HSV, COLOR_RGB2HSV, COLOR_HSV2BGR, COLOR_HSV2RGB``     )
       In case of 8-bit and 16-bit images,
       R, G, and B are converted to the floating-point format and scaled to fit the 0 to 1 range.
 
@@ -237,7 +235,7 @@ The function can do the following transformations:
         H, S, and V are left as is
 
 *
-    RGB :math:`\leftrightarrow` HLS ( ``CV_BGR2HLS, CV_RGB2HLS, CV_HLS2BGR, CV_HLS2RGB`` ).
+    RGB :math:`\leftrightarrow` HLS ( ``COLOR_BGR2HLS, COLOR_RGB2HLS, COLOR_HLS2BGR, COLOR_HLS2RGB`` ).
       In case of 8-bit and 16-bit images,
       R, G, and B are converted to the floating-point format and scaled to fit the 0 to 1 range.
 
@@ -287,7 +285,7 @@ The function can do the following transformations:
         H, S, V are left as is
 
 *
-    RGB :math:`\leftrightarrow` CIE L*a*b* ( ``CV_BGR2Lab, CV_RGB2Lab, CV_Lab2BGR, CV_Lab2RGB`` ).
+    RGB :math:`\leftrightarrow` CIE L*a*b* ( ``COLOR_BGR2Lab, COLOR_RGB2Lab, COLOR_Lab2BGR, COLOR_Lab2RGB`` ).
       In case of 8-bit and 16-bit images,
       R, G, and B are converted to the floating-point format and scaled to fit the 0 to 1 range.
 
@@ -343,7 +341,7 @@ The function can do the following transformations:
         L, a, and b are left as is
 
 *
-    RGB :math:`\leftrightarrow` CIE L*u*v* ( ``CV_BGR2Luv, CV_RGB2Luv, CV_Luv2BGR, CV_Luv2RGB`` ).
+    RGB :math:`\leftrightarrow` CIE L*u*v* ( ``COLOR_BGR2Luv, COLOR_RGB2Luv, COLOR_Luv2BGR, COLOR_Luv2RGB`` ).
       In case of 8-bit and 16-bit images,
       R, G, and B are converted to the floating-point format and scaled to fit 0 to 1 range.
 
@@ -392,7 +390,7 @@ The function can do the following transformations:
     http://www.poynton.com/ColorFAQ.html
 
 *
-    Bayer :math:`\rightarrow`     RGB ( ``CV_BayerBG2BGR, CV_BayerGB2BGR, CV_BayerRG2BGR, CV_BayerGR2BGR, CV_BayerBG2RGB, CV_BayerGB2RGB, CV_BayerRG2RGB, CV_BayerGR2RGB``     ). The Bayer pattern is widely used in CCD and CMOS cameras. It enables you to get color pictures from a single plane where R,G, and B pixels (sensors of a particular component) are interleaved as follows:
+    Bayer :math:`\rightarrow`     RGB ( ``COLOR_BayerBG2BGR, COLOR_BayerGB2BGR, COLOR_BayerRG2BGR, COLOR_BayerGR2BGR, COLOR_BayerBG2RGB, COLOR_BayerGB2RGB, COLOR_BayerRG2RGB, COLOR_BayerGR2RGB``     ). The Bayer pattern is widely used in CCD and CMOS cameras. It enables you to get color pictures from a single plane where R,G, and B pixels (sensors of a particular component) are interleaved as follows:
 
     .. image:: pics/bayer.png
 
@@ -419,8 +417,6 @@ Calculates the distance to the closest zero pixel for each pixel of the source i
 
 .. ocv:cfunction:: void cvDistTransform( const CvArr* src, CvArr* dst, int distance_type=CV_DIST_L2, int mask_size=3, const float* mask=NULL, CvArr* labels=NULL, int labelType=CV_DIST_LABEL_CCOMP )
 
-.. ocv:pyoldfunction:: cv.DistTransform(src, dst, distance_type=CV_DIST_L2, mask_size=3, mask=None, labels=None) -> None
-
     :param src: 8-bit, single-channel (binary) source image.
 
     :param dst: Output image with calculated distances. It is a 32-bit floating-point, single-channel image of the same size as  ``src`` .
@@ -480,6 +476,12 @@ In this mode, the complexity is still linear.
 That is, the function provides a very fast way to compute the Voronoi diagram for a binary image.
 Currently, the second variant can use only the approximate distance transform algorithm, i.e. ``maskSize=CV_DIST_MASK_PRECISE`` is not supported yet.
 
+.. note::
+
+   * An example on using the distance transform can be found at opencv_source_code/samples/cpp/distrans.cpp
+
+   * (Python) An example on using the distance transform can be found at opencv_source/samples/python2/distrans.py
+
 floodFill
 ---------
 Fills a connected component with the given color.
@@ -491,7 +493,6 @@ Fills a connected component with the given color.
 .. ocv:pyfunction:: cv2.floodFill(image, mask, seedPoint, newVal[, loDiff[, upDiff[, flags]]]) -> retval, image, mask, rect
 
 .. ocv:cfunction:: void cvFloodFill( CvArr* image, CvPoint seed_point, CvScalar new_val, CvScalar lo_diff=cvScalarAll(0), CvScalar up_diff=cvScalarAll(0), CvConnectedComp* comp=NULL, int flags=4, CvArr* mask=NULL )
-.. ocv:pyoldfunction:: cv.FloodFill(image, seed_point, new_val, lo_diff=(0, 0, 0, 0), up_diff=(0, 0, 0, 0), flags=4, mask=None)-> comp
 
     :param image: Input/output 1- or 3-channel, 8-bit, or floating-point image. It is modified by the function unless the  ``FLOODFILL_MASK_ONLY``  flag is set in the second variant of the function. See the details below.
 
@@ -513,7 +514,7 @@ Fills a connected component with the given color.
 
             * **FLOODFILL_FIXED_RANGE** If set, the difference between the current pixel and seed pixel is considered. Otherwise, the difference between neighbor pixels is considered (that is, the range is floating).
 
-            * **FLOODFILL_MASK_ONLY**  If set, the function does not change the image ( ``newVal``  is ignored), but fills the mask.  The flag can be used for the second variant only.
+            * **FLOODFILL_MASK_ONLY**  If set, the function does not change the image ( ``newVal``  is ignored), but fills the mask with the value in bits 8-16 of ``flags`` (that is, the fill value is set to newValue by adding (newValue << 8) to the ``flags``).  The flag can be used for the second variant only.
 
 The functions ``floodFill`` fill a connected component starting from the seed point with the specified color. The connectivity is determined by the color/brightness closeness of the neighbor pixels. The pixel at
 :math:`(x,y)` is considered to belong to the repainted domain if:
@@ -579,11 +580,15 @@ where
 *
     Color/brightness of the seed point in case of a fixed range.
 
-Use these functions to either mark a connected component with the specified color in-place, or build a mask and then extract the contour, or copy the region to another image, and so on. Various modes of the function are demonstrated in the ``floodfill.cpp`` sample.
+Use these functions to either mark a connected component with the specified color in-place, or build a mask and then extract the contour, or copy the region to another image, and so on.
 
 .. seealso:: :ocv:func:`findContours`
 
+.. note::
 
+   * An example using the FloodFill technique can be found at opencv_source_code/samples/cpp/ffilldemo.cpp
+
+   * (Python) An example using the FloodFill technique can be found at opencv_source_code/samples/python2/floodfill.cpp
 
 integral
 --------
@@ -603,8 +608,6 @@ Calculates the integral of an image.
 
 .. ocv:cfunction:: void cvIntegral( const CvArr* image, CvArr* sum, CvArr* sqsum=NULL, CvArr* tilted_sum=NULL )
 
-.. ocv:pyoldfunction:: cv.Integral(image, sum, sqsum=None, tiltedSum=None)-> None
-
     :param image: input image as :math:`W \times H`, 8-bit or floating-point (32f or 64f).
 
     :param sum: integral image as  :math:`(W+1)\times (H+1)` , 32-bit integer or floating-point (32f or 64f).
@@ -655,8 +658,6 @@ Applies a fixed-level threshold to each array element.
 
 .. ocv:cfunction:: double cvThreshold( const CvArr* src, CvArr* dst, double threshold, double max_value, int threshold_type )
 
-.. ocv:pyoldfunction:: cv.Threshold(src, dst, threshold, maxValue, thresholdType)-> None
-
     :param src: input array (single-channel, 8-bit or 32-bit floating point).
 
     :param dst: output array of the same size and type as ``src``.
@@ -747,6 +748,12 @@ Visual demonstration and usage example of the function can be found in the OpenC
 
 .. seealso:: :ocv:func:`findContours`
 
+.. note::
+
+   * An example using the watershed algorithm can be found at opencv_source_code/samples/cpp/watershed.cpp
+
+   * (Python) An example using the watershed algorithm can be found at opencv_source_code/samples/python2/watershed.py
+
 grabCut
 -------
 Runs the GrabCut algorithm.
@@ -765,7 +772,7 @@ Runs the GrabCut algorithm.
 
         * **GC_PR_BGD** defines a possible background pixel.
 
-        * **GC_PR_BGD** defines a possible foreground pixel.
+        * **GC_PR_FGD** defines a possible foreground pixel.
 
     :param rect: ROI containing a segmented object. The pixels outside of the ROI are marked as "obvious background". The parameter is only used when  ``mode==GC_INIT_WITH_RECT`` .
 
@@ -792,4 +799,9 @@ See the sample ``grabcut.cpp`` to learn how to use the function.
 
 .. [Meyer92] Meyer, F. *Color Image Segmentation*, ICIP92, 1992
 
-.. [Telea04] Alexandru Telea, *An Image Inpainting Technique Based on the Fast Marching Method*. Journal of Graphics, GPU, and Game Tools 9 1, pp 23-34 (2004)
+
+.. note::
+
+   * An example using the GrabCut algorithm can be found at opencv_source_code/samples/cpp/grabcut.cpp
+
+   * (Python) An example using the GrabCut algorithm can be found at opencv_source_code/samples/python2/grabcut.py
diff --git a/modules/imgproc/doc/motion_analysis_and_object_tracking.rst b/modules/imgproc/doc/motion_analysis_and_object_tracking.rst
index bcc372a6a2..c02125e618 100644
--- a/modules/imgproc/doc/motion_analysis_and_object_tracking.rst
+++ b/modules/imgproc/doc/motion_analysis_and_object_tracking.rst
@@ -13,8 +13,6 @@ Adds an image to the accumulator.
 
 .. ocv:cfunction:: void cvAcc( const CvArr* image, CvArr* sum, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.Acc(image, sum, mask=None) -> None
-
     :param src: Input image as 1- or 3-channel, 8-bit or 32-bit floating point.
 
     :param dst: Accumulator image with the same number of channels as input image, 32-bit or 64-bit floating-point.
@@ -49,8 +47,6 @@ Adds the square of a source image to the accumulator.
 
 .. ocv:cfunction:: void cvSquareAcc( const CvArr* image, CvArr* sqsum, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.SquareAcc(image, sqsum, mask=None) -> None
-
     :param src: Input image as 1- or 3-channel, 8-bit or 32-bit floating point.
 
     :param dst: Accumulator image with the same number of channels as input image, 32-bit or 64-bit floating-point.
@@ -83,8 +79,6 @@ Adds the per-element product of two input images to the accumulator.
 
 .. ocv:cfunction:: void cvMultiplyAcc( const CvArr* image1, const CvArr* image2, CvArr* acc, const CvArr* mask=NULL )
 
-.. ocv:pyoldfunction:: cv.MultiplyAcc(image1, image2, acc, mask=None)-> None
-
     :param src1: First input image, 1- or 3-channel, 8-bit or 32-bit floating point.
 
     :param src2: Second input image of the same type and the same size as  ``src1`` .
@@ -118,7 +112,6 @@ Updates a running average.
 .. ocv:pyfunction:: cv2.accumulateWeighted(src, dst, alpha[, mask]) -> dst
 
 .. ocv:cfunction:: void cvRunningAvg( const CvArr* image, CvArr* acc, double alpha, const CvArr* mask=NULL )
-.. ocv:pyoldfunction:: cv.RunningAvg(image, acc, alpha, mask=None)-> None
 
     :param src: Input image as 1- or 3-channel, 8-bit or 32-bit floating point.
 
diff --git a/modules/imgproc/doc/object_detection.rst b/modules/imgproc/doc/object_detection.rst
index c6231a05e0..12996b5a48 100644
--- a/modules/imgproc/doc/object_detection.rst
+++ b/modules/imgproc/doc/object_detection.rst
@@ -12,7 +12,6 @@ Compares a template against overlapped image regions.
 .. ocv:pyfunction:: cv2.matchTemplate(image, templ, method[, result]) -> result
 
 .. ocv:cfunction:: void cvMatchTemplate( const CvArr* image, const CvArr* templ, CvArr* result, int method )
-.. ocv:pyoldfunction:: cv.MatchTemplate(image, templ, result, method)-> None
 
     :param image: Image where the search is running. It must be 8-bit or 32-bit floating-point.
 
@@ -74,3 +73,6 @@ image patch:
 After the function finishes the comparison, the best matches can be found as global minimums (when ``CV_TM_SQDIFF`` was used) or maximums (when ``CV_TM_CCORR`` or ``CV_TM_CCOEFF`` was used) using the
 :ocv:func:`minMaxLoc` function. In case of a color image, template summation in the numerator and each sum in the denominator is done over all of the channels and separate mean values are used for each channel. That is, the function can take a color template and a color image. The result will still be a single-channel image, which is easier to analyze.
 
+.. note::
+
+   * (Python) An example on how to match mouse selected regions in an image can be found at opencv_source_code/samples/python2/mouse_and_match.py
diff --git a/modules/imgproc/doc/pics/building_lsd.png b/modules/imgproc/doc/pics/building_lsd.png
new file mode 100644
index 0000000000..747029a65d
Binary files /dev/null and b/modules/imgproc/doc/pics/building_lsd.png differ
diff --git a/modules/imgproc/doc/structural_analysis_and_shape_descriptors.rst b/modules/imgproc/doc/structural_analysis_and_shape_descriptors.rst
index 4c1911a77d..d34600669a 100644
--- a/modules/imgproc/doc/structural_analysis_and_shape_descriptors.rst
+++ b/modules/imgproc/doc/structural_analysis_and_shape_descriptors.rst
@@ -13,8 +13,6 @@ Calculates all of the moments up to the third order of a polygon or rasterized s
 
 .. ocv:cfunction:: void cvMoments( const CvArr* arr, CvMoments* moments, int binary=0 )
 
-.. ocv:pyoldfunction:: cv.Moments(arr, binary=0) -> moments
-
     :param array: Raster image (single-channel, 8-bit or floating-point 2D array) or an array ( :math:`1 \times N`  or  :math:`N \times 1` ) of 2D points (``Point``  or  ``Point2f`` ).
 
     :param binaryImage: If it is true, all non-zero image pixels are treated as 1's. The parameter is used for images only.
@@ -98,8 +96,6 @@ Calculates seven Hu invariants.
 
 .. ocv:cfunction:: void cvGetHuMoments( CvMoments* moments, CvHuMoments* hu_moments )
 
-.. ocv:pyoldfunction:: cv.GetHuMoments(moments) -> hu
-
     :param moments: Input moments computed with  :ocv:func:`moments` .
     :param hu: Output Hu invariants.
 
@@ -163,8 +159,6 @@ Finds contours in a binary image.
 
 .. ocv:cfunction:: int cvFindContours( CvArr* image, CvMemStorage* storage, CvSeq** first_contour, int header_size=sizeof(CvContour), int mode=CV_RETR_LIST, int method=CV_CHAIN_APPROX_SIMPLE, CvPoint offset=cvPoint(0,0) )
 
-.. ocv:pyoldfunction:: cv.FindContours(image, storage, mode=CV_RETR_LIST, method=CV_CHAIN_APPROX_SIMPLE, offset=(0, 0)) -> contours
-
     :param image: Source, an 8-bit single-channel image. Non-zero pixels are treated as 1's. Zero pixels remain 0's, so the image is treated as  ``binary`` . You can use  :ocv:func:`compare` ,  :ocv:func:`inRange` ,  :ocv:func:`threshold` ,  :ocv:func:`adaptiveThreshold` ,  :ocv:func:`Canny` , and others to create a binary image out of a grayscale or color one. The function modifies the  ``image``  while extracting the contours.
 
     :param contours: Detected contours. Each contour is stored as a vector of points.
@@ -198,6 +192,14 @@ The function retrieves contours from the binary image using the algorithm
 
 .. note:: If you use the new Python interface then the ``CV_`` prefix has to be omitted in contour retrieval mode and contour approximation method parameters (for example, use ``cv2.RETR_LIST`` and ``cv2.CHAIN_APPROX_NONE`` parameters). If you use the old Python interface then these parameters have the ``CV_`` prefix (for example, use ``cv.CV_RETR_LIST`` and ``cv.CV_CHAIN_APPROX_NONE``).
 
+.. note::
+
+   * An example using the findContour functionality can be found at opencv_source_code/samples/cpp/contours2.cpp
+   * An example using findContours to clean up a background segmentation result at opencv_source_code/samples/cpp/segment_objects.cpp
+
+   * (Python) An example using the findContour functionality can be found at opencv_source/samples/python2/contours.py
+   * (Python) An example of detecting squares in an image can be found at opencv_source/samples/python2/squares.py
+
 
 approxPolyDP
 ----------------
@@ -243,8 +245,6 @@ Approximates Freeman chain(s) with a polygonal curve.
 
 .. ocv:cfunction:: CvSeq* cvApproxChains( CvSeq* src_seq, CvMemStorage* storage, int method=CV_CHAIN_APPROX_SIMPLE, double parameter=0, int minimal_perimeter=0, int recursive=0 )
 
-.. ocv:pyoldfunction:: cv.ApproxChains(src_seq, storage, method=CV_CHAIN_APPROX_SIMPLE, parameter=0, minimal_perimeter=0, recursive=0)-> contours
-
     :param src_seq: Pointer to the approximated Freeman chain that can refer to other chains.
 
     :param storage: Storage location for the resulting polylines.
@@ -270,8 +270,6 @@ Calculates a contour perimeter or a curve length.
 
 .. ocv:cfunction:: double cvArcLength( const void* curve, CvSlice slice=CV_WHOLE_SEQ, int is_closed=-1 )
 
-.. ocv:pyoldfunction:: cv.ArcLength(curve, slice=CV_WHOLE_SEQ, isClosed=-1) -> float
-
     :param curve: Input vector of 2D points, stored in ``std::vector`` or ``Mat``.
 
     :param closed: Flag indicating whether the curve is closed or not.
@@ -289,7 +287,6 @@ Calculates the up-right bounding rectangle of a point set.
 .. ocv:pyfunction:: cv2.boundingRect(points) -> retval
 
 .. ocv:cfunction:: CvRect cvBoundingRect( CvArr* points, int update=0 )
-.. ocv:pyoldfunction:: cv.BoundingRect(points, update=0)-> CvRect
 
     :param points: Input 2D point set, stored in ``std::vector`` or ``Mat``.
 
@@ -308,8 +305,6 @@ Calculates a contour area.
 
 .. ocv:cfunction:: double cvContourArea( const CvArr* contour, CvSlice slice=CV_WHOLE_SEQ, int oriented=0 )
 
-.. ocv:pyoldfunction:: cv.ContourArea(contour, slice=CV_WHOLE_SEQ) -> float
-
     :param contour: Input vector of 2D points (contour vertices), stored in ``std::vector`` or ``Mat``.
 
     :param oriented: Oriented area flag. If it is true, the function returns a signed area value, depending on the contour orientation (clockwise or counter-clockwise). Using this feature you can determine orientation of a contour by taking the sign of an area. By default, the parameter is ``false``, which means that the absolute value is returned.
@@ -349,15 +344,13 @@ Finds the convex hull of a point set.
 
 .. ocv:cfunction:: CvSeq* cvConvexHull2( const CvArr* input, void* hull_storage=NULL, int orientation=CV_CLOCKWISE, int return_points=0 )
 
-.. ocv:pyoldfunction:: cv.ConvexHull2(points, storage, orientation=CV_CLOCKWISE, return_points=0) -> convexHull
-
     :param points: Input 2D point set, stored in ``std::vector`` or ``Mat``.
 
     :param hull: Output convex hull. It is either an integer vector of indices or vector of points. In the first case, the ``hull`` elements are 0-based indices of the convex hull points in the original array (since the set of convex hull points is a subset of the original point set). In the second case, ``hull`` elements are the convex hull points themselves.
 
     :param hull_storage: Output memory storage in the old API (``cvConvexHull2`` returns a sequence containing the convex hull points or their indices).
 
-    :param clockwise: Orientation flag. If it is true, the output convex hull is oriented clockwise. Otherwise, it is oriented counter-clockwise. The usual screen coordinate system is assumed so that the origin is at the top-left corner, x axis is oriented to the right, and y axis is oriented downwards.
+    :param clockwise: Orientation flag. If it is true, the output convex hull is oriented clockwise. Otherwise, it is oriented counter-clockwise. The assumed coordinate system has its X axis pointing to the right, and its Y axis pointing upwards.
 
     :param orientation: Convex hull orientation parameter in the old API, ``CV_CLOCKWISE`` or ``CV_COUNTERCLOCKWISE``.
 
@@ -368,6 +361,10 @@ The functions find the convex hull of a 2D point set using the Sklansky's algori
 that has
 *O(N logN)* complexity in the current implementation. See the OpenCV sample ``convexhull.cpp`` that demonstrates the usage of different function variants.
 
+.. note::
+
+   * An example using the convexHull functionality can be found at opencv_source_code/samples/cpp/convexhull.cpp
+
 
 convexityDefects
 ----------------
@@ -379,8 +376,6 @@ Finds the convexity defects of a contour.
 
 .. ocv:cfunction:: CvSeq* cvConvexityDefects(  const CvArr* contour, const CvArr* convexhull, CvMemStorage* storage=NULL )
 
-.. ocv:pyoldfunction:: cv.ConvexityDefects(contour, convexhull, storage)-> convexityDefects
-
     :param contour: Input contour.
 
     :param convexhull: Convex hull obtained using  :ocv:func:`convexHull`  that should contain indices of the contour points that make the hull.
@@ -412,7 +407,6 @@ Fits an ellipse around a set of 2D points.
 .. ocv:pyfunction:: cv2.fitEllipse(points) -> retval
 
 .. ocv:cfunction:: CvBox2D cvFitEllipse2( const CvArr* points )
-.. ocv:pyoldfunction:: cv.FitEllipse2(points)-> Box2D
 
     :param points: Input 2D point set, stored in:
 
@@ -424,6 +418,11 @@ Fits an ellipse around a set of 2D points.
 
 The function calculates the ellipse that fits (in a least-squares sense) a set of 2D points best of all. It returns the rotated rectangle in which the ellipse is inscribed. The algorithm [Fitzgibbon95]_ is used.
 
+.. note::
+
+   * An example using the fitEllipse technique can be found at opencv_source_code/samples/cpp/fitellipse.cpp
+
+
 fitLine
 -----------
 Fits a line to a 2D or 3D point set.
@@ -434,8 +433,6 @@ Fits a line to a 2D or 3D point set.
 
 .. ocv:cfunction:: void cvFitLine( const CvArr* points, int dist_type, double param, double reps, double aeps, float* line )
 
-.. ocv:pyoldfunction:: cv.FitLine(points, dist_type, param, reps, aeps) -> line
-
     :param points: Input vector of 2D or 3D points, stored in ``std::vector<>`` or ``Mat``.
 
     :param line: Output line parameters. In case of 2D fitting, it should be a vector of 4 elements (like ``Vec4f``) - ``(vx, vy, x0, y0)``,  where  ``(vx, vy)``  is a normalized vector collinear to the line and  ``(x0, y0)``  is a point on the line. In case of 3D fitting, it should be a vector of 6 elements (like  ``Vec6f``) - ``(vx, vy, vz, x0, y0, z0)``, where ``(vx, vy, vz)`` is a normalized vector collinear to the line and ``(x0, y0, z0)`` is a point on the line.
@@ -496,6 +493,9 @@ http://en.wikipedia.org/wiki/M-estimator
 :math:`w_i` are adjusted to be inversely proportional to
 :math:`\rho(r_i)` .
 
+.. Sample code:
+
+   * (Python) An example of robust line fitting can be found at opencv_source_code/samples/python2/fitline.py
 
 
 isContourConvex
@@ -507,7 +507,6 @@ Tests a contour convexity.
 .. ocv:pyfunction:: cv2.isContourConvex(contour) -> retval
 
 .. ocv:cfunction:: int cvCheckContourConvexity( const CvArr* contour )
-.. ocv:pyoldfunction:: cv.CheckContourConvexity(contour)-> int
 
     :param contour: Input vector of 2D points, stored in:
 
@@ -531,8 +530,6 @@ Finds a rotated rectangle of the minimum area enclosing the input 2D point set.
 
 .. ocv:cfunction:: CvBox2D cvMinAreaRect2( const CvArr* points, CvMemStorage* storage=NULL )
 
-.. ocv:pyoldfunction:: cv.MinAreaRect2(points, storage=None) -> Box2D
-
     :param points: Input vector of 2D points, stored in:
 
         * ``std::vector<>`` or ``Mat`` (C++ interface)
@@ -545,6 +542,24 @@ The function calculates and returns the minimum-area bounding rectangle (possibl
 
 
 
+boxPoints
+-----------
+Finds the four vertices of a rotated rect. Useful to draw the rotated rectangle.
+
+.. ocv:function:: void boxPoints(RotatedRect box, OutputArray points)
+
+.. ocv:pyfunction:: cv2.boxPoints(box[, points]) -> points
+
+.. ocv:cfunction:: void cvBoxPoints( CvBox2D box, CvPoint2D32f pt[4] )
+
+    :param box: The input rotated rectangle. It may be the output of .. ocv:function:: minAreaRect.
+
+    :param points: The output array of four vertices of rectangles.
+
+The function finds the four vertices of a rotated rectangle. This function is useful to draw the rectangle. In C++, instead of using this function, you can directly use box.points() method. Please visit the `tutorial on bounding rectangle <http://docs.opencv.org/doc/tutorials/imgproc/shapedescriptors/bounding_rects_circles/bounding_rects_circles.html#bounding-rects-circles>`_ for more information.
+
+
+
 minEnclosingCircle
 ----------------------
 Finds a circle of the minimum area enclosing a 2D point set.
@@ -555,8 +570,6 @@ Finds a circle of the minimum area enclosing a 2D point set.
 
 .. ocv:cfunction:: int cvMinEnclosingCircle( const CvArr* points, CvPoint2D32f* center, float* radius )
 
-.. ocv:pyoldfunction:: cv.MinEnclosingCircle(points)-> (int, center, radius)
-
     :param points: Input vector of 2D points, stored in:
 
         * ``std::vector<>`` or ``Mat`` (C++ interface)
@@ -582,7 +595,6 @@ Compares two shapes.
 .. ocv:pyfunction:: cv2.matchShapes(contour1, contour2, method, parameter) -> retval
 
 .. ocv:cfunction:: double cvMatchShapes( const void* object1, const void* object2, int method, double parameter=0 )
-.. ocv:pyoldfunction:: cv.MatchShapes(object1, object2, method, parameter=0) -> float
 
     :param object1: First contour or grayscale image.
 
@@ -637,7 +649,6 @@ Performs a point-in-contour test.
 .. ocv:pyfunction:: cv2.pointPolygonTest(contour, pt, measureDist) -> retval
 
 .. ocv:cfunction:: double cvPointPolygonTest( const CvArr* contour, CvPoint2D32f pt, int measure_dist )
-.. ocv:pyoldfunction:: cv.PointPolygonTest(contour, pt, measure_dist) -> float
 
     :param contour: Input contour.
 
diff --git a/modules/imgproc/include/opencv2/imgproc.hpp b/modules/imgproc/include/opencv2/imgproc.hpp
index df456b5cd0..5ff1b7449c 100644
--- a/modules/imgproc/include/opencv2/imgproc.hpp
+++ b/modules/imgproc/include/opencv2/imgproc.hpp
@@ -44,9 +44,6 @@
 #define __OPENCV_IMGPROC_HPP__
 
 #include "opencv2/core.hpp"
-#include "opencv2/imgproc/types_c.h"
-
-#ifdef __cplusplus
 
 /*! \namespace cv
  Namespace where all the C++ OpenCV functionality resides
@@ -54,15 +51,418 @@
 namespace cv
 {
 
-//! various border interpolation methods
-enum { BORDER_REPLICATE=IPL_BORDER_REPLICATE, BORDER_CONSTANT=IPL_BORDER_CONSTANT,
-       BORDER_REFLECT=IPL_BORDER_REFLECT, BORDER_WRAP=IPL_BORDER_WRAP,
-       BORDER_REFLECT_101=IPL_BORDER_REFLECT_101, BORDER_REFLECT101=BORDER_REFLECT_101,
-       BORDER_TRANSPARENT=IPL_BORDER_TRANSPARENT,
-       BORDER_DEFAULT=BORDER_REFLECT_101, BORDER_ISOLATED=16 };
+//! type of the kernel
+enum { KERNEL_GENERAL      = 0, // the kernel is generic. No any type of symmetry or other properties.
+       KERNEL_SYMMETRICAL  = 1, // kernel[i] == kernel[ksize-i-1] , and the anchor is at the center
+       KERNEL_ASYMMETRICAL = 2, // kernel[i] == -kernel[ksize-i-1] , and the anchor is at the center
+       KERNEL_SMOOTH       = 4, // all the kernel elements are non-negative and summed to 1
+       KERNEL_INTEGER      = 8  // all the kernel coefficients are integer numbers
+     };
+
+//! type of morphological operation
+enum { MORPH_ERODE    = 0,
+       MORPH_DILATE   = 1,
+       MORPH_OPEN     = 2,
+       MORPH_CLOSE    = 3,
+       MORPH_GRADIENT = 4,
+       MORPH_TOPHAT   = 5,
+       MORPH_BLACKHAT = 6
+     };
+
+//! shape of the structuring element
+enum { MORPH_RECT    = 0,
+       MORPH_CROSS   = 1,
+       MORPH_ELLIPSE = 2
+     };
+
+//! interpolation algorithm
+enum { INTER_NEAREST    = 0, //!< nearest neighbor interpolation
+       INTER_LINEAR     = 1, //!< bilinear interpolation
+       INTER_CUBIC      = 2, //!< bicubic interpolation
+       INTER_AREA       = 3, //!< area-based (or super) interpolation
+       INTER_LANCZOS4   = 4, //!< Lanczos interpolation over 8x8 neighborhood
+
+       INTER_MAX        = 7, //!< mask for interpolation codes
+       WARP_INVERSE_MAP = 16
+     };
+
+enum { INTER_BITS      = 5,
+       INTER_BITS2     = INTER_BITS * 2,
+       INTER_TAB_SIZE  = 1 << INTER_BITS,
+       INTER_TAB_SIZE2 = INTER_TAB_SIZE * INTER_TAB_SIZE
+     };
+
+//! Distance types for Distance Transform and M-estimators
+enum { DIST_USER    = -1,  // User defined distance
+       DIST_L1      = 1,   // distance = |x1-x2| + |y1-y2|
+       DIST_L2      = 2,   // the simple euclidean distance
+       DIST_C       = 3,   // distance = max(|x1-x2|,|y1-y2|)
+       DIST_L12     = 4,   // L1-L2 metric: distance = 2(sqrt(1+x*x/2) - 1))
+       DIST_FAIR    = 5,   // distance = c^2(|x|/c-log(1+|x|/c)), c = 1.3998
+       DIST_WELSCH  = 6,   // distance = c^2/2(1-exp(-(x/c)^2)), c = 2.9846
+       DIST_HUBER   = 7    // distance = |x|<c ? x^2/2 : c(|x|-c/2), c=1.345
+};
+
+//! Mask size for distance transform
+enum { DIST_MASK_3       = 3,
+       DIST_MASK_5       = 5,
+       DIST_MASK_PRECISE = 0
+     };
+
+//! type of the threshold operation
+enum { THRESH_BINARY     = 0, // value = value > threshold ? max_value : 0
+       THRESH_BINARY_INV = 1, // value = value > threshold ? 0 : max_value
+       THRESH_TRUNC      = 2, // value = value > threshold ? threshold : value
+       THRESH_TOZERO     = 3, // value = value > threshold ? value : 0
+       THRESH_TOZERO_INV = 4, // value = value > threshold ? 0 : value
+       THRESH_MASK       = 7,
+       THRESH_OTSU       = 8  // use Otsu algorithm to choose the optimal threshold value
+     };
+
+//! adaptive threshold algorithm
+enum { ADAPTIVE_THRESH_MEAN_C     = 0,
+       ADAPTIVE_THRESH_GAUSSIAN_C = 1
+     };
+
+enum { PROJ_SPHERICAL_ORTHO  = 0,
+       PROJ_SPHERICAL_EQRECT = 1
+     };
+
+//! class of the pixel in GrabCut algorithm
+enum { GC_BGD    = 0,  //!< background
+       GC_FGD    = 1,  //!< foreground
+       GC_PR_BGD = 2,  //!< most probably background
+       GC_PR_FGD = 3   //!< most probably foreground
+     };
+
+//! GrabCut algorithm flags
+enum { GC_INIT_WITH_RECT  = 0,
+       GC_INIT_WITH_MASK  = 1,
+       GC_EVAL            = 2
+};
+
+//! distanceTransform algorithm flags
+enum { DIST_LABEL_CCOMP = 0,
+       DIST_LABEL_PIXEL = 1
+     };
+
+//! floodfill algorithm flags
+enum { FLOODFILL_FIXED_RANGE = 1 << 16,
+       FLOODFILL_MASK_ONLY   = 1 << 17
+     };
+
+//! type of the template matching operation
+enum { TM_SQDIFF        = 0,
+       TM_SQDIFF_NORMED = 1,
+       TM_CCORR         = 2,
+       TM_CCORR_NORMED  = 3,
+       TM_CCOEFF        = 4,
+       TM_CCOEFF_NORMED = 5
+     };
+
+//! connected components algorithm output formats
+enum { CC_STAT_LEFT   = 0,
+       CC_STAT_TOP    = 1,
+       CC_STAT_WIDTH  = 2,
+       CC_STAT_HEIGHT = 3,
+       CC_STAT_AREA   = 4,
+       CC_STAT_MAX    = 5
+     };
+
+//! mode of the contour retrieval algorithm
+enum { RETR_EXTERNAL  = 0, //!< retrieve only the most external (top-level) contours
+       RETR_LIST      = 1, //!< retrieve all the contours without any hierarchical information
+       RETR_CCOMP     = 2, //!< retrieve the connected components (that can possibly be nested)
+       RETR_TREE      = 3, //!< retrieve all the contours and the whole hierarchy
+       RETR_FLOODFILL = 4
+     };
+
+//! the contour approximation algorithm
+enum { CHAIN_APPROX_NONE      = 1,
+       CHAIN_APPROX_SIMPLE    = 2,
+       CHAIN_APPROX_TC89_L1   = 3,
+       CHAIN_APPROX_TC89_KCOS = 4
+     };
+
+//! Variants of a Hough transform
+enum { HOUGH_STANDARD      = 0,
+       HOUGH_PROBABILISTIC = 1,
+       HOUGH_MULTI_SCALE   = 2,
+       HOUGH_GRADIENT      = 3
+     };
+
+//! Variants of Line Segment Detector
+enum { LSD_REFINE_NONE = 0,
+       LSD_REFINE_STD  = 1,
+       LSD_REFINE_ADV  = 2
+     };
+
+//! Histogram comparison methods
+enum { HISTCMP_CORREL        = 0,
+       HISTCMP_CHISQR        = 1,
+       HISTCMP_INTERSECT     = 2,
+       HISTCMP_BHATTACHARYYA = 3,
+       HISTCMP_HELLINGER     = HISTCMP_BHATTACHARYYA
+     };
+
+//! the color conversion code
+enum { COLOR_BGR2BGRA     = 0,
+       COLOR_RGB2RGBA     = COLOR_BGR2BGRA,
+
+       COLOR_BGRA2BGR     = 1,
+       COLOR_RGBA2RGB     = COLOR_BGRA2BGR,
+
+       COLOR_BGR2RGBA     = 2,
+       COLOR_RGB2BGRA     = COLOR_BGR2RGBA,
+
+       COLOR_RGBA2BGR     = 3,
+       COLOR_BGRA2RGB     = COLOR_RGBA2BGR,
+
+       COLOR_BGR2RGB      = 4,
+       COLOR_RGB2BGR      = COLOR_BGR2RGB,
+
+       COLOR_BGRA2RGBA    = 5,
+       COLOR_RGBA2BGRA    = COLOR_BGRA2RGBA,
+
+       COLOR_BGR2GRAY     = 6,
+       COLOR_RGB2GRAY     = 7,
+       COLOR_GRAY2BGR     = 8,
+       COLOR_GRAY2RGB     = COLOR_GRAY2BGR,
+       COLOR_GRAY2BGRA    = 9,
+       COLOR_GRAY2RGBA    = COLOR_GRAY2BGRA,
+       COLOR_BGRA2GRAY    = 10,
+       COLOR_RGBA2GRAY    = 11,
+
+       COLOR_BGR2BGR565   = 12,
+       COLOR_RGB2BGR565   = 13,
+       COLOR_BGR5652BGR   = 14,
+       COLOR_BGR5652RGB   = 15,
+       COLOR_BGRA2BGR565  = 16,
+       COLOR_RGBA2BGR565  = 17,
+       COLOR_BGR5652BGRA  = 18,
+       COLOR_BGR5652RGBA  = 19,
+
+       COLOR_GRAY2BGR565  = 20,
+       COLOR_BGR5652GRAY  = 21,
+
+       COLOR_BGR2BGR555   = 22,
+       COLOR_RGB2BGR555   = 23,
+       COLOR_BGR5552BGR   = 24,
+       COLOR_BGR5552RGB   = 25,
+       COLOR_BGRA2BGR555  = 26,
+       COLOR_RGBA2BGR555  = 27,
+       COLOR_BGR5552BGRA  = 28,
+       COLOR_BGR5552RGBA  = 29,
+
+       COLOR_GRAY2BGR555  = 30,
+       COLOR_BGR5552GRAY  = 31,
+
+       COLOR_BGR2XYZ      = 32,
+       COLOR_RGB2XYZ      = 33,
+       COLOR_XYZ2BGR      = 34,
+       COLOR_XYZ2RGB      = 35,
+
+       COLOR_BGR2YCrCb    = 36,
+       COLOR_RGB2YCrCb    = 37,
+       COLOR_YCrCb2BGR    = 38,
+       COLOR_YCrCb2RGB    = 39,
+
+       COLOR_BGR2HSV      = 40,
+       COLOR_RGB2HSV      = 41,
+
+       COLOR_BGR2Lab      = 44,
+       COLOR_RGB2Lab      = 45,
+
+       COLOR_BGR2Luv      = 50,
+       COLOR_RGB2Luv      = 51,
+       COLOR_BGR2HLS      = 52,
+       COLOR_RGB2HLS      = 53,
+
+       COLOR_HSV2BGR      = 54,
+       COLOR_HSV2RGB      = 55,
+
+       COLOR_Lab2BGR      = 56,
+       COLOR_Lab2RGB      = 57,
+       COLOR_Luv2BGR      = 58,
+       COLOR_Luv2RGB      = 59,
+       COLOR_HLS2BGR      = 60,
+       COLOR_HLS2RGB      = 61,
+
+       COLOR_BGR2HSV_FULL = 66,
+       COLOR_RGB2HSV_FULL = 67,
+       COLOR_BGR2HLS_FULL = 68,
+       COLOR_RGB2HLS_FULL = 69,
+
+       COLOR_HSV2BGR_FULL = 70,
+       COLOR_HSV2RGB_FULL = 71,
+       COLOR_HLS2BGR_FULL = 72,
+       COLOR_HLS2RGB_FULL = 73,
+
+       COLOR_LBGR2Lab     = 74,
+       COLOR_LRGB2Lab     = 75,
+       COLOR_LBGR2Luv     = 76,
+       COLOR_LRGB2Luv     = 77,
+
+       COLOR_Lab2LBGR     = 78,
+       COLOR_Lab2LRGB     = 79,
+       COLOR_Luv2LBGR     = 80,
+       COLOR_Luv2LRGB     = 81,
+
+       COLOR_BGR2YUV      = 82,
+       COLOR_RGB2YUV      = 83,
+       COLOR_YUV2BGR      = 84,
+       COLOR_YUV2RGB      = 85,
+
+       // YUV 4:2:0 family to RGB
+       COLOR_YUV2RGB_NV12  = 90,
+       COLOR_YUV2BGR_NV12  = 91,
+       COLOR_YUV2RGB_NV21  = 92,
+       COLOR_YUV2BGR_NV21  = 93,
+       COLOR_YUV420sp2RGB  = COLOR_YUV2RGB_NV21,
+       COLOR_YUV420sp2BGR  = COLOR_YUV2BGR_NV21,
+
+       COLOR_YUV2RGBA_NV12 = 94,
+       COLOR_YUV2BGRA_NV12 = 95,
+       COLOR_YUV2RGBA_NV21 = 96,
+       COLOR_YUV2BGRA_NV21 = 97,
+       COLOR_YUV420sp2RGBA = COLOR_YUV2RGBA_NV21,
+       COLOR_YUV420sp2BGRA = COLOR_YUV2BGRA_NV21,
+
+       COLOR_YUV2RGB_YV12  = 98,
+       COLOR_YUV2BGR_YV12  = 99,
+       COLOR_YUV2RGB_IYUV  = 100,
+       COLOR_YUV2BGR_IYUV  = 101,
+       COLOR_YUV2RGB_I420  = COLOR_YUV2RGB_IYUV,
+       COLOR_YUV2BGR_I420  = COLOR_YUV2BGR_IYUV,
+       COLOR_YUV420p2RGB   = COLOR_YUV2RGB_YV12,
+       COLOR_YUV420p2BGR   = COLOR_YUV2BGR_YV12,
+
+       COLOR_YUV2RGBA_YV12 = 102,
+       COLOR_YUV2BGRA_YV12 = 103,
+       COLOR_YUV2RGBA_IYUV = 104,
+       COLOR_YUV2BGRA_IYUV = 105,
+       COLOR_YUV2RGBA_I420 = COLOR_YUV2RGBA_IYUV,
+       COLOR_YUV2BGRA_I420 = COLOR_YUV2BGRA_IYUV,
+       COLOR_YUV420p2RGBA  = COLOR_YUV2RGBA_YV12,
+       COLOR_YUV420p2BGRA  = COLOR_YUV2BGRA_YV12,
+
+       COLOR_YUV2GRAY_420  = 106,
+       COLOR_YUV2GRAY_NV21 = COLOR_YUV2GRAY_420,
+       COLOR_YUV2GRAY_NV12 = COLOR_YUV2GRAY_420,
+       COLOR_YUV2GRAY_YV12 = COLOR_YUV2GRAY_420,
+       COLOR_YUV2GRAY_IYUV = COLOR_YUV2GRAY_420,
+       COLOR_YUV2GRAY_I420 = COLOR_YUV2GRAY_420,
+       COLOR_YUV420sp2GRAY = COLOR_YUV2GRAY_420,
+       COLOR_YUV420p2GRAY  = COLOR_YUV2GRAY_420,
+
+       // YUV 4:2:2 family to RGB
+       COLOR_YUV2RGB_UYVY = 107,
+       COLOR_YUV2BGR_UYVY = 108,
+     //COLOR_YUV2RGB_VYUY = 109,
+     //COLOR_YUV2BGR_VYUY = 110,
+       COLOR_YUV2RGB_Y422 = COLOR_YUV2RGB_UYVY,
+       COLOR_YUV2BGR_Y422 = COLOR_YUV2BGR_UYVY,
+       COLOR_YUV2RGB_UYNV = COLOR_YUV2RGB_UYVY,
+       COLOR_YUV2BGR_UYNV = COLOR_YUV2BGR_UYVY,
+
+       COLOR_YUV2RGBA_UYVY = 111,
+       COLOR_YUV2BGRA_UYVY = 112,
+     //COLOR_YUV2RGBA_VYUY = 113,
+     //COLOR_YUV2BGRA_VYUY = 114,
+       COLOR_YUV2RGBA_Y422 = COLOR_YUV2RGBA_UYVY,
+       COLOR_YUV2BGRA_Y422 = COLOR_YUV2BGRA_UYVY,
+       COLOR_YUV2RGBA_UYNV = COLOR_YUV2RGBA_UYVY,
+       COLOR_YUV2BGRA_UYNV = COLOR_YUV2BGRA_UYVY,
+
+       COLOR_YUV2RGB_YUY2 = 115,
+       COLOR_YUV2BGR_YUY2 = 116,
+       COLOR_YUV2RGB_YVYU = 117,
+       COLOR_YUV2BGR_YVYU = 118,
+       COLOR_YUV2RGB_YUYV = COLOR_YUV2RGB_YUY2,
+       COLOR_YUV2BGR_YUYV = COLOR_YUV2BGR_YUY2,
+       COLOR_YUV2RGB_YUNV = COLOR_YUV2RGB_YUY2,
+       COLOR_YUV2BGR_YUNV = COLOR_YUV2BGR_YUY2,
+
+       COLOR_YUV2RGBA_YUY2 = 119,
+       COLOR_YUV2BGRA_YUY2 = 120,
+       COLOR_YUV2RGBA_YVYU = 121,
+       COLOR_YUV2BGRA_YVYU = 122,
+       COLOR_YUV2RGBA_YUYV = COLOR_YUV2RGBA_YUY2,
+       COLOR_YUV2BGRA_YUYV = COLOR_YUV2BGRA_YUY2,
+       COLOR_YUV2RGBA_YUNV = COLOR_YUV2RGBA_YUY2,
+       COLOR_YUV2BGRA_YUNV = COLOR_YUV2BGRA_YUY2,
+
+       COLOR_YUV2GRAY_UYVY = 123,
+       COLOR_YUV2GRAY_YUY2 = 124,
+     //CV_YUV2GRAY_VYUY    = CV_YUV2GRAY_UYVY,
+       COLOR_YUV2GRAY_Y422 = COLOR_YUV2GRAY_UYVY,
+       COLOR_YUV2GRAY_UYNV = COLOR_YUV2GRAY_UYVY,
+       COLOR_YUV2GRAY_YVYU = COLOR_YUV2GRAY_YUY2,
+       COLOR_YUV2GRAY_YUYV = COLOR_YUV2GRAY_YUY2,
+       COLOR_YUV2GRAY_YUNV = COLOR_YUV2GRAY_YUY2,
+
+       // alpha premultiplication
+       COLOR_RGBA2mRGBA    = 125,
+       COLOR_mRGBA2RGBA    = 126,
+
+       // RGB to YUV 4:2:0 family
+       COLOR_RGB2YUV_I420  = 127,
+       COLOR_BGR2YUV_I420  = 128,
+       COLOR_RGB2YUV_IYUV  = COLOR_RGB2YUV_I420,
+       COLOR_BGR2YUV_IYUV  = COLOR_BGR2YUV_I420,
+
+       COLOR_RGBA2YUV_I420 = 129,
+       COLOR_BGRA2YUV_I420 = 130,
+       COLOR_RGBA2YUV_IYUV = COLOR_RGBA2YUV_I420,
+       COLOR_BGRA2YUV_IYUV = COLOR_BGRA2YUV_I420,
+       COLOR_RGB2YUV_YV12  = 131,
+       COLOR_BGR2YUV_YV12  = 132,
+       COLOR_RGBA2YUV_YV12 = 133,
+       COLOR_BGRA2YUV_YV12 = 134,
+
+       // Demosaicing
+       COLOR_BayerBG2BGR = 46,
+       COLOR_BayerGB2BGR = 47,
+       COLOR_BayerRG2BGR = 48,
+       COLOR_BayerGR2BGR = 49,
+
+       COLOR_BayerBG2RGB = COLOR_BayerRG2BGR,
+       COLOR_BayerGB2RGB = COLOR_BayerGR2BGR,
+       COLOR_BayerRG2RGB = COLOR_BayerBG2BGR,
+       COLOR_BayerGR2RGB = COLOR_BayerGB2BGR,
+
+       COLOR_BayerBG2GRAY = 86,
+       COLOR_BayerGB2GRAY = 87,
+       COLOR_BayerRG2GRAY = 88,
+       COLOR_BayerGR2GRAY = 89,
+
+       // Demosaicing using Variable Number of Gradients
+       COLOR_BayerBG2BGR_VNG = 62,
+       COLOR_BayerGB2BGR_VNG = 63,
+       COLOR_BayerRG2BGR_VNG = 64,
+       COLOR_BayerGR2BGR_VNG = 65,
+
+       COLOR_BayerBG2RGB_VNG = COLOR_BayerRG2BGR_VNG,
+       COLOR_BayerGB2RGB_VNG = COLOR_BayerGR2BGR_VNG,
+       COLOR_BayerRG2RGB_VNG = COLOR_BayerBG2BGR_VNG,
+       COLOR_BayerGR2RGB_VNG = COLOR_BayerGB2BGR_VNG,
+
+       // Edge-Aware Demosaicing
+       COLOR_BayerBG2BGR_EA  = 135,
+       COLOR_BayerGB2BGR_EA  = 136,
+       COLOR_BayerRG2BGR_EA  = 137,
+       COLOR_BayerGR2BGR_EA  = 138,
+
+       COLOR_BayerBG2RGB_EA  = COLOR_BayerRG2BGR_EA,
+       COLOR_BayerGB2RGB_EA  = COLOR_BayerGR2BGR_EA,
+       COLOR_BayerRG2RGB_EA  = COLOR_BayerBG2BGR_EA,
+       COLOR_BayerGR2RGB_EA  = COLOR_BayerGB2BGR_EA,
+
+
+       COLOR_COLORCVT_MAX  = 139
+};
+
 
-//! 1D interpolation function: returns coordinate of the "donor" pixel for the specified location p.
-CV_EXPORTS_W int borderInterpolate( int p, int len, int borderType );
 
 /*!
  The Base Class for 1D or Row-wise Filters
@@ -81,9 +481,10 @@ public:
     //! the destructor
     virtual ~BaseRowFilter();
     //! the filtering operator. Must be overrided in the derived classes. The horizontal border interpolation is done outside of the class.
-    virtual void operator()(const uchar* src, uchar* dst,
-                            int width, int cn) = 0;
-    int ksize, anchor;
+    virtual void operator()(const uchar* src, uchar* dst, int width, int cn) = 0;
+
+    int ksize;
+    int anchor;
 };
 
 
@@ -108,13 +509,15 @@ public:
     //! the destructor
     virtual ~BaseColumnFilter();
     //! the filtering operator. Must be overrided in the derived classes. The vertical border interpolation is done outside of the class.
-    virtual void operator()(const uchar** src, uchar* dst, int dststep,
-                            int dstcount, int width) = 0;
+    virtual void operator()(const uchar** src, uchar* dst, int dststep, int dstcount, int width) = 0;
     //! resets the internal buffers, if any
     virtual void reset();
-    int ksize, anchor;
+
+    int ksize;
+    int anchor;
 };
 
+
 /*!
  The Base Class for Non-Separable 2D Filters.
 
@@ -134,14 +537,15 @@ public:
     //! the destructor
     virtual ~BaseFilter();
     //! the filtering operator. The horizontal and the vertical border interpolation is done outside of the class.
-    virtual void operator()(const uchar** src, uchar* dst, int dststep,
-                            int dstcount, int width, int cn) = 0;
+    virtual void operator()(const uchar** src, uchar* dst, int dststep, int dstcount, int width, int cn) = 0;
     //! resets the internal buffers, if any
     virtual void reset();
+
     Size ksize;
     Point anchor;
 };
 
+
 /*!
  The Main Class for Image Filtering.
 
@@ -225,9 +629,9 @@ public:
                  const Ptr<BaseRowFilter>& _rowFilter,
                  const Ptr<BaseColumnFilter>& _columnFilter,
                  int srcType, int dstType, int bufType,
-                 int _rowBorderType=BORDER_REPLICATE,
-                 int _columnBorderType=-1,
-                 const Scalar& _borderValue=Scalar());
+                 int _rowBorderType = BORDER_REPLICATE,
+                 int _columnBorderType = -1,
+                 const Scalar& _borderValue = Scalar());
     //! the destructor
     virtual ~FilterEngine();
     //! reinitializes the engine. The previously assigned filters are released.
@@ -235,42 +639,52 @@ public:
               const Ptr<BaseRowFilter>& _rowFilter,
               const Ptr<BaseColumnFilter>& _columnFilter,
               int srcType, int dstType, int bufType,
-              int _rowBorderType=BORDER_REPLICATE, int _columnBorderType=-1,
-              const Scalar& _borderValue=Scalar());
+              int _rowBorderType = BORDER_REPLICATE,
+              int _columnBorderType = -1,
+              const Scalar& _borderValue = Scalar());
     //! starts filtering of the specified ROI of an image of size wholeSize.
-    virtual int start(Size wholeSize, Rect roi, int maxBufRows=-1);
+    virtual int start(Size wholeSize, Rect roi, int maxBufRows = -1);
     //! starts filtering of the specified ROI of the specified image.
-    virtual int start(const Mat& src, const Rect& srcRoi=Rect(0,0,-1,-1),
-                      bool isolated=false, int maxBufRows=-1);
+    virtual int start(const Mat& src, const Rect& srcRoi = Rect(0,0,-1,-1),
+                      bool isolated = false, int maxBufRows = -1);
     //! processes the next srcCount rows of the image.
     virtual int proceed(const uchar* src, int srcStep, int srcCount,
                         uchar* dst, int dstStep);
     //! applies filter to the specified ROI of the image. if srcRoi=(0,0,-1,-1), the whole image is filtered.
     virtual void apply( const Mat& src, Mat& dst,
-                        const Rect& srcRoi=Rect(0,0,-1,-1),
-                        Point dstOfs=Point(0,0),
-                        bool isolated=false);
+                        const Rect& srcRoi = Rect(0,0,-1,-1),
+                        Point dstOfs = Point(0,0),
+                        bool isolated = false);
     //! returns true if the filter is separable
-    bool isSeparable() const { return (const BaseFilter*)filter2D == 0; }
+    bool isSeparable() const { return !filter2D; }
     //! returns the number
     int remainingInputRows() const;
     int remainingOutputRows() const;
 
-    int srcType, dstType, bufType;
+    int srcType;
+    int dstType;
+    int bufType;
     Size ksize;
     Point anchor;
     int maxWidth;
     Size wholeSize;
     Rect roi;
-    int dx1, dx2;
-    int rowBorderType, columnBorderType;
+    int dx1;
+    int dx2;
+    int rowBorderType;
+    int columnBorderType;
     std::vector<int> borderTab;
     int borderElemSize;
     std::vector<uchar> ringBuf;
     std::vector<uchar> srcRow;
     std::vector<uchar> constBorderValue;
     std::vector<uchar> constBorderRow;
-    int bufStep, startY, startY0, endY, rowCount, dstY;
+    int bufStep;
+    int startY;
+    int startY0;
+    int endY;
+    int rowCount;
+    int dstY;
     std::vector<uchar*> rows;
 
     Ptr<BaseFilter> filter2D;
@@ -278,947 +692,143 @@ public:
     Ptr<BaseColumnFilter> columnFilter;
 };
 
-//! type of the kernel
-enum { KERNEL_GENERAL=0, KERNEL_SYMMETRICAL=1, KERNEL_ASYMMETRICAL=2,
-       KERNEL_SMOOTH=4, KERNEL_INTEGER=8 };
-
-//! returns type (one of KERNEL_*) of 1D or 2D kernel specified by its coefficients.
-CV_EXPORTS int getKernelType(InputArray kernel, Point anchor);
-
-//! returns the primitive row filter with the specified kernel
-CV_EXPORTS Ptr<BaseRowFilter> getLinearRowFilter(int srcType, int bufType,
-                                            InputArray kernel, int anchor,
-                                            int symmetryType);
-
-//! returns the primitive column filter with the specified kernel
-CV_EXPORTS Ptr<BaseColumnFilter> getLinearColumnFilter(int bufType, int dstType,
-                                            InputArray kernel, int anchor,
-                                            int symmetryType, double delta=0,
-                                            int bits=0);
-
-//! returns 2D filter with the specified kernel
-CV_EXPORTS Ptr<BaseFilter> getLinearFilter(int srcType, int dstType,
-                                           InputArray kernel,
-                                           Point anchor=Point(-1,-1),
-                                           double delta=0, int bits=0);
-
-//! returns the separable linear filter engine
-CV_EXPORTS Ptr<FilterEngine> createSeparableLinearFilter(int srcType, int dstType,
-                          InputArray rowKernel, InputArray columnKernel,
-                          Point anchor=Point(-1,-1), double delta=0,
-                          int rowBorderType=BORDER_DEFAULT,
-                          int columnBorderType=-1,
-                          const Scalar& borderValue=Scalar());
-
-//! returns the non-separable linear filter engine
-CV_EXPORTS Ptr<FilterEngine> createLinearFilter(int srcType, int dstType,
-                 InputArray kernel, Point _anchor=Point(-1,-1),
-                 double delta=0, int rowBorderType=BORDER_DEFAULT,
-                 int columnBorderType=-1, const Scalar& borderValue=Scalar());
-
-//! returns the Gaussian kernel with the specified parameters
-CV_EXPORTS_W Mat getGaussianKernel( int ksize, double sigma, int ktype=CV_64F );
-
-//! returns the Gaussian filter engine
-CV_EXPORTS Ptr<FilterEngine> createGaussianFilter( int type, Size ksize,
-                                    double sigma1, double sigma2=0,
-                                    int borderType=BORDER_DEFAULT);
-//! initializes kernels of the generalized Sobel operator
-CV_EXPORTS_W void getDerivKernels( OutputArray kx, OutputArray ky,
-                                   int dx, int dy, int ksize,
-                                   bool normalize=false, int ktype=CV_32F );
-//! returns filter engine for the generalized Sobel operator
-CV_EXPORTS Ptr<FilterEngine> createDerivFilter( int srcType, int dstType,
-                                        int dx, int dy, int ksize,
-                                        int borderType=BORDER_DEFAULT );
-//! returns horizontal 1D box filter
-CV_EXPORTS Ptr<BaseRowFilter> getRowSumFilter(int srcType, int sumType,
-                                              int ksize, int anchor=-1);
-//! returns vertical 1D box filter
-CV_EXPORTS Ptr<BaseColumnFilter> getColumnSumFilter( int sumType, int dstType,
-                                                     int ksize, int anchor=-1,
-                                                     double scale=1);
-//! returns box filter engine
-CV_EXPORTS Ptr<FilterEngine> createBoxFilter( int srcType, int dstType, Size ksize,
-                                              Point anchor=Point(-1,-1),
-                                              bool normalize=true,
-                                              int borderType=BORDER_DEFAULT);
-
-//! returns the Gabor kernel with the specified parameters
-CV_EXPORTS_W Mat getGaborKernel( Size ksize, double sigma, double theta, double lambd,
-                                 double gamma, double psi=CV_PI*0.5, int ktype=CV_64F );
-
-//! type of morphological operation
-enum { MORPH_ERODE=CV_MOP_ERODE, MORPH_DILATE=CV_MOP_DILATE,
-       MORPH_OPEN=CV_MOP_OPEN, MORPH_CLOSE=CV_MOP_CLOSE,
-       MORPH_GRADIENT=CV_MOP_GRADIENT, MORPH_TOPHAT=CV_MOP_TOPHAT,
-       MORPH_BLACKHAT=CV_MOP_BLACKHAT };
-
-//! returns horizontal 1D morphological filter
-CV_EXPORTS Ptr<BaseRowFilter> getMorphologyRowFilter(int op, int type, int ksize, int anchor=-1);
-//! returns vertical 1D morphological filter
-CV_EXPORTS Ptr<BaseColumnFilter> getMorphologyColumnFilter(int op, int type, int ksize, int anchor=-1);
-//! returns 2D morphological filter
-CV_EXPORTS Ptr<BaseFilter> getMorphologyFilter(int op, int type, InputArray kernel,
-                                               Point anchor=Point(-1,-1));
-
-//! returns "magic" border value for erosion and dilation. It is automatically transformed to Scalar::all(-DBL_MAX) for dilation.
-static inline Scalar morphologyDefaultBorderValue() { return Scalar::all(DBL_MAX); }
-
-//! returns morphological filter engine. Only MORPH_ERODE and MORPH_DILATE are supported.
-CV_EXPORTS Ptr<FilterEngine> createMorphologyFilter(int op, int type, InputArray kernel,
-                    Point anchor=Point(-1,-1), int rowBorderType=BORDER_CONSTANT,
-                    int columnBorderType=-1,
-                    const Scalar& borderValue=morphologyDefaultBorderValue());
-
-//! shape of the structuring element
-enum { MORPH_RECT=0, MORPH_CROSS=1, MORPH_ELLIPSE=2 };
-//! returns structuring element of the specified shape and size
-CV_EXPORTS_W Mat getStructuringElement(int shape, Size ksize, Point anchor=Point(-1,-1));
-
-template<> CV_EXPORTS void Ptr<IplConvKernel>::delete_obj();
-
-//! copies 2D array to a larger destination array with extrapolation of the outer part of src using the specified border mode
-CV_EXPORTS_W void copyMakeBorder( InputArray src, OutputArray dst,
-                                int top, int bottom, int left, int right,
-                                int borderType, const Scalar& value=Scalar() );
-
-//! smooths the image using median filter.
-CV_EXPORTS_W void medianBlur( InputArray src, OutputArray dst, int ksize );
-//! smooths the image using Gaussian filter.
-CV_EXPORTS_W void GaussianBlur( InputArray src,
-                                               OutputArray dst, Size ksize,
-                                               double sigmaX, double sigmaY=0,
-                                               int borderType=BORDER_DEFAULT );
-//! smooths the image using bilateral filter
-CV_EXPORTS_W void bilateralFilter( InputArray src, OutputArray dst, int d,
-                                   double sigmaColor, double sigmaSpace,
-                                   int borderType=BORDER_DEFAULT );
-//! smooths the image using the box filter. Each pixel is processed in O(1) time
-CV_EXPORTS_W void boxFilter( InputArray src, OutputArray dst, int ddepth,
-                             Size ksize, Point anchor=Point(-1,-1),
-                             bool normalize=true,
-                             int borderType=BORDER_DEFAULT );
-//! a synonym for normalized box filter
-CV_EXPORTS_W void blur( InputArray src, OutputArray dst,
-                        Size ksize, Point anchor=Point(-1,-1),
-                        int borderType=BORDER_DEFAULT );
-
-//! applies non-separable 2D linear filter to the image
-CV_EXPORTS_W void filter2D( InputArray src, OutputArray dst, int ddepth,
-                            InputArray kernel, Point anchor=Point(-1,-1),
-                            double delta=0, int borderType=BORDER_DEFAULT );
-
-//! applies separable 2D linear filter to the image
-CV_EXPORTS_W void sepFilter2D( InputArray src, OutputArray dst, int ddepth,
-                               InputArray kernelX, InputArray kernelY,
-                               Point anchor=Point(-1,-1),
-                               double delta=0, int borderType=BORDER_DEFAULT );
-
-//! applies generalized Sobel operator to the image
-CV_EXPORTS_W void Sobel( InputArray src, OutputArray dst, int ddepth,
-                         int dx, int dy, int ksize=3,
-                         double scale=1, double delta=0,
-                         int borderType=BORDER_DEFAULT );
-
-//! applies the vertical or horizontal Scharr operator to the image
-CV_EXPORTS_W void Scharr( InputArray src, OutputArray dst, int ddepth,
-                          int dx, int dy, double scale=1, double delta=0,
-                          int borderType=BORDER_DEFAULT );
-
-//! applies Laplacian operator to the image
-CV_EXPORTS_W void Laplacian( InputArray src, OutputArray dst, int ddepth,
-                             int ksize=1, double scale=1, double delta=0,
-                             int borderType=BORDER_DEFAULT );
-
-//! applies Canny edge detector and produces the edge map.
-CV_EXPORTS_W void Canny( InputArray image, OutputArray edges,
-                         double threshold1, double threshold2,
-                         int apertureSize=3, bool L2gradient=false );
-
-//! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria
-CV_EXPORTS_W void cornerMinEigenVal( InputArray src, OutputArray dst,
-                                   int blockSize, int ksize=3,
-                                   int borderType=BORDER_DEFAULT );
-
-//! computes Harris cornerness criteria at each image pixel
-CV_EXPORTS_W void cornerHarris( InputArray src, OutputArray dst, int blockSize,
-                                int ksize, double k,
-                                int borderType=BORDER_DEFAULT );
-
-// low-level function for computing eigenvalues and eigenvectors of 2x2 matrices
-CV_EXPORTS void eigen2x2( const float* a, float* e, int n );
-
-//! computes both eigenvalues and the eigenvectors of 2x2 derivative covariation matrix  at each pixel. The output is stored as 6-channel matrix.
-CV_EXPORTS_W void cornerEigenValsAndVecs( InputArray src, OutputArray dst,
-                                          int blockSize, int ksize,
-                                          int borderType=BORDER_DEFAULT );
-
-//! computes another complex cornerness criteria at each pixel
-CV_EXPORTS_W void preCornerDetect( InputArray src, OutputArray dst, int ksize,
-                                   int borderType=BORDER_DEFAULT );
-
-//! adjusts the corner locations with sub-pixel accuracy to maximize the certain cornerness criteria
-CV_EXPORTS_W void cornerSubPix( InputArray image, InputOutputArray corners,
-                                Size winSize, Size zeroZone,
-                                TermCriteria criteria );
-
-//! finds the strong enough corners where the cornerMinEigenVal() or cornerHarris() report the local maxima
-CV_EXPORTS_W void goodFeaturesToTrack( InputArray image, OutputArray corners,
-                                     int maxCorners, double qualityLevel, double minDistance,
-                                     InputArray mask=noArray(), int blockSize=3,
-                                     bool useHarrisDetector=false, double k=0.04 );
-
-//! finds lines in the black-n-white image using the standard or pyramid Hough transform
-CV_EXPORTS_W void HoughLines( InputArray image, OutputArray lines,
-                              double rho, double theta, int threshold,
-                              double srn=0, double stn=0 );
-
-//! finds line segments in the black-n-white image using probabalistic Hough transform
-CV_EXPORTS_W void HoughLinesP( InputArray image, OutputArray lines,
-                               double rho, double theta, int threshold,
-                               double minLineLength=0, double maxLineGap=0 );
-
-//! finds circles in the grayscale image using 2+1 gradient Hough transform
-CV_EXPORTS_W void HoughCircles( InputArray image, OutputArray circles,
-                               int method, double dp, double minDist,
-                               double param1=100, double param2=100,
-                               int minRadius=0, int maxRadius=0 );
-
-enum
-{
-    GHT_POSITION = 0,
-    GHT_SCALE = 1,
-    GHT_ROTATION = 2
-};
 
 //! finds arbitrary template in the grayscale image using Generalized Hough Transform
-//! Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
-//! Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
 class CV_EXPORTS GeneralizedHough : public Algorithm
 {
 public:
-    static Ptr<GeneralizedHough> create(int method);
-
-    virtual ~GeneralizedHough();
-
     //! set template to search
-    void setTemplate(InputArray templ, int cannyThreshold = 100, Point templCenter = Point(-1, -1));
-    void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter = Point(-1, -1));
+    virtual void setTemplate(InputArray templ, Point templCenter = Point(-1, -1)) = 0;
+    virtual void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter = Point(-1, -1)) = 0;
 
     //! find template on image
-    void detect(InputArray image, OutputArray positions, OutputArray votes = cv::noArray(), int cannyThreshold = 100);
-    void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes = cv::noArray());
+    virtual void detect(InputArray image, OutputArray positions, OutputArray votes = noArray()) = 0;
+    virtual void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes = noArray()) = 0;
 
-    void release();
+    //! Canny low threshold.
+    virtual void setCannyLowThresh(int cannyLowThresh) = 0;
+    virtual int getCannyLowThresh() const = 0;
 
-protected:
-    virtual void setTemplateImpl(const Mat& edges, const Mat& dx, const Mat& dy, Point templCenter) = 0;
-    virtual void detectImpl(const Mat& edges, const Mat& dx, const Mat& dy, OutputArray positions, OutputArray votes) = 0;
-    virtual void releaseImpl() = 0;
+    //! Canny high threshold.
+    virtual void setCannyHighThresh(int cannyHighThresh) = 0;
+    virtual int getCannyHighThresh() const = 0;
 
-private:
-    Mat edges_, dx_, dy_;
+    //! Minimum distance between the centers of the detected objects.
+    virtual void setMinDist(double minDist) = 0;
+    virtual double getMinDist() const = 0;
+
+    //! Inverse ratio of the accumulator resolution to the image resolution.
+    virtual void setDp(double dp) = 0;
+    virtual double getDp() const = 0;
+
+    //! Maximal size of inner buffers.
+    virtual void setMaxBufferSize(int maxBufferSize) = 0;
+    virtual int getMaxBufferSize() const = 0;
 };
 
-//! erodes the image (applies the local minimum operator)
-CV_EXPORTS_W void erode( InputArray src, OutputArray dst, InputArray kernel,
-                         Point anchor=Point(-1,-1), int iterations=1,
-                         int borderType=BORDER_CONSTANT,
-                         const Scalar& borderValue=morphologyDefaultBorderValue() );
-
-//! dilates the image (applies the local maximum operator)
-CV_EXPORTS_W void dilate( InputArray src, OutputArray dst, InputArray kernel,
-                          Point anchor=Point(-1,-1), int iterations=1,
-                          int borderType=BORDER_CONSTANT,
-                          const Scalar& borderValue=morphologyDefaultBorderValue() );
-
-//! applies an advanced morphological operation to the image
-CV_EXPORTS_W void morphologyEx( InputArray src, OutputArray dst,
-                                int op, InputArray kernel,
-                                Point anchor=Point(-1,-1), int iterations=1,
-                                int borderType=BORDER_CONSTANT,
-                                const Scalar& borderValue=morphologyDefaultBorderValue() );
-
-//! interpolation algorithm
-enum
-{
-    INTER_NEAREST=CV_INTER_NN, //!< nearest neighbor interpolation
-    INTER_LINEAR=CV_INTER_LINEAR, //!< bilinear interpolation
-    INTER_CUBIC=CV_INTER_CUBIC, //!< bicubic interpolation
-    INTER_AREA=CV_INTER_AREA, //!< area-based (or super) interpolation
-    INTER_LANCZOS4=CV_INTER_LANCZOS4, //!< Lanczos interpolation over 8x8 neighborhood
-    INTER_MAX=7,
-    WARP_INVERSE_MAP=CV_WARP_INVERSE_MAP
-};
-
-//! resizes the image
-CV_EXPORTS_W void resize( InputArray src, OutputArray dst,
-                          Size dsize, double fx=0, double fy=0,
-                          int interpolation=INTER_LINEAR );
-
-//! warps the image using affine transformation
-CV_EXPORTS_W void warpAffine( InputArray src, OutputArray dst,
-                              InputArray M, Size dsize,
-                              int flags=INTER_LINEAR,
-                              int borderMode=BORDER_CONSTANT,
-                              const Scalar& borderValue=Scalar());
-
-//! warps the image using perspective transformation
-CV_EXPORTS_W void warpPerspective( InputArray src, OutputArray dst,
-                                   InputArray M, Size dsize,
-                                   int flags=INTER_LINEAR,
-                                   int borderMode=BORDER_CONSTANT,
-                                   const Scalar& borderValue=Scalar());
-
-enum
-{
-    INTER_BITS=5, INTER_BITS2=INTER_BITS*2,
-    INTER_TAB_SIZE=(1<<INTER_BITS),
-    INTER_TAB_SIZE2=INTER_TAB_SIZE*INTER_TAB_SIZE
-};
-
-//! warps the image using the precomputed maps. The maps are stored in either floating-point or integer fixed-point format
-CV_EXPORTS_W void remap( InputArray src, OutputArray dst,
-                         InputArray map1, InputArray map2,
-                         int interpolation, int borderMode=BORDER_CONSTANT,
-                         const Scalar& borderValue=Scalar());
-
-//! converts maps for remap from floating-point to fixed-point format or backwards
-CV_EXPORTS_W void convertMaps( InputArray map1, InputArray map2,
-                               OutputArray dstmap1, OutputArray dstmap2,
-                               int dstmap1type, bool nninterpolation=false );
-
-//! returns 2x3 affine transformation matrix for the planar rotation.
-CV_EXPORTS_W Mat getRotationMatrix2D( Point2f center, double angle, double scale );
-//! returns 3x3 perspective transformation for the corresponding 4 point pairs.
-CV_EXPORTS Mat getPerspectiveTransform( const Point2f src[], const Point2f dst[] );
-//! returns 2x3 affine transformation for the corresponding 3 point pairs.
-CV_EXPORTS Mat getAffineTransform( const Point2f src[], const Point2f dst[] );
-//! computes 2x3 affine transformation matrix that is inverse to the specified 2x3 affine transformation.
-CV_EXPORTS_W void invertAffineTransform( InputArray M, OutputArray iM );
-
-CV_EXPORTS_W Mat getPerspectiveTransform( InputArray src, InputArray dst );
-CV_EXPORTS_W Mat getAffineTransform( InputArray src, InputArray dst );
-
-//! extracts rectangle from the image at sub-pixel location
-CV_EXPORTS_W void getRectSubPix( InputArray image, Size patchSize,
-                                 Point2f center, OutputArray patch, int patchType=-1 );
-
-//! computes the integral image
-CV_EXPORTS_W void integral( InputArray src, OutputArray sum, int sdepth=-1 );
-
-//! computes the integral image and integral for the squared image
-CV_EXPORTS_AS(integral2) void integral( InputArray src, OutputArray sum,
-                                        OutputArray sqsum, int sdepth=-1 );
-//! computes the integral image, integral for the squared image and the tilted integral image
-CV_EXPORTS_AS(integral3) void integral( InputArray src, OutputArray sum,
-                                        OutputArray sqsum, OutputArray tilted,
-                                        int sdepth=-1 );
-
-//! adds image to the accumulator (dst += src). Unlike cv::add, dst and src can have different types.
-CV_EXPORTS_W void accumulate( InputArray src, InputOutputArray dst,
-                              InputArray mask=noArray() );
-//! adds squared src image to the accumulator (dst += src*src).
-CV_EXPORTS_W void accumulateSquare( InputArray src, InputOutputArray dst,
-                                    InputArray mask=noArray() );
-//! adds product of the 2 images to the accumulator (dst += src1*src2).
-CV_EXPORTS_W void accumulateProduct( InputArray src1, InputArray src2,
-                                     InputOutputArray dst, InputArray mask=noArray() );
-//! updates the running average (dst = dst*(1-alpha) + src*alpha)
-CV_EXPORTS_W void accumulateWeighted( InputArray src, InputOutputArray dst,
-                                      double alpha, InputArray mask=noArray() );
-
-//! computes PSNR image/video quality metric
-CV_EXPORTS_W double PSNR(InputArray src1, InputArray src2);
-
-CV_EXPORTS_W Point2d phaseCorrelate(InputArray src1, InputArray src2,
-                                    InputArray window = noArray(), CV_OUT double* response=0);
-CV_EXPORTS_W void createHanningWindow(OutputArray dst, Size winSize, int type);
-
-//! type of the threshold operation
-enum { THRESH_BINARY=CV_THRESH_BINARY, THRESH_BINARY_INV=CV_THRESH_BINARY_INV,
-       THRESH_TRUNC=CV_THRESH_TRUNC, THRESH_TOZERO=CV_THRESH_TOZERO,
-       THRESH_TOZERO_INV=CV_THRESH_TOZERO_INV, THRESH_MASK=CV_THRESH_MASK,
-       THRESH_OTSU=CV_THRESH_OTSU };
-
-//! applies fixed threshold to the image
-CV_EXPORTS_W double threshold( InputArray src, OutputArray dst,
-                               double thresh, double maxval, int type );
-
-//! adaptive threshold algorithm
-enum { ADAPTIVE_THRESH_MEAN_C=0, ADAPTIVE_THRESH_GAUSSIAN_C=1 };
-
-//! applies variable (adaptive) threshold to the image
-CV_EXPORTS_W void adaptiveThreshold( InputArray src, OutputArray dst,
-                                     double maxValue, int adaptiveMethod,
-                                     int thresholdType, int blockSize, double C );
-
-//! smooths and downsamples the image
-CV_EXPORTS_W void pyrDown( InputArray src, OutputArray dst,
-                           const Size& dstsize=Size(), int borderType=BORDER_DEFAULT );
-//! upsamples and smoothes the image
-CV_EXPORTS_W void pyrUp( InputArray src, OutputArray dst,
-                         const Size& dstsize=Size(), int borderType=BORDER_DEFAULT );
-
-//! builds the gaussian pyramid using pyrDown() as a basic operation
-CV_EXPORTS void buildPyramid( InputArray src, OutputArrayOfArrays dst,
-                              int maxlevel, int borderType=BORDER_DEFAULT );
-
-//! corrects lens distortion for the given camera matrix and distortion coefficients
-CV_EXPORTS_W void undistort( InputArray src, OutputArray dst,
-                             InputArray cameraMatrix,
-                             InputArray distCoeffs,
-                             InputArray newCameraMatrix=noArray() );
-
-//! initializes maps for cv::remap() to correct lens distortion and optionally rectify the image
-CV_EXPORTS_W void initUndistortRectifyMap( InputArray cameraMatrix, InputArray distCoeffs,
-                           InputArray R, InputArray newCameraMatrix,
-                           Size size, int m1type, OutputArray map1, OutputArray map2 );
-
-enum
-{
-    PROJ_SPHERICAL_ORTHO = 0,
-    PROJ_SPHERICAL_EQRECT = 1
-};
-
-//! initializes maps for cv::remap() for wide-angle
-CV_EXPORTS_W float initWideAngleProjMap( InputArray cameraMatrix, InputArray distCoeffs,
-                                         Size imageSize, int destImageWidth,
-                                         int m1type, OutputArray map1, OutputArray map2,
-                                         int projType=PROJ_SPHERICAL_EQRECT, double alpha=0);
-
-//! returns the default new camera matrix (by default it is the same as cameraMatrix unless centerPricipalPoint=true)
-CV_EXPORTS_W Mat getDefaultNewCameraMatrix( InputArray cameraMatrix, Size imgsize=Size(),
-                                            bool centerPrincipalPoint=false );
-
-//! returns points' coordinates after lens distortion correction
-CV_EXPORTS_W void undistortPoints( InputArray src, OutputArray dst,
-                                   InputArray cameraMatrix, InputArray distCoeffs,
-                                   InputArray R=noArray(), InputArray P=noArray());
-
-template<> CV_EXPORTS void Ptr<CvHistogram>::delete_obj();
-
-//! computes the joint dense histogram for a set of images.
-CV_EXPORTS void calcHist( const Mat* images, int nimages,
-                          const int* channels, InputArray mask,
-                          OutputArray hist, int dims, const int* histSize,
-                          const float** ranges, bool uniform=true, bool accumulate=false );
-
-//! computes the joint sparse histogram for a set of images.
-CV_EXPORTS void calcHist( const Mat* images, int nimages,
-                          const int* channels, InputArray mask,
-                          SparseMat& hist, int dims,
-                          const int* histSize, const float** ranges,
-                          bool uniform=true, bool accumulate=false );
-
-CV_EXPORTS_W void calcHist( InputArrayOfArrays images,
-                            const std::vector<int>& channels,
-                            InputArray mask, OutputArray hist,
-                            const std::vector<int>& histSize,
-                            const std::vector<float>& ranges,
-                            bool accumulate=false );
-
-//! computes back projection for the set of images
-CV_EXPORTS void calcBackProject( const Mat* images, int nimages,
-                                 const int* channels, InputArray hist,
-                                 OutputArray backProject, const float** ranges,
-                                 double scale=1, bool uniform=true );
-
-//! computes back projection for the set of images
-CV_EXPORTS void calcBackProject( const Mat* images, int nimages,
-                                 const int* channels, const SparseMat& hist,
-                                 OutputArray backProject, const float** ranges,
-                                 double scale=1, bool uniform=true );
-
-CV_EXPORTS_W void calcBackProject( InputArrayOfArrays images, const std::vector<int>& channels,
-                                   InputArray hist, OutputArray dst,
-                                   const std::vector<float>& ranges,
-                                   double scale );
-
-/*CV_EXPORTS void calcBackProjectPatch( const Mat* images, int nimages, const int* channels,
-                                      InputArray hist, OutputArray dst, Size patchSize,
-                                      int method, double factor=1 );
-
-CV_EXPORTS_W void calcBackProjectPatch( InputArrayOfArrays images, const std::vector<int>& channels,
-                                        InputArray hist, OutputArray dst, Size patchSize,
-                                        int method, double factor=1 );*/
-
-//! compares two histograms stored in dense arrays
-CV_EXPORTS_W double compareHist( InputArray H1, InputArray H2, int method );
-
-//! compares two histograms stored in sparse arrays
-CV_EXPORTS double compareHist( const SparseMat& H1, const SparseMat& H2, int method );
-
-//! normalizes the grayscale image brightness and contrast by normalizing its histogram
-CV_EXPORTS_W void equalizeHist( InputArray src, OutputArray dst );
-
-CV_EXPORTS float EMD( InputArray signature1, InputArray signature2,
-                      int distType, InputArray cost=noArray(),
-                      float* lowerBound=0, OutputArray flow=noArray() );
-
-//! segments the image using watershed algorithm
-CV_EXPORTS_W void watershed( InputArray image, InputOutputArray markers );
-
-//! filters image using meanshift algorithm
-CV_EXPORTS_W void pyrMeanShiftFiltering( InputArray src, OutputArray dst,
-                                         double sp, double sr, int maxLevel=1,
-                                         TermCriteria termcrit=TermCriteria(
-                                            TermCriteria::MAX_ITER+TermCriteria::EPS,5,1) );
-
-//! class of the pixel in GrabCut algorithm
-enum
-{
-    GC_BGD    = 0,  //!< background
-    GC_FGD    = 1,  //!< foreground
-    GC_PR_BGD = 2,  //!< most probably background
-    GC_PR_FGD = 3   //!< most probably foreground
-};
-
-//! GrabCut algorithm flags
-enum
-{
-    GC_INIT_WITH_RECT  = 0,
-    GC_INIT_WITH_MASK  = 1,
-    GC_EVAL            = 2
-};
-
-//! segments the image using GrabCut algorithm
-CV_EXPORTS_W void grabCut( InputArray img, InputOutputArray mask, Rect rect,
-                           InputOutputArray bgdModel, InputOutputArray fgdModel,
-                           int iterCount, int mode = GC_EVAL );
-
-enum
-{
-    DIST_LABEL_CCOMP = 0,
-    DIST_LABEL_PIXEL = 1
-};
-
-//! builds the discrete Voronoi diagram
-CV_EXPORTS_AS(distanceTransformWithLabels) void distanceTransform( InputArray src, OutputArray dst,
-                                     OutputArray labels, int distanceType, int maskSize,
-                                     int labelType=DIST_LABEL_CCOMP );
-
-//! computes the distance transform map
-CV_EXPORTS_W void distanceTransform( InputArray src, OutputArray dst,
-                                     int distanceType, int maskSize );
-
-enum { FLOODFILL_FIXED_RANGE = 1 << 16, FLOODFILL_MASK_ONLY = 1 << 17 };
-
-//! fills the semi-uniform image region starting from the specified seed point
-CV_EXPORTS int floodFill( InputOutputArray image,
-                          Point seedPoint, Scalar newVal, CV_OUT Rect* rect=0,
-                          Scalar loDiff=Scalar(), Scalar upDiff=Scalar(),
-                          int flags=4 );
-
-//! fills the semi-uniform image region and/or the mask starting from the specified seed point
-CV_EXPORTS_W int floodFill( InputOutputArray image, InputOutputArray mask,
-                            Point seedPoint, Scalar newVal, CV_OUT Rect* rect=0,
-                            Scalar loDiff=Scalar(), Scalar upDiff=Scalar(),
-                            int flags=4 );
-
-
-enum
-{
-    COLOR_BGR2BGRA    =0,
-    COLOR_RGB2RGBA    =COLOR_BGR2BGRA,
-
-    COLOR_BGRA2BGR    =1,
-    COLOR_RGBA2RGB    =COLOR_BGRA2BGR,
-
-    COLOR_BGR2RGBA    =2,
-    COLOR_RGB2BGRA    =COLOR_BGR2RGBA,
-
-    COLOR_RGBA2BGR    =3,
-    COLOR_BGRA2RGB    =COLOR_RGBA2BGR,
-
-    COLOR_BGR2RGB     =4,
-    COLOR_RGB2BGR     =COLOR_BGR2RGB,
-
-    COLOR_BGRA2RGBA   =5,
-    COLOR_RGBA2BGRA   =COLOR_BGRA2RGBA,
-
-    COLOR_BGR2GRAY    =6,
-    COLOR_RGB2GRAY    =7,
-    COLOR_GRAY2BGR    =8,
-    COLOR_GRAY2RGB    =COLOR_GRAY2BGR,
-    COLOR_GRAY2BGRA   =9,
-    COLOR_GRAY2RGBA   =COLOR_GRAY2BGRA,
-    COLOR_BGRA2GRAY   =10,
-    COLOR_RGBA2GRAY   =11,
-
-    COLOR_BGR2BGR565  =12,
-    COLOR_RGB2BGR565  =13,
-    COLOR_BGR5652BGR  =14,
-    COLOR_BGR5652RGB  =15,
-    COLOR_BGRA2BGR565 =16,
-    COLOR_RGBA2BGR565 =17,
-    COLOR_BGR5652BGRA =18,
-    COLOR_BGR5652RGBA =19,
-
-    COLOR_GRAY2BGR565 =20,
-    COLOR_BGR5652GRAY =21,
-
-    COLOR_BGR2BGR555  =22,
-    COLOR_RGB2BGR555  =23,
-    COLOR_BGR5552BGR  =24,
-    COLOR_BGR5552RGB  =25,
-    COLOR_BGRA2BGR555 =26,
-    COLOR_RGBA2BGR555 =27,
-    COLOR_BGR5552BGRA =28,
-    COLOR_BGR5552RGBA =29,
-
-    COLOR_GRAY2BGR555 =30,
-    COLOR_BGR5552GRAY =31,
-
-    COLOR_BGR2XYZ     =32,
-    COLOR_RGB2XYZ     =33,
-    COLOR_XYZ2BGR     =34,
-    COLOR_XYZ2RGB     =35,
-
-    COLOR_BGR2YCrCb   =36,
-    COLOR_RGB2YCrCb   =37,
-    COLOR_YCrCb2BGR   =38,
-    COLOR_YCrCb2RGB   =39,
-
-    COLOR_BGR2HSV     =40,
-    COLOR_RGB2HSV     =41,
-
-    COLOR_BGR2Lab     =44,
-    COLOR_RGB2Lab     =45,
-
-    COLOR_BayerBG2BGR =46,
-    COLOR_BayerGB2BGR =47,
-    COLOR_BayerRG2BGR =48,
-    COLOR_BayerGR2BGR =49,
-
-    COLOR_BayerBG2RGB =COLOR_BayerRG2BGR,
-    COLOR_BayerGB2RGB =COLOR_BayerGR2BGR,
-    COLOR_BayerRG2RGB =COLOR_BayerBG2BGR,
-    COLOR_BayerGR2RGB =COLOR_BayerGB2BGR,
-
-    COLOR_BGR2Luv     =50,
-    COLOR_RGB2Luv     =51,
-    COLOR_BGR2HLS     =52,
-    COLOR_RGB2HLS     =53,
-
-    COLOR_HSV2BGR     =54,
-    COLOR_HSV2RGB     =55,
-
-    COLOR_Lab2BGR     =56,
-    COLOR_Lab2RGB     =57,
-    COLOR_Luv2BGR     =58,
-    COLOR_Luv2RGB     =59,
-    COLOR_HLS2BGR     =60,
-    COLOR_HLS2RGB     =61,
-
-    COLOR_BayerBG2BGR_VNG =62,
-    COLOR_BayerGB2BGR_VNG =63,
-    COLOR_BayerRG2BGR_VNG =64,
-    COLOR_BayerGR2BGR_VNG =65,
-
-    COLOR_BayerBG2RGB_VNG =COLOR_BayerRG2BGR_VNG,
-    COLOR_BayerGB2RGB_VNG =COLOR_BayerGR2BGR_VNG,
-    COLOR_BayerRG2RGB_VNG =COLOR_BayerBG2BGR_VNG,
-    COLOR_BayerGR2RGB_VNG =COLOR_BayerGB2BGR_VNG,
-
-    COLOR_BGR2HSV_FULL = 66,
-    COLOR_RGB2HSV_FULL = 67,
-    COLOR_BGR2HLS_FULL = 68,
-    COLOR_RGB2HLS_FULL = 69,
-
-    COLOR_HSV2BGR_FULL = 70,
-    COLOR_HSV2RGB_FULL = 71,
-    COLOR_HLS2BGR_FULL = 72,
-    COLOR_HLS2RGB_FULL = 73,
-
-    COLOR_LBGR2Lab     = 74,
-    COLOR_LRGB2Lab     = 75,
-    COLOR_LBGR2Luv     = 76,
-    COLOR_LRGB2Luv     = 77,
-
-    COLOR_Lab2LBGR     = 78,
-    COLOR_Lab2LRGB     = 79,
-    COLOR_Luv2LBGR     = 80,
-    COLOR_Luv2LRGB     = 81,
-
-    COLOR_BGR2YUV      = 82,
-    COLOR_RGB2YUV      = 83,
-    COLOR_YUV2BGR      = 84,
-    COLOR_YUV2RGB      = 85,
-
-    COLOR_BayerBG2GRAY = 86,
-    COLOR_BayerGB2GRAY = 87,
-    COLOR_BayerRG2GRAY = 88,
-    COLOR_BayerGR2GRAY = 89,
-
-    //YUV 4:2:0 formats family
-    COLOR_YUV2RGB_NV12 = 90,
-    COLOR_YUV2BGR_NV12 = 91,
-    COLOR_YUV2RGB_NV21 = 92,
-    COLOR_YUV2BGR_NV21 = 93,
-    COLOR_YUV420sp2RGB = COLOR_YUV2RGB_NV21,
-    COLOR_YUV420sp2BGR = COLOR_YUV2BGR_NV21,
-
-    COLOR_YUV2RGBA_NV12 = 94,
-    COLOR_YUV2BGRA_NV12 = 95,
-    COLOR_YUV2RGBA_NV21 = 96,
-    COLOR_YUV2BGRA_NV21 = 97,
-    COLOR_YUV420sp2RGBA = COLOR_YUV2RGBA_NV21,
-    COLOR_YUV420sp2BGRA = COLOR_YUV2BGRA_NV21,
-
-    COLOR_YUV2RGB_YV12 = 98,
-    COLOR_YUV2BGR_YV12 = 99,
-    COLOR_YUV2RGB_IYUV = 100,
-    COLOR_YUV2BGR_IYUV = 101,
-    COLOR_YUV2RGB_I420 = COLOR_YUV2RGB_IYUV,
-    COLOR_YUV2BGR_I420 = COLOR_YUV2BGR_IYUV,
-    COLOR_YUV420p2RGB = COLOR_YUV2RGB_YV12,
-    COLOR_YUV420p2BGR = COLOR_YUV2BGR_YV12,
-
-    COLOR_YUV2RGBA_YV12 = 102,
-    COLOR_YUV2BGRA_YV12 = 103,
-    COLOR_YUV2RGBA_IYUV = 104,
-    COLOR_YUV2BGRA_IYUV = 105,
-    COLOR_YUV2RGBA_I420 = COLOR_YUV2RGBA_IYUV,
-    COLOR_YUV2BGRA_I420 = COLOR_YUV2BGRA_IYUV,
-    COLOR_YUV420p2RGBA = COLOR_YUV2RGBA_YV12,
-    COLOR_YUV420p2BGRA = COLOR_YUV2BGRA_YV12,
-
-    COLOR_YUV2GRAY_420 = 106,
-    COLOR_YUV2GRAY_NV21 = COLOR_YUV2GRAY_420,
-    COLOR_YUV2GRAY_NV12 = COLOR_YUV2GRAY_420,
-    COLOR_YUV2GRAY_YV12 = COLOR_YUV2GRAY_420,
-    COLOR_YUV2GRAY_IYUV = COLOR_YUV2GRAY_420,
-    COLOR_YUV2GRAY_I420 = COLOR_YUV2GRAY_420,
-    COLOR_YUV420sp2GRAY = COLOR_YUV2GRAY_420,
-    COLOR_YUV420p2GRAY = COLOR_YUV2GRAY_420,
-
-    //YUV 4:2:2 formats family
-    COLOR_YUV2RGB_UYVY = 107,
-    COLOR_YUV2BGR_UYVY = 108,
-    //COLOR_YUV2RGB_VYUY = 109,
-    //COLOR_YUV2BGR_VYUY = 110,
-    COLOR_YUV2RGB_Y422 = COLOR_YUV2RGB_UYVY,
-    COLOR_YUV2BGR_Y422 = COLOR_YUV2BGR_UYVY,
-    COLOR_YUV2RGB_UYNV = COLOR_YUV2RGB_UYVY,
-    COLOR_YUV2BGR_UYNV = COLOR_YUV2BGR_UYVY,
-
-    COLOR_YUV2RGBA_UYVY = 111,
-    COLOR_YUV2BGRA_UYVY = 112,
-    //COLOR_YUV2RGBA_VYUY = 113,
-    //COLOR_YUV2BGRA_VYUY = 114,
-    COLOR_YUV2RGBA_Y422 = COLOR_YUV2RGBA_UYVY,
-    COLOR_YUV2BGRA_Y422 = COLOR_YUV2BGRA_UYVY,
-    COLOR_YUV2RGBA_UYNV = COLOR_YUV2RGBA_UYVY,
-    COLOR_YUV2BGRA_UYNV = COLOR_YUV2BGRA_UYVY,
-
-    COLOR_YUV2RGB_YUY2 = 115,
-    COLOR_YUV2BGR_YUY2 = 116,
-    COLOR_YUV2RGB_YVYU = 117,
-    COLOR_YUV2BGR_YVYU = 118,
-    COLOR_YUV2RGB_YUYV = COLOR_YUV2RGB_YUY2,
-    COLOR_YUV2BGR_YUYV = COLOR_YUV2BGR_YUY2,
-    COLOR_YUV2RGB_YUNV = COLOR_YUV2RGB_YUY2,
-    COLOR_YUV2BGR_YUNV = COLOR_YUV2BGR_YUY2,
-
-    COLOR_YUV2RGBA_YUY2 = 119,
-    COLOR_YUV2BGRA_YUY2 = 120,
-    COLOR_YUV2RGBA_YVYU = 121,
-    COLOR_YUV2BGRA_YVYU = 122,
-    COLOR_YUV2RGBA_YUYV = COLOR_YUV2RGBA_YUY2,
-    COLOR_YUV2BGRA_YUYV = COLOR_YUV2BGRA_YUY2,
-    COLOR_YUV2RGBA_YUNV = COLOR_YUV2RGBA_YUY2,
-    COLOR_YUV2BGRA_YUNV = COLOR_YUV2BGRA_YUY2,
-
-    COLOR_YUV2GRAY_UYVY = 123,
-    COLOR_YUV2GRAY_YUY2 = 124,
-    //COLOR_YUV2GRAY_VYUY = COLOR_YUV2GRAY_UYVY,
-    COLOR_YUV2GRAY_Y422 = COLOR_YUV2GRAY_UYVY,
-    COLOR_YUV2GRAY_UYNV = COLOR_YUV2GRAY_UYVY,
-    COLOR_YUV2GRAY_YVYU = COLOR_YUV2GRAY_YUY2,
-    COLOR_YUV2GRAY_YUYV = COLOR_YUV2GRAY_YUY2,
-    COLOR_YUV2GRAY_YUNV = COLOR_YUV2GRAY_YUY2,
-
-    // alpha premultiplication
-    COLOR_RGBA2mRGBA = 125,
-    COLOR_mRGBA2RGBA = 126,
-
-    COLOR_RGB2YUV_I420 = 127,
-    COLOR_BGR2YUV_I420 = 128,
-    COLOR_RGB2YUV_IYUV = COLOR_RGB2YUV_I420,
-    COLOR_BGR2YUV_IYUV = COLOR_BGR2YUV_I420,
-
-    COLOR_RGBA2YUV_I420 = 129,
-    COLOR_BGRA2YUV_I420 = 130,
-    COLOR_RGBA2YUV_IYUV = COLOR_RGBA2YUV_I420,
-    COLOR_BGRA2YUV_IYUV = COLOR_BGRA2YUV_I420,
-    COLOR_RGB2YUV_YV12  = 131,
-    COLOR_BGR2YUV_YV12  = 132,
-    COLOR_RGBA2YUV_YV12 = 133,
-    COLOR_BGRA2YUV_YV12 = 134,
-
-    // Edge-Aware Demosaicing
-    COLOR_BayerBG2BGR_EA = 135,
-    COLOR_BayerGB2BGR_EA = 136,
-    COLOR_BayerRG2BGR_EA = 137,
-    COLOR_BayerGR2BGR_EA = 138,
-
-    COLOR_BayerBG2RGB_EA = COLOR_BayerRG2BGR_EA,
-    COLOR_BayerGB2RGB_EA = COLOR_BayerGR2BGR_EA,
-    COLOR_BayerRG2RGB_EA = COLOR_BayerBG2BGR_EA,
-    COLOR_BayerGR2RGB_EA = COLOR_BayerGB2BGR_EA,
-
-    COLOR_COLORCVT_MAX  = 139
-};
-
-
-//! converts image from one color space to another
-CV_EXPORTS_W void cvtColor( InputArray src, OutputArray dst, int code, int dstCn=0 );
-
-//! raster image moments
-class CV_EXPORTS_W_MAP Moments
+//! Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
+//! Detects position only without traslation and rotation
+class CV_EXPORTS GeneralizedHoughBallard : public GeneralizedHough
 {
 public:
-    //! the default constructor
-    Moments();
-    //! the full constructor
-    Moments(double m00, double m10, double m01, double m20, double m11,
-            double m02, double m30, double m21, double m12, double m03 );
-    //! the conversion from CvMoments
-    Moments( const CvMoments& moments );
-    //! the conversion to CvMoments
-    operator CvMoments() const;
+    //! R-Table levels.
+    virtual void setLevels(int levels) = 0;
+    virtual int getLevels() const = 0;
 
-    //! spatial moments
-    CV_PROP_RW double  m00, m10, m01, m20, m11, m02, m30, m21, m12, m03;
-    //! central moments
-    CV_PROP_RW double  mu20, mu11, mu02, mu30, mu21, mu12, mu03;
-    //! central normalized moments
-    CV_PROP_RW double  nu20, nu11, nu02, nu30, nu21, nu12, nu03;
+    //! The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.
+    virtual void setVotesThreshold(int votesThreshold) = 0;
+    virtual int getVotesThreshold() const = 0;
 };
 
-//! computes moments of the rasterized shape or a vector of points
-CV_EXPORTS_W Moments moments( InputArray array, bool binaryImage=false );
-
-//! computes 7 Hu invariants from the moments
-CV_EXPORTS void HuMoments( const Moments& moments, double hu[7] );
-CV_EXPORTS_W void HuMoments( const Moments& m, OutputArray hu );
-
-//! type of the template matching operation
-enum { TM_SQDIFF=0, TM_SQDIFF_NORMED=1, TM_CCORR=2, TM_CCORR_NORMED=3, TM_CCOEFF=4, TM_CCOEFF_NORMED=5 };
-
-//! computes the proximity map for the raster template and the image where the template is searched for
-CV_EXPORTS_W void matchTemplate( InputArray image, InputArray templ,
-                                 OutputArray result, int method );
-
-enum { CC_STAT_LEFT=0, CC_STAT_TOP=1, CC_STAT_WIDTH=2, CC_STAT_HEIGHT=3, CC_STAT_AREA=4, CC_STAT_MAX = 5};
-
-// computes the connected components labeled image of boolean image ``image``
-// with 4 or 8 way connectivity - returns N, the total
-// number of labels [0, N-1] where 0 represents the background label.
-// ltype specifies the output label image type, an important
-// consideration based on the total number of labels or
-// alternatively the total number of pixels in the source image.
-CV_EXPORTS_W int connectedComponents(InputArray image, OutputArray labels,
-                                     int connectivity = 8, int ltype=CV_32S);
-CV_EXPORTS_W int connectedComponentsWithStats(InputArray image, OutputArray labels,
-                                              OutputArray stats, OutputArray centroids,
-                                              int connectivity = 8, int ltype=CV_32S);
-
-//! mode of the contour retrieval algorithm
-enum
+//! Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
+//! Detects position, traslation and rotation
+class CV_EXPORTS GeneralizedHoughGuil : public GeneralizedHough
 {
-    RETR_EXTERNAL=CV_RETR_EXTERNAL, //!< retrieve only the most external (top-level) contours
-    RETR_LIST=CV_RETR_LIST, //!< retrieve all the contours without any hierarchical information
-    RETR_CCOMP=CV_RETR_CCOMP, //!< retrieve the connected components (that can possibly be nested)
-    RETR_TREE=CV_RETR_TREE, //!< retrieve all the contours and the whole hierarchy
-    RETR_FLOODFILL=CV_RETR_FLOODFILL
+public:
+    //! Angle difference in degrees between two points in feature.
+    virtual void setXi(double xi) = 0;
+    virtual double getXi() const = 0;
+
+    //! Feature table levels.
+    virtual void setLevels(int levels) = 0;
+    virtual int getLevels() const = 0;
+
+    //! Maximal difference between angles that treated as equal.
+    virtual void setAngleEpsilon(double angleEpsilon) = 0;
+    virtual double getAngleEpsilon() const = 0;
+
+    //! Minimal rotation angle to detect in degrees.
+    virtual void setMinAngle(double minAngle) = 0;
+    virtual double getMinAngle() const = 0;
+
+    //! Maximal rotation angle to detect in degrees.
+    virtual void setMaxAngle(double maxAngle) = 0;
+    virtual double getMaxAngle() const = 0;
+
+    //! Angle step in degrees.
+    virtual void setAngleStep(double angleStep) = 0;
+    virtual double getAngleStep() const = 0;
+
+    //! Angle votes threshold.
+    virtual void setAngleThresh(int angleThresh) = 0;
+    virtual int getAngleThresh() const = 0;
+
+    //! Minimal scale to detect.
+    virtual void setMinScale(double minScale) = 0;
+    virtual double getMinScale() const = 0;
+
+    //! Maximal scale to detect.
+    virtual void setMaxScale(double maxScale) = 0;
+    virtual double getMaxScale() const = 0;
+
+    //! Scale step.
+    virtual void setScaleStep(double scaleStep) = 0;
+    virtual double getScaleStep() const = 0;
+
+    //! Scale votes threshold.
+    virtual void setScaleThresh(int scaleThresh) = 0;
+    virtual int getScaleThresh() const = 0;
+
+    //! Position votes threshold.
+    virtual void setPosThresh(int posThresh) = 0;
+    virtual int getPosThresh() const = 0;
 };
 
-//! the contour approximation algorithm
-enum
+
+class CV_EXPORTS_W CLAHE : public Algorithm
 {
-    CHAIN_APPROX_NONE=CV_CHAIN_APPROX_NONE,
-    CHAIN_APPROX_SIMPLE=CV_CHAIN_APPROX_SIMPLE,
-    CHAIN_APPROX_TC89_L1=CV_CHAIN_APPROX_TC89_L1,
-    CHAIN_APPROX_TC89_KCOS=CV_CHAIN_APPROX_TC89_KCOS
+public:
+    CV_WRAP virtual void apply(InputArray src, OutputArray dst) = 0;
+
+    CV_WRAP virtual void setClipLimit(double clipLimit) = 0;
+    CV_WRAP virtual double getClipLimit() const = 0;
+
+    CV_WRAP virtual void setTilesGridSize(Size tileGridSize) = 0;
+    CV_WRAP virtual Size getTilesGridSize() const = 0;
+
+    CV_WRAP virtual void collectGarbage() = 0;
 };
 
-//! retrieves contours and the hierarchical information from black-n-white image.
-CV_EXPORTS_W void findContours( InputOutputArray image, OutputArrayOfArrays contours,
-                              OutputArray hierarchy, int mode,
-                              int method, Point offset=Point());
-
-//! retrieves contours from black-n-white image.
-CV_EXPORTS void findContours( InputOutputArray image, OutputArrayOfArrays contours,
-                              int mode, int method, Point offset=Point());
-
-//! approximates contour or a curve using Douglas-Peucker algorithm
-CV_EXPORTS_W void approxPolyDP( InputArray curve,
-                                OutputArray approxCurve,
-                                double epsilon, bool closed );
-
-//! computes the contour perimeter (closed=true) or a curve length
-CV_EXPORTS_W double arcLength( InputArray curve, bool closed );
-//! computes the bounding rectangle for a contour
-CV_EXPORTS_W Rect boundingRect( InputArray points );
-//! computes the contour area
-CV_EXPORTS_W double contourArea( InputArray contour, bool oriented=false );
-//! computes the minimal rotated rectangle for a set of points
-CV_EXPORTS_W RotatedRect minAreaRect( InputArray points );
-//! computes the minimal enclosing circle for a set of points
-CV_EXPORTS_W void minEnclosingCircle( InputArray points,
-                                      CV_OUT Point2f& center, CV_OUT float& radius );
-//! matches two contours using one of the available algorithms
-CV_EXPORTS_W double matchShapes( InputArray contour1, InputArray contour2,
-                                 int method, double parameter );
-//! computes convex hull for a set of 2D points.
-CV_EXPORTS_W void convexHull( InputArray points, OutputArray hull,
-                              bool clockwise=false, bool returnPoints=true );
-//! computes the contour convexity defects
-CV_EXPORTS_W void convexityDefects( InputArray contour, InputArray convexhull, OutputArray convexityDefects );
-
-//! returns true if the contour is convex. Does not support contours with self-intersection
-CV_EXPORTS_W bool isContourConvex( InputArray contour );
-
-//! finds intersection of two convex polygons
-CV_EXPORTS_W float intersectConvexConvex( InputArray _p1, InputArray _p2,
-                                          OutputArray _p12, bool handleNested=true );
-
-//! fits ellipse to the set of 2D points
-CV_EXPORTS_W RotatedRect fitEllipse( InputArray points );
-
-//! fits line to the set of 2D points using M-estimator algorithm
-CV_EXPORTS_W void fitLine( InputArray points, OutputArray line, int distType,
-                           double param, double reps, double aeps );
-//! checks if the point is inside the contour. Optionally computes the signed distance from the point to the contour boundary
-CV_EXPORTS_W double pointPolygonTest( InputArray contour, Point2f pt, bool measureDist );
-
 
 class CV_EXPORTS_W Subdiv2D
 {
 public:
-    enum
-    {
-        PTLOC_ERROR = -2,
-        PTLOC_OUTSIDE_RECT = -1,
-        PTLOC_INSIDE = 0,
-        PTLOC_VERTEX = 1,
-        PTLOC_ON_EDGE = 2
-    };
+    enum { PTLOC_ERROR        = -2,
+           PTLOC_OUTSIDE_RECT = -1,
+           PTLOC_INSIDE       = 0,
+           PTLOC_VERTEX       = 1,
+           PTLOC_ON_EDGE      = 2
+         };
 
-    enum
-    {
-        NEXT_AROUND_ORG   = 0x00,
-        NEXT_AROUND_DST   = 0x22,
-        PREV_AROUND_ORG   = 0x11,
-        PREV_AROUND_DST   = 0x33,
-        NEXT_AROUND_LEFT  = 0x13,
-        NEXT_AROUND_RIGHT = 0x31,
-        PREV_AROUND_LEFT  = 0x20,
-        PREV_AROUND_RIGHT = 0x02
-    };
+    enum { NEXT_AROUND_ORG   = 0x00,
+           NEXT_AROUND_DST   = 0x22,
+           PREV_AROUND_ORG   = 0x11,
+           PREV_AROUND_DST   = 0x33,
+           NEXT_AROUND_LEFT  = 0x13,
+           NEXT_AROUND_RIGHT = 0x31,
+           PREV_AROUND_LEFT  = 0x20,
+           PREV_AROUND_RIGHT = 0x02
+         };
 
     CV_WRAP Subdiv2D();
     CV_WRAP Subdiv2D(Rect rect);
@@ -1228,25 +838,25 @@ public:
     CV_WRAP void insert(const std::vector<Point2f>& ptvec);
     CV_WRAP int locate(Point2f pt, CV_OUT int& edge, CV_OUT int& vertex);
 
-    CV_WRAP int findNearest(Point2f pt, CV_OUT Point2f* nearestPt=0);
+    CV_WRAP int findNearest(Point2f pt, CV_OUT Point2f* nearestPt = 0);
     CV_WRAP void getEdgeList(CV_OUT std::vector<Vec4f>& edgeList) const;
     CV_WRAP void getTriangleList(CV_OUT std::vector<Vec6f>& triangleList) const;
     CV_WRAP void getVoronoiFacetList(const std::vector<int>& idx, CV_OUT std::vector<std::vector<Point2f> >& facetList,
                                      CV_OUT std::vector<Point2f>& facetCenters);
 
-    CV_WRAP Point2f getVertex(int vertex, CV_OUT int* firstEdge=0) const;
+    CV_WRAP Point2f getVertex(int vertex, CV_OUT int* firstEdge = 0) const;
 
     CV_WRAP int getEdge( int edge, int nextEdgeType ) const;
     CV_WRAP int nextEdge(int edge) const;
     CV_WRAP int rotateEdge(int edge, int rotate) const;
     CV_WRAP int symEdge(int edge) const;
-    CV_WRAP int edgeOrg(int edge, CV_OUT Point2f* orgpt=0) const;
-    CV_WRAP int edgeDst(int edge, CV_OUT Point2f* dstpt=0) const;
+    CV_WRAP int edgeOrg(int edge, CV_OUT Point2f* orgpt = 0) const;
+    CV_WRAP int edgeDst(int edge, CV_OUT Point2f* dstpt = 0) const;
 
 protected:
     int newEdge();
     void deleteEdge(int edge);
-    int newPoint(Point2f pt, bool isvirtual, int firstEdge=0);
+    int newPoint(Point2f pt, bool isvirtual, int firstEdge = 0);
     void deletePoint(int vtx);
     void setEdgePoints( int edge, int orgPt, int dstPt );
     void splice( int edgeA, int edgeB );
@@ -1263,15 +873,18 @@ protected:
         Vertex(Point2f pt, bool _isvirtual, int _firstEdge=0);
         bool isvirtual() const;
         bool isfree() const;
+
         int firstEdge;
         int type;
         Point2f pt;
     };
+
     struct CV_EXPORTS QuadEdge
     {
         QuadEdge();
         QuadEdge(int edgeidx);
         bool isfree() const;
+
         int next[4];
         int pt[4];
     };
@@ -1287,13 +900,599 @@ protected:
     Point2f bottomRight;
 };
 
+class LineSegmentDetector : public Algorithm
+{
+public:
+/**
+ * Detect lines in the input image.
+ *
+ * @param _image    A grayscale(CV_8UC1) input image.
+ *                  If only a roi needs to be selected, use
+ *                  lsd_ptr->detect(image(roi), ..., lines);
+ *                  lines += Scalar(roi.x, roi.y, roi.x, roi.y);
+ * @param _lines    Return: A vector of Vec4i elements specifying the beginning and ending point of a line.
+ *                          Where Vec4i is (x1, y1, x2, y2), point 1 is the start, point 2 - end.
+ *                          Returned lines are strictly oriented depending on the gradient.
+ * @param width     Return: Vector of widths of the regions, where the lines are found. E.g. Width of line.
+ * @param prec      Return: Vector of precisions with which the lines are found.
+ * @param nfa       Return: Vector containing number of false alarms in the line region, with precision of 10%.
+ *                          The bigger the value, logarithmically better the detection.
+ *                              * -1 corresponds to 10 mean false alarms
+ *                              * 0 corresponds to 1 mean false alarm
+ *                              * 1 corresponds to 0.1 mean false alarms
+ *                          This vector will be calculated _only_ when the objects type is REFINE_ADV
+ */
+    virtual void detect(InputArray _image, OutputArray _lines,
+                        OutputArray width = noArray(), OutputArray prec = noArray(),
+                        OutputArray nfa = noArray()) = 0;
+
+/**
+ * Draw lines on the given canvas.
+ *
+ * @param image     The image, where lines will be drawn.
+ *                  Should have the size of the image, where the lines were found
+ * @param lines     The lines that need to be drawn
+ */
+    virtual void drawSegments(InputOutputArray _image, InputArray lines) = 0;
+
+/**
+ * Draw both vectors on the image canvas. Uses blue for lines 1 and red for lines 2.
+ *
+ * @param size      The size of the image, where lines were found.
+ * @param lines1    The first lines that need to be drawn. Color - Blue.
+ * @param lines2    The second lines that need to be drawn. Color - Red.
+ * @param image     Optional image, where lines will be drawn.
+ *                  Should have the size of the image, where the lines were found
+ * @return          The number of mismatching pixels between lines1 and lines2.
+ */
+    virtual int compareSegments(const Size& size, InputArray lines1, InputArray lines2, InputOutputArray _image = noArray()) = 0;
+
+    virtual ~LineSegmentDetector() {};
+};
+
+//! Returns a pointer to a LineSegmentDetector class.
+CV_EXPORTS Ptr<LineSegmentDetector> createLineSegmentDetectorPtr(
+    int _refine = LSD_REFINE_STD, double _scale = 0.8,
+    double _sigma_scale = 0.6, double _quant = 2.0, double _ang_th = 22.5,
+    double _log_eps = 0, double _density_th = 0.7, int _n_bins = 1024);
+
+//! returns type (one of KERNEL_*) of 1D or 2D kernel specified by its coefficients.
+CV_EXPORTS int getKernelType(InputArray kernel, Point anchor);
+
+//! returns the primitive row filter with the specified kernel
+CV_EXPORTS Ptr<BaseRowFilter> getLinearRowFilter(int srcType, int bufType,
+                                            InputArray kernel, int anchor,
+                                            int symmetryType);
+
+//! returns the primitive column filter with the specified kernel
+CV_EXPORTS Ptr<BaseColumnFilter> getLinearColumnFilter(int bufType, int dstType,
+                                            InputArray kernel, int anchor,
+                                            int symmetryType, double delta = 0,
+                                            int bits = 0);
+
+//! returns 2D filter with the specified kernel
+CV_EXPORTS Ptr<BaseFilter> getLinearFilter(int srcType, int dstType,
+                                           InputArray kernel,
+                                           Point anchor = Point(-1,-1),
+                                           double delta = 0, int bits = 0);
+
+//! returns the separable linear filter engine
+CV_EXPORTS Ptr<FilterEngine> createSeparableLinearFilter(int srcType, int dstType,
+                          InputArray rowKernel, InputArray columnKernel,
+                          Point anchor = Point(-1,-1), double delta = 0,
+                          int rowBorderType = BORDER_DEFAULT,
+                          int columnBorderType = -1,
+                          const Scalar& borderValue = Scalar());
+
+//! returns the non-separable linear filter engine
+CV_EXPORTS Ptr<FilterEngine> createLinearFilter(int srcType, int dstType,
+                 InputArray kernel, Point _anchor = Point(-1,-1),
+                 double delta = 0, int rowBorderType = BORDER_DEFAULT,
+                 int columnBorderType = -1, const Scalar& borderValue = Scalar());
+
+//! returns the Gaussian kernel with the specified parameters
+CV_EXPORTS_W Mat getGaussianKernel( int ksize, double sigma, int ktype = CV_64F );
+
+//! returns the Gaussian filter engine
+CV_EXPORTS Ptr<FilterEngine> createGaussianFilter( int type, Size ksize,
+                                    double sigma1, double sigma2 = 0,
+                                    int borderType = BORDER_DEFAULT);
+
+//! initializes kernels of the generalized Sobel operator
+CV_EXPORTS_W void getDerivKernels( OutputArray kx, OutputArray ky,
+                                   int dx, int dy, int ksize,
+                                   bool normalize = false, int ktype = CV_32F );
+
+//! returns filter engine for the generalized Sobel operator
+CV_EXPORTS Ptr<FilterEngine> createDerivFilter( int srcType, int dstType,
+                                        int dx, int dy, int ksize,
+                                        int borderType = BORDER_DEFAULT );
+
+//! returns horizontal 1D box filter
+CV_EXPORTS Ptr<BaseRowFilter> getRowSumFilter(int srcType, int sumType,
+                                              int ksize, int anchor = -1);
+
+//! returns vertical 1D box filter
+CV_EXPORTS Ptr<BaseColumnFilter> getColumnSumFilter( int sumType, int dstType,
+                                                     int ksize, int anchor = -1,
+                                                     double scale = 1);
+//! returns box filter engine
+CV_EXPORTS Ptr<FilterEngine> createBoxFilter( int srcType, int dstType, Size ksize,
+                                              Point anchor = Point(-1,-1),
+                                              bool normalize = true,
+                                              int borderType = BORDER_DEFAULT);
+
+//! returns the Gabor kernel with the specified parameters
+CV_EXPORTS_W Mat getGaborKernel( Size ksize, double sigma, double theta, double lambd,
+                                 double gamma, double psi = CV_PI*0.5, int ktype = CV_64F );
+
+//! returns horizontal 1D morphological filter
+CV_EXPORTS Ptr<BaseRowFilter> getMorphologyRowFilter(int op, int type, int ksize, int anchor = -1);
+
+//! returns vertical 1D morphological filter
+CV_EXPORTS Ptr<BaseColumnFilter> getMorphologyColumnFilter(int op, int type, int ksize, int anchor = -1);
+
+//! returns 2D morphological filter
+CV_EXPORTS Ptr<BaseFilter> getMorphologyFilter(int op, int type, InputArray kernel,
+                                               Point anchor = Point(-1,-1));
+
+//! returns "magic" border value for erosion and dilation. It is automatically transformed to Scalar::all(-DBL_MAX) for dilation.
+static inline Scalar morphologyDefaultBorderValue() { return Scalar::all(DBL_MAX); }
+
+//! returns morphological filter engine. Only MORPH_ERODE and MORPH_DILATE are supported.
+CV_EXPORTS Ptr<FilterEngine> createMorphologyFilter(int op, int type, InputArray kernel,
+                    Point anchor = Point(-1,-1), int rowBorderType = BORDER_CONSTANT,
+                    int columnBorderType = -1, const Scalar& borderValue = morphologyDefaultBorderValue());
+
+//! returns structuring element of the specified shape and size
+CV_EXPORTS_W Mat getStructuringElement(int shape, Size ksize, Point anchor = Point(-1,-1));
+
+//! smooths the image using median filter.
+CV_EXPORTS_W void medianBlur( InputArray src, OutputArray dst, int ksize );
+
+//! smooths the image using Gaussian filter.
+CV_EXPORTS_W void GaussianBlur( InputArray src, OutputArray dst, Size ksize,
+                                double sigmaX, double sigmaY = 0,
+                                int borderType = BORDER_DEFAULT );
+
+//! smooths the image using bilateral filter
+CV_EXPORTS_W void bilateralFilter( InputArray src, OutputArray dst, int d,
+                                   double sigmaColor, double sigmaSpace,
+                                   int borderType = BORDER_DEFAULT );
+
+//! smooths the image using adaptive bilateral filter
+CV_EXPORTS_W void adaptiveBilateralFilter( InputArray src, OutputArray dst, Size ksize,
+                                           double sigmaSpace, Point anchor=Point(-1, -1),
+                                           int borderType=BORDER_DEFAULT );
+
+//! smooths the image using the box filter. Each pixel is processed in O(1) time
+CV_EXPORTS_W void boxFilter( InputArray src, OutputArray dst, int ddepth,
+                             Size ksize, Point anchor = Point(-1,-1),
+                             bool normalize = true,
+                             int borderType = BORDER_DEFAULT );
+
+//! a synonym for normalized box filter
+CV_EXPORTS_W void blur( InputArray src, OutputArray dst,
+                        Size ksize, Point anchor = Point(-1,-1),
+                        int borderType = BORDER_DEFAULT );
+
+//! applies non-separable 2D linear filter to the image
+CV_EXPORTS_W void filter2D( InputArray src, OutputArray dst, int ddepth,
+                            InputArray kernel, Point anchor = Point(-1,-1),
+                            double delta = 0, int borderType = BORDER_DEFAULT );
+
+//! applies separable 2D linear filter to the image
+CV_EXPORTS_W void sepFilter2D( InputArray src, OutputArray dst, int ddepth,
+                               InputArray kernelX, InputArray kernelY,
+                               Point anchor = Point(-1,-1),
+                               double delta = 0, int borderType = BORDER_DEFAULT );
+
+//! applies generalized Sobel operator to the image
+CV_EXPORTS_W void Sobel( InputArray src, OutputArray dst, int ddepth,
+                         int dx, int dy, int ksize = 3,
+                         double scale = 1, double delta = 0,
+                         int borderType = BORDER_DEFAULT );
+
+//! applies the vertical or horizontal Scharr operator to the image
+CV_EXPORTS_W void Scharr( InputArray src, OutputArray dst, int ddepth,
+                          int dx, int dy, double scale = 1, double delta = 0,
+                          int borderType = BORDER_DEFAULT );
+
+//! applies Laplacian operator to the image
+CV_EXPORTS_W void Laplacian( InputArray src, OutputArray dst, int ddepth,
+                             int ksize = 1, double scale = 1, double delta = 0,
+                             int borderType = BORDER_DEFAULT );
+
+//! applies Canny edge detector and produces the edge map.
+CV_EXPORTS_W void Canny( InputArray image, OutputArray edges,
+                         double threshold1, double threshold2,
+                         int apertureSize = 3, bool L2gradient = false );
+
+//! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria
+CV_EXPORTS_W void cornerMinEigenVal( InputArray src, OutputArray dst,
+                                     int blockSize, int ksize = 3,
+                                     int borderType = BORDER_DEFAULT );
+
+//! computes Harris cornerness criteria at each image pixel
+CV_EXPORTS_W void cornerHarris( InputArray src, OutputArray dst, int blockSize,
+                                int ksize, double k,
+                                int borderType = BORDER_DEFAULT );
+
+//! computes both eigenvalues and the eigenvectors of 2x2 derivative covariation matrix  at each pixel. The output is stored as 6-channel matrix.
+CV_EXPORTS_W void cornerEigenValsAndVecs( InputArray src, OutputArray dst,
+                                          int blockSize, int ksize,
+                                          int borderType = BORDER_DEFAULT );
+
+//! computes another complex cornerness criteria at each pixel
+CV_EXPORTS_W void preCornerDetect( InputArray src, OutputArray dst, int ksize,
+                                   int borderType = BORDER_DEFAULT );
+
+//! adjusts the corner locations with sub-pixel accuracy to maximize the certain cornerness criteria
+CV_EXPORTS_W void cornerSubPix( InputArray image, InputOutputArray corners,
+                                Size winSize, Size zeroZone,
+                                TermCriteria criteria );
+
+//! finds the strong enough corners where the cornerMinEigenVal() or cornerHarris() report the local maxima
+CV_EXPORTS_W void goodFeaturesToTrack( InputArray image, OutputArray corners,
+                                     int maxCorners, double qualityLevel, double minDistance,
+                                     InputArray mask = noArray(), int blockSize = 3,
+                                     bool useHarrisDetector = false, double k = 0.04 );
+
+//! finds lines in the black-n-white image using the standard or pyramid Hough transform
+CV_EXPORTS_W void HoughLines( InputArray image, OutputArray lines,
+                              double rho, double theta, int threshold,
+                              double srn = 0, double stn = 0 );
+
+//! finds line segments in the black-n-white image using probabalistic Hough transform
+CV_EXPORTS_W void HoughLinesP( InputArray image, OutputArray lines,
+                               double rho, double theta, int threshold,
+                               double minLineLength = 0, double maxLineGap = 0 );
+
+//! finds circles in the grayscale image using 2+1 gradient Hough transform
+CV_EXPORTS_W void HoughCircles( InputArray image, OutputArray circles,
+                               int method, double dp, double minDist,
+                               double param1 = 100, double param2 = 100,
+                               int minRadius = 0, int maxRadius = 0 );
+
+//! erodes the image (applies the local minimum operator)
+CV_EXPORTS_W void erode( InputArray src, OutputArray dst, InputArray kernel,
+                         Point anchor = Point(-1,-1), int iterations = 1,
+                         int borderType = BORDER_CONSTANT,
+                         const Scalar& borderValue = morphologyDefaultBorderValue() );
+
+//! dilates the image (applies the local maximum operator)
+CV_EXPORTS_W void dilate( InputArray src, OutputArray dst, InputArray kernel,
+                          Point anchor = Point(-1,-1), int iterations = 1,
+                          int borderType = BORDER_CONSTANT,
+                          const Scalar& borderValue = morphologyDefaultBorderValue() );
+
+//! applies an advanced morphological operation to the image
+CV_EXPORTS_W void morphologyEx( InputArray src, OutputArray dst,
+                                int op, InputArray kernel,
+                                Point anchor = Point(-1,-1), int iterations = 1,
+                                int borderType = BORDER_CONSTANT,
+                                const Scalar& borderValue = morphologyDefaultBorderValue() );
+
+//! resizes the image
+CV_EXPORTS_W void resize( InputArray src, OutputArray dst,
+                          Size dsize, double fx = 0, double fy = 0,
+                          int interpolation = INTER_LINEAR );
+
+//! warps the image using affine transformation
+CV_EXPORTS_W void warpAffine( InputArray src, OutputArray dst,
+                              InputArray M, Size dsize,
+                              int flags = INTER_LINEAR,
+                              int borderMode = BORDER_CONSTANT,
+                              const Scalar& borderValue = Scalar());
+
+//! warps the image using perspective transformation
+CV_EXPORTS_W void warpPerspective( InputArray src, OutputArray dst,
+                                   InputArray M, Size dsize,
+                                   int flags = INTER_LINEAR,
+                                   int borderMode = BORDER_CONSTANT,
+                                   const Scalar& borderValue = Scalar());
+
+//! warps the image using the precomputed maps. The maps are stored in either floating-point or integer fixed-point format
+CV_EXPORTS_W void remap( InputArray src, OutputArray dst,
+                         InputArray map1, InputArray map2,
+                         int interpolation, int borderMode = BORDER_CONSTANT,
+                         const Scalar& borderValue = Scalar());
+
+//! converts maps for remap from floating-point to fixed-point format or backwards
+CV_EXPORTS_W void convertMaps( InputArray map1, InputArray map2,
+                               OutputArray dstmap1, OutputArray dstmap2,
+                               int dstmap1type, bool nninterpolation = false );
+
+//! returns 2x3 affine transformation matrix for the planar rotation.
+CV_EXPORTS_W Mat getRotationMatrix2D( Point2f center, double angle, double scale );
+
+//! returns 3x3 perspective transformation for the corresponding 4 point pairs.
+CV_EXPORTS Mat getPerspectiveTransform( const Point2f src[], const Point2f dst[] );
+
+//! returns 2x3 affine transformation for the corresponding 3 point pairs.
+CV_EXPORTS Mat getAffineTransform( const Point2f src[], const Point2f dst[] );
+
+//! computes 2x3 affine transformation matrix that is inverse to the specified 2x3 affine transformation.
+CV_EXPORTS_W void invertAffineTransform( InputArray M, OutputArray iM );
+
+CV_EXPORTS_W Mat getPerspectiveTransform( InputArray src, InputArray dst );
+
+CV_EXPORTS_W Mat getAffineTransform( InputArray src, InputArray dst );
+
+//! extracts rectangle from the image at sub-pixel location
+CV_EXPORTS_W void getRectSubPix( InputArray image, Size patchSize,
+                                 Point2f center, OutputArray patch, int patchType = -1 );
+
+//! computes the integral image
+CV_EXPORTS_W void integral( InputArray src, OutputArray sum, int sdepth = -1 );
+
+//! computes the integral image and integral for the squared image
+CV_EXPORTS_AS(integral2) void integral( InputArray src, OutputArray sum,
+                                        OutputArray sqsum, int sdepth = -1 );
+
+//! computes the integral image, integral for the squared image and the tilted integral image
+CV_EXPORTS_AS(integral3) void integral( InputArray src, OutputArray sum,
+                                        OutputArray sqsum, OutputArray tilted,
+                                        int sdepth = -1 );
+
+//! adds image to the accumulator (dst += src). Unlike cv::add, dst and src can have different types.
+CV_EXPORTS_W void accumulate( InputArray src, InputOutputArray dst,
+                              InputArray mask = noArray() );
+
+//! adds squared src image to the accumulator (dst += src*src).
+CV_EXPORTS_W void accumulateSquare( InputArray src, InputOutputArray dst,
+                                    InputArray mask = noArray() );
+//! adds product of the 2 images to the accumulator (dst += src1*src2).
+CV_EXPORTS_W void accumulateProduct( InputArray src1, InputArray src2,
+                                     InputOutputArray dst, InputArray mask=noArray() );
+
+//! updates the running average (dst = dst*(1-alpha) + src*alpha)
+CV_EXPORTS_W void accumulateWeighted( InputArray src, InputOutputArray dst,
+                                      double alpha, InputArray mask = noArray() );
+
+CV_EXPORTS_W Point2d phaseCorrelate(InputArray src1, InputArray src2,
+                                    InputArray window = noArray(), CV_OUT double* response = 0);
+
+CV_EXPORTS_W void createHanningWindow(OutputArray dst, Size winSize, int type);
+
+//! applies fixed threshold to the image
+CV_EXPORTS_W double threshold( InputArray src, OutputArray dst,
+                               double thresh, double maxval, int type );
+
+
+//! applies variable (adaptive) threshold to the image
+CV_EXPORTS_W void adaptiveThreshold( InputArray src, OutputArray dst,
+                                     double maxValue, int adaptiveMethod,
+                                     int thresholdType, int blockSize, double C );
+
+//! smooths and downsamples the image
+CV_EXPORTS_W void pyrDown( InputArray src, OutputArray dst,
+                           const Size& dstsize = Size(), int borderType = BORDER_DEFAULT );
+
+//! upsamples and smoothes the image
+CV_EXPORTS_W void pyrUp( InputArray src, OutputArray dst,
+                         const Size& dstsize = Size(), int borderType = BORDER_DEFAULT );
+
+//! builds the gaussian pyramid using pyrDown() as a basic operation
+CV_EXPORTS void buildPyramid( InputArray src, OutputArrayOfArrays dst,
+                              int maxlevel, int borderType = BORDER_DEFAULT );
+
+//! corrects lens distortion for the given camera matrix and distortion coefficients
+CV_EXPORTS_W void undistort( InputArray src, OutputArray dst,
+                             InputArray cameraMatrix,
+                             InputArray distCoeffs,
+                             InputArray newCameraMatrix = noArray() );
+
+//! initializes maps for cv::remap() to correct lens distortion and optionally rectify the image
+CV_EXPORTS_W void initUndistortRectifyMap( InputArray cameraMatrix, InputArray distCoeffs,
+                           InputArray R, InputArray newCameraMatrix,
+                           Size size, int m1type, OutputArray map1, OutputArray map2 );
+
+//! initializes maps for cv::remap() for wide-angle
+CV_EXPORTS_W float initWideAngleProjMap( InputArray cameraMatrix, InputArray distCoeffs,
+                                         Size imageSize, int destImageWidth,
+                                         int m1type, OutputArray map1, OutputArray map2,
+                                         int projType = PROJ_SPHERICAL_EQRECT, double alpha = 0);
+
+//! returns the default new camera matrix (by default it is the same as cameraMatrix unless centerPricipalPoint=true)
+CV_EXPORTS_W Mat getDefaultNewCameraMatrix( InputArray cameraMatrix, Size imgsize = Size(),
+                                            bool centerPrincipalPoint = false );
+
+//! returns points' coordinates after lens distortion correction
+CV_EXPORTS_W void undistortPoints( InputArray src, OutputArray dst,
+                                   InputArray cameraMatrix, InputArray distCoeffs,
+                                   InputArray R = noArray(), InputArray P = noArray());
+
+//! computes the joint dense histogram for a set of images.
+CV_EXPORTS void calcHist( const Mat* images, int nimages,
+                          const int* channels, InputArray mask,
+                          OutputArray hist, int dims, const int* histSize,
+                          const float** ranges, bool uniform = true, bool accumulate = false );
+
+//! computes the joint sparse histogram for a set of images.
+CV_EXPORTS void calcHist( const Mat* images, int nimages,
+                          const int* channels, InputArray mask,
+                          SparseMat& hist, int dims,
+                          const int* histSize, const float** ranges,
+                          bool uniform = true, bool accumulate = false );
+
+CV_EXPORTS_W void calcHist( InputArrayOfArrays images,
+                            const std::vector<int>& channels,
+                            InputArray mask, OutputArray hist,
+                            const std::vector<int>& histSize,
+                            const std::vector<float>& ranges,
+                            bool accumulate = false );
+
+//! computes back projection for the set of images
+CV_EXPORTS void calcBackProject( const Mat* images, int nimages,
+                                 const int* channels, InputArray hist,
+                                 OutputArray backProject, const float** ranges,
+                                 double scale = 1, bool uniform = true );
+
+//! computes back projection for the set of images
+CV_EXPORTS void calcBackProject( const Mat* images, int nimages,
+                                 const int* channels, const SparseMat& hist,
+                                 OutputArray backProject, const float** ranges,
+                                 double scale = 1, bool uniform = true );
+
+CV_EXPORTS_W void calcBackProject( InputArrayOfArrays images, const std::vector<int>& channels,
+                                   InputArray hist, OutputArray dst,
+                                   const std::vector<float>& ranges,
+                                   double scale );
+
+//! compares two histograms stored in dense arrays
+CV_EXPORTS_W double compareHist( InputArray H1, InputArray H2, int method );
+
+//! compares two histograms stored in sparse arrays
+CV_EXPORTS double compareHist( const SparseMat& H1, const SparseMat& H2, int method );
+
+//! normalizes the grayscale image brightness and contrast by normalizing its histogram
+CV_EXPORTS_W void equalizeHist( InputArray src, OutputArray dst );
+
+CV_EXPORTS float EMD( InputArray signature1, InputArray signature2,
+                      int distType, InputArray cost=noArray(),
+                      float* lowerBound = 0, OutputArray flow = noArray() );
+
+//! segments the image using watershed algorithm
+CV_EXPORTS_W void watershed( InputArray image, InputOutputArray markers );
+
+//! filters image using meanshift algorithm
+CV_EXPORTS_W void pyrMeanShiftFiltering( InputArray src, OutputArray dst,
+                                         double sp, double sr, int maxLevel = 1,
+                                         TermCriteria termcrit=TermCriteria(TermCriteria::MAX_ITER+TermCriteria::EPS,5,1) );
+
+//! segments the image using GrabCut algorithm
+CV_EXPORTS_W void grabCut( InputArray img, InputOutputArray mask, Rect rect,
+                           InputOutputArray bgdModel, InputOutputArray fgdModel,
+                           int iterCount, int mode = GC_EVAL );
+
+
+//! builds the discrete Voronoi diagram
+CV_EXPORTS_AS(distanceTransformWithLabels) void distanceTransform( InputArray src, OutputArray dst,
+                                     OutputArray labels, int distanceType, int maskSize,
+                                     int labelType = DIST_LABEL_CCOMP );
+
+//! computes the distance transform map
+CV_EXPORTS_W void distanceTransform( InputArray src, OutputArray dst,
+                                     int distanceType, int maskSize );
+
+
+//! fills the semi-uniform image region starting from the specified seed point
+CV_EXPORTS int floodFill( InputOutputArray image,
+                          Point seedPoint, Scalar newVal, CV_OUT Rect* rect = 0,
+                          Scalar loDiff = Scalar(), Scalar upDiff = Scalar(),
+                          int flags = 4 );
+
+//! fills the semi-uniform image region and/or the mask starting from the specified seed point
+CV_EXPORTS_W int floodFill( InputOutputArray image, InputOutputArray mask,
+                            Point seedPoint, Scalar newVal, CV_OUT Rect* rect=0,
+                            Scalar loDiff = Scalar(), Scalar upDiff = Scalar(),
+                            int flags = 4 );
+
+//! converts image from one color space to another
+CV_EXPORTS_W void cvtColor( InputArray src, OutputArray dst, int code, int dstCn = 0 );
+
 // main function for all demosaicing procceses
 CV_EXPORTS_W void demosaicing(InputArray _src, OutputArray _dst, int code, int dcn = 0);
 
-}
+//! computes moments of the rasterized shape or a vector of points
+CV_EXPORTS_W Moments moments( InputArray array, bool binaryImage = false );
 
-#endif /* __cplusplus */
+//! computes 7 Hu invariants from the moments
+CV_EXPORTS void HuMoments( const Moments& moments, double hu[7] );
+
+CV_EXPORTS_W void HuMoments( const Moments& m, OutputArray hu );
+
+//! computes the proximity map for the raster template and the image where the template is searched for
+CV_EXPORTS_W void matchTemplate( InputArray image, InputArray templ,
+                                 OutputArray result, int method );
+
+
+// computes the connected components labeled image of boolean image ``image``
+// with 4 or 8 way connectivity - returns N, the total
+// number of labels [0, N-1] where 0 represents the background label.
+// ltype specifies the output label image type, an important
+// consideration based on the total number of labels or
+// alternatively the total number of pixels in the source image.
+CV_EXPORTS_W int connectedComponents(InputArray image, OutputArray labels,
+                                     int connectivity = 8, int ltype = CV_32S);
+
+CV_EXPORTS_W int connectedComponentsWithStats(InputArray image, OutputArray labels,
+                                              OutputArray stats, OutputArray centroids,
+                                              int connectivity = 8, int ltype = CV_32S);
+
+
+//! retrieves contours and the hierarchical information from black-n-white image.
+CV_EXPORTS_W void findContours( InputOutputArray image, OutputArrayOfArrays contours,
+                              OutputArray hierarchy, int mode,
+                              int method, Point offset = Point());
+
+//! retrieves contours from black-n-white image.
+CV_EXPORTS void findContours( InputOutputArray image, OutputArrayOfArrays contours,
+                              int mode, int method, Point offset = Point());
+
+//! approximates contour or a curve using Douglas-Peucker algorithm
+CV_EXPORTS_W void approxPolyDP( InputArray curve,
+                                OutputArray approxCurve,
+                                double epsilon, bool closed );
+
+//! computes the contour perimeter (closed=true) or a curve length
+CV_EXPORTS_W double arcLength( InputArray curve, bool closed );
+
+//! computes the bounding rectangle for a contour
+CV_EXPORTS_W Rect boundingRect( InputArray points );
+
+//! computes the contour area
+CV_EXPORTS_W double contourArea( InputArray contour, bool oriented = false );
+
+//! computes the minimal rotated rectangle for a set of points
+CV_EXPORTS_W RotatedRect minAreaRect( InputArray points );
+
+//! computes boxpoints
+CV_EXPORTS_W void boxPoints(RotatedRect box, OutputArray points);
+
+//! computes the minimal enclosing circle for a set of points
+CV_EXPORTS_W void minEnclosingCircle( InputArray points,
+                                      CV_OUT Point2f& center, CV_OUT float& radius );
+
+//! matches two contours using one of the available algorithms
+CV_EXPORTS_W double matchShapes( InputArray contour1, InputArray contour2,
+                                 int method, double parameter );
+
+//! computes convex hull for a set of 2D points.
+CV_EXPORTS_W void convexHull( InputArray points, OutputArray hull,
+                              bool clockwise = false, bool returnPoints = true );
+
+//! computes the contour convexity defects
+CV_EXPORTS_W void convexityDefects( InputArray contour, InputArray convexhull, OutputArray convexityDefects );
+
+//! returns true if the contour is convex. Does not support contours with self-intersection
+CV_EXPORTS_W bool isContourConvex( InputArray contour );
+
+//! finds intersection of two convex polygons
+CV_EXPORTS_W float intersectConvexConvex( InputArray _p1, InputArray _p2,
+                                          OutputArray _p12, bool handleNested = true );
+
+//! fits ellipse to the set of 2D points
+CV_EXPORTS_W RotatedRect fitEllipse( InputArray points );
+
+//! fits line to the set of 2D points using M-estimator algorithm
+CV_EXPORTS_W void fitLine( InputArray points, OutputArray line, int distType,
+                           double param, double reps, double aeps );
+
+//! checks if the point is inside the contour. Optionally computes the signed distance from the point to the contour boundary
+CV_EXPORTS_W double pointPolygonTest( InputArray contour, Point2f pt, bool measureDist );
+
+CV_EXPORTS_W Ptr<CLAHE> createCLAHE(double clipLimit = 40.0, Size tileGridSize = Size(8, 8));
+
+//! Ballard, D.H. (1981). Generalizing the Hough transform to detect arbitrary shapes. Pattern Recognition 13 (2): 111-122.
+//! Detects position only without traslation and rotation
+CV_EXPORTS Ptr<GeneralizedHoughBallard> createGeneralizedHoughBallard();
+
+//! Guil, N., González-Linares, J.M. and Zapata, E.L. (1999). Bidimensional shape detection using an invariant approach. Pattern Recognition 32 (6): 1025-1038.
+//! Detects position, traslation and rotation
+CV_EXPORTS Ptr<GeneralizedHoughGuil> createGeneralizedHoughGuil();
+
+} // cv
 
 #endif
-
-/* End of file. */
diff --git a/modules/imgproc/include/opencv2/imgproc/imgproc.hpp b/modules/imgproc/include/opencv2/imgproc/imgproc.hpp
index a7905d7d3a..4175bd0bc0 100644
--- a/modules/imgproc/include/opencv2/imgproc/imgproc.hpp
+++ b/modules/imgproc/include/opencv2/imgproc/imgproc.hpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/imgproc.hpp"
\ No newline at end of file
+#include "opencv2/imgproc.hpp"
diff --git a/modules/imgproc/include/opencv2/imgproc/imgproc_c.h b/modules/imgproc/include/opencv2/imgproc/imgproc_c.h
index c7b525c96d..4e2dc7142a 100644
--- a/modules/imgproc/include/opencv2/imgproc/imgproc_c.h
+++ b/modules/imgproc/include/opencv2/imgproc/imgproc_c.h
@@ -43,7 +43,6 @@
 #ifndef __OPENCV_IMGPROC_IMGPROC_C_H__
 #define __OPENCV_IMGPROC_IMGPROC_C_H__
 
-#include "opencv2/core/core_c.h"
 #include "opencv2/imgproc/types_c.h"
 
 #ifdef __cplusplus
diff --git a/modules/imgproc/include/opencv2/imgproc/types_c.h b/modules/imgproc/include/opencv2/imgproc/types_c.h
index 8e65269866..dd0d8b8a6e 100644
--- a/modules/imgproc/include/opencv2/imgproc/types_c.h
+++ b/modules/imgproc/include/opencv2/imgproc/types_c.h
@@ -383,6 +383,24 @@ typedef struct CvMoments
     double  m00, m10, m01, m20, m11, m02, m30, m21, m12, m03; /* spatial moments */
     double  mu20, mu11, mu02, mu30, mu21, mu12, mu03; /* central moments */
     double  inv_sqrt_m00; /* m00 != 0 ? 1/sqrt(m00) : 0 */
+
+#ifdef __cplusplus
+    CvMoments(){}
+    CvMoments(const cv::Moments& m)
+    {
+        m00 = m.m00; m10 = m.m10; m01 = m.m01;
+        m20 = m.m20; m11 = m.m11; m02 = m.m02;
+        m30 = m.m30; m21 = m.m21; m12 = m.m12; m03 = m.m03;
+        mu20 = m.mu20; mu11 = m.mu11; mu02 = m.mu02;
+        mu30 = m.mu30; mu21 = m.mu21; mu12 = m.mu12; mu03 = m.mu03;
+        double am00 = std::abs(m.m00);
+        inv_sqrt_m00 = am00 > DBL_EPSILON ? 1./std::sqrt(am00) : 0;
+    }
+    operator cv::Moments() const
+    {
+        return cv::Moments(m00, m10, m01, m20, m11, m02, m30, m21, m12, m03);
+    }
+#endif
 }
 CvMoments;
 
@@ -490,7 +508,8 @@ enum
     CV_COMP_CHISQR        =1,
     CV_COMP_INTERSECT     =2,
     CV_COMP_BHATTACHARYYA =3,
-    CV_COMP_HELLINGER     =CV_COMP_BHATTACHARYYA
+    CV_COMP_HELLINGER     =CV_COMP_BHATTACHARYYA,
+    CV_COMP_CHISQR_ALT    =4
 };
 
 /* Mask size for distance transform */
diff --git a/modules/imgproc/perf/perf_bilateral.cpp b/modules/imgproc/perf/perf_bilateral.cpp
index d6d9a63a65..af565f8a5e 100644
--- a/modules/imgproc/perf/perf_bilateral.cpp
+++ b/modules/imgproc/perf/perf_bilateral.cpp
@@ -15,7 +15,7 @@ PERF_TEST_P( TestBilateralFilter, BilateralFilter,
              Combine(
                 Values( szVGA, sz1080p ), // image size
                 Values( 3, 5 ), // d
-                ValuesIn( Mat_Type::all() ) // image type
+                Mat_Type::all() // image type
              )
 )
 {
diff --git a/modules/imgproc/perf/perf_blur.cpp b/modules/imgproc/perf/perf_blur.cpp
index 6604608b04..a6e31d3a8d 100644
--- a/modules/imgproc/perf/perf_blur.cpp
+++ b/modules/imgproc/perf/perf_blur.cpp
@@ -1,5 +1,4 @@
 #include "perf_precomp.hpp"
-#include "opencv2/core/internal.hpp"
 
 using namespace std;
 using namespace cv;
@@ -48,7 +47,7 @@ PERF_TEST_P(Size_MatType_BorderType3x3, gaussianBlur3x3,
             testing::Combine(
                 testing::Values(szODD, szQVGA, szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_16SC1, CV_32FC1),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
                 )
             )
 {
@@ -70,7 +69,7 @@ PERF_TEST_P(Size_MatType_BorderType3x3, blur3x3,
             testing::Combine(
                 testing::Values(szODD, szQVGA, szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_16SC1, CV_32FC1),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
                 )
             )
 {
@@ -92,7 +91,7 @@ PERF_TEST_P(Size_MatType_BorderType, blur16x16,
             testing::Combine(
                 testing::Values(szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_16SC1, CV_32FC1),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
                 )
             )
 {
@@ -114,7 +113,7 @@ PERF_TEST_P(Size_MatType_BorderType3x3, box3x3,
             testing::Combine(
                 testing::Values(szODD, szQVGA, szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_16SC1, CV_32SC1, CV_32FC1, CV_32FC3),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
                 )
             )
 {
@@ -136,7 +135,7 @@ PERF_TEST_P(Size_MatType_BorderType3x3, box3x3_inplace,
             testing::Combine(
                 testing::Values(szODD, szQVGA, szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_16SC1, CV_32SC1, CV_32FC1, CV_32FC3),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
                 )
             )
 {
@@ -164,7 +163,7 @@ PERF_TEST_P(Size_MatType_BorderType, gaussianBlur5x5,
             testing::Combine(
                 testing::Values(szODD, szQVGA, szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_16SC1, CV_32FC1),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
                 )
             )
 {
@@ -186,7 +185,7 @@ PERF_TEST_P(Size_MatType_BorderType, blur5x5,
             testing::Combine(
                 testing::Values(szVGA, sz720p),
                 testing::Values(CV_8UC1, CV_8UC4, CV_16UC1, CV_16SC1, CV_32FC1, CV_32FC3),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
                 )
             )
 {
diff --git a/modules/imgproc/perf/perf_cornerEigenValsAndVecs.cpp b/modules/imgproc/perf/perf_cornerEigenValsAndVecs.cpp
index 1fbf7655f5..5a323cc2ab 100644
--- a/modules/imgproc/perf/perf_cornerEigenValsAndVecs.cpp
+++ b/modules/imgproc/perf/perf_cornerEigenValsAndVecs.cpp
@@ -16,7 +16,7 @@ PERF_TEST_P(Img_BlockSize_ApertureSize_BorderType, cornerEigenValsAndVecs,
                 testing::Values( "stitching/a1.png", "cv/shared/pic5.png"),
                 testing::Values( 3, 5 ),
                 testing::Values( 3, 5 ),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
             )
           )
 {
@@ -37,4 +37,4 @@ PERF_TEST_P(Img_BlockSize_ApertureSize_BorderType, cornerEigenValsAndVecs,
     extractChannel(dst, l1, 0);
 
     SANITY_CHECK(l1, 2e-5);
-}
\ No newline at end of file
+}
diff --git a/modules/imgproc/perf/perf_cornerHarris.cpp b/modules/imgproc/perf/perf_cornerHarris.cpp
index 4006a7be3a..832845e7e7 100644
--- a/modules/imgproc/perf/perf_cornerHarris.cpp
+++ b/modules/imgproc/perf/perf_cornerHarris.cpp
@@ -17,7 +17,7 @@ PERF_TEST_P(Img_BlockSize_ApertureSize_k_BorderType, cornerHarris,
                 testing::Values( 3, 5 ),
                 testing::Values( 3, 5 ),
                 testing::Values( 0.04, 0.1 ),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
                 )
           )
 {
@@ -36,4 +36,4 @@ PERF_TEST_P(Img_BlockSize_ApertureSize_k_BorderType, cornerHarris,
     TEST_CYCLE() cornerHarris(src, dst, blockSize, apertureSize, k, borderType);
 
     SANITY_CHECK(dst, 2e-5);
-}
\ No newline at end of file
+}
diff --git a/modules/imgproc/perf/perf_cvt_color.cpp b/modules/imgproc/perf/perf_cvt_color.cpp
index 2038fc88c9..4ec464717b 100644
--- a/modules/imgproc/perf/perf_cvt_color.cpp
+++ b/modules/imgproc/perf/perf_cvt_color.cpp
@@ -9,114 +9,114 @@ using std::tr1::get;
 //extra color conversions supported implicitly
 enum
 {
-    CX_BGRA2HLS      = CV_COLORCVT_MAX + CV_BGR2HLS,
-    CX_BGRA2HLS_FULL = CV_COLORCVT_MAX + CV_BGR2HLS_FULL,
-    CX_BGRA2HSV      = CV_COLORCVT_MAX + CV_BGR2HSV,
-    CX_BGRA2HSV_FULL = CV_COLORCVT_MAX + CV_BGR2HSV_FULL,
-    CX_BGRA2Lab      = CV_COLORCVT_MAX + CV_BGR2Lab,
-    CX_BGRA2Luv      = CV_COLORCVT_MAX + CV_BGR2Luv,
-    CX_BGRA2XYZ      = CV_COLORCVT_MAX + CV_BGR2XYZ,
-    CX_BGRA2YCrCb    = CV_COLORCVT_MAX + CV_BGR2YCrCb,
-    CX_BGRA2YUV      = CV_COLORCVT_MAX + CV_BGR2YUV,
-    CX_HLS2BGRA      = CV_COLORCVT_MAX + CV_HLS2BGR,
-    CX_HLS2BGRA_FULL = CV_COLORCVT_MAX + CV_HLS2BGR_FULL,
-    CX_HLS2RGBA      = CV_COLORCVT_MAX + CV_HLS2RGB,
-    CX_HLS2RGBA_FULL = CV_COLORCVT_MAX + CV_HLS2RGB_FULL,
-    CX_HSV2BGRA      = CV_COLORCVT_MAX + CV_HSV2BGR,
-    CX_HSV2BGRA_FULL = CV_COLORCVT_MAX + CV_HSV2BGR_FULL,
-    CX_HSV2RGBA      = CV_COLORCVT_MAX + CV_HSV2RGB,
-    CX_HSV2RGBA_FULL = CV_COLORCVT_MAX + CV_HSV2RGB_FULL,
-    CX_Lab2BGRA      = CV_COLORCVT_MAX + CV_Lab2BGR,
-    CX_Lab2LBGRA     = CV_COLORCVT_MAX + CV_Lab2LBGR,
-    CX_Lab2LRGBA     = CV_COLORCVT_MAX + CV_Lab2LRGB,
-    CX_Lab2RGBA      = CV_COLORCVT_MAX + CV_Lab2RGB,
-    CX_LBGRA2Lab     = CV_COLORCVT_MAX + CV_LBGR2Lab,
-    CX_LBGRA2Luv     = CV_COLORCVT_MAX + CV_LBGR2Luv,
-    CX_LRGBA2Lab     = CV_COLORCVT_MAX + CV_LRGB2Lab,
-    CX_LRGBA2Luv     = CV_COLORCVT_MAX + CV_LRGB2Luv,
-    CX_Luv2BGRA      = CV_COLORCVT_MAX + CV_Luv2BGR,
-    CX_Luv2LBGRA     = CV_COLORCVT_MAX + CV_Luv2LBGR,
-    CX_Luv2LRGBA     = CV_COLORCVT_MAX + CV_Luv2LRGB,
-    CX_Luv2RGBA      = CV_COLORCVT_MAX + CV_Luv2RGB,
-    CX_RGBA2HLS      = CV_COLORCVT_MAX + CV_RGB2HLS,
-    CX_RGBA2HLS_FULL = CV_COLORCVT_MAX + CV_RGB2HLS_FULL,
-    CX_RGBA2HSV      = CV_COLORCVT_MAX + CV_RGB2HSV,
-    CX_RGBA2HSV_FULL = CV_COLORCVT_MAX + CV_RGB2HSV_FULL,
-    CX_RGBA2Lab      = CV_COLORCVT_MAX + CV_RGB2Lab,
-    CX_RGBA2Luv      = CV_COLORCVT_MAX + CV_RGB2Luv,
-    CX_RGBA2XYZ      = CV_COLORCVT_MAX + CV_RGB2XYZ,
-    CX_RGBA2YCrCb    = CV_COLORCVT_MAX + CV_RGB2YCrCb,
-    CX_RGBA2YUV      = CV_COLORCVT_MAX + CV_RGB2YUV,
-    CX_XYZ2BGRA      = CV_COLORCVT_MAX + CV_XYZ2BGR,
-    CX_XYZ2RGBA      = CV_COLORCVT_MAX + CV_XYZ2RGB,
-    CX_YCrCb2BGRA    = CV_COLORCVT_MAX + CV_YCrCb2BGR,
-    CX_YCrCb2RGBA    = CV_COLORCVT_MAX + CV_YCrCb2RGB,
-    CX_YUV2BGRA      = CV_COLORCVT_MAX + CV_YUV2BGR,
-    CX_YUV2RGBA      = CV_COLORCVT_MAX + CV_YUV2RGB
+    CX_BGRA2HLS      = COLOR_COLORCVT_MAX + COLOR_BGR2HLS,
+    CX_BGRA2HLS_FULL = COLOR_COLORCVT_MAX + COLOR_BGR2HLS_FULL,
+    CX_BGRA2HSV      = COLOR_COLORCVT_MAX + COLOR_BGR2HSV,
+    CX_BGRA2HSV_FULL = COLOR_COLORCVT_MAX + COLOR_BGR2HSV_FULL,
+    CX_BGRA2Lab      = COLOR_COLORCVT_MAX + COLOR_BGR2Lab,
+    CX_BGRA2Luv      = COLOR_COLORCVT_MAX + COLOR_BGR2Luv,
+    CX_BGRA2XYZ      = COLOR_COLORCVT_MAX + COLOR_BGR2XYZ,
+    CX_BGRA2YCrCb    = COLOR_COLORCVT_MAX + COLOR_BGR2YCrCb,
+    CX_BGRA2YUV      = COLOR_COLORCVT_MAX + COLOR_BGR2YUV,
+    CX_HLS2BGRA      = COLOR_COLORCVT_MAX + COLOR_HLS2BGR,
+    CX_HLS2BGRA_FULL = COLOR_COLORCVT_MAX + COLOR_HLS2BGR_FULL,
+    CX_HLS2RGBA      = COLOR_COLORCVT_MAX + COLOR_HLS2RGB,
+    CX_HLS2RGBA_FULL = COLOR_COLORCVT_MAX + COLOR_HLS2RGB_FULL,
+    CX_HSV2BGRA      = COLOR_COLORCVT_MAX + COLOR_HSV2BGR,
+    CX_HSV2BGRA_FULL = COLOR_COLORCVT_MAX + COLOR_HSV2BGR_FULL,
+    CX_HSV2RGBA      = COLOR_COLORCVT_MAX + COLOR_HSV2RGB,
+    CX_HSV2RGBA_FULL = COLOR_COLORCVT_MAX + COLOR_HSV2RGB_FULL,
+    CX_Lab2BGRA      = COLOR_COLORCVT_MAX + COLOR_Lab2BGR,
+    CX_Lab2LBGRA     = COLOR_COLORCVT_MAX + COLOR_Lab2LBGR,
+    CX_Lab2LRGBA     = COLOR_COLORCVT_MAX + COLOR_Lab2LRGB,
+    CX_Lab2RGBA      = COLOR_COLORCVT_MAX + COLOR_Lab2RGB,
+    CX_LBGRA2Lab     = COLOR_COLORCVT_MAX + COLOR_LBGR2Lab,
+    CX_LBGRA2Luv     = COLOR_COLORCVT_MAX + COLOR_LBGR2Luv,
+    CX_LRGBA2Lab     = COLOR_COLORCVT_MAX + COLOR_LRGB2Lab,
+    CX_LRGBA2Luv     = COLOR_COLORCVT_MAX + COLOR_LRGB2Luv,
+    CX_Luv2BGRA      = COLOR_COLORCVT_MAX + COLOR_Luv2BGR,
+    CX_Luv2LBGRA     = COLOR_COLORCVT_MAX + COLOR_Luv2LBGR,
+    CX_Luv2LRGBA     = COLOR_COLORCVT_MAX + COLOR_Luv2LRGB,
+    CX_Luv2RGBA      = COLOR_COLORCVT_MAX + COLOR_Luv2RGB,
+    CX_RGBA2HLS      = COLOR_COLORCVT_MAX + COLOR_RGB2HLS,
+    CX_RGBA2HLS_FULL = COLOR_COLORCVT_MAX + COLOR_RGB2HLS_FULL,
+    CX_RGBA2HSV      = COLOR_COLORCVT_MAX + COLOR_RGB2HSV,
+    CX_RGBA2HSV_FULL = COLOR_COLORCVT_MAX + COLOR_RGB2HSV_FULL,
+    CX_RGBA2Lab      = COLOR_COLORCVT_MAX + COLOR_RGB2Lab,
+    CX_RGBA2Luv      = COLOR_COLORCVT_MAX + COLOR_RGB2Luv,
+    CX_RGBA2XYZ      = COLOR_COLORCVT_MAX + COLOR_RGB2XYZ,
+    CX_RGBA2YCrCb    = COLOR_COLORCVT_MAX + COLOR_RGB2YCrCb,
+    CX_RGBA2YUV      = COLOR_COLORCVT_MAX + COLOR_RGB2YUV,
+    CX_XYZ2BGRA      = COLOR_COLORCVT_MAX + COLOR_XYZ2BGR,
+    CX_XYZ2RGBA      = COLOR_COLORCVT_MAX + COLOR_XYZ2RGB,
+    CX_YCrCb2BGRA    = COLOR_COLORCVT_MAX + COLOR_YCrCb2BGR,
+    CX_YCrCb2RGBA    = COLOR_COLORCVT_MAX + COLOR_YCrCb2RGB,
+    CX_YUV2BGRA      = COLOR_COLORCVT_MAX + COLOR_YUV2BGR,
+    CX_YUV2RGBA      = COLOR_COLORCVT_MAX + COLOR_YUV2RGB
 };
 
 CV_ENUM(CvtMode,
-    CV_BGR2BGR555, CV_BGR2BGR565, CV_BGR2BGRA, CV_BGR2GRAY,
-    CV_BGR2HLS, CV_BGR2HLS_FULL, CV_BGR2HSV, CV_BGR2HSV_FULL,
-    CV_BGR2Lab, CV_BGR2Luv, CV_BGR2RGB, CV_BGR2RGBA, CV_BGR2XYZ,
-    CV_BGR2YCrCb, CV_BGR2YUV, CV_BGR5552BGR, CV_BGR5552BGRA,
+    COLOR_BGR2BGR555, COLOR_BGR2BGR565, COLOR_BGR2BGRA, COLOR_BGR2GRAY,
+    COLOR_BGR2HLS, COLOR_BGR2HLS_FULL, COLOR_BGR2HSV, COLOR_BGR2HSV_FULL,
+    COLOR_BGR2Lab, COLOR_BGR2Luv, COLOR_BGR2RGB, COLOR_BGR2RGBA, COLOR_BGR2XYZ,
+    COLOR_BGR2YCrCb, COLOR_BGR2YUV, COLOR_BGR5552BGR, COLOR_BGR5552BGRA,
 
-    CV_BGR5552GRAY, CV_BGR5552RGB, CV_BGR5552RGBA, CV_BGR5652BGR,
-    CV_BGR5652BGRA, CV_BGR5652GRAY, CV_BGR5652RGB, CV_BGR5652RGBA,
+    COLOR_BGR5552GRAY, COLOR_BGR5552RGB, COLOR_BGR5552RGBA, COLOR_BGR5652BGR,
+    COLOR_BGR5652BGRA, COLOR_BGR5652GRAY, COLOR_BGR5652RGB, COLOR_BGR5652RGBA,
 
-    CV_BGRA2BGR, CV_BGRA2BGR555, CV_BGRA2BGR565, CV_BGRA2GRAY, CV_BGRA2RGBA,
+    COLOR_BGRA2BGR, COLOR_BGRA2BGR555, COLOR_BGRA2BGR565, COLOR_BGRA2GRAY, COLOR_BGRA2RGBA,
     CX_BGRA2HLS, CX_BGRA2HLS_FULL, CX_BGRA2HSV, CX_BGRA2HSV_FULL,
     CX_BGRA2Lab, CX_BGRA2Luv, CX_BGRA2XYZ,
     CX_BGRA2YCrCb, CX_BGRA2YUV,
 
-    CV_GRAY2BGR, CV_GRAY2BGR555, CV_GRAY2BGR565, CV_GRAY2BGRA,
+    COLOR_GRAY2BGR, COLOR_GRAY2BGR555, COLOR_GRAY2BGR565, COLOR_GRAY2BGRA,
 
-    CV_HLS2BGR, CV_HLS2BGR_FULL, CV_HLS2RGB, CV_HLS2RGB_FULL,
+    COLOR_HLS2BGR, COLOR_HLS2BGR_FULL, COLOR_HLS2RGB, COLOR_HLS2RGB_FULL,
     CX_HLS2BGRA, CX_HLS2BGRA_FULL, CX_HLS2RGBA, CX_HLS2RGBA_FULL,
 
-    CV_HSV2BGR, CV_HSV2BGR_FULL, CV_HSV2RGB, CV_HSV2RGB_FULL,
+    COLOR_HSV2BGR, COLOR_HSV2BGR_FULL, COLOR_HSV2RGB, COLOR_HSV2RGB_FULL,
     CX_HSV2BGRA, CX_HSV2BGRA_FULL, CX_HSV2RGBA,    CX_HSV2RGBA_FULL,
 
-    CV_Lab2BGR, CV_Lab2LBGR, CV_Lab2LRGB, CV_Lab2RGB,
+    COLOR_Lab2BGR, COLOR_Lab2LBGR, COLOR_Lab2LRGB, COLOR_Lab2RGB,
     CX_Lab2BGRA, CX_Lab2LBGRA, CX_Lab2LRGBA, CX_Lab2RGBA,
 
-    CV_LBGR2Lab, CV_LBGR2Luv, CV_LRGB2Lab, CV_LRGB2Luv,
+    COLOR_LBGR2Lab, COLOR_LBGR2Luv, COLOR_LRGB2Lab, COLOR_LRGB2Luv,
     CX_LBGRA2Lab, CX_LBGRA2Luv, CX_LRGBA2Lab, CX_LRGBA2Luv,
 
-    CV_Luv2BGR, CV_Luv2LBGR, CV_Luv2LRGB, CV_Luv2RGB,
+    COLOR_Luv2BGR, COLOR_Luv2LBGR, COLOR_Luv2LRGB, COLOR_Luv2RGB,
     CX_Luv2BGRA, CX_Luv2LBGRA, CX_Luv2LRGBA, CX_Luv2RGBA,
 
-    CV_RGB2BGR555, CV_RGB2BGR565, CV_RGB2GRAY,
-    CV_RGB2HLS, CV_RGB2HLS_FULL, CV_RGB2HSV, CV_RGB2HSV_FULL,
-    CV_RGB2Lab, CV_RGB2Luv, CV_RGB2XYZ, CV_RGB2YCrCb, CV_RGB2YUV,
+    COLOR_RGB2BGR555, COLOR_RGB2BGR565, COLOR_RGB2GRAY,
+    COLOR_RGB2HLS, COLOR_RGB2HLS_FULL, COLOR_RGB2HSV, COLOR_RGB2HSV_FULL,
+    COLOR_RGB2Lab, COLOR_RGB2Luv, COLOR_RGB2XYZ, COLOR_RGB2YCrCb, COLOR_RGB2YUV,
 
-    CV_RGBA2BGR, CV_RGBA2BGR555, CV_RGBA2BGR565, CV_RGBA2GRAY,
+    COLOR_RGBA2BGR, COLOR_RGBA2BGR555, COLOR_RGBA2BGR565, COLOR_RGBA2GRAY,
     CX_RGBA2HLS, CX_RGBA2HLS_FULL, CX_RGBA2HSV, CX_RGBA2HSV_FULL,
     CX_RGBA2Lab, CX_RGBA2Luv, CX_RGBA2XYZ,
     CX_RGBA2YCrCb, CX_RGBA2YUV,
 
-    CV_XYZ2BGR, CV_XYZ2RGB, CX_XYZ2BGRA, CX_XYZ2RGBA,
+    COLOR_XYZ2BGR, COLOR_XYZ2RGB, CX_XYZ2BGRA, CX_XYZ2RGBA,
 
-    CV_YCrCb2BGR, CV_YCrCb2RGB, CX_YCrCb2BGRA, CX_YCrCb2RGBA,
-    CV_YUV2BGR, CV_YUV2RGB, CX_YUV2BGRA, CX_YUV2RGBA
+    COLOR_YCrCb2BGR, COLOR_YCrCb2RGB, CX_YCrCb2BGRA, CX_YCrCb2RGBA,
+    COLOR_YUV2BGR, COLOR_YUV2RGB, CX_YUV2BGRA, CX_YUV2RGBA
     )
 
 
 CV_ENUM(CvtModeBayer,
-    CV_BayerBG2BGR, CV_BayerBG2BGR_VNG, CV_BayerBG2GRAY,
-    CV_BayerGB2BGR, CV_BayerGB2BGR_VNG, CV_BayerGB2GRAY,
-    CV_BayerGR2BGR, CV_BayerGR2BGR_VNG, CV_BayerGR2GRAY,
-    CV_BayerRG2BGR, CV_BayerRG2BGR_VNG, CV_BayerRG2GRAY
+    COLOR_BayerBG2BGR, COLOR_BayerBG2BGR_VNG, COLOR_BayerBG2GRAY,
+    COLOR_BayerGB2BGR, COLOR_BayerGB2BGR_VNG, COLOR_BayerGB2GRAY,
+    COLOR_BayerGR2BGR, COLOR_BayerGR2BGR_VNG, COLOR_BayerGR2GRAY,
+    COLOR_BayerRG2BGR, COLOR_BayerRG2BGR_VNG, COLOR_BayerRG2GRAY
     )
 
 
-CV_ENUM(CvtMode2, CV_YUV2BGR_NV12, CV_YUV2BGRA_NV12, CV_YUV2RGB_NV12, CV_YUV2RGBA_NV12, CV_YUV2BGR_NV21, CV_YUV2BGRA_NV21, CV_YUV2RGB_NV21, CV_YUV2RGBA_NV21,
-                  CV_YUV2BGR_YV12, CV_YUV2BGRA_YV12, CV_YUV2RGB_YV12, CV_YUV2RGBA_YV12, CV_YUV2BGR_IYUV, CV_YUV2BGRA_IYUV, CV_YUV2RGB_IYUV, CV_YUV2RGBA_IYUV,
-                  COLOR_YUV2GRAY_420, CV_YUV2RGB_UYVY, CV_YUV2BGR_UYVY, CV_YUV2RGBA_UYVY, CV_YUV2BGRA_UYVY, CV_YUV2RGB_YUY2, CV_YUV2BGR_YUY2, CV_YUV2RGB_YVYU,
-                  CV_YUV2BGR_YVYU, CV_YUV2RGBA_YUY2, CV_YUV2BGRA_YUY2, CV_YUV2RGBA_YVYU, CV_YUV2BGRA_YVYU)
+CV_ENUM(CvtMode2, COLOR_YUV2BGR_NV12, COLOR_YUV2BGRA_NV12, COLOR_YUV2RGB_NV12, COLOR_YUV2RGBA_NV12, COLOR_YUV2BGR_NV21, COLOR_YUV2BGRA_NV21, COLOR_YUV2RGB_NV21, COLOR_YUV2RGBA_NV21,
+                  COLOR_YUV2BGR_YV12, COLOR_YUV2BGRA_YV12, COLOR_YUV2RGB_YV12, COLOR_YUV2RGBA_YV12, COLOR_YUV2BGR_IYUV, COLOR_YUV2BGRA_IYUV, COLOR_YUV2RGB_IYUV, COLOR_YUV2RGBA_IYUV,
+                  COLOR_YUV2GRAY_420, COLOR_YUV2RGB_UYVY, COLOR_YUV2BGR_UYVY, COLOR_YUV2RGBA_UYVY, COLOR_YUV2BGRA_UYVY, COLOR_YUV2RGB_YUY2, COLOR_YUV2BGR_YUY2, COLOR_YUV2RGB_YVYU,
+                  COLOR_YUV2BGR_YVYU, COLOR_YUV2RGBA_YUY2, COLOR_YUV2BGRA_YUY2, COLOR_YUV2RGBA_YVYU, COLOR_YUV2BGRA_YVYU)
 
-CV_ENUM(CvtMode3, CV_RGB2YUV_IYUV, CV_BGR2YUV_IYUV, CV_RGBA2YUV_IYUV, CV_BGRA2YUV_IYUV,
-                  CV_RGB2YUV_YV12, CV_BGR2YUV_YV12, CV_RGBA2YUV_YV12, CV_BGRA2YUV_YV12)
+CV_ENUM(CvtMode3, COLOR_RGB2YUV_IYUV, COLOR_BGR2YUV_IYUV, COLOR_RGBA2YUV_IYUV, COLOR_BGRA2YUV_IYUV,
+                  COLOR_RGB2YUV_YV12, COLOR_BGR2YUV_YV12, COLOR_RGBA2YUV_YV12, COLOR_BGRA2YUV_YV12)
 
 struct ChPair
 {
@@ -128,74 +128,74 @@ ChPair getConversionInfo(int cvtMode)
 {
     switch(cvtMode)
     {
-    case CV_BayerBG2GRAY: case CV_BayerGB2GRAY:
-    case CV_BayerGR2GRAY: case CV_BayerRG2GRAY:
-    case CV_YUV2GRAY_420:
+    case COLOR_BayerBG2GRAY: case COLOR_BayerGB2GRAY:
+    case COLOR_BayerGR2GRAY: case COLOR_BayerRG2GRAY:
+    case COLOR_YUV2GRAY_420:
         return ChPair(1,1);
-    case CV_GRAY2BGR555: case CV_GRAY2BGR565:
+    case COLOR_GRAY2BGR555: case COLOR_GRAY2BGR565:
         return ChPair(1,2);
-    case CV_BayerBG2BGR: case CV_BayerBG2BGR_VNG:
-    case CV_BayerGB2BGR: case CV_BayerGB2BGR_VNG:
-    case CV_BayerGR2BGR: case CV_BayerGR2BGR_VNG:
-    case CV_BayerRG2BGR: case CV_BayerRG2BGR_VNG:
-    case CV_GRAY2BGR:
-    case CV_YUV2BGR_NV12: case CV_YUV2RGB_NV12:
-    case CV_YUV2BGR_NV21: case CV_YUV2RGB_NV21:
-    case CV_YUV2BGR_YV12: case CV_YUV2RGB_YV12:
-    case CV_YUV2BGR_IYUV: case CV_YUV2RGB_IYUV:
+    case COLOR_BayerBG2BGR: case COLOR_BayerBG2BGR_VNG:
+    case COLOR_BayerGB2BGR: case COLOR_BayerGB2BGR_VNG:
+    case COLOR_BayerGR2BGR: case COLOR_BayerGR2BGR_VNG:
+    case COLOR_BayerRG2BGR: case COLOR_BayerRG2BGR_VNG:
+    case COLOR_GRAY2BGR:
+    case COLOR_YUV2BGR_NV12: case COLOR_YUV2RGB_NV12:
+    case COLOR_YUV2BGR_NV21: case COLOR_YUV2RGB_NV21:
+    case COLOR_YUV2BGR_YV12: case COLOR_YUV2RGB_YV12:
+    case COLOR_YUV2BGR_IYUV: case COLOR_YUV2RGB_IYUV:
         return ChPair(1,3);
-    case CV_GRAY2BGRA:
-    case CV_YUV2BGRA_NV12: case CV_YUV2RGBA_NV12:
-    case CV_YUV2BGRA_NV21: case CV_YUV2RGBA_NV21:
-    case CV_YUV2BGRA_YV12: case CV_YUV2RGBA_YV12:
-    case CV_YUV2BGRA_IYUV: case CV_YUV2RGBA_IYUV:
+    case COLOR_GRAY2BGRA:
+    case COLOR_YUV2BGRA_NV12: case COLOR_YUV2RGBA_NV12:
+    case COLOR_YUV2BGRA_NV21: case COLOR_YUV2RGBA_NV21:
+    case COLOR_YUV2BGRA_YV12: case COLOR_YUV2RGBA_YV12:
+    case COLOR_YUV2BGRA_IYUV: case COLOR_YUV2RGBA_IYUV:
         return ChPair(1,4);
-    case CV_BGR5552GRAY: case CV_BGR5652GRAY:
+    case COLOR_BGR5552GRAY: case COLOR_BGR5652GRAY:
         return ChPair(2,1);
-    case CV_BGR5552BGR: case CV_BGR5552RGB:
-    case CV_BGR5652BGR: case CV_BGR5652RGB:
-    case CV_YUV2RGB_UYVY: case CV_YUV2BGR_UYVY:
-    case CV_YUV2RGBA_UYVY: case CV_YUV2BGRA_UYVY:
-    case CV_YUV2RGB_YUY2: case CV_YUV2BGR_YUY2:
-    case CV_YUV2RGB_YVYU: case CV_YUV2BGR_YVYU:
-    case CV_YUV2RGBA_YUY2: case CV_YUV2BGRA_YUY2:
-    case CV_YUV2RGBA_YVYU: case CV_YUV2BGRA_YVYU:
+    case COLOR_BGR5552BGR: case COLOR_BGR5552RGB:
+    case COLOR_BGR5652BGR: case COLOR_BGR5652RGB:
+    case COLOR_YUV2RGB_UYVY: case COLOR_YUV2BGR_UYVY:
+    case COLOR_YUV2RGBA_UYVY: case COLOR_YUV2BGRA_UYVY:
+    case COLOR_YUV2RGB_YUY2: case COLOR_YUV2BGR_YUY2:
+    case COLOR_YUV2RGB_YVYU: case COLOR_YUV2BGR_YVYU:
+    case COLOR_YUV2RGBA_YUY2: case COLOR_YUV2BGRA_YUY2:
+    case COLOR_YUV2RGBA_YVYU: case COLOR_YUV2BGRA_YVYU:
         return ChPair(2,3);
-    case CV_BGR5552BGRA: case CV_BGR5552RGBA:
-    case CV_BGR5652BGRA: case CV_BGR5652RGBA:
+    case COLOR_BGR5552BGRA: case COLOR_BGR5552RGBA:
+    case COLOR_BGR5652BGRA: case COLOR_BGR5652RGBA:
         return ChPair(2,4);
-    case CV_BGR2GRAY: case CV_RGB2GRAY:
-    case CV_RGB2YUV_IYUV: case CV_RGB2YUV_YV12:
-    case CV_BGR2YUV_IYUV: case CV_BGR2YUV_YV12:
+    case COLOR_BGR2GRAY: case COLOR_RGB2GRAY:
+    case COLOR_RGB2YUV_IYUV: case COLOR_RGB2YUV_YV12:
+    case COLOR_BGR2YUV_IYUV: case COLOR_BGR2YUV_YV12:
         return ChPair(3,1);
-    case CV_BGR2BGR555: case CV_BGR2BGR565:
-    case CV_RGB2BGR555: case CV_RGB2BGR565:
+    case COLOR_BGR2BGR555: case COLOR_BGR2BGR565:
+    case COLOR_RGB2BGR555: case COLOR_RGB2BGR565:
         return ChPair(3,2);
-    case CV_BGR2HLS: case CV_BGR2HLS_FULL:
-    case CV_BGR2HSV: case CV_BGR2HSV_FULL:
-    case CV_BGR2Lab: case CV_BGR2Luv:
-    case CV_BGR2RGB: case CV_BGR2XYZ:
-    case CV_BGR2YCrCb: case CV_BGR2YUV:
-    case CV_HLS2BGR: case CV_HLS2BGR_FULL:
-    case CV_HLS2RGB: case CV_HLS2RGB_FULL:
-    case CV_HSV2BGR: case CV_HSV2BGR_FULL:
-    case CV_HSV2RGB: case CV_HSV2RGB_FULL:
-    case CV_Lab2BGR: case CV_Lab2LBGR:
-    case CV_Lab2LRGB: case CV_Lab2RGB:
-    case CV_LBGR2Lab: case CV_LBGR2Luv:
-    case CV_LRGB2Lab: case CV_LRGB2Luv:
-    case CV_Luv2BGR: case CV_Luv2LBGR:
-    case CV_Luv2LRGB: case CV_Luv2RGB:
-    case CV_RGB2HLS: case CV_RGB2HLS_FULL:
-    case CV_RGB2HSV: case CV_RGB2HSV_FULL:
-    case CV_RGB2Lab: case CV_RGB2Luv:
-    case CV_RGB2XYZ: case CV_RGB2YCrCb:
-    case CV_RGB2YUV: case CV_XYZ2BGR:
-    case CV_XYZ2RGB: case CV_YCrCb2BGR:
-    case CV_YCrCb2RGB: case CV_YUV2BGR:
-    case CV_YUV2RGB:
+    case COLOR_BGR2HLS: case COLOR_BGR2HLS_FULL:
+    case COLOR_BGR2HSV: case COLOR_BGR2HSV_FULL:
+    case COLOR_BGR2Lab: case COLOR_BGR2Luv:
+    case COLOR_BGR2RGB: case COLOR_BGR2XYZ:
+    case COLOR_BGR2YCrCb: case COLOR_BGR2YUV:
+    case COLOR_HLS2BGR: case COLOR_HLS2BGR_FULL:
+    case COLOR_HLS2RGB: case COLOR_HLS2RGB_FULL:
+    case COLOR_HSV2BGR: case COLOR_HSV2BGR_FULL:
+    case COLOR_HSV2RGB: case COLOR_HSV2RGB_FULL:
+    case COLOR_Lab2BGR: case COLOR_Lab2LBGR:
+    case COLOR_Lab2LRGB: case COLOR_Lab2RGB:
+    case COLOR_LBGR2Lab: case COLOR_LBGR2Luv:
+    case COLOR_LRGB2Lab: case COLOR_LRGB2Luv:
+    case COLOR_Luv2BGR: case COLOR_Luv2LBGR:
+    case COLOR_Luv2LRGB: case COLOR_Luv2RGB:
+    case COLOR_RGB2HLS: case COLOR_RGB2HLS_FULL:
+    case COLOR_RGB2HSV: case COLOR_RGB2HSV_FULL:
+    case COLOR_RGB2Lab: case COLOR_RGB2Luv:
+    case COLOR_RGB2XYZ: case COLOR_RGB2YCrCb:
+    case COLOR_RGB2YUV: case COLOR_XYZ2BGR:
+    case COLOR_XYZ2RGB: case COLOR_YCrCb2BGR:
+    case COLOR_YCrCb2RGB: case COLOR_YUV2BGR:
+    case COLOR_YUV2RGB:
         return ChPair(3,3);
-    case CV_BGR2BGRA: case CV_BGR2RGBA:
+    case COLOR_BGR2BGRA: case COLOR_BGR2RGBA:
     case CX_HLS2BGRA: case CX_HLS2BGRA_FULL:
     case CX_HLS2RGBA: case CX_HLS2RGBA_FULL:
     case CX_HSV2BGRA: case CX_HSV2BGRA_FULL:
@@ -208,27 +208,27 @@ ChPair getConversionInfo(int cvtMode)
     case CX_YCrCb2BGRA: case CX_YCrCb2RGBA:
     case CX_YUV2BGRA: case CX_YUV2RGBA:
         return ChPair(3,4);
-    case CV_BGRA2GRAY: case CV_RGBA2GRAY:
-    case CV_RGBA2YUV_IYUV: case CV_RGBA2YUV_YV12:
-    case CV_BGRA2YUV_IYUV: case CV_BGRA2YUV_YV12:
+    case COLOR_BGRA2GRAY: case COLOR_RGBA2GRAY:
+    case COLOR_RGBA2YUV_IYUV: case COLOR_RGBA2YUV_YV12:
+    case COLOR_BGRA2YUV_IYUV: case COLOR_BGRA2YUV_YV12:
         return ChPair(4,1);
-    case CV_BGRA2BGR555: case CV_BGRA2BGR565:
-    case CV_RGBA2BGR555: case CV_RGBA2BGR565:
+    case COLOR_BGRA2BGR555: case COLOR_BGRA2BGR565:
+    case COLOR_RGBA2BGR555: case COLOR_RGBA2BGR565:
         return ChPair(4,2);
-    case CV_BGRA2BGR: case CX_BGRA2HLS:
+    case COLOR_BGRA2BGR: case CX_BGRA2HLS:
     case CX_BGRA2HLS_FULL: case CX_BGRA2HSV:
     case CX_BGRA2HSV_FULL: case CX_BGRA2Lab:
     case CX_BGRA2Luv: case CX_BGRA2XYZ:
     case CX_BGRA2YCrCb: case CX_BGRA2YUV:
     case CX_LBGRA2Lab: case CX_LBGRA2Luv:
     case CX_LRGBA2Lab: case CX_LRGBA2Luv:
-    case CV_RGBA2BGR: case CX_RGBA2HLS:
+    case COLOR_RGBA2BGR: case CX_RGBA2HLS:
     case CX_RGBA2HLS_FULL: case CX_RGBA2HSV:
     case CX_RGBA2HSV_FULL: case CX_RGBA2Lab:
     case CX_RGBA2Luv: case CX_RGBA2XYZ:
     case CX_RGBA2YCrCb: case CX_RGBA2YUV:
         return ChPair(4,3);
-    case CV_BGRA2RGBA:
+    case COLOR_BGRA2RGBA:
         return ChPair(4,4);
     default:
         ADD_FAILURE() << "Unknown conversion type";
@@ -243,14 +243,14 @@ typedef perf::TestBaseWithParam<Size_CvtMode_t> Size_CvtMode;
 PERF_TEST_P(Size_CvtMode, cvtColor8u,
             testing::Combine(
                 testing::Values(::perf::szODD, ::perf::szVGA, ::perf::sz1080p),
-                testing::ValuesIn(CvtMode::all())
+                CvtMode::all()
                 )
             )
 {
     Size sz = get<0>(GetParam());
     int mode = get<1>(GetParam());
     ChPair ch = getConversionInfo(mode);
-    mode %= CV_COLORCVT_MAX;
+    mode %= COLOR_COLORCVT_MAX;
 
     Mat src(sz, CV_8UC(ch.scn));
     Mat dst(sz, CV_8UC(ch.dcn));
@@ -258,7 +258,8 @@ PERF_TEST_P(Size_CvtMode, cvtColor8u,
     declare.time(100);
     declare.in(src, WARMUP_RNG).out(dst);
 
-    TEST_CYCLE() cvtColor(src, dst, mode, ch.dcn);
+    int runs = sz.width <= 320 ? 100 : 5;
+    TEST_CYCLE_MULTIRUN(runs) cvtColor(src, dst, mode, ch.dcn);
 
     SANITY_CHECK(dst, 1);
 }
@@ -269,14 +270,14 @@ typedef perf::TestBaseWithParam<Size_CvtMode_Bayer_t> Size_CvtMode_Bayer;
 PERF_TEST_P(Size_CvtMode_Bayer, cvtColorBayer8u,
             testing::Combine(
                 testing::Values(::perf::szODD, ::perf::szVGA),
-                testing::ValuesIn(CvtModeBayer::all())
+                CvtModeBayer::all()
                 )
             )
 {
     Size sz = get<0>(GetParam());
     int mode = get<1>(GetParam());
     ChPair ch = getConversionInfo(mode);
-    mode %= CV_COLORCVT_MAX;
+    mode %= COLOR_COLORCVT_MAX;
 
     Mat src(sz, CV_8UC(ch.scn));
     Mat dst(sz, CV_8UC(ch.dcn));
@@ -295,7 +296,7 @@ typedef perf::TestBaseWithParam<Size_CvtMode2_t> Size_CvtMode2;
 PERF_TEST_P(Size_CvtMode2, cvtColorYUV420,
             testing::Combine(
                 testing::Values(szVGA, sz1080p, Size(130, 60)),
-                testing::ValuesIn(CvtMode2::all())
+                CvtMode2::all()
                 )
             )
 {
@@ -320,7 +321,7 @@ typedef perf::TestBaseWithParam<Size_CvtMode3_t> Size_CvtMode3;
 PERF_TEST_P(Size_CvtMode3, cvtColorRGB2YUV420p,
             testing::Combine(
                 testing::Values(szVGA, sz720p, sz1080p, Size(130, 60)),
-                testing::ValuesIn(CvtMode3::all())
+                CvtMode3::all()
                 )
             )
 {
@@ -334,7 +335,8 @@ PERF_TEST_P(Size_CvtMode3, cvtColorRGB2YUV420p,
     declare.time(100);
     declare.in(src, WARMUP_RNG).out(dst);
 
-    TEST_CYCLE() cvtColor(src, dst, mode, ch.dcn);
+    int runs = (sz.width <= 640) ? 10 : 1;
+    TEST_CYCLE_MULTIRUN(runs) cvtColor(src, dst, mode, ch.dcn);
 
     SANITY_CHECK(dst, 1);
 }
@@ -347,7 +349,7 @@ typedef perf::TestBaseWithParam<EdgeAwareParams> EdgeAwareDemosaicingTest;
 PERF_TEST_P(EdgeAwareDemosaicingTest, demosaicingEA,
             testing::Combine(
                 testing::Values(szVGA, sz720p, sz1080p, Size(130, 60)),
-                testing::ValuesIn(EdgeAwareBayerMode::all())
+                EdgeAwareBayerMode::all()
                 )
             )
 {
diff --git a/modules/imgproc/perf/perf_filter2d.cpp b/modules/imgproc/perf/perf_filter2d.cpp
index 5bdd470176..cfce852b71 100644
--- a/modules/imgproc/perf/perf_filter2d.cpp
+++ b/modules/imgproc/perf/perf_filter2d.cpp
@@ -17,7 +17,7 @@ PERF_TEST_P( TestFilter2d, Filter2d,
              Combine(
                 Values( Size(320, 240), sz1080p ),
                 Values( 3, 5 ),
-                ValuesIn( BorderMode::all() )
+                BorderMode::all()
              )
 )
 {
@@ -72,5 +72,3 @@ PERF_TEST_P( Image_KernelSize, GaborFilter2d,
 
     SANITY_CHECK(filteredImage, 1e-3);
 }
-
-
diff --git a/modules/imgproc/perf/perf_histogram.cpp b/modules/imgproc/perf/perf_histogram.cpp
index 3b320633ae..1789470c20 100644
--- a/modules/imgproc/perf/perf_histogram.cpp
+++ b/modules/imgproc/perf/perf_histogram.cpp
@@ -28,14 +28,14 @@ PERF_TEST_P(Size_Source, calcHist1d,
     int dims = 1;
     int numberOfImages = 1;
 
-    const float r[] = {rangeLow, rangeHight};
-    const float* ranges[] = {r};
+    const float range[] = {rangeLow, rangeHight};
+    const float* ranges[] = {range};
 
     randu(source, rangeLow, rangeHight);
 
     declare.in(source);
 
-    TEST_CYCLE()
+    TEST_CYCLE_MULTIRUN(3)
     {
         calcHist(&source, numberOfImages, channels, Mat(), hist, dims, histSize, ranges);
     }
@@ -115,3 +115,25 @@ PERF_TEST_P(MatSize, equalizeHist,
 
     SANITY_CHECK(destination);
 }
+
+typedef tr1::tuple<Size, double> Sz_ClipLimit_t;
+typedef TestBaseWithParam<Sz_ClipLimit_t> Sz_ClipLimit;
+
+PERF_TEST_P(Sz_ClipLimit, CLAHE,
+            testing::Combine(testing::Values(::perf::szVGA, ::perf::sz720p, ::perf::sz1080p),
+                             testing::Values(0.0, 40.0))
+            )
+{
+    const Size size = get<0>(GetParam());
+    const double clipLimit = get<1>(GetParam());
+
+    Mat src(size, CV_8UC1);
+    declare.in(src, WARMUP_RNG);
+
+    Ptr<CLAHE> clahe = createCLAHE(clipLimit);
+    Mat dst;
+
+    TEST_CYCLE() clahe->apply(src, dst);
+
+    SANITY_CHECK(dst);
+}
diff --git a/modules/imgproc/perf/perf_matchTemplate.cpp b/modules/imgproc/perf/perf_matchTemplate.cpp
index a89435b535..a1f0ddeab1 100644
--- a/modules/imgproc/perf/perf_matchTemplate.cpp
+++ b/modules/imgproc/perf/perf_matchTemplate.cpp
@@ -6,7 +6,7 @@ using namespace perf;
 using std::tr1::make_tuple;
 using std::tr1::get;
 
-CV_ENUM(MethodType, CV_TM_SQDIFF, CV_TM_SQDIFF_NORMED, CV_TM_CCORR, CV_TM_CCORR_NORMED, CV_TM_CCOEFF, CV_TM_CCOEFF_NORMED)
+CV_ENUM(MethodType, TM_SQDIFF, TM_SQDIFF_NORMED, TM_CCORR, TM_CCORR_NORMED, TM_CCOEFF, TM_CCOEFF_NORMED)
 
 typedef std::tr1::tuple<Size, Size, MethodType> ImgSize_TmplSize_Method_t;
 typedef perf::TestBaseWithParam<ImgSize_TmplSize_Method_t> ImgSize_TmplSize_Method;
@@ -18,7 +18,7 @@ PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateSmall,
                                 cv::Size(1024, 768), cv::Size(1280, 1024)),
                 testing::Values(cv::Size(12, 12), cv::Size(28, 9),
                                 cv::Size(8, 30), cv::Size(16, 16)),
-                testing::ValuesIn(MethodType::all())
+                MethodType::all()
                 )
             )
 {
@@ -39,9 +39,9 @@ PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateSmall,
     TEST_CYCLE() matchTemplate(img, tmpl, result, method);
 
     bool isNormed =
-        method == CV_TM_CCORR_NORMED ||
-        method == CV_TM_SQDIFF_NORMED ||
-        method == CV_TM_CCOEFF_NORMED;
+        method == TM_CCORR_NORMED ||
+        method == TM_SQDIFF_NORMED ||
+        method == TM_CCOEFF_NORMED;
     double eps = isNormed ? 1e-6
         : 255 * 255 * tmpl.total() * 1e-6;
 
@@ -52,7 +52,7 @@ PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateBig,
             testing::Combine(
                 testing::Values(cv::Size(1280, 1024)),
                 testing::Values(cv::Size(1260, 1000), cv::Size(1261, 1013)),
-                testing::ValuesIn(MethodType::all())
+                MethodType::all()
                 )
     )
 {
@@ -73,9 +73,9 @@ PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateBig,
     TEST_CYCLE() matchTemplate(img, tmpl, result, method);
 
     bool isNormed =
-        method == CV_TM_CCORR_NORMED ||
-        method == CV_TM_SQDIFF_NORMED ||
-        method == CV_TM_CCOEFF_NORMED;
+        method == TM_CCORR_NORMED ||
+        method == TM_SQDIFF_NORMED ||
+        method == TM_CCOEFF_NORMED;
     double eps = isNormed ? 1e-6
         : 255 * 255 * tmpl.total() * 1e-6;
 
diff --git a/modules/imgproc/perf/perf_morph.cpp b/modules/imgproc/perf/perf_morph.cpp
index 9aadeaff52..d3dbba38fb 100644
--- a/modules/imgproc/perf/perf_morph.cpp
+++ b/modules/imgproc/perf/perf_morph.cpp
@@ -19,7 +19,8 @@ PERF_TEST_P(Size_MatType, erode, TYPICAL_MATS_MORPH)
 
     declare.in(src, WARMUP_RNG).out(dst);
 
-    TEST_CYCLE() erode(src, dst, noArray());
+    int runs = (sz.width <= 320) ? 15 : 1;
+    TEST_CYCLE_MULTIRUN(runs) erode(src, dst, noArray());
 
     SANITY_CHECK(dst);
 }
diff --git a/modules/imgproc/perf/perf_precomp.cpp b/modules/imgproc/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/imgproc/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/imgproc/perf/perf_remap.cpp b/modules/imgproc/perf/perf_remap.cpp
index f1c8956b31..92c6007a2b 100644
--- a/modules/imgproc/perf/perf_remap.cpp
+++ b/modules/imgproc/perf/perf_remap.cpp
@@ -16,7 +16,7 @@ PERF_TEST_P( TestRemap, Remap,
                 Values( szVGA, sz1080p ),
                 Values( CV_16UC1, CV_16SC1, CV_32FC1 ),
                 Values( CV_16SC2, CV_32FC1, CV_32FC2 ),
-                ValuesIn( InterType::all() )
+                InterType::all()
              )
 )
 {
@@ -63,7 +63,8 @@ PERF_TEST_P( TestRemap, Remap,
 
     declare.in(src, WARMUP_RNG).out(dst).time(20);
 
-    TEST_CYCLE() remap(src, dst, map1, map2, inter_type);
+    int runs = (sz.width <= 640) ? 3 : 1;
+    TEST_CYCLE_MULTIRUN(runs) remap(src, dst, map1, map2, inter_type);
 
     SANITY_CHECK(dst);
 }
diff --git a/modules/imgproc/perf/perf_resize.cpp b/modules/imgproc/perf/perf_resize.cpp
index 3e6ecbf1fd..2212880177 100644
--- a/modules/imgproc/perf/perf_resize.cpp
+++ b/modules/imgproc/perf/perf_resize.cpp
@@ -25,7 +25,7 @@ PERF_TEST_P(MatInfo_Size_Size, resizeUpLinear,
     cvtest::fillGradient(src);
     declare.in(src).out(dst);
 
-    TEST_CYCLE() resize(src, dst, to);
+    TEST_CYCLE_MULTIRUN(10) resize(src, dst, to);
 
 #ifdef ANDROID
     SANITY_CHECK(dst, 5);
@@ -52,7 +52,7 @@ PERF_TEST_P(MatInfo_Size_Size, resizeDownLinear,
     cvtest::fillGradient(src);
     declare.in(src).out(dst);
 
-    TEST_CYCLE() resize(src, dst, to);
+    TEST_CYCLE_MULTIRUN(10) resize(src, dst, to);
 
 #ifdef ANDROID
     SANITY_CHECK(dst, 5);
@@ -85,7 +85,8 @@ PERF_TEST_P(MatInfo_Size_Scale, ResizeAreaFast,
 
     declare.in(src, WARMUP_RNG).out(dst);
 
-    TEST_CYCLE() resize(src, dst, dst.size(), 0, 0, INTER_AREA);
+    int runs = 15;
+    TEST_CYCLE_MULTIRUN(runs) resize(src, dst, dst.size(), 0, 0, INTER_AREA);
 
     //difference equal to 1 is allowed because of different possible rounding modes: round-to-nearest vs bankers' rounding
     SANITY_CHECK(dst, 1);
diff --git a/modules/imgproc/perf/perf_sepfilters.cpp b/modules/imgproc/perf/perf_sepfilters.cpp
index 9a5bd4e711..6147b30eba 100644
--- a/modules/imgproc/perf/perf_sepfilters.cpp
+++ b/modules/imgproc/perf/perf_sepfilters.cpp
@@ -34,7 +34,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3, sobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0), make_tuple(1, 1), make_tuple(0, 2), make_tuple(2, 0), make_tuple(2, 2)),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
             )
           )
 {
@@ -59,7 +59,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3ROI, sobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0), make_tuple(1, 1), make_tuple(0, 2), make_tuple(2, 0), make_tuple(2, 2)),
-                testing::ValuesIn(BorderType3x3ROI::all())
+                BorderType3x3ROI::all()
             )
           )
 {
@@ -87,7 +87,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border5x5, sobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0), make_tuple(1, 1), make_tuple(0, 2), make_tuple(2, 0)),
-                testing::ValuesIn(BorderType::all())
+                BorderType::all()
             )
           )
 {
@@ -112,7 +112,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border5x5ROI, sobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0), make_tuple(1, 1), make_tuple(0, 2), make_tuple(2, 0)),
-                testing::ValuesIn(BorderTypeROI::all())
+                BorderTypeROI::all()
             )
           )
 {
@@ -142,7 +142,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3, scharrFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0)),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
             )
           )
 {
@@ -167,7 +167,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3ROI, scharrFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0)),
-                testing::ValuesIn(BorderType3x3ROI::all())
+                BorderType3x3ROI::all()
             )
           )
 {
@@ -195,7 +195,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3, scharrViaSobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0)),
-                testing::ValuesIn(BorderType3x3::all())
+                BorderType3x3::all()
             )
           )
 {
@@ -220,7 +220,7 @@ PERF_TEST_P(Size_MatType_dx_dy_Border3x3ROI, scharrViaSobelFilter,
                 testing::Values(FILTER_SRC_SIZES),
                 testing::Values(CV_16S, CV_32F),
                 testing::Values(make_tuple(0, 1), make_tuple(1, 0)),
-                testing::ValuesIn(BorderType3x3ROI::all())
+                BorderType3x3ROI::all()
             )
           )
 {
diff --git a/modules/imgproc/perf/perf_threshold.cpp b/modules/imgproc/perf/perf_threshold.cpp
index 8727b6339e..9ccafd6b54 100644
--- a/modules/imgproc/perf/perf_threshold.cpp
+++ b/modules/imgproc/perf/perf_threshold.cpp
@@ -15,7 +15,7 @@ PERF_TEST_P(Size_MatType_ThreshType, threshold,
             testing::Combine(
                 testing::Values(TYPICAL_MAT_SIZES),
                 testing::Values(CV_8UC1, CV_16SC1),
-                testing::ValuesIn(ThreshType::all())
+                ThreshType::all()
                 )
             )
 {
@@ -32,7 +32,7 @@ PERF_TEST_P(Size_MatType_ThreshType, threshold,
 
     declare.in(src, WARMUP_RNG).out(dst);
 
-    int runs = (sz.width <= 640) ? 8 : 1;
+    int runs = (sz.width <= 640) ? 40 : 1;
     TEST_CYCLE_MULTIRUN(runs) threshold(src, dst, thresh, maxval, threshType);
 
     SANITY_CHECK(dst);
@@ -51,7 +51,8 @@ PERF_TEST_P(Size_Only, threshold_otsu, testing::Values(TYPICAL_MAT_SIZES))
 
     declare.in(src, WARMUP_RNG).out(dst);
 
-    TEST_CYCLE() threshold(src, dst, 0, maxval, THRESH_BINARY|THRESH_OTSU);
+    int runs = 15;
+    TEST_CYCLE_MULTIRUN(runs) threshold(src, dst, 0, maxval, THRESH_BINARY|THRESH_OTSU);
 
     SANITY_CHECK(dst);
 }
@@ -65,8 +66,8 @@ typedef perf::TestBaseWithParam<Size_AdaptThreshType_AdaptThreshMethod_BlockSize
 PERF_TEST_P(Size_AdaptThreshType_AdaptThreshMethod_BlockSize, adaptiveThreshold,
             testing::Combine(
                 testing::Values(TYPICAL_MAT_SIZES),
-                testing::ValuesIn(AdaptThreshType::all()),
-                testing::ValuesIn(AdaptThreshMethod::all()),
+                AdaptThreshType::all(),
+                AdaptThreshMethod::all(),
                 testing::Values(3, 5)
                 )
             )
diff --git a/modules/imgproc/perf/perf_warp.cpp b/modules/imgproc/perf/perf_warp.cpp
index 3d90b3e902..146528b6bb 100644
--- a/modules/imgproc/perf/perf_warp.cpp
+++ b/modules/imgproc/perf/perf_warp.cpp
@@ -23,8 +23,8 @@ void update_map(const Mat& src, Mat& map_x, Mat& map_y, const int remapMode );
 PERF_TEST_P( TestWarpAffine, WarpAffine,
              Combine(
                 Values( szVGA, sz720p, sz1080p ),
-                ValuesIn( InterType::all() ),
-                ValuesIn( BorderMode::all() )
+                InterType::all(),
+                BorderMode::all()
              )
 )
 {
@@ -53,8 +53,8 @@ PERF_TEST_P( TestWarpAffine, WarpAffine,
 PERF_TEST_P( TestWarpPerspective, WarpPerspective,
              Combine(
                 Values( szVGA, sz720p, sz1080p ),
-                ValuesIn( InterType::all() ),
-                ValuesIn( BorderMode::all() )
+                InterType::all(),
+                BorderMode::all()
              )
 )
 {
@@ -91,8 +91,8 @@ PERF_TEST_P( TestWarpPerspective, WarpPerspective,
 PERF_TEST_P( TestWarpPerspectiveNear_t, WarpPerspectiveNear,
              Combine(
                  Values( Size(640,480), Size(1920,1080), Size(2592,1944) ),
-                 ValuesIn( InterType::all() ),
-                 ValuesIn( BorderMode::all() ),
+                 InterType::all(),
+                 BorderMode::all(),
                  Values( CV_8UC1, CV_8UC4 )
                  )
              )
@@ -138,9 +138,9 @@ PERF_TEST_P( TestRemap, remap,
              Combine(
                  Values( TYPICAL_MAT_TYPES ),
                  Values( szVGA, sz720p, sz1080p ),
-                 ValuesIn( InterType::all() ),
-                 ValuesIn( BorderMode::all() ),
-                 ValuesIn( RemapMode::all() )
+                 InterType::all(),
+                 BorderMode::all(),
+                 RemapMode::all()
                  )
              )
 {
@@ -221,4 +221,3 @@ PERF_TEST(Transform, getPerspectiveTransform)
 
     SANITY_CHECK(transformCoefficient, 1e-5);
 }
-
diff --git a/modules/imgproc/src/canny.cpp b/modules/imgproc/src/canny.cpp
index 4dae01343d..fb6afaf2b3 100644
--- a/modules/imgproc/src/canny.cpp
+++ b/modules/imgproc/src/canny.cpp
@@ -41,6 +41,50 @@
 
 #include "precomp.hpp"
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+#define USE_IPP_CANNY 1
+#else
+#undef USE_IPP_CANNY
+#endif
+
+#ifdef USE_IPP_CANNY
+namespace cv
+{
+static bool ippCanny(const Mat& _src, Mat& _dst, float low,  float high)
+{
+    int size = 0, size1 = 0;
+    IppiSize roi = { _src.cols, _src.rows };
+
+    ippiFilterSobelNegVertGetBufferSize_8u16s_C1R(roi, ippMskSize3x3, &size);
+    ippiFilterSobelHorizGetBufferSize_8u16s_C1R(roi, ippMskSize3x3, &size1);
+    size = std::max(size, size1);
+    ippiCannyGetSize(roi, &size1);
+    size = std::max(size, size1);
+
+    AutoBuffer<uchar> buf(size + 64);
+    uchar* buffer = alignPtr((uchar*)buf, 32);
+
+    Mat _dx(_src.rows, _src.cols, CV_16S);
+    if( ippiFilterSobelNegVertBorder_8u16s_C1R(_src.data, (int)_src.step,
+                    _dx.ptr<short>(), (int)_dx.step, roi,
+                    ippMskSize3x3, ippBorderRepl, 0, buffer) < 0 )
+        return false;
+
+    Mat _dy(_src.rows, _src.cols, CV_16S);
+    if( ippiFilterSobelHorizBorder_8u16s_C1R(_src.data, (int)_src.step,
+                    _dy.ptr<short>(), (int)_dy.step, roi,
+                    ippMskSize3x3, ippBorderRepl, 0, buffer) < 0 )
+        return false;
+
+    if( ippiCanny_16s8u_C1R(_dx.ptr<short>(), (int)_dx.step,
+                            _dy.ptr<short>(), (int)_dy.step,
+                            _dst.data, (int)_dst.step, roi, low, high, buffer) < 0 )
+        return false;
+    return true;
+}
+}
+#endif
+
 void cv::Canny( InputArray _src, OutputArray _dst,
                 double low_thresh, double high_thresh,
                 int aperture_size, bool L2gradient )
@@ -61,20 +105,26 @@ void cv::Canny( InputArray _src, OutputArray _dst,
     if ((aperture_size & 1) == 0 || (aperture_size != -1 && (aperture_size < 3 || aperture_size > 7)))
         CV_Error(CV_StsBadFlag, "");
 
+    if (low_thresh > high_thresh)
+        std::swap(low_thresh, high_thresh);
+
 #ifdef HAVE_TEGRA_OPTIMIZATION
     if (tegra::canny(src, dst, low_thresh, high_thresh, aperture_size, L2gradient))
         return;
 #endif
 
+#ifdef USE_IPP_CANNY
+    if( aperture_size == 3 && !L2gradient &&
+        ippCanny(src, dst, low_thresh, high_thresh) >= 0 )
+        return;
+#endif
+
     const int cn = src.channels();
-    cv::Mat dx(src.rows, src.cols, CV_16SC(cn));
-    cv::Mat dy(src.rows, src.cols, CV_16SC(cn));
+    Mat dx(src.rows, src.cols, CV_16SC(cn));
+    Mat dy(src.rows, src.cols, CV_16SC(cn));
 
-    cv::Sobel(src, dx, CV_16S, 1, 0, aperture_size, 1, 0, cv::BORDER_REPLICATE);
-    cv::Sobel(src, dy, CV_16S, 0, 1, aperture_size, 1, 0, cv::BORDER_REPLICATE);
-
-    if (low_thresh > high_thresh)
-        std::swap(low_thresh, high_thresh);
+    Sobel(src, dx, CV_16S, 1, 0, aperture_size, 1, 0, cv::BORDER_REPLICATE);
+    Sobel(src, dy, CV_16S, 0, 1, aperture_size, 1, 0, cv::BORDER_REPLICATE);
 
     if (L2gradient)
     {
@@ -88,7 +138,7 @@ void cv::Canny( InputArray _src, OutputArray _dst,
     int high = cvFloor(high_thresh);
 
     ptrdiff_t mapstep = src.cols + 2;
-    cv::AutoBuffer<uchar> buffer((src.cols+2)*(src.rows+2) + cn * mapstep * 3 * sizeof(int));
+    AutoBuffer<uchar> buffer((src.cols+2)*(src.rows+2) + cn * mapstep * 3 * sizeof(int));
 
     int* mag_buf[3];
     mag_buf[0] = (int*)(uchar*)buffer;
diff --git a/modules/imgproc/src/clahe.cpp b/modules/imgproc/src/clahe.cpp
new file mode 100644
index 0000000000..89fb62bd01
--- /dev/null
+++ b/modules/imgproc/src/clahe.cpp
@@ -0,0 +1,334 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the copyright holders or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+// ----------------------------------------------------------------------
+// CLAHE
+
+namespace
+{
+    class CLAHE_CalcLut_Body : public cv::ParallelLoopBody
+    {
+    public:
+        CLAHE_CalcLut_Body(const cv::Mat& src, cv::Mat& lut, cv::Size tileSize, int tilesX, int tilesY, int clipLimit, float lutScale) :
+            src_(src), lut_(lut), tileSize_(tileSize), tilesX_(tilesX), tilesY_(tilesY), clipLimit_(clipLimit), lutScale_(lutScale)
+        {
+        }
+
+        void operator ()(const cv::Range& range) const;
+
+    private:
+        cv::Mat src_;
+        mutable cv::Mat lut_;
+
+        cv::Size tileSize_;
+        int tilesX_;
+        int tilesY_;
+        int clipLimit_;
+        float lutScale_;
+    };
+
+    void CLAHE_CalcLut_Body::operator ()(const cv::Range& range) const
+    {
+        const int histSize = 256;
+
+        uchar* tileLut = lut_.ptr(range.start);
+        const size_t lut_step = lut_.step;
+
+        for (int k = range.start; k < range.end; ++k, tileLut += lut_step)
+        {
+            const int ty = k / tilesX_;
+            const int tx = k % tilesX_;
+
+            // retrieve tile submatrix
+
+            cv::Rect tileROI;
+            tileROI.x = tx * tileSize_.width;
+            tileROI.y = ty * tileSize_.height;
+            tileROI.width = tileSize_.width;
+            tileROI.height = tileSize_.height;
+
+            const cv::Mat tile = src_(tileROI);
+
+            // calc histogram
+
+            int tileHist[histSize] = {0, };
+
+            int height = tileROI.height;
+            const size_t sstep = tile.step;
+            for (const uchar* ptr = tile.ptr<uchar>(0); height--; ptr += sstep)
+            {
+                int x = 0;
+                for (; x <= tileROI.width - 4; x += 4)
+                {
+                    int t0 = ptr[x], t1 = ptr[x+1];
+                    tileHist[t0]++; tileHist[t1]++;
+                    t0 = ptr[x+2]; t1 = ptr[x+3];
+                    tileHist[t0]++; tileHist[t1]++;
+                }
+
+                for (; x < tileROI.width; ++x)
+                    tileHist[ptr[x]]++;
+            }
+
+            // clip histogram
+
+            if (clipLimit_ > 0)
+            {
+                // how many pixels were clipped
+                int clipped = 0;
+                for (int i = 0; i < histSize; ++i)
+                {
+                    if (tileHist[i] > clipLimit_)
+                    {
+                        clipped += tileHist[i] - clipLimit_;
+                        tileHist[i] = clipLimit_;
+                    }
+                }
+
+                // redistribute clipped pixels
+                int redistBatch = clipped / histSize;
+                int residual = clipped - redistBatch * histSize;
+
+                for (int i = 0; i < histSize; ++i)
+                    tileHist[i] += redistBatch;
+
+                for (int i = 0; i < residual; ++i)
+                    tileHist[i]++;
+            }
+
+            // calc Lut
+
+            int sum = 0;
+            for (int i = 0; i < histSize; ++i)
+            {
+                sum += tileHist[i];
+                tileLut[i] = cv::saturate_cast<uchar>(sum * lutScale_);
+            }
+        }
+    }
+
+    class CLAHE_Interpolation_Body : public cv::ParallelLoopBody
+    {
+    public:
+        CLAHE_Interpolation_Body(const cv::Mat& src, cv::Mat& dst, const cv::Mat& lut, cv::Size tileSize, int tilesX, int tilesY) :
+            src_(src), dst_(dst), lut_(lut), tileSize_(tileSize), tilesX_(tilesX), tilesY_(tilesY)
+        {
+        }
+
+        void operator ()(const cv::Range& range) const;
+
+    private:
+        cv::Mat src_;
+        mutable cv::Mat dst_;
+        cv::Mat lut_;
+
+        cv::Size tileSize_;
+        int tilesX_;
+        int tilesY_;
+    };
+
+    void CLAHE_Interpolation_Body::operator ()(const cv::Range& range) const
+    {
+        const size_t lut_step = lut_.step;
+
+        for (int y = range.start; y < range.end; ++y)
+        {
+            const uchar* srcRow = src_.ptr<uchar>(y);
+            uchar* dstRow = dst_.ptr<uchar>(y);
+
+            const float tyf = (static_cast<float>(y) / tileSize_.height) - 0.5f;
+
+            int ty1 = cvFloor(tyf);
+            int ty2 = ty1 + 1;
+
+            const float ya = tyf - ty1;
+
+            ty1 = std::max(ty1, 0);
+            ty2 = std::min(ty2, tilesY_ - 1);
+
+            const uchar* lutPlane1 = lut_.ptr(ty1 * tilesX_);
+            const uchar* lutPlane2 = lut_.ptr(ty2 * tilesX_);
+
+            for (int x = 0; x < src_.cols; ++x)
+            {
+                const float txf = (static_cast<float>(x) / tileSize_.width) - 0.5f;
+
+                int tx1 = cvFloor(txf);
+                int tx2 = tx1 + 1;
+
+                const float xa = txf - tx1;
+
+                tx1 = std::max(tx1, 0);
+                tx2 = std::min(tx2, tilesX_ - 1);
+
+                const int srcVal = srcRow[x];
+
+                const size_t ind1 = tx1 * lut_step + srcVal;
+                const size_t ind2 = tx2 * lut_step + srcVal;
+
+                float res = 0;
+
+                res += lutPlane1[ind1] * ((1.0f - xa) * (1.0f - ya));
+                res += lutPlane1[ind2] * ((xa) * (1.0f - ya));
+                res += lutPlane2[ind1] * ((1.0f - xa) * (ya));
+                res += lutPlane2[ind2] * ((xa) * (ya));
+
+                dstRow[x] = cv::saturate_cast<uchar>(res);
+            }
+        }
+    }
+
+    class CLAHE_Impl : public cv::CLAHE
+    {
+    public:
+        CLAHE_Impl(double clipLimit = 40.0, int tilesX = 8, int tilesY = 8);
+
+        cv::AlgorithmInfo* info() const;
+
+        void apply(cv::InputArray src, cv::OutputArray dst);
+
+        void setClipLimit(double clipLimit);
+        double getClipLimit() const;
+
+        void setTilesGridSize(cv::Size tileGridSize);
+        cv::Size getTilesGridSize() const;
+
+        void collectGarbage();
+
+    private:
+        double clipLimit_;
+        int tilesX_;
+        int tilesY_;
+
+        cv::Mat srcExt_;
+        cv::Mat lut_;
+    };
+
+    CLAHE_Impl::CLAHE_Impl(double clipLimit, int tilesX, int tilesY) :
+        clipLimit_(clipLimit), tilesX_(tilesX), tilesY_(tilesY)
+    {
+    }
+
+    CV_INIT_ALGORITHM(CLAHE_Impl, "CLAHE",
+        obj.info()->addParam(obj, "clipLimit", obj.clipLimit_);
+        obj.info()->addParam(obj, "tilesX", obj.tilesX_);
+        obj.info()->addParam(obj, "tilesY", obj.tilesY_))
+
+    void CLAHE_Impl::apply(cv::InputArray _src, cv::OutputArray _dst)
+    {
+        cv::Mat src = _src.getMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+
+        _dst.create( src.size(), src.type() );
+        cv::Mat dst = _dst.getMat();
+
+        const int histSize = 256;
+
+        lut_.create(tilesX_ * tilesY_, histSize, CV_8UC1);
+
+        cv::Size tileSize;
+        cv::Mat srcForLut;
+
+        if (src.cols % tilesX_ == 0 && src.rows % tilesY_ == 0)
+        {
+            tileSize = cv::Size(src.cols / tilesX_, src.rows / tilesY_);
+            srcForLut = src;
+        }
+        else
+        {
+            cv::copyMakeBorder(src, srcExt_, 0, tilesY_ - (src.rows % tilesY_), 0, tilesX_ - (src.cols % tilesX_), cv::BORDER_REFLECT_101);
+
+            tileSize = cv::Size(srcExt_.cols / tilesX_, srcExt_.rows / tilesY_);
+            srcForLut = srcExt_;
+        }
+
+        const int tileSizeTotal = tileSize.area();
+        const float lutScale = static_cast<float>(histSize - 1) / tileSizeTotal;
+
+        int clipLimit = 0;
+        if (clipLimit_ > 0.0)
+        {
+            clipLimit = static_cast<int>(clipLimit_ * tileSizeTotal / histSize);
+            clipLimit = std::max(clipLimit, 1);
+        }
+
+        CLAHE_CalcLut_Body calcLutBody(srcForLut, lut_, tileSize, tilesX_, tilesY_, clipLimit, lutScale);
+        cv::parallel_for_(cv::Range(0, tilesX_ * tilesY_), calcLutBody);
+
+        CLAHE_Interpolation_Body interpolationBody(src, dst, lut_, tileSize, tilesX_, tilesY_);
+        cv::parallel_for_(cv::Range(0, src.rows), interpolationBody);
+    }
+
+    void CLAHE_Impl::setClipLimit(double clipLimit)
+    {
+        clipLimit_ = clipLimit;
+    }
+
+    double CLAHE_Impl::getClipLimit() const
+    {
+        return clipLimit_;
+    }
+
+    void CLAHE_Impl::setTilesGridSize(cv::Size tileGridSize)
+    {
+        tilesX_ = tileGridSize.width;
+        tilesY_ = tileGridSize.height;
+    }
+
+    cv::Size CLAHE_Impl::getTilesGridSize() const
+    {
+        return cv::Size(tilesX_, tilesY_);
+    }
+
+    void CLAHE_Impl::collectGarbage()
+    {
+        srcExt_.release();
+        lut_.release();
+    }
+}
+
+cv::Ptr<cv::CLAHE> cv::createCLAHE(double clipLimit, cv::Size tileGridSize)
+{
+    return makePtr<CLAHE_Impl>(clipLimit, tileGridSize.width, tileGridSize.height);
+}
diff --git a/modules/imgproc/src/color.cpp b/modules/imgproc/src/color.cpp
index 4f1ce82034..14703cce28 100644
--- a/modules/imgproc/src/color.cpp
+++ b/modules/imgproc/src/color.cpp
@@ -92,6 +92,15 @@
 #include "precomp.hpp"
 #include <limits>
 
+#define  CV_DESCALE(x,n)     (((x) + (1 << ((n)-1))) >> (n))
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+#define MAX_IPP8u   255
+#define MAX_IPP16u  65535
+#define MAX_IPP32f  1.0
+static IppStatus sts = ippInit();
+#endif
+
 namespace cv
 {
 
@@ -189,6 +198,301 @@ void CvtColorLoop(const Mat& src, Mat& dst, const Cvt& cvt)
     parallel_for_(Range(0, src.rows), CvtColorLoop_Invoker<Cvt>(src, dst, cvt), src.total()/(double)(1<<16) );
 }
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+typedef IppStatus (CV_STDCALL* ippiReorderFunc)(const void *, int, void *, int, IppiSize, const int *);
+typedef IppStatus (CV_STDCALL* ippiGeneralFunc)(const void *, int, void *, int, IppiSize);
+typedef IppStatus (CV_STDCALL* ippiColor2GrayFunc)(const void *, int, void *, int, IppiSize, const Ipp32f *);
+
+template <typename Cvt>
+class CvtColorIPPLoop_Invoker : public ParallelLoopBody
+{
+public:
+
+    CvtColorIPPLoop_Invoker(const Mat& _src, Mat& _dst, const Cvt& _cvt, bool *_ok) :
+        ParallelLoopBody(), src(_src), dst(_dst), cvt(_cvt), ok(_ok)
+    {
+        *ok = true;
+    }
+
+    virtual void operator()(const Range& range) const
+    {
+        const void *yS = src.ptr<uchar>(range.start);
+        void *yD = dst.ptr<uchar>(range.start);
+        if( cvt(yS, (int)src.step[0], yD, (int)dst.step[0], src.cols, range.end - range.start) < 0 )
+            *ok = false;
+    }
+
+private:
+    const Mat& src;
+    Mat& dst;
+    const Cvt& cvt;
+    bool *ok;
+
+    const CvtColorIPPLoop_Invoker& operator= (const CvtColorIPPLoop_Invoker&);
+};
+
+template <typename Cvt>
+bool CvtColorIPPLoop(const Mat& src, Mat& dst, const Cvt& cvt)
+{
+    bool ok;
+    parallel_for_(Range(0, src.rows), CvtColorIPPLoop_Invoker<Cvt>(src, dst, cvt, &ok), src.total()/(double)(1<<16) );
+    return ok;
+}
+
+template <typename Cvt>
+bool CvtColorIPPLoopCopy(Mat& src, Mat& dst, const Cvt& cvt)
+{
+    Mat temp;
+    Mat &source = src;
+    if( src.data == dst.data )
+    {
+        src.copyTo(temp);
+        source = temp;
+    }
+    bool ok;
+    parallel_for_(Range(0, source.rows), CvtColorIPPLoop_Invoker<Cvt>(source, dst, cvt, &ok), source.total()/(double)(1<<16) );
+    return ok;
+}
+
+static IppStatus CV_STDCALL ippiSwapChannels_8u_C3C4Rf(const Ipp8u* pSrc, int srcStep, Ipp8u* pDst, int dstStep,
+         IppiSize roiSize, const int *dstOrder)
+{
+    return ippiSwapChannels_8u_C3C4R(pSrc, srcStep, pDst, dstStep, roiSize, dstOrder, MAX_IPP8u);
+}
+
+static IppStatus CV_STDCALL ippiSwapChannels_16u_C3C4Rf(const Ipp16u* pSrc, int srcStep, Ipp16u* pDst, int dstStep,
+         IppiSize roiSize, const int *dstOrder)
+{
+    return ippiSwapChannels_16u_C3C4R(pSrc, srcStep, pDst, dstStep, roiSize, dstOrder, MAX_IPP16u);
+}
+
+static IppStatus CV_STDCALL ippiSwapChannels_32f_C3C4Rf(const Ipp32f* pSrc, int srcStep, Ipp32f* pDst, int dstStep,
+         IppiSize roiSize, const int *dstOrder)
+{
+    return ippiSwapChannels_32f_C3C4R(pSrc, srcStep, pDst, dstStep, roiSize, dstOrder, MAX_IPP32f);
+}
+
+static ippiReorderFunc ippiSwapChannelsC3C4RTab[] =
+{
+    (ippiReorderFunc)ippiSwapChannels_8u_C3C4Rf, 0, (ippiReorderFunc)ippiSwapChannels_16u_C3C4Rf, 0,
+    0, (ippiReorderFunc)ippiSwapChannels_32f_C3C4Rf, 0, 0
+};
+
+static ippiGeneralFunc ippiCopyAC4C3RTab[] =
+{
+    (ippiGeneralFunc)ippiCopy_8u_AC4C3R, 0, (ippiGeneralFunc)ippiCopy_16u_AC4C3R, 0,
+    0, (ippiGeneralFunc)ippiCopy_32f_AC4C3R, 0, 0
+};
+
+static ippiReorderFunc ippiSwapChannelsC4C3RTab[] =
+{
+    (ippiReorderFunc)ippiSwapChannels_8u_C4C3R, 0, (ippiReorderFunc)ippiSwapChannels_16u_C4C3R, 0,
+    0, (ippiReorderFunc)ippiSwapChannels_32f_C4C3R, 0, 0
+};
+
+static ippiReorderFunc ippiSwapChannelsC3RTab[] =
+{
+    (ippiReorderFunc)ippiSwapChannels_8u_C3R, 0, (ippiReorderFunc)ippiSwapChannels_16u_C3R, 0,
+    0, (ippiReorderFunc)ippiSwapChannels_32f_C3R, 0, 0
+};
+
+static ippiReorderFunc ippiSwapChannelsC4RTab[] =
+{
+    (ippiReorderFunc)ippiSwapChannels_8u_AC4R, 0, (ippiReorderFunc)ippiSwapChannels_16u_AC4R, 0,
+    0, (ippiReorderFunc)ippiSwapChannels_32f_AC4R, 0, 0
+};
+
+static ippiColor2GrayFunc ippiColor2GrayC3Tab[] =
+{
+    (ippiColor2GrayFunc)ippiColorToGray_8u_C3C1R, 0, (ippiColor2GrayFunc)ippiColorToGray_16u_C3C1R, 0,
+    0, (ippiColor2GrayFunc)ippiColorToGray_32f_C3C1R, 0, 0
+};
+
+static ippiColor2GrayFunc ippiColor2GrayC4Tab[] =
+{
+    (ippiColor2GrayFunc)ippiColorToGray_8u_AC4C1R, 0, (ippiColor2GrayFunc)ippiColorToGray_16u_AC4C1R, 0,
+    0, (ippiColor2GrayFunc)ippiColorToGray_32f_AC4C1R, 0, 0
+};
+
+static ippiGeneralFunc ippiRGB2GrayC3Tab[] =
+{
+    (ippiGeneralFunc)ippiRGBToGray_8u_C3C1R, 0, (ippiGeneralFunc)ippiRGBToGray_16u_C3C1R, 0,
+    0, (ippiGeneralFunc)ippiRGBToGray_32f_C3C1R, 0, 0
+};
+
+static ippiGeneralFunc ippiRGB2GrayC4Tab[] =
+{
+    (ippiGeneralFunc)ippiRGBToGray_8u_AC4C1R, 0, (ippiGeneralFunc)ippiRGBToGray_16u_AC4C1R, 0,
+    0, (ippiGeneralFunc)ippiRGBToGray_32f_AC4C1R, 0, 0
+};
+
+static ippiGeneralFunc ippiCopyP3C3RTab[] =
+{
+    (ippiGeneralFunc)ippiCopy_8u_P3C3R, 0, (ippiGeneralFunc)ippiCopy_16u_P3C3R, 0,
+    0, (ippiGeneralFunc)ippiCopy_32f_P3C3R, 0, 0
+};
+
+static ippiGeneralFunc ippiRGB2XYZTab[] =
+{
+    (ippiGeneralFunc)ippiRGBToXYZ_8u_C3R, 0, (ippiGeneralFunc)ippiRGBToXYZ_16u_C3R, 0,
+    0, (ippiGeneralFunc)ippiRGBToXYZ_32f_C3R, 0, 0
+};
+
+static ippiGeneralFunc ippiXYZ2RGBTab[] =
+{
+    (ippiGeneralFunc)ippiXYZToRGB_8u_C3R, 0, (ippiGeneralFunc)ippiXYZToRGB_16u_C3R, 0,
+    0, (ippiGeneralFunc)ippiXYZToRGB_32f_C3R, 0, 0
+};
+
+static ippiGeneralFunc ippiRGB2HSVTab[] =
+{
+    (ippiGeneralFunc)ippiRGBToHSV_8u_C3R, 0, (ippiGeneralFunc)ippiRGBToHSV_16u_C3R, 0,
+    0, 0, 0, 0
+};
+
+static ippiGeneralFunc ippiHSV2RGBTab[] =
+{
+    (ippiGeneralFunc)ippiHSVToRGB_8u_C3R, 0, (ippiGeneralFunc)ippiHSVToRGB_16u_C3R, 0,
+    0, 0, 0, 0
+};
+
+static ippiGeneralFunc ippiRGB2HLSTab[] =
+{
+    (ippiGeneralFunc)ippiRGBToHLS_8u_C3R, 0, (ippiGeneralFunc)ippiRGBToHLS_16u_C3R, 0,
+    0, (ippiGeneralFunc)ippiRGBToHLS_32f_C3R, 0, 0
+};
+
+static ippiGeneralFunc ippiHLS2RGBTab[] =
+{
+    (ippiGeneralFunc)ippiHLSToRGB_8u_C3R, 0, (ippiGeneralFunc)ippiHLSToRGB_16u_C3R, 0,
+    0, (ippiGeneralFunc)ippiHLSToRGB_32f_C3R, 0, 0
+};
+
+struct IPPGeneralFunctor
+{
+    IPPGeneralFunctor(ippiGeneralFunc _func) : func(_func){}
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        return func(src, srcStep, dst, dstStep, ippiSize(cols, rows)) >= 0;
+    }
+private:
+    ippiGeneralFunc func;
+};
+
+struct IPPReorderFunctor
+{
+    IPPReorderFunctor(ippiReorderFunc _func, int _order0, int _order1, int _order2) : func(_func)
+    {
+        order[0] = _order0;
+        order[1] = _order1;
+        order[2] = _order2;
+        order[3] = 3;
+    }
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        return func(src, srcStep, dst, dstStep, ippiSize(cols, rows), order) >= 0;
+    }
+private:
+    ippiReorderFunc func;
+    int order[4];
+};
+
+struct IPPColor2GrayFunctor
+{
+    IPPColor2GrayFunctor(ippiColor2GrayFunc _func) : func(_func)
+    {
+        coeffs[0] = 0.114f;
+        coeffs[1] = 0.587f;
+        coeffs[2] = 0.299f;
+    }
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        return func(src, srcStep, dst, dstStep, ippiSize(cols, rows), coeffs) >= 0;
+    }
+private:
+    ippiColor2GrayFunc func;
+    Ipp32f coeffs[3];
+};
+
+struct IPPGray2BGRFunctor
+{
+    IPPGray2BGRFunctor(ippiGeneralFunc _func) : func(_func){}
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        const void* srcarray[3] = { src, src, src };
+        return func(srcarray, srcStep, dst, dstStep, ippiSize(cols, rows)) >= 0;
+    }
+private:
+    ippiGeneralFunc func;
+};
+
+struct IPPGray2BGRAFunctor
+{
+    IPPGray2BGRAFunctor(ippiGeneralFunc _func1, ippiReorderFunc _func2, int _depth) : func1(_func1), func2(_func2), depth(_depth){}
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        const void* srcarray[3] = { src, src, src };
+        Mat temp(rows, cols, CV_MAKETYPE(depth, 3));
+        if(func1(srcarray, srcStep, temp.data, (int)temp.step[0], ippiSize(cols, rows)) < 0)
+            return false;
+        int order[4] = {0, 1, 2, 3};
+        return func2(temp.data, (int)temp.step[0], dst, dstStep, ippiSize(cols, rows), order) >= 0;
+    }
+private:
+    ippiGeneralFunc func1;
+    ippiReorderFunc func2;
+    int depth;
+};
+
+struct IPPReorderGeneralFunctor
+{
+    IPPReorderGeneralFunctor(ippiReorderFunc _func1, ippiGeneralFunc _func2, int _order0, int _order1, int _order2, int _depth) : func1(_func1), func2(_func2), depth(_depth)
+    {
+        order[0] = _order0;
+        order[1] = _order1;
+        order[2] = _order2;
+        order[3] = 3;
+    }
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        Mat temp;
+        temp.create(rows, cols, CV_MAKETYPE(depth, 3));
+        if(func1(src, srcStep, temp.data, (int)temp.step[0], ippiSize(cols, rows), order) < 0)
+            return false;
+        return func2(temp.data, (int)temp.step[0], dst, dstStep, ippiSize(cols, rows)) >= 0;
+    }
+private:
+    ippiReorderFunc func1;
+    ippiGeneralFunc func2;
+    int order[4];
+    int depth;
+};
+
+struct IPPGeneralReorderFunctor
+{
+    IPPGeneralReorderFunctor(ippiGeneralFunc _func1, ippiReorderFunc _func2, int _order0, int _order1, int _order2, int _depth) : func1(_func1), func2(_func2), depth(_depth)
+    {
+        order[0] = _order0;
+        order[1] = _order1;
+        order[2] = _order2;
+        order[3] = 3;
+    }
+    bool operator()(const void *src, int srcStep, void *dst, int dstStep, int cols, int rows) const
+    {
+        Mat temp;
+        temp.create(rows, cols, CV_MAKETYPE(depth, 3));
+        if(func1(src, srcStep, temp.data, (int)temp.step[0], ippiSize(cols, rows)) < 0)
+            return false;
+        return func2(temp.data, (int)temp.step[0], dst, dstStep, ippiSize(cols, rows), order) >= 0;
+    }
+private:
+    ippiGeneralFunc func1;
+    ippiReorderFunc func2;
+    int order[4];
+    int depth;
+};
+#endif
+
 ////////////////// Various 3/4-channel to 3/4-channel RGB transformations /////////////////
 
 template<typename _Tp> struct RGB2RGB
@@ -1813,7 +2117,7 @@ const int ITUR_BT_601_CGV = -385875;
 const int ITUR_BT_601_CBV = -74448;
 
 template<int bIdx, int uIdx>
-struct YUV420sp2RGB888Invoker
+struct YUV420sp2RGB888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* my1, *muv;
@@ -1822,10 +2126,10 @@ struct YUV420sp2RGB888Invoker
     YUV420sp2RGB888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _uv)
         : dst(_dst), my1(_y1), muv(_uv), width(_dst->cols), stride(_stride) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int rangeBegin = range.begin() * 2;
-        int rangeEnd = range.end() * 2;
+        int rangeBegin = range.start * 2;
+        int rangeEnd = range.end * 2;
 
         //R = 1.164(Y - 16) + 1.596(V - 128)
         //G = 1.164(Y - 16) - 0.813(V - 128) - 0.391(U - 128)
@@ -1882,7 +2186,7 @@ struct YUV420sp2RGB888Invoker
 };
 
 template<int bIdx, int uIdx>
-struct YUV420sp2RGBA8888Invoker
+struct YUV420sp2RGBA8888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* my1, *muv;
@@ -1891,10 +2195,10 @@ struct YUV420sp2RGBA8888Invoker
     YUV420sp2RGBA8888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _uv)
         : dst(_dst), my1(_y1), muv(_uv), width(_dst->cols), stride(_stride) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int rangeBegin = range.begin() * 2;
-        int rangeEnd = range.end() * 2;
+        int rangeBegin = range.start * 2;
+        int rangeEnd = range.end * 2;
 
         //R = 1.164(Y - 16) + 1.596(V - 128)
         //G = 1.164(Y - 16) - 0.813(V - 128) - 0.391(U - 128)
@@ -1955,7 +2259,7 @@ struct YUV420sp2RGBA8888Invoker
 };
 
 template<int bIdx>
-struct YUV420p2RGB888Invoker
+struct YUV420p2RGB888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* my1, *mu, *mv;
@@ -1965,19 +2269,19 @@ struct YUV420p2RGB888Invoker
     YUV420p2RGB888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int _ustepIdx, int _vstepIdx)
         : dst(_dst), my1(_y1), mu(_u), mv(_v), width(_dst->cols), stride(_stride), ustepIdx(_ustepIdx), vstepIdx(_vstepIdx) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        const int rangeBegin = range.begin() * 2;
-        const int rangeEnd = range.end() * 2;
+        const int rangeBegin = range.start * 2;
+        const int rangeEnd = range.end * 2;
 
         size_t uvsteps[2] = {width/2, stride - width/2};
         int usIdx = ustepIdx, vsIdx = vstepIdx;
 
         const uchar* y1 = my1 + rangeBegin * stride;
-        const uchar* u1 = mu + (range.begin() / 2) * stride;
-        const uchar* v1 = mv + (range.begin() / 2) * stride;
+        const uchar* u1 = mu + (range.start / 2) * stride;
+        const uchar* v1 = mv + (range.start / 2) * stride;
 
-        if(range.begin() % 2 == 1)
+        if(range.start % 2 == 1)
         {
             u1 += uvsteps[(usIdx++) & 1];
             v1 += uvsteps[(vsIdx++) & 1];
@@ -2023,7 +2327,7 @@ struct YUV420p2RGB888Invoker
 };
 
 template<int bIdx>
-struct YUV420p2RGBA8888Invoker
+struct YUV420p2RGBA8888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* my1, *mu, *mv;
@@ -2033,19 +2337,19 @@ struct YUV420p2RGBA8888Invoker
     YUV420p2RGBA8888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int _ustepIdx, int _vstepIdx)
         : dst(_dst), my1(_y1), mu(_u), mv(_v), width(_dst->cols), stride(_stride), ustepIdx(_ustepIdx), vstepIdx(_vstepIdx) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int rangeBegin = range.begin() * 2;
-        int rangeEnd = range.end() * 2;
+        int rangeBegin = range.start * 2;
+        int rangeEnd = range.end * 2;
 
         size_t uvsteps[2] = {width/2, stride - width/2};
         int usIdx = ustepIdx, vsIdx = vstepIdx;
 
         const uchar* y1 = my1 + rangeBegin * stride;
-        const uchar* u1 = mu + (range.begin() / 2) * stride;
-        const uchar* v1 = mv + (range.begin() / 2) * stride;
+        const uchar* u1 = mu + (range.start / 2) * stride;
+        const uchar* v1 = mv + (range.start / 2) * stride;
 
-        if(range.begin() % 2 == 1)
+        if(range.start % 2 == 1)
         {
             u1 += uvsteps[(usIdx++) & 1];
             v1 += uvsteps[(vsIdx++) & 1];
@@ -2100,48 +2404,40 @@ template<int bIdx, int uIdx>
 inline void cvtYUV420sp2RGB(Mat& _dst, int _stride, const uchar* _y1, const uchar* _uv)
 {
     YUV420sp2RGB888Invoker<bIdx, uIdx> converter(&_dst, _stride, _y1,  _uv);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows/2), converter);
+        parallel_for_(Range(0, _dst.rows/2), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows/2));
+        converter(Range(0, _dst.rows/2));
 }
 
 template<int bIdx, int uIdx>
 inline void cvtYUV420sp2RGBA(Mat& _dst, int _stride, const uchar* _y1, const uchar* _uv)
 {
     YUV420sp2RGBA8888Invoker<bIdx, uIdx> converter(&_dst, _stride, _y1,  _uv);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows/2), converter);
+        parallel_for_(Range(0, _dst.rows/2), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows/2));
+        converter(Range(0, _dst.rows/2));
 }
 
 template<int bIdx>
 inline void cvtYUV420p2RGB(Mat& _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int ustepIdx, int vstepIdx)
 {
     YUV420p2RGB888Invoker<bIdx> converter(&_dst, _stride, _y1,  _u, _v, ustepIdx, vstepIdx);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows/2), converter);
+        parallel_for_(Range(0, _dst.rows/2), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows/2));
+        converter(Range(0, _dst.rows/2));
 }
 
 template<int bIdx>
 inline void cvtYUV420p2RGBA(Mat& _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int ustepIdx, int vstepIdx)
 {
     YUV420p2RGBA8888Invoker<bIdx> converter(&_dst, _stride, _y1,  _u, _v, ustepIdx, vstepIdx);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows/2), converter);
+        parallel_for_(Range(0, _dst.rows/2), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows/2));
+        converter(Range(0, _dst.rows/2));
 }
 
 ///////////////////////////////////// RGB -> YUV420p /////////////////////////////////////
@@ -2225,7 +2521,7 @@ static void cvtRGBtoYUV420p(const Mat& src, Mat& dst)
 ///////////////////////////////////// YUV422 -> RGB /////////////////////////////////////
 
 template<int bIdx, int uIdx, int yIdx>
-struct YUV422toRGB888Invoker
+struct YUV422toRGB888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* src;
@@ -2234,10 +2530,10 @@ struct YUV422toRGB888Invoker
     YUV422toRGB888Invoker(Mat* _dst, int _stride, const uchar* _yuv)
         : dst(_dst), src(_yuv), width(_dst->cols), stride(_stride) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int rangeBegin = range.begin();
-        int rangeEnd = range.end();
+        int rangeBegin = range.start;
+        int rangeEnd = range.end;
 
         const int uidx = 1 - yIdx + uIdx * 2;
         const int vidx = (2 + uidx) % 4;
@@ -2271,7 +2567,7 @@ struct YUV422toRGB888Invoker
 };
 
 template<int bIdx, int uIdx, int yIdx>
-struct YUV422toRGBA8888Invoker
+struct YUV422toRGBA8888Invoker : ParallelLoopBody
 {
     Mat* dst;
     const uchar* src;
@@ -2280,10 +2576,10 @@ struct YUV422toRGBA8888Invoker
     YUV422toRGBA8888Invoker(Mat* _dst, int _stride, const uchar* _yuv)
         : dst(_dst), src(_yuv), width(_dst->cols), stride(_stride) {}
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int rangeBegin = range.begin();
-        int rangeEnd = range.end();
+        int rangeBegin = range.start;
+        int rangeEnd = range.end;
 
         const int uidx = 1 - yIdx + uIdx * 2;
         const int vidx = (2 + uidx) % 4;
@@ -2324,24 +2620,20 @@ template<int bIdx, int uIdx, int yIdx>
 inline void cvtYUV422toRGB(Mat& _dst, int _stride, const uchar* _yuv)
 {
     YUV422toRGB888Invoker<bIdx, uIdx, yIdx> converter(&_dst, _stride, _yuv);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV422_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows), converter);
+        parallel_for_(Range(0, _dst.rows), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows));
+        converter(Range(0, _dst.rows));
 }
 
 template<int bIdx, int uIdx, int yIdx>
 inline void cvtYUV422toRGBA(Mat& _dst, int _stride, const uchar* _yuv)
 {
     YUV422toRGBA8888Invoker<bIdx, uIdx, yIdx> converter(&_dst, _stride, _yuv);
-#ifdef HAVE_TBB
     if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV422_CONVERSION)
-        parallel_for(BlockedRange(0, _dst.rows), converter);
+        parallel_for_(Range(0, _dst.rows), converter);
     else
-#endif
-        converter(BlockedRange(0, _dst.rows));
+        converter(Range(0, _dst.rows));
 }
 
 /////////////////////////// RGBA <-> mRGBA (alpha premultiplied) //////////////
@@ -2420,6 +2712,39 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create( sz, CV_MAKETYPE(depth, dcn));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( code == CV_BGR2BGRA || code == CV_RGB2RGBA)
+            {
+                if ( CvtColorIPPLoop(src, dst, IPPReorderFunctor(ippiSwapChannelsC3C4RTab[depth], 0, 1, 2)) )
+                    return;
+            }
+            else if( code == CV_BGRA2BGR )
+            {
+                if ( CvtColorIPPLoop(src, dst, IPPGeneralFunctor(ippiCopyAC4C3RTab[depth])) )
+                    return;
+            }
+            else if( code == CV_BGR2RGBA )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPReorderFunctor(ippiSwapChannelsC3C4RTab[depth], 2, 1, 0)) )
+                    return;
+            }
+            else if( code == CV_RGBA2BGR )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPReorderFunctor(ippiSwapChannelsC4C3RTab[depth], 2, 1, 0)) )
+                    return;
+            }
+            else if( code == CV_RGB2BGR )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPReorderFunctor(ippiSwapChannelsC3RTab[depth], 2, 1, 0)) )
+                    return;
+            }
+            else if( code == CV_RGBA2BGRA )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPReorderFunctor(ippiSwapChannelsC4RTab[depth], 2, 1, 0)) )
+                    return;
+            }
+#endif
+
             if( depth == CV_8U )
             {
 #ifdef HAVE_TEGRA_OPTIMIZATION
@@ -2473,6 +2798,29 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, 1));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( code == CV_BGR2GRAY )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPColor2GrayFunctor(ippiColor2GrayC3Tab[depth])) )
+                    return;
+            }
+            else if( code == CV_RGB2GRAY )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGeneralFunctor(ippiRGB2GrayC3Tab[depth])) )
+                    return;
+            }
+            else if( code == CV_BGRA2GRAY )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPColor2GrayFunctor(ippiColor2GrayC4Tab[depth])) )
+                    return;
+            }
+            else if( code == CV_RGBA2GRAY )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGeneralFunctor(ippiRGB2GrayC4Tab[depth])) )
+                    return;
+            }
+#endif
+
             bidx = code == CV_BGR2GRAY || code == CV_BGRA2GRAY ? 0 : 2;
 
             if( depth == CV_8U )
@@ -2502,6 +2850,20 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, dcn));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( code == CV_GRAY2BGR )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGray2BGRFunctor(ippiCopyP3C3RTab[depth])) )
+                    return;
+            }
+            else if( code == CV_GRAY2BGRA )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGray2BGRAFunctor(ippiCopyP3C3RTab[depth], ippiSwapChannelsC3C4RTab[depth], depth)) )
+                    return;
+            }
+#endif
+
+
             if( depth == CV_8U )
             {
 #ifdef HAVE_TEGRA_OPTIMIZATION
@@ -2527,7 +2889,7 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
         case CV_BGR2YUV: case CV_RGB2YUV:
             {
             CV_Assert( scn == 3 || scn == 4 );
-            bidx = code == CV_BGR2YCrCb || code == CV_RGB2YUV ? 0 : 2;
+            bidx = code == CV_BGR2YCrCb || code == CV_BGR2YUV ? 0 : 2;
             static const float yuv_f[] = { 0.114f, 0.587f, 0.299f, 0.492f, 0.877f };
             static const int yuv_i[] = { B2Y, G2Y, R2Y, 8061, 14369 };
             const float* coeffs_f = code == CV_BGR2YCrCb || code == CV_RGB2YCrCb ? 0 : yuv_f;
@@ -2556,7 +2918,7 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             {
             if( dcn <= 0 ) dcn = 3;
             CV_Assert( scn == 3 && (dcn == 3 || dcn == 4) );
-            bidx = code == CV_YCrCb2BGR || code == CV_YUV2RGB ? 0 : 2;
+            bidx = code == CV_YCrCb2BGR || code == CV_YUV2BGR ? 0 : 2;
             static const float yuv_f[] = { 2.032f, -0.395f, -0.581f, 1.140f };
             static const int yuv_i[] = { 33292, -6472, -9519, 18678 };
             const float* coeffs_f = code == CV_YCrCb2BGR || code == CV_YCrCb2RGB ? 0 : yuv_f;
@@ -2581,6 +2943,29 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, 3));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( code == CV_BGR2XYZ && scn == 3 )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC3RTab[depth], ippiRGB2XYZTab[depth], 2, 1, 0, depth)) )
+                    return;
+            }
+            else if( code == CV_BGR2XYZ && scn == 4 )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2XYZTab[depth], 2, 1, 0, depth)) )
+                    return;
+            }
+            else if( code == CV_RGB2XYZ && scn == 3 )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiRGB2XYZTab[depth])) )
+                    return;
+            }
+            else if( code == CV_RGB2XYZ && scn == 4 )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2XYZTab[depth], 0, 1, 2, depth)) )
+                    return;
+            }
+#endif
+
             if( depth == CV_8U )
                 CvtColorLoop(src, dst, RGB2XYZ_i<uchar>(scn, bidx, 0));
             else if( depth == CV_16U )
@@ -2597,6 +2982,29 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, dcn));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( code == CV_XYZ2BGR && dcn == 3 )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPGeneralReorderFunctor(ippiXYZ2RGBTab[depth], ippiSwapChannelsC3RTab[depth], 2, 1, 0, depth)) )
+                    return;
+            }
+            else if( code == CV_XYZ2BGR && dcn == 4 )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiXYZ2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 2, 1, 0, depth)) )
+                    return;
+            }
+            if( code == CV_XYZ2RGB && dcn == 3 )
+            {
+                if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiXYZ2RGBTab[depth])) )
+                    return;
+            }
+            else if( code == CV_XYZ2RGB && dcn == 4 )
+            {
+                if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiXYZ2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 0, 1, 2, depth)) )
+                    return;
+            }
+#endif
+
             if( depth == CV_8U )
                 CvtColorLoop(src, dst, XYZ2RGB_i<uchar>(dcn, bidx, 0));
             else if( depth == CV_16U )
@@ -2617,6 +3025,52 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, 3));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( depth == CV_8U || depth == CV_16U )
+            {
+                if( code == CV_BGR2HSV_FULL && scn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC3RTab[depth], ippiRGB2HSVTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_BGR2HSV_FULL && scn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2HSVTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_RGB2HSV_FULL && scn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiRGB2HSVTab[depth])) )
+                        return;
+                }
+                else if( code == CV_RGB2HSV_FULL && scn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2HSVTab[depth], 0, 1, 2, depth)) )
+                        return;
+                }
+                else if( code == CV_BGR2HLS_FULL && scn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC3RTab[depth], ippiRGB2HLSTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_BGR2HLS_FULL && scn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2HLSTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_RGB2HLS_FULL && scn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiRGB2HLSTab[depth])) )
+                        return;
+                }
+                else if( code == CV_RGB2HLS_FULL && scn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPReorderGeneralFunctor(ippiSwapChannelsC4C3RTab[depth], ippiRGB2HLSTab[depth], 0, 1, 2, depth)) )
+                        return;
+                }
+            }
+#endif
+
             if( code == CV_BGR2HSV || code == CV_RGB2HSV ||
                 code == CV_BGR2HSV_FULL || code == CV_RGB2HSV_FULL )
             {
@@ -2652,6 +3106,52 @@ void cv::cvtColor( InputArray _src, OutputArray _dst, int code, int dcn )
             _dst.create(sz, CV_MAKETYPE(depth, dcn));
             dst = _dst.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( depth == CV_8U || depth == CV_16U )
+            {
+                if( code == CV_HSV2BGR_FULL && dcn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralReorderFunctor(ippiHSV2RGBTab[depth], ippiSwapChannelsC3RTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_HSV2BGR_FULL && dcn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiHSV2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_HSV2RGB_FULL && dcn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiHSV2RGBTab[depth])) )
+                        return;
+                }
+                else if( code == CV_HSV2RGB_FULL && dcn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiHSV2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 0, 1, 2, depth)) )
+                        return;
+                }
+                else if( code == CV_HLS2BGR_FULL && dcn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralReorderFunctor(ippiHLS2RGBTab[depth], ippiSwapChannelsC3RTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_HLS2BGR_FULL && dcn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiHLS2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 2, 1, 0, depth)) )
+                        return;
+                }
+                else if( code == CV_HLS2RGB_FULL && dcn == 3 )
+                {
+                    if( CvtColorIPPLoopCopy(src, dst, IPPGeneralFunctor(ippiHLS2RGBTab[depth])) )
+                        return;
+                }
+                else if( code == CV_HLS2RGB_FULL && dcn == 4 )
+                {
+                    if( CvtColorIPPLoop(src, dst, IPPGeneralReorderFunctor(ippiHLS2RGBTab[depth], ippiSwapChannelsC3C4RTab[depth], 0, 1, 2, depth)) )
+                        return;
+                }
+            }
+#endif
+
             if( code == CV_HSV2BGR || code == CV_HSV2RGB ||
                 code == CV_HSV2BGR_FULL || code == CV_HSV2RGB_FULL )
             {
diff --git a/modules/imgproc/src/connectedcomponents.cpp b/modules/imgproc/src/connectedcomponents.cpp
index 4fee0aacd1..41cb480945 100644
--- a/modules/imgproc/src/connectedcomponents.cpp
+++ b/modules/imgproc/src/connectedcomponents.cpp
@@ -92,20 +92,10 @@ namespace cv{
         }
         void operator()(int r, int c, int l){
             int *row = &statsv.at<int>(l, 0);
-            if(c > row[CC_STAT_WIDTH]){
-                row[CC_STAT_WIDTH] = c;
-            }else{
-                if(c < row[CC_STAT_LEFT]){
-                    row[CC_STAT_LEFT] = c;
-                }
-            }
-            if(r > row[CC_STAT_HEIGHT]){
-                row[CC_STAT_HEIGHT] = r;
-            }else{
-                if(r < row[CC_STAT_TOP]){
-                    row[CC_STAT_TOP] = r;
-                }
-            }
+            row[CC_STAT_LEFT] = MIN(row[CC_STAT_LEFT], c);
+            row[CC_STAT_WIDTH] = MAX(row[CC_STAT_WIDTH], c);
+            row[CC_STAT_TOP] = MIN(row[CC_STAT_TOP], r);
+            row[CC_STAT_HEIGHT] = MAX(row[CC_STAT_HEIGHT], r);
             row[CC_STAT_AREA]++;
             Point2ui64 &integral = integrals[l];
             integral.x += c;
@@ -114,9 +104,7 @@ namespace cv{
         void finish(){
             for(int l = 0; l < statsv.rows; ++l){
                 int *row = &statsv.at<int>(l, 0);
-                row[CC_STAT_LEFT] = std::min(row[CC_STAT_LEFT], row[CC_STAT_WIDTH]);
                 row[CC_STAT_WIDTH] = row[CC_STAT_WIDTH] - row[CC_STAT_LEFT] + 1;
-                row[CC_STAT_TOP] = std::min(row[CC_STAT_TOP], row[CC_STAT_HEIGHT]);
                 row[CC_STAT_HEIGHT] = row[CC_STAT_HEIGHT] - row[CC_STAT_TOP] + 1;
 
                 Point2ui64 &integral = integrals[l];
@@ -399,7 +387,7 @@ int cv::connectedComponentsWithStats(InputArray _img, OutputArray _labels, Outpu
     const cv::Mat img = _img.getMat();
     _labels.create(img.size(), CV_MAT_DEPTH(ltype));
     cv::Mat labels = _labels.getMat();
-    connectedcomponents::CCStatsOp sop(statsv, centroids); 
+    connectedcomponents::CCStatsOp sop(statsv, centroids);
     if(ltype == CV_16U){
         return connectedComponents_sub1(img, labels, connectivity, sop);
     }else if(ltype == CV_32S){
diff --git a/modules/imgproc/src/contours.cpp b/modules/imgproc/src/contours.cpp
index 75e18b2d84..797f8c56c8 100644
--- a/modules/imgproc/src/contours.cpp
+++ b/modules/imgproc/src/contours.cpp
@@ -48,7 +48,7 @@
      (deltas)[6] =  (step), (deltas)[7] =  (step) + (nch))
 
 static const CvPoint icvCodeDeltas[8] =
-    { {1, 0}, {1, -1}, {0, -1}, {-1, -1}, {-1, 0}, {-1, 1}, {0, 1}, {1, 1} };
+    { CvPoint(1, 0), CvPoint(1, -1), CvPoint(0, -1), CvPoint(-1, -1), CvPoint(-1, 0), CvPoint(-1, 1), CvPoint(0, 1), CvPoint(1, 1) };
 
 CV_IMPL void
 cvStartReadChainPoints( CvChain * chain, CvChainPtReader * reader )
@@ -78,7 +78,7 @@ cvReadChainPoint( CvChainPtReader * reader )
 {
     schar *ptr;
     int code;
-    CvPoint pt = { 0, 0 };
+    CvPoint pt;
 
     if( !reader )
         CV_Error( CV_StsNullPtr, "" );
@@ -1339,8 +1339,8 @@ icvFindContoursInInterval( const CvArr* src,
     if( contourHeaderSize < (int)sizeof(CvContour))
         CV_Error( CV_StsBadSize, "Contour header size must be >= sizeof(CvContour)" );
 
-    storage00 = cvCreateChildMemStorage(storage);
-    storage01 = cvCreateChildMemStorage(storage);
+    storage00.reset(cvCreateChildMemStorage(storage));
+    storage01.reset(cvCreateChildMemStorage(storage));
 
     CvMat stub, *mat;
 
diff --git a/modules/imgproc/src/convhull.cpp b/modules/imgproc/src/convhull.cpp
index 019c996320..f4f4193e5e 100644
--- a/modules/imgproc/src/convhull.cpp
+++ b/modules/imgproc/src/convhull.cpp
@@ -400,7 +400,7 @@ cvConvexHull2( const CvArr* array, void* hull_storage,
 
     CvMat* mat = 0;
     CvContour contour_header;
-    union { CvContour c; CvSeq s; } hull_header;
+    CvSeq hull_header;
     CvSeqBlock block, hullblock;
     CvSeq* ptseq = 0;
     CvSeq* hullseq = 0;
@@ -455,8 +455,8 @@ cvConvexHull2( const CvArr* array, void* hull_storage,
 
         hullseq = cvMakeSeqHeaderForArray(
                                           CV_SEQ_KIND_CURVE|CV_MAT_TYPE(mat->type)|CV_SEQ_FLAG_CLOSED,
-                                          sizeof(contour_header), CV_ELEM_SIZE(mat->type), mat->data.ptr,
-                                          mat->cols + mat->rows - 1, &hull_header.s, &hullblock );
+                                          sizeof(hull_header), CV_ELEM_SIZE(mat->type), mat->data.ptr,
+                                          mat->cols + mat->rows - 1, &hull_header, &hullblock );
         cvClearSeq( hullseq );
     }
 
@@ -475,6 +475,7 @@ cvConvexHull2( const CvArr* array, void* hull_storage,
     cv::convexHull(cv::cvarrToMat(ptseq, false, false, 0, &_ptbuf), h0,
                    orientation == CV_CLOCKWISE, CV_MAT_CN(hulltype) == 2);
 
+
     if( hulltype == CV_SEQ_ELTYPE_PPOINT )
     {
         const int* idx = h0.ptr<int>();
@@ -522,7 +523,7 @@ CV_IMPL CvSeq* cvConvexityDefects( const CvArr* array,
     int rev_orientation;
 
     CvContour contour_header;
-    union { CvContour c; CvSeq s; } hull_header;
+    CvSeq hull_header;
     CvSeqBlock block, hullblock;
     CvSeq *ptseq = (CvSeq*)array, *hull = (CvSeq*)hullarray;
 
@@ -575,7 +576,7 @@ CV_IMPL CvSeq* cvConvexityDefects( const CvArr* array,
         hull = cvMakeSeqHeaderForArray(
                                        CV_SEQ_KIND_CURVE|CV_MAT_TYPE(mat->type)|CV_SEQ_FLAG_CLOSED,
                                        sizeof(CvContour), CV_ELEM_SIZE(mat->type), mat->data.ptr,
-                                       mat->cols + mat->rows - 1, &hull_header.s, &hullblock );
+                                       mat->cols + mat->rows - 1, &hull_header, &hullblock );
     }
 
     is_index = CV_SEQ_ELTYPE(hull) == CV_SEQ_ELTYPE_INDEX;
diff --git a/modules/imgproc/src/corner.cpp b/modules/imgproc/src/corner.cpp
index fa3021ab8c..8f8c770068 100644
--- a/modules/imgproc/src/corner.cpp
+++ b/modules/imgproc/src/corner.cpp
@@ -162,7 +162,7 @@ calcHarris( const Mat& _cov, Mat& _dst, double k )
 }
 
 
-void eigen2x2( const float* cov, float* dst, int n )
+static void eigen2x2( const float* cov, float* dst, int n )
 {
     for( int j = 0; j < n; j++ )
     {
diff --git a/modules/imgproc/src/demosaicing.cpp b/modules/imgproc/src/demosaicing.cpp
index a3156dd49c..9326fa1932 100644
--- a/modules/imgproc/src/demosaicing.cpp
+++ b/modules/imgproc/src/demosaicing.cpp
@@ -44,6 +44,8 @@
 
 #include <limits>
 
+#define  CV_DESCALE(x,n)     (((x) + (1 << ((n)-1))) >> (n))
+
 namespace cv
 {
 
@@ -496,7 +498,7 @@ public:
         srcmat(_srcmat), dstmat(_dstmat), Start_with_green(_start_with_green), Blue(_blue), size(_size)
     {
     }
-    
+
     virtual void operator() (const Range& range) const
     {
         SIMDInterpolator vecOp;
@@ -506,24 +508,24 @@ public:
 
         int bayer_step = (int)(srcmat.step/sizeof(T));
         const T* bayer0 = reinterpret_cast<const T*>(srcmat.data) + bayer_step * range.start;
-        
+
         int dst_step = (int)(dstmat.step/sizeof(T));
         T* dst0 = reinterpret_cast<T*>(dstmat.data) + (range.start + 1) * dst_step + dcn + 1;
-        
+
         int blue = Blue, start_with_green = Start_with_green;
         if (range.start % 2)
         {
             blue = -blue;
             start_with_green = !start_with_green;
         }
-        
+
         for (int i = range.start; i < range.end; bayer0 += bayer_step, dst0 += dst_step, ++i )
         {
             int t0, t1;
             const T* bayer = bayer0;
             T* dst = dst0;
             const T* bayer_end = bayer + size.width;
-            
+
             // in case of when size.width <= 2
             if( size.width <= 0 )
             {
@@ -540,27 +542,27 @@ public:
                 }
                 continue;
             }
-            
+
             if( start_with_green )
             {
                 t0 = (bayer[1] + bayer[bayer_step*2+1] + 1) >> 1;
                 t1 = (bayer[bayer_step] + bayer[bayer_step+2] + 1) >> 1;
-                
+
                 dst[-blue] = (T)t0;
                 dst[0] = bayer[bayer_step+1];
                 dst[blue] = (T)t1;
                 if (dcn == 4)
                     dst[2] = alpha; // alpha channel
-                
+
                 bayer++;
                 dst += dcn;
             }
-            
+
             // simd optimization only for dcn == 3
             int delta = dcn == 4 ? 0 : vecOp.bayer2RGB(bayer, bayer_step, dst, size.width, blue);
             bayer += delta;
             dst += delta*dcn;
-            
+
             if (dcn == 3) // Bayer to BGR
             {
                 if( blue > 0 )
@@ -574,7 +576,7 @@ public:
                         dst[-1] = (T)t0;
                         dst[0] = (T)t1;
                         dst[1] = bayer[bayer_step+1];
-                        
+
                         t0 = (bayer[2] + bayer[bayer_step*2+2] + 1) >> 1;
                         t1 = (bayer[bayer_step+1] + bayer[bayer_step+3] + 1) >> 1;
                         dst[2] = (T)t0;
@@ -593,7 +595,7 @@ public:
                         dst[1] = (T)t0;
                         dst[0] = (T)t1;
                         dst[-1] = bayer[bayer_step+1];
-                        
+
                         t0 = (bayer[2] + bayer[bayer_step*2+2] + 1) >> 1;
                         t1 = (bayer[bayer_step+1] + bayer[bayer_step+3] + 1) >> 1;
                         dst[4] = (T)t0;
@@ -617,7 +619,7 @@ public:
                         dst[0] = (T)t1;
                         dst[1] = bayer[bayer_step+1];
                         dst[2] = alpha; // alpha channel
-                        
+
                         t0 = (bayer[2] + bayer[bayer_step*2+2] + 1) >> 1;
                         t1 = (bayer[bayer_step+1] + bayer[bayer_step+3] + 1) >> 1;
                         dst[3] = (T)t0;
@@ -638,7 +640,7 @@ public:
                         dst[0] = (T)t1;
                         dst[1] = (T)t0;
                         dst[2] = alpha; // alpha channel
-                        
+
                         t0 = (bayer[2] + bayer[bayer_step*2+2] + 1) >> 1;
                         t1 = (bayer[bayer_step+1] + bayer[bayer_step+3] + 1) >> 1;
                         dst[3] = (T)t1;
@@ -648,7 +650,7 @@ public:
                     }
                 }
             }
-            
+
             // if skip one pixel at the end of row
             if( bayer < bayer_end )
             {
@@ -664,7 +666,7 @@ public:
                 bayer++;
                 dst += dcn;
             }
-            
+
             // fill the last and the first pixels of row accordingly
             if (dcn == 3)
             {
@@ -686,12 +688,12 @@ public:
                 dst0[size.width*dcn+1] = dst0[size.width*dcn-3];
                 dst0[size.width*dcn+2] = dst0[size.width*dcn-2]; // alpha channel
             }
-            
+
             blue = -blue;
             start_with_green = !start_with_green;
         }
     }
-    
+
 private:
     Mat srcmat;
     Mat dstmat;
@@ -706,7 +708,7 @@ static void Bayer2RGB_( const Mat& srcmat, Mat& dstmat, int code )
     Size size = srcmat.size();
     int blue = code == CV_BayerBG2BGR || code == CV_BayerGB2BGR ? -1 : 1;
     int start_with_green = code == CV_BayerGB2BGR || code == CV_BayerGR2BGR;
- 
+
     int dcn = dstmat.channels();
     size.height -= 2;
     size.width -= 2;
@@ -717,7 +719,7 @@ static void Bayer2RGB_( const Mat& srcmat, Mat& dstmat, int code )
         Bayer2RGB_Invoker<T, SIMDInterpolator> invoker(srcmat, dstmat, start_with_green, blue, size);
         parallel_for_(range, invoker, dstmat.total()/static_cast<double>(1<<16));
     }
-    
+
     // filling the first and the last rows
     size = dstmat.size();
     T* dst0 = (T*)dstmat.data;
@@ -1018,9 +1020,9 @@ static void Bayer2RGB_VNG_8u( const Mat& srcmat, Mat& dstmat, int code )
                     R = G + cvRound((Rs - Gs)*scale[ng]);
                     B = G + cvRound((Bs - Gs)*scale[ng]);
                 }
-                dstrow[blueIdx] = CV_CAST_8U(B);
-                dstrow[1] = CV_CAST_8U(G);
-                dstrow[blueIdx^2] = CV_CAST_8U(R);
+                dstrow[blueIdx] = cv::saturate_cast<uchar>(B);
+                dstrow[1] = cv::saturate_cast<uchar>(G);
+                dstrow[blueIdx^2] = cv::saturate_cast<uchar>(R);
                 greenCell = !greenCell;
             }
 
diff --git a/modules/imgproc/src/distransform.cpp b/modules/imgproc/src/distransform.cpp
index bdb2a1b1b8..29e3fc10a2 100644
--- a/modules/imgproc/src/distransform.cpp
+++ b/modules/imgproc/src/distransform.cpp
@@ -46,6 +46,7 @@ namespace cv
 
 static const int DIST_SHIFT = 16;
 static const int INIT_DIST0 = (INT_MAX >> 2);
+#define  CV_FLT_TO_FIX(x,n)  cvRound((x)*(1<<(n)))
 
 static void
 initTopBottom( Mat& temp, int border )
@@ -441,7 +442,7 @@ static void getDistanceTransformMask( int maskType, float *metrics )
     }
 }
 
-struct DTColumnInvoker
+struct DTColumnInvoker : ParallelLoopBody
 {
     DTColumnInvoker( const Mat* _src, Mat* _dst, const int* _sat_tab, const float* _sqr_tab)
     {
@@ -451,9 +452,9 @@ struct DTColumnInvoker
         sqr_tab = _sqr_tab;
     }
 
-    void operator()( const BlockedRange& range ) const
+    void operator()( const Range& range ) const
     {
-        int i, i1 = range.begin(), i2 = range.end();
+        int i, i1 = range.start, i2 = range.end;
         int m = src->rows;
         size_t sstep = src->step, dstep = dst->step/sizeof(float);
         AutoBuffer<int> _d(m);
@@ -488,7 +489,7 @@ struct DTColumnInvoker
 };
 
 
-struct DTRowInvoker
+struct DTRowInvoker : ParallelLoopBody
 {
     DTRowInvoker( Mat* _dst, const float* _sqr_tab, const float* _inv_tab )
     {
@@ -497,10 +498,10 @@ struct DTRowInvoker
         inv_tab = _inv_tab;
     }
 
-    void operator()( const BlockedRange& range ) const
+    void operator()( const Range& range ) const
     {
         const float inf = 1e15f;
-        int i, i1 = range.begin(), i2 = range.end();
+        int i, i1 = range.start, i2 = range.end;
         int n = dst->cols;
         AutoBuffer<uchar> _buf((n+2)*2*sizeof(float) + (n+2)*sizeof(int));
         float* f = (float*)(uchar*)_buf;
@@ -577,7 +578,7 @@ trueDistTrans( const Mat& src, Mat& dst )
     for( ; i <= m*3; i++ )
         sat_tab[i] = i - shift;
 
-    cv::parallel_for(cv::BlockedRange(0, n), cv::DTColumnInvoker(&src, &dst, sat_tab, sqr_tab));
+    cv::parallel_for_(cv::Range(0, n), cv::DTColumnInvoker(&src, &dst, sat_tab, sqr_tab));
 
     // stage 2: compute modified distance transform for each row
     float* inv_tab = sqr_tab + n;
@@ -589,7 +590,7 @@ trueDistTrans( const Mat& src, Mat& dst )
         sqr_tab[i] = (float)(i*i);
     }
 
-    cv::parallel_for(cv::BlockedRange(0, m), cv::DTRowInvoker(&dst, sqr_tab, inv_tab));
+    cv::parallel_for_(cv::Range(0, m), cv::DTRowInvoker(&dst, sqr_tab, inv_tab));
 }
 
 
@@ -619,7 +620,7 @@ distanceATS_L1_8u( const Mat& src, Mat& dst )
 
     ////////////////////// forward scan ////////////////////////
     for( x = 0; x < 256; x++ )
-        lut[x] = CV_CAST_8U(x+1);
+        lut[x] = cv::saturate_cast<uchar>(x+1);
 
     //init first pixel to max (we're going to be skipping it)
     dbase[0] = (uchar)(sbase[0] == 0 ? 0 : 255);
@@ -663,7 +664,7 @@ distanceATS_L1_8u( const Mat& src, Mat& dst )
         // do right edge
         a = lut[dbase[width-1+dststep]];
         dbase[width-1] = (uchar)(MIN(a, dbase[width-1]));
-        
+
         for( x = width - 2; x >= 0; x-- )
         {
             int b = dbase[x+dststep];
@@ -743,6 +744,16 @@ void cv::distanceTransform( InputArray _src, OutputArray _dst, OutputArray _labe
 
         if( labelType == CV_DIST_LABEL_CCOMP )
         {
+        #if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+            if( maskSize == CV_DIST_MASK_5 )
+            {
+                IppiSize roi = { src->cols, src->rows };
+                if( ippiDistanceTransform_5x5_8u32f_C1R(
+                        src->data.ptr, src->step,
+                        dst->data.fl, dst->step, roi, _mask) >= 0 )
+                    return;
+            }
+        #endif
             Mat zpix = src == 0;
             connectedComponents(zpix, labels, 8, CV_32S);
         }
diff --git a/modules/imgproc/src/featureselect.cpp b/modules/imgproc/src/featureselect.cpp
index bcb9bd7091..8c740382f2 100644
--- a/modules/imgproc/src/featureselect.cpp
+++ b/modules/imgproc/src/featureselect.cpp
@@ -92,7 +92,7 @@ void cv::goodFeaturesToTrack( InputArray _image, OutputArray _corners,
         }
     }
 
-    sort( tmpCorners, greaterThanPtr<float>() );
+    std::sort( tmpCorners.begin(), tmpCorners.end(), greaterThanPtr<float>() );
     std::vector<Point2f> corners;
     size_t i, j, total = tmpCorners.size(), ncorners = 0;
 
diff --git a/modules/imgproc/src/filter.cpp b/modules/imgproc/src/filter.cpp
index bef3fe29f7..f7e7efa854 100644
--- a/modules/imgproc/src/filter.cpp
+++ b/modules/imgproc/src/filter.cpp
@@ -46,50 +46,6 @@
                                     Base Image Filter
 \****************************************************************************************/
 
-/*
- Various border types, image boundaries are denoted with '|'
-
- * BORDER_REPLICATE:     aaaaaa|abcdefgh|hhhhhhh
- * BORDER_REFLECT:       fedcba|abcdefgh|hgfedcb
- * BORDER_REFLECT_101:   gfedcb|abcdefgh|gfedcba
- * BORDER_WRAP:          cdefgh|abcdefgh|abcdefg
- * BORDER_CONSTANT:      iiiiii|abcdefgh|iiiiiii  with some specified 'i'
- */
-int cv::borderInterpolate( int p, int len, int borderType )
-{
-    if( (unsigned)p < (unsigned)len )
-        ;
-    else if( borderType == BORDER_REPLICATE )
-        p = p < 0 ? 0 : len - 1;
-    else if( borderType == BORDER_REFLECT || borderType == BORDER_REFLECT_101 )
-    {
-        int delta = borderType == BORDER_REFLECT_101;
-        if( len == 1 )
-            return 0;
-        do
-        {
-            if( p < 0 )
-                p = -p - 1 + delta;
-            else
-                p = len - 1 - (p - len) - delta;
-        }
-        while( (unsigned)p >= (unsigned)len );
-    }
-    else if( borderType == BORDER_WRAP )
-    {
-        if( p < 0 )
-            p -= ((p-len+1)/len)*len;
-        if( p >= len )
-            p %= len;
-    }
-    else if( borderType == BORDER_CONSTANT )
-        p = -1;
-    else
-        CV_Error( CV_StsBadArg, "Unknown/unsupported border type" );
-    return p;
-}
-
-
 namespace cv
 {
 
@@ -161,7 +117,7 @@ void FilterEngine::init( const Ptr<BaseFilter>& _filter2D,
 
     if( isSeparable() )
     {
-        CV_Assert( !rowFilter.empty() && !columnFilter.empty() );
+        CV_Assert( rowFilter && columnFilter );
         ksize = Size(rowFilter->ksize, columnFilter->ksize);
         anchor = Point(rowFilter->anchor, columnFilter->anchor);
     }
@@ -193,7 +149,7 @@ void FilterEngine::init( const Ptr<BaseFilter>& _filter2D,
     wholeSize = Size(-1,-1);
 }
 
-static const int VEC_ALIGN = CV_MALLOC_ALIGN;
+#define VEC_ALIGN CV_MALLOC_ALIGN
 
 int FilterEngine::start(Size _wholeSize, Rect _roi, int _maxBufRows)
 {
@@ -288,9 +244,9 @@ int FilterEngine::start(Size _wholeSize, Rect _roi, int _maxBufRows)
     rowCount = dstY = 0;
     startY = startY0 = std::max(roi.y - anchor.y, 0);
     endY = std::min(roi.y + roi.height + ksize.height - anchor.y - 1, wholeSize.height);
-    if( !columnFilter.empty() )
+    if( columnFilter )
         columnFilter->reset();
-    if( !filter2D.empty() )
+    if( filter2D )
         filter2D->reset();
 
     return startY;
@@ -2779,42 +2735,42 @@ cv::Ptr<cv::BaseRowFilter> cv::getLinearRowFilter( int srcType, int bufType,
     if( (symmetryType & (KERNEL_SYMMETRICAL|KERNEL_ASYMMETRICAL)) != 0 && ksize <= 5 )
     {
         if( sdepth == CV_8U && ddepth == CV_32S )
-            return Ptr<BaseRowFilter>(new SymmRowSmallFilter<uchar, int, SymmRowSmallVec_8u32s>
-                (kernel, anchor, symmetryType, SymmRowSmallVec_8u32s(kernel, symmetryType)));
+            return makePtr<SymmRowSmallFilter<uchar, int, SymmRowSmallVec_8u32s> >
+                (kernel, anchor, symmetryType, SymmRowSmallVec_8u32s(kernel, symmetryType));
         if( sdepth == CV_32F && ddepth == CV_32F )
-            return Ptr<BaseRowFilter>(new SymmRowSmallFilter<float, float, SymmRowSmallVec_32f>
-                (kernel, anchor, symmetryType, SymmRowSmallVec_32f(kernel, symmetryType)));
+            return makePtr<SymmRowSmallFilter<float, float, SymmRowSmallVec_32f> >
+                (kernel, anchor, symmetryType, SymmRowSmallVec_32f(kernel, symmetryType));
     }
 
     if( sdepth == CV_8U && ddepth == CV_32S )
-        return Ptr<BaseRowFilter>(new RowFilter<uchar, int, RowVec_8u32s>
-            (kernel, anchor, RowVec_8u32s(kernel)));
+        return makePtr<RowFilter<uchar, int, RowVec_8u32s> >
+            (kernel, anchor, RowVec_8u32s(kernel));
     if( sdepth == CV_8U && ddepth == CV_32F )
-        return Ptr<BaseRowFilter>(new RowFilter<uchar, float, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<uchar, float, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_8U && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowFilter<uchar, double, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<uchar, double, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_16U && ddepth == CV_32F )
-        return Ptr<BaseRowFilter>(new RowFilter<ushort, float, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<ushort, float, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_16U && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowFilter<ushort, double, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<ushort, double, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_16S && ddepth == CV_32F )
-        return Ptr<BaseRowFilter>(new RowFilter<short, float, RowVec_16s32f>
-                                  (kernel, anchor, RowVec_16s32f(kernel)));
+        return makePtr<RowFilter<short, float, RowVec_16s32f> >
+                                  (kernel, anchor, RowVec_16s32f(kernel));
     if( sdepth == CV_16S && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowFilter<short, double, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<short, double, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_32F && ddepth == CV_32F )
-        return Ptr<BaseRowFilter>(new RowFilter<float, float, RowVec_32f>
-            (kernel, anchor, RowVec_32f(kernel)));
+        return makePtr<RowFilter<float, float, RowVec_32f> >
+            (kernel, anchor, RowVec_32f(kernel));
     if( sdepth == CV_32F && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowFilter<float, double, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<float, double, RowNoVec> >(kernel, anchor);
     if( sdepth == CV_64F && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowFilter<double, double, RowNoVec>(kernel, anchor));
+        return makePtr<RowFilter<double, double, RowNoVec> >(kernel, anchor);
 
     CV_Error_( CV_StsNotImplemented,
         ("Unsupported combination of source format (=%d), and buffer format (=%d)",
         srcType, bufType));
 
-    return Ptr<BaseRowFilter>(0);
+    return Ptr<BaseRowFilter>();
 }
 
 
@@ -2833,24 +2789,24 @@ cv::Ptr<cv::BaseColumnFilter> cv::getLinearColumnFilter( int bufType, int dstTyp
     if( !(symmetryType & (KERNEL_SYMMETRICAL|KERNEL_ASYMMETRICAL)) )
     {
         if( ddepth == CV_8U && sdepth == CV_32S )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<FixedPtCastEx<int, uchar>, ColumnNoVec>
-            (kernel, anchor, delta, FixedPtCastEx<int, uchar>(bits)));
+            return makePtr<ColumnFilter<FixedPtCastEx<int, uchar>, ColumnNoVec> >
+            (kernel, anchor, delta, FixedPtCastEx<int, uchar>(bits));
         if( ddepth == CV_8U && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<float, uchar>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<float, uchar>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_8U && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<double, uchar>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<double, uchar>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_16U && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<float, ushort>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<float, ushort>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_16U && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<double, ushort>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<double, ushort>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_16S && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<float, short>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<float, short>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_16S && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<double, short>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<double, short>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_32F && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<float, float>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<float, float>, ColumnNoVec> >(kernel, anchor, delta);
         if( ddepth == CV_64F && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new ColumnFilter<Cast<double, double>, ColumnNoVec>(kernel, anchor, delta));
+            return makePtr<ColumnFilter<Cast<double, double>, ColumnNoVec> >(kernel, anchor, delta);
     }
     else
     {
@@ -2858,60 +2814,60 @@ cv::Ptr<cv::BaseColumnFilter> cv::getLinearColumnFilter( int bufType, int dstTyp
         if( ksize == 3 )
         {
             if( ddepth == CV_8U && sdepth == CV_32S )
-                return Ptr<BaseColumnFilter>(new SymmColumnSmallFilter<
-                    FixedPtCastEx<int, uchar>, SymmColumnVec_32s8u>
+                return makePtr<SymmColumnSmallFilter<
+                    FixedPtCastEx<int, uchar>, SymmColumnVec_32s8u> >
                     (kernel, anchor, delta, symmetryType, FixedPtCastEx<int, uchar>(bits),
-                    SymmColumnVec_32s8u(kernel, symmetryType, bits, delta)));
+                    SymmColumnVec_32s8u(kernel, symmetryType, bits, delta));
             if( ddepth == CV_16S && sdepth == CV_32S && bits == 0 )
-                return Ptr<BaseColumnFilter>(new SymmColumnSmallFilter<Cast<int, short>,
-                    SymmColumnSmallVec_32s16s>(kernel, anchor, delta, symmetryType,
-                        Cast<int, short>(), SymmColumnSmallVec_32s16s(kernel, symmetryType, bits, delta)));
+                return makePtr<SymmColumnSmallFilter<Cast<int, short>,
+                    SymmColumnSmallVec_32s16s> >(kernel, anchor, delta, symmetryType,
+                        Cast<int, short>(), SymmColumnSmallVec_32s16s(kernel, symmetryType, bits, delta));
             if( ddepth == CV_32F && sdepth == CV_32F )
-                return Ptr<BaseColumnFilter>(new SymmColumnSmallFilter<
-                    Cast<float, float>,SymmColumnSmallVec_32f>
+                return makePtr<SymmColumnSmallFilter<
+                    Cast<float, float>,SymmColumnSmallVec_32f> >
                     (kernel, anchor, delta, symmetryType, Cast<float, float>(),
-                    SymmColumnSmallVec_32f(kernel, symmetryType, 0, delta)));
+                    SymmColumnSmallVec_32f(kernel, symmetryType, 0, delta));
         }
         if( ddepth == CV_8U && sdepth == CV_32S )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<FixedPtCastEx<int, uchar>, SymmColumnVec_32s8u>
+            return makePtr<SymmColumnFilter<FixedPtCastEx<int, uchar>, SymmColumnVec_32s8u> >
                 (kernel, anchor, delta, symmetryType, FixedPtCastEx<int, uchar>(bits),
-                SymmColumnVec_32s8u(kernel, symmetryType, bits, delta)));
+                SymmColumnVec_32s8u(kernel, symmetryType, bits, delta));
         if( ddepth == CV_8U && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<float, uchar>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<float, uchar>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_8U && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<double, uchar>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<double, uchar>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_16U && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<float, ushort>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<float, ushort>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_16U && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<double, ushort>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<double, ushort>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_16S && sdepth == CV_32S )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<int, short>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<int, short>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_16S && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<float, short>, SymmColumnVec_32f16s>
+            return makePtr<SymmColumnFilter<Cast<float, short>, SymmColumnVec_32f16s> >
                  (kernel, anchor, delta, symmetryType, Cast<float, short>(),
-                  SymmColumnVec_32f16s(kernel, symmetryType, 0, delta)));
+                  SymmColumnVec_32f16s(kernel, symmetryType, 0, delta));
         if( ddepth == CV_16S && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<double, short>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<double, short>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
         if( ddepth == CV_32F && sdepth == CV_32F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<float, float>, SymmColumnVec_32f>
+            return makePtr<SymmColumnFilter<Cast<float, float>, SymmColumnVec_32f> >
                 (kernel, anchor, delta, symmetryType, Cast<float, float>(),
-                SymmColumnVec_32f(kernel, symmetryType, 0, delta)));
+                SymmColumnVec_32f(kernel, symmetryType, 0, delta));
         if( ddepth == CV_64F && sdepth == CV_64F )
-            return Ptr<BaseColumnFilter>(new SymmColumnFilter<Cast<double, double>, ColumnNoVec>
-                (kernel, anchor, delta, symmetryType));
+            return makePtr<SymmColumnFilter<Cast<double, double>, ColumnNoVec> >
+                (kernel, anchor, delta, symmetryType);
     }
 
     CV_Error_( CV_StsNotImplemented,
         ("Unsupported combination of buffer format (=%d), and destination format (=%d)",
         bufType, dstType));
 
-    return Ptr<BaseColumnFilter>(0);
+    return Ptr<BaseColumnFilter>();
 }
 
 
@@ -2977,7 +2933,7 @@ cv::Ptr<cv::FilterEngine> cv::createSeparableLinearFilter(
     Ptr<BaseColumnFilter> _columnFilter = getLinearColumnFilter(
         _bufType, _dstType, columnKernel, _anchor.y, ctype, _delta, bits );
 
-    return Ptr<FilterEngine>( new FilterEngine(Ptr<BaseFilter>(0), _rowFilter, _columnFilter,
+    return Ptr<FilterEngine>( new FilterEngine(Ptr<BaseFilter>(), _rowFilter, _columnFilter,
         _srcType, _dstType, _bufType, _rowBorderType, _columnBorderType, _borderValue ));
 }
 
@@ -3129,13 +3085,13 @@ cv::Ptr<cv::BaseFilter> cv::getLinearFilter(int srcType, int dstType,
     anchor = normalizeAnchor(anchor, _kernel.size());
 
     /*if( sdepth == CV_8U && ddepth == CV_8U && kdepth == CV_32S )
-        return Ptr<BaseFilter>(new Filter2D<uchar, FixedPtCastEx<int, uchar>, FilterVec_8u>
+        return makePtr<Filter2D<uchar, FixedPtCastEx<int, uchar>, FilterVec_8u> >
             (_kernel, anchor, delta, FixedPtCastEx<int, uchar>(bits),
-            FilterVec_8u(_kernel, bits, delta)));
+            FilterVec_8u(_kernel, bits, delta));
     if( sdepth == CV_8U && ddepth == CV_16S && kdepth == CV_32S )
-        return Ptr<BaseFilter>(new Filter2D<uchar, FixedPtCastEx<int, short>, FilterVec_8u16s>
+        return makePtr<Filter2D<uchar, FixedPtCastEx<int, short>, FilterVec_8u16s> >
             (_kernel, anchor, delta, FixedPtCastEx<int, short>(bits),
-            FilterVec_8u16s(_kernel, bits, delta)));*/
+            FilterVec_8u16s(_kernel, bits, delta));*/
 
     kdepth = sdepth == CV_64F || ddepth == CV_64F ? CV_64F : CV_32F;
     Mat kernel;
@@ -3145,53 +3101,53 @@ cv::Ptr<cv::BaseFilter> cv::getLinearFilter(int srcType, int dstType,
         _kernel.convertTo(kernel, kdepth, _kernel.type() == CV_32S ? 1./(1 << bits) : 1.);
 
     if( sdepth == CV_8U && ddepth == CV_8U )
-        return Ptr<BaseFilter>(new Filter2D<uchar, Cast<float, uchar>, FilterVec_8u>
-            (kernel, anchor, delta, Cast<float, uchar>(), FilterVec_8u(kernel, 0, delta)));
+        return makePtr<Filter2D<uchar, Cast<float, uchar>, FilterVec_8u> >
+            (kernel, anchor, delta, Cast<float, uchar>(), FilterVec_8u(kernel, 0, delta));
     if( sdepth == CV_8U && ddepth == CV_16U )
-        return Ptr<BaseFilter>(new Filter2D<uchar,
-            Cast<float, ushort>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<uchar,
+            Cast<float, ushort>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_8U && ddepth == CV_16S )
-        return Ptr<BaseFilter>(new Filter2D<uchar, Cast<float, short>, FilterVec_8u16s>
-            (kernel, anchor, delta, Cast<float, short>(), FilterVec_8u16s(kernel, 0, delta)));
+        return makePtr<Filter2D<uchar, Cast<float, short>, FilterVec_8u16s> >
+            (kernel, anchor, delta, Cast<float, short>(), FilterVec_8u16s(kernel, 0, delta));
     if( sdepth == CV_8U && ddepth == CV_32F )
-        return Ptr<BaseFilter>(new Filter2D<uchar,
-            Cast<float, float>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<uchar,
+            Cast<float, float>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_8U && ddepth == CV_64F )
-        return Ptr<BaseFilter>(new Filter2D<uchar,
-            Cast<double, double>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<uchar,
+            Cast<double, double>, FilterNoVec> >(kernel, anchor, delta);
 
     if( sdepth == CV_16U && ddepth == CV_16U )
-        return Ptr<BaseFilter>(new Filter2D<ushort,
-            Cast<float, ushort>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<ushort,
+            Cast<float, ushort>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_16U && ddepth == CV_32F )
-        return Ptr<BaseFilter>(new Filter2D<ushort,
-            Cast<float, float>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<ushort,
+            Cast<float, float>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_16U && ddepth == CV_64F )
-        return Ptr<BaseFilter>(new Filter2D<ushort,
-            Cast<double, double>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<ushort,
+            Cast<double, double>, FilterNoVec> >(kernel, anchor, delta);
 
     if( sdepth == CV_16S && ddepth == CV_16S )
-        return Ptr<BaseFilter>(new Filter2D<short,
-            Cast<float, short>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<short,
+            Cast<float, short>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_16S && ddepth == CV_32F )
-        return Ptr<BaseFilter>(new Filter2D<short,
-            Cast<float, float>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<short,
+            Cast<float, float>, FilterNoVec> >(kernel, anchor, delta);
     if( sdepth == CV_16S && ddepth == CV_64F )
-        return Ptr<BaseFilter>(new Filter2D<short,
-            Cast<double, double>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<short,
+            Cast<double, double>, FilterNoVec> >(kernel, anchor, delta);
 
     if( sdepth == CV_32F && ddepth == CV_32F )
-        return Ptr<BaseFilter>(new Filter2D<float, Cast<float, float>, FilterVec_32f>
-            (kernel, anchor, delta, Cast<float, float>(), FilterVec_32f(kernel, 0, delta)));
+        return makePtr<Filter2D<float, Cast<float, float>, FilterVec_32f> >
+            (kernel, anchor, delta, Cast<float, float>(), FilterVec_32f(kernel, 0, delta));
     if( sdepth == CV_64F && ddepth == CV_64F )
-        return Ptr<BaseFilter>(new Filter2D<double,
-            Cast<double, double>, FilterNoVec>(kernel, anchor, delta));
+        return makePtr<Filter2D<double,
+            Cast<double, double>, FilterNoVec> >(kernel, anchor, delta);
 
     CV_Error_( CV_StsNotImplemented,
         ("Unsupported combination of source format (=%d), and destination format (=%d)",
         srcType, dstType));
 
-    return Ptr<BaseFilter>(0);
+    return Ptr<BaseFilter>();
 }
 
 
@@ -3222,9 +3178,9 @@ cv::Ptr<cv::FilterEngine> cv::createLinearFilter( int _srcType, int _dstType,
     Ptr<BaseFilter> _filter2D = getLinearFilter(_srcType, _dstType,
         kernel, _anchor, _delta, bits);
 
-    return Ptr<FilterEngine>(new FilterEngine(_filter2D, Ptr<BaseRowFilter>(0),
-        Ptr<BaseColumnFilter>(0), _srcType, _dstType, _srcType,
-        _rowBorderType, _columnBorderType, _borderValue ));
+    return makePtr<FilterEngine>(_filter2D, Ptr<BaseRowFilter>(),
+        Ptr<BaseColumnFilter>(), _srcType, _dstType, _srcType,
+        _rowBorderType, _columnBorderType, _borderValue );
 }
 
 
diff --git a/modules/imgproc/src/floodfill.cpp b/modules/imgproc/src/floodfill.cpp
index 5436c7b02d..40556c48dd 100644
--- a/modules/imgproc/src/floodfill.cpp
+++ b/modules/imgproc/src/floodfill.cpp
@@ -127,7 +127,6 @@ floodFill_CnIR( Mat& image, Point seed,
                _Tp newVal, ConnectedComp* region, int flags,
                std::vector<FFillSegment>* buffer )
 {
-    typedef typename DataType<_Tp>::channel_type _CTp;
     _Tp* img = (_Tp*)(image.data + image.step * seed.y);
     Size roi = image.size();
     int i, L, R;
@@ -279,7 +278,6 @@ floodFillGrad_CnIR( Mat& image, Mat& msk,
                    Diff diff, ConnectedComp* region, int flags,
                    std::vector<FFillSegment>* buffer )
 {
-    typedef typename DataType<_Tp>::channel_type _CTp;
     int step = (int)image.step, maskStep = (int)msk.step;
     uchar* pImage = image.data;
     _Tp* img = (_Tp*)(pImage + step*seed.y);
@@ -610,7 +608,7 @@ int cv::floodFill( InputOutputArray _image, InputOutputArray _mask,
                 &comp, flags, &buffer);
     else
         CV_Error(CV_StsUnsupportedFormat, "");
-    
+
     if( rect )
         *rect = comp.rect;
     return comp.area;
diff --git a/modules/imgproc/src/gcgraph.hpp b/modules/imgproc/src/gcgraph.hpp
index 59c9744e7b..f93bd19a20 100644
--- a/modules/imgproc/src/gcgraph.hpp
+++ b/modules/imgproc/src/gcgraph.hpp
@@ -241,7 +241,7 @@ TWeight GCGraph<TWeight>::maxFlow()
 
         // find the minimum edge weight along the path
         minWeight = edgePtr[e0].weight;
-        assert( minWeight > 0 );
+        CV_Assert( minWeight > 0 );
         // k = 1: source tree, k = 0: destination tree
         for( int k = 1; k >= 0; k-- )
         {
@@ -251,11 +251,11 @@ TWeight GCGraph<TWeight>::maxFlow()
                     break;
                 weight = edgePtr[ei^k].weight;
                 minWeight = MIN(minWeight, weight);
-                assert( minWeight > 0 );
+                CV_Assert( minWeight > 0 );
             }
             weight = fabs(v->weight);
             minWeight = MIN(minWeight, weight);
-            assert( minWeight > 0 );
+            CV_Assert( minWeight > 0 );
         }
 
         // modify weights of the edges along the path and collect orphans
diff --git a/modules/imgproc/src/generalized_hough.cpp b/modules/imgproc/src/generalized_hough.cpp
index ab0890ef83..a261d644b8 100644
--- a/modules/imgproc/src/generalized_hough.cpp
+++ b/modules/imgproc/src/generalized_hough.cpp
@@ -41,20 +41,14 @@
 
 #include "precomp.hpp"
 #include <functional>
+#include <limits>
 
 using namespace cv;
 
+// common
+
 namespace
 {
-    /////////////////////////////////////
-    // Common
-
-    template <typename T, class A> void releaseVector(std::vector<T, A>& v)
-    {
-        std::vector<T, A> empty;
-        empty.swap(v);
-    }
-
     double toRad(double a)
     {
         return a * CV_PI / 180.0;
@@ -65,70 +59,112 @@ namespace
         return fabs(v) > std::numeric_limits<float>::epsilon();
     }
 
-    class GHT_Pos : public GeneralizedHough
+    class GeneralizedHoughBase
     {
-    public:
-        GHT_Pos();
-
     protected:
-        void setTemplateImpl(const Mat& edges, const Mat& dx, const Mat& dy, Point templCenter);
-        void detectImpl(const Mat& edges, const Mat& dx, const Mat& dy, OutputArray positions, OutputArray votes);
-        void releaseImpl();
+        GeneralizedHoughBase();
+        virtual ~GeneralizedHoughBase() {}
+
+        void setTemplateImpl(InputArray templ, Point templCenter);
+        void setTemplateImpl(InputArray edges, InputArray dx, InputArray dy, Point templCenter);
+
+        void detectImpl(InputArray image, OutputArray positions, OutputArray votes);
+        void detectImpl(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes);
 
         virtual void processTempl() = 0;
         virtual void processImage() = 0;
 
+        int cannyLowThresh_;
+        int cannyHighThresh_;
+        double minDist_;
+        double dp_;
+
+        Size templSize_;
+        Point templCenter_;
+        Mat templEdges_;
+        Mat templDx_;
+        Mat templDy_;
+
+        Size imageSize_;
+        Mat imageEdges_;
+        Mat imageDx_;
+        Mat imageDy_;
+
+        std::vector<Vec4f> posOutBuf_;
+        std::vector<Vec3i> voteOutBuf_;
+
+    private:
+        void calcEdges(InputArray src, Mat& edges, Mat& dx, Mat& dy);
         void filterMinDist();
         void convertTo(OutputArray positions, OutputArray votes);
-
-        double minDist;
-
-        Size templSize;
-        Point templCenter;
-        Mat templEdges;
-        Mat templDx;
-        Mat templDy;
-
-        Size imageSize;
-        Mat imageEdges;
-        Mat imageDx;
-        Mat imageDy;
-
-        std::vector<Vec4f> posOutBuf;
-        std::vector<Vec3i> voteOutBuf;
     };
 
-    GHT_Pos::GHT_Pos()
+    GeneralizedHoughBase::GeneralizedHoughBase()
     {
-        minDist = 1.0;
+        cannyLowThresh_ = 50;
+        cannyHighThresh_ = 100;
+        minDist_ = 1.0;
+        dp_ = 1.0;
     }
 
-    void GHT_Pos::setTemplateImpl(const Mat& edges, const Mat& dx, const Mat& dy, Point templCenter_)
+    void GeneralizedHoughBase::calcEdges(InputArray _src, Mat& edges, Mat& dx, Mat& dy)
     {
-        templSize = edges.size();
-        templCenter = templCenter_;
-        edges.copyTo(templEdges);
-        dx.copyTo(templDx);
-        dy.copyTo(templDy);
+        Mat src = _src.getMat();
+
+        CV_Assert( src.type() == CV_8UC1 );
+        CV_Assert( cannyLowThresh_ > 0 && cannyLowThresh_ < cannyHighThresh_ );
+
+        Canny(src, edges, cannyLowThresh_, cannyHighThresh_);
+        Sobel(src, dx, CV_32F, 1, 0);
+        Sobel(src, dy, CV_32F, 0, 1);
+    }
+
+    void GeneralizedHoughBase::setTemplateImpl(InputArray templ, Point templCenter)
+    {
+        calcEdges(templ, templEdges_, templDx_, templDy_);
+
+        if (templCenter == Point(-1, -1))
+            templCenter = Point(templEdges_.cols / 2, templEdges_.rows / 2);
+
+        templSize_ = templEdges_.size();
+        templCenter_ = templCenter;
 
         processTempl();
     }
 
-    void GHT_Pos::detectImpl(const Mat& edges, const Mat& dx, const Mat& dy, OutputArray positions, OutputArray votes)
+    void GeneralizedHoughBase::setTemplateImpl(InputArray edges, InputArray dx, InputArray dy, Point templCenter)
     {
-        imageSize = edges.size();
-        edges.copyTo(imageEdges);
-        dx.copyTo(imageDx);
-        dy.copyTo(imageDy);
+        edges.getMat().copyTo(templEdges_);
+        dx.getMat().copyTo(templDx_);
+        dy.getMat().copyTo(templDy_);
 
-        posOutBuf.clear();
-        voteOutBuf.clear();
+        CV_Assert( templEdges_.type() == CV_8UC1 );
+        CV_Assert( templDx_.type() == CV_32FC1 && templDx_.size() == templEdges_.size() );
+        CV_Assert( templDy_.type() == templDx_.type() && templDy_.size() == templEdges_.size() );
+
+        if (templCenter == Point(-1, -1))
+            templCenter = Point(templEdges_.cols / 2, templEdges_.rows / 2);
+
+        templSize_ = templEdges_.size();
+        templCenter_ = templCenter;
+
+        processTempl();
+    }
+
+    void GeneralizedHoughBase::detectImpl(InputArray image, OutputArray positions, OutputArray votes)
+    {
+        calcEdges(image, imageEdges_, imageDx_, imageDy_);
+
+        imageSize_ = imageEdges_.size();
+
+        posOutBuf_.clear();
+        voteOutBuf_.clear();
 
         processImage();
 
-        if (!posOutBuf.empty())
+        if (!posOutBuf_.empty())
         {
-            if (minDist > 1)
+            if (minDist_ > 1)
                 filterMinDist();
             convertTo(positions, votes);
         }
@@ -140,51 +176,70 @@ namespace
         }
     }
 
-    void GHT_Pos::releaseImpl()
+    void GeneralizedHoughBase::detectImpl(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes)
     {
-        templSize = Size();
-        templCenter = Point(-1, -1);
-        templEdges.release();
-        templDx.release();
-        templDy.release();
+        edges.getMat().copyTo(imageEdges_);
+        dx.getMat().copyTo(imageDx_);
+        dy.getMat().copyTo(imageDy_);
 
-        imageSize = Size();
-        imageEdges.release();
-        imageDx.release();
-        imageDy.release();
+        CV_Assert( imageEdges_.type() == CV_8UC1 );
+        CV_Assert( imageDx_.type() == CV_32FC1 && imageDx_.size() == imageEdges_.size() );
+        CV_Assert( imageDy_.type() == imageDx_.type() && imageDy_.size() == imageEdges_.size() );
 
-        releaseVector(posOutBuf);
-        releaseVector(voteOutBuf);
+        imageSize_ = imageEdges_.size();
+
+        posOutBuf_.clear();
+        voteOutBuf_.clear();
+
+        processImage();
+
+        if (!posOutBuf_.empty())
+        {
+            if (minDist_ > 1)
+                filterMinDist();
+            convertTo(positions, votes);
+        }
+        else
+        {
+            positions.release();
+            if (votes.needed())
+                votes.release();
+        }
     }
 
-    #define votes_cmp_gt(l1, l2) (aux[l1][0] > aux[l2][0])
-    static CV_IMPLEMENT_QSORT_EX( sortIndexies, size_t, votes_cmp_gt, const Vec3i* )
-
-    void GHT_Pos::filterMinDist()
+    class Vec3iGreaterThanIdx
     {
-        size_t oldSize = posOutBuf.size();
-        const bool hasVotes = !voteOutBuf.empty();
+    public:
+        Vec3iGreaterThanIdx( const Vec3i* _arr ) : arr(_arr) {}
+        bool operator()(size_t a, size_t b) const { return arr[a][0] > arr[b][0]; }
+        const Vec3i* arr;
+    };
 
-        CV_Assert(!hasVotes || voteOutBuf.size() == oldSize);
+    void GeneralizedHoughBase::filterMinDist()
+    {
+        size_t oldSize = posOutBuf_.size();
+        const bool hasVotes = !voteOutBuf_.empty();
 
-        std::vector<Vec4f> oldPosBuf(posOutBuf);
-        std::vector<Vec3i> oldVoteBuf(voteOutBuf);
+        CV_Assert( !hasVotes || voteOutBuf_.size() == oldSize );
+
+        std::vector<Vec4f> oldPosBuf(posOutBuf_);
+        std::vector<Vec3i> oldVoteBuf(voteOutBuf_);
 
         std::vector<size_t> indexies(oldSize);
         for (size_t i = 0; i < oldSize; ++i)
             indexies[i] = i;
-        sortIndexies(&indexies[0], oldSize, &oldVoteBuf[0]);
+        std::sort(indexies.begin(), indexies.end(), Vec3iGreaterThanIdx(&oldVoteBuf[0]));
 
-        posOutBuf.clear();
-        voteOutBuf.clear();
+        posOutBuf_.clear();
+        voteOutBuf_.clear();
 
-        const int cellSize = cvRound(minDist);
-        const int gridWidth = (imageSize.width + cellSize - 1) / cellSize;
-        const int gridHeight = (imageSize.height + cellSize - 1) / cellSize;
+        const int cellSize = cvRound(minDist_);
+        const int gridWidth = (imageSize_.width + cellSize - 1) / cellSize;
+        const int gridHeight = (imageSize_.height + cellSize - 1) / cellSize;
 
         std::vector< std::vector<Point2f> > grid(gridWidth * gridHeight);
 
-        const double minDist2 = minDist * minDist;
+        const double minDist2 = minDist_ * minDist_;
 
         for (size_t i = 0; i < oldSize; ++i)
         {
@@ -233,108 +288,112 @@ namespace
             {
                 grid[yCell * gridWidth + xCell].push_back(p);
 
-                posOutBuf.push_back(oldPosBuf[ind]);
+                posOutBuf_.push_back(oldPosBuf[ind]);
                 if (hasVotes)
-                    voteOutBuf.push_back(oldVoteBuf[ind]);
+                    voteOutBuf_.push_back(oldVoteBuf[ind]);
             }
         }
     }
 
-    void GHT_Pos::convertTo(OutputArray _positions, OutputArray _votes)
+    void GeneralizedHoughBase::convertTo(OutputArray _positions, OutputArray _votes)
     {
-        const int total = static_cast<int>(posOutBuf.size());
-        const bool hasVotes = !voteOutBuf.empty();
+        const int total = static_cast<int>(posOutBuf_.size());
+        const bool hasVotes = !voteOutBuf_.empty();
 
-        CV_Assert(!hasVotes || voteOutBuf.size() == posOutBuf.size());
+        CV_Assert( !hasVotes || voteOutBuf_.size() == posOutBuf_.size() );
 
         _positions.create(1, total, CV_32FC4);
         Mat positions = _positions.getMat();
-        Mat(1, total, CV_32FC4, &posOutBuf[0]).copyTo(positions);
+        Mat(1, total, CV_32FC4, &posOutBuf_[0]).copyTo(positions);
 
         if (_votes.needed())
         {
             if (!hasVotes)
+            {
                 _votes.release();
+            }
             else
             {
                 _votes.create(1, total, CV_32SC3);
                 Mat votes = _votes.getMat();
-                Mat(1, total, CV_32SC3, &voteOutBuf[0]).copyTo(votes);
+                Mat(1, total, CV_32SC3, &voteOutBuf_[0]).copyTo(votes);
             }
         }
     }
+}
 
-    /////////////////////////////////////
-    // POSITION Ballard
+// GeneralizedHoughBallard
 
-    class GHT_Ballard_Pos : public GHT_Pos
+namespace
+{
+    class GeneralizedHoughBallardImpl : public GeneralizedHoughBallard, private GeneralizedHoughBase
     {
     public:
-        AlgorithmInfo* info() const;
+        GeneralizedHoughBallardImpl();
 
-        GHT_Ballard_Pos();
+        void setTemplate(InputArray templ, Point templCenter) { setTemplateImpl(templ, templCenter); }
+        void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter) { setTemplateImpl(edges, dx, dy, templCenter); }
 
-    protected:
-        void releaseImpl();
+        void detect(InputArray image, OutputArray positions, OutputArray votes) { detectImpl(image, positions, votes); }
+        void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes) { detectImpl(edges, dx, dy, positions, votes); }
 
+        void setCannyLowThresh(int cannyLowThresh) { cannyLowThresh_ = cannyLowThresh; }
+        int getCannyLowThresh() const { return cannyLowThresh_; }
+
+        void setCannyHighThresh(int cannyHighThresh) { cannyHighThresh_ = cannyHighThresh; }
+        int getCannyHighThresh() const { return cannyHighThresh_; }
+
+        void setMinDist(double minDist) { minDist_ = minDist; }
+        double getMinDist() const { return minDist_; }
+
+        void setDp(double dp) { dp_ = dp; }
+        double getDp() const { return dp_; }
+
+        void setMaxBufferSize(int) {  }
+        int getMaxBufferSize() const { return 0; }
+
+        void setLevels(int levels) { levels_ = levels; }
+        int getLevels() const { return levels_; }
+
+        void setVotesThreshold(int votesThreshold) { votesThreshold_ = votesThreshold; }
+        int getVotesThreshold() const { return votesThreshold_; }
+
+    private:
         void processTempl();
         void processImage();
 
-        virtual void calcHist();
-        virtual void findPosInHist();
+        void calcHist();
+        void findPosInHist();
 
-        int levels;
-        int votesThreshold;
-        double dp;
+        int levels_;
+        int votesThreshold_;
 
-        std::vector< std::vector<Point> > r_table;
-        Mat hist;
+        std::vector< std::vector<Point> > r_table_;
+        Mat hist_;
     };
 
-    CV_INIT_ALGORITHM(GHT_Ballard_Pos, "GeneralizedHough.POSITION",
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution."));
-
-    GHT_Ballard_Pos::GHT_Ballard_Pos()
+    GeneralizedHoughBallardImpl::GeneralizedHoughBallardImpl()
     {
-        levels = 360;
-        votesThreshold = 100;
-        dp = 1.0;
+        levels_ = 360;
+        votesThreshold_ = 100;
     }
 
-    void GHT_Ballard_Pos::releaseImpl()
+    void GeneralizedHoughBallardImpl::processTempl()
     {
-        GHT_Pos::releaseImpl();
+        CV_Assert( levels_ > 0 );
 
-        releaseVector(r_table);
-        hist.release();
-    }
+        const double thetaScale = levels_ / 360.0;
 
-    void GHT_Ballard_Pos::processTempl()
-    {
-        CV_Assert(templEdges.type() == CV_8UC1);
-        CV_Assert(templDx.type() == CV_32FC1 && templDx.size() == templSize);
-        CV_Assert(templDy.type() == templDx.type() && templDy.size() == templSize);
-        CV_Assert(levels > 0);
+        r_table_.resize(levels_ + 1);
+        std::for_each(r_table_.begin(), r_table_.end(), std::mem_fun_ref(&std::vector<Point>::clear));
 
-        const double thetaScale = levels / 360.0;
-
-        r_table.resize(levels + 1);
-        for_each(r_table.begin(), r_table.end(), mem_fun_ref(&std::vector<Point>::clear));
-
-        for (int y = 0; y < templSize.height; ++y)
+        for (int y = 0; y < templSize_.height; ++y)
         {
-            const uchar* edgesRow = templEdges.ptr(y);
-            const float* dxRow = templDx.ptr<float>(y);
-            const float* dyRow = templDy.ptr<float>(y);
+            const uchar* edgesRow = templEdges_.ptr(y);
+            const float* dxRow = templDx_.ptr<float>(y);
+            const float* dyRow = templDy_.ptr<float>(y);
 
-            for (int x = 0; x < templSize.width; ++x)
+            for (int x = 0; x < templSize_.width; ++x)
             {
                 const Point p(x, y);
 
@@ -342,42 +401,42 @@ namespace
                 {
                     const float theta = fastAtan2(dyRow[x], dxRow[x]);
                     const int n = cvRound(theta * thetaScale);
-                    r_table[n].push_back(p - templCenter);
+                    r_table_[n].push_back(p - templCenter_);
                 }
             }
         }
     }
 
-    void GHT_Ballard_Pos::processImage()
+    void GeneralizedHoughBallardImpl::processImage()
     {
         calcHist();
         findPosInHist();
     }
 
-    void GHT_Ballard_Pos::calcHist()
+    void GeneralizedHoughBallardImpl::calcHist()
     {
-        CV_Assert(imageEdges.type() == CV_8UC1);
-        CV_Assert(imageDx.type() == CV_32FC1 && imageDx.size() == imageSize);
-        CV_Assert(imageDy.type() == imageDx.type() && imageDy.size() == imageSize);
-        CV_Assert(levels > 0 && r_table.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(dp > 0.0);
+        CV_Assert( imageEdges_.type() == CV_8UC1 );
+        CV_Assert( imageDx_.type() == CV_32FC1 && imageDx_.size() == imageSize_);
+        CV_Assert( imageDy_.type() == imageDx_.type() && imageDy_.size() == imageSize_);
+        CV_Assert( levels_ > 0 && r_table_.size() == static_cast<size_t>(levels_ + 1) );
+        CV_Assert( dp_ > 0.0 );
 
-        const double thetaScale = levels / 360.0;
-        const double idp = 1.0 / dp;
+        const double thetaScale = levels_ / 360.0;
+        const double idp = 1.0 / dp_;
 
-        hist.create(cvCeil(imageSize.height * idp) + 2, cvCeil(imageSize.width * idp) + 2, CV_32SC1);
-        hist.setTo(0);
+        hist_.create(cvCeil(imageSize_.height * idp) + 2, cvCeil(imageSize_.width * idp) + 2, CV_32SC1);
+        hist_.setTo(0);
 
-        const int rows = hist.rows - 2;
-        const int cols = hist.cols - 2;
+        const int rows = hist_.rows - 2;
+        const int cols = hist_.cols - 2;
 
-        for (int y = 0; y < imageSize.height; ++y)
+        for (int y = 0; y < imageSize_.height; ++y)
         {
-            const uchar* edgesRow = imageEdges.ptr(y);
-            const float* dxRow = imageDx.ptr<float>(y);
-            const float* dyRow = imageDy.ptr<float>(y);
+            const uchar* edgesRow = imageEdges_.ptr(y);
+            const float* dxRow = imageDx_.ptr<float>(y);
+            const float* dyRow = imageDy_.ptr<float>(y);
 
-            for (int x = 0; x < imageSize.width; ++x)
+            for (int x = 0; x < imageSize_.width; ++x)
             {
                 const Point p(x, y);
 
@@ -386,7 +445,7 @@ namespace
                     const float theta = fastAtan2(dyRow[x], dxRow[x]);
                     const int n = cvRound(theta * thetaScale);
 
-                    const std::vector<Point>& r_row = r_table[n];
+                    const std::vector<Point>& r_row = r_table_[n];
 
                     for (size_t j = 0; j < r_row.size(); ++j)
                     {
@@ -396,407 +455,132 @@ namespace
                         c.y = cvRound(c.y * idp);
 
                         if (c.x >= 0 && c.x < cols && c.y >= 0 && c.y < rows)
-                            ++hist.at<int>(c.y + 1, c.x + 1);
+                            ++hist_.at<int>(c.y + 1, c.x + 1);
                     }
                 }
             }
         }
     }
 
-    void GHT_Ballard_Pos::findPosInHist()
+    void GeneralizedHoughBallardImpl::findPosInHist()
     {
-        CV_Assert(votesThreshold > 0);
+        CV_Assert( votesThreshold_ > 0 );
 
-        const int histRows = hist.rows - 2;
-        const int histCols = hist.cols - 2;
+        const int histRows = hist_.rows - 2;
+        const int histCols = hist_.cols - 2;
 
         for(int y = 0; y < histRows; ++y)
         {
-            const int* prevRow = hist.ptr<int>(y);
-            const int* curRow = hist.ptr<int>(y + 1);
-            const int* nextRow = hist.ptr<int>(y + 2);
+            const int* prevRow = hist_.ptr<int>(y);
+            const int* curRow = hist_.ptr<int>(y + 1);
+            const int* nextRow = hist_.ptr<int>(y + 2);
 
             for(int x = 0; x < histCols; ++x)
             {
                 const int votes = curRow[x + 1];
 
-                if (votes > votesThreshold && votes > curRow[x] && votes >= curRow[x + 2] && votes > prevRow[x + 1] && votes >= nextRow[x + 1])
+                if (votes > votesThreshold_ && votes > curRow[x] && votes >= curRow[x + 2] && votes > prevRow[x + 1] && votes >= nextRow[x + 1])
                 {
-                    posOutBuf.push_back(Vec4f(static_cast<float>(x * dp), static_cast<float>(y * dp), 1.0f, 0.0f));
-                    voteOutBuf.push_back(Vec3i(votes, 0, 0));
+                    posOutBuf_.push_back(Vec4f(static_cast<float>(x * dp_), static_cast<float>(y * dp_), 1.0f, 0.0f));
+                    voteOutBuf_.push_back(Vec3i(votes, 0, 0));
                 }
             }
         }
     }
+}
 
-    /////////////////////////////////////
-    // POSITION & SCALE
+Ptr<GeneralizedHoughBallard> cv::createGeneralizedHoughBallard()
+{
+    return makePtr<GeneralizedHoughBallardImpl>();
+}
 
-    class GHT_Ballard_PosScale : public GHT_Ballard_Pos
+// GeneralizedHoughGuil
+
+namespace
+{
+    class GeneralizedHoughGuilImpl : public GeneralizedHoughGuil, private GeneralizedHoughBase
     {
     public:
-        AlgorithmInfo* info() const;
+        GeneralizedHoughGuilImpl();
 
-        GHT_Ballard_PosScale();
+        void setTemplate(InputArray templ, Point templCenter) { setTemplateImpl(templ, templCenter); }
+        void setTemplate(InputArray edges, InputArray dx, InputArray dy, Point templCenter) { setTemplateImpl(edges, dx, dy, templCenter); }
 
-    protected:
-        void calcHist();
-        void findPosInHist();
+        void detect(InputArray image, OutputArray positions, OutputArray votes) { detectImpl(image, positions, votes); }
+        void detect(InputArray edges, InputArray dx, InputArray dy, OutputArray positions, OutputArray votes) { detectImpl(edges, dx, dy, positions, votes); }
 
-        double minScale;
-        double maxScale;
-        double scaleStep;
+        void setCannyLowThresh(int cannyLowThresh) { cannyLowThresh_ = cannyLowThresh; }
+        int getCannyLowThresh() const { return cannyLowThresh_; }
 
-        class Worker;
-        friend class Worker;
-    };
+        void setCannyHighThresh(int cannyHighThresh) { cannyHighThresh_ = cannyHighThresh; }
+        int getCannyHighThresh() const { return cannyHighThresh_; }
 
-    CV_INIT_ALGORITHM(GHT_Ballard_PosScale, "GeneralizedHough.POSITION_SCALE",
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "minScale", obj.minScale, false, 0, 0,
-                                           "Minimal scale to detect.");
-                      obj.info()->addParam(obj, "maxScale", obj.maxScale, false, 0, 0,
-                                           "Maximal scale to detect.");
-                      obj.info()->addParam(obj, "scaleStep", obj.scaleStep, false, 0, 0,
-                                           "Scale step."));
+        void setMinDist(double minDist) { minDist_ = minDist; }
+        double getMinDist() const { return minDist_; }
 
-    GHT_Ballard_PosScale::GHT_Ballard_PosScale()
-    {
-        minScale = 0.5;
-        maxScale = 2.0;
-        scaleStep = 0.05;
-    }
+        void setDp(double dp) { dp_ = dp; }
+        double getDp() const { return dp_; }
 
-    class GHT_Ballard_PosScale::Worker : public ParallelLoopBody
-    {
-    public:
-        explicit Worker(GHT_Ballard_PosScale* base_) : base(base_) {}
+        void setMaxBufferSize(int maxBufferSize) { maxBufferSize_ = maxBufferSize; }
+        int getMaxBufferSize() const { return maxBufferSize_; }
 
-        void operator ()(const Range& range) const;
+        void setXi(double xi) { xi_ = xi; }
+        double getXi() const { return xi_; }
+
+        void setLevels(int levels) { levels_ = levels; }
+        int getLevels() const { return levels_; }
+
+        void setAngleEpsilon(double angleEpsilon) { angleEpsilon_ = angleEpsilon; }
+        double getAngleEpsilon() const { return angleEpsilon_; }
+
+        void setMinAngle(double minAngle) { minAngle_ = minAngle; }
+        double getMinAngle() const { return minAngle_; }
+
+        void setMaxAngle(double maxAngle) { maxAngle_ = maxAngle; }
+        double getMaxAngle() const { return maxAngle_; }
+
+        void setAngleStep(double angleStep) { angleStep_ = angleStep; }
+        double getAngleStep() const { return angleStep_; }
+
+        void setAngleThresh(int angleThresh) { angleThresh_ = angleThresh; }
+        int getAngleThresh() const { return angleThresh_; }
+
+        void setMinScale(double minScale) { minScale_ = minScale; }
+        double getMinScale() const { return minScale_; }
+
+        void setMaxScale(double maxScale) { maxScale_ = maxScale; }
+        double getMaxScale() const { return maxScale_; }
+
+        void setScaleStep(double scaleStep) { scaleStep_ = scaleStep; }
+        double getScaleStep() const { return scaleStep_; }
+
+        void setScaleThresh(int scaleThresh) { scaleThresh_ = scaleThresh; }
+        int getScaleThresh() const { return scaleThresh_; }
+
+        void setPosThresh(int posThresh) { posThresh_ = posThresh; }
+        int getPosThresh() const { return posThresh_; }
 
     private:
-        GHT_Ballard_PosScale* base;
-    };
-
-    void GHT_Ballard_PosScale::Worker::operator ()(const Range& range) const
-    {
-        const double thetaScale = base->levels / 360.0;
-        const double idp = 1.0 / base->dp;
-
-        for (int s = range.start; s < range.end; ++s)
-        {
-            const double scale = base->minScale + s * base->scaleStep;
-
-            Mat curHist(base->hist.size[1], base->hist.size[2], CV_32SC1, base->hist.ptr(s + 1), base->hist.step[1]);
-
-            for (int y = 0; y < base->imageSize.height; ++y)
-            {
-                const uchar* edgesRow = base->imageEdges.ptr(y);
-                const float* dxRow = base->imageDx.ptr<float>(y);
-                const float* dyRow = base->imageDy.ptr<float>(y);
-
-                for (int x = 0; x < base->imageSize.width; ++x)
-                {
-                    const Point2d p(x, y);
-
-                    if (edgesRow[x] && (notNull(dyRow[x]) || notNull(dxRow[x])))
-                    {
-                        const float theta = fastAtan2(dyRow[x], dxRow[x]);
-                        const int n = cvRound(theta * thetaScale);
-
-                        const std::vector<Point>& r_row = base->r_table[n];
-
-                        for (size_t j = 0; j < r_row.size(); ++j)
-                        {
-                            Point2d d = r_row[j];
-                            Point2d c = p - d * scale;
-
-                            c.x *= idp;
-                            c.y *= idp;
-
-                            if (c.x >= 0 && c.x < base->hist.size[2] - 2 && c.y >= 0 && c.y < base->hist.size[1] - 2)
-                                ++curHist.at<int>(cvRound(c.y + 1), cvRound(c.x + 1));
-                        }
-                    }
-                }
-            }
-        }
-    }
-
-    void GHT_Ballard_PosScale::calcHist()
-    {
-        CV_Assert(imageEdges.type() == CV_8UC1);
-        CV_Assert(imageDx.type() == CV_32FC1 && imageDx.size() == imageSize);
-        CV_Assert(imageDy.type() == imageDx.type() && imageDy.size() == imageSize);
-        CV_Assert(levels > 0 && r_table.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(dp > 0.0);
-        CV_Assert(minScale > 0.0 && minScale < maxScale);
-        CV_Assert(scaleStep > 0.0);
-
-        const double idp = 1.0 / dp;
-        const int scaleRange = cvCeil((maxScale - minScale) / scaleStep);
-
-        const int sizes[] = {scaleRange + 2, cvCeil(imageSize.height * idp) + 2, cvCeil(imageSize.width * idp) + 2};
-        hist.create(3, sizes, CV_32SC1);
-        hist.setTo(0);
-
-        parallel_for_(Range(0, scaleRange), Worker(this));
-    }
-
-    void GHT_Ballard_PosScale::findPosInHist()
-    {
-        CV_Assert(votesThreshold > 0);
-
-        const int scaleRange = hist.size[0] - 2;
-        const int histRows = hist.size[1] - 2;
-        const int histCols = hist.size[2] - 2;
-
-        for (int s = 0; s < scaleRange; ++s)
-        {
-            const float scale = static_cast<float>(minScale + s * scaleStep);
-
-            const Mat prevHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(s), hist.step[1]);
-            const Mat curHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(s + 1), hist.step[1]);
-            const Mat nextHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(s + 2), hist.step[1]);
-
-            for(int y = 0; y < histRows; ++y)
-            {
-                const int* prevHistRow = prevHist.ptr<int>(y + 1);
-                const int* prevRow = curHist.ptr<int>(y);
-                const int* curRow = curHist.ptr<int>(y + 1);
-                const int* nextRow = curHist.ptr<int>(y + 2);
-                const int* nextHistRow = nextHist.ptr<int>(y + 1);
-
-                for(int x = 0; x < histCols; ++x)
-                {
-                    const int votes = curRow[x + 1];
-
-                    if (votes > votesThreshold &&
-                        votes > curRow[x] &&
-                        votes >= curRow[x + 2] &&
-                        votes > prevRow[x + 1] &&
-                        votes >= nextRow[x + 1] &&
-                        votes > prevHistRow[x + 1] &&
-                        votes >= nextHistRow[x + 1])
-                    {
-                        posOutBuf.push_back(Vec4f(static_cast<float>(x * dp), static_cast<float>(y * dp), scale, 0.0f));
-                        voteOutBuf.push_back(Vec3i(votes, votes, 0));
-                    }
-                }
-            }
-        }
-    }
-
-    /////////////////////////////////////
-    // POSITION & ROTATION
-
-    class GHT_Ballard_PosRotation : public GHT_Ballard_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Ballard_PosRotation();
-
-    protected:
-        void calcHist();
-        void findPosInHist();
-
-        double minAngle;
-        double maxAngle;
-        double angleStep;
-
-        class Worker;
-        friend class Worker;
-    };
-
-    CV_INIT_ALGORITHM(GHT_Ballard_PosRotation, "GeneralizedHough.POSITION_ROTATION",
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "R-Table levels.");
-                      obj.info()->addParam(obj, "votesThreshold", obj.votesThreshold, false, 0, 0,
-                                           "The accumulator threshold for the template centers at the detection stage. The smaller it is, the more false positions may be detected.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "minAngle", obj.minAngle, false, 0, 0,
-                                           "Minimal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "maxAngle", obj.maxAngle, false, 0, 0,
-                                           "Maximal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "angleStep", obj.angleStep, false, 0, 0,
-                                           "Angle step in degrees."));
-
-    GHT_Ballard_PosRotation::GHT_Ballard_PosRotation()
-    {
-        minAngle = 0.0;
-        maxAngle = 360.0;
-        angleStep = 1.0;
-    }
-
-    class GHT_Ballard_PosRotation::Worker : public ParallelLoopBody
-    {
-    public:
-        explicit Worker(GHT_Ballard_PosRotation* base_) : base(base_) {}
-
-        void operator ()(const Range& range) const;
-
-    private:
-        GHT_Ballard_PosRotation* base;
-    };
-
-    void GHT_Ballard_PosRotation::Worker::operator ()(const Range& range) const
-    {
-        const double thetaScale = base->levels / 360.0;
-        const double idp = 1.0 / base->dp;
-
-        for (int a = range.start; a < range.end; ++a)
-        {
-            const double angle = base->minAngle + a * base->angleStep;
-
-            const double sinA = ::sin(toRad(angle));
-            const double cosA = ::cos(toRad(angle));
-
-            Mat curHist(base->hist.size[1], base->hist.size[2], CV_32SC1, base->hist.ptr(a + 1), base->hist.step[1]);
-
-            for (int y = 0; y < base->imageSize.height; ++y)
-            {
-                const uchar* edgesRow = base->imageEdges.ptr(y);
-                const float* dxRow = base->imageDx.ptr<float>(y);
-                const float* dyRow = base->imageDy.ptr<float>(y);
-
-                for (int x = 0; x < base->imageSize.width; ++x)
-                {
-                    const Point2d p(x, y);
-
-                    if (edgesRow[x] && (notNull(dyRow[x]) || notNull(dxRow[x])))
-                    {
-                        double theta = fastAtan2(dyRow[x], dxRow[x]) - angle;
-                        if (theta < 0)
-                            theta += 360.0;
-                        const int n = cvRound(theta * thetaScale);
-
-                        const std::vector<Point>& r_row = base->r_table[n];
-
-                        for (size_t j = 0; j < r_row.size(); ++j)
-                        {
-                            Point2d d = r_row[j];
-                            Point2d c = p - Point2d(d.x * cosA - d.y * sinA, d.x * sinA + d.y * cosA);
-
-                            c.x *= idp;
-                            c.y *= idp;
-
-                            if (c.x >= 0 && c.x < base->hist.size[2] - 2 && c.y >= 0 && c.y < base->hist.size[1] - 2)
-                                ++curHist.at<int>(cvRound(c.y + 1), cvRound(c.x + 1));
-                        }
-                    }
-                }
-            }
-        }
-    }
-
-    void GHT_Ballard_PosRotation::calcHist()
-    {
-        CV_Assert(imageEdges.type() == CV_8UC1);
-        CV_Assert(imageDx.type() == CV_32FC1 && imageDx.size() == imageSize);
-        CV_Assert(imageDy.type() == imageDx.type() && imageDy.size() == imageSize);
-        CV_Assert(levels > 0 && r_table.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(dp > 0.0);
-        CV_Assert(minAngle >= 0.0 && minAngle < maxAngle && maxAngle <= 360.0);
-        CV_Assert(angleStep > 0.0 && angleStep < 360.0);
-
-        const double idp = 1.0 / dp;
-        const int angleRange = cvCeil((maxAngle - minAngle) / angleStep);
-
-        const int sizes[] = {angleRange + 2, cvCeil(imageSize.height * idp) + 2, cvCeil(imageSize.width * idp) + 2};
-        hist.create(3, sizes, CV_32SC1);
-        hist.setTo(0);
-
-        parallel_for_(Range(0, angleRange), Worker(this));
-    }
-
-    void GHT_Ballard_PosRotation::findPosInHist()
-    {
-        CV_Assert(votesThreshold > 0);
-
-        const int angleRange = hist.size[0] - 2;
-        const int histRows = hist.size[1] - 2;
-        const int histCols = hist.size[2] - 2;
-
-        for (int a = 0; a < angleRange; ++a)
-        {
-            const float angle = static_cast<float>(minAngle + a * angleStep);
-
-            const Mat prevHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(a), hist.step[1]);
-            const Mat curHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(a + 1), hist.step[1]);
-            const Mat nextHist(histRows + 2, histCols + 2, CV_32SC1, hist.ptr(a + 2), hist.step[1]);
-
-            for(int y = 0; y < histRows; ++y)
-            {
-                const int* prevHistRow = prevHist.ptr<int>(y + 1);
-                const int* prevRow = curHist.ptr<int>(y);
-                const int* curRow = curHist.ptr<int>(y + 1);
-                const int* nextRow = curHist.ptr<int>(y + 2);
-                const int* nextHistRow = nextHist.ptr<int>(y + 1);
-
-                for(int x = 0; x < histCols; ++x)
-                {
-                    const int votes = curRow[x + 1];
-
-                    if (votes > votesThreshold &&
-                        votes > curRow[x] &&
-                        votes >= curRow[x + 2] &&
-                        votes > prevRow[x + 1] &&
-                        votes >= nextRow[x + 1] &&
-                        votes > prevHistRow[x + 1] &&
-                        votes >= nextHistRow[x + 1])
-                    {
-                        posOutBuf.push_back(Vec4f(static_cast<float>(x * dp), static_cast<float>(y * dp), 1.0f, angle));
-                        voteOutBuf.push_back(Vec3i(votes, 0, votes));
-                    }
-                }
-            }
-        }
-    }
-
-    /////////////////////////////////////////
-    // POSITION & SCALE & ROTATION
-
-    double clampAngle(double a)
-    {
-        double res = a;
-
-        while (res > 360.0)
-            res -= 360.0;
-        while (res < 0)
-            res += 360.0;
-
-        return res;
-    }
-
-    bool angleEq(double a, double b, double eps = 1.0)
-    {
-        return (fabs(clampAngle(a - b)) <= eps);
-    }
-
-    class GHT_Guil_Full : public GHT_Pos
-    {
-    public:
-        AlgorithmInfo* info() const;
-
-        GHT_Guil_Full();
-
-    protected:
-        void releaseImpl();
-
         void processTempl();
         void processImage();
 
+        int maxBufferSize_;
+        double xi_;
+        int levels_;
+        double angleEpsilon_;
+
+        double minAngle_;
+        double maxAngle_;
+        double angleStep_;
+        int angleThresh_;
+
+        double minScale_;
+        double maxScale_;
+        double scaleStep_;
+        int scaleThresh_;
+
+        int posThresh_;
+
         struct ContourPoint
         {
             Point2d pos;
@@ -822,137 +606,92 @@ namespace
         void calcScale(double angle);
         void calcPosition(double angle, int angleVotes, double scale, int scaleVotes);
 
-        int maxSize;
-        double xi;
-        int levels;
-        double angleEpsilon;
+        std::vector< std::vector<Feature> > templFeatures_;
+        std::vector< std::vector<Feature> > imageFeatures_;
 
-        double minAngle;
-        double maxAngle;
-        double angleStep;
-        int angleThresh;
-
-        double minScale;
-        double maxScale;
-        double scaleStep;
-        int scaleThresh;
-
-        double dp;
-        int posThresh;
-
-        std::vector< std::vector<Feature> > templFeatures;
-        std::vector< std::vector<Feature> > imageFeatures;
-
-        std::vector< std::pair<double, int> > angles;
-        std::vector< std::pair<double, int> > scales;
+        std::vector< std::pair<double, int> > angles_;
+        std::vector< std::pair<double, int> > scales_;
     };
 
-    CV_INIT_ALGORITHM(GHT_Guil_Full, "GeneralizedHough.POSITION_SCALE_ROTATION",
-                      obj.info()->addParam(obj, "minDist", obj.minDist, false, 0, 0,
-                                           "Minimum distance between the centers of the detected objects.");
-                      obj.info()->addParam(obj, "maxSize", obj.maxSize, false, 0, 0,
-                                           "Maximal size of inner buffers.");
-                      obj.info()->addParam(obj, "xi", obj.xi, false, 0, 0,
-                                           "Angle difference in degrees between two points in feature.");
-                      obj.info()->addParam(obj, "levels", obj.levels, false, 0, 0,
-                                           "Feature table levels.");
-                      obj.info()->addParam(obj, "angleEpsilon", obj.angleEpsilon, false, 0, 0,
-                                           "Maximal difference between angles that treated as equal.");
-                      obj.info()->addParam(obj, "minAngle", obj.minAngle, false, 0, 0,
-                                           "Minimal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "maxAngle", obj.maxAngle, false, 0, 0,
-                                           "Maximal rotation angle to detect in degrees.");
-                      obj.info()->addParam(obj, "angleStep", obj.angleStep, false, 0, 0,
-                                           "Angle step in degrees.");
-                      obj.info()->addParam(obj, "angleThresh", obj.angleThresh, false, 0, 0,
-                                           "Angle threshold.");
-                      obj.info()->addParam(obj, "minScale", obj.minScale, false, 0, 0,
-                                           "Minimal scale to detect.");
-                      obj.info()->addParam(obj, "maxScale", obj.maxScale, false, 0, 0,
-                                           "Maximal scale to detect.");
-                      obj.info()->addParam(obj, "scaleStep", obj.scaleStep, false, 0, 0,
-                                           "Scale step.");
-                      obj.info()->addParam(obj, "scaleThresh", obj.scaleThresh, false, 0, 0,
-                                           "Scale threshold.");
-                      obj.info()->addParam(obj, "dp", obj.dp, false, 0, 0,
-                                           "Inverse ratio of the accumulator resolution to the image resolution.");
-                      obj.info()->addParam(obj, "posThresh", obj.posThresh, false, 0, 0,
-                                           "Position threshold."));
-
-    GHT_Guil_Full::GHT_Guil_Full()
+    double clampAngle(double a)
     {
-        maxSize = 1000;
-        xi = 90.0;
-        levels = 360;
-        angleEpsilon = 1.0;
+        double res = a;
 
-        minAngle = 0.0;
-        maxAngle = 360.0;
-        angleStep = 1.0;
-        angleThresh = 15000;
+        while (res > 360.0)
+            res -= 360.0;
+        while (res < 0)
+            res += 360.0;
 
-        minScale = 0.5;
-        maxScale = 2.0;
-        scaleStep = 0.05;
-        scaleThresh = 1000;
-
-        dp = 1.0;
-        posThresh = 100;
+        return res;
     }
 
-    void GHT_Guil_Full::releaseImpl()
+    bool angleEq(double a, double b, double eps = 1.0)
     {
-        GHT_Pos::releaseImpl();
-
-        releaseVector(templFeatures);
-        releaseVector(imageFeatures);
-
-        releaseVector(angles);
-        releaseVector(scales);
+        return (fabs(clampAngle(a - b)) <= eps);
     }
 
-    void GHT_Guil_Full::processTempl()
+    GeneralizedHoughGuilImpl::GeneralizedHoughGuilImpl()
     {
-        buildFeatureList(templEdges, templDx, templDy, templFeatures, templCenter);
+        maxBufferSize_ = 1000;
+        xi_ = 90.0;
+        levels_ = 360;
+        angleEpsilon_ = 1.0;
+
+        minAngle_ = 0.0;
+        maxAngle_ = 360.0;
+        angleStep_ = 1.0;
+        angleThresh_ = 15000;
+
+        minScale_ = 0.5;
+        maxScale_ = 2.0;
+        scaleStep_ = 0.05;
+        scaleThresh_ = 1000;
+
+        posThresh_ = 100;
     }
 
-    void GHT_Guil_Full::processImage()
+    void GeneralizedHoughGuilImpl::processTempl()
     {
-        buildFeatureList(imageEdges, imageDx, imageDy, imageFeatures);
+        buildFeatureList(templEdges_, templDx_, templDy_, templFeatures_, templCenter_);
+    }
+
+    void GeneralizedHoughGuilImpl::processImage()
+    {
+        buildFeatureList(imageEdges_, imageDx_, imageDy_, imageFeatures_);
 
         calcOrientation();
 
-        for (size_t i = 0; i < angles.size(); ++i)
+        for (size_t i = 0; i < angles_.size(); ++i)
         {
-            const double angle = angles[i].first;
-            const int angleVotes = angles[i].second;
+            const double angle = angles_[i].first;
+            const int angleVotes = angles_[i].second;
 
             calcScale(angle);
 
-            for (size_t j = 0; j < scales.size(); ++j)
+            for (size_t j = 0; j < scales_.size(); ++j)
             {
-                const double scale = scales[j].first;
-                const int scaleVotes = scales[j].second;
+                const double scale = scales_[j].first;
+                const int scaleVotes = scales_[j].second;
 
                 calcPosition(angle, angleVotes, scale, scaleVotes);
             }
         }
     }
 
-    void GHT_Guil_Full::buildFeatureList(const Mat& edges, const Mat& dx, const Mat& dy, std::vector< std::vector<Feature> >& features, Point2d center)
+    void GeneralizedHoughGuilImpl::buildFeatureList(const Mat& edges, const Mat& dx, const Mat& dy, std::vector< std::vector<Feature> >& features, Point2d center)
     {
-        CV_Assert(levels > 0);
+        CV_Assert( levels_ > 0 );
 
-        const double maxDist = sqrt((double) templSize.width * templSize.width + templSize.height * templSize.height) * maxScale;
+        const double maxDist = sqrt((double) templSize_.width * templSize_.width + templSize_.height * templSize_.height) * maxScale_;
 
-        const double alphaScale = levels / 360.0;
+        const double alphaScale = levels_ / 360.0;
 
         std::vector<ContourPoint> points;
         getContourPoints(edges, dx, dy, points);
 
-        features.resize(levels + 1);
-        for_each(features.begin(), features.end(), mem_fun_ref(&std::vector<Feature>::clear));
-        for_each(features.begin(), features.end(), bind2nd(mem_fun_ref(&std::vector<Feature>::reserve), maxSize));
+        features.resize(levels_ + 1);
+        std::for_each(features.begin(), features.end(), std::mem_fun_ref(&std::vector<Feature>::clear));
+        std::for_each(features.begin(), features.end(), std::bind2nd(std::mem_fun_ref(&std::vector<Feature>::reserve), maxBufferSize_));
 
         for (size_t i = 0; i < points.size(); ++i)
         {
@@ -962,7 +701,7 @@ namespace
             {
                 ContourPoint p2 = points[j];
 
-                if (angleEq(p1.theta - p2.theta, xi, angleEpsilon))
+                if (angleEq(p1.theta - p2.theta, xi_, angleEpsilon_))
                 {
                     const Point2d d = p1.pos - p2.pos;
 
@@ -982,18 +721,18 @@ namespace
 
                     const int n = cvRound(f.alpha12 * alphaScale);
 
-                    if (features[n].size() < static_cast<size_t>(maxSize))
+                    if (features[n].size() < static_cast<size_t>(maxBufferSize_))
                         features[n].push_back(f);
                 }
             }
         }
     }
 
-    void GHT_Guil_Full::getContourPoints(const Mat& edges, const Mat& dx, const Mat& dy, std::vector<ContourPoint>& points)
+    void GeneralizedHoughGuilImpl::getContourPoints(const Mat& edges, const Mat& dx, const Mat& dy, std::vector<ContourPoint>& points)
     {
-        CV_Assert(edges.type() == CV_8UC1);
-        CV_Assert(dx.type() == CV_32FC1 && dx.size == edges.size);
-        CV_Assert(dy.type() == dx.type() && dy.size == edges.size);
+        CV_Assert( edges.type() == CV_8UC1 );
+        CV_Assert( dx.type() == CV_32FC1 && dx.size == edges.size );
+        CV_Assert( dy.type() == dx.type() && dy.size == edges.size );
 
         points.clear();
         points.reserve(edges.size().area());
@@ -1019,23 +758,23 @@ namespace
         }
     }
 
-    void GHT_Guil_Full::calcOrientation()
+    void GeneralizedHoughGuilImpl::calcOrientation()
     {
-        CV_Assert(levels > 0);
-        CV_Assert(templFeatures.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(imageFeatures.size() == templFeatures.size());
-        CV_Assert(minAngle >= 0.0 && minAngle < maxAngle && maxAngle <= 360.0);
-        CV_Assert(angleStep > 0.0 && angleStep < 360.0);
-        CV_Assert(angleThresh > 0);
+        CV_Assert( levels_ > 0 );
+        CV_Assert( templFeatures_.size() == static_cast<size_t>(levels_ + 1) );
+        CV_Assert( imageFeatures_.size() == templFeatures_.size() );
+        CV_Assert( minAngle_ >= 0.0 && minAngle_ < maxAngle_ && maxAngle_ <= 360.0 );
+        CV_Assert( angleStep_ > 0.0 && angleStep_ < 360.0 );
+        CV_Assert( angleThresh_ > 0 );
 
-        const double iAngleStep = 1.0 / angleStep;
-        const int angleRange = cvCeil((maxAngle - minAngle) * iAngleStep);
+        const double iAngleStep = 1.0 / angleStep_;
+        const int angleRange = cvCeil((maxAngle_ - minAngle_) * iAngleStep);
 
         std::vector<int> OHist(angleRange + 1, 0);
-        for (int i = 0; i <= levels; ++i)
+        for (int i = 0; i <= levels_; ++i)
         {
-            const std::vector<Feature>& templRow = templFeatures[i];
-            const std::vector<Feature>& imageRow = imageFeatures[i];
+            const std::vector<Feature>& templRow = templFeatures_[i];
+            const std::vector<Feature>& imageRow = imageFeatures_[i];
 
             for (size_t j = 0; j < templRow.size(); ++j)
             {
@@ -1046,45 +785,45 @@ namespace
                     Feature imF = imageRow[k];
 
                     const double angle = clampAngle(imF.p1.theta - templF.p1.theta);
-                    if (angle >= minAngle && angle <= maxAngle)
+                    if (angle >= minAngle_ && angle <= maxAngle_)
                     {
-                        const int n = cvRound((angle - minAngle) * iAngleStep);
+                        const int n = cvRound((angle - minAngle_) * iAngleStep);
                         ++OHist[n];
                     }
                 }
             }
         }
 
-        angles.clear();
+        angles_.clear();
 
         for (int n = 0; n < angleRange; ++n)
         {
-            if (OHist[n] >= angleThresh)
+            if (OHist[n] >= angleThresh_)
             {
-                const double angle = minAngle + n * angleStep;
-                angles.push_back(std::make_pair(angle, OHist[n]));
+                const double angle = minAngle_ + n * angleStep_;
+                angles_.push_back(std::make_pair(angle, OHist[n]));
             }
         }
     }
 
-    void GHT_Guil_Full::calcScale(double angle)
+    void GeneralizedHoughGuilImpl::calcScale(double angle)
     {
-        CV_Assert(levels > 0);
-        CV_Assert(templFeatures.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(imageFeatures.size() == templFeatures.size());
-        CV_Assert(minScale > 0.0 && minScale < maxScale);
-        CV_Assert(scaleStep > 0.0);
-        CV_Assert(scaleThresh > 0);
+        CV_Assert( levels_ > 0 );
+        CV_Assert( templFeatures_.size() == static_cast<size_t>(levels_ + 1) );
+        CV_Assert( imageFeatures_.size() == templFeatures_.size() );
+        CV_Assert( minScale_ > 0.0 && minScale_ < maxScale_ );
+        CV_Assert( scaleStep_ > 0.0 );
+        CV_Assert( scaleThresh_ > 0 );
 
-        const double iScaleStep = 1.0 / scaleStep;
-        const int scaleRange = cvCeil((maxScale - minScale) * iScaleStep);
+        const double iScaleStep = 1.0 / scaleStep_;
+        const int scaleRange = cvCeil((maxScale_ - minScale_) * iScaleStep);
 
         std::vector<int> SHist(scaleRange + 1, 0);
 
-        for (int i = 0; i <= levels; ++i)
+        for (int i = 0; i <= levels_; ++i)
         {
-            const std::vector<Feature>& templRow = templFeatures[i];
-            const std::vector<Feature>& imageRow = imageFeatures[i];
+            const std::vector<Feature>& templRow = templFeatures_[i];
+            const std::vector<Feature>& imageRow = imageFeatures_[i];
 
             for (size_t j = 0; j < templRow.size(); ++j)
             {
@@ -1096,12 +835,12 @@ namespace
                 {
                     Feature imF = imageRow[k];
 
-                    if (angleEq(imF.p1.theta, templF.p1.theta, angleEpsilon))
+                    if (angleEq(imF.p1.theta, templF.p1.theta, angleEpsilon_))
                     {
                         const double scale = imF.d12 / templF.d12;
-                        if (scale >= minScale && scale <= maxScale)
+                        if (scale >= minScale_ && scale <= maxScale_)
                         {
-                            const int s = cvRound((scale - minScale) * iScaleStep);
+                            const int s = cvRound((scale - minScale_) * iScaleStep);
                             ++SHist[s];
                         }
                     }
@@ -1109,39 +848,39 @@ namespace
             }
         }
 
-        scales.clear();
+        scales_.clear();
 
         for (int s = 0; s < scaleRange; ++s)
         {
-            if (SHist[s] >= scaleThresh)
+            if (SHist[s] >= scaleThresh_)
             {
-                const double scale = minScale + s * scaleStep;
-                scales.push_back(std::make_pair(scale, SHist[s]));
+                const double scale = minScale_ + s * scaleStep_;
+                scales_.push_back(std::make_pair(scale, SHist[s]));
             }
         }
     }
 
-    void GHT_Guil_Full::calcPosition(double angle, int angleVotes, double scale, int scaleVotes)
+    void GeneralizedHoughGuilImpl::calcPosition(double angle, int angleVotes, double scale, int scaleVotes)
     {
-        CV_Assert(levels > 0);
-        CV_Assert(templFeatures.size() == static_cast<size_t>(levels + 1));
-        CV_Assert(imageFeatures.size() == templFeatures.size());
-        CV_Assert(dp > 0.0);
-        CV_Assert(posThresh > 0);
+        CV_Assert( levels_ > 0 );
+        CV_Assert( templFeatures_.size() == static_cast<size_t>(levels_ + 1) );
+        CV_Assert( imageFeatures_.size() == templFeatures_.size() );
+        CV_Assert( dp_ > 0.0 );
+        CV_Assert( posThresh_ > 0 );
 
         const double sinVal = sin(toRad(angle));
         const double cosVal = cos(toRad(angle));
-        const double idp = 1.0 / dp;
+        const double idp = 1.0 / dp_;
 
-        const int histRows = cvCeil(imageSize.height * idp);
-        const int histCols = cvCeil(imageSize.width * idp);
+        const int histRows = cvCeil(imageSize_.height * idp);
+        const int histCols = cvCeil(imageSize_.width * idp);
 
         Mat DHist(histRows + 2, histCols + 2, CV_32SC1, Scalar::all(0));
 
-        for (int i = 0; i <= levels; ++i)
+        for (int i = 0; i <= levels_; ++i)
         {
-            const std::vector<Feature>& templRow = templFeatures[i];
-            const std::vector<Feature>& imageRow = imageFeatures[i];
+            const std::vector<Feature>& templRow = templFeatures_[i];
+            const std::vector<Feature>& imageRow = imageFeatures_[i];
 
             for (size_t j = 0; j < templRow.size(); ++j)
             {
@@ -1159,7 +898,7 @@ namespace
                 {
                     Feature imF = imageRow[k];
 
-                    if (angleEq(imF.p1.theta, templF.p1.theta, angleEpsilon))
+                    if (angleEq(imF.p1.theta, templF.p1.theta, angleEpsilon_))
                     {
                         Point2d c1, c2;
 
@@ -1189,101 +928,17 @@ namespace
             {
                 const int votes = curRow[x + 1];
 
-                if (votes > posThresh && votes > curRow[x] && votes >= curRow[x + 2] && votes > prevRow[x + 1] && votes >= nextRow[x + 1])
+                if (votes > posThresh_ && votes > curRow[x] && votes >= curRow[x + 2] && votes > prevRow[x + 1] && votes >= nextRow[x + 1])
                 {
-                    posOutBuf.push_back(Vec4f(static_cast<float>(x * dp), static_cast<float>(y * dp), static_cast<float>(scale), static_cast<float>(angle)));
-                    voteOutBuf.push_back(Vec3i(votes, scaleVotes, angleVotes));
+                    posOutBuf_.push_back(Vec4f(static_cast<float>(x * dp_), static_cast<float>(y * dp_), static_cast<float>(scale), static_cast<float>(angle)));
+                    voteOutBuf_.push_back(Vec3i(votes, scaleVotes, angleVotes));
                 }
             }
         }
     }
 }
 
-Ptr<GeneralizedHough> cv::GeneralizedHough::create(int method)
+Ptr<GeneralizedHoughGuil> cv::createGeneralizedHoughGuil()
 {
-    switch (method)
-    {
-    case GHT_POSITION:
-        CV_Assert( !GHT_Ballard_Pos_info_auto.name().empty() );
-        return new GHT_Ballard_Pos();
-
-    case (GHT_POSITION | GHT_SCALE):
-        CV_Assert( !GHT_Ballard_PosScale_info_auto.name().empty() );
-        return new GHT_Ballard_PosScale();
-
-    case (GHT_POSITION | GHT_ROTATION):
-        CV_Assert( !GHT_Ballard_PosRotation_info_auto.name().empty() );
-        return new GHT_Ballard_PosRotation();
-
-    case (GHT_POSITION | GHT_SCALE | GHT_ROTATION):
-        CV_Assert( !GHT_Guil_Full_info_auto.name().empty() );
-        return new GHT_Guil_Full();
-    }
-
-    CV_Error(CV_StsBadArg, "Unsupported method");
-    return Ptr<GeneralizedHough>();
-}
-
-cv::GeneralizedHough::~GeneralizedHough()
-{
-}
-
-void cv::GeneralizedHough::setTemplate(InputArray _templ, int cannyThreshold, Point templCenter)
-{
-    Mat templ = _templ.getMat();
-
-    CV_Assert(templ.type() == CV_8UC1);
-    CV_Assert(cannyThreshold > 0);
-
-    Canny(templ, edges_, cannyThreshold / 2, cannyThreshold);
-    Sobel(templ, dx_, CV_32F, 1, 0);
-    Sobel(templ, dy_, CV_32F, 0, 1);
-
-    if (templCenter == Point(-1, -1))
-        templCenter = Point(templ.cols / 2, templ.rows / 2);
-
-    setTemplateImpl(edges_, dx_, dy_, templCenter);
-}
-
-void cv::GeneralizedHough::setTemplate(InputArray _edges, InputArray _dx, InputArray _dy, Point templCenter)
-{
-    Mat edges = _edges.getMat();
-    Mat dx = _dx.getMat();
-    Mat dy = _dy.getMat();
-
-    if (templCenter == Point(-1, -1))
-        templCenter = Point(edges.cols / 2, edges.rows / 2);
-
-    setTemplateImpl(edges, dx, dy, templCenter);
-}
-
-void cv::GeneralizedHough::detect(InputArray _image, OutputArray positions, OutputArray votes, int cannyThreshold)
-{
-    Mat image = _image.getMat();
-
-    CV_Assert(image.type() == CV_8UC1);
-    CV_Assert(cannyThreshold > 0);
-
-    Canny(image, edges_, cannyThreshold / 2, cannyThreshold);
-    Sobel(image, dx_, CV_32F, 1, 0);
-    Sobel(image, dy_, CV_32F, 0, 1);
-
-    detectImpl(edges_, dx_, dy_, positions, votes);
-}
-
-void cv::GeneralizedHough::detect(InputArray _edges, InputArray _dx, InputArray _dy, OutputArray positions, OutputArray votes)
-{
-    cv::Mat edges = _edges.getMat();
-    cv::Mat dx = _dx.getMat();
-    cv::Mat dy = _dy.getMat();
-
-    detectImpl(edges, dx, dy, positions, votes);
-}
-
-void cv::GeneralizedHough::release()
-{
-    edges_.release();
-    dx_.release();
-    dy_.release();
-    releaseImpl();
+    return makePtr<GeneralizedHoughGuilImpl>();
 }
diff --git a/modules/imgproc/src/geometry.cpp b/modules/imgproc/src/geometry.cpp
index 5c529e5b56..5d1fa789b9 100644
--- a/modules/imgproc/src/geometry.cpp
+++ b/modules/imgproc/src/geometry.cpp
@@ -589,124 +589,3 @@ float cv::intersectConvexConvex( InputArray _p1, InputArray _p2, OutputArray _p1
     }
     return (float)fabs(area);
 }
-
-/*
-static void testConvConv()
-{
-    static const int P1[] =
-    {
-        0,	0,
-        100, 0,
-        100, 100,
-        0, 100,
-    };
-
-    static const int Q1[] =
-    {
-        100, 80,
-        50, 80,
-        50, 50,
-        100, 50
-    };
-
-    static const int P2[] =
-    {
-        0, 0,
-        200, 0,
-        200, 100,
-        100, 200,
-        0, 100
-    };
-
-    static const int Q2[] =
-    {
-        100, 100,
-        300, 100,
-        300, 200,
-        100, 200
-    };
-
-    static const int P3[] =
-    {
-        0,   0,
-        100, 0,
-        100, 100,
-        0, 100
-    };
-
-    static const int Q3[] =
-    {
-        50, 50,
-        150, 50,
-        150, 150,
-        50, 150
-    };
-
-    static const int P4[] =
-    {
-        0, 160,
-        50, 80,
-        130, 0,
-        190, 20,
-        240, 100,
-        240, 260,
-        190, 290,
-        130, 320,
-        70, 320,
-        30, 290
-    };
-
-    static const int Q4[] =
-    {
-        160, -30,
-        280, 160,
-        160, 320,
-        0, 220,
-        30, 100
-    };
-
-    static const void* PQs[] =
-    {
-        P1, Q1, P2, Q2, P3, Q3, P4, Q4
-    };
-
-    static const int lens[] =
-    {
-        CV_DIM(P1), CV_DIM(Q1),
-        CV_DIM(P2), CV_DIM(Q2),
-        CV_DIM(P3), CV_DIM(Q3),
-        CV_DIM(P4), CV_DIM(Q4)
-    };
-
-    Mat img(800, 800, CV_8UC3);
-
-    for( int i = 0; i < CV_DIM(PQs)/2; i++ )
-    {
-        Mat Pm = Mat(lens[i*2]/2, 1, CV_32SC2, (void*)PQs[i*2]) + Scalar(100, 100);
-        Mat Qm = Mat(lens[i*2+1]/2, 1, CV_32SC2, (void*)PQs[i*2+1]) + Scalar(100, 100);
-        Point* P = Pm.ptr<Point>();
-        Point* Q = Qm.ptr<Point>();
-
-        flip(Pm, Pm, 0);
-        flip(Qm, Qm, 0);
-
-        Mat Rm;
-        intersectConvexConvex(Pm, Qm, Rm);
-        std::cout << Rm << std::endl << std::endl;
-
-        img = Scalar::all(0);
-
-        polylines(img, Pm, true, Scalar(0,255,0), 1, CV_AA, 0);
-        polylines(img, Qm, true, Scalar(0,0,255), 1, CV_AA, 0);
-        Mat temp;
-        Rm.convertTo(temp, CV_32S, 256);
-        polylines(img, temp, true, Scalar(128, 255, 255), 3, CV_AA, 8);
-
-        namedWindow("test", 1);
-        imshow("test", img);
-        waitKey();
-    }
-}
-*/
-
-/* End of file. */
diff --git a/modules/imgproc/src/histogram.cpp b/modules/imgproc/src/histogram.cpp
index 8b8f05cd74..f0a78187d0 100644
--- a/modules/imgproc/src/histogram.cpp
+++ b/modules/imgproc/src/histogram.cpp
@@ -43,10 +43,6 @@
 namespace cv
 {
 
-template<> void Ptr<CvHistogram>::delete_obj()
-{ cvReleaseHist(&obj); }
-
-
 ////////////////// Helper functions //////////////////////
 
 static const size_t OUT_OF_RANGE = (size_t)1 << (sizeof(size_t)*8 - 2);
@@ -270,6 +266,8 @@ public:
     }
 
 private:
+    calcHist1D_Invoker operator=(const calcHist1D_Invoker&);
+
     T* p_[one];
     uchar* mask_;
     int step_[one];
@@ -342,6 +340,8 @@ public:
     }
 
 private:
+    calcHist2D_Invoker operator=(const calcHist2D_Invoker&);
+
     T* p_[two];
     uchar* mask_;
     int step_[two];
@@ -432,6 +432,8 @@ public:
     }
 
 private:
+    calcHist3D_Invoker operator=(const calcHist3D_Invoker&);
+
     T* p_[three];
     uchar* mask_;
     int step_[three];
@@ -771,8 +773,7 @@ calcHist_( std::vector<uchar*>& _ptrs, const std::vector<int>& _deltas,
 #ifdef HAVE_TBB
             calcHist1D_Invoker<T> body(_ptrs, _deltas, hist, _uniranges, size[0], dims, imsize);
             parallel_for(BlockedRange(0, imsize.height), body);
-            return;
-#endif
+#else
             double a = uniranges[0], b = uniranges[1];
             int sz = size[0], d0 = deltas[0], step0 = deltas[1];
             const T* p0 = (const T*)ptrs[0];
@@ -795,14 +796,15 @@ calcHist_( std::vector<uchar*>& _ptrs, const std::vector<int>& _deltas,
                                 ((int*)H)[idx]++;
                         }
             }
+#endif //HAVE_TBB
+            return;
         }
         else if( dims == 2 )
         {
 #ifdef HAVE_TBB
             calcHist2D_Invoker<T> body(_ptrs, _deltas, hist, _uniranges, size, dims, imsize, hstep);
             parallel_for(BlockedRange(0, imsize.height), body);
-            return;
-#endif
+#else
             double a0 = uniranges[0], b0 = uniranges[1], a1 = uniranges[2], b1 = uniranges[3];
             int sz0 = size[0], sz1 = size[1];
             int d0 = deltas[0], step0 = deltas[1],
@@ -831,6 +833,8 @@ calcHist_( std::vector<uchar*>& _ptrs, const std::vector<int>& _deltas,
                                 ((int*)(H + hstep0*idx0))[idx1]++;
                         }
             }
+#endif //HAVE_TBB
+            return;
         }
         else if( dims == 3 )
         {
@@ -1974,7 +1978,7 @@ double cv::compareHist( InputArray _H1, InputArray _H2, int method )
     double result = 0;
     int j, len = (int)it.size;
 
-    CV_Assert( H1.type() == H2.type() && H1.type() == CV_32F );
+    CV_Assert( H1.type() == H2.type() && H1.depth() == CV_32F );
 
     double s1 = 0, s2 = 0, s11 = 0, s12 = 0, s22 = 0;
 
@@ -1984,14 +1988,14 @@ double cv::compareHist( InputArray _H1, InputArray _H2, int method )
     {
         const float* h1 = (const float*)it.planes[0].data;
         const float* h2 = (const float*)it.planes[1].data;
-        len = it.planes[0].rows*it.planes[0].cols;
+        len = it.planes[0].rows*it.planes[0].cols*H1.channels();
 
-        if( method == CV_COMP_CHISQR )
+        if( (method == CV_COMP_CHISQR) || (method == CV_COMP_CHISQR_ALT))
         {
             for( j = 0; j < len; j++ )
             {
                 double a = h1[j] - h2[j];
-                double b = h1[j];
+                double b = (method == CV_COMP_CHISQR) ? h1[j] : h1[j] + h2[j];
                 if( fabs(b) > DBL_EPSILON )
                     result += a*a/b;
             }
@@ -2030,7 +2034,9 @@ double cv::compareHist( InputArray _H1, InputArray _H2, int method )
             CV_Error( CV_StsBadArg, "Unknown comparison method" );
     }
 
-    if( method == CV_COMP_CORREL )
+    if( method == CV_COMP_CHISQR_ALT )
+        result *= 2;
+    else if( method == CV_COMP_CORREL )
     {
         size_t total = H1.total();
         double scale = 1./total;
@@ -2059,13 +2065,13 @@ double cv::compareHist( const SparseMat& H1, const SparseMat& H2, int method )
         CV_Assert( H1.size(i) == H2.size(i) );
 
     const SparseMat *PH1 = &H1, *PH2 = &H2;
-    if( PH1->nzcount() > PH2->nzcount() && method != CV_COMP_CHISQR )
+    if( PH1->nzcount() > PH2->nzcount() && method != CV_COMP_CHISQR && method != CV_COMP_CHISQR_ALT)
         std::swap(PH1, PH2);
 
     SparseMatConstIterator it = PH1->begin();
     int N1 = (int)PH1->nzcount(), N2 = (int)PH2->nzcount();
 
-    if( method == CV_COMP_CHISQR )
+    if( (method == CV_COMP_CHISQR) || (method == CV_COMP_CHISQR_ALT) )
     {
         for( i = 0; i < N1; i++, ++it )
         {
@@ -2073,7 +2079,7 @@ double cv::compareHist( const SparseMat& H1, const SparseMat& H2, int method )
             const SparseMat::Node* node = it.node();
             float v2 = PH2->value<float>(node->idx, (size_t*)&node->hashval);
             double a = v1 - v2;
-            double b = v1;
+            double b = (method == CV_COMP_CHISQR) ? v1 : v1 + v2;
             if( fabs(b) > DBL_EPSILON )
                 result += a*a/b;
         }
@@ -2142,6 +2148,9 @@ double cv::compareHist( const SparseMat& H1, const SparseMat& H2, int method )
     else
         CV_Error( CV_StsBadArg, "Unknown comparison method" );
 
+    if( method == CV_COMP_CHISQR_ALT )
+        result *= 2;
+
     return result;
 }
 
@@ -2456,7 +2465,8 @@ cvCompareHist( const CvHistogram* hist1,
 
     if( !CV_IS_SPARSE_MAT(hist1->bins) )
     {
-        cv::Mat H1((const CvMatND*)hist1->bins), H2((const CvMatND*)hist2->bins);
+        cv::Mat H1 = cv::cvarrToMat(hist1->bins);
+        cv::Mat H2 = cv::cvarrToMat(hist2->bins);
         return cv::compareHist(H1, H2, method);
     }
 
@@ -2480,13 +2490,13 @@ cvCompareHist( const CvHistogram* hist1,
     CvSparseMatIterator iterator;
     CvSparseNode *node1, *node2;
 
-    if( mat1->heap->active_count > mat2->heap->active_count && method != CV_COMP_CHISQR )
+    if( mat1->heap->active_count > mat2->heap->active_count && method != CV_COMP_CHISQR && method != CV_COMP_CHISQR_ALT)
     {
         CvSparseMat* t;
         CV_SWAP( mat1, mat2, t );
     }
 
-    if( method == CV_COMP_CHISQR )
+    if( (method == CV_COMP_CHISQR) || (method == CV_COMP_CHISQR_ALT) )
     {
         for( node1 = cvInitSparseMatIterator( mat1, &iterator );
              node1 != 0; node1 = cvGetNextSparseNode( &iterator ))
@@ -2495,7 +2505,7 @@ cvCompareHist( const CvHistogram* hist1,
             uchar* node2_data = cvPtrND( mat2, CV_NODE_IDX(mat1,node1), 0, 0, &node1->hashval );
             double v2 = node2_data ? *(float*)node2_data : 0.f;
             double a = v1 - v2;
-            double b = v1;
+            double b = (method == CV_COMP_CHISQR) ? v1 : v1 + v2;
             if( fabs(b) > DBL_EPSILON )
                 result += a*a/b;
         }
@@ -2585,6 +2595,9 @@ cvCompareHist( const CvHistogram* hist1,
     else
         CV_Error( CV_StsBadArg, "Unknown comparison method" );
 
+    if( method == CV_COMP_CHISQR_ALT )
+        result *= 2;
+
     return result;
 }
 
@@ -2604,7 +2617,7 @@ cvCopyHist( const CvHistogram* src, CvHistogram** _dst )
     int size1[CV_MAX_DIM];
     bool is_sparse = CV_IS_SPARSE_MAT(src->bins);
     int dims1 = cvGetDims( src->bins, size1 );
-    
+
     if( dst && (is_sparse == CV_IS_SPARSE_MAT(dst->bins)))
     {
         int size2[CV_MAX_DIM];
@@ -2613,14 +2626,14 @@ cvCopyHist( const CvHistogram* src, CvHistogram** _dst )
         if( dims1 == dims2 )
         {
             int i;
-            
+
             for( i = 0; i < dims1; i++ )
             {
                 if( size1[i] != size2[i] )
                     break;
             }
-                   
-	        eq = (i == dims1);
+
+            eq = (i == dims1);
         }
     }
 
@@ -2635,19 +2648,19 @@ cvCopyHist( const CvHistogram* src, CvHistogram** _dst )
     {
         float* ranges[CV_MAX_DIM];
         float** thresh = 0;
-        
+
         if( CV_IS_UNIFORM_HIST( src ))
         {
             for( int i = 0; i < dims1; i++ )
                 ranges[i] = (float*)src->thresh[i];
-                
+
             thresh = ranges;
         }
         else
         {
             thresh = src->thresh2;
         }
-            
+
         cvSetHistBinRanges( dst, thresh, CV_IS_UNIFORM_HIST(src));
     }
 
@@ -2758,7 +2771,7 @@ cvCalcArrHist( CvArr** img, CvHistogram* hist, int accumulate, const CvArr* mask
 
     if( !CV_IS_SPARSE_HIST(hist) )
     {
-        cv::Mat H((const CvMatND*)hist->bins);
+        cv::Mat H = cv::cvarrToMat(hist->bins);
         cv::calcHist( &images[0], (int)images.size(), 0, _mask,
                       H, cvGetDims(hist->bins), H.size, ranges, uniform, accumulate != 0 );
     }
@@ -2768,7 +2781,8 @@ cvCalcArrHist( CvArr** img, CvHistogram* hist, int accumulate, const CvArr* mask
 
         if( !accumulate )
             cvZero( hist->bins );
-        cv::SparseMat sH(sparsemat);
+        cv::SparseMat sH;
+        sparsemat->copyToSparseMat(sH);
         cv::calcHist( &images[0], (int)images.size(), 0, _mask, sH, sH.dims(),
                       sH.dims() > 0 ? sH.hdr->size : 0, ranges, uniform, accumulate != 0, true );
 
@@ -2820,13 +2834,14 @@ cvCalcArrBackProject( CvArr** img, CvArr* dst, const CvHistogram* hist )
 
     if( !CV_IS_SPARSE_HIST(hist) )
     {
-        cv::Mat H((const CvMatND*)hist->bins);
+        cv::Mat H = cv::cvarrToMat(hist->bins);
         cv::calcBackProject( &images[0], (int)images.size(),
                             0, H, _dst, ranges, 1, uniform );
     }
     else
     {
-        cv::SparseMat sH((const CvSparseMat*)hist->bins);
+        cv::SparseMat sH;
+        ((const CvSparseMat*)hist->bins)->copyToSparseMat(sH);
         cv::calcBackProject( &images[0], (int)images.size(),
                              0, sH, _dst, ranges, 1, uniform );
     }
@@ -2986,29 +3001,23 @@ cvCalcProbDensity( const CvHistogram* hist, const CvHistogram* hist_mask,
     }
 }
 
-class EqualizeHistCalcHist_Invoker
+class EqualizeHistCalcHist_Invoker : public cv::ParallelLoopBody
 {
 public:
     enum {HIST_SZ = 256};
 
-#ifdef HAVE_TBB
-    typedef tbb::mutex* MutextPtr;
-#else
-    typedef void* MutextPtr;
-#endif
-
-    EqualizeHistCalcHist_Invoker(cv::Mat& src, int* histogram, MutextPtr histogramLock)
+    EqualizeHistCalcHist_Invoker(cv::Mat& src, int* histogram, cv::Mutex* histogramLock)
         : src_(src), globalHistogram_(histogram), histogramLock_(histogramLock)
     { }
 
-    void operator()( const cv::BlockedRange& rowRange ) const
+    void operator()( const cv::Range& rowRange ) const
     {
         int localHistogram[HIST_SZ] = {0, };
 
         const size_t sstep = src_.step;
 
         int width = src_.cols;
-        int height = rowRange.end() - rowRange.begin();
+        int height = rowRange.end - rowRange.start;
 
         if (src_.isContinuous())
         {
@@ -3016,7 +3025,7 @@ public:
             height = 1;
         }
 
-        for (const uchar* ptr = src_.ptr<uchar>(rowRange.begin()); height--; ptr += sstep)
+        for (const uchar* ptr = src_.ptr<uchar>(rowRange.start); height--; ptr += sstep)
         {
             int x = 0;
             for (; x <= width - 4; x += 4)
@@ -3031,9 +3040,7 @@ public:
                 localHistogram[ptr[x]]++;
         }
 
-#ifdef HAVE_TBB
-        tbb::mutex::scoped_lock lock(*histogramLock_);
-#endif
+        cv::AutoLock lock(*histogramLock_);
 
         for( int i = 0; i < HIST_SZ; i++ )
             globalHistogram_[i] += localHistogram[i];
@@ -3041,12 +3048,7 @@ public:
 
     static bool isWorthParallel( const cv::Mat& src )
     {
-#ifdef HAVE_TBB
         return ( src.total() >= 640*480 );
-#else
-        (void)src;
-        return false;
-#endif
     }
 
 private:
@@ -3054,10 +3056,10 @@ private:
 
     cv::Mat& src_;
     int* globalHistogram_;
-    MutextPtr histogramLock_;
+    cv::Mutex* histogramLock_;
 };
 
-class EqualizeHistLut_Invoker
+class EqualizeHistLut_Invoker : public cv::ParallelLoopBody
 {
 public:
     EqualizeHistLut_Invoker( cv::Mat& src, cv::Mat& dst, int* lut )
@@ -3066,13 +3068,13 @@ public:
           lut_(lut)
     { }
 
-    void operator()( const cv::BlockedRange& rowRange ) const
+    void operator()( const cv::Range& rowRange ) const
     {
         const size_t sstep = src_.step;
         const size_t dstep = dst_.step;
 
         int width = src_.cols;
-        int height = rowRange.end() - rowRange.begin();
+        int height = rowRange.end - rowRange.start;
         int* lut = lut_;
 
         if (src_.isContinuous() && dst_.isContinuous())
@@ -3081,8 +3083,8 @@ public:
             height = 1;
         }
 
-        const uchar* sptr = src_.ptr<uchar>(rowRange.begin());
-        uchar* dptr = dst_.ptr<uchar>(rowRange.begin());
+        const uchar* sptr = src_.ptr<uchar>(rowRange.start);
+        uchar* dptr = dst_.ptr<uchar>(rowRange.start);
 
         for (; height--; sptr += sstep, dptr += dstep)
         {
@@ -3111,12 +3113,7 @@ public:
 
     static bool isWorthParallel( const cv::Mat& src )
     {
-#ifdef HAVE_TBB
         return ( src.total() >= 640*480 );
-#else
-        (void)src;
-        return false;
-#endif
     }
 
 private:
@@ -3143,23 +3140,18 @@ void cv::equalizeHist( InputArray _src, OutputArray _dst )
     if(src.empty())
         return;
 
-#ifdef HAVE_TBB
-    tbb::mutex histogramLockInstance;
-    EqualizeHistCalcHist_Invoker::MutextPtr histogramLock = &histogramLockInstance;
-#else
-    EqualizeHistCalcHist_Invoker::MutextPtr histogramLock = 0;
-#endif
+    Mutex histogramLockInstance;
 
     const int hist_sz = EqualizeHistCalcHist_Invoker::HIST_SZ;
     int hist[hist_sz] = {0,};
     int lut[hist_sz];
 
-    EqualizeHistCalcHist_Invoker calcBody(src, hist, histogramLock);
+    EqualizeHistCalcHist_Invoker calcBody(src, hist, &histogramLockInstance);
     EqualizeHistLut_Invoker      lutBody(src, dst, lut);
-    cv::BlockedRange heightRange(0, src.rows);
+    cv::Range heightRange(0, src.rows);
 
     if(EqualizeHistCalcHist_Invoker::isWorthParallel(src))
-        parallel_for(heightRange, calcBody);
+        parallel_for_(heightRange, calcBody);
     else
         calcBody(heightRange);
 
@@ -3183,11 +3175,13 @@ void cv::equalizeHist( InputArray _src, OutputArray _dst )
     }
 
     if(EqualizeHistLut_Invoker::isWorthParallel(src))
-        parallel_for(heightRange, lutBody);
+        parallel_for_(heightRange, lutBody);
     else
         lutBody(heightRange);
 }
 
+// ----------------------------------------------------------------------
+
 /* Implementation of RTTI and Generic Functions for CvHistogram */
 #define CV_TYPE_NAME_HIST "opencv-hist"
 
@@ -3339,4 +3333,3 @@ CvType hist_type( CV_TYPE_NAME_HIST, icvIsHist, (CvReleaseFunc)cvReleaseHist,
                   icvReadHist, icvWriteHist, (CvCloneFunc)icvCloneHist );
 
 /* End of file. */
-
diff --git a/modules/imgproc/src/hough.cpp b/modules/imgproc/src/hough.cpp
index 42901c2434..9c8eaca8f5 100644
--- a/modules/imgproc/src/hough.cpp
+++ b/modules/imgproc/src/hough.cpp
@@ -56,7 +56,10 @@ struct LinePolar
 struct hough_cmp_gt
 {
     hough_cmp_gt(const int* _aux) : aux(_aux) {}
-    bool operator()(int l1, int l2) const { return aux[l1] > aux[l2]; }
+    bool operator()(int l1, int l2) const
+    {
+        return aux[l1] > aux[l2] || (aux[l1] == aux[l2] && l1 < l2);
+    }
     const int* aux;
 };
 
@@ -128,7 +131,7 @@ HoughLinesStandard( const Mat& img, float rho, float theta,
         }
 
     // stage 3. sort the detected lines by accumulator value
-    cv::sort(_sort_buf, hough_cmp_gt(accum));
+    std::sort(_sort_buf.begin(), _sort_buf.end(), hough_cmp_gt(accum));
 
     // stage 4. store the first min(total,linesMax) lines to the output buffer
     linesMax = std::min(linesMax, (int)_sort_buf.size());
@@ -763,21 +766,21 @@ icvHoughCirclesGradient( CvMat* img, float dp, float min_dist,
     float idp, dr;
     CvSeqReader reader;
 
-    edges = cvCreateMat( img->rows, img->cols, CV_8UC1 );
+    edges.reset(cvCreateMat( img->rows, img->cols, CV_8UC1 ));
     cvCanny( img, edges, MAX(canny_threshold/2,1), canny_threshold, 3 );
 
-    dx = cvCreateMat( img->rows, img->cols, CV_16SC1 );
-    dy = cvCreateMat( img->rows, img->cols, CV_16SC1 );
+    dx.reset(cvCreateMat( img->rows, img->cols, CV_16SC1 ));
+    dy.reset(cvCreateMat( img->rows, img->cols, CV_16SC1 ));
     cvSobel( img, dx, 1, 0, 3 );
     cvSobel( img, dy, 0, 1, 3 );
 
     if( dp < 1.f )
         dp = 1.f;
     idp = 1.f/dp;
-    accum = cvCreateMat( cvCeil(img->rows*idp)+2, cvCeil(img->cols*idp)+2, CV_32SC1 );
+    accum.reset(cvCreateMat( cvCeil(img->rows*idp)+2, cvCeil(img->cols*idp)+2, CV_32SC1 ));
     cvZero(accum);
 
-    storage = cvCreateMemStorage();
+    storage.reset(cvCreateMemStorage());
     nz = cvCreateSeq( CV_32SC2, sizeof(CvSeq), sizeof(CvPoint), storage );
     centers = cvCreateSeq( CV_32SC1, sizeof(CvSeq), sizeof(int), storage );
 
@@ -863,7 +866,7 @@ icvHoughCirclesGradient( CvMat* img, float dp, float min_dist,
     cvClearSeq( centers );
     cvSeqPushMulti( centers, &sort_buf[0], center_count );
 
-    dist_buf = cvCreateMat( 1, nz_count, CV_32FC1 );
+    dist_buf.reset(cvCreateMat( 1, nz_count, CV_32FC1 ));
     ddata = dist_buf->data.fl;
 
     dr = dp;
@@ -1057,7 +1060,7 @@ void cv::HoughCircles( InputArray _image, OutputArray _circles,
                        double param1, double param2,
                        int minRadius, int maxRadius )
 {
-    Ptr<CvMemStorage> storage = cvCreateMemStorage(STORAGE_SIZE);
+    Ptr<CvMemStorage> storage(cvCreateMemStorage(STORAGE_SIZE));
     Mat image = _image.getMat();
     CvMat c_image = image;
     CvSeq* seq = cvHoughCircles( &c_image, storage, method,
diff --git a/modules/imgproc/src/imgwarp.cpp b/modules/imgproc/src/imgwarp.cpp
index f344f8b1aa..9e2048407b 100644
--- a/modules/imgproc/src/imgwarp.cpp
+++ b/modules/imgproc/src/imgwarp.cpp
@@ -50,9 +50,73 @@
 #include <iostream>
 #include <vector>
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+static IppStatus sts = ippInit();
+#endif
+
 namespace cv
 {
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    typedef IppStatus (CV_STDCALL* ippiSetFunc)(const void*, void *, int, IppiSize);
+    typedef IppStatus (CV_STDCALL* ippiWarpPerspectiveBackFunc)(const void*, IppiSize, int, IppiRect, void *, int, IppiRect, double [3][3], int);
+    typedef IppStatus (CV_STDCALL* ippiWarpAffineBackFunc)(const void*, IppiSize, int, IppiRect, void *, int, IppiRect, double [2][3], int);
+    typedef IppStatus (CV_STDCALL* ippiResizeSqrPixelFunc)(const void*, IppiSize, int, IppiRect, void*, int, IppiRect, double, double, double, double, int, Ipp8u *);
+
+    template <int channels, typename Type>
+    bool IPPSetSimple(cv::Scalar value, void *dataPointer, int step, IppiSize &size, ippiSetFunc func)
+    {
+        Type values[channels];
+        for( int i = 0; i < channels; i++ )
+            values[i] = (Type)value[i];
+        return func(values, dataPointer, step, size) >= 0;
+    }
+
+    bool IPPSet(const cv::Scalar &value, void *dataPointer, int step, IppiSize &size, int channels, int depth)
+    {
+        if( channels == 1 )
+        {
+            switch( depth )
+            {
+            case CV_8U:
+                return ippiSet_8u_C1R((Ipp8u)value[0], (Ipp8u *)dataPointer, step, size) >= 0;
+            case CV_16U:
+                return ippiSet_16u_C1R((Ipp16u)value[0], (Ipp16u *)dataPointer, step, size) >= 0;
+            case CV_32F:
+                return ippiSet_32f_C1R((Ipp32f)value[0], (Ipp32f *)dataPointer, step, size) >= 0;
+            }
+        }
+        else
+        {
+            if( channels == 3 )
+            {
+                switch( depth )
+                {
+                case CV_8U:
+                    return IPPSetSimple<3, Ipp8u>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_8u_C3R);
+                case CV_16U:
+                    return IPPSetSimple<3, Ipp16u>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_16u_C3R);
+                case CV_32F:
+                    return IPPSetSimple<3, Ipp32f>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_32f_C3R);
+                }
+            }
+            else if( channels == 4 )
+            {
+                switch( depth )
+                {
+                case CV_8U:
+                    return IPPSetSimple<4, Ipp8u>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_8u_C4R);
+                case CV_16U:
+                    return IPPSetSimple<4, Ipp16u>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_16u_C4R);
+                case CV_32F:
+                    return IPPSetSimple<4, Ipp32f>(value, dataPointer, step, size, (ippiSetFunc)ippiSet_32f_C4R);
+                }
+            }
+        }
+        return false;
+    }
+#endif
+
 /************** interpolation formulas and tables ***************/
 
 const int INTER_RESIZE_COEF_BITS=11;
@@ -1219,8 +1283,6 @@ static void resizeGeneric_( const Mat& src, Mat& dst,
                             const int* yofs, const void* _beta,
                             int xmin, int xmax, int ksize )
 {
-    typedef typename HResize::value_type T;
-    typedef typename HResize::buf_type WT;
     typedef typename HResize::alpha_type AT;
 
     const AT* beta = (const AT*)_beta;
@@ -1797,6 +1859,45 @@ static int computeResizeAreaTab( int ssize, int dsize, int cn, double scale, Dec
     return k;
 }
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+class IPPresizeInvoker :
+    public ParallelLoopBody
+{
+public:
+    IPPresizeInvoker(Mat &_src, Mat &_dst, double &_inv_scale_x, double &_inv_scale_y, int _mode, ippiResizeSqrPixelFunc _func, bool *_ok) :
+      ParallelLoopBody(), src(_src), dst(_dst), inv_scale_x(_inv_scale_x), inv_scale_y(_inv_scale_y), mode(_mode), func(_func), ok(_ok)
+      {
+          *ok = true;
+      }
+
+      virtual void operator() (const Range& range) const
+      {
+          int cn = src.channels();
+          IppiRect srcroi = { 0, range.start, src.cols, range.end - range.start };
+          int dsty = CV_IMIN(cvRound(range.start * inv_scale_y), dst.rows);
+          int dstwidth = CV_IMIN(cvRound(src.cols * inv_scale_x), dst.cols);
+          int dstheight = CV_IMIN(cvRound(range.end * inv_scale_y), dst.rows);
+          IppiRect dstroi = { 0, dsty, dstwidth, dstheight - dsty };
+          int bufsize;
+          ippiResizeGetBufSize( srcroi, dstroi, cn, mode, &bufsize );
+          Ipp8u *buf;
+          buf = ippsMalloc_8u( bufsize );
+          IppStatus sts;
+          if( func( src.data, ippiSize(src.cols, src.rows), (int)src.step[0], srcroi, dst.data, (int)dst.step[0], dstroi, inv_scale_x, inv_scale_y, 0, 0, mode, buf ) < 0 )
+              *ok = false;
+          ippsFree(buf);
+      }
+private:
+    Mat &src;
+    Mat &dst;
+    double inv_scale_x;
+    double inv_scale_y;
+    int mode;
+    ippiResizeSqrPixelFunc func;
+    bool *ok;
+    const IPPresizeInvoker& operator= (const IPPresizeInvoker&);
+};
+#endif
 
 }
 
@@ -1931,7 +2032,7 @@ void cv::resize( InputArray _src, OutputArray _dst, Size dsize,
 
 
 #ifdef HAVE_TEGRA_OPTIMIZATION
-    if (tegra::resize(src, dst, inv_scale_x, inv_scale_y, interpolation))
+    if (tegra::resize(src, dst, (float)inv_scale_x, (float)inv_scale_y, interpolation))
         return;
 #endif
 
@@ -1939,6 +2040,34 @@ void cv::resize( InputArray _src, OutputArray _dst, Size dsize,
     double scale_x = 1./inv_scale_x, scale_y = 1./inv_scale_y;
     int k, sx, sy, dx, dy;
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    int mode = interpolation == INTER_LINEAR ? IPPI_INTER_LINEAR : 0;
+    int type = src.type();
+    ippiResizeSqrPixelFunc ippFunc =
+        type == CV_8UC1 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_8u_C1R :
+        type == CV_8UC3 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_8u_C3R :
+        type == CV_8UC4 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_8u_C4R :
+        type == CV_16UC1 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16u_C1R :
+        type == CV_16UC3 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16u_C3R :
+        type == CV_16UC4 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16u_C4R :
+        type == CV_16SC1 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16s_C1R :
+        type == CV_16SC3 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16s_C3R :
+        type == CV_16SC4 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_16s_C4R :
+        type == CV_32FC1 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_32f_C1R :
+        type == CV_32FC3 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_32f_C3R :
+        type == CV_32FC4 ? (ippiResizeSqrPixelFunc)ippiResizeSqrPixel_32f_C4R :
+        0;
+    if( ippFunc && mode != 0 )
+    {
+        bool ok;
+        Range range(0, src.rows);
+        IPPresizeInvoker invoker(src, dst, inv_scale_x, inv_scale_y, mode, ippFunc, &ok);
+        parallel_for_(range, invoker, dst.total()/(double)(1<<16));
+        if( ok )
+            return;
+    }
+#endif
+
     if( interpolation == INTER_NEAREST )
     {
         resizeNN( src, dst, inv_scale_x, inv_scale_y );
@@ -3448,6 +3577,49 @@ private:
     double *M;
 };
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+class IPPwarpAffineInvoker :
+    public ParallelLoopBody
+{
+public:
+    IPPwarpAffineInvoker(Mat &_src, Mat &_dst, double (&_coeffs)[2][3], int &_interpolation, int &_borderType, const Scalar &_borderValue, ippiWarpAffineBackFunc _func, bool *_ok) :
+      ParallelLoopBody(), src(_src), dst(_dst), mode(_interpolation), coeffs(_coeffs), borderType(_borderType), borderValue(_borderValue), func(_func), ok(_ok)
+      {
+          *ok = true;
+      }
+
+      virtual void operator() (const Range& range) const
+      {
+          IppiSize srcsize = { src.cols, src.rows };
+          IppiRect srcroi = { 0, 0, src.cols, src.rows };
+          IppiRect dstroi = { 0, range.start, dst.cols, range.end - range.start };
+          int cnn = src.channels();
+          if( borderType == BORDER_CONSTANT )
+          {
+              IppiSize setSize = { dst.cols, range.end - range.start };
+              void *dataPointer = dst.data + dst.step[0] * range.start;
+              if( !IPPSet( borderValue, dataPointer, (int)dst.step[0], setSize, cnn, src.depth() ) )
+              {
+                  *ok = false;
+                  return;
+              }
+          }
+          if( func( src.data, srcsize, (int)src.step[0], srcroi, dst.data, (int)dst.step[0], dstroi, coeffs, mode ) < 0) ////Aug 2013: problem in IPP 7.1, 8.0 : sometimes function return ippStsCoeffErr
+              *ok = false;
+      }
+private:
+    Mat &src;
+    Mat &dst;
+    double (&coeffs)[2][3];
+    int mode;
+    int borderType;
+    Scalar borderValue;
+    ippiWarpAffineBackFunc func;
+    bool *ok;
+    const IPPwarpAffineInvoker& operator= (const IPPwarpAffineInvoker&);
+};
+#endif
+
 }
 
 
@@ -3494,6 +3666,50 @@ void cv::warpAffine( InputArray _src, OutputArray _dst,
     const int AB_BITS = MAX(10, (int)INTER_BITS);
     const int AB_SCALE = 1 << AB_BITS;
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    int depth = src.depth();
+    int channels = src.channels();
+    if( ( depth == CV_8U || depth == CV_16U || depth == CV_32F ) &&
+        ( channels == 1 || channels == 3 || channels == 4 ) &&
+        ( borderType == cv::BORDER_TRANSPARENT || ( borderType == cv::BORDER_CONSTANT ) ) )
+    {
+        int type = src.type();
+        ippiWarpAffineBackFunc ippFunc =
+            type == CV_8UC1 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_8u_C1R :
+            type == CV_8UC3 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_8u_C3R :
+            type == CV_8UC4 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_8u_C4R :
+            type == CV_16UC1 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_16u_C1R :
+            type == CV_16UC3 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_16u_C3R :
+            type == CV_16UC4 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_16u_C4R :
+            type == CV_32FC1 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_32f_C1R :
+            type == CV_32FC3 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_32f_C3R :
+            type == CV_32FC4 ? (ippiWarpAffineBackFunc)ippiWarpAffineBack_32f_C4R :
+            0;
+        int mode =
+            flags == INTER_LINEAR ? IPPI_INTER_LINEAR :
+            flags == INTER_NEAREST ? IPPI_INTER_NN :
+            flags == INTER_CUBIC ? IPPI_INTER_CUBIC :
+            0;
+        if( mode && ippFunc )
+        {
+            double coeffs[2][3];
+            for( int i = 0; i < 2; i++ )
+            {
+                for( int j = 0; j < 3; j++ )
+                {
+                    coeffs[i][j] = matM.at<double>(i, j);
+                }
+            }
+            bool ok;
+            Range range(0, dst.rows);
+            IPPwarpAffineInvoker invoker(src, dst, coeffs, mode, borderType, borderValue, ippFunc, &ok);
+            parallel_for_(range, invoker, dst.total()/(double)(1<<16));
+            if( ok )
+                return;
+        }
+    }
+#endif
+
     for( x = 0; x < dst.cols; x++ )
     {
         adelta[x] = saturate_cast<int>(M[0]*x*AB_SCALE);
@@ -3601,6 +3817,50 @@ private:
     Scalar borderValue;
 };
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+class IPPwarpPerspectiveInvoker :
+    public ParallelLoopBody
+{
+public:
+    IPPwarpPerspectiveInvoker(Mat &_src, Mat &_dst, double (&_coeffs)[3][3], int &_interpolation, int &_borderType, const Scalar &_borderValue, ippiWarpPerspectiveBackFunc _func, bool *_ok) :
+      ParallelLoopBody(), src(_src), dst(_dst), mode(_interpolation), coeffs(_coeffs), borderType(_borderType), borderValue(_borderValue), func(_func), ok(_ok)
+      {
+          *ok = true;
+      }
+
+      virtual void operator() (const Range& range) const
+      {
+          IppiSize srcsize = {src.cols, src.rows};
+          IppiRect srcroi = {0, 0, src.cols, src.rows};
+          IppiRect dstroi = {0, range.start, dst.cols, range.end - range.start};
+          int cnn = src.channels();
+
+          if( borderType == BORDER_CONSTANT )
+          {
+              IppiSize setSize = {dst.cols, range.end - range.start};
+              void *dataPointer = dst.data + dst.step[0] * range.start;
+              if( !IPPSet( borderValue, dataPointer, (int)dst.step[0], setSize, cnn, src.depth() ) )
+              {
+                  *ok = false;
+                  return;
+              }
+          }
+          if( func(src.data, srcsize, (int)src.step[0], srcroi, dst.data, (int)dst.step[0], dstroi, coeffs, mode) < 0)
+              *ok = false;
+      }
+private:
+    Mat &src;
+    Mat &dst;
+    double (&coeffs)[3][3];
+    int mode;
+    int borderType;
+    const Scalar borderValue;
+    ippiWarpPerspectiveBackFunc func;
+    bool *ok;
+    const IPPwarpPerspectiveInvoker& operator= (const IPPwarpPerspectiveInvoker&);
+};
+#endif
+
 }
 
 void cv::warpPerspective( InputArray _src, OutputArray _dst, InputArray _M0,
@@ -3631,6 +3891,50 @@ void cv::warpPerspective( InputArray _src, OutputArray _dst, InputArray _M0,
     if( !(flags & WARP_INVERSE_MAP) )
          invert(matM, matM);
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    int depth = src.depth();
+    int channels = src.channels();
+    if( ( depth == CV_8U || depth == CV_16U || depth == CV_32F ) &&
+        ( channels == 1 || channels == 3 || channels == 4 ) &&
+        ( borderType == cv::BORDER_TRANSPARENT || borderType == cv::BORDER_CONSTANT ) )
+    {
+        int type = src.type();
+        ippiWarpPerspectiveBackFunc ippFunc =
+            type == CV_8UC1 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_8u_C1R :
+            type == CV_8UC3 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_8u_C3R :
+            type == CV_8UC4 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_8u_C4R :
+            type == CV_16UC1 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_16u_C1R :
+            type == CV_16UC3 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_16u_C3R :
+            type == CV_16UC4 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_16u_C4R :
+            type == CV_32FC1 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_32f_C1R :
+            type == CV_32FC3 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_32f_C3R :
+            type == CV_32FC4 ? (ippiWarpPerspectiveBackFunc)ippiWarpPerspectiveBack_32f_C4R :
+            0;
+        int mode =
+            flags == INTER_LINEAR ? IPPI_INTER_LINEAR :
+            flags == INTER_NEAREST ? IPPI_INTER_NN :
+            flags == INTER_CUBIC ? IPPI_INTER_CUBIC :
+            0;
+        if( mode && ippFunc )
+        {
+            double coeffs[3][3];
+            for( int i = 0; i < 3; i++ )
+            {
+                for( int j = 0; j < 3; j++ )
+                {
+                    coeffs[i][j] = matM.at<double>(i, j);
+                }
+            }
+            bool ok;
+            Range range(0, dst.rows);
+            IPPwarpPerspectiveInvoker invoker(src, dst, coeffs, mode, borderType, borderValue, ippFunc, &ok);
+            parallel_for_(range, invoker, dst.total()/(double)(1<<16));
+            if( ok )
+                return;
+        }
+    }
+#endif
+
     Range range(0, dst.rows);
     warpPerspectiveInvoker invoker(src, dst, M, interpolation, borderType, borderValue);
     parallel_for_(range, invoker, dst.total()/(double)(1<<16));
@@ -3858,7 +4162,7 @@ cv2DRotationMatrix( CvPoint2D32f center, double angle,
                     double scale, CvMat* matrix )
 {
     cv::Mat M0 = cv::cvarrToMat(matrix), M = cv::getRotationMatrix2D(center, angle, scale);
-    CV_Assert( M.size() == M.size() );
+    CV_Assert( M.size() == M0.size() );
     M.convertTo(M0, M0.type());
     return matrix;
 }
@@ -3871,7 +4175,7 @@ cvGetPerspectiveTransform( const CvPoint2D32f* src,
 {
     cv::Mat M0 = cv::cvarrToMat(matrix),
         M = cv::getPerspectiveTransform((const cv::Point2f*)src, (const cv::Point2f*)dst);
-    CV_Assert( M.size() == M.size() );
+    CV_Assert( M.size() == M0.size() );
     M.convertTo(M0, M0.type());
     return matrix;
 }
@@ -3933,8 +4237,8 @@ cvLogPolar( const CvArr* srcarr, CvArr* dstarr,
     ssize = cvGetMatSize(src);
     dsize = cvGetMatSize(dst);
 
-    mapx = cvCreateMat( dsize.height, dsize.width, CV_32F );
-    mapy = cvCreateMat( dsize.height, dsize.width, CV_32F );
+    mapx.reset(cvCreateMat( dsize.height, dsize.width, CV_32F ));
+    mapy.reset(cvCreateMat( dsize.height, dsize.width, CV_32F ));
 
     if( !(flags & CV_WARP_INVERSE_MAP) )
     {
@@ -4051,8 +4355,8 @@ void cvLinearPolar( const CvArr* srcarr, CvArr* dstarr,
     dsize.width = dst->cols;
     dsize.height = dst->rows;
 
-    mapx = cvCreateMat( dsize.height, dsize.width, CV_32F );
-    mapy = cvCreateMat( dsize.height, dsize.width, CV_32F );
+    mapx.reset(cvCreateMat( dsize.height, dsize.width, CV_32F ));
+    mapy.reset(cvCreateMat( dsize.height, dsize.width, CV_32F ));
 
     if( !(flags & CV_WARP_INVERSE_MAP) )
     {
diff --git a/modules/imgproc/src/linefit.cpp b/modules/imgproc/src/linefit.cpp
index 61969b5028..dc71d888ae 100644
--- a/modules/imgproc/src/linefit.cpp
+++ b/modules/imgproc/src/linefit.cpp
@@ -571,14 +571,14 @@ static void fitLine3D( Point3f * points, int count, int dist,
                 for( j = 0; j < count; j++ )
                     w[j] = 1.f;
             }
-            
+
             /* save the line parameters */
             memcpy( _lineprev, _line, 6 * sizeof( float ));
-            
+
             /* Run again... */
             fitLine3D_wods( points, count, w, _line );
         }
-        
+
         if( err < min_err )
         {
             min_err = err;
@@ -595,27 +595,27 @@ void cv::fitLine( InputArray _points, OutputArray _line, int distType,
                  double param, double reps, double aeps )
 {
     Mat points = _points.getMat();
-    
+
     float linebuf[6]={0.f};
     int npoints2 = points.checkVector(2, -1, false);
     int npoints3 = points.checkVector(3, -1, false);
-    
+
     CV_Assert( npoints2 >= 0 || npoints3 >= 0 );
-    
+
     if( points.depth() != CV_32F || !points.isContinuous() )
     {
         Mat temp;
         points.convertTo(temp, CV_32F);
         points = temp;
     }
-    
+
     if( npoints2 >= 0 )
         fitLine2D( points.ptr<Point2f>(), npoints2, distType,
                    (float)param, (float)reps, (float)aeps, linebuf);
     else
         fitLine3D( points.ptr<Point3f>(), npoints3, distType,
                    (float)param, (float)reps, (float)aeps, linebuf);
-    
+
     Mat(npoints2 >= 0 ? 4 : 6, 1, CV_32F, linebuf).copyTo(_line);
 }
 
diff --git a/modules/imgproc/src/lsd.cpp b/modules/imgproc/src/lsd.cpp
new file mode 100644
index 0000000000..bb38954480
--- /dev/null
+++ b/modules/imgproc/src/lsd.cpp
@@ -0,0 +1,1246 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include <vector>
+
+/////////////////////////////////////////////////////////////////////////////////////////
+// Default LSD parameters
+// SIGMA_SCALE 0.6    - Sigma for Gaussian filter is computed as sigma = sigma_scale/scale.
+// QUANT       2.0    - Bound to the quantization error on the gradient norm.
+// ANG_TH      22.5   - Gradient angle tolerance in degrees.
+// LOG_EPS     0.0    - Detection threshold: -log10(NFA) > log_eps
+// DENSITY_TH  0.7    - Minimal density of region points in rectangle.
+// N_BINS      1024   - Number of bins in pseudo-ordering of gradient modulus.
+
+#define M_3_2_PI    (3 * CV_PI) / 2   // 3/2 pi
+#define M_2__PI     (2 * CV_PI)         // 2 pi
+
+#ifndef M_LN10
+#define M_LN10      2.30258509299404568402
+#endif
+
+#define NOTDEF      double(-1024.0) // Label for pixels with undefined gradient.
+
+#define NOTUSED     0   // Label for pixels not used in yet.
+#define USED        1   // Label for pixels already used in detection.
+
+#define RELATIVE_ERROR_FACTOR 100.0
+
+const double DEG_TO_RADS = CV_PI / 180;
+
+#define log_gamma(x) ((x)>15.0?log_gamma_windschitl(x):log_gamma_lanczos(x))
+
+struct edge
+{
+    cv::Point p;
+    bool taken;
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////
+
+inline double distSq(const double x1, const double y1,
+                     const double x2, const double y2)
+{
+    return (x2 - x1)*(x2 - x1) + (y2 - y1)*(y2 - y1);
+}
+
+inline double dist(const double x1, const double y1,
+                   const double x2, const double y2)
+{
+    return sqrt(distSq(x1, y1, x2, y2));
+}
+
+// Signed angle difference
+inline double angle_diff_signed(const double& a, const double& b)
+{
+    double diff = a - b;
+    while(diff <= -CV_PI) diff += M_2__PI;
+    while(diff >   CV_PI) diff -= M_2__PI;
+    return diff;
+}
+
+// Absolute value angle difference
+inline double angle_diff(const double& a, const double& b)
+{
+    return std::fabs(angle_diff_signed(a, b));
+}
+
+// Compare doubles by relative error.
+inline bool double_equal(const double& a, const double& b)
+{
+    // trivial case
+    if(a == b) return true;
+
+    double abs_diff = fabs(a - b);
+    double aa = fabs(a);
+    double bb = fabs(b);
+    double abs_max = (aa > bb)? aa : bb;
+
+    if(abs_max < DBL_MIN) abs_max = DBL_MIN;
+
+    return (abs_diff / abs_max) <= (RELATIVE_ERROR_FACTOR * DBL_EPSILON);
+}
+
+inline bool AsmallerB_XoverY(const edge& a, const edge& b)
+{
+    if (a.p.x == b.p.x) return a.p.y < b.p.y;
+    else return a.p.x < b.p.x;
+}
+
+/**
+ *   Computes the natural logarithm of the absolute value of
+ *   the gamma function of x using Windschitl method.
+ *   See http://www.rskey.org/gamma.htm
+ */
+inline double log_gamma_windschitl(const double& x)
+{
+    return 0.918938533204673 + (x-0.5)*log(x) - x
+         + 0.5*x*log(x*sinh(1/x) + 1/(810.0*pow(x, 6.0)));
+}
+
+/**
+ *   Computes the natural logarithm of the absolute value of
+ *   the gamma function of x using the Lanczos approximation.
+ *   See http://www.rskey.org/gamma.htm
+ */
+inline double log_gamma_lanczos(const double& x)
+{
+    static double q[7] = { 75122.6331530, 80916.6278952, 36308.2951477,
+                         8687.24529705, 1168.92649479, 83.8676043424,
+                         2.50662827511 };
+    double a = (x + 0.5) * log(x + 5.5) - (x + 5.5);
+    double b = 0;
+    for(int n = 0; n < 7; ++n)
+    {
+        a -= log(x + double(n));
+        b += q[n] * pow(x, double(n));
+    }
+    return a + log(b);
+}
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+namespace cv{
+
+class LineSegmentDetectorImpl : public LineSegmentDetector
+{
+public:
+
+/**
+ * Create a LineSegmentDetectorImpl object. Specifying scale, number of subdivisions for the image, should the lines be refined and other constants as follows:
+ *
+ * @param _refine       How should the lines found be refined?
+ *                      LSD_REFINE_NONE - No refinement applied.
+ *                      LSD_REFINE_STD  - Standard refinement is applied. E.g. breaking arches into smaller line approximations.
+ *                      LSD_REFINE_ADV  - Advanced refinement. Number of false alarms is calculated,
+ *                                    lines are refined through increase of precision, decrement in size, etc.
+ * @param _scale        The scale of the image that will be used to find the lines. Range (0..1].
+ * @param _sigma_scale  Sigma for Gaussian filter is computed as sigma = _sigma_scale/_scale.
+ * @param _quant        Bound to the quantization error on the gradient norm.
+ * @param _ang_th       Gradient angle tolerance in degrees.
+ * @param _log_eps      Detection threshold: -log10(NFA) > _log_eps
+ * @param _density_th   Minimal density of aligned region points in rectangle.
+ * @param _n_bins       Number of bins in pseudo-ordering of gradient modulus.
+ */
+    LineSegmentDetectorImpl(int _refine = LSD_REFINE_STD, double _scale = 0.8,
+        double _sigma_scale = 0.6, double _quant = 2.0, double _ang_th = 22.5,
+        double _log_eps = 0, double _density_th = 0.7, int _n_bins = 1024);
+
+/**
+ * Detect lines in the input image.
+ *
+ * @param _image    A grayscale(CV_8UC1) input image.
+ *                  If only a roi needs to be selected, use
+ *                  lsd_ptr->detect(image(roi), ..., lines);
+ *                  lines += Scalar(roi.x, roi.y, roi.x, roi.y);
+ * @param _lines    Return: A vector of Vec4i elements specifying the beginning and ending point of a line.
+ *                          Where Vec4i is (x1, y1, x2, y2), point 1 is the start, point 2 - end.
+ *                          Returned lines are strictly oriented depending on the gradient.
+ * @param width     Return: Vector of widths of the regions, where the lines are found. E.g. Width of line.
+ * @param prec      Return: Vector of precisions with which the lines are found.
+ * @param nfa       Return: Vector containing number of false alarms in the line region, with precision of 10%.
+ *                          The bigger the value, logarithmically better the detection.
+ *                              * -1 corresponds to 10 mean false alarms
+ *                              * 0 corresponds to 1 mean false alarm
+ *                              * 1 corresponds to 0.1 mean false alarms
+ *                          This vector will be calculated _only_ when the objects type is REFINE_ADV
+ */
+    void detect(InputArray _image, OutputArray _lines,
+                OutputArray width = noArray(), OutputArray prec = noArray(),
+                OutputArray nfa = noArray());
+
+/**
+ * Draw lines on the given canvas.
+ *
+ * @param image     The image, where lines will be drawn.
+ *                  Should have the size of the image, where the lines were found
+ * @param lines     The lines that need to be drawn
+ */
+    void drawSegments(InputOutputArray _image, InputArray lines);
+
+/**
+ * Draw both vectors on the image canvas. Uses blue for lines 1 and red for lines 2.
+ *
+ * @param size      The size of the image, where lines1 and lines2 were found.
+ * @param lines1    The first lines that need to be drawn. Color - Blue.
+ * @param lines2    The second lines that need to be drawn. Color - Red.
+ * @param image     An optional image, where lines will be drawn.
+ *                  Should have the size of the image, where the lines were found
+ * @return          The number of mismatching pixels between lines1 and lines2.
+ */
+    int compareSegments(const Size& size, InputArray lines1, InputArray lines2, InputOutputArray _image = noArray());
+
+private:
+    Mat image;
+    Mat_<double> scaled_image;
+    double *scaled_image_data;
+    Mat_<double> angles;     // in rads
+    double *angles_data;
+    Mat_<double> modgrad;
+    double *modgrad_data;
+    Mat_<uchar> used;
+
+    int img_width;
+    int img_height;
+    double LOG_NT;
+
+    bool w_needed;
+    bool p_needed;
+    bool n_needed;
+
+    const double SCALE;
+    const int doRefine;
+    const double SIGMA_SCALE;
+    const double QUANT;
+    const double ANG_TH;
+    const double LOG_EPS;
+    const double DENSITY_TH;
+    const int N_BINS;
+
+    struct RegionPoint {
+        int x;
+        int y;
+        uchar* used;
+        double angle;
+        double modgrad;
+    };
+
+
+    struct coorlist
+    {
+        Point2i p;
+        struct coorlist* next;
+    };
+
+    struct rect
+    {
+        double x1, y1, x2, y2;    // first and second point of the line segment
+        double width;             // rectangle width
+        double x, y;              // center of the rectangle
+        double theta;             // angle
+        double dx,dy;             // (dx,dy) is vector oriented as the line segment
+        double prec;              // tolerance angle
+        double p;                 // probability of a point with angle within 'prec'
+    };
+
+    LineSegmentDetectorImpl& operator= (const LineSegmentDetectorImpl&); // to quiet MSVC
+
+/**
+ * Detect lines in the whole input image.
+ *
+ * @param lines         Return: A vector of Vec4i elements specifying the beginning and ending point of a line.
+ *                              Where Vec4i is (x1, y1, x2, y2), point 1 is the start, point 2 - end.
+ *                              Returned lines are strictly oriented depending on the gradient.
+ * @param widths        Return: Vector of widths of the regions, where the lines are found. E.g. Width of line.
+ * @param precisions    Return: Vector of precisions with which the lines are found.
+ * @param nfas          Return: Vector containing number of false alarms in the line region, with precision of 10%.
+ *                              The bigger the value, logarithmically better the detection.
+ *                                  * -1 corresponds to 10 mean false alarms
+ *                                  * 0 corresponds to 1 mean false alarm
+ *                                  * 1 corresponds to 0.1 mean false alarms
+ */
+    void flsd(std::vector<Vec4i>& lines,
+              std::vector<double>& widths, std::vector<double>& precisions,
+              std::vector<double>& nfas);
+
+/**
+ * Finds the angles and the gradients of the image. Generates a list of pseudo ordered points.
+ *
+ * @param threshold The minimum value of the angle that is considered defined, otherwise NOTDEF
+ * @param n_bins    The number of bins with which gradients are ordered by, using bucket sort.
+ * @param list      Return: Vector of coordinate points that are pseudo ordered by magnitude.
+ *                  Pixels would be ordered by norm value, up to a precision given by max_grad/n_bins.
+ */
+    void ll_angle(const double& threshold, const unsigned int& n_bins, std::vector<coorlist>& list);
+
+/**
+ * Grow a region starting from point s with a defined precision,
+ * returning the containing points size and the angle of the gradients.
+ *
+ * @param s         Starting point for the region.
+ * @param reg       Return: Vector of points, that are part of the region
+ * @param reg_size  Return: The size of the region.
+ * @param reg_angle Return: The mean angle of the region.
+ * @param prec      The precision by which each region angle should be aligned to the mean.
+ */
+    void region_grow(const Point2i& s, std::vector<RegionPoint>& reg,
+                     int& reg_size, double& reg_angle, const double& prec);
+
+/**
+ * Finds the bounding rotated rectangle of a region.
+ *
+ * @param reg       The region of points, from which the rectangle to be constructed from.
+ * @param reg_size  The number of points in the region.
+ * @param reg_angle The mean angle of the region.
+ * @param prec      The precision by which points were found.
+ * @param p         Probability of a point with angle within 'prec'.
+ * @param rec       Return: The generated rectangle.
+ */
+    void region2rect(const std::vector<RegionPoint>& reg, const int reg_size, const double reg_angle,
+                     const double prec, const double p, rect& rec) const;
+
+/**
+ * Compute region's angle as the principal inertia axis of the region.
+ * @return          Regions angle.
+ */
+    double get_theta(const std::vector<RegionPoint>& reg, const int& reg_size, const double& x,
+                     const double& y, const double& reg_angle, const double& prec) const;
+
+/**
+ * An estimation of the angle tolerance is performed by the standard deviation of the angle at points
+ * near the region's starting point. Then, a new region is grown starting from the same point, but using the
+ * estimated angle tolerance. If this fails to produce a rectangle with the right density of region points,
+ * 'reduce_region_radius' is called to try to satisfy this condition.
+ */
+    bool refine(std::vector<RegionPoint>& reg, int& reg_size, double reg_angle,
+                const double prec, double p, rect& rec, const double& density_th);
+
+/**
+ * Reduce the region size, by elimination the points far from the starting point, until that leads to
+ * rectangle with the right density of region points or to discard the region if too small.
+ */
+    bool reduce_region_radius(std::vector<RegionPoint>& reg, int& reg_size, double reg_angle,
+                const double prec, double p, rect& rec, double density, const double& density_th);
+
+/**
+ * Try some rectangles variations to improve NFA value. Only if the rectangle is not meaningful (i.e., log_nfa <= log_eps).
+ * @return      The new NFA value.
+ */
+    double rect_improve(rect& rec) const;
+
+/**
+ * Calculates the number of correctly aligned points within the rectangle.
+ * @return      The new NFA value.
+ */
+    double rect_nfa(const rect& rec) const;
+
+/**
+ * Computes the NFA values based on the total number of points, points that agree.
+ * n, k, p are the binomial parameters.
+ * @return      The new NFA value.
+ */
+    double nfa(const int& n, const int& k, const double& p) const;
+
+/**
+ * Is the point at place 'address' aligned to angle theta, up to precision 'prec'?
+ * @return      Whether the point is aligned.
+ */
+    bool isAligned(const int& address, const double& theta, const double& prec) const;
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////
+
+CV_EXPORTS Ptr<LineSegmentDetector> createLineSegmentDetectorPtr(
+        int _refine, double _scale, double _sigma_scale, double _quant, double _ang_th,
+        double _log_eps, double _density_th, int _n_bins)
+{
+    return makePtr<LineSegmentDetectorImpl>(
+            _refine, _scale, _sigma_scale, _quant, _ang_th,
+            _log_eps, _density_th, _n_bins);
+}
+
+/////////////////////////////////////////////////////////////////////////////////////////
+
+LineSegmentDetectorImpl::LineSegmentDetectorImpl(int _refine, double _scale, double _sigma_scale, double _quant,
+        double _ang_th, double _log_eps, double _density_th, int _n_bins)
+        :SCALE(_scale), doRefine(_refine), SIGMA_SCALE(_sigma_scale), QUANT(_quant),
+        ANG_TH(_ang_th), LOG_EPS(_log_eps), DENSITY_TH(_density_th), N_BINS(_n_bins)
+{
+    CV_Assert(_scale > 0 && _sigma_scale > 0 && _quant >= 0 &&
+              _ang_th > 0 && _ang_th < 180 && _density_th >= 0 && _density_th < 1 &&
+              _n_bins > 0);
+}
+
+void LineSegmentDetectorImpl::detect(InputArray _image, OutputArray _lines,
+                OutputArray _width, OutputArray _prec, OutputArray _nfa)
+{
+    Mat_<double> img = _image.getMat();
+    CV_Assert(!img.empty() && img.channels() == 1);
+
+    // Convert image to double
+    img.convertTo(image, CV_64FC1);
+
+    std::vector<Vec4i> lines;
+    std::vector<double> w, p, n;
+    w_needed = _width.needed();
+    p_needed = _prec.needed();
+    n_needed = _nfa.needed();
+
+    CV_Assert((!_nfa.needed()) ||                              // NFA InputArray will be filled _only_ when
+              (_nfa.needed() && doRefine >= LSD_REFINE_ADV));  // REFINE_ADV type LineSegmentDetectorImpl object is created.
+
+    flsd(lines, w, p, n);
+
+    Mat(lines).copyTo(_lines);
+    if(w_needed) Mat(w).copyTo(_width);
+    if(p_needed) Mat(p).copyTo(_prec);
+    if(n_needed) Mat(n).copyTo(_nfa);
+}
+
+void LineSegmentDetectorImpl::flsd(std::vector<Vec4i>& lines,
+    std::vector<double>& widths, std::vector<double>& precisions,
+    std::vector<double>& nfas)
+{
+    // Angle tolerance
+    const double prec = CV_PI * ANG_TH / 180;
+    const double p = ANG_TH / 180;
+    const double rho = QUANT / sin(prec);    // gradient magnitude threshold
+
+    std::vector<coorlist> list;
+    if(SCALE != 1)
+    {
+        Mat gaussian_img;
+        const double sigma = (SCALE < 1)?(SIGMA_SCALE / SCALE):(SIGMA_SCALE);
+        const double sprec = 3;
+        const unsigned int h =  (unsigned int)(ceil(sigma * sqrt(2 * sprec * log(10.0))));
+        Size ksize(1 + 2 * h, 1 + 2 * h); // kernel size
+        GaussianBlur(image, gaussian_img, ksize, sigma);
+        // Scale image to needed size
+        resize(gaussian_img, scaled_image, Size(), SCALE, SCALE);
+        ll_angle(rho, N_BINS, list);
+    }
+    else
+    {
+        scaled_image = image;
+        ll_angle(rho, N_BINS, list);
+    }
+
+    LOG_NT = 5 * (log10(double(img_width)) + log10(double(img_height))) / 2 + log10(11.0);
+    const int min_reg_size = int(-LOG_NT/log10(p)); // minimal number of points in region that can give a meaningful event
+
+    // // Initialize region only when needed
+    // Mat region = Mat::zeros(scaled_image.size(), CV_8UC1);
+    used = Mat_<uchar>::zeros(scaled_image.size()); // zeros = NOTUSED
+    std::vector<RegionPoint> reg(img_width * img_height);
+
+    // Search for line segments
+    unsigned int ls_count = 0;
+    unsigned int list_size = list.size();
+    for(unsigned int i = 0; i < list_size; ++i)
+    {
+        unsigned int adx = list[i].p.x + list[i].p.y * img_width;
+        if((used.data[adx] == NOTUSED) && (angles_data[adx] != NOTDEF))
+        {
+            int reg_size;
+            double reg_angle;
+            region_grow(list[i].p, reg, reg_size, reg_angle, prec);
+
+            // Ignore small regions
+            if(reg_size < min_reg_size) { continue; }
+
+            // Construct rectangular approximation for the region
+            rect rec;
+            region2rect(reg, reg_size, reg_angle, prec, p, rec);
+
+            double log_nfa = -1;
+            if(doRefine > LSD_REFINE_NONE)
+            {
+                // At least REFINE_STANDARD lvl.
+                if(!refine(reg, reg_size, reg_angle, prec, p, rec, DENSITY_TH)) { continue; }
+
+                if(doRefine >= LSD_REFINE_ADV)
+                {
+                    // Compute NFA
+                    log_nfa = rect_improve(rec);
+                    if(log_nfa <= LOG_EPS) { continue; }
+                }
+            }
+            // Found new line
+            ++ls_count;
+
+            // Add the offset
+            rec.x1 += 0.5; rec.y1 += 0.5;
+            rec.x2 += 0.5; rec.y2 += 0.5;
+
+            // scale the result values if a sub-sampling was performed
+            if(SCALE != 1)
+            {
+                rec.x1 /= SCALE; rec.y1 /= SCALE;
+                rec.x2 /= SCALE; rec.y2 /= SCALE;
+                rec.width /= SCALE;
+            }
+
+            //Store the relevant data
+            lines.push_back(Vec4i(int(rec.x1), int(rec.y1), int(rec.x2), int(rec.y2)));
+            if(w_needed) widths.push_back(rec.width);
+            if(p_needed) precisions.push_back(rec.p);
+            if(n_needed && doRefine >= LSD_REFINE_ADV) nfas.push_back(log_nfa);
+
+
+            // //Add the linesID to the region on the image
+            // for(unsigned int el = 0; el < reg_size; el++)
+            // {
+            //     region.data[reg[i].x + reg[i].y * width] = ls_count;
+            // }
+        }
+    }
+}
+
+void LineSegmentDetectorImpl::ll_angle(const double& threshold,
+                                   const unsigned int& n_bins,
+                                   std::vector<coorlist>& list)
+{
+    //Initialize data
+    angles = Mat_<double>(scaled_image.size());
+    modgrad = Mat_<double>(scaled_image.size());
+
+    angles_data = angles.ptr<double>(0);
+    modgrad_data = modgrad.ptr<double>(0);
+    scaled_image_data = scaled_image.ptr<double>(0);
+
+    img_width = scaled_image.cols;
+    img_height = scaled_image.rows;
+
+    // Undefined the down and right boundaries
+    angles.row(img_height - 1).setTo(NOTDEF);
+    angles.col(img_width - 1).setTo(NOTDEF);
+
+    // Computing gradient for remaining pixels
+    CV_Assert(scaled_image.isContinuous() &&
+              modgrad.isContinuous() &&
+              angles.isContinuous());   // Accessing image data linearly
+
+    double max_grad = -1;
+    for(int y = 0; y < img_height - 1; ++y)
+    {
+        for(int addr = y * img_width, addr_end = addr + img_width - 1; addr < addr_end; ++addr)
+        {
+            double DA = scaled_image_data[addr + img_width + 1] - scaled_image_data[addr];
+            double BC = scaled_image_data[addr + 1] - scaled_image_data[addr + img_width];
+            double gx = DA + BC;    // gradient x component
+            double gy = DA - BC;    // gradient y component
+            double norm = std::sqrt((gx * gx + gy * gy) / 4); // gradient norm
+
+            modgrad_data[addr] = norm;    // store gradient
+
+            if (norm <= threshold)  // norm too small, gradient no defined
+            {
+                angles_data[addr] = NOTDEF;
+            }
+            else
+            {
+                angles_data[addr] = fastAtan2(float(gx), float(-gy)) * DEG_TO_RADS;  // gradient angle computation
+                if (norm > max_grad) { max_grad = norm; }
+            }
+
+        }
+    }
+
+    // Compute histogram of gradient values
+    list = std::vector<coorlist>(img_width * img_height);
+    std::vector<coorlist*> range_s(n_bins);
+    std::vector<coorlist*> range_e(n_bins);
+    unsigned int count = 0;
+    double bin_coef = (max_grad > 0) ? double(n_bins - 1) / max_grad : 0; // If all image is smooth, max_grad <= 0
+
+    for(int y = 0; y < img_height - 1; ++y)
+    {
+        const double* norm = modgrad_data + y * img_width;
+        for(int x = 0; x < img_width - 1; ++x, ++norm)
+        {
+            // Store the point in the right bin according to its norm
+            int i = int((*norm) * bin_coef);
+            if(!range_e[i])
+            {
+                range_e[i] = range_s[i] = &list[count];
+                ++count;
+            }
+            else
+            {
+                range_e[i]->next = &list[count];
+                range_e[i] = &list[count];
+                ++count;
+            }
+            range_e[i]->p = Point(x, y);
+            range_e[i]->next = 0;
+        }
+    }
+
+    // Sort
+    int idx = n_bins - 1;
+    for(;idx > 0 && !range_s[idx]; --idx);
+    coorlist* start = range_s[idx];
+    coorlist* end = range_e[idx];
+    if(start)
+    {
+        while(idx > 0)
+        {
+            --idx;
+            if(range_s[idx])
+            {
+                end->next = range_s[idx];
+                end = range_e[idx];
+            }
+        }
+    }
+}
+
+void LineSegmentDetectorImpl::region_grow(const Point2i& s, std::vector<RegionPoint>& reg,
+                                      int& reg_size, double& reg_angle, const double& prec)
+{
+    // Point to this region
+    reg_size = 1;
+    reg[0].x = s.x;
+    reg[0].y = s.y;
+    int addr = s.x + s.y * img_width;
+    reg[0].used = used.data + addr;
+    reg_angle = angles_data[addr];
+    reg[0].angle = reg_angle;
+    reg[0].modgrad = modgrad_data[addr];
+
+    float sumdx = float(std::cos(reg_angle));
+    float sumdy = float(std::sin(reg_angle));
+    *reg[0].used = USED;
+
+    //Try neighboring regions
+    for(int i = 0; i < reg_size; ++i)
+    {
+        const RegionPoint& rpoint = reg[i];
+        int xx_min = std::max(rpoint.x - 1, 0), xx_max = std::min(rpoint.x + 1, img_width - 1);
+        int yy_min = std::max(rpoint.y - 1, 0), yy_max = std::min(rpoint.y + 1, img_height - 1);
+        for(int yy = yy_min; yy <= yy_max; ++yy)
+        {
+            int c_addr = xx_min + yy * img_width;
+            for(int xx = xx_min; xx <= xx_max; ++xx, ++c_addr)
+            {
+                if((used.data[c_addr] != USED) &&
+                   (isAligned(c_addr, reg_angle, prec)))
+                {
+                    // Add point
+                    used.data[c_addr] = USED;
+                    RegionPoint& region_point = reg[reg_size];
+                    region_point.x = xx;
+                    region_point.y = yy;
+                    region_point.used = &(used.data[c_addr]);
+                    region_point.modgrad = modgrad_data[c_addr];
+                    const double& angle = angles_data[c_addr];
+                    region_point.angle = angle;
+                    ++reg_size;
+
+                    // Update region's angle
+                    sumdx += cos(float(angle));
+                    sumdy += sin(float(angle));
+                    // reg_angle is used in the isAligned, so it needs to be updates?
+                    reg_angle = fastAtan2(sumdy, sumdx) * DEG_TO_RADS;
+                }
+            }
+        }
+    }
+}
+
+void LineSegmentDetectorImpl::region2rect(const std::vector<RegionPoint>& reg, const int reg_size,
+                                      const double reg_angle, const double prec, const double p, rect& rec) const
+{
+    double x = 0, y = 0, sum = 0;
+    for(int i = 0; i < reg_size; ++i)
+    {
+        const RegionPoint& pnt = reg[i];
+        const double& weight = pnt.modgrad;
+        x += double(pnt.x) * weight;
+        y += double(pnt.y) * weight;
+        sum += weight;
+    }
+
+    // Weighted sum must differ from 0
+    CV_Assert(sum > 0);
+
+    x /= sum;
+    y /= sum;
+
+    double theta = get_theta(reg, reg_size, x, y, reg_angle, prec);
+
+    // Find length and width
+    double dx = cos(theta);
+    double dy = sin(theta);
+    double l_min = 0, l_max = 0, w_min = 0, w_max = 0;
+
+    for(int i = 0; i < reg_size; ++i)
+    {
+        double regdx = double(reg[i].x) - x;
+        double regdy = double(reg[i].y) - y;
+
+        double l = regdx * dx + regdy * dy;
+        double w = -regdx * dy + regdy * dx;
+
+        if(l > l_max) l_max = l;
+        else if(l < l_min) l_min = l;
+        if(w > w_max) w_max = w;
+        else if(w < w_min) w_min = w;
+    }
+
+    // Store values
+    rec.x1 = x + l_min * dx;
+    rec.y1 = y + l_min * dy;
+    rec.x2 = x + l_max * dx;
+    rec.y2 = y + l_max * dy;
+    rec.width = w_max - w_min;
+    rec.x = x;
+    rec.y = y;
+    rec.theta = theta;
+    rec.dx = dx;
+    rec.dy = dy;
+    rec.prec = prec;
+    rec.p = p;
+
+    // Min width of 1 pixel
+    if(rec.width < 1.0) rec.width = 1.0;
+}
+
+double LineSegmentDetectorImpl::get_theta(const std::vector<RegionPoint>& reg, const int& reg_size, const double& x,
+                                      const double& y, const double& reg_angle, const double& prec) const
+{
+    double Ixx = 0.0;
+    double Iyy = 0.0;
+    double Ixy = 0.0;
+
+    // Compute inertia matrix
+    for(int i = 0; i < reg_size; ++i)
+    {
+        const double& regx = reg[i].x;
+        const double& regy = reg[i].y;
+        const double& weight = reg[i].modgrad;
+        double dx = regx - x;
+        double dy = regy - y;
+        Ixx += dy * dy * weight;
+        Iyy += dx * dx * weight;
+        Ixy -= dx * dy * weight;
+    }
+
+    // Check if inertia matrix is null
+    CV_Assert(!(double_equal(Ixx, 0) && double_equal(Iyy, 0) && double_equal(Ixy, 0)));
+
+    // Compute smallest eigenvalue
+    double lambda = 0.5 * (Ixx + Iyy - sqrt((Ixx - Iyy) * (Ixx - Iyy) + 4.0 * Ixy * Ixy));
+
+    // Compute angle
+    double theta = (fabs(Ixx)>fabs(Iyy))?
+                    double(fastAtan2(float(lambda - Ixx), float(Ixy))):
+                    double(fastAtan2(float(Ixy), float(lambda - Iyy))); // in degs
+    theta *= DEG_TO_RADS;
+
+    // Correct angle by 180 deg if necessary
+    if(angle_diff(theta, reg_angle) > prec) { theta += CV_PI; }
+
+    return theta;
+}
+
+bool LineSegmentDetectorImpl::refine(std::vector<RegionPoint>& reg, int& reg_size, double reg_angle,
+                                 const double prec, double p, rect& rec, const double& density_th)
+{
+    double density = double(reg_size) / (dist(rec.x1, rec.y1, rec.x2, rec.y2) * rec.width);
+
+    if (density >= density_th) { return true; }
+
+    // Try to reduce angle tolerance
+    double xc = double(reg[0].x);
+    double yc = double(reg[0].y);
+    const double& ang_c = reg[0].angle;
+    double sum = 0, s_sum = 0;
+    int n = 0;
+
+    for (int i = 0; i < reg_size; ++i)
+    {
+        *(reg[i].used) = NOTUSED;
+        if (dist(xc, yc, reg[i].x, reg[i].y) < rec.width)
+        {
+            const double& angle = reg[i].angle;
+            double ang_d = angle_diff_signed(angle, ang_c);
+            sum += ang_d;
+            s_sum += ang_d * ang_d;
+            ++n;
+        }
+    }
+    double mean_angle = sum / double(n);
+    // 2 * standard deviation
+    double tau = 2.0 * sqrt((s_sum - 2.0 * mean_angle * sum) / double(n) + mean_angle * mean_angle);
+
+    // Try new region
+    region_grow(Point(reg[0].x, reg[0].y), reg, reg_size, reg_angle, tau);
+
+    if (reg_size < 2) { return false; }
+
+    region2rect(reg, reg_size, reg_angle, prec, p, rec);
+    density = double(reg_size) / (dist(rec.x1, rec.y1, rec.x2, rec.y2) * rec.width);
+
+    if (density < density_th)
+    {
+        return reduce_region_radius(reg, reg_size, reg_angle, prec, p, rec, density, density_th);
+    }
+    else
+    {
+        return true;
+    }
+}
+
+bool LineSegmentDetectorImpl::reduce_region_radius(std::vector<RegionPoint>& reg, int& reg_size, double reg_angle,
+                const double prec, double p, rect& rec, double density, const double& density_th)
+{
+    // Compute region's radius
+    double xc = double(reg[0].x);
+    double yc = double(reg[0].y);
+    double radSq1 = distSq(xc, yc, rec.x1, rec.y1);
+    double radSq2 = distSq(xc, yc, rec.x2, rec.y2);
+    double radSq = radSq1 > radSq2 ? radSq1 : radSq2;
+
+    while(density < density_th)
+    {
+        radSq *= 0.75*0.75; // Reduce region's radius to 75% of its value
+        // Remove points from the region and update 'used' map
+        for(int i = 0; i < reg_size; ++i)
+        {
+            if(distSq(xc, yc, double(reg[i].x), double(reg[i].y)) > radSq)
+            {
+                // Remove point from the region
+                *(reg[i].used) = NOTUSED;
+                std::swap(reg[i], reg[reg_size - 1]);
+                --reg_size;
+                --i; // To avoid skipping one point
+            }
+        }
+
+        if(reg_size < 2) { return false; }
+
+        // Re-compute rectangle
+        region2rect(reg, reg_size ,reg_angle, prec, p, rec);
+
+        // Re-compute region points density
+        density = double(reg_size) /
+                  (dist(rec.x1, rec.y1, rec.x2, rec.y2) * rec.width);
+    }
+
+    return true;
+}
+
+double LineSegmentDetectorImpl::rect_improve(rect& rec) const
+{
+    double delta = 0.5;
+    double delta_2 = delta / 2.0;
+
+    double log_nfa = rect_nfa(rec);
+
+    if(log_nfa > LOG_EPS) return log_nfa; // Good rectangle
+
+    // Try to improve
+    // Finer precision
+    rect r = rect(rec); // Copy
+    for(int n = 0; n < 5; ++n)
+    {
+        r.p /= 2;
+        r.prec = r.p * CV_PI;
+        double log_nfa_new = rect_nfa(r);
+        if(log_nfa_new > log_nfa)
+        {
+            log_nfa = log_nfa_new;
+            rec = rect(r);
+        }
+    }
+    if(log_nfa > LOG_EPS) return log_nfa;
+
+    // Try to reduce width
+    r = rect(rec);
+    for(unsigned int n = 0; n < 5; ++n)
+    {
+        if((r.width - delta) >= 0.5)
+        {
+            r.width -= delta;
+            double log_nfa_new = rect_nfa(r);
+            if(log_nfa_new > log_nfa)
+            {
+                rec = rect(r);
+                log_nfa = log_nfa_new;
+            }
+        }
+    }
+    if(log_nfa > LOG_EPS) return log_nfa;
+
+    // Try to reduce one side of rectangle
+    r = rect(rec);
+    for(unsigned int n = 0; n < 5; ++n)
+    {
+        if((r.width - delta) >= 0.5)
+        {
+            r.x1 += -r.dy * delta_2;
+            r.y1 +=  r.dx * delta_2;
+            r.x2 += -r.dy * delta_2;
+            r.y2 +=  r.dx * delta_2;
+            r.width -= delta;
+            double log_nfa_new = rect_nfa(r);
+            if(log_nfa_new > log_nfa)
+            {
+                rec = rect(r);
+                log_nfa = log_nfa_new;
+            }
+        }
+    }
+    if(log_nfa > LOG_EPS) return log_nfa;
+
+    // Try to reduce other side of rectangle
+    r = rect(rec);
+    for(unsigned int n = 0; n < 5; ++n)
+    {
+        if((r.width - delta) >= 0.5)
+        {
+            r.x1 -= -r.dy * delta_2;
+            r.y1 -=  r.dx * delta_2;
+            r.x2 -= -r.dy * delta_2;
+            r.y2 -=  r.dx * delta_2;
+            r.width -= delta;
+            double log_nfa_new = rect_nfa(r);
+            if(log_nfa_new > log_nfa)
+            {
+                rec = rect(r);
+                log_nfa = log_nfa_new;
+            }
+        }
+    }
+    if(log_nfa > LOG_EPS) return log_nfa;
+
+    // Try finer precision
+    r = rect(rec);
+    for(unsigned int n = 0; n < 5; ++n)
+    {
+        if((r.width - delta) >= 0.5)
+        {
+            r.p /= 2;
+            r.prec = r.p * CV_PI;
+            double log_nfa_new = rect_nfa(r);
+            if(log_nfa_new > log_nfa)
+            {
+                rec = rect(r);
+                log_nfa = log_nfa_new;
+            }
+        }
+    }
+
+    return log_nfa;
+}
+
+double LineSegmentDetectorImpl::rect_nfa(const rect& rec) const
+{
+    int total_pts = 0, alg_pts = 0;
+    double half_width = rec.width / 2.0;
+    double dyhw = rec.dy * half_width;
+    double dxhw = rec.dx * half_width;
+
+    std::vector<edge> ordered_x(4);
+    edge* min_y = &ordered_x[0];
+    edge* max_y = &ordered_x[0]; // Will be used for loop range
+
+    ordered_x[0].p.x = int(rec.x1 - dyhw); ordered_x[0].p.y = int(rec.y1 + dxhw); ordered_x[0].taken = false;
+    ordered_x[1].p.x = int(rec.x2 - dyhw); ordered_x[1].p.y = int(rec.y2 + dxhw); ordered_x[1].taken = false;
+    ordered_x[2].p.x = int(rec.x2 + dyhw); ordered_x[2].p.y = int(rec.y2 - dxhw); ordered_x[2].taken = false;
+    ordered_x[3].p.x = int(rec.x1 + dyhw); ordered_x[3].p.y = int(rec.y1 - dxhw); ordered_x[3].taken = false;
+
+    std::sort(ordered_x.begin(), ordered_x.end(), AsmallerB_XoverY);
+
+    // Find min y. And mark as taken. find max y.
+    for(unsigned int i = 1; i < 4; ++i)
+    {
+        if(min_y->p.y > ordered_x[i].p.y) {min_y = &ordered_x[i]; }
+        if(max_y->p.y < ordered_x[i].p.y) {max_y = &ordered_x[i]; }
+    }
+    min_y->taken = true;
+
+    // Find leftmost untaken point;
+    edge* leftmost = 0;
+    for(unsigned int i = 0; i < 4; ++i)
+    {
+        if(!ordered_x[i].taken)
+        {
+            if(!leftmost) // if uninitialized
+            {
+                leftmost = &ordered_x[i];
+            }
+            else if (leftmost->p.x > ordered_x[i].p.x)
+            {
+                leftmost = &ordered_x[i];
+            }
+        }
+    }
+    leftmost->taken = true;
+
+    // Find rightmost untaken point;
+    edge* rightmost = 0;
+    for(unsigned int i = 0; i < 4; ++i)
+    {
+        if(!ordered_x[i].taken)
+        {
+            if(!rightmost) // if uninitialized
+            {
+                rightmost = &ordered_x[i];
+            }
+            else if (rightmost->p.x < ordered_x[i].p.x)
+            {
+                rightmost = &ordered_x[i];
+            }
+        }
+    }
+    rightmost->taken = true;
+
+    // Find last untaken point;
+    edge* tailp = 0;
+    for(unsigned int i = 0; i < 4; ++i)
+    {
+        if(!ordered_x[i].taken)
+        {
+            if(!tailp) // if uninitialized
+            {
+                tailp = &ordered_x[i];
+            }
+            else if (tailp->p.x > ordered_x[i].p.x)
+            {
+                tailp = &ordered_x[i];
+            }
+        }
+    }
+    tailp->taken = true;
+
+    double flstep = (min_y->p.y != leftmost->p.y) ?
+                    (min_y->p.x - leftmost->p.x) / (min_y->p.y - leftmost->p.y) : 0; //first left step
+    double slstep = (leftmost->p.y != tailp->p.x) ?
+                    (leftmost->p.x - tailp->p.x) / (leftmost->p.y - tailp->p.x) : 0; //second left step
+
+    double frstep = (min_y->p.y != rightmost->p.y) ?
+                    (min_y->p.x - rightmost->p.x) / (min_y->p.y - rightmost->p.y) : 0; //first right step
+    double srstep = (rightmost->p.y != tailp->p.x) ?
+                    (rightmost->p.x - tailp->p.x) / (rightmost->p.y - tailp->p.x) : 0; //second right step
+
+    double lstep = flstep, rstep = frstep;
+
+    double left_x = min_y->p.x, right_x = min_y->p.x;
+
+    // Loop around all points in the region and count those that are aligned.
+    int min_iter = std::max(min_y->p.y, 0);
+    int max_iter = std::min(max_y->p.y, img_height - 1);
+    for(int y = min_iter; y <= max_iter; ++y)
+    {
+        int adx = y * img_width + int(left_x);
+        for(int x = int(left_x); x <= int(right_x); ++x, ++adx)
+        {
+            ++total_pts;
+            if(isAligned(adx, rec.theta, rec.prec))
+            {
+                ++alg_pts;
+            }
+        }
+
+        if(y >= leftmost->p.y) { lstep = slstep; }
+        if(y >= rightmost->p.y) { rstep = srstep; }
+
+        left_x += lstep;
+        right_x += rstep;
+    }
+
+    return nfa(total_pts, alg_pts, rec.p);
+}
+
+double LineSegmentDetectorImpl::nfa(const int& n, const int& k, const double& p) const
+{
+    // Trivial cases
+    if(n == 0 || k == 0) { return -LOG_NT; }
+    if(n == k) { return -LOG_NT - double(n) * log10(p); }
+
+    double p_term = p / (1 - p);
+
+    double log1term = (double(n) + 1) - log_gamma(double(k) + 1)
+                - log_gamma(double(n-k) + 1)
+                + double(k) * log(p) + double(n-k) * log(1.0 - p);
+    double term = exp(log1term);
+
+    if(double_equal(term, 0))
+    {
+        if(k > n * p) return -log1term / M_LN10 - LOG_NT;
+        else return -LOG_NT;
+    }
+
+    // Compute more terms if needed
+    double bin_tail = term;
+    double tolerance = 0.1; // an error of 10% in the result is accepted
+    for(int i = k + 1; i <= n; ++i)
+    {
+        double bin_term = double(n - i + 1) / double(i);
+        double mult_term = bin_term * p_term;
+        term *= mult_term;
+        bin_tail += term;
+        if(bin_term < 1)
+        {
+            double err = term * ((1 - pow(mult_term, double(n-i+1))) / (1 - mult_term) - 1);
+            if(err < tolerance * fabs(-log10(bin_tail) - LOG_NT) * bin_tail) break;
+        }
+
+    }
+    return -log10(bin_tail) - LOG_NT;
+}
+
+inline bool LineSegmentDetectorImpl::isAligned(const int& address, const double& theta, const double& prec) const
+{
+    if(address < 0) { return false; }
+    const double& a = angles_data[address];
+    if(a == NOTDEF) { return false; }
+
+    // It is assumed that 'theta' and 'a' are in the range [-pi,pi]
+    double n_theta = theta - a;
+    if(n_theta < 0) { n_theta = -n_theta; }
+    if(n_theta > M_3_2_PI)
+    {
+        n_theta -= M_2__PI;
+        if(n_theta < 0) n_theta = -n_theta;
+    }
+
+    return n_theta <= prec;
+}
+
+
+void LineSegmentDetectorImpl::drawSegments(InputOutputArray _image, InputArray lines)
+{
+    CV_Assert(!_image.empty() && (_image.channels() == 1 || _image.channels() == 3));
+
+    Mat gray;
+    if (_image.channels() == 1)
+    {
+        gray = _image.getMatRef();
+    }
+    else if (_image.channels() == 3)
+    {
+        cvtColor(_image, gray, CV_BGR2GRAY);
+    }
+
+    // Create a 3 channel image in order to draw colored lines
+    std::vector<Mat> planes;
+    planes.push_back(gray);
+    planes.push_back(gray);
+    planes.push_back(gray);
+
+    merge(planes, _image);
+
+    Mat _lines;
+    _lines = lines.getMat();
+
+    // Draw segments
+    for(int i = 0; i < _lines.size().width; ++i)
+    {
+        const Vec4i& v = _lines.at<Vec4i>(i);
+        Point b(v[0], v[1]);
+        Point e(v[2], v[3]);
+        line(_image.getMatRef(), b, e, Scalar(0, 0, 255), 1);
+    }
+}
+
+
+int LineSegmentDetectorImpl::compareSegments(const Size& size, InputArray lines1, InputArray lines2, InputOutputArray _image)
+{
+    Size sz = size;
+    if (_image.needed() && _image.size() != size) sz = _image.size();
+    CV_Assert(sz.area());
+
+    Mat_<uchar> I1 = Mat_<uchar>::zeros(sz);
+    Mat_<uchar> I2 = Mat_<uchar>::zeros(sz);
+
+    Mat _lines1;
+    Mat _lines2;
+    _lines1 = lines1.getMat();
+    _lines2 = lines2.getMat();
+    // Draw segments
+    std::vector<Mat> _lines;
+    for(int i = 0; i < _lines1.size().width; ++i)
+    {
+        Point b(_lines1.at<Vec4i>(i)[0], _lines1.at<Vec4i>(i)[1]);
+        Point e(_lines1.at<Vec4i>(i)[2], _lines1.at<Vec4i>(i)[3]);
+        line(I1, b, e, Scalar::all(255), 1);
+    }
+    for(int i = 0; i < _lines2.size().width; ++i)
+    {
+        Point b(_lines2.at<Vec4i>(i)[0], _lines2.at<Vec4i>(i)[1]);
+        Point e(_lines2.at<Vec4i>(i)[2], _lines2.at<Vec4i>(i)[3]);
+        line(I2, b, e, Scalar::all(255), 1);
+    }
+
+    // Count the pixels that don't agree
+    Mat Ixor;
+    bitwise_xor(I1, I2, Ixor);
+    int N = countNonZero(Ixor);
+
+    if (_image.needed())
+    {
+        CV_Assert(_image.channels() == 3);
+        Mat img = _image.getMatRef();
+        CV_Assert(img.isContinuous() && I1.isContinuous() && I2.isContinuous());
+
+        for (unsigned int i = 0; i < I1.total(); ++i)
+        {
+            uchar i1 = I1.data[i];
+            uchar i2 = I2.data[i];
+            if (i1 || i2)
+            {
+                unsigned int base_idx = i * 3;
+                if (i1) img.data[base_idx] = 255;
+                else img.data[base_idx] = 0;
+                img.data[base_idx + 1] = 0;
+                if (i2) img.data[base_idx + 2] = 255;
+                else img.data[base_idx + 2] = 0;
+            }
+        }
+    }
+
+    return N;
+}
+
+} // namespace cv
diff --git a/modules/imgproc/src/matchcontours.cpp b/modules/imgproc/src/matchcontours.cpp
index eca3859001..1ac6c16627 100644
--- a/modules/imgproc/src/matchcontours.cpp
+++ b/modules/imgproc/src/matchcontours.cpp
@@ -142,7 +142,7 @@ double cv::matchShapes(InputArray contour1, InputArray contour2, int method, dou
     default:
         CV_Error( CV_StsBadArg, "Unknown comparison method" );
     }
-    
+
     return result;
 }
 
diff --git a/modules/imgproc/src/moments.cpp b/modules/imgproc/src/moments.cpp
index c5c9c218a5..14e672abdb 100644
--- a/modules/imgproc/src/moments.cpp
+++ b/modules/imgproc/src/moments.cpp
@@ -48,12 +48,12 @@ static void completeMomentState( Moments* moments )
 {
     double cx = 0, cy = 0;
     double mu20, mu11, mu02;
-
+    double inv_m00 = 0.0;
     assert( moments != 0 );
 
     if( fabs(moments->m00) > DBL_EPSILON )
     {
-        double inv_m00 = 1. / moments->m00;
+        inv_m00 = 1. / moments->m00;
         cx = moments->m10 * inv_m00;
         cy = moments->m01 * inv_m00;
     }
@@ -78,6 +78,14 @@ static void completeMomentState( Moments* moments )
     moments->mu12 = moments->m12 - cy * (mu11 + cy * moments->m10) - cx * mu02;
     // mu03 = m03 - cy*(3*mu02 + cy*m01)
     moments->mu03 = moments->m03 - cy * (3 * mu02 + cy * moments->m01);
+
+
+    double inv_sqrt_m00 = std::sqrt(std::abs(inv_m00));
+    double s2 = inv_m00*inv_m00, s3 = s2*inv_sqrt_m00;
+
+    moments->nu20 = moments->mu20*s2; moments->nu11 = moments->mu11*s2; moments->nu02 = moments->mu02*s2;
+    moments->nu30 = moments->mu30*s3; moments->nu21 = moments->mu21*s3; moments->nu12 = moments->mu12*s3; moments->nu03 = moments->mu03*s3;
+
 }
 
 
@@ -151,7 +159,7 @@ static Moments contourMoments( const Mat& contour )
     if( fabs(a00) > FLT_EPSILON )
     {
         double db1_2, db1_6, db1_12, db1_24, db1_20, db1_60;
-        
+
         if( a00 > 0 )
         {
             db1_2 = 0.5;
@@ -354,25 +362,6 @@ Moments::Moments( double _m00, double _m10, double _m01, double _m20, double _m1
     nu30 = mu30*s3; nu21 = mu21*s3; nu12 = mu12*s3; nu03 = mu03*s3;
 }
 
-Moments::Moments( const CvMoments& m )
-{
-    *this = Moments(m.m00, m.m10, m.m01, m.m20, m.m11, m.m02, m.m30, m.m21, m.m12, m.m03);
-}
-
-Moments::operator CvMoments() const
-{
-    CvMoments m;
-    m.m00 = m00; m.m10 = m10; m.m01 = m01;
-    m.m20 = m20; m.m11 = m11; m.m02 = m02;
-    m.m30 = m30; m.m21 = m21; m.m12 = m12; m.m03 = m03;
-    m.mu20 = mu20; m.mu11 = mu11; m.mu02 = mu02;
-    m.mu30 = mu30; m.mu21 = mu21; m.mu12 = mu12; m.mu03 = mu03;
-    double am00 = std::abs(m00);
-    m.inv_sqrt_m00 = am00 > DBL_EPSILON ? 1./std::sqrt(am00) : 0;
-
-    return m;
-}
-
 }
 
 
@@ -475,7 +464,7 @@ cv::Moments cv::moments( InputArray _src, bool binary )
             m.m03 += mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
         }
     }
-    
+
     completeMomentState( &m );
     return m;
 }
diff --git a/modules/imgproc/src/morph.cpp b/modules/imgproc/src/morph.cpp
index 4a939d9bb7..845e001249 100644
--- a/modules/imgproc/src/morph.cpp
+++ b/modules/imgproc/src/morph.cpp
@@ -857,42 +857,42 @@ cv::Ptr<cv::BaseRowFilter> cv::getMorphologyRowFilter(int op, int type, int ksiz
     if( op == MORPH_ERODE )
     {
         if( depth == CV_8U )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MinOp<uchar>,
-                                      ErodeRowVec8u>(ksize, anchor));
+            return makePtr<MorphRowFilter<MinOp<uchar>,
+                                      ErodeRowVec8u> >(ksize, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MinOp<ushort>,
-                                      ErodeRowVec16u>(ksize, anchor));
+            return makePtr<MorphRowFilter<MinOp<ushort>,
+                                      ErodeRowVec16u> >(ksize, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MinOp<short>,
-                                      ErodeRowVec16s>(ksize, anchor));
+            return makePtr<MorphRowFilter<MinOp<short>,
+                                      ErodeRowVec16s> >(ksize, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MinOp<float>,
-                                      ErodeRowVec32f>(ksize, anchor));
+            return makePtr<MorphRowFilter<MinOp<float>,
+                                      ErodeRowVec32f> >(ksize, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MinOp<double>,
-                                      ErodeRowVec64f>(ksize, anchor));
+            return makePtr<MorphRowFilter<MinOp<double>,
+                                      ErodeRowVec64f> >(ksize, anchor);
     }
     else
     {
         if( depth == CV_8U )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MaxOp<uchar>,
-                                      DilateRowVec8u>(ksize, anchor));
+            return makePtr<MorphRowFilter<MaxOp<uchar>,
+                                      DilateRowVec8u> >(ksize, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MaxOp<ushort>,
-                                      DilateRowVec16u>(ksize, anchor));
+            return makePtr<MorphRowFilter<MaxOp<ushort>,
+                                      DilateRowVec16u> >(ksize, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MaxOp<short>,
-                                      DilateRowVec16s>(ksize, anchor));
+            return makePtr<MorphRowFilter<MaxOp<short>,
+                                      DilateRowVec16s> >(ksize, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MaxOp<float>,
-                                      DilateRowVec32f>(ksize, anchor));
+            return makePtr<MorphRowFilter<MaxOp<float>,
+                                      DilateRowVec32f> >(ksize, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseRowFilter>(new MorphRowFilter<MaxOp<double>,
-                                      DilateRowVec64f>(ksize, anchor));
+            return makePtr<MorphRowFilter<MaxOp<double>,
+                                      DilateRowVec64f> >(ksize, anchor);
     }
 
     CV_Error_( CV_StsNotImplemented, ("Unsupported data type (=%d)", type));
-    return Ptr<BaseRowFilter>(0);
+    return Ptr<BaseRowFilter>();
 }
 
 cv::Ptr<cv::BaseColumnFilter> cv::getMorphologyColumnFilter(int op, int type, int ksize, int anchor)
@@ -904,42 +904,42 @@ cv::Ptr<cv::BaseColumnFilter> cv::getMorphologyColumnFilter(int op, int type, in
     if( op == MORPH_ERODE )
     {
         if( depth == CV_8U )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MinOp<uchar>,
-                                         ErodeColumnVec8u>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MinOp<uchar>,
+                                         ErodeColumnVec8u> >(ksize, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MinOp<ushort>,
-                                         ErodeColumnVec16u>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MinOp<ushort>,
+                                         ErodeColumnVec16u> >(ksize, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MinOp<short>,
-                                         ErodeColumnVec16s>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MinOp<short>,
+                                         ErodeColumnVec16s> >(ksize, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MinOp<float>,
-                                         ErodeColumnVec32f>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MinOp<float>,
+                                         ErodeColumnVec32f> >(ksize, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MinOp<double>,
-                                         ErodeColumnVec64f>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MinOp<double>,
+                                         ErodeColumnVec64f> >(ksize, anchor);
     }
     else
     {
         if( depth == CV_8U )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MaxOp<uchar>,
-                                         DilateColumnVec8u>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MaxOp<uchar>,
+                                         DilateColumnVec8u> >(ksize, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MaxOp<ushort>,
-                                         DilateColumnVec16u>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MaxOp<ushort>,
+                                         DilateColumnVec16u> >(ksize, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MaxOp<short>,
-                                         DilateColumnVec16s>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MaxOp<short>,
+                                         DilateColumnVec16s> >(ksize, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MaxOp<float>,
-                                         DilateColumnVec32f>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MaxOp<float>,
+                                         DilateColumnVec32f> >(ksize, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseColumnFilter>(new MorphColumnFilter<MaxOp<double>,
-                                         DilateColumnVec64f>(ksize, anchor));
+            return makePtr<MorphColumnFilter<MaxOp<double>,
+                                         DilateColumnVec64f> >(ksize, anchor);
     }
 
     CV_Error_( CV_StsNotImplemented, ("Unsupported data type (=%d)", type));
-    return Ptr<BaseColumnFilter>(0);
+    return Ptr<BaseColumnFilter>();
 }
 
 
@@ -952,32 +952,32 @@ cv::Ptr<cv::BaseFilter> cv::getMorphologyFilter(int op, int type, InputArray _ke
     if( op == MORPH_ERODE )
     {
         if( depth == CV_8U )
-            return Ptr<BaseFilter>(new MorphFilter<MinOp<uchar>, ErodeVec8u>(kernel, anchor));
+            return makePtr<MorphFilter<MinOp<uchar>, ErodeVec8u> >(kernel, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseFilter>(new MorphFilter<MinOp<ushort>, ErodeVec16u>(kernel, anchor));
+            return makePtr<MorphFilter<MinOp<ushort>, ErodeVec16u> >(kernel, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseFilter>(new MorphFilter<MinOp<short>, ErodeVec16s>(kernel, anchor));
+            return makePtr<MorphFilter<MinOp<short>, ErodeVec16s> >(kernel, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseFilter>(new MorphFilter<MinOp<float>, ErodeVec32f>(kernel, anchor));
+            return makePtr<MorphFilter<MinOp<float>, ErodeVec32f> >(kernel, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseFilter>(new MorphFilter<MinOp<double>, ErodeVec64f>(kernel, anchor));
+            return makePtr<MorphFilter<MinOp<double>, ErodeVec64f> >(kernel, anchor);
     }
     else
     {
         if( depth == CV_8U )
-            return Ptr<BaseFilter>(new MorphFilter<MaxOp<uchar>, DilateVec8u>(kernel, anchor));
+            return makePtr<MorphFilter<MaxOp<uchar>, DilateVec8u> >(kernel, anchor);
         if( depth == CV_16U )
-            return Ptr<BaseFilter>(new MorphFilter<MaxOp<ushort>, DilateVec16u>(kernel, anchor));
+            return makePtr<MorphFilter<MaxOp<ushort>, DilateVec16u> >(kernel, anchor);
         if( depth == CV_16S )
-            return Ptr<BaseFilter>(new MorphFilter<MaxOp<short>, DilateVec16s>(kernel, anchor));
+            return makePtr<MorphFilter<MaxOp<short>, DilateVec16s> >(kernel, anchor);
         if( depth == CV_32F )
-            return Ptr<BaseFilter>(new MorphFilter<MaxOp<float>, DilateVec32f>(kernel, anchor));
+            return makePtr<MorphFilter<MaxOp<float>, DilateVec32f> >(kernel, anchor);
         if( depth == CV_64F )
-            return Ptr<BaseFilter>(new MorphFilter<MaxOp<double>, DilateVec64f>(kernel, anchor));
+            return makePtr<MorphFilter<MaxOp<double>, DilateVec64f> >(kernel, anchor);
     }
 
     CV_Error_( CV_StsNotImplemented, ("Unsupported data type (=%d)", type));
-    return Ptr<BaseFilter>(0);
+    return Ptr<BaseFilter>();
 }
 
 
@@ -1020,8 +1020,8 @@ cv::Ptr<cv::FilterEngine> cv::createMorphologyFilter( int op, int type, InputArr
                                        depth == CV_32F ? (double)-FLT_MAX : -DBL_MAX);
     }
 
-    return Ptr<FilterEngine>(new FilterEngine(filter2D, rowFilter, columnFilter,
-                                              type, type, type, _rowBorderType, _columnBorderType, borderValue ));
+    return makePtr<FilterEngine>(filter2D, rowFilter, columnFilter,
+                                 type, type, type, _rowBorderType, _columnBorderType, borderValue );
 }
 
 
@@ -1081,7 +1081,7 @@ cv::Mat cv::getStructuringElement(int shape, Size ksize, Point anchor)
 namespace cv
 {
 
-class MorphologyRunner
+class MorphologyRunner : public ParallelLoopBody
 {
 public:
     MorphologyRunner(Mat _src, Mat _dst, int _nStripes, int _iterations,
@@ -1102,14 +1102,14 @@ public:
         columnBorderType = _columnBorderType;
     }
 
-    void operator () ( const BlockedRange& range ) const
+    void operator () ( const Range& range ) const
     {
-        int row0 = std::min(cvRound(range.begin() * src.rows / nStripes), src.rows);
-        int row1 = std::min(cvRound(range.end() * src.rows / nStripes), src.rows);
+        int row0 = std::min(cvRound(range.start * src.rows / nStripes), src.rows);
+        int row1 = std::min(cvRound(range.end * src.rows / nStripes), src.rows);
 
         /*if(0)
             printf("Size = (%d, %d), range[%d,%d), row0 = %d, row1 = %d\n",
-                   src.rows, src.cols, range.begin(), range.end(), row0, row1);*/
+                   src.rows, src.cols, range.start, range.end, row0, row1);*/
 
         Mat srcStripe = src.rowRange(row0, row1);
         Mat dstStripe = dst.rowRange(row0, row1);
@@ -1136,17 +1136,171 @@ private:
     Scalar borderValue;
 };
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+static bool IPPMorphReplicate(int op, const Mat &src, Mat &dst, const Mat &kernel,
+                              const Size& ksize, const Point &anchor, bool rectKernel)
+{
+    int type = src.type();
+    const Mat* _src = &src;
+    Mat temp;
+    if( src.data == dst.data )
+    {
+        src.copyTo(temp);
+        _src = &temp;
+    }
+    //DEPRECATED. Allocates and initializes morphology state structure for erosion or dilation operation.
+    typedef IppStatus (CV_STDCALL* ippiMorphologyInitAllocFunc)(int, const void*, IppiSize, IppiPoint, IppiMorphState **);
+    typedef IppStatus (CV_STDCALL* ippiMorphologyBorderReplicateFunc)(const void*, int, void *, int,
+                                                                      IppiSize, IppiBorderType, IppiMorphState *);
+    typedef IppStatus (CV_STDCALL* ippiFilterMinMaxGetBufferSizeFunc)(int, IppiSize, int*);
+    typedef IppStatus (CV_STDCALL* ippiFilterMinMaxBorderReplicateFunc)(const void*, int, void*, int,
+                                                                        IppiSize, IppiSize, IppiPoint, void*);
+
+    ippiMorphologyInitAllocFunc initAllocFunc = 0;
+    ippiMorphologyBorderReplicateFunc morphFunc = 0;
+    ippiFilterMinMaxGetBufferSizeFunc getBufSizeFunc = 0;
+    ippiFilterMinMaxBorderReplicateFunc morphRectFunc = 0;
+
+    #define IPP_MORPH_CASE(type, flavor) \
+    case type: \
+        initAllocFunc = (ippiMorphologyInitAllocFunc)ippiMorphologyInitAlloc_##flavor; \
+        morphFunc = op == MORPH_ERODE ? (ippiMorphologyBorderReplicateFunc)ippiErodeBorderReplicate_##flavor : \
+                                        (ippiMorphologyBorderReplicateFunc)ippiDilateBorderReplicate_##flavor; \
+        getBufSizeFunc = (ippiFilterMinMaxGetBufferSizeFunc)ippiFilterMinGetBufferSize_##flavor; \
+        morphRectFunc = op == MORPH_ERODE ? (ippiFilterMinMaxBorderReplicateFunc)ippiFilterMinBorderReplicate_##flavor : \
+                                            (ippiFilterMinMaxBorderReplicateFunc)ippiFilterMaxBorderReplicate_##flavor; \
+        break
+
+    switch( type )
+    {
+    IPP_MORPH_CASE(CV_8UC1, 8u_C1R);
+    IPP_MORPH_CASE(CV_8UC3, 8u_C3R);
+    IPP_MORPH_CASE(CV_8UC4, 8u_C4R);
+    IPP_MORPH_CASE(CV_32FC1, 32f_C1R);
+    IPP_MORPH_CASE(CV_32FC3, 32f_C3R);
+    IPP_MORPH_CASE(CV_32FC4, 32f_C4R);
+    default:
+        return false;
+    }
+    #undef IPP_MORPH_CASE
+
+    IppiSize roiSize = {src.cols, src.rows};
+    IppiSize kernelSize = {ksize.width, ksize.height};
+    IppiPoint point = {anchor.x, anchor.y};
+
+    if( !rectKernel && morphFunc && initAllocFunc )
+    {
+        IppiMorphState* pState;
+        if( initAllocFunc( roiSize.width, kernel.data, kernelSize, point, &pState ) < 0 )
+            return false;
+        bool is_ok = morphFunc( _src->data, (int)_src->step[0],
+                               dst.data, (int)dst.step[0],
+                               roiSize, ippBorderRepl, pState ) >= 0;
+        ippiMorphologyFree(pState);
+        return is_ok;
+    }
+    else if( rectKernel && morphRectFunc && getBufSizeFunc )
+    {
+        int bufSize = 0;
+        if( getBufSizeFunc( src.cols, kernelSize, &bufSize) < 0 )
+            return false;
+        AutoBuffer<uchar> buf(bufSize + 64);
+        uchar* buffer = alignPtr((uchar*)buf, 32);
+        return morphRectFunc(_src->data, (int)_src->step[0], dst.data, (int)dst.step[0],
+                             roiSize, kernelSize, point, buffer) >= 0;
+    }
+    return false;
+}
+
+static bool IPPMorphOp(int op, InputArray _src, OutputArray _dst,
+    const Mat& _kernel, Point anchor, int iterations,
+    int borderType, const Scalar &borderValue)
+{
+    Mat src = _src.getMat(), kernel = _kernel;
+    if( !( src.depth() == CV_8U || src.depth() == CV_32F ) || ( iterations > 1 ) ||
+        !( borderType == cv::BORDER_REPLICATE || (borderType == cv::BORDER_CONSTANT && borderValue == morphologyDefaultBorderValue()) )
+        || !( op == MORPH_DILATE || op == MORPH_ERODE) )
+        return false;
+    if( borderType == cv::BORDER_CONSTANT && kernel.data )
+    {
+        int x, y;
+        for( y = 0; y < kernel.rows; y++ )
+        {
+            if( kernel.at<uchar>(y, anchor.x) != 0 )
+                continue;
+            for( x = 0; x < kernel.cols; x++ )
+            {
+                if( kernel.at<uchar>(y,x) != 0 )
+                    return false;
+            }
+        }
+        for( x = 0; y < kernel.cols; x++ )
+        {
+            if( kernel.at<uchar>(anchor.y, x) != 0 )
+                continue;
+            for( y = 0; y < kernel.rows; y++ )
+            {
+                if( kernel.at<uchar>(y,x) != 0 )
+                    return false;
+            }
+        }
+
+    }
+    Size ksize = kernel.data ? kernel.size() : Size(3,3);
+
+    _dst.create( src.size(), src.type() );
+    Mat dst = _dst.getMat();
+
+    if( iterations == 0 || kernel.rows*kernel.cols == 1 )
+    {
+        src.copyTo(dst);
+        return true;
+    }
+
+    bool rectKernel = false;
+    if( !kernel.data )
+    {
+        ksize = Size(1+iterations*2,1+iterations*2);
+        anchor = Point(iterations, iterations);
+        rectKernel = true;
+        iterations = 1;
+    }
+    else if( iterations >= 1 && countNonZero(kernel) == kernel.rows*kernel.cols )
+    {
+        ksize = Size(ksize.width + (iterations-1)*(ksize.width-1),
+             ksize.height + (iterations-1)*(ksize.height-1)),
+        anchor = Point(anchor.x*iterations, anchor.y*iterations);
+        kernel = Mat();
+        rectKernel = true;
+        iterations = 1;
+    }
+
+    // TODO: implement the case of iterations > 1.
+    if( iterations > 1 )
+        return false;
+
+    return IPPMorphReplicate( op, src, dst, kernel, ksize, anchor, rectKernel );
+}
+#endif
+
 static void morphOp( int op, InputArray _src, OutputArray _dst,
                      InputArray _kernel,
                      Point anchor, int iterations,
                      int borderType, const Scalar& borderValue )
 {
-    Mat src = _src.getMat(), kernel = _kernel.getMat();
+    Mat kernel = _kernel.getMat();
     Size ksize = kernel.data ? kernel.size() : Size(3,3);
     anchor = normalizeAnchor(anchor, ksize);
 
     CV_Assert( anchor.inside(Rect(0, 0, ksize.width, ksize.height)) );
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    if( IPPMorphOp(op, _src, _dst, kernel, anchor, iterations, borderType, borderValue) )
+        return;
+#endif
+
+    Mat src = _src.getMat();
+
     _dst.create( src.size(), src.type() );
     Mat dst = _dst.getMat();
 
@@ -1173,15 +1327,15 @@ static void morphOp( int op, InputArray _src, OutputArray _dst,
     }
 
     int nStripes = 1;
-#if defined HAVE_TBB && defined HAVE_TEGRA_OPTIMIZATION
+#if defined HAVE_TEGRA_OPTIMIZATION
     if (src.data != dst.data && iterations == 1 &&  //NOTE: threads are not used for inplace processing
         (borderType & BORDER_ISOLATED) == 0 && //TODO: check border types
         src.rows >= 64 ) //NOTE: just heuristics
         nStripes = 4;
 #endif
 
-    parallel_for(BlockedRange(0, nStripes),
-                 MorphologyRunner(src, dst, nStripes, iterations, op, kernel, anchor, borderType, borderType, borderValue));
+    parallel_for_(Range(0, nStripes),
+                  MorphologyRunner(src, dst, nStripes, iterations, op, kernel, anchor, borderType, borderType, borderValue));
 
     //Ptr<FilterEngine> f = createMorphologyFilter(op, src.type(),
     //                                             kernel, anchor, borderType, borderType, borderValue );
@@ -1191,9 +1345,6 @@ static void morphOp( int op, InputArray _src, OutputArray _dst,
     //    f->apply( dst, dst );
 }
 
-template<> void Ptr<IplConvKernel>::delete_obj()
-{ cvReleaseStructuringElement(&obj); }
-
 }
 
 void cv::erode( InputArray src, OutputArray dst, InputArray kernel,
@@ -1319,7 +1470,7 @@ static void convertConvKernel( const IplConvKernel* src, cv::Mat& dst, cv::Point
 
     int i, size = src->nRows*src->nCols;
     for( i = 0; i < size; i++ )
-        dst.data[i] = (uchar)src->values[i];
+        dst.data[i] = (uchar)(src->values[i] != 0);
 }
 
 
diff --git a/modules/imgproc/src/precomp.cpp b/modules/imgproc/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/imgproc/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/imgproc/src/precomp.hpp b/modules/imgproc/src/precomp.hpp
index c3e693c31c..a3bbd65dbb 100644
--- a/modules/imgproc/src/precomp.hpp
+++ b/modules/imgproc/src/precomp.hpp
@@ -43,13 +43,12 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/imgproc.hpp"
+#include "opencv2/core/utility.hpp"
+
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/private.hpp"
+
 #include <math.h>
 #include <assert.h>
 #include <string.h>
diff --git a/modules/imgproc/src/pyramids.cpp b/modules/imgproc/src/pyramids.cpp
index e7d315cb52..9ebf11ea89 100644
--- a/modules/imgproc/src/pyramids.cpp
+++ b/modules/imgproc/src/pyramids.cpp
@@ -191,6 +191,7 @@ pyrDown_( const Mat& _src, Mat& _dst, int borderType )
     typedef typename CastOp::type1 WT;
     typedef typename CastOp::rtype T;
 
+    CV_Assert( !_src.empty() );
     Size ssize = _src.size(), dsize = _dst.size();
     int cn = _src.channels();
     int bufstep = (int)alignSize(dsize.width*cn, 16);
@@ -203,7 +204,8 @@ pyrDown_( const Mat& _src, Mat& _dst, int borderType )
     CastOp castOp;
     VecOp vecOp;
 
-    CV_Assert( std::abs(dsize.width*2 - ssize.width) <= 2 &&
+    CV_Assert( ssize.width > 0 && ssize.height > 0 &&
+               std::abs(dsize.width*2 - ssize.width) <= 2 &&
                std::abs(dsize.height*2 - ssize.height) <= 2 );
     int k, x, sy0 = -PD_SZ/2, sy = sy0, width0 = std::min((ssize.width-PD_SZ/2-1)/2 + 1, dsize.width);
 
diff --git a/modules/imgproc/src/rotcalipers.cpp b/modules/imgproc/src/rotcalipers.cpp
index cc43732c26..0983d4d35b 100644
--- a/modules/imgproc/src/rotcalipers.cpp
+++ b/modules/imgproc/src/rotcalipers.cpp
@@ -398,3 +398,9 @@ cvMinAreaRect2( const CvArr* array, CvMemStorage* /*storage*/ )
     return (CvBox2D)rr;
 }
 
+void cv::boxPoints(cv::RotatedRect box, OutputArray _pts)
+{
+    _pts.create(4, 2, CV_32F);
+    Mat pts = _pts.getMat();
+    box.points((Point2f*)pts.data);
+}
diff --git a/modules/imgproc/src/samplers.cpp b/modules/imgproc/src/samplers.cpp
index 80e2706ba9..64b7b6296b 100644
--- a/modules/imgproc/src/samplers.cpp
+++ b/modules/imgproc/src/samplers.cpp
@@ -267,6 +267,11 @@ static void getRectSubPix_8u32f
     }
 }
 
+typedef CvStatus (CV_STDCALL *CvIPPGetRectSubPixFunc)( const void* src, int src_step,
+                                                       CvSize src_size, void* dst,
+                                                       int dst_step, CvSize win_size,
+                                                       CvPoint2D32f center,
+                                                       CvPoint* minpt, CvPoint* maxpt );
 
 static void
 getQuadrangleSubPix_8u32f_CnR( const uchar* src, size_t src_step, Size src_size,
@@ -336,12 +341,12 @@ getQuadrangleSubPix_8u32f_CnR( const uchar* src, size_t src_step, Size src_size,
                 float w00 = a1*b1, w01 = a*b1, w10 = a1*b, w11 = a*b;
                 const uchar *ptr0, *ptr1;
                 xs += A11; ys += A21;
-                
+
                 if( (unsigned)iys < (unsigned)(src_size.height-1) )
                     ptr0 = src + src_step*iys, ptr1 = ptr0 + src_step;
                 else
                     ptr0 = ptr1 = src + (iys < 0 ? 0 : src_size.height-1)*src_step;
-                
+
                 if( (unsigned)ixs < (unsigned)(src_size.width-1) )
                 {
                     ptr0 += ixs*cn; ptr1 += ixs*cn;
@@ -374,6 +379,19 @@ void cv::getRectSubPix( InputArray _image, Size patchSize, Point2f center,
     _patch.create(patchSize, CV_MAKETYPE(ddepth, cn));
     Mat patch = _patch.getMat();
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    CvPoint minpt, maxpt;
+    int srctype = CV_MAT_TYPE(src->type), dsttype = CV_MAT_TYPE(dst->type);
+    CvIPPGetRectSubPixFunc ippfunc =
+        srctype == CV_8UC1 && dsttype == CV_8UC1 ? (CvIPPGetRectSubPixFunc)ippiCopySubpixIntersect_8u_C1R :
+        srctype == CV_8UC1 && dsttype == CV_32FC1 ? (CvIPPGetRectSubPixFunc)ippiCopySubpixIntersect_8u32f_C1R :
+        srctype == CV_32FC1 && dsttype == CV_32FC1 ? (CvIPPGetRectSubPixFunc)ippiCopySubpixIntersect_32f_C1R : 0;
+
+    if( ippfunc && ippfunc(src->data.ptr, src->step, src_size, dst->data.ptr,
+                           dst->step, dst_size, center, &minpt, &maxpt) >= 0 )
+        return;
+#endif
+
     if( depth == CV_8U && ddepth == CV_8U )
         getRectSubPix_Cn_<uchar, uchar, int, scale_fixpt, cast_8u>
         (image.data, image.step, image.size(), patch.data, patch.step, patch.size(), center, cn);
@@ -438,7 +456,7 @@ cvSampleLine( const void* _img, CvPoint pt1, CvPoint pt2,
     cv::LineIterator li(img, pt1, pt2, connectivity, false);
     uchar* buffer = (uchar*)_buffer;
     size_t pixsize = img.elemSize();
-    
+
     if( !buffer )
         CV_Error( CV_StsNullPtr, "" );
 
diff --git a/modules/imgproc/src/segmentation.cpp b/modules/imgproc/src/segmentation.cpp
index e4cd43401c..320f2273af 100644
--- a/modules/imgproc/src/segmentation.cpp
+++ b/modules/imgproc/src/segmentation.cpp
@@ -327,7 +327,7 @@ void cv::pyrMeanShiftFiltering( InputArray _src, OutputArray _dst,
     double sr2 = sr * sr;
     int isr2 = cvRound(sr2), isr22 = MAX(isr2,16);
     int tab[768];
-    
+
 
     if( src0.type() != CV_8UC3 )
         CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 3-channel images are supported" );
@@ -534,4 +534,3 @@ cvPyrMeanShiftFiltering( const CvArr* srcarr, CvArr* dstarr,
 
     cv::pyrMeanShiftFiltering(src, dst, sp0, sr, max_level, termcrit);
 }
-
diff --git a/modules/imgproc/src/shapedescr.cpp b/modules/imgproc/src/shapedescr.cpp
index 93368be965..4258d473a7 100644
--- a/modules/imgproc/src/shapedescr.cpp
+++ b/modules/imgproc/src/shapedescr.cpp
@@ -381,7 +381,7 @@ cv::RotatedRect cv::fitEllipse( InputArray _points )
     // New fitellipse algorithm, contributed by Dr. Daniel Weiss
     Point2f c(0,0);
     double gfp[5], rp[5], t;
-    const double min_eps = 1e-6;
+    const double min_eps = 1e-8;
     bool is_float = depth == CV_32F;
     const Point* ptsi = (const Point*)points.data;
     const Point2f* ptsf = (const Point2f*)points.data;
@@ -1046,7 +1046,7 @@ cvFitEllipse2( const CvArr* array )
 CV_IMPL  CvRect
 cvBoundingRect( CvArr* array, int update )
 {
-    CvRect  rect = { 0, 0, 0, 0 };
+    CvRect  rect;
     CvContour contour_header;
     CvSeq* ptseq = 0;
     CvSeqBlock block;
diff --git a/modules/imgproc/src/smooth.cpp b/modules/imgproc/src/smooth.cpp
index 2986ae20d0..4c53e33458 100644
--- a/modules/imgproc/src/smooth.cpp
+++ b/modules/imgproc/src/smooth.cpp
@@ -622,29 +622,29 @@ cv::Ptr<cv::BaseRowFilter> cv::getRowSumFilter(int srcType, int sumType, int ksi
         anchor = ksize/2;
 
     if( sdepth == CV_8U && ddepth == CV_32S )
-        return Ptr<BaseRowFilter>(new RowSum<uchar, int>(ksize, anchor));
+        return makePtr<RowSum<uchar, int> >(ksize, anchor);
     if( sdepth == CV_8U && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowSum<uchar, double>(ksize, anchor));
+        return makePtr<RowSum<uchar, double> >(ksize, anchor);
     if( sdepth == CV_16U && ddepth == CV_32S )
-        return Ptr<BaseRowFilter>(new RowSum<ushort, int>(ksize, anchor));
+        return makePtr<RowSum<ushort, int> >(ksize, anchor);
     if( sdepth == CV_16U && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowSum<ushort, double>(ksize, anchor));
+        return makePtr<RowSum<ushort, double> >(ksize, anchor);
     if( sdepth == CV_16S && ddepth == CV_32S )
-        return Ptr<BaseRowFilter>(new RowSum<short, int>(ksize, anchor));
+        return makePtr<RowSum<short, int> >(ksize, anchor);
     if( sdepth == CV_32S && ddepth == CV_32S )
-        return Ptr<BaseRowFilter>(new RowSum<int, int>(ksize, anchor));
+        return makePtr<RowSum<int, int> >(ksize, anchor);
     if( sdepth == CV_16S && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowSum<short, double>(ksize, anchor));
+        return makePtr<RowSum<short, double> >(ksize, anchor);
     if( sdepth == CV_32F && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowSum<float, double>(ksize, anchor));
+        return makePtr<RowSum<float, double> >(ksize, anchor);
     if( sdepth == CV_64F && ddepth == CV_64F )
-        return Ptr<BaseRowFilter>(new RowSum<double, double>(ksize, anchor));
+        return makePtr<RowSum<double, double> >(ksize, anchor);
 
     CV_Error_( CV_StsNotImplemented,
         ("Unsupported combination of source format (=%d), and buffer format (=%d)",
         srcType, sumType));
 
-    return Ptr<BaseRowFilter>(0);
+    return Ptr<BaseRowFilter>();
 }
 
 
@@ -658,33 +658,33 @@ cv::Ptr<cv::BaseColumnFilter> cv::getColumnSumFilter(int sumType, int dstType, i
         anchor = ksize/2;
 
     if( ddepth == CV_8U && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, uchar>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, uchar> >(ksize, anchor, scale);
     if( ddepth == CV_8U && sdepth == CV_64F )
-        return Ptr<BaseColumnFilter>(new ColumnSum<double, uchar>(ksize, anchor, scale));
+        return makePtr<ColumnSum<double, uchar> >(ksize, anchor, scale);
     if( ddepth == CV_16U && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, ushort>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, ushort> >(ksize, anchor, scale);
     if( ddepth == CV_16U && sdepth == CV_64F )
-        return Ptr<BaseColumnFilter>(new ColumnSum<double, ushort>(ksize, anchor, scale));
+        return makePtr<ColumnSum<double, ushort> >(ksize, anchor, scale);
     if( ddepth == CV_16S && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, short>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, short> >(ksize, anchor, scale);
     if( ddepth == CV_16S && sdepth == CV_64F )
-        return Ptr<BaseColumnFilter>(new ColumnSum<double, short>(ksize, anchor, scale));
+        return makePtr<ColumnSum<double, short> >(ksize, anchor, scale);
     if( ddepth == CV_32S && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, int>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, int> >(ksize, anchor, scale);
     if( ddepth == CV_32F && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, float>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, float> >(ksize, anchor, scale);
     if( ddepth == CV_32F && sdepth == CV_64F )
-        return Ptr<BaseColumnFilter>(new ColumnSum<double, float>(ksize, anchor, scale));
+        return makePtr<ColumnSum<double, float> >(ksize, anchor, scale);
     if( ddepth == CV_64F && sdepth == CV_32S )
-        return Ptr<BaseColumnFilter>(new ColumnSum<int, double>(ksize, anchor, scale));
+        return makePtr<ColumnSum<int, double> >(ksize, anchor, scale);
     if( ddepth == CV_64F && sdepth == CV_64F )
-        return Ptr<BaseColumnFilter>(new ColumnSum<double, double>(ksize, anchor, scale));
+        return makePtr<ColumnSum<double, double> >(ksize, anchor, scale);
 
     CV_Error_( CV_StsNotImplemented,
         ("Unsupported combination of sum format (=%d), and destination format (=%d)",
         sumType, dstType));
 
-    return Ptr<BaseColumnFilter>(0);
+    return Ptr<BaseColumnFilter>();
 }
 
 
@@ -703,8 +703,8 @@ cv::Ptr<cv::FilterEngine> cv::createBoxFilter( int srcType, int dstType, Size ks
     Ptr<BaseColumnFilter> columnFilter = getColumnSumFilter(sumType,
         dstType, ksize.height, anchor.y, normalize ? 1./(ksize.width*ksize.height) : 1);
 
-    return Ptr<FilterEngine>(new FilterEngine(Ptr<BaseFilter>(0), rowFilter, columnFilter,
-           srcType, dstType, sumType, borderType ));
+    return makePtr<FilterEngine>(Ptr<BaseFilter>(), rowFilter, columnFilter,
+           srcType, dstType, sumType, borderType );
 }
 
 
@@ -856,6 +856,22 @@ void cv::GaussianBlur( InputArray _src, OutputArray _dst, Size ksize,
         return;
 #endif
 
+#if defined HAVE_IPP && (IPP_VERSION_MAJOR >= 7)
+    if(src.type() == CV_32FC1 && sigma1 == sigma2 && ksize.width == ksize.height && sigma1 != 0.0 )
+    {
+        IppiSize roi = {src.cols, src.rows};
+        int bufSize = 0;
+        ippiFilterGaussGetBufferSize_32f_C1R(roi, ksize.width, &bufSize);
+        AutoBuffer<uchar> buf(bufSize+128);
+        if( ippiFilterGaussBorder_32f_C1R((const Ipp32f *)src.data, (int)src.step,
+                                          (Ipp32f *)dst.data, (int)dst.step,
+                                          roi, ksize.width, (Ipp32f)sigma1,
+                                          (IppiBorderType)borderType, 0.0,
+                                          alignPtr(&buf[0],32)) >= 0 )
+            return;
+    }
+#endif
+
     Ptr<FilterEngine> f = createGaussianFilter( src.type(), ksize, sigma1, sigma2, borderType );
     f->apply( src, dst );
 }
@@ -1772,7 +1788,7 @@ public:
                         sum += val*w;
                         wsum += w;
                     }
-                    // overflow is not possible here => there is no need to use CV_CAST_8U
+                    // overflow is not possible here => there is no need to use cv::saturate_cast
                     dptr[j] = (uchar)cvRound(sum/wsum);
                 }
             }
@@ -1863,6 +1879,41 @@ private:
     float *space_weight, *color_weight;
 };
 
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+class IPPBilateralFilter_8u_Invoker :
+    public ParallelLoopBody
+{
+public:
+    IPPBilateralFilter_8u_Invoker(Mat &_src, Mat &_dst, double _sigma_color, double _sigma_space, int _radius, bool *_ok) :
+      ParallelLoopBody(), src(_src), dst(_dst), sigma_color(_sigma_color), sigma_space(_sigma_space), radius(_radius), ok(_ok)
+      {
+          *ok = true;
+      }
+
+      virtual void operator() (const Range& range) const
+      {
+          int d = radius * 2 + 1;
+          IppiSize kernel = {d, d};
+          IppiSize roi={dst.cols, range.end - range.start};
+          int bufsize=0;
+          ippiFilterBilateralGetBufSize_8u_C1R( ippiFilterBilateralGauss, roi, kernel, &bufsize);
+          AutoBuffer<uchar> buf(bufsize);
+          IppiFilterBilateralSpec *pSpec = (IppiFilterBilateralSpec *)alignPtr(&buf[0], 32);
+          ippiFilterBilateralInit_8u_C1R( ippiFilterBilateralGauss, kernel, (Ipp32f)sigma_color, (Ipp32f)sigma_space, 1, pSpec );
+          if( ippiFilterBilateral_8u_C1R( src.ptr<uchar>(range.start) + radius * ((int)src.step[0] + 1), (int)src.step[0], dst.ptr<uchar>(range.start), (int)dst.step[0], roi, kernel, pSpec ) < 0)
+              *ok = false;
+      }
+private:
+    Mat &src;
+    Mat &dst;
+    double sigma_color;
+    double sigma_space;
+    int radius;
+    bool *ok;
+    const IPPBilateralFilter_8u_Invoker& operator= (const IPPBilateralFilter_8u_Invoker&);
+};
+#endif
+
 static void
 bilateralFilter_8u( const Mat& src, Mat& dst, int d,
     double sigma_color, double sigma_space,
@@ -1895,6 +1946,16 @@ bilateralFilter_8u( const Mat& src, Mat& dst, int d,
     Mat temp;
     copyMakeBorder( src, temp, radius, radius, radius, radius, borderType );
 
+#if defined HAVE_IPP && (IPP_VERSION_MAJOR >= 7)
+    if( cn == 1 )
+    {
+        bool ok;
+        IPPBilateralFilter_8u_Invoker body(temp, dst, sigma_color * sigma_color, sigma_space * sigma_space, radius, &ok );
+        parallel_for_(Range(0, dst.rows), body, dst.total()/(double)(1<<16));
+        if( ok ) return;
+    }
+#endif
+
     std::vector<float> _color_weight(cn*256);
     std::vector<float> _space_weight(d*d);
     std::vector<int> _space_ofs(d*d);
@@ -2219,6 +2280,236 @@ void cv::bilateralFilter( InputArray _src, OutputArray _dst, int d,
         "Bilateral filtering is only implemented for 8u and 32f images" );
 }
 
+
+/****************************************************************************************\
+                                  Adaptive Bilateral Filtering
+\****************************************************************************************/
+
+namespace cv
+{
+#define CALCVAR 1
+#define FIXED_WEIGHT 0
+
+class adaptiveBilateralFilter_8u_Invoker :
+    public ParallelLoopBody
+{
+public:
+    adaptiveBilateralFilter_8u_Invoker(Mat& _dest, const Mat& _temp, Size _ksize, double _sigma_space, Point _anchor) :
+        temp(&_temp), dest(&_dest), ksize(_ksize), sigma_space(_sigma_space), anchor(_anchor)
+    {
+        if( sigma_space <= 0 )
+            sigma_space = 1;
+        CV_Assert((ksize.width & 1) && (ksize.height & 1));
+        space_weight.resize(ksize.width * ksize.height);
+        double sigma2 = sigma_space * sigma_space;
+        int idx = 0;
+        int w = ksize.width / 2;
+        int h = ksize.height / 2;
+        for(int y=-h; y<=h; y++)
+            for(int x=-w; x<=w; x++)
+        {
+            space_weight[idx++] = (float)(sigma2 / (sigma2 + x * x + y * y));
+        }
+    }
+    virtual void operator()(const Range& range) const
+    {
+        int cn = dest->channels();
+        int anX = anchor.x;
+
+        const uchar *tptr;
+
+        for(int i = range.start;i < range.end; i++)
+        {
+            int startY = i;
+            if(cn == 1)
+            {
+                float var;
+                int currVal;
+                int sumVal = 0;
+                int sumValSqr = 0;
+                int currValCenter;
+                int currWRTCenter;
+                float weight;
+                float totalWeight = 0.;
+                float tmpSum = 0.;
+
+                for(int j = 0;j < dest->cols *cn; j+=cn)
+                {
+                    sumVal = 0;
+                    sumValSqr= 0;
+                    totalWeight = 0.;
+                    tmpSum = 0.;
+
+                    // Top row: don't sum the very last element
+                    int startLMJ = 0;
+                    int endLMJ  = ksize.width  - 1;
+                    int howManyAll = (anX *2 +1)*(ksize.width );
+#if CALCVAR
+                    for(int x = startLMJ; x< endLMJ; x++)
+                    {
+                        tptr = temp->ptr(startY + x) +j;
+                        for(int y=-anX; y<=anX; y++)
+                        {
+                            currVal = tptr[cn*(y+anX)];
+                            sumVal += currVal;
+                            sumValSqr += (currVal *currVal);
+                        }
+                    }
+                    var = ( (sumValSqr * howManyAll)- sumVal * sumVal )  /  ( (float)(howManyAll*howManyAll));
+#else
+                    var = 900.0;
+#endif
+                    startLMJ = 0;
+                    endLMJ = ksize.width;
+                    tptr = temp->ptr(startY + (startLMJ+ endLMJ)/2);
+                    currValCenter =tptr[j+cn*anX];
+                    for(int x = startLMJ; x< endLMJ; x++)
+                    {
+                        tptr = temp->ptr(startY + x) +j;
+                        for(int y=-anX; y<=anX; y++)
+                        {
+#if FIXED_WEIGHT
+                            weight = 1.0;
+#else
+                            currVal = tptr[cn*(y+anX)];
+                            currWRTCenter = currVal - currValCenter;
+
+                            weight = var / ( var + (currWRTCenter * currWRTCenter) ) * space_weight[x*ksize.width+y+anX];;
+#endif
+                            tmpSum += ((float)tptr[cn*(y+anX)] * weight);
+                            totalWeight += weight;
+                        }
+                    }
+                    tmpSum /= totalWeight;
+
+                   dest->at<uchar>(startY ,j)= static_cast<uchar>(tmpSum);
+                }
+            }
+            else
+            {
+                assert(cn == 3);
+                float var_b, var_g, var_r;
+                int currVal_b, currVal_g, currVal_r;
+                int sumVal_b= 0, sumVal_g= 0, sumVal_r= 0;
+                int sumValSqr_b= 0, sumValSqr_g= 0, sumValSqr_r= 0;
+                int currValCenter_b= 0, currValCenter_g= 0, currValCenter_r= 0;
+                int currWRTCenter_b, currWRTCenter_g, currWRTCenter_r;
+                float weight_b, weight_g, weight_r;
+                float totalWeight_b= 0., totalWeight_g= 0., totalWeight_r= 0.;
+                float tmpSum_b = 0., tmpSum_g= 0., tmpSum_r = 0.;
+
+                for(int j = 0;j < dest->cols *cn; j+=cn)
+                {
+                    sumVal_b= 0, sumVal_g= 0, sumVal_r= 0;
+                    sumValSqr_b= 0, sumValSqr_g= 0, sumValSqr_r= 0;
+                    totalWeight_b= 0., totalWeight_g= 0., totalWeight_r= 0.;
+                    tmpSum_b = 0., tmpSum_g= 0., tmpSum_r = 0.;
+
+                    // Top row: don't sum the very last element
+                    int startLMJ = 0;
+                    int endLMJ  = ksize.width - 1;
+                    int howManyAll = (anX *2 +1)*(ksize.width);
+#if CALCVAR
+                    for(int x = startLMJ; x< endLMJ; x++)
+                    {
+                        tptr = temp->ptr(startY + x) +j;
+                        for(int y=-anX; y<=anX; y++)
+                        {
+                            currVal_b = tptr[cn*(y+anX)], currVal_g = tptr[cn*(y+anX)+1], currVal_r =tptr[cn*(y+anX)+2];
+                            sumVal_b += currVal_b;
+                            sumVal_g += currVal_g;
+                            sumVal_r += currVal_r;
+                            sumValSqr_b += (currVal_b *currVal_b);
+                            sumValSqr_g += (currVal_g *currVal_g);
+                            sumValSqr_r += (currVal_r *currVal_r);
+                        }
+                    }
+                    var_b = ( (sumValSqr_b * howManyAll)- sumVal_b * sumVal_b )  /  ( (float)(howManyAll*howManyAll));
+                    var_g = ( (sumValSqr_g * howManyAll)- sumVal_g * sumVal_g )  /  ( (float)(howManyAll*howManyAll));
+                    var_r = ( (sumValSqr_r * howManyAll)- sumVal_r * sumVal_r )  /  ( (float)(howManyAll*howManyAll));
+#else
+                    var_b = 900.0; var_g = 900.0;var_r = 900.0;
+#endif
+                    startLMJ = 0;
+                    endLMJ = ksize.width;
+                    tptr = temp->ptr(startY + (startLMJ+ endLMJ)/2) + j;
+                    currValCenter_b =tptr[cn*anX], currValCenter_g =tptr[cn*anX+1], currValCenter_r =tptr[cn*anX+2];
+                    for(int x = startLMJ; x< endLMJ; x++)
+                    {
+                        tptr = temp->ptr(startY + x) +j;
+                        for(int y=-anX; y<=anX; y++)
+                        {
+#if FIXED_WEIGHT
+                            weight_b = 1.0;
+                            weight_g = 1.0;
+                            weight_r = 1.0;
+#else
+                            currVal_b = tptr[cn*(y+anX)];currVal_g=tptr[cn*(y+anX)+1];currVal_r=tptr[cn*(y+anX)+2];
+                            currWRTCenter_b = currVal_b - currValCenter_b;
+                            currWRTCenter_g = currVal_g - currValCenter_g;
+                            currWRTCenter_r = currVal_r - currValCenter_r;
+
+                            float cur_spw = space_weight[x*ksize.width+y+anX];
+                            weight_b = var_b / ( var_b + (currWRTCenter_b * currWRTCenter_b) ) * cur_spw;
+                            weight_g = var_g / ( var_g + (currWRTCenter_g * currWRTCenter_g) ) * cur_spw;
+                            weight_r = var_r / ( var_r + (currWRTCenter_r * currWRTCenter_r) ) * cur_spw;
+#endif
+                            tmpSum_b += ((float)tptr[cn*(y+anX)]   * weight_b);
+                            tmpSum_g += ((float)tptr[cn*(y+anX)+1] * weight_g);
+                            tmpSum_r += ((float)tptr[cn*(y+anX)+2] * weight_r);
+                            totalWeight_b += weight_b, totalWeight_g += weight_g, totalWeight_r += weight_r;
+                        }
+                    }
+                    tmpSum_b /= totalWeight_b;
+                    tmpSum_g /= totalWeight_g;
+                    tmpSum_r /= totalWeight_r;
+
+                    dest->at<uchar>(startY,j  )= static_cast<uchar>(tmpSum_b);
+                    dest->at<uchar>(startY,j+1)= static_cast<uchar>(tmpSum_g);
+                    dest->at<uchar>(startY,j+2)= static_cast<uchar>(tmpSum_r);
+                }
+            }
+        }
+    }
+private:
+    const Mat *temp;
+    Mat *dest;
+    Size ksize;
+    double sigma_space;
+    Point anchor;
+    std::vector<float> space_weight;
+};
+static void adaptiveBilateralFilter_8u( const Mat& src, Mat& dst, Size ksize, double sigmaSpace, Point anchor, int borderType )
+{
+    Size size = src.size();
+
+    CV_Assert( (src.type() == CV_8UC1 || src.type() == CV_8UC3) &&
+              src.type() == dst.type() && src.size() == dst.size() &&
+              src.data != dst.data );
+    Mat temp;
+    copyMakeBorder(src, temp, anchor.x, anchor.y, anchor.x, anchor.y, borderType);
+
+    adaptiveBilateralFilter_8u_Invoker body(dst, temp, ksize, sigmaSpace, anchor);
+    parallel_for_(Range(0, size.height), body, dst.total()/(double)(1<<16));
+}
+}
+void cv::adaptiveBilateralFilter( InputArray _src, OutputArray _dst, Size ksize,
+                                  double sigmaSpace, Point anchor, int borderType )
+{
+    Mat src = _src.getMat();
+    _dst.create(src.size(), src.type());
+    Mat dst = _dst.getMat();
+
+    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC3);
+
+    anchor = normalizeAnchor(anchor,ksize);
+    if( src.depth() == CV_8U )
+        adaptiveBilateralFilter_8u( src, dst, ksize, sigmaSpace, anchor, borderType );
+    else
+        CV_Error( CV_StsUnsupportedFormat,
+        "Adaptive Bilateral filtering is only implemented for 8u images" );
+}
+
 //////////////////////////////////////////////////////////////////////////////////////////
 
 CV_IMPL void
diff --git a/modules/imgproc/src/sumpixels.cpp b/modules/imgproc/src/sumpixels.cpp
index b44197047a..229bbcb8c6 100644
--- a/modules/imgproc/src/sumpixels.cpp
+++ b/modules/imgproc/src/sumpixels.cpp
@@ -41,6 +41,9 @@
 //M*/
 
 #include "precomp.hpp"
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+static IppStatus sts = ippInit();
+#endif
 
 namespace cv
 {
@@ -234,6 +237,53 @@ void cv::integral( InputArray _src, OutputArray _sum, OutputArray _sqsum, Output
     if( sdepth <= 0 )
         sdepth = depth == CV_8U ? CV_32S : CV_64F;
     sdepth = CV_MAT_DEPTH(sdepth);
+
+#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
+    if( ( depth == CV_8U ) && ( !_tilted.needed() ) )
+    {
+        if( sdepth == CV_32F )
+        {
+            if( cn == 1 )
+            {
+                IppiSize srcRoiSize = ippiSize( src.cols, src.rows );
+                _sum.create( isize, CV_MAKETYPE( sdepth, cn ) );
+                sum = _sum.getMat();
+                if( _sqsum.needed() )
+                {
+                    _sqsum.create( isize, CV_MAKETYPE( CV_64F, cn ) );
+                    sqsum = _sqsum.getMat();
+                    ippiSqrIntegral_8u32f64f_C1R( (const Ipp8u*)src.data, src.step, (Ipp32f*)sum.data, sum.step, (Ipp64f*)sqsum.data, sqsum.step, srcRoiSize, 0, 0 );
+                }
+                else
+                {
+                    ippiIntegral_8u32f_C1R( (const Ipp8u*)src.data, src.step, (Ipp32f*)sum.data, sum.step, srcRoiSize, 0 );
+                }
+                return;
+            }
+        }
+        if( sdepth == CV_32S )
+        {
+            if( cn == 1 )
+            {
+                IppiSize srcRoiSize = ippiSize( src.cols, src.rows );
+                _sum.create( isize, CV_MAKETYPE( sdepth, cn ) );
+                sum = _sum.getMat();
+                if( _sqsum.needed() )
+                {
+                    _sqsum.create( isize, CV_MAKETYPE( CV_64F, cn ) );
+                    sqsum = _sqsum.getMat();
+                    ippiSqrIntegral_8u32s64f_C1R( (const Ipp8u*)src.data, src.step, (Ipp32s*)sum.data, sum.step, (Ipp64f*)sqsum.data, sqsum.step, srcRoiSize, 0, 0 );
+                }
+                else
+                {
+                    ippiIntegral_8u32s_C1R( (const Ipp8u*)src.data, src.step, (Ipp32s*)sum.data, sum.step, srcRoiSize, 0 );
+                }
+                return;
+            }
+        }
+    }
+#endif
+
     _sum.create( isize, CV_MAKETYPE(sdepth, cn) );
     sum = _sum.getMat();
 
diff --git a/modules/imgproc/src/templmatch.cpp b/modules/imgproc/src/templmatch.cpp
index f87c296a3f..bfe7ce6000 100644
--- a/modules/imgproc/src/templmatch.cpp
+++ b/modules/imgproc/src/templmatch.cpp
@@ -248,6 +248,8 @@ void cv::matchTemplate( InputArray _img, InputArray _templ, OutputArray _result,
     CV_Assert( (img.depth() == CV_8U || img.depth() == CV_32F) &&
                img.type() == templ.type() );
 
+    CV_Assert( img.rows >= templ.rows && img.cols >= templ.cols);
+
     Size corrSize(img.cols - templ.cols + 1, img.rows - templ.rows + 1);
     _result.create(corrSize, CV_32F);
     Mat result = _result.getMat();
@@ -280,8 +282,7 @@ void cv::matchTemplate( InputArray _img, InputArray _templ, OutputArray _result,
         integral(img, sum, sqsum, CV_64F);
         meanStdDev( templ, templMean, templSdv );
 
-        templNorm = CV_SQR(templSdv[0]) + CV_SQR(templSdv[1]) +
-                    CV_SQR(templSdv[2]) + CV_SQR(templSdv[3]);
+        templNorm = templSdv[0]*templSdv[0] + templSdv[1]*templSdv[1] + templSdv[2]*templSdv[2] + templSdv[3]*templSdv[3];
 
         if( templNorm < DBL_EPSILON && method == CV_TM_CCOEFF_NORMED )
         {
@@ -289,9 +290,7 @@ void cv::matchTemplate( InputArray _img, InputArray _templ, OutputArray _result,
             return;
         }
 
-        templSum2 = templNorm +
-                     CV_SQR(templMean[0]) + CV_SQR(templMean[1]) +
-                     CV_SQR(templMean[2]) + CV_SQR(templMean[3]);
+        templSum2 = templNorm + templMean[0]*templMean[0] + templMean[1]*templMean[1] + templMean[2]*templMean[2] + templMean[3]*templMean[3];
 
         if( numType != 1 )
         {
@@ -335,7 +334,7 @@ void cv::matchTemplate( InputArray _img, InputArray _templ, OutputArray _result,
                 for( k = 0; k < cn; k++ )
                 {
                     t = p0[idx+k] - p1[idx+k] - p2[idx+k] + p3[idx+k];
-                    wndMean2 += CV_SQR(t);
+                    wndMean2 += t*t;
                     num -= t*templMean[k];
                 }
 
diff --git a/modules/imgproc/src/utils.cpp b/modules/imgproc/src/utils.cpp
index 72f4978e93..4f45cde5e0 100644
--- a/modules/imgproc/src/utils.cpp
+++ b/modules/imgproc/src/utils.cpp
@@ -75,196 +75,6 @@ CV_IMPL CvSeq* cvPointSeqFromMat( int seq_kind, const CvArr* arr,
     return (CvSeq*)contour_header;
 }
 
-namespace cv
-{
-
-static void copyMakeBorder_8u( const uchar* src, size_t srcstep, Size srcroi,
-                               uchar* dst, size_t dststep, Size dstroi,
-                               int top, int left, int cn, int borderType )
-{
-    const int isz = (int)sizeof(int);
-    int i, j, k, elemSize = 1;
-    bool intMode = false;
-
-    if( (cn | srcstep | dststep | (size_t)src | (size_t)dst) % isz == 0 )
-    {
-        cn /= isz;
-        elemSize = isz;
-        intMode = true;
-    }
-
-    AutoBuffer<int> _tab((dstroi.width - srcroi.width)*cn);
-    int* tab = _tab;
-    int right = dstroi.width - srcroi.width - left;
-    int bottom = dstroi.height - srcroi.height - top;
-
-    for( i = 0; i < left; i++ )
-    {
-        j = borderInterpolate(i - left, srcroi.width, borderType)*cn;
-        for( k = 0; k < cn; k++ )
-            tab[i*cn + k] = j + k;
-    }
-
-    for( i = 0; i < right; i++ )
-    {
-        j = borderInterpolate(srcroi.width + i, srcroi.width, borderType)*cn;
-        for( k = 0; k < cn; k++ )
-            tab[(i+left)*cn + k] = j + k;
-    }
-
-    srcroi.width *= cn;
-    dstroi.width *= cn;
-    left *= cn;
-    right *= cn;
-
-    uchar* dstInner = dst + dststep*top + left*elemSize;
-
-    for( i = 0; i < srcroi.height; i++, dstInner += dststep, src += srcstep )
-    {
-        if( dstInner != src )
-            memcpy(dstInner, src, srcroi.width*elemSize);
-
-        if( intMode )
-        {
-            const int* isrc = (int*)src;
-            int* idstInner = (int*)dstInner;
-            for( j = 0; j < left; j++ )
-                idstInner[j - left] = isrc[tab[j]];
-            for( j = 0; j < right; j++ )
-                idstInner[j + srcroi.width] = isrc[tab[j + left]];
-        }
-        else
-        {
-            for( j = 0; j < left; j++ )
-                dstInner[j - left] = src[tab[j]];
-            for( j = 0; j < right; j++ )
-                dstInner[j + srcroi.width] = src[tab[j + left]];
-        }
-    }
-
-    dstroi.width *= elemSize;
-    dst += dststep*top;
-
-    for( i = 0; i < top; i++ )
-    {
-        j = borderInterpolate(i - top, srcroi.height, borderType);
-        memcpy(dst + (i - top)*dststep, dst + j*dststep, dstroi.width);
-    }
-
-    for( i = 0; i < bottom; i++ )
-    {
-        j = borderInterpolate(i + srcroi.height, srcroi.height, borderType);
-        memcpy(dst + (i + srcroi.height)*dststep, dst + j*dststep, dstroi.width);
-    }
-}
-
-
-static void copyMakeConstBorder_8u( const uchar* src, size_t srcstep, Size srcroi,
-                                    uchar* dst, size_t dststep, Size dstroi,
-                                    int top, int left, int cn, const uchar* value )
-{
-    int i, j;
-    AutoBuffer<uchar> _constBuf(dstroi.width*cn);
-    uchar* constBuf = _constBuf;
-    int right = dstroi.width - srcroi.width - left;
-    int bottom = dstroi.height - srcroi.height - top;
-
-    for( i = 0; i < dstroi.width; i++ )
-    {
-        for( j = 0; j < cn; j++ )
-            constBuf[i*cn + j] = value[j];
-    }
-
-    srcroi.width *= cn;
-    dstroi.width *= cn;
-    left *= cn;
-    right *= cn;
-
-    uchar* dstInner = dst + dststep*top + left;
-
-    for( i = 0; i < srcroi.height; i++, dstInner += dststep, src += srcstep )
-    {
-        if( dstInner != src )
-            memcpy( dstInner, src, srcroi.width );
-        memcpy( dstInner - left, constBuf, left );
-        memcpy( dstInner + srcroi.width, constBuf, right );
-    }
-
-    dst += dststep*top;
-
-    for( i = 0; i < top; i++ )
-        memcpy(dst + (i - top)*dststep, constBuf, dstroi.width);
-
-    for( i = 0; i < bottom; i++ )
-        memcpy(dst + (i + srcroi.height)*dststep, constBuf, dstroi.width);
-}
-
-}
-
-void cv::copyMakeBorder( InputArray _src, OutputArray _dst, int top, int bottom,
-                         int left, int right, int borderType, const Scalar& value )
-{
-    Mat src = _src.getMat();
-    CV_Assert( top >= 0 && bottom >= 0 && left >= 0 && right >= 0 );
-
-    if( src.isSubmatrix() && (borderType & BORDER_ISOLATED) == 0 )
-    {
-        Size wholeSize;
-        Point ofs;
-        src.locateROI(wholeSize, ofs);
-        int dtop = std::min(ofs.y, top);
-        int dbottom = std::min(wholeSize.height - src.rows - ofs.y, bottom);
-        int dleft = std::min(ofs.x, left);
-        int dright = std::min(wholeSize.width - src.cols - ofs.x, right);
-        src.adjustROI(dtop, dbottom, dleft, dright);
-        top -= dtop;
-        left -= dleft;
-        bottom -= dbottom;
-        right -= dright;
-    }
-
-    _dst.create( src.rows + top + bottom, src.cols + left + right, src.type() );
-    Mat dst = _dst.getMat();
-
-    if(top == 0 && left == 0 && bottom == 0 && right == 0)
-    {
-        if(src.data != dst.data || src.step != dst.step)
-            src.copyTo(dst);
-        return;
-    }
-
-    borderType &= ~BORDER_ISOLATED;
-
-    if( borderType != BORDER_CONSTANT )
-        copyMakeBorder_8u( src.data, src.step, src.size(),
-                           dst.data, dst.step, dst.size(),
-                           top, left, (int)src.elemSize(), borderType );
-    else
-    {
-        int cn = src.channels(), cn1 = cn;
-        AutoBuffer<double> buf(cn);
-        if( cn > 4 )
-        {
-            CV_Assert( value[0] == value[1] && value[0] == value[2] && value[0] == value[3] );
-            cn1 = 1;
-        }
-        scalarToRawData(value, buf, CV_MAKETYPE(src.depth(), cn1), cn);
-        copyMakeConstBorder_8u( src.data, src.step, src.size(),
-                                dst.data, dst.step, dst.size(),
-                                top, left, (int)src.elemSize(), (uchar*)(double*)buf );
-    }
-}
-
-
-double cv::PSNR(InputArray _src1, InputArray _src2)
-{
-    Mat src1 = _src1.getMat(), src2 = _src2.getMat();
-    CV_Assert( src1.depth() == CV_8U );
-    double diff = std::sqrt(norm(src1, src2, NORM_L2SQR)/(src1.total()*src1.channels()));
-    return 20*log10(255./(diff+DBL_EPSILON));
-}
-
-
 CV_IMPL void
 cvCopyMakeBorder( const CvArr* srcarr, CvArr* dstarr, CvPoint offset,
                   int borderType, CvScalar value )
diff --git a/modules/imgproc/test/test_approxpoly.cpp b/modules/imgproc/test/test_approxpoly.cpp
index d087d541e1..c4973c31e6 100644
--- a/modules/imgproc/test/test_approxpoly.cpp
+++ b/modules/imgproc/test/test_approxpoly.cpp
@@ -356,4 +356,3 @@ _exit_:
 }
 
 TEST(Imgproc_ApproxPoly, accuracy) { CV_ApproxPolyTest test; test.safe_run(); }
-
diff --git a/modules/imgproc/test/test_bilateral_filter.cpp b/modules/imgproc/test/test_bilateral_filter.cpp
index 2d45fdcf71..0bfc3dc4cc 100644
--- a/modules/imgproc/test/test_bilateral_filter.cpp
+++ b/modules/imgproc/test/test_bilateral_filter.cpp
@@ -251,7 +251,7 @@ namespace cvtest
 
     int CV_BilateralFilterTest::validate_test_results(int test_case_index)
     {
-        static const double eps = 1;
+        static const double eps = 4;
 
         Mat reference_dst, reference_src;
         if (_src.depth() == CV_32F)
diff --git a/modules/imgproc/test/test_color.cpp b/modules/imgproc/test/test_color.cpp
index 86f9c67564..0c94f8f502 100644
--- a/modules/imgproc/test/test_color.cpp
+++ b/modules/imgproc/test/test_color.cpp
@@ -1687,10 +1687,10 @@ TEST(Imgproc_ColorBayer, regression)
 {
     cvtest::TS* ts = cvtest::TS::ptr();
 
-    Mat given = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", CV_LOAD_IMAGE_GRAYSCALE);
-    Mat gold = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_gold.png", CV_LOAD_IMAGE_UNCHANGED);
+    Mat given = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", IMREAD_GRAYSCALE);
+    Mat gold = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_gold.png", IMREAD_UNCHANGED);
     Mat result;
-    
+
     CV_Assert(given.data != NULL && gold.data != NULL);
 
     cvtColor(given, result, CV_BayerBG2GRAY);
@@ -1709,9 +1709,9 @@ TEST(Imgproc_ColorBayerVNG, regression)
 {
     cvtest::TS* ts = cvtest::TS::ptr();
 
-    Mat given = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", CV_LOAD_IMAGE_GRAYSCALE);
+    Mat given = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", IMREAD_GRAYSCALE);
     string goldfname = string(ts->get_data_path()) + "/cvtcolor/bayerVNG_gold.png";
-    Mat gold = imread(goldfname, CV_LOAD_IMAGE_UNCHANGED);
+    Mat gold = imread(goldfname, IMREAD_UNCHANGED);
     Mat result;
 
     CV_Assert(given.data != NULL);
@@ -1804,7 +1804,7 @@ TEST(Imgproc_ColorBayerVNG_Strict, regression)
 
     Mat src, dst, bayer, reference;
     std::string full_path = parent_path + image_name;
-    src = imread(full_path, CV_LOAD_IMAGE_UNCHANGED);
+    src = imread(full_path, IMREAD_UNCHANGED);
 
     if (src.data == NULL)
     {
@@ -1824,7 +1824,7 @@ TEST(Imgproc_ColorBayerVNG_Strict, regression)
 
         // reading a reference image
         full_path = parent_path + pattern[i] + image_name;
-        reference = imread(full_path, CV_LOAD_IMAGE_UNCHANGED);
+        reference = imread(full_path, IMREAD_UNCHANGED);
         if (reference.data == NULL)
         {
             imwrite(full_path, dst);
@@ -2091,7 +2091,7 @@ TEST(ImgProc_BayerEdgeAwareDemosaicing, accuracy)
 
     Mat src, bayer;
     std::string full_path = parent_path + image_name;
-    src = imread(full_path, CV_LOAD_IMAGE_UNCHANGED);
+    src = imread(full_path, IMREAD_UNCHANGED);
 
     if (src.data == NULL)
     {
@@ -2141,7 +2141,7 @@ TEST(ImgProc_BayerEdgeAwareDemosaicing, accuracy)
 TEST(ImgProc_Bayer2RGBA, accuracy)
 {
     cvtest::TS* ts = cvtest::TS::ptr();
-    Mat raw = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", CV_LOAD_IMAGE_GRAYSCALE);
+    Mat raw = imread(string(ts->get_data_path()) + "/cvtcolor/bayer_input.png", IMREAD_GRAYSCALE);
     Mat rgb, reference;
 
     CV_Assert(raw.channels() == 1);
diff --git a/modules/imgproc/test/test_connectedcomponents.cpp b/modules/imgproc/test/test_connectedcomponents.cpp
index c428cc0742..a9567da970 100644
--- a/modules/imgproc/test/test_connectedcomponents.cpp
+++ b/modules/imgproc/test/test_connectedcomponents.cpp
@@ -105,4 +105,3 @@ void CV_ConnectedComponentsTest::run( int /* start_from */)
 }
 
 TEST(Imgproc_ConnectedComponents, regression) { CV_ConnectedComponentsTest test; test.safe_run(); }
-
diff --git a/modules/imgproc/test/test_convhull.cpp b/modules/imgproc/test/test_convhull.cpp
index 19f536a944..49456c6cc9 100644
--- a/modules/imgproc/test/test_convhull.cpp
+++ b/modules/imgproc/test/test_convhull.cpp
@@ -1626,7 +1626,7 @@ CV_PerimeterAreaSliceTest::~CV_PerimeterAreaSliceTest() {}
 
 void CV_PerimeterAreaSliceTest::run( int )
 {
-    Ptr<CvMemStorage> storage = cvCreateMemStorage();
+    Ptr<CvMemStorage> storage(cvCreateMemStorage());
     RNG& rng = theRNG();
     const double min_r = 90, max_r = 120;
 
@@ -1700,4 +1700,3 @@ TEST(Imgproc_ContourPerimeterSlice, accuracy) { CV_PerimeterAreaSliceTest test;
 TEST(Imgproc_FitEllipse, small) { CV_FitEllipseSmallTest test; test.safe_run(); }
 
 /* End of file. */
-
diff --git a/modules/imgproc/test/test_distancetransform.cpp b/modules/imgproc/test/test_distancetransform.cpp
index 9ce9c6c423..dd3c2e8b41 100644
--- a/modules/imgproc/test/test_distancetransform.cpp
+++ b/modules/imgproc/test/test_distancetransform.cpp
@@ -283,5 +283,3 @@ void CV_DisTransTest::prepare_to_validation( int /*test_case_idx*/ )
 
 
 TEST(Imgproc_DistanceTransform, accuracy) { CV_DisTransTest test; test.safe_run(); }
-
-
diff --git a/modules/imgproc/test/test_filter.cpp b/modules/imgproc/test/test_filter.cpp
index 6bf529b0e7..efbad99749 100644
--- a/modules/imgproc/test/test_filter.cpp
+++ b/modules/imgproc/test/test_filter.cpp
@@ -1886,4 +1886,3 @@ protected:
 };
 
 TEST(Imgproc_Filtering, supportedFormats) { CV_FilterSupportedFormatsTest test; test.safe_run(); }
-
diff --git a/modules/imgproc/test/test_floodfill.cpp b/modules/imgproc/test/test_floodfill.cpp
index 558aa322b6..07e0bcbc4e 100644
--- a/modules/imgproc/test/test_floodfill.cpp
+++ b/modules/imgproc/test/test_floodfill.cpp
@@ -259,16 +259,17 @@ cvTsFloodFill( CvMat* _img, CvPoint seed_pt, CvScalar new_val,
     if( CV_MAT_DEPTH(_img->type) == CV_8U || CV_MAT_DEPTH(_img->type) == CV_32S )
     {
         tmp = cvCreateMat( rows, cols, CV_MAKETYPE(CV_32F,CV_MAT_CN(_img->type)) );
-        cvTsConvert(_img, tmp);
+        cvtest::convert(cvarrToMat(_img), cvarrToMat(tmp), -1);
     }
 
     mask = cvCreateMat( rows + 2, cols + 2, CV_16UC1 );
 
     if( _mask )
-        cvTsConvert( _mask, mask );
+        cvtest::convert(cvarrToMat(_mask), cvarrToMat(mask), -1);
     else
     {
-        cvTsZero( mask );
+        Mat m_mask = cvarrToMat(mask);
+        cvtest::set( m_mask, Scalar::all(0), Mat() );
         cvRectangle( mask, cvPoint(0,0), cvPoint(mask->cols-1,mask->rows-1), Scalar::all(1.), 1, 8, 0 );
     }
 
@@ -481,7 +482,7 @@ _exit_:
     if( tmp != _img )
     {
         if( !mask_only )
-            cvTsConvert(tmp, _img);
+            cvtest::convert(cvarrToMat(tmp), cvarrToMat(_img), -1);
         cvReleaseMat( &tmp );
     }
 
diff --git a/modules/imgproc/test/test_grabcut.cpp b/modules/imgproc/test/test_grabcut.cpp
index f1b1c080df..cfc17dc946 100644
--- a/modules/imgproc/test/test_grabcut.cpp
+++ b/modules/imgproc/test/test_grabcut.cpp
@@ -143,8 +143,8 @@ TEST(Imgproc_GrabCut, repeatability)
 {
     cvtest::TS& ts = *cvtest::TS::ptr();
 
-    Mat image_1 = imread(string(ts.get_data_path()) + "grabcut/image1652.ppm", CV_LOAD_IMAGE_COLOR);
-    Mat mask_1 = imread(string(ts.get_data_path()) + "grabcut/mask1652.ppm", CV_LOAD_IMAGE_GRAYSCALE);
+    Mat image_1 = imread(string(ts.get_data_path()) + "grabcut/image1652.ppm", IMREAD_COLOR);
+    Mat mask_1 = imread(string(ts.get_data_path()) + "grabcut/mask1652.ppm", IMREAD_GRAYSCALE);
     Rect roi_1(0, 0, 150, 150);
 
     Mat image_2 = image_1.clone();
diff --git a/modules/imgproc/test/test_histograms.cpp b/modules/imgproc/test/test_histograms.cpp
index 3ac1e94be0..19ccc656b5 100644
--- a/modules/imgproc/test/test_histograms.cpp
+++ b/modules/imgproc/test/test_histograms.cpp
@@ -630,7 +630,8 @@ void CV_MinMaxHistTest::run_func(void)
 {
     if( hist_type != CV_HIST_ARRAY && test_cpp )
     {
-        cv::SparseMat h((CvSparseMat*)hist[0]->bins);
+        cv::SparseMat h;
+        ((CvSparseMat*)hist[0]->bins)->copyToSparseMat(h);
         double _min_val = 0, _max_val = 0;
         cv::minMaxLoc(h, &_min_val, &_max_val, min_idx, max_idx );
         min_val = (float)_min_val;
@@ -727,10 +728,11 @@ void CV_NormHistTest::run_func(void)
 {
     if( hist_type != CV_HIST_ARRAY && test_cpp )
     {
-        cv::SparseMat h((CvSparseMat*)hist[0]->bins);
+        cv::SparseMat h;
+        ((CvSparseMat*)hist[0]->bins)->copyToSparseMat(h);
         cv::normalize(h, h, factor, CV_L1);
         cvReleaseSparseMat((CvSparseMat**)&hist[0]->bins);
-        hist[0]->bins = (CvSparseMat*)h;
+        hist[0]->bins = cvCreateSparseMat(h);
     }
     else
         cvNormalizeHist( hist[0], factor );
@@ -946,7 +948,7 @@ int CV_ThreshHistTest::validate_test_results( int /*test_case_idx*/ )
 class CV_CompareHistTest : public CV_BaseHistTest
 {
 public:
-    enum { MAX_METHOD = 4 };
+    enum { MAX_METHOD = 5 };
 
     CV_CompareHistTest();
 protected:
@@ -978,8 +980,9 @@ void CV_CompareHistTest::run_func(void)
     int k;
     if( hist_type != CV_HIST_ARRAY && test_cpp )
     {
-        cv::SparseMat h0((CvSparseMat*)hist[0]->bins);
-        cv::SparseMat h1((CvSparseMat*)hist[1]->bins);
+        cv::SparseMat h0, h1;
+        ((CvSparseMat*)hist[0]->bins)->copyToSparseMat(h0);
+        ((CvSparseMat*)hist[1]->bins)->copyToSparseMat(h1);
         for( k = 0; k < MAX_METHOD; k++ )
             result[k] = cv::compareHist(h0, h1, k);
     }
@@ -1011,6 +1014,8 @@ int CV_CompareHistTest::validate_test_results( int /*test_case_idx*/ )
             result0[CV_COMP_INTERSECT] += MIN(v0,v1);
             if( fabs(v0) > DBL_EPSILON )
                 result0[CV_COMP_CHISQR] += (v0 - v1)*(v0 - v1)/v0;
+            if( fabs(v0 + v1) > DBL_EPSILON )
+                result0[CV_COMP_CHISQR_ALT] += (v0 - v1)*(v0 - v1)/(v0 + v1);
             s0 += v0;
             s1 += v1;
             sq0 += v0*v0;
@@ -1036,6 +1041,8 @@ int CV_CompareHistTest::validate_test_results( int /*test_case_idx*/ )
             result0[CV_COMP_INTERSECT] += MIN(v0,v1);
             if( fabs(v0) > DBL_EPSILON )
                 result0[CV_COMP_CHISQR] += (v0 - v1)*(v0 - v1)/v0;
+            if( fabs(v0 + v1) > DBL_EPSILON )
+                result0[CV_COMP_CHISQR_ALT] += (v0 - v1)*(v0 - v1)/(v0 + v1);
             s0 += v0;
             sq0 += v0*v0;
             result0[CV_COMP_BHATTACHARYYA] += sqrt(v0*v1);
@@ -1050,6 +1057,8 @@ int CV_CompareHistTest::validate_test_results( int /*test_case_idx*/ )
         }
     }
 
+    result0[CV_COMP_CHISQR_ALT] *= 2;
+
     t = (sq0 - s0*s0/total_size)*(sq1 - s1*s1/total_size);
     result0[CV_COMP_CORREL] = fabs(t) > DBL_EPSILON ?
         (result0[CV_COMP_CORREL] - s0*s1/total_size)/sqrt(t) : 1;
@@ -1064,6 +1073,7 @@ int CV_CompareHistTest::validate_test_results( int /*test_case_idx*/ )
         double v = result[i], v0 = result0[i];
         const char* method_name =
             i == CV_COMP_CHISQR ? "Chi-Square" :
+            i == CV_COMP_CHISQR_ALT ? "Alternative Chi-Square" :
             i == CV_COMP_CORREL ? "Correlation" :
             i == CV_COMP_INTERSECT ? "Intersection" :
             i == CV_COMP_BHATTACHARYYA ? "Bhattacharyya" : "Unknown";
diff --git a/modules/imgproc/test/test_imgwarp.cpp b/modules/imgproc/test/test_imgwarp.cpp
index 4ea3b76ba8..53f7d5c2fc 100644
--- a/modules/imgproc/test/test_imgwarp.cpp
+++ b/modules/imgproc/test/test_imgwarp.cpp
@@ -1424,7 +1424,7 @@ TEST(Imgproc_fitLine_vector_2d, regression)
 
 TEST(Imgproc_fitLine_Mat_2dC2, regression)
 {
-    cv::Mat mat1(3, 1, CV_32SC2);
+    cv::Mat mat1 = Mat::zeros(3, 1, CV_32SC2);
     std::vector<float> line1;
 
     cv::fitLine(mat1, line1, CV_DIST_L2, 0 ,0 ,0);
@@ -1444,7 +1444,7 @@ TEST(Imgproc_fitLine_Mat_2dC1, regression)
 
 TEST(Imgproc_fitLine_Mat_3dC3, regression)
 {
-    cv::Mat mat1(2, 1, CV_32SC3);
+    cv::Mat mat1 = Mat::zeros(2, 1, CV_32SC3);
     std::vector<float> line1;
 
     cv::fitLine(mat1, line1, CV_DIST_L2, 0 ,0 ,0);
@@ -1454,7 +1454,7 @@ TEST(Imgproc_fitLine_Mat_3dC3, regression)
 
 TEST(Imgproc_fitLine_Mat_3dC1, regression)
 {
-    cv::Mat mat2(2, 3, CV_32SC1);
+    cv::Mat mat2 = Mat::zeros(2, 3, CV_32SC1);
     std::vector<float> line2;
 
     cv::fitLine(mat2, line2, CV_DIST_L2, 0 ,0 ,0);
diff --git a/modules/imgproc/test/test_imgwarp_strict.cpp b/modules/imgproc/test/test_imgwarp_strict.cpp
index 59b851c45e..76d65b198b 100644
--- a/modules/imgproc/test/test_imgwarp_strict.cpp
+++ b/modules/imgproc/test/test_imgwarp_strict.cpp
@@ -47,7 +47,7 @@
 
 using namespace cv;
 
-namespace internal
+namespace
 {
     void __wrap_printf_func(const char* fmt, ...)
     {
@@ -62,7 +62,6 @@ namespace internal
     #define PRINT_TO_LOG __wrap_printf_func
 }
 
-using internal::PRINT_TO_LOG;
 #define SHOW_IMAGE
 #undef SHOW_IMAGE
 
@@ -359,7 +358,7 @@ CV_Resize_Test::~CV_Resize_Test()
 {
 }
 
-namespace internal
+namespace
 {
     void interpolateLinear(float x, float* coeffs)
     {
@@ -523,7 +522,7 @@ void CV_Resize_Test::resize_1d(const Mat& _src, Mat& _dst, int dy, const dim& _d
     }
     else if (interpolation == INTER_LINEAR || interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4)
     {
-        internal::interpolate_method inter_func = internal::inter_array[interpolation - (interpolation == INTER_LANCZOS4 ? 2 : 1)];
+        interpolate_method inter_func = inter_array[interpolation - (interpolation == INTER_LANCZOS4 ? 2 : 1)];
         size_t elemsize = _src.elemSize();
 
         int ofs = 0, ksize = 2;
@@ -679,8 +678,8 @@ void CV_Remap_Test::generate_test_data()
             MatIterator_<Vec2s> begin_x = mapx.begin<Vec2s>(), end_x = mapx.end<Vec2s>();
             for ( ; begin_x != end_x; ++begin_x)
             {
-                begin_x[0] = static_cast<short>(rng.uniform(static_cast<int>(_n), std::max(src.cols + n - 1, 0)));
-                begin_x[1] = static_cast<short>(rng.uniform(static_cast<int>(_n), std::max(src.rows + n - 1, 0)));
+                (*begin_x)[0] = static_cast<short>(rng.uniform(static_cast<int>(_n), std::max(src.cols + n - 1, 0)));
+                (*begin_x)[1] = static_cast<short>(rng.uniform(static_cast<int>(_n), std::max(src.rows + n - 1, 0)));
             }
 
             if (interpolation != INTER_NEAREST)
@@ -874,7 +873,7 @@ void CV_Remap_Test::remap_generic(const Mat& _src, Mat& _dst)
         height1 = std::max(ssize.height - ksize + 1, 0);
 
     float ix[8], w[16];
-    internal::interpolate_method inter_func = internal::inter_array[interpolation - (interpolation == INTER_LANCZOS4 ? 2 : 1)];
+    interpolate_method inter_func = inter_array[interpolation - (interpolation == INTER_LANCZOS4 ? 2 : 1)];
 
     for (int dy = 0; dy < dsize.height; ++dy)
     {
diff --git a/modules/imgproc/test/test_lsd.cpp b/modules/imgproc/test/test_lsd.cpp
new file mode 100644
index 0000000000..7d8b7f243e
--- /dev/null
+++ b/modules/imgproc/test/test_lsd.cpp
@@ -0,0 +1,265 @@
+#include "test_precomp.hpp"
+
+#include <vector>
+
+using namespace cv;
+using namespace std;
+
+const Size img_size(640, 480);
+const int LSD_TEST_SEED = 0x134679;
+const int EPOCHS = 20;
+
+class LSDBase : public testing::Test
+{
+public:
+    LSDBase() {};
+
+protected:
+    Mat test_image;
+    vector<Vec4i> lines;
+    RNG rng;
+    int passedtests;
+
+    void GenerateWhiteNoise(Mat& image);
+    void GenerateConstColor(Mat& image);
+    void GenerateLines(Mat& image, const unsigned int numLines);
+    void GenerateRotatedRect(Mat& image);
+    virtual void SetUp();
+};
+
+class Imgproc_LSD_ADV: public LSDBase
+{
+public:
+    Imgproc_LSD_ADV() {};
+protected:
+
+};
+
+class Imgproc_LSD_STD: public LSDBase
+{
+public:
+    Imgproc_LSD_STD() {};
+protected:
+
+};
+
+class Imgproc_LSD_NONE: public LSDBase
+{
+public:
+    Imgproc_LSD_NONE() {};
+protected:
+
+};
+
+void LSDBase::GenerateWhiteNoise(Mat& image)
+{
+    image = Mat(img_size, CV_8UC1);
+    rng.fill(image, RNG::UNIFORM, 0, 256);
+}
+
+void LSDBase::GenerateConstColor(Mat& image)
+{
+    image = Mat(img_size, CV_8UC1, Scalar::all(rng.uniform(0, 256)));
+}
+
+void LSDBase::GenerateLines(Mat& image, const unsigned int numLines)
+{
+    image = Mat(img_size, CV_8UC1, Scalar::all(rng.uniform(0, 128)));
+
+    for(unsigned int i = 0; i < numLines; ++i)
+    {
+        int y = rng.uniform(10, img_size.width - 10);
+        Point p1(y, 10);
+        Point p2(y, img_size.height - 10);
+        line(image, p1, p2, Scalar(255), 3);
+    }
+}
+
+void LSDBase::GenerateRotatedRect(Mat& image)
+{
+    image = Mat::zeros(img_size, CV_8UC1);
+
+    Point center(rng.uniform(img_size.width/4, img_size.width*3/4),
+                 rng.uniform(img_size.height/4, img_size.height*3/4));
+    Size rect_size(rng.uniform(img_size.width/8, img_size.width/6),
+                   rng.uniform(img_size.height/8, img_size.height/6));
+    float angle = rng.uniform(0.f, 360.f);
+
+    Point2f vertices[4];
+
+    RotatedRect rRect = RotatedRect(center, rect_size, angle);
+
+    rRect.points(vertices);
+    for (int i = 0; i < 4; i++)
+    {
+        line(image, vertices[i], vertices[(i + 1) % 4], Scalar(255), 3);
+    }
+}
+
+void LSDBase::SetUp()
+{
+    lines.clear();
+    test_image = Mat();
+    rng = RNG(LSD_TEST_SEED);
+    passedtests = 0;
+}
+
+
+TEST_F(Imgproc_LSD_ADV, whiteNoise)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateWhiteNoise(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_ADV);
+        detector->detect(test_image, lines);
+
+        if(40u >= lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_ADV, constColor)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateConstColor(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_ADV);
+        detector->detect(test_image, lines);
+
+        if(0u == lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_ADV, lines)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        const unsigned int numOfLines = 1;
+        GenerateLines(test_image, numOfLines);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_ADV);
+        detector->detect(test_image, lines);
+
+        if(numOfLines * 2 == lines.size()) ++passedtests;  // * 2 because of Gibbs effect
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_ADV, rotatedRect)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateRotatedRect(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_ADV);
+        detector->detect(test_image, lines);
+
+        if(2u <= lines.size())  ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_STD, whiteNoise)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateWhiteNoise(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+        detector->detect(test_image, lines);
+
+        if(50u >= lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_STD, constColor)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateConstColor(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+        detector->detect(test_image, lines);
+
+        if(0u == lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_STD, lines)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        const unsigned int numOfLines = 1;
+        GenerateLines(test_image, numOfLines);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+        detector->detect(test_image, lines);
+
+        if(numOfLines * 2 == lines.size()) ++passedtests;  // * 2 because of Gibbs effect
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_STD, rotatedRect)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateRotatedRect(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+        detector->detect(test_image, lines);
+
+        if(4u <= lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_NONE, whiteNoise)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateWhiteNoise(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+        detector->detect(test_image, lines);
+
+        if(50u >= lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_NONE, constColor)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateConstColor(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_NONE);
+        detector->detect(test_image, lines);
+
+        if(0u == lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_NONE, lines)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        const unsigned int numOfLines = 1;
+        GenerateLines(test_image, numOfLines);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_NONE);
+        detector->detect(test_image, lines);
+
+        if(numOfLines * 2 == lines.size()) ++passedtests;  // * 2 because of Gibbs effect
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
+
+TEST_F(Imgproc_LSD_NONE, rotatedRect)
+{
+    for (int i = 0; i < EPOCHS; ++i)
+    {
+        GenerateRotatedRect(test_image);
+        Ptr<LineSegmentDetector> detector = createLineSegmentDetectorPtr(LSD_REFINE_NONE);
+        detector->detect(test_image, lines);
+
+        if(8u <= lines.size()) ++passedtests;
+    }
+    ASSERT_EQ(EPOCHS, passedtests);
+}
diff --git a/modules/imgproc/test/test_precomp.cpp b/modules/imgproc/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/imgproc/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/imgproc/test/test_precomp.hpp b/modules/imgproc/test/test_precomp.hpp
index 9e937b9416..9650b7fcc8 100644
--- a/modules/imgproc/test/test_precomp.hpp
+++ b/modules/imgproc/test/test_precomp.hpp
@@ -9,11 +9,11 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/highgui.hpp"
-#include "opencv2/highgui/highgui_c.h"
-#include <iostream>
+
+#include "opencv2/imgproc/imgproc_c.h"
 
 #endif
diff --git a/modules/imgproc/test/test_templmatch.cpp b/modules/imgproc/test/test_templmatch.cpp
index 1b38e5614d..92410cd901 100644
--- a/modules/imgproc/test/test_templmatch.cpp
+++ b/modules/imgproc/test/test_templmatch.cpp
@@ -192,8 +192,8 @@ static void cvTsMatchTemplate( const CvMat* img, const CvMat* templ, CvMat* resu
     {
         for( j = 0; j < result->cols; j++ )
         {
-            CvScalar a_sum = {{ 0, 0, 0, 0 }}, a_sum2 = {{ 0, 0, 0, 0 }};
-            CvScalar ccorr = {{ 0, 0, 0, 0 }};
+            CvScalar a_sum(0), a_sum2(0);
+            CvScalar ccorr(0);
             double value = 0.;
 
             if( depth == CV_8U )
diff --git a/modules/imgproc/test/test_thresh.cpp b/modules/imgproc/test/test_thresh.cpp
index 30a760da5c..f59fec19e1 100644
--- a/modules/imgproc/test/test_thresh.cpp
+++ b/modules/imgproc/test/test_thresh.cpp
@@ -318,4 +318,3 @@ void CV_ThreshTest::prepare_to_validation( int /*test_case_idx*/ )
 }
 
 TEST(Imgproc_Threshold, accuracy) { CV_ThreshTest test; test.safe_run(); }
-
diff --git a/modules/imgproc/test/test_watershed.cpp b/modules/imgproc/test/test_watershed.cpp
index 52cb37a4fc..0edb18e02e 100644
--- a/modules/imgproc/test/test_watershed.cpp
+++ b/modules/imgproc/test/test_watershed.cpp
@@ -130,4 +130,3 @@ void CV_WatershedTest::run( int /* start_from */)
 }
 
 TEST(Imgproc_Watershed, regression) { CV_WatershedTest test; test.safe_run(); }
-
diff --git a/modules/java/CMakeLists.txt b/modules/java/CMakeLists.txt
index cf99226ee1..ad10aafc52 100644
--- a/modules/java/CMakeLists.txt
+++ b/modules/java/CMakeLists.txt
@@ -269,9 +269,15 @@ else(ANDROID)
 endif(ANDROID)
 
 # step 5: build native part
+
+# workarounding lack of `__attribute__ ((visibility("default")))` in jni_md.h/JNIEXPORT
+string(REPLACE "-fvisibility=hidden" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+
 add_library(${the_module} SHARED ${handwrittren_h_sources} ${handwrittren_cpp_sources} ${generated_cpp_sources}
                                  ${copied_files}
                                 "${JAR_FILE}" "${JAR_FILE}.dephelper")
+set_target_properties(${the_module} PROPERTIES COMPILE_DEFINITIONS OPENCV_NOSTL)
+
 if(BUILD_FAT_JAVA_LIB)
   set(__deps ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_MODULES_BUILD})
   list(REMOVE_ITEM __deps ${the_module} opencv_ts)
@@ -281,7 +287,14 @@ if(BUILD_FAT_JAVA_LIB)
   if(__extradeps)
     list(REMOVE_ITEM __deps ${__extradeps})
   endif()
-  target_link_libraries(${the_module} -Wl,-whole-archive ${__deps} -Wl,-no-whole-archive ${__extradeps} ${OPENCV_LINKER_LIBS})
+  if(APPLE)
+    foreach(_dep ${__deps})
+      target_link_libraries(${the_module} -Wl,-force_load "${_dep}")
+    endforeach()
+  else()
+    target_link_libraries(${the_module} -Wl,-whole-archive ${__deps} -Wl,-no-whole-archive)
+  endif()
+  target_link_libraries(${the_module} ${__extradeps} ${OPENCV_LINKER_LIBS})
 else()
   target_link_libraries(${the_module} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_LINKER_LIBS})
 endif()
diff --git a/modules/java/android_lib/lint.xml b/modules/java/android_lib/lint.xml
index e54ced1dc4..a95e509d22 100644
--- a/modules/java/android_lib/lint.xml
+++ b/modules/java/android_lib/lint.xml
@@ -1,5 +1,8 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <lint>
+    <issue id="InlinedApi">
+        <ignore path="src\org\opencv\android\JavaCameraView.java" />
+    </issue>
     <issue id="NewApi">
         <ignore path="src\org\opencv\android\JavaCameraView.java" />
     </issue>
diff --git a/modules/java/android_lib/res/values/attrs.xml b/modules/java/android_lib/res/values/attrs.xml
index 0cdf1097a0..6902621f66 100644
--- a/modules/java/android_lib/res/values/attrs.xml
+++ b/modules/java/android_lib/res/values/attrs.xml
@@ -4,8 +4,8 @@
        <attr name="show_fps" format="boolean"/>
        <attr name="camera_id" format="integer" >
           <enum name="any" value="-1" />
-          <enum name="back" value="0" />
-          <enum name="front" value="1" />
+          <enum name="back" value="99" />
+          <enum name="front" value="98" />
        </attr>
     </declare-styleable>
 </resources>
diff --git a/modules/java/android_test/.gitignore b/modules/java/android_test/.gitignore
index 2d406cbbcc..a21956836e 100644
--- a/modules/java/android_test/.gitignore
+++ b/modules/java/android_test/.gitignore
@@ -4,4 +4,4 @@ build.xml
 local.properties
 proguard-project.txt
 project.properties
-default.properties
\ No newline at end of file
+default.properties
diff --git a/modules/java/android_test/AndroidManifest.xml b/modules/java/android_test/AndroidManifest.xml
index e261e2d706..81f2bc134c 100644
--- a/modules/java/android_test/AndroidManifest.xml
+++ b/modules/java/android_test/AndroidManifest.xml
@@ -3,8 +3,8 @@
       package="org.opencv.test"
       android:versionCode="1"
       android:versionName="1.0">
-    
-	<uses-sdk android:minSdkVersion="8" />
+
+    <uses-sdk android:minSdkVersion="8" />
 
     <!-- We add an application tag here just so that we can indicate that
          this package needs to link against the android.test library,
@@ -20,7 +20,7 @@
     <instrumentation android:name="org.opencv.test.OpenCVTestRunner"
                      android:targetPackage="org.opencv.test"
                      android:label="Tests for org.opencv"/>
-    
+
     <uses-permission android:name="android.permission.CAMERA"/>
     <uses-feature android:name="android.hardware.camera" />
     <uses-feature android:name="android.hardware.camera.autofocus" />
diff --git a/modules/java/android_test/CMakeLists.txt b/modules/java/android_test/CMakeLists.txt
index 06ebf4aa10..41f69e6ca7 100644
--- a/modules/java/android_test/CMakeLists.txt
+++ b/modules/java/android_test/CMakeLists.txt
@@ -14,7 +14,7 @@ ocv_list_filterout(opencv_test_java_files ".svn")
 
 # copy sources out from the build tree
 set(opencv_test_java_file_deps "")
-foreach(f ${opencv_test_java_files} ${ANDROID_MANIFEST_FILE})
+foreach(f ${opencv_test_java_files} ${ANDROID_MANIFEST_FILE} ".classpath" ".project")
   add_custom_command(
       OUTPUT "${opencv_test_java_bin_dir}/${f}"
       COMMAND ${CMAKE_COMMAND} -E copy "${CMAKE_CURRENT_SOURCE_DIR}/${f}" "${opencv_test_java_bin_dir}/${f}"
diff --git a/modules/java/android_test/res/layout/main.xml b/modules/java/android_test/res/layout/main.xml
index 3a5f117d3c..1fb4479c21 100644
--- a/modules/java/android_test/res/layout/main.xml
+++ b/modules/java/android_test/res/layout/main.xml
@@ -4,9 +4,9 @@
     android:layout_width="fill_parent"
     android:layout_height="fill_parent"
     >
-<TextView  
-    android:layout_width="fill_parent" 
-    android:layout_height="wrap_content" 
+<TextView
+    android:layout_width="fill_parent"
+    android:layout_height="wrap_content"
     android:text="@string/hello"
     />
 </LinearLayout>
diff --git a/modules/java/android_test/src/org/opencv/test/OpenCVTestCase.java b/modules/java/android_test/src/org/opencv/test/OpenCVTestCase.java
index 5c932110d3..97cdca783a 100644
--- a/modules/java/android_test/src/org/opencv/test/OpenCVTestCase.java
+++ b/modules/java/android_test/src/org/opencv/test/OpenCVTestCase.java
@@ -19,8 +19,8 @@ import org.opencv.core.Point3;
 import org.opencv.core.Rect;
 import org.opencv.core.Scalar;
 import org.opencv.core.Size;
-import org.opencv.features2d.DMatch;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.DMatch;
+import org.opencv.core.KeyPoint;
 import org.opencv.highgui.Highgui;
 
 import android.util.Log;
diff --git a/modules/java/android_test/src/org/opencv/test/OpenCVTestRunner.java b/modules/java/android_test/src/org/opencv/test/OpenCVTestRunner.java
index 2dc4aead8d..9425e2644c 100644
--- a/modules/java/android_test/src/org/opencv/test/OpenCVTestRunner.java
+++ b/modules/java/android_test/src/org/opencv/test/OpenCVTestRunner.java
@@ -17,8 +17,6 @@ import android.util.Log;
 
 /**
  * This only class is Android specific.
- *
- * @see <a href="http://opencv.itseez.com">OpenCV</a>
  */
 
 public class OpenCVTestRunner extends InstrumentationTestRunner {
diff --git a/modules/java/android_test/src/org/opencv/test/calib3d/Calib3dTest.java b/modules/java/android_test/src/org/opencv/test/calib3d/Calib3dTest.java
index 8bcaf58a05..69b8d0fcbd 100644
--- a/modules/java/android_test/src/org/opencv/test/calib3d/Calib3dTest.java
+++ b/modules/java/android_test/src/org/opencv/test/calib3d/Calib3dTest.java
@@ -188,13 +188,13 @@ public class Calib3dTest extends OpenCVTestCase {
         assertTrue(!corners.empty());
     }
 
-    public void testFindCirclesGridDefaultMatSizeMat() {
+    public void testFindCirclesGridMatSizeMat() {
         int size = 300;
         Mat img = new Mat(size, size, CvType.CV_8U);
         img.setTo(new Scalar(255));
         Mat centers = new Mat();
 
-        assertFalse(Calib3d.findCirclesGridDefault(img, new Size(5, 5), centers));
+        assertFalse(Calib3d.findCirclesGrid(img, new Size(5, 5), centers));
 
         for (int i = 0; i < 5; i++)
             for (int j = 0; j < 5; j++) {
@@ -202,20 +202,20 @@ public class Calib3dTest extends OpenCVTestCase {
                 Core.circle(img, pt, 10, new Scalar(0), -1);
             }
 
-        assertTrue(Calib3d.findCirclesGridDefault(img, new Size(5, 5), centers));
+        assertTrue(Calib3d.findCirclesGrid(img, new Size(5, 5), centers));
 
         assertEquals(25, centers.rows());
         assertEquals(1, centers.cols());
         assertEquals(CvType.CV_32FC2, centers.type());
     }
 
-    public void testFindCirclesGridDefaultMatSizeMatInt() {
+    public void testFindCirclesGridMatSizeMatInt() {
         int size = 300;
         Mat img = new Mat(size, size, CvType.CV_8U);
         img.setTo(new Scalar(255));
         Mat centers = new Mat();
 
-        assertFalse(Calib3d.findCirclesGridDefault(img, new Size(3, 5), centers, Calib3d.CALIB_CB_CLUSTERING
+        assertFalse(Calib3d.findCirclesGrid(img, new Size(3, 5), centers, Calib3d.CALIB_CB_CLUSTERING
                 | Calib3d.CALIB_CB_ASYMMETRIC_GRID));
 
         int step = size * 2 / 15;
@@ -227,7 +227,7 @@ public class Calib3dTest extends OpenCVTestCase {
                 Core.circle(img, pt, 10, new Scalar(0), -1);
             }
 
-        assertTrue(Calib3d.findCirclesGridDefault(img, new Size(3, 5), centers, Calib3d.CALIB_CB_CLUSTERING
+        assertTrue(Calib3d.findCirclesGrid(img, new Size(3, 5), centers, Calib3d.CALIB_CB_CLUSTERING
                 | Calib3d.CALIB_CB_ASYMMETRIC_GRID));
 
         assertEquals(15, centers.rows());
@@ -585,4 +585,18 @@ public class Calib3dTest extends OpenCVTestCase {
     public void testValidateDisparityMatMatIntIntInt() {
         fail("Not yet implemented");
     }
+
+    public void testComputeCorrespondEpilines()
+    {
+        Mat fundamental = new Mat(3, 3, CvType.CV_64F);
+        fundamental.put(0, 0, 0, -0.577, 0.288, 0.577, 0, 0.288, -0.288, -0.288, 0);
+        MatOfPoint2f left = new MatOfPoint2f();
+        left.alloc(1);
+        left.put(0, 0, 2, 3); //add(new Point(x, y));
+        Mat lines = new Mat();
+        Mat truth = new Mat(1, 1, CvType.CV_32FC3);
+        truth.put(0, 0, -0.70735186, 0.70686162, -0.70588124);
+        Calib3d.computeCorrespondEpilines(left, 1, fundamental, lines);
+        assertMatEqual(truth, lines, EPS);
+    }
 }
diff --git a/modules/java/android_test/src/org/opencv/test/core/CoreTest.java b/modules/java/android_test/src/org/opencv/test/core/CoreTest.java
index a6023c7cc6..c13df193ef 100644
--- a/modules/java/android_test/src/org/opencv/test/core/CoreTest.java
+++ b/modules/java/android_test/src/org/opencv/test/core/CoreTest.java
@@ -460,7 +460,7 @@ public class CoreTest extends OpenCVTestCase {
         Mat eigenVals = new Mat();
         Mat eigenVecs = new Mat();
 
-        Core.eigen(src, true, eigenVals, eigenVecs);
+        Core.eigen(src, eigenVals, eigenVecs);
 
         Mat expectedEigenVals = new Mat(3, 1, CvType.CV_32FC1) {
             {
@@ -491,20 +491,6 @@ public class CoreTest extends OpenCVTestCase {
 
         Point truth[] = {
                 new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(5, 6),
-                new Point(4, 6),
-                new Point(4, 6),
-                new Point(4, 6),
-                new Point(4, 6),
-                new Point(4, 6),
-                new Point(4, 6),
-                new Point(4, 6),
                 new Point(4, 6)
         };
         assertArrayPointsEquals(truth, pts.toArray(), EPS);
@@ -1113,18 +1099,6 @@ public class CoreTest extends OpenCVTestCase {
         assertMatEqual(gray255, dst);
     }
 
-    public void testLUTMatMatMatInt() {
-        Mat lut = new Mat(1, 256, CvType.CV_8UC1);
-        // TODO: ban this overload
-        try
-        {
-            Core.LUT(grayRnd, lut, dst, 1);
-            fail("Last parameter for LUT was not supported");
-        } catch (CvException e) {
-            // expected
-        }
-    }
-
     public void testMagnitude() {
         Mat x = new Mat(1, 4, CvType.CV_32F);
         Mat y = new Mat(1, 4, CvType.CV_32F);
@@ -1839,9 +1813,9 @@ public class CoreTest extends OpenCVTestCase {
     }
 
     public void testRandShuffleMat() {
-        Mat original = new Mat(1, 5, CvType.CV_32F) {
+        Mat original = new Mat(1, 10, CvType.CV_32F) {
             {
-                put(0, 0, 7, 5, 2, 8, 1);
+                put(0, 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9);
             }
         };
         Mat shuffled = original.clone();
@@ -1857,9 +1831,9 @@ public class CoreTest extends OpenCVTestCase {
     }
 
     public void testRandShuffleMatDouble() {
-        Mat original = new Mat(1, 5, CvType.CV_32F) {
+        Mat original = new Mat(1, 10, CvType.CV_32F) {
             {
-                put(0, 0, 7, 5, 2, 8, 1);
+                put(0, 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9);
             }
         };
         Mat shuffled = original.clone();
@@ -2268,4 +2242,35 @@ public class CoreTest extends OpenCVTestCase {
 
     }
 
+    public void testCopyMakeBorderMatMatIntIntIntIntInt() {
+        Mat src = new Mat(2, 2, CvType.CV_32F, new Scalar(1));
+        int border = 2;
+
+        Core.copyMakeBorder(src, dst, border, border, border, border, Core.BORDER_REPLICATE);
+
+        truth = new Mat(6, 6, CvType.CV_32F, new Scalar(1));
+        assertMatEqual(truth, dst, EPS);
+    }
+
+    public void testCopyMakeBorderMatMatIntIntIntIntIntScalar() {
+        Mat src = new Mat(2, 2, CvType.CV_32F, new Scalar(1));
+
+        Scalar value = new Scalar(0);
+        int border = 2;
+
+        Core.copyMakeBorder(src, dst, border, border, border, border, Core.BORDER_REPLICATE, value);
+        // TODO_: write better test (use Core.BORDER_CONSTANT)
+
+        truth = new Mat(6, 6, CvType.CV_32F, new Scalar(1));
+        assertMatEqual(truth, dst, EPS);
+    }
+
+    public void testBorderInterpolate() {
+        float val1 = Core.borderInterpolate(100, 150, Core.BORDER_REFLECT_101);
+        assertEquals(100f, val1);
+
+        float val2 = Core.borderInterpolate(-5, 10, Core.BORDER_WRAP);
+        assertEquals(5f, val2);
+    }
+
 }
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/DMatchTest.java b/modules/java/android_test/src/org/opencv/test/core/DMatchTest.java
similarity index 93%
rename from modules/java/android_test/src/org/opencv/test/features2d/DMatchTest.java
rename to modules/java/android_test/src/org/opencv/test/core/DMatchTest.java
index ef3443de3f..b523e0cd64 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/DMatchTest.java
+++ b/modules/java/android_test/src/org/opencv/test/core/DMatchTest.java
@@ -1,6 +1,6 @@
-package org.opencv.test.features2d;
+package org.opencv.test.core;
 
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 
 import junit.framework.TestCase;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/KeyPointTest.java b/modules/java/android_test/src/org/opencv/test/core/KeyPointTest.java
similarity index 96%
rename from modules/java/android_test/src/org/opencv/test/features2d/KeyPointTest.java
rename to modules/java/android_test/src/org/opencv/test/core/KeyPointTest.java
index 2a0b9253ef..71f3a320b6 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/KeyPointTest.java
+++ b/modules/java/android_test/src/org/opencv/test/core/KeyPointTest.java
@@ -1,7 +1,7 @@
-package org.opencv.test.features2d;
+package org.opencv.test.core;
 
 import org.opencv.core.Point;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 
 public class KeyPointTest extends OpenCVTestCase {
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BRIEFDescriptorExtractorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BRIEFDescriptorExtractorTest.java
index 4fdaa2db62..060175234f 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BRIEFDescriptorExtractorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BRIEFDescriptorExtractorTest.java
@@ -7,7 +7,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.DescriptorExtractor;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceDescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceDescriptorMatcherTest.java
index 2d43307352..ae8645e488 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceDescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceDescriptorMatcherTest.java
@@ -11,11 +11,11 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingDescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingDescriptorMatcherTest.java
index 08ff220ede..5ed99df824 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingDescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingDescriptorMatcherTest.java
@@ -1,5 +1,6 @@
 package org.opencv.test.features2d;
 
+import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.List;
 
@@ -10,7 +11,7 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
@@ -204,7 +205,17 @@ public class BruteForceHammingDescriptorMatcherTest extends OpenCVTestCase {
     }
 
     public void testRadiusMatchMatListOfListOfDMatchFloat() {
-        fail("Not yet implemented");
+        Mat train = getTrainDescriptors();
+        Mat query = getQueryDescriptors();
+        ArrayList<MatOfDMatch> matches = new ArrayList<MatOfDMatch>();
+
+        matcher.radiusMatch(query, train, matches, 50.f);
+
+        assertEquals(matches.size(), 4);
+        assertTrue(matches.get(0).empty());
+        assertMatEqual(matches.get(1), new MatOfDMatch(truth[1]), EPS);
+        assertMatEqual(matches.get(2), new MatOfDMatch(truth[2]), EPS);
+        assertTrue(matches.get(3).empty());
     }
 
     public void testRadiusMatchMatListOfListOfDMatchFloatListOfMat() {
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingLUTDescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingLUTDescriptorMatcherTest.java
index 2d99c75b79..961d93605e 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingLUTDescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceHammingLUTDescriptorMatcherTest.java
@@ -10,7 +10,7 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceL1DescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceL1DescriptorMatcherTest.java
index 5d0ab51e4b..485f9686e3 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceL1DescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceL1DescriptorMatcherTest.java
@@ -10,11 +10,11 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceSL2DescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceSL2DescriptorMatcherTest.java
index 6a097b4fa5..c19d1d5f5d 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/BruteForceSL2DescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/BruteForceSL2DescriptorMatcherTest.java
@@ -10,11 +10,11 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/FASTFeatureDetectorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/FASTFeatureDetectorTest.java
index 0b7d289d53..df123d2a18 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/FASTFeatureDetectorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/FASTFeatureDetectorTest.java
@@ -9,7 +9,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/Features2dTest.java b/modules/java/android_test/src/org/opencv/test/features2d/Features2dTest.java
index 5b815fd145..2118b53065 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/Features2dTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/Features2dTest.java
@@ -12,12 +12,12 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.MatOfPoint2f;
 import org.opencv.core.Point;
 import org.opencv.core.Range;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
 import org.opencv.features2d.Features2d;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.highgui.Highgui;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/FlannBasedDescriptorMatcherTest.java b/modules/java/android_test/src/org/opencv/test/features2d/FlannBasedDescriptorMatcherTest.java
index 70c994b288..8239473697 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/FlannBasedDescriptorMatcherTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/FlannBasedDescriptorMatcherTest.java
@@ -11,11 +11,11 @@ import org.opencv.core.MatOfDMatch;
 import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 import org.opencv.features2d.DescriptorExtractor;
 import org.opencv.features2d.DescriptorMatcher;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/ORBDescriptorExtractorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/ORBDescriptorExtractorTest.java
index 4ccf09808a..2a14dd99aa 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/ORBDescriptorExtractorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/ORBDescriptorExtractorTest.java
@@ -7,7 +7,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.DescriptorExtractor;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/SIFTDescriptorExtractorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/SIFTDescriptorExtractorTest.java
index 8add9f4933..c793a6c3ba 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/SIFTDescriptorExtractorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/SIFTDescriptorExtractorTest.java
@@ -7,7 +7,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.DescriptorExtractor;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/STARFeatureDetectorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/STARFeatureDetectorTest.java
index 0b7e4e1031..35e332c5cf 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/STARFeatureDetectorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/STARFeatureDetectorTest.java
@@ -9,7 +9,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/SURFDescriptorExtractorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/SURFDescriptorExtractorTest.java
index af2e26efbe..6d63e36d59 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/SURFDescriptorExtractorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/SURFDescriptorExtractorTest.java
@@ -7,7 +7,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.DescriptorExtractor;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/features2d/SURFFeatureDetectorTest.java b/modules/java/android_test/src/org/opencv/test/features2d/SURFFeatureDetectorTest.java
index 778d6f3735..b3c59b016f 100644
--- a/modules/java/android_test/src/org/opencv/test/features2d/SURFFeatureDetectorTest.java
+++ b/modules/java/android_test/src/org/opencv/test/features2d/SURFFeatureDetectorTest.java
@@ -13,7 +13,7 @@ import org.opencv.core.MatOfKeyPoint;
 import org.opencv.core.Point;
 import org.opencv.core.Scalar;
 import org.opencv.features2d.FeatureDetector;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.test.OpenCVTestRunner;
 
diff --git a/modules/java/android_test/src/org/opencv/test/imgproc/ImgprocTest.java b/modules/java/android_test/src/org/opencv/test/imgproc/ImgprocTest.java
index 9d4d1346aa..7971ea2f1f 100644
--- a/modules/java/android_test/src/org/opencv/test/imgproc/ImgprocTest.java
+++ b/modules/java/android_test/src/org/opencv/test/imgproc/ImgprocTest.java
@@ -176,7 +176,7 @@ public class ImgprocTest extends OpenCVTestCase {
     }
 
     public void testBilateralFilterMatMatIntDoubleDoubleInt() {
-        Imgproc.bilateralFilter(gray255, dst, 5, 10, 5, Imgproc.BORDER_REFLECT);
+        Imgproc.bilateralFilter(gray255, dst, 5, 10, 5, Core.BORDER_REFLECT);
 
         assertMatEqual(gray255, dst);
         // TODO_: write better test
@@ -198,19 +198,11 @@ public class ImgprocTest extends OpenCVTestCase {
     }
 
     public void testBlurMatMatSizePointInt() {
-        Imgproc.blur(gray0, dst, size, anchorPoint, Imgproc.BORDER_REFLECT);
+        Imgproc.blur(gray0, dst, size, anchorPoint, Core.BORDER_REFLECT);
         assertMatEqual(gray0, dst);
         // TODO_: write better test
     }
 
-    public void testBorderInterpolate() {
-        float val1 = Imgproc.borderInterpolate(100, 150, Imgproc.BORDER_REFLECT_101);
-        assertEquals(100f, val1);
-
-        float val2 = Imgproc.borderInterpolate(-5, 10, Imgproc.BORDER_WRAP);
-        assertEquals(5f, val2);
-    }
-
     public void testBoundingRect() {
         MatOfPoint points = new MatOfPoint(new Point(0, 0), new Point(0, 4), new Point(4, 0), new Point(4, 4));
         Point p1 = new Point(1, 1);
@@ -236,7 +228,7 @@ public class ImgprocTest extends OpenCVTestCase {
     }
 
     public void testBoxFilterMatMatIntSizePointBooleanInt() {
-        Imgproc.boxFilter(gray255, dst, 8, size, anchorPoint, false, Imgproc.BORDER_REFLECT);
+        Imgproc.boxFilter(gray255, dst, 8, size, anchorPoint, false, Core.BORDER_REFLECT);
         assertMatEqual(gray255, dst);
         // TODO_: write better test
     }
@@ -479,29 +471,6 @@ public class ImgprocTest extends OpenCVTestCase {
         assertMatEqual(new MatOfInt4(3, 0, 5, 3620), convexityDefects);
     }
 
-    public void testCopyMakeBorderMatMatIntIntIntIntInt() {
-        Mat src = new Mat(imgprocSz, imgprocSz, CvType.CV_32F, new Scalar(1));
-        int border = 2;
-
-        Imgproc.copyMakeBorder(src, dst, border, border, border, border, Imgproc.BORDER_REPLICATE);
-
-        truth = new Mat(6, 6, CvType.CV_32F, new Scalar(1));
-        assertMatEqual(truth, dst, EPS);
-    }
-
-    public void testCopyMakeBorderMatMatIntIntIntIntIntScalar() {
-        Mat src = new Mat(imgprocSz, imgprocSz, CvType.CV_32F, new Scalar(1));
-
-        Scalar value = new Scalar(0);
-        int border = 2;
-
-        Imgproc.copyMakeBorder(src, dst, border, border, border, border, Imgproc.BORDER_REPLICATE, value);
-        // TODO_: write better test (use Imgproc.BORDER_CONSTANT)
-
-        truth = new Mat(6, 6, CvType.CV_32F, new Scalar(1));
-        assertMatEqual(truth, dst, EPS);
-    }
-
     public void testCornerEigenValsAndVecsMatMatIntInt() {
         fail("Not yet implemented");
         // TODO: write better test
@@ -528,7 +497,7 @@ public class ImgprocTest extends OpenCVTestCase {
 
         truth = new Mat(4, 4, CvType.CV_32FC(6), new Scalar(0));
 
-        Imgproc.cornerEigenValsAndVecs(src, dst, blockSize, ksize, Imgproc.BORDER_REFLECT);
+        Imgproc.cornerEigenValsAndVecs(src, dst, blockSize, ksize, Core.BORDER_REFLECT);
         assertMatEqual(truth, dst, EPS);
     }
 
@@ -552,7 +521,7 @@ public class ImgprocTest extends OpenCVTestCase {
         int blockSize = 5;
         int ksize = 7;
         double k = 0.1;
-        Imgproc.cornerHarris(gray255, dst, blockSize, ksize, k, Imgproc.BORDER_REFLECT);
+        Imgproc.cornerHarris(gray255, dst, blockSize, ksize, k, Core.BORDER_REFLECT);
         assertMatEqual(truth, dst, EPS);
     }
 
@@ -598,7 +567,7 @@ public class ImgprocTest extends OpenCVTestCase {
         int blockSize = 3;
         int ksize = 5;
 
-        Imgproc.cornerMinEigenVal(src, dst, blockSize, ksize, Imgproc.BORDER_REFLECT);
+        Imgproc.cornerMinEigenVal(src, dst, blockSize, ksize, Core.BORDER_REFLECT);
 
         truth = new Mat(3, 3, CvType.CV_32FC1) {
             {
@@ -742,7 +711,7 @@ public class ImgprocTest extends OpenCVTestCase {
         Mat kernel = new Mat();
         Scalar sc = new Scalar(3, 3);
 
-        Imgproc.erode(src, dst, kernel, anchorPoint, 10, Imgproc.BORDER_REFLECT, sc);
+        Imgproc.erode(src, dst, kernel, anchorPoint, 10, Core.BORDER_REFLECT, sc);
 
         truth = new Mat(3, 3, CvType.CV_8U, new Scalar(8));
         assertMatEqual(truth, dst);
@@ -773,7 +742,7 @@ public class ImgprocTest extends OpenCVTestCase {
         Mat kernel = new Mat(imgprocSz, imgprocSz, CvType.CV_32F, new Scalar(0));
         Point point = new Point(0, 0);
 
-        Imgproc.filter2D(gray128, dst, -1, kernel, point, 2, Imgproc.BORDER_CONSTANT);
+        Imgproc.filter2D(gray128, dst, -1, kernel, point, 2, Core.BORDER_CONSTANT);
 
         assertMatEqual(gray2, dst);
     }
@@ -901,7 +870,7 @@ public class ImgprocTest extends OpenCVTestCase {
     }
 
     public void testGaussianBlurMatMatSizeDoubleDoubleInt() {
-        Imgproc.GaussianBlur(gray2, dst, size, 1, 3, Imgproc.BORDER_REFLECT);
+        Imgproc.GaussianBlur(gray2, dst, size, 1, 3, Core.BORDER_REFLECT);
 
         assertMatEqual(gray2, dst);
         // TODO_: write better test
@@ -1384,7 +1353,7 @@ public class ImgprocTest extends OpenCVTestCase {
     public void testLaplacianMatMatIntIntDoubleDoubleInt() {
         Mat src = new Mat(3, 3, CvType.CV_32F, new Scalar(2));
 
-        Imgproc.Laplacian(src, dst, CvType.CV_32F, 1, 2, EPS, Imgproc.BORDER_REFLECT);
+        Imgproc.Laplacian(src, dst, CvType.CV_32F, 1, 2, EPS, Core.BORDER_REFLECT);
 
         truth = new Mat(3, 3, CvType.CV_32F, new Scalar(0.00099945068));
         assertMatEqual(truth, dst, EPS);
@@ -1486,7 +1455,7 @@ public class ImgprocTest extends OpenCVTestCase {
         Point point = new Point(1, 1);
         Scalar sc = new Scalar(3, 3);
 
-        Imgproc.morphologyEx(src, dst, Imgproc.MORPH_TOPHAT, kernel, point, 10, Imgproc.BORDER_REFLECT, sc);
+        Imgproc.morphologyEx(src, dst, Imgproc.MORPH_TOPHAT, kernel, point, 10, Core.BORDER_REFLECT, sc);
         truth = new Mat(imgprocSz, imgprocSz, CvType.CV_8U) {
             {
                 put(0, 0, 1, 0);
@@ -1520,7 +1489,7 @@ public class ImgprocTest extends OpenCVTestCase {
         Mat src = new Mat(4, 4, CvType.CV_32F, new Scalar(1));
         int ksize = 3;
 
-        Imgproc.preCornerDetect(src, dst, ksize, Imgproc.BORDER_REFLECT);
+        Imgproc.preCornerDetect(src, dst, ksize, Core.BORDER_REFLECT);
 
         truth = new Mat(4, 4, CvType.CV_32F, new Scalar(0));
         assertMatEqual(truth, dst, EPS);
@@ -1640,7 +1609,7 @@ public class ImgprocTest extends OpenCVTestCase {
 
         truth = new Mat(1, 3, CvType.CV_32F, new Scalar(2));
 
-        Imgproc.remap(src, dst, map1, map2, Imgproc.INTER_LINEAR, Imgproc.BORDER_REFLECT, sc);
+        Imgproc.remap(src, dst, map1, map2, Imgproc.INTER_LINEAR, Core.BORDER_REFLECT, sc);
         assertMatEqual(truth, dst, EPS);
     }
 
@@ -1683,7 +1652,7 @@ public class ImgprocTest extends OpenCVTestCase {
     public void testScharrMatMatIntIntIntDoubleDoubleInt() {
         Mat src = Mat.eye(3, 3, CvType.CV_32F);
 
-        Imgproc.Scharr(src, dst, CvType.CV_32F, 1, 0, 1.5, 0, Imgproc.BORDER_REFLECT);
+        Imgproc.Scharr(src, dst, CvType.CV_32F, 1, 0, 1.5, 0, Core.BORDER_REFLECT);
 
         truth = new Mat(3, 3, CvType.CV_32F) {
             {
@@ -1728,7 +1697,7 @@ public class ImgprocTest extends OpenCVTestCase {
         Mat kernelY = new Mat(1, 3, CvType.CV_32FC1);
         kernelY.put(0, 0, 1, 1, 1);
 
-        Imgproc.sepFilter2D(gray0, dst, CvType.CV_32F, kernelX, kernelY, anchorPoint, weakEPS, Imgproc.BORDER_REFLECT);
+        Imgproc.sepFilter2D(gray0, dst, CvType.CV_32F, kernelX, kernelY, anchorPoint, weakEPS, Core.BORDER_REFLECT);
 
         truth = new Mat(10, 10, CvType.CV_32F, new Scalar(weakEPS));
         assertMatEqual(truth, dst, EPS);
@@ -1756,7 +1725,7 @@ public class ImgprocTest extends OpenCVTestCase {
             }
         };
 
-        Imgproc.Sobel(src, dst, CvType.CV_32F, 1, 0, 3, 2, 0, Imgproc.BORDER_REPLICATE);
+        Imgproc.Sobel(src, dst, CvType.CV_32F, 1, 0, 3, 2, 0, Core.BORDER_REPLICATE);
 
         truth = new Mat(3, 3, CvType.CV_32F) {
             {
@@ -1963,4 +1932,4 @@ public class ImgprocTest extends OpenCVTestCase {
         assertMatEqual(truth, markers);
     }
 
-}
\ No newline at end of file
+}
diff --git a/modules/java/android_test/src/org/opencv/test/utils/ConvertersTest.java b/modules/java/android_test/src/org/opencv/test/utils/ConvertersTest.java
index 7a07115234..117bbd8083 100644
--- a/modules/java/android_test/src/org/opencv/test/utils/ConvertersTest.java
+++ b/modules/java/android_test/src/org/opencv/test/utils/ConvertersTest.java
@@ -5,8 +5,8 @@ import org.opencv.core.Mat;
 import org.opencv.core.Point;
 import org.opencv.core.Point3;
 import org.opencv.core.Rect;
-import org.opencv.features2d.DMatch;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.DMatch;
+import org.opencv.core.KeyPoint;
 import org.opencv.test.OpenCVTestCase;
 import org.opencv.utils.Converters;
 
diff --git a/modules/java/check-tests.py b/modules/java/check-tests.py
index 4cb80ff724..394b377ef9 100755
--- a/modules/java/check-tests.py
+++ b/modules/java/check-tests.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys, os, re
 
@@ -162,4 +162,3 @@ if __name__ == '__main__':
     print "Not tested methods found:", len(funcs)
     print "Total tests found:", parser.get_tests_count()
     print "Empty test stubs found:", parser.get_empty_stubs_count()
-
diff --git a/modules/java/generator/config/core.filelist b/modules/java/generator/config/core.filelist
index 96c377786d..27a46dc6e0 100644
--- a/modules/java/generator/config/core.filelist
+++ b/modules/java/generator/config/core.filelist
@@ -1,2 +1,4 @@
+include/opencv2/core/base.hpp
 include/opencv2/core.hpp
+include/opencv2/core/utility.hpp
 ../java/generator/src/cpp/core_manual.hpp
diff --git a/modules/java/generator/gen_java.py b/modules/java/generator/gen_java.py
index 030f6c897a..d0238458eb 100755
--- a/modules/java/generator/gen_java.py
+++ b/modules/java/generator/gen_java.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys, re, os.path
 from string import Template
@@ -10,9 +10,9 @@ except:
 
 class_ignore_list = (
     #core
-    "FileNode", "FileStorage", "KDTree",
+    "FileNode", "FileStorage", "KDTree", "KeyPoint", "DMatch",
     #highgui
-    "VideoWriter", "VideoCapture",
+    "VideoWriter",
 )
 
 const_ignore_list = (
@@ -285,13 +285,13 @@ type_dict = {
     "CvSlice" : { "j_type" : "Range",  "jn_args" : (("int", ".start"), ("int", ".end")),
                   "jni_var" : "Range %(n)s(%(n)s_start, %(n)s_end)", "jni_type" : "jdoubleArray",
                   "suffix" : "II"},
-    "string"  : { "j_type" : "String",  "jn_type" : "String",
+    "String"  : { "j_type" : "String",  "jn_type" : "String",
                   "jni_type" : "jstring", "jni_name" : "n_%(n)s",
-                  "jni_var" : 'const char* utf_%(n)s = env->GetStringUTFChars(%(n)s, 0); std::string n_%(n)s( utf_%(n)s ? utf_%(n)s : "" ); env->ReleaseStringUTFChars(%(n)s, utf_%(n)s)',
+                  "jni_var" : 'const char* utf_%(n)s = env->GetStringUTFChars(%(n)s, 0); String n_%(n)s( utf_%(n)s ? utf_%(n)s : "" ); env->ReleaseStringUTFChars(%(n)s, utf_%(n)s)',
                   "suffix" : "Ljava_lang_String_2"},
     "c_string": { "j_type" : "String",  "jn_type" : "String",
                   "jni_type" : "jstring", "jni_name" : "n_%(n)s.c_str()",
-                  "jni_var" : 'const char* utf_%(n)s = env->GetStringUTFChars(%(n)s, 0); std::string n_%(n)s( utf_%(n)s ? utf_%(n)s : "" ); env->ReleaseStringUTFChars(%(n)s, utf_%(n)s)',
+                  "jni_var" : 'const char* utf_%(n)s = env->GetStringUTFChars(%(n)s, 0); String n_%(n)s( utf_%(n)s ? utf_%(n)s : "" ); env->ReleaseStringUTFChars(%(n)s, utf_%(n)s)',
                   "suffix" : "Ljava_lang_String_2"},
 "TermCriteria": { "j_type" : "TermCriteria",  "jn_args" : (("int", ".type"), ("int", ".maxCount"), ("double", ".epsilon")),
                   "jni_var" : "TermCriteria %(n)s(%(n)s_type, %(n)s_maxCount, %(n)s_epsilon)", "jni_type" : "jdoubleArray",
@@ -417,7 +417,7 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1minMaxLocManual
         {
             'j_code'   :
 """
-    // C++: Size getTextSize(const std::string& text, int fontFace, double fontScale, int thickness, int* baseLine);
+    // C++: Size getTextSize(const String& text, int fontFace, double fontScale, int thickness, int* baseLine);
     //javadoc:getTextSize(text, fontFace, fontScale, thickness, baseLine)
     public static Size getTextSize(String text, int fontFace, double fontScale, int thickness, int[] baseLine) {
         if(baseLine != null && baseLine.length != 1)
@@ -430,7 +430,7 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1minMaxLocManual
 """    private static native double[] n_getTextSize(String text, int fontFace, double fontScale, int thickness, int[] baseLine);\n""",
             'cpp_code' :
 """
-// C++: Size getTextSize(const std::string& text, int fontFace, double fontScale, int thickness, int* baseLine);
+// C++: Size getTextSize(const String& text, int fontFace, double fontScale, int thickness, int* baseLine);
 JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1getTextSize (JNIEnv*, jclass, jstring, jint, jdouble, jint, jintArray);
 
 JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1getTextSize
@@ -445,7 +445,7 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1getTextSize
         }
 
         const char* utf_text = env->GetStringUTFChars(text, 0);
-        std::string n_text( utf_text ? utf_text : "" );
+        String n_text( utf_text ? utf_text : "" );
         env->ReleaseStringUTFChars(text, utf_text);
 
         int _baseLine;
@@ -510,50 +510,118 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Core_n_1getTextSize
         "resizeWindow"      : {'j_code' : '', 'jn_code' : '', 'cpp_code' : '' },
     }, # Highgui
 
+    'VideoCapture' :
+    {
+        "getSupportedPreviewSizes" :
+        {
+            'j_code' :
+"""
+    public java.util.List<org.opencv.core.Size> getSupportedPreviewSizes()
+    {
+        String[] sizes_str = getSupportedPreviewSizes_0(nativeObj).split(",");
+        java.util.List<org.opencv.core.Size> sizes = new java.util.ArrayList<org.opencv.core.Size>(sizes_str.length);
+
+        for (String str : sizes_str) {
+            String[] wh = str.split("x");
+            sizes.add(new org.opencv.core.Size(Double.parseDouble(wh[0]), Double.parseDouble(wh[1])));
+        }
+
+        return sizes;
+    }
+
+""",
+            'jn_code' :
+"""\n    private static native String getSupportedPreviewSizes_0(long nativeObj);\n""",
+            'cpp_code' :
+"""
+JNIEXPORT jstring JNICALL Java_org_opencv_highgui_VideoCapture_getSupportedPreviewSizes_10
+  (JNIEnv *env, jclass, jlong self);
+
+JNIEXPORT jstring JNICALL Java_org_opencv_highgui_VideoCapture_getSupportedPreviewSizes_10
+  (JNIEnv *env, jclass, jlong self)
+{
+    static const char method_name[] = "highgui::VideoCapture_getSupportedPreviewSizes_10()";
+    try {
+        LOGD("%s", method_name);
+        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
+        union {double prop; const char* name;} u;
+        u.prop = me->get(CAP_PROP_ANDROID_PREVIEW_SIZES_STRING);
+        return env->NewStringUTF(u.name);
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
+    } catch (...) {
+        throwJavaException(env, 0, method_name);
+    }
+    return env->NewStringUTF("");
 }
 
-# { class : { func : {arg_name : ctype} } }
+""",
+        }, # getSupportedPreviewSizes
+    }, # VideoCapture
+}
+
+# { class : { func : { arg_name : {"ctype" : ctype, "attrib" : [attrib]} } } }
 func_arg_fix = {
     '' : {
-        'randu'    : { 'low'     : 'double', 'high'   : 'double', },
-        'randn'    : { 'mean'    : 'double', 'stddev' : 'double', },
-        'inRange'  : { 'lowerb'  : 'Scalar', 'upperb' : 'Scalar', },
-        'goodFeaturesToTrack' : { 'corners' : 'vector_Point', },
-        'findFundamentalMat' : { 'points1' : 'vector_Point2f', 'points2' : 'vector_Point2f', },
-        'cornerSubPix' : { 'corners' : 'vector_Point2f', },
-        'minEnclosingCircle' : { 'points' : 'vector_Point2f', },
-        'findHomography' : { 'srcPoints' : 'vector_Point2f', 'dstPoints' : 'vector_Point2f', },
-        'solvePnP' : { 'objectPoints' : 'vector_Point3f', 'imagePoints' : 'vector_Point2f',
-                       'distCoeffs' : 'vector_double' },
-        'solvePnPRansac' : { 'objectPoints' : 'vector_Point3f', 'imagePoints' : 'vector_Point2f',
-                             'distCoeffs' : 'vector_double' },
-        'calcOpticalFlowPyrLK' : { 'prevPts' : 'vector_Point2f', 'nextPts' : 'vector_Point2f',
-                                   'status' : 'vector_uchar', 'err' : 'vector_float', },
-        'fitEllipse' : { 'points' : 'vector_Point2f', },
-        'fillPoly' : { 'pts' : 'vector_vector_Point', },
-        'polylines' : { 'pts' : 'vector_vector_Point', },
-        'fillConvexPoly' : { 'points' : 'vector_Point', },
-        'boundingRect' : { 'points' : 'vector_Point', },
-        'approxPolyDP' : { 'curve' : 'vector_Point2f', 'approxCurve' : 'vector_Point2f', },
-        'arcLength' : { 'curve' : 'vector_Point2f', },
-        'pointPolygonTest' : { 'contour' : 'vector_Point2f', },
-        'minAreaRect' : { 'points' : 'vector_Point2f', },
-        'getAffineTransform' : { 'src' : 'vector_Point2f', 'dst' : 'vector_Point2f', },
-        'hconcat' : { 'src' : 'vector_Mat', },
-        'vconcat' : { 'src' : 'vector_Mat', },
-        'undistortPoints' : { 'src' : 'vector_Point2f', 'dst' : 'vector_Point2f' },
-        'checkRange' : {'pos' : '*'},
-        'meanStdDev' : {'mean' : 'vector_double', 'stddev' : 'vector_double'},
-        'drawContours' : {'contours' : 'vector_vector_Point'},
-        'findContours' : {'contours' : 'vector_vector_Point'},
-        'convexityDefects' : {'contour' : 'vector_Point', 'convexhull' : 'vector_int', 'convexityDefects' : 'vector_Vec4i'},
-        'isContourConvex' : { 'contour' : 'vector_Point', },
-        'convexHull' : {'points' : 'vector_Point', 'hull' : 'vector_int', 'returnPoints' : ''},
-        'projectPoints' : { 'objectPoints' : 'vector_Point3f', 'imagePoints' : 'vector_Point2f',
-                            'distCoeffs' : 'vector_double' },
-        'initCameraMatrix2D' : {'objectPoints' : 'vector_vector_Point3f', 'imagePoints' : 'vector_vector_Point2f', },
-        'findChessboardCorners' : { 'corners' : 'vector_Point2f' },
-        'drawChessboardCorners' : { 'corners' : 'vector_Point2f' },
+        'randu'    : { 'low'  : {"ctype" : 'double'},
+                       'high' : {"ctype"    : 'double'} },
+        'randn'    : { 'mean'   : {"ctype" : 'double'},
+                       'stddev' : {"ctype"  : 'double'} },
+        'inRange'  : { 'lowerb' : {"ctype" : 'Scalar'},
+                       'upperb' : {"ctype" : 'Scalar'} },
+        'goodFeaturesToTrack' : { 'corners' : {"ctype" : 'vector_Point'} },
+        'findFundamentalMat'  : { 'points1' : {"ctype" : 'vector_Point2f'},
+                                  'points2' : {"ctype" : 'vector_Point2f'} },
+        'cornerSubPix' : { 'corners' : {"ctype" : 'vector_Point2f'} },
+        'minEnclosingCircle' : { 'points' : {"ctype" : 'vector_Point2f'} },
+        'findHomography' : { 'srcPoints' : {"ctype" : 'vector_Point2f'},
+                             'dstPoints' : {"ctype" : 'vector_Point2f'} },
+        'solvePnP' : { 'objectPoints' : {"ctype" : 'vector_Point3f'},
+                      'imagePoints'   : {"ctype" : 'vector_Point2f'},
+                      'distCoeffs'    : {"ctype" : 'vector_double' } },
+        'solvePnPRansac' : { 'objectPoints' : {"ctype" : 'vector_Point3f'},
+                             'imagePoints'  : {"ctype" : 'vector_Point2f'},
+                             'distCoeffs'   : {"ctype" : 'vector_double' } },
+        'calcOpticalFlowPyrLK' : { 'prevPts' : {"ctype" : 'vector_Point2f'},
+                                   'nextPts' : {"ctype" : 'vector_Point2f'},
+                                   'status'  : {"ctype" : 'vector_uchar'},
+                                   'err'     : {"ctype" : 'vector_float'} },
+        'fitEllipse' : { 'points' : {"ctype" : 'vector_Point2f'} },
+        'fillPoly'   : { 'pts' : {"ctype" : 'vector_vector_Point'} },
+        'polylines'  : { 'pts' : {"ctype" : 'vector_vector_Point'} },
+        'fillConvexPoly' : { 'points' : {"ctype" : 'vector_Point'} },
+        'boundingRect'   : { 'points' : {"ctype" : 'vector_Point'} },
+        'approxPolyDP' : { 'curve'       : {"ctype" : 'vector_Point2f'},
+                           'approxCurve' : {"ctype" : 'vector_Point2f'} },
+        'arcLength' : { 'curve' : {"ctype" : 'vector_Point2f'} },
+        'pointPolygonTest' : { 'contour' : {"ctype" : 'vector_Point2f'} },
+        'minAreaRect' : { 'points' : {"ctype" : 'vector_Point2f'} },
+        'getAffineTransform' : { 'src' : {"ctype" : 'vector_Point2f'},
+                                 'dst' : {"ctype" : 'vector_Point2f'} },
+        'hconcat' : { 'src' : {"ctype" : 'vector_Mat'} },
+        'vconcat' : { 'src' : {"ctype" : 'vector_Mat'} },
+        'undistortPoints' : { 'src' : {"ctype" : 'vector_Point2f'},
+                              'dst' : {"ctype" : 'vector_Point2f'} },
+        'checkRange' : {'pos' : {"ctype" : '*'} },
+        'meanStdDev' : { 'mean'   : {"ctype" : 'vector_double'},
+                         'stddev' : {"ctype" : 'vector_double'} },
+        'drawContours' : {'contours' : {"ctype" : 'vector_vector_Point'} },
+        'findContours' : {'contours' : {"ctype" : 'vector_vector_Point'} },
+        'convexityDefects' : { 'contour'          : {"ctype" : 'vector_Point'},
+                               'convexhull'       : {"ctype" : 'vector_int'},
+                               'convexityDefects' : {"ctype" : 'vector_Vec4i'} },
+        'isContourConvex' : { 'contour' : {"ctype" : 'vector_Point'} },
+        'convexHull' : { 'points' : {"ctype" : 'vector_Point'},
+                         'hull'   : {"ctype" : 'vector_int'},
+                         'returnPoints' : {"ctype" : ''} },
+        'projectPoints' : { 'objectPoints' : {"ctype" : 'vector_Point3f'},
+                            'imagePoints'  : {"ctype" : 'vector_Point2f'},
+                            'distCoeffs'   : {"ctype" : 'vector_double' } },
+        'initCameraMatrix2D' : { 'objectPoints' : {"ctype" : 'vector_vector_Point3f'},
+                                 'imagePoints'  : {"ctype" : 'vector_vector_Point2f'} },
+        'findChessboardCorners' : { 'corners' : {"ctype" : 'vector_Point2f'} },
+        'drawChessboardCorners' : { 'corners' : {"ctype" : 'vector_Point2f'} },
+        'mixChannels' : { 'dst' : {"attrib" : []} },
     }, # '', i.e. no class
 } # func_arg_fix
 
@@ -564,7 +632,8 @@ def getLibVersion(version_hpp_path):
     major = re.search("^W*#\W*define\W+CV_VERSION_MAJOR\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
     minor = re.search("^W*#\W*define\W+CV_VERSION_MINOR\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
     revision = re.search("^W*#\W*define\W+CV_VERSION_REVISION\W+(\d+)\W*$", version_file, re.MULTILINE).group(1)
-    return (epoch, major, minor, revision)
+    status = re.search("^W*#\W*define\W+CV_VERSION_STATUS\W+\"(.*?)\"\W*$", version_file, re.MULTILINE).group(1)
+    return (epoch, major, minor, revision, status)
 
 class ConstInfo(object):
     def __init__(self, cname, name, val, addedManually=False):
@@ -641,10 +710,12 @@ class FuncInfo(object):
         self.static = ["","static"][ "/S" in decl[2] ]
         self.ctype = re.sub(r"^CvTermCriteria", "TermCriteria", decl[1] or "")
         self.args = []
-        arg_fix_map = func_arg_fix.get(classname, {}).get(self.jname, {})
+        func_fix_map = func_arg_fix.get(classname, {}).get(self.jname, {})
         for a in decl[3]:
             arg = a[:]
-            arg[0] = arg_fix_map.get(arg[1], arg[0])
+            arg_fix_map = func_fix_map.get(arg[1], {})
+            arg[0] = arg_fix_map.get('ctype',  arg[0]) #fixing arg type
+            arg[3] = arg_fix_map.get('attrib', arg[3]) #fixing arg attrib
             ai = ArgInfo(arg)
             self.args.append(ai)
 
@@ -729,15 +800,19 @@ public class %(jc)s {
 """ % { 'm' : self.module, 'jc' : jname } )
 
         if class_name == 'Core':
-            (epoch, major, minor, revision) = getLibVersion(
+            (epoch, major, minor, revision, status) = getLibVersion(
                 (os.path.dirname(__file__) or '.') + '/../../core/include/opencv2/core/version.hpp')
-            version_str    = '.'.join( (epoch, major, minor, revision) )
+            version_str    = '.'.join( (epoch, major, minor, revision) ) + status
             version_suffix =  ''.join( (epoch, major, minor) )
             self.classes[class_name].imports.add("java.lang.String")
             self.java_code[class_name]["j_code"].write("""
     public static final String VERSION = "%(v)s", NATIVE_LIBRARY_NAME = "opencv_java%(vs)s";
-    public static final int VERSION_EPOCH = %(ep)s, VERSION_MAJOR = %(ma)s, VERSION_MINOR = %(mi)s, VERSION_REVISION = %(re)s;
-""" % { 'v' : version_str, 'vs' : version_suffix, 'ep' : epoch, 'ma' : major, 'mi' : minor, 're' : revision } )
+    public static final int VERSION_EPOCH = %(ep)s;
+    public static final int VERSION_MAJOR = %(ma)s;
+    public static final int VERSION_MINOR = %(mi)s;
+    public static final int VERSION_REVISION = %(re)s;
+    public static final String VERSION_STATUS = "%(st)s";
+""" % { 'v' : version_str, 'vs' : version_suffix, 'ep' : epoch, 'ma' : major, 'mi' : minor, 're' : revision, 'st': status } )
 
 
     def add_class(self, decl):
@@ -876,21 +951,51 @@ public class %(jc)s {
                     self.add_func(decl)
 
         self.cpp_code = StringIO()
-        self.cpp_code.write("""
+        self.cpp_code.write(Template("""
 //
 // This file is auto-generated, please don't edit!
 //
 
-#define LOG_TAG "org.opencv.%(m)s"
+#define LOG_TAG "org.opencv.$m"
 
 #include "common.h"
-#include "opencv2/%(m)s.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+#ifdef HAVE_OPENCV_$M
+
+#include <string>
+
+#include "opencv2/$m.hpp"
 
 using namespace cv;
 
+/// throw java exception
+static void throwJavaException(JNIEnv *env, const std::exception *e, const char *method) {
+  std::string what = "unknown exception";
+  jclass je = 0;
+
+  if(e) {
+    std::string exception_type = "std::exception";
+
+    if(dynamic_cast<const cv::Exception*>(e)) {
+      exception_type = "cv::Exception";
+      je = env->FindClass("org/opencv/core/CvException");
+    }
+
+    what = exception_type + ": " + e->what();
+  }
+
+  if(!je) je = env->FindClass("java/lang/Exception");
+  env->ThrowNew(je, what.c_str());
+
+  LOGE("%s caught %s", method, what.c_str());
+  (void)method;        // avoid "unused" warning
+}
+
+
 extern "C" {
 
-""" % {'m' : module} )
+""").substitute( m = module, M = module.upper() ) )
 
         # generate code for the classes
         for name in self.classes.keys():
@@ -905,7 +1010,7 @@ extern "C" {
             java_code = Template(java_code).substitute(imports = imports)
             self.save("%s/%s+%s.java" % (output_path, module, self.classes[name].jname), java_code)
 
-        self.cpp_code.write( '\n} // extern "C"\n' )
+        self.cpp_code.write( '\n} // extern "C"\n\n#endif // HAVE_OPENCV_%s\n' % module.upper() )
         self.save(output_path+"/"+module+".cpp",  self.cpp_code.getvalue())
 
         # report
@@ -996,6 +1101,11 @@ extern "C" {
             return
         for a in fi.args:
             if a.ctype not in type_dict:
+                if not a.defval and a.ctype.endswith("*"):
+                    a.defval = 0
+                if a.defval:
+                    a.ctype = ''
+                    continue
                 msg = "// Unknown type '%s' (%s), skipping the function\n\n" % (a.ctype, a.out or "I")
                 self.skipped_func_list.append(c_decl + "\n" + msg)
                 j_code.write( " "*4 + msg )
@@ -1196,7 +1306,7 @@ extern "C" {
                 ret = "return (jlong) _retval_;"
             elif fi.ctype.startswith('vector'): # c-tor
                 ret = "return (jlong) _retval_;"
-            elif fi.ctype == "string":
+            elif fi.ctype == "String":
                 ret = "return env->NewStringUTF(_retval_.c_str());"
                 default = 'return env->NewStringUTF("");'
             elif fi.ctype in self.classes: # wrapped class:
@@ -1218,8 +1328,8 @@ extern "C" {
             retval = fi.ctype + " _retval_ = "
             if fi.ctype == "void":
                 retval = ""
-            elif fi.ctype == "string":
-                retval = "std::" + retval
+            elif fi.ctype == "String":
+                retval = "cv::" + retval
             elif fi.ctype.startswith('vector'):
                 retval = type_dict[fi.ctype]['jni_var'] % {"n" : '_ret_val_vector_'} + " = "
                 c_epilogue.append("Mat* _retval_ = new Mat();")
@@ -1266,23 +1376,18 @@ JNIEXPORT $rtype JNICALL Java_org_opencv_${module}_${clazz}_$fname ($argst);
 JNIEXPORT $rtype JNICALL Java_org_opencv_${module}_${clazz}_$fname
   ($args)
 {
+    static const char method_name[] = "$module::$fname()";
     try {
-        LOGD("$module::$fname()");
+        LOGD("%s", method_name);
         $prologue
         $retval$cvname( $cvargs );
         $epilogue$ret
-    } catch(cv::Exception e) {
-        LOGD("$module::$fname() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        $default
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("$module::$fname() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {$module::$fname()}");
-        $default
+        throwJavaException(env, 0, method_name);
     }
+    $default
 }
 
 
@@ -1415,4 +1520,3 @@ if __name__ == "__main__":
     #print "Generating module '" + module + "' from headers:\n\t" + "\n\t".join(srcfiles)
     generator = JavaWrapperGenerator()
     generator.gen(srcfiles, module, dstdir)
-
diff --git a/modules/java/generator/gen_javadoc.py b/modules/java/generator/gen_javadoc.py
index 71372d3a2c..dfa591a959 100755
--- a/modules/java/generator/gen_javadoc.py
+++ b/modules/java/generator/gen_javadoc.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import os, sys, re, string, glob
 from optparse import OptionParser
diff --git a/modules/java/generator/rst_parser.py b/modules/java/generator/rst_parser.py
index 358392da65..86625f396c 100755
--- a/modules/java/generator/rst_parser.py
+++ b/modules/java/generator/rst_parser.py
@@ -1,7 +1,7 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import os, sys, re, string, fnmatch
-allmodules = ["core", "flann", "imgproc", "ml", "highgui", "video", "features2d", "calib3d", "objdetect", "legacy", "contrib", "gpu", "androidcamera", "java", "python", "stitching", "ts", "photo", "nonfree", "videostab", "ocl", "softcascade"]
+allmodules = ["core", "flann", "imgproc", "ml", "highgui", "video", "features2d", "calib3d", "objdetect", "legacy", "contrib", "gpu", "androidcamera", "java", "python", "stitching", "ts", "photo", "nonfree", "videostab", "ocl", "softcascade", "superres"]
 verbose = False
 show_warnings = True
 show_errors = True
@@ -79,8 +79,6 @@ class DeclarationParser(object):
             return "C"
         if line.startswith(".. ocv:pyfunction::"):
             return "Python2"
-        if line.startswith(".. ocv:pyoldfunction::"):
-            return "Python1"
         if line.startswith(".. ocv:jfunction::"):
             return "Java"
         return None
@@ -380,7 +378,7 @@ class RstParser(object):
 
     @classmethod
     def parse_namespace(cls, func, section_name):
-        known_namespaces = ["cv", "gpu", "flann"]
+        known_namespaces = ["cv", "gpu", "flann", "superres"]
         l = section_name.strip()
         for namespace in known_namespaces:
             if l.startswith(namespace + "::"):
diff --git a/modules/java/generator/src/cpp/Mat.cpp b/modules/java/generator/src/cpp/Mat.cpp
index 6b1d10038b..b3b0f66e72 100644
--- a/modules/java/generator/src/cpp/Mat.cpp
+++ b/modules/java/generator/src/cpp/Mat.cpp
@@ -1,10 +1,36 @@
 #define LOG_TAG "org.opencv.core.Mat"
 
+#include <stdexcept>
+#include <string>
+
 #include "common.h"
 #include "opencv2/core.hpp"
 
 using namespace cv;
 
+/// throw java exception
+static void throwJavaException(JNIEnv *env, const std::exception *e, const char *method) {
+  std::string what = "unknown exception";
+  jclass je = 0;
+
+  if(e) {
+    std::string exception_type = "std::exception";
+
+    if(dynamic_cast<const cv::Exception*>(e)) {
+      exception_type = "cv::Exception";
+      je = env->FindClass("org/opencv/core/CvException");
+    }
+
+    what = exception_type + ": " + e->what();
+  }
+
+  if(!je) je = env->FindClass("java/lang/Exception");
+  env->ThrowNew(je, what.c_str());
+
+  LOGE("%s caught %s", method, what.c_str());
+  (void)method;        // avoid "unused" warning
+}
+
 extern "C" {
 
 
@@ -35,24 +61,17 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__III
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__III
   (JNIEnv* env, jclass, jint rows, jint cols, jint type)
 {
+    static const char method_name[] = "Mat::n_1Mat__III()";
     try {
-        LOGD("Mat::n_1Mat__III()");
-
-        Mat* _retval_ = new Mat( rows, cols, type );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__III() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        LOGD("%s", method_name);
+        return (jlong) new Mat( rows, cols, type );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__III() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__III()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -67,24 +86,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__DDI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__DDI
   (JNIEnv* env, jclass, jdouble size_width, jdouble size_height, jint type)
 {
+    static const char method_name[] = "Mat::n_1Mat__DDI()";
     try {
-        LOGD("Mat::n_1Mat__DDI()");
+        LOGD("%s", method_name);
         Size size((int)size_width, (int)size_height);
-        Mat* _retval_ = new Mat( size, type );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__DDI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return (jlong) new Mat( size, type );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__DDI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__DDI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -100,24 +113,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__IIIDDDD
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__IIIDDDD
   (JNIEnv* env, jclass, jint rows, jint cols, jint type, jdouble s_val0, jdouble s_val1, jdouble s_val2, jdouble s_val3)
 {
+    static const char method_name[] = "Mat::n_1Mat__IIIDDDD()";
     try {
-        LOGD("Mat::n_1Mat__IIIDDDD()");
+        LOGD("%s", method_name);
         Scalar s(s_val0, s_val1, s_val2, s_val3);
-        Mat* _retval_ = new Mat( rows, cols, type, s );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__IIIDDDD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return (jlong) new Mat( rows, cols, type, s );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__IIIDDDD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__IIIDDDD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -132,25 +139,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__DDIDDDD
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__DDIDDDD
   (JNIEnv* env, jclass, jdouble size_width, jdouble size_height, jint type, jdouble s_val0, jdouble s_val1, jdouble s_val2, jdouble s_val3)
 {
+    static const char method_name[] = "Mat::n_1Mat__DDIDDDD()";
     try {
-        LOGD("Mat::n_1Mat__DDIDDDD()");
+        LOGD("%s", method_name);
         Size size((int)size_width, (int)size_height);
         Scalar s(s_val0, s_val1, s_val2, s_val3);
-        Mat* _retval_ = new Mat( size, type, s );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__DDIDDDD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return (jlong) new Mat( size, type, s );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__DDIDDDD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__DDIDDDD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -165,25 +166,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__JIIII
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__JIIII
   (JNIEnv* env, jclass, jlong m_nativeObj, jint rowRange_start, jint rowRange_end, jint colRange_start, jint colRange_end)
 {
+    static const char method_name[] = "Mat::n_1Mat__JIIII()";
     try {
-        LOGD("Mat::n_1Mat__JIIII()");
+        LOGD("%s", method_name);
         Range rowRange(rowRange_start, rowRange_end);
         Range colRange(colRange_start, colRange_end);
-        Mat* _retval_ = new Mat( (*(Mat*)m_nativeObj), rowRange, colRange );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__JIIII() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return (jlong) new Mat( (*(Mat*)m_nativeObj), rowRange, colRange );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__JIIII() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__JIIII()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -194,24 +189,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__JII
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1Mat__JII
   (JNIEnv* env, jclass, jlong m_nativeObj, jint rowRange_start, jint rowRange_end)
 {
+    static const char method_name[] = "Mat::n_1Mat__JII()";
     try {
-        LOGD("Mat::n_1Mat__JII()");
+        LOGD("%s", method_name);
         Range rowRange(rowRange_start, rowRange_end);
-        Mat* _retval_ = new Mat( (*(Mat*)m_nativeObj), rowRange );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1Mat__JII() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return (jlong) new Mat( (*(Mat*)m_nativeObj), rowRange );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1Mat__JII() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1Mat__JII()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -225,24 +214,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1adjustROI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1adjustROI
   (JNIEnv* env, jclass, jlong self, jint dtop, jint dbottom, jint dleft, jint dright)
 {
+    static const char method_name[] = "Mat::n_1adjustROI()";
     try {
-        LOGD("Mat::n_1adjustROI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->adjustROI( dtop, dbottom, dleft, dright );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1adjustROI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1adjustROI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1adjustROI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -257,23 +241,15 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1assignTo__JJI
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1assignTo__JJI
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jint type)
 {
+    static const char method_name[] = "Mat::n_1assignTo__JJI()";
     try {
-        LOGD("Mat::n_1assignTo__JJI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         me->assignTo( (*(Mat*)m_nativeObj), type );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1assignTo__JJI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1assignTo__JJI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1assignTo__JJI()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -284,23 +260,15 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1assignTo__JJ
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1assignTo__JJ
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1assignTo__JJ()";
     try {
-        LOGD("Mat::n_1assignTo__JJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         me->assignTo( (*(Mat*)m_nativeObj) );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1assignTo__JJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1assignTo__JJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1assignTo__JJ()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -316,24 +284,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1channels
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1channels
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1channels()";
     try {
-        LOGD("Mat::n_1channels()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->channels(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1channels() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->channels(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1channels() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1channels()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -348,24 +310,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JIIZ
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JIIZ
   (JNIEnv* env, jclass, jlong self, jint elemChannels, jint depth, jboolean requireContinuous)
 {
+    static const char method_name[] = "Mat::n_1checkVector__JIIZ()";
     try {
-        LOGD("Mat::n_1checkVector__JIIZ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->checkVector( elemChannels, depth, requireContinuous );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1checkVector__JIIZ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->checkVector( elemChannels, depth, requireContinuous );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1checkVector__JIIZ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1checkVector__JIIZ()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -376,24 +332,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JII
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JII
   (JNIEnv* env, jclass, jlong self, jint elemChannels, jint depth)
 {
+    static const char method_name[] = "Mat::n_1checkVector__JII()";
     try {
-        LOGD("Mat::n_1checkVector__JII()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->checkVector( elemChannels, depth );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1checkVector__JII() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->checkVector( elemChannels, depth );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1checkVector__JII() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1checkVector__JII()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -404,24 +354,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JI
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1checkVector__JI
   (JNIEnv* env, jclass, jlong self, jint elemChannels)
 {
+    static const char method_name[] = "Mat::n_1checkVector__JI()";
     try {
-        LOGD("Mat::n_1checkVector__JI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->checkVector( elemChannels );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1checkVector__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->checkVector( elemChannels );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1checkVector__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1checkVector__JI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -437,24 +381,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1clone
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1clone
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1clone()";
     try {
-        LOGD("Mat::n_1clone()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->clone(  );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1clone() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1clone() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1clone()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -469,24 +408,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1col
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1col
   (JNIEnv* env, jclass, jlong self, jint x)
 {
+    static const char method_name[] = "Mat::n_1col()";
     try {
-        LOGD("Mat::n_1col()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->col( x );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1col() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1col() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1col()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -501,24 +435,45 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1colRange
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1colRange
   (JNIEnv* env, jclass, jlong self, jint startcol, jint endcol)
 {
+    static const char method_name[] = "Mat::n_1colRange()";
     try {
-        LOGD("Mat::n_1colRange()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->colRange( startcol, endcol );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1colRange() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1colRange() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1colRange()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
+}
+
+
+
+//
+//  int Mat::dims()
+//
+
+JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1dims
+  (JNIEnv* env, jclass, jlong self);
+
+JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1dims
+  (JNIEnv* env, jclass, jlong self)
+{
+    static const char method_name[] = "Mat::n_1dims()";
+    try {
+        LOGD("%s", method_name);
+        Mat* me = (Mat*) self; //TODO: check for NULL
+        return me->dims;
+    } catch(cv::Exception e) {
+        throwJavaException(env, &e, method_name);
+    } catch (...) {
+        throwJavaException(env, 0, method_name);
+    }
+
+    return 0;
 }
 
 
@@ -533,24 +488,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1cols
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1cols
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1cols()";
     try {
-        LOGD("Mat::n_1cols()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->cols;
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1cols() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->cols;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1cols() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1cols()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -565,24 +514,16 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJIDD
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJIDD
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jint rtype, jdouble alpha, jdouble beta)
 {
+    static const char method_name[] = "Mat::n_1convertTo__JJIDD()";
     try {
-        LOGD("Mat::n_1convertTo__JJIDD()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         me->convertTo( m, rtype, alpha, beta );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1convertTo__JJIDD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1convertTo__JJIDD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1convertTo__JJIDD()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -593,24 +534,16 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJID
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJID
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jint rtype, jdouble alpha)
 {
+    static const char method_name[] = "Mat::n_1convertTo__JJID()";
     try {
-        LOGD("Mat::n_1convertTo__JJID()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         me->convertTo( m, rtype, alpha );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1convertTo__JJID() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1convertTo__JJID() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1convertTo__JJID()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -621,24 +554,16 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJI
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1convertTo__JJI
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jint rtype)
 {
+    static const char method_name[] = "Mat::n_1convertTo__JJI()";
     try {
-        LOGD("Mat::n_1convertTo__JJI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         me->convertTo( m, rtype );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1convertTo__JJI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1convertTo__JJI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1convertTo__JJI()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -654,24 +579,16 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1copyTo__JJ
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1copyTo__JJ
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1copyTo__JJ()";
     try {
-        LOGD("Mat::n_1copyTo__JJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         me->copyTo( m );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1copyTo__JJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1copyTo__JJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1copyTo__JJ()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -687,25 +604,17 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1copyTo__JJJ
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1copyTo__JJJ
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jlong mask_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1copyTo__JJJ()";
     try {
-        LOGD("Mat::n_1copyTo__JJJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         Mat& mask = *((Mat*)mask_nativeObj);
         me->copyTo( m, mask );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1copyTo__JJJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1copyTo__JJJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1copyTo__JJJ()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -721,23 +630,15 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1create__JIII
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1create__JIII
   (JNIEnv* env, jclass, jlong self, jint rows, jint cols, jint type)
 {
+    static const char method_name[] = "Mat::n_1create__JIII()";
     try {
-        LOGD("Mat::n_1create__JIII()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         me->create( rows, cols, type );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1create__JIII() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1create__JIII() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1create__JIII()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -753,24 +654,16 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1create__JDDI
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1create__JDDI
   (JNIEnv* env, jclass, jlong self, jdouble size_width, jdouble size_height, jint type)
 {
+    static const char method_name[] = "Mat::n_1create__JDDI()";
     try {
-        LOGD("Mat::n_1create__JDDI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Size size((int)size_width, (int)size_height);
         me->create( size, type );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1create__JDDI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1create__JDDI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1create__JDDI()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -786,25 +679,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1cross
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1cross
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1cross()";
     try {
-        LOGD("Mat::n_1cross()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         Mat _retval_ = me->cross( m );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1cross() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1cross() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1cross()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -836,24 +724,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1depth
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1depth
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1depth()";
     try {
-        LOGD("Mat::n_1depth()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->depth(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1depth() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->depth(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1depth() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1depth()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -868,24 +750,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1diag__JI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1diag__JI
   (JNIEnv* env, jclass, jlong self, jint d)
 {
+    static const char method_name[] = "Mat::n_1diag__JI()";
     try {
-        LOGD("Mat::n_1diag__JI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->diag( d );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1diag__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1diag__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1diag__JI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -901,24 +778,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1diag__J
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1diag__J
   (JNIEnv* env, jclass, jlong d_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1diag__J()";
     try {
-        LOGD("Mat::n_1diag__J()");
-
+        LOGD("%s", method_name);
         Mat _retval_ = Mat::diag( (*(Mat*)d_nativeObj) );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1diag__J() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1diag__J() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1diag__J()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -933,25 +804,19 @@ JNIEXPORT jdouble JNICALL Java_org_opencv_core_Mat_n_1dot
 JNIEXPORT jdouble JNICALL Java_org_opencv_core_Mat_n_1dot
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1dot()";
     try {
-        LOGD("Mat::n_1dot()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
-        double _retval_ = me->dot( m );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1dot() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->dot( m );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1dot() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1dot()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -966,24 +831,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1elemSize
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1elemSize
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1elemSize()";
     try {
-        LOGD("Mat::n_1elemSize()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        size_t _retval_ = me->elemSize(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1elemSize() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->elemSize(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1elemSize() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1elemSize()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -998,24 +857,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1elemSize1
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1elemSize1
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1elemSize1()";
     try {
-        LOGD("Mat::n_1elemSize1()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        size_t _retval_ = me->elemSize1(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1elemSize1() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->elemSize1(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1elemSize1() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1elemSize1()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1030,24 +883,18 @@ JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1empty
 JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1empty
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1empty()";
     try {
-        LOGD("Mat::n_1empty()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        bool _retval_ = me->empty(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1empty() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->empty(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1empty() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1empty()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1062,24 +909,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1eye__III
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1eye__III
   (JNIEnv* env, jclass, jint rows, jint cols, jint type)
 {
+    static const char method_name[] = "Mat::n_1eye__III()";
     try {
-        LOGD("Mat::n_1eye__III()");
-
+        LOGD("%s", method_name);
         Mat _retval_ = Mat::eye( rows, cols, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1eye__III() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1eye__III() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1eye__III()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1094,24 +935,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1eye__DDI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1eye__DDI
   (JNIEnv* env, jclass, jdouble size_width, jdouble size_height, jint type)
 {
+    static const char method_name[] = "Mat::n_1eye__DDI()";
     try {
-        LOGD("Mat::n_1eye__DDI()");
+        LOGD("%s", method_name);
         Size size((int)size_width, (int)size_height);
         Mat _retval_ = Mat::eye( size, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1eye__DDI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1eye__DDI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1eye__DDI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1126,24 +962,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1inv__JI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1inv__JI
   (JNIEnv* env, jclass, jlong self, jint method)
 {
+    static const char method_name[] = "Mat::n_1inv__JI()";
     try {
-        LOGD("Mat::n_1inv__JI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->inv( method );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1inv__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1inv__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1inv__JI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1153,24 +984,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1inv__J
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1inv__J
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1inv__J()";
     try {
-        LOGD("Mat::n_1inv__J()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->inv(  );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1inv__J() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1inv__J() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1inv__J()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1185,24 +1011,18 @@ JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1isContinuous
 JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1isContinuous
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1isContinuous()";
     try {
-        LOGD("Mat::n_1isContinuous()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        bool _retval_ = me->isContinuous(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1isContinuous() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->isContinuous(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1isContinuous() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1isContinuous()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1217,24 +1037,18 @@ JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1isSubmatrix
 JNIEXPORT jboolean JNICALL Java_org_opencv_core_Mat_n_1isSubmatrix
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1isSubmatrix()";
     try {
-        LOGD("Mat::n_1isSubmatrix()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        bool _retval_ = me->isSubmatrix(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1isSubmatrix() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->isSubmatrix(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1isSubmatrix() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1isSubmatrix()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1249,25 +1063,18 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_locateROI_10
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_locateROI_10
   (JNIEnv* env, jclass, jlong self, jdoubleArray wholeSize_out, jdoubleArray ofs_out)
 {
+    static const char method_name[] = "core::locateROI_10()";
     try {
-        LOGD("core::locateROI_10()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Size wholeSize;
         Point ofs;
         me->locateROI( wholeSize, ofs );
         jdouble tmp_wholeSize[2] = {wholeSize.width, wholeSize.height}; env->SetDoubleArrayRegion(wholeSize_out, 0, 2, tmp_wholeSize);  jdouble tmp_ofs[2] = {ofs.x, ofs.y}; env->SetDoubleArrayRegion(ofs_out, 0, 2, tmp_ofs);
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::locateROI_10() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::locateROI_10() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::locateROI_10()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -1283,25 +1090,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1mul__JJD
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1mul__JJD
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj, jdouble scale)
 {
+    static const char method_name[] = "Mat::n_1mul__JJD()";
     try {
-        LOGD("Mat::n_1mul__JJD()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         Mat _retval_ = me->mul( m, scale );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1mul__JJD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1mul__JJD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1mul__JJD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1312,25 +1114,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1mul__JJ
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1mul__JJ
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1mul__JJ()";
     try {
-        LOGD("Mat::n_1mul__JJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& m = *((Mat*)m_nativeObj);
         Mat _retval_ = me->mul( m );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1mul__JJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1mul__JJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1mul__JJ()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1345,24 +1142,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1ones__III
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1ones__III
   (JNIEnv* env, jclass, jint rows, jint cols, jint type)
 {
+    static const char method_name[] = "Mat::n_1ones__III()";
     try {
-        LOGD("Mat::n_1ones__III()");
-
+        LOGD("%s", method_name);
         Mat _retval_ = Mat::ones( rows, cols, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1ones__III() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1ones__III() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1ones__III()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1377,24 +1168,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1ones__DDI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1ones__DDI
   (JNIEnv* env, jclass, jdouble size_width, jdouble size_height, jint type)
 {
+    static const char method_name[] = "Mat::n_1ones__DDI()";
     try {
-        LOGD("Mat::n_1ones__DDI()");
+        LOGD("%s", method_name);
         Size size((int)size_width, (int)size_height);
         Mat _retval_ = Mat::ones( size, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1ones__DDI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1ones__DDI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1ones__DDI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1409,23 +1195,15 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1push_1back
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1push_1back
   (JNIEnv* env, jclass, jlong self, jlong m_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1push_1back()";
     try {
-        LOGD("Mat::n_1push_1back()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         me->push_back( (*(Mat*)m_nativeObj) );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1push_1back() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1push_1back() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1push_1back()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -1441,23 +1219,15 @@ JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1release
 JNIEXPORT void JNICALL Java_org_opencv_core_Mat_n_1release
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1release()";
     try {
-        LOGD("Mat::n_1release()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         me->release(  );
-
-        return;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1release() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1release() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1release()}");
-        return;
+        throwJavaException(env, 0, method_name);
     }
 }
 
@@ -1473,24 +1243,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1reshape__JII
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1reshape__JII
   (JNIEnv* env, jclass, jlong self, jint cn, jint rows)
 {
+    static const char method_name[] = "Mat::n_1reshape__JII()";
     try {
-        LOGD("Mat::n_1reshape__JII()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->reshape( cn, rows );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1reshape__JII() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1reshape__JII() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1reshape__JII()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1501,24 +1266,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1reshape__JI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1reshape__JI
   (JNIEnv* env, jclass, jlong self, jint cn)
 {
+    static const char method_name[] = "Mat::n_1reshape__JI()";
     try {
-        LOGD("Mat::n_1reshape__JI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->reshape( cn );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1reshape__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1reshape__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1reshape__JI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1533,24 +1293,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1row
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1row
   (JNIEnv* env, jclass, jlong self, jint y)
 {
+    static const char method_name[] = "Mat::n_1row()";
     try {
-        LOGD("Mat::n_1row()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->row( y );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1row() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1row() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1row()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1565,24 +1320,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1rowRange
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1rowRange
   (JNIEnv* env, jclass, jlong self, jint startrow, jint endrow)
 {
+    static const char method_name[] = "Mat::n_1rowRange()";
     try {
-        LOGD("Mat::n_1rowRange()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->rowRange( startrow, endrow );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1rowRange() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1rowRange() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1rowRange()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1597,24 +1347,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1rows
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1rows
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1rows()";
     try {
-        LOGD("Mat::n_1rows()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->rows;
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1rows() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->rows;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1rows() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1rows()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1629,25 +1373,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JDDDD
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JDDDD
   (JNIEnv* env, jclass, jlong self, jdouble s_val0, jdouble s_val1, jdouble s_val2, jdouble s_val3)
 {
+    static const char method_name[] = "Mat::n_1setTo__JDDDD()";
     try {
-        LOGD("Mat::n_1setTo__JDDDD()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Scalar s(s_val0, s_val1, s_val2, s_val3);
         Mat _retval_ = me->operator =( s );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1setTo__JDDDD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1setTo__JDDDD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1setTo__JDDDD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1662,26 +1401,21 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JDDDDJ
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JDDDDJ
   (JNIEnv* env, jclass, jlong self, jdouble s_val0, jdouble s_val1, jdouble s_val2, jdouble s_val3, jlong mask_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1setTo__JDDDDJ()";
     try {
-        LOGD("Mat::n_1setTo__JDDDDJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Scalar s(s_val0, s_val1, s_val2, s_val3);
         Mat& mask = *((Mat*)mask_nativeObj);
         Mat _retval_ = me->setTo( s, mask );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1setTo__JDDDDJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1setTo__JDDDDJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1setTo__JDDDDJ()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1696,26 +1430,21 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JJJ
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JJJ
   (JNIEnv* env, jclass, jlong self, jlong value_nativeObj, jlong mask_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1setTo__JJJ()";
     try {
-        LOGD("Mat::n_1setTo__JJJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& value = *((Mat*)value_nativeObj);
         Mat& mask = *((Mat*)mask_nativeObj);
         Mat _retval_ = me->setTo( value, mask );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1setTo__JJJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1setTo__JJJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1setTo__JJJ()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1726,25 +1455,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JJ
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1setTo__JJ
   (JNIEnv* env, jclass, jlong self, jlong value_nativeObj)
 {
+    static const char method_name[] = "Mat::n_1setTo__JJ()";
     try {
-        LOGD("Mat::n_1setTo__JJ()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat& value = *((Mat*)value_nativeObj);
         Mat _retval_ = me->setTo( value );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1setTo__JJ() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1setTo__JJ() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1setTo__JJ()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1759,24 +1483,22 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_n_1size
 JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_n_1size
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1size()";
     try {
-        LOGD("Mat::n_1size()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Size _retval_ = me->size(  );
-        jdoubleArray _da_retval_ = env->NewDoubleArray(2);  jdouble _tmp_retval_[2] = {_retval_.width, _retval_.height}; env->SetDoubleArrayRegion(_da_retval_, 0, 2, _tmp_retval_);
+        jdoubleArray _da_retval_ = env->NewDoubleArray(2);
+        jdouble _tmp_retval_[2] = {_retval_.width, _retval_.height};
+        env->SetDoubleArrayRegion(_da_retval_, 0, 2, _tmp_retval_);
         return _da_retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1size() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1size() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1size()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1791,24 +1513,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1step1__JI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1step1__JI
   (JNIEnv* env, jclass, jlong self, jint i)
 {
+    static const char method_name[] = "Mat::n_1step1__JI()";
     try {
-        LOGD("Mat::n_1step1__JI()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        size_t _retval_ = me->step1( i );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1step1__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->step1( i );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1step1__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1step1__JI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1819,24 +1535,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1step1__J
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1step1__J
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1step1__J()";
     try {
-        LOGD("Mat::n_1step1__J()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        size_t _retval_ = me->step1(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1step1__J() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->step1(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1step1__J() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1step1__J()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 //
@@ -1849,26 +1559,21 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1submat_1rr
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1submat_1rr
   (JNIEnv* env, jclass, jlong self, jint rowRange_start, jint rowRange_end, jint colRange_start, jint colRange_end)
 {
+    static const char method_name[] = "Mat::n_1submat_1rr()";
     try {
-        LOGD("Mat::n_1submat_1rr()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Range rowRange(rowRange_start, rowRange_end);
         Range colRange(colRange_start, colRange_end);
         Mat _retval_ = me->operator()( rowRange, colRange );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1submat_1rr() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1submat_1rr() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1submat_1rr()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1883,25 +1588,20 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1submat
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1submat
   (JNIEnv* env, jclass, jlong self, jint roi_x, jint roi_y, jint roi_width, jint roi_height)
 {
+    static const char method_name[] = "Mat::n_1submat()";
     try {
-        LOGD("Mat::n_1submat()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Rect roi(roi_x, roi_y, roi_width, roi_height);
         Mat _retval_ = me->operator()( roi );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1submat() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1submat() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1submat()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1916,24 +1616,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1t
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1t
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1t()";
     try {
-        LOGD("Mat::n_1t()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
         Mat _retval_ = me->t(  );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1t() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1t() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1t()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1948,24 +1643,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1total
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1total
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1total()";
     try {
-        LOGD("Mat::n_1total()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        size_t _retval_ = me->total(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1total() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->total(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1total() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1total()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -1980,24 +1669,18 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1type
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_n_1type
   (JNIEnv* env, jclass, jlong self)
 {
+    static const char method_name[] = "Mat::n_1type()";
     try {
-        LOGD("Mat::n_1type()");
+        LOGD("%s", method_name);
         Mat* me = (Mat*) self; //TODO: check for NULL
-        int _retval_ = me->type(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1type() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+        return me->type(  );
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1type() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1type()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -2012,24 +1695,18 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1zeros__III
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1zeros__III
   (JNIEnv* env, jclass, jint rows, jint cols, jint type)
 {
+    static const char method_name[] = "Mat::n_1zeros__III()";
     try {
-        LOGD("Mat::n_1zeros__III()");
-
+        LOGD("%s", method_name);
         Mat _retval_ = Mat::zeros( rows, cols, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1zeros__III() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1zeros__III() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1zeros__III()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -2044,24 +1721,19 @@ JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1zeros__DDI
 JNIEXPORT jlong JNICALL Java_org_opencv_core_Mat_n_1zeros__DDI
   (JNIEnv* env, jclass, jdouble size_width, jdouble size_height, jint type)
 {
+    static const char method_name[] = "Mat::n_1zeros__DDI()";
     try {
-        LOGD("Mat::n_1zeros__DDI()");
+        LOGD("%s", method_name);
         Size size((int)size_width, (int)size_height);
         Mat _retval_ = Mat::zeros( size, type );
-
         return (jlong) new Mat(_retval_);
-    } catch(cv::Exception e) {
-        LOGD("Mat::n_1zeros__DDI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::n_1zeros__DDI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::n_1zeros__DDI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -2089,8 +1761,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutD
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutD
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jdoubleArray vals)
 {
+    static const char method_name[] = "Mat::nPutD()";
     try {
-        LOGD("Mat::nPutD()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(!me || !me->data) return 0;  // no native object behind
         if(me->rows<=row || me->cols<=col) return 0; // indexes out of range
@@ -2130,18 +1803,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutD
 
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nPutD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nPutD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nPutD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -2186,8 +1854,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutB
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutB
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jbyteArray vals)
 {
+    static const char method_name[] = "Mat::nPutB()";
     try {
-        LOGD("Mat::nPutB()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_8U && me->depth() != CV_8S) return 0; // incompatible type
@@ -2197,18 +1866,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutB
         int res = mat_put<char>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nPutB() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nPutB() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nPutB()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutS
@@ -2217,8 +1881,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutS
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutS
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jshortArray vals)
 {
+    static const char method_name[] = "Mat::nPutS()";
     try {
-        LOGD("Mat::nPutS()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_16U && me->depth() != CV_16S) return 0; // incompatible type
@@ -2228,18 +1893,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutS
         int res = mat_put<short>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nPutS() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nPutS() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nPutS()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutI
@@ -2248,8 +1908,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutI
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutI
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jintArray vals)
 {
+    static const char method_name[] = "Mat::nPutI()";
     try {
-        LOGD("Mat::nPutI()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_32S) return 0; // incompatible type
@@ -2259,18 +1920,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutI
         int res = mat_put<int>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nPutI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nPutI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nPutI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutF
@@ -2279,8 +1935,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutF
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutF
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jfloatArray vals)
 {
+    static const char method_name[] = "Mat::nPutF()";
     try {
-        LOGD("Mat::nPutF()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_32F) return 0; // incompatible type
@@ -2290,18 +1947,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nPutF
         int res = mat_put<float>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nPutF() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nPutF() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nPutF()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 
@@ -2345,8 +1997,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetB
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetB
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jbyteArray vals)
 {
+    static const char method_name[] = "Mat::nGetB()";
     try {
-        LOGD("Mat::nGetB()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_8U && me->depth() != CV_8S) return 0; // incompatible type
@@ -2356,18 +2009,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetB
         int res = mat_get<char>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGetB() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGetB() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGetB()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetS
@@ -2376,8 +2024,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetS
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetS
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jshortArray vals)
 {
+    static const char method_name[] = "Mat::nGetS()";
     try {
-        LOGD("Mat::nGetS()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_16U && me->depth() != CV_16S) return 0; // incompatible type
@@ -2387,18 +2036,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetS
         int res = mat_get<short>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGetS() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGetS() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGetS()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetI
@@ -2407,8 +2051,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetI
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetI
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jintArray vals)
 {
+    static const char method_name[] = "Mat::nGetI()";
     try {
-        LOGD("Mat::nGetI()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_32S) return 0; // incompatible type
@@ -2418,18 +2063,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetI
         int res = mat_get<int>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGetI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGetI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGetI()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetF
@@ -2438,8 +2078,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetF
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetF
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jfloatArray vals)
 {
+    static const char method_name[] = "Mat::nGetF()";
     try {
-        LOGD("Mat::nGetF()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_32F) return 0; // incompatible type
@@ -2449,18 +2090,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetF
         int res = mat_get<float>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGetF() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGetF() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGetF()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetD
@@ -2469,8 +2105,9 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetD
 JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetD
     (JNIEnv* env, jclass, jlong self, jint row, jint col, jint count, jdoubleArray vals)
 {
+    static const char method_name[] = "Mat::nGetD()";
     try {
-        LOGD("Mat::nGetD()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->depth() != CV_64F) return 0; // incompatible type
@@ -2480,18 +2117,13 @@ JNIEXPORT jint JNICALL Java_org_opencv_core_Mat_nGetD
         int res = mat_get<double>(me, row, col, count, values);
         env->ReleasePrimitiveArrayCritical(vals, values, 0);
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGetD() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGetD() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGetD()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_nGet
@@ -2500,8 +2132,9 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_nGet
 JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_nGet
     (JNIEnv* env, jclass, jlong self, jint row, jint col)
 {
+    static const char method_name[] = "Mat::nGet()";
     try {
-        LOGD("Mat::nGet()");
+        LOGD("%s", method_name);
         cv::Mat* me = (cv::Mat*) self;
         if(! self) return 0; // no native object behind
         if(me->rows<=row || me->cols<=col) return 0; // indexes out of range
@@ -2522,18 +2155,13 @@ JNIEXPORT jdoubleArray JNICALL Java_org_opencv_core_Mat_nGet
             env->SetDoubleArrayRegion(res, 0, me->channels(), buff);
         }
         return res;
-    } catch(cv::Exception e) {
-        LOGD("Mat::nGet() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
     } catch (...) {
-        LOGD("Mat::nGet() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {Mat::nGet()}");
-        return 0;
+        throwJavaException(env, 0, method_name);
     }
+
+    return 0;
 }
 
 JNIEXPORT jstring JNICALL Java_org_opencv_core_Mat_nDump
@@ -2542,25 +2170,24 @@ JNIEXPORT jstring JNICALL Java_org_opencv_core_Mat_nDump
 JNIEXPORT jstring JNICALL Java_org_opencv_core_Mat_nDump
   (JNIEnv *env, jclass, jlong self)
 {
-    cv::Mat* me = (cv::Mat*) self; //TODO: check for NULL
-    std::stringstream s;
+    static const char method_name[] = "Mat::nDump()";
     try {
-            LOGD("Mat::nDump()");
-
-            s << *me;
-            return env->NewStringUTF(s.str().c_str());
-        } catch(cv::Exception e) {
-            LOGE("Mat::nDump() catched cv::Exception: %s", e.what());
-            jclass je = env->FindClass("org/opencv/core/CvException");
-            if(!je) je = env->FindClass("java/lang/Exception");
-            env->ThrowNew(je, e.what());
-            return env->NewStringUTF("ERROR");
-        } catch (...) {
-            LOGE("Mat::nDump() catched unknown exception (...)");
-            jclass je = env->FindClass("java/lang/Exception");
-            env->ThrowNew(je, "Unknown exception in JNI code {Mat::nDump()}");
-            return env->NewStringUTF("ERROR");
+        LOGD("%s", method_name);
+        cv::Mat* me = (cv::Mat*) self; //TODO: check for NULL
+        String s;
+        Ptr<Formatted> fmtd = Formatter::get()->format(*me);
+        for(const char* str = fmtd->next(); str; str = fmtd->next())
+        {
+            s = s + String(str);
         }
+        return env->NewStringUTF(s.c_str());
+    } catch(const std::exception &e) {
+        throwJavaException(env, &e, method_name);
+    } catch (...) {
+        throwJavaException(env, 0, method_name);
+    }
+
+    return 0;
 }
 
 
diff --git a/modules/java/generator/src/cpp/VideoCapture.cpp b/modules/java/generator/src/cpp/VideoCapture.cpp
deleted file mode 100644
index f0e5c233b6..0000000000
--- a/modules/java/generator/src/cpp/VideoCapture.cpp
+++ /dev/null
@@ -1,435 +0,0 @@
-#define LOG_TAG "org.opencv.highgui.VideoCapture"
-#include "common.h"
-
-#include "opencv2/opencv_modules.hpp"
-#ifdef HAVE_OPENCV_HIGHGUI
-
-#include "opencv2/highgui/highgui_c.h"
-#include "opencv2/highgui.hpp"
-using namespace cv;
-
-
-extern "C" {
-
-//
-//   VideoCapture::VideoCapture()
-//
-
-JNIEXPORT jlong JNICALL Java_org_opencv_highgui_VideoCapture_n_1VideoCapture__
-  (JNIEnv* env, jclass);
-
-JNIEXPORT jlong JNICALL Java_org_opencv_highgui_VideoCapture_n_1VideoCapture__
-  (JNIEnv* env, jclass)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__()");
-
-        VideoCapture* _retval_ = new VideoCapture(  );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1VideoCapture__()}");
-        return 0;
-    }
-}
-
-
-//
-//   VideoCapture::VideoCapture(int device)
-//
-
-JNIEXPORT jlong JNICALL Java_org_opencv_highgui_VideoCapture_n_1VideoCapture__I
-  (JNIEnv* env, jclass, jint device);
-
-JNIEXPORT jlong JNICALL Java_org_opencv_highgui_VideoCapture_n_1VideoCapture__I
-  (JNIEnv* env, jclass, jint device)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__I()");
-
-        VideoCapture* _retval_ = new VideoCapture( device );
-
-        return (jlong) _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__I() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1VideoCapture__I() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1VideoCapture__I()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  double VideoCapture::get(int propId)
-//
-
-JNIEXPORT jdouble JNICALL Java_org_opencv_highgui_VideoCapture_n_1get
-  (JNIEnv* env, jclass, jlong self, jint propId);
-
-JNIEXPORT jdouble JNICALL Java_org_opencv_highgui_VideoCapture_n_1get
-  (JNIEnv* env, jclass, jlong self, jint propId)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1get()");
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        double _retval_ = me->get( propId );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1get() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1get() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1get()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  bool VideoCapture::grab()
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1grab
-  (JNIEnv* env, jclass, jlong self);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1grab
-  (JNIEnv* env, jclass, jlong self)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1grab()");
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        bool _retval_ = me->grab(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1grab() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1grab() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1grab()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  bool VideoCapture::isOpened()
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1isOpened
-  (JNIEnv* env, jclass, jlong self);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1isOpened
-  (JNIEnv* env, jclass, jlong self)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1isOpened()");
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        bool _retval_ = me->isOpened(  );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1isOpened() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1isOpened() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1isOpened()}");
-        return 0;
-    }
-}
-
-
-//
-//  bool VideoCapture::open(int device)
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1open__JI
-  (JNIEnv* env, jclass, jlong self, jint device);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1open__JI
-  (JNIEnv* env, jclass, jlong self, jint device)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1open__JI()");
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        bool _retval_ = me->open( device );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1open__JI() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1open__JI() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1open__JI()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  bool VideoCapture::read(Mat image)
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1read
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1read
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj)
-{
-    try {
-        LOGD("highgui::VideoCapture_n_1read()");
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        Mat& image = *((Mat*)image_nativeObj);
-        bool _retval_ = me->read( image );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-        LOGD("highgui::VideoCapture_n_1read() catched cv::Exception: %s", e.what());
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-        LOGD("highgui::VideoCapture_n_1read() catched unknown exception (...)");
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1read()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  void VideoCapture::release()
-//
-
-JNIEXPORT void JNICALL Java_org_opencv_highgui_VideoCapture_n_1release
-  (JNIEnv* env, jclass, jlong self);
-
-JNIEXPORT void JNICALL Java_org_opencv_highgui_VideoCapture_n_1release
-  (JNIEnv* env, jclass, jlong self)
-{
-    try {
-
-        LOGD("highgui::VideoCapture_n_1release()");
-
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        me->release(  );
-
-        return;
-    } catch(cv::Exception e) {
-
-        LOGD("highgui::VideoCapture_n_1release() catched cv::Exception: %s", e.what());
-
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return;
-    } catch (...) {
-
-        LOGD("highgui::VideoCapture_n_1release() catched unknown exception (...)");
-
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1release()}");
-        return;
-    }
-}
-
-
-
-//
-//  bool VideoCapture::retrieve(Mat image, int channel = 0)
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1retrieve__JJI
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj, jint channel);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1retrieve__JJI
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj, jint channel)
-{
-    try {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJI()");
-
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        Mat& image = *((Mat*)image_nativeObj);
-        bool _retval_ = me->retrieve( image, channel );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJI() catched cv::Exception: %s", e.what());
-
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJI() catched unknown exception (...)");
-
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1retrieve__JJI()}");
-        return 0;
-    }
-}
-
-
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1retrieve__JJ
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1retrieve__JJ
-  (JNIEnv* env, jclass, jlong self, jlong image_nativeObj)
-{
-    try {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJ()");
-
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        Mat& image = *((Mat*)image_nativeObj);
-        bool _retval_ = me->retrieve( image );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJ() catched cv::Exception: %s", e.what());
-
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-
-        LOGD("highgui::VideoCapture_n_1retrieve__JJ() catched unknown exception (...)");
-
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1retrieve__JJ()}");
-        return 0;
-    }
-}
-
-
-
-//
-//  bool VideoCapture::set(int propId, double value)
-//
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1set
-  (JNIEnv* env, jclass, jlong self, jint propId, jdouble value);
-
-JNIEXPORT jboolean JNICALL Java_org_opencv_highgui_VideoCapture_n_1set
-  (JNIEnv* env, jclass, jlong self, jint propId, jdouble value)
-{
-    try {
-
-        LOGD("highgui::VideoCapture_n_1set()");
-
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        bool _retval_ = me->set( propId, value );
-
-        return _retval_;
-    } catch(cv::Exception e) {
-
-        LOGD("highgui::VideoCapture_n_1set() catched cv::Exception: %s", e.what());
-
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return 0;
-    } catch (...) {
-
-        LOGD("highgui::VideoCapture_n_1set() catched unknown exception (...)");
-
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1set()}");
-        return 0;
-    }
-}
-
-JNIEXPORT jstring JNICALL Java_org_opencv_highgui_VideoCapture_n_1getSupportedPreviewSizes
-  (JNIEnv *env, jclass, jlong self);
-
-JNIEXPORT jstring JNICALL Java_org_opencv_highgui_VideoCapture_n_1getSupportedPreviewSizes
-  (JNIEnv *env, jclass, jlong self)
-{
-    try {
-
-        LOGD("highgui::VideoCapture_n_1set()");
-
-        VideoCapture* me = (VideoCapture*) self; //TODO: check for NULL
-        union {double prop; const char* name;} u;
-        u.prop = me->get(CV_CAP_PROP_SUPPORTED_PREVIEW_SIZES_STRING);
-        return env->NewStringUTF(u.name);
-    } catch(cv::Exception e) {
-
-        LOGD("highgui::VideoCapture_n_1getSupportedPreviewSizes() catched cv::Exception: %s", e.what());
-
-        jclass je = env->FindClass("org/opencv/core/CvException");
-        if(!je) je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, e.what());
-        return env->NewStringUTF("");
-    } catch (...) {
-
-        LOGD("highgui::VideoCapture_n_1getSupportedPreviewSizes() catched unknown exception (...)");
-
-        jclass je = env->FindClass("java/lang/Exception");
-        env->ThrowNew(je, "Unknown exception in JNI code {highgui::VideoCapture_n_1getSupportedPreviewSizes()}");
-        return env->NewStringUTF("");
-    }
-}
-
-
-
-//
-//  native support for java finalize()
-//  static void VideoCapture::n_delete( __int64 self )
-//
-
-JNIEXPORT void JNICALL Java_org_opencv_highgui_VideoCapture_n_1delete
-  (JNIEnv*, jclass, jlong self);
-
-JNIEXPORT void JNICALL Java_org_opencv_highgui_VideoCapture_n_1delete
-  (JNIEnv*, jclass, jlong self)
-{
-    delete (VideoCapture*) self;
-}
-
-} // extern "C"
-
-#endif // HAVE_OPENCV_HIGHGUI
\ No newline at end of file
diff --git a/modules/java/generator/src/cpp/common.h b/modules/java/generator/src/cpp/common.h
index b67f633f7b..0c0b025f57 100644
--- a/modules/java/generator/src/cpp/common.h
+++ b/modules/java/generator/src/cpp/common.h
@@ -20,6 +20,8 @@
 #  define LOGD(...)
 #endif
 
+#include "opencv2/core/utility.hpp"
+
 #include "converters.h"
 
 #include "core_manual.hpp"
@@ -30,4 +32,4 @@
 #  pragma warning(disable:4800 4244)
 #endif
 
-#endif //__JAVA_COMMON_H__
\ No newline at end of file
+#endif //__JAVA_COMMON_H__
diff --git a/modules/java/generator/src/cpp/core_manual.cpp b/modules/java/generator/src/cpp/core_manual.cpp
index d7ba5b9947..97ce27ee37 100644
--- a/modules/java/generator/src/cpp/core_manual.cpp
+++ b/modules/java/generator/src/cpp/core_manual.cpp
@@ -1,6 +1,8 @@
 #define LOG_TAG "org.opencv.core.Core"
 #include "common.h"
 
+#include "opencv2/core/utility.hpp"
+
 static int quietCallback( int, const char*, const char*, const char*, int, void* )
 {
     return 0;
@@ -12,4 +14,4 @@ void cv::setErrorVerbosity(bool verbose)
         cv::redirectError(0);
     else
         cv::redirectError((cv::ErrorCallback)quietCallback);
-}
\ No newline at end of file
+}
diff --git a/modules/java/generator/src/cpp/features2d_manual.hpp b/modules/java/generator/src/cpp/features2d_manual.hpp
index 39fcafa2dd..723f820e08 100644
--- a/modules/java/generator/src/cpp/features2d_manual.hpp
+++ b/modules/java/generator/src/cpp/features2d_manual.hpp
@@ -11,15 +11,17 @@
 namespace cv
 {
 
-class CV_EXPORTS_AS(FeatureDetector) javaFeatureDetector : public FeatureDetector
+class CV_EXPORTS_AS(FeatureDetector) javaFeatureDetector
 {
 public:
-#if 0
-    //DO NOT REMOVE! The block is required for sources parser
-    CV_WRAP void detect( const Mat& image, CV_OUT std::vector<KeyPoint>& keypoints, const Mat& mask=Mat() ) const;
-    CV_WRAP void detect( const std::vector<Mat>& images, CV_OUT std::vector<std::vector<KeyPoint> >& keypoints, const std::vector<Mat>& masks=std::vector<Mat>() ) const;
-    CV_WRAP virtual bool empty() const;
-#endif
+    CV_WRAP void detect( const Mat& image, CV_OUT std::vector<KeyPoint>& keypoints, const Mat& mask=Mat() ) const
+    { return wrapped->detect(image, keypoints, mask); }
+
+    CV_WRAP void detect( const std::vector<Mat>& images, CV_OUT std::vector<std::vector<KeyPoint> >& keypoints, const std::vector<Mat>& masks=std::vector<Mat>() ) const
+    { return wrapped->detect(images, keypoints, masks); }
+
+    CV_WRAP bool empty() const
+    { return wrapped->empty(); }
 
     enum
     {
@@ -84,7 +86,7 @@ public:
     //not supported: SimpleBlob, Dense
     CV_WRAP static javaFeatureDetector* create( int detectorType )
     {
-        std::string name;
+        String name;
         if (detectorType > DYNAMICDETECTOR)
         {
             name = "Dynamic";
@@ -104,89 +106,111 @@ public:
         switch(detectorType)
         {
         case FAST:
-            name += "FAST";
+            name = name + "FAST";
             break;
         case STAR:
-            name += "STAR";
+            name = name + "STAR";
             break;
         case SIFT:
-            name += "SIFT";
+            name = name + "SIFT";
             break;
         case SURF:
-            name += "SURF";
+            name = name + "SURF";
             break;
         case ORB:
-            name += "ORB";
+            name = name + "ORB";
             break;
         case MSER:
-            name += "MSER";
+            name = name + "MSER";
             break;
         case GFTT:
-            name += "GFTT";
+            name = name + "GFTT";
             break;
         case HARRIS:
-            name += "HARRIS";
+            name = name + "HARRIS";
             break;
         case SIMPLEBLOB:
-            name += "SimpleBlob";
+            name = name + "SimpleBlob";
             break;
         case DENSE:
-            name += "Dense";
+            name = name + "Dense";
             break;
         case BRISK:
-            name += "BRISK";
+            name = name + "BRISK";
             break;
         default:
-            CV_Error( CV_StsBadArg, "Specified feature detector type is not supported." );
+            CV_Error( Error::StsBadArg, "Specified feature detector type is not supported." );
             break;
         }
 
-        Ptr<FeatureDetector> detector = FeatureDetector::create(name);
-        detector.addref();
-        return (javaFeatureDetector*)((FeatureDetector*) detector);
+        return new javaFeatureDetector(FeatureDetector::create(name));
     }
 
-    CV_WRAP void write( const std::string& fileName ) const
+    CV_WRAP void write( const String& fileName ) const
     {
         FileStorage fs(fileName, FileStorage::WRITE);
-        ((FeatureDetector*)this)->write(fs);
-        fs.release();
+        wrapped->write(fs);
     }
 
-    CV_WRAP void read( const std::string& fileName )
+    CV_WRAP void read( const String& fileName )
     {
         FileStorage fs(fileName, FileStorage::READ);
-        ((FeatureDetector*)this)->read(fs.root());
-        fs.release();
+        wrapped->read(fs.root());
     }
+
+private:
+    javaFeatureDetector(Ptr<FeatureDetector> _wrapped) : wrapped(_wrapped)
+    {}
+
+    Ptr<FeatureDetector> wrapped;
 };
 
-class CV_EXPORTS_AS(DescriptorMatcher) javaDescriptorMatcher : public DescriptorMatcher
+class CV_EXPORTS_AS(DescriptorMatcher) javaDescriptorMatcher
 {
 public:
-#if 0
-    //DO NOT REMOVE! The block is required for sources parser
-    CV_WRAP virtual bool isMaskSupported() const;
-    CV_WRAP virtual void add( const std::vector<Mat>& descriptors );
-    CV_WRAP const std::vector<Mat>& getTrainDescriptors() const;
-    CV_WRAP virtual void clear();
-    CV_WRAP virtual bool empty() const;
-    CV_WRAP virtual void train();
+    CV_WRAP bool isMaskSupported() const
+    { return wrapped->isMaskSupported(); }
+
+    CV_WRAP void add( const std::vector<Mat>& descriptors )
+    { return wrapped->add(descriptors); }
+
+    CV_WRAP const std::vector<Mat>& getTrainDescriptors() const
+    { return wrapped->getTrainDescriptors(); }
+
+    CV_WRAP void clear()
+    { return wrapped->clear(); }
+
+    CV_WRAP bool empty() const
+    { return wrapped->empty(); }
+
+    CV_WRAP void train()
+    { return wrapped->train(); }
+
     CV_WRAP void match( const Mat& queryDescriptors, const Mat& trainDescriptors,
-                CV_OUT std::vector<DMatch>& matches, const Mat& mask=Mat() ) const;
+                CV_OUT std::vector<DMatch>& matches, const Mat& mask=Mat() ) const
+    { return wrapped->match(queryDescriptors, trainDescriptors, matches, mask); }
+
     CV_WRAP void knnMatch( const Mat& queryDescriptors, const Mat& trainDescriptors,
                    CV_OUT std::vector<std::vector<DMatch> >& matches, int k,
-                   const Mat& mask=Mat(), bool compactResult=false ) const;
+                   const Mat& mask=Mat(), bool compactResult=false ) const
+    { return wrapped->knnMatch(queryDescriptors, trainDescriptors, matches, k, mask, compactResult); }
+
     CV_WRAP void radiusMatch( const Mat& queryDescriptors, const Mat& trainDescriptors,
                       CV_OUT std::vector<std::vector<DMatch> >& matches, float maxDistance,
-                      const Mat& mask=Mat(), bool compactResult=false ) const;
+                      const Mat& mask=Mat(), bool compactResult=false ) const
+    { return wrapped->radiusMatch(queryDescriptors, trainDescriptors, matches, maxDistance, mask, compactResult); }
+
     CV_WRAP void match( const Mat& queryDescriptors, CV_OUT std::vector<DMatch>& matches,
-                const std::vector<Mat>& masks=std::vector<Mat>() );
+                const std::vector<Mat>& masks=std::vector<Mat>() )
+    { return wrapped->match(queryDescriptors, matches, masks); }
+
     CV_WRAP void knnMatch( const Mat& queryDescriptors, CV_OUT std::vector<std::vector<DMatch> >& matches, int k,
-           const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false );
+           const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false )
+    { return wrapped->knnMatch(queryDescriptors, matches, k, masks, compactResult); }
+
     CV_WRAP void radiusMatch( const Mat& queryDescriptors, CV_OUT std::vector<std::vector<DMatch> >& matches, float maxDistance,
-                   const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false );
-#endif
+                   const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false )
+    { return wrapped->radiusMatch(queryDescriptors, matches, maxDistance, masks, compactResult); }
 
     enum
     {
@@ -200,15 +224,13 @@ public:
 
     CV_WRAP_AS(clone) javaDescriptorMatcher* jclone( bool emptyTrainData=false ) const
     {
-        Ptr<DescriptorMatcher> matcher = this->clone(emptyTrainData);
-        matcher.addref();
-        return (javaDescriptorMatcher*)((DescriptorMatcher*) matcher);
+        return new javaDescriptorMatcher(wrapped->clone(emptyTrainData));
     }
 
     //supported: FlannBased, BruteForce, BruteForce-L1, BruteForce-Hamming, BruteForce-HammingLUT
     CV_WRAP static javaDescriptorMatcher* create( int matcherType )
     {
-        std::string name;
+        String name;
 
         switch(matcherType)
         {
@@ -231,42 +253,49 @@ public:
             name = "BruteForce-SL2";
             break;
         default:
-            CV_Error( CV_StsBadArg, "Specified descriptor matcher type is not supported." );
+            CV_Error( Error::StsBadArg, "Specified descriptor matcher type is not supported." );
             break;
         }
 
-        Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create(name);
-        matcher.addref();
-        return (javaDescriptorMatcher*)((DescriptorMatcher*) matcher);
+        return new javaDescriptorMatcher(DescriptorMatcher::create(name));
     }
 
-    CV_WRAP void write( const std::string& fileName ) const
+    CV_WRAP void write( const String& fileName ) const
     {
         FileStorage fs(fileName, FileStorage::WRITE);
-        ((DescriptorMatcher*)this)->write(fs);
-        fs.release();
+        wrapped->write(fs);
     }
 
-    CV_WRAP void read( const std::string& fileName )
+    CV_WRAP void read( const String& fileName )
     {
         FileStorage fs(fileName, FileStorage::READ);
-        ((DescriptorMatcher*)this)->read(fs.root());
-        fs.release();
+        wrapped->read(fs.root());
     }
+
+private:
+    javaDescriptorMatcher(Ptr<DescriptorMatcher> _wrapped) : wrapped(_wrapped)
+    {}
+
+    Ptr<DescriptorMatcher> wrapped;
 };
 
-class CV_EXPORTS_AS(DescriptorExtractor) javaDescriptorExtractor : public DescriptorExtractor
+class CV_EXPORTS_AS(DescriptorExtractor) javaDescriptorExtractor
 {
 public:
-#if 0
-    //DO NOT REMOVE! The block is required for sources parser
-    CV_WRAP void compute( const Mat& image, std::vector<KeyPoint>& keypoints, Mat& descriptors ) const;
-    CV_WRAP void compute( const std::vector<Mat>& images, std::vector<std::vector<KeyPoint> >& keypoints, CV_OUT std::vector<Mat>& descriptors ) const;
-    CV_WRAP virtual int descriptorSize() const;
-    CV_WRAP virtual int descriptorType() const;
+    CV_WRAP void compute( const Mat& image, CV_IN_OUT std::vector<KeyPoint>& keypoints, Mat& descriptors ) const
+    { return wrapped->compute(image, keypoints, descriptors); }
 
-    CV_WRAP virtual bool empty() const;
-#endif
+    CV_WRAP void compute( const std::vector<Mat>& images, CV_IN_OUT std::vector<std::vector<KeyPoint> >& keypoints, CV_OUT std::vector<Mat>& descriptors ) const
+    { return wrapped->compute(images, keypoints, descriptors); }
+
+    CV_WRAP int descriptorSize() const
+    { return wrapped->descriptorSize(); }
+
+    CV_WRAP int descriptorType() const
+    { return wrapped->descriptorType(); }
+
+    CV_WRAP bool empty() const
+    { return wrapped->empty(); }
 
     enum
     {
@@ -294,7 +323,7 @@ public:
     //not supported: Calonder
     CV_WRAP static javaDescriptorExtractor* create( int extractorType )
     {
-        std::string name;
+        String name;
 
         if (extractorType > OPPONENTEXTRACTOR)
         {
@@ -305,84 +334,115 @@ public:
         switch(extractorType)
         {
         case SIFT:
-            name += "SIFT";
+            name = name + "SIFT";
             break;
         case SURF:
-            name += "SURF";
+            name = name + "SURF";
             break;
         case ORB:
-            name += "ORB";
+            name = name + "ORB";
             break;
         case BRIEF:
-            name += "BRIEF";
+            name = name + "BRIEF";
             break;
         case BRISK:
-            name += "BRISK";
+            name = name + "BRISK";
             break;
         case FREAK:
-            name += "FREAK";
+            name = name + "FREAK";
             break;
         default:
-            CV_Error( CV_StsBadArg, "Specified descriptor extractor type is not supported." );
+            CV_Error( Error::StsBadArg, "Specified descriptor extractor type is not supported." );
             break;
         }
 
-        Ptr<DescriptorExtractor> extractor = DescriptorExtractor::create(name);
-        extractor.addref();
-        return (javaDescriptorExtractor*)((DescriptorExtractor*) extractor);
+        return new javaDescriptorExtractor(DescriptorExtractor::create(name));
     }
 
-    CV_WRAP void write( const std::string& fileName ) const
+    CV_WRAP void write( const String& fileName ) const
     {
         FileStorage fs(fileName, FileStorage::WRITE);
-        ((DescriptorExtractor*)this)->write(fs);
-        fs.release();
+        wrapped->write(fs);
     }
 
-    CV_WRAP void read( const std::string& fileName )
+    CV_WRAP void read( const String& fileName )
     {
         FileStorage fs(fileName, FileStorage::READ);
-        ((DescriptorExtractor*)this)->read(fs.root());
-        fs.release();
+        wrapped->read(fs.root());
     }
+
+private:
+    javaDescriptorExtractor(Ptr<DescriptorExtractor> _wrapped) : wrapped(_wrapped)
+    {}
+
+    Ptr<DescriptorExtractor> wrapped;
 };
 
-class CV_EXPORTS_AS(GenericDescriptorMatcher) javaGenericDescriptorMatcher : public GenericDescriptorMatcher
+class CV_EXPORTS_AS(GenericDescriptorMatcher) javaGenericDescriptorMatcher
 {
 public:
-#if 0
-    //DO NOT REMOVE! The block is required for sources parser
-    CV_WRAP virtual void add( const std::vector<Mat>& images,
-                      std::vector<std::vector<KeyPoint> >& keypoints );
-    CV_WRAP const std::vector<Mat>& getTrainImages() const;
-    CV_WRAP const std::vector<std::vector<KeyPoint> >& getTrainKeypoints() const;
-    CV_WRAP virtual void clear();
-    CV_WRAP virtual bool isMaskSupported();
-    CV_WRAP virtual void train();
+    CV_WRAP void add( const std::vector<Mat>& images,
+                      std::vector<std::vector<KeyPoint> >& keypoints )
+    { return wrapped->add(images, keypoints); }
+
+    CV_WRAP const std::vector<Mat>& getTrainImages() const
+    { return wrapped->getTrainImages(); }
+
+    CV_WRAP const std::vector<std::vector<KeyPoint> >& getTrainKeypoints() const
+    { return wrapped->getTrainKeypoints(); }
+
+    CV_WRAP void clear()
+    { return wrapped->clear(); }
+
+    CV_WRAP bool isMaskSupported()
+    { return wrapped->isMaskSupported(); }
+
+    CV_WRAP void train()
+    { return wrapped->train(); }
+
     CV_WRAP void classify( const Mat& queryImage, CV_IN_OUT std::vector<KeyPoint>& queryKeypoints,
-                           const Mat& trainImage, std::vector<KeyPoint>& trainKeypoints ) const;
-    CV_WRAP void classify( const Mat& queryImage, CV_IN_OUT std::vector<KeyPoint>& queryKeypoints );
+                           const Mat& trainImage, std::vector<KeyPoint>& trainKeypoints ) const
+    { return wrapped->classify(queryImage, queryKeypoints, trainImage, trainKeypoints); }
+
+    CV_WRAP void classify( const Mat& queryImage, CV_IN_OUT std::vector<KeyPoint>& queryKeypoints )
+    { return wrapped->classify(queryImage, queryKeypoints); }
+
     CV_WRAP void match( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
                 const Mat& trainImage, std::vector<KeyPoint>& trainKeypoints,
-                CV_OUT std::vector<DMatch>& matches, const Mat& mask=Mat() ) const;
+                CV_OUT std::vector<DMatch>& matches, const Mat& mask=Mat() ) const
+    { return wrapped->match(queryImage, queryKeypoints, trainImage, trainKeypoints, matches, mask); }
+
     CV_WRAP void knnMatch( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
                    const Mat& trainImage, std::vector<KeyPoint>& trainKeypoints,
                    CV_OUT std::vector<std::vector<DMatch> >& matches, int k,
-                   const Mat& mask=Mat(), bool compactResult=false ) const;
+                   const Mat& mask=Mat(), bool compactResult=false ) const
+    { return wrapped->knnMatch(queryImage, queryKeypoints, trainImage, trainKeypoints,
+                               matches, k, mask, compactResult); }
+
     CV_WRAP void radiusMatch( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
                       const Mat& trainImage, std::vector<KeyPoint>& trainKeypoints,
                       CV_OUT std::vector<std::vector<DMatch> >& matches, float maxDistance,
-                      const Mat& mask=Mat(), bool compactResult=false ) const;
+                      const Mat& mask=Mat(), bool compactResult=false ) const
+    { return wrapped->radiusMatch(queryImage, queryKeypoints, trainImage, trainKeypoints,
+                                   matches, maxDistance, mask, compactResult); }
+
     CV_WRAP void match( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
-                CV_OUT std::vector<DMatch>& matches, const std::vector<Mat>& masks=std::vector<Mat>() );
+                CV_OUT std::vector<DMatch>& matches, const std::vector<Mat>& masks=std::vector<Mat>() )
+    { return wrapped->match(queryImage, queryKeypoints, matches, masks); }
+
     CV_WRAP void knnMatch( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
                    CV_OUT std::vector<std::vector<DMatch> >& matches, int k,
-                   const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false );
+                   const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false )
+    { return wrapped->knnMatch(queryImage, queryKeypoints, matches, k, masks, compactResult); }
+
     CV_WRAP void radiusMatch( const Mat& queryImage, std::vector<KeyPoint>& queryKeypoints,
                       CV_OUT std::vector<std::vector<DMatch> >& matches, float maxDistance,
-                      const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false );
-    CV_WRAP virtual bool empty() const;
-#endif
+                      const std::vector<Mat>& masks=std::vector<Mat>(), bool compactResult=false )
+    { return wrapped->radiusMatch(queryImage, queryKeypoints, matches, maxDistance, masks, compactResult); }
+
+    CV_WRAP bool empty() const
+    { return wrapped->empty(); }
+
 
     enum
     {
@@ -392,16 +452,14 @@ public:
 
     CV_WRAP_AS(clone) javaGenericDescriptorMatcher* jclone( bool emptyTrainData=false ) const
     {
-        Ptr<GenericDescriptorMatcher> matcher = this->clone(emptyTrainData);
-        matcher.addref();
-        return (javaGenericDescriptorMatcher*)((GenericDescriptorMatcher*) matcher);
+        return new javaGenericDescriptorMatcher(wrapped->clone(emptyTrainData));
     }
 
     //supported: OneWay, Fern
     //unsupported: Vector
     CV_WRAP static javaGenericDescriptorMatcher* create( int matcherType )
     {
-        std::string name;
+        String name;
 
         switch(matcherType)
         {
@@ -412,28 +470,30 @@ public:
             name = "FERN";
             break;
         default:
-            CV_Error( CV_StsBadArg, "Specified generic descriptor matcher type is not supported." );
+            CV_Error( Error::StsBadArg, "Specified generic descriptor matcher type is not supported." );
             break;
         }
 
-        Ptr<GenericDescriptorMatcher> matcher = GenericDescriptorMatcher::create(name);
-        matcher.addref();
-        return (javaGenericDescriptorMatcher*)((GenericDescriptorMatcher*) matcher);
+        return new javaGenericDescriptorMatcher(GenericDescriptorMatcher::create(name));
     }
 
-    CV_WRAP void write( const std::string& fileName ) const
+    CV_WRAP void write( const String& fileName ) const
     {
         FileStorage fs(fileName, FileStorage::WRITE);
-        ((GenericDescriptorMatcher*)this)->write(fs);
-        fs.release();
+        wrapped->write(fs);
     }
 
-    CV_WRAP void read( const std::string& fileName )
+    CV_WRAP void read( const String& fileName )
     {
         FileStorage fs(fileName, FileStorage::READ);
-        ((GenericDescriptorMatcher*)this)->read(fs.root());
-        fs.release();
+        wrapped->read(fs.root());
     }
+
+private:
+    javaGenericDescriptorMatcher(Ptr<GenericDescriptorMatcher> _wrapped) : wrapped(_wrapped)
+    {}
+
+    Ptr<GenericDescriptorMatcher> wrapped;
 };
 
 #if 0
diff --git a/modules/java/generator/src/cpp/jni_part.cpp b/modules/java/generator/src/cpp/jni_part.cpp
index 8f6eb81618..ccd870cdf9 100644
--- a/modules/java/generator/src/cpp/jni_part.cpp
+++ b/modules/java/generator/src/cpp/jni_part.cpp
@@ -63,4 +63,4 @@ JNI_OnUnload(JavaVM*, void*)
   //do nothing
 }
 
-} // extern "C"
\ No newline at end of file
+} // extern "C"
diff --git a/modules/java/generator/src/cpp/utils.cpp b/modules/java/generator/src/cpp/utils.cpp
index ed964e377e..40811e8f9e 100644
--- a/modules/java/generator/src/cpp/utils.cpp
+++ b/modules/java/generator/src/cpp/utils.cpp
@@ -44,7 +44,7 @@ JNIEXPORT void JNICALL Java_org_opencv_android_Utils_nBitmapToMat2
                 // info.format == ANDROID_BITMAP_FORMAT_RGB_565
                 LOGD("nBitmapToMat: RGB_565 -> CV_8UC4");
                 Mat tmp(info.height, info.width, CV_8UC2, pixels);
-                cvtColor(tmp, dst, CV_BGR5652RGBA);
+                cvtColor(tmp, dst, COLOR_BGR5652RGBA);
             }
             AndroidBitmap_unlockPixels(env, bitmap);
             return;
@@ -104,10 +104,10 @@ JNIEXPORT void JNICALL Java_org_opencv_android_Utils_nMatToBitmap2
                 if(src.type() == CV_8UC1)
                 {
                     LOGD("nMatToBitmap: CV_8UC1 -> RGBA_8888");
-                    cvtColor(src, tmp, CV_GRAY2RGBA);
+                    cvtColor(src, tmp, COLOR_GRAY2RGBA);
                 } else if(src.type() == CV_8UC3){
                     LOGD("nMatToBitmap: CV_8UC3 -> RGBA_8888");
-                    cvtColor(src, tmp, CV_RGB2RGBA);
+                    cvtColor(src, tmp, COLOR_RGB2RGBA);
                 } else if(src.type() == CV_8UC4){
                     LOGD("nMatToBitmap: CV_8UC4 -> RGBA_8888");
                     if(needPremultiplyAlpha) cvtColor(src, tmp, COLOR_RGBA2mRGBA);
@@ -119,13 +119,13 @@ JNIEXPORT void JNICALL Java_org_opencv_android_Utils_nMatToBitmap2
                 if(src.type() == CV_8UC1)
                 {
                     LOGD("nMatToBitmap: CV_8UC1 -> RGB_565");
-                    cvtColor(src, tmp, CV_GRAY2BGR565);
+                    cvtColor(src, tmp, COLOR_GRAY2BGR565);
                 } else if(src.type() == CV_8UC3){
                     LOGD("nMatToBitmap: CV_8UC3 -> RGB_565");
-                    cvtColor(src, tmp, CV_RGB2BGR565);
+                    cvtColor(src, tmp, COLOR_RGB2BGR565);
                 } else if(src.type() == CV_8UC4){
                     LOGD("nMatToBitmap: CV_8UC4 -> RGB_565");
-                    cvtColor(src, tmp, CV_RGBA2BGR565);
+                    cvtColor(src, tmp, COLOR_RGBA2BGR565);
                 }
             }
             AndroidBitmap_unlockPixels(env, bitmap);
@@ -158,4 +158,4 @@ JNIEXPORT void JNICALL Java_org_opencv_android_Utils_nMatToBitmap
 
 } // extern "C"
 
-#endif //__ANDROID__
\ No newline at end of file
+#endif //__ANDROID__
diff --git a/modules/java/generator/src/java/android+CameraBridgeViewBase.java b/modules/java/generator/src/java/android+CameraBridgeViewBase.java
index 4dd3af96f1..c0c9f5bde7 100644
--- a/modules/java/generator/src/java/android+CameraBridgeViewBase.java
+++ b/modules/java/generator/src/java/android+CameraBridgeViewBase.java
@@ -15,6 +15,7 @@ import android.content.DialogInterface;
 import android.content.res.TypedArray;
 import android.graphics.Bitmap;
 import android.graphics.Canvas;
+import android.graphics.Rect;
 import android.util.AttributeSet;
 import android.util.Log;
 import android.view.SurfaceHolder;
@@ -44,14 +45,22 @@ public abstract class CameraBridgeViewBase extends SurfaceView implements Surfac
     protected int mFrameHeight;
     protected int mMaxHeight;
     protected int mMaxWidth;
+    protected float mScale = 0;
     protected int mPreviewFormat = Highgui.CV_CAP_ANDROID_COLOR_FRAME_RGBA;
-    protected int mCameraIndex = -1;
+    protected int mCameraIndex = CAMERA_ID_ANY;
     protected boolean mEnabled;
     protected FpsMeter mFpsMeter = null;
 
+    public static final int CAMERA_ID_ANY   = -1;
+    public static final int CAMERA_ID_BACK  = 99;
+    public static final int CAMERA_ID_FRONT = 98;
+
     public CameraBridgeViewBase(Context context, int cameraId) {
         super(context);
         mCameraIndex = cameraId;
+        getHolder().addCallback(this);
+        mMaxWidth = MAX_UNSPECIFIED;
+        mMaxHeight = MAX_UNSPECIFIED;
     }
 
     public CameraBridgeViewBase(Context context, AttributeSet attrs) {
@@ -69,6 +78,15 @@ public abstract class CameraBridgeViewBase extends SurfaceView implements Surfac
         getHolder().addCallback(this);
         mMaxWidth = MAX_UNSPECIFIED;
         mMaxHeight = MAX_UNSPECIFIED;
+        styledAttrs.recycle();
+    }
+
+    /**
+     * Sets the camera index
+     * @param cameraIndex new camera index
+     */
+    public void setCameraIndex(int cameraIndex) {
+        this.mCameraIndex = cameraIndex;
     }
 
     public interface CvCameraViewListener {
@@ -150,8 +168,6 @@ public abstract class CameraBridgeViewBase extends SurfaceView implements Surfac
             mPreviewFormat = format;
         }
 
-        private CvCameraViewListenerAdapter() {}
-
         private int mPreviewFormat = Highgui.CV_CAP_ANDROID_COLOR_FRAME_RGBA;
         private CvCameraViewListener mOldStyleListener;
     };
@@ -389,7 +405,22 @@ public abstract class CameraBridgeViewBase extends SurfaceView implements Surfac
             Canvas canvas = getHolder().lockCanvas();
             if (canvas != null) {
                 canvas.drawColor(0, android.graphics.PorterDuff.Mode.CLEAR);
-                canvas.drawBitmap(mCacheBitmap, (canvas.getWidth() - mCacheBitmap.getWidth()) / 2, (canvas.getHeight() - mCacheBitmap.getHeight()) / 2, null);
+                Log.d(TAG, "mStretch value: " + mScale);
+
+                if (mScale != 0) {
+                    canvas.drawBitmap(mCacheBitmap, new Rect(0,0,mCacheBitmap.getWidth(), mCacheBitmap.getHeight()),
+                         new Rect((int)((canvas.getWidth() - mScale*mCacheBitmap.getWidth()) / 2),
+                         (int)((canvas.getHeight() - mScale*mCacheBitmap.getHeight()) / 2),
+                         (int)((canvas.getWidth() - mScale*mCacheBitmap.getWidth()) / 2 + mScale*mCacheBitmap.getWidth()),
+                         (int)((canvas.getHeight() - mScale*mCacheBitmap.getHeight()) / 2 + mScale*mCacheBitmap.getHeight())), null);
+                } else {
+                     canvas.drawBitmap(mCacheBitmap, new Rect(0,0,mCacheBitmap.getWidth(), mCacheBitmap.getHeight()),
+                         new Rect((canvas.getWidth() - mCacheBitmap.getWidth()) / 2,
+                         (canvas.getHeight() - mCacheBitmap.getHeight()) / 2,
+                         (canvas.getWidth() - mCacheBitmap.getWidth()) / 2 + mCacheBitmap.getWidth(),
+                         (canvas.getHeight() - mCacheBitmap.getHeight()) / 2 + mCacheBitmap.getHeight()), null);
+                }
+
                 if (mFpsMeter != null) {
                     mFpsMeter.measure();
                     mFpsMeter.draw(canvas, 20, 30);
diff --git a/modules/java/generator/src/java/android+JavaCameraView.java b/modules/java/generator/src/java/android+JavaCameraView.java
index 962899ad8f..f864e5370b 100644
--- a/modules/java/generator/src/java/android+JavaCameraView.java
+++ b/modules/java/generator/src/java/android+JavaCameraView.java
@@ -10,6 +10,7 @@ import android.hardware.Camera.PreviewCallback;
 import android.os.Build;
 import android.util.AttributeSet;
 import android.util.Log;
+import android.view.ViewGroup.LayoutParams;
 
 import org.opencv.core.CvType;
 import org.opencv.core.Mat;
@@ -37,7 +38,7 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
     private boolean mStopThread;
 
     protected Camera mCamera;
-    protected JavaCameraFrame mCameraFrame;
+    protected JavaCameraFrame[] mCameraFrame;
     private SurfaceTexture mSurfaceTexture;
 
     public static class JavaCameraSizeAccessor implements ListItemAccessor {
@@ -59,7 +60,6 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
 
     public JavaCameraView(Context context, AttributeSet attrs) {
         super(context, attrs);
-        Log.d(TAG, "Java camera view ctor");
     }
 
     protected boolean initializeCamera(int width, int height) {
@@ -68,7 +68,7 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
         synchronized (this) {
             mCamera = null;
 
-            if (mCameraIndex == -1) {
+            if (mCameraIndex == CAMERA_ID_ANY) {
                 Log.d(TAG, "Trying to open camera with old open()");
                 try {
                     mCamera = Camera.open();
@@ -92,11 +92,39 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
                 }
             } else {
                 if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.GINGERBREAD) {
-                    Log.d(TAG, "Trying to open camera with new open(" + Integer.valueOf(mCameraIndex) + ")");
-                    try {
-                        mCamera = Camera.open(mCameraIndex);
-                    } catch (RuntimeException e) {
-                        Log.e(TAG, "Camera #" + mCameraIndex + "failed to open: " + e.getLocalizedMessage());
+                    int localCameraIndex = mCameraIndex;
+                    if (mCameraIndex == CAMERA_ID_BACK) {
+                        Log.i(TAG, "Trying to open back camera");
+                        Camera.CameraInfo cameraInfo = new Camera.CameraInfo();
+                        for (int camIdx = 0; camIdx < Camera.getNumberOfCameras(); ++camIdx) {
+                            Camera.getCameraInfo( camIdx, cameraInfo );
+                            if (cameraInfo.facing == Camera.CameraInfo.CAMERA_FACING_BACK) {
+                                localCameraIndex = camIdx;
+                                break;
+                            }
+                        }
+                    } else if (mCameraIndex == CAMERA_ID_FRONT) {
+                        Log.i(TAG, "Trying to open front camera");
+                        Camera.CameraInfo cameraInfo = new Camera.CameraInfo();
+                        for (int camIdx = 0; camIdx < Camera.getNumberOfCameras(); ++camIdx) {
+                            Camera.getCameraInfo( camIdx, cameraInfo );
+                            if (cameraInfo.facing == Camera.CameraInfo.CAMERA_FACING_FRONT) {
+                                localCameraIndex = camIdx;
+                                break;
+                            }
+                        }
+                    }
+                    if (localCameraIndex == CAMERA_ID_BACK) {
+                        Log.e(TAG, "Back camera not found!");
+                    } else if (localCameraIndex == CAMERA_ID_FRONT) {
+                        Log.e(TAG, "Front camera not found!");
+                    } else {
+                        Log.d(TAG, "Trying to open camera with new open(" + Integer.valueOf(localCameraIndex) + ")");
+                        try {
+                            mCamera = Camera.open(localCameraIndex);
+                        } catch (RuntimeException e) {
+                            Log.e(TAG, "Camera #" + localCameraIndex + "failed to open: " + e.getLocalizedMessage());
+                        }
                     }
                 }
             }
@@ -130,6 +158,11 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
                     mFrameWidth = params.getPreviewSize().width;
                     mFrameHeight = params.getPreviewSize().height;
 
+                    if ((getLayoutParams().width == LayoutParams.MATCH_PARENT) && (getLayoutParams().height == LayoutParams.MATCH_PARENT))
+                        mScale = Math.min(((float)height)/mFrameHeight, ((float)width)/mFrameWidth);
+                    else
+                        mScale = 0;
+
                     if (mFpsMeter != null) {
                         mFpsMeter.setResolution(mFrameWidth, mFrameHeight);
                     }
@@ -147,7 +180,9 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
 
                     AllocateCache();
 
-                    mCameraFrame = new JavaCameraFrame(mFrameChain[mChainIdx], mFrameWidth, mFrameHeight);
+                    mCameraFrame = new JavaCameraFrame[2];
+                    mCameraFrame[0] = new JavaCameraFrame(mFrameChain[0], mFrameWidth, mFrameHeight);
+                    mCameraFrame[1] = new JavaCameraFrame(mFrameChain[1], mFrameWidth, mFrameHeight);
 
                     if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.HONEYCOMB) {
                         mSurfaceTexture = new SurfaceTexture(MAGIC_TEXTURE_ID);
@@ -174,6 +209,8 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
         synchronized (this) {
             if (mCamera != null) {
                 mCamera.stopPreview();
+                mCamera.setPreviewCallback(null);
+
                 mCamera.release();
             }
             mCamera = null;
@@ -181,8 +218,10 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
                 mFrameChain[0].release();
                 mFrameChain[1].release();
             }
-            if (mCameraFrame != null)
-                mCameraFrame.release();
+            if (mCameraFrame != null) {
+                mCameraFrame[0].release();
+                mCameraFrame[1].release();
+            }
         }
     }
 
@@ -231,10 +270,8 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
     }
 
     public void onPreviewFrame(byte[] frame, Camera arg1) {
-        Log.i(TAG, "Preview Frame received. Need to create MAT and deliver it to clients");
-        Log.i(TAG, "Frame size  is " + frame.length);
-        synchronized (this)
-        {
+        Log.d(TAG, "Preview Frame received. Frame size: " + frame.length);
+        synchronized (this) {
             mFrameChain[1 - mChainIdx].put(0, 0, frame);
             this.notify();
         }
@@ -242,8 +279,7 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
             mCamera.addCallbackBuffer(mBuffer);
     }
 
-    private class JavaCameraFrame implements CvCameraViewFrame
-    {
+    private class JavaCameraFrame implements CvCameraViewFrame {
         public Mat gray() {
             return mYuvFrameData.submat(0, mHeight, 0, mWidth);
         }
@@ -265,9 +301,6 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
             mRgba.release();
         }
 
-        private JavaCameraFrame(CvCameraViewFrame obj) {
-        }
-
         private Mat mYuvFrameData;
         private Mat mRgba;
         private int mWidth;
@@ -289,7 +322,7 @@ public class JavaCameraView extends CameraBridgeViewBase implements PreviewCallb
 
                 if (!mStopThread) {
                     if (!mFrameChain[mChainIdx].empty())
-                        deliverAndDrawFrame(mCameraFrame);
+                        deliverAndDrawFrame(mCameraFrame[mChainIdx]);
                     mChainIdx = 1 - mChainIdx;
                 }
             } while (!mStopThread);
diff --git a/modules/java/generator/src/java/android+NativeCameraView.java b/modules/java/generator/src/java/android+NativeCameraView.java
index 3802167b35..62d0775801 100644
--- a/modules/java/generator/src/java/android+NativeCameraView.java
+++ b/modules/java/generator/src/java/android+NativeCameraView.java
@@ -8,6 +8,7 @@ import org.opencv.highgui.VideoCapture;
 import android.content.Context;
 import android.util.AttributeSet;
 import android.util.Log;
+import android.view.ViewGroup.LayoutParams;
 
 /**
  * This class is an implementation of a bridge between SurfaceView and native OpenCV camera.
@@ -52,14 +53,16 @@ public class NativeCameraView extends CameraBridgeViewBase {
         /* 1. We need to stop thread which updating the frames
          * 2. Stop camera and release it
          */
-        try {
-            mStopThread = true;
-            mThread.join();
-        } catch (InterruptedException e) {
-            e.printStackTrace();
-        } finally {
-            mThread =  null;
-            mStopThread = false;
+        if (mThread != null) {
+            try {
+                mStopThread = true;
+                mThread.join();
+            } catch (InterruptedException e) {
+                e.printStackTrace();
+            } finally {
+                mThread =  null;
+                mStopThread = false;
+            }
         }
 
         /* Now release camera */
@@ -102,6 +105,11 @@ public class NativeCameraView extends CameraBridgeViewBase {
             mFrameWidth = (int)frameSize.width;
             mFrameHeight = (int)frameSize.height;
 
+            if ((getLayoutParams().width == LayoutParams.MATCH_PARENT) && (getLayoutParams().height == LayoutParams.MATCH_PARENT))
+                mScale = Math.min(((float)height)/mFrameHeight, ((float)width)/mFrameWidth);
+            else
+                mScale = 0;
+
             if (mFpsMeter != null) {
                 mFpsMeter.setResolution(mFrameWidth, mFrameHeight);
             }
@@ -125,17 +133,17 @@ public class NativeCameraView extends CameraBridgeViewBase {
         }
     }
 
-    private class NativeCameraFrame implements CvCameraViewFrame {
+    private static class NativeCameraFrame implements CvCameraViewFrame {
 
         @Override
         public Mat rgba() {
-            mCamera.retrieve(mRgba, Highgui.CV_CAP_ANDROID_COLOR_FRAME_RGBA);
+            mCapture.retrieve(mRgba, Highgui.CV_CAP_ANDROID_COLOR_FRAME_RGBA);
             return mRgba;
         }
 
         @Override
         public Mat gray() {
-            mCamera.retrieve(mGray, Highgui.CV_CAP_ANDROID_GREY_FRAME);
+            mCapture.retrieve(mGray, Highgui.CV_CAP_ANDROID_GREY_FRAME);
             return mGray;
         }
 
@@ -152,9 +160,6 @@ public class NativeCameraView extends CameraBridgeViewBase {
 
     private class CameraWorker implements Runnable {
 
-        private Mat mRgba = new Mat();
-        private Mat mGray = new Mat();
-
         public void run() {
             do {
                 if (!mCamera.grab()) {
diff --git a/modules/java/generator/src/java/android+OpenCVLoader.java b/modules/java/generator/src/java/android+OpenCVLoader.java
index 70e94944dd..a76471eac9 100644
--- a/modules/java/generator/src/java/android+OpenCVLoader.java
+++ b/modules/java/generator/src/java/android+OpenCVLoader.java
@@ -22,6 +22,17 @@ public class OpenCVLoader
      */
     public static final String OPENCV_VERSION_2_4_4 = "2.4.4";
 
+    /**
+     * OpenCV Library version 2.4.5.
+     */
+    public static final String OPENCV_VERSION_2_4_5 = "2.4.5";
+
+    /**
+     * OpenCV Library version 2.4.6.
+     */
+    public static final String OPENCV_VERSION_2_4_6 = "2.4.6";
+
+
     /**
      * Loads and initializes OpenCV library from current application package. Roughly, it's an analog of system.loadLibrary("opencv_java").
      * @return Returns true is initialization of OpenCV was successful.
diff --git a/modules/java/generator/src/java/android+sync.py b/modules/java/generator/src/java/android+sync.py
index baf95cb6e1..ed78537f08 100755
--- a/modules/java/generator/src/java/android+sync.py
+++ b/modules/java/generator/src/java/android+sync.py
@@ -4,4 +4,4 @@ import os
 import shutil
 
 for f in os.listdir("."):
-    shutil.copyfile(f, os.path.join("../../../../../../modules/java/generator/src/java/", "android+" + f));
\ No newline at end of file
+    shutil.copyfile(f, os.path.join("../../../../../../modules/java/generator/src/java/", "android+" + f));
diff --git a/modules/java/generator/src/java/features2d+DMatch.java b/modules/java/generator/src/java/core+DMatch.java
similarity index 97%
rename from modules/java/generator/src/java/features2d+DMatch.java
rename to modules/java/generator/src/java/core+DMatch.java
index 4b1615ac5b..12bd86ec01 100644
--- a/modules/java/generator/src/java/features2d+DMatch.java
+++ b/modules/java/generator/src/java/core+DMatch.java
@@ -1,4 +1,4 @@
-package org.opencv.features2d;
+package org.opencv.core;
 
 //C++: class DMatch
 
diff --git a/modules/java/generator/src/java/features2d+KeyPoint.java b/modules/java/generator/src/java/core+KeyPoint.java
similarity index 98%
rename from modules/java/generator/src/java/features2d+KeyPoint.java
rename to modules/java/generator/src/java/core+KeyPoint.java
index ac65097d8b..de5b2151ff 100644
--- a/modules/java/generator/src/java/features2d+KeyPoint.java
+++ b/modules/java/generator/src/java/core+KeyPoint.java
@@ -1,4 +1,4 @@
-package org.opencv.features2d;
+package org.opencv.core;
 
 import org.opencv.core.Point;
 
diff --git a/modules/java/generator/src/java/core+Mat.java b/modules/java/generator/src/java/core+Mat.java
index 8dd72fb681..6db255417c 100644
--- a/modules/java/generator/src/java/core+Mat.java
+++ b/modules/java/generator/src/java/core+Mat.java
@@ -245,6 +245,19 @@ public class Mat {
         return retVal;
     }
 
+    //
+    // C++: int Mat::dims()
+    //
+
+    // javadoc: Mat::dims()
+    public int dims()
+    {
+
+        int retVal = n_dims(nativeObj);
+
+        return retVal;
+    }
+
     //
     // C++: int Mat::cols()
     //
@@ -1130,6 +1143,9 @@ public class Mat {
     // C++: Mat Mat::colRange(int startcol, int endcol)
     private static native long n_colRange(long nativeObj, int startcol, int endcol);
 
+    // C++: int Mat::dims()
+    private static native int n_dims(long nativeObj);
+
     // C++: int Mat::cols()
     private static native int n_cols(long nativeObj);
 
diff --git a/modules/java/generator/src/java/core+MatOfByte.java b/modules/java/generator/src/java/core+MatOfByte.java
index 0ebdb66733..7756eb94f9 100644
--- a/modules/java/generator/src/java/core+MatOfByte.java
+++ b/modules/java/generator/src/java/core+MatOfByte.java
@@ -14,8 +14,8 @@ public class MatOfByte extends Mat {
 
     protected MatOfByte(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfByte extends Mat {
 
     public MatOfByte(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfDMatch.java b/modules/java/generator/src/java/core+MatOfDMatch.java
index 7a3094933c..2c99e14d1b 100644
--- a/modules/java/generator/src/java/core+MatOfDMatch.java
+++ b/modules/java/generator/src/java/core+MatOfDMatch.java
@@ -3,7 +3,7 @@ package org.opencv.core;
 import java.util.Arrays;
 import java.util.List;
 
-import org.opencv.features2d.DMatch;
+import org.opencv.core.DMatch;
 
 public class MatOfDMatch extends Mat {
     // 32FC4
@@ -16,8 +16,8 @@ public class MatOfDMatch extends Mat {
 
     protected MatOfDMatch(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat: " + toString());
         //FIXME: do we need release() here?
     }
 
@@ -27,8 +27,8 @@ public class MatOfDMatch extends Mat {
 
     public MatOfDMatch(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat: " + toString());
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfDouble.java b/modules/java/generator/src/java/core+MatOfDouble.java
index cca5251105..1a8e23ca30 100644
--- a/modules/java/generator/src/java/core+MatOfDouble.java
+++ b/modules/java/generator/src/java/core+MatOfDouble.java
@@ -14,8 +14,8 @@ public class MatOfDouble extends Mat {
 
     protected MatOfDouble(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfDouble extends Mat {
 
     public MatOfDouble(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfFloat.java b/modules/java/generator/src/java/core+MatOfFloat.java
index ce73b6f638..dfd6b84048 100644
--- a/modules/java/generator/src/java/core+MatOfFloat.java
+++ b/modules/java/generator/src/java/core+MatOfFloat.java
@@ -14,8 +14,8 @@ public class MatOfFloat extends Mat {
 
     protected MatOfFloat(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfFloat extends Mat {
 
     public MatOfFloat(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfFloat4.java b/modules/java/generator/src/java/core+MatOfFloat4.java
index 8a3e51014f..96f9e5cde4 100644
--- a/modules/java/generator/src/java/core+MatOfFloat4.java
+++ b/modules/java/generator/src/java/core+MatOfFloat4.java
@@ -14,8 +14,8 @@ public class MatOfFloat4 extends Mat {
 
     protected MatOfFloat4(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfFloat4 extends Mat {
 
     public MatOfFloat4(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfFloat6.java b/modules/java/generator/src/java/core+MatOfFloat6.java
index 1e23101a72..140bb57a9b 100644
--- a/modules/java/generator/src/java/core+MatOfFloat6.java
+++ b/modules/java/generator/src/java/core+MatOfFloat6.java
@@ -14,8 +14,8 @@ public class MatOfFloat6 extends Mat {
 
     protected MatOfFloat6(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfFloat6 extends Mat {
 
     public MatOfFloat6(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfInt.java b/modules/java/generator/src/java/core+MatOfInt.java
index 80c5b3a5c2..7d076878fc 100644
--- a/modules/java/generator/src/java/core+MatOfInt.java
+++ b/modules/java/generator/src/java/core+MatOfInt.java
@@ -15,8 +15,8 @@ public class MatOfInt extends Mat {
 
     protected MatOfInt(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -26,8 +26,8 @@ public class MatOfInt extends Mat {
 
     public MatOfInt(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfInt4.java b/modules/java/generator/src/java/core+MatOfInt4.java
index 60277103cc..4ca4d9e318 100644
--- a/modules/java/generator/src/java/core+MatOfInt4.java
+++ b/modules/java/generator/src/java/core+MatOfInt4.java
@@ -15,8 +15,8 @@ public class MatOfInt4 extends Mat {
 
     protected MatOfInt4(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -26,8 +26,8 @@ public class MatOfInt4 extends Mat {
 
     public MatOfInt4(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfKeyPoint.java b/modules/java/generator/src/java/core+MatOfKeyPoint.java
index b91fedcee8..24b9a817f2 100644
--- a/modules/java/generator/src/java/core+MatOfKeyPoint.java
+++ b/modules/java/generator/src/java/core+MatOfKeyPoint.java
@@ -3,7 +3,7 @@ package org.opencv.core;
 import java.util.Arrays;
 import java.util.List;
 
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.KeyPoint;
 
 public class MatOfKeyPoint extends Mat {
     // 32FC7
@@ -16,8 +16,8 @@ public class MatOfKeyPoint extends Mat {
 
     protected MatOfKeyPoint(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -27,8 +27,8 @@ public class MatOfKeyPoint extends Mat {
 
     public MatOfKeyPoint(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfPoint.java b/modules/java/generator/src/java/core+MatOfPoint.java
index 23eeed0ebb..f4d573bb73 100644
--- a/modules/java/generator/src/java/core+MatOfPoint.java
+++ b/modules/java/generator/src/java/core+MatOfPoint.java
@@ -14,8 +14,8 @@ public class MatOfPoint extends Mat {
 
     protected MatOfPoint(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfPoint extends Mat {
 
     public MatOfPoint(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfPoint2f.java b/modules/java/generator/src/java/core+MatOfPoint2f.java
index ba4be4ac5e..4b8c926fff 100644
--- a/modules/java/generator/src/java/core+MatOfPoint2f.java
+++ b/modules/java/generator/src/java/core+MatOfPoint2f.java
@@ -14,8 +14,8 @@ public class MatOfPoint2f extends Mat {
 
     protected MatOfPoint2f(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfPoint2f extends Mat {
 
     public MatOfPoint2f(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfPoint3.java b/modules/java/generator/src/java/core+MatOfPoint3.java
index 16e21301ef..3b50561e91 100644
--- a/modules/java/generator/src/java/core+MatOfPoint3.java
+++ b/modules/java/generator/src/java/core+MatOfPoint3.java
@@ -14,8 +14,8 @@ public class MatOfPoint3 extends Mat {
 
     protected MatOfPoint3(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfPoint3 extends Mat {
 
     public MatOfPoint3(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfPoint3f.java b/modules/java/generator/src/java/core+MatOfPoint3f.java
index 97e2a95702..fc5fee436d 100644
--- a/modules/java/generator/src/java/core+MatOfPoint3f.java
+++ b/modules/java/generator/src/java/core+MatOfPoint3f.java
@@ -14,8 +14,8 @@ public class MatOfPoint3f extends Mat {
 
     protected MatOfPoint3f(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -25,8 +25,8 @@ public class MatOfPoint3f extends Mat {
 
     public MatOfPoint3f(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/core+MatOfRect.java b/modules/java/generator/src/java/core+MatOfRect.java
index 2e58bfe897..ec0fb01e4a 100644
--- a/modules/java/generator/src/java/core+MatOfRect.java
+++ b/modules/java/generator/src/java/core+MatOfRect.java
@@ -15,8 +15,8 @@ public class MatOfRect extends Mat {
 
     protected MatOfRect(long addr) {
         super(addr);
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
@@ -26,8 +26,8 @@ public class MatOfRect extends Mat {
 
     public MatOfRect(Mat m) {
         super(m, Range.all());
-        if(checkVector(_channels, _depth) < 0 )
-            throw new IllegalArgumentException("Incomatible Mat");
+        if( !empty() && checkVector(_channels, _depth) < 0 )
+            throw new IllegalArgumentException("Incompatible Mat");
         //FIXME: do we need release() here?
     }
 
diff --git a/modules/java/generator/src/java/engine+OpenCVEngineInterface.aidl b/modules/java/generator/src/java/engine+OpenCVEngineInterface.aidl
index 7a47a90732..21fe5f716b 100644
--- a/modules/java/generator/src/java/engine+OpenCVEngineInterface.aidl
+++ b/modules/java/generator/src/java/engine+OpenCVEngineInterface.aidl
@@ -30,4 +30,4 @@ interface OpenCVEngineInterface
     * @return Returns names of OpenCV libraries, separated by semicolon.
     */
     String getLibraryList(String version);
-}
\ No newline at end of file
+}
diff --git a/modules/java/generator/src/java/highgui+VideoCapture.java b/modules/java/generator/src/java/highgui+VideoCapture.java
deleted file mode 100644
index b8569bb9c0..0000000000
--- a/modules/java/generator/src/java/highgui+VideoCapture.java
+++ /dev/null
@@ -1,240 +0,0 @@
-package org.opencv.highgui;
-
-import java.util.List;
-import java.util.LinkedList;
-
-import org.opencv.core.Mat;
-import org.opencv.core.Size;
-
-// C++: class VideoCapture
-//javadoc: VideoCapture
-public class VideoCapture {
-
-    protected final long nativeObj;
-
-    protected VideoCapture(long addr) {
-        nativeObj = addr;
-    }
-
-    //
-    // C++: VideoCapture::VideoCapture()
-    //
-
-    // javadoc: VideoCapture::VideoCapture()
-    public VideoCapture()
-    {
-
-        nativeObj = n_VideoCapture();
-
-        return;
-    }
-
-    //
-    // C++: VideoCapture::VideoCapture(int device)
-    //
-
-    // javadoc: VideoCapture::VideoCapture(device)
-    public VideoCapture(int device)
-    {
-
-        nativeObj = n_VideoCapture(device);
-
-        return;
-    }
-
-    //
-    // C++: double VideoCapture::get(int propId)
-    //
-
-/**
- * Returns the specified "VideoCapture" property.
- *
- * Note: When querying a property that is not supported by the backend used by
- * the "VideoCapture" class, value 0 is returned.
- *
- * @param propId property identifier; it can be one of the following:
- *   * CV_CAP_PROP_FRAME_WIDTH width of the frames in the video stream.
- *   * CV_CAP_PROP_FRAME_HEIGHT height of the frames in the video stream.
- *
- * @see <a href="http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#videocapture-get">org.opencv.highgui.VideoCapture.get</a>
- */
-    public double get(int propId)
-    {
-
-        double retVal = n_get(nativeObj, propId);
-
-        return retVal;
-    }
-
-    public List<Size> getSupportedPreviewSizes()
-    {
-        String[] sizes_str = n_getSupportedPreviewSizes(nativeObj).split(",");
-        List<Size> sizes = new LinkedList<Size>();
-
-        for (String str : sizes_str) {
-            String[] wh = str.split("x");
-            sizes.add(new Size(Double.parseDouble(wh[0]), Double.parseDouble(wh[1])));
-        }
-
-        return sizes;
-    }
-
-    //
-    // C++: bool VideoCapture::grab()
-    //
-
-    // javadoc: VideoCapture::grab()
-    public boolean grab()
-    {
-
-        boolean retVal = n_grab(nativeObj);
-
-        return retVal;
-    }
-
-    //
-    // C++: bool VideoCapture::isOpened()
-    //
-
-    // javadoc: VideoCapture::isOpened()
-    public boolean isOpened()
-    {
-
-        boolean retVal = n_isOpened(nativeObj);
-
-        return retVal;
-    }
-
-    //
-    // C++: bool VideoCapture::open(int device)
-    //
-
-    // javadoc: VideoCapture::open(device)
-    public boolean open(int device)
-    {
-
-        boolean retVal = n_open(nativeObj, device);
-
-        return retVal;
-    }
-
-    //
-    // C++: bool VideoCapture::read(Mat image)
-    //
-
-    // javadoc: VideoCapture::read(image)
-    public boolean read(Mat image)
-    {
-
-        boolean retVal = n_read(nativeObj, image.nativeObj);
-
-        return retVal;
-    }
-
-    //
-    // C++: void VideoCapture::release()
-    //
-
-    // javadoc: VideoCapture::release()
-    public void release()
-    {
-
-        n_release(nativeObj);
-
-        return;
-    }
-
-    //
-    // C++: bool VideoCapture::retrieve(Mat image, int channel = 0)
-    //
-
-    // javadoc: VideoCapture::retrieve(image, channel)
-    public boolean retrieve(Mat image, int channel)
-    {
-
-        boolean retVal = n_retrieve(nativeObj, image.nativeObj, channel);
-
-        return retVal;
-    }
-
-    // javadoc: VideoCapture::retrieve(image)
-    public boolean retrieve(Mat image)
-    {
-
-        boolean retVal = n_retrieve(nativeObj, image.nativeObj);
-
-        return retVal;
-    }
-
-    //
-    // C++: bool VideoCapture::set(int propId, double value)
-    //
-
-/**
- * Sets a property in the "VideoCapture".
- *
- * @param propId property identifier; it can be one of the following:
- *   * CV_CAP_PROP_FRAME_WIDTH width of the frames in the video stream.
- *   * CV_CAP_PROP_FRAME_HEIGHT height of the frames in the video stream.
- * @param value value of the property.
- *
- * @see <a href="http://opencv.itseez.com/modules/highgui/doc/reading_and_writing_images_and_video.html#videocapture-set">org.opencv.highgui.VideoCapture.set</a>
- */
-    public boolean set(int propId, double value)
-    {
-
-        boolean retVal = n_set(nativeObj, propId, value);
-
-        return retVal;
-    }
-
-    @Override
-    protected void finalize() throws Throwable {
-        n_delete(nativeObj);
-        super.finalize();
-    }
-
-    // C++: VideoCapture::VideoCapture()
-    private static native long n_VideoCapture();
-
-    // C++: VideoCapture::VideoCapture(string filename)
-    private static native long n_VideoCapture(java.lang.String filename);
-
-    // C++: VideoCapture::VideoCapture(int device)
-    private static native long n_VideoCapture(int device);
-
-    // C++: double VideoCapture::get(int propId)
-    private static native double n_get(long nativeObj, int propId);
-
-    // C++: bool VideoCapture::grab()
-    private static native boolean n_grab(long nativeObj);
-
-    // C++: bool VideoCapture::isOpened()
-    private static native boolean n_isOpened(long nativeObj);
-
-    // C++: bool VideoCapture::open(string filename)
-    private static native boolean n_open(long nativeObj, java.lang.String filename);
-
-    // C++: bool VideoCapture::open(int device)
-    private static native boolean n_open(long nativeObj, int device);
-
-    // C++: bool VideoCapture::read(Mat image)
-    private static native boolean n_read(long nativeObj, long image_nativeObj);
-
-    // C++: void VideoCapture::release()
-    private static native void n_release(long nativeObj);
-
-    // C++: bool VideoCapture::retrieve(Mat image, int channel = 0)
-    private static native boolean n_retrieve(long nativeObj, long image_nativeObj, int channel);
-
-    private static native boolean n_retrieve(long nativeObj, long image_nativeObj);
-
-    // C++: bool VideoCapture::set(int propId, double value)
-    private static native boolean n_set(long nativeObj, int propId, double value);
-
-    private static native String n_getSupportedPreviewSizes(long nativeObj);
-
-    // native support for java finalize()
-    private static native void n_delete(long nativeObj);
-
-}
diff --git a/modules/java/generator/src/java/utils+Converters.java b/modules/java/generator/src/java/utils+Converters.java
index 49c0844a87..cd4fb9863a 100644
--- a/modules/java/generator/src/java/utils+Converters.java
+++ b/modules/java/generator/src/java/utils+Converters.java
@@ -14,8 +14,8 @@ import org.opencv.core.MatOfPoint3f;
 import org.opencv.core.Point;
 import org.opencv.core.Point3;
 import org.opencv.core.Rect;
-import org.opencv.features2d.DMatch;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.DMatch;
+import org.opencv.core.KeyPoint;
 
 public class Converters {
 
diff --git a/modules/java/test/CMakeLists.txt b/modules/java/test/CMakeLists.txt
index 8f3021991d..122aae22a0 100644
--- a/modules/java/test/CMakeLists.txt
+++ b/modules/java/test/CMakeLists.txt
@@ -75,3 +75,7 @@ add_custom_command(OUTPUT "${opencv_test_java_bin_dir}/build/jar/opencv-test.jar
 
 add_custom_target(${PROJECT_NAME} ALL SOURCES "${opencv_test_java_bin_dir}/build/jar/opencv-test.jar")
 add_dependencies(${PROJECT_NAME} ${the_module})
+
+if(ENABLE_SOLUTION_FOLDERS)
+  set_target_properties(${PROJECT_NAME} PROPERTIES FOLDER "tests accuracy")
+endif()
diff --git a/modules/java/test/src/org/opencv/test/OpenCVTestCase.java b/modules/java/test/src/org/opencv/test/OpenCVTestCase.java
index ac1bf863ce..496f96242b 100644
--- a/modules/java/test/src/org/opencv/test/OpenCVTestCase.java
+++ b/modules/java/test/src/org/opencv/test/OpenCVTestCase.java
@@ -23,8 +23,8 @@ import org.opencv.core.Point3;
 import org.opencv.core.Rect;
 import org.opencv.core.Scalar;
 import org.opencv.core.Size;
-import org.opencv.features2d.DMatch;
-import org.opencv.features2d.KeyPoint;
+import org.opencv.core.DMatch;
+import org.opencv.core.KeyPoint;
 import org.opencv.highgui.Highgui;
 
 public class OpenCVTestCase extends TestCase {
@@ -97,7 +97,7 @@ public class OpenCVTestCase extends TestCase {
         super.setUp();
 
         try {
-            System.loadLibrary("opencv_java");
+            System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
         } catch (SecurityException e) {
             System.out.println(e.toString());
             System.exit(-1);
diff --git a/modules/legacy/doc/common_interfaces_of_descriptor_extractors.rst b/modules/legacy/doc/common_interfaces_of_descriptor_extractors.rst
index dc5d81db34..4f13385dc8 100644
--- a/modules/legacy/doc/common_interfaces_of_descriptor_extractors.rst
+++ b/modules/legacy/doc/common_interfaces_of_descriptor_extractors.rst
@@ -22,7 +22,7 @@ Wrapping class for computing descriptors by using the
     class CalonderDescriptorExtractor : public DescriptorExtractor
     {
     public:
-        CalonderDescriptorExtractor( const string& classifierFile );
+        CalonderDescriptorExtractor( const String& classifierFile );
 
         virtual void read( const FileNode &fn );
         virtual void write( FileStorage &fs ) const;
@@ -30,4 +30,4 @@ Wrapping class for computing descriptors by using the
         virtual int descriptorType() const;
     protected:
         ...
-    }
\ No newline at end of file
+    }
diff --git a/modules/legacy/doc/common_interfaces_of_generic_descriptor_matchers.rst b/modules/legacy/doc/common_interfaces_of_generic_descriptor_matchers.rst
index a2abfa3e49..e959bbf69a 100644
--- a/modules/legacy/doc/common_interfaces_of_generic_descriptor_matchers.rst
+++ b/modules/legacy/doc/common_interfaces_of_generic_descriptor_matchers.rst
@@ -26,7 +26,7 @@ and finding the nearest closest descriptor to an input feature. ::
                             const char* pca_hr_config = 0, const char* pca_desc_config = 0, int pyr_levels = 1,
                             int pca_dim_high = 100, int pca_dim_low = 100);
 
-        OneWayDescriptorBase(CvSize patch_size, int pose_count, const string &pca_filename, const string &train_path = string(), const string &images_list = string(),
+        OneWayDescriptorBase(CvSize patch_size, int pose_count, const String &pca_filename, const String &train_path = String(), const String &images_list = String(),
                             float _scale_min = 0.7f, float _scale_max=1.5f, float _scale_step=1.2f, int pyr_levels = 1,
                             int pca_dim_high = 100, int pca_dim_low = 100);
 
@@ -160,7 +160,7 @@ and finding the nearest closest descriptor to an input feature. ::
         void ConvertDescriptorsArrayToTree(); // Converting pca_descriptors array to KD tree
 
         // GetPCAFilename: get default PCA filename
-        static string GetPCAFilename () { return "pca.yml"; }
+        static String GetPCAFilename () { return "pca.yml"; }
 
         virtual bool empty() const { return m_train_feature_count <= 0 ? true : false; }
 
@@ -190,16 +190,16 @@ Wrapping class for computing, matching, and classifying descriptors using the
 
             Params( int poseCount = POSE_COUNT,
                     Size patchSize = Size(PATCH_WIDTH, PATCH_HEIGHT),
-                    string pcaFilename = string(),
-                    string trainPath = string(), string trainImagesList = string(),
+                    String pcaFilename = String(),
+                    String trainPath = String(), String trainImagesList = String(),
                     float minScale = GET_MIN_SCALE(), float maxScale = GET_MAX_SCALE(),
                     float stepScale = GET_STEP_SCALE() );
 
             int poseCount;
             Size patchSize;
-            string pcaFilename;
-            string trainPath;
-            string trainImagesList;
+            String pcaFilename;
+            String trainPath;
+            String trainImagesList;
 
             float minScale, maxScale, stepScale;
         };
@@ -330,7 +330,7 @@ Wrapping class for computing, matching, and classifying descriptors using the
                     int compressionMethod=FernClassifier::COMPRESSION_NONE,
                     const PatchGenerator& patchGenerator=PatchGenerator() );
 
-            Params( const string& filename );
+            Params( const String& filename );
 
             int nclasses;
             int patchSize;
@@ -341,7 +341,7 @@ Wrapping class for computing, matching, and classifying descriptors using the
             int compressionMethod;
             PatchGenerator patchGenerator;
 
-            string filename;
+            String filename;
         };
 
         FernDescriptorMatcher( const Params& params=Params() );
@@ -361,4 +361,3 @@ Wrapping class for computing, matching, and classifying descriptors using the
     protected:
             ...
     };
-
diff --git a/modules/legacy/doc/expectation_maximization.rst b/modules/legacy/doc/expectation_maximization.rst
index d97d152e7b..6a5418f1fa 100644
--- a/modules/legacy/doc/expectation_maximization.rst
+++ b/modules/legacy/doc/expectation_maximization.rst
@@ -5,6 +5,12 @@ This section describes obsolete ``C`` interface of EM algorithm. Details of the
 
 .. highlight:: cpp
 
+.. note::
+
+   * An example on using the Expectation Maximalization algorithm can be found at opencv_source_code/samples/cpp/em.cpp
+
+   * (Python) An example using Expectation Maximalization for Gaussian Mixing can be found at opencv_source_code/samples/python2/gaussian_mix.py
+
 
 CvEMParams
 ----------
@@ -185,4 +191,3 @@ Reads the trained Gaussian mixture model from the file storage.
     :param fs: A file storage with the trained model.
 
     :param node: The parent map. If it is NULL, the function searches a node with parameters in all the top-level nodes (streams), starting with the first one.
-
diff --git a/modules/legacy/doc/feature_detection_and_description.rst b/modules/legacy/doc/feature_detection_and_description.rst
index 222d7a03a6..3dc1ae5792 100644
--- a/modules/legacy/doc/feature_detection_and_description.rst
+++ b/modules/legacy/doc/feature_detection_and_description.rst
@@ -75,7 +75,9 @@ Class containing a base structure for ``RTreeClassifier``. ::
             void estimateQuantPercForPosteriors(float perc[2]);
     };
 
+.. note::
 
+   * : PYTHON : An example using Randomized Tree training for letter recognition can be found at opencv_source_code/samples/python2/letter_recog.py
 
 RandomizedTree::train
 -------------------------
@@ -99,7 +101,9 @@ Trains a randomized tree using an input set of keypoints.
 
     :param num_quant_bits: Number of bits used for quantization.
 
+.. note::
 
+   * : An example on training a Random Tree Classifier for letter recognition can be found at opencv_source_code\samples\cpp\letter_recog.cpp
 
 RandomizedTree::read
 ------------------------
diff --git a/modules/legacy/doc/histograms.rst b/modules/legacy/doc/histograms.rst
index 0405bf2fe0..fba348c3e6 100644
--- a/modules/legacy/doc/histograms.rst
+++ b/modules/legacy/doc/histograms.rst
@@ -39,11 +39,6 @@ Queries the value of the histogram bin.
 .. ocv:cfunction:: float cvQueryHistValue_3D(CvHistogram hist, int idx0, int idx1, int idx2)
 .. ocv:cfunction:: float cvQueryHistValue_nD(CvHistogram hist, const int* idx)
 
-.. ocv:pyoldfunction:: cv.QueryHistValue_1D(hist, idx0) -> float
-.. ocv:pyoldfunction:: cv.QueryHistValue_2D(hist, idx0, idx1) -> float
-.. ocv:pyoldfunction:: cv.QueryHistValue_3D(hist, idx0, idx1, idx2) -> float
-.. ocv:pyoldfunction:: cv.QueryHistValue_nD(hist, idx) -> float
-
     :param hist: Histogram.
 
     :param idx0: 0-th index.
@@ -92,4 +87,3 @@ Returns a pointer to the histogram bin.
 ..
 
 The macros ``GetHistValue`` return a pointer to the specified bin of the 1D, 2D, 3D, or N-D histogram. In case of a sparse histogram, the function creates a new bin and sets it to 0, unless it exists already.
-
diff --git a/modules/legacy/doc/motion_analysis.rst b/modules/legacy/doc/motion_analysis.rst
index 005e186214..8c046bf245 100644
--- a/modules/legacy/doc/motion_analysis.rst
+++ b/modules/legacy/doc/motion_analysis.rst
@@ -10,8 +10,6 @@ Calculates the optical flow for two images by using the block matching method.
 
 .. ocv:cfunction:: void cvCalcOpticalFlowBM( const CvArr* prev, const CvArr* curr, CvSize block_size, CvSize shift_size, CvSize max_range, int use_previous, CvArr* velx, CvArr* vely )
 
-.. ocv:pyoldfunction:: cv.CalcOpticalFlowBM(prev, curr, blockSize, shiftSize, max_range, usePrevious, velx, vely)-> None
-
         :param prev: First image, 8-bit, single-channel
 
         :param curr: Second image, 8-bit, single-channel
@@ -45,8 +43,6 @@ Calculates the optical flow for two images using Horn-Schunck algorithm.
 
 .. ocv:cfunction:: void cvCalcOpticalFlowHS(const CvArr* prev, const CvArr* curr, int use_previous, CvArr* velx, CvArr* vely, double lambda, CvTermCriteria criteria)
 
-.. ocv:pyoldfunction:: cv.CalcOpticalFlowHS(prev, curr, usePrevious, velx, vely, lambda, criteria)-> None
-
     :param prev: First image, 8-bit, single-channel
 
     :param curr: Second image, 8-bit, single-channel
@@ -71,8 +67,6 @@ Calculates the optical flow for two images using Lucas-Kanade algorithm.
 
 .. ocv:cfunction:: void cvCalcOpticalFlowLK( const CvArr* prev, const CvArr* curr, CvSize win_size, CvArr* velx, CvArr* vely )
 
-.. ocv:pyoldfunction:: cv.CalcOpticalFlowLK(prev, curr, winSize, velx, vely)-> None
-
     :param prev: First image, 8-bit, single-channel
 
     :param curr: Second image, 8-bit, single-channel
@@ -84,5 +78,3 @@ Calculates the optical flow for two images using Lucas-Kanade algorithm.
     :param vely: Vertical component of the optical flow of the same size as input images, 32-bit floating-point, single-channel
 
 The function computes the flow for every pixel of the first input image using the Lucas and Kanade algorithm [Lucas81]_. The function is obsolete. To track sparse features, use :ocv:func:`calcOpticalFlowPyrLK`. To track all the pixels, use :ocv:func:`calcOpticalFlowFarneback`.
-
-
diff --git a/modules/legacy/doc/planar_subdivisions.rst b/modules/legacy/doc/planar_subdivisions.rst
index f01e11ae8d..f4ce8cbb59 100644
--- a/modules/legacy/doc/planar_subdivisions.rst
+++ b/modules/legacy/doc/planar_subdivisions.rst
@@ -116,7 +116,6 @@ CalcSubdivVoronoi2D
 Calculates the coordinates of the Voronoi diagram cells.
 
 .. ocv:cfunction:: void cvCalcSubdivVoronoi2D(  CvSubdiv2D* subdiv )
-.. ocv:pyoldfunction:: cv.CalcSubdivVoronoi2D(subdiv)-> None
 
     :param subdiv: Delaunay subdivision, in which all the points are already added.
 
@@ -130,7 +129,6 @@ ClearSubdivVoronoi2D
 Removes all virtual points.
 
 .. ocv:cfunction:: void cvClearSubdivVoronoi2D( CvSubdiv2D* subdiv )
-.. ocv:pyoldfunction:: cv.ClearSubdivVoronoi2D(subdiv)-> None
 
     :param subdiv: Delaunay subdivision.
 
@@ -145,7 +143,6 @@ CreateSubdivDelaunay2D
 Creates an empty Delaunay triangulation.
 
 .. ocv:cfunction:: CvSubdiv2D* cvCreateSubdivDelaunay2D(  CvRect rect, CvMemStorage* storage )
-.. ocv:pyoldfunction:: cv.CreateSubdivDelaunay2D(rect, storage) -> CvSubdiv2D
 
     :param rect: Rectangle that includes all of the 2D points that are to be added to the subdivision.
 
@@ -166,7 +163,6 @@ FindNearestPoint2D
 Finds the subdivision vertex closest to the given point.
 
 .. ocv:cfunction:: CvSubdiv2DPoint* cvFindNearestPoint2D(  CvSubdiv2D* subdiv, CvPoint2D32f pt )
-.. ocv:pyoldfunction:: cv.FindNearestPoint2D(subdiv, pt)-> point
 
     :param subdiv: Delaunay or another subdivision.
 
@@ -185,7 +181,6 @@ Subdiv2DEdgeDst
 Returns the edge destination.
 
 .. ocv:cfunction:: CvSubdiv2DPoint* cvSubdiv2DEdgeDst(  CvSubdiv2DEdge edge )
-.. ocv:pyoldfunction:: cv.Subdiv2DEdgeDst(edge)-> point
 
     :param edge: Subdivision edge (not a quad-edge).
 
@@ -200,7 +195,6 @@ Subdiv2DGetEdge
 Returns one of the edges related to the given edge.
 
 .. ocv:cfunction:: CvSubdiv2DEdge  cvSubdiv2DGetEdge( CvSubdiv2DEdge edge, CvNextEdgeType type )
-.. ocv:pyoldfunction:: cv.Subdiv2DGetEdge(edge, type)-> CvSubdiv2DEdge
 
     :param edge: Subdivision edge (not a quad-edge).
 
@@ -231,7 +225,6 @@ Subdiv2DNextEdge
 Returns next edge around the edge origin.
 
 .. ocv:cfunction:: CvSubdiv2DEdge  cvSubdiv2DNextEdge( CvSubdiv2DEdge edge )
-.. ocv:pyoldfunction:: cv.Subdiv2DNextEdge(edge)-> CvSubdiv2DEdge
 
     :param edge: Subdivision edge (not a quad-edge).
 
@@ -246,7 +239,6 @@ Subdiv2DLocate
 Returns the location of a point within a Delaunay triangulation.
 
 .. ocv:cfunction:: CvSubdiv2DPointLocation  cvSubdiv2DLocate(  CvSubdiv2D* subdiv, CvPoint2D32f pt, CvSubdiv2DEdge* edge, CvSubdiv2DPoint** vertex=NULL )
-.. ocv:pyoldfunction:: cv.Subdiv2DLocate(subdiv, pt) -> (loc, where)
 
     :param subdiv: Delaunay or another subdivision.
 
@@ -294,7 +286,6 @@ Subdiv2DRotateEdge
 Returns another edge of the same quad-edge.
 
 .. ocv:cfunction:: CvSubdiv2DEdge  cvSubdiv2DRotateEdge(  CvSubdiv2DEdge edge, int rotate )
-.. ocv:pyoldfunction:: cv.Subdiv2DRotateEdge(edge, rotate)-> CvSubdiv2DEdge
 
     :param edge: Subdivision edge (not a quad-edge).
 
@@ -315,11 +306,9 @@ SubdivDelaunay2DInsert
 Inserts a single point into a Delaunay triangulation.
 
 .. ocv:cfunction:: CvSubdiv2DPoint*  cvSubdivDelaunay2DInsert(  CvSubdiv2D* subdiv, CvPoint2D32f pt)
-.. ocv:pyoldfunction:: cv.SubdivDelaunay2DInsert(subdiv, pt)-> point
 
     :param subdiv: Delaunay subdivision created by the function  :ocv:cfunc:`CreateSubdivDelaunay2D`.
 
     :param pt: Inserted point.
 
 The function inserts a single point into a subdivision and modifies the subdivision topology appropriately. If a point with the same coordinates exists already, no new point is added. The function returns a pointer to the allocated point. No virtual point coordinates are calculated at this stage.
-
diff --git a/modules/legacy/include/opencv2/legacy.hpp b/modules/legacy/include/opencv2/legacy.hpp
index 7b765b0c8f..ed0ac9cc34 100644
--- a/modules/legacy/include/opencv2/legacy.hpp
+++ b/modules/legacy/include/opencv2/legacy.hpp
@@ -42,13 +42,12 @@
 #ifndef __OPENCV_LEGACY_HPP__
 #define __OPENCV_LEGACY_HPP__
 
-#include "opencv2/imgproc.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/features2d.hpp"
-#include "opencv2/calib3d.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 #include "opencv2/ml.hpp"
 
 #ifdef __cplusplus
+#include "opencv2/features2d.hpp"
 extern "C" {
 #endif
 
@@ -1887,7 +1886,7 @@ public:
     void setVerbose(bool verbose);
 
     void read(const FileNode& node);
-    void write(FileStorage& fs, const std::string& name=std::string()) const;
+    void write(FileStorage& fs, const String& name=String()) const;
 
     int radius;
     int threshold;
@@ -1918,7 +1917,7 @@ public:
                    const PatchGenerator& patchGenerator=PatchGenerator());
     virtual ~FernClassifier();
     virtual void read(const FileNode& n);
-    virtual void write(FileStorage& fs, const std::string& name=std::string()) const;
+    virtual void write(FileStorage& fs, const String& name=String()) const;
     virtual void trainFromSingleView(const Mat& image,
                                      const std::vector<KeyPoint>& keypoints,
                                      int _patchSize=PATCH_SIZE,
@@ -2062,8 +2061,8 @@ public:
     inline void applyQuantization(int num_quant_bits) { makePosteriors2(num_quant_bits); }
 
     // debug
-    void savePosteriors(std::string url, bool append=false);
-    void savePosteriors2(std::string url, bool append=false);
+    void savePosteriors(String url, bool append=false);
+    void savePosteriors2(String url, bool append=false);
 
 private:
     int classes_;
@@ -2181,9 +2180,9 @@ public:
     void write(std::ostream &os) const;
 
     // experimental and debug
-    void saveAllFloatPosteriors(std::string file_url);
-    void saveAllBytePosteriors(std::string file_url);
-    void setFloatPosteriorsFromTextfile_176(std::string url);
+    void saveAllFloatPosteriors(String file_url);
+    void saveAllBytePosteriors(String file_url);
+    void setFloatPosteriorsFromTextfile_176(String url);
     float countZeroElements();
 
     std::vector<RandomizedTree> trees_;
@@ -2356,7 +2355,7 @@ protected:
     CvAffinePose* m_affine_poses; // an array of poses
     CvMat** m_transforms; // an array of affine transforms corresponding to poses
 
-    std::string m_feature_name; // the name of the feature associated with the descriptor
+    String m_feature_name; // the name of the feature associated with the descriptor
     CvPoint m_center; // the coordinates of the feature (the center of the input image ROI)
 
     int m_pca_dim_high; // the number of descriptor pca components to use for generating affine poses
@@ -2382,7 +2381,7 @@ public:
                          const char* pca_hr_config = 0, const char* pca_desc_config = 0, int pyr_levels = 1,
                          int pca_dim_high = 100, int pca_dim_low = 100);
 
-    OneWayDescriptorBase(CvSize patch_size, int pose_count, const std::string &pca_filename, const std::string &train_path = std::string(), const std::string &images_list = std::string(),
+    OneWayDescriptorBase(CvSize patch_size, int pose_count, const String &pca_filename, const String &train_path = String(), const String &images_list = String(),
                          float _scale_min = 0.7f, float _scale_max=1.5f, float _scale_step=1.2f, int pyr_levels = 1,
                          int pca_dim_high = 100, int pca_dim_low = 100);
 
@@ -2516,7 +2515,7 @@ public:
     void ConvertDescriptorsArrayToTree(); // Converting pca_descriptors array to KD tree
 
     // GetPCAFilename: get default PCA filename
-    static std::string GetPCAFilename () { return "pca.yml"; }
+    static String GetPCAFilename () { return "pca.yml"; }
 
     virtual bool empty() const { return m_train_feature_count <= 0 ? true : false; }
 
@@ -2568,8 +2567,8 @@ public:
     OneWayDescriptorObject(CvSize patch_size, int pose_count, const char* train_path, const char* pca_config,
                            const char* pca_hr_config = 0, const char* pca_desc_config = 0, int pyr_levels = 1);
 
-    OneWayDescriptorObject(CvSize patch_size, int pose_count, const std::string &pca_filename,
-                           const std::string &train_path = std::string (), const std::string &images_list = std::string (),
+    OneWayDescriptorObject(CvSize patch_size, int pose_count, const String &pca_filename,
+                           const String &train_path = String (), const String &images_list = String (),
                            float _scale_min = 0.7f, float _scale_max=1.5f, float _scale_step=1.2f, int pyr_levels = 1);
 
 
@@ -2633,16 +2632,16 @@ public:
 
         Params( int poseCount = POSE_COUNT,
                Size patchSize = Size(PATCH_WIDTH, PATCH_HEIGHT),
-               std::string pcaFilename = std::string(),
-               std::string trainPath = std::string(), std::string trainImagesList = std::string(),
+               String pcaFilename = String(),
+               String trainPath = String(), String trainImagesList = String(),
                float minScale = GET_MIN_SCALE(), float maxScale = GET_MAX_SCALE(),
                float stepScale = GET_STEP_SCALE() );
 
         int poseCount;
         Size patchSize;
-        std::string pcaFilename;
-        std::string trainPath;
-        std::string trainImagesList;
+        String pcaFilename;
+        String trainPath;
+        String trainImagesList;
 
         float minScale, maxScale, stepScale;
     };
@@ -2706,7 +2705,7 @@ public:
                int compressionMethod=FernClassifier::COMPRESSION_NONE,
                const PatchGenerator& patchGenerator=PatchGenerator() );
 
-        Params( const std::string& filename );
+        Params( const String& filename );
 
         int nclasses;
         int patchSize;
@@ -2717,7 +2716,7 @@ public:
         int compressionMethod;
         PatchGenerator patchGenerator;
 
-        std::string filename;
+        String filename;
     };
 
     FernDescriptorMatcher( const Params& params=Params() );
@@ -2759,7 +2758,7 @@ template<typename T>
 class CV_EXPORTS CalonderDescriptorExtractor : public DescriptorExtractor
 {
 public:
-    CalonderDescriptorExtractor( const std::string& classifierFile );
+    CalonderDescriptorExtractor( const String& classifierFile );
 
     virtual void read( const FileNode &fn );
     virtual void write( FileStorage &fs ) const;
@@ -2777,7 +2776,7 @@ protected:
 };
 
 template<typename T>
-CalonderDescriptorExtractor<T>::CalonderDescriptorExtractor(const std::string& classifier_file)
+CalonderDescriptorExtractor<T>::CalonderDescriptorExtractor(const String& classifier_file)
 {
     classifier_.read( classifier_file.c_str() );
 }
@@ -2867,7 +2866,7 @@ public:
     void setVerbose(bool verbose);
 
     void read(const FileNode& node);
-    void write(FileStorage& fs, const std::string& name=std::string()) const;
+    void write(FileStorage& fs, const String& name=String()) const;
     bool operator()(const Mat& image, CV_OUT Mat& H, CV_OUT std::vector<Point2f>& corners) const;
     bool operator()(const std::vector<Mat>& pyr, const std::vector<KeyPoint>& keypoints,
                     CV_OUT Mat& H, CV_OUT std::vector<Point2f>& corners,
diff --git a/modules/legacy/include/opencv2/legacy/compat.hpp b/modules/legacy/include/opencv2/legacy/compat.hpp
index 5b5495edcb..5840d747a0 100644
--- a/modules/legacy/include/opencv2/legacy/compat.hpp
+++ b/modules/legacy/include/opencv2/legacy/compat.hpp
@@ -39,14 +39,6 @@
 //
 //M*/
 
-/*
-   A few macros and definitions for backward compatibility
-   with the previous versions of OpenCV. They are obsolete and
-   are likely to be removed in future. To check whether your code
-   uses any of these, define CV_NO_BACKWARD_COMPATIBILITY before
-   including cv.h.
-*/
-
 #ifndef __OPENCV_COMPAT_HPP__
 #define __OPENCV_COMPAT_HPP__
 
@@ -56,6 +48,8 @@
 #include <math.h>
 #include <string.h>
 
+#define CV_NOOP(...)
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -190,7 +184,8 @@ CV_EXPORTS double cvPseudoInverse( const CvArr* src, CvArr* dst );
 #define cvMinMaxLocMask(img, mask, min_val, max_val, min_loc, max_loc) \
         cvMinMaxLoc(img, min_val, max_val, min_loc, max_loc, mask)
 
-#define cvRemoveMemoryManager  cvSetMemoryManager
+#define cvRemoveMemoryManager  CV_NOOP
+#define cvSetMemoryManager     CV_NOOP
 
 #define cvmSetZero( mat )               cvSetZero( mat )
 #define cvmSetIdentity( mat )           cvSetIdentity( mat )
diff --git a/modules/legacy/include/opencv2/legacy/legacy.hpp b/modules/legacy/include/opencv2/legacy/legacy.hpp
index fafad10a1c..adea8da005 100644
--- a/modules/legacy/include/opencv2/legacy/legacy.hpp
+++ b/modules/legacy/include/opencv2/legacy/legacy.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/legacy.hpp"
\ No newline at end of file
+#include "opencv2/legacy.hpp"
diff --git a/modules/legacy/include/opencv2/legacy/streams.hpp b/modules/legacy/include/opencv2/legacy/streams.hpp
index 6935b00c3d..e164bf4a83 100644
--- a/modules/legacy/include/opencv2/legacy/streams.hpp
+++ b/modules/legacy/include/opencv2/legacy/streams.hpp
@@ -90,4 +90,3 @@ and DirectX SDK (2006 April or later).
 #endif
 
 #endif
-
diff --git a/modules/legacy/src/3dtracker.cpp b/modules/legacy/src/3dtracker.cpp
index 536a534fe2..a55a9f8516 100644
--- a/modules/legacy/src/3dtracker.cpp
+++ b/modules/legacy/src/3dtracker.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 #include <stdio.h>
 #include <map>
@@ -321,17 +322,19 @@ static void DrawEtalon(IplImage *img, CvPoint2D32f *corners,
     const int r = 4;
     int i;
     int x, y;
-    CvPoint prev_pt = { 0, 0 };
-    static const CvScalar rgb_colors[] = {
-        {{0,0,255}},
-        {{0,128,255}},
-        {{0,200,200}},
-        {{0,255,0}},
-        {{200,200,0}},
-        {{255,0,0}},
-        {{255,0,255}} };
+    CvPoint prev_pt;
+    static const CvScalar rgb_colors[] =
+    {
+        CvScalar(0,0,255),
+        CvScalar(0,128,255),
+        CvScalar(0,200,200),
+        CvScalar(0,255,0),
+        CvScalar(200,200,0),
+        CvScalar(255,0,0),
+        CvScalar(255,0,255)
+    };
     static const CvScalar gray_colors[] = {
-        {{80}}, {{120}}, {{160}}, {{200}}, {{100}}, {{140}}, {{180}}
+        CvScalar(80), CvScalar(120), CvScalar(160), CvScalar(200), CvScalar(100), CvScalar(140), CvScalar(180)
     };
     const CvScalar* colors = img->nChannels == 3 ? rgb_colors : gray_colors;
 
diff --git a/modules/legacy/src/_vectrack.h b/modules/legacy/src/_vectrack.h
index 77df1e3c0c..9c90cc7e8a 100644
--- a/modules/legacy/src/_vectrack.h
+++ b/modules/legacy/src/_vectrack.h
@@ -87,8 +87,8 @@ int PointInRect(const CvPoint& p, const CvRect& r)
 inline
 int RectInRect(const CvRect& r1, const CvRect& r2)
 {
-        CvPoint plt = {r1.x, r1.y};
-        CvPoint prb = {r1.x + r1.width, r1.y + r1.height};
+        CvPoint plt(r1.x, r1.y);
+        CvPoint prb(r1.x + r1.width, r1.y + r1.height);
         return (PointInRect(plt, r2) && PointInRect(prb, r2));
 }
 
diff --git a/modules/legacy/src/_vm.h b/modules/legacy/src/_vm.h
index df247fffb5..7964cd35b0 100644
--- a/modules/legacy/src/_vm.h
+++ b/modules/legacy/src/_vm.h
@@ -293,4 +293,3 @@ int icvSingularValueDecomposition(	int		M,
 
 /*======================================================================================*/
 #endif/*_CV_VM_H_*/
-
diff --git a/modules/legacy/src/bgfg_codebook.cpp b/modules/legacy/src/bgfg_codebook.cpp
index 03e7e09784..2a53b29bf6 100644
--- a/modules/legacy/src/bgfg_codebook.cpp
+++ b/modules/legacy/src/bgfg_codebook.cpp
@@ -359,4 +359,3 @@ void cvBGCodeBookClearStale( CvBGCodeBookModel* model, int staleThresh,
 }
 
 /* End of file. */
-
diff --git a/modules/legacy/src/bgfg_common.cpp b/modules/legacy/src/bgfg_common.cpp
index 5431c2703b..2bcea5573b 100644
--- a/modules/legacy/src/bgfg_common.cpp
+++ b/modules/legacy/src/bgfg_common.cpp
@@ -135,4 +135,3 @@ cvSegmentFGMask( CvArr* _mask, int poly1Hull0, float perimScale,
 }
 
 /* End of file. */
-
diff --git a/modules/legacy/src/bgfg_gaussmix.cpp b/modules/legacy/src/bgfg_gaussmix.cpp
index f4c65f55c3..ae75e40c26 100644
--- a/modules/legacy/src/bgfg_gaussmix.cpp
+++ b/modules/legacy/src/bgfg_gaussmix.cpp
@@ -50,7 +50,7 @@ icvReleaseGaussianBGModel( CvGaussBGModel** bg_model )
 
     if( *bg_model )
     {
-        delete (cv::BackgroundSubtractorMOG*)((*bg_model)->mog);
+        delete (cv::Ptr<cv::BackgroundSubtractor>*)((*bg_model)->mog);
         cvReleaseImage( &(*bg_model)->background );
         cvReleaseImage( &(*bg_model)->foreground );
         memset( *bg_model, 0, sizeof(**bg_model) );
@@ -65,10 +65,10 @@ icvUpdateGaussianBGModel( IplImage* curr_frame, CvGaussBGModel*  bg_model, doubl
 {
     cv::Mat image = cv::cvarrToMat(curr_frame), mask = cv::cvarrToMat(bg_model->foreground);
 
-    cv::BackgroundSubtractorMOG* mog = (cv::BackgroundSubtractorMOG*)(bg_model->mog);
+    cv::Ptr<cv::BackgroundSubtractor>* mog = (cv::Ptr<cv::BackgroundSubtractor>*)(bg_model->mog);
     CV_Assert(mog != 0);
 
-    (*mog)(image, mask, learningRate);
+    (*mog)->apply(image, mask, learningRate);
     bg_model->countFrames++;
 
     return 0;
@@ -105,13 +105,11 @@ cvCreateGaussianBGModel( IplImage* first_frame, CvGaussBGStatModelParams* parame
 
     bg_model->params = params;
 
-    cv::BackgroundSubtractorMOG* mog =
-        new cv::BackgroundSubtractorMOG(params.win_size,
-                                        params.n_gauss,
-                                        params.bg_threshold,
-                                        params.variance_init);
-
-    bg_model->mog = mog;
+    cv::Ptr<cv::BackgroundSubtractor> mog = cv::createBackgroundSubtractorMOG(params.win_size, params.n_gauss,
+                                                                              params.bg_threshold);
+    cv::Ptr<cv::BackgroundSubtractor>* pmog = new cv::Ptr<cv::BackgroundSubtractor>;
+    *pmog = mog;
+    bg_model->mog = pmog;
 
     CvSize sz = cvGetSize(first_frame);
     bg_model->background = cvCreateImage(sz, IPL_DEPTH_8U, first_frame->nChannels);
@@ -1319,4 +1317,3 @@ icvUpdateGaussianBGModel2( IplImage* curr_frame, CvGaussBGModel2*  bg_model )
 }
 
 /* End of file. */
-
diff --git a/modules/legacy/src/blobtrack.cpp b/modules/legacy/src/blobtrack.cpp
index 6d0a5e3f93..48b83ef914 100644
--- a/modules/legacy/src/blobtrack.cpp
+++ b/modules/legacy/src/blobtrack.cpp
@@ -637,4 +637,3 @@ void cvReleaseBlobTracker(CvBlobTracker**ppT )
     ppT[0]->Release();
     ppT[0] = 0;
 }
-
diff --git a/modules/legacy/src/blobtrackanalysis.cpp b/modules/legacy/src/blobtrackanalysis.cpp
index 5095d3aee2..6416ae688e 100644
--- a/modules/legacy/src/blobtrackanalysis.cpp
+++ b/modules/legacy/src/blobtrackanalysis.cpp
@@ -125,4 +125,3 @@ CvBlobTrackAnalysis* cvCreateBlobTrackAnalysisList(CvBlobTrackAnalysisOne* (*cre
 }
 
 /* ======================== Analyser modules ============================= */
-
diff --git a/modules/legacy/src/blobtrackanalysisior.cpp b/modules/legacy/src/blobtrackanalysisior.cpp
index 571ca9ef06..c5b2222e66 100644
--- a/modules/legacy/src/blobtrackanalysisior.cpp
+++ b/modules/legacy/src/blobtrackanalysisior.cpp
@@ -171,4 +171,3 @@ CvBlobTrackAnalysis* cvCreateModuleBlobTrackAnalysisIOR()
     return (CvBlobTrackAnalysis*)pIOR;
 }/* cvCreateCvBlobTrackAnalysisIOR */
 /* ======================== Analyser modules ============================= */
-
diff --git a/modules/legacy/src/blobtrackanalysistrackdist.cpp b/modules/legacy/src/blobtrackanalysistrackdist.cpp
index 9e93ca27a4..2adba7e3a0 100644
--- a/modules/legacy/src/blobtrackanalysistrackdist.cpp
+++ b/modules/legacy/src/blobtrackanalysistrackdist.cpp
@@ -578,4 +578,3 @@ public:
 
 CvBlobTrackAnalysis* cvCreateModuleBlobTrackAnalysisTrackDist()
 {return (CvBlobTrackAnalysis*) new CvBlobTrackAnalysisTrackDist;}
-
diff --git a/modules/legacy/src/blobtrackgen1.cpp b/modules/legacy/src/blobtrackgen1.cpp
index b76086e3eb..7114f2b4ce 100644
--- a/modules/legacy/src/blobtrackgen1.cpp
+++ b/modules/legacy/src/blobtrackgen1.cpp
@@ -181,5 +181,3 @@ CvBlobTrackGen* cvCreateModuleBlobTrackGen1()
 {
     return (CvBlobTrackGen*) new CvBlobTrackGen1(0);
 }
-
-
diff --git a/modules/legacy/src/blobtrackgenyml.cpp b/modules/legacy/src/blobtrackgenyml.cpp
index 5d0e870fef..0d9de45ee3 100644
--- a/modules/legacy/src/blobtrackgenyml.cpp
+++ b/modules/legacy/src/blobtrackgenyml.cpp
@@ -61,7 +61,7 @@ protected:
     {
         int     ObjNum = m_TrackList.GetBlobNum();
         int     i;
-        char    video_name[1024];
+        char    video_name[1024+1];
         char*   struct_name = NULL;
         CvFileStorage* storage = cvOpenFileStorage(m_pFileName,NULL,CV_STORAGE_WRITE_TEXT);
 
@@ -215,5 +215,3 @@ CvBlobTrackGen* cvCreateModuleBlobTrackGenYML()
 {
     return (CvBlobTrackGen*) new CvBlobTrackGenYML;
 }
-
-
diff --git a/modules/legacy/src/blobtrackingauto.cpp b/modules/legacy/src/blobtrackingauto.cpp
index fab0503bd0..ada0bdff44 100644
--- a/modules/legacy/src/blobtrackingauto.cpp
+++ b/modules/legacy/src/blobtrackingauto.cpp
@@ -429,10 +429,11 @@ void CvBlobTrackerAuto1::Process(IplImage* pImg, IplImage* pMask)
             for(i=0; i<NewBlobList.GetBlobNum(); ++i)
             {
                 CvBlob* pBN = NewBlobList.GetBlob(i);
-                pBN->ID = m_NextBlobID;
 
                 if(pBN && pBN->w >= CV_BLOB_MINW && pBN->h >= CV_BLOB_MINH)
                 {
+                    pBN->ID = m_NextBlobID;
+
                     CvBlob* pB = m_pBT->AddBlob(pBN, pImg, pmask );
                     if(pB)
                     {
@@ -478,4 +479,3 @@ void CvBlobTrackerAuto1::Process(IplImage* pImg, IplImage* pMask)
     TIME_END("TrackAnalysis",m_BlobList.GetBlobNum())
 
 } /* CvBlobTrackerAuto1::Process */
-
diff --git a/modules/legacy/src/blobtrackingmsfg.cpp b/modules/legacy/src/blobtrackingmsfg.cpp
index 8444964447..5fd9634d4d 100644
--- a/modules/legacy/src/blobtrackingmsfg.cpp
+++ b/modules/legacy/src/blobtrackingmsfg.cpp
@@ -1179,4 +1179,3 @@ CvBlobTracker* cvCreateBlobTrackerMSPF()
 {
     return cvCreateBlobTrackerList(cvCreateBlobTrackerOneMSPF);
 }
-
diff --git a/modules/legacy/src/blobtrackingmsfgs.cpp b/modules/legacy/src/blobtrackingmsfgs.cpp
index 3e1c7a066e..f3b3cf3753 100644
--- a/modules/legacy/src/blobtrackingmsfgs.cpp
+++ b/modules/legacy/src/blobtrackingmsfgs.cpp
@@ -454,4 +454,3 @@ CvBlobTracker* cvCreateBlobTrackerMSFGS()
 {
     return cvCreateBlobTrackerList(cvCreateBlobTrackerOneMSFGS);
 }
-
diff --git a/modules/legacy/src/blobtrackpostprockalman.cpp b/modules/legacy/src/blobtrackpostprockalman.cpp
index debe89d3f5..378e1e78fe 100644
--- a/modules/legacy/src/blobtrackpostprockalman.cpp
+++ b/modules/legacy/src/blobtrackpostprockalman.cpp
@@ -39,6 +39,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/video/tracking_c.h"
 
 /*======================= KALMAN FILTER =========================*/
 /* State vector is (x,y,w,h,dx,dy,dw,dh). */
@@ -324,4 +325,3 @@ CvBlobTrackPredictor* cvCreateModuleBlobTrackPredictKalman()
     return (CvBlobTrackPredictor*) new CvBlobTrackPredictKalman;
 }
 /*======================= KALMAN PREDICTOR =========================*/
-
diff --git a/modules/legacy/src/calcimagehomography.cpp b/modules/legacy/src/calcimagehomography.cpp
index 502c24d983..928e056b45 100644
--- a/modules/legacy/src/calcimagehomography.cpp
+++ b/modules/legacy/src/calcimagehomography.cpp
@@ -111,4 +111,3 @@ cvCalcImageHomography( float* line, CvPoint3D32f* _center,
 }
 
 /* End of file. */
-
diff --git a/modules/legacy/src/calibfilter.cpp b/modules/legacy/src/calibfilter.cpp
index f1a4d7cd2b..99e9ec7c12 100644
--- a/modules/legacy/src/calibfilter.cpp
+++ b/modules/legacy/src/calibfilter.cpp
@@ -235,7 +235,7 @@ void CvCalibFilter::SetCameraCount( int count )
             cvReleaseMat( &rectMap[i][1] );
         }
 
-        memset( latestCounts, 0, sizeof(latestPoints) );
+        memset( latestCounts, 0, sizeof(latestCounts) );
         maxPoints = 0;
         cameraCount = count;
     }
@@ -546,19 +546,19 @@ void CvCalibFilter::DrawPoints( CvMat** dstarr )
 
                 static const CvScalar line_colors[] =
                 {
-                    {{0,0,255}},
-                    {{0,128,255}},
-                    {{0,200,200}},
-                    {{0,255,0}},
-                    {{200,200,0}},
-                    {{255,0,0}},
-                    {{255,0,255}}
+                    CvScalar(0,0,255),
+                    CvScalar(0,128,255),
+                    CvScalar(0,200,200),
+                    CvScalar(0,255,0),
+                    CvScalar(200,200,0),
+                    CvScalar(255,0,0),
+                    CvScalar(255,0,255)
                 };
 
                 const int colorCount = sizeof(line_colors)/sizeof(line_colors[0]);
                 const int r = 4;
                 CvScalar color = line_colors[0];
-                CvPoint prev_pt = { 0, 0};
+                CvPoint prev_pt;
 
                 for( j = 0; j < count; j++ )
                 {
@@ -919,5 +919,3 @@ bool CvCalibFilter::Undistort( CvMat** srcarr, CvMat** dstarr )
 
     return true;
 }
-
-
diff --git a/modules/legacy/src/calonder.cpp b/modules/legacy/src/calonder.cpp
index 73f274862d..c5a8a6ec2b 100644
--- a/modules/legacy/src/calonder.cpp
+++ b/modules/legacy/src/calonder.cpp
@@ -334,7 +334,7 @@ void RandomizedTree::train(std::vector<BaseKeypoint> const& base_set,
   Size patchSize(PATCH_SIZE, PATCH_SIZE);
   for (keypt_it = base_set.begin(); keypt_it != base_set.end(); ++keypt_it, ++class_id) {
     for (int i = 0; i < views; ++i) {
-      make_patch( Mat(keypt_it->image), Point(keypt_it->x, keypt_it->y ), patch, patchSize, rng );
+      make_patch( cv::cvarrToMat(keypt_it->image), Point(keypt_it->x, keypt_it->y ), patch, patchSize, rng );
       IplImage iplPatch = patch;
       addExample(class_id, getData(&iplPatch));
     }
@@ -626,7 +626,7 @@ void RandomizedTree::write(std::ostream &os) const
 }
 
 
-void RandomizedTree::savePosteriors(std::string url, bool append)
+void RandomizedTree::savePosteriors(String url, bool append)
 {
    std::ofstream file(url.c_str(), (append?std::ios::app:std::ios::out));
    for (int i=0; i<num_leaves_; i++) {
@@ -641,7 +641,7 @@ void RandomizedTree::savePosteriors(std::string url, bool append)
    file.close();
 }
 
-void RandomizedTree::savePosteriors2(std::string url, bool append)
+void RandomizedTree::savePosteriors2(String url, bool append)
 {
    std::ofstream file(url.c_str(), (append?std::ios::app:std::ios::out));
    for (int i=0; i<num_leaves_; i++) {
@@ -915,7 +915,7 @@ printf("RTreeClassifier::write: num_quant_bits_=%i\n", num_quant_bits_);
     tree_it->write(os);
 }
 
-void RTreeClassifier::saveAllFloatPosteriors(std::string url)
+void RTreeClassifier::saveAllFloatPosteriors(String url)
 {
   printf("[DEBUG] writing all float posteriors to %s...\n", url.c_str());
   for (int i=0; i<(int)trees_.size(); ++i)
@@ -923,7 +923,7 @@ void RTreeClassifier::saveAllFloatPosteriors(std::string url)
   printf("[DEBUG] done\n");
 }
 
-void RTreeClassifier::saveAllBytePosteriors(std::string url)
+void RTreeClassifier::saveAllBytePosteriors(String url)
 {
   printf("[DEBUG] writing all byte posteriors to %s...\n", url.c_str());
   for (int i=0; i<(int)trees_.size(); ++i)
@@ -932,7 +932,7 @@ void RTreeClassifier::saveAllBytePosteriors(std::string url)
 }
 
 
-void RTreeClassifier::setFloatPosteriorsFromTextfile_176(std::string url)
+void RTreeClassifier::setFloatPosteriorsFromTextfile_176(String url)
 {
    std::ifstream ifs(url.c_str());
 
diff --git a/modules/legacy/src/camshift.cpp b/modules/legacy/src/camshift.cpp
index 830e198249..b3cd9974a4 100644
--- a/modules/legacy/src/camshift.cpp
+++ b/modules/legacy/src/camshift.cpp
@@ -39,6 +39,7 @@
 //
 //M*/
 #include "precomp.hpp"
+#include "opencv2/video/tracking_c.h"
 
 CvCamShiftTracker::CvCamShiftTracker()
 {
diff --git a/modules/legacy/src/clique.cpp b/modules/legacy/src/clique.cpp
index 9280cb17b7..d8f2f59da2 100644
--- a/modules/legacy/src/clique.cpp
+++ b/modules/legacy/src/clique.cpp
@@ -706,4 +706,3 @@ void cvBronKerbosch( CvGraph* graph )
 }//end cvBronKerbosch
 
 #endif
-
diff --git a/modules/legacy/src/compat.cpp b/modules/legacy/src/compat.cpp
index 5e7a0d6c6a..52e05b99ff 100644
--- a/modules/legacy/src/compat.cpp
+++ b/modules/legacy/src/compat.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
 CvMat cvMatArray( int rows, int cols, int type,
                   int count, void* data)
diff --git a/modules/legacy/src/corrimages.cpp b/modules/legacy/src/corrimages.cpp
index 027dc6e97c..019fb01f35 100644
--- a/modules/legacy/src/corrimages.cpp
+++ b/modules/legacy/src/corrimages.cpp
@@ -40,11 +40,6 @@
 //M*/
 
 #include "precomp.hpp"
-//#include "cvtypes.h"
-//#include <float.h>
-//#include <limits.h>
-//#include "cv.h"
-//#include "highgui.h"
 #if 0
 #include <stdio.h>
 
diff --git a/modules/legacy/src/createhandmask.cpp b/modules/legacy/src/createhandmask.cpp
index d87754843b..7936718325 100644
--- a/modules/legacy/src/createhandmask.cpp
+++ b/modules/legacy/src/createhandmask.cpp
@@ -124,7 +124,7 @@ cvCreateHandMask( CvSeq * numbers, IplImage * img_mask, CvRect * roi )
     __BEGIN__;
 
     if( img_mask->depth != IPL_DEPTH_8U )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
 
     if( img_mask->nChannels != 1 )
         CV_ERROR( CV_BadNumChannels, "output image have wrong number of channels" );
diff --git a/modules/legacy/src/decomppoly.cpp b/modules/legacy/src/decomppoly.cpp
index c577a9c1f5..0cb10551c1 100644
--- a/modules/legacy/src/decomppoly.cpp
+++ b/modules/legacy/src/decomppoly.cpp
@@ -626,4 +626,3 @@ void cvDecompPoly( CvContour* cont,
 #endif
 
 // End of file decomppoly.cpp
-
diff --git a/modules/legacy/src/dpstereo.cpp b/modules/legacy/src/dpstereo.cpp
index a55e1cafbb..838cf8ae09 100644
--- a/modules/legacy/src/dpstereo.cpp
+++ b/modules/legacy/src/dpstereo.cpp
@@ -76,7 +76,7 @@ typedef struct _CvRightImData
     uchar min_val, max_val;
 } _CvRightImData;
 
-#define CV_IMAX3(a,b,c) ((temp3 = (a) >= (b) ? (a) : (b)),(temp3 >= (c) ? temp3 : (c)))
+#define CV_IMAX3(a,b,c) ((temp2 = (a) >= (b) ? (a) : (b)),(temp2 >= (c) ? temp2 : (c)))
 #define CV_IMIN3(a,b,c) ((temp3 = (a) <= (b) ? (a) : (b)),(temp3 <= (c) ? temp3 : (c)))
 
 static void icvFindStereoCorrespondenceByBirchfieldDP( uchar* src1, uchar* src2,
@@ -87,7 +87,7 @@ static void icvFindStereoCorrespondenceByBirchfieldDP( uchar* src1, uchar* src2,
                                                 float  _param3, float _param4,
                                                 float  _param5 )
 {
-    int     x, y, i, j, temp3;
+    int     x, y, i, j, temp2, temp3;
     int     d, s;
     int     dispH =  maxDisparity + 3;
     uchar  *dispdata;
@@ -551,4 +551,3 @@ cvFindStereoCorrespondence(
 }
 
 /* End of file. */
-
diff --git a/modules/legacy/src/eigenobjects.cpp b/modules/legacy/src/eigenobjects.cpp
index 87d5d4277a..4c8b973636 100644
--- a/modules/legacy/src/eigenobjects.cpp
+++ b/modules/legacy/src/eigenobjects.cpp
@@ -1192,9 +1192,9 @@ cvCalcCovarMatrixEx( int  nObjects, void*  input, int  ioFlags,
 
     cvGetImageRawData( avg, (uchar **) & avg_data, &avg_step, &avg_size );
     if( avg->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( avg->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     if( ioFlags == CV_EIGOBJ_NO_CALLBACK )
     {
@@ -1213,11 +1213,11 @@ cvCalcCovarMatrixEx( int  nObjects, void*  input, int  ioFlags,
 
             cvGetImageRawData( img, &img_data, &img_step, &img_size );
             if( img->depth != IPL_DEPTH_8U )
-                CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                CV_ERROR( CV_BadDepth, "Unsupported format" );
             if( img_size != avg_size || img_size != old_size )
                 CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
             if( img->nChannels != 1 )
-                CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                CV_ERROR( CV_BadNumChannels, "Unsupported format" );
             if( i > 0 && img_step != old_step )
                 CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1327,9 +1327,9 @@ cvCalcEigenObjects( int       nObjects,
 
     cvGetImageRawData( avg, (uchar **) & avg_data, &avg_step, &avg_size );
     if( avg->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( avg->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     if( nEigens > calcLimit->max_iter && calcLimit->type != CV_TERMCRIT_EPS )
         nEigens = calcLimit->max_iter;
@@ -1358,11 +1358,11 @@ cvCalcEigenObjects( int       nObjects,
 
                 cvGetImageRawData( img, &obj_data, &obj_step, &obj_size );
                 if( img->depth != IPL_DEPTH_8U )
-                    CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadDepth, "Unsupported format" );
                 if( obj_size != avg_size || obj_size != old_size )
                     CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
                 if( img->nChannels != 1 )
-                    CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadNumChannels, "Unsupported format" );
                 if( i > 0 && obj_step != old_step )
                     CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1377,11 +1377,11 @@ cvCalcEigenObjects( int       nObjects,
 
                 cvGetImageRawData( eig, (uchar **) & eig_data, &eig_step, &eig_size );
                 if( eig->depth != IPL_DEPTH_32F )
-                    CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadDepth, "Unsupported format" );
                 if( eig_size != avg_size || eig_size != oldeig_size )
                     CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
                 if( eig->nChannels != 1 )
-                    CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadNumChannels, "Unsupported format" );
                 if( i > 0 && eig_step != oldeig_step )
                     CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1415,11 +1415,11 @@ cvCalcEigenObjects( int       nObjects,
 
                 cvGetImageRawData( img, &obj_data, &obj_step, &obj_size );
                 if( img->depth != IPL_DEPTH_8U )
-                    CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadDepth, "Unsupported format" );
                 if( obj_size != avg_size || obj_size != old_size )
                     CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
                 if( img->nChannels != 1 )
-                    CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadNumChannels, "Unsupported format" );
                 if( i > 0 && obj_step != old_step )
                     CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1461,11 +1461,11 @@ cvCalcEigenObjects( int       nObjects,
 
                 cvGetImageRawData( eig, (uchar **) & eig_data, &eig_step, &eig_size );
                 if( eig->depth != IPL_DEPTH_32F )
-                    CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadDepth, "Unsupported format" );
                 if( eig_size != avg_size || eig_size != oldeig_size )
                     CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
                 if( eig->nChannels != 1 )
-                    CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                    CV_ERROR( CV_BadNumChannels, "Unsupported format" );
                 if( i > 0 && eig_step != oldeig_step )
                     CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1545,21 +1545,21 @@ cvCalcDecompCoeff( IplImage * obj, IplImage * eigObj, IplImage * avg )
 
     cvGetImageRawData( obj, &obj_data, &obj_step, &obj_size );
     if( obj->depth != IPL_DEPTH_8U )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( obj->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     cvGetImageRawData( eigObj, (uchar **) & eig_data, &eig_step, &eig_size );
     if( eigObj->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( eigObj->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     cvGetImageRawData( avg, (uchar **) & avg_data, &avg_step, &avg_size );
     if( avg->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( avg->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     if( obj_size != eig_size || obj_size != avg_size )
         CV_ERROR( CV_StsBadArg, "different sizes of images" );
@@ -1614,15 +1614,15 @@ cvEigenDecomposite( IplImage* obj,
 
     cvGetImageRawData( avg, (uchar **) & avg_data, &avg_step, &avg_size );
     if( avg->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( avg->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     cvGetImageRawData( obj, &obj_data, &obj_step, &obj_size );
     if( obj->depth != IPL_DEPTH_8U )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( obj->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     if( obj_size != avg_size )
         CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
@@ -1644,11 +1644,11 @@ cvEigenDecomposite( IplImage* obj,
 
             cvGetImageRawData( eig, (uchar **) & eig_data, &eig_step, &eig_size );
             if( eig->depth != IPL_DEPTH_32F )
-                CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                CV_ERROR( CV_BadDepth, "Unsupported format" );
             if( eig_size != avg_size || eig_size != old_size )
                 CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
             if( eig->nChannels != 1 )
-                CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                CV_ERROR( CV_BadNumChannels, "Unsupported format" );
             if( i > 0 && eig_step != old_step )
                 CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
@@ -1731,15 +1731,15 @@ cvEigenProjection( void*     eigInput,
 
     cvGetImageRawData( avg, (uchar **) & avg_data, &avg_step, &avg_size );
     if( avg->depth != IPL_DEPTH_32F )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( avg->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     cvGetImageRawData( proj, &proj_data, &proj_step, &proj_size );
     if( proj->depth != IPL_DEPTH_8U )
-        CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+        CV_ERROR( CV_BadDepth, "Unsupported format" );
     if( proj->nChannels != 1 )
-        CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_ERROR( CV_BadNumChannels, "Unsupported format" );
 
     if( proj_size != avg_size )
         CV_ERROR( CV_StsBadArg, "Different sizes of projects" );
@@ -1761,11 +1761,11 @@ cvEigenProjection( void*     eigInput,
 
             cvGetImageRawData( eig, (uchar **) & eig_data, &eig_step, &eig_size );
             if( eig->depth != IPL_DEPTH_32F )
-                CV_ERROR( CV_BadDepth, cvUnsupportedFormat );
+                CV_ERROR( CV_BadDepth, "Unsupported format" );
             if( eig_size != avg_size || eig_size != old_size )
                 CV_ERROR( CV_StsBadArg, "Different sizes of objects" );
             if( eig->nChannels != 1 )
-                CV_ERROR( CV_BadNumChannels, cvUnsupportedFormat );
+                CV_ERROR( CV_BadNumChannels, "Unsupported format" );
             if( i > 0 && eig_step != old_step )
                 CV_ERROR( CV_StsBadArg, "Different steps of objects" );
 
diff --git a/modules/legacy/src/em.cpp b/modules/legacy/src/em.cpp
index 200f743138..b49eb91316 100644
--- a/modules/legacy/src/em.cpp
+++ b/modules/legacy/src/em.cpp
@@ -85,13 +85,12 @@ void CvEM::read( CvFileStorage* fs, CvFileNode* node )
 
 void CvEM::write( CvFileStorage* _fs, const char* name ) const
 {
-    FileStorage fs = _fs;
+    FileStorage fs(_fs, false);
     if(name)
         fs << name << "{";
     emObj.write(fs);
     if(name)
         fs << "}";
-    fs.fs.obj = 0;
 }
 
 double CvEM::calcLikelihood( const Mat &input_sample ) const
@@ -139,15 +138,15 @@ void init_params(const CvEMParams& src,
                  Mat& prbs, Mat& weights,
                  Mat& means, std::vector<Mat>& covsHdrs)
 {
-    prbs = src.probs;
-    weights = src.weights;
-    means = src.means;
+    prbs = cv::cvarrToMat(src.probs);
+    weights = cv::cvarrToMat(src.weights);
+    means = cv::cvarrToMat(src.means);
 
     if(src.covs)
     {
         covsHdrs.resize(src.nclusters);
         for(size_t i = 0; i < covsHdrs.size(); i++)
-            covsHdrs[i] = src.covs[i];
+            covsHdrs[i] = cv::cvarrToMat(src.covs[i]);
     }
 }
 
diff --git a/modules/legacy/src/enteringblobdetection.cpp b/modules/legacy/src/enteringblobdetection.cpp
index 32a83bf0fa..d66a997a70 100644
--- a/modules/legacy/src/enteringblobdetection.cpp
+++ b/modules/legacy/src/enteringblobdetection.cpp
@@ -1060,5 +1060,3 @@ int CvBlobDetectorCC::DetectNewBlob(IplImage* /*pImg*/, IplImage* pFGMask, CvBlo
     return result;
 
 }   /* cvDetectNewBlob */
-
-
diff --git a/modules/legacy/src/enteringblobdetectionreal.cpp b/modules/legacy/src/enteringblobdetectionreal.cpp
index b0ec792a75..9458ab53ee 100644
--- a/modules/legacy/src/enteringblobdetectionreal.cpp
+++ b/modules/legacy/src/enteringblobdetectionreal.cpp
@@ -161,4 +161,3 @@ public:
 
 /* Blob detector constructor: */
 //CvBlobDetector* cvCreateBlobDetectorReal(CvTestSeq* pTestSeq){return new CvBlobDetectorReal(pTestSeq);}
-
diff --git a/modules/legacy/src/epilines.cpp b/modules/legacy/src/epilines.cpp
index 309ca0b1d5..8407e646bb 100644
--- a/modules/legacy/src/epilines.cpp
+++ b/modules/legacy/src/epilines.cpp
@@ -1826,7 +1826,7 @@ void icvGetCutPiece(   CvVect64d areaLineCoef1,CvVect64d areaLineCoef2,
     /* Collect all candidate point */
 
     CvPoint2D64d candPoints[8];
-    CvPoint2D64d midPoint = {0, 0};
+    CvPoint2D64d midPoint;
     int numPoints = 0;
     int res;
     int i;
@@ -2115,7 +2115,7 @@ CV_IMPL IplImage* icvCreateIsometricImage( IplImage* src, IplImage* dst,
     if( !dst || dst->depth != desired_depth ||
         dst->nChannels != desired_num_channels ||
         dst_size.width != src_size.width ||
-        dst_size.height != dst_size.height )
+        dst_size.height != src_size.height )
     {
         cvReleaseImage( &dst );
         dst = cvCreateImage( src_size, desired_depth, desired_num_channels );
@@ -3698,5 +3698,3 @@ int cvComputeEssentialMatrix(  CvMatr32f rotMatr,
 
     return CV_OK;
 }
-
-
diff --git a/modules/legacy/src/extendededges.cpp b/modules/legacy/src/extendededges.cpp
index 8ade446e1e..aa2488f832 100644
--- a/modules/legacy/src/extendededges.cpp
+++ b/modules/legacy/src/extendededges.cpp
@@ -263,7 +263,3 @@ CvSeq* icvCutContourRaster( CvSeq* current, CvMemStorage* storage, IplImage* ima
 
 */
 #endif /* WIN32 */
-
-
-
-
diff --git a/modules/legacy/src/face.cpp b/modules/legacy/src/face.cpp
index 7d2a5f13e1..b188a10def 100644
--- a/modules/legacy/src/face.cpp
+++ b/modules/legacy/src/face.cpp
@@ -351,4 +351,3 @@ inline void RFace::ResizeRect(CvRect Rect,CvRect * lpRect,long lDir,long lD)
     }
 
 }// inline void RFace::ResizeRect(CvRect * lpRect,long lDir,long lD)
-
diff --git a/modules/legacy/src/face.h b/modules/legacy/src/face.h
index e7bda1a0c6..74a96f8de5 100644
--- a/modules/legacy/src/face.h
+++ b/modules/legacy/src/face.h
@@ -133,4 +133,3 @@ private:
 
 
 #endif //__FACE_H__
-
diff --git a/modules/legacy/src/facedetection.cpp b/modules/legacy/src/facedetection.cpp
index 4487fdad15..b93d9aa6b2 100644
--- a/modules/legacy/src/facedetection.cpp
+++ b/modules/legacy/src/facedetection.cpp
@@ -499,5 +499,3 @@ Face * FaceDetectionList::GetData()
     m_pCurElem = m_pCurElem->m_pNext;
     return m_pCurElem->m_pFace;
 }//Face * FaceDetectionList::GetData()
-
-
diff --git a/modules/legacy/src/facetemplate.h b/modules/legacy/src/facetemplate.h
index b0ff47c2ce..31a3a8303c 100644
--- a/modules/legacy/src/facetemplate.h
+++ b/modules/legacy/src/facetemplate.h
@@ -199,4 +199,3 @@ inline MouthFaceTemplate::MouthFaceTemplate(long lNumber,CvRect rect,double dEye
 
 
 #endif//__FACETEMPLATE_H__
-
diff --git a/modules/legacy/src/features2d.cpp b/modules/legacy/src/features2d.cpp
index e985ec2263..f313875ac4 100644
--- a/modules/legacy/src/features2d.cpp
+++ b/modules/legacy/src/features2d.cpp
@@ -73,7 +73,7 @@ cvExtractSURF( const CvArr* _img, const CvArr* _mask,
     Mat descr;
 
     Ptr<Feature2D> surf = Algorithm::create<Feature2D>("Feature2D.SURF");
-    if( surf.empty() )
+    if( !surf )
         CV_Error(CV_StsNotImplemented, "OpenCV was built without SURF support");
 
     surf->set("hessianThreshold", params.hessianThreshold);
@@ -107,10 +107,10 @@ CV_IMPL CvSeq*
 cvGetStarKeypoints( const CvArr* _img, CvMemStorage* storage,
                     CvStarDetectorParams params )
 {
-    Ptr<StarDetector> star = new StarDetector(params.maxSize, params.responseThreshold,
-                                              params.lineThresholdProjected,
-                                              params.lineThresholdBinarized,
-                                              params.suppressNonmaxSize);
+    Ptr<StarDetector> star(new StarDetector(params.maxSize, params.responseThreshold,
+                                            params.lineThresholdProjected,
+                                            params.lineThresholdBinarized,
+                                            params.suppressNonmaxSize));
     std::vector<KeyPoint> kpts;
     star->detect(cvarrToMat(_img), kpts, Mat());
 
@@ -122,5 +122,3 @@ cvGetStarKeypoints( const CvArr* _img, CvMemStorage* storage,
     }
     return seq;
 }
-
-
diff --git a/modules/legacy/src/findface.cpp b/modules/legacy/src/findface.cpp
index 8190994819..a41c579d4e 100644
--- a/modules/legacy/src/findface.cpp
+++ b/modules/legacy/src/findface.cpp
@@ -65,4 +65,3 @@ CvSeq * cvPostBoostingFindFace(IplImage * Image,CvMemStorage* lpStorage)
 
     return lpSeq;
 }//cvPostBoostingFindFace(IplImage * Image)
-
diff --git a/modules/legacy/src/findhandregion.cpp b/modules/legacy/src/findhandregion.cpp
index 8a568833ef..23fc41935e 100644
--- a/modules/legacy/src/findhandregion.cpp
+++ b/modules/legacy/src/findhandregion.cpp
@@ -631,4 +631,3 @@ cvFindHandRegionA( CvPoint3D32f * points, int count,
     IPPI_CALL( icvFindHandRegionA( points, count, indexs, line, size, jc,
                                     center, storage, numbers ));
 }
-
diff --git a/modules/legacy/src/hmm.cpp b/modules/legacy/src/hmm.cpp
index d1af33618d..d1196feaca 100644
--- a/modules/legacy/src/hmm.cpp
+++ b/modules/legacy/src/hmm.cpp
@@ -1695,4 +1695,3 @@ cvMixSegmL2( CvImgObsInfo ** obs_info_array, int num_img, CvEHMM * hmm )
 }
 
 /* End of file */
-
diff --git a/modules/legacy/src/hmm1d.cpp b/modules/legacy/src/hmm1d.cpp
index 056bbf8c73..5daee8693b 100644
--- a/modules/legacy/src/hmm1d.cpp
+++ b/modules/legacy/src/hmm1d.cpp
@@ -1148,4 +1148,3 @@ CvStatus icvEstimate1DHMMStateParams(CvImgObsInfo** obs_info_array, int num_img,
 
 
 #endif
-
diff --git a/modules/legacy/src/image.cpp b/modules/legacy/src/image.cpp
index 3f691bc27f..a3fc8b9141 100644
--- a/modules/legacy/src/image.cpp
+++ b/modules/legacy/src/image.cpp
@@ -330,4 +330,3 @@ void CvMatrix::show( const char* window_name )
 
 
 /* End of file. */
-
diff --git a/modules/legacy/src/kdtree.cpp b/modules/legacy/src/kdtree.cpp
index 12a4acd692..4183950759 100644
--- a/modules/legacy/src/kdtree.cpp
+++ b/modules/legacy/src/kdtree.cpp
@@ -117,10 +117,10 @@ class CvKDTreeWrap : public CvFeatureTree {
            CvMat* results) {
     int rn = results->rows * results->cols;
     std::vector<int> inbounds;
-    dispatch_cvtype(mat, ((__treetype*)data)->
-        find_ortho_range((typename __treetype::scalar_type*)bounds_min->data.ptr,
+    assert(CV_MAT_DEPTH(mat->type) == CV_32F || CV_MAT_DEPTH(mat->type) == CV_64F);
+    ((__treetype*)data)->find_ortho_range((typename __treetype::scalar_type*)bounds_min->data.ptr,
              (typename __treetype::scalar_type*)bounds_max->data.ptr,
-             inbounds));
+             inbounds);
     std::copy(inbounds.begin(),
         inbounds.begin() + std::min((int)inbounds.size(), rn),
         (int*) results->data.ptr);
@@ -172,7 +172,7 @@ public:
       CV_Error(CV_StsUnsupportedFormat, "dist must be CV_64FC1");
 
     if (CV_MAT_TYPE(type()) != CV_MAT_TYPE(desc->type)) {
-      tmp_desc = cvCreateMat(desc->rows, desc->cols, type());
+      tmp_desc.reset(cvCreateMat(desc->rows, desc->cols, type()));
       cvConvert(desc, tmp_desc);
       desc = tmp_desc;
     }
diff --git a/modules/legacy/src/lcm.cpp b/modules/legacy/src/lcm.cpp
index 58abcdb7dc..95ab2c6d6a 100644
--- a/modules/legacy/src/lcm.cpp
+++ b/modules/legacy/src/lcm.cpp
@@ -694,7 +694,7 @@ void _cvProjectionPointToSegment(CvPoint2D32f* PointO,
                                  float* dist)
 {
     float scal_AO_AB, scal_AB_AB;
-    CvPoint2D32f VectorAB = {PointB->x - PointA->x, PointB->y - PointA->y};
+    CvPoint2D32f VectorAB(PointB->x - PointA->x, PointB->y - PointA->y);
     scal_AB_AB = VectorAB.x*VectorAB.x + VectorAB.y*VectorAB.y;
     if(scal_AB_AB < LCM_CONST_ZERO)
     {
@@ -704,7 +704,7 @@ void _cvProjectionPointToSegment(CvPoint2D32f* PointO,
         return;
     }
 
-    CvPoint2D32f VectorAO = {PointO->x - PointA->x, PointO->y - PointA->y};
+    CvPoint2D32f VectorAO(PointO->x - PointA->x, PointO->y - PointA->y);
     scal_AO_AB = VectorAO.x*VectorAB.x + VectorAO.y*VectorAB.y;
 
     if(dist)
@@ -718,7 +718,3 @@ void _cvProjectionPointToSegment(CvPoint2D32f* PointO,
     PrPoint->y = PointO->y - VectorAO.y + alfa*VectorAB.y;
     return;
 }//end of _cvProjectionPointToSegment
-
-
-
-
diff --git a/modules/legacy/src/levmar.cpp b/modules/legacy/src/levmar.cpp
index aa2514e335..9b16b5a0fb 100644
--- a/modules/legacy/src/levmar.cpp
+++ b/modules/legacy/src/levmar.cpp
@@ -315,6 +315,3 @@ void TestOptimX2Y2()
 }
 
 #endif
-
-
-
diff --git a/modules/legacy/src/levmarprojbandle.cpp b/modules/legacy/src/levmarprojbandle.cpp
index 162ea528c4..545a96ef3e 100644
--- a/modules/legacy/src/levmarprojbandle.cpp
+++ b/modules/legacy/src/levmarprojbandle.cpp
@@ -41,11 +41,8 @@
 
 
 #include "precomp.hpp"
-//#include "cvtypes.h"
 #include <float.h>
 #include <limits.h>
-//#include "cv.h"
-
 #include <stdio.h>
 
 void icvReconstructPoints4DStatus(CvMat** projPoints, CvMat **projMatrs, CvMat** presPoints, CvMat *points4D,int numImages,CvMat **projError=0);
@@ -745,6 +742,7 @@ void icvReconstructPoints4DStatus(CvMat** projPoints, CvMat **projMatrs, CvMat**
     double* matrW_dat = 0;
 
     CV_FUNCNAME( "icvReconstructPoints4DStatus" );
+
     __BEGIN__;
 
     /* ----- Test input params for errors ----- */
@@ -770,6 +768,7 @@ void icvReconstructPoints4DStatus(CvMat** projPoints, CvMat **projMatrs, CvMat**
         CV_ERROR( CV_StsOutOfRange, "Points must have 4 cordinates" );
     }
 
+
     /* !!! Not tested all input parameters */
     /* ----- End test ----- */
 
@@ -778,75 +777,75 @@ void icvReconstructPoints4DStatus(CvMat** projPoints, CvMat **projMatrs, CvMat**
 
     /* Allocate maximum data */
 
-
-    CvMat matrV;
-    double matrV_dat[4*4];
-    matrV = cvMat(4,4,CV_64F,matrV_dat);
-
-    CV_CALL(matrA_dat = (double*)cvAlloc(3*numImages * 4 * sizeof(double)));
-    CV_CALL(matrW_dat = (double*)cvAlloc(3*numImages * 4 * sizeof(double)));
-
-    /* reconstruct each point */
-    for( currPoint = 0; currPoint < numPoints; currPoint++ )
     {
-        /* Reconstruct current point */
-        /* Define number of visible projections */
-        int numVisProj = 0;
-        for( currImage = 0; currImage < numImages; currImage++ )
+        double matrV_dat[4*4];
+        CvMat matrV = cvMat(4,4,CV_64F,matrV_dat);
+
+        CV_CALL(matrA_dat = (double*)cvAlloc(3*numImages * 4 * sizeof(double)));
+        CV_CALL(matrW_dat = (double*)cvAlloc(3*numImages * 4 * sizeof(double)));
+
+        /* reconstruct each point */
+        for( currPoint = 0; currPoint < numPoints; currPoint++ )
         {
-            if( cvmGet(presPoints[currImage],0,currPoint) > 0 )
+            /* Reconstruct current point */
+            /* Define number of visible projections */
+            int numVisProj = 0;
+            for( currImage = 0; currImage < numImages; currImage++ )
             {
-                numVisProj++;
-            }
-        }
-
-        if( numVisProj < 2 )
-        {
-            /* This point can't be reconstructed */
-            continue;
-        }
-
-        /* Allocate memory and create matrices */
-        CvMat matrA;
-        matrA = cvMat(3*numVisProj,4,CV_64F,matrA_dat);
-
-        CvMat matrW;
-        matrW = cvMat(3*numVisProj,4,CV_64F,matrW_dat);
-
-        int currVisProj = 0;
-        for( currImage = 0; currImage < numImages; currImage++ )/* For each view */
-        {
-            if( cvmGet(presPoints[currImage],0,currPoint) > 0 )
-            {
-                double x,y;
-                x = cvmGet(projPoints[currImage],0,currPoint);
-                y = cvmGet(projPoints[currImage],1,currPoint);
-                for( int k = 0; k < 4; k++ )
+                if( cvmGet(presPoints[currImage],0,currPoint) > 0 )
                 {
-                    matrA_dat[currVisProj*12   + k] =
-                           x * cvmGet(projMatrs[currImage],2,k) -     cvmGet(projMatrs[currImage],0,k);
-
-                    matrA_dat[currVisProj*12+4 + k] =
-                           y * cvmGet(projMatrs[currImage],2,k) -     cvmGet(projMatrs[currImage],1,k);
-
-                    matrA_dat[currVisProj*12+8 + k] =
-                           x * cvmGet(projMatrs[currImage],1,k) - y * cvmGet(projMatrs[currImage],0,k);
+                    numVisProj++;
                 }
-                currVisProj++;
             }
+
+            if( numVisProj < 2 )
+            {
+                /* This point can't be reconstructed */
+                continue;
+            }
+
+            /* Allocate memory and create matrices */
+            CvMat matrA;
+            matrA = cvMat(3*numVisProj,4,CV_64F,matrA_dat);
+
+            CvMat matrW;
+            matrW = cvMat(3*numVisProj,4,CV_64F,matrW_dat);
+
+            int currVisProj = 0;
+            for( currImage = 0; currImage < numImages; currImage++ )/* For each view */
+            {
+                if( cvmGet(presPoints[currImage],0,currPoint) > 0 )
+                {
+                    double x,y;
+                    x = cvmGet(projPoints[currImage],0,currPoint);
+                    y = cvmGet(projPoints[currImage],1,currPoint);
+                    for( int k = 0; k < 4; k++ )
+                    {
+                        matrA_dat[currVisProj*12   + k] =
+                               x * cvmGet(projMatrs[currImage],2,k) -     cvmGet(projMatrs[currImage],0,k);
+
+                        matrA_dat[currVisProj*12+4 + k] =
+                               y * cvmGet(projMatrs[currImage],2,k) -     cvmGet(projMatrs[currImage],1,k);
+
+                        matrA_dat[currVisProj*12+8 + k] =
+                               x * cvmGet(projMatrs[currImage],1,k) - y * cvmGet(projMatrs[currImage],0,k);
+                    }
+                    currVisProj++;
+                }
+            }
+
+            /* Solve system for current point */
+            {
+                cvSVD(&matrA,&matrW,0,&matrV,CV_SVD_V_T);
+
+                /* Copy computed point */
+                cvmSet(points4D,0,currPoint,cvmGet(&matrV,3,0));//X
+                cvmSet(points4D,1,currPoint,cvmGet(&matrV,3,1));//Y
+                cvmSet(points4D,2,currPoint,cvmGet(&matrV,3,2));//Z
+                cvmSet(points4D,3,currPoint,cvmGet(&matrV,3,3));//W
+            }
+
         }
-
-        /* Solve system for current point */
-        {
-            cvSVD(&matrA,&matrW,0,&matrV,CV_SVD_V_T);
-
-            /* Copy computed point */
-            cvmSet(points4D,0,currPoint,cvmGet(&matrV,3,0));//X
-            cvmSet(points4D,1,currPoint,cvmGet(&matrV,3,1));//Y
-            cvmSet(points4D,2,currPoint,cvmGet(&matrV,3,2));//Z
-            cvmSet(points4D,3,currPoint,cvmGet(&matrV,3,3));//W
-        }
-
     }
 
     {/* Compute projection error */
@@ -913,7 +912,7 @@ static void icvProjPointsStatusFunc( int numImages, CvMat *points4D, CvMat **pro
     {
         CV_ERROR( CV_StsNullPtr, "Some of parameters is a NULL pointer" );
     }
-
+    {
     int numPoints;
     numPoints = points4D->cols;
     if( numPoints < 1 )
@@ -994,7 +993,7 @@ static void icvProjPointsStatusFunc( int numImages, CvMat *points4D, CvMat **pro
             }
         }
     }
-
+    }
     __END__;
 }
 
diff --git a/modules/legacy/src/lines.cpp b/modules/legacy/src/lines.cpp
index d0e243dfb4..48eaf797d8 100644
--- a/modules/legacy/src/lines.cpp
+++ b/modules/legacy/src/lines.cpp
@@ -477,4 +477,3 @@ cvPostWarpImage( int numLines,  /* number of scanlines  */
 }
 
 /* End of file */
-
diff --git a/modules/legacy/src/lmeds.cpp b/modules/legacy/src/lmeds.cpp
index 42272345c3..7794640769 100644
--- a/modules/legacy/src/lmeds.cpp
+++ b/modules/legacy/src/lmeds.cpp
@@ -1683,4 +1683,3 @@ icvPoint7( int *ml, int *mr, double *F, int *amount )
 
     return error;
 }                               /* icvPoint7 */
-
diff --git a/modules/legacy/src/morphcontours.cpp b/modules/legacy/src/morphcontours.cpp
index 851b00bbe8..b04cdd48ad 100644
--- a/modules/legacy/src/morphcontours.cpp
+++ b/modules/legacy/src/morphcontours.cpp
@@ -78,7 +78,7 @@ void _cvWorkEast     (int i, int j, _CvWork** W, CvPoint2D32f* edges1, CvPoint2D
 void _cvWorkSouthEast(int i, int j, _CvWork** W, CvPoint2D32f* edges1, CvPoint2D32f* edges2);
 void _cvWorkSouth    (int i, int j, _CvWork** W, CvPoint2D32f* edges1, CvPoint2D32f* edges2);
 
-static CvPoint2D32f null_edge = {0,0};
+static CvPoint2D32f null_edge;
 
 double _cvStretchingWork(CvPoint2D32f* P1,
                          CvPoint2D32f* P2)
@@ -106,7 +106,7 @@ double _cvBendingWork(  CvPoint2D32f* B0,
                         CvPoint* K*/)
 {
     CvPoint2D32f Q0, Q1, Q2;
-    CvPoint2D32f Q1_nm = { 0, 0 }, Q2_nm = { 0, 0 };
+    CvPoint2D32f Q1_nm, Q2_nm;
     double d0, d1, d2, des, t_zero;
     double k_zero, k_nonmon;
     CvPoint2D32f center;
diff --git a/modules/legacy/src/morphing.cpp b/modules/legacy/src/morphing.cpp
index 44451414db..ab85b3b939 100644
--- a/modules/legacy/src/morphing.cpp
+++ b/modules/legacy/src/morphing.cpp
@@ -389,4 +389,3 @@ cvMorphEpilinesMulti( int lines,        /* number of lines             */
                                           second_corr   /* correspond information for the 2nd seq     */
                 ));
 }
-
diff --git a/modules/legacy/src/oneway.cpp b/modules/legacy/src/oneway.cpp
index eae17a5d47..43ded85939 100644
--- a/modules/legacy/src/oneway.cpp
+++ b/modules/legacy/src/oneway.cpp
@@ -465,7 +465,7 @@ namespace cv{
 
     void OneWayDescriptor::Initialize(int pose_count, IplImage* frontal, const char* feature_name, int norm)
     {
-        m_feature_name = std::string(feature_name);
+        m_feature_name = String(feature_name);
         CvRect roi = cvGetImageROI(frontal);
         m_center = rect_center(roi);
 
@@ -482,7 +482,7 @@ namespace cv{
             Initialize(pose_count, frontal, feature_name, 1);
             return;
         }
-        m_feature_name = std::string(feature_name);
+        m_feature_name = String(feature_name);
         CvRect roi = cvGetImageROI(frontal);
         m_center = rect_center(roi);
 
@@ -529,7 +529,7 @@ namespace cv{
             }
             return;
         }
-        CvRect roi={0,0,0,0};
+        CvRect roi;
         if (!CV_IS_MAT(patch))
         {
             roi = cvGetImageROI((IplImage*)patch);
@@ -669,7 +669,7 @@ namespace cv{
             cvConvertScale(m_samples[i], patch, 255/maxval);
 
 #ifdef HAVE_OPENCV_HIGHGUI
-            cvSaveImage(buf, patch);
+            cv::imwrite(buf, cv::cvarrToMat(patch));
 #else
             CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without image I/O support");
 #endif
@@ -1287,8 +1287,8 @@ namespace cv{
 
     }
 
-    OneWayDescriptorBase::OneWayDescriptorBase(CvSize patch_size, int pose_count, const std::string &pca_filename,
-                                               const std::string &train_path, const std::string &images_list, float _scale_min, float _scale_max,
+    OneWayDescriptorBase::OneWayDescriptorBase(CvSize patch_size, int pose_count, const String &pca_filename,
+                                               const String &train_path, const String &images_list, float _scale_min, float _scale_max,
                                                float _scale_step, int pyr_levels,
                                                int pca_dim_high, int pca_dim_low)
     : m_pca_dim_high(pca_dim_high), m_pca_dim_low(pca_dim_low), scale_min(_scale_min), scale_max(_scale_max), scale_step(_scale_step)
@@ -1736,11 +1736,11 @@ namespace cv{
     {
         std::vector<KeyPoint> features;
         Ptr<FeatureDetector> surf_extractor = FeatureDetector::create("SURF");
-        if( surf_extractor.empty() )
+        if( !surf_extractor )
             CV_Error(CV_StsNotImplemented, "OpenCV was built without SURF support");
         surf_extractor->set("hessianThreshold", 1.0);
         //printf("Extracting SURF features...");
-        surf_extractor->detect(Mat(img), features);
+        surf_extractor->detect(cv::cvarrToMat(img), features);
         //printf("done\n");
 
         for (int j = 0; j < (int)features.size(); j++)
@@ -1801,17 +1801,16 @@ namespace cv{
             sprintf(filename, "%s/%s", path, imagename);
 
             //printf("Reading image %s...", filename);
-            IplImage* img = 0;
+            IplImage img;
 #ifdef HAVE_OPENCV_HIGHGUI
-            img = cvLoadImage(filename, CV_LOAD_IMAGE_GRAYSCALE);
+            Mat img2 = cv::imread(filename, IMREAD_GRAYSCALE);
+            img = img2;
 #else
             CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without image I/O support");
 #endif
             //printf("done\n");
 
-            extractPatches (img, patches, patch_size);
-
-            cvReleaseImage(&img);
+            extractPatches (&img, patches, patch_size);
         }
         fclose(pFile);
     }
@@ -2080,8 +2079,8 @@ namespace cv{
         m_part_id = 0;
     }
 
-    OneWayDescriptorObject::OneWayDescriptorObject(CvSize patch_size, int pose_count, const std::string &pca_filename,
-                                                   const std::string &train_path, const std::string &images_list, float _scale_min, float _scale_max, float _scale_step, int pyr_levels) :
+    OneWayDescriptorObject::OneWayDescriptorObject(CvSize patch_size, int pose_count, const String &pca_filename,
+                                                   const String &train_path, const String &images_list, float _scale_min, float _scale_max, float _scale_step, int pyr_levels) :
     OneWayDescriptorBase(patch_size, pose_count, pca_filename, train_path, images_list, _scale_min, _scale_max, _scale_step, pyr_levels)
     {
         m_part_id = 0;
@@ -2145,7 +2144,7 @@ namespace cv{
 
     void readPCAFeatures(const FileNode &fn, CvMat** avg, CvMat** eigenvectors, const char* postfix)
     {
-        std::string str = std::string ("avg") + postfix;
+        String str = String ("avg") + postfix;
         CvMat* _avg = reinterpret_cast<CvMat*> (fn[str].readObj());
         if (_avg != 0)
         {
@@ -2153,7 +2152,7 @@ namespace cv{
             cvReleaseMat(&_avg);
         }
 
-        str = std::string ("eigenvectors") + postfix;
+        str = String ("eigenvectors") + postfix;
         CvMat* _eigenvectors = reinterpret_cast<CvMat*> (fn[str].readObj());
         if (_eigenvectors != 0)
         {
@@ -2166,8 +2165,8 @@ namespace cv{
      *                                OneWayDescriptorMatcher                                  *
      \****************************************************************************************/
 
-    OneWayDescriptorMatcher::Params::Params( int _poseCount, Size _patchSize, std::string _pcaFilename,
-                                            std::string _trainPath, std::string _trainImagesList,
+    OneWayDescriptorMatcher::Params::Params( int _poseCount, Size _patchSize, String _pcaFilename,
+                                            String _trainPath, String _trainImagesList,
                                             float _minScale, float _maxScale, float _stepScale ) :
     poseCount(_poseCount), patchSize(_patchSize), pcaFilename(_pcaFilename),
     trainPath(_trainPath), trainImagesList(_trainImagesList),
@@ -2187,7 +2186,7 @@ namespace cv{
     {
         clear();
 
-        if( _base.empty() )
+        if( !_base )
             base = _base;
 
         params = _params;
@@ -2198,16 +2197,17 @@ namespace cv{
         GenericDescriptorMatcher::clear();
 
         prevTrainCount = 0;
-        if( !base.empty() )
+        if( base )
             base->clear();
     }
 
     void OneWayDescriptorMatcher::train()
     {
-        if( base.empty() || prevTrainCount < (int)trainPointCollection.keypointCount() )
+        if( !base || prevTrainCount < (int)trainPointCollection.keypointCount() )
         {
-            base = new OneWayDescriptorObject( params.patchSize, params.poseCount, params.pcaFilename,
-                                              params.trainPath, params.trainImagesList, params.minScale, params.maxScale, params.stepScale );
+            base.reset(
+                new OneWayDescriptorObject( params.patchSize, params.poseCount, params.pcaFilename,
+                                            params.trainPath, params.trainImagesList, params.minScale, params.maxScale, params.stepScale ));
 
             base->Allocate( (int)trainPointCollection.keypointCount() );
             prevTrainCount = (int)trainPointCollection.keypointCount();
@@ -2271,8 +2271,9 @@ namespace cv{
 
     void OneWayDescriptorMatcher::read( const FileNode &fn )
     {
-        base = new OneWayDescriptorObject( params.patchSize, params.poseCount, std::string (), std::string (), std::string (),
-                                          params.minScale, params.maxScale, params.stepScale );
+        base.reset(
+            new OneWayDescriptorObject( params.patchSize, params.poseCount, String (), String (), String (),
+                                        params.minScale, params.maxScale, params.stepScale ));
         base->Read (fn);
     }
 
@@ -2283,12 +2284,12 @@ namespace cv{
 
     bool OneWayDescriptorMatcher::empty() const
     {
-        return base.empty() || base->empty();
+        return !base || base->empty();
     }
 
     Ptr<GenericDescriptorMatcher> OneWayDescriptorMatcher::clone( bool emptyTrainData ) const
     {
-        OneWayDescriptorMatcher* matcher = new OneWayDescriptorMatcher( params );
+        Ptr<OneWayDescriptorMatcher> matcher = makePtr<OneWayDescriptorMatcher>( params );
 
         if( !emptyTrainData )
         {
diff --git a/modules/legacy/src/optflowbm.cpp b/modules/legacy/src/optflowbm.cpp
index afab0c1fe8..91b55cd584 100644
--- a/modules/legacy/src/optflowbm.cpp
+++ b/modules/legacy/src/optflowbm.cpp
@@ -75,11 +75,10 @@ cvCalcOpticalFlowBM( const void* srcarrA, const void* srcarrB,
     if( !CV_ARE_TYPES_EQ( velx, vely ))
         CV_Error( CV_StsUnmatchedFormats, "Destination images have different formats" );
 
-    CvSize velSize =
-    {
+    CvSize velSize(
         (srcA->width - blockSize.width + shiftSize.width)/shiftSize.width,
         (srcA->height - blockSize.height + shiftSize.height)/shiftSize.height
-    };
+    );
 
     if( !CV_ARE_SIZES_EQ( srcA, srcB ) ||
         !CV_ARE_SIZES_EQ( velx, vely ) ||
diff --git a/modules/legacy/src/planardetect.cpp b/modules/legacy/src/planardetect.cpp
index 6c5e96ff8c..fa9152d474 100644
--- a/modules/legacy/src/planardetect.cpp
+++ b/modules/legacy/src/planardetect.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d.hpp"
 #include <stdio.h>
 
 namespace cv
@@ -389,7 +390,7 @@ void LDetector::getMostStable2D(const Mat& image, std::vector<KeyPoint>& keypoin
 
     if( (int)keypoints.size() > maxPoints )
     {
-        sort(keypoints, CmpKeypointScores());
+        std::sort(keypoints.begin(), keypoints.end(), CmpKeypointScores());
         keypoints.resize(maxPoints);
     }
 }
@@ -602,7 +603,7 @@ void LDetector::operator()(const std::vector<Mat>& pyr, std::vector<KeyPoint>& k
 
     if( maxCount > 0 && keypoints.size() > (size_t)maxCount )
     {
-        sort(keypoints, CmpKeypointScores());
+        std::sort(keypoints.begin(), keypoints.end(), CmpKeypointScores());
         keypoints.resize(maxCount);
     }
 }
@@ -617,9 +618,9 @@ void LDetector::read(const FileNode& objnode)
     clusteringDistance = (int)objnode["clustering-distance"];
 }
 
-void LDetector::write(FileStorage& fs, const std::string& name) const
+void LDetector::write(FileStorage& fs, const String& name) const
 {
-    WriteStructContext ws(fs, name, CV_NODE_MAP);
+    internal::WriteStructContext ws(fs, name, CV_NODE_MAP);
 
     fs << "radius" << radius
     << "threshold" << threshold
@@ -707,9 +708,9 @@ FernClassifier::FernClassifier(const std::vector<std::vector<Point2f> >& points,
 }
 
 
-void FernClassifier::write(FileStorage& fs, const std::string& objname) const
+void FernClassifier::write(FileStorage& fs, const String& objname) const
 {
-    WriteStructContext ws(fs, objname, CV_NODE_MAP);
+    internal::WriteStructContext ws(fs, objname, CV_NODE_MAP);
 
     cv::write(fs, "nstructs", nstructs);
     cv::write(fs, "struct-size", structSize);
@@ -718,7 +719,7 @@ void FernClassifier::write(FileStorage& fs, const std::string& objname) const
     cv::write(fs, "compression-method", compressionMethod);
     cv::write(fs, "patch-size", patchSize.width);
     {
-        WriteStructContext wsf(fs, "features", CV_NODE_SEQ + CV_NODE_FLOW);
+        internal::WriteStructContext wsf(fs, "features", CV_NODE_SEQ + CV_NODE_FLOW);
         int i, nfeatures = (int)features.size();
         for( i = 0; i < nfeatures; i++ )
         {
@@ -727,7 +728,7 @@ void FernClassifier::write(FileStorage& fs, const std::string& objname) const
         }
     }
     {
-        WriteStructContext wsp(fs, "posteriors", CV_NODE_SEQ + CV_NODE_FLOW);
+        internal::WriteStructContext wsp(fs, "posteriors", CV_NODE_SEQ + CV_NODE_FLOW);
         cv::write(fs, posteriors);
     }
 }
@@ -1228,7 +1229,7 @@ nstructs(_nstructs), structSize(_structSize), nviews(_nviews),
 compressionMethod(_compressionMethod), patchGenerator(_patchGenerator)
 {}
 
-FernDescriptorMatcher::Params::Params( const std::string& _filename )
+FernDescriptorMatcher::Params::Params( const String& _filename )
 {
     filename = _filename;
 }
@@ -1239,7 +1240,7 @@ FernDescriptorMatcher::FernDescriptorMatcher( const Params& _params )
     params = _params;
     if( !params.filename.empty() )
     {
-        classifier = new FernClassifier;
+        classifier = makePtr<FernClassifier>();
         FileStorage fs(params.filename, FileStorage::READ);
         if( fs.isOpened() )
             classifier->read( fs.getFirstTopLevelNode() );
@@ -1259,7 +1260,7 @@ void FernDescriptorMatcher::clear()
 
 void FernDescriptorMatcher::train()
 {
-    if( classifier.empty() || prevTrainCount < (int)trainPointCollection.keypointCount() )
+    if( !classifier || prevTrainCount < (int)trainPointCollection.keypointCount() )
     {
         assert( params.filename.empty() );
 
@@ -1267,9 +1268,10 @@ void FernDescriptorMatcher::train()
         for( size_t imgIdx = 0; imgIdx < trainPointCollection.imageCount(); imgIdx++ )
             KeyPoint::convert( trainPointCollection.getKeypoints((int)imgIdx), points[imgIdx] );
 
-        classifier = new FernClassifier( points, trainPointCollection.getImages(), std::vector<std::vector<int> >(), 0, // each points is a class
-                                        params.patchSize, params.signatureSize, params.nstructs, params.structSize,
-                                        params.nviews, params.compressionMethod, params.patchGenerator );
+        classifier.reset(
+            new FernClassifier( points, trainPointCollection.getImages(), std::vector<std::vector<int> >(), 0, // each points is a class
+                                params.patchSize, params.signatureSize, params.nstructs, params.structSize,
+                                params.nviews, params.compressionMethod, params.patchGenerator ));
     }
 }
 
@@ -1383,12 +1385,12 @@ void FernDescriptorMatcher::write( FileStorage& fs ) const
 
 bool FernDescriptorMatcher::empty() const
 {
-    return classifier.empty() || classifier->empty();
+    return !classifier || classifier->empty();
 }
 
 Ptr<GenericDescriptorMatcher> FernDescriptorMatcher::clone( bool emptyTrainData ) const
 {
-    FernDescriptorMatcher* matcher = new FernDescriptorMatcher( params );
+    Ptr<FernDescriptorMatcher> matcher = makePtr<FernDescriptorMatcher>( params );
     if( !emptyTrainData )
     {
         CV_Error( CV_StsNotImplemented, "deep clone dunctionality is not implemented, because "
@@ -1476,12 +1478,12 @@ void PlanarObjectDetector::read(const FileNode& node)
 }
 
 
-void PlanarObjectDetector::write(FileStorage& fs, const std::string& objname) const
+void PlanarObjectDetector::write(FileStorage& fs, const String& objname) const
 {
-    WriteStructContext ws(fs, objname, CV_NODE_MAP);
+    internal::WriteStructContext ws(fs, objname, CV_NODE_MAP);
 
     {
-        WriteStructContext wsroi(fs, "model-roi", CV_NODE_SEQ + CV_NODE_FLOW);
+        internal::WriteStructContext wsroi(fs, "model-roi", CV_NODE_SEQ + CV_NODE_FLOW);
         cv::write(fs, modelROI.x);
         cv::write(fs, modelROI.y);
         cv::write(fs, modelROI.width);
diff --git a/modules/legacy/src/precomp.cpp b/modules/legacy/src/precomp.cpp
deleted file mode 100644
index a9477b85ba..0000000000
--- a/modules/legacy/src/precomp.cpp
+++ /dev/null
@@ -1,42 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
diff --git a/modules/legacy/src/precomp.hpp b/modules/legacy/src/precomp.hpp
index 6e40e3ffd0..7093f9fa90 100644
--- a/modules/legacy/src/precomp.hpp
+++ b/modules/legacy/src/precomp.hpp
@@ -41,22 +41,21 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/legacy.hpp"
 
-#include "opencv2/core/internal.hpp"
 #include "opencv2/video.hpp"
-
 #include "opencv2/legacy/blobtrack.hpp"
 #include "opencv2/legacy/compat.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/core/private.hpp"
+
+#define __BEGIN__ __CV_BEGIN__
+#define __END__ __CV_END__
+#define EXIT __CV_EXIT__
 
 #include "_matrix.h"
 
-typedef unsigned short ushort;
-
 CV_INLINE bool operator == (CvSize size1, CvSize size2 );
 CV_INLINE bool operator == (CvSize size1, CvSize size2 )
 {
diff --git a/modules/legacy/src/pyrsegmentation.cpp b/modules/legacy/src/pyrsegmentation.cpp
index c64b7bcde0..03a4678a95 100644
--- a/modules/legacy/src/pyrsegmentation.cpp
+++ b/modules/legacy/src/pyrsegmentation.cpp
@@ -287,7 +287,7 @@ icvPyrSegmentation8uC1R( uchar * src_image, int src_step,
     /* calculate initial pyramid */
     for( l = 1; l <= level; l++ )
     {
-        CvSize dst_size = { size.width/2+1, size.height/2+1 };
+        CvSize dst_size(size.width/2+1, size.height/2+1);
         CvMat prev_level = cvMat( size.height, size.width, CV_32FC1 );
         CvMat next_level = cvMat( dst_size.height, dst_size.width, CV_32FC1 );
 
@@ -706,7 +706,7 @@ icvPyrSegmentation8uC3R( uchar * src_image, int src_step,
     /* calculate initial pyramid */
     for( l = 1; l <= level; l++ )
     {
-        CvSize dst_size = { size.width/2 + 1, size.height/2 + 1 };
+        CvSize dst_size(size.width/2 + 1, size.height/2 + 1);
         CvMat prev_level = cvMat( size.height, size.width, CV_32FC3 );
         CvMat next_level = cvMat( dst_size.height, dst_size.width, CV_32FC3 );
 
@@ -1843,7 +1843,7 @@ cvPyrSegmentation( IplImage * src,
     int thresh2 = cvRound( threshold2 );
 
     if( src->depth != IPL_DEPTH_8U )
-        CV_Error( CV_BadDepth, cvUnsupportedFormat );
+        CV_Error( CV_BadDepth, "Unsupported format" );
 
     if( src->depth != dst->depth || src->nChannels != dst->nChannels )
         CV_Error( CV_StsBadArg, "src and dst have different formats" );
@@ -1872,7 +1872,7 @@ cvPyrSegmentation( IplImage * src,
                                             comp, storage, level, thresh1, thresh2 ));
         break;
     default:
-        CV_Error( CV_BadNumChannels, cvUnsupportedFormat );
+        CV_Error( CV_BadNumChannels, "Unsupported format" );
     }
 }
 
diff --git a/modules/legacy/src/segment.cpp b/modules/legacy/src/segment.cpp
index 07e6082988..e61eeadf42 100644
--- a/modules/legacy/src/segment.cpp
+++ b/modules/legacy/src/segment.cpp
@@ -328,8 +328,8 @@ static void color_derv( const CvArr* srcArr, CvArr* dstArr, int thresh )
 }
 #endif
 
-const CvPoint icvCodeDeltas[8] =
-    { {1, 0}, {1, -1}, {0, -1}, {-1, -1}, {-1, 0}, {-1, 1}, {0, 1}, {1, 1} };
+static const CvPoint icvCodeDeltas[8] =
+    { CvPoint(1, 0), CvPoint(1, -1), CvPoint(0, -1), CvPoint(-1, -1), CvPoint(-1, 0), CvPoint(-1, 1), CvPoint(0, 1), CvPoint(1, 1) };
 
 static CvSeq*
 icvGetComponent( uchar* img, int step, CvRect rect,
@@ -366,7 +366,7 @@ icvGetComponent( uchar* img, int step, CvRect rect,
                 CvSeqWriter  writer;
                 char  *i0, *i1, *i3, *i4 = 0;
                 int  prev_s = -1, s, s_end;
-                CvPoint pt = { x, y };
+                CvPoint pt(x, y);
 
                 if( !(prev == 0 && p == 2) )    /* if not external contour */
                 {
diff --git a/modules/legacy/src/snakes.cpp b/modules/legacy/src/snakes.cpp
index 4c1a553389..68c1bc76a0 100644
--- a/modules/legacy/src/snakes.cpp
+++ b/modules/legacy/src/snakes.cpp
@@ -292,8 +292,8 @@ icvSnake8uC1R( unsigned char *src,
                             int leftshift = x ? 1 : 0;
                             int bottomshift = MIN( 1, roi.height - (y + 1)*WTILE_SIZE );
                             int rightshift = MIN( 1, roi.width - (x + 1)*WTILE_SIZE );
-                            CvRect g_roi = { x*WTILE_SIZE - leftshift, y*WTILE_SIZE - upshift,
-                                leftshift + WTILE_SIZE + rightshift, upshift + WTILE_SIZE + bottomshift };
+                            CvRect g_roi(x*WTILE_SIZE - leftshift, y*WTILE_SIZE - upshift,
+                                leftshift + WTILE_SIZE + rightshift, upshift + WTILE_SIZE + bottomshift);
                             CvMat _src1;
                             cvGetSubArr( &_src, &_src1, g_roi );
 
@@ -416,7 +416,7 @@ cvSnakeImage( const IplImage* src, CvPoint* points,
         CV_Error( CV_BadNumChannels, "input image has more than one channel" );
 
     if( src->depth != IPL_DEPTH_8U )
-        CV_Error( CV_BadDepth, cvUnsupportedFormat );
+        CV_Error( CV_BadDepth, "Unsupported format" );
 
     cvGetRawData( src, &data, &step, &size );
 
diff --git a/modules/legacy/src/subdiv2.cpp b/modules/legacy/src/subdiv2.cpp
index fb48e05a92..b2fbcf6223 100644
--- a/modules/legacy/src/subdiv2.cpp
+++ b/modules/legacy/src/subdiv2.cpp
@@ -833,7 +833,7 @@ draw_subdiv_facet( CvSubdiv2D * subdiv, IplImage * dst, IplImage * src, CvSubdiv
     {
         CvSubdiv2DPoint *pt = cvSubdiv2DEdgeDst( cvSubdiv2DRotateEdge( edge, 1 ));
         CvPoint ip = cvPoint( cvRound( pt->pt.x ), cvRound( pt->pt.y ));
-        CvScalar color = {{0,0,0,0}};
+        CvScalar color(0);
 
         //printf("count = %d, (%d,%d)\n", ip.x, ip.y );
 
diff --git a/modules/legacy/src/testseq.cpp b/modules/legacy/src/testseq.cpp
index d3521a5ae2..af4da33840 100644
--- a/modules/legacy/src/testseq.cpp
+++ b/modules/legacy/src/testseq.cpp
@@ -850,13 +850,13 @@ CvTestSeq* cvCreateTestSeq(char* pConfigfile, char** videos, int numvideo, float
     {   /* Calculate elements and image size and video length: */
         CvTestSeqElem*  p = pTS->pElemList;
         int             num = 0;
-        CvSize          MaxSize = {0,0};
+        CvSize          MaxSize;
         int             MaxFN = 0;
 
         for(p = pTS->pElemList; p; p=p->next, num++)
         {
             int     FN = p->FrameBegin+p->FrameNum;
-            CvSize  S = {0,0};
+            CvSize  S;
 
             if(p->pImg && p->BG)
             {
@@ -927,7 +927,7 @@ IplImage* cvTestSeqQueryFrame(CvTestSeq* pTestSeq)
 
     for(p=pTS->pElemList; p; p=p->next)
     {
-        int             DirectCopy = FALSE;
+        int             DirectCopy = 0;
         int             frame = pTS->CurFrame - p->FrameBegin;
         //float           t = p->FrameNum>1?((float)frame/(p->FrameNum-1)):0;
         CvTSTrans*      pTrans = p->pTrans + frame%p->TransNum;
@@ -996,13 +996,13 @@ IplImage* cvTestSeqQueryFrame(CvTestSeq* pTestSeq)
 
 
             {   /* Check for direct copy: */
-                DirectCopy = TRUE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,0,0)-1) > 0.00001) DirectCopy = FALSE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,1,0)) > 0.00001) DirectCopy = FALSE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,0,1)) > 0.00001) DirectCopy = FALSE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,0,1)) > 0.00001) DirectCopy = FALSE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,0,2)-(pImg->width-1)*0.5) > 0.5) DirectCopy = FALSE;
-                if( fabs(CV_MAT_ELEM(pT[0],float,1,2)-(pImg->height-1)*0.5) > 0.5) DirectCopy = FALSE;
+                DirectCopy = 1;
+                if( fabs(CV_MAT_ELEM(pT[0],float,0,0)-1) > 0.00001) DirectCopy = 0;
+                if( fabs(CV_MAT_ELEM(pT[0],float,1,0)) > 0.00001) DirectCopy = 0;
+                if( fabs(CV_MAT_ELEM(pT[0],float,0,1)) > 0.00001) DirectCopy = 0;
+                if( fabs(CV_MAT_ELEM(pT[0],float,0,1)) > 0.00001) DirectCopy = 0;
+                if( fabs(CV_MAT_ELEM(pT[0],float,0,2)-(pImg->width-1)*0.5) > 0.5) DirectCopy = 0;
+                if( fabs(CV_MAT_ELEM(pT[0],float,1,2)-(pImg->height-1)*0.5) > 0.5) DirectCopy = 0;
             }
 
             /* Extract image and mask: */
@@ -1367,4 +1367,3 @@ void cvAddNoise(IplImage* pImg, int noise_type, double Ampl, CvRandState* rnd_st
     cvReleaseImage(&pImgAdd);
 
 }   /* cvAddNoise */
-
diff --git a/modules/legacy/src/texture.cpp b/modules/legacy/src/texture.cpp
index 4b7cb061cc..07fd087269 100644
--- a/modules/legacy/src/texture.cpp
+++ b/modules/legacy/src/texture.cpp
@@ -645,4 +645,3 @@ cvCreateGLCMImage( CvGLCM* GLCM, int step )
 
     return dest;
 }
-
diff --git a/modules/legacy/src/trifocal.cpp b/modules/legacy/src/trifocal.cpp
index 87aadf696d..b049ccccfe 100644
--- a/modules/legacy/src/trifocal.cpp
+++ b/modules/legacy/src/trifocal.cpp
@@ -40,13 +40,10 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
 
-//#include "cvtypes.h"
 #include <float.h>
 #include <limits.h>
-//#include "cv.h"
-//#include "windows.h"
-
 #include <stdio.h>
 
 /* Valery Mosyagin */
@@ -223,7 +220,7 @@ int icvComputeProjectMatrices6Points( CvMat* points1,CvMat* points2,CvMat* point
         CV_ERROR( CV_StsUnmatchedSizes, "Number of coordinates of points4D  must be 4" );
     }
 #endif
-
+    {
     /* Find transform matrix for each camera */
     int i;
     CvMat* points[3];
@@ -400,7 +397,7 @@ int icvComputeProjectMatrices6Points( CvMat* points1,CvMat* points2,CvMat* point
 #endif
 
     }/* for all sollutions */
-
+    }
     __END__;
     return numSol;
 }
@@ -1362,7 +1359,7 @@ void icvFindBaseTransform(CvMat* points,CvMat* resultT)
     /* Function gets four points and compute transformation to e1=(100) e2=(010) e3=(001) e4=(111) */
 
     /* !!! test each three points not collinear. Need to test */
-
+    {
     /* Create matrices */
     CvMat matrA;
     CvMat vectB;
@@ -1410,7 +1407,7 @@ void icvFindBaseTransform(CvMat* points,CvMat* resultT)
     cvInvert(&matrA,&tmpRes);
 
     cvConvert(&tmpRes,resultT);
-
+    }
     __END__;
 
     return;
@@ -1459,7 +1456,7 @@ void GetGeneratorReduceFundSolution(CvMat* points1,CvMat* points2,CvMat* fundRed
     }
 
     /* Using 3 corr. points compute reduce */
-
+    {
     /* Create matrix */
     CvMat matrA;
     double matrA_dat[3*5];
@@ -1507,7 +1504,7 @@ void GetGeneratorReduceFundSolution(CvMat* points1,CvMat* points2,CvMat* fundRed
         cvmSet(fundReduceCoef1,0,i,cvmGet(&matrV,3,i));
         cvmSet(fundReduceCoef2,0,i,cvmGet(&matrV,4,i));
     }
-
+    }
     __END__;
     return;
 
@@ -1551,7 +1548,7 @@ int GetGoodReduceFundamMatrFromTwo(CvMat* fundReduceCoef1,CvMat* fundReduceCoef2
     {
         CV_ERROR( CV_StsUnmatchedSizes, "Size of resFundReduceCoef must be 1x5" );
     }
-
+    {
     double p1,q1,r1,s1,t1;
     double p2,q2,r2,s2,t2;
     p1 = cvmGet(fundReduceCoef1,0,0);
@@ -1599,7 +1596,7 @@ int GetGoodReduceFundamMatrFromTwo(CvMat* fundReduceCoef1,CvMat* fundReduceCoef2
             numRoots++;
         }
     }
-
+    }
     __END__;
     return numRoots;
 }
@@ -1636,7 +1633,7 @@ void GetProjMatrFromReducedFundamental(CvMat* fundReduceCoefs,CvMat* projMatrCoe
     /* Computes project matrix from given reduced matrix */
     /* we have p,q,r,s,t and need get a,b,c,d */
     /* Fill matrix to compute ratio a:b:c as A:B:C */
-
+    {
     CvMat matrA;
     double matrA_dat[3*3];
     matrA = cvMat(3,3,CV_64F,matrA_dat);
@@ -1752,7 +1749,7 @@ void GetProjMatrFromReducedFundamental(CvMat* fundReduceCoefs,CvMat* projMatrCoe
         cvmSet(projMatrCoefs,0,3,d);
 
     }
-
+    }
     __END__;
     return;
 }
@@ -2106,7 +2103,7 @@ void icvReconstructPointsFor3View( CvMat* projMatr1,CvMat* projMatr2,CvMat* proj
     {
         CV_ERROR( CV_StsUnmatchedSizes, "Size of projection matrices must be 3x4" );
     }
-
+    {
     CvMat matrA;
     double matrA_dat[36];
     matrA = cvMat(9,4,CV_64F,matrA_dat);
@@ -2203,7 +2200,7 @@ void icvReconstructPointsFor3View( CvMat* projMatr1,CvMat* projMatr2,CvMat* proj
                 }
         }
     }*/
-
+    }
     __END__;
     return;
 }
@@ -2785,4 +2782,3 @@ static void icvDecomposeQ(CvMat* /*matrQ*/,CvMat* /*matrH*/)
 }
 
 #endif
-
diff --git a/modules/legacy/src/vecfacetracking.cpp b/modules/legacy/src/vecfacetracking.cpp
index 3a11ea9256..81d1a04cb2 100644
--- a/modules/legacy/src/vecfacetracking.cpp
+++ b/modules/legacy/src/vecfacetracking.cpp
@@ -152,12 +152,12 @@ struct CvFaceTracker
         if ((NULL == imgray) ||
             (NULL == imgThresh) ||
             (NULL == mstgContours))
-            return FALSE;
-        return TRUE;
+            return 0;
+        return 1;
     };
     int InitNextImage(IplImage* img)
     {
-        CvSize sz = {img->width, img->height};
+        CvSize sz(img->width, img->height);
         ReallocImage(&imgGray, sz, 1);
         ReallocImage(&imgThresh, sz, 1);
         ptRotate = face[MOUTH].ptCenter;
@@ -165,7 +165,7 @@ struct CvFaceTracker
         CvMat mat = cvMat( 2, 3, CV_32FC1, m );
 
         if (NULL == imgGray || NULL == imgThresh)
-            return FALSE;
+            return 0;
 
         /*m[0] = (float)cos(-dbRotateAngle*CV_PI/180.);
         m[1] = (float)sin(-dbRotateAngle*CV_PI/180.);
@@ -181,8 +181,8 @@ struct CvFaceTracker
         else
             cvClearMemStorage(mstgContours);
         if (NULL == mstgContours)
-            return FALSE;
-        return TRUE;
+            return 0;
+        return 1;
     }
 };
 
@@ -209,14 +209,14 @@ public:
         if (NULL != mstg)
             m_mstgRects = mstg;
         if (NULL == m_mstgRects)
-            return FALSE;
+            return 0;
         if (NULL == m_seqRects)
             m_seqRects = cvCreateSeq(0, sizeof(CvSeq), sizeof(CvTrackingRect), m_mstgRects);
         else
             cvClearSeq(m_seqRects);
         if (NULL == m_seqRects)
-            return FALSE;
-        return TRUE;
+            return 0;
+        return 1;
     };
     void FindRects(IplImage* img, IplImage* thresh, int nLayers, int dMinSize);
 protected:
@@ -433,14 +433,14 @@ cvInitFaceTracker(CvFaceTracker* pFaceTracker, const IplImage* imgGray, CvRect*
         (nRects < NUM_FACE_ELEMENTS))
         return NULL;
 
-    //int new_face = FALSE;
+    //int new_face = 0;
     CvFaceTracker* pFace = pFaceTracker;
     if (NULL == pFace)
     {
         pFace = new CvFaceTracker;
         if (NULL == pFace)
             return NULL;
-        //new_face = TRUE;
+        //new_face = 1;
     }
     pFace->Init(pRects, (IplImage*)imgGray);
     return pFace;
@@ -464,7 +464,7 @@ cvTrackFace(CvFaceTracker* pFaceTracker, IplImage* imgGray, CvRect* pRects, int
     assert(NULL != pRects && nRects >= NUM_FACE_ELEMENTS);
     if ((NULL == pFaceTracker) ||
         (NULL == imgGray))
-        return FALSE;
+        return 0;
     pFaceTracker->InitNextImage(imgGray);
     *ptRotate = pFaceTracker->ptRotate;
     *dbAngleRotate = pFaceTracker->dbRotateAngle;
@@ -505,7 +505,7 @@ START:
         if (r.y + r.height > pFaceTracker->imgGray->height - 2)
             r.height = pFaceTracker->imgGray->height - 2 - r.y;
         if (!big_face[elem].Init(r, pFaceTracker->face[elem], pFaceTracker->mstgContours))
-            return FALSE;
+            return 0;
     }
     // find contours
     for (elem = 0; elem < NUM_FACE_ELEMENTS; elem++)
@@ -514,13 +514,13 @@ START:
     CvTrackingRect new_face[NUM_FACE_ELEMENTS];
     int new_energy = 0;
     int found = ChoiceTrackingFace3(pFaceTracker, nElements, big_face, new_face, new_energy);
-    int restart = FALSE;
-    int find2 = FALSE;
+    int restart = 0;
+    int find2 = 0;
     int noel = -1;
     if (found)
     {
         if (new_energy > 100000 && -1 != pFaceTracker->iTrackingFaceType)
-            find2 = TRUE;
+            find2 = 1;
         else if (new_energy > 150000)
         {
             int elements = 0;
@@ -532,17 +532,17 @@ START:
                     noel = el;
             }
             if (2 == elements)
-                find2 = TRUE;
+                find2 = 1;
             else
-                restart = TRUE;
+                restart = 1;
         }
     }
     else
     {
         if (-1 != pFaceTracker->iTrackingFaceType)
-            find2 = TRUE;
+            find2 = 1;
         else
-            restart = TRUE;
+            restart = 1;
     }
 RESTART:
     if (restart)
@@ -561,11 +561,11 @@ RESTART:
         if (found2 && new_energy < 100000)
         {
             pFaceTracker->iTrackingFaceType = noel;
-            found = TRUE;
+            found = 1;
         }
         else
         {
-            restart = TRUE;
+            restart = 1;
             goto RESTART;
         }
     }
@@ -599,7 +599,7 @@ RESTART:
     {
         pFaceTracker->dbRotateDelta = 0;
         pFaceTracker->dbRotateAngle = 0;
-        found = FALSE;
+        found = 0;
     }
     if (found)
     {
@@ -663,7 +663,7 @@ int ChoiceTrackingFace3(CvFaceTracker* pTF, const int nElements, const CvFaceEle
     CvTrackingRect* new_face[NUM_FACE_ELEMENTS] = {NULL};
     new_energy = 0x7fffffff;
     int curr_energy = 0x7fffffff;
-    int found = FALSE;
+    int found = 0;
     int N = 0;
     CvSeqReader reader_m, reader_l, reader_r;
     cvStartReadSeq( big_face[MOUTH].m_seqRects, &reader_m );
@@ -689,7 +689,7 @@ int ChoiceTrackingFace3(CvFaceTracker* pTF, const int nElements, const CvFaceEle
                             for (int elem = 0; elem < NUM_FACE_ELEMENTS; elem++)
                                 new_face[elem] = curr_face[elem];
                             new_energy = curr_energy;
-                            found = TRUE;
+                            found = 1;
                         }
                         N++;
                     }
@@ -722,7 +722,7 @@ int ChoiceTrackingFace2(CvFaceTracker* pTF, const int nElements, const CvFaceEle
     CvTrackingRect* new_face[NUM_FACE_ELEMENTS] = {NULL};
     new_energy = 0x7fffffff;
     int curr_energy = 0x7fffffff;
-    int found = FALSE;
+    int found = 0;
     int N = 0;
     CvSeqReader reader0, reader1;
     cvStartReadSeq( big_face[element[0]].m_seqRects, &reader0 );
@@ -739,7 +739,7 @@ int ChoiceTrackingFace2(CvFaceTracker* pTF, const int nElements, const CvFaceEle
                 for (int elem = 0; elem < NUM_FACE_ELEMENTS; elem++)
                     new_face[elem] = curr_face[elem];
                 new_energy = curr_energy;
-                found = TRUE;
+                found = 1;
             }
             N++;
         }
@@ -749,11 +749,11 @@ int ChoiceTrackingFace2(CvFaceTracker* pTF, const int nElements, const CvFaceEle
         face[element[0]] = *(new_face[element[0]]);
         face[element[1]] = *(new_face[element[1]]);
         // 3 element find by template
-        CvPoint templ_v01 = {pTF->ptTempl[element[1]].x - pTF->ptTempl[element[0]].x, pTF->ptTempl[element[1]].y - pTF->ptTempl[element[0]].y};
-        CvPoint templ_v02 = {pTF->ptTempl[element[2]].x - pTF->ptTempl[element[0]].x, pTF->ptTempl[element[2]].y - pTF->ptTempl[element[0]].y};
-        CvPoint prev_v01 = {pTF->face[element[1]].ptCenter.x - pTF->face[element[0]].ptCenter.x, pTF->face[element[1]].ptCenter.y - pTF->face[element[0]].ptCenter.y};
-        CvPoint prev_v02 = {pTF->face[element[2]].ptCenter.x - pTF->face[element[0]].ptCenter.x, pTF->face[element[2]].ptCenter.y - pTF->face[element[0]].ptCenter.y};
-        CvPoint new_v01 = {new_face[element[1]]->ptCenter.x - new_face[element[0]]->ptCenter.x, new_face[element[1]]->ptCenter.y - new_face[element[0]]->ptCenter.y};
+        CvPoint templ_v01(pTF->ptTempl[element[1]].x - pTF->ptTempl[element[0]].x, pTF->ptTempl[element[1]].y - pTF->ptTempl[element[0]].y);
+        CvPoint templ_v02(pTF->ptTempl[element[2]].x - pTF->ptTempl[element[0]].x, pTF->ptTempl[element[2]].y - pTF->ptTempl[element[0]].y);
+        CvPoint prev_v01(pTF->face[element[1]].ptCenter.x - pTF->face[element[0]].ptCenter.x, pTF->face[element[1]].ptCenter.y - pTF->face[element[0]].ptCenter.y);
+        CvPoint prev_v02(pTF->face[element[2]].ptCenter.x - pTF->face[element[0]].ptCenter.x, pTF->face[element[2]].ptCenter.y - pTF->face[element[0]].ptCenter.y);
+        CvPoint new_v01(new_face[element[1]]->ptCenter.x - new_face[element[0]]->ptCenter.x, new_face[element[1]]->ptCenter.y - new_face[element[0]]->ptCenter.y);
         double templ_d01 = sqrt((double)templ_v01.x*templ_v01.x + templ_v01.y*templ_v01.y);
         double templ_d02 = sqrt((double)templ_v02.x*templ_v02.x + templ_v02.y*templ_v02.y);
         double prev_d01 = sqrt((double)prev_v01.x*prev_v01.x + prev_v01.y*prev_v01.y);
@@ -767,7 +767,7 @@ int ChoiceTrackingFace2(CvFaceTracker* pTF, const int nElements, const CvFaceEle
         double y = double(new_v01.x) * sin_a + double(new_v01.y) * cos_a;
         x = x * new_d02 / new_d01;
         y = y * new_d02 / new_d01;
-        CvPoint new_v02 = {int(x + 0.5), int(y + 0.5)};
+        CvPoint new_v02(int(x + 0.5), int(y + 0.5));
         face[element[2]].iColor = 0;
         face[element[2]].iEnergy = 0;
         face[element[2]].nRectsInThis = 0;
@@ -818,10 +818,10 @@ inline int GetEnergy(CvTrackingRect** ppNew, const CvTrackingRect* pPrev, CvPoin
 
 inline int GetEnergy2(CvTrackingRect** ppNew, const CvTrackingRect* pPrev, CvPoint* ptTempl, CvRect* rTempl, int* element)
 {
-    CvPoint new_v = {ppNew[element[0]]->ptCenter.x - ppNew[element[1]]->ptCenter.x,
-        ppNew[element[0]]->ptCenter.y - ppNew[element[1]]->ptCenter.y};
-    CvPoint prev_v = {pPrev[element[0]].ptCenter.x - pPrev[element[1]].ptCenter.x,
-        pPrev[element[0]].ptCenter.y - pPrev[element[1]].ptCenter.y};
+    CvPoint new_v(ppNew[element[0]]->ptCenter.x - ppNew[element[1]]->ptCenter.x,
+        ppNew[element[0]]->ptCenter.y - ppNew[element[1]]->ptCenter.y);
+    CvPoint prev_v(pPrev[element[0]].ptCenter.x - pPrev[element[1]].ptCenter.x,
+        pPrev[element[0]].ptCenter.y - pPrev[element[1]].ptCenter.y);
     double new_d = sqrt((double)new_v.x*new_v.x + new_v.y*new_v.y);
     double prev_d = sqrt((double)prev_v.x*prev_v.x + prev_v.y*prev_v.y);
     double dx = ptTempl[element[0]].x - ptTempl[element[1]].x;
@@ -971,4 +971,3 @@ inline double CalculateTransformationLMS3_0( CvPoint* pTemplPoints, CvPoint* pSr
         dbLMS = dbXtXt + dbYtYt - ((double)pow(dbXtXs + dbYtYs,2) + (double)pow(dbXtYs - dbYtXs,2)) / del;
     return dbLMS;
 }
-
diff --git a/modules/legacy/src/video.cpp b/modules/legacy/src/video.cpp
index 3c6191ac95..f0bb6bca4d 100644
--- a/modules/legacy/src/video.cpp
+++ b/modules/legacy/src/video.cpp
@@ -80,5 +80,3 @@ cvDeInterlace( const CvArr* framearr, CvArr* fieldEven, CvArr* fieldOdd )
 }
 
 /* End of file. */
-
-
diff --git a/modules/legacy/test/test_bruteforcematcher.cpp b/modules/legacy/test/test_bruteforcematcher.cpp
index c6c40f845b..3259563e56 100644
--- a/modules/legacy/test/test_bruteforcematcher.cpp
+++ b/modules/legacy/test/test_bruteforcematcher.cpp
@@ -113,4 +113,3 @@ protected:
 };
 
 TEST(Legacy_BruteForceMatcher, accuracy) { CV_BruteForceMatcherTest test; test.safe_run(); }
-
diff --git a/modules/legacy/test/test_optflow.cpp b/modules/legacy/test/test_optflow.cpp
index f72d4c36af..c5b35e9637 100644
--- a/modules/legacy/test/test_optflow.cpp
+++ b/modules/legacy/test/test_optflow.cpp
@@ -42,6 +42,7 @@
 
 #include "test_precomp.hpp"
 #include "opencv2/video/tracking.hpp"
+#include "opencv2/video/tracking_c.h"
 
 #include <string>
 #include <iostream>
@@ -352,5 +353,3 @@ void CV_OptFlowTest::run( int /* start_from */)
 
 
 TEST(Legacy_OpticalFlow, accuracy) { CV_OptFlowTest test; test.safe_run(); }
-
-
diff --git a/modules/legacy/test/test_precomp.cpp b/modules/legacy/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/legacy/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/legacy/test/test_precomp.hpp b/modules/legacy/test/test_precomp.hpp
index ccfb359fb7..52b92fb6c1 100644
--- a/modules/legacy/test/test_precomp.hpp
+++ b/modules/legacy/test/test_precomp.hpp
@@ -9,12 +9,12 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/highgui.hpp"
 #include "opencv2/highgui/highgui_c.h"
 #include "opencv2/legacy.hpp"
-#include <iostream>
 
 #endif
diff --git a/modules/legacy/test/test_pyrsegmentation.cpp b/modules/legacy/test/test_pyrsegmentation.cpp
index 674265a7cf..3bd278c095 100644
--- a/modules/legacy/test/test_pyrsegmentation.cpp
+++ b/modules/legacy/test/test_pyrsegmentation.cpp
@@ -66,12 +66,12 @@ void CV_PyrSegmentationTest::run( int /*start_from*/ )
 
     int code = cvtest::TS::OK;
 
-    CvPoint _cp[] ={{33,33}, {43,33}, {43,43}, {33,43}};
-    CvPoint _cp2[] ={{50,50}, {70,50}, {70,70}, {50,70}};
+    CvPoint _cp[] ={ CvPoint(33,33), CvPoint(43,33), CvPoint(43,43), CvPoint(33,43)};
+    CvPoint _cp2[] ={CvPoint(50,50), CvPoint(70,50), CvPoint(70,70), CvPoint(50,70)};
     CvPoint* cp = _cp;
     CvPoint* cp2 = _cp2;
     CvConnectedComp *dst_comp[3];
-    CvRect rect[3] = {{50,50,21,21}, {0,0,128,128}, {33,33,11,11}};
+    CvRect rect[3] = {CvRect(50,50,21,21), CvRect(0,0,128,128), CvRect(33,33,11,11)};
     double a[3] = {441.0, 15822.0, 121.0};
 
 /*    ippiPoint cp3[] ={130,130, 150,130, 150,150, 130,150};  */
@@ -86,7 +86,7 @@ void CV_PyrSegmentationTest::run( int /*start_from*/ )
     int i, j, iter;
 
     IplImage *image, *image_f, *image_s;
-    CvSize size = {128, 128};
+    CvSize size(128, 128);
     const int threshold1 = 50, threshold2 = 50;
 
     rect[1].width = size.width;
diff --git a/modules/legacy/test/test_stereomatching.cpp b/modules/legacy/test/test_stereomatching.cpp
index a6524bb712..7262249844 100644
--- a/modules/legacy/test/test_stereomatching.cpp
+++ b/modules/legacy/test/test_stereomatching.cpp
@@ -593,10 +593,10 @@ int CV_StereoMatchingTest::readDatasetsParams( FileStorage& fs )
     assert(fn.isSeq());
     for( int i = 0; i < (int)fn.size(); i+=3 )
     {
-        string nm = fn[i];
+        String nm = fn[i];
         DatasetParams params;
-        string sf = fn[i+1]; params.dispScaleFactor = atoi(sf.c_str());
-        string uv = fn[i+2]; params.dispUnknVal = atoi(uv.c_str());
+        String sf = fn[i+1]; params.dispScaleFactor = atoi(sf.c_str());
+        String uv = fn[i+2]; params.dispUnknVal = atoi(uv.c_str());
         datasetsParams[nm] = params;
     }
     return cvtest::TS::OK;
@@ -684,10 +684,10 @@ protected:
         assert(fn.isSeq());
         for( int i = 0; i < (int)fn.size(); i+=4 )
         {
-            string caseName = fn[i], datasetName = fn[i+1];
+            String caseName = fn[i], datasetName = fn[i+1];
             RunParams params;
-            string ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
-            string iterCount = fn[i+3]; params.iterCount = atoi(iterCount.c_str());
+            String ndisp = fn[i+2]; params.ndisp = atoi(ndisp.c_str());
+            String iterCount = fn[i+3]; params.iterCount = atoi(iterCount.c_str());
             caseNames.push_back( caseName );
             caseDatasets.push_back( datasetName );
             caseRunParams.push_back( params );
diff --git a/modules/legacy/test/test_subdivisions.cpp b/modules/legacy/test/test_subdivisions.cpp
index 3dafd17ba3..0df71ab5f6 100644
--- a/modules/legacy/test/test_subdivisions.cpp
+++ b/modules/legacy/test/test_subdivisions.cpp
@@ -338,4 +338,3 @@ _exit_:
 TEST(Legacy_Subdiv, correctness) { CV_SubdivTest test; test.safe_run(); }
 
 /* End of file. */
-
diff --git a/modules/ml/doc/decision_trees.rst b/modules/ml/doc/decision_trees.rst
index 4bf9c116b9..de6fc99d63 100644
--- a/modules/ml/doc/decision_trees.rst
+++ b/modules/ml/doc/decision_trees.rst
@@ -318,4 +318,3 @@ decision tree.
 
 
 .. [Breiman84] Breiman, L., Friedman, J. Olshen, R. and Stone, C. (1984), *Classification and Regression Trees*, Wadsworth.
-
diff --git a/modules/ml/doc/k_nearest_neighbors.rst b/modules/ml/doc/k_nearest_neighbors.rst
index e018717716..05413c7785 100644
--- a/modules/ml/doc/k_nearest_neighbors.rst
+++ b/modules/ml/doc/k_nearest_neighbors.rst
@@ -11,6 +11,12 @@ CvKNearest
 
 The class implements K-Nearest Neighbors model as described in the beginning of this section.
 
+.. note::
+
+   * (Python) An example of digit recognition using KNearest can be found at opencv_source/samples/python2/digits.py
+   * (Python) An example of grid search digit recognition using KNearest can be found at opencv_source/samples/python2/digits_adjust.py
+   * (Python) An example of video digit recognition using KNearest can be found at opencv_source/samples/python2/digits_video.py
+
 CvKNearest::CvKNearest
 ----------------------
 Default and training constructors.
@@ -188,5 +194,3 @@ The sample below (currently using the obsolete ``CvMat`` structures) demonstrate
         cvReleaseMat( &trainData );
         return 0;
     }
-
-
diff --git a/modules/ml/doc/mldata.rst b/modules/ml/doc/mldata.rst
index bae342e498..c3092d1490 100644
--- a/modules/ml/doc/mldata.rst
+++ b/modules/ml/doc/mldata.rst
@@ -48,7 +48,7 @@ Class for loading the data from a ``.csv`` file.
         void set_miss_ch( char ch );
         char get_miss_ch() const;
 
-        const std::map<std::string, int>& get_class_labels_map() const;
+        const std::map<String, int>& get_class_labels_map() const;
 
     protected:
         ...
@@ -245,7 +245,7 @@ CvMLData::get_class_labels_map
 -------------------------------
 Returns a map that converts strings to labels.
 
-.. ocv:function:: const std::map<std::string, int>& CvMLData::get_class_labels_map() const
+.. ocv:function:: const std::map<String, int>& CvMLData::get_class_labels_map() const
 
 The method returns a map that converts string class labels to the numerical class labels. It can be used to get an original class label as in a file.
 
diff --git a/modules/ml/doc/neural_networks.rst b/modules/ml/doc/neural_networks.rst
index f955f15b11..0496e2201a 100644
--- a/modules/ml/doc/neural_networks.rst
+++ b/modules/ml/doc/neural_networks.rst
@@ -278,4 +278,3 @@ Returns neurons weights of the particular layer.
 .. ocv:function:: double* CvANN_MLP::get_weights(int layer)
 
     :param layer: Index of the particular layer.
-
diff --git a/modules/ml/doc/statistical_models.rst b/modules/ml/doc/statistical_models.rst
index 76da927b19..af250b7864 100644
--- a/modules/ml/doc/statistical_models.rst
+++ b/modules/ml/doc/statistical_models.rst
@@ -161,4 +161,3 @@ Predicts the response for a sample.
 The method is used to predict the response for a new sample. In case of a classification, the method returns the class label. In case of a regression, the method returns the output function value. The input sample must have as many components as the ``train_data`` passed to ``train`` contains. If the ``var_idx`` parameter is passed to ``train``, it is remembered and then is used to extract only the necessary components from the input sample in the method ``predict``.
 
 The suffix ``const`` means that prediction does not affect the internal model state, so the method can be safely called from within different threads.
-
diff --git a/modules/ml/doc/support_vector_machines.rst b/modules/ml/doc/support_vector_machines.rst
index 723dda0873..14ee12bb4e 100644
--- a/modules/ml/doc/support_vector_machines.rst
+++ b/modules/ml/doc/support_vector_machines.rst
@@ -158,6 +158,12 @@ CvSVM
 
 Support Vector Machines.
 
+.. note::
+
+   * (Python) An example of digit recognition using SVM can be found at opencv_source/samples/python2/digits.py
+   * (Python) An example of grid search digit recognition using SVM can be found at opencv_source/samples/python2/digits_adjust.py
+   * (Python) An example of video digit recognition using SVM can be found at opencv_source/samples/python2/digits_video.py
+
 CvSVM::CvSVM
 ------------
 Default and training constructors.
diff --git a/modules/ml/include/opencv2/ml.hpp b/modules/ml/include/opencv2/ml.hpp
index 9592fe7dfb..2b84d67ca7 100644
--- a/modules/ml/include/opencv2/ml.hpp
+++ b/modules/ml/include/opencv2/ml.hpp
@@ -41,13 +41,16 @@
 #ifndef __OPENCV_ML_HPP__
 #define __OPENCV_ML_HPP__
 
-#include "opencv2/core.hpp"
+#ifdef __cplusplus
+#  include "opencv2/core.hpp"
+#endif
+
+#include "opencv2/core/core_c.h"
 #include <limits.h>
 
 #ifdef __cplusplus
 
 #include <map>
-#include <string>
 #include <iostream>
 
 // Apple defines a check() macro somewhere in the debug headers
@@ -537,6 +540,8 @@ protected:
     virtual void write_params( CvFileStorage* fs ) const;
     virtual void read_params( CvFileStorage* fs, CvFileNode* node );
 
+    void optimize_linear_svm();
+
     CvSVMParams params;
     CvMat* class_labels;
     int var_all;
@@ -937,6 +942,8 @@ protected:
     CvDTreeNode* root;
     CvMat* var_importance;
     CvDTreeTrainData* data;
+    CvMat train_data_hdr, responses_hdr;
+    cv::Mat train_data_mat, responses_mat;
 
 public:
     int pruned_tree_idx;
@@ -1041,13 +1048,15 @@ public:
     CvForestTree* get_tree(int i) const;
 
 protected:
-    virtual std::string getName() const;
+    virtual cv::String getName() const;
 
     virtual bool grow_forest( const CvTermCriteria term_crit );
 
     // array of the trees of the forest
     CvForestTree** trees;
     CvDTreeTrainData* data;
+    CvMat train_data_hdr, responses_hdr;
+    cv::Mat train_data_mat, responses_mat;
     int ntrees;
     int nclasses;
     double oob_error;
@@ -1113,7 +1122,7 @@ public:
                        CvRTParams params=CvRTParams());
     virtual bool train( CvMLData* data, CvRTParams params=CvRTParams() );
 protected:
-    virtual std::string getName() const;
+    virtual cv::String getName() const;
     virtual bool grow_forest( const CvTermCriteria term_crit );
 };
 
@@ -1263,6 +1272,8 @@ protected:
     virtual void initialize_weights(double (&p)[2]);
 
     CvDTreeTrainData* data;
+    CvMat train_data_hdr, responses_hdr;
+    cv::Mat train_data_mat, responses_mat;
     CvBoostParams params;
     CvSeq* weak;
 
@@ -2070,7 +2081,7 @@ public:
     void set_miss_ch( char ch );
     char get_miss_ch() const;
 
-    const std::map<std::string, int>& get_class_labels_map() const;
+    const std::map<cv::String, int>& get_class_labels_map() const;
 
 protected:
     virtual void clear();
@@ -2099,7 +2110,7 @@ protected:
     bool mix;
 
     int total_class_count;
-    std::map<std::string, int> class_map;
+    std::map<cv::String, int> class_map;
 
     CvMat* train_sample_idx;
     CvMat* test_sample_idx;
@@ -2137,7 +2148,7 @@ typedef CvANN_MLP NeuralNet_MLP;
 typedef CvGBTreesParams GradientBoostingTreeParams;
 typedef CvGBTrees GradientBoostingTrees;
 
-template<> CV_EXPORTS void Ptr<CvDTreeSplit>::delete_obj();
+template<> CV_EXPORTS void DefaultDeleter<CvDTreeSplit>::operator ()(CvDTreeSplit* obj) const;
 
 CV_EXPORTS bool initModule_ml(void);
 }
diff --git a/modules/ml/include/opencv2/ml/ml.hpp b/modules/ml/include/opencv2/ml/ml.hpp
index 676d861b69..f6f9cd8f89 100644
--- a/modules/ml/include/opencv2/ml/ml.hpp
+++ b/modules/ml/include/opencv2/ml/ml.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/ml.hpp"
\ No newline at end of file
+#include "opencv2/ml.hpp"
diff --git a/modules/ml/src/ann_mlp.cpp b/modules/ml/src/ann_mlp.cpp
index 438872ae8c..7323ab57a7 100644
--- a/modules/ml/src/ann_mlp.cpp
+++ b/modules/ml/src/ann_mlp.cpp
@@ -40,10 +40,6 @@
 
 #include "precomp.hpp"
 
-#ifdef HAVE_TBB
-#include <tbb/tbb.h>
-#endif
-
 CvANN_MLP_TrainParams::CvANN_MLP_TrainParams()
 {
     term_crit = cvTermCriteria( CV_TERMCRIT_ITER + CV_TERMCRIT_EPS, 1000, 0.01 );
@@ -255,7 +251,7 @@ void CvANN_MLP::create( const CvMat* _layer_sizes, int _activ_func,
     buf_sz += (l_dst[0] + l_dst[l_count-1]*2)*2;
 
     CV_CALL( wbuf = cvCreateMat( 1, buf_sz, CV_64F ));
-    CV_CALL( weights = (double**)cvAlloc( (l_count+1)*sizeof(weights[0]) ));
+    CV_CALL( weights = (double**)cvAlloc( (l_count+2)*sizeof(weights[0]) ));
 
     weights[0] = wbuf->data.db;
     weights[1] = weights[0] + l_dst[0]*2;
@@ -1022,7 +1018,7 @@ int CvANN_MLP::train_backprop( CvVectors x0, CvVectors u, const double* sw )
     return iter;
 }
 
-struct rprop_loop {
+struct rprop_loop : cv::ParallelLoopBody {
   rprop_loop(const CvANN_MLP* _point, double**& _weights, int& _count, int& _ivcount, CvVectors* _x0,
      int& _l_count, CvMat*& _layer_sizes, int& _ovcount, int& _max_count,
      CvVectors* _u, const double*& _sw, double& _inv_count, CvMat*& _dEdw, int& _dcount0, double* _E, int _buf_sz)
@@ -1063,7 +1059,7 @@ struct rprop_loop {
   int buf_sz;
 
 
-  void operator()( const cv::BlockedRange& range ) const
+  void operator()( const cv::Range& range ) const
   {
     double* buf_ptr;
     double** x = 0;
@@ -1084,7 +1080,7 @@ struct rprop_loop {
         buf_ptr += (df[i] - x[i])*2;
     }
 
-    for(int si = range.begin(); si < range.end(); si++ )
+    for(int si = range.start; si < range.end; si++ )
     {
         if (si % dcount0 != 0) continue;
         int n1, n2, k;
@@ -1170,36 +1166,33 @@ struct rprop_loop {
             }
 
         // backward pass, update dEdw
-        #ifdef HAVE_TBB
-        static tbb::spin_mutex mutex;
-        tbb::spin_mutex::scoped_lock lock;
-        #endif
+        static cv::Mutex mutex;
+
         for(int i = l_count-1; i > 0; i-- )
         {
             n1 = layer_sizes->data.i[i-1]; n2 = layer_sizes->data.i[i];
             cvInitMatHeader( &_df, dcount, n2, CV_64F, df[i] );
             cvMul( grad1, &_df, grad1 );
-            #ifdef HAVE_TBB
-            lock.acquire(mutex);
-            #endif
-            cvInitMatHeader( &_dEdw, n1, n2, CV_64F, dEdw->data.db+(weights[i]-weights[0]) );
-            cvInitMatHeader( x1, dcount, n1, CV_64F, x[i-1] );
-            cvGEMM( x1, grad1, 1, &_dEdw, 1, &_dEdw, CV_GEMM_A_T );
 
-            // update bias part of dEdw
-           for( k = 0; k < dcount; k++ )
-           {
-               double* dst = _dEdw.data.db + n1*n2;
-               const double* src = grad1->data.db + k*n2;
-               for(int j = 0; j < n2; j++ )
-                   dst[j] += src[j];
+            {
+                cv::AutoLock lock(mutex);
+                cvInitMatHeader( &_dEdw, n1, n2, CV_64F, dEdw->data.db+(weights[i]-weights[0]) );
+                cvInitMatHeader( x1, dcount, n1, CV_64F, x[i-1] );
+                cvGEMM( x1, grad1, 1, &_dEdw, 1, &_dEdw, CV_GEMM_A_T );
+
+                // update bias part of dEdw
+                for( k = 0; k < dcount; k++ )
+                {
+                    double* dst = _dEdw.data.db + n1*n2;
+                    const double* src = grad1->data.db + k*n2;
+                    for(int j = 0; j < n2; j++ )
+                        dst[j] += src[j];
+                }
+
+                if (i > 1)
+                    cvInitMatHeader( &_w, n1, n2, CV_64F, weights[i] );
            }
 
-           if (i > 1)
-               cvInitMatHeader( &_w, n1, n2, CV_64F, weights[i] );
-           #ifdef HAVE_TBB
-           lock.release();
-           #endif
            cvInitMatHeader( grad2, dcount, n1, CV_64F, grad2->data.db );
            if( i > 1 )
                cvGEMM( grad1, &_w, 1, 0, 0, grad2, CV_GEMM_B_T );
@@ -1297,7 +1290,7 @@ int CvANN_MLP::train_rprop( CvVectors x0, CvVectors u, const double* sw )
         double E = 0;
 
         // first, iterate through all the samples and compute dEdw
-        cv::parallel_for(cv::BlockedRange(0, count),
+        cv::parallel_for_(cv::Range(0, count),
             rprop_loop(this, weights, count, ivcount, &x0, l_count, layer_sizes,
                        ovcount, max_count, &u, sw, inv_count, dEdw, dcount0, &E, buf_sz)
         );
diff --git a/modules/ml/src/boost.cpp b/modules/ml/src/boost.cpp
index 3d58ac0398..a22e13a532 100644
--- a/modules/ml/src/boost.cpp
+++ b/modules/ml/src/boost.cpp
@@ -351,9 +351,12 @@ CvBoostTree::find_split_ord_class( CvDTreeNode* node, int vi, float init_quality
     return split;
 }
 
-
-#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
-static CV_IMPLEMENT_QSORT_EX( icvSortDblPtr, double*, CV_CMP_NUM_PTR, int )
+template<typename T>
+class LessThanPtr
+{
+public:
+    bool operator()(T* a, T* b) const { return *a < *b; }
+};
 
 CvDTreeSplit*
 CvBoostTree::find_split_cat_class( CvDTreeNode* node, int vi, float init_quality, CvDTreeSplit* _split, uchar* _ext_buf )
@@ -412,7 +415,7 @@ CvBoostTree::find_split_cat_class( CvDTreeNode* node, int vi, float init_quality
 
     // sort rows of c_jk by increasing c_j,1
     // (i.e. by the weight of samples in j-th category that belong to class 1)
-    icvSortDblPtr( dbl_ptr, mi, 0 );
+    std::sort(dbl_ptr, dbl_ptr + mi, LessThanPtr<double>());
 
     for( subset_i = 0; subset_i < mi-1; subset_i++ )
     {
@@ -594,7 +597,7 @@ CvBoostTree::find_split_cat_reg( CvDTreeNode* node, int vi, float init_quality,
         sum_ptr[i] = sum + i;
     }
 
-    icvSortDblPtr( sum_ptr, mi, 0 );
+    std::sort(sum_ptr, sum_ptr + mi, LessThanPtr<double>());
 
     // revert back to unnormalized sums
     // (there should be a very little loss in accuracy)
@@ -1421,9 +1424,6 @@ CvBoost::update_weights( CvBoostTree* tree )
 }
 
 
-static CV_IMPLEMENT_QSORT_EX( icvSort_64f, double, CV_LT, int )
-
-
 void
 CvBoost::trim_weights()
 {
@@ -1440,7 +1440,7 @@ CvBoost::trim_weights()
     // use weak_eval as temporary buffer for sorted weights
     cvCopy( weights, weak_eval );
 
-    icvSort_64f( weak_eval->data.db, count, 0 );
+    std::sort(weak_eval->data.db, weak_eval->data.db + count);
 
     // as weight trimming occurs immediately after updating the weights,
     // where they are renormalized, we assume that the weight sum = 1.
@@ -1655,10 +1655,10 @@ CvBoost::predict( const CvMat* _sample, const CvMat* _missing,
     const int* cmap = data->cat_map->data.i;
     const int* cofs = data->cat_ofs->data.i;
 
-    cv::Mat sample = _sample;
+    cv::Mat sample = cv::cvarrToMat(_sample);
     cv::Mat missing;
     if(!_missing)
-        missing = _missing;
+        missing = cv::cvarrToMat(_missing);
 
     // if need, preprocess the input vector
     if( !raw_mode )
@@ -2122,9 +2122,14 @@ CvBoost::train( const Mat& _train_data, int _tflag,
                const Mat& _missing_mask,
                CvBoostParams _params, bool _update )
 {
-    CvMat tdata = _train_data, responses = _responses, vidx = _var_idx,
-        sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
-    return train(&tdata, _tflag, &responses, vidx.data.ptr ? &vidx : 0,
+    train_data_hdr = _train_data;
+    train_data_mat = _train_data;
+    responses_hdr = _responses;
+    responses_mat = _responses;
+
+    CvMat vidx = _var_idx, sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
+
+    return train(&train_data_hdr, _tflag, &responses_hdr, vidx.data.ptr ? &vidx : 0,
           sidx.data.ptr ? &sidx : 0, vtype.data.ptr ? &vtype : 0,
           mmask.data.ptr ? &mmask : 0, _params, _update);
 }
diff --git a/modules/ml/src/data.cpp b/modules/ml/src/data.cpp
index 7d9cceac4e..9171dd713c 100644
--- a/modules/ml/src/data.cpp
+++ b/modules/ml/src/data.cpp
@@ -170,8 +170,13 @@ int CvMLData::read_csv(const char* filename)
 
     // skip header lines
     for( int i = 0; i < header_lines_number; i++ )
+    {
         if( fgets( buf, M, file ) == 0 )
+        {
+            fclose(file);
             return -1;
+        }
+    }
 
     // read the first data line and determine the number of variables
     if( !fgets_chomp( buf, M, file ))
@@ -285,7 +290,7 @@ const CvMat* CvMLData::get_missing() const
     return missing;
 }
 
-const std::map<std::string, int>& CvMLData::get_class_labels_map() const
+const std::map<cv::String, int>& CvMLData::get_class_labels_map() const
 {
     return class_map;
 }
diff --git a/modules/ml/src/ertrees.cpp b/modules/ml/src/ertrees.cpp
index 31d20398a9..0201deb0c6 100644
--- a/modules/ml/src/ertrees.cpp
+++ b/modules/ml/src/ertrees.cpp
@@ -44,13 +44,18 @@ static const float ord_nan = FLT_MAX*0.5f;
 static const int min_block_size = 1 << 16;
 static const int block_size_delta = 1 << 10;
 
-#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
-static CV_IMPLEMENT_QSORT_EX( icvSortIntPtr, int*, CV_CMP_NUM_PTR, int )
+template<typename T>
+class LessThanPtr
+{
+public:
+    bool operator()(T* a, T* b) const { return *a < *b; }
+};
 
-#define CV_CMP_PAIRS(a,b) (*((a).i) < *((b).i))
-static CV_IMPLEMENT_QSORT_EX( icvSortPairs, CvPair16u32s, CV_CMP_PAIRS, int )
-
-///
+class LessThanPairs
+{
+public:
+    bool operator()(const CvPair16u32s& a, const CvPair16u32s& b) const { return *a.i < *b.i; }
+};
 
 void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
     const CvMat* _responses, const CvMat* _var_idx, const CvMat* _sample_idx,
@@ -353,7 +358,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
 
             if (is_buf_16u)
             {
-                icvSortPairs( pair16u32s_ptr, sample_count, 0 );
+                std::sort(pair16u32s_ptr, pair16u32s_ptr + sample_count, LessThanPairs());
                 // count the categories
                 for( i = 1; i < num_valid; i++ )
                     if (*pair16u32s_ptr[i].i != *pair16u32s_ptr[i-1].i)
@@ -361,7 +366,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
             }
             else
             {
-                icvSortIntPtr( int_ptr, sample_count, 0 );
+                std::sort(int_ptr, int_ptr + sample_count, LessThanPtr<int>());
                 // count the categories
                 for( i = 1; i < num_valid; i++ )
                     c_count += *int_ptr[i] != *int_ptr[i-1];
@@ -1537,7 +1542,7 @@ CvERTrees::~CvERTrees()
 {
 }
 
-std::string CvERTrees::getName() const
+cv::String CvERTrees::getName() const
 {
     return CV_TYPE_NAME_ML_ERTREES;
 }
@@ -1839,12 +1844,16 @@ bool CvERTrees::train( const Mat& _train_data, int _tflag,
                       const Mat& _sample_idx, const Mat& _var_type,
                       const Mat& _missing_mask, CvRTParams params )
 {
-    CvMat tdata = _train_data, responses = _responses, vidx = _var_idx,
-    sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
-    return train(&tdata, _tflag, &responses, vidx.data.ptr ? &vidx : 0,
+    train_data_hdr = _train_data;
+    train_data_mat = _train_data;
+    responses_hdr = _responses;
+    responses_mat = _responses;
+
+    CvMat vidx = _var_idx, sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
+
+    return train(&train_data_hdr, _tflag, &responses_hdr, vidx.data.ptr ? &vidx : 0,
                  sidx.data.ptr ? &sidx : 0, vtype.data.ptr ? &vtype : 0,
                  mmask.data.ptr ? &mmask : 0, params);
 }
 
 // End of file.
-
diff --git a/modules/ml/src/gbt.cpp b/modules/ml/src/gbt.cpp
index a409b7ab32..42d0d4f3af 100644
--- a/modules/ml/src/gbt.cpp
+++ b/modules/ml/src/gbt.cpp
@@ -5,9 +5,6 @@
 #define pCvSeq CvSeq*
 #define pCvDTreeNode CvDTreeNode*
 
-#define CV_CMP_FLOAT(a,b) ((a) < (b))
-static CV_IMPLEMENT_QSORT_EX( icvSortFloat, float, CV_CMP_FLOAT, float)
-
 //===========================================================================
 //----------------------------- CvGBTreesParams -----------------------------
 //===========================================================================
@@ -285,7 +282,7 @@ CvGBTrees::train( const CvMat* _train_data, int _tflag,
             } break;
             default: CV_Error(CV_StsUnmatchedFormats, "_sample_idx should be a 32sC1, 8sC1 or 8uC1 vector.");
         }
-        icvSortFloat(sample_idx->data.fl, sample_idx_len, 0);
+        std::sort(sample_idx->data.fl, sample_idx->data.fl + sample_idx_len);
     }
     else
     {
@@ -470,7 +467,7 @@ void CvGBTrees::find_gradient(const int k)
                 int idx = *(sample_data + subsample_data[i]*s_step);
                 residuals[i] = fabs(resp_data[idx] - current_data[idx]);
             }
-            icvSortFloat(residuals, n, 0.0f);
+            std::sort(residuals, residuals + n);
 
             delta = residuals[int(ceil(n*alpha))];
 
@@ -693,7 +690,7 @@ float CvGBTrees::find_optimal_value( const CvMat* _Idx )
             float* residuals = new float[n];
             for (int i=0; i<n; ++i, ++idx)
                 residuals[i] = (resp_data[*idx] - cur_data[*idx]);
-            icvSortFloat(residuals, n, 0.0f);
+            std::sort(residuals, residuals + n);
             if (n % 2)
                 gamma = residuals[n/2];
             else gamma = (residuals[n/2-1] + residuals[n/2]) / 2.0f;
@@ -705,7 +702,7 @@ float CvGBTrees::find_optimal_value( const CvMat* _Idx )
             float* residuals = new float[n];
             for (int i=0; i<n; ++i, ++idx)
                 residuals[i] = (resp_data[*idx] - cur_data[*idx]);
-            icvSortFloat(residuals, n, 0.0f);
+            std::sort(residuals, residuals + n);
 
             int n_half = n >> 1;
             float r_median = (n == n_half<<1) ?
@@ -887,7 +884,7 @@ float CvGBTrees::predict_serial( const CvMat* _sample, const CvMat* _missing,
 }
 
 
-class Tree_predictor
+class Tree_predictor : public cv::ParallelLoopBody
 {
 private:
     pCvSeq* weak;
@@ -897,9 +894,7 @@ private:
     const CvMat* missing;
     const float shrinkage;
 
-#ifdef HAVE_TBB
-    static tbb::spin_mutex SumMutex;
-#endif
+    static cv::Mutex SumMutex;
 
 
 public:
@@ -918,14 +913,11 @@ public:
     Tree_predictor& operator=( const Tree_predictor& )
     { return *this; }
 
-    virtual void operator()(const cv::BlockedRange& range) const
+    virtual void operator()(const cv::Range& range) const
     {
-#ifdef HAVE_TBB
-        tbb::spin_mutex::scoped_lock lock;
-#endif
         CvSeqReader reader;
-        int begin = range.begin();
-        int end = range.end();
+        int begin = range.start;
+        int end = range.end;
 
         int weak_count = end - begin;
         CvDTree* tree;
@@ -943,13 +935,11 @@ public:
                     tmp_sum += shrinkage*(float)(tree->predict(sample, missing)->value);
                 }
             }
-#ifdef HAVE_TBB
-            lock.acquire(SumMutex);
-            sum[i] += tmp_sum;
-            lock.release();
-#else
-            sum[i] += tmp_sum;
-#endif
+
+            {
+                cv::AutoLock lock(SumMutex);
+                sum[i] += tmp_sum;
+            }
         }
     } // Tree_predictor::operator()
 
@@ -957,11 +947,7 @@ public:
 
 }; // class Tree_predictor
 
-
-#ifdef HAVE_TBB
-tbb::spin_mutex Tree_predictor::SumMutex;
-#endif
-
+cv::Mutex Tree_predictor::SumMutex;
 
 
 float CvGBTrees::predict( const CvMat* _sample, const CvMat* _missing,
@@ -979,12 +965,7 @@ float CvGBTrees::predict( const CvMat* _sample, const CvMat* _missing,
         Tree_predictor predictor = Tree_predictor(weak_seq, class_count,
                                     params.shrinkage, _sample, _missing, sum);
 
-//#ifdef HAVE_TBB
-//      tbb::parallel_for(cv::BlockedRange(begin, end), predictor,
-//                          tbb::auto_partitioner());
-//#else
-        cv::parallel_for(cv::BlockedRange(begin, end), predictor);
-//#endif
+        cv::parallel_for_(cv::Range(begin, end), predictor);
 
         for (int i=0; i<class_count; ++i)
             sum[i] = sum[i] /** params.shrinkage*/ + base_value;
@@ -1117,7 +1098,7 @@ void CvGBTrees::write( CvFileStorage* fs, const char* name ) const
 
     CvSeqReader reader;
     int i;
-    std::string s;
+    cv::String s;
 
     cvStartWriteStruct( fs, name, CV_NODE_MAP, CV_TYPE_NAME_ML_GBT );
 
@@ -1167,7 +1148,7 @@ void CvGBTrees::read( CvFileStorage* fs, CvFileNode* node )
     CvFileNode* trees_fnode;
     CvMemStorage* storage;
     int i, ntrees;
-    std::string s;
+    cv::String s;
 
     clear();
     read_params( fs, node );
@@ -1213,7 +1194,7 @@ void CvGBTrees::read( CvFileStorage* fs, CvFileNode* node )
 
 //===========================================================================
 
-class Sample_predictor
+class Sample_predictor : public cv::ParallelLoopBody
 {
 private:
     const CvGBTrees* gbt;
@@ -1243,10 +1224,10 @@ public:
     {}
 
 
-    virtual void operator()(const cv::BlockedRange& range) const
+    virtual void operator()(const cv::Range& range) const
     {
-        int begin = range.begin();
-        int end = range.end();
+        int begin = range.start;
+        int end = range.end;
 
         CvMat x;
         CvMat miss;
@@ -1302,11 +1283,7 @@ CvGBTrees::calc_error( CvMLData* _data, int type, std::vector<float> *resp )
     Sample_predictor predictor = Sample_predictor(this, pred_resp, _data->get_values(),
             _data->get_missing(), _sample_idx);
 
-//#ifdef HAVE_TBB
-//    tbb::parallel_for(cv::BlockedRange(0,n), predictor, tbb::auto_partitioner());
-//#else
-    cv::parallel_for(cv::BlockedRange(0,n), predictor);
-//#endif
+    cv::parallel_for_(cv::Range(0,n), predictor);
 
     int* sidx = _sample_idx ? _sample_idx->data.i : 0;
     int r_step = CV_IS_MAT_CONT(response->type) ?
diff --git a/modules/ml/src/knearest.cpp b/modules/ml/src/knearest.cpp
index 3c2f9ebada..a05a30da86 100644
--- a/modules/ml/src/knearest.cpp
+++ b/modules/ml/src/knearest.cpp
@@ -116,7 +116,7 @@ bool CvKNearest::train( const CvMat* _train_data, const CvMat* _responses,
 
     if( !responses )
         CV_ERROR( CV_StsNoMem, "Could not allocate memory for responses" );
-        
+
     if( _update_base && _dims != var_count )
         CV_ERROR( CV_StsBadArg, "The newly added data have different dimensionality" );
 
@@ -306,7 +306,7 @@ float CvKNearest::write_results( int k, int k1, int start, int end,
     return result;
 }
 
-struct P1 {
+struct P1 : cv::ParallelLoopBody {
   P1(const CvKNearest* _pointer, int _buf_sz, int _k, const CvMat* __samples, const float** __neighbors,
      int _k1, CvMat* __results, CvMat* __neighbor_responses, CvMat* __dist, float* _result)
   {
@@ -333,10 +333,10 @@ struct P1 {
   float* result;
   int buf_sz;
 
-  void operator()( const cv::BlockedRange& range ) const
+  void operator()( const cv::Range& range ) const
   {
     cv::AutoBuffer<float> buf(buf_sz);
-    for(int i = range.begin(); i < range.end(); i += 1 )
+    for(int i = range.start; i < range.end; i += 1 )
     {
         float* neighbor_responses = &buf[0];
         float* dist = neighbor_responses + 1*k;
@@ -410,8 +410,8 @@ float CvKNearest::find_nearest( const CvMat* _samples, int k, CvMat* _results,
     int k1 = get_sample_count();
     k1 = MIN( k1, k );
 
-    cv::parallel_for(cv::BlockedRange(0, count), P1(this, buf_sz, k, _samples, _neighbors, k1,
-                                                    _results, _neighbor_responses, _dist, &result)
+    cv::parallel_for_(cv::Range(0, count), P1(this, buf_sz, k, _samples, _neighbors, k1,
+                                             _results, _neighbor_responses, _dist, &result)
     );
 
     return result;
@@ -480,4 +480,3 @@ float CvKNearest::find_nearest( const cv::Mat& _samples, int k, CV_OUT cv::Mat&
 }
 
 /* End of file */
-
diff --git a/modules/ml/src/ml_init.cpp b/modules/ml/src/ml_init.cpp
index 276f6f5d47..3ff7d9baee 100644
--- a/modules/ml/src/ml_init.cpp
+++ b/modules/ml/src/ml_init.cpp
@@ -56,7 +56,7 @@ CV_INIT_ALGORITHM(EM, "StatModel.EM",
 
 bool initModule_ml(void)
 {
-    Ptr<Algorithm> em = createEM();
+    Ptr<Algorithm> em = createEM_ptr_hidden();
     return em->info() != 0;
 }
 
diff --git a/modules/ml/src/nbayes.cpp b/modules/ml/src/nbayes.cpp
index 15146d6f4e..b9a966cd1d 100644
--- a/modules/ml/src/nbayes.cpp
+++ b/modules/ml/src/nbayes.cpp
@@ -277,7 +277,7 @@ bool CvNormalBayesClassifier::train( const CvMat* _train_data, const CvMat* _res
     return result;
 }
 
-struct predict_body {
+struct predict_body : cv::ParallelLoopBody {
   predict_body(CvMat* _c, CvMat** _cov_rotate_mats, CvMat** _inv_eigen_values, CvMat** _avg,
      const CvMat* _samples, const int* _vidx, CvMat* _cls_labels,
      CvMat* _results, float* _value, int _var_count1
@@ -307,7 +307,7 @@ struct predict_body {
   float* value;
   int var_count1;
 
-  void operator()( const cv::BlockedRange& range ) const
+  void operator()( const cv::Range& range ) const
   {
 
     int cls = -1;
@@ -324,7 +324,7 @@ struct predict_body {
     cv::AutoBuffer<double> buffer(nclasses + var_count1);
     CvMat diff = cvMat( 1, var_count1, CV_64FC1, &buffer[0] );
 
-    for(int k = range.begin(); k < range.end(); k += 1 )
+    for(int k = range.start; k < range.end; k += 1 )
     {
         int ival;
         double opt = FLT_MAX;
@@ -397,9 +397,9 @@ float CvNormalBayesClassifier::predict( const CvMat* samples, CvMat* results ) c
 
     const int* vidx = var_idx ? var_idx->data.i : 0;
 
-    cv::parallel_for(cv::BlockedRange(0, samples->rows), predict_body(c, cov_rotate_mats, inv_eigen_values, avg, samples,
-                                                                      vidx, cls_labels, results, &value, var_count
-    ));
+    cv::parallel_for_(cv::Range(0, samples->rows),
+                      predict_body(c, cov_rotate_mats, inv_eigen_values, avg, samples,
+                                   vidx, cls_labels, results, &value, var_count));
 
     return value;
 }
@@ -623,4 +623,3 @@ float CvNormalBayesClassifier::predict( const Mat& _samples, Mat* _results ) con
 }
 
 /* End of file. */
-
diff --git a/modules/ml/src/precomp.hpp b/modules/ml/src/precomp.hpp
index 112e077911..06b8f4f355 100644
--- a/modules/ml/src/precomp.hpp
+++ b/modules/ml/src/precomp.hpp
@@ -41,13 +41,12 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
+#include "opencv2/core.hpp"
 #include "opencv2/ml.hpp"
 #include "opencv2/core/core_c.h"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/core/private.hpp"
 
 #include <assert.h>
 #include <float.h>
@@ -59,6 +58,9 @@
 #include <time.h>
 
 #define ML_IMPL CV_IMPL
+#define __BEGIN__ __CV_BEGIN__
+#define __END__ __CV_END__
+#define EXIT __CV_EXIT__
 
 #define CV_MAT_ELEM_FLAG( mat, type, comp, vect, tflag )    \
     (( tflag == CV_ROW_SAMPLE )                             \
diff --git a/modules/ml/src/rtrees.cpp b/modules/ml/src/rtrees.cpp
index d88611bef5..c41b842142 100644
--- a/modules/ml/src/rtrees.cpp
+++ b/modules/ml/src/rtrees.cpp
@@ -126,7 +126,7 @@ void ForestTreeBestSplitFinder::operator()(const BlockedRange& range)
         }
 
         if( res && bestSplit->quality < split->quality )
-                memcpy( (CvDTreeSplit*)bestSplit, (CvDTreeSplit*)split, splitSize );
+            memcpy( bestSplit.get(), split.get(), splitSize );
     }
 }
 }
@@ -246,7 +246,7 @@ CvRTrees::~CvRTrees()
     clear();
 }
 
-std::string CvRTrees::getName() const
+cv::String CvRTrees::getName() const
 {
     return CV_TYPE_NAME_ML_RTREES;
 }
@@ -730,7 +730,7 @@ void CvRTrees::write( CvFileStorage* fs, const char* name ) const
     if( ntrees < 1 || !trees || nsamples < 1 )
         CV_Error( CV_StsBadArg, "Invalid CvRTrees object" );
 
-    std::string modelNodeName = this->getName();
+    cv::String modelNodeName = this->getName();
     cvStartWriteStruct( fs, name, CV_NODE_MAP, modelNodeName.c_str() );
 
     cvWriteInt( fs, "nclasses", nclasses );
@@ -839,9 +839,14 @@ bool CvRTrees::train( const Mat& _train_data, int _tflag,
                      const Mat& _sample_idx, const Mat& _var_type,
                      const Mat& _missing_mask, CvRTParams _params )
 {
-    CvMat tdata = _train_data, responses = _responses, vidx = _var_idx,
-    sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
-    return train(&tdata, _tflag, &responses, vidx.data.ptr ? &vidx : 0,
+    train_data_hdr = _train_data;
+    train_data_mat = _train_data;
+    responses_hdr = _responses;
+    responses_mat = _responses;
+
+    CvMat vidx = _var_idx, sidx = _sample_idx, vtype = _var_type, mmask = _missing_mask;
+
+    return train(&train_data_hdr, _tflag, &responses_hdr, vidx.data.ptr ? &vidx : 0,
                  sidx.data.ptr ? &sidx : 0, vtype.data.ptr ? &vtype : 0,
                  mmask.data.ptr ? &mmask : 0, _params);
 }
@@ -861,7 +866,7 @@ float CvRTrees::predict_prob( const Mat& _sample, const Mat& _missing) const
 
 Mat CvRTrees::getVarImportance()
 {
-    return Mat(get_var_importance());
+    return cvarrToMat(get_var_importance());
 }
 
 // End of file.
diff --git a/modules/ml/src/svm.cpp b/modules/ml/src/svm.cpp
index 7183d62b1e..50c618e5d9 100644
--- a/modules/ml/src/svm.cpp
+++ b/modules/ml/src/svm.cpp
@@ -1566,6 +1566,7 @@ bool CvSVM::do_train( int svm_type, int sample_count, int var_count, const float
         }
     }
 
+    optimize_linear_svm();
     ok = true;
 
     __END__;
@@ -1573,6 +1574,62 @@ bool CvSVM::do_train( int svm_type, int sample_count, int var_count, const float
     return ok;
 }
 
+
+void CvSVM::optimize_linear_svm()
+{
+    // we optimize only linear SVM: compress all the support vectors into one.
+    if( params.kernel_type != LINEAR )
+        return;
+
+    int class_count = class_labels ? class_labels->cols :
+            params.svm_type == CvSVM::ONE_CLASS ? 1 : 0;
+
+    int i, df_count = class_count > 1 ? class_count*(class_count-1)/2 : 1;
+    CvSVMDecisionFunc* df = decision_func;
+
+    for( i = 0; i < df_count; i++ )
+    {
+        int sv_count = df[i].sv_count;
+        if( sv_count != 1 )
+            break;
+    }
+
+    // if every decision functions uses a single support vector;
+    // it's already compressed. skip it then.
+    if( i == df_count )
+        return;
+
+    int var_count = get_var_count();
+    cv::AutoBuffer<double> vbuf(var_count);
+    double* v = vbuf;
+    float** new_sv = (float**)cvMemStorageAlloc(storage, df_count*sizeof(new_sv[0]));
+
+    for( i = 0; i < df_count; i++ )
+    {
+        new_sv[i] = (float*)cvMemStorageAlloc(storage, var_count*sizeof(new_sv[i][0]));
+        float* dst = new_sv[i];
+        memset(v, 0, var_count*sizeof(v[0]));
+        int j, k, sv_count = df[i].sv_count;
+        for( j = 0; j < sv_count; j++ )
+        {
+            const float* src = class_count > 1 && df[i].sv_index ? sv[df[i].sv_index[j]] : sv[j];
+            double a = df[i].alpha[j];
+            for( k = 0; k < var_count; k++ )
+                v[k] += src[k]*a;
+        }
+        for( k = 0; k < var_count; k++ )
+            dst[k] = (float)v[k];
+        df[i].sv_count = 1;
+        df[i].alpha[0] = 1.;
+        if( class_count > 1 && df[i].sv_index )
+            df[i].sv_index[0] = i;
+    }
+
+    sv = new_sv;
+    sv_total = df_count;
+}
+
+
 bool CvSVM::train( const CvMat* _train_data, const CvMat* _responses,
     const CvMat* _var_idx, const CvMat* _sample_idx, CvSVMParams _params )
 {
@@ -1870,7 +1927,7 @@ bool CvSVM::train_auto( const CvMat* _train_data, const CvMat* _responses,
         qsort(ratios, k_fold, sizeof(ratios[0]), icvCmpIndexedratio);
         double old_dist = 0.0;
         for (int k=0; k<k_fold; ++k)
-            old_dist += abs(ratios[k].val-class_ratio);
+            old_dist += cv::abs(ratios[k].val-class_ratio);
         double new_dist = 1.0;
         // iterate to make the folds more balanced
         while (new_dist > 0.0)
@@ -1887,7 +1944,7 @@ bool CvSVM::train_auto( const CvMat* _train_data, const CvMat* _responses,
             qsort(ratios, k_fold, sizeof(ratios[0]), icvCmpIndexedratio);
             new_dist = 0.0;
             for (int k=0; k<k_fold; ++k)
-                new_dist += abs(ratios[k].val-class_ratio);
+                new_dist += cv::abs(ratios[k].val-class_ratio);
             if (new_dist < old_dist)
             {
                 // swapping really improves, so swap the samples
@@ -2135,7 +2192,7 @@ float CvSVM::predict( const CvMat* sample, bool returnDFVal ) const
     return result;
 }
 
-struct predict_body_svm {
+struct predict_body_svm : ParallelLoopBody {
     predict_body_svm(const CvSVM* _pointer, float* _result, const CvMat* _samples, CvMat* _results)
     {
         pointer = _pointer;
@@ -2149,9 +2206,9 @@ struct predict_body_svm {
     const CvMat* samples;
     CvMat* results;
 
-    void operator()( const cv::BlockedRange& range ) const
+    void operator()( const cv::Range& range ) const
     {
-        for(int i = range.begin(); i < range.end(); i++ )
+        for(int i = range.start; i < range.end; i++ )
         {
             CvMat sample;
             cvGetRow( samples, &sample, i );
@@ -2167,7 +2224,7 @@ struct predict_body_svm {
 float CvSVM::predict(const CvMat* samples, CV_OUT CvMat* results) const
 {
     float result = 0;
-    cv::parallel_for(cv::BlockedRange(0, samples->rows),
+    cv::parallel_for_(cv::Range(0, samples->rows),
              predict_body_svm(this, &result, samples, results)
     );
     return result;
@@ -2565,6 +2622,8 @@ void CvSVM::read( CvFileStorage* fs, CvFileNode* svm_node )
         CV_NEXT_SEQ_ELEM( df_node->data.seq->elem_size, reader );
     }
 
+    if( cvReadIntByName(fs, svm_node, "optimize_linear", 1) != 0 )
+        optimize_linear_svm();
     create_kernel();
 
     __END__;
@@ -2939,4 +2998,3 @@ cvTrainSVM_CrossValidation( const CvMat* train_data, int tflag,
 #endif
 
 /* End of file. */
-
diff --git a/modules/ml/src/tree.cpp b/modules/ml/src/tree.cpp
index 6f128ebdd1..3a67cdd375 100644
--- a/modules/ml/src/tree.cpp
+++ b/modules/ml/src/tree.cpp
@@ -120,16 +120,27 @@ bool CvDTreeTrainData::set_params( const CvDTreeParams& _params )
     return ok;
 }
 
-#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
-static CV_IMPLEMENT_QSORT_EX( icvSortIntPtr, int*, CV_CMP_NUM_PTR, int )
-static CV_IMPLEMENT_QSORT_EX( icvSortDblPtr, double*, CV_CMP_NUM_PTR, int )
+template<typename T>
+class LessThanPtr
+{
+public:
+    bool operator()(T* a, T* b) const { return *a < *b; }
+};
 
-#define CV_CMP_NUM_IDX(i,j) (aux[i] < aux[j])
-static CV_IMPLEMENT_QSORT_EX( icvSortIntAux, int, CV_CMP_NUM_IDX, const float* )
-static CV_IMPLEMENT_QSORT_EX( icvSortUShAux, unsigned short, CV_CMP_NUM_IDX, const float* )
+template<typename T, typename Idx>
+class LessThanIdx
+{
+public:
+    LessThanIdx( const T* _arr ) : arr(_arr) {}
+    bool operator()(Idx a, Idx b) const { return arr[a] < arr[b]; }
+    const T* arr;
+};
 
-#define CV_CMP_PAIRS(a,b) (*((a).i) < *((b).i))
-static CV_IMPLEMENT_QSORT_EX( icvSortPairs, CvPair16u32s, CV_CMP_PAIRS, int )
+class LessThanPairs
+{
+public:
+    bool operator()(const CvPair16u32s& a, const CvPair16u32s& b) const { return *a.i < *b.i; }
+};
 
 void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
     const CvMat* _responses, const CvMat* _var_idx, const CvMat* _sample_idx,
@@ -461,7 +472,7 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
             c_count = num_valid > 0;
             if (is_buf_16u)
             {
-                icvSortPairs( pair16u32s_ptr, sample_count, 0 );
+                std::sort(pair16u32s_ptr, pair16u32s_ptr + sample_count, LessThanPairs());
                 // count the categories
                 for( i = 1; i < num_valid; i++ )
                     if (*pair16u32s_ptr[i].i != *pair16u32s_ptr[i-1].i)
@@ -469,7 +480,7 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
             }
             else
             {
-                icvSortIntPtr( int_ptr, sample_count, 0 );
+                std::sort(int_ptr, int_ptr + sample_count, LessThanPtr<int>());
                 // count the categories
                 for( i = 1; i < num_valid; i++ )
                     c_count += *int_ptr[i] != *int_ptr[i-1];
@@ -561,9 +572,9 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
 
             }
             if (is_buf_16u)
-                icvSortUShAux( udst, sample_count, _fdst);
+                std::sort(udst, udst + sample_count, LessThanIdx<float, unsigned short>(_fdst));
             else
-                icvSortIntAux( idst, sample_count, _fdst );
+                std::sort(idst, idst + sample_count, LessThanIdx<float, int>(_fdst));
         }
 
         if( vi < var_count )
@@ -1583,9 +1594,14 @@ bool CvDTree::train( const Mat& _train_data, int _tflag,
                     const Mat& _sample_idx, const Mat& _var_type,
                     const Mat& _missing_mask, CvDTreeParams _params )
 {
-    CvMat tdata = _train_data, responses = _responses, vidx=_var_idx,
-        sidx=_sample_idx, vtype=_var_type, mmask=_missing_mask;
-    return train(&tdata, _tflag, &responses, vidx.data.ptr ? &vidx : 0, sidx.data.ptr ? &sidx : 0,
+    train_data_hdr = _train_data;
+    train_data_mat = _train_data;
+    responses_hdr = _responses;
+    responses_mat = _responses;
+
+    CvMat vidx=_var_idx, sidx=_sample_idx, vtype=_var_type, mmask=_missing_mask;
+
+    return train(&train_data_hdr, _tflag, &responses_hdr, vidx.data.ptr ? &vidx : 0, sidx.data.ptr ? &sidx : 0,
                  vtype.data.ptr ? &vtype : 0, mmask.data.ptr ? &mmask : 0, _params);
 }
 
@@ -1866,7 +1882,7 @@ double CvDTree::calc_node_dir( CvDTreeNode* node )
 namespace cv
 {
 
-template<> CV_EXPORTS void Ptr<CvDTreeSplit>::delete_obj()
+template<> CV_EXPORTS void DefaultDeleter<CvDTreeSplit>::operator ()(CvDTreeSplit* obj) const
 {
     fastFree(obj);
 }
@@ -1877,12 +1893,12 @@ DTreeBestSplitFinder::DTreeBestSplitFinder( CvDTree* _tree, CvDTreeNode* _node)
     node = _node;
     splitSize = tree->get_data()->split_heap->elem_size;
 
-    bestSplit = (CvDTreeSplit*)fastMalloc(splitSize);
-    memset((CvDTreeSplit*)bestSplit, 0, splitSize);
+    bestSplit.reset((CvDTreeSplit*)fastMalloc(splitSize));
+    memset(bestSplit.get(), 0, splitSize);
     bestSplit->quality = -1;
     bestSplit->condensed_idx = INT_MIN;
-    split = (CvDTreeSplit*)fastMalloc(splitSize);
-    memset((CvDTreeSplit*)split, 0, splitSize);
+    split.reset((CvDTreeSplit*)fastMalloc(splitSize));
+    memset(split.get(), 0, splitSize);
     //haveSplit = false;
 }
 
@@ -1892,10 +1908,10 @@ DTreeBestSplitFinder::DTreeBestSplitFinder( const DTreeBestSplitFinder& finder,
     node = finder.node;
     splitSize = tree->get_data()->split_heap->elem_size;
 
-    bestSplit = (CvDTreeSplit*)fastMalloc(splitSize);
-    memcpy((CvDTreeSplit*)(bestSplit), (const CvDTreeSplit*)finder.bestSplit, splitSize);
-    split = (CvDTreeSplit*)fastMalloc(splitSize);
-    memset((CvDTreeSplit*)split, 0, splitSize);
+    bestSplit.reset((CvDTreeSplit*)fastMalloc(splitSize));
+    memcpy(bestSplit.get(), finder.bestSplit.get(), splitSize);
+    split.reset((CvDTreeSplit*)fastMalloc(splitSize));
+    memset(split.get(), 0, splitSize);
 }
 
 void DTreeBestSplitFinder::operator()(const BlockedRange& range)
@@ -1928,14 +1944,14 @@ void DTreeBestSplitFinder::operator()(const BlockedRange& range)
         }
 
         if( res && bestSplit->quality < split->quality )
-                memcpy( (CvDTreeSplit*)bestSplit, (CvDTreeSplit*)split, splitSize );
+                memcpy( bestSplit.get(), split.get(), splitSize );
     }
 }
 
 void DTreeBestSplitFinder::join( DTreeBestSplitFinder& rhs )
 {
     if( bestSplit->quality < rhs.bestSplit->quality )
-        memcpy( (CvDTreeSplit*)bestSplit, (CvDTreeSplit*)rhs.bestSplit, splitSize );
+        memcpy( bestSplit.get(), rhs.bestSplit.get(), splitSize );
 }
 }
 
@@ -2239,7 +2255,7 @@ CvDTreeSplit* CvDTree::find_split_cat_class( CvDTreeNode* node, int vi, float in
         int_ptr = (int**)(c_weights + _mi);
         for( j = 0; j < mi; j++ )
             int_ptr[j] = cjk + j*2 + 1;
-        icvSortIntPtr( int_ptr, mi, 0 );
+        std::sort(int_ptr, int_ptr + mi, LessThanPtr<int>());
         subset_i = 0;
         subset_n = mi;
     }
@@ -2466,7 +2482,7 @@ CvDTreeSplit* CvDTree::find_split_cat_reg( CvDTreeNode* node, int vi, float init
         sum_ptr[i] = sum + i;
     }
 
-    icvSortDblPtr( sum_ptr, mi, 0 );
+    std::sort(sum_ptr, sum_ptr + mi, LessThanPtr<double>());
 
     // revert back to unnormalized sums
     // (there should be a very little loss of accuracy)
@@ -4131,7 +4147,7 @@ void CvDTree::read( CvFileStorage* fs, CvFileNode* node, CvDTreeTrainData* _data
 
 Mat CvDTree::getVarImportance()
 {
-    return Mat(get_var_importance());
+    return cvarrToMat(get_var_importance());
 }
 
 /* End of file. */
diff --git a/modules/ml/test/test_emknearestkmeans.cpp b/modules/ml/test/test_emknearestkmeans.cpp
index b75e477f3a..a14b636061 100644
--- a/modules/ml/test/test_emknearestkmeans.cpp
+++ b/modules/ml/test/test_emknearestkmeans.cpp
@@ -597,7 +597,7 @@ protected:
             ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
         }
 
-        Mat values = data.get_values();
+        Mat values = cv::cvarrToMat(data.get_values());
         CV_Assert(values.cols == 58);
         int responseIndex = 57;
 
@@ -678,4 +678,3 @@ TEST(ML_KNearest, accuracy) { CV_KNearestTest test; test.safe_run(); }
 TEST(ML_EM, accuracy) { CV_EMTest test; test.safe_run(); }
 TEST(ML_EM, save_load) { CV_EMTest_SaveLoad test; test.safe_run(); }
 TEST(ML_EM, classification) { CV_EMTest_Classification test; test.safe_run(); }
-
diff --git a/modules/ml/test/test_mltests2.cpp b/modules/ml/test/test_mltests2.cpp
index 80776b4ced..4f6c95921e 100644
--- a/modules/ml/test/test_mltests2.cpp
+++ b/modules/ml/test/test_mltests2.cpp
@@ -179,7 +179,7 @@ float knearest_calc_error( CvKNearest* knearest, CvMLData* _data, int k, int typ
 }
 
 // 3. svm
-int str_to_svm_type(string& str)
+int str_to_svm_type(String& str)
 {
     if( !str.compare("C_SVC") )
         return CvSVM::C_SVC;
@@ -194,7 +194,7 @@ int str_to_svm_type(string& str)
     CV_Error( CV_StsBadArg, "incorrect svm type string" );
     return -1;
 }
-int str_to_svm_kernel_type( string& str )
+int str_to_svm_kernel_type( String& str )
 {
     if( !str.compare("LINEAR") )
         return CvSVM::LINEAR;
@@ -297,7 +297,7 @@ float svm_calc_error( CvSVM* svm, CvMLData* _data, int type, vector<float> *resp
 
 // 4. em
 // 5. ann
-int str_to_ann_train_method( string& str )
+int str_to_ann_train_method( String& str )
 {
     if( !str.compare("BACKPROP") )
         return CvANN_MLP_TrainParams::BACKPROP;
@@ -397,7 +397,7 @@ float ann_calc_error( CvANN_MLP* ann, CvMLData* _data, map<int, int>& cls_map, i
         int si = sidx ? sidx[i] : i;
         cvGetRow( &predictors, &sample, si );
         ann->predict( &sample, &_output );
-        CvPoint best_cls = {0,0};
+        CvPoint best_cls;
         cvMinMaxLoc( &_output, 0, 0, 0, &best_cls, 0 );
         int r = cvRound(responses->data.fl[si*r_step]);
         CV_DbgAssert( fabs(responses->data.fl[si*r_step]-r) < FLT_EPSILON );
@@ -412,7 +412,7 @@ float ann_calc_error( CvANN_MLP* ann, CvMLData* _data, map<int, int>& cls_map, i
 
 // 6. dtree
 // 7. boost
-int str_to_boost_type( string& str )
+int str_to_boost_type( String& str )
 {
     if ( !str.compare("DISCRETE") )
         return CvBoost::DISCRETE;
@@ -546,7 +546,7 @@ void CV_MLBaseTest::run( int )
 int CV_MLBaseTest::prepare_test_case( int test_case_idx )
 {
     int trainSampleCount, respIdx;
-    string varTypes;
+    String varTypes;
     clear();
 
     string dataPath = ts->get_data_path();
@@ -605,7 +605,7 @@ int CV_MLBaseTest::train( int testCaseIdx )
     }
     else if( !modelName.compare(CV_SVM) )
     {
-        string svm_type_str, kernel_type_str;
+        String svm_type_str, kernel_type_str;
         modelParamsNode["svm_type"] >> svm_type_str;
         modelParamsNode["kernel_type"] >> kernel_type_str;
         CvSVMParams params;
@@ -625,7 +625,7 @@ int CV_MLBaseTest::train( int testCaseIdx )
     }
     else if( !modelName.compare(CV_ANN) )
     {
-        string train_method_str;
+        String train_method_str;
         double param1, param2;
         modelParamsNode["train_method"] >> train_method_str;
         modelParamsNode["param1"] >> param1;
@@ -659,7 +659,7 @@ int CV_MLBaseTest::train( int testCaseIdx )
         int BOOST_TYPE, WEAK_COUNT, MAX_DEPTH;
         float WEIGHT_TRIM_RATE;
         bool USE_SURROGATE;
-        string typeStr;
+        String typeStr;
         modelParamsNode["type"] >> typeStr;
         BOOST_TYPE = str_to_boost_type( typeStr );
         modelParamsNode["weak_count"] >> WEAK_COUNT;
@@ -769,7 +769,11 @@ void CV_MLBaseTest::load( const char* filename )
     else if( !modelName.compare(CV_KNEAREST) )
         knearest->load( filename );
     else if( !modelName.compare(CV_SVM) )
+    {
+        delete svm;
+        svm = new CvSVM;
         svm->load( filename );
+    }
     else if( !modelName.compare(CV_ANN) )
         ann->load( filename );
     else if( !modelName.compare(CV_DTREE) )
diff --git a/modules/ml/test/test_precomp.cpp b/modules/ml/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/ml/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/ml/test/test_precomp.hpp b/modules/ml/test/test_precomp.hpp
index f311d48786..e68e5513ba 100644
--- a/modules/ml/test/test_precomp.hpp
+++ b/modules/ml/test/test_precomp.hpp
@@ -9,11 +9,11 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
+#include <map>
 #include "opencv2/ts.hpp"
 #include "opencv2/ml.hpp"
 #include "opencv2/core/core_c.h"
-#include <iostream>
-#include <map>
 
 #define CV_NBAYES   "nbayes"
 #define CV_KNEAREST "knearest"
diff --git a/modules/ml/test/test_save_load.cpp b/modules/ml/test/test_save_load.cpp
index 889b98b62b..9fd31b9f24 100644
--- a/modules/ml/test/test_save_load.cpp
+++ b/modules/ml/test/test_save_load.cpp
@@ -82,32 +82,53 @@ int CV_SLMLTest::validate_test_results( int testCaseIdx )
     int code = cvtest::TS::OK;
 
     // 1. compare files
-    ifstream f1( fname1.c_str() ), f2( fname2.c_str() );
-    string s1, s2;
-    int lineIdx = 0;
-    CV_Assert( f1.is_open() && f2.is_open() );
-    for( ; !f1.eof() && !f2.eof(); lineIdx++ )
+    FILE *fs1 = fopen(fname1.c_str(), "rb"), *fs2 = fopen(fname2.c_str(), "rb");
+    size_t sz1 = 0, sz2 = 0;
+    if( !fs1 || !fs2 )
+        code = cvtest::TS::FAIL_MISSING_TEST_DATA;
+    if( code >= 0 )
     {
-        getline( f1, s1 );
-        getline( f2, s2 );
-        if( s1.compare(s2) )
+        fseek(fs1, 0, SEEK_END); fseek(fs2, 0, SEEK_END);
+        sz1 = ftell(fs1);
+        sz2 = ftell(fs2);
+        fseek(fs1, 0, SEEK_SET); fseek(fs2, 0, SEEK_SET);
+    }
+
+    if( sz1 != sz2 )
+        code = cvtest::TS::FAIL_INVALID_OUTPUT;
+
+    if( code >= 0 )
+    {
+        const int BUFSZ = 1024;
+        uchar buf1[BUFSZ], buf2[BUFSZ];
+        for( size_t pos = 0; pos < sz1;  )
         {
-            ts->printf( cvtest::TS::LOG, "first and second saved files differ in %n-line; first %n line: %s; second %n-line: %s",
-               lineIdx, lineIdx, s1.c_str(), lineIdx, s2.c_str() );
-            code = cvtest::TS::FAIL_INVALID_OUTPUT;
+            size_t r1 = fread(buf1, 1, BUFSZ, fs1);
+            size_t r2 = fread(buf2, 1, BUFSZ, fs2);
+            if( r1 != r2 || memcmp(buf1, buf2, r1) != 0 )
+            {
+                ts->printf( cvtest::TS::LOG,
+                           "in test case %d first (%s) and second (%s) saved files differ in %d-th kb\n",
+                           testCaseIdx, fname1.c_str(), fname2.c_str(),
+                           (int)pos );
+                code = cvtest::TS::FAIL_INVALID_OUTPUT;
+                break;
+            }
+            pos += r1;
         }
     }
-    if( !f1.eof() || !f2.eof() )
-    {
-        ts->printf( cvtest::TS::LOG, "in test case %d first and second saved files differ in %n-line; first %n line: %s; second %n-line: %s",
-            testCaseIdx, lineIdx, lineIdx, s1.c_str(), lineIdx, s2.c_str() );
-        code = cvtest::TS::FAIL_INVALID_OUTPUT;
-    }
-    f1.close();
-    f2.close();
+
+    if(fs1)
+        fclose(fs1);
+    if(fs2)
+        fclose(fs2);
+
     // delete temporary files
-    remove( fname1.c_str() );
-    remove( fname2.c_str() );
+    if( code >= 0 )
+    {
+        remove( fname1.c_str() );
+        remove( fname2.c_str() );
+    }
 
     // 2. compare responses
     CV_Assert( test_resps1.size() == test_resps2.size() );
@@ -133,4 +154,32 @@ TEST(ML_Boost, save_load) { CV_SLMLTest test( CV_BOOST ); test.safe_run(); }
 TEST(ML_RTrees, save_load) { CV_SLMLTest test( CV_RTREES ); test.safe_run(); }
 TEST(ML_ERTrees, save_load) { CV_SLMLTest test( CV_ERTREES ); test.safe_run(); }
 
+
+TEST(DISABLED_ML_SVM, linear_save_load)
+{
+    CvSVM svm1, svm2, svm3;
+    svm1.load("SVM45_X_38-1.xml");
+    svm2.load("SVM45_X_38-2.xml");
+    string tname = tempfile("a.xml");
+    svm2.save(tname.c_str());
+    svm3.load(tname.c_str());
+
+    ASSERT_EQ(svm1.get_var_count(), svm2.get_var_count());
+    ASSERT_EQ(svm1.get_var_count(), svm3.get_var_count());
+
+    int m = 10000, n = svm1.get_var_count();
+    Mat samples(m, n, CV_32F), r1, r2, r3;
+    randu(samples, 0., 1.);
+
+    svm1.predict(samples, r1);
+    svm2.predict(samples, r2);
+    svm3.predict(samples, r3);
+
+    double eps = 1e-4;
+    EXPECT_LE(norm(r1, r2, NORM_INF), eps);
+    EXPECT_LE(norm(r1, r3, NORM_INF), eps);
+
+    remove(tname.c_str());
+}
+
 /* End of file. */
diff --git a/modules/nonfree/CMakeLists.txt b/modules/nonfree/CMakeLists.txt
index eeaf53a6f8..fa73931aab 100644
--- a/modules/nonfree/CMakeLists.txt
+++ b/modules/nonfree/CMakeLists.txt
@@ -3,4 +3,5 @@ if(BUILD_ANDROID_PACKAGE)
 endif()
 
 set(the_description "Functionality with possible limitations on the use")
-ocv_define_module(nonfree opencv_imgproc opencv_features2d opencv_calib3d)
+ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef)
+ocv_define_module(nonfree opencv_imgproc opencv_features2d opencv_calib3d OPTIONAL opencv_gpuarithm opencv_ocl)
diff --git a/modules/nonfree/doc/feature_detection.rst b/modules/nonfree/doc/feature_detection.rst
index e4ac357429..9f66271aac 100644
--- a/modules/nonfree/doc/feature_detection.rst
+++ b/modules/nonfree/doc/feature_detection.rst
@@ -16,6 +16,8 @@ The SIFT constructors.
 
 .. ocv:function:: SIFT::SIFT( int nfeatures=0, int nOctaveLayers=3, double contrastThreshold=0.04, double edgeThreshold=10, double sigma=1.6)
 
+.. ocv:pyfunction:: cv2.SIFT([, nfeatures[, nOctaveLayers[, contrastThreshold[, edgeThreshold[, sigma]]]]]) -> <SIFT object>
+
     :param nfeatures: The number of best features to retain. The features are ranked by their scores (measured in SIFT algorithm as the local contrast)
 
     :param nOctaveLayers: The number of layers in each octave. 3 is the value used in D. Lowe paper. The number of octaves is computed automatically from the image resolution.
@@ -33,6 +35,12 @@ Extract features and computes their descriptors using SIFT algorithm
 
 .. ocv:function:: void SIFT::operator()(InputArray img, InputArray mask, vector<KeyPoint>& keypoints, OutputArray descriptors, bool useProvidedKeypoints=false)
 
+.. ocv:pyfunction:: cv2.SIFT.detect(image[, mask]) -> keypoints
+
+.. ocv:pyfunction:: cv2.SIFT.compute(image, keypoints[, descriptors]) -> keypoints, descriptors
+
+.. ocv:pyfunction:: cv2.SIFT.detectAndCompute(image, mask[, descriptors[, useProvidedKeypoints]]) -> keypoints, descriptors
+
     :param img: Input 8-bit grayscale image
 
     :param mask: Optional input mask that marks the regions where we should detect features.
@@ -43,6 +51,7 @@ Extract features and computes their descriptors using SIFT algorithm
 
     :param useProvidedKeypoints: Boolean flag. If it is true, the keypoint detector is not run. Instead, the provided vector of keypoints is used and the algorithm just computes their descriptors.
 
+.. note:: Python API provides three functions. First one finds keypoints only. Second function computes the descriptors based on the keypoints we provide. Third function detects the keypoints and computes their descriptors. If you want both keypoints and descriptors, directly use third function as ``kp, des = cv2.SIFT.detectAndCompute(image, None)``
 
 SURF
 ----
@@ -75,6 +84,10 @@ SURF
 
 .. [Bay06] Bay, H. and Tuytelaars, T. and Van Gool, L. "SURF: Speeded Up Robust Features", 9th European Conference on Computer Vision, 2006
 
+.. note::
+
+   * An example using the SURF feature detector can be found at opencv_source_code/samples/cpp/generic_descriptor_match.cpp
+   * Another example using the SURF feature detector, extractor and matcher can be found at opencv_source_code/samples/cpp/matcher_simple.cpp
 
 SURF::SURF
 ----------
@@ -105,11 +118,11 @@ Detects keypoints and computes SURF descriptors for them.
 .. ocv:function:: void SURF::operator()(InputArray img, InputArray mask, vector<KeyPoint>& keypoints, OutputArray descriptors, bool useProvidedKeypoints=false)
 
 .. ocv:pyfunction:: cv2.SURF.detect(image[, mask]) -> keypoints
+.. ocv:pyfunction:: cv2.SURF.compute(image, keypoints[, descriptors]) -> keypoints, descriptors
+.. ocv:pyfunction:: cv2.SURF.detectAndCompute(image, mask[, descriptors[, useProvidedKeypoints]]) -> keypoints, descriptors
 
 .. ocv:cfunction:: void cvExtractSURF( const CvArr* image, const CvArr* mask, CvSeq** keypoints, CvSeq** descriptors, CvMemStorage* storage, CvSURFParams params )
 
-.. ocv:pyoldfunction:: cv.ExtractSURF(image, mask, storage, params)-> (keypoints, descriptors)
-
     :param image: Input 8-bit grayscale image
 
     :param mask: Optional input mask that marks the regions where we should detect features.
@@ -127,3 +140,211 @@ Detects keypoints and computes SURF descriptors for them.
 The function is parallelized with the TBB library.
 
 If you are using the C version, make sure you call ``cv::initModule_nonfree()`` from ``nonfree/nonfree.hpp``.
+
+
+gpu::SURF_GPU
+-------------
+.. ocv:class:: gpu::SURF_GPU
+
+Class used for extracting Speeded Up Robust Features (SURF) from an image. ::
+
+    class SURF_GPU
+    {
+    public:
+        enum KeypointLayout
+        {
+            X_ROW = 0,
+            Y_ROW,
+            LAPLACIAN_ROW,
+            OCTAVE_ROW,
+            SIZE_ROW,
+            ANGLE_ROW,
+            HESSIAN_ROW,
+            ROWS_COUNT
+        };
+
+        //! the default constructor
+        SURF_GPU();
+        //! the full constructor taking all the necessary parameters
+        explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
+             int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f);
+
+        //! returns the descriptor size in float's (64 or 128)
+        int descriptorSize() const;
+
+        //! upload host keypoints to device memory
+        void uploadKeypoints(const vector<KeyPoint>& keypoints,
+            GpuMat& keypointsGPU);
+        //! download keypoints from device to host memory
+        void downloadKeypoints(const GpuMat& keypointsGPU,
+            vector<KeyPoint>& keypoints);
+
+        //! download descriptors from device to host memory
+        void downloadDescriptors(const GpuMat& descriptorsGPU,
+            vector<float>& descriptors);
+
+        void operator()(const GpuMat& img, const GpuMat& mask,
+            GpuMat& keypoints);
+
+        void operator()(const GpuMat& img, const GpuMat& mask,
+            GpuMat& keypoints, GpuMat& descriptors,
+            bool useProvidedKeypoints = false,
+            bool calcOrientation = true);
+
+        void operator()(const GpuMat& img, const GpuMat& mask,
+            std::vector<KeyPoint>& keypoints);
+
+        void operator()(const GpuMat& img, const GpuMat& mask,
+            std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
+            bool useProvidedKeypoints = false,
+            bool calcOrientation = true);
+
+        void operator()(const GpuMat& img, const GpuMat& mask,
+            std::vector<KeyPoint>& keypoints,
+            std::vector<float>& descriptors,
+            bool useProvidedKeypoints = false,
+            bool calcOrientation = true);
+
+        void releaseMemory();
+
+        // SURF parameters
+        double hessianThreshold;
+        int nOctaves;
+        int nOctaveLayers;
+        bool extended;
+        bool upright;
+
+        //! max keypoints = keypointsRatio * img.size().area()
+        float keypointsRatio;
+
+        GpuMat sum, mask1, maskSum, intBuffer;
+
+        GpuMat det, trace;
+
+        GpuMat maxPosBuffer;
+    };
+
+
+The class ``SURF_GPU`` implements Speeded Up Robust Features descriptor. There is a fast multi-scale Hessian keypoint detector that can be used to find the keypoints (which is the default option). But the descriptors can also be computed for the user-specified keypoints. Only 8-bit grayscale images are supported.
+
+The class ``SURF_GPU`` can store results in the GPU and CPU memory. It provides functions to convert results between CPU and GPU version ( ``uploadKeypoints``, ``downloadKeypoints``, ``downloadDescriptors`` ). The format of CPU results is the same as ``SURF`` results. GPU results are stored in ``GpuMat``. The ``keypoints`` matrix is :math:`\texttt{nFeatures} \times 7` matrix with the ``CV_32FC1`` type.
+
+* ``keypoints.ptr<float>(X_ROW)[i]`` contains x coordinate of the i-th feature.
+* ``keypoints.ptr<float>(Y_ROW)[i]`` contains y coordinate of the i-th feature.
+* ``keypoints.ptr<float>(LAPLACIAN_ROW)[i]``  contains the laplacian sign of the i-th feature.
+* ``keypoints.ptr<float>(OCTAVE_ROW)[i]`` contains the octave of the i-th feature.
+* ``keypoints.ptr<float>(SIZE_ROW)[i]`` contains the size of the i-th feature.
+* ``keypoints.ptr<float>(ANGLE_ROW)[i]`` contain orientation of the i-th feature.
+* ``keypoints.ptr<float>(HESSIAN_ROW)[i]`` contains the response of the i-th feature.
+
+The ``descriptors`` matrix is :math:`\texttt{nFeatures} \times \texttt{descriptorSize}` matrix with the ``CV_32FC1`` type.
+
+The class ``SURF_GPU`` uses some buffers and provides access to it. All buffers can be safely released between function calls.
+
+.. seealso:: :ocv:class:`SURF`
+
+.. note::
+
+   * An example for using the SURF keypoint matcher on GPU can be found at opencv_source_code/samples/gpu/surf_keypoint_matcher.cpp
+
+ocl::SURF_OCL
+-------------
+.. ocv:class:: ocl::SURF_OCL
+
+Class used for extracting Speeded Up Robust Features (SURF) from an image. ::
+
+    class SURF_OCL
+    {
+    public:
+        enum KeypointLayout
+        {
+            X_ROW = 0,
+            Y_ROW,
+            LAPLACIAN_ROW,
+            OCTAVE_ROW,
+            SIZE_ROW,
+            ANGLE_ROW,
+            HESSIAN_ROW,
+            ROWS_COUNT
+        };
+
+        //! the default constructor
+        SURF_OCL();
+        //! the full constructor taking all the necessary parameters
+        explicit SURF_OCL(double _hessianThreshold, int _nOctaves=4,
+             int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
+
+        //! returns the descriptor size in float's (64 or 128)
+        int descriptorSize() const;
+
+        //! upload host keypoints to device memory
+        void uploadKeypoints(const vector<KeyPoint>& keypoints,
+            oclMat& keypointsocl);
+        //! download keypoints from device to host memory
+        void downloadKeypoints(const oclMat& keypointsocl,
+            vector<KeyPoint>& keypoints);
+
+        //! download descriptors from device to host memory
+        void downloadDescriptors(const oclMat& descriptorsocl,
+            vector<float>& descriptors);
+
+        void operator()(const oclMat& img, const oclMat& mask,
+            oclMat& keypoints);
+
+        void operator()(const oclMat& img, const oclMat& mask,
+            oclMat& keypoints, oclMat& descriptors,
+            bool useProvidedKeypoints = false);
+
+        void operator()(const oclMat& img, const oclMat& mask,
+            std::vector<KeyPoint>& keypoints);
+
+        void operator()(const oclMat& img, const oclMat& mask,
+            std::vector<KeyPoint>& keypoints, oclMat& descriptors,
+            bool useProvidedKeypoints = false);
+
+        void operator()(const oclMat& img, const oclMat& mask,
+            std::vector<KeyPoint>& keypoints,
+            std::vector<float>& descriptors,
+            bool useProvidedKeypoints = false);
+
+        void releaseMemory();
+
+        // SURF parameters
+        double hessianThreshold;
+        int nOctaves;
+        int nOctaveLayers;
+        bool extended;
+        bool upright;
+
+        //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
+        float keypointsRatio;
+
+        oclMat sum, mask1, maskSum, intBuffer;
+
+        oclMat det, trace;
+
+        oclMat maxPosBuffer;
+    };
+
+
+The class ``SURF_OCL`` implements Speeded Up Robust Features descriptor. There is a fast multi-scale Hessian keypoint detector that can be used to find the keypoints (which is the default option). But the descriptors can also be computed for the user-specified keypoints. Only 8-bit grayscale images are supported.
+
+The class ``SURF_OCL`` can store results in the GPU and CPU memory. It provides functions to convert results between CPU and GPU version ( ``uploadKeypoints``, ``downloadKeypoints``, ``downloadDescriptors`` ). The format of CPU results is the same as ``SURF`` results. GPU results are stored in ``oclMat``. The ``keypoints`` matrix is :math:`\texttt{nFeatures} \times 7` matrix with the ``CV_32FC1`` type.
+
+* ``keypoints.ptr<float>(X_ROW)[i]`` contains x coordinate of the i-th feature.
+* ``keypoints.ptr<float>(Y_ROW)[i]`` contains y coordinate of the i-th feature.
+* ``keypoints.ptr<float>(LAPLACIAN_ROW)[i]``  contains the laplacian sign of the i-th feature.
+* ``keypoints.ptr<float>(OCTAVE_ROW)[i]`` contains the octave of the i-th feature.
+* ``keypoints.ptr<float>(SIZE_ROW)[i]`` contains the size of the i-th feature.
+* ``keypoints.ptr<float>(ANGLE_ROW)[i]`` contain orientation of the i-th feature.
+* ``keypoints.ptr<float>(HESSIAN_ROW)[i]`` contains the response of the i-th feature.
+
+The ``descriptors`` matrix is :math:`\texttt{nFeatures} \times \texttt{descriptorSize}` matrix with the ``CV_32FC1`` type.
+
+The class ``SURF_OCL`` uses some buffers and provides access to it. All buffers can be safely released between function calls.
+
+.. seealso:: :ocv:class:`SURF`
+
+.. note::
+
+   * OCL : An example of the SURF detector can be found at opencv_source_code/samples/ocl/surf_matcher.cpp
diff --git a/modules/nonfree/include/opencv2/nonfree/features2d.hpp b/modules/nonfree/include/opencv2/nonfree/features2d.hpp
index 57b9f1e688..4acd747c15 100644
--- a/modules/nonfree/include/opencv2/nonfree/features2d.hpp
+++ b/modules/nonfree/include/opencv2/nonfree/features2d.hpp
@@ -45,8 +45,6 @@
 
 #include "opencv2/features2d.hpp"
 
-#ifdef __cplusplus
-
 namespace cv
 {
 
@@ -58,9 +56,9 @@ namespace cv
 class CV_EXPORTS_W SIFT : public Feature2D
 {
 public:
-    CV_WRAP explicit SIFT( int nfeatures=0, int nOctaveLayers=3,
-          double contrastThreshold=0.04, double edgeThreshold=10,
-          double sigma=1.6);
+    CV_WRAP explicit SIFT( int nfeatures = 0, int nOctaveLayers = 3,
+          double contrastThreshold = 0.04, double edgeThreshold = 10,
+          double sigma = 1.6);
 
     //! returns the descriptor size in floats (128)
     CV_WRAP int descriptorSize() const;
@@ -76,7 +74,7 @@ public:
     void operator()(InputArray img, InputArray mask,
                     std::vector<KeyPoint>& keypoints,
                     OutputArray descriptors,
-                    bool useProvidedKeypoints=false) const;
+                    bool useProvidedKeypoints = false) const;
 
     AlgorithmInfo* info() const;
 
@@ -86,7 +84,7 @@ public:
                                 std::vector<KeyPoint>& keypoints ) const;
 
 protected:
-    void detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask=Mat() ) const;
+    void detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask = Mat() ) const;
     void computeImpl( const Mat& image, std::vector<KeyPoint>& keypoints, Mat& descriptors ) const;
 
     CV_PROP_RW int nfeatures;
@@ -111,8 +109,8 @@ public:
     CV_WRAP SURF();
     //! the full constructor taking all the necessary parameters
     explicit CV_WRAP SURF(double hessianThreshold,
-                  int nOctaves=4, int nOctaveLayers=2,
-                  bool extended=true, bool upright=false);
+                  int nOctaves = 4, int nOctaveLayers = 2,
+                  bool extended = true, bool upright = false);
 
     //! returns the descriptor size in float's (64 or 128)
     CV_WRAP int descriptorSize() const;
@@ -127,7 +125,7 @@ public:
     void operator()(InputArray img, InputArray mask,
                     CV_OUT std::vector<KeyPoint>& keypoints,
                     OutputArray descriptors,
-                    bool useProvidedKeypoints=false) const;
+                    bool useProvidedKeypoints = false) const;
 
     AlgorithmInfo* info() const;
 
@@ -139,7 +137,7 @@ public:
 
 protected:
 
-    void detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask=Mat() ) const;
+    void detectImpl( const Mat& image, std::vector<KeyPoint>& keypoints, const Mat& mask = Mat() ) const;
     void computeImpl( const Mat& image, std::vector<KeyPoint>& keypoints, Mat& descriptors ) const;
 };
 
@@ -148,8 +146,4 @@ typedef SURF SurfDescriptorExtractor;
 
 } /* namespace cv */
 
-#endif /* __cplusplus */
-
 #endif
-
-/* End of file. */
diff --git a/modules/nonfree/include/opencv2/nonfree/gpu.hpp b/modules/nonfree/include/opencv2/nonfree/gpu.hpp
new file mode 100644
index 0000000000..b94a99cf72
--- /dev/null
+++ b/modules/nonfree/include/opencv2/nonfree/gpu.hpp
@@ -0,0 +1,126 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_NONFREE_GPU_HPP__
+#define __OPENCV_NONFREE_GPU_HPP__
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpu {
+
+class CV_EXPORTS SURF_GPU
+{
+public:
+    enum KeypointLayout
+    {
+        X_ROW = 0,
+        Y_ROW,
+        LAPLACIAN_ROW,
+        OCTAVE_ROW,
+        SIZE_ROW,
+        ANGLE_ROW,
+        HESSIAN_ROW,
+        ROWS_COUNT
+    };
+
+    //! the default constructor
+    SURF_GPU();
+    //! the full constructor taking all the necessary parameters
+    explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
+         int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
+
+    //! returns the descriptor size in float's (64 or 128)
+    int descriptorSize() const;
+
+    //! upload host keypoints to device memory
+    void uploadKeypoints(const std::vector<KeyPoint>& keypoints, GpuMat& keypointsGPU);
+    //! download keypoints from device to host memory
+    void downloadKeypoints(const GpuMat& keypointsGPU, std::vector<KeyPoint>& keypoints);
+
+    //! download descriptors from device to host memory
+    void downloadDescriptors(const GpuMat& descriptorsGPU, std::vector<float>& descriptors);
+
+    //! finds the keypoints using fast hessian detector used in SURF
+    //! supports CV_8UC1 images
+    //! keypoints will have nFeature cols and 6 rows
+    //! keypoints.ptr<float>(X_ROW)[i] will contain x coordinate of i'th feature
+    //! keypoints.ptr<float>(Y_ROW)[i] will contain y coordinate of i'th feature
+    //! keypoints.ptr<float>(LAPLACIAN_ROW)[i] will contain laplacian sign of i'th feature
+    //! keypoints.ptr<float>(OCTAVE_ROW)[i] will contain octave of i'th feature
+    //! keypoints.ptr<float>(SIZE_ROW)[i] will contain size of i'th feature
+    //! keypoints.ptr<float>(ANGLE_ROW)[i] will contain orientation of i'th feature
+    //! keypoints.ptr<float>(HESSIAN_ROW)[i] will contain response of i'th feature
+    void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints);
+    //! finds the keypoints and computes their descriptors.
+    //! Optionally it can compute descriptors for the user-provided keypoints and recompute keypoints direction
+    void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors,
+        bool useProvidedKeypoints = false);
+
+    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
+    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
+        bool useProvidedKeypoints = false);
+
+    void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, std::vector<float>& descriptors,
+        bool useProvidedKeypoints = false);
+
+    void releaseMemory();
+
+    // SURF parameters
+    double hessianThreshold;
+    int nOctaves;
+    int nOctaveLayers;
+    bool extended;
+    bool upright;
+
+    //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
+    float keypointsRatio;
+
+    GpuMat sum, mask1, maskSum, intBuffer;
+
+    GpuMat det, trace;
+
+    GpuMat maxPosBuffer;
+};
+
+}} // namespace cv { namespace gpu {
+
+#endif // __OPENCV_NONFREE_GPU_HPP__
diff --git a/modules/nonfree/include/opencv2/nonfree/nonfree.hpp b/modules/nonfree/include/opencv2/nonfree/nonfree.hpp
index 5601147ae6..c1bb651f07 100644
--- a/modules/nonfree/include/opencv2/nonfree/nonfree.hpp
+++ b/modules/nonfree/include/opencv2/nonfree/nonfree.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/nonfree.hpp"
\ No newline at end of file
+#include "opencv2/nonfree.hpp"
diff --git a/modules/nonfree/include/opencv2/nonfree/ocl.hpp b/modules/nonfree/include/opencv2/nonfree/ocl.hpp
new file mode 100644
index 0000000000..2b788ba3c8
--- /dev/null
+++ b/modules/nonfree/include/opencv2/nonfree/ocl.hpp
@@ -0,0 +1,124 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+// By downloading, copying, installing or using the software you agree to this license.
+// If you do not agree to this license, do not download, install,
+// copy or use the software.
+//
+//
+// License Agreement
+// For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+// * Redistribution's of source code must retain the above copyright notice,
+// this list of conditions and the following disclaimer.
+//
+// * Redistribution's in binary form must reproduce the above copyright notice,
+// this list of conditions and the following disclaimer in the documentation
+// and/or other materials provided with the distribution.
+//
+// * The name of the copyright holders may not be used to endorse or promote products
+// derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_NONFREE_OCL_HPP__
+#define __OPENCV_NONFREE_OCL_HPP__
+
+#include "opencv2/ocl.hpp"
+
+namespace cv
+{
+    namespace ocl
+    {
+        //! Speeded up robust features, port from GPU module.
+        ////////////////////////////////// SURF //////////////////////////////////////////
+
+        class CV_EXPORTS SURF_OCL
+        {
+        public:
+            enum KeypointLayout
+            {
+                X_ROW = 0,
+                Y_ROW,
+                LAPLACIAN_ROW,
+                OCTAVE_ROW,
+                SIZE_ROW,
+                ANGLE_ROW,
+                HESSIAN_ROW,
+                ROWS_COUNT
+            };
+
+            //! the default constructor
+            SURF_OCL();
+            //! the full constructor taking all the necessary parameters
+            explicit SURF_OCL(double _hessianThreshold, int _nOctaves = 4,
+                              int _nOctaveLayers = 2, bool _extended = false, float _keypointsRatio = 0.01f, bool _upright = false);
+
+            //! returns the descriptor size in float's (64 or 128)
+            int descriptorSize() const;
+            //! upload host keypoints to device memory
+            void uploadKeypoints(const std::vector<cv::KeyPoint> &keypoints, oclMat &keypointsocl);
+            //! download keypoints from device to host memory
+            void downloadKeypoints(const oclMat &keypointsocl, std::vector<KeyPoint> &keypoints);
+            //! download descriptors from device to host memory
+            void downloadDescriptors(const oclMat &descriptorsocl, std::vector<float> &descriptors);
+            //! finds the keypoints using fast hessian detector used in SURF
+            //! supports CV_8UC1 images
+            //! keypoints will have nFeature cols and 6 rows
+            //! keypoints.ptr<float>(X_ROW)[i] will contain x coordinate of i'th feature
+            //! keypoints.ptr<float>(Y_ROW)[i] will contain y coordinate of i'th feature
+            //! keypoints.ptr<float>(LAPLACIAN_ROW)[i] will contain laplacian sign of i'th feature
+            //! keypoints.ptr<float>(OCTAVE_ROW)[i] will contain octave of i'th feature
+            //! keypoints.ptr<float>(SIZE_ROW)[i] will contain size of i'th feature
+            //! keypoints.ptr<float>(ANGLE_ROW)[i] will contain orientation of i'th feature
+            //! keypoints.ptr<float>(HESSIAN_ROW)[i] will contain response of i'th feature
+            void operator()(const oclMat &img, const oclMat &mask, oclMat &keypoints);
+            //! finds the keypoints and computes their descriptors.
+            //! Optionally it can compute descriptors for the user-provided keypoints and recompute keypoints direction
+            void operator()(const oclMat &img, const oclMat &mask, oclMat &keypoints, oclMat &descriptors,
+                            bool useProvidedKeypoints = false);
+            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints);
+            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints, oclMat &descriptors,
+                            bool useProvidedKeypoints = false);
+            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints, std::vector<float> &descriptors,
+                            bool useProvidedKeypoints = false);
+
+            void releaseMemory();
+
+            // SURF parameters
+            float hessianThreshold;
+            int nOctaves;
+            int nOctaveLayers;
+            bool extended;
+            bool upright;
+            //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
+            float keypointsRatio;
+            oclMat sum, mask1, maskSum, intBuffer;
+            oclMat det, trace;
+            oclMat maxPosBuffer;
+
+        };
+    }
+}
+
+#endif //__OPENCV_NONFREE_OCL_HPP__
diff --git a/modules/nonfree/perf/perf_gpu.cpp b/modules/nonfree/perf/perf_gpu.cpp
new file mode 100644
index 0000000000..95708227ae
--- /dev/null
+++ b/modules/nonfree/perf/perf_gpu.cpp
@@ -0,0 +1,103 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+#include "opencv2/ts/gpu_perf.hpp"
+
+using namespace std;
+using namespace testing;
+using namespace perf;
+
+//////////////////////////////////////////////////////////////////////
+// SURF
+
+#ifdef HAVE_OPENCV_GPUARITHM
+
+DEF_PARAM_TEST_1(Image, string);
+
+PERF_TEST_P(Image, GPU_SURF,
+            Values<std::string>("gpu/perf/aloe.png"))
+{
+    declare.time(50.0);
+
+    const cv::Mat img = readImage(GetParam(), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::SURF_GPU d_surf;
+
+        const cv::gpu::GpuMat d_img(img);
+        cv::gpu::GpuMat d_keypoints, d_descriptors;
+
+        TEST_CYCLE() d_surf(d_img, cv::gpu::GpuMat(), d_keypoints, d_descriptors);
+
+        std::vector<cv::KeyPoint> gpu_keypoints;
+        d_surf.downloadKeypoints(d_keypoints, gpu_keypoints);
+
+        cv::Mat gpu_descriptors(d_descriptors);
+
+        sortKeyPoints(gpu_keypoints, gpu_descriptors);
+
+        SANITY_CHECK_KEYPOINTS(gpu_keypoints);
+        SANITY_CHECK(gpu_descriptors, 1e-3);
+    }
+    else
+    {
+        cv::SURF surf;
+
+        std::vector<cv::KeyPoint> cpu_keypoints;
+        cv::Mat cpu_descriptors;
+
+        TEST_CYCLE() surf(img, cv::noArray(), cpu_keypoints, cpu_descriptors);
+
+        SANITY_CHECK_KEYPOINTS(cpu_keypoints);
+        SANITY_CHECK(cpu_descriptors);
+    }
+}
+
+#endif // HAVE_OPENCV_GPUARITHM
+
+#endif // HAVE_CUDA
diff --git a/modules/nonfree/perf/perf_main.cpp b/modules/nonfree/perf/perf_main.cpp
index 3fcca19c3c..d5f4a1a512 100644
--- a/modules/nonfree/perf/perf_main.cpp
+++ b/modules/nonfree/perf/perf_main.cpp
@@ -1,3 +1,11 @@
 #include "perf_precomp.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-CV_PERF_TEST_MAIN(nonfree)
+static const char * impls[] = {
+#ifdef HAVE_CUDA
+    "cuda",
+#endif
+    "plain"
+};
+
+CV_PERF_TEST_MAIN_WITH_IMPLS(nonfree, impls, perf::printCudaInfo())
diff --git a/modules/nonfree/perf/perf_precomp.cpp b/modules/nonfree/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/nonfree/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/nonfree/perf/perf_precomp.hpp b/modules/nonfree/perf/perf_precomp.hpp
index bb03c28b60..e8fad3b7be 100644
--- a/modules/nonfree/perf/perf_precomp.hpp
+++ b/modules/nonfree/perf/perf_precomp.hpp
@@ -13,6 +13,17 @@
 #include "opencv2/nonfree.hpp"
 #include "opencv2/highgui.hpp"
 
+#include "opencv2/opencv_modules.hpp"
+#include "cvconfig.h"
+
+#ifdef HAVE_OPENCV_OCL
+#  include "opencv2/nonfree/ocl.hpp"
+#endif
+
+#ifdef HAVE_CUDA
+#  include "opencv2/nonfree/gpu.hpp"
+#endif
+
 #ifdef GTEST_CREATE_SHARED_LIBRARY
 #error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
 #endif
diff --git a/modules/ocl/perf/perf_surf.cpp b/modules/nonfree/perf/perf_surf.ocl.cpp
similarity index 63%
rename from modules/ocl/perf/perf_surf.cpp
rename to modules/nonfree/perf/perf_surf.ocl.cpp
index 6aa4f512a2..fdd1931bd8 100644
--- a/modules/ocl/perf/perf_surf.cpp
+++ b/modules/nonfree/perf/perf_surf.ocl.cpp
@@ -43,61 +43,69 @@
 //
 //M*/
 
-#include "precomp.hpp"
-#include <iomanip>
+#include "perf_precomp.hpp"
 
-#ifdef HAVE_OPENCL
+#ifdef HAVE_OPENCV_OCL
 
 using namespace cv;
 using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
 using namespace std;
 
-#define FILTER_IMAGE "../../../samples/gpu/road.png"
+typedef perf::TestBaseWithParam<std::string> OCL_SURF;
 
-TEST(SURF, Performance)
+#define SURF_IMAGES \
+    "cv/detectors_descriptors_evaluation/images_datasets/leuven/img1.png",\
+    "stitching/a3.png"
+
+PERF_TEST_P(OCL_SURF, DISABLED_with_data_transfer, testing::Values(SURF_IMAGES))
 {
-    cv::Mat img = readImage(FILTER_IMAGE, cv::IMREAD_GRAYSCALE);
+    string filename = getDataPath(GetParam());
+    Mat img = imread(filename, IMREAD_GRAYSCALE);
     ASSERT_FALSE(img.empty());
 
-    ocl::SURF_OCL d_surf;
-    ocl::oclMat d_keypoints;
-    ocl::oclMat d_descriptors;
+    SURF_OCL d_surf;
+    oclMat d_keypoints;
+    oclMat d_descriptors;
+    Mat cpu_kp;
+    Mat cpu_dp;
 
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
+    declare.time(60);
 
-    double t1 = 0;
-    double t2 = 0;
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    TEST_CYCLE()
     {
-        t1 = (double)cvGetTickCount();//gpu start1
+        oclMat d_src(img);
 
-        ocl::oclMat d_src(img);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        d_surf(d_src, ocl::oclMat(), d_keypoints, d_descriptors);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_kp, cpu_dp;
-        d_keypoints.download (cpu_kp);//download
-        d_descriptors.download (cpu_dp);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
+        d_surf(d_src, oclMat(), d_keypoints, d_descriptors);
 
+        d_keypoints.download(cpu_kp);
+        d_descriptors.download(cpu_dp);
     }
 
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
+    SANITY_CHECK(cpu_kp, 1);
+    SANITY_CHECK(cpu_dp, 1);
 }
-#endif  //Have opencl
\ No newline at end of file
+
+PERF_TEST_P(OCL_SURF, DISABLED_without_data_transfer, testing::Values(SURF_IMAGES))
+{
+    string filename = getDataPath(GetParam());
+    Mat img = imread(filename, IMREAD_GRAYSCALE);
+    ASSERT_FALSE(img.empty());
+
+    SURF_OCL d_surf;
+    oclMat d_keypoints;
+    oclMat d_descriptors;
+    oclMat d_src(img);
+
+    declare.time(60);
+
+    TEST_CYCLE() d_surf(d_src, oclMat(), d_keypoints, d_descriptors);
+
+    Mat cpu_kp;
+    Mat cpu_dp;
+    d_keypoints.download(cpu_kp);
+    d_descriptors.download(cpu_dp);
+    SANITY_CHECK(cpu_kp, 1);
+    SANITY_CHECK(cpu_dp, 1);
+}
+
+#endif // HAVE_OPENCV_OCL
diff --git a/modules/gpu/src/cuda/surf.cu b/modules/nonfree/src/cuda/surf.cu
similarity index 95%
rename from modules/gpu/src/cuda/surf.cu
rename to modules/nonfree/src/cuda/surf.cu
index 5dc2c82104..cdd54e4caf 100644
--- a/modules/gpu/src/cuda/surf.cu
+++ b/modules/nonfree/src/cuda/surf.cu
@@ -38,24 +38,48 @@
 // or tort (including negligence or otherwise) arising in any way out of
 // the use of this software, even if advised of the possibility of such damage.
 //
-// Copyright (c) 2010, Paul Furgale, Chi Hay Tong
-//
-// The original code was written by Paul Furgale and Chi Hay Tong
-// and later optimized and prepared for integration into OpenCV by Itseez.
-//
 //M*/
 
-#if !defined CUDA_DISABLER
+#include "opencv2/opencv_modules.hpp"
 
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/limits.hpp"
-#include "opencv2/gpu/device/saturate_cast.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/utility.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/filters.hpp"
+#ifdef HAVE_OPENCV_GPUARITHM
 
-namespace cv { namespace gpu { namespace device
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/limits.hpp"
+#include "opencv2/core/cuda/saturate_cast.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/filters.hpp"
+
+namespace cv { namespace gpu { namespace cudev
+{
+    namespace surf
+    {
+        void loadGlobalConstants(int maxCandidates, int maxFeatures, int img_rows, int img_cols, int nOctaveLayers, float hessianThreshold);
+        void loadOctaveConstants(int octave, int layer_rows, int layer_cols);
+
+        void bindImgTex(PtrStepSzb img);
+        size_t bindSumTex(PtrStepSz<unsigned int> sum);
+        size_t bindMaskSumTex(PtrStepSz<unsigned int> maskSum);
+
+        void icvCalcLayerDetAndTrace_gpu(const PtrStepf& det, const PtrStepf& trace, int img_rows, int img_cols,
+            int octave, int nOctaveLayer);
+
+        void icvFindMaximaInLayer_gpu(const PtrStepf& det, const PtrStepf& trace, int4* maxPosBuffer, unsigned int* maxCounter,
+            int img_rows, int img_cols, int octave, bool use_mask, int nLayers);
+
+        void icvInterpolateKeypoint_gpu(const PtrStepf& det, const int4* maxPosBuffer, unsigned int maxCounter,
+            float* featureX, float* featureY, int* featureLaplacian, int* featureOctave, float* featureSize, float* featureHessian,
+            unsigned int* featureCounter);
+
+        void icvCalcOrientation_gpu(const float* featureX, const float* featureY, const float* featureSize, float* featureDir, int nFeatures);
+
+        void compute_descriptors_gpu(PtrStepSz<float4> descriptors, const float* featureX, const float* featureY, const float* featureSize, const float* featureDir, int nFeatures);
+    }
+}}}
+
+namespace cv { namespace gpu { namespace cudev
 {
     namespace surf
     {
@@ -602,7 +626,7 @@ namespace cv { namespace gpu { namespace device
                 }
 
                 plus<float> op;
-                device::reduce<32>(smem_tuple(s_sumx + threadIdx.y * 32, s_sumy + threadIdx.y * 32),
+                cudev::reduce<32>(smem_tuple(s_sumx + threadIdx.y * 32, s_sumy + threadIdx.y * 32),
                                    thrust::tie(sumx, sumy), threadIdx.x, thrust::make_tuple(op, op));
 
                 const float temp_mod = sumx * sumx + sumy * sumy;
@@ -717,6 +741,9 @@ namespace cv { namespace gpu { namespace device
             int height;
         };
 
+        __device__ void calc_dx_dy(const float* featureX, const float* featureY, const float* featureSize, const float* featureDir,
+                                   float& dx, float& dy);
+
         __device__ void calc_dx_dy(const float* featureX, const float* featureY, const float* featureSize, const float* featureDir,
                                    float& dx, float& dy)
         {
@@ -928,7 +955,6 @@ namespace cv { namespace gpu { namespace device
             }
         }
     } // namespace surf
-}}} // namespace cv { namespace gpu { namespace device
+}}} // namespace cv { namespace gpu { namespace cudev
 
-
-#endif /* CUDA_DISABLER */
+#endif // HAVE_OPENCV_GPUARITHM
diff --git a/modules/nonfree/src/nonfree_init.cpp b/modules/nonfree/src/nonfree_init.cpp
index 3c530e3257..ac804dd9a6 100644
--- a/modules/nonfree/src/nonfree_init.cpp
+++ b/modules/nonfree/src/nonfree_init.cpp
@@ -67,7 +67,7 @@ CV_INIT_ALGORITHM(SIFT, "Feature2D.SIFT",
 
 bool initModule_nonfree(void)
 {
-    Ptr<Algorithm> sift = createSIFT(), surf = createSURF();
+    Ptr<Algorithm> sift = createSIFT_ptr_hidden(), surf = createSURF_ptr_hidden();
     return sift->info() != 0 && surf->info() != 0;
 }
 
diff --git a/modules/ocl/src/kernels/nonfree_surf.cl b/modules/nonfree/src/opencl/surf.cl
similarity index 66%
rename from modules/ocl/src/kernels/nonfree_surf.cl
rename to modules/nonfree/src/opencl/surf.cl
index e6acba2b0b..aace143d53 100644
--- a/modules/ocl/src/kernels/nonfree_surf.cl
+++ b/modules/nonfree/src/opencl/surf.cl
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Peng Xiao, pengxiao@multicorewareinc.com
+//    Sen Liu, swjtuls1987@126.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -43,9 +44,6 @@
 //
 //M*/
 
-#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable
-#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable
-
 // specialized for non-image2d_t supported platform, intel HD4000, for example
 #ifdef DISABLE_IMAGE2D
 #define IMAGE_INT32 __global uint  *
@@ -78,7 +76,12 @@ uchar read_imgTex(IMAGE_INT8 img, sampler_t sam, float2 coord, int rows, int col
 
 // dynamically change the precision used for floating type
 
-#if defined DOUBLE_SUPPORT
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #define F double
 #else
 #define F float
@@ -100,7 +103,7 @@ __constant sampler_t sampler    = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAM
 // for simple haar paatern
 float icvCalcHaarPatternSum_2(
     IMAGE_INT32 sumTex,
-    __constant float src[2][5],
+    __constant float2 *src,
     int oldSize,
     int newSize,
     int y, int x,
@@ -111,21 +114,24 @@ float icvCalcHaarPatternSum_2(
 
     F d = 0;
 
-#pragma unroll
-    for (int k = 0; k < 2; ++k)
-    {
-        int dx1 = convert_int_rte(ratio * src[k][0]);
-        int dy1 = convert_int_rte(ratio * src[k][1]);
-        int dx2 = convert_int_rte(ratio * src[k][2]);
-        int dy2 = convert_int_rte(ratio * src[k][3]);
+    int2 dx1 = convert_int2_rte(ratio * src[0]);
+    int2 dy1 = convert_int2_rte(ratio * src[1]);
+    int2 dx2 = convert_int2_rte(ratio * src[2]);
+    int2 dy2 = convert_int2_rte(ratio * src[3]);
 
-        F t = 0;
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy1), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy2), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy1), rows, cols, elemPerRow );
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy2), rows, cols, elemPerRow );
-        d += t * src[k][4] / ((dx2 - dx1) * (dy2 - dy1));
-    }
+    F t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy1.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy2.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy1.x), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy2.x), rows, cols, elemPerRow );
+    d += t * src[4].x / ((dx2.x - dx1.x) * (dy2.x - dy1.x));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy1.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy2.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy1.y), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy2.y), rows, cols, elemPerRow );
+    d += t * src[4].y / ((dx2.y - dx1.y) * (dy2.y - dy1.y));
 
     return (float)d;
 }
@@ -133,7 +139,7 @@ float icvCalcHaarPatternSum_2(
 // N = 3
 float icvCalcHaarPatternSum_3(
     IMAGE_INT32 sumTex,
-    __constant float src[2][5],
+    __constant float4 *src,
     int oldSize,
     int newSize,
     int y, int x,
@@ -144,21 +150,31 @@ float icvCalcHaarPatternSum_3(
 
     F d = 0;
 
-#pragma unroll
-    for (int k = 0; k < 3; ++k)
-    {
-        int dx1 = convert_int_rte(ratio * src[k][0]);
-        int dy1 = convert_int_rte(ratio * src[k][1]);
-        int dx2 = convert_int_rte(ratio * src[k][2]);
-        int dy2 = convert_int_rte(ratio * src[k][3]);
+    int4 dx1 = convert_int4_rte(ratio * src[0]);
+    int4 dy1 = convert_int4_rte(ratio * src[1]);
+    int4 dx2 = convert_int4_rte(ratio * src[2]);
+    int4 dy2 = convert_int4_rte(ratio * src[3]);
 
-        F t = 0;
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy1), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy2), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy1), rows, cols, elemPerRow );
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy2), rows, cols, elemPerRow );
-        d += t * src[k][4] / ((dx2 - dx1) * (dy2 - dy1));
-    }
+    F t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy1.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy2.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy1.x), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy2.x), rows, cols, elemPerRow );
+    d += t * src[4].x / ((dx2.x - dx1.x) * (dy2.x - dy1.x));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy1.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy2.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy1.y), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy2.y), rows, cols, elemPerRow );
+    d += t * src[4].y / ((dx2.y - dx1.y) * (dy2.y - dy1.y));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.z, y + dy1.z), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.z, y + dy2.z), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.z, y + dy1.z), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.z, y + dy2.z), rows, cols, elemPerRow );
+    d += t * src[4].z / ((dx2.z - dx1.z) * (dy2.z - dy1.z));
 
     return (float)d;
 }
@@ -166,7 +182,7 @@ float icvCalcHaarPatternSum_3(
 // N = 4
 float icvCalcHaarPatternSum_4(
     IMAGE_INT32 sumTex,
-    __constant float src[2][5],
+    __constant float4 *src,
     int oldSize,
     int newSize,
     int y, int x,
@@ -177,21 +193,38 @@ float icvCalcHaarPatternSum_4(
 
     F d = 0;
 
-#pragma unroll
-    for (int k = 0; k < 4; ++k)
-    {
-        int dx1 = convert_int_rte(ratio * src[k][0]);
-        int dy1 = convert_int_rte(ratio * src[k][1]);
-        int dx2 = convert_int_rte(ratio * src[k][2]);
-        int dy2 = convert_int_rte(ratio * src[k][3]);
+    int4 dx1 = convert_int4_rte(ratio * src[0]);
+    int4 dy1 = convert_int4_rte(ratio * src[1]);
+    int4 dx2 = convert_int4_rte(ratio * src[2]);
+    int4 dy2 = convert_int4_rte(ratio * src[3]);
 
-        F t = 0;
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy1), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx1, y + dy2), rows, cols, elemPerRow );
-        t -= read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy1), rows, cols, elemPerRow );
-        t += read_sumTex( sumTex, sampler, (int2)(x + dx2, y + dy2), rows, cols, elemPerRow );
-        d += t * src[k][4] / ((dx2 - dx1) * (dy2 - dy1));
-    }
+    F t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy1.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.x, y + dy2.x), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy1.x), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.x, y + dy2.x), rows, cols, elemPerRow );
+    d += t * src[4].x / ((dx2.x - dx1.x) * (dy2.x - dy1.x));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy1.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.y, y + dy2.y), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy1.y), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.y, y + dy2.y), rows, cols, elemPerRow );
+    d += t * src[4].y / ((dx2.y - dx1.y) * (dy2.y - dy1.y));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.z, y + dy1.z), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.z, y + dy2.z), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.z, y + dy1.z), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.z, y + dy2.z), rows, cols, elemPerRow );
+    d += t * src[4].z / ((dx2.z - dx1.z) * (dy2.z - dy1.z));
+
+    t = 0;
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx1.w, y + dy1.w), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx1.w, y + dy2.w), rows, cols, elemPerRow );
+    t -= read_sumTex( sumTex, sampler, (int2)(x + dx2.w, y + dy1.w), rows, cols, elemPerRow );
+    t += read_sumTex( sumTex, sampler, (int2)(x + dx2.w, y + dy2.w), rows, cols, elemPerRow );
+    d += t * src[4].w / ((dx2.w - dx1.w) * (dy2.w - dy1.w));
 
     return (float)d;
 }
@@ -199,9 +232,9 @@ float icvCalcHaarPatternSum_4(
 ////////////////////////////////////////////////////////////////////////
 // Hessian
 
-__constant float c_DX [3][5] = { {0, 2, 3, 7, 1}, {3, 2, 6, 7, -2}, {6, 2, 9, 7, 1} };
-__constant float c_DY [3][5] = { {2, 0, 7, 3, 1}, {2, 3, 7, 6, -2}, {2, 6, 7, 9, 1} };
-__constant float c_DXY[4][5] = { {1, 1, 4, 4, 1}, {5, 1, 8, 4, -1}, {1, 5, 4, 8, -1}, {5, 5, 8, 8, 1} };
+__constant float4 c_DX[5] = { (float4)(0, 3, 6, 0), (float4)(2, 2, 2, 0), (float4)(3, 6, 9, 0), (float4)(7, 7, 7, 0), (float4)(1, -2, 1, 0) };
+__constant float4 c_DY[5] = { (float4)(2, 2, 2, 0), (float4)(0, 3, 6, 0), (float4)(7, 7, 7, 0), (float4)(3, 6, 9, 0), (float4)(1, -2, 1, 0) };
+__constant float4 c_DXY[5] = { (float4)(1, 5, 1, 5), (float4)(1, 1, 5, 5), (float4)(4, 8, 4, 8), (float4)(4, 4, 8, 8), (float4)(1, -1, -1, 1) };// Use integral image to calculate haar wavelets.
 
 __inline int calcSize(int octave, int layer)
 {
@@ -231,7 +264,7 @@ __kernel void icvCalcLayerDetAndTrace(
     int c_octave,
     int c_layer_rows,
     int sumTex_step
-    )
+)
 {
     det_step   /= sizeof(*det);
     trace_step /= sizeof(*trace);
@@ -295,7 +328,7 @@ bool within_check(IMAGE_INT32 maskSumTex, int sum_i, int sum_j, int size, int ro
 
 // Non-maximal suppression to further filtering the candidates from previous step
 __kernel
-    void icvFindMaximaInLayer_withmask(
+void icvFindMaximaInLayer_withmask(
     __global const float * det,
     __global const float * trace,
     __global int4 * maxPosBuffer,
@@ -313,7 +346,7 @@ __kernel
     float c_hessianThreshold,
     IMAGE_INT32 maskSumTex,
     int mask_step
-    )
+)
 {
     volatile __local  float N9[768]; // threads.x * threads.y * 3
 
@@ -342,26 +375,26 @@ __kernel
     const int localLin = get_local_id(0) + get_local_id(1) * get_local_size(0) + zoff;
     N9[localLin - zoff] =
         det[det_step *
-        (c_layer_rows * (layer - 1) + min(max(i, 0), c_img_rows - 1)) // y
-        + min(max(j, 0), c_img_cols - 1)];                            // x
+            (c_layer_rows * (layer - 1) + min(max(i, 0), c_img_rows - 1)) // y
+            + min(max(j, 0), c_img_cols - 1)];                            // x
     N9[localLin       ] =
         det[det_step *
-        (c_layer_rows * (layer    ) + min(max(i, 0), c_img_rows - 1)) // y
-        + min(max(j, 0), c_img_cols - 1)];                            // x
+            (c_layer_rows * (layer    ) + min(max(i, 0), c_img_rows - 1)) // y
+            + min(max(j, 0), c_img_cols - 1)];                            // x
     N9[localLin + zoff] =
         det[det_step *
-        (c_layer_rows * (layer + 1) + min(max(i, 0), c_img_rows - 1)) // y
-        + min(max(j, 0), c_img_cols - 1)];                            // x
+            (c_layer_rows * (layer + 1) + min(max(i, 0), c_img_rows - 1)) // y
+            + min(max(j, 0), c_img_cols - 1)];                            // x
 
     barrier(CLK_LOCAL_MEM_FENCE);
 
     if (i < c_layer_rows - margin
-        && j < c_layer_cols - margin
-        && get_local_id(0) > 0
-        && get_local_id(0) < get_local_size(0) - 1
-        && get_local_id(1) > 0
-        && get_local_id(1) < get_local_size(1) - 1 // these are unnecessary conditions ported from CUDA
-        )
+            && j < c_layer_cols - margin
+            && get_local_id(0) > 0
+            && get_local_id(0) < get_local_size(0) - 1
+            && get_local_id(1) > 0
+            && get_local_id(1) < get_local_size(1) - 1 // these are unnecessary conditions ported from CUDA
+       )
     {
         float val0 = N9[localLin];
 
@@ -377,34 +410,34 @@ __kernel
             {
                 // Check to see if we have a max (in its 26 neighbours)
                 const bool condmax = val0 > N9[localLin - 1 - get_local_size(0) - zoff]
-                &&                   val0 > N9[localLin     - get_local_size(0) - zoff]
-                &&                   val0 > N9[localLin + 1 - get_local_size(0) - zoff]
-                &&                   val0 > N9[localLin - 1                     - zoff]
-                &&                   val0 > N9[localLin                         - zoff]
-                &&                   val0 > N9[localLin + 1                     - zoff]
-                &&                   val0 > N9[localLin - 1 + get_local_size(0) - zoff]
-                &&                   val0 > N9[localLin     + get_local_size(0) - zoff]
-                &&                   val0 > N9[localLin + 1 + get_local_size(0) - zoff]
+                                     &&                   val0 > N9[localLin     - get_local_size(0) - zoff]
+                                     &&                   val0 > N9[localLin + 1 - get_local_size(0) - zoff]
+                                     &&                   val0 > N9[localLin - 1                     - zoff]
+                                     &&                   val0 > N9[localLin                         - zoff]
+                                     &&                   val0 > N9[localLin + 1                     - zoff]
+                                     &&                   val0 > N9[localLin - 1 + get_local_size(0) - zoff]
+                                     &&                   val0 > N9[localLin     + get_local_size(0) - zoff]
+                                     &&                   val0 > N9[localLin + 1 + get_local_size(0) - zoff]
 
-                &&                   val0 > N9[localLin - 1 - get_local_size(0)]
-                &&                   val0 > N9[localLin     - get_local_size(0)]
-                &&                   val0 > N9[localLin + 1 - get_local_size(0)]
-                &&                   val0 > N9[localLin - 1                    ]
-                &&                   val0 > N9[localLin + 1                    ]
-                &&                   val0 > N9[localLin - 1 + get_local_size(0)]
-                &&                   val0 > N9[localLin     + get_local_size(0)]
-                &&                   val0 > N9[localLin + 1 + get_local_size(0)]
+                                     &&                   val0 > N9[localLin - 1 - get_local_size(0)]
+                                     &&                   val0 > N9[localLin     - get_local_size(0)]
+                                     &&                   val0 > N9[localLin + 1 - get_local_size(0)]
+                                     &&                   val0 > N9[localLin - 1                    ]
+                                     &&                   val0 > N9[localLin + 1                    ]
+                                     &&                   val0 > N9[localLin - 1 + get_local_size(0)]
+                                     &&                   val0 > N9[localLin     + get_local_size(0)]
+                                     &&                   val0 > N9[localLin + 1 + get_local_size(0)]
 
-                &&                   val0 > N9[localLin - 1 - get_local_size(0) + zoff]
-                &&                   val0 > N9[localLin     - get_local_size(0) + zoff]
-                &&                   val0 > N9[localLin + 1 - get_local_size(0) + zoff]
-                &&                   val0 > N9[localLin - 1                     + zoff]
-                &&                   val0 > N9[localLin                         + zoff]
-                &&                   val0 > N9[localLin + 1                     + zoff]
-                &&                   val0 > N9[localLin - 1 + get_local_size(0) + zoff]
-                &&                   val0 > N9[localLin     + get_local_size(0) + zoff]
-                &&                   val0 > N9[localLin + 1 + get_local_size(0) + zoff]
-                ;
+                                     &&                   val0 > N9[localLin - 1 - get_local_size(0) + zoff]
+                                     &&                   val0 > N9[localLin     - get_local_size(0) + zoff]
+                                     &&                   val0 > N9[localLin + 1 - get_local_size(0) + zoff]
+                                     &&                   val0 > N9[localLin - 1                     + zoff]
+                                     &&                   val0 > N9[localLin                         + zoff]
+                                     &&                   val0 > N9[localLin + 1                     + zoff]
+                                     &&                   val0 > N9[localLin - 1 + get_local_size(0) + zoff]
+                                     &&                   val0 > N9[localLin     + get_local_size(0) + zoff]
+                                     &&                   val0 > N9[localLin + 1 + get_local_size(0) + zoff]
+                                     ;
 
                 if(condmax)
                 {
@@ -423,7 +456,7 @@ __kernel
 }
 
 __kernel
-    void icvFindMaximaInLayer(
+void icvFindMaximaInLayer(
     __global float * det,
     __global float * trace,
     __global int4 * maxPosBuffer,
@@ -439,7 +472,7 @@ __kernel
     int c_layer_cols,
     int c_max_candidates,
     float c_hessianThreshold
-    )
+)
 {
     volatile __local  float N9[768]; // threads.x * threads.y * 3
 
@@ -478,12 +511,12 @@ __kernel
     barrier(CLK_LOCAL_MEM_FENCE);
 
     if (i < c_layer_rows - margin
-        && j < c_layer_cols - margin
-        && get_local_id(0) > 0
-        && get_local_id(0) < get_local_size(0) - 1
-        && get_local_id(1) > 0
-        && get_local_id(1) < get_local_size(1) - 1 // these are unnecessary conditions ported from CUDA
-        )
+            && j < c_layer_cols - margin
+            && get_local_id(0) > 0
+            && get_local_id(0) < get_local_size(0) - 1
+            && get_local_id(1) > 0
+            && get_local_id(1) < get_local_size(1) - 1 // these are unnecessary conditions ported from CUDA
+       )
     {
         float val0 = N9[localLin];
         if (val0 > c_hessianThreshold)
@@ -494,38 +527,38 @@ __kernel
 
             // Check to see if we have a max (in its 26 neighbours)
             const bool condmax = val0 > N9[localLin - 1 - get_local_size(0) - zoff]
-            &&                   val0 > N9[localLin     - get_local_size(0) - zoff]
-            &&                   val0 > N9[localLin + 1 - get_local_size(0) - zoff]
-            &&                   val0 > N9[localLin - 1                     - zoff]
-            &&                   val0 > N9[localLin                         - zoff]
-            &&                   val0 > N9[localLin + 1                     - zoff]
-            &&                   val0 > N9[localLin - 1 + get_local_size(0) - zoff]
-            &&                   val0 > N9[localLin     + get_local_size(0) - zoff]
-            &&                   val0 > N9[localLin + 1 + get_local_size(0) - zoff]
+                                 &&                   val0 > N9[localLin     - get_local_size(0) - zoff]
+                                 &&                   val0 > N9[localLin + 1 - get_local_size(0) - zoff]
+                                 &&                   val0 > N9[localLin - 1                     - zoff]
+                                 &&                   val0 > N9[localLin                         - zoff]
+                                 &&                   val0 > N9[localLin + 1                     - zoff]
+                                 &&                   val0 > N9[localLin - 1 + get_local_size(0) - zoff]
+                                 &&                   val0 > N9[localLin     + get_local_size(0) - zoff]
+                                 &&                   val0 > N9[localLin + 1 + get_local_size(0) - zoff]
 
-            &&                   val0 > N9[localLin - 1 - get_local_size(0)]
-            &&                   val0 > N9[localLin     - get_local_size(0)]
-            &&                   val0 > N9[localLin + 1 - get_local_size(0)]
-            &&                   val0 > N9[localLin - 1                    ]
-            &&                   val0 > N9[localLin + 1                    ]
-            &&                   val0 > N9[localLin - 1 + get_local_size(0)]
-            &&                   val0 > N9[localLin     + get_local_size(0)]
-            &&                   val0 > N9[localLin + 1 + get_local_size(0)]
+                                 &&                   val0 > N9[localLin - 1 - get_local_size(0)]
+                                 &&                   val0 > N9[localLin     - get_local_size(0)]
+                                 &&                   val0 > N9[localLin + 1 - get_local_size(0)]
+                                 &&                   val0 > N9[localLin - 1                    ]
+                                 &&                   val0 > N9[localLin + 1                    ]
+                                 &&                   val0 > N9[localLin - 1 + get_local_size(0)]
+                                 &&                   val0 > N9[localLin     + get_local_size(0)]
+                                 &&                   val0 > N9[localLin + 1 + get_local_size(0)]
 
-            &&                   val0 > N9[localLin - 1 - get_local_size(0) + zoff]
-            &&                   val0 > N9[localLin     - get_local_size(0) + zoff]
-            &&                   val0 > N9[localLin + 1 - get_local_size(0) + zoff]
-            &&                   val0 > N9[localLin - 1                     + zoff]
-            &&                   val0 > N9[localLin                         + zoff]
-            &&                   val0 > N9[localLin + 1                     + zoff]
-            &&                   val0 > N9[localLin - 1 + get_local_size(0) + zoff]
-            &&                   val0 > N9[localLin     + get_local_size(0) + zoff]
-            &&                   val0 > N9[localLin + 1 + get_local_size(0) + zoff]
-            ;
+                                 &&                   val0 > N9[localLin - 1 - get_local_size(0) + zoff]
+                                 &&                   val0 > N9[localLin     - get_local_size(0) + zoff]
+                                 &&                   val0 > N9[localLin + 1 - get_local_size(0) + zoff]
+                                 &&                   val0 > N9[localLin - 1                     + zoff]
+                                 &&                   val0 > N9[localLin                         + zoff]
+                                 &&                   val0 > N9[localLin + 1                     + zoff]
+                                 &&                   val0 > N9[localLin - 1 + get_local_size(0) + zoff]
+                                 &&                   val0 > N9[localLin     + get_local_size(0) + zoff]
+                                 &&                   val0 > N9[localLin + 1 + get_local_size(0) + zoff]
+                                 ;
 
             if(condmax)
             {
-                 int ind = atomic_inc(maxCounter);
+                int ind = atomic_inc(maxCounter);
 
                 if (ind < c_max_candidates)
                 {
@@ -539,30 +572,30 @@ __kernel
 }
 
 // solve 3x3 linear system Ax=b for floating point input
-inline bool solve3x3_float(volatile __local  const float A[3][3], volatile __local  const float b[3], volatile __local  float x[3])
+inline bool solve3x3_float(volatile __local  const float4 *A, volatile __local  const float *b, volatile __local  float *x)
 {
-    float det = A[0][0] * (A[1][1] * A[2][2] - A[1][2] * A[2][1])
-        - A[0][1] * (A[1][0] * A[2][2] - A[1][2] * A[2][0])
-        + A[0][2] * (A[1][0] * A[2][1] - A[1][1] * A[2][0]);
+    float det = A[0].x * (A[1].y * A[2].z - A[1].z * A[2].y)
+                - A[0].y * (A[1].x * A[2].z - A[1].z * A[2].x)
+                + A[0].z * (A[1].x * A[2].y - A[1].y * A[2].x);
 
     if (det != 0)
     {
         F invdet = 1.0 / det;
 
         x[0] = invdet *
-            (b[0]    * (A[1][1] * A[2][2] - A[1][2] * A[2][1]) -
-            A[0][1] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) +
-            A[0][2] * (b[1]    * A[2][1] - A[1][1] * b[2]   ));
+               (b[0]    * (A[1].y * A[2].z - A[1].z * A[2].y) -
+                A[0].y * (b[1]    * A[2].z - A[1].z * b[2]   ) +
+                A[0].z * (b[1]    * A[2].y - A[1].y * b[2]   ));
 
         x[1] = invdet *
-            (A[0][0] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) -
-            b[0]    * (A[1][0] * A[2][2] - A[1][2] * A[2][0]) +
-            A[0][2] * (A[1][0] * b[2]    - b[1]    * A[2][0]));
+               (A[0].x * (b[1]    * A[2].z - A[1].z * b[2]   ) -
+                b[0]    * (A[1].x * A[2].z - A[1].z * A[2].x) +
+                A[0].z * (A[1].x * b[2]    - b[1]    * A[2].x));
 
         x[2] = invdet *
-            (A[0][0] * (A[1][1] * b[2]    - b[1]    * A[2][1]) -
-            A[0][1] * (A[1][0] * b[2]    - b[1]    * A[2][0]) +
-            b[0]    * (A[1][0] * A[2][1] - A[1][1] * A[2][0]));
+               (A[0].x * (A[1].y * b[2]    - b[1]    * A[2].y) -
+                A[0].y * (A[1].x * b[2]    - b[1]    * A[2].x) +
+                b[0]    * (A[1].x * A[2].y - A[1].y * A[2].x));
 
         return true;
     }
@@ -581,7 +614,7 @@ inline bool solve3x3_float(volatile __local  const float A[3][3], volatile __loc
 ////////////////////////////////////////////////////////////////////////
 // INTERPOLATION
 __kernel
-    void icvInterpolateKeypoint(
+void icvInterpolateKeypoint(
     __global const float * det,
     __global const int4 * maxPosBuffer,
     __global float * keypoints,
@@ -593,7 +626,7 @@ __kernel
     int c_octave,
     int c_layer_rows,
     int c_max_features
-    )
+)
 {
     det_step /= sizeof(*det);
     keypoints_step /= sizeof(*keypoints);
@@ -627,26 +660,26 @@ __kernel
         //ds
         dD[2] = -0.5f * (N9[2][1][1] - N9[0][1][1]);
 
-        volatile __local  float H[3][3];
+        volatile __local  float4 H[3];
 
         //dxx
-        H[0][0] = N9[1][1][0] - 2.0f * N9[1][1][1] + N9[1][1][2];
+        H[0].x = N9[1][1][0] - 2.0f * N9[1][1][1] + N9[1][1][2];
         //dxy
-        H[0][1]= 0.25f * (N9[1][2][2] - N9[1][2][0] - N9[1][0][2] + N9[1][0][0]);
+        H[0].y= 0.25f * (N9[1][2][2] - N9[1][2][0] - N9[1][0][2] + N9[1][0][0]);
         //dxs
-        H[0][2]= 0.25f * (N9[2][1][2] - N9[2][1][0] - N9[0][1][2] + N9[0][1][0]);
+        H[0].z= 0.25f * (N9[2][1][2] - N9[2][1][0] - N9[0][1][2] + N9[0][1][0]);
         //dyx = dxy
-        H[1][0] = H[0][1];
+        H[1].x = H[0].y;
         //dyy
-        H[1][1] = N9[1][0][1] - 2.0f * N9[1][1][1] + N9[1][2][1];
+        H[1].y = N9[1][0][1] - 2.0f * N9[1][1][1] + N9[1][2][1];
         //dys
-        H[1][2]= 0.25f * (N9[2][2][1] - N9[2][0][1] - N9[0][2][1] + N9[0][0][1]);
+        H[1].z= 0.25f * (N9[2][2][1] - N9[2][0][1] - N9[0][2][1] + N9[0][0][1]);
         //dsx = dxs
-        H[2][0] = H[0][2];
+        H[2].x = H[0].z;
         //dsy = dys
-        H[2][1] = H[1][2];
+        H[2].y = H[1].z;
         //dss
-        H[2][2] = N9[0][1][1] - 2.0f * N9[1][1][1] + N9[2][1][1];
+        H[2].z = N9[0][1][1] - 2.0f * N9[1][1][1] + N9[2][1][1];
 
         volatile __local  float x[3];
 
@@ -684,7 +717,7 @@ __kernel
                 if ((c_img_rows + 1) >= grad_wav_size && (c_img_cols + 1) >= grad_wav_size)
                 {
                     // Get a new feature index.
-                     int ind = atomic_inc(featureCounter);
+                    int ind = atomic_inc(featureCounter);
 
                     if (ind < c_max_features)
                     {
@@ -711,58 +744,86 @@ __kernel
 __constant float c_aptX[ORI_SAMPLES] = {-6, -5, -5, -5, -5, -5, -5, -5, -4, -4, -4, -4, -4, -4, -4, -4, -4, -3, -3, -3, -3, -3, -3, -3, -3, -3, -3, -3, -2, -2, -2, -2, -2, -2, -2, -2, -2, -2, -2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 6};
 __constant float c_aptY[ORI_SAMPLES] = {0, -3, -2, -1, 0, 1, 2, 3, -4, -3, -2, -1, 0, 1, 2, 3, 4, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, 6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, -4, -3, -2, -1, 0, 1, 2, 3, 4, -3, -2, -1, 0, 1, 2, 3, 0};
 __constant float c_aptW[ORI_SAMPLES] = {0.001455130288377404f, 0.001707611023448408f, 0.002547456417232752f, 0.003238451667129993f, 0.0035081731621176f,
-    0.003238451667129993f, 0.002547456417232752f, 0.001707611023448408f, 0.002003900473937392f, 0.0035081731621176f, 0.005233579315245152f,
-    0.00665318313986063f, 0.00720730796456337f, 0.00665318313986063f, 0.005233579315245152f, 0.0035081731621176f,
-    0.002003900473937392f, 0.001707611023448408f, 0.0035081731621176f, 0.006141661666333675f, 0.009162282571196556f,
-    0.01164754293859005f, 0.01261763460934162f, 0.01164754293859005f, 0.009162282571196556f, 0.006141661666333675f,
-    0.0035081731621176f, 0.001707611023448408f, 0.002547456417232752f, 0.005233579315245152f, 0.009162282571196556f,
-    0.01366852037608624f, 0.01737609319388866f, 0.0188232995569706f, 0.01737609319388866f, 0.01366852037608624f,
-    0.009162282571196556f, 0.005233579315245152f, 0.002547456417232752f, 0.003238451667129993f, 0.00665318313986063f,
-    0.01164754293859005f, 0.01737609319388866f, 0.02208934165537357f, 0.02392910048365593f, 0.02208934165537357f,
-    0.01737609319388866f, 0.01164754293859005f, 0.00665318313986063f, 0.003238451667129993f, 0.001455130288377404f,
-    0.0035081731621176f, 0.00720730796456337f, 0.01261763460934162f, 0.0188232995569706f, 0.02392910048365593f,
-    0.02592208795249462f, 0.02392910048365593f, 0.0188232995569706f, 0.01261763460934162f, 0.00720730796456337f,
-    0.0035081731621176f, 0.001455130288377404f, 0.003238451667129993f, 0.00665318313986063f, 0.01164754293859005f,
-    0.01737609319388866f, 0.02208934165537357f, 0.02392910048365593f, 0.02208934165537357f, 0.01737609319388866f,
-    0.01164754293859005f, 0.00665318313986063f, 0.003238451667129993f, 0.002547456417232752f, 0.005233579315245152f,
-    0.009162282571196556f, 0.01366852037608624f, 0.01737609319388866f, 0.0188232995569706f, 0.01737609319388866f,
-    0.01366852037608624f, 0.009162282571196556f, 0.005233579315245152f, 0.002547456417232752f, 0.001707611023448408f,
-    0.0035081731621176f, 0.006141661666333675f, 0.009162282571196556f, 0.01164754293859005f, 0.01261763460934162f,
-    0.01164754293859005f, 0.009162282571196556f, 0.006141661666333675f, 0.0035081731621176f, 0.001707611023448408f,
-    0.002003900473937392f, 0.0035081731621176f, 0.005233579315245152f, 0.00665318313986063f, 0.00720730796456337f,
-    0.00665318313986063f, 0.005233579315245152f, 0.0035081731621176f, 0.002003900473937392f, 0.001707611023448408f,
-    0.002547456417232752f, 0.003238451667129993f, 0.0035081731621176f, 0.003238451667129993f, 0.002547456417232752f,
-    0.001707611023448408f, 0.001455130288377404f};
+                                        0.003238451667129993f, 0.002547456417232752f, 0.001707611023448408f, 0.002003900473937392f, 0.0035081731621176f, 0.005233579315245152f,
+                                        0.00665318313986063f, 0.00720730796456337f, 0.00665318313986063f, 0.005233579315245152f, 0.0035081731621176f,
+                                        0.002003900473937392f, 0.001707611023448408f, 0.0035081731621176f, 0.006141661666333675f, 0.009162282571196556f,
+                                        0.01164754293859005f, 0.01261763460934162f, 0.01164754293859005f, 0.009162282571196556f, 0.006141661666333675f,
+                                        0.0035081731621176f, 0.001707611023448408f, 0.002547456417232752f, 0.005233579315245152f, 0.009162282571196556f,
+                                        0.01366852037608624f, 0.01737609319388866f, 0.0188232995569706f, 0.01737609319388866f, 0.01366852037608624f,
+                                        0.009162282571196556f, 0.005233579315245152f, 0.002547456417232752f, 0.003238451667129993f, 0.00665318313986063f,
+                                        0.01164754293859005f, 0.01737609319388866f, 0.02208934165537357f, 0.02392910048365593f, 0.02208934165537357f,
+                                        0.01737609319388866f, 0.01164754293859005f, 0.00665318313986063f, 0.003238451667129993f, 0.001455130288377404f,
+                                        0.0035081731621176f, 0.00720730796456337f, 0.01261763460934162f, 0.0188232995569706f, 0.02392910048365593f,
+                                        0.02592208795249462f, 0.02392910048365593f, 0.0188232995569706f, 0.01261763460934162f, 0.00720730796456337f,
+                                        0.0035081731621176f, 0.001455130288377404f, 0.003238451667129993f, 0.00665318313986063f, 0.01164754293859005f,
+                                        0.01737609319388866f, 0.02208934165537357f, 0.02392910048365593f, 0.02208934165537357f, 0.01737609319388866f,
+                                        0.01164754293859005f, 0.00665318313986063f, 0.003238451667129993f, 0.002547456417232752f, 0.005233579315245152f,
+                                        0.009162282571196556f, 0.01366852037608624f, 0.01737609319388866f, 0.0188232995569706f, 0.01737609319388866f,
+                                        0.01366852037608624f, 0.009162282571196556f, 0.005233579315245152f, 0.002547456417232752f, 0.001707611023448408f,
+                                        0.0035081731621176f, 0.006141661666333675f, 0.009162282571196556f, 0.01164754293859005f, 0.01261763460934162f,
+                                        0.01164754293859005f, 0.009162282571196556f, 0.006141661666333675f, 0.0035081731621176f, 0.001707611023448408f,
+                                        0.002003900473937392f, 0.0035081731621176f, 0.005233579315245152f, 0.00665318313986063f, 0.00720730796456337f,
+                                        0.00665318313986063f, 0.005233579315245152f, 0.0035081731621176f, 0.002003900473937392f, 0.001707611023448408f,
+                                        0.002547456417232752f, 0.003238451667129993f, 0.0035081731621176f, 0.003238451667129993f, 0.002547456417232752f,
+                                        0.001707611023448408f, 0.001455130288377404f
+                                       };
 
-__constant float c_NX[2][5] = {{0, 0, 2, 4, -1}, {2, 0, 4, 4, 1}};
-__constant float c_NY[2][5] = {{0, 0, 4, 2, 1}, {0, 2, 4, 4, -1}};
+__constant float2 c_NX[5] = { (float2)(0, 2), (float2)(0, 0), (float2)(2, 4), (float2)(4, 4), (float2)(-1, 1) };
+__constant float2 c_NY[5] = { (float2)(0, 0), (float2)(0, 2), (float2)(4, 4), (float2)(2, 4), (float2)(1, -1) };
 
 void reduce_32_sum(volatile __local  float * data, volatile float* partial_reduction, int tid)
 {
 #define op(A, B) (*A)+(B)
     data[tid] = *partial_reduction;
     barrier(CLK_LOCAL_MEM_FENCE);
-
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
     if (tid < 16)
     {
         data[tid] = *partial_reduction = op(partial_reduction, data[tid + 16]);
-        data[tid] = *partial_reduction = op(partial_reduction, data[tid + 8 ]);
-        data[tid] = *partial_reduction = op(partial_reduction, data[tid + 4 ]);
+#if WAVE_SIZE < 16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        data[tid] = *partial_reduction = op(partial_reduction, data[tid + 8]);
+#if WAVE_SIZE < 8
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 4)
+    {
+#endif
+        data[tid] = *partial_reduction = op(partial_reduction, data[tid + 4]);
+#if WAVE_SIZE < 4
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 2)
+    {
+#endif
         data[tid] = *partial_reduction = op(partial_reduction, data[tid + 2 ]);
+#if WAVE_SIZE < 2
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 1)
+    {
+#endif
         data[tid] = *partial_reduction = op(partial_reduction, data[tid + 1 ]);
     }
+#undef WAVE_SIZE
 #undef op
 }
 
 __kernel
-    void icvCalcOrientation(
+void icvCalcOrientation(
     IMAGE_INT32 sumTex,
     __global float * keypoints,
     int keypoints_step,
     int c_img_rows,
     int c_img_cols,
     int sum_step
-    )
+)
 {
     keypoints_step /= sizeof(*keypoints);
     sum_step       /= sizeof(uint);
@@ -806,7 +867,7 @@ __kernel
         const int y = convert_int_rte(featureY[get_group_id(0)] + c_aptY[tid] * s - margin);
 
         if (y >= 0 && y < (c_img_rows + 1) - grad_wav_size &&
-            x >= 0 && x < (c_img_cols + 1) - grad_wav_size)
+                x >= 0 && x < (c_img_cols + 1) - grad_wav_size)
         {
             X = c_aptW[tid] * icvCalcHaarPatternSum_2(sumTex, c_NX, 4, grad_wav_size, y, x, c_img_rows, c_img_cols, sum_step);
             Y = c_aptW[tid] * icvCalcHaarPatternSum_2(sumTex, c_NY, 4, grad_wav_size, y, x, c_img_rows, c_img_cols, sum_step);
@@ -892,9 +953,9 @@ __kernel
             kp_dir += 2.0f * CV_PI_F;
         kp_dir *= 180.0f / CV_PI_F;
 
-        //kp_dir = 360.0f - kp_dir;
-        //if (fabs(kp_dir - 360.f) < FLT_EPSILON)
-        //    kp_dir = 0.f;
+        kp_dir = 360.0f - kp_dir;
+        if (fabs(kp_dir - 360.f) < FLT_EPSILON)
+            kp_dir = 0.f;
 
         featureDir[get_group_id(0)] = kp_dir;
     }
@@ -902,18 +963,18 @@ __kernel
 
 
 __kernel
-    void icvSetUpright(
+void icvSetUpright(
     __global float * keypoints,
     int keypoints_step,
     int nFeatures
-    )
+)
 {
     keypoints_step /= sizeof(*keypoints);
     __global float* featureDir  = keypoints + ANGLE_ROW * keypoints_step;
 
     if(get_global_id(0) <= nFeatures)
     {
-        featureDir[get_global_id(0)] = 90.0f;
+        featureDir[get_global_id(0)] = 270.0f;
     }
 }
 
@@ -956,7 +1017,7 @@ inline uchar readerGet(
     IMAGE_INT8 src,
     const float centerX, const float centerY, const float win_offset, const float cos_dir, const float sin_dir,
     int i, int j, int rows, int cols, int elemPerRow
-    )
+)
 {
     float pixel_x = centerX + (win_offset + j) * cos_dir + (win_offset + i) * sin_dir;
     float pixel_y = centerY - (win_offset + j) * sin_dir + (win_offset + i) * cos_dir;
@@ -967,7 +1028,7 @@ inline float linearFilter(
     IMAGE_INT8 src,
     const float centerX, const float centerY, const float win_offset, const float cos_dir, const float sin_dir,
     float y, float x, int rows, int cols, int elemPerRow
-    )
+)
 {
     x -= 0.5f;
     y -= 0.5f;
@@ -996,9 +1057,9 @@ inline float linearFilter(
 
 void calc_dx_dy(
     IMAGE_INT8 imgTex,
-    volatile __local  float s_dx_bin[25],
-    volatile __local  float s_dy_bin[25],
-    volatile __local  float s_PATCH[6][6],
+    volatile __local  float *s_dx_bin,
+    volatile __local  float *s_dy_bin,
+    volatile __local  float *s_PATCH,
     __global const float* featureX,
     __global const float* featureY,
     __global const float* featureSize,
@@ -1006,12 +1067,18 @@ void calc_dx_dy(
     int rows,
     int cols,
     int elemPerRow
-    )
+)
 {
     const float centerX = featureX[get_group_id(0)];
     const float centerY = featureY[get_group_id(0)];
     const float size = featureSize[get_group_id(0)];
-    float descriptor_dir = featureDir[get_group_id(0)] * (float)(CV_PI_F / 180.0f);
+    float descriptor_dir = 360.0f - featureDir[get_group_id(0)];
+    if(fabs(descriptor_dir - 360.0f) < FLT_EPSILON)
+    {
+        descriptor_dir = 0.0f;
+    }
+
+    descriptor_dir *= (float)(CV_PI_F / 180.0f);
 
     /* The sampling intervals and wavelet sized for selecting an orientation
     and building the keypoint descriptor are defined relative to 's' */
@@ -1037,7 +1104,7 @@ void calc_dx_dy(
     const float icoo = ((float)yIndex / (PATCH_SZ + 1)) * win_size;
     const float jcoo = ((float)xIndex / (PATCH_SZ + 1)) * win_size;
 
-    s_PATCH[get_local_id(1)][get_local_id(0)] = linearFilter(imgTex, centerX, centerY, win_offset, cos_dir, sin_dir, icoo, jcoo, rows, cols, elemPerRow);
+    s_PATCH[get_local_id(1) * 6 + get_local_id(0)] = linearFilter(imgTex, centerX, centerY, win_offset, cos_dir, sin_dir, icoo, jcoo, rows, cols, elemPerRow);
 
     barrier(CLK_LOCAL_MEM_FENCE);
 
@@ -1048,17 +1115,17 @@ void calc_dx_dy(
         const float dw = c_DW[yIndex * PATCH_SZ + xIndex];
 
         const float vx = (
-            s_PATCH[get_local_id(1)    ][get_local_id(0) + 1] -
-            s_PATCH[get_local_id(1)    ][get_local_id(0)    ] +
-            s_PATCH[get_local_id(1) + 1][get_local_id(0) + 1] -
-            s_PATCH[get_local_id(1) + 1][get_local_id(0)    ])
-            * dw;
+                             s_PATCH[      get_local_id(1) * 6 + get_local_id(0) + 1] -
+                             s_PATCH[      get_local_id(1) * 6 + get_local_id(0)    ] +
+                             s_PATCH[(get_local_id(1) + 1) * 6 + get_local_id(0) + 1] -
+                             s_PATCH[(get_local_id(1) + 1) * 6 + get_local_id(0)    ])
+                         * dw;
         const float vy = (
-            s_PATCH[get_local_id(1) + 1][get_local_id(0)    ] -
-            s_PATCH[get_local_id(1)    ][get_local_id(0)    ] +
-            s_PATCH[get_local_id(1) + 1][get_local_id(0) + 1] -
-            s_PATCH[get_local_id(1)    ][get_local_id(0) + 1])
-            * dw;
+                             s_PATCH[(get_local_id(1) + 1) * 6 + get_local_id(0)    ] -
+                             s_PATCH[      get_local_id(1) * 6 + get_local_id(0)    ] +
+                             s_PATCH[(get_local_id(1) + 1) * 6 + get_local_id(0) + 1] -
+                             s_PATCH[      get_local_id(1) * 6 + get_local_id(0) + 1])
+                         * dw;
         s_dx_bin[tid] = vx;
         s_dy_bin[tid] = vy;
     }
@@ -1069,53 +1136,73 @@ void reduce_sum25(
     volatile __local  float* sdata3,
     volatile __local  float* sdata4,
     int tid
-    )
+)
 {
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
     // first step is to reduce from 25 to 16
-    if (tid < 9) // use 9 threads
+    if (tid < 9)
     {
         sdata1[tid] += sdata1[tid + 16];
         sdata2[tid] += sdata2[tid + 16];
         sdata3[tid] += sdata3[tid + 16];
         sdata4[tid] += sdata4[tid + 16];
+#if WAVE_SIZE < 16
     }
-
-    // sum (reduce) from 16 to 1 (unrolled - aligned to a half-warp)
+    barrier(CLK_LOCAL_MEM_FENCE);
     if (tid < 8)
     {
+#endif
         sdata1[tid] += sdata1[tid + 8];
-        sdata1[tid] += sdata1[tid + 4];
-        sdata1[tid] += sdata1[tid + 2];
-        sdata1[tid] += sdata1[tid + 1];
-
         sdata2[tid] += sdata2[tid + 8];
-        sdata2[tid] += sdata2[tid + 4];
-        sdata2[tid] += sdata2[tid + 2];
-        sdata2[tid] += sdata2[tid + 1];
-
         sdata3[tid] += sdata3[tid + 8];
-        sdata3[tid] += sdata3[tid + 4];
-        sdata3[tid] += sdata3[tid + 2];
-        sdata3[tid] += sdata3[tid + 1];
-
         sdata4[tid] += sdata4[tid + 8];
+#if WAVE_SIZE < 8
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 4)
+    {
+#endif
+        sdata1[tid] += sdata1[tid + 4];
+        sdata2[tid] += sdata2[tid + 4];
+        sdata3[tid] += sdata3[tid + 4];
         sdata4[tid] += sdata4[tid + 4];
+#if WAVE_SIZE < 4
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 2)
+    {
+#endif
+        sdata1[tid] += sdata1[tid + 2];
+        sdata2[tid] += sdata2[tid + 2];
+        sdata3[tid] += sdata3[tid + 2];
         sdata4[tid] += sdata4[tid + 2];
+#if WAVE_SIZE < 2
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 1)
+    {
+#endif
+        sdata1[tid] += sdata1[tid + 1];
+        sdata2[tid] += sdata2[tid + 1];
+        sdata3[tid] += sdata3[tid + 1];
         sdata4[tid] += sdata4[tid + 1];
     }
+#undef WAVE_SIZE
 }
 
 __kernel
-    void compute_descriptors64(
+void compute_descriptors64(
     IMAGE_INT8 imgTex,
-    volatile __global float * descriptors,
+    __global float * descriptors,
     __global const float * keypoints,
     int descriptors_step,
     int keypoints_step,
     int rows,
     int cols,
     int img_step
-    )
+)
 {
     descriptors_step /= sizeof(float);
     keypoints_step   /= sizeof(float);
@@ -1129,7 +1216,7 @@ __kernel
     volatile __local  float sdy[25];
     volatile __local  float sdxabs[25];
     volatile __local  float sdyabs[25];
-    volatile __local  float s_PATCH[6][6];
+    volatile __local  float s_PATCH[6*6];
 
     calc_dx_dy(imgTex, sdx, sdy, s_PATCH, featureX, featureY, featureSize, featureDir, rows, cols, img_step);
     barrier(CLK_LOCAL_MEM_FENCE);
@@ -1142,14 +1229,13 @@ __kernel
         sdyabs[tid] = fabs(sdy[tid]); // |dy| array
     }
     barrier(CLK_LOCAL_MEM_FENCE);
-    if (tid < 25)
-    {
-        reduce_sum25(sdx, sdy, sdxabs, sdyabs, tid);
-    }
+
+    reduce_sum25(sdx, sdy, sdxabs, sdyabs, tid);
+
     barrier(CLK_LOCAL_MEM_FENCE);
     if (tid < 25)
     {
-        volatile __global float* descriptors_block = descriptors + descriptors_step * get_group_id(0) + (get_group_id(1) << 2);
+        __global float* descriptors_block = descriptors + descriptors_step * get_group_id(0) + (get_group_id(1) << 2);
 
         // write dx, dy, |dx|, |dy|
         if (tid == 0)
@@ -1162,16 +1248,16 @@ __kernel
     }
 }
 __kernel
-    void compute_descriptors128(
+void compute_descriptors128(
     IMAGE_INT8 imgTex,
-    __global volatile float * descriptors,
+    __global float * descriptors,
     __global float * keypoints,
     int descriptors_step,
     int keypoints_step,
     int rows,
     int cols,
     int img_step
-    )
+)
 {
     descriptors_step /= sizeof(*descriptors);
     keypoints_step   /= sizeof(*keypoints);
@@ -1190,7 +1276,7 @@ __kernel
     volatile __local  float sd2[25];
     volatile __local  float sdabs1[25];
     volatile __local  float sdabs2[25];
-    volatile __local  float s_PATCH[6][6];
+    volatile __local  float s_PATCH[6*6];
 
     calc_dx_dy(imgTex, sdx, sdy, s_PATCH, featureX, featureY, featureSize, featureDir, rows, cols, img_step);
     barrier(CLK_LOCAL_MEM_FENCE);
@@ -1213,13 +1299,15 @@ __kernel
             sd2[tid] = sdx[tid];
             sdabs2[tid] = fabs(sdx[tid]);
         }
-        //barrier(CLK_LOCAL_MEM_FENCE);
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
 
-        reduce_sum25(sd1, sd2, sdabs1, sdabs2, tid);
-        //barrier(CLK_LOCAL_MEM_FENCE);
-
-        volatile __global float* descriptors_block = descriptors + descriptors_step * get_group_id(0) + (get_group_id(1) << 3);
+    reduce_sum25(sd1, sd2, sdabs1, sdabs2, tid);
+    barrier(CLK_LOCAL_MEM_FENCE);
 
+    __global float* descriptors_block = descriptors + descriptors_step * get_group_id(0) + (get_group_id(1) << 3);
+    if (tid < 25)
+    {
         // write dx (dy >= 0), |dx| (dy >= 0), dx (dy < 0), |dx| (dy < 0)
         if (tid == 0)
         {
@@ -1243,11 +1331,13 @@ __kernel
             sd2[tid] = sdy[tid];
             sdabs2[tid] = fabs(sdy[tid]);
         }
-        //barrier(CLK_LOCAL_MEM_FENCE);
-
-        reduce_sum25(sd1, sd2, sdabs1, sdabs2, tid);
-        //barrier(CLK_LOCAL_MEM_FENCE);
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    reduce_sum25(sd1, sd2, sdabs1, sdabs2, tid);
+    barrier(CLK_LOCAL_MEM_FENCE);
 
+    if (tid < 25)
+    {
         // write dy (dx >= 0), |dy| (dx >= 0), dy (dx < 0), |dy| (dx < 0)
         if (tid == 0)
         {
@@ -1259,8 +1349,109 @@ __kernel
     }
 }
 
+void reduce_sum128(volatile __local  float* smem, int tid)
+{
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
+
+    if (tid < 64)
+    {
+        smem[tid] += smem[tid + 64];
+#if WAVE_SIZE < 64
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 32)
+    {
+#endif
+        smem[tid] += smem[tid + 32];
+#if WAVE_SIZE < 32
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16)
+    {
+#endif
+        smem[tid] += smem[tid + 16];
+#if WAVE_SIZE < 16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        smem[tid] += smem[tid + 8];
+#if WAVE_SIZE < 8
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 4)
+    {
+#endif
+        smem[tid] += smem[tid + 4];
+#if WAVE_SIZE < 4
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 2)
+    {
+#endif
+        smem[tid] += smem[tid + 2];
+#if WAVE_SIZE < 2
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 1)
+    {
+#endif
+        smem[tid] += smem[tid + 1];
+    }
+}
+
+
+void reduce_sum64(volatile __local  float* smem, int tid)
+{
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
+    if (tid < 32)
+    {
+        smem[tid] += smem[tid + 32];
+#if WAVE_SIZE < 32
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16)
+    {
+#endif
+        smem[tid] += smem[tid + 16];
+#if WAVE_SIZE < 16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        smem[tid] += smem[tid + 8];
+#if WAVE_SIZE < 8
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 4)
+    {
+#endif
+        smem[tid] += smem[tid + 4];
+#if WAVE_SIZE < 4
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 2)
+    {
+#endif
+        smem[tid] += smem[tid + 2];
+#if WAVE_SIZE < 2
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 1)
+    {
+#endif
+        smem[tid] += smem[tid + 1];
+    }
+}
+
 __kernel
-    void normalize_descriptors128(__global float * descriptors, int descriptors_step)
+void normalize_descriptors128(__global float * descriptors, int descriptors_step)
 {
     descriptors_step /= sizeof(*descriptors);
     // no need for thread ID
@@ -1272,23 +1463,9 @@ __kernel
     sqDesc[get_local_id(0)] = lookup * lookup;
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    if (get_local_id(0) < 64)
-        sqDesc[get_local_id(0)] += sqDesc[get_local_id(0) + 64];
+    reduce_sum128(sqDesc, get_local_id(0));
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    // reduction to get total
-    if (get_local_id(0) < 32)
-    {
-        volatile __local  float* smem = sqDesc;
-
-        smem[get_local_id(0)] += smem[get_local_id(0) + 32];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 16];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 8];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 4];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 2];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 1];
-    }
-
     // compute length (square root)
     volatile __local  float len;
     if (get_local_id(0) == 0)
@@ -1301,7 +1478,7 @@ __kernel
     descriptor_base[get_local_id(0)] = lookup / len;
 }
 __kernel
-    void normalize_descriptors64(__global float * descriptors, int descriptors_step)
+void normalize_descriptors64(__global float * descriptors, int descriptors_step)
 {
     descriptors_step /= sizeof(*descriptors);
     // no need for thread ID
@@ -1313,18 +1490,8 @@ __kernel
     sqDesc[get_local_id(0)] = lookup * lookup;
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    // reduction to get total
-    if (get_local_id(0) < 32)
-    {
-        volatile __local  float* smem = sqDesc;
-
-        smem[get_local_id(0)] += smem[get_local_id(0) + 32];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 16];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 8];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 4];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 2];
-        smem[get_local_id(0)] += smem[get_local_id(0) + 1];
-    }
+    reduce_sum64(sqDesc, get_local_id(0));
+    barrier(CLK_LOCAL_MEM_FENCE);
 
     // compute length (square root)
     volatile __local  float len;
diff --git a/modules/nonfree/src/precomp.hpp b/modules/nonfree/src/precomp.hpp
index ea0aa6bd4a..7563be2284 100644
--- a/modules/nonfree/src/precomp.hpp
+++ b/modules/nonfree/src/precomp.hpp
@@ -43,12 +43,26 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/nonfree.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/core/internal.hpp"
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
+
+#include "opencv2/nonfree/gpu.hpp"
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_OCL
+#  include "opencv2/nonfree/ocl.hpp"
+#  include "opencv2/ocl/private/util.hpp"
+#endif
+
+#include "opencv2/core/private.hpp"
 
 #endif
diff --git a/modules/nonfree/src/sift.cpp b/modules/nonfree/src/sift.cpp
index 4161c4911e..68216f58bb 100644
--- a/modules/nonfree/src/sift.cpp
+++ b/modules/nonfree/src/sift.cpp
@@ -732,10 +732,10 @@ void SIFT::operator()(InputArray _image, InputArray _mask,
     Mat image = _image.getMat(), mask = _mask.getMat();
 
     if( image.empty() || image.depth() != CV_8U )
-        CV_Error( CV_StsBadArg, "image is empty or has incorrect depth (!=CV_8U)" );
+        CV_Error( Error::StsBadArg, "image is empty or has incorrect depth (!=CV_8U)" );
 
     if( !mask.empty() && mask.type() != CV_8UC1 )
-        CV_Error( CV_StsBadArg, "mask has incorrect type (!=CV_8UC1)" );
+        CV_Error( Error::StsBadArg, "mask has incorrect type (!=CV_8UC1)" );
 
     if( useProvidedKeypoints )
     {
@@ -774,9 +774,6 @@ void SIFT::operator()(InputArray _image, InputArray _mask,
         findScaleSpaceExtrema(gpyr, dogpyr, keypoints);
         KeyPointsFilter::removeDuplicated( keypoints );
 
-        if( !mask.empty() )
-            KeyPointsFilter::runByPixelsMask( keypoints, mask );
-
         if( nfeatures > 0 )
             KeyPointsFilter::retainBest(keypoints, nfeatures);
         //t = (double)getTickCount() - t;
@@ -791,6 +788,9 @@ void SIFT::operator()(InputArray _image, InputArray _mask,
                 kpt.pt *= scale;
                 kpt.size *= scale;
             }
+
+        if( !mask.empty() )
+            KeyPointsFilter::runByPixelsMask( keypoints, mask );
     }
     else
     {
diff --git a/modules/nonfree/src/surf.cpp b/modules/nonfree/src/surf.cpp
index ea310478bd..98d1449828 100644
--- a/modules/nonfree/src/surf.cpp
+++ b/modules/nonfree/src/surf.cpp
@@ -258,7 +258,7 @@ interpolateKeypoint( float N9[3][9], int dx, int dy, int ds, KeyPoint& kpt )
 }
 
 // Multi-threaded construction of the scale-space pyramid
-struct SURFBuildInvoker
+struct SURFBuildInvoker : ParallelLoopBody
 {
     SURFBuildInvoker( const Mat& _sum, const std::vector<int>& _sizes,
                       const std::vector<int>& _sampleSteps,
@@ -271,9 +271,9 @@ struct SURFBuildInvoker
         traces = &_traces;
     }
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        for( int i=range.begin(); i<range.end(); i++ )
+        for( int i=range.start; i<range.end; i++ )
             calcLayerDetAndTrace( *sum, (*sizes)[i], (*sampleSteps)[i], (*dets)[i], (*traces)[i] );
     }
 
@@ -285,7 +285,7 @@ struct SURFBuildInvoker
 };
 
 // Multi-threaded search of the scale-space pyramid for keypoints
-struct SURFFindInvoker
+struct SURFFindInvoker : ParallelLoopBody
 {
     SURFFindInvoker( const Mat& _sum, const Mat& _mask_sum,
                      const std::vector<Mat>& _dets, const std::vector<Mat>& _traces,
@@ -310,9 +310,9 @@ struct SURFFindInvoker
                    const std::vector<int>& sizes, std::vector<KeyPoint>& keypoints,
                    int octave, int layer, float hessianThreshold, int sampleStep );
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        for( int i=range.begin(); i<range.end(); i++ )
+        for( int i=range.start; i<range.end; i++ )
         {
             int layer = (*middleIndices)[i];
             int octave = i / nOctaveLayers;
@@ -333,14 +333,10 @@ struct SURFFindInvoker
     int nOctaveLayers;
     float hessianThreshold;
 
-#ifdef HAVE_TBB
-    static tbb::mutex findMaximaInLayer_m;
-#endif
+    static Mutex findMaximaInLayer_m;
 };
 
-#ifdef HAVE_TBB
-tbb::mutex SURFFindInvoker::findMaximaInLayer_m;
-#endif
+Mutex SURFFindInvoker::findMaximaInLayer_m;
 
 
 /*
@@ -427,7 +423,7 @@ void SURFFindInvoker::findMaximaInLayer( const Mat& sum, const Mat& mask_sum,
                     float center_j = sum_j + (size-1)*0.5f;
 
                     KeyPoint kpt( center_j, center_i, (float)sizes[layer],
-                                  -1, val0, octave, CV_SIGN(trace_ptr[j]) );
+                                  -1, val0, octave, (trace_ptr[j] > 0) - (trace_ptr[j] < 0) );
 
                     /* Interpolate maxima location within the 3x3x3 neighbourhood  */
                     int ds = size - sizes[layer-1];
@@ -437,9 +433,7 @@ void SURFFindInvoker::findMaximaInLayer( const Mat& sum, const Mat& mask_sum,
                     if( interp_ok  )
                     {
                         /*printf( "KeyPoint %f %f %d\n", point.pt.x, point.pt.y, point.size );*/
-#ifdef HAVE_TBB
-                        tbb::mutex::scoped_lock lock(findMaximaInLayer_m);
-#endif
+                        cv::AutoLock lock(findMaximaInLayer_m);
                         keypoints.push_back(kpt);
                     }
                 }
@@ -505,20 +499,20 @@ static void fastHessianDetector( const Mat& sum, const Mat& mask_sum, std::vecto
     }
 
     // Calculate hessian determinant and trace samples in each layer
-    parallel_for( BlockedRange(0, nTotalLayers),
-                      SURFBuildInvoker(sum, sizes, sampleSteps, dets, traces) );
+    parallel_for_( Range(0, nTotalLayers),
+                   SURFBuildInvoker(sum, sizes, sampleSteps, dets, traces) );
 
     // Find maxima in the determinant of the hessian
-    parallel_for( BlockedRange(0, nMiddleLayers),
-                      SURFFindInvoker(sum, mask_sum, dets, traces, sizes,
-                                      sampleSteps, middleIndices, keypoints,
-                                      nOctaveLayers, hessianThreshold) );
+    parallel_for_( Range(0, nMiddleLayers),
+                   SURFFindInvoker(sum, mask_sum, dets, traces, sizes,
+                                   sampleSteps, middleIndices, keypoints,
+                                   nOctaveLayers, hessianThreshold) );
 
     std::sort(keypoints.begin(), keypoints.end(), KeypointGreater());
 }
 
 
-struct SURFInvoker
+struct SURFInvoker : ParallelLoopBody
 {
     enum { ORI_RADIUS = 6, ORI_WIN = 60, PATCH_SZ = 20 };
 
@@ -550,7 +544,7 @@ struct SURFInvoker
             {
                 if( i*i + j*j <= ORI_RADIUS*ORI_RADIUS )
                 {
-                    apt[nOriSamples] = cvPoint(i,j);
+                    apt[nOriSamples] = Point(i,j);
                     aptw[nOriSamples++] = G_ori.at<float>(i+ORI_RADIUS,0) * G_ori.at<float>(j+ORI_RADIUS,0);
                 }
             }
@@ -566,7 +560,7 @@ struct SURFInvoker
         }
     }
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
         /* X and Y gradient wavelet data */
         const int NX=2, NY=2;
@@ -580,21 +574,19 @@ struct SURFInvoker
         float X[nOriSampleBound], Y[nOriSampleBound], angle[nOriSampleBound];
         uchar PATCH[PATCH_SZ+1][PATCH_SZ+1];
         float DX[PATCH_SZ][PATCH_SZ], DY[PATCH_SZ][PATCH_SZ];
-        CvMat matX = cvMat(1, nOriSampleBound, CV_32F, X);
-        CvMat matY = cvMat(1, nOriSampleBound, CV_32F, Y);
-        CvMat _angle = cvMat(1, nOriSampleBound, CV_32F, angle);
         Mat _patch(PATCH_SZ+1, PATCH_SZ+1, CV_8U, PATCH);
 
         int dsize = extended ? 128 : 64;
 
-        int k, k1 = range.begin(), k2 = range.end();
+        int k, k1 = range.start, k2 = range.end;
         float maxSize = 0;
         for( k = k1; k < k2; k++ )
         {
             maxSize = std::max(maxSize, (*keypoints)[k].size);
         }
         int imaxSize = std::max(cvCeil((PATCH_SZ+1)*maxSize*1.2f/9.0f), 1);
-        Ptr<CvMat> winbuf = cvCreateMat( 1, imaxSize*imaxSize, CV_8U );
+        cv::AutoBuffer<uchar> winbuf(imaxSize*imaxSize);
+
         for( k = k1; k < k2; k++ )
         {
             int i, j, kk, nangle;
@@ -648,8 +640,8 @@ struct SURFInvoker
                     kp.size = -1;
                     continue;
                 }
-                matX.cols = matY.cols = _angle.cols = nangle;
-                cvCartToPolar( &matX, &matY, 0, &_angle, 1 );
+
+                phase( Mat(1, nangle, CV_32F, X), Mat(1, nangle, CV_32F, Y), Mat(1, nangle, CV_32F, angle), true );
 
                 float bestx = 0, besty = 0, descriptor_mod = 0;
                 for( i = 0; i < 360; i += SURF_ORI_SEARCH_INC )
@@ -680,8 +672,8 @@ struct SURFInvoker
 
             /* Extract a window of pixels around the keypoint of size 20s */
             int win_size = (int)((PATCH_SZ+1)*s);
-            CV_Assert( winbuf->cols >= win_size*win_size );
-            Mat win(win_size, win_size, CV_8U, winbuf->data.ptr);
+            CV_Assert( imaxSize >= win_size );
+            Mat win(win_size, win_size, CV_8U, winbuf);
 
             if( !upright )
             {
@@ -954,7 +946,7 @@ void SURF::operator()(InputArray _img, InputArray _mask,
 
         // we call SURFInvoker in any case, even if we do not need descriptors,
         // since it computes orientation of each feature.
-        parallel_for(BlockedRange(0, N), SURFInvoker(img, sum, keypoints, descriptors, extended, upright) );
+        parallel_for_(Range(0, N), SURFInvoker(img, sum, keypoints, descriptors, extended, upright) );
 
         // remove keypoints that were marked for deletion
         for( i = j = 0; i < N; i++ )
diff --git a/modules/ocl/src/surf.cpp b/modules/nonfree/src/surf.ocl.cpp
similarity index 89%
rename from modules/ocl/src/surf.cpp
rename to modules/nonfree/src/surf.ocl.cpp
index 5d14392286..698168f0b6 100644
--- a/modules/ocl/src/surf.cpp
+++ b/modules/nonfree/src/surf.ocl.cpp
@@ -42,9 +42,10 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
-#include <iomanip>
 #include "precomp.hpp"
-#include "mcwutil.hpp"
+#include <cstdio>
+
+#ifdef HAVE_OPENCV_OCL
 
 using namespace cv;
 using namespace cv::ocl;
@@ -54,30 +55,37 @@ namespace cv
     namespace ocl
     {
         ///////////////////////////OpenCL kernel strings///////////////////////////
-        extern const char *nonfree_surf;
+        extern const char *surf;
 
-        const char* noImage2dOption = "-D DISABLE_IMAGE2D";
+        const char noImage2dOption [] = "-D DISABLE_IMAGE2D";
 
-        static void openCLExecuteKernelSURF(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
-            size_t localThreads[3], std::vector< std::pair<size_t, const void *> > &args, int channels, int depth)
+        static bool use_image2d = false;
+
+        static void openCLExecuteKernelSURF(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
+            size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth)
         {
-            if(support_image2d())
+            char optBuf [100] = {0};
+            char * optBufPtr = optBuf;
+            if( !use_image2d )
             {
-                openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args, channels, depth);
-            }
-            else
-            {
-                openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args, channels, depth, noImage2dOption);
+                strcat(optBufPtr, noImage2dOption);
+                optBufPtr += strlen(noImage2dOption);
             }
+            cl_kernel kernel;
+            kernel = openCLGetKernelFromSource(clCxt, source, kernelName, optBufPtr);
+            size_t wave_size = queryDeviceInfo<WAVEFRONT_SIZE, size_t>(kernel);
+            CV_Assert(clReleaseKernel(kernel) == CL_SUCCESS);
+            sprintf(optBufPtr, "-D WAVE_SIZE=%d", static_cast<int>(wave_size));
+            openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args, channels, depth, optBufPtr);
         }
     }
 }
 
-
-static inline int divUp(int total, int grain)
+static inline size_t divUp(size_t total, size_t grain)
 {
     return (total + grain - 1) / grain;
 }
+
 static inline int calcSize(int octave, int layer)
 {
     /* Wavelet size at first layer of first octave. */
@@ -148,17 +156,19 @@ public:
         counters.setTo(Scalar::all(0));
 
         integral(img, surf_.sum);
-        if(support_image2d())
+        use_image2d = support_image2d();
+        if(use_image2d)
         {
-        bindImgTex(img, imgTex);
+            bindImgTex(img, imgTex);
             bindImgTex(surf_.sum, sumTex);
+            finish();
         }
 
         maskSumTex = 0;
 
         if (use_mask)
         {
-            throw std::exception();
+            CV_Error(Error::StsBadFunc, "Masked SURF detector is not implemented yet");
             //!FIXME
             // temp fix for missing min overload
             //oclMat temp(mask.size(), mask.type());
@@ -479,7 +489,7 @@ void SURF_OCL_Invoker::icvCalcLayerDetAndTrace_gpu(oclMat &det, oclMat &trace, i
     const int max_samples_j = 1 + ((img_cols - min_size) >> octave);
 
     Context *clCxt = det.clCxt;
-    std::string kernelName = "icvCalcLayerDetAndTrace";
+    String kernelName = "icvCalcLayerDetAndTrace";
     std::vector< std::pair<size_t, const void *> > args;
 
     if(sumTex)
@@ -508,16 +518,16 @@ void SURF_OCL_Invoker::icvCalcLayerDetAndTrace_gpu(oclMat &det, oclMat &trace, i
         divUp(max_samples_i, localThreads[1]) *localThreads[1] *(nOctaveLayers + 2),
         1
     };
-    openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void SURF_OCL_Invoker::icvFindMaximaInLayer_gpu(const oclMat &det, const oclMat &trace, oclMat &maxPosBuffer, oclMat &maxCounter, int counterOffset,
-        int octave, bool use_mask, int nLayers, int layer_rows, int layer_cols)
+        int octave, bool useMask, int nLayers, int layer_rows, int layer_cols)
 {
     const int min_margin = ((calcSize(octave, 2) >> 1) >> octave) + 1;
 
     Context *clCxt = det.clCxt;
-    std::string kernelName = use_mask ? "icvFindMaximaInLayer_withmask" : "icvFindMaximaInLayer";
+    String kernelName = use_mask ? "icvFindMaximaInLayer_withmask" : "icvFindMaximaInLayer";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&det.data));
@@ -536,7 +546,7 @@ void SURF_OCL_Invoker::icvFindMaximaInLayer_gpu(const oclMat &det, const oclMat
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&maxCandidates));
     args.push_back( std::make_pair( sizeof(cl_float), (void *)&surf_.hessianThreshold));
 
-    if(use_mask)
+    if(useMask)
     {
         if(maskSumTex)
         {
@@ -554,38 +564,38 @@ void SURF_OCL_Invoker::icvFindMaximaInLayer_gpu(const oclMat &det, const oclMat
                                1
                               };
 
-    openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void SURF_OCL_Invoker::icvInterpolateKeypoint_gpu(const oclMat &det, const oclMat &maxPosBuffer, int maxCounter,
-        oclMat &keypoints, oclMat &counters, int octave, int layer_rows, int maxFeatures)
+        oclMat &keypoints, oclMat &counters_, int octave, int layer_rows, int max_features)
 {
     Context *clCxt = det.clCxt;
-    std::string kernelName = "icvInterpolateKeypoint";
+    String kernelName = "icvInterpolateKeypoint";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&det.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&maxPosBuffer.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&keypoints.data));
-    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&counters.data));
+    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&counters_.data));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&det.step));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&keypoints.step));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_rows));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_cols));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&octave));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&layer_rows));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&maxFeatures));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&max_features));
 
     size_t localThreads[3]  = {3, 3, 3};
     size_t globalThreads[3] = {maxCounter *localThreads[0], localThreads[1], 1};
 
-    openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void SURF_OCL_Invoker::icvCalcOrientation_gpu(const oclMat &keypoints, int nFeatures)
 {
     Context *clCxt = counters.clCxt;
-    std::string kernelName = "icvCalcOrientation";
+    String kernelName = "icvCalcOrientation";
 
     std::vector< std::pair<size_t, const void *> > args;
 
@@ -606,13 +616,13 @@ void SURF_OCL_Invoker::icvCalcOrientation_gpu(const oclMat &keypoints, int nFeat
     size_t localThreads[3]  = {32, 4, 1};
     size_t globalThreads[3] = {nFeatures *localThreads[0], localThreads[1], 1};
 
-    openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void SURF_OCL_Invoker::icvSetUpright_gpu(const oclMat &keypoints, int nFeatures)
 {
     Context *clCxt = counters.clCxt;
-    std::string kernelName = "icvSetUpright";
+    String kernelName = "icvSetUpright";
 
     std::vector< std::pair<size_t, const void *> > args;
 
@@ -621,9 +631,9 @@ void SURF_OCL_Invoker::icvSetUpright_gpu(const oclMat &keypoints, int nFeatures)
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&nFeatures));
 
     size_t localThreads[3]  = {256, 1, 1};
-    size_t globalThreads[3] = {nFeatures, 1, 1};
+    size_t globalThreads[3] = {saturate_cast<size_t>(nFeatures), 1, 1};
 
-    openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 
@@ -631,7 +641,7 @@ void SURF_OCL_Invoker::compute_descriptors_gpu(const oclMat &descriptors, const
 {
     // compute unnormalized descriptors, then normalize them - odd indexing since grid must be 2D
     Context *clCxt = descriptors.clCxt;
-    std::string kernelName = "";
+    String kernelName;
     std::vector< std::pair<size_t, const void *> > args;
     size_t localThreads[3]  = {1, 1, 1};
     size_t globalThreads[3] = {1, 1, 1};
@@ -663,7 +673,7 @@ void SURF_OCL_Invoker::compute_descriptors_gpu(const oclMat &descriptors, const
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&_img.cols));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&_img.step));
 
-        openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+        openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 
         kernelName = "normalize_descriptors64";
 
@@ -677,7 +687,7 @@ void SURF_OCL_Invoker::compute_descriptors_gpu(const oclMat &descriptors, const
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&descriptors.data));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&descriptors.step));
 
-        openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+        openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
     }
     else
     {
@@ -706,7 +716,7 @@ void SURF_OCL_Invoker::compute_descriptors_gpu(const oclMat &descriptors, const
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&_img.cols));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&_img.step));
 
-        openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+        openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
 
         kernelName = "normalize_descriptors128";
 
@@ -720,7 +730,8 @@ void SURF_OCL_Invoker::compute_descriptors_gpu(const oclMat &descriptors, const
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&descriptors.data));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&descriptors.step));
 
-        openCLExecuteKernelSURF(clCxt, &nonfree_surf, kernelName, globalThreads, localThreads, args, -1, -1);
+        openCLExecuteKernelSURF(clCxt, &surf, kernelName, globalThreads, localThreads, args, -1, -1);
     }
 }
 
+#endif //HAVE_OPENCV_OCL
diff --git a/modules/gpu/src/surf.cpp b/modules/nonfree/src/surf_gpu.cpp
similarity index 92%
rename from modules/gpu/src/surf.cpp
rename to modules/nonfree/src/surf_gpu.cpp
index 123c8ce6ea..35805470b2 100644
--- a/modules/gpu/src/surf.cpp
+++ b/modules/nonfree/src/surf_gpu.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -45,24 +45,24 @@
 using namespace cv;
 using namespace cv::gpu;
 
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
+#if !defined (HAVE_CUDA) || !defined (HAVE_OPENCV_GPUARITHM)
 
-cv::gpu::SURF_GPU::SURF_GPU() { throw_nogpu(); }
-cv::gpu::SURF_GPU::SURF_GPU(double, int, int, bool, float, bool) { throw_nogpu(); }
-int cv::gpu::SURF_GPU::descriptorSize() const { throw_nogpu(); return 0;}
-void cv::gpu::SURF_GPU::uploadKeypoints(const std::vector<KeyPoint>&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::downloadKeypoints(const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::downloadDescriptors(const GpuMat&, std::vector<float>&) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, GpuMat&, bool) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, std::vector<float>&, bool) { throw_nogpu(); }
-void cv::gpu::SURF_GPU::releaseMemory() { throw_nogpu(); }
+cv::gpu::SURF_GPU::SURF_GPU() { throw_no_cuda(); }
+cv::gpu::SURF_GPU::SURF_GPU(double, int, int, bool, float, bool) { throw_no_cuda(); }
+int cv::gpu::SURF_GPU::descriptorSize() const { throw_no_cuda(); return 0;}
+void cv::gpu::SURF_GPU::uploadKeypoints(const std::vector<KeyPoint>&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::downloadKeypoints(const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::downloadDescriptors(const GpuMat&, std::vector<float>&) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, GpuMat&, bool) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::operator()(const GpuMat&, const GpuMat&, std::vector<KeyPoint>&, std::vector<float>&, bool) { throw_no_cuda(); }
+void cv::gpu::SURF_GPU::releaseMemory() { throw_no_cuda(); }
 
-#else /* !defined (HAVE_CUDA) */
+#else // !defined (HAVE_CUDA)
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace surf
     {
@@ -89,7 +89,7 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-using namespace ::cv::gpu::device::surf;
+using namespace ::cv::gpu::cudev::surf;
 
 namespace
 {
@@ -142,13 +142,13 @@ namespace
 
             bindImgTex(img);
 
-            integralBuffered(img, surf_.sum, surf_.intBuffer);
+            gpu::integral(img, surf_.sum, surf_.intBuffer);
             sumOffset = bindSumTex(surf_.sum);
 
             if (use_mask)
             {
-                min(mask, 1.0, surf_.mask1);
-                integralBuffered(surf_.mask1, surf_.maskSum, surf_.intBuffer);
+                gpu::min(mask, 1.0, surf_.mask1);
+                gpu::integral(surf_.mask1, surf_.maskSum, surf_.intBuffer);
                 maskOffset = bindMaskSumTex(surf_.maskSum);
             }
         }
@@ -221,6 +221,9 @@ namespace
         }
 
     private:
+        SURF_GPU_Invoker(const SURF_GPU_Invoker&);
+        SURF_GPU_Invoker& operator =(const SURF_GPU_Invoker&);
+
         SURF_GPU& surf_;
 
         int img_cols, img_rows;
@@ -415,4 +418,4 @@ void cv::gpu::SURF_GPU::releaseMemory()
     maxPosBuffer.release();
 }
 
-#endif /* !defined (HAVE_CUDA) */
+#endif // !defined (HAVE_CUDA)
diff --git a/modules/nonfree/test/test_detectors.cpp b/modules/nonfree/test/test_detectors.cpp
index e675d3f99e..5854207dcd 100644
--- a/modules/nonfree/test/test_detectors.cpp
+++ b/modules/nonfree/test/test_detectors.cpp
@@ -113,10 +113,10 @@ void showOrig(const Mat& img, const vector<KeyPoint>& orig_pts)
 {
 
     Mat img_color;
-    cvtColor(img, img_color, CV_GRAY2BGR);
+    cvtColor(img, img_color, COLOR_GRAY2BGR);
 
     for(size_t i = 0; i < orig_pts.size(); ++i)
-        circle(img_color, orig_pts[i].pt, (int)orig_pts[i].size/2, CV_RGB(0, 255, 0));
+        circle(img_color, orig_pts[i].pt, (int)orig_pts[i].size/2, Scalar(0, 255, 0));
 
     namedWindow("O"); imshow("O", img_color);
 }
@@ -125,13 +125,13 @@ void show(const string& name, const Mat& new_img, const vector<KeyPoint>& new_pt
 {
 
     Mat new_img_color;
-    cvtColor(new_img, new_img_color, CV_GRAY2BGR);
+    cvtColor(new_img, new_img_color, COLOR_GRAY2BGR);
 
     for(size_t i = 0; i < transf_pts.size(); ++i)
-        circle(new_img_color, transf_pts[i].pt, (int)transf_pts[i].size/2, CV_RGB(255, 0, 0));
+        circle(new_img_color, transf_pts[i].pt, (int)transf_pts[i].size/2, Scalar(255, 0, 0));
 
     for(size_t i = 0; i < new_pts.size(); ++i)
-        circle(new_img_color, new_pts[i].pt, (int)new_pts[i].size/2, CV_RGB(0, 0, 255));
+        circle(new_img_color, new_pts[i].pt, (int)new_pts[i].size/2, Scalar(0, 0, 255));
 
     namedWindow(name + "_T"); imshow(name + "_T", new_img_color);
 }
@@ -312,6 +312,3 @@ void CV_DetectorsTest::run( int /*start_from*/ )
 
 
 TEST(Features2d_Detectors, regression) { CV_DetectorsTest test; test.safe_run(); }
-
-
-
diff --git a/modules/nonfree/test/test_features2d.cpp b/modules/nonfree/test/test_features2d.cpp
index 5de09f2a09..7c888e261e 100644
--- a/modules/nonfree/test/test_features2d.cpp
+++ b/modules/nonfree/test/test_features2d.cpp
@@ -231,7 +231,7 @@ void CV_FeatureDetectorTest::regressionTest()
 
 void CV_FeatureDetectorTest::run( int /*start_from*/ )
 {
-    if( fdetector.empty() )
+    if( !fdetector )
     {
         ts->printf( cvtest::TS::LOG, "Feature detector is empty.\n" );
         ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
@@ -275,12 +275,16 @@ static Mat readMatFromBin( const string& filename )
         size_t elements_read4 = fread( (void*)&dataSize, sizeof(int), 1, f );
         CV_Assert(elements_read1 == 1 && elements_read2 == 1 && elements_read3 == 1 && elements_read4 == 1);
 
-        uchar* data = (uchar*)cvAlloc(dataSize);
-        size_t elements_read = fread( (void*)data, 1, dataSize, f );
+        size_t step = dataSize / rows / CV_ELEM_SIZE(type);
+        CV_Assert(step >= (size_t)cols);
+
+        Mat m = Mat( rows, step, type).colRange(0, cols);
+
+        size_t elements_read = fread( m.ptr(), 1, dataSize, f );
         CV_Assert(elements_read == (size_t)(dataSize));
         fclose(f);
 
-        return Mat( rows, cols, type, data );
+        return m;
     }
     return Mat();
 }
@@ -402,7 +406,7 @@ protected:
             double t = (double)getTickCount();
             dextractor->compute( img, keypoints, calcDescriptors );
             t = getTickCount() - t;
-            ts->printf(cvtest::TS::LOG, "\nAverage time of computing one descriptor = %g ms.\n", t/((double)cvGetTickFrequency()*1000.)/calcDescriptors.rows );
+            ts->printf(cvtest::TS::LOG, "\nAverage time of computing one descriptor = %g ms.\n", t/((double)getTickFrequency()*1000.)/calcDescriptors.rows );
 
             if( calcDescriptors.rows != (int)keypoints.size() )
             {
@@ -460,7 +464,7 @@ protected:
     void run(int)
     {
         createDescriptorExtractor();
-        if( dextractor.empty() )
+        if( !dextractor )
         {
             ts->printf(cvtest::TS::LOG, "Descriptor extractor is empty.\n");
             ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
@@ -1097,7 +1101,7 @@ protected:
     void run(int)
     {
         Ptr<Feature2D> f = Algorithm::create<Feature2D>("Feature2D." + fname);
-        if(f.empty())
+        if(!f)
             return;
         string path = string(ts->get_data_path()) + "detectors_descriptors_evaluation/planar/";
         string imgname1 = path + "box.png";
@@ -1145,3 +1149,76 @@ protected:
 
 TEST(Features2d_SIFTHomographyTest, regression) { CV_DetectPlanarTest test("SIFT", 80); test.safe_run(); }
 TEST(Features2d_SURFHomographyTest, regression) { CV_DetectPlanarTest test("SURF", 80); test.safe_run(); }
+
+class FeatureDetectorUsingMaskTest : public cvtest::BaseTest
+{
+public:
+    FeatureDetectorUsingMaskTest(const Ptr<FeatureDetector>& featureDetector) :
+        featureDetector_(featureDetector)
+    {
+        CV_Assert(featureDetector_);
+    }
+
+protected:
+
+    void run(int)
+    {
+        const int nStepX = 2;
+        const int nStepY = 2;
+
+        const string imageFilename = string(ts->get_data_path()) + "/features2d/tsukuba.png";
+
+        Mat image = imread(imageFilename);
+        if(image.empty())
+        {
+            ts->printf(cvtest::TS::LOG, "Image %s can not be read.\n", imageFilename.c_str());
+            ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_TEST_DATA);
+            return;
+        }
+
+        Mat mask(image.size(), CV_8U);
+
+        const int stepX = image.size().width / nStepX;
+        const int stepY = image.size().height / nStepY;
+
+        vector<KeyPoint> keyPoints;
+        vector<Point2f> points;
+        for(int i=0; i<nStepX; ++i)
+            for(int j=0; j<nStepY; ++j)
+            {
+
+                mask.setTo(0);
+                Rect whiteArea(i * stepX, j * stepY, stepX, stepY);
+                mask(whiteArea).setTo(255);
+
+                featureDetector_->detect(image, keyPoints, mask);
+                KeyPoint::convert(keyPoints, points);
+
+                for(size_t k=0; k<points.size(); ++k)
+                {
+                    if ( !whiteArea.contains(points[k]) )
+                    {
+                        ts->printf(cvtest::TS::LOG, "The feature point is outside of the mask.");
+                        ts->set_failed_test_info(cvtest::TS::FAIL_INVALID_OUTPUT);
+                        return;
+                    }
+                }
+            }
+
+        ts->set_failed_test_info( cvtest::TS::OK );
+    }
+
+    Ptr<FeatureDetector> featureDetector_;
+};
+
+TEST(Features2d_SIFT_using_mask, regression)
+{
+    FeatureDetectorUsingMaskTest test(Algorithm::create<FeatureDetector>("Feature2D.SIFT"));
+    test.safe_run();
+}
+
+TEST(DISABLED_Features2d_SURF_using_mask, regression)
+{
+    FeatureDetectorUsingMaskTest test(Algorithm::create<FeatureDetector>("Feature2D.SURF"));
+    test.safe_run();
+}
diff --git a/modules/nonfree/test/test_gpu.cpp b/modules/nonfree/test/test_gpu.cpp
new file mode 100644
index 0000000000..2836d6b116
--- /dev/null
+++ b/modules/nonfree/test/test_gpu.cpp
@@ -0,0 +1,198 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+using namespace cvtest;
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// SURF
+
+#ifdef HAVE_OPENCV_GPUARITHM
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(SURF_HessianThreshold, double)
+    IMPLEMENT_PARAM_CLASS(SURF_Octaves, int)
+    IMPLEMENT_PARAM_CLASS(SURF_OctaveLayers, int)
+    IMPLEMENT_PARAM_CLASS(SURF_Extended, bool)
+    IMPLEMENT_PARAM_CLASS(SURF_Upright, bool)
+}
+
+PARAM_TEST_CASE(SURF, SURF_HessianThreshold, SURF_Octaves, SURF_OctaveLayers, SURF_Extended, SURF_Upright)
+{
+    double hessianThreshold;
+    int nOctaves;
+    int nOctaveLayers;
+    bool extended;
+    bool upright;
+
+    virtual void SetUp()
+    {
+        hessianThreshold = GET_PARAM(0);
+        nOctaves = GET_PARAM(1);
+        nOctaveLayers = GET_PARAM(2);
+        extended = GET_PARAM(3);
+        upright = GET_PARAM(4);
+    }
+};
+
+GPU_TEST_P(SURF, Detector)
+{
+    cv::Mat image = readImage("../gpu/features2d/aloe.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::gpu::SURF_GPU surf;
+    surf.hessianThreshold = hessianThreshold;
+    surf.nOctaves = nOctaves;
+    surf.nOctaveLayers = nOctaveLayers;
+    surf.extended = extended;
+    surf.upright = upright;
+    surf.keypointsRatio = 0.05f;
+
+    std::vector<cv::KeyPoint> keypoints;
+    surf(loadMat(image), cv::gpu::GpuMat(), keypoints);
+
+    cv::SURF surf_gold;
+    surf_gold.hessianThreshold = hessianThreshold;
+    surf_gold.nOctaves = nOctaves;
+    surf_gold.nOctaveLayers = nOctaveLayers;
+    surf_gold.extended = extended;
+    surf_gold.upright = upright;
+
+    std::vector<cv::KeyPoint> keypoints_gold;
+    surf_gold(image, cv::noArray(), keypoints_gold);
+
+    ASSERT_EQ(keypoints_gold.size(), keypoints.size());
+    int matchedCount = getMatchedPointsCount(keypoints_gold, keypoints);
+    double matchedRatio = static_cast<double>(matchedCount) / keypoints_gold.size();
+
+    EXPECT_GT(matchedRatio, 0.95);
+}
+
+GPU_TEST_P(SURF, Detector_Masked)
+{
+    cv::Mat image = readImage("../gpu/features2d/aloe.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::Mat mask(image.size(), CV_8UC1, cv::Scalar::all(1));
+    mask(cv::Range(0, image.rows / 2), cv::Range(0, image.cols / 2)).setTo(cv::Scalar::all(0));
+
+    cv::gpu::SURF_GPU surf;
+    surf.hessianThreshold = hessianThreshold;
+    surf.nOctaves = nOctaves;
+    surf.nOctaveLayers = nOctaveLayers;
+    surf.extended = extended;
+    surf.upright = upright;
+    surf.keypointsRatio = 0.05f;
+
+    std::vector<cv::KeyPoint> keypoints;
+    surf(loadMat(image), loadMat(mask), keypoints);
+
+    cv::SURF surf_gold;
+    surf_gold.hessianThreshold = hessianThreshold;
+    surf_gold.nOctaves = nOctaves;
+    surf_gold.nOctaveLayers = nOctaveLayers;
+    surf_gold.extended = extended;
+    surf_gold.upright = upright;
+
+    std::vector<cv::KeyPoint> keypoints_gold;
+    surf_gold(image, mask, keypoints_gold);
+
+    ASSERT_EQ(keypoints_gold.size(), keypoints.size());
+    int matchedCount = getMatchedPointsCount(keypoints_gold, keypoints);
+    double matchedRatio = static_cast<double>(matchedCount) / keypoints_gold.size();
+
+    EXPECT_GT(matchedRatio, 0.95);
+}
+
+GPU_TEST_P(SURF, Descriptor)
+{
+    cv::Mat image = readImage("../gpu/features2d/aloe.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::gpu::SURF_GPU surf;
+    surf.hessianThreshold = hessianThreshold;
+    surf.nOctaves = nOctaves;
+    surf.nOctaveLayers = nOctaveLayers;
+    surf.extended = extended;
+    surf.upright = upright;
+    surf.keypointsRatio = 0.05f;
+
+    cv::SURF surf_gold;
+    surf_gold.hessianThreshold = hessianThreshold;
+    surf_gold.nOctaves = nOctaves;
+    surf_gold.nOctaveLayers = nOctaveLayers;
+    surf_gold.extended = extended;
+    surf_gold.upright = upright;
+
+    std::vector<cv::KeyPoint> keypoints;
+    surf_gold(image, cv::noArray(), keypoints);
+
+    cv::gpu::GpuMat descriptors;
+    surf(loadMat(image), cv::gpu::GpuMat(), keypoints, descriptors, true);
+
+    cv::Mat descriptors_gold;
+    surf_gold(image, cv::noArray(), keypoints, descriptors_gold, true);
+
+    cv::BFMatcher matcher(cv::NORM_L2);
+    std::vector<cv::DMatch> matches;
+    matcher.match(descriptors_gold, cv::Mat(descriptors), matches);
+
+    int matchedCount = getMatchedPointsCount(keypoints, keypoints, matches);
+    double matchedRatio = static_cast<double>(matchedCount) / keypoints.size();
+
+    EXPECT_GT(matchedRatio, 0.6);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Features2D, SURF, testing::Combine(
+    testing::Values(SURF_HessianThreshold(100.0), SURF_HessianThreshold(500.0), SURF_HessianThreshold(1000.0)),
+    testing::Values(SURF_Octaves(3), SURF_Octaves(4)),
+    testing::Values(SURF_OctaveLayers(2), SURF_OctaveLayers(3)),
+    testing::Values(SURF_Extended(false), SURF_Extended(true)),
+    testing::Values(SURF_Upright(false), SURF_Upright(true))));
+
+#endif // HAVE_OPENCV_GPUARITHM
+
+#endif // HAVE_CUDA
diff --git a/modules/nonfree/test/test_keypoints.cpp b/modules/nonfree/test/test_keypoints.cpp
index 3984f19d89..b046d7534c 100644
--- a/modules/nonfree/test/test_keypoints.cpp
+++ b/modules/nonfree/test/test_keypoints.cpp
@@ -62,7 +62,7 @@ protected:
     virtual void run(int)
     {
         cv::initModule_features2d();
-        CV_Assert(!detector.empty());
+        CV_Assert(detector);
         string imgFilename = string(ts->get_data_path()) + FEATURES2D_DIR + "/" + IMAGE_FILENAME;
 
         // Read the test image.
diff --git a/modules/nonfree/test/test_precomp.cpp b/modules/nonfree/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/nonfree/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/nonfree/test/test_precomp.hpp b/modules/nonfree/test/test_precomp.hpp
index daa79c90d3..965e651e56 100644
--- a/modules/nonfree/test/test_precomp.hpp
+++ b/modules/nonfree/test/test_precomp.hpp
@@ -9,10 +9,24 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_test.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/highgui.hpp"
 #include "opencv2/nonfree.hpp"
-#include <iostream>
+
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+#include "cvconfig.h"
+
+#ifdef HAVE_OPENCV_OCL
+#  include "opencv2/nonfree/ocl.hpp"
+#endif
+
+#ifdef HAVE_CUDA
+#  include "opencv2/nonfree/gpu.hpp"
+#endif
 
 #endif
diff --git a/modules/nonfree/test/test_rotation_and_scale_invariance.cpp b/modules/nonfree/test/test_rotation_and_scale_invariance.cpp
index 9ca0225c2d..47efc60eb8 100644
--- a/modules/nonfree/test/test_rotation_and_scale_invariance.cpp
+++ b/modules/nonfree/test/test_rotation_and_scale_invariance.cpp
@@ -210,7 +210,7 @@ public:
         minKeyPointMatchesRatio(_minKeyPointMatchesRatio),
         minAngleInliersRatio(_minAngleInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
+        CV_Assert(featureDetector);
     }
 
 protected:
@@ -232,7 +232,7 @@ protected:
         featureDetector->detect(image0, keypoints0);
         removeVerySmallKeypoints(keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
         const int maxAngle = 360, angleStep = 15;
         for(int angle = 0; angle < maxAngle; angle += angleStep)
@@ -262,7 +262,7 @@ protected:
                 float angle0 = keypoints0[matches[m].queryIdx].angle;
                 float angle1 = keypoints1[matches[m].trainIdx].angle;
                 if(angle0 == -1 || angle1 == -1)
-                    CV_Error(CV_StsBadArg, "Given FeatureDetector is not rotation invariant, it can not be tested here.\n");
+                    CV_Error(Error::StsBadArg, "Given FeatureDetector is not rotation invariant, it can not be tested here.\n");
                 CV_Assert(angle0 >= 0.f && angle0 < 360.f);
                 CV_Assert(angle1 >= 0.f && angle1 < 360.f);
 
@@ -323,8 +323,8 @@ public:
         normType(_normType),
         minDescInliersRatio(_minDescInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
-        CV_Assert(!descriptorExtractor.empty());
+        CV_Assert(featureDetector);
+        CV_Assert(descriptorExtractor);
     }
 
 protected:
@@ -347,7 +347,7 @@ protected:
         featureDetector->detect(image0, keypoints0);
         removeVerySmallKeypoints(keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
         descriptorExtractor->compute(image0, keypoints0, descriptors0);
 
         BFMatcher bfmatcher(normType);
@@ -410,7 +410,7 @@ public:
         minKeyPointMatchesRatio(_minKeyPointMatchesRatio),
         minScaleInliersRatio(_minScaleInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
+        CV_Assert(featureDetector);
     }
 
 protected:
@@ -432,7 +432,7 @@ protected:
         featureDetector->detect(image0, keypoints0);
         removeVerySmallKeypoints(keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
         for(int scaleIdx = 1; scaleIdx <= 3; scaleIdx++)
         {
@@ -444,7 +444,7 @@ protected:
             featureDetector->detect(image1, keypoints1);
             removeVerySmallKeypoints(keypoints1);
             if(keypoints1.size() < 15)
-                CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+                CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
 
             if(keypoints1.size() > keypoints0.size())
             {
@@ -530,8 +530,8 @@ public:
         normType(_normType),
         minDescInliersRatio(_minDescInliersRatio)
     {
-        CV_Assert(!featureDetector.empty());
-        CV_Assert(!descriptorExtractor.empty());
+        CV_Assert(featureDetector);
+        CV_Assert(descriptorExtractor);
     }
 
 protected:
@@ -553,7 +553,7 @@ protected:
         featureDetector->detect(image0, keypoints0);
         removeVerySmallKeypoints(keypoints0);
         if(keypoints0.size() < 15)
-            CV_Error(CV_StsAssert, "Detector gives too few points in a test image\n");
+            CV_Error(Error::StsAssert, "Detector gives too few points in a test image\n");
         Mat descriptors0;
         descriptorExtractor->compute(image0, keypoints0, descriptors0);
 
diff --git a/modules/nonfree/test/test_surf.ocl.cpp b/modules/nonfree/test/test_surf.ocl.cpp
new file mode 100644
index 0000000000..d6a877bc80
--- /dev/null
+++ b/modules/nonfree/test/test_surf.ocl.cpp
@@ -0,0 +1,215 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_OPENCV_OCL
+
+using namespace std;
+using std::tr1::get;
+
+static bool keyPointsEquals(const cv::KeyPoint& p1, const cv::KeyPoint& p2)
+{
+    const double maxPtDif = 0.1;
+    const double maxSizeDif = 0.1;
+    const double maxAngleDif = 0.1;
+    const double maxResponseDif = 0.01;
+
+    double dist = cv::norm(p1.pt - p2.pt);
+
+    if (dist < maxPtDif &&
+        fabs(p1.size - p2.size) < maxSizeDif &&
+        abs(p1.angle - p2.angle) < maxAngleDif &&
+        abs(p1.response - p2.response) < maxResponseDif &&
+        p1.octave == p2.octave &&
+        p1.class_id == p2.class_id)
+    {
+        return true;
+    }
+
+    return false;
+}
+
+static int getMatchedPointsCount(std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual)
+{
+    std::sort(actual.begin(), actual.end(), perf::comparators::KeypointGreater());
+    std::sort(gold.begin(), gold.end(), perf::comparators::KeypointGreater());
+
+    int validCount = 0;
+
+    for (size_t i = 0; i < gold.size(); ++i)
+    {
+        const cv::KeyPoint& p1 = gold[i];
+        const cv::KeyPoint& p2 = actual[i];
+
+        if (keyPointsEquals(p1, p2))
+            ++validCount;
+    }
+
+    return validCount;
+}
+
+static int getMatchedPointsCount(const std::vector<cv::KeyPoint>& keypoints1, const std::vector<cv::KeyPoint>& keypoints2, const std::vector<cv::DMatch>& matches)
+{
+    int validCount = 0;
+
+    for (size_t i = 0; i < matches.size(); ++i)
+    {
+        const cv::DMatch& m = matches[i];
+
+        const cv::KeyPoint& p1 = keypoints1[m.queryIdx];
+        const cv::KeyPoint& p2 = keypoints2[m.trainIdx];
+
+        if (keyPointsEquals(p1, p2))
+            ++validCount;
+    }
+
+    return validCount;
+}
+
+IMPLEMENT_PARAM_CLASS(HessianThreshold, double)
+IMPLEMENT_PARAM_CLASS(Octaves, int)
+IMPLEMENT_PARAM_CLASS(OctaveLayers, int)
+IMPLEMENT_PARAM_CLASS(Extended, bool)
+IMPLEMENT_PARAM_CLASS(Upright, bool)
+
+PARAM_TEST_CASE(SURF, HessianThreshold, Octaves, OctaveLayers, Extended, Upright)
+{
+    double hessianThreshold;
+    int nOctaves;
+    int nOctaveLayers;
+    bool extended;
+    bool upright;
+
+    virtual void SetUp()
+    {
+        hessianThreshold = get<0>(GetParam());
+        nOctaves = get<1>(GetParam());
+        nOctaveLayers = get<2>(GetParam());
+        extended = get<3>(GetParam());
+        upright = get<4>(GetParam());
+    }
+};
+
+TEST_P(SURF, DISABLED_Detector)
+{
+    cv::Mat image  = cv::imread(string(cvtest::TS::ptr()->get_data_path()) + "shared/fruits.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::ocl::SURF_OCL surf;
+    surf.hessianThreshold = static_cast<float>(hessianThreshold);
+    surf.nOctaves = nOctaves;
+    surf.nOctaveLayers = nOctaveLayers;
+    surf.extended = extended;
+    surf.upright = upright;
+    surf.keypointsRatio = 0.05f;
+
+    std::vector<cv::KeyPoint> keypoints;
+    surf(cv::ocl::oclMat(image), cv::ocl::oclMat(), keypoints);
+
+    cv::SURF surf_gold;
+    surf_gold.hessianThreshold = hessianThreshold;
+    surf_gold.nOctaves = nOctaves;
+    surf_gold.nOctaveLayers = nOctaveLayers;
+    surf_gold.extended = extended;
+    surf_gold.upright = upright;
+
+    std::vector<cv::KeyPoint> keypoints_gold;
+    surf_gold(image, cv::noArray(), keypoints_gold);
+
+    ASSERT_EQ(keypoints_gold.size(), keypoints.size());
+    int matchedCount = getMatchedPointsCount(keypoints_gold, keypoints);
+    double matchedRatio = static_cast<double>(matchedCount) / keypoints_gold.size();
+
+    EXPECT_GT(matchedRatio, 0.99);
+}
+
+TEST_P(SURF, DISABLED_Descriptor)
+{
+    cv::Mat image  = cv::imread(string(cvtest::TS::ptr()->get_data_path()) + "shared/fruits.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(image.empty());
+
+    cv::ocl::SURF_OCL surf;
+    surf.hessianThreshold = static_cast<float>(hessianThreshold);
+    surf.nOctaves = nOctaves;
+    surf.nOctaveLayers = nOctaveLayers;
+    surf.extended = extended;
+    surf.upright = upright;
+    surf.keypointsRatio = 0.05f;
+
+    cv::SURF surf_gold;
+    surf_gold.hessianThreshold = hessianThreshold;
+    surf_gold.nOctaves = nOctaves;
+    surf_gold.nOctaveLayers = nOctaveLayers;
+    surf_gold.extended = extended;
+    surf_gold.upright = upright;
+
+    std::vector<cv::KeyPoint> keypoints;
+    surf_gold(image, cv::noArray(), keypoints);
+
+    cv::ocl::oclMat descriptors;
+    surf(cv::ocl::oclMat(image), cv::ocl::oclMat(), keypoints, descriptors, true);
+
+    cv::Mat descriptors_gold;
+    surf_gold(image, cv::noArray(), keypoints, descriptors_gold, true);
+
+    cv::BFMatcher matcher(cv::NORM_L2);
+    std::vector<cv::DMatch> matches;
+    matcher.match(descriptors_gold, cv::Mat(descriptors), matches);
+
+    int matchedCount = getMatchedPointsCount(keypoints, keypoints, matches);
+    double matchedRatio = static_cast<double>(matchedCount) / keypoints.size();
+
+    EXPECT_GT(matchedRatio, 0.35);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Features2D, SURF, testing::Combine(
+    testing::Values(HessianThreshold(500.0), HessianThreshold(1000.0)),
+    testing::Values(Octaves(3), Octaves(4)),
+    testing::Values(OctaveLayers(2), OctaveLayers(3)),
+    testing::Values(Extended(false), Extended(true)),
+    testing::Values(Upright(false), Upright(true))));
+
+#endif // HAVE_OPENCV_OCL
diff --git a/modules/objdetect/CMakeLists.txt b/modules/objdetect/CMakeLists.txt
index 78aa4aa234..5739ee9e38 100644
--- a/modules/objdetect/CMakeLists.txt
+++ b/modules/objdetect/CMakeLists.txt
@@ -1,2 +1,2 @@
 set(the_description "Object Detection")
-ocv_define_module(objdetect opencv_core opencv_imgproc OPTIONAL opencv_highgui)
+ocv_define_module(objdetect opencv_core opencv_imgproc opencv_ml OPTIONAL opencv_highgui)
diff --git a/modules/objdetect/doc/cascade_classification.rst b/modules/objdetect/doc/cascade_classification.rst
index 60b9a26966..453f05285b 100644
--- a/modules/objdetect/doc/cascade_classification.rst
+++ b/modules/objdetect/doc/cascade_classification.rst
@@ -146,7 +146,7 @@ CascadeClassifier::CascadeClassifier
 ----------------------------------------
 Loads a classifier from a file.
 
-.. ocv:function:: CascadeClassifier::CascadeClassifier(const string& filename)
+.. ocv:function:: CascadeClassifier::CascadeClassifier(const String& filename)
 
 .. ocv:pyfunction:: cv2.CascadeClassifier([filename]) -> <CascadeClassifier object>
 
@@ -167,7 +167,7 @@ CascadeClassifier::load
 ---------------------------
 Loads a classifier from a file.
 
-.. ocv:function:: bool CascadeClassifier::load(const string& filename)
+.. ocv:function:: bool CascadeClassifier::load(const String& filename)
 
 .. ocv:pyfunction:: cv2.CascadeClassifier.load(filename) -> retval
 
@@ -189,20 +189,21 @@ CascadeClassifier::detectMultiScale
 Detects objects of different sizes in the input image. The detected objects are returned as a list of rectangles.
 
 .. ocv:function:: void CascadeClassifier::detectMultiScale( const Mat& image, vector<Rect>& objects, double scaleFactor=1.1, int minNeighbors=3, int flags=0, Size minSize=Size(), Size maxSize=Size())
+.. ocv:function:: void CascadeClassifier::detectMultiScale( const Mat& image, vector<Rect>& objects, vector<int>& numDetections, double scaleFactor=1.1, int minNeighbors=3, int flags=0, Size minSize=Size(), Size maxSize=Size())
 
 .. ocv:pyfunction:: cv2.CascadeClassifier.detectMultiScale(image[, scaleFactor[, minNeighbors[, flags[, minSize[, maxSize]]]]]) -> objects
-.. ocv:pyfunction:: cv2.CascadeClassifier.detectMultiScale(image, rejectLevels, levelWeights[, scaleFactor[, minNeighbors[, flags[, minSize[, maxSize[, outputRejectLevels]]]]]]) -> objects
+.. ocv:pyfunction:: cv2.CascadeClassifier.detectMultiScale(image[, scaleFactor[, minNeighbors[, flags[, minSize[, maxSize[, outputRejectLevels]]]]]]) -> objects, rejectLevels, levelWeights
 
 .. ocv:cfunction:: CvSeq* cvHaarDetectObjects( const CvArr* image, CvHaarClassifierCascade* cascade, CvMemStorage* storage, double scale_factor=1.1, int min_neighbors=3, int flags=0, CvSize min_size=cvSize(0,0), CvSize max_size=cvSize(0,0) )
 
-.. ocv:pyoldfunction:: cv.HaarDetectObjects(image, cascade, storage, scale_factor=1.1, min_neighbors=3, flags=0, min_size=(0, 0)) -> detectedObjects
-
     :param cascade: Haar classifier cascade (OpenCV 1.x API only). It can be loaded from XML or YAML file using :ocv:cfunc:`Load`. When the cascade is not needed anymore, release it using ``cvReleaseHaarClassifierCascade(&cascade)``.
 
     :param image: Matrix of the type   ``CV_8U``  containing an image where objects are detected.
 
     :param objects: Vector of rectangles where each rectangle contains the detected object, the rectangles may be partially outside the original image.
 
+    :param numDetections: Vector of detection numbers for the corresponding objects. An object's number of detections is the number of neighboring positively classified rectangles that were joined together to form the object.
+
     :param scaleFactor: Parameter specifying how much the image size is reduced at each image scale.
 
     :param minNeighbors: Parameter specifying how many neighbors each candidate rectangle should have to retain it.
@@ -215,6 +216,10 @@ Detects objects of different sizes in the input image. The detected objects are
 
 The function is parallelized with the TBB library.
 
+.. note::
+
+   * (Python) A face detection example using cascade classifiers can be found at opencv_source_code/samples/python2/facedetect.py
+
 
 CascadeClassifier::setImage
 -------------------------------
diff --git a/modules/objdetect/doc/latent_svm.rst b/modules/objdetect/doc/latent_svm.rst
index 86fb0edc1e..4b4ff117fa 100644
--- a/modules/objdetect/doc/latent_svm.rst
+++ b/modules/objdetect/doc/latent_svm.rst
@@ -204,7 +204,7 @@ Two types of constructors.
 
 .. ocv:function:: LatentSvmDetector::LatentSvmDetector()
 
-.. ocv:function:: LatentSvmDetector::LatentSvmDetector(const vector<string>& filenames, const vector<string>& classNames=vector<string>())
+.. ocv:function:: LatentSvmDetector::LatentSvmDetector(const vector<String>& filenames, const vector<String>& classNames=vector<String>())
 
 
 
@@ -228,7 +228,7 @@ LatentSvmDetector::load
 -----------------------
 Load the trained models from given ``.xml`` files and return ``true`` if at least one model was loaded.
 
-.. ocv:function:: bool LatentSvmDetector::load( const vector<string>& filenames, const vector<string>& classNames=vector<string>() )
+.. ocv:function:: bool LatentSvmDetector::load( const vector<String>& filenames, const vector<String>& classNames=vector<String>() )
 
     :param filenames: A set of filenames storing the trained detectors (models). Each file contains one model. See examples of such files here /opencv_extra/testdata/cv/latentsvmdetector/models_VOC2007/.
 
@@ -250,7 +250,7 @@ LatentSvmDetector::getClassNames
 --------------------------------
 Return the class (model) names that were passed in constructor or method ``load`` or extracted from models filenames in those methods.
 
-.. ocv:function:: const vector<string>& LatentSvmDetector::getClassNames() const
+.. ocv:function:: const vector<String>& LatentSvmDetector::getClassNames() const
 
 LatentSvmDetector::getClassCount
 --------------------------------
@@ -260,5 +260,3 @@ Return a count of loaded models (classes).
 
 
 .. [Felzenszwalb2010] Felzenszwalb, P. F. and Girshick, R. B. and McAllester, D. and Ramanan, D. *Object Detection with Discriminatively Trained Part Based Models*. PAMI, vol. 32, no. 9, pp. 1627-1645, September 2010
-
-
diff --git a/modules/objdetect/include/opencv2/objdetect.hpp b/modules/objdetect/include/opencv2/objdetect.hpp
index a8c2955e15..f1b371610d 100644
--- a/modules/objdetect/include/opencv2/objdetect.hpp
+++ b/modules/objdetect/include/opencv2/objdetect.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -45,243 +46,8 @@
 
 #include "opencv2/core.hpp"
 
-#ifdef __cplusplus
-#include <map>
-#include <deque>
-
-extern "C" {
-#endif
-
-/****************************************************************************************\
-*                         Haar-like Object Detection functions                           *
-\****************************************************************************************/
-
-#define CV_HAAR_MAGIC_VAL    0x42500000
-#define CV_TYPE_NAME_HAAR    "opencv-haar-classifier"
-
-#define CV_IS_HAAR_CLASSIFIER( haar )                                                    \
-    ((haar) != NULL &&                                                                   \
-    (((const CvHaarClassifierCascade*)(haar))->flags & CV_MAGIC_MASK)==CV_HAAR_MAGIC_VAL)
-
-#define CV_HAAR_FEATURE_MAX  3
-
-typedef struct CvHaarFeature
-{
-    int tilted;
-    struct
-    {
-        CvRect r;
-        float weight;
-    } rect[CV_HAAR_FEATURE_MAX];
-} CvHaarFeature;
-
-typedef struct CvHaarClassifier
-{
-    int count;
-    CvHaarFeature* haar_feature;
-    float* threshold;
-    int* left;
-    int* right;
-    float* alpha;
-} CvHaarClassifier;
-
-typedef struct CvHaarStageClassifier
-{
-    int  count;
-    float threshold;
-    CvHaarClassifier* classifier;
-
-    int next;
-    int child;
-    int parent;
-} CvHaarStageClassifier;
-
-typedef struct CvHidHaarClassifierCascade CvHidHaarClassifierCascade;
-
-typedef struct CvHaarClassifierCascade
-{
-    int  flags;
-    int  count;
-    CvSize orig_window_size;
-    CvSize real_window_size;
-    double scale;
-    CvHaarStageClassifier* stage_classifier;
-    CvHidHaarClassifierCascade* hid_cascade;
-} CvHaarClassifierCascade;
-
-typedef struct CvAvgComp
-{
-    CvRect rect;
-    int neighbors;
-} CvAvgComp;
-
-/* Loads haar classifier cascade from a directory.
-   It is obsolete: convert your cascade to xml and use cvLoad instead */
-CVAPI(CvHaarClassifierCascade*) cvLoadHaarClassifierCascade(
-                    const char* directory, CvSize orig_window_size);
-
-CVAPI(void) cvReleaseHaarClassifierCascade( CvHaarClassifierCascade** cascade );
-
-#define CV_HAAR_DO_CANNY_PRUNING    1
-#define CV_HAAR_SCALE_IMAGE         2
-#define CV_HAAR_FIND_BIGGEST_OBJECT 4
-#define CV_HAAR_DO_ROUGH_SEARCH     8
-
-//CVAPI(CvSeq*) cvHaarDetectObjectsForROC( const CvArr* image,
-//                     CvHaarClassifierCascade* cascade, CvMemStorage* storage,
-//                     CvSeq** rejectLevels, CvSeq** levelWeightds,
-//                     double scale_factor CV_DEFAULT(1.1),
-//                     int min_neighbors CV_DEFAULT(3), int flags CV_DEFAULT(0),
-//                     CvSize min_size CV_DEFAULT(cvSize(0,0)), CvSize max_size CV_DEFAULT(cvSize(0,0)),
-//                     bool outputRejectLevels = false );
-
-
-CVAPI(CvSeq*) cvHaarDetectObjects( const CvArr* image,
-                     CvHaarClassifierCascade* cascade, CvMemStorage* storage,
-                     double scale_factor CV_DEFAULT(1.1),
-                     int min_neighbors CV_DEFAULT(3), int flags CV_DEFAULT(0),
-                     CvSize min_size CV_DEFAULT(cvSize(0,0)), CvSize max_size CV_DEFAULT(cvSize(0,0)));
-
-/* sets images for haar classifier cascade */
-CVAPI(void) cvSetImagesForHaarClassifierCascade( CvHaarClassifierCascade* cascade,
-                                                const CvArr* sum, const CvArr* sqsum,
-                                                const CvArr* tilted_sum, double scale );
-
-/* runs the cascade on the specified window */
-CVAPI(int) cvRunHaarClassifierCascade( const CvHaarClassifierCascade* cascade,
-                                       CvPoint pt, int start_stage CV_DEFAULT(0));
-
-
-/****************************************************************************************\
-*                         Latent SVM Object Detection functions                          *
-\****************************************************************************************/
-
-// DataType: STRUCT position
-// Structure describes the position of the filter in the feature pyramid
-// l - level in the feature pyramid
-// (x, y) - coordinate in level l
-typedef struct CvLSVMFilterPosition
-{
-    int x;
-    int y;
-    int l;
-} CvLSVMFilterPosition;
-
-// DataType: STRUCT filterObject
-// Description of the filter, which corresponds to the part of the object
-// V               - ideal (penalty = 0) position of the partial filter
-//                   from the root filter position (V_i in the paper)
-// penaltyFunction - vector describes penalty function (d_i in the paper)
-//                   pf[0] * x + pf[1] * y + pf[2] * x^2 + pf[3] * y^2
-// FILTER DESCRIPTION
-//   Rectangular map (sizeX x sizeY),
-//   every cell stores feature vector (dimension = p)
-// H               - matrix of feature vectors
-//                   to set and get feature vectors (i,j)
-//                   used formula H[(j * sizeX + i) * p + k], where
-//                   k - component of feature vector in cell (i, j)
-// END OF FILTER DESCRIPTION
-typedef struct CvLSVMFilterObject{
-    CvLSVMFilterPosition V;
-    float fineFunction[4];
-    int sizeX;
-    int sizeY;
-    int numFeatures;
-    float *H;
-} CvLSVMFilterObject;
-
-// data type: STRUCT CvLatentSvmDetector
-// structure contains internal representation of trained Latent SVM detector
-// num_filters			- total number of filters (root plus part) in model
-// num_components		- number of components in model
-// num_part_filters		- array containing number of part filters for each component
-// filters				- root and part filters for all model components
-// b					- biases for all model components
-// score_threshold		- confidence level threshold
-typedef struct CvLatentSvmDetector
-{
-    int num_filters;
-    int num_components;
-    int* num_part_filters;
-    CvLSVMFilterObject** filters;
-    float* b;
-    float score_threshold;
-}
-CvLatentSvmDetector;
-
-// data type: STRUCT CvObjectDetection
-// structure contains the bounding box and confidence level for detected object
-// rect					- bounding box for a detected object
-// score				- confidence level
-typedef struct CvObjectDetection
-{
-    CvRect rect;
-    float score;
-} CvObjectDetection;
-
-//////////////// Object Detection using Latent SVM //////////////
-
-
-/*
-// load trained detector from a file
-//
-// API
-// CvLatentSvmDetector* cvLoadLatentSvmDetector(const char* filename);
-// INPUT
-// filename				- path to the file containing the parameters of
-                        - trained Latent SVM detector
-// OUTPUT
-// trained Latent SVM detector in internal representation
-*/
-CVAPI(CvLatentSvmDetector*) cvLoadLatentSvmDetector(const char* filename);
-
-/*
-// release memory allocated for CvLatentSvmDetector structure
-//
-// API
-// void cvReleaseLatentSvmDetector(CvLatentSvmDetector** detector);
-// INPUT
-// detector				- CvLatentSvmDetector structure to be released
-// OUTPUT
-*/
-CVAPI(void) cvReleaseLatentSvmDetector(CvLatentSvmDetector** detector);
-
-/*
-// find rectangular regions in the given image that are likely
-// to contain objects and corresponding confidence levels
-//
-// API
-// CvSeq* cvLatentSvmDetectObjects(const IplImage* image,
-//									CvLatentSvmDetector* detector,
-//									CvMemStorage* storage,
-//									float overlap_threshold = 0.5f,
-//                                  int numThreads = -1);
-// INPUT
-// image				- image to detect objects in
-// detector				- Latent SVM detector in internal representation
-// storage				- memory storage to store the resultant sequence
-//							of the object candidate rectangles
-// overlap_threshold	- threshold for the non-maximum suppression algorithm
-                           = 0.5f [here will be the reference to original paper]
-// OUTPUT
-// sequence of detected objects (bounding boxes and confidence levels stored in CvObjectDetection structures)
-*/
-CVAPI(CvSeq*) cvLatentSvmDetectObjects(IplImage* image,
-                                CvLatentSvmDetector* detector,
-                                CvMemStorage* storage,
-                                float overlap_threshold CV_DEFAULT(0.5f),
-                                int numThreads CV_DEFAULT(-1));
-
-#ifdef __cplusplus
-}
-
-CV_EXPORTS CvSeq* cvHaarDetectObjectsForROC( const CvArr* image,
-                     CvHaarClassifierCascade* cascade, CvMemStorage* storage,
-                     std::vector<int>& rejectLevels, std::vector<double>& levelWeightds,
-                     double scale_factor CV_DEFAULT(1.1),
-                     int min_neighbors CV_DEFAULT(3), int flags CV_DEFAULT(0),
-                     CvSize min_size CV_DEFAULT(cvSize(0,0)), CvSize max_size CV_DEFAULT(cvSize(0,0)),
-                     bool outputRejectLevels = false );
+typedef struct CvLatentSvmDetector CvLatentSvmDetector;
+typedef struct CvHaarClassifierCascade CvHaarClassifierCascade;
 
 namespace cv
 {
@@ -300,46 +66,66 @@ public:
     struct CV_EXPORTS ObjectDetection
     {
         ObjectDetection();
-        ObjectDetection( const Rect& rect, float score, int classID=-1 );
+        ObjectDetection( const Rect& rect, float score, int classID = -1 );
         Rect rect;
         float score;
         int classID;
     };
 
     LatentSvmDetector();
-    LatentSvmDetector( const std::vector<std::string>& filenames, const std::vector<std::string>& classNames=std::vector<std::string>() );
+    LatentSvmDetector( const std::vector<String>& filenames, const std::vector<String>& classNames = std::vector<String>() );
     virtual ~LatentSvmDetector();
 
     virtual void clear();
     virtual bool empty() const;
-    bool load( const std::vector<std::string>& filenames, const std::vector<std::string>& classNames=std::vector<std::string>() );
+    bool load( const std::vector<String>& filenames, const std::vector<String>& classNames = std::vector<String>() );
 
     virtual void detect( const Mat& image,
                          std::vector<ObjectDetection>& objectDetections,
-                         float overlapThreshold=0.5f,
-                         int numThreads=-1 );
+                         float overlapThreshold = 0.5f,
+                         int numThreads = -1 );
 
-    const std::vector<std::string>& getClassNames() const;
+    const std::vector<String>& getClassNames() const;
     size_t getClassCount() const;
 
 private:
     std::vector<CvLatentSvmDetector*> detectors;
-    std::vector<std::string> classNames;
+    std::vector<String> classNames;
 };
 
-CV_EXPORTS void groupRectangles(CV_OUT CV_IN_OUT std::vector<Rect>& rectList, int groupThreshold, double eps=0.2);
-CV_EXPORTS_W void groupRectangles(CV_OUT CV_IN_OUT std::vector<Rect>& rectList, CV_OUT std::vector<int>& weights, int groupThreshold, double eps=0.2);
-CV_EXPORTS void groupRectangles( std::vector<Rect>& rectList, int groupThreshold, double eps, std::vector<int>* weights, std::vector<double>* levelWeights );
-CV_EXPORTS void groupRectangles(std::vector<Rect>& rectList, std::vector<int>& rejectLevels,
-                                std::vector<double>& levelWeights, int groupThreshold, double eps=0.2);
-CV_EXPORTS void groupRectangles_meanshift(std::vector<Rect>& rectList, std::vector<double>& foundWeights, std::vector<double>& foundScales,
-                                          double detectThreshold = 0.0, Size winDetSize = Size(64, 128));
+// class for grouping object candidates, detected by Cascade Classifier, HOG etc.
+// instance of the class is to be passed to cv::partition (see cxoperations.hpp)
+class CV_EXPORTS SimilarRects
+{
+public:
+    SimilarRects(double _eps) : eps(_eps) {}
+    inline bool operator()(const Rect& r1, const Rect& r2) const
+    {
+        double delta = eps*(std::min(r1.width, r2.width) + std::min(r1.height, r2.height))*0.5;
+        return std::abs(r1.x - r2.x) <= delta &&
+            std::abs(r1.y - r2.y) <= delta &&
+            std::abs(r1.x + r1.width - r2.x - r2.width) <= delta &&
+            std::abs(r1.y + r1.height - r2.y - r2.height) <= delta;
+    }
+    double eps;
+};
 
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps = 0.2);
+CV_EXPORTS_W void groupRectangles(CV_IN_OUT std::vector<Rect>& rectList, CV_OUT std::vector<int>& weights, int groupThreshold, double eps = 0.2);
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps, std::vector<int>* weights, std::vector<double>* levelWeights );
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, std::vector<int>& rejectLevels,
+                                  std::vector<double>& levelWeights, int groupThreshold, double eps = 0.2);
+CV_EXPORTS   void groupRectangles_meanshift(std::vector<Rect>& rectList, std::vector<double>& foundWeights, std::vector<double>& foundScales,
+                                          double detectThreshold = 0.0, Size winDetSize = Size(64, 128));
 
 class CV_EXPORTS FeatureEvaluator
 {
 public:
-    enum { HAAR = 0, LBP = 1, HOG = 2 };
+    enum { HAAR = 0,
+           LBP  = 1,
+           HOG  = 2
+         };
+
     virtual ~FeatureEvaluator();
 
     virtual bool read(const FileNode& node);
@@ -355,28 +141,34 @@ public:
     static Ptr<FeatureEvaluator> create(int type);
 };
 
-template<> CV_EXPORTS void Ptr<CvHaarClassifierCascade>::delete_obj();
+template<> CV_EXPORTS void DefaultDeleter<CvHaarClassifierCascade>::operator ()(CvHaarClassifierCascade* obj) const;
 
-enum
-{
-    CASCADE_DO_CANNY_PRUNING=1,
-    CASCADE_SCALE_IMAGE=2,
-    CASCADE_FIND_BIGGEST_OBJECT=4,
-    CASCADE_DO_ROUGH_SEARCH=8
-};
+enum { CASCADE_DO_CANNY_PRUNING    = 1,
+       CASCADE_SCALE_IMAGE         = 2,
+       CASCADE_FIND_BIGGEST_OBJECT = 4,
+       CASCADE_DO_ROUGH_SEARCH     = 8
+     };
 
 class CV_EXPORTS_W CascadeClassifier
 {
 public:
     CV_WRAP CascadeClassifier();
-    CV_WRAP CascadeClassifier( const std::string& filename );
+    CV_WRAP CascadeClassifier( const String& filename );
     virtual ~CascadeClassifier();
 
     CV_WRAP virtual bool empty() const;
-    CV_WRAP bool load( const std::string& filename );
+    CV_WRAP bool load( const String& filename );
     virtual bool read( const FileNode& node );
     CV_WRAP virtual void detectMultiScale( const Mat& image,
                                    CV_OUT std::vector<Rect>& objects,
+                                   double scaleFactor = 1.1,
+                                   int minNeighbors = 3, int flags = 0,
+                                   Size minSize = Size(),
+                                   Size maxSize = Size() );
+
+    CV_WRAP virtual void detectMultiScale( const Mat& image,
+                                   CV_OUT std::vector<Rect>& objects,
+                                   CV_OUT std::vector<int>& numDetections,
                                    double scaleFactor=1.1,
                                    int minNeighbors=3, int flags=0,
                                    Size minSize=Size(),
@@ -384,13 +176,13 @@ public:
 
     CV_WRAP virtual void detectMultiScale( const Mat& image,
                                    CV_OUT std::vector<Rect>& objects,
-                                   std::vector<int>& rejectLevels,
-                                   std::vector<double>& levelWeights,
-                                   double scaleFactor=1.1,
-                                   int minNeighbors=3, int flags=0,
-                                   Size minSize=Size(),
-                                   Size maxSize=Size(),
-                                   bool outputRejectLevels=false );
+                                   CV_OUT std::vector<int>& rejectLevels,
+                                   CV_OUT std::vector<double>& levelWeights,
+                                   double scaleFactor = 1.1,
+                                   int minNeighbors = 3, int flags = 0,
+                                   Size minSize = Size(),
+                                   Size maxSize = Size(),
+                                   bool outputRejectLevels = false );
 
 
     bool isOldFormatCascade() const;
@@ -399,17 +191,23 @@ public:
     bool setImage( const Mat& );
 
 protected:
-    //virtual bool detectSingleScale( const Mat& image, int stripCount, Size processingRectSize,
-    //                                int stripSize, int yStep, double factor, std::vector<Rect>& candidates );
-
     virtual bool detectSingleScale( const Mat& image, int stripCount, Size processingRectSize,
                                     int stripSize, int yStep, double factor, std::vector<Rect>& candidates,
-                                    std::vector<int>& rejectLevels, std::vector<double>& levelWeights, bool outputRejectLevels=false);
+                                    std::vector<int>& rejectLevels, std::vector<double>& levelWeights, bool outputRejectLevels = false );
+
+    virtual void detectMultiScaleNoGrouping( const Mat& image, std::vector<Rect>& candidates,
+                                             std::vector<int>& rejectLevels, std::vector<double>& levelWeights,
+                                             double scaleFactor, Size minObjectSize, Size maxObjectSize,
+                                             bool outputRejectLevels = false );
 
 protected:
-    enum { BOOST = 0 };
-    enum { DO_CANNY_PRUNING = 1, SCALE_IMAGE = 2,
-           FIND_BIGGEST_OBJECT = 4, DO_ROUGH_SEARCH = 8 };
+    enum { BOOST = 0
+         };
+    enum { DO_CANNY_PRUNING    = CASCADE_DO_CANNY_PRUNING,
+           SCALE_IMAGE         = CASCADE_SCALE_IMAGE,
+           FIND_BIGGEST_OBJECT = CASCADE_FIND_BIGGEST_OBJECT,
+           DO_ROUGH_SEARCH     = CASCADE_DO_ROUGH_SEARCH
+         };
 
     friend class CascadeClassifierInvoker;
 
@@ -504,8 +302,10 @@ struct DetectionROI
 struct CV_EXPORTS_W HOGDescriptor
 {
 public:
-    enum { L2Hys=0 };
-    enum { DEFAULT_NLEVELS=64 };
+    enum { L2Hys = 0
+         };
+    enum { DEFAULT_NLEVELS = 64
+         };
 
     CV_WRAP HOGDescriptor() : winSize(64,128), blockSize(16,16), blockStride(8,8),
         cellSize(8,8), nbins(9), derivAperture(1), winSigma(-1),
@@ -524,7 +324,7 @@ public:
     gammaCorrection(_gammaCorrection), nlevels(_nlevels)
     {}
 
-    CV_WRAP HOGDescriptor(const std::string& filename)
+    CV_WRAP HOGDescriptor(const String& filename)
     {
         load(filename);
     }
@@ -543,40 +343,40 @@ public:
     CV_WRAP virtual void setSVMDetector(InputArray _svmdetector);
 
     virtual bool read(FileNode& fn);
-    virtual void write(FileStorage& fs, const std::string& objname) const;
+    virtual void write(FileStorage& fs, const String& objname) const;
 
-    CV_WRAP virtual bool load(const std::string& filename, const std::string& objname=std::string());
-    CV_WRAP virtual void save(const std::string& filename, const std::string& objname=std::string()) const;
+    CV_WRAP virtual bool load(const String& filename, const String& objname = String());
+    CV_WRAP virtual void save(const String& filename, const String& objname = String()) const;
     virtual void copyTo(HOGDescriptor& c) const;
 
     CV_WRAP virtual void compute(const Mat& img,
                          CV_OUT std::vector<float>& descriptors,
-                         Size winStride=Size(), Size padding=Size(),
-                         const std::vector<Point>& locations=std::vector<Point>()) const;
+                         Size winStride = Size(), Size padding = Size(),
+                         const std::vector<Point>& locations = std::vector<Point>()) const;
     //with found weights output
     CV_WRAP virtual void detect(const Mat& img, CV_OUT std::vector<Point>& foundLocations,
                         CV_OUT std::vector<double>& weights,
-                        double hitThreshold=0, Size winStride=Size(),
-                        Size padding=Size(),
-                        const std::vector<Point>& searchLocations=std::vector<Point>()) const;
+                        double hitThreshold = 0, Size winStride = Size(),
+                        Size padding = Size(),
+                        const std::vector<Point>& searchLocations = std::vector<Point>()) const;
     //without found weights output
     virtual void detect(const Mat& img, CV_OUT std::vector<Point>& foundLocations,
-                        double hitThreshold=0, Size winStride=Size(),
-                        Size padding=Size(),
+                        double hitThreshold = 0, Size winStride = Size(),
+                        Size padding = Size(),
                         const std::vector<Point>& searchLocations=std::vector<Point>()) const;
     //with result weights output
     CV_WRAP virtual void detectMultiScale(const Mat& img, CV_OUT std::vector<Rect>& foundLocations,
-                                  CV_OUT std::vector<double>& foundWeights, double hitThreshold=0,
-                                  Size winStride=Size(), Size padding=Size(), double scale=1.05,
-                                  double finalThreshold=2.0,bool useMeanshiftGrouping = false) const;
+                                  CV_OUT std::vector<double>& foundWeights, double hitThreshold = 0,
+                                  Size winStride = Size(), Size padding = Size(), double scale = 1.05,
+                                  double finalThreshold = 2.0,bool useMeanshiftGrouping = false) const;
     //without found weights output
     virtual void detectMultiScale(const Mat& img, CV_OUT std::vector<Rect>& foundLocations,
-                                  double hitThreshold=0, Size winStride=Size(),
-                                  Size padding=Size(), double scale=1.05,
-                                  double finalThreshold=2.0, bool useMeanshiftGrouping = false) const;
+                                  double hitThreshold = 0, Size winStride = Size(),
+                                  Size padding = Size(), double scale = 1.05,
+                                  double finalThreshold = 2.0, bool useMeanshiftGrouping = false) const;
 
     CV_WRAP virtual void computeGradient(const Mat& img, CV_OUT Mat& grad, CV_OUT Mat& angleOfs,
-                                 Size paddingTL=Size(), Size paddingBR=Size()) const;
+                                 Size paddingTL = Size(), Size paddingBR = Size()) const;
 
     CV_WRAP static std::vector<float> getDefaultPeopleDetector();
     CV_WRAP static std::vector<float> getDaimlerPeopleDetector();
@@ -609,436 +409,22 @@ public:
                                                        int groupThreshold = 0) const;
 
    // read/parse Dalal's alt model file
-   void readALTModel(std::string modelfile);
+   void readALTModel(String modelfile);
+   void groupRectangles(std::vector<cv::Rect>& rectList, std::vector<double>& weights, int groupThreshold, double eps) const;
 };
 
 
 CV_EXPORTS_W void findDataMatrix(InputArray image,
-                                 CV_OUT std::vector<std::string>& codes,
-                                 OutputArray corners=noArray(),
-                                 OutputArrayOfArrays dmtx=noArray());
+                                 CV_OUT std::vector<String>& codes,
+                                 OutputArray corners = noArray(),
+                                 OutputArrayOfArrays dmtx = noArray());
+
 CV_EXPORTS_W void drawDataMatrixCodes(InputOutputArray image,
-                                      const std::vector<std::string>& codes,
+                                      const std::vector<String>& codes,
                                       InputArray corners);
 }
 
-/****************************************************************************************\
-*                                Datamatrix                                              *
-\****************************************************************************************/
-
-struct CV_EXPORTS CvDataMatrixCode {
-  char msg[4];
-  CvMat *original;
-  CvMat *corners;
-};
-
-CV_EXPORTS std::deque<CvDataMatrixCode> cvFindDataMatrix(CvMat *im);
-
-/****************************************************************************************\
-*                                 LINE-MOD                                               *
-\****************************************************************************************/
-
-namespace cv {
-namespace linemod {
-
-/// @todo Convert doxy comments to rst
-
-/**
- * \brief Discriminant feature described by its location and label.
- */
-struct CV_EXPORTS Feature
-{
-  int x; ///< x offset
-  int y; ///< y offset
-  int label; ///< Quantization
-
-  Feature() : x(0), y(0), label(0) {}
-  Feature(int x, int y, int label);
-
-  void read(const FileNode& fn);
-  void write(FileStorage& fs) const;
-};
-
-inline Feature::Feature(int _x, int _y, int _label) : x(_x), y(_y), label(_label) {}
-
-struct CV_EXPORTS Template
-{
-  int width;
-  int height;
-  int pyramid_level;
-  std::vector<Feature> features;
-
-  void read(const FileNode& fn);
-  void write(FileStorage& fs) const;
-};
-
-/**
- * \brief Represents a modality operating over an image pyramid.
- */
-class QuantizedPyramid
-{
-public:
-  // Virtual destructor
-  virtual ~QuantizedPyramid() {}
-
-  /**
-   * \brief Compute quantized image at current pyramid level for online detection.
-   *
-   * \param[out] dst The destination 8-bit image. For each pixel at most one bit is set,
-   *                 representing its classification.
-   */
-  virtual void quantize(Mat& dst) const =0;
-
-  /**
-   * \brief Extract most discriminant features at current pyramid level to form a new template.
-   *
-   * \param[out] templ The new template.
-   */
-  virtual bool extractTemplate(Template& templ) const =0;
-
-  /**
-   * \brief Go to the next pyramid level.
-   *
-   * \todo Allow pyramid scale factor other than 2
-   */
-  virtual void pyrDown() =0;
-
-protected:
-  /// Candidate feature with a score
-  struct Candidate
-  {
-    Candidate(int x, int y, int label, float score);
-
-    /// Sort candidates with high score to the front
-    bool operator<(const Candidate& rhs) const
-    {
-      return score > rhs.score;
-    }
-
-    Feature f;
-    float score;
-  };
-
-  /**
-   * \brief Choose candidate features so that they are not bunched together.
-   *
-   * \param[in]  candidates   Candidate features sorted by score.
-   * \param[out] features     Destination vector of selected features.
-   * \param[in]  num_features Number of candidates to select.
-   * \param[in]  distance     Hint for desired distance between features.
-   */
-  static void selectScatteredFeatures(const std::vector<Candidate>& candidates,
-                                      std::vector<Feature>& features,
-                                      size_t num_features, float distance);
-};
-
-inline QuantizedPyramid::Candidate::Candidate(int x, int y, int label, float _score) : f(x, y, label), score(_score) {}
-
-/**
- * \brief Interface for modalities that plug into the LINE template matching representation.
- *
- * \todo Max response, to allow optimization of summing (255/MAX) features as uint8
- */
-class CV_EXPORTS Modality
-{
-public:
-  // Virtual destructor
-  virtual ~Modality() {}
-
-  /**
-   * \brief Form a quantized image pyramid from a source image.
-   *
-   * \param[in] src  The source image. Type depends on the modality.
-   * \param[in] mask Optional mask. If not empty, unmasked pixels are set to zero
-   *                 in quantized image and cannot be extracted as features.
-   */
-  Ptr<QuantizedPyramid> process(const Mat& src,
-                    const Mat& mask = Mat()) const
-  {
-    return processImpl(src, mask);
-  }
-
-  virtual std::string name() const =0;
-
-  virtual void read(const FileNode& fn) =0;
-  virtual void write(FileStorage& fs) const =0;
-
-  /**
-   * \brief Create modality by name.
-   *
-   * The following modality types are supported:
-   * - "ColorGradient"
-   * - "DepthNormal"
-   */
-  static Ptr<Modality> create(const std::string& modality_type);
-
-  /**
-   * \brief Load a modality from file.
-   */
-  static Ptr<Modality> create(const FileNode& fn);
-
-protected:
-  // Indirection is because process() has a default parameter.
-  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
-                        const Mat& mask) const =0;
-};
-
-/**
- * \brief Modality that computes quantized gradient orientations from a color image.
- */
-class CV_EXPORTS ColorGradient : public Modality
-{
-public:
-  /**
-   * \brief Default constructor. Uses reasonable default parameter values.
-   */
-  ColorGradient();
-
-  /**
-   * \brief Constructor.
-   *
-   * \param weak_threshold   When quantizing, discard gradients with magnitude less than this.
-   * \param num_features     How many features a template must contain.
-   * \param strong_threshold Consider as candidate features only gradients whose norms are
-   *                         larger than this.
-   */
-  ColorGradient(float weak_threshold, size_t num_features, float strong_threshold);
-
-  virtual std::string name() const;
-
-  virtual void read(const FileNode& fn);
-  virtual void write(FileStorage& fs) const;
-
-  float weak_threshold;
-  size_t num_features;
-  float strong_threshold;
-
-protected:
-  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
-                        const Mat& mask) const;
-};
-
-/**
- * \brief Modality that computes quantized surface normals from a dense depth map.
- */
-class CV_EXPORTS DepthNormal : public Modality
-{
-public:
-  /**
-   * \brief Default constructor. Uses reasonable default parameter values.
-   */
-  DepthNormal();
-
-  /**
-   * \brief Constructor.
-   *
-   * \param distance_threshold   Ignore pixels beyond this distance.
-   * \param difference_threshold When computing normals, ignore contributions of pixels whose
-   *                             depth difference with the central pixel is above this threshold.
-   * \param num_features         How many features a template must contain.
-   * \param extract_threshold    Consider as candidate feature only if there are no differing
-   *                             orientations within a distance of extract_threshold.
-   */
-  DepthNormal(int distance_threshold, int difference_threshold, size_t num_features,
-              int extract_threshold);
-
-  virtual std::string name() const;
-
-  virtual void read(const FileNode& fn);
-  virtual void write(FileStorage& fs) const;
-
-  int distance_threshold;
-  int difference_threshold;
-  size_t num_features;
-  int extract_threshold;
-
-protected:
-  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
-                        const Mat& mask) const;
-};
-
-/**
- * \brief Debug function to colormap a quantized image for viewing.
- */
-void colormap(const Mat& quantized, Mat& dst);
-
-/**
- * \brief Represents a successful template match.
- */
-struct CV_EXPORTS Match
-{
-  Match()
-  {
-  }
-
-  Match(int x, int y, float similarity, const std::string& class_id, int template_id);
-
-  /// Sort matches with high similarity to the front
-  bool operator<(const Match& rhs) const
-  {
-    // Secondarily sort on template_id for the sake of duplicate removal
-    if (similarity != rhs.similarity)
-      return similarity > rhs.similarity;
-    else
-      return template_id < rhs.template_id;
-  }
-
-  bool operator==(const Match& rhs) const
-  {
-    return x == rhs.x && y == rhs.y && similarity == rhs.similarity && class_id == rhs.class_id;
-  }
-
-  int x;
-  int y;
-  float similarity;
-  std::string class_id;
-  int template_id;
-};
-
-inline  Match::Match(int _x, int _y, float _similarity, const std::string& _class_id, int _template_id)
-    : x(_x), y(_y), similarity(_similarity), class_id(_class_id), template_id(_template_id)
-  {
-  }
-
-/**
- * \brief Object detector using the LINE template matching algorithm with any set of
- * modalities.
- */
-class CV_EXPORTS Detector
-{
-public:
-  /**
-   * \brief Empty constructor, initialize with read().
-   */
-  Detector();
-
-  /**
-   * \brief Constructor.
-   *
-   * \param modalities       Modalities to use (color gradients, depth normals, ...).
-   * \param T_pyramid        Value of the sampling step T at each pyramid level. The
-   *                         number of pyramid levels is T_pyramid.size().
-   */
-  Detector(const std::vector< Ptr<Modality> >& modalities, const std::vector<int>& T_pyramid);
-
-  /**
-   * \brief Detect objects by template matching.
-   *
-   * Matches globally at the lowest pyramid level, then refines locally stepping up the pyramid.
-   *
-   * \param      sources   Source images, one for each modality.
-   * \param      threshold Similarity threshold, a percentage between 0 and 100.
-   * \param[out] matches   Template matches, sorted by similarity score.
-   * \param      class_ids If non-empty, only search for the desired object classes.
-   * \param[out] quantized_images Optionally return vector<Mat> of quantized images.
-   * \param      masks     The masks for consideration during matching. The masks should be CV_8UC1
-   *                       where 255 represents a valid pixel.  If non-empty, the vector must be
-   *                       the same size as sources.  Each element must be
-   *                       empty or the same size as its corresponding source.
-   */
-  void match(const std::vector<Mat>& sources, float threshold, std::vector<Match>& matches,
-             const std::vector<std::string>& class_ids = std::vector<std::string>(),
-             OutputArrayOfArrays quantized_images = noArray(),
-             const std::vector<Mat>& masks = std::vector<Mat>()) const;
-
-  /**
-   * \brief Add new object template.
-   *
-   * \param      sources      Source images, one for each modality.
-   * \param      class_id     Object class ID.
-   * \param      object_mask  Mask separating object from background.
-   * \param[out] bounding_box Optionally return bounding box of the extracted features.
-   *
-   * \return Template ID, or -1 if failed to extract a valid template.
-   */
-  int addTemplate(const std::vector<Mat>& sources, const std::string& class_id,
-          const Mat& object_mask, Rect* bounding_box = NULL);
-
-  /**
-   * \brief Add a new object template computed by external means.
-   */
-  int addSyntheticTemplate(const std::vector<Template>& templates, const std::string& class_id);
-
-  /**
-   * \brief Get the modalities used by this detector.
-   *
-   * You are not permitted to add/remove modalities, but you may dynamic_cast them to
-   * tweak parameters.
-   */
-  const std::vector< Ptr<Modality> >& getModalities() const { return modalities; }
-
-  /**
-   * \brief Get sampling step T at pyramid_level.
-   */
-  int getT(int pyramid_level) const { return T_at_level[pyramid_level]; }
-
-  /**
-   * \brief Get number of pyramid levels used by this detector.
-   */
-  int pyramidLevels() const { return pyramid_levels; }
-
-  /**
-   * \brief Get the template pyramid identified by template_id.
-   *
-   * For example, with 2 modalities (Gradient, Normal) and two pyramid levels
-   * (L0, L1), the order is (GradientL0, NormalL0, GradientL1, NormalL1).
-   */
-  const std::vector<Template>& getTemplates(const std::string& class_id, int template_id) const;
-
-  int numTemplates() const;
-  int numTemplates(const std::string& class_id) const;
-  int numClasses() const { return static_cast<int>(class_templates.size()); }
-
-  std::vector<std::string> classIds() const;
-
-  void read(const FileNode& fn);
-  void write(FileStorage& fs) const;
-
-  std::string readClass(const FileNode& fn, const std::string &class_id_override = "");
-  void writeClass(const std::string& class_id, FileStorage& fs) const;
-
-  void readClasses(const std::vector<std::string>& class_ids,
-                   const std::string& format = "templates_%s.yml.gz");
-  void writeClasses(const std::string& format = "templates_%s.yml.gz") const;
-
-protected:
-  std::vector< Ptr<Modality> > modalities;
-  int pyramid_levels;
-  std::vector<int> T_at_level;
-
-  typedef std::vector<Template> TemplatePyramid;
-  typedef std::map<std::string, std::vector<TemplatePyramid> > TemplatesMap;
-  TemplatesMap class_templates;
-
-  typedef std::vector<Mat> LinearMemories;
-  // Indexed as [pyramid level][modality][quantized label]
-  typedef std::vector< std::vector<LinearMemories> > LinearMemoryPyramid;
-
-  void matchClass(const LinearMemoryPyramid& lm_pyramid,
-                  const std::vector<Size>& sizes,
-                  float threshold, std::vector<Match>& matches,
-                  const std::string& class_id,
-                  const std::vector<TemplatePyramid>& template_pyramids) const;
-};
-
-/**
- * \brief Factory function for detector using LINE algorithm with color gradients.
- *
- * Default parameter settings suitable for VGA images.
- */
-CV_EXPORTS Ptr<Detector> getDefaultLINE();
-
-/**
- * \brief Factory function for detector using LINE-MOD algorithm with color gradients
- * and depth normals.
- *
- * Default parameter settings suitable for VGA images.
- */
-CV_EXPORTS Ptr<Detector> getDefaultLINEMOD();
-
-} // namespace linemod
-} // namespace cv
-
-#endif
+#include "opencv2/objdetect/linemod.hpp"
+#include "opencv2/objdetect/erfilter.hpp"
 
 #endif
diff --git a/modules/objdetect/include/opencv2/objdetect/erfilter.hpp b/modules/objdetect/include/opencv2/objdetect/erfilter.hpp
new file mode 100644
index 0000000000..78b5847662
--- /dev/null
+++ b/modules/objdetect/include/opencv2/objdetect/erfilter.hpp
@@ -0,0 +1,220 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_OBJDETECT_ERFILTER_HPP__
+#define __OPENCV_OBJDETECT_ERFILTER_HPP__
+
+#include "opencv2/core.hpp"
+#include <vector>
+#include <deque>
+
+namespace cv
+{
+
+/*!
+    Extremal Region Stat structure
+
+    The ERStat structure represents a class-specific Extremal Region (ER).
+
+    An ER is a 4-connected set of pixels with all its grey-level values smaller than the values
+    in its outer boundary. A class-specific ER is selected (using a classifier) from all the ER's
+    in the component tree of the image.
+*/
+struct CV_EXPORTS ERStat
+{
+public:
+    //! Constructor
+    explicit ERStat(int level = 256, int pixel = 0, int x = 0, int y = 0);
+    //! Destructor
+    ~ERStat(){};
+
+    //! seed point and the threshold (max grey-level value)
+    int pixel;
+    int level;
+
+    //! incrementally computable features
+    int area;
+    int perimeter;
+    int euler;                 //!< euler number
+    Rect rect;
+    double raw_moments[2];     //!< order 1 raw moments to derive the centroid
+    double central_moments[3]; //!< order 2 central moments to construct the covariance matrix
+    std::deque<int> *crossings;//!< horizontal crossings
+    float med_crossings;       //!< median of the crossings at three different height levels
+
+    //! 2nd stage features
+    float hole_area_ratio;
+    float convex_hull_ratio;
+    float num_inflexion_points;
+
+    // TODO Other features can be added (average color, standard deviation, and such)
+
+
+    // TODO shall we include the pixel list whenever available (i.e. after 2nd stage) ?
+    std::vector<int> *pixels;
+
+    //! probability that the ER belongs to the class we are looking for
+    double probability;
+
+    //! pointers preserving the tree structure of the component tree
+    ERStat* parent;
+    ERStat* child;
+    ERStat* next;
+    ERStat* prev;
+
+    //! wenever the regions is a local maxima of the probability
+    bool local_maxima;
+    ERStat* max_probability_ancestor;
+    ERStat* min_probability_ancestor;
+};
+
+/*!
+    Base class for 1st and 2nd stages of Neumann and Matas scene text detection algorithms
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    Extracts the component tree (if needed) and filter the extremal regions (ER's) by using a given classifier.
+*/
+class CV_EXPORTS ERFilter : public Algorithm
+{
+public:
+
+    //! callback with the classifier is made a class. By doing it we hide SVM, Boost etc.
+    class CV_EXPORTS Callback
+    {
+    public:
+        virtual ~Callback(){};
+        //! The classifier must return probability measure for the region.
+        virtual double eval(const ERStat& stat) = 0; //const = 0; //TODO why cannot use const = 0 here?
+    };
+
+    /*!
+        the key method. Takes image on input and returns the selected regions in a vector of ERStat
+        only distinctive ERs which correspond to characters are selected by a sequential classifier
+        \param image   is the input image
+        \param regions is output for the first stage, input/output for the second one.
+    */
+    virtual void run( InputArray image, std::vector<ERStat>& regions ) = 0;
+
+
+    //! set/get methods to set the algorithm properties,
+    virtual void setCallback(const Ptr<ERFilter::Callback>& cb) = 0;
+    virtual void setThresholdDelta(int thresholdDelta) = 0;
+    virtual void setMinArea(float minArea) = 0;
+    virtual void setMaxArea(float maxArea) = 0;
+    virtual void setMinProbability(float minProbability) = 0;
+    virtual void setMinProbabilityDiff(float minProbabilityDiff) = 0;
+    virtual void setNonMaxSuppression(bool nonMaxSuppression) = 0;
+    virtual int  getNumRejected() = 0;
+};
+
+
+/*!
+    Create an Extremal Region Filter for the 1st stage classifier of N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    The component tree of the image is extracted by a threshold increased step by step
+    from 0 to 255, incrementally computable descriptors (aspect_ratio, compactness,
+    number of holes, and number of horizontal crossings) are computed for each ER
+    and used as features for a classifier which estimates the class-conditional
+    probability P(er|character). The value of P(er|character) is tracked using the inclusion
+    relation of ER across all thresholds and only the ERs which correspond to local maximum
+    of the probability P(er|character) are selected (if the local maximum of the
+    probability is above a global limit pmin and the difference between local maximum and
+    local minimum is greater than minProbabilityDiff).
+
+    \param  cb                Callback with the classifier.
+                              if omitted tries to load a default classifier from file trained_classifierNM1.xml
+    \param  thresholdDelta    Threshold step in subsequent thresholds when extracting the component tree
+    \param  minArea           The minimum area (% of image size) allowed for retreived ER's
+    \param  minArea           The maximum area (% of image size) allowed for retreived ER's
+    \param  minProbability    The minimum probability P(er|character) allowed for retreived ER's
+    \param  nonMaxSuppression Whenever non-maximum suppression is done over the branch probabilities
+    \param  minProbability    The minimum probability difference between local maxima and local minima ERs
+*/
+CV_EXPORTS Ptr<ERFilter> createERFilterNM1(const Ptr<ERFilter::Callback>& cb = Ptr<ERFilter::Callback>(),
+                                                  int thresholdDelta = 1, float minArea = 0.00025,
+                                                  float maxArea = 0.13, float minProbability = 0.4,
+                                                  bool nonMaxSuppression = true,
+                                                  float minProbabilityDiff = 0.1);
+
+/*!
+    Create an Extremal Region Filter for the 2nd stage classifier of N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    In the second stage, the ERs that passed the first stage are classified into character
+    and non-character classes using more informative but also more computationally expensive
+    features. The classifier uses all the features calculated in the first stage and the following
+    additional features: hole area ratio, convex hull ratio, and number of outer inflexion points.
+
+    \param  cb             Callback with the classifier
+                           if omitted tries to load a default classifier from file trained_classifierNM2.xml
+    \param  minProbability The minimum probability P(er|character) allowed for retreived ER's
+*/
+CV_EXPORTS Ptr<ERFilter> createERFilterNM2(const Ptr<ERFilter::Callback>& cb = Ptr<ERFilter::Callback>(),
+                                                  float minProbability = 0.3);
+
+
+// computeNMChannels operation modes
+enum { ERFILTER_NM_RGBLGrad = 0,
+       ERFILTER_NM_IHSGrad  = 1
+     };
+
+/*!
+    Compute the different channels to be processed independently in the N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    In N&M algorithm, the combination of intensity (I), hue (H), saturation (S), and gradient
+    magnitude channels (Grad) are used in order to obtain high localization recall.
+    This implementation also provides an alternative combination of red (R), green (G), blue (B),
+    lightness (L), and gradient magnitude (Grad).
+
+    \param  _src           Source image. Must be RGB CV_8UC3.
+    \param  _channels      Output vector<Mat> where computed channels are stored.
+    \param  _mode          Mode of operation. Currently the only available options are
+                           ERFILTER_NM_RGBLGrad (by default) and ERFILTER_NM_IHSGrad.
+
+*/
+CV_EXPORTS void computeNMChannels(InputArray _src, OutputArrayOfArrays _channels, int _mode = ERFILTER_NM_RGBLGrad);
+
+}
+#endif // _OPENCV_ERFILTER_HPP_
diff --git a/modules/objdetect/include/opencv2/objdetect/linemod.hpp b/modules/objdetect/include/opencv2/objdetect/linemod.hpp
new file mode 100644
index 0000000000..46d869926f
--- /dev/null
+++ b/modules/objdetect/include/opencv2/objdetect/linemod.hpp
@@ -0,0 +1,455 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_OBJDETECT_LINEMOD_HPP__
+#define __OPENCV_OBJDETECT_LINEMOD_HPP__
+
+#include "opencv2/core.hpp"
+#include <map>
+
+/****************************************************************************************\
+*                                 LINE-MOD                                               *
+\****************************************************************************************/
+
+namespace cv {
+namespace linemod {
+
+/// @todo Convert doxy comments to rst
+
+/**
+ * \brief Discriminant feature described by its location and label.
+ */
+struct CV_EXPORTS Feature
+{
+  int x; ///< x offset
+  int y; ///< y offset
+  int label; ///< Quantization
+
+  Feature() : x(0), y(0), label(0) {}
+  Feature(int x, int y, int label);
+
+  void read(const FileNode& fn);
+  void write(FileStorage& fs) const;
+};
+
+inline Feature::Feature(int _x, int _y, int _label) : x(_x), y(_y), label(_label) {}
+
+struct CV_EXPORTS Template
+{
+  int width;
+  int height;
+  int pyramid_level;
+  std::vector<Feature> features;
+
+  void read(const FileNode& fn);
+  void write(FileStorage& fs) const;
+};
+
+/**
+ * \brief Represents a modality operating over an image pyramid.
+ */
+class QuantizedPyramid
+{
+public:
+  // Virtual destructor
+  virtual ~QuantizedPyramid() {}
+
+  /**
+   * \brief Compute quantized image at current pyramid level for online detection.
+   *
+   * \param[out] dst The destination 8-bit image. For each pixel at most one bit is set,
+   *                 representing its classification.
+   */
+  virtual void quantize(Mat& dst) const =0;
+
+  /**
+   * \brief Extract most discriminant features at current pyramid level to form a new template.
+   *
+   * \param[out] templ The new template.
+   */
+  virtual bool extractTemplate(Template& templ) const =0;
+
+  /**
+   * \brief Go to the next pyramid level.
+   *
+   * \todo Allow pyramid scale factor other than 2
+   */
+  virtual void pyrDown() =0;
+
+protected:
+  /// Candidate feature with a score
+  struct Candidate
+  {
+    Candidate(int x, int y, int label, float score);
+
+    /// Sort candidates with high score to the front
+    bool operator<(const Candidate& rhs) const
+    {
+      return score > rhs.score;
+    }
+
+    Feature f;
+    float score;
+  };
+
+  /**
+   * \brief Choose candidate features so that they are not bunched together.
+   *
+   * \param[in]  candidates   Candidate features sorted by score.
+   * \param[out] features     Destination vector of selected features.
+   * \param[in]  num_features Number of candidates to select.
+   * \param[in]  distance     Hint for desired distance between features.
+   */
+  static void selectScatteredFeatures(const std::vector<Candidate>& candidates,
+                                      std::vector<Feature>& features,
+                                      size_t num_features, float distance);
+};
+
+inline QuantizedPyramid::Candidate::Candidate(int x, int y, int label, float _score) : f(x, y, label), score(_score) {}
+
+/**
+ * \brief Interface for modalities that plug into the LINE template matching representation.
+ *
+ * \todo Max response, to allow optimization of summing (255/MAX) features as uint8
+ */
+class CV_EXPORTS Modality
+{
+public:
+  // Virtual destructor
+  virtual ~Modality() {}
+
+  /**
+   * \brief Form a quantized image pyramid from a source image.
+   *
+   * \param[in] src  The source image. Type depends on the modality.
+   * \param[in] mask Optional mask. If not empty, unmasked pixels are set to zero
+   *                 in quantized image and cannot be extracted as features.
+   */
+  Ptr<QuantizedPyramid> process(const Mat& src,
+                    const Mat& mask = Mat()) const
+  {
+    return processImpl(src, mask);
+  }
+
+  virtual String name() const =0;
+
+  virtual void read(const FileNode& fn) =0;
+  virtual void write(FileStorage& fs) const =0;
+
+  /**
+   * \brief Create modality by name.
+   *
+   * The following modality types are supported:
+   * - "ColorGradient"
+   * - "DepthNormal"
+   */
+  static Ptr<Modality> create(const String& modality_type);
+
+  /**
+   * \brief Load a modality from file.
+   */
+  static Ptr<Modality> create(const FileNode& fn);
+
+protected:
+  // Indirection is because process() has a default parameter.
+  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
+                        const Mat& mask) const =0;
+};
+
+/**
+ * \brief Modality that computes quantized gradient orientations from a color image.
+ */
+class CV_EXPORTS ColorGradient : public Modality
+{
+public:
+  /**
+   * \brief Default constructor. Uses reasonable default parameter values.
+   */
+  ColorGradient();
+
+  /**
+   * \brief Constructor.
+   *
+   * \param weak_threshold   When quantizing, discard gradients with magnitude less than this.
+   * \param num_features     How many features a template must contain.
+   * \param strong_threshold Consider as candidate features only gradients whose norms are
+   *                         larger than this.
+   */
+  ColorGradient(float weak_threshold, size_t num_features, float strong_threshold);
+
+  virtual String name() const;
+
+  virtual void read(const FileNode& fn);
+  virtual void write(FileStorage& fs) const;
+
+  float weak_threshold;
+  size_t num_features;
+  float strong_threshold;
+
+protected:
+  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
+                        const Mat& mask) const;
+};
+
+/**
+ * \brief Modality that computes quantized surface normals from a dense depth map.
+ */
+class CV_EXPORTS DepthNormal : public Modality
+{
+public:
+  /**
+   * \brief Default constructor. Uses reasonable default parameter values.
+   */
+  DepthNormal();
+
+  /**
+   * \brief Constructor.
+   *
+   * \param distance_threshold   Ignore pixels beyond this distance.
+   * \param difference_threshold When computing normals, ignore contributions of pixels whose
+   *                             depth difference with the central pixel is above this threshold.
+   * \param num_features         How many features a template must contain.
+   * \param extract_threshold    Consider as candidate feature only if there are no differing
+   *                             orientations within a distance of extract_threshold.
+   */
+  DepthNormal(int distance_threshold, int difference_threshold, size_t num_features,
+              int extract_threshold);
+
+  virtual String name() const;
+
+  virtual void read(const FileNode& fn);
+  virtual void write(FileStorage& fs) const;
+
+  int distance_threshold;
+  int difference_threshold;
+  size_t num_features;
+  int extract_threshold;
+
+protected:
+  virtual Ptr<QuantizedPyramid> processImpl(const Mat& src,
+                        const Mat& mask) const;
+};
+
+/**
+ * \brief Debug function to colormap a quantized image for viewing.
+ */
+void colormap(const Mat& quantized, Mat& dst);
+
+/**
+ * \brief Represents a successful template match.
+ */
+struct CV_EXPORTS Match
+{
+  Match()
+  {
+  }
+
+  Match(int x, int y, float similarity, const String& class_id, int template_id);
+
+  /// Sort matches with high similarity to the front
+  bool operator<(const Match& rhs) const
+  {
+    // Secondarily sort on template_id for the sake of duplicate removal
+    if (similarity != rhs.similarity)
+      return similarity > rhs.similarity;
+    else
+      return template_id < rhs.template_id;
+  }
+
+  bool operator==(const Match& rhs) const
+  {
+    return x == rhs.x && y == rhs.y && similarity == rhs.similarity && class_id == rhs.class_id;
+  }
+
+  int x;
+  int y;
+  float similarity;
+  String class_id;
+  int template_id;
+};
+
+inline
+Match::Match(int _x, int _y, float _similarity, const String& _class_id, int _template_id)
+    : x(_x), y(_y), similarity(_similarity), class_id(_class_id), template_id(_template_id)
+{}
+
+/**
+ * \brief Object detector using the LINE template matching algorithm with any set of
+ * modalities.
+ */
+class CV_EXPORTS Detector
+{
+public:
+  /**
+   * \brief Empty constructor, initialize with read().
+   */
+  Detector();
+
+  /**
+   * \brief Constructor.
+   *
+   * \param modalities       Modalities to use (color gradients, depth normals, ...).
+   * \param T_pyramid        Value of the sampling step T at each pyramid level. The
+   *                         number of pyramid levels is T_pyramid.size().
+   */
+  Detector(const std::vector< Ptr<Modality> >& modalities, const std::vector<int>& T_pyramid);
+
+  /**
+   * \brief Detect objects by template matching.
+   *
+   * Matches globally at the lowest pyramid level, then refines locally stepping up the pyramid.
+   *
+   * \param      sources   Source images, one for each modality.
+   * \param      threshold Similarity threshold, a percentage between 0 and 100.
+   * \param[out] matches   Template matches, sorted by similarity score.
+   * \param      class_ids If non-empty, only search for the desired object classes.
+   * \param[out] quantized_images Optionally return vector<Mat> of quantized images.
+   * \param      masks     The masks for consideration during matching. The masks should be CV_8UC1
+   *                       where 255 represents a valid pixel.  If non-empty, the vector must be
+   *                       the same size as sources.  Each element must be
+   *                       empty or the same size as its corresponding source.
+   */
+  void match(const std::vector<Mat>& sources, float threshold, std::vector<Match>& matches,
+             const std::vector<String>& class_ids = std::vector<String>(),
+             OutputArrayOfArrays quantized_images = noArray(),
+             const std::vector<Mat>& masks = std::vector<Mat>()) const;
+
+  /**
+   * \brief Add new object template.
+   *
+   * \param      sources      Source images, one for each modality.
+   * \param      class_id     Object class ID.
+   * \param      object_mask  Mask separating object from background.
+   * \param[out] bounding_box Optionally return bounding box of the extracted features.
+   *
+   * \return Template ID, or -1 if failed to extract a valid template.
+   */
+  int addTemplate(const std::vector<Mat>& sources, const String& class_id,
+          const Mat& object_mask, Rect* bounding_box = NULL);
+
+  /**
+   * \brief Add a new object template computed by external means.
+   */
+  int addSyntheticTemplate(const std::vector<Template>& templates, const String& class_id);
+
+  /**
+   * \brief Get the modalities used by this detector.
+   *
+   * You are not permitted to add/remove modalities, but you may dynamic_cast them to
+   * tweak parameters.
+   */
+  const std::vector< Ptr<Modality> >& getModalities() const { return modalities; }
+
+  /**
+   * \brief Get sampling step T at pyramid_level.
+   */
+  int getT(int pyramid_level) const { return T_at_level[pyramid_level]; }
+
+  /**
+   * \brief Get number of pyramid levels used by this detector.
+   */
+  int pyramidLevels() const { return pyramid_levels; }
+
+  /**
+   * \brief Get the template pyramid identified by template_id.
+   *
+   * For example, with 2 modalities (Gradient, Normal) and two pyramid levels
+   * (L0, L1), the order is (GradientL0, NormalL0, GradientL1, NormalL1).
+   */
+  const std::vector<Template>& getTemplates(const String& class_id, int template_id) const;
+
+  int numTemplates() const;
+  int numTemplates(const String& class_id) const;
+  int numClasses() const { return static_cast<int>(class_templates.size()); }
+
+  std::vector<String> classIds() const;
+
+  void read(const FileNode& fn);
+  void write(FileStorage& fs) const;
+
+  String readClass(const FileNode& fn, const String &class_id_override = "");
+  void writeClass(const String& class_id, FileStorage& fs) const;
+
+  void readClasses(const std::vector<String>& class_ids,
+                   const String& format = "templates_%s.yml.gz");
+  void writeClasses(const String& format = "templates_%s.yml.gz") const;
+
+protected:
+  std::vector< Ptr<Modality> > modalities;
+  int pyramid_levels;
+  std::vector<int> T_at_level;
+
+  typedef std::vector<Template> TemplatePyramid;
+  typedef std::map<String, std::vector<TemplatePyramid> > TemplatesMap;
+  TemplatesMap class_templates;
+
+  typedef std::vector<Mat> LinearMemories;
+  // Indexed as [pyramid level][modality][quantized label]
+  typedef std::vector< std::vector<LinearMemories> > LinearMemoryPyramid;
+
+  void matchClass(const LinearMemoryPyramid& lm_pyramid,
+                  const std::vector<Size>& sizes,
+                  float threshold, std::vector<Match>& matches,
+                  const String& class_id,
+                  const std::vector<TemplatePyramid>& template_pyramids) const;
+};
+
+/**
+ * \brief Factory function for detector using LINE algorithm with color gradients.
+ *
+ * Default parameter settings suitable for VGA images.
+ */
+CV_EXPORTS Ptr<Detector> getDefaultLINE();
+
+/**
+ * \brief Factory function for detector using LINE-MOD algorithm with color gradients
+ * and depth normals.
+ *
+ * Default parameter settings suitable for VGA images.
+ */
+CV_EXPORTS Ptr<Detector> getDefaultLINEMOD();
+
+} // namespace linemod
+} // namespace cv
+
+#endif // __OPENCV_OBJDETECT_LINEMOD_HPP__
diff --git a/modules/objdetect/include/opencv2/objdetect/objdetect.hpp b/modules/objdetect/include/opencv2/objdetect/objdetect.hpp
index 1a35996146..3ee284f427 100644
--- a/modules/objdetect/include/opencv2/objdetect/objdetect.hpp
+++ b/modules/objdetect/include/opencv2/objdetect/objdetect.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/objdetect.hpp"
\ No newline at end of file
+#include "opencv2/objdetect.hpp"
diff --git a/modules/objdetect/include/opencv2/objdetect/objdetect_c.h b/modules/objdetect/include/opencv2/objdetect/objdetect_c.h
new file mode 100644
index 0000000000..1d6fd2bcec
--- /dev/null
+++ b/modules/objdetect/include/opencv2/objdetect/objdetect_c.h
@@ -0,0 +1,289 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_OBJDETECT_C_H__
+#define __OPENCV_OBJDETECT_C_H__
+
+#include "opencv2/core/core_c.h"
+
+#ifdef __cplusplus
+#include <deque>
+#include <vector>
+
+extern "C" {
+#endif
+
+/****************************************************************************************\
+*                         Haar-like Object Detection functions                           *
+\****************************************************************************************/
+
+#define CV_HAAR_MAGIC_VAL    0x42500000
+#define CV_TYPE_NAME_HAAR    "opencv-haar-classifier"
+
+#define CV_IS_HAAR_CLASSIFIER( haar )                                                    \
+    ((haar) != NULL &&                                                                   \
+    (((const CvHaarClassifierCascade*)(haar))->flags & CV_MAGIC_MASK)==CV_HAAR_MAGIC_VAL)
+
+#define CV_HAAR_FEATURE_MAX  3
+
+typedef struct CvHaarFeature
+{
+    int tilted;
+    struct
+    {
+        CvRect r;
+        float weight;
+    } rect[CV_HAAR_FEATURE_MAX];
+} CvHaarFeature;
+
+typedef struct CvHaarClassifier
+{
+    int count;
+    CvHaarFeature* haar_feature;
+    float* threshold;
+    int* left;
+    int* right;
+    float* alpha;
+} CvHaarClassifier;
+
+typedef struct CvHaarStageClassifier
+{
+    int  count;
+    float threshold;
+    CvHaarClassifier* classifier;
+
+    int next;
+    int child;
+    int parent;
+} CvHaarStageClassifier;
+
+typedef struct CvHidHaarClassifierCascade CvHidHaarClassifierCascade;
+
+typedef struct CvHaarClassifierCascade
+{
+    int  flags;
+    int  count;
+    CvSize orig_window_size;
+    CvSize real_window_size;
+    double scale;
+    CvHaarStageClassifier* stage_classifier;
+    CvHidHaarClassifierCascade* hid_cascade;
+} CvHaarClassifierCascade;
+
+typedef struct CvAvgComp
+{
+    CvRect rect;
+    int neighbors;
+} CvAvgComp;
+
+/* Loads haar classifier cascade from a directory.
+   It is obsolete: convert your cascade to xml and use cvLoad instead */
+CVAPI(CvHaarClassifierCascade*) cvLoadHaarClassifierCascade(
+                    const char* directory, CvSize orig_window_size);
+
+CVAPI(void) cvReleaseHaarClassifierCascade( CvHaarClassifierCascade** cascade );
+
+#define CV_HAAR_DO_CANNY_PRUNING    1
+#define CV_HAAR_SCALE_IMAGE         2
+#define CV_HAAR_FIND_BIGGEST_OBJECT 4
+#define CV_HAAR_DO_ROUGH_SEARCH     8
+
+CVAPI(CvSeq*) cvHaarDetectObjects( const CvArr* image,
+                     CvHaarClassifierCascade* cascade, CvMemStorage* storage,
+                     double scale_factor CV_DEFAULT(1.1),
+                     int min_neighbors CV_DEFAULT(3), int flags CV_DEFAULT(0),
+                     CvSize min_size CV_DEFAULT(cvSize(0,0)), CvSize max_size CV_DEFAULT(cvSize(0,0)));
+
+/* sets images for haar classifier cascade */
+CVAPI(void) cvSetImagesForHaarClassifierCascade( CvHaarClassifierCascade* cascade,
+                                                const CvArr* sum, const CvArr* sqsum,
+                                                const CvArr* tilted_sum, double scale );
+
+/* runs the cascade on the specified window */
+CVAPI(int) cvRunHaarClassifierCascade( const CvHaarClassifierCascade* cascade,
+                                       CvPoint pt, int start_stage CV_DEFAULT(0));
+
+
+/****************************************************************************************\
+*                         Latent SVM Object Detection functions                          *
+\****************************************************************************************/
+
+// DataType: STRUCT position
+// Structure describes the position of the filter in the feature pyramid
+// l - level in the feature pyramid
+// (x, y) - coordinate in level l
+typedef struct CvLSVMFilterPosition
+{
+    int x;
+    int y;
+    int l;
+} CvLSVMFilterPosition;
+
+// DataType: STRUCT filterObject
+// Description of the filter, which corresponds to the part of the object
+// V               - ideal (penalty = 0) position of the partial filter
+//                   from the root filter position (V_i in the paper)
+// penaltyFunction - vector describes penalty function (d_i in the paper)
+//                   pf[0] * x + pf[1] * y + pf[2] * x^2 + pf[3] * y^2
+// FILTER DESCRIPTION
+//   Rectangular map (sizeX x sizeY),
+//   every cell stores feature vector (dimension = p)
+// H               - matrix of feature vectors
+//                   to set and get feature vectors (i,j)
+//                   used formula H[(j * sizeX + i) * p + k], where
+//                   k - component of feature vector in cell (i, j)
+// END OF FILTER DESCRIPTION
+typedef struct CvLSVMFilterObject{
+    CvLSVMFilterPosition V;
+    float fineFunction[4];
+    int sizeX;
+    int sizeY;
+    int numFeatures;
+    float *H;
+} CvLSVMFilterObject;
+
+// data type: STRUCT CvLatentSvmDetector
+// structure contains internal representation of trained Latent SVM detector
+// num_filters          - total number of filters (root plus part) in model
+// num_components       - number of components in model
+// num_part_filters     - array containing number of part filters for each component
+// filters              - root and part filters for all model components
+// b                    - biases for all model components
+// score_threshold      - confidence level threshold
+typedef struct CvLatentSvmDetector
+{
+    int num_filters;
+    int num_components;
+    int* num_part_filters;
+    CvLSVMFilterObject** filters;
+    float* b;
+    float score_threshold;
+} CvLatentSvmDetector;
+
+// data type: STRUCT CvObjectDetection
+// structure contains the bounding box and confidence level for detected object
+// rect                 - bounding box for a detected object
+// score                - confidence level
+typedef struct CvObjectDetection
+{
+    CvRect rect;
+    float score;
+} CvObjectDetection;
+
+//////////////// Object Detection using Latent SVM //////////////
+
+
+/*
+// load trained detector from a file
+//
+// API
+// CvLatentSvmDetector* cvLoadLatentSvmDetector(const char* filename);
+// INPUT
+// filename             - path to the file containing the parameters of
+                        - trained Latent SVM detector
+// OUTPUT
+// trained Latent SVM detector in internal representation
+*/
+CVAPI(CvLatentSvmDetector*) cvLoadLatentSvmDetector(const char* filename);
+
+/*
+// release memory allocated for CvLatentSvmDetector structure
+//
+// API
+// void cvReleaseLatentSvmDetector(CvLatentSvmDetector** detector);
+// INPUT
+// detector             - CvLatentSvmDetector structure to be released
+// OUTPUT
+*/
+CVAPI(void) cvReleaseLatentSvmDetector(CvLatentSvmDetector** detector);
+
+/*
+// find rectangular regions in the given image that are likely
+// to contain objects and corresponding confidence levels
+//
+// API
+// CvSeq* cvLatentSvmDetectObjects(const IplImage* image,
+//                                  CvLatentSvmDetector* detector,
+//                                  CvMemStorage* storage,
+//                                  float overlap_threshold = 0.5f,
+//                                  int numThreads = -1);
+// INPUT
+// image                - image to detect objects in
+// detector             - Latent SVM detector in internal representation
+// storage              - memory storage to store the resultant sequence
+//                          of the object candidate rectangles
+// overlap_threshold    - threshold for the non-maximum suppression algorithm
+                           = 0.5f [here will be the reference to original paper]
+// OUTPUT
+// sequence of detected objects (bounding boxes and confidence levels stored in CvObjectDetection structures)
+*/
+CVAPI(CvSeq*) cvLatentSvmDetectObjects(IplImage* image,
+                                CvLatentSvmDetector* detector,
+                                CvMemStorage* storage,
+                                float overlap_threshold CV_DEFAULT(0.5f),
+                                int numThreads CV_DEFAULT(-1));
+
+#ifdef __cplusplus
+}
+
+CV_EXPORTS CvSeq* cvHaarDetectObjectsForROC( const CvArr* image,
+                     CvHaarClassifierCascade* cascade, CvMemStorage* storage,
+                     std::vector<int>& rejectLevels, std::vector<double>& levelWeightds,
+                     double scale_factor = 1.1,
+                     int min_neighbors = 3, int flags = 0,
+                     CvSize min_size = cvSize(0, 0), CvSize max_size = cvSize(0, 0),
+                     bool outputRejectLevels = false );
+
+struct CvDataMatrixCode
+{
+  char msg[4];
+  CvMat* original;
+  CvMat* corners;
+};
+
+CV_EXPORTS std::deque<CvDataMatrixCode> cvFindDataMatrix(CvMat *im);
+
+#endif
+
+
+#endif /* __OPENCV_OBJDETECT_C_H__ */
diff --git a/modules/objdetect/perf/perf_cascadeclassifier.cpp b/modules/objdetect/perf/perf_cascadeclassifier.cpp
index 2823302c77..1d5bff11f2 100644
--- a/modules/objdetect/perf/perf_cascadeclassifier.cpp
+++ b/modules/objdetect/perf/perf_cascadeclassifier.cpp
@@ -47,4 +47,4 @@ PERF_TEST_P(ImageName_MinSize, CascadeClassifierLBPFrontalFace,
 
     std::sort(faces.begin(), faces.end(), comparators::RectLess());
     SANITY_CHECK(faces, 3.001 * faces.size());
-}
\ No newline at end of file
+}
diff --git a/modules/objdetect/perf/perf_precomp.cpp b/modules/objdetect/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/objdetect/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/objdetect/src/_lsvm_routine.h b/modules/objdetect/src/_lsvm_routine.h
index 6000afae2d..5a95b68b16 100644
--- a/modules/objdetect/src/_lsvm_routine.h
+++ b/modules/objdetect/src/_lsvm_routine.h
@@ -1,6 +1,8 @@
 #ifndef _LSVM_ROUTINE_H_
 #define _LSVM_ROUTINE_H_
 
+#include "opencv2/objdetect/objdetect_c.h"
+
 #include "_lsvm_types.h"
 #include "_lsvm_error.h"
 
diff --git a/modules/objdetect/src/_lsvm_tbbversion.h b/modules/objdetect/src/_lsvm_tbbversion.h
index fb02d8ac04..8bfca1bc11 100644
--- a/modules/objdetect/src/_lsvm_tbbversion.h
+++ b/modules/objdetect/src/_lsvm_tbbversion.h
@@ -48,4 +48,4 @@ int tbbTasksThresholdFunctionalScore(const CvLSVMFilterObject **filters, const i
                                      int *kPoints,
                                      CvPoint ****partsDisplacement);
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/objdetect/src/_lsvmparser.h b/modules/objdetect/src/_lsvmparser.h
index 35d4bf6866..0c34ede8ad 100644
--- a/modules/objdetect/src/_lsvmparser.h
+++ b/modules/objdetect/src/_lsvmparser.h
@@ -1,5 +1,6 @@
 #ifndef LSVM_PARSER
 #define LSVM_PARSER
+#include "opencv2/objdetect/objdetect_c.h"
 
 #include "_lsvm_types.h"
 
diff --git a/modules/objdetect/src/cascadedetect.cpp b/modules/objdetect/src/cascadedetect.cpp
index 2e5d1b18aa..88f463faa0 100644
--- a/modules/objdetect/src/cascadedetect.cpp
+++ b/modules/objdetect/src/cascadedetect.cpp
@@ -43,8 +43,7 @@
 #include <cstdio>
 
 #include "cascadedetect.hpp"
-
-#include <string>
+#include "opencv2/objdetect/objdetect_c.h"
 
 #if defined (LOG_CASCADE_STATISTIC)
 struct Logger
@@ -114,24 +113,6 @@ struct Logger
 namespace cv
 {
 
-// class for grouping object candidates, detected by Cascade Classifier, HOG etc.
-// instance of the class is to be passed to cv::partition (see cxoperations.hpp)
-class CV_EXPORTS SimilarRects
-{
-public:
-    SimilarRects(double _eps) : eps(_eps) {}
-    inline bool operator()(const Rect& r1, const Rect& r2) const
-    {
-        double delta = eps*(std::min(r1.width, r2.width) + std::min(r1.height, r2.height))*0.5;
-        return std::abs(r1.x - r2.x) <= delta &&
-        std::abs(r1.y - r2.y) <= delta &&
-        std::abs(r1.x + r1.width - r2.x - r2.width) <= delta &&
-        std::abs(r1.y + r1.height - r2.y - r2.height) <= delta;
-    }
-    double eps;
-};
-
-
 void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps, std::vector<int>* weights, std::vector<double>* levelWeights)
 {
     if( groupThreshold <= 0 || rectList.empty() )
@@ -197,8 +178,11 @@ void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps
     for( i = 0; i < nclasses; i++ )
     {
         Rect r1 = rrects[i];
-        int n1 = levelWeights ? rejectLevels[i] : rweights[i];
+        int n1 = rweights[i];
         double w1 = rejectWeights[i];
+        int l1 = rejectLevels[i];
+
+        // filter out rectangles which don't have enough similar rectangles
         if( n1 <= groupThreshold )
             continue;
         // filter out small face rectangles inside large rectangles
@@ -226,7 +210,7 @@ void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps
         {
             rectList.push_back(r1);
             if( weights )
-                weights->push_back(n1);
+                weights->push_back(l1);
             if( levelWeights )
                 levelWeights->push_back(w1);
         }
@@ -483,7 +467,7 @@ bool HaarEvaluator::Feature :: read( const FileNode& node )
 
 HaarEvaluator::HaarEvaluator()
 {
-    features = new std::vector<Feature>();
+    features = makePtr<std::vector<Feature> >();
 }
 HaarEvaluator::~HaarEvaluator()
 {
@@ -508,7 +492,7 @@ bool HaarEvaluator::read(const FileNode& node)
 
 Ptr<FeatureEvaluator> HaarEvaluator::clone() const
 {
-    HaarEvaluator* ret = new HaarEvaluator;
+    Ptr<HaarEvaluator> ret = makePtr<HaarEvaluator>();
     ret->origWinSize = origWinSize;
     ret->features = features;
     ret->featuresPtr = &(*ret->features)[0];
@@ -598,7 +582,7 @@ bool LBPEvaluator::Feature :: read(const FileNode& node )
 
 LBPEvaluator::LBPEvaluator()
 {
-    features = new std::vector<Feature>();
+    features = makePtr<std::vector<Feature> >();
 }
 LBPEvaluator::~LBPEvaluator()
 {
@@ -619,7 +603,7 @@ bool LBPEvaluator::read( const FileNode& node )
 
 Ptr<FeatureEvaluator> LBPEvaluator::clone() const
 {
-    LBPEvaluator* ret = new LBPEvaluator;
+    Ptr<LBPEvaluator> ret = makePtr<LBPEvaluator>();
     ret->origWinSize = origWinSize;
     ret->features = features;
     ret->featuresPtr = &(*ret->features)[0];
@@ -678,7 +662,7 @@ bool HOGEvaluator::Feature :: read( const FileNode& node )
 
 HOGEvaluator::HOGEvaluator()
 {
-    features = new std::vector<Feature>();
+    features = makePtr<std::vector<Feature> >();
 }
 
 HOGEvaluator::~HOGEvaluator()
@@ -700,7 +684,7 @@ bool HOGEvaluator::read( const FileNode& node )
 
 Ptr<FeatureEvaluator> HOGEvaluator::clone() const
 {
-    HOGEvaluator* ret = new HOGEvaluator;
+    Ptr<HOGEvaluator> ret = makePtr<HOGEvaluator>();
     ret->origWinSize = origWinSize;
     ret->features = features;
     ret->featuresPtr = &(*ret->features)[0];
@@ -854,7 +838,7 @@ CascadeClassifier::CascadeClassifier()
 {
 }
 
-CascadeClassifier::CascadeClassifier(const std::string& filename)
+CascadeClassifier::CascadeClassifier(const String& filename)
 {
     load(filename);
 }
@@ -865,10 +849,10 @@ CascadeClassifier::~CascadeClassifier()
 
 bool CascadeClassifier::empty() const
 {
-    return oldCascade.empty() && data.stages.empty();
+    return !oldCascade && data.stages.empty();
 }
 
-bool CascadeClassifier::load(const std::string& filename)
+bool CascadeClassifier::load(const String& filename)
 {
     oldCascade.release();
     data = Data();
@@ -883,13 +867,13 @@ bool CascadeClassifier::load(const std::string& filename)
 
     fs.release();
 
-    oldCascade = Ptr<CvHaarClassifierCascade>((CvHaarClassifierCascade*)cvLoad(filename.c_str(), 0, 0, 0));
+    oldCascade.reset((CvHaarClassifierCascade*)cvLoad(filename.c_str(), 0, 0, 0));
     return !oldCascade.empty();
 }
 
 int CascadeClassifier::runAt( Ptr<FeatureEvaluator>& evaluator, Point pt, double& weight )
 {
-    CV_Assert( oldCascade.empty() );
+    CV_Assert( !oldCascade );
 
     assert( data.featureType == FeatureEvaluator::HAAR ||
             data.featureType == FeatureEvaluator::LBP ||
@@ -989,7 +973,7 @@ public:
                 {
                     if( result == 1 )
                         result =  -(int)classifier->data.stages.size();
-                    if( classifier->data.stages.size() + result < 4 )
+                    if( classifier->data.stages.size() + result == 0 )
                     {
                         mtx->lock();
                         rectangles->push_back(Rect(cvRound(x*scalingFactor), cvRound(y*scalingFactor), winSize.width, winSize.height));
@@ -1023,6 +1007,7 @@ public:
 };
 
 struct getRect { Rect operator ()(const CvAvgComp& e) const { return e.rect; } };
+struct getNeighbors { int operator ()(const CvAvgComp& e) const { return e.neighbors; } };
 
 
 bool CascadeClassifier::detectSingleScale( const Mat& image, int stripCount, Size processingRectSize,
@@ -1037,7 +1022,7 @@ bool CascadeClassifier::detectSingleScale( const Mat& image, int stripCount, Siz
 #endif
 
     Mat currentMask;
-    if (!maskGenerator.empty()) {
+    if (maskGenerator) {
         currentMask=maskGenerator->generateMask(image);
     }
 
@@ -1087,39 +1072,33 @@ bool CascadeClassifier::setImage(const Mat& image)
     return featureEvaluator->setImage(image, data.origWinSize);
 }
 
-void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& objects,
-                                          std::vector<int>& rejectLevels,
-                                          std::vector<double>& levelWeights,
-                                          double scaleFactor, int minNeighbors,
-                                          int flags, Size minObjectSize, Size maxObjectSize,
-                                          bool outputRejectLevels )
+static void detectMultiScaleOldFormat( const Mat& image, Ptr<CvHaarClassifierCascade> oldCascade,
+                                       std::vector<Rect>& objects,
+                                       std::vector<int>& rejectLevels,
+                                       std::vector<double>& levelWeights,
+                                       std::vector<CvAvgComp>& vecAvgComp,
+                                       double scaleFactor, int minNeighbors,
+                                       int flags, Size minObjectSize, Size maxObjectSize,
+                                       bool outputRejectLevels = false )
 {
-    const double GROUP_EPS = 0.2;
+    MemStorage storage(cvCreateMemStorage(0));
+    CvMat _image = image;
+    CvSeq* _objects = cvHaarDetectObjectsForROC( &_image, oldCascade, storage, rejectLevels, levelWeights, scaleFactor,
+                                                 minNeighbors, flags, minObjectSize, maxObjectSize, outputRejectLevels );
+    Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
+    objects.resize(vecAvgComp.size());
+    std::transform(vecAvgComp.begin(), vecAvgComp.end(), objects.begin(), getRect());
+}
 
-    CV_Assert( scaleFactor > 1 && image.depth() == CV_8U );
+void CascadeClassifier::detectMultiScaleNoGrouping( const Mat& image, std::vector<Rect>& candidates,
+                                                    std::vector<int>& rejectLevels, std::vector<double>& levelWeights,
+                                                    double scaleFactor, Size minObjectSize, Size maxObjectSize,
+                                                    bool outputRejectLevels )
+{
+    candidates.clear();
 
-    if( empty() )
-        return;
-
-    if( isOldFormatCascade() )
-    {
-        MemStorage storage(cvCreateMemStorage(0));
-        CvMat _image = image;
-        CvSeq* _objects = cvHaarDetectObjectsForROC( &_image, oldCascade, storage, rejectLevels, levelWeights, scaleFactor,
-                                              minNeighbors, flags, minObjectSize, maxObjectSize, outputRejectLevels );
-        std::vector<CvAvgComp> vecAvgComp;
-        Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
-        objects.resize(vecAvgComp.size());
-        std::transform(vecAvgComp.begin(), vecAvgComp.end(), objects.begin(), getRect());
-        return;
-    }
-
-    objects.clear();
-
-    if (!maskGenerator.empty()) {
+    if (maskGenerator)
         maskGenerator->initializeMask(image);
-    }
-
 
     if( maxObjectSize.height == 0 || maxObjectSize.width == 0 )
         maxObjectSize = image.size();
@@ -1128,12 +1107,11 @@ void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& o
     if( grayImage.channels() > 1 )
     {
         Mat temp;
-        cvtColor(grayImage, temp, CV_BGR2GRAY);
+        cvtColor(grayImage, temp, COLOR_BGR2GRAY);
         grayImage = temp;
     }
 
     Mat imageBuffer(image.rows + 1, image.cols + 1, CV_8U);
-    std::vector<Rect> candidates;
 
     for( double factor = 1; ; factor *= scaleFactor )
     {
@@ -1141,7 +1119,7 @@ void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& o
 
         Size windowSize( cvRound(originalWindowSize.width*factor), cvRound(originalWindowSize.height*factor) );
         Size scaledImageSize( cvRound( grayImage.cols/factor ), cvRound( grayImage.rows/factor ) );
-        Size processingRectSize( scaledImageSize.width - originalWindowSize.width + 1, scaledImageSize.height - originalWindowSize.height + 1 );
+        Size processingRectSize( scaledImageSize.width - originalWindowSize.width, scaledImageSize.height - originalWindowSize.height );
 
         if( processingRectSize.width <= 0 || processingRectSize.height <= 0 )
             break;
@@ -1151,7 +1129,7 @@ void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& o
             continue;
 
         Mat scaledImage( scaledImageSize, CV_8U, imageBuffer.data );
-        resize( grayImage, scaledImage, scaledImageSize, 0, 0, CV_INTER_LINEAR );
+        resize( grayImage, scaledImage, scaledImageSize, 0, 0, INTER_LINEAR );
 
         int yStep;
         if( getFeatureType() == cv::FeatureEvaluator::HOG )
@@ -1165,32 +1143,48 @@ void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& o
 
         int stripCount, stripSize;
 
-    #ifdef HAVE_TBB
         const int PTS_PER_THREAD = 1000;
         stripCount = ((processingRectSize.width/yStep)*(processingRectSize.height + yStep-1)/yStep + PTS_PER_THREAD/2)/PTS_PER_THREAD;
         stripCount = std::min(std::max(stripCount, 1), 100);
         stripSize = (((processingRectSize.height + stripCount - 1)/stripCount + yStep-1)/yStep)*yStep;
-    #else
-        stripCount = 1;
-        stripSize = processingRectSize.height;
-    #endif
 
         if( !detectSingleScale( scaledImage, stripCount, processingRectSize, stripSize, yStep, factor, candidates,
             rejectLevels, levelWeights, outputRejectLevels ) )
             break;
     }
+}
 
+void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& objects,
+                                          std::vector<int>& rejectLevels,
+                                          std::vector<double>& levelWeights,
+                                          double scaleFactor, int minNeighbors,
+                                          int flags, Size minObjectSize, Size maxObjectSize,
+                                          bool outputRejectLevels )
+{
+    CV_Assert( scaleFactor > 1 && image.depth() == CV_8U );
 
-    objects.resize(candidates.size());
-    std::copy(candidates.begin(), candidates.end(), objects.begin());
+    if( empty() )
+        return;
 
-    if( outputRejectLevels )
+    if( isOldFormatCascade() )
     {
-        groupRectangles( objects, rejectLevels, levelWeights, minNeighbors, GROUP_EPS );
+        std::vector<CvAvgComp> fakeVecAvgComp;
+        detectMultiScaleOldFormat( image, oldCascade, objects, rejectLevels, levelWeights, fakeVecAvgComp, scaleFactor,
+                                   minNeighbors, flags, minObjectSize, maxObjectSize, outputRejectLevels );
     }
     else
     {
-        groupRectangles( objects, minNeighbors, GROUP_EPS );
+        detectMultiScaleNoGrouping( image, objects, rejectLevels, levelWeights, scaleFactor, minObjectSize, maxObjectSize,
+                                    outputRejectLevels );
+        const double GROUP_EPS = 0.2;
+        if( outputRejectLevels )
+        {
+            groupRectangles( objects, rejectLevels, levelWeights, minNeighbors, GROUP_EPS );
+        }
+        else
+        {
+            groupRectangles( objects, minNeighbors, GROUP_EPS );
+        }
     }
 }
 
@@ -1201,7 +1195,35 @@ void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& o
     std::vector<int> fakeLevels;
     std::vector<double> fakeWeights;
     detectMultiScale( image, objects, fakeLevels, fakeWeights, scaleFactor,
-        minNeighbors, flags, minObjectSize, maxObjectSize, false );
+        minNeighbors, flags, minObjectSize, maxObjectSize );
+}
+
+void CascadeClassifier::detectMultiScale( const Mat& image, std::vector<Rect>& objects,
+                                          std::vector<int>& numDetections, double scaleFactor,
+                                          int minNeighbors, int flags, Size minObjectSize,
+                                          Size maxObjectSize )
+{
+    CV_Assert( scaleFactor > 1 && image.depth() == CV_8U );
+
+    if( empty() )
+        return;
+
+    std::vector<int> fakeLevels;
+    std::vector<double> fakeWeights;
+    if( isOldFormatCascade() )
+    {
+        std::vector<CvAvgComp> vecAvgComp;
+        detectMultiScaleOldFormat( image, oldCascade, objects, fakeLevels, fakeWeights, vecAvgComp, scaleFactor,
+                                   minNeighbors, flags, minObjectSize, maxObjectSize );
+        numDetections.resize(vecAvgComp.size());
+        std::transform(vecAvgComp.begin(), vecAvgComp.end(), numDetections.begin(), getNeighbors());
+    }
+    else
+    {
+        detectMultiScaleNoGrouping( image, objects, fakeLevels, fakeWeights, scaleFactor, minObjectSize, maxObjectSize );
+        const double GROUP_EPS = 0.2;
+        groupRectangles( objects, numDetections, minNeighbors, GROUP_EPS );
+    }
 }
 
 bool CascadeClassifier::Data::read(const FileNode &root)
@@ -1209,13 +1231,13 @@ bool CascadeClassifier::Data::read(const FileNode &root)
     static const float THRESHOLD_EPS = 1e-5f;
 
     // load stage params
-    std::string stageTypeStr = (std::string)root[CC_STAGE_TYPE];
+    String stageTypeStr = (String)root[CC_STAGE_TYPE];
     if( stageTypeStr == CC_BOOST )
         stageType = BOOST;
     else
         return false;
 
-    std::string featureTypeStr = (std::string)root[CC_FEATURE_TYPE];
+    String featureTypeStr = (String)root[CC_FEATURE_TYPE];
     if( featureTypeStr == CC_HAAR )
         featureType = FeatureEvaluator::HAAR;
     else if( featureTypeStr == CC_LBP )
@@ -1328,7 +1350,7 @@ bool CascadeClassifier::read(const FileNode& root)
     return featureEvaluator->read(fn);
 }
 
-template<> void Ptr<CvHaarClassifierCascade>::delete_obj()
+template<> void DefaultDeleter<CvHaarClassifierCascade>::operator ()(CvHaarClassifierCascade* obj) const
 { cvReleaseHaarClassifierCascade(&obj); }
 
 } // namespace cv
diff --git a/modules/objdetect/src/cascadedetect.hpp b/modules/objdetect/src/cascadedetect.hpp
index c6da4b9ffe..de3303e4aa 100644
--- a/modules/objdetect/src/cascadedetect.hpp
+++ b/modules/objdetect/src/cascadedetect.hpp
@@ -487,4 +487,3 @@ inline int predictCategoricalStump( CascadeClassifier& cascade, Ptr<FeatureEvalu
     return 1;
 }
 }
-
diff --git a/modules/objdetect/src/datamatrix.cpp b/modules/objdetect/src/datamatrix.cpp
index 325032c533..754dd11980 100644
--- a/modules/objdetect/src/datamatrix.cpp
+++ b/modules/objdetect/src/datamatrix.cpp
@@ -1,6 +1,7 @@
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
 
-#include <deque>
 #include <algorithm>
 
 class Sampler {
@@ -43,7 +44,7 @@ unsigned char ccblk[256] = { 34,17,2,17,19,19,2,17,36,36,2,36,19,19,2,17,51,51,2
     36,19,19,2,32,66,66,2,66,19,19,2,66,36,36,2,36,19,19,2,66,51,51,2,51,19,19,2,51,36,36,2,36,19,19,2,32,49,49,2,49,
     19,19,2,49,36,36,2,36,19,19,2,49,51,51,2,51,19,19,2,51,36,36,2,36,19,19,2,49,66,66,2,66,19,19,2,66,36,36,2,36,19,
     19,2,66,51,51,2,51,19,19,2,51,36,36,2,36,19,19,2,34 };
-static const CvPoint pickup[64] = { {7,6},{8,6},{7,5},{8,5},{1,5},{7,4},{8,4},{1,4},{1,8},{2,8},{1,7},{2,7},{3,7},
+static const int pickup[64][2] = { {7,6},{8,6},{7,5},{8,5},{1,5},{7,4},{8,4},{1,4},{1,8},{2,8},{1,7},{2,7},{3,7},
     {1,6},{2,6},{3,6},{3,2},{4,2},{3,1},{4,1},{5,1},{3,8},{4,8},{5,8},{6,1},{7,1},{6,8},{7,8},{8,8},{6,7},{7,7},{8,7},
     {4,7},{5,7},{4,6},{5,6},{6,6},{4,5},{5,5},{6,5},{2,5},{3,5},{2,4},{3,4},{4,4},{2,3},{3,3},{4,3},{8,3},{1,3},{8,2},
     {1,2},{2,2},{8,1},{1,1},{2,1},{5,4},{6,4},{5,3},{6,3},{7,3},{5,2},{6,2},{7,2} };
@@ -255,7 +256,7 @@ static int decode(Sampler &sa, code &cc)
     sum += sa.getpixel(1 + (i & 7), 1 + (i >> 3));
   uchar mean = (uchar)(sum / 64);
   for (int i = 0; i < 64; i++) {
-    b = (b << 1) + (sa.getpixel(pickup[i].x, pickup[i].y) <= mean);
+    b = (b << 1) + (sa.getpixel(pickup[i][0], pickup[i][1]) <= mean);
     if ((i & 7) == 7) {
       binary[i >> 3] = b;
       b = 0;
@@ -489,7 +490,7 @@ namespace cv
 {
 
 void findDataMatrix(InputArray _image,
-                    std::vector<std::string>& codes,
+                    std::vector<String>& codes,
                     OutputArray _corners,
                     OutputArrayOfArrays _dmtx)
 {
@@ -513,7 +514,7 @@ void findDataMatrix(InputArray _image,
     for( i = 0; i < n; i++ )
     {
         CvDataMatrixCode& rc_i = rc[i];
-        codes[i] = std::string(rc_i.msg);
+        codes[i] = String(rc_i.msg);
 
         if( corners.data )
         {
@@ -528,14 +529,14 @@ void findDataMatrix(InputArray _image,
         {
             _dmtx.create(rc_i.original->rows, rc_i.original->cols, rc_i.original->type, i);
             Mat dst = _dmtx.getMat(i);
-            Mat(rc_i.original).copyTo(dst);
+            cv::cvarrToMat(rc_i.original).copyTo(dst);
         }
         cvReleaseMat(&rc_i.original);
     }
 }
 
 void drawDataMatrixCodes(InputOutputArray _image,
-                         const std::vector<std::string>& codes,
+                         const std::vector<String>& codes,
                          InputArray _corners)
 {
     Mat image = _image.getMat();
diff --git a/modules/objdetect/src/erfilter.cpp b/modules/objdetect/src/erfilter.cpp
new file mode 100644
index 0000000000..c231a513c5
--- /dev/null
+++ b/modules/objdetect/src/erfilter.cpp
@@ -0,0 +1,1258 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include <fstream>
+
+using namespace std;
+
+namespace cv
+{
+
+ERStat::ERStat(int init_level, int init_pixel, int init_x, int init_y) : pixel(init_pixel),
+               level(init_level), area(0), perimeter(0), euler(0), probability(1.0),
+               parent(0), child(0), next(0), prev(0), local_maxima(0),
+               max_probability_ancestor(0), min_probability_ancestor(0)
+{
+    rect = Rect(init_x,init_y,1,1);
+    raw_moments[0] = 0.0;
+    raw_moments[1] = 0.0;
+    central_moments[0] = 0.0;
+    central_moments[1] = 0.0;
+    central_moments[2] = 0.0;
+    crossings = new std::deque<int>();
+    crossings->push_back(0);
+}
+
+
+// derivative classes
+
+
+// the classe implementing the interface for the 1st and 2nd stages of Neumann and Matas algorithm
+class CV_EXPORTS ERFilterNM : public ERFilter
+{
+public:
+    //Constructor
+    ERFilterNM();
+    //Destructor
+    ~ERFilterNM() {};
+
+    float minProbability;
+    bool  nonMaxSuppression;
+    float minProbabilityDiff;
+
+    // the key method. Takes image on input, vector of ERStat is output for the first stage,
+    // input/output - for the second one.
+    void run( InputArray image, std::vector<ERStat>& regions );
+
+protected:
+    int thresholdDelta;
+    float maxArea;
+    float minArea;
+
+    Ptr<ERFilter::Callback> classifier;
+
+    // count of the rejected/accepted regions
+    int num_rejected_regions;
+    int num_accepted_regions;
+
+public:
+
+    // set/get methods to set the algorithm properties,
+    void setCallback(const Ptr<ERFilter::Callback>& cb);
+    void setThresholdDelta(int thresholdDelta);
+    void setMinArea(float minArea);
+    void setMaxArea(float maxArea);
+    void setMinProbability(float minProbability);
+    void setMinProbabilityDiff(float minProbabilityDiff);
+    void setNonMaxSuppression(bool nonMaxSuppression);
+    int  getNumRejected();
+
+private:
+    // pointer to the input/output regions vector
+    std::vector<ERStat> *regions;
+    // image mask used for feature calculations
+    Mat region_mask;
+
+    // extract the component tree and store all the ER regions
+    void er_tree_extract( InputArray image );
+    // accumulate a pixel into an ER
+    void er_add_pixel( ERStat *parent, int x, int y, int non_boundary_neighbours,
+                       int non_boundary_neighbours_horiz,
+                       int d_C1, int d_C2, int d_C3 );
+    // merge an ER with its nested parent
+    void er_merge( ERStat *parent, ERStat *child );
+    // recursively walk the tree and clean memory
+    void er_tree_clean( ERStat *er );
+    // copy extracted regions into the output vector
+    ERStat* er_save( ERStat *er, ERStat *parent, ERStat *prev );
+    // recursively walk the tree and filter (remove) regions using the callback classifier
+    ERStat* er_tree_filter( InputArray image, ERStat *stat, ERStat *parent, ERStat *prev );
+    // recursively walk the tree selecting only regions with local maxima probability
+    ERStat* er_tree_nonmax_suppression( ERStat *er, ERStat *parent, ERStat *prev );
+};
+
+
+// default 1st stage classifier
+class CV_EXPORTS ERClassifierNM1 : public ERFilter::Callback
+{
+public:
+    //Constructor
+    ERClassifierNM1();
+    // Destructor
+    ~ERClassifierNM1() {};
+
+    // The classifier must return probability measure for the region.
+    double eval(const ERStat& stat);
+
+private:
+    CvBoost boost;
+};
+
+// default 2nd stage classifier
+class CV_EXPORTS ERClassifierNM2 : public ERFilter::Callback
+{
+public:
+    //constructor
+    ERClassifierNM2();
+    // Destructor
+    ~ERClassifierNM2() {};
+
+    // The classifier must return probability measure for the region.
+    double eval(const ERStat& stat);
+
+private:
+    CvBoost boost;
+};
+
+
+
+
+
+// default constructor
+ERFilterNM::ERFilterNM()
+{
+    thresholdDelta = 1;
+    minArea = 0.;
+    maxArea = 1.;
+    minProbability = 0.;
+    nonMaxSuppression = false;
+    minProbabilityDiff = 1.;
+    num_accepted_regions = 0;
+    num_rejected_regions = 0;
+}
+
+// the key method. Takes image on input, vector of ERStat is output for the first stage,
+// input/output for the second one.
+void ERFilterNM::run( InputArray image, std::vector<ERStat>& _regions )
+{
+
+    // assert correct image type
+    CV_Assert( image.getMat().type() == CV_8UC1 );
+
+    regions = &_regions;
+    region_mask = Mat::zeros(image.getMat().rows+2, image.getMat().cols+2, CV_8UC1);
+
+    // if regions vector is empty we must extract the entire component tree
+    if ( regions->size() == 0 )
+    {
+        er_tree_extract( image );
+        if (nonMaxSuppression)
+        {
+            vector<ERStat> aux_regions;
+            regions->swap(aux_regions);
+            regions->reserve(aux_regions.size());
+            er_tree_nonmax_suppression( &aux_regions.front(), NULL, NULL );
+            aux_regions.clear();
+        }
+    }
+    else // if regions vector is already filled we'll just filter the current regions
+    {
+        // the tree root must have no parent
+        CV_Assert( regions->front().parent == NULL );
+
+        vector<ERStat> aux_regions;
+        regions->swap(aux_regions);
+        regions->reserve(aux_regions.size());
+        er_tree_filter( image, &aux_regions.front(), NULL, NULL );
+        aux_regions.clear();
+    }
+}
+
+// extract the component tree and store all the ER regions
+// uses the algorithm described in
+// Linear time maximally stable extremal regions, D Nistér, H Stewénius – ECCV 2008
+void ERFilterNM::er_tree_extract( InputArray image )
+{
+
+    Mat src = image.getMat();
+    // assert correct image type
+    CV_Assert( src.type() == CV_8UC1 );
+
+    if (thresholdDelta > 1)
+    {
+        Mat tmp;
+        src.copyTo(tmp);
+        src.release();
+        src = (image.getMat() / thresholdDelta) -1;
+    }
+
+    const unsigned char * image_data = src.data;
+    int width = src.cols, height = src.rows;
+
+    // the component stack
+    vector<ERStat*> er_stack;
+
+    //the quads for euler number calculation
+    unsigned char quads[3][4];
+    quads[0][0] = 1 << 3;
+    quads[0][1] = 1 << 2;
+    quads[0][2] = 1 << 1;
+    quads[0][3] = 1;
+    quads[1][0] = (1<<2)|(1<<1)|(1);
+    quads[1][1] = (1<<3)|(1<<1)|(1);
+    quads[1][2] = (1<<3)|(1<<2)|(1);
+    quads[1][3] = (1<<3)|(1<<2)|(1<<1);
+    quads[2][0] = (1<<2)|(1<<1);
+    quads[2][1] = (1<<3)|(1);
+    quads[2][3] = 255;
+
+
+    // masks to know if a pixel is accessible and if it has been already added to some region
+    vector<bool> accessible_pixel_mask(width * height);
+    vector<bool> accumulated_pixel_mask(width * height);
+
+    // heap of boundary pixels
+    vector<int> boundary_pixes[256];
+    vector<int> boundary_edges[256];
+
+    // add a dummy-component before start
+    er_stack.push_back(new ERStat);
+
+    // we'll look initially for all pixels with grey-level lower than a grey-level higher than any allowed in the image
+    int threshold_level = (255/thresholdDelta)+1;
+
+    // starting from the first pixel (0,0)
+    int current_pixel = 0;
+    int current_edge = 0;
+    int current_level = image_data[0];
+    accessible_pixel_mask[0] = true;
+
+    bool push_new_component = true;
+
+    for (;;) {
+
+        int x = current_pixel % width;
+        int y = current_pixel / width;
+
+        // push a component with current level in the component stack
+        if (push_new_component)
+            er_stack.push_back(new ERStat(current_level, current_pixel, x, y));
+        push_new_component = false;
+
+        // explore the (remaining) edges to the neighbors to the current pixel
+        for (current_edge = current_edge; current_edge < 4; current_edge++)
+        {
+
+            int neighbour_pixel = current_pixel;
+
+            switch (current_edge)
+            {
+                    case 0: if (x < width - 1) neighbour_pixel = current_pixel + 1;  break;
+                    case 1: if (y < height - 1) neighbour_pixel = current_pixel + width; break;
+                    case 2: if (x > 0) neighbour_pixel = current_pixel - 1; break;
+                    default: if (y > 0) neighbour_pixel = current_pixel - width; break;
+            }
+
+            // if neighbour is not accessible, mark it accessible and retreive its grey-level value
+            if ( !accessible_pixel_mask[neighbour_pixel] && (neighbour_pixel != current_pixel) )
+            {
+
+                int neighbour_level = image_data[neighbour_pixel];
+                accessible_pixel_mask[neighbour_pixel] = true;
+
+                // if neighbour level is not lower than current level add neighbour to the boundary heap
+                if (neighbour_level >= current_level)
+                {
+
+                    boundary_pixes[neighbour_level].push_back(neighbour_pixel);
+                    boundary_edges[neighbour_level].push_back(0);
+
+                    // if neighbour level is lower than our threshold_level set threshold_level to neighbour level
+                    if (neighbour_level < threshold_level)
+                        threshold_level = neighbour_level;
+
+                }
+                else // if neighbour level is lower than current add current_pixel (and next edge)
+                     // to the boundary heap for later processing
+                {
+
+                    boundary_pixes[current_level].push_back(current_pixel);
+                    boundary_edges[current_level].push_back(current_edge + 1);
+
+                    // if neighbour level is lower than threshold_level set threshold_level to neighbour level
+                    if (current_level < threshold_level)
+                        threshold_level = current_level;
+
+                    // consider the new pixel and its grey-level as current pixel
+                    current_pixel = neighbour_pixel;
+                    current_edge = 0;
+                    current_level = neighbour_level;
+
+                    // and push a new component
+                    push_new_component = true;
+                    break;
+                }
+            }
+
+        } // else neigbor was already accessible
+
+        if (push_new_component) continue;
+
+
+        // once here we can add the current pixel to the component at the top of the stack
+        // but first we find how many of its neighbours are part of the region boundary (needed for
+        // perimeter and crossings calc.) and the increment in quads counts for euler number calc.
+        int non_boundary_neighbours = 0;
+        int non_boundary_neighbours_horiz = 0;
+
+        unsigned char quad_before[4] = {0,0,0,0};
+        unsigned char quad_after[4] = {0,0,0,0};
+        quad_after[0] = 1<<1;
+        quad_after[1] = 1<<3;
+        quad_after[2] = 1<<2;
+        quad_after[3] = 1;
+
+        for (int edge = 0; edge < 8; edge++)
+        {
+            int neighbour4 = -1;
+            int neighbour8 = -1;
+            int cell = 0;
+            switch (edge)
+            {
+                    case 0: if (x < width - 1) { neighbour4 = neighbour8 = current_pixel + 1;} cell = 5; break;
+                    case 1: if ((x < width - 1)&&(y < height - 1)) { neighbour8 = current_pixel + 1 + width;} cell = 8; break;
+                    case 2: if (y < height - 1) { neighbour4 = neighbour8 = current_pixel + width;} cell = 7; break;
+                    case 3: if ((x > 0)&&(y < height - 1)) { neighbour8 = current_pixel - 1 + width;} cell = 6; break;
+                    case 4: if (x > 0) { neighbour4 = neighbour8 = current_pixel - 1;} cell = 3; break;
+                    case 5: if ((x > 0)&&(y > 0)) { neighbour8 = current_pixel - 1 - width;} cell = 0; break;
+                    case 6: if (y > 0) { neighbour4 = neighbour8 = current_pixel - width;} cell = 1; break;
+                    default: if ((x < width - 1)&&(y > 0)) { neighbour8 = current_pixel + 1 - width;} cell = 2; break;
+            }
+            if ((neighbour4 != -1)&&(accumulated_pixel_mask[neighbour4])&&(image_data[neighbour4]<=image_data[current_pixel]))
+            {
+                non_boundary_neighbours++;
+                if ((edge == 0) || (edge == 4))
+                    non_boundary_neighbours_horiz++;
+            }
+
+            int pix_value = image_data[current_pixel] + 1;
+            if (neighbour8 != -1)
+            {
+                if (accumulated_pixel_mask[neighbour8])
+                    pix_value = image_data[neighbour8];
+            }
+
+            if (pix_value<=image_data[current_pixel])
+            {
+                switch(cell)
+                {
+                    case 0:
+                        quad_before[3] = quad_before[3] | (1<<3);
+                        quad_after[3]  = quad_after[3]  | (1<<3);
+                        break;
+                    case 1:
+                        quad_before[3] = quad_before[3] | (1<<2);
+                        quad_after[3]  = quad_after[3]  | (1<<2);
+                        quad_before[0] = quad_before[0] | (1<<3);
+                        quad_after[0]  = quad_after[0]  | (1<<3);
+                        break;
+                    case 2:
+                        quad_before[0] = quad_before[0] | (1<<2);
+                        quad_after[0]  = quad_after[0]  | (1<<2);
+                        break;
+                    case 3:
+                        quad_before[3] = quad_before[3] | (1<<1);
+                        quad_after[3]  = quad_after[3]  | (1<<1);
+                        quad_before[2] = quad_before[2] | (1<<3);
+                        quad_after[2]  = quad_after[2]  | (1<<3);
+                        break;
+                    case 5:
+                        quad_before[0] = quad_before[0] | (1);
+                        quad_after[0]  = quad_after[0]  | (1);
+                        quad_before[1] = quad_before[1] | (1<<2);
+                        quad_after[1]  = quad_after[1]  | (1<<2);
+                        break;
+                    case 6:
+                        quad_before[2] = quad_before[2] | (1<<1);
+                        quad_after[2]  = quad_after[2]  | (1<<1);
+                        break;
+                    case 7:
+                        quad_before[2] = quad_before[2] | (1);
+                        quad_after[2]  = quad_after[2]  | (1);
+                        quad_before[1] = quad_before[1] | (1<<1);
+                        quad_after[1]  = quad_after[1]  | (1<<1);
+                        break;
+                    default:
+                        quad_before[1] = quad_before[1] | (1);
+                        quad_after[1]  = quad_after[1]  | (1);
+                        break;
+                }
+            }
+
+        }
+
+        int C_before[3] = {0, 0, 0};
+        int C_after[3] = {0, 0, 0};
+
+        for (int p=0; p<3; p++)
+        {
+            for (int q=0; q<4; q++)
+            {
+                if ( (quad_before[0] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_before[p]++;
+                if ( (quad_before[1] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_before[p]++;
+                if ( (quad_before[2] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_before[p]++;
+                if ( (quad_before[3] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_before[p]++;
+
+                if ( (quad_after[0] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_after[p]++;
+                if ( (quad_after[1] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_after[p]++;
+                if ( (quad_after[2] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_after[p]++;
+                if ( (quad_after[3] == quads[p][q]) && ((p<2)||(q<2)) )
+                    C_after[p]++;
+            }
+        }
+
+        int d_C1 = C_after[0]-C_before[0];
+        int d_C2 = C_after[1]-C_before[1];
+        int d_C3 = C_after[2]-C_before[2];
+
+        er_add_pixel(er_stack.back(), x, y, non_boundary_neighbours, non_boundary_neighbours_horiz, d_C1, d_C2, d_C3);
+        accumulated_pixel_mask[current_pixel] = true;
+
+        // if we have processed all the possible threshold levels (the hea is empty) we are done!
+        if (threshold_level == (255/thresholdDelta)+1)
+        {
+
+            // save the extracted regions into the output vector
+            regions->reserve(num_accepted_regions+1);
+            er_save(er_stack.back(), NULL, NULL);
+
+            // clean memory
+            er_tree_clean(er_stack.back());
+            er_stack.clear();
+
+            return;
+        }
+
+
+        // pop the heap of boundary pixels
+        current_pixel = boundary_pixes[threshold_level].back();
+        boundary_pixes[threshold_level].erase(boundary_pixes[threshold_level].end()-1);
+        current_edge  = boundary_edges[threshold_level].back();
+        boundary_edges[threshold_level].erase(boundary_edges[threshold_level].end()-1);
+
+        while (boundary_pixes[threshold_level].empty() && (threshold_level < (255/thresholdDelta)+1))
+            threshold_level++;
+
+
+        int new_level = image_data[current_pixel];
+
+        // if the new pixel has higher grey value than the current one
+        if (new_level != current_level) {
+
+            current_level = new_level;
+
+            // process components on the top of the stack until we reach the higher grey-level
+            while (er_stack.back()->level < new_level)
+            {
+                ERStat* er = er_stack.back();
+                er_stack.erase(er_stack.end()-1);
+
+                if (new_level < er_stack.back()->level)
+                {
+                    er_stack.push_back(new ERStat(new_level, current_pixel, current_pixel%width, current_pixel/width));
+                    er_merge(er_stack.back(), er);
+                    break;
+                }
+
+                er_merge(er_stack.back(), er);
+            }
+
+        }
+
+    }
+}
+
+// accumulate a pixel into an ER
+void ERFilterNM::er_add_pixel(ERStat *parent, int x, int y, int non_border_neighbours,
+                                                            int non_border_neighbours_horiz,
+                                                            int d_C1, int d_C2, int d_C3)
+{
+    parent->area++;
+    parent->perimeter += 4 - 2*non_border_neighbours;
+
+    if (parent->crossings->size()>0)
+    {
+        if (y<parent->rect.y) parent->crossings->push_front(2);
+        else if (y>parent->rect.br().y-1) parent->crossings->push_back(2);
+        else {
+            parent->crossings->at(y - parent->rect.y) += 2-2*non_border_neighbours_horiz;
+        }
+    } else {
+        parent->crossings->push_back(2);
+    }
+
+    parent->euler += (d_C1 - d_C2 + 2*d_C3) / 4;
+
+    int new_x1 = min(parent->rect.x,x);
+    int new_y1 = min(parent->rect.y,y);
+    int new_x2 = max(parent->rect.br().x-1,x);
+    int new_y2 = max(parent->rect.br().y-1,y);
+    parent->rect.x = new_x1;
+    parent->rect.y = new_y1;
+    parent->rect.width  = new_x2-new_x1+1;
+    parent->rect.height = new_y2-new_y1+1;
+
+    parent->raw_moments[0] += x;
+    parent->raw_moments[1] += y;
+
+    parent->central_moments[0] += x * x;
+    parent->central_moments[1] += x * y;
+    parent->central_moments[2] += y * y;
+}
+
+// merge an ER with its nested parent
+void ERFilterNM::er_merge(ERStat *parent, ERStat *child)
+{
+
+    parent->area += child->area;
+
+    parent->perimeter += child->perimeter;
+
+
+    for (int i=parent->rect.y; i<=min(parent->rect.br().y-1,child->rect.br().y-1); i++)
+        if (i-child->rect.y >= 0)
+            parent->crossings->at(i-parent->rect.y) += child->crossings->at(i-child->rect.y);
+
+    for (int i=parent->rect.y-1; i>=child->rect.y; i--)
+        if (i-child->rect.y < (int)child->crossings->size())
+            parent->crossings->push_front(child->crossings->at(i-child->rect.y));
+        else
+            parent->crossings->push_front(0);
+
+    for (int i=parent->rect.br().y; i<child->rect.y; i++)
+        parent->crossings->push_back(0);
+
+    for (int i=max(parent->rect.br().y,child->rect.y); i<=child->rect.br().y-1; i++)
+        parent->crossings->push_back(child->crossings->at(i-child->rect.y));
+
+    parent->euler += child->euler;
+
+    int new_x1 = min(parent->rect.x,child->rect.x);
+    int new_y1 = min(parent->rect.y,child->rect.y);
+    int new_x2 = max(parent->rect.br().x-1,child->rect.br().x-1);
+    int new_y2 = max(parent->rect.br().y-1,child->rect.br().y-1);
+    parent->rect.x = new_x1;
+    parent->rect.y = new_y1;
+    parent->rect.width  = new_x2-new_x1+1;
+    parent->rect.height = new_y2-new_y1+1;
+
+    parent->raw_moments[0] += child->raw_moments[0];
+    parent->raw_moments[1] += child->raw_moments[1];
+
+    parent->central_moments[0] += child->central_moments[0];
+    parent->central_moments[1] += child->central_moments[1];
+    parent->central_moments[2] += child->central_moments[2];
+
+    vector<int> m_crossings;
+    m_crossings.push_back(child->crossings->at((int)(child->rect.height)/6));
+    m_crossings.push_back(child->crossings->at((int)3*(child->rect.height)/6));
+    m_crossings.push_back(child->crossings->at((int)5*(child->rect.height)/6));
+    std::sort(m_crossings.begin(), m_crossings.end());
+    child->med_crossings = (float)m_crossings.at(1);
+
+    // free unnecessary mem
+    child->crossings->clear();
+    delete(child->crossings);
+    child->crossings = NULL;
+
+    // recover the original grey-level
+    child->level = child->level*thresholdDelta;
+
+    // before saving calculate P(child|character) and filter if possible
+    if (classifier != NULL)
+    {
+        child->probability = classifier->eval(*child);
+    }
+
+    if ( (((classifier!=NULL)?(child->probability >= minProbability):true)||(nonMaxSuppression)) &&
+         ((child->area >= (minArea*region_mask.rows*region_mask.cols)) &&
+          (child->area <= (maxArea*region_mask.rows*region_mask.cols)) &&
+          (child->rect.width > 2) && (child->rect.height > 2)) )
+    {
+
+        num_accepted_regions++;
+
+        child->next = parent->child;
+        if (parent->child)
+            parent->child->prev = child;
+        parent->child = child;
+        child->parent = parent;
+
+    } else {
+
+        num_rejected_regions++;
+
+        if (child->prev !=NULL)
+            child->prev->next = child->next;
+
+        ERStat *new_child = child->child;
+        if (new_child != NULL)
+        {
+            while (new_child->next != NULL)
+                new_child = new_child->next;
+            new_child->next = parent->child;
+            if (parent->child)
+                parent->child->prev = new_child;
+            parent->child   = child->child;
+            child->child->parent = parent;
+        }
+
+        // free mem
+        if(child->crossings)
+        {
+            child->crossings->clear();
+            delete(child->crossings);
+            child->crossings = NULL;
+        }
+        delete(child);
+    }
+
+}
+
+// recursively walk the tree and clean memory
+void ERFilterNM::er_tree_clean( ERStat *stat )
+{
+        for (ERStat * child = stat->child; child; child = child->next)
+        {
+            er_tree_clean(child);
+        }
+        if (stat->crossings)
+        {
+            stat->crossings->clear();
+            delete(stat->crossings);
+            stat->crossings = NULL;
+        }
+        delete stat;
+}
+
+// copy extracted regions into the output vector
+ERStat* ERFilterNM::er_save( ERStat *er, ERStat *parent, ERStat *prev )
+{
+
+    regions->push_back(*er);
+
+    regions->back().parent = parent;
+    if (prev != NULL)
+    {
+      prev->next = &(regions->back());
+    }
+    else if (parent != NULL)
+      parent->child = &(regions->back());
+
+    ERStat *old_prev = NULL;
+    ERStat *this_er  = &regions->back();
+
+    if (this_er->parent == NULL)
+    {
+       this_er->probability = 0;
+    }
+
+    if (nonMaxSuppression)
+    {
+        if (this_er->parent == NULL)
+        {
+            this_er->max_probability_ancestor = this_er;
+            this_er->min_probability_ancestor = this_er;
+        }
+        else
+        {
+            this_er->max_probability_ancestor = (this_er->probability > parent->max_probability_ancestor->probability)? this_er :  parent->max_probability_ancestor;
+
+            this_er->min_probability_ancestor = (this_er->probability < parent->min_probability_ancestor->probability)? this_er :  parent->min_probability_ancestor;
+
+            if ( (this_er->max_probability_ancestor->probability > minProbability) && (this_er->max_probability_ancestor->probability - this_er->min_probability_ancestor->probability > minProbabilityDiff))
+            {
+              this_er->max_probability_ancestor->local_maxima = true;
+              if ((this_er->max_probability_ancestor == this_er) && (this_er->parent->local_maxima))
+              {
+                this_er->parent->local_maxima = false;
+              }
+            }
+            else if (this_er->probability < this_er->parent->probability)
+            {
+              this_er->min_probability_ancestor = this_er;
+            }
+            else if (this_er->probability > this_er->parent->probability)
+            {
+              this_er->max_probability_ancestor = this_er;
+            }
+
+
+        }
+    }
+
+    for (ERStat * child = er->child; child; child = child->next)
+    {
+        old_prev = er_save(child, this_er, old_prev);
+    }
+
+    return this_er;
+}
+
+// recursively walk the tree and filter (remove) regions using the callback classifier
+ERStat* ERFilterNM::er_tree_filter ( InputArray image, ERStat * stat, ERStat *parent, ERStat *prev )
+{
+    Mat src = image.getMat();
+    // assert correct image type
+    CV_Assert( src.type() == CV_8UC1 );
+
+    //Fill the region and calculate 2nd stage features
+    Mat region = region_mask(Rect(Point(stat->rect.x,stat->rect.y),Point(stat->rect.br().x+2,stat->rect.br().y+2)));
+    region = Scalar(0);
+    int newMaskVal = 255;
+    int flags = 4 + (newMaskVal << 8) + FLOODFILL_FIXED_RANGE + FLOODFILL_MASK_ONLY;
+    Rect rect;
+
+    floodFill( src(Rect(Point(stat->rect.x,stat->rect.y),Point(stat->rect.br().x,stat->rect.br().y))),
+               region, Point(stat->pixel%src.cols - stat->rect.x, stat->pixel/src.cols - stat->rect.y),
+               Scalar(255), &rect, Scalar(stat->level), Scalar(0), flags );
+    rect.width += 2;
+    rect.height += 2;
+    region = region(rect);
+
+    vector<vector<Point> > contours;
+    vector<Point> contour_poly;
+    vector<Vec4i> hierarchy;
+    findContours( region, contours, hierarchy, RETR_TREE, CHAIN_APPROX_NONE, Point(0, 0) );
+    //TODO check epsilon parameter of approxPolyDP (set empirically) : we want more precission
+    //     if the region is very small because otherwise we'll loose all the convexities
+    approxPolyDP( Mat(contours[0]), contour_poly, (float)min(rect.width,rect.height)/17, true );
+
+    bool was_convex = false;
+    int  num_inflexion_points = 0;
+
+    for (int p = 0 ; p<(int)contour_poly.size(); p++)
+    {
+        int p_prev = p-1;
+        int p_next = p+1;
+        if (p_prev == -1)
+            p_prev = contour_poly.size()-1;
+        if (p_next == (int)contour_poly.size())
+            p_next = 0;
+
+        double angle_next = atan2((double)(contour_poly[p_next].y-contour_poly[p].y),
+                                  (double)(contour_poly[p_next].x-contour_poly[p].x));
+        double angle_prev = atan2((double)(contour_poly[p_prev].y-contour_poly[p].y),
+                                  (double)(contour_poly[p_prev].x-contour_poly[p].x));
+        if ( angle_next < 0 )
+            angle_next = 2.*CV_PI + angle_next;
+
+        double angle = (angle_next - angle_prev);
+        if (angle > 2.*CV_PI)
+            angle = angle - 2.*CV_PI;
+        else if (angle < 0)
+            angle = 2.*CV_PI + std::abs(angle);
+
+        if (p>0)
+        {
+            if ( ((angle > CV_PI)&&(!was_convex)) || ((angle < CV_PI)&&(was_convex)) )
+                num_inflexion_points++;
+        }
+        was_convex = (angle > CV_PI);
+
+    }
+
+    floodFill(region, Point(0,0), Scalar(255), 0);
+    int holes_area = region.cols*region.rows-countNonZero(region);
+
+    int hull_area = 0;
+
+    {
+
+        vector<Point> hull;
+        convexHull(contours[0], hull, false);
+        hull_area = (int)contourArea(hull);
+    }
+
+
+    stat->hole_area_ratio = (float)holes_area / stat->area;
+    stat->convex_hull_ratio = (float)hull_area / (float)contourArea(contours[0]);
+    stat->num_inflexion_points = (float)num_inflexion_points;
+
+
+    // calculate P(child|character) and filter if possible
+    if ( (classifier != NULL) && (stat->parent != NULL) )
+    {
+        stat->probability = classifier->eval(*stat);
+    }
+
+    if ( ( ((classifier != NULL)?(stat->probability >= minProbability):true) &&
+          ((stat->area >= minArea*region_mask.rows*region_mask.cols) &&
+           (stat->area <= maxArea*region_mask.rows*region_mask.cols)) ) ||
+        (stat->parent == NULL) )
+    {
+
+        num_accepted_regions++;
+        regions->push_back(*stat);
+
+        regions->back().parent = parent;
+        regions->back().next   = NULL;
+        regions->back().child  = NULL;
+
+        if (prev != NULL)
+            prev->next = &(regions->back());
+        else if (parent != NULL)
+            parent->child = &(regions->back());
+
+        ERStat *old_prev = NULL;
+        ERStat *this_er  = &regions->back();
+
+        for (ERStat * child = stat->child; child; child = child->next)
+        {
+            old_prev = er_tree_filter(image, child, this_er, old_prev);
+        }
+
+        return this_er;
+
+    } else {
+
+        num_rejected_regions++;
+
+        ERStat *old_prev = prev;
+
+        for (ERStat * child = stat->child; child; child = child->next)
+        {
+            old_prev = er_tree_filter(image, child, parent, old_prev);
+        }
+
+        return old_prev;
+    }
+
+}
+
+// recursively walk the tree selecting only regions with local maxima probability
+ERStat* ERFilterNM::er_tree_nonmax_suppression ( ERStat * stat, ERStat *parent, ERStat *prev )
+{
+
+    if ( ( stat->local_maxima ) || ( stat->parent == NULL ) )
+    {
+
+        regions->push_back(*stat);
+
+        regions->back().parent = parent;
+        regions->back().next   = NULL;
+        regions->back().child  = NULL;
+
+        if (prev != NULL)
+            prev->next = &(regions->back());
+        else if (parent != NULL)
+            parent->child = &(regions->back());
+
+        ERStat *old_prev = NULL;
+        ERStat *this_er  = &regions->back();
+
+        for (ERStat * child = stat->child; child; child = child->next)
+        {
+            old_prev = er_tree_nonmax_suppression( child, this_er, old_prev );
+        }
+
+        return this_er;
+
+    } else {
+
+        num_rejected_regions++;
+        num_accepted_regions--;
+
+        ERStat *old_prev = prev;
+
+        for (ERStat * child = stat->child; child; child = child->next)
+        {
+            old_prev = er_tree_nonmax_suppression( child, parent, old_prev );
+        }
+
+        return old_prev;
+    }
+
+}
+
+void ERFilterNM::setCallback(const Ptr<ERFilter::Callback>& cb)
+{
+    classifier = cb;
+};
+
+void ERFilterNM::setMinArea(float _minArea)
+{
+    CV_Assert( (_minArea >= 0) && (_minArea < maxArea) );
+    minArea = _minArea;
+    return;
+};
+
+void ERFilterNM::setMaxArea(float _maxArea)
+{
+    CV_Assert(_maxArea <= 1);
+    CV_Assert(minArea < _maxArea);
+    maxArea = _maxArea;
+    return;
+};
+
+void ERFilterNM::setThresholdDelta(int _thresholdDelta)
+{
+    CV_Assert( (_thresholdDelta > 0) && (_thresholdDelta <= 128) );
+    thresholdDelta = _thresholdDelta;
+    return;
+};
+
+void ERFilterNM::setMinProbability(float _minProbability)
+{
+    CV_Assert( (_minProbability >= 0.0) && (_minProbability <= 1.0) );
+    minProbability = _minProbability;
+    return;
+};
+
+void ERFilterNM::setMinProbabilityDiff(float _minProbabilityDiff)
+{
+    CV_Assert( (_minProbabilityDiff >= 0.0) && (_minProbabilityDiff <= 1.0) );
+    minProbabilityDiff = _minProbabilityDiff;
+    return;
+};
+
+void ERFilterNM::setNonMaxSuppression(bool _nonMaxSuppression)
+{
+    nonMaxSuppression = _nonMaxSuppression;
+    return;
+};
+
+int ERFilterNM::getNumRejected()
+{
+    return num_rejected_regions;
+};
+
+
+
+
+// load default 1st stage classifier if found
+ERClassifierNM1::ERClassifierNM1()
+{
+
+    if (ifstream("./trained_classifierNM1.xml"))
+    {
+        // The file with default classifier exists
+        boost.load("./trained_classifierNM1.xml", "boost");
+    }
+    else if (ifstream("./training/trained_classifierNM1.xml"))
+    {
+        // The file with default classifier exists
+        boost.load("./training/trained_classifierNM1.xml", "boost");
+    }
+    else
+    {
+        // File not found
+        CV_Error(CV_StsBadArg, "Default classifier ./trained_classifierNM1.xml not found!");
+    }
+};
+
+double ERClassifierNM1::eval(const ERStat& stat)
+{
+    //Classify
+    float arr[] = {0,(float)(stat.rect.width)/(stat.rect.height), // aspect ratio
+                     sqrt((float)(stat.area))/stat.perimeter, // compactness
+                     (float)(1-stat.euler), //number of holes
+                     stat.med_crossings};
+
+    vector<float> sample (arr, arr + sizeof(arr) / sizeof(arr[0]) );
+
+    float votes = boost.predict( Mat(sample), Mat(), Range::all(), false, true );
+
+    // Logistic Correction returns a probability value (in the range(0,1))
+    return (double)1-(double)1/(1+exp(-2*votes));
+};
+
+
+// load default 2nd stage classifier if found
+ERClassifierNM2::ERClassifierNM2()
+{
+
+    if (ifstream("./trained_classifierNM2.xml"))
+    {
+        // The file with default classifier exists
+        boost.load("./trained_classifierNM2.xml", "boost");
+    }
+    else if (ifstream("./training/trained_classifierNM2.xml"))
+    {
+        // The file with default classifier exists
+        boost.load("./training/trained_classifierNM2.xml", "boost");
+    }
+    else
+    {
+        // File not found
+        CV_Error(CV_StsBadArg, "Default classifier ./trained_classifierNM2.xml not found!");
+    }
+};
+
+double ERClassifierNM2::eval(const ERStat& stat)
+{
+    //Classify
+    float arr[] = {0,(float)(stat.rect.width)/(stat.rect.height), // aspect ratio
+                     sqrt((float)(stat.area))/stat.perimeter, // compactness
+                     (float)(1-stat.euler), //number of holes
+                     stat.med_crossings, stat.hole_area_ratio,
+                     stat.convex_hull_ratio, stat.num_inflexion_points};
+
+    vector<float> sample (arr, arr + sizeof(arr) / sizeof(arr[0]) );
+
+    float votes = boost.predict( Mat(sample), Mat(), Range::all(), false, true );
+
+    // Logistic Correction returns a probability value (in the range(0,1))
+    return (double)1-(double)1/(1+exp(-2*votes));
+};
+
+
+/*!
+    Create an Extremal Region Filter for the 1st stage classifier of N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    The component tree of the image is extracted by a threshold increased step by step
+    from 0 to 255, incrementally computable descriptors (aspect_ratio, compactness,
+    number of holes, and number of horizontal crossings) are computed for each ER
+    and used as features for a classifier which estimates the class-conditional
+    probability P(er|character). The value of P(er|character) is tracked using the inclusion
+    relation of ER across all thresholds and only the ERs which correspond to local maximum
+    of the probability P(er|character) are selected (if the local maximum of the
+    probability is above a global limit pmin and the difference between local maximum and
+    local minimum is greater than minProbabilityDiff).
+
+    \param  cb                Callback with the classifier.
+                              if omitted tries to load a default classifier from file trained_classifierNM1.xml
+    \param  thresholdDelta    Threshold step in subsequent thresholds when extracting the component tree
+    \param  minArea           The minimum area (% of image size) allowed for retreived ER's
+    \param  minArea           The maximum area (% of image size) allowed for retreived ER's
+    \param  minProbability    The minimum probability P(er|character) allowed for retreived ER's
+    \param  nonMaxSuppression Whenever non-maximum suppression is done over the branch probabilities
+    \param  minProbability    The minimum probability difference between local maxima and local minima ERs
+*/
+Ptr<ERFilter> createERFilterNM1(const Ptr<ERFilter::Callback>& cb, int thresholdDelta,
+                                float minArea, float maxArea, float minProbability,
+                                bool nonMaxSuppression, float minProbabilityDiff)
+{
+
+    CV_Assert( (minProbability >= 0.) && (minProbability <= 1.) );
+    CV_Assert( (minArea < maxArea) && (minArea >=0.) && (maxArea <= 1.) );
+    CV_Assert( (thresholdDelta >= 0) && (thresholdDelta <= 128) );
+    CV_Assert( (minProbabilityDiff >= 0.) && (minProbabilityDiff <= 1.) );
+
+    Ptr<ERFilterNM> filter = makePtr<ERFilterNM>();
+
+    if (cb == NULL)
+        filter->setCallback(makePtr<ERClassifierNM1>());
+    else
+        filter->setCallback(cb);
+
+    filter->setThresholdDelta(thresholdDelta);
+    filter->setMinArea(minArea);
+    filter->setMaxArea(maxArea);
+    filter->setMinProbability(minProbability);
+    filter->setNonMaxSuppression(nonMaxSuppression);
+    filter->setMinProbabilityDiff(minProbabilityDiff);
+    return (Ptr<ERFilter>)filter;
+}
+
+/*!
+    Create an Extremal Region Filter for the 2nd stage classifier of N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    In the second stage, the ERs that passed the first stage are classified into character
+    and non-character classes using more informative but also more computationally expensive
+    features. The classifier uses all the features calculated in the first stage and the following
+    additional features: hole area ratio, convex hull ratio, and number of outer inflexion points.
+
+    \param  cb             Callback with the classifier
+                           if omitted tries to load a default classifier from file trained_classifierNM2.xml
+    \param  minProbability The minimum probability P(er|character) allowed for retreived ER's
+*/
+Ptr<ERFilter> createERFilterNM2(const Ptr<ERFilter::Callback>& cb, float minProbability)
+{
+
+    CV_Assert( (minProbability >= 0.) && (minProbability <= 1.) );
+
+    Ptr<ERFilterNM> filter = makePtr<ERFilterNM>();
+
+    if (cb == NULL)
+        filter->setCallback(makePtr<ERClassifierNM2>());
+    else
+        filter->setCallback(cb);
+
+    filter->setMinProbability(minProbability);
+    return (Ptr<ERFilter>)filter;
+}
+
+
+/* ------------------------------------------------------------------------------------*/
+/* -------------------------------- Compute Channels NM -------------------------------*/
+/* ------------------------------------------------------------------------------------*/
+
+
+void  get_gradient_magnitude(Mat& _grey_img, Mat& _gradient_magnitude);
+
+void get_gradient_magnitude(Mat& _grey_img, Mat& _gradient_magnitude)
+{
+    Mat C = Mat_<float>(_grey_img);
+
+    Mat kernel = (Mat_<float>(1,3) << -1,0,1);
+    Mat grad_x;
+    filter2D(C, grad_x, -1, kernel, Point(-1,-1), 0, BORDER_DEFAULT);
+
+    Mat kernel2 = (Mat_<float>(3,1) << -1,0,1);
+    Mat grad_y;
+    filter2D(C, grad_y, -1, kernel2, Point(-1,-1), 0, BORDER_DEFAULT);
+
+    magnitude( grad_x, grad_y, _gradient_magnitude);
+}
+
+
+/*!
+    Compute the diferent channels to be processed independently in the N&M algorithm
+    Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+
+    In N&M algorithm, the combination of intensity (I), hue (H), saturation (S), and gradient
+    magnitude channels (Grad) are used in order to obatin high localization recall.
+    This implementation also the alternative combination of red (R), grren (G), blue (B),
+    lightness (L), and gradient magnitude (Grad).
+
+    \param  _src           Source image. Must be RGB CV_8UC3.
+    \param  _channels      Output vector<Mat> where computed channels are stored.
+    \param  _mode          Mode of operation. Currently the only available options are
+                           ERFILTER_NM_RGBLGrad and ERFILTER_NM_IHSGrad.
+
+*/
+void computeNMChannels(InputArray _src, OutputArrayOfArrays _channels, int _mode)
+{
+
+    CV_Assert( ( _mode == ERFILTER_NM_RGBLGrad ) || ( _mode == ERFILTER_NM_IHSGrad ) );
+
+    Mat src = _src.getMat();
+    if( src.empty() )
+    {
+        _channels.release();
+        return;
+    }
+
+    // assert RGB image
+    CV_Assert(src.type() == CV_8UC3);
+
+    if (_mode == ERFILTER_NM_IHSGrad)
+    {
+        _channels.create( 4, 1, src.depth());
+
+        Mat hsv;
+        cvtColor(src, hsv, COLOR_RGB2HSV);
+        vector<Mat> channelsHSV;
+        split(hsv, channelsHSV);
+
+        for (int i = 0; i < src.channels(); i++)
+        {
+            _channels.create(src.rows, src.cols, CV_8UC1, i);
+            Mat channel = _channels.getMat(i);
+            channelsHSV.at(i).copyTo(channel);
+        }
+
+        Mat grey;
+        cvtColor(src, grey, COLOR_RGB2GRAY);
+        Mat gradient_magnitude = Mat_<float>(grey.size());
+        get_gradient_magnitude( grey, gradient_magnitude);
+        gradient_magnitude.convertTo(gradient_magnitude, CV_8UC1);
+
+        _channels.create(src.rows, src.cols, CV_8UC1, 3);
+        Mat channelGrad = _channels.getMat(3);
+        gradient_magnitude.copyTo(channelGrad);
+
+    } else if (_mode == ERFILTER_NM_RGBLGrad) {
+
+        _channels.create( 5, 1, src.depth());
+
+        vector<Mat> channelsRGB;
+        split(src, channelsRGB);
+        for (int i = 0; i < src.channels(); i++)
+        {
+            _channels.create(src.rows, src.cols, CV_8UC1, i);
+            Mat channel = _channels.getMat(i);
+            channelsRGB.at(i).copyTo(channel);
+        }
+
+        Mat hls;
+        cvtColor(src, hls, COLOR_RGB2HLS);
+        vector<Mat> channelsHLS;
+        split(hls, channelsHLS);
+
+        _channels.create(src.rows, src.cols, CV_8UC1, 3);
+        Mat channelL = _channels.getMat(3);
+        channelsHLS.at(1).copyTo(channelL);
+
+        Mat grey;
+        cvtColor(src, grey, COLOR_RGB2GRAY);
+        Mat gradient_magnitude = Mat_<float>(grey.size());
+        get_gradient_magnitude( grey, gradient_magnitude);
+        gradient_magnitude.convertTo(gradient_magnitude, CV_8UC1);
+
+        _channels.create(src.rows, src.cols, CV_8UC1, 4);
+        Mat channelGrad = _channels.getMat(4);
+        gradient_magnitude.copyTo(channelGrad);
+    }
+}
+}
diff --git a/modules/objdetect/src/featurepyramid.cpp b/modules/objdetect/src/featurepyramid.cpp
index c6bc6c0e8a..35cda95394 100644
--- a/modules/objdetect/src/featurepyramid.cpp
+++ b/modules/objdetect/src/featurepyramid.cpp
@@ -1,4 +1,5 @@
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 #include "_latentsvm.h"
 #include "_lsvm_resizeimg.h"
 
@@ -511,4 +512,4 @@ int getFeaturePyramid(IplImage * image, CvLSVMFeaturePyramid **maps)
     }
 
     return LATENT_SVM_OK;
-}
\ No newline at end of file
+}
diff --git a/modules/objdetect/src/fft.cpp b/modules/objdetect/src/fft.cpp
index 6db38a214f..5c5902f001 100644
--- a/modules/objdetect/src/fft.cpp
+++ b/modules/objdetect/src/fft.cpp
@@ -244,4 +244,3 @@ int fftInverse2d(float *x_in, float *x_out, int numRows, int numColls)
     free(x_outTmp);
     return FFT_OK;
 }
-
diff --git a/modules/objdetect/src/haar.cpp b/modules/objdetect/src/haar.cpp
index 4166162739..cbb60b0917 100644
--- a/modules/objdetect/src/haar.cpp
+++ b/modules/objdetect/src/haar.cpp
@@ -42,8 +42,9 @@
 /* Haar features calculation */
 
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
 #include <stdio.h>
-#include "opencv2/core/internal.hpp"
 
 #if CV_SSE2
 #   if 1 /*!CV_SSE4_1 && !CV_SSE4_2*/
@@ -1535,15 +1536,15 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
         maxSize.width = img->cols;
     }
 
-    temp = cvCreateMat( img->rows, img->cols, CV_8UC1 );
-    sum = cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 );
-    sqsum = cvCreateMat( img->rows + 1, img->cols + 1, CV_64FC1 );
+    temp.reset(cvCreateMat( img->rows, img->cols, CV_8UC1 ));
+    sum.reset(cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 ));
+    sqsum.reset(cvCreateMat( img->rows + 1, img->cols + 1, CV_64FC1 ));
 
     if( !cascade->hid_cascade )
         icvCreateHidHaarClassifierCascade(cascade);
 
     if( cascade->hid_cascade->has_tilted_features )
-        tilted = cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 );
+        tilted.reset(cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 ));
 
     result_seq = cvCreateSeq( 0, sizeof(CvSeq), sizeof(CvAvgComp), storage );
 
@@ -1565,18 +1566,18 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
         if( use_ipp )
             normImg = cvCreateMat( img->rows, img->cols, CV_32FC1 );
 #endif
-        imgSmall = cvCreateMat( img->rows + 1, img->cols + 1, CV_8UC1 );
+        imgSmall.reset(cvCreateMat( img->rows + 1, img->cols + 1, CV_8UC1 ));
 
         for( factor = 1; ; factor *= scaleFactor )
         {
-            CvSize winSize = { cvRound(winSize0.width*factor),
-                                cvRound(winSize0.height*factor) };
-            CvSize sz = { cvRound( img->cols/factor ), cvRound( img->rows/factor ) };
-            CvSize sz1 = { sz.width - winSize0.width + 1, sz.height - winSize0.height + 1 };
+            CvSize winSize(cvRound(winSize0.width*factor),
+                                cvRound(winSize0.height*factor));
+            CvSize sz(cvRound( img->cols/factor ), cvRound( img->rows/factor ));
+            CvSize sz1(sz.width - winSize0.width + 1, sz.height - winSize0.height + 1);
 
-            CvRect equRect = { icv_object_win_border, icv_object_win_border,
+            CvRect equRect(icv_object_win_border, icv_object_win_border,
                 winSize0.width - icv_object_win_border*2,
-                winSize0.height - icv_object_win_border*2 };
+                winSize0.height - icv_object_win_border*2);
 
             CvMat img1, sum1, sqsum1, norm1, tilted1, mask1;
             CvMat* _tilted = 0;
@@ -1611,17 +1612,17 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
             if( use_ipp )
             {
                 cv::Mat fsum(sum1.rows, sum1.cols, CV_32F, sum1.data.ptr, sum1.step);
-                cv::Mat(&sum1).convertTo(fsum, CV_32F, 1, -(1<<24));
+                cv::cvarrToMat(&sum1).convertTo(fsum, CV_32F, 1, -(1<<24));
             }
             else
 #endif
                 cvSetImagesForHaarClassifierCascade( cascade, &sum1, &sqsum1, _tilted, 1. );
 
-            cv::Mat _norm1(&norm1), _mask1(&mask1);
+            cv::Mat _norm1 = cv::cvarrToMat(&norm1), _mask1 = cv::cvarrToMat(&mask1);
             cv::parallel_for_(cv::Range(0, stripCount),
                          cv::HaarDetectObjects_ScaleImage_Invoker(cascade,
                                 (((sz1.height + stripCount - 1)/stripCount + ystep-1)/ystep)*ystep,
-                                factor, cv::Mat(&sum1), cv::Mat(&sqsum1), &_norm1, &_mask1,
+                                factor, cv::cvarrToMat(&sum1), cv::cvarrToMat(&sqsum1), &_norm1, &_mask1,
                                 cv::Rect(equRect), allCandidates, rejectLevels, levelWeights, outputRejectLevels, &mtx));
         }
     }
@@ -1634,7 +1635,7 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
 
         if( doCannyPruning )
         {
-            sumcanny = cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 );
+            sumcanny.reset(cvCreateMat( img->rows + 1, img->cols + 1, CV_32SC1 ));
             cvCanny( img, temp, 0, 50, 3 );
             cvIntegral( temp, sumcanny );
         }
@@ -1656,9 +1657,9 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
         for( ; n_factors-- > 0; factor *= scaleFactor )
         {
             const double ystep = std::max( 2., factor );
-            CvSize winSize = { cvRound( cascade->orig_window_size.width * factor ),
-                                cvRound( cascade->orig_window_size.height * factor )};
-            CvRect equRect = { 0, 0, 0, 0 };
+            CvSize winSize(cvRound( cascade->orig_window_size.width * factor ),
+                                cvRound( cascade->orig_window_size.height * factor ));
+            CvRect equRect;
             int *p[4] = {0,0,0,0};
             int *pq[4] = {0,0,0,0};
             int startX = 0, startY = 0;
@@ -1775,7 +1776,7 @@ cvHaarDetectObjectsForROC( const CvArr* _img,
 
     if( findBiggestObject && rectList.size() )
     {
-        CvAvgComp result_comp = {{0,0,0,0},0};
+        CvAvgComp result_comp = {CvRect(),0};
 
         for( size_t i = 0; i < rectList.size(); i++ )
         {
@@ -1964,10 +1965,10 @@ cvLoadHaarClassifierCascade( const char* directory, CvSize orig_window_size )
 
     size += (n+1)*sizeof(char*);
     const char** input_cascade = (const char**)cvAlloc( size );
-    
+
     if( !input_cascade )
       CV_Error( CV_StsNoMem, "Could not allocate memory for input_cascade" );
-      
+
     char* ptr = (char*)(input_cascade + n + 1);
 
     for( int i = 0; i < n; i++ )
@@ -1988,7 +1989,7 @@ cvLoadHaarClassifierCascade( const char* directory, CvSize orig_window_size )
     }
 
     input_cascade[n] = 0;
-    
+
     CvHaarClassifierCascade* cascade = icvLoadCascadeCART( input_cascade, n, orig_window_size );
 
     if( input_cascade )
diff --git a/modules/objdetect/src/hog.cpp b/modules/objdetect/src/hog.cpp
index 676d6f4b81..5cc7f6a61b 100644
--- a/modules/objdetect/src/hog.cpp
+++ b/modules/objdetect/src/hog.cpp
@@ -41,6 +41,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/core/core_c.h"
 
 #include <cstdio>
 #include <iterator>
@@ -121,7 +122,7 @@ bool HOGDescriptor::read(FileNode& obj)
     return true;
 }
 
-void HOGDescriptor::write(FileStorage& fs, const std::string& objName) const
+void HOGDescriptor::write(FileStorage& fs, const String& objName) const
 {
     if( !objName.empty() )
         fs << objName;
@@ -143,14 +144,14 @@ void HOGDescriptor::write(FileStorage& fs, const std::string& objName) const
     fs << "}";
 }
 
-bool HOGDescriptor::load(const std::string& filename, const std::string& objname)
+bool HOGDescriptor::load(const String& filename, const String& objname)
 {
     FileStorage fs(filename, FileStorage::READ);
     FileNode obj = !objname.empty() ? fs[objname] : fs.getFirstTopLevelNode();
     return read(obj);
 }
 
-void HOGDescriptor::save(const std::string& filename, const std::string& objName) const
+void HOGDescriptor::save(const String& filename, const String& objName) const
 {
     FileStorage fs(filename, FileStorage::WRITE);
     write(fs, !objName.empty() ? objName : FileStorage::getDefaultObjectName(filename));
@@ -362,27 +363,27 @@ void HOGDescriptor::computeGradient(const Mat& img, Mat& grad, Mat& qangle,
         __m128 fhalf = _mm_set1_ps(0.5f), fzero = _mm_setzero_ps();
         __m128 _angleScale = _mm_set1_ps(angleScale), fone = _mm_set1_ps(1.0f);
         __m128i ione = _mm_set1_epi32(1), _nbins = _mm_set1_epi32(nbins), izero = _mm_setzero_si128();
-        
+
         for ( ; x <= width - 4; x += 4)
         {
             int x2 = x << 1;
             __m128 _mag = _mm_loadu_ps(dbuf + x + (width << 1));
             __m128 _angle = _mm_loadu_ps(dbuf + x + width * 3);
             _angle = _mm_sub_ps(_mm_mul_ps(_angleScale, _angle), fhalf);
-            
+
             __m128 sign = _mm_and_ps(fone, _mm_cmplt_ps(_angle, fzero));
             __m128i _hidx = _mm_cvttps_epi32(_angle);
             _hidx = _mm_sub_epi32(_hidx, _mm_cvtps_epi32(sign));
             _angle = _mm_sub_ps(_angle, _mm_cvtepi32_ps(_hidx));
-            
+
             __m128 ft0 = _mm_mul_ps(_mag, _mm_sub_ps(fone, _angle));
             __m128 ft1 = _mm_mul_ps(_mag, _angle);
             __m128 ft2 = _mm_unpacklo_ps(ft0, ft1);
             __m128 ft3 = _mm_unpackhi_ps(ft0, ft1);
-            
+
             _mm_storeu_ps(gradPtr + x2, ft2);
             _mm_storeu_ps(gradPtr + x2 + 4, ft3);
-            
+
             __m128i mask0 = _mm_sub_epi32(izero, _mm_srli_epi32(_hidx, 31));
             __m128i it0 = _mm_and_si128(mask0, _nbins);
             mask0 = _mm_cmplt_epi32(_hidx, _nbins);
@@ -428,7 +429,7 @@ struct HOGCache
         BlockData() :
             histOfs(0), imgOffset()
         { }
-        
+
         int histOfs;
         Point imgOffset;
     };
@@ -455,7 +456,7 @@ struct HOGCache
 
     const float* getBlock(Point pt, float* buf);
     virtual void normalizeBlockHistogram(float* histogram) const;
-    
+
     std::vector<PixData> pixData;
     std::vector<BlockData> blockData;
 
@@ -540,7 +541,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
         __m128i idx = _mm_loadu_si128((__m128i*)a);
         __m128 _bw = _mm_set1_ps(bw), _bh = _mm_set1_ps(bh);
         __m128i ifour = _mm_set1_epi32(4);
-        
+
         for (; i <= blockSize.height - 4; i += 4)
         {
             __m128 t = _mm_sub_ps(_mm_cvtepi32_ps(idx), _bh);
@@ -617,7 +618,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
             int icellX1 = icellX0 + 1, icellY1 = icellY0 + 1;
             cellX -= icellX0;
             cellY -= icellY0;
-            
+
             if( (unsigned)icellX0 < (unsigned)ncells.width &&
                (unsigned)icellX1 < (unsigned)ncells.width )
             {
@@ -657,7 +658,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
                     icellX1 = icellX0;
                     cellX = 1.f - cellX;
                 }
-                
+
                 if( (unsigned)icellY0 < (unsigned)ncells.height &&
                    (unsigned)icellY1 < (unsigned)ncells.height )
                 {
@@ -687,7 +688,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
             data->qangleOfs = (qangle.cols*i + j)*2;
             data->gradWeight = weights(i,j);
         }
-    
+
     assert( count1 + count2 + count4 == rawBlockSize );
     // defragment pixData
     for( j = 0; j < count2; j++ )
@@ -696,7 +697,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
         pixData[j + count1 + count2] = pixData[j + rawBlockSize*2];
     count2 += count1;
     count4 += count2;
-    
+
     // initialize blockData
     for( j = 0; j < nblocks.width; j++ )
         for( i = 0; i < nblocks.height; i++ )
@@ -744,9 +745,9 @@ const float* HOGCache::getBlock(Point pt, float* buf)
 
 //    CV_Assert( blockHist != 0 );
     memset(blockHist, 0, sizeof(float) * blockHistogramSize);
-    
+
     const PixData* _pixData = &pixData[0];
-    
+
     for( k = 0; k < C1; k++ )
     {
         const PixData& pk = _pixData[k];
@@ -754,7 +755,7 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         float w = pk.gradWeight*pk.histWeights[0];
         const uchar* h = qanglePtr + pk.qangleOfs;
         int h0 = h[0], h1 = h[1];
-        
+
         float* hist = blockHist + pk.histOfs[0];
         float t0 = hist[h0] + a[0]*w;
         float t1 = hist[h1] + a[1]*w;
@@ -769,19 +770,19 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         const float* const a = gradPtr + pk.gradOfs;
         const uchar* const h = qanglePtr + pk.qangleOfs;
         int h0 = h[0], h1 = h[1];
-        
+
         __m128 _a0 = _mm_set1_ps(a[0]), _a1 = _mm_set1_ps(a[1]);
         __m128 _w = _mm_mul_ps(_mm_set1_ps(pk.gradWeight), _mm_loadu_ps(pk.histWeights));
         __m128 _t0 = _mm_mul_ps(_a0, _w), _t1 = _mm_mul_ps(_a1, _w);
-        
+
         _mm_storeu_ps(hist0, _t0);
         _mm_storeu_ps(hist1, _t1);
-        
+
         float* hist = blockHist + pk.histOfs[0];
         float t0 = hist[h0] + hist0[0];
         float t1 = hist[h1] + hist1[0];
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[1];
         t0 = hist[h0] + hist0[1];
         t1 = hist[h1] + hist1[1];
@@ -795,13 +796,13 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         float w, t0, t1, a0 = a[0], a1 = a[1];
         const uchar* const h = qanglePtr + pk.qangleOfs;
         int h0 = h[0], h1 = h[1];
-        
+
         float* hist = blockHist + pk.histOfs[0];
         w = pk.gradWeight*pk.histWeights[0];
         t0 = hist[h0] + a0*w;
         t1 = hist[h1] + a1*w;
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[1];
         w = pk.gradWeight*pk.histWeights[1];
         t0 = hist[h0] + a0*w;
@@ -809,7 +810,7 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         hist[h0] = t0; hist[h1] = t1;
     }
 #endif
-    
+
 #if CV_SSE2
     for( ; k < C4; k++ )
     {
@@ -817,34 +818,34 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         const float* const a = gradPtr + pk.gradOfs;
         const uchar* const h = qanglePtr + pk.qangleOfs;
         int h0 = h[0], h1 = h[1];
-        
+
         __m128 _a0 = _mm_set1_ps(a[0]), _a1 = _mm_set1_ps(a[1]);
-		__m128 _w = _mm_mul_ps(_mm_set1_ps(pk.gradWeight), _mm_loadu_ps(pk.histWeights));
+        __m128 _w = _mm_mul_ps(_mm_set1_ps(pk.gradWeight), _mm_loadu_ps(pk.histWeights));
         __m128 _t0 = _mm_mul_ps(_a0, _w), _t1 = _mm_mul_ps(_a1, _w);
-        
+
         _mm_storeu_ps(hist0, _t0);
         _mm_storeu_ps(hist1, _t1);
-        
+
         float* hist = blockHist + pk.histOfs[0];
         float t0 = hist[h0] + hist0[0];
         float t1 = hist[h1] + hist1[0];
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[1];
         t0 = hist[h0] + hist0[1];
         t1 = hist[h1] + hist1[1];
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[2];
         t0 = hist[h0] + hist0[2];
         t1 = hist[h1] + hist1[2];
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[3];
         t0 = hist[h0] + hist0[3];
         t1 = hist[h1] + hist1[3];
         hist[h0] = t0; hist[h1] = t1;
-        
+
 //        __m128 _hist0 = _mm_set_ps((blockHist + pk.histOfs[3])[h0], (blockHist + pk.histOfs[2])[h0],
 //            (blockHist + pk.histOfs[1])[h0], (blockHist + pk.histOfs[0])[h0]);
 //        __m128 _hist1 = _mm_set_ps((blockHist + pk.histOfs[3])[h1], (blockHist + pk.histOfs[2])[h1],
@@ -860,7 +861,7 @@ const float* HOGCache::getBlock(Point pt, float* buf)
 //        (pk.histOfs[1] + blockHist)[h0] = hist0[1];
 //        (pk.histOfs[2] + blockHist)[h0] = hist0[2];
 //        (pk.histOfs[3] + blockHist)[h0] = hist0[3];
-//        
+//
 //        (pk.histOfs[0] + blockHist)[h1] = hist1[0];
 //        (pk.histOfs[1] + blockHist)[h1] = hist1[1];
 //        (pk.histOfs[2] + blockHist)[h1] = hist1[2];
@@ -874,25 +875,25 @@ const float* HOGCache::getBlock(Point pt, float* buf)
         float w, t0, t1, a0 = a[0], a1 = a[1];
         const uchar* h = qanglePtr + pk.qangleOfs;
         int h0 = h[0], h1 = h[1];
-        
+
         float* hist = blockHist + pk.histOfs[0];
         w = pk.gradWeight*pk.histWeights[0];
         t0 = hist[h0] + a0*w;
         t1 = hist[h1] + a1*w;
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[1];
         w = pk.gradWeight*pk.histWeights[1];
         t0 = hist[h0] + a0*w;
         t1 = hist[h1] + a1*w;
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[2];
         w = pk.gradWeight*pk.histWeights[2];
         t0 = hist[h0] + a0*w;
         t1 = hist[h1] + a1*w;
         hist[h0] = t0; hist[h1] = t1;
-        
+
         hist = blockHist + pk.histOfs[3];
         w = pk.gradWeight*pk.histWeights[3];
         t0 = hist[h0] + a0*w;
@@ -1015,6 +1016,19 @@ Rect HOGCache::getWindow(const Size& imageSize, const Size& winStride, int idx)
     return Rect( x*winStride.width, y*winStride.height, winSize.width, winSize.height );
 }
 
+static inline int gcd(int a, int b)
+{
+    if( a < b )
+        std::swap(a, b);
+    while( b > 0 )
+    {
+        int r = a % b;
+        a = b;
+        b = r;
+    }
+    return a;
+}
+
 void HOGDescriptor::compute(const Mat& img, std::vector<float>& descriptors,
     Size winStride, Size padding, const std::vector<Point>& locations) const
 {
@@ -1289,7 +1303,7 @@ void HOGDescriptor::detectMultiScale(
     if ( useMeanshiftGrouping )
         groupRectangles_meanshift(foundLocations, foundWeights, foundScales, finalThreshold, winSize);
     else
-        groupRectangles(foundLocations, (int)finalThreshold, 0.2);
+        groupRectangles(foundLocations, foundWeights, (int)finalThreshold, 0.2);
 }
 
 void HOGDescriptor::detectMultiScale(const Mat& img, std::vector<Rect>& foundLocations,
@@ -1301,6 +1315,57 @@ void HOGDescriptor::detectMultiScale(const Mat& img, std::vector<Rect>& foundLoc
                 padding, scale0, finalThreshold, useMeanshiftGrouping);
 }
 
+template<typename _ClsName> struct RTTIImpl
+{
+public:
+    static int isInstance(const void* ptr)
+    {
+        static _ClsName dummy;
+        static void* dummyp = &dummy;
+        union
+        {
+            const void* p;
+            const void** pp;
+        } a, b;
+        a.p = dummyp;
+        b.p = ptr;
+        return *a.pp == *b.pp;
+    }
+    static void release(void** dbptr)
+    {
+        if(dbptr && *dbptr)
+        {
+            delete (_ClsName*)*dbptr;
+            *dbptr = 0;
+        }
+    }
+    static void* read(CvFileStorage* fs, CvFileNode* n)
+    {
+        FileNode fn(fs, n);
+        _ClsName* obj = new _ClsName;
+        if(obj->read(fn))
+            return obj;
+        delete obj;
+        return 0;
+    }
+
+    static void write(CvFileStorage* _fs, const char* name, const void* ptr, CvAttrList)
+    {
+        if(ptr && _fs)
+        {
+            FileStorage fs(_fs, false);
+            ((const _ClsName*)ptr)->write(fs, String(name));
+        }
+    }
+
+    static void* clone(const void* ptr)
+    {
+        if(!ptr)
+            return 0;
+        return new _ClsName(*(const _ClsName*)ptr);
+    }
+};
+
 typedef RTTIImpl<HOGDescriptor> HOGRTTI;
 
 CvType hog_type( CV_TYPE_NAME_HOG_DESCRIPTOR, HOGRTTI::isInstance,
@@ -2788,30 +2853,30 @@ void HOGDescriptor::detectMultiScaleROI(const cv::Mat& img,
     cv::groupRectangles(foundLocations, groupThreshold, 0.2);
 }
 
-void HOGDescriptor::readALTModel(std::string modelfile)
+void HOGDescriptor::readALTModel(String modelfile)
 {
     // read model from SVMlight format..
     FILE *modelfl;
     if ((modelfl = fopen(modelfile.c_str(), "rb")) == NULL)
     {
-        std::string eerr("file not exist");
-        std::string efile(__FILE__);
-        std::string efunc(__FUNCTION__);
-        throw Exception(CV_StsError, eerr, efile, efunc, __LINE__);
+        String eerr("file not exist");
+        String efile(__FILE__);
+        String efunc(__FUNCTION__);
+        throw Exception(Error::StsError, eerr, efile, efunc, __LINE__);
     }
     char version_buffer[10];
     if (!fread (&version_buffer,sizeof(char),10,modelfl))
     {
-        std::string eerr("version?");
-        std::string efile(__FILE__);
-        std::string efunc(__FUNCTION__);
-        throw Exception(CV_StsError, eerr, efile, efunc, __LINE__);
+        String eerr("version?");
+        String efile(__FILE__);
+        String efunc(__FUNCTION__);
+        throw Exception(Error::StsError, eerr, efile, efunc, __LINE__);
     }
     if(strcmp(version_buffer,"V6.01")) {
-        std::string eerr("version doesnot match");
-        std::string efile(__FILE__);
-        std::string efunc(__FUNCTION__);
-        throw Exception(CV_StsError, eerr, efile, efunc, __LINE__);
+        String eerr("version doesnot match");
+        String efile(__FILE__);
+        String efunc(__FUNCTION__);
+        throw Exception(Error::StsError, eerr, efile, efunc, __LINE__);
     }
     /* read version number */
     int version = 0;
@@ -2819,9 +2884,9 @@ void HOGDescriptor::readALTModel(std::string modelfile)
     { throw Exception(); }
     if (version < 200)
     {
-        std::string eerr("version doesnot match");
-        std::string efile(__FILE__);
-        std::string efunc(__FUNCTION__);
+        String eerr("version doesnot match");
+        String efile(__FILE__);
+        String efunc(__FUNCTION__);
         throw Exception();
     }
     int kernel_type;
@@ -2871,12 +2936,89 @@ void HOGDescriptor::readALTModel(std::string modelfile)
 
         detector.push_back((float)-linearbias);
         setSVMDetector(detector);
-        delete linearwt;
+        delete [] linearwt;
     } else {
         throw Exception();
     }
     fclose(modelfl);
 }
 
-}
+void HOGDescriptor::groupRectangles(std::vector<cv::Rect>& rectList, std::vector<double>& weights, int groupThreshold, double eps) const
+{
+    if( groupThreshold <= 0 || rectList.empty() )
+    {
+        return;
+    }
 
+    CV_Assert(rectList.size() == weights.size());
+
+    std::vector<int> labels;
+    int nclasses = partition(rectList, labels, SimilarRects(eps));
+
+    std::vector<cv::Rect_<double> > rrects(nclasses);
+    std::vector<int> numInClass(nclasses, 0);
+    std::vector<double> foundWeights(nclasses, DBL_MIN);
+    std::vector<double> totalFactorsPerClass(nclasses, 1);
+    int i, j, nlabels = (int)labels.size();
+
+    for( i = 0; i < nlabels; i++ )
+    {
+        int cls = labels[i];
+        rrects[cls].x += rectList[i].x;
+        rrects[cls].y += rectList[i].y;
+        rrects[cls].width += rectList[i].width;
+        rrects[cls].height += rectList[i].height;
+        foundWeights[cls] = max(foundWeights[cls], weights[i]);
+        numInClass[cls]++;
+    }
+
+    for( i = 0; i < nclasses; i++ )
+    {
+        // find the average of all ROI in the cluster
+        cv::Rect_<double> r = rrects[i];
+        double s = 1.0/numInClass[i];
+        rrects[i] = cv::Rect_<double>(cv::saturate_cast<double>(r.x*s),
+            cv::saturate_cast<double>(r.y*s),
+            cv::saturate_cast<double>(r.width*s),
+            cv::saturate_cast<double>(r.height*s));
+    }
+
+    rectList.clear();
+    weights.clear();
+
+    for( i = 0; i < nclasses; i++ )
+    {
+        cv::Rect r1 = rrects[i];
+        int n1 = numInClass[i];
+        double w1 = foundWeights[i];
+        if( n1 <= groupThreshold )
+            continue;
+        // filter out small rectangles inside large rectangles
+        for( j = 0; j < nclasses; j++ )
+        {
+            int n2 = numInClass[j];
+
+            if( j == i || n2 <= groupThreshold )
+                continue;
+
+            cv::Rect r2 = rrects[j];
+
+            int dx = cv::saturate_cast<int>( r2.width * eps );
+            int dy = cv::saturate_cast<int>( r2.height * eps );
+
+            if( r1.x >= r2.x - dx &&
+                r1.y >= r2.y - dy &&
+                r1.x + r1.width <= r2.x + r2.width + dx &&
+                r1.y + r1.height <= r2.y + r2.height + dy &&
+                (n2 > std::max(3, n1) || n1 < 3) )
+                break;
+        }
+
+        if( j == nclasses )
+        {
+            rectList.push_back(r1);
+            weights.push_back(w1);
+        }
+    }
+}
+}
diff --git a/modules/objdetect/src/latentsvm.cpp b/modules/objdetect/src/latentsvm.cpp
index 521f0fdf56..613da1ceaa 100644
--- a/modules/objdetect/src/latentsvm.cpp
+++ b/modules/objdetect/src/latentsvm.cpp
@@ -389,7 +389,7 @@ int showRootFilterBoxes(IplImage *image,
                     color, thickness, line_type, shift);
     }
 #ifdef HAVE_OPENCV_HIGHGUI
-    cvShowImage("Initial image", image);
+    cv::imshow("Initial image", cv::cvarrToMat(image));
 #endif
     return LATENT_SVM_OK;
 }
@@ -445,7 +445,7 @@ int showPartFilterBoxes(IplImage *image,
         }
     }
 #ifdef HAVE_OPENCV_HIGHGUI
-    cvShowImage("Initial image", image);
+    cv::imshow("Initial image", cv::cvarrToMat(image));
 #endif
     return LATENT_SVM_OK;
 }
@@ -481,7 +481,7 @@ int showBoxes(IplImage *img,
                     color, thickness, line_type, shift);
     }
 #ifdef HAVE_OPENCV_HIGHGUI
-    cvShowImage("Initial image", img);
+    cv::imshow("Initial image", cv::cvarrToMat(img));
 #endif
     return LATENT_SVM_OK;
 }
@@ -582,7 +582,6 @@ int searchObjectThresholdSomeComponents(const CvLSVMFeaturePyramid *H,
     // For each component perform searching
     for (i = 0; i < kComponents; i++)
     {
-#ifdef HAVE_TBB
         int error = searchObjectThreshold(H, &(filters[componentIndex]), kPartFilters[i],
             b[i], maxXBorder, maxYBorder, scoreThreshold,
             &(pointsArr[i]), &(levelsArr[i]), &(kPointsArr[i]),
@@ -598,13 +597,6 @@ int searchObjectThresholdSomeComponents(const CvLSVMFeaturePyramid *H,
             free(partsDisplacementArr);
             return LATENT_SVM_SEARCH_OBJECT_FAILED;
         }
-#else
-    (void)numThreads;
-        searchObjectThreshold(H, &(filters[componentIndex]), kPartFilters[i],
-            b[i], maxXBorder, maxYBorder, scoreThreshold,
-            &(pointsArr[i]), &(levelsArr[i]), &(kPointsArr[i]),
-            &(scoreArr[i]), &(partsDisplacementArr[i]));
-#endif
         estimateBoxes(pointsArr[i], levelsArr[i], kPointsArr[i],
             filters[componentIndex]->sizeX, filters[componentIndex]->sizeY, &(oppPointsArr[i]));
         componentIndex += (kPartFilters[i] + 1);
diff --git a/modules/objdetect/src/latentsvmdetector.cpp b/modules/objdetect/src/latentsvmdetector.cpp
index 492e73f7cb..c7bfb2481f 100644
--- a/modules/objdetect/src/latentsvmdetector.cpp
+++ b/modules/objdetect/src/latentsvmdetector.cpp
@@ -1,4 +1,6 @@
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/objdetect/objdetect_c.h"
 #include "_lsvmparser.h"
 #include "_lsvm_matching.h"
 
@@ -124,9 +126,9 @@ CvSeq* cvLatentSvmDetectObjects(IplImage* image,
 
     for (int i = 0; i < numBoxesOut; i++)
     {
-        CvObjectDetection detection = {{0, 0, 0, 0}, 0};
+        CvObjectDetection detection = {CvRect(), 0};
         detection.score = scoreOut[i];
-        CvRect bounding_box = {0, 0, 0, 0};
+        CvRect bounding_box;
         bounding_box.x = pointsOut[i].x;
         bounding_box.y = pointsOut[i].y;
         bounding_box.width = oppPointsOut[i].x - pointsOut[i].x;
@@ -158,7 +160,7 @@ LatentSvmDetector::ObjectDetection::ObjectDetection( const Rect& _rect, float _s
 LatentSvmDetector::LatentSvmDetector()
 {}
 
-LatentSvmDetector::LatentSvmDetector( const std::vector<std::string>& filenames, const std::vector<std::string>& _classNames )
+LatentSvmDetector::LatentSvmDetector( const std::vector<String>& filenames, const std::vector<String>& _classNames )
 {
     load( filenames, _classNames );
 }
@@ -182,7 +184,7 @@ bool LatentSvmDetector::empty() const
     return detectors.empty();
 }
 
-const std::vector<std::string>& LatentSvmDetector::getClassNames() const
+const std::vector<String>& LatentSvmDetector::getClassNames() const
 {
     return classNames;
 }
@@ -192,13 +194,13 @@ size_t LatentSvmDetector::getClassCount() const
     return classNames.size();
 }
 
-static std::string extractModelName( const std::string& filename )
+static String extractModelName( const String& filename )
 {
     size_t startPos = filename.rfind('/');
-    if( startPos == std::string::npos )
+    if( startPos == String::npos )
         startPos = filename.rfind('\\');
 
-    if( startPos == std::string::npos )
+    if( startPos == String::npos )
         startPos = 0;
     else
         startPos++;
@@ -210,7 +212,7 @@ static std::string extractModelName( const std::string& filename )
     return filename.substr(startPos, substrLength);
 }
 
-bool LatentSvmDetector::load( const std::vector<std::string>& filenames, const std::vector<std::string>& _classNames )
+bool LatentSvmDetector::load( const std::vector<String>& filenames, const std::vector<String>& _classNames )
 {
     clear();
 
@@ -218,7 +220,7 @@ bool LatentSvmDetector::load( const std::vector<std::string>& filenames, const s
 
     for( size_t i = 0; i < filenames.size(); i++ )
     {
-        const std::string filename = filenames[i];
+        const String filename = filenames[i];
         if( filename.length() < 5 || filename.substr(filename.length()-4, 4) != ".xml" )
             continue;
 
diff --git a/modules/objdetect/src/linemod.cpp b/modules/objdetect/src/linemod.cpp
index d6d0227d97..e8fc8e4aac 100644
--- a/modules/objdetect/src/linemod.cpp
+++ b/modules/objdetect/src/linemod.cpp
@@ -66,7 +66,7 @@ static inline int getLabel(int quantized)
     case 64:  return 6;
     case 128: return 7;
     default:
-      CV_Error(CV_StsBadArg, "Invalid value of quantized parameter");
+      CV_Error(Error::StsBadArg, "Invalid value of quantized parameter");
       return -1; //avoid warning
   }
 }
@@ -175,7 +175,7 @@ void QuantizedPyramid::selectScatteredFeatures(const std::vector<Candidate>& can
                                                size_t num_features, float distance)
 {
   features.clear();
-  float distance_sq = CV_SQR(distance);
+  float distance_sq = distance * distance;
   int i = 0;
   while (features.size() < num_features)
   {
@@ -186,7 +186,7 @@ void QuantizedPyramid::selectScatteredFeatures(const std::vector<Candidate>& can
     for (int j = 0; (j < (int)features.size()) && keep; ++j)
     {
       Feature f = features[j];
-      keep = CV_SQR(c.f.x - f.x) + CV_SQR(c.f.y - f.y) >= distance_sq;
+      keep = (c.f.x - f.x)*(c.f.x - f.x) + (c.f.y - f.y)*(c.f.y - f.y) >= distance_sq;
     }
     if (keep)
       features.push_back(c.f);
@@ -196,24 +196,24 @@ void QuantizedPyramid::selectScatteredFeatures(const std::vector<Candidate>& can
       // Start back at beginning, and relax required distance
       i = 0;
       distance -= 1.0f;
-      distance_sq = CV_SQR(distance);
+      distance_sq = distance * distance;
     }
   }
 }
 
-Ptr<Modality> Modality::create(const std::string& modality_type)
+Ptr<Modality> Modality::create(const String& modality_type)
 {
   if (modality_type == "ColorGradient")
-    return new ColorGradient();
+    return makePtr<ColorGradient>();
   else if (modality_type == "DepthNormal")
-    return new DepthNormal();
+    return makePtr<DepthNormal>();
   else
-    return NULL;
+    return Ptr<Modality>();
 }
 
 Ptr<Modality> Modality::create(const FileNode& fn)
 {
-  std::string type = fn["type"];
+  String type = fn["type"];
   Ptr<Modality> modality = create(type);
   modality->read(fn);
   return modality;
@@ -306,9 +306,9 @@ static void quantizedOrientations(const Mat& src, Mat& magnitude,
     for (int i = 0; i < length0; i += 3)
     {
       // Use the gradient orientation of the channel whose magnitude is largest
-      int mag1 = CV_SQR(ptrx[i]) + CV_SQR(ptry[i]);
-      int mag2 = CV_SQR(ptrx[i + 1]) + CV_SQR(ptry[i + 1]);
-      int mag3 = CV_SQR(ptrx[i + 2]) + CV_SQR(ptry[i + 2]);
+      int mag1 = ptrx[i+0] * ptrx[i + 0] + ptry[i + 0] * ptry[i + 0];
+      int mag2 = ptrx[i+1] * ptrx[i + 1] + ptry[i + 1] * ptry[i + 1];
+      int mag3 = ptrx[i+2] * ptrx[i + 2] + ptry[i + 2] * ptry[i + 2];
 
       if (mag1 >= mag2 && mag1 >= mag3)
       {
@@ -339,7 +339,7 @@ static void quantizedOrientations(const Mat& src, Mat& magnitude,
 
   // Calculate the final gradient orientations
   phase(sobel_dx, sobel_dy, sobel_ag, true);
-  hysteresisGradient(magnitude, angle, sobel_ag, CV_SQR(threshold));
+  hysteresisGradient(magnitude, angle, sobel_ag, threshold * threshold);
 }
 
 void hysteresisGradient(Mat& magnitude, Mat& quantized_angle,
@@ -483,7 +483,7 @@ void ColorGradientPyramid::pyrDown()
   if (!mask.empty())
   {
     Mat next_mask;
-    resize(mask, next_mask, size, 0.0, 0.0, CV_INTER_NN);
+    resize(mask, next_mask, size, 0.0, 0.0, INTER_NEAREST);
     mask = next_mask;
   }
 
@@ -509,7 +509,7 @@ bool ColorGradientPyramid::extractTemplate(Template& templ) const
   // Create sorted list of all pixels with magnitude greater than a threshold
   std::vector<Candidate> candidates;
   bool no_mask = local_mask.empty();
-  float threshold_sq = CV_SQR(strong_threshold);
+  float threshold_sq = strong_threshold*strong_threshold;
   for (int r = 0; r < magnitude.rows; ++r)
   {
     const uchar* angle_r = angle.ptr<uchar>(r);
@@ -566,7 +566,7 @@ ColorGradient::ColorGradient(float _weak_threshold, size_t _num_features, float
 
 static const char CG_NAME[] = "ColorGradient";
 
-std::string ColorGradient::name() const
+String ColorGradient::name() const
 {
   return CG_NAME;
 }
@@ -574,12 +574,12 @@ std::string ColorGradient::name() const
 Ptr<QuantizedPyramid> ColorGradient::processImpl(const Mat& src,
                                                      const Mat& mask) const
 {
-  return new ColorGradientPyramid(src, mask, weak_threshold, num_features, strong_threshold);
+  return makePtr<ColorGradientPyramid>(src, mask, weak_threshold, num_features, strong_threshold);
 }
 
 void ColorGradient::read(const FileNode& fn)
 {
-  std::string type = fn["type"];
+  String type = fn["type"];
   CV_Assert(type == CG_NAME);
 
   weak_threshold = fn["weak_threshold"];
@@ -635,17 +635,11 @@ static void quantizedNormals(const Mat& src, Mat& dst, int distance_threshold,
 {
   dst = Mat::zeros(src.size(), CV_8U);
 
-  IplImage src_ipl = src;
-  IplImage* ap_depth_data = &src_ipl;
-  IplImage dst_ipl = dst;
-  IplImage* dst_ipl_ptr = &dst_ipl;
-  IplImage** m_dep = &dst_ipl_ptr;
+  const unsigned short * lp_depth   = src.ptr<ushort>();
+  unsigned char  * lp_normals = dst.ptr<uchar>();
 
-  unsigned short * lp_depth   = (unsigned short *)ap_depth_data->imageData;
-  unsigned char  * lp_normals = (unsigned char *)m_dep[0]->imageData;
-
-  const int l_W = ap_depth_data->width;
-  const int l_H = ap_depth_data->height;
+  const int l_W = src.cols;
+  const int l_H = src.rows;
 
   const int l_r = 5; // used to be 7
   const int l_offset0 = -l_r - l_r * l_W;
@@ -662,7 +656,7 @@ static void quantizedNormals(const Mat& src, Mat& dst, int distance_threshold,
 
   for (int l_y = l_r; l_y < l_H - l_r - 1; ++l_y)
   {
-    unsigned short * lp_line = lp_depth + (l_y * l_W + l_r);
+    const unsigned short * lp_line = lp_depth + (l_y * l_W + l_r);
     unsigned char * lp_norm = lp_normals + (l_y * l_W + l_r);
 
     for (int l_x = l_r; l_x < l_W - l_r - 1; ++l_x)
@@ -725,7 +719,7 @@ static void quantizedNormals(const Mat& src, Mat& dst, int distance_threshold,
       ++lp_norm;
     }
   }
-  cvSmooth(m_dep[0], m_dep[0], CV_MEDIAN, 5, 5);
+  medianBlur(dst, dst, 5);
 }
 
 class DepthNormalPyramid : public QuantizedPyramid
@@ -772,12 +766,12 @@ void DepthNormalPyramid::pyrDown()
   // In this case, NN-downsample the quantized image
   Mat next_normal;
   Size size(normal.cols / 2, normal.rows / 2);
-  resize(normal, next_normal, size, 0.0, 0.0, CV_INTER_NN);
+  resize(normal, next_normal, size, 0.0, 0.0, INTER_NEAREST);
   normal = next_normal;
   if (!mask.empty())
   {
     Mat next_mask;
-    resize(mask, next_mask, size, 0.0, 0.0, CV_INTER_NN);
+    resize(mask, next_mask, size, 0.0, 0.0, INTER_NEAREST);
     mask = next_mask;
   }
 }
@@ -805,7 +799,7 @@ bool DepthNormalPyramid::extractTemplate(Template& templ) const
     temp.setTo(1 << i, local_mask);
     bitwise_and(temp, normal, temp);
     // temp is now non-zero at pixels in the mask with quantized orientation i
-    distanceTransform(temp, distances[i], CV_DIST_C, 3);
+    distanceTransform(temp, distances[i], DIST_C, 3);
   }
 
   // Count how many features taken for each label
@@ -887,7 +881,7 @@ DepthNormal::DepthNormal(int _distance_threshold, int _difference_threshold, siz
 
 static const char DN_NAME[] = "DepthNormal";
 
-std::string DepthNormal::name() const
+String DepthNormal::name() const
 {
   return DN_NAME;
 }
@@ -895,13 +889,13 @@ std::string DepthNormal::name() const
 Ptr<QuantizedPyramid> DepthNormal::processImpl(const Mat& src,
                                                    const Mat& mask) const
 {
-  return new DepthNormalPyramid(src, mask, distance_threshold, difference_threshold,
-                                num_features, extract_threshold);
+  return makePtr<DepthNormalPyramid>(src, mask, distance_threshold, difference_threshold,
+                                     num_features, extract_threshold);
 }
 
 void DepthNormal::read(const FileNode& fn)
 {
-  std::string type = fn["type"];
+  String type = fn["type"];
   CV_Assert(type == DN_NAME);
 
   distance_threshold = fn["distance_threshold"];
@@ -1397,24 +1391,24 @@ Detector::Detector(const std::vector< Ptr<Modality> >& _modalities,
 }
 
 void Detector::match(const std::vector<Mat>& sources, float threshold, std::vector<Match>& matches,
-                     const std::vector<std::string>& class_ids, OutputArrayOfArrays quantized_images,
+                     const std::vector<String>& class_ids, OutputArrayOfArrays quantized_images,
                      const std::vector<Mat>& masks) const
 {
   matches.clear();
   if (quantized_images.needed())
     quantized_images.create(1, static_cast<int>(pyramid_levels * modalities.size()), CV_8U);
 
-  assert(sources.size() == modalities.size());
+  CV_Assert(sources.size() == modalities.size());
   // Initialize each modality with our sources
   std::vector< Ptr<QuantizedPyramid> > quantizers;
   for (int i = 0; i < (int)modalities.size(); ++i){
     Mat mask, source;
     source = sources[i];
     if(!masks.empty()){
-      assert(masks.size() == modalities.size());
+      CV_Assert(masks.size() == modalities.size());
       mask = masks[i];
     }
-    assert(mask.empty() || mask.size() == source.size());
+    CV_Assert(mask.empty() || mask.size() == source.size());
     quantizers.push_back(modalities[i]->process(source, mask));
   }
   // pyramid level -> modality -> quantization
@@ -1488,7 +1482,7 @@ struct MatchPredicate
 void Detector::matchClass(const LinearMemoryPyramid& lm_pyramid,
                           const std::vector<Size>& sizes,
                           float threshold, std::vector<Match>& matches,
-                          const std::string& class_id,
+                          const String& class_id,
                           const std::vector<TemplatePyramid>& template_pyramids) const
 {
   // For each template...
@@ -1613,7 +1607,7 @@ void Detector::matchClass(const LinearMemoryPyramid& lm_pyramid,
   }
 }
 
-int Detector::addTemplate(const std::vector<Mat>& sources, const std::string& class_id,
+int Detector::addTemplate(const std::vector<Mat>& sources, const String& class_id,
                           const Mat& object_mask, Rect* bounding_box)
 {
   int num_modalities = static_cast<int>(modalities.size());
@@ -1649,7 +1643,7 @@ int Detector::addTemplate(const std::vector<Mat>& sources, const std::string& cl
   return template_id;
 }
 
-int Detector::addSyntheticTemplate(const std::vector<Template>& templates, const std::string& class_id)
+int Detector::addSyntheticTemplate(const std::vector<Template>& templates, const String& class_id)
 {
   std::vector<TemplatePyramid>& template_pyramids = class_templates[class_id];
   int template_id = static_cast<int>(template_pyramids.size());
@@ -1657,7 +1651,7 @@ int Detector::addSyntheticTemplate(const std::vector<Template>& templates, const
   return template_id;
 }
 
-const std::vector<Template>& Detector::getTemplates(const std::string& class_id, int template_id) const
+const std::vector<Template>& Detector::getTemplates(const String& class_id, int template_id) const
 {
   TemplatesMap::const_iterator i = class_templates.find(class_id);
   CV_Assert(i != class_templates.end());
@@ -1674,7 +1668,7 @@ int Detector::numTemplates() const
   return ret;
 }
 
-int Detector::numTemplates(const std::string& class_id) const
+int Detector::numTemplates(const String& class_id) const
 {
   TemplatesMap::const_iterator i = class_templates.find(class_id);
   if (i == class_templates.end())
@@ -1682,9 +1676,9 @@ int Detector::numTemplates(const std::string& class_id) const
   return static_cast<int>(i->second.size());
 }
 
-std::vector<std::string> Detector::classIds() const
+std::vector<String> Detector::classIds() const
 {
-  std::vector<std::string> ids;
+  std::vector<String> ids;
   TemplatesMap::const_iterator i = class_templates.begin(), iend = class_templates.end();
   for ( ; i != iend; ++i)
   {
@@ -1724,7 +1718,7 @@ void Detector::write(FileStorage& fs) const
   fs << "]"; // modalities
 }
 
-  std::string Detector::readClass(const FileNode& fn, const std::string &class_id_override)
+  String Detector::readClass(const FileNode& fn, const String &class_id_override)
   {
   // Verify compatible with Detector settings
   FileNode mod_fn = fn["modalities"];
@@ -1732,14 +1726,14 @@ void Detector::write(FileStorage& fs) const
   FileNodeIterator mod_it = mod_fn.begin(), mod_it_end = mod_fn.end();
   int i = 0;
   for ( ; mod_it != mod_it_end; ++mod_it, ++i)
-    CV_Assert(modalities[i]->name() == (std::string)(*mod_it));
+    CV_Assert(modalities[i]->name() == (String)(*mod_it));
   CV_Assert((int)fn["pyramid_levels"] == pyramid_levels);
 
   // Detector should not already have this class
-    std::string class_id;
+    String class_id;
     if (class_id_override.empty())
     {
-      std::string class_id_tmp = fn["class_id"];
+      String class_id_tmp = fn["class_id"];
       CV_Assert(class_templates.find(class_id_tmp) == class_templates.end());
       class_id = class_id_tmp;
     }
@@ -1774,7 +1768,7 @@ void Detector::write(FileStorage& fs) const
   return class_id;
 }
 
-void Detector::writeClass(const std::string& class_id, FileStorage& fs) const
+void Detector::writeClass(const String& class_id, FileStorage& fs) const
 {
   TemplatesMap::const_iterator it = class_templates.find(class_id);
   CV_Assert(it != class_templates.end());
@@ -1805,25 +1799,25 @@ void Detector::writeClass(const std::string& class_id, FileStorage& fs) const
   fs << "]"; // pyramids
 }
 
-void Detector::readClasses(const std::vector<std::string>& class_ids,
-                           const std::string& format)
+void Detector::readClasses(const std::vector<String>& class_ids,
+                           const String& format)
 {
   for (size_t i = 0; i < class_ids.size(); ++i)
   {
-    const std::string& class_id = class_ids[i];
-    std::string filename = cv::format(format.c_str(), class_id.c_str());
+    const String& class_id = class_ids[i];
+    String filename = cv::format(format.c_str(), class_id.c_str());
     FileStorage fs(filename, FileStorage::READ);
     readClass(fs.root());
   }
 }
 
-void Detector::writeClasses(const std::string& format) const
+void Detector::writeClasses(const String& format) const
 {
   TemplatesMap::const_iterator it = class_templates.begin(), it_end = class_templates.end();
   for ( ; it != it_end; ++it)
   {
-    const std::string& class_id = it->first;
-    std::string filename = cv::format(format.c_str(), class_id.c_str());
+    const String& class_id = it->first;
+    String filename = cv::format(format.c_str(), class_id.c_str());
     FileStorage fs(filename, FileStorage::WRITE);
     writeClass(class_id, fs);
   }
@@ -1834,16 +1828,16 @@ static const int T_DEFAULTS[] = {5, 8};
 Ptr<Detector> getDefaultLINE()
 {
   std::vector< Ptr<Modality> > modalities;
-  modalities.push_back(new ColorGradient);
-  return new Detector(modalities, std::vector<int>(T_DEFAULTS, T_DEFAULTS + 2));
+  modalities.push_back(makePtr<ColorGradient>());
+  return makePtr<Detector>(modalities, std::vector<int>(T_DEFAULTS, T_DEFAULTS + 2));
 }
 
 Ptr<Detector> getDefaultLINEMOD()
 {
   std::vector< Ptr<Modality> > modalities;
-  modalities.push_back(new ColorGradient);
-  modalities.push_back(new DepthNormal);
-  return new Detector(modalities, std::vector<int>(T_DEFAULTS, T_DEFAULTS + 2));
+  modalities.push_back(makePtr<ColorGradient>());
+  modalities.push_back(makePtr<DepthNormal>());
+  return makePtr<Detector>(modalities, std::vector<int>(T_DEFAULTS, T_DEFAULTS + 2));
 }
 
 } // namespace linemod
diff --git a/modules/objdetect/src/lsvmtbbversion.cpp b/modules/objdetect/src/lsvmtbbversion.cpp
index 908c685b85..26e85bd1bf 100644
--- a/modules/objdetect/src/lsvmtbbversion.cpp
+++ b/modules/objdetect/src/lsvmtbbversion.cpp
@@ -120,4 +120,3 @@ int tbbTasksThresholdFunctionalScore(const CvLSVMFilterObject **filters, const i
     return LATENT_SVM_OK;
 };
 #endif
-
diff --git a/modules/objdetect/src/matching.cpp b/modules/objdetect/src/matching.cpp
index 382f6312a9..87ad8382f3 100644
--- a/modules/objdetect/src/matching.cpp
+++ b/modules/objdetect/src/matching.cpp
@@ -1,4 +1,5 @@
 #include "precomp.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
 #include "_lsvm_matching.h"
 #include <stdio.h>
 
diff --git a/modules/objdetect/src/precomp.cpp b/modules/objdetect/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/objdetect/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/objdetect/src/precomp.hpp b/modules/objdetect/src/precomp.hpp
index 2b237a1563..e5157d022f 100644
--- a/modules/objdetect/src/precomp.hpp
+++ b/modules/objdetect/src/precomp.hpp
@@ -43,21 +43,20 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/objdetect.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/core_c.h"
-#include "opencv2/core/internal.hpp"
+
+#include "opencv2/ml.hpp"
+
+#include "opencv2/core/utility.hpp"
 
 #include "opencv2/opencv_modules.hpp"
 #ifdef HAVE_OPENCV_HIGHGUI
 #  include "opencv2/highgui.hpp"
 #endif
 
+#include "opencv2/core/private.hpp"
+
 #ifdef HAVE_TEGRA_OPTIMIZATION
 #include "opencv2/objdetect/objdetect_tegra.hpp"
 #endif
diff --git a/modules/objdetect/src/resizeimg.cpp b/modules/objdetect/src/resizeimg.cpp
index 59f0b8cb8d..d999d12846 100644
--- a/modules/objdetect/src/resizeimg.cpp
+++ b/modules/objdetect/src/resizeimg.cpp
@@ -1,4 +1,5 @@
 #include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 #include "_lsvm_resizeimg.h"
 #include <stdio.h>
 #include <assert.h>
diff --git a/modules/objdetect/test/test_cascadeandhog.cpp b/modules/objdetect/test/test_cascadeandhog.cpp
index a165d9deb1..a301099058 100644
--- a/modules/objdetect/test/test_cascadeandhog.cpp
+++ b/modules/objdetect/test/test_cascadeandhog.cpp
@@ -41,6 +41,7 @@
 
 #include "test_precomp.hpp"
 #include "opencv2/imgproc.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
 
 using namespace cv;
 using namespace std;
@@ -117,12 +118,12 @@ int CV_DetectorTest::prepareData( FileStorage& _fs )
 //        fn[TOTAL_NO_PAIR_E] >> eps.totalNoPair;
 
         // read detectors
-        if( fn[DETECTOR_NAMES].node->data.seq != 0 )
+        if( fn[DETECTOR_NAMES].size() != 0 )
         {
             FileNodeIterator it = fn[DETECTOR_NAMES].begin();
             for( ; it != fn[DETECTOR_NAMES].end(); )
             {
-                string _name;
+                String _name;
                 it >> _name;
                 detectorNames.push_back(_name);
                 readDetector(fn[DETECTORS][_name]);
@@ -132,11 +133,11 @@ int CV_DetectorTest::prepareData( FileStorage& _fs )
 
         // read images filenames and images
         string dataPath = ts->get_data_path();
-        if( fn[IMAGE_FILENAMES].node->data.seq != 0 )
+        if( fn[IMAGE_FILENAMES].size() != 0 )
         {
             for( FileNodeIterator it = fn[IMAGE_FILENAMES].begin(); it != fn[IMAGE_FILENAMES].end(); )
             {
-                string filename;
+                String filename;
                 it >> filename;
                 imageFilenames.push_back(filename);
                 Mat img = imread( dataPath+filename, 1 );
@@ -210,7 +211,7 @@ void CV_DetectorTest::run( int )
         {
             char buf[10];
             sprintf( buf, "%s%d", "img_", ii );
-            cvWriteComment( validationFS.fs, buf, 0 );
+            //cvWriteComment( validationFS.fs, buf, 0 );
             validationFS << *it;
         }
         validationFS << "]"; // IMAGE_FILENAMES
@@ -316,7 +317,7 @@ int CV_DetectorTest::validate( int detectorIdx, vector<vector<Rect> >& objects )
         string imageIdxStr = buf;
         FileNode node = validationFS.getFirstTopLevelNode()[VALIDATION][detectorNames[detectorIdx]][imageIdxStr];
         vector<Rect> valRects;
-        if( node.node->data.seq != 0 )
+        if( node.size() != 0 )
         {
             for( FileNodeIterator it2 = node.begin(); it2 != node.end(); )
             {
@@ -402,7 +403,7 @@ CV_CascadeDetectorTest::CV_CascadeDetectorTest()
 
 void CV_CascadeDetectorTest::readDetector( const FileNode& fn )
 {
-    string filename;
+    String filename;
     int flag;
     fn[FILENAME] >> filename;
     detectorFilenames.push_back(filename);
@@ -410,12 +411,12 @@ void CV_CascadeDetectorTest::readDetector( const FileNode& fn )
     if( flag )
         flags.push_back( 0 );
     else
-        flags.push_back( CV_HAAR_SCALE_IMAGE );
+        flags.push_back( CASCADE_SCALE_IMAGE );
 }
 
 void CV_CascadeDetectorTest::writeDetector( FileStorage& fs, int di )
 {
-    int sc = flags[di] & CV_HAAR_SCALE_IMAGE ? 0 : 1;
+    int sc = flags[di] & CASCADE_SCALE_IMAGE ? 0 : 1;
     fs << FILENAME << detectorFilenames[di];
     fs << C_SCALE_CASCADE << sc;
 }
@@ -425,21 +426,21 @@ int CV_CascadeDetectorTest::detectMultiScale_C( const string& filename,
                                                 int di, const Mat& img,
                                                 vector<Rect>& objects )
 {
-    Ptr<CvHaarClassifierCascade> c_cascade = cvLoadHaarClassifierCascade(filename.c_str(), cvSize(0,0));
-    Ptr<CvMemStorage> storage = cvCreateMemStorage();
+    Ptr<CvHaarClassifierCascade> c_cascade(cvLoadHaarClassifierCascade(filename.c_str(), cvSize(0,0)));
+    Ptr<CvMemStorage> storage(cvCreateMemStorage());
 
-    if( c_cascade.empty() )
+    if( !c_cascade )
     {
         ts->printf( cvtest::TS::LOG, "cascade %s can not be opened");
         return cvtest::TS::FAIL_INVALID_TEST_DATA;
     }
     Mat grayImg;
-    cvtColor( img, grayImg, CV_BGR2GRAY );
+    cvtColor( img, grayImg, COLOR_BGR2GRAY );
     equalizeHist( grayImg, grayImg );
 
     CvMat c_gray = grayImg;
     CvSeq* rs = cvHaarDetectObjects(&c_gray, c_cascade, storage, 1.1, 3, flags[di] );
-    
+
     objects.clear();
     for( int i = 0; i < rs->total; i++ )
     {
@@ -469,7 +470,7 @@ int CV_CascadeDetectorTest::detectMultiScale( int di, const Mat& img,
         return cvtest::TS::FAIL_INVALID_TEST_DATA;
     }
     Mat grayImg;
-    cvtColor( img, grayImg, CV_BGR2GRAY );
+    cvtColor( img, grayImg, COLOR_BGR2GRAY );
     equalizeHist( grayImg, grayImg );
     cascade.detectMultiScale( grayImg, objects, 1.1, 3, flags[di] );
     return cvtest::TS::OK;
@@ -493,8 +494,8 @@ CV_HOGDetectorTest::CV_HOGDetectorTest()
 
 void CV_HOGDetectorTest::readDetector( const FileNode& fn )
 {
-    string filename;
-    if( fn[FILENAME].node->data.seq != 0 )
+    String filename;
+    if( fn[FILENAME].size() != 0 )
         fn[FILENAME] >> filename;
     detectorFilenames.push_back( filename);
 }
@@ -995,6 +996,8 @@ inline bool HOGDescriptorTester::is_failed() const
     return failed;
 }
 
+static inline int gcd(int a, int b) { return (a % b == 0) ? b : gcd (b, a % b); }
+
 void HOGDescriptorTester::detect(const Mat& img,
     vector<Point>& hits, vector<double>& weights, double hitThreshold,
     Size winStride, Size padding, const vector<Point>& locations) const
@@ -1083,7 +1086,7 @@ void HOGDescriptorTester::detect(const Mat& img,
     }
 
     const double eps = 0.0;
-    double diff_norm = norm(Mat(actual_weights) - Mat(weights), CV_L2);
+    double diff_norm = norm(Mat(actual_weights) - Mat(weights), NORM_L2);
     if (diff_norm > eps)
     {
         ts->printf(cvtest::TS::SUMMARY, "Weights for found locations aren't equal.\n"
@@ -1162,7 +1165,7 @@ void HOGDescriptorTester::compute(const Mat& img, vector<float>& descriptors,
     std::vector<float> actual_descriptors;
     actual_hog->compute(img, actual_descriptors, winStride, padding, locations);
 
-    double diff_norm = cv::norm(Mat(actual_descriptors) - Mat(descriptors), CV_L2);
+    double diff_norm = cv::norm(Mat(actual_descriptors) - Mat(descriptors), NORM_L2);
     const double eps = 0.0;
     if (diff_norm > eps)
     {
@@ -1312,7 +1315,7 @@ void HOGDescriptorTester::computeGradient(const Mat& img, Mat& grad, Mat& qangle
     const double eps = 0.0;
     for (i = 0; i < 2; ++i)
     {
-       double diff_norm = norm(reference_mats[i] - actual_mats[i], CV_L2);
+       double diff_norm = norm(reference_mats[i] - actual_mats[i], NORM_L2);
        if (diff_norm > eps)
        {
            ts->printf(cvtest::TS::LOG, "%s matrices are not equal\n"
diff --git a/modules/objdetect/test/test_latentsvmdetector.cpp b/modules/objdetect/test/test_latentsvmdetector.cpp
index fbd0a5d785..30ab157f01 100644
--- a/modules/objdetect/test/test_latentsvmdetector.cpp
+++ b/modules/objdetect/test/test_latentsvmdetector.cpp
@@ -41,13 +41,9 @@
 //M*/
 
 #include "test_precomp.hpp"
-
+#include "opencv2/objdetect/objdetect_c.h"
 #include <string>
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #ifdef HAVE_TBB
 #include "tbb/task_scheduler_init.h"
 #endif
@@ -87,8 +83,9 @@ void CV_LatentSVMDetectorTest::run( int /* start_from */)
     init.initialize(numThreads);
 #endif
 
-    IplImage* image = cvLoadImage(img_path.c_str());
-    if (!image)
+    Mat image2 = cv::imread(img_path.c_str());
+    IplImage image = image2;
+    if (image2.empty())
     {
         ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
         return;
@@ -98,13 +95,12 @@ void CV_LatentSVMDetectorTest::run( int /* start_from */)
     if (!detector)
     {
         ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
-        cvReleaseImage(&image);
         return;
     }
 
     CvMemStorage* storage = cvCreateMemStorage(0);
     CvSeq* detections = 0;
-    detections = cvLatentSvmDetectObjects(image, detector, storage, 0.5f, numThreads);
+    detections = cvLatentSvmDetectObjects(&image, detector, storage, 0.5f, numThreads);
     if (detections->total != num_detections)
     {
         ts->set_failed_test_info( cvtest::TS::FAIL_MISMATCH );
@@ -129,7 +125,6 @@ void CV_LatentSVMDetectorTest::run( int /* start_from */)
 #endif
     cvReleaseMemStorage( &storage );
     cvReleaseLatentSvmDetector( &detector );
-    cvReleaseImage( &image );
 }
 
 // Test for c++ version of Latent SVM
@@ -232,9 +227,9 @@ void LatentSVMDetectorTest::run( int /* start_from */)
     // detector12 - to test case of two (several) classes 'cat' and car
 
     // Load detectors
-    LatentSvmDetector detector1( vector<string>(1,model_path_cat) );
+    LatentSvmDetector detector1( vector<String>(1,model_path_cat) );
 
-    vector<string> models_pathes(2);
+    vector<String> models_pathes(2);
     models_pathes[0] = model_path_cat;
     models_pathes[1] = model_path_car;
     LatentSvmDetector detector12( models_pathes );
diff --git a/modules/objdetect/test/test_precomp.cpp b/modules/objdetect/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/objdetect/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/ocl/CMakeLists.txt b/modules/ocl/CMakeLists.txt
index d0e254d3a2..05b28b83fe 100644
--- a/modules/ocl/CMakeLists.txt
+++ b/modules/ocl/CMakeLists.txt
@@ -1,69 +1,7 @@
-# Will be modified later
 if(NOT HAVE_OPENCL)
   ocv_module_disable(ocl)
 endif()
 
 set(the_description "OpenCL-accelerated Computer Vision")
-ocv_add_module(ocl opencv_core opencv_imgproc opencv_features2d opencv_objdetect opencv_video)
-
-ocv_module_include_directories()
-
-file(GLOB CL_FILES "${CMAKE_CURRENT_SOURCE_DIR}/src/kernels/*.cl")
-set(kernels_cpp "${CMAKE_CURRENT_BINARY_DIR}/kernels.cpp")
-set(cl2cpp_script "${CMAKE_CURRENT_SOURCE_DIR}/cl2cpp.cmake")
-
-add_custom_command(
-  OUTPUT ${kernels_cpp}
-  COMMAND ${CMAKE_COMMAND} -DCL_DIR="${CMAKE_CURRENT_SOURCE_DIR}/src/kernels" -DOUTPUT="${kernels_cpp}" -P ${cl2cpp_script}
-  DEPENDS ${CL_FILES} ${cl2cpp_script})
-
-file(GLOB lib_hdrs     "include/opencv2/${name}/*.hpp" "include/opencv2/${name}/*.h")
-file(GLOB lib_srcs     "src/*.cpp")
-file(GLOB lib_int_hdrs "src/*.h*")
-
-source_group("Include"   FILES ${lib_hdrs})
-source_group("Src\\Host" FILES ${lib_srcs} ${lib_int_hdrs} ${kernels_cpp})
-
-if (HAVE_OPENCL)
-  set(ocl_link_libs ${OPENCL_LIBRARIES})
-  if(OPENCL_INCLUDE_DIR)
-    ocv_include_directories(${OPENCL_INCLUDE_DIR})
-  endif()
-  if (HAVE_CLAMDFFT)
-    set(ocl_link_libs ${ocl_link_libs} ${CLAMDFFT_LIBRARIES})
-    ocv_include_directories(${CLAMDFFT_INCLUDE_DIR})
-  endif()
-  if (HAVE_CLAMDBLAS)
-    set(ocl_link_libs ${ocl_link_libs} ${CLAMDBLAS_LIBRARIES})
-    ocv_include_directories(${CLAMDBLAS_INCLUDE_DIR})
-  endif()
-endif()
-
+ocv_define_module(ocl opencv_core opencv_imgproc opencv_features2d opencv_objdetect opencv_video opencv_calib3d)
 ocv_warnings_disable(CMAKE_CXX_FLAGS -Wshadow)
-
-ocv_set_module_sources(HEADERS ${lib_hdrs} SOURCES ${lib_int_hdrs} ${lib_srcs} ${kernels_cpp})
-ocv_create_module(${ocl_link_libs})
-
-install(FILES ${lib_hdrs}
-  DESTINATION include/opencv2/${name}
-  COMPONENT main)
-
-ocv_add_precompiled_headers(${the_module})
-
-################################################################################################################
-################################      OpenCL Module Tests     ##################################################
-################################################################################################################
-file(GLOB test_srcs "test/*.cpp")
-file(GLOB test_hdrs "test/*.hpp" "test/*.h")
-
-ocv_add_accuracy_tests(FILES "Include" ${test_hdrs}
-                       FILES "Src" ${test_srcs})
-
-################################################################################################################
-################################   OpenCL Module Performance  ##################################################
-################################################################################################################
-file(GLOB perf_srcs "perf/*.cpp")
-file(GLOB perf_hdrs "perf/*.hpp" "perf/*.h")
-
-ocv_add_perf_tests(FILES "Include" ${perf_hdrs}
-                   FILES "Src" ${perf_srcs})
diff --git a/modules/ocl/doc/data_structures.rst b/modules/ocl/doc/data_structures.rst
index 8519c40a3f..556efa961e 100644
--- a/modules/ocl/doc/data_structures.rst
+++ b/modules/ocl/doc/data_structures.rst
@@ -186,4 +186,4 @@ Only basic flags are supported in oclMat(i.e. depth number of channels)
 
 All the 3-channel matrix(i.e. RGB image) are represented by 4-channel matrix in oclMat. It means 3-channel image have 4-channel space with the last channel unused. We provide a transparent interface to handle the difference between OpenCV Mat and oclMat.
 
-For example: If a oclMat has 3 channels, channels() returns 3 and oclchannels() returns 4
\ No newline at end of file
+For example: If a oclMat has 3 channels, channels() returns 3 and oclchannels() returns 4
diff --git a/modules/ocl/doc/feature_detection_and_description.rst b/modules/ocl/doc/feature_detection_and_description.rst
index d4fd6e65fa..11fb27242c 100644
--- a/modules/ocl/doc/feature_detection_and_description.rst
+++ b/modules/ocl/doc/feature_detection_and_description.rst
@@ -363,7 +363,9 @@ The class implements Histogram of Oriented Gradients ([Dalal2005]_) object detec
 
 Interfaces of all methods are kept similar to the ``CPU HOG`` descriptor and detector analogues as much as possible.
 
+.. note::
 
+   (Ocl) An example using the HOG descriptor can be found at opencv_source_code/samples/ocl/hog.cpp
 
 ocl::HOGDescriptor::HOGDescriptor
 -------------------------------------
@@ -497,4 +499,4 @@ Returns block descriptors computed for the whole image.
 
         * **DESCR_FORMAT_COL_BY_COL** - Column-major order.
 
-The function is mainly used to learn the classifier.
\ No newline at end of file
+The function is mainly used to learn the classifier.
diff --git a/modules/ocl/doc/image_filtering.rst b/modules/ocl/doc/image_filtering.rst
index ca97d3a93b..4798bddaa7 100644
--- a/modules/ocl/doc/image_filtering.rst
+++ b/modules/ocl/doc/image_filtering.rst
@@ -109,25 +109,60 @@ Returns void
 
 The function calculates the Laplacian of the source image by adding up the second x and y derivatives calculated using the Sobel operator.
 
+ocl::ConvolveBuf
+----------------
+.. ocv:struct:: ocl::ConvolveBuf
+
+Class providing a memory buffer for :ocv:func:`ocl::convolve` function, plus it allows to adjust some specific parameters. ::
+
+    struct CV_EXPORTS ConvolveBuf
+    {
+        Size result_size;
+        Size block_size;
+        Size user_block_size;
+        Size dft_size;
+        int spect_len;
+
+        oclMat image_spect, templ_spect, result_spect;
+        oclMat image_block, templ_block, result_data;
+
+        void create(Size image_size, Size templ_size);
+        static Size estimateBlockSize(Size result_size, Size templ_size);
+    };
+
+You can use field `user_block_size` to set specific block size for :ocv:func:`ocl::convolve` function. If you leave its default value `Size(0,0)` then automatic estimation of block size will be used (which is optimized for speed). By varying `user_block_size` you can reduce memory requirements at the cost of speed.
+
+ocl::ConvolveBuf::create
+------------------------
+.. ocv:function:: ocl::ConvolveBuf::create(Size image_size, Size templ_size)
+
+Constructs a buffer for :ocv:func:`ocl::convolve` function with respective arguments.
+
 ocl::convolve
 ------------------
 Returns void
 
-.. ocv:function:: void ocl::convolve(const oclMat &image, const oclMat &temp1, oclMat &result)
+.. ocv:function:: void ocl::convolve(const oclMat &image, const oclMat &temp1, oclMat &result, bool ccorr=false)
 
-    :param image: The source image
+.. ocv:function:: void ocl::convolve(const oclMat &image, const oclMat &temp1, oclMat &result, bool ccorr, ConvolveBuf& buf)
 
-    :param temp1: Convolution kernel, a single-channel floating point matrix.
+    :param image: The source image. Only  ``CV_32FC1`` images are supported for now.
+
+    :param temp1: Convolution kernel, a single-channel floating point matrix. The size is not greater than the  ``image`` size. The type is the same as  ``image``.
 
     :param result: The destination image
 
+    :param ccorr: Flags to evaluate cross-correlation instead of convolution.
+
+    :param buf: Optional buffer to avoid extra memory allocations and to adjust some specific parameters. See :ocv:struct:`ocl::ConvolveBuf`.
+
 Convolves an image with the kernel. Supports only CV_32FC1 data types and do not support ROI.
 
 ocl::bilateralFilter
 --------------------
 Returns void
 
-.. ocv:function:: void ocl::bilateralFilter(const oclMat &src, oclMat &dst, int d, double sigmaColor, double sigmaSpave, int borderType=BORDER_DEFAULT)
+.. ocv:function:: void ocl::bilateralFilter(const oclMat &src, oclMat &dst, int d, double sigmaColor, double sigmaSpace, int borderType=BORDER_DEFAULT)
 
     :param src: The source image
 
@@ -280,4 +315,4 @@ Performs linear blending of two images.
 
     :param weights2: Weights for second image. Must have tha same size as ``img2`` . Supports only ``CV_32F`` type.
 
-    :param result: Destination image.
\ No newline at end of file
+    :param result: Destination image.
diff --git a/modules/ocl/doc/image_processing.rst b/modules/ocl/doc/image_processing.rst
index 3f97301749..98f334762a 100644
--- a/modules/ocl/doc/image_processing.rst
+++ b/modules/ocl/doc/image_processing.rst
@@ -257,7 +257,10 @@ The class can calculate an optical flow for a sparse feature set or dense optica
 
 .. seealso:: :ocv:func:`calcOpticalFlowPyrLK`
 
+.. note::
 
+   (Ocl) An example the Lucas Kanade optical flow pyramid method can be found at opencv_source_code/samples/ocl/pyrlk_optical_flow.cpp
+   (Ocl) An example for square detection can be found at opencv_source_code/samples/ocl/squares.cpp
 
 ocl::PyrLKOpticalFlow::sparse
 -----------------------------
@@ -363,4 +366,3 @@ Finds circles in a grayscale image using the Hough transform.
 
 .. note:: Currently only non-ROI oclMat is supported for src.
 .. seealso:: :ocv:func:`HoughCircles`
-
diff --git a/modules/ocl/doc/matrix_reductions.rst b/modules/ocl/doc/matrix_reductions.rst
index 69b27eb658..aed9fa5644 100644
--- a/modules/ocl/doc/matrix_reductions.rst
+++ b/modules/ocl/doc/matrix_reductions.rst
@@ -67,4 +67,4 @@ Returns the squared sum of matrix elements for each channel
 
     :param m: The Source image of all depth
 
-Counts the squared sum of matrix elements for each channel.
\ No newline at end of file
+Counts the squared sum of matrix elements for each channel.
diff --git a/modules/ocl/doc/object_detection.rst b/modules/ocl/doc/object_detection.rst
index 0104da5930..024ce52680 100644
--- a/modules/ocl/doc/object_detection.rst
+++ b/modules/ocl/doc/object_detection.rst
@@ -12,23 +12,24 @@ Cascade classifier class used for object detection. Supports HAAR cascade classi
     class CV_EXPORTS OclCascadeClassifier : public CascadeClassifier
     {
     public:
-          OclCascadeClassifier() {};
-          ~OclCascadeClassifier() {};
-           CvSeq *oclHaarDetectObjects(oclMat &gimg, CvMemStorage *storage,
-                                      double scaleFactor,int minNeighbors,
-                                      int flags, CvSize minSize = cvSize(0, 0),
-                                      CvSize maxSize = cvSize(0, 0));
+            void detectMultiScale(oclMat &image, CV_OUT std::vector<cv::Rect>& faces,
+                                              double scaleFactor = 1.1, int minNeighbors = 3, int flags = 0,
+                                              Size minSize = Size(), Size maxSize = Size());
     };
 
+.. note::
+
+   (Ocl) A face detection example using cascade classifiers can be found at opencv_source_code/samples/ocl/facedetect.cpp
+
 ocl::OclCascadeClassifier::oclHaarDetectObjects
 ------------------------------------------------------
-Returns the detected objects by a list of rectangles
+Detects objects of different sizes in the input image.
 
-.. ocv:function:: CvSeq* ocl::OclCascadeClassifier::oclHaarDetectObjects(oclMat &gimg, CvMemStorage *storage, double scaleFactor,int minNeighbors, int flags, CvSize minSize = cvSize(0, 0), CvSize maxSize = cvSize(0, 0))
+.. ocv:function:: void ocl::OclCascadeClassifier::detectMultiScale(oclMat &image, std::vector<cv::Rect>& faces, double scaleFactor = 1.1, int minNeighbors = 3, int flags = 0, Size minSize = Size(), Size maxSize = Size())
 
     :param image:  Matrix of type CV_8U containing an image where objects should be detected.
 
-    :param imageobjectsBuff: Buffer to store detected objects (rectangles). If it is empty, it is allocated with the defaultsize. If not empty, the function searches not more than N  objects, where N = sizeof(objectsBufers data)/sizeof(cv::Rect).
+    :param faces: Vector of rectangles where each rectangle contains the detected object.
 
     :param scaleFactor: Parameter specifying how much the image size is reduced at each image scale.
 
@@ -36,7 +37,9 @@ Returns the detected objects by a list of rectangles
 
     :param minSize: Minimum possible object size. Objects smaller than that are ignored.
 
-Detects objects of different sizes in the input image,only tested for face detection now. The function returns the number of detected objects.
+    :param maxSize: Maximum possible object size. Objects larger than that are ignored.
+
+The function provides a very similar interface with that in CascadeClassifier class, except using oclMat as input image.
 
 ocl::MatchTemplateBuf
 ---------------------
@@ -88,102 +91,3 @@ Computes a proximity map for a raster template and an image where the template i
     * ``CV_TM_CCORR``
 
 .. seealso:: :ocv:func:`matchTemplate`
-
-
-ocl::SURF_OCL
--------------
-.. ocv:class:: ocl::SURF_OCL
-
-Class used for extracting Speeded Up Robust Features (SURF) from an image. ::
-
-    class SURF_OCL
-    {
-    public:
-        enum KeypointLayout
-        {
-            X_ROW = 0,
-            Y_ROW,
-            LAPLACIAN_ROW,
-            OCTAVE_ROW,
-            SIZE_ROW,
-            ANGLE_ROW,
-            HESSIAN_ROW,
-            ROWS_COUNT
-        };
-
-        //! the default constructor
-        SURF_OCL();
-        //! the full constructor taking all the necessary parameters
-        explicit SURF_OCL(double _hessianThreshold, int _nOctaves=4,
-             int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
-
-        //! returns the descriptor size in float's (64 or 128)
-        int descriptorSize() const;
-
-        //! upload host keypoints to device memory
-        void uploadKeypoints(const vector<KeyPoint>& keypoints,
-            oclMat& keypointsocl);
-        //! download keypoints from device to host memory
-        void downloadKeypoints(const oclMat& keypointsocl,
-            vector<KeyPoint>& keypoints);
-
-        //! download descriptors from device to host memory
-        void downloadDescriptors(const oclMat& descriptorsocl,
-            vector<float>& descriptors);
-
-        void operator()(const oclMat& img, const oclMat& mask,
-            oclMat& keypoints);
-
-        void operator()(const oclMat& img, const oclMat& mask,
-            oclMat& keypoints, oclMat& descriptors,
-            bool useProvidedKeypoints = false);
-
-        void operator()(const oclMat& img, const oclMat& mask,
-            std::vector<KeyPoint>& keypoints);
-
-        void operator()(const oclMat& img, const oclMat& mask,
-            std::vector<KeyPoint>& keypoints, oclMat& descriptors,
-            bool useProvidedKeypoints = false);
-
-        void operator()(const oclMat& img, const oclMat& mask,
-            std::vector<KeyPoint>& keypoints,
-            std::vector<float>& descriptors,
-            bool useProvidedKeypoints = false);
-
-        void releaseMemory();
-
-        // SURF parameters
-        double hessianThreshold;
-        int nOctaves;
-        int nOctaveLayers;
-        bool extended;
-        bool upright;
-
-        //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
-        float keypointsRatio;
-
-        oclMat sum, mask1, maskSum, intBuffer;
-
-        oclMat det, trace;
-
-        oclMat maxPosBuffer;
-    };
-
-
-The class ``SURF_OCL`` implements Speeded Up Robust Features descriptor. There is a fast multi-scale Hessian keypoint detector that can be used to find the keypoints (which is the default option). But the descriptors can also be computed for the user-specified keypoints. Only 8-bit grayscale images are supported.
-
-The class ``SURF_OCL`` can store results in the GPU and CPU memory. It provides functions to convert results between CPU and GPU version ( ``uploadKeypoints``, ``downloadKeypoints``, ``downloadDescriptors`` ). The format of CPU results is the same as ``SURF`` results. GPU results are stored in ``oclMat``. The ``keypoints`` matrix is :math:`\texttt{nFeatures} \times 7` matrix with the ``CV_32FC1`` type.
-
-* ``keypoints.ptr<float>(X_ROW)[i]`` contains x coordinate of the i-th feature.
-* ``keypoints.ptr<float>(Y_ROW)[i]`` contains y coordinate of the i-th feature.
-* ``keypoints.ptr<float>(LAPLACIAN_ROW)[i]``  contains the laplacian sign of the i-th feature.
-* ``keypoints.ptr<float>(OCTAVE_ROW)[i]`` contains the octave of the i-th feature.
-* ``keypoints.ptr<float>(SIZE_ROW)[i]`` contains the size of the i-th feature.
-* ``keypoints.ptr<float>(ANGLE_ROW)[i]`` contain orientation of the i-th feature.
-* ``keypoints.ptr<float>(HESSIAN_ROW)[i]`` contains the response of the i-th feature.
-
-The ``descriptors`` matrix is :math:`\texttt{nFeatures} \times \texttt{descriptorSize}` matrix with the ``CV_32FC1`` type.
-
-The class ``SURF_OCL`` uses some buffers and provides access to it. All buffers can be safely released between function calls.
-
-.. seealso:: :ocv:class:`SURF`
\ No newline at end of file
diff --git a/modules/ocl/doc/operations_on_matrices.rst b/modules/ocl/doc/operations_on_matrices.rst
index 145659b991..e47e720922 100644
--- a/modules/ocl/doc/operations_on_matrices.rst
+++ b/modules/ocl/doc/operations_on_matrices.rst
@@ -481,4 +481,40 @@ Performs generalized matrix multiplication.
             * **GEMM_1_T** transpose  ``src1``
             * **GEMM_2_T** transpose  ``src2``
 
-.. seealso:: :ocv:func:`gemm`
\ No newline at end of file
+.. seealso:: :ocv:func:`gemm`
+
+ocl::sortByKey
+------------------
+Returns void
+
+.. ocv:function:: void ocl::sortByKey(oclMat& keys, oclMat& values, int method, bool isGreaterThan = false)
+
+    :param keys:   The keys to be used as sorting indices.
+
+    :param values: The array of values.
+
+    :param isGreaterThan: Determine sorting order.
+
+    :param method: supported sorting methods:
+            * **SORT_BITONIC**   bitonic sort, only support power-of-2 buffer size
+            * **SORT_SELECTION** selection sort, currently cannot sort duplicate keys
+            * **SORT_MERGE**     merge sort
+            * **SORT_RADIX**     radix sort, only support signed int/float keys(``CV_32S``/``CV_32F``)
+
+Returns the sorted result of all the elements in values based on equivalent keys.
+
+The element unit in the values to be sorted is determined from the data type,
+i.e., a ``CV_32FC2`` input ``{a1a2, b1b2}`` will be considered as two elements, regardless its matrix dimension.
+
+Both keys and values will be sorted inplace.
+
+Keys needs to be a **single** channel `oclMat`.
+
+Example::
+    input -
+    keys   = {2,    3,   1}   (CV_8UC1)
+    values = {10,5, 4,3, 6,2} (CV_8UC2)
+    sortByKey(keys, values, SORT_SELECTION, false);
+    output -
+    keys   = {1,    2,   3}   (CV_8UC1)
+    values = {6,2, 10,5, 4,3} (CV_8UC2)
diff --git a/modules/ocl/doc/structures_and_utility_functions.rst b/modules/ocl/doc/structures_and_utility_functions.rst
index 32aa8b9eae..3810d7e2d2 100644
--- a/modules/ocl/doc/structures_and_utility_functions.rst
+++ b/modules/ocl/doc/structures_and_utility_functions.rst
@@ -55,4 +55,4 @@ Returns the pointer to the opencl command queue
 
 .. ocv:function:: void* ocl::getoclCommandQueue()
 
-Thefunction are used to get opencl command queue so that opencv can interactive with other opencl program.
\ No newline at end of file
+Thefunction are used to get opencl command queue so that opencv can interactive with other opencl program.
diff --git a/modules/ocl/include/opencv2/ocl.hpp b/modules/ocl/include/opencv2/ocl.hpp
index 3cf09dc840..fc7c114d9f 100644
--- a/modules/ocl/include/opencv2/ocl.hpp
+++ b/modules/ocl/include/opencv2/ocl.hpp
@@ -50,7 +50,6 @@
 #include "opencv2/core.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/objdetect.hpp"
-#include "opencv2/features2d.hpp"
 
 namespace cv
 {
@@ -103,7 +102,8 @@ namespace cv
             ~Info();
             void release();
             Info &operator = (const Info &m);
-            std::vector<std::string> DeviceName;
+            std::vector<String> DeviceName;
+            String PlatformName;
         };
         //////////////////////////////// Initialization & Info ////////////////////////
         //this function may be obsoleted
@@ -117,42 +117,52 @@ namespace cv
         //the devnum is the index of the selected device in DeviceName vector of INfo
         CV_EXPORTS void setDevice(Info &oclinfo, int devnum = 0);
 
-        //optional function, if you want save opencl binary kernel to the file, set its path
-        CV_EXPORTS  void setBinpath(const char *path);
-
         //The two functions below enable other opencl program to use ocl module's cl_context and cl_command_queue
+        //returns cl_context *
         CV_EXPORTS void* getoclContext();
-
+        //returns cl_command_queue *
         CV_EXPORTS void* getoclCommandQueue();
 
+        //explicit call clFinish. The global command queue will be used.
+        CV_EXPORTS void finish();
+
         //this function enable ocl module to use customized cl_context and cl_command_queue
         //getDevice also need to be called before this function
         CV_EXPORTS void setDeviceEx(Info &oclinfo, void *ctx, void *qu, int devnum = 0);
 
-        //////////////////////////////// Error handling ////////////////////////
-        CV_EXPORTS void error(const char *error_string, const char *file, const int line, const char *func);
+        //returns true when global OpenCL context is initialized
+        CV_EXPORTS bool initialized();
 
         //////////////////////////////// OpenCL context ////////////////////////
         //This is a global singleton class used to represent a OpenCL context.
-        class Context
+        class CV_EXPORTS Context
         {
         protected:
             Context();
             friend class std::auto_ptr<Context>;
+            friend bool initialized();
+        private:
             static std::auto_ptr<Context> clCxt;
-
+            static int val;
         public:
             ~Context();
-            static int val;
+            void release();
+            Info::Impl* impl;
+
             static Context *getContext();
             static void setContext(Info &oclinfo);
-            struct Impl;
-            Impl *impl;
+
+            enum {CL_DOUBLE, CL_UNIFIED_MEM, CL_VER_1_2};
+            bool supportsFeature(int ftype);
+            size_t computeUnits();
+            size_t maxWorkGroupSize();
+            void* oclContext();
+            void* oclCommandQueue();
         };
 
         //! Calls a kernel, by string. Pass globalThreads = NULL, and cleanUp = true, to finally clean-up without executing.
         CV_EXPORTS double openCLExecuteKernelInterop(Context *clCxt ,
-                                                        const char **source, std::string kernelName,
+                                                        const char **source, String kernelName,
                                                         size_t globalThreads[3], size_t localThreads[3],
                                                         std::vector< std::pair<size_t, const void *> > &args,
                                                         int channels, int depth, const char *build_options,
@@ -161,13 +171,36 @@ namespace cv
 
         //! Calls a kernel, by file. Pass globalThreads = NULL, and cleanUp = true, to finally clean-up without executing.
         CV_EXPORTS double openCLExecuteKernelInterop(Context *clCxt ,
-                                                        const char **fileName, const int numFiles, std::string kernelName,
+                                                        const char **fileName, const int numFiles, String kernelName,
                                                         size_t globalThreads[3], size_t localThreads[3],
                                                         std::vector< std::pair<size_t, const void *> > &args,
                                                         int channels, int depth, const char *build_options,
                                                         bool finish = true, bool measureKernelTime = false,
                                                         bool cleanUp = true);
 
+        //! Enable or disable OpenCL program binary caching onto local disk
+        // After a program (*.cl files in opencl/ folder) is built at runtime, we allow the
+        // compiled OpenCL program to be cached to the path automatically as "path/*.clb"
+        // binary file, which will be reused when the OpenCV executable is started again.
+        //
+        // Caching mode is controlled by the following enums
+        // Notes
+        //   1. the feature is by default enabled when OpenCV is built in release mode.
+        //   2. the CACHE_DEBUG / CACHE_RELEASE flags only effectively work with MSVC compiler;
+        //      for GNU compilers, the function always treats the build as release mode (enabled by default).
+        enum
+        {
+            CACHE_NONE    = 0,        // do not cache OpenCL binary
+            CACHE_DEBUG   = 0x1 << 0, // cache OpenCL binary when built in debug mode (only work with MSVC)
+            CACHE_RELEASE = 0x1 << 1, // default behavior, only cache when built in release mode (only work with MSVC)
+            CACHE_ALL     = CACHE_DEBUG | CACHE_RELEASE, // always cache opencl binary
+            CACHE_UPDATE  = 0x1 << 2  // if the binary cache file with the same name is already on the disk, it will be updated.
+        };
+        CV_EXPORTS void setBinaryDiskCache(int mode = CACHE_RELEASE, cv::String path = "./");
+
+        //! set where binary cache to be saved to
+        CV_EXPORTS void setBinpath(const char *path);
+
         class CV_EXPORTS oclMatExpr;
         //////////////////////////////// oclMat ////////////////////////////////
         class CV_EXPORTS oclMat
@@ -212,6 +245,11 @@ namespace cv
             operator Mat() const;
             void download(cv::Mat &m) const;
 
+            //! convert to _InputArray
+            operator _InputArray();
+
+            //! convert to _OutputArray
+            operator _OutputArray();
 
             //! returns a new oclMatrix header for the specified row
             oclMat row(int y) const;
@@ -255,8 +293,10 @@ namespace cv
             void create(Size size, int type);
 
             //! allocates new oclMatrix with specified device memory type.
-            void createEx(int rows, int cols, int type, DevMemRW rw_type, DevMemType mem_type);
-            void createEx(Size size, int type, DevMemRW rw_type, DevMemType mem_type);
+            void createEx(int rows, int cols, int type,
+                          DevMemRW rw_type, DevMemType mem_type, void* hptr = 0);
+            void createEx(Size size, int type, DevMemRW rw_type,
+                          DevMemType mem_type, void* hptr = 0);
 
             //! decreases reference counter;
             // deallocate the data when reference counter reaches 0.
@@ -351,6 +391,9 @@ namespace cv
             int wholecols;
         };
 
+        // convert InputArray/OutputArray to oclMat references
+        CV_EXPORTS oclMat& getOclMatRef(InputArray src);
+        CV_EXPORTS oclMat& getOclMatRef(OutputArray src);
 
         ///////////////////// mat split and merge /////////////////////////////////
         //! Compose a multi-channel array from several single-channel arrays
@@ -395,6 +438,9 @@ namespace cv
         //! computes element-wise product of the two arrays (c = a * b)
         // supports all types except CV_8SC1,CV_8SC2,CV8SC3 and CV_8SC4
         CV_EXPORTS void multiply(const oclMat &a, const oclMat &b, oclMat &c, double scale = 1);
+        //! multiplies matrix to a number (dst = scalar * src)
+        // supports CV_32FC1 only
+        CV_EXPORTS void multiply(double scalar, const oclMat &src, oclMat &dst);
         //! computes element-wise quotient of the two arrays (c = a / b)
         // supports all types except CV_8SC1,CV_8SC2,CV8SC3 and CV_8SC4
         CV_EXPORTS void divide(const oclMat &a, const oclMat &b, oclMat &c, double scale = 1);
@@ -446,6 +492,7 @@ namespace cv
         // support all C1 types
 
         CV_EXPORTS void minMax(const oclMat &src, double *minVal, double *maxVal = 0, const oclMat &mask = oclMat());
+        CV_EXPORTS void minMax_buf(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask, oclMat& buf);
 
         //! finds global minimum and maximum array elements and returns their values with locations
         // support all C1 types
@@ -466,9 +513,21 @@ namespace cv
         CV_EXPORTS void calcHist(const oclMat &mat_src, oclMat &mat_hist);
         //! only 8UC1 and 256 bins is supported now
         CV_EXPORTS void equalizeHist(const oclMat &mat_src, oclMat &mat_dst);
+
+        //! only 8UC1 is supported now
+        CV_EXPORTS Ptr<cv::CLAHE> createCLAHE(double clipLimit = 40.0, Size tileGridSize = Size(8, 8));
+
         //! bilateralFilter
         // supports 8UC1 8UC4
-        CV_EXPORTS void bilateralFilter(const oclMat& src, oclMat& dst, int d, double sigmaColor, double sigmaSpave, int borderType=BORDER_DEFAULT);
+        CV_EXPORTS void bilateralFilter(const oclMat& src, oclMat& dst, int d, double sigmaColor, double sigmaSpace, int borderType=BORDER_DEFAULT);
+
+        //! Applies an adaptive bilateral filter to the input image
+        //  This is not truly a bilateral filter. Instead of using user provided fixed parameters,
+        //  the function calculates a constant at each window based on local standard deviation,
+        //  and use this constant to do filtering.
+        //  supports 8UC1 8UC3
+        CV_EXPORTS void adaptiveBilateralFilter(const oclMat& src, oclMat& dst, Size ksize, double sigmaSpace, Point anchor = Point(-1, -1), int borderType=BORDER_DEFAULT);
+
         //! computes exponent of each matrix element (b = e**a)
         // supports only CV_32FC1 type
         CV_EXPORTS void exp(const oclMat &a, oclMat &b);
@@ -529,9 +588,29 @@ namespace cv
         CV_EXPORTS oclMatExpr operator * (const oclMat &src1, const oclMat &src2);
         CV_EXPORTS oclMatExpr operator / (const oclMat &src1, const oclMat &src2);
 
-        //! computes convolution of two images
+        struct CV_EXPORTS ConvolveBuf
+        {
+            Size result_size;
+            Size block_size;
+            Size user_block_size;
+            Size dft_size;
+
+            oclMat image_spect, templ_spect, result_spect;
+            oclMat image_block, templ_block, result_data;
+
+            void create(Size image_size, Size templ_size);
+            static Size estimateBlockSize(Size result_size, Size templ_size);
+        };
+
+        //! computes convolution of two images, may use discrete Fourier transform
         //! support only CV_32FC1 type
-        CV_EXPORTS void convolve(const oclMat &image, const oclMat &temp1, oclMat &result);
+        CV_EXPORTS void convolve(const oclMat &image, const oclMat &temp1, oclMat &result, bool ccorr = false);
+        CV_EXPORTS void convolve(const oclMat &image, const oclMat &temp1, oclMat &result, bool ccorr, ConvolveBuf& buf);
+
+        //! Performs a per-element multiplication of two Fourier spectrums.
+        //! Only full (not packed) CV_32FC2 complex spectrums in the interleaved format are supported for now.
+        //! support only CV_32FC2 type
+        CV_EXPORTS void mulSpectrums(const oclMat &a, const oclMat &b, oclMat &c, int flags, float scale, bool conjB = false);
 
         CV_EXPORTS void cvtColor(const oclMat &src, oclMat &dst, int code , int dcn = 0);
 
@@ -639,7 +718,7 @@ namespace cv
         //! returns 2D filter with the specified kernel
         // supports CV_8UC1 and CV_8UC4 types
         CV_EXPORTS Ptr<BaseFilter_GPU> getLinearFilter_GPU(int srcType, int dstType, const Mat &kernel, const Size &ksize,
-                Point anchor = Point(-1, -1), int borderType = BORDER_DEFAULT);
+                const Point &anchor = Point(-1, -1), int borderType = BORDER_DEFAULT);
 
         //! returns the non-separable linear filter engine
         CV_EXPORTS Ptr<FilterEngine_GPU> createLinearFilter_GPU(int srcType, int dstType, const Mat &kernel,
@@ -672,6 +751,8 @@ namespace cv
         }
 
         //! applies non-separable 2D linear filter to the image
+        //  Note, at the moment this function only works when anchor point is in the kernel center
+        //  and kernel size supported is either 3x3 or 5x5; otherwise the function will fail to output valid result
         CV_EXPORTS void filter2D(const oclMat &src, oclMat &dst, int ddepth, const Mat &kernel,
                                  Point anchor = Point(-1, -1), int borderType = BORDER_DEFAULT);
 
@@ -774,24 +855,36 @@ namespace cv
         CV_EXPORTS void integral(const oclMat &src, oclMat &sum, oclMat &sqsum);
         CV_EXPORTS void integral(const oclMat &src, oclMat &sum);
         CV_EXPORTS void cornerHarris(const oclMat &src, oclMat &dst, int blockSize, int ksize, double k, int bordertype = cv::BORDER_DEFAULT);
+        CV_EXPORTS void cornerHarris_dxdy(const oclMat &src, oclMat &dst, oclMat &Dx, oclMat &Dy,
+            int blockSize, int ksize, double k, int bordertype = cv::BORDER_DEFAULT);
         CV_EXPORTS void cornerMinEigenVal(const oclMat &src, oclMat &dst, int blockSize, int ksize, int bordertype = cv::BORDER_DEFAULT);
+        CV_EXPORTS void cornerMinEigenVal_dxdy(const oclMat &src, oclMat &dst, oclMat &Dx, oclMat &Dy,
+            int blockSize, int ksize, int bordertype = cv::BORDER_DEFAULT);
+
+
+        /////////////////////////////////// ML ///////////////////////////////////////////
+
+        //! Compute closest centers for each lines in source and lable it after center's index
+        // supports CV_32FC1/CV_32FC2/CV_32FC4 data type
+        CV_EXPORTS void distanceToCenters(oclMat &dists, oclMat &labels, const oclMat &src, const oclMat &centers);
+
+        //!Does k-means procedure on GPU
+        // supports CV_32FC1/CV_32FC2/CV_32FC4 data type
+        CV_EXPORTS double kmeans(const oclMat &src, int K, oclMat &bestLabels,
+                                     TermCriteria criteria, int attemps, int flags, oclMat &centers);
+
 
         ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
         ///////////////////////////////////////////CascadeClassifier//////////////////////////////////////////////////////////////////
         ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-        class CV_EXPORTS_W OclCascadeClassifier : public  cv::CascadeClassifier
+        class CV_EXPORTS OclCascadeClassifier : public  cv::CascadeClassifier
         {
         public:
-            OclCascadeClassifier() {};
-            ~OclCascadeClassifier() {};
-
-            CvSeq* oclHaarDetectObjects(oclMat &gimg, CvMemStorage *storage, double scaleFactor,
-                                        int minNeighbors, int flags, CvSize minSize = cvSize(0, 0), CvSize maxSize = cvSize(0, 0));
+            void detectMultiScale(oclMat &image, CV_OUT std::vector<cv::Rect>& faces,
+                                  double scaleFactor = 1.1, int minNeighbors = 3, int flags = 0,
+                                  Size minSize = Size(), Size maxSize = Size());
         };
 
-
-
         /////////////////////////////// Pyramid /////////////////////////////////////
         CV_EXPORTS void pyrDown(const oclMat &src, oclMat &dst);
 
@@ -816,7 +909,6 @@ namespace cv
             std::vector<oclMat> image_sqsums;
         };
 
-
         //! computes the proximity map for the raster template and the image where the template is searched for
         // Supports TM_SQDIFF, TM_SQDIFF_NORMED, TM_CCORR, TM_CCORR_NORMED, TM_CCOEFF, TM_CCOEFF_NORMED for type 8UC1 and 8UC4
         // Supports TM_SQDIFF, TM_CCORR for type 32FC1 and 32FC4
@@ -830,68 +922,36 @@ namespace cv
 
 
         ///////////////////////////////////////////// Canny /////////////////////////////////////////////
-
         struct CV_EXPORTS CannyBuf;
 
-
-
         //! compute edges of the input image using Canny operator
-
         // Support CV_8UC1 only
-
         CV_EXPORTS void Canny(const oclMat &image, oclMat &edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false);
-
         CV_EXPORTS void Canny(const oclMat &image, CannyBuf &buf, oclMat &edges, double low_thresh, double high_thresh, int apperture_size = 3, bool L2gradient = false);
-
         CV_EXPORTS void Canny(const oclMat &dx, const oclMat &dy, oclMat &edges, double low_thresh, double high_thresh, bool L2gradient = false);
-
         CV_EXPORTS void Canny(const oclMat &dx, const oclMat &dy, CannyBuf &buf, oclMat &edges, double low_thresh, double high_thresh, bool L2gradient = false);
 
-
-
         struct CV_EXPORTS CannyBuf
-
         {
-
             CannyBuf() : counter(NULL) {}
-
             ~CannyBuf()
             {
                 release();
             }
-
             explicit CannyBuf(const Size &image_size, int apperture_size = 3) : counter(NULL)
-
             {
-
                 create(image_size, apperture_size);
-
             }
-
             CannyBuf(const oclMat &dx_, const oclMat &dy_);
-
-
-
             void create(const Size &image_size, int apperture_size = 3);
-
-
-
             void release();
 
-
-
             oclMat dx, dy;
-
             oclMat dx_buf, dy_buf;
-
-            oclMat edgeBuf;
-
+            oclMat magBuf, mapBuf;
             oclMat trackBuf1, trackBuf2;
-
             void *counter;
-
             Ptr<FilterEngine_GPU> filterDX, filterDY;
-
         };
 
         ///////////////////////////////////////// Hough Transform /////////////////////////////////////////
@@ -1088,167 +1148,15 @@ namespace cv
         };
 
 
-
-        //! Speeded up robust features, port from GPU module.
-        ////////////////////////////////// SURF //////////////////////////////////////////
-
-        class CV_EXPORTS SURF_OCL
-
-        {
-
-        public:
-
-            enum KeypointLayout
-
-            {
-
-                X_ROW = 0,
-
-                Y_ROW,
-
-                LAPLACIAN_ROW,
-
-                OCTAVE_ROW,
-
-                SIZE_ROW,
-
-                ANGLE_ROW,
-
-                HESSIAN_ROW,
-
-                ROWS_COUNT
-
-            };
-
-
-
-            //! the default constructor
-
-            SURF_OCL();
-
-            //! the full constructor taking all the necessary parameters
-
-            explicit SURF_OCL(double _hessianThreshold, int _nOctaves = 4,
-
-                              int _nOctaveLayers = 2, bool _extended = false, float _keypointsRatio = 0.01f, bool _upright = false);
-
-
-
-            //! returns the descriptor size in float's (64 or 128)
-
-            int descriptorSize() const;
-
-
-
-            //! upload host keypoints to device memory
-
-            void uploadKeypoints(const std::vector<cv::KeyPoint> &keypoints, oclMat &keypointsocl);
-
-            //! download keypoints from device to host memory
-
-            void downloadKeypoints(const oclMat &keypointsocl, std::vector<KeyPoint> &keypoints);
-
-
-
-            //! download descriptors from device to host memory
-
-            void downloadDescriptors(const oclMat &descriptorsocl, std::vector<float> &descriptors);
-
-
-
-            //! finds the keypoints using fast hessian detector used in SURF
-
-            //! supports CV_8UC1 images
-
-            //! keypoints will have nFeature cols and 6 rows
-
-            //! keypoints.ptr<float>(X_ROW)[i] will contain x coordinate of i'th feature
-
-            //! keypoints.ptr<float>(Y_ROW)[i] will contain y coordinate of i'th feature
-
-            //! keypoints.ptr<float>(LAPLACIAN_ROW)[i] will contain laplacian sign of i'th feature
-
-            //! keypoints.ptr<float>(OCTAVE_ROW)[i] will contain octave of i'th feature
-
-            //! keypoints.ptr<float>(SIZE_ROW)[i] will contain size of i'th feature
-
-            //! keypoints.ptr<float>(ANGLE_ROW)[i] will contain orientation of i'th feature
-
-            //! keypoints.ptr<float>(HESSIAN_ROW)[i] will contain response of i'th feature
-
-            void operator()(const oclMat &img, const oclMat &mask, oclMat &keypoints);
-
-            //! finds the keypoints and computes their descriptors.
-
-            //! Optionally it can compute descriptors for the user-provided keypoints and recompute keypoints direction
-
-            void operator()(const oclMat &img, const oclMat &mask, oclMat &keypoints, oclMat &descriptors,
-
-                            bool useProvidedKeypoints = false);
-
-
-
-            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints);
-
-            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints, oclMat &descriptors,
-
-                            bool useProvidedKeypoints = false);
-
-
-
-            void operator()(const oclMat &img, const oclMat &mask, std::vector<KeyPoint> &keypoints, std::vector<float> &descriptors,
-
-                            bool useProvidedKeypoints = false);
-
-
-
-            void releaseMemory();
-
-
-
-            // SURF parameters
-
-            float hessianThreshold;
-
-            int nOctaves;
-
-            int nOctaveLayers;
-
-            bool extended;
-
-            bool upright;
-
-
-
-            //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
-
-            float keypointsRatio;
-
-
-
-            oclMat sum, mask1, maskSum, intBuffer;
-
-
-
-            oclMat det, trace;
-
-
-
-            oclMat maxPosBuffer;
-
-        };
-
         ////////////////////////feature2d_ocl/////////////////
         /****************************************************************************************\
         *                                      Distance                                          *
         \****************************************************************************************/
-
         template<typename T>
         struct CV_EXPORTS Accumulator
         {
             typedef T Type;
         };
-
         template<> struct Accumulator<unsigned char>
         {
             typedef float Type;
@@ -1322,470 +1230,329 @@ namespace cv
         {
         public:
             enum DistType {L1Dist = 0, L2Dist, HammingDist};
-
             explicit BruteForceMatcher_OCL_base(DistType distType = L2Dist);
 
-
-
             // Add descriptors to train descriptor collection
-
             void add(const std::vector<oclMat> &descCollection);
 
-
-
             // Get train descriptors collection
-
             const std::vector<oclMat> &getTrainDescriptors() const;
 
-
-
             // Clear train descriptors collection
-
             void clear();
 
-
-
             // Return true if there are not train descriptors in collection
-
             bool empty() const;
 
-
-
             // Return true if the matcher supports mask in match methods
-
             bool isMaskSupported() const;
 
-
-
             // Find one best match for each query descriptor
-
             void matchSingle(const oclMat &query, const oclMat &train,
-
                              oclMat &trainIdx, oclMat &distance,
-
                              const oclMat &mask = oclMat());
 
-
-
             // Download trainIdx and distance and convert it to CPU vector with DMatch
-
             static void matchDownload(const oclMat &trainIdx, const oclMat &distance, std::vector<DMatch> &matches);
-
             // Convert trainIdx and distance to vector with DMatch
-
             static void matchConvert(const Mat &trainIdx, const Mat &distance, std::vector<DMatch> &matches);
 
-
-
             // Find one best match for each query descriptor
-
             void match(const oclMat &query, const oclMat &train, std::vector<DMatch> &matches, const oclMat &mask = oclMat());
 
-
-
             // Make gpu collection of trains and masks in suitable format for matchCollection function
-
             void makeGpuCollection(oclMat &trainCollection, oclMat &maskCollection, const std::vector<oclMat> &masks = std::vector<oclMat>());
 
-
-
             // Find one best match from train collection for each query descriptor
-
             void matchCollection(const oclMat &query, const oclMat &trainCollection,
-
                                  oclMat &trainIdx, oclMat &imgIdx, oclMat &distance,
-
                                  const oclMat &masks = oclMat());
 
-
-
             // Download trainIdx, imgIdx and distance and convert it to vector with DMatch
-
             static void matchDownload(const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance, std::vector<DMatch> &matches);
-
             // Convert trainIdx, imgIdx and distance to vector with DMatch
-
             static void matchConvert(const Mat &trainIdx, const Mat &imgIdx, const Mat &distance, std::vector<DMatch> &matches);
 
-
-
             // Find one best match from train collection for each query descriptor.
-
             void match(const oclMat &query, std::vector<DMatch> &matches, const std::vector<oclMat> &masks = std::vector<oclMat>());
 
-
-
             // Find k best matches for each query descriptor (in increasing order of distances)
-
             void knnMatchSingle(const oclMat &query, const oclMat &train,
-
                                 oclMat &trainIdx, oclMat &distance, oclMat &allDist, int k,
-
                                 const oclMat &mask = oclMat());
 
-
-
             // Download trainIdx and distance and convert it to vector with DMatch
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             static void knnMatchDownload(const oclMat &trainIdx, const oclMat &distance,
-
                                          std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
-
             // Convert trainIdx and distance to vector with DMatch
-
             static void knnMatchConvert(const Mat &trainIdx, const Mat &distance,
-
                                         std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
 
-
-
             // Find k best matches for each query descriptor (in increasing order of distances).
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             void knnMatch(const oclMat &query, const oclMat &train,
-
                           std::vector< std::vector<DMatch> > &matches, int k, const oclMat &mask = oclMat(),
-
                           bool compactResult = false);
 
-
-
             // Find k best matches from train collection for each query descriptor (in increasing order of distances)
-
             void knnMatch2Collection(const oclMat &query, const oclMat &trainCollection,
-
                                      oclMat &trainIdx, oclMat &imgIdx, oclMat &distance,
-
                                      const oclMat &maskCollection = oclMat());
 
-
-
             // Download trainIdx and distance and convert it to vector with DMatch
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             static void knnMatch2Download(const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance,
-
                                           std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
-
             // Convert trainIdx and distance to vector with DMatch
-
             static void knnMatch2Convert(const Mat &trainIdx, const Mat &imgIdx, const Mat &distance,
-
                                          std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
 
-
-
             // Find k best matches  for each query descriptor (in increasing order of distances).
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             void knnMatch(const oclMat &query, std::vector< std::vector<DMatch> > &matches, int k,
-
                           const std::vector<oclMat> &masks = std::vector<oclMat>(), bool compactResult = false);
 
-
-
             // Find best matches for each query descriptor which have distance less than maxDistance.
-
             // nMatches.at<int>(0, queryIdx) will contain matches count for queryIdx.
-
             // carefully nMatches can be greater than trainIdx.cols - it means that matcher didn't find all matches,
-
             // because it didn't have enough memory.
-
             // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nTrain / 100), 10),
-
             // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
-
             // Matches doesn't sorted.
-
             void radiusMatchSingle(const oclMat &query, const oclMat &train,
-
                                    oclMat &trainIdx, oclMat &distance, oclMat &nMatches, float maxDistance,
-
                                    const oclMat &mask = oclMat());
 
-
-
             // Download trainIdx, nMatches and distance and convert it to vector with DMatch.
-
             // matches will be sorted in increasing order of distances.
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             static void radiusMatchDownload(const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches,
-
                                             std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
-
             // Convert trainIdx, nMatches and distance to vector with DMatch.
-
             static void radiusMatchConvert(const Mat &trainIdx, const Mat &distance, const Mat &nMatches,
-
                                            std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
 
-
-
             // Find best matches for each query descriptor which have distance less than maxDistance
-
             // in increasing order of distances).
-
             void radiusMatch(const oclMat &query, const oclMat &train,
-
                              std::vector< std::vector<DMatch> > &matches, float maxDistance,
-
                              const oclMat &mask = oclMat(), bool compactResult = false);
 
-
-
             // Find best matches for each query descriptor which have distance less than maxDistance.
-
             // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x max((nQuery / 100), 10),
-
             // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
-
             // Matches doesn't sorted.
-
             void radiusMatchCollection(const oclMat &query, oclMat &trainIdx, oclMat &imgIdx, oclMat &distance, oclMat &nMatches, float maxDistance,
-
                                        const std::vector<oclMat> &masks = std::vector<oclMat>());
 
-
-
             // Download trainIdx, imgIdx, nMatches and distance and convert it to vector with DMatch.
-
             // matches will be sorted in increasing order of distances.
-
             // compactResult is used when mask is not empty. If compactResult is false matches
-
             // vector will have the same size as queryDescriptors rows. If compactResult is true
-
             // matches vector will not contain matches for fully masked out query descriptors.
-
             static void radiusMatchDownload(const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance, const oclMat &nMatches,
-
                                             std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
-
             // Convert trainIdx, nMatches and distance to vector with DMatch.
-
             static void radiusMatchConvert(const Mat &trainIdx, const Mat &imgIdx, const Mat &distance, const Mat &nMatches,
-
                                            std::vector< std::vector<DMatch> > &matches, bool compactResult = false);
 
-
-
             // Find best matches from train collection for each query descriptor which have distance less than
-
             // maxDistance (in increasing order of distances).
-
             void radiusMatch(const oclMat &query, std::vector< std::vector<DMatch> > &matches, float maxDistance,
-
                              const std::vector<oclMat> &masks = std::vector<oclMat>(), bool compactResult = false);
 
-
-
             DistType distType;
 
-
-
         private:
-
             std::vector<oclMat> trainDescCollection;
-
         };
 
-
-
         template <class Distance>
-
         class CV_EXPORTS BruteForceMatcher_OCL;
 
-
-
         template <typename T>
-
         class CV_EXPORTS BruteForceMatcher_OCL< L1<T> > : public BruteForceMatcher_OCL_base
-
         {
-
         public:
-
             explicit BruteForceMatcher_OCL() : BruteForceMatcher_OCL_base(L1Dist) {}
-
             explicit BruteForceMatcher_OCL(L1<T> /*d*/) : BruteForceMatcher_OCL_base(L1Dist) {}
-
         };
-
         template <typename T>
-
         class CV_EXPORTS BruteForceMatcher_OCL< L2<T> > : public BruteForceMatcher_OCL_base
-
         {
-
         public:
-
             explicit BruteForceMatcher_OCL() : BruteForceMatcher_OCL_base(L2Dist) {}
-
             explicit BruteForceMatcher_OCL(L2<T> /*d*/) : BruteForceMatcher_OCL_base(L2Dist) {}
-
         };
-
         template <> class CV_EXPORTS BruteForceMatcher_OCL< Hamming > : public BruteForceMatcher_OCL_base
-
         {
-
         public:
-
             explicit BruteForceMatcher_OCL() : BruteForceMatcher_OCL_base(HammingDist) {}
-
             explicit BruteForceMatcher_OCL(Hamming /*d*/) : BruteForceMatcher_OCL_base(HammingDist) {}
-
         };
 
+        class CV_EXPORTS BFMatcher_OCL : public BruteForceMatcher_OCL_base
+        {
+        public:
+            explicit BFMatcher_OCL(int norm = NORM_L2) : BruteForceMatcher_OCL_base(norm == NORM_L1 ? L1Dist : norm == NORM_L2 ? L2Dist : HammingDist) {}
+        };
 
+        class CV_EXPORTS GoodFeaturesToTrackDetector_OCL
+        {
+        public:
+            explicit GoodFeaturesToTrackDetector_OCL(int maxCorners = 1000, double qualityLevel = 0.01, double minDistance = 0.0,
+                int blockSize = 3, bool useHarrisDetector = false, double harrisK = 0.04);
+
+            //! return 1 rows matrix with CV_32FC2 type
+            void operator ()(const oclMat& image, oclMat& corners, const oclMat& mask = oclMat());
+            //! download points of type Point2f to a vector. the vector's content will be erased
+            void downloadPoints(const oclMat &points, std::vector<Point2f> &points_v);
+
+            int maxCorners;
+            double qualityLevel;
+            double minDistance;
+
+            int blockSize;
+            bool useHarrisDetector;
+            double harrisK;
+            void releaseMemory()
+            {
+                Dx_.release();
+                Dy_.release();
+                eig_.release();
+                minMaxbuf_.release();
+                tmpCorners_.release();
+            }
+        private:
+            oclMat Dx_;
+            oclMat Dy_;
+            oclMat eig_;
+            oclMat minMaxbuf_;
+            oclMat tmpCorners_;
+        };
+
+        inline GoodFeaturesToTrackDetector_OCL::GoodFeaturesToTrackDetector_OCL(int maxCorners_, double qualityLevel_, double minDistance_,
+            int blockSize_, bool useHarrisDetector_, double harrisK_)
+        {
+            maxCorners = maxCorners_;
+            qualityLevel = qualityLevel_;
+            minDistance = minDistance_;
+            blockSize = blockSize_;
+            useHarrisDetector = useHarrisDetector_;
+            harrisK = harrisK_;
+        }
 
         /////////////////////////////// PyrLKOpticalFlow /////////////////////////////////////
 
         class CV_EXPORTS PyrLKOpticalFlow
-
         {
-
         public:
-
             PyrLKOpticalFlow()
-
             {
-
                 winSize = Size(21, 21);
-
                 maxLevel = 3;
-
                 iters = 30;
-
                 derivLambda = 0.5;
-
                 useInitialFlow = false;
-
                 minEigThreshold = 1e-4f;
-
                 getMinEigenVals = false;
-
                 isDeviceArch11_ = false;
-
             }
 
-
-
             void sparse(const oclMat &prevImg, const oclMat &nextImg, const oclMat &prevPts, oclMat &nextPts,
-
                         oclMat &status, oclMat *err = 0);
 
-
-
             void dense(const oclMat &prevImg, const oclMat &nextImg, oclMat &u, oclMat &v, oclMat *err = 0);
 
-
-
             Size winSize;
-
             int maxLevel;
-
             int iters;
-
             double derivLambda;
-
             bool useInitialFlow;
-
             float minEigThreshold;
-
             bool getMinEigenVals;
 
-
-
             void releaseMemory()
-
             {
-
                 dx_calcBuf_.release();
-
                 dy_calcBuf_.release();
 
-
-
                 prevPyr_.clear();
-
                 nextPyr_.clear();
 
-
-
                 dx_buf_.release();
-
                 dy_buf_.release();
-
             }
 
-
-
         private:
-
             void calcSharrDeriv(const oclMat &src, oclMat &dx, oclMat &dy);
 
-
-
             void buildImagePyramid(const oclMat &img0, std::vector<oclMat> &pyr, bool withBorder);
 
-
-
             oclMat dx_calcBuf_;
-
             oclMat dy_calcBuf_;
 
-
-
             std::vector<oclMat> prevPyr_;
-
             std::vector<oclMat> nextPyr_;
 
-
-
             oclMat dx_buf_;
-
             oclMat dy_buf_;
 
-
-
             oclMat uPyr_[2];
-
             oclMat vPyr_[2];
 
-
-
             bool isDeviceArch11_;
-
         };
+
+        class CV_EXPORTS FarnebackOpticalFlow
+        {
+        public:
+            FarnebackOpticalFlow();
+
+            int numLevels;
+            double pyrScale;
+            bool fastPyramids;
+            int winSize;
+            int numIters;
+            int polyN;
+            double polySigma;
+            int flags;
+
+            void operator ()(const oclMat &frame0, const oclMat &frame1, oclMat &flowx, oclMat &flowy);
+
+            void releaseMemory();
+
+        private:
+            void prepareGaussian(
+                int n, double sigma, float *g, float *xg, float *xxg,
+                double &ig11, double &ig03, double &ig33, double &ig55);
+
+            void setPolynomialExpansionConsts(int n, double sigma);
+
+            void updateFlow_boxFilter(
+                const oclMat& R0, const oclMat& R1, oclMat& flowx, oclMat &flowy,
+                oclMat& M, oclMat &bufM, int blockSize, bool updateMatrices);
+
+            void updateFlow_gaussianBlur(
+                const oclMat& R0, const oclMat& R1, oclMat& flowx, oclMat& flowy,
+                oclMat& M, oclMat &bufM, int blockSize, bool updateMatrices);
+
+            oclMat frames_[2];
+            oclMat pyrLevel_[2], M_, bufM_, R_[2], blurredFrame_[2];
+            std::vector<oclMat> pyramid0_, pyramid1_;
+        };
+
         //////////////// build warping maps ////////////////////
         //! builds plane warping maps
         CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat &R, const Mat &T, float scale, oclMat &map_x, oclMat &map_y);
@@ -1821,6 +1588,372 @@ namespace cv
 
         //! computes moments of the rasterized shape or a vector of points
         CV_EXPORTS Moments ocl_moments(InputArray _array, bool binaryImage);
+
+        class CV_EXPORTS StereoBM_OCL
+        {
+        public:
+            enum { BASIC_PRESET = 0, PREFILTER_XSOBEL = 1 };
+
+            enum { DEFAULT_NDISP = 64, DEFAULT_WINSZ = 19 };
+
+            //! the default constructor
+            StereoBM_OCL();
+            //! the full constructor taking the camera-specific preset, number of disparities and the SAD window size. ndisparities must be multiple of 8.
+            StereoBM_OCL(int preset, int ndisparities = DEFAULT_NDISP, int winSize = DEFAULT_WINSZ);
+
+            //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair
+            //! Output disparity has CV_8U type.
+            void operator() ( const oclMat &left, const oclMat &right, oclMat &disparity);
+
+            //! Some heuristics that tries to estmate
+            // if current GPU will be faster then CPU in this algorithm.
+            // It queries current active device.
+            static bool checkIfGpuCallReasonable();
+
+            int preset;
+            int ndisp;
+            int winSize;
+
+            // If avergeTexThreshold  == 0 => post procesing is disabled
+            // If avergeTexThreshold != 0 then disparity is set 0 in each point (x,y) where for left image
+            // SumOfHorizontalGradiensInWindow(x, y, winSize) < (winSize * winSize) * avergeTexThreshold
+            // i.e. input left image is low textured.
+            float avergeTexThreshold;
+        private:
+            oclMat minSSD, leBuf, riBuf;
+        };
+
+        class CV_EXPORTS StereoBeliefPropagation
+        {
+        public:
+            enum { DEFAULT_NDISP  = 64 };
+            enum { DEFAULT_ITERS  = 5  };
+            enum { DEFAULT_LEVELS = 5  };
+            static void estimateRecommendedParams(int width, int height, int &ndisp, int &iters, int &levels);
+            explicit StereoBeliefPropagation(int ndisp  = DEFAULT_NDISP,
+                                             int iters  = DEFAULT_ITERS,
+                                             int levels = DEFAULT_LEVELS,
+                                             int msg_type = CV_16S);
+            StereoBeliefPropagation(int ndisp, int iters, int levels,
+                                    float max_data_term, float data_weight,
+                                    float max_disc_term, float disc_single_jump,
+                                    int msg_type = CV_32F);
+            void operator()(const oclMat &left, const oclMat &right, oclMat &disparity);
+            void operator()(const oclMat &data, oclMat &disparity);
+            int ndisp;
+            int iters;
+            int levels;
+            float max_data_term;
+            float data_weight;
+            float max_disc_term;
+            float disc_single_jump;
+            int msg_type;
+        private:
+            oclMat u, d, l, r, u2, d2, l2, r2;
+            std::vector<oclMat> datas;
+            oclMat out;
+        };
+
+        class CV_EXPORTS StereoConstantSpaceBP
+        {
+        public:
+            enum { DEFAULT_NDISP    = 128 };
+            enum { DEFAULT_ITERS    = 8   };
+            enum { DEFAULT_LEVELS   = 4   };
+            enum { DEFAULT_NR_PLANE = 4   };
+            static void estimateRecommendedParams(int width, int height, int &ndisp, int &iters, int &levels, int &nr_plane);
+            explicit StereoConstantSpaceBP(
+                int ndisp    = DEFAULT_NDISP,
+                int iters    = DEFAULT_ITERS,
+                int levels   = DEFAULT_LEVELS,
+                int nr_plane = DEFAULT_NR_PLANE,
+                int msg_type = CV_32F);
+            StereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane,
+                float max_data_term, float data_weight, float max_disc_term, float disc_single_jump,
+                int min_disp_th = 0,
+                int msg_type = CV_32F);
+            void operator()(const oclMat &left, const oclMat &right, oclMat &disparity);
+            int ndisp;
+            int iters;
+            int levels;
+            int nr_plane;
+            float max_data_term;
+            float data_weight;
+            float max_disc_term;
+            float disc_single_jump;
+            int min_disp_th;
+            int msg_type;
+            bool use_local_init_data_cost;
+        private:
+            oclMat u[2], d[2], l[2], r[2];
+            oclMat disp_selected_pyr[2];
+            oclMat data_cost;
+            oclMat data_cost_selected;
+            oclMat temp;
+            oclMat out;
+        };
+
+        // Implementation of the Zach, Pock and Bischof Dual TV-L1 Optical Flow method
+        //
+        // see reference:
+        //   [1] C. Zach, T. Pock and H. Bischof, "A Duality Based Approach for Realtime TV-L1 Optical Flow".
+        //   [2] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
+        class CV_EXPORTS OpticalFlowDual_TVL1_OCL
+        {
+        public:
+            OpticalFlowDual_TVL1_OCL();
+
+            void operator ()(const oclMat& I0, const oclMat& I1, oclMat& flowx, oclMat& flowy);
+
+            void collectGarbage();
+
+            /**
+            * Time step of the numerical scheme.
+            */
+            double tau;
+
+            /**
+            * Weight parameter for the data term, attachment parameter.
+            * This is the most relevant parameter, which determines the smoothness of the output.
+            * The smaller this parameter is, the smoother the solutions we obtain.
+            * It depends on the range of motions of the images, so its value should be adapted to each image sequence.
+            */
+            double lambda;
+
+            /**
+            * Weight parameter for (u - v)^2, tightness parameter.
+            * It serves as a link between the attachment and the regularization terms.
+            * In theory, it should have a small value in order to maintain both parts in correspondence.
+            * The method is stable for a large range of values of this parameter.
+            */
+            double theta;
+
+            /**
+            * Number of scales used to create the pyramid of images.
+            */
+            int nscales;
+
+            /**
+            * Number of warpings per scale.
+            * Represents the number of times that I1(x+u0) and grad( I1(x+u0) ) are computed per scale.
+            * This is a parameter that assures the stability of the method.
+            * It also affects the running time, so it is a compromise between speed and accuracy.
+            */
+            int warps;
+
+            /**
+            * Stopping criterion threshold used in the numerical scheme, which is a trade-off between precision and running time.
+            * A small value will yield more accurate solutions at the expense of a slower convergence.
+            */
+            double epsilon;
+
+            /**
+            * Stopping criterion iterations number used in the numerical scheme.
+            */
+            int iterations;
+
+            bool useInitialFlow;
+
+        private:
+            void procOneScale(const oclMat& I0, const oclMat& I1, oclMat& u1, oclMat& u2);
+
+            std::vector<oclMat> I0s;
+            std::vector<oclMat> I1s;
+            std::vector<oclMat> u1s;
+            std::vector<oclMat> u2s;
+
+            oclMat I1x_buf;
+            oclMat I1y_buf;
+
+            oclMat I1w_buf;
+            oclMat I1wx_buf;
+            oclMat I1wy_buf;
+
+            oclMat grad_buf;
+            oclMat rho_c_buf;
+
+            oclMat p11_buf;
+            oclMat p12_buf;
+            oclMat p21_buf;
+            oclMat p22_buf;
+
+            oclMat diff_buf;
+            oclMat norm_buf;
+        };
+        // current supported sorting methods
+        enum
+        {
+            SORT_BITONIC,   // only support power-of-2 buffer size
+            SORT_SELECTION, // cannot sort duplicate keys
+            SORT_MERGE,
+            SORT_RADIX      // only support signed int/float keys(CV_32S/CV_32F)
+        };
+        //! Returns the sorted result of all the elements in input based on equivalent keys.
+        //
+        //  The element unit in the values to be sorted is determined from the data type,
+        //  i.e., a CV_32FC2 input {a1a2, b1b2} will be considered as two elements, regardless its
+        //  matrix dimension.
+        //  both keys and values will be sorted inplace
+        //  Key needs to be single channel oclMat.
+        //
+        //  Example:
+        //  input -
+        //    keys   = {2,    3,   1}   (CV_8UC1)
+        //    values = {10,5, 4,3, 6,2} (CV_8UC2)
+        //  sortByKey(keys, values, SORT_SELECTION, false);
+        //  output -
+        //    keys   = {1,    2,   3}   (CV_8UC1)
+        //    values = {6,2, 10,5, 4,3} (CV_8UC2)
+        void CV_EXPORTS sortByKey(oclMat& keys, oclMat& values, int method, bool isGreaterThan = false);
+        /*!Base class for MOG and MOG2!*/
+        class CV_EXPORTS BackgroundSubtractor
+        {
+        public:
+            //! the virtual destructor
+            virtual ~BackgroundSubtractor();
+            //! the update operator that takes the next video frame and returns the current foreground mask as 8-bit binary image.
+            virtual void operator()(const oclMat& image, oclMat& fgmask, float learningRate);
+
+            //! computes a background image
+            virtual void getBackgroundImage(oclMat& backgroundImage) const = 0;
+        };
+                /*!
+        Gaussian Mixture-based Backbround/Foreground Segmentation Algorithm
+
+        The class implements the following algorithm:
+        "An improved adaptive background mixture model for real-time tracking with shadow detection"
+        P. KadewTraKuPong and R. Bowden,
+        Proc. 2nd European Workshp on Advanced Video-Based Surveillance Systems, 2001."
+        http://personal.ee.surrey.ac.uk/Personal/R.Bowden/publications/avbs01/avbs01.pdf
+        */
+        class CV_EXPORTS MOG: public cv::ocl::BackgroundSubtractor
+        {
+        public:
+            //! the default constructor
+            MOG(int nmixtures = -1);
+
+            //! re-initiaization method
+            void initialize(Size frameSize, int frameType);
+
+            //! the update operator
+            void operator()(const oclMat& frame, oclMat& fgmask, float learningRate = 0.f);
+
+            //! computes a background image which are the mean of all background gaussians
+            void getBackgroundImage(oclMat& backgroundImage) const;
+
+            //! releases all inner buffers
+            void release();
+
+            int history;
+            float varThreshold;
+            float backgroundRatio;
+            float noiseSigma;
+
+        private:
+            int nmixtures_;
+
+            Size frameSize_;
+            int frameType_;
+            int nframes_;
+
+            oclMat weight_;
+            oclMat sortKey_;
+            oclMat mean_;
+            oclMat var_;
+        };
+
+        /*!
+        The class implements the following algorithm:
+        "Improved adaptive Gausian mixture model for background subtraction"
+        Z.Zivkovic
+        International Conference Pattern Recognition, UK, August, 2004.
+        http://www.zoranz.net/Publications/zivkovic2004ICPR.pdf
+        */
+        class CV_EXPORTS MOG2: public cv::ocl::BackgroundSubtractor
+        {
+        public:
+            //! the default constructor
+            MOG2(int nmixtures = -1);
+
+            //! re-initiaization method
+            void initialize(Size frameSize, int frameType);
+
+            //! the update operator
+            void operator()(const oclMat& frame, oclMat& fgmask, float learningRate = -1.0f);
+
+            //! computes a background image which are the mean of all background gaussians
+            void getBackgroundImage(oclMat& backgroundImage) const;
+
+            //! releases all inner buffers
+            void release();
+
+            // parameters
+            // you should call initialize after parameters changes
+
+            int history;
+
+            //! here it is the maximum allowed number of mixture components.
+            //! Actual number is determined dynamically per pixel
+            float varThreshold;
+            // threshold on the squared Mahalanobis distance to decide if it is well described
+            // by the background model or not. Related to Cthr from the paper.
+            // This does not influence the update of the background. A typical value could be 4 sigma
+            // and that is varThreshold=4*4=16; Corresponds to Tb in the paper.
+
+            /////////////////////////
+            // less important parameters - things you might change but be carefull
+            ////////////////////////
+
+            float backgroundRatio;
+            // corresponds to fTB=1-cf from the paper
+            // TB - threshold when the component becomes significant enough to be included into
+            // the background model. It is the TB=1-cf from the paper. So I use cf=0.1 => TB=0.
+            // For alpha=0.001 it means that the mode should exist for approximately 105 frames before
+            // it is considered foreground
+            // float noiseSigma;
+            float varThresholdGen;
+
+            //correspondts to Tg - threshold on the squared Mahalan. dist. to decide
+            //when a sample is close to the existing components. If it is not close
+            //to any a new component will be generated. I use 3 sigma => Tg=3*3=9.
+            //Smaller Tg leads to more generated components and higher Tg might make
+            //lead to small number of components but they can grow too large
+            float fVarInit;
+            float fVarMin;
+            float fVarMax;
+
+            //initial variance  for the newly generated components.
+            //It will will influence the speed of adaptation. A good guess should be made.
+            //A simple way is to estimate the typical standard deviation from the images.
+            //I used here 10 as a reasonable value
+            // min and max can be used to further control the variance
+            float fCT; //CT - complexity reduction prior
+            //this is related to the number of samples needed to accept that a component
+            //actually exists. We use CT=0.05 of all the samples. By setting CT=0 you get
+            //the standard Stauffer&Grimson algorithm (maybe not exact but very similar)
+
+            //shadow detection parameters
+            bool bShadowDetection; //default 1 - do shadow detection
+            unsigned char nShadowDetection; //do shadow detection - insert this value as the detection result - 127 default value
+            float fTau;
+            // Tau - shadow threshold. The shadow is detected if the pixel is darker
+            //version of the background. Tau is a threshold on how much darker the shadow can be.
+            //Tau= 0.5 means that if pixel is more than 2 times darker then it is not shadow
+            //See: Prati,Mikic,Trivedi,Cucchiarra,"Detecting Moving Shadows...",IEEE PAMI,2003.
+
+        private:
+            int nmixtures_;
+
+            Size frameSize_;
+            int frameType_;
+            int nframes_;
+
+            oclMat weight_;
+            oclMat variance_;
+            oclMat mean_;
+
+            oclMat bgmodelUsedModes_; //keep track of number of modes per pixel
+        };
     }
 }
 #if defined _MSC_VER && _MSC_VER >= 1200
diff --git a/modules/ocl/include/opencv2/ocl/matrix_operations.hpp b/modules/ocl/include/opencv2/ocl/matrix_operations.hpp
index fcd847cb6b..1361367fca 100644
--- a/modules/ocl/include/opencv2/ocl/matrix_operations.hpp
+++ b/modules/ocl/include/opencv2/ocl/matrix_operations.hpp
@@ -41,8 +41,8 @@
 //
 //M*/
 
-#ifndef __OPENCV_GPU_MATRIX_OPERATIONS_HPP__
-#define __OPENCV_GPU_MATRIX_OPERATIONS_HPP__
+#ifndef __OPENCV_OCL_MATRIX_OPERATIONS_HPP__
+#define __OPENCV_OCL_MATRIX_OPERATIONS_HPP__
 
 namespace cv
 {
@@ -211,8 +211,8 @@ namespace cv
         {
             flags &= roi.width < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
             offset += roi.y * step + roi.x * elemSize();
-            CV_Assert( 0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols &&
-                       0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows );
+            CV_Assert( 0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.wholecols &&
+                       0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.wholerows );
             if( refcount )
                 CV_XADD(refcount, 1);
             if( rows <= 0 || cols <= 0 )
@@ -459,28 +459,28 @@ namespace cv
         inline uchar *oclMat::ptr(int y)
         {
             CV_DbgAssert( (unsigned)y < (unsigned)rows );
-            CV_Error(CV_GpuNotSupported, "This function hasn't been supported yet.\n");
+            CV_Error(Error::GpuNotSupported, "This function hasn't been supported yet.\n");
             return data + step * y;
         }
 
         inline const uchar *oclMat::ptr(int y) const
         {
             CV_DbgAssert( (unsigned)y < (unsigned)rows );
-            CV_Error(CV_GpuNotSupported, "This function hasn't been supported yet.\n");
+            CV_Error(Error::GpuNotSupported, "This function hasn't been supported yet.\n");
             return data + step * y;
         }
 
         template<typename _Tp> inline _Tp *oclMat::ptr(int y)
         {
             CV_DbgAssert( (unsigned)y < (unsigned)rows );
-            CV_Error(CV_GpuNotSupported, "This function hasn't been supported yet.\n");
+            CV_Error(Error::GpuNotSupported, "This function hasn't been supported yet.\n");
             return (_Tp *)(data + step * y);
         }
 
         template<typename _Tp> inline const _Tp *oclMat::ptr(int y) const
         {
             CV_DbgAssert( (unsigned)y < (unsigned)rows );
-            CV_Error(CV_GpuNotSupported, "This function hasn't been supported yet.\n");
+            CV_Error(Error::GpuNotSupported, "This function hasn't been supported yet.\n");
             return (const _Tp *)(data + step * y);
         }
 
@@ -514,4 +514,4 @@ namespace cv
 
 } /* end of namespace cv */
 
-#endif /* __OPENCV_GPU_MATRIX_OPERATIONS_HPP__ */
+#endif /* __OPENCV_OCL_MATRIX_OPERATIONS_HPP__ */
diff --git a/modules/ocl/include/opencv2/ocl/ocl.hpp b/modules/ocl/include/opencv2/ocl/ocl.hpp
index e982774a56..3dd46545ae 100644
--- a/modules/ocl/include/opencv2/ocl/ocl.hpp
+++ b/modules/ocl/include/opencv2/ocl/ocl.hpp
@@ -7,7 +7,7 @@
 //  copy or use the software.
 //
 //
-//                          License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/ocl.hpp"
\ No newline at end of file
+#include "opencv2/ocl.hpp"
diff --git a/modules/ocl/include/opencv2/ocl/private/util.hpp b/modules/ocl/include/opencv2/ocl/private/util.hpp
new file mode 100644
index 0000000000..f8172e0c2e
--- /dev/null
+++ b/modules/ocl/include/opencv2/ocl/private/util.hpp
@@ -0,0 +1,181 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_OCL_PRIVATE_UTIL__
+#define __OPENCV_OCL_PRIVATE_UTIL__
+
+#if defined __APPLE__
+#include <OpenCL/opencl.h>
+#else
+#include <CL/opencl.h>
+#endif
+
+#include "opencv2/ocl.hpp"
+
+namespace cv
+{
+    namespace ocl
+    {
+        enum openCLMemcpyKind
+        {
+            clMemcpyHostToDevice = 0,
+            clMemcpyDeviceToHost,
+            clMemcpyDeviceToDevice
+        };
+        ///////////////////////////OpenCL call wrappers////////////////////////////
+        //void CV_EXPORTS openCLMallocPitch(Context *clCxt, void **dev_ptr, size_t *pitch,
+        //                                  size_t widthInBytes, size_t height);
+        void CV_EXPORTS openCLMallocPitch(Context *clCxt, void **dev_ptr, size_t *pitch,
+                                            size_t widthInBytes, size_t height,
+                                            DevMemRW rw_type, DevMemType mem_type, void* hptr = 0);
+        void CV_EXPORTS openCLMemcpy2D(Context *clCxt, void *dst, size_t dpitch,
+                                       const void *src, size_t spitch,
+                                       size_t width, size_t height, openCLMemcpyKind kind, int channels = -1);
+        void CV_EXPORTS openCLCopyBuffer2D(Context *clCxt, void *dst, size_t dpitch, int dst_offset,
+                                           const void *src, size_t spitch,
+                                           size_t width, size_t height, int src_offset);
+        void CV_EXPORTS openCLFree(void *devPtr);
+        cl_mem CV_EXPORTS openCLCreateBuffer(Context *clCxt, size_t flag, size_t size);
+        void CV_EXPORTS openCLReadBuffer(Context *clCxt, cl_mem dst_buffer, void *host_buffer, size_t size);
+        cl_kernel CV_EXPORTS openCLGetKernelFromSource(const Context *clCxt,
+                                                       const char **source, String kernelName);
+        cl_kernel CV_EXPORTS openCLGetKernelFromSource(const Context *clCxt,
+                                                       const char **source, String kernelName, const char *build_options);
+        void CV_EXPORTS openCLVerifyKernel(const Context *clCxt, cl_kernel kernel, size_t *localThreads);
+        void CV_EXPORTS openCLExecuteKernel(Context *clCxt , const char **source, String kernelName, std::vector< std::pair<size_t, const void *> > &args,
+                                 int globalcols , int globalrows, size_t blockSize = 16, int kernel_expand_depth = -1, int kernel_expand_channel = -1);
+        void CV_EXPORTS openCLExecuteKernel_(Context *clCxt , const char **source, String kernelName,
+                                  size_t globalThreads[3], size_t localThreads[3],
+                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options);
+        void CV_EXPORTS openCLExecuteKernel(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
+                                 size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth);
+        void CV_EXPORTS openCLExecuteKernel(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
+                                 size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
+                                 int depth, const char *build_options);
+
+        cl_mem CV_EXPORTS load_constant(cl_context context, cl_command_queue command_queue, const void *value,
+                             const size_t size);
+
+        cl_mem CV_EXPORTS openCLMalloc(cl_context clCxt, size_t size, cl_mem_flags flags, void *host_ptr);
+
+        int CV_EXPORTS savetofile(const Context *clcxt,  cl_program &program, const char *fileName);
+
+        enum FLUSH_MODE
+        {
+            CLFINISH = 0,
+            CLFLUSH,
+            DISABLE
+        };
+
+        void CV_EXPORTS openCLExecuteKernel2(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
+                                  size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, FLUSH_MODE finish_mode = DISABLE);
+        void CV_EXPORTS openCLExecuteKernel2(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
+                                  size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
+                                  int depth, const char *build_options, FLUSH_MODE finish_mode = DISABLE);
+        // bind oclMat to OpenCL image textures
+        // note:
+        //   1. there is no memory management. User need to explicitly release the resource
+        //   2. for faster clamping, there is no buffer padding for the constructed texture
+        cl_mem CV_EXPORTS bindTexture(const oclMat &mat);
+        void CV_EXPORTS releaseTexture(cl_mem& texture);
+
+        //Represents an image texture object
+        class CV_EXPORTS TextureCL
+        {
+        public:
+            TextureCL(cl_mem tex, int r, int c, int t)
+                : tex_(tex), rows(r), cols(c), type(t) {}
+            ~TextureCL()
+            {
+                openCLFree(tex_);
+            }
+            operator cl_mem()
+            {
+                return tex_;
+            }
+            cl_mem const tex_;
+            const int rows;
+            const int cols;
+            const int type;
+        private:
+            //disable assignment
+            void operator=(const TextureCL&);
+        };
+        // bind oclMat to OpenCL image textures and retunrs an TextureCL object
+        // note:
+        //   for faster clamping, there is no buffer padding for the constructed texture
+        Ptr<TextureCL> CV_EXPORTS bindTexturePtr(const oclMat &mat);
+
+        // returns whether the current context supports image2d_t format or not
+        bool CV_EXPORTS support_image2d(Context *clCxt = Context::getContext());
+
+        // the enums are used to query device information
+        enum DEVICE_INFO
+        {
+            WAVEFRONT_SIZE,             //in AMD speak
+            IS_CPU_DEVICE               //check if the device is CPU
+        };
+        template<DEVICE_INFO _it, typename _ty>
+        _ty queryDeviceInfo(cl_kernel kernel = NULL);
+
+        template<>
+        int CV_EXPORTS queryDeviceInfo<WAVEFRONT_SIZE, int>(cl_kernel kernel);
+        template<>
+        size_t CV_EXPORTS queryDeviceInfo<WAVEFRONT_SIZE, size_t>(cl_kernel kernel);
+        template<>
+        bool CV_EXPORTS queryDeviceInfo<IS_CPU_DEVICE, bool>(cl_kernel kernel);
+
+        //only these three specializations are implemented at the moment
+        template<>
+        int CV_EXPORTS queryDeviceInfo<WAVEFRONT_SIZE, int>(cl_kernel kernel);
+        template<>
+        size_t CV_EXPORTS queryDeviceInfo<WAVEFRONT_SIZE, size_t>(cl_kernel kernel);
+        template<>
+        bool CV_EXPORTS queryDeviceInfo<IS_CPU_DEVICE, bool>(cl_kernel kernel);
+    }//namespace ocl
+
+}//namespace cv
+
+#endif //__OPENCV_OCL_PRIVATE_UTIL__
diff --git a/modules/ocl/perf/interpolation.hpp b/modules/ocl/perf/interpolation.hpp
deleted file mode 100644
index fb89e701d7..0000000000
--- a/modules/ocl/perf/interpolation.hpp
+++ /dev/null
@@ -1,120 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_TEST_INTERPOLATION_HPP__
-#define __OPENCV_TEST_INTERPOLATION_HPP__
-
-template <typename T> T readVal(const cv::Mat &src, int y, int x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-{
-    if (border_type == cv::BORDER_CONSTANT)
-        return (y >= 0 && y < src.rows && x >= 0 && x < src.cols) ? src.at<T>(y, x * src.channels() + c) : cv::saturate_cast<T>(borderVal.val[c]);
-
-    return src.at<T>(cv::borderInterpolate(y, src.rows, border_type), cv::borderInterpolate(x, src.cols, border_type) * src.channels() + c);
-}
-
-template <typename T> struct NearestInterpolator
-{
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        return readVal<T>(src, cvFloor(y), cvFloor(x), c, border_type, borderVal);
-    }
-};
-
-template <typename T> struct LinearInterpolator
-{
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        x -= 0.5f;
-        y -= 0.5f;
-
-        int x1 = cvFloor(x);
-        int y1 = cvFloor(y);
-        int x2 = x1 + 1;
-        int y2 = y1 + 1;
-
-        float res = 0;
-
-        res += readVal<T>(src, y1, x1, c, border_type, borderVal) * ((x2 - x) * (y2 - y));
-        res += readVal<T>(src, y1, x2, c, border_type, borderVal) * ((x - x1) * (y2 - y));
-        res += readVal<T>(src, y2, x1, c, border_type, borderVal) * ((x2 - x) * (y - y1));
-        res += readVal<T>(src, y2, x2, c, border_type, borderVal) * ((x - x1) * (y - y1));
-
-        return cv::saturate_cast<T>(res);
-    }
-};
-
-template <typename T> struct CubicInterpolator
-{
-    static float getValue(float p[4], float x)
-    {
-        return p[1] + 0.5 * x * (p[2] - p[0] + x * (2.0 * p[0] - 5.0 * p[1] + 4.0 * p[2] - p[3] + x * (3.0 * (p[1] - p[2]) + p[3] - p[0])));
-    }
-
-    static float getValue(float p[4][4], float x, float y)
-    {
-        float arr[4];
-
-        arr[0] = getValue(p[0], x);
-        arr[1] = getValue(p[1], x);
-        arr[2] = getValue(p[2], x);
-        arr[3] = getValue(p[3], x);
-
-        return getValue(arr, y);
-    }
-
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        int ix = cvRound(x);
-        int iy = cvRound(y);
-
-        float vals[4][4] =
-        {
-            {readVal<T>(src, iy - 2, ix - 2, c, border_type, borderVal), readVal<T>(src, iy - 2, ix - 1, c, border_type, borderVal), readVal<T>(src, iy - 2, ix, c, border_type, borderVal), readVal<T>(src, iy - 2, ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy - 1, ix - 2, c, border_type, borderVal), readVal<T>(src, iy - 1, ix - 1, c, border_type, borderVal), readVal<T>(src, iy - 1, ix, c, border_type, borderVal), readVal<T>(src, iy - 1, ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy    , ix - 2, c, border_type, borderVal), readVal<T>(src, iy    , ix - 1, c, border_type, borderVal), readVal<T>(src, iy    , ix, c, border_type, borderVal), readVal<T>(src, iy    , ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy + 1, ix - 2, c, border_type, borderVal), readVal<T>(src, iy + 1, ix - 1, c, border_type, borderVal), readVal<T>(src, iy + 1, ix, c, border_type, borderVal), readVal<T>(src, iy + 1, ix + 1, c, border_type, borderVal)},
-        };
-
-        return cv::saturate_cast<T>(getValue(vals, (x - ix + 2.0) / 4.0, (y - iy + 2.0) / 4.0));
-    }
-};
-
-#endif // __OPENCV_TEST_INTERPOLATION_HPP__
diff --git a/modules/ocl/perf/main.cpp b/modules/ocl/perf/main.cpp
index f1b6884279..d1f24f165a 100644
--- a/modules/ocl/perf/main.cpp
+++ b/modules/ocl/perf/main.cpp
@@ -7,12 +7,13 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
-//
+
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -21,12 +22,12 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
+//     and/or other oclMaterials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -39,101 +40,56 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "perf_precomp.hpp"
 
-#ifdef HAVE_OPENCL
-
-using namespace std;
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-
-void print_info()
+const char * impls[] =
 {
-    printf("\n");
-#if defined _WIN32
-#   if defined _WIN64
-    puts("OS: Windows 64");
-#   else
-    puts("OS: Windows 32");
-#   endif
-#elif defined linux
-#   if defined _LP64
-    puts("OS: Linux 64");
-#   else
-    puts("OS: Linux 32");
-#   endif
-#elif defined __APPLE__
-#   if defined _LP64
-    puts("OS: Apple 64");
-#   else
-    puts("OS: Apple 32");
-#   endif
+    IMPL_OCL,
+    IMPL_PLAIN,
+#ifdef HAVE_OPENCV_GPU
+    IMPL_GPU
 #endif
+};
 
-}
-std::string workdir;
-int main(int argc, char **argv)
+int main(int argc, char ** argv)
 {
-    TS::ptr()->init("ocl");
-    InitGoogleTest(&argc, argv);
-    const char *keys =
-        "{ h | false              | print help message }"
-		"{ w | ../../../samples/c/| set working directory i.e. -w=C:\\}"
-        "{ t | gpu                | set device type:i.e. -t=cpu or gpu}"
-        "{ p | 0                  | set platform id i.e. -p=0}"
-        "{ d | 0                  | set device id i.e. -d=0}";
+    const char * keys =
+        "{ h help     | false              | print help message }"
+        "{ t type     | gpu                | set device type:cpu or gpu}"
+        "{ p platform | 0                  | set platform id }"
+        "{ d device   | 0                  | set device id }";
 
     CommandLineParser cmd(argc, argv, keys);
-    if (cmd.get<string>("h")=="true")
+    if (cmd.has("help"))
     {
-        cout << "Avaible options besides goole test option:" << endl;
+        cout << "Available options besides google test option:" << endl;
         cmd.printMessage();
         return 0;
     }
-    workdir = cmd.get<string>("w");
-    string type = cmd.get<string>("t");
-    unsigned int pid = cmd.get<unsigned int>("p");
-    int device = cmd.get<int>("d");
-    print_info();
-    // int flag = CVCL_DEVICE_TYPE_GPU;
 
-    // if(type == "cpu")
-    // {
-    //     flag = CVCL_DEVICE_TYPE_CPU;
-    // }
+    string type = cmd.get<string>("type");
+    unsigned int pid = cmd.get<unsigned int>("platform");
+    int device = cmd.get<int>("device");
+
+    int flag = type == "cpu" ? cv::ocl::CVCL_DEVICE_TYPE_CPU :
+                               cv::ocl::CVCL_DEVICE_TYPE_GPU;
+
     std::vector<cv::ocl::Info> oclinfo;
-    int devnums = getDevice(oclinfo);
-    if(devnums <= device || device < 0)
+    int devnums = cv::ocl::getDevice(oclinfo, flag);
+    if (devnums <= device || device < 0)
     {
         std::cout << "device invalid\n";
         return -1;
     }
 
-    if(pid >= oclinfo.size())
+    if (pid >= oclinfo.size())
     {
         std::cout << "platform invalid\n";
         return -1;
     }
 
-    if(pid != 0 || device != 0)
-    {
-        setDevice(oclinfo[pid], device);
-    }
+    cv::ocl::setDevice(oclinfo[pid], device);
+    cv::ocl::setBinaryDiskCache(cv::ocl::CACHE_UPDATE);
 
-    cout << "Device type:" << type << endl << "Device name:" << oclinfo[pid].DeviceName[device] << endl;
-    setBinpath(CLBINPATH);
-    return RUN_ALL_TESTS();
+    CV_PERF_TEST_MAIN_INTERNALS(ocl, impls)
 }
-
-#else // DON'T HAVE_OPENCL
-
-int main()
-{
-    printf("OpenCV was built without OpenCL support\n");
-    return 0;
-}
-
-
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/perf/perf_arithm.cpp b/modules/ocl/perf/perf_arithm.cpp
index b7f82b685d..814b272f02 100644
--- a/modules/ocl/perf/perf_arithm.cpp
+++ b/modules/ocl/perf/perf_arithm.cpp
@@ -1,4 +1,4 @@
-///////////////////////////////////////////////////////////////////////////////////////
+/*M///////////////////////////////////////////////////////////////////////////////////////
 //
 //  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
 //
@@ -10,17 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Niko Li, newlife20080214@gmail.com
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//    Shengen Yan, yanshengen@gmail.com
-//    Jiang Liyuan,jlyuan001.good@163.com
-//    Rock Li, Rock.Li@amd.com
-//    Zailong Wu, bullet@yeah.net
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -35,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -48,4371 +44,884 @@
 //
 //M*/
 
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-#include <iomanip>
+using namespace perf;
+using std::tr1::get;
+using std::tr1::tuple;
 
-#ifdef HAVE_OPENCL
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-PARAM_TEST_CASE(ArithmTestBase, MatType, bool)
+///////////// Lut ////////////////////////
+
+typedef Size_MatType LUTFixture;
+
+PERF_TEST_P(LUTFixture, LUT,
+          ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                             OCL_PERF_ENUM(CV_8UC1, CV_8UC3)))
 {
-    int type;
-    cv::Scalar val;
+    // getting params
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mask;
-    cv::Mat dst;
-    cv::Mat dst1; //bak, for two outputs
+    // creating src data
+    Mat src(srcSize, type), lut(1, 256, CV_8UC1);
+    int dstType = CV_MAKETYPE(lut.depth(), src.channels());
+    Mat dst(srcSize, dstType);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int dstx;
-    int dsty;
-    int maskx;
-    int masky;
+    randu(lut, 0, 2);
+    declare.in(src, WARMUP_RNG).in(lut).out(dst);
 
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-    cv::ocl::oclMat gdst1_whole; //bak
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gmat2;
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gdst1;   //bak
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
+        ocl::oclMat oclSrc(src), oclLut(lut), oclDst(srcSize, dstType);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
+        OCL_TEST_CYCLE() cv::ocl::LUT(oclSrc, oclLut, oclDst);
+        oclDst.download(dst);
 
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        //mat2 = randomMat(rng, cv::Size(512,3), type, 5, 16, false);
-        mat2 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        dst1  = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
-
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums>0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src2x   = 1;
-            src1y   = 1;
-            src2y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-            maskx	 = 1;
-            masky	= 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src2x = 0;
-            src1y = 0;
-            src2y = 0;
-            dstx = 0;
-            dsty = 0;
-            maskx	 = 0;
-            masky	= 0;
-        };
+        TEST_CYCLE() cv::LUT(src, lut, dst);
 
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        //mat2_roi = mat2(Rect(src2x,src2y,256,1));
-        mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-        dst1_roi = dst1(Rect(dstx, dsty, roicols, roirows));
-
-        //gdst_whole = dst;
-        //gdst = gdst_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gdst1_whole = dst1;
-        //gdst1 = gdst1_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gmat1 = mat1_roi;
-        //gmat2 = mat2_roi;
-        //gmask = mask_roi;
+        SANITY_CHECK(dst);
     }
-
-};
-////////////////////////////////lut/////////////////////////////////////////////////
-
-struct Lut : ArithmTestBase {};
-
-TEST_P(Lut, Mat)
-{
-
-    cv::Mat mat2(3, 512, CV_8UC1);
-    cv::RNG &rng = TS::ptr()->get_rng();
-    rng.fill(mat2, cv::RNG::UNIFORM, cv::Scalar::all(0), cv::Scalar::all(256));
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            mat2 = randomMat(rng, cv::Size(512, 3), type, 5, 16, false);
-            mat2_roi = mat2(Rect(src2x, src2y, 256, 1));
-
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::LUT(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::LUT(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        // s=GetParam();
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        //  src2x = rng.uniform( 0,mat2.cols - 256);
-        // src2y = rng.uniform (0,mat2.rows - 1);
-
-        // cv::Mat mat2_roi = mat2(Rect(src2x,src2y,256,1));
-        mat2 = randomMat(rng, cv::Size(512, 3), type, 5, 16, false);
-        mat2_roi = mat2(Rect(src2x, src2y, 256, 1));
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        //   gdst1_whole = dst1;
-        //     gdst1 = gdst1_whole(Rect(dstx,dsty,roicols,roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        //     gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::LUT(gmat1, gmat2, gdst);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Exp ////////////////////////
 
+typedef TestBaseWithParam<Size> ExpFixture;
 
-////////////////////////////////exp/////////////////////////////////////////////////
-
-struct Exp : ArithmTestBase {};
-
-TEST_P(Exp, Mat)
+PERF_TEST_P(ExpFixture, Exp, OCL_TYPICAL_MAT_SIZES)
 {
+    // getting params
+    const Size srcSize = GetParam();
+    const double eps = 3e-1;
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    // creating src data
+    Mat src(srcSize, CV_32FC1), dst(srcSize, CV_32FC1);
+    declare.in(src).out(dst);
+    randu(src, 5, 16);
+
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, src.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::exp(mat1_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::exp(oclSrc, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
+        oclDst.download(dst);
 
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::exp(gmat1, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download(cpu_dst);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-            //EXPECT_MAT_NEAR(dst, cpu_dst, 0,"");
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, eps);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::exp(gmat1, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::exp(src, dst);
 
+        SANITY_CHECK(dst, eps);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// LOG ////////////////////////
 
-////////////////////////////////log/////////////////////////////////////////////////
+typedef TestBaseWithParam<Size> LogFixture;
 
-struct Log : ArithmTestBase {};
-
-TEST_P(Log, Mat)
+PERF_TEST_P(LogFixture, Log, OCL_TYPICAL_MAT_SIZES)
 {
+    // getting params
+    const Size srcSize = GetParam();
+    const double eps = 1e-5;
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    // creating src data
+    Mat src(srcSize, CV_32F), dst(srcSize, src.type());
+    randu(src, 1, 10);
+    declare.in(src).out(dst);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(3.6);
+
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, src.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::log(mat1_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::log(oclSrc, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::log(gmat1, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, eps);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::log(gmat1, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::log(src, dst);
 
+        SANITY_CHECK(dst, eps);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Add ////////////////////////
 
+typedef Size_MatType AddFixture;
 
-
-////////////////////////////////add/////////////////////////////////////////////////
-
-struct Add : ArithmTestBase {};
-
-TEST_P(Add, Mat)
+PERF_TEST_P(AddFixture, Add,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
+    // getting params
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    // creating src data
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    randu(src1, 0, 1);
+    randu(src2, 0, 1);
+    declare.in(src1, src2).out(dst);
+
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::add(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::add(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::add(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::add(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::add(src1, src2, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Add, Mat_Mask)
+///////////// Mul ////////////////////////
+
+typedef Size_MatType MulFixture;
+
+PERF_TEST_P(MulFixture, Mul, ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                                                OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    // getting params
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    // creating src data
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+    declare.in(src1, src2).out(dst);
+
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::add(mat1_roi, mat2_roi, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::multiply(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::add(gmat1, gmat2, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::add(gmat1, gmat2, gdst, gmask);
-    };
-#endif
-}
-TEST_P(Add, Scalar)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        TEST_CYCLE() cv::multiply(src1, src2, dst);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::add(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::add(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::add(gmat1, val, gdst);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Add, Scalar_Mask)
+///////////// Div ////////////////////////
+
+typedef Size_MatType DivFixture;
+
+PERF_TEST_P(DivFixture, Div,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    // getting params
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    // creating src data
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    declare.in(src1, src2).out(dst);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+    if ((srcSize == OCL_SIZE_4000 && type == CV_8UC1) ||
+            (srcSize == OCL_SIZE_2000 && type == CV_8UC4))
+        declare.time(4.2);
+    else if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(16.6);
+
+    // select implementation
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::add(mat1_roi, val, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::divide(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::add(gmat1, val, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
+        oclDst.download(dst);
 
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::add(gmat1, val, gdst, gmask);
-    };
-#endif
+        TEST_CYCLE() cv::divide(src1, src2, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Absdiff ////////////////////////
 
-////////////////////////////////sub/////////////////////////////////////////////////
-struct Sub : ArithmTestBase {};
+typedef Size_MatType AbsDiffFixture;
 
-TEST_P(Sub, Mat)
+PERF_TEST_P(AbsDiffFixture, Absdiff,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    declare.in(src1, src2).in(dst);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::subtract(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::absdiff(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::subtract(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::subtract(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::absdiff(src1, src2, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Sub, Mat_Mask)
+///////////// CartToPolar ////////////////////////
+
+typedef TestBaseWithParam<Size> CartToPolarFixture;
+
+PERF_TEST_P(CartToPolarFixture, CartToPolar, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+    const double eps = 8e-3;
+
+    Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            dst1(srcSize, CV_32FC1), dst2(srcSize, CV_32FC1);
+    declare.in(src1, src2).out(dst1, dst2);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(3.6);
+
+   if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2),
+                oclDst1(srcSize, src1.type()), oclDst2(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::subtract(mat1_roi, mat2_roi, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::cartToPolar(oclSrc1, oclSrc2, oclDst1, oclDst2);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst1.download(dst1);
+        oclDst2.download(dst2);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::subtract(gmat1, gmat2, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst1, eps);
+        SANITY_CHECK(dst2, eps);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::subtract(gmat1, gmat2, gdst, gmask);
-    };
-#endif
-}
-TEST_P(Sub, Scalar)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        TEST_CYCLE() cv::cartToPolar(src1, src2, dst1, dst2);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::subtract(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::subtract(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst1, eps);
+        SANITY_CHECK(dst2, eps);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::subtract(gmat1, val, gdst);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Sub, Scalar_Mask)
+///////////// PolarToCart ////////////////////////
+
+typedef TestBaseWithParam<Size> PolarToCartFixture;
+
+PERF_TEST_P(PolarToCartFixture, PolarToCart, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+   Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            dst1(srcSize, CV_32FC1), dst2(srcSize, CV_32FC1);
+    declare.in(src1, src2).out(dst1, dst2);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(5.4);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2),
+                oclDst1(srcSize, src1.type()), oclDst2(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::subtract(mat1_roi, val, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::polarToCart(oclSrc1, oclSrc2, oclDst1, oclDst2);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst1.download(dst1);
+        oclDst2.download(dst2);
 
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::subtract(gmat1, val, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst1, 5e-5);
+        SANITY_CHECK(dst2, 5e-5);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::subtract(gmat1, val, gdst, gmask);
-    };
-#endif
+        TEST_CYCLE() cv::polarToCart(src1, src2, dst1, dst2);
+
+        SANITY_CHECK(dst1, 5e-5);
+        SANITY_CHECK(dst2, 5e-5);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Magnitude ////////////////////////
 
-////////////////////////////////Mul/////////////////////////////////////////////////
-struct Mul : ArithmTestBase {};
+typedef TestBaseWithParam<Size> MagnitudeFixture;
 
-TEST_P(Mul, Mat)
+PERF_TEST_P(MagnitudeFixture, Magnitude, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+    Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            dst(srcSize, CV_32FC1);
+    randu(src1, 0, 1);
+    randu(src2, 0, 1);
+    declare.in(src1, src2).out(dst);
+
+   if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2),
+                oclDst(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::multiply(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::magnitude(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::multiply(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, 1e-6);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::multiply(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::magnitude(src1, src2, dst);
+
+        SANITY_CHECK(dst, 1e-6);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Mul, Mat_Scalar)
+///////////// Transpose ////////////////////////
+
+typedef Size_MatType TransposeFixture;
+
+PERF_TEST_P(TransposeFixture, Transpose,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+   if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            cv::RNG &rng = TS::ptr()->get_rng();
-            double s = rng.uniform(-10.0, 10.0);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::multiply(mat1_roi, mat2_roi, dst_roi, s);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        OCL_TEST_CYCLE() cv::ocl::transpose(oclSrc, oclDst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::multiply(gmat1, gmat2, gdst, s);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
+        oclDst.download(dst);
 
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        cv::RNG &rng = TS::ptr()->get_rng();
-        double s = rng.uniform(-10.0, 10.0);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::multiply(gmat1, gmat2, gdst, s);
-    };
-#endif
+        TEST_CYCLE() cv::transpose(src, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Flip ////////////////////////
 
-struct Div : ArithmTestBase {};
+typedef Size_MatType FlipFixture;
 
-TEST_P(Div, Mat)
+PERF_TEST_P(FlipFixture, Flip,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::divide(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::flip(oclSrc, oclDst, 0);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::divide(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::divide(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::flip(src, dst, 0);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Div, Mat_Scalar)
+///////////// minMax ////////////////////////
+
+typedef Size_MatType minMaxFixture;
+
+PERF_TEST_P(minMaxFixture, minMax,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type);
+    declare.in(src, WARMUP_RNG);
+
+    double min_val = std::numeric_limits<double>::max(),
+            max_val = std::numeric_limits<double>::min();
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            cv::RNG &rng = TS::ptr()->get_rng();
-            double s = rng.uniform(-10.0, 10.0);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::divide(mat1_roi, mat2_roi, dst_roi, s);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        ocl::oclMat oclSrc(src);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        OCL_TEST_CYCLE() cv::ocl::minMax(oclSrc, &min_val, &max_val);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::divide(gmat1, gmat2, gdst, s);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        ASSERT_GE(max_val, min_val);
+        SANITY_CHECK(min_val);
+        SANITY_CHECK(max_val);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        cv::RNG &rng = TS::ptr()->get_rng();
-        double s = rng.uniform(-10.0, 10.0);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::divide(gmat1, gmat2, gdst, s);
-    };
-#endif
+        Point min_loc, max_loc;
+
+        TEST_CYCLE() cv::minMaxLoc(src, &min_val, &max_val, &min_loc, &max_loc);
+
+        ASSERT_GE(max_val, min_val);
+        SANITY_CHECK(min_val);
+        SANITY_CHECK(max_val);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// minMaxLoc ////////////////////////
 
-struct Absdiff : ArithmTestBase {};
+typedef Size_MatType minMaxLocFixture;
 
-TEST_P(Absdiff, Mat)
+PERF_TEST_P(minMaxLocFixture, minMaxLoc,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
+   const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    Mat src(srcSize, type);
+    randu(src, 0, 1);
+    declare.in(src);
+
+    double min_val = 0.0, max_val = 0.0;
+    Point min_loc, max_loc;
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::absdiff(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::minMaxLoc(oclSrc, &min_val, &max_val, &min_loc, &max_loc);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::absdiff(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        ASSERT_GE(max_val, min_val);
+        SANITY_CHECK(min_val);
+        SANITY_CHECK(max_val);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::absdiff(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::minMaxLoc(src, &min_val, &max_val, &min_loc, &max_loc);
+
+        ASSERT_GE(max_val, min_val);
+        SANITY_CHECK(min_val);
+        SANITY_CHECK(max_val);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(Absdiff, Mat_Scalar)
+///////////// Sum ////////////////////////
+
+typedef Size_MatType SumFixture;
+
+PERF_TEST_P(SumFixture, Sum,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32SC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+   Mat src(srcSize, type);
+    Scalar result;
+    randu(src, 0, 60);
+    declare.in(src);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::absdiff(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() result = cv::ocl::sum(oclSrc);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::absdiff(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(result);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::absdiff(gmat1, val, gdst);
-    };
-#endif
+        TEST_CYCLE() result = cv::sum(src);
+
+        SANITY_CHECK(result);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// countNonZero ////////////////////////
 
+typedef Size_MatType countNonZeroFixture;
 
-struct CartToPolar : ArithmTestBase {};
-
-TEST_P(CartToPolar, angleInDegree)
+PERF_TEST_P(countNonZeroFixture, countNonZero,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type);
+    int result = 0;
+    randu(src, 0, 256);
+    declare.in(src);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::cartToPolar(mat1_roi, mat2_roi, dst_roi, dst1_roi, 1);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() result = cv::ocl::countNonZero(oclSrc);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gdst1_whole = dst1;
-            gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            cv::Mat cpu_dst1;
-            gdst1_whole.download(cpu_dst1);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(result);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 1);
-    };
-#endif
+        TEST_CYCLE() result = cv::countNonZero(src);
+
+        SANITY_CHECK(result);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(CartToPolar, angleInRadians)
+///////////// Phase ////////////////////////
+
+typedef TestBaseWithParam<Size> PhaseFixture;
+
+PERF_TEST_P(PhaseFixture, Phase, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+    Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            dst(srcSize, CV_32FC1);
+    declare.in(src1, src2).out(dst);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2),
+                oclDst(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::cartToPolar(mat1_roi, mat2_roi, dst_roi, dst1_roi, 0);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::phase(oclSrc1, oclSrc2, oclDst, 1);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gdst1_whole = dst1;
-            gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 0);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            cv::Mat cpu_dst1;
-            gdst1_whole.download(cpu_dst1);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
+        oclDst.download(dst);
 
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, 1e-2);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 0);
-    };
-#endif
+        TEST_CYCLE() cv::phase(src1, src2, dst, 1);
+
+        SANITY_CHECK(dst, 1e-2);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// bitwise_and////////////////////////
 
-struct PolarToCart : ArithmTestBase {};
+typedef Size_MatType BitwiseAndFixture;
 
-TEST_P(PolarToCart, angleInDegree)
+PERF_TEST_P(BitwiseAndFixture, bitwise_and,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32SC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+   const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+   Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    declare.in(src1, src2).out(dst);
+    randu(src1, 0, 256);
+    randu(src2, 0, 256);
+
+   if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::polarToCart(mat1_roi, mat2_roi, dst_roi, dst1_roi, 1);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::bitwise_and(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gdst1_whole = dst1;
-            gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            cv::Mat cpu_dst1;
-            gdst1_whole.download(cpu_dst1);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 1);
-    };
-#endif
+        TEST_CYCLE() cv::bitwise_and(src1, src2, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(PolarToCart, angleInRadians)
+///////////// bitwise_not////////////////////////
+
+typedef Size_MatType BitwiseNotFixture;
+
+PERF_TEST_P(BitwiseAndFixture, bitwise_not,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32SC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::polarToCart(mat1_roi, mat2_roi, dst_roi, dst1_roi, 0);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::bitwise_not(oclSrc, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gdst1_whole = dst1;
-            gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 0);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            cv::Mat cpu_dst1;
-            gdst1_whole.download(cpu_dst1);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 0);
-    };
-#endif
+        TEST_CYCLE() cv::bitwise_not(src, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// compare////////////////////////
 
+typedef Size_MatType CompareFixture;
 
-struct Magnitude : ArithmTestBase {};
-
-TEST_P(Magnitude, Mat)
+PERF_TEST_P(CompareFixture, compare,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, CV_8UC1);
+    declare.in(src1, src2, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, CV_8UC1);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::magnitude(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::compare(oclSrc1, oclSrc2, oclDst, CMP_EQ);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::magnitude(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::magnitude(gmat1, gmat2, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::compare(src1, src2, dst, CMP_EQ);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-struct Transpose : ArithmTestBase {};
+///////////// pow ////////////////////////
 
-TEST_P(Transpose, Mat)
+typedef TestBaseWithParam<Size> PowFixture;
+
+PERF_TEST_P(PowFixture, pow, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+   Mat src(srcSize, CV_32F), dst(srcSize, CV_32F);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, src.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::transpose(mat1_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::pow(oclSrc, -2.0, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::transpose(gmat1, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, 5e-2);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::transpose(gmat1, gdst);
-    };
-#endif
+        TEST_CYCLE() cv::pow(src, -2.0, dst);
+
+        SANITY_CHECK(dst, 5e-2);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// MagnitudeSqr////////////////////////
 
-struct Flip : ArithmTestBase {};
+typedef TestBaseWithParam<Size> MagnitudeSqrFixture;
 
-TEST_P(Flip, X)
+PERF_TEST_P(MagnitudeSqrFixture, MagnitudeSqr, OCL_TYPICAL_MAT_SIZES)
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+    Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            dst(srcSize, CV_32FC1);
+    declare.in(src1, src2, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, src1.type());
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::flip(mat1_roi, dst_roi, 0);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::magnitudeSqr(oclSrc1, oclSrc2, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::flip(gmat1, gdst, 0);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
+        ASSERT_EQ(1, src1.channels());
+
+        TEST_CYCLE()
         {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::flip(gmat1, gdst, 0);
-    };
-#endif
-}
-
-TEST_P(Flip, Y)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::flip(mat1_roi, dst_roi, 1);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::flip(gmat1, gdst, 1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::flip(gmat1, gdst, 1);
-    };
-#endif
-}
-
-TEST_P(Flip, BOTH)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::flip(mat1_roi, dst_roi, -1);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::flip(gmat1, gdst, -1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::flip(gmat1, gdst, -1);
-    };
-#endif
-}
-
-
-
-struct MinMax : ArithmTestBase {};
-
-TEST_P(MinMax, MAT)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            t0 = (double)cvGetTickCount();//cpu start
-            if (mat1.depth() != CV_8S)
+            for (int y = 0; y < srcSize.height; ++y)
             {
-                cv::minMaxLoc(mat1_roi, &minVal, &maxVal, &minLoc, &maxLoc);
-            }
-            else
-            {
-                minVal = std::numeric_limits<double>::max();
-                maxVal = -std::numeric_limits<double>::max();
-                for (int i = 0; i < mat1_roi.rows; ++i)
-                    for (int j = 0; j < mat1_roi.cols; ++j)
-                    {
-                        signed char val = mat1_roi.at<signed char>(i, j);
-                        if (val < minVal) minVal = val;
-                        if (val > maxVal) maxVal = val;
-                    }
-            }
-
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            double minVal_, maxVal_;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::minMax(gmat1, &minVal_, &maxVal_);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        double minVal_, maxVal_;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::minMax(gmat1, &minVal_, &maxVal_);
-    };
-#endif
-}
-
-TEST_P(MinMax, MASK)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            t0 = (double)cvGetTickCount();//cpu start
-            if (mat1.depth() != CV_8S)
-            {
-                cv::minMaxLoc(mat1_roi, &minVal, &maxVal, &minLoc, &maxLoc, mask_roi);
-            }
-            else
-            {
-                minVal = std::numeric_limits<double>::max();
-                maxVal = -std::numeric_limits<double>::max();
-                for (int i = 0; i < mat1_roi.rows; ++i)
-                    for (int j = 0; j < mat1_roi.cols; ++j)
-                    {
-                        signed char val = mat1_roi.at<signed char>(i, j);
-                        unsigned char m = mask_roi.at<unsigned char>(i, j);
-                        if (val < minVal && m) minVal = val;
-                        if (val > maxVal && m) maxVal = val;
-                    }
-            }
-
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            double minVal_, maxVal_;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::minMax(gmat1, &minVal_, &maxVal_, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-        double minVal_, maxVal_;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::minMax(gmat1, &minVal_, &maxVal_, gmask);
-    };
-#endif
-}
-
-
-struct MinMaxLoc : ArithmTestBase {};
-
-TEST_P(MinMaxLoc, MAT)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            int depth = mat1.depth();
-            t0 = (double)cvGetTickCount();//cpu start
-            if (depth != CV_8S)
-            {
-                cv::minMaxLoc(mat1_roi, &minVal, &maxVal, &minLoc, &maxLoc);
-            }
-            else
-            {
-                minVal = std::numeric_limits<double>::max();
-                maxVal = -std::numeric_limits<double>::max();
-                for (int i = 0; i < mat1_roi.rows; ++i)
-                    for (int j = 0; j < mat1_roi.cols; ++j)
-                    {
-                        signed char val = mat1_roi.at<signed char>(i, j);
-                        if (val < minVal)
-                        {
-                            minVal = val;
-                            minLoc.x = j;
-                            minLoc.y = i;
-                        }
-                        if (val > maxVal)
-                        {
-                            maxVal = val;
-                            maxLoc.x = j;
-                            maxLoc.y = i;
-                        }
-                    }
-            }
-
-
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            double minVal_, maxVal_;
-            cv::Point minLoc_, maxLoc_;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::minMaxLoc(gmat1, &minVal_, &maxVal_, &minLoc_, &maxLoc_, cv::ocl::oclMat());
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        double minVal_, maxVal_;
-        cv::Point minLoc_, maxLoc_;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::minMaxLoc(gmat1, &minVal_, &maxVal_, &minLoc_, &maxLoc_, cv::ocl::oclMat());
-    };
-#endif
-
-}
-
-
-TEST_P(MinMaxLoc, MASK)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double minVal, maxVal;
-            cv::Point minLoc, maxLoc;
-            int depth = mat1.depth();
-            t0 = (double)cvGetTickCount();//cpu start
-            if (depth != CV_8S)
-            {
-                cv::minMaxLoc(mat1_roi, &minVal, &maxVal, &minLoc, &maxLoc, mask_roi);
-            }
-            else
-            {
-                minVal = std::numeric_limits<double>::max();
-                maxVal = -std::numeric_limits<double>::max();
-                for (int i = 0; i < mat1_roi.rows; ++i)
-                    for (int j = 0; j < mat1_roi.cols; ++j)
-                    {
-                        signed char val = mat1_roi.at<signed char>(i, j);
-                        unsigned char m = mask_roi.at<unsigned char>(i , j);
-                        if (val < minVal && m)
-                        {
-                            minVal = val;
-                            minLoc.x = j;
-                            minLoc.y = i;
-                        }
-                        if (val > maxVal && m)
-                        {
-                            maxVal = val;
-                            maxLoc.x = j;
-                            maxLoc.y = i;
-                        }
-                    }
-            }
-
-
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            double minVal_, maxVal_;
-            cv::Point minLoc_, maxLoc_;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::minMaxLoc(gmat1, &minVal_, &maxVal_, &minLoc_, &maxLoc_, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-        double minVal_, maxVal_;
-        cv::Point minLoc_, maxLoc_;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::minMaxLoc(gmat1, &minVal_, &maxVal_, &minLoc_, &maxLoc_, gmask);
-    };
-#endif
-}
-
-
-struct Sum : ArithmTestBase {};
-
-TEST_P(Sum, MAT)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::sum(mat1_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::sum(gmat1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        Scalar gpures = cv::ocl::sum(gmat1);
-    };
-#endif
-}
-
-//TEST_P(Sum, MASK)
-//{
-//    for(int j=0; j<LOOP_TIMES; j++)
-//    {
-//
-//    }
-//}
-
-struct CountNonZero : ArithmTestBase {};
-
-TEST_P(CountNonZero, MAT)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::countNonZero(mat1_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::countNonZero(gmat1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::countNonZero(gmat1);
-    };
-#endif
-
-}
-
-
-
-////////////////////////////////phase/////////////////////////////////////////////////
-struct Phase : ArithmTestBase {};
-
-TEST_P(Phase, Mat)
-{
-    if(mat1.depth() != CV_32F && mat1.depth() != CV_64F)
-    {
-        cout << "\tUnsupported type\t\n";
-    }
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::phase(mat1_roi, mat2_roi, dst_roi, 0);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::phase(gmat1, gmat2, gdst, 0);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::phase(gmat1, gmat2, gdst, 0);
-    };
-#endif
-
-}
-
-
-////////////////////////////////bitwise_and/////////////////////////////////////////////////
-struct Bitwise_and : ArithmTestBase {};
-
-TEST_P(Bitwise_and, Mat)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_and(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_and(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_and(gmat1, gmat2, gdst);
-    };
-#endif
-
-}
-
-TEST_P(Bitwise_and, Mat_Mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_and(mat1_roi, mat2_roi, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_and(gmat1, gmat2, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_and(gmat1, gmat2, gdst, gmask);
-    };
-#endif
-}
-
-TEST_P(Bitwise_and, Scalar)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_and(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_and(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_and(gmat1, val, gdst);
-    };
-#endif
-}
-
-TEST_P(Bitwise_and, Scalar_Mask)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_and(mat1_roi, val, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_and(gmat1, val, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_and(gmat1, val, gdst, gmask);
-    };
-#endif
-}
-
-
-
-////////////////////////////////bitwise_or/////////////////////////////////////////////////
-
-struct Bitwise_or : ArithmTestBase {};
-
-TEST_P(Bitwise_or, Mat)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_or(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_or(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_or(gmat1, gmat2, gdst);
-    };
-#endif
-}
-
-TEST_P(Bitwise_or, Mat_Mask)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_or(mat1_roi, mat2_roi, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_or(gmat1, gmat2, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_or(gmat1, gmat2, gdst, gmask);
-    };
-#endif
-}
-TEST_P(Bitwise_or, Scalar)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_or(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_or(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_or(gmat1, val, gdst);
-    };
-#endif
-}
-
-TEST_P(Bitwise_or, Scalar_Mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_or(mat1_roi, val, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_or(gmat1, val, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_or(gmat1, val, gdst, gmask);
-    };
-#endif
-}
-
-
-////////////////////////////////bitwise_xor/////////////////////////////////////////////////
-
-struct Bitwise_xor : ArithmTestBase {};
-
-TEST_P(Bitwise_xor, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_xor(mat1_roi, mat2_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_xor(gmat1, gmat2, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_xor(gmat1, gmat2, gdst);
-    };
-#endif
-}
-
-TEST_P(Bitwise_xor, Mat_Mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_xor(mat1_roi, mat2_roi, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_xor(gmat1, gmat2, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_xor(gmat1, gmat2, gdst, gmask);
-    };
-#endif
-}
-
-TEST_P(Bitwise_xor, Scalar)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_xor(mat1_roi, val, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_xor(gmat1, val, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_xor(gmat1, val, gdst);
-    };
-#endif
-}
-
-TEST_P(Bitwise_xor, Scalar_Mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_xor(mat1_roi, val, dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_xor(gmat1, val, gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_xor(gmat1, val, gdst, gmask);
-    };
-#endif
-}
-
-
-////////////////////////////////bitwise_not/////////////////////////////////////////////////
-
-struct Bitwise_not : ArithmTestBase {};
-
-TEST_P(Bitwise_not, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::bitwise_not(mat1_roi, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::bitwise_not(gmat1, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::bitwise_not(gmat1, gdst);
-    };
-#endif
-}
-
-////////////////////////////////compare/////////////////////////////////////////////////
-PARAM_TEST_CASE ( CompareTestBase, MatType, bool)
-{
-    int type;
-    cv::Scalar val;
-
-    //src mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mask;
-    cv::Mat dst;
-    cv::Mat dst1; //bak, for two outputs
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int dstx;
-    int dsty;
-    int maskx;
-    int masky;
-
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-    cv::ocl::oclMat gdst1_whole; //bak
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gmat2;
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gdst1;   //bak
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
-    {
-        //type = GET_PARAM(0);
-        type = CV_8UC1;
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        //mat2 = randomMat(rng, cv::Size(512,3), type, 5, 16, false);
-        mat2 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        dst1  = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
-
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums>0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
-    }
-
-    void Has_roi(int b)
-    {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src2x   = 1;
-            src1y   = 1;
-            src2y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-            maskx	 = 1;
-            masky	= 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src2x = 0;
-            src1y = 0;
-            src2y = 0;
-            dstx = 0;
-            dsty = 0;
-            maskx	 = 0;
-            masky	= 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        //mat2_roi = mat2(Rect(src2x,src2y,256,1));
-        mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-        dst1_roi = dst1(Rect(dstx, dsty, roicols, roirows));
-
-        //gdst_whole = dst;
-        //gdst = gdst_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gdst1_whole = dst1;
-        //gdst1 = gdst1_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gmat1 = mat1_roi;
-        //gmat2 = mat2_roi;
-        //gmask = mask_roi;
-    }
-
-};
-struct Compare : CompareTestBase {};
-
-TEST_P(Compare, Mat)
-{
-    if(mat1.type() == CV_8SC1)
-    {
-        cout << "\tUnsupported type\t\n";
-    }
-
-    int cmp_codes[] = {CMP_EQ, CMP_GT, CMP_GE, CMP_LT, CMP_LE, CMP_NE};
-    const char *cmp_str[] = {"CMP_EQ", "CMP_GT", "CMP_GE", "CMP_LT", "CMP_LE", "CMP_NE"};
-    int cmp_num = sizeof(cmp_codes) / sizeof(int);
-    for (int i = 0; i < cmp_num; ++i)
-    {
-
-#ifndef PRINT_KERNEL_RUN_TIME
-        double totalcputick = 0;
-        double totalgputick = 0;
-        double totalgputick_kernel = 0;
-        double t0 = 0;
-        double t1 = 0;
-        double t2 = 0;
-        for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-        {
-            totalcputick = 0;
-            totalgputick = 0;
-            totalgputick_kernel = 0;
-            for(int j = 0; j < LOOP_TIMES + 1; j ++)
-            {
-                Has_roi(k);
-
-                t0 = (double)cvGetTickCount();//cpu start
-                cv::compare(mat1_roi, mat2_roi, dst_roi, cmp_codes[i]);
-                t0 = (double)cvGetTickCount() - t0;//cpu end
-
-                t1 = (double)cvGetTickCount();//gpu start1
-                gdst_whole = dst;
-                gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-                gmat1 = mat1_roi;
-                gmat2 = mat2_roi;
-                t2 = (double)cvGetTickCount(); //kernel
-                cv::ocl::compare(gmat1, gmat2, gdst, cmp_codes[i]);
-                t2 = (double)cvGetTickCount() - t2;//kernel
-                cv::Mat cpu_dst;
-                gdst_whole.download (cpu_dst);//download
-                t1 = (double)cvGetTickCount() - t1;//gpu end1
-                if(j == 0)
-                    continue;
-                totalgputick = t1 + totalgputick;
-                totalcputick = t0 + totalcputick;
-                totalgputick_kernel = t2 + totalgputick_kernel;
-
-            }
-            cout << cmp_str[i] << endl;
-            if(k == 0)
-            {
-                cout << "no roi\n";
-            }
-            else
-            {
-                cout << "with roi\n";
-            };
-            cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        }
-#else
-        for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-        {
-            Has_roi(j);
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            if(j == 0)
-            {
-                cout << "no roi:";
-            }
-            else
-            {
-                cout << "\nwith roi:";
-            };
-            cv::ocl::compare(gmat1, gmat2, gdst, cmp_codes[i]);
-        };
-#endif
-    }
-
-}
-
-struct Pow : ArithmTestBase {};
-
-TEST_P(Pow, Mat)
-{
-    if(mat1.depth() != CV_32F && mat1.depth() != CV_64F)
-    {
-        cout << "\tUnsupported type\t\n";
-    }
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double p = 4.5;
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::pow(mat1_roi, p, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::pow(gmat1, p, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        double p = 4.5;
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::pow(gmat1, p, gdst);
-    };
-#endif
-}
-
-
-struct MagnitudeSqr : ArithmTestBase {};
-
-TEST_P(MagnitudeSqr, Mat)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            for(int i = 0; i < mat1.rows; ++i)
-                for(int j = 0; j < mat1.cols; ++j)
+                const float * const src1Data = reinterpret_cast<float *>(src1.data + src1.step * y);
+                const float * const src2Data = reinterpret_cast<float *>(src2.data + src2.step * y);
+                float * const dstData = reinterpret_cast<float *>(dst.data + dst.step * y);
+                for (int x = 0; x < srcSize.width; ++x)
                 {
-                    float val1 = mat1.at<float>(i, j);
-                    float val2 = mat2.at<float>(i, j);
-
-                    ((float *)(dst.data))[i * dst.step / 4 + j] = val1 * val1 + val2 * val2;
-
+                    float t0 = src1Data[x] * src1Data[x];
+                    float t1 = src2Data[x] * src2Data[x];
+                    dstData[x] = t0 + t1;
                 }
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            cv::ocl::oclMat clmat1(mat1), clmat2(mat2), cldst;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::magnitudeSqr(clmat1, clmat2, cldst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            cldst.download(cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
+            }
         }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+
+        SANITY_CHECK(dst, 1e-6, ERROR_RELATIVE);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        cv::ocl::oclMat clmat1(mat1), clmat2(mat2), cldst;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::magnitudeSqr(clmat1, clmat2, cldst);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// AddWeighted////////////////////////
 
-struct AddWeighted : ArithmTestBase {};
+typedef Size_MatType AddWeightedFixture;
 
-TEST_P(AddWeighted, Mat)
+PERF_TEST_P(AddWeightedFixture, AddWeighted,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src1(srcSize, type), src2(srcSize, type), dst(srcSize, type);
+    declare.in(src1, src2, WARMUP_RNG).out(dst);
+    double alpha = 2.0, beta = 1.0, gama = 3.0;
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            double alpha = 2.0, beta = 1.0, gama = 3.0;
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::addWeighted(mat1_roi, alpha, mat2_roi, beta, gama, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::addWeighted(oclSrc1, alpha, oclSrc2, beta, gama, oclDst);
 
-            t1 = (double)cvGetTickCount();//gpu start1
+        oclDst.download(dst);
 
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::addWeighted(gmat1, alpha, gmat2, beta, gama, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download(cpu_dst);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        double alpha = 2.0, beta = 1.0, gama = 3.0;
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::addWeighted(gmat1, alpha, gmat2, beta, gama, gdst);
-        // double alpha=2.0,beta=1.0,gama=3.0;
-        // cv::ocl::oclMat clmat1(mat1),clmat2(mat2),cldst;
-        // if(j==0){cout<<"no roi:";}else{cout<<"\nwith roi:";};
-        // cv::ocl::addWeighted(clmat1,alpha,clmat2,beta,gama, cldst);
-    };
-#endif
-
-}
-/*
-struct AddWeighted : ArithmTestBase {};
-
-TEST_P(AddWeighted, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick=0;
-    double totalgputick=0;
-    double totalgputick_kernel=0;
-    double t0=0;
-    double t1=0;
-    double t2=0;
-    for(int j = 0; j < LOOP_TIMES+1; j ++)
-    {
-        double alpha=2.0,beta=1.0,gama=3.0;
-
-        t0 = (double)cvGetTickCount();//cpu start
-        cv::addWeighted(mat1,alpha,mat2,beta,gama,dst);
-        t0 = (double)cvGetTickCount() - t0;//cpu end
-
-        t1 = (double)cvGetTickCount();//gpu start1
-        cv::ocl::oclMat clmat1(mat1),clmat2(mat2),cldst;
-
-        t2=(double)cvGetTickCount();//kernel
-        cv::ocl::addWeighted(clmat1,alpha,clmat2,beta,gama, cldst);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-        cv::Mat cpu_dst;
-        cldst.download(cpu_dst);
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-        if(j == 0)
-            continue;
-        totalgputick=t1+totalgputick;
-        totalcputick=t0+totalcputick;
-        totalgputick_kernel=t2+totalgputick_kernel;
+        TEST_CYCLE() cv::addWeighted(src1, alpha, src2, beta, gama, dst);
 
+        SANITY_CHECK(dst);
     }
-    cout << "average cpu runtime is  " << totalcputick/((double)cvGetTickFrequency()* LOOP_TIMES *1000.) << "ms" << endl;
-    cout << "average gpu runtime is  " << totalgputick/((double)cvGetTickFrequency()* LOOP_TIMES *1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel/((double)cvGetTickFrequency()* LOOP_TIMES *1000.) << "ms" << endl;
-
-#else
-    //for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    //	{
-    double alpha=2.0,beta=1.0,gama=3.0;
-    cv::ocl::oclMat clmat1(mat1),clmat2(mat2),cldst;
-    //if(j==0){cout<<"no roi:";}else{cout<<"\nwith roi:";};
-    cv::ocl::addWeighted(clmat1,alpha,clmat2,beta,gama, cldst);
-    //	};
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
-
-*/
-//********test****************
-
-INSTANTIATE_TEST_CASE_P(Arithm, Lut, Combine(
-                            Values(CV_8UC1, CV_8UC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Exp, Combine(
-                            Values(CV_32FC1, CV_32FC1),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Log, Combine(
-                            Values(CV_32FC1, CV_32FC1),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Add, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1,  CV_32FC4),
-                            Values(false)));
-
-INSTANTIATE_TEST_CASE_P(Arithm, Mul, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Div, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-
-INSTANTIATE_TEST_CASE_P(Arithm, Absdiff, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, CartToPolar, Combine(
-                            Values(CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, PolarToCart, Combine(
-                            Values(CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Magnitude, Combine(
-                            Values(CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Transpose, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Flip, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, MinMax, Combine(
-                            Values(CV_8UC1, CV_32FC1),
-                            Values(false)));
-
-INSTANTIATE_TEST_CASE_P(Arithm, MinMaxLoc, Combine(
-                            Values(CV_8UC1, CV_32FC1),
-                            Values(false)));
-
-INSTANTIATE_TEST_CASE_P(Arithm, Sum, Combine(
-                            Values(CV_8U, CV_32S, CV_32F),
-                            Values(false)));
-
-INSTANTIATE_TEST_CASE_P(Arithm, CountNonZero, Combine(
-                            Values(CV_8U, CV_32S, CV_32F),
-                            Values(false)));
-
-
-INSTANTIATE_TEST_CASE_P(Arithm, Phase, Combine(Values(CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-
-INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_and, Combine(
-                            Values(CV_8UC1, CV_32SC1, CV_32SC4, CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_or, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_xor, Combine(
-                            Values(CV_8UC1, CV_32SC1, CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_not, Combine(
-                            Values(CV_8UC1, CV_32SC1, CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Compare, Combine(Values(CV_8UC1, CV_16UC1, CV_16SC1, CV_32SC1, CV_32FC1, CV_64FC1), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, Pow, Combine(Values(CV_32FC1, CV_32FC4), Values(false)));
-//Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, MagnitudeSqr, Combine(
-                            Values(CV_32FC1, CV_32FC1),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(Arithm, AddWeighted, Combine(
-                            Values(CV_8UC1, CV_32SC1, CV_32FC1),
-                            Values(false))); // Values(false) is the reserved parameter
-
-
-
-
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/perf/perf_bgfg.cpp b/modules/ocl/perf/perf_bgfg.cpp
new file mode 100644
index 0000000000..9ccd1657e4
--- /dev/null
+++ b/modules/ocl/perf/perf_bgfg.cpp
@@ -0,0 +1,282 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "perf_precomp.hpp"
+using namespace perf;
+using namespace std;
+using namespace cv::ocl;
+using namespace cv;
+using std::tr1::tuple;
+using std::tr1::get;
+#if defined(HAVE_XINE)         || \
+    defined(HAVE_GSTREAMER)    || \
+    defined(HAVE_QUICKTIME)    || \
+    defined(HAVE_AVFOUNDATION) || \
+    defined(HAVE_FFMPEG)       || \
+    defined(WIN32)
+
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 1
+#else
+#  define BUILD_WITH_VIDEO_INPUT_SUPPORT 0
+#endif
+
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+static void cvtFrameFmt(vector<Mat>& input, vector<Mat>& output)
+{
+    for(int i = 0; i< (int)(input.size()); i++)
+    {
+        cvtColor(input[i], output[i], COLOR_RGB2GRAY);
+    }
+}
+//prepare data for CPU
+static void prepareData(VideoCapture& cap, int cn, vector<Mat>& frame_buffer)
+{
+    cv::Mat frame;
+    std::vector<Mat> frame_buffer_init;
+    int nFrame = (int)frame_buffer.size();
+    for(int i = 0; i < nFrame; i++)
+    {
+        cap >> frame;
+        ASSERT_FALSE(frame.empty());
+        frame_buffer_init.push_back(frame);
+    }
+
+    if(cn == 1)
+        cvtFrameFmt(frame_buffer_init, frame_buffer);
+    else
+        frame_buffer = frame_buffer_init;
+}
+//copy CPU data to GPU
+static void prepareData(vector<Mat>& frame_buffer, vector<oclMat>& frame_buffer_ocl)
+{
+    for(int i = 0; i < (int)frame_buffer.size(); i++)
+        frame_buffer_ocl.push_back(cv::ocl::oclMat(frame_buffer[i]));
+}
+#endif
+///////////// MOG ////////////////////////
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+typedef tuple<string, int, double> VideoMOGParamType;
+typedef TestBaseWithParam<VideoMOGParamType> VideoMOGFixture;
+
+PERF_TEST_P(VideoMOGFixture, MOG,
+            ::testing::Combine(::testing::Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+            ::testing::Values(1, 3),
+            ::testing::Values(0.0, 0.01)))
+{
+    VideoMOGParamType params = GetParam();
+
+    const string inputFile = perf::TestBase::getDataPath(get<0>(params));
+    const int cn = get<1>(params);
+    const float learningRate = static_cast<float>(get<2>(params));
+
+    const int nFrame = 5;
+
+    Mat foreground_cpu;
+    std::vector<Mat> frame_buffer(nFrame);
+    std::vector<oclMat> frame_buffer_ocl;
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    prepareData(cap, cn, frame_buffer);
+
+    cv::Mat foreground;
+    cv::ocl::oclMat foreground_d;
+    if(RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE()
+        {
+            cv::Ptr<cv::BackgroundSubtractorMOG> mog = createBackgroundSubtractorMOG();
+            foreground.release();
+            for (int i = 0; i < nFrame; i++)
+            {
+                mog->apply(frame_buffer[i], foreground, learningRate);
+            }
+        }
+        SANITY_CHECK(foreground);
+    }else if(RUN_OCL_IMPL)
+    {
+        prepareData(frame_buffer, frame_buffer_ocl);
+        CV_Assert((int)(frame_buffer_ocl.size()) == nFrame);
+        OCL_TEST_CYCLE()
+        {
+            cv::ocl::MOG d_mog;
+            foreground_d.release();
+            for (int i = 0; i < nFrame; ++i)
+            {
+                d_mog(frame_buffer_ocl[i], foreground_d, learningRate);
+            }
+        }
+        foreground_d.download(foreground);
+        SANITY_CHECK(foreground);
+    }else
+        OCL_PERF_ELSE
+}
+#endif
+
+///////////// MOG2 ////////////////////////
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+typedef tuple<string, int> VideoMOG2ParamType;
+typedef TestBaseWithParam<VideoMOG2ParamType> VideoMOG2Fixture;
+
+PERF_TEST_P(VideoMOG2Fixture, MOG2,
+            ::testing::Combine(::testing::Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+            ::testing::Values(1, 3)))
+{
+    VideoMOG2ParamType params = GetParam();
+
+    const string inputFile = perf::TestBase::getDataPath(get<0>(params));
+    const int cn = get<1>(params);
+    int nFrame = 5;
+
+    std::vector<cv::Mat> frame_buffer(nFrame);
+    std::vector<cv::ocl::oclMat> frame_buffer_ocl;
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+    prepareData(cap, cn, frame_buffer);
+    cv::Mat foreground;
+    cv::ocl::oclMat foreground_d;
+
+    if(RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE()
+        {
+            cv::Ptr<cv::BackgroundSubtractorMOG2> mog2 = createBackgroundSubtractorMOG2();
+            mog2->set("detectShadows", false);
+            foreground.release();
+
+            for (int i = 0; i < nFrame; i++)
+            {
+                mog2->apply(frame_buffer[i], foreground);
+            }
+        }
+        SANITY_CHECK(foreground);
+    }else if(RUN_OCL_IMPL)
+    {
+        prepareData(frame_buffer, frame_buffer_ocl);
+        CV_Assert((int)(frame_buffer_ocl.size()) == nFrame);
+        OCL_TEST_CYCLE()
+        {
+            cv::ocl::MOG2 d_mog2;
+            foreground_d.release();
+            for (int i = 0; i < nFrame; i++)
+            {
+                d_mog2(frame_buffer_ocl[i], foreground_d);
+            }
+        }
+        foreground_d.download(foreground);
+        SANITY_CHECK(foreground);
+    }else
+        OCL_PERF_ELSE
+}
+#endif
+
+///////////// MOG2_GetBackgroundImage //////////////////
+#if BUILD_WITH_VIDEO_INPUT_SUPPORT
+
+typedef TestBaseWithParam<VideoMOG2ParamType> Video_MOG2GetBackgroundImage;
+
+PERF_TEST_P(Video_MOG2GetBackgroundImage, MOG2,
+            ::testing::Combine(::testing::Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"),
+            ::testing::Values(3)))
+{
+    VideoMOG2ParamType params = GetParam();
+
+    const string inputFile = perf::TestBase::getDataPath(get<0>(params));
+    const int cn = get<1>(params);
+    int nFrame = 5;
+
+    std::vector<cv::Mat> frame_buffer(nFrame);
+    std::vector<cv::ocl::oclMat> frame_buffer_ocl;
+
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    prepareData(cap, cn, frame_buffer);
+
+    cv::Mat foreground;
+    cv::Mat background;
+    cv::ocl::oclMat foreground_d;
+    cv::ocl::oclMat background_d;
+
+    if(RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE()
+        {
+            cv::Ptr<cv::BackgroundSubtractorMOG2> mog2 = createBackgroundSubtractorMOG2();
+            mog2->set("detectShadows", false);
+            foreground.release();
+            background.release();
+            for (int i = 0; i < nFrame; i++)
+            {
+                mog2->apply(frame_buffer[i], foreground);
+            }
+            mog2->getBackgroundImage(background);
+        }
+        SANITY_CHECK(background);
+    }else if(RUN_OCL_IMPL)
+    {
+        prepareData(frame_buffer, frame_buffer_ocl);
+        CV_Assert((int)(frame_buffer_ocl.size()) == nFrame);
+        OCL_TEST_CYCLE()
+        {
+            cv::ocl::MOG2 d_mog2;
+            foreground_d.release();
+            background_d.release();
+            for (int i = 0; i < nFrame; i++)
+            {
+                d_mog2(frame_buffer_ocl[i], foreground_d);
+            }
+            d_mog2.getBackgroundImage(background_d);
+        }
+        background_d.download(background);
+        SANITY_CHECK(background);
+    }else
+        OCL_PERF_ELSE
+}
+#endif
diff --git a/modules/ocl/perf/perf_blend.cpp b/modules/ocl/perf/perf_blend.cpp
index f78f7d6b2c..018ec63154 100644
--- a/modules/ocl/perf/perf_blend.cpp
+++ b/modules/ocl/perf/perf_blend.cpp
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -43,80 +44,71 @@
 //
 //M*/
 
-#include "precomp.hpp"
-#include <iomanip>
+#include "perf_precomp.hpp"
 
-#ifdef HAVE_OPENCL
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+using namespace perf;
 
-PARAM_TEST_CASE(Blend, MatType, int)
+///////////// blend ////////////////////////
+
+template <typename T>
+static void blendLinearGold(const cv::Mat &img1, const cv::Mat &img2,
+                            const cv::Mat &weights1, const cv::Mat &weights2,
+                            cv::Mat &result_gold)
 {
-    int type;
-    int channels;
-    std::vector<cv::ocl::Info> oclinfo;
+    result_gold.create(img1.size(), img1.type());
 
-    virtual void SetUp()
+    int cn = img1.channels();
+
+    for (int y = 0; y < img1.rows; ++y)
     {
+        const float *weights1_row = weights1.ptr<float>(y);
+        const float *weights2_row = weights2.ptr<float>(y);
+        const T *img1_row = img1.ptr<T>(y);
+        const T *img2_row = img2.ptr<T>(y);
+        T *result_gold_row = result_gold.ptr<T>(y);
 
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-};
-
-TEST_P(Blend, Performance)
-{
-    cv::Size size(MWIDTH, MHEIGHT);
-    cv::Mat img1_host = randomMat(size, CV_MAKETYPE(type, channels), 0, type == CV_8U ? 255.0 : 1.0);
-    cv::Mat img2_host = randomMat(size, CV_MAKETYPE(type, channels), 0, type == CV_8U ? 255.0 : 1.0);
-    cv::Mat weights1 = randomMat(size, CV_32F, 0, 1);
-    cv::Mat weights2 = randomMat(size, CV_32F, 0, 1);
-    cv::ocl::oclMat gimg1(size, CV_MAKETYPE(type, channels)), gimg2(size, CV_MAKETYPE(type, channels)), gweights1(size, CV_32F), gweights2(size, CV_32F);
-    cv::ocl::oclMat gdst(size, CV_MAKETYPE(type, channels));
-
-
-    double totalgputick_all = 0;
-    double totalgputick_kernel = 0;
-    double t1 = 0;
-    double t2 = 0;
-
-    for (int j = 0; j < LOOP_TIMES + 1; j ++) //LOOP_TIMES=100
-    {
-        t1 = (double)cvGetTickCount();
-        cv::ocl::oclMat gimg1 = cv::ocl::oclMat(img1_host);
-        cv::ocl::oclMat gimg2 = cv::ocl::oclMat(img2_host);
-        cv::ocl::oclMat gweights1 = cv::ocl::oclMat(weights1);
-        cv::ocl::oclMat gweights2 = cv::ocl::oclMat(weights1);
-
-        t2 = (double)cvGetTickCount();
-        cv::ocl::blendLinear(gimg1, gimg2, gweights1, gweights2, gdst);
-        t2 = (double)cvGetTickCount() - t2;
-
-        cv::Mat m;
-        gdst.download(m);
-        t1 = (double)cvGetTickCount() - t1;
-
-        if (j == 0)
+        for (int x = 0; x < img1.cols * cn; ++x)
         {
-            continue;
+            int x1 = x * cn;
+            float w1 = weights1_row[x];
+            float w2 = weights2_row[x];
+            result_gold_row[x] = static_cast<T>((img1_row[x1] * w1
+                                                 + img2_row[x1] * w2) / (w1 + w2 + 1e-5f));
         }
-
-        totalgputick_all = t1 + totalgputick_all;
-        totalgputick_kernel = t2 + totalgputick_kernel;
-    };
-
-    cout << "average gpu total  runtime is  " << totalgputick_all / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-    cout << "average gpu runtime without data transfering  is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
+    }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Blend, Combine(
-                            Values(CV_8U, CV_32F), Values(1, 4)));
-#endif
\ No newline at end of file
+typedef TestBaseWithParam<Size> blendLinearFixture;
+
+PERF_TEST_P(blendLinearFixture, blendLinear, OCL_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+    const int type = CV_8UC1;
+
+    Mat src1(srcSize, type), src2(srcSize, CV_8UC1), dst;
+    Mat weights1(srcSize, CV_32FC1), weights2(srcSize, CV_32FC1);
+
+    declare.in(src1, src2, WARMUP_RNG);
+    randu(weights1, 0.0f, 1.0f);
+    randu(weights2, 0.0f, 1.0f);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2), oclDst;
+        ocl::oclMat oclWeights1(weights1), oclWeights2(weights2);
+
+        OCL_TEST_CYCLE() cv::ocl::blendLinear(oclSrc1, oclSrc2, oclWeights1, oclWeights2, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() blendLinearGold<uchar>(src1, src2, weights1, weights2, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_brute_force_matcher.cpp b/modules/ocl/perf/perf_brute_force_matcher.cpp
new file mode 100644
index 0000000000..33c42c72dc
--- /dev/null
+++ b/modules/ocl/perf/perf_brute_force_matcher.cpp
@@ -0,0 +1,177 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+#define OCL_BFMATCHER_TYPICAL_MAT_SIZES ::testing::Values(cv::Size(128, 500), cv::Size(128, 1000), cv::Size(128, 2000))
+
+//////////////////// BruteForceMatch /////////////////
+
+typedef TestBaseWithParam<Size> BruteForceMatcherFixture;
+
+PERF_TEST_P(BruteForceMatcherFixture, DISABLED_match,
+            OCL_BFMATCHER_TYPICAL_MAT_SIZES) // TODO too big difference between implementations
+{
+    const Size srcSize = GetParam();
+
+    vector<DMatch> matches;
+    Mat query(srcSize, CV_32F), train(srcSize, CV_32F);
+    declare.in(query, train).time(srcSize.height == 2000 ? 9 : 4 );
+    randu(query, 0.0f, 1.0f);
+    randu(train, 0.0f, 1.0f);
+
+    if (RUN_PLAIN_IMPL)
+    {
+        BFMatcher matcher(NORM_L2);
+        TEST_CYCLE() matcher.match(query, train, matches);
+
+        SANITY_CHECK_MATCHES(matches);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::BruteForceMatcher_OCL_base oclMatcher(ocl::BruteForceMatcher_OCL_base::L2Dist);
+        ocl::oclMat oclQuery(query), oclTrain(train);
+        ocl::oclMat oclTrainIdx, oclDistance;
+
+        OCL_TEST_CYCLE()
+            oclMatcher.matchSingle(oclQuery, oclTrain, oclTrainIdx, oclDistance);
+
+        oclMatcher.matchDownload(oclTrainIdx, oclDistance, matches);
+
+        SANITY_CHECK_MATCHES(matches);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+PERF_TEST_P(BruteForceMatcherFixture, DISABLED_knnMatch,
+            OCL_BFMATCHER_TYPICAL_MAT_SIZES) // TODO too big difference between implementations
+{
+    const Size srcSize = GetParam();
+
+    vector<vector<DMatch> > matches(2);
+    Mat query(srcSize, CV_32F), train(srcSize, CV_32F);
+    randu(query, 0.0f, 1.0f);
+    randu(train, 0.0f, 1.0f);
+
+    declare.in(query, train);
+    if (srcSize.height == 2000)
+        declare.time(9);
+
+    if (RUN_PLAIN_IMPL)
+    {
+        BFMatcher matcher(NORM_L2);
+        TEST_CYCLE() matcher.knnMatch(query, train, matches, 2);
+
+        std::vector<DMatch> & matches0 = matches[0], & matches1 = matches[1];
+        SANITY_CHECK_MATCHES(matches0);
+        SANITY_CHECK_MATCHES(matches1);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::BruteForceMatcher_OCL_base oclMatcher(ocl::BruteForceMatcher_OCL_base::L2Dist);
+        ocl::oclMat oclQuery(query), oclTrain(train);
+        ocl::oclMat oclTrainIdx, oclDistance, oclAllDist;
+
+        OCL_TEST_CYCLE()
+                oclMatcher.knnMatchSingle(oclQuery, oclTrain, oclTrainIdx, oclDistance, oclAllDist, 2);
+
+        oclMatcher.knnMatchDownload(oclTrainIdx, oclDistance, matches);
+
+        std::vector<DMatch> & matches0 = matches[0], & matches1 = matches[1];
+        SANITY_CHECK_MATCHES(matches0);
+        SANITY_CHECK_MATCHES(matches1);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+PERF_TEST_P(BruteForceMatcherFixture, radiusMatch,
+            OCL_BFMATCHER_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+
+    const float max_distance = 2.0f;
+    vector<vector<DMatch> > matches(2);
+    Mat query(srcSize, CV_32F), train(srcSize, CV_32F);
+    declare.in(query, train);
+
+    randu(query, 0.0f, 1.0f);
+    randu(train, 0.0f, 1.0f);
+
+    if (srcSize.height == 2000)
+        declare.time(9.15);
+
+    if (RUN_PLAIN_IMPL)
+    {
+        cv::BFMatcher matcher(NORM_L2);
+        TEST_CYCLE() matcher.radiusMatch(query, train, matches, max_distance);
+
+        std::vector<DMatch> & matches0 = matches[0], & matches1 = matches[1];
+        SANITY_CHECK_MATCHES(matches0);
+        SANITY_CHECK_MATCHES(matches1);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclQuery(query), oclTrain(train);
+        ocl::BruteForceMatcher_OCL_base oclMatcher(ocl::BruteForceMatcher_OCL_base::L2Dist);
+        ocl::oclMat oclTrainIdx, oclDistance, oclNMatches;
+
+        OCL_TEST_CYCLE()
+                oclMatcher.radiusMatchSingle(oclQuery, oclTrain, oclTrainIdx, oclDistance, oclNMatches, max_distance);
+
+        oclMatcher.radiusMatchDownload(oclTrainIdx, oclDistance, oclNMatches, matches);
+
+        std::vector<DMatch> & matches0 = matches[0], & matches1 = matches[1];
+        SANITY_CHECK_MATCHES(matches0);
+        SANITY_CHECK_MATCHES(matches1);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+#undef OCL_BFMATCHER_TYPICAL_MAT_SIZES
diff --git a/modules/ocl/test/test_columnsum.cpp b/modules/ocl/perf/perf_calib3d.cpp
similarity index 60%
rename from modules/ocl/test/test_columnsum.cpp
rename to modules/ocl/perf/perf_calib3d.cpp
index 9bd2e6f8ef..997e848561 100644
--- a/modules/ocl/test/test_columnsum.cpp
+++ b/modules/ocl/perf/perf_calib3d.cpp
@@ -15,8 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//	   Chunpeng Zhang chunpeng@multicorewareinc.com
-//
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -31,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -44,63 +44,42 @@
 //
 //M*/
 
-#include "precomp.hpp"
-#include <iomanip>
+#include "perf_precomp.hpp"
 
-///////////////////////////////////////////////////////////////////////////////
-/// ColumnSum
+///////////// StereoMatchBM ////////////////////////
 
-#ifdef HAVE_OPENCL
-
-////////////////////////////////////////////////////////////////////////
-// ColumnSum
-
-PARAM_TEST_CASE(ColumnSum, cv::Size, bool )
+PERF_TEST(StereoMatchBMFixture, StereoMatchBM)
 {
-    cv::Size size;
-    cv::Mat src;
-    bool useRoi;
-    //std::vector<cv::ocl::Info> oclinfo;
+    Mat left_image = imread(getDataPath("gpu/stereobm/aloe-L.png"), cv::IMREAD_GRAYSCALE);
+    Mat right_image = imread(getDataPath("gpu/stereobm/aloe-R.png"), cv::IMREAD_GRAYSCALE);
 
-    virtual void SetUp()
+    ASSERT_TRUE(!left_image.empty()) << "no input image";
+    ASSERT_TRUE(!right_image.empty()) << "no input image";
+    ASSERT_TRUE(right_image.size() == left_image.size());
+    ASSERT_TRUE(right_image.size() == left_image.size());
+
+    const int n_disp = 128, winSize = 19;
+    Mat disp(left_image.size(), CV_16SC1);
+
+    declare.in(left_image, right_image).out(disp);
+
+    if (RUN_OCL_IMPL)
     {
-        size = GET_PARAM(0);
-        useRoi = GET_PARAM(1);
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
+        ocl::oclMat oclLeft(left_image), oclRight(right_image),
+                oclDisp(left_image.size(), CV_16SC1);
+        ocl::StereoBM_OCL oclBM(0, n_disp, winSize);
+
+        OCL_TEST_CYCLE() oclBM(oclLeft, oclRight, oclDisp);
     }
-};
-
-TEST_P(ColumnSum, Accuracy)
-{
-    cv::Mat src = randomMat(size, CV_32FC1);
-    cv::ocl::oclMat d_dst;
-    cv::ocl::oclMat d_src(src);
-
-    cv::ocl::columnSum(d_src, d_dst);
-
-    cv::Mat dst(d_dst);
-
-    for (int j = 0; j < src.cols; ++j)
+    else if (RUN_PLAIN_IMPL)
     {
-        float gold = src.at<float>(0, j);
-        float res = dst.at<float>(0, j);
-        ASSERT_NEAR(res, gold, 1e-5);
-    }
+        Ptr<StereoBM> bm = createStereoBM(n_disp, winSize);
 
-    for (int i = 1; i < src.rows; ++i)
-    {
-        for (int j = 0; j < src.cols; ++j)
-        {
-            float gold = src.at<float>(i, j) += src.at<float>(i - 1, j);
-            float res = dst.at<float>(i, j);
-            ASSERT_NEAR(res, gold, 1e-5);
-        }
+        TEST_CYCLE() bm->compute(left_image, right_image, disp);
     }
+    else
+        OCL_PERF_ELSE
+
+    int value = 0;
+    SANITY_CHECK(value);
 }
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, ColumnSum, testing::Combine(
-                            DIFFERENT_SIZES, testing::Values(Inverse(false), Inverse(true))));
-
-
-#endif
diff --git a/modules/ocl/perf/perf_canny.cpp b/modules/ocl/perf/perf_canny.cpp
index eb895df5ec..259684092f 100644
--- a/modules/ocl/perf/perf_canny.cpp
+++ b/modules/ocl/perf/perf_canny.cpp
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -42,112 +43,34 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-#include <iomanip>
-#ifdef HAVE_OPENCL
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+using namespace perf;
 
-#ifndef MWC_TEST_UTILITY
-#define MWC_TEST_UTILITY
+///////////// Canny ////////////////////////
 
-// Param class
-#ifndef IMPLEMENT_PARAM_CLASS
-#define IMPLEMENT_PARAM_CLASS(name, type) \
-class name \
-    { \
-    public: \
-    name ( type arg = type ()) : val_(arg) {} \
-    operator type () const {return val_;} \
-    private: \
-    type val_; \
-    }; \
-    inline void PrintTo( name param, std::ostream* os) \
-    { \
-    *os << #name <<  "(" << testing::PrintToString(static_cast< type >(param)) << ")"; \
-    }
-
-IMPLEMENT_PARAM_CLASS(Channels, int)
-#endif // IMPLEMENT_PARAM_CLASS
-#endif // MWC_TEST_UTILITY
-
-////////////////////////////////////////////////////////
-// Canny1
-extern std::string workdir;
-IMPLEMENT_PARAM_CLASS(AppertureSize, int);
-IMPLEMENT_PARAM_CLASS(L2gradient, bool);
-
-PARAM_TEST_CASE(Canny1, AppertureSize, L2gradient)
+PERF_TEST(CannyFixture, Canny)
 {
-    int apperture_size;
-    bool useL2gradient;
-    //std::vector<cv::ocl::Info> oclinfo;
+    Mat img = imread(getDataPath("gpu/stereobm/aloe-L.png"), cv::IMREAD_GRAYSCALE),
+            edges(img.size(), CV_8UC1);
+    ASSERT_TRUE(!img.empty()) << "can't open aloe-L.png";
 
-    virtual void SetUp()
+    declare.in(img).out(edges);
+
+    if (RUN_OCL_IMPL)
     {
-        apperture_size = GET_PARAM(0);
-        useL2gradient = GET_PARAM(1);
+        ocl::oclMat oclImg(img), oclEdges(img.size(), CV_8UC1);
 
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
+        OCL_TEST_CYCLE() ocl::Canny(oclImg, oclEdges, 50.0, 100.0);
+        oclEdges.download(edges);
     }
-};
-
-TEST_P(Canny1, Performance)
-{
-    cv::Mat img = readImage(workdir + "fruits.jpg", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    double low_thresh = 100.0;
-    double high_thresh = 150.0;
-
-    cv::Mat edges_gold;
-    cv::ocl::oclMat edges;
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat ocl_img = cv::ocl::oclMat(img);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::Canny(ocl_img, edges, low_thresh, high_thresh, apperture_size, useL2gradient);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        edges.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
+        TEST_CYCLE() Canny(img, edges, 50.0, 100.0);
     }
+    else
+        OCL_PERF_ELSE
 
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
+    int value = 0;
+    SANITY_CHECK(value);
 }
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Canny1, testing::Combine(
-                            testing::Values(AppertureSize(3), AppertureSize(5)),
-                            testing::Values(L2gradient(false), L2gradient(true))));
-
-
-
-#endif  //Have opencl
\ No newline at end of file
diff --git a/modules/ocl/src/precomp.cpp b/modules/ocl/perf/perf_color.cpp
similarity index 69%
rename from modules/ocl/src/precomp.cpp
rename to modules/ocl/perf/perf_color.cpp
index 766138c577..b66fc2b0ae 100644
--- a/modules/ocl/src/precomp.cpp
+++ b/modules/ocl/perf/perf_color.cpp
@@ -10,11 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -29,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -41,7 +43,36 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-//CriticalSection cs;
-/* End of file. */
+using namespace perf;
+
+///////////// cvtColor////////////////////////
+
+typedef TestBaseWithParam<Size> cvtColorFixture;
+
+PERF_TEST_P(cvtColorFixture, cvtColor, OCL_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+
+    Mat src(srcSize, CV_8UC4), dst(srcSize, CV_8UC4);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(src.size(), CV_8UC4);
+
+        OCL_TEST_CYCLE() ocl::cvtColor(oclSrc, oclDst, COLOR_RGBA2GRAY, 4);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::cvtColor(src, dst, COLOR_RGBA2GRAY, 4);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_fft.cpp b/modules/ocl/perf/perf_fft.cpp
index c9c19d0d49..4cba47e960 100644
--- a/modules/ocl/perf/perf_fft.cpp
+++ b/modules/ocl/perf/perf_fft.cpp
@@ -15,7 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Fangfangbai, fangfang@multicorewareinc.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -43,84 +44,45 @@
 //
 //M*/
 
-#include "precomp.hpp"
-using namespace std;
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+///////////// dft ////////////////////////
+
+typedef TestBaseWithParam<Size> dftFixture;
+
 #ifdef HAVE_CLAMDFFT
-////////////////////////////////////////////////////////////////////////////
-// Dft
-PARAM_TEST_CASE(Dft, cv::Size, bool)
+
+PERF_TEST_P(dftFixture, dft, OCL_TYPICAL_MAT_SIZES)
 {
-    cv::Size dft_size;
-    bool	 dft_rows;
-    vector<cv::ocl::Info> info;
-    virtual void SetUp()
+    const Size srcSize = GetParam();
+
+    Mat src(srcSize, CV_32FC2), dst;
+    randu(src, 0.0f, 1.0f);
+    declare.in(src);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(7.4);
+
+    if (RUN_OCL_IMPL)
     {
-        dft_size = GET_PARAM(0);
-        dft_rows = GET_PARAM(1);
-        cv::ocl::getDevice(info);
+        ocl::oclMat oclSrc(src), oclDst;
+
+        OCL_TEST_CYCLE() cv::ocl::dft(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1.5);
     }
-};
-
-TEST_P(Dft, C2C)
-{
-    cv::Mat a = randomMat(dft_size, CV_32FC2, 0.0, 10.0);
-    int flags = 0;
-    flags |= dft_rows ? cv::DFT_ROWS : 0;
-
-    cv::ocl::oclMat d_b;
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t1 = 0;
-    double t2 = 0;
-
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
+        TEST_CYCLE() cv::dft(src, dst);
 
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat ga = cv::ocl::oclMat(a); //upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::dft(ga, d_b, a.size(), flags);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        d_b.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
+        SANITY_CHECK(dst);
     }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+    else
+        OCL_PERF_ELSE
 }
 
-
-
-TEST_P(Dft, R2CthenC2R)
-{
-    cv::Mat a = randomMat(dft_size, CV_32FC1, 0.0, 10.0);
-
-    int flags = 0;
-    //flags |= dft_rows ? cv::DFT_ROWS : 0; // not supported yet
-
-    cv::ocl::oclMat d_b, d_c;
-
-    cv::ocl::dft(cv::ocl::oclMat(a), d_b, a.size(), flags);
-    cv::ocl::dft(d_b, d_c, a.size(), flags + cv::DFT_INVERSE + cv::DFT_REAL_OUTPUT);
-
-    EXPECT_MAT_NEAR(a, d_c, a.size().area() * 1e-4, "");
-}
-
-//INSTANTIATE_TEST_CASE_P(ocl_DFT, Dft, testing::Combine(
-//						testing::Values(cv::Size(1280, 1024), cv::Size(1920, 1080),cv::Size(1800, 1500)),
-//						testing::Values(false, true)));
-
-#endif // HAVE_CLAMDFFT
+#endif
diff --git a/modules/ocl/perf/perf_filters.cpp b/modules/ocl/perf/perf_filters.cpp
index 100a1c59d9..aa562412bc 100644
--- a/modules/ocl/perf/perf_filters.cpp
+++ b/modules/ocl/perf/perf_filters.cpp
@@ -10,15 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Niko Li, newlife20080214@gmail.com
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//    Zero Lin, Zero.Lin@amd.com
-//    Zhang Ying, zhangying913@gmail.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -33,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -45,1165 +43,360 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
+using namespace perf;
+using std::tr1::get;
+using std::tr1::tuple;
 
-#ifdef HAVE_OPENCL
+///////////// Blur////////////////////////
 
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-//using namespace cv::ocl;
+typedef Size_MatType BlurFixture;
 
-PARAM_TEST_CASE(FilterTestBase, MatType, bool)
+PERF_TEST_P(BlurFixture, Blur,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    cv::Scalar val;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params), ksize(3, 3);
+    const int type = get<1>(params), bordertype = BORDER_CONSTANT;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mask;
-    cv::Mat dst;
-    cv::Mat dst1; //bak, for two outputs
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int dstx;
-    int dsty;
-    int maskx;
-    int masky;
+    if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(5);
 
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-    cv::ocl::oclMat gdst1_whole; //bak
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gmat2;
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gdst1;   //bak
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::blur(oclSrc, oclDst, ksize, Point(-1, -1), bordertype);
 
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        mat2 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        dst1  = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
+        oclDst.download(dst);
 
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
     }
-
-    void random_roi()
+    else if (RUN_PLAIN_IMPL)
     {
-        cv::RNG &rng = TS::ptr()->get_rng();
+        TEST_CYCLE() cv::blur(src, dst, ksize, Point(-1, -1), bordertype);
 
-        //randomize ROI
-        roicols = rng.uniform(1, mat1.cols);
-        roirows = rng.uniform(1, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        src2x   = rng.uniform(0, mat2.cols - roicols);
-        src2y   = rng.uniform(0, mat2.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-        maskx   = rng.uniform(0, mask.cols - roicols);
-        masky   = rng.uniform(0, mask.rows - roirows);
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-        dst1_roi = dst1(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
     }
-
-};
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// blur
-
-PARAM_TEST_CASE(Blur, MatType, cv::Size, int)
-{
-    int type;
-    cv::Size ksize;
-    int bordertype;
-
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        ksize = GET_PARAM(1);
-        bordertype = GET_PARAM(2);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-
-
-    void Has_roi(int b)
-    {
-        if(b)
-        {
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-    }
-
-};
-
-TEST_P(Blur, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::blur(mat1_roi, dst_roi, ksize, Point(-1, -1), bordertype);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::blur(gmat1, gdst, ksize, Point(-1, -1), bordertype);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::blur(gmat1, gdst, ksize, Point(-1, -1), bordertype);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-//Laplacian
+///////////// Laplacian////////////////////////
 
-PARAM_TEST_CASE(LaplacianTestBase, MatType, int)
+typedef Size_MatType LaplacianFixture;
+
+PERF_TEST_P(LaplacianFixture, Laplacian,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    int ksize;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), ksize = 3;
 
-    //src mat
-    cv::Mat mat;
-    cv::Mat dst;
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
+    if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(6);
 
-    //src mat with roi
-    cv::Mat mat_roi;
-    cv::Mat dst_roi;
-    std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        ksize = GET_PARAM(1);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::Laplacian(oclSrc, oclDst, -1, ksize, 1);
 
-        mat  = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        if(b)
-        {
-            roicols =  mat.cols - 1;
-            roirows = mat.rows - 1;
-            srcx   = 1;
-            srcy   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat.cols;
-            roirows = mat.rows;
-            srcx = 0;
-            srcy = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat_roi = mat(Rect(srcx, srcy, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() cv::Laplacian(src, dst, -1, ksize, 1);
 
+        SANITY_CHECK(dst);
     }
-
-};
-
-struct Laplacian : LaplacianTestBase {};
-
-TEST_P(Laplacian, Accuracy)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::Laplacian(mat_roi, dst_roi, -1, ksize, 1);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat = mat_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::Laplacian(gmat, gdst, -1, ksize, 1);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat = mat_roi;
-
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::Laplacian(gmat, gdst, -1, ksize, 1);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Erode ////////////////////
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// erode & dilate
+typedef Size_MatType ErodeFixture;
 
-PARAM_TEST_CASE(ErodeDilateBase, MatType, bool)
+PERF_TEST_P(ErodeFixture, Erode,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4)))
 {
-    int type;
-    //int iterations;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), ksize = 3;
+    const Mat ker = getStructuringElement(MORPH_RECT, Size(ksize, ksize));
 
-    //erode or dilate kernel
-    cv::Mat kernel;
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst).in(ker);
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
+    if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(5);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        //  iterations = GET_PARAM(1);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type), oclKer(ker);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::erode(oclSrc, oclDst, oclKer);
 
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //		rng.fill(kernel, cv::RNG::UNIFORM, cv::Scalar::all(0), cv::Scalar::all(3));
-        kernel = randomMat(rng, Size(3, 3), CV_8UC1, 0, 3, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        if(b)
-        {
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() cv::erode(src, dst, ker);
 
+        SANITY_CHECK(dst);
     }
-
-};
-
-// erode
-
-struct Erode : ErodeDilateBase {};
-
-TEST_P(Erode, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::erode(mat1_roi, dst_roi, kernel);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::erode(gmat1, gdst, kernel);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::erode(gmat1, gdst, kernel);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
-// dilate
+///////////// Sobel ////////////////////////
 
-struct Dilate : ErodeDilateBase {};
+typedef Size_MatType SobelFixture;
 
-TEST_P(Dilate, Mat)
+PERF_TEST_P(SobelFixture, Sobel,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), dx = 1, dy = 1;
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if ((srcSize == OCL_SIZE_2000 && type == CV_8UC4) ||
+            (srcSize == OCL_SIZE_4000 && type == CV_8UC1))
+        declare.time(5.5);
+    else if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(20);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::dilate(mat1_roi, dst_roi, kernel);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        OCL_TEST_CYCLE() cv::ocl::Sobel(oclSrc, oclDst, -1, dx, dy);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::dilate(gmat1, gdst, kernel);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
+        oclDst.download(dst);
 
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::dilate(gmat1, gdst, kernel);
-    };
-#endif
+        TEST_CYCLE() cv::Sobel(src, dst, -1, dx, dy);
 
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// Sobel
+///////////// Scharr ////////////////////////
 
-PARAM_TEST_CASE(Sobel, MatType, int, int, int, int)
+typedef Size_MatType ScharrFixture;
+
+PERF_TEST_P(ScharrFixture, Scharr,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    int dx, dy, ksize, bordertype;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), dx = 1, dy = 0;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
+    if ((srcSize == OCL_SIZE_2000 && type == CV_8UC4) ||
+            (srcSize == OCL_SIZE_4000 && type == CV_8UC1))
+        declare.time(5.5);
+    else if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(21);
 
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        dx = GET_PARAM(1);
-        dy = GET_PARAM(2);
-        ksize = GET_PARAM(3);
-        bordertype = GET_PARAM(4);
-        dx = 2;
-        dy = 0;
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::Scharr(oclSrc, oclDst, -1, dx, dy);
 
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        if(b)
-        {
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() cv::Scharr(src, dst, -1, dx, dy);
 
+        SANITY_CHECK(dst);
     }
-
-};
-
-TEST_P(Sobel, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::Sobel(mat1_roi, dst_roi, -1, dx, dy, ksize, /*scale*/0.00001,/*delta*/0, bordertype);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::Sobel(gmat1, gdst, -1, dx, dy, ksize,/*scale*/0.00001,/*delta*/0, bordertype);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::Sobel(gmat1, gdst, -1, dx, dy, ksize,/*scale*/0.00001,/*delta*/0, bordertype);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// Scharr
+///////////// GaussianBlur ////////////////////////
 
-PARAM_TEST_CASE(Scharr, MatType, int, int, int)
+typedef Size_MatType GaussianBlurFixture;
+
+PERF_TEST_P(GaussianBlurFixture, GaussianBlur,
+            ::testing::Combine(::testing::Values(OCL_SIZE_1000, OCL_SIZE_2000),
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4)))
 {
-    int type;
-    int dx, dy, bordertype;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), ksize = 7;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
+    const double eps = src.depth() == CV_8U ? 1 + DBL_EPSILON : 3e-4;
 
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        dx = GET_PARAM(1);
-        dy = GET_PARAM(2);
-        bordertype = GET_PARAM(3);
-        dx = 1;
-        dy = 0;
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::GaussianBlur(oclSrc, oclDst, Size(ksize, ksize), 0);
 
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, eps);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        if(b)
-        {
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() cv::GaussianBlur(src, dst, Size(ksize, ksize), 0);
 
+        SANITY_CHECK(dst, eps);
     }
-};
-
-TEST_P(Scharr, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::Scharr(mat1_roi, dst_roi, -1, dx, dy, /*scale*/1,/*delta*/0, bordertype);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::Scharr(gmat1, gdst, -1, dx, dy,/*scale*/1,/*delta*/0, bordertype);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::Scharr(gmat1, gdst, -1, dx, dy,/*scale*/1,/*delta*/0, bordertype);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// GaussianBlur
+///////////// filter2D////////////////////////
 
-PARAM_TEST_CASE(GaussianBlur, MatType, cv::Size, int)
+typedef Size_MatType filter2DFixture;
+
+PERF_TEST_P(filter2DFixture, filter2D,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    cv::Size ksize;
-    int bordertype;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), ksize = 3;
 
-    double sigma1, sigma2;
+    Mat src(srcSize, type), dst(srcSize, type), kernel(ksize, ksize, CV_32SC1);
+    declare.in(src, WARMUP_RNG).in(kernel).out(dst);
+    randu(kernel, -3.0, 3.0);
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
+    if (srcSize == OCL_SIZE_4000 && type == CV_8UC4)
+        declare.time(8);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        ksize = GET_PARAM(1);
-        bordertype = GET_PARAM(2);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type), oclKernel(kernel);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::filter2D(oclSrc, oclDst, -1, oclKernel);
 
-        sigma1 = rng.uniform(0.1, 1.0);
-        sigma2 = rng.uniform(0.1, 1.0);
+        oclDst.download(dst);
 
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        if(b)
-        {
-            roicols =  mat1.cols - 1;
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() cv::filter2D(src, dst, -1, kernel);
 
+        SANITY_CHECK(dst);
     }
-
-};
-
-TEST_P(GaussianBlur, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::GaussianBlur(mat1_roi, dst_roi, ksize, sigma1, sigma2, bordertype);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::GaussianBlur(gmat1, gdst, ksize, sigma1, sigma2, bordertype);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::GaussianBlur(gmat1, gdst, ksize, sigma1, sigma2, bordertype);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
-//************test**********
+///////////// Bilateral////////////////////////
 
-INSTANTIATE_TEST_CASE_P(Filter, Blur, Combine(Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                        Values(cv::Size(3, 3)/*, cv::Size(5, 5), cv::Size(7, 7)*/),
-                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE, (MatType)cv::BORDER_REFLECT, (MatType)cv::BORDER_REFLECT_101)));
+typedef Size_MatType BilateralFixture;
 
+PERF_TEST_P(BilateralFixture, Bilateral,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC3)))
+{
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), d = 7;
+    double sigmacolor = 50.0, sigmaspace = 50.0;
 
-INSTANTIATE_TEST_CASE_P(Filters, Laplacian, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(1/*, 3*/)));
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-//INSTANTIATE_TEST_CASE_P(Filter, ErodeDilate, Combine(Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1, 2, 3)));
+    if (srcSize == OCL_SIZE_4000 && type == CV_8UC3)
+        declare.time(8);
 
-INSTANTIATE_TEST_CASE_P(Filter, Erode, Combine(Values(CV_8UC1, CV_8UC1), Values(false)));
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-//INSTANTIATE_TEST_CASE_P(Filter, ErodeDilate, Combine(Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1, 2, 3)));
+        OCL_TEST_CYCLE() cv::ocl::bilateralFilter(oclSrc, oclDst, d, sigmacolor, sigmaspace);
 
-INSTANTIATE_TEST_CASE_P(Filter, Dilate, Combine(Values(CV_8UC1, CV_8UC1), Values(false)));
+        oclDst.download(dst);
 
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::bilateralFilter(src, dst, d, sigmacolor, sigmaspace);
 
-INSTANTIATE_TEST_CASE_P(Filter, Sobel, Combine(Values(CV_8UC1, CV_32FC1),
-                        Values(1, 2), Values(0, 1), Values(3, 5), Values((MatType)cv::BORDER_CONSTANT,
-                                (MatType)cv::BORDER_REPLICATE)));
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
 
+///////////// adaptiveBilateral////////////////////////
 
-INSTANTIATE_TEST_CASE_P(Filter, Scharr, Combine(
-                            Values(CV_8UC1,  CV_32FC1), Values(0, 1), Values(0, 1),
-                            Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
+typedef Size_MatType adaptiveBilateralFixture;
 
-INSTANTIATE_TEST_CASE_P(Filter, GaussianBlur, Combine(
-                            Values(CV_8UC1,  CV_32FC1),
-                            Values(cv::Size(3, 3), cv::Size(5, 5)),
-                            Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
+PERF_TEST_P(adaptiveBilateralFixture, adaptiveBilateral,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC3)))
+{
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+    double sigmaspace = 10.0;
+    Size ksize(9,9);
 
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-#endif // HAVE_OPENCL
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(15);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
+
+        OCL_TEST_CYCLE() cv::ocl::adaptiveBilateralFilter(oclSrc, oclDst, ksize, sigmaspace);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1.);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::adaptiveBilateralFilter(src, dst, ksize, sigmaspace);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_gemm.cpp b/modules/ocl/perf/perf_gemm.cpp
index c3dcab34fe..803e1f91b6 100644
--- a/modules/ocl/perf/perf_gemm.cpp
+++ b/modules/ocl/perf/perf_gemm.cpp
@@ -16,6 +16,8 @@
 //
 // @Authors
 //    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -41,73 +43,46 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
+using namespace perf;
+
+///////////// gemm ////////////////////////
+
+typedef TestBaseWithParam<Size> gemmFixture;
 
-#include "precomp.hpp"
-using namespace std;
 #ifdef HAVE_CLAMDBLAS
-////////////////////////////////////////////////////////////////////////////
-// GEMM
-PARAM_TEST_CASE(Gemm, int, cv::Size, int)
+
+PERF_TEST_P(gemmFixture, gemm, ::testing::Values(OCL_SIZE_1000, OCL_SIZE_2000))
 {
-    int      type;
-    cv::Size mat_size;
-    int		 flags;
-    vector<cv::ocl::Info> info;
-    virtual void SetUp()
+    const Size srcSize = GetParam();
+
+    Mat src1(srcSize, CV_32FC1), src2(srcSize, CV_32FC1),
+            src3(srcSize, CV_32FC1), dst(srcSize, CV_32FC1);
+    declare.in(src1, src2, src3).out(dst).time(srcSize == OCL_SIZE_2000 ? 65 : 8);
+    randu(src1, -10.0f, 10.0f);
+    randu(src2, -10.0f, 10.0f);
+    randu(src3, -10.0f, 10.0f);
+
+    if (RUN_OCL_IMPL)
     {
-        type     = GET_PARAM(0);
-        mat_size = GET_PARAM(1);
-        flags    = GET_PARAM(2);
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2),
+                oclSrc3(src3), oclDst(srcSize, CV_32FC1);
 
-        cv::ocl::getDevice(info);
+        OCL_TEST_CYCLE() cv::ocl::gemm(oclSrc1, oclSrc2, 1.0, oclSrc3, 1.0, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 0.01);
     }
-};
-
-TEST_P(Gemm, Performance)
-{
-    cv::Mat a = randomMat(mat_size, type, 0.0, 10.0);
-    cv::Mat b = randomMat(mat_size, type, 0.0, 10.0);
-    cv::Mat c = randomMat(mat_size, type, 0.0, 10.0);
-    cv::ocl::oclMat ocl_dst;
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t1 = 0;
-    double t2 = 0;
-
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
+        TEST_CYCLE() cv::gemm(src1, src2, 1.0, src3, 1.0, dst);
 
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat ga = cv::ocl::oclMat(a);//upload
-        cv::ocl::oclMat gb = cv::ocl::oclMat(b);//upload
-        cv::ocl::oclMat gc = cv::ocl::oclMat(c);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::gemm(ga, gb, 1.0, gc, 1.0, ocl_dst, flags);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        ocl_dst.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
+        SANITY_CHECK(dst, 0.01);
     }
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+    else
+        OCL_PERF_ELSE
 }
 
-
-INSTANTIATE_TEST_CASE_P(ocl_gemm, Gemm, testing::Combine(
-                            testing::Values(CV_32FC1, CV_32FC2/* , CV_64FC1, CV_64FC2*/),
-                            testing::Values(cv::Size(512, 512), cv::Size(1024, 1024)),
-                            testing::Values(0, (int)cv::GEMM_1_T, (int)cv::GEMM_2_T, (int)(cv::GEMM_1_T + cv::GEMM_2_T))));
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/ocl/src/mcwutil.hpp b/modules/ocl/perf/perf_gftt.cpp
similarity index 58%
rename from modules/ocl/src/mcwutil.hpp
rename to modules/ocl/perf/perf_gftt.cpp
index a199b70d4e..8a29adc0cc 100644
--- a/modules/ocl/src/mcwutil.hpp
+++ b/modules/ocl/perf/perf_gftt.cpp
@@ -15,7 +15,7 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Peng Xiao, pengxiao@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -43,38 +43,53 @@
 //
 //M*/
 
-#ifndef _OPENCV_MCWUTIL_
-#define _OPENCV_MCWUTIL_
 
-#include "precomp.hpp"
+#include "perf_precomp.hpp"
 
-namespace cv
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
+
+///////////// GoodFeaturesToTrack ////////////////////////
+
+typedef tuple<string, double> GoodFeaturesToTrackParams;
+typedef TestBaseWithParam<GoodFeaturesToTrackParams> GoodFeaturesToTrackFixture;
+
+PERF_TEST_P(GoodFeaturesToTrackFixture, GoodFeaturesToTrack,
+            ::testing::Combine(::testing::Values(string("gpu/opticalflow/rubberwhale1.png"),
+                                                 string("gpu/stereobm/aloe-L.png")),
+                               ::testing::Range(0.0, 4.0, 3.0)))
 {
-    namespace ocl
+
+    const GoodFeaturesToTrackParams param = GetParam();
+    const string fileName = getDataPath(get<0>(param));
+    const int maxCorners = 2000;
+    const double qualityLevel = 0.01, minDistance = get<1>(param);
+
+    Mat frame = imread(fileName, IMREAD_GRAYSCALE);
+    ASSERT_TRUE(!frame.empty()) << "no input image";
+
+    vector<Point2f> pts_gold;
+    declare.in(frame);
+
+    if (RUN_OCL_IMPL)
     {
-        enum FLUSH_MODE
-        {
-            CLFINISH = 0,
-            CLFLUSH,
-            DISABLE
-        };
-        void openCLExecuteKernel2(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
-                                  size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, FLUSH_MODE finish_mode = DISABLE);
-        void openCLExecuteKernel2(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
-                                  size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
-                                  int depth, char *build_options, FLUSH_MODE finish_mode = DISABLE);
-        // bind oclMat to OpenCL image textures
-        // note:
-        //   1. there is no memory management. User need to explicitly release the resource
-        //   2. for faster clamping, there is no buffer padding for the constructed texture
-        cl_mem bindTexture(const oclMat &mat);
-        void releaseTexture(cl_mem& texture);
+        ocl::oclMat oclFrame(frame), pts_oclmat;
+        ocl::GoodFeaturesToTrackDetector_OCL detector(maxCorners, qualityLevel, minDistance);
 
-        // returns whether the current context supports image2d_t format or not
-        bool support_image2d(Context *clCxt = Context::getContext());
+        OCL_TEST_CYCLE() detector(oclFrame, pts_oclmat);
 
-    }//namespace ocl
+        detector.downloadPoints(pts_oclmat, pts_gold);
 
-}//namespace cv
+        SANITY_CHECK(pts_gold);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::goodFeaturesToTrack(frame, pts_gold,
+                                             maxCorners, qualityLevel, minDistance);
 
-#endif //_OPENCV_MCWUTIL_
+        SANITY_CHECK(pts_gold);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_haar.cpp b/modules/ocl/perf/perf_haar.cpp
index 08c066a311..9c258fe259 100644
--- a/modules/ocl/perf/perf_haar.cpp
+++ b/modules/ocl/perf/perf_haar.cpp
@@ -10,12 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -30,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -42,132 +43,46 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
 
-#ifdef HAVE_OPENCL
+using namespace perf;
 
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-using namespace cv;
-extern std::string workdir;
-struct getRect
+///////////// Haar ////////////////////////
+
+PERF_TEST(HaarFixture, Haar)
 {
-    Rect operator ()(const CvAvgComp &e) const
+    vector<Rect> faces;
+
+    Mat img = imread(getDataPath("gpu/haarcascade/basketball1.png"), IMREAD_GRAYSCALE);
+    ASSERT_TRUE(!img.empty()) << "can't open basketball1.png";
+    declare.in(img);
+
+    if (RUN_PLAIN_IMPL)
     {
-        return e.rect;
+        CascadeClassifier faceCascade;
+        ASSERT_TRUE(faceCascade.load(getDataPath("gpu/haarcascade/haarcascade_frontalface_alt.xml")))
+                << "can't load haarcascade_frontalface_alt.xml";
+
+        TEST_CYCLE() faceCascade.detectMultiScale(img, faces,
+                                                     1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
+
+        SANITY_CHECK(faces, 4 + 1e-4);
     }
-};
-
-PARAM_TEST_CASE(HaarTestBase, int, int)
-{
-    //std::vector<cv::ocl::Info> oclinfo;
-    cv::ocl::OclCascadeClassifier cascade, nestedCascade;
-    cv::CascadeClassifier cpucascade, cpunestedCascade;
-    //    Mat img;
-
-    double scale;
-    int index;
-
-    virtual void SetUp()
+    else if (RUN_OCL_IMPL)
     {
-        scale = 1.0;
-        index = 0;
-        string cascadeName = "../../../data/haarcascades/haarcascade_frontalface_alt.xml";
+        ocl::OclCascadeClassifier faceCascade;
+        ocl::oclMat oclImg(img);
 
-        if( (!cascade.load( cascadeName )) || (!cpucascade.load(cascadeName)))
-        {
-            cout << "ERROR: Could not load classifier cascade" << endl;
-            return;
-        }
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums>0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath("E:\\");
+        ASSERT_TRUE(faceCascade.load(getDataPath("gpu/haarcascade/haarcascade_frontalface_alt.xml")))
+                << "can't load haarcascade_frontalface_alt.xml";
+
+        OCL_TEST_CYCLE() faceCascade.detectMultiScale(oclImg, faces,
+                                     1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
+
+        SANITY_CHECK(faces, 4 + 1e-4);
     }
-};
-
-////////////////////////////////faceDetect/////////////////////////////////////////////////
-
-struct Haar : HaarTestBase {};
-
-TEST_F(Haar, FaceDetect)
-{
-    string imgName = workdir + "lena.jpg";
-    Mat img = imread( imgName, 1 );
-
-    if(img.empty())
-    {
-        std::cout << imgName << std::endl;
-        return ;
-    }
-
-    //int i = 0;
-    double t = 0;
-    vector<Rect> faces, oclfaces;
-
-    // const static Scalar colors[] =  { CV_RGB(0, 0, 255),
-    //                                   CV_RGB(0, 128, 255),
-    //                                   CV_RGB(0, 255, 255),
-    //                                   CV_RGB(0, 255, 0),
-    //                                   CV_RGB(255, 128, 0),
-    //                                   CV_RGB(255, 255, 0),
-    //                                   CV_RGB(255, 0, 0),
-    //                                   CV_RGB(255, 0, 255)
-    //                                 } ;
-
-    Mat gray, smallImg(cvRound (img.rows / scale), cvRound(img.cols / scale), CV_8UC1 );
-    MemStorage storage(cvCreateMemStorage(0));
-    cvtColor( img, gray, CV_BGR2GRAY );
-    resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
-    equalizeHist( smallImg, smallImg );
-
-    t = (double)cvGetTickCount();
-    for(int k = 0; k < LOOP_TIMES; k++)
-    {
-        cpucascade.detectMultiScale( smallImg, faces,  1.1,
-                                     3, 0
-                                     | CV_HAAR_SCALE_IMAGE
-                                     , Size(30, 30), Size(0, 0) );
-    }
-    t = (double)cvGetTickCount() - t ;
-    printf( "cpudetection time = %g ms\n", t / (LOOP_TIMES * (double)cvGetTickFrequency() * 1000.) );
-
-    cv::ocl::oclMat image;
-    CvSeq *_objects=NULL;
-    t = (double)cvGetTickCount();
-    for(int k = 0; k < LOOP_TIMES; k++)
-    {
-        image.upload(smallImg);
-        _objects = cascade.oclHaarDetectObjects( image, storage, 1.1,
-                   3, 0
-                   | CV_HAAR_SCALE_IMAGE
-                   , Size(30, 30), Size(0, 0) );
-    }
-    t = (double)cvGetTickCount() - t ;
-    printf( "ocldetection time = %g ms\n", t / (LOOP_TIMES * (double)cvGetTickFrequency() * 1000.) );
-    vector<CvAvgComp> vecAvgComp;
-    Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
-    oclfaces.resize(vecAvgComp.size());
-    std::transform(vecAvgComp.begin(), vecAvgComp.end(), oclfaces.begin(), getRect());
-
-    //for( vector<Rect>::const_iterator r = faces.begin(); r != faces.end(); r++, i++ )
-    //{
-    //	Mat smallImgROI;
-    //	Point center;
-    //	Scalar color = colors[i%8];
-    //	int radius;
-    //	center.x = cvRound((r->x + r->width*0.5)*scale);
-    //	center.y = cvRound((r->y + r->height*0.5)*scale);
-    //	radius = cvRound((r->width + r->height)*0.25*scale);
-    //	circle( img, center, radius, color, 3, 8, 0 );
-    //}
-    //namedWindow("result");
-    //imshow("result",img);
-    //waitKey(0);
-    //destroyAllWindows();
-
+    else
+        OCL_PERF_ELSE
 }
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/perf/perf_hog.cpp b/modules/ocl/perf/perf_hog.cpp
index fd58808a88..15846d8315 100644
--- a/modules/ocl/perf/perf_hog.cpp
+++ b/modules/ocl/perf/perf_hog.cpp
@@ -15,7 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Peng Xiao, pengxiao@multicorewareinc.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -42,125 +43,39 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-#include <iomanip>
+using namespace perf;
 
-#ifdef HAVE_OPENCL
+///////////// HOG////////////////////////
 
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-extern std::string workdir;
-
-#ifndef MWC_TEST_UTILITY
-#define MWC_TEST_UTILITY
-
-// Param class
-#ifndef IMPLEMENT_PARAM_CLASS
-#define IMPLEMENT_PARAM_CLASS(name, type) \
-class name \
-    { \
-    public: \
-    name ( type arg = type ()) : val_(arg) {} \
-    operator type () const {return val_;} \
-    private: \
-    type val_; \
-    }; \
-    inline void PrintTo( name param, std::ostream* os) \
-    { \
-    *os << #name <<  "(" << testing::PrintToString(static_cast< type >(param)) << ")"; \
-    }
-
-#endif // IMPLEMENT_PARAM_CLASS
-#endif // MWC_TEST_UTILITY
-
-IMPLEMENT_PARAM_CLASS(WinSizw48, bool);
-
-PARAM_TEST_CASE(HOG, WinSizw48, bool)
+PERF_TEST(HOGFixture, HOG)
 {
-    bool is48;
-    vector<float> detector;
-    virtual void SetUp()
+    Mat src = imread(getDataPath("gpu/hog/road.png"), cv::IMREAD_GRAYSCALE);
+    ASSERT_TRUE(!src.empty()) << "can't open input image road.png";
+
+    vector<cv::Rect> found_locations;
+    declare.in(src).time(5);
+
+    if (RUN_PLAIN_IMPL)
     {
-        is48 = GET_PARAM(0);
-        if(is48)
-        {
-            detector = cv::ocl::HOGDescriptor::getPeopleDetector48x96();
-        }
-        else
-        {
-            detector = cv::ocl::HOGDescriptor::getPeopleDetector64x128();
-        }
+        HOGDescriptor hog;
+        hog.setSVMDetector(hog.getDefaultPeopleDetector());
+
+        TEST_CYCLE() hog.detectMultiScale(src, found_locations);
+
+        SANITY_CHECK(found_locations, 1 + DBL_EPSILON);
     }
-};
-
-TEST_P(HOG, Performance)
-{
-    cv::Mat img = readImage(workdir + "lena.jpg", cv::IMREAD_GRAYSCALE);
-    ASSERT_FALSE(img.empty());
-
-    // define HOG related arguments
-    float scale = 1.05f;
-    //int nlevels = 13;
-    int gr_threshold = 8;
-    float hit_threshold = 1.4f;
-    //bool hit_threshold_auto = true;
-
-    int win_width = is48 ? 48 : 64;
-    int win_stride_width = 8;
-    int win_stride_height = 8;
-
-    bool gamma_corr = true;
-
-    Size win_size(win_width, win_width * 2); //(64, 128) or (48, 96)
-    Size win_stride(win_stride_width, win_stride_height);
-
-    cv::ocl::HOGDescriptor gpu_hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9,
-                                   cv::ocl::HOGDescriptor::DEFAULT_WIN_SIGMA, 0.2, gamma_corr,
-                                   cv::ocl::HOGDescriptor::DEFAULT_NLEVELS);
-
-    gpu_hog.setSVMDetector(detector);
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_OCL_IMPL)
     {
-        t1 = (double)cvGetTickCount();//gpu start1
+        ocl::HOGDescriptor ocl_hog;
+        ocl_hog.setSVMDetector(ocl_hog.getDefaultPeopleDetector());
+        ocl::oclMat oclSrc(src);
 
-        ocl::oclMat d_src(img);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-
-        vector<Rect> found;
-        gpu_hog.detectMultiScale(d_src, found, hit_threshold, win_stride,
-                                 Size(0, 0), scale, gr_threshold);
-
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        // no download time for HOG
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
+        OCL_TEST_CYCLE() ocl_hog.detectMultiScale(oclSrc, found_locations);
 
+        SANITY_CHECK(found_locations, 1 + DBL_EPSILON);
     }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+    else
+        OCL_PERF_ELSE
 }
-
-
-INSTANTIATE_TEST_CASE_P(GPU_ObjDetect, HOG, testing::Combine(testing::Values(WinSizw48(false), WinSizw48(true)), testing::Values(false)));
-
-#endif  //Have opencl
\ No newline at end of file
diff --git a/modules/ocl/perf/perf_hough.cpp b/modules/ocl/perf/perf_hough.cpp
index d0d130e922..61e8e48d2f 100644
--- a/modules/ocl/perf/perf_hough.cpp
+++ b/modules/ocl/perf/perf_hough.cpp
@@ -40,7 +40,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "perf_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -80,18 +80,18 @@ PERF_TEST_P(Size_Dp_MinDist, OCL_HoughCircles,
 
         cv::circle(src, center, radius, cv::Scalar::all(255), -1);
     }
-    
+
     cv::ocl::oclMat ocl_src(src);
     cv::ocl::oclMat ocl_circles;
 
     declare.time(10.0).iterations(25);
-    
+
     TEST_CYCLE()
     {
-        cv::ocl::HoughCircles(ocl_src, ocl_circles, CV_HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+        cv::ocl::HoughCircles(ocl_src, ocl_circles, HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
     }
-    
-    cv::Mat circles(ocl_circles);    
+
+    cv::Mat circles(ocl_circles);
     SANITY_CHECK(circles);
 }
 
diff --git a/modules/ocl/perf/perf_imgproc.cpp b/modules/ocl/perf/perf_imgproc.cpp
index bc54cb275c..cb1f8efa40 100644
--- a/modules/ocl/perf/perf_imgproc.cpp
+++ b/modules/ocl/perf/perf_imgproc.cpp
@@ -10,18 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Niko Li, newlife20080214@gmail.com
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//    Shengen Yan, yanshengen@gmail.com
-//    Jiang Liyuan, lyuan001.good@163.com
-//    Rock Li, Rock.Li@amd.com
-//    Zailong Wu, bullet@yeah.net
-//    Xu Pang, pangxu010@163.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -36,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -48,2104 +43,820 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
-#ifdef HAVE_OPENCL
+///////////// equalizeHist ////////////////////////
 
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+typedef TestBaseWithParam<Size> equalizeHistFixture;
 
-
-MatType nulltype = -1;
-
-#define ONE_TYPE(type)  testing::ValuesIn(typeVector(type))
-#define NULL_TYPE  testing::ValuesIn(typeVector(nulltype))
-
-
-vector<MatType> typeVector(MatType type)
+PERF_TEST_P(equalizeHistFixture, equalizeHist, OCL_TYPICAL_MAT_SIZES)
 {
-    vector<MatType> v;
-    v.push_back(type);
-    return v;
-}
+    const Size srcSize = GetParam();
 
+    Mat src(srcSize, CV_8UC1), dst(srcSize, CV_8UC1);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-PARAM_TEST_CASE(ImgprocTestBase, MatType, MatType, MatType, MatType, MatType, bool)
-{
-    int type1, type2, type3, type4, type5;
-    cv::Scalar val;
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int dstx;
-    int dsty;
-    int dst1x;
-    int dst1y;
-    int maskx;
-    int masky;
-
-    //mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mask;
-    cv::Mat dst;
-    cv::Mat dst1; //bak, for two outputs
-
-    //mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl mat
-    cv::ocl::oclMat clmat1;
-    cv::ocl::oclMat clmat2;
-    cv::ocl::oclMat clmask;
-    cv::ocl::oclMat cldst;
-    cv::ocl::oclMat cldst1; //bak
-
-    //ocl mat with roi
-    cv::ocl::oclMat clmat1_roi;
-    cv::ocl::oclMat clmat2_roi;
-    cv::ocl::oclMat clmask_roi;
-    cv::ocl::oclMat cldst_roi;
-    cv::ocl::oclMat cldst1_roi;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type1 = GET_PARAM(0);
-        type2 = GET_PARAM(1);
-        type3 = GET_PARAM(2);
-        type4 = GET_PARAM(3);
-        type5 = GET_PARAM(4);
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-        double min = 1, max = 20;
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums>0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-        if(type1 != nulltype)
-        {
-            mat1 = randomMat(rng, size, type1, min, max, false);
-            clmat1 = mat1;
-        }
-        if(type2 != nulltype)
-        {
-            mat2 = randomMat(rng, size, type2, min, max, false);
-            clmat2 = mat2;
-        }
-        if(type3 != nulltype)
-        {
-            dst  = randomMat(rng, size, type3, min, max, false);
-            cldst = dst;
-        }
-        if(type4 != nulltype)
-        {
-            dst1 = randomMat(rng, size, type4, min, max, false);
-            cldst1 = dst1;
-        }
-        if(type5 != nulltype)
-        {
-            mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
-            cv::threshold(mask, mask, 0.5, 255., type5);
-            clmask = mask;
-        }
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
+        ocl::oclMat oclSrc(src), oclDst(srcSize, src.type());
+
+        OCL_TEST_CYCLE() cv::ocl::equalizeHist(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
     }
-
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat1.cols - 1; //start
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src2x   = 1;
-            src1y   = 1;
-            src2y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-            dst1x    = 1;
-            dst1y    = 1;
-            maskx	 = 1;
-            masky	= 1;
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src2x = 0;
-            src1y = 0;
-            src2y = 0;
-            dstx = 0;
-            dsty = 0;
-            dst1x  = 0;
-            dst1y  = 0;
-            maskx	 = 0;
-            masky	= 0;
-        };
+        TEST_CYCLE() cv::equalizeHist(src, dst);
 
-        if(type1 != nulltype)
-        {
-            mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-            //clmat1_roi = clmat1(Rect(src1x,src1y,roicols,roirows));
-        }
-        if(type2 != nulltype)
-        {
-            mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-            //clmat2_roi = clmat2(Rect(src2x,src2y,roicols,roirows));
-        }
-        if(type3 != nulltype)
-        {
-            dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-            //cldst_roi = cldst(Rect(dstx,dsty,roicols,roirows));
-        }
-        if(type4 != nulltype)
-        {
-            dst1_roi = dst1(Rect(dst1x, dst1y, roicols, roirows));
-            //cldst1_roi = cldst1(Rect(dst1x,dst1y,roicols,roirows));
-        }
-        if(type5 != nulltype)
-        {
-            mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-            //clmask_roi = clmask(Rect(maskx,masky,roicols,roirows));
-        }
-    }
-
-    void random_roi()
-    {
-        cv::RNG &rng = TS::ptr()->get_rng();
-
-        //randomize ROI
-        roicols = rng.uniform(1, mat1.cols);
-        roirows = rng.uniform(1, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        src2x   = rng.uniform(0, mat2.cols - roicols);
-        src2y   = rng.uniform(0, mat2.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-        dst1x    = rng.uniform(0, dst1.cols  - roicols);
-        dst1y    = rng.uniform(0, dst1.rows  - roirows);
-        maskx   = rng.uniform(0, mask.cols - roicols);
-        masky   = rng.uniform(0, mask.rows - roirows);
-
-        if(type1 != nulltype)
-        {
-            mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-            //clmat1_roi = clmat1(Rect(src1x,src1y,roicols,roirows));
-        }
-        if(type2 != nulltype)
-        {
-            mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-            //clmat2_roi = clmat2(Rect(src2x,src2y,roicols,roirows));
-        }
-        if(type3 != nulltype)
-        {
-            dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-            //cldst_roi = cldst(Rect(dstx,dsty,roicols,roirows));
-        }
-        if(type4 != nulltype)
-        {
-            dst1_roi = dst1(Rect(dst1x, dst1y, roicols, roirows));
-            //cldst1_roi = cldst1(Rect(dst1x,dst1y,roicols,roirows));
-        }
-        if(type5 != nulltype)
-        {
-            mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-            //clmask_roi = clmask(Rect(maskx,masky,roicols,roirows));
-        }
-    }
-};
-////////////////////////////////equalizeHist//////////////////////////////////////////
-
-struct equalizeHist : ImgprocTestBase {};
-
-TEST_P(equalizeHist, MatType)
-{
-    if (mat1.type() != CV_8UC1 || mat1.type() != dst.type())
-    {
-        cout << "Unsupported type" << endl;
-        EXPECT_DOUBLE_EQ(0.0, 0.0);
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
     }
     else
-    {
-#ifndef PRINT_KERNEL_RUN_TIME
-        double totalcputick = 0;
-        double totalgputick = 0;
-        double totalgputick_kernel = 0;
-        double t0 = 0;
-        double t1 = 0;
-        double t2 = 0;
-        for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-        {
-            totalcputick = 0;
-            totalgputick = 0;
-            totalgputick_kernel = 0;
-            for(int j = 0; j < LOOP_TIMES + 1; j ++)
-            {
-                Has_roi(k);
-
-                t0 = (double)cvGetTickCount();//cpu start
-                cv::equalizeHist(mat1_roi, dst_roi);
-                t0 = (double)cvGetTickCount() - t0;//cpu end
-
-                t1 = (double)cvGetTickCount();//gpu start1
-                if(type1 != nulltype)
-                {
-                    clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-                }
-                cldst_roi = cldst(Rect(dstx, dsty, roicols, roirows));
-                t2 = (double)cvGetTickCount(); //kernel
-                cv::ocl::equalizeHist(clmat1_roi, cldst_roi);
-                t2 = (double)cvGetTickCount() - t2;//kernel
-                cv::Mat cpu_cldst;
-                //cldst.download(cpu_cldst);//download
-                t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-                if(j == 0)
-                    continue;
-
-                totalgputick = t1 + totalgputick;
-                totalcputick = t0 + totalcputick;
-                totalgputick_kernel = t2 + totalgputick_kernel;
-
-            }
-            if(k == 0)
-            {
-                cout << "no roi\n";
-            }
-            else
-            {
-                cout << "with roi\n";
-            };
-            cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        }
-#else
-        for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-        {
-            Has_roi(j);
-            if(type1 != nulltype)
-            {
-                clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-            }
-            if(j == 0)
-            {
-                cout << "no roi:";
-            }
-            else
-            {
-                cout << "\nwith roi:";
-            };
-            cv::ocl::equalizeHist(clmat1_roi, cldst_roi);
-        };
-#endif
-    }
+        OCL_PERF_ELSE
 }
 
+/////////// CopyMakeBorder //////////////////////
 
-////////////////////////////////bilateralFilter////////////////////////////////////////////
+typedef Size_MatType CopyMakeBorderFixture;
 
-struct bilateralFilter : ImgprocTestBase {};
-
-TEST_P(bilateralFilter, Mat)
+PERF_TEST_P(CopyMakeBorderFixture, CopyMakeBorder,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    double sigmacolor = 50.0;
-    int radius = 9;
-    int d = 2 * radius + 1;
-    double sigmaspace = 20.0;
-    int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE/*,cv::BORDER_REFLECT,cv::BORDER_WRAP,cv::BORDER_REFLECT_101*/};
-    const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE"/*, "BORDER_REFLECT","BORDER_WRAP","BORDER_REFLECT_101"*/};
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), borderType = BORDER_CONSTANT;
 
-    if (mat1.depth() != CV_8U || mat1.type() != dst.type())
+    Mat src(srcSize, type), dst;
+    const Size dstSize = srcSize + Size(12, 12);
+    dst.create(dstSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        cout << "Unsupported type" << endl;
-        EXPECT_DOUBLE_EQ(0.0, 0.0);
+        ocl::oclMat oclSrc(src), oclDst(dstSize, type);
+
+        OCL_TEST_CYCLE() cv::ocl::copyMakeBorder(oclSrc, oclDst, 7, 5, 5, 7, borderType, cv::Scalar(1.0));
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::copyMakeBorder(src, dst, 7, 5, 5, 7, borderType, cv::Scalar(1.0));
+
+        SANITY_CHECK(dst);
     }
     else
-    {
-        for(size_t i = 0; i < sizeof(bordertype) / sizeof(int); i++)
-        {
-            cout << borderstr[i] << endl;
-#ifndef PRINT_KERNEL_RUN_TIME
-            double totalcputick = 0;
-            double totalgputick = 0;
-            double totalgputick_kernel = 0;
-            double t0 = 0;
-            double t1 = 0;
-            double t2 = 0;
-            for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-            {
-                totalcputick = 0;
-                totalgputick = 0;
-                totalgputick_kernel = 0;
-                for(int j = 0; j < LOOP_TIMES + 1; j ++)
-                {
-                    Has_roi(k);
-                    if(((bordertype[i] != cv::BORDER_CONSTANT) && (bordertype[i] != cv::BORDER_REPLICATE) && (mat1_roi.cols <= radius)) || (mat1_roi.cols <= radius) || (mat1_roi.rows <= radius) || (mat1_roi.rows <= radius))
-                    {
-                        continue;
-                    }
-                    t0 = (double)cvGetTickCount();//cpu start
-                    cv::bilateralFilter(mat1_roi, dst_roi, d, sigmacolor, sigmaspace, bordertype[i]);
-                    t0 = (double)cvGetTickCount() - t0;//cpu end
-
-                    t1 = (double)cvGetTickCount();//gpu start1
-                    if(type1 != nulltype)
-                    {
-                        clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-                    }
-                    t2 = (double)cvGetTickCount(); //kernel
-                    cv::ocl::bilateralFilter(clmat1_roi, cldst_roi, d, sigmacolor, sigmaspace, bordertype[i]);
-                    t2 = (double)cvGetTickCount() - t2;//kernel
-                    cv::Mat cpu_cldst;
-                    cldst.download(cpu_cldst);//download
-                    t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-                    if(j == 0)
-                        continue;
-
-                    totalgputick = t1 + totalgputick;
-                    totalcputick = t0 + totalcputick;
-                    totalgputick_kernel = t2 + totalgputick_kernel;
-
-                }
-                if(k == 0)
-                {
-                    cout << "no roi\n";
-                }
-                else
-                {
-                    cout << "with roi\n";
-                };
-                cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-                cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-                cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            }
-
-#else
-            for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-            {
-                Has_roi(j);
-                if(type1 != nulltype)
-                {
-                    clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-                };
-                if(j == 0)
-                {
-                    cout << "no roi:";
-                }
-                else
-                {
-                    cout << "\nwith roi:";
-                };
-                cv::ocl::bilateralFilter(clmat1_roi, cldst_roi, d, sigmacolor, sigmaspace, bordertype[i]);
-            };
-
-#endif
-        };
-
-    }
+        OCL_PERF_ELSE
 }
 
-////////////////////////////////copyMakeBorder////////////////////////////////////////////
+///////////// cornerMinEigenVal ////////////////////////
 
-struct CopyMakeBorder : ImgprocTestBase {};
+typedef Size_MatType cornerMinEigenValFixture;
 
-TEST_P(CopyMakeBorder, Mat)
+PERF_TEST_P(cornerMinEigenValFixture, cornerMinEigenVal,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-    int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE, cv::BORDER_REFLECT, cv::BORDER_WRAP, cv::BORDER_REFLECT_101};
-    //const char* borderstr[]={"BORDER_CONSTANT", "BORDER_REPLICATE"/*, "BORDER_REFLECT","BORDER_WRAP","BORDER_REFLECT_101"*/};
-    int top = 5;
-    int bottom = 5;
-    int left = 6;
-    int right = 6;
-    if (mat1.type() != dst.type())
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), borderType = BORDER_REFLECT;
+    const int blockSize = 7, apertureSize = 1 + 2 * 3;
+
+    Mat src(srcSize, type), dst(srcSize, CV_32FC1);
+    declare.in(src, WARMUP_RNG).out(dst)
+            .time(srcSize == OCL_SIZE_4000 ? 20 : srcSize == OCL_SIZE_2000 ? 5 : 3);
+
+    const int depth = CV_MAT_DEPTH(type);
+    const ERROR_TYPE errorType = depth == CV_8U ? ERROR_ABSOLUTE : ERROR_RELATIVE;
+
+    if (RUN_OCL_IMPL)
     {
-        cout << "Unsupported type" << endl;
-        EXPECT_DOUBLE_EQ(0.0, 0.0);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, CV_32FC1);
+
+        OCL_TEST_CYCLE() cv::ocl::cornerMinEigenVal(oclSrc, oclDst, blockSize, apertureSize, borderType);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1e-6, errorType);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::cornerMinEigenVal(src, dst, blockSize, apertureSize, borderType);
+
+        SANITY_CHECK(dst, 1e-6, errorType);
     }
     else
-    {
-        for(size_t i = 0; i < sizeof(bordertype) / sizeof(int); i++)
-        {
-#ifndef PRINT_KERNEL_RUN_TIME
-            double totalcputick = 0;
-            double totalgputick = 0;
-            double totalgputick_kernel = 0;
-            double t0 = 0;
-            double t1 = 0;
-            double t2 = 0;
-            for(int k = LOOPROISTART; k < 1; k++) //don't support roi perf test
-            {
-                totalcputick = 0;
-                totalgputick = 0;
-                totalgputick_kernel = 0;
-                for(int j = 0; j < LOOP_TIMES + 1; j ++)
-                {
-                    Has_roi(k);
-
-                    t0 = (double)cvGetTickCount();//cpu start
-                    cv::copyMakeBorder(mat1_roi, dst_roi, top, bottom, left, right, bordertype[i] | cv::BORDER_ISOLATED, cv::Scalar(1.0));
-                    t0 = (double)cvGetTickCount() - t0;//cpu end
-
-                    t1 = (double)cvGetTickCount();//gpu start1
-                    if(type1 != nulltype)
-                    {
-                        clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-                    }
-                    t2 = (double)cvGetTickCount(); //kernel
-                    cv::ocl::copyMakeBorder(clmat1_roi, cldst_roi, top, bottom, left, right,  bordertype[i] | cv::BORDER_ISOLATED, cv::Scalar(1.0));
-                    t2 = (double)cvGetTickCount() - t2;//kernel
-                    cv::Mat cpu_cldst;
-                    cldst.download(cpu_cldst);//download
-                    t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-                    if(j == 0)
-                        continue;
-
-                    totalgputick = t1 + totalgputick;
-                    totalcputick = t0 + totalcputick;
-                    totalgputick_kernel = t2 + totalgputick_kernel;
-
-                }
-                if(k == 0)
-                {
-                    cout << "no roi\n";
-                }
-                else
-                {
-                    cout << "with roi\n";
-                };
-                cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-                cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-                cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-            }
-#else
-            for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-            {
-                Has_roi(j);
-                if(type1 != nulltype)
-                {
-                    clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-                };
-                if(j == 0)
-                {
-                    cout << "no roi:";
-                }
-                else
-                {
-                    cout << "\nwith roi:";
-                };
-                cv::ocl::copyMakeBorder(clmat1_roi, cldst_roi, top, bottom, left, right,  bordertype[i] | cv::BORDER_ISOLATED, cv::Scalar(1.0));
-            };
-#endif
-        };
-    }
+        OCL_PERF_ELSE
 }
 
-////////////////////////////////cornerMinEigenVal//////////////////////////////////////////
+///////////// cornerHarris ////////////////////////
 
-struct cornerMinEigenVal : ImgprocTestBase {};
+typedef Size_MatType cornerHarrisFixture;
 
-TEST_P(cornerMinEigenVal, Mat)
+PERF_TEST_P(cornerHarrisFixture, cornerHarris,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_32FC1)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), borderType = BORDER_REFLECT;
+
+    Mat src(srcSize, type), dst(srcSize, CV_32FC1);
+    randu(src, 0, 1);
+    declare.in(src).out(dst)
+            .time(srcSize == OCL_SIZE_4000 ? 20 : srcSize == OCL_SIZE_2000 ? 5 : 3);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            int blockSize = 7, apertureSize = 3; //1 + 2 * (rand() % 4);
-            int borderType = cv::BORDER_REFLECT;
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::cornerMinEigenVal(mat1_roi, dst_roi, blockSize, apertureSize, borderType);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        ocl::oclMat oclSrc(src), oclDst(srcSize, CV_32FC1);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            if(type1 != nulltype)
-            {
-                clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-            }
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::cornerMinEigenVal(clmat1_roi, cldst_roi, blockSize, apertureSize, borderType);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_cldst;
-            cldst.download(cpu_cldst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
+        OCL_TEST_CYCLE() cv::ocl::cornerHarris(oclSrc, oclDst, 5, 7, 0.1, borderType);
 
-            if(j == 0)
-                continue;
+        oclDst.download(dst);
 
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst, 3e-5);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        int blockSize = 7, apertureSize = 1 + 2 * (rand() % 4);
-        int borderType = cv::BORDER_REFLECT;
-        if(type1 != nulltype)
-        {
-            clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-        };
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::cornerMinEigenVal(clmat1_roi, cldst_roi, blockSize, apertureSize, borderType);
-    };
-#endif
+        TEST_CYCLE() cv::cornerHarris(src, dst, 5, 7, 0.1, borderType);
+
+        SANITY_CHECK(dst, 3e-5);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// integral ////////////////////////
 
-////////////////////////////////cornerHarris//////////////////////////////////////////
+typedef TestBaseWithParam<Size> integralFixture;
 
-struct cornerHarris : ImgprocTestBase {};
-
-TEST_P(cornerHarris, Mat)
+PERF_TEST_P(integralFixture, DISABLED_integral, OCL_TYPICAL_MAT_SIZES) // TODO does not work properly
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size srcSize = GetParam();
+
+    Mat src(srcSize, CV_8UC1), dst;
+    declare.in(src, WARMUP_RNG);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            int blockSize = 7, apertureSize = 3;
-            int borderType = cv::BORDER_REFLECT;
-            double kk = 2;
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::cornerHarris(mat1_roi, dst_roi, blockSize, apertureSize, kk, borderType);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        ocl::oclMat oclSrc(src), oclDst;
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            if(type1 != nulltype)
-            {
-                clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-            }
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::cornerHarris(clmat1_roi, cldst_roi, blockSize, apertureSize, kk, borderType);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_cldst;
-            cldst.download(cpu_cldst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
+        OCL_TEST_CYCLE() cv::ocl::integral(oclSrc, oclDst);
 
-            if(j == 0)
-                continue;
+        oclDst.download(dst);
 
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        double kk = 2;
-        int blockSize = 7, apertureSize = 3;
-        int borderType = cv::BORDER_REFLECT;
-        if(type1 != nulltype)
-        {
-            clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-        };
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::cornerHarris(clmat1_roi, cldst_roi, blockSize, apertureSize, kk, borderType);
-    };
-#endif
+        TEST_CYCLE() cv::integral(src, dst);
 
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// WarpAffine ////////////////////////
 
-////////////////////////////////integral/////////////////////////////////////////////////
+typedef Size_MatType WarpAffineFixture;
 
-struct integral : ImgprocTestBase {};
-
-TEST_P(integral, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::integral(mat1_roi, dst_roi, dst1_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            if(type1 != nulltype)
-            {
-                clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-            }
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::integral(clmat1_roi, cldst_roi, cldst1_roi);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_cldst;
-            cv::Mat cpu_cldst1;
-            cldst.download(cpu_cldst);//download
-            cldst1.download(cpu_cldst1);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        if(type1 != nulltype)
-        {
-            clmat1_roi = clmat1(Rect(src1x, src1y, roicols, roirows));
-        };
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::integral(clmat1_roi, cldst_roi, cldst1_roi);
-    };
-#endif
-}
-
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// warpAffine  & warpPerspective
-
-PARAM_TEST_CASE(WarpTestBase, MatType, int)
-{
-    int type;
-    cv::Size size;
-    int interpolation;
-
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int src_roicols;
-    int src_roirows;
-    int dst_roicols;
-    int dst_roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        //dsize = GET_PARAM(1);
-        interpolation = GET_PARAM(1);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        size = cv::Size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-    void Has_roi(int b)
-    {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            src_roicols =  mat1.cols - 1; //start
-            src_roirows = mat1.rows - 1;
-            dst_roicols = dst.cols - 1;
-            dst_roirows = dst.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-
-        }
-        else
-        {
-            src_roicols = mat1.cols;
-            src_roirows = mat1.rows;
-            dst_roicols = dst.cols;
-            dst_roirows = dst.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-
-        };
-        mat1_roi = mat1(Rect(src1x, src1y, src_roicols, src_roirows));
-        dst_roi  = dst(Rect(dstx, dsty, dst_roicols, dst_roirows));
-
-
-    }
-
-};
-
-/////warpAffine
-
-struct WarpAffine : WarpTestBase {};
-
-TEST_P(WarpAffine, Mat)
+PERF_TEST_P(WarpAffineFixture, WarpAffine,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
     static const double coeffs[2][3] =
     {
-        {cos(3.14 / 6), -sin(3.14 / 6), 100.0},
-        {sin(3.14 / 6), cos(3.14 / 6), -100.0}
+        { cos(CV_PI / 6), -sin(CV_PI / 6), 100.0 },
+        { sin(CV_PI / 6), cos(CV_PI / 6), -100.0 }
     };
     Mat M(2, 3, CV_64F, (void *)coeffs);
+    const int interpolation = INTER_NEAREST;
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::warpAffine(mat1_roi, dst_roi, M, size, interpolation);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::warpAffine(oclSrc, oclDst, M, srcSize, interpolation);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::warpAffine(gmat1, gdst, M, size, interpolation);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::warpAffine(gmat1, gdst, M, size, interpolation);
-    };
-#endif
+        TEST_CYCLE() cv::warpAffine(src, dst, M, srcSize, interpolation);
 
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// WarpPerspective ////////////////////////
 
-// warpPerspective
+typedef Size_MatType WarpPerspectiveFixture;
 
-struct WarpPerspective : WarpTestBase {};
-
-TEST_P(WarpPerspective, Mat)
+PERF_TEST_P(WarpPerspectiveFixture, WarpPerspective,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
     static const double coeffs[3][3] =
     {
-        {cos(3.14 / 6), -sin(3.14 / 6), 100.0},
-        {sin(3.14 / 6), cos(3.14 / 6), -100.0},
+        {cos(CV_PI / 6), -sin(CV_PI / 6), 100.0},
+        {sin(CV_PI / 6), cos(CV_PI / 6), -100.0},
         {0.0, 0.0, 1.0}
     };
     Mat M(3, 3, CV_64F, (void *)coeffs);
+    const int interpolation = INTER_LINEAR;
 
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst)
+            .time(srcSize == OCL_SIZE_4000 ? 18 : srcSize == OCL_SIZE_2000 ? 5 : 2);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::warpPerspective(mat1_roi, dst_roi, M, size, interpolation);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() cv::ocl::warpPerspective(oclSrc, oclDst, M, srcSize, interpolation);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
+        oclDst.download(dst);
 
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::warpPerspective(gmat1, gdst, M, size, interpolation);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::warpPerspective(gmat1, gdst, M, size, interpolation);
-    };
-#endif
+        TEST_CYCLE() cv::warpPerspective(src, dst, M, srcSize, interpolation);
 
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// remap
-//////////////////////////////////////////////////////////////////////////////////////////////////
+///////////// resize ////////////////////////
 
-PARAM_TEST_CASE(Remap, MatType, MatType, MatType, int, int)
+CV_ENUM(resizeInterType, INTER_NEAREST, INTER_LINEAR)
+
+typedef tuple<Size, MatType, resizeInterType, double> resizeParams;
+typedef TestBaseWithParam<resizeParams> resizeFixture;
+
+PERF_TEST_P(resizeFixture, resize,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4),
+                               resizeInterType::all(),
+                               ::testing::Values(0.5, 2.0)))
 {
-    int srcType;
-    int map1Type;
-    int map2Type;
-    cv::Scalar val;
+    const resizeParams params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), interType = get<2>(params);
+    double scale = get<3>(params);
 
-    int interpolation;
-    int bordertype;
+    Mat src(srcSize, type), dst;
+    const Size dstSize(cvRound(srcSize.width * scale), cvRound(srcSize.height * scale));
+    dst.create(dstSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+    if (interType == INTER_LINEAR && type == CV_8UC4 && OCL_SIZE_4000 == srcSize)
+        declare.time(11);
 
-    cv::Mat src;
-    cv::Mat dst;
-    cv::Mat map1;
-    cv::Mat map2;
-
-
-    int src_roicols;
-    int src_roirows;
-    int dst_roicols;
-    int dst_roirows;
-    int map1_roicols;
-    int map1_roirows;
-    int map2_roicols;
-    int map2_roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
-    int map1x;
-    int map1y;
-    int map2x;
-    int map2y;
-
-    cv::Mat src_roi;
-    cv::Mat dst_roi;
-    cv::Mat map1_roi;
-    cv::Mat map2_roi;
-
-    //ocl mat for testing
-    cv::ocl::oclMat gdst;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gsrc_roi;
-    cv::ocl::oclMat gdst_roi;
-    cv::ocl::oclMat gmap1_roi;
-    cv::ocl::oclMat gmap2_roi;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        srcType = GET_PARAM(0);
-        map1Type = GET_PARAM(1);
-        map2Type = GET_PARAM(2);
-        interpolation = GET_PARAM(3);
-        bordertype = GET_PARAM(4);
+        ocl::oclMat oclSrc(src), oclDst(dstSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size srcSize = cv::Size(MWIDTH, MHEIGHT);
-        cv::Size map1Size = cv::Size(MWIDTH, MHEIGHT);
-        double min = 5, max = 16;
+        OCL_TEST_CYCLE() cv::ocl::resize(oclSrc, oclDst, Size(), scale, scale, interType);
 
-        if(srcType != nulltype)
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::resize(src, dst, Size(), scale, scale, interType);
+
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+///////////// threshold////////////////////////
+
+CV_ENUM(ThreshType, THRESH_BINARY, THRESH_TRUNC)
+
+typedef tuple<Size, ThreshType> ThreshParams;
+typedef TestBaseWithParam<ThreshParams> ThreshFixture;
+
+PERF_TEST_P(ThreshFixture, threshold,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               ThreshType::all()))
+{
+    const ThreshParams params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int threshType = get<1>(params);
+
+    Mat src(srcSize, CV_8U), dst(srcSize, CV_8U);
+    randu(src, 0, 100);
+    declare.in(src).out(dst);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(srcSize, CV_8U);
+
+        OCL_TEST_CYCLE() cv::ocl::threshold(oclSrc, oclDst, 50.0, 0.0, threshType);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::threshold(src, dst, 50.0, 0.0, threshType);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+///////////// meanShiftFiltering////////////////////////
+
+typedef struct _COOR
+{
+    short x;
+    short y;
+} COOR;
+
+static COOR do_meanShift(int x0, int y0, uchar *sptr, uchar *dptr, int sstep, cv::Size size, int sp, int sr, int maxIter, float eps, int *tab)
+{
+
+    int isr2 = sr * sr;
+    int c0, c1, c2, c3;
+    int iter;
+    uchar *ptr = NULL;
+    uchar *pstart = NULL;
+    int revx = 0, revy = 0;
+    c0 = sptr[0];
+    c1 = sptr[1];
+    c2 = sptr[2];
+    c3 = sptr[3];
+    // iterate meanshift procedure
+    for(iter = 0; iter < maxIter; iter++ )
+    {
+        int count = 0;
+        int s0 = 0, s1 = 0, s2 = 0, sx = 0, sy = 0;
+
+        //mean shift: process pixels in window (p-sigmaSp)x(p+sigmaSp)
+        int minx = x0 - sp;
+        int miny = y0 - sp;
+        int maxx = x0 + sp;
+        int maxy = y0 + sp;
+
+        //deal with the image boundary
+        if(minx < 0) minx = 0;
+        if(miny < 0) miny = 0;
+        if(maxx >= size.width) maxx = size.width - 1;
+        if(maxy >= size.height) maxy = size.height - 1;
+        if(iter == 0)
         {
-            src = randomMat(rng, srcSize, srcType, min, max, false);
+            pstart = sptr;
         }
-        if((map1Type == CV_16SC2 && map2Type == nulltype) || (map1Type == CV_32FC2 && map2Type == nulltype))
-        {
-            map1 = randomMat(rng, map1Size, map1Type, min, max, false);
-
-        }
-        else if (map1Type == CV_32FC1 && map2Type == CV_32FC1)
-        {
-            map1 = randomMat(rng, map1Size, map1Type, min, max, false);
-            map2 = randomMat(rng, map1Size, map1Type, min, max, false);
-        }
-
         else
-            cout << "The wrong input type" << endl;
-
-        dst = randomMat(rng, map1Size, srcType, min, max, false);
-        switch (src.channels())
         {
-        case 1:
-            val = cv::Scalar(rng.uniform(0.0, 10.0), 0, 0, 0);
+            pstart = pstart + revy * sstep + (revx << 2); //point to the new position
+        }
+        ptr = pstart;
+        ptr = ptr + (miny - y0) * sstep + ((minx - x0) << 2); //point to the start in the row
+
+        for( int y = miny; y <= maxy; y++, ptr += sstep - ((maxx - minx + 1) << 2))
+        {
+            int rowCount = 0;
+            int x = minx;
+#if CV_ENABLE_UNROLLED
+            for( ; x + 4 <= maxx; x += 4, ptr += 16)
+            {
+                int t0, t1, t2;
+                t0 = ptr[0], t1 = ptr[1], t2 = ptr[2];
+                if(tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
+                {
+                    s0 += t0;
+                    s1 += t1;
+                    s2 += t2;
+                    sx += x;
+                    rowCount++;
+                }
+                t0 = ptr[4], t1 = ptr[5], t2 = ptr[6];
+                if(tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
+                {
+                    s0 += t0;
+                    s1 += t1;
+                    s2 += t2;
+                    sx += x + 1;
+                    rowCount++;
+                }
+                t0 = ptr[8], t1 = ptr[9], t2 = ptr[10];
+                if(tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
+                {
+                    s0 += t0;
+                    s1 += t1;
+                    s2 += t2;
+                    sx += x + 2;
+                    rowCount++;
+                }
+                t0 = ptr[12], t1 = ptr[13], t2 = ptr[14];
+                if(tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
+                {
+                    s0 += t0;
+                    s1 += t1;
+                    s2 += t2;
+                    sx += x + 3;
+                    rowCount++;
+                }
+            }
+#endif
+            for(; x <= maxx; x++, ptr += 4)
+            {
+                int t0 = ptr[0], t1 = ptr[1], t2 = ptr[2];
+                if(tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
+                {
+                    s0 += t0;
+                    s1 += t1;
+                    s2 += t2;
+                    sx += x;
+                    rowCount++;
+                }
+            }
+            if(rowCount == 0)
+                continue;
+            count += rowCount;
+            sy += y * rowCount;
+        }
+
+        if( count == 0 )
             break;
-        case 2:
-            val = cv::Scalar(rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0), 0, 0);
+
+        int x1 = sx / count;
+        int y1 = sy / count;
+        s0 = s0 / count;
+        s1 = s1 / count;
+        s2 = s2 / count;
+
+        bool stopFlag = (x0 == x1 && y0 == y1) || (abs(x1 - x0) + abs(y1 - y0) +
+            tab[s0 - c0 + 255] + tab[s1 - c1 + 255] + tab[s2 - c2 + 255] <= eps);
+
+        //revise the pointer corresponding to the new (y0,x0)
+        revx = x1 - x0;
+        revy = y1 - y0;
+
+        x0 = x1;
+        y0 = y1;
+        c0 = s0;
+        c1 = s1;
+        c2 = s2;
+
+        if( stopFlag )
             break;
-        case 3:
-            val = cv::Scalar(rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0), 0);
-            break;
-        case 4:
-            val = cv::Scalar(rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0), rng.uniform(0.0, 10.0));
-            break;
-        }
+    } //for iter
 
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        //if you want to use undefault device, set it here
-        //setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-    void Has_roi(int b)
-    {
-        if(b)
-        {
-            //randomize ROI
-            dst_roicols = dst.cols - 1;
-            dst_roirows = dst.rows - 1;
-
-            src_roicols = src.cols - 1;
-            src_roirows = src.rows - 1;
-
-
-            srcx = 1;
-            srcy = 1;
-            dstx = 1;
-            dsty = 1;
-        }
-        else
-        {
-            dst_roicols = dst.cols;
-            dst_roirows = dst.rows;
-
-            src_roicols = src.cols;
-            src_roirows = src.rows;
-
-
-            srcx = 0;
-            srcy = 0;
-            dstx = 0;
-            dsty = 0;
-        }
-        map1_roicols = dst_roicols;
-        map1_roirows = dst_roirows;
-        map2_roicols = dst_roicols;
-        map2_roirows = dst_roirows;
-        map1x = dstx;
-        map1y = dsty;
-        map2x = dstx;
-        map2y = dsty;
-
-        if((map1Type == CV_16SC2 && map2Type == nulltype) || (map1Type == CV_32FC2 && map2Type == nulltype))
-        {
-            map1_roi = map1(Rect(map1x, map1y, map1_roicols, map1_roirows));
-            gmap1_roi = map1_roi;
-        }
-
-        else if (map1Type == CV_32FC1 && map2Type == CV_32FC1)
-        {
-            map1_roi = map1(Rect(map1x, map1y, map1_roicols, map1_roirows));
-            map2_roi = map2(Rect(map2x, map2y, map2_roicols, map2_roirows));
-            gmap1_roi = map1_roi;
-            gmap2_roi = map2_roi;
-        }
-        dst_roi = dst(Rect(dstx, dsty, dst_roicols, dst_roirows));
-        src_roi = dst(Rect(srcx, srcy, src_roicols, src_roirows));
-
-    }
-};
-
-TEST_P(Remap, Mat)
-{
-    if((interpolation == 1 && map1Type == CV_16SC2) || (map1Type == CV_32FC1 && map2Type == nulltype) || (map1Type == CV_16SC2 && map2Type == CV_32FC1) || (map1Type == CV_32FC2 && map2Type == CV_32FC1))
-    {
-        cout << "LINEAR don't support the map1Type and map2Type" << endl;
-        return;
-    }
-    int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE/*,BORDER_REFLECT,BORDER_WRAP,BORDER_REFLECT_101*/};
-    const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE"/*, "BORDER_REFLECT","BORDER_WRAP","BORDER_REFLECT_101"*/};
-    cout << borderstr[0] << endl;
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = 0; k < 2; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::remap(src_roi, dst_roi, map1_roi, map2_roi, interpolation, bordertype[0], val);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start
-            gsrc_roi = src_roi;
-            gdst = dst;
-            gdst_roi = gdst(Rect(dstx, dsty, dst_roicols, dst_roirows));
-
-            t2 = (double)cvGetTickCount();//kernel
-            cv::ocl::remap(gsrc_roi, gdst_roi, gmap1_roi, gmap2_roi, interpolation, bordertype[0], val);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-
-            cv::Mat cpu_dst;
-            gdst.download(cpu_dst);
-
-            t1 = (double)cvGetTickCount() - t1;//gpu end
-
-            if (j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = 0; j < 2; j ++)
-    {
-        Has_roi(j);
-        gdst = dst;
-        gdst_roi = gdst(Rect(dstx, dsty, dst_roicols, dst_roirows));
-        gsrc_roi = src_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::remap(gsrc_roi, gdst_roi, gmap1_roi, gmap2_roi, interpolation, bordertype[0], val);
-    };
-#endif
+    dptr[0] = (uchar)c0;
+    dptr[1] = (uchar)c1;
+    dptr[2] = (uchar)c2;
+    dptr[3] = (uchar)c3;
 
+    COOR coor;
+    coor.x = static_cast<short>(x0);
+    coor.y = static_cast<short>(y0);
+    return coor;
 }
 
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// resize
-
-PARAM_TEST_CASE(Resize, MatType, cv::Size, double, double, int)
+static void meanShiftFiltering_(const Mat &src_roi, Mat &dst_roi, int sp, int sr, cv::TermCriteria crit)
 {
-    int type;
-    cv::Size dsize;
-    double fx, fy;
-    int interpolation;
+    if( src_roi.empty() )
+        CV_Error( Error::StsBadArg, "The input image is empty" );
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
+    if( src_roi.depth() != CV_8U || src_roi.channels() != 4 )
+        CV_Error( Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
 
-    // set up roi
-    int src_roicols;
-    int src_roirows;
-    int dst_roicols;
-    int dst_roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
+    dst_roi.create(src_roi.size(), src_roi.type());
 
+    CV_Assert( (src_roi.cols == dst_roi.cols) && (src_roi.rows == dst_roi.rows) );
+    CV_Assert( !(dst_roi.step & 0x3) );
 
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
+    if( !(crit.type & cv::TermCriteria::MAX_ITER) )
+        crit.maxCount = 5;
+    int maxIter = std::min(std::max(crit.maxCount, 1), 100);
+    float eps;
+    if( !(crit.type & cv::TermCriteria::EPS) )
+        eps = 1.f;
+    eps = (float)std::max(crit.epsilon, 0.0);
 
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
+    int tab[512];
+    for(int i = 0; i < 512; i++)
+        tab[i] = (i - 255) * (i - 255);
+    uchar *sptr = src_roi.data;
+    uchar *dptr = dst_roi.data;
+    int sstep = (int)src_roi.step;
+    int dstep = (int)dst_roi.step;
+    cv::Size size = src_roi.size();
 
-    virtual void SetUp()
+    for(int i = 0; i < size.height; i++, sptr += sstep - (size.width << 2),
+        dptr += dstep - (size.width << 2))
     {
-        type = GET_PARAM(0);
-        dsize = GET_PARAM(1);
-        fx = GET_PARAM(2);
-        fy = GET_PARAM(3);
-        interpolation = GET_PARAM(4);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        if(dsize == cv::Size() && !(fx > 0 && fy > 0))
+        for(int j = 0; j < size.width; j++, sptr += 4, dptr += 4)
         {
-            cout << "invalid dsize and fx fy" << endl;
-            return;
+            do_meanShift(j, i, sptr, dptr, sstep, size, sp, sr, maxIter, eps, tab);
         }
-
-        if(dsize == cv::Size())
-        {
-            dsize.width = (int)(size.width * fx);
-            dsize.height = (int)(size.height * fy);
-        }
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, dsize, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-    void Has_roi(int b)
-    {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            src_roicols =  mat1.cols - 1; //start
-            src_roirows = mat1.rows - 1;
-            dst_roicols = dst.cols - 1;
-            dst_roirows = dst.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-
-        }
-        else
-        {
-            src_roicols = mat1.cols;
-            src_roirows = mat1.rows;
-            dst_roicols = dst.cols;
-            dst_roirows = dst.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-
-        };
-        mat1_roi = mat1(Rect(src1x, src1y, src_roicols, src_roirows));
-        dst_roi  = dst(Rect(dstx, dsty, dst_roicols, dst_roirows));
-
-
-    }
-
-};
-
-TEST_P(Resize, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::resize(mat1_roi, dst_roi, dsize, fx, fy, interpolation);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
-
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::resize(gmat1, gdst, dsize, fx, fy, interpolation);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, dst_roicols, dst_roirows));
-        gmat1 = mat1_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::resize(gmat1, gdst, dsize, fx, fy, interpolation);
-    };
-#endif
-
-}
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-//threshold
-
-PARAM_TEST_CASE(Threshold, MatType, ThreshOp)
-{
-    int type;
-    int threshOp;
-
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        threshOp = GET_PARAM(1);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-    void Has_roi(int b)
-    {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat1.cols - 1; //start
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            dstx    = 1;
-            dsty    = 1;
-
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x = 0;
-            src1y = 0;
-            dstx = 0;
-            dsty = 0;
-
-        };
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-
-    }
-};
-
-TEST_P(Threshold, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            double maxVal = randomDouble(20.0, 127.0);
-            double thresh = randomDouble(0.0, maxVal);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::threshold(mat1_roi, dst_roi, thresh, maxVal, threshOp);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            gmat1 = mat1_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::threshold(gmat1, gdst, thresh, maxVal, threshOp);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        double maxVal = randomDouble(20.0, 127.0);
-        double thresh = randomDouble(0.0, maxVal);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        gmat1 = mat1_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::threshold(gmat1, gdst, thresh, maxVal, threshOp);
-    };
-#endif
-
-}
-///////////////////////////////////////////////////////////////////////////////////////////////////
-//meanShift
-
-PARAM_TEST_CASE(meanShiftTestBase, MatType, MatType, int, int, cv::TermCriteria)
-{
-    int type, typeCoor;
-    int sp, sr;
-    cv::TermCriteria crit;
-    //src mat
-    cv::Mat src;
-    cv::Mat dst;
-    cv::Mat dstCoor;
-
-    //set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat src_roi;
-    cv::Mat dst_roi;
-    cv::Mat dstCoor_roi;
-
-    //ocl dst mat
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gdstCoor;
-
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl mat with roi
-    cv::ocl::oclMat gsrc_roi;
-    cv::ocl::oclMat gdst_roi;
-    cv::ocl::oclMat gdstCoor_roi;
-
-    virtual void SetUp()
-    {
-        type     = GET_PARAM(0);
-        typeCoor = GET_PARAM(1);
-        sp       = GET_PARAM(2);
-        sr       = GET_PARAM(3);
-        crit     = GET_PARAM(4);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-
-        // MWIDTH=256, MHEIGHT=256. defined in utility.hpp
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
-
-        src = randomMat(rng, size, type, 5, 16, false);
-        dst = randomMat(rng, size, type, 5, 16, false);
-        dstCoor = randomMat(rng, size, typeCoor, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath(CLBINPATH);
-    }
-
-    void Has_roi(int b)
-    {
-        if(b)
-        {
-            //randomize ROI
-            roicols = src.cols - 1;
-            roirows = src.rows - 1;
-            srcx = 1;
-            srcy = 1;
-            dstx = 1;
-            dsty = 1;
-        }
-        else
-        {
-            roicols = src.cols;
-            roirows = src.rows;
-            srcx = 0;
-            srcy = 0;
-            dstx = 0;
-            dsty = 0;
-        };
-
-        src_roi = src(Rect(srcx, srcy, roicols, roirows));
-        dst_roi = dst(Rect(dstx, dsty, roicols, roirows));
-        dstCoor_roi = dstCoor(Rect(dstx, dsty, roicols, roirows));
-
-        gdst = dst;
-        gdstCoor = dstCoor;
-    }
-};
-
-/////////////////////////meanShiftFiltering/////////////////////////////
-struct meanShiftFiltering : meanShiftTestBase {};
-
-TEST_P(meanShiftFiltering, Mat)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = 0; k < 2; k++)
-    {
-        double totalgputick = 0;
-        double totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t1 = (double)cvGetTickCount();//gpu start1
-
-            gsrc_roi = src_roi;
-            gdst_roi = gdst(Rect(dstx, dsty, roicols, roirows));  //gdst_roi
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::meanShiftFiltering(gsrc_roi, gdst_roi, sp, sr, crit);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-
-            cv::Mat cpu_gdst;
-            gdst.download(cpu_gdst);//download
-
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-
-        gsrc_roi = src_roi;
-        gdst_roi = gdst(Rect(dstx, dsty, roicols, roirows));  //gdst_roi
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::meanShiftFiltering(gsrc_roi, gdst_roi, sp, sr, crit);
-    };
-#endif
-
-}
-
-///////////////////////////meanShiftProc//////////////////////////////////
-struct meanShiftProc : meanShiftTestBase {};
-
-TEST_P(meanShiftProc, Mat)
-{
-
-#ifndef PRINT_KERNEL_RUN_TIME
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = 0; k < 2; k++)
-    {
-        double totalgputick = 0;
-        double totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t1 = (double)cvGetTickCount();//gpu start1
-
-            gsrc_roi = src_roi;
-            gdst_roi = gdst(Rect(dstx, dsty, roicols, roirows));  //gdst_roi
-            gdstCoor_roi = gdstCoor(Rect(dstx, dsty, roicols, roirows));
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::meanShiftProc(gsrc_roi, gdst_roi, gdstCoor_roi, sp, sr, crit);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-
-            cv::Mat cpu_gdstCoor;
-            gdstCoor.download(cpu_gdstCoor);//download
-
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-
-        gsrc_roi = src_roi;
-        gdst_roi = gdst(Rect(dstx, dsty, roicols, roirows));  //gdst_roi
-        gdstCoor_roi = gdstCoor(Rect(dstx, dsty, roicols, roirows));
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::meanShiftProc(gsrc_roi, gdst_roi, gdstCoor_roi, sp, sr, crit);
-    };
-#endif
-
-}
-
-///////////////////////////////////////////////////////////////////////////////////////////
-//hist
-
-void calcHistGold(const cv::Mat &src, cv::Mat &hist)
-{
-    hist.create(1, 256, CV_32SC1);
-    hist.setTo(cv::Scalar::all(0));
-
-    int *hist_row = hist.ptr<int>();
-    for (int y = 0; y < src.rows; ++y)
-    {
-        const uchar *src_row = src.ptr(y);
-
-        for (int x = 0; x < src.cols; ++x)
-            ++hist_row[src_row[x]];
     }
 }
 
-PARAM_TEST_CASE(histTestBase, MatType, MatType)
+typedef TestBaseWithParam<Size> meanShiftFilteringFixture;
+
+PERF_TEST_P(meanShiftFilteringFixture, meanShiftFiltering,
+            OCL_TYPICAL_MAT_SIZES)
 {
-    int type_src;
+    const Size srcSize = GetParam();
+    const int sp = 5, sr = 6;
+    cv::TermCriteria crit(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, 5, 1);
 
-    //src mat
-    cv::Mat src;
-    cv::Mat dst_hist;
-    //set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    //src mat with roi
-    cv::Mat src_roi;
-    //ocl dst mat, dst_hist and gdst_hist don't have roi
-    cv::ocl::oclMat gdst_hist;
+    Mat src(srcSize, CV_8UC4), dst(srcSize, CV_8UC4);
+    declare.in(src, WARMUP_RNG).out(dst)
+            .time(srcSize == OCL_SIZE_4000 ?
+                      56 : srcSize == OCL_SIZE_2000 ? 15 : 3.8);
 
-    //ocl mat with roi
-    cv::ocl::oclMat gsrc_roi;
-
-    //    std::vector<cv::ocl::Info> oclinfo;
-
-    virtual void SetUp()
+    if (RUN_PLAIN_IMPL)
     {
-        type_src   = GET_PARAM(0);
+        TEST_CYCLE() meanShiftFiltering_(src, dst, sp, sr, crit);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
-
-        src = randomMat(rng, size, type_src, 0, 256, false);
-
-        //        int devnums = getDevice(oclinfo);
-        //        CV_Assert(devnums > 0);
-        //if you want to use undefault device, set it here
-        //setDevice(oclinfo[0]);
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_OCL_IMPL)
     {
-        if(b)
-        {
-            //randomize ROI
-            roicols = src.cols - 1;
-            roirows = src.rows - 1;
-            srcx = 1;
-            srcy = 1;
-        }
-        else
-        {
-            roicols = src.cols;
-            roirows = src.rows;
-            srcx = 0;
-            srcy = 0;
-        };
-        src_roi = src(Rect(srcx, srcy, roicols, roirows));
+        ocl::oclMat oclSrc(src), oclDst(srcSize, CV_8UC4);
+
+        OCL_TEST_CYCLE() ocl::meanShiftFiltering(oclSrc, oclDst, sp, sr, crit);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-};
-
-///////////////////////////calcHist///////////////////////////////////////
-struct calcHist : histTestBase {};
-
-TEST_P(calcHist, Mat)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = 0; k < 2; k++)
-    {
-        double totalcputick = 0;
-        double totalgputick = 0;
-        double totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            calcHistGold(src_roi, dst_hist);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-
-            gsrc_roi = src_roi;
-
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::calcHist(gsrc_roi, gdst_hist);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-
-            cv::Mat cpu_hist;
-            gdst_hist.download(cpu_hist);//download
-
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalcputick = t0 + totalcputick;
-            totalgputick = t1 + totalgputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = 0; j < 2; j ++)
-    {
-        Has_roi(j);
-
-        gsrc_roi = src_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::calcHist(gsrc_roi, gdst_hist);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
+static void meanShiftProc_(const Mat &src_roi, Mat &dst_roi, Mat &dstCoor_roi, int sp, int sr, cv::TermCriteria crit)
+{
+    if (src_roi.empty())
+    {
+        CV_Error(Error::StsBadArg, "The input image is empty");
+    }
+    if (src_roi.depth() != CV_8U || src_roi.channels() != 4)
+    {
+        CV_Error(Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported");
+    }
 
-//************test*******************
+    dst_roi.create(src_roi.size(), src_roi.type());
+    dstCoor_roi.create(src_roi.size(), CV_16SC2);
 
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, equalizeHist, Combine(
-                            ONE_TYPE(CV_8UC1),
-                            NULL_TYPE,
-                            ONE_TYPE(CV_8UC1),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
+    CV_Assert((src_roi.cols == dst_roi.cols) && (src_roi.rows == dst_roi.rows) &&
+              (src_roi.cols == dstCoor_roi.cols) && (src_roi.rows == dstCoor_roi.rows));
+    CV_Assert(!(dstCoor_roi.step & 0x3));
 
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, bilateralFilter, Combine(
-                            Values(CV_8UC1, CV_8UC3),
-                            NULL_TYPE,
-                            Values(CV_8UC1, CV_8UC3),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
+    if (!(crit.type & cv::TermCriteria::MAX_ITER))
+    {
+        crit.maxCount = 5;
+    }
 
+    int maxIter = std::min(std::max(crit.maxCount, 1), 100);
+    float eps;
 
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, CopyMakeBorder, Combine(
-                            Values(CV_8UC1, CV_8UC4/*, CV_32SC1*/),
-                            NULL_TYPE,
-                            Values(CV_8UC1, CV_8UC4/*,CV_32SC1*/),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, cornerMinEigenVal, Combine(
-                            Values(CV_8UC1, CV_32FC1),
-                            NULL_TYPE,
-                            ONE_TYPE(CV_32FC1),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
+    if (!(crit.type & cv::TermCriteria::EPS))
+    {
+        eps = 1.f;
+    }
 
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, cornerHarris, Combine(
-                            Values(CV_8UC1, CV_32FC1),
-                            NULL_TYPE,
-                            ONE_TYPE(CV_32FC1),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
+    eps = (float)std::max(crit.epsilon, 0.0);
 
+    int tab[512];
 
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, integral, Combine(
-                            ONE_TYPE(CV_8UC1),
-                            NULL_TYPE,
-                            ONE_TYPE(CV_32SC1),
-                            ONE_TYPE(CV_32FC1),
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
+    for (int i = 0; i < 512; i++)
+    {
+        tab[i] = (i - 255) * (i - 255);
+    }
 
-INSTANTIATE_TEST_CASE_P(Imgproc, WarpAffine, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values((MatType)cv::INTER_NEAREST, (MatType)cv::INTER_LINEAR,
-                                   (MatType)cv::INTER_CUBIC, (MatType)(cv::INTER_NEAREST | cv::WARP_INVERSE_MAP),
-                                   (MatType)(cv::INTER_LINEAR | cv::WARP_INVERSE_MAP), (MatType)(cv::INTER_CUBIC | cv::WARP_INVERSE_MAP))));
+    uchar *sptr = src_roi.data;
+    uchar *dptr = dst_roi.data;
+    short *dCoorptr = (short *)dstCoor_roi.data;
+    int sstep = (int)src_roi.step;
+    int dstep = (int)dst_roi.step;
+    int dCoorstep = (int)dstCoor_roi.step >> 1;
+    cv::Size size = src_roi.size();
 
+    for (int i = 0; i < size.height; i++, sptr += sstep - (size.width << 2),
+            dptr += dstep - (size.width << 2), dCoorptr += dCoorstep - (size.width << 1))
+    {
+        for (int j = 0; j < size.width; j++, sptr += 4, dptr += 4, dCoorptr += 2)
+        {
+            *((COOR *)dCoorptr) = do_meanShift(j, i, sptr, dptr, sstep, size, sp, sr, maxIter, eps, tab);
+        }
+    }
 
-INSTANTIATE_TEST_CASE_P(Imgproc, WarpPerspective, Combine
-                        (Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                         Values((MatType)cv::INTER_NEAREST, (MatType)cv::INTER_LINEAR,
-                                (MatType)cv::INTER_CUBIC, (MatType)(cv::INTER_NEAREST | cv::WARP_INVERSE_MAP),
-                                (MatType)(cv::INTER_LINEAR | cv::WARP_INVERSE_MAP), (MatType)(cv::INTER_CUBIC | cv::WARP_INVERSE_MAP))));
+}
 
+typedef TestBaseWithParam<Size> meanShiftProcFixture;
 
-INSTANTIATE_TEST_CASE_P(Imgproc, Resize, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),  Values(cv::Size()),
-                            Values(0.5/*, 1.5, 2*/), Values(0.5/*, 1.5, 2*/), Values((MatType)cv::INTER_NEAREST, (MatType)cv::INTER_LINEAR)));
+PERF_TEST_P(meanShiftProcFixture, meanShiftProc,
+            OCL_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+    TermCriteria crit(TermCriteria::COUNT + TermCriteria::EPS, 5, 1);
 
+    Mat src(srcSize, CV_8UC4), dst1(srcSize, CV_8UC4),
+            dst2(srcSize, CV_16SC2);
+    declare.in(src, WARMUP_RNG).out(dst1, dst2)
+            .time(srcSize == OCL_SIZE_4000 ?
+                      56 : srcSize == OCL_SIZE_2000 ? 15 : 3.8);;
 
-INSTANTIATE_TEST_CASE_P(Imgproc, Threshold, Combine(
-                            Values(CV_8UC1, CV_32FC1), Values(ThreshOp(cv::THRESH_BINARY),
-                                    ThreshOp(cv::THRESH_BINARY_INV), ThreshOp(cv::THRESH_TRUNC),
-                                    ThreshOp(cv::THRESH_TOZERO), ThreshOp(cv::THRESH_TOZERO_INV))));
+    if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() meanShiftProc_(src, dst1, dst2, 5, 6, crit);
 
-INSTANTIATE_TEST_CASE_P(Imgproc, meanShiftFiltering, Combine(
-                            ONE_TYPE(CV_8UC4),
-                            ONE_TYPE(CV_16SC2),//it is no use in meanShiftFiltering
-                            Values(5),
-                            Values(6),
-                            Values(cv::TermCriteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, 5, 1))
-                        ));
+        SANITY_CHECK(dst1);
+        SANITY_CHECK(dst2);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst1(srcSize, CV_8UC4),
+                oclDst2(srcSize, CV_16SC2);
 
-INSTANTIATE_TEST_CASE_P(Imgproc, meanShiftProc, Combine(
-                            ONE_TYPE(CV_8UC4),
-                            ONE_TYPE(CV_16SC2),
-                            Values(5),
-                            Values(6),
-                            Values(cv::TermCriteria(cv::TermCriteria::COUNT + cv::TermCriteria::EPS, 5, 1))
-                        ));
+        OCL_TEST_CYCLE() ocl::meanShiftProc(oclSrc, oclDst1, oclDst2, 5, 6, crit);
 
-INSTANTIATE_TEST_CASE_P(Imgproc, Remap, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(CV_32FC1, CV_16SC2, CV_32FC2), Values(-1, CV_32FC1),
-                            Values((int)cv::INTER_NEAREST, (int)cv::INTER_LINEAR),
-                            Values((int)cv::BORDER_CONSTANT)));
+        oclDst1.download(dst1);
+        oclDst2.download(dst2);
 
-INSTANTIATE_TEST_CASE_P(histTestBase, calcHist, Combine(
-                            ONE_TYPE(CV_8UC1),
-                            ONE_TYPE(CV_32SC1) //no use
-                        ));
+        SANITY_CHECK(dst1);
+        SANITY_CHECK(dst2);
+    }
+    else
+        OCL_PERF_ELSE
+}
 
-#endif // HAVE_OPENCL
+///////////// remap////////////////////////
+
+CV_ENUM(RemapInterType, INTER_NEAREST, INTER_LINEAR)
+
+typedef tuple<Size, MatType, RemapInterType> remapParams;
+typedef TestBaseWithParam<remapParams> remapFixture;
+
+PERF_TEST_P(remapFixture, remap,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4),
+                               RemapInterType::all()))
+{
+    const remapParams params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params), interpolation = get<2>(params);
+
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (srcSize == OCL_SIZE_4000 && interpolation == INTER_LINEAR)
+        declare.time(9);
+
+    Mat xmap, ymap;
+    xmap.create(srcSize, CV_32FC1);
+    ymap.create(srcSize, CV_32FC1);
+
+    for (int i = 0; i < srcSize.height; ++i)
+    {
+        float * const xmap_row = xmap.ptr<float>(i);
+        float * const ymap_row = ymap.ptr<float>(i);
+
+        for (int j = 0; j < srcSize.width; ++j)
+        {
+            xmap_row[j] = (j - srcSize.width * 0.5f) * 0.75f + srcSize.width * 0.5f;
+            ymap_row[j] = (i - srcSize.height * 0.5f) * 0.75f + srcSize.height * 0.5f;
+        }
+    }
+
+    const int borderMode = BORDER_CONSTANT;
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
+        ocl::oclMat oclXMap(xmap), oclYMap(ymap);
+
+        OCL_TEST_CYCLE() cv::ocl::remap(oclSrc, oclDst, oclXMap, oclYMap, interpolation, borderMode);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::remap(src, dst, xmap, ymap, interpolation, borderMode);
+
+        SANITY_CHECK(dst, 1 + DBL_EPSILON);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+///////////// CLAHE ////////////////////////
+
+typedef TestBaseWithParam<Size> CLAHEFixture;
+
+PERF_TEST_P(CLAHEFixture, CLAHE, OCL_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+    const string impl = getSelectedImpl();
+
+    Mat src(srcSize, CV_8UC1), dst;
+    const double clipLimit = 40.0;
+    declare.in(src, WARMUP_RNG);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(11);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst;
+        cv::Ptr<cv::CLAHE> oclClahe = cv::ocl::createCLAHE(clipLimit);
+
+        OCL_TEST_CYCLE() oclClahe->apply(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        cv::Ptr<cv::CLAHE> clahe = cv::createCLAHE(clipLimit);
+        TEST_CYCLE() clahe->apply(src, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+///////////// columnSum////////////////////////
+
+typedef TestBaseWithParam<Size> columnSumFixture;
+
+static void columnSumPerfTest(const Mat & src, Mat & dst)
+{
+    for (int j = 0; j < src.cols; j++)
+        dst.at<float>(0, j) = src.at<float>(0, j);
+
+    for (int i = 1; i < src.rows; ++i)
+        for (int j = 0; j < src.cols; ++j)
+            dst.at<float>(i, j) = dst.at<float>(i - 1 , j) + src.at<float>(i , j);
+}
+
+PERF_TEST_P(columnSumFixture, columnSum, OCL_TYPICAL_MAT_SIZES)
+{
+    const Size srcSize = GetParam();
+
+    Mat src(srcSize, CV_32FC1), dst(srcSize, CV_32FC1);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (srcSize == OCL_SIZE_4000)
+        declare.time(5);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(srcSize, CV_32FC1);
+
+        OCL_TEST_CYCLE() cv::ocl::columnSum(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() columnSumPerfTest(src, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_match_template.cpp b/modules/ocl/perf/perf_match_template.cpp
index cb5e86bab9..869e01e602 100644
--- a/modules/ocl/perf/perf_match_template.cpp
+++ b/modules/ocl/perf/perf_match_template.cpp
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -42,191 +43,79 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-#include <iomanip>
-#ifdef HAVE_OPENCL
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
-#ifndef MWC_TEST_UTILITY
-#define MWC_TEST_UTILITY
-//////// Utility
-#ifndef DIFFERENT_SIZES
-#else
-#undef DIFFERENT_SIZES
-#endif
-#define DIFFERENT_SIZES testing::Values(cv::Size(256, 256), cv::Size(3000, 3000))
+/////////// matchTemplate ////////////////////////
 
-// Param class
-#ifndef IMPLEMENT_PARAM_CLASS
-#define IMPLEMENT_PARAM_CLASS(name, type) \
-class name \
-{ \
-public: \
-    name ( type arg = type ()) : val_(arg) {} \
-    operator type () const {return val_;} \
-private: \
-    type val_; \
-}; \
-    inline void PrintTo( name param, std::ostream* os) \
-{ \
-    *os << #name <<  "(" << testing::PrintToString(static_cast< type >(param)) << ")"; \
+typedef Size_MatType CV_TM_CCORRFixture;
+
+PERF_TEST_P(CV_TM_CCORRFixture, matchTemplate,
+            ::testing::Combine(::testing::Values(OCL_SIZE_1000, OCL_SIZE_2000),
+                               OCL_PERF_ENUM(CV_32FC1, CV_32FC4)))
+{
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params), templSize(5, 5);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), templ(templSize, type);
+    const Size dstSize(src.cols - templ.cols + 1, src.rows - templ.rows + 1);
+    Mat dst(dstSize, CV_32F);
+    randu(src, 0.0f, 1.0f);
+    randu(templ, 0.0f, 1.0f);
+    declare.time(srcSize == OCL_SIZE_2000 ? 20 : 6).in(src, templ).out(dst);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclTempl(templ), oclDst(dstSize, CV_32F);
+
+        OCL_TEST_CYCLE() cv::ocl::matchTemplate(oclSrc, oclTempl, oclDst, TM_CCORR);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 1e-4);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() cv::matchTemplate(src, templ, dst, TM_CCORR);
+
+        SANITY_CHECK(dst, 1e-4);
+    }
+    else
+        OCL_PERF_ELSE
 }
 
-IMPLEMENT_PARAM_CLASS(Channels, int)
-#endif // IMPLEMENT_PARAM_CLASS
-#endif // MWC_TEST_UTILITY
+typedef TestBaseWithParam<Size> CV_TM_CCORR_NORMEDFixture;
 
-////////////////////////////////////////////////////////////////////////////////
-// MatchTemplate
-#define ALL_TEMPLATE_METHODS testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR), TemplateMethod(cv::TM_CCOEFF), TemplateMethod(cv::TM_SQDIFF_NORMED), TemplateMethod(cv::TM_CCORR_NORMED), TemplateMethod(cv::TM_CCOEFF_NORMED))
-
-IMPLEMENT_PARAM_CLASS(TemplateSize, cv::Size);
-
-const char *TEMPLATE_METHOD_NAMES[6] = {"TM_SQDIFF", "TM_SQDIFF_NORMED", "TM_CCORR", "TM_CCORR_NORMED", "TM_CCOEFF", "TM_CCOEFF_NORMED"};
-
-PARAM_TEST_CASE(MatchTemplate, cv::Size, TemplateSize, Channels, TemplateMethod)
+PERF_TEST_P(CV_TM_CCORR_NORMEDFixture, matchTemplate, OCL_TYPICAL_MAT_SIZES)
 {
-    cv::Size size;
-    cv::Size templ_size;
-    int cn;
-    int method;
-    //vector<cv::ocl::Info> oclinfo;
+    const Size srcSize = GetParam(), templSize(5, 5);
 
-    virtual void SetUp()
+    Mat src(srcSize, CV_8UC1), templ(templSize, CV_8UC1), dst;
+    const Size dstSize(src.cols - templ.cols + 1, src.rows - templ.rows + 1);
+    dst.create(dstSize, CV_8UC1);
+    declare.in(src, templ, WARMUP_RNG).out(dst)
+            .time(srcSize == OCL_SIZE_2000 ? 10 : srcSize == OCL_SIZE_4000 ? 23 : 2);
+
+    if (RUN_OCL_IMPL)
     {
-        size = GET_PARAM(0);
-        templ_size = GET_PARAM(1);
-        cn = GET_PARAM(2);
-        method = GET_PARAM(3);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
+        ocl::oclMat oclSrc(src), oclTempl(templ), oclDst(dstSize, CV_8UC1);
+
+        OCL_TEST_CYCLE() cv::ocl::matchTemplate(oclSrc, oclTempl, oclDst, TM_CCORR_NORMED);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst, 2e-2);
     }
-};
-struct MatchTemplate8U : MatchTemplate {};
-
-TEST_P(MatchTemplate8U, Performance)
-{
-    std::cout << "Method: " << TEMPLATE_METHOD_NAMES[method] << std::endl;
-    std::cout << "Image Size: (" << size.width << ", " << size.height << ")" << std::endl;
-    std::cout << "Template Size: (" << templ_size.width << ", " << templ_size.height << ")" << std::endl;
-    std::cout << "Channels: " << cn << std::endl;
-
-    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_8U, cn));
-    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_8U, cn));
-    cv::Mat dst_gold;
-    cv::ocl::oclMat dst;
-
-
-
-
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
+        TEST_CYCLE() cv::matchTemplate(src, templ, dst, TM_CCORR_NORMED);
 
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat ocl_image = cv::ocl::oclMat(image);//upload
-        cv::ocl::oclMat ocl_templ = cv::ocl::oclMat(templ);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::matchTemplate(ocl_image, ocl_templ, dst, method);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        dst.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
+        SANITY_CHECK(dst, 2e-2);
     }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
+    else
+        OCL_PERF_ELSE
 }
-
-
-struct MatchTemplate32F : MatchTemplate {};
-TEST_P(MatchTemplate32F, Performance)
-{
-    std::cout << "Method: " << TEMPLATE_METHOD_NAMES[method] << std::endl;
-    std::cout << "Image Size: (" << size.width << ", " << size.height << ")" << std::endl;
-    std::cout << "Template Size: (" << templ_size.width << ", " << templ_size.height << ")" << std::endl;
-    std::cout << "Channels: " << cn << std::endl;
-    cv::Mat image = randomMat(size, CV_MAKETYPE(CV_32F, cn));
-    cv::Mat templ = randomMat(templ_size, CV_MAKETYPE(CV_32F, cn));
-
-    cv::Mat dst_gold;
-    cv::ocl::oclMat dst;
-
-
-
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-    for(int j = 0; j < LOOP_TIMES; j ++)
-    {
-
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat ocl_image = cv::ocl::oclMat(image);//upload
-        cv::ocl::oclMat ocl_templ = cv::ocl::oclMat(templ);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::matchTemplate(ocl_image, ocl_templ, dst, method);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        dst.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
-    }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
-
-}
-
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate8U,
-                        testing::Combine(
-                            testing::Values(cv::Size(1280, 1024), cv::Size(MWIDTH, MHEIGHT), cv::Size(1800, 1500)),
-                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16))/*, TemplateSize(cv::Size(30, 30))*/),
-                            testing::Values(Channels(1), Channels(4)/*, Channels(3)*/),
-                            ALL_TEMPLATE_METHODS
-                        )
-                       );
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate32F, testing::Combine(
-                            testing::Values(cv::Size(1280, 1024), cv::Size(MWIDTH, MHEIGHT), cv::Size(1800, 1500)),
-                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16))/*, TemplateSize(cv::Size(30, 30))*/),
-                            testing::Values(Channels(1), Channels(4) /*, Channels(3)*/),
-                            testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))));
-
-#endif //HAVE_OPENCL
\ No newline at end of file
diff --git a/modules/ocl/perf/perf_matrix_operation.cpp b/modules/ocl/perf/perf_matrix_operation.cpp
index ba011f8dfc..ad13275030 100644
--- a/modules/ocl/perf/perf_matrix_operation.cpp
+++ b/modules/ocl/perf/perf_matrix_operation.cpp
@@ -10,12 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -30,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -42,697 +43,115 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
-#ifdef HAVE_OPENCL
+///////////// ConvertTo////////////////////////
 
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-using namespace cv::ocl;
-////////////////////////////////converto/////////////////////////////////////////////////
-PARAM_TEST_CASE(ConvertToTestBase, MatType, MatType)
+typedef Size_MatType ConvertToFixture;
+
+PERF_TEST_P(ConvertToFixture, ConvertTo,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    int dst_type;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-    //src mat
-    cv::Mat mat;
-    cv::Mat dst;
+    Mat src(srcSize, type), dst;
+    const int dstType = CV_MAKE_TYPE(CV_32F, src.channels());
+    dst.create(srcSize, dstType);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type     = GET_PARAM(0);
-        dst_type = GET_PARAM(1);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, dstType);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() oclSrc.convertTo(oclDst, dstType);
 
-        mat = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat.cols - 1; //start
-            roirows = mat.rows - 1;
-            srcx   = 1;
-            srcy   = 1;
-            dstx    = 1;
-            dsty    = 1;
-        }
-        else
-        {
-            roicols = mat.cols;
-            roirows = mat.rows;
-            srcx   = 0;
-            srcy   = 0;
-            dstx   = 0;
-            dsty   = 0;
-        };
+        TEST_CYCLE() src.convertTo(dst, dstType);
 
-        mat_roi = mat(Rect(srcx, srcy, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        //gdst_whole = dst;
-        //gdst = gdst_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gmat = mat_roi;
+        SANITY_CHECK(dst);
     }
-};
-
-
-struct ConvertTo : ConvertToTestBase {};
-
-TEST_P(ConvertTo, Accuracy)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            mat_roi.convertTo(dst_roi, dst_type);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat = mat_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            gmat.convertTo(gdst, dst_type);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gmat = mat_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        gmat.convertTo(gdst, dst_type);
-    };
-#endif
-
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// copyTo////////////////////////
 
-///////////////////////////////////////////copyto/////////////////////////////////////////////////////////////
+typedef Size_MatType copyToFixture;
 
-PARAM_TEST_CASE(CopyToTestBase, MatType, bool)
+PERF_TEST_P(copyToFixture, copyTo,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-    cv::Mat mat;
-    cv::Mat mask;
-    cv::Mat dst;
+    Mat src(srcSize, type), dst(srcSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
 
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
-    int maskx;
-    int masky;
-
-    //src mat with roi
-    cv::Mat mat_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
+        ocl::oclMat oclSrc(src), oclDst(srcSize, type);
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() oclSrc.copyTo(oclDst);
 
-        mat = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
+        oclDst.download(dst);
 
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
+        SANITY_CHECK(dst);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat.cols - 1; //start
-            roirows = mat.rows - 1;
-            srcx   = 1;
-            srcy   = 1;
-            dstx    = 1;
-            dsty    = 1;
-            maskx   = 1;
-            masky   = 1;
-        }
-        else
-        {
-            roicols = mat.cols;
-            roirows = mat.rows;
-            srcx   = 0;
-            srcy   = 0;
-            dstx   = 0;
-            dsty   = 0;
-            maskx   = 0;
-            masky   = 0;
-        };
+        TEST_CYCLE() src.copyTo(dst);
 
-        mat_roi = mat(Rect(srcx, srcy, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        //gdst_whole = dst;
-        //gdst = gdst_whole(Rect(dstx,dsty,roicols,roirows));
-
-        //gmat = mat_roi;
-        //gmask = mask_roi;
+        SANITY_CHECK(dst);
     }
-};
-
-struct CopyTo : CopyToTestBase {};
-
-TEST_P(CopyTo, Without_mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            mat_roi.copyTo(dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat = mat_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            gmat.copyTo(gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gmat = mat_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        gmat.copyTo(gdst);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
-TEST_P(CopyTo, With_mask)
+///////////// setTo////////////////////////
+
+typedef Size_MatType setToFixture;
+
+PERF_TEST_P(setToFixture, setTo,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+    const Scalar val(1, 2, 3, 4);
+
+    Mat src(srcSize, type);
+    declare.in(src);
+
+    if (RUN_OCL_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
+        ocl::oclMat oclSrc(srcSize, type);
 
-            t0 = (double)cvGetTickCount();//cpu start
-            mat_roi.copyTo(dst_roi, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        OCL_TEST_CYCLE() oclSrc.setTo(val);
+        oclSrc.download(src);
 
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-            gmat = mat_roi;
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            gmat.copyTo(gdst, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(src);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        Has_roi(j);
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
+        TEST_CYCLE() src.setTo(val);
 
-        gmat = mat_roi;
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        gmat.copyTo(gdst, gmask);
-    };
-#endif
+        SANITY_CHECK(src);
+    }
+    else
+        OCL_PERF_ELSE
 }
-
-///////////////////////////////////////////copyto/////////////////////////////////////////////////////////////
-
-PARAM_TEST_CASE(SetToTestBase, MatType, bool)
-{
-    int type;
-    cv::Scalar val;
-
-    cv::Mat mat;
-    cv::Mat mask;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int maskx;
-    int masky;
-
-    //src mat with roi
-    cv::Mat mat_roi;
-    cv::Mat mask_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gmat_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
-
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
-    }
-
-    void Has_roi(int b)
-    {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat.cols - 1; //start
-            roirows = mat.rows - 1;
-            srcx   = 1;
-            srcy   = 1;
-            maskx   = 1;
-            masky   = 1;
-        }
-        else
-        {
-            roicols = mat.cols;
-            roirows = mat.rows;
-            srcx   = 0;
-            srcy   = 0;
-            maskx   = 0;
-            masky   = 0;
-        };
-
-        mat_roi = mat(Rect(srcx, srcy, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-
-        //gmat_whole = mat;
-        //gmat = gmat_whole(Rect(srcx,srcy,roicols,roirows));
-
-        //gmask = mask_roi;
-    }
-};
-
-struct SetTo : SetToTestBase {};
-
-TEST_P(SetTo, Without_mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            mat_roi.setTo(val);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat_whole = mat;
-            gmat = gmat_whole(Rect(srcx, srcy, roicols, roirows));
-            t2 = (double)cvGetTickCount(); //kernel
-            gmat.setTo(val);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gmat_whole.download(cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat_whole = mat;
-        gmat = gmat_whole(Rect(srcx, srcy, roicols, roirows));
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        gmat.setTo(val);
-    };
-#endif
-}
-
-TEST_P(SetTo, With_mask)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-
-            t0 = (double)cvGetTickCount();//cpu start
-            mat_roi.setTo(val, mask_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gmat_whole = mat;
-            gmat = gmat_whole(Rect(srcx, srcy, roicols, roirows));
-
-            gmask = mask_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            gmat.setTo(val, gmask);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gmat_whole.download(cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat_whole = mat;
-        gmat = gmat_whole(Rect(srcx, srcy, roicols, roirows));
-
-        gmask = mask_roi;
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        gmat.setTo(val, gmask);
-    };
-#endif
-}
-PARAM_TEST_CASE(DataTransfer, MatType, bool)
-{
-    int type;
-    cv::Mat mat;
-    cv::ocl::oclMat gmat_whole;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-        mat = randomMat(rng, size, type, 5, 16, false);
-    }
-};
-TEST_P(DataTransfer, perf)
-{
-    double totaluploadtick = 0;
-    double totaldownloadtick = 0;
-    double totaltick = 0;
-    double t0 = 0;
-    double t1 = 0;
-    cv::Mat cpu_dst;
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
-    {
-        t0 = (double)cvGetTickCount();
-        gmat_whole.upload(mat);//upload
-        t0 = (double)cvGetTickCount() - t0;
-
-        t1 = (double)cvGetTickCount();
-        gmat_whole.download(cpu_dst);//download
-        t1 = (double)cvGetTickCount() - t1;
-
-        if(j == 0)
-            continue;
-        totaluploadtick = t0 + totaluploadtick;
-        totaldownloadtick = t1 + totaldownloadtick;
-    }
-    totaltick = totaluploadtick + totaldownloadtick;
-    cout << "average upload time is  " << totaluploadtick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average download time is  " << totaldownloadtick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average data transfer time is  " << totaltick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-}
-//**********test************
-
-INSTANTIATE_TEST_CASE_P(MatrixOperation, ConvertTo, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4)));
-
-INSTANTIATE_TEST_CASE_P(MatrixOperation, CopyTo, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-
-INSTANTIATE_TEST_CASE_P(MatrixOperation, SetTo, Combine(
-                            Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-INSTANTIATE_TEST_CASE_P(MatrixOperation, DataTransfer, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
-                            Values(false))); // Values(false) is the reserved parameter
-#endif
diff --git a/modules/ocl/test/test_pyrup.cpp b/modules/ocl/perf/perf_moments.cpp
similarity index 65%
rename from modules/ocl/test/test_pyrup.cpp
rename to modules/ocl/perf/perf_moments.cpp
index 2e0cb92cce..6ecc766517 100644
--- a/modules/ocl/test/test_pyrup.cpp
+++ b/modules/ocl/perf/perf_moments.cpp
@@ -15,8 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Zhang Chunpeng chunpeng@multicorewareinc.com
-//    Yao Wang yao@multicorewareinc.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -31,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -44,53 +44,46 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "perf_precomp.hpp"
 
-#ifdef HAVE_OPENCL
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
-using namespace cv;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+///////////// Moments ////////////////////////
 
-PARAM_TEST_CASE(PyrUp, MatType, int)
+typedef Size_MatType MomentsFixture;
+
+PERF_TEST_P(MomentsFixture, DISABLED_Moments,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_16SC1, CV_32FC1, CV_64FC1)))  // TODO does not work properly (see below)
 {
-    int type;
-    int channels;
-    //std::vector<cv::ocl::Info> oclinfo;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-    virtual void SetUp()
+    Mat src(srcSize, type), dst(7, 1, CV_64F);
+    const bool binaryImage = false;
+    cv::Moments mom;
+
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        //int devnums = cv::ocl::getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
-    }
-};
+        ocl::oclMat oclSrc(src);
 
-TEST_P(PyrUp, Accuracy)
-{
-    for(int j = 0; j < LOOP_TIMES; j++)
+        OCL_TEST_CYCLE() mom = cv::ocl::ocl_moments(oclSrc, binaryImage); // TODO Use oclSrc
+        cv::HuMoments(mom, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
     {
-        Size size(MWIDTH, MHEIGHT);
-        Mat src = randomMat(size, CV_MAKETYPE(type, channels));
-        Mat dst_gold;
-        pyrUp(src, dst_gold);
-        ocl::oclMat dst;
-        ocl::oclMat srcMat(src);
-        ocl::pyrUp(srcMat, dst);
-        Mat cpu_dst;
-        dst.download(cpu_dst);
-        char s[100] = {0};
+        TEST_CYCLE() mom = cv::moments(src, binaryImage);
+        cv::HuMoments(mom, dst);
 
-        EXPECT_MAT_NEAR(dst_gold, cpu_dst, (src.depth() == CV_32F ? 1e-4f : 1.0), s);
+        SANITY_CHECK(dst);
     }
-
+    else
+        OCL_PERF_ELSE
 }
-
-
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrUp, testing::Combine(
-                            Values(CV_8U, CV_32F), Values(1, 3, 4)));
-
-
-#endif // HAVE_OPENCL
\ No newline at end of file
diff --git a/modules/softcascade/perf/perf_precomp.cpp b/modules/ocl/perf/perf_norm.cpp
similarity index 62%
rename from modules/softcascade/perf/perf_precomp.cpp
rename to modules/ocl/perf/perf_norm.cpp
index cc728b050f..363bcd2b33 100644
--- a/modules/softcascade/perf/perf_precomp.cpp
+++ b/modules/ocl/perf/perf_norm.cpp
@@ -10,10 +10,14 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2008-2013, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
+// @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -22,12 +26,12 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and / or other materials provided with the distribution.
+//     and/or other oclMaterials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -39,5 +43,41 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
-
 #include "perf_precomp.hpp"
+
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
+
+///////////// norm////////////////////////
+
+typedef TestBaseWithParam<Size> normFixture;
+
+PERF_TEST_P(normFixture, DISABLED_norm, OCL_TYPICAL_MAT_SIZES) // TODO doesn't work properly
+{
+    const Size srcSize = GetParam();
+    const std::string impl = getSelectedImpl();
+    double value = 0.0;
+
+    Mat src1(srcSize, CV_8UC1), src2(srcSize, CV_8UC1);
+    declare.in(src1, src2);
+    randu(src1, 0, 1);
+    randu(src2, 0, 1);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc1(src1), oclSrc2(src2);
+
+        OCL_TEST_CYCLE() value = cv::ocl::norm(oclSrc1, oclSrc2, NORM_INF);
+
+        SANITY_CHECK(value);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() value = cv::norm(src1, src2, NORM_INF);
+
+        SANITY_CHECK(value);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_opticalflow.cpp b/modules/ocl/perf/perf_opticalflow.cpp
new file mode 100644
index 0000000000..8613075261
--- /dev/null
+++ b/modules/ocl/perf/perf_opticalflow.cpp
@@ -0,0 +1,255 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "perf_precomp.hpp"
+
+///////////// PyrLKOpticalFlow ////////////////////////
+
+using namespace perf;
+using std::tr1::get;
+using std::tr1::tuple;
+using std::tr1::make_tuple;
+
+CV_ENUM(LoadMode, IMREAD_GRAYSCALE, IMREAD_COLOR)
+
+typedef tuple<int, tuple<string, string, LoadMode> > PyrLKOpticalFlowParamType;
+typedef TestBaseWithParam<PyrLKOpticalFlowParamType> PyrLKOpticalFlowFixture;
+
+PERF_TEST_P(PyrLKOpticalFlowFixture,
+            PyrLKOpticalFlow,
+            ::testing::Combine(
+                ::testing::Values(1000, 2000, 4000),
+                ::testing::Values(
+                    make_tuple<string, string, LoadMode>
+                    (
+                        string("gpu/opticalflow/rubberwhale1.png"),
+                        string("gpu/opticalflow/rubberwhale2.png"),
+                        LoadMode(IMREAD_COLOR)
+                        ),
+                    make_tuple<string, string, LoadMode>
+                    (
+                        string("gpu/stereobm/aloe-L.png"),
+                        string("gpu/stereobm/aloe-R.png"),
+                        LoadMode(IMREAD_GRAYSCALE)
+                        )
+                    )
+                )
+            )
+{
+    PyrLKOpticalFlowParamType params = GetParam();
+    tuple<string, string, LoadMode> fileParam = get<1>(params);
+    const int pointsCount = get<0>(params);
+    const int openMode = static_cast<int>(get<2>(fileParam));
+    const string fileName0 = get<0>(fileParam), fileName1 = get<1>(fileParam);
+    Mat frame0 = imread(getDataPath(fileName0), openMode);
+    Mat frame1 = imread(getDataPath(fileName1), openMode);
+
+    declare.in(frame0, frame1);
+
+    ASSERT_FALSE(frame0.empty()) << "can't load " << fileName0;
+    ASSERT_FALSE(frame1.empty()) << "can't load " << fileName1;
+
+    Mat grayFrame;
+    if (openMode == IMREAD_COLOR)
+        cvtColor(frame0, grayFrame, COLOR_BGR2GRAY);
+    else
+        grayFrame = frame0;
+
+    vector<Point2f> pts, nextPts;
+    vector<unsigned char> status;
+    vector<float> err;
+    goodFeaturesToTrack(grayFrame, pts, pointsCount, 0.01, 0.0);
+    Mat ptsMat(1, static_cast<int>(pts.size()), CV_32FC2, (void *)&pts[0]);
+
+    if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE()
+                cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, err);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::PyrLKOpticalFlow oclPyrLK;
+        ocl::oclMat oclFrame0(frame0), oclFrame1(frame1);
+        ocl::oclMat oclPts(ptsMat);
+        ocl::oclMat oclNextPts, oclStatus, oclErr;
+
+        OCL_TEST_CYCLE()
+                oclPyrLK.sparse(oclFrame0, oclFrame1, oclPts, oclNextPts, oclStatus, &oclErr);
+    }
+    else
+        OCL_PERF_ELSE
+
+    int value = 0;
+    SANITY_CHECK(value);
+}
+
+PERF_TEST(tvl1flowFixture, tvl1flow)
+{
+    Mat frame0 = imread(getDataPath("gpu/opticalflow/rubberwhale1.png"), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty()) << "can't load rubberwhale1.png";
+
+    Mat frame1 = imread(getDataPath("gpu/opticalflow/rubberwhale2.png"), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty()) << "can't load rubberwhale2.png";
+
+    const Size srcSize = frame0.size();
+    const double eps = 1.2;
+    Mat flow(srcSize, CV_32FC2), flow1(srcSize, CV_32FC1), flow2(srcSize, CV_32FC1);
+    declare.in(frame0, frame1).out(flow1, flow2).time(159);
+
+    if (RUN_PLAIN_IMPL)
+    {
+        Ptr<DenseOpticalFlow> alg = createOptFlow_DualTVL1();
+
+        TEST_CYCLE() alg->calc(frame0, frame1, flow);
+
+        alg->collectGarbage();
+        Mat flows[2] = { flow1, flow2 };
+        split(flow, flows);
+
+        SANITY_CHECK(flow1, eps);
+        SANITY_CHECK(flow2, eps);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::OpticalFlowDual_TVL1_OCL oclAlg;
+        ocl::oclMat oclFrame0(frame0), oclFrame1(frame1), oclFlow1(srcSize, CV_32FC1),
+                oclFlow2(srcSize, CV_32FC1);
+
+        OCL_TEST_CYCLE() oclAlg(oclFrame0, oclFrame1, oclFlow1, oclFlow2);
+
+        oclAlg.collectGarbage();
+
+        oclFlow1.download(flow1);
+        oclFlow2.download(flow2);
+
+        SANITY_CHECK(flow1, eps);
+        SANITY_CHECK(flow2, eps);
+    }
+    else
+        OCL_PERF_ELSE
+}
+
+///////////// FarnebackOpticalFlow ////////////////////////
+
+CV_ENUM(farneFlagType, 0, OPTFLOW_FARNEBACK_GAUSSIAN)
+
+typedef tuple<tuple<int, double>, farneFlagType, bool> FarnebackOpticalFlowParams;
+typedef TestBaseWithParam<FarnebackOpticalFlowParams> FarnebackOpticalFlowFixture;
+
+PERF_TEST_P(FarnebackOpticalFlowFixture, FarnebackOpticalFlow,
+            ::testing::Combine(
+                ::testing::Values(make_tuple<int, double>(5, 1.1),
+                                  make_tuple<int, double>(7, 1.5)),
+                farneFlagType::all(),
+                ::testing::Bool()))
+{
+    Mat frame0 = imread(getDataPath("gpu/opticalflow/rubberwhale1.png"), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty()) << "can't load rubberwhale1.png";
+
+    Mat frame1 = imread(getDataPath("gpu/opticalflow/rubberwhale2.png"), cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty()) << "can't load rubberwhale2.png";
+
+    const Size srcSize = frame0.size();
+
+    const FarnebackOpticalFlowParams params = GetParam();
+    const tuple<int, double> polyParams = get<0>(params);
+    const int polyN = get<0>(polyParams), flags = get<1>(params);
+    const double polySigma = get<1>(polyParams), pyrScale = 0.5;
+    const bool useInitFlow = get<2>(params);
+    const double eps = 1.5;
+
+    Mat flowx(srcSize, CV_32FC1), flowy(srcSize, CV_32FC1), flow(srcSize, CV_32FC2);
+    declare.in(frame0, frame1).out(flowx, flowy);
+
+    ocl::FarnebackOpticalFlow farn;
+    farn.pyrScale = pyrScale;
+    farn.polyN = polyN;
+    farn.polySigma = polySigma;
+    farn.flags = flags;
+
+    if (RUN_PLAIN_IMPL)
+    {
+        if (useInitFlow)
+        {
+            calcOpticalFlowFarneback(
+                        frame0, frame1, flow, farn.pyrScale, farn.numLevels, farn.winSize,
+                        farn.numIters, farn.polyN, farn.polySigma, farn.flags);
+            farn.flags |= OPTFLOW_USE_INITIAL_FLOW;
+        }
+
+        TEST_CYCLE()
+                calcOpticalFlowFarneback(
+                    frame0, frame1, flow, farn.pyrScale, farn.numLevels, farn.winSize,
+                    farn.numIters, farn.polyN, farn.polySigma, farn.flags);
+
+        Mat flowxy[2] = { flowx, flowy };
+        split(flow, flowxy);
+
+        SANITY_CHECK(flowx, eps);
+        SANITY_CHECK(flowy, eps);
+    }
+    else if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclFrame0(frame0), oclFrame1(frame1),
+                oclFlowx(srcSize, CV_32FC1), oclFlowy(srcSize, CV_32FC1);
+
+        if (useInitFlow)
+        {
+            farn(oclFrame0, oclFrame1, oclFlowx, oclFlowy);
+            farn.flags |= OPTFLOW_USE_INITIAL_FLOW;
+        }
+
+        OCL_TEST_CYCLE()
+                farn(oclFrame0, oclFrame1, oclFlowx, oclFlowy);
+
+        oclFlowx.download(flowx);
+        oclFlowy.download(flowy);
+
+        SANITY_CHECK(flowx, eps);
+        SANITY_CHECK(flowy, eps);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_columnsum.cpp b/modules/ocl/perf/perf_precomp.hpp
similarity index 50%
rename from modules/ocl/perf/perf_columnsum.cpp
rename to modules/ocl/perf/perf_precomp.hpp
index 96ea26a503..0235b3d045 100644
--- a/modules/ocl/perf/perf_columnsum.cpp
+++ b/modules/ocl/perf/perf_precomp.hpp
@@ -13,11 +13,7 @@
 // Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
-//
-// @Authors
-//	   Fangfang Bai fangfang@multicorewareinc.com
-//
-//
+
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -31,7 +27,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -44,77 +40,80 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
 #include <iomanip>
+#include <stdexcept>
+#include <string>
+#include <iostream>
+#include <cstdio>
+#include <vector>
+#include <numeric>
+
+#include "cvconfig.h"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/ocl.hpp"
+#include "opencv2/ts.hpp"
 
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
 using namespace std;
+using namespace cv;
 
-///////////////////////////////////////////////////////////////////////////////
-/// ColumnSum
+#define OCL_SIZE_1000 Size(1000, 1000)
+#define OCL_SIZE_2000 Size(2000, 2000)
+#define OCL_SIZE_4000 Size(4000, 4000)
 
-#ifdef HAVE_OPENCL
+#define OCL_TYPICAL_MAT_SIZES ::testing::Values(OCL_SIZE_1000, OCL_SIZE_2000, OCL_SIZE_4000)
 
-////////////////////////////////////////////////////////////////////////
-// ColumnSum
+#define OCL_PERF_ENUM(type, ...) ::testing::Values(type, ## __VA_ARGS__ )
 
-PARAM_TEST_CASE(ColumnSum)
-{
-    cv::Mat src;
-    //std::vector<cv::ocl::Info> oclinfo;
+#define IMPL_OCL "ocl"
+#define IMPL_GPU "gpu"
+#define IMPL_PLAIN "plain"
 
-    virtual void SetUp()
-    {
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-    }
-};
+#define RUN_OCL_IMPL (IMPL_OCL == getSelectedImpl())
+#define RUN_PLAIN_IMPL (IMPL_PLAIN == getSelectedImpl())
 
-TEST_F(ColumnSum, Performance)
-{
-    cv::Size size(MWIDTH, MHEIGHT);
-    cv::Mat src = randomMat(size, CV_32FC1);
-    cv::ocl::oclMat d_dst;
+#ifdef HAVE_OPENCV_GPU
+# define RUN_GPU_IMPL (IMPL_GPU == getSelectedImpl())
+#endif
 
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t1 = 0;
-    double t2 = 0;
+#ifdef HAVE_OPENCV_GPU
+#define OCL_PERF_ELSE               \
+        if (RUN_GPU_IMPL)           \
+            CV_TEST_FAIL_NO_IMPL(); \
+        else                        \
+            CV_TEST_FAIL_NO_IMPL();
+#else
+#define OCL_PERF_ELSE               \
+            CV_TEST_FAIL_NO_IMPL();
+#endif
 
-    for(int j = 0; j < LOOP_TIMES + 1; j ++)
-    {
+#define OCL_TEST_CYCLE_N(n) for(declare.iterations(n); startTimer(), next(); ocl::finish(), stopTimer())
+#define OCL_TEST_CYCLE() for(; startTimer(), next(); ocl::finish(), stopTimer())
+#define OCL_TEST_CYCLE_MULTIRUN(runsNum) for(declare.runs(runsNum); startTimer(), next(); stopTimer()) for(int r = 0; r < runsNum; ocl::finish(), ++r)
 
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat d_src(src);
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::columnSum(d_src, d_dst);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        d_dst.download (cpu_dst);//download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if(j == 0)
-            continue;
-
-        totalgputick = t1 + totalgputick;
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
-    }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
-
-}
-
-
-
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/ocl/perf/perf_pyrup.cpp b/modules/ocl/perf/perf_pyramid.cpp
similarity index 53%
rename from modules/ocl/perf/perf_pyrup.cpp
rename to modules/ocl/perf/perf_pyramid.cpp
index f2400610ad..19c728bb77 100644
--- a/modules/ocl/perf/perf_pyrup.cpp
+++ b/modules/ocl/perf/perf_pyramid.cpp
@@ -15,7 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    fangfang bai fangfang@multicorewareinc.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -30,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -42,80 +43,82 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
-#include <iomanip>
-#ifdef HAVE_OPENCL
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
+///////////// pyrDown //////////////////////
 
-PARAM_TEST_CASE(PyrUp, MatType, int)
+typedef Size_MatType pyrDownFixture;
+
+PERF_TEST_P(pyrDownFixture, pyrDown,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
 {
-    int type;
-    int channels;
-    //std::vector<cv::ocl::Info> oclinfo;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
 
-    virtual void SetUp()
+    Mat src(srcSize, type), dst;
+    Size dstSize((srcSize.height + 1) >> 1, (srcSize.width + 1) >> 1);
+    dst.create(dstSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
+        ocl::oclMat oclSrc(src), oclDst(dstSize, type);
+
+        OCL_TEST_CYCLE() ocl::pyrDown(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
     }
-};
-
-TEST_P(PyrUp, Performance)
-{
-    cv::Size size(MWIDTH, MHEIGHT);
-    cv::Mat src = randomMat(size, CV_MAKETYPE(type, channels));
-    cv::Mat dst_gold;
-    cv::ocl::oclMat dst;
-
-
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-
-    for (int j = 0; j < LOOP_TIMES + 1; j ++)
+    else if (RUN_PLAIN_IMPL)
     {
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat srcMat = cv::ocl::oclMat(src);//upload
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::pyrUp(srcMat, dst);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        dst.download(cpu_dst); //download
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if (j == 0)
-        {
-            continue;
-        }
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
+        TEST_CYCLE() pyrDown(src, dst);
 
+        SANITY_CHECK(dst);
     }
-
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-
+    else
+        OCL_PERF_ELSE
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrUp, Combine(
-                            Values(CV_8U, CV_32F), Values(1, 4)));
+///////////// pyrUp ////////////////////////
 
-#endif // HAVE_OPENCL
\ No newline at end of file
+typedef Size_MatType pyrUpFixture;
+
+PERF_TEST_P(pyrUpFixture, pyrUp,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8UC1, CV_8UC4)))
+{
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int type = get<1>(params);
+
+    Mat src(srcSize, type), dst;
+    Size dstSize(srcSize.height << 1, srcSize.width << 1);
+    dst.create(dstSize, type);
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    if (RUN_OCL_IMPL)
+    {
+        ocl::oclMat oclSrc(src), oclDst(dstSize, type);
+
+        OCL_TEST_CYCLE() ocl::pyrDown(oclSrc, oclDst);
+
+        oclDst.download(dst);
+
+        SANITY_CHECK(dst);
+    }
+    else if (RUN_PLAIN_IMPL)
+    {
+        TEST_CYCLE() pyrDown(src, dst);
+
+        SANITY_CHECK(dst);
+    }
+    else
+        OCL_PERF_ELSE
+}
diff --git a/modules/ocl/perf/perf_pyrdown.cpp b/modules/ocl/perf/perf_pyrdown.cpp
deleted file mode 100644
index 2cea4df4a3..0000000000
--- a/modules/ocl/perf/perf_pyrdown.cpp
+++ /dev/null
@@ -1,137 +0,0 @@
-///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    fangfang bai, fangfang@multicorewareinc.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#include <iomanip>
-
-#ifdef HAVE_OPENCL
-
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-
-PARAM_TEST_CASE(PyrDown, MatType, int)
-{
-    int type;
-    int channels;
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-    }
-
-
-};
-
-#define VARNAME(A) string(#A);
-
-////////////////////////////////PyrDown/////////////////////////////////////////////////
-TEST_P(PyrDown, Mat)
-{
-    cv::Size size(MWIDTH, MHEIGHT);
-    cv::RNG &rng = TS::ptr()->get_rng();
-    mat1 = randomMat(rng, size, CV_MAKETYPE(type, channels), 5, 16, false);
-
-
-    cv::ocl::oclMat gdst;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-
-    double t1 = 0;
-    double t2 = 0;
-
-    for (int j = 0; j < LOOP_TIMES + 1; j ++)
-    {
-
-        t1 = (double)cvGetTickCount();//gpu start1
-
-        cv::ocl::oclMat gmat1(mat1);
-
-        t2 = (double)cvGetTickCount(); //kernel
-        cv::ocl::pyrDown(gmat1, gdst);
-        t2 = (double)cvGetTickCount() - t2;//kernel
-
-        cv::Mat cpu_dst;
-        gdst.download(cpu_dst);
-
-        t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-        if (j == 0)
-        {
-            continue;
-        }
-
-        totalgputick = t1 + totalgputick;
-
-        totalgputick_kernel = t2 + totalgputick_kernel;
-
-    }
-
-    cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-
-}
-
-//********test****************
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrDown, Combine(
-                            Values(CV_8U, CV_32F), Values(1, 4)));
-
-
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/perf/perf_split_merge.cpp b/modules/ocl/perf/perf_split_merge.cpp
index 67a3d24aea..3821a8e163 100644
--- a/modules/ocl/perf/perf_split_merge.cpp
+++ b/modules/ocl/perf/perf_split_merge.cpp
@@ -10,12 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -30,7 +31,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -42,446 +43,99 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
+#include "perf_precomp.hpp"
 
-#include "precomp.hpp"
+using namespace perf;
+using std::tr1::tuple;
+using std::tr1::get;
 
-#ifdef HAVE_OPENCL
+///////////// Merge////////////////////////
 
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-using namespace cv::ocl;
-PARAM_TEST_CASE(MergeTestBase, MatType, int)
+typedef Size_MatType MergeFixture;
+
+PERF_TEST_P(MergeFixture, Merge,
+            ::testing::Combine(::testing::Values(OCL_SIZE_1000, OCL_SIZE_2000),
+                               OCL_PERF_ENUM(CV_8U, CV_32F)))
 {
-    int type;
-    int channels;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int depth = get<1>(params), channels = 3;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mat3;
-    cv::Mat mat4;
-
-    //dst mat
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int src3x;
-    int src3y;
-    int src4x;
-    int src4y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mat3_roi;
-    cv::Mat mat4_roi;
-
-    //dst mat with roi
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gmat2;
-    cv::ocl::oclMat gmat3;
-    cv::ocl::oclMat gmat4;
-    cv::ocl::oclMat gdst;
-
-    virtual void SetUp()
+    const int dstType = CV_MAKE_TYPE(depth, channels);
+    Mat dst(srcSize, dstType);
+    vector<Mat> src(channels);
+    for (vector<Mat>::iterator i = src.begin(), end = src.end(); i != end; ++i)
     {
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        mat2 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        mat3 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        mat4 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        dst  = randomMat(rng, size, CV_MAKETYPE(type, channels), 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
+        i->create(srcSize, CV_MAKE_TYPE(depth, 1));
+        declare.in(*i, WARMUP_RNG);
     }
-    void Has_roi(int b)
+    declare.out(dst);
+
+    if (RUN_OCL_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat1.cols - 1; //start
-            roirows = mat1.rows - 1;
-            src1x   = 1;
-            src1y   = 1;
-            src2x   = 1;
-            src2y   = 1;
-            src3x   = 1;
-            src3y   = 1;
-            src4x   = 1;
-            src4y   = 1;
-            dstx    = 1;
-            dsty    = 1;
+        ocl::oclMat oclDst(srcSize, dstType);
+        vector<ocl::oclMat> oclSrc(src.size());
+        for (vector<ocl::oclMat>::size_type i = 0, end = src.size(); i < end; ++i)
+            oclSrc[i] = src[i];
 
-        }
-        else
-        {
-            roicols = mat1.cols;
-            roirows = mat1.rows;
-            src1x   = 0;
-            src1y   = 0;
-            src2x   = 0;
-            src2y   = 0;
-            src3x   = 0;
-            src3y   = 0;
-            src4x   = 0;
-            src4y   = 0;
-            dstx    = 0;
-            dsty    = 0;
-        };
+        OCL_TEST_CYCLE() cv::ocl::merge(oclSrc, oclDst);
 
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-        mat3_roi = mat3(Rect(src3x, src3y, roicols, roirows));
-        mat4_roi = mat4(Rect(src4x, src4y, roicols, roirows));
+        oclDst.download(dst);
 
-
-        dst_roi = dst(Rect(dstx, dsty, roicols, roirows));
+        SANITY_CHECK(dst);
     }
-
-};
-
-struct Merge : MergeTestBase {};
-
-TEST_P(Merge, Accuracy)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
+    else if (RUN_PLAIN_IMPL)
     {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            std::vector<cv::Mat> dev_src;
-            dev_src.push_back(mat1_roi);
-            dev_src.push_back(mat2_roi);
-            dev_src.push_back(mat3_roi);
-            dev_src.push_back(mat4_roi);
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::merge(dev_src, dst_roi);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
+        TEST_CYCLE() cv::merge(src, dst);
 
-            t1 = (double)cvGetTickCount();//gpu start1	]
-            gmat1 = mat1_roi;
-            gmat2 = mat2_roi;
-            gmat3 = mat3_roi;
-            gmat4 = mat4_roi;
-            gdst_whole = dst;
-            gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-            std::vector<cv::ocl::oclMat> dev_gsrc;
-            dev_gsrc.push_back(gmat1);
-            dev_gsrc.push_back(gmat2);
-            dev_gsrc.push_back(gmat3);
-            dev_gsrc.push_back(gmat4);
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::merge(dev_gsrc, gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst;
-            gdst_whole.download (cpu_dst);//download
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-
-            if(j == 0)
-                continue;
-
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
+        SANITY_CHECK(dst);
     }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmat3 = mat3_roi;
-        gmat4 = mat4_roi;
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-        std::vector<cv::ocl::oclMat> dev_gsrc;
-        dev_gsrc.push_back(gmat1);
-        dev_gsrc.push_back(gmat2);
-        dev_gsrc.push_back(gmat3);
-        dev_gsrc.push_back(gmat4);
-
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::merge(dev_gsrc, gdst);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
 
+///////////// Split////////////////////////
 
-PARAM_TEST_CASE(SplitTestBase, MatType, int)
+typedef Size_MatType SplitFixture;
+
+PERF_TEST_P(SplitFixture, Split,
+            ::testing::Combine(OCL_TYPICAL_MAT_SIZES,
+                               OCL_PERF_ENUM(CV_8U, CV_32F)))
 {
-    int type;
-    int channels;
+    const Size_MatType_t params = GetParam();
+    const Size srcSize = get<0>(params);
+    const int depth = get<1>(params), channels = 3;
 
-    //src mat
-    cv::Mat mat;
+    Mat src(srcSize, CV_MAKE_TYPE(depth, channels));
+    declare.in(src, WARMUP_RNG);
 
-    //dstmat
-    cv::Mat dst1;
-    cv::Mat dst2;
-    cv::Mat dst3;
-    cv::Mat dst4;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dst1x;
-    int dst1y;
-    int dst2x;
-    int dst2y;
-    int dst3x;
-    int dst3y;
-    int dst4x;
-    int dst4y;
-
-    //src mat with roi
-    cv::Mat mat_roi;
-
-    //dst mat with roi
-    cv::Mat dst1_roi;
-    cv::Mat dst2_roi;
-    cv::Mat dst3_roi;
-    cv::Mat dst4_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst1_whole;
-    cv::ocl::oclMat gdst2_whole;
-    cv::ocl::oclMat gdst3_whole;
-    cv::ocl::oclMat gdst4_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gdst1;
-    cv::ocl::oclMat gdst2;
-    cv::ocl::oclMat gdst3;
-    cv::ocl::oclMat gdst4;
-
-    virtual void SetUp()
+    if (RUN_OCL_IMPL)
     {
-        type = GET_PARAM(0);
-        channels = GET_PARAM(1);
+        ocl::oclMat oclSrc(src);
+        vector<ocl::oclMat> oclDst(channels, ocl::oclMat(srcSize, CV_MAKE_TYPE(depth, 1)));
 
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
+        OCL_TEST_CYCLE() cv::ocl::split(oclSrc, oclDst);
 
-        mat  = randomMat(rng, size, CV_MAKETYPE(type, channels), 5, 16, false);
-        dst1 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        dst2 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        dst3 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        dst4 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //setBinpath(CLBINPATH);
+        ASSERT_EQ(3, channels);
+        Mat dst0, dst1, dst2;
+        oclDst[0].download(dst0);
+        oclDst[1].download(dst1);
+        oclDst[2].download(dst2);
+        SANITY_CHECK(dst0);
+        SANITY_CHECK(dst1);
+        SANITY_CHECK(dst2);
     }
-
-    void Has_roi(int b)
+    else if (RUN_PLAIN_IMPL)
     {
-        //cv::RNG& rng = TS::ptr()->get_rng();
-        if(b)
-        {
-            //randomize ROI
-            roicols =  mat.cols - 1; //start
-            roirows = mat.rows - 1;
-            srcx   = 1;
-            srcx   = 1;
-            dst1x    = 1;
-            dst1y    = 1;
-            dst2x    = 1;
-            dst2y    = 1;
-            dst3x    = 1;
-            dst3y    = 1;
-            dst4x    = 1;
-            dst4y    = 1;
-        }
-        else
-        {
-            roicols = mat.cols;
-            roirows = mat.rows;
-            srcx = 0;
-            srcy = 0;
-            dst1x = 0;
-            dst1y = 0;
-            dst2x    = 0;
-            dst2y    = 0;
-            dst3x    = 0;
-            dst3y    = 0;
-            dst4x    = 0;
-            dst4y    = 0;
-        };
+        vector<Mat> dst(channels, Mat(srcSize, CV_MAKE_TYPE(depth, 1)));
+        TEST_CYCLE() cv::split(src, dst);
 
-        mat_roi = mat(Rect(srcx, srcy, roicols, roirows));
-
-        dst1_roi = dst1(Rect(dst1x, dst1y, roicols, roirows));
-        dst2_roi = dst2(Rect(dst2x, dst2y, roicols, roirows));
-        dst3_roi = dst3(Rect(dst3x, dst3y, roicols, roirows));
-        dst4_roi = dst4(Rect(dst4x, dst4y, roicols, roirows));
+        ASSERT_EQ(3, channels);
+        Mat & dst0 = dst[0], & dst1 = dst[1], & dst2 = dst[2];
+        SANITY_CHECK(dst0);
+        SANITY_CHECK(dst1);
+        SANITY_CHECK(dst2);
     }
-
-};
-
-struct Split : SplitTestBase {};
-
-TEST_P(Split, Accuracy)
-{
-#ifndef PRINT_KERNEL_RUN_TIME
-    double totalcputick = 0;
-    double totalgputick = 0;
-    double totalgputick_kernel = 0;
-    double t0 = 0;
-    double t1 = 0;
-    double t2 = 0;
-    for(int k = LOOPROISTART; k < LOOPROIEND; k++)
-    {
-        totalcputick = 0;
-        totalgputick = 0;
-        totalgputick_kernel = 0;
-        for(int j = 0; j < LOOP_TIMES + 1; j ++)
-        {
-            Has_roi(k);
-            cv::Mat         dev_dst[4]  = {dst1_roi, dst2_roi, dst3_roi, dst4_roi};
-            cv::ocl::oclMat dev_gdst[4] = {gdst1, gdst2, gdst3, gdst4};
-            t0 = (double)cvGetTickCount();//cpu start
-            cv::split(mat_roi, dev_dst);
-            t0 = (double)cvGetTickCount() - t0;//cpu end
-
-            t1 = (double)cvGetTickCount();//gpu start1
-            gdst1_whole = dst1;
-            gdst1 = gdst1_whole(Rect(dst1x, dst1y, roicols, roirows));
-
-            gdst2_whole = dst2;
-            gdst2 = gdst2_whole(Rect(dst2x, dst2y, roicols, roirows));
-
-            gdst3_whole = dst3;
-            gdst3 = gdst3_whole(Rect(dst3x, dst3y, roicols, roirows));
-
-            gdst4_whole = dst4;
-            gdst4 = gdst4_whole(Rect(dst4x, dst4y, roicols, roirows));
-
-            gmat = mat_roi;
-            t2 = (double)cvGetTickCount(); //kernel
-            cv::ocl::split(gmat, dev_gdst);
-            t2 = (double)cvGetTickCount() - t2;//kernel
-            cv::Mat cpu_dst1;
-            cv::Mat cpu_dst2;
-            cv::Mat cpu_dst3;
-            cv::Mat cpu_dst4;
-            gdst1_whole.download(cpu_dst1);
-            gdst2_whole.download(cpu_dst2);
-            gdst3_whole.download(cpu_dst3);
-            gdst4_whole.download(cpu_dst4);
-            t1 = (double)cvGetTickCount() - t1;//gpu end1
-            if(j == 0)
-                continue;
-            totalgputick = t1 + totalgputick;
-            totalcputick = t0 + totalcputick;
-            totalgputick_kernel = t2 + totalgputick_kernel;
-
-        }
-        if(k == 0)
-        {
-            cout << "no roi\n";
-        }
-        else
-        {
-            cout << "with roi\n";
-        };
-        cout << "average cpu runtime is  " << totalcputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime is  " << totalgputick / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-        cout << "average gpu runtime without data transfer is  " << totalgputick_kernel / ((double)cvGetTickFrequency()* LOOP_TIMES * 1000.) << "ms" << endl;
-    }
-#else
-    for(int j = LOOPROISTART; j < LOOPROIEND; j ++)
-    {
-        Has_roi(j);
-        //cv::Mat         dev_dst[4]  = {dst1_roi, dst2_roi, dst3_roi, dst4_roi};
-        cv::ocl::oclMat dev_gdst[4] = {gdst1, gdst2, gdst3, gdst4};
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dst1x, dst1y, roicols, roirows));
-
-        gdst2_whole = dst2;
-        gdst2 = gdst2_whole(Rect(dst2x, dst2y, roicols, roirows));
-
-        gdst3_whole = dst3;
-        gdst3 = gdst3_whole(Rect(dst3x, dst3y, roicols, roirows));
-
-        gdst4_whole = dst4;
-        gdst4 = gdst4_whole(Rect(dst4x, dst4y, roicols, roirows));
-        gmat = mat_roi;
-        if(j == 0)
-        {
-            cout << "no roi:";
-        }
-        else
-        {
-            cout << "\nwith roi:";
-        };
-        cv::ocl::split(gmat, dev_gdst);
-    };
-#endif
+    else
+        OCL_PERF_ELSE
 }
-
-//*************test*****************
-INSTANTIATE_TEST_CASE_P(SplitMerge, Merge, Combine(
-                            Values(CV_8UC4, CV_32FC4), Values(1, 4)));
-
-INSTANTIATE_TEST_CASE_P(SplitMerge, Split , Combine(
-                            Values(CV_8U, CV_32S, CV_32F), Values(1, 4)));
-
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/perf/precomp.cpp b/modules/ocl/perf/precomp.cpp
deleted file mode 100644
index 7d287004ee..0000000000
--- a/modules/ocl/perf/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-
diff --git a/modules/ocl/perf/utility.cpp b/modules/ocl/perf/utility.cpp
deleted file mode 100644
index b7fbe4fa0d..0000000000
--- a/modules/ocl/perf/utility.cpp
+++ /dev/null
@@ -1,265 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#define VARNAME(A) #A
-using namespace std;
-using namespace cv;
-using namespace cv::gpu;
-using namespace cvtest;
-
-
-//std::string generateVarList(int first,...)
-//{
-//	vector<std::string> varname;
-//
-//	va_list argp;
-//	string s;
-//	stringstream ss;
-//	va_start(argp,first);
-//	int i=first;
-//	while(i!=-1)
-//	{
-//		ss<<i<<",";
-//		i=va_arg(argp,int);
-//	};
-//	s=ss.str();
-//	va_end(argp);
-//	return s;
-//};
-
-//std::string generateVarList(int& p1,int& p2)
-//{
-//	stringstream ss;
-//	ss<<VARNAME(p1)<<":"<<src1x<<","<<VARNAME(p2)<<":"<<src1y;
-//	return ss.str();
-//};
-
-int randomInt(int minVal, int maxVal)
-{
-    RNG &rng = TS::ptr()->get_rng();
-    return rng.uniform(minVal, maxVal);
-}
-
-double randomDouble(double minVal, double maxVal)
-{
-    RNG &rng = TS::ptr()->get_rng();
-    return rng.uniform(minVal, maxVal);
-}
-
-Size randomSize(int minVal, int maxVal)
-{
-    return cv::Size(randomInt(minVal, maxVal), randomInt(minVal, maxVal));
-}
-
-Scalar randomScalar(double minVal, double maxVal)
-{
-    return Scalar(randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal));
-}
-
-Mat randomMat(Size size, int type, double minVal, double maxVal)
-{
-    return randomMat(TS::ptr()->get_rng(), size, type, minVal, maxVal, false);
-}
-
-
-
-
-
-
-
-/*
-void showDiff(InputArray gold_, InputArray actual_, double eps)
-{
-    Mat gold;
-    if (gold_.kind() == _InputArray::MAT)
-        gold = gold_.getMat();
-    else
-        gold_.getGpuMat().download(gold);
-
-    Mat actual;
-    if (actual_.kind() == _InputArray::MAT)
-        actual = actual_.getMat();
-    else
-        actual_.getGpuMat().download(actual);
-
-    Mat diff;
-    absdiff(gold, actual, diff);
-    threshold(diff, diff, eps, 255.0, cv::THRESH_BINARY);
-
-    namedWindow("gold", WINDOW_NORMAL);
-    namedWindow("actual", WINDOW_NORMAL);
-    namedWindow("diff", WINDOW_NORMAL);
-
-    imshow("gold", gold);
-    imshow("actual", actual);
-    imshow("diff", diff);
-
-    waitKey();
-}
-*/
-
-/*
-bool supportFeature(const DeviceInfo& info, FeatureSet feature)
-{
-    return TargetArchs::builtWith(feature) && info.supports(feature);
-}
-
-const vector<DeviceInfo>& devices()
-{
-    static vector<DeviceInfo> devs;
-    static bool first = true;
-
-    if (first)
-    {
-        int deviceCount = getCudaEnabledDeviceCount();
-
-        devs.reserve(deviceCount);
-
-        for (int i = 0; i < deviceCount; ++i)
-        {
-            DeviceInfo info(i);
-            if (info.isCompatible())
-                devs.push_back(info);
-        }
-
-        first = false;
-    }
-
-    return devs;
-}
-
-vector<DeviceInfo> devices(FeatureSet feature)
-{
-    const vector<DeviceInfo>& d = devices();
-
-    vector<DeviceInfo> devs_filtered;
-
-    if (TargetArchs::builtWith(feature))
-    {
-        devs_filtered.reserve(d.size());
-
-        for (size_t i = 0, size = d.size(); i < size; ++i)
-        {
-            const DeviceInfo& info = d[i];
-
-            if (info.supports(feature))
-                devs_filtered.push_back(info);
-        }
-    }
-
-    return devs_filtered;
-}
-*/
-
-vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end)
-{
-    vector<MatType> v;
-
-    v.reserve((depth_end - depth_start + 1) * (cn_end - cn_start + 1));
-
-    for (int depth = depth_start; depth <= depth_end; ++depth)
-    {
-        for (int cn = cn_start; cn <= cn_end; ++cn)
-        {
-            v.push_back(CV_MAKETYPE(depth, cn));
-        }
-    }
-
-    return v;
-}
-
-const vector<MatType> &all_types()
-{
-    static vector<MatType> v = types(CV_8U, CV_64F, 1, 4);
-
-    return v;
-}
-
-Mat readImage(const string &fileName, int flags)
-{
-    return imread(string(cvtest::TS::ptr()->get_data_path()) + fileName, flags);
-}
-
-Mat readImageType(const string &fname, int type)
-{
-    Mat src = readImage(fname, CV_MAT_CN(type) == 1 ? IMREAD_GRAYSCALE : IMREAD_COLOR);
-    if (CV_MAT_CN(type) == 4)
-    {
-        Mat temp;
-        cvtColor(src, temp, cv::COLOR_BGR2BGRA);
-        swap(src, temp);
-    }
-    src.convertTo(src, CV_MAT_DEPTH(type));
-    return src;
-}
-
-double checkNorm(const Mat &m)
-{
-    return norm(m, NORM_INF);
-}
-
-double checkNorm(const Mat &m1, const Mat &m2)
-{
-    return norm(m1, m2, NORM_INF);
-}
-
-double checkSimilarity(const Mat &m1, const Mat &m2)
-{
-    Mat diff;
-    matchTemplate(m1, m2, diff, CV_TM_CCORR_NORMED);
-    return std::abs(diff.at<float>(0, 0) - 1.f);
-}
-
-/*
-void cv::ocl::PrintTo(const DeviceInfo& info, ostream* os)
-{
-    (*os) << info.name();
-}
-*/
-
-void PrintTo(const Inverse &inverse, std::ostream *os)
-{
-    if (inverse)
-        (*os) << "inverse";
-    else
-        (*os) << "direct";
-}
diff --git a/modules/ocl/perf/utility.hpp b/modules/ocl/perf/utility.hpp
deleted file mode 100644
index 7d34b6731a..0000000000
--- a/modules/ocl/perf/utility.hpp
+++ /dev/null
@@ -1,182 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_TEST_UTILITY_HPP__
-#define __OPENCV_TEST_UTILITY_HPP__
-//#define PRINT_KERNEL_RUN_TIME
-#ifdef PRINT_KERNEL_RUN_TIME
-#define LOOP_TIMES 1
-#else
-#define LOOP_TIMES 1
-#endif
-#define MWIDTH 1920
-#define MHEIGHT 1080
-#define CLBINPATH ".\\"
-#define LOOPROISTART 0
-#define LOOPROIEND 1
-int randomInt(int minVal, int maxVal);
-double randomDouble(double minVal, double maxVal);
-
-//std::string generateVarList(int first,...);
-std::string generateVarList(int &p1, int &p2);
-cv::Size randomSize(int minVal, int maxVal);
-cv::Scalar randomScalar(double minVal, double maxVal);
-cv::Mat randomMat(cv::Size size, int type, double minVal = 0.0, double maxVal = 255.0);
-
-void showDiff(cv::InputArray gold, cv::InputArray actual, double eps);
-
-//! return true if device supports specified feature and gpu module was built with support the feature.
-//bool supportFeature(const cv::gpu::DeviceInfo& info, cv::gpu::FeatureSet feature);
-
-//! return all devices compatible with current gpu module build.
-//const std::vector<cv::ocl::DeviceInfo>& devices();
-//! return all devices compatible with current gpu module build which support specified feature.
-//std::vector<cv::ocl::DeviceInfo> devices(cv::gpu::FeatureSet feature);
-
-//! read image from testdata folder.
-cv::Mat readImage(const std::string &fileName, int flags = cv::IMREAD_COLOR);
-cv::Mat readImageType(const std::string &fname, int type);
-
-double checkNorm(const cv::Mat &m);
-double checkNorm(const cv::Mat &m1, const cv::Mat &m2);
-double checkSimilarity(const cv::Mat &m1, const cv::Mat &m2);
-
-#define EXPECT_MAT_NORM(mat, eps) \
-{ \
-    EXPECT_LE(checkNorm(cv::Mat(mat)), eps) \
-}
-
-/*#define EXPECT_MAT_NEAR(mat1, mat2, eps) \
-{ \
-   ASSERT_EQ(mat1.type(), mat2.type()); \
-   ASSERT_EQ(mat1.size(), mat2.size()); \
-   EXPECT_LE(checkNorm(cv::Mat(mat1), cv::Mat(mat2)), eps); \
-}*/
-
-#define EXPECT_MAT_NEAR(mat1, mat2, eps,s) \
-{ \
-    ASSERT_EQ(mat1.type(), mat2.type()); \
-    ASSERT_EQ(mat1.size(), mat2.size()); \
-    EXPECT_LE(checkNorm(cv::Mat(mat1), cv::Mat(mat2)), eps)<<s; \
-}
-
-#define EXPECT_MAT_SIMILAR(mat1, mat2, eps) \
-{ \
-    ASSERT_EQ(mat1.type(), mat2.type()); \
-    ASSERT_EQ(mat1.size(), mat2.size()); \
-    EXPECT_LE(checkSimilarity(cv::Mat(mat1), cv::Mat(mat2)), eps); \
-}
-
-namespace cv
-{
-    namespace ocl
-    {
-        // void PrintTo(const DeviceInfo& info, std::ostream* os);
-    }
-}
-
-using perf::MatDepth;
-using perf::MatType;
-
-//! return vector with types from specified range.
-std::vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end);
-
-//! return vector with all types (depth: CV_8U-CV_64F, channels: 1-4).
-const std::vector<MatType> &all_types();
-
-class Inverse
-{
-public:
-    inline Inverse(bool val = false) : val_(val) {}
-
-    inline operator bool() const
-    {
-        return val_;
-    }
-
-private:
-    bool val_;
-};
-
-void PrintTo(const Inverse &useRoi, std::ostream *os);
-
-CV_ENUM(CmpCode, cv::CMP_EQ, cv::CMP_GT, cv::CMP_GE, cv::CMP_LT, cv::CMP_LE, cv::CMP_NE)
-
-CV_ENUM(NormCode, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_TYPE_MASK, cv::NORM_RELATIVE, cv::NORM_MINMAX)
-
-enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
-CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
-
-CV_ENUM(ReduceOp, CV_REDUCE_SUM, CV_REDUCE_AVG, CV_REDUCE_MAX, CV_REDUCE_MIN)
-
-CV_FLAGS(GemmFlags, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T);
-
-CV_ENUM(MorphOp, cv::MORPH_OPEN, cv::MORPH_CLOSE, cv::MORPH_GRADIENT, cv::MORPH_TOPHAT, cv::MORPH_BLACKHAT)
-
-CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
-
-CV_ENUM(Interpolation, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC)
-
-CV_ENUM(Border, cv::BORDER_REFLECT101, cv::BORDER_REPLICATE, cv::BORDER_CONSTANT, cv::BORDER_REFLECT, cv::BORDER_WRAP)
-
-CV_FLAGS(WarpFlags, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::WARP_INVERSE_MAP)
-
-CV_ENUM(TemplateMethod, cv::TM_SQDIFF, cv::TM_SQDIFF_NORMED, cv::TM_CCORR, cv::TM_CCORR_NORMED, cv::TM_CCOEFF, cv::TM_CCOEFF_NORMED)
-
-CV_FLAGS(DftFlags, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
-
-void  run_perf_test();
-
-#define PARAM_TEST_CASE(name, ...) struct name : testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
-
-#define GET_PARAM(k) std::tr1::get< k >(GetParam())
-
-#define ALL_DEVICES testing::ValuesIn(devices())
-#define DEVICES(feature) testing::ValuesIn(devices(feature))
-
-#define ALL_TYPES testing::ValuesIn(all_types())
-#define TYPES(depth_start, depth_end, cn_start, cn_end) testing::ValuesIn(types(depth_start, depth_end, cn_start, cn_end))
-
-#define DIFFERENT_SIZES testing::Values(cv::Size(128, 128), cv::Size(113, 113))
-
-#define DIRECT_INVERSE testing::Values(Inverse(false), Inverse(true))
-
-#endif // __OPENCV_TEST_UTILITY_HPP__
diff --git a/modules/ocl/src/arithm.cpp b/modules/ocl/src/arithm.cpp
index 232688895a..c9ee6f9f6f 100644
--- a/modules/ocl/src/arithm.cpp
+++ b/modules/ocl/src/arithm.cpp
@@ -22,6 +22,7 @@
 //    Jiang Liyuan, jlyuan001.good@163.com
 //    Rock Li, Rock.Li@amd.com
 //    Zailong Wu, bullet@yeah.net
+//    Peng Xiao, pengxiao@outlook.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -61,8 +62,6 @@ namespace cv
     namespace ocl
     {
         ////////////////////////////////OpenCL kernel strings/////////////////////
-        extern const char *bitwise;
-        extern const char *bitwiseM;
         extern const char *transpose_kernel;
         extern const char *arithm_nonzero;
         extern const char *arithm_sum;
@@ -76,24 +75,13 @@ namespace cv
         extern const char *arithm_add;
         extern const char *arithm_add_scalar;
         extern const char *arithm_add_scalar_mask;
+        extern const char *arithm_bitwise_binary;
+        extern const char *arithm_bitwise_binary_mask;
+        extern const char *arithm_bitwise_binary_scalar;
+        extern const char *arithm_bitwise_binary_scalar_mask;
         extern const char *arithm_bitwise_not;
-        extern const char *arithm_bitwise_and;
-        extern const char *arithm_bitwise_and_mask;
-        extern const char *arithm_bitwise_and_scalar;
-        extern const char *arithm_bitwise_and_scalar_mask;
-        extern const char *arithm_bitwise_or;
-        extern const char *arithm_bitwise_or_mask;
-        extern const char *arithm_bitwise_or_scalar;
-        extern const char *arithm_bitwise_or_scalar_mask;
-        extern const char *arithm_bitwise_xor;
-        extern const char *arithm_bitwise_xor_mask;
-        extern const char *arithm_bitwise_xor_scalar;
-        extern const char *arithm_bitwise_xor_scalar_mask;
         extern const char *arithm_compare_eq;
         extern const char *arithm_compare_ne;
-        extern const char *arithm_sub;
-        extern const char *arithm_sub_scalar;
-        extern const char *arithm_sub_scalar_mask;
         extern const char *arithm_mul;
         extern const char *arithm_div;
         extern const char *arithm_absdiff;
@@ -129,11 +117,12 @@ inline int divUp(int total, int grain)
 /////////////////////// add subtract multiply divide /////////////////////////
 //////////////////////////////////////////////////////////////////////////////
 template<typename T>
-void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString, void *_scalar)
+void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
+                    String kernelName, const char **kernelString, void *_scalar, int op_type = 0)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -185,18 +174,29 @@ void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::st
         scalar = (T)scalar1;
         args.push_back( std::make_pair( sizeof(T), (void *)&scalar ));
     }
-
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
-}
-static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString)
-{
-    arithmetic_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL);
-}
-static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString)
-{
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    switch(op_type)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        case MAT_ADD:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, "-D ARITHM_ADD");
+            break;
+        case MAT_SUB:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, "-D ARITHM_SUB");
+            break;
+        default:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
+    }
+}
+static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
+                           String kernelName, const char **kernelString, int op_type = 0)
+{
+    arithmetic_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL, op_type);
+}
+static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask,
+                           String kernelName, const char **kernelString, int op_type = 0)
+{
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
+    {
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -247,50 +247,61 @@ static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
 
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
+    switch (op_type)
+    {
+        case MAT_ADD:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_ADD");
+            break;
+        case MAT_SUB:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_SUB");
+            break;
+        default:
+            openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
+    }
 }
 void cv::ocl::add(const oclMat &src1, const oclMat &src2, oclMat &dst)
 {
-    arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add);
+    arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add, MAT_ADD);
 }
 void cv::ocl::add(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
 {
-    arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add);
+    arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add, MAT_ADD);
 }
 
 void cv::ocl::subtract(const oclMat &src1, const oclMat &src2, oclMat &dst)
 {
-    arithmetic_run(src1, src2, dst, "arithm_sub", &arithm_sub);
+    arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add, MAT_SUB);
 }
 void cv::ocl::subtract(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
 {
-    arithmetic_run(src1, src2, dst, mask, "arithm_sub_with_mask", &arithm_sub);
+    arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add, MAT_SUB);
 }
-typedef void (*MulDivFunc)(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName,
+typedef void (*MulDivFunc)(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName,
                            const char **kernelString, void *scalar);
 
 void cv::ocl::multiply(const oclMat &src1, const oclMat &src2, oclMat &dst, double scalar)
 {
-    if((src1.clCxt -> impl -> double_support != 0) && (src1.depth() == CV_64F))
+    if(src1.clCxt->supportsFeature(Context::CL_DOUBLE) && (src1.depth() == CV_64F))
         arithmetic_run<double>(src1, src2, dst, "arithm_mul", &arithm_mul, (void *)(&scalar));
     else
         arithmetic_run<float>(src1, src2, dst, "arithm_mul", &arithm_mul, (void *)(&scalar));
 }
+
 void cv::ocl::divide(const oclMat &src1, const oclMat &src2, oclMat &dst, double scalar)
 {
 
-    if(src1.clCxt -> impl -> double_support != 0)
+    if(src1.clCxt->supportsFeature(Context::CL_DOUBLE))
         arithmetic_run<double>(src1, src2, dst, "arithm_div", &arithm_div, (void *)(&scalar));
     else
         arithmetic_run<float>(src1, src2, dst, "arithm_div", &arithm_div, (void *)(&scalar));
 
 }
 template <typename WT , typename CL_WT>
-void arithmetic_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar)
+void arithmetic_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -350,19 +361,16 @@ void arithmetic_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst,
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step1 ));
     if(isMatSubScalar != 0)
-    {
-        isMatSubScalar = isMatSubScalar > 0 ? 1 : 0;
-        args.push_back( std::make_pair( sizeof(cl_int) , (void *)&isMatSubScalar));
-    }
-
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
+        openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_SUB");
+    else
+        openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_ADD");
 }
 
-static void arithmetic_scalar_run(const oclMat &src, oclMat &dst, std::string kernelName, const char **kernelString, double scalar)
+static void arithmetic_scalar_run(const oclMat &src, oclMat &dst, String kernelName, const char **kernelString, double scalar)
 {
-    if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -404,21 +412,21 @@ static void arithmetic_scalar_run(const oclMat &src, oclMat &dst, std::string ke
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
 
-    if(src.clCxt -> impl -> double_support != 0)
+    float f_scalar = (float)scalar;
+    if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
         args.push_back( std::make_pair( sizeof(cl_double), (void *)&scalar ));
     else
     {
-        float f_scalar = (float)scalar;
         args.push_back( std::make_pair( sizeof(cl_float), (void *)&f_scalar));
     }
 
     openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
 }
 
-typedef void (*ArithmeticFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar);
+typedef void (*ArithmeticFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar);
 
 
-static void arithmetic_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar)
+static void arithmetic_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar)
 {
     static ArithmeticFuncS tab[8] =
     {
@@ -433,17 +441,17 @@ static void arithmetic_scalar(const oclMat &src1, const Scalar &src2, oclMat &ds
     };
     ArithmeticFuncS func = tab[src1.depth()];
     if(func == 0)
-        cv::ocl::error("Unsupported arithmetic operation", __FILE__, __LINE__);
+        cv::error(Error::StsBadArg, "Unsupported arithmetic operation", "", __FILE__, __LINE__);
     func(src1, src2, dst, mask, kernelName, kernelString, isMatSubScalar);
 }
-static void arithmetic_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString)
+static void arithmetic_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, String kernelName, const char **kernelString)
 {
     arithmetic_scalar(src1, src2, dst, mask, kernelName, kernelString, 0);
 }
 
 void cv::ocl::add(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
 {
-    std::string kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add";
+    String kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add";
     const char **kernelString = mask.data ? &arithm_add_scalar_mask : &arithm_add_scalar;
 
     arithmetic_scalar( src1, src2, dst, mask, kernelName, kernelString);
@@ -451,25 +459,32 @@ void cv::ocl::add(const oclMat &src1, const Scalar &src2, oclMat &dst, const ocl
 
 void cv::ocl::subtract(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
 {
-    std::string kernelName = mask.data ? "arithm_s_sub_with_mask" : "arithm_s_sub";
-    const char **kernelString = mask.data ? &arithm_sub_scalar_mask : &arithm_sub_scalar;
+    String kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add";
+    const char **kernelString = mask.data ? &arithm_add_scalar_mask : &arithm_add_scalar;
+
     arithmetic_scalar( src1, src2, dst, mask, kernelName, kernelString, 1);
 }
 void cv::ocl::subtract(const Scalar &src2, const oclMat &src1, oclMat &dst, const oclMat &mask)
 {
-    std::string kernelName = mask.data ? "arithm_s_sub_with_mask" : "arithm_s_sub";
-    const char **kernelString = mask.data ? &arithm_sub_scalar_mask : &arithm_sub_scalar;
+    String kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add";
+    const char **kernelString = mask.data ? &arithm_add_scalar_mask : &arithm_add_scalar;
+
     arithmetic_scalar( src1, src2, dst, mask, kernelName, kernelString, -1);
 }
+void cv::ocl::multiply(double scalar, const oclMat &src, oclMat &dst)
+{
+    String kernelName = "arithm_muls";
+    arithmetic_scalar_run( src, dst, kernelName, &arithm_mul, scalar);
+}
 void cv::ocl::divide(double scalar, const oclMat &src,  oclMat &dst)
 {
-    if(src.clCxt -> impl -> double_support == 0)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE))
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
-    std::string kernelName =  "arithm_s_div";
+    String kernelName =  "arithm_s_div";
     arithmetic_scalar_run(src, dst, kernelName, &arithm_div, scalar);
 }
 //////////////////////////////////////////////////////////////////////////////
@@ -481,14 +496,14 @@ void cv::ocl::absdiff(const oclMat &src1, const oclMat &src2, oclMat &dst)
 }
 void cv::ocl::absdiff(const oclMat &src1, const Scalar &src2, oclMat &dst)
 {
-    std::string kernelName = "arithm_s_absdiff";
+    String kernelName = "arithm_s_absdiff";
     oclMat mask;
     arithmetic_scalar( src1, src2, dst, mask, kernelName, &arithm_absdiff);
 }
 //////////////////////////////////////////////////////////////////////////////
 /////////////////////////////////  compare ///////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void compare_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString)
+static void compare_run(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName, const char **kernelString)
 {
     dst.create(src1.size(), CV_8UC1);
     CV_Assert(src1.oclchannels() == 1);
@@ -523,12 +538,12 @@ static void compare_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std
 
 void cv::ocl::compare(const oclMat &src1, const oclMat &src2, oclMat &dst , int cmpOp)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    std::string kernelName;
+    String kernelName;
     const char **kernelString = NULL;
     switch( cmpOp )
     {
@@ -557,7 +572,7 @@ void cv::ocl::compare(const oclMat &src1, const oclMat &src2, oclMat &dst , int
         kernelString = &arithm_compare_ne;
         break;
     default:
-        CV_Error(CV_StsBadArg, "Unknown comparison method");
+        CV_Error(Error::StsBadArg, "Unknown comparison method");
     }
     compare_run(src1, src2, dst, kernelName, kernelString);
 }
@@ -598,7 +613,7 @@ static void arithmetic_sum_buffer_run(const oclMat &src, cl_mem &dst, int vlen ,
 template <typename T>
 Scalar arithmetic_sum(const oclMat &src, int type = 0)
 {
-    size_t groupnum = src.clCxt->impl->maxComputeUnits;
+    size_t groupnum = src.clCxt->computeUnits();
     CV_Assert(groupnum != 0);
     int vlen = src.oclchannels() == 3 ? 12 : 8, dbsize = groupnum * vlen;
     Context *clCxt = src.clCxt;
@@ -626,9 +641,9 @@ Scalar arithmetic_sum(const oclMat &src, int type = 0)
 typedef Scalar (*sumFunc)(const oclMat &src, int type);
 Scalar cv::ocl::sum(const oclMat &src)
 {
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     static sumFunc functab[2] =
     {
@@ -637,15 +652,15 @@ Scalar cv::ocl::sum(const oclMat &src)
     };
 
     sumFunc func;
-    func = functab[src.clCxt->impl->double_support];
+    func = functab[(int)src.clCxt->supportsFeature(Context::CL_DOUBLE)];
     return func(src, 0);
 }
 
 Scalar cv::ocl::absSum(const oclMat &src)
 {
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     static sumFunc functab[2] =
     {
@@ -654,15 +669,15 @@ Scalar cv::ocl::absSum(const oclMat &src)
     };
 
     sumFunc func;
-    func = functab[src.clCxt->impl->double_support];
+    func = functab[(int)src.clCxt->supportsFeature(Context::CL_DOUBLE)];
     return func(src, 1);
 }
 
 Scalar cv::ocl::sqrSum(const oclMat &src)
 {
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     static sumFunc functab[2] =
     {
@@ -671,7 +686,7 @@ Scalar cv::ocl::sqrSum(const oclMat &src)
     };
 
     sumFunc func;
-    func = functab[src.clCxt->impl->double_support];
+    func = functab[(int)src.clCxt->supportsFeature(Context::CL_DOUBLE)];
     return func(src, 2);
 }
 //////////////////////////////////////////////////////////////////////////////
@@ -700,7 +715,7 @@ void cv::ocl::meanStdDev(const oclMat &src, Scalar &mean, Scalar &stddev)
 //////////////////////////////////////////////////////////////////////////////
 //////////////////////////////////// minMax  /////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_minMax_run(const oclMat &src, const oclMat &mask, cl_mem &dst, int vlen , int groupnum, std::string kernelName)
+static void arithmetic_minMax_run(const oclMat &src, const oclMat &mask, cl_mem &dst, int vlen , int groupnum, String kernelName)
 {
     std::vector<std::pair<size_t , const void *> > args;
     int all_cols = src.step / (vlen * src.elemSize1());
@@ -737,7 +752,7 @@ static void arithmetic_minMax_run(const oclMat &src, const oclMat &mask, cl_mem
 }
 
 
-static void arithmetic_minMax_mask_run(const oclMat &src, const oclMat &mask, cl_mem &dst, int vlen, int groupnum, std::string kernelName)
+static void arithmetic_minMax_mask_run(const oclMat &src, const oclMat &mask, cl_mem &dst, int vlen, int groupnum, String kernelName)
 {
     std::vector<std::pair<size_t , const void *> > args;
     size_t gt[3] = {groupnum * 256, 1, 1}, lt[3] = {256, 1, 1};
@@ -768,50 +783,60 @@ static void arithmetic_minMax_mask_run(const oclMat &src, const oclMat &mask, cl
     }
 }
 
-template <typename T> void arithmetic_minMax(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask)
+template <typename T> void arithmetic_minMax(const oclMat &src, double *minVal, double *maxVal,
+                                             const oclMat &mask, oclMat &buf)
 {
-    size_t groupnum = src.clCxt->impl->maxComputeUnits;
+    size_t groupnum = src.clCxt->computeUnits();
     CV_Assert(groupnum != 0);
     groupnum = groupnum * 2;
     int vlen = 8;
     int dbsize = groupnum * 2 * vlen * sizeof(T) ;
-    Context *clCxt = src.clCxt;
-    cl_mem dstBuffer = openCLCreateBuffer(clCxt, CL_MEM_WRITE_ONLY, dbsize);
-    *minVal = std::numeric_limits<double>::max() , *maxVal = -std::numeric_limits<double>::max();
+
+    ensureSizeIsEnough(1, dbsize, CV_8UC1, buf);
+
+    cl_mem buf_data = reinterpret_cast<cl_mem>(buf.data);
+
     if (mask.empty())
     {
-        arithmetic_minMax_run(src, mask, dstBuffer, vlen, groupnum, "arithm_op_minMax");
+        arithmetic_minMax_run(src, mask, buf_data, vlen, groupnum, "arithm_op_minMax");
     }
     else
     {
-        arithmetic_minMax_mask_run(src, mask, dstBuffer, vlen, groupnum, "arithm_op_minMax_mask");
+        arithmetic_minMax_mask_run(src, mask, buf_data, vlen, groupnum, "arithm_op_minMax_mask");
     }
-    T *p = new T[groupnum * vlen * 2];
-    memset(p, 0, dbsize);
-    openCLReadBuffer(clCxt, dstBuffer, (void *)p, dbsize);
-    if(minVal != NULL){
+
+    Mat matbuf = Mat(buf);
+    T *p = matbuf.ptr<T>();
+    if(minVal != NULL)
+    {
+        *minVal = std::numeric_limits<double>::max();
         for(int i = 0; i < vlen * (int)groupnum; i++)
         {
             *minVal = *minVal < p[i] ? *minVal : p[i];
         }
     }
-    if(maxVal != NULL){
+    if(maxVal != NULL)
+    {
+        *maxVal = -std::numeric_limits<double>::max();
         for(int i = vlen * (int)groupnum; i < 2 * vlen * (int)groupnum; i++)
         {
             *maxVal = *maxVal > p[i] ? *maxVal : p[i];
         }
     }
-    delete[] p;
-    openCLFree(dstBuffer);
 }
 
-typedef void (*minMaxFunc)(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask);
+typedef void (*minMaxFunc)(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask, oclMat &buf);
 void cv::ocl::minMax(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask)
+{
+    oclMat buf;
+    minMax_buf(src, minVal, maxVal, mask, buf);
+}
+void cv::ocl::minMax_buf(const oclMat &src, double *minVal, double *maxVal, const oclMat &mask, oclMat &buf)
 {
     CV_Assert(src.oclchannels() == 1);
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     static minMaxFunc functab[8] =
     {
@@ -826,7 +851,7 @@ void cv::ocl::minMax(const oclMat &src, double *minVal, double *maxVal, const oc
     };
     minMaxFunc func;
     func = functab[src.depth()];
-    func(src, minVal, maxVal, mask);
+    func(src, minVal, maxVal, mask, buf);
 }
 
 //////////////////////////////////////////////////////////////////////////////
@@ -891,11 +916,11 @@ double cv::ocl::norm(const oclMat &src1, const oclMat &src2, int normType)
 //////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////// flip //////////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_flip_rows_run(const oclMat &src, oclMat &dst, std::string kernelName)
+static void arithmetic_flip_rows_run(const oclMat &src, oclMat &dst, String kernelName)
 {
-    if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -940,11 +965,11 @@ static void arithmetic_flip_rows_run(const oclMat &src, oclMat &dst, std::string
 
     openCLExecuteKernel(clCxt, &arithm_flip, kernelName, globalThreads, localThreads, args, -1, depth);
 }
-static void arithmetic_flip_cols_run(const oclMat &src, oclMat &dst, std::string kernelName, bool isVertical)
+static void arithmetic_flip_cols_run(const oclMat &src, oclMat &dst, String kernelName, bool isVertical)
 {
-    if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1011,7 +1036,7 @@ void cv::ocl::flip(const oclMat &src, oclMat &dst, int flipCode)
 //////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////// LUT  //////////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_lut_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName)
+static void arithmetic_lut_run(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName)
 {
     Context *clCxt = src1.clCxt;
     int channels = src1.oclchannels();
@@ -1105,14 +1130,14 @@ void cv::ocl::LUT(const oclMat &src, const oclMat &lut, oclMat &dst)
     CV_Assert((lut.oclchannels() == 1 || lut.oclchannels() == cn) && lut.rows == 1 && lut.cols == 256);
     dst.create(src.size(), CV_MAKETYPE(lut.depth(), cn));
     //oclMat _lut(lut);
-    std::string kernelName = "LUT";
+    String kernelName = "LUT";
     arithmetic_lut_run(src, lut, dst, kernelName);
 }
 
 //////////////////////////////////////////////////////////////////////////////
 //////////////////////////////// exp log /////////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_exp_log_run(const oclMat &src, oclMat &dst, std::string kernelName, const char **kernelString)
+static void arithmetic_exp_log_run(const oclMat &src, oclMat &dst, String kernelName, const char **kernelString)
 {
     dst.create(src.size(), src.type());
     CV_Assert(src.cols == dst.cols &&
@@ -1122,9 +1147,9 @@ static void arithmetic_exp_log_run(const oclMat &src, oclMat &dst, std::string k
     CV_Assert( src.type() == CV_32F || src.type() == CV_64F);
 
     Context  *clCxt = src.clCxt;
-    if(clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
     //int channels = dst.oclchannels();
@@ -1161,11 +1186,11 @@ void cv::ocl::log(const oclMat &src, oclMat &dst)
 //////////////////////////////////////////////////////////////////////////////
 ////////////////////////////// magnitude phase ///////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_magnitude_phase_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName)
+static void arithmetic_magnitude_phase_run(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1209,11 +1234,11 @@ void cv::ocl::magnitude(const oclMat &src1, const oclMat &src2, oclMat &dst)
     arithmetic_magnitude_phase_run(src1, src2, dst, "arithm_magnitude");
 }
 
-static void arithmetic_phase_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString)
+static void arithmetic_phase_run(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName, const char **kernelString)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1256,7 +1281,7 @@ void cv::ocl::phase(const oclMat &x, const oclMat &y, oclMat &Angle , bool angle
 {
     CV_Assert(x.type() == y.type() && x.size() == y.size() && (x.depth() == CV_32F || x.depth() == CV_64F));
     Angle.create(x.size(), x.type());
-    std::string kernelName = angleInDegrees ? "arithm_phase_indegrees" : "arithm_phase_inradians";
+    String kernelName = angleInDegrees ? "arithm_phase_indegrees" : "arithm_phase_inradians";
     if(angleInDegrees)
     {
         arithmetic_phase_run(x, y, Angle, kernelName, &arithm_phase);
@@ -1273,11 +1298,11 @@ void cv::ocl::phase(const oclMat &x, const oclMat &y, oclMat &Angle , bool angle
 ////////////////////////////////// cartToPolar ///////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
 static void arithmetic_cartToPolar_run(const oclMat &src1, const oclMat &src2, oclMat &dst_mag, oclMat &dst_cart,
-                                std::string kernelName, bool angleInDegrees)
+                                String kernelName, bool angleInDegrees)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1328,11 +1353,11 @@ void cv::ocl::cartToPolar(const oclMat &x, const oclMat &y, oclMat &mag, oclMat
 ////////////////////////////////// polarToCart ///////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
 static void arithmetic_ptc_run(const oclMat &src1, const oclMat &src2, oclMat &dst1, oclMat &dst2, bool angleInDegrees,
-                        std::string kernelName)
+                        String kernelName)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1451,7 +1476,7 @@ void arithmetic_minMaxLoc(const oclMat &src, double *minVal, double *maxVal,
                           Point *minLoc, Point *maxLoc, const oclMat &mask)
 {
     CV_Assert(src.oclchannels() == 1);
-    size_t groupnum = src.clCxt->impl->maxComputeUnits;
+    size_t groupnum = src.clCxt->computeUnits();
     CV_Assert(groupnum != 0);
     int minloc = -1 , maxloc = -1;
     int vlen = 4, dbsize = groupnum * vlen * 4 * sizeof(T) ;
@@ -1512,9 +1537,9 @@ typedef void (*minMaxLocFunc)(const oclMat &src, double *minVal, double *maxVal,
 void cv::ocl::minMaxLoc(const oclMat &src, double *minVal, double *maxVal,
                         Point *minLoc, Point *maxLoc, const oclMat &mask)
 {
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     static minMaxLocFunc functab[2] =
     {
@@ -1523,14 +1548,14 @@ void cv::ocl::minMaxLoc(const oclMat &src, double *minVal, double *maxVal,
     };
 
     minMaxLocFunc func;
-    func = functab[src.clCxt->impl->double_support];
+    func = functab[(int)src.clCxt->supportsFeature(Context::CL_DOUBLE)];
     func(src, minVal, maxVal, minLoc, maxLoc, mask);
 }
 
 //////////////////////////////////////////////////////////////////////////////
 ///////////////////////////// countNonZero ///////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void arithmetic_countNonZero_run(const oclMat &src, cl_mem &dst, int vlen , int groupnum, std::string kernelName)
+static void arithmetic_countNonZero_run(const oclMat &src, cl_mem &dst, int vlen , int groupnum, String kernelName)
 {
     std::vector<std::pair<size_t , const void *> > args;
     int all_cols = src.step / (vlen * src.elemSize1());
@@ -1558,17 +1583,17 @@ static void arithmetic_countNonZero_run(const oclMat &src, cl_mem &dst, int vlen
 
 int cv::ocl::countNonZero(const oclMat &src)
 {
-    size_t groupnum = src.clCxt->impl->maxComputeUnits;
-    if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+    size_t groupnum = src.clCxt->computeUnits();
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "select device don't support double");
+        CV_Error(Error::GpuNotSupported, "select device don't support double");
     }
     CV_Assert(groupnum != 0);
     groupnum = groupnum * 2;
     int vlen = 8 , dbsize = groupnum * vlen;
     //cl_ulong start, end;
     Context *clCxt = src.clCxt;
-    std::string kernelName = "arithm_op_nonzero";
+    String kernelName = "arithm_op_nonzero";
     int *p = new int[dbsize], nonzero = 0;
     cl_mem dstBuffer = openCLCreateBuffer(clCxt, CL_MEM_WRITE_ONLY, dbsize * sizeof(int));
     arithmetic_countNonZero_run(src, dstBuffer, vlen, groupnum, kernelName);
@@ -1587,7 +1612,7 @@ int cv::ocl::countNonZero(const oclMat &src)
 //////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////bitwise_op////////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
-static void bitwise_run(const oclMat &src1, oclMat &dst, std::string kernelName, const char **kernelString)
+static void bitwise_run(const oclMat &src1, oclMat &dst, String kernelName, const char **kernelString)
 {
     dst.create(src1.size(), src1.type());
 
@@ -1629,7 +1654,8 @@ static void bitwise_run(const oclMat &src1, oclMat &dst, std::string kernelName,
 
 
 template<typename T>
-void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString, void *_scalar)
+void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, String kernelName,
+ const char **kernelString, void *_scalar, const char* _opt = NULL)
 {
     dst.create(src1.size(), src1.type());
     CV_Assert(src1.cols == src2.cols && src2.cols == dst.cols &&
@@ -1672,20 +1698,23 @@ void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::strin
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
 
+    T scalar;
     if(_scalar != NULL)
     {
         double scalar1 = *((double *)_scalar);
-        T scalar = (T)scalar1;
+        scalar = (T)scalar1;
         args.push_back( std::make_pair( sizeof(T), (void *)&scalar ));
     }
 
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
+    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, _opt);
 }
-static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, std::string kernelName, const char **kernelString)
+static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
+ String kernelName, const char **kernelString, const char* _opt = NULL)
 {
-    bitwise_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL);
+    bitwise_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL, _opt);
 }
-static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString)
+static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
+ const oclMat &mask, String kernelName, const char **kernelString, const char* _opt = NULL)
 {
     dst.create(src1.size(), src1.type());
     CV_Assert(src1.cols == src2.cols && src2.cols == dst.cols &&
@@ -1733,12 +1762,13 @@ static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, con
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
 
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
+    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, _opt);
 }
 
 
 template <typename WT , typename CL_WT>
-void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar)
+void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst,
+ const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar, const char* opt = NULL)
 {
     dst.create(src1.size(), src1.type());
 
@@ -1800,14 +1830,16 @@ void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, con
         args.push_back( std::make_pair( sizeof(cl_int) , (void *)&isMatSubScalar));
     }
 
-    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
+    openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, opt);
 }
 
 
-typedef void (*BitwiseFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar);
+typedef void (*BitwiseFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst,
+ const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar, const char* opt);
 
 
-static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString, int isMatSubScalar)
+static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
+ const oclMat &mask, String kernelName, const char **kernelString, int isMatSubScalar, const char* opt)
 {
     static BitwiseFuncS tab[8] =
     {
@@ -1834,120 +1866,126 @@ static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
     };
     BitwiseFuncS func = tab[src1.depth()];
     if(func == 0)
-        cv::ocl::error("Unsupported arithmetic operation", __FILE__, __LINE__);
-    func(src1, src2, dst, mask, kernelName, kernelString, isMatSubScalar);
+        cv::error(Error::StsBadArg, "Unsupported arithmetic operation", "", __FILE__, __LINE__);
+    func(src1, src2, dst, mask, kernelName, kernelString, isMatSubScalar, opt);
 }
-static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, std::string kernelName, const char **kernelString)
+static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
+ const oclMat &mask, String kernelName, const char **kernelString, const char * opt = NULL)
 {
-    bitwise_scalar(src1, src2, dst, mask, kernelName, kernelString, 0);
+    bitwise_scalar(src1, src2, dst, mask, kernelName, kernelString, 0, opt);
 }
 
 void cv::ocl::bitwise_not(const oclMat &src, oclMat &dst)
 {
-    if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
     dst.create(src.size(), src.type());
-    std::string kernelName =  "arithm_bitwise_not";
+    String kernelName =  "arithm_bitwise_not";
     bitwise_run(src, dst, kernelName, &arithm_bitwise_not);
 }
 
 void cv::ocl::bitwise_or(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
 {
     // dst.create(src1.size(),src1.type());
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    oclMat emptyMat;
-    std::string kernelName = mask.empty() ? "arithm_bitwise_or" : "arithm_bitwise_or_with_mask";
+
+    String kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
+    static const char opt [] = "-D OP_BINARY=|";
     if (mask.empty())
-        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_or);
+        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
     else
-        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_or_mask);
+        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
 }
 
 
 void cv::ocl::bitwise_or(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    std::string kernelName = mask.data ? "arithm_s_bitwise_or_with_mask" : "arithm_s_bitwise_or";
+    static const char opt [] = "-D OP_BINARY=|";
+    String kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
     if (mask.data)
-        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_or_scalar_mask);
+        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
     else
-        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_or_scalar);
+        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
 }
 
 void cv::ocl::bitwise_and(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
 {
     //    dst.create(src1.size(),src1.type());
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
     oclMat emptyMat;
 
-    std::string kernelName = mask.empty() ? "arithm_bitwise_and" : "arithm_bitwise_and_with_mask";
+    String kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
 
+    static const char opt [] = "-D OP_BINARY=&";
     if (mask.empty())
-        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_and);
+        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
     else
-        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_mask);
+        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
 }
 
 void cv::ocl::bitwise_and(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    std::string kernelName = mask.data ? "arithm_s_bitwise_and_with_mask" : "arithm_s_bitwise_and";
+    static const char opt [] = "-D OP_BINARY=&";
+    String kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
     if (mask.data)
-        bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_scalar_mask);
+        bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
     else
-        bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_scalar);
+        bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
 }
 
 void cv::ocl::bitwise_xor(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
 {
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    oclMat emptyMat;
-    std::string kernelName = mask.empty() ? "arithm_bitwise_xor" : "arithm_bitwise_xor_with_mask";
+    String kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
 
+    static const char opt [] = "-D OP_BINARY=^";
 
     if (mask.empty())
-        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_xor);
+        bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
     else
-        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_mask);
+        bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
 }
 
 
 void cv::ocl::bitwise_xor(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
 {
 
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
-    std::string kernelName = mask.data ? "arithm_s_bitwise_xor_with_mask" : "arithm_s_bitwise_xor";
+    String kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
+    static const char opt [] = "-D OP_BINARY=^";
     if (mask.data)
-        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_scalar_mask);
+        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
     else
-        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_scalar);
+        bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
 }
 
 oclMat cv::ocl::operator ~ (const oclMat &src)
@@ -2033,11 +2071,11 @@ oclMatExpr::operator oclMat() const
 //////////////////////////////////////////////////////////////////////////////
 #define TILE_DIM      (32)
 #define BLOCK_ROWS    (256/TILE_DIM)
-static void transpose_run(const oclMat &src, oclMat &dst, std::string kernelName)
+static void transpose_run(const oclMat &src, oclMat &dst, String kernelName)
 {
-    if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -2134,7 +2172,7 @@ void cv::ocl::addWeighted(const oclMat &src1, double alpha, const oclMat &src2,
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&src2_step ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&src2.offset));
 
-    if(src1.clCxt -> impl -> double_support != 0)
+    if(src1.clCxt->supportsFeature(Context::CL_DOUBLE))
     {
         args.push_back( std::make_pair( sizeof(cl_double), (void *)&alpha ));
         args.push_back( std::make_pair( sizeof(cl_double), (void *)&beta ));
@@ -2250,7 +2288,7 @@ void cv::ocl::magnitudeSqr(const oclMat &src1, oclMat &dst)
     openCLExecuteKernel(clCxt, &arithm_magnitudeSqr, "magnitudeSqr", globalThreads, localThreads, args, 2, depth);
 }
 
-static void arithmetic_pow_run(const oclMat &src1, double p, oclMat &dst, std::string kernelName, const char **kernelString)
+static void arithmetic_pow_run(const oclMat &src1, double p, oclMat &dst, String kernelName, const char **kernelString)
 {
     CV_Assert(src1.cols == dst.cols && src1.rows == dst.rows);
     CV_Assert(src1.type() == dst.type());
@@ -2281,9 +2319,9 @@ static void arithmetic_pow_run(const oclMat &src1, double p, oclMat &dst, std::s
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst.rows ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
-    if(src1.clCxt -> impl -> double_support == 0)
+    float pf = p;
+    if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE))
     {
-        float pf = p;
         args.push_back( std::make_pair( sizeof(cl_float), (void *)&pf ));
     }
     else
@@ -2293,15 +2331,15 @@ static void arithmetic_pow_run(const oclMat &src1, double p, oclMat &dst, std::s
 }
 void cv::ocl::pow(const oclMat &x, double p, oclMat &y)
 {
-    if(x.clCxt -> impl -> double_support == 0 && x.type() == CV_64F)
+    if(!x.clCxt->supportsFeature(Context::CL_DOUBLE) && x.type() == CV_64F)
     {
         std::cout << "Selected device do not support double" << std::endl;
         return;
     }
 
-    CV_Assert((x.type() == y.type() && x.size() == y.size() && x.depth() == CV_32F) || x.depth() == CV_64F);
+    CV_Assert(x.depth() == CV_32F || x.depth() == CV_64F);
     y.create(x.size(), x.type());
-    std::string kernelName = "arithm_pow";
+    String kernelName = "arithm_pow";
 
     arithmetic_pow_run(x, p, y, kernelName, &arithm_pow);
 }
diff --git a/modules/ocl/src/bgfg_mog.cpp b/modules/ocl/src/bgfg_mog.cpp
new file mode 100644
index 0000000000..a92365f0db
--- /dev/null
+++ b/modules/ocl/src/bgfg_mog.cpp
@@ -0,0 +1,638 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jin Ma, jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+using namespace cv;
+using namespace cv::ocl;
+namespace cv
+{
+    namespace ocl
+    {
+        extern const char* bgfg_mog;
+
+        typedef struct _contant_struct
+        {
+            cl_float c_Tb;
+            cl_float c_TB;
+            cl_float c_Tg;
+            cl_float c_varInit;
+            cl_float c_varMin;
+            cl_float c_varMax;
+            cl_float c_tau;
+            cl_uchar c_shadowVal;
+        }contant_struct;
+
+        cl_mem cl_constants = NULL;
+        float c_TB;
+    }
+}
+
+#if defined _MSC_VER
+#define snprintf sprintf_s
+#endif
+
+namespace cv { namespace ocl { namespace device
+{
+    namespace mog
+    {
+        void mog_ocl(const oclMat& frame, int cn, oclMat& fgmask, oclMat& weight, oclMat& sortKey, oclMat& mean, oclMat& var,
+            int nmixtures, float varThreshold, float learningRate, float backgroundRatio, float noiseSigma);
+
+        void getBackgroundImage_ocl(int cn, const oclMat& weight, const oclMat& mean, oclMat& dst, int nmixtures, float backgroundRatio);
+
+        void loadConstants(float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau,
+                            unsigned char shadowVal);
+
+        void mog2_ocl(const oclMat& frame, int cn, oclMat& fgmask, oclMat& modesUsed, oclMat& weight, oclMat& variance, oclMat& mean,
+                      float alphaT, float prune, bool detectShadows, int nmixtures);
+
+        void getBackgroundImage2_ocl(int cn, const oclMat& modesUsed, const oclMat& weight, const oclMat& mean, oclMat& dst, int nmixtures);
+    }
+}}}
+
+namespace mog
+{
+    const int defaultNMixtures = 5;
+    const int defaultHistory = 200;
+    const float defaultBackgroundRatio = 0.7f;
+    const float defaultVarThreshold = 2.5f * 2.5f;
+    const float defaultNoiseSigma = 30.0f * 0.5f;
+    const float defaultInitialWeight = 0.05f;
+}
+void cv::ocl::BackgroundSubtractor::operator()(const oclMat&, oclMat&, float)
+{
+
+}
+cv::ocl::BackgroundSubtractor::~BackgroundSubtractor()
+{
+
+}
+
+cv::ocl::MOG::MOG(int nmixtures) :
+frameSize_(0, 0), frameType_(0), nframes_(0)
+{
+    nmixtures_ = std::min(nmixtures > 0 ? nmixtures : mog::defaultNMixtures, 8);
+    history = mog::defaultHistory;
+    varThreshold = mog::defaultVarThreshold;
+    backgroundRatio = mog::defaultBackgroundRatio;
+    noiseSigma = mog::defaultNoiseSigma;
+}
+
+void cv::ocl::MOG::initialize(cv::Size frameSize, int frameType)
+{
+    CV_Assert(frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4);
+
+    frameSize_ = frameSize;
+    frameType_ = frameType;
+
+    int ch = CV_MAT_CN(frameType);
+    int work_ch = ch;
+
+    // for each gaussian mixture of each pixel bg model we store
+    // the mixture sort key (w/sum_of_variances), the mixture weight (w),
+    // the mean (nchannels values) and
+    // the diagonal covariance matrix (another nchannels values)
+
+    weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+    sortKey_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+    mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
+    var_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch));
+
+    weight_.setTo(cv::Scalar::all(0));
+    sortKey_.setTo(cv::Scalar::all(0));
+    mean_.setTo(cv::Scalar::all(0));
+    var_.setTo(cv::Scalar::all(0));
+
+    nframes_ = 0;
+}
+
+void cv::ocl::MOG::operator()(const cv::ocl::oclMat& frame, cv::ocl::oclMat& fgmask, float learningRate)
+{
+    using namespace cv::ocl::device::mog;
+
+    CV_Assert(frame.depth() == CV_8U);
+
+    int ch = frame.oclchannels();
+    int work_ch = ch;
+
+    if (nframes_ == 0 || learningRate >= 1.0 || frame.size() != frameSize_ || work_ch != mean_.oclchannels())
+        initialize(frame.size(), frame.type());
+
+    fgmask.create(frameSize_, CV_8UC1);
+
+    ++nframes_;
+    learningRate = learningRate >= 0.0f && nframes_ > 1 ? learningRate : 1.0f / std::min(nframes_, history);
+    CV_Assert(learningRate >= 0.0f);
+
+    mog_ocl(frame, ch, fgmask, weight_, sortKey_, mean_, var_, nmixtures_,
+        varThreshold, learningRate, backgroundRatio, noiseSigma);
+}
+
+void cv::ocl::MOG::getBackgroundImage(oclMat& backgroundImage) const
+{
+    using namespace cv::ocl::device::mog;
+
+    backgroundImage.create(frameSize_, frameType_);
+
+    cv::ocl::device::mog::getBackgroundImage_ocl(backgroundImage.oclchannels(), weight_, mean_, backgroundImage, nmixtures_, backgroundRatio);
+}
+
+void cv::ocl::MOG::release()
+{
+    frameSize_ = Size(0, 0);
+    frameType_ = 0;
+    nframes_ = 0;
+
+    weight_.release();
+    sortKey_.release();
+    mean_.release();
+    var_.release();
+    clReleaseMemObject(cl_constants);
+}
+
+static void mog_withoutLearning(const oclMat& frame, int cn, oclMat& fgmask, oclMat& weight, oclMat& mean, oclMat& var,
+    int nmixtures, float varThreshold, float backgroundRatio)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {frame.cols, frame.rows, 1};
+
+    int frame_step = (int)(frame.step/frame.elemSize());
+    int fgmask_step = (int)(fgmask.step/fgmask.elemSize());
+    int weight_step = (int)(weight.step/weight.elemSize());
+    int mean_step = (int)(mean.step/mean.elemSize());
+    int var_step = (int)(var.step/var.elemSize());
+
+    int fgmask_offset_y = (int)(fgmask.offset/fgmask.step);
+    int fgmask_offset_x = (int)(fgmask.offset%fgmask.step);
+    fgmask_offset_x = fgmask_offset_x/(int)fgmask.elemSize();
+
+    int frame_offset_y = (int)(frame.offset/frame.step);
+    int frame_offset_x = (int)(frame.offset%frame.step);
+    frame_offset_x = frame_offset_x/(int)frame.elemSize();
+
+    char build_option[50];
+    if(cn == 1)
+    {
+        snprintf(build_option, 50, "-D CN1 -D NMIXTURES=%d", nmixtures);
+    }else
+    {
+        snprintf(build_option, 50, "-D NMIXTURES=%d", nmixtures);
+    }
+
+    String kernel_name = "mog_withoutLearning_kernel";
+    std::vector<std::pair<size_t, const void*> > args;
+
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&frame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&fgmask.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&weight.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&mean.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&var.data));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.cols));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&weight_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&mean_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&var_step));
+
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&varThreshold));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&backgroundRatio));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_y));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_y));
+
+    openCLExecuteKernel(clCxt, &bgfg_mog, kernel_name, global_thread, local_thread, args, -1, -1, build_option);
+}
+
+
+static void mog_withLearning(const oclMat& frame, int cn, oclMat& fgmask_raw, oclMat& weight, oclMat& sortKey, oclMat& mean, oclMat& var,
+    int nmixtures, float varThreshold, float backgroundRatio, float learningRate, float minVar)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {frame.cols, frame.rows, 1};
+
+    oclMat fgmask(fgmask_raw.size(), CV_32SC1);
+
+    int frame_step = (int)(frame.step/frame.elemSize());
+    int fgmask_step = (int)(fgmask.step/fgmask.elemSize());
+    int weight_step = (int)(weight.step/weight.elemSize());
+    int sortKey_step = (int)(sortKey.step/sortKey.elemSize());
+    int mean_step = (int)(mean.step/mean.elemSize());
+    int var_step = (int)(var.step/var.elemSize());
+
+    int fgmask_offset_y = (int)(fgmask.offset/fgmask.step);
+    int fgmask_offset_x = (int)(fgmask.offset%fgmask.step);
+    fgmask_offset_x = fgmask_offset_x/(int)fgmask.elemSize();
+
+    int frame_offset_y = (int)(frame.offset/frame.step);
+    int frame_offset_x = (int)(frame.offset%frame.step);
+    frame_offset_x = frame_offset_x/(int)frame.elemSize();
+
+    char build_option[50];
+    if(cn == 1)
+    {
+        snprintf(build_option, 50, "-D CN1 -D NMIXTURES=%d", nmixtures);
+    }else
+    {
+        snprintf(build_option, 50, "-D NMIXTURES=%d", nmixtures);
+    }
+
+    String kernel_name = "mog_withLearning_kernel";
+    std::vector<std::pair<size_t, const void*> > args;
+
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&frame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&fgmask.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&weight.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&sortKey.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&mean.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&var.data));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.cols));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&weight_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&sortKey_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&mean_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&var_step));
+
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&varThreshold));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&backgroundRatio));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&learningRate));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&minVar));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_y));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_y));
+
+    openCLExecuteKernel(clCxt, &bgfg_mog, kernel_name, global_thread, local_thread, args, -1, -1, build_option);
+    fgmask.convertTo(fgmask, CV_8U);
+    fgmask.copyTo(fgmask_raw);
+}
+
+void cv::ocl::device::mog::mog_ocl(const oclMat& frame, int cn, oclMat& fgmask, oclMat& weight, oclMat& sortKey, oclMat& mean, oclMat& var,
+    int nmixtures, float varThreshold, float learningRate, float backgroundRatio, float noiseSigma)
+{
+    const float minVar = noiseSigma * noiseSigma;
+
+    if(learningRate > 0.0f)
+        mog_withLearning(frame, cn, fgmask, weight, sortKey, mean, var, nmixtures,
+                         varThreshold, backgroundRatio, learningRate, minVar);
+    else
+        mog_withoutLearning(frame, cn, fgmask, weight, mean, var, nmixtures, varThreshold, backgroundRatio);
+}
+
+void cv::ocl::device::mog::getBackgroundImage_ocl(int cn, const oclMat& weight, const oclMat& mean, oclMat& dst, int nmixtures, float backgroundRatio)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {dst.cols, dst.rows, 1};
+
+    int weight_step = (int)(weight.step/weight.elemSize());
+    int mean_step = (int)(mean.step/mean.elemSize());
+    int dst_step = (int)(dst.step/dst.elemSize());
+
+    char build_option[50];
+    if(cn == 1)
+    {
+        snprintf(build_option, 50, "-D CN1 -D NMIXTURES=%d", nmixtures);
+    }else
+    {
+        snprintf(build_option, 50, "-D NMIXTURES=%d", nmixtures);
+    }
+
+    String kernel_name = "getBackgroundImage_kernel";
+    std::vector<std::pair<size_t, const void*> > args;
+
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&weight.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&mean.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&dst.data));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst.cols));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&weight_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&mean_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst_step));
+
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&backgroundRatio));
+
+    openCLExecuteKernel(clCxt, &bgfg_mog, kernel_name, global_thread, local_thread, args, -1, -1, build_option);
+}
+
+void cv::ocl::device::mog::loadConstants(float Tb, float TB, float Tg, float varInit, float varMin, float varMax, float tau, unsigned char shadowVal)
+{
+    varMin = cv::min(varMin, varMax);
+    varMax = cv::max(varMin, varMax);
+
+    c_TB = TB;
+
+    _contant_struct *constants = new _contant_struct;
+    constants->c_Tb = Tb;
+    constants->c_TB = TB;
+    constants->c_Tg = Tg;
+    constants->c_varInit = varInit;
+    constants->c_varMin = varMin;
+    constants->c_varMax = varMax;
+    constants->c_tau = tau;
+    constants->c_shadowVal = shadowVal;
+
+    cl_constants = load_constant(*((cl_context*)getoclContext()), *((cl_command_queue*)getoclCommandQueue()),
+        (void *)constants, sizeof(_contant_struct));
+}
+
+void cv::ocl::device::mog::mog2_ocl(const oclMat& frame, int cn, oclMat& fgmaskRaw, oclMat& modesUsed, oclMat& weight, oclMat& variance,
+                                oclMat& mean, float alphaT, float prune, bool detectShadows, int nmixtures)
+{
+    oclMat fgmask(fgmaskRaw.size(), CV_32SC1);
+
+    Context* clCxt = Context::getContext();
+
+    const float alpha1 = 1.0f - alphaT;
+
+    cl_int detectShadows_flag = 0;
+    if(detectShadows)
+        detectShadows_flag = 1;
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {frame.cols, frame.rows, 1};
+
+    int frame_step = (int)(frame.step/frame.elemSize());
+    int fgmask_step = (int)(fgmask.step/fgmask.elemSize());
+    int weight_step = (int)(weight.step/weight.elemSize());
+    int modesUsed_step = (int)(modesUsed.step/modesUsed.elemSize());
+    int mean_step = (int)(mean.step/mean.elemSize());
+    int var_step = (int)(variance.step/variance.elemSize());
+
+    int fgmask_offset_y = (int)(fgmask.offset/fgmask.step);
+    int fgmask_offset_x = (int)(fgmask.offset%fgmask.step);
+    fgmask_offset_x = fgmask_offset_x/(int)fgmask.elemSize();
+
+    int frame_offset_y = (int)(frame.offset/frame.step);
+    int frame_offset_x = (int)(frame.offset%frame.step);
+    frame_offset_x = frame_offset_x/(int)frame.elemSize();
+
+    String kernel_name = "mog2_kernel";
+    std::vector<std::pair<size_t, const void*> > args;
+
+    char build_option[50];
+    if(cn == 1)
+    {
+        snprintf(build_option, 50, "-D CN1 -D NMIXTURES=%d", nmixtures);
+    }else
+    {
+        snprintf(build_option, 50, "-D NMIXTURES=%d", nmixtures);
+    }
+
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&frame.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&fgmask.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&weight.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&mean.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&modesUsed.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&variance.data));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame.cols));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&weight_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&mean_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&modesUsed_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&var_step));
+
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&alphaT));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&alpha1));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&prune));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&detectShadows_flag));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&fgmask_offset_y));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&frame_offset_y));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&cl_constants));
+
+    openCLExecuteKernel(clCxt, &bgfg_mog, kernel_name, global_thread, local_thread, args, -1, -1, build_option);
+
+    fgmask.convertTo(fgmask, CV_8U);
+    fgmask.copyTo(fgmaskRaw);
+}
+
+void cv::ocl::device::mog::getBackgroundImage2_ocl(int cn, const oclMat& modesUsed, const oclMat& weight, const oclMat& mean, oclMat& dst, int nmixtures)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {modesUsed.cols, modesUsed.rows, 1};
+
+    int weight_step = (int)(weight.step/weight.elemSize());
+    int modesUsed_step = (int)(modesUsed.step/modesUsed.elemSize());
+    int mean_step = (int)(mean.step/mean.elemSize());
+    int dst_step = (int)(dst.step/dst.elemSize());
+
+    int dst_y = (int)(dst.offset/dst.step);
+    int dst_x = (int)(dst.offset%dst.step);
+    dst_x = dst_x/(int)dst.elemSize();
+
+    String kernel_name = "getBackgroundImage2_kernel";
+    std::vector<std::pair<size_t, const void*> > args;
+
+    char build_option[50];
+    if(cn == 1)
+    {
+        snprintf(build_option, 50, "-D CN1 -D NMIXTURES=%d", nmixtures);
+    }else
+    {
+        snprintf(build_option, 50, "-D NMIXTURES=%d", nmixtures);
+    }
+
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&modesUsed.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&weight.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&mean.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_float), (void*)&c_TB));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&modesUsed.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&modesUsed.cols));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&modesUsed_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&weight_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&mean_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst_step));
+
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst_x));
+    args.push_back(std::make_pair(sizeof(cl_int), (void*)&dst_y));
+
+    openCLExecuteKernel(clCxt, &bgfg_mog, kernel_name, global_thread, local_thread, args, -1, -1, build_option);
+}
+
+/////////////////////////////////////////////////////////////////
+// MOG2
+
+namespace mog2
+{
+    // default parameters of gaussian background detection algorithm
+    const int defaultHistory = 500; // Learning rate; alpha = 1/defaultHistory2
+    const float defaultVarThreshold = 4.0f * 4.0f;
+    const int defaultNMixtures = 5; // maximal number of Gaussians in mixture
+    const float defaultBackgroundRatio = 0.9f; // threshold sum of weights for background test
+    const float defaultVarThresholdGen = 3.0f * 3.0f;
+    const float defaultVarInit = 15.0f; // initial variance for new components
+    const float defaultVarMax = 5.0f * defaultVarInit;
+    const float defaultVarMin = 4.0f;
+
+    // additional parameters
+    const float defaultfCT = 0.05f; // complexity reduction prior constant 0 - no reduction of number of components
+    const unsigned char defaultnShadowDetection = 127; // value to use in the segmentation mask for shadows, set 0 not to do shadow detection
+    const float defaultfTau = 0.5f; // Tau - shadow threshold, see the paper for explanation
+}
+
+cv::ocl::MOG2::MOG2(int nmixtures) : frameSize_(0, 0), frameType_(0), nframes_(0)
+{
+    nmixtures_ = nmixtures > 0 ? nmixtures : mog2::defaultNMixtures;
+
+    history = mog2::defaultHistory;
+    varThreshold = mog2::defaultVarThreshold;
+    bShadowDetection = true;
+
+    backgroundRatio = mog2::defaultBackgroundRatio;
+    fVarInit = mog2::defaultVarInit;
+    fVarMax  = mog2::defaultVarMax;
+    fVarMin = mog2::defaultVarMin;
+
+    varThresholdGen = mog2::defaultVarThresholdGen;
+    fCT = mog2::defaultfCT;
+    nShadowDetection =  mog2::defaultnShadowDetection;
+    fTau = mog2::defaultfTau;
+}
+
+void cv::ocl::MOG2::initialize(cv::Size frameSize, int frameType)
+{
+    using namespace cv::ocl::device::mog;
+    CV_Assert(frameType == CV_8UC1 || frameType == CV_8UC3 || frameType == CV_8UC4);
+
+    frameSize_ = frameSize;
+    frameType_ = frameType;
+    nframes_ = 0;
+
+    int ch = CV_MAT_CN(frameType);
+    int work_ch = ch;
+
+    // for each gaussian mixture of each pixel bg model we store ...
+    // the mixture weight (w),
+    // the mean (nchannels values) and
+    // the covariance
+    weight_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+    weight_.setTo(Scalar::all(0));
+
+    variance_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC1);
+    variance_.setTo(Scalar::all(0));
+
+    mean_.create(frameSize.height * nmixtures_, frameSize_.width, CV_32FC(work_ch)); //4 channels
+    mean_.setTo(Scalar::all(0));
+
+    //make the array for keeping track of the used modes per pixel - all zeros at start
+    bgmodelUsedModes_.create(frameSize_, CV_32FC1);
+    bgmodelUsedModes_.setTo(cv::Scalar::all(0));
+
+    loadConstants(varThreshold, backgroundRatio, varThresholdGen, fVarInit, fVarMin, fVarMax, fTau, nShadowDetection);
+}
+
+void cv::ocl::MOG2::operator()(const oclMat& frame, oclMat& fgmask, float learningRate)
+{
+    using namespace cv::ocl::device::mog;
+
+    int ch = frame.oclchannels();
+    int work_ch = ch;
+
+    if (nframes_ == 0 || learningRate >= 1.0f || frame.size() != frameSize_ || work_ch != mean_.oclchannels())
+        initialize(frame.size(), frame.type());
+
+    fgmask.create(frameSize_, CV_8UC1);
+    fgmask.setTo(cv::Scalar::all(0));
+
+    ++nframes_;
+    learningRate = learningRate >= 0.0f && nframes_ > 1 ? learningRate : 1.0f / std::min(2 * nframes_, history);
+    CV_Assert(learningRate >= 0.0f);
+
+    mog2_ocl(frame, frame.oclchannels(), fgmask, bgmodelUsedModes_, weight_, variance_, mean_, learningRate, -learningRate * fCT, bShadowDetection, nmixtures_);
+}
+
+void cv::ocl::MOG2::getBackgroundImage(oclMat& backgroundImage) const
+{
+    using namespace cv::ocl::device::mog;
+
+    backgroundImage.create(frameSize_, frameType_);
+
+    cv::ocl::device::mog::getBackgroundImage2_ocl(backgroundImage.oclchannels(), bgmodelUsedModes_, weight_, mean_, backgroundImage, nmixtures_);
+}
+
+void cv::ocl::MOG2::release()
+{
+    frameSize_ = Size(0, 0);
+    frameType_ = 0;
+    nframes_ = 0;
+
+    weight_.release();
+    variance_.release();
+    mean_.release();
+
+    bgmodelUsedModes_.release();
+}
diff --git a/modules/ocl/src/binarycaching.hpp b/modules/ocl/src/binarycaching.hpp
index 6c7bee94d4..1b1a7138dd 100644
--- a/modules/ocl/src/binarycaching.hpp
+++ b/modules/ocl/src/binarycaching.hpp
@@ -68,13 +68,13 @@ namespace cv
             }
 
             //lookup the binary given the file name
-            cl_program progLookup(std::string srcsign);
+            cl_program progLookup(String srcsign);
 
             //add program to the cache
-            void addProgram(std::string srcsign, cl_program program);
+            void addProgram(String srcsign, cl_program program);
             void releaseProgram();
 
-            std::map <std::string, cl_program> codeCache;
+            std::map <String, cl_program> codeCache;
             unsigned int cacheSize;
             //The presumed watermark for the cache volume (256MB). Is it enough?
             //We may need more delicate algorithms when necessary later.
diff --git a/modules/ocl/src/blend.cpp b/modules/ocl/src/blend.cpp
index d0af27c026..265998e753 100644
--- a/modules/ocl/src/blend.cpp
+++ b/modules/ocl/src/blend.cpp
@@ -62,7 +62,7 @@ void cv::ocl::blendLinear(const oclMat &img1, const oclMat &img2, const oclMat &
                           oclMat &result)
 {
     cv::ocl::Context *ctx = img1.clCxt;
-    assert(ctx == img2.clCxt && ctx == weights1.clCxt && ctx == weights2.clCxt);
+    CV_Assert(ctx == img2.clCxt && ctx == weights1.clCxt && ctx == weights2.clCxt);
     int channels = img1.oclchannels();
     int depth = img1.depth();
     int rows = img1.rows;
@@ -85,8 +85,8 @@ void cv::ocl::blendLinear(const oclMat &img1, const oclMat &img2, const oclMat &
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&istep ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&wstep ));
-        std::string kernelName = "BlendLinear";
+        String kernelName = "BlendLinear";
 
         openCLExecuteKernel(ctx, &blend_linear, kernelName, globalSize, localSize, args, channels, depth);
     }
-}
\ No newline at end of file
+}
diff --git a/modules/ocl/src/brute_force_matcher.cpp b/modules/ocl/src/brute_force_matcher.cpp
index da9e5412aa..d7255afd0d 100644
--- a/modules/ocl/src/brute_force_matcher.cpp
+++ b/modules/ocl/src/brute_force_matcher.cpp
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Nathan, liujun@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -44,6 +45,7 @@
 //M*/
 
 #include "precomp.hpp"
+
 #include <functional>
 #include <iterator>
 #include <vector>
@@ -59,7 +61,18 @@ namespace cv
     }
 }
 
-template < int BLOCK_SIZE, int MAX_DESC_LEN,  typename T/*, typename Mask*/ >
+static const int OPT_SIZE = 100;
+
+static const char * T_ARR [] = {
+    "uchar",
+    "char",
+    "ushort",
+    "short",
+    "int",
+    "float -D T_FLOAT",
+    "double"};
+
+template < int BLOCK_SIZE, int MAX_DESC_LEN/*, typename Mask*/ >
 void matchUnrolledCached(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
                          const oclMat &trainIdx, const oclMat &distance, int distType)
 {
@@ -71,6 +84,11 @@ void matchUnrolledCached(const oclMat &query, const oclMat &train, const oclMat
     int m_size = MAX_DESC_LEN;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d -D MAX_DESC_LEN=%d",
+        T_ARR[query.depth()], distType, block_size, m_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -79,28 +97,25 @@ void matchUnrolledCached(const oclMat &query, const oclMat &train, const oclMat
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&trainIdx.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&m_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_UnrollMatch";
+        String kernelName = "BruteForceMatch_UnrollMatch";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-template < int BLOCK_SIZE, int MAX_DESC_LEN,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE, int MAX_DESC_LEN/*, typename Mask*/ >
 void matchUnrolledCached(const oclMat /*query*/, const oclMat * /*trains*/, int /*n*/, const oclMat /*mask*/,
                          const oclMat &/*bestTrainIdx*/, const oclMat & /*bestImgIdx*/, const oclMat & /*bestDistance*/, int /*distType*/)
 {
 }
 
-template < int BLOCK_SIZE,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE/*, typename Mask*/ >
 void match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
            const oclMat &trainIdx, const oclMat &distance, int distType)
 {
@@ -111,6 +126,10 @@ void match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
     int block_size = BLOCK_SIZE;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d",
+        T_ARR[query.depth()], distType, block_size);
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -119,28 +138,26 @@ void match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&trainIdx.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_Match";
+        String kernelName = "BruteForceMatch_Match";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-template < int BLOCK_SIZE,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE/*, typename Mask*/ >
 void match(const oclMat /*query*/, const oclMat * /*trains*/, int /*n*/, const oclMat /*mask*/,
            const oclMat &/*bestTrainIdx*/, const oclMat & /*bestImgIdx*/, const oclMat & /*bestDistance*/, int /*distType*/)
 {
 }
 
 //radius_matchUnrolledCached
-template < int BLOCK_SIZE, int MAX_DESC_LEN,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE, int MAX_DESC_LEN/*, typename Mask*/ >
 void matchUnrolledCached(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &/*mask*/,
                          const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches, int distType)
 {
@@ -152,6 +169,11 @@ void matchUnrolledCached(const oclMat &query, const oclMat &train, float maxDist
     int m_size = MAX_DESC_LEN;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d -D MAX_DESC_LEN=%d",
+        T_ARR[query.depth()], distType, block_size, m_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -162,8 +184,6 @@ void matchUnrolledCached(const oclMat &query, const oclMat &train, float maxDist
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&nMatches.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&m_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
@@ -171,16 +191,15 @@ void matchUnrolledCached(const oclMat &query, const oclMat &train, float maxDist
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&trainIdx.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&trainIdx.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_RadiusUnrollMatch";
+        String kernelName = "BruteForceMatch_RadiusUnrollMatch";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
 //radius_match
-template < int BLOCK_SIZE, typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE/*, typename Mask*/ >
 void radius_match(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &/*mask*/,
                   const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches, int distType)
 {
@@ -191,6 +210,11 @@ void radius_match(const oclMat &query, const oclMat &train, float maxDistance, c
     int block_size = BLOCK_SIZE;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d",
+        T_ARR[query.depth()], distType, block_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -201,7 +225,6 @@ void radius_match(const oclMat &query, const oclMat &train, float maxDistance, c
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&nMatches.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
@@ -209,267 +232,76 @@ void radius_match(const oclMat &query, const oclMat &train, float maxDistance, c
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&trainIdx.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&trainIdx.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_RadiusMatch";
+        String kernelName = "BruteForceMatch_RadiusMatch";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
-        //float *dis = (float *)clEnqueueMapBuffer(ctx->impl->clCmdQueue, (cl_mem)distance.data, CL_TRUE, CL_MAP_READ, 0, 8, 0, NULL, NULL, NULL);
-        //printf("%f, %f\n", dis[0], dis[1]);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-// with mask
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
+static void matchDispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
                      const oclMat &trainIdx, const oclMat &distance, int distType)
 {
+    const oclMat zeroMask;
+    const oclMat &tempMask = mask.data ? mask : zeroMask;
+    bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
     if (query.cols <= 64)
     {
-        matchUnrolledCached<16, 64, T>(query, train, mask, trainIdx, distance, distType);
+        matchUnrolledCached<16, 64>(query, train, tempMask, trainIdx, distance, distType);
     }
-    else if (query.cols <= 128)
+    else if (query.cols <= 128 && !is_cpu)
     {
-        matchUnrolledCached<16, 128, T>(query, train, mask, trainIdx,  distance, distType);
+        matchUnrolledCached<16, 128>(query, train, tempMask, trainIdx,  distance, distType);
     }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, train, mask, trainIdx, distance, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, train, mask, trainIdx, distance, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, train, mask, trainIdx, distance, stream);
-    }*/
     else
     {
-        match<16, T>(query, train, mask, trainIdx, distance, distType);
+        match<16>(query, train, tempMask, trainIdx, distance, distType);
     }
 }
 
-// without mask
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, const oclMat &trainIdx, const oclMat &distance, int distType)
-{
-    oclMat mask;
-    if (query.cols <= 64)
-    {
-        matchUnrolledCached<16, 64, T>(query, train, mask, trainIdx, distance, distType);
-    }
-    else if (query.cols <= 128)
-    {
-        matchUnrolledCached<16, 128, T>(query, train, mask, trainIdx,  distance, distType);
-    }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance);
-    }*/
-    else
-    {
-        match<16, T>(query, train, mask, trainIdx, distance, distType);
-    }
-}
-
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat *trains, int n, const oclMat &mask,
+static void matchDispatcher(const oclMat &query, const oclMat *trains, int n, const oclMat &mask,
                      const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance, int distType)
 {
+    const oclMat zeroMask;
+    const oclMat &tempMask = mask.data ? mask : zeroMask;
+    bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
     if (query.cols <= 64)
     {
-        matchUnrolledCached<16, 64, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
+        matchUnrolledCached<16, 64>(query, trains, n, tempMask, trainIdx, imgIdx, distance, distType);
     }
-    else if (query.cols <= 128)
+    else if (query.cols <= 128 && !is_cpu)
     {
-        matchUnrolledCached<16, 128, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
+        matchUnrolledCached<16, 128>(query, trains, n, tempMask, trainIdx, imgIdx, distance, distType);
     }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }*/
     else
     {
-        match<16, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
-    }
-}
-
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat *trains, int n, const oclMat &trainIdx,
-                     const oclMat &imgIdx, const oclMat &distance, int distType)
-{
-    oclMat mask;
-    if (query.cols <= 64)
-    {
-        matchUnrolledCached<16, 64, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
-    }
-    else if (query.cols <= 128)
-    {
-        matchUnrolledCached<16, 128, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
-    }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, trains, n, mask, trainIdx, imgIdx, distance, stream);
-    }*/
-    else
-    {
-        match<16, T>(query, trains, n, mask, trainIdx, imgIdx, distance, distType);
+        match<16>(query, trains, n, tempMask, trainIdx, imgIdx, distance, distType);
     }
 }
 
 //radius matchDispatcher
-// with mask
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &mask,
+static void matchDispatcher(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &mask,
                      const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches, int distType)
 {
+    const oclMat zeroMask;
+    const oclMat &tempMask = mask.data ? mask : zeroMask;
+    bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
     if (query.cols <= 64)
     {
-        matchUnrolledCached<16, 64, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
+        matchUnrolledCached<16, 64>(query, train, maxDistance, tempMask, trainIdx, distance, nMatches, distType);
     }
-    else if (query.cols <= 128)
+    else if (query.cols <= 128 && !is_cpu)
     {
-        matchUnrolledCached<16, 128, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
+        matchUnrolledCached<16, 128>(query, train, maxDistance, tempMask, trainIdx, distance, nMatches, distType);
     }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }*/
     else
     {
-        radius_match<16, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-}
-
-// without mask
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &trainIdx,
-                     const oclMat &distance, const oclMat &nMatches, int distType)
-{
-    oclMat mask;
-    if (query.cols <= 64)
-    {
-        matchUnrolledCached<16, 64, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    else if (query.cols <= 128)
-    {
-        matchUnrolledCached<16, 128, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, train, maxDistance, mask, trainIdx, distance, nMatches, stream);
-    }*/
-    else
-    {
-        radius_match<16, T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-}
-
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, int n, float maxDistance, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches, int distType)
-{
-    if (query.cols <= 64)
-    {
-        matchUnrolledCached<16, 64, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    else if (query.cols <= 128)
-    {
-        matchUnrolledCached<16, 128, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }*/
-    else
-    {
-        match<16, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-}
-
-// without mask
-template < typename T/*, typename Mask*/ >
-void matchDispatcher(const oclMat &query, const oclMat &train, int n, float maxDistance, const oclMat &trainIdx,
-                     const oclMat &distance, const oclMat &nMatches, int distType)
-{
-    oclMat mask;
-    if (query.cols <= 64)
-    {
-        matchUnrolledCached<16, 64, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    else if (query.cols <= 128)
-    {
-        matchUnrolledCached<16, 128, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, trains, n, maxDistance, masks, trainIdx, imgIdx, distance, nMatches, stream);
-    }*/
-    else
-    {
-        match<16, T>(query, train, n, maxDistance, mask, trainIdx, distance, nMatches, distType);
+        radius_match<16>(query, train, maxDistance, tempMask, trainIdx, distance, nMatches, distType);
     }
 }
 
 //knn match Dispatcher
-template < int BLOCK_SIZE, int MAX_DESC_LEN,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE, int MAX_DESC_LEN/*, typename Mask*/ >
 void knn_matchUnrolledCached(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
                              const oclMat &trainIdx, const oclMat &distance, int distType)
 {
@@ -481,6 +313,11 @@ void knn_matchUnrolledCached(const oclMat &query, const oclMat &train, const ocl
     int m_size = MAX_DESC_LEN;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d -D MAX_DESC_LEN=%d",
+        T_ARR[query.depth()], distType, block_size, m_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -489,22 +326,19 @@ void knn_matchUnrolledCached(const oclMat &query, const oclMat &train, const ocl
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&trainIdx.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&m_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_knnUnrollMatch";
+        String kernelName = "BruteForceMatch_knnUnrollMatch";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-template < int BLOCK_SIZE,  typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE/*, typename Mask*/ >
 void knn_match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
                const oclMat &trainIdx, const oclMat &distance, int distType)
 {
@@ -515,6 +349,11 @@ void knn_match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
     int block_size = BLOCK_SIZE;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d",
+        T_ARR[query.depth()], distType, block_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -523,21 +362,19 @@ void knn_match(const oclMat &query, const oclMat &train, const oclMat &/*mask*/,
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&trainIdx.data ));
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&distance.data ));
         args.push_back( std::make_pair( smemSize, (void *)NULL));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_size ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_knnMatch";
+        String kernelName = "BruteForceMatch_knnMatch";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-template < int BLOCK_SIZE, int MAX_DESC_LEN, typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE, int MAX_DESC_LEN/*, typename Mask*/ >
 void calcDistanceUnrolled(const oclMat &query, const oclMat &train, const oclMat &/*mask*/, const oclMat &allDist, int distType)
 {
     cv::ocl::Context *ctx = query.clCxt;
@@ -548,6 +385,11 @@ void calcDistanceUnrolled(const oclMat &query, const oclMat &train, const oclMat
     int m_size = MAX_DESC_LEN;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d -D MAX_DESC_LEN=%d",
+        T_ARR[query.depth()], distType, block_size, m_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -562,15 +404,14 @@ void calcDistanceUnrolled(const oclMat &query, const oclMat &train, const oclMat
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_calcDistanceUnrolled";
+        String kernelName = "BruteForceMatch_calcDistanceUnrolled";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
-template < int BLOCK_SIZE, typename T/*, typename Mask*/ >
+template < int BLOCK_SIZE/*, typename Mask*/ >
 void calcDistance(const oclMat &query, const oclMat &train, const oclMat &/*mask*/, const oclMat &allDist, int distType)
 {
     cv::ocl::Context *ctx = query.clCxt;
@@ -580,6 +421,11 @@ void calcDistance(const oclMat &query, const oclMat &train, const oclMat &/*mask
     int block_size = BLOCK_SIZE;
     std::vector< std::pair<size_t, const void *> > args;
 
+    char opt [OPT_SIZE] = "";
+    sprintf(opt,
+        "-D T=%s -D DIST_TYPE=%d -D BLOCK_SIZE=%d",
+        T_ARR[query.depth()], distType, block_size);
+
     if(globalSize[0] != 0)
     {
         args.push_back( std::make_pair( sizeof(cl_mem), (void *)&query.data ));
@@ -593,73 +439,47 @@ void calcDistance(const oclMat &query, const oclMat &train, const oclMat &/*mask
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.rows ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&train.cols ));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&query.step ));
-        args.push_back( std::make_pair( sizeof(cl_int), (void *)&distType ));
 
-        std::string kernelName = "BruteForceMatch_calcDistance";
+        String kernelName = "BruteForceMatch_calcDistance";
 
-        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1);
+        openCLExecuteKernel(ctx, &brute_force_match, kernelName, globalSize, localSize, args, -1, -1, opt);
     }
 }
 
 ///////////////////////////////////////////////////////////////////////////////
 // Calc Distance dispatcher
-template < typename T/*, typename Mask*/ >
-void calcDistanceDispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
+static void calcDistanceDispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
                             const oclMat &allDist, int distType)
 {
     if (query.cols <= 64)
     {
-        calcDistanceUnrolled<16, 64, T>(query, train, mask, allDist, distType);
+        calcDistanceUnrolled<16, 64>(query, train, mask, allDist, distType);
     }
     else if (query.cols <= 128)
     {
-        calcDistanceUnrolled<16, 128, T>(query, train, mask, allDist, distType);
+        calcDistanceUnrolled<16, 128>(query, train, mask, allDist, distType);
     }
-    /*else if (query.cols <= 256)
-    {
-        calcDistanceUnrolled<16, 256, Dist>(query, train, mask, allDist, stream);
-    }
-    else if (query.cols <= 512)
-    {
-        calcDistanceUnrolled<16, 512, Dist>(query, train, mask, allDist, stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        calcDistanceUnrolled<16, 1024, Dist>(query, train, mask, allDist, stream);
-    }*/
     else
     {
-        calcDistance<16, T>(query, train, mask, allDist, distType);
+        calcDistance<16>(query, train, mask, allDist, distType);
     }
 }
 
-template < typename T/*, typename Mask*/ >
-void match2Dispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
+static void match2Dispatcher(const oclMat &query, const oclMat &train, const oclMat &mask,
                       const oclMat &trainIdx, const oclMat &distance, int distType)
 {
+    bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
     if (query.cols <= 64)
     {
-        knn_matchUnrolledCached<16, 64, T>(query, train, mask, trainIdx, distance, distType);
+        knn_matchUnrolledCached<16, 64>(query, train, mask, trainIdx, distance, distType);
     }
-    else if (query.cols <= 128)
+    else if (query.cols <= 128 && !is_cpu)
     {
-        knn_matchUnrolledCached<16, 128, T>(query, train, mask, trainIdx, distance, distType);
+        knn_matchUnrolledCached<16, 128>(query, train, mask, trainIdx, distance, distType);
     }
-    /*else if (query.cols <= 256)
-    {
-        matchUnrolled<16, 256, Dist>(query, train, mask, static_cast< DevMem2D_<int2> >(trainIdx), static_cast< DevMem2D_<float2> > (distance), stream);
-    }
-    else if (query.cols <= 512)
-    {
-        matchUnrolled<16, 512, Dist>(query, train, mask, static_cast< DevMem2D_<int2> >(trainIdx), static_cast< DevMem2D_<float2> > (distance), stream);
-    }
-    else if (query.cols <= 1024)
-    {
-        matchUnrolled<16, 1024, Dist>(query, train, mask, static_cast< DevMem2D_<int2> >(trainIdx), static_cast< DevMem2D_<float2> > (distance), stream);
-    }*/
     else
     {
-        knn_match<16, T>(query, train, mask, trainIdx, distance, distType);
+        knn_match<16>(query, train, mask, trainIdx, distance, distType);
     }
 }
 
@@ -670,7 +490,7 @@ void findKnnMatch(int k, const oclMat &trainIdx, const oclMat &distance, const o
     size_t globalSize[] = {trainIdx.rows * BLOCK_SIZE, 1, 1};
     size_t localSize[] = {BLOCK_SIZE, 1, 1};
     int block_size = BLOCK_SIZE;
-    std::string kernelName = "BruteForceMatch_findBestMatch";
+    String kernelName = "BruteForceMatch_findBestMatch";
 
     for (int i = 0; i < k; ++i)
     {
@@ -694,206 +514,22 @@ static void findKnnMatchDispatcher(int k, const oclMat &trainIdx, const oclMat &
     findKnnMatch<256>(k, trainIdx, distance, allDist, distType);
 }
 
-//with mask
-template < typename T/*, typename Mask*/ >
-void kmatchDispatcher(const oclMat &query, const oclMat &train, int k, const oclMat &mask,
+static void kmatchDispatcher(const oclMat &query, const oclMat &train, int k, const oclMat &mask,
                       const oclMat &trainIdx, const oclMat &distance, const oclMat &allDist, int distType)
 {
+    const oclMat zeroMask;
+    const oclMat &tempMask = mask.data ? mask : zeroMask;
     if (k == 2)
     {
-        match2Dispatcher<T>(query, train, mask, trainIdx, distance, distType);
+        match2Dispatcher(query, train, tempMask, trainIdx, distance, distType);
     }
     else
     {
-        calcDistanceDispatcher<T>(query, train, mask, allDist, distType);
+        calcDistanceDispatcher(query, train, tempMask, allDist, distType);
         findKnnMatchDispatcher(k, trainIdx, distance, allDist, distType);
     }
 }
 
-//without mask
-template < typename T/*, typename Mask*/ >
-void kmatchDispatcher(const oclMat &query, const oclMat &train, int k,
-                      const oclMat &trainIdx, const oclMat &distance, const oclMat &allDist, int distType)
-{
-    oclMat mask;
-    if (k == 2)
-    {
-        match2Dispatcher<T>(query, train, mask, trainIdx, distance, distType);
-    }
-    else
-    {
-        calcDistanceDispatcher<T>(query, train, mask, allDist, distType);
-        findKnnMatchDispatcher(k, trainIdx, distance, allDist, distType);
-    }
-}
-
-
-
-template <typename T>
-void ocl_matchL1_gpu(const oclMat &query, const oclMat &train, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance)
-{
-    int distType = 0;
-    if (mask.data)
-    {
-        matchDispatcher<T>(query, train, mask, trainIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher< T >(query, train, trainIdx, distance, distType);
-    }
-}
-
-template <typename T>
-void ocl_matchL1_gpu(const oclMat &query, const oclMat &trains, const oclMat &masks,
-                     const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance)
-{
-    int distType = 0;
-
-    if (masks.data)
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, masks, trainIdx, imgIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, trainIdx, imgIdx, distance, distType);
-    }
-}
-
-template <typename T>
-void ocl_matchL2_gpu(const oclMat &query, const oclMat &train, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance)
-{
-    int distType = 1;
-    if (mask.data)
-    {
-        matchDispatcher<T>(query, train, mask, trainIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher<T >(query, train, trainIdx, distance, distType);
-    }
-}
-
-template <typename T>
-void ocl_matchL2_gpu(const oclMat &query, const oclMat &trains, const oclMat &masks,
-                     const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance)
-{
-    int distType = 1;
-    if (masks.data)
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, masks, trainIdx, imgIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, trainIdx, imgIdx, distance, distType);
-    }
-}
-
-template <typename T>
-void ocl_matchHamming_gpu(const oclMat &query, const oclMat &train, const oclMat &mask,
-                          const oclMat &trainIdx, const oclMat &distance)
-{
-    int distType = 2;
-    if (mask.data)
-    {
-        matchDispatcher<T>(query, train, mask, trainIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher< T >(query, train, trainIdx, distance, distType);
-    }
-}
-
-template <typename T>
-void ocl_matchHamming_gpu(const oclMat &query, const oclMat &trains, const oclMat &masks,
-                          const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance)
-{
-    int distType = 2;
-    if (masks.data)
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, masks, trainIdx, imgIdx, distance, distType);
-    }
-    else
-    {
-        matchDispatcher<T>(query, (const oclMat *)trains.ptr(), trains.cols, trainIdx, imgIdx, distance, distType);
-    }
-}
-
-// knn caller
-template <typename T>
-void ocl_matchL1_gpu(const oclMat &query, const oclMat &train, int k, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance, const oclMat &allDist)
-{
-    int distType = 0;
-
-    if (mask.data)
-        kmatchDispatcher<T>(query, train, k, mask, trainIdx, distance, allDist, distType);
-    else
-        kmatchDispatcher<T>(query, train, k, trainIdx, distance, allDist, distType);
-}
-
-template <typename T>
-void ocl_matchL2_gpu(const oclMat &query, const oclMat &train, int k, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance, const oclMat &allDist)
-{
-    int distType = 1;
-
-    if (mask.data)
-        kmatchDispatcher<T>(query, train, k, mask, trainIdx, distance, allDist, distType);
-    else
-        kmatchDispatcher<T>(query, train, k, trainIdx, distance, allDist, distType);
-}
-
-template <typename T>
-void ocl_matchHamming_gpu(const oclMat &query, const oclMat &train, int k, const oclMat &mask,
-                          const oclMat &trainIdx, const oclMat &distance, const oclMat &allDist)
-{
-    int distType = 2;
-
-    if (mask.data)
-        kmatchDispatcher<T>(query, train, k, mask, trainIdx, distance, allDist, distType);
-    else
-        kmatchDispatcher<T>(query, train, k,  trainIdx, distance, allDist, distType);
-}
-
-//radius caller
-template <typename T>
-void ocl_matchL1_gpu(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches)
-{
-    int distType = 0;
-
-    if (mask.data)
-        matchDispatcher<T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    else
-        matchDispatcher<T>(query, train, maxDistance, trainIdx, distance, nMatches, distType);
-}
-
-template <typename T>
-void ocl_matchL2_gpu(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &mask,
-                     const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches)
-{
-    int distType = 1;
-
-    if (mask.data)
-        matchDispatcher<T>(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
-    else
-        matchDispatcher<T>(query, train, maxDistance, trainIdx, distance, nMatches, distType);
-}
-
-template <typename T>
-void ocl_matchHamming_gpu(const oclMat &query, const oclMat &train, float maxDistance, const oclMat &mask,
-                          const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches)
-{
-    int distType = 2;
-
-    if (mask.data)
-        matchDispatcher<T>(query, train, maxDistance, mask, trainIdx, distance,  nMatches, distType);
-    else
-        matchDispatcher<T>(query, train, maxDistance, trainIdx, distance, nMatches, distType);
-}
-
 cv::ocl::BruteForceMatcher_OCL_base::BruteForceMatcher_OCL_base(DistType distType_) : distType(distType_)
 {
 }
@@ -929,37 +565,15 @@ void cv::ocl::BruteForceMatcher_OCL_base::matchSingle(const oclMat &query, const
     if (query.empty() || train.empty())
         return;
 
-    typedef void (*caller_t)(const oclMat & query, const oclMat & train, const oclMat & mask,
-                             const oclMat & trainIdx, const oclMat & distance);
-
-    static const caller_t callers[3][6] =
-    {
-        {
-            ocl_matchL1_gpu<unsigned char>, 0/*ocl_matchL1_gpu<signed char>*/,
-            ocl_matchL1_gpu<unsigned short>, ocl_matchL1_gpu<short>,
-            ocl_matchL1_gpu<int>, ocl_matchL1_gpu<float>
-        },
-        {
-            0/*ocl_matchL2_gpu<unsigned char>*/, 0/*ocl_matchL2_gpu<signed char>*/,
-            0/*ocl_matchL2_gpu<unsigned short>*/, 0/*ocl_matchL2_gpu<short>*/,
-            0/*ocl_matchL2_gpu<int>*/, ocl_matchL2_gpu<float>
-        },
-        {
-            ocl_matchHamming_gpu<unsigned char>, 0/*ocl_matchHamming_gpu<signed char>*/,
-            ocl_matchHamming_gpu<unsigned short>, 0/*ocl_matchHamming_gpu<short>*/,
-            ocl_matchHamming_gpu<int>, 0/*ocl_matchHamming_gpu<float>*/
-        }
-    };
-
     CV_Assert(query.channels() == 1 && query.depth() < CV_64F);
     CV_Assert(train.cols == query.cols && train.type() == query.type());
 
-    const int nQuery = query.rows;
-    trainIdx.create(1, nQuery, CV_32S);
-    distance.create(1, nQuery, CV_32F);
+    ensureSizeIsEnough(1, query.rows, CV_32S, trainIdx);
+    ensureSizeIsEnough(1, query.rows, CV_32F, distance);
 
-    caller_t func = callers[distType][query.depth()];
-    func(query, train, mask, trainIdx, distance);
+    matchDispatcher(query, train, mask, trainIdx, distance, distType);
+
+    return;
 }
 
 void cv::ocl::BruteForceMatcher_OCL_base::matchDownload(const oclMat &trainIdx, const oclMat &distance, std::vector<DMatch> &matches)
@@ -1005,7 +619,7 @@ void cv::ocl::BruteForceMatcher_OCL_base::matchConvert(const Mat &trainIdx, cons
 
 void cv::ocl::BruteForceMatcher_OCL_base::match(const oclMat &query, const oclMat &train, std::vector<DMatch> &matches, const oclMat &mask)
 {
-    assert(mask.empty()); // mask is not supported at the moment
+    CV_Assert(mask.empty()); // mask is not supported at the moment
     oclMat trainIdx, distance;
     matchSingle(query, train, trainIdx, distance, mask);
     matchDownload(trainIdx, distance, matches);
@@ -1061,40 +675,17 @@ void cv::ocl::BruteForceMatcher_OCL_base::matchCollection(const oclMat &query, c
     if (query.empty() || trainCollection.empty())
         return;
 
-    typedef void (*caller_t)(const oclMat & query, const oclMat & trains, const oclMat & masks,
-                             const oclMat & trainIdx, const oclMat & imgIdx, const oclMat & distance);
-
-    static const caller_t callers[3][6] =
-    {
-        {
-            ocl_matchL1_gpu<unsigned char>, 0/*matchL1_gpu<signed char>*/,
-            ocl_matchL1_gpu<unsigned short>, ocl_matchL1_gpu<short>,
-            ocl_matchL1_gpu<int>, ocl_matchL1_gpu<float>
-        },
-        {
-            0/*matchL2_gpu<unsigned char>*/, 0/*matchL2_gpu<signed char>*/,
-            0/*matchL2_gpu<unsigned short>*/, 0/*matchL2_gpu<short>*/,
-            0/*matchL2_gpu<int>*/, ocl_matchL2_gpu<float>
-        },
-        {
-            ocl_matchHamming_gpu<unsigned char>, 0/*matchHamming_gpu<signed char>*/,
-            ocl_matchHamming_gpu<unsigned short>, 0/*matchHamming_gpu<short>*/,
-            ocl_matchHamming_gpu<int>, 0/*matchHamming_gpu<float>*/
-        }
-    };
-
     CV_Assert(query.channels() == 1 && query.depth() < CV_64F);
 
     const int nQuery = query.rows;
 
-    trainIdx.create(1, nQuery, CV_32S);
-    imgIdx.create(1, nQuery, CV_32S);
-    distance.create(1, nQuery, CV_32F);
+    ensureSizeIsEnough(1, nQuery, CV_32S, trainIdx);
+    ensureSizeIsEnough(1, nQuery, CV_32S, imgIdx);
+    ensureSizeIsEnough(1, nQuery, CV_32F, distance);
 
-    caller_t func = callers[distType][query.depth()];
-    CV_Assert(func != 0);
+    matchDispatcher(query, (const oclMat *)trainCollection.ptr(), trainCollection.cols, masks, trainIdx, imgIdx, distance, distType);
 
-    func(query, trainCollection, masks, trainIdx, imgIdx, distance);
+    return;
 }
 
 void cv::ocl::BruteForceMatcher_OCL_base::matchDownload(const oclMat &trainIdx, const oclMat &imgIdx, const oclMat &distance, std::vector<DMatch> &matches)
@@ -1163,28 +754,6 @@ void cv::ocl::BruteForceMatcher_OCL_base::knnMatchSingle(const oclMat &query, co
     if (query.empty() || train.empty())
         return;
 
-    typedef void (*caller_t)(const oclMat & query, const oclMat & train, int k, const oclMat & mask,
-                             const oclMat & trainIdx, const oclMat & distance, const oclMat & allDist);
-
-    static const caller_t callers[3][6] =
-    {
-        {
-            ocl_matchL1_gpu<unsigned char>, 0/*ocl_matchL1_gpu<signed char>*/,
-            ocl_matchL1_gpu<unsigned short>, ocl_matchL1_gpu<short>,
-            ocl_matchL1_gpu<int>, ocl_matchL1_gpu<float>
-        },
-        {
-            0/*ocl_matchL2_gpu<unsigned char>*/, 0/*ocl_matchL2_gpu<signed char>*/,
-            0/*ocl_matchL2_gpu<unsigned short>*/, 0/*ocl_matchL2_gpu<short>*/,
-            0/*ocl_matchL2_gpu<int>*/, ocl_matchL2_gpu<float>
-        },
-        {
-            ocl_matchHamming_gpu<unsigned char>, 0/*ocl_matchHamming_gpu<signed char>*/,
-            ocl_matchHamming_gpu<unsigned short>, 0/*ocl_matchHamming_gpu<short>*/,
-            ocl_matchHamming_gpu<int>, 0/*ocl_matchHamming_gpu<float>*/
-        }
-    };
-
     CV_Assert(query.channels() == 1 && query.depth() < CV_64F);
     CV_Assert(train.type() == query.type() && train.cols == query.cols);
 
@@ -1193,22 +762,21 @@ void cv::ocl::BruteForceMatcher_OCL_base::knnMatchSingle(const oclMat &query, co
 
     if (k == 2)
     {
-        trainIdx.create(1, nQuery, CV_32SC2);
-        distance.create(1, nQuery, CV_32FC2);
+        ensureSizeIsEnough(1, nQuery, CV_32SC2, trainIdx);
+        ensureSizeIsEnough(1, nQuery, CV_32FC2, distance);
     }
     else
     {
-        trainIdx.create(nQuery, k, CV_32S);
-        distance.create(nQuery, k, CV_32F);
-        allDist.create(nQuery, nTrain, CV_32FC1);
+        ensureSizeIsEnough(nQuery, k, CV_32S, trainIdx);
+        ensureSizeIsEnough(nQuery, k, CV_32F, distance);
+        ensureSizeIsEnough(nQuery, nTrain, CV_32FC1, allDist);
     }
 
     trainIdx.setTo(Scalar::all(-1));
 
-    caller_t func = callers[distType][query.depth()];
-    CV_Assert(func != 0);
+    kmatchDispatcher(query, train, k, mask, trainIdx, distance, allDist, distType);
 
-    func(query, train, k, mask, trainIdx, distance, allDist);
+    return;
 }
 
 void cv::ocl::BruteForceMatcher_OCL_base::knnMatchDownload(const oclMat &trainIdx, const oclMat &distance, std::vector< std::vector<DMatch> > &matches, bool compactResult)
@@ -1280,35 +848,16 @@ void cv::ocl::BruteForceMatcher_OCL_base::knnMatch2Collection(const oclMat &quer
     if (query.empty() || trainCollection.empty())
         return;
 
-    typedef void (*caller_t)(const oclMat & query, const oclMat & trains, const oclMat & masks,
-                             const oclMat & trainIdx, const oclMat & imgIdx, const oclMat & distance);
-#if 0
-    static const caller_t callers[3][6] =
-    {
-        {
-            ocl_match2L1_gpu<unsigned char>, 0/*match2L1_gpu<signed char>*/,
-            ocl_match2L1_gpu<unsigned short>, ocl_match2L1_gpu<short>,
-            ocl_match2L1_gpu<int>, ocl_match2L1_gpu<float>
-        },
-        {
-            0/*match2L2_gpu<unsigned char>*/, 0/*match2L2_gpu<signed char>*/,
-            0/*match2L2_gpu<unsigned short>*/, 0/*match2L2_gpu<short>*/,
-            0/*match2L2_gpu<int>*/, ocl_match2L2_gpu<float>
-        },
-        {
-            ocl_match2Hamming_gpu<unsigned char>, 0/*match2Hamming_gpu<signed char>*/,
-            ocl_match2Hamming_gpu<unsigned short>, 0/*match2Hamming_gpu<short>*/,
-            ocl_match2Hamming_gpu<int>, 0/*match2Hamming_gpu<float>*/
-        }
-    };
-#endif
+    // typedef void (*caller_t)(const oclMat & query, const oclMat & trains, const oclMat & masks,
+    //                          const oclMat & trainIdx, const oclMat & imgIdx, const oclMat & distance);
+
     CV_Assert(query.channels() == 1 && query.depth() < CV_64F);
 
     const int nQuery = query.rows;
 
-    trainIdx.create(1, nQuery, CV_32SC2);
-    imgIdx.create(1, nQuery, CV_32SC2);
-    distance.create(1, nQuery, CV_32SC2);
+    ensureSizeIsEnough(1, nQuery, CV_32SC2, trainIdx);
+    ensureSizeIsEnough(1, nQuery, CV_32SC2, imgIdx);
+    ensureSizeIsEnough(1, nQuery, CV_32FC2, distance);
 
     trainIdx.setTo(Scalar::all(-1));
 
@@ -1393,8 +942,6 @@ namespace
 void cv::ocl::BruteForceMatcher_OCL_base::knnMatch(const oclMat &query, std::vector< std::vector<DMatch> > &matches, int k,
         const std::vector<oclMat> &masks, bool compactResult)
 {
-
-
     if (k == 2)
     {
         oclMat trainCollection;
@@ -1417,7 +964,7 @@ void cv::ocl::BruteForceMatcher_OCL_base::knnMatch(const oclMat &query, std::vec
         temp.reserve(2 * k);
 
         matches.resize(query.rows);
-        std::for_each(matches.begin(), matches.end(), std::bind2nd(std::mem_fun_ref(&std::vector<DMatch>::reserve), k));
+        for_each(matches.begin(), matches.end(), bind2nd(mem_fun_ref(&std::vector<DMatch>::reserve), k));
 
         for (size_t imgIdx = 0, size = trainDescCollection.size(); imgIdx < size; ++imgIdx)
         {
@@ -1441,7 +988,7 @@ void cv::ocl::BruteForceMatcher_OCL_base::knnMatch(const oclMat &query, std::vec
 
         if (compactResult)
         {
-            std::vector< std::vector<DMatch> >::iterator new_end = std::remove_if(matches.begin(), matches.end(), std::mem_fun_ref(&std::vector<DMatch>::empty));
+            std::vector< std::vector<DMatch> >::iterator new_end = remove_if(matches.begin(), matches.end(), mem_fun_ref(&std::vector<DMatch>::empty));
             matches.erase(new_end, matches.end());
         }
     }
@@ -1454,50 +1001,25 @@ void cv::ocl::BruteForceMatcher_OCL_base::radiusMatchSingle(const oclMat &query,
     if (query.empty() || train.empty())
         return;
 
-    typedef void (*caller_t)(const oclMat & query, const oclMat & train, float maxDistance, const oclMat & mask,
-                             const oclMat & trainIdx, const oclMat & distance, const oclMat & nMatches);
-
-    //#if 0
-    static const caller_t callers[3][6] =
-    {
-        {
-            ocl_matchL1_gpu<unsigned char>, 0/*ocl_matchL1_gpu<signed char>*/,
-            ocl_matchL1_gpu<unsigned short>, ocl_matchL1_gpu<short>,
-            ocl_matchL1_gpu<int>, ocl_matchL1_gpu<float>
-        },
-        {
-            0/*ocl_matchL2_gpu<unsigned char>*/, 0/*ocl_matchL2_gpu<signed char>*/,
-            0/*ocl_matchL2_gpu<unsigned short>*/, 0/*ocl_matchL2_gpu<short>*/,
-            0/*ocl_matchL2_gpu<int>*/, ocl_matchL2_gpu<float>
-        },
-        {
-            ocl_matchHamming_gpu<unsigned char>, 0/*ocl_matchHamming_gpu<signed char>*/,
-            ocl_matchHamming_gpu<unsigned short>, 0/*ocl_matchHamming_gpu<short>*/,
-            ocl_matchHamming_gpu<int>, 0/*ocl_matchHamming_gpu<float>*/
-        }
-    };
-    //#endif
-
     const int nQuery = query.rows;
     const int nTrain = train.rows;
 
     CV_Assert(query.channels() == 1 && query.depth() < CV_64F);
     CV_Assert(train.type() == query.type() && train.cols == query.cols);
-    CV_Assert(trainIdx.empty() || (trainIdx.rows == nQuery && trainIdx.size() == distance.size()));
+    CV_Assert(trainIdx.empty() || (trainIdx.rows == query.rows && trainIdx.size() == distance.size()));
 
-    nMatches.create(1, nQuery, CV_32SC1);
+    ensureSizeIsEnough(1, nQuery, CV_32SC1, nMatches);
     if (trainIdx.empty())
     {
-        trainIdx.create(nQuery, std::max((nTrain / 100), 10), CV_32SC1);
-        distance.create(nQuery, std::max((nTrain / 100), 10), CV_32FC1);
+        ensureSizeIsEnough(nQuery, std::max((nTrain / 100), 10), CV_32SC1, trainIdx);
+        ensureSizeIsEnough(nQuery, std::max((nTrain / 100), 10), CV_32FC1, distance);
     }
 
     nMatches.setTo(Scalar::all(0));
 
-    caller_t func = callers[distType][query.depth()];
-    //CV_Assert(func != 0);
-    //func(query, train, maxDistance, mask, trainIdx, distance, nMatches, cc, StreamAccessor::getStream(stream));
-    func(query, train, maxDistance, mask, trainIdx, distance, nMatches);
+    matchDispatcher(query, train, maxDistance, mask, trainIdx, distance, nMatches, distType);
+
+    return;
 }
 
 void cv::ocl::BruteForceMatcher_OCL_base::radiusMatchDownload(const oclMat &trainIdx, const oclMat &distance, const oclMat &nMatches,
@@ -1576,9 +1098,9 @@ void cv::ocl::BruteForceMatcher_OCL_base::radiusMatchCollection(const oclMat &qu
     if (query.empty() || empty())
         return;
 
+#if 0
     typedef void (*caller_t)(const oclMat & query, const oclMat * trains, int n, float maxDistance, const oclMat * masks,
                              const oclMat & trainIdx, const oclMat & imgIdx, const oclMat & distance, const oclMat & nMatches);
-#if 0
     static const caller_t callers[3][6] =
     {
         {
diff --git a/modules/ocl/src/build_warps.cpp b/modules/ocl/src/build_warps.cpp
index 9c9e7e7e0c..dcbd6f0755 100644
--- a/modules/ocl/src/build_warps.cpp
+++ b/modules/ocl/src/build_warps.cpp
@@ -83,7 +83,7 @@ void cv::ocl::buildWarpPlaneMaps(Size /*src_size*/, Rect dst_roi, const Mat &K,
     int tl_v = dst_roi.tl().y;
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "buildWarpPlaneMaps";
+    String kernelName = "buildWarpPlaneMaps";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&map_x.data));
@@ -123,7 +123,7 @@ void cv::ocl::buildWarpCylindricalMaps(Size /*src_size*/, Rect dst_roi, const Ma
     int tl_v = dst_roi.tl().y;
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "buildWarpCylindricalMaps";
+    String kernelName = "buildWarpCylindricalMaps";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&map_x.data));
@@ -162,7 +162,7 @@ void cv::ocl::buildWarpSphericalMaps(Size /*src_size*/, Rect dst_roi, const Mat
     int tl_v = dst_roi.tl().y;
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "buildWarpSphericalMaps";
+    String kernelName = "buildWarpSphericalMaps";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&map_x.data));
@@ -205,7 +205,7 @@ void cv::ocl::buildWarpAffineMaps(const Mat &M, bool inverse, Size dsize, oclMat
     oclMat coeffsOclMat(coeffsMat.reshape(1, 1));
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "buildWarpAffineMaps";
+    String kernelName = "buildWarpAffineMaps";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&xmap.data));
@@ -244,7 +244,7 @@ void cv::ocl::buildWarpPerspectiveMaps(const Mat &M, bool inverse, Size dsize, o
     oclMat coeffsOclMat(coeffsMat.reshape(1, 1));
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "buildWarpPerspectiveMaps";
+    String kernelName = "buildWarpPerspectiveMaps";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&xmap.data));
diff --git a/modules/ocl/src/canny.cpp b/modules/ocl/src/canny.cpp
index a7871d4a3e..abe98cf590 100644
--- a/modules/ocl/src/canny.cpp
+++ b/modules/ocl/src/canny.cpp
@@ -43,9 +43,7 @@
 //
 //M*/
 
-#include <iomanip>
 #include "precomp.hpp"
-#include "mcwutil.hpp"
 
 using namespace cv;
 using namespace cv::ocl;
@@ -88,7 +86,8 @@ void cv::ocl::CannyBuf::create(const Size &image_size, int apperture_size)
             filterDY = createDerivFilter_GPU(CV_8U, CV_32S, 0, 1, apperture_size, BORDER_REPLICATE);
         }
     }
-    ensureSizeIsEnough(image_size.height + 2, image_size.width + 2, CV_32FC1, edgeBuf);
+    ensureSizeIsEnough(image_size.height + 2, image_size.width + 2, CV_32FC1, magBuf);
+    ensureSizeIsEnough(image_size.height + 2, image_size.width + 2, CV_32FC1, mapBuf);
 
     ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf1);
     ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf2);
@@ -99,7 +98,7 @@ void cv::ocl::CannyBuf::create(const Size &image_size, int apperture_size)
     {
         openCLFree(counter);
     }
-    counter = clCreateBuffer( Context::getContext()->impl->clContext, CL_MEM_COPY_HOST_PTR, sizeof(int), counter_i, &err );
+    counter = clCreateBuffer( *((cl_context*)getoclContext()), CL_MEM_COPY_HOST_PTR, sizeof(int), counter_i, &err );
     openCLSafeCall(err);
 }
 
@@ -109,10 +108,15 @@ void cv::ocl::CannyBuf::release()
     dy.release();
     dx_buf.release();
     dy_buf.release();
-    edgeBuf.release();
+    magBuf.release();
+    mapBuf.release();
     trackBuf1.release();
     trackBuf2.release();
-    openCLFree(counter);
+    if(counter)
+    {
+        openCLFree(counter);
+        counter = NULL;
+    }
 }
 
 namespace cv
@@ -142,13 +146,13 @@ namespace
     void CannyCaller(CannyBuf &buf, oclMat &dst, float low_thresh, float high_thresh)
     {
         using namespace ::cv::ocl::canny;
-        calcMap_gpu(buf.dx, buf.dy, buf.edgeBuf, buf.edgeBuf, dst.rows, dst.cols, low_thresh, high_thresh);
+        calcMap_gpu(buf.dx, buf.dy, buf.magBuf, buf.mapBuf, dst.rows, dst.cols, low_thresh, high_thresh);
 
-        edgesHysteresisLocal_gpu(buf.edgeBuf, buf.trackBuf1, buf.counter, dst.rows, dst.cols);
+        edgesHysteresisLocal_gpu(buf.mapBuf, buf.trackBuf1, buf.counter, dst.rows, dst.cols);
 
-        edgesHysteresisGlobal_gpu(buf.edgeBuf, buf.trackBuf1, buf.trackBuf2, buf.counter, dst.rows, dst.cols);
+        edgesHysteresisGlobal_gpu(buf.mapBuf, buf.trackBuf1, buf.trackBuf2, buf.counter, dst.rows, dst.cols);
 
-        getEdges_gpu(buf.edgeBuf, dst, dst.rows, dst.cols);
+        getEdges_gpu(buf.mapBuf, dst, dst.rows, dst.cols);
     }
 }
 
@@ -171,20 +175,20 @@ void cv::ocl::Canny(const oclMat &src, CannyBuf &buf, oclMat &dst, double low_th
     dst.setTo(Scalar::all(0));
 
     buf.create(src.size(), apperture_size);
-    buf.edgeBuf.setTo(Scalar::all(0));
+    buf.magBuf.setTo(Scalar::all(0));
 
     if (apperture_size == 3)
     {
         calcSobelRowPass_gpu(src, buf.dx_buf, buf.dy_buf, src.rows, src.cols);
 
-        calcMagnitude_gpu(buf.dx_buf, buf.dy_buf, buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
+        calcMagnitude_gpu(buf.dx_buf, buf.dy_buf, buf.dx, buf.dy, buf.magBuf, src.rows, src.cols, L2gradient);
     }
     else
     {
         buf.filterDX->apply(src, buf.dx);
         buf.filterDY->apply(src, buf.dy);
 
-        calcMagnitude_gpu(buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
+        calcMagnitude_gpu(buf.dx, buf.dy, buf.magBuf, src.rows, src.cols, L2gradient);
     }
     CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
 }
@@ -209,8 +213,8 @@ void cv::ocl::Canny(const oclMat &dx, const oclMat &dy, CannyBuf &buf, oclMat &d
     buf.dx = dx;
     buf.dy = dy;
     buf.create(dx.size(), -1);
-    buf.edgeBuf.setTo(Scalar::all(0));
-    calcMagnitude_gpu(buf.dx, buf.dy, buf.edgeBuf, dx.rows, dx.cols, L2gradient);
+    buf.magBuf.setTo(Scalar::all(0));
+    calcMagnitude_gpu(buf.dx, buf.dy, buf.magBuf, dx.rows, dx.cols, L2gradient);
 
     CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
 }
@@ -218,7 +222,7 @@ void cv::ocl::Canny(const oclMat &dx, const oclMat &dy, CannyBuf &buf, oclMat &d
 void canny::calcSobelRowPass_gpu(const oclMat &src, oclMat &dx_buf, oclMat &dy_buf, int rows, int cols)
 {
     Context *clCxt = src.clCxt;
-    std::string kernelName = "calcSobelRowPass";
+    String kernelName = "calcSobelRowPass";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src.data));
@@ -235,13 +239,13 @@ void canny::calcSobelRowPass_gpu(const oclMat &src, oclMat &dx_buf, oclMat &dy_b
 
     size_t globalThreads[3] = {cols, rows, 1};
     size_t localThreads[3]  = {16, 16, 1};
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void canny::calcMagnitude_gpu(const oclMat &dx_buf, const oclMat &dy_buf, oclMat &dx, oclMat &dy, oclMat &mag, int rows, int cols, bool L2Grad)
 {
     Context *clCxt = dx_buf.clCxt;
-    std::string kernelName = "calcMagnitude_buf";
+    String kernelName = "calcMagnitude_buf";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&dx_buf.data));
@@ -265,17 +269,13 @@ void canny::calcMagnitude_gpu(const oclMat &dx_buf, const oclMat &dy_buf, oclMat
     size_t globalThreads[3] = {cols, rows, 1};
     size_t localThreads[3]  = {16, 16, 1};
 
-    char build_options [15] = "";
-    if(L2Grad)
-    {
-        strcat(build_options, "-D L2GRAD");
-    }
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
+    const char * build_options = L2Grad ? "-D L2GRAD":"";
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
 }
 void canny::calcMagnitude_gpu(const oclMat &dx, const oclMat &dy, oclMat &mag, int rows, int cols, bool L2Grad)
 {
     Context *clCxt = dx.clCxt;
-    std::string kernelName = "calcMagnitude";
+    String kernelName = "calcMagnitude";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&dx.data));
@@ -293,12 +293,8 @@ void canny::calcMagnitude_gpu(const oclMat &dx, const oclMat &dy, oclMat &mag, i
     size_t globalThreads[3] = {cols, rows, 1};
     size_t localThreads[3]  = {16, 16, 1};
 
-    char build_options [15] = "";
-    if(L2Grad)
-    {
-        strcat(build_options, "-D L2GRAD");
-    }
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
+    const char * build_options = L2Grad ? "-D L2GRAD":"";
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
 }
 
 void canny::calcMap_gpu(oclMat &dx, oclMat &dy, oclMat &mag, oclMat &map, int rows, int cols, float low_thresh, float high_thresh)
@@ -326,16 +322,16 @@ void canny::calcMap_gpu(oclMat &dx, oclMat &dy, oclMat &mag, oclMat &map, int ro
 
 
     size_t globalThreads[3] = {cols, rows, 1};
-    std::string kernelName = "calcMap";
+    String kernelName = "calcMap";
     size_t localThreads[3]  = {16, 16, 1};
 
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void canny::edgesHysteresisLocal_gpu(oclMat &map, oclMat &st1, void *counter, int rows, int cols)
 {
     Context *clCxt = map.clCxt;
-    std::string kernelName = "edgesHysteresisLocal";
+    String kernelName = "edgesHysteresisLocal";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&map.data));
@@ -349,15 +345,15 @@ void canny::edgesHysteresisLocal_gpu(oclMat &map, oclMat &st1, void *counter, in
     size_t globalThreads[3] = {cols, rows, 1};
     size_t localThreads[3]  = {16, 16, 1};
 
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
 }
 
 void canny::edgesHysteresisGlobal_gpu(oclMat &map, oclMat &st1, oclMat &st2, void *counter, int rows, int cols)
 {
     unsigned int count;
-    openCLSafeCall(clEnqueueReadBuffer(Context::getContext()->impl->clCmdQueue, (cl_mem)counter, 1, 0, sizeof(float), &count, 0, NULL, NULL));
+    openCLSafeCall(clEnqueueReadBuffer(*(cl_command_queue*)getoclCommandQueue(), (cl_mem)counter, 1, 0, sizeof(float), &count, 0, NULL, NULL));
     Context *clCxt = map.clCxt;
-    std::string kernelName = "edgesHysteresisGlobal";
+    String kernelName = "edgesHysteresisGlobal";
     std::vector< std::pair<size_t, const void *> > args;
     size_t localThreads[3]  = {128, 1, 1};
 
@@ -365,7 +361,7 @@ void canny::edgesHysteresisGlobal_gpu(oclMat &map, oclMat &st1, oclMat &st2, voi
     int count_i[1] = {0};
     while(count > 0)
     {
-        openCLSafeCall(clEnqueueWriteBuffer(Context::getContext()->impl->clCmdQueue, (cl_mem)counter, 1, 0, sizeof(int), &count_i, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueWriteBuffer(*(cl_command_queue*)getoclCommandQueue(), (cl_mem)counter, 1, 0, sizeof(int), &count_i, 0, NULL, NULL));
 
         args.clear();
         size_t globalThreads[3] = {std::min(count, 65535u) * 128, DIVUP(count, 65535), 1};
@@ -379,8 +375,8 @@ void canny::edgesHysteresisGlobal_gpu(oclMat &map, oclMat &st1, oclMat &st2, voi
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&map.step));
         args.push_back( std::make_pair( sizeof(cl_int), (void *)&map.offset));
 
-        openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, DISABLE);
-        openCLSafeCall(clEnqueueReadBuffer(Context::getContext()->impl->clCmdQueue, (cl_mem)counter, 1, 0, sizeof(int), &count, 0, NULL, NULL));
+        openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
+        openCLSafeCall(clEnqueueReadBuffer(*(cl_command_queue*)getoclCommandQueue(), (cl_mem)counter, 1, 0, sizeof(int), &count, 0, NULL, NULL));
         std::swap(st1, st2);
     }
 #undef DIVUP
@@ -389,7 +385,7 @@ void canny::edgesHysteresisGlobal_gpu(oclMat &map, oclMat &st1, oclMat &st2, voi
 void canny::getEdges_gpu(oclMat &map, oclMat &dst, int rows, int cols)
 {
     Context *clCxt = map.clCxt;
-    std::string kernelName = "getEdges";
+    String kernelName = "getEdges";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&map.data));
@@ -404,5 +400,5 @@ void canny::getEdges_gpu(oclMat &map, oclMat &dst, int rows, int cols)
     size_t globalThreads[3] = {cols, rows, 1};
     size_t localThreads[3]  = {16, 16, 1};
 
-    openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
 }
diff --git a/modules/ocl/src/color.cpp b/modules/ocl/src/color.cpp
index a98549c400..13136e0d33 100644
--- a/modules/ocl/src/color.cpp
+++ b/modules/ocl/src/color.cpp
@@ -182,93 +182,93 @@ void cvtColor_caller(const oclMat &src, oclMat &dst, int code, int dcn)
     switch (code)
     {
         /*
-        case CV_BGR2BGRA: case CV_RGB2BGRA: case CV_BGRA2BGR:
-        case CV_RGBA2BGR: case CV_RGB2BGR: case CV_BGRA2RGBA:
-        case CV_BGR2BGR565: case CV_BGR2BGR555: case CV_RGB2BGR565: case CV_RGB2BGR555:
-        case CV_BGRA2BGR565: case CV_BGRA2BGR555: case CV_RGBA2BGR565: case CV_RGBA2BGR555:
-        case CV_BGR5652BGR: case CV_BGR5552BGR: case CV_BGR5652RGB: case CV_BGR5552RGB:
-        case CV_BGR5652BGRA: case CV_BGR5552BGRA: case CV_BGR5652RGBA: case CV_BGR5552RGBA:
+        case COLOR_BGR2BGRA: case COLOR_RGB2BGRA: case COLOR_BGRA2BGR:
+        case COLOR_RGBA2BGR: case COLOR_RGB2BGR: case COLOR_BGRA2RGBA:
+        case COLOR_BGR2BGR565: case COLOR_BGR2BGR555: case COLOR_RGB2BGR565: case COLOR_RGB2BGR555:
+        case COLOR_BGRA2BGR565: case COLOR_BGRA2BGR555: case COLOR_RGBA2BGR565: case COLOR_RGBA2BGR555:
+        case COLOR_BGR5652BGR: case COLOR_BGR5552BGR: case COLOR_BGR5652RGB: case COLOR_BGR5552RGB:
+        case COLOR_BGR5652BGRA: case COLOR_BGR5552BGRA: case COLOR_BGR5652RGBA: case COLOR_BGR5552RGBA:
         */
-    case CV_BGR2GRAY:
-    case CV_BGRA2GRAY:
-    case CV_RGB2GRAY:
-    case CV_RGBA2GRAY:
+    case COLOR_BGR2GRAY:
+    case COLOR_BGRA2GRAY:
+    case COLOR_RGB2GRAY:
+    case COLOR_RGBA2GRAY:
     {
         CV_Assert(scn == 3 || scn == 4);
-        bidx = code == CV_BGR2GRAY || code == CV_BGRA2GRAY ? 0 : 2;
+        bidx = code == COLOR_BGR2GRAY || code == COLOR_BGRA2GRAY ? 0 : 2;
         dst.create(sz, CV_MAKETYPE(depth, 1));
         RGB2Gray_caller(src, dst, bidx);
         break;
     }
-    case CV_GRAY2BGR:
-    case CV_GRAY2BGRA:
+    case COLOR_GRAY2BGR:
+    case COLOR_GRAY2BGRA:
     {
         CV_Assert(scn == 1);
-        dcn  = code == CV_GRAY2BGRA ? 4 : 3;
+        dcn  = code == COLOR_GRAY2BGRA ? 4 : 3;
         dst.create(sz, CV_MAKETYPE(depth, dcn));
         Gray2RGB_caller(src, dst);
         break;
     }
-    case CV_BGR2YUV:
-    case CV_RGB2YUV:
+    case COLOR_BGR2YUV:
+    case COLOR_RGB2YUV:
     {
         CV_Assert(scn == 3 || scn == 4);
-        bidx = code == CV_BGR2YUV ? 0 : 2;
+        bidx = code == COLOR_RGB2YUV ? 0 : 2;
         dst.create(sz, CV_MAKETYPE(depth, 3));
         RGB2YUV_caller(src, dst, bidx);
         break;
     }
-    case CV_YUV2BGR:
-    case CV_YUV2RGB:
+    case COLOR_YUV2BGR:
+    case COLOR_YUV2RGB:
     {
         CV_Assert(scn == 3 || scn == 4);
-        bidx = code == CV_YUV2BGR ? 0 : 2;
+        bidx = code == COLOR_YUV2RGB ? 0 : 2;
         dst.create(sz, CV_MAKETYPE(depth, 3));
         YUV2RGB_caller(src, dst, bidx);
         break;
     }
-    case CV_YUV2RGB_NV12:
-    case CV_YUV2BGR_NV12:
-    case CV_YUV2RGBA_NV12:
-    case CV_YUV2BGRA_NV12:
+    case COLOR_YUV2RGB_NV12:
+    case COLOR_YUV2BGR_NV12:
+    case COLOR_YUV2RGBA_NV12:
+    case COLOR_YUV2BGRA_NV12:
     {
         CV_Assert(scn == 1);
         CV_Assert( sz.width % 2 == 0 && sz.height % 3 == 0 && depth == CV_8U );
-        dcn  = code == CV_YUV2BGRA_NV12 || code == CV_YUV2RGBA_NV12 ? 4 : 3;
-        bidx = code == CV_YUV2BGRA_NV12 || code == CV_YUV2BGR_NV12 ? 0 : 2;
+        dcn  = code == COLOR_YUV2BGRA_NV12 || code == COLOR_YUV2RGBA_NV12 ? 4 : 3;
+        bidx = code == COLOR_YUV2BGRA_NV12 || code == COLOR_YUV2BGR_NV12 ? 0 : 2;
 
         Size dstSz(sz.width, sz.height * 2 / 3);
         dst.create(dstSz, CV_MAKETYPE(depth, dcn));
         YUV2RGB_NV12_caller(src, dst, bidx);
         break;
     }
-    case CV_BGR2YCrCb:
-    case CV_RGB2YCrCb:
+    case COLOR_BGR2YCrCb:
+    case COLOR_RGB2YCrCb:
     {
         CV_Assert(scn == 3 || scn == 4);
-        bidx = code == CV_BGR2YCrCb ? 0 : 2;
+        bidx = code == COLOR_BGR2YCrCb ? 0 : 2;
         dst.create(sz, CV_MAKETYPE(depth, 3));
         RGB2YCrCb_caller(src, dst, bidx);
         break;
     }
-    case CV_YCrCb2BGR:
-    case CV_YCrCb2RGB:
+    case COLOR_YCrCb2BGR:
+    case COLOR_YCrCb2RGB:
     {
         break;
     }
     /*
-    case CV_BGR5652GRAY: case CV_BGR5552GRAY:
-    case CV_GRAY2BGR565: case CV_GRAY2BGR555:
-    case CV_BGR2YCrCb: case CV_RGB2YCrCb:
-    case CV_BGR2XYZ: case CV_RGB2XYZ:
-    case CV_XYZ2BGR: case CV_XYZ2RGB:
-    case CV_BGR2HSV: case CV_RGB2HSV: case CV_BGR2HSV_FULL: case CV_RGB2HSV_FULL:
-    case CV_BGR2HLS: case CV_RGB2HLS: case CV_BGR2HLS_FULL: case CV_RGB2HLS_FULL:
-    case CV_HSV2BGR: case CV_HSV2RGB: case CV_HSV2BGR_FULL: case CV_HSV2RGB_FULL:
-    case CV_HLS2BGR: case CV_HLS2RGB: case CV_HLS2BGR_FULL: case CV_HLS2RGB_FULL:
+    case COLOR_BGR5652GRAY: case COLOR_BGR5552GRAY:
+    case COLOR_GRAY2BGR565: case COLOR_GRAY2BGR555:
+    case COLOR_BGR2YCrCb: case COLOR_RGB2YCrCb:
+    case COLOR_BGR2XYZ: case COLOR_RGB2XYZ:
+    case COLOR_XYZ2BGR: case COLOR_XYZ2RGB:
+    case COLOR_BGR2HSV: case COLOR_RGB2HSV: case COLOR_BGR2HSV_FULL: case COLOR_RGB2HSV_FULL:
+    case COLOR_BGR2HLS: case COLOR_RGB2HLS: case COLOR_BGR2HLS_FULL: case COLOR_RGB2HLS_FULL:
+    case COLOR_HSV2BGR: case COLOR_HSV2RGB: case COLOR_HSV2BGR_FULL: case COLOR_HSV2RGB_FULL:
+    case COLOR_HLS2BGR: case COLOR_HLS2RGB: case COLOR_HLS2BGR_FULL: case COLOR_HLS2RGB_FULL:
     */
     default:
-        CV_Error( CV_StsBadFlag, "Unknown/unsupported color conversion code" );
+        CV_Error(Error::StsBadFlag, "Unknown/unsupported color conversion code" );
     }
 }
 }
diff --git a/modules/ocl/src/columnsum.cpp b/modules/ocl/src/columnsum.cpp
index 397c5356a5..a336ec754a 100644
--- a/modules/ocl/src/columnsum.cpp
+++ b/modules/ocl/src/columnsum.cpp
@@ -65,7 +65,7 @@ void cv::ocl::columnSum(const oclMat &src, oclMat &dst)
 
     Context *clCxt = src.clCxt;
 
-    const std::string kernelName = "columnSum";
+    const String kernelName = "columnSum";
 
     std::vector< std::pair<size_t, const void *> > args;
 
@@ -81,4 +81,4 @@ void cv::ocl::columnSum(const oclMat &src, oclMat &dst)
 
     openCLExecuteKernel(clCxt, &imgproc_columnsum, kernelName, globalThreads, localThreads, args, src.channels(), src.depth());
 
-}
\ No newline at end of file
+}
diff --git a/modules/ocl/src/error.cpp b/modules/ocl/src/error.cpp
index e854e70cd0..55bcd0a23f 100644
--- a/modules/ocl/src/error.cpp
+++ b/modules/ocl/src/error.cpp
@@ -170,21 +170,5 @@ namespace cv
             sprintf(buf, "%d", err);
             return buf;
         }
-
-        void error(const char *error_string, const char *file, const int line, const char *func)
-        {
-            int code = CV_GpuApiCallError;
-
-            if (std::uncaught_exception())
-            {
-                const char *errorStr = cvErrorStr(code);
-                const char *function = func ? func : "unknown function";
-
-                std::cerr << "OpenCV Error: " << errorStr << "(" << error_string << ") in " << function << ", file " << file << ", line " << line;
-                std::cerr.flush();
-            }
-            else
-                cv::error( cv::Exception(code, error_string, func, file, line) );
-        }
     }
 }
diff --git a/modules/ocl/src/fft.cpp b/modules/ocl/src/fft.cpp
index 076fab9874..40bfa257f8 100644
--- a/modules/ocl/src/fft.cpp
+++ b/modules/ocl/src/fft.cpp
@@ -51,7 +51,7 @@ using namespace cv::ocl;
 #if !defined HAVE_CLAMDFFT
 void cv::ocl::dft(const oclMat&, oclMat&, Size, int)
 {
-    CV_Error(CV_StsNotImplemented, "OpenCL DFT is not implemented");
+    CV_Error(Error::StsNotImplemented, "OpenCL DFT is not implemented");
 }
 namespace cv { namespace ocl {
     void fft_teardown();
@@ -180,19 +180,19 @@ cv::ocl::FftPlan::FftPlan(Size _dft_size, int _src_step, int _dst_step, int _fla
     case C2C:
         inLayout        = CLFFT_COMPLEX_INTERLEAVED;
         outLayout       = CLFFT_COMPLEX_INTERLEAVED;
-        clStridesIn[1]  = src_step / sizeof(std::complex<float>);
+        clStridesIn[1]  = src_step / (2*sizeof(float));
         clStridesOut[1] = clStridesIn[1];
         break;
     case R2C:
         inLayout        = CLFFT_REAL;
         outLayout       = CLFFT_HERMITIAN_INTERLEAVED;
         clStridesIn[1]  = src_step / sizeof(float);
-        clStridesOut[1] = dst_step / sizeof(std::complex<float>);
+        clStridesOut[1] = dst_step / (2*sizeof(float));
         break;
     case C2R:
         inLayout        = CLFFT_HERMITIAN_INTERLEAVED;
         outLayout       = CLFFT_REAL;
-        clStridesIn[1]  = src_step / sizeof(std::complex<float>);
+        clStridesIn[1]  = src_step / (2*sizeof(float));
         clStridesOut[1] = dst_step / sizeof(float);
         break;
     default:
@@ -205,7 +205,7 @@ cv::ocl::FftPlan::FftPlan(Size _dft_size, int _src_step, int _dst_step, int _fla
     clStridesIn[2]  = is_row_dft ? clStridesIn[1]  : dft_size.width * clStridesIn[1];
     clStridesOut[2] = is_row_dft ? clStridesOut[1] : dft_size.width * clStridesOut[1];
 
-    openCLSafeCall( clAmdFftCreateDefaultPlan( &plHandle, Context::getContext()->impl->clContext, dim, clLengthsIn ) );
+    openCLSafeCall( clAmdFftCreateDefaultPlan( &plHandle, *(cl_context*)getoclContext(), dim, clLengthsIn ) );
 
     openCLSafeCall( clAmdFftSetResultLocation( plHandle, CLFFT_OUTOFPLACE ) );
     openCLSafeCall( clAmdFftSetLayout( plHandle, inLayout, outLayout ) );
@@ -219,7 +219,7 @@ cv::ocl::FftPlan::FftPlan(Size _dft_size, int _src_step, int _dst_step, int _fla
     openCLSafeCall( clAmdFftSetPlanScale  ( plHandle, is_inverse ? CLFFT_BACKWARD : CLFFT_FORWARD, scale_ ) );
 
     //ready to bake
-    openCLSafeCall( clAmdFftBakePlan( plHandle, 1, &(Context::getContext()->impl->clCmdQueue), NULL, NULL ) );
+    openCLSafeCall( clAmdFftBakePlan( plHandle, 1, (cl_command_queue*)getoclCommandQueue(), NULL, NULL ) );
 }
 cv::ocl::FftPlan::~FftPlan()
 {
@@ -337,16 +337,17 @@ void cv::ocl::dft(const oclMat &src, oclMat &dst, Size dft_size, int flags)
     if (buffersize)
     {
         cl_int medstatus;
-        clMedBuffer = clCreateBuffer ( src.clCxt->impl->clContext, CL_MEM_READ_WRITE, buffersize, 0, &medstatus);
+        clMedBuffer = clCreateBuffer ( (cl_context)src.clCxt->oclContext(), CL_MEM_READ_WRITE, buffersize, 0, &medstatus);
         openCLSafeCall( medstatus );
     }
+    cl_command_queue clq = (cl_command_queue)src.clCxt->oclCommandQueue();
     openCLSafeCall( clAmdFftEnqueueTransform( plHandle,
         is_inverse ? CLFFT_BACKWARD : CLFFT_FORWARD,
         1,
-        &src.clCxt->impl->clCmdQueue,
+        &clq,
         0, NULL, NULL,
         (cl_mem *)&src.data, (cl_mem *)&dst.data, clMedBuffer ) );
-    openCLSafeCall( clFinish(src.clCxt->impl->clCmdQueue) );
+    openCLSafeCall( clFinish(clq) );
     if(clMedBuffer)
     {
         openCLFree(clMedBuffer);
diff --git a/modules/ocl/src/filtering.cpp b/modules/ocl/src/filtering.cpp
index 61e899c17e..bbbdf8bc1b 100644
--- a/modules/ocl/src/filtering.cpp
+++ b/modules/ocl/src/filtering.cpp
@@ -48,8 +48,7 @@
 //M*/
 
 #include "precomp.hpp"
-#include "mcwutil.hpp"
-#include <iostream>
+
 using namespace cv;
 using namespace cv::ocl;
 
@@ -64,6 +63,7 @@ extern const char *filter_sep_row;
 extern const char *filter_sep_col;
 extern const char *filtering_laplacian;
 extern const char *filtering_morph;
+extern const char *filtering_adaptive_bilateral;
 }
 }
 
@@ -164,7 +164,7 @@ public:
 
 Ptr<FilterEngine_GPU> cv::ocl::createFilter2D_GPU(const Ptr<BaseFilter_GPU> filter2D)
 {
-    return Ptr<FilterEngine_GPU>(new Filter2DEngine_GPU(filter2D));
+    return makePtr<Filter2DEngine_GPU>(filter2D);
 }
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -194,7 +194,7 @@ public:
 
 namespace
 {
-typedef void (*GPUMorfFilter_t)(const oclMat & , oclMat & , oclMat & , Size &, const Point, bool rectKernel, bool usrROI);
+typedef void (*GPUMorfFilter_t)(const oclMat & , oclMat & , oclMat & , Size &, const Point, bool rectKernel);
 
 class MorphFilter_GPU : public BaseFilter_GPU
 {
@@ -204,7 +204,7 @@ public:
 
     virtual void operator()(const oclMat &src, oclMat &dst)
     {
-        func(src, dst, kernel, ksize, anchor, rectKernel, false) ;
+        func(src, dst, kernel, ksize, anchor, rectKernel) ;
     }
 
     oclMat kernel;
@@ -219,7 +219,7 @@ public:
 **Note that the kernel need to be further refined.
 */
 static void GPUErode(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
-                         Size &ksize, const Point anchor, bool rectKernel, bool useROI)
+                         Size &ksize, const Point anchor, bool rectKernel)
 {
     //Normalize the result by default
     //float alpha = ksize.height * ksize.width;
@@ -236,7 +236,7 @@ static void GPUErode(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
     int srcOffset_x = srcOffset % srcStep;
     int srcOffset_y = srcOffset / srcStep;
     Context *clCxt = src.clCxt;
-    std::string kernelName;
+    String kernelName;
     size_t localThreads[3] = {16, 16, 1};
     size_t globalThreads[3] = {(src.cols + localThreads[0] - 1) / localThreads[0] *localThreads[0], (src.rows + localThreads[1] - 1) / localThreads[1] *localThreads[1], 1};
 
@@ -271,14 +271,13 @@ static void GPUErode(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
         sprintf(s, "-D VAL=FLT_MAX -D GENTYPE=float4");
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unsupported type");
+        CV_Error(Error::StsUnsupportedFormat, "unsupported type");
     }
 
     char compile_option[128];
-    sprintf(compile_option, "-D RADIUSX=%d -D RADIUSY=%d -D LSIZE0=%d -D LSIZE1=%d -D ERODE %s %s %s",
+    sprintf(compile_option, "-D RADIUSX=%d -D RADIUSY=%d -D LSIZE0=%d -D LSIZE1=%d -D ERODE %s %s",
         anchor.x, anchor.y, (int)localThreads[0], (int)localThreads[1],
         rectKernel?"-D RECTKERNEL":"",
-        useROI?"-D USEROI":"",
         s);
     std::vector< std::pair<size_t, const void *> > args;
     args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
@@ -299,7 +298,7 @@ static void GPUErode(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
 
 //! data type supported: CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4
 static void GPUDilate(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
-                          Size &ksize, const Point anchor, bool rectKernel, bool useROI)
+                          Size &ksize, const Point anchor, bool rectKernel)
 {
     //Normalize the result by default
     //float alpha = ksize.height * ksize.width;
@@ -316,7 +315,7 @@ static void GPUDilate(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
     int srcOffset_x = srcOffset % srcStep;
     int srcOffset_y = srcOffset / srcStep;
     Context *clCxt = src.clCxt;
-    std::string kernelName;
+    String kernelName;
     size_t localThreads[3] = {16, 16, 1};
     size_t globalThreads[3] = {(src.cols + localThreads[0] - 1) / localThreads[0] *localThreads[0],
                                (src.rows + localThreads[1] - 1) / localThreads[1] *localThreads[1], 1};
@@ -352,15 +351,13 @@ static void GPUDilate(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
         sprintf(s, "-D VAL=-FLT_MAX -D GENTYPE=float4");
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unsupported type");
+        CV_Error(Error::StsUnsupportedFormat, "unsupported type");
     }
 
     char compile_option[128];
-    sprintf(compile_option, "-D RADIUSX=%d -D RADIUSY=%d -D LSIZE0=%d -D LSIZE1=%d -D DILATE %s %s %s",
+    sprintf(compile_option, "-D RADIUSX=%d -D RADIUSY=%d -D LSIZE0=%d -D LSIZE1=%d -D DILATE %s %s",
         anchor.x, anchor.y, (int)localThreads[0], (int)localThreads[1],
-        rectKernel?"-D RECTKERNEL":"",
-        useROI?"-D USEROI":"",
-        s);
+        s, rectKernel?"-D RECTKERNEL":"");
     std::vector< std::pair<size_t, const void *> > args;
     args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
     args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
@@ -456,7 +453,7 @@ Ptr<FilterEngine_GPU> cv::ocl::createMorphologyFilter_GPU(int op, int type, cons
 
     Ptr<BaseFilter_GPU> filter2D = getMorphologyFilter_GPU(op, type, kernel, ksize, anchor);
 
-    return Ptr<FilterEngine_GPU>(new MorphologyFilterEngine_GPU(filter2D, iterations));
+    return makePtr<MorphologyFilterEngine_GPU>(filter2D, iterations);
 }
 
 namespace
@@ -465,7 +462,7 @@ void morphOp(int op, const oclMat &src, oclMat &dst, const Mat &_kernel, Point a
 {
     if ((borderType != cv::BORDER_CONSTANT) || (borderValue != morphologyDefaultBorderValue()))
     {
-        CV_Error(CV_StsBadArg, "unsupported border type");
+        CV_Error(Error::StsBadArg, "unsupported border type");
     }
 
     Mat kernel;
@@ -547,27 +544,27 @@ void cv::ocl::morphologyEx(const oclMat &src, oclMat &dst, int op, const Mat &ke
         erode(src, temp, kernel, anchor, iterations, borderType, borderValue);
         dilate(temp, dst, kernel, anchor, iterations, borderType, borderValue);
         break;
-    case CV_MOP_CLOSE:
+    case MORPH_CLOSE:
         dilate(src, temp, kernel, anchor, iterations, borderType, borderValue);
         erode(temp, dst, kernel, anchor, iterations, borderType, borderValue);
         break;
-    case CV_MOP_GRADIENT:
+    case MORPH_GRADIENT:
         erode(src, temp, kernel, anchor, iterations, borderType, borderValue);
         dilate(src, dst, kernel, anchor, iterations, borderType, borderValue);
         subtract(dst, temp, dst);
         break;
-    case CV_MOP_TOPHAT:
+    case MORPH_TOPHAT:
         erode(src, dst, kernel, anchor, iterations, borderType, borderValue);
         dilate(dst, temp, kernel, anchor, iterations, borderType, borderValue);
         subtract(src, temp, dst);
         break;
-    case CV_MOP_BLACKHAT:
+    case MORPH_BLACKHAT:
         dilate(src, dst, kernel, anchor, iterations, borderType, borderValue);
         erode(dst, temp, kernel, anchor, iterations, borderType, borderValue);
         subtract(temp, src, dst);
         break;
     default:
-        CV_Error(CV_StsBadArg, "unknown morphological operation");
+        CV_Error(Error::StsBadArg, "unknown morphological operation");
     }
 }
 
@@ -576,7 +573,7 @@ void cv::ocl::morphologyEx(const oclMat &src, oclMat &dst, int op, const Mat &ke
 
 namespace
 {
-typedef void (*GPUFilter2D_t)(const oclMat & , oclMat & , oclMat & , Size &, const Point, const int);
+typedef void (*GPUFilter2D_t)(const oclMat & , oclMat & , const oclMat & , const Size &, const Point&, const int);
 
 class LinearFilter_GPU : public BaseFilter_GPU
 {
@@ -595,21 +592,22 @@ public:
 };
 }
 
-static void GPUFilter2D(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
-                 Size &ksize, const Point anchor, const int borderType)
+static void GPUFilter2D(const oclMat &src, oclMat &dst, const oclMat &mat_kernel,
+    const Size &ksize, const Point& anchor, const int borderType)
 {
     CV_Assert(src.clCxt == dst.clCxt);
     CV_Assert((src.cols == dst.cols) &&
               (src.rows == dst.rows));
     CV_Assert((src.oclchannels() == dst.oclchannels()));
-    CV_Assert((borderType != 0));
     CV_Assert(ksize.height > 0 && ksize.width > 0 && ((ksize.height & 1) == 1) && ((ksize.width & 1) == 1));
     CV_Assert((anchor.x == -1 && anchor.y == -1) || (anchor.x == ksize.width >> 1 && anchor.y == ksize.height >> 1));
+    CV_Assert(ksize.width == ksize.height);
     Context *clCxt = src.clCxt;
-    int cn =  src.oclchannels();
-    int depth = src.depth();
 
-    std::string kernelName = "filter2D";
+    int filterWidth = ksize.width;
+    bool ksize_3x3 = filterWidth == 3 && src.type() != CV_32FC4; // CV_32FC4 is not tuned up with filter2d_3x3 kernel
+
+    String kernelName = ksize_3x3 ? "filter2D_3x3" : "filter2D";
 
     size_t src_offset_x = (src.offset % src.step) / src.elemSize();
     size_t src_offset_y = src.offset / src.step;
@@ -617,54 +615,84 @@ static void GPUFilter2D(const oclMat &src, oclMat &dst, oclMat &mat_kernel,
     size_t dst_offset_x = (dst.offset % dst.step) / dst.elemSize();
     size_t dst_offset_y = dst.offset / dst.step;
 
+    int paddingPixels = filterWidth & (-2);
+
+    size_t localThreads[3]  = {ksize_3x3 ? 256 : 16, ksize_3x3 ? 1 : 16, 1};
+    size_t globalThreads[3] = {src.wholecols, src.wholerows, 1};
+
+    int cn =  src.oclchannels();
+    int src_step = (int)(src.step/src.elemSize());
+    int dst_step = (int)(dst.step/src.elemSize());
+
+    int localWidth = localThreads[0] + paddingPixels;
+    int localHeight = localThreads[1] + paddingPixels;
+
+    // 260 = divup((localThreads[0] + filterWidth * 2), 4) * 4
+    // 6   = (ROWS_PER_GROUP_WHICH_IS_4 + filterWidth * 2)
+    size_t localMemSize = ksize_3x3 ? 260 * 6 * src.elemSize() : (localWidth * localHeight) * src.elemSize();
+
     int vector_lengths[4][7] = {{4, 4, 4, 4, 4, 4, 4},
-        {4, 4, 1, 1, 1, 1, 1},
-        {1, 1, 1, 1, 1, 1, 1},
-        {4, 4, 4, 4, 1, 1, 4}
+    {4, 4, 1, 1, 1, 1, 1},
+    {1, 1, 1, 1, 1, 1, 1},
+    {4, 4, 4, 4, 1, 1, 4}
     };
-
-    int vector_length = vector_lengths[cn - 1][depth];
-    int offset_cols = (dst_offset_x) & (vector_length - 1);
-    int cols = dst.cols + offset_cols;
-    int rows = divUp(dst.rows, vector_length);
-
-    size_t localThreads[3] = {256, 1, 1};
-    size_t globalThreads[3] = { divUp(cols, localThreads[0]) *localThreads[0],
-                                divUp(rows, localThreads[1]) *localThreads[1], 1
-                              };
+    int cols = dst.cols + ((dst_offset_x) & (vector_lengths[cn - 1][src.depth()] - 1));
 
     std::vector< std::pair<size_t, const void *> > args;
     args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
-    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst_step));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&mat_kernel.data));
+    args.push_back(std::make_pair(localMemSize,   (void *)NULL));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholerows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholecols));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&src_offset_x));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&src_offset_y));
-    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
-    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst_offset_x));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst_offset_y));
-    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&mat_kernel.data));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.cols));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.rows));
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&cols));
-    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholecols));
-    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholerows));
-
-    openCLExecuteKernel(clCxt, &filtering_laplacian, kernelName, globalThreads, localThreads, args, cn, depth);
+    char btype[30];
+    switch (borderType)
+    {
+    case 0:
+        sprintf(btype, "BORDER_CONSTANT");
+        break;
+    case 1:
+        sprintf(btype, "BORDER_REPLICATE");
+        break;
+    case 2:
+        sprintf(btype, "BORDER_REFLECT");
+        break;
+    case 3:
+        CV_Error(CV_StsUnsupportedFormat, "BORDER_WRAP is not supported!");
+        return;
+    case 4:
+        sprintf(btype, "BORDER_REFLECT_101");
+        break;
+    }
+    int type = src.depth();
+    char build_options[150];
+    sprintf(build_options, "-D %s -D IMG_C_%d_%d -D CN=%d -D FILTER_SIZE=%d", btype, cn, type, cn, ksize.width);
+    openCLExecuteKernel(clCxt, &filtering_laplacian, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
 }
+
 Ptr<BaseFilter_GPU> cv::ocl::getLinearFilter_GPU(int srcType, int dstType, const Mat &kernel, const Size &ksize,
-        Point anchor, int borderType)
+        const Point &anchor, int borderType)
 {
     static const GPUFilter2D_t GPUFilter2D_callers[] = {0, GPUFilter2D, 0, GPUFilter2D, GPUFilter2D};
 
     CV_Assert((srcType == CV_8UC1 || srcType == CV_8UC3 || srcType == CV_8UC4 || srcType == CV_32FC1 || srcType == CV_32FC3 || srcType == CV_32FC4) && dstType == srcType);
 
     oclMat gpu_krnl;
-    int nDivisor;
-    normalizeKernel(kernel, gpu_krnl, CV_32S, &nDivisor, true);
-    normalizeAnchor(anchor, ksize);
+    Point norm_archor = anchor;
+    normalizeKernel(kernel, gpu_krnl, CV_32FC1);
+    normalizeAnchor(norm_archor, ksize);
 
-    return Ptr<BaseFilter_GPU>(new LinearFilter_GPU(ksize, anchor, gpu_krnl, GPUFilter2D_callers[CV_MAT_CN(srcType)],
-                               borderType));
+    return makePtr<LinearFilter_GPU>(ksize, anchor, gpu_krnl, GPUFilter2D_callers[CV_MAT_CN(srcType)],
+        borderType);
 }
 
 Ptr<FilterEngine_GPU> cv::ocl::createLinearFilter_GPU(int srcType, int dstType, const Mat &kernel, const Point &anchor,
@@ -746,7 +774,7 @@ public:
 Ptr<FilterEngine_GPU> cv::ocl::createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU> &rowFilter,
         const Ptr<BaseColumnFilter_GPU> &columnFilter)
 {
-    return Ptr<FilterEngine_GPU>(new SeparableFilterEngine_GPU(rowFilter, columnFilter));
+    return makePtr<SeparableFilterEngine_GPU>(rowFilter, columnFilter);
 }
 
 /*
@@ -765,7 +793,7 @@ static void GPUFilterBox_8u_C1R(const oclMat &src, oclMat &dst,
               (src.rows == dst.rows));
     Context *clCxt = src.clCxt;
 
-    std::string kernelName = "boxFilter_C1_D0";
+    String kernelName = "boxFilter_C1_D0";
 
     char btype[30];
 
@@ -781,7 +809,7 @@ static void GPUFilterBox_8u_C1R(const oclMat &src, oclMat &dst,
         sprintf(btype, "BORDER_REFLECT");
         break;
     case 3:
-        CV_Error(CV_StsUnsupportedFormat, "BORDER_WRAP is not supported!");
+        CV_Error(Error::StsUnsupportedFormat, "BORDER_WRAP is not supported!");
         return;
     case 4:
         sprintf(btype, "BORDER_REFLECT_101");
@@ -827,7 +855,7 @@ static void GPUFilterBox_8u_C4R(const oclMat &src, oclMat &dst,
               (src.rows == dst.rows));
     Context *clCxt = src.clCxt;
 
-    std::string kernelName = "boxFilter_C4_D0";
+    String kernelName = "boxFilter_C4_D0";
 
     char btype[30];
 
@@ -843,7 +871,7 @@ static void GPUFilterBox_8u_C4R(const oclMat &src, oclMat &dst,
         sprintf(btype, "BORDER_REFLECT");
         break;
     case 3:
-        CV_Error(CV_StsUnsupportedFormat, "BORDER_WRAP is not supported!");
+        CV_Error(Error::StsUnsupportedFormat, "BORDER_WRAP is not supported!");
         return;
     case 4:
         sprintf(btype, "BORDER_REFLECT_101");
@@ -889,7 +917,7 @@ static void GPUFilterBox_32F_C1R(const oclMat &src, oclMat &dst,
               (src.rows == dst.rows));
     Context *clCxt = src.clCxt;
 
-    std::string kernelName = "boxFilter_C1_D5";
+    String kernelName = "boxFilter_C1_D5";
 
     char btype[30];
 
@@ -905,7 +933,7 @@ static void GPUFilterBox_32F_C1R(const oclMat &src, oclMat &dst,
         sprintf(btype, "BORDER_REFLECT");
         break;
     case 3:
-        CV_Error(CV_StsUnsupportedFormat, "BORDER_WRAP is not supported!");
+        CV_Error(Error::StsUnsupportedFormat, "BORDER_WRAP is not supported!");
         return;
     case 4:
         sprintf(btype, "BORDER_REFLECT_101");
@@ -952,7 +980,7 @@ static void GPUFilterBox_32F_C4R(const oclMat &src, oclMat &dst,
               (src.rows == dst.rows));
     Context *clCxt = src.clCxt;
 
-    std::string kernelName = "boxFilter_C4_D5";
+    String kernelName = "boxFilter_C4_D5";
 
     char btype[30];
 
@@ -968,7 +996,7 @@ static void GPUFilterBox_32F_C4R(const oclMat &src, oclMat &dst,
         sprintf(btype, "BORDER_REFLECT");
         break;
     case 3:
-        CV_Error(CV_StsUnsupportedFormat, "BORDER_WRAP is not supported!");
+        CV_Error(Error::StsUnsupportedFormat, "BORDER_WRAP is not supported!");
         return;
     case 4:
         sprintf(btype, "BORDER_REFLECT_101");
@@ -1017,8 +1045,8 @@ Ptr<BaseFilter_GPU> cv::ocl::getBoxFilter_GPU(int srcType, int dstType,
 
     normalizeAnchor(anchor, ksize);
 
-    return Ptr<BaseFilter_GPU>(new GPUBoxFilter(ksize, anchor,
-                               borderType, FilterBox_callers[(CV_MAT_DEPTH(srcType) == CV_32F)][CV_MAT_CN(srcType)]));
+    return makePtr<GPUBoxFilter>(ksize, anchor,
+        borderType, FilterBox_callers[(CV_MAT_DEPTH(srcType) == CV_32F)][CV_MAT_CN(srcType)]);
 }
 
 Ptr<FilterEngine_GPU> cv::ocl::createBoxFilter_GPU(int srcType, int dstType,
@@ -1098,7 +1126,7 @@ void linearRowFilter_gpu(const oclMat &src, const oclMat &dst, oclMat mat_kernel
     int channels = src.oclchannels();
 
     size_t localThreads[3] = {16, 16, 1};
-    std::string kernelName = "row_filter";
+    String kernelName = "row_filter";
 
     char btype[30];
 
@@ -1176,7 +1204,7 @@ void linearRowFilter_gpu(const oclMat &src, const oclMat &dst, oclMat mat_kernel
     args.push_back(std::make_pair(sizeof(cl_int), (void *)&ridusy));
     args.push_back(std::make_pair(sizeof(cl_mem), (void *)&mat_kernel.data));
 
-    openCLExecuteKernel2(clCxt, &filter_sep_row, kernelName, globalThreads, localThreads, args, channels, src.depth(), compile_option, CLFLUSH);
+    openCLExecuteKernel(clCxt, &filter_sep_row, kernelName, globalThreads, localThreads, args, channels, src.depth(), compile_option);
 }
 
 Ptr<BaseRowFilter_GPU> cv::ocl::getLinearRowFilter_GPU(int srcType, int /*bufType*/, const Mat &rowKernel, int anchor, int bordertype)
@@ -1201,8 +1229,8 @@ Ptr<BaseRowFilter_GPU> cv::ocl::getLinearRowFilter_GPU(int srcType, int /*bufTyp
 
     normalizeAnchor(anchor, ksize);
 
-    return Ptr<BaseRowFilter_GPU>(new GpuLinearRowFilter(ksize, anchor, mat_kernel,
-                                  gpuFilter1D_callers[CV_MAT_DEPTH(srcType)], bordertype));
+    return makePtr<GpuLinearRowFilter>(ksize, anchor, mat_kernel,
+        gpuFilter1D_callers[CV_MAT_DEPTH(srcType)], bordertype);
 }
 
 namespace
@@ -1230,7 +1258,7 @@ void linearColumnFilter_gpu(const oclMat &src, const oclMat &dst, oclMat mat_ker
     int channels = src.oclchannels();
 
     size_t localThreads[3] = {16, 16, 1};
-    std::string kernelName = "col_filter";
+    String kernelName = "col_filter";
 
     char btype[30];
 
@@ -1370,8 +1398,8 @@ Ptr<BaseColumnFilter_GPU> cv::ocl::getLinearColumnFilter_GPU(int /*bufType*/, in
 
     normalizeAnchor(anchor, ksize);
 
-    return Ptr<BaseColumnFilter_GPU>(new GpuLinearColumnFilter(ksize, anchor, mat_kernel,
-                                     gpuFilter1D_callers[CV_MAT_DEPTH(dstType)], bordertype));
+    return makePtr<GpuLinearColumnFilter>(ksize, anchor, mat_kernel,
+        gpuFilter1D_callers[CV_MAT_DEPTH(dstType)], bordertype);
 }
 
 Ptr<FilterEngine_GPU> cv::ocl::createSeparableLinearFilter_GPU(int srcType, int dstType,
@@ -1399,7 +1427,7 @@ void cv::ocl::sepFilter2D(const oclMat &src, oclMat &dst, int ddepth, const Mat
             if ((bordertype != cv::BORDER_CONSTANT) &&
                     (bordertype != cv::BORDER_REPLICATE))
             {
-                CV_Error(CV_StsBadArg, "unsupported border type");
+                CV_Error(Error::StsBadArg, "unsupported border type");
             }
         }
     }
@@ -1480,9 +1508,9 @@ void cv::ocl::Scharr(const oclMat &src, oclMat &dst, int ddepth, int dx, int dy,
 
 void cv::ocl::Laplacian(const oclMat &src, oclMat &dst, int ddepth, int ksize, double scale)
 {
-    if (src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    if (!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.type() == CV_64F)
     {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+        CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
         return;
     }
 
@@ -1566,7 +1594,7 @@ void cv::ocl::GaussianBlur(const oclMat &src, oclMat &dst, Size ksize, double si
             if ((bordertype != cv::BORDER_CONSTANT) &&
                     (bordertype != cv::BORDER_REPLICATE))
             {
-                CV_Error(CV_StsBadArg, "unsupported border type");
+                CV_Error(Error::StsBadArg, "unsupported border type");
             }
         }
     }
@@ -1589,3 +1617,100 @@ void cv::ocl::GaussianBlur(const oclMat &src, oclMat &dst, Size ksize, double si
     Ptr<FilterEngine_GPU> f = createGaussianFilter_GPU(src.type(), ksize, sigma1, sigma2, bordertype);
     f->apply(src, dst);
 }
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Adaptive Bilateral Filter
+
+void cv::ocl::adaptiveBilateralFilter(const oclMat& src, oclMat& dst, Size ksize, double sigmaSpace, Point anchor, int borderType)
+{
+    CV_Assert((ksize.width & 1) && (ksize.height & 1));  // ksize must be odd
+    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC3);  // source must be 8bit RGB image
+    if( sigmaSpace <= 0 )
+        sigmaSpace = 1;
+    Mat lut(Size(ksize.width, ksize.height), CV_32FC1);
+    double sigma2 = sigmaSpace * sigmaSpace;
+    int idx = 0;
+    int w = ksize.width / 2;
+    int h = ksize.height / 2;
+    for(int y=-h; y<=h; y++)
+        for(int x=-w; x<=w; x++)
+    {
+        lut.at<float>(idx++) = sigma2 / (sigma2 + x * x + y * y);
+    }
+    oclMat dlut(lut);
+    int depth = src.depth();
+    int cn = src.oclchannels();
+
+    normalizeAnchor(anchor, ksize);
+    const static String kernelName = "edgeEnhancingFilter";
+
+    dst.create(src.size(), src.type());
+
+    char btype[30];
+    switch(borderType)
+    {
+    case BORDER_CONSTANT:
+        sprintf(btype, "BORDER_CONSTANT");
+        break;
+    case BORDER_REPLICATE:
+        sprintf(btype, "BORDER_REPLICATE");
+        break;
+    case BORDER_REFLECT:
+        sprintf(btype, "BORDER_REFLECT");
+        break;
+    case BORDER_WRAP:
+        sprintf(btype, "BORDER_WRAP");
+        break;
+    case BORDER_REFLECT101:
+        sprintf(btype, "BORDER_REFLECT_101");
+        break;
+    default:
+        CV_Error(CV_StsBadArg, "This border type is not supported");
+        break;
+    }
+
+    //the following constants may be adjusted for performance concerns
+    const static size_t blockSizeX = 64, blockSizeY = 1, EXTRA = ksize.height - 1;
+
+    //Normalize the result by default
+    const float alpha = ksize.height * ksize.width;
+
+    const size_t gSize = blockSizeX - ksize.width / 2 * 2;
+    const size_t globalSizeX = (src.cols) % gSize == 0 ?
+        src.cols / gSize * blockSizeX :
+        (src.cols / gSize + 1) * blockSizeX;
+    const size_t rows_per_thread = 1 + EXTRA;
+    const size_t globalSizeY = ((src.rows + rows_per_thread - 1) / rows_per_thread) % blockSizeY == 0 ?
+        ((src.rows + rows_per_thread - 1) / rows_per_thread) :
+        (((src.rows + rows_per_thread - 1) / rows_per_thread) / blockSizeY + 1) * blockSizeY;
+
+    size_t globalThreads[3] = { globalSizeX, globalSizeY, 1};
+    size_t localThreads[3]  = { blockSizeX, blockSizeY, 1};
+
+    char build_options[250];
+
+    //LDATATYPESIZE is sizeof local data store. This is to exemplify effect of LDS on kernel performance
+    sprintf(build_options,
+        "-D VAR_PER_CHANNEL=1 -D CALCVAR=1 -D FIXED_WEIGHT=0 -D EXTRA=%d"
+        " -D THREADS=%d -D anX=%d -D anY=%d -D ksX=%d -D ksY=%d -D %s",
+        static_cast<int>(EXTRA), static_cast<int>(blockSizeX), anchor.x, anchor.y, ksize.width, ksize.height, btype);
+
+    std::vector<std::pair<size_t , const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), &src.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), &dst.data));
+    args.push_back(std::make_pair(sizeof(cl_float), (void *)&alpha));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.offset));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholerows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.wholecols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.offset));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
+    args.push_back(std::make_pair(sizeof(cl_mem), &dlut.data));
+    int lut_step = dlut.step1();
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&lut_step));
+
+    openCLExecuteKernel(Context::getContext(), &filtering_adaptive_bilateral, kernelName,
+        globalThreads, localThreads, args, cn, depth, build_options);
+}
diff --git a/modules/ocl/src/gemm.cpp b/modules/ocl/src/gemm.cpp
index be7e79cce3..a74a24be19 100644
--- a/modules/ocl/src/gemm.cpp
+++ b/modules/ocl/src/gemm.cpp
@@ -46,21 +46,67 @@
 #include <iomanip>
 #include "precomp.hpp"
 
+namespace cv { namespace ocl {
+
+// used for clAmdBlas library to avoid redundant setup/teardown
+void clBlasSetup();
+void clBlasTeardown();
+
+}} /* namespace cv { namespace ocl */
+
+
 #if !defined HAVE_CLAMDBLAS
 void cv::ocl::gemm(const oclMat&, const oclMat&, double,
                    const oclMat&, double, oclMat&, int)
+{
+    CV_Error(Error::StsNotImplemented, "OpenCL BLAS is not implemented");
+}
+
+void cv::ocl::clBlasSetup()
 {
     CV_Error(CV_StsNotImplemented, "OpenCL BLAS is not implemented");
 }
+
+void cv::ocl::clBlasTeardown()
+{
+    //intentionally do nothing
+}
+
 #else
 #include "clAmdBlas.h"
 using namespace cv;
 
+static bool clBlasInitialized = false;
+static Mutex cs;
+
+void cv::ocl::clBlasSetup()
+{
+    if(!clBlasInitialized)
+    {
+        AutoLock al(cs);
+        if(!clBlasInitialized)
+        {
+            openCLSafeCall(clAmdBlasSetup());
+            clBlasInitialized = true;
+        }
+    }
+}
+
+void cv::ocl::clBlasTeardown()
+{
+    AutoLock al(cs);
+    if(clBlasInitialized)
+    {
+        clAmdBlasTeardown();
+        clBlasInitialized = false;
+    }
+}
+
 void cv::ocl::gemm(const oclMat &src1, const oclMat &src2, double alpha,
                    const oclMat &src3, double beta, oclMat &dst, int flags)
 {
     CV_Assert(src1.cols == src2.rows &&
-              (src3.empty() || src1.rows == src3.rows && src2.cols == src3.cols));
+              (src3.empty() || (src1.rows == src3.rows && src2.cols == src3.cols)));
     CV_Assert(!(cv::GEMM_3_T & flags)); // cv::GEMM_3_T is not supported
     if(!src3.empty())
     {
@@ -71,7 +117,8 @@ void cv::ocl::gemm(const oclMat &src1, const oclMat &src2, double alpha,
         dst.create(src1.rows, src2.cols, src1.type());
         dst.setTo(Scalar::all(0));
     }
-    openCLSafeCall( clAmdBlasSetup() );
+
+    clBlasSetup();
 
     const clAmdBlasTranspose transA = (cv::GEMM_1_T & flags) ? clAmdBlasTrans : clAmdBlasNoTrans;
     const clAmdBlasTranspose transB = (cv::GEMM_2_T & flags) ? clAmdBlasTrans : clAmdBlasNoTrans;
@@ -87,7 +134,7 @@ void cv::ocl::gemm(const oclMat &src1, const oclMat &src2, double alpha,
     int offb    = src2.offset;
     int offc    = dst.offset;
 
-
+    cl_command_queue clq = (cl_command_queue)src1.clCxt->oclCommandQueue();
     switch(src1.type())
     {
     case CV_32FC1:
@@ -97,11 +144,12 @@ void cv::ocl::gemm(const oclMat &src1, const oclMat &src2, double alpha,
         offa /= sizeof(float);
         offb /= sizeof(float);
         offc /= sizeof(float);
+
         openCLSafeCall
         (
             clAmdBlasSgemmEx(order, transA, transB, M, N, K,
                              alpha, (const cl_mem)src1.data, offa, lda, (const cl_mem)src2.data, offb, ldb,
-                             beta, (cl_mem)dst.data, offc, ldc, 1, &src1.clCxt->impl->clCmdQueue, 0, NULL, NULL)
+                             beta, (cl_mem)dst.data, offc, ldc, 1, &clq, 0, NULL, NULL)
         );
         break;
     case CV_64FC1:
@@ -115,46 +163,45 @@ void cv::ocl::gemm(const oclMat &src1, const oclMat &src2, double alpha,
         (
             clAmdBlasDgemmEx(order, transA, transB, M, N, K,
                              alpha, (const cl_mem)src1.data, offa, lda, (const cl_mem)src2.data, offb, ldb,
-                             beta, (cl_mem)dst.data, offc, ldc, 1, &src1.clCxt->impl->clCmdQueue, 0, NULL, NULL)
+                             beta, (cl_mem)dst.data, offc, ldc, 1, &clq, 0, NULL, NULL)
         );
         break;
     case CV_32FC2:
     {
-        lda  /= sizeof(std::complex<float>);
-        ldb  /= sizeof(std::complex<float>);
-        ldc  /= sizeof(std::complex<float>);
-        offa /= sizeof(std::complex<float>);
-        offb /= sizeof(std::complex<float>);
-        offc /= sizeof(std::complex<float>);
+        lda  /= (2*sizeof(float));
+        ldb  /= (2*sizeof(float));
+        ldc  /= (2*sizeof(float));
+        offa /= (2*sizeof(float));
+        offb /= (2*sizeof(float));
+        offc /= (2*sizeof(float));
         cl_float2 alpha_2 = {{alpha, 0}};
         cl_float2 beta_2  = {{beta, 0}};
         openCLSafeCall
         (
             clAmdBlasCgemmEx(order, transA, transB, M, N, K,
                              alpha_2, (const cl_mem)src1.data, offa, lda, (const cl_mem)src2.data, offb, ldb,
-                             beta_2, (cl_mem)dst.data, offc, ldc, 1, &src1.clCxt->impl->clCmdQueue, 0, NULL, NULL)
+                             beta_2, (cl_mem)dst.data, offc, ldc, 1, &clq, 0, NULL, NULL)
         );
     }
     break;
     case CV_64FC2:
     {
-        lda  /= sizeof(std::complex<double>);
-        ldb  /= sizeof(std::complex<double>);
-        ldc  /= sizeof(std::complex<double>);
-        offa /= sizeof(std::complex<double>);
-        offb /= sizeof(std::complex<double>);
-        offc /= sizeof(std::complex<double>);
+        lda  /= (2*sizeof(double));
+        ldb  /= (2*sizeof(double));
+        ldc  /= (2*sizeof(double));
+        offa /= (2*sizeof(double));
+        offb /= (2*sizeof(double));
+        offc /= (2*sizeof(double));
         cl_double2 alpha_2 = {{alpha, 0}};
         cl_double2 beta_2  = {{beta, 0}};
         openCLSafeCall
         (
             clAmdBlasZgemmEx(order, transA, transB, M, N, K,
                              alpha_2, (const cl_mem)src1.data, offa, lda, (const cl_mem)src2.data, offb, ldb,
-                             beta_2, (cl_mem)dst.data, offc, ldc, 1, &src1.clCxt->impl->clCmdQueue, 0, NULL, NULL)
+                             beta_2, (cl_mem)dst.data, offc, ldc, 1, &clq, 0, NULL, NULL)
         );
     }
     break;
     }
-    clAmdBlasTeardown();
 }
 #endif
diff --git a/modules/ocl/src/gftt.cpp b/modules/ocl/src/gftt.cpp
new file mode 100644
index 0000000000..0af3039b31
--- /dev/null
+++ b/modules/ocl/src/gftt.cpp
@@ -0,0 +1,350 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include <iomanip>
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::ocl;
+
+static bool use_cpu_sorter = true;
+
+namespace cv
+{
+    namespace ocl
+    {
+        ///////////////////////////OpenCL kernel strings///////////////////////////
+        extern const char *imgproc_gftt;
+    }
+}
+
+namespace
+{
+enum SortMethod
+{
+    CPU_STL,
+    BITONIC,
+    SELECTION
+};
+
+const int GROUP_SIZE = 256;
+
+template<SortMethod method>
+struct Sorter
+{
+    //typedef EigType;
+};
+
+//TODO(pengx): optimize GPU sorter's performance thus CPU sorter is removed.
+template<>
+struct Sorter<CPU_STL>
+{
+    typedef oclMat EigType;
+    static cv::Mutex cs;
+    static Mat mat_eig;
+
+    //prototype
+    static int clfloat2Gt(cl_float2 pt1, cl_float2 pt2)
+    {
+        float v1 = mat_eig.at<float>(cvRound(pt1.s[1]), cvRound(pt1.s[0]));
+        float v2 = mat_eig.at<float>(cvRound(pt2.s[1]), cvRound(pt2.s[0]));
+        return v1 > v2;
+    }
+    static void sortCorners_caller(const EigType& eig_tex, oclMat& corners, const int count)
+    {
+        cv::AutoLock lock(cs);
+        //temporarily use STL's sort function
+        Mat mat_corners = corners;
+        mat_eig = eig_tex;
+        std::sort(mat_corners.begin<cl_float2>(), mat_corners.begin<cl_float2>() + count, clfloat2Gt);
+        corners = mat_corners;
+    }
+};
+cv::Mutex Sorter<CPU_STL>::cs;
+cv::Mat   Sorter<CPU_STL>::mat_eig;
+
+template<>
+struct Sorter<BITONIC>
+{
+    typedef TextureCL EigType;
+
+    static void sortCorners_caller(const EigType& eig_tex, oclMat& corners, const int count)
+    {
+        Context * cxt = Context::getContext();
+        size_t globalThreads[3] = {count / 2, 1, 1};
+        size_t localThreads[3]  = {GROUP_SIZE, 1, 1};
+
+        // 2^numStages should be equal to count or the output is invalid
+        int numStages = 0;
+        for(int i = count; i > 1; i >>= 1)
+        {
+            ++numStages;
+        }
+        const int argc = 5;
+        std::vector< std::pair<size_t, const void *> > args(argc);
+        String kernelname = "sortCorners_bitonicSort";
+        args[0] = std::make_pair(sizeof(cl_mem), (void *)&eig_tex);
+        args[1] = std::make_pair(sizeof(cl_mem), (void *)&corners.data);
+        args[2] = std::make_pair(sizeof(cl_int), (void *)&count);
+        for(int stage = 0; stage < numStages; ++stage)
+        {
+            args[3] = std::make_pair(sizeof(cl_int), (void *)&stage);
+            for(int passOfStage = 0; passOfStage < stage + 1; ++passOfStage)
+            {
+                args[4] = std::make_pair(sizeof(cl_int), (void *)&passOfStage);
+                openCLExecuteKernel(cxt, &imgproc_gftt, kernelname, globalThreads, localThreads, args, -1, -1);
+            }
+        }
+    }
+};
+
+template<>
+struct Sorter<SELECTION>
+{
+    typedef TextureCL EigType;
+
+    static void sortCorners_caller(const EigType& eig_tex, oclMat& corners, const int count)
+    {
+        Context * cxt = Context::getContext();
+
+        size_t globalThreads[3] = {count, 1, 1};
+        size_t localThreads[3]  = {GROUP_SIZE, 1, 1};
+
+        std::vector< std::pair<size_t, const void *> > args;
+        //local
+        String kernelname = "sortCorners_selectionSortLocal";
+        int lds_size = GROUP_SIZE * sizeof(cl_float2);
+        args.push_back( std::make_pair( sizeof(cl_mem), (void*)&eig_tex) );
+        args.push_back( std::make_pair( sizeof(cl_mem), (void*)&corners.data) );
+        args.push_back( std::make_pair( sizeof(cl_int), (void*)&count) );
+        args.push_back( std::make_pair( lds_size,       (void*)NULL) );
+
+        openCLExecuteKernel(cxt, &imgproc_gftt, kernelname, globalThreads, localThreads, args, -1, -1);
+
+        //final
+        kernelname = "sortCorners_selectionSortFinal";
+        args.pop_back();
+        openCLExecuteKernel(cxt, &imgproc_gftt, kernelname, globalThreads, localThreads, args, -1, -1);
+    }
+};
+
+int findCorners_caller(
+    const TextureCL& eig,
+    const float threshold,
+    const oclMat& mask,
+    oclMat& corners,
+    const int max_count)
+{
+    std::vector<int> k;
+    Context * cxt = Context::getContext();
+
+    std::vector< std::pair<size_t, const void*> > args;
+    String kernelname = "findCorners";
+
+    const int mask_strip = mask.step / mask.elemSize1();
+
+    oclMat g_counter(1, 1, CV_32SC1);
+    g_counter.setTo(0);
+
+    args.push_back(std::make_pair( sizeof(cl_mem),   (void*)&eig  ));
+    args.push_back(std::make_pair( sizeof(cl_mem),   (void*)&mask.data ));
+    args.push_back(std::make_pair( sizeof(cl_mem),   (void*)&corners.data ));
+    args.push_back(std::make_pair( sizeof(cl_int),   (void*)&mask_strip));
+    args.push_back(std::make_pair( sizeof(cl_float), (void*)&threshold ));
+    args.push_back(std::make_pair( sizeof(cl_int), (void*)&eig.rows ));
+    args.push_back(std::make_pair( sizeof(cl_int), (void*)&eig.cols ));
+    args.push_back(std::make_pair( sizeof(cl_int), (void*)&max_count ));
+    args.push_back(std::make_pair( sizeof(cl_mem), (void*)&g_counter.data ));
+
+    size_t globalThreads[3] = {eig.cols, eig.rows, 1};
+    size_t localThreads[3]  = {16, 16, 1};
+
+    const char * opt = mask.empty() ? "" : "-D WITH_MASK";
+    openCLExecuteKernel(cxt, &imgproc_gftt, kernelname, globalThreads, localThreads, args, -1, -1, opt);
+    return std::min(Mat(g_counter).at<int>(0), max_count);
+}
+}//unnamed namespace
+
+void cv::ocl::GoodFeaturesToTrackDetector_OCL::operator ()(const oclMat& image, oclMat& corners, const oclMat& mask)
+{
+    CV_Assert(qualityLevel > 0 && minDistance >= 0 && maxCorners >= 0);
+    CV_Assert(mask.empty() || (mask.type() == CV_8UC1 && mask.size() == image.size()));
+
+    CV_DbgAssert(support_image2d());
+
+    ensureSizeIsEnough(image.size(), CV_32F, eig_);
+
+    if (useHarrisDetector)
+        cornerMinEigenVal_dxdy(image, eig_, Dx_, Dy_, blockSize, 3, harrisK);
+    else
+        cornerMinEigenVal_dxdy(image, eig_, Dx_, Dy_, blockSize, 3);
+
+    double maxVal = 0;
+    minMax_buf(eig_, 0, &maxVal, oclMat(), minMaxbuf_);
+
+    ensureSizeIsEnough(1, std::max(1000, static_cast<int>(image.size().area() * 0.05)), CV_32FC2, tmpCorners_);
+
+    Ptr<TextureCL> eig_tex = bindTexturePtr(eig_);
+    int total = findCorners_caller(
+        *eig_tex,
+        static_cast<float>(maxVal * qualityLevel),
+        mask,
+        tmpCorners_,
+        tmpCorners_.cols);
+
+    if (total == 0)
+    {
+        corners.release();
+        return;
+    }
+    if(use_cpu_sorter)
+    {
+        Sorter<CPU_STL>::sortCorners_caller(eig_, tmpCorners_, total);
+    }
+    else
+    {
+        //if total is power of 2
+        if(((total - 1) & (total)) == 0)
+        {
+            Sorter<BITONIC>::sortCorners_caller(*eig_tex, tmpCorners_, total);
+        }
+        else
+        {
+            Sorter<SELECTION>::sortCorners_caller(*eig_tex, tmpCorners_, total);
+        }
+    }
+
+    if (minDistance < 1)
+    {
+        Rect roi_range(0, 0, maxCorners > 0 ? std::min(maxCorners, total) : total, 1);
+        tmpCorners_(roi_range).copyTo(corners);
+    }
+    else
+    {
+        std::vector<Point2f> tmp(total);
+        downloadPoints(tmpCorners_, tmp);
+
+        std::vector<Point2f> tmp2;
+        tmp2.reserve(total);
+
+        const int cell_size = cvRound(minDistance);
+        const int grid_width = (image.cols + cell_size - 1) / cell_size;
+        const int grid_height = (image.rows + cell_size - 1) / cell_size;
+
+        std::vector< std::vector<Point2f> > grid(grid_width * grid_height);
+
+        for (int i = 0; i < total; ++i)
+        {
+            Point2f p = tmp[i];
+
+            bool good = true;
+
+            int x_cell = static_cast<int>(p.x / cell_size);
+            int y_cell = static_cast<int>(p.y / cell_size);
+
+            int x1 = x_cell - 1;
+            int y1 = y_cell - 1;
+            int x2 = x_cell + 1;
+            int y2 = y_cell + 1;
+
+            // boundary check
+            x1 = std::max(0, x1);
+            y1 = std::max(0, y1);
+            x2 = std::min(grid_width - 1, x2);
+            y2 = std::min(grid_height - 1, y2);
+
+            for (int yy = y1; yy <= y2; yy++)
+            {
+                for (int xx = x1; xx <= x2; xx++)
+                {
+                    std::vector<Point2f>& m = grid[yy * grid_width + xx];
+
+                    if (!m.empty())
+                    {
+                        for(size_t j = 0; j < m.size(); j++)
+                        {
+                            float dx = p.x - m[j].x;
+                            float dy = p.y - m[j].y;
+
+                            if (dx * dx + dy * dy < minDistance * minDistance)
+                            {
+                                good = false;
+                                goto break_out;
+                            }
+                        }
+                    }
+                }
+            }
+
+            break_out:
+
+            if(good)
+            {
+                grid[y_cell * grid_width + x_cell].push_back(p);
+
+                tmp2.push_back(p);
+
+                if (maxCorners > 0 && tmp2.size() == static_cast<size_t>(maxCorners))
+                    break;
+            }
+        }
+
+        corners.upload(Mat(1, static_cast<int>(tmp2.size()), CV_32FC2, &tmp2[0]));
+    }
+}
+void cv::ocl::GoodFeaturesToTrackDetector_OCL::downloadPoints(const oclMat &points, std::vector<Point2f> &points_v)
+{
+    CV_DbgAssert(points.type() == CV_32FC2);
+    points_v.resize(points.cols);
+    openCLSafeCall(clEnqueueReadBuffer(
+        *reinterpret_cast<cl_command_queue*>(getoclCommandQueue()),
+        reinterpret_cast<cl_mem>(points.data),
+        CL_TRUE,
+        0,
+        points.cols * sizeof(Point2f),
+        &points_v[0],
+        0,
+        NULL,
+        NULL));
+}
diff --git a/modules/ocl/src/haar.cpp b/modules/ocl/src/haar.cpp
index 31da6ec2de..ac6a6e0a4e 100644
--- a/modules/ocl/src/haar.cpp
+++ b/modules/ocl/src/haar.cpp
@@ -49,7 +49,6 @@
 
 #include "precomp.hpp"
 #include <stdio.h>
-#include <string>
 
 using namespace cv;
 using namespace cv::ocl;
@@ -66,9 +65,9 @@ extern const char *haarobjectdetect_scaled2;
 }
 
 /* these settings affect the quality of detection: change with care */
-#define CV_ADJUST_FEATURES 1
-#define CV_ADJUST_WEIGHTS  0
-
+#define CV_ADJUST_FEATURES  1
+#define CV_ADJUST_WEIGHTS   0
+#define CV_HAAR_FEATURE_MAX 3
 typedef int sumtype;
 typedef double sqsumtype;
 
@@ -134,46 +133,22 @@ struct CvHidHaarClassifierCascade
 };
 typedef struct
 {
-    //int rows;
-    //int ystep;
     int width_height;
-    //int height;
     int grpnumperline_totalgrp;
-    //int totalgrp;
     int imgoff;
     float factor;
 } detect_piramid_info;
-#ifdef WIN32
+#ifdef _MSC_VER
 #define _ALIGNED_ON(_ALIGNMENT) __declspec(align(_ALIGNMENT))
-typedef _ALIGNED_ON(128) struct  GpuHidHaarFeature
-{
-    _ALIGNED_ON(32) struct
-    {
-        _ALIGNED_ON(4)  int    p0 ;
-        _ALIGNED_ON(4)  int    p1 ;
-        _ALIGNED_ON(4)  int    p2 ;
-        _ALIGNED_ON(4)  int    p3 ;
-        _ALIGNED_ON(4)  float weight  ;
-    }
-    /*_ALIGNED_ON(32)*/ rect[CV_HAAR_FEATURE_MAX] ;
-}
-GpuHidHaarFeature;
-
 
 typedef _ALIGNED_ON(128) struct  GpuHidHaarTreeNode
 {
     _ALIGNED_ON(64) int p[CV_HAAR_FEATURE_MAX][4];
-    //_ALIGNED_ON(16) int p1[CV_HAAR_FEATURE_MAX] ;
-    //_ALIGNED_ON(16) int p2[CV_HAAR_FEATURE_MAX] ;
-    //_ALIGNED_ON(16) int p3[CV_HAAR_FEATURE_MAX] ;
-    /*_ALIGNED_ON(16)*/
     float weight[CV_HAAR_FEATURE_MAX] ;
-    /*_ALIGNED_ON(4)*/
     float threshold ;
-    _ALIGNED_ON(8) float alpha[2] ;
+    _ALIGNED_ON(16) float alpha[3] ;
     _ALIGNED_ON(4) int left ;
     _ALIGNED_ON(4) int right ;
-    // GpuHidHaarFeature feature __attribute__((aligned (128)));
 }
 GpuHidHaarTreeNode;
 
@@ -181,7 +156,6 @@ GpuHidHaarTreeNode;
 typedef  _ALIGNED_ON(32) struct  GpuHidHaarClassifier
 {
     _ALIGNED_ON(4) int count;
-    //CvHaarFeature* orig_feature;
     _ALIGNED_ON(8) GpuHidHaarTreeNode *node ;
     _ALIGNED_ON(8) float *alpha ;
 }
@@ -216,32 +190,16 @@ typedef _ALIGNED_ON(64) struct  GpuHidHaarClassifierCascade
     _ALIGNED_ON(4) int p2 ;
     _ALIGNED_ON(4) int p3 ;
     _ALIGNED_ON(4) float inv_window_area ;
-    // GpuHidHaarStageClassifier* stage_classifier __attribute__((aligned (8)));
 } GpuHidHaarClassifierCascade;
 #else
 #define _ALIGNED_ON(_ALIGNMENT) __attribute__((aligned(_ALIGNMENT) ))
 
-typedef struct _ALIGNED_ON(128) GpuHidHaarFeature
-{
-    struct _ALIGNED_ON(32)
-{
-    int    p0 _ALIGNED_ON(4);
-    int    p1 _ALIGNED_ON(4);
-    int    p2 _ALIGNED_ON(4);
-    int    p3 _ALIGNED_ON(4);
-    float weight  _ALIGNED_ON(4);
-}
-rect[CV_HAAR_FEATURE_MAX] _ALIGNED_ON(32);
-}
-GpuHidHaarFeature;
-
-
 typedef struct _ALIGNED_ON(128) GpuHidHaarTreeNode
 {
     int p[CV_HAAR_FEATURE_MAX][4] _ALIGNED_ON(64);
     float weight[CV_HAAR_FEATURE_MAX];// _ALIGNED_ON(16);
     float threshold;// _ALIGNED_ON(4);
-    float alpha[2] _ALIGNED_ON(8);
+    float alpha[3] _ALIGNED_ON(16);
     int left _ALIGNED_ON(4);
     int right _ALIGNED_ON(4);
 }
@@ -284,7 +242,6 @@ typedef struct _ALIGNED_ON(64) GpuHidHaarClassifierCascade
     int p2 _ALIGNED_ON(4);
     int p3 _ALIGNED_ON(4);
     float inv_window_area _ALIGNED_ON(4);
-    // GpuHidHaarStageClassifier* stage_classifier __attribute__((aligned (8)));
 } GpuHidHaarClassifierCascade;
 #endif
 
@@ -292,36 +249,6 @@ const int icv_object_win_border = 1;
 const float icv_stage_threshold_bias = 0.0001f;
 double globaltime = 0;
 
-
-// static CvHaarClassifierCascade * gpuCreateHaarClassifierCascade( int stage_count )
-// {
-//     CvHaarClassifierCascade *cascade = 0;
-
-//     int block_size = sizeof(*cascade) + stage_count * sizeof(*cascade->stage_classifier);
-
-//     if( stage_count <= 0 )
-//         CV_Error( CV_StsOutOfRange, "Number of stages should be positive" );
-
-//     cascade = (CvHaarClassifierCascade *)cvAlloc( block_size );
-//     memset( cascade, 0, block_size );
-
-//     cascade->stage_classifier = (CvHaarStageClassifier *)(cascade + 1);
-//     cascade->flags = CV_HAAR_MAGIC_VAL;
-//     cascade->count = stage_count;
-
-//     return cascade;
-// }
-
-//static int globalcounter = 0;
-
-// static void gpuReleaseHidHaarClassifierCascade( GpuHidHaarClassifierCascade **_cascade )
-// {
-//     if( _cascade && *_cascade )
-//     {
-//         cvFree( _cascade );
-//     }
-// }
-
 /* create more efficient internal representation of haar classifier cascade */
 static GpuHidHaarClassifierCascade * gpuCreateHidHaarClassifierCascade( CvHaarClassifierCascade *cascade, int *size, int *totalclassifier)
 {
@@ -331,7 +258,7 @@ static GpuHidHaarClassifierCascade * gpuCreateHidHaarClassifierCascade( CvHaarCl
     int datasize;
     int total_classifiers = 0;
     int total_nodes = 0;
-    char errorstr[100];
+    char errorstr[256];
 
     GpuHidHaarStageClassifier *stage_classifier_ptr;
     GpuHidHaarClassifier *haar_classifier_ptr;
@@ -437,24 +364,12 @@ static GpuHidHaarClassifierCascade * gpuCreateHidHaarClassifierCascade( CvHaarCl
         hid_stage_classifier->two_rects = 1;
         haar_classifier_ptr += stage_classifier->count;
 
-        /*
-        hid_stage_classifier->parent = (stage_classifier->parent == -1)
-        ? NULL : stage_classifier_ptr + stage_classifier->parent;
-        hid_stage_classifier->next = (stage_classifier->next == -1)
-        ? NULL : stage_classifier_ptr + stage_classifier->next;
-        hid_stage_classifier->child = (stage_classifier->child == -1)
-        ? NULL : stage_classifier_ptr + stage_classifier->child;
-
-        out->is_tree |= hid_stage_classifier->next != NULL;
-        */
-
         for( j = 0; j < stage_classifier->count; j++ )
         {
             CvHaarClassifier *classifier         = stage_classifier->classifier + j;
             GpuHidHaarClassifier *hid_classifier = hid_stage_classifier->classifier + j;
             int node_count = classifier->count;
 
-            //   float* alpha_ptr = (float*)(haar_node_ptr + node_count);
             float *alpha_ptr = &haar_node_ptr->alpha[0];
 
             hid_classifier->count = node_count;
@@ -481,16 +396,12 @@ static GpuHidHaarClassifierCascade * gpuCreateHidHaarClassifierCascade( CvHaarCl
                     node->p[2][3] = 0;
                     node->weight[2] = 0;
                 }
-                //   memset( &(node->feature.rect[2]), 0, sizeof(node->feature.rect[2]) );
                 else
                     hid_stage_classifier->two_rects = 0;
+
+                memcpy( node->alpha, classifier->alpha, (node_count + 1)*sizeof(alpha_ptr[0]));
+                haar_node_ptr = haar_node_ptr + 1;
             }
-
-            memcpy( alpha_ptr, classifier->alpha, (node_count + 1)*sizeof(alpha_ptr[0]));
-            haar_node_ptr = haar_node_ptr + 1;
-            // (GpuHidHaarTreeNode*)cvAlignPtr(alpha_ptr+node_count+1, sizeof(void*));
-            //   (GpuHidHaarTreeNode*)(alpha_ptr+node_count+1);
-
             out->is_stump_based &= node_count == 1;
         }
     }
@@ -503,25 +414,19 @@ static GpuHidHaarClassifierCascade * gpuCreateHidHaarClassifierCascade( CvHaarCl
 
 
 #define sum_elem_ptr(sum,row,col)  \
-	((sumtype*)CV_MAT_ELEM_PTR_FAST((sum),(row),(col),sizeof(sumtype)))
+    ((sumtype*)CV_MAT_ELEM_PTR_FAST((sum),(row),(col),sizeof(sumtype)))
 
 #define sqsum_elem_ptr(sqsum,row,col)  \
-	((sqsumtype*)CV_MAT_ELEM_PTR_FAST((sqsum),(row),(col),sizeof(sqsumtype)))
+    ((sqsumtype*)CV_MAT_ELEM_PTR_FAST((sqsum),(row),(col),sizeof(sqsumtype)))
 
 #define calc_sum(rect,offset) \
-	((rect).p0[offset] - (rect).p1[offset] - (rect).p2[offset] + (rect).p3[offset])
+    ((rect).p0[offset] - (rect).p1[offset] - (rect).p2[offset] + (rect).p3[offset])
 
 
 static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_cascade,
-                                      /*   const CvArr* _sum,
-                                      const CvArr* _sqsum,
-                                      const CvArr* _tilted_sum,*/
                                       double scale,
                                       int step)
 {
-    //   CvMat sum_stub, *sum = (CvMat*)_sum;
-    //   CvMat sqsum_stub, *sqsum = (CvMat*)_sqsum;
-    //   CvMat tilted_stub, *tilted = (CvMat*)_tilted_sum;
     GpuHidHaarClassifierCascade *cascade;
     int coi0 = 0, coi1 = 0;
     int i;
@@ -537,61 +442,25 @@ static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_casc
     if( scale <= 0 )
         CV_Error( CV_StsOutOfRange, "Scale must be positive" );
 
-    //   sum = cvGetMat( sum, &sum_stub, &coi0 );
-    //   sqsum = cvGetMat( sqsum, &sqsum_stub, &coi1 );
-
     if( coi0 || coi1 )
         CV_Error( CV_BadCOI, "COI is not supported" );
 
-    //   if( !CV_ARE_SIZES_EQ( sum, sqsum ))
-    //       CV_Error( CV_StsUnmatchedSizes, "All integral images must have the same size" );
-
-    //   if( CV_MAT_TYPE(sqsum->type) != CV_64FC1 ||
-    //       CV_MAT_TYPE(sum->type) != CV_32SC1 )
-    //       CV_Error( CV_StsUnsupportedFormat,
-    //       "Only (32s, 64f, 32s) combination of (sum,sqsum,tilted_sum) formats is allowed" );
-
     if( !_cascade->hid_cascade )
         gpuCreateHidHaarClassifierCascade(_cascade, &datasize, &total);
 
     cascade = (GpuHidHaarClassifierCascade *) _cascade->hid_cascade;
     stage_classifier = (GpuHidHaarStageClassifier *) (cascade + 1);
 
-    if( cascade->has_tilted_features )
-    {
-        //    tilted = cvGetMat( tilted, &tilted_stub, &coi1 );
-
-        //    if( CV_MAT_TYPE(tilted->type) != CV_32SC1 )
-        //        CV_Error( CV_StsUnsupportedFormat,
-        //        "Only (32s, 64f, 32s) combination of (sum,sqsum,tilted_sum) formats is allowed" );
-
-        //    if( sum->step != tilted->step )
-        //        CV_Error( CV_StsUnmatchedSizes,
-        //        "Sum and tilted_sum must have the same stride (step, widthStep)" );
-
-        //    if( !CV_ARE_SIZES_EQ( sum, tilted ))
-        //        CV_Error( CV_StsUnmatchedSizes, "All integral images must have the same size" );
-        //  cascade->tilted = *tilted;
-    }
-
     _cascade->scale = scale;
     _cascade->real_window_size.width = cvRound( _cascade->orig_window_size.width * scale );
     _cascade->real_window_size.height = cvRound( _cascade->orig_window_size.height * scale );
 
-    //cascade->sum = *sum;
-    //cascade->sqsum = *sqsum;
-
     equRect.x = equRect.y = cvRound(scale);
     equRect.width = cvRound((_cascade->orig_window_size.width - 2) * scale);
     equRect.height = cvRound((_cascade->orig_window_size.height - 2) * scale);
     weight_scale = 1. / (equRect.width * equRect.height);
     cascade->inv_window_area = weight_scale;
 
-    //	cascade->pq0 = equRect.y * step + equRect.x;
-    //	cascade->pq1 = equRect.y * step + equRect.x + equRect.width ;
-    //	cascade->pq2 = (equRect.y + equRect.height)*step + equRect.x;
-    //	cascade->pq3 = (equRect.y + equRect.height)*step + equRect.x + equRect.width ;
-
     cascade->pq0 = equRect.x;
     cascade->pq1 = equRect.y;
     cascade->pq2 = equRect.x + equRect.width;
@@ -614,10 +483,6 @@ static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_casc
             {
                 CvHaarFeature *feature =
                     &_cascade->stage_classifier[i].classifier[j].haar_feature[l];
-                /*  GpuHidHaarClassifier* classifier =
-                cascade->stage_classifier[i].classifier + j; */
-                //GpuHidHaarFeature* hidfeature =
-                //    &cascade->stage_classifier[i].classifier[j].node[l].feature;
                 GpuHidHaarTreeNode *hidnode = &stage_classifier[i].classifier[j].node[l];
                 double sum0 = 0, area0 = 0;
                 CvRect r[3];
@@ -632,8 +497,6 @@ static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_casc
                 /* align blocks */
                 for( k = 0; k < CV_HAAR_FEATURE_MAX; k++ )
                 {
-                    //if( !hidfeature->rect[k].p0 )
-                    //    break;
                     if(!hidnode->p[k][0])
                         break;
                     r[k] = feature->rect[k].r;
@@ -713,15 +576,6 @@ static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_casc
 
                     if( !feature->tilted )
                     {
-                        /*     hidfeature->rect[k].p0 = tr.y * sum->cols + tr.x;
-                        hidfeature->rect[k].p1 = tr.y * sum->cols + tr.x + tr.width;
-                        hidfeature->rect[k].p2 = (tr.y + tr.height) * sum->cols + tr.x;
-                        hidfeature->rect[k].p3 = (tr.y + tr.height) * sum->cols + tr.x + tr.width;
-                        */
-                        /*hidnode->p0[k] = tr.y * step + tr.x;
-                        hidnode->p1[k] = tr.y * step + tr.x + tr.width;
-                        hidnode->p2[k] = (tr.y + tr.height) * step + tr.x;
-                        hidnode->p3[k] = (tr.y + tr.height) * step + tr.x + tr.width;*/
                         hidnode->p[k][0] = tr.x;
                         hidnode->p[k][1] = tr.y;
                         hidnode->p[k][2] = tr.x + tr.width;
@@ -729,37 +583,24 @@ static void gpuSetImagesForHaarClassifierCascade( CvHaarClassifierCascade *_casc
                     }
                     else
                     {
-                        /*    hidfeature->rect[k].p2 = (tr.y + tr.width) * tilted->cols + tr.x + tr.width;
-                        hidfeature->rect[k].p3 = (tr.y + tr.width + tr.height) * tilted->cols + tr.x + tr.width - tr.height;
-                        hidfeature->rect[k].p0 = tr.y * tilted->cols + tr.x;
-                        hidfeature->rect[k].p1 = (tr.y + tr.height) * tilted->cols + tr.x - tr.height;
-                        */
-
                         hidnode->p[k][2] = (tr.y + tr.width) * step + tr.x + tr.width;
                         hidnode->p[k][3] = (tr.y + tr.width + tr.height) * step + tr.x + tr.width - tr.height;
                         hidnode->p[k][0] = tr.y * step + tr.x;
                         hidnode->p[k][1] = (tr.y + tr.height) * step + tr.x - tr.height;
                     }
-
-                    //hidfeature->rect[k].weight = (float)(feature->rect[k].weight * correction_ratio);
                     hidnode->weight[k] = (float)(feature->rect[k].weight * correction_ratio);
                     if( k == 0 )
                         area0 = tr.width * tr.height;
                     else
-                        //sum0 += hidfeature->rect[k].weight * tr.width * tr.height;
                         sum0 += hidnode->weight[k] * tr.width * tr.height;
                 }
-
-                // hidfeature->rect[0].weight = (float)(-sum0/area0);
                 hidnode->weight[0] = (float)(-sum0 / area0);
             } /* l */
         } /* j */
     }
 }
 
-static void gpuSetHaarClassifierCascade( CvHaarClassifierCascade *_cascade
-                             /*double scale=0.0,*/
-                             /*int step*/)
+static void gpuSetHaarClassifierCascade( CvHaarClassifierCascade *_cascade)
 {
     GpuHidHaarClassifierCascade *cascade;
     int i;
@@ -813,11 +654,7 @@ static void gpuSetHaarClassifierCascade( CvHaarClassifierCascade *_cascade
                     if(!hidnode->p[k][0])
                         break;
                     r[k] = feature->rect[k].r;
-                    // 					base_w = (int)CV_IMIN( (unsigned)base_w, (unsigned)(r[k].width-1) );
-                    // 					base_w = (int)CV_IMIN( (unsigned)base_w, (unsigned)(r[k].x - r[0].x-1) );
-                    // 					base_h = (int)CV_IMIN( (unsigned)base_h, (unsigned)(r[k].height-1) );
-                    // 					base_h = (int)CV_IMIN( (unsigned)base_h, (unsigned)(r[k].y - r[0].y-1) );
-                }
+               }
 
                 nr = k;
                 for( k = 0; k < nr; k++ )
@@ -835,22 +672,19 @@ static void gpuSetHaarClassifierCascade( CvHaarClassifierCascade *_cascade
                     hidnode->p[k][3] = tr.height;
                     hidnode->weight[k] = (float)(feature->rect[k].weight * correction_ratio);
                 }
-                //hidnode->weight[0]=(float)(-sum0/area0);
             } /* l */
         } /* j */
     }
 }
-CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemStorage *storage, double scaleFactor,
-        int minNeighbors, int flags, CvSize minSize, CvSize maxSize)
+void OclCascadeClassifier::detectMultiScale(oclMat &gimg, CV_OUT std::vector<cv::Rect>& faces,
+                                            double scaleFactor, int minNeighbors, int flags,
+                                            Size minSize, Size maxSize)
+//CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemStorage *storage, double scaleFactor,
+//        int minNeighbors, int flags, CvSize minSize, CvSize maxSize)
 {
     CvHaarClassifierCascade *cascade = oldCascade;
 
-    //double alltime = (double)cvGetTickCount();
-    //double t = (double)cvGetTickCount();
     const double GROUP_EPS = 0.2;
-    oclMat gtemp, gsum1, gtilted1, gsqsum1, gnormImg, gsumcanny;
-    CvSeq *result_seq = 0;
-    cv::Ptr<CvMemStorage> temp_storage;
 
     cv::ConcurrentRectVector allCandidates;
     std::vector<cv::Rect> rectList;
@@ -859,7 +693,6 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
     int datasize=0;
     int totalclassifier=0;
 
-    //void *out;
     GpuHidHaarClassifierCascade *gcascade;
     GpuHidHaarStageClassifier    *stage;
     GpuHidHaarClassifier         *classifier;
@@ -868,11 +701,8 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
     int *candidate;
     cl_int status;
 
-    //    bool doCannyPruning = (flags & CV_HAAR_DO_CANNY_PRUNING) != 0;
     bool findBiggestObject = (flags & CV_HAAR_FIND_BIGGEST_OBJECT) != 0;
-    //    bool roughSearch = (flags & CV_HAAR_DO_ROUGH_SEARCH) != 0;
 
-    //double t = 0;
     if( maxSize.height == 0 || maxSize.width == 0 )
     {
         maxSize.height = gimg.rows;
@@ -882,8 +712,8 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
     if( !CV_IS_HAAR_CLASSIFIER(cascade) )
         CV_Error( !cascade ? CV_StsNullPtr : CV_StsBadArg, "Invalid classifier cascade" );
 
-    if( !storage )
-        CV_Error( CV_StsNullPtr, "Null storage pointer" );
+    //if( !storage )
+    //    CV_Error( CV_StsNullPtr, "Null storage pointer" );
 
     if( CV_MAT_DEPTH(gimg.type()) != CV_8U )
         CV_Error( CV_StsUnsupportedFormat, "Only 8-bit images are supported" );
@@ -894,48 +724,39 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
     if( findBiggestObject )
         flags &= ~CV_HAAR_SCALE_IMAGE;
 
-    //gtemp = oclMat( gimg.rows, gimg.cols, CV_8UC1);
-    //gsum1 =  oclMat( gimg.rows + 1, gimg.cols + 1, CV_32SC1 );
-    //gsqsum1 = oclMat( gimg.rows + 1, gimg.cols + 1, CV_32FC1 );
-
     if( !cascade->hid_cascade )
-        /*out = (void *)*/gpuCreateHidHaarClassifierCascade(cascade, &datasize, &totalclassifier);
-    if( cascade->hid_cascade->has_tilted_features )
-        gtilted1 = oclMat( gimg.rows + 1, gimg.cols + 1, CV_32SC1 );
+        gpuCreateHidHaarClassifierCascade(cascade, &datasize, &totalclassifier);
 
-    result_seq = cvCreateSeq( 0, sizeof(CvSeq), sizeof(CvAvgComp), storage );
+    //result_seq = cvCreateSeq( 0, sizeof(CvSeq), sizeof(CvAvgComp), storage );
 
     if( CV_MAT_CN(gimg.type()) > 1 )
     {
-        cvtColor( gimg, gtemp, CV_BGR2GRAY );
+        oclMat gtemp;
+        cvtColor( gimg, gtemp, COLOR_BGR2GRAY );
         gimg = gtemp;
     }
 
     if( findBiggestObject )
         flags &= ~(CV_HAAR_SCALE_IMAGE | CV_HAAR_DO_CANNY_PRUNING);
-    //t = (double)cvGetTickCount() - t;
-    //printf( "before if time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
 
     if( gimg.cols < minSize.width || gimg.rows < minSize.height )
         CV_Error(CV_StsError, "Image too small");
 
+    cl_command_queue qu = reinterpret_cast<cl_command_queue>(Context::getContext()->oclCommandQueue());
     if( (flags & CV_HAAR_SCALE_IMAGE) )
     {
         CvSize winSize0 = cascade->orig_window_size;
-        //float scalefactor = 1.1f;
-        //float factor = 1.f;
         int totalheight = 0;
         int indexy = 0;
         CvSize sz;
-        //t = (double)cvGetTickCount();
         std::vector<CvSize> sizev;
         std::vector<float> scalev;
         for(factor = 1.f;; factor *= scaleFactor)
         {
-            CvSize winSize = { cvRound(winSize0.width * factor), cvRound(winSize0.height * factor) };
+            CvSize winSize( cvRound(winSize0.width * factor), cvRound(winSize0.height * factor) );
             sz.width     = cvRound( gimg.cols / factor ) + 1;
             sz.height    = cvRound( gimg.rows / factor ) + 1;
-            CvSize sz1     = { sz.width - winSize0.width - 1,      sz.height - winSize0.height - 1 };
+            CvSize sz1( sz.width - winSize0.width - 1,      sz.height - winSize0.height - 1 );
 
             if( sz1.width <= 0 || sz1.height <= 0 )
                 break;
@@ -948,20 +769,15 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
             sizev.push_back(sz);
             scalev.push_back(factor);
         }
-        //int flag = 0;
 
         oclMat gimg1(gimg.rows, gimg.cols, CV_8UC1);
-        oclMat gsum(totalheight, gimg.cols + 1, CV_32SC1);
-        oclMat gsqsum(totalheight, gimg.cols + 1, CV_32FC1);
+        oclMat gsum(totalheight + 4, gimg.cols + 1, CV_32SC1);
+        oclMat gsqsum(totalheight + 4, gimg.cols + 1, CV_32FC1);
 
-        //cl_mem cascadebuffer;
         cl_mem stagebuffer;
-        //cl_mem classifierbuffer;
         cl_mem nodebuffer;
         cl_mem candidatebuffer;
         cl_mem scaleinfobuffer;
-        //cl_kernel kernel;
-        //kernel = openCLGetKernelFromSource(gimg.clCxt, &haarobjectdetect, "gpuRunHaarClassifierCascade");
         cv::Rect roi, roi2;
         cv::Mat imgroi, imgroisq;
         cv::ocl::oclMat resizeroi, gimgroi, gimgroisq;
@@ -969,18 +785,13 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
 
         size_t blocksize = 8;
         size_t localThreads[3] = { blocksize, blocksize , 1 };
-        size_t globalThreads[3] = { grp_per_CU *((gsum.clCxt)->impl->maxComputeUnits) *localThreads[0],
+        size_t globalThreads[3] = { grp_per_CU *(gsum.clCxt->computeUnits()) *localThreads[0],
                                     localThreads[1], 1
                                   };
         int outputsz = 256 * globalThreads[0] / localThreads[0];
         int loopcount = sizev.size();
         detect_piramid_info *scaleinfo = (detect_piramid_info *)malloc(sizeof(detect_piramid_info) * loopcount);
 
-        //t = (double)cvGetTickCount() - t;
-        // printf( "pre time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-        //int *it =scaleinfo;
-        // t = (double)cvGetTickCount();
-
         for( int i = 0; i < loopcount; i++ )
         {
             sz = sizev[i];
@@ -990,7 +801,6 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
             resizeroi = gimg1(roi2);
             gimgroi = gsum(roi);
             gimgroisq = gsqsum(roi);
-            //scaleinfo[i].rows = gimgroi.rows;
             int width = gimgroi.cols - 1 - cascade->orig_window_size.width;
             int height = gimgroi.rows - 1 - cascade->orig_window_size.height;
             scaleinfo[i].width_height = (width << 16) | height;
@@ -998,76 +808,39 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
 
             int grpnumperline = (width + localThreads[0] - 1) / localThreads[0];
             int totalgrp = ((height + localThreads[1] - 1) / localThreads[1]) * grpnumperline;
-            //outputsz +=width*height;
 
             scaleinfo[i].grpnumperline_totalgrp = (grpnumperline << 16) | totalgrp;
             scaleinfo[i].imgoff = gimgroi.offset >> 2;
             scaleinfo[i].factor = factor;
-            //printf("rows = %d,ystep = %d,width = %d,height = %d,grpnumperline = %d,totalgrp = %d,imgoff = %d,factor = %f\n",
-            //	scaleinfo[i].rows,scaleinfo[i].ystep,scaleinfo[i].width,scaleinfo[i].height,scaleinfo[i].grpnumperline,
-            //	scaleinfo[i].totalgrp,scaleinfo[i].imgoff,scaleinfo[i].factor);
             cv::ocl::resize(gimg, resizeroi, Size(sz.width - 1, sz.height - 1), 0, 0, INTER_LINEAR);
-            //cv::imwrite("D:\\1.jpg",gimg1);
             cv::ocl::integral(resizeroi, gimgroi, gimgroisq);
-            //cv::ocl::oclMat chk(sz.height,sz.width,CV_32SC1),chksq(sz.height,sz.width,CV_32FC1);
-            //cv::ocl::integral(gimg1, chk, chksq);
-            //double r = cv::norm(chk,gimgroi,NORM_INF);
-            //if(r > std::numeric_limits<double>::epsilon())
-            //{
-            //	printf("failed");
-            //}
             indexy += sz.height;
         }
-        //int ystep = factor > 2 ? 1 : 2;
-        // t = (double)cvGetTickCount() - t;
-        //printf( "resize integral time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-        //t = (double)cvGetTickCount();
+
         gcascade   = (GpuHidHaarClassifierCascade *)cascade->hid_cascade;
         stage      = (GpuHidHaarStageClassifier *)(gcascade + 1);
         classifier = (GpuHidHaarClassifier *)(stage + gcascade->count);
         node       = (GpuHidHaarTreeNode *)(classifier->node);
 
-        //int m,n;
-        //m = (gsum.cols - 1 - cascade->orig_window_size.width  + ystep - 1)/ystep;
-        //n = (gsum.rows - 1 - cascade->orig_window_size.height + ystep - 1)/ystep;
-        //int counter = m*n;
-
         int nodenum = (datasize - sizeof(GpuHidHaarClassifierCascade) -
                        sizeof(GpuHidHaarStageClassifier) * gcascade->count - sizeof(GpuHidHaarClassifier) * totalclassifier) / sizeof(GpuHidHaarTreeNode);
-        //if(flag == 0){
-        candidate = (int *)malloc(4 * sizeof(int) * outputsz);
-        //memset((char*)candidate,0,4*sizeof(int)*outputsz);
-        gpuSetImagesForHaarClassifierCascade( cascade,/* &sum1, &sqsum1, _tilted,*/ 1., gsum.step / 4 );
 
-        //cascadebuffer = clCreateBuffer(gsum.clCxt->clContext,CL_MEM_READ_ONLY,sizeof(GpuHidHaarClassifierCascade),NULL,&status);
-        //openCLVerifyCall(status);
-        //openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->clCmdQueue,cascadebuffer,1,0,sizeof(GpuHidHaarClassifierCascade),gcascade,0,NULL,NULL));
+        candidate = (int *)malloc(4 * sizeof(int) * outputsz);
+
+        gpuSetImagesForHaarClassifierCascade( cascade, 1., gsum.step / 4 );
 
         stagebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(GpuHidHaarStageClassifier) * gcascade->count);
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, stagebuffer, 1, 0, sizeof(GpuHidHaarStageClassifier)*gcascade->count, stage, 0, NULL, NULL));
-
-        //classifierbuffer = clCreateBuffer(gsum.clCxt->clContext,CL_MEM_READ_ONLY,sizeof(GpuHidHaarClassifier)*totalclassifier,NULL,&status);
-        //status = clEnqueueWriteBuffer(gsum.clCxt->clCmdQueue,classifierbuffer,1,0,sizeof(GpuHidHaarClassifier)*totalclassifier,classifier,0,NULL,NULL);
+        openCLSafeCall(clEnqueueWriteBuffer(qu, stagebuffer, 1, 0, sizeof(GpuHidHaarStageClassifier)*gcascade->count, stage, 0, NULL, NULL));
 
         nodebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, nodenum * sizeof(GpuHidHaarTreeNode));
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, nodebuffer, 1, 0,
-                                            nodenum * sizeof(GpuHidHaarTreeNode),
+
+        openCLSafeCall(clEnqueueWriteBuffer(qu, nodebuffer, 1, 0, nodenum * sizeof(GpuHidHaarTreeNode),
                                             node, 0, NULL, NULL));
         candidatebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_WRITE_ONLY, 4 * sizeof(int) * outputsz);
-        //openCLVerifyCall(status);
+
         scaleinfobuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(detect_piramid_info) * loopcount);
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, scaleinfobuffer, 1, 0, sizeof(detect_piramid_info)*loopcount, scaleinfo, 0, NULL, NULL));
-        //flag  = 1;
-        //}
+        openCLSafeCall(clEnqueueWriteBuffer(qu, scaleinfobuffer, 1, 0, sizeof(detect_piramid_info)*loopcount, scaleinfo, 0, NULL, NULL));
 
-        //t = (double)cvGetTickCount() - t;
-        //printf( "update time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-
-        //size_t globalThreads[3] = { counter+blocksize*blocksize-counter%(blocksize*blocksize),1,1};
-        //t = (double)cvGetTickCount();
         int startstage = 0;
         int endstage = gcascade->count;
         int startnode = 0;
@@ -1085,11 +858,6 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         pq.s[3] = gcascade->pq3;
         float correction = gcascade->inv_window_area;
 
-        //int grpnumperline = ((m + localThreads[0] - 1) / localThreads[0]);
-        //int totalgrp = ((n + localThreads[1] - 1) / localThreads[1])*grpnumperline;
-        //   openCLVerifyKernel(gsum.clCxt, kernel, &blocksize, globalThreads, localThreads);
-        //openCLSafeCall(clSetKernelArg(kernel,argcount++,sizeof(cl_mem),(void*)&cascadebuffer));
-
         std::vector<std::pair<size_t, const void *> > args;
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&stagebuffer ));
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&scaleinfobuffer ));
@@ -1108,29 +876,24 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         args.push_back ( std::make_pair(sizeof(cl_int4) , (void *)&pq ));
         args.push_back ( std::make_pair(sizeof(cl_float) , (void *)&correction ));
 
-        openCLExecuteKernel(gsum.clCxt, &haarobjectdetect, "gpuRunHaarClassifierCascade", globalThreads, localThreads, args, -1, -1);
-        //t = (double)cvGetTickCount() - t;
-        //printf( "detection time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-        //t = (double)cvGetTickCount();
-        //openCLSafeCall(clEnqueueReadBuffer(gsum.clCxt->impl->clCmdQueue, candidatebuffer, 1, 0, 4 * sizeof(int)*outputsz, candidate, 0, NULL, NULL));
+        const char * build_options = gcascade->is_stump_based ? "-D STUMP_BASED=1" : "-D STUMP_BASED=0";
+
+        openCLExecuteKernel(gsum.clCxt, &haarobjectdetect, "gpuRunHaarClassifierCascade", globalThreads, localThreads, args, -1, -1, build_options);
+
         openCLReadBuffer( gsum.clCxt, candidatebuffer, candidate, 4 * sizeof(int)*outputsz );
 
         for(int i = 0; i < outputsz; i++)
             if(candidate[4 * i + 2] != 0)
-                allCandidates.push_back(Rect(candidate[4 * i], candidate[4 * i + 1], candidate[4 * i + 2], candidate[4 * i + 3]));
-        // t = (double)cvGetTickCount() - t;
-        //printf( "post time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-        //t = (double)cvGetTickCount();
+                allCandidates.push_back(Rect(candidate[4 * i], candidate[4 * i + 1],
+                candidate[4 * i + 2], candidate[4 * i + 3]));
+
         free(scaleinfo);
         free(candidate);
-        //openCLSafeCall(clReleaseMemObject(cascadebuffer));
         openCLSafeCall(clReleaseMemObject(stagebuffer));
         openCLSafeCall(clReleaseMemObject(scaleinfobuffer));
         openCLSafeCall(clReleaseMemObject(nodebuffer));
         openCLSafeCall(clReleaseMemObject(candidatebuffer));
-        // openCLSafeCall(clReleaseKernel(kernel));
-        //t = (double)cvGetTickCount() - t;
-        //printf( "release time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
+
     }
     else
     {
@@ -1148,7 +911,6 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         classifier = (GpuHidHaarClassifier *)(stage + gcascade->count);
         node       = (GpuHidHaarTreeNode *)(classifier->node);
         cl_mem stagebuffer;
-        //cl_mem classifierbuffer;
         cl_mem nodebuffer;
         cl_mem candidatebuffer;
         cl_mem scaleinfobuffer;
@@ -1159,9 +921,7 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
                 cvRound(factor * winsize0.height) < gimg.rows - 10;
                 n_factors++, factor *= scaleFactor )
         {
-            CvSize winSize = { cvRound( winsize0.width * factor ),
-                               cvRound( winsize0.height * factor )
-                             };
+            CvSize winSize( cvRound( winsize0.width * factor ), cvRound( winsize0.height * factor ) );
             if( winSize.width < minSize.width || winSize.height < minSize.height )
             {
                 continue;
@@ -1184,25 +944,20 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         int grp_per_CU = 12;
         size_t blocksize = 8;
         size_t localThreads[3] = { blocksize, blocksize , 1 };
-        size_t globalThreads[3] = { grp_per_CU *gsum.clCxt->impl->maxComputeUnits *localThreads[0],
-                                    localThreads[1], 1
-                                  };
+        size_t globalThreads[3] = { grp_per_CU *gsum.clCxt->computeUnits() *localThreads[0],
+                                    localThreads[1], 1 };
         int outputsz = 256 * globalThreads[0] / localThreads[0];
         int nodenum = (datasize - sizeof(GpuHidHaarClassifierCascade) -
                        sizeof(GpuHidHaarStageClassifier) * gcascade->count - sizeof(GpuHidHaarClassifier) * totalclassifier) / sizeof(GpuHidHaarTreeNode);
         nodebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY,
                                         nodenum * sizeof(GpuHidHaarTreeNode));
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, nodebuffer, 1, 0,
+        openCLSafeCall(clEnqueueWriteBuffer(qu, nodebuffer, 1, 0,
                                             nodenum * sizeof(GpuHidHaarTreeNode),
                                             node, 0, NULL, NULL));
         cl_mem newnodebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_WRITE,
                                loopcount * nodenum * sizeof(GpuHidHaarTreeNode));
         int startstage = 0;
         int endstage = gcascade->count;
-        //cl_kernel kernel;
-        //kernel = openCLGetKernelFromSource(gsum.clCxt, &haarobjectdetect_scaled2, "gpuRunHaarClassifierCascade_scaled2");
-        //cl_kernel kernel2 = openCLGetKernelFromSource(gimg.clCxt, &haarobjectdetect_scaled2, "gpuscaleclassifier");
         for(int i = 0; i < loopcount; i++)
         {
             sz = sizev[i];
@@ -1221,7 +976,7 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
             int height = (gsum.rows - 1 - sz.height + ystep - 1) / ystep;
             int grpnumperline = (width + localThreads[0] - 1) / localThreads[0];
             int totalgrp = ((height + localThreads[1] - 1) / localThreads[1]) * grpnumperline;
-            //outputsz +=width*height;
+
             scaleinfo[i].width_height = (width << 16) | height;
             scaleinfo[i].grpnumperline_totalgrp = (grpnumperline << 16) | totalgrp;
             scaleinfo[i].imgoff = 0;
@@ -1237,30 +992,21 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
             args1.push_back ( std::make_pair(sizeof(cl_int) , (void *)&startnodenum ));
 
             size_t globalThreads2[3] = {nodenum, 1, 1};
-
             openCLExecuteKernel(gsum.clCxt, &haarobjectdetect_scaled2, "gpuscaleclassifier", globalThreads2, NULL/*localThreads2*/, args1, -1, -1);
-
-            //clEnqueueNDRangeKernel(gsum.clCxt->impl->clCmdQueue, kernel2, 1, NULL, globalThreads2, 0, 0, NULL, NULL);
-            //clFinish(gsum.clCxt->impl->clCmdQueue);
         }
-        //clReleaseKernel(kernel2);
+
         int step = gsum.step / 4;
         int startnode = 0;
         int splitstage = 3;
-        int splitnode = stage[0].count + stage[1].count + stage[2].count;
         stagebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(GpuHidHaarStageClassifier) * gcascade->count);
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, stagebuffer, 1, 0, sizeof(GpuHidHaarStageClassifier)*gcascade->count, stage, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueWriteBuffer(qu, stagebuffer, 1, 0, sizeof(GpuHidHaarStageClassifier)*gcascade->count, stage, 0, NULL, NULL));
         candidatebuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_WRITE_ONLY | CL_MEM_ALLOC_HOST_PTR, 4 * sizeof(int) * outputsz);
-        //openCLVerifyCall(status);
         scaleinfobuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(detect_piramid_info) * loopcount);
-        //openCLVerifyCall(status);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, scaleinfobuffer, 1, 0, sizeof(detect_piramid_info)*loopcount, scaleinfo, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueWriteBuffer(qu, scaleinfobuffer, 1, 0, sizeof(detect_piramid_info)*loopcount, scaleinfo, 0, NULL, NULL));
         pbuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(cl_int4) * loopcount);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, pbuffer, 1, 0, sizeof(cl_int4)*loopcount, p, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueWriteBuffer(qu, pbuffer, 1, 0, sizeof(cl_int4)*loopcount, p, 0, NULL, NULL));
         correctionbuffer = openCLCreateBuffer(gsum.clCxt, CL_MEM_READ_ONLY, sizeof(cl_float) * loopcount);
-        openCLSafeCall(clEnqueueWriteBuffer(gsum.clCxt->impl->clCmdQueue, correctionbuffer, 1, 0, sizeof(cl_float)*loopcount, correction, 0, NULL, NULL));
-        //int argcount = 0;
+        openCLSafeCall(clEnqueueWriteBuffer(qu, correctionbuffer, 1, 0, sizeof(cl_float)*loopcount, correction, 0, NULL, NULL));
 
         std::vector<std::pair<size_t, const void *> > args;
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&stagebuffer ));
@@ -1269,22 +1015,21 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&gsum.data ));
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&gsqsum.data ));
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&candidatebuffer ));
+        args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&gsum.rows ));
+        args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&gsum.cols ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&step ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&loopcount ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&startstage ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&splitstage ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&endstage ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&startnode ));
-        args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&splitnode ));
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&pbuffer ));
         args.push_back ( std::make_pair(sizeof(cl_mem) , (void *)&correctionbuffer ));
         args.push_back ( std::make_pair(sizeof(cl_int) , (void *)&nodenum ));
+        const char * build_options = gcascade->is_stump_based ? "-D STUMP_BASED=1" : "-D STUMP_BASED=0";
+        openCLExecuteKernel(gsum.clCxt, &haarobjectdetect_scaled2, "gpuRunHaarClassifierCascade_scaled2", globalThreads, localThreads, args, -1, -1, build_options);
 
-
-        openCLExecuteKernel(gsum.clCxt, &haarobjectdetect_scaled2, "gpuRunHaarClassifierCascade_scaled2", globalThreads, localThreads, args, -1, -1);
-
-        //openCLSafeCall(clEnqueueReadBuffer(gsum.clCxt->clCmdQueue,candidatebuffer,1,0,4*sizeof(int)*outputsz,candidate,0,NULL,NULL));
-        candidate = (int *)clEnqueueMapBuffer(gsum.clCxt->impl->clCmdQueue, candidatebuffer, 1, CL_MAP_READ, 0, 4 * sizeof(int), 0, 0, 0, &status);
+        candidate = (int *)clEnqueueMapBuffer(qu, candidatebuffer, 1, CL_MAP_READ, 0, 4 * sizeof(int) * outputsz, 0, 0, 0, &status);
 
         for(int i = 0; i < outputsz; i++)
         {
@@ -1295,7 +1040,7 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         free(scaleinfo);
         free(p);
         free(correction);
-        clEnqueueUnmapMemObject(gsum.clCxt->impl->clCmdQueue, candidatebuffer, candidate, 0, 0, 0);
+        clEnqueueUnmapMemObject(qu, candidatebuffer, candidate, 0, 0, 0);
         openCLSafeCall(clReleaseMemObject(stagebuffer));
         openCLSafeCall(clReleaseMemObject(scaleinfobuffer));
         openCLSafeCall(clReleaseMemObject(nodebuffer));
@@ -1304,195 +1049,33 @@ CvSeq *cv::ocl::OclCascadeClassifier::oclHaarDetectObjects( oclMat &gimg, CvMemS
         openCLSafeCall(clReleaseMemObject(pbuffer));
         openCLSafeCall(clReleaseMemObject(correctionbuffer));
     }
-    //t = (double)cvGetTickCount() ;
+
     cvFree(&cascade->hid_cascade);
-    //    printf("%d\n",globalcounter);
     rectList.resize(allCandidates.size());
     if(!allCandidates.empty())
         std::copy(allCandidates.begin(), allCandidates.end(), rectList.begin());
 
-    //cout << "count = " << rectList.size()<< endl;
-
     if( minNeighbors != 0 || findBiggestObject )
         groupRectangles(rectList, rweights, std::max(minNeighbors, 1), GROUP_EPS);
     else
         rweights.resize(rectList.size(), 0);
 
-
+    faces.clear();
     if( findBiggestObject && rectList.size() )
     {
-        CvAvgComp result_comp = {{0, 0, 0, 0}, 0};
-
+        Rect result_comp(0, 0, 0, 0);
         for( size_t i = 0; i < rectList.size(); i++ )
         {
             cv::Rect r = rectList[i];
-            if( r.area() > cv::Rect(result_comp.rect).area() )
+            if( r.area() > result_comp.area() )
             {
-                result_comp.rect = r;
-                result_comp.neighbors = rweights[i];
+                result_comp = r;
             }
         }
-        cvSeqPush( result_seq, &result_comp );
+        faces.push_back(result_comp);
     }
     else
     {
-        for( size_t i = 0; i < rectList.size(); i++ )
-        {
-            CvAvgComp c;
-            c.rect = rectList[i];
-            c.neighbors = rweights[i];
-            cvSeqPush( result_seq, &c );
-        }
+        faces = rectList;
     }
-    //t = (double)cvGetTickCount() - t;
-    //printf( "get face time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
-    //alltime = (double)cvGetTickCount() - alltime;
-    //printf( "all time = %g ms\n", alltime/((double)cvGetTickFrequency()*1000.) );
-    return result_seq;
-}
-
-
-#ifndef _MAX_PATH
-#define _MAX_PATH 1024
-#endif
-
-
-/****************************************************************************************\
-*                                  Persistence functions                                 *
-\****************************************************************************************/
-
-/* field names */
-
-#define ICV_HAAR_SIZE_NAME            "size"
-#define ICV_HAAR_STAGES_NAME          "stages"
-#define ICV_HAAR_TREES_NAME             "trees"
-#define ICV_HAAR_FEATURE_NAME             "feature"
-#define ICV_HAAR_RECTS_NAME                 "rects"
-#define ICV_HAAR_TILTED_NAME                "tilted"
-#define ICV_HAAR_THRESHOLD_NAME           "threshold"
-#define ICV_HAAR_LEFT_NODE_NAME           "left_node"
-#define ICV_HAAR_LEFT_VAL_NAME            "left_val"
-#define ICV_HAAR_RIGHT_NODE_NAME          "right_node"
-#define ICV_HAAR_RIGHT_VAL_NAME           "right_val"
-#define ICV_HAAR_STAGE_THRESHOLD_NAME   "stage_threshold"
-#define ICV_HAAR_PARENT_NAME            "parent"
-#define ICV_HAAR_NEXT_NAME              "next"
-
-static int gpuRunHaarClassifierCascade( /*const CvHaarClassifierCascade *_cascade, CvPoint pt, int start_stage */)
-{
-    return 1;
-}
-
-namespace cv
-{
-namespace ocl
-{
-
-struct gpuHaarDetectObjects_ScaleImage_Invoker
-{
-    gpuHaarDetectObjects_ScaleImage_Invoker( const CvHaarClassifierCascade *_cascade,
-            int _stripSize, double _factor,
-            const Mat &_sum1, const Mat &_sqsum1, Mat *_norm1,
-            Mat *_mask1, Rect _equRect, ConcurrentRectVector &_vec )
-    {
-        cascade = _cascade;
-        stripSize = _stripSize;
-        factor = _factor;
-        sum1 = _sum1;
-        sqsum1 = _sqsum1;
-        norm1 = _norm1;
-        mask1 = _mask1;
-        equRect = _equRect;
-        vec = &_vec;
-    }
-
-    void operator()( const BlockedRange &range ) const
-    {
-        Size winSize0 = cascade->orig_window_size;
-        Size winSize(cvRound(winSize0.width * factor), cvRound(winSize0.height * factor));
-        int y1 = range.begin() * stripSize, y2 = std::min(range.end() * stripSize, sum1.rows - 1 - winSize0.height);
-        Size ssz(sum1.cols - 1 - winSize0.width, y2 - y1);
-        int x, y, ystep = factor > 2 ? 1 : 2;
-
-        for( y = y1; y < y2; y += ystep )
-            for( x = 0; x < ssz.width; x += ystep )
-            {
-                if( gpuRunHaarClassifierCascade( /*cascade, cvPoint(x, y), 0*/ ) > 0 )
-                    vec->push_back(Rect(cvRound(x * factor), cvRound(y * factor),
-                                        winSize.width, winSize.height));
-            }
-    }
-
-    const CvHaarClassifierCascade *cascade;
-    int stripSize;
-    double factor;
-    Mat sum1, sqsum1, *norm1, *mask1;
-    Rect equRect;
-    ConcurrentRectVector *vec;
-};
-
-
-struct gpuHaarDetectObjects_ScaleCascade_Invoker
-{
-    gpuHaarDetectObjects_ScaleCascade_Invoker( const CvHaarClassifierCascade *_cascade,
-            Size _winsize, const Range &_xrange, double _ystep,
-            size_t _sumstep, const int **_p, const int **_pq,
-            ConcurrentRectVector &_vec )
-    {
-        cascade = _cascade;
-        winsize = _winsize;
-        xrange = _xrange;
-        ystep = _ystep;
-        sumstep = _sumstep;
-        p = _p;
-        pq = _pq;
-        vec = &_vec;
-    }
-
-    void operator()( const BlockedRange &range ) const
-    {
-        int iy, startY = range.begin(), endY = range.end();
-        const int *p0 = p[0], *p1 = p[1], *p2 = p[2], *p3 = p[3];
-        const int *pq0 = pq[0], *pq1 = pq[1], *pq2 = pq[2], *pq3 = pq[3];
-        bool doCannyPruning = p0 != 0;
-        int sstep = (int)(sumstep / sizeof(p0[0]));
-
-        for( iy = startY; iy < endY; iy++ )
-        {
-            int ix, y = cvRound(iy * ystep), ixstep = 1;
-            for( ix = xrange.start; ix < xrange.end; ix += ixstep )
-            {
-                int x = cvRound(ix * ystep); // it should really be ystep, not ixstep
-
-                if( doCannyPruning )
-                {
-                    int offset = y * sstep + x;
-                    int s = p0[offset] - p1[offset] - p2[offset] + p3[offset];
-                    int sq = pq0[offset] - pq1[offset] - pq2[offset] + pq3[offset];
-                    if( s < 100 || sq < 20 )
-                    {
-                        ixstep = 2;
-                        continue;
-                    }
-                }
-
-                int result = gpuRunHaarClassifierCascade(/* cascade, cvPoint(x, y), 0 */);
-                if( result > 0 )
-                    vec->push_back(Rect(x, y, winsize.width, winsize.height));
-                ixstep = result != 0 ? 1 : 2;
-            }
-        }
-    }
-
-    const CvHaarClassifierCascade *cascade;
-    double ystep;
-    size_t sumstep;
-    Size winsize;
-    Range xrange;
-    const int **p;
-    const int **pq;
-    ConcurrentRectVector *vec;
-};
-
-}
 }
diff --git a/modules/ocl/src/hog.cpp b/modules/ocl/src/hog.cpp
index db92205837..ac113d27b5 100644
--- a/modules/ocl/src/hog.cpp
+++ b/modules/ocl/src/hog.cpp
@@ -15,7 +15,7 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//		Wenju He, wenju@multicorewareinc.com
+//     Wenju He, wenju@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -44,17 +44,19 @@
 //M*/
 
 #include "precomp.hpp"
-#include "mcwutil.hpp"
+
 using namespace cv;
 using namespace cv::ocl;
 
-
 #define CELL_WIDTH 8
 #define CELL_HEIGHT 8
 #define CELLS_PER_BLOCK_X 2
 #define CELLS_PER_BLOCK_Y 2
 #define NTHREADS 256
 
+static oclMat gauss_w_lut;
+static bool hog_device_cpu;
+
 namespace cv
 {
     namespace ocl
@@ -78,38 +80,43 @@ namespace cv
                 int cnblocks_win_x;
                 int cnblocks_win_y;
                 int cblock_hist_size;
-                int cblock_hist_size_2up;
                 int cdescr_size;
                 int cdescr_width;
+                int cdescr_height;
 
                 void set_up_constants(int nbins, int block_stride_x, int block_stride_y,
                                       int nblocks_win_x, int nblocks_win_y);
 
                 void compute_hists(int nbins, int block_stride_x, int blovck_stride_y,
                                    int height, int width, const cv::ocl::oclMat &grad,
-                                   const cv::ocl::oclMat &qangle, float sigma, cv::ocl::oclMat &block_hists);
+                                   const cv::ocl::oclMat &qangle,
+                                   const cv::ocl::oclMat &gauss_w_lut, cv::ocl::oclMat &block_hists);
 
                 void normalize_hists(int nbins, int block_stride_x, int block_stride_y,
-                                     int height, int width, cv::ocl::oclMat &block_hists, float threshold);
+                                     int height, int width, cv::ocl::oclMat &block_hists,
+                                     float threshold);
 
                 void classify_hists(int win_height, int win_width, int block_stride_y,
-                                    int block_stride_x, int win_stride_y, int win_stride_x, int height,
-                                    int width, const cv::ocl::oclMat &block_hists, const cv::ocl::oclMat &coefs, float free_coef,
+                                    int block_stride_x, int win_stride_y, int win_stride_x,
+                                    int height, int width, const cv::ocl::oclMat &block_hists,
+                                    const cv::ocl::oclMat &coefs, float free_coef,
                                     float threshold, cv::ocl::oclMat &labels);
 
-                void extract_descrs_by_rows(int win_height, int win_width, int block_stride_y, int block_stride_x,
-                                            int win_stride_y, int win_stride_x, int height, int width, const cv::ocl::oclMat &block_hists,
+                void extract_descrs_by_rows(int win_height, int win_width, int block_stride_y,
+                                            int block_stride_x, int win_stride_y, int win_stride_x,
+                                            int height, int width, const cv::ocl::oclMat &block_hists,
                                             cv::ocl::oclMat &descriptors);
-                void extract_descrs_by_cols(int win_height, int win_width, int block_stride_y, int block_stride_x,
-                                            int win_stride_y, int win_stride_x, int height, int width, const cv::ocl::oclMat &block_hists,
+                void extract_descrs_by_cols(int win_height, int win_width, int block_stride_y,
+                                            int block_stride_x, int win_stride_y, int win_stride_x,
+                                            int height, int width, const cv::ocl::oclMat &block_hists,
                                             cv::ocl::oclMat &descriptors);
 
                 void compute_gradients_8UC1(int height, int width, const cv::ocl::oclMat &img,
-                                            float angle_scale, cv::ocl::oclMat &grad, cv::ocl::oclMat &qangle, bool correct_gamma);
+                                            float angle_scale, cv::ocl::oclMat &grad,
+                                            cv::ocl::oclMat &qangle, bool correct_gamma);
                 void compute_gradients_8UC4(int height, int width, const cv::ocl::oclMat &img,
-                                            float angle_scale, cv::ocl::oclMat &grad, cv::ocl::oclMat &qangle, bool correct_gamma);
-
-                void resize( const oclMat &src, oclMat &dst, const Size sz);
+                                            float angle_scale, cv::ocl::oclMat &grad,
+                                            cv::ocl::oclMat &qangle, bool correct_gamma);
             }
         }
     }
@@ -117,8 +124,14 @@ namespace cv
 
 using namespace ::cv::ocl::device;
 
-cv::ocl::HOGDescriptor::HOGDescriptor(Size win_size_, Size block_size_, Size block_stride_, Size cell_size_,
-                                      int nbins_, double win_sigma_, double threshold_L2hys_, bool gamma_correction_, int nlevels_)
+static inline int divUp(int total, int grain)
+{
+    return (total + grain - 1) / grain;
+}
+
+cv::ocl::HOGDescriptor::HOGDescriptor(Size win_size_, Size block_size_, Size block_stride_,
+                                      Size cell_size_, int nbins_, double win_sigma_,
+                                      double threshold_L2hys_, bool gamma_correction_, int nlevels_)
     : win_size(win_size_),
       block_size(block_size_),
       block_stride(block_stride_),
@@ -132,19 +145,27 @@ cv::ocl::HOGDescriptor::HOGDescriptor(Size win_size_, Size block_size_, Size blo
     CV_Assert((win_size.width  - block_size.width ) % block_stride.width  == 0 &&
               (win_size.height - block_size.height) % block_stride.height == 0);
 
-    CV_Assert(block_size.width % cell_size.width == 0 && block_size.height % cell_size.height == 0);
+    CV_Assert(block_size.width % cell_size.width == 0 &&
+        block_size.height % cell_size.height == 0);
 
     CV_Assert(block_stride == cell_size);
 
     CV_Assert(cell_size == Size(8, 8));
 
-    Size cells_per_block = Size(block_size.width / cell_size.width, block_size.height / cell_size.height);
+    Size cells_per_block(block_size.width / cell_size.width,
+        block_size.height / cell_size.height);
     CV_Assert(cells_per_block == Size(2, 2));
 
     cv::Size blocks_per_win = numPartsWithin(win_size, block_size, block_stride);
-    hog::set_up_constants(nbins, block_stride.width, block_stride.height, blocks_per_win.width, blocks_per_win.height);
+    hog::set_up_constants(nbins, block_stride.width, block_stride.height,
+        blocks_per_win.width, blocks_per_win.height);
 
     effect_size = Size(0, 0);
+
+    if (queryDeviceInfo<IS_CPU_DEVICE, bool>())
+        hog_device_cpu = true;
+    else
+        hog_device_cpu = false;
 }
 
 size_t cv::ocl::HOGDescriptor::getDescriptorSize() const
@@ -154,7 +175,8 @@ size_t cv::ocl::HOGDescriptor::getDescriptorSize() const
 
 size_t cv::ocl::HOGDescriptor::getBlockHistogramSize() const
 {
-    Size cells_per_block = Size(block_size.width / cell_size.width, block_size.height / cell_size.height);
+    Size cells_per_block = Size(block_size.width / cell_size.width,
+        block_size.height / cell_size.height);
     return (size_t)(nbins * cells_per_block.area());
 }
 
@@ -167,7 +189,8 @@ bool cv::ocl::HOGDescriptor::checkDetectorSize() const
 {
     size_t detector_size = detector.rows * detector.cols;
     size_t descriptor_size = getDescriptorSize();
-    return detector_size == 0 || detector_size == descriptor_size || detector_size == descriptor_size + 1;
+    return detector_size == 0 || detector_size == descriptor_size ||
+        detector_size == descriptor_size + 1;
 }
 
 void cv::ocl::HOGDescriptor::setSVMDetector(const std::vector<float> &_detector)
@@ -207,10 +230,24 @@ void cv::ocl::HOGDescriptor::init_buffer(const oclMat &img, Size win_stride)
 
     const size_t block_hist_size = getBlockHistogramSize();
     const Size blocks_per_img = numPartsWithin(img.size(), block_size, block_stride);
-    block_hists.create(1, static_cast<int>(block_hist_size * blocks_per_img.area()), CV_32F);
+    block_hists.create(1,
+        static_cast<int>(block_hist_size * blocks_per_img.area()) + 256, CV_32F);
 
     Size wins_per_img = numPartsWithin(img.size(), win_size, win_stride);
     labels.create(1, wins_per_img.area(), CV_8U);
+
+    float sigma = getWinSigma();
+    float scale = 1.f / (2.f * sigma * sigma);
+    Mat gaussian_lut(1, 512, CV_32FC1);
+    int idx = 0;
+    for(int i=-8; i<8; i++)
+        for(int j=-8; j<8; j++)
+            gaussian_lut.at<float>(idx++) = std::exp(-(j * j + i * i) * scale);
+    for(int i=-8; i<8; i++)
+        for(int j=-8; j<8; j++)
+            gaussian_lut.at<float>(idx++) = (8.f - fabs(j + 0.5f)) * (8.f - fabs(i + 0.5f)) / 64.f;
+
+    gauss_w_lut.upload(gaussian_lut);
 }
 
 void cv::ocl::HOGDescriptor::computeGradient(const oclMat &img, oclMat &grad, oclMat &qangle)
@@ -221,10 +258,12 @@ void cv::ocl::HOGDescriptor::computeGradient(const oclMat &img, oclMat &grad, oc
     switch (img.type())
     {
     case CV_8UC1:
-        hog::compute_gradients_8UC1(effect_size.height, effect_size.width, img, angleScale, grad, qangle, gamma_correction);
+        hog::compute_gradients_8UC1(effect_size.height, effect_size.width, img,
+            angleScale, grad, qangle, gamma_correction);
         break;
     case CV_8UC4:
-        hog::compute_gradients_8UC4(effect_size.height, effect_size.width, img, angleScale, grad, qangle, gamma_correction);
+        hog::compute_gradients_8UC4(effect_size.height, effect_size.width, img,
+            angleScale, grad, qangle, gamma_correction);
         break;
     }
 }
@@ -232,19 +271,21 @@ void cv::ocl::HOGDescriptor::computeGradient(const oclMat &img, oclMat &grad, oc
 
 void cv::ocl::HOGDescriptor::computeBlockHistograms(const oclMat &img)
 {
-    computeGradient(img, grad, qangle);
+    computeGradient(img, this->grad, this->qangle);
 
-    hog::compute_hists(nbins, block_stride.width, block_stride.height, effect_size.height, effect_size.width,
-                       grad, qangle, (float)getWinSigma(), block_hists);
+    hog::compute_hists(nbins, block_stride.width, block_stride.height, effect_size.height,
+        effect_size.width, grad, qangle, gauss_w_lut, block_hists);
 
-    hog::normalize_hists(nbins, block_stride.width, block_stride.height, effect_size.height, effect_size.width,
-                         block_hists, (float)threshold_L2hys);
+    hog::normalize_hists(nbins, block_stride.width, block_stride.height, effect_size.height,
+        effect_size.width, block_hists, (float)threshold_L2hys);
 }
 
 
-void cv::ocl::HOGDescriptor::getDescriptors(const oclMat &img, Size win_stride, oclMat &descriptors, int descr_format)
+void cv::ocl::HOGDescriptor::getDescriptors(const oclMat &img, Size win_stride,
+                                            oclMat &descriptors, int descr_format)
 {
-    CV_Assert(win_stride.width % block_stride.width == 0 && win_stride.height % block_stride.height == 0);
+    CV_Assert(win_stride.width % block_stride.width == 0 &&
+        win_stride.height % block_stride.height == 0);
 
     init_buffer(img, win_stride);
 
@@ -254,25 +295,29 @@ void cv::ocl::HOGDescriptor::getDescriptors(const oclMat &img, Size win_stride,
     Size blocks_per_win = numPartsWithin(win_size, block_size, block_stride);
     Size wins_per_img   = numPartsWithin(effect_size, win_size, win_stride);
 
-    descriptors.create(wins_per_img.area(), static_cast<int>(blocks_per_win.area() * block_hist_size), CV_32F);
+    descriptors.create(wins_per_img.area(),
+        static_cast<int>(blocks_per_win.area() * block_hist_size), CV_32F);
 
     switch (descr_format)
     {
     case DESCR_FORMAT_ROW_BY_ROW:
-        hog::extract_descrs_by_rows(win_size.height, win_size.width, block_stride.height, block_stride.width,
-                                    win_stride.height, win_stride.width, effect_size.height, effect_size.width, block_hists, descriptors);
+        hog::extract_descrs_by_rows(win_size.height, win_size.width,
+            block_stride.height, block_stride.width, win_stride.height, win_stride.width,
+            effect_size.height, effect_size.width, block_hists, descriptors);
         break;
     case DESCR_FORMAT_COL_BY_COL:
-        hog::extract_descrs_by_cols(win_size.height, win_size.width, block_stride.height, block_stride.width,
-                                    win_stride.height, win_stride.width, effect_size.height, effect_size.width, block_hists, descriptors);
+        hog::extract_descrs_by_cols(win_size.height, win_size.width,
+            block_stride.height, block_stride.width, win_stride.height, win_stride.width,
+            effect_size.height, effect_size.width, block_hists, descriptors);
         break;
     default:
-        CV_Error(CV_StsBadArg, "Unknown descriptor format");
+        CV_Error(Error::StsBadArg, "Unknown descriptor format");
     }
 }
 
 
-void cv::ocl::HOGDescriptor::detect(const oclMat &img, std::vector<Point> &hits, double hit_threshold, Size win_stride, Size padding)
+void cv::ocl::HOGDescriptor::detect(const oclMat &img, std::vector<Point> &hits,
+                                    double hit_threshold, Size win_stride, Size padding)
 {
     CV_Assert(img.type() == CV_8UC1 || img.type() == CV_8UC4);
     CV_Assert(padding == Size(0, 0));
@@ -284,14 +329,16 @@ void cv::ocl::HOGDescriptor::detect(const oclMat &img, std::vector<Point> &hits,
     if (win_stride == Size())
         win_stride = block_stride;
     else
-        CV_Assert(win_stride.width % block_stride.width == 0 && win_stride.height % block_stride.height == 0);
+        CV_Assert(win_stride.width % block_stride.width == 0 &&
+            win_stride.height % block_stride.height == 0);
     init_buffer(img, win_stride);
 
     computeBlockHistograms(img);
 
-    hog::classify_hists(win_size.height, win_size.width, block_stride.height, block_stride.width,
-                        win_stride.height, win_stride.width, effect_size.height, effect_size.width, block_hists,
-                        detector, (float)free_coef, (float)hit_threshold, labels);
+    hog::classify_hists(win_size.height, win_size.width, block_stride.height,
+        block_stride.width, win_stride.height, win_stride.width,
+        effect_size.height, effect_size.width, block_hists, detector,
+        (float)free_coef, (float)hit_threshold, labels);
 
     labels.download(labels_host);
     unsigned char *vec = labels_host.ptr();
@@ -307,8 +354,9 @@ void cv::ocl::HOGDescriptor::detect(const oclMat &img, std::vector<Point> &hits,
 
 
 
-void cv::ocl::HOGDescriptor::detectMultiScale(const oclMat &img, std::vector<Rect> &found_locations, double hit_threshold,
-        Size win_stride, Size padding, double scale0, int group_threshold)
+void cv::ocl::HOGDescriptor::detectMultiScale(const oclMat &img, std::vector<Rect> &found_locations,
+                                              double hit_threshold, Size win_stride, Size padding,
+                                              double scale0, int group_threshold)
 {
     CV_Assert(img.type() == CV_8UC1 || img.type() == CV_8UC4);
     CV_Assert(scale0 > 1);
@@ -334,7 +382,8 @@ void cv::ocl::HOGDescriptor::detectMultiScale(const oclMat &img, std::vector<Rec
     if (win_stride == Size())
         win_stride = block_stride;
     else
-        CV_Assert(win_stride.width % block_stride.width == 0 && win_stride.height % block_stride.height == 0);
+        CV_Assert(win_stride.width % block_stride.width == 0 &&
+            win_stride.height % block_stride.height == 0);
     init_buffer(img, win_stride);
     image_scale.create(img.size(), img.type());
 
@@ -348,16 +397,17 @@ void cv::ocl::HOGDescriptor::detectMultiScale(const oclMat &img, std::vector<Rec
         }
         else
         {
-            hog::resize( img, image_scale, effect_size);
+            resize(img, image_scale, effect_size);
             detect(image_scale, locations, hit_threshold, win_stride, padding);
         }
-        Size scaled_win_size(cvRound(win_size.width * scale), cvRound(win_size.height * scale));
+        Size scaled_win_size(cvRound(win_size.width * scale),
+            cvRound(win_size.height * scale));
         for (size_t j = 0; j < locations.size(); j++)
-            all_candidates.push_back(Rect(Point2d((CvPoint)locations[j]) * scale, scaled_win_size));
+            all_candidates.push_back(Rect(Point2d(locations[j]) * scale, scaled_win_size));
     }
 
     found_locations.assign(all_candidates.begin(), all_candidates.end());
-    groupRectangles(found_locations, group_threshold, 0.2/*magic number copied from CPU version*/);
+    groupRectangles(found_locations, group_threshold, 0.2);
 }
 
 int cv::ocl::HOGDescriptor::numPartsWithin(int size, int part_size, int stride)
@@ -365,9 +415,11 @@ int cv::ocl::HOGDescriptor::numPartsWithin(int size, int part_size, int stride)
     return (size - part_size + stride) / stride;
 }
 
-cv::Size cv::ocl::HOGDescriptor::numPartsWithin(cv::Size size, cv::Size part_size, cv::Size stride)
+cv::Size cv::ocl::HOGDescriptor::numPartsWithin(cv::Size size, cv::Size part_size,
+                                                cv::Size stride)
 {
-    return Size(numPartsWithin(size.width, part_size.width, stride.width), numPartsWithin(size.height, part_size.height, stride.height));
+    return Size(numPartsWithin(size.width, part_size.width, stride.width),
+        numPartsWithin(size.height, part_size.height, stride.height));
 }
 
 std::vector<float> cv::ocl::HOGDescriptor::getDefaultPeopleDetector()
@@ -1548,8 +1600,9 @@ static int power_2up(unsigned int n)
     return -1; // Input is too big
 }
 
-void cv::ocl::device::hog::set_up_constants(int nbins, int block_stride_x, int block_stride_y,
-        int nblocks_win_x, int nblocks_win_y)
+void cv::ocl::device::hog::set_up_constants(int nbins,
+                                            int block_stride_x, int block_stride_y,
+                                            int nblocks_win_x, int nblocks_win_y)
 {
     cnbins = nbins;
     cblock_stride_x = block_stride_x;
@@ -1560,106 +1613,184 @@ void cv::ocl::device::hog::set_up_constants(int nbins, int block_stride_x, int b
     int block_hist_size = nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y;
     cblock_hist_size = block_hist_size;
 
-    int block_hist_size_2up = power_2up(block_hist_size);
-    cblock_hist_size_2up = block_hist_size_2up;
-
     int descr_width = nblocks_win_x * block_hist_size;
     cdescr_width = descr_width;
+    cdescr_height = nblocks_win_y;
 
     int descr_size = descr_width * nblocks_win_y;
     cdescr_size = descr_size;
 }
 
-void cv::ocl::device::hog::compute_hists(int nbins, int block_stride_x, int block_stride_y,
-        int height, int width, const cv::ocl::oclMat &grad,
-        const cv::ocl::oclMat &qangle, float sigma, cv::ocl::oclMat &block_hists)
+void cv::ocl::device::hog::compute_hists(int nbins,
+                                         int block_stride_x, int block_stride_y,
+                                         int height, int width,
+                                         const cv::ocl::oclMat &grad,
+                                         const cv::ocl::oclMat &qangle,
+                                         const cv::ocl::oclMat &gauss_w_lut,
+                                         cv::ocl::oclMat &block_hists)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "compute_hists_kernel";
     std::vector< std::pair<size_t, const void *> > args;
+    String kernelName = "compute_hists_lut_kernel";
 
-    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
-    int img_block_height = (height - CELLS_PER_BLOCK_Y * CELL_HEIGHT + block_stride_y) / block_stride_y;
-
-    size_t globalThreads[3] = { img_block_width * 32, img_block_height * 2, 1 };
-    size_t localThreads[3] = { 32, 2, 1 };
+    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x)
+        / block_stride_x;
+    int img_block_height = (height - CELLS_PER_BLOCK_Y * CELL_HEIGHT + block_stride_y)
+        / block_stride_y;
+    int blocks_total = img_block_width * img_block_height;
 
     int grad_quadstep = grad.step >> 2;
     int qangle_step = qangle.step;
 
-    // Precompute gaussian spatial window parameter
-    float scale = 1.f / (2.f * sigma * sigma);
+    int blocks_in_group = 4;
+    size_t localThreads[3] = { blocks_in_group * 24, 2, 1 };
+    size_t globalThreads[3] = {
+        divUp(img_block_width * img_block_height, blocks_in_group) * localThreads[0], 2, 1 };
 
     int hists_size = (nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y * 12) * sizeof(float);
     int final_hists_size = (nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y) * sizeof(float);
-    int smem = hists_size + final_hists_size;
 
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&width));
+    int smem = (hists_size + final_hists_size) * blocks_in_group;
+
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cblock_stride_x));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cblock_stride_y));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cnbins));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cblock_hist_size));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_block_width));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&blocks_in_group));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&blocks_total));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&grad_quadstep));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&qangle_step));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&grad.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&qangle.data));
-    args.push_back( std::make_pair( sizeof(cl_float), (void *)&scale));
+    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&gauss_w_lut.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&block_hists.data));
     args.push_back( std::make_pair( smem, (void *)NULL));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    if(hog_device_cpu)
+    {
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+            localThreads, args, -1, -1, "-D CPU");
+    }else
+    {
+        cl_kernel kernel = openCLGetKernelFromSource(clCxt, &objdetect_hog, kernelName);
+        int wave_size = queryDeviceInfo<WAVEFRONT_SIZE, int>(kernel);
+        char opt[32] = {0};
+        sprintf(opt, "-D WAVE_SIZE=%d", wave_size);
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+            localThreads, args, -1, -1, opt);
+    }
 }
 
-void cv::ocl::device::hog::normalize_hists(int nbins, int block_stride_x, int block_stride_y,
-        int height, int width, cv::ocl::oclMat &block_hists, float threshold)
+void cv::ocl::device::hog::normalize_hists(int nbins,
+                                           int block_stride_x, int block_stride_y,
+                                           int height, int width,
+                                           cv::ocl::oclMat &block_hists,
+                                           float threshold)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "normalize_hists_kernel";
     std::vector< std::pair<size_t, const void *> > args;
+    String kernelName;
 
     int block_hist_size = nbins * CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y;
-    int nthreads = power_2up(block_hist_size);
+    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x)
+        / block_stride_x;
+    int img_block_height = (height - CELLS_PER_BLOCK_Y * CELL_HEIGHT + block_stride_y)
+        / block_stride_y;
+    int nthreads;
+    size_t globalThreads[3] = { 1, 1, 1  };
+    size_t localThreads[3] = { 1, 1, 1  };
 
-    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
-    int img_block_height = (height - CELLS_PER_BLOCK_Y * CELL_HEIGHT + block_stride_y) / block_stride_y;
-    size_t globalThreads[3] = { img_block_width * nthreads, img_block_height, 1 };
-    size_t localThreads[3] = { nthreads, 1, 1  };
+    if ( nbins == 9 )
+    {
+        /* optimized for the case of 9 bins */
+        kernelName = "normalize_hists_36_kernel";
+        int blocks_in_group = NTHREADS / block_hist_size;
+        nthreads = blocks_in_group * block_hist_size;
+        int num_groups = divUp( img_block_width * img_block_height, blocks_in_group);
+        globalThreads[0] = nthreads * num_groups;
+        localThreads[0] = nthreads;
+    }
+    else
+    {
+        kernelName = "normalize_hists_kernel";
+        nthreads = power_2up(block_hist_size);
+        globalThreads[0] = img_block_width * nthreads;
+        globalThreads[1] = img_block_height;
+        localThreads[0] = nthreads;
 
-    if ((nthreads < 32) || (nthreads > 512) )
-        cv::ocl::error("normalize_hists: histogram's size is too small or too big", __FILE__, __LINE__, "normalize_hists");
+        if ((nthreads < 32) || (nthreads > 512) )
+            cv::error(Error::StsBadArg, "normalize_hists: histogram's size is too small or too big",
+                "normalize_hists", __FILE__, __LINE__);
+
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&nthreads));
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_hist_size));
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_block_width));
+    }
 
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&nthreads));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&block_hist_size));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_block_width));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&block_hists.data));
     args.push_back( std::make_pair( sizeof(cl_float), (void *)&threshold));
     args.push_back( std::make_pair( nthreads * sizeof(float), (void *)NULL));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    if(hog_device_cpu)
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+                             localThreads, args, -1, -1, "-D CPU");
+    else
+    {
+        cl_kernel kernel = openCLGetKernelFromSource(clCxt, &objdetect_hog, kernelName);
+        int wave_size = queryDeviceInfo<WAVEFRONT_SIZE, int>(kernel);
+        char opt[32] = {0};
+        sprintf(opt, "-D WAVE_SIZE=%d", wave_size);
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+                             localThreads, args, -1, -1, opt);
+    }
 }
 
-void cv::ocl::device::hog::classify_hists(int win_height, int win_width, int block_stride_y,
-        int block_stride_x, int win_stride_y, int win_stride_x, int height,
-        int width, const cv::ocl::oclMat &block_hists, const cv::ocl::oclMat &coefs, float free_coef,
-        float threshold, cv::ocl::oclMat &labels)
+void cv::ocl::device::hog::classify_hists(int win_height, int win_width,
+                                          int block_stride_y, int block_stride_x,
+                                          int win_stride_y, int win_stride_x,
+                                          int height, int width,
+                                          const cv::ocl::oclMat &block_hists,
+                                          const cv::ocl::oclMat &coefs,
+                                          float free_coef, float threshold,
+                                          cv::ocl::oclMat &labels)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "classify_hists_kernel";
     std::vector< std::pair<size_t, const void *> > args;
 
+    int nthreads;
+    String kernelName;
+    switch (cdescr_width)
+    {
+    case 180:
+        nthreads = 180;
+        kernelName = "classify_hists_180_kernel";
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_width));
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_height));
+        break;
+    case 252:
+        nthreads = 256;
+        kernelName = "classify_hists_252_kernel";
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_width));
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_height));
+        break;
+    default:
+        nthreads = 256;
+        kernelName = "classify_hists_kernel";
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_size));
+        args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_width));
+    }
+
     int win_block_stride_x = win_stride_x / block_stride_x;
     int win_block_stride_y = win_stride_y / block_stride_y;
     int img_win_width = (width - win_width + win_stride_x) / win_stride_x;
     int img_win_height = (height - win_height + win_stride_y) / win_stride_y;
-    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
-
-    size_t globalThreads[3] = { img_win_width * NTHREADS, img_win_height, 1 };
-    size_t localThreads[3] = { NTHREADS, 1, 1 };
+    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) /
+        block_stride_x;
 
+    size_t globalThreads[3] = { img_win_width * nthreads, img_win_height, 1 };
+    size_t localThreads[3] = { nthreads, 1, 1 };
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cblock_hist_size));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_size));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&cdescr_width));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_win_width));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&img_block_width));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&win_block_stride_x));
@@ -1670,22 +1801,37 @@ void cv::ocl::device::hog::classify_hists(int win_height, int win_width, int blo
     args.push_back( std::make_pair( sizeof(cl_float), (void *)&threshold));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&labels.data));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    if(hog_device_cpu)
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+                             localThreads, args, -1, -1, "-D CPU");
+    else
+    {
+        cl_kernel kernel = openCLGetKernelFromSource(clCxt, &objdetect_hog, kernelName);
+        int wave_size = queryDeviceInfo<WAVEFRONT_SIZE, int>(kernel);
+        char opt[32] = {0};
+        sprintf(opt, "-D WAVE_SIZE=%d", wave_size);
+        openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+                             localThreads, args, -1, -1, opt);
+    }
 }
 
-void cv::ocl::device::hog::extract_descrs_by_rows(int win_height, int win_width, int block_stride_y, int block_stride_x,
-        int win_stride_y, int win_stride_x, int height, int width,
-        const cv::ocl::oclMat &block_hists, cv::ocl::oclMat &descriptors)
+void cv::ocl::device::hog::extract_descrs_by_rows(int win_height, int win_width,
+                                                  int block_stride_y, int block_stride_x,
+                                                  int win_stride_y, int win_stride_x,
+                                                  int height, int width,
+                                                  const cv::ocl::oclMat &block_hists,
+                                                  cv::ocl::oclMat &descriptors)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "extract_descrs_by_rows_kernel";
+    String kernelName = "extract_descrs_by_rows_kernel";
     std::vector< std::pair<size_t, const void *> > args;
 
     int win_block_stride_x = win_stride_x / block_stride_x;
     int win_block_stride_y = win_stride_y / block_stride_y;
     int img_win_width = (width - win_width + win_stride_x) / win_stride_x;
     int img_win_height = (height - win_height + win_stride_y) / win_stride_y;
-    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
+    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) /
+        block_stride_x;
     int descriptors_quadstep = descriptors.step >> 2;
 
     size_t globalThreads[3] = { img_win_width * NTHREADS, img_win_height, 1 };
@@ -1701,22 +1847,27 @@ void cv::ocl::device::hog::extract_descrs_by_rows(int win_height, int win_width,
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&block_hists.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&descriptors.data));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+        localThreads, args, -1, -1);
 }
 
-void cv::ocl::device::hog::extract_descrs_by_cols(int win_height, int win_width, int block_stride_y, int block_stride_x,
-        int win_stride_y, int win_stride_x, int height, int width,
-        const cv::ocl::oclMat &block_hists, cv::ocl::oclMat &descriptors)
+void cv::ocl::device::hog::extract_descrs_by_cols(int win_height, int win_width,
+                                                  int block_stride_y, int block_stride_x,
+                                                  int win_stride_y, int win_stride_x,
+                                                  int height, int width,
+                                                  const cv::ocl::oclMat &block_hists,
+                                                  cv::ocl::oclMat &descriptors)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "extract_descrs_by_cols_kernel";
+    String kernelName = "extract_descrs_by_cols_kernel";
     std::vector< std::pair<size_t, const void *> > args;
 
     int win_block_stride_x = win_stride_x / block_stride_x;
     int win_block_stride_y = win_stride_y / block_stride_y;
     int img_win_width = (width - win_width + win_stride_x) / win_stride_x;
     int img_win_height = (height - win_height + win_stride_y) / win_stride_y;
-    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) / block_stride_x;
+    int img_block_width = (width - CELLS_PER_BLOCK_X * CELL_WIDTH + block_stride_x) /
+        block_stride_x;
     int descriptors_quadstep = descriptors.step >> 2;
 
     size_t globalThreads[3] = { img_win_width * NTHREADS, img_win_height, 1 };
@@ -1733,19 +1884,19 @@ void cv::ocl::device::hog::extract_descrs_by_cols(int win_height, int win_width,
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&block_hists.data));
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&descriptors.data));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+        localThreads, args, -1, -1);
 }
 
-static inline int divUp(int total, int grain)
-{
-    return (total + grain - 1) / grain;
-}
-
-void cv::ocl::device::hog::compute_gradients_8UC1(int height, int width, const cv::ocl::oclMat &img,
-        float angle_scale, cv::ocl::oclMat &grad, cv::ocl::oclMat &qangle, bool correct_gamma)
+void cv::ocl::device::hog::compute_gradients_8UC1(int height, int width,
+                                                  const cv::ocl::oclMat &img,
+                                                  float angle_scale,
+                                                  cv::ocl::oclMat &grad,
+                                                  cv::ocl::oclMat &qangle,
+                                                  bool correct_gamma)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "compute_gradients_8UC1_kernel";
+    String kernelName = "compute_gradients_8UC1_kernel";
     std::vector< std::pair<size_t, const void *> > args;
 
     size_t localThreads[3] = { NTHREADS, 1, 1 };
@@ -1767,14 +1918,19 @@ void cv::ocl::device::hog::compute_gradients_8UC1(int height, int width, const c
     args.push_back( std::make_pair( sizeof(cl_char), (void *)&correctGamma));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cnbins));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+        localThreads, args, -1, -1);
 }
 
-void cv::ocl::device::hog::compute_gradients_8UC4(int height, int width, const cv::ocl::oclMat &img,
-        float angle_scale, cv::ocl::oclMat &grad, cv::ocl::oclMat &qangle, bool correct_gamma)
+void cv::ocl::device::hog::compute_gradients_8UC4(int height, int width,
+                                                  const cv::ocl::oclMat &img,
+                                                  float angle_scale,
+                                                  cv::ocl::oclMat &grad,
+                                                  cv::ocl::oclMat &qangle,
+                                                  bool correct_gamma)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "compute_gradients_8UC4_kernel";
+    String kernelName = "compute_gradients_8UC4_kernel";
     std::vector< std::pair<size_t, const void *> > args;
 
     size_t localThreads[3] = { NTHREADS, 1, 1 };
@@ -1797,37 +1953,6 @@ void cv::ocl::device::hog::compute_gradients_8UC4(int height, int width, const c
     args.push_back( std::make_pair( sizeof(cl_char), (void *)&correctGamma));
     args.push_back( std::make_pair( sizeof(cl_int), (void *)&cnbins));
 
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
-}
-
-void cv::ocl::device::hog::resize( const oclMat &src, oclMat &dst, const Size sz)
-{
-    CV_Assert( (src.channels() == dst.channels()) );
-    Context *clCxt = Context::getContext();
-
-    std::string kernelName = (src.type() == CV_8UC1) ? "resize_8UC1_kernel" : "resize_8UC4_kernel";
-    size_t blkSizeX = 16, blkSizeY = 16;
-    size_t glbSizeX = sz.width % blkSizeX == 0 ? sz.width : (sz.width / blkSizeX + 1) * blkSizeX;
-    size_t glbSizeY = sz.height % blkSizeY == 0 ? sz.height : (sz.height / blkSizeY + 1) * blkSizeY;
-    size_t globalThreads[3] = {glbSizeX, glbSizeY, 1};
-    size_t localThreads[3] = {blkSizeX, blkSizeY, 1};
-
-    float ifx = (float)src.cols / sz.width;
-    float ify = (float)src.rows / sz.height;
-
-    std::vector< std::pair<size_t, const void *> > args;
-    args.push_back( std::make_pair(sizeof(cl_mem), (void *)&dst.data));
-    args.push_back( std::make_pair(sizeof(cl_mem), (void *)&src.data));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&dst.offset));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&src.offset));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&dst.step));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&src.step));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&src.cols));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&src.rows));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&sz.width));
-    args.push_back( std::make_pair(sizeof(cl_int), (void *)&sz.height));
-    args.push_back( std::make_pair(sizeof(cl_float), (void *)&ifx));
-    args.push_back( std::make_pair(sizeof(cl_float), (void *)&ify));
-
-    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads, localThreads, args, -1, -1);
+    openCLExecuteKernel2(clCxt, &objdetect_hog, kernelName, globalThreads,
+        localThreads, args, -1, -1);
 }
diff --git a/modules/ocl/src/hough.cpp b/modules/ocl/src/hough.cpp
index f5d770c952..66e48875c5 100644
--- a/modules/ocl/src/hough.cpp
+++ b/modules/ocl/src/hough.cpp
@@ -73,7 +73,7 @@ namespace
 
         int totalCount = 0;
         int err = CL_SUCCESS;
-        cl_mem counter = clCreateBuffer(src.clCxt->impl->clContext,
+        cl_mem counter = clCreateBuffer((cl_context)src.clCxt->oclContext(),
                                         CL_MEM_COPY_HOST_PTR,
                                         sizeof(int),
                                         &totalCount,
@@ -98,7 +98,7 @@ namespace
         args.push_back( std::make_pair( sizeof(cl_mem)  , (void *)&counter ));
 
         openCLExecuteKernel(src.clCxt, &imgproc_hough, "buildPointList", globalThreads, localThreads, args, -1, -1);
-        openCLSafeCall(clEnqueueReadBuffer(src.clCxt->impl->clCmdQueue, counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueReadBuffer((cl_command_queue)src.clCxt->oclCommandQueue(), counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
         openCLSafeCall(clReleaseMemObject(counter));
 
         return totalCount;
@@ -143,7 +143,7 @@ namespace
     {
         int totalCount = 0;
         int err = CL_SUCCESS;
-        cl_mem counter = clCreateBuffer(accum.clCxt->impl->clContext,
+        cl_mem counter = clCreateBuffer((cl_context)accum.clCxt->oclContext(),
                                         CL_MEM_COPY_HOST_PTR,
                                         sizeof(int),
                                         &totalCount,
@@ -169,7 +169,7 @@ namespace
 
         openCLExecuteKernel(accum.clCxt, &imgproc_hough, "buildCentersList", globalThreads, localThreads, args, -1, -1);
 
-        openCLSafeCall(clEnqueueReadBuffer(accum.clCxt->impl->clCmdQueue, counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueReadBuffer((cl_command_queue)accum.clCxt->oclCommandQueue(), counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
         openCLSafeCall(clReleaseMemObject(counter));
 
         return totalCount;
@@ -182,14 +182,14 @@ namespace
     {
         int totalCount = 0;
         int err = CL_SUCCESS;
-        cl_mem counter = clCreateBuffer(circles.clCxt->impl->clContext,
+        cl_mem counter = clCreateBuffer((cl_context)circles.clCxt->oclContext(),
                                         CL_MEM_COPY_HOST_PTR,
                                         sizeof(int),
                                         &totalCount,
                                         &err);
         openCLSafeCall(err);
 
-        const size_t blkSizeX = circles.clCxt->impl->maxWorkGroupSize;
+        const size_t blkSizeX = circles.clCxt->maxWorkGroupSize();
         size_t localThreads[3] = { blkSizeX, 1, 1 };
 
         const size_t glbSizeX = centersCount * blkSizeX;
@@ -216,7 +216,7 @@ namespace
 
         openCLExecuteKernel(circles.clCxt, &imgproc_hough, "circlesAccumRadius", globalThreads, localThreads, args, -1, -1);
 
-        openCLSafeCall(clEnqueueReadBuffer(circles.clCxt->impl->clCmdQueue, counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
+        openCLSafeCall(clEnqueueReadBuffer((cl_command_queue)circles.clCxt->oclCommandQueue(), counter, CL_TRUE, 0, sizeof(int), &totalCount, 0, NULL, NULL));
 
         openCLSafeCall(clReleaseMemObject(counter));
 
@@ -242,7 +242,7 @@ void cv::ocl::HoughCircles(const oclMat& src, oclMat& circles, HoughCirclesBuf&
     CV_Assert(src.type() == CV_8UC1);
     CV_Assert(src.cols < std::numeric_limits<unsigned short>::max());
     CV_Assert(src.rows < std::numeric_limits<unsigned short>::max());
-    CV_Assert(method == CV_HOUGH_GRADIENT);
+    CV_Assert(method == HOUGH_GRADIENT);
     CV_Assert(dp > 0);
     CV_Assert(minRadius > 0 && maxRadius > minRadius);
     CV_Assert(cannyThreshold > 0);
@@ -283,7 +283,7 @@ void cv::ocl::HoughCircles(const oclMat& src, oclMat& circles, HoughCirclesBuf&
         unsigned int* oldBuf = oldBuf_;
         unsigned int* newBuf = newBuf_;
 
-        openCLSafeCall(clEnqueueReadBuffer(buf.centers.clCxt->impl->clCmdQueue,
+        openCLSafeCall(clEnqueueReadBuffer((cl_command_queue)buf.centers.clCxt->oclCommandQueue(),
                                            (cl_mem)buf.centers.data,
                                            CL_TRUE,
                                            0,
@@ -357,7 +357,7 @@ void cv::ocl::HoughCircles(const oclMat& src, oclMat& circles, HoughCirclesBuf&
             }
         }
 
-        openCLSafeCall(clEnqueueWriteBuffer(buf.centers.clCxt->impl->clCmdQueue,
+        openCLSafeCall(clEnqueueWriteBuffer((cl_command_queue)buf.centers.clCxt->oclCommandQueue(),
                                             (cl_mem)buf.centers.data,
                                             CL_TRUE,
                                             0,
@@ -385,7 +385,7 @@ void cv::ocl::HoughCircles(const oclMat& src, oclMat& circles, HoughCirclesBuf&
 void cv::ocl::HoughCirclesDownload(const oclMat& d_circles, cv::OutputArray h_circles_)
 {
     // FIX ME: garbage values are copied!
-    CV_Error(CV_StsNotImplemented, "HoughCirclesDownload is not implemented");
+    CV_Error(Error::StsNotImplemented, "HoughCirclesDownload is not implemented");
 
     if (d_circles.empty())
     {
diff --git a/modules/ocl/src/imgproc.cpp b/modules/ocl/src/imgproc.cpp
index 3501636de5..0dc7fe9cea 100644
--- a/modules/ocl/src/imgproc.cpp
+++ b/modules/ocl/src/imgproc.cpp
@@ -25,6 +25,8 @@
 //    Xu Pang, pangxu010@163.com
 //    Wu Zailong, bullet@yeah.net
 //    Wenju He, wenju@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
+//    Sen Liu, swjtuls1987@126.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -79,6 +81,8 @@ namespace cv
         extern const char *imgproc_calcHarris;
         extern const char *imgproc_calcMinEigenVal;
         extern const char *imgproc_convolve;
+        extern const char *imgproc_mulAndScaleSpectrums;
+        extern const char *imgproc_clahe;
         ////////////////////////////////////OpenCL call wrappers////////////////////////////
 
         template <typename T> struct index_and_sizeof;
@@ -123,7 +127,7 @@ namespace cv
 
             uchar thresh_uchar = cvFloor(thresh);
             uchar max_val = cvRound(maxVal);
-            std::string kernelName = "threshold";
+            String kernelName = "threshold";
 
             size_t cols = (dst.cols + (dst.offset % 16) + 15) / 16;
             size_t bSizeX = 16, bSizeY = 16;
@@ -159,7 +163,7 @@ namespace cv
             int src_offset = (src.offset >> 2);
             int src_step = (src.step >> 2);
 
-            std::string kernelName = "threshold";
+            String kernelName = "threshold";
 
             size_t cols = (dst.cols + (dst_offset & 3) + 3) / 4;
             //size_t cols = dst.cols;
@@ -216,7 +220,7 @@ namespace cv
             dst.create(map1.size(), src.type());
 
 
-            std::string kernelName;
+            String kernelName;
 
             if( map1.type() == CV_32FC2 && !map2.data )
             {
@@ -268,7 +272,7 @@ namespace cv
             size_t globalThreads[3] = {glbSizeX, glbSizeY, 1};
             size_t localThreads[3] = {blkSizeX, blkSizeY, 1};
 
-
+            float borderFloat[4] = {(float)borderValue[0], (float)borderValue[1], (float)borderValue[2], (float)borderValue[3]};
             std::vector< std::pair<size_t, const void *> > args;
             if(map1.channels() == 2)
             {
@@ -290,7 +294,7 @@ namespace cv
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&cols));
                 float borderFloat[4] = {(float)borderValue[0], (float)borderValue[1], (float)borderValue[2], (float)borderValue[3]};
 
-               if(src.clCxt -> impl -> double_support != 0)
+                if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
                 {
                     args.push_back( std::make_pair(sizeof(cl_double4), (void *)&borderValue));
                 }
@@ -318,13 +322,12 @@ namespace cv
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&map1.cols));
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&map1.rows));
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&cols));
-                if(src.clCxt -> impl -> double_support != 0)
+                if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
                 {
                     args.push_back( std::make_pair(sizeof(cl_double4), (void *)&borderValue));
                 }
                 else
                 {
-                    float borderFloat[4] = {(float)borderValue[0], (float)borderValue[1], (float)borderValue[2], (float)borderValue[3]};
                     args.push_back( std::make_pair(sizeof(cl_float4), (void *)&borderFloat));
                 }
             }
@@ -347,7 +350,7 @@ namespace cv
             int dstStep_in_pixel = dst.step1() / dst.oclchannels();
             int dstoffset_in_pixel = dst.offset / dst.elemSize();
             //printf("%d %d\n",src.step1() , dst.elemSize());
-            std::string kernelName;
+            String kernelName;
             if(interpolation == INTER_LINEAR)
                 kernelName = "resizeLN";
             else if(interpolation == INTER_NEAREST)
@@ -382,7 +385,7 @@ namespace cv
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&src.rows));
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&dst.cols));
                 args.push_back( std::make_pair(sizeof(cl_int), (void *)&dst.rows));
-                if(src.clCxt -> impl -> double_support != 0)
+                if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
                 {
                     args.push_back( std::make_pair(sizeof(cl_double), (void *)&ifx_d));
                     args.push_back( std::make_pair(sizeof(cl_double), (void *)&ify_d));
@@ -426,7 +429,7 @@ namespace cv
             {
                 if(dsize.width != (int)(src.cols * fx) || dsize.height != (int)(src.rows * fy))
                 {
-                    CV_Error(CV_StsUnmatchedSizes, "invalid dsize and fx, fy!");
+                    CV_Error(Error::StsUnmatchedSizes, "invalid dsize and fx, fy!");
                 }
             }
             if( dsize == Size() )
@@ -446,7 +449,7 @@ namespace cv
                 resize_gpu( src, dst, fx, fy, interpolation);
                 return;
             }
-            CV_Error(CV_StsUnsupportedFormat, "Non-supported interpolation method");
+            CV_Error(Error::StsUnsupportedFormat, "Non-supported interpolation method");
         }
 
 
@@ -471,7 +474,7 @@ namespace cv
             int dstOffset = dst.offset / dst.oclchannels() / dst.elemSize1();
 
             Context *clCxt = src.clCxt;
-            std::string kernelName = "medianFilter";
+            String kernelName = "medianFilter";
 
 
             std::vector< std::pair<size_t, const void *> > args;
@@ -489,18 +492,18 @@ namespace cv
 
             if(m == 3)
             {
-                std::string kernelName = "medianFilter3";
+                String kernelName = "medianFilter3";
                 openCLExecuteKernel(clCxt, &imgproc_median, kernelName, globalThreads, localThreads, args, src.oclchannels(), src.depth());
             }
             else if(m == 5)
             {
-                std::string kernelName = "medianFilter5";
+                String kernelName = "medianFilter5";
                 openCLExecuteKernel(clCxt, &imgproc_median, kernelName, globalThreads, localThreads, args, src.oclchannels(), src.depth());
             }
             else
             {
-                CV_Error(CV_StsUnsupportedFormat, "Non-supported filter length");
-                //std::string kernelName = "medianFilter";
+                CV_Error(Error::StsUnsupportedFormat, "Non-supported filter length");
+                //String kernelName = "medianFilter";
                 //args.push_back( std::make_pair( sizeof(cl_int),(void*)&m));
 
                 //openCLExecuteKernel(clCxt,&imgproc_median,kernelName,globalThreads,localThreads,args,src.oclchannels(),-1);
@@ -520,7 +523,7 @@ namespace cv
                         (bordertype != cv::BORDER_CONSTANT) &&
                         (bordertype != cv::BORDER_REPLICATE))
                 {
-                    CV_Error(CV_StsBadArg, "unsupported border type");
+                    CV_Error(Error::StsBadArg, "unsupported border type");
                 }
             }
             bordertype &= ~cv::BORDER_ISOLATED;
@@ -547,9 +550,9 @@ namespace cv
             }
             if(bordertype_index == sizeof(__bordertype) / sizeof(int))
             {
-                CV_Error(CV_StsBadArg, "unsupported border type");
+                CV_Error(Error::StsBadArg, "unsupported border type");
             }
-            std::string kernelName = "copymakeborder";
+            String kernelName = "copymakeborder";
             size_t localThreads[3] = {16, 16, 1};
             size_t globalThreads[3] = {(dst.cols + localThreads[0] - 1) / localThreads[0] *localThreads[0],
                                        (dst.rows + localThreads[1] - 1) / localThreads[1] *localThreads[1], 1
@@ -602,7 +605,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_8S:
@@ -621,7 +624,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_char4) , (void *)&val.cval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_16U:
@@ -640,7 +643,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_16S:
@@ -659,7 +662,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_short4) , (void *)&val.shval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_32S:
@@ -685,7 +688,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_int4) , (void *)&val.ival ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_32F:
@@ -704,7 +707,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_float4) , (void *)&val.fval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             case CV_64F:
@@ -723,11 +726,11 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_double4) , (void *)&val.dval ));
                     break;
                 default:
-                    CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+                    CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
                 }
                 break;
             default:
-                CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+                CV_Error(Error::StsUnsupportedFormat, "unknown depth");
             }
 
             openCLExecuteKernel(src.clCxt, &imgproc_copymakeboder, kernelName, globalThreads, localThreads, args, -1, -1, compile_option);
@@ -819,16 +822,16 @@ namespace cv
                 cl_mem coeffs_cm;
 
                 Context *clCxt = src.clCxt;
-                std::string s[3] = {"NN", "Linear", "Cubic"};
-                std::string kernelName = "warpAffine" + s[interpolation];
+                String s[3] = {"NN", "Linear", "Cubic"};
+                String kernelName = "warpAffine" + s[interpolation];
 
 
-                if(src.clCxt -> impl -> double_support != 0)
+                if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
                 {
                     cl_int st;
-                    coeffs_cm = clCreateBuffer( clCxt->impl->clContext, CL_MEM_READ_WRITE, sizeof(F) * 2 * 3, NULL, &st );
+                    coeffs_cm = clCreateBuffer( (cl_context)clCxt->oclContext(), CL_MEM_READ_WRITE, sizeof(F) * 2 * 3, NULL, &st );
                     openCLVerifyCall(st);
-                    openCLSafeCall(clEnqueueWriteBuffer(clCxt->impl->clCmdQueue, (cl_mem)coeffs_cm, 1, 0, sizeof(F) * 2 * 3, coeffs, 0, 0, 0));
+                    openCLSafeCall(clEnqueueWriteBuffer((cl_command_queue)clCxt->oclCommandQueue(), (cl_mem)coeffs_cm, 1, 0, sizeof(F) * 2 * 3, coeffs, 0, 0, 0));
                 }
                 else
                 {
@@ -838,8 +841,8 @@ namespace cv
                         {
                             float_coeffs[m][n] = coeffs[m][n];
                         }
-                    coeffs_cm = clCreateBuffer( clCxt->impl->clContext, CL_MEM_READ_WRITE, sizeof(float) * 2 * 3, NULL, &st );
-                    openCLSafeCall(clEnqueueWriteBuffer(clCxt->impl->clCmdQueue, (cl_mem)coeffs_cm, 1, 0, sizeof(float) * 2 * 3, float_coeffs, 0, 0, 0));
+                        coeffs_cm = clCreateBuffer( (cl_context)clCxt->oclContext(), CL_MEM_READ_WRITE, sizeof(float) * 2 * 3, NULL, &st );
+                        openCLSafeCall(clEnqueueWriteBuffer((cl_command_queue)clCxt->oclCommandQueue(), (cl_mem)coeffs_cm, 1, 0, sizeof(float) * 2 * 3, float_coeffs, 0, 0, 0));
 
                 }
                 //TODO: improve this kernel
@@ -890,15 +893,15 @@ namespace cv
                 cl_mem coeffs_cm;
 
                 Context *clCxt = src.clCxt;
-                std::string s[3] = {"NN", "Linear", "Cubic"};
-                std::string kernelName = "warpPerspective" + s[interpolation];
+                String s[3] = {"NN", "Linear", "Cubic"};
+                String kernelName = "warpPerspective" + s[interpolation];
 
-                if(src.clCxt -> impl -> double_support != 0)
+                if(src.clCxt->supportsFeature(Context::CL_DOUBLE))
                 {
                     cl_int st;
-                    coeffs_cm = clCreateBuffer( clCxt->impl->clContext, CL_MEM_READ_WRITE, sizeof(double) * 3 * 3, NULL, &st );
+                    coeffs_cm = clCreateBuffer((cl_context) clCxt->oclContext(), CL_MEM_READ_WRITE, sizeof(double) * 3 * 3, NULL, &st );
                     openCLVerifyCall(st);
-                    openCLSafeCall(clEnqueueWriteBuffer(clCxt->impl->clCmdQueue, (cl_mem)coeffs_cm, 1, 0, sizeof(double) * 3 * 3, coeffs, 0, 0, 0));
+                    openCLSafeCall(clEnqueueWriteBuffer((cl_command_queue)clCxt->oclCommandQueue(), (cl_mem)coeffs_cm, 1, 0, sizeof(double) * 3 * 3, coeffs, 0, 0, 0));
                 }
                 else
                 {
@@ -907,9 +910,9 @@ namespace cv
                         for(int n = 0; n < 3; n++)
                             float_coeffs[m][n] = coeffs[m][n];
 
-                    coeffs_cm = clCreateBuffer( clCxt->impl->clContext, CL_MEM_READ_WRITE, sizeof(float) * 3 * 3, NULL, &st );
+                    coeffs_cm = clCreateBuffer((cl_context) clCxt->oclContext(), CL_MEM_READ_WRITE, sizeof(float) * 3 * 3, NULL, &st );
                     openCLVerifyCall(st);
-                    openCLSafeCall(clEnqueueWriteBuffer(clCxt->impl->clCmdQueue, (cl_mem)coeffs_cm, 1, 0, sizeof(float) * 3 * 3, float_coeffs, 0, 0, 0));
+                    openCLSafeCall(clEnqueueWriteBuffer((cl_command_queue)clCxt->oclCommandQueue(), (cl_mem)coeffs_cm, 1, 0, sizeof(float) * 3 * 3, float_coeffs, 0, 0, 0));
                 }
                 //TODO: improve this kernel
                 size_t blkSizeX = 16, blkSizeY = 16;
@@ -1010,16 +1013,14 @@ namespace cv
             warpPerspective_gpu(src, dst, coeffs, interpolation);
         }
 
-
         ////////////////////////////////////////////////////////////////////////
         // integral
-
         void integral(const oclMat &src, oclMat &sum, oclMat &sqsum)
         {
             CV_Assert(src.type() == CV_8UC1);
-            if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+            if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
             {
-                CV_Error(CV_GpuNotSupported, "select device don't support double");
+                CV_Error(Error::GpuNotSupported, "select device don't support double");
             }
             int vlen = 4;
             int offset = src.offset / vlen;
@@ -1027,13 +1028,24 @@ namespace cv
             int vcols = (pre_invalid + src.cols + vlen - 1) / vlen;
 
             oclMat t_sum , t_sqsum;
-            t_sum.create(src.cols, src.rows, CV_32SC1);
-            t_sqsum.create(src.cols, src.rows, CV_32FC1);
-
             int w = src.cols + 1, h = src.rows + 1;
-            sum.create(h, w, CV_32SC1);
+            int depth;
+            if( src.cols * src.rows <= 2901 * 2901 ) //2901 is the maximum size for int when all values are 255
+            {
+                t_sum.create(src.cols, src.rows, CV_32SC1);
+                sum.create(h, w, CV_32SC1);
+            }
+            else
+            {
+                 //Use float to prevent overflow
+                t_sum.create(src.cols, src.rows, CV_32FC1);
+                sum.create(h, w, CV_32FC1);
+            }
+            t_sqsum.create(src.cols, src.rows, CV_32FC1);
             sqsum.create(h, w, CV_32FC1);
-            int sum_offset = sum.offset / vlen, sqsum_offset = sqsum.offset / vlen;
+            depth = sum.depth();
+            int sum_offset = sum.offset / vlen;
+            int sqsum_offset = sqsum.offset / vlen;
 
             std::vector<std::pair<size_t , const void *> > args;
             args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data ));
@@ -1046,7 +1058,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.step ));
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&t_sum.step));
             size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
-            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_cols", gt, lt, args, -1, -1);
+            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_cols", gt, lt, args, -1, depth);
             args.clear();
             args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
             args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&t_sqsum.data ));
@@ -1060,9 +1072,9 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&sum_offset));
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&sqsum_offset));
             size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
-            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_rows", gt2, lt2, args, -1, -1);
-            //std::cout << "tested" << std::endl;
+            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_rows", gt2, lt2, args, -1, depth);
         }
+
         void integral(const oclMat &src, oclMat &sum)
         {
             CV_Assert(src.type() == CV_8UC1);
@@ -1072,10 +1084,18 @@ namespace cv
             int vcols = (pre_invalid + src.cols + vlen - 1) / vlen;
 
             oclMat t_sum;
-            t_sum.create(src.cols, src.rows, CV_32SC1);
-
             int w = src.cols + 1, h = src.rows + 1;
-            sum.create(h, w, CV_32SC1);
+            int depth;
+            if(src.cols * src.rows <= 2901 * 2901)
+            {
+                t_sum.create(src.cols, src.rows, CV_32SC1);
+                sum.create(h, w, CV_32SC1);
+            }else
+            {
+                 t_sum.create(src.cols, src.rows, CV_32FC1);
+                 sum.create(h, w, CV_32FC1);
+            }
+            depth = sum.depth();
             int sum_offset = sum.offset / vlen;
 
             std::vector<std::pair<size_t , const void *> > args;
@@ -1088,7 +1108,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.step ));
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&t_sum.step));
             size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
-            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_cols", gt, lt, args, -1, -1);
+            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_cols", gt, lt, args, -1, depth);
             args.clear();
             args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
             args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&sum.data ));
@@ -1098,7 +1118,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&sum.step));
             args.push_back( std::make_pair( sizeof(cl_int) , (void *)&sum_offset));
             size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
-            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_rows", gt2, lt2, args, -1, -1);
+            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_rows", gt2, lt2, args, -1, depth);
             //std::cout << "tested" << std::endl;
         }
 
@@ -1133,7 +1153,7 @@ namespace cv
             CV_Assert(Dx.offset == 0 && Dy.offset == 0);
         }
 
-        static void corner_ocl(const char *src_str, std::string kernelName, int block_size, float k, oclMat &Dx, oclMat &Dy,
+        static void corner_ocl(const char *src_str, String kernelName, int block_size, float k, oclMat &Dx, oclMat &Dy,
                         oclMat &dst, int border_type)
         {
             char borderType[30];
@@ -1191,30 +1211,41 @@ namespace cv
         void cornerHarris(const oclMat &src, oclMat &dst, int blockSize, int ksize,
                           double k, int borderType)
         {
-            if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+            oclMat dx, dy;
+            cornerHarris_dxdy(src, dst, dx, dy, blockSize, ksize, k, borderType);
+        }
+
+        void cornerHarris_dxdy(const oclMat &src, oclMat &dst, oclMat &dx, oclMat &dy, int blockSize, int ksize,
+                          double k, int borderType)
+        {
+            if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
             {
-                CV_Error(CV_GpuNotSupported, "select device don't support double");
+                CV_Error(Error::GpuNotSupported, "select device don't support double");
             }
             CV_Assert(src.cols >= blockSize / 2 && src.rows >= blockSize / 2);
-            oclMat Dx, Dy;
             CV_Assert(borderType == cv::BORDER_CONSTANT || borderType == cv::BORDER_REFLECT101 || borderType == cv::BORDER_REPLICATE || borderType == cv::BORDER_REFLECT);
-            extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
+            extractCovData(src, dx, dy, blockSize, ksize, borderType);
             dst.create(src.size(), CV_32F);
-            corner_ocl(imgproc_calcHarris, "calcHarris", blockSize, static_cast<float>(k), Dx, Dy, dst, borderType);
+            corner_ocl(imgproc_calcHarris, "calcHarris", blockSize, static_cast<float>(k), dx, dy, dst, borderType);
         }
 
         void cornerMinEigenVal(const oclMat &src, oclMat &dst, int blockSize, int ksize, int borderType)
         {
-            if(src.clCxt->impl->double_support == 0 && src.depth() == CV_64F)
+            oclMat dx, dy;
+            cornerMinEigenVal_dxdy(src, dst, dx, dy, blockSize, ksize, borderType);
+        }
+
+        void cornerMinEigenVal_dxdy(const oclMat &src, oclMat &dst, oclMat &dx, oclMat &dy, int blockSize, int ksize, int borderType)
+        {
+            if(!src.clCxt->supportsFeature(Context::CL_DOUBLE) && src.depth() == CV_64F)
             {
-                CV_Error(CV_GpuNotSupported, "select device don't support double");
+                CV_Error(Error::GpuNotSupported, "select device don't support double");
             }
             CV_Assert(src.cols >= blockSize / 2 && src.rows >= blockSize / 2);
-            oclMat Dx, Dy;
             CV_Assert(borderType == cv::BORDER_CONSTANT || borderType == cv::BORDER_REFLECT101 || borderType == cv::BORDER_REPLICATE || borderType == cv::BORDER_REFLECT);
-            extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
+            extractCovData(src, dx, dy, blockSize, ksize, borderType);
             dst.create(src.size(), CV_32F);
-            corner_ocl(imgproc_calcMinEigenVal, "calcMinEigenVal", blockSize, 0, Dx, Dy, dst, borderType);
+            corner_ocl(imgproc_calcMinEigenVal, "calcMinEigenVal", blockSize, 0, dx, dy, dst, borderType);
         }
         /////////////////////////////////// MeanShiftfiltering ///////////////////////////////////////////////
         static void meanShiftFiltering_gpu(const oclMat &src, oclMat dst, int sp, int sr, int maxIter, float eps)
@@ -1254,15 +1285,10 @@ namespace cv
         void meanShiftFiltering(const oclMat &src, oclMat &dst, int sp, int sr, TermCriteria criteria)
         {
             if( src.empty() )
-                CV_Error( CV_StsBadArg, "The input image is empty" );
+                CV_Error(Error::StsBadArg, "The input image is empty" );
 
             if( src.depth() != CV_8U || src.oclchannels() != 4 )
-                CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
-
-            //            if(src.clCxt->impl->double_support == 0)
-            //            {
-            //                CV_Error( CV_GpuNotSupported, "Selected device doesn't support double, so a deviation exists.\nIf the accuracy is acceptable, the error can be ignored.\n");
-            //            }
+                CV_Error(Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
 
             dst.create( src.size(), CV_8UC4 );
 
@@ -1322,15 +1348,10 @@ namespace cv
         void meanShiftProc(const oclMat &src, oclMat &dstr, oclMat &dstsp, int sp, int sr, TermCriteria criteria)
         {
             if( src.empty() )
-                CV_Error( CV_StsBadArg, "The input image is empty" );
+                CV_Error(Error::StsBadArg, "The input image is empty" );
 
             if( src.depth() != CV_8U || src.oclchannels() != 4 )
-                CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
-
-            //            if(src.clCxt->impl->double_support == 0)
-            //            {
-            //                CV_Error( CV_GpuNotSupported, "Selected device doesn't support double, so a deviation exists.\nIf the accuracy is acceptable, the error can be ignored.\n");
-            //            }
+                CV_Error(Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
 
             dstr.create( src.size(), CV_8UC4 );
             dstsp.create( src.size(), CV_16SC2 );
@@ -1364,7 +1385,7 @@ namespace cv
             Context  *clCxt = mat_src.clCxt;
             int depth = mat_src.depth();
 
-            std::string kernelName = "calc_sub_hist";
+            String kernelName = "calc_sub_hist";
 
             size_t localThreads[3]  = { HISTOGRAM256_BIN_COUNT, 1, 1 };
             size_t globalThreads[3] = { PARTIAL_HISTOGRAM256_COUNT *localThreads[0], 1, 1};
@@ -1441,7 +1462,7 @@ namespace cv
             using namespace histograms;
 
             Context  *clCxt = sub_hist.clCxt;
-            std::string kernelName = "merge_hist";
+            String kernelName = "merge_hist";
 
             size_t localThreads[3]  = { 256, 1, 1 };
             size_t globalThreads[3] = { HISTOGRAM256_BIN_COUNT *localThreads[0], 1, 1};
@@ -1474,7 +1495,7 @@ namespace cv
             calcHist(mat_src, mat_hist);
 
             Context *clCxt = mat_src.clCxt;
-            std::string kernelName = "calLUT";
+            String kernelName = "calLUT";
             size_t localThreads[3] = { 256, 1, 1};
             size_t globalThreads[3] = { 256, 1, 1};
             oclMat lut(1, 256, CV_8UC1);
@@ -1486,6 +1507,194 @@ namespace cv
             openCLExecuteKernel(clCxt, &imgproc_histogram, kernelName, globalThreads, localThreads, args, -1, -1);
             LUT(mat_src, lut, mat_dst);
         }
+
+        ////////////////////////////////////////////////////////////////////////
+        // CLAHE
+        namespace clahe
+        {
+            inline int divUp(int total, int grain)
+            {
+                return (total + grain - 1) / grain * grain;
+            }
+
+            static void calcLut(const oclMat &src, oclMat &dst,
+                const int tilesX, const int tilesY, const cv::Size tileSize,
+                const int clipLimit, const float lutScale)
+            {
+                cl_int2 tile_size;
+                tile_size.s[0] = tileSize.width;
+                tile_size.s[1] = tileSize.height;
+
+                std::vector<std::pair<size_t , const void *> > args;
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src.data ));
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&dst.data ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&src.step ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst.step ));
+                args.push_back( std::make_pair( sizeof(cl_int2), (void *)&tile_size ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&tilesX ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&clipLimit ));
+                args.push_back( std::make_pair( sizeof(cl_float), (void *)&lutScale ));
+
+                String kernelName = "calcLut";
+                size_t localThreads[3]  = { 32, 8, 1 };
+                size_t globalThreads[3] = { tilesX * localThreads[0], tilesY * localThreads[1], 1 };
+                bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
+                if (is_cpu)
+                {
+                    openCLExecuteKernel(Context::getContext(), &imgproc_clahe, kernelName, globalThreads, localThreads, args, -1, -1, (char*)" -D CPU");
+                }
+                else
+                {
+                    cl_kernel kernel = openCLGetKernelFromSource(Context::getContext(), &imgproc_clahe, kernelName);
+                    int wave_size = queryDeviceInfo<WAVEFRONT_SIZE, int>(kernel);
+                    openCLSafeCall(clReleaseKernel(kernel));
+
+                    static char opt[20] = {0};
+                    sprintf(opt, " -D WAVE_SIZE=%d", wave_size);
+                    openCLExecuteKernel(Context::getContext(), &imgproc_clahe, kernelName, globalThreads, localThreads, args, -1, -1, opt);
+                }
+            }
+
+            static void transform(const oclMat &src, oclMat &dst, const oclMat &lut,
+                const int tilesX, const int tilesY, const cv::Size tileSize)
+            {
+                cl_int2 tile_size;
+                tile_size.s[0] = tileSize.width;
+                tile_size.s[1] = tileSize.height;
+
+                std::vector<std::pair<size_t , const void *> > args;
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src.data ));
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&dst.data ));
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&lut.data ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&src.step ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst.step ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&lut.step ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&src.cols ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&src.rows ));
+                args.push_back( std::make_pair( sizeof(cl_int2), (void *)&tile_size ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&tilesX ));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&tilesY ));
+
+                String kernelName = "transform";
+                size_t localThreads[3]  = { 32, 8, 1 };
+                size_t globalThreads[3] = { divUp(src.cols, localThreads[0]), divUp(src.rows, localThreads[1]), 1 };
+
+                openCLExecuteKernel(Context::getContext(), &imgproc_clahe, kernelName, globalThreads, localThreads, args, -1, -1);
+            }
+        }
+
+        namespace
+        {
+            class CLAHE_Impl : public cv::CLAHE
+            {
+            public:
+                CLAHE_Impl(double clipLimit = 40.0, int tilesX = 8, int tilesY = 8);
+
+                cv::AlgorithmInfo* info() const;
+
+                void apply(cv::InputArray src, cv::OutputArray dst);
+
+                void setClipLimit(double clipLimit);
+                double getClipLimit() const;
+
+                void setTilesGridSize(cv::Size tileGridSize);
+                cv::Size getTilesGridSize() const;
+
+                void collectGarbage();
+
+            private:
+                double clipLimit_;
+                int tilesX_;
+                int tilesY_;
+
+                oclMat srcExt_;
+                oclMat lut_;
+            };
+            CLAHE_Impl::CLAHE_Impl(double clipLimit, int tilesX, int tilesY) :
+            clipLimit_(clipLimit), tilesX_(tilesX), tilesY_(tilesY)
+            {
+            }
+
+            CV_INIT_ALGORITHM(CLAHE_Impl, "CLAHE_OCL",
+                obj.info()->addParam(obj, "clipLimit", obj.clipLimit_);
+                obj.info()->addParam(obj, "tilesX", obj.tilesX_);
+                obj.info()->addParam(obj, "tilesY", obj.tilesY_))
+            void CLAHE_Impl::apply(cv::InputArray src_raw, cv::OutputArray dst_raw)
+            {
+                oclMat& src = getOclMatRef(src_raw);
+                oclMat& dst = getOclMatRef(dst_raw);
+                CV_Assert( src.type() == CV_8UC1 );
+
+                dst.create( src.size(), src.type() );
+
+                const int histSize = 256;
+
+                ensureSizeIsEnough(tilesX_ * tilesY_, histSize, CV_8UC1, lut_);
+
+                cv::Size tileSize;
+                oclMat srcForLut;
+
+                if (src.cols % tilesX_ == 0 && src.rows % tilesY_ == 0)
+                {
+                    tileSize = cv::Size(src.cols / tilesX_, src.rows / tilesY_);
+                    srcForLut = src;
+                }
+                else
+                {
+                    cv::ocl::copyMakeBorder(src, srcExt_, 0, tilesY_ - (src.rows % tilesY_), 0, tilesX_ - (src.cols % tilesX_), cv::BORDER_REFLECT_101, cv::Scalar());
+
+                    tileSize = cv::Size(srcExt_.cols / tilesX_, srcExt_.rows / tilesY_);
+                    srcForLut = srcExt_;
+                }
+
+                const int tileSizeTotal = tileSize.area();
+                const float lutScale = static_cast<float>(histSize - 1) / tileSizeTotal;
+
+                int clipLimit = 0;
+                if (clipLimit_ > 0.0)
+                {
+                    clipLimit = static_cast<int>(clipLimit_ * tileSizeTotal / histSize);
+                    clipLimit = std::max(clipLimit, 1);
+                }
+
+                clahe::calcLut(srcForLut, lut_, tilesX_, tilesY_, tileSize, clipLimit, lutScale);
+                //finish();
+                clahe::transform(src, dst, lut_, tilesX_, tilesY_, tileSize);
+            }
+
+            void CLAHE_Impl::setClipLimit(double clipLimit)
+            {
+                clipLimit_ = clipLimit;
+            }
+
+            double CLAHE_Impl::getClipLimit() const
+            {
+                return clipLimit_;
+            }
+
+            void CLAHE_Impl::setTilesGridSize(cv::Size tileGridSize)
+            {
+                tilesX_ = tileGridSize.width;
+                tilesY_ = tileGridSize.height;
+            }
+
+            cv::Size CLAHE_Impl::getTilesGridSize() const
+            {
+                return cv::Size(tilesX_, tilesY_);
+            }
+
+            void CLAHE_Impl::collectGarbage()
+            {
+                srcExt_.release();
+                lut_.release();
+            }
+        }
+
+        cv::Ptr<cv::CLAHE> createCLAHE(double clipLimit, cv::Size tileGridSize)
+        {
+            return makePtr<CLAHE_Impl>(clipLimit, tileGridSize.width, tileGridSize.height);
+        }
+
         //////////////////////////////////bilateralFilter////////////////////////////////////////////////////
         static void
         oclbilateralFilter_8u( const oclMat &src, oclMat &dst, int d,
@@ -1544,7 +1753,7 @@ namespace cv
             oclMat oclspace_weight(1, d * d, CV_32FC1, space_weight);
             oclMat oclspace_ofs(1, d * d, CV_32SC1, space_ofs);
 
-            std::string kernelName = "bilateral";
+            String kernelName = "bilateral";
             size_t localThreads[3]  = { 16, 16, 1 };
             size_t globalThreads[3] = { (dst.cols + localThreads[0] - 1) / localThreads[0] *localThreads[0],
                                         (dst.rows + localThreads[1] - 1) / localThreads[1] *localThreads[1],
@@ -1579,18 +1788,157 @@ namespace cv
             if( src.depth() == CV_8U )
                 oclbilateralFilter_8u( src, dst, radius, sigmaclr, sigmaspc, borderType );
             else
-                CV_Error( CV_StsUnsupportedFormat,
-                          "Bilateral filtering is only implemented for 8uimages" );
+                CV_Error(Error::StsUnsupportedFormat, "Bilateral filtering is only implemented for 8uimages" );
         }
 
     }
 }
+//////////////////////////////////mulSpectrums////////////////////////////////////////////////////
+void cv::ocl::mulSpectrums(const oclMat &a, const oclMat &b, oclMat &c, int /*flags*/, float scale, bool conjB)
+{
+    CV_Assert(a.type() == CV_32FC2);
+    CV_Assert(b.type() == CV_32FC2);
+
+    c.create(a.size(), CV_32FC2);
+
+    size_t lt[3]  = { 16, 16, 1 };
+    size_t gt[3]  = { a.cols, a.rows, 1 };
+
+    String kernelName = conjB ? "mulAndScaleSpectrumsKernel_CONJ":"mulAndScaleSpectrumsKernel";
+
+    std::vector<std::pair<size_t , const void *> > args;
+    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&a.data ));
+    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&b.data ));
+    args.push_back( std::make_pair( sizeof(cl_float), (void *)&scale));
+    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&c.data ));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&a.cols ));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&a.rows));
+    args.push_back( std::make_pair( sizeof(cl_int), (void *)&a.step ));
+
+    Context *clCxt = Context::getContext();
+    openCLExecuteKernel(clCxt, &imgproc_mulAndScaleSpectrums, kernelName, gt, lt, args, -1, -1);
+}
 //////////////////////////////////convolve////////////////////////////////////////////////////
 inline int divUp(int total, int grain)
 {
     return (total + grain - 1) / grain;
 }
-static void convolve_run(const oclMat &src, const oclMat &temp1, oclMat &dst, std::string kernelName, const char **kernelString)
+
+// ported from CUDA module
+void cv::ocl::ConvolveBuf::create(Size image_size, Size templ_size)
+{
+    result_size = Size(image_size.width - templ_size.width + 1,
+                       image_size.height - templ_size.height + 1);
+
+    block_size = user_block_size;
+    if (user_block_size.width == 0 || user_block_size.height == 0)
+        block_size = estimateBlockSize(result_size, templ_size);
+
+    dft_size.width  = 1 << int(ceil(std::log(block_size.width + templ_size.width - 1.) / std::log(2.)));
+    dft_size.height = 1 << int(ceil(std::log(block_size.height + templ_size.height - 1.) / std::log(2.)));
+
+    // CUFFT has hard-coded kernels for power-of-2 sizes (up to 8192),
+    // see CUDA Toolkit 4.1 CUFFT Library Programming Guide
+    //if (dft_size.width > 8192)
+    dft_size.width = getOptimalDFTSize(block_size.width + templ_size.width - 1.);
+    //if (dft_size.height > 8192)
+    dft_size.height = getOptimalDFTSize(block_size.height + templ_size.height - 1.);
+
+    // To avoid wasting time doing small DFTs
+    dft_size.width = std::max(dft_size.width, 512);
+    dft_size.height = std::max(dft_size.height, 512);
+
+    image_block.create(dft_size, CV_32F);
+    templ_block.create(dft_size, CV_32F);
+    result_data.create(dft_size, CV_32F);
+
+    //spect_len = dft_size.height * (dft_size.width / 2 + 1);
+    image_spect.create(dft_size.height, dft_size.width / 2 + 1, CV_32FC2);
+    templ_spect.create(dft_size.height, dft_size.width / 2 + 1, CV_32FC2);
+    result_spect.create(dft_size.height, dft_size.width / 2 + 1, CV_32FC2);
+
+    // Use maximum result matrix block size for the estimated DFT block size
+    block_size.width = std::min(dft_size.width - templ_size.width + 1, result_size.width);
+    block_size.height = std::min(dft_size.height - templ_size.height + 1, result_size.height);
+}
+
+Size cv::ocl::ConvolveBuf::estimateBlockSize(Size result_size, Size /*templ_size*/)
+{
+    int width = (result_size.width + 2) / 3;
+    int height = (result_size.height + 2) / 3;
+    width = std::min(width, result_size.width);
+    height = std::min(height, result_size.height);
+    return Size(width, height);
+}
+
+static void convolve_run_fft(const oclMat &image, const oclMat &templ, oclMat &result, bool ccorr, ConvolveBuf& buf)
+{
+#if defined HAVE_CLAMDFFT
+    CV_Assert(image.type() == CV_32F);
+    CV_Assert(templ.type() == CV_32F);
+
+    buf.create(image.size(), templ.size());
+    result.create(buf.result_size, CV_32F);
+
+    Size& block_size = buf.block_size;
+    Size& dft_size = buf.dft_size;
+
+    oclMat& image_block = buf.image_block;
+    oclMat& templ_block = buf.templ_block;
+    oclMat& result_data = buf.result_data;
+
+    oclMat& image_spect = buf.image_spect;
+    oclMat& templ_spect = buf.templ_spect;
+    oclMat& result_spect = buf.result_spect;
+
+    oclMat templ_roi = templ;
+    copyMakeBorder(templ_roi, templ_block, 0, templ_block.rows - templ_roi.rows, 0,
+                   templ_block.cols - templ_roi.cols, 0, Scalar());
+
+    cv::ocl::dft(templ_block, templ_spect, dft_size);
+
+    // Process all blocks of the result matrix
+    for (int y = 0; y < result.rows; y += block_size.height)
+    {
+        for (int x = 0; x < result.cols; x += block_size.width)
+        {
+            Size image_roi_size(std::min(x + dft_size.width, image.cols) - x,
+                                std::min(y + dft_size.height, image.rows) - y);
+            Rect roi0(x, y, image_roi_size.width, image_roi_size.height);
+
+            oclMat image_roi(image, roi0);
+
+            copyMakeBorder(image_roi, image_block, 0, image_block.rows - image_roi.rows,
+                           0, image_block.cols - image_roi.cols, 0, Scalar());
+
+            cv::ocl::dft(image_block, image_spect, dft_size);
+
+            mulSpectrums(image_spect, templ_spect, result_spect, 0,
+                                 1.f / dft_size.area(), ccorr);
+
+            cv::ocl::dft(result_spect, result_data, dft_size, cv::DFT_INVERSE | cv::DFT_REAL_OUTPUT);
+
+            Size result_roi_size(std::min(x + block_size.width, result.cols) - x,
+                                 std::min(y + block_size.height, result.rows) - y);
+
+            Rect roi1(x, y, result_roi_size.width, result_roi_size.height);
+            Rect roi2(0, 0, result_roi_size.width, result_roi_size.height);
+
+            oclMat result_roi(result, roi1);
+            oclMat result_block(result_data, roi2);
+
+            result_block.copyTo(result_roi);
+        }
+    }
+
+#else
+    CV_Error(Error::StsNotImplemented, "OpenCL DFT is not implemented");
+#define UNUSED(x) (void)(x);
+    UNUSED(image) UNUSED(templ) UNUSED(result) UNUSED(ccorr) UNUSED(buf)
+#undef UNUSED
+#endif
+}
+static void convolve_run(const oclMat &src, const oclMat &temp1, oclMat &dst, String kernelName, const char **kernelString)
 {
     CV_Assert(src.depth() == CV_32FC1);
     CV_Assert(temp1.depth() == CV_32F);
@@ -1630,13 +1978,25 @@ static void convolve_run(const oclMat &src, const oclMat &temp1, oclMat &dst, st
 
     openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
 }
-void cv::ocl::convolve(const oclMat &x, const oclMat &t, oclMat &y)
+void cv::ocl::convolve(const oclMat &x, const oclMat &t, oclMat &y, bool ccorr)
 {
     CV_Assert(x.depth() == CV_32F);
     CV_Assert(t.depth() == CV_32F);
-    CV_Assert(x.type() == y.type() && x.size() == y.size());
     y.create(x.size(), x.type());
-    std::string kernelName = "convolve";
-
-    convolve_run(x, t, y, kernelName, &imgproc_convolve);
+    String kernelName = "convolve";
+    if(t.cols > 17 || t.rows > 17)
+    {
+        ConvolveBuf buf;
+        convolve_run_fft(x, t, y, ccorr, buf);
+    }
+    else
+    {
+        CV_Assert(ccorr == false);
+        convolve_run(x, t, y, kernelName, &imgproc_convolve);
+    }
+}
+void cv::ocl::convolve(const oclMat &image, const oclMat &templ, oclMat &result, bool ccorr, ConvolveBuf& buf)
+{
+    result.create(image.size(), image.type());
+    convolve_run_fft(image, templ, result, ccorr, buf);
 }
diff --git a/modules/ocl/src/initialization.cpp b/modules/ocl/src/initialization.cpp
index c0796d0add..827b7d495a 100644
--- a/modules/ocl/src/initialization.cpp
+++ b/modules/ocl/src/initialization.cpp
@@ -65,6 +65,7 @@ namespace cv
     namespace ocl
     {
         extern void fft_teardown();
+        extern void clBlasTeardown();
         /*
          * The binary caching system to eliminate redundant program source compilation.
          * Strictly, this is not a cache because we do not implement evictions right now.
@@ -92,9 +93,9 @@ namespace cv
             releaseProgram();
         }
 
-        cl_program ProgramCache::progLookup(std::string srcsign)
+        cl_program ProgramCache::progLookup(String srcsign)
         {
-            std::map<std::string, cl_program>::iterator iter;
+            std::map<String, cl_program>::iterator iter;
             iter = codeCache.find(srcsign);
             if(iter != codeCache.end())
                 return iter->second;
@@ -102,17 +103,17 @@ namespace cv
                 return NULL;
         }
 
-        void ProgramCache::addProgram(std::string srcsign , cl_program program)
+        void ProgramCache::addProgram(String srcsign , cl_program program)
         {
             if(!progLookup(srcsign))
             {
-                codeCache.insert(std::map<std::string, cl_program>::value_type(srcsign, program));
+                codeCache.insert(std::map<String, cl_program>::value_type(srcsign, program));
             }
         }
 
         void ProgramCache::releaseProgram()
         {
-            std::map<std::string, cl_program>::iterator iter;
+            std::map<String, cl_program>::iterator iter;
             for(iter = codeCache.begin(); iter != codeCache.end(); iter++)
             {
                 openCLSafeCall(clReleaseProgram(iter->second));
@@ -120,6 +121,148 @@ namespace cv
             codeCache.clear();
             cacheSize = 0;
         }
+        struct Info::Impl
+        {
+            cl_platform_id oclplatform;
+            std::vector<cl_device_id> devices;
+            std::vector<String> devName;
+            String platName;
+            String clVersion;
+
+            cl_context oclcontext;
+            cl_command_queue clCmdQueue;
+            int devnum;
+            size_t maxWorkGroupSize;
+            cl_uint maxDimensions; // == maxWorkItemSizes.size()
+            std::vector<size_t> maxWorkItemSizes;
+            cl_uint maxComputeUnits;
+            char extra_options[512];
+            int  double_support;
+            int unified_memory; //1 means integrated GPU, otherwise this value is 0
+            int refcounter;
+
+            Impl();
+
+            void setDevice(void *ctx, void *q, int devnum);
+
+            void release()
+            {
+                if(1 == CV_XADD(&refcounter, -1))
+                {
+                    releaseResources();
+                    delete this;
+                }
+            }
+
+            Impl* copy()
+            {
+                CV_XADD(&refcounter, 1);
+                return this;
+            }
+
+        private:
+            Impl(const Impl&);
+            Impl& operator=(const Impl&);
+            void releaseResources();
+        };
+
+        // global variables to hold binary cache properties
+        static int enable_disk_cache =
+#ifdef _DEBUG
+            false;
+#else
+            true;
+#endif
+        static int update_disk_cache = false;
+        static String binpath = "";
+
+        Info::Impl::Impl()
+            :oclplatform(0),
+            oclcontext(0),
+            clCmdQueue(0),
+            devnum(-1),
+            maxWorkGroupSize(0),
+            maxDimensions(0),
+            maxComputeUnits(0),
+            double_support(0),
+            unified_memory(0),
+            refcounter(1)
+        {
+            memset(extra_options, 0, 512);
+        }
+
+        void Info::Impl::releaseResources()
+        {
+            devnum = -1;
+
+            if(clCmdQueue)
+            {
+                //temporarily disable command queue release as it causes program hang at exit
+                //openCLSafeCall(clReleaseCommandQueue(clCmdQueue));
+                clCmdQueue = 0;
+            }
+
+            if(oclcontext)
+            {
+                openCLSafeCall(clReleaseContext(oclcontext));
+                oclcontext = 0;
+            }
+        }
+
+        void Info::Impl::setDevice(void *ctx, void *q, int dnum)
+        {
+            if((ctx && q) || devnum != dnum)
+                releaseResources();
+
+            CV_Assert(dnum >= 0 && dnum < (int)devices.size());
+            devnum = dnum;
+            if(ctx && q)
+            {
+                oclcontext = (cl_context)ctx;
+                clCmdQueue = (cl_command_queue)q;
+                clRetainContext(oclcontext);
+                clRetainCommandQueue(clCmdQueue);
+            }
+            else
+            {
+                cl_int status = 0;
+                cl_context_properties cps[3] = { CL_CONTEXT_PLATFORM, (cl_context_properties)(oclplatform), 0 };
+                oclcontext = clCreateContext(cps, 1, &devices[devnum], 0, 0, &status);
+                openCLVerifyCall(status);
+                clCmdQueue = clCreateCommandQueue(oclcontext, devices[devnum], CL_QUEUE_PROFILING_ENABLE, &status);
+                openCLVerifyCall(status);
+            }
+
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_MAX_WORK_GROUP_SIZE, sizeof(size_t), (void *)&maxWorkGroupSize, 0));
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS, sizeof(cl_uint), (void *)&maxDimensions, 0));
+            maxWorkItemSizes.resize(maxDimensions);
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_MAX_WORK_ITEM_SIZES, sizeof(size_t)*maxDimensions, (void *)&maxWorkItemSizes[0], 0));
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_MAX_COMPUTE_UNITS, sizeof(cl_uint), (void *)&maxComputeUnits, 0));
+
+            cl_bool unfymem = false;
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_HOST_UNIFIED_MEMORY, sizeof(cl_bool), (void *)&unfymem, 0));
+            unified_memory = unfymem ? 1 : 0;
+
+            //initialize extra options for compilation. Currently only fp64 is included.
+            //Assume 4KB is enough to store all possible extensions.
+            const int EXT_LEN = 4096 + 1 ;
+            char extends_set[EXT_LEN];
+            size_t extends_size;
+            openCLSafeCall(clGetDeviceInfo(devices[devnum], CL_DEVICE_EXTENSIONS, EXT_LEN, (void *)extends_set, &extends_size));
+            extends_set[EXT_LEN - 1] = 0;
+            size_t fp64_khr = String(extends_set).find("cl_khr_fp64");
+
+            if(fp64_khr != String::npos)
+            {
+                sprintf(extra_options, "-D DOUBLE_SUPPORT");
+                double_support = 1;
+            }
+            else
+            {
+                memset(extra_options, 0, 512);
+                double_support = 0;
+            }
+        }
 
         ////////////////////////Common OpenCL specific calls///////////////
         int getDevMemType(DevMemRW& rw_type, DevMemType& mem_type)
@@ -131,36 +274,14 @@ namespace cv
 
         int setDevMemType(DevMemRW rw_type, DevMemType mem_type)
         {
-            if( (mem_type == DEVICE_MEM_PM && Context::getContext()->impl->unified_memory == 0) ||
-                 mem_type == DEVICE_MEM_UHP ||
-                 mem_type == DEVICE_MEM_CHP )
+            if( (mem_type == DEVICE_MEM_PM &&
+                 Context::getContext()->impl->unified_memory == 0) )
                 return -1;
             gDeviceMemRW = rw_type;
             gDeviceMemType = mem_type;
             return 0;
         }
 
-       struct Info::Impl
-        {
-            cl_platform_id oclplatform;
-            std::vector<cl_device_id> devices;
-            std::vector<std::string> devName;
-
-            cl_context oclcontext;
-            cl_command_queue clCmdQueue;
-            int devnum;
-            cl_uint maxDimensions;
-            size_t maxWorkGroupSize;
-            size_t *maxWorkItemSizes;
-            cl_uint maxComputeUnits;
-            char extra_options[512];
-            int  double_support;
-            Impl()
-            {
-                memset(extra_options, 0, 512);
-            }
-        };
-
         inline int divUp(int total, int grain)
         {
             return (total + grain - 1) / grain;
@@ -168,6 +289,9 @@ namespace cv
 
         int getDevice(std::vector<Info> &oclinfo, int devicetype)
         {
+            //TODO: cache oclinfo vector
+            oclinfo.clear();
+
             switch(devicetype)
             {
             case CVCL_DEVICE_TYPE_DEFAULT:
@@ -177,46 +301,53 @@ namespace cv
             case CVCL_DEVICE_TYPE_ALL:
                 break;
             default:
-                CV_Error(CV_GpuApiCallError, "Unkown device type");
+                return 0;
             }
-            int devcienums = 0;
-            // Platform info
-            cl_int status = 0;
-            cl_uint numPlatforms;
-            Info ocltmpinfo;
-            openCLSafeCall(clGetPlatformIDs(0, NULL, &numPlatforms));
-            CV_Assert(numPlatforms > 0);
-            cl_platform_id *platforms = new cl_platform_id[numPlatforms];
 
-            openCLSafeCall(clGetPlatformIDs(numPlatforms, platforms, NULL));
-            char deviceName[256];
+            // Platform info
+            cl_uint numPlatforms;
+            openCLSafeCall(clGetPlatformIDs(0, 0, &numPlatforms));
+            if(numPlatforms < 1) return 0;
+
+            std::vector<cl_platform_id> platforms(numPlatforms);
+            openCLSafeCall(clGetPlatformIDs(numPlatforms, &platforms[0], 0));
+
+            int devcienums = 0;
+
+            const static int max_name_length = 256;
+            char deviceName[max_name_length];
+            char plfmName[max_name_length];
+            char clVersion[256];
             for (unsigned i = 0; i < numPlatforms; ++i)
             {
-                cl_uint numsdev;
-                status = clGetDeviceIDs(platforms[i], devicetype, 0, NULL, &numsdev);
+                cl_uint numsdev = 0;
+                cl_int status = clGetDeviceIDs(platforms[i], devicetype, 0, NULL, &numsdev);
                 if(status != CL_DEVICE_NOT_FOUND)
-                {
                     openCLVerifyCall(status);
-                }
+
                 if(numsdev > 0)
                 {
                     devcienums += numsdev;
-                    cl_device_id *devices = new cl_device_id[numsdev];
-                    openCLSafeCall(clGetDeviceIDs(platforms[i], devicetype, numsdev, devices, NULL));
+                    std::vector<cl_device_id> devices(numsdev);
+                    openCLSafeCall(clGetDeviceIDs(platforms[i], devicetype, numsdev, &devices[0], 0));
+
+                    Info ocltmpinfo;
+                    openCLSafeCall(clGetPlatformInfo(platforms[i], CL_PLATFORM_NAME, sizeof(plfmName), plfmName, NULL));
+                    ocltmpinfo.PlatformName = String(plfmName);
+                    ocltmpinfo.impl->platName = String(plfmName);
                     ocltmpinfo.impl->oclplatform = platforms[i];
-                    for(unsigned j = 0; j < numsdev; j++)
+                    openCLSafeCall(clGetPlatformInfo(platforms[i], CL_PLATFORM_VERSION, sizeof(clVersion), clVersion, NULL));
+                    ocltmpinfo.impl->clVersion = clVersion;
+                    for(unsigned j = 0; j < numsdev; ++j)
                     {
                         ocltmpinfo.impl->devices.push_back(devices[j]);
-                        openCLSafeCall(clGetDeviceInfo(devices[j], CL_DEVICE_NAME, 256, deviceName, NULL));
-                        ocltmpinfo.impl->devName.push_back(std::string(deviceName));
-                        ocltmpinfo.DeviceName.push_back(std::string(deviceName));
+                        openCLSafeCall(clGetDeviceInfo(devices[j], CL_DEVICE_NAME, sizeof(deviceName), deviceName, 0));
+                        ocltmpinfo.impl->devName.push_back(deviceName);
+                        ocltmpinfo.DeviceName.push_back(deviceName);
                     }
-                    delete[] devices;
                     oclinfo.push_back(ocltmpinfo);
-                    ocltmpinfo.release();
                 }
             }
-            delete[] platforms;
             if(devcienums > 0)
             {
                 setDevice(oclinfo[0]);
@@ -224,84 +355,72 @@ namespace cv
             return devcienums;
         }
 
-        static void fillClcontext(Info &oclinfo)
-        {
-            //get device information
-            size_t devnum = oclinfo.impl->devnum;
-
-            openCLSafeCall(clGetDeviceInfo(oclinfo.impl->devices[devnum], CL_DEVICE_MAX_WORK_GROUP_SIZE,
-                                           sizeof(size_t), (void *)&oclinfo.impl->maxWorkGroupSize, NULL));
-            openCLSafeCall(clGetDeviceInfo(oclinfo.impl->devices[devnum], CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS,
-                                           sizeof(cl_uint), (void *)&oclinfo.impl->maxDimensions, NULL));
-            oclinfo.impl->maxWorkItemSizes = new size_t[oclinfo.impl->maxDimensions];
-            openCLSafeCall(clGetDeviceInfo(oclinfo.impl->devices[devnum], CL_DEVICE_MAX_WORK_ITEM_SIZES,
-                                           sizeof(size_t)*oclinfo.impl->maxDimensions, (void *)oclinfo.impl->maxWorkItemSizes, NULL));
-            openCLSafeCall(clGetDeviceInfo(oclinfo.impl->devices[devnum], CL_DEVICE_MAX_COMPUTE_UNITS,
-                                           sizeof(cl_uint), (void *)&oclinfo.impl->maxComputeUnits, NULL));
-            //initialize extra options for compilation. Currently only fp64 is included.
-            //Assume 4KB is enough to store all possible extensions.
-
-            const int EXT_LEN = 4096 + 1 ;
-            char extends_set[EXT_LEN];
-            size_t extends_size;
-            openCLSafeCall(clGetDeviceInfo(oclinfo.impl->devices[devnum], CL_DEVICE_EXTENSIONS,
-                                           EXT_LEN, (void *)extends_set, &extends_size));
-            CV_Assert(extends_size < (size_t)EXT_LEN);
-            extends_set[EXT_LEN - 1] = 0;
-            memset(oclinfo.impl->extra_options, 0, 512);
-            oclinfo.impl->double_support = 0;
-            int fp64_khr = std::string(extends_set).find("cl_khr_fp64");
-
-            if(fp64_khr >= 0 && fp64_khr < EXT_LEN)
-            {
-                sprintf(oclinfo.impl->extra_options , "-D DOUBLE_SUPPORT");
-                oclinfo.impl -> double_support = 1;
-            }
-            Context::setContext(oclinfo);
-
-        }
-
         void setDevice(Info &oclinfo, int devnum)
         {
-            CV_Assert(devnum >= 0);
-            cl_int status = 0;
-            cl_context_properties cps[3] =
-            {
-                CL_CONTEXT_PLATFORM, (cl_context_properties)(oclinfo.impl->oclplatform), 0
-            };
-            oclinfo.impl->devnum = devnum;
-            oclinfo.impl->oclcontext = clCreateContext(cps, 1, &oclinfo.impl->devices[devnum], NULL, NULL, &status);
-            openCLVerifyCall(status);
-            //create the command queue using the first device of the list
-            oclinfo.impl->clCmdQueue = clCreateCommandQueue(oclinfo.impl->oclcontext, oclinfo.impl->devices[devnum],
-                                       CL_QUEUE_PROFILING_ENABLE, &status);
-            openCLVerifyCall(status);
-            fillClcontext(oclinfo);
+            oclinfo.impl->setDevice(0, 0, devnum);
+            Context::setContext(oclinfo);
         }
 
         void setDeviceEx(Info &oclinfo, void *ctx, void *q, int devnum)
         {
-            CV_Assert(devnum >= 0);
-            oclinfo.impl->devnum = devnum;
-            if(ctx && q)
-            {
-                oclinfo.impl->oclcontext = (cl_context)ctx;
-                oclinfo.impl->clCmdQueue = (cl_command_queue)q;
-                clRetainContext((cl_context)ctx);
-                clRetainCommandQueue((cl_command_queue)q);
-                fillClcontext(oclinfo);
-             }
+            oclinfo.impl->setDevice(ctx, q, devnum);
+            Context::setContext(oclinfo);
          }
 
         void *getoclContext()
         {
-            return &(Context::getContext()->impl->clContext);
+            return &(Context::getContext()->impl->oclcontext);
         }
 
         void *getoclCommandQueue()
         {
             return &(Context::getContext()->impl->clCmdQueue);
         }
+
+        void finish()
+        {
+            clFinish(Context::getContext()->impl->clCmdQueue);
+        }
+
+        //template specializations of queryDeviceInfo
+        template<>
+        bool queryDeviceInfo<IS_CPU_DEVICE, bool>(cl_kernel)
+        {
+            Info::Impl* impl = Context::getContext()->impl;
+            cl_device_type devicetype;
+            openCLSafeCall(clGetDeviceInfo(impl->devices[impl->devnum],
+                CL_DEVICE_TYPE, sizeof(cl_device_type),
+                &devicetype, NULL));
+            return (devicetype == CVCL_DEVICE_TYPE_CPU);
+        }
+
+        template<typename _ty>
+        static _ty queryWavesize(cl_kernel kernel)
+        {
+            size_t info = 0;
+            Info::Impl* impl = Context::getContext()->impl;
+            bool is_cpu = queryDeviceInfo<IS_CPU_DEVICE, bool>();
+            if(is_cpu)
+            {
+                return 1;
+            }
+            CV_Assert(kernel != NULL);
+            openCLSafeCall(clGetKernelWorkGroupInfo(kernel, impl->devices[impl->devnum],
+                CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE, sizeof(size_t), &info, NULL));
+            return static_cast<_ty>(info);
+        }
+
+        template<>
+        size_t queryDeviceInfo<WAVEFRONT_SIZE, size_t>(cl_kernel kernel)
+        {
+            return queryWavesize<size_t>(kernel);
+        }
+        template<>
+        int queryDeviceInfo<WAVEFRONT_SIZE, int>(cl_kernel kernel)
+        {
+            return queryWavesize<int>(kernel);
+        }
+
         void openCLReadBuffer(Context *clCxt, cl_mem dst_buffer, void *host_buffer, size_t size)
         {
             cl_int status;
@@ -313,31 +432,30 @@ namespace cv
         cl_mem openCLCreateBuffer(Context *clCxt, size_t flag , size_t size)
         {
             cl_int status;
-            cl_mem buffer = clCreateBuffer(clCxt->impl->clContext, (cl_mem_flags)flag, size, NULL, &status);
+            cl_mem buffer = clCreateBuffer(clCxt->impl->oclcontext, (cl_mem_flags)flag, size, NULL, &status);
             openCLVerifyCall(status);
             return buffer;
         }
 
         void openCLMallocPitch(Context *clCxt, void **dev_ptr, size_t *pitch,
-                               size_t widthInBytes, size_t height)
-        {
-            openCLMallocPitchEx(clCxt, dev_ptr, pitch, widthInBytes, height, gDeviceMemRW, gDeviceMemType);
-        }
-
-        void openCLMallocPitchEx(Context *clCxt, void **dev_ptr, size_t *pitch,
-                               size_t widthInBytes, size_t height, DevMemRW rw_type, DevMemType mem_type)
+                                 size_t widthInBytes, size_t height,
+                                 DevMemRW rw_type, DevMemType mem_type, void* hptr)
         {
             cl_int status;
-
-            *dev_ptr = clCreateBuffer(clCxt->impl->clContext, gDevMemRWValueMap[rw_type]|gDevMemTypeValueMap[mem_type],
-                                      widthInBytes * height, 0, &status);
+            if(hptr && (mem_type==DEVICE_MEM_UHP || mem_type==DEVICE_MEM_CHP))
+                *dev_ptr = clCreateBuffer(clCxt->impl->oclcontext,
+                                          gDevMemRWValueMap[rw_type]|gDevMemTypeValueMap[mem_type],
+                                          widthInBytes * height, hptr, &status);
+            else
+                *dev_ptr = clCreateBuffer(clCxt->impl->oclcontext, gDevMemRWValueMap[rw_type]|gDevMemTypeValueMap[mem_type],
+                                          widthInBytes * height, 0, &status);
             openCLVerifyCall(status);
             *pitch = widthInBytes;
         }
 
         void openCLMemcpy2D(Context *clCxt, void *dst, size_t dpitch,
                             const void *src, size_t spitch,
-                            size_t width, size_t height, enum openCLMemcpyKind kind, int channels)
+                            size_t width, size_t height, openCLMemcpyKind kind, int channels)
         {
             size_t buffer_origin[3] = {0, 0, 0};
             size_t host_origin[3] = {0, 0, 0};
@@ -386,16 +504,35 @@ namespace cv
         {
             openCLSafeCall(clReleaseMemObject((cl_mem)devPtr));
         }
-        cl_kernel openCLGetKernelFromSource(const Context *clCxt, const char **source, std::string kernelName)
+        cl_kernel openCLGetKernelFromSource(const Context *clCxt, const char **source, String kernelName)
         {
             return openCLGetKernelFromSource(clCxt, source, kernelName, NULL);
         }
 
+        void setBinaryDiskCache(int mode, String path)
+        {
+            if(mode == CACHE_NONE)
+            {
+                update_disk_cache = 0;
+                enable_disk_cache = 0;
+                return;
+            }
+            update_disk_cache |= (mode & CACHE_UPDATE) == CACHE_UPDATE;
+            enable_disk_cache |=
+#ifdef _DEBUG
+                (mode & CACHE_DEBUG)   == CACHE_DEBUG;
+#else
+                (mode & CACHE_RELEASE) == CACHE_RELEASE;
+#endif
+            if(enable_disk_cache && !path.empty())
+            {
+                binpath = path;
+            }
+        }
 
         void setBinpath(const char *path)
         {
-            Context *clcxt = Context::getContext();
-            clcxt->impl->Binpath = path;
+            binpath = path;
         }
 
         int savetofile(const Context*,  cl_program &program, const char *fileName)
@@ -408,7 +545,7 @@ namespace cv
             char* binary = (char*)malloc(binarySize);
             if(binary == NULL)
             {
-                CV_Error(CV_StsNoMem, "Failed to allocate host memory.");
+                CV_Error(Error::StsNoMem, "Failed to allocate host memory.");
             }
             openCLSafeCall(clGetProgramInfo(program,
                                     CL_PROGRAM_BINARIES,
@@ -426,26 +563,26 @@ namespace cv
             return 1;
         }
 
-        cl_kernel openCLGetKernelFromSource(const Context *clCxt, const char **source, std::string kernelName,
+        cl_kernel openCLGetKernelFromSource(const Context *clCxt, const char **source, String kernelName,
                                             const char *build_options)
         {
             cl_kernel kernel;
             cl_program program ;
             cl_int status = 0;
             std::stringstream src_sign;
-            std::string srcsign;
-            std::string filename;
+            String srcsign;
+            String filename;
             CV_Assert(programCache != NULL);
 
             if(NULL != build_options)
             {
-                src_sign << (int64)(*source) << clCxt->impl->clContext << "_" << build_options;
+                src_sign << (int64)(*source) << clCxt->impl->oclcontext << "_" << build_options;
             }
             else
             {
-                src_sign << (int64)(*source) << clCxt->impl->clContext;
+                src_sign << (int64)(*source) << clCxt->impl->oclcontext;
             }
-            srcsign = src_sign.str();
+            srcsign = src_sign.str().c_str();
 
             program = NULL;
             program = programCache->progLookup(srcsign);
@@ -463,24 +600,24 @@ namespace cv
                     strcat(all_build_options, build_options);
                 if(all_build_options != NULL)
                 {
-                    filename = clCxt->impl->Binpath  + kernelName + "_" + clCxt->impl->devName + all_build_options + ".clb";
+                    filename = binpath + kernelName + "_" + clCxt->impl->devName[clCxt->impl->devnum] + all_build_options + ".clb";
                 }
                 else
                 {
-                    filename = clCxt->impl->Binpath  + kernelName + "_" + clCxt->impl->devName + ".clb";
+                    filename = binpath + kernelName + "_" + clCxt->impl->devName[clCxt->impl->devnum] + ".clb";
                 }
 
-                FILE *fp = fopen(filename.c_str(), "rb");
-                if(fp == NULL || clCxt->impl->Binpath.size() == 0)    //we should generate a binary file for the first time.
+                FILE *fp = enable_disk_cache ? fopen(filename.c_str(), "rb") : NULL;
+                if(fp == NULL || update_disk_cache)
                 {
                     if(fp != NULL)
                         fclose(fp);
 
                     program = clCreateProgramWithSource(
-                                  clCxt->impl->clContext, 1, source, NULL, &status);
+                                  clCxt->impl->oclcontext, 1, source, NULL, &status);
                     openCLVerifyCall(status);
-                    status = clBuildProgram(program, 1, &(clCxt->impl->devices), all_build_options, NULL, NULL);
-                    if(status == CL_SUCCESS && clCxt->impl->Binpath.size())
+                    status = clBuildProgram(program, 1, &(clCxt->impl->devices[clCxt->impl->devnum]), all_build_options, NULL, NULL);
+                    if(status == CL_SUCCESS && enable_disk_cache)
                         savetofile(clCxt, program, filename.c_str());
                 }
                 else
@@ -492,15 +629,15 @@ namespace cv
                     CV_Assert(1 == fread(binary, binarySize, 1, fp));
                     fclose(fp);
                     cl_int status = 0;
-                    program = clCreateProgramWithBinary(clCxt->impl->clContext,
+                    program = clCreateProgramWithBinary(clCxt->impl->oclcontext,
                                                         1,
-                                                        &(clCxt->impl->devices),
+                                                        &(clCxt->impl->devices[clCxt->impl->devnum]),
                                                         (const size_t *)&binarySize,
                                                         (const unsigned char **)&binary,
                                                         NULL,
                                                         &status);
                     openCLVerifyCall(status);
-                    status = clBuildProgram(program, 1, &(clCxt->impl->devices), all_build_options, NULL, NULL);
+                    status = clBuildProgram(program, 1, &(clCxt->impl->devices[clCxt->impl->devnum]), all_build_options, NULL, NULL);
                     delete[] binary;
                 }
 
@@ -512,14 +649,14 @@ namespace cv
                         char *buildLog = NULL;
                         size_t buildLogSize = 0;
                         logStatus = clGetProgramBuildInfo(program,
-                                                          clCxt->impl->devices, CL_PROGRAM_BUILD_LOG, buildLogSize,
+                                                          clCxt->impl->devices[clCxt->impl->devnum], CL_PROGRAM_BUILD_LOG, buildLogSize,
                                                           buildLog, &buildLogSize);
                         if(logStatus != CL_SUCCESS)
                             std::cout << "Failed to build the program and get the build info." << std::endl;
                         buildLog = new char[buildLogSize];
                         CV_DbgAssert(!!buildLog);
                         memset(buildLog, 0, buildLogSize);
-                        openCLSafeCall(clGetProgramBuildInfo(program, clCxt->impl->devices,
+                        openCLSafeCall(clGetProgramBuildInfo(program, clCxt->impl->devices[clCxt->impl->devnum],
                                                              CL_PROGRAM_BUILD_LOG, buildLogSize, buildLog, NULL));
                         std::cout << "\n\t\t\tBUILD LOG\n";
                         std::cout << buildLog << std::endl;
@@ -541,20 +678,20 @@ namespace cv
         void openCLVerifyKernel(const Context *clCxt, cl_kernel kernel, size_t *localThreads)
         {
             size_t kernelWorkGroupSize;
-            openCLSafeCall(clGetKernelWorkGroupInfo(kernel, clCxt->impl->devices,
+            openCLSafeCall(clGetKernelWorkGroupInfo(kernel, clCxt->impl->devices[clCxt->impl->devnum],
                                                     CL_KERNEL_WORK_GROUP_SIZE, sizeof(size_t), &kernelWorkGroupSize, 0));
-            CV_Assert( (localThreads[0] <= clCxt->impl->maxWorkItemSizes[0]) &&
-                          (localThreads[1] <= clCxt->impl->maxWorkItemSizes[1]) &&
-                          (localThreads[2] <= clCxt->impl->maxWorkItemSizes[2]) &&
-                          ((localThreads[0] * localThreads[1] * localThreads[2]) <= kernelWorkGroupSize) &&
-                          (localThreads[0] * localThreads[1] * localThreads[2]) <= clCxt->impl->maxWorkGroupSize);
+            CV_Assert( localThreads[0] <= clCxt->impl->maxWorkItemSizes[0] );
+            CV_Assert( localThreads[1] <= clCxt->impl->maxWorkItemSizes[1] );
+            CV_Assert( localThreads[2] <= clCxt->impl->maxWorkItemSizes[2] );
+            CV_Assert( localThreads[0] * localThreads[1] * localThreads[2] <= kernelWorkGroupSize );
+            CV_Assert( localThreads[0] * localThreads[1] * localThreads[2] <= clCxt->impl->maxWorkGroupSize );
         }
 
 #ifdef PRINT_KERNEL_RUN_TIME
         static double total_execute_time = 0;
         static double total_kernel_time = 0;
 #endif
-        void openCLExecuteKernel_(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
+        void openCLExecuteKernel_(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
                                   size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
                                   int depth, const char *build_options)
         {
@@ -566,7 +703,7 @@ namespace cv
                 idxStr << "_C" << channels;
             if(depth != -1)
                 idxStr << "_D" << depth;
-            kernelName += idxStr.str();
+            kernelName = kernelName + idxStr.str().c_str();
 
             cl_kernel kernel;
             kernel = openCLGetKernelFromSource(clCxt, source, kernelName, build_options);
@@ -617,18 +754,18 @@ namespace cv
             clReleaseEvent(event);
 #endif
 
-            clFinish(clCxt->impl->clCmdQueue);
+            clFlush(clCxt->impl->clCmdQueue);
             openCLSafeCall(clReleaseKernel(kernel));
         }
 
-        void openCLExecuteKernel(Context *clCxt , const char **source, std::string kernelName,
+        void openCLExecuteKernel(Context *clCxt , const char **source, String kernelName,
                                  size_t globalThreads[3], size_t localThreads[3],
                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth)
         {
             openCLExecuteKernel(clCxt, source, kernelName, globalThreads, localThreads, args,
                                 channels, depth, NULL);
         }
-        void openCLExecuteKernel(Context *clCxt , const char **source, std::string kernelName,
+        void openCLExecuteKernel(Context *clCxt , const char **source, String kernelName,
                                  size_t globalThreads[3], size_t localThreads[3],
                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options)
 
@@ -637,7 +774,7 @@ namespace cv
             openCLExecuteKernel_(clCxt, source, kernelName, globalThreads, localThreads, args, channels, depth,
                                  build_options);
 #else
-            std::string data_type[] = { "uchar", "char", "ushort", "short", "int", "float", "double"};
+            String data_type[] = { "uchar", "char", "ushort", "short", "int", "float", "double"};
             std::cout << std::endl;
             std::cout << "Function Name: " << kernelName;
             if(depth >= 0)
@@ -662,7 +799,7 @@ namespace cv
 #endif
         }
 
-       double openCLExecuteKernelInterop(Context *clCxt , const char **source, std::string kernelName,
+       double openCLExecuteKernelInterop(Context *clCxt , const char **source, String kernelName,
                                  size_t globalThreads[3], size_t localThreads[3],
                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options,
                                  bool finish, bool measureKernelTime, bool cleanUp)
@@ -676,7 +813,7 @@ namespace cv
                 idxStr << "_C" << channels;
             if(depth != -1)
                 idxStr << "_D" << depth;
-            kernelName += idxStr.str();
+            kernelName = kernelName + idxStr.str().c_str();
 
             cl_kernel kernel;
             kernel = openCLGetKernelFromSource(clCxt, source, kernelName, build_options);
@@ -738,7 +875,7 @@ namespace cv
         }
 
         // Converts the contents of a file into a string
-        static int convertToString(const char *filename, std::string& s)
+        static int convertToString(const char *filename, String& s)
         {
             size_t size;
             char*  str;
@@ -770,16 +907,16 @@ namespace cv
             return -1;
         }
 
-        double openCLExecuteKernelInterop(Context *clCxt , const char **fileName, const int numFiles, std::string kernelName,
+        double openCLExecuteKernelInterop(Context *clCxt , const char **fileName, const int numFiles, String kernelName,
                                  size_t globalThreads[3], size_t localThreads[3],
                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options,
                                  bool finish, bool measureKernelTime, bool cleanUp)
 
         {
-            std::vector<std::string> fsource;
+            std::vector<String> fsource;
             for (int i = 0 ; i < numFiles ; i++)
             {
-                std::string str;
+                String str;
                 if (convertToString(fileName[i], str) >= 0)
                     fsource.push_back(str);
             }
@@ -812,141 +949,144 @@ namespace cv
         /////////////////////////////OpenCL initialization/////////////////
         std::auto_ptr<Context> Context::clCxt;
         int Context::val = 0;
-        Mutex cs;
-        Context *Context::getContext()
+        static Mutex cs;
+        static volatile int context_tear_down = 0;
+
+        bool initialized()
         {
-            if(val == 0)
+            return *((volatile int*)&Context::val) != 0 &&
+                Context::clCxt->impl->clCmdQueue != NULL&&
+                Context::clCxt->impl->oclcontext != NULL;
+        }
+
+        Context* Context::getContext()
+        {
+            if(*((volatile int*)&val) != 1)
             {
                 AutoLock al(cs);
-                if( NULL == clCxt.get())
+                if(*((volatile int*)&val) != 1)
+                {
+                    if (context_tear_down)
+                        return clCxt.get();
+                    if( 0 == clCxt.get())
+                    clCxt.reset(new Context);
+                    std::vector<Info> oclinfo;
+                    CV_Assert(getDevice(oclinfo, CVCL_DEVICE_TYPE_ALL) > 0);
+
+                    *((volatile int*)&val) = 1;
+                }
+            }
+                return clCxt.get();
+            }
+
+        void Context::setContext(Info &oclinfo)
+        {
+            AutoLock guard(cs);
+            if(*((volatile int*)&val) != 1)
+            {
+                if( 0 == clCxt.get())
                     clCxt.reset(new Context);
 
-                val = 1;
-                return clCxt.get();
+                clCxt.get()->impl = oclinfo.impl->copy();
+
+                *((volatile int*)&val) = 1;
             }
             else
             {
-                return clCxt.get();
+                clCxt.get()->impl->release();
+                clCxt.get()->impl = oclinfo.impl->copy();
             }
         }
-        void Context::setContext(Info &oclinfo)
-        {
-            Context *clcxt = getContext();
-            clcxt->impl->clContext = oclinfo.impl->oclcontext;
-            clcxt->impl->clCmdQueue = oclinfo.impl->clCmdQueue;
-            clcxt->impl->devices = oclinfo.impl->devices[oclinfo.impl->devnum];
-            clcxt->impl->devName = oclinfo.impl->devName[oclinfo.impl->devnum];
-            clcxt->impl->maxDimensions = oclinfo.impl->maxDimensions;
-            clcxt->impl->maxWorkGroupSize = oclinfo.impl->maxWorkGroupSize;
-            for(size_t i=0; i<clcxt->impl->maxDimensions && i<4; i++)
-                clcxt->impl->maxWorkItemSizes[i] = oclinfo.impl->maxWorkItemSizes[i];
-            clcxt->impl->maxComputeUnits = oclinfo.impl->maxComputeUnits;
-            clcxt->impl->double_support = oclinfo.impl->double_support;
-            //extra options to recognize compiler options
-            memcpy(clcxt->impl->extra_options, oclinfo.impl->extra_options, 512);
-            cl_bool unfymem = false;
-            openCLSafeCall(clGetDeviceInfo(clcxt->impl->devices, CL_DEVICE_HOST_UNIFIED_MEMORY,
-                                           sizeof(cl_bool), (void *)&unfymem, NULL));
-            if(unfymem)
-                clcxt->impl->unified_memory = 1;
-        }
+
         Context::Context()
         {
-            impl = new Impl;
-            //Information of the OpenCL context
-            impl->clContext = NULL;
-            impl->clCmdQueue = NULL;
-            impl->devices = NULL;
-            impl->maxDimensions = 0;
-            impl->maxWorkGroupSize = 0;
-            for(int i=0; i<4; i++)
-                impl->maxWorkItemSizes[i] = 0;
-            impl->maxComputeUnits = 0;
-            impl->double_support = 0;
-            //extra options to recognize vendor specific fp64 extensions
-            memset(impl->extra_options, 0, 512);
-            impl->unified_memory = 0;
+            impl = 0;
             programCache = ProgramCache::getProgramCache();
         }
 
         Context::~Context()
         {
-            delete impl;
+            release();
+        }
+
+        void Context::release()
+        {
+            if (impl)
+                impl->release();
             programCache->releaseProgram();
         }
+
+        bool Context::supportsFeature(int ftype)
+        {
+            switch(ftype)
+            {
+            case CL_DOUBLE:
+                return impl->double_support == 1;
+            case CL_UNIFIED_MEM:
+                return impl->unified_memory == 1;
+            case CL_VER_1_2:
+                return impl->clVersion.find("OpenCL 1.2") != String::npos;
+            default:
+                return false;
+            }
+        }
+
+        size_t Context::computeUnits()
+        {
+            return impl->maxComputeUnits;
+        }
+
+        size_t Context::maxWorkGroupSize()
+        {
+            return impl->maxWorkGroupSize;
+        }
+
+        void* Context::oclContext()
+        {
+            return impl->oclcontext;
+        }
+
+        void* Context::oclCommandQueue()
+        {
+            return impl->clCmdQueue;
+        }
+
         Info::Info()
         {
             impl = new Impl;
-            impl->oclplatform = 0;
-            impl->oclcontext = 0;
-            impl->clCmdQueue = 0;
-            impl->devnum = 0;
-            impl->maxDimensions = 0;
-            impl->maxWorkGroupSize = 0;
-            impl->maxWorkItemSizes = 0;
-            impl->maxComputeUnits = 0;
-            impl->double_support = 0;
-            //extra_options = 0;
         }
+
         void Info::release()
         {
             fft_teardown();
-            if(impl->oclplatform)
-            {
-                impl->oclplatform = 0;
-            }
-            if(impl->clCmdQueue)
-            {
-                openCLSafeCall(clReleaseCommandQueue(impl->clCmdQueue));
-            }
-            ProgramCache::getProgramCache()->releaseProgram();
-            if(impl->oclcontext)
-            {
-                openCLSafeCall(clReleaseContext(impl->oclcontext));
-            }
-            if(impl->maxWorkItemSizes)
-            {
-                delete[] impl->maxWorkItemSizes;
-                impl->maxWorkItemSizes = 0;
-            }
-            //if(extra_options)
-            //{
-            //	delete[] extra_options;
-            //	extra_options = 0;
-            //}
-            impl->devices.clear();
-            impl->devName.clear();
+            clBlasTeardown();
+            impl->release();
+            impl = new Impl;
             DeviceName.clear();
+            PlatformName.clear();
         }
+
         Info::~Info()
         {
-            release();
-            delete impl;
+            fft_teardown();
+            clBlasTeardown();
+            impl->release();
         }
+
         Info &Info::operator = (const Info &m)
         {
-            impl->oclplatform = m.impl->oclplatform;
-            impl->oclcontext = m.impl->oclcontext;
-            impl->clCmdQueue = m.impl->clCmdQueue;
-            impl->devnum = m.impl->devnum;
-            impl->maxDimensions = m.impl->maxDimensions;
-            impl->maxWorkGroupSize = m.impl->maxWorkGroupSize;
-            impl->maxWorkItemSizes = m.impl->maxWorkItemSizes;
-            impl->maxComputeUnits = m.impl->maxComputeUnits;
-            impl->double_support = m.impl->double_support;
-            memcpy(impl->extra_options, m.impl->extra_options, 512);
-            for(size_t i = 0; i < m.impl->devices.size(); i++)
-            {
-                impl->devices.push_back(m.impl->devices[i]);
-                impl->devName.push_back(m.impl->devName[i]);
-                DeviceName.push_back(m.DeviceName[i]);
-            }
+            impl->release();
+            impl = m.impl->copy();
+            DeviceName = m.DeviceName;
+            PlatformName = m.PlatformName;
             return *this;
         }
+
         Info::Info(const Info &m)
         {
-            impl = new Impl;
-            *this = m;
+            impl = m.impl->copy();
+            DeviceName = m.DeviceName;
+            PlatformName = m.PlatformName;
         }
     }//namespace ocl
 
diff --git a/modules/ocl/src/interpolate_frames.cpp b/modules/ocl/src/interpolate_frames.cpp
index fe06187b5c..deca8c8a32 100644
--- a/modules/ocl/src/interpolate_frames.cpp
+++ b/modules/ocl/src/interpolate_frames.cpp
@@ -43,9 +43,7 @@
 //
 //M*/
 
-#include <iomanip>
 #include "precomp.hpp"
-#include "mcwutil.hpp"
 
 using namespace cv;
 using namespace cv::ocl;
@@ -133,7 +131,7 @@ void cv::ocl::interpolateFrames(const oclMat &frame0, const oclMat &frame1,
 void interpolate::memsetKernel(float val, oclMat &img, int height, int offset)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "memsetKernel";
+    String kernelName = "memsetKernel";
     std::vector< std::pair<size_t, const void *> > args;
     int step = img.step / sizeof(float);
     offset = step * height * offset;
@@ -152,7 +150,7 @@ void interpolate::memsetKernel(float val, oclMat &img, int height, int offset)
 void interpolate::normalizeKernel(oclMat &buffer, int height, int factor_offset, int dst_offset)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "normalizeKernel";
+    String kernelName = "normalizeKernel";
     std::vector< std::pair<size_t, const void *> > args;
     int step   = buffer.step / sizeof(float);
     factor_offset = step * height * factor_offset;
@@ -174,7 +172,7 @@ void interpolate::forwardWarpKernel(const oclMat &src, oclMat &buffer, const ocl
                                     int b_offset, int d_offset)
 {
     Context *clCxt = Context::getContext();
-    std::string kernelName = "forwardWarpKernel";
+    String kernelName = "forwardWarpKernel";
     std::vector< std::pair<size_t, const void *> > args;
     int f_step  = u.step / sizeof(float); // flow step
     int b_step  = buffer.step / sizeof(float);
@@ -212,7 +210,7 @@ void interpolate::blendFrames(const oclMat &frame0, const oclMat &/*frame1*/, co
     int step = buffer.step / sizeof(float);
 
     Context *clCxt = Context::getContext();
-    std::string kernelName = "blendFramesKernel";
+    String kernelName = "blendFramesKernel";
     std::vector< std::pair<size_t, const void *> > args;
 
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&tex_src0));
@@ -237,4 +235,3 @@ void interpolate::bindImgTex(const oclMat &img, cl_mem &texture)
     }
     texture = bindTexture(img);
 }
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_and.cl b/modules/ocl/src/kernels/arithm_bitwise_and.cl
deleted file mode 100644
index f954452b1f..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_and.cl
+++ /dev/null
@@ -1,323 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_and without mask**************************************/
-__kernel void arithm_bitwise_and_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
-     uchar4 src1_data = vload4(0, src1 + src1_index_fix);
-     uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-
-     if(src1_index < 0)
-     {     
-        uchar4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        uchar4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
-
-        uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = dst_data;
-    }
-}
-
-
-__kernel void arithm_bitwise_and_D1 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
-     char4 src1_data = vload4(0, src1 + src1_index_fix);
-     char4 src2_data = vload4(0, src2 + src2_index_fix);
-
-     if(src1_index < 0)
-     {     
-        char4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        char4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
-        char4 dst_data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global char4 *)(dst + dst_index)) = dst_data;
-    }
-}
-
-
-__kernel void arithm_bitwise_and_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global ushort *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
-        ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index_fix));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index_fix));
-
-     if(src1_index < 0)
-     {     
-        ushort4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        ushort4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
-        ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-        ushort4 tmp_data = src1_data & src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global short *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
-        short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index_fix));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index_fix));
-
-     if(src1_index < 0)
-     {     
-        short4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        short4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
-        short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-        short4 tmp_data = src1_data & src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global int *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int tmp  = data1 & data2;
-
-        *((__global int *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-__kernel void arithm_bitwise_and_D5 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 tmp = data1 & data2;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_and_D6 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data1 & data2;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_and_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_and_mask.cl
deleted file mode 100644
index d1f745ff29..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_and_mask.cl
+++ /dev/null
@@ -1,1138 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_and with mask**************************************/
-__kernel void arithm_bitwise_and_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C1_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data & src2_data;
-
-        data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-    short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-    short2 tmp_data = src1_data & src2_data;
-
-        data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C1_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C1_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_and_with_mask_C1_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_and_with_mask_C2_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_and_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C2_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int    *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C2_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_and_with_mask_C2_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 =  src1_data_0 & src2_data_0;
-        uchar4 tmp_data_1 =  src1_data_1 & src2_data_1;
-        uchar4 tmp_data_2 =  src1_data_2 & src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_bitwise_and_with_mask_C3_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        char4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        char4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 =  src1_data_0 & src2_data_0;
-        char4 tmp_data_1 =  src1_data_1 & src2_data_1;
-        char4 tmp_data_2 =  src1_data_2 & src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_bitwise_and_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0));
-        ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4));
-        ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 =  src1_data_0 & src2_data_0 ;
-        ushort2 tmp_data_1 =  src1_data_1 & src2_data_1 ;
-        ushort2 tmp_data_2 =  src1_data_2 & src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0));
-        short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4));
-        short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 =  src1_data_0 & src2_data_0 ;
-        short2 tmp_data_1 =  src1_data_1 & src2_data_1 ;
-        short2 tmp_data_2 =  src1_data_2 & src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C3_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0));
-        int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4));
-        int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 =  src1_data_0 &  src2_data_0 ;
-        int tmp_data_1 =  src1_data_1 &  src2_data_1 ;
-        int tmp_data_2 =  src1_data_2 &  src2_data_2 ;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C3_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0));
-        char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4));
-        char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 & src2_data_0;
-        char4 tmp_data_1 = src1_data_1 & src2_data_1;
-        char4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_and_with_mask_C3_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 24) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 ));
-        char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 ));
-        char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-
-        char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 ));
-        char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 ));
-        char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 ));
-        char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 ));
-        char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
-
-        char8 tmp_data_0 = src1_data_0 & src2_data_0;
-        char8 tmp_data_1 = src1_data_1 & src2_data_1;
-        char8 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_and_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_and_with_mask_C4_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)(src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_and_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C4_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_and_with_mask_C4_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_and_with_mask_C4_D6 (__global char *src1, int src1_step, int src1_offset,
-                                                  __global char *src2, int src2_step, int src2_offset,
-                                                  __global uchar  *mask, int mask_step, int mask_offset,
-                                                  __global char *dst,  int dst_step,  int dst_offset,
-                                                  int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
-        char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
-        char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-        char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
-
-        char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
-        char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
-        char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-        char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
-
-        char8 dst_data_0  = *((__global char8 *)((__global char *)dst  + dst_index + 0));
-        char8 dst_data_1  = *((__global char8 *)((__global char *)dst  + dst_index + 8));
-        char8 dst_data_2  = *((__global char8 *)((__global char *)dst  + dst_index + 16));
-        char8 dst_data_3  = *((__global char8 *)((__global char *)dst  + dst_index + 24));
-
-        char8 data_0 = src_data1_0 & src_data2_0;
-        char8 data_1 = src_data1_1 & src_data2_1;
-        char8 data_2 = src_data1_2 & src_data2_2;
-        char8 data_3 = src_data1_3 & src_data2_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_and_scalar.cl b/modules/ocl/src/kernels/arithm_bitwise_and_scalar.cl
deleted file mode 100644
index a5152ce0bf..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_and_scalar.cl
+++ /dev/null
@@ -1,907 +0,0 @@
-////////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//
-#if defined (__ATI__)
-#pragma OPENCL EXTENSION cl_amd_fp64:enable
-#elif defined (__NVIDIA__)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************and with scalar without mask**************************************/
-__kernel void arithm_s_bitwise_and_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data & src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-
-        short2 tmp_data = src1_data & src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C1_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-
-        int data = src_data1 & src_data2;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C1_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        char4 data  = *((__global char4 *)((__global char *)dst  + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        short4 tmp_data = src1_data & src2_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-
-        ushort2 data = src_data1 & src_data2;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-
-        short2 data = src_data1 & src_data2;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-
-        int2 data = src_data1 & src_data2;
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        char8 tmp_data = src1_data & src2_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = tmp_data;
-      }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_C2_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        short8 tmp_data = src1_data & src2_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = src1_data_0 & src2_data_0;
-        uchar4 tmp_data_1 = src1_data_1 & src2_data_1;
-        uchar4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 = convert_char4_sat(convert_uchar4_sat(src1_data_0) & convert_uchar4_sat(src2_data_0));
-        char4 tmp_data_1 = convert_char4_sat(convert_uchar4_sat(src1_data_1) & convert_uchar4_sat(src2_data_1));
-        char4 tmp_data_2 = convert_char4_sat(convert_uchar4_sat(src1_data_2) & convert_uchar4_sat(src2_data_2));
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0 & src2_data_0;
-        ushort2 tmp_data_1 = src1_data_1 & src2_data_1;
-        ushort2 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = src1_data_0 & src2_data_0;
-        short2 tmp_data_1 = src1_data_1 & src2_data_1;
-        short2 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 & src2_data_0;
-        int tmp_data_1 = src1_data_1 & src2_data_1;
-        int tmp_data_2 = src1_data_2 & src2_data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 & src2_data_0;
-        char4 tmp_data_1 = src1_data_1 & src2_data_1;
-        char4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 & src2_data_0;
-        short4 tmp_data_1 = src1_data_1 & src2_data_1;
-        short4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-
-        uchar4 data = src_data1 & src2;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-
-        char4 data = src_data1 & src2;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-
-        ushort4 data = src_data1 & src2;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-
-        short4 data = src_data1 & src2;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-
-        int4 data = src_data1 & src2;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-
-        char16 tmp_data = src1_data & src2_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 tmp_data_0 = src1_data_0 & src2_data_0;
-        short4 tmp_data_1 = src1_data_1 & src2_data_1;
-        short4 tmp_data_2 = src1_data_2 & src2_data_2;
-        short4 tmp_data_3 = src1_data_3 & src2_data_3;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-       *((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
-
-    }
-}
-#endif
diff --git a/modules/ocl/src/kernels/arithm_bitwise_and_scalar_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_and_scalar_mask.cl
deleted file mode 100644
index 50304aa34a..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_and_scalar_mask.cl
+++ /dev/null
@@ -1,1058 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (__ATI__)
-#pragma OPENCL EXTENSION cl_amd_fp64:enable
-#elif defined (__NVIDIA__)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_and with scalar with mask**************************************/
-__kernel void arithm_s_bitwise_and_with_mask_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_with_mask_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_with_mask_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data & src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        short2 tmp_data = src1_data & src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C1_D4 (__global   int   *src1, int src1_step, int src1_offset,
-                                            __global   int   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_with_mask_C1_D5 (__global char *src1, int src1_step, int src1_offset,
-                                                    __global char *dst,  int dst_step,  int dst_offset,
-                                                    __global   uchar *mask, int mask_step, int mask_offset,
-                                                    char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src1_data & src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_with_mask_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                            __global short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src1_data & src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_with_mask_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data & src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_with_mask_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data & src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_with_mask_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 & src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global  char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        char8 dst_data = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src1_data & src2_data;
-
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_with_mask_C2_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        short8 dst_data = *((__global short8 *)((__global char *)dst  + dst_index));
-
-        short8 data = src1_data & src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_with_mask_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = src1_data_0 & src2_data_0;
-        uchar4 tmp_data_1 = src1_data_1 & src2_data_1;
-        uchar4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_with_mask_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 & src2_data_0;
-        char4 tmp_data_1 = src1_data_1 & src2_data_1;
-        char4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_with_mask_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0 & src2_data_0;
-        ushort2 tmp_data_1 = src1_data_1 & src2_data_1;
-        ushort2 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = src1_data_0 & src2_data_0;
-        short2 tmp_data_1 = src1_data_1 & src2_data_1;
-        short2 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 & src2_data_0;
-        int tmp_data_1 = src1_data_1 & src2_data_1;
-        int tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 & src2_data_0;
-        char4 tmp_data_1 = src1_data_1 & src2_data_1;
-        char4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_with_mask_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar  *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        uchar mask_data = * (mask + mask_index);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 & src2_data_0;
-        short4 tmp_data_1 = src1_data_1 & src2_data_1;
-        short4 tmp_data_2 = src1_data_2 & src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_and_with_mask_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 & src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_and_with_mask_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 & src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_and_with_mask_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 & src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 & src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 & src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_and_with_mask_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                                    __global   char *dst,  int dst_step,  int dst_offset,
-                                                    __global   uchar *mask, int mask_step, int mask_offset,
-                                                    char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src1_data & src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_and_with_mask_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 dst_data_0  = *((__global short4 *)((__global char *)dst  + dst_index + 0));
-        short4 dst_data_1  = *((__global short4 *)((__global char *)dst  + dst_index + 8));
-        short4 dst_data_2  = *((__global short4 *)((__global char *)dst  + dst_index + 16));
-        short4 dst_data_3  = *((__global short4 *)((__global char *)dst  + dst_index + 24));
-
-        short4 data_0 = src1_data_0 & src2_data_0;
-        short4 data_1 = src1_data_1 & src2_data_1;
-        short4 data_2 = src1_data_2 & src2_data_2;
-        short4 data_3 = src1_data_3 & src2_data_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global short4 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global short4 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global short4 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global short4 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_or.cl b/modules/ocl/src/kernels/arithm_bitwise_or.cl
deleted file mode 100644
index 01e3a2f998..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_or.cl
+++ /dev/null
@@ -1,279 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_or without mask**************************************/
-__kernel void arithm_bitwise_or_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-      int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-      int src2_index_fix = src2_index < 0 ? 0 : src2_index;
-        uchar4 src1_data = vload4(0, src1 + src1_index_fix);
-        uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-      if(src1_index < 0)
-      {
-        uchar4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-      }
-      if(src2_index < 0)
-      {
-        uchar4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-      }
-        uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = dst_data;
-    }
-}
-
-
-__kernel void arithm_bitwise_or_D1 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-
-        char4 dst_data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global char4 *)(dst + dst_index)) = dst_data;
-    }
-}
-
-
-__kernel void arithm_bitwise_or_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global ushort *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-        ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
-
-        ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-        ushort4 tmp_data = src1_data | src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global short *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-        short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
-
-        short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-        short4 tmp_data = src1_data | src2_data;
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global int *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int tmp  = data1 | data2;
-
-        *((__global int *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-__kernel void arithm_bitwise_or_D5 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 tmp = data1 | data2;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_or_D6 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data1 | data2;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_or_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_or_mask.cl
deleted file mode 100644
index 92d98ec01c..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_or_mask.cl
+++ /dev/null
@@ -1,1138 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_or with mask**************************************/
-__kernel void arithm_bitwise_or_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C1_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data | src2_data;
-
-        data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-    short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-    short2 tmp_data = src1_data | src2_data;
-
-        data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C1_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C1_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_or_with_mask_C1_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_or_with_mask_C2_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_or_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C2_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int    *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C2_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_or_with_mask_C2_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 =  src1_data_0 | src2_data_0;
-        uchar4 tmp_data_1 =  src1_data_1 | src2_data_1;
-        uchar4 tmp_data_2 =  src1_data_2 | src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_bitwise_or_with_mask_C3_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        char4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        char4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 =  src1_data_0 | src2_data_0;
-        char4 tmp_data_1 =  src1_data_1 | src2_data_1;
-        char4 tmp_data_2 =  src1_data_2 | src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_bitwise_or_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0));
-        ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4));
-        ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 =  src1_data_0 | src2_data_0 ;
-        ushort2 tmp_data_1 =  src1_data_1 | src2_data_1 ;
-        ushort2 tmp_data_2 =  src1_data_2 | src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0));
-        short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4));
-        short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 =  src1_data_0 | src2_data_0 ;
-        short2 tmp_data_1 =  src1_data_1 | src2_data_1 ;
-        short2 tmp_data_2 =  src1_data_2 | src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C3_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0));
-        int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4));
-        int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 =  src1_data_0 |  src2_data_0 ;
-        int tmp_data_1 =  src1_data_1 |  src2_data_1 ;
-        int tmp_data_2 =  src1_data_2 |  src2_data_2 ;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C3_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0));
-        char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4));
-        char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 | src2_data_0;
-        char4 tmp_data_1 = src1_data_1 | src2_data_1;
-        char4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_or_with_mask_C3_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 24) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 ));
-        char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 ));
-        char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-
-        char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 ));
-        char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 ));
-        char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 ));
-        char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 ));
-        char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
-
-        char8 tmp_data_0 = src1_data_0 | src2_data_0;
-        char8 tmp_data_1 = src1_data_1 | src2_data_1;
-        char8 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_or_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_or_with_mask_C4_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)(src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_or_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C4_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_or_with_mask_C4_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_or_with_mask_C4_D6 (__global char *src1, int src1_step, int src1_offset,
-                                                  __global char *src2, int src2_step, int src2_offset,
-                                                  __global uchar  *mask, int mask_step, int mask_offset,
-                                                  __global char *dst,  int dst_step,  int dst_offset,
-                                                  int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
-        char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
-        char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-        char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
-
-        char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
-        char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
-        char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-        char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
-
-        char8 dst_data_0  = *((__global char8 *)((__global char *)dst  + dst_index + 0));
-        char8 dst_data_1  = *((__global char8 *)((__global char *)dst  + dst_index + 8));
-        char8 dst_data_2  = *((__global char8 *)((__global char *)dst  + dst_index + 16));
-        char8 dst_data_3  = *((__global char8 *)((__global char *)dst  + dst_index + 24));
-
-        char8 data_0 = src_data1_0 | src_data2_0;
-        char8 data_1 = src_data1_1 | src_data2_1;
-        char8 data_2 = src_data1_2 | src_data2_2;
-        char8 data_3 = src_data1_3 | src_data2_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_or_scalar.cl b/modules/ocl/src/kernels/arithm_bitwise_or_scalar.cl
deleted file mode 100644
index bbd5f3fb2e..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_or_scalar.cl
+++ /dev/null
@@ -1,914 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************and with scalar without mask**************************************/
-__kernel void arithm_s_bitwise_or_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data | src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-
-        short2 tmp_data = src1_data | src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C1_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-
-        int data = src_data1 | src_data2;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C1_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src_data2 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        char4 data = src_data1 | src_data2;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        short4 tmp_data = src1_data | src2_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-
-__kernel void arithm_s_bitwise_or_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-
-        ushort2 data = src_data1 | src_data2;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-
-        short2 data = src_data1 | src_data2;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-
-        int2 data = src_data1 | src_data2;
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        char8 data = src_data1 | src_data2;
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_C2_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        short8 tmp_data = src1_data & src2_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_or_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 =  src1_data_0  |  src2_data_0  ;
-        uchar4 tmp_data_1 =  src1_data_1  |  src2_data_1  ;
-        uchar4 tmp_data_2 =  src1_data_2  |  src2_data_2  ;
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 =  src1_data_0  |  src2_data_0;
-        char4 tmp_data_1 =  src1_data_1  |  src2_data_1;
-        char4 tmp_data_2 =  src1_data_2  |  src2_data_2;
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0  |  src2_data_0  ;
-        ushort2 tmp_data_1 = src1_data_1  |  src2_data_1  ;
-        ushort2 tmp_data_2 = src1_data_2  |  src2_data_2  ;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 =  src1_data_0  |  src2_data_0  ;
-        short2 tmp_data_1 =  src1_data_1  |  src2_data_1  ;
-        short2 tmp_data_2 =  src1_data_2  |  src2_data_2  ;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 | src2_data_0;
-        int tmp_data_1 = src1_data_1 | src2_data_1;
-        int tmp_data_2 = src1_data_2 | src2_data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        char4 tmp_data_0 = src1_data_0 | src2_data_0;
-        char4 tmp_data_1 = src1_data_1 | src2_data_1;
-        char4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 | src2_data_0;
-        short4 tmp_data_1 = src1_data_1 | src2_data_1;
-        short4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_or_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-
-        uchar4 data = src_data1 | src2;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-
-        char4 data = src_data1 | src2;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-
-        ushort4 data = src_data1 | src2;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-
-        short4 data = src_data1 | src2;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-
-        int4 data = src_data1 | src2;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-
-        char16 data = src_data1 | src_data2;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 tmp_data_0 = src1_data_0 | src2_data_0;
-        short4 tmp_data_1 = src1_data_1 | src2_data_1;
-        short4 tmp_data_2 = src1_data_2 | src2_data_2;
-        short4 tmp_data_3 = src1_data_3 | src2_data_3;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-       *((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
-
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_or_scalar_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_or_scalar_mask.cl
deleted file mode 100644
index 153398706f..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_or_scalar_mask.cl
+++ /dev/null
@@ -1,1081 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_or with scalar with mask**************************************/
-__kernel void arithm_s_bitwise_or_with_mask_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_with_mask_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_with_mask_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data | src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        short2 tmp_data = src1_data | src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C1_D4 (__global   int   *src1, int src1_step, int src1_offset,
-                                            __global   int   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_with_mask_C1_D5 (__global   char   *src1, int src1_step, int src1_offset,
-                                            __global   char   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src_data2 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_with_mask_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                            __global short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src1_data | src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_or_with_mask_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data | src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_with_mask_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data | src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_with_mask_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        char8 dst_data = *((__global char8 *)((__global char *)dst  + dst_index));
-          char8 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-
-      }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_with_mask_C2_D6 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        short8 dst_data = *((__global short8 *)((__global char *)dst  + dst_index));
-
-        short8 data = src1_data | src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_or_with_mask_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = src1_data_0 | src2_data_0;
-        uchar4 tmp_data_1 = src1_data_1 | src2_data_1;
-        uchar4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_with_mask_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 | src2_data_0;
-        char4 tmp_data_1 = src1_data_1 | src2_data_1;
-        char4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_with_mask_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0 | src2_data_0;
-        ushort2 tmp_data_1 = src1_data_1 | src2_data_1;
-        ushort2 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = src1_data_0 | src2_data_0;
-        short2 tmp_data_1 = src1_data_1 | src2_data_1;
-        short2 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 | src2_data_0;
-        int tmp_data_1 = src1_data_1 | src2_data_1;
-        int tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 | src2_data_0;
-        char4 tmp_data_1 = src1_data_1 | src2_data_1;
-        char4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-
-       }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_with_mask_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar  *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        uchar mask_data = * (mask + mask_index);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 | src2_data_0;
-        short4 tmp_data_1 = src1_data_1 | src2_data_1;
-        short4 tmp_data_2 = src1_data_2 | src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_or_with_mask_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 | src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_or_with_mask_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 | src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_or_with_mask_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 | src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 | src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 | src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_or_with_mask_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 | src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_or_with_mask_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 dst_data_0  = *((__global short4 *)((__global char *)dst  + dst_index + 0));
-        short4 dst_data_1  = *((__global short4 *)((__global char *)dst  + dst_index + 8));
-        short4 dst_data_2  = *((__global short4 *)((__global char *)dst  + dst_index + 16));
-        short4 dst_data_3  = *((__global short4 *)((__global char *)dst  + dst_index + 24));
-
-        short4 data_0 = src1_data_0 | src2_data_0;
-        short4 data_1 = src1_data_1 | src2_data_1;
-        short4 data_2 = src1_data_2 | src2_data_2;
-        short4 data_3 = src1_data_3 | src2_data_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global short4 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global short4 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global short4 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global short4 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_xor_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_xor_mask.cl
deleted file mode 100644
index 248654ef74..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_xor_mask.cl
+++ /dev/null
@@ -1,1138 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_xor with mask**************************************/
-__kernel void arithm_bitwise_xor_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C1_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data ^ src2_data;
-
-        data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-    short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-    short2 tmp_data = src1_data ^ src2_data;
-
-        data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C1_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C1_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_xor_with_mask_C1_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_xor_with_mask_C2_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_xor_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C2_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int    *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C2_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
-        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_xor_with_mask_C2_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 =  src1_data_0 ^ src2_data_0;
-        uchar4 tmp_data_1 =  src1_data_1 ^ src2_data_1;
-        uchar4 tmp_data_2 =  src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_bitwise_xor_with_mask_C3_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        char4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        char4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 =  src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 =  src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 =  src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_bitwise_xor_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0));
-        ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4));
-        ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 =  src1_data_0 ^ src2_data_0 ;
-        ushort2 tmp_data_1 =  src1_data_1 ^ src2_data_1 ;
-        ushort2 tmp_data_2 =  src1_data_2 ^ src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0));
-        short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4));
-        short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 =  src1_data_0 ^ src2_data_0 ;
-        short2 tmp_data_1 =  src1_data_1 ^ src2_data_1 ;
-        short2 tmp_data_2 =  src1_data_2 ^ src2_data_2 ;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C3_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0));
-        int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4));
-        int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 =  src1_data_0 ^  src2_data_0 ;
-        int tmp_data_1 =  src1_data_1 ^  src2_data_1 ;
-        int tmp_data_2 =  src1_data_2 ^  src2_data_2 ;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C3_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0));
-        char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4));
-        char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_xor_with_mask_C3_D6 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 24) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 ));
-        char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 ));
-        char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-
-        char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 ));
-        char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 ));
-        char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-
-        uchar mask_data = * (mask + mask_index);
-
-        char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 ));
-        char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 ));
-        char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
-
-        char8 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char8 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char8 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-
-
-
-__kernel void arithm_bitwise_xor_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_bitwise_xor_with_mask_C4_D1 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 src_data2 = *((__global char4 *)(src2 + src2_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_bitwise_xor_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C4_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_bitwise_xor_with_mask_C4_D5 (__global char *src1, int src1_step, int src1_offset,
-                                          __global char *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global char *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_xor_with_mask_C4_D6 (__global char *src1, int src1_step, int src1_offset,
-                                                  __global char *src2, int src2_step, int src2_offset,
-                                                  __global uchar  *mask, int mask_step, int mask_offset,
-                                                  __global char *dst,  int dst_step,  int dst_offset,
-                                                  int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
-        char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
-        char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
-        char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
-
-        char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
-        char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
-        char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
-        char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
-
-        char8 dst_data_0  = *((__global char8 *)((__global char *)dst  + dst_index + 0));
-        char8 dst_data_1  = *((__global char8 *)((__global char *)dst  + dst_index + 8));
-        char8 dst_data_2  = *((__global char8 *)((__global char *)dst  + dst_index + 16));
-        char8 dst_data_3  = *((__global char8 *)((__global char *)dst  + dst_index + 24));
-
-        char8 data_0 = src_data1_0 ^ src_data2_0;
-        char8 data_1 = src_data1_1 ^ src_data2_1;
-        char8 data_2 = src_data1_2 ^ src_data2_2;
-        char8 data_3 = src_data1_3 ^ src_data2_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_bitwise_xor_scalar.cl b/modules/ocl/src/kernels/arithm_bitwise_xor_scalar.cl
deleted file mode 100644
index 318432a186..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_xor_scalar.cl
+++ /dev/null
@@ -1,907 +0,0 @@
-////////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//
-#if defined (__ATI__)
-#pragma OPENCL EXTENSION cl_amd_fp64:enable
-#elif defined (__NVIDIA__)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************xor with scalar without mask**************************************/
-__kernel void arithm_s_bitwise_xor_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data ^ src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-
-        short2 tmp_data = src1_data ^ src2_data;
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C1_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-
-        int data = src_data1 ^ src_data2;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C1_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        char4 data  = *((__global char4 *)((__global char *)dst  + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-
-        short4 tmp_data = src1_data ^ src2_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-
-        ushort2 data = src_data1 ^ src_data2;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-
-        short2 data = src_data1 ^ src_data2;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-
-        int2 data = src_data1 ^ src_data2;
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        char8 tmp_data = src1_data ^ src2_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = tmp_data;
-      }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_C2_D6 (__global short *src1, int src1_step, int src1_offset,
-                                  __global short *dst,  int dst_step,  int dst_offset,
-                                  short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-
-        short8 tmp_data = src1_data ^ src2_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        uchar4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        uchar4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        ushort2 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        ushort2 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        short2 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        short2 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 ^ src2_data_0;
-        int tmp_data_1 = src1_data_1 ^ src2_data_1;
-        int tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        short4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        short4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-
-        uchar4 data = src_data1 ^ src2;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-
-        char4 data = src_data1 ^ src2;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-
-        ushort4 data = src_data1 ^ src2;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-
-        short4 data = src_data1 ^ src2;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-
-        int4 data = src_data1 ^ src2;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                  __global   char *dst,  int dst_step,  int dst_offset,
-                                  char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-
-        char16 tmp_data = src1_data ^ src2_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = tmp_data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        short4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        short4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-        short4 tmp_data_3 = src1_data_3 ^ src2_data_3;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-       *((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
-
-    }
-}
-#endif
diff --git a/modules/ocl/src/kernels/arithm_bitwise_xor_scalar_mask.cl b/modules/ocl/src/kernels/arithm_bitwise_xor_scalar_mask.cl
deleted file mode 100644
index 4efa2dac6c..0000000000
--- a/modules/ocl/src/kernels/arithm_bitwise_xor_scalar_mask.cl
+++ /dev/null
@@ -1,1058 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jiang Liyuan, jlyuan001.good@163.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#if defined (__ATI__)
-#pragma OPENCL EXTENSION cl_amd_fp64:enable
-#elif defined (__NVIDIA__)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_xor with scalar with mask**************************************/
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = (ushort2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        ushort2 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = (short2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        short2 tmp_data = src1_data ^ src2_data;
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D4 (__global   int   *src1, int src1_step, int src1_offset,
-                                            __global   int   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D5 (__global char *src1, int src1_step, int src1_offset,
-                                                    __global char *dst,  int dst_step,  int dst_offset,
-                                                    __global   uchar *mask, int mask_step, int mask_offset,
-                                                    char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
-        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
-
-        char4 data = src1_data ^ src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_with_mask_C1_D6 (__global short *src1, int src1_step, int src1_offset,
-                                            __global short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src1_data ^ src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        char4 src1_data = vload4(0, src1 + src1_index);
-        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        char4 data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        ushort2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = (short2)(src2.x, src2.y);
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        short2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = src_data1 ^ src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global  char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
-        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        char8 dst_data = *((__global char8 *)((__global char *)dst  + dst_index));
-
-        char8 data = src1_data ^ src2_data;
-
-        data = mask_data ? data : dst_data;
-
-        *((__global char8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_with_mask_C2_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
-        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
-        short8 dst_data = *((__global short8 *)((__global char *)dst  + dst_index));
-
-        short8 data = src1_data ^ src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global short8 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x);
-        uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y);
-        uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        uchar4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        uchar4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        char4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        char4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        char4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x);
-        char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y);
-        char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)(dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)(dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)(dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global char4 *)(dst + dst_index + 0)) = data_0;
-        *((__global char4 *)(dst + dst_index + 4)) = data_1;
-        *((__global char4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = (ushort2)(src2.x, src2.y);
-        ushort2 src2_data_1 = (ushort2)(src2.z, src2.x);
-        ushort2 src2_data_2 = (ushort2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        ushort2 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        ushort2 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = (short2)(src2.x, src2.y);
-        short2 src2_data_1 = (short2)(src2.z, src2.x);
-        short2 src2_data_2 = (short2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        short2 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        short2 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = src1_data_0 ^ src2_data_0;
-        int tmp_data_1 = src1_data_1 ^ src2_data_1;
-        int tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0));
-        char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4));
-        char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8));
-
-        char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB);
-
-        uchar mask_data = * (mask + mask_index);
-
-        char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0));
-        char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4));
-        char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8));
-
-        char4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        char4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        char4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_with_mask_C3_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar  *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 ));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 ));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-
-        uchar mask_data = * (mask + mask_index);
-
-        short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 ));
-        short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 ));
-        short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
-
-        short4 tmp_data_0 = src1_data_0 ^ src2_data_0;
-        short4 tmp_data_1 = src1_data_1 ^ src2_data_1;
-        short4 tmp_data_2 = src1_data_2 ^ src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global short4 *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global short4 *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global short4 *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            uchar4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = src_data1 ^ src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-
-
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D1 (__global   char *src1, int src1_step, int src1_offset,
-                                            __global   char *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            char4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
-        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
-
-        char4 data = src_data1 ^ src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global char4 *)(dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            ushort4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = src_data1 ^ src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            short4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = src_data1 ^ src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = src_data1 ^ src2;
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D5 (__global   char *src1, int src1_step, int src1_offset,
-                                                    __global   char *dst,  int dst_step,  int dst_offset,
-                                                    __global   uchar *mask, int mask_step, int mask_offset,
-                                                    char16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
-        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
-                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
-        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
-
-        char16 data = src1_data ^ src2_data;
-        data = mask_data ? data : dst_data;
-
-        *((__global char16 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_bitwise_xor_with_mask_C4_D6 (__global short *src1, int src1_step, int src1_offset,
-                                                    __global short *dst,  int dst_step,  int dst_offset,
-                                                    __global uchar *mask, int mask_step, int mask_offset,
-                                                    short16 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
-        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
-        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
-        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
-
-        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
-        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
-        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
-        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
-
-        short4 dst_data_0  = *((__global short4 *)((__global char *)dst  + dst_index + 0));
-        short4 dst_data_1  = *((__global short4 *)((__global char *)dst  + dst_index + 8));
-        short4 dst_data_2  = *((__global short4 *)((__global char *)dst  + dst_index + 16));
-        short4 dst_data_3  = *((__global short4 *)((__global char *)dst  + dst_index + 24));
-
-        short4 data_0 = src1_data_0 ^ src2_data_0;
-        short4 data_1 = src1_data_1 ^ src2_data_1;
-        short4 data_2 = src1_data_2 ^ src2_data_2;
-        short4 data_3 = src1_data_3 ^ src2_data_3;
-
-        data_0 = mask_data ? data_0 : dst_data_0;
-        data_1 = mask_data ? data_1 : dst_data_1;
-        data_2 = mask_data ? data_2 : dst_data_2;
-        data_3 = mask_data ? data_3 : dst_data_3;
-
-        *((__global short4 *)((__global char *)dst + dst_index + 0)) = data_0;
-        *((__global short4 *)((__global char *)dst + dst_index + 8)) = data_1;
-        *((__global short4 *)((__global char *)dst + dst_index + 16)) = data_2;
-        *((__global short4 *)((__global char *)dst + dst_index + 24)) = data_3;
-    }
-}
-#endif
-
diff --git a/modules/ocl/src/kernels/arithm_sub.cl b/modules/ocl/src/kernels/arithm_sub.cl
deleted file mode 100644
index 9cf37970b2..0000000000
--- a/modules/ocl/src/kernels/arithm_sub.cl
+++ /dev/null
@@ -1,1104 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////////////////SUB////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************sub without mask**************************************/
-__kernel void arithm_sub_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-
-        uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
-        short4 tmp      = convert_short4_sat(src1_data) - convert_short4_sat(src2_data);
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = dst_data;
-    }
-}
-__kernel void arithm_sub_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global ushort *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-        ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
-
-        ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-        int4    tmp = convert_int4_sat(src1_data) - convert_int4_sat(src2_data);
-        ushort4 tmp_data = convert_ushort4_sat(tmp);
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-__kernel void arithm_sub_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global short *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
-
-        short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
-
-        short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-        int4   tmp = convert_int4_sat(src1_data) - convert_int4_sat(src2_data);
-        short4 tmp_data = convert_short4_sat(tmp);
-
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = dst_data;
-    }
-}
-
-__kernel void arithm_sub_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global int *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int data2 = *((__global int *)((__global char *)src2 + src2_index));
-        long tmp  = (long)(data1) - (long)(data2);
-
-        *((__global int *)((__global char *)dst + dst_index)) = convert_int_sat(tmp);
-    }
-}
-__kernel void arithm_sub_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global float *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        float data1 = *((__global float *)((__global char *)src1 + src1_index));
-        float data2 = *((__global float *)((__global char *)src2 + src2_index));
-        float tmp = data1 - data2;
-
-        *((__global float *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_sub_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global double *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        double data1 = *((__global double *)((__global char *)src1 + src1_index));
-        double data2 = *((__global double *)((__global char *)src2 + src2_index));
-
-        *((__global double *)((__global char *)dst + dst_index)) = data1 - data2;
-    }
-}
-#endif
-
-/**************************************sub with mask**************************************/
-__kernel void arithm_sub_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        short4 tmp = convert_short4_sat(src1_data) - convert_short4_sat(src2_data);
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) - convert_int2_sat(src2_data);
-        ushort2 tmp_data = convert_ushort2_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) - convert_int2_sat(src2_data);
-        short2 tmp_data = convert_short2_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C1_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        int data = convert_int_sat((long)src_data1 - (long)src_data2);
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_sub_with_mask_C1_D5 (__global float *src1, int src1_step, int src1_offset,
-                                          __global float *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global float *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float src_data1 = *((__global float *)((__global char *)src1 + src1_index));
-        float src_data2 = *((__global float *)((__global char *)src2 + src2_index));
-        float dst_data  = *((__global float *)((__global char *)dst  + dst_index));
-
-        float data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global float *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_sub_with_mask_C1_D6 (__global double *src1, int src1_step, int src1_offset,
-                                          __global double *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global double *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double src_data1 = *((__global double *)((__global char *)src1 + src1_index));
-        double src_data2 = *((__global double *)((__global char *)src2 + src2_index));
-        double dst_data  = *((__global double *)((__global char *)dst  + dst_index));
-
-        double data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global double *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-__kernel void arithm_sub_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        short4   tmp = convert_short4_sat(src1_data) - convert_short4_sat(src2_data);
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - convert_int2_sat(src_data2);
-        ushort2 data = convert_ushort2_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - convert_int2_sat(src_data2);
-        short2 data = convert_short2_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C2_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int    *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        int2 data = convert_int2_sat(convert_long2_sat(src_data1) - convert_long2_sat(src_data2));
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C2_D5 (__global float *src1, int src1_step, int src1_offset,
-                                          __global float *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global float *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index));
-        float2 src_data2 = *((__global float2 *)((__global char *)src2 + src2_index));
-        float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
-
-        float2 data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global float2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_sub_with_mask_C2_D6 (__global double *src1, int src1_step, int src1_offset,
-                                          __global double *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global double *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index));
-        double2 src_data2 = *((__global double2 *)((__global char *)src2 + src2_index));
-        double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
-
-        double2 data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global double2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_sub_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(convert_short4_sat(src1_data_0) - convert_short4_sat(src2_data_0));
-        uchar4 tmp_data_1 = convert_uchar4_sat(convert_short4_sat(src1_data_1) - convert_short4_sat(src2_data_1));
-        uchar4 tmp_data_2 = convert_uchar4_sat(convert_short4_sat(src1_data_2) - convert_short4_sat(src2_data_2));
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_sub_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0));
-        ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4));
-        ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) - convert_int2_sat(src2_data_0));
-        ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) - convert_int2_sat(src2_data_1));
-        ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) - convert_int2_sat(src2_data_2));
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_sub_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0));
-        short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4));
-        short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) - convert_int2_sat(src2_data_0));
-        short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) - convert_int2_sat(src2_data_1));
-        short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) - convert_int2_sat(src2_data_2));
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_sub_with_mask_C3_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0));
-        int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4));
-        int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = convert_int_sat((long)src1_data_0 - (long)src2_data_0);
-        int tmp_data_1 = convert_int_sat((long)src1_data_1 - (long)src2_data_1);
-        int tmp_data_2 = convert_int_sat((long)src1_data_2 - (long)src2_data_2);
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_sub_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset,
-                                          __global float *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global float *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = *((__global float *)((__global char *)src2 + src2_index + 0));
-        float src2_data_1 = *((__global float *)((__global char *)src2 + src2_index + 4));
-        float src2_data_2 = *((__global float *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_data_0 = src1_data_0 - src2_data_0;
-        float tmp_data_1 = src1_data_1 - src2_data_1;
-        float tmp_data_2 = src1_data_2 - src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_sub_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset,
-                                          __global double *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global double *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 24) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = *((__global double *)((__global char *)src2 + src2_index + 0 ));
-        double src2_data_1 = *((__global double *)((__global char *)src2 + src2_index + 8 ));
-        double src2_data_2 = *((__global double *)((__global char *)src2 + src2_index + 16));
-
-        uchar mask_data = * (mask + mask_index);
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 - src2_data_0;
-        double tmp_data_1 = src1_data_1 - src2_data_1;
-        double tmp_data_2 = src1_data_2 - src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-__kernel void arithm_sub_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        uchar4 data = convert_uchar4_sat(convert_short4_sat(src_data1) - convert_short4_sat(src_data2));
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) - convert_int4_sat(src_data2));
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        short4 data = convert_short4_sat(convert_int4_sat(src_data1) - convert_int4_sat(src_data2));
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C4_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        int4 data = convert_int4_sat(convert_long4_sat(src_data1) - convert_long4_sat(src_data2));
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_sub_with_mask_C4_D5 (__global float *src1, int src1_step, int src1_offset,
-                                          __global float *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global float *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index));
-        float4 src_data2 = *((__global float4 *)((__global char *)src2 + src2_index));
-        float4 dst_data  = *((__global float4 *)((__global char *)dst  + dst_index));
-
-        float4 data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global float4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_sub_with_mask_C4_D6 (__global double *src1, int src1_step, int src1_offset,
-                                          __global double *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global double *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index));
-        double4 src_data2 = *((__global double4 *)((__global char *)src2 + src2_index));
-        double4 dst_data  = *((__global double4 *)((__global char *)dst  + dst_index));
-
-        double4 data = src_data1 - src_data2;
-        data = mask_data ? data : dst_data;
-
-        *((__global double4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
diff --git a/modules/ocl/src/kernels/arithm_sub_scalar.cl b/modules/ocl/src/kernels/arithm_sub_scalar.cl
deleted file mode 100644
index 782bcd0607..0000000000
--- a/modules/ocl/src/kernels/arithm_sub_scalar.cl
+++ /dev/null
@@ -1,806 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-/**************************************sub with scalar without mask**************************************/
-__kernel void arithm_s_sub_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        int4 src2_data = (int4)(src2.x, src2.x, src2.x, src2.x);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        int2 src2_data = (int2)(src2.x, src2.x);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort2 tmp_data = convert_ushort2_sat(tmp);
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        int2 src2_data = (int2)(src2.x, src2.x);
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-
-        int2   tmp = convert_int2_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short2 tmp_data = convert_short2_sat(tmp);
-
-        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
-        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C1_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-
-        long tmp = (long)src_data1 - (long)src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        int data = convert_int_sat(tmp);
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C1_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                  __global   float *dst,  int dst_step,  int dst_offset,
-                                  float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        float src_data1 = *((__global float *)((__global char *)src1 + src1_index));
-        float src_data2 = src2.x;
-
-        float tmp = src_data1 - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-
-        *((__global float *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_C1_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                  __global   double *dst,  int dst_step,  int dst_offset,
-                                  double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        double src_data1 = *((__global double *)((__global char *)src1 + src1_index));
-        double src2_data = src2.x;
-
-        double data = src_data1 - src2_data;
-        data = isMatSubScalar ? data : -data;
-
-        *((__global double *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-
-__kernel void arithm_s_sub_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
-        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        int2 src_data2    = (int2)(src2.x, src2.y);
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort2 data = convert_ushort2_sat(tmp);
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short2 data = convert_short2_sat(tmp);
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        long2 tmp = convert_long2_sat(src_data1) - convert_long2_sat(src_data2);
-        tmp = isMatSubScalar ? tmp : -tmp;
-        int2 data = convert_int2_sat(tmp);
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C2_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                  __global   float *dst,  int dst_step,  int dst_offset,
-                                  float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index));
-        float2 src_data2 = (float2)(src2.x, src2.y);
-        float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
-
-        float2 tmp = src_data1 - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-
-        *((__global float2 *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_C2_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                  __global   double *dst,  int dst_step,  int dst_offset,
-                                  double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index));
-        double2 src_data2 = (double2)(src2.x, src2.y);
-        double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
-
-        double2 data = src_data1 - src_data2;
-        data = isMatSubScalar ? data : -data;
-
-        *((__global double2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_sub_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x);
-        int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y);
-        int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        int4 tmp_0 = convert_int4_sat(src1_data_0) - src2_data_0;
-        int4 tmp_1 = convert_int4_sat(src1_data_1) - src2_data_1;
-        int4 tmp_2 = convert_int4_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(tmp_0);
-        uchar4 tmp_data_1 = convert_uchar4_sat(tmp_1);
-        uchar4 tmp_data_2 = convert_uchar4_sat(tmp_2);
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_s_sub_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0;
-        int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1;
-        int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(tmp_0);
-        ushort2 tmp_data_1 = convert_ushort2_sat(tmp_1);
-        ushort2 tmp_data_2 = convert_ushort2_sat(tmp_2);
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_sub_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0;
-        int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1;
-        int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        short2 tmp_data_0 = convert_short2_sat(tmp_0);
-        short2 tmp_data_1 = convert_short2_sat(tmp_1);
-        short2 tmp_data_2 = convert_short2_sat(tmp_2);
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_sub_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        long tmp_0 = (long)src1_data_0 - (long)src2_data_0;
-        long tmp_1 = (long)src1_data_1 - (long)src2_data_1;
-        long tmp_2 = (long)src1_data_2 - (long)src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        int tmp_data_0 = convert_int_sat(tmp_0);
-        int tmp_data_1 = convert_int_sat(tmp_1);
-        int tmp_data_2 = convert_int_sat(tmp_2);
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-__kernel void arithm_s_sub_C3_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                  __global   float *dst,  int dst_step,  int dst_offset,
-                                  float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = src2.x;
-        float src2_data_1 = src2.y;
-        float src2_data_2 = src2.z;
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_0 = src1_data_0 - src2_data_0;
-        float tmp_1 = src1_data_1 - src2_data_1;
-        float tmp_2 = src1_data_2 - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= tmp_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= tmp_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= tmp_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_C3_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                  __global   double *dst,  int dst_step,  int dst_offset,
-                                  double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = src2.x;
-        double src2_data_1 = src2.y;
-        double src2_data_2 = src2.z;
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 - src2_data_0;
-        double tmp_data_1 = src1_data_1 - src2_data_1;
-        double tmp_data_2 = src1_data_2 - src2_data_2;
-
-        tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0;
-        tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1;
-        tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-    }
-}
-#endif
-__kernel void arithm_s_sub_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-
-        int4 tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 data = convert_uchar4_sat(tmp);
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-
-        int4 tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort4 data = convert_ushort4_sat(tmp);
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-
-        int4 tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short4 data = convert_short4_sat(tmp);
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-
-        long4 tmp = convert_long4_sat(src_data1) - convert_long4_sat(src2);
-        tmp = isMatSubScalar ? tmp : -tmp;
-        int4 data = convert_int4_sat(tmp);
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_C4_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                  __global   float *dst,  int dst_step,  int dst_offset,
-                                  float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index));
-
-        float4 tmp = src_data1 - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-
-        *((__global float4 *)((__global char *)dst + dst_index)) = tmp;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_C4_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                  __global   double *dst,  int dst_step,  int dst_offset,
-                                  double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index));
-
-        double4 data = src_data1 - src2;
-        data = isMatSubScalar ? data : -data;
-
-        *((__global double4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
diff --git a/modules/ocl/src/kernels/arithm_sub_scalar_mask.cl b/modules/ocl/src/kernels/arithm_sub_scalar_mask.cl
deleted file mode 100644
index 135354993c..0000000000
--- a/modules/ocl/src/kernels/arithm_sub_scalar_mask.cl
+++ /dev/null
@@ -1,941 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#if defined (DOUBLE_SUPPORT)
-#pragma OPENCL EXTENSION cl_khr_fp64:enable
-#endif
-
-/**************************************sub with scalar with mask**************************************/
-__kernel void arithm_s_sub_with_mask_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        int4 src2_data = (int4)(src2.x, src2.x, src2.x, src2.x);
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
-        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
-        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C1_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
-        int2 src2_data = (int2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort2 tmp_data = convert_ushort2_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C1_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
-        int2 src2_data = (int2)(src2.x, src2.x);
-        uchar2  mask_data = vload2(0, mask + mask_index);
-
-        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short2 tmp_data = convert_short2_sat(tmp);
-
-        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
-        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
-
-        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C1_D4 (__global   int   *src1, int src1_step, int src1_offset,
-                                            __global   int   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
-        int src_data2 = src2.x;
-        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
-
-        long tmp = (long)src_data1 - (long)src_data2;
-        tmp = isMatSubScalar ? tmp : - tmp;
-        int data = convert_int_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global int *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-__kernel void arithm_s_sub_with_mask_C1_D5 (__global   float   *src1, int src1_step, int src1_offset,
-                                            __global   float   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float src_data1 = *((__global float *)((__global char *)src1 + src1_index));
-        float src_data2 = src2.x;
-        float dst_data  = *((__global float *)((__global char *)dst  + dst_index));
-
-        float data = src_data1 - src_data2;
-        data = isMatSubScalar ? data : -data;
-        data = mask_data ? data : dst_data;
-
-        *((__global float *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_with_mask_C1_D6 (__global   double   *src1, int src1_step, int src1_offset,
-                                            __global   double   *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double src_data1 = *((__global double *)((__global char *)src1 + src1_index));
-        double src_data2 = src2.x;
-        double dst_data  = *((__global double *)((__global char *)dst  + dst_index));
-
-        double data = src_data1 - src_data2;
-        data = isMatSubScalar ? data : -data;
-        data = mask_data ? data : dst_data;
-
-        *((__global double *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_sub_with_mask_C2_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align ((dst_offset >> 1) & 1)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
-
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y);
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4   tmp = convert_int4_sat(src1_data) - src2_data;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 tmp_data = convert_uchar4_sat(tmp);
-
-        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
-        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C2_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort2 data = convert_ushort2_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
-
-        int2    tmp = convert_int2_sat(src_data1) - src_data2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short2 data = convert_short2_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global short2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C2_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
-        int2 src_data2 = (int2)(src2.x, src2.y);
-        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
-
-        long2 tmp = convert_long2_sat(src_data1) - convert_long2_sat(src_data2);
-        tmp = isMatSubScalar ? tmp : -tmp;
-        int2 data = convert_int2_sat(tmp);
-        data = mask_data ? data : dst_data;
-
-        *((__global int2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C2_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                            __global   float *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index));
-        float2 src_data2 = (float2)(src2.x, src2.y);
-        float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
-
-        float2 data = src_data1 - src_data2;
-        data = isMatSubScalar ? data : -data;
-        data = mask_data ? data : dst_data;
-
-        *((__global float2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_with_mask_C2_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                            __global   double *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index));
-        double2 src_data2 = (double2)(src2.x, src2.y);
-        double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
-
-        double2 data = src_data1 - src_data2;
-        data = isMatSubScalar ? data : -data;
-        data = mask_data ? data : dst_data;
-
-        *((__global double2 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
-__kernel void arithm_s_sub_with_mask_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x);
-        int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y);
-        int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        int4 tmp_0 = convert_int4_sat(src1_data_0) - src2_data_0;
-        int4 tmp_1 = convert_int4_sat(src1_data_1) - src2_data_1;
-        int4 tmp_2 = convert_int4_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(tmp_0);
-        uchar4 tmp_data_1 = convert_uchar4_sat(tmp_1);
-        uchar4 tmp_data_2 = convert_uchar4_sat(tmp_2);
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0;
-        int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1;
-        int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(tmp_0);
-        ushort2 tmp_data_1 = convert_ushort2_sat(tmp_1);
-        ushort2 tmp_data_2 = convert_ushort2_sat(tmp_2);
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0;
-        int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1;
-        int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        short2 tmp_data_0 = convert_short2_sat(tmp_0);
-        short2 tmp_data_1 = convert_short2_sat(tmp_1);
-        short2 tmp_data_2 = convert_short2_sat(tmp_2);
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        long tmp_0 = (long)src1_data_0 - (long)src2_data_0;
-        long tmp_1 = (long)src1_data_1 - (long)src2_data_1;
-        long tmp_2 = (long)src1_data_2 - (long)src2_data_2;
-
-        tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0;
-        tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1;
-        tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2;
-
-        int tmp_data_0 = convert_int_sat(tmp_0);
-        int tmp_data_1 = convert_int_sat(tmp_1);
-        int tmp_data_2 = convert_int_sat(tmp_2);
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C3_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                            __global   float *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = src2.x;
-        float src2_data_1 = src2.y;
-        float src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_data_0 = src1_data_0 - src2_data_0;
-        float tmp_data_1 = src1_data_1 - src2_data_1;
-        float tmp_data_2 = src1_data_2 - src2_data_2;
-
-        tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0;
-        tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1;
-        tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_with_mask_C3_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                            __global   double *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = src2.x;
-        double src2_data_1 = src2.y;
-        double src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 - src2_data_0;
-        double tmp_data_1 = src1_data_1 - src2_data_1;
-        double tmp_data_2 = src1_data_2 - src2_data_2;
-
-        tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0;
-        tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1;
-        tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-__kernel void arithm_s_sub_with_mask_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
-        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
-
-        int4 tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        uchar4 data = convert_uchar4_sat(tmp);
-
-        data = mask_data ? data : dst_data;
-
-        *((__global uchar4 *)(dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C4_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
-        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
-
-        int4    tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        ushort4 data = convert_ushort4_sat(tmp);
-
-        data = mask_data ? data : dst_data;
-
-        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C4_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
-        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
-
-        int4    tmp = convert_int4_sat(src_data1) - src2;
-        tmp = isMatSubScalar ? tmp : -tmp;
-        short4 data = convert_short4_sat(tmp);
-
-        data = mask_data ? data : dst_data;
-
-        *((__global short4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C4_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
-        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
-
-        long4 tmp = convert_long4_sat(src_data1) - convert_long4_sat(src2);
-        tmp = isMatSubScalar ? tmp : -tmp;
-        int4 data = convert_int4_sat(tmp);
-
-        data = mask_data ? data : dst_data;
-
-        *((__global int4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-__kernel void arithm_s_sub_with_mask_C4_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                            __global   float *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index));
-        float4 dst_data  = *((__global float4 *)((__global char *)dst  + dst_index));
-
-        float4 data = src_data1 - src2;
-        data = isMatSubScalar ? data : -data;
-
-        data = mask_data ? data : dst_data;
-
-        *((__global float4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_sub_with_mask_C4_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                            __global   double *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
-        int mask_index = mad24(y, mask_step,  x       + mask_offset);
-        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
-
-        uchar mask_data = *(mask + mask_index);
-
-        double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index));
-        double4 dst_data  = *((__global double4 *)((__global char *)dst  + dst_index));
-
-        double4 data = src_data1 - src2;
-        data = isMatSubScalar ? data : -data;
-        data = mask_data ? data : dst_data;
-
-        *((__global double4 *)((__global char *)dst + dst_index)) = data;
-    }
-}
-#endif
diff --git a/modules/ocl/src/kernels/brute_force_match.cl b/modules/ocl/src/kernels/brute_force_match.cl
deleted file mode 100644
index 0730ac5ac7..0000000000
--- a/modules/ocl/src/kernels/brute_force_match.cl
+++ /dev/null
@@ -1,865 +0,0 @@
-#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics:enable
-#define MAX_FLOAT 1e7f
-
-int bit1Count(float x)
-{
-    int c = 0;
-    int ix = (int)x;
-
-    for (int i = 0 ; i < 32 ; i++)
-    {
-        c += ix & 0x1;
-        ix >>= 1;
-    }
-
-    return (float)c;
-}
-/* 2dim launch, global size: dim0 is (query rows + block_size - 1) / block_size * block_size, dim1 is block_size
-local size: dim0 is block_size, dim1 is block_size.
-*/
-__kernel void BruteForceMatch_UnrollMatch(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global int *bestTrainIdx,
-    __global float *bestDistance,
-    __local float *sharebuffer,
-    int block_size,
-    int max_desc_len,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-
-    __local float *s_query = sharebuffer;
-    __local float *s_train = sharebuffer + block_size * max_desc_len;
-
-    int queryIdx = groupidx * block_size + lidy;
-
-    // load the query into local memory.
-    for (int i = 0 ;  i <  max_desc_len / block_size; i ++)
-    {
-        int loadx = lidx + i * block_size;
-        s_query[lidy * max_desc_len + loadx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-    }
-
-    float myBestDistance = MAX_FLOAT;
-    int myBestTrainIdx = -1;
-
-    // loopUnrolledCached to find the best trainIdx and best distance.
-    volatile int imgIdx = 0;
-
-    for (int t = 0 ; t < (train_rows + block_size - 1) / block_size ; t++)
-    {
-        float result = 0;
-
-        for (int i = 0 ; i < max_desc_len / block_size ; i++)
-        {
-            //load a block_size * block_size block into local train.
-            const int loadx = lidx + i * block_size;
-            s_train[lidx * block_size + lidy] = loadx < train_cols ? train[min(t * block_size + lidy, train_rows - 1) * (step / sizeof(float)) + loadx] : 0;
-
-            //synchronize to make sure each elem for reduceIteration in share memory is written already.
-            barrier(CLK_LOCAL_MEM_FENCE);
-
-            /* there are threee types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-            sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-            switch (distType)
-            {
-                case 0:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        result += fabs(s_query[lidy * max_desc_len + i * block_size + j] -  s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-                case 1:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        float qr = s_query[lidy * max_desc_len + i * block_size + j] -  s_train[j * block_size + lidx];
-                        result += qr * qr;
-                    }
-
-                    break;
-                case 2:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        //result += popcount((uint)s_query[lidy * max_desc_len + i * block_size + j] ^ (uint)s_train[j * block_size + lidx]);
-                        result += bit1Count((uint)s_query[lidy * max_desc_len + i * block_size + j] ^(uint)s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-        }
-
-        int trainIdx = t * block_size + lidx;
-
-        if (queryIdx < query_rows && trainIdx < train_rows && result < myBestDistance/* && mask(queryIdx, trainIdx)*/)
-        {
-            //bestImgIdx = imgIdx;
-            myBestDistance = result;
-            myBestTrainIdx = trainIdx;
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    __local float *s_distance = (__local float *)(sharebuffer);
-    __local int *s_trainIdx = (__local int *)(sharebuffer + block_size * block_size);
-
-    //find BestMatch
-    s_distance += lidy * block_size;
-    s_trainIdx += lidy * block_size;
-    s_distance[lidx] = myBestDistance;
-    s_trainIdx[lidx] = myBestTrainIdx;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    //reduce -- now all reduce implement in each threads.
-    for (int k = 0 ; k < block_size; k++)
-    {
-        if (myBestDistance > s_distance[k])
-        {
-            myBestDistance = s_distance[k];
-            myBestTrainIdx = s_trainIdx[k];
-        }
-    }
-
-    if (queryIdx < query_rows && lidx == 0)
-    {
-        bestTrainIdx[queryIdx] = myBestTrainIdx;
-        bestDistance[queryIdx] = myBestDistance;
-    }
-}
-
-__kernel void BruteForceMatch_Match(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global int *bestTrainIdx,
-    __global float *bestDistance,
-    __local float *sharebuffer,
-    int block_size,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-
-    const int queryIdx = groupidx * block_size + lidy;
-
-    float myBestDistance = MAX_FLOAT;
-    int myBestTrainIdx = -1;
-
-    __local float *s_query = sharebuffer;
-    __local float *s_train = sharebuffer + block_size * block_size;
-
-    // loop
-    for (int t = 0 ;  t < (train_rows + block_size - 1) / block_size ; t++)
-    {
-        //Dist dist;
-        float result = 0;
-
-        for (int i = 0 ; i < (query_cols + block_size - 1) / block_size ; i++)
-        {
-            const int loadx = lidx + i * block_size;
-            //load query and train into local memory
-            s_query[lidy * block_size + lidx] = 0;
-            s_train[lidx * block_size + lidy] = 0;
-
-            if (loadx < query_cols)
-            {
-                s_query[lidy * block_size + lidx] = query[min(queryIdx, query_rows - 1) * (step / sizeof(float)) + loadx];
-                s_train[lidx * block_size + lidy] = train[min(t * block_size + lidy, train_rows - 1) * (step / sizeof(float)) + loadx];
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-
-            /* there are threee types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-            sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-            switch (distType)
-            {
-                case 0:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        result += fabs(s_query[lidy * block_size + j] -  s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-                case 1:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        float qr = s_query[lidy * block_size + j] -  s_train[j * block_size + lidx];
-                        result += qr * qr;
-                    }
-
-                    break;
-                case 2:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        //result += popcount((uint)s_query[lidy * block_size + j] ^ (uint)s_train[j * block_size + lidx]);
-                        result += bit1Count((uint)s_query[lidy * block_size + j] ^(uint)s_train[(uint)j * block_size + lidx]);
-                    }
-
-                    break;
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-        }
-
-        const int trainIdx = t * block_size + lidx;
-
-        if (queryIdx < query_rows && trainIdx < train_rows && result < myBestDistance /*&& mask(queryIdx, trainIdx)*/)
-        {
-            //myBestImgidx = imgIdx;
-            myBestDistance = result;
-            myBestTrainIdx = trainIdx;
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    __local float *s_distance = (__local float *)sharebuffer;
-    __local int *s_trainIdx = (__local int *)(sharebuffer + block_size * block_size);
-
-    //findBestMatch
-    s_distance += lidy * block_size;
-    s_trainIdx += lidy * block_size;
-    s_distance[lidx] = myBestDistance;
-    s_trainIdx[lidx] = myBestTrainIdx;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    //reduce -- now all reduce implement in each threads.
-    for (int k = 0 ; k < block_size; k++)
-    {
-        if (myBestDistance > s_distance[k])
-        {
-            myBestDistance = s_distance[k];
-            myBestTrainIdx = s_trainIdx[k];
-        }
-    }
-
-    if (queryIdx < query_rows && lidx == 0)
-    {
-        bestTrainIdx[queryIdx] = myBestTrainIdx;
-        bestDistance[queryIdx] = myBestDistance;
-    }
-}
-
-//radius_unrollmatch
-__kernel void BruteForceMatch_RadiusUnrollMatch(
-    __global float *query,
-    __global float *train,
-    float maxDistance,
-    //__global float *mask,
-    __global int *bestTrainIdx,
-    __global float *bestDistance,
-    __global int *nMatches,
-    __local float *sharebuffer,
-    int block_size,
-    int max_desc_len,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int bestTrainIdx_cols,
-    int step,
-    int ostep,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-    const int groupidy = get_group_id(1);
-
-    const int queryIdx = groupidy * block_size + lidy;
-    const int trainIdx = groupidx * block_size + lidx;
-
-    __local float *s_query = sharebuffer;
-    __local float *s_train = sharebuffer + block_size * block_size;
-
-    float result = 0;
-
-    for (int i = 0 ; i < max_desc_len / block_size ; ++i)
-    {
-        //load a block_size * block_size block into local train.
-        const int loadx = lidx + i * block_size;
-
-        s_query[lidy * block_size + lidx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-        s_train[lidx * block_size + lidy] = loadx < query_cols ? train[min(groupidx * block_size + lidy, train_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-
-        //synchronize to make sure each elem for reduceIteration in share memory is written already.
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        /* there are three types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-        sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-        switch (distType)
-        {
-            case 0:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    result += fabs(s_query[lidy * block_size + j] - s_train[j * block_size + lidx]);
-                }
-
-                break;
-            case 1:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    float qr = s_query[lidy * block_size + j] - s_train[j * block_size + lidx];
-                    result += qr * qr;
-                }
-
-                break;
-            case 2:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    result += bit1Count((uint)s_query[lidy * block_size + j] ^(uint)s_train[j * block_size + lidx]);
-                }
-
-                break;
-        }
-
-        barrier(CLK_LOCAL_MEM_FENCE);
-    }
-
-    if (queryIdx < query_rows && trainIdx < train_rows && result < maxDistance/* && mask(queryIdx, trainIdx)*/)
-    {
-        unsigned int ind = atom_inc(nMatches + queryIdx/*, (unsigned int) -1*/);
-
-        if (ind < bestTrainIdx_cols)
-        {
-            //bestImgIdx = imgIdx;
-            bestTrainIdx[queryIdx * (ostep / sizeof(int)) + ind] = trainIdx;
-            bestDistance[queryIdx * (ostep / sizeof(float)) + ind] = result;
-        }
-    }
-}
-
-//radius_match
-__kernel void BruteForceMatch_RadiusMatch(
-    __global float *query,
-    __global float *train,
-    float maxDistance,
-    //__global float *mask,
-    __global int *bestTrainIdx,
-    __global float *bestDistance,
-    __global int *nMatches,
-    __local float *sharebuffer,
-    int block_size,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int bestTrainIdx_cols,
-    int step,
-    int ostep,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-    const int groupidy = get_group_id(1);
-
-    const int queryIdx = groupidy * block_size + lidy;
-    const int trainIdx = groupidx * block_size + lidx;
-
-    __local float *s_query = sharebuffer;
-    __local float *s_train = sharebuffer + block_size * block_size;
-
-    float result = 0;
-
-    for (int i = 0 ; i < (query_cols + block_size - 1) / block_size ; ++i)
-    {
-        //load a block_size * block_size block into local train.
-        const int loadx = lidx + i * block_size;
-
-        s_query[lidy * block_size + lidx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-        s_train[lidx * block_size + lidy] = loadx < query_cols ? train[min(groupidx * block_size + lidy, train_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-
-        //synchronize to make sure each elem for reduceIteration in share memory is written already.
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        /* there are three types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-        sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-        switch (distType)
-        {
-            case 0:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    result += fabs(s_query[lidy * block_size + j] - s_train[j * block_size + lidx]);
-                }
-
-                break;
-            case 1:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    float qr = s_query[lidy * block_size + j] - s_train[j * block_size + lidx];
-                    result += qr * qr;
-                }
-
-                break;
-            case 2:
-
-                for (int j = 0 ; j < block_size ; ++j)
-                {
-                    result += bit1Count((uint)s_query[lidy * block_size + j] ^(uint)s_train[j * block_size + lidx]);
-                }
-
-                break;
-        }
-
-        barrier(CLK_LOCAL_MEM_FENCE);
-    }
-
-    if (queryIdx < query_rows && trainIdx < train_rows && result < maxDistance/* && mask(queryIdx, trainIdx)*/)
-    {
-        unsigned int ind = atom_inc(nMatches + queryIdx/*, (unsigned int) -1*/);
-
-        if (ind < bestTrainIdx_cols)
-        {
-            //bestImgIdx = imgIdx;
-            bestTrainIdx[queryIdx * (ostep / sizeof(int)) + ind] = trainIdx;
-            bestDistance[queryIdx * (ostep / sizeof(float)) + ind] = result;
-        }
-    }
-}
-
-
-__kernel void BruteForceMatch_knnUnrollMatch(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global int2 *bestTrainIdx,
-    __global float2 *bestDistance,
-    __local float *sharebuffer,
-    int block_size,
-    int max_desc_len,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-
-    const int queryIdx = groupidx * block_size + lidy;
-    local float *s_query = sharebuffer;
-    local float *s_train = sharebuffer + block_size * max_desc_len;
-
-    // load the query into local memory.
-    for (int i = 0 ;  i <  max_desc_len / block_size; i ++)
-    {
-        int loadx = lidx + i * block_size;
-        s_query[lidy * max_desc_len + loadx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
-    }
-
-    float myBestDistance1 = MAX_FLOAT;
-    float myBestDistance2 = MAX_FLOAT;
-    int myBestTrainIdx1 = -1;
-    int myBestTrainIdx2 = -1;
-
-    //loopUnrolledCached
-    volatile int imgIdx = 0;
-
-    for (int t = 0 ; t < (train_rows + block_size - 1) / block_size ; t++)
-    {
-        float result = 0;
-
-        for (int i = 0 ; i < max_desc_len / block_size ; i++)
-        {
-            const int loadX = lidx + i * block_size;
-            //load a block_size * block_size block into local train.
-            const int loadx = lidx + i * block_size;
-            s_train[lidx * block_size + lidy] = loadx < train_cols ? train[min(t * block_size + lidy, train_rows - 1) * (step / sizeof(float)) + loadx] : 0;
-
-            //synchronize to make sure each elem for reduceIteration in share memory is written already.
-            barrier(CLK_LOCAL_MEM_FENCE);
-
-            /* there are threee types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-            sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-            switch (distType)
-            {
-                case 0:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        result += fabs(s_query[lidy * max_desc_len + i * block_size + j] -  s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-                case 1:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        float qr = s_query[lidy * max_desc_len + i * block_size + j] -  s_train[j * block_size + lidx];
-                        result += qr * qr;
-                    }
-
-                    break;
-                case 2:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        //result += popcount((uint)s_query[lidy * max_desc_len + i * block_size + j] ^ (uint)s_train[j * block_size + lidx]);
-                        result += bit1Count((uint)s_query[lidy * max_desc_len + i * block_size + j] ^(uint)s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-        }
-
-        const int trainIdx = t * block_size + lidx;
-
-        if (queryIdx < query_rows && trainIdx < train_rows)
-        {
-            if (result < myBestDistance1)
-            {
-                myBestDistance2 = myBestDistance1;
-                myBestTrainIdx2 = myBestTrainIdx1;
-                myBestDistance1 = result;
-                myBestTrainIdx1 = trainIdx;
-            }
-            else if (result < myBestDistance2)
-            {
-                myBestDistance2 = result;
-                myBestTrainIdx2 = trainIdx;
-            }
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    local float *s_distance = (local float *)sharebuffer;
-    local int *s_trainIdx = (local int *)(sharebuffer + block_size * block_size);
-
-    // find BestMatch
-    s_distance += lidy * block_size;
-    s_trainIdx += lidy * block_size;
-
-    s_distance[lidx] = myBestDistance1;
-    s_trainIdx[lidx] = myBestTrainIdx1;
-
-    float bestDistance1 = MAX_FLOAT;
-    float bestDistance2 = MAX_FLOAT;
-    int bestTrainIdx1 = -1;
-    int bestTrainIdx2 = -1;
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (lidx == 0)
-    {
-        for (int i = 0 ; i < block_size ; i++)
-        {
-            float val = s_distance[i];
-
-            if (val < bestDistance1)
-            {
-                bestDistance2 = bestDistance1;
-                bestTrainIdx2 = bestTrainIdx1;
-
-                bestDistance1 = val;
-                bestTrainIdx1 = s_trainIdx[i];
-            }
-            else if (val < bestDistance2)
-            {
-                bestDistance2 = val;
-                bestTrainIdx2 = s_trainIdx[i];
-            }
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    s_distance[lidx] = myBestDistance2;
-    s_trainIdx[lidx] = myBestTrainIdx2;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (lidx == 0)
-    {
-        for (int i = 0 ; i < block_size ; i++)
-        {
-            float val = s_distance[i];
-
-            if (val < bestDistance2)
-            {
-                bestDistance2 = val;
-                bestTrainIdx2 = s_trainIdx[i];
-            }
-        }
-    }
-
-    myBestDistance1 = bestDistance1;
-    myBestDistance2 = bestDistance2;
-
-    myBestTrainIdx1 = bestTrainIdx1;
-    myBestTrainIdx2 = bestTrainIdx2;
-
-    if (queryIdx < query_rows && lidx == 0)
-    {
-        bestTrainIdx[queryIdx] = (int2)(myBestTrainIdx1, myBestTrainIdx2);
-        bestDistance[queryIdx] = (float2)(myBestDistance1, myBestDistance2);
-    }
-}
-
-__kernel void BruteForceMatch_knnMatch(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global int2 *bestTrainIdx,
-    __global float2 *bestDistance,
-    __local float *sharebuffer,
-    int block_size,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType
-)
-{
-    const int lidx = get_local_id(0);
-    const int lidy = get_local_id(1);
-    const int groupidx = get_group_id(0);
-
-    const int queryIdx = groupidx * block_size + lidy;
-    local float *s_query = sharebuffer;
-    local float *s_train = sharebuffer + block_size * block_size;
-
-    float myBestDistance1 = MAX_FLOAT;
-    float myBestDistance2 = MAX_FLOAT;
-    int myBestTrainIdx1 = -1;
-    int myBestTrainIdx2 = -1;
-
-    //loop
-    for (int  t = 0 ; t < (train_rows + block_size - 1) / block_size ; t++)
-    {
-        float result = 0.0f;
-
-        for (int i = 0 ; i < (query_cols + block_size - 1) / block_size ; i++)
-        {
-            const int loadx = lidx + i * block_size;
-            //load query and train into local memory
-            s_query[lidy * block_size + lidx] = 0;
-            s_train[lidx * block_size + lidy] = 0;
-
-            if (loadx < query_cols)
-            {
-                s_query[lidy * block_size + lidx] = query[min(queryIdx, query_rows - 1) * (step / sizeof(float)) + loadx];
-                s_train[lidx * block_size + lidy] = train[min(t * block_size + lidy, train_rows - 1) * (step / sizeof(float)) + loadx];
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-
-            /* there are threee types in the reducer. the first is L1Dist, which to sum the abs(v1, v2), the second is L2Dist, which to
-            sum the (v1 - v2) * (v1 - v2), the third is humming, which to popc(v1 ^ v2), popc is to count the bits are set to 1*/
-
-            switch (distType)
-            {
-                case 0:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        result += fabs(s_query[lidy * block_size + j] -  s_train[j * block_size + lidx]);
-                    }
-
-                    break;
-                case 1:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        float qr = s_query[lidy * block_size + j] -  s_train[j * block_size + lidx];
-                        result += qr * qr;
-                    }
-
-                    break;
-                case 2:
-
-                    for (int j = 0 ; j < block_size ; j++)
-                    {
-                        //result += popcount((uint)s_query[lidy * block_size + j] ^ (uint)s_train[j * block_size + lidx]);
-                        result += bit1Count((uint)s_query[lidy * block_size + j] ^(uint)s_train[(uint)j * block_size + lidx]);
-                    }
-
-                    break;
-            }
-
-            barrier(CLK_LOCAL_MEM_FENCE);
-        }
-
-        const int trainIdx = t * block_size + lidx;
-
-        if (queryIdx < query_rows && trainIdx < train_rows /*&& mask(queryIdx, trainIdx)*/)
-        {
-            if (result < myBestDistance1)
-            {
-                myBestDistance2 = myBestDistance1;
-                myBestTrainIdx2 = myBestTrainIdx1;
-                myBestDistance1 = result;
-                myBestTrainIdx1 = trainIdx;
-            }
-            else if (result < myBestDistance2)
-            {
-                myBestDistance2 = result;
-                myBestTrainIdx2 = trainIdx;
-            }
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    __local float *s_distance = (__local float *)sharebuffer;
-    __local int *s_trainIdx = (__local int *)(sharebuffer + block_size * block_size);
-
-    //findBestMatch
-    s_distance += lidy * block_size;
-    s_trainIdx += lidy * block_size;
-
-    s_distance[lidx] = myBestDistance1;
-    s_trainIdx[lidx] = myBestTrainIdx1;
-
-    float bestDistance1 = MAX_FLOAT;
-    float bestDistance2 = MAX_FLOAT;
-    int bestTrainIdx1 = -1;
-    int bestTrainIdx2 = -1;
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (lidx == 0)
-    {
-        for (int i = 0 ; i < block_size ; i++)
-        {
-            float val = s_distance[i];
-
-            if (val < bestDistance1)
-            {
-                bestDistance2 = bestDistance1;
-                bestTrainIdx2 = bestTrainIdx1;
-
-                bestDistance1 = val;
-                bestTrainIdx1 = s_trainIdx[i];
-            }
-            else if (val < bestDistance2)
-            {
-                bestDistance2 = val;
-                bestTrainIdx2 = s_trainIdx[i];
-            }
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    s_distance[lidx] = myBestDistance2;
-    s_trainIdx[lidx] = myBestTrainIdx2;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (lidx == 0)
-    {
-        for (int i = 0 ; i < block_size ; i++)
-        {
-            float val = s_distance[i];
-
-            if (val < bestDistance2)
-            {
-                bestDistance2 = val;
-                bestTrainIdx2 = s_trainIdx[i];
-            }
-        }
-    }
-
-    myBestDistance1 = bestDistance1;
-    myBestDistance2 = bestDistance2;
-
-    myBestTrainIdx1 = bestTrainIdx1;
-    myBestTrainIdx2 = bestTrainIdx2;
-
-    if (queryIdx < query_rows && lidx == 0)
-    {
-        bestTrainIdx[queryIdx] = (int2)(myBestTrainIdx1, myBestTrainIdx2);
-        bestDistance[queryIdx] = (float2)(myBestDistance1, myBestDistance2);
-    }
-}
-
-kernel void BruteForceMatch_calcDistanceUnrolled(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global float *allDist,
-    __local float *sharebuffer,
-    int block_size,
-    int max_desc_len,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType)
-{
-    /* Todo */
-}
-
-kernel void BruteForceMatch_calcDistance(
-    __global float *query,
-    __global float *train,
-    //__global float *mask,
-    __global float *allDist,
-    __local float *sharebuffer,
-    int block_size,
-    int query_rows,
-    int query_cols,
-    int train_rows,
-    int train_cols,
-    int step,
-    int distType)
-{
-    /* Todo */
-}
-
-kernel void BruteForceMatch_findBestMatch(
-    __global float *allDist,
-    __global int *bestTrainIdx,
-    __global float *bestDistance,
-    int k,
-    int block_size
-)
-{
-    /* Todo */
-}
\ No newline at end of file
diff --git a/modules/ocl/src/kernels/filtering_laplacian.cl b/modules/ocl/src/kernels/filtering_laplacian.cl
deleted file mode 100644
index 27c15a788c..0000000000
--- a/modules/ocl/src/kernels/filtering_laplacian.cl
+++ /dev/null
@@ -1,531 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-#define BORDER_REFLECT_101
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////Macro for border type////////////////////////////////////////////
-/////////////////////////////////////////////////////////////////////////////////////////////////
-#ifdef BORDER_REPLICATE
-//BORDER_REPLICATE:     aaaaaa|abcdefgh|hhhhhhh
-#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? (l_edge)   : (i))
-#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? (r_edge)-1 : (addr))
-#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? (t_edge)   :(i))
-#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (b_edge)-1 :(addr))
-#endif
-
-#ifdef BORDER_REFLECT
-//BORDER_REFLECT:       fedcba|abcdefgh|hgfedcb
-#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? -(i)-1               : (i))
-#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-1+((r_edge)<<1) : (addr))
-#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? -(i)-1 : (i))
-#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-1+((b_edge)<<1) : (addr))
-#endif
-
-#ifdef BORDER_REFLECT_101
-//BORDER_REFLECT_101:   gfedcb|abcdefgh|gfedcba
-#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? -(i)                 : (i))
-#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-2+((r_edge)<<1) : (addr))
-#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? -(i)                 : (i))
-#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-2+((b_edge)<<1) : (addr))
-#endif
-
-#ifdef BORDER_WRAP
-//BORDER_WRAP:          cdefgh|abcdefgh|abcdefg
-#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? (i)+(r_edge) : (i))
-#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? (i)-(r_edge) : (addr))
-#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? (i)+(b_edge) : (i))
-#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (i)-(b_edge) : (addr))
-#endif
-
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////Macro for define elements number per thread/////////////////////////////
-////////////////////////////////////////////////////////////////////////////////////////////////////
-#define ANCHOR                  3
-#define ANX                     1
-#define ANY                     1
-
-#define ROWS_PER_GROUP          4
-#define ROWS_PER_GROUP_BITS     2
-#define ROWS_FETCH              (ROWS_PER_GROUP + ANY + ANY)   //(ROWS_PER_GROUP + anY * 2)
-
-#define THREADS_PER_ROW         64
-#define THREADS_PER_ROW_BIT     6
-
-#define ELEMENTS_PER_THREAD     4
-#define ELEMENTS_PER_THREAD_BIT 2
-
-#define LOCAL_MEM_STEP          260 //divup((get_local_size(0) + anX * 2), 4) * 4
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////////////8uC1////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////////////////////////////////////////////
-__kernel void filter2D_C1_D0(__global uchar *src, int src_step, int src_offset_x, int src_offset_y,
-                             __global uchar *dst, int dst_step, int dst_offset_x, int dst_offset_y,
-                             __constant int *mat_kernel __attribute__((max_constant_size (16384))),
-                             int cols,int rows, int operate_cols, int wholecols, int wholerows)
-{
-    int gX = get_global_id(0);
-    int gY = get_global_id(1);
-
-    int lX = get_local_id(0);
-
-    int groupX_size = get_local_size(0);
-    int groupX_id   = get_group_id(0);
-
-    #define dst_align (dst_offset_x & 3)
-    int cols_start_index_group = src_offset_x - dst_align + groupX_size * groupX_id - ANX;
-    int rows_start_index       = src_offset_y + (gY << ROWS_PER_GROUP_BITS) - ANY;
-
-    __local uchar local_data[LOCAL_MEM_STEP * ROWS_FETCH];
-    if((gY << 2) < rows)
-    {
-        for(int i = 0; i < ROWS_FETCH; ++i)
-        {
-            if((rows_start_index - src_offset_y) + i < rows + ANY)
-            {
-                #ifdef BORDER_CONSTANT
-                int selected_row  = rows_start_index + i;
-                int selected_cols = cols_start_index_group + lX;
-
-                uchar data = *(src + selected_row * src_step + selected_cols);
-                int con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                data = con ? data : 0;
-                local_data[i * LOCAL_MEM_STEP + lX ] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-
-                    data = *(src + selected_row * src_step + selected_cols);
-                    con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                    data = con ? data : 0;
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #else
-                int selected_row = ADDR_H(rows_start_index + i,  0, wholerows);
-                selected_row     = ADDR_B(rows_start_index + i, wholerows, selected_row);
-
-                int selected_cols = ADDR_L(cols_start_index_group + lX, 0, wholecols);
-                selected_cols     = ADDR_R(cols_start_index_group + lX, wholecols, selected_cols);
-
-                uchar data = *(src + selected_row * src_step + selected_cols);
-
-                local_data[i * LOCAL_MEM_STEP + lX ] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-                    selected_cols = ADDR_R(selected_cols, wholecols, selected_cols);
-
-                    data = *(src + selected_row * src_step + selected_cols);
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #endif
-            }
-        }
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    int process_col = groupX_size * groupX_id + ((lX % THREADS_PER_ROW) << 2);
-    if(((gY << 2) < rows) && (process_col < operate_cols))
-    {
-        int dst_cols_start = dst_offset_x;
-        int dst_cols_end   = dst_offset_x + cols;
-        int dst_cols_index = (dst_offset_x + process_col) & 0xfffffffc;
-
-        int dst_rows_end   = dst_offset_y + rows;
-        int dst_rows_index = dst_offset_y + (gY << ROWS_PER_GROUP_BITS) + (lX >> THREADS_PER_ROW_BIT);
-
-        uchar4 dst_data = *((__global uchar4 *)(dst + dst_rows_index * dst_step + dst_cols_index));
-
-        int4 sum = (int4)(0);
-        uchar4 data;
-
-        for(int i = 0; i < ANCHOR; i++)
-        {
-           #pragma unroll 3
-           for(int j = 0; j < ANCHOR; j++)
-           {
-                if(dst_rows_index < dst_rows_end)
-                {
-                     int local_row = (lX >> THREADS_PER_ROW_BIT) + i;
-                     int local_cols = ((lX % THREADS_PER_ROW) << ELEMENTS_PER_THREAD_BIT) + j;
-
-                     data = vload4(0, local_data+local_row * LOCAL_MEM_STEP + local_cols);
-                     sum = sum + (mat_kernel[i * ANCHOR + j] * convert_int4_sat(data));
-                 }
-            }
-        }
-
-        if(dst_rows_index < dst_rows_end)
-        {
-            sum.x = ((dst_cols_index + 0 >= dst_cols_start) && (dst_cols_index + 0 < dst_cols_end)) ? sum.x : dst_data.x;
-            sum.y = ((dst_cols_index + 1 >= dst_cols_start) && (dst_cols_index + 1 < dst_cols_end)) ? sum.y : dst_data.y;
-            sum.z = ((dst_cols_index + 2 >= dst_cols_start) && (dst_cols_index + 2 < dst_cols_end)) ? sum.z : dst_data.z;
-            sum.w = ((dst_cols_index + 3 >= dst_cols_start) && (dst_cols_index + 3 < dst_cols_end)) ? sum.w : dst_data.w;
-            *((__global uchar4 *)(dst + dst_rows_index * dst_step + dst_cols_index)) = convert_uchar4_sat(sum);
-        }
-   }
-}
-///////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////////////32FC1////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////////////////////////////////////////////
-__kernel void filter2D_C1_D5(__global float *src, int src_step, int src_offset_x, int src_offset_y,
-                             __global float *dst, int dst_step, int dst_offset_x, int dst_offset_y,
-                             __constant int *mat_kernel __attribute__((max_constant_size (16384))),
-                             int cols,int rows, int operate_cols, int wholecols, int wholerows)
-{
-    int gX = get_global_id(0);
-    int gY = get_global_id(1);
-
-    int lX = get_local_id(0);
-
-    int groupX_size = get_local_size(0);
-    int groupX_id   = get_group_id(0);
-
-    #define dst_align (dst_offset_x & 3)
-    int cols_start_index_group = src_offset_x - dst_align + groupX_size * groupX_id - ANX;
-    int rows_start_index       = src_offset_y + (gY << ROWS_PER_GROUP_BITS) - ANY;
-
-    __local float local_data[LOCAL_MEM_STEP * ROWS_FETCH];
-    if(((gY << 2) < rows))
-    {
-        for(int i = 0; i < ROWS_FETCH; ++i)
-        {
-            if((rows_start_index - src_offset_y) + i < rows + ANY)
-            {
-                #ifdef BORDER_CONSTANT
-                int selected_row  = rows_start_index + i;
-                int selected_cols = cols_start_index_group + lX;
-
-                float data = *((__global float *)((__global char *)src + selected_row * src_step + (selected_cols << 2)));
-                int con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                data = con ? data : 0;
-                local_data[i * LOCAL_MEM_STEP + lX ] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-
-                    data = *((__global float *)((__global char *)src + selected_row * src_step + (selected_cols << 2)));
-                    con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                    data = con ? data : 0;
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #else
-                int selected_row = ADDR_H(rows_start_index + i,  0, wholerows);
-                selected_row     = ADDR_B(rows_start_index + i, wholerows, selected_row);
-
-                int selected_cols = ADDR_L(cols_start_index_group + lX, 0, wholecols);
-                selected_cols     = ADDR_R(cols_start_index_group + lX, wholecols, selected_cols);
-
-                float data = *((__global float *)((__global char *)src + selected_row * src_step + (selected_cols << 2)));
-                local_data[i * LOCAL_MEM_STEP + lX] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-                    selected_cols = ADDR_R(selected_cols, wholecols, selected_cols);
-
-                    data = *((__global float *)((__global char *)src + selected_row * src_step + (selected_cols << 2)));
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #endif
-            }
-        }
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    int process_col = groupX_size * groupX_id + ((lX % THREADS_PER_ROW) << 2);
-    if(((gY << 2) < rows) && (process_col < operate_cols))
-    {
-        int dst_cols_start = dst_offset_x;
-        int dst_cols_end   = dst_offset_x + cols;
-        int dst_cols_index = (dst_offset_x + process_col) & 0xfffffffc;
-
-        int dst_rows_end   = dst_offset_y + rows;
-        int dst_rows_index = dst_offset_y + (gY << ROWS_PER_GROUP_BITS) + (lX >> THREADS_PER_ROW_BIT);
-
-        float4 dst_data = *((__global float4*)((__global char *)dst + dst_rows_index * dst_step + (dst_cols_index << 2)));
-
-        float4 sum = (float4)(0);
-        float4 data;
-
-        for(int i = 0; i < ANCHOR; i++)
-        {
-           #pragma unroll 3
-           for(int j = 0; j < ANCHOR; j++)
-           {
-                if(dst_rows_index < dst_rows_end)
-                {
-                     int local_row = (lX >> THREADS_PER_ROW_BIT) + i;
-                     int local_cols = ((lX % THREADS_PER_ROW) << ELEMENTS_PER_THREAD_BIT) + j;
-
-                     data = vload4(0, local_data+local_row * LOCAL_MEM_STEP + local_cols);
-                     sum = sum + (mat_kernel[i * ANCHOR + j] * data);
-                 }
-            }
-        }
-
-        if(dst_rows_index < dst_rows_end)
-        {
-            sum.x = ((dst_cols_index + 0 >= dst_cols_start) && (dst_cols_index + 0 < dst_cols_end)) ? sum.x : dst_data.x;
-            sum.y = ((dst_cols_index + 1 >= dst_cols_start) && (dst_cols_index + 1 < dst_cols_end)) ? sum.y : dst_data.y;
-            sum.z = ((dst_cols_index + 2 >= dst_cols_start) && (dst_cols_index + 2 < dst_cols_end)) ? sum.z : dst_data.z;
-            sum.w = ((dst_cols_index + 3 >= dst_cols_start) && (dst_cols_index + 3 < dst_cols_end)) ? sum.w : dst_data.w;
-
-            *((__global float4 *)((__global char *)dst + dst_rows_index * dst_step + (dst_cols_index << 2))) = sum;
-        }
-   }
-}
-
-///////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////////////8uC4////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////////////////////////////////////////////
-__kernel void filter2D_C4_D0(__global uchar4 *src, int src_step, int src_offset_x, int src_offset_y,
-                             __global uchar4 *dst, int dst_step, int dst_offset_x, int dst_offset_y,
-                             __constant int *mat_kernel __attribute__((max_constant_size (16384))),
-                             int cols,int rows, int operate_cols, int wholecols, int wholerows)
-{
-    int gX = get_global_id(0);
-    int gY = get_global_id(1);
-
-    int lX = get_local_id(0);
-
-    int groupX_size = get_local_size(0);
-    int groupX_id   = get_group_id(0);
-
-    #define dst_align (dst_offset_x & 3)
-    int cols_start_index_group = src_offset_x - dst_align + groupX_size * groupX_id - ANX;
-    int rows_start_index       = src_offset_y + (gY << ROWS_PER_GROUP_BITS) - ANY;
-
-    __local uchar4 local_data[LOCAL_MEM_STEP * ROWS_FETCH];
-
-    if(((gY << 2) < rows))
-    {
-        for(int i = 0; i < ROWS_FETCH; ++i)
-        {
-            if((rows_start_index - src_offset_y) + i < rows + ANY)
-            {
-                #ifdef BORDER_CONSTANT
-                int selected_row  = rows_start_index + i;
-                int selected_cols = cols_start_index_group + lX;
-
-                uchar4 data = *((__global uchar4*)((__global char*)src + selected_row * src_step + (selected_cols << 2)));
-                int con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                data = con ? data : 0;
-                local_data[i * LOCAL_MEM_STEP + lX ] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-
-                    data = *((__global uchar4*)((__global char*)src + selected_row * src_step + (selected_cols << 2)));
-                    con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                    data = con ? data : 0;
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #else
-                int selected_row = ADDR_H(rows_start_index + i,  0, wholerows);
-                selected_row     = ADDR_B(rows_start_index + i, wholerows, selected_row);
-
-                int selected_cols = ADDR_L(cols_start_index_group + lX, 0, wholecols);
-                selected_cols     = ADDR_R(cols_start_index_group + lX, wholecols, selected_cols);
-
-                uchar4 data = *((__global uchar4*)((__global char*)src + selected_row * src_step + (selected_cols << 2)));
-
-                local_data[i * LOCAL_MEM_STEP + lX] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-                    selected_cols = ADDR_R(selected_cols, wholecols, selected_cols);
-
-                    data = *((__global uchar4*)((__global char*)src + selected_row * src_step + (selected_cols << 2)));
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #endif
-            }
-        }
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    int process_col = groupX_size * groupX_id + ((lX % THREADS_PER_ROW) << 2);
-    if(((gY << 2) < rows) && (process_col < operate_cols))
-    {
-        int dst_cols_start = dst_offset_x;
-        int dst_cols_end   = dst_offset_x + cols;
-        int dst_cols_index = (dst_offset_x + process_col) & 0xfffffffc;
-
-        int dst_rows_end   = dst_offset_y + rows;
-        int dst_rows_index = dst_offset_y + (gY << ROWS_PER_GROUP_BITS) + (lX >> THREADS_PER_ROW_BIT);
-
-        uchar16 dst_data;
-        dst_data = *((__global uchar16*)((__global char *)dst + dst_rows_index * dst_step + (dst_cols_index << 2)));
-
-        int16 sum = (int16)(0);
-        uchar16 data;
-
-        for(int i = 0; i < ANCHOR; i++)
-        {
-           #pragma unroll 3
-           for(int j = 0; j < ANCHOR; j++)
-           {
-                if(dst_rows_index < dst_rows_end)
-                {
-                     int local_row = (lX >> THREADS_PER_ROW_BIT) + i;
-                     int local_cols = ((lX % THREADS_PER_ROW) << ELEMENTS_PER_THREAD_BIT) + j;
-
-                     data = vload16(0, (__local uchar *)(local_data+local_row * LOCAL_MEM_STEP + local_cols));
-                     sum = sum + (mat_kernel[i * ANCHOR + j] * convert_int16_sat(data));
-                 }
-            }
-        }
-
-        if(dst_rows_index < dst_rows_end)
-        {
-            uchar16 sum1 = convert_uchar16_sat(sum);
-            sum1.s0123 = ((dst_cols_index + 0 >= dst_cols_start) && (dst_cols_index + 0 < dst_cols_end))?
-                         sum1.s0123 : dst_data.s0123;
-            sum1.s4567 = ((dst_cols_index + 1 >= dst_cols_start) && (dst_cols_index + 1 < dst_cols_end))?
-                         sum1.s4567 : dst_data.s4567;
-            sum1.s89ab = ((dst_cols_index + 2 >= dst_cols_start) && (dst_cols_index + 2 < dst_cols_end))?
-                         sum1.s89ab : dst_data.s89ab;
-            sum1.scdef = ((dst_cols_index + 3 >= dst_cols_start) && (dst_cols_index + 3 < dst_cols_end))?
-                         sum1.scdef : dst_data.scdef;
-
-            *((__global uchar16*)((__global char *)dst + dst_rows_index * dst_step + (dst_cols_index << 2))) = sum1;
-        }
-    }
-}
-///////////////////////////////////////////////////////////////////////////////////////////////////
-/////////////////////////////////////////32FC4////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////////////////////////////////////////////
-#define ROWS_FETCH_C4              (1 + ANY + ANY)   //(ROWS_PER_GROUP + anY * 2)
-#define LOCAL_MEM_STEP_C4           260 //divup((get_local_size(0) + anX * 2), 4) * 4)
-__kernel void filter2D_C4_D5(__global float4 *src, int src_step, int src_offset_x, int src_offset_y,
-                             __global float4 *dst, int dst_step, int dst_offset_x, int dst_offset_y,
-                             __constant int *mat_kernel __attribute__((max_constant_size (16384))),
-                             int cols,int rows, int operate_cols, int wholecols, int wholerows)
-{
-    int gX = get_global_id(0);
-    int gY = get_global_id(1);
-
-    int lX = get_local_id(0);
-
-    int groupX_size = get_local_size(0);
-    int groupX_id   = get_group_id(0);
-
-    int cols_start_index_group = src_offset_x + groupX_size * groupX_id - ANX;
-    int rows_start_index       = src_offset_y + gY - ANY;
-
-    __local float4 local_data[LOCAL_MEM_STEP_C4 * ROWS_FETCH_C4];
-    if((gY < rows) && (gX < (operate_cols + ANX + ANX)))
-    {
-        for(int i = 0; i < ROWS_FETCH_C4; ++i)
-        {
-            if((rows_start_index - src_offset_y) + i < rows + ANY)
-            {
-                #ifdef BORDER_CONSTANT
-                int selected_row  = rows_start_index + i;
-                int selected_cols = cols_start_index_group + lX;
-
-                float4 data = *((__global float4*)((__global char*)src + selected_row * src_step + (selected_cols << 4)));
-                int con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                data = con ? data : 0;
-                local_data[i * LOCAL_MEM_STEP + lX ] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-
-                    data = *((__global float4*)((__global char*)src + selected_row * src_step + (selected_cols << 4)));
-                    con = selected_row >=0 && selected_row < wholerows && selected_cols >=0 && selected_cols < wholecols;
-                    data = con ? data : 0;
-                    local_data[i * LOCAL_MEM_STEP + lX + groupX_size] =data;
-                }
-                #else
-                int selected_row = ADDR_H(rows_start_index + i,  0, wholerows);
-                selected_row     = ADDR_B(rows_start_index + i, wholerows, selected_row);
-
-                int selected_cols = ADDR_L(cols_start_index_group + lX, 0, wholecols);
-                selected_cols     = ADDR_R(cols_start_index_group + lX, wholecols, selected_cols);
-
-                float4 data = *((__global float4*)((__global char*)src + selected_row * src_step + (selected_cols << 4)));
-                local_data[i * LOCAL_MEM_STEP_C4 + lX] =data;
-
-                if(lX < (ANX << 1))
-                {
-                    selected_cols = cols_start_index_group + lX + groupX_size;
-                    selected_cols = ADDR_R(selected_cols, wholecols, selected_cols);
-
-                    data = *((__global float4*)((__global char*)src + selected_row * src_step + (selected_cols << 4)));
-                    local_data[i * LOCAL_MEM_STEP_C4 + lX + groupX_size] =data;
-                }
-                #endif
-            }
-        }
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if((gY < rows) && (gX < operate_cols))
-    {
-        int dst_cols_index = dst_offset_x + gX;
-        int dst_rows_index = dst_offset_y + gY;
-
-        float4 sum = (float4)(0);
-
-        for(int i = 0; i < ANCHOR; i++)
-        {
-           for(int j = 0; j < ANCHOR; j++)
-           {
-               int local_cols = lX + j;
-               sum = sum + mat_kernel[i * ANCHOR + j] * local_data[i * LOCAL_MEM_STEP_C4 + local_cols];
-            }
-        }
-
-        *((__global float4*)((__global char *)dst + dst_rows_index * dst_step + (dst_cols_index << 4))) = sum;
-    }
-}
diff --git a/modules/ocl/src/kernels/haarobjectdetect.cl b/modules/ocl/src/kernels/haarobjectdetect.cl
deleted file mode 100644
index 7835b4bcc5..0000000000
--- a/modules/ocl/src/kernels/haarobjectdetect.cl
+++ /dev/null
@@ -1,565 +0,0 @@
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Niko Li, newlife20080214@gmail.com
-//    Wang Weiyan, wangweiyanster@gmail.com
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//    Nathan, liujun@multicorewareinc.com
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//
-
-#pragma OPENCL EXTENSION cl_amd_printf : enable
-#define CV_HAAR_FEATURE_MAX           3
-
-#define calc_sum(rect,offset)        (sum[(rect).p0+offset] - sum[(rect).p1+offset] - sum[(rect).p2+offset] + sum[(rect).p3+offset])
-#define calc_sum1(rect,offset,i)     (sum[(rect).p0[i]+offset] - sum[(rect).p1[i]+offset] - sum[(rect).p2[i]+offset] + sum[(rect).p3[i]+offset])
-
-typedef int   sumtype;
-typedef float sqsumtype;
-
-typedef struct  __attribute__((aligned (128)))  GpuHidHaarFeature
-{
-    struct __attribute__((aligned (32)))
-{
-    int p0 __attribute__((aligned (4)));
-    int p1 __attribute__((aligned (4)));
-    int p2 __attribute__((aligned (4)));
-    int p3 __attribute__((aligned (4)));
-    float weight __attribute__((aligned (4)));
-}
-rect[CV_HAAR_FEATURE_MAX] __attribute__((aligned (32)));
-}
-GpuHidHaarFeature;
-
-
-typedef struct __attribute__((aligned (128) )) GpuHidHaarTreeNode
-{
-    int p[CV_HAAR_FEATURE_MAX][4] __attribute__((aligned (64)));
-    float weight[CV_HAAR_FEATURE_MAX] /*__attribute__((aligned (16)))*/;
-    float threshold /*__attribute__((aligned (4)))*/;
-    float alpha[2] __attribute__((aligned (8)));
-    int left __attribute__((aligned (4)));
-    int right __attribute__((aligned (4)));
-}
-GpuHidHaarTreeNode;
-
-
-typedef struct __attribute__((aligned (32))) GpuHidHaarClassifier
-{
-    int count __attribute__((aligned (4)));
-    GpuHidHaarTreeNode* node __attribute__((aligned (8)));
-    float* alpha __attribute__((aligned (8)));
-}
-GpuHidHaarClassifier;
-
-
-typedef struct __attribute__((aligned (64))) GpuHidHaarStageClassifier
-{
-    int  count __attribute__((aligned (4)));
-    float threshold __attribute__((aligned (4)));
-    int two_rects __attribute__((aligned (4)));
-    int reserved0 __attribute__((aligned (8)));
-    int reserved1 __attribute__((aligned (8)));
-    int reserved2 __attribute__((aligned (8)));
-    int reserved3 __attribute__((aligned (8)));
-}
-GpuHidHaarStageClassifier;
-
-
-typedef struct __attribute__((aligned (64))) GpuHidHaarClassifierCascade
-{
-    int  count __attribute__((aligned (4)));
-    int  is_stump_based __attribute__((aligned (4)));
-    int  has_tilted_features __attribute__((aligned (4)));
-    int  is_tree __attribute__((aligned (4)));
-    int pq0 __attribute__((aligned (4)));
-    int pq1 __attribute__((aligned (4)));
-    int pq2 __attribute__((aligned (4)));
-    int pq3 __attribute__((aligned (4)));
-    int p0 __attribute__((aligned (4)));
-    int p1 __attribute__((aligned (4)));
-    int p2 __attribute__((aligned (4)));
-    int p3 __attribute__((aligned (4)));
-    float inv_window_area __attribute__((aligned (4)));
-} GpuHidHaarClassifierCascade;
-
-
-__kernel void __attribute__((reqd_work_group_size(8,8,1)))gpuRunHaarClassifierCascade(//constant GpuHidHaarClassifierCascade * cascade,
-    global GpuHidHaarStageClassifier * stagecascadeptr,
-    global int4 * info,
-    global GpuHidHaarTreeNode * nodeptr,
-    global const int * restrict sum1,
-    global const float * restrict sqsum1,
-    global int4 * candidate,
-    const int pixelstep,
-    const int loopcount,
-    const int start_stage,
-    const int split_stage,
-    const int end_stage,
-    const int startnode,
-    const int splitnode,
-    const int4 p,
-    const int4 pq,
-    const float correction
-    //const int width,
-    //const int height,
-    //const int grpnumperline,
-    //const int totalgrp
-)
-{
-    int grpszx = get_local_size(0);
-    int grpszy = get_local_size(1);
-    int grpnumx = get_num_groups(0);
-    int grpidx = get_group_id(0);
-    int lclidx = get_local_id(0);
-    int lclidy = get_local_id(1);
-
-    int lcl_sz = mul24(grpszx,grpszy);
-    int lcl_id = mad24(lclidy,grpszx,lclidx);
-
-    //assume lcl_sz == 256 or 128 or 64
-    //int lcl_sz_shift = (lcl_sz == 256) ? 8 : 7;
-    //lcl_sz_shift = (lcl_sz == 64) ? 6 : lcl_sz_shift;
-    __local int lclshare[1024];
-
-#define OFF 0
-    __local int* lcldata = lclshare + OFF;//for save win data
-    __local int* glboutindex = lcldata + 28*28;//for save global out index
-    __local int* lclcount = glboutindex + 1;//for save the numuber of temp pass pixel
-    __local int* lcloutindex = lclcount + 1;//for save info of temp pass pixel
-    __local float* partialsum = (__local float*)(lcloutindex + (lcl_sz<<1));
-    glboutindex[0]=0;
-    int outputoff = mul24(grpidx,256);
-
-    //assume window size is 20X20
-#define WINDOWSIZE 20+1
-    //make sure readwidth is the multiple of 4
-    //ystep =1, from host code
-    int readwidth = ((grpszx-1 + WINDOWSIZE+3)>>2)<<2;
-    int readheight = grpszy-1+WINDOWSIZE;
-    int read_horiz_cnt = readwidth >> 2;//each read int4
-    int total_read = mul24(read_horiz_cnt,readheight);
-    int read_loop = (total_read + lcl_sz - 1) >> 6;
-    candidate[outputoff+(lcl_id<<2)] = (int4)0;
-    candidate[outputoff+(lcl_id<<2)+1] = (int4)0;
-    candidate[outputoff+(lcl_id<<2)+2] = (int4)0;
-    candidate[outputoff+(lcl_id<<2)+3] = (int4)0;
-    for(int scalei = 0; scalei <loopcount; scalei++)
-    {
-        int4 scaleinfo1= info[scalei];
-        int width = (scaleinfo1.x & 0xffff0000) >> 16;
-        int height = scaleinfo1.x & 0xffff;
-        int grpnumperline =(scaleinfo1.y & 0xffff0000) >> 16;
-        int totalgrp = scaleinfo1.y & 0xffff;
-        int imgoff = scaleinfo1.z;
-        float factor = as_float(scaleinfo1.w);
-        //int ystep =1;// factor > 2.0 ? 1 : 2;
-
-        __global const int * sum = sum1 + imgoff;
-        __global const float * sqsum = sqsum1 + imgoff;
-        for(int grploop=grpidx; grploop<totalgrp; grploop+=grpnumx)
-        {
-            int grpidy = grploop / grpnumperline;
-            int grpidx = grploop - mul24(grpidy, grpnumperline);
-            int x = mad24(grpidx,grpszx,lclidx);
-            int y = mad24(grpidy,grpszy,lclidy);
-            //candidate_result.x = convert_int_rtn(x*factor);
-            //candidate_result.y = convert_int_rtn(y*factor);
-            int grpoffx = x-lclidx;
-            int grpoffy = y-lclidy;
-
-            for(int i=0; i<read_loop; i++)
-            {
-                int pos_id = mad24(i,lcl_sz,lcl_id);
-                pos_id = pos_id < total_read ? pos_id : 0;
-
-                int lcl_y = pos_id / read_horiz_cnt;
-                int lcl_x = pos_id - mul24(lcl_y, read_horiz_cnt);
-
-                int glb_x = grpoffx + (lcl_x<<2);
-                int glb_y = grpoffy + lcl_y;
-
-                int glb_off = mad24(glb_y,pixelstep,glb_x);
-                int4 data = *(__global int4*)&sum[glb_off];
-                int lcl_off = mad24(lcl_y, readwidth, lcl_x<<2);
-
-                lcldata[lcl_off] = data.x;
-                lcldata[lcl_off+1] = data.y;
-                lcldata[lcl_off+2] = data.z;
-                lcldata[lcl_off+3] = data.w;
-            }
-
-            lcloutindex[lcl_id] = 0;
-            lclcount[0] = 0;
-            int result = 1;
-            int nodecounter= startnode;
-            float mean, variance_norm_factor;
-            barrier(CLK_LOCAL_MEM_FENCE);
-
-            int lcl_off = mad24(lclidy,readwidth,lclidx);
-            int4 cascadeinfo1, cascadeinfo2;
-            cascadeinfo1 = p;
-            cascadeinfo2 = pq;// + mad24(y, pixelstep, x);
-
-
-            //if((x < width) && (y < height))
-            {
-                cascadeinfo1.x +=lcl_off;
-                cascadeinfo1.z +=lcl_off;
-                mean = (lcldata[mad24(cascadeinfo1.y,readwidth,cascadeinfo1.x)] - lcldata[mad24(cascadeinfo1.y,readwidth,cascadeinfo1.z)] -
-                        lcldata[mad24(cascadeinfo1.w,readwidth,cascadeinfo1.x)] + lcldata[mad24(cascadeinfo1.w,readwidth,cascadeinfo1.z)])
-                       *correction;
-
-                int p_offset = mad24(y, pixelstep, x);
-
-                cascadeinfo2.x +=p_offset;
-                cascadeinfo2.z +=p_offset;
-                variance_norm_factor =sqsum[mad24(cascadeinfo2.y, pixelstep, cascadeinfo2.x)] - sqsum[mad24(cascadeinfo2.y, pixelstep, cascadeinfo2.z)] -
-                                      sqsum[mad24(cascadeinfo2.w, pixelstep, cascadeinfo2.x)] + sqsum[mad24(cascadeinfo2.w, pixelstep, cascadeinfo2.z)];
-
-                variance_norm_factor = variance_norm_factor * correction - mean * mean;
-                variance_norm_factor = variance_norm_factor >=0.f ? sqrt(variance_norm_factor) : 1.f;
-                //if( cascade->is_stump_based )
-                //{
-                for(int stageloop = start_stage; (stageloop < split_stage)  && result; stageloop++ )
-                {
-                    float stage_sum = 0.f;
-                    int2 stageinfo = *(global int2*)(stagecascadeptr+stageloop);
-                    float stagethreshold = as_float(stageinfo.y);
-                    for(int nodeloop = 0; nodeloop < stageinfo.x; nodeloop++ )
-                    {
-                        __global GpuHidHaarTreeNode* currentnodeptr = (nodeptr + nodecounter);
-
-                        int4 info1 = *(__global int4*)(&(currentnodeptr->p[0][0]));
-                        int4 info2 = *(__global int4*)(&(currentnodeptr->p[1][0]));
-                        int4 info3 = *(__global int4*)(&(currentnodeptr->p[2][0]));
-                        float4 w = *(__global float4*)(&(currentnodeptr->weight[0]));
-                        float2 alpha2 = *(__global float2*)(&(currentnodeptr->alpha[0]));
-                        float nodethreshold  = w.w * variance_norm_factor;
-
-                        info1.x +=lcl_off;
-                        info1.z +=lcl_off;
-                        info2.x +=lcl_off;
-                        info2.z +=lcl_off;
-
-                        float classsum = (lcldata[mad24(info1.y,readwidth,info1.x)] - lcldata[mad24(info1.y,readwidth,info1.z)] -
-                                          lcldata[mad24(info1.w,readwidth,info1.x)] + lcldata[mad24(info1.w,readwidth,info1.z)]) * w.x;
-
-
-                        classsum += (lcldata[mad24(info2.y,readwidth,info2.x)] - lcldata[mad24(info2.y,readwidth,info2.z)] -
-                                     lcldata[mad24(info2.w,readwidth,info2.x)] + lcldata[mad24(info2.w,readwidth,info2.z)]) * w.y;
-
-
-                        //if((info3.z - info3.x) && (!stageinfo.z))
-                        //{
-                        info3.x +=lcl_off;
-                        info3.z +=lcl_off;
-                        classsum += (lcldata[mad24(info3.y,readwidth,info3.x)] - lcldata[mad24(info3.y,readwidth,info3.z)] -
-                                     lcldata[mad24(info3.w,readwidth,info3.x)] + lcldata[mad24(info3.w,readwidth,info3.z)]) * w.z;
-                        //}
-                        stage_sum += classsum >= nodethreshold ? alpha2.y : alpha2.x;
-                        nodecounter++;
-                    }
-
-                    result = (stage_sum >= stagethreshold);
-                }
-
-                if(result && (x < width) && (y < height))
-                {
-                    int queueindex = atomic_inc(lclcount);
-                    lcloutindex[queueindex<<1] = (lclidy << 16) | lclidx;
-                    lcloutindex[(queueindex<<1)+1] = as_int(variance_norm_factor);
-                }
-                barrier(CLK_LOCAL_MEM_FENCE);
-                int queuecount  = lclcount[0];
-                barrier(CLK_LOCAL_MEM_FENCE);
-                nodecounter = splitnode;
-                for(int stageloop = split_stage; stageloop< end_stage && queuecount>0; stageloop++)
-                {
-                    //barrier(CLK_LOCAL_MEM_FENCE);
-                    //if(lcl_id == 0)
-                    lclcount[0]=0;
-                    barrier(CLK_LOCAL_MEM_FENCE);
-
-                    int2 stageinfo = *(global int2*)(stagecascadeptr+stageloop);
-                    float stagethreshold = as_float(stageinfo.y);
-
-                    int perfscale = queuecount > 4 ? 3 : 2;
-                    int queuecount_loop = (queuecount + (1<<perfscale)-1) >> perfscale;
-                    int lcl_compute_win = lcl_sz >> perfscale;
-                    int lcl_compute_win_id = (lcl_id >>(6-perfscale));
-                    int lcl_loops = (stageinfo.x + lcl_compute_win -1) >> (6-perfscale);
-                    int lcl_compute_id = lcl_id - (lcl_compute_win_id << (6-perfscale));
-                    for(int queueloop=0; queueloop<queuecount_loop/* && lcl_compute_win_id < queuecount*/; queueloop++)
-                    {
-                        float stage_sum = 0.f;
-                        int temp_coord = lcloutindex[lcl_compute_win_id<<1];
-                        float variance_norm_factor = as_float(lcloutindex[(lcl_compute_win_id<<1)+1]);
-                        int queue_pixel = mad24(((temp_coord  & (int)0xffff0000)>>16),readwidth,temp_coord & 0xffff);
-
-                        //barrier(CLK_LOCAL_MEM_FENCE);
-                        if(lcl_compute_win_id < queuecount)
-                        {
-
-                            int tempnodecounter = lcl_compute_id;
-                            float part_sum = 0.f;
-                            for(int lcl_loop=0; lcl_loop<lcl_loops && tempnodecounter<stageinfo.x; lcl_loop++)
-                            {
-                                __global GpuHidHaarTreeNode* currentnodeptr = (nodeptr + nodecounter + tempnodecounter);
-
-                                int4 info1 = *(__global int4*)(&(currentnodeptr->p[0][0]));
-                                int4 info2 = *(__global int4*)(&(currentnodeptr->p[1][0]));
-                                int4 info3 = *(__global int4*)(&(currentnodeptr->p[2][0]));
-                                float4 w = *(__global float4*)(&(currentnodeptr->weight[0]));
-                                float2 alpha2 = *(__global float2*)(&(currentnodeptr->alpha[0]));
-                                float nodethreshold  = w.w * variance_norm_factor;
-
-                                info1.x +=queue_pixel;
-                                info1.z +=queue_pixel;
-                                info2.x +=queue_pixel;
-                                info2.z +=queue_pixel;
-
-                                float classsum = (lcldata[mad24(info1.y,readwidth,info1.x)] - lcldata[mad24(info1.y,readwidth,info1.z)] -
-                                                  lcldata[mad24(info1.w,readwidth,info1.x)] + lcldata[mad24(info1.w,readwidth,info1.z)]) * w.x;
-
-
-                                classsum += (lcldata[mad24(info2.y,readwidth,info2.x)] - lcldata[mad24(info2.y,readwidth,info2.z)] -
-                                             lcldata[mad24(info2.w,readwidth,info2.x)] + lcldata[mad24(info2.w,readwidth,info2.z)]) * w.y;
-                                //if((info3.z - info3.x) && (!stageinfo.z))
-                                //{
-                                info3.x +=queue_pixel;
-                                info3.z +=queue_pixel;
-                                classsum += (lcldata[mad24(info3.y,readwidth,info3.x)] - lcldata[mad24(info3.y,readwidth,info3.z)] -
-                                             lcldata[mad24(info3.w,readwidth,info3.x)] + lcldata[mad24(info3.w,readwidth,info3.z)]) * w.z;
-                                //}
-                                part_sum += classsum >= nodethreshold ? alpha2.y : alpha2.x;
-                                tempnodecounter +=lcl_compute_win;
-                            }//end for(int lcl_loop=0;lcl_loop<lcl_loops;lcl_loop++)
-                            partialsum[lcl_id]=part_sum;
-                        }
-                        barrier(CLK_LOCAL_MEM_FENCE);
-                        if(lcl_compute_win_id < queuecount)
-                        {
-                            for(int i=0; i<lcl_compute_win && (lcl_compute_id==0); i++)
-                            {
-                                stage_sum += partialsum[lcl_id+i];
-                            }
-                            if(stage_sum >= stagethreshold && (lcl_compute_id==0))
-                            {
-                                int queueindex = atomic_inc(lclcount);
-                                lcloutindex[queueindex<<1] = temp_coord;
-                                lcloutindex[(queueindex<<1)+1] = as_int(variance_norm_factor);
-                            }
-                            lcl_compute_win_id +=(1<<perfscale);
-                        }
-                        barrier(CLK_LOCAL_MEM_FENCE);
-                    }//end for(int queueloop=0;queueloop<queuecount_loop;queueloop++)
-                    //barrier(CLK_LOCAL_MEM_FENCE);
-                    queuecount = lclcount[0];
-                    barrier(CLK_LOCAL_MEM_FENCE);
-                    nodecounter += stageinfo.x;
-                }//end for(int stageloop = splitstage; stageloop< endstage && queuecount>0;stageloop++)
-                //barrier(CLK_LOCAL_MEM_FENCE);
-                if(lcl_id<queuecount)
-                {
-                    int temp = lcloutindex[lcl_id<<1];
-                    int x = mad24(grpidx,grpszx,temp & 0xffff);
-                    int y = mad24(grpidy,grpszy,((temp & (int)0xffff0000) >> 16));
-                    temp = glboutindex[0];
-                    int4 candidate_result;
-                    candidate_result.zw = (int2)convert_int_rtn(factor*20.f);
-                    candidate_result.x = convert_int_rtn(x*factor);
-                    candidate_result.y = convert_int_rtn(y*factor);
-                    atomic_inc(glboutindex);
-                    candidate[outputoff+temp+lcl_id] = candidate_result;
-                }
-                barrier(CLK_LOCAL_MEM_FENCE);
-            }//end if((x < width) && (y < height))
-        }//end for(int grploop=grpidx;grploop<totalgrp;grploop+=grpnumx)
-        //outputoff +=mul24(width,height);
-    }//end for(int scalei = 0; scalei <loopcount; scalei++)
-}
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-/*
-if(stagecascade->two_rects)
-{
-    #pragma unroll
-    for( n = 0; n < stagecascade->count; n++ )
-    {
-        t1 = *(node + counter);
-        t = t1.threshold * variance_norm_factor;
-        classsum = calc_sum1(t1,p_offset,0) * t1.weight[0];
-
-        classsum  += calc_sum1(t1, p_offset,1) * t1.weight[1];
-        stage_sum += classsum >= t ? t1.alpha[1]:t1.alpha[0];
-
-        counter++;
-    }
-}
-else
-{
-    #pragma unroll
-    for( n = 0; n < stagecascade->count; n++ )
-    {
-        t = node[counter].threshold*variance_norm_factor;
-        classsum = calc_sum1(node[counter],p_offset,0) * node[counter].weight[0];
-        classsum += calc_sum1(node[counter],p_offset,1) * node[counter].weight[1];
-
-        if( node[counter].p0[2] )
-            classsum += calc_sum1(node[counter],p_offset,2) * node[counter].weight[2];
-
-        stage_sum += classsum >= t ? node[counter].alpha[1]:node[counter].alpha[0];// modify
-
-        counter++;
-    }
-}
-*/
-/*
-__kernel void gpuRunHaarClassifierCascade_ScaleWindow(
-                          constant GpuHidHaarClassifierCascade * _cascade,
-                          global GpuHidHaarStageClassifier * stagecascadeptr,
-                          //global GpuHidHaarClassifier * classifierptr,
-                          global GpuHidHaarTreeNode * nodeptr,
-                          global int * sum,
-                          global float * sqsum,
-                          global int * _candidate,
-                          int pixel_step,
-                          int cols,
-                          int rows,
-                          int start_stage,
-                          int end_stage,
-                          //int counts,
-                          int nodenum,
-                          int ystep,
-                          int detect_width,
-                          //int detect_height,
-                          int loopcount,
-                          int outputstep)
-                          //float scalefactor)
-{
-unsigned int x1 = get_global_id(0);
-unsigned int y1 = get_global_id(1);
-int p_offset;
-int m, n;
-int result;
-int counter;
-float mean, variance_norm_factor;
-for(int i=0;i<loopcount;i++)
-{
-constant GpuHidHaarClassifierCascade * cascade = _cascade + i;
-global int * candidate = _candidate + i*outputstep;
-int window_width = cascade->p1 - cascade->p0;
-int window_height = window_width;
-result = 1;
-counter = 0;
-unsigned int x = mul24(x1,ystep);
-unsigned int y = mul24(y1,ystep);
-if((x < cols - window_width - 1) && (y < rows - window_height -1))
-{
-global GpuHidHaarStageClassifier *stagecascade = stagecascadeptr +cascade->count*i+ start_stage;
-//global GpuHidHaarClassifier      *classifier   = classifierptr;
-global GpuHidHaarTreeNode        *node         = nodeptr + nodenum*i;
-
-p_offset = mad24(y, pixel_step, x);// modify
-
-mean = (*(sum + p_offset + (int)cascade->p0) - *(sum + p_offset + (int)cascade->p1) -
-    *(sum + p_offset + (int)cascade->p2) + *(sum + p_offset + (int)cascade->p3))
-    *cascade->inv_window_area;
-
-variance_norm_factor = *(sqsum + p_offset + cascade->p0) - *(sqsum + cascade->p1 + p_offset) -
-                    *(sqsum + p_offset + cascade->p2) + *(sqsum + cascade->p3 + p_offset);
-variance_norm_factor = variance_norm_factor * cascade->inv_window_area - mean * mean;
-variance_norm_factor = variance_norm_factor >=0.f ? sqrt(variance_norm_factor) : 1;//modify
-
-// if( cascade->is_stump_based )
-//{
-for( m = start_stage; m < end_stage; m++ )
-{
-float stage_sum = 0.f;
-float t,  classsum;
-GpuHidHaarTreeNode t1;
-
-//#pragma unroll
-for( n = 0; n < stagecascade->count; n++ )
-{
-     t1 = *(node + counter);
-     t  = t1.threshold * variance_norm_factor;
-     classsum = calc_sum1(t1, p_offset ,0) * t1.weight[0] + calc_sum1(t1, p_offset ,1) * t1.weight[1];
-
-     if((t1.p0[2]) && (!stagecascade->two_rects))
-         classsum += calc_sum1(t1, p_offset, 2) * t1.weight[2];
-
-     stage_sum += classsum >= t ? t1.alpha[1] : t1.alpha[0];// modify
-     counter++;
-}
-
-if (stage_sum < stagecascade->threshold)
-{
-    result = 0;
-    break;
-}
-
-stagecascade++;
-
-}
-if(result)
-{
-    candidate[4 * (y1 * detect_width + x1)]     = x;
-    candidate[4 * (y1 * detect_width + x1) + 1] = y;
-    candidate[4 * (y1 * detect_width + x1)+2]     = window_width;
-    candidate[4 * (y1 * detect_width + x1) + 3] = window_height;
-}
-//}
-}
-}
-}
-*/
-
-
-
-
diff --git a/modules/ocl/src/kernels/haarobjectdetect_scaled2.cl b/modules/ocl/src/kernels/haarobjectdetect_scaled2.cl
deleted file mode 100644
index 22d3004e29..0000000000
--- a/modules/ocl/src/kernels/haarobjectdetect_scaled2.cl
+++ /dev/null
@@ -1,286 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Wu Xinglong, wxl370@126.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-// Enter your kernel in this window
-//#pragma OPENCL EXTENSION cl_amd_printf:enable
-#define CV_HAAR_FEATURE_MAX           3
-typedef int   sumtype;
-typedef float sqsumtype;
-typedef struct  __attribute__((aligned(128)))  GpuHidHaarFeature
-{
-    struct __attribute__((aligned(32)))
-{
-        int p0 __attribute__((aligned(4)));
-        int p1 __attribute__((aligned(4)));
-        int p2 __attribute__((aligned(4)));
-        int p3 __attribute__((aligned(4)));
-        float weight __attribute__((aligned(4)));
-}
-rect[CV_HAAR_FEATURE_MAX] __attribute__((aligned(32)));
-}
-GpuHidHaarFeature;
-typedef struct __attribute__((aligned(128))) GpuHidHaarTreeNode
-{
-    int p[CV_HAAR_FEATURE_MAX][4] __attribute__((aligned(64)));
-    float weight[CV_HAAR_FEATURE_MAX] /*__attribute__((aligned (16)))*/;
-    float threshold /*__attribute__((aligned (4)))*/;
-    float alpha[2] __attribute__((aligned(8)));
-    int left __attribute__((aligned(4)));
-    int right __attribute__((aligned(4)));
-}
-GpuHidHaarTreeNode;
-typedef struct __attribute__((aligned(32))) GpuHidHaarClassifier
-{
-    int count __attribute__((aligned(4)));
-    GpuHidHaarTreeNode *node __attribute__((aligned(8)));
-    float *alpha __attribute__((aligned(8)));
-}
-GpuHidHaarClassifier;
-typedef struct __attribute__((aligned(64))) GpuHidHaarStageClassifier
-{
-    int  count __attribute__((aligned(4)));
-    float threshold __attribute__((aligned(4)));
-    int two_rects __attribute__((aligned(4)));
-    int reserved0 __attribute__((aligned(8)));
-    int reserved1 __attribute__((aligned(8)));
-    int reserved2 __attribute__((aligned(8)));
-    int reserved3 __attribute__((aligned(8)));
-}
-GpuHidHaarStageClassifier;
-typedef struct __attribute__((aligned(64))) GpuHidHaarClassifierCascade
-{
-    int  count __attribute__((aligned(4)));
-    int  is_stump_based __attribute__((aligned(4)));
-    int  has_tilted_features __attribute__((aligned(4)));
-    int  is_tree __attribute__((aligned(4)));
-    int pq0 __attribute__((aligned(4)));
-    int pq1 __attribute__((aligned(4)));
-    int pq2 __attribute__((aligned(4)));
-    int pq3 __attribute__((aligned(4)));
-    int p0 __attribute__((aligned(4)));
-    int p1 __attribute__((aligned(4)));
-    int p2 __attribute__((aligned(4)));
-    int p3 __attribute__((aligned(4)));
-    float inv_window_area __attribute__((aligned(4)));
-} GpuHidHaarClassifierCascade;
-
-__kernel void gpuRunHaarClassifierCascade_scaled2(
-    global GpuHidHaarStageClassifier *stagecascadeptr,
-    global int4 *info,
-    global GpuHidHaarTreeNode *nodeptr,
-    global const int *restrict sum,
-    global const float   *restrict sqsum,
-    global int4 *candidate,
-    const int step,
-    const int loopcount,
-    const int start_stage,
-    const int split_stage,
-    const int end_stage,
-    const int startnode,
-    const int splitnode,
-    global int4 *p,
-    //const int4 * pq,
-    global float *correction,
-    const int nodecount)
-{
-        int grpszx = get_local_size(0);
-        int grpszy = get_local_size(1);
-        int grpnumx = get_num_groups(0);
-        int grpidx = get_group_id(0);
-        int lclidx = get_local_id(0);
-        int lclidy = get_local_id(1);
-        int lcl_sz = mul24(grpszx, grpszy);
-        int lcl_id = mad24(lclidy, grpszx, lclidx);
-        __local int lclshare[1024];
-        __local int *glboutindex = lclshare + 0;
-        __local int *lclcount = glboutindex + 1;
-        __local int *lcloutindex = lclcount + 1;
-        __local float *partialsum = (__local float *)(lcloutindex + (lcl_sz << 1));
-        glboutindex[0] = 0;
-        int outputoff = mul24(grpidx, 256);
-        candidate[outputoff + (lcl_id << 2)] = (int4)0;
-        candidate[outputoff + (lcl_id << 2) + 1] = (int4)0;
-        candidate[outputoff + (lcl_id << 2) + 2] = (int4)0;
-        candidate[outputoff + (lcl_id << 2) + 3] = (int4)0;
-
-        for (int scalei = 0; scalei < loopcount; scalei++)
-        {
-                int4 scaleinfo1;
-                scaleinfo1 = info[scalei];
-                int width = (scaleinfo1.x & 0xffff0000) >> 16;
-                int height = scaleinfo1.x & 0xffff;
-                int grpnumperline = (scaleinfo1.y & 0xffff0000) >> 16;
-                int totalgrp = scaleinfo1.y & 0xffff;
-                float factor = as_float(scaleinfo1.w);
-                float correction_t = correction[scalei];
-                int ystep = (int)(max(2.0f, factor) + 0.5f);
-
-                for (int grploop = get_group_id(0); grploop < totalgrp; grploop += grpnumx)
-                {
-                        int4 cascadeinfo = p[scalei];
-                        int grpidy = grploop / grpnumperline;
-                        int grpidx = grploop - mul24(grpidy, grpnumperline);
-                        int ix = mad24(grpidx, grpszx, lclidx);
-                        int iy = mad24(grpidy, grpszy, lclidy);
-                        int x = ix * ystep;
-                        int y = iy * ystep;
-                        lcloutindex[lcl_id] = 0;
-                        lclcount[0] = 0;
-                        int result = 1, nodecounter;
-                        float mean, variance_norm_factor;
-                        //if((ix < width) && (iy < height))
-                        {
-                                const int p_offset = mad24(y, step, x);
-                                cascadeinfo.x += p_offset;
-                                cascadeinfo.z += p_offset;
-                                mean = (sum[mad24(cascadeinfo.y, step, cascadeinfo.x)] - sum[mad24(cascadeinfo.y, step, cascadeinfo.z)] -
-                                        sum[mad24(cascadeinfo.w, step, cascadeinfo.x)] + sum[mad24(cascadeinfo.w, step, cascadeinfo.z)])
-                                       * correction_t;
-                                variance_norm_factor = sqsum[mad24(cascadeinfo.y, step, cascadeinfo.x)] - sqsum[mad24(cascadeinfo.y, step, cascadeinfo.z)] -
-                                                       sqsum[mad24(cascadeinfo.w, step, cascadeinfo.x)] + sqsum[mad24(cascadeinfo.w, step, cascadeinfo.z)];
-                                variance_norm_factor = variance_norm_factor * correction_t - mean * mean;
-                                variance_norm_factor = variance_norm_factor >= 0.f ? sqrt(variance_norm_factor) : 1.f;
-                                result = 1;
-                                nodecounter = startnode + nodecount * scalei;
-
-                                for (int stageloop = start_stage; stageloop < end_stage && result; stageloop++)
-                                {
-                                        float stage_sum = 0.f;
-                                        int4 stageinfo = *(global int4 *)(stagecascadeptr + stageloop);
-                                        float stagethreshold = as_float(stageinfo.y);
-
-                                        for (int nodeloop = 0; nodeloop < stageinfo.x; nodeloop++)
-                                        {
-                                                __global GpuHidHaarTreeNode *currentnodeptr = (nodeptr + nodecounter);
-                                                int4 info1 = *(__global int4 *)(&(currentnodeptr->p[0][0]));
-                                                int4 info2 = *(__global int4 *)(&(currentnodeptr->p[1][0]));
-                                                int4 info3 = *(__global int4 *)(&(currentnodeptr->p[2][0]));
-                                                float4 w = *(__global float4 *)(&(currentnodeptr->weight[0]));
-                                                float2 alpha2 = *(__global float2 *)(&(currentnodeptr->alpha[0]));
-                                                float nodethreshold  = w.w * variance_norm_factor;
-                                                info1.x += p_offset;
-                                                info1.z += p_offset;
-                                                info2.x += p_offset;
-                                                info2.z += p_offset;
-                                                float classsum = (sum[mad24(info1.y, step, info1.x)] - sum[mad24(info1.y, step, info1.z)] -
-                                                                  sum[mad24(info1.w, step, info1.x)] + sum[mad24(info1.w, step, info1.z)]) * w.x;
-                                                classsum += (sum[mad24(info2.y, step, info2.x)] - sum[mad24(info2.y, step, info2.z)] -
-                                                             sum[mad24(info2.w, step, info2.x)] + sum[mad24(info2.w, step, info2.z)]) * w.y;
-                                                info3.x += p_offset;
-                                                info3.z += p_offset;
-                                                classsum += (sum[mad24(info3.y, step, info3.x)] - sum[mad24(info3.y, step, info3.z)] -
-                                                             sum[mad24(info3.w, step, info3.x)] + sum[mad24(info3.w, step, info3.z)]) * w.z;
-                                                stage_sum += classsum >= nodethreshold ? alpha2.y : alpha2.x;
-                                                nodecounter++;
-                                        }
-
-                                        result = (stage_sum >= stagethreshold);
-                                }
-
-                                if (result && (ix < width) && (iy < height))
-                                {
-                                        int queueindex = atomic_inc(lclcount);
-                                        lcloutindex[queueindex << 1] = (y << 16) | x;
-                                        lcloutindex[(queueindex << 1) + 1] = as_int(variance_norm_factor);
-                                }
-
-                                barrier(CLK_LOCAL_MEM_FENCE);
-                                int queuecount = lclcount[0];
-                                nodecounter = splitnode + nodecount * scalei;
-
-                                if (lcl_id < queuecount)
-                                {
-                                        int temp = lcloutindex[lcl_id << 1];
-                                        int x = temp & 0xffff;
-                                        int y = (temp & (int)0xffff0000) >> 16;
-                                        temp = glboutindex[0];
-                                        int4 candidate_result;
-                                        candidate_result.zw = (int2)convert_int_rtn(factor * 20.f);
-                                        candidate_result.x = x;
-                                        candidate_result.y = y;
-                                        atomic_inc(glboutindex);
-                                        candidate[outputoff + temp + lcl_id] = candidate_result;
-                                }
-
-                                barrier(CLK_LOCAL_MEM_FENCE);
-                        }
-                }
-        }
-}
-__kernel void gpuscaleclassifier(global GpuHidHaarTreeNode *orinode, global GpuHidHaarTreeNode *newnode, float scale, float weight_scale, int nodenum)
-{
-        int counter = get_global_id(0);
-        int tr_x[3], tr_y[3], tr_h[3], tr_w[3], i = 0;
-        GpuHidHaarTreeNode t1 = *(orinode + counter);
-#pragma unroll
-
-        for (i = 0; i < 3; i++)
-        {
-                tr_x[i] = (int)(t1.p[i][0] * scale + 0.5f);
-                tr_y[i] = (int)(t1.p[i][1] * scale + 0.5f);
-                tr_w[i] = (int)(t1.p[i][2] * scale + 0.5f);
-                tr_h[i] = (int)(t1.p[i][3] * scale + 0.5f);
-        }
-
-        t1.weight[0] = t1.p[2][0] ? -(t1.weight[1] * tr_h[1] * tr_w[1] + t1.weight[2] * tr_h[2] * tr_w[2]) / (tr_h[0] * tr_w[0]) : -t1.weight[1] * tr_h[1] * tr_w[1] / (tr_h[0] * tr_w[0]);
-        counter += nodenum;
-#pragma unroll
-
-        for (i = 0; i < 3; i++)
-        {
-                newnode[counter].p[i][0] = tr_x[i];
-                newnode[counter].p[i][1] = tr_y[i];
-                newnode[counter].p[i][2] = tr_x[i] + tr_w[i];
-                newnode[counter].p[i][3] = tr_y[i] + tr_h[i];
-                newnode[counter].weight[i] = t1.weight[i] * weight_scale;
-        }
-
-        newnode[counter].left = t1.left;
-        newnode[counter].right = t1.right;
-        newnode[counter].threshold = t1.threshold;
-        newnode[counter].alpha[0] = t1.alpha[0];
-        newnode[counter].alpha[1] = t1.alpha[1];
-}
-
diff --git a/modules/ocl/src/kernels/objdetect_hog.cl b/modules/ocl/src/kernels/objdetect_hog.cl
deleted file mode 100644
index db11ed1410..0000000000
--- a/modules/ocl/src/kernels/objdetect_hog.cl
+++ /dev/null
@@ -1,498 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Wenju He, wenju@multicorewareinc.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-
-#define CELL_WIDTH 8
-#define CELL_HEIGHT 8
-#define CELLS_PER_BLOCK_X 2
-#define CELLS_PER_BLOCK_Y 2
-#define NTHREADS 256
-#define CV_PI_F 3.1415926535897932384626433832795f
-
-//----------------------------------------------------------------------------
-// Histogram computation
-
-__kernel void compute_hists_kernel(const int width, const int cblock_stride_x, const int cblock_stride_y,
-                                   const int cnbins, const int cblock_hist_size, const int img_block_width,
-                                   const int grad_quadstep, const int qangle_step,
-                                   __global const float* grad, __global const uchar* qangle,
-                                   const float scale, __global float* block_hists, __local float* smem)
-{
-    const int lidX = get_local_id(0);
-    const int lidY = get_local_id(1);
-    const int gidX = get_group_id(0);
-    const int gidY = get_group_id(1);
-
-    const int cell_x = lidX / 16;
-    const int cell_y = lidY;
-    const int cell_thread_x = lidX & 0xF;
-
-    __local float* hists = smem;
-    __local float* final_hist = smem + cnbins * 48;
-
-    const int offset_x = gidX * cblock_stride_x + (cell_x << 2) + cell_thread_x;
-    const int offset_y = gidY * cblock_stride_y + (cell_y << 2);
-
-    __global const float* grad_ptr = grad + offset_y * grad_quadstep + (offset_x << 1);
-    __global const uchar* qangle_ptr = qangle + offset_y * qangle_step + (offset_x << 1);
-
-    // 12 means that 12 pixels affect on block's cell (in one row)
-    if (cell_thread_x < 12)
-    {
-        __local float* hist = hists + 12 * (cell_y * CELLS_PER_BLOCK_Y + cell_x) + cell_thread_x;
-        for (int bin_id = 0; bin_id < cnbins; ++bin_id)
-            hist[bin_id * 48] = 0.f;
-
-        const int dist_x = -4 + cell_thread_x - 4 * cell_x;
-
-        const int dist_y_begin = -4 - 4 * lidY;
-        for (int dist_y = dist_y_begin; dist_y < dist_y_begin + 12; ++dist_y)
-        {
-            float2 vote = (float2) (grad_ptr[0], grad_ptr[1]);
-            uchar2 bin = (uchar2) (qangle_ptr[0], qangle_ptr[1]);
-
-            grad_ptr += grad_quadstep;
-            qangle_ptr += qangle_step;
-
-            int dist_center_y = dist_y - 4 * (1 - 2 * cell_y);
-            int dist_center_x = dist_x - 4 * (1 - 2 * cell_x);
-
-            float gaussian = exp(-(dist_center_y * dist_center_y + dist_center_x * dist_center_x) * scale);
-            float interp_weight = (8.f - fabs(dist_y + 0.5f)) * (8.f - fabs(dist_x + 0.5f)) / 64.f;
-
-            hist[bin.x * 48] += gaussian * interp_weight * vote.x;
-            hist[bin.y * 48] += gaussian * interp_weight * vote.y;
-        }
-
-        volatile __local float* hist_ = hist;
-        for (int bin_id = 0; bin_id < cnbins; ++bin_id, hist_ += 48)
-        {
-            if (cell_thread_x < 6) hist_[0] += hist_[6];
-            if (cell_thread_x < 3) hist_[0] += hist_[3];
-            if (cell_thread_x == 0)
-                final_hist[(cell_x * 2 + cell_y) * cnbins + bin_id] = hist_[0] + hist_[1] + hist_[2];
-        }
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    __global float* block_hist = block_hists + (gidY * img_block_width + gidX) * cblock_hist_size;
-
-    int tid = (cell_y * CELLS_PER_BLOCK_Y + cell_x) * 16 + cell_thread_x;
-    if (tid < cblock_hist_size)
-        block_hist[tid] = final_hist[tid];
-}
-
-//-------------------------------------------------------------
-//  Normalization of histograms via L2Hys_norm
-//
-float reduce_smem(volatile __local float* smem, int size)
-{
-    unsigned int tid = get_local_id(0);
-    float sum = smem[tid];
-
-    if (size >= 512) { if (tid < 256) smem[tid] = sum = sum + smem[tid + 256]; barrier(CLK_LOCAL_MEM_FENCE); }
-    if (size >= 256) { if (tid < 128) smem[tid] = sum = sum + smem[tid + 128]; barrier(CLK_LOCAL_MEM_FENCE); }
-    if (size >= 128) { if (tid < 64) smem[tid] = sum = sum + smem[tid + 64]; barrier(CLK_LOCAL_MEM_FENCE); }
-
-    if (tid < 32)
-    {
-        if (size >= 64) smem[tid] = sum = sum + smem[tid + 32];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-    if (tid < 16)
-    {
-        if (size >= 32) smem[tid] = sum = sum + smem[tid + 16];
-        if (size >= 16) smem[tid] = sum = sum + smem[tid + 8];
-        if (size >= 8) smem[tid] = sum = sum + smem[tid + 4];
-        if (size >= 4) smem[tid] = sum = sum + smem[tid + 2];
-        if (size >= 2) smem[tid] = sum = sum + smem[tid + 1];
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    sum = smem[0];
-
-    return sum;
-}
-
-__kernel void normalize_hists_kernel(const int nthreads, const int block_hist_size, const int img_block_width,
-                                     __global float* block_hists, const float threshold, __local float *squares)
-{
-    const int tid = get_local_id(0);
-    const int gidX = get_group_id(0);
-    const int gidY = get_group_id(1);
-
-    __global float* hist = block_hists + (gidY * img_block_width + gidX) * block_hist_size + tid;
-
-    float elem = 0.f;
-    if (tid < block_hist_size)
-        elem = hist[0];
-
-    squares[tid] = elem * elem;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    float sum = reduce_smem(squares, nthreads);
-
-    float scale = 1.0f / (sqrt(sum) + 0.1f * block_hist_size);
-    elem = min(elem * scale, threshold);
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    squares[tid] = elem * elem;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    sum = reduce_smem(squares, nthreads);
-    scale = 1.0f / (sqrt(sum) + 1e-3f);
-
-    if (tid < block_hist_size)
-        hist[0] = elem * scale;
-}
-
-//---------------------------------------------------------------------
-//  Linear SVM based classification
-//
-__kernel void classify_hists_kernel(const int cblock_hist_size, const int cdescr_size, const int cdescr_width,
-                                    const int img_win_width, const int img_block_width,
-                                    const int win_block_stride_x, const int win_block_stride_y,
-                                    __global const float * block_hists, __global const float* coefs,
-                                    float free_coef, float threshold, __global uchar* labels)
-{
-    const int tid = get_local_id(0);
-    const int gidX = get_group_id(0);
-    const int gidY = get_group_id(1);
-
-    __global const float* hist = block_hists + (gidY * win_block_stride_y * img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
-
-    float product = 0.f;
-    for (int i = tid; i < cdescr_size; i += NTHREADS)
-    {
-        int offset_y = i / cdescr_width;
-        int offset_x = i - offset_y * cdescr_width;
-        product += coefs[i] * hist[offset_y * img_block_width * cblock_hist_size + offset_x];
-    }
-
-    __local float products[NTHREADS];
-
-    products[tid] = product;
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (tid < 128) products[tid] = product = product + products[tid + 128];
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (tid < 64) products[tid] = product = product + products[tid + 64];
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if (tid < 32)
-    {
-        volatile __local float* smem = products;
-        smem[tid] = product = product + smem[tid + 32];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-    if (tid < 16)
-    {
-        volatile __local float* smem = products;
-        smem[tid] = product = product + smem[tid + 16];
-        smem[tid] = product = product + smem[tid + 8];
-        smem[tid] = product = product + smem[tid + 4];
-        smem[tid] = product = product + smem[tid + 2];
-        smem[tid] = product = product + smem[tid + 1];
-    }
-
-    if (tid == 0)
-        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
-}
-
-//----------------------------------------------------------------------------
-// Extract descriptors
-
-__kernel void extract_descrs_by_rows_kernel(const int cblock_hist_size, const int descriptors_quadstep, const int cdescr_size, const int cdescr_width,
-                                            const int img_block_width, const int win_block_stride_x, const int win_block_stride_y,
-                                            __global const float* block_hists, __global float* descriptors)
-{
-    int tid = get_local_id(0);
-    int gidX = get_group_id(0);
-    int gidY = get_group_id(1);
-
-    // Get left top corner of the window in src
-    __global const float* hist = block_hists + (gidY * win_block_stride_y * img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
-
-    // Get left top corner of the window in dst
-    __global float* descriptor = descriptors + (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
-
-    // Copy elements from src to dst
-    for (int i = tid; i < cdescr_size; i += NTHREADS)
-    {
-        int offset_y = i / cdescr_width;
-        int offset_x = i - offset_y * cdescr_width;
-        descriptor[i] = hist[offset_y * img_block_width * cblock_hist_size + offset_x];
-    }
-}
-
-__kernel void extract_descrs_by_cols_kernel(const int cblock_hist_size, const int descriptors_quadstep, const int cdescr_size,
-                                            const int cnblocks_win_x, const int cnblocks_win_y, const int img_block_width, const int win_block_stride_x,
-                                            const int win_block_stride_y, __global const float* block_hists, __global float* descriptors)
-{
-    int tid = get_local_id(0);
-    int gidX = get_group_id(0);
-    int gidY = get_group_id(1);
-
-    // Get left top corner of the window in src
-    __global const float* hist = block_hists + (gidY * win_block_stride_y * img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
-
-    // Get left top corner of the window in dst
-    __global float* descriptor = descriptors + (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
-
-    // Copy elements from src to dst
-    for (int i = tid; i < cdescr_size; i += NTHREADS)
-    {
-        int block_idx = i / cblock_hist_size;
-        int idx_in_block = i - block_idx * cblock_hist_size;
-
-        int y = block_idx / cnblocks_win_x;
-        int x = block_idx - y * cnblocks_win_x;
-
-        descriptor[(x * cnblocks_win_y + y) * cblock_hist_size + idx_in_block] = hist[(y * img_block_width  + x) * cblock_hist_size + idx_in_block];
-    }
-}
-
-//----------------------------------------------------------------------------
-// Gradients computation
-
-__kernel void compute_gradients_8UC4_kernel(const int height, const int width, const int img_step, const int grad_quadstep, const int qangle_step,
-                                            const __global uchar4 * img, __global float * grad, __global uchar * qangle,
-                                            const float angle_scale, const char correct_gamma, const int cnbins)
-{
-    const int x = get_global_id(0);
-    const int tid = get_local_id(0);
-    const int gSizeX = get_local_size(0);
-    const int gidX = get_group_id(0);
-    const int gidY = get_group_id(1);
-
-    __global const uchar4* row = img + gidY * img_step;
-
-    __local float sh_row[(NTHREADS + 2) * 3];
-
-    uchar4 val;
-    if (x < width)
-        val = row[x];
-    else
-        val = row[width - 2];
-
-    sh_row[tid + 1] = val.x;
-    sh_row[tid + 1 + (NTHREADS + 2)] = val.y;
-    sh_row[tid + 1 + 2 * (NTHREADS + 2)] = val.z;
-
-    if (tid == 0)
-    {
-        val = row[max(x - 1, 1)];
-        sh_row[0] = val.x;
-        sh_row[(NTHREADS + 2)] = val.y;
-        sh_row[2 * (NTHREADS + 2)] = val.z;
-    }
-
-    if (tid == gSizeX - 1)
-    {
-        val = row[min(x + 1, width - 2)];
-        sh_row[gSizeX + 1] = val.x;
-        sh_row[gSizeX + 1 + (NTHREADS + 2)] = val.y;
-        sh_row[gSizeX + 1 + 2 * (NTHREADS + 2)] = val.z;
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    if (x < width)
-    {
-        float3 a = (float3) (sh_row[tid], sh_row[tid + (NTHREADS + 2)], sh_row[tid + 2 * (NTHREADS + 2)]);
-        float3 b = (float3) (sh_row[tid + 2], sh_row[tid + 2 + (NTHREADS + 2)], sh_row[tid + 2 + 2 * (NTHREADS + 2)]);
-
-        float3 dx;
-        if (correct_gamma == 1)
-            dx = sqrt(b) - sqrt(a);
-        else
-            dx = b - a;
-
-        float3 dy = (float3) 0.f;
-
-        if (gidY > 0 && gidY < height - 1)
-        {
-            a = convert_float3(img[(gidY - 1) * img_step + x].xyz);
-            b = convert_float3(img[(gidY + 1) * img_step + x].xyz);
-
-            if (correct_gamma == 1)
-                dy = sqrt(b) - sqrt(a);
-            else
-                dy = b - a;
-        }
-
-        float best_dx = dx.x;
-        float best_dy = dy.x;
-
-        float mag0 = dx.x * dx.x + dy.x * dy.x;
-        float mag1 = dx.y * dx.y + dy.y * dy.y;
-        if (mag0 < mag1)
-        {
-            best_dx = dx.y;
-            best_dy = dy.y;
-            mag0 = mag1;
-        }
-
-        mag1 = dx.z * dx.z + dy.z * dy.z;
-        if (mag0 < mag1)
-        {
-            best_dx = dx.z;
-            best_dy = dy.z;
-            mag0 = mag1;
-        }
-
-        mag0 = sqrt(mag0);
-
-        float ang = (atan2(best_dy, best_dx) + CV_PI_F) * angle_scale - 0.5f;
-        int hidx = (int)floor(ang);
-        ang -= hidx;
-        hidx = (hidx + cnbins) % cnbins;
-
-        qangle[(gidY * qangle_step + x) << 1] = hidx;
-        qangle[((gidY * qangle_step + x) << 1) + 1] = (hidx + 1) % cnbins;
-        grad[(gidY * grad_quadstep + x) << 1] = mag0 * (1.f - ang);
-        grad[((gidY * grad_quadstep + x) << 1) + 1] = mag0 * ang;
-    }
-}
-
-__kernel void compute_gradients_8UC1_kernel(const int height, const int width, const int img_step, const int grad_quadstep, const int qangle_step,
-                                            __global const uchar * img, __global float * grad, __global uchar * qangle,
-                                            const float angle_scale, const char correct_gamma, const int cnbins)
-{
-    const int x = get_global_id(0);
-    const int tid = get_local_id(0);
-    const int gSizeX = get_local_size(0);
-    const int gidX = get_group_id(0);
-    const int gidY = get_group_id(1);
-
-    __global const uchar* row = img + gidY * img_step;
-
-    __local float sh_row[NTHREADS + 2];
-
-    if (x < width)
-        sh_row[tid + 1] = row[x];
-    else
-        sh_row[tid + 1] = row[width - 2];
-
-    if (tid == 0)
-        sh_row[0] = row[max(x - 1, 1)];
-
-    if (tid == gSizeX - 1)
-        sh_row[gSizeX + 1] = row[min(x + 1, width - 2)];
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-    if (x < width)
-    {
-        float dx;
-
-        if (correct_gamma == 1)
-            dx = sqrt(sh_row[tid + 2]) - sqrt(sh_row[tid]);
-        else
-            dx = sh_row[tid + 2] - sh_row[tid];
-
-        float dy = 0.f;
-        if (gidY > 0 && gidY < height - 1)
-        {
-            float a = (float) img[ (gidY + 1) * img_step + x ];
-            float b = (float) img[ (gidY - 1) * img_step + x ];
-            if (correct_gamma == 1)
-                dy = sqrt(a) - sqrt(b);
-            else
-                dy = a - b;
-        }
-        float mag = sqrt(dx * dx + dy * dy);
-
-        float ang = (atan2(dy, dx) + CV_PI_F) * angle_scale - 0.5f;
-        int hidx = (int)floor(ang);
-        ang -= hidx;
-        hidx = (hidx + cnbins) % cnbins;
-
-        qangle[ (gidY * qangle_step + x) << 1 ]     = hidx;
-        qangle[ ((gidY * qangle_step + x) << 1) + 1 ] = (hidx + 1) % cnbins;
-        grad[ (gidY * grad_quadstep + x) << 1 ]       = mag * (1.f - ang);
-        grad[ ((gidY * grad_quadstep + x) << 1) + 1 ]   = mag * ang;
-    }
-}
-
-//----------------------------------------------------------------------------
-// Resize
-
-__kernel void resize_8UC4_kernel(__global uchar4 * dst, __global const uchar4 * src,
-                                 int dst_offset, int src_offset, int dst_step, int src_step,
-                                 int src_cols, int src_rows, int dst_cols, int dst_rows, float ifx, float ify )
-{
-    int dx = get_global_id(0);
-    int dy = get_global_id(1);
-
-    int sx = (int)floor(dx*ifx+0.5f);
-    int sy = (int)floor(dy*ify+0.5f);
-    sx = min(sx, src_cols-1);
-    sy = min(sy, src_rows-1);
-    int dpos = (dst_offset>>2) + dy * (dst_step>>2) + dx;
-    int spos = (src_offset>>2) + sy * (src_step>>2) + sx;
-
-    if(dx<dst_cols && dy<dst_rows)
-        dst[dpos] = src[spos];
-}
-
-__kernel void resize_8UC1_kernel(__global uchar * dst, __global const uchar * src,
-                                 int dst_offset, int src_offset, int dst_step, int src_step,
-                                 int src_cols, int src_rows, int dst_cols, int dst_rows, float ifx, float ify )
-{
-    int dx = get_global_id(0);
-    int dy = get_global_id(1);
-
-    int sx = (int)floor(dx*ifx+0.5f);
-    int sy = (int)floor(dy*ify+0.5f);
-    sx = min(sx, src_cols-1);
-    sy = min(sy, src_rows-1);
-    int dpos = dst_offset + dy * dst_step + dx;
-    int spos = src_offset + sy * src_step + sx;
-
-    if(dx<dst_cols && dy<dst_rows)
-        dst[dpos] = src[spos];
-}
\ No newline at end of file
diff --git a/modules/ocl/src/kernels/pyr_up.cl b/modules/ocl/src/kernels/pyr_up.cl
deleted file mode 100644
index d603ad6bce..0000000000
--- a/modules/ocl/src/kernels/pyr_up.cl
+++ /dev/null
@@ -1,750 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Zhang Chunpeng	chunpeng@multicorewareinc.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-//#pragma OPENCL EXTENSION cl_amd_printf : enable
-
-uchar get_valid_uchar(uchar data)
-{
-    return (uchar)(data <= 255 ? data : data > 0 ? 255 : 0);
-}
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_8UC1  //////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-__kernel void pyrUp_C1_D0(__global uchar* src,__global uchar* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float s_srcPatch[10][10];
-    __local float s_dstPatch[20][16];
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = (float)(src[srcx + srcy * srcStep]);
-
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float sum = 0;
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    if(eveny)
-    {
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)];
-        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)];
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)];
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy - 2][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][get_local_id(0)];
-    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][get_local_id(0)];
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][get_local_id(0)];
-
-    if ((x < dstCols) && (y < dstRows))
-        dst[x + y * dstStep] = (float)(4.0f * sum);
-
-}
-
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_16UC1  /////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-__kernel void pyrUp_C1_D2(__global ushort* src,__global ushort* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float s_srcPatch[10][10];
-    __local float s_dstPatch[20][16];
-
-    srcStep = srcStep >> 1;
-    dstStep = dstStep >> 1;
-    srcOffset = srcOffset >> 1;
-    dstOffset = dstOffset >> 1;
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = (float)(src[srcx + srcy * srcStep]);
-
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float sum = 0;
-
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    if(eveny)
-    {
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)];
-        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)];
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)];
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy - 2][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][get_local_id(0)];
-    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][get_local_id(0)];
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][get_local_id(0)];
-
-    if ((x < dstCols) && (y < dstRows))
-        dst[x + y * dstStep] = (float)(4.0f * sum);
-
-}
-
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_32FC1  /////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-__kernel void pyrUp_C1_D5(__global float* src,__global float* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float s_srcPatch[10][10];
-    __local float s_dstPatch[20][16];
-
-    srcOffset = srcOffset >> 2;
-    dstOffset = dstOffset >> 2;
-    srcStep = srcStep >> 2;
-    dstStep = dstStep >> 2;
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = (float)(src[srcx + srcy * srcStep]);
-
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float sum = 0;
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    if(eveny)
-    {
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)];
-        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)];
-        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)];
-        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)];
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
-            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
-            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
-            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy - 2][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][get_local_id(0)];
-    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][get_local_id(0)];
-    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][get_local_id(0)];
-    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][get_local_id(0)];
-
-    if ((x < dstCols) && (y < dstRows))
-        dst[x + y * dstStep] = (float)(4.0f * sum);
-
-}
-
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_8UC4  //////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-float4 covert_uchar4_to_float4(uchar4 data)
-{
-    float4 f4Data = {0,0,0,0};
-
-    f4Data.x = (float)data.x;
-    f4Data.y = (float)data.y;
-    f4Data.z = (float)data.z;
-    f4Data.w = (float)data.w;
-
-    return f4Data;
-}
-
-
-uchar4 convert_float4_to_uchar4(float4 data)
-{
-    uchar4 u4Data;
-
-    u4Data.x = get_valid_uchar(data.x);
-    u4Data.y = get_valid_uchar(data.y);
-    u4Data.z = get_valid_uchar(data.z);
-    u4Data.w = get_valid_uchar(data.w);
-
-    return u4Data;
-}
-
-float4 int_x_float4(int leftOpr,float4 rightOpr)
-{
-    float4 result = {0,0,0,0};
-
-    result.x = rightOpr.x * leftOpr;
-    result.y = rightOpr.y * leftOpr;
-    result.z = rightOpr.z * leftOpr;
-    result.w = rightOpr.w * leftOpr;
-
-    return result;
-}
-
-float4 float4_x_float4(float4 leftOpr,float4 rightOpr)
-{
-    float4 result;
-
-    result.x = leftOpr.x * rightOpr.x;
-    result.y = leftOpr.y * rightOpr.y;
-    result.z = leftOpr.z * rightOpr.z;
-    result.w = leftOpr.w * rightOpr.w;
-
-    return result;
-}
-
-__kernel void pyrUp_C4_D0(__global uchar4* src,__global uchar4* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float4 s_srcPatch[10][10];
-    __local float4 s_dstPatch[20][16];
-
-    srcOffset >>= 2;
-    dstOffset >>= 2;
-    srcStep >>= 2;
-    dstStep >>= 2;
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = covert_uchar4_to_float4(src[srcx + srcy * srcStep]);
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float4 sum = (float4)(0,0,0,0);
-
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
-    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
-    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
-
-
-    if(eveny)
-    {
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co1 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)]);
-
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1 ) , s_srcPatch[0][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx + 2) >> 1)]);
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1) , s_srcPatch[9][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx + 2) >> 1)]);
-
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy - 2][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy - 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co1 , s_dstPatch[2 + tidy    ][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy + 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy + 2][get_local_id(0)]);
-
-    if ((x < dstCols) && (y < dstRows))
-    {
-        dst[x + y * dstStep] = convert_float4_to_uchar4(int_x_float4(4.0f,sum));
-    }
-}
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_16UC4 //////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-float4 covert_ushort4_to_float4(ushort4 data)
-{
-    float4 f4Data = {0,0,0,0};
-
-    f4Data.x = (float)data.x;
-    f4Data.y = (float)data.y;
-    f4Data.z = (float)data.z;
-    f4Data.w = (float)data.w;
-
-    return f4Data;
-}
-
-
-ushort4 convert_float4_to_ushort4(float4 data)
-{
-    ushort4 u4Data;
-
-    u4Data.x = (float)data.x;
-    u4Data.y = (float)data.y;
-    u4Data.z = (float)data.z;
-    u4Data.w = (float)data.w;
-
-    return u4Data;
-}
-
-
-__kernel void pyrUp_C4_D2(__global ushort4* src,__global ushort4* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float4 s_srcPatch[10][10];
-    __local float4 s_dstPatch[20][16];
-
-    srcOffset >>= 3;
-    dstOffset >>= 3;
-    srcStep >>= 3;
-    dstStep >>= 3;
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = covert_ushort4_to_float4(src[srcx + srcy * srcStep]);
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float4 sum = (float4)(0,0,0,0);
-
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
-    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
-    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
-
-
-    if(eveny)
-    {
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co1 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)]);
-
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1 ) , s_srcPatch[0][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx + 2) >> 1)]);
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1) , s_srcPatch[9][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx + 2) >> 1)]);
-
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy - 2][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy - 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co1 , s_dstPatch[2 + tidy    ][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy + 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy + 2][get_local_id(0)]);
-
-    if ((x < dstCols) && (y < dstRows))
-    {
-        dst[x + y * dstStep] = convert_float4_to_ushort4(int_x_float4(4.0f,sum));
-    }
-}
-
-///////////////////////////////////////////////////////////////////////
-//////////////////////////  CV_32FC4 //////////////////////////////////
-///////////////////////////////////////////////////////////////////////
-__kernel void pyrUp_C4_D5(__global float4* src,__global float4* dst,
-                int srcRows,int dstRows,int srcCols,int dstCols,
-                int srcOffset,int dstOffset,int srcStep,int dstStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    __local float4 s_srcPatch[10][10];
-    __local float4 s_dstPatch[20][16];
-
-    srcOffset >>= 4;
-    dstOffset >>= 4;
-    srcStep >>= 4;
-    dstStep >>= 4;
-
-
-    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
-    {
-        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
-        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
-
-        srcx = abs(srcx);
-        srcx = min(srcCols - 1,srcx);
-
-        srcy = abs(srcy);
-        srcy = min(srcRows -1 ,srcy);
-
-        s_srcPatch[get_local_id(1)][get_local_id(0)] = (float4)(src[srcx + srcy * srcStep]);
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    float4 sum = (float4)(0,0,0,0);
-
-    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
-    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
-    const bool  eveny = ((get_local_id(1) & 1) == 0);
-    const int tidx = get_local_id(0);
-
-    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
-    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
-    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
-
-
-    if(eveny)
-    {
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co1 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( oddFlag , co2 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)]);
-        sum = sum + float4_x_float4(int_x_float4( evenFlag, co3 ) , s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)]);
-
-    }
-
-    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
-
-    if (get_local_id(1) < 2)
-    {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1 ) , s_srcPatch[0][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2  ) , s_srcPatch[0][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[0][1 + ((tidx + 2) >> 1)]);
-         }
-
-         s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-     if (get_local_id(1) > 13)
-     {
-        sum = 0;
-
-        if (eveny)
-        {
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx - 2) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx - 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co1) , s_srcPatch[9][1 + ((tidx    ) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4( oddFlag , co2) , s_srcPatch[9][1 + ((tidx + 1) >> 1)]);
-            sum = sum + float4_x_float4(int_x_float4(evenFlag , co3) , s_srcPatch[9][1 + ((tidx + 2) >> 1)]);
-
-        }
-        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
-     }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    sum = 0;
-
-    const int tidy = get_local_id(1);
-
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy - 2][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy - 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co1 , s_dstPatch[2 + tidy    ][get_local_id(0)]);
-    sum = sum + float4_x_float4(co2 , s_dstPatch[2 + tidy + 1][get_local_id(0)]);
-    sum = sum + float4_x_float4(co3 , s_dstPatch[2 + tidy + 2][get_local_id(0)]);
-
-    if ((x < dstCols) && (y < dstRows))
-    {
-        dst[x + y * dstStep] = 4.0f * sum;
-    }
-}
\ No newline at end of file
diff --git a/modules/ocl/src/kernels/pyrlk.cl b/modules/ocl/src/kernels/pyrlk.cl
deleted file mode 100644
index c772be78ac..0000000000
--- a/modules/ocl/src/kernels/pyrlk.cl
+++ /dev/null
@@ -1,1127 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Dachuan Zhao, dachuan@multicorewareinc.com
-//    Yao Wang, bitwangyaoyao@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-//#pragma OPENCL EXTENSION cl_amd_printf : enable
-
-__kernel void calcSharrDeriv_vertical_C1_D0(__global const uchar* src, int srcStep, int rows, int cols, int cn, __global short* dx_buf, int dx_bufStep, __global short* dy_buf, int dy_bufStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    if (y < rows && x < cols * cn)
-    {
-        const uchar src_val0 = (src + (y > 0 ? y-1 : rows > 1 ? 1 : 0) * srcStep)[x];
-        const uchar src_val1 = (src + y * srcStep)[x];
-        const uchar src_val2 = (src + (y < rows-1 ? y+1 : rows > 1 ? rows-2 : 0) * srcStep)[x];
-
-        ((__global short*)((__global char*)dx_buf + y * dx_bufStep / 2))[x] = (src_val0 + src_val2) * 3 + src_val1 * 10;
-        ((__global short*)((__global char*)dy_buf + y * dy_bufStep / 2))[x] = src_val2 - src_val0;
-    }
-}
-
-__kernel void calcSharrDeriv_vertical_C4_D0(__global const uchar* src, int srcStep, int rows, int cols, int cn, __global short* dx_buf, int dx_bufStep, __global short* dy_buf, int dy_bufStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    if (y < rows && x < cols * cn)
-    {
-        const uchar src_val0 = (src + (y > 0 ? y - 1 : 1) * srcStep)[x];
-        const uchar src_val1 = (src + y * srcStep)[x];
-        const uchar src_val2 = (src + (y < rows - 1 ? y + 1 : rows - 2) * srcStep)[x];
-
-        ((__global short*)((__global char*)dx_buf + y * dx_bufStep / 2))[x] = (src_val0 + src_val2) * 3 + src_val1 * 10;
-        ((__global short*)((__global char*)dy_buf + y * dy_bufStep / 2))[x] = src_val2 - src_val0;
-    }
-}
-
-__kernel void calcSharrDeriv_horizontal_C1_D0(int rows, int cols, int cn, __global const short* dx_buf, int dx_bufStep, __global const short* dy_buf, int dy_bufStep, __global short* dIdx, int dIdxStep, __global short* dIdy, int dIdyStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    const int colsn = cols * cn;
-
-    if (y < rows && x < colsn)
-    {
-        __global const short* dx_buf_row = dx_buf + y * dx_bufStep;
-        __global const short* dy_buf_row = dy_buf + y * dy_bufStep;
-
-        const int xr = x + cn < colsn ? x + cn : (cols - 2) * cn + x + cn - colsn;
-        const int xl = x - cn >= 0 ? x - cn : cn + x;
-
-        ((__global short*)((__global char*)dIdx + y * dIdxStep / 2))[x] = dx_buf_row[xr] - dx_buf_row[xl];
-        ((__global short*)((__global char*)dIdy + y * dIdyStep / 2))[x] = (dy_buf_row[xr] + dy_buf_row[xl]) * 3 + dy_buf_row[x] * 10;
-    }
-}
-
-__kernel void calcSharrDeriv_horizontal_C4_D0(int rows, int cols, int cn, __global const short* dx_buf, int dx_bufStep, __global const short* dy_buf, int dy_bufStep, __global short* dIdx, int dIdxStep, __global short* dIdy, int dIdyStep)
-{
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    const int colsn = cols * cn;
-
-    if (y < rows && x < colsn)
-    {
-        __global const short* dx_buf_row = dx_buf + y * dx_bufStep;
-        __global const short* dy_buf_row = dy_buf + y * dy_bufStep;
-
-        const int xr = x + cn < colsn ? x + cn : (cols - 2) * cn + x + cn - colsn;
-        const int xl = x - cn >= 0 ? x - cn : cn + x;
-
-        ((__global short*)((__global char*)dIdx + y * dIdxStep / 2))[x] = dx_buf_row[xr] - dx_buf_row[xl];
-        ((__global short*)((__global char*)dIdy + y * dIdyStep / 2))[x] = (dy_buf_row[xr] + dy_buf_row[xl]) * 3 + dy_buf_row[x] * 10;
-    }
-}
-
-#define W_BITS 14
-#define W_BITS1 14
-
-#define  CV_DESCALE(x, n)     (((x) + (1 << ((n)-1))) >> (n))
-
-int linearFilter_uchar(__global const uchar* src, int srcStep, int cn, float2 pt, int x, int y)
-{
-    int2 ipt;
-    ipt.x = convert_int_sat_rtn(pt.x);
-    ipt.y = convert_int_sat_rtn(pt.y);
-
-    float a = pt.x - ipt.x;
-    float b = pt.y - ipt.y;
-
-    int iw00 = convert_int_sat_rte((1.0f - a) * (1.0f - b) * (1 << W_BITS));
-    int iw01 = convert_int_sat_rte(a * (1.0f - b) * (1 << W_BITS));
-    int iw10 = convert_int_sat_rte((1.0f - a) * b * (1 << W_BITS));
-    int iw11 = (1 << W_BITS) - iw00 - iw01 - iw10;
-
-    __global const uchar* src_row = src + (ipt.y + y) * srcStep + ipt.x * cn;
-    __global const uchar* src_row1 = src + (ipt.y + y + 1) * srcStep + ipt.x * cn;
-
-    return CV_DESCALE(src_row[x] * iw00 + src_row[x + cn] * iw01 + src_row1[x] * iw10 + src_row1[x + cn] * iw11, W_BITS1 - 5);
-}
-
-int linearFilter_short(__global const short* src, int srcStep, int cn, float2 pt, int x, int y)
-{
-    int2 ipt;
-    ipt.x = convert_int_sat_rtn(pt.x);
-    ipt.y = convert_int_sat_rtn(pt.y);
-
-    float a = pt.x - ipt.x;
-    float b = pt.y - ipt.y;
-
-    int iw00 = convert_int_sat_rte((1.0f - a) * (1.0f - b) * (1 << W_BITS));
-    int iw01 = convert_int_sat_rte(a * (1.0f - b) * (1 << W_BITS));
-    int iw10 = convert_int_sat_rte((1.0f - a) * b * (1 << W_BITS));
-    int iw11 = (1 << W_BITS) - iw00 - iw01 - iw10;
-
-    __global const short* src_row = src + (ipt.y + y) * srcStep + ipt.x * cn;
-    __global const short* src_row1 = src + (ipt.y + y + 1) * srcStep + ipt.x * cn;
-
-    return CV_DESCALE(src_row[x] * iw00 + src_row[x + cn] * iw01 + src_row1[x] * iw10 + src_row1[x + cn] * iw11, W_BITS1);
-}
-
-float linearFilter_float(__global const float* src, int srcStep, int cn, float2 pt, float x, float y)
-{
-    int2 ipt;
-    ipt.x = convert_int_sat_rtn(pt.x);
-    ipt.y = convert_int_sat_rtn(pt.y);
-
-    float a = pt.x - ipt.x;
-    float b = pt.y - ipt.y;
-
-    float iw00 = ((1.0f - a) * (1.0f - b) * (1 << W_BITS));
-    float iw01 = (a * (1.0f - b) * (1 << W_BITS));
-    float iw10 = ((1.0f - a) * b * (1 << W_BITS));
-    float iw11 = (1 << W_BITS) - iw00 - iw01 - iw10;
-
-    __global const float* src_row = src + (int)(ipt.y + y) * srcStep / 4 + ipt.x * cn;
-    __global const float* src_row1 = src + (int)(ipt.y + y + 1) * srcStep / 4 + ipt.x * cn;
-
-    return src_row[(int)x] * iw00 + src_row[(int)x + cn] * iw01 + src_row1[(int)x] * iw10 + src_row1[(int)x + cn] * iw11, W_BITS1 - 5;
-}
-
-#define	BUFFER	64
-void reduce3(float val1, float val2, float val3, __local float* smem1, __local float* smem2, __local float* smem3, int tid)
-{
-    smem1[tid] = val1;
-    smem2[tid] = val2;
-    smem3[tid] = val3;
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-#if	BUFFER > 128
-    if (tid < 128)
-    {
-        smem1[tid] = val1 += smem1[tid + 128];
-        smem2[tid] = val2 += smem2[tid + 128];
-        smem3[tid] = val3 += smem3[tid + 128];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-#if	BUFFER > 64
-    if (tid < 64)
-    {
-        smem1[tid] = val1 += smem1[tid + 64];
-        smem2[tid] = val2 += smem2[tid + 64];
-        smem3[tid] = val3 += smem3[tid + 64];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-    if (tid < 32)
-    {
-        volatile __local float* vmem1 = smem1;
-        volatile __local float* vmem2 = smem2;
-        volatile __local float* vmem3 = smem3;
-
-        vmem1[tid] = val1 += vmem1[tid + 32];
-        vmem2[tid] = val2 += vmem2[tid + 32];
-        vmem3[tid] = val3 += vmem3[tid + 32];
-
-        vmem1[tid] = val1 += vmem1[tid + 16];
-        vmem2[tid] = val2 += vmem2[tid + 16];
-        vmem3[tid] = val3 += vmem3[tid + 16];
-
-        vmem1[tid] = val1 += vmem1[tid + 8];
-        vmem2[tid] = val2 += vmem2[tid + 8];
-        vmem3[tid] = val3 += vmem3[tid + 8];
-
-        vmem1[tid] = val1 += vmem1[tid + 4];
-        vmem2[tid] = val2 += vmem2[tid + 4];
-        vmem3[tid] = val3 += vmem3[tid + 4];
-
-        vmem1[tid] = val1 += vmem1[tid + 2];
-        vmem2[tid] = val2 += vmem2[tid + 2];
-        vmem3[tid] = val3 += vmem3[tid + 2];
-
-        vmem1[tid] = val1 += vmem1[tid + 1];
-        vmem2[tid] = val2 += vmem2[tid + 1];
-        vmem3[tid] = val3 += vmem3[tid + 1];
-    }
-}
-
-void reduce2(float val1, float val2, __local float* smem1, __local float* smem2, int tid)
-{
-    smem1[tid] = val1;
-    smem2[tid] = val2;
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-#if	BUFFER > 128
-    if (tid < 128)
-    {
-        smem1[tid] = val1 += smem1[tid + 128];
-        smem2[tid] = val2 += smem2[tid + 128];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-#if	BUFFER > 64
-    if (tid < 64)
-    {
-        smem1[tid] = val1 += smem1[tid + 64];
-        smem2[tid] = val2 += smem2[tid + 64];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-    if (tid < 32)
-    {
-        volatile __local float* vmem1 = smem1;
-        volatile __local float* vmem2 = smem2;
-
-        vmem1[tid] = val1 += vmem1[tid + 32];
-        vmem2[tid] = val2 += vmem2[tid + 32];
-
-        vmem1[tid] = val1 += vmem1[tid + 16];
-        vmem2[tid] = val2 += vmem2[tid + 16];
-
-        vmem1[tid] = val1 += vmem1[tid + 8];
-        vmem2[tid] = val2 += vmem2[tid + 8];
-
-        vmem1[tid] = val1 += vmem1[tid + 4];
-        vmem2[tid] = val2 += vmem2[tid + 4];
-
-        vmem1[tid] = val1 += vmem1[tid + 2];
-        vmem2[tid] = val2 += vmem2[tid + 2];
-
-        vmem1[tid] = val1 += vmem1[tid + 1];
-        vmem2[tid] = val2 += vmem2[tid + 1];
-    }
-}
-
-void reduce1(float val1, __local float* smem1, int tid)
-{
-    smem1[tid] = val1;
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-#if	BUFFER > 128
-    if (tid < 128)
-    {
-        smem1[tid] = val1 += smem1[tid + 128];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-#if	BUFFER > 64
-    if (tid < 64)
-    {
-        smem1[tid] = val1 += smem1[tid + 64];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-#endif
-
-    if (tid < 32)
-    {
-        volatile __local float* vmem1 = smem1;
-
-        vmem1[tid] = val1 += vmem1[tid + 32];
-        vmem1[tid] = val1 += vmem1[tid + 16];
-        vmem1[tid] = val1 += vmem1[tid + 8];
-        vmem1[tid] = val1 += vmem1[tid + 4];
-        vmem1[tid] = val1 += vmem1[tid + 2];
-        vmem1[tid] = val1 += vmem1[tid + 1];
-    }
-}
-
-#define SCALE (1.0f / (1 << 20))
-#define	THRESHOLD	0.01f
-#define	DIMENSION	21
-
-// Image read mode
-__constant sampler_t sampler    = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_LINEAR;
-
-void SetPatch(image2d_t I, float x, float y,
-                                float* Pch, float* Dx, float* Dy,
-                                float* A11, float* A12, float* A22)
-{
-            *Pch = read_imagef(I, sampler, (float2)(x, y)).x;
-
-            float dIdx = 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x + 1, y)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)).x -
-                             (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x - 1, y)).x + 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)).x);
-
-            float dIdy = 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x, y + 1)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)).x -
-                            (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x, y - 1)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)).x);
-
-
-            *Dx = dIdx;
-            *Dy = dIdy;
-
-            *A11 += dIdx * dIdx;
-            *A12 += dIdx * dIdy;
-            *A22 += dIdy * dIdy;
-}
-
-void GetPatch(image2d_t J, float x, float y,
-                                float* Pch, float* Dx, float* Dy,
-                                float* b1, float* b2)
-{
-                float J_val = read_imagef(J, sampler, (float2)(x, y)).x;
-                float diff = (J_val - *Pch) * 32.0f;
-                *b1 += diff**Dx;
-                *b2 += diff**Dy;
-}
-
-void GetError(image2d_t J, const float x, const float y, const float* Pch, float* errval)
-{
-        float diff = read_imagef(J, sampler, (float2)(x,y)).x-*Pch;
-        *errval += fabs(diff);
-}
-
-void SetPatch4(image2d_t I, const float x, const float y,
-                                float4* Pch, float4* Dx, float4* Dy,
-                                float* A11, float* A12, float* A22)
-{
-            *Pch = read_imagef(I, sampler, (float2)(x, y));
-
-            float4 dIdx = 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x + 1, y)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)) -
-                             (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x - 1, y)) + 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)));
-
-            float4 dIdy = 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)) + 10.0f * read_imagef(I, sampler, (float2)(x, y + 1)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)) -
-                            (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x, y - 1)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)));
-
-
-            *Dx = dIdx;
-            *Dy = dIdy;
-                        float4 sqIdx = dIdx * dIdx;
-                        *A11 += sqIdx.x + sqIdx.y + sqIdx.z;
-                        sqIdx = dIdx * dIdy;
-                        *A12 += sqIdx.x + sqIdx.y + sqIdx.z;
-                        sqIdx = dIdy * dIdy;
-                        *A22 += sqIdx.x + sqIdx.y + sqIdx.z;
-}
-
-void GetPatch4(image2d_t J, const float x, const float y,
-                                const float4* Pch, const float4* Dx, const float4* Dy,
-                                float* b1, float* b2)
-{
-                float4 J_val = read_imagef(J, sampler, (float2)(x, y));
-                float4 diff = (J_val - *Pch) * 32.0f;
-                                float4 xdiff = diff* *Dx;
-                                *b1 += xdiff.x + xdiff.y + xdiff.z;
-                                xdiff = diff* *Dy;
-                                *b2 += xdiff.x + xdiff.y + xdiff.z;
-}
-
-void GetError4(image2d_t J, const float x, const float y, const float4* Pch, float* errval)
-{
-        float4 diff = read_imagef(J, sampler, (float2)(x,y))-*Pch;
-        *errval += fabs(diff.x) + fabs(diff.y) + fabs(diff.z);
-}
-
-
-__kernel void lkSparse_C1_D5(image2d_t I, image2d_t J,
-    __global const float2* prevPts, int prevPtsStep, __global float2* nextPts, int nextPtsStep, __global uchar* status, __global float* err,
-        const int level, const int rows, const int cols, int PATCH_X, int PATCH_Y, int cn, int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
-{
-    __local float smem1[BUFFER];
-    __local float smem2[BUFFER];
-    __local float smem3[BUFFER];
-
-        unsigned int xid=get_local_id(0);
-        unsigned int yid=get_local_id(1);
-        unsigned int gid=get_group_id(0);
-        unsigned int xsize=get_local_size(0);
-        unsigned int ysize=get_local_size(1);
-        int xBase, yBase, i, j, k;
-
-        float2 c_halfWin = (float2)((c_winSize_x - 1)>>1, (c_winSize_y - 1)>>1);
-
-    const int tid = mad24(yid, xsize, xid);
-
-    float2 prevPt = prevPts[gid] / (1 << level);
-
-    if (prevPt.x < 0 || prevPt.x >= cols || prevPt.y < 0 || prevPt.y >= rows)
-    {
-        if (tid == 0 && level == 0)
-        {
-            status[gid] = 0;
-        }
-
-        return;
-    }
-    prevPt -= c_halfWin;
-
-    // extract the patch from the first image, compute covariation matrix of derivatives
-
-    float A11 = 0;
-    float A12 = 0;
-    float A22 = 0;
-
-    float I_patch[3][3];
-    float dIdx_patch[3][3];
-    float dIdy_patch[3][3];
-
-        yBase=yid;
-        {
-                xBase=xid;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[0][0], &dIdx_patch[0][0], &dIdy_patch[0][0],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[0][1], &dIdx_patch[0][1], &dIdy_patch[0][1],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[0][2], &dIdx_patch[0][2], &dIdy_patch[0][2],
-                                        &A11, &A12, &A22);
-        }
-        yBase+=ysize;
-        {
-                xBase=xid;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[1][0], &dIdx_patch[1][0], &dIdy_patch[1][0],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[1][1], &dIdx_patch[1][1], &dIdy_patch[1][1],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[1][2], &dIdx_patch[1][2], &dIdy_patch[1][2],
-                                        &A11, &A12, &A22);
-        }
-        yBase+=ysize;
-        if(yBase<c_winSize_y)
-        {
-                xBase=xid;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[2][0], &dIdx_patch[2][0], &dIdy_patch[2][0],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[2][1], &dIdx_patch[2][1], &dIdy_patch[2][1],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                        &I_patch[2][2], &dIdx_patch[2][2], &dIdy_patch[2][2],
-                                        &A11, &A12, &A22);
-        }
-    reduce3(A11, A12, A22, smem1, smem2, smem3, tid);
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    A11 = smem1[0];
-    A12 = smem2[0];
-    A22 = smem3[0];
-
-    float D = A11 * A22 - A12 * A12;
-
-    if (D < 1.192092896e-07f)
-    {
-        if (tid == 0 && level == 0)
-            status[gid] = 0;
-
-        return;
-    }
-
-    A11 /= D;
-    A12 /= D;
-    A22 /= D;
-
-    prevPt = nextPts[gid] * 2.0f - c_halfWin;
-
-    for (k = 0; k < c_iters; ++k)
-    {
-        if (prevPt.x < -c_halfWin.x || prevPt.x >= cols || prevPt.y < -c_halfWin.y || prevPt.y >= rows)
-        {
-            if (tid == 0 && level == 0)
-                status[gid] = 0;
-            return;
-        }
-
-        float b1 = 0;
-        float b2 = 0;
-
-                yBase=yid;
-                {
-                        xBase=xid;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][0], &dIdx_patch[0][0], &dIdy_patch[0][0],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][1], &dIdx_patch[0][1], &dIdy_patch[0][1],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][2], &dIdx_patch[0][2], &dIdy_patch[0][2],
-                                                &b1, &b2);
-                }
-                yBase+=ysize;
-                {
-                        xBase=xid;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][0], &dIdx_patch[1][0], &dIdy_patch[1][0],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][1], &dIdx_patch[1][1], &dIdy_patch[1][1],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][2], &dIdx_patch[1][2], &dIdy_patch[1][2],
-                                                &b1, &b2);
-                }
-                yBase+=ysize;
-                if(yBase<c_winSize_y)
-                {
-                        xBase=xid;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][0], &dIdx_patch[2][0], &dIdy_patch[2][0],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][1], &dIdx_patch[2][1], &dIdy_patch[2][1],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][2], &dIdx_patch[2][2], &dIdy_patch[2][2],
-                                                &b1, &b2);
-                }
-
-        reduce2(b1, b2, smem1, smem2, tid);
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        b1 = smem1[0];
-        b2 = smem2[0];
-
-        float2 delta;
-        delta.x = A12 * b2 - A22 * b1;
-        delta.y = A12 * b1 - A11 * b2;
-
-                prevPt += delta;
-
-        if (fabs(delta.x) < THRESHOLD && fabs(delta.y) < THRESHOLD)
-            break;
-    }
-
-    D = 0.0f;
-    if (calcErr)
-    {
-                yBase=yid;
-                {
-                        xBase=xid;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][0], &D);
-
-
-                        xBase+=xsize;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][1], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[0][2], &D);
-                }
-                yBase+=ysize;
-                {
-                        xBase=xid;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][0], &D);
-
-
-                        xBase+=xsize;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][1], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[1][2], &D);
-                }
-                yBase+=ysize;
-                if(yBase<c_winSize_y)
-                {
-                        xBase=xid;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][0], &D);
-
-
-                        xBase+=xsize;
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][1], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
-                                                &I_patch[2][2], &D);
-                }
-
-        reduce1(D, smem1, tid);
-    }
-
-    if (tid == 0)
-    {
-                prevPt += c_halfWin;
-
-        nextPts[gid] = prevPt;
-
-        if (calcErr)
-            err[gid] = smem1[0] / (c_winSize_x * c_winSize_y);
-    }
-
-}
-
-__kernel void lkSparse_C4_D5(image2d_t I, image2d_t J,
-    __global const float2* prevPts, int prevPtsStep, __global float2* nextPts, int nextPtsStep, __global uchar* status, __global float* err,
-        const int level, const int rows, const int cols, int PATCH_X, int PATCH_Y, int cn, int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
-{
-    __local float smem1[BUFFER];
-    __local float smem2[BUFFER];
-    __local float smem3[BUFFER];
-
-        unsigned int xid=get_local_id(0);
-        unsigned int yid=get_local_id(1);
-        unsigned int gid=get_group_id(0);
-        unsigned int xsize=get_local_size(0);
-        unsigned int ysize=get_local_size(1);
-        int xBase, yBase, i, j, k;
-
-        float2 c_halfWin = (float2)((c_winSize_x - 1)>>1, (c_winSize_y - 1)>>1);
-
-    const int tid = mad24(yid, xsize, xid);
-
-    float2 nextPt = prevPts[gid]/(1<<level);
-
-    if (nextPt.x < 0 || nextPt.x >= cols || nextPt.y < 0 || nextPt.y >= rows)
-    {
-        if (tid == 0 && level == 0)
-        {
-            status[gid] = 0;
-        }
-
-        return;
-    }
-
-        nextPt -= c_halfWin;
-
-    // extract the patch from the first image, compute covariation matrix of derivatives
-
-    float A11 = 0;
-    float A12 = 0;
-    float A22 = 0;
-
-    float4 I_patch[8];
-    float4 dIdx_patch[8];
-    float4 dIdy_patch[8];
-        float4 I_add,Dx_add,Dy_add;
-
-        yBase=yid;
-        {
-                xBase=xid;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[0], &dIdx_patch[0], &dIdy_patch[0],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[1], &dIdx_patch[1], &dIdy_patch[1],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[2], &dIdx_patch[2], &dIdy_patch[2],
-                                        &A11, &A12, &A22);
-
-        }
-        yBase+=ysize;
-        {
-                xBase=xid;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[3], &dIdx_patch[3], &dIdy_patch[3],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[4], &dIdx_patch[4], &dIdy_patch[4],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[5], &dIdx_patch[5], &dIdy_patch[5],
-                                        &A11, &A12, &A22);
-        }
-        yBase+=ysize;
-        if(yBase<c_winSize_y)
-        {
-                xBase=xid;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[6], &dIdx_patch[6], &dIdy_patch[6],
-                                        &A11, &A12, &A22);
-
-
-                xBase+=xsize;
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_patch[7], &dIdx_patch[7], &dIdy_patch[7],
-                                        &A11, &A12, &A22);
-
-                xBase+=xsize;
-                if(xBase<c_winSize_x)
-                SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                        &I_add, &Dx_add, &Dy_add,
-                                        &A11, &A12, &A22);
-        }
-
-    reduce3(A11, A12, A22, smem1, smem2, smem3, tid);
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    A11 = smem1[0];
-    A12 = smem2[0];
-    A22 = smem3[0];
-
-    float D = A11 * A22 - A12 * A12;
-
-    if (D < 1.192092896e-07f)
-    {
-        if (tid == 0 && level == 0)
-            status[gid] = 0;
-
-        return;
-    }
-
-    A11 /= D;
-    A12 /= D;
-    A22 /= D;
-
-        nextPt = nextPts[gid] * 2.0f - c_halfWin;
-
-    for (k = 0; k < c_iters; ++k)
-    {
-        if (nextPt.x < -c_halfWin.x || nextPt.x >= cols || nextPt.y < -c_halfWin.y || nextPt.y >= rows)
-        {
-            if (tid == 0 && level == 0)
-                status[gid] = 0;
-            return;
-        }
-
-        float b1 = 0;
-        float b2 = 0;
-
-                yBase=yid;
-                {
-                        xBase=xid;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[0], &dIdx_patch[0], &dIdy_patch[0],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[1], &dIdx_patch[1], &dIdy_patch[1],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[2], &dIdx_patch[2], &dIdy_patch[2],
-                                                &b1, &b2);
-                }
-                yBase+=ysize;
-                {
-                        xBase=xid;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[3], &dIdx_patch[3], &dIdy_patch[3],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[4], &dIdx_patch[4], &dIdy_patch[4],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[5], &dIdx_patch[5], &dIdy_patch[5],
-                                                &b1, &b2);
-                }
-                yBase+=ysize;
-                if(yBase<c_winSize_y)
-                {
-                        xBase=xid;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[6], &dIdx_patch[6], &dIdy_patch[6],
-                                                &b1, &b2);
-
-
-                        xBase+=xsize;
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[7], &dIdx_patch[7], &dIdy_patch[7],
-                                                &b1, &b2);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_add, &Dx_add, &Dy_add,
-                                                &b1, &b2);
-                }
-
-
-        reduce2(b1, b2, smem1, smem2, tid);
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        b1 = smem1[0];
-        b2 = smem2[0];
-
-        float2 delta;
-        delta.x = A12 * b2 - A22 * b1;
-        delta.y = A12 * b1 - A11 * b2;
-
-                nextPt +=delta;
-
-        if (fabs(delta.x) < THRESHOLD && fabs(delta.y) < THRESHOLD)
-            break;
-    }
-
-    D = 0.0f;
-    if (calcErr)
-    {
-                yBase=yid;
-                {
-                        xBase=xid;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[0], &D);
-
-
-                        xBase+=xsize;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[1], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[2], &D);
-                }
-                yBase+=ysize;
-                {
-                        xBase=xid;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[3], &D);
-
-
-                        xBase+=xsize;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[4], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[5], &D);
-                }
-                yBase+=ysize;
-                if(yBase<c_winSize_y)
-                {
-                        xBase=xid;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[6], &D);
-
-
-                        xBase+=xsize;
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_patch[7], &D);
-
-                        xBase+=xsize;
-                        if(xBase<c_winSize_x)
-                        GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
-                                                &I_add, &D);
-                }
-
-        reduce1(D, smem1, tid);
-    }
-
-    if (tid == 0)
-    {
-                nextPt += c_halfWin;
-        nextPts[gid] = nextPt;
-
-        if (calcErr)
-            err[gid] = smem1[0] / (3 * c_winSize_x * c_winSize_y);
-    }
-}
-
-__kernel void lkDense_C1_D0(image2d_t I, image2d_t J, __global float* u, int uStep, __global float* v, int vStep, __global const float* prevU, int prevUStep, __global const float* prevV, int prevVStep,
-    const int rows, const int cols, /*__global float* err, int errStep, int cn,*/ int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
-{
-        int c_halfWin_x = (c_winSize_x - 1) / 2;
-        int c_halfWin_y = (c_winSize_y - 1) / 2;
-
-    const int patchWidth  = get_local_size(0) + 2 * c_halfWin_x;
-    const int patchHeight = get_local_size(1) + 2 * c_halfWin_y;
-
-    __local int smem[8192];
-
-    __local int* I_patch = smem;
-    __local int* dIdx_patch = I_patch + patchWidth * patchHeight;
-    __local int* dIdy_patch = dIdx_patch + patchWidth * patchHeight;
-
-    const int xBase = get_group_id(0) * get_local_size(0);
-    const int yBase = get_group_id(1) * get_local_size(1);
-
-        sampler_t sampleri    = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_NEAREST;
-
-    for (int i = get_local_id(1); i < patchHeight; i += get_local_size(1))
-    {
-        for (int j = get_local_id(0); j < patchWidth; j += get_local_size(0))
-        {
-            float x = xBase - c_halfWin_x + j + 0.5f;
-            float y = yBase - c_halfWin_y + i + 0.5f;
-
-            I_patch[i * patchWidth + j] = read_imagei(I, sampleri, (float2)(x, y)).x;
-
-            // Sharr Deriv
-
-            dIdx_patch[i * patchWidth + j] = 3 * read_imagei(I, sampleri, (float2)(x+1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x+1, y)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y+1)).x -
-                                            (3 * read_imagei(I, sampleri, (float2)(x-1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x-1, y)).x + 3 * read_imagei(I, sampleri, (float2)(x-1, y+1)).x);
-
-            dIdy_patch[i * patchWidth + j] = 3 * read_imagei(I, sampleri, (float2)(x-1, y+1)).x + 10 * read_imagei(I, sampleri, (float2)(x, y+1)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y+1)).x -
-                                            (3 * read_imagei(I, sampleri, (float2)(x-1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x, y-1)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y-1)).x);
-        }
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    // extract the patch from the first image, compute covariation matrix of derivatives
-
-    const int x = get_global_id(0);
-    const int y = get_global_id(1);
-
-    if (x >= cols || y >= rows)
-        return;
-
-    int A11i = 0;
-    int A12i = 0;
-    int A22i = 0;
-
-    for (int i = 0; i < c_winSize_y; ++i)
-    {
-        for (int j = 0; j < c_winSize_x; ++j)
-        {
-            int dIdx = dIdx_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
-            int dIdy = dIdy_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
-
-            A11i += dIdx * dIdx;
-            A12i += dIdx * dIdy;
-            A22i += dIdy * dIdy;
-        }
-    }
-
-    float A11 = A11i;
-    float A12 = A12i;
-    float A22 = A22i;
-
-    float D = A11 * A22 - A12 * A12;
-
-    //if (calcErr && GET_MIN_EIGENVALS)
-    //    (err + y * errStep)[x] = minEig;
-
-    if (D < 1.192092896e-07f)
-    {
-        //if (calcErr)
-        //    err(y, x) = 3.402823466e+38f;
-
-        return;
-    }
-
-    D = 1.f / D;
-
-    A11 *= D;
-    A12 *= D;
-    A22 *= D;
-
-    float2 nextPt;
-    nextPt.x = x + prevU[y/2 * prevUStep / 4 + x/2] * 2.0f;
-    nextPt.y = y + prevV[y/2 * prevVStep / 4 + x/2] * 2.0f;
-
-    for (int k = 0; k < c_iters; ++k)
-    {
-        if (nextPt.x < 0 || nextPt.x >= cols || nextPt.y < 0 || nextPt.y >= rows)
-        {
-            //if (calcErr)
-            //    err(y, x) = 3.402823466e+38f;
-
-            return;
-        }
-
-        int b1 = 0;
-        int b2 = 0;
-
-        for (int i = 0; i < c_winSize_y; ++i)
-        {
-            for (int j = 0; j < c_winSize_x; ++j)
-            {
-                int iI = I_patch[(get_local_id(1) + i) * patchWidth + get_local_id(0) + j];
-                int iJ = read_imagei(J, sampler, (float2)(nextPt.x - c_halfWin_x + j + 0.5f, nextPt.y - c_halfWin_y + i + 0.5f)).x;
-
-                int diff = (iJ - iI) * 32;
-
-                int dIdx = dIdx_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
-                int dIdy = dIdy_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
-
-                b1 += diff * dIdx;
-                b2 += diff * dIdy;
-            }
-        }
-
-        float2 delta;
-        delta.x = A12 * b2 - A22 * b1;
-        delta.y = A12 * b1 - A11 * b2;
-
-        nextPt.x += delta.x;
-        nextPt.y += delta.y;
-
-        if (fabs(delta.x) < 0.01f && fabs(delta.y) < 0.01f)
-            break;
-    }
-
-    u[y * uStep / 4 + x] = nextPt.x - x;
-    v[y * vStep / 4 + x] = nextPt.y - y;
-
-    if (calcErr)
-    {
-        int errval = 0;
-
-        for (int i = 0; i < c_winSize_y; ++i)
-        {
-            for (int j = 0; j < c_winSize_x; ++j)
-            {
-                int iI = I_patch[(get_local_id(1) + i) * patchWidth + get_local_id(0) + j];
-                int iJ = read_imagei(J, sampler, (float2)(nextPt.x - c_halfWin_x + j + 0.5f, nextPt.y - c_halfWin_y + i + 0.5f)).x;
-
-                errval += abs(iJ - iI);
-            }
-        }
-
-        //err[y * errStep / 4 + x] = static_cast<float>(errval) / (c_winSize_x * c_winSize_y);
-    }
-}
diff --git a/modules/ocl/src/kmeans.cpp b/modules/ocl/src/kmeans.cpp
new file mode 100644
index 0000000000..0866117a63
--- /dev/null
+++ b/modules/ocl/src/kmeans.cpp
@@ -0,0 +1,437 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//     Xiaopeng Fu, fuxiaopeng2222@163.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include <iomanip>
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace ocl;
+
+namespace cv
+{
+namespace ocl
+{
+////////////////////////////////////OpenCL kernel strings//////////////////////////
+extern const char *kmeans_kernel;
+}
+}
+
+static void generateRandomCenter(const std::vector<Vec2f>& box, float* center, RNG& rng)
+{
+    size_t j, dims = box.size();
+    float margin = 1.f/dims;
+    for( j = 0; j < dims; j++ )
+        center[j] = ((float)rng*(1.f+margin*2.f)-margin)*(box[j][1] - box[j][0]) + box[j][0];
+}
+
+// This class is copied from matrix.cpp in core module.
+class KMeansPPDistanceComputer : public ParallelLoopBody
+{
+public:
+    KMeansPPDistanceComputer( float *_tdist2,
+                              const float *_data,
+                              const float *_dist,
+                              int _dims,
+                              size_t _step,
+                              size_t _stepci )
+        : tdist2(_tdist2),
+          data(_data),
+          dist(_dist),
+          dims(_dims),
+          step(_step),
+          stepci(_stepci) { }
+
+    void operator()( const cv::Range& range ) const
+    {
+        const int begin = range.start;
+        const int end = range.end;
+
+        for ( int i = begin; i<end; i++ )
+        {
+            tdist2[i] = std::min(normL2Sqr_(data + step*i, data + stepci, dims), dist[i]);
+        }
+    }
+
+private:
+    KMeansPPDistanceComputer& operator=(const KMeansPPDistanceComputer&); // to quiet MSVC
+
+    float *tdist2;
+    const float *data;
+    const float *dist;
+    const int dims;
+    const size_t step;
+    const size_t stepci;
+};
+/*
+k-means center initialization using the following algorithm:
+Arthur & Vassilvitskii (2007) k-means++: The Advantages of Careful Seeding
+*/
+static void generateCentersPP(const Mat& _data, Mat& _out_centers,
+                              int K, RNG& rng, int trials)
+{
+    int i, j, k, dims = _data.cols, N = _data.rows;
+    const float* data = (float*)_data.data;
+    size_t step = _data.step/sizeof(data[0]);
+    std::vector<int> _centers(K);
+    int* centers = &_centers[0];
+    std::vector<float> _dist(N*3);
+    float* dist = &_dist[0], *tdist = dist + N, *tdist2 = tdist + N;
+    double sum0 = 0;
+
+    centers[0] = (unsigned)rng % N;
+
+    for( i = 0; i < N; i++ )
+    {
+        dist[i] = normL2Sqr_(data + step*i, data + step*centers[0], dims);
+        sum0 += dist[i];
+    }
+
+    for( k = 1; k < K; k++ )
+    {
+        double bestSum = DBL_MAX;
+        int bestCenter = -1;
+
+        for( j = 0; j < trials; j++ )
+        {
+            double p = (double)rng*sum0, s = 0;
+            for( i = 0; i < N-1; i++ )
+                if( (p -= dist[i]) <= 0 )
+                    break;
+            int ci = i;
+
+            parallel_for_(Range(0, N),
+                          KMeansPPDistanceComputer(tdist2, data, dist, dims, step, step*ci));
+            for( i = 0; i < N; i++ )
+            {
+                s += tdist2[i];
+            }
+
+            if( s < bestSum )
+            {
+                bestSum = s;
+                bestCenter = ci;
+                std::swap(tdist, tdist2);
+            }
+        }
+        centers[k] = bestCenter;
+        sum0 = bestSum;
+        std::swap(dist, tdist);
+    }
+
+    for( k = 0; k < K; k++ )
+    {
+        const float* src = data + step*centers[k];
+        float* dst = _out_centers.ptr<float>(k);
+        for( j = 0; j < dims; j++ )
+            dst[j] = src[j];
+    }
+}
+
+void cv::ocl::distanceToCenters(oclMat &dists, oclMat &labels, const oclMat &src, const oclMat &centers)
+{
+    //if(src.clCxt -> impl -> double_support == 0 && src.type() == CV_64F)
+    //{
+    //    CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+    //    return;
+    //}
+
+    Context  *clCxt = src.clCxt;
+    int labels_step = (int)(labels.step/labels.elemSize());
+    String kernelname = "distanceToCenters";
+    int threadNum = src.rows > 256 ? 256 : src.rows;
+    size_t localThreads[3]  = {1, threadNum, 1};
+    size_t globalThreads[3] = {1, src.rows, 1};
+
+    std::vector<std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&labels_step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&centers.rows));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&labels.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&centers.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.rows));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&centers.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void*)&dists.data));
+
+    openCLExecuteKernel(clCxt, &kmeans_kernel, kernelname, globalThreads, localThreads, args, -1, -1, NULL);
+}
+///////////////////////////////////k - means /////////////////////////////////////////////////////////
+double cv::ocl::kmeans(const oclMat &_src, int K, oclMat &_bestLabels,
+                       TermCriteria criteria, int attempts, int flags, oclMat &_centers)
+{
+    const int SPP_TRIALS = 3;
+    bool isrow = _src.rows == 1 && _src.oclchannels() > 1;
+    int N = !isrow ? _src.rows : _src.cols;
+    int dims = (!isrow ? _src.cols : 1) * _src.oclchannels();
+    int type = _src.depth();
+
+    attempts = std::max(attempts, 1);
+    CV_Assert(type == CV_32F && K > 0 );
+    CV_Assert( N >= K );
+
+    Mat _labels;
+    if( flags & KMEANS_USE_INITIAL_LABELS )
+    {
+        CV_Assert( (_bestLabels.cols == 1 || _bestLabels.rows == 1) &&
+                   _bestLabels.cols * _bestLabels.rows == N &&
+                   _bestLabels.type() == CV_32S );
+        _bestLabels.download(_labels);
+    }
+    else
+    {
+        if( !((_bestLabels.cols == 1 || _bestLabels.rows == 1) &&
+                _bestLabels.cols * _bestLabels.rows == N &&
+                _bestLabels.type() == CV_32S &&
+                _bestLabels.isContinuous()))
+            _bestLabels.create(N, 1, CV_32S);
+        _labels.create(_bestLabels.size(), _bestLabels.type());
+    }
+    int* labels = _labels.ptr<int>();
+
+    Mat data;
+    _src.download(data);
+    Mat centers(K, dims, type), old_centers(K, dims, type), temp(1, dims, type);
+    std::vector<int> counters(K);
+    std::vector<Vec2f> _box(dims);
+    Vec2f* box = &_box[0];
+    double best_compactness = DBL_MAX, compactness = 0;
+    RNG& rng = theRNG();
+    int a, iter, i, j, k;
+
+    if( criteria.type & TermCriteria::EPS )
+        criteria.epsilon = std::max(criteria.epsilon, 0.);
+    else
+        criteria.epsilon = FLT_EPSILON;
+    criteria.epsilon *= criteria.epsilon;
+
+    if( criteria.type & TermCriteria::COUNT )
+        criteria.maxCount = std::min(std::max(criteria.maxCount, 2), 100);
+    else
+        criteria.maxCount = 100;
+
+    if( K == 1 )
+    {
+        attempts = 1;
+        criteria.maxCount = 2;
+    }
+
+    const float* sample = data.ptr<float>();
+    for( j = 0; j < dims; j++ )
+        box[j] = Vec2f(sample[j], sample[j]);
+
+    for( i = 1; i < N; i++ )
+    {
+        sample = data.ptr<float>(i);
+        for( j = 0; j < dims; j++ )
+        {
+            float v = sample[j];
+            box[j][0] = std::min(box[j][0], v);
+            box[j][1] = std::max(box[j][1], v);
+        }
+    }
+
+    for( a = 0; a < attempts; a++ )
+    {
+        double max_center_shift = DBL_MAX;
+        for( iter = 0;; )
+        {
+            swap(centers, old_centers);
+
+            if( iter == 0 && (a > 0 || !(flags & KMEANS_USE_INITIAL_LABELS)) )
+            {
+                if( flags & KMEANS_PP_CENTERS )
+                    generateCentersPP(data, centers, K, rng, SPP_TRIALS);
+                else
+                {
+                    for( k = 0; k < K; k++ )
+                        generateRandomCenter(_box, centers.ptr<float>(k), rng);
+                }
+            }
+            else
+            {
+                if( iter == 0 && a == 0 && (flags & KMEANS_USE_INITIAL_LABELS) )
+                {
+                    for( i = 0; i < N; i++ )
+                        CV_Assert( (unsigned)labels[i] < (unsigned)K );
+                }
+
+                // compute centers
+                centers = Scalar(0);
+                for( k = 0; k < K; k++ )
+                    counters[k] = 0;
+
+                for( i = 0; i < N; i++ )
+                {
+                    sample = data.ptr<float>(i);
+                    k = labels[i];
+                    float* center = centers.ptr<float>(k);
+                    j=0;
+#if CV_ENABLE_UNROLLED
+                    for(; j <= dims - 4; j += 4 )
+                    {
+                        float t0 = center[j] + sample[j];
+                        float t1 = center[j+1] + sample[j+1];
+
+                        center[j] = t0;
+                        center[j+1] = t1;
+
+                        t0 = center[j+2] + sample[j+2];
+                        t1 = center[j+3] + sample[j+3];
+
+                        center[j+2] = t0;
+                        center[j+3] = t1;
+                    }
+#endif
+                    for( ; j < dims; j++ )
+                        center[j] += sample[j];
+                    counters[k]++;
+                }
+
+                if( iter > 0 )
+                    max_center_shift = 0;
+
+                for( k = 0; k < K; k++ )
+                {
+                    if( counters[k] != 0 )
+                        continue;
+
+                    // if some cluster appeared to be empty then:
+                    //   1. find the biggest cluster
+                    //   2. find the farthest from the center point in the biggest cluster
+                    //   3. exclude the farthest point from the biggest cluster and form a new 1-point cluster.
+                    int max_k = 0;
+                    for( int k1 = 1; k1 < K; k1++ )
+                    {
+                        if( counters[max_k] < counters[k1] )
+                            max_k = k1;
+                    }
+
+                    double max_dist = 0;
+                    int farthest_i = -1;
+                    float* new_center =  centers.ptr<float>(k);
+                    float* old_center =  centers.ptr<float>(max_k);
+                    float* _old_center = temp.ptr<float>(); // normalized
+                    float scale = 1.f/counters[max_k];
+                    for( j = 0; j < dims; j++ )
+                        _old_center[j] = old_center[j]*scale;
+
+                    for( i = 0; i < N; i++ )
+                    {
+                        if( labels[i] != max_k )
+                            continue;
+                        sample = data.ptr<float>(i);
+                        double dist = normL2Sqr_(sample, _old_center, dims);
+
+                        if( max_dist <= dist )
+                        {
+                            max_dist = dist;
+                            farthest_i = i;
+                        }
+                    }
+
+                    counters[max_k]--;
+                    counters[k]++;
+                    labels[farthest_i] = k;
+                    sample = data.ptr<float>(farthest_i);
+
+                    for( j = 0; j < dims; j++ )
+                    {
+                        old_center[j] -= sample[j];
+                        new_center[j] += sample[j];
+                    }
+                }
+
+                for( k = 0; k < K; k++ )
+                {
+                    float* center = centers.ptr<float>(k);
+                    CV_Assert( counters[k] != 0 );
+
+                    float scale = 1.f/counters[k];
+                    for( j = 0; j < dims; j++ )
+                        center[j] *= scale;
+
+                    if( iter > 0 )
+                    {
+                        double dist = 0;
+                        const float* old_center = old_centers.ptr<float>(k);
+                        for( j = 0; j < dims; j++ )
+                        {
+                            double t = center[j] - old_center[j];
+                            dist += t*t;
+                        }
+                        max_center_shift = std::max(max_center_shift, dist);
+                    }
+                }
+            }
+
+            if( ++iter == MAX(criteria.maxCount, 2) || max_center_shift <= criteria.epsilon )
+                break;
+
+            // assign labels
+            oclMat _dists(1, N, CV_64F);
+
+            _bestLabels.upload(_labels);
+            _centers.upload(centers);
+            distanceToCenters(_dists, _bestLabels, _src, _centers);
+
+            Mat dists;
+            _dists.download(dists);
+            _bestLabels.download(_labels);
+
+            double* dist = dists.ptr<double>(0);
+            compactness = 0;
+            for( i = 0; i < N; i++ )
+            {
+                compactness += dist[i];
+            }
+        }
+
+        if( compactness < best_compactness )
+        {
+            best_compactness = compactness;
+        }
+    }
+
+    return best_compactness;
+}
diff --git a/modules/ocl/src/match_template.cpp b/modules/ocl/src/match_template.cpp
index 9d20f5204f..016a49a797 100644
--- a/modules/ocl/src/match_template.cpp
+++ b/modules/ocl/src/match_template.cpp
@@ -70,6 +70,9 @@ namespace cv
         void matchTemplate_SQDIFF_NORMED(
             const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
 
+        void convolve_32F(
+            const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
+
         void matchTemplate_CCORR(
             const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
 
@@ -89,41 +92,79 @@ namespace cv
         void matchTemplateNaive_CCORR(
             const oclMat &image, const oclMat &templ, oclMat &result, int cn);
 
+        void extractFirstChannel_32F(
+            const oclMat &image, oclMat &result);
+
         // Evaluates optimal template's area threshold. If
         // template's area is less  than the threshold, we use naive match
         // template version, otherwise FFT-based (if available)
-        static int getTemplateThreshold(int method, int depth)
+        static bool useNaive(int method, int depth, Size size)
         {
-            switch (method)
+#ifdef HAVE_CLAMDFFT
+            if (method == TM_SQDIFF && (depth == CV_32F || !Context::getContext()->supportsFeature(Context::CL_DOUBLE)))
             {
-            case CV_TM_CCORR:
-                if (depth == CV_32F) return 250;
-                if (depth == CV_8U) return 300;
-                break;
-            case CV_TM_SQDIFF:
-                if (depth == CV_32F) return 0x7fffffff; // do naive SQDIFF for CV_32F
-                if (depth == CV_8U) return 300;
-                break;
+                return true;
             }
-            CV_Error(CV_StsBadArg, "getTemplateThreshold: unsupported match template mode");
-            return 0;
+            else if(method == TM_CCORR || (method == TM_SQDIFF && depth == CV_8U))
+            {
+                return size.height < 18 && size.width < 18;
+            }
+            else
+                return false;
+#else
+#define UNUSED(x) (void)(x);
+            UNUSED(method) UNUSED(depth) UNUSED(size)
+#undef  UNUSED
+            return true;
+#endif
         }
 
         //////////////////////////////////////////////////////////////////////
         // SQDIFF
         void matchTemplate_SQDIFF(
-            const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &)
+            const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf & buf)
         {
             result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-            if (templ.size().area() < getTemplateThreshold(CV_TM_SQDIFF, image.depth()))
+            if (useNaive(TM_SQDIFF, image.depth(), templ.size()))
             {
                 matchTemplateNaive_SQDIFF(image, templ, result, image.oclchannels());
                 return;
             }
             else
             {
-                // TODO
-                CV_Error(CV_StsBadArg, "Not supported yet for this size template");
+                buf.image_sqsums.resize(1);
+
+                // TODO, add double support for ocl::integral
+                // use CPU integral temporarily
+                Mat sums, sqsums;
+                cv::integral(Mat(image.reshape(1)), sums, sqsums);
+                buf.image_sqsums[0] = sqsums;
+
+                unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
+                matchTemplate_CCORR(image, templ, result, buf);
+
+                //port CUDA's matchTemplatePrepared_SQDIFF_8U
+                Context *clCxt = image.clCxt;
+                String kernelName = "matchTemplate_Prepared_SQDIFF";
+                std::vector< std::pair<size_t, const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[0].data));
+                args.push_back( std::make_pair( sizeof(cl_mem), (void *)&result.data));
+                args.push_back( std::make_pair( sizeof(cl_ulong), (void *)&templ_sqsum));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.rows));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.cols));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&templ.rows));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&templ.cols));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].offset));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].step));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.offset));
+                args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
+
+                size_t globalThreads[3] = {result.cols, result.rows, 1};
+                size_t localThreads[3]  = {16, 16, 1};
+
+                const char * build_opt = image.oclchannels() == 4 ? "-D CN4" : "";
+                openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, 1, CV_8U, build_opt);
             }
         }
 
@@ -133,13 +174,12 @@ namespace cv
             matchTemplate_CCORR(image, templ, result, buf);
             buf.image_sums.resize(1);
 
-
             integral(image.reshape(1), buf.image_sums[0]);
 
             unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
 
             Context *clCxt = image.clCxt;
-            std::string kernelName = "matchTemplate_Prepared_SQDIFF_NORMED";
+            String kernelName = "matchTemplate_Prepared_SQDIFF_NORMED";
             std::vector< std::pair<size_t, const void *> > args;
 
             args.push_back( std::make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data));
@@ -155,7 +195,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
 
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
             openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, 1, CV_8U);
         }
 
@@ -169,7 +209,7 @@ namespace cv
             CV_Assert(result.rows == image.rows - templ.rows + 1 && result.cols == image.cols - templ.cols + 1);
 
             Context *clCxt = image.clCxt;
-            std::string kernelName = "matchTemplate_Naive_SQDIFF";
+            String kernelName = "matchTemplate_Naive_SQDIFF";
 
             std::vector< std::pair<size_t, const void *> > args;
 
@@ -190,33 +230,48 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
 
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
             openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
         }
 
         //////////////////////////////////////////////////////////////////////
         // CCORR
+        void convolve_32F(
+            const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
+        {
+            ConvolveBuf convolve_buf;
+            convolve_buf.user_block_size = buf.user_block_size;
+            if (image.oclchannels() == 1)
+                convolve(image, templ, result, true, convolve_buf);
+            else
+            {
+                oclMat result_;
+                convolve(image.reshape(1), templ.reshape(1), result_, true, convolve_buf);
+                extractFirstChannel_32F(result_, result);
+            }
+        }
+
         void matchTemplate_CCORR(
             const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
         {
             result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
-            if (templ.size().area() < getTemplateThreshold(CV_TM_SQDIFF, image.depth()))
+            if (useNaive(TM_CCORR, image.depth(), templ.size()))
             {
                 matchTemplateNaive_CCORR(image, templ, result, image.oclchannels());
                 return;
             }
             else
             {
-                CV_Error(CV_StsBadArg, "Not supported yet for this size template");
                 if(image.depth() == CV_8U && templ.depth() == CV_8U)
                 {
                     image.convertTo(buf.imagef, CV_32F);
                     templ.convertTo(buf.templf, CV_32F);
+                    convolve_32F(buf.imagef, buf.templf, result, buf);
+                }
+                else
+                {
+                    convolve_32F(image, templ, result, buf);
                 }
-                CV_Assert(image.oclchannels() == 1);
-                oclMat o_result(image.size(), CV_MAKETYPE(CV_32F, image.oclchannels()));
-                filter2D(buf.imagef, o_result, CV_32F, buf.templf, Point(0, 0));
-                result = o_result(Rect(0, 0, image.rows - templ.rows + 1, image.cols - templ.cols + 1));
             }
         }
 
@@ -232,7 +287,7 @@ namespace cv
             unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
 
             Context *clCxt = image.clCxt;
-            std::string kernelName = "normalizeKernel";
+            String kernelName = "normalizeKernel";
             std::vector< std::pair<size_t, const void *> > args;
 
             args.push_back( std::make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[0].data));
@@ -248,7 +303,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
 
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
             openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, 1, CV_8U);
         }
 
@@ -262,7 +317,7 @@ namespace cv
             CV_Assert(result.rows == image.rows - templ.rows + 1 && result.cols == image.cols - templ.cols + 1);
 
             Context *clCxt = image.clCxt;
-            std::string kernelName = "matchTemplate_Naive_CCORR";
+            String kernelName = "matchTemplate_Naive_CCORR";
 
             std::vector< std::pair<size_t, const void *> > args;
 
@@ -283,7 +338,7 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
 
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
             openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
         }
         //////////////////////////////////////////////////////////////////////
@@ -296,11 +351,11 @@ namespace cv
             matchTemplate_CCORR(image, templ, result, buf);
 
             Context *clCxt = image.clCxt;
-            std::string kernelName;
+            String kernelName;
 
             kernelName = "matchTemplate_Prepared_CCOFF";
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
 
             std::vector< std::pair<size_t, const void *> > args;
             args.push_back( std::make_pair( sizeof(cl_mem), (void *)&result.data) );
@@ -312,22 +367,22 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.cols) );
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.offset));
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
+            Vec4f templ_sum = Vec4f::all(0);
             // to be continued in the following section
             if(image.oclchannels() == 1)
             {
                 buf.image_sums.resize(1);
                 integral(image, buf.image_sums[0]);
 
-                float templ_sum = 0;
-                templ_sum = (float)sum(templ)[0] / templ.size().area();
+                templ_sum[0] = (float)sum(templ)[0] / templ.size().area();
                 args.push_back( std::make_pair( sizeof(cl_mem),  (void *)&buf.image_sums[0].data) );
                 args.push_back( std::make_pair( sizeof(cl_int),  (void *)&buf.image_sums[0].offset) );
                 args.push_back( std::make_pair( sizeof(cl_int),  (void *)&buf.image_sums[0].step) );
-                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sum) );
+                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sum[0]) );
             }
             else
             {
-                Vec4f templ_sum = Vec4f::all(0);
+
                 split(image, buf.images);
                 templ_sum = sum(templ) / templ.size().area();
                 buf.image_sums.resize(buf.images.size());
@@ -352,7 +407,7 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sum[3]) );
                     break;
                 default:
-                    CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
+                    CV_Error(Error::StsBadArg, "matchTemplate: unsupported number of channels");
                     break;
                 }
             }
@@ -369,11 +424,11 @@ namespace cv
             float scale = 1.f / templ.size().area();
 
             Context *clCxt = image.clCxt;
-            std::string kernelName;
+            String kernelName;
 
             kernelName = "matchTemplate_Prepared_CCOFF_NORMED";
             size_t globalThreads[3] = {result.cols, result.rows, 1};
-            size_t localThreads[3]  = {32, 8, 1};
+            size_t localThreads[3]  = {16, 16, 1};
 
             std::vector< std::pair<size_t, const void *> > args;
             args.push_back( std::make_pair( sizeof(cl_mem), (void *)&result.data) );
@@ -386,20 +441,22 @@ namespace cv
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.offset));
             args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
             args.push_back( std::make_pair( sizeof(cl_float), (void *)&scale) );
+
+            Vec4f templ_sum   = Vec4f::all(0);
+            Vec4f templ_sqsum = Vec4f::all(0);
             // to be continued in the following section
             if(image.oclchannels() == 1)
             {
                 buf.image_sums.resize(1);
                 buf.image_sqsums.resize(1);
                 integral(image, buf.image_sums[0], buf.image_sqsums[0]);
-                float templ_sum = 0;
-                float templ_sqsum = 0;
-                templ_sum   = (float)sum(templ)[0];
 
-                templ_sqsum = sqrSum(templ)[0];
+                templ_sum[0]   = (float)sum(templ)[0];
 
-                templ_sqsum -= scale * templ_sum * templ_sum;
-                templ_sum   *= scale;
+                templ_sqsum[0] = sqrSum(templ)[0];
+
+                templ_sqsum[0] -= scale * templ_sum[0] * templ_sum[0];
+                templ_sum[0]   *= scale;
 
                 args.push_back( std::make_pair( sizeof(cl_mem),  (void *)&buf.image_sums[0].data) );
                 args.push_back( std::make_pair( sizeof(cl_int),  (void *)&buf.image_sums[0].offset) );
@@ -407,13 +464,11 @@ namespace cv
                 args.push_back( std::make_pair( sizeof(cl_mem),  (void *)&buf.image_sqsums[0].data) );
                 args.push_back( std::make_pair( sizeof(cl_int),  (void *)&buf.image_sqsums[0].offset) );
                 args.push_back( std::make_pair( sizeof(cl_int),  (void *)&buf.image_sqsums[0].step) );
-                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sum) );
-                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sqsum) );
+                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sum[0]) );
+                args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sqsum[0]) );
             }
             else
             {
-                Vec4f templ_sum   = Vec4f::all(0);
-                Vec4f templ_sqsum = Vec4f::all(0);
 
                 split(image, buf.images);
                 templ_sum   = sum(templ);
@@ -458,13 +513,33 @@ namespace cv
                     args.push_back( std::make_pair( sizeof(cl_float), (void *)&templ_sqsum_sum) );
                     break;
                 default:
-                    CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
+                    CV_Error(Error::StsBadArg, "matchTemplate: unsupported number of channels");
                     break;
                 }
             }
             openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
         }
+        void extractFirstChannel_32F(const oclMat &image, oclMat &result)
+        {
+            Context *clCxt = image.clCxt;
+            String kernelName;
 
+            kernelName = "extractFirstChannel";
+            size_t globalThreads[3] = {result.cols, result.rows, 1};
+            size_t localThreads[3]  = {16, 16, 1};
+
+            std::vector< std::pair<size_t, const void *> > args;
+            args.push_back( std::make_pair( sizeof(cl_mem), (void *)&image.data) );
+            args.push_back( std::make_pair( sizeof(cl_mem), (void *)&result.data) );
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.rows) );
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.cols) );
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&image.offset));
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.offset));
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&image.step));
+            args.push_back( std::make_pair( sizeof(cl_int), (void *)&result.step));
+
+            openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, -1, -1);
+        }
     }/*ocl*/
 } /*cv*/
 
diff --git a/modules/ocl/src/matrix_operations.cpp b/modules/ocl/src/matrix_operations.cpp
index 189a7dc9a8..ddbd76db44 100644
--- a/modules/ocl/src/matrix_operations.cpp
+++ b/modules/ocl/src/matrix_operations.cpp
@@ -73,6 +73,7 @@ namespace cv
     }
 }
 
+
 ////////////////////////////////////////////////////////////////////////
 // convert_C3C4
 static void convert_C3C4(const cl_mem &src, oclMat &dst)
@@ -80,7 +81,7 @@ static void convert_C3C4(const cl_mem &src, oclMat &dst)
     int dstStep_in_pixel = dst.step1() / dst.oclchannels();
     int pixel_end = dst.wholecols * dst.wholerows - 1;
     Context *clCxt = dst.clCxt;
-    std::string kernelName = "convertC3C4";
+    String kernelName = "convertC3C4";
     char compile_option[32];
     switch(dst.depth())
     {
@@ -106,7 +107,7 @@ static void convert_C3C4(const cl_mem &src, oclMat &dst)
         sprintf(compile_option, "-D GENTYPE4=double4");
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+        CV_Error(Error::StsUnsupportedFormat, "unknown depth");
     }
     std::vector< std::pair<size_t, const void *> > args;
     args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src));
@@ -128,7 +129,7 @@ static void convert_C4C3(const oclMat &src, cl_mem &dst)
     int srcStep_in_pixel = src.step1() / src.oclchannels();
     int pixel_end = src.wholecols * src.wholerows - 1;
     Context *clCxt = src.clCxt;
-    std::string kernelName = "convertC4C3";
+    String kernelName = "convertC4C3";
     char compile_option[32];
     switch(src.depth())
     {
@@ -154,7 +155,7 @@ static void convert_C4C3(const oclMat &src, cl_mem &dst)
         sprintf(compile_option, "-D GENTYPE4=double4");
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+        CV_Error(Error::StsUnsupportedFormat, "unknown depth");
     }
 
     std::vector< std::pair<size_t, const void *> > args;
@@ -177,53 +178,70 @@ void cv::ocl::oclMat::upload(const Mat &m)
     Size wholeSize;
     Point ofs;
     m.locateROI(wholeSize, ofs);
-    //   int type = m.type();
-    //   if(m.oclchannels() == 3)
-    //{
-    //	type = CV_MAKETYPE(m.depth(), 4);
-    //}
-    create(wholeSize, m.type());
-
     if(m.channels() == 3)
     {
+        create(wholeSize, m.type());
         int pitch = wholeSize.width * 3 * m.elemSize1();
         int tail_padding = m.elemSize1() * 3072;
         int err;
-        cl_mem temp = clCreateBuffer(clCxt->impl->clContext, CL_MEM_READ_WRITE,
-                                     (pitch * wholeSize.height + tail_padding - 1) / tail_padding * tail_padding, 0, &err);
-        openCLVerifyCall(err);
+        cl_mem temp;
+        if(gDeviceMemType!=DEVICE_MEM_UHP && gDeviceMemType!=DEVICE_MEM_CHP){
+            temp = clCreateBuffer((cl_context)clCxt->oclContext(), CL_MEM_READ_WRITE,
+                                  (pitch * wholeSize.height + tail_padding - 1) / tail_padding * tail_padding, 0, &err);
+            openCLVerifyCall(err);
+            openCLMemcpy2D(clCxt, temp, pitch, m.datastart, m.step,
+                           wholeSize.width * m.elemSize(), wholeSize.height, clMemcpyHostToDevice, 3);
+        }
+        else{
+            temp = clCreateBuffer((cl_context)clCxt->oclContext(), CL_MEM_READ_WRITE|CL_MEM_USE_HOST_PTR,
+                                  (pitch * wholeSize.height + tail_padding - 1) / tail_padding * tail_padding, m.datastart, &err);
+            openCLVerifyCall(err);
+        }
+
 
-        openCLMemcpy2D(clCxt, temp, pitch, m.datastart, m.step, wholeSize.width * m.elemSize(), wholeSize.height, clMemcpyHostToDevice, 3);
         convert_C3C4(temp, *this);
-        //int* cputemp=new int[wholeSize.height*wholeSize.width * 3];
-        //int* cpudata=new int[this->step*this->wholerows/sizeof(int)];
-        //openCLSafeCall(clEnqueueReadBuffer(clCxt->impl->clCmdQueue, temp, CL_TRUE,
-        //						0, wholeSize.height*wholeSize.width * 3* sizeof(int), cputemp, 0, NULL, NULL));
-        //openCLSafeCall(clEnqueueReadBuffer(clCxt->impl->clCmdQueue, (cl_mem)data, CL_TRUE,
-        //						0, this->step*this->wholerows, cpudata, 0, NULL, NULL));
-        //for(int i=0;i<wholeSize.height;i++)
-        //{
-        //	int *a = cputemp+i*wholeSize.width * 3,*b = cpudata + i*this->step/sizeof(int);
-        //	for(int j=0;j<wholeSize.width;j++)
-        //	{
-        //		if((a[3*j] != b[4*j])||(a[3*j+1] != b[4*j+1])||(a[3*j+2] != b[4*j+2]))
-        //			printf("rows=%d,cols=%d,cputtemp=%d,%d,%d;cpudata=%d,%d,%d\n",
-        //			i,j,a[3*j],a[3*j+1],a[3*j+2],b[4*j],b[4*j+1],b[4*j+2]);
-        //	}
-        //}
-        //delete []cputemp;
-        //delete []cpudata;
         openCLSafeCall(clReleaseMemObject(temp));
     }
     else
     {
-        openCLMemcpy2D(clCxt, data, step, m.datastart, m.step, wholeSize.width * elemSize(), wholeSize.height, clMemcpyHostToDevice);
+        // try to use host ptr
+        createEx(wholeSize, m.type(), gDeviceMemRW, gDeviceMemType, m.datastart);
+        if(gDeviceMemType!=DEVICE_MEM_UHP && gDeviceMemType!=DEVICE_MEM_CHP)
+            openCLMemcpy2D(clCxt, data, step, m.datastart, m.step,
+                           wholeSize.width * elemSize(), wholeSize.height, clMemcpyHostToDevice);
     }
 
     rows = m.rows;
     cols = m.cols;
     offset = ofs.y * step + ofs.x * elemSize();
-    //download_channels = m.channels();
+}
+
+cv::ocl::oclMat::operator cv::_InputArray()
+{
+    _InputArray newInputArray;
+    newInputArray.flags = cv::_InputArray::OCL_MAT;
+    newInputArray.obj   = reinterpret_cast<void *>(this);
+    return newInputArray;
+}
+
+cv::ocl::oclMat::operator cv::_OutputArray()
+{
+    _OutputArray newOutputArray;
+    newOutputArray.flags = cv::_InputArray::OCL_MAT;
+    newOutputArray.obj   = reinterpret_cast<void *>(this);
+    return newOutputArray;
+}
+
+cv::ocl::oclMat& cv::ocl::getOclMatRef(InputArray src)
+{
+    CV_Assert(src.flags & cv::_InputArray::OCL_MAT);
+    return *reinterpret_cast<oclMat*>(src.obj);
+}
+
+cv::ocl::oclMat& cv::ocl::getOclMatRef(OutputArray src)
+{
+    CV_Assert(src.flags & cv::_InputArray::OCL_MAT);
+    return *reinterpret_cast<oclMat*>(src.obj);
 }
 
 void cv::ocl::oclMat::download(cv::Mat &m) const
@@ -241,7 +259,7 @@ void cv::ocl::oclMat::download(cv::Mat &m) const
         int pitch = wholecols * 3 * m.elemSize1();
         int tail_padding = m.elemSize1() * 3072;
         int err;
-        cl_mem temp = clCreateBuffer(clCxt->impl->clContext, CL_MEM_READ_WRITE,
+        cl_mem temp = clCreateBuffer((cl_context)clCxt->oclContext(), CL_MEM_READ_WRITE,
                                      (pitch * wholerows + tail_padding - 1) / tail_padding * tail_padding, 0, &err);
         openCLVerifyCall(err);
 
@@ -285,7 +303,7 @@ inline int divUp(int total, int grain)
 ///////////////////////////////////////////////////////////////////////////
 ////////////////////////////////// CopyTo /////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////
-static void copy_to_with_mask(const oclMat &src, oclMat &dst, const oclMat &mask, std::string kernelName)
+static void copy_to_with_mask(const oclMat &src, oclMat &dst, const oclMat &mask, String kernelName)
 {
     CV_DbgAssert( dst.rows == mask.rows && dst.cols == mask.cols &&
                   src.rows == dst.rows && src.cols == dst.cols
@@ -293,7 +311,7 @@ static void copy_to_with_mask(const oclMat &src, oclMat &dst, const oclMat &mask
 
     std::vector<std::pair<size_t , const void *> > args;
 
-    std::string string_types[4][7] = {{"uchar", "char", "ushort", "short", "int", "float", "double"},
+    String string_types[4][7] = {{"uchar", "char", "ushort", "short", "int", "float", "double"},
         {"uchar2", "char2", "ushort2", "short2", "int2", "float2", "double2"},
         {"uchar3", "char3", "ushort3", "short3", "int3", "float3", "double3"},
         {"uchar4", "char4", "ushort4", "short4", "int4", "float4", "double4"}
@@ -352,10 +370,10 @@ void cv::ocl::oclMat::copyTo( oclMat &mat, const oclMat &mask) const
 ///////////////////////////////////////////////////////////////////////////
 static void convert_run(const oclMat &src, oclMat &dst, double alpha, double beta)
 {
-    std::string kernelName = "convert_to_S";
+    String kernelName = "convert_to_S";
     std::stringstream idxStr;
     idxStr << src.depth();
-    kernelName += idxStr.str();
+    kernelName = kernelName + idxStr.str().c_str();
     float alpha_f = alpha, beta_f = beta;
     CV_DbgAssert(src.rows == dst.rows && src.cols == dst.cols);
     std::vector<std::pair<size_t , const void *> > args;
@@ -421,7 +439,7 @@ oclMat &cv::ocl::oclMat::operator = (const Scalar &s)
     setTo(s);
     return *this;
 }
-static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std::string kernelName)
+static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, String kernelName)
 {
     std::vector<std::pair<size_t , const void *> > args;
 
@@ -464,7 +482,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_8S:
@@ -483,7 +501,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_char4) , (void *)&val.cval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_16U:
@@ -502,7 +520,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_16S:
@@ -521,7 +539,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_short4) , (void *)&val.shval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_32S:
@@ -547,7 +565,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_int4) , (void *)&val.ival ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_32F:
@@ -566,7 +584,7 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_float4) , (void *)&val.fval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_64F:
@@ -585,18 +603,23 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
             args.push_back( std::make_pair( sizeof(cl_double4) , (void *)&val.dval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+        CV_Error(Error::StsUnsupportedFormat, "unknown depth");
     }
 #ifdef CL_VERSION_1_2
-    if(dst.offset == 0 && dst.cols == dst.wholecols)
+    //this enables backwards portability to
+    //run on OpenCL 1.1 platform if library binaries are compiled with OpenCL 1.2 support
+    if(Context::getContext()->supportsFeature(Context::CL_VER_1_2) &&
+        dst.offset == 0 && dst.cols == dst.wholecols)
     {
-        clEnqueueFillBuffer(dst.clCxt->impl->clCmdQueue, (cl_mem)dst.data, args[0].second, args[0].first, 0, dst.step * dst.rows, 0, NULL, NULL);
+        clEnqueueFillBuffer((cl_command_queue)dst.clCxt->oclCommandQueue(),
+            (cl_mem)dst.data, args[0].second, args[0].first, 0, dst.step * dst.rows, 0, NULL, NULL);
     }
     else
+#endif
     {
         args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data ));
         args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.cols ));
@@ -604,20 +627,11 @@ static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, std:
         args.push_back( std::make_pair( sizeof(cl_int) , (void *)&step_in_pixel ));
         args.push_back( std::make_pair( sizeof(cl_int) , (void *)&offset_in_pixel));
         openCLExecuteKernel(dst.clCxt , &operator_setTo, kernelName, globalThreads,
-                            localThreads, args, -1, -1, compile_option);
+            localThreads, args, -1, -1, compile_option);
     }
-#else
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.cols ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.rows ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&step_in_pixel ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&offset_in_pixel));
-    openCLExecuteKernel(dst.clCxt , &operator_setTo, kernelName, globalThreads,
-                        localThreads, args, -1, -1, compile_option);
-#endif
 }
 
-static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const oclMat &mask, std::string kernelName)
+static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const oclMat &mask, String kernelName)
 {
     CV_DbgAssert( dst.rows == mask.rows && dst.cols == mask.cols);
     std::vector<std::pair<size_t , const void *> > args;
@@ -656,7 +670,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_8S:
@@ -675,7 +689,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_char4) , (void *)&val.cval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_16U:
@@ -694,7 +708,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_16S:
@@ -713,7 +727,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_short4) , (void *)&val.shval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_32S:
@@ -732,7 +746,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_int4) , (void *)&val.ival ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_32F:
@@ -751,7 +765,7 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_float4) , (void *)&val.fval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case CV_64F:
@@ -770,11 +784,11 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
             args.push_back( std::make_pair( sizeof(cl_double4) , (void *)&val.dval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+        CV_Error(Error::StsUnsupportedFormat, "unknown depth");
     }
     args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.cols ));
@@ -824,13 +838,8 @@ oclMat &cv::ocl::oclMat::setTo(const Scalar &scalar, const oclMat &mask)
 oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
 {
     if( new_rows != 0 && new_rows != rows)
-
     {
-
-        CV_Error( CV_StsBadFunc,
-
-                  "oclMat's number of rows can not be changed for current version" );
-
+        CV_Error( Error::StsBadFunc, "oclMat's number of rows can not be changed for current version" );
     }
 
     oclMat hdr = *this;
@@ -863,13 +872,13 @@ oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
 
         if (!isContinuous())
 
-            CV_Error(CV_BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
+            CV_Error(Error::BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
 
 
 
         if ((unsigned)new_rows > (unsigned)total_size)
 
-            CV_Error(CV_StsOutOfRange, "Bad new number of rows");
+            CV_Error(Error::StsOutOfRange, "Bad new number of rows");
 
 
 
@@ -879,7 +888,7 @@ oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
 
         if (total_width * new_rows != total_size)
 
-            CV_Error(CV_StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
+            CV_Error(Error::StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
 
 
 
@@ -897,7 +906,7 @@ oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
 
     if (new_width * new_cn != total_width)
 
-        CV_Error(CV_BadNumChannels, "The total width is not divisible by the new number of channels");
+        CV_Error(Error::BadNumChannels, "The total width is not divisible by the new number of channels");
 
 
 
@@ -913,9 +922,10 @@ oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
 
 }
 
-void cv::ocl::oclMat::createEx(Size size, int type, DevMemRW rw_type, DevMemType mem_type)
+void cv::ocl::oclMat::createEx(Size size, int type,
+                               DevMemRW rw_type, DevMemType mem_type, void* hptr)
 {
-    createEx(size.height, size.width, type, rw_type, mem_type);
+    createEx(size.height, size.width, type, rw_type, mem_type, hptr);
 }
 
 void cv::ocl::oclMat::create(int _rows, int _cols, int _type)
@@ -923,16 +933,12 @@ void cv::ocl::oclMat::create(int _rows, int _cols, int _type)
     createEx(_rows, _cols, _type, gDeviceMemRW, gDeviceMemType);
 }
 
-void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type, DevMemRW rw_type, DevMemType mem_type)
+void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type,
+                               DevMemRW rw_type, DevMemType mem_type, void* hptr)
 {
     clCxt = Context::getContext();
     /* core logic */
-    _type &= TYPE_MASK;
-    //download_channels = CV_MAT_CN(_type);
-    //if(download_channels==3)
-    //{
-    //	_type = CV_MAKE_TYPE((CV_MAT_DEPTH(_type)),4);
-    //}
+    _type &= Mat::TYPE_MASK;
     if( rows == _rows && cols == _cols && type() == _type && data )
         return;
     if( data )
@@ -948,8 +954,8 @@ void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type, DevMemRW rw_type
         size_t esz = elemSize();
 
         void *dev_ptr;
-        openCLMallocPitchEx(clCxt, &dev_ptr, &step, GPU_MATRIX_MALLOC_STEP(esz * cols), rows, rw_type, mem_type);
-        //openCLMallocPitch(clCxt,&dev_ptr, &step, esz * cols, rows);
+        openCLMallocPitch(clCxt, &dev_ptr, &step, GPU_MATRIX_MALLOC_STEP(esz * cols),
+                            rows, rw_type, mem_type, hptr);
 
         if (esz * cols == step)
             flags |= Mat::CONTINUOUS_FLAG;
diff --git a/modules/ocl/src/mcwutil.cpp b/modules/ocl/src/mcwutil.cpp
index ebdcbf73e1..c520cb8d48 100644
--- a/modules/ocl/src/mcwutil.cpp
+++ b/modules/ocl/src/mcwutil.cpp
@@ -43,12 +43,29 @@
 //
 //M*/
 
-#include "mcwutil.hpp"
+#include "precomp.hpp"
 
-#if defined (HAVE_OPENCL)
-#ifndef CL_VERSION_1_2
-#define CL_VERSION_1_2 0
+#ifdef __GNUC__
+#if ((__GNUC__ * 100) + __GNUC_MINOR__) >= 402
+#define GCC_DIAG_STR(s) #s
+#define GCC_DIAG_JOINSTR(x,y) GCC_DIAG_STR(x ## y)
+# define GCC_DIAG_DO_PRAGMA(x) _Pragma (#x)
+# define GCC_DIAG_PRAGMA(x) GCC_DIAG_DO_PRAGMA(GCC diagnostic x)
+# if ((__GNUC__ * 100) + __GNUC_MINOR__) >= 406
+#  define GCC_DIAG_OFF(x) GCC_DIAG_PRAGMA(push) \
+GCC_DIAG_PRAGMA(ignored GCC_DIAG_JOINSTR(-W,x))
+#  define GCC_DIAG_ON(x) GCC_DIAG_PRAGMA(pop)
+# else
+#  define GCC_DIAG_OFF(x) GCC_DIAG_PRAGMA(ignored GCC_DIAG_JOINSTR(-W,x))
+#  define GCC_DIAG_ON(x)  GCC_DIAG_PRAGMA(warning GCC_DIAG_JOINSTR(-W,x))
+# endif
+#else
+# define GCC_DIAG_OFF(x)
+# define GCC_DIAG_ON(x)
 #endif
+#endif /* __GNUC__ */
+
+using namespace std;
 
 namespace cv
 {
@@ -61,9 +78,9 @@ namespace cv
         }
 
         // provide additional methods for the user to interact with the command queue after a task is fired
-        static void openCLExecuteKernel_2(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
+        static void openCLExecuteKernel_2(Context *clCxt , const char **source, String kernelName, size_t globalThreads[3],
                                    size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
-                                   int depth, char *build_options, FLUSH_MODE finish_mode)
+                                   int depth, const char *build_options, FLUSH_MODE finish_mode)
         {
             //construct kernel name
             //The rule is functionName_Cn_Dn, C represent Channels, D Represent DataType Depth, n represent an integer number
@@ -73,7 +90,7 @@ namespace cv
                 idxStr << "_C" << channels;
             if(depth != -1)
                 idxStr << "_D" << depth;
-            kernelName += idxStr.str();
+            kernelName = kernelName + idxStr.str().c_str();
 
             cl_kernel kernel;
             kernel = openCLGetKernelFromSource(clCxt, source, kernelName, build_options);
@@ -90,15 +107,15 @@ namespace cv
             for(size_t i = 0; i < args.size(); i ++)
                 openCLSafeCall(clSetKernelArg(kernel, i, args[i].first, args[i].second));
 
-            openCLSafeCall(clEnqueueNDRangeKernel(clCxt->impl->clCmdQueue, kernel, 3, NULL, globalThreads,
+            openCLSafeCall(clEnqueueNDRangeKernel((cl_command_queue)clCxt->oclCommandQueue(), kernel, 3, NULL, globalThreads,
                                                   localThreads, 0, NULL, NULL));
 
             switch(finish_mode)
             {
             case CLFINISH:
-                clFinish(clCxt->impl->clCmdQueue);
+                clFinish((cl_command_queue)clCxt->oclCommandQueue());
             case CLFLUSH:
-                clFlush(clCxt->impl->clCmdQueue);
+                clFlush((cl_command_queue)clCxt->oclCommandQueue());
                 break;
             case DISABLE:
             default:
@@ -107,29 +124,32 @@ namespace cv
             openCLSafeCall(clReleaseKernel(kernel));
         }
 
-        void openCLExecuteKernel2(Context *clCxt , const char **source, std::string kernelName,
+        void openCLExecuteKernel2(Context *clCxt , const char **source, String kernelName,
                                   size_t globalThreads[3], size_t localThreads[3],
                                   std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, FLUSH_MODE finish_mode)
         {
             openCLExecuteKernel2(clCxt, source, kernelName, globalThreads, localThreads, args,
                                  channels, depth, NULL, finish_mode);
         }
-        void openCLExecuteKernel2(Context *clCxt , const char **source, std::string kernelName,
+        void openCLExecuteKernel2(Context *clCxt , const char **source, String kernelName,
                                   size_t globalThreads[3], size_t localThreads[3],
-                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, char *build_options, FLUSH_MODE finish_mode)
+                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options, FLUSH_MODE finish_mode)
 
         {
             openCLExecuteKernel_2(clCxt, source, kernelName, globalThreads, localThreads, args, channels, depth,
                                   build_options, finish_mode);
         }
 
-       cl_mem bindTexture(const oclMat &mat)
+#ifdef __GNUC__
+        GCC_DIAG_OFF(deprecated-declarations)
+#endif
+        cl_mem bindTexture(const oclMat &mat)
         {
             cl_mem texture;
             cl_image_format format;
             int err;
             int depth    = mat.depth();
-            int channels = mat.channels();
+            int channels = mat.oclchannels();
 
             switch(depth)
             {
@@ -143,7 +163,7 @@ namespace cv
                 format.image_channel_data_type = CL_FLOAT;
                 break;
             default:
-                throw std::exception();
+                CV_Error(-1, "Image forma is not supported");
                 break;
             }
             switch(channels)
@@ -158,61 +178,76 @@ namespace cv
                 format.image_channel_order     = CL_RGBA;
                 break;
             default:
-                throw std::exception();
+                CV_Error(-1, "Image format is not supported");
                 break;
             }
-#if CL_VERSION_1_2
-            cl_image_desc desc;
-            desc.image_type       = CL_MEM_OBJECT_IMAGE2D;
-            desc.image_width      = mat.cols;
-            desc.image_height     = mat.rows;
-            desc.image_depth      = 0;
-            desc.image_array_size = 1;
-            desc.image_row_pitch  = 0;
-            desc.image_slice_pitch = 0;
-            desc.buffer           = NULL;
-            desc.num_mip_levels   = 0;
-            desc.num_samples      = 0;
-            texture = clCreateImage(mat.clCxt->impl->clContext, CL_MEM_READ_WRITE, &format, &desc, NULL, &err);
-#else
-            texture = clCreateImage2D(
-                mat.clCxt->impl->clContext,
-                CL_MEM_READ_WRITE,
-                &format,
-                mat.cols,
-                mat.rows,
-                0,
-                NULL,
-                &err);
+#ifdef CL_VERSION_1_2
+            //this enables backwards portability to
+            //run on OpenCL 1.1 platform if library binaries are compiled with OpenCL 1.2 support
+            if(Context::getContext()->supportsFeature(Context::CL_VER_1_2))
+            {
+                cl_image_desc desc;
+                desc.image_type       = CL_MEM_OBJECT_IMAGE2D;
+                desc.image_width      = mat.cols;
+                desc.image_height     = mat.rows;
+                desc.image_depth      = 0;
+                desc.image_array_size = 1;
+                desc.image_row_pitch  = 0;
+                desc.image_slice_pitch = 0;
+                desc.buffer           = NULL;
+                desc.num_mip_levels   = 0;
+                desc.num_samples      = 0;
+                texture = clCreateImage((cl_context)mat.clCxt->oclContext(), CL_MEM_READ_WRITE, &format, &desc, NULL, &err);
+            }
+            else
 #endif
+            {
+                texture = clCreateImage2D(
+                    (cl_context)mat.clCxt->oclContext(),
+                    CL_MEM_READ_WRITE,
+                    &format,
+                    mat.cols,
+                    mat.rows,
+                    0,
+                    NULL,
+                    &err);
+            }
             size_t origin[] = { 0, 0, 0 };
             size_t region[] = { mat.cols, mat.rows, 1 };
 
             cl_mem devData;
             if (mat.cols * mat.elemSize() != mat.step)
             {
-                devData = clCreateBuffer(mat.clCxt->impl->clContext, CL_MEM_READ_ONLY, mat.cols * mat.rows
+                devData = clCreateBuffer((cl_context)mat.clCxt->oclContext(), CL_MEM_READ_ONLY, mat.cols * mat.rows
                     * mat.elemSize(), NULL, NULL);
                 const size_t regin[3] = {mat.cols * mat.elemSize(), mat.rows, 1};
-                clEnqueueCopyBufferRect(mat.clCxt->impl->clCmdQueue, (cl_mem)mat.data, devData, origin, origin,
+                clEnqueueCopyBufferRect((cl_command_queue)mat.clCxt->oclCommandQueue(), (cl_mem)mat.data, devData, origin, origin,
                     regin, mat.step, 0, mat.cols * mat.elemSize(), 0, 0, NULL, NULL);
+                clFlush((cl_command_queue)mat.clCxt->oclCommandQueue());
             }
             else
             {
                 devData = (cl_mem)mat.data;
             }
 
-            clEnqueueCopyBufferToImage(mat.clCxt->impl->clCmdQueue, devData, texture, 0, origin, region, 0, NULL, 0);
+            clEnqueueCopyBufferToImage((cl_command_queue)mat.clCxt->oclCommandQueue(), devData, texture, 0, origin, region, 0, NULL, 0);
             if ((mat.cols * mat.elemSize() != mat.step))
             {
-                clFinish(mat.clCxt->impl->clCmdQueue);
+                clFlush((cl_command_queue)mat.clCxt->oclCommandQueue());
                 clReleaseMemObject(devData);
             }
 
             openCLSafeCall(err);
             return texture;
         }
+#ifdef __GNUC__
+        GCC_DIAG_ON(deprecated-declarations)
+#endif
 
+        Ptr<TextureCL> bindTexturePtr(const oclMat &mat)
+        {
+            return makePtr<TextureCL>(bindTexture(mat), mat.rows, mat.cols, mat.type());
+        }
         void releaseTexture(cl_mem& texture)
         {
             openCLFree(texture);
@@ -230,6 +265,7 @@ namespace cv
             try
             {
                 cv::ocl::openCLGetKernelFromSource(clCxt, &_kernel_string, "test_func");
+                finish();
                 _support = true;
             }
             catch (const cv::Exception& e)
@@ -250,4 +286,3 @@ namespace cv
     }//namespace ocl
 
 }//namespace cv
-#endif
\ No newline at end of file
diff --git a/modules/ocl/src/moments.cpp b/modules/ocl/src/moments.cpp
index 4bf3d0032b..60d6ad7c53 100644
--- a/modules/ocl/src/moments.cpp
+++ b/modules/ocl/src/moments.cpp
@@ -16,7 +16,7 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Sen Liu, sen@multicorewareinc.com
+//    Sen Liu, swjtuls1987@126.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -45,6 +45,10 @@
 //M*/
 #include "precomp.hpp"
 #include <iostream>
+
+#include "opencv2/imgproc/types_c.h"
+#include "opencv2/imgproc/imgproc_c.h"
+
 namespace cv
 {
 namespace ocl
@@ -98,25 +102,19 @@ static void icvContourMoments( CvSeq* contour, CvMoments* mom )
         CvSeqReader reader;
         int lpt = contour->total;
         double a00, a10, a01, a20, a11, a02, a30, a21, a12, a03;
-        int dst_type = cv::ocl::Context::getContext()->impl->double_support ? CV_64FC1 : CV_32FC1;
 
         cvStartReadSeq( contour, &reader, 0 );
 
-        cv::ocl::oclMat dst_a00(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a10(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a01(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a20(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a11(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a02(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a30(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a21(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a12(1,lpt,dst_type);
-        cv::ocl::oclMat dst_a03(1,lpt,dst_type);
         size_t reader_size = lpt << 1;
         cv::Mat reader_mat(1,reader_size,CV_32FC1);
 
         bool is_float = CV_SEQ_ELTYPE(contour) == CV_32FC2;
 
+        if (!cv::ocl::Context::getContext()->supportsFeature(Context::CL_DOUBLE) && is_float)
+        {
+            CV_Error(CV_StsUnsupportedFormat, "Moments - double is not supported by your GPU!");
+        }
+
         if( is_float )
         {
             for(size_t i = 0; i < reader_size; ++i)
@@ -136,6 +134,7 @@ static void icvContourMoments( CvSeq* contour, CvMoments* mom )
             }
         }
 
+        cv::ocl::oclMat dst_a(10, lpt, CV_64FC1);
         cv::ocl::oclMat reader_oclmat(reader_mat);
         int llength = std::min(lpt,128);
         size_t localThreads[3]  = { llength, 1, 1};
@@ -143,48 +142,43 @@ static void icvContourMoments( CvSeq* contour, CvMoments* mom )
         std::vector<std::pair<size_t , const void *> > args;
         args.push_back( std::make_pair( sizeof(cl_int) , (void *)&contour->total ));
         args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&reader_oclmat.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a00.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a10.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a01.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a20.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a11.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a02.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a30.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a21.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a12.data ));
-        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a03.data ));
-        openCLExecuteKernel(dst_a00.clCxt, &moments, "icvContourMoments", globalThreads, localThreads, args, -1, -1);
+        args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_a.data ));
+        cl_int dst_step = (cl_int)dst_a.step;
+        args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step ));
 
-        cv::Mat dst(dst_a00);
-        cv::Scalar s = cv::sum(dst);
-        a00 = s[0];
-        dst = dst_a10;
-        s = cv::sum(dst);
-        a10 = s[0];//dstsum[1];
-        dst = dst_a01;
-        s = cv::sum(dst);
-        a01 = s[0];//dstsum[2];
-        dst = dst_a20;
-        s = cv::sum(dst);
-        a20 = s[0];//dstsum[3];
-        dst = dst_a11;
-        s = cv::sum(dst);
-        a11 = s[0];//dstsum[4];
-        dst = dst_a02;
-        s = cv::sum(dst);
-        a02 = s[0];//dstsum[5];
-        dst = dst_a30;
-        s = cv::sum(dst);
-        a30 = s[0];//dstsum[6];
-        dst = dst_a21;
-        s = cv::sum(dst);
-        a21 = s[0];//dstsum[7];
-        dst = dst_a12;
-        s = cv::sum(dst);
-        a12 = s[0];//dstsum[8];
-        dst = dst_a03;
-        s = cv::sum(dst);
-        a03 = s[0];//dstsum[9];
+        openCLExecuteKernel2(dst_a.clCxt, &moments, "icvContourMoments", globalThreads, localThreads, args, -1, -1);
+
+        cv::Mat dst(dst_a);
+        a00 = a10 = a01 = a20 = a11 = a02 = a30 = a21 = a12 = a03 = 0.0;
+        if (!cv::ocl::Context::getContext()->supportsFeature(Context::CL_DOUBLE))
+        {
+            for (int i = 0; i < contour->total; ++i)
+            {
+                a00 += dst.at<cl_long>(0, i);
+                a10 += dst.at<cl_long>(1, i);
+                a01 += dst.at<cl_long>(2, i);
+                a20 += dst.at<cl_long>(3, i);
+                a11 += dst.at<cl_long>(4, i);
+                a02 += dst.at<cl_long>(5, i);
+                a30 += dst.at<cl_long>(6, i);
+                a21 += dst.at<cl_long>(7, i);
+                a12 += dst.at<cl_long>(8, i);
+                a03 += dst.at<cl_long>(9, i);
+            }
+        }
+        else
+        {
+            a00 = cv::sum(dst.row(0))[0];
+            a10 = cv::sum(dst.row(1))[0];
+            a01 = cv::sum(dst.row(2))[0];
+            a20 = cv::sum(dst.row(3))[0];
+            a11 = cv::sum(dst.row(4))[0];
+            a02 = cv::sum(dst.row(5))[0];
+            a30 = cv::sum(dst.row(6))[0];
+            a21 = cv::sum(dst.row(7))[0];
+            a12 = cv::sum(dst.row(8))[0];
+            a03 = cv::sum(dst.row(9))[0];
+        }
 
         double db1_2, db1_6, db1_12, db1_24, db1_20, db1_60;
         if( fabs(a00) > FLT_EPSILON )
@@ -275,7 +269,7 @@ static void ocl_cvMoments( const void* array, CvMoments* mom, int binary )
     if( size.width <= 0 || size.height <= 0 )
         return;
 
-    cv::Mat src0(mat);
+    cv::Mat src0 = cv::cvarrToMat(mat);
     cv::ocl::oclMat src(src0);
     cv::Size tileSize;
     int blockx,blocky;
@@ -287,17 +281,8 @@ static void ocl_cvMoments( const void* array, CvMoments* mom, int binary )
         blocky = size.height/TILE_SIZE;
     else
         blocky = size.height/TILE_SIZE + 1;
-    cv::ocl::oclMat dst_m00(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m10(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m01(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m20(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m11(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m02(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m30(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m21(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m12(blocky, blockx, CV_64FC1);
-    cv::ocl::oclMat dst_m03(blocky, blockx, CV_64FC1);
-    cl_mem sum = openCLCreateBuffer(src.clCxt,CL_MEM_READ_WRITE,10*sizeof(double));
+    oclMat dst_m(blocky * 10, blockx, CV_64FC1);
+    oclMat sum(1, 10, CV_64FC1);
     int tile_width  = std::min(size.width,TILE_SIZE);
     int tile_height = std::min(size.height,TILE_SIZE);
     size_t localThreads[3]  = { tile_height, 1, 1};
@@ -307,27 +292,16 @@ static void ocl_cvMoments( const void* array, CvMoments* mom, int binary )
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.step ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&tileSize.width ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&tileSize.height ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m00.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m10.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m01.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m20.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m11.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m02.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m30.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m21.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m12.data ));
-    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m03.data ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_m00.cols ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_m00.step ));
-    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&type ));
+    args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m.data ));
+    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_m.cols ));
+    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_m.step ));
+    args.push_back( std::make_pair( sizeof(cl_int) , (void *)&blocky ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&depth ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&cn ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&coi ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&binary ));
     args.push_back( std::make_pair( sizeof(cl_int) , (void *)&TILE_SIZE ));
-    openCLExecuteKernel(dst_m00.clCxt, &moments, "CvMoments", globalThreads, localThreads, args, -1, depth);
+    openCLExecuteKernel2(Context::getContext(), &moments, "CvMoments", globalThreads, localThreads, args, -1, depth);
 
     size_t localThreadss[3]  = { 128, 1, 1};
     size_t globalThreadss[3] = { 128, 1, 1};
@@ -336,35 +310,27 @@ static void ocl_cvMoments( const void* array, CvMoments* mom, int binary )
     args_sum.push_back( std::make_pair( sizeof(cl_int) , (void *)&tile_height ));
     args_sum.push_back( std::make_pair( sizeof(cl_int) , (void *)&tile_width ));
     args_sum.push_back( std::make_pair( sizeof(cl_int) , (void *)&TILE_SIZE ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&sum ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m00.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m10.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m01.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m20.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m11.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m02.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m30.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m21.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m12.data ));
-    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m03.data ));
-    openCLExecuteKernel(dst_m00.clCxt, &moments, "dst_sum", globalThreadss, localThreadss, args_sum, -1, -1);
-    double* dstsum = new double[10];
-    memset(dstsum,0,10*sizeof(double));
-    openCLReadBuffer(dst_m00.clCxt,sum,(void *)dstsum,10*sizeof(double));
-    mom->m00 = dstsum[0];
-    mom->m10 = dstsum[1];
-    mom->m01 = dstsum[2];
-    mom->m20 = dstsum[3];
-    mom->m11 = dstsum[4];
-    mom->m02 = dstsum[5];
-    mom->m30 = dstsum[6];
-    mom->m21 = dstsum[7];
-    mom->m12 = dstsum[8];
-    mom->m03 = dstsum[9];
+    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&sum.data ));
+    args_sum.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst_m.data ));
+    args_sum.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_m.step ));
+    openCLExecuteKernel2(Context::getContext(), &moments, "dst_sum", globalThreadss, localThreadss, args_sum, -1, -1);
+
+    Mat dstsum(sum);
+    mom->m00 = dstsum.at<double>(0, 0);
+    mom->m10 = dstsum.at<double>(0, 1);
+    mom->m01 = dstsum.at<double>(0, 2);
+    mom->m20 = dstsum.at<double>(0, 3);
+    mom->m11 = dstsum.at<double>(0, 4);
+    mom->m02 = dstsum.at<double>(0, 5);
+    mom->m30 = dstsum.at<double>(0, 6);
+    mom->m21 = dstsum.at<double>(0, 7);
+    mom->m12 = dstsum.at<double>(0, 8);
+    mom->m03 = dstsum.at<double>(0, 9);
 
     icvCompleteMomentState( mom );
 }
 
+
 Moments ocl_moments( InputArray _array, bool binaryImage )
 {
     CvMoments om;
diff --git a/modules/ocl/src/kernels/arithm_2_mat.cl b/modules/ocl/src/opencl/arithm_2_mat.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_2_mat.cl
rename to modules/ocl/src/opencl/arithm_2_mat.cl
diff --git a/modules/ocl/src/kernels/arithm_LUT.cl b/modules/ocl/src/opencl/arithm_LUT.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_LUT.cl
rename to modules/ocl/src/opencl/arithm_LUT.cl
diff --git a/modules/ocl/src/kernels/arithm_absdiff.cl b/modules/ocl/src/opencl/arithm_absdiff.cl
similarity index 93%
rename from modules/ocl/src/kernels/arithm_absdiff.cl
rename to modules/ocl/src/opencl/arithm_absdiff.cl
index 37f154216a..341a0048ff 100644
--- a/modules/ocl/src/kernels/arithm_absdiff.cl
+++ b/modules/ocl/src/opencl/arithm_absdiff.cl
@@ -44,7 +44,11 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -63,6 +67,9 @@ __kernel void arithm_absdiff_D0 (__global uchar *src1, int src1_step, int src1_o
     {
         x = x << 2;
 
+#ifdef dst_align
+#undef dst_align
+#endif
         #define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
         int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
@@ -111,7 +118,10 @@ __kernel void arithm_absdiff_D2 (__global ushort *src1, int src1_step, int src1_
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -145,7 +155,10 @@ __kernel void arithm_absdiff_D3 (__global short *src1, int src1_step, int src1_o
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -249,7 +262,10 @@ __kernel void arithm_s_absdiff_C1_D0 (__global   uchar *src1, int src1_step, int
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -288,7 +304,10 @@ __kernel void arithm_s_absdiff_C1_D2 (__global   ushort *src1, int src1_step, in
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -319,7 +338,10 @@ __kernel void arithm_s_absdiff_C1_D3 (__global   short *src1, int src1_step, int
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -387,8 +409,8 @@ __kernel void arithm_s_absdiff_C1_D5 (__global   float *src1, int src1_step, int
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_s_absdiff_C1_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                     __global   double *dst,  int dst_step,  int dst_offset,
-                                     double4 src2, int rows, int cols, int dst_step1)
+                                      __global   double *dst,  int dst_step,  int dst_offset,
+                                      double4 src2, int rows, int cols, int dst_step1)
 {
 
     int x = get_global_id(0);
@@ -422,7 +444,10 @@ __kernel void arithm_s_absdiff_C2_D0 (__global   uchar *src1, int src1_step, int
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -465,7 +490,7 @@ __kernel void arithm_s_absdiff_C2_D2 (__global   ushort *src1, int src1_step, in
 }
 __kernel void arithm_s_absdiff_C2_D3 (__global   short *src1, int src1_step, int src1_offset,
                                       __global   short *dst,  int dst_step,  int dst_offset,
-                                     int4 src2, int rows, int cols, int dst_step1)
+                                      int4 src2, int rows, int cols, int dst_step1)
 {
 
     int x = get_global_id(0);
@@ -509,7 +534,7 @@ __kernel void arithm_s_absdiff_C2_D4 (__global   int *src1, int src1_step, int s
 }
 __kernel void arithm_s_absdiff_C2_D5 (__global   float *src1, int src1_step, int src1_offset,
                                       __global   float *dst,  int dst_step,  int dst_offset,
-                                     float4 src2, int rows, int cols, int dst_step1)
+                                      float4 src2, int rows, int cols, int dst_step1)
 {
 
     int x = get_global_id(0);
@@ -564,7 +589,10 @@ __kernel void arithm_s_absdiff_C3_D0 (__global   uchar *src1, int src1_step, int
     {
         x = x << 2;
 
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (((dst_offset % dst_step) / 3 ) & 3)
         int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -618,7 +646,10 @@ __kernel void arithm_s_absdiff_C3_D2 (__global   ushort *src1, int src1_step, in
     {
         x = x << 1;
 
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (((dst_offset % dst_step) / 6 ) & 1)
         int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -644,16 +675,16 @@ __kernel void arithm_s_absdiff_C3_D2 (__global   ushort *src1, int src1_step, in
         data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
 
         data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
+                    ? tmp_data_1.x : data_1.x;
         data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
+                    ? tmp_data_1.y : data_1.y;
 
         data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
+                    ? tmp_data_2.xy : data_2.xy;
 
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
+        *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
+        *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
+        *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
     }
 }
 __kernel void arithm_s_absdiff_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
@@ -668,7 +699,10 @@ __kernel void arithm_s_absdiff_C3_D3 (__global   short *src1, int src1_step, int
     {
         x = x << 1;
 
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (((dst_offset % dst_step) / 6 ) & 1)
         int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -694,16 +728,16 @@ __kernel void arithm_s_absdiff_C3_D3 (__global   short *src1, int src1_step, int
         data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
 
         data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
+                    ? tmp_data_1.x : data_1.x;
         data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
+                    ? tmp_data_1.y : data_1.y;
 
         data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
+                    ? tmp_data_2.xy : data_2.xy;
 
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
+        *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
+        *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
+        *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
     }
 }
 __kernel void arithm_s_absdiff_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
@@ -735,9 +769,9 @@ __kernel void arithm_s_absdiff_C3_D4 (__global   int *src1, int src1_step, int s
         int tmp_data_1 = convert_int_sat(abs_diff(src1_data_1, src2_data_1));
         int tmp_data_2 = convert_int_sat(abs_diff(src1_data_2, src2_data_2));
 
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
+        *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
+        *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
+        *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
     }
 }
 __kernel void arithm_s_absdiff_C3_D5 (__global   float *src1, int src1_step, int src1_offset,
@@ -769,9 +803,9 @@ __kernel void arithm_s_absdiff_C3_D5 (__global   float *src1, int src1_step, int
         float tmp_data_1 = fabs(src1_data_1 - src2_data_1);
         float tmp_data_2 = fabs(src1_data_2 - src2_data_2);
 
-       *((__global float *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= tmp_data_2;
+        *((__global float *)((__global char *)dst + dst_index + 0))= tmp_data_0;
+        *((__global float *)((__global char *)dst + dst_index + 4))= tmp_data_1;
+        *((__global float *)((__global char *)dst + dst_index + 8))= tmp_data_2;
     }
 }
 
@@ -805,9 +839,9 @@ __kernel void arithm_s_absdiff_C3_D6 (__global   double *src1, int src1_step, in
         double tmp_data_1 = fabs(src1_data_1 - src2_data_1);
         double tmp_data_2 = fabs(src1_data_2 - src2_data_2);
 
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2;
+        *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
+        *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
+        *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2;
     }
 }
 #endif
diff --git a/modules/ocl/src/kernels/arithm_add.cl b/modules/ocl/src/opencl/arithm_add.cl
similarity index 66%
rename from modules/ocl/src/kernels/arithm_add.cl
rename to modules/ocl/src/opencl/arithm_add.cl
index 789a42444c..070ced4731 100644
--- a/modules/ocl/src/kernels/arithm_add.cl
+++ b/modules/ocl/src/opencl/arithm_add.cl
@@ -45,9 +45,18 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
+#ifdef ARITHM_ADD
+  #define ARITHM_OP(A,B) ((A)+(B))
+#elif defined ARITHM_SUB
+  #define ARITHM_OP(A,B) ((A)-(B))
+#endif
 //////////////////////////////////////////////////////////////////////////////////////////////////////
 /////////////////////////////////////////////ADD////////////////////////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -64,7 +73,10 @@ __kernel void arithm_add_D0 (__global uchar *src1, int src1_step, int src1_offse
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
         int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
@@ -88,7 +100,7 @@ __kernel void arithm_add_D0 (__global uchar *src1, int src1_step, int src1_offse
             src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
         }
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
-        short4 tmp      = convert_short4_sat(src1_data) + convert_short4_sat(src2_data);
+        short4 tmp      = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data));
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -112,7 +124,10 @@ __kernel void arithm_add_D2 (__global ushort *src1, int src1_step, int src1_offs
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -124,7 +139,7 @@ __kernel void arithm_add_D2 (__global ushort *src1, int src1_step, int src1_offs
         ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
 
         ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-        int4    tmp = convert_int4_sat(src1_data) + convert_int4_sat(src2_data);
+        int4    tmp = ARITHM_OP(convert_int4_sat(src1_data), convert_int4_sat(src2_data));
         ushort4 tmp_data = convert_ushort4_sat(tmp);
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -147,7 +162,10 @@ __kernel void arithm_add_D3 (__global short *src1, int src1_step, int src1_offse
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -159,7 +177,7 @@ __kernel void arithm_add_D3 (__global short *src1, int src1_step, int src1_offse
         short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
 
         short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-        int4   tmp = convert_int4_sat(src1_data) + convert_int4_sat(src2_data);
+        int4   tmp = ARITHM_OP(convert_int4_sat(src1_data), convert_int4_sat(src2_data));
         short4 tmp_data = convert_short4_sat(tmp);
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -187,7 +205,7 @@ __kernel void arithm_add_D4 (__global int *src1, int src1_step, int src1_offset,
 
         int data1 = *((__global int *)((__global char *)src1 + src1_index));
         int data2 = *((__global int *)((__global char *)src2 + src2_index));
-        long tmp  = (long)(data1) + (long)(data2);
+        long tmp  = ARITHM_OP((long)(data1), (long)(data2));
 
         *((__global int *)((__global char *)dst + dst_index)) = convert_int_sat(tmp);
     }
@@ -208,7 +226,7 @@ __kernel void arithm_add_D5 (__global float *src1, int src1_step, int src1_offse
 
         float data1 = *((__global float *)((__global char *)src1 + src1_index));
         float data2 = *((__global float *)((__global char *)src2 + src2_index));
-        float tmp = data1 + data2;
+        float tmp = ARITHM_OP(data1, data2);
 
         *((__global float *)((__global char *)dst + dst_index)) = tmp;
     }
@@ -232,7 +250,7 @@ __kernel void arithm_add_D6 (__global double *src1, int src1_step, int src1_offs
         double data1 = *((__global double *)((__global char *)src1 + src1_index));
         double data2 = *((__global double *)((__global char *)src2 + src2_index));
 
-        *((__global double *)((__global char *)dst + dst_index)) = data1 + data2;
+        *((__global double *)((__global char *)dst + dst_index)) = ARITHM_OP(data1, data2);
     }
 }
 #endif
@@ -252,7 +270,10 @@ __kernel void arithm_add_with_mask_C1_D0 (__global uchar *src1, int src1_step, i
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
         int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@@ -286,7 +307,7 @@ __kernel void arithm_add_with_mask_C1_D0 (__global uchar *src1, int src1_step, i
         }
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        short4 tmp = convert_short4_sat(src1_data) + convert_short4_sat(src2_data);
+        short4 tmp = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data));
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
@@ -311,7 +332,10 @@ __kernel void arithm_add_with_mask_C1_D2 (__global ushort *src1, int src1_step,
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@@ -325,7 +349,7 @@ __kernel void arithm_add_with_mask_C1_D2 (__global ushort *src1, int src1_step,
         uchar2  mask_data = vload2(0, mask + mask_index);
 
         ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) + convert_int2_sat(src2_data);
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), convert_int2_sat(src2_data));
         ushort2 tmp_data = convert_ushort2_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
@@ -348,7 +372,10 @@ __kernel void arithm_add_with_mask_C1_D3 (__global short *src1, int src1_step, i
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@@ -362,7 +389,7 @@ __kernel void arithm_add_with_mask_C1_D3 (__global short *src1, int src1_step, i
         uchar2  mask_data = vload2(0, mask + mask_index);
 
         short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) + convert_int2_sat(src2_data);
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), convert_int2_sat(src2_data));
         short2 tmp_data = convert_short2_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
@@ -394,7 +421,7 @@ __kernel void arithm_add_with_mask_C1_D4 (__global int   *src1, int src1_step, i
         int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
         int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
 
-        int data = convert_int_sat((long)src_data1 + (long)src_data2);
+        int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2));
         data = mask_data ? data : dst_data;
 
         *((__global int *)((__global char *)dst + dst_index)) = data;
@@ -424,7 +451,7 @@ __kernel void arithm_add_with_mask_C1_D5 (__global float *src1, int src1_step, i
         float src_data2 = *((__global float *)((__global char *)src2 + src2_index));
         float dst_data  = *((__global float *)((__global char *)dst  + dst_index));
 
-        float data = src_data1 + src_data2;
+        float data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global float *)((__global char *)dst + dst_index)) = data;
@@ -455,7 +482,7 @@ __kernel void arithm_add_with_mask_C1_D6 (__global double *src1, int src1_step,
         double src_data2 = *((__global double *)((__global char *)src2 + src2_index));
         double dst_data  = *((__global double *)((__global char *)dst  + dst_index));
 
-        double data = src_data1 + src_data2;
+        double data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global double *)((__global char *)dst + dst_index)) = data;
@@ -477,7 +504,10 @@ __kernel void arithm_add_with_mask_C2_D0 (__global uchar *src1, int src1_step, i
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@@ -491,7 +521,7 @@ __kernel void arithm_add_with_mask_C2_D0 (__global uchar *src1, int src1_step, i
         uchar2 mask_data = vload2(0, mask + mask_index);
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        short4   tmp = convert_short4_sat(src1_data) + convert_short4_sat(src2_data);
+        short4   tmp = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data));
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
@@ -523,7 +553,7 @@ __kernel void arithm_add_with_mask_C2_D2 (__global ushort *src1, int src1_step,
         ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
         ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + convert_int2_sat(src_data2);
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), convert_int2_sat(src_data2));
         ushort2 data = convert_ushort2_sat(tmp);
         data = mask_data ? data : dst_data;
 
@@ -553,7 +583,7 @@ __kernel void arithm_add_with_mask_C2_D3 (__global short *src1, int src1_step, i
         short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
         short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + convert_int2_sat(src_data2);
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), convert_int2_sat(src_data2));
         short2 data = convert_short2_sat(tmp);
         data = mask_data ? data : dst_data;
 
@@ -583,7 +613,7 @@ __kernel void arithm_add_with_mask_C2_D4 (__global int   *src1, int src1_step, i
         int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
         int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
 
-        int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2));
+        int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global int2 *)((__global char *)dst + dst_index)) = data;
@@ -612,7 +642,7 @@ __kernel void arithm_add_with_mask_C2_D5 (__global float *src1, int src1_step, i
         float2 src_data2 = *((__global float2 *)((__global char *)src2 + src2_index));
         float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
 
-        float2 data = src_data1 + src_data2;
+        float2 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global float2 *)((__global char *)dst + dst_index)) = data;
@@ -643,320 +673,14 @@ __kernel void arithm_add_with_mask_C2_D6 (__global double *src1, int src1_step,
         double2 src_data2 = *((__global double2 *)((__global char *)src2 + src2_index));
         double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
 
-        double2 data = src_data1 + src_data2;
+        double2 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global double2 *)((__global char *)dst + dst_index)) = data;
     }
 }
 #endif
-__kernel void arithm_add_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                                          __global uchar *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global uchar *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
 
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0);
-        uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4);
-        uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(convert_short4_sat(src1_data_0) + convert_short4_sat(src2_data_0));
-        uchar4 tmp_data_1 = convert_uchar4_sat(convert_short4_sat(src1_data_1) + convert_short4_sat(src2_data_1));
-        uchar4 tmp_data_2 = convert_uchar4_sat(convert_short4_sat(src1_data_2) + convert_short4_sat(src2_data_2));
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_add_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                                          __global ushort *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global ushort *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0));
-        ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4));
-        ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + convert_int2_sat(src2_data_0));
-        ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + convert_int2_sat(src2_data_1));
-        ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + convert_int2_sat(src2_data_2));
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_add_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset,
-                                          __global short *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global short *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0));
-        short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4));
-        short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8));
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + convert_int2_sat(src2_data_0));
-        short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + convert_int2_sat(src2_data_1));
-        short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + convert_int2_sat(src2_data_2));
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_add_with_mask_C3_D4 (__global int   *src1, int src1_step, int src1_offset,
-                                          __global int   *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global int   *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0));
-        int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4));
-        int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0);
-        int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1);
-        int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2);
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_add_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset,
-                                          __global float *src2, int src2_step, int src2_offset,
-                                          __global uchar *mask, int mask_step, int mask_offset,
-                                          __global float *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 12) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = *((__global float *)((__global char *)src2 + src2_index + 0));
-        float src2_data_1 = *((__global float *)((__global char *)src2 + src2_index + 4));
-        float src2_data_2 = *((__global float *)((__global char *)src2 + src2_index + 8));
-
-        uchar mask_data = * (mask + mask_index);
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_data_0 = src1_data_0 + src2_data_0;
-        float tmp_data_1 = src1_data_1 + src2_data_1;
-        float tmp_data_2 = src1_data_2 + src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_add_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset,
-                                          __global double *src2, int src2_step, int src2_offset,
-                                          __global uchar  *mask, int mask_step, int mask_offset,
-                                          __global double *dst,  int dst_step,  int dst_offset,
-                                          int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int src2_index = mad24(y, src2_step, (x * 24) + src2_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = *((__global double *)((__global char *)src2 + src2_index + 0 ));
-        double src2_data_1 = *((__global double *)((__global char *)src2 + src2_index + 8 ));
-        double src2_data_2 = *((__global double *)((__global char *)src2 + src2_index + 16));
-
-        uchar mask_data = * (mask + mask_index);
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 + src2_data_0;
-        double tmp_data_1 = src1_data_1 + src2_data_1;
-        double tmp_data_2 = src1_data_2 + src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
 __kernel void arithm_add_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset,
                                           __global uchar *src2, int src2_step, int src2_offset,
                                           __global uchar *mask, int mask_step, int mask_offset,
@@ -980,7 +704,7 @@ __kernel void arithm_add_with_mask_C4_D0 (__global uchar *src1, int src1_step, i
         uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
 
-        uchar4 data = convert_uchar4_sat(convert_ushort4_sat(src_data1) + convert_ushort4_sat(src_data2));
+        uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_short4_sat(src_data1), convert_short4_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global uchar4 *)(dst + dst_index)) = data;
@@ -1009,7 +733,7 @@ __kernel void arithm_add_with_mask_C4_D2 (__global ushort *src1, int src1_step,
         ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
         ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
 
-        ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + convert_int4_sat(src_data2));
+        ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), convert_int4_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
@@ -1038,7 +762,7 @@ __kernel void arithm_add_with_mask_C4_D3 (__global short *src1, int src1_step, i
         short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
         short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
 
-        short4 data = convert_short4_sat(convert_int4_sat(src_data1) + convert_int4_sat(src_data2));
+        short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), convert_int4_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global short4 *)((__global char *)dst + dst_index)) = data;
@@ -1067,7 +791,7 @@ __kernel void arithm_add_with_mask_C4_D4 (__global int   *src1, int src1_step, i
         int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
         int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
 
-        int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src_data2));
+        int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global int4 *)((__global char *)dst + dst_index)) = data;
@@ -1096,7 +820,7 @@ __kernel void arithm_add_with_mask_C4_D5 (__global float *src1, int src1_step, i
         float4 src_data2 = *((__global float4 *)((__global char *)src2 + src2_index));
         float4 dst_data  = *((__global float4 *)((__global char *)dst  + dst_index));
 
-        float4 data = src_data1 + src_data2;
+        float4 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global float4 *)((__global char *)dst + dst_index)) = data;
@@ -1127,7 +851,7 @@ __kernel void arithm_add_with_mask_C4_D6 (__global double *src1, int src1_step,
         double4 src_data2 = *((__global double4 *)((__global char *)src2 + src2_index));
         double4 dst_data  = *((__global double4 *)((__global char *)dst  + dst_index));
 
-        double4 data = src_data1 + src_data2;
+        double4 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global double4 *)((__global char *)dst + dst_index)) = data;
diff --git a/modules/ocl/src/kernels/arithm_addWeighted.cl b/modules/ocl/src/opencl/arithm_addWeighted.cl
similarity index 69%
rename from modules/ocl/src/kernels/arithm_addWeighted.cl
rename to modules/ocl/src/opencl/arithm_addWeighted.cl
index 7e9df6f253..e7ed289281 100644
--- a/modules/ocl/src/kernels/arithm_addWeighted.cl
+++ b/modules/ocl/src/opencl/arithm_addWeighted.cl
@@ -42,8 +42,12 @@
 // the use of this software, even if advised of the possibility of such damage.
 //
 //M*/
-#if defined DOUBLE_SUPPORT
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 typedef double F;
 #else
 typedef float F;
@@ -52,42 +56,45 @@ typedef float F;
 /////////////////////////////////////////////addWeighted//////////////////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 __kernel void addWeighted_D0 (__global uchar *src1,int src1_step,int src1_offset,
-                           __global uchar *src2, int src2_step,int src2_offset,
-                           F alpha,F beta,F gama,
-                           __global uchar *dst,  int dst_step,int dst_offset,
-                           int rows,  int cols,int dst_step1)
+                              __global uchar *src2, int src2_step,int src2_offset,
+                              F alpha,F beta,F gama,
+                              __global uchar *dst,  int dst_step,int dst_offset,
+                              int rows,  int cols,int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
 
         x = x << 2;
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		uchar4 src1_data ,src2_data;
+        uchar4 src1_data ,src2_data;
 
-		src1_data.x= src1_index+0 >= 0 ? src1[src1_index+0] : 0;
-		src1_data.y= src1_index+1 >= 0 ? src1[src1_index+1] : 0;
-		src1_data.z= src1_index+2 >= 0 ? src1[src1_index+2] : 0;
-		src1_data.w= src1_index+3 >= 0 ? src1[src1_index+3] : 0;
+        src1_data.x= src1_index+0 >= 0 ? src1[src1_index+0] : 0;
+        src1_data.y= src1_index+1 >= 0 ? src1[src1_index+1] : 0;
+        src1_data.z= src1_index+2 >= 0 ? src1[src1_index+2] : 0;
+        src1_data.w= src1_index+3 >= 0 ? src1[src1_index+3] : 0;
 
-		src2_data.x= src2_index+0 >= 0 ? src2[src2_index+0] : 0;
-		src2_data.y= src2_index+1 >= 0 ? src2[src2_index+1] : 0;
-		src2_data.z= src2_index+2 >= 0 ? src2[src2_index+2] : 0;
-		src2_data.w= src2_index+3 >= 0 ? src2[src2_index+3] : 0;
+        src2_data.x= src2_index+0 >= 0 ? src2[src2_index+0] : 0;
+        src2_data.y= src2_index+1 >= 0 ? src2[src2_index+1] : 0;
+        src2_data.z= src2_index+2 >= 0 ? src2[src2_index+2] : 0;
+        src2_data.w= src2_index+3 >= 0 ? src2[src2_index+3] : 0;
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
 //        short4 tmp      = convert_short4_sat(src1_data) * alpha + convert_short4_sat(src2_data) * beta + gama;
-         short4 tmp;
+        short4 tmp;
         tmp.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp.z = src1_data.z * alpha + src2_data.z * beta + gama;
@@ -100,7 +107,7 @@ __kernel void addWeighted_D0 (__global uchar *src1,int src1_step,int src1_offset
         dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
 
         *((__global uchar4 *)(dst + dst_index)) = dst_data;
-       // dst[x + y * dst_step] = src1[x + y * src1_step] * alpha + src2[x + y * src2_step] * beta + gama;
+        // dst[x + y * dst_step] = src1[x + y * src1_step] * alpha + src2[x + y * src2_step] * beta + gama;
     }
 
 }
@@ -108,48 +115,51 @@ __kernel void addWeighted_D0 (__global uchar *src1,int src1_step,int src1_offset
 
 
 __kernel void addWeighted_D2 (__global ushort *src1, int src1_step,int src1_offset,
-                           __global ushort *src2, int src2_step,int src2_offset,
-                           F alpha,F beta,F gama,
-                           __global ushort *dst,  int dst_step,int dst_offset,
-                           int rows,  int cols,int dst_step1)
+                              __global ushort *src2, int src2_step,int src2_offset,
+                              F alpha,F beta,F gama,
+                              __global ushort *dst,  int dst_step,int dst_offset,
+                              int rows,  int cols,int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
 
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset -( dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset -( dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset -( dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset -( dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset +( x<< 1) & (int)0xfffffff8);
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-    int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index_fix));
         ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index_fix));
-    if(src1_index < 0)
-    {
-        ushort4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-    if(src2_index < 0)
-    {
-        ushort4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-    }
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
 
         ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-       // int4 tmp      = convert_int4_sat(src1_data) * alpha + convert_int4_sat(src2_data) * beta + gama;
-         int4 tmp;
+        // int4 tmp      = convert_int4_sat(src1_data) * alpha + convert_int4_sat(src2_data) * beta + gama;
+        int4 tmp;
         tmp.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp.z = src1_data.z * alpha + src2_data.z * beta + gama;
@@ -177,39 +187,42 @@ __kernel void addWeighted_D3 (__global short *src1, int src1_step,int src1_offse
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
 
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset -( dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset -( dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset -( dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset -( dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset +( x<< 1) - (dst_align << 1 ));
 
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-    int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index_fix));
         short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index_fix));
 
-    if(src1_index < 0)
-    {
-        short4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-    if(src2_index < 0)
-    {
-        short4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-    }
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
         short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-       // int4 tmp      = convert_int4_sat(src1_data) * alpha + convert_int4_sat(src2_data) * beta + gama;
-         int4 tmp;
+        // int4 tmp      = convert_int4_sat(src1_data) * alpha + convert_int4_sat(src2_data) * beta + gama;
+        int4 tmp;
         tmp.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp.z = src1_data.z * alpha + src2_data.z * beta + gama;
@@ -228,7 +241,7 @@ __kernel void addWeighted_D3 (__global short *src1, int src1_step,int src1_offse
 
 __kernel void addWeighted_D4 (__global int *src1, int src1_step,int src1_offset,
                               __global int *src2, int src2_step,int src2_offset,
-                             F alpha,F beta, F gama,
+                              F alpha,F beta, F gama,
                               __global int *dst,  int dst_step,int dst_offset,
                               int rows,  int cols,int dst_step1)
 {
@@ -236,42 +249,45 @@ __kernel void addWeighted_D4 (__global int *src1, int src1_step,int src1_offset,
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
-            
+
         x = x << 2;
 
-        #define bitOfInt  (sizeof(int)== 4 ? 2: 3)
+#define bitOfInt  (sizeof(int)== 4 ? 2: 3)
 
-        #define dst_align ((dst_offset >> bitOfInt) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> bitOfInt) & 3)
+
+        int src1_index = mad24(y, src1_step, (x << bitOfInt) + src1_offset - (dst_align << bitOfInt));
+        int src2_index = mad24(y, src2_step, (x << bitOfInt) + src2_offset - (dst_align << bitOfInt));
 
-        int src1_index = mad24(y, src1_step, (x << bitOfInt) + src1_offset - (dst_align << bitOfInt)); 
-        int src2_index = mad24(y, src2_step, (x << bitOfInt) + src2_offset - (dst_align << bitOfInt)); 
-       
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << bitOfInt) -(dst_align << bitOfInt));
 
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-    int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index_fix));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index_fix));
-        
-    if(src1_index < 0)
-    {
-        int4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-    if(src2_index < 0)
-    {
-        int4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-    }
+
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
         int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
-       // double4   tmp = convert_double4(src1_data) * alpha + convert_double4(src2_data) * beta + gama ;
-         float4 tmp;
+        // double4   tmp = convert_double4(src1_data) * alpha + convert_double4(src2_data) * beta + gama ;
+        float4 tmp;
         tmp.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp.z = src1_data.z * alpha + src2_data.z * beta + gama;
@@ -291,7 +307,7 @@ __kernel void addWeighted_D4 (__global int *src1, int src1_step,int src1_offset,
 
 __kernel void addWeighted_D5 (__global float *src1,int src1_step,int src1_offset,
                               __global float *src2, int src2_step,int src2_offset,
-                             F alpha,F beta, F gama,
+                              F alpha,F beta, F gama,
                               __global float *dst,  int dst_step,int dst_offset,
                               int rows,  int cols,int dst_step1)
 {
@@ -299,46 +315,49 @@ __kernel void addWeighted_D5 (__global float *src1,int src1_step,int src1_offset
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
-            
+
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 2) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
-       
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 2) -(dst_align << 2));
 
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-    int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float  *)((__global char *)src1 + src1_index_fix));
         float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
         float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index));
-    if(src1_index < 0)
-    {
-        float4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-    if(src2_index < 0)
-    {
-        float4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-    }
-    //    double4   tmp = convert_double4(src1_data) * alpha + convert_double4(src2_data) * beta + gama ;
+        if(src1_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+        //    double4   tmp = convert_double4(src1_data) * alpha + convert_double4(src2_data) * beta + gama ;
 
-       // float4   tmp_data =(src1_data) * alpha + (src2_data) * beta + gama ;
-         float4 tmp_data;
+        // float4   tmp_data =(src1_data) * alpha + (src2_data) * beta + gama ;
+        float4 tmp_data;
         tmp_data.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp_data.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp_data.z = src1_data.z * alpha + src2_data.z * beta + gama;
         tmp_data.w = src1_data.w * alpha + src2_data.w * beta + gama;
-       // float4 tmp_data = convert_float4(tmp);
+        // float4 tmp_data = convert_float4(tmp);
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.y : dst_data.y;
@@ -353,7 +372,7 @@ __kernel void addWeighted_D5 (__global float *src1,int src1_step,int src1_offset
 #if defined (DOUBLE_SUPPORT)
 __kernel void addWeighted_D6 (__global double *src1, int src1_step,int src1_offset,
                               __global double *src2, int src2_step,int src2_offset,
-                             F alpha,F beta, F gama,
+                              F alpha,F beta, F gama,
                               __global double *dst,  int dst_step,int dst_offset,
                               int rows,  int cols,int dst_step1)
 {
@@ -361,39 +380,42 @@ __kernel void addWeighted_D6 (__global double *src1, int src1_step,int src1_offs
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
     {
-            
+
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 3) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3) & 3)
+
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
-       
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 3) -(dst_align << 3));
 
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-    int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double  *)((__global char *)src1 + src1_index_fix));
         double4 src2_data = vload4(0, (__global double  *)((__global char *)src2 + src2_index_fix));
         double4 dst_data = *((__global double4 *)((__global char *)dst + dst_index));
-    if(src1_index < 0)
-    {
-        double4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-    if(src2_index < 0)
-    {
-        double4 tmp;
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-    }
-      //  double4   tmp_data = (src1_data) * alpha + (src2_data) * beta + gama ;
-         double4 tmp_data;
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+        //  double4   tmp_data = (src1_data) * alpha + (src2_data) * beta + gama ;
+        double4 tmp_data;
         tmp_data.x = src1_data.x * alpha + src2_data.x * beta + gama;
         tmp_data.y = src1_data.y * alpha + src2_data.y * beta + gama;
         tmp_data.z = src1_data.z * alpha + src2_data.z * beta + gama;
diff --git a/modules/ocl/src/kernels/arithm_add_scalar.cl b/modules/ocl/src/opencl/arithm_add_scalar.cl
similarity index 57%
rename from modules/ocl/src/kernels/arithm_add_scalar.cl
rename to modules/ocl/src/opencl/arithm_add_scalar.cl
index 05b813dc8c..cdb79f37ed 100644
--- a/modules/ocl/src/kernels/arithm_add_scalar.cl
+++ b/modules/ocl/src/opencl/arithm_add_scalar.cl
@@ -44,9 +44,18 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
+#ifdef ARITHM_ADD
+  #define ARITHM_OP(A,B) ((A)+(B))
+#elif defined ARITHM_SUB
+  #define ARITHM_OP(A,B) ((A)-(B))
+#endif
 /**************************************add with scalar without mask**************************************/
 __kernel void arithm_s_add_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
                                   __global   uchar *dst,  int dst_step,  int dst_offset,
@@ -59,7 +68,10 @@ __kernel void arithm_s_add_C1_D0 (__global   uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -76,7 +88,7 @@ __kernel void arithm_s_add_C1_D0 (__global   uchar *src1, int src1_step, int src
         }
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) + src2_data;
+        int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data);
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
@@ -99,7 +111,10 @@ __kernel void arithm_s_add_C1_D2 (__global   ushort *src1, int src1_step, int sr
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -110,7 +125,7 @@ __kernel void arithm_s_add_C1_D2 (__global   ushort *src1, int src1_step, int sr
         int2 src2_data = (int2)(src2.x, src2.x);
 
         ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) + src2_data;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data);
         ushort2 tmp_data = convert_ushort2_sat(tmp);
 
         data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
@@ -131,7 +146,10 @@ __kernel void arithm_s_add_C1_D3 (__global   short *src1, int src1_step, int src
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -142,7 +160,7 @@ __kernel void arithm_s_add_C1_D3 (__global   short *src1, int src1_step, int src
         int2 src2_data = (int2)(src2.x, src2.x);
         short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
 
-        int2    tmp = convert_int2_sat(src1_data) + src2_data;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data);
         short2 tmp_data = convert_short2_sat(tmp);
 
         data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
@@ -168,7 +186,7 @@ __kernel void arithm_s_add_C1_D4 (__global   int *src1, int src1_step, int src1_
         int src_data2 = src2.x;
         int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
 
-        int data = convert_int_sat((long)src_data1 + (long)src_data2);
+        int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2));
 
         *((__global int *)((__global char *)dst + dst_index)) = data;
     }
@@ -190,7 +208,7 @@ __kernel void arithm_s_add_C1_D5 (__global   float *src1, int src1_step, int src
         float src_data2 = src2.x;
         float dst_data  = *((__global float *)((__global char *)dst  + dst_index));
 
-        float data = src_data1 + src_data2;
+        float data = ARITHM_OP(src_data1, src_data2);
 
         *((__global float *)((__global char *)dst + dst_index)) = data;
     }
@@ -214,7 +232,7 @@ __kernel void arithm_s_add_C1_D6 (__global   double *src1, int src1_step, int sr
         double src2_data = src2.x;
         double dst_data  = *((__global double *)((__global char *)dst  + dst_index));
 
-        double data = src_data1 + src2_data;
+        double data = ARITHM_OP(src_data1, src2_data);
 
         *((__global double *)((__global char *)dst + dst_index)) = data;
     }
@@ -233,7 +251,10 @@ __kernel void arithm_s_add_C2_D0 (__global   uchar *src1, int src1_step, int src
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -244,7 +265,7 @@ __kernel void arithm_s_add_C2_D0 (__global   uchar *src1, int src1_step, int src
         int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y);
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) + src2_data;
+        int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data);
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
@@ -270,7 +291,7 @@ __kernel void arithm_s_add_C2_D2 (__global   ushort *src1, int src1_step, int sr
         int2 src_data2 = (int2)(src2.x, src2.y);
         ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + src_data2;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2);
         ushort2 data = convert_ushort2_sat(tmp);
 
         *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
@@ -293,7 +314,7 @@ __kernel void arithm_s_add_C2_D3 (__global   short *src1, int src1_step, int src
         int2 src_data2 = (int2)(src2.x, src2.y);
         short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + src_data2;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2);
         short2 data = convert_short2_sat(tmp);
 
         *((__global short2 *)((__global char *)dst + dst_index)) = data;
@@ -316,7 +337,7 @@ __kernel void arithm_s_add_C2_D4 (__global   int *src1, int src1_step, int src1_
         int2 src_data2 = (int2)(src2.x, src2.y);
         int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
 
-        int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2));
+        int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2)));
         *((__global int2 *)((__global char *)dst + dst_index)) = data;
     }
 }
@@ -337,7 +358,7 @@ __kernel void arithm_s_add_C2_D5 (__global   float *src1, int src1_step, int src
         float2 src_data2 = (float2)(src2.x, src2.y);
         float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
 
-        float2 data = src_data1 + src_data2;
+        float2 data = ARITHM_OP(src_data1, src_data2);
         *((__global float2 *)((__global char *)dst + dst_index)) = data;
     }
 }
@@ -360,271 +381,13 @@ __kernel void arithm_s_add_C2_D6 (__global   double *src1, int src1_step, int sr
         double2 src_data2 = (double2)(src2.x, src2.y);
         double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
 
-        double2 data = src_data1 + src_data2;
+        double2 data = ARITHM_OP(src_data1, src_data2);
 
         *((__global double2 *)((__global char *)dst + dst_index)) = data;
     }
 }
 #endif
-__kernel void arithm_s_add_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                  __global   uchar *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
 
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x);
-        int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y);
-        int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(convert_int4_sat(src1_data_0) + src2_data_0);
-        uchar4 tmp_data_1 = convert_uchar4_sat(convert_int4_sat(src1_data_1) + src2_data_1);
-        uchar4 tmp_data_2 = convert_uchar4_sat(convert_int4_sat(src1_data_2) + src2_data_2);
-
-        data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_s_add_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                  __global   ushort *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + src2_data_0);
-        ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + src2_data_1);
-        ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + src2_data_2);
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_add_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                  __global   short *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + src2_data_0);
-        short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + src2_data_1);
-        short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + src2_data_2);
-
-        data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_add_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                  __global   int *dst,  int dst_step,  int dst_offset,
-                                  int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0);
-        int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1);
-        int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2);
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-__kernel void arithm_s_add_C3_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                  __global   float *dst,  int dst_step,  int dst_offset,
-                                  float4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = src2.x;
-        float src2_data_1 = src2.y;
-        float src2_data_2 = src2.z;
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_data_0 = src1_data_0 + src2_data_0;
-        float tmp_data_1 = src1_data_1 + src2_data_1;
-        float tmp_data_2 = src1_data_2 + src2_data_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= tmp_data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= tmp_data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= tmp_data_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_add_C3_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                  __global   double *dst,  int dst_step,  int dst_offset,
-                                  double4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = src2.x;
-        double src2_data_1 = src2.y;
-        double src2_data_2 = src2.z;
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 + src2_data_0;
-        double tmp_data_1 = src1_data_1 + src2_data_1;
-        double tmp_data_2 = src1_data_2 + src2_data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2;
-    }
-}
-#endif
 __kernel void arithm_s_add_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
                                   __global   uchar *dst,  int dst_step,  int dst_offset,
                                   int4 src2, int rows, int cols, int dst_step1)
@@ -640,7 +403,7 @@ __kernel void arithm_s_add_C4_D0 (__global   uchar *src1, int src1_step, int src
 
         uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
 
-        uchar4 data = convert_uchar4_sat(convert_int4_sat(src_data1) + src2);
+        uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
 
         *((__global uchar4 *)(dst + dst_index)) = data;
     }
@@ -660,7 +423,7 @@ __kernel void arithm_s_add_C4_D2 (__global   ushort *src1, int src1_step, int sr
 
         ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
 
-        ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + src2);
+        ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
 
         *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
     }
@@ -680,7 +443,7 @@ __kernel void arithm_s_add_C4_D3 (__global   short *src1, int src1_step, int src
 
         short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
 
-        short4 data = convert_short4_sat(convert_int4_sat(src_data1) + src2);
+        short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
 
         *((__global short4 *)((__global char *)dst + dst_index)) = data;
     }
@@ -700,7 +463,7 @@ __kernel void arithm_s_add_C4_D4 (__global   int *src1, int src1_step, int src1_
 
         int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
 
-        int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src2));
+        int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src2)));
 
         *((__global int4 *)((__global char *)dst + dst_index)) = data;
     }
@@ -720,7 +483,7 @@ __kernel void arithm_s_add_C4_D5 (__global   float *src1, int src1_step, int src
 
         float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index));
 
-        float4 data = src_data1 + src2;
+        float4 data = ARITHM_OP(src_data1, src2);
 
         *((__global float4 *)((__global char *)dst + dst_index)) = data;
     }
@@ -742,7 +505,7 @@ __kernel void arithm_s_add_C4_D6 (__global   double *src1, int src1_step, int sr
 
         double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index));
 
-        double4 data = src_data1 + src2;
+        double4 data = ARITHM_OP(src_data1, src2);
 
         *((__global double4 *)((__global char *)dst + dst_index)) = data;
     }
diff --git a/modules/ocl/src/kernels/arithm_add_scalar_mask.cl b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl
similarity index 60%
rename from modules/ocl/src/kernels/arithm_add_scalar_mask.cl
rename to modules/ocl/src/opencl/arithm_add_scalar_mask.cl
index 4acb5be6a2..a0cb7dacb4 100644
--- a/modules/ocl/src/kernels/arithm_add_scalar_mask.cl
+++ b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl
@@ -44,9 +44,18 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
+#ifdef ARITHM_ADD
+  #define ARITHM_OP(A,B) ((A)+(B))
+#elif defined ARITHM_SUB
+  #define ARITHM_OP(A,B) ((A)-(B))
+#endif
 /**************************************add with scalar with mask**************************************/
 __kernel void arithm_s_add_with_mask_C1_D0 (__global   uchar *src1, int src1_step, int src1_offset,
                                             __global   uchar *dst,  int dst_step,  int dst_offset,
@@ -61,7 +70,10 @@ __kernel void arithm_s_add_with_mask_C1_D0 (__global   uchar *src1, int src1_ste
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
 
@@ -87,7 +99,7 @@ __kernel void arithm_s_add_with_mask_C1_D0 (__global   uchar *src1, int src1_ste
         }
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4 tmp = convert_int4_sat(src1_data) + src2_data;
+        int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data);
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
@@ -111,7 +123,10 @@ __kernel void arithm_s_add_with_mask_C1_D2 (__global   ushort *src1, int src1_st
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
 
@@ -124,7 +139,7 @@ __kernel void arithm_s_add_with_mask_C1_D2 (__global   ushort *src1, int src1_st
         uchar2  mask_data = vload2(0, mask + mask_index);
 
         ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) + src2_data;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data);
         ushort2 tmp_data = convert_ushort2_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
@@ -146,7 +161,10 @@ __kernel void arithm_s_add_with_mask_C1_D3 (__global   short *src1, int src1_ste
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
 
@@ -159,7 +177,7 @@ __kernel void arithm_s_add_with_mask_C1_D3 (__global   short *src1, int src1_ste
         uchar2  mask_data = vload2(0, mask + mask_index);
 
         short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
-        int2    tmp = convert_int2_sat(src1_data) + src2_data;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data);
         short2 tmp_data = convert_short2_sat(tmp);
 
         data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
@@ -189,7 +207,7 @@ __kernel void arithm_s_add_with_mask_C1_D4 (__global   int   *src1, int src1_ste
         int src_data2 = src2.x;
         int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
 
-        int data = convert_int_sat((long)src_data1 + (long)src_data2);
+        int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2));
         data = mask_data ? data : dst_data;
 
         *((__global int *)((__global char *)dst + dst_index)) = data;
@@ -217,7 +235,7 @@ __kernel void arithm_s_add_with_mask_C1_D5 (__global   float   *src1, int src1_s
         float src_data2 = src2.x;
         float dst_data  = *((__global float *)((__global char *)dst  + dst_index));
 
-        float data = src_data1 + src_data2;
+        float data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global float *)((__global char *)dst + dst_index)) = data;
@@ -247,7 +265,7 @@ __kernel void arithm_s_add_with_mask_C1_D6 (__global   double   *src1, int src1_
         double src_data2 = src2.x;
         double dst_data  = *((__global double *)((__global char *)dst  + dst_index));
 
-        double data = src_data1 + src_data2;
+        double data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global double *)((__global char *)dst + dst_index)) = data;
@@ -267,7 +285,10 @@ __kernel void arithm_s_add_with_mask_C2_D0 (__global   uchar *src1, int src1_ste
     {
         x = x << 1;
 
-        #define dst_align ((dst_offset >> 1) & 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
 
@@ -280,7 +301,7 @@ __kernel void arithm_s_add_with_mask_C2_D0 (__global   uchar *src1, int src1_ste
         uchar2 mask_data = vload2(0, mask + mask_index);
 
         uchar4 data = *((__global uchar4 *)(dst + dst_index));
-        int4  tmp = convert_int4_sat(src1_data) + src2_data;
+        int4  tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data);
         uchar4 tmp_data = convert_uchar4_sat(tmp);
 
         data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
@@ -310,7 +331,7 @@ __kernel void arithm_s_add_with_mask_C2_D2 (__global   ushort *src1, int src1_st
         int2 src_data2 = (int2)(src2.x, src2.y);
         ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + src_data2;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2);
         ushort2 data = convert_ushort2_sat(tmp);
         data = mask_data ? data : dst_data;
 
@@ -338,7 +359,7 @@ __kernel void arithm_s_add_with_mask_C2_D3 (__global   short *src1, int src1_ste
         int2 src_data2 = (int2)(src2.x, src2.y);
         short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
 
-        int2    tmp = convert_int2_sat(src_data1) + src_data2;
+        int2    tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2);
         short2 data = convert_short2_sat(tmp);
         data = mask_data ? data : dst_data;
 
@@ -366,7 +387,7 @@ __kernel void arithm_s_add_with_mask_C2_D4 (__global   int *src1, int src1_step,
         int2 src_data2 = (int2)(src2.x, src2.y);
         int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
 
-        int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2));
+        int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2)));
         data = mask_data ? data : dst_data;
 
         *((__global int2 *)((__global char *)dst + dst_index)) = data;
@@ -393,7 +414,7 @@ __kernel void arithm_s_add_with_mask_C2_D5 (__global   float *src1, int src1_ste
         float2 src_data2 = (float2)(src2.x, src2.y);
         float2 dst_data  = *((__global float2 *)((__global char *)dst  + dst_index));
 
-        float2 data = src_data1 + src_data2;
+        float2 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global float2 *)((__global char *)dst + dst_index)) = data;
@@ -422,7 +443,7 @@ __kernel void arithm_s_add_with_mask_C2_D6 (__global   double *src1, int src1_st
         double2 src_data2 = (double2)(src2.x, src2.y);
         double2 dst_data  = *((__global double2 *)((__global char *)dst  + dst_index));
 
-        double2 data = src_data1 + src_data2;
+        double2 data = ARITHM_OP(src_data1, src_data2);
         data = mask_data ? data : dst_data;
 
         *((__global double2 *)((__global char *)dst + dst_index)) = data;
@@ -430,308 +451,11 @@ __kernel void arithm_s_add_with_mask_C2_D6 (__global   double *src1, int src1_st
 }
 #endif
 
-__kernel void arithm_s_add_with_mask_C3_D0 (__global   uchar *src1, int src1_step, int src1_offset,
-                                            __global   uchar *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 2;
-
-        #define dst_align (((dst_offset % dst_step) / 3 ) & 3)
-        int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3));
-
-        uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0);
-        uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4);
-        uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8);
-
-        int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x);
-        int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y);
-        int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z);
-
-        uchar4 mask_data = vload4(0, mask + mask_index);
-
-        uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0));
-        uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4));
-        uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8));
-
-        uchar4 tmp_data_0 = convert_uchar4_sat(convert_int4_sat(src1_data_0) + src2_data_0);
-        uchar4 tmp_data_1 = convert_uchar4_sat(convert_int4_sat(src1_data_1) + src2_data_1);
-        uchar4 tmp_data_2 = convert_uchar4_sat(convert_int4_sat(src1_data_2) + src2_data_2);
-
-        data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz;
-        data_0.w   = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_0.w : data_0.w;
-
-        data_1.xy  = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end))
-                     ? tmp_data_1.xy : data_1.xy;
-        data_1.zw  = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.zw : data_1.zw;
-
-        data_2.x   = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.x : data_2.x;
-        data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end))
-                     ? tmp_data_2.yzw : data_2.yzw;
-
-        *((__global uchar4 *)(dst + dst_index + 0)) = data_0;
-        *((__global uchar4 *)(dst + dst_index + 4)) = data_1;
-        *((__global uchar4 *)(dst + dst_index + 8)) = data_2;
-    }
-}
-__kernel void arithm_s_add_with_mask_C3_D2 (__global   ushort *src1, int src1_step, int src1_offset,
-                                            __global   ushort *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0));
-        ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4));
-        ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0));
-        ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4));
-        ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8));
-
-        ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + src2_data_0);
-        ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + src2_data_1);
-        ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + src2_data_2);
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_add_with_mask_C3_D3 (__global   short *src1, int src1_step, int src1_offset,
-                                            __global   short *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        x = x << 1;
-
-        #define dst_align (((dst_offset % dst_step) / 6 ) & 1)
-        int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6));
-        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
-
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6));
-
-        short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0));
-        short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4));
-        short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8));
-
-        int2 src2_data_0 = (int2)(src2.x, src2.y);
-        int2 src2_data_1 = (int2)(src2.z, src2.x);
-        int2 src2_data_2 = (int2)(src2.y, src2.z);
-
-        uchar2 mask_data = vload2(0, mask + mask_index);
-
-        short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0));
-        short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4));
-        short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8));
-
-        short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + src2_data_0);
-        short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + src2_data_1);
-        short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + src2_data_2);
-
-        data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy;
-
-        data_1.x  = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end))
-                     ? tmp_data_1.x : data_1.x;
-        data_1.y  = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_1.y : data_1.y;
-
-        data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end))
-                     ? tmp_data_2.xy : data_2.xy;
-
-       *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_add_with_mask_C3_D4 (__global   int *src1, int src1_step, int src1_offset,
-                                            __global   int *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            int4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0));
-        int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4));
-        int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8));
-
-        int src2_data_0 = src2.x;
-        int src2_data_1 = src2.y;
-        int src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        int data_0 = *((__global int *)((__global char *)dst + dst_index + 0));
-        int data_1 = *((__global int *)((__global char *)dst + dst_index + 4));
-        int data_2 = *((__global int *)((__global char *)dst + dst_index + 8));
-
-        int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0);
-        int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1);
-        int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2);
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global int *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global int *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global int *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-__kernel void arithm_s_add_with_mask_C3_D5 (__global   float *src1, int src1_step, int src1_offset,
-                                            __global   float *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar *mask, int mask_step, int mask_offset,
-                                            float4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 12) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 12));
-
-        float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0));
-        float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4));
-        float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8));
-
-        float src2_data_0 = src2.x;
-        float src2_data_1 = src2.y;
-        float src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        float data_0 = *((__global float *)((__global char *)dst + dst_index + 0));
-        float data_1 = *((__global float *)((__global char *)dst + dst_index + 4));
-        float data_2 = *((__global float *)((__global char *)dst + dst_index + 8));
-
-        float tmp_data_0 = src1_data_0 + src2_data_0;
-        float tmp_data_1 = src1_data_1 + src2_data_1;
-        float tmp_data_2 = src1_data_2 + src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global float *)((__global char *)dst + dst_index + 0))= data_0;
-       *((__global float *)((__global char *)dst + dst_index + 4))= data_1;
-       *((__global float *)((__global char *)dst + dst_index + 8))= data_2;
-    }
-}
-
-#if defined (DOUBLE_SUPPORT)
-__kernel void arithm_s_add_with_mask_C3_D6 (__global   double *src1, int src1_step, int src1_offset,
-                                            __global   double *dst,  int dst_step,  int dst_offset,
-                                            __global   uchar  *mask, int mask_step, int mask_offset,
-                                            double4 src2, int rows, int cols, int dst_step1)
-{
-
-    int x = get_global_id(0);
-    int y = get_global_id(1);
-
-    if (x < cols && y < rows)
-    {
-        int src1_index = mad24(y, src1_step, (x * 24) + src1_offset);
-        int mask_index = mad24(y, mask_step, x + mask_offset);
-        int dst_index  = mad24(y, dst_step, dst_offset + (x * 24));
-
-        double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 ));
-        double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 ));
-        double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16));
-
-        double src2_data_0 = src2.x;
-        double src2_data_1 = src2.y;
-        double src2_data_2 = src2.z;
-
-        uchar mask_data = * (mask + mask_index);
-
-        double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 ));
-        double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 ));
-        double data_2 = *((__global double *)((__global char *)dst + dst_index + 16));
-
-        double tmp_data_0 = src1_data_0 + src2_data_0;
-        double tmp_data_1 = src1_data_1 + src2_data_1;
-        double tmp_data_2 = src1_data_2 + src2_data_2;
-
-        data_0 = mask_data ? tmp_data_0 : data_0;
-        data_1 = mask_data ? tmp_data_1 : data_1;
-        data_2 = mask_data ? tmp_data_2 : data_2;
-
-       *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0;
-       *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1;
-       *((__global double *)((__global char *)dst + dst_index + 16))= data_2;
-    }
-}
-#endif
-
 __kernel void arithm_s_add_with_mask_C4_D0 (__global   uchar *src1, int src1_step, int src1_offset,
                                             __global   uchar *dst,  int dst_step,  int dst_offset,
                                             __global   uchar *mask, int mask_step, int mask_offset,
                                             int4 src2, int rows, int cols, int dst_step1)
 {
-
     int x = get_global_id(0);
     int y = get_global_id(1);
 
@@ -746,7 +470,7 @@ __kernel void arithm_s_add_with_mask_C4_D0 (__global   uchar *src1, int src1_ste
         uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
 
-        uchar4 data = convert_uchar4_sat(convert_int4_sat(src_data1) + src2);
+        uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
         data = mask_data ? data : dst_data;
 
         *((__global uchar4 *)(dst + dst_index)) = data;
@@ -772,7 +496,7 @@ __kernel void arithm_s_add_with_mask_C4_D2 (__global   ushort *src1, int src1_st
         ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
         ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
 
-        ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + src2);
+        ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
         data = mask_data ? data : dst_data;
 
         *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
@@ -798,7 +522,7 @@ __kernel void arithm_s_add_with_mask_C4_D3 (__global   short *src1, int src1_ste
         short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
         short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
 
-        short4 data = convert_short4_sat(convert_int4_sat(src_data1) + src2);
+        short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2));
         data = mask_data ? data : dst_data;
 
         *((__global short4 *)((__global char *)dst + dst_index)) = data;
@@ -824,7 +548,7 @@ __kernel void arithm_s_add_with_mask_C4_D4 (__global   int *src1, int src1_step,
         int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
         int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
 
-        int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src2));
+        int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src2)));
         data = mask_data ? data : dst_data;
 
         *((__global int4 *)((__global char *)dst + dst_index)) = data;
@@ -850,7 +574,7 @@ __kernel void arithm_s_add_with_mask_C4_D5 (__global   float *src1, int src1_ste
         float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index));
         float4 dst_data  = *((__global float4 *)((__global char *)dst  + dst_index));
 
-        float4 data = src_data1 + src2;
+        float4 data = ARITHM_OP(src_data1, src2);
         data = mask_data ? data : dst_data;
 
         *((__global float4 *)((__global char *)dst + dst_index)) = data;
@@ -878,7 +602,7 @@ __kernel void arithm_s_add_with_mask_C4_D6 (__global   double *src1, int src1_st
         double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index));
         double4 dst_data  = *((__global double4 *)((__global char *)dst  + dst_index));
 
-        double4 data = src_data1 + src2;
+        double4 data = ARITHM_OP(src_data1, src2);
         data = mask_data ? data : dst_data;
 
         *((__global double4 *)((__global char *)dst + dst_index)) = data;
diff --git a/modules/ocl/src/kernels/arithm_bitwise_xor.cl b/modules/ocl/src/opencl/arithm_bitwise_binary.cl
similarity index 60%
rename from modules/ocl/src/kernels/arithm_bitwise_xor.cl
rename to modules/ocl/src/opencl/arithm_bitwise_binary.cl
index 6e83ef50ec..8bdd23c177 100644
--- a/modules/ocl/src/kernels/arithm_bitwise_xor.cl
+++ b/modules/ocl/src/opencl/arithm_bitwise_binary.cl
@@ -16,6 +16,7 @@
 //
 // @Authors
 //    Jiang Liyuan, jlyuan001.good@163.com
+//    Peng Xiao,    pengxiao@outlook.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -25,7 +26,7 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other oclMaterials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
@@ -43,17 +44,27 @@
 //
 //M*/
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
-//////////////////////////////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////////////////////////////////////////////////
-/**************************************bitwise_xor without mask**************************************/
-__kernel void arithm_bitwise_xor_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+//bitwise_binary without mask for and, or, xor operators
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////bitwise_binary///////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#ifndef OP_BINARY
+#define OP_BINARY &
+#endif
+
+__kernel void arithm_bitwise_binary_D0 (__global uchar *src1, int src1_step, int src1_offset,
+                                     __global uchar *src2, int src2_step, int src2_offset,
+                                     __global uchar *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -62,33 +73,36 @@ __kernel void arithm_bitwise_xor_D0 (__global uchar *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
 
-     if(src1_index < 0)
-     {     
-        uchar4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        uchar4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
-        uchar4 tmp_data = src1_data ^ src2_data;
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
@@ -100,10 +114,10 @@ __kernel void arithm_bitwise_xor_D0 (__global uchar *src1, int src1_step, int sr
 }
 
 
-__kernel void arithm_bitwise_xor_D1 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D1 (__global char *src1, int src1_step, int src1_offset,
+                                     __global char *src2, int src2_step, int src2_offset,
+                                     __global char *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -112,33 +126,36 @@ __kernel void arithm_bitwise_xor_D1 (__global char *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         char4 src1_data = vload4(0, src1 + src1_index_fix);
         char4 src2_data = vload4(0, src2 + src2_index_fix);
 
-     if(src1_index < 0)
-     {     
-        char4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        char4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
+        if(src1_index < 0)
+        {
+            char4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            char4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
         char4 dst_data = *((__global char4 *)(dst + dst_index));
-        char4 tmp_data = src1_data ^ src2_data;
+        char4 tmp_data = src1_data OP_BINARY src2_data;
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
@@ -150,10 +167,10 @@ __kernel void arithm_bitwise_xor_D1 (__global char *src1, int src1_step, int src
 }
 
 
-__kernel void arithm_bitwise_xor_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global ushort *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D2 (__global ushort *src1, int src1_step, int src1_offset,
+                                     __global ushort *src2, int src2_step, int src2_offset,
+                                     __global ushort *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -163,33 +180,36 @@ __kernel void arithm_bitwise_xor_D2 (__global ushort *src1, int src1_step, int s
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
 
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index_fix));
         ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index_fix));
 
-     if(src1_index < 0)
-     {     
-        ushort4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        ushort4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
         ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
-        ushort4 tmp_data = src1_data ^ src2_data;
+        ushort4 tmp_data = src1_data OP_BINARY src2_data;
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
@@ -202,10 +222,10 @@ __kernel void arithm_bitwise_xor_D2 (__global ushort *src1, int src1_step, int s
 
 
 
-__kernel void arithm_bitwise_xor_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global short *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D3 (__global short *src1, int src1_step, int src1_offset,
+                                     __global short *src2, int src2_step, int src2_offset,
+                                     __global short *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -215,37 +235,36 @@ __kernel void arithm_bitwise_xor_D3 (__global short *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
 
-     int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-     int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index_fix));
         short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index_fix));
 
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
         short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
-
-     if(src1_index < 0)
-     {     
-        short4 tmp;
-        tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-        src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-     }                 
-     if(src2_index < 0)  
-     {                         
-        short4 tmp;                   
-        tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-        src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-     }
-
-
-
-        short4 tmp_data = src1_data ^ src2_data;
+        short4 tmp_data = src1_data OP_BINARY src2_data;
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
@@ -258,10 +277,10 @@ __kernel void arithm_bitwise_xor_D3 (__global short *src1, int src1_step, int sr
 
 
 
-__kernel void arithm_bitwise_xor_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global int *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D4 (__global int *src1, int src1_step, int src1_offset,
+                                     __global int *src2, int src2_step, int src2_offset,
+                                     __global int *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -274,16 +293,16 @@ __kernel void arithm_bitwise_xor_D4 (__global int *src1, int src1_step, int src1
 
         int data1 = *((__global int *)((__global char *)src1 + src1_index));
         int data2 = *((__global int *)((__global char *)src2 + src2_index));
-        int tmp  = data1 ^ data2;
+        int tmp  = data1 OP_BINARY data2;
 
         *((__global int *)((__global char *)dst + dst_index)) = tmp;
     }
 }
 
-__kernel void arithm_bitwise_xor_D5 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D5 (__global char *src1, int src1_step, int src1_offset,
+                                     __global char *src2, int src2_step, int src2_offset,
+                                     __global char *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -296,17 +315,17 @@ __kernel void arithm_bitwise_xor_D5 (__global char *src1, int src1_step, int src
 
         char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
         char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
-        char4 tmp = data1 ^ data2;
+        char4 tmp = data1 OP_BINARY data2;
 
         *((__global char4 *)((__global char *)dst + dst_index)) = tmp;
     }
 }
 
 #if defined (DOUBLE_SUPPORT)
-__kernel void arithm_bitwise_xor_D6 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *src2, int src2_step, int src2_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+__kernel void arithm_bitwise_binary_D6 (__global char *src1, int src1_step, int src1_offset,
+                                     __global char *src2, int src2_step, int src2_offset,
+                                     __global char *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -320,8 +339,7 @@ __kernel void arithm_bitwise_xor_D6 (__global char *src1, int src1_step, int src
         char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
         char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
 
-        *((__global char8 *)((__global char *)dst + dst_index)) = data1 ^ data2;
+        *((__global char8 *)((__global char *)dst + dst_index)) = data1 OP_BINARY data2;
     }
 }
 #endif
-
diff --git a/modules/ocl/src/opencl/arithm_bitwise_binary_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_binary_mask.cl
new file mode 100644
index 0000000000..60cd188203
--- /dev/null
+++ b/modules/ocl/src/opencl/arithm_bitwise_binary_mask.cl
@@ -0,0 +1,809 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jiang Liyuan, jlyuan001.good@163.com
+//    Peng Xiao,    pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
+#endif
+
+#ifndef OP_BINARY
+#define OP_BINARY &
+#endif
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////bitwise_binary////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+/**************************************bitwise_binary with mask**************************************/
+__kernel void arithm_bitwise_binary_with_mask_C1_D0 (
+        __global uchar *src1, int src1_step, int src1_offset,
+        __global uchar *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global uchar *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = vload4(0, src2 + src2_index);
+        uchar4 mask_data = vload4(0, mask + mask_index);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D1 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = vload4(0, src2 + src2_index);
+        uchar4 mask_data = vload4(0, mask + mask_index);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D2 (
+        __global ushort *src1, int src1_step, int src1_offset,
+        __global ushort *src2, int src2_step, int src2_offset,
+        __global uchar  *mask, int mask_step, int mask_offset,
+        __global ushort *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
+        ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
+        uchar2  mask_data = vload2(0, mask + mask_index);
+
+        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
+        ushort2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
+
+        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D3 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
+        short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
+        uchar2  mask_data = vload2(0, mask + mask_index);
+
+        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
+        short2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
+
+        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D4 (
+        __global int   *src1, int src1_step, int src1_offset,
+        __global int   *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global int   *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
+        int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
+        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
+
+        int data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D5 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
+        char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
+        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
+
+        char4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C1_D6 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
+        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
+        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
+
+        char8 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char8 *)((__global char *)dst + dst_index)) = data;
+    }
+
+}
+
+
+
+__kernel void arithm_bitwise_binary_with_mask_C2_D0 (
+        __global uchar *src1, int src1_step, int src1_offset,
+        __global uchar *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global uchar *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = vload4(0, src2 + src2_index);
+        uchar2 mask_data = vload2(0, mask + mask_index);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
+        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_bitwise_binary_with_mask_C2_D1 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = vload4(0, src2 + src2_index);
+        uchar2 mask_data = vload2(0, mask + mask_index);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
+        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_bitwise_binary_with_mask_C2_D2 (
+        __global ushort *src1, int src1_step, int src1_offset,
+        __global ushort *src2, int src2_step, int src2_offset,
+        __global uchar  *mask, int mask_step, int mask_offset,
+        __global ushort *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
+        ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
+        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
+
+        ushort2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C2_D3 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
+        short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
+        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
+
+        short2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global short2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C2_D4 (
+        __global int   *src1, int src1_step, int src1_offset,
+        __global int   *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global int    *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
+        int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
+        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
+
+        int2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C2_D5 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
+        char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
+        char8 dst_data  = *((__global char8 *)((__global char *)dst  + dst_index));
+
+        char8 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char8 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_bitwise_binary_with_mask_C2_D6 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
+        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
+        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
+
+        char16 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char16 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_bitwise_binary_with_mask_C4_D0 (
+        __global uchar *src1, int src1_step, int src1_offset,
+        __global uchar *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global uchar *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
+        uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+
+        uchar4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_bitwise_binary_with_mask_C4_D1 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
+        char4 src_data2 = *((__global char4 *)(src2 + src2_index));
+        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
+
+        char4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_bitwise_binary_with_mask_C4_D2 (
+        __global ushort *src1, int src1_step, int src1_offset,
+        __global ushort *src2, int src2_step, int src2_offset,
+        __global uchar  *mask, int mask_step, int mask_offset,
+        __global ushort *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
+        ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
+        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
+
+        ushort4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C4_D3 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
+        short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
+        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
+
+        short4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global short4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C4_D4 (
+        __global int   *src1, int src1_step, int src1_offset,
+        __global int   *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global int   *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
+        int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
+        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
+
+        int4 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_bitwise_binary_with_mask_C4_D5 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
+        char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
+        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
+
+        char16 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char16 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_bitwise_binary_with_mask_C4_D6 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *src2, int src2_step, int src2_offset,
+        __global uchar  *mask, int mask_step, int mask_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
+        int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
+        char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
+        char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
+        char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
+
+        char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
+        char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
+        char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
+        char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
+
+        char8 dst_data_0  = *((__global char8 *)((__global char *)dst  + dst_index + 0));
+        char8 dst_data_1  = *((__global char8 *)((__global char *)dst  + dst_index + 8));
+        char8 dst_data_2  = *((__global char8 *)((__global char *)dst  + dst_index + 16));
+        char8 dst_data_3  = *((__global char8 *)((__global char *)dst  + dst_index + 24));
+
+        char8 data_0 = src_data1_0 OP_BINARY src_data2_0;
+        char8 data_1 = src_data1_1 OP_BINARY src_data2_1;
+        char8 data_2 = src_data1_2 OP_BINARY src_data2_2;
+        char8 data_3 = src_data1_3 OP_BINARY src_data2_3;
+
+        data_0 = mask_data ? data_0 : dst_data_0;
+        data_1 = mask_data ? data_1 : dst_data_1;
+        data_2 = mask_data ? data_2 : dst_data_2;
+        data_3 = mask_data ? data_3 : dst_data_3;
+
+        *((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
+        *((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
+        *((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
+        *((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
+    }
+}
+#endif
diff --git a/modules/ocl/src/opencl/arithm_bitwise_binary_scalar.cl b/modules/ocl/src/opencl/arithm_bitwise_binary_scalar.cl
new file mode 100644
index 0000000000..5fa25004d5
--- /dev/null
+++ b/modules/ocl/src/opencl/arithm_bitwise_binary_scalar.cl
@@ -0,0 +1,638 @@
+////////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jiang Liyuan, jlyuan001.good@163.com
+//    Peng Xiao,    pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
+#endif
+
+#ifndef OP_BINARY
+#define OP_BINARY &
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////bitwise_binary/////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+/******************************bitwise binary with scalar without mask********************************/
+__kernel void arithm_s_bitwise_binary_C1_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_C1_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_C1_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
+        ushort2 src2_data = (ushort2)(src2.x, src2.x);
+
+        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
+        ushort2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
+        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
+
+        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C1_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
+        short2 src2_data = (short2)(src2.x, src2.x);
+        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
+
+        short2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
+        data.y = (dst_index + 2 <  dst_end  ) ? tmp_data.y : data.y;
+
+        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C1_D4 (
+        __global   int *src1, int src1_step, int src1_offset,
+        __global   int *dst,  int dst_step,  int dst_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
+        int src_data2 = src2.x;
+
+        int data = src_data1 OP_BINARY src_data2;
+
+        *((__global int *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C1_D5 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
+        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
+
+        char4 data  = *((__global char4 *)((__global char *)dst  + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global char4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_C1_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
+        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
+
+        short4 tmp_data = src1_data OP_BINARY src2_data;
+
+        *((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
+    }
+}
+#endif
+__kernel void arithm_s_bitwise_binary_C2_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+
+        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
+        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_C2_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
+        data.zw = (dst_index + 2 <  dst_end  ) ? tmp_data.zw : data.zw;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_C2_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
+        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
+
+        ushort2 data = src_data1 OP_BINARY src_data2;
+
+        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C2_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
+        short2 src_data2 = (short2)(src2.x, src2.y);
+
+        short2 data = src_data1 OP_BINARY src_data2;
+
+        *((__global short2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C2_D4 (
+        __global   int *src1, int src1_step, int src1_offset,
+        __global   int *dst,  int dst_step,  int dst_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
+        int2 src_data2 = (int2)(src2.x, src2.y);
+
+        int2 data = src_data1 OP_BINARY src_data2;
+        *((__global int2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C2_D5 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
+        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
+
+        char8 tmp_data = src1_data OP_BINARY src2_data;
+
+        *((__global char8 *)((__global char *)dst + dst_index)) = tmp_data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_C2_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
+        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
+
+        short8 tmp_data = src1_data OP_BINARY src2_data;
+
+        *((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
+    }
+}
+#endif
+
+__kernel void arithm_s_bitwise_binary_C4_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
+
+        uchar4 data = src_data1 OP_BINARY src2;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_C4_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
+
+        char4 data = src_data1 OP_BINARY src2;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_C4_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
+
+        ushort4 data = src_data1 OP_BINARY src2;
+
+        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C4_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
+
+        short4 data = src_data1 OP_BINARY src2;
+
+        *((__global short4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C4_D4 (
+        __global   int *src1, int src1_step, int src1_offset,
+        __global   int *dst,  int dst_step,  int dst_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
+
+        int4 data = src_data1 OP_BINARY src2;
+
+        *((__global int4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_C4_D5 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
+        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
+                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
+
+        char16 tmp_data = src1_data OP_BINARY src2_data;
+
+        *((__global char16 *)((__global char *)dst + dst_index)) = tmp_data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_C4_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
+
+        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
+        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
+        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
+        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
+
+        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
+        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
+        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
+        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
+
+        short4 tmp_data_0 = src1_data_0 OP_BINARY src2_data_0;
+        short4 tmp_data_1 = src1_data_1 OP_BINARY src2_data_1;
+        short4 tmp_data_2 = src1_data_2 OP_BINARY src2_data_2;
+        short4 tmp_data_3 = src1_data_3 OP_BINARY src2_data_3;
+
+        *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
+        *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
+        *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
+        *((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
+
+    }
+}
+#endif
diff --git a/modules/ocl/src/opencl/arithm_bitwise_binary_scalar_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_binary_scalar_mask.cl
new file mode 100644
index 0000000000..9af6589adf
--- /dev/null
+++ b/modules/ocl/src/opencl/arithm_bitwise_binary_scalar_mask.cl
@@ -0,0 +1,747 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jiang Liyuan, jlyuan001.good@163.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other GpuMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
+#endif
+
+#ifndef OP_BINARY
+#define OP_BINARY &
+#endif
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////bitwise_binary////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+/**************************************bitwise_binary with scalar with mask**************************************/
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
+        uchar4 mask_data = vload4(0, mask + mask_index);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 2;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
+        uchar4 mask_data = vload4(0, mask + mask_index);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
+        data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
+        data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
+        data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        __global   uchar  *mask, int mask_step, int mask_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
+        ushort2 src2_data = (ushort2)(src2.x, src2.x);
+        uchar2  mask_data = vload2(0, mask + mask_index);
+
+        ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
+        ushort2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
+        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
+
+        *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
+        short2 src2_data = (short2)(src2.x, src2.x);
+        uchar2  mask_data = vload2(0, mask + mask_index);
+
+        short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
+        short2 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
+        data.y = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.y : data.y;
+
+        *((__global short2 *)((__global uchar *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D4 (
+        __global   int   *src1, int src1_step, int src1_offset,
+        __global   int   *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
+        int src_data2 = src2.x;
+        int dst_data  = *((__global int *)((__global char *)dst  + dst_index));
+
+        int data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D5 (
+        __global char *src1, int src1_step, int src1_offset,
+        __global char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
+        char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
+        char4 dst_data  = *((__global char4 *)((__global char *)dst  + dst_index));
+
+        char4 data = src1_data OP_BINARY src2_data;
+        data = mask_data ? data : dst_data;
+
+        *((__global char4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_with_mask_C1_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
+        short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
+        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
+
+        short4 data = src1_data OP_BINARY src2_data;
+        data = mask_data ? data : dst_data;
+
+        *((__global short4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#endif
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        uchar4 src1_data = vload4(0, src1 + src1_index);
+        uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
+        uchar2 mask_data = vload2(0, mask + mask_index);
+
+        uchar4 data = *((__global uchar4 *)(dst + dst_index));
+        uchar4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
+        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        x = x << 1;
+
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset / 2) & 1)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
+
+        int dst_start  = mad24(y, dst_step, dst_offset);
+        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
+        int dst_index  = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
+
+        char4 src1_data = vload4(0, src1 + src1_index);
+        char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
+        uchar2 mask_data = vload2(0, mask + mask_index);
+
+        char4 data = *((__global char4 *)(dst + dst_index));
+        char4 tmp_data = src1_data OP_BINARY src2_data;
+
+        data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
+        data.zw = ((mask_data.y) && (dst_index + 2 <  dst_end  )) ? tmp_data.zw : data.zw;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
+        ushort2 src_data2 = (ushort2)(src2.x, src2.y);
+        ushort2 dst_data  = *((__global ushort2 *)((__global char *)dst  + dst_index));
+
+        ushort2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global ushort2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
+        short2 src_data2 = (short2)(src2.x, src2.y);
+        short2 dst_data  = *((__global short2 *)((__global char *)dst  + dst_index));
+
+        short2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global short2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D4 (
+        __global   int *src1, int src1_step, int src1_offset,
+        __global   int *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
+        int2 src_data2 = (int2)(src2.x, src2.y);
+        int2 dst_data  = *((__global int2 *)((__global char *)dst  + dst_index));
+
+        int2 data = src_data1 OP_BINARY src_data2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int2 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D5 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global  char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
+        char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
+        char8 dst_data = *((__global char8 *)((__global char *)dst  + dst_index));
+
+        char8 data = src1_data OP_BINARY src2_data;
+
+        data = mask_data ? data : dst_data;
+
+        *((__global char8 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_with_mask_C2_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
+        short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
+        short8 dst_data = *((__global short8 *)((__global char *)dst  + dst_index));
+
+        short8 data = src1_data OP_BINARY src2_data;
+        data = mask_data ? data : dst_data;
+
+        *((__global short8 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#endif
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D0 (
+        __global   uchar *src1, int src1_step, int src1_offset,
+        __global   uchar *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        uchar4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+
+        uchar4 data = src_data1 OP_BINARY src2;
+        data = mask_data ? data : dst_data;
+
+        *((__global uchar4 *)(dst + dst_index)) = data;
+    }
+}
+
+
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D1 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 2) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char4 src_data1 = *((__global char4 *)(src1 + src1_index));
+        char4 dst_data  = *((__global char4 *)(dst  + dst_index));
+
+        char4 data = src_data1 OP_BINARY src2;
+        data = mask_data ? data : dst_data;
+
+        *((__global char4 *)(dst + dst_index)) = data;
+    }
+}
+
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D2 (
+        __global   ushort *src1, int src1_step, int src1_offset,
+        __global   ushort *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        ushort4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
+        ushort4 dst_data  = *((__global ushort4 *)((__global char *)dst  + dst_index));
+
+        ushort4 data = src_data1 OP_BINARY src2;
+        data = mask_data ? data : dst_data;
+
+        *((__global ushort4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D3 (
+        __global   short *src1, int src1_step, int src1_offset,
+        __global   short *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        short4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 3) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
+        short4 dst_data  = *((__global short4 *)((__global char *)dst  + dst_index));
+
+        short4 data = src_data1 OP_BINARY src2;
+        data = mask_data ? data : dst_data;
+
+        *((__global short4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D4 (
+        __global   int *src1, int src1_step, int src1_offset,
+        __global   int *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        int4 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
+        int4 dst_data  = *((__global int4 *)((__global char *)dst  + dst_index));
+
+        int4 data = src_data1 OP_BINARY src2;
+        data = mask_data ? data : dst_data;
+
+        *((__global int4 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D5 (
+        __global   char *src1, int src1_step, int src1_offset,
+        __global   char *dst,  int dst_step,  int dst_offset,
+        __global   uchar *mask, int mask_step, int mask_offset,
+        char16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 4) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
+        char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
+                                    src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
+        char16 dst_data  = *((__global char16 *)((__global char *)dst  + dst_index));
+
+        char16 data = src1_data OP_BINARY src2_data;
+        data = mask_data ? data : dst_data;
+
+        *((__global char16 *)((__global char *)dst + dst_index)) = data;
+    }
+}
+#if defined (DOUBLE_SUPPORT)
+__kernel void arithm_s_bitwise_binary_with_mask_C4_D6 (
+        __global short *src1, int src1_step, int src1_offset,
+        __global short *dst,  int dst_step,  int dst_offset,
+        __global uchar *mask, int mask_step, int mask_offset,
+        short16 src2, int rows, int cols, int dst_step1)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < cols && y < rows)
+    {
+        int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
+        int mask_index = mad24(y, mask_step,  x       + mask_offset);
+        int dst_index  = mad24(y, dst_step,  (x << 5) + dst_offset);
+
+        uchar mask_data = *(mask + mask_index);
+
+        short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
+        short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
+        short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
+        short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
+
+        short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
+        short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
+        short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
+        short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
+
+        short4 dst_data_0  = *((__global short4 *)((__global char *)dst  + dst_index + 0));
+        short4 dst_data_1  = *((__global short4 *)((__global char *)dst  + dst_index + 8));
+        short4 dst_data_2  = *((__global short4 *)((__global char *)dst  + dst_index + 16));
+        short4 dst_data_3  = *((__global short4 *)((__global char *)dst  + dst_index + 24));
+
+        short4 data_0 = src1_data_0 OP_BINARY src2_data_0;
+        short4 data_1 = src1_data_1 OP_BINARY src2_data_1;
+        short4 data_2 = src1_data_2 OP_BINARY src2_data_2;
+        short4 data_3 = src1_data_3 OP_BINARY src2_data_3;
+
+        data_0 = mask_data ? data_0 : dst_data_0;
+        data_1 = mask_data ? data_1 : dst_data_1;
+        data_2 = mask_data ? data_2 : dst_data_2;
+        data_3 = mask_data ? data_3 : dst_data_3;
+
+        *((__global short4 *)((__global char *)dst + dst_index + 0)) = data_0;
+        *((__global short4 *)((__global char *)dst + dst_index + 8)) = data_1;
+        *((__global short4 *)((__global char *)dst + dst_index + 16)) = data_2;
+        *((__global short4 *)((__global char *)dst + dst_index + 24)) = data_3;
+    }
+}
+#endif
diff --git a/modules/ocl/src/kernels/arithm_bitwise_not.cl b/modules/ocl/src/opencl/arithm_bitwise_not.cl
similarity index 86%
rename from modules/ocl/src/kernels/arithm_bitwise_not.cl
rename to modules/ocl/src/opencl/arithm_bitwise_not.cl
index 64bcc1799a..9905130013 100644
--- a/modules/ocl/src/kernels/arithm_bitwise_not.cl
+++ b/modules/ocl/src/opencl/arithm_bitwise_not.cl
@@ -43,9 +43,12 @@
 //
 //M*/
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
-
 //////////////////////////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////BITWISE_NOT////////////////////////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -61,25 +64,28 @@ __kernel void arithm_bitwise_not_D0 (__global uchar *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-    int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = ~ src1_data;
-        
-  /*  if(src1_index < 0)
-    {
-      uchar4 tmp;
-      tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-      src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-    }
-  */
+
+        /*  if(src1_index < 0)
+          {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+          }
+        */
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
         dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
         dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
@@ -91,8 +97,8 @@ __kernel void arithm_bitwise_not_D0 (__global uchar *src1, int src1_step, int sr
 
 
 __kernel void arithm_bitwise_not_D1 (__global char *src1, int src1_step, int src1_offset,
-                             __global char *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                     __global char *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -101,8 +107,11 @@ __kernel void arithm_bitwise_not_D1 (__global char *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
@@ -124,8 +133,8 @@ __kernel void arithm_bitwise_not_D1 (__global char *src1, int src1_step, int src
 
 
 __kernel void arithm_bitwise_not_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                     __global ushort *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -135,8 +144,11 @@ __kernel void arithm_bitwise_not_D2 (__global ushort *src1, int src1_step, int s
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
@@ -159,8 +171,8 @@ __kernel void arithm_bitwise_not_D2 (__global ushort *src1, int src1_step, int s
 
 
 __kernel void arithm_bitwise_not_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                     __global short *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -170,8 +182,11 @@ __kernel void arithm_bitwise_not_D3 (__global short *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
@@ -194,8 +209,8 @@ __kernel void arithm_bitwise_not_D3 (__global short *src1, int src1_step, int sr
 
 
 __kernel void arithm_bitwise_not_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                     __global int *dst,  int dst_step,  int dst_offset,
+                                     int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -245,14 +260,13 @@ __kernel void arithm_bitwise_not_D6 (__global char *src, int src_step, int src_o
     {
         int src_index = mad24(y, src_step, (x << 3) + src_offset);
         int dst_index = mad24(y, dst_step,  (x << 3) + dst_offset);
-         
+
         char8 data;
 
         data = *((__global char8 *)((__global char *)src + src_index));
         data = ~ data;
-        
+
         *((__global char8 *)((__global char *)dst + dst_index)) = data;
     }
 }
 #endif
-
diff --git a/modules/ocl/src/kernels/arithm_cartToPolar.cl b/modules/ocl/src/opencl/arithm_cartToPolar.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_cartToPolar.cl
rename to modules/ocl/src/opencl/arithm_cartToPolar.cl
diff --git a/modules/ocl/src/kernels/arithm_compare_eq.cl b/modules/ocl/src/opencl/arithm_compare_eq.cl
similarity index 60%
rename from modules/ocl/src/kernels/arithm_compare_eq.cl
rename to modules/ocl/src/opencl/arithm_compare_eq.cl
index 1db0b7dd14..16a56acef3 100644
--- a/modules/ocl/src/kernels/arithm_compare_eq.cl
+++ b/modules/ocl/src/opencl/arithm_compare_eq.cl
@@ -43,7 +43,11 @@
 //
 //M*/
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -51,9 +55,9 @@
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 
 __kernel void arithm_compare_eq_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -62,32 +66,35 @@ __kernel void arithm_compare_eq_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+
 
-  
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
 
@@ -102,9 +109,9 @@ __kernel void arithm_compare_eq_D0 (__global uchar *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -114,30 +121,33 @@ __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1)& 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1)& 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
- 		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
@@ -153,9 +163,9 @@ __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_eq_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -165,33 +175,36 @@ __kernel void arithm_compare_eq_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+
 
 
-  
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
 
@@ -207,40 +220,43 @@ __kernel void arithm_compare_eq_D3 (__global short *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_eq_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    {   
+    {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
-         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
+        int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
@@ -255,9 +271,9 @@ __kernel void arithm_compare_eq_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_eq_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -265,23 +281,27 @@ __kernel void arithm_compare_eq_D5 (__global float *src1, int src1_step, int src
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
-        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
@@ -297,9 +317,9 @@ __kernel void arithm_compare_eq_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_eq_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -307,30 +327,33 @@ __kernel void arithm_compare_eq_D6 (__global double *src1, int src1_step, int sr
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 3) & 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3) & 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
         double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data == src2_data));
@@ -347,9 +370,9 @@ __kernel void arithm_compare_eq_D6 (__global double *src1, int src1_step, int sr
 
 /***********************************Compare GT**************************/
 __kernel void arithm_compare_gt_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -358,32 +381,35 @@ __kernel void arithm_compare_gt_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+
 
- 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data > src2_data));
 
@@ -397,9 +423,9 @@ __kernel void arithm_compare_gt_D0 (__global uchar *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_gt_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -409,32 +435,35 @@ __kernel void arithm_compare_gt_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
-  
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+
 
- 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data > src2_data));
 
@@ -450,9 +479,9 @@ __kernel void arithm_compare_gt_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_gt_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -462,30 +491,33 @@ __kernel void arithm_compare_gt_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -501,9 +533,9 @@ __kernel void arithm_compare_gt_D3 (__global short *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_gt_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -511,32 +543,35 @@ __kernel void arithm_compare_gt_D4 (__global int *src1, int src1_step, int src1_
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
-         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
+        int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
- 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data > src2_data));
 
@@ -550,9 +585,9 @@ __kernel void arithm_compare_gt_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_gt_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -560,30 +595,33 @@ __kernel void arithm_compare_gt_D5 (__global float *src1, int src1_step, int src
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
         float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
-		if(src1_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data > src2_data));
@@ -599,9 +637,9 @@ __kernel void arithm_compare_gt_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_gt_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -609,30 +647,33 @@ __kernel void arithm_compare_gt_D6 (__global double *src1, int src1_step, int sr
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 3) & 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3) & 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
-        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));	
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data > src2_data));
@@ -649,9 +690,9 @@ __kernel void arithm_compare_gt_D6 (__global double *src1, int src1_step, int sr
 
 /***********************************Compare GE**************************/
 __kernel void arithm_compare_ge_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -660,31 +701,34 @@ __kernel void arithm_compare_ge_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -702,9 +746,9 @@ __kernel void arithm_compare_ge_D0 (__global uchar *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_ge_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -714,31 +758,34 @@ __kernel void arithm_compare_ge_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
- 		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
-  
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
 
@@ -757,9 +804,9 @@ __kernel void arithm_compare_ge_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_ge_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -769,31 +816,34 @@ __kernel void arithm_compare_ge_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1)& 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1)& 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -809,9 +859,9 @@ __kernel void arithm_compare_ge_D3 (__global short *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_ge_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -820,32 +870,35 @@ __kernel void arithm_compare_ge_D4 (__global int *src1, int src1_step, int src1_
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 2)& 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2)& 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
-       uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data >= src2_data));
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -858,9 +911,9 @@ __kernel void arithm_compare_ge_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_ge_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -869,31 +922,34 @@ __kernel void arithm_compare_ge_D5 (__global float *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 2)& 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2)& 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-  		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
         float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
-		if(src1_index < 0)
-		{
+        if(src1_index < 0)
+        {
 
-			float4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data >= src2_data));
@@ -909,9 +965,9 @@ __kernel void arithm_compare_ge_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_ge_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -920,29 +976,33 @@ __kernel void arithm_compare_ge_D6 (__global double *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 3)& 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3)& 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
-        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));	
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data >= src2_data));
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -954,4 +1014,3 @@ __kernel void arithm_compare_ge_D6 (__global double *src1, int src1_step, int sr
     }
 }
 #endif
-
diff --git a/modules/ocl/src/kernels/arithm_compare_ne.cl b/modules/ocl/src/opencl/arithm_compare_ne.cl
similarity index 60%
rename from modules/ocl/src/kernels/arithm_compare_ne.cl
rename to modules/ocl/src/opencl/arithm_compare_ne.cl
index 1c5063a460..fb5859d3b2 100644
--- a/modules/ocl/src/kernels/arithm_compare_ne.cl
+++ b/modules/ocl/src/opencl/arithm_compare_ne.cl
@@ -43,13 +43,17 @@
 //
 //M*/
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 /***********************************Compare NE*******************************/
 __kernel void arithm_compare_ne_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -58,30 +62,33 @@ __kernel void arithm_compare_ne_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
@@ -98,9 +105,9 @@ __kernel void arithm_compare_ne_D0 (__global uchar *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -110,30 +117,33 @@ __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1)& 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1)& 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
- 		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
@@ -150,9 +160,9 @@ __kernel void arithm_compare_ne_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_ne_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -162,30 +172,33 @@ __kernel void arithm_compare_ne_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1)& 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1)& 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
@@ -200,9 +213,9 @@ __kernel void arithm_compare_ne_D3 (__global short *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_ne_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -210,31 +223,34 @@ __kernel void arithm_compare_ne_D4 (__global int *src1, int src1_step, int src1_
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2)& 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2)& 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-	
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
@@ -249,9 +265,9 @@ __kernel void arithm_compare_ne_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_ne_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -259,30 +275,34 @@ __kernel void arithm_compare_ne_D5 (__global float *src1, int src1_step, int src
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
-        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));		if(src1_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
-       uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -296,9 +316,9 @@ __kernel void arithm_compare_ne_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_ne_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -306,30 +326,33 @@ __kernel void arithm_compare_ne_D6 (__global double *src1, int src1_step, int sr
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 3) & 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3) & 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
-        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));	
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data != src2_data));
@@ -344,12 +367,12 @@ __kernel void arithm_compare_ne_D6 (__global double *src1, int src1_step, int sr
 }
 #endif
 
-   
+
 /***********************************Compare LT*******************************/
 __kernel void arithm_compare_lt_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global  uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global  uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -358,30 +381,33 @@ __kernel void arithm_compare_lt_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data < src2_data));
@@ -398,9 +424,9 @@ __kernel void arithm_compare_lt_D0 (__global uchar *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_lt_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -410,31 +436,34 @@ __kernel void arithm_compare_lt_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
-  
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data < src2_data));
@@ -451,9 +480,9 @@ __kernel void arithm_compare_lt_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_lt_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -463,30 +492,33 @@ __kernel void arithm_compare_lt_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -502,9 +534,9 @@ __kernel void arithm_compare_lt_D3 (__global short *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_lt_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -512,35 +544,38 @@ __kernel void arithm_compare_lt_D4 (__global int *src1, int src1_step, int src1_
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
 
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
+
 
 
- 
-   
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data < src2_data));
 
@@ -554,9 +589,9 @@ __kernel void arithm_compare_lt_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_lt_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -564,32 +599,35 @@ __kernel void arithm_compare_lt_D5 (__global float *src1, int src1_step, int src
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2) & 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2) & 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
         float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
-		if(src1_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
-       uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data < src2_data));
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -603,9 +641,9 @@ __kernel void arithm_compare_lt_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_lt_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -613,32 +651,35 @@ __kernel void arithm_compare_lt_D6 (__global double *src1, int src1_step, int sr
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 3) & 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3) & 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
-        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));	
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
-       uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
+
+        uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data < src2_data));
 
         dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
@@ -653,9 +694,9 @@ __kernel void arithm_compare_lt_D6 (__global double *src1, int src1_step, int sr
 
 /***********************************Compare LE*******************************/
 __kernel void arithm_compare_le_D0 (__global uchar *src1, int src1_step, int src1_offset,
-                             __global uchar *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global uchar *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -664,30 +705,33 @@ __kernel void arithm_compare_le_D0 (__global uchar *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); 
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
+        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         uchar4 src1_data = vload4(0, src1 + src1_index_fix);
         uchar4 src2_data = vload4(0, src2 + src2_index_fix);
-		if(src1_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			uchar4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        if(src1_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            uchar4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -705,9 +749,9 @@ __kernel void arithm_compare_le_D0 (__global uchar *src1, int src1_step, int src
 
 
 __kernel void arithm_compare_le_D2 (__global ushort *src1, int src1_step, int src1_offset,
-                             __global ushort *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global ushort *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -717,30 +761,33 @@ __kernel void arithm_compare_le_D2 (__global ushort *src1, int src1_step, int sr
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
-        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
-  
+        ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -758,9 +805,9 @@ __kernel void arithm_compare_le_D2 (__global ushort *src1, int src1_step, int sr
 
 
 __kernel void arithm_compare_le_D3 (__global short *src1, int src1_step, int src1_offset,
-                             __global short *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global short *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
@@ -770,30 +817,33 @@ __kernel void arithm_compare_le_D3 (__global short *src1, int src1_step, int src
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
-        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); 
-        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
+        int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
+        int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
-        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));		
-		if(src1_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
+        if(src1_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
 
         uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
@@ -809,9 +859,9 @@ __kernel void arithm_compare_le_D3 (__global short *src1, int src1_step, int src
 }
 
 __kernel void arithm_compare_le_D4 (__global int *src1, int src1_step, int src1_offset,
-                             __global int *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global int *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -819,30 +869,33 @@ __kernel void arithm_compare_le_D4 (__global int *src1, int src1_step, int src1_
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2)& 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2)& 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
 
         int4 src1_data = vload4(0, (__global int *)((__global char *)src1 + src1_index));
         int4 src2_data = vload4(0, (__global int *)((__global char *)src2 + src2_index));
-		if(src1_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			int4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        if(src1_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            int4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data =convert_uchar4((src1_data <= src2_data));
@@ -857,9 +910,9 @@ __kernel void arithm_compare_le_D4 (__global int *src1, int src1_step, int src1_
 }
 
 __kernel void arithm_compare_le_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global float *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -867,29 +920,32 @@ __kernel void arithm_compare_le_D5 (__global float *src1, int src1_step, int src
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 2)& 3)
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 2)& 3)
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         float4 src1_data = vload4(0, (__global float *)((__global char *)src1 + src1_index_fix));
-        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));		
-		if(src1_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			float4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
+        float4 src2_data = vload4(0, (__global float *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            float4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data <= src2_data));
@@ -905,9 +961,9 @@ __kernel void arithm_compare_le_D5 (__global float *src1, int src1_step, int src
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void arithm_compare_le_D6 (__global double *src1, int src1_step, int src1_offset,
-                             __global double *src2, int src2_step, int src2_offset,
-                             __global uchar *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1)
+                                    __global double *src2, int src2_step, int src2_offset,
+                                    __global uchar *dst,  int dst_step,  int dst_offset,
+                                    int rows, int cols, int dst_step1)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -915,30 +971,33 @@ __kernel void arithm_compare_le_D6 (__global double *src1, int src1_step, int sr
     if (x < cols && y < rows)
     {
         x = x << 2;
-        #define dst_align ((dst_offset >> 3)& 3)
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3)); 
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 3)& 3)
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+        int src2_index = mad24(y, src2_step, (x << 3) + src2_offset - (dst_align << 3));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
-		int src1_index_fix = src1_index < 0 ? 0 : src1_index;
-		int src2_index_fix = src2_index < 0 ? 0 : src2_index;
+        int src1_index_fix = src1_index < 0 ? 0 : src1_index;
+        int src2_index_fix = src2_index < 0 ? 0 : src2_index;
         double4 src1_data = vload4(0, (__global double *)((__global char *)src1 + src1_index_fix));
-        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));	
-		if(src1_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
-			src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
-		}
-		if(src2_index < 0)
-		{
-			double4 tmp;
-			tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
-			src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
-		}		
- 
+        double4 src2_data = vload4(0, (__global double *)((__global char *)src2 + src2_index_fix));
+        if(src1_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
+            src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
+        }
+        if(src2_index < 0)
+        {
+            double4 tmp;
+            tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
+            src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
+        }
+
 
         uchar4 dst_data  = *((__global uchar4 *)(dst  + dst_index));
         uchar4 tmp_data = convert_uchar4((src1_data <= src2_data));
@@ -952,5 +1011,3 @@ __kernel void arithm_compare_le_D6 (__global double *src1, int src1_step, int sr
     }
 }
 #endif
-
-
diff --git a/modules/ocl/src/kernels/arithm_div.cl b/modules/ocl/src/opencl/arithm_div.cl
similarity index 88%
rename from modules/ocl/src/kernels/arithm_div.cl
rename to modules/ocl/src/opencl/arithm_div.cl
index 54fe3cdc15..1dce3853ff 100644
--- a/modules/ocl/src/kernels/arithm_div.cl
+++ b/modules/ocl/src/opencl/arithm_div.cl
@@ -44,7 +44,11 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 typedef double F ;
 typedef double4 F4;
 #define convert_F4 convert_double4
@@ -56,34 +60,24 @@ typedef float4 F4;
 #define convert_F  float
 #endif
 
-uchar round2_uchar(F v){
-
-    uchar v1 = convert_uchar_sat(round(v));
-    //uchar v2 = convert_uchar_sat(v+(v>=0 ? 0.5 : -0.5));
-
-    return v1;//(((v-v1)==0.5) && (v1%2==0)) ? v1 : v2;
+inline uchar round2_uchar(F v)
+{
+    return convert_uchar_sat(round(v));
 }
 
-ushort round2_ushort(F v){
-
-    ushort v1 = convert_ushort_sat(round(v));
-    //ushort v2 = convert_ushort_sat(v+(v>=0 ? 0.5 : -0.5));
-
-    return v1;//(((v-v1)==0.5) && (v1%2==0)) ? v1 : v2;
+inline ushort round2_ushort(F v)
+{
+    return convert_ushort_sat(round(v));
 }
-short round2_short(F v){
 
-    short v1 = convert_short_sat(round(v));
-    //short v2 = convert_short_sat(v+(v>=0 ? 0.5 : -0.5));
-
-    return v1;//(((v-v1)==0.5) && (v1%2==0)) ? v1 : v2;
+inline short round2_short(F v)
+{
+    return convert_short_sat(round(v));
 }
-int round2_int(F v){
 
-    int v1 = convert_int_sat(round(v));
-    //int v2 = convert_int_sat(v+(v>=0 ? 0.5 : -0.5));
-
-    return v1;//(((v-v1)==0.5) && (v1%2==0)) ? v1 : v2;
+inline int round2_int(F v)
+{
+    return convert_int_sat(round(v));
 }
 ///////////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////divide///////////////////////////////////////////////////
@@ -94,39 +88,41 @@ __kernel void arithm_div_D0 (__global uchar *src1, int src1_step, int src1_offse
                              __global uchar *dst,  int dst_step,  int dst_offset,
                              int rows, int cols, int dst_step1, F scalar)
 {
-    int x = get_global_id(0);
-    int y = get_global_id(1);
+    int2 coor = (int2)(get_global_id(0), get_global_id(1));
 
-    if (x < cols && y < rows)
+    if (coor.x < cols && coor.y < rows)
     {
-        x = x << 2;
+        coor.x = coor.x << 2;
 
-        #define dst_align (dst_offset & 3)
-        int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
-        int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
+        int2 src_index = (int2)(mad24(coor.y, src1_step, coor.x + src1_offset - dst_align),
+                                mad24(coor.y, src2_step, coor.x + src2_offset - dst_align));
 
-        int dst_start  = mad24(y, dst_step, dst_offset);
-        int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
-        int dst_index  = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
+        int4 dst_args  = (int4)(mad24(coor.y, dst_step, dst_offset),
+                                mad24(coor.y, dst_step, dst_offset + dst_step1),
+                                mad24(coor.y, dst_step, dst_offset + coor.x & (int)0xfffffffc),
+                                0);
 
-        uchar4 src1_data = vload4(0, src1 + src1_index);
-        uchar4 src2_data = vload4(0, src2 + src2_index);
-        uchar4 dst_data  = *((__global uchar4 *)(dst + dst_index));
+        uchar4 src1_data = vload4(0, src1 + src_index.x);
+        uchar4 src2_data = vload4(0, src2 + src_index.y);
+        uchar4 dst_data  = *((__global uchar4 *)(dst + dst_args.z));
 
         F4 tmp      = convert_F4(src1_data) * scalar;
-
         uchar4 tmp_data;
-        tmp_data.x = ((tmp.x == 0) || (src2_data.x == 0)) ? 0 : round2_uchar(tmp.x / (F)src2_data.x);
-        tmp_data.y = ((tmp.y == 0) || (src2_data.y == 0)) ? 0 : round2_uchar(tmp.y / (F)src2_data.y);
-        tmp_data.z = ((tmp.z == 0) || (src2_data.z == 0)) ? 0 : round2_uchar(tmp.z / (F)src2_data.z);
-        tmp_data.w = ((tmp.w == 0) || (src2_data.w == 0)) ? 0 : round2_uchar(tmp.w / (F)src2_data.w);
+        tmp_data.x = ((tmp.x == 0) || (src2_data.x == 0)) ? 0 : round2_uchar(tmp.x / src2_data.x);
+        tmp_data.y = ((tmp.y == 0) || (src2_data.y == 0)) ? 0 : round2_uchar(tmp.y / src2_data.y);
+        tmp_data.z = ((tmp.z == 0) || (src2_data.z == 0)) ? 0 : round2_uchar(tmp.z / src2_data.z);
+        tmp_data.w = ((tmp.w == 0) || (src2_data.w == 0)) ? 0 : round2_uchar(tmp.w / src2_data.w);
 
-        dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
-        dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
-        dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
-        dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
+        dst_data.x = ((dst_args.z + 0 >= dst_args.x) && (dst_args.z + 0 < dst_args.y)) ? tmp_data.x : dst_data.x;
+        dst_data.y = ((dst_args.z + 1 >= dst_args.x) && (dst_args.z + 1 < dst_args.y)) ? tmp_data.y : dst_data.y;
+        dst_data.z = ((dst_args.z + 2 >= dst_args.x) && (dst_args.z + 2 < dst_args.y)) ? tmp_data.z : dst_data.z;
+        dst_data.w = ((dst_args.z + 3 >= dst_args.x) && (dst_args.z + 3 < dst_args.y)) ? tmp_data.w : dst_data.w;
 
-        *((__global uchar4 *)(dst + dst_index)) = dst_data;
+        *((__global uchar4 *)(dst + dst_args.z)) = dst_data;
     }
 }
 
@@ -142,7 +138,10 @@ __kernel void arithm_div_D2 (__global ushort *src1, int src1_step, int src1_offs
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -182,7 +181,10 @@ __kernel void arithm_div_D3 (__global short *src1, int src1_step, int src1_offse
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -297,7 +299,10 @@ __kernel void arithm_s_div_D0 (__global uchar *src, int src_step, int src_offset
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src_index = mad24(y, src_step, x + src_offset - dst_align);
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -333,7 +338,10 @@ __kernel void arithm_s_div_D2 (__global ushort *src, int src_step, int src_offse
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src_index = mad24(y, src_step, (x << 1) + src_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -368,7 +376,10 @@ __kernel void arithm_s_div_D3 (__global short *src, int src_step, int src_offset
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src_index = mad24(y, src_step, (x << 1) + src_offset - (dst_align << 1));
 
         int dst_start  = mad24(y, dst_step, dst_offset);
@@ -455,5 +466,3 @@ __kernel void arithm_s_div_D6 (__global double *src, int src_step, int src_offse
     }
 }
 #endif
-
-
diff --git a/modules/ocl/src/kernels/arithm_exp.cl b/modules/ocl/src/opencl/arithm_exp.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_exp.cl
rename to modules/ocl/src/opencl/arithm_exp.cl
diff --git a/modules/ocl/src/kernels/arithm_flip.cl b/modules/ocl/src/opencl/arithm_flip.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_flip.cl
rename to modules/ocl/src/opencl/arithm_flip.cl
index f4925244a5..49242d07c7 100644
--- a/modules/ocl/src/kernels/arithm_flip.cl
+++ b/modules/ocl/src/opencl/arithm_flip.cl
@@ -44,7 +44,11 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -61,7 +65,10 @@ __kernel void arithm_flip_rows_D0 (__global uchar *src, int src_step, int src_of
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src_index_0 = mad24(y,            src_step, x + src_offset - dst_align);
         int src_index_1 = mad24(rows - y - 1, src_step, x + src_offset - dst_align);
 
@@ -116,7 +123,10 @@ __kernel void arithm_flip_rows_D1 (__global char *src, int src_step, int src_off
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src_index_0 = mad24(y,            src_step, x + src_offset - dst_align);
         int src_index_1 = mad24(rows - y - 1, src_step, x + src_offset - dst_align);
 
@@ -158,7 +168,10 @@ __kernel void arithm_flip_rows_D2 (__global ushort *src, int src_step, int src_o
     {
         x = x << 2;
 
-        #define dst_align (((dst_offset >> 1) & 3) << 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (((dst_offset >> 1) & 3) << 1)
         int src_index_0 = mad24(y,            src_step, (x << 1) + src_offset - dst_align);
         int src_index_1 = mad24(rows - y - 1, src_step, (x << 1) + src_offset - dst_align);
 
@@ -200,7 +213,10 @@ __kernel void arithm_flip_rows_D3 (__global short *src, int src_step, int src_of
     {
         x = x << 2;
 
-        #define dst_align (((dst_offset >> 1) & 3) << 1)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (((dst_offset >> 1) & 3) << 1)
         int src_index_0 = mad24(y,            src_step, (x << 1) + src_offset - dst_align);
         int src_index_1 = mad24(rows - y - 1, src_step, (x << 1) + src_offset - dst_align);
 
@@ -314,16 +330,14 @@ __kernel void arithm_flip_cols_C1_D0 (__global uchar *src, int src_step, int src
     if (x < thread_cols && y < rows)
     {
         int src_index_0 = mad24(y, src_step, (x)           + src_offset);
-        int src_index_1 = mad24(y, src_step, (cols - x -1) + src_offset);
-
-        int dst_index_0 = mad24(y, dst_step, (x)           + dst_offset);
         int dst_index_1 = mad24(y, dst_step, (cols - x -1) + dst_offset);
-
         uchar data0 = *(src + src_index_0);
-        uchar data1 = *(src + src_index_1);
-
-        *(dst + dst_index_0) = data1;
         *(dst + dst_index_1) = data0;
+
+        int src_index_1 = mad24(y, src_step, (cols - x -1) + src_offset);
+        int dst_index_0 = mad24(y, dst_step, (x)           + dst_offset);
+        uchar data1 = *(src + src_index_1);
+        *(dst + dst_index_0) = data1;
     }
 }
 __kernel void arithm_flip_cols_C1_D1 (__global char *src, int src_step, int src_offset,
diff --git a/modules/ocl/src/kernels/arithm_flip_rc.cl b/modules/ocl/src/opencl/arithm_flip_rc.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_flip_rc.cl
rename to modules/ocl/src/opencl/arithm_flip_rc.cl
diff --git a/modules/ocl/src/kernels/arithm_log.cl b/modules/ocl/src/opencl/arithm_log.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_log.cl
rename to modules/ocl/src/opencl/arithm_log.cl
diff --git a/modules/ocl/src/kernels/arithm_magnitude.cl b/modules/ocl/src/opencl/arithm_magnitude.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_magnitude.cl
rename to modules/ocl/src/opencl/arithm_magnitude.cl
diff --git a/modules/ocl/src/kernels/arithm_magnitudeSqr.cl b/modules/ocl/src/opencl/arithm_magnitudeSqr.cl
similarity index 98%
rename from modules/ocl/src/kernels/arithm_magnitudeSqr.cl
rename to modules/ocl/src/opencl/arithm_magnitudeSqr.cl
index f1d0aa5733..3fd697ff1f 100644
--- a/modules/ocl/src/kernels/arithm_magnitudeSqr.cl
+++ b/modules/ocl/src/opencl/arithm_magnitudeSqr.cl
@@ -60,17 +60,17 @@ __kernel void magnitudeSqr_C1_D5 (__global float *src1,int src1_step,int src1_of
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
 
     {
-            
+
         x = x << 2;
 
         #define dst_align ((dst_offset >> 2) & 3)
 
-        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2)); 
-        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2)); 
-       
+        int src1_index = mad24(y, src1_step, (x << 2) + src1_offset - (dst_align << 2));
+        int src2_index = mad24(y, src2_step, (x << 2) + src2_offset - (dst_align << 2));
+
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 2) -(dst_align << 2));
@@ -125,16 +125,16 @@ __kernel void magnitudeSqr_C2_D5 (__global float *src1,int src1_step,int src1_of
     int y = get_global_id(1);
 
     if (x < cols && y < rows)
-    
+
 
     {
-            
+
         x = x << 2;
 
         #define dst_align ((dst_offset >> 2) & 3)
 
-        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3)); 
-       
+        int src1_index = mad24(y, src1_step, (x << 3) + src1_offset - (dst_align << 3));
+
         int dst_start  = mad24(y, dst_step, dst_offset);
         int dst_end    = mad24(y, dst_step, dst_offset + dst_step1);
         int dst_index  = mad24(y, dst_step, dst_offset + (x << 2) -(dst_align << 2));
@@ -148,8 +148,8 @@ __kernel void magnitudeSqr_C2_D5 (__global float *src1,int src1_step,int src1_of
           src1_data.s01234567 = src1_data.s45670123;
     if(src1_index== -2)
           src1_data.s01234567 = src1_data.s23456701;
-        
-    
+
+
 
         float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index));
 
diff --git a/modules/ocl/src/kernels/arithm_minMax.cl b/modules/ocl/src/opencl/arithm_minMax.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_minMax.cl
rename to modules/ocl/src/opencl/arithm_minMax.cl
diff --git a/modules/ocl/src/kernels/arithm_minMaxLoc.cl b/modules/ocl/src/opencl/arithm_minMaxLoc.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_minMaxLoc.cl
rename to modules/ocl/src/opencl/arithm_minMaxLoc.cl
index 761cd70f8b..94cc14d258 100644
--- a/modules/ocl/src/kernels/arithm_minMaxLoc.cl
+++ b/modules/ocl/src/opencl/arithm_minMaxLoc.cl
@@ -377,4 +377,4 @@ __kernel void arithm_op_minMaxLoc_mask (int cols,int invalid_cols,int offset,int
    }
 }
 
-*/
\ No newline at end of file
+*/
diff --git a/modules/ocl/src/kernels/arithm_minMaxLoc_mask.cl b/modules/ocl/src/opencl/arithm_minMaxLoc_mask.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_minMaxLoc_mask.cl
rename to modules/ocl/src/opencl/arithm_minMaxLoc_mask.cl
index f87b928cec..0af4f7ba03 100644
--- a/modules/ocl/src/kernels/arithm_minMaxLoc_mask.cl
+++ b/modules/ocl/src/opencl/arithm_minMaxLoc_mask.cl
@@ -240,4 +240,3 @@ __kernel void arithm_op_minMaxLoc_mask (int cols,int invalid_cols,int offset,int
        dst[gid + 3 * groupnum] = CONVERT_RES_TYPE(lm_maxloc[0]);
    }
 }
-
diff --git a/modules/ocl/src/kernels/arithm_minMax_mask.cl b/modules/ocl/src/opencl/arithm_minMax_mask.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_minMax_mask.cl
rename to modules/ocl/src/opencl/arithm_minMax_mask.cl
index 4097762331..734ccab750 100644
--- a/modules/ocl/src/kernels/arithm_minMax_mask.cl
+++ b/modules/ocl/src/opencl/arithm_minMax_mask.cl
@@ -194,4 +194,3 @@ __kernel void arithm_op_minMax_mask (int cols,int invalid_cols,int offset,int el
        dst[gid + groupnum] = localmem_max[0];
    }
 }
-
diff --git a/modules/ocl/src/kernels/arithm_mul.cl b/modules/ocl/src/opencl/arithm_mul.cl
similarity index 94%
rename from modules/ocl/src/kernels/arithm_mul.cl
rename to modules/ocl/src/opencl/arithm_mul.cl
index f9f3936a46..bfbb5942e0 100644
--- a/modules/ocl/src/kernels/arithm_mul.cl
+++ b/modules/ocl/src/opencl/arithm_mul.cl
@@ -16,7 +16,6 @@
 //
 // @Authors
 //    Jia Haipeng, jiahaipeng95@gmail.com
-//    Dachuan Zhao, dachuan@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -44,11 +43,16 @@
 //
 //M*/
 
-#if defined DOUBLE_SUPPORT
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 
-int4 round_int4(float4 v){
+int4 round_int4(float4 v)
+{
     v.s0 = v.s0 + (v.s0 > 0 ? 0.5 : -0.5);
     v.s1 = v.s1 + (v.s1 > 0 ? 0.5 : -0.5);
     v.s2 = v.s2 + (v.s2 > 0 ? 0.5 : -0.5);
@@ -56,7 +60,8 @@ int4 round_int4(float4 v){
 
     return convert_int4_sat(v);
 }
-uint4 round_uint4(float4 v){
+uint4 round_uint4(float4 v)
+{
     v.s0 = v.s0 + (v.s0 > 0 ? 0.5 : -0.5);
     v.s1 = v.s1 + (v.s1 > 0 ? 0.5 : -0.5);
     v.s2 = v.s2 + (v.s2 > 0 ? 0.5 : -0.5);
@@ -64,7 +69,8 @@ uint4 round_uint4(float4 v){
 
     return convert_uint4_sat(v);
 }
-long round_int(float v){
+long round_int(float v)
+{
     v = v + (v > 0 ? 0.5 : -0.5);
 
     return convert_int_sat(v);
@@ -85,7 +91,10 @@ __kernel void arithm_mul_D0 (__global uchar *src1, int src1_step, int src1_offse
     {
         x = x << 2;
 
-        #define dst_align (dst_offset & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align (dst_offset & 3)
         int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
         int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
 
@@ -130,7 +139,10 @@ __kernel void arithm_mul_D2 (__global ushort *src1, int src1_step, int src1_offs
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -166,7 +178,10 @@ __kernel void arithm_mul_D3 (__global short *src1, int src1_step, int src1_offse
     {
         x = x << 2;
 
-        #define dst_align ((dst_offset >> 1) & 3)
+#ifdef dst_align
+#undef dst_align
+#endif
+#define dst_align ((dst_offset >> 1) & 3)
         int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
         int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
 
@@ -262,9 +277,15 @@ __kernel void arithm_mul_D6 (__global double *src1, int src1_step, int src1_offs
 }
 #endif
 
+#ifdef DOUBLE_SUPPORT
+#define SCALAR_TYPE double
+#else
+#define SCALAR_TYPE float
+#endif
+
 __kernel void arithm_muls_D5 (__global float *src1, int src1_step, int src1_offset,
-                             __global float *dst,  int dst_step,  int dst_offset,
-                             int rows, int cols, int dst_step1, float scalar)
+                              __global float *dst,  int dst_step,  int dst_offset,
+                              int rows, int cols, int dst_step1, SCALAR_TYPE scalar)
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -279,4 +300,4 @@ __kernel void arithm_muls_D5 (__global float *src1, int src1_step, int src1_offs
 
         *((__global float *)((__global char *)dst + dst_index)) = tmp;
     }
-}
\ No newline at end of file
+}
diff --git a/modules/ocl/src/kernels/arithm_nonzero.cl b/modules/ocl/src/opencl/arithm_nonzero.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_nonzero.cl
rename to modules/ocl/src/opencl/arithm_nonzero.cl
diff --git a/modules/ocl/src/kernels/arithm_phase.cl b/modules/ocl/src/opencl/arithm_phase.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_phase.cl
rename to modules/ocl/src/opencl/arithm_phase.cl
diff --git a/modules/ocl/src/kernels/arithm_polarToCart.cl b/modules/ocl/src/opencl/arithm_polarToCart.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_polarToCart.cl
rename to modules/ocl/src/opencl/arithm_polarToCart.cl
diff --git a/modules/ocl/src/kernels/arithm_pow.cl b/modules/ocl/src/opencl/arithm_pow.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_pow.cl
rename to modules/ocl/src/opencl/arithm_pow.cl
diff --git a/modules/ocl/src/kernels/arithm_sum.cl b/modules/ocl/src/opencl/arithm_sum.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_sum.cl
rename to modules/ocl/src/opencl/arithm_sum.cl
index d29a71c699..280b0a5111 100644
--- a/modules/ocl/src/kernels/arithm_sum.cl
+++ b/modules/ocl/src/opencl/arithm_sum.cl
@@ -203,4 +203,3 @@ __kernel void arithm_op_sum (int cols,int invalid_cols,int offset,int elemnum,in
        dst[gid] = localmem_sum[0];
    }
 }
-
diff --git a/modules/ocl/src/kernels/arithm_sum_3.cl b/modules/ocl/src/opencl/arithm_sum_3.cl
similarity index 99%
rename from modules/ocl/src/kernels/arithm_sum_3.cl
rename to modules/ocl/src/opencl/arithm_sum_3.cl
index 1401889a73..3f6ed08803 100644
--- a/modules/ocl/src/kernels/arithm_sum_3.cl
+++ b/modules/ocl/src/opencl/arithm_sum_3.cl
@@ -245,4 +245,3 @@ __kernel void arithm_op_sum_3 (int cols,int invalid_cols,int offset,int elemnum,
        dst[gid*3+2] = localmem_sum3[0];
    }
 }
-
diff --git a/modules/ocl/src/kernels/arithm_transpose.cl b/modules/ocl/src/opencl/arithm_transpose.cl
similarity index 100%
rename from modules/ocl/src/kernels/arithm_transpose.cl
rename to modules/ocl/src/opencl/arithm_transpose.cl
diff --git a/modules/ocl/src/opencl/bgfg_mog.cl b/modules/ocl/src/opencl/bgfg_mog.cl
new file mode 100644
index 0000000000..2e269999ac
--- /dev/null
+++ b/modules/ocl/src/opencl/bgfg_mog.cl
@@ -0,0 +1,535 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jin Ma jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if defined (CN1)
+#define T_FRAME uchar
+#define T_MEAN_VAR float
+#define CONVERT_TYPE convert_uchar_sat
+#define F_ZERO (0.0f)
+float cvt(uchar val)
+{
+    return val;
+}
+
+float sqr(float val)
+{
+    return val * val;
+}
+
+float sum(float val)
+{
+    return val;
+}
+
+float clamp1(float var, float learningRate, float diff, float minVar)
+{
+    return fmax(var + learningRate * (diff * diff - var), minVar);
+}
+#else
+#define T_FRAME uchar4
+#define T_MEAN_VAR float4
+#define CONVERT_TYPE convert_uchar4_sat
+#define F_ZERO (0.0f, 0.0f, 0.0f, 0.0f)
+float4 cvt(const uchar4 val)
+{
+    float4 result;
+    result.x = val.x;
+    result.y = val.y;
+    result.z = val.z;
+    result.w = val.w;
+
+    return result;
+}
+
+float sqr(const float4 val)
+{
+    return val.x * val.x + val.y * val.y + val.z * val.z;
+}
+
+float sum(const float4 val)
+{
+    return (val.x + val.y + val.z);
+}
+
+float4 clamp1(const float4 var, float learningRate, const float4 diff, float minVar)
+{
+    float4 result;
+    result.x = fmax(var.x + learningRate * (diff.x * diff.x - var.x), minVar);
+    result.y = fmax(var.y + learningRate * (diff.y * diff.y - var.y), minVar);
+    result.z = fmax(var.z + learningRate * (diff.z * diff.z - var.z), minVar);
+    result.w = 0.0f;
+    return result;
+}
+#endif
+
+typedef struct
+{
+    float c_Tb;
+    float c_TB;
+    float c_Tg;
+    float c_varInit;
+    float c_varMin;
+    float c_varMax;
+    float c_tau;
+    uchar c_shadowVal;
+}con_srtuct_t;
+
+void swap(__global float* ptr, int x, int y, int k, int rows, int ptr_step)
+{
+    float val = ptr[(k * rows + y) * ptr_step + x];
+    ptr[(k * rows + y) * ptr_step + x] = ptr[((k + 1) * rows + y) * ptr_step + x];
+    ptr[((k + 1) * rows + y) * ptr_step + x] = val;
+}
+
+void swap4(__global float4* ptr, int x, int y, int k, int rows, int ptr_step)
+{
+    float4 val = ptr[(k * rows + y) * ptr_step + x];
+    ptr[(k * rows + y) * ptr_step + x] = ptr[((k + 1) * rows + y) * ptr_step + x];
+    ptr[((k + 1) * rows + y) * ptr_step + x] = val;
+}
+
+__kernel void mog_withoutLearning_kernel(__global T_FRAME* frame, __global uchar* fgmask,
+    __global float* weight, __global T_MEAN_VAR* mean, __global T_MEAN_VAR* var,
+    int frame_row, int frame_col, int frame_step, int fgmask_step,
+    int weight_step, int mean_step, int var_step,
+    float varThreshold, float backgroundRatio, int fgmask_offset_x,
+    int fgmask_offset_y, int frame_offset_x, int frame_offset_y)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (x < frame_col && y < frame_row)
+    {
+        T_MEAN_VAR pix = cvt(frame[(y + frame_offset_y) * frame_step + (x + frame_offset_x)]);
+
+        int kHit = -1;
+        int kForeground = -1;
+
+        for (int k = 0; k < (NMIXTURES); ++k)
+        {
+            if (weight[(k * frame_row + y) * weight_step + x] < 1.192092896e-07f)
+                break;
+
+            T_MEAN_VAR mu = mean[(k * frame_row + y) * mean_step + x];
+            T_MEAN_VAR _var = var[(k * frame_row + y) + var_step + x];
+
+            T_MEAN_VAR diff = pix - mu;
+
+            if (sqr(diff) < varThreshold * sum(_var))
+            {
+                kHit = k;
+                break;
+            }
+        }
+
+        if (kHit >= 0)
+        {
+            float wsum = 0.0f;
+            for (int k = 0; k < (NMIXTURES); ++k)
+            {
+                wsum += weight[(k * frame_row + y) * weight_step + x];
+
+                if (wsum > backgroundRatio)
+                {
+                    kForeground = k + 1;
+                    break;
+                }
+            }
+        }
+        if(kHit < 0 || kHit >= kForeground)
+            fgmask[(y + fgmask_offset_y) * fgmask_step + (x + fgmask_offset_x)] = (uchar) (-1);
+        else
+            fgmask[(y + fgmask_offset_y) * fgmask_step + (x + fgmask_offset_x)] = (uchar) (0);
+    }
+}
+
+__kernel void mog_withLearning_kernel(__global T_FRAME* frame, __global int* fgmask,
+    __global float* weight, __global float* sortKey, __global T_MEAN_VAR* mean,
+    __global T_MEAN_VAR* var, int frame_row, int frame_col, int frame_step, int fgmask_step,
+    int weight_step, int sortKey_step, int mean_step, int var_step,
+    float varThreshold, float backgroundRatio, float learningRate, float minVar,
+    int fgmask_offset_x, int fgmask_offset_y, int frame_offset_x, int frame_offset_y)
+{
+    const float w0 = 0.05f;
+    const float sk0 = w0 / 30.0f;
+    const float var0 = 900.f;
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x >= frame_col || y >= frame_row) return;
+    float wsum = 0.0f;
+    int kHit = -1;
+    int kForeground = -1;
+    int k = 0;
+
+    T_MEAN_VAR pix = cvt(frame[(y + frame_offset_y) * frame_step + (x + frame_offset_x)]);
+
+    for (; k < (NMIXTURES); ++k)
+    {
+        float w = weight[(k * frame_row + y) * weight_step + x];
+        wsum += w;
+
+        if (w < 1.192092896e-07f)
+            break;
+
+        T_MEAN_VAR mu = mean[(k * frame_row + y) * mean_step + x];
+        T_MEAN_VAR _var = var[(k * frame_row + y) * var_step + x];
+
+        float sortKey_prev, weight_prev;
+        T_MEAN_VAR mean_prev, var_prev;
+        if (sqr(pix - mu) < varThreshold * sum(_var))
+        {
+            wsum -= w;
+            float dw = learningRate * (1.0f - w);
+
+            _var = clamp1(_var, learningRate, pix - mu, minVar);
+
+            sortKey_prev = w / sqr(sum(_var));
+            sortKey[(k * frame_row + y) * sortKey_step + x] = sortKey_prev;
+
+            weight_prev = w + dw;
+            weight[(k * frame_row + y) * weight_step + x] = weight_prev;
+
+            mean_prev = mu + learningRate * (pix - mu);
+            mean[(k * frame_row + y) * mean_step + x] = mean_prev;
+
+            var_prev = _var;
+            var[(k * frame_row + y) * var_step + x] = var_prev;
+        }
+
+        int k1 = k - 1;
+
+        if (k1 >= 0 && sqr(pix - mu) < varThreshold * sum(_var))
+        {
+            float sortKey_next = sortKey[(k1 * frame_row + y) * sortKey_step + x];
+            float weight_next = weight[(k1 * frame_row + y) * weight_step + x];
+            T_MEAN_VAR mean_next = mean[(k1 * frame_row + y) * mean_step + x];
+            T_MEAN_VAR var_next = var[(k1 * frame_row + y) * var_step + x];
+
+            for (; sortKey_next < sortKey_prev && k1 >= 0; --k1)
+            {
+                sortKey[(k1 * frame_row + y) * sortKey_step + x] = sortKey_prev;
+                sortKey[((k1 + 1) * frame_row + y) * sortKey_step + x] = sortKey_next;
+
+                weight[(k1 * frame_row + y) * weight_step + x] = weight_prev;
+                weight[((k1 + 1) * frame_row + y) * weight_step + x] = weight_next;
+
+                mean[(k1 * frame_row + y) * mean_step + x] = mean_prev;
+                mean[((k1 + 1) * frame_row + y) * mean_step + x] = mean_next;
+
+                var[(k1 * frame_row + y) * var_step + x] = var_prev;
+                var[((k1 + 1) * frame_row + y) * var_step + x] = var_next;
+
+                sortKey_prev = sortKey_next;
+                sortKey_next = k1 > 0 ? sortKey[((k1 - 1) * frame_row + y) * sortKey_step + x] : 0.0f;
+
+                weight_prev = weight_next;
+                weight_next = k1 > 0 ? weight[((k1 - 1) * frame_row + y) * weight_step + x] : 0.0f;
+
+                mean_prev = mean_next;
+                mean_next = k1 > 0 ? mean[((k1 - 1) * frame_row + y) * mean_step + x] : (T_MEAN_VAR)F_ZERO;
+
+                var_prev = var_next;
+                var_next = k1 > 0 ? var[((k1 - 1) * frame_row + y) * var_step + x] : (T_MEAN_VAR)F_ZERO;
+            }
+        }
+
+        kHit = k1 + 1;
+        break;
+    }
+
+    if (kHit < 0)
+    {
+        kHit = k = k < ((NMIXTURES) - 1) ? k : ((NMIXTURES) - 1);
+        wsum += w0 - weight[(k * frame_row + y) * weight_step + x];
+
+        weight[(k * frame_row + y) * weight_step + x] = w0;
+        mean[(k * frame_row + y) * mean_step + x] = pix;
+#if defined (CN1)
+        var[(k * frame_row + y) * var_step + x] = (T_MEAN_VAR)(var0);
+#else
+        var[(k * frame_row + y) * var_step + x] = (T_MEAN_VAR)(var0, var0, var0, var0);
+#endif
+        sortKey[(k * frame_row + y) * sortKey_step + x] = sk0;
+    }
+    else
+    {
+        for( ; k < (NMIXTURES); k++)
+            wsum += weight[(k * frame_row + y) * weight_step + x];
+    }
+
+    float wscale = 1.0f / wsum;
+    wsum = 0;
+    for (k = 0; k < (NMIXTURES); ++k)
+    {
+        float w = weight[(k * frame_row + y) * weight_step + x];
+        w *= wscale;
+        wsum += w;
+
+        weight[(k * frame_row + y) * weight_step + x] = w;
+        sortKey[(k * frame_row + y) * sortKey_step + x] *= wscale;
+
+        kForeground = select(kForeground, k + 1, wsum > backgroundRatio && kForeground < 0);
+    }
+    fgmask[(y + fgmask_offset_y) * fgmask_step + (x + fgmask_offset_x)] = (uchar)(-(kHit >= kForeground));
+}
+
+
+__kernel void getBackgroundImage_kernel(__global float* weight, __global T_MEAN_VAR* mean, __global T_FRAME* dst,
+    int dst_row, int dst_col, int weight_step, int mean_step, int dst_step,
+    float backgroundRatio)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < dst_col && y < dst_row)
+    {
+        T_MEAN_VAR meanVal = (T_MEAN_VAR)F_ZERO;
+        float totalWeight = 0.0f;
+
+        for (int mode = 0; mode < (NMIXTURES); ++mode)
+        {
+            float _weight = weight[(mode * dst_row + y) * weight_step + x];
+
+            T_MEAN_VAR _mean = mean[(mode * dst_row + y) * mean_step + x];
+            meanVal = meanVal + _weight * _mean;
+
+            totalWeight += _weight;
+
+            if(totalWeight > backgroundRatio)
+                break;
+        }
+        meanVal = meanVal * (1.f / totalWeight);
+        dst[y * dst_step + x] = CONVERT_TYPE(meanVal);
+    }
+}
+
+__kernel void mog2_kernel(__global T_FRAME * frame, __global int* fgmask, __global float* weight, __global T_MEAN_VAR * mean,
+        __global int* modesUsed, __global float* variance, int frame_row, int frame_col, int frame_step,
+        int fgmask_step, int weight_step, int mean_step, int modesUsed_step, int var_step, float alphaT, float alpha1, float prune,
+        int detectShadows_flag, int fgmask_offset_x, int fgmask_offset_y, int frame_offset_x, int frame_offset_y, __constant con_srtuct_t* constants)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < frame_col && y < frame_row)
+    {
+        T_MEAN_VAR pix = cvt(frame[(y + frame_offset_y) * frame_step + x + frame_offset_x]);
+
+        bool background = false; // true - the pixel classified as background
+
+        bool fitsPDF = false; //if it remains zero a new GMM mode will be added
+
+        int nmodes = modesUsed[y * modesUsed_step + x];
+        int nNewModes = nmodes; //current number of modes in GMM
+
+        float totalWeight = 0.0f;
+
+        for (int mode = 0; mode < nmodes; ++mode)
+        {
+            float _weight = alpha1 * weight[(mode * frame_row + y) * weight_step + x] + prune;
+
+            if (!fitsPDF)
+            {
+                float var = variance[(mode * frame_row + y) * var_step + x];
+
+                T_MEAN_VAR _mean = mean[(mode * frame_row + y) * mean_step + x];
+
+                T_MEAN_VAR diff = _mean - pix;
+                float dist2 = sqr(diff);
+
+                if (totalWeight < constants -> c_TB && dist2 < constants -> c_Tb * var)
+                    background = true;
+
+                if (dist2 < constants -> c_Tg * var)
+                {
+                    fitsPDF = true;
+                    _weight += alphaT;
+                    float k = alphaT / _weight;
+                    mean[(mode * frame_row + y) * mean_step + x] = _mean - k * diff;
+                    float varnew = var + k * (dist2 - var);
+                    varnew = fmax(varnew, constants -> c_varMin);
+                    varnew = fmin(varnew, constants -> c_varMax);
+
+                    variance[(mode * frame_row + y) * var_step + x] = varnew;
+                    for (int i = mode; i > 0; --i)
+                    {
+                        if (_weight < weight[((i - 1) * frame_row + y) * weight_step + x])
+                            break;
+                        swap(weight, x, y, i - 1, frame_row, weight_step);
+                        swap(variance, x, y, i - 1, frame_row, var_step);
+                        #if defined (CN1)
+                        swap(mean, x, y, i - 1, frame_row, mean_step);
+                        #else
+                        swap4(mean, x, y, i - 1, frame_row, mean_step);
+                        #endif
+                    }
+                }
+            } // !fitsPDF
+
+            if (_weight < -prune)
+            {
+                _weight = 0.0;
+                nmodes--;
+            }
+
+            weight[(mode * frame_row + y) * weight_step + x] = _weight; //update weight by the calculated value
+            totalWeight += _weight;
+        }
+
+        totalWeight = 1.f / totalWeight;
+        for (int mode = 0; mode < nmodes; ++mode)
+            weight[(mode * frame_row + y) * weight_step + x] *= totalWeight;
+
+        nmodes = nNewModes;
+
+        if (!fitsPDF)
+        {
+            int mode = nmodes == (NMIXTURES) ? (NMIXTURES) - 1 : nmodes++;
+
+            if (nmodes == 1)
+                weight[(mode * frame_row + y) * weight_step + x] = 1.f;
+            else
+            {
+                weight[(mode * frame_row + y) * weight_step + x] = alphaT;
+
+                for (int i = 0; i < nmodes - 1; ++i)
+                    weight[(i * frame_row + y) * weight_step + x] *= alpha1;
+            }
+
+            mean[(mode * frame_row + y) * mean_step + x] = pix;
+            variance[(mode * frame_row + y) * var_step + x] = constants -> c_varInit;
+
+            for (int i = nmodes - 1; i > 0; --i)
+            {
+                // check one up
+                if (alphaT < weight[((i - 1) * frame_row + y) * weight_step + x])
+                    break;
+
+                swap(weight, x, y, i - 1, frame_row, weight_step);
+                swap(variance, x, y, i - 1, frame_row, var_step);
+                #if defined (CN1)
+                swap(mean, x, y, i - 1, frame_row, mean_step);
+                #else
+                swap4(mean, x, y, i - 1, frame_row, mean_step);
+                #endif
+            }
+        }
+
+        modesUsed[y * modesUsed_step + x] = nmodes;
+
+        bool isShadow = false;
+        if (detectShadows_flag && !background)
+        {
+            float tWeight = 0.0f;
+
+            for (int mode = 0; mode < nmodes; ++mode)
+            {
+                T_MEAN_VAR _mean = mean[(mode * frame_row + y) * mean_step + x];
+
+                T_MEAN_VAR pix_mean = pix * _mean;
+
+                float numerator = sum(pix_mean);
+                float denominator = sqr(_mean);
+
+                if (denominator == 0)
+                    break;
+
+                if (numerator <= denominator && numerator >= constants -> c_tau * denominator)
+                {
+                    float a = numerator / denominator;
+
+                    T_MEAN_VAR dD = a * _mean - pix;
+
+                    if (sqr(dD) < constants -> c_Tb * variance[(mode * frame_row + y) * var_step + x] * a * a)
+                    {
+                        isShadow = true;
+                        break;
+                    }
+                }
+
+                tWeight += weight[(mode * frame_row + y) * weight_step + x];
+                if (tWeight > constants -> c_TB)
+                    break;
+            }
+        }
+
+        fgmask[(y + fgmask_offset_y) * fgmask_step + x + fgmask_offset_x] = background ? 0 : isShadow ? constants -> c_shadowVal : 255;
+    }
+}
+
+__kernel void getBackgroundImage2_kernel(__global int* modesUsed, __global float* weight, __global T_MEAN_VAR* mean,
+    __global T_FRAME* dst, float c_TB, int modesUsed_row, int modesUsed_col, int modesUsed_step, int weight_step,
+    int mean_step, int dst_step, int dst_x, int dst_y)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < modesUsed_col && y < modesUsed_row)
+    {
+        int nmodes = modesUsed[y * modesUsed_step + x];
+
+        T_MEAN_VAR meanVal = (T_MEAN_VAR)F_ZERO;
+
+        float totalWeight = 0.0f;
+
+        for (int mode = 0; mode < nmodes; ++mode)
+        {
+            float _weight = weight[(mode * modesUsed_row + y) * weight_step + x];
+
+            T_MEAN_VAR _mean = mean[(mode * modesUsed_row + y) * mean_step + x];
+            meanVal = meanVal + _weight * _mean;
+
+            totalWeight += _weight;
+
+            if(totalWeight > c_TB)
+                break;
+        }
+
+        meanVal = meanVal * (1.f / totalWeight);
+        dst[(y + dst_y) * dst_step + x + dst_x] = CONVERT_TYPE(meanVal);
+    }
+}
diff --git a/modules/ocl/src/kernels/blend_linear.cl b/modules/ocl/src/opencl/blend_linear.cl
similarity index 99%
rename from modules/ocl/src/kernels/blend_linear.cl
rename to modules/ocl/src/opencl/blend_linear.cl
index acea75bd2d..50c5c39c5f 100644
--- a/modules/ocl/src/kernels/blend_linear.cl
+++ b/modules/ocl/src/opencl/blend_linear.cl
@@ -138,4 +138,3 @@ __kernel void BlendLinear_C4_D5(
         dst[pos] = (img1[pos] * w1 + img2[pos] * w2) / (w1 + w2 + 1e-5f);
     }
 }
-
diff --git a/modules/ocl/src/opencl/brute_force_match.cl b/modules/ocl/src/opencl/brute_force_match.cl
new file mode 100644
index 0000000000..a05c98ee03
--- /dev/null
+++ b/modules/ocl/src/opencl/brute_force_match.cl
@@ -0,0 +1,769 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Nathan, liujun@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics:enable
+#define MAX_FLOAT 3.40282e+038f
+
+#ifndef T
+#define T float
+#endif
+
+#ifndef BLOCK_SIZE
+#define BLOCK_SIZE 16
+#endif
+#ifndef MAX_DESC_LEN
+#define MAX_DESC_LEN 64
+#endif
+
+#ifndef DIST_TYPE
+#define DIST_TYPE 0
+#endif
+
+//http://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
+int bit1Count(int v)
+{
+    v = v - ((v >> 1) & 0x55555555);                    // reuse input as temporary
+    v = (v & 0x33333333) + ((v >> 2) & 0x33333333);     // temp
+    return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24; // count
+}
+
+// dirty fix for non-template support
+#if   (DIST_TYPE == 0) // L1Dist
+#   ifdef T_FLOAT
+#       define DIST(x, y) fabs((x) - (y))
+        typedef float value_type;
+        typedef float result_type;
+#   else
+#       define DIST(x, y) abs((x) - (y))
+        typedef int value_type;
+        typedef int result_type;
+#   endif
+#define DIST_RES(x) (x)
+#elif (DIST_TYPE == 1) // L2Dist
+#define DIST(x, y) (((x) - (y)) * ((x) - (y)))
+typedef float value_type;
+typedef float result_type;
+#define DIST_RES(x) sqrt(x)
+#elif (DIST_TYPE == 2) // Hamming
+#define DIST(x, y) bit1Count( (x) ^ (y) )
+typedef int value_type;
+typedef int result_type;
+#define DIST_RES(x) (x)
+#endif
+
+result_type reduce_block(
+    __local value_type *s_query,
+    __local value_type *s_train,
+    int lidx,
+    int lidy
+    )
+{
+    result_type result = 0;
+    #pragma unroll
+    for (int j = 0 ; j < BLOCK_SIZE ; j++)
+    {
+        result += DIST(
+            s_query[lidy * BLOCK_SIZE + j],
+            s_train[j * BLOCK_SIZE + lidx]);
+    }
+    return DIST_RES(result);
+}
+
+result_type reduce_multi_block(
+    __local value_type *s_query,
+    __local value_type *s_train,
+    int block_index,
+    int lidx,
+    int lidy
+    )
+{
+    result_type result = 0;
+    #pragma unroll
+    for (int j = 0 ; j < BLOCK_SIZE ; j++)
+    {
+        result += DIST(
+            s_query[lidy * MAX_DESC_LEN + block_index * BLOCK_SIZE + j],
+            s_train[j * BLOCK_SIZE + lidx]);
+    }
+    return DIST_RES(result);
+}
+
+/* 2dim launch, global size: dim0 is (query rows + BLOCK_SIZE - 1) / BLOCK_SIZE * BLOCK_SIZE, dim1 is BLOCK_SIZE
+local size: dim0 is BLOCK_SIZE, dim1 is BLOCK_SIZE.
+*/
+__kernel void BruteForceMatch_UnrollMatch(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global int *bestTrainIdx,
+    __global float *bestDistance,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * MAX_DESC_LEN;
+
+    int queryIdx = groupidx * BLOCK_SIZE + lidy;
+    // load the query into local memory.
+    #pragma unroll
+    for (int i = 0 ;  i <  MAX_DESC_LEN / BLOCK_SIZE; i ++)
+    {
+        int loadx = lidx + i * BLOCK_SIZE;
+        s_query[lidy * MAX_DESC_LEN + loadx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+    }
+
+    float myBestDistance = MAX_FLOAT;
+    int myBestTrainIdx = -1;
+
+    // loopUnrolledCached to find the best trainIdx and best distance.
+    volatile int imgIdx = 0;
+    for (int t = 0, endt = (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt; t++)
+    {
+        result_type result = 0;
+        #pragma unroll
+        for (int i = 0 ; i < MAX_DESC_LEN / BLOCK_SIZE ; i++)
+        {
+            //load a BLOCK_SIZE * BLOCK_SIZE block into local train.
+            const int loadx = lidx + i * BLOCK_SIZE;
+            s_train[lidx * BLOCK_SIZE + lidy] = loadx < train_cols ? train[min(t * BLOCK_SIZE + lidy, train_rows - 1) * (step / sizeof(float)) + loadx] : 0;
+
+            //synchronize to make sure each elem for reduceIteration in share memory is written already.
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            result += reduce_multi_block(s_query, s_train, i, lidx, lidy);
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+        }
+
+        int trainIdx = t * BLOCK_SIZE + lidx;
+
+        if (queryIdx < query_rows && trainIdx < train_rows && result < myBestDistance/* && mask(queryIdx, trainIdx)*/)
+        {
+            //bestImgIdx = imgIdx;
+            myBestDistance = result;
+            myBestTrainIdx = trainIdx;
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    __local float *s_distance = (__local float*)(sharebuffer);
+    __local int* s_trainIdx = (__local int *)(sharebuffer + BLOCK_SIZE * BLOCK_SIZE);
+
+    //find BestMatch
+    s_distance += lidy * BLOCK_SIZE;
+    s_trainIdx += lidy * BLOCK_SIZE;
+    s_distance[lidx] = myBestDistance;
+    s_trainIdx[lidx] = myBestTrainIdx;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    //reduce -- now all reduce implement in each threads.
+    #pragma unroll
+    for (int k = 0 ; k < BLOCK_SIZE; k++)
+    {
+        if (myBestDistance > s_distance[k])
+        {
+            myBestDistance = s_distance[k];
+            myBestTrainIdx = s_trainIdx[k];
+        }
+    }
+
+    if (queryIdx < query_rows && lidx == 0)
+    {
+        bestTrainIdx[queryIdx] = myBestTrainIdx;
+        bestDistance[queryIdx] = myBestDistance;
+    }
+}
+
+__kernel void BruteForceMatch_Match(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global int *bestTrainIdx,
+    __global float *bestDistance,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+
+    const int queryIdx = groupidx * BLOCK_SIZE + lidy;
+
+    float myBestDistance = MAX_FLOAT;
+    int myBestTrainIdx = -1;
+
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * BLOCK_SIZE;
+
+    // loop
+    for (int t = 0 ;  t < (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE ; t++)
+    {
+        result_type result = 0;
+        for (int i = 0 ; i < (query_cols + BLOCK_SIZE - 1) / BLOCK_SIZE ; i++)
+        {
+            const int loadx = lidx + i * BLOCK_SIZE;
+            //load query and train into local memory
+            s_query[lidy * BLOCK_SIZE + lidx] = 0;
+            s_train[lidx * BLOCK_SIZE + lidy] = 0;
+
+            if (loadx < query_cols)
+            {
+                s_query[lidy * BLOCK_SIZE + lidx] = query[min(queryIdx, query_rows - 1) * (step / sizeof(float)) + loadx];
+                s_train[lidx * BLOCK_SIZE + lidy] = train[min(t * BLOCK_SIZE + lidy, train_rows - 1) * (step / sizeof(float)) + loadx];
+            }
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            result += reduce_block(s_query, s_train, lidx, lidy);
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+        }
+
+        const int trainIdx = t * BLOCK_SIZE + lidx;
+
+        if (queryIdx < query_rows && trainIdx < train_rows && result < myBestDistance /*&& mask(queryIdx, trainIdx)*/)
+        {
+            //myBestImgidx = imgIdx;
+            myBestDistance = result;
+            myBestTrainIdx = trainIdx;
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local float *s_distance = (__local float *)sharebuffer;
+    __local int *s_trainIdx = (__local int *)(sharebuffer + BLOCK_SIZE * BLOCK_SIZE);
+
+    //findBestMatch
+    s_distance += lidy * BLOCK_SIZE;
+    s_trainIdx += lidy * BLOCK_SIZE;
+    s_distance[lidx] = myBestDistance;
+    s_trainIdx[lidx] = myBestTrainIdx;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    //reduce -- now all reduce implement in each threads.
+    for (int k = 0 ; k < BLOCK_SIZE; k++)
+    {
+        if (myBestDistance > s_distance[k])
+        {
+            myBestDistance = s_distance[k];
+            myBestTrainIdx = s_trainIdx[k];
+        }
+    }
+
+    if (queryIdx < query_rows && lidx == 0)
+    {
+        bestTrainIdx[queryIdx] = myBestTrainIdx;
+        bestDistance[queryIdx] = myBestDistance;
+    }
+}
+
+//radius_unrollmatch
+__kernel void BruteForceMatch_RadiusUnrollMatch(
+    __global T *query,
+    __global T *train,
+    float maxDistance,
+    //__global float *mask,
+    __global int *bestTrainIdx,
+    __global float *bestDistance,
+    __global int *nMatches,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int bestTrainIdx_cols,
+    int step,
+    int ostep
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+    const int groupidy = get_group_id(1);
+
+    const int queryIdx = groupidy * BLOCK_SIZE + lidy;
+    const int trainIdx = groupidx * BLOCK_SIZE + lidx;
+
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * BLOCK_SIZE;
+
+    result_type result = 0;
+    for (int i = 0 ; i < MAX_DESC_LEN / BLOCK_SIZE ; ++i)
+    {
+        //load a BLOCK_SIZE * BLOCK_SIZE block into local train.
+        const int loadx = lidx + i * BLOCK_SIZE;
+
+        s_query[lidy * BLOCK_SIZE + lidx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+        s_train[lidx * BLOCK_SIZE + lidy] = loadx < query_cols ? train[min(groupidx * BLOCK_SIZE + lidy, train_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+
+        //synchronize to make sure each elem for reduceIteration in share memory is written already.
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        result += reduce_block(s_query, s_train, lidx, lidy);
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+
+    if (queryIdx < query_rows && trainIdx < train_rows &&
+        convert_float(result) < maxDistance/* && mask(queryIdx, trainIdx)*/)
+    {
+        unsigned int ind = atom_inc(nMatches + queryIdx/*, (unsigned int) -1*/);
+
+        if(ind < bestTrainIdx_cols)
+        {
+            //bestImgIdx = imgIdx;
+            bestTrainIdx[queryIdx * (ostep / sizeof(int)) + ind] = trainIdx;
+            bestDistance[queryIdx * (ostep / sizeof(float)) + ind] = result;
+        }
+    }
+}
+
+//radius_match
+__kernel void BruteForceMatch_RadiusMatch(
+    __global T *query,
+    __global T *train,
+    float maxDistance,
+    //__global float *mask,
+    __global int *bestTrainIdx,
+    __global float *bestDistance,
+    __global int *nMatches,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int bestTrainIdx_cols,
+    int step,
+    int ostep
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+    const int groupidy = get_group_id(1);
+
+    const int queryIdx = groupidy * BLOCK_SIZE + lidy;
+    const int trainIdx = groupidx * BLOCK_SIZE + lidx;
+
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * BLOCK_SIZE;
+
+    result_type result = 0;
+    for (int i = 0 ; i < (query_cols + BLOCK_SIZE - 1) / BLOCK_SIZE ; ++i)
+    {
+        //load a BLOCK_SIZE * BLOCK_SIZE block into local train.
+        const int loadx = lidx + i * BLOCK_SIZE;
+
+        s_query[lidy * BLOCK_SIZE + lidx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+        s_train[lidx * BLOCK_SIZE + lidy] = loadx < query_cols ? train[min(groupidx * BLOCK_SIZE + lidy, train_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+
+        //synchronize to make sure each elem for reduceIteration in share memory is written already.
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        result += reduce_block(s_query, s_train, lidx, lidy);
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+
+    if (queryIdx < query_rows && trainIdx < train_rows &&
+        convert_float(result) < maxDistance/* && mask(queryIdx, trainIdx)*/)
+    {
+        unsigned int ind = atom_inc(nMatches + queryIdx);
+
+        if(ind < bestTrainIdx_cols)
+        {
+            //bestImgIdx = imgIdx;
+            bestTrainIdx[queryIdx * (ostep / sizeof(int)) + ind] = trainIdx;
+            bestDistance[queryIdx * (ostep / sizeof(float)) + ind] = result;
+        }
+    }
+}
+
+
+__kernel void BruteForceMatch_knnUnrollMatch(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global int2 *bestTrainIdx,
+    __global float2 *bestDistance,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+
+    const int queryIdx = groupidx * BLOCK_SIZE + lidy;
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * MAX_DESC_LEN;
+
+    // load the query into local memory.
+    for (int i = 0 ;  i <  MAX_DESC_LEN / BLOCK_SIZE; i ++)
+    {
+        int loadx = lidx + i * BLOCK_SIZE;
+        s_query[lidy * MAX_DESC_LEN + loadx] = loadx < query_cols ? query[min(queryIdx, query_rows - 1)  * (step / sizeof(float)) + loadx] : 0;
+    }
+
+    float myBestDistance1 = MAX_FLOAT;
+    float myBestDistance2 = MAX_FLOAT;
+    int myBestTrainIdx1 = -1;
+    int myBestTrainIdx2 = -1;
+
+    //loopUnrolledCached
+    volatile int imgIdx = 0;
+    for (int t = 0 ; t < (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE ; t++)
+    {
+        result_type result = 0;
+        for (int i = 0 ; i < MAX_DESC_LEN / BLOCK_SIZE ; i++)
+        {
+            //load a BLOCK_SIZE * BLOCK_SIZE block into local train.
+            const int loadx = lidx + i * BLOCK_SIZE;
+            s_train[lidx * BLOCK_SIZE + lidy] = loadx < train_cols ? train[min(t * BLOCK_SIZE + lidy, train_rows - 1) * (step / sizeof(float)) + loadx] : 0;
+
+            //synchronize to make sure each elem for reduceIteration in share memory is written already.
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            result += reduce_multi_block(s_query, s_train, i, lidx, lidy);
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+        }
+
+        const int trainIdx = t * BLOCK_SIZE + lidx;
+
+        if (queryIdx < query_rows && trainIdx < train_rows)
+        {
+            if (result < myBestDistance1)
+            {
+                myBestDistance2 = myBestDistance1;
+                myBestTrainIdx2 = myBestTrainIdx1;
+                myBestDistance1 = result;
+                myBestTrainIdx1 = trainIdx;
+            }
+            else if (result < myBestDistance2)
+            {
+                myBestDistance2 = result;
+                myBestTrainIdx2 = trainIdx;
+            }
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local float *s_distance = (local float *)sharebuffer;
+    __local int *s_trainIdx = (local int *)(sharebuffer + BLOCK_SIZE * BLOCK_SIZE);
+
+    // find BestMatch
+    s_distance += lidy * BLOCK_SIZE;
+    s_trainIdx += lidy * BLOCK_SIZE;
+
+    s_distance[lidx] = myBestDistance1;
+    s_trainIdx[lidx] = myBestTrainIdx1;
+
+    float bestDistance1 = MAX_FLOAT;
+    float bestDistance2 = MAX_FLOAT;
+    int bestTrainIdx1 = -1;
+    int bestTrainIdx2 = -1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (lidx == 0)
+    {
+        for (int i = 0 ; i < BLOCK_SIZE ; i++)
+        {
+            float val = s_distance[i];
+            if (val < bestDistance1)
+            {
+                bestDistance2 = bestDistance1;
+                bestTrainIdx2 = bestTrainIdx1;
+
+                bestDistance1 = val;
+                bestTrainIdx1 = s_trainIdx[i];
+            }
+            else if (val < bestDistance2)
+            {
+                bestDistance2 = val;
+                bestTrainIdx2 = s_trainIdx[i];
+            }
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    s_distance[lidx] = myBestDistance2;
+    s_trainIdx[lidx] = myBestTrainIdx2;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (lidx == 0)
+    {
+        for (int i = 0 ; i < BLOCK_SIZE ; i++)
+        {
+            float val = s_distance[i];
+
+            if (val < bestDistance2)
+            {
+                bestDistance2 = val;
+                bestTrainIdx2 = s_trainIdx[i];
+            }
+        }
+    }
+
+    myBestDistance1 = bestDistance1;
+    myBestDistance2 = bestDistance2;
+
+    myBestTrainIdx1 = bestTrainIdx1;
+    myBestTrainIdx2 = bestTrainIdx2;
+
+    if (queryIdx < query_rows && lidx == 0)
+    {
+        bestTrainIdx[queryIdx] = (int2)(myBestTrainIdx1, myBestTrainIdx2);
+        bestDistance[queryIdx] = (float2)(myBestDistance1, myBestDistance2);
+    }
+}
+
+__kernel void BruteForceMatch_knnMatch(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global int2 *bestTrainIdx,
+    __global float2 *bestDistance,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step
+)
+{
+    const int lidx = get_local_id(0);
+    const int lidy = get_local_id(1);
+    const int groupidx = get_group_id(0);
+
+    const int queryIdx = groupidx * BLOCK_SIZE + lidy;
+    __local value_type *s_query = (__local value_type *)sharebuffer;
+    __local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * BLOCK_SIZE;
+
+    float myBestDistance1 = MAX_FLOAT;
+    float myBestDistance2 = MAX_FLOAT;
+    int myBestTrainIdx1 = -1;
+    int myBestTrainIdx2 = -1;
+
+    //loop
+    for (int  t = 0 ; t < (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE ; t++)
+    {
+        result_type result = 0.0f;
+        for (int i = 0 ; i < (query_cols + BLOCK_SIZE -1) / BLOCK_SIZE ; i++)
+        {
+            const int loadx = lidx + i * BLOCK_SIZE;
+            //load query and train into local memory
+            s_query[lidy * BLOCK_SIZE + lidx] = 0;
+            s_train[lidx * BLOCK_SIZE + lidy] = 0;
+
+            if (loadx < query_cols)
+            {
+                s_query[lidy * BLOCK_SIZE + lidx] = query[min(queryIdx, query_rows - 1) * (step / sizeof(float)) + loadx];
+                s_train[lidx * BLOCK_SIZE + lidy] = train[min(t * BLOCK_SIZE + lidy, train_rows - 1) * (step / sizeof(float)) + loadx];
+            }
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            result += reduce_block(s_query, s_train, lidx, lidy);
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+        }
+
+        const int trainIdx = t * BLOCK_SIZE + lidx;
+
+        if (queryIdx < query_rows && trainIdx < train_rows /*&& mask(queryIdx, trainIdx)*/)
+        {
+            if (result < myBestDistance1)
+            {
+                myBestDistance2 = myBestDistance1;
+                myBestTrainIdx2 = myBestTrainIdx1;
+                myBestDistance1 = result;
+                myBestTrainIdx1 = trainIdx;
+            }
+            else if (result < myBestDistance2)
+            {
+                myBestDistance2 = result;
+                myBestTrainIdx2 = trainIdx;
+            }
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local float *s_distance = (__local float *)sharebuffer;
+    __local int *s_trainIdx = (__local int *)(sharebuffer + BLOCK_SIZE * BLOCK_SIZE);
+
+    //findBestMatch
+    s_distance += lidy * BLOCK_SIZE;
+    s_trainIdx += lidy * BLOCK_SIZE;
+
+    s_distance[lidx] = myBestDistance1;
+    s_trainIdx[lidx] = myBestTrainIdx1;
+
+    float bestDistance1 = MAX_FLOAT;
+    float bestDistance2 = MAX_FLOAT;
+    int bestTrainIdx1 = -1;
+    int bestTrainIdx2 = -1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (lidx == 0)
+    {
+        for (int i = 0 ; i < BLOCK_SIZE ; i++)
+        {
+            float val = s_distance[i];
+            if (val < bestDistance1)
+            {
+                bestDistance2 = bestDistance1;
+                bestTrainIdx2 = bestTrainIdx1;
+
+                bestDistance1 = val;
+                bestTrainIdx1 = s_trainIdx[i];
+            }
+            else if (val < bestDistance2)
+            {
+                bestDistance2 = val;
+                bestTrainIdx2 = s_trainIdx[i];
+            }
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    s_distance[lidx] = myBestDistance2;
+    s_trainIdx[lidx] = myBestTrainIdx2;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (lidx == 0)
+    {
+        for (int i = 0 ; i < BLOCK_SIZE ; i++)
+        {
+            float val = s_distance[i];
+
+            if (val < bestDistance2)
+            {
+                bestDistance2 = val;
+                bestTrainIdx2 = s_trainIdx[i];
+            }
+        }
+    }
+
+    myBestDistance1 = bestDistance1;
+    myBestDistance2 = bestDistance2;
+
+    myBestTrainIdx1 = bestTrainIdx1;
+    myBestTrainIdx2 = bestTrainIdx2;
+
+    if (queryIdx < query_rows && lidx == 0)
+    {
+        bestTrainIdx[queryIdx] = (int2)(myBestTrainIdx1, myBestTrainIdx2);
+        bestDistance[queryIdx] = (float2)(myBestDistance1, myBestDistance2);
+    }
+}
+
+kernel void BruteForceMatch_calcDistanceUnrolled(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global float *allDist,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step)
+{
+    /* Todo */
+}
+
+kernel void BruteForceMatch_calcDistance(
+    __global T *query,
+    __global T *train,
+    //__global float *mask,
+    __global float *allDist,
+    __local float *sharebuffer,
+    int query_rows,
+    int query_cols,
+    int train_rows,
+    int train_cols,
+    int step)
+{
+    /* Todo */
+}
+
+kernel void BruteForceMatch_findBestMatch(
+    __global float *allDist,
+    __global int *bestTrainIdx,
+    __global float *bestDistance,
+    int k
+)
+{
+    /* Todo */
+}
diff --git a/modules/ocl/src/kernels/build_warps.cl b/modules/ocl/src/opencl/build_warps.cl
similarity index 99%
rename from modules/ocl/src/kernels/build_warps.cl
rename to modules/ocl/src/opencl/build_warps.cl
index 13d7bb95ca..07cccee1a3 100644
--- a/modules/ocl/src/kernels/build_warps.cl
+++ b/modules/ocl/src/opencl/build_warps.cl
@@ -234,4 +234,3 @@ __kernel
         map_y[y * step_y + x] = ycoo;
     }
 }
-
diff --git a/modules/ocl/src/kernels/convertC3C4.cl b/modules/ocl/src/opencl/convertC3C4.cl
similarity index 100%
rename from modules/ocl/src/kernels/convertC3C4.cl
rename to modules/ocl/src/opencl/convertC3C4.cl
diff --git a/modules/ocl/src/kernels/cvt_color.cl b/modules/ocl/src/opencl/cvt_color.cl
similarity index 100%
rename from modules/ocl/src/kernels/cvt_color.cl
rename to modules/ocl/src/opencl/cvt_color.cl
diff --git a/modules/ocl/src/kernels/filter_sep_col.cl b/modules/ocl/src/opencl/filter_sep_col.cl
similarity index 100%
rename from modules/ocl/src/kernels/filter_sep_col.cl
rename to modules/ocl/src/opencl/filter_sep_col.cl
diff --git a/modules/ocl/src/kernels/filter_sep_row.cl b/modules/ocl/src/opencl/filter_sep_row.cl
similarity index 77%
rename from modules/ocl/src/kernels/filter_sep_row.cl
rename to modules/ocl/src/opencl/filter_sep_row.cl
index dbca8bd3a6..9dc498399d 100644
--- a/modules/ocl/src/kernels/filter_sep_row.cl
+++ b/modules/ocl/src/opencl/filter_sep_row.cl
@@ -96,18 +96,18 @@ The info above maybe obsolete.
 ***********************************************************************************/
 
 __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_C1_D0
-                        (__global const uchar * restrict src,
-                         __global float * dst,
-                         const int dst_cols,
-                         const int dst_rows,
-                         const int src_whole_cols,
-                         const int src_whole_rows,
-                         const int src_step_in_pixel,
-                         const int src_offset_x,
-                         const int src_offset_y,
-                         const int dst_step_in_pixel,
-                         const int radiusy,
-                         __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
+(__global const uchar * restrict src,
+ __global float * dst,
+ const int dst_cols,
+ const int dst_rows,
+ const int src_whole_cols,
+ const int src_whole_rows,
+ const int src_step_in_pixel,
+ const int src_offset_x,
+ const int src_offset_y,
+ const int dst_step_in_pixel,
+ const int radiusy,
+ __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
 {
     int x = get_global_id(0)<<2;
     int y = get_global_id(1);
@@ -122,17 +122,17 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     uchar4 temp[READ_TIMES_ROW];
 
     __local uchar4 LDS_DAT[LSIZE1][READ_TIMES_ROW*LSIZE0+1];
-    #ifdef BORDER_CONSTANT
+#ifdef BORDER_CONSTANT
     int end_addr = mad24(src_whole_rows - 1,src_step_in_pixel,src_whole_cols);
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         int current_addr = start_addr+i*LSIZE0*4;
         current_addr = ((current_addr < end_addr) && (current_addr > 0)) ? current_addr : 0;
         temp[i] = *(__global uchar4*)&src[current_addr];
     }
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         temp[i].x= ELEM(start_x+i*LSIZE0*4,0,src_whole_cols,0,temp[i].x);
         temp[i].y= ELEM(start_x+i*LSIZE0*4+1,0,src_whole_cols,0,temp[i].y);
@@ -140,7 +140,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         temp[i].w= ELEM(start_x+i*LSIZE0*4+3,0,src_whole_cols,0,temp[i].w);
         temp[i]= ELEM(start_y,0,src_whole_rows,(uchar4)0,temp[i]);
     }
-    #else
+#else
     int not_all_in_range = (start_x<0) | (start_x + READ_TIMES_ROW*LSIZE0*4+4>src_whole_cols)| (start_y<0) | (start_y >= src_whole_rows);
     int4 index[READ_TIMES_ROW];
     int4 addr;
@@ -148,7 +148,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     if(not_all_in_range)
     {
         //judge if read out of boundary
-        for(i = 0;i<READ_TIMES_ROW;i++)
+        for(i = 0; i<READ_TIMES_ROW; i++)
         {
             index[i].x= ADDR_L(start_x+i*LSIZE0*4,0,src_whole_cols,start_x+i*LSIZE0*4);
             index[i].x= ADDR_R(start_x+i*LSIZE0*4,src_whole_cols,index[i].x);
@@ -162,7 +162,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         s_y= ADDR_L(start_y,0,src_whole_rows,start_y);
         s_y= ADDR_R(start_y,src_whole_rows,s_y);
         //read pixels from src
-        for(i = 0;i<READ_TIMES_ROW;i++)
+        for(i = 0; i<READ_TIMES_ROW; i++)
         {
             addr = mad24((int4)s_y,(int4)src_step_in_pixel,index[i]);
             temp[i].x = src[addr.x];
@@ -174,15 +174,15 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     else
     {
         //read pixels from src
-        for(i = 0;i<READ_TIMES_ROW;i++)
+        for(i = 0; i<READ_TIMES_ROW; i++)
         {
             temp[i] = *(__global uchar4*)&src[start_addr+i*LSIZE0*4];
         }
     }
-    #endif
+#endif
 
     //save pixels to lds
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         LDS_DAT[l_y][l_x+i*LSIZE0]=temp[i];
     }
@@ -190,7 +190,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 
     //read pixels from lds and calculate the result
     sum =convert_float4(vload4(0,(__local uchar*)&LDS_DAT[l_y][l_x]+RADIUSX+offset))*mat_kernel[RADIUSX];
-    for(i=1;i<=RADIUSX;i++)
+    for(i=1; i<=RADIUSX; i++)
     {
         temp[0]=vload4(0,(__local uchar*)&LDS_DAT[l_y][l_x]+RADIUSX+offset-i);
         temp[1]=vload4(0,(__local uchar*)&LDS_DAT[l_y][l_x]+RADIUSX+offset+i);
@@ -219,18 +219,18 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     }
 }
 __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_C4_D0
-                        (__global const uchar4 * restrict src,
-                         __global float4 * dst,
-                         const int dst_cols,
-                         const int dst_rows,
-                         const int src_whole_cols,
-                         const int src_whole_rows,
-                         const int src_step_in_pixel,
-                         const int src_offset_x,
-                         const int src_offset_y,
-                         const int dst_step_in_pixel,
-                         const int radiusy,
-                         __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
+(__global const uchar4 * restrict src,
+ __global float4 * dst,
+ const int dst_cols,
+ const int dst_rows,
+ const int src_whole_cols,
+ const int src_whole_rows,
+ const int src_step_in_pixel,
+ const int src_offset_x,
+ const int src_offset_y,
+ const int dst_step_in_pixel,
+ const int radiusy,
+ __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -244,26 +244,26 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     uchar4 temp[READ_TIMES_ROW];
 
     __local uchar4 LDS_DAT[LSIZE1][READ_TIMES_ROW*LSIZE0+1];
-    #ifdef BORDER_CONSTANT
+#ifdef BORDER_CONSTANT
     int end_addr = mad24(src_whole_rows - 1,src_step_in_pixel,src_whole_cols);
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         int current_addr = start_addr+i*LSIZE0;
         current_addr = ((current_addr < end_addr) && (current_addr > 0)) ? current_addr : 0;
         temp[i] = src[current_addr];
     }
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         temp[i]= ELEM(start_x+i*LSIZE0,0,src_whole_cols,(uchar4)0,temp[i]);
         temp[i]= ELEM(start_y,0,src_whole_rows,(uchar4)0,temp[i]);
     }
-    #else
+#else
     int index[READ_TIMES_ROW];
     int s_x,s_y;
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         s_x= ADDR_L(start_x+i*LSIZE0,0,src_whole_cols,start_x+i*LSIZE0);
         s_x= ADDR_R(start_x+i*LSIZE0,src_whole_cols,s_x);
@@ -272,14 +272,14 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         index[i]=mad24(s_y,src_step_in_pixel,s_x);
     }
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         temp[i] = src[index[i]];
     }
-    #endif
+#endif
 
     //save pixels to lds
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         LDS_DAT[l_y][l_x+i*LSIZE0]=temp[i];
     }
@@ -287,7 +287,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 
     //read pixels from lds and calculate the result
     sum =convert_float4(LDS_DAT[l_y][l_x+RADIUSX])*mat_kernel[RADIUSX];
-    for(i=1;i<=RADIUSX;i++)
+    for(i=1; i<=RADIUSX; i++)
     {
         temp[0]=LDS_DAT[l_y][l_x+RADIUSX-i];
         temp[1]=LDS_DAT[l_y][l_x+RADIUSX+i];
@@ -302,18 +302,18 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 }
 
 __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_C1_D5
-                        (__global const float * restrict src,
-                         __global float * dst,
-                         const int dst_cols,
-                         const int dst_rows,
-                         const int src_whole_cols,
-                         const int src_whole_rows,
-                         const int src_step_in_pixel,
-                         const int src_offset_x,
-                         const int src_offset_y,
-                         const int dst_step_in_pixel,
-                         const int radiusy,
-                         __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
+(__global const float * restrict src,
+ __global float * dst,
+ const int dst_cols,
+ const int dst_rows,
+ const int src_whole_cols,
+ const int src_whole_rows,
+ const int src_step_in_pixel,
+ const int src_offset_x,
+ const int src_offset_y,
+ const int dst_step_in_pixel,
+ const int radiusy,
+ __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -327,26 +327,26 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     float temp[READ_TIMES_ROW];
 
     __local float LDS_DAT[LSIZE1][READ_TIMES_ROW*LSIZE0+1];
-    #ifdef BORDER_CONSTANT
+#ifdef BORDER_CONSTANT
     int end_addr = mad24(src_whole_rows - 1,src_step_in_pixel,src_whole_cols);
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         int current_addr = start_addr+i*LSIZE0;
         current_addr = ((current_addr < end_addr) && (current_addr > 0)) ? current_addr : 0;
         temp[i] = src[current_addr];
     }
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
-        temp[i]= ELEM(start_x+i*LSIZE0,0,src_whole_cols,0,temp[i]);
-        temp[i]= ELEM(start_y,0,src_whole_rows,0,temp[i]);
+        temp[i]= ELEM(start_x+i*LSIZE0,0,src_whole_cols,(float)0,temp[i]);
+        temp[i]= ELEM(start_y,0,src_whole_rows,(float)0,temp[i]);
     }
-    #else
+#else
     int index[READ_TIMES_ROW];
     int s_x,s_y;
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         s_x= ADDR_L(start_x+i*LSIZE0,0,src_whole_cols,start_x+i*LSIZE0);
         s_x= ADDR_R(start_x+i*LSIZE0,src_whole_cols,s_x);
@@ -355,14 +355,14 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         index[i]=mad24(s_y,src_step_in_pixel,s_x);
     }
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         temp[i] = src[index[i]];
     }
-    #endif
+#endif
 
     //save pixels to lds
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         LDS_DAT[l_y][l_x+i*LSIZE0]=temp[i];
     }
@@ -370,7 +370,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 
     //read pixels from lds and calculate the result
     sum =LDS_DAT[l_y][l_x+RADIUSX]*mat_kernel[RADIUSX];
-    for(i=1;i<=RADIUSX;i++)
+    for(i=1; i<=RADIUSX; i++)
     {
         temp[0]=LDS_DAT[l_y][l_x+RADIUSX-i];
         temp[1]=LDS_DAT[l_y][l_x+RADIUSX+i];
@@ -385,18 +385,18 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 }
 
 __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_C4_D5
-                        (__global const float4 * restrict src,
-                         __global float4 * dst,
-                         const int dst_cols,
-                         const int dst_rows,
-                         const int src_whole_cols,
-                         const int src_whole_rows,
-                         const int src_step_in_pixel,
-                         const int src_offset_x,
-                         const int src_offset_y,
-                         const int dst_step_in_pixel,
-                         const int radiusy,
-                         __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
+(__global const float4 * restrict src,
+ __global float4 * dst,
+ const int dst_cols,
+ const int dst_rows,
+ const int src_whole_cols,
+ const int src_whole_rows,
+ const int src_step_in_pixel,
+ const int src_offset_x,
+ const int src_offset_y,
+ const int dst_step_in_pixel,
+ const int radiusy,
+ __constant float * mat_kernel __attribute__((max_constant_size(4*(2*RADIUSX+1)))))
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
@@ -410,26 +410,26 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
     float4 temp[READ_TIMES_ROW];
 
     __local float4 LDS_DAT[LSIZE1][READ_TIMES_ROW*LSIZE0+1];
-    #ifdef BORDER_CONSTANT
+#ifdef BORDER_CONSTANT
     int end_addr = mad24(src_whole_rows - 1,src_step_in_pixel,src_whole_cols);
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         int current_addr = start_addr+i*LSIZE0;
         current_addr = ((current_addr < end_addr) && (current_addr > 0)) ? current_addr : 0;
         temp[i] = src[current_addr];
     }
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
-        temp[i]= ELEM(start_x+i*LSIZE0,0,src_whole_cols,0,temp[i]);
-        temp[i]= ELEM(start_y,0,src_whole_rows,0,temp[i]);
+        temp[i]= ELEM(start_x+i*LSIZE0,0,src_whole_cols,(float4)0,temp[i]);
+        temp[i]= ELEM(start_y,0,src_whole_rows,(float4)0,temp[i]);
     }
-    #else
+#else
     int index[READ_TIMES_ROW];
     int s_x,s_y;
     //judge if read out of boundary
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         s_x= ADDR_L(start_x+i*LSIZE0,0,src_whole_cols,start_x+i*LSIZE0);
         s_x= ADDR_R(start_x+i*LSIZE0,src_whole_cols,s_x);
@@ -438,14 +438,14 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         index[i]=mad24(s_y,src_step_in_pixel,s_x);
     }
     //read pixels from src
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         temp[i] = src[index[i]];
     }
-    #endif
+#endif
 
     //save pixels to lds
-    for(i = 0;i<READ_TIMES_ROW;i++)
+    for(i = 0; i<READ_TIMES_ROW; i++)
     {
         LDS_DAT[l_y][l_x+i*LSIZE0]=temp[i];
     }
@@ -453,7 +453,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
 
     //read pixels from lds and calculate the result
     sum =LDS_DAT[l_y][l_x+RADIUSX]*mat_kernel[RADIUSX];
-    for(i=1;i<=RADIUSX;i++)
+    for(i=1; i<=RADIUSX; i++)
     {
         temp[0]=LDS_DAT[l_y][l_x+RADIUSX-i];
         temp[1]=LDS_DAT[l_y][l_x+RADIUSX+i];
@@ -465,6 +465,5 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
         start_addr = mad24(y,dst_step_in_pixel,x);
         dst[start_addr] = sum;
     }
+
 }
-
-
diff --git a/modules/ocl/src/opencl/filtering_adaptive_bilateral.cl b/modules/ocl/src/opencl/filtering_adaptive_bilateral.cl
new file mode 100644
index 0000000000..a8e0fd17ec
--- /dev/null
+++ b/modules/ocl/src/opencl/filtering_adaptive_bilateral.cl
@@ -0,0 +1,424 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Harris Gasparakis, harris.gasparakis@amd.com
+//    Xiaopeng Fu, fuxiaopeng2222@163.com
+//    Yao Wang, bitwangyaoyao@gmail.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#ifdef BORDER_REPLICATE
+//BORDER_REPLICATE:     aaaaaa|abcdefgh|hhhhhhh
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? (l_edge)   : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? (r_edge)-1 : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? (t_edge)   :(i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (b_edge)-1 :(addr))
+#endif
+
+#ifdef BORDER_REFLECT
+//BORDER_REFLECT:       fedcba|abcdefgh|hgfedcb
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? -(i)-1               : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-1+((r_edge)<<1) : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? -(i)-1 : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-1+((b_edge)<<1) : (addr))
+#endif
+
+#ifdef BORDER_REFLECT_101
+//BORDER_REFLECT_101:   gfedcb|abcdefgh|gfedcba
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? -(i)                 : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-2+((r_edge)<<1) : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? -(i)                 : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-2+((b_edge)<<1) : (addr))
+#endif
+
+//blur function does not support BORDER_WRAP
+#ifdef BORDER_WRAP
+//BORDER_WRAP:          cdefgh|abcdefgh|abcdefg
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? (i)+(r_edge) : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? (i)-(r_edge) : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? (i)+(b_edge) : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (i)-(b_edge) : (addr))
+#endif
+
+__kernel void
+edgeEnhancingFilter_C4_D0(
+    __global const uchar4 * restrict src,
+    __global uchar4 *dst,
+    float alpha,
+    int src_offset,
+    int src_whole_rows,
+    int src_whole_cols,
+    int src_step,
+    int dst_offset,
+    int dst_rows,
+    int dst_cols,
+    int dst_step,
+    __global const float* lut,
+    int lut_step)
+{
+    int col = get_local_id(0);
+    const int gX = get_group_id(0);
+    const int gY = get_group_id(1);
+
+    int src_x_off = (src_offset % src_step) >> 2;
+    int src_y_off = src_offset / src_step;
+    int dst_x_off = (dst_offset % dst_step) >> 2;
+    int dst_y_off = dst_offset / dst_step;
+
+    int startX = gX * (THREADS-ksX+1) - anX + src_x_off;
+    int startY = (gY * (1+EXTRA)) - anY + src_y_off;
+
+    int dst_startX = gX * (THREADS-ksX+1) + dst_x_off;
+    int dst_startY = (gY * (1+EXTRA)) + dst_y_off;
+
+    int posX = dst_startX - dst_x_off + col;
+    int posY = (gY * (1+EXTRA))	;
+
+    __local uchar4 data[ksY+EXTRA][THREADS];
+
+    float4 tmp_sum[1+EXTRA];
+    for(int tmpint = 0; tmpint < 1+EXTRA; tmpint++)
+    {
+        tmp_sum[tmpint] = (float4)(0,0,0,0);
+    }
+
+#ifdef BORDER_CONSTANT
+    bool con;
+    uchar4 ss;
+    for(int j = 0;	j < ksY+EXTRA; j++)
+    {
+        con = (startX+col >= 0 && startX+col < src_whole_cols && startY+j >= 0 && startY+j < src_whole_rows);
+
+        int cur_col = clamp(startX + col, 0, src_whole_cols);
+        if(con)
+        {
+            ss = src[(startY+j)*(src_step>>2) + cur_col];
+        }
+
+        data[j][col] = con ? ss : (uchar4)0;
+    }
+#else
+    for(int j= 0; j < ksY+EXTRA; j++)
+    {
+        int selected_row;
+        int selected_col;
+        selected_row = ADDR_H(startY+j, 0, src_whole_rows);
+        selected_row = ADDR_B(startY+j, src_whole_rows, selected_row);
+
+        selected_col = ADDR_L(startX+col, 0, src_whole_cols);
+        selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
+
+        data[j][col] = src[selected_row * (src_step>>2) + selected_col];
+    }
+#endif
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float4 var[1+EXTRA];
+
+#if VAR_PER_CHANNEL
+    float4 weight;
+    float4 totalWeight = (float4)(0,0,0,0);
+#else
+    float weight;
+    float totalWeight = 0;
+#endif
+
+    int4 currValCenter;
+    int4 currWRTCenter;
+
+    int4 sumVal = 0;
+    int4 sumValSqr = 0;
+
+    if(col < (THREADS-(ksX-1)))
+    {
+        int4 currVal;
+
+        int howManyAll = (2*anX+1)*(ksY);
+
+        //find variance of all data
+        int startLMj;
+        int endLMj ;
+#if CALCVAR
+        // Top row: don't sum the very last element
+        for(int extraCnt = 0; extraCnt <=EXTRA; extraCnt++)
+        {
+            startLMj = extraCnt;
+            endLMj =  ksY+extraCnt-1;
+            sumVal =0;
+            sumValSqr=0;
+            for(int j = startLMj; j < endLMj; j++)
+            {
+                for(int i=-anX; i<=anX; i++)
+                {
+                    currVal	= convert_int4(data[j][col+anX+i])	;
+
+                    sumVal += currVal;
+                    sumValSqr += mul24(currVal, currVal);
+                }
+            }
+            var[extraCnt] = convert_float4( ( (sumValSqr * howManyAll)- mul24(sumVal , sumVal) ) ) /  ( (float)(howManyAll*howManyAll) ) ;
+#else
+        var[extraCnt] = (float4)(900.0, 900.0, 900.0, 0.0);
+#endif
+        }
+
+        for(int extraCnt = 0; extraCnt <= EXTRA; extraCnt++)
+        {
+
+            // top row: include the very first element, even on first time
+            startLMj = extraCnt;
+            // go all the way, unless this is the last local mem chunk,
+            // then stay within limits - 1
+            endLMj =  extraCnt + ksY;
+
+            // Top row: don't sum the very last element
+            currValCenter = convert_int4( data[ (startLMj + endLMj)/2][col+anX] );
+
+            for(int j = startLMj, lut_j = 0; j < endLMj; j++, lut_j++)
+            {
+                for(int i=-anX; i<=anX; i++)
+                {
+#if FIXED_WEIGHT
+#if VAR_PER_CHANNEL
+                    weight.x = 1.0f;
+                    weight.y = 1.0f;
+                    weight.z = 1.0f;
+                    weight.w = 1.0f;
+#else
+                    weight = 1.0f;
+#endif
+#else
+                    currVal	= convert_int4(data[j][col+anX+i])	;
+                    currWRTCenter = currVal-currValCenter;
+
+#if VAR_PER_CHANNEL
+                    weight = var[extraCnt] / (var[extraCnt] + convert_float4(currWRTCenter * currWRTCenter)) * (float4)(lut[lut_j*lut_step+anX+i]);
+                    //weight.x = var[extraCnt].x / ( var[extraCnt].x + (float) mul24(currWRTCenter.x , currWRTCenter.x) ) ;
+                    //weight.y = var[extraCnt].y / ( var[extraCnt].y + (float) mul24(currWRTCenter.y , currWRTCenter.y) ) ;
+                    //weight.z = var[extraCnt].z / ( var[extraCnt].z + (float) mul24(currWRTCenter.z , currWRTCenter.z) ) ;
+                    //weight.w = 0;
+#else
+                    weight = 1.0f/(1.0f+( mul24(currWRTCenter.x, currWRTCenter.x) + mul24(currWRTCenter.y, currWRTCenter.y) +  mul24(currWRTCenter.z, currWRTCenter.z))/(var.x+var.y+var.z));
+#endif
+#endif
+                    tmp_sum[extraCnt] += convert_float4(data[j][col+anX+i]) * weight;
+                    totalWeight += weight;
+                }
+            }
+
+            tmp_sum[extraCnt] /= totalWeight;
+
+            if(posX >= 0 && posX < dst_cols && (posY+extraCnt) >= 0 && (posY+extraCnt) < dst_rows)
+            {
+                dst[(dst_startY+extraCnt) * (dst_step>>2)+ dst_startX + col] = convert_uchar4(tmp_sum[extraCnt]);
+            }
+
+#if VAR_PER_CHANNEL
+            totalWeight = (float4)(0,0,0,0);
+#else
+            totalWeight = 0;
+#endif
+        }
+    }
+}
+
+
+__kernel void
+edgeEnhancingFilter_C1_D0(
+    __global const uchar * restrict src,
+    __global uchar *dst,
+    float alpha,
+    int src_offset,
+    int src_whole_rows,
+    int src_whole_cols,
+    int src_step,
+    int dst_offset,
+    int dst_rows,
+    int dst_cols,
+    int dst_step,
+    __global const float * lut,
+    int lut_step)
+{
+    int col = get_local_id(0);
+    const int gX = get_group_id(0);
+    const int gY = get_group_id(1);
+
+    int src_x_off = (src_offset % src_step);
+    int src_y_off = src_offset / src_step;
+    int dst_x_off = (dst_offset % dst_step);
+    int dst_y_off = dst_offset / dst_step;
+
+    int startX = gX * (THREADS-ksX+1) - anX + src_x_off;
+    int startY = (gY * (1+EXTRA)) - anY + src_y_off;
+
+    int dst_startX = gX * (THREADS-ksX+1) + dst_x_off;
+    int dst_startY = (gY * (1+EXTRA)) + dst_y_off;
+
+    int posX = dst_startX - dst_x_off + col;
+    int posY = (gY * (1+EXTRA))	;
+
+    __local uchar data[ksY+EXTRA][THREADS];
+
+    float tmp_sum[1+EXTRA];
+    for(int tmpint = 0; tmpint < 1+EXTRA; tmpint++)
+    {
+        tmp_sum[tmpint] = (float)(0);
+    }
+
+#ifdef BORDER_CONSTANT
+    bool con;
+    uchar ss;
+    for(int j = 0;	j < ksY+EXTRA; j++)
+    {
+        con = (startX+col >= 0 && startX+col < src_whole_cols && startY+j >= 0 && startY+j < src_whole_rows);
+
+        int cur_col = clamp(startX + col, 0, src_whole_cols);
+        if(con)
+        {
+            ss = src[(startY+j)*(src_step) + cur_col];
+        }
+
+        data[j][col] = con ? ss : 0;
+    }
+#else
+    for(int j= 0; j < ksY+EXTRA; j++)
+    {
+        int selected_row;
+        int selected_col;
+        selected_row = ADDR_H(startY+j, 0, src_whole_rows);
+        selected_row = ADDR_B(startY+j, src_whole_rows, selected_row);
+
+        selected_col = ADDR_L(startX+col, 0, src_whole_cols);
+        selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
+
+        data[j][col] = src[selected_row * (src_step) + selected_col];
+    }
+#endif
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float var[1+EXTRA];
+
+    float weight;
+    float totalWeight = 0;
+
+    int currValCenter;
+    int currWRTCenter;
+
+    int sumVal = 0;
+    int sumValSqr = 0;
+
+    if(col < (THREADS-(ksX-1)))
+    {
+        int currVal;
+
+        int howManyAll = (2*anX+1)*(ksY);
+
+        //find variance of all data
+        int startLMj;
+        int endLMj;
+#if CALCVAR
+        // Top row: don't sum the very last element
+        for(int extraCnt=0; extraCnt<=EXTRA; extraCnt++)
+        {
+            startLMj = extraCnt;
+            endLMj =  ksY+extraCnt-1;
+            sumVal = 0;
+            sumValSqr =0;
+            for(int j = startLMj; j < endLMj; j++)
+            {
+                for(int i=-anX; i<=anX; i++)
+                {
+                    currVal	= (uint)(data[j][col+anX+i])	;
+
+                    sumVal += currVal;
+                    sumValSqr += mul24(currVal, currVal);
+                }
+            }
+            var[extraCnt] = (float)( ( (sumValSqr * howManyAll)- mul24(sumVal , sumVal) ) ) /  ( (float)(howManyAll*howManyAll) ) ;
+#else
+        var[extraCnt] = (float)(900.0);
+#endif
+        }
+
+        for(int extraCnt = 0; extraCnt <= EXTRA; extraCnt++)
+        {
+
+            // top row: include the very first element, even on first time
+            startLMj = extraCnt;
+            // go all the way, unless this is the last local mem chunk,
+            // then stay within limits - 1
+            endLMj =  extraCnt + ksY;
+
+            // Top row: don't sum the very last element
+            currValCenter = (int)( data[ (startLMj + endLMj)/2][col+anX] );
+
+            for(int j = startLMj, lut_j = 0; j < endLMj; j++, lut_j++)
+            {
+                for(int i=-anX; i<=anX; i++)
+                {
+#if FIXED_WEIGHT
+                    weight = 1.0f;
+#else
+                    currVal	= (int)(data[j][col+anX+i])	;
+                    currWRTCenter = currVal-currValCenter;
+
+                    weight = var[extraCnt] / (var[extraCnt] + (float)mul24(currWRTCenter,currWRTCenter)) * lut[lut_j*lut_step+anX+i] ;
+#endif
+                    tmp_sum[extraCnt] += (float)(data[j][col+anX+i] * weight);
+                    totalWeight += weight;
+                }
+            }
+
+            tmp_sum[extraCnt] /= totalWeight;
+
+
+            if(posX >= 0 && posX < dst_cols && (posY+extraCnt) >= 0 && (posY+extraCnt) < dst_rows)
+            {
+                dst[(dst_startY+extraCnt) * (dst_step)+ dst_startX + col] = (uchar)(tmp_sum[extraCnt]);
+            }
+
+            totalWeight = 0;
+        }
+    }
+}
diff --git a/modules/ocl/src/kernels/filtering_boxFilter.cl b/modules/ocl/src/opencl/filtering_boxFilter.cl
similarity index 56%
rename from modules/ocl/src/kernels/filtering_boxFilter.cl
rename to modules/ocl/src/opencl/filtering_boxFilter.cl
index 79ca8d735b..d163ebe76a 100644
--- a/modules/ocl/src/kernels/filtering_boxFilter.cl
+++ b/modules/ocl/src/opencl/filtering_boxFilter.cl
@@ -79,15 +79,73 @@
 #define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (i)-(b_edge) : (addr))
 #endif
 
+#define THREADS 256
+#define ELEM(i, l_edge, r_edge, elem1, elem2) (i) >= (l_edge) && (i) < (r_edge) ? (elem1) : (elem2)
+
+inline void update_dst_C1_D0(__global uchar *dst, __local uint* temp,
+                             int dst_rows, int dst_cols,
+                             int dst_startX, int dst_x_off,
+                             float alpha)
+{
+    if(get_local_id(0) < anX || get_local_id(0) >= (THREADS-ksX+anX+1))
+    {
+        return;
+    }
+
+    uint4 tmp_sum = 0;
+    int posX = dst_startX - dst_x_off + (get_local_id(0)-anX)*4;
+    int posY = (get_group_id(1) << 1);
+
+    for(int i=-anX; i<=anX; i++)
+    {
+        tmp_sum += vload4(get_local_id(0), temp+i);
+    }
+
+    if(posY < dst_rows && posX < dst_cols)
+    {
+        tmp_sum /= (uint4) alpha;
+        if(posX >= 0 && posX < dst_cols)
+            *(dst) = tmp_sum.x;
+        if(posX+1 >= 0 && posX+1 < dst_cols)
+            *(dst + 1) = tmp_sum.y;
+        if(posX+2 >= 0 && posX+2 < dst_cols)
+            *(dst + 2) = tmp_sum.z;
+        if(posX+3 >= 0 && posX+3 < dst_cols)
+            *(dst + 3) = tmp_sum.w;
+    }
+}
+
+
+inline void update_dst_C4_D0(__global uchar4 *dst, __local uint4* temp,
+                             int dst_rows, int dst_cols,
+                             int dst_startX, int dst_x_off,
+                             float alpha)
+{
+    if(get_local_id(0) >= (THREADS-ksX+1))
+    {
+        return;
+    }
+
+    int posX = dst_startX - dst_x_off + get_local_id(0);
+    int posY = (get_group_id(1) << 1);
+
+    uint4 temp_sum = 0;
+    for(int i=-anX; i<=anX; i++)
+    {
+        temp_sum += temp[get_local_id(0) + anX + i];
+    }
+
+    if(posX >= 0 && posX < dst_cols && posY >= 0 && posY < dst_rows)
+        *dst = convert_uchar4(convert_float4(temp_sum)/alpha);
+}
+
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 /////////////////////////////////////////8uC1////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////////////////////////
-#define THREADS 256
-#define ELEM(i, l_edge, r_edge, elem1, elem2) (i) >= (l_edge) && (i) < (r_edge) ? (elem1) : (elem2)
 __kernel void boxFilter_C1_D0(__global const uchar * restrict src, __global uchar *dst, float alpha,
-                                     int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
-                                     int dst_offset, int dst_rows, int dst_cols, int dst_step
-                                     )
+                              int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
+                              int dst_offset, int dst_rows, int dst_cols, int dst_step
+                             )
 {
 
     int col = get_local_id(0);
@@ -105,115 +163,84 @@ __kernel void boxFilter_C1_D0(__global const uchar * restrict src, __global ucha
     int dst_startY = (gY << 1) + dst_y_off;
 
     uint4 data[ksY+1];
-    __local uint4 temp[(THREADS<<1)];
+    __local uint4 temp[2][THREADS];
 
 #ifdef BORDER_CONSTANT
 
-        for(int i=0; i < ksY+1; i++)
+    for(int i=0; i < ksY+1; i++)
+    {
+        if(startY+i >=0 && startY+i < src_whole_rows && startX+col*4 >=0 && startX+col*4+3<src_whole_cols)
         {
-            if(startY+i >=0 && startY+i < src_whole_rows && startX+col*4 >=0 && startX+col*4+3<src_whole_cols)
-                data[i] = convert_uint4(vload4(col,(__global uchar*)(src+(startY+i)*src_step + startX)));
-            else
-            {
-                data[i]=0;
-                int con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4 >=0 && startX+col*4<src_whole_cols;
-                if(con)data[i].s0 = *(src+(startY+i)*src_step + startX + col*4);
-                con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+1 >=0 && startX+col*4+1<src_whole_cols;
-                if(con)data[i].s1 = *(src+(startY+i)*src_step + startX + col*4+1) ;
-                con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+2 >=0 && startX+col*4+2<src_whole_cols;
-                if(con)data[i].s2 = *(src+(startY+i)*src_step + startX + col*4+2);
-                con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+3 >=0 && startX+col*4+3<src_whole_cols;
-                if(con)data[i].s3 = *(src+(startY+i)*src_step + startX + col*4+3);
-            }
+            data[i].x = *(src+(startY+i)*src_step + startX + col * 4);
+            data[i].y = *(src+(startY+i)*src_step + startX + col * 4 + 1);
+            data[i].z = *(src+(startY+i)*src_step + startX + col * 4 + 2);
+            data[i].w = *(src+(startY+i)*src_step + startX + col * 4 + 3);
         }
+        else
+        {
+            data[i]=0;
+            int con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4 >=0 && startX+col*4<src_whole_cols;
+            if(con)data[i].s0 = *(src+(startY+i)*src_step + startX + col*4);
+            con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+1 >=0 && startX+col*4+1<src_whole_cols;
+            if(con)data[i].s1 = *(src+(startY+i)*src_step + startX + col*4+1) ;
+            con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+2 >=0 && startX+col*4+2<src_whole_cols;
+            if(con)data[i].s2 = *(src+(startY+i)*src_step + startX + col*4+2);
+            con = startY+i >=0 && startY+i < src_whole_rows && startX+col*4+3 >=0 && startX+col*4+3<src_whole_cols;
+            if(con)data[i].s3 = *(src+(startY+i)*src_step + startX + col*4+3);
+        }
+    }
 
 #else
-   int not_all_in_range;
-   for(int i=0; i < ksY+1; i++)
-   {
-      not_all_in_range = (startX+col*4<0) | (startX+col*4+3>src_whole_cols-1)
-                        | (startY+i<0) | (startY+i>src_whole_rows-1);
-      if(not_all_in_range)
-      {
-          int selected_row;
-          int4 selected_col;
-          selected_row = ADDR_H(startY+i, 0, src_whole_rows);
-          selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
+    int not_all_in_range;
+    for(int i=0; i < ksY+1; i++)
+    {
+        not_all_in_range = (startX+col*4<0) | (startX+col*4+3>src_whole_cols-1)
+                           | (startY+i<0) | (startY+i>src_whole_rows-1);
+        if(not_all_in_range)
+        {
+            int selected_row;
+            int4 selected_col;
+            selected_row = ADDR_H(startY+i, 0, src_whole_rows);
+            selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
 
-          selected_col.x = ADDR_L(startX+col*4, 0, src_whole_cols);
-          selected_col.x = ADDR_R(startX+col*4, src_whole_cols, selected_col.x);
+            selected_col.x = ADDR_L(startX+col*4, 0, src_whole_cols);
+            selected_col.x = ADDR_R(startX+col*4, src_whole_cols, selected_col.x);
 
-          selected_col.y = ADDR_L(startX+col*4+1, 0, src_whole_cols);
-          selected_col.y = ADDR_R(startX+col*4+1, src_whole_cols, selected_col.y);
+            selected_col.y = ADDR_L(startX+col*4+1, 0, src_whole_cols);
+            selected_col.y = ADDR_R(startX+col*4+1, src_whole_cols, selected_col.y);
 
-          selected_col.z = ADDR_L(startX+col*4+2, 0, src_whole_cols);
-          selected_col.z = ADDR_R(startX+col*4+2, src_whole_cols, selected_col.z);
+            selected_col.z = ADDR_L(startX+col*4+2, 0, src_whole_cols);
+            selected_col.z = ADDR_R(startX+col*4+2, src_whole_cols, selected_col.z);
 
-          selected_col.w = ADDR_L(startX+col*4+3, 0, src_whole_cols);
-          selected_col.w = ADDR_R(startX+col*4+3, src_whole_cols, selected_col.w);
+            selected_col.w = ADDR_L(startX+col*4+3, 0, src_whole_cols);
+            selected_col.w = ADDR_R(startX+col*4+3, src_whole_cols, selected_col.w);
 
-          data[i].x = *(src + selected_row * src_step + selected_col.x);
-          data[i].y = *(src + selected_row * src_step + selected_col.y);
-          data[i].z = *(src + selected_row * src_step + selected_col.z);
-          data[i].w = *(src + selected_row * src_step + selected_col.w);
-      }
-      else
-      {
-          data[i] =  convert_uint4(vload4(col,(__global uchar*)(src+(startY+i)*src_step + startX)));
-      }
-   }
+            data[i].x = *(src + selected_row * src_step + selected_col.x);
+            data[i].y = *(src + selected_row * src_step + selected_col.y);
+            data[i].z = *(src + selected_row * src_step + selected_col.z);
+            data[i].w = *(src + selected_row * src_step + selected_col.w);
+        }
+        else
+        {
+            data[i] =  convert_uint4(vload4(col,(__global uchar*)(src+(startY+i)*src_step + startX)));
+        }
+    }
 #endif
-    uint4 sum0 = 0, sum1 = 0, sum2 = 0;
+    uint4 tmp_sum = 0;
     for(int i=1; i < ksY; i++)
     {
-        sum0 += (data[i]);
+        tmp_sum += (data[i]);
     }
-    sum1 = sum0 + (data[0]);
-    sum2 = sum0 + (data[ksY]);
 
-    temp[col] = sum1;
-    temp[col+THREADS] = sum2;
+    int index = dst_startY * dst_step + dst_startX + (col-anX)*4;
+
+    temp[0][col] = tmp_sum + (data[0]);
+    temp[1][col] = tmp_sum + (data[ksY]);
     barrier(CLK_LOCAL_MEM_FENCE);
-
-    if(col >= anX && col < (THREADS-ksX+anX+1))
-    {
-        int posX = dst_startX - dst_x_off + (col-anX)*4;
-        int posY = (gY << 1);
-        uint4 tmp_sum1=0, tmp_sum2=0;
-        for(int i=-anX; i<=anX; i++)
-        {
-           tmp_sum1 += vload4(col, (__local uint*)temp+i);
-        }
-
-        for(int i=-anX; i<=anX; i++)
-        {
-           tmp_sum2 += vload4(col, (__local uint*)(temp+THREADS)+i);
-        }
-
-        if(posY < dst_rows && posX < dst_cols)
-        {
-           if(posX >= 0 && posX < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX + (col-anX)*4) = tmp_sum1.x/alpha;
-           if(posX+1 >= 0 && posX+1 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+1 + (col-anX)*4) = tmp_sum1.y/alpha;
-           if(posX+2 >= 0 && posX+2 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+2 + (col-anX)*4) = tmp_sum1.z/alpha;
-           if(posX+3 >= 0 && posX+3 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+3 + (col-anX)*4) = tmp_sum1.w/alpha;
-        }
-        if(posY+1 < dst_rows && posX < dst_cols)
-        {
-           dst_startY+=1;
-           if(posX >= 0 && posX < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX + (col-anX)*4) = tmp_sum2.x/alpha;
-           if(posX+1 >= 0 && posX+1 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+1 + (col-anX)*4) = tmp_sum2.y/alpha;
-           if(posX+2 >= 0 && posX+2 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+2 + (col-anX)*4) = tmp_sum2.z/alpha;
-           if(posX+3 >= 0 && posX+3 < dst_cols)
-               *(dst+dst_startY * dst_step + dst_startX+3 + (col-anX)*4) = tmp_sum2.w/alpha;
-        }
-    }
+    update_dst_C1_D0(dst+index, (__local uint *)(temp[0]),
+                     dst_rows, dst_cols, dst_startX, dst_x_off, alpha);
+    update_dst_C1_D0(dst+index+dst_step, (__local uint *)(temp[1]),
+                     dst_rows, dst_cols, dst_startX, dst_x_off, alpha);
 
 }
 
@@ -221,9 +248,9 @@ __kernel void boxFilter_C1_D0(__global const uchar * restrict src, __global ucha
 /////////////////////////////////////////8uC4////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 __kernel void boxFilter_C4_D0(__global const uchar4 * restrict src, __global uchar4 *dst, float alpha,
-                                     int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
-                                     int dst_offset, int dst_rows, int dst_cols, int dst_step
-                                     )
+                              int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
+                              int dst_offset, int dst_rows, int dst_cols, int dst_step
+                             )
 {
     int col = get_local_id(0);
     const int gX = get_group_id(0);
@@ -238,81 +265,63 @@ __kernel void boxFilter_C4_D0(__global const uchar4 * restrict src, __global uch
     int startY = (gY << 1) - anY + src_y_off;
     int dst_startX = gX * (THREADS-ksX+1) + dst_x_off;
     int dst_startY = (gY << 1) + dst_y_off;
-      //int end_addr = (src_whole_rows-1)*(src_step>>2) + src_whole_cols-4;
 
-      int end_addr = src_whole_cols-4;
     uint4 data[ksY+1];
     __local uint4 temp[2][THREADS];
+
 #ifdef BORDER_CONSTANT
     bool con;
-    uint4 ss;
     for(int i=0; i < ksY+1; i++)
     {
         con = startX+col >= 0 && startX+col < src_whole_cols && startY+i >= 0 && startY+i < src_whole_rows;
-
-            //int cur_addr = clamp((startY+i)*(src_step>>2)+(startX+col),0,end_addr);
-        //ss = convert_uint4(src[cur_addr]);
-
         int cur_col = clamp(startX + col, 0, src_whole_cols);
-        if(con)
-          ss = convert_uint4(src[(startY+i)*(src_step>>2) + cur_col]);
 
-        data[i] = con ? ss : 0;
+        data[i].x = con ? src[(startY+i)*(src_step>>2) + cur_col].x : 0;
+        data[i].y = con ? src[(startY+i)*(src_step>>2) + cur_col].y : 0;
+        data[i].z = con ? src[(startY+i)*(src_step>>2) + cur_col].z : 0;
+        data[i].w = con ? src[(startY+i)*(src_step>>2) + cur_col].w : 0;
     }
 #else
-   for(int i=0; i < ksY+1; i++)
-   {
-          int selected_row;
-          int selected_col;
-          selected_row = ADDR_H(startY+i, 0, src_whole_rows);
-          selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
+    for(int i=0; i < ksY+1; i++)
+    {
+        int selected_row;
+        int selected_col;
+        selected_row = ADDR_H(startY+i, 0, src_whole_rows);
+        selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
 
-          selected_col = ADDR_L(startX+col, 0, src_whole_cols);
-          selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
+        selected_col = ADDR_L(startX+col, 0, src_whole_cols);
+        selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
 
 
-          data[i] = convert_uint4(src[selected_row * (src_step>>2) + selected_col]);
-   }
+        data[i] = convert_uint4(src[selected_row * (src_step>>2) + selected_col]);
+    }
 
 #endif
-    uint4 sum0 = 0, sum1 = 0, sum2 = 0;
+    uint4 tmp_sum = 0;
     for(int i=1; i < ksY; i++)
     {
-        sum0 += (data[i]);
+        tmp_sum += (data[i]);
     }
-    sum1 = sum0 + (data[0]);
-    sum2 = sum0 + (data[ksY]);
-    temp[0][col] = sum1;
-    temp[1][col] = sum2;
+
+    int index = dst_startY * (dst_step>>2)+ dst_startX + col;
+
+    temp[0][col] = tmp_sum + (data[0]);
+    temp[1][col] = tmp_sum + (data[ksY]);
     barrier(CLK_LOCAL_MEM_FENCE);
-    if(col < (THREADS-(ksX-1)))
-    {
-        col += anX;
-        int posX = dst_startX - dst_x_off + col - anX;
-        int posY = (gY << 1);
+    update_dst_C4_D0(dst+index, (__local uint4 *)(temp[0]),
+                     dst_rows, dst_cols, dst_startX, dst_x_off, alpha);
+    update_dst_C4_D0(dst+index+(dst_step>>2), (__local uint4 *)(temp[1]),
+                     dst_rows, dst_cols, dst_startX, dst_x_off, alpha);
 
-        uint4 tmp_sum[2]={(uint4)(0,0,0,0),(uint4)(0,0,0,0)};
-        for(int k=0; k<2; k++)
-            for(int i=-anX; i<=anX; i++)
-            {
-                tmp_sum[k] += temp[k][col+i];
-            }
-        for(int i=0; i<2; i++)
-        {
-            if(posX >= 0 && posX < dst_cols && (posY+i) >= 0 && (posY+i) < dst_rows)
-                dst[(dst_startY+i) * (dst_step>>2)+ dst_startX + col - anX] = convert_uchar4(convert_float4(tmp_sum[i])/alpha);
-        }
-
-    }
 }
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 /////////////////////////////////////////32fC1////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 __kernel void boxFilter_C1_D5(__global const float *restrict src, __global float *dst, float alpha,
-                                     int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
-                                     int dst_offset, int dst_rows, int dst_cols, int dst_step
-                                     )
+                              int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
+                              int dst_offset, int dst_rows, int dst_cols, int dst_step
+                             )
 {
     int col = get_local_id(0);
     const int gX = get_group_id(0);
@@ -327,7 +336,6 @@ __kernel void boxFilter_C1_D5(__global const float *restrict src, __global float
     int startY = (gY << 1) - anY + src_y_off;
     int dst_startX = gX * (THREADS-ksX+1) + dst_x_off;
     int dst_startY = (gY << 1) + dst_y_off;
-    int end_addr = (src_whole_rows-1)*(src_step>>2) + src_whole_cols-4;
     float data[ksY+1];
     __local float temp[2][THREADS];
 #ifdef BORDER_CONSTANT
@@ -336,28 +344,25 @@ __kernel void boxFilter_C1_D5(__global const float *restrict src, __global float
     for(int i=0; i < ksY+1; i++)
     {
         con = startX+col >= 0 && startX+col < src_whole_cols && startY+i >= 0 && startY+i < src_whole_rows;
-          //int cur_addr = clamp((startY+i)*(src_step>>2)+(startX+col),0,end_addr);
-        //ss = src[cur_addr];
 
         int cur_col = clamp(startX + col, 0, src_whole_cols);
-        //ss = src[(startY+i)*(src_step>>2) + cur_col];
-        ss = (startY+i)<src_whole_rows&&(startY+i)>=0&&cur_col>=0&&cur_col<src_whole_cols?src[(startY+i)*(src_step>>2) + cur_col]:0;
+        ss = (startY+i)<src_whole_rows&&(startY+i)>=0&&cur_col>=0&&cur_col<src_whole_cols?src[(startY+i)*(src_step>>2) + cur_col]:(float)0;
 
         data[i] = con ? ss : 0.f;
     }
 #else
-   for(int i=0; i < ksY+1; i++)
-   {
-          int selected_row;
-          int selected_col;
-          selected_row = ADDR_H(startY+i, 0, src_whole_rows);
-          selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
+    for(int i=0; i < ksY+1; i++)
+    {
+        int selected_row;
+        int selected_col;
+        selected_row = ADDR_H(startY+i, 0, src_whole_rows);
+        selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
 
-          selected_col = ADDR_L(startX+col, 0, src_whole_cols);
-          selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
+        selected_col = ADDR_L(startX+col, 0, src_whole_cols);
+        selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
 
-          data[i] = src[selected_row * (src_step>>2) + selected_col];
-   }
+        data[i] = src[selected_row * (src_step>>2) + selected_col];
+    }
 
 #endif
     float sum0 = 0.0, sum1 = 0.0, sum2 = 0.0;
@@ -376,7 +381,7 @@ __kernel void boxFilter_C1_D5(__global const float *restrict src, __global float
         int posX = dst_startX - dst_x_off + col - anX;
         int posY = (gY << 1);
 
-        float tmp_sum[2]={0.0, 0.0};
+        float tmp_sum[2]= {0.0, 0.0};
         for(int k=0; k<2; k++)
             for(int i=-anX; i<=anX; i++)
             {
@@ -395,9 +400,9 @@ __kernel void boxFilter_C1_D5(__global const float *restrict src, __global float
 /////////////////////////////////////////32fC4////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 __kernel void boxFilter_C4_D5(__global const float4 *restrict src, __global float4 *dst, float alpha,
-                                     int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
-                                     int dst_offset, int dst_rows, int dst_cols, int dst_step
-                                     )
+                              int src_offset, int src_whole_rows, int src_whole_cols, int src_step,
+                              int dst_offset, int dst_rows, int dst_cols, int dst_step
+                             )
 {
     int col = get_local_id(0);
     const int gX = get_group_id(0);
@@ -412,7 +417,6 @@ __kernel void boxFilter_C4_D5(__global const float4 *restrict src, __global floa
     int startY = (gY << 1) - anY + src_y_off;
     int dst_startX = gX * (THREADS-ksX+1) + dst_x_off;
     int dst_startY = (gY << 1) + dst_y_off;
-    int end_addr = (src_whole_rows-1)*(src_step>>4) + src_whole_cols-16;
     float4 data[ksY+1];
     __local float4 temp[2][THREADS];
 #ifdef BORDER_CONSTANT
@@ -421,28 +425,25 @@ __kernel void boxFilter_C4_D5(__global const float4 *restrict src, __global floa
     for(int i=0; i < ksY+1; i++)
     {
         con = startX+col >= 0 && startX+col < src_whole_cols && startY+i >= 0 && startY+i < src_whole_rows;
-            //int cur_addr = clamp((startY+i)*(src_step>>4)+(startX+col),0,end_addr);
-        //ss = src[cur_addr];
 
         int cur_col = clamp(startX + col, 0, src_whole_cols);
-        //ss = src[(startY+i)*(src_step>>4) + cur_col];
-        ss = (startY+i)<src_whole_rows&&(startY+i)>=0&&cur_col>=0&&cur_col<src_whole_cols?src[(startY+i)*(src_step>>4) + cur_col]:0;
+        ss = (startY+i)<src_whole_rows&&(startY+i)>=0&&cur_col>=0&&cur_col<src_whole_cols?src[(startY+i)*(src_step>>4) + cur_col]:(float4)0;
 
         data[i] = con ? ss : (float4)(0.0,0.0,0.0,0.0);
     }
 #else
-   for(int i=0; i < ksY+1; i++)
-   {
-          int selected_row;
-          int selected_col;
-          selected_row = ADDR_H(startY+i, 0, src_whole_rows);
-          selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
+    for(int i=0; i < ksY+1; i++)
+    {
+        int selected_row;
+        int selected_col;
+        selected_row = ADDR_H(startY+i, 0, src_whole_rows);
+        selected_row = ADDR_B(startY+i, src_whole_rows, selected_row);
 
-          selected_col = ADDR_L(startX+col, 0, src_whole_cols);
-          selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
+        selected_col = ADDR_L(startX+col, 0, src_whole_cols);
+        selected_col = ADDR_R(startX+col, src_whole_cols, selected_col);
 
-          data[i] = src[selected_row * (src_step>>4) + selected_col];
-   }
+        data[i] = src[selected_row * (src_step>>4) + selected_col];
+    }
 
 #endif
     float4 sum0 = 0.0, sum1 = 0.0, sum2 = 0.0;
@@ -461,7 +462,7 @@ __kernel void boxFilter_C4_D5(__global const float4 *restrict src, __global floa
         int posX = dst_startX - dst_x_off + col - anX;
         int posY = (gY << 1);
 
-        float4 tmp_sum[2]={(float4)(0.0,0.0,0.0,0.0), (float4)(0.0,0.0,0.0,0.0)};
+        float4 tmp_sum[2]= {(float4)(0.0,0.0,0.0,0.0), (float4)(0.0,0.0,0.0,0.0)};
         for(int k=0; k<2; k++)
             for(int i=-anX; i<=anX; i++)
             {
diff --git a/modules/ocl/src/opencl/filtering_laplacian.cl b/modules/ocl/src/opencl/filtering_laplacian.cl
new file mode 100644
index 0000000000..f7430d5332
--- /dev/null
+++ b/modules/ocl/src/opencl/filtering_laplacian.cl
@@ -0,0 +1,376 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Pang Erping, erping@multicorewareinc.com
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+///////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////Macro for border type////////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef BORDER_REPLICATE
+
+//BORDER_REPLICATE:     aaaaaa|abcdefgh|hhhhhhh
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? (l_edge)   : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? (r_edge)-1 : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? (t_edge)   : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? (b_edge)-1 :(addr))
+#endif
+
+#ifdef BORDER_REFLECT
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? ((l_edge)<<1)-(i)-1                 : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-1+((r_edge)<<1) : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? ((t_edge)<<1)-(i)-1                 : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-1+((b_edge)<<1) : (addr))
+#endif
+
+#ifdef BORDER_REFLECT_101
+//BORDER_REFLECT_101:   gfedcb|abcdefgh|gfedcba
+#define ADDR_L(i, l_edge, r_edge)  ((i) <  (l_edge) ? ((l_edge)<<1)-(i)                 : (i))
+#define ADDR_R(i, r_edge, addr)    ((i) >= (r_edge) ? -(i)-2+((r_edge)<<1) : (addr))
+#define ADDR_H(i, t_edge, b_edge)  ((i) <  (t_edge) ? ((t_edge)<<1)-(i)                 : (i))
+#define ADDR_B(i, b_edge, addr)    ((i) >= (b_edge) ? -(i)-2+((b_edge)<<1) : (addr))
+#endif
+
+#ifdef IMG_C_1_0
+#define T_IMG   uchar
+#define T_IMGx4 uchar4
+#define T_IMG_C1 uchar
+#define CONVERT_TYPE   convert_uchar_sat
+#define CONVERT_TYPEx4 convert_uchar4_sat
+#endif
+#ifdef IMG_C_4_0
+#define T_IMG   uchar4
+#define T_IMGx4 uchar16
+#define T_IMG_C1 uchar
+#define CONVERT_TYPE   convert_uchar4_sat
+#define CONVERT_TYPEx4 convert_uchar16_sat
+#endif
+#ifdef IMG_C_1_5
+#define T_IMG   float
+#define T_IMGx4 float4
+#define T_IMG_C1 float
+#define CONVERT_TYPE   convert_float
+#define CONVERT_TYPEx4 convert_float4
+#endif
+#ifdef IMG_C_4_5
+#define T_IMG   float4
+#define T_IMGx4 float16
+#define T_IMG_C1 float
+#define CONVERT_TYPE   convert_float4
+#define CONVERT_TYPEx4 convert_float16
+#endif
+
+#ifndef CN
+#define CN 1
+#endif
+
+#if CN == 1
+#define T_SUM   float
+#define T_SUMx4 float4
+#define CONVERT_TYPE_SUM   convert_float
+#define CONVERT_TYPE_SUMx4 convert_float4
+#define SUM_ZERO   (0.0f)
+#define SUM_ZEROx4 (0.0f, 0.0f, 0.0f, 0.0f)
+#define VLOAD4 vload4
+#define SX x
+#define SY y
+#define SZ z
+#define SW w
+#elif CN == 4
+#define T_SUM float4
+#define T_SUMx4 float16
+#define CONVERT_TYPE_SUM   convert_float4
+#define CONVERT_TYPE_SUMx4 convert_float16
+#define SUM_ZERO   (0.0f, 0.0f, 0.0f, 0.0f)
+#define SUM_ZEROx4 (0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f)
+#define VLOAD4 vload16
+#define SX s0123
+#define SY s4567
+#define SZ s89ab
+#define SW scdef
+#endif
+
+#ifndef FILTER_SIZE
+#define FILTER_SIZE 3
+#endif
+
+#define LOCAL_GROUP_SIZE 16
+
+#define LOCAL_WIDTH  ((FILTER_SIZE/2)*2 + LOCAL_GROUP_SIZE)
+#define LOCAL_HEIGHT ((FILTER_SIZE/2)*2 + LOCAL_GROUP_SIZE)
+
+#define FILTER_RADIUS (FILTER_SIZE >> 1)
+
+__kernel void filter2D(
+    __global T_IMG *src,
+    __global T_IMG *dst,
+    int src_step,
+    int dst_step,
+    __constant float *mat_kernel,
+    __local T_IMG *local_data,
+    int wholerows,
+    int wholecols,
+    int src_offset_x,
+    int src_offset_y,
+    int dst_offset_x,
+    int dst_offset_y,
+    int cols,
+    int rows,
+    int operate_cols
+)
+{
+    int groupStartCol = get_group_id(0) * get_local_size(0);
+    int groupStartRow = get_group_id(1) * get_local_size(1);
+
+    int localCol = get_local_id(0);
+    int localRow = get_local_id(1);
+    int globalCol = groupStartCol + localCol;
+    int globalRow = groupStartRow + localRow;
+    const int src_offset = mad24(src_offset_y, src_step, src_offset_x);
+    const int dst_offset = mad24(dst_offset_y, dst_step, dst_offset_x);
+#ifdef BORDER_CONSTANT
+    for(int i = localRow; i < LOCAL_HEIGHT; i += get_local_size(1))
+    {
+        int curRow = groupStartRow + i;
+        for(int j = localCol; j < LOCAL_WIDTH; j += get_local_size(0))
+        {
+            int curCol = groupStartCol + j;
+            if(curRow < FILTER_RADIUS - src_offset_y || (curRow - FILTER_RADIUS) >= wholerows - src_offset_y||
+                curCol < FILTER_RADIUS - src_offset_x || (curCol - FILTER_RADIUS) >= wholecols - src_offset_x)
+            {
+                local_data[(i) * LOCAL_WIDTH + j] = 0;
+            }
+            else
+            {
+                local_data[(i) * LOCAL_WIDTH + j] = src[(curRow - FILTER_RADIUS) * src_step + curCol - FILTER_RADIUS + src_offset];
+            }
+        }
+    }
+#else
+    for(int i = localRow; i < LOCAL_HEIGHT; i += get_local_size(1))
+    {
+        int curRow = groupStartRow + i;
+
+        curRow = ADDR_H(curRow, FILTER_RADIUS - src_offset_y, wholerows - src_offset_y);
+
+        curRow = ADDR_B(curRow - FILTER_RADIUS, wholerows - src_offset_y, curRow - FILTER_RADIUS);
+
+        for(int j = localCol; j < LOCAL_WIDTH; j += get_local_size(0))
+        {
+            int curCol = groupStartCol + j;
+            curCol = ADDR_L(curCol, FILTER_RADIUS - src_offset_x, wholecols - src_offset_x);
+            curCol = ADDR_R(curCol - FILTER_RADIUS, wholecols - src_offset_x, curCol - FILTER_RADIUS);
+            if(curRow < wholerows  && curCol < wholecols)
+            {
+                local_data[(i) * LOCAL_WIDTH + j] = src[(curRow) * src_step + curCol + src_offset];
+            }
+        }
+    }
+#endif
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(globalRow < rows && globalCol < cols)
+    {
+        T_SUM sum = (T_SUM)SUM_ZERO;
+        int filterIdx = 0;
+        for(int i = 0; i < FILTER_SIZE; i++)
+        {
+            int offset = (i + localRow) * LOCAL_WIDTH;
+
+            for(int j = 0; j < FILTER_SIZE; j++)
+            {
+                sum += CONVERT_TYPE_SUM(local_data[offset + j + localCol]) * mat_kernel[filterIdx++];
+            }
+        }
+        dst[(globalRow)*dst_step + (globalCol) + dst_offset] = CONVERT_TYPE(sum);
+    }
+}
+
+/// following is specific for 3x3 kernels
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////Macro for define elements number per thread/////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////
+#define ANX                     1
+#define ANY                     1
+
+#define ROWS_PER_GROUP          4
+#define ROWS_PER_GROUP_BITS     2
+#define ROWS_FETCH              (ROWS_PER_GROUP + ANY + ANY)   //(ROWS_PER_GROUP + anY * 2)
+
+#define THREADS_PER_ROW         64
+#define THREADS_PER_ROW_BIT     6
+
+#define ELEMENTS_PER_THREAD     4
+#define ELEMENTS_PER_THREAD_BIT 2
+
+#define LOCAL_MEM_STEP          260 //divup((get_local_size(0) + anX * 2), 4) * 4
+
+///////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////////8uC1////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void filter2D_3x3(
+    __global T_IMG *src,
+    __global T_IMG *dst,
+    int src_step,
+    int dst_step,
+    __constant float *mat_kernel,
+    __local T_IMG *local_data,
+    int wholerows,
+    int wholecols,
+    int src_offset_x,
+    int src_offset_y,
+    int dst_offset_x,
+    int dst_offset_y,
+    int cols,
+    int rows,
+    int operate_cols
+)
+{
+    int gX = get_global_id(0);
+    int gY = get_global_id(1);
+
+    int lX = get_local_id(0);
+
+    int groupX_size = get_local_size(0);
+    int groupX_id   = get_group_id(0);
+
+#define dst_align (dst_offset_x & 3)
+    int cols_start_index_group = src_offset_x - dst_align + groupX_size * groupX_id - ANX;
+    int rows_start_index       = src_offset_y + (gY << ROWS_PER_GROUP_BITS) - ANY;
+
+    if((gY << 2) < rows)
+    {
+        for(int i = 0; i < ROWS_FETCH; ++i)
+        {
+            if((rows_start_index - src_offset_y) + i < rows + ANY)
+            {
+#ifdef BORDER_CONSTANT
+                int selected_row  = rows_start_index + i;
+                int selected_cols = cols_start_index_group + lX;
+
+                T_IMG data = src[mad24(selected_row, src_step, selected_cols)];
+                int con = selected_row >= 0 && selected_row < wholerows && selected_cols >= 0 && selected_cols < wholecols;
+                data = con ? data : 0;
+                local_data[mad24(i, LOCAL_MEM_STEP, lX)] = data;
+
+                if(lX < (ANX << 1))
+                {
+                    selected_cols = cols_start_index_group + lX + groupX_size;
+
+                    data  = src[mad24(selected_row, src_step, selected_cols)];
+                    con = selected_row >= 0 && selected_row < wholerows && selected_cols >= 0 && selected_cols < wholecols;
+                    data = con ? data : 0;
+                    local_data[mad24(i, LOCAL_MEM_STEP, lX) + groupX_size] = data;
+                }
+#else
+                int selected_row = ADDR_H(rows_start_index + i,  0, wholerows);
+                selected_row     = ADDR_B(rows_start_index + i, wholerows, selected_row);
+
+                int selected_cols = ADDR_L(cols_start_index_group + lX, 0, wholecols);
+                selected_cols     = ADDR_R(cols_start_index_group + lX, wholecols, selected_cols);
+
+                T_IMG data = src[mad24(selected_row, src_step, selected_cols)];
+
+                local_data[mad24(i, LOCAL_MEM_STEP, lX)] = data;
+
+                if(lX < (ANX << 1))
+                {
+                    selected_cols = cols_start_index_group + lX + groupX_size;
+                    selected_cols = ADDR_R(selected_cols, wholecols, selected_cols);
+
+                    data = src[mad24(selected_row, src_step, selected_cols)];
+                    local_data[mad24(i, LOCAL_MEM_STEP, lX) + groupX_size] = data;
+                }
+#endif
+            }
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    int process_col = groupX_size * groupX_id + ((lX % THREADS_PER_ROW) << 2);
+    if(((gY << 2) < rows) && (process_col < operate_cols))
+    {
+        int dst_cols_start = dst_offset_x;
+        int dst_cols_end   = dst_offset_x + cols;
+        int dst_cols_index = (dst_offset_x + process_col) & 0xfffffffc;
+
+        int dst_rows_end   = dst_offset_y + rows;
+        int dst_rows_index = dst_offset_y + (gY << ROWS_PER_GROUP_BITS) + (lX >> THREADS_PER_ROW_BIT);
+        dst = dst + mad24(dst_rows_index, dst_step, dst_cols_index);
+
+        T_IMGx4 dst_data = *(__global T_IMGx4 *)dst;
+
+        T_SUMx4 sum = (T_SUMx4)SUM_ZEROx4;
+        T_IMGx4 data;
+
+        for(int i = 0; i < FILTER_SIZE; i++)
+        {
+#pragma unroll
+            for(int j = 0; j < FILTER_SIZE; j++)
+            {
+                if(dst_rows_index < dst_rows_end)
+                {
+                    int local_row = (lX >> THREADS_PER_ROW_BIT) + i;
+                    int local_cols = ((lX % THREADS_PER_ROW) << ELEMENTS_PER_THREAD_BIT) + j;
+
+                    data = VLOAD4(0, (__local T_IMG_C1 *)(local_data + local_row * LOCAL_MEM_STEP + local_cols));
+                    sum = sum + (mat_kernel[i * FILTER_SIZE + j] * CONVERT_TYPE_SUMx4(data));
+                }
+            }
+        }
+        if(dst_rows_index < dst_rows_end)
+        {
+            T_IMGx4 tmp_dst = CONVERT_TYPEx4(sum);
+            tmp_dst.SX = ((dst_cols_index + 0 >= dst_cols_start) && (dst_cols_index + 0 < dst_cols_end)) ?
+                         tmp_dst.SX : dst_data.SX;
+            tmp_dst.SY = ((dst_cols_index + 1 >= dst_cols_start) && (dst_cols_index + 1 < dst_cols_end)) ?
+                         tmp_dst.SY : dst_data.SY;
+            tmp_dst.SZ = ((dst_cols_index + 2 >= dst_cols_start) && (dst_cols_index + 2 < dst_cols_end)) ?
+                         tmp_dst.SZ : dst_data.SZ;
+            tmp_dst.SW = ((dst_cols_index + 3 >= dst_cols_start) && (dst_cols_index + 3 < dst_cols_end)) ?
+                         tmp_dst.SW : dst_data.SW;
+            *(__global T_IMGx4 *)dst = tmp_dst;
+        }
+    }
+}
diff --git a/modules/ocl/src/kernels/filtering_morph.cl b/modules/ocl/src/opencl/filtering_morph.cl
similarity index 98%
rename from modules/ocl/src/kernels/filtering_morph.cl
rename to modules/ocl/src/opencl/filtering_morph.cl
index f60d76a8b7..e659a59f51 100644
--- a/modules/ocl/src/kernels/filtering_morph.cl
+++ b/modules/ocl/src/opencl/filtering_morph.cl
@@ -120,8 +120,7 @@ __kernel void morph_C1_D0(__global const uchar * restrict src,
     int gidy = get_global_id(1);
     int out_addr = mad24(gidy,dst_step_in_pixel,gidx+dst_offset_in_pixel);
 
-#ifdef USEROI
-    if(gidx+3<cols && gidy<rows && (dst_offset_in_pixel&3==0))
+    if(gidx+3<cols && gidy<rows && ((dst_offset_in_pixel&3)==0))
     {
         *(__global uchar4*)&dst[out_addr] = res;
     }
@@ -150,9 +149,6 @@ __kernel void morph_C1_D0(__global const uchar * restrict src,
             dst[out_addr] = res.x;
         }
     }
-#else
-    *(__global uchar4*)&dst[out_addr] = res;
-#endif
 }
 #else
 __kernel void morph(__global const GENTYPE * restrict src,
diff --git a/modules/ocl/src/opencl/haarobjectdetect.cl b/modules/ocl/src/opencl/haarobjectdetect.cl
new file mode 100644
index 0000000000..22a7fe7cbf
--- /dev/null
+++ b/modules/ocl/src/opencl/haarobjectdetect.cl
@@ -0,0 +1,423 @@
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Niko Li, newlife20080214@gmail.com
+//    Wang Weiyan, wangweiyanster@gmail.com
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Nathan, liujun@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//
+
+#pragma OPENCL EXTENSION cl_amd_printf : enable
+#define CV_HAAR_FEATURE_MAX           3
+
+#define calc_sum(rect,offset)        (sum[(rect).p0+offset] - sum[(rect).p1+offset] - sum[(rect).p2+offset] + sum[(rect).p3+offset])
+#define calc_sum1(rect,offset,i)     (sum[(rect).p0[i]+offset] - sum[(rect).p1[i]+offset] - sum[(rect).p2[i]+offset] + sum[(rect).p3[i]+offset])
+
+typedef int   sumtype;
+typedef float sqsumtype;
+
+#ifndef STUMP_BASED
+#define STUMP_BASED 1
+#endif
+
+typedef struct __attribute__((aligned (128) )) GpuHidHaarTreeNode
+{
+    int p[CV_HAAR_FEATURE_MAX][4] __attribute__((aligned (64)));
+    float weight[CV_HAAR_FEATURE_MAX];
+    float threshold;
+    float alpha[3] __attribute__((aligned (16)));
+    int left __attribute__((aligned (4)));
+    int right __attribute__((aligned (4)));
+}
+GpuHidHaarTreeNode;
+
+
+typedef struct __attribute__((aligned (32))) GpuHidHaarClassifier
+{
+    int count __attribute__((aligned (4)));
+    GpuHidHaarTreeNode* node __attribute__((aligned (8)));
+    float* alpha __attribute__((aligned (8)));
+}
+GpuHidHaarClassifier;
+
+
+typedef struct __attribute__((aligned (64))) GpuHidHaarStageClassifier
+{
+    int  count __attribute__((aligned (4)));
+    float threshold __attribute__((aligned (4)));
+    int two_rects __attribute__((aligned (4)));
+    int reserved0 __attribute__((aligned (8)));
+    int reserved1 __attribute__((aligned (8)));
+    int reserved2 __attribute__((aligned (8)));
+    int reserved3 __attribute__((aligned (8)));
+}
+GpuHidHaarStageClassifier;
+
+
+typedef struct __attribute__((aligned (64))) GpuHidHaarClassifierCascade
+{
+    int  count __attribute__((aligned (4)));
+    int  is_stump_based __attribute__((aligned (4)));
+    int  has_tilted_features __attribute__((aligned (4)));
+    int  is_tree __attribute__((aligned (4)));
+    int pq0 __attribute__((aligned (4)));
+    int pq1 __attribute__((aligned (4)));
+    int pq2 __attribute__((aligned (4)));
+    int pq3 __attribute__((aligned (4)));
+    int p0 __attribute__((aligned (4)));
+    int p1 __attribute__((aligned (4)));
+    int p2 __attribute__((aligned (4)));
+    int p3 __attribute__((aligned (4)));
+    float inv_window_area __attribute__((aligned (4)));
+} GpuHidHaarClassifierCascade;
+
+__kernel void __attribute__((reqd_work_group_size(8,8,1)))gpuRunHaarClassifierCascade(
+    global GpuHidHaarStageClassifier * stagecascadeptr,
+    global int4 * info,
+    global GpuHidHaarTreeNode * nodeptr,
+    global const int * restrict sum1,
+    global const float * restrict sqsum1,
+    global int4 * candidate,
+    const int pixelstep,
+    const int loopcount,
+    const int start_stage,
+    const int split_stage,
+    const int end_stage,
+    const int startnode,
+    const int splitnode,
+    const int4 p,
+    const int4 pq,
+    const float correction)
+{
+    int grpszx = get_local_size(0);
+    int grpszy = get_local_size(1);
+    int grpnumx = get_num_groups(0);
+    int grpidx = get_group_id(0);
+    int lclidx = get_local_id(0);
+    int lclidy = get_local_id(1);
+
+    int lcl_sz = mul24(grpszx,grpszy);
+    int lcl_id = mad24(lclidy,grpszx,lclidx);
+
+    __local int lclshare[1024];
+    __local int* lcldata = lclshare;//for save win data
+    __local int* glboutindex = lcldata + 28*28;//for save global out index
+    __local int* lclcount = glboutindex + 1;//for save the numuber of temp pass pixel
+    __local int* lcloutindex = lclcount + 1;//for save info of temp pass pixel
+    __local float* partialsum = (__local float*)(lcloutindex + (lcl_sz<<1));
+    glboutindex[0]=0;
+    int outputoff = mul24(grpidx,256);
+
+    //assume window size is 20X20
+#define WINDOWSIZE 20+1
+    //make sure readwidth is the multiple of 4
+    //ystep =1, from host code
+    int readwidth = ((grpszx-1 + WINDOWSIZE+3)>>2)<<2;
+    int readheight = grpszy-1+WINDOWSIZE;
+    int read_horiz_cnt = readwidth >> 2;//each read int4
+    int total_read = mul24(read_horiz_cnt,readheight);
+    int read_loop = (total_read + lcl_sz - 1) >> 6;
+    candidate[outputoff+(lcl_id<<2)] = (int4)0;
+    candidate[outputoff+(lcl_id<<2)+1] = (int4)0;
+    candidate[outputoff+(lcl_id<<2)+2] = (int4)0;
+    candidate[outputoff+(lcl_id<<2)+3] = (int4)0;
+    for(int scalei = 0; scalei <loopcount; scalei++)
+    {
+        int4 scaleinfo1= info[scalei];
+        int width = (scaleinfo1.x & 0xffff0000) >> 16;
+        int height = scaleinfo1.x & 0xffff;
+        int grpnumperline =(scaleinfo1.y & 0xffff0000) >> 16;
+        int totalgrp = scaleinfo1.y & 0xffff;
+        int imgoff = scaleinfo1.z;
+        float factor = as_float(scaleinfo1.w);
+
+        __global const int * sum = sum1 + imgoff;
+        __global const float * sqsum = sqsum1 + imgoff;
+        for(int grploop=grpidx; grploop<totalgrp; grploop+=grpnumx)
+        {
+            int grpidy = grploop / grpnumperline;
+            int grpidx = grploop - mul24(grpidy, grpnumperline);
+            int x = mad24(grpidx,grpszx,lclidx);
+            int y = mad24(grpidy,grpszy,lclidy);
+            int grpoffx = x-lclidx;
+            int grpoffy = y-lclidy;
+
+            for(int i=0; i<read_loop; i++)
+            {
+                int pos_id = mad24(i,lcl_sz,lcl_id);
+                pos_id = pos_id < total_read ? pos_id : 0;
+
+                int lcl_y = pos_id / read_horiz_cnt;
+                int lcl_x = pos_id - mul24(lcl_y, read_horiz_cnt);
+
+                int glb_x = grpoffx + (lcl_x<<2);
+                int glb_y = grpoffy + lcl_y;
+
+                int glb_off = mad24(min(glb_y, height - 1),pixelstep,glb_x);
+                int4 data = *(__global int4*)&sum[glb_off];
+                int lcl_off = mad24(lcl_y, readwidth, lcl_x<<2);
+
+                vstore4(data, 0, &lcldata[lcl_off]);
+            }
+
+            lcloutindex[lcl_id] = 0;
+            lclcount[0] = 0;
+            int result = 1;
+            int nodecounter= startnode;
+            float mean, variance_norm_factor;
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            int lcl_off = mad24(lclidy,readwidth,lclidx);
+            int4 cascadeinfo1, cascadeinfo2;
+            cascadeinfo1 = p;
+            cascadeinfo2 = pq;
+
+            cascadeinfo1.x +=lcl_off;
+            cascadeinfo1.z +=lcl_off;
+            mean = (lcldata[mad24(cascadeinfo1.y,readwidth,cascadeinfo1.x)] - lcldata[mad24(cascadeinfo1.y,readwidth,cascadeinfo1.z)] -
+                    lcldata[mad24(cascadeinfo1.w,readwidth,cascadeinfo1.x)] + lcldata[mad24(cascadeinfo1.w,readwidth,cascadeinfo1.z)])
+                    *correction;
+
+            int p_offset = mad24(y, pixelstep, x);
+
+            cascadeinfo2.x +=p_offset;
+            cascadeinfo2.z +=p_offset;
+            variance_norm_factor =sqsum[mad24(cascadeinfo2.y, pixelstep, cascadeinfo2.x)] - sqsum[mad24(cascadeinfo2.y, pixelstep, cascadeinfo2.z)] -
+                                    sqsum[mad24(cascadeinfo2.w, pixelstep, cascadeinfo2.x)] + sqsum[mad24(cascadeinfo2.w, pixelstep, cascadeinfo2.z)];
+
+            variance_norm_factor = variance_norm_factor * correction - mean * mean;
+            variance_norm_factor = variance_norm_factor >=0.f ? sqrt(variance_norm_factor) : 1.f;
+
+            for(int stageloop = start_stage; (stageloop < split_stage)  && result; stageloop++ )
+            {
+                float stage_sum = 0.f;
+                int2 stageinfo = *(global int2*)(stagecascadeptr+stageloop);
+                float stagethreshold = as_float(stageinfo.y);
+                for(int nodeloop = 0; nodeloop < stageinfo.x; )
+                {
+                    __global GpuHidHaarTreeNode* currentnodeptr = (nodeptr + nodecounter);
+
+                    int4 info1 = *(__global int4*)(&(currentnodeptr->p[0][0]));
+                    int4 info2 = *(__global int4*)(&(currentnodeptr->p[1][0]));
+                    int4 info3 = *(__global int4*)(&(currentnodeptr->p[2][0]));
+                    float4 w = *(__global float4*)(&(currentnodeptr->weight[0]));
+                    float3 alpha3 = *(__global float3*)(&(currentnodeptr->alpha[0]));
+
+                    float nodethreshold  = w.w * variance_norm_factor;
+
+                    info1.x +=lcl_off;
+                    info1.z +=lcl_off;
+                    info2.x +=lcl_off;
+                    info2.z +=lcl_off;
+
+                    float classsum = (lcldata[mad24(info1.y,readwidth,info1.x)] - lcldata[mad24(info1.y,readwidth,info1.z)] -
+                                        lcldata[mad24(info1.w,readwidth,info1.x)] + lcldata[mad24(info1.w,readwidth,info1.z)]) * w.x;
+
+                    classsum += (lcldata[mad24(info2.y,readwidth,info2.x)] - lcldata[mad24(info2.y,readwidth,info2.z)] -
+                                    lcldata[mad24(info2.w,readwidth,info2.x)] + lcldata[mad24(info2.w,readwidth,info2.z)]) * w.y;
+
+                    info3.x +=lcl_off;
+                    info3.z +=lcl_off;
+                    classsum += (lcldata[mad24(info3.y,readwidth,info3.x)] - lcldata[mad24(info3.y,readwidth,info3.z)] -
+                                    lcldata[mad24(info3.w,readwidth,info3.x)] + lcldata[mad24(info3.w,readwidth,info3.z)]) * w.z;
+
+                    bool passThres = classsum >= nodethreshold;
+#if STUMP_BASED
+                    stage_sum += passThres ? alpha3.y : alpha3.x;
+                    nodecounter++;
+                    nodeloop++;
+#else
+                    bool isRootNode = (nodecounter & 1) == 0;
+                    if(isRootNode)
+                    {
+                        if( (passThres && currentnodeptr->right) ||
+                            (!passThres && currentnodeptr->left))
+                        {
+                            nodecounter ++;
+                        }
+                        else
+                        {
+                            stage_sum += alpha3.x;
+                            nodecounter += 2;
+                            nodeloop ++;
+                        }
+                    }
+                    else
+                    {
+                        stage_sum += passThres ? alpha3.z : alpha3.y;
+                        nodecounter ++;
+                        nodeloop ++;
+                    }
+#endif
+                }
+
+                result = (stage_sum >= stagethreshold);
+            }
+
+            if(result && (x < width) && (y < height))
+            {
+                int queueindex = atomic_inc(lclcount);
+                lcloutindex[queueindex<<1] = (lclidy << 16) | lclidx;
+                lcloutindex[(queueindex<<1)+1] = as_int(variance_norm_factor);
+            }
+            barrier(CLK_LOCAL_MEM_FENCE);
+            int queuecount  = lclcount[0];
+            barrier(CLK_LOCAL_MEM_FENCE);
+            nodecounter = splitnode;
+            for(int stageloop = split_stage; stageloop< end_stage && queuecount>0; stageloop++)
+            {
+                lclcount[0]=0;
+                barrier(CLK_LOCAL_MEM_FENCE);
+
+                int2 stageinfo = *(global int2*)(stagecascadeptr+stageloop);
+                float stagethreshold = as_float(stageinfo.y);
+
+                int perfscale = queuecount > 4 ? 3 : 2;
+                int queuecount_loop = (queuecount + (1<<perfscale)-1) >> perfscale;
+                int lcl_compute_win = lcl_sz >> perfscale;
+                int lcl_compute_win_id = (lcl_id >>(6-perfscale));
+                int lcl_loops = (stageinfo.x + lcl_compute_win -1) >> (6-perfscale);
+                int lcl_compute_id = lcl_id - (lcl_compute_win_id << (6-perfscale));
+                for(int queueloop=0; queueloop<queuecount_loop; queueloop++)
+                {
+                    float stage_sum = 0.f;
+                    int temp_coord = lcloutindex[lcl_compute_win_id<<1];
+                    float variance_norm_factor = as_float(lcloutindex[(lcl_compute_win_id<<1)+1]);
+                    int queue_pixel = mad24(((temp_coord  & (int)0xffff0000)>>16),readwidth,temp_coord & 0xffff);
+
+                    if(lcl_compute_win_id < queuecount)
+                    {
+                        int tempnodecounter = lcl_compute_id;
+                        float part_sum = 0.f;
+                        const int stump_factor = STUMP_BASED ? 1 : 2;
+                        int root_offset = 0;
+                        for(int lcl_loop=0; lcl_loop<lcl_loops && tempnodecounter<stageinfo.x;)
+                        {
+                            __global GpuHidHaarTreeNode* currentnodeptr =
+                                nodeptr + (nodecounter + tempnodecounter) * stump_factor + root_offset;
+
+                            int4 info1 = *(__global int4*)(&(currentnodeptr->p[0][0]));
+                            int4 info2 = *(__global int4*)(&(currentnodeptr->p[1][0]));
+                            int4 info3 = *(__global int4*)(&(currentnodeptr->p[2][0]));
+                            float4 w = *(__global float4*)(&(currentnodeptr->weight[0]));
+                            float3 alpha3 = *(__global float3*)(&(currentnodeptr->alpha[0]));
+                            float nodethreshold  = w.w * variance_norm_factor;
+
+                            info1.x +=queue_pixel;
+                            info1.z +=queue_pixel;
+                            info2.x +=queue_pixel;
+                            info2.z +=queue_pixel;
+
+                            float classsum = (lcldata[mad24(info1.y,readwidth,info1.x)] - lcldata[mad24(info1.y,readwidth,info1.z)] -
+                                                lcldata[mad24(info1.w,readwidth,info1.x)] + lcldata[mad24(info1.w,readwidth,info1.z)]) * w.x;
+
+
+                            classsum += (lcldata[mad24(info2.y,readwidth,info2.x)] - lcldata[mad24(info2.y,readwidth,info2.z)] -
+                                            lcldata[mad24(info2.w,readwidth,info2.x)] + lcldata[mad24(info2.w,readwidth,info2.z)]) * w.y;
+
+                            info3.x +=queue_pixel;
+                            info3.z +=queue_pixel;
+                            classsum += (lcldata[mad24(info3.y,readwidth,info3.x)] - lcldata[mad24(info3.y,readwidth,info3.z)] -
+                                            lcldata[mad24(info3.w,readwidth,info3.x)] + lcldata[mad24(info3.w,readwidth,info3.z)]) * w.z;
+
+                            bool passThres = classsum >= nodethreshold;
+#if STUMP_BASED
+                            part_sum += passThres ? alpha3.y : alpha3.x;
+                            tempnodecounter += lcl_compute_win;
+                            lcl_loop++;
+#else
+                            if(root_offset == 0)
+                            {
+                                if( (passThres && currentnodeptr->right) ||
+                                    (!passThres && currentnodeptr->left))
+                                {
+                                    root_offset = 1;
+                                }
+                                else
+                                {
+                                    part_sum += alpha3.x;
+                                    tempnodecounter += lcl_compute_win;
+                                    lcl_loop++;
+                                }
+                            }
+                            else
+                            {
+                                part_sum += passThres ? alpha3.z : alpha3.y;
+                                tempnodecounter += lcl_compute_win;
+                                lcl_loop++;
+                                root_offset = 0;
+                            }
+#endif
+                        }//end for(int lcl_loop=0;lcl_loop<lcl_loops;lcl_loop++)
+                        partialsum[lcl_id]=part_sum;
+                    }
+                    barrier(CLK_LOCAL_MEM_FENCE);
+                    if(lcl_compute_win_id < queuecount)
+                    {
+                        for(int i=0; i<lcl_compute_win && (lcl_compute_id==0); i++)
+                        {
+                            stage_sum += partialsum[lcl_id+i];
+                        }
+                        if(stage_sum >= stagethreshold && (lcl_compute_id==0))
+                        {
+                            int queueindex = atomic_inc(lclcount);
+                            lcloutindex[queueindex<<1] = temp_coord;
+                            lcloutindex[(queueindex<<1)+1] = as_int(variance_norm_factor);
+                        }
+                        lcl_compute_win_id +=(1<<perfscale);
+                    }
+                    barrier(CLK_LOCAL_MEM_FENCE);
+                }//end for(int queueloop=0;queueloop<queuecount_loop;queueloop++)
+
+                queuecount = lclcount[0];
+                barrier(CLK_LOCAL_MEM_FENCE);
+                nodecounter += stageinfo.x;
+            }//end for(int stageloop = splitstage; stageloop< endstage && queuecount>0;stageloop++)
+
+            if(lcl_id<queuecount)
+            {
+                int temp = lcloutindex[lcl_id<<1];
+                int x = mad24(grpidx,grpszx,temp & 0xffff);
+                int y = mad24(grpidy,grpszy,((temp & (int)0xffff0000) >> 16));
+                temp = glboutindex[0];
+                int4 candidate_result;
+                candidate_result.zw = (int2)convert_int_rtn(factor*20.f);
+                candidate_result.x = convert_int_rtn(x*factor);
+                candidate_result.y = convert_int_rtn(y*factor);
+                atomic_inc(glboutindex);
+                candidate[outputoff+temp+lcl_id] = candidate_result;
+            }
+            barrier(CLK_LOCAL_MEM_FENCE);
+        }//end for(int grploop=grpidx;grploop<totalgrp;grploop+=grpnumx)
+    }//end for(int scalei = 0; scalei <loopcount; scalei++)
+}
diff --git a/modules/ocl/src/opencl/haarobjectdetect_scaled2.cl b/modules/ocl/src/opencl/haarobjectdetect_scaled2.cl
new file mode 100644
index 0000000000..c12ab59506
--- /dev/null
+++ b/modules/ocl/src/opencl/haarobjectdetect_scaled2.cl
@@ -0,0 +1,306 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Wu Xinglong, wxl370@126.com
+//    Sen Liu, swjtuls1987@126.com
+//    Peng Xiao, pengxiao@outlook.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+// Enter your kernel in this window
+//#pragma OPENCL EXTENSION cl_amd_printf:enable
+#define CV_HAAR_FEATURE_MAX           3
+typedef int   sumtype;
+typedef float sqsumtype;
+
+typedef struct __attribute__((aligned(128))) GpuHidHaarTreeNode
+{
+    int p[CV_HAAR_FEATURE_MAX][4] __attribute__((aligned(64)));
+    float weight[CV_HAAR_FEATURE_MAX] /*__attribute__((aligned (16)))*/;
+    float threshold /*__attribute__((aligned (4)))*/;
+    float alpha[3] __attribute__((aligned(16)));
+    int left __attribute__((aligned(4)));
+    int right __attribute__((aligned(4)));
+}
+GpuHidHaarTreeNode;
+typedef struct __attribute__((aligned(32))) GpuHidHaarClassifier
+{
+    int count __attribute__((aligned(4)));
+    GpuHidHaarTreeNode *node __attribute__((aligned(8)));
+    float *alpha __attribute__((aligned(8)));
+}
+GpuHidHaarClassifier;
+typedef struct __attribute__((aligned(64))) GpuHidHaarStageClassifier
+{
+    int  count __attribute__((aligned(4)));
+    float threshold __attribute__((aligned(4)));
+    int two_rects __attribute__((aligned(4)));
+    int reserved0 __attribute__((aligned(8)));
+    int reserved1 __attribute__((aligned(8)));
+    int reserved2 __attribute__((aligned(8)));
+    int reserved3 __attribute__((aligned(8)));
+}
+GpuHidHaarStageClassifier;
+typedef struct __attribute__((aligned(64))) GpuHidHaarClassifierCascade
+{
+    int  count __attribute__((aligned(4)));
+    int  is_stump_based __attribute__((aligned(4)));
+    int  has_tilted_features __attribute__((aligned(4)));
+    int  is_tree __attribute__((aligned(4)));
+    int pq0 __attribute__((aligned(4)));
+    int pq1 __attribute__((aligned(4)));
+    int pq2 __attribute__((aligned(4)));
+    int pq3 __attribute__((aligned(4)));
+    int p0 __attribute__((aligned(4)));
+    int p1 __attribute__((aligned(4)));
+    int p2 __attribute__((aligned(4)));
+    int p3 __attribute__((aligned(4)));
+    float inv_window_area __attribute__((aligned(4)));
+} GpuHidHaarClassifierCascade;
+
+__kernel void gpuRunHaarClassifierCascade_scaled2(
+    global GpuHidHaarStageClassifier *stagecascadeptr,
+    global int4 *info,
+    global GpuHidHaarTreeNode *nodeptr,
+    global const int *restrict sum,
+    global const float   *restrict sqsum,
+    global int4 *candidate,
+    const int rows,
+    const int cols,
+    const int step,
+    const int loopcount,
+    const int start_stage,
+    const int split_stage,
+    const int end_stage,
+    const int startnode,
+    global int4 *p,
+    global float *correction,
+    const int nodecount)
+{
+    int grpszx = get_local_size(0);
+    int grpszy = get_local_size(1);
+    int grpnumx = get_num_groups(0);
+    int grpidx = get_group_id(0);
+    int lclidx = get_local_id(0);
+    int lclidy = get_local_id(1);
+    int lcl_sz = mul24(grpszx, grpszy);
+    int lcl_id = mad24(lclidy, grpszx, lclidx);
+    __local int glboutindex[1];
+    __local int lclcount[1];
+    __local int lcloutindex[64];
+    glboutindex[0] = 0;
+    int outputoff = mul24(grpidx, 256);
+    candidate[outputoff + (lcl_id << 2)] = (int4)0;
+    candidate[outputoff + (lcl_id << 2) + 1] = (int4)0;
+    candidate[outputoff + (lcl_id << 2) + 2] = (int4)0;
+    candidate[outputoff + (lcl_id << 2) + 3] = (int4)0;
+    int max_idx = rows * cols - 1;
+    for (int scalei = 0; scalei < loopcount; scalei++)
+    {
+        int4 scaleinfo1;
+        scaleinfo1 = info[scalei];
+        int width = (scaleinfo1.x & 0xffff0000) >> 16;
+        int height = scaleinfo1.x & 0xffff;
+        int grpnumperline = (scaleinfo1.y & 0xffff0000) >> 16;
+        int totalgrp = scaleinfo1.y & 0xffff;
+        float factor = as_float(scaleinfo1.w);
+        float correction_t = correction[scalei];
+        int ystep = (int)(max(2.0f, factor) + 0.5f);
+
+        for (int grploop = get_group_id(0); grploop < totalgrp; grploop += grpnumx)
+        {
+            int4 cascadeinfo = p[scalei];
+            int grpidy = grploop / grpnumperline;
+            int grpidx = grploop - mul24(grpidy, grpnumperline);
+            int ix = mad24(grpidx, grpszx, lclidx);
+            int iy = mad24(grpidy, grpszy, lclidy);
+            int x = ix * ystep;
+            int y = iy * ystep;
+            lcloutindex[lcl_id] = 0;
+            lclcount[0] = 0;
+            int nodecounter;
+            float mean, variance_norm_factor;
+            //if((ix < width) && (iy < height))
+            {
+                const int p_offset = mad24(y, step, x);
+                cascadeinfo.x += p_offset;
+                cascadeinfo.z += p_offset;
+                mean = (sum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.x), 0, max_idx)]
+                - sum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.z), 0, max_idx)] -
+                        sum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.x), 0, max_idx)]
+                + sum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.z), 0, max_idx)])
+                       * correction_t;
+                variance_norm_factor = sqsum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.x), 0, max_idx)]
+                - sqsum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.z), 0, max_idx)] -
+                                       sqsum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.x), 0, max_idx)]
+                + sqsum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.z), 0, max_idx)];
+                variance_norm_factor = variance_norm_factor * correction_t - mean * mean;
+                variance_norm_factor = variance_norm_factor >= 0.f ? sqrt(variance_norm_factor) : 1.f;
+                bool result = true;
+                nodecounter = startnode + nodecount * scalei;
+                for (int stageloop = start_stage; (stageloop < end_stage) && result; stageloop++)
+                {
+                    float stage_sum = 0.f;
+                    int   stagecount = stagecascadeptr[stageloop].count;
+                    for (int nodeloop = 0; nodeloop < stagecount;)
+                    {
+                        __global GpuHidHaarTreeNode *currentnodeptr = (nodeptr + nodecounter);
+                        int4 info1 = *(__global int4 *)(&(currentnodeptr->p[0][0]));
+                        int4 info2 = *(__global int4 *)(&(currentnodeptr->p[1][0]));
+                        int4 info3 = *(__global int4 *)(&(currentnodeptr->p[2][0]));
+                        float4 w = *(__global float4 *)(&(currentnodeptr->weight[0]));
+                        float3 alpha3 = *(__global float3 *)(&(currentnodeptr->alpha[0]));
+                        float nodethreshold  = w.w * variance_norm_factor;
+
+                        info1.x += p_offset;
+                        info1.z += p_offset;
+                        info2.x += p_offset;
+                        info2.z += p_offset;
+                        info3.x += p_offset;
+                        info3.z += p_offset;
+                        float classsum = (sum[clamp(mad24(info1.y, step, info1.x), 0, max_idx)]
+                        - sum[clamp(mad24(info1.y, step, info1.z), 0, max_idx)] -
+                                          sum[clamp(mad24(info1.w, step, info1.x), 0, max_idx)]
+                        + sum[clamp(mad24(info1.w, step, info1.z), 0, max_idx)]) * w.x;
+                        classsum += (sum[clamp(mad24(info2.y, step, info2.x), 0, max_idx)]
+                        - sum[clamp(mad24(info2.y, step, info2.z), 0, max_idx)] -
+                                     sum[clamp(mad24(info2.w, step, info2.x), 0, max_idx)]
+                        + sum[clamp(mad24(info2.w, step, info2.z), 0, max_idx)]) * w.y;
+                        classsum += (sum[clamp(mad24(info3.y, step, info3.x), 0, max_idx)]
+                        - sum[clamp(mad24(info3.y, step, info3.z), 0, max_idx)] -
+                                     sum[clamp(mad24(info3.w, step, info3.x), 0, max_idx)]
+                        + sum[clamp(mad24(info3.w, step, info3.z), 0, max_idx)]) * w.z;
+
+                        bool passThres = classsum >= nodethreshold;
+
+#if STUMP_BASED
+                        stage_sum += passThres ? alpha3.y : alpha3.x;
+                        nodecounter++;
+                        nodeloop++;
+#else
+                        bool isRootNode = (nodecounter & 1) == 0;
+                        if(isRootNode)
+                        {
+                            if( (passThres && currentnodeptr->right) ||
+                                (!passThres && currentnodeptr->left))
+                            {
+                                nodecounter ++;
+                            }
+                            else
+                            {
+                                stage_sum += alpha3.x;
+                                nodecounter += 2;
+                                nodeloop ++;
+                            }
+                        }
+                        else
+                        {
+                            stage_sum += (passThres ? alpha3.z : alpha3.y);
+                            nodecounter ++;
+                            nodeloop ++;
+                        }
+#endif
+                    }
+                    result = (int)(stage_sum >= stagecascadeptr[stageloop].threshold);
+                }
+
+                barrier(CLK_LOCAL_MEM_FENCE);
+
+                if (result && (ix < width) && (iy < height))
+                {
+                    int queueindex = atomic_inc(lclcount);
+                    lcloutindex[queueindex] = (y << 16) | x;
+                }
+                barrier(CLK_LOCAL_MEM_FENCE);
+                int queuecount = lclcount[0];
+
+                if (lcl_id < queuecount)
+                {
+                    int temp = lcloutindex[lcl_id];
+                    int x = temp & 0xffff;
+                    int y = (temp & (int)0xffff0000) >> 16;
+                    temp = atomic_inc(glboutindex);
+                    int4 candidate_result;
+                    candidate_result.zw = (int2)convert_int_rtn(factor * 20.f);
+                    candidate_result.x = x;
+                    candidate_result.y = y;
+                    candidate[outputoff + temp + lcl_id] = candidate_result;
+                }
+
+                barrier(CLK_LOCAL_MEM_FENCE);
+            }
+        }
+    }
+}
+__kernel void gpuscaleclassifier(global GpuHidHaarTreeNode *orinode, global GpuHidHaarTreeNode *newnode, float scale, float weight_scale, int nodenum)
+{
+    int counter = get_global_id(0);
+    int tr_x[3], tr_y[3], tr_h[3], tr_w[3], i = 0;
+    GpuHidHaarTreeNode t1 = *(orinode + counter);
+#pragma unroll
+
+    for (i = 0; i < 3; i++)
+    {
+        tr_x[i] = (int)(t1.p[i][0] * scale + 0.5f);
+        tr_y[i] = (int)(t1.p[i][1] * scale + 0.5f);
+        tr_w[i] = (int)(t1.p[i][2] * scale + 0.5f);
+        tr_h[i] = (int)(t1.p[i][3] * scale + 0.5f);
+    }
+
+    t1.weight[0] = t1.p[2][0] ? -(t1.weight[1] * tr_h[1] * tr_w[1] + t1.weight[2] * tr_h[2] * tr_w[2]) / (tr_h[0] * tr_w[0]) : -t1.weight[1] * tr_h[1] * tr_w[1] / (tr_h[0] * tr_w[0]);
+    counter += nodenum;
+#pragma unroll
+
+    for (i = 0; i < 3; i++)
+    {
+        newnode[counter].p[i][0] = tr_x[i];
+        newnode[counter].p[i][1] = tr_y[i];
+        newnode[counter].p[i][2] = tr_x[i] + tr_w[i];
+        newnode[counter].p[i][3] = tr_y[i] + tr_h[i];
+        newnode[counter].weight[i] = t1.weight[i] * weight_scale;
+    }
+
+    newnode[counter].left = t1.left;
+    newnode[counter].right = t1.right;
+    newnode[counter].threshold = t1.threshold;
+    newnode[counter].alpha[0] = t1.alpha[0];
+    newnode[counter].alpha[1] = t1.alpha[1];
+    newnode[counter].alpha[2] = t1.alpha[2];
+}
diff --git a/modules/ocl/src/kernels/imgproc_bilateral.cl b/modules/ocl/src/opencl/imgproc_bilateral.cl
similarity index 100%
rename from modules/ocl/src/kernels/imgproc_bilateral.cl
rename to modules/ocl/src/opencl/imgproc_bilateral.cl
diff --git a/modules/ocl/src/kernels/imgproc_calcHarris.cl b/modules/ocl/src/opencl/imgproc_calcHarris.cl
similarity index 95%
rename from modules/ocl/src/kernels/imgproc_calcHarris.cl
rename to modules/ocl/src/opencl/imgproc_calcHarris.cl
index 15742d6c5e..1911a72016 100644
--- a/modules/ocl/src/kernels/imgproc_calcHarris.cl
+++ b/modules/ocl/src/opencl/imgproc_calcHarris.cl
@@ -130,28 +130,29 @@ __kernel void calcHarris(__global const float *Dx,__global const float *Dy, __gl
         data[2][i] = dy_data[i] * dy_data[i];
     }
 #else
-   for(int i=0; i < ksY+1; i++)
-   {
+    int clamped_col = min(dst_cols, col);
+    for(int i=0; i < ksY+1; i++)
+    {
         int dx_selected_row;
         int dx_selected_col;
         dx_selected_row = ADDR_H(dx_startY+i, 0, dx_whole_rows);
         dx_selected_row = ADDR_B(dx_startY+i, dx_whole_rows, dx_selected_row);
-        dx_selected_col = ADDR_L(dx_startX+col, 0, dx_whole_cols);
-        dx_selected_col = ADDR_R(dx_startX+col, dx_whole_cols, dx_selected_col);
+        dx_selected_col = ADDR_L(dx_startX+clamped_col, 0, dx_whole_cols);
+        dx_selected_col = ADDR_R(dx_startX+clamped_col, dx_whole_cols, dx_selected_col);
         dx_data[i] = Dx[dx_selected_row * (dx_step>>2) + dx_selected_col];
 
         int dy_selected_row;
         int dy_selected_col;
         dy_selected_row = ADDR_H(dy_startY+i, 0, dy_whole_rows);
         dy_selected_row = ADDR_B(dy_startY+i, dy_whole_rows, dy_selected_row);
-        dy_selected_col = ADDR_L(dy_startX+col, 0, dy_whole_cols);
-        dy_selected_col = ADDR_R(dy_startX+col, dy_whole_cols, dy_selected_col);
+        dy_selected_col = ADDR_L(dy_startX+clamped_col, 0, dy_whole_cols);
+        dy_selected_col = ADDR_R(dy_startX+clamped_col, dy_whole_cols, dy_selected_col);
         dy_data[i] = Dy[dy_selected_row * (dy_step>>2) + dy_selected_col];
 
         data[0][i] = dx_data[i] * dx_data[i];
         data[1][i] = dx_data[i] * dy_data[i];
         data[2][i] = dy_data[i] * dy_data[i];
-   }
+    }
 #endif
     float sum0 = 0.0, sum1 = 0.0, sum2 = 0.0;
     for(int i=1; i < ksY; i++)
diff --git a/modules/ocl/src/kernels/imgproc_calcMinEigenVal.cl b/modules/ocl/src/opencl/imgproc_calcMinEigenVal.cl
similarity index 95%
rename from modules/ocl/src/kernels/imgproc_calcMinEigenVal.cl
rename to modules/ocl/src/opencl/imgproc_calcMinEigenVal.cl
index 662fbb07b9..462ec77925 100644
--- a/modules/ocl/src/kernels/imgproc_calcMinEigenVal.cl
+++ b/modules/ocl/src/opencl/imgproc_calcMinEigenVal.cl
@@ -130,28 +130,30 @@ __kernel void calcMinEigenVal(__global const float *Dx,__global const float *Dy,
         data[2][i] = dy_data[i] * dy_data[i];
     }
 #else
-   for(int i=0; i < ksY+1; i++)
-   {
+    int clamped_col = min(dst_cols, col);
+
+    for(int i=0; i < ksY+1; i++)
+    {
         int dx_selected_row;
         int dx_selected_col;
         dx_selected_row = ADDR_H(dx_startY+i, 0, dx_whole_rows);
         dx_selected_row = ADDR_B(dx_startY+i, dx_whole_rows, dx_selected_row);
-        dx_selected_col = ADDR_L(dx_startX+col, 0, dx_whole_cols);
-        dx_selected_col = ADDR_R(dx_startX+col, dx_whole_cols, dx_selected_col);
+        dx_selected_col = ADDR_L(dx_startX+clamped_col, 0, dx_whole_cols);
+        dx_selected_col = ADDR_R(dx_startX+clamped_col, dx_whole_cols, dx_selected_col);
         dx_data[i] = Dx[dx_selected_row * (dx_step>>2) + dx_selected_col];
 
         int dy_selected_row;
         int dy_selected_col;
         dy_selected_row = ADDR_H(dy_startY+i, 0, dy_whole_rows);
         dy_selected_row = ADDR_B(dy_startY+i, dy_whole_rows, dy_selected_row);
-        dy_selected_col = ADDR_L(dy_startX+col, 0, dy_whole_cols);
-        dy_selected_col = ADDR_R(dy_startX+col, dy_whole_cols, dy_selected_col);
+        dy_selected_col = ADDR_L(dy_startX+clamped_col, 0, dy_whole_cols);
+        dy_selected_col = ADDR_R(dy_startX+clamped_col, dy_whole_cols, dy_selected_col);
         dy_data[i] = Dy[dy_selected_row * (dy_step>>2) + dy_selected_col];
 
         data[0][i] = dx_data[i] * dx_data[i];
         data[1][i] = dx_data[i] * dy_data[i];
         data[2][i] = dy_data[i] * dy_data[i];
-   }
+    }
 #endif
     float sum0 = 0.0, sum1 = 0.0, sum2 = 0.0;
     for(int i=1; i < ksY; i++)
diff --git a/modules/ocl/src/kernels/imgproc_canny.cl b/modules/ocl/src/opencl/imgproc_canny.cl
similarity index 67%
rename from modules/ocl/src/kernels/imgproc_canny.cl
rename to modules/ocl/src/opencl/imgproc_canny.cl
index 5ec4465238..5402759e3c 100644
--- a/modules/ocl/src/kernels/imgproc_canny.cl
+++ b/modules/ocl/src/opencl/imgproc_canny.cl
@@ -69,8 +69,10 @@ inline float calc(int x, int y)
 // dx_buf	output dx buffer
 // dy_buf	output dy buffer
 __kernel
-    void calcSobelRowPass
-    (
+void
+__attribute__((reqd_work_group_size(16,16,1)))
+calcSobelRowPass
+(
     __global const uchar * src,
     __global int * dx_buf,
     __global int * dy_buf,
@@ -82,10 +84,8 @@ __kernel
     int dx_buf_offset,
     int dy_buf_step,
     int dy_buf_offset
-    )
+)
 {
-    //src_step   /= sizeof(*src);
-    //src_offset /= sizeof(*src);
     dx_buf_step   /= sizeof(*dx_buf);
     dx_buf_offset /= sizeof(*dx_buf);
     dy_buf_step   /= sizeof(*dy_buf);
@@ -99,24 +99,23 @@ __kernel
 
     __local int smem[16][18];
 
-    smem[lidy][lidx + 1] = src[gidx + gidy * src_step + src_offset];
+    smem[lidy][lidx + 1] =
+        src[gidx + min(gidy, rows - 1) * src_step + src_offset];
     if(lidx == 0)
     {
-        smem[lidy][0]  = src[max(gidx - 1,  0)        + gidy * src_step + src_offset];
-        smem[lidy][17] = src[min(gidx + 16, cols - 1) + gidy * src_step + src_offset];
+        smem[lidy][0]  =
+            src[max(gidx - 1,  0)        + min(gidy, rows - 1) * src_step + src_offset];
+        smem[lidy][17] =
+            src[min(gidx + 16, cols - 1) + min(gidy, rows - 1) * src_step + src_offset];
     }
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    if(gidy < rows)
+    if(gidy < rows && gidx < cols)
     {
-
-        if(gidx < cols)
-        {
-            dx_buf[gidx + gidy * dx_buf_step + dx_buf_offset] =
-                -smem[lidy][lidx] + smem[lidy][lidx + 2];
-            dy_buf[gidx + gidy * dy_buf_step + dy_buf_offset] =
-                smem[lidy][lidx] + 2 * smem[lidy][lidx + 1] + smem[lidy][lidx + 2];
-        }
+        dx_buf[gidx + gidy * dx_buf_step + dx_buf_offset] =
+            -smem[lidy][lidx] + smem[lidy][lidx + 2];
+        dy_buf[gidx + gidy * dy_buf_step + dy_buf_offset] =
+            smem[lidy][lidx] + 2 * smem[lidy][lidx + 1] + smem[lidy][lidx + 2];
     }
 }
 
@@ -129,8 +128,10 @@ __kernel
 // dy			direvitive in y direction output
 // mag			magnitude direvitive of xy output
 __kernel
-    void calcMagnitude_buf
-    (
+void
+__attribute__((reqd_work_group_size(16,16,1)))
+calcMagnitude_buf
+(
     __global const int * dx_buf,
     __global const int * dy_buf,
     __global int * dx,
@@ -148,7 +149,7 @@ __kernel
     int dy_offset,
     int mag_step,
     int mag_offset
-    )
+)
 {
     dx_buf_step    /= sizeof(*dx_buf);
     dx_buf_offset  /= sizeof(*dx_buf);
@@ -170,30 +171,33 @@ __kernel
     __local int sdx[18][16];
     __local int sdy[18][16];
 
-    sdx[lidy + 1][lidx] = dx_buf[gidx + gidy * dx_buf_step + dx_buf_offset];
-    sdy[lidy + 1][lidx] = dy_buf[gidx + gidy * dy_buf_step + dy_buf_offset];
+    sdx[lidy + 1][lidx] =
+        dx_buf[gidx + min(gidy, rows - 1) * dx_buf_step + dx_buf_offset];
+    sdy[lidy + 1][lidx] =
+        dy_buf[gidx + min(gidy, rows - 1) * dy_buf_step + dy_buf_offset];
     if(lidy == 0)
     {
-        sdx[0][lidx]  = dx_buf[gidx + max(gidy - 1,  0)        * dx_buf_step + dx_buf_offset];
-        sdx[17][lidx] = dx_buf[gidx + min(gidy + 16, rows - 1) * dx_buf_step + dx_buf_offset];
+        sdx[0][lidx]  =
+            dx_buf[gidx + min(max(gidy-1,0),rows-1) * dx_buf_step + dx_buf_offset];
+        sdx[17][lidx] =
+            dx_buf[gidx + min(gidy + 16, rows - 1)  * dx_buf_step + dx_buf_offset];
 
-        sdy[0][lidx]  = dy_buf[gidx + max(gidy - 1,  0)        * dy_buf_step + dy_buf_offset];
-        sdy[17][lidx] = dy_buf[gidx + min(gidy + 16, rows - 1) * dy_buf_step + dy_buf_offset];
+        sdy[0][lidx]  =
+            dy_buf[gidx + min(max(gidy-1,0),rows-1) * dy_buf_step + dy_buf_offset];
+        sdy[17][lidx] =
+            dy_buf[gidx + min(gidy + 16, rows - 1)  * dy_buf_step + dy_buf_offset];
     }
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    if(gidx < cols)
+    if(gidx < cols && gidy < rows)
     {
-        if(gidy < rows)
-        {
-            int x =  sdx[lidy][lidx] + 2 * sdx[lidy + 1][lidx] + sdx[lidy + 2][lidx];
-            int y = -sdy[lidy][lidx] + sdy[lidy + 2][lidx];
+        int x =  sdx[lidy][lidx] + 2 * sdx[lidy + 1][lidx] + sdx[lidy + 2][lidx];
+        int y = -sdy[lidy][lidx] + sdy[lidy + 2][lidx];
 
-            dx[gidx + gidy * dx_step + dx_offset] = x;
-            dy[gidx + gidy * dy_step + dy_offset] = y;
+        dx[gidx + gidy * dx_step + dx_offset] = x;
+        dy[gidx + gidy * dy_step + dy_offset] = y;
 
-            mag[(gidx + 1) + (gidy + 1) * mag_step + mag_offset] = calc(x, y);
-        }
+        mag[(gidx + 1) + (gidy + 1) * mag_step + mag_offset] = calc(x, y);
     }
 }
 
@@ -206,8 +210,8 @@ __kernel
 // dy			direvitive in y direction output
 // mag			magnitude direvitive of xy output
 __kernel
-    void calcMagnitude
-    (
+void calcMagnitude
+(
     __global const int * dx,
     __global const int * dy,
     __global float * mag,
@@ -219,7 +223,7 @@ __kernel
     int dy_offset,
     int mag_step,
     int mag_offset
-    )
+)
 {
     dx_step    /= sizeof(*dx);
     dx_offset  /= sizeof(*dx);
@@ -235,9 +239,9 @@ __kernel
     {
         mag[(gidx + 1) + (gidy + 1) * mag_step + mag_offset] =
             calc(
-            dx[gidx + gidy * dx_step + dx_offset],
-            dy[gidx + gidy * dy_step + dy_offset]
-        );
+                dx[gidx + gidy * dx_step + dx_offset],
+                dy[gidx + gidy * dy_step + dy_offset]
+            );
     }
 }
 
@@ -262,8 +266,10 @@ __kernel
 // mag			magnitudes calculated from calcMagnitude function
 // map			output containing raw edge types
 __kernel
-    void calcMap
-    (
+void
+__attribute__((reqd_work_group_size(16,16,1)))
+calcMap
+(
     __global const int * dx,
     __global const int * dy,
     __global const float * mag,
@@ -280,7 +286,7 @@ __kernel
     int mag_offset,
     int map_step,
     int map_offset
-    )
+)
 {
     dx_step    /= sizeof(*dx);
     dx_offset  /= sizeof(*dx);
@@ -291,6 +297,9 @@ __kernel
     map_step   /= sizeof(*map);
     map_offset /= sizeof(*map);
 
+    mag += mag_offset;
+    map += map_offset;
+
     __local float smem[18][18];
 
     int gidx = get_global_id(0);
@@ -307,193 +316,13 @@ __kernel
     int ly = tid / 18;
     if(ly < 14)
     {
-        smem[ly][lx] = mag[grp_idx + lx + (grp_idy + ly) * mag_step];
+        smem[ly][lx] =
+            mag[grp_idx + lx + min(grp_idy + ly, rows - 1) * mag_step];
     }
     if(ly < 4 && grp_idy + ly + 14 <= rows && grp_idx + lx <= cols)
     {
-        smem[ly + 14][lx] = mag[grp_idx + lx + (grp_idy + ly + 14) * mag_step];
-    }
-
-    barrier(CLK_LOCAL_MEM_FENCE);
-
-    if(gidy < rows && gidx < cols)
-    {
-        int x = dx[gidx + gidy * dx_step];
-        int y = dy[gidx + gidy * dy_step];
-        const int s = (x ^ y) < 0 ? -1 : 1;
-        const float m = smem[lidy + 1][lidx + 1];
-        x = abs(x);
-        y = abs(y);
-
-        // 0 - the pixel can not belong to an edge
-        // 1 - the pixel might belong to an edge
-        // 2 - the pixel does belong to an edge
-        int edge_type = 0;
-        if(m > low_thresh)
-        {
-            const int tg22x = x * TG22;
-            const int tg67x = tg22x + (x << (1 + CANNY_SHIFT));
-            y <<= CANNY_SHIFT;
-            if(y < tg22x)
-            {
-                if(m > smem[lidy + 1][lidx] && m >= smem[lidy + 1][lidx + 2])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-            else if (y > tg67x)
-            {
-                if(m > smem[lidy][lidx + 1]&& m >= smem[lidy + 2][lidx + 1])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-            else
-            {
-                if(m > smem[lidy][lidx + 1 - s]&& m > smem[lidy + 2][lidx + 1 + s])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-        }
-        map[gidx + 1 + (gidy + 1) * map_step] = edge_type;
-    }
-}
-
-// non local memory version
-__kernel
-    void calcMap_2
-    (
-    __global const int * dx,
-    __global const int * dy,
-    __global const float * mag,
-    __global int * map,
-    int rows,
-    int cols,
-    float low_thresh,
-    float high_thresh,
-    int dx_step,
-    int dx_offset,
-    int dy_step,
-    int dy_offset,
-    int mag_step,
-    int mag_offset,
-    int map_step,
-    int map_offset
-    )
-{
-    dx_step    /= sizeof(*dx);
-    dx_offset  /= sizeof(*dx);
-    dy_step    /= sizeof(*dy);
-    dy_offset  /= sizeof(*dy);
-    mag_step   /= sizeof(*mag);
-    mag_offset /= sizeof(*mag);
-    map_step   /= sizeof(*map);
-    map_offset /= sizeof(*map);
-
-
-    int gidx = get_global_id(0);
-    int gidy = get_global_id(1);
-
-    if(gidy < rows && gidx < cols)
-    {
-        int x = dx[gidx + gidy * dx_step];
-        int y = dy[gidx + gidy * dy_step];
-        const int s = (x ^ y) < 0 ? -1 : 1;
-        const float m = mag[gidx + 1 + (gidy + 1) * mag_step];
-        x = abs(x);
-        y = abs(y);
-
-        // 0 - the pixel can not belong to an edge
-        // 1 - the pixel might belong to an edge
-        // 2 - the pixel does belong to an edge
-        int edge_type = 0;
-        if(m > low_thresh)
-        {
-            const int tg22x = x * TG22;
-            const int tg67x = tg22x + (x << (1 + CANNY_SHIFT));
-            y <<= CANNY_SHIFT;
-            if(y < tg22x)
-            {
-                if(m > mag[gidx + (gidy + 1) * mag_step] && m >= mag[gidx + 2 + (gidy + 1) * mag_step])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-            else if (y > tg67x)
-            {
-                if(m > mag[gidx + 1 + gidy* mag_step] && m >= mag[gidx + 1 + (gidy + 2) * mag_step])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-            else
-            {
-                if(m > mag[gidx + 1 - s + gidy * mag_step] && m > mag[gidx + 1 + s + (gidy + 2) * mag_step])
-                {
-                    edge_type = 1 + (int)(m > high_thresh);
-                }
-            }
-        }
-        map[gidx + 1 + (gidy + 1) * map_step] = edge_type;
-    }
-}
-
-// [256, 1, 1] threaded, local memory version
-__kernel
-    void calcMap_3
-    (
-    __global const int * dx,
-    __global const int * dy,
-    __global const float * mag,
-    __global int * map,
-    int rows,
-    int cols,
-    float low_thresh,
-    float high_thresh,
-    int dx_step,
-    int dx_offset,
-    int dy_step,
-    int dy_offset,
-    int mag_step,
-    int mag_offset,
-    int map_step,
-    int map_offset
-    )
-{
-    dx_step    /= sizeof(*dx);
-    dx_offset  /= sizeof(*dx);
-    dy_step    /= sizeof(*dy);
-    dy_offset  /= sizeof(*dy);
-    mag_step   /= sizeof(*mag);
-    mag_offset /= sizeof(*mag);
-    map_step   /= sizeof(*map);
-    map_offset /= sizeof(*map);
-
-    __local float smem[18][18];
-
-    int lidx = get_local_id(0) % 16;
-    int lidy = get_local_id(0) / 16;
-
-    int grp_pix = get_global_id(0); // identifies which pixel is processing currently in the target block
-    int grp_ind = get_global_id(1); // identifies which block of pixels is currently processing
-
-    int grp_idx = (grp_ind % (cols/16)) * 16;
-    int grp_idy = (grp_ind / (cols/16)) * 16; //(grp_ind / (cols/16)) * 16
-
-    int gidx = grp_idx + lidx;
-    int gidy = grp_idy + lidy;
-
-    int tid = get_global_id(0) % 256;
-    int lx = tid % 18;
-    int ly = tid / 18;
-    if(ly < 14)
-    {
-        smem[ly][lx] = mag[grp_idx + lx + (grp_idy + ly) * mag_step];
-    }
-    if(ly < 4 && grp_idy + ly + 14 <= rows && grp_idx + lx <= cols)
-    {
-        smem[ly + 14][lx] = mag[grp_idx + lx + (grp_idy + ly + 14) * mag_step];
+        smem[ly + 14][lx] =
+            mag[grp_idx + lx + min(grp_idy + ly + 14, rows -1) * mag_step];
     }
 
     barrier(CLK_LOCAL_MEM_FENCE);
@@ -557,20 +386,24 @@ __kernel
 // st		the potiential edge points found in this kernel call
 // counter	the number of potiential edge points
 __kernel
-    void edgesHysteresisLocal
-    (
+void
+__attribute__((reqd_work_group_size(16,16,1)))
+edgesHysteresisLocal
+(
     __global int * map,
     __global ushort2 * st,
-    volatile __global unsigned int * counter,
+    __global unsigned int * counter,
     int rows,
     int cols,
     int map_step,
     int map_offset
-    )
+)
 {
     map_step   /= sizeof(*map);
     map_offset /= sizeof(*map);
 
+    map += map_offset;
+
     __local int smem[18][18];
 
     int gidx = get_global_id(0);
@@ -587,11 +420,13 @@ __kernel
     int ly = tid / 18;
     if(ly < 14)
     {
-        smem[ly][lx] = map[grp_idx + lx + (grp_idy + ly) * map_step + map_offset];
+        smem[ly][lx] =
+            map[grp_idx + lx + min(grp_idy + ly, rows - 1) * map_step];
     }
     if(ly < 4 && grp_idy + ly + 14 <= rows && grp_idx + lx <= cols)
     {
-        smem[ly + 14][lx] = map[grp_idx + lx + (grp_idy + ly + 14) * map_step + map_offset];
+        smem[ly + 14][lx] =
+            map[grp_idx + lx + min(grp_idy + ly + 14, rows - 1) * map_step];
     }
 
     barrier(CLK_LOCAL_MEM_FENCE);
@@ -652,25 +487,30 @@ __kernel
 __constant int c_dx[8] = {-1,  0,  1, -1, 1, -1, 0, 1};
 __constant int c_dy[8] = {-1, -1, -1,  0, 0,  1, 1, 1};
 
+
 #define stack_size 512
 __kernel
-    void edgesHysteresisGlobal
-    (
+void
+__attribute__((reqd_work_group_size(128,1,1)))
+edgesHysteresisGlobal
+(
     __global int * map,
     __global ushort2 * st1,
     __global ushort2 * st2,
-    volatile __global int * counter,
+    __global int * counter,
     int rows,
     int cols,
     int count,
     int map_step,
     int map_offset
-    )
+)
 {
 
     map_step   /= sizeof(*map);
     map_offset /= sizeof(*map);
 
+    map += map_offset;
+
     int gidx = get_global_id(0);
     int gidy = get_global_id(1);
 
@@ -680,7 +520,7 @@ __kernel
     int grp_idx = get_group_id(0);
     int grp_idy = get_group_id(1);
 
-    volatile __local unsigned int s_counter;
+    __local unsigned int s_counter;
     __local unsigned int s_ind;
 
     __local ushort2 s_st[stack_size];
@@ -717,7 +557,7 @@ __kernel
             while (s_counter > 0 && s_counter <= stack_size - get_local_size(0))
             {
                 const int subTaskIdx = lidx >> 3;
-                const int portion = min(s_counter, get_local_size(0)>> 3);
+                const int portion = min(s_counter, (uint)(get_local_size(0)>> 3));
 
                 pos.x = pos.y = 0;
 
@@ -734,9 +574,9 @@ __kernel
                     pos.x += c_dx[lidx & 7];
                     pos.y += c_dy[lidx & 7];
 
-                    if (map[pos.x + map_offset + pos.y * map_step] == 1)
+                    if (map[pos.x + pos.y * map_step] == 1)
                     {
-                        map[pos.x + map_offset + pos.y * map_step] = 2;
+                        map[pos.x + pos.y * map_step] = 2;
 
                         ind = atomic_inc(&s_counter);
 
@@ -771,8 +611,8 @@ __kernel
 // map		edge type mappings
 // dst		edge output
 __kernel
-    void getEdges
-    (
+void getEdges
+(
     __global const int * map,
     __global uchar * dst,
     int rows,
@@ -781,19 +621,16 @@ __kernel
     int map_offset,
     int dst_step,
     int dst_offset
-    )
+)
 {
     map_step   /= sizeof(*map);
     map_offset /= sizeof(*map);
-    //dst_step   /= sizeof(*dst);
-    //dst_offset /= sizeof(*dst);
 
     int gidx = get_global_id(0);
     int gidy = get_global_id(1);
 
     if(gidy < rows && gidx < cols)
     {
-        //dst[gidx + gidy * dst_step] = map[gidx + 1 + (gidy + 1) * map_step] == 2 ? 255: 0;
-        dst[gidx + gidy * dst_step] = (uchar)(-(map[gidx + 1 + (gidy + 1) * map_step] / 2));
+        dst[gidx + gidy * dst_step] = (uchar)(-(map[gidx + 1 + (gidy + 1) * map_step + map_offset] >> 1));
     }
 }
diff --git a/modules/ocl/src/opencl/imgproc_clahe.cl b/modules/ocl/src/opencl/imgproc_clahe.cl
new file mode 100644
index 0000000000..49c7096927
--- /dev/null
+++ b/modules/ocl/src/opencl/imgproc_clahe.cl
@@ -0,0 +1,275 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Sen Liu, swjtuls1987@126.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
+
+int calc_lut(__local int* smem, int val, int tid)
+{
+    smem[tid] = val;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid == 0)
+    {
+        for (int i = 1; i < 256; ++i)
+        {
+            smem[i] += smem[i - 1];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    return smem[tid];
+}
+
+#ifdef CPU
+void reduce(volatile __local int* smem, int val, int tid)
+{
+    smem[tid] = val;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 128)
+    {
+        smem[tid] = val += smem[tid + 128];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 64)
+    {
+        smem[tid] = val += smem[tid + 64];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 32)
+    {
+        smem[tid] += smem[tid + 32];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 16)
+    {
+        smem[tid] += smem[tid + 16];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 8)
+    {
+        smem[tid] += smem[tid + 8];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 4)
+    {
+        smem[tid] += smem[tid + 4];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 2)
+    {
+        smem[tid] += smem[tid + 2];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 1)
+    {
+        smem[256] = smem[tid] + smem[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+}
+#else
+void reduce(__local volatile int* smem, int val, int tid)
+{
+    smem[tid] = val;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 128)
+    {
+        smem[tid] = val += smem[tid + 128];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 64)
+    {
+        smem[tid] = val += smem[tid + 64];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 32)
+    {
+        smem[tid] += smem[tid + 32];
+#if WAVE_SIZE < 32
+    } barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16) {
+#endif
+        smem[tid] += smem[tid + 16];
+#if WAVE_SIZE < 16
+    } barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8) {
+#endif
+        smem[tid] += smem[tid + 8];
+        smem[tid] += smem[tid + 4];
+        smem[tid] += smem[tid + 2];
+        smem[tid] += smem[tid + 1];
+    }
+}
+#endif
+
+__kernel void calcLut(__global __const uchar * src, __global uchar * lut,
+                      const int srcStep, const int dstStep,
+                      const int2 tileSize, const int tilesX,
+                      const int clipLimit, const float lutScale)
+{
+    __local int smem[512];
+
+    const int tx = get_group_id(0);
+    const int ty = get_group_id(1);
+    const unsigned int tid = get_local_id(1) * get_local_size(0)
+                             + get_local_id(0);
+
+    smem[tid] = 0;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for (int i = get_local_id(1); i < tileSize.y; i += get_local_size(1))
+    {
+        __global const uchar* srcPtr = src + mad24( ty * tileSize.y + i,
+                                                    srcStep, tx * tileSize.x );
+        for (int j = get_local_id(0); j < tileSize.x; j += get_local_size(0))
+        {
+            const int data = srcPtr[j];
+            atomic_inc(&smem[data]);
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    int tHistVal = smem[tid];
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (clipLimit > 0)
+    {
+        // clip histogram bar
+
+        int clipped = 0;
+        if (tHistVal > clipLimit)
+        {
+            clipped = tHistVal - clipLimit;
+            tHistVal = clipLimit;
+        }
+
+        // find number of overall clipped samples
+
+        reduce(smem, clipped, tid);
+        barrier(CLK_LOCAL_MEM_FENCE);
+#ifdef CPU
+        clipped = smem[256];
+#else
+        clipped = smem[0];
+#endif
+
+        // broadcast evaluated value
+
+        __local int totalClipped;
+
+        if (tid == 0)
+            totalClipped = clipped;
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        // redistribute clipped samples evenly
+
+        int redistBatch = totalClipped / 256;
+        tHistVal += redistBatch;
+
+        int residual = totalClipped - redistBatch * 256;
+        if (tid < residual)
+            ++tHistVal;
+    }
+
+    const int lutVal = calc_lut(smem, tHistVal, tid);
+    uint ires = (uint)convert_int_rte(lutScale * lutVal);
+    lut[(ty * tilesX + tx) * dstStep + tid] =
+        convert_uchar(clamp(ires, (uint)0, (uint)255));
+}
+
+__kernel void transform(__global __const uchar * src,
+                        __global uchar * dst,
+                        __global uchar * lut,
+                        const int srcStep, const int dstStep, const int lutStep,
+                        const int cols, const int rows,
+                        const int2 tileSize,
+                        const int tilesX, const int tilesY)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if (x >= cols || y >= rows)
+        return;
+
+    const float tyf = (convert_float(y) / tileSize.y) - 0.5f;
+    int ty1 = convert_int_rtn(tyf);
+    int ty2 = ty1 + 1;
+    const float ya = tyf - ty1;
+    ty1 = max(ty1, 0);
+    ty2 = min(ty2, tilesY - 1);
+
+    const float txf = (convert_float(x) / tileSize.x) - 0.5f;
+    int tx1 = convert_int_rtn(txf);
+    int tx2 = tx1 + 1;
+    const float xa = txf - tx1;
+    tx1 = max(tx1, 0);
+    tx2 = min(tx2, tilesX - 1);
+
+    const int srcVal = src[mad24(y, srcStep, x)];
+
+    float res = 0;
+
+    res += lut[mad24(ty1 * tilesX + tx1, lutStep, srcVal)] * ((1.0f - xa) * (1.0f - ya));
+    res += lut[mad24(ty1 * tilesX + tx2, lutStep, srcVal)] * ((xa) * (1.0f - ya));
+    res += lut[mad24(ty2 * tilesX + tx1, lutStep, srcVal)] * ((1.0f - xa) * (ya));
+    res += lut[mad24(ty2 * tilesX + tx2, lutStep, srcVal)] * ((xa) * (ya));
+
+    uint ires = (uint)convert_int_rte(res);
+    dst[mad24(y, dstStep, x)] = convert_uchar(clamp(ires, (uint)0, (uint)255));
+}
diff --git a/modules/ocl/src/kernels/imgproc_columnsum.cl b/modules/ocl/src/opencl/imgproc_columnsum.cl
similarity index 100%
rename from modules/ocl/src/kernels/imgproc_columnsum.cl
rename to modules/ocl/src/opencl/imgproc_columnsum.cl
diff --git a/modules/ocl/src/kernels/imgproc_convolve.cl b/modules/ocl/src/opencl/imgproc_convolve.cl
similarity index 99%
rename from modules/ocl/src/kernels/imgproc_convolve.cl
rename to modules/ocl/src/opencl/imgproc_convolve.cl
index d113eb8169..76e7cfc55b 100644
--- a/modules/ocl/src/kernels/imgproc_convolve.cl
+++ b/modules/ocl/src/opencl/imgproc_convolve.cl
@@ -107,5 +107,3 @@ __kernel void convolve_D5 (__global float *src, __global float *temp1, __global
         dst[gy*(dst_step >> 2)+gx] = res;
    }
 }
-
-
diff --git a/modules/ocl/src/kernels/imgproc_copymakeboder.cl b/modules/ocl/src/opencl/imgproc_copymakeboder.cl
similarity index 100%
rename from modules/ocl/src/kernels/imgproc_copymakeboder.cl
rename to modules/ocl/src/opencl/imgproc_copymakeboder.cl
diff --git a/modules/ocl/src/opencl/imgproc_gftt.cl b/modules/ocl/src/opencl/imgproc_gftt.cl
new file mode 100644
index 0000000000..bfa8ee3f8f
--- /dev/null
+++ b/modules/ocl/src/opencl/imgproc_gftt.cl
@@ -0,0 +1,275 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef WITH_MASK
+#define WITH_MASK 0
+#endif
+
+__constant sampler_t sampler = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_NEAREST;
+
+inline float ELEM_INT2(image2d_t _eig, int _x, int _y)
+{
+    return read_imagef(_eig, sampler, (int2)(_x, _y)).x;
+}
+
+inline float ELEM_FLT2(image2d_t _eig, float2 pt)
+{
+    return read_imagef(_eig, sampler, pt).x;
+}
+
+__kernel
+    void findCorners
+    (
+        image2d_t eig,
+        __global const char * mask,
+        __global float2 * corners,
+        const int mask_strip,// in pixels
+        const float threshold,
+        const int rows,
+        const int cols,
+        const int max_count,
+        __global int * g_counter
+    )
+{
+    const int j = get_global_id(0);
+    const int i = get_global_id(1);
+
+    if (i > 0 && i < rows - 1 && j > 0 && j < cols - 1
+#if WITH_MASK
+        && mask[i * mask_strip + j] != 0
+#endif
+        )
+    {
+        const float val = ELEM_INT2(eig, j, i);
+
+        if (val > threshold)
+        {
+            float maxVal = val;
+
+            maxVal = fmax(ELEM_INT2(eig, j - 1, i - 1), maxVal);
+            maxVal = fmax(ELEM_INT2(eig, j    , i - 1), maxVal);
+            maxVal = fmax(ELEM_INT2(eig, j + 1, i - 1), maxVal);
+
+            maxVal = fmax(ELEM_INT2(eig, j - 1, i), maxVal);
+            maxVal = fmax(ELEM_INT2(eig, j + 1, i), maxVal);
+
+            maxVal = fmax(ELEM_INT2(eig, j - 1, i + 1), maxVal);
+            maxVal = fmax(ELEM_INT2(eig, j    , i + 1), maxVal);
+            maxVal = fmax(ELEM_INT2(eig, j + 1, i + 1), maxVal);
+
+            if (val == maxVal)
+            {
+                const int ind = atomic_inc(g_counter);
+
+                if (ind < max_count)
+                    corners[ind] = (float2)(j, i);
+            }
+        }
+    }
+}
+
+//bitonic sort
+__kernel
+    void sortCorners_bitonicSort
+    (
+        image2d_t eig,
+        __global float2 * corners,
+        const int count,
+        const int stage,
+        const int passOfStage
+    )
+{
+    const int threadId = get_global_id(0);
+    if(threadId >= count / 2)
+    {
+        return;
+    }
+
+    const int sortOrder = (((threadId/(1 << stage)) % 2)) == 1 ? 1 : 0; // 0 is descent
+
+    const int pairDistance = 1 << (stage - passOfStage);
+    const int blockWidth   = 2 * pairDistance;
+
+    const int leftId = min( (threadId % pairDistance)
+                   + (threadId / pairDistance) * blockWidth, count );
+
+    const int rightId = min( leftId + pairDistance, count );
+
+    const float2 leftPt  = corners[leftId];
+    const float2 rightPt = corners[rightId];
+
+    const float leftVal  = ELEM_FLT2(eig, leftPt);
+    const float rightVal = ELEM_FLT2(eig, rightPt);
+
+    const bool compareResult = leftVal > rightVal;
+
+    float2 greater = compareResult ? leftPt:rightPt;
+    float2 lesser  = compareResult ? rightPt:leftPt;
+
+    corners[leftId]  = sortOrder ? lesser : greater;
+    corners[rightId] = sortOrder ? greater : lesser;
+}
+
+//selection sort for gfft
+//kernel is ported from Bolt library:
+//https://github.com/HSA-Libraries/Bolt/blob/master/include/bolt/cl/sort_kernels.cl
+//  Local sort will firstly sort elements of each workgroup using selection sort
+//  its performance is O(n)
+__kernel
+    void sortCorners_selectionSortLocal
+    (
+        image2d_t eig,
+        __global float2 * corners,
+        const int count,
+        __local float2 * scratch
+    )
+{
+    int          i  = get_local_id(0); // index in workgroup
+    int numOfGroups = get_num_groups(0); // index in workgroup
+    int groupID     = get_group_id(0);
+    int         wg  = get_local_size(0); // workgroup size = block size
+    int n; // number of elements to be processed for this work group
+
+    int offset   = groupID * wg;
+    int same     = 0;
+    corners      += offset;
+    n = (groupID == (numOfGroups-1))? (count - wg*(numOfGroups-1)) : wg;
+    float2 pt1, pt2;
+
+    pt1 = corners[min(i, n)];
+    scratch[i] = pt1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(i >= n)
+    {
+        return;
+    }
+
+    float val1 = ELEM_FLT2(eig, pt1);
+    float val2;
+
+    int pos = 0;
+    for (int j=0;j<n;++j)
+    {
+        pt2  = scratch[j];
+        val2 = ELEM_FLT2(eig, pt2);
+        if(val2 > val1)
+            pos++;//calculate the rank of this element in this work group
+        else
+        {
+            if(val1 > val2)
+                continue;
+            else
+            {
+                // val1 and val2 are same
+                same++;
+            }
+        }
+    }
+    for (int j=0; j< same; j++)
+        corners[pos + j] = pt1;
+}
+__kernel
+    void sortCorners_selectionSortFinal
+    (
+        image2d_t eig,
+        __global float2 * corners,
+        const int count
+    )
+{
+    const int          i  = get_local_id(0); // index in workgroup
+    const int numOfGroups = get_num_groups(0); // index in workgroup
+    const int groupID     = get_group_id(0);
+    const int         wg  = get_local_size(0); // workgroup size = block size
+    int pos = 0, same = 0;
+    const int offset = get_group_id(0) * wg;
+    const int remainder = count - wg*(numOfGroups-1);
+
+    if((offset + i ) >= count)
+        return;
+    float2 pt1, pt2;
+    pt1 = corners[groupID*wg + i];
+
+    float val1 = ELEM_FLT2(eig, pt1);
+    float val2;
+
+    for(int j=0; j<numOfGroups-1; j++ )
+    {
+        for(int k=0; k<wg; k++)
+        {
+            pt2  = corners[j*wg + k];
+            val2 = ELEM_FLT2(eig, pt2);
+            if(val1 > val2)
+                break;
+            else
+            {
+                //Increment only if the value is not the same.
+                if( val2 > val1 )
+                    pos++;
+                else
+                    same++;
+            }
+        }
+    }
+
+    for(int k=0; k<remainder; k++)
+    {
+        pt2  = corners[(numOfGroups-1)*wg + k];
+        val2 = ELEM_FLT2(eig, pt2);
+        if(val1 > val2)
+            break;
+        else
+        {
+            //Don't increment if the value is the same.
+            //Two elements are same if (*userComp)(jData, iData)  and (*userComp)(iData, jData) are both false
+            if(val2 > val1)
+                pos++;
+            else
+                same++;
+        }
+    }
+    for (int j=0; j< same; j++)
+        corners[pos + j] = pt1;
+}
diff --git a/modules/ocl/src/kernels/imgproc_histogram.cl b/modules/ocl/src/opencl/imgproc_histogram.cl
similarity index 99%
rename from modules/ocl/src/kernels/imgproc_histogram.cl
rename to modules/ocl/src/opencl/imgproc_histogram.cl
index 01e333fbc1..6bfa095f30 100644
--- a/modules/ocl/src/kernels/imgproc_histogram.cl
+++ b/modules/ocl/src/opencl/imgproc_histogram.cl
@@ -267,4 +267,3 @@ __kernel __attribute__((reqd_work_group_size(256,1,1)))void equalizeHist(
     }
 }
 */
-
diff --git a/modules/ocl/src/kernels/imgproc_hough.cl b/modules/ocl/src/opencl/imgproc_hough.cl
similarity index 99%
rename from modules/ocl/src/kernels/imgproc_hough.cl
rename to modules/ocl/src/opencl/imgproc_hough.cl
index 06655dea31..fd1c5b9a8a 100644
--- a/modules/ocl/src/kernels/imgproc_hough.cl
+++ b/modules/ocl/src/opencl/imgproc_hough.cl
@@ -66,7 +66,7 @@ __kernel void buildPointList(__global const uchar* src,
     if (get_local_id(0) == 0)
         s_qsize[get_local_id(1)] = 0;
     barrier(CLK_LOCAL_MEM_FENCE);
-        
+
     if (y < rows)
     {
         // fill the queue
@@ -102,7 +102,7 @@ __kernel void buildPointList(__global const uchar* src,
     }
 
     barrier(CLK_GLOBAL_MEM_FENCE);
-    
+
     // copy local queues to global queue
     const int qsize = s_qsize[get_local_id(1)];
     int gidx = s_globStart[get_local_id(1)] + get_local_id(0);
@@ -131,7 +131,7 @@ __kernel void circlesAccumCenters(__global const unsigned int* list,
     const int dxStepInPixel    = dxStep    / sizeof(int);
     const int dyStepInPixel    = dyStep    / sizeof(int);
     const int accumStepInPixel = accumStep / sizeof(int);
-    
+
     const int SHIFT = 10;
     const int ONE = 1 << SHIFT;
 
@@ -195,7 +195,7 @@ __kernel void buildCentersList(__global const int* accum,
                                __global int* counter)
 {
     const int accumStepInPixel = accumStep/sizeof(int);
-    
+
     const int x = get_global_id(0);
     const int y = get_global_id(1);
 
@@ -206,7 +206,7 @@ __kernel void buildCentersList(__global const int* accum,
         const int left   = accum[mad24(y + 1, accumStepInPixel, x)];
         const int cur    = accum[mad24(y + 1, accumStepInPixel, x + 1)];
         const int right  = accum[mad24(y + 1, accumStepInPixel, x + 2)];
-        
+
         const int bottom = accum[mad24(y + 2, accumStepInPixel, x + 1)];;
 
         if (cur > threshold && cur > top && cur >= bottom && cur >  left && cur >= right)
@@ -268,7 +268,7 @@ __kernel void circlesAccumRadius(__global const unsigned int* centers,
         const int curVotes = smem[i + 1];
 
         if (curVotes >= threshold && curVotes > smem[i] && curVotes >= smem[i + 2])
-            
+
         {
             const int ind = atomic_add(counter, 1);
             if (ind < maxCircles)
diff --git a/modules/ocl/src/kernels/imgproc_integral.cl b/modules/ocl/src/opencl/imgproc_integral.cl
similarity index 51%
rename from modules/ocl/src/kernels/imgproc_integral.cl
rename to modules/ocl/src/opencl/imgproc_integral.cl
index 80f460b86e..9ced01d020 100644
--- a/modules/ocl/src/kernels/imgproc_integral.cl
+++ b/modules/ocl/src/opencl/imgproc_integral.cl
@@ -44,7 +44,11 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 #endif
 #define LSIZE 256
 #define LSIZE_1 255
@@ -56,7 +60,7 @@
 #define GET_CONFLICT_OFFSET(lid) ((lid) >> LOG_NUM_BANKS)
 
 
-kernel void integral_cols(__global uchar4 *src,__global int *sum ,__global float *sqsum,
+kernel void integral_cols_D4(__global uchar4 *src,__global int *sum ,__global float *sqsum,
                           int src_offset,int pre_invalid,int rows,int cols,int src_step,int dst_step)
 {
     unsigned int lid = get_local_id(0);
@@ -71,13 +75,13 @@ kernel void integral_cols(__global uchar4 *src,__global int *sum ,__global float
     gid = gid << 1;
     for(int i = 0; i < rows; i =i + LSIZE_1)
     {
-        src_t[0] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + gid]) : 0);
-        src_t[1] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + gid + 1]) : 0);
+        src_t[0] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + min(gid, (uint)cols - 1)]) : 0);
+        src_t[1] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + min(gid + 1, (uint)cols - 1)]) : 0);
 
         sum_t[0] = (i == 0 ? 0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sqsum_t[0] = (i == 0 ? 0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[0] = (i == 0 ? (float4)0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
         sum_t[1] =  (i == 0 ? 0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        sqsum_t[1] =  (i == 0 ? 0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[1] =  (i == 0 ? (float4)0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
         barrier(CLK_LOCAL_MEM_FENCE);
 
         int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
@@ -127,7 +131,8 @@ kernel void integral_cols(__global uchar4 *src,__global int *sum ,__global float
         }
         barrier(CLK_LOCAL_MEM_FENCE);
         int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
-        if(lid > 0 && (i+lid) <= rows){
+        if(lid > 0 && (i+lid) <= rows)
+        {
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
             lm_sqsum[0][bf_loc] += sqsum_t[0];
@@ -154,7 +159,7 @@ kernel void integral_cols(__global uchar4 *src,__global int *sum ,__global float
 }
 
 
-kernel void integral_rows(__global int4 *srcsum,__global float4 * srcsqsum,__global int *sum ,
+kernel void integral_rows_D4(__global int4 *srcsum,__global float4 * srcsqsum,__global int *sum ,
                           __global float *sqsum,int rows,int cols,int src_step,int sum_step,
                           int sqsum_step,int sum_offset,int sqsum_offset)
 {
@@ -169,15 +174,15 @@ kernel void integral_rows(__global int4 *srcsum,__global float4 * srcsqsum,__glo
     src_step = src_step >> 4;
     for(int i = 0; i < rows; i =i + LSIZE_1)
     {
-        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : 0;
-        sqsrc_t[0] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2] : 0;
-        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : 0;
-        sqsrc_t[1] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2 + 1] : 0;
+        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : (int4)0;
+        sqsrc_t[0] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2] : (float4)0;
+        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : (int4)0;
+        sqsrc_t[1] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2 + 1] : (float4)0;
 
         sum_t[0] =  (i == 0 ? 0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sqsum_t[0] =  (i == 0 ? 0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[0] =  (i == 0 ? (float4)0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
         sum_t[1] =  (i == 0 ? 0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        sqsum_t[1] =  (i == 0 ? 0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[1] =  (i == 0 ? (float4)0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
         barrier(CLK_LOCAL_MEM_FENCE);
 
         int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
@@ -228,14 +233,14 @@ kernel void integral_rows(__global int4 *srcsum,__global float4 * srcsqsum,__glo
         barrier(CLK_LOCAL_MEM_FENCE);
         if(gid == 0 && (i + lid) <= rows)
         {
-           sum[sum_offset + i + lid] = 0;
-           sqsum[sqsum_offset + i + lid] = 0;
+            sum[sum_offset + i + lid] = 0;
+            sqsum[sqsum_offset + i + lid] = 0;
         }
         if(i + lid == 0)
         {
             int loc0 = gid * 2 * sum_step;
             int loc1 = gid * 2 * sqsum_step;
-            for(int k = 1;k <= 8;k++)
+            for(int k = 1; k <= 8; k++)
             {
                 if(gid * 8 + k > cols) break;
                 sum[sum_offset + loc0 + k * sum_step / 4] = 0;
@@ -244,7 +249,8 @@ kernel void integral_rows(__global int4 *srcsum,__global float4 * srcsqsum,__glo
         }
         int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
         int loc_sq0 = sqsum_offset + gid * 2 * sqsum_step + sqsum_step / 4 + i + lid, loc_sq1 = loc_sq0 + sqsum_step ;
-        if(lid > 0 && (i+lid) <= rows){
+        if(lid > 0 && (i+lid) <= rows)
+        {
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
             lm_sqsum[0][bf_loc] += sqsum_t[0];
@@ -269,3 +275,219 @@ kernel void integral_rows(__global int4 *srcsum,__global float4 * srcsqsum,__glo
         barrier(CLK_LOCAL_MEM_FENCE);
     }
 }
+
+kernel void integral_cols_D5(__global uchar4 *src,__global float *sum ,__global float *sqsum,
+                          int src_offset,int pre_invalid,int rows,int cols,int src_step,int dst_step)
+{
+    unsigned int lid = get_local_id(0);
+    unsigned int gid = get_group_id(0);
+    float4 src_t[2], sum_t[2];
+    float4 sqsum_t[2];
+    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
+    __local float4 lm_sqsum[2][LSIZE + LOG_LSIZE];
+    __local float* sum_p;
+    __local float* sqsum_p;
+    src_step = src_step >> 2;
+    gid = gid << 1;
+    for(int i = 0; i < rows; i =i + LSIZE_1)
+    {
+        src_t[0] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + min(gid, (uint)cols - 1)]) : (float4)0);
+        src_t[1] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + min(gid + 1, (uint)cols - 1)]) : (float4)0);
+
+        sum_t[0] = (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[0] = (i == 0 ? (float4)0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
+        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[1] =  (i == 0 ? (float4)0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
+        lm_sum[0][bf_loc] = src_t[0];
+        lm_sqsum[0][bf_loc] = convert_float4(src_t[0] * src_t[0]);
+
+        lm_sum[1][bf_loc] = src_t[1];
+        lm_sqsum[1][bf_loc] = convert_float4(src_t[1] * src_t[1]);
+
+        int offset = 1;
+        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
+                lm_sqsum[lid >> 7][bi]  +=  lm_sqsum[lid >> 7][ai];
+            }
+            offset <<= 1;
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lid < 2)
+        {
+            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+            lm_sqsum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+        }
+        for(int d = 1;  d < LSIZE; d <<= 1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            offset >>= 1;
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
+                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
+
+                lm_sqsum[lid >> 7][bi] += lm_sqsum[lid >> 7][ai];
+                lm_sqsum[lid >> 7][ai] = lm_sqsum[lid >> 7][bi] - lm_sqsum[lid >> 7][ai];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
+        if(lid > 0 && (i+lid) <= rows)
+        {
+            lm_sum[0][bf_loc] += sum_t[0];
+            lm_sum[1][bf_loc] += sum_t[1];
+            lm_sqsum[0][bf_loc] += sqsum_t[0];
+            lm_sqsum[1][bf_loc] += sqsum_t[1];
+            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
+            sqsum_p = (__local float*)(&(lm_sqsum[0][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 4 + k >= cols + pre_invalid || gid * 4 + k < pre_invalid) continue;
+                sum[loc_s0 + k * dst_step / 4] = sum_p[k];
+                sqsum[loc_s0 + k * dst_step / 4] = sqsum_p[k];
+            }
+            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
+            sqsum_p = (__local float*)(&(lm_sqsum[1][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 4 + k + 4 >= cols + pre_invalid) break;
+                sum[loc_s1 + k * dst_step / 4] = sum_p[k];
+                sqsum[loc_s1 + k * dst_step / 4] = sqsum_p[k];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
+
+
+kernel void integral_rows_D5(__global float4 *srcsum,__global float4 * srcsqsum,__global float *sum ,
+                          __global float *sqsum,int rows,int cols,int src_step,int sum_step,
+                          int sqsum_step,int sum_offset,int sqsum_offset)
+{
+    unsigned int lid = get_local_id(0);
+    unsigned int gid = get_group_id(0);
+    float4 src_t[2], sum_t[2];
+    float4 sqsrc_t[2],sqsum_t[2];
+    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
+    __local float4 lm_sqsum[2][LSIZE + LOG_LSIZE];
+    __local float *sum_p;
+    __local float *sqsum_p;
+    src_step = src_step >> 4;
+    for(int i = 0; i < rows; i =i + LSIZE_1)
+    {
+        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : (float4)0;
+        sqsrc_t[0] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2] : (float4)0;
+        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : (float4)0;
+        sqsrc_t[1] = i + lid < rows ? srcsqsum[(lid+i) * src_step + gid * 2 + 1] : (float4)0;
+
+        sum_t[0] =  (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[0] =  (i == 0 ? (float4)0 : lm_sqsum[0][LSIZE_2 + LOG_LSIZE]);
+        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
+        sqsum_t[1] =  (i == 0 ? (float4)0 : lm_sqsum[1][LSIZE_2 + LOG_LSIZE]);
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
+        lm_sum[0][bf_loc] = src_t[0];
+        lm_sqsum[0][bf_loc] = sqsrc_t[0];
+
+        lm_sum[1][bf_loc] = src_t[1];
+        lm_sqsum[1][bf_loc] = sqsrc_t[1];
+
+        int offset = 1;
+        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
+                lm_sqsum[lid >> 7][bi]  +=  lm_sqsum[lid >> 7][ai];
+            }
+            offset <<= 1;
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lid < 2)
+        {
+            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+            lm_sqsum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+        }
+        for(int d = 1;  d < LSIZE; d <<= 1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            offset >>= 1;
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
+                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
+
+                lm_sqsum[lid >> 7][bi] += lm_sqsum[lid >> 7][ai];
+                lm_sqsum[lid >> 7][ai] = lm_sqsum[lid >> 7][bi] - lm_sqsum[lid >> 7][ai];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(gid == 0 && (i + lid) <= rows)
+        {
+            sum[sum_offset + i + lid] = 0;
+            sqsum[sqsum_offset + i + lid] = 0;
+        }
+        if(i + lid == 0)
+        {
+            int loc0 = gid * 2 * sum_step;
+            int loc1 = gid * 2 * sqsum_step;
+            for(int k = 1; k <= 8; k++)
+            {
+                if(gid * 8 + k > cols) break;
+                sum[sum_offset + loc0 + k * sum_step / 4] = 0;
+                sqsum[sqsum_offset + loc1 + k * sqsum_step / 4] = 0;
+            }
+        }
+        int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
+        int loc_sq0 = sqsum_offset + gid * 2 * sqsum_step + sqsum_step / 4 + i + lid, loc_sq1 = loc_sq0 + sqsum_step ;
+        if(lid > 0 && (i+lid) <= rows)
+        {
+            lm_sum[0][bf_loc] += sum_t[0];
+            lm_sum[1][bf_loc] += sum_t[1];
+            lm_sqsum[0][bf_loc] += sqsum_t[0];
+            lm_sqsum[1][bf_loc] += sqsum_t[1];
+            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
+            sqsum_p = (__local float*)(&(lm_sqsum[0][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 8 + k >= cols) break;
+                sum[loc_s0 + k * sum_step / 4] = sum_p[k];
+                sqsum[loc_sq0 + k * sqsum_step / 4] = sqsum_p[k];
+            }
+            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
+            sqsum_p = (__local float*)(&(lm_sqsum[1][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 8 + 4 + k >= cols) break;
+                sum[loc_s1 + k * sum_step / 4] = sum_p[k];
+                sqsum[loc_sq1 + k * sqsum_step / 4] = sqsum_p[k];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
diff --git a/modules/ocl/src/kernels/imgproc_integral_sum.cl b/modules/ocl/src/opencl/imgproc_integral_sum.cl
similarity index 54%
rename from modules/ocl/src/kernels/imgproc_integral_sum.cl
rename to modules/ocl/src/opencl/imgproc_integral_sum.cl
index b7b3f2ff0e..70f0c63df2 100644
--- a/modules/ocl/src/kernels/imgproc_integral_sum.cl
+++ b/modules/ocl/src/opencl/imgproc_integral_sum.cl
@@ -44,8 +44,13 @@
 //M*/
 
 #if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
 #endif
+#endif
+
 #define LSIZE 256
 #define LSIZE_1 255
 #define LSIZE_2 254
@@ -56,8 +61,8 @@
 #define GET_CONFLICT_OFFSET(lid) ((lid) >> LOG_NUM_BANKS)
 
 
-kernel void integral_sum_cols(__global uchar4 *src,__global int *sum ,
-                          int src_offset,int pre_invalid,int rows,int cols,int src_step,int dst_step)
+kernel void integral_sum_cols_D4(__global uchar4 *src,__global int *sum ,
+                              int src_offset,int pre_invalid,int rows,int cols,int src_step,int dst_step)
 {
     unsigned int lid = get_local_id(0);
     unsigned int gid = get_group_id(0);
@@ -114,7 +119,8 @@ kernel void integral_sum_cols(__global uchar4 *src,__global int *sum ,
             }
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid > 0 && (i+lid) <= rows){
+        if(lid > 0 && (i+lid) <= rows)
+        {
             int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
@@ -136,9 +142,9 @@ kernel void integral_sum_cols(__global uchar4 *src,__global int *sum ,
 }
 
 
-kernel void integral_sum_rows(__global int4 *srcsum,__global int *sum ,
-                          int rows,int cols,int src_step,int sum_step,
-                          int sum_offset)
+kernel void integral_sum_rows_D4(__global int4 *srcsum,__global int *sum ,
+                              int rows,int cols,int src_step,int sum_step,
+                              int sum_offset)
 {
     unsigned int lid = get_local_id(0);
     unsigned int gid = get_group_id(0);
@@ -196,19 +202,20 @@ kernel void integral_sum_rows(__global int4 *srcsum,__global int *sum ,
         barrier(CLK_LOCAL_MEM_FENCE);
         if(gid == 0 && (i + lid) <= rows)
         {
-           sum[sum_offset + i + lid] = 0;
+            sum[sum_offset + i + lid] = 0;
         }
         if(i + lid == 0)
         {
             int loc0 = gid * 2 * sum_step;
-            for(int k = 1;k <= 8;k++)
+            for(int k = 1; k <= 8; k++)
             {
                 if(gid * 8 + k > cols) break;
                 sum[sum_offset + loc0 + k * sum_step / 4] = 0;
             }
         }
 
-        if(lid > 0 && (i+lid) <= rows){
+        if(lid > 0 && (i+lid) <= rows)
+        {
             int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
@@ -228,3 +235,178 @@ kernel void integral_sum_rows(__global int4 *srcsum,__global int *sum ,
         barrier(CLK_LOCAL_MEM_FENCE);
     }
 }
+
+kernel void integral_sum_cols_D5(__global uchar4 *src,__global float *sum ,
+                              int src_offset,int pre_invalid,int rows,int cols,int src_step,int dst_step)
+{
+    unsigned int lid = get_local_id(0);
+    unsigned int gid = get_group_id(0);
+    float4 src_t[2], sum_t[2];
+    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
+    __local float* sum_p;
+    src_step = src_step >> 2;
+    gid = gid << 1;
+    for(int i = 0; i < rows; i =i + LSIZE_1)
+    {
+        src_t[0] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + gid]) : (float4)0);
+        src_t[1] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + gid + 1]) : (float4)0);
+
+        sum_t[0] =  (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
+        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
+        lm_sum[0][bf_loc] = src_t[0];
+
+        lm_sum[1][bf_loc] = src_t[1];
+
+        int offset = 1;
+        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
+            }
+            offset <<= 1;
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lid < 2)
+        {
+            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+        }
+        for(int d = 1;  d < LSIZE; d <<= 1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            offset >>= 1;
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
+                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lid > 0 && (i+lid) <= rows)
+        {
+            int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
+            lm_sum[0][bf_loc] += sum_t[0];
+            lm_sum[1][bf_loc] += sum_t[1];
+            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 4 + k >= cols + pre_invalid || gid * 4 + k < pre_invalid) continue;
+                sum[loc_s0 + k * dst_step / 4] = sum_p[k];
+            }
+            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 4 + k + 4 >= cols + pre_invalid) break;
+                sum[loc_s1 + k * dst_step / 4] = sum_p[k];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
+
+
+kernel void integral_sum_rows_D5(__global float4 *srcsum,__global float *sum ,
+                              int rows,int cols,int src_step,int sum_step,
+                              int sum_offset)
+{
+    unsigned int lid = get_local_id(0);
+    unsigned int gid = get_group_id(0);
+    float4 src_t[2], sum_t[2];
+    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
+    __local float *sum_p;
+    src_step = src_step >> 4;
+    for(int i = 0; i < rows; i =i + LSIZE_1)
+    {
+        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : (float4)0;
+        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : (float4)0;
+
+        sum_t[0] =  (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
+        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
+        lm_sum[0][bf_loc] = src_t[0];
+
+        lm_sum[1][bf_loc] = src_t[1];
+
+        int offset = 1;
+        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
+            }
+            offset <<= 1;
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lid < 2)
+        {
+            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+        }
+        for(int d = 1;  d < LSIZE; d <<= 1)
+        {
+            barrier(CLK_LOCAL_MEM_FENCE);
+            offset >>= 1;
+            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            ai += GET_CONFLICT_OFFSET(ai);
+            bi += GET_CONFLICT_OFFSET(bi);
+
+            if((lid & 127) < d)
+            {
+                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
+                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(gid == 0 && (i + lid) <= rows)
+        {
+            sum[sum_offset + i + lid] = 0;
+        }
+        if(i + lid == 0)
+        {
+            int loc0 = gid * 2 * sum_step;
+            for(int k = 1; k <= 8; k++)
+            {
+                if(gid * 8 + k > cols) break;
+                sum[sum_offset + loc0 + k * sum_step / 4] = 0;
+            }
+        }
+
+        if(lid > 0 && (i+lid) <= rows)
+        {
+            int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
+            lm_sum[0][bf_loc] += sum_t[0];
+            lm_sum[1][bf_loc] += sum_t[1];
+            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 8 + k >= cols) break;
+                sum[loc_s0 + k * sum_step / 4] = sum_p[k];
+            }
+            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
+            for(int k = 0; k < 4; k++)
+            {
+                if(gid * 8 + 4 + k >= cols) break;
+                sum[loc_s1 + k * sum_step / 4] = sum_p[k];
+            }
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
diff --git a/modules/ocl/src/kernels/imgproc_median.cl b/modules/ocl/src/opencl/imgproc_median.cl
similarity index 99%
rename from modules/ocl/src/kernels/imgproc_median.cl
rename to modules/ocl/src/opencl/imgproc_median.cl
index 2d9cd45f67..b87af96891 100644
--- a/modules/ocl/src/kernels/imgproc_median.cl
+++ b/modules/ocl/src/opencl/imgproc_median.cl
@@ -484,4 +484,3 @@ __kernel void medianFilter5_C1_D5(__global float * src, __global float * dst,  i
         dst[dstOffset + get_global_id(1)*dstStep + get_global_id(0)]=p12;
 }
 #undef op(a,b)
-
diff --git a/modules/ocl/src/opencl/imgproc_mulAndScaleSpectrums.cl b/modules/ocl/src/opencl/imgproc_mulAndScaleSpectrums.cl
new file mode 100644
index 0000000000..86d4e5d520
--- /dev/null
+++ b/modules/ocl/src/opencl/imgproc_mulAndScaleSpectrums.cl
@@ -0,0 +1,96 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the uintel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business uinterruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+typedef float2 cfloat;
+inline cfloat cmulf(cfloat a, cfloat b)
+{
+    return (cfloat)( a.x*b.x - a.y*b.y, a.x*b.y + a.y*b.x);
+}
+
+inline cfloat conjf(cfloat a)
+{
+    return (cfloat)( a.x, - a.y );
+}
+
+__kernel void
+mulAndScaleSpectrumsKernel(
+    __global const cfloat* a,
+    __global const cfloat* b,
+    float scale,
+    __global cfloat* dst,
+    uint cols,
+    uint rows,
+    uint mstep
+)
+{
+    const uint x = get_global_id(0);
+    const uint y = get_global_id(1);
+    const uint idx = mad24(y, mstep / sizeof(cfloat), x);
+    if (x < cols && y < rows)
+    {
+        cfloat v = cmulf(a[idx], b[idx]);
+        dst[idx] = (cfloat)( v.x * scale, v.y * scale );
+    }
+}
+__kernel void
+mulAndScaleSpectrumsKernel_CONJ(
+    __global const cfloat* a,
+    __global const cfloat* b,
+    float scale,
+    __global cfloat* dst,
+    uint cols,
+    uint rows,
+    uint mstep
+)
+{
+    const uint x = get_global_id(0);
+    const uint y = get_global_id(1);
+    const uint idx = mad24(y, mstep / sizeof(cfloat), x);
+    if (x < cols && y < rows)
+    {
+        cfloat v = cmulf(a[idx], conjf(b[idx]));
+        dst[idx] = (cfloat)( v.x * scale, v.y * scale );
+    }
+}
diff --git a/modules/ocl/src/kernels/imgproc_remap.cl b/modules/ocl/src/opencl/imgproc_remap.cl
similarity index 98%
rename from modules/ocl/src/kernels/imgproc_remap.cl
rename to modules/ocl/src/opencl/imgproc_remap.cl
index 4917749561..ee40e935cc 100644
--- a/modules/ocl/src/kernels/imgproc_remap.cl
+++ b/modules/ocl/src/opencl/imgproc_remap.cl
@@ -48,7 +48,7 @@
 #if defined DOUBLE_SUPPORT
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
 typedef double4 F4 ;
-#else 
+#else
 typedef float4 F4;
 #endif
 
@@ -62,7 +62,7 @@ __kernel void remapNNSConstant_C1_D0(__global unsigned char* dst, __global unsig
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         x = x << 2;
@@ -79,7 +79,7 @@ __kernel void remapNNSConstant_C1_D0(__global unsigned char* dst, __global unsig
 
         map1_data = *((__global short8 *)((__global char*)map1 + map1Start));
         int4 srcIdx = convert_int4(map1_data.odd) * src_step + convert_int4(map1_data.even) + src_offset;
-   
+
         uchar4 con = convert_uchar4(convert_int4(map1_data.even) >= (int4)(src_cols) || convert_int4(map1_data.odd) >= (int4)(src_rows) || convert_int4(map1_data.even) < (int4)(0) || convert_int4(map1_data.odd) < (int4)(0));
         uchar4 src_data = val;
 
@@ -91,12 +91,12 @@ __kernel void remapNNSConstant_C1_D0(__global unsigned char* dst, __global unsig
         src_data.s2 = *(src + srcIdx.s2);
         if (con.s3 == 0)
         src_data.s3 = *(src + srcIdx.s3);
-        
+
         uchar4 dst_data;
- 
+
         __global uchar4* d = (__global uchar4 *)(dst + dstStart);
 
-        uchar4 dVal = *d;      
+        uchar4 dVal = *d;
 
         int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
         dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal;
@@ -113,7 +113,7 @@ __kernel void remapNNFConstant_C1_D0(__global unsigned char* dst, __global unsig
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         x = x << 2;
@@ -131,9 +131,9 @@ __kernel void remapNNFConstant_C1_D0(__global unsigned char* dst, __global unsig
         map1_data = *((__global float8 *)((__global char*)map1 + map1Start));
         int8 map1_dataZ = convert_int8_sat_rte(map1_data);
         int4 srcIdx = map1_dataZ.odd * src_step + map1_dataZ.even + src_offset;
-    
+
         uchar4 src_data = val;
-        uchar4 con = convert_uchar4(map1_dataZ.even >= (int4)(src_cols) || map1_dataZ.odd >= (int4)(src_rows) || map1_dataZ.even < (int4)(0) || map1_dataZ.odd < (int4)(0)); 
+        uchar4 con = convert_uchar4(map1_dataZ.even >= (int4)(src_cols) || map1_dataZ.odd >= (int4)(src_rows) || map1_dataZ.even < (int4)(0) || map1_dataZ.odd < (int4)(0));
 
         if (con.s0 == 0)
         src_data.s0 = *(src + srcIdx.s0);
@@ -147,10 +147,10 @@ __kernel void remapNNFConstant_C1_D0(__global unsigned char* dst, __global unsig
        // dst_data = convert_uchar4(map1_dataZ.even >= (int4)(src_cols) || map1_dataZ.odd >= (int4)(src_rows)) ? (uchar4)(val) : src_data;
         __global uchar4* d = (__global uchar4 *)(dst + dstStart);
 
-        uchar4 dVal = *d;      
+        uchar4 dVal = *d;
 
         int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
-  
+
         dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal;
         *d = dst_data;
     }
@@ -162,7 +162,7 @@ __kernel void remapNNF1Constant_C1_D0(__global unsigned char* dst, __global unsi
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         x = x << 2;
@@ -183,9 +183,9 @@ __kernel void remapNNF1Constant_C1_D0(__global unsigned char* dst, __global unsi
         float8 map_data = (float8)(map1_data.s0, map2_data.s0, map1_data.s1, map2_data.s1, map1_data.s2, map2_data.s2, map1_data.s3, map2_data.s3);
         int8 map_dataZ = convert_int8_sat_rte(map_data);
         int4 srcIdx = map_dataZ.odd * src_step + map_dataZ.even + src_offset;
-     
+
         uchar4 src_data = val;
-        uchar4 con = convert_uchar4(map_dataZ.even >= (int4)(src_cols) || map_dataZ.odd >= (int4)(src_rows)|| map_dataZ.even < (int4)(0) || map_dataZ.odd < (int4)(0)); 
+        uchar4 con = convert_uchar4(map_dataZ.even >= (int4)(src_cols) || map_dataZ.odd >= (int4)(src_rows)|| map_dataZ.even < (int4)(0) || map_dataZ.odd < (int4)(0));
 
         if (con.s0 == 0)
         src_data.s0 = *(src + srcIdx.s0);
@@ -196,14 +196,14 @@ __kernel void remapNNF1Constant_C1_D0(__global unsigned char* dst, __global unsi
         if (con.s3 == 0)
         src_data.s3 = *(src + srcIdx.s3);
         uchar4 dst_data;
-    
+
     //    dst_data = convert_uchar4(map_dataZ.even >= (int4)(src_cols) || map_dataZ.odd >= (int4)(src_rows)) ? (uchar4)(val) : src_data;
         __global uchar4* d = (__global uchar4 *)(dst + dstStart);
 
-        uchar4 dVal = *d;      
+        uchar4 dVal = *d;
 
         int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
-  
+
         dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal;
         *d = dst_data;
     }
@@ -272,7 +272,7 @@ __kernel void remapNNF1Constant_C4_D0(__global unsigned char* dst, __global unsi
     int y = get_global_id(1);
 
     if(x < threadCols && y < dst_rows)
-    { 
+    {
          int dstIdx = y * dst_step + (x << 2) + dst_offset;
         int mapIdx = y * map1_step + (x << 2) + map1_offset;
         float map1_data = *((__global float *)((__global char*)map1 + mapIdx));
@@ -294,7 +294,7 @@ __kernel void remapNNSConstant_C1_D5(__global float* dst, __global float const *
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         int dstIdx = y * dst_step + (x << 2) + dst_offset;
@@ -309,7 +309,7 @@ __kernel void remapNNSConstant_C1_D5(__global float* dst, __global float const *
            src_data = *((__global float *)((__global uchar *)src + srcIdx));
         *((__global float *)((__global uchar*)dst + dstIdx)) = src_data;
 
- 
+
     }
 
 
@@ -321,7 +321,7 @@ __kernel void remapNNFConstant_C1_D5(__global float* dst, __global float const *
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         int dstIdx = y * dst_step + (x << 2) + dst_offset;
@@ -337,7 +337,7 @@ __kernel void remapNNFConstant_C1_D5(__global float* dst, __global float const *
            src_data = *((__global float *)((__global uchar *)src + srcIdx));
         *((__global float *)((__global uchar*)dst + dstIdx)) = src_data;
 
- 
+
     }
 
 }
@@ -348,7 +348,7 @@ __kernel void remapNNF1Constant_C1_D5(__global float* dst, __global float const
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
-     
+
     if(x < threadCols && y < dst_rows)
     {
         int dstIdx = y * dst_step + (x << 2) + dst_offset;
@@ -367,7 +367,7 @@ __kernel void remapNNF1Constant_C1_D5(__global float* dst, __global float const
            src_data = *((__global float *)((__global uchar *)src + srcIdx));
         *((__global float *)((__global uchar*)dst + dstIdx)) = src_data;
 
- 
+
     }
 
 }
@@ -391,9 +391,9 @@ __kernel void remapNNSConstant_C4_D5(__global float * dst, __global float const
           src_data = nval;
       else
           src_data = *((__global float4 *)((__global uchar *)src + srcIdx));
-      *((__global float4 *)((__global uchar*)dst + dstIdx)) = src_data; 
+      *((__global float4 *)((__global uchar*)dst + dstIdx)) = src_data;
+
 
-      
     }
 }
 
@@ -454,13 +454,13 @@ __kernel void remapLNFConstant_C1_D0(__global unsigned char* dst, __global unsig
     int y = get_global_id(1);
     if(x < threadCols && y < dst_rows)
     {
-      x = x << 2; 
+      x = x << 2;
       int gx = x - (dst_offset&3);
       int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
 
       uchar4 nval =convert_uchar4(nVal);
       uchar4 val = (uchar4)(nval.s0);
-  
+
 
       int dstStart = (y * dst_step + x  + dst_offset) - (dst_offset&3);
 
@@ -518,12 +518,12 @@ __kernel void remapLNFConstant_C1_D0(__global unsigned char* dst, __global unsig
           d.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx1.s2 + src_offset));
       if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
           d.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx1.s3 + src_offset));
- 
+
       uchar4 dst_data = convert_uchar4_sat_rte((convert_float4(a))* ud * vd +(convert_float4(b))* u * vd + (convert_float4(c))* ud * v + (convert_float4(d)) * u * v );
-    
+
       __global uchar4* D = (__global uchar4 *)(dst + dstStart);
 
-      uchar4 dVal = *D;      
+      uchar4 dVal = *D;
       int4 con = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
       dst_data = (convert_uchar4(con) != (uchar4)(0)) ? dst_data : dVal;
 
@@ -540,13 +540,13 @@ __kernel void remapLNF1Constant_C1_D0(__global unsigned char* dst, __global unsi
     int y = get_global_id(1);
     if(x < threadCols && y < dst_rows)
     {
-      x = x << 2; 
+      x = x << 2;
       int gx = x - (dst_offset&3);
       int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
 
       uchar4 nval =convert_uchar4(nVal);
       uchar4 val = (uchar4)(nval.s0);
-  
+
 
       int dstStart = (y * dst_step + x  + dst_offset) - (dst_offset&3);
 
@@ -607,13 +607,13 @@ __kernel void remapLNF1Constant_C1_D0(__global unsigned char* dst, __global unsi
           d.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx1.s2 + src_offset));
       if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
           d.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx1.s3 + src_offset));
- 
+
 
       uchar4 dst_data = convert_uchar4_sat_rte((convert_float4(a))* ud * vd +(convert_float4(b))* u * vd + (convert_float4(c))* ud * v + (convert_float4(d)) * u * v );
-    
+
       __global uchar4* D = (__global uchar4 *)(dst + dstStart);
 
-      uchar4 dVal = *D;      
+      uchar4 dVal = *D;
       int4 con = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
       dst_data = (convert_uchar4(con) != (uchar4)(0)) ? dst_data : dVal;
 
@@ -725,13 +725,13 @@ __kernel void remapLNFConstant_C1_D5(__global float* dst, __global float const *
     int y = get_global_id(1);
     if(x < threadCols && y < dst_rows)
     {
-      x = x << 4; 
+      x = x << 4;
       int gx = x - (dst_offset&15);
       int4 Gx = (int4)(gx, gx+4, gx+8, gx+12);
 
       float4 nval =convert_float4(nVal);
       float4 val = (float4)(nval.s0);
-  
+
       int dstStart = (y * dst_step + x  + dst_offset) - (dst_offset&15);
       int map1Start = y * map1_step + (x << 1) + map1_offset - ((dst_offset & 15) << 1);
       float8 map1_data;
@@ -787,12 +787,12 @@ __kernel void remapLNFConstant_C1_D5(__global float* dst, __global float const *
           d.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
       if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
           d.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
-    
+
       float4 dst_data = a * ud * vd + b * u * vd + c * ud * v + d * u * v ;
-    
+
       __global float4* D = (__global float4 *)((__global char*)dst + dstStart);
 
-      float4 dVal = *D;      
+      float4 dVal = *D;
       int4 con = (Gx >= 0 && Gx < (dst_cols << 2) && y >= 0 && y < dst_rows);
       dst_data = (convert_float4(con) != (float4)(0)) ? dst_data : dVal;
 
@@ -809,13 +809,13 @@ __kernel void remapLNF1Constant_C1_D5(__global float* dst, __global float const
     int y = get_global_id(1);
     if(x < threadCols && y < dst_rows)
     {
-      x = x << 4; 
+      x = x << 4;
       int gx = x - (dst_offset&15);
       int4 Gx = (int4)(gx, gx+4, gx+8, gx+12);
 
       float4 nval =convert_float4(nVal);
       float4 val = (float4)(nval.s0);
-  
+
       int dstStart = y * dst_step + x  + dst_offset - (dst_offset & 15);
       int map1Start = y * map1_step + x + map1_offset - (dst_offset & 15);
       float4 map1_data;
@@ -874,13 +874,13 @@ __kernel void remapLNF1Constant_C1_D5(__global float* dst, __global float const
           d.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
       if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
           d.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
- 
-      
+
+
       float4 dst_data = a * ud * vd + b * u * vd + c * ud * v + d * u * v ;
-    
+
       __global float4* D = (__global float4 *)((__global char*)dst + dstStart);
 
-      float4 dVal = *D;      
+      float4 dVal = *D;
       int4 con = (Gx >= 0 && Gx < (dst_cols << 2) && y >= 0 && y < dst_rows);
       dst_data = (convert_float4(con) != (float4)(0)) ? dst_data : dVal;
 
@@ -928,7 +928,7 @@ __kernel void remapLNFConstant_C4_D5(__global float * dst, __global float const
       else
       d = *((__global float4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<4) + src_offset ));
 
-      float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y)); 
+      float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y));
       *((__global float4 *)((__global uchar*)dst + dstIdx)) =  dst_data ;
 
     }
@@ -974,12 +974,9 @@ __kernel void remapLNF1Constant_C4_D5(__global float * dst, __global float const
       else
       d = *((__global float4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<4) + src_offset ));
 
-      float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y)); 
+      float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y));
       *((__global float4 *)((__global uchar*)dst + dstIdx)) =  dst_data ;
 
 
     }
 }
-
-
-
diff --git a/modules/ocl/src/kernels/imgproc_resize.cl b/modules/ocl/src/opencl/imgproc_resize.cl
similarity index 99%
rename from modules/ocl/src/kernels/imgproc_resize.cl
rename to modules/ocl/src/opencl/imgproc_resize.cl
index b6a25d3827..fd486de40a 100644
--- a/modules/ocl/src/kernels/imgproc_resize.cl
+++ b/modules/ocl/src/opencl/imgproc_resize.cl
@@ -411,4 +411,3 @@ __kernel void resizeNN_C4_D5(__global float4 * dst, __global float4 * src,
         dst[dpos] = src[spos];
 
 }
-
diff --git a/modules/ocl/src/kernels/imgproc_threshold.cl b/modules/ocl/src/opencl/imgproc_threshold.cl
similarity index 98%
rename from modules/ocl/src/kernels/imgproc_threshold.cl
rename to modules/ocl/src/opencl/imgproc_threshold.cl
index e046b49a75..9162abb7ef 100644
--- a/modules/ocl/src/kernels/imgproc_threshold.cl
+++ b/modules/ocl/src/opencl/imgproc_threshold.cl
@@ -143,11 +143,10 @@ __kernel void threshold_C1_D5(__global const float * restrict src, __global floa
         int4 dpos = (int4)(dstart, dstart+1, dstart+2, dstart+3);
         float4 dVal = *(__global float4*)(dst+dst_offset+gy*dst_step+dstart);
         int4 con = dpos >= 0 && dpos < dst_cols;
-        ddata = convert_float4(con) != 0 ? ddata : dVal;
+        ddata = convert_float4(con) != (float4)(0) ? ddata : dVal;
         if(dstart < dst_cols)
         {
             *(__global float4*)(dst+dst_offset+gy*dst_step+dstart) = ddata;
         }
     }
 }
-
diff --git a/modules/ocl/src/kernels/imgproc_warpAffine.cl b/modules/ocl/src/opencl/imgproc_warpAffine.cl
similarity index 86%
rename from modules/ocl/src/kernels/imgproc_warpAffine.cl
rename to modules/ocl/src/opencl/imgproc_warpAffine.cl
index 8aee1838c4..16971e252b 100644
--- a/modules/ocl/src/kernels/imgproc_warpAffine.cl
+++ b/modules/ocl/src/opencl/imgproc_warpAffine.cl
@@ -47,8 +47,12 @@
 //warpAffine kernel
 //support data types: CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4, and three interpolation methods: NN, Linear, Cubic.
 
-#if defined DOUBLE_SUPPORT
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 typedef double F;
 typedef double4 F4;
 #define convert_F4 convert_double4
@@ -58,7 +62,6 @@ typedef float4 F4;
 #define convert_F4 convert_float4
 #endif
 
-
 #define INTER_BITS 5
 #define INTER_TAB_SIZE (1 << INTER_BITS)
 #define INTER_SCALE 1.f/INTER_TAB_SIZE
@@ -81,8 +84,8 @@ inline void interpolateCubic( float x, float* coeffs )
 /**********************************************8UC1*********************************************
 ***********************************************************************************************/
 __kernel void warpAffineNN_C1_D0(__global uchar const * restrict src, __global uchar * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                 int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                 int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -123,14 +126,14 @@ __kernel void warpAffineNN_C1_D0(__global uchar const * restrict src, __global u
         sval.s1 = scon.s1 ? src[spos.s1] : 0;
         sval.s2 = scon.s2 ? src[spos.s2] : 0;
         sval.s3 = scon.s3 ? src[spos.s3] : 0;
-        dval = convert_uchar4(dcon != 0) ? sval : dval;
+        dval = convert_uchar4(dcon) != (uchar4)(0,0,0,0) ? sval : dval;
         *d = dval;
     }
 }
 
 __kernel void warpAffineLinear_C1_D0(__global const uchar * restrict src, __global uchar * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                     int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                     int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -206,16 +209,16 @@ __kernel void warpAffineLinear_C1_D0(__global const uchar * restrict src, __glob
         taby = INTER_SCALE * convert_float4(ay);
         tabx = INTER_SCALE * convert_float4(ax);
 
-        itab0 = convert_short4_sat(( (1.0f-taby)*(1.0f-tabx) * INTER_REMAP_COEF_SCALE ));
-        itab1 = convert_short4_sat(( (1.0f-taby)*tabx * INTER_REMAP_COEF_SCALE ));
-        itab2 = convert_short4_sat(( taby*(1.0f-tabx) * INTER_REMAP_COEF_SCALE ));
-        itab3 = convert_short4_sat(( taby*tabx * INTER_REMAP_COEF_SCALE ));
+        itab0 = convert_short4_sat(( (1.0f-taby)*(1.0f-tabx) * (float4)INTER_REMAP_COEF_SCALE ));
+        itab1 = convert_short4_sat(( (1.0f-taby)*tabx * (float4)INTER_REMAP_COEF_SCALE ));
+        itab2 = convert_short4_sat(( taby*(1.0f-tabx) * (float4)INTER_REMAP_COEF_SCALE ));
+        itab3 = convert_short4_sat(( taby*tabx * (float4)INTER_REMAP_COEF_SCALE ));
 
 
         int4 val;
         uchar4 tval;
         val = convert_int4(v0) * convert_int4(itab0) + convert_int4(v1) * convert_int4(itab1)
-            + convert_int4(v2) * convert_int4(itab2) + convert_int4(v3) * convert_int4(itab3);
+              + convert_int4(v2) * convert_int4(itab2) + convert_int4(v3) * convert_int4(itab3);
         tval = convert_uchar4_sat ( (val + (1 << (INTER_REMAP_COEF_BITS-1))) >> INTER_REMAP_COEF_BITS ) ;
 
         __global uchar4 * d =(__global uchar4 *)(dst+dst_offset+dy*dstStep+dx);
@@ -228,8 +231,8 @@ __kernel void warpAffineLinear_C1_D0(__global const uchar * restrict src, __glob
 }
 
 __kernel void warpAffineCubic_C1_D0(__global uchar * src, __global uchar * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                    int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                    int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -255,10 +258,10 @@ __kernel void warpAffineCubic_C1_D0(__global uchar * src, __global uchar * dst,
 
 #pragma unroll 4
         for(i=0; i<4;  i++)
-        for(j=0; j<4;  j++)
-        {
-            v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? src[src_offset+(sy+i) * srcStep + (sx+j)] : 0;
-        }
+            for(j=0; j<4;  j++)
+            {
+                v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? src[src_offset+(sy+i) * srcStep + (sx+j)] : 0;
+            }
 
         short itab[16];
         float tab1y[4], tab1x[4];
@@ -288,7 +291,7 @@ __kernel void warpAffineCubic_C1_D0(__global uchar * src, __global uchar * dst,
                     if( itab[(k1<<2)+k2] < itab[(mk1<<2)+mk2] )
                         mk1 = k1, mk2 = k2;
                     else if( itab[(k1<<2)+k2] > itab[(Mk1<<2)+Mk2] )
-                         Mk1 = k1, Mk2 = k2;
+                        Mk1 = k1, Mk2 = k2;
                 }
             diff<0 ? (itab[(Mk1<<2)+Mk2]=(short)(itab[(Mk1<<2)+Mk2]-diff)) : (itab[(mk1<<2)+mk2]=(short)(itab[(mk1<<2)+mk2]-diff));
         }
@@ -309,8 +312,8 @@ __kernel void warpAffineCubic_C1_D0(__global uchar * src, __global uchar * dst,
 ***********************************************************************************************/
 
 __kernel void warpAffineNN_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                 int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                 int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -333,8 +336,8 @@ __kernel void warpAffineNN_C4_D0(__global uchar4 const * restrict src, __global
 }
 
 __kernel void warpAffineLinear_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                     int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                     int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -386,8 +389,8 @@ __kernel void warpAffineLinear_C4_D0(__global uchar4 const * restrict src, __glo
 }
 
 __kernel void warpAffineCubic_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                    int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                    int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -418,10 +421,10 @@ __kernel void warpAffineCubic_C4_D0(__global uchar4 const * restrict src, __glob
         int i,j;
 #pragma unroll 4
         for(i=0; i<4; i++)
-        for(j=0; j<4; j++)
-        {
-            v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? (src[src_offset+(sy+i) * srcStep + (sx+j)])  : (uchar4)0;
-        }
+            for(j=0; j<4; j++)
+            {
+                v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? (src[src_offset+(sy+i) * srcStep + (sx+j)])  : (uchar4)0;
+            }
         int itab[16];
         float tab1y[4], tab1x[4];
         float axx, ayy;
@@ -447,14 +450,14 @@ __kernel void warpAffineCubic_C4_D0(__global uchar4 const * restrict src, __glob
             int diff = isum - INTER_REMAP_COEF_SCALE;
             int Mk1=2, Mk2=2, mk1=2, mk2=2;
 
-               for( k1 = 2; k1 < 4; k1++ )
+            for( k1 = 2; k1 < 4; k1++ )
                 for( k2 = 2; k2 < 4; k2++ )
                 {
 
                     if( itab[(k1<<2)+k2] < itab[(mk1<<2)+mk2] )
                         mk1 = k1, mk2 = k2;
                     else if( itab[(k1<<2)+k2] > itab[(Mk1<<2)+Mk2] )
-                         Mk1 = k1, Mk2 = k2;
+                        Mk1 = k1, Mk2 = k2;
                 }
 
             diff<0 ? (itab[(Mk1<<2)+Mk2]=(short)(itab[(Mk1<<2)+Mk2]-diff)) : (itab[(mk1<<2)+mk2]=(short)(itab[(mk1<<2)+mk2]-diff));
@@ -477,8 +480,8 @@ __kernel void warpAffineCubic_C4_D0(__global uchar4 const * restrict src, __glob
 ***********************************************************************************************/
 
 __kernel void warpAffineNN_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                 int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                 int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -501,8 +504,8 @@ __kernel void warpAffineNN_C1_D5(__global float * src, __global float * dst, int
 }
 
 __kernel void warpAffineLinear_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                     int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                     int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -548,12 +551,12 @@ __kernel void warpAffineLinear_C1_D5(__global float * src, __global float * dst,
         sum += v0 * tab[0] +  v1 * tab[1] +  v2 * tab[2] +  v3 * tab[3];
         if(dx >= 0 && dx < dst_cols && dy >= 0 && dy < dst_rows)
             dst[(dst_offset>>2)+dy*dstStep+dx] = sum;
-   }
+    }
 }
 
 __kernel void warpAffineCubic_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                    int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                    int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -617,8 +620,8 @@ __kernel void warpAffineCubic_C1_D5(__global float * src, __global float * dst,
 ***********************************************************************************************/
 
 __kernel void warpAffineNN_C4_D5(__global float4 * src, __global float4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                 int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                 int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -636,13 +639,13 @@ __kernel void warpAffineNN_C4_D5(__global float4 * src, __global float4 * dst, i
         short sy0 = (short)(Y0 >> AB_BITS);
 
         if(dx >= 0 && dx < dst_cols && dy >= 0 && dy < dst_rows)
-            dst[(dst_offset>>4)+dy*(dstStep>>2)+dx]= (sx0>=0 && sx0<src_cols && sy0>=0 && sy0<src_rows) ? src[(src_offset>>4)+sy0*(srcStep>>2)+sx0] : 0;
+            dst[(dst_offset>>4)+dy*(dstStep>>2)+dx]= (sx0>=0 && sx0<src_cols && sy0>=0 && sy0<src_rows) ? src[(src_offset>>4)+sy0*(srcStep>>2)+sx0] : (float4)0;
     }
 }
 
 __kernel void warpAffineLinear_C4_D5(__global float4 * src, __global float4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                     int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                     int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -670,10 +673,10 @@ __kernel void warpAffineLinear_C4_D5(__global float4 * src, __global float4 * ds
 
         float4 v0, v1, v2, v3;
 
-        v0 = (sx0 >= 0 && sx0 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0] : 0;
-        v1 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0+1] : 0;
-        v2 = (sx0 >= 0 && sx0 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0] : 0;
-        v3 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0+1] : 0;
+        v0 = (sx0 >= 0 && sx0 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0] : (float4)0;
+        v1 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0+1] : (float4)0;
+        v2 = (sx0 >= 0 && sx0 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0] : (float4)0;
+        v3 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0+1] : (float4)0;
 
         float tab[4];
         float taby[2], tabx[2];
@@ -691,12 +694,12 @@ __kernel void warpAffineLinear_C4_D5(__global float4 * src, __global float4 * ds
         sum += v0 * tab[0] +  v1 * tab[1] +  v2 * tab[2] +  v3 * tab[3];
         if(dx >= 0 && dx < dst_cols && dy >= 0 && dy < dst_rows)
             dst[dst_offset+dy*dstStep+dx] = sum;
-  }
+    }
 }
 
 __kernel void warpAffineCubic_C4_D5(__global float4 * src, __global float4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                    int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                    int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -726,7 +729,7 @@ __kernel void warpAffineCubic_C4_D5(__global float4 * src, __global float4 * dst
         int i;
 
         for(i=0; i<16;  i++)
-            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : 0;
+            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : (float4)0;
 
         float tab[16];
         float tab1y[4], tab1x[4];
@@ -754,5 +757,5 @@ __kernel void warpAffineCubic_C4_D5(__global float4 * src, __global float4 * dst
             dst[dst_offset+dy*dstStep+dx] = sum;
 
         }
-   }
+    }
 }
diff --git a/modules/ocl/src/kernels/imgproc_warpPerspective.cl b/modules/ocl/src/opencl/imgproc_warpPerspective.cl
similarity index 85%
rename from modules/ocl/src/kernels/imgproc_warpPerspective.cl
rename to modules/ocl/src/opencl/imgproc_warpPerspective.cl
index 9a5ec83edd..ef9e77058c 100644
--- a/modules/ocl/src/kernels/imgproc_warpPerspective.cl
+++ b/modules/ocl/src/opencl/imgproc_warpPerspective.cl
@@ -47,8 +47,12 @@
 //wrapPerspective kernel
 //support data types: CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4, and three interpolation methods: NN, Linear, Cubic.
 
-#if defined DOUBLE_SUPPORT
+#if defined (DOUBLE_SUPPORT)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
 typedef double F;
 typedef double4 F4;
 #define convert_F4 convert_double4
@@ -81,8 +85,8 @@ inline void interpolateCubic( float x, float* coeffs )
 /**********************************************8UC1*********************************************
 ***********************************************************************************************/
 __kernel void warpPerspectiveNN_C1_D0(__global uchar const * restrict src, __global uchar * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                      int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                      int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -112,14 +116,14 @@ __kernel void warpPerspectiveNN_C1_D0(__global uchar const * restrict src, __glo
         sval.s1 = scon.s1 ? src[spos.s1] : 0;
         sval.s2 = scon.s2 ? src[spos.s2] : 0;
         sval.s3 = scon.s3 ? src[spos.s3] : 0;
-        dval = convert_uchar4(dcon != 0) ? sval : dval;
+        dval = convert_uchar4(dcon) != (uchar4)(0,0,0,0) ? sval : dval;
         *d = dval;
     }
 }
 
 __kernel void warpPerspectiveLinear_C1_D0(__global const uchar * restrict src, __global uchar * dst,
-                            int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
-                            int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
+        int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -142,7 +146,7 @@ __kernel void warpPerspectiveLinear_C1_D0(__global const uchar * restrict src, _
         int i;
 #pragma unroll 4
         for(i=0; i<4;  i++)
-           v[i] = (sx+(i&1) >= 0 && sx+(i&1) < src_cols && sy+(i>>1) >= 0 && sy+(i>>1) < src_rows) ? src[src_offset + (sy+(i>>1)) * srcStep + (sx+(i&1))] : 0;
+            v[i] = (sx+(i&1) >= 0 && sx+(i&1) < src_cols && sy+(i>>1) >= 0 && sy+(i>>1) < src_rows) ? src[src_offset + (sy+(i>>1)) * srcStep + (sx+(i&1))] : (uchar)0;
 
         short itab[4];
         float tab1y[2], tab1x[2];
@@ -170,8 +174,8 @@ __kernel void warpPerspectiveLinear_C1_D0(__global const uchar * restrict src, _
 }
 
 __kernel void warpPerspectiveCubic_C1_D0(__global uchar * src, __global uchar * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int dst_cols, int dst_rows, int srcStep, int dstStep,
+        int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -190,15 +194,15 @@ __kernel void warpPerspectiveCubic_C1_D0(__global uchar * src, __global uchar *
         short ay = (short)(Y & (INTER_TAB_SIZE-1));
         short ax = (short)(X & (INTER_TAB_SIZE-1));
 
-      uchar v[16];
+        uchar v[16];
         int i, j;
 
 #pragma unroll 4
         for(i=0; i<4;  i++)
-        for(j=0; j<4;  j++)
-        {
-            v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? src[src_offset+(sy+i) * srcStep + (sx+j)] : 0;
-        }
+            for(j=0; j<4;  j++)
+            {
+                v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? src[src_offset+(sy+i) * srcStep + (sx+j)] : (uchar)0;
+            }
 
         short itab[16];
         float tab1y[4], tab1x[4];
@@ -227,7 +231,7 @@ __kernel void warpPerspectiveCubic_C1_D0(__global uchar * src, __global uchar *
                     if( itab[(k1<<2)+k2] < itab[(mk1<<2)+mk2] )
                         mk1 = k1, mk2 = k2;
                     else if( itab[(k1<<2)+k2] > itab[(Mk1<<2)+Mk2] )
-                         Mk1 = k1, Mk2 = k2;
+                        Mk1 = k1, Mk2 = k2;
                 }
             diff<0 ? (itab[(Mk1<<2)+Mk2]=(short)(itab[(Mk1<<2)+Mk2]-diff)) : (itab[(mk1<<2)+mk2]=(short)(itab[(mk1<<2)+mk2]-diff));
         }
@@ -249,8 +253,8 @@ __kernel void warpPerspectiveCubic_C1_D0(__global uchar * src, __global uchar *
 ***********************************************************************************************/
 
 __kernel void warpPerspectiveNN_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst,
-                            int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
-                            int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                      int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
+                                      int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -273,8 +277,8 @@ __kernel void warpPerspectiveNN_C4_D0(__global uchar4 const * restrict src, __gl
 }
 
 __kernel void warpPerspectiveLinear_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst,
-                            int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
-                            int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
+        int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -299,10 +303,10 @@ __kernel void warpPerspectiveLinear_C4_D0(__global uchar4 const * restrict src,
 
         int4 v0, v1, v2, v3;
 
-        v0 = (sx >= 0 && sx < src_cols && sy >= 0 && sy < src_rows) ? convert_int4(src[src_offset+sy * srcStep + sx]) : 0;
-        v1 = (sx+1 >= 0 && sx+1 < src_cols && sy >= 0 && sy < src_rows) ? convert_int4(src[src_offset+sy * srcStep + sx+1]) : 0;
-        v2 = (sx >= 0 && sx < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? convert_int4(src[src_offset+(sy+1) * srcStep + sx]) : 0;
-        v3 = (sx+1 >= 0 && sx+1 < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? convert_int4(src[src_offset+(sy+1) * srcStep + sx+1]) : 0;
+        v0 = (sx >= 0 && sx < src_cols && sy >= 0 && sy < src_rows) ? convert_int4(src[src_offset+sy * srcStep + sx]) : (int4)0;
+        v1 = (sx+1 >= 0 && sx+1 < src_cols && sy >= 0 && sy < src_rows) ? convert_int4(src[src_offset+sy * srcStep + sx+1]) : (int4)0;
+        v2 = (sx >= 0 && sx < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? convert_int4(src[src_offset+(sy+1) * srcStep + sx]) : (int4)0;
+        v3 = (sx+1 >= 0 && sx+1 < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? convert_int4(src[src_offset+(sy+1) * srcStep + sx+1]) : (int4)0;
 
         int itab0, itab1, itab2, itab3;
         float taby, tabx;
@@ -323,8 +327,8 @@ __kernel void warpPerspectiveLinear_C4_D0(__global uchar4 const * restrict src,
 }
 
 __kernel void warpPerspectiveCubic_C4_D0(__global uchar4 const * restrict src, __global uchar4 * dst,
-                            int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
-                            int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
+        int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -352,10 +356,10 @@ __kernel void warpPerspectiveCubic_C4_D0(__global uchar4 const * restrict src, _
         int i,j;
 #pragma unroll 4
         for(i=0; i<4; i++)
-        for(j=0; j<4; j++)
-        {
-            v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? (src[src_offset+(sy+i) * srcStep + (sx+j)])  : (uchar4)0;
-        }
+            for(j=0; j<4; j++)
+            {
+                v[i*4+j] = (sx+j >= 0 && sx+j < src_cols && sy+i >= 0 && sy+i < src_rows) ? (src[src_offset+(sy+i) * srcStep + (sx+j)])  : (uchar4)0;
+            }
         int itab[16];
         float tab1y[4], tab1x[4];
         float axx, ayy;
@@ -381,14 +385,14 @@ __kernel void warpPerspectiveCubic_C4_D0(__global uchar4 const * restrict src, _
             int diff = isum - INTER_REMAP_COEF_SCALE;
             int Mk1=2, Mk2=2, mk1=2, mk2=2;
 
-               for( k1 = 2; k1 < 4; k1++ )
+            for( k1 = 2; k1 < 4; k1++ )
                 for( k2 = 2; k2 < 4; k2++ )
                 {
 
                     if( itab[(k1<<2)+k2] < itab[(mk1<<2)+mk2] )
                         mk1 = k1, mk2 = k2;
                     else if( itab[(k1<<2)+k2] > itab[(Mk1<<2)+Mk2] )
-                         Mk1 = k1, Mk2 = k2;
+                        Mk1 = k1, Mk2 = k2;
                 }
 
             diff<0 ? (itab[(Mk1<<2)+Mk2]=(short)(itab[(Mk1<<2)+Mk2]-diff)) : (itab[(mk1<<2)+mk2]=(short)(itab[(mk1<<2)+mk2]-diff));
@@ -411,8 +415,8 @@ __kernel void warpPerspectiveCubic_C4_D0(__global uchar4 const * restrict src, _
 ***********************************************************************************************/
 
 __kernel void warpPerspectiveNN_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                      int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                      int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -434,8 +438,8 @@ __kernel void warpPerspectiveNN_C1_D5(__global float * src, __global float * dst
 }
 
 __kernel void warpPerspectiveLinear_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int dst_cols, int dst_rows, int srcStep, int dstStep,
+        int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -458,10 +462,10 @@ __kernel void warpPerspectiveLinear_C1_D5(__global float * src, __global float *
 
         float v0, v1, v2, v3;
 
-        v0 = (sx >= 0 && sx < src_cols && sy >= 0 && sy < src_rows) ? src[src_offset+sy * srcStep + sx] : 0;
-        v1 = (sx+1 >= 0 && sx+1 < src_cols && sy >= 0 && sy < src_rows) ? src[src_offset+sy * srcStep + sx+1] : 0;
-        v2 = (sx >= 0 && sx < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? src[src_offset+(sy+1) * srcStep + sx] : 0;
-        v3 = (sx+1 >= 0 && sx+1 < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? src[src_offset+(sy+1) * srcStep + sx+1] : 0;
+        v0 = (sx >= 0 && sx < src_cols && sy >= 0 && sy < src_rows) ? src[src_offset+sy * srcStep + sx] : (float)0;
+        v1 = (sx+1 >= 0 && sx+1 < src_cols && sy >= 0 && sy < src_rows) ? src[src_offset+sy * srcStep + sx+1] : (float)0;
+        v2 = (sx >= 0 && sx < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? src[src_offset+(sy+1) * srcStep + sx] : (float)0;
+        v3 = (sx+1 >= 0 && sx+1 < src_cols && sy+1 >= 0 && sy+1 < src_rows) ? src[src_offset+(sy+1) * srcStep + sx+1] : (float)0;
 
         float tab[4];
         float taby[2], tabx[2];
@@ -483,8 +487,8 @@ __kernel void warpPerspectiveLinear_C1_D5(__global float * src, __global float *
 }
 
 __kernel void warpPerspectiveCubic_C1_D5(__global float * src, __global float * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int dst_cols, int dst_rows, int srcStep, int dstStep,
+        int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -510,7 +514,7 @@ __kernel void warpPerspectiveCubic_C1_D5(__global float * src, __global float *
         int i;
 
         for(i=0; i<16;  i++)
-            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : 0;
+            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : (float)0;
 
         float tab[16];
         float tab1y[4], tab1x[4];
@@ -546,8 +550,8 @@ __kernel void warpPerspectiveCubic_C1_D5(__global float * src, __global float *
 ***********************************************************************************************/
 
 __kernel void warpPerspectiveNN_C4_D5(__global float4 * src, __global float4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+                                      int dst_cols, int dst_rows, int srcStep, int dstStep,
+                                      int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -564,13 +568,13 @@ __kernel void warpPerspectiveNN_C4_D5(__global float4 * src, __global float4 * d
         short sy = (short)Y;
 
         if(dx >= 0 && dx < dst_cols && dy >= 0 && dy < dst_rows)
-            dst[(dst_offset>>4)+dy*(dstStep>>2)+dx]= (sx>=0 && sx<src_cols && sy>=0 && sy<src_rows) ? src[(src_offset>>4)+sy*(srcStep>>2)+sx] : 0;
+            dst[(dst_offset>>4)+dy*(dstStep>>2)+dx]= (sx>=0 && sx<src_cols && sy>=0 && sy<src_rows) ? src[(src_offset>>4)+sy*(srcStep>>2)+sx] : (float)0;
     }
 }
 
 __kernel void warpPerspectiveLinear_C4_D5(__global float4 * src, __global float4 * dst, int src_cols, int src_rows,
-                            int dst_cols, int dst_rows, int srcStep, int dstStep,
-                            int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int dst_cols, int dst_rows, int srcStep, int dstStep,
+        int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -597,10 +601,10 @@ __kernel void warpPerspectiveLinear_C4_D5(__global float4 * src, __global float4
 
         float4 v0, v1, v2, v3;
 
-        v0 = (sx0 >= 0 && sx0 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0] : 0;
-        v1 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0+1] : 0;
-        v2 = (sx0 >= 0 && sx0 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0] : 0;
-        v3 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0+1] : 0;
+        v0 = (sx0 >= 0 && sx0 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0] : (float4)0;
+        v1 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0 >= 0 && sy0 < src_rows) ? src[src_offset+sy0 * srcStep + sx0+1] : (float4)0;
+        v2 = (sx0 >= 0 && sx0 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0] : (float4)0;
+        v3 = (sx0+1 >= 0 && sx0+1 < src_cols && sy0+1 >= 0 && sy0+1 < src_rows) ? src[src_offset+(sy0+1) * srcStep + sx0+1] : (float4)0;
 
         float tab[4];
         float taby[2], tabx[2];
@@ -622,8 +626,8 @@ __kernel void warpPerspectiveLinear_C4_D5(__global float4 * src, __global float4
 }
 
 __kernel void warpPerspectiveCubic_C4_D5(__global float4 * src, __global float4 * dst,
-                            int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
-                            int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
+        int src_cols, int src_rows, int dst_cols, int dst_rows, int srcStep,
+        int dstStep, int src_offset, int dst_offset,  __constant F * M, int threadCols )
 {
     int dx = get_global_id(0);
     int dy = get_global_id(1);
@@ -652,7 +656,7 @@ __kernel void warpPerspectiveCubic_C4_D5(__global float4 * src, __global float4
         int i;
 
         for(i=0; i<16;  i++)
-            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : 0;
+            v[i] = (sx+(i&3) >= 0 && sx+(i&3) < src_cols && sy+(i>>2) >= 0 && sy+(i>>2) < src_rows) ? src[src_offset+(sy+(i>>2)) * srcStep + (sx+(i&3))] : (float4)0;
 
         float tab[16];
         float tab1y[4], tab1x[4];
@@ -680,6 +684,5 @@ __kernel void warpPerspectiveCubic_C4_D5(__global float4 * src, __global float4
             dst[dst_offset+dy*dstStep+dx] = sum;
 
         }
-   }
+    }
 }
-
diff --git a/modules/ocl/src/kernels/interpolate_frames.cl b/modules/ocl/src/opencl/interpolate_frames.cl
similarity index 100%
rename from modules/ocl/src/kernels/interpolate_frames.cl
rename to modules/ocl/src/opencl/interpolate_frames.cl
diff --git a/modules/ocl/src/opencl/kernel_radix_sort_by_key.cl b/modules/ocl/src/opencl/kernel_radix_sort_by_key.cl
new file mode 100644
index 0000000000..3c3eb98c80
--- /dev/null
+++ b/modules/ocl/src/opencl/kernel_radix_sort_by_key.cl
@@ -0,0 +1,176 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#pragma OPENCL EXTENSION cl_khr_byte_addressable_store : enable
+
+#ifndef N   // number of radices
+#define N 4
+#endif
+
+#ifndef K_T
+#define K_T float
+#endif
+
+#ifndef V_T
+#define V_T float
+#endif
+
+#ifndef IS_GT
+#define IS_GT 0
+#endif
+
+
+// from Thrust::b40c, link:
+// https://github.com/thrust/thrust/blob/master/thrust/system/cuda/detail/detail/b40c/radixsort_key_conversion.h
+__inline uint convertKey(uint converted_key)
+{
+#ifdef K_FLT
+    unsigned int mask = (converted_key & 0x80000000) ? 0xffffffff : 0x80000000;
+    converted_key ^= mask;
+#elif defined(K_INT)
+    const uint SIGN_MASK = 1u << ((sizeof(int) * 8) - 1);
+    converted_key ^= SIGN_MASK;
+#else
+
+#endif
+    return converted_key;
+}
+
+//FIXME(pengx17):
+// exclusive scan, need to be optimized as this is too naive...
+kernel
+    void naiveScanAddition(
+    __global int * input,
+    __global int * output,
+    int size
+    )
+{
+    if(get_global_id(0) == 0)
+    {
+        output[0] = 0;
+        for(int i = 1; i < size; i ++)
+        {
+            output[i] = output[i - 1] + input[i - 1];
+        }
+    }
+}
+
+// following is ported from
+// https://github.com/HSA-Libraries/Bolt/blob/master/include/bolt/cl/sort_uint_kernels.cl
+kernel
+    void histogramRadixN (
+    __global K_T* unsortedKeys,
+    __global int * buckets,
+    uint shiftCount
+    )
+{
+    const int RADIX_T     = N;
+    const int RADICES_T   = (1 << RADIX_T);
+    const int NUM_OF_ELEMENTS_PER_WORK_ITEM_T = RADICES_T;
+    const int MASK_T      = (1 << RADIX_T) - 1;
+    int localBuckets[16] = {0,0,0,0,0,0,0,0,
+                            0,0,0,0,0,0,0,0};
+    int globalId    = get_global_id(0);
+    int numOfGroups = get_num_groups(0);
+
+    /* Calculate thread-histograms */
+    for(int i = 0; i < NUM_OF_ELEMENTS_PER_WORK_ITEM_T; ++i)
+    {
+        uint value = convertKey(as_uint(unsortedKeys[mad24(globalId, NUM_OF_ELEMENTS_PER_WORK_ITEM_T, i)]));
+        value = (value >> shiftCount) & MASK_T;
+#if IS_GT
+        localBuckets[RADICES_T - value - 1]++;
+#else
+        localBuckets[value]++;
+#endif
+    }
+
+    for(int i = 0; i < NUM_OF_ELEMENTS_PER_WORK_ITEM_T; ++i)
+    {
+        buckets[mad24(i, RADICES_T * numOfGroups, globalId) ] = localBuckets[i];
+    }
+}
+
+kernel
+    void permuteRadixN (
+    __global K_T*  unsortedKeys,
+    __global V_T*  unsortedVals,
+    __global int* scanedBuckets,
+    uint shiftCount,
+    __global K_T*  sortedKeys,
+    __global V_T*  sortedVals
+    )
+{
+    const int RADIX_T     = N;
+    const int RADICES_T   = (1 << RADIX_T);
+    const int MASK_T = (1<<RADIX_T)  -1;
+
+    int globalId  = get_global_id(0);
+    int numOfGroups = get_num_groups(0);
+    const int NUM_OF_ELEMENTS_PER_WORK_GROUP_T = numOfGroups << N;
+    int  localIndex[16];
+
+    /*Load the index to local memory*/
+    for(int i = 0; i < RADICES_T; ++i)
+    {
+#if IS_GT
+        localIndex[i] = scanedBuckets[mad24(RADICES_T - i - 1, NUM_OF_ELEMENTS_PER_WORK_GROUP_T, globalId)];
+#else
+        localIndex[i] = scanedBuckets[mad24(i, NUM_OF_ELEMENTS_PER_WORK_GROUP_T, globalId)];
+#endif
+    }
+    /* Permute elements to appropriate location */
+    for(int i = 0; i < RADICES_T; ++i)
+    {
+        int old_idx = mad24(globalId, RADICES_T, i);
+        K_T  ovalue = unsortedKeys[old_idx];
+        uint value = convertKey(as_uint(ovalue));
+        uint maskedValue = (value >> shiftCount) & MASK_T;
+        uint index = localIndex[maskedValue];
+        sortedKeys[index] = ovalue;
+        sortedVals[index] = unsortedVals[old_idx];
+        localIndex[maskedValue] = index + 1;
+    }
+}
diff --git a/modules/ocl/src/opencl/kernel_sort_by_key.cl b/modules/ocl/src/opencl/kernel_sort_by_key.cl
new file mode 100644
index 0000000000..2e85e5a888
--- /dev/null
+++ b/modules/ocl/src/opencl/kernel_sort_by_key.cl
@@ -0,0 +1,245 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef K_T
+#define K_T float
+#endif
+
+#ifndef V_T
+#define V_T float
+#endif
+
+#ifndef IS_GT
+#define IS_GT false
+#endif
+
+#if IS_GT
+#define my_comp(x,y) ((x) > (y))
+#else
+#define my_comp(x,y) ((x) < (y))
+#endif
+
+/////////////////////// Bitonic sort ////////////////////////////
+// ported from
+// https://github.com/HSA-Libraries/Bolt/blob/master/include/bolt/cl/sort_by_key_kernels.cl
+__kernel
+    void bitonicSort
+    (
+        __global K_T * keys,
+        __global V_T * vals,
+        int count,
+        int stage,
+        int passOfStage
+    )
+{
+    const int threadId = get_global_id(0);
+    if(threadId >= count / 2)
+    {
+        return;
+    }
+    const int pairDistance = 1 << (stage - passOfStage);
+    const int blockWidth   = 2 * pairDistance;
+
+    int leftId = min( (threadId % pairDistance)
+                   + (threadId / pairDistance) * blockWidth, count );
+
+    int rightId = min( leftId + pairDistance, count );
+
+    int temp;
+
+    const V_T lval = vals[leftId];
+    const V_T rval = vals[rightId];
+
+    const K_T lkey = keys[leftId];
+    const K_T rkey = keys[rightId];
+
+    int sameDirectionBlockWidth = 1 << stage;
+
+    if((threadId/sameDirectionBlockWidth) % 2 == 1)
+    {
+        temp = rightId;
+        rightId = leftId;
+        leftId = temp;
+    }
+
+    const bool compareResult = my_comp(lkey, rkey);
+
+    if(compareResult)
+    {
+        keys[rightId] = rkey;
+        keys[leftId]  = lkey;
+        vals[rightId] = rval;
+        vals[leftId]  = lval;
+    }
+    else
+    {
+        keys[rightId] = lkey;
+        keys[leftId]  = rkey;
+        vals[rightId] = lval;
+        vals[leftId]  = rval;
+    }
+}
+
+/////////////////////// Selection sort ////////////////////////////
+//kernel is ported from Bolt library:
+//https://github.com/HSA-Libraries/Bolt/blob/master/include/bolt/cl/sort_kernels.cl
+__kernel
+    void selectionSortLocal
+    (
+        __global K_T * keys,
+        __global V_T * vals,
+        const int count,
+        __local  K_T * scratch
+    )
+{
+    int          i  = get_local_id(0); // index in workgroup
+    int numOfGroups = get_num_groups(0); // index in workgroup
+    int groupID     = get_group_id(0);
+    int         wg  = get_local_size(0); // workgroup size = block size
+    int n; // number of elements to be processed for this work group
+
+    int offset   = groupID * wg;
+    int same     = 0;
+
+    vals      += offset;
+    keys      += offset;
+    n = (groupID == (numOfGroups-1))? (count - wg*(numOfGroups-1)) : wg;
+
+    int clamped_i= min(i, n - 1);
+
+    K_T key1 = keys[clamped_i], key2;
+    V_T val1 = vals[clamped_i];
+    scratch[i] = key1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(i >= n)
+    {
+        return;
+    }
+
+    int pos = 0;
+    for (int j=0;j<n;++j)
+    {
+        key2  = scratch[j];
+        if(my_comp(key2, key1))
+            pos++;//calculate the rank of this element in this work group
+        else
+        {
+            if(my_comp(key1, key2))
+                continue;
+            else
+            {
+                // key1 and key2 are same
+                same++;
+            }
+        }
+    }
+    for (int j=0; j< same; j++)
+    {
+        vals[pos + j] = val1;
+        keys[pos + j] = key1;
+    }
+}
+__kernel
+    void selectionSortFinal
+    (
+        __global K_T * keys,
+        __global V_T * vals,
+        const int count
+    )
+{
+    const int          i  = get_local_id(0); // index in workgroup
+    const int numOfGroups = get_num_groups(0); // index in workgroup
+    const int groupID     = get_group_id(0);
+    const int         wg  = get_local_size(0); // workgroup size = block size
+    int pos = 0, same = 0;
+    const int offset = get_group_id(0) * wg;
+    const int remainder = count - wg*(numOfGroups-1);
+
+    if((offset + i ) >= count)
+        return;
+    V_T val1 = vals[offset + i];
+
+    K_T key1 = keys[offset + i];
+    K_T key2;
+
+    for(int j=0; j<numOfGroups-1; j++ )
+    {
+        for(int k=0; k<wg; k++)
+        {
+            key2 = keys[j*wg + k];
+            if(my_comp(key1, key2))
+                break;
+            else
+            {
+                //Increment only if the value is not the same.
+                if(my_comp(key2, key1))
+                    pos++;
+                else
+                    same++;
+            }
+        }
+    }
+
+    for(int k=0; k<remainder; k++)
+    {
+        key2 = keys[(numOfGroups-1)*wg + k];
+        if(my_comp(key1, key2))
+            break;
+        else
+        {
+            //Don't increment if the value is the same.
+            if(my_comp(key2, key1))
+                pos++;
+            else
+                same++;
+        }
+    }
+    for (int j=0; j< same; j++)
+    {
+        vals[pos + j] = val1;
+        keys[pos + j] = key1;
+    }
+}
diff --git a/modules/ocl/src/opencl/kernel_stablesort_by_key.cl b/modules/ocl/src/opencl/kernel_stablesort_by_key.cl
new file mode 100644
index 0000000000..9596d8cc85
--- /dev/null
+++ b/modules/ocl/src/opencl/kernel_stablesort_by_key.cl
@@ -0,0 +1,296 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef K_T
+#define K_T float
+#endif
+
+#ifndef V_T
+#define V_T float
+#endif
+
+#ifndef IS_GT
+#define IS_GT false
+#endif
+
+#if IS_GT
+#define my_comp(x,y) ((x) > (y))
+#else
+#define my_comp(x,y) ((x) < (y))
+#endif
+
+///////////// parallel merge sort ///////////////
+// ported from https://github.com/HSA-Libraries/Bolt/blob/master/include/bolt/cl/stablesort_by_key_kernels.cl
+uint lowerBoundLinear( global K_T* data, uint left, uint right, K_T searchVal)
+{
+    //  The values firstIndex and lastIndex get modified within the loop, narrowing down the potential sequence
+    uint firstIndex = left;
+    uint lastIndex = right;
+
+    //  This loops through [firstIndex, lastIndex)
+    //  Since firstIndex and lastIndex will be different for every thread depending on the nested branch,
+    //  this while loop will be divergent within a wavefront
+    while( firstIndex < lastIndex )
+    {
+        K_T dataVal = data[ firstIndex ];
+
+        //  This branch will create divergent wavefronts
+        if( my_comp( dataVal, searchVal ) )
+        {
+            firstIndex = firstIndex+1;
+        }
+        else
+        {
+            break;
+        }
+    }
+
+    return firstIndex;
+}
+
+//  This implements a binary search routine to look for an 'insertion point' in a sequence, denoted
+//  by a base pointer and left and right index for a particular candidate value.  The comparison operator is
+//  passed as a functor parameter my_comp
+//  This function returns an index that is the first index whos value would be equal to the searched value
+uint lowerBoundBinary( global K_T* data, uint left, uint right, K_T searchVal)
+{
+    //  The values firstIndex and lastIndex get modified within the loop, narrowing down the potential sequence
+    uint firstIndex = left;
+    uint lastIndex = right;
+
+    //  This loops through [firstIndex, lastIndex)
+    //  Since firstIndex and lastIndex will be different for every thread depending on the nested branch,
+    //  this while loop will be divergent within a wavefront
+    while( firstIndex < lastIndex )
+    {
+        //  midIndex is the average of first and last, rounded down
+        uint midIndex = ( firstIndex + lastIndex ) / 2;
+        K_T midValue = data[ midIndex ];
+
+        //  This branch will create divergent wavefronts
+        if( my_comp( midValue, searchVal ) )
+        {
+            firstIndex = midIndex+1;
+            // printf( "lowerBound: lastIndex[ %i ]=%i\n", get_local_id( 0 ), lastIndex );
+        }
+        else
+        {
+            lastIndex = midIndex;
+            // printf( "lowerBound: firstIndex[ %i ]=%i\n", get_local_id( 0 ), firstIndex );
+        }
+    }
+
+    return firstIndex;
+}
+
+//  This implements a binary search routine to look for an 'insertion point' in a sequence, denoted
+//  by a base pointer and left and right index for a particular candidate value.  The comparison operator is
+//  passed as a functor parameter my_comp
+//  This function returns an index that is the first index whos value would be greater than the searched value
+//  If the search value is not found in the sequence, upperbound returns the same result as lowerbound
+uint upperBoundBinary( global K_T* data, uint left, uint right, K_T searchVal)
+{
+    uint upperBound = lowerBoundBinary( data, left, right, searchVal );
+
+    // printf( "upperBoundBinary: upperBound[ %i, %i ]= %i\n", left, right, upperBound );
+    //  If upperBound == right, then  searchVal was not found in the sequence.  Just return.
+    if( upperBound != right )
+    {
+        //  While the values are equal i.e. !(x < y) && !(y < x) increment the index
+        K_T upperValue = data[ upperBound ];
+        while( !my_comp( upperValue, searchVal ) && !my_comp( searchVal, upperValue) && (upperBound != right) )
+        {
+            upperBound++;
+            upperValue = data[ upperBound ];
+        }
+    }
+
+    return upperBound;
+}
+
+//  This kernel implements merging of blocks of sorted data.  The input to this kernel most likely is
+//  the output of blockInsertionSortTemplate.  It is expected that the source array contains multiple
+//  blocks, each block is independently sorted.  The goal is to write into the output buffer half as
+//  many blocks, of double the size.  The even and odd blocks are stably merged together to form
+//  a new sorted block of twice the size.  The algorithm is out-of-place.
+kernel void merge(
+    global K_T*   iKey_ptr,
+    global V_T*   iValue_ptr,
+    global K_T*   oKey_ptr,
+    global V_T*   oValue_ptr,
+    const uint    srcVecSize,
+    const uint    srcLogicalBlockSize,
+    local K_T*    key_lds,
+    local V_T*    val_lds
+)
+{
+    size_t globalID     = get_global_id( 0 );
+    size_t groupID      = get_group_id( 0 );
+    size_t localID      = get_local_id( 0 );
+    size_t wgSize       = get_local_size( 0 );
+
+    //  Abort threads that are passed the end of the input vector
+    if( globalID >= srcVecSize )
+        return; // on SI this doesn't mess-up barriers
+
+    //  For an element in sequence A, find the lowerbound index for it in sequence B
+    uint srcBlockNum   = globalID / srcLogicalBlockSize;
+    uint srcBlockIndex = globalID % srcLogicalBlockSize;
+
+    // printf( "mergeTemplate: srcBlockNum[%i]=%i\n", srcBlockNum, srcBlockIndex );
+
+    //  Pairs of even-odd blocks will be merged together
+    //  An even block should search for an insertion point in the next odd block,
+    //  and the odd block should look for an insertion point in the corresponding previous even block
+    uint dstLogicalBlockSize = srcLogicalBlockSize<<1;
+    uint leftBlockIndex = globalID & ~((dstLogicalBlockSize) - 1 );
+    leftBlockIndex += (srcBlockNum & 0x1) ? 0 : srcLogicalBlockSize;
+    leftBlockIndex = min( leftBlockIndex, srcVecSize );
+    uint rightBlockIndex = min( leftBlockIndex + srcLogicalBlockSize, srcVecSize );
+
+    // if( localID == 0 )
+    // {
+    // printf( "mergeTemplate: wavefront[ %i ] logicalBlock[ %i ] logicalIndex[ %i ] leftBlockIndex[ %i ] <=> rightBlockIndex[ %i ]\n", groupID, srcBlockNum, srcBlockIndex, leftBlockIndex, rightBlockIndex );
+    // }
+
+    //  For a particular element in the input array, find the lowerbound index for it in the search sequence given by leftBlockIndex & rightBlockIndex
+    // uint insertionIndex = lowerBoundLinear( iKey_ptr, leftBlockIndex, rightBlockIndex, iKey_ptr[ globalID ], my_comp ) - leftBlockIndex;
+    uint insertionIndex = 0;
+    if( (srcBlockNum & 0x1) == 0 )
+    {
+        insertionIndex = lowerBoundBinary( iKey_ptr, leftBlockIndex, rightBlockIndex, iKey_ptr[ globalID ] ) - leftBlockIndex;
+    }
+    else
+    {
+        insertionIndex = upperBoundBinary( iKey_ptr, leftBlockIndex, rightBlockIndex, iKey_ptr[ globalID ] ) - leftBlockIndex;
+    }
+
+    //  The index of an element in the result sequence is the summation of it's indixes in the two input
+    //  sequences
+    uint dstBlockIndex = srcBlockIndex + insertionIndex;
+    uint dstBlockNum = srcBlockNum/2;
+
+    // if( (dstBlockNum*dstLogicalBlockSize)+dstBlockIndex == 395 )
+    // {
+    // printf( "mergeTemplate: (dstBlockNum[ %i ] * dstLogicalBlockSize[ %i ]) + dstBlockIndex[ %i ] = srcBlockIndex[ %i ] + insertionIndex[ %i ]\n", dstBlockNum, dstLogicalBlockSize, dstBlockIndex, srcBlockIndex, insertionIndex );
+    // printf( "mergeTemplate: dstBlockIndex[ %i ] = iKey_ptr[ %i ] ( %i )\n", (dstBlockNum*dstLogicalBlockSize)+dstBlockIndex, globalID, iKey_ptr[ globalID ] );
+    // }
+    oKey_ptr[ (dstBlockNum*dstLogicalBlockSize)+dstBlockIndex ] = iKey_ptr[ globalID ];
+    oValue_ptr[ (dstBlockNum*dstLogicalBlockSize)+dstBlockIndex ] = iValue_ptr[ globalID ];
+    // printf( "mergeTemplate: leftResultIndex[ %i ]=%i + %i\n", leftResultIndex, srcBlockIndex, leftInsertionIndex );
+}
+
+kernel void blockInsertionSort(
+    global K_T*   key_ptr,
+    global V_T*   value_ptr,
+    const uint    vecSize,
+    local K_T*    key_lds,
+    local V_T*    val_lds
+)
+{
+    size_t gloId    = get_global_id( 0 );
+    size_t groId    = get_group_id( 0 );
+    size_t locId    = get_local_id( 0 );
+    size_t wgSize   = get_local_size( 0 );
+
+    bool in_range = gloId < vecSize;
+    K_T key;
+    V_T val;
+    //  Abort threads that are passed the end of the input vector
+    if (in_range)
+    {
+        //  Make a copy of the entire input array into fast local memory
+        key = key_ptr[ gloId ];
+        val = value_ptr[ gloId ];
+        key_lds[ locId ] = key;
+        val_lds[ locId ] = val;
+    }
+    barrier( CLK_LOCAL_MEM_FENCE );
+    //  Sorts a workgroup using a naive insertion sort
+    //  The sort uses one thread within a workgroup to sort the entire workgroup
+    if( locId == 0 && in_range )
+    {
+        //  The last workgroup may have an irregular size, so we calculate a per-block endIndex
+        //  endIndex is essentially emulating a mod operator with subtraction and multiply
+        size_t endIndex = vecSize - ( groId * wgSize );
+        endIndex = min( endIndex, wgSize );
+
+        // printf( "Debug: endIndex[%i]=%i\n", groId, endIndex );
+
+        //  Indices are signed because the while loop will generate a -1 index inside of the max function
+        for( int currIndex = 1; currIndex < endIndex; ++currIndex )
+        {
+            key = key_lds[ currIndex ];
+            val = val_lds[ currIndex ];
+            int scanIndex = currIndex;
+            K_T ldsKey = key_lds[scanIndex - 1];
+            while( scanIndex > 0 && my_comp( key, ldsKey ) )
+            {
+                V_T ldsVal = val_lds[scanIndex - 1];
+
+                //  If the keys are being swapped, make sure the values are swapped identicaly
+                key_lds[ scanIndex ] = ldsKey;
+                val_lds[ scanIndex ] = ldsVal;
+
+                scanIndex = scanIndex - 1;
+                ldsKey = key_lds[ max( 0, scanIndex - 1 ) ];  // scanIndex-1 may be -1
+            }
+            key_lds[ scanIndex ] = key;
+            val_lds[ scanIndex ] = val;
+        }
+    }
+    barrier( CLK_LOCAL_MEM_FENCE );
+
+    if(in_range)
+    {
+        key = key_lds[ locId ];
+        key_ptr[ gloId ] = key;
+
+        val = val_lds[ locId ];
+        value_ptr[ gloId ] = val;
+    }
+}
+
+///////////// Radix sort from b40c library /////////////
diff --git a/modules/gpu/include/opencv2/gpu/devmem2d.hpp b/modules/ocl/src/opencl/kmeans_kernel.cl
similarity index 65%
rename from modules/gpu/include/opencv2/gpu/devmem2d.hpp
rename to modules/ocl/src/opencl/kmeans_kernel.cl
index fc2d2f2757..c6af0ad249 100644
--- a/modules/gpu/include/opencv2/gpu/devmem2d.hpp
+++ b/modules/ocl/src/opencl/kmeans_kernel.cl
@@ -10,10 +10,13 @@
 //                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
+// @Authors
+//    Xiaopeng Fu, fuxiaopeng2222@163.com
+//
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -27,7 +30,7 @@
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -40,4 +43,42 @@
 //
 //M*/
 
-#include "opencv2/core/cuda_devptrs.hpp"
+__kernel void distanceToCenters(
+    int label_step, int K,
+    __global float *src,
+    __global int *labels, int dims, int rows,
+    __global float *centers,
+    __global float *dists)
+{
+    int gid = get_global_id(1);
+
+    float dist, euDist, min;
+    int minCentroid;
+
+    if(gid >= rows)
+        return;
+
+    for(int i = 0 ; i < K; i++)
+    {
+        euDist = 0;
+        for(int j = 0; j < dims; j++)
+        {
+            dist = (src[j + gid * dims]
+                    - centers[j + i * dims]);
+            euDist += dist * dist;
+        }
+
+        if(i == 0)
+        {
+            min = euDist;
+            minCentroid = 0;
+        }
+        else if(euDist < min)
+        {
+            min = euDist;
+            minCentroid = i;
+        }
+    }
+    dists[gid] = min;
+    labels[label_step * gid] = minCentroid;
+}
diff --git a/modules/ocl/src/kernels/match_template.cl b/modules/ocl/src/opencl/match_template.cl
similarity index 77%
rename from modules/ocl/src/kernels/match_template.cl
rename to modules/ocl/src/opencl/match_template.cl
index ddbd86ba49..0dd3e69c40 100644
--- a/modules/ocl/src/kernels/match_template.cl
+++ b/modules/ocl/src/opencl/match_template.cl
@@ -45,22 +45,28 @@
 
 #pragma OPENCL EXTENSION cl_amd_printf : enable
 
-#if defined (__ATI__)
-#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#if defined (DOUBLE_SUPPORT)
 
-#elif defined (__NVIDIA__)
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
 #endif
 
-#if !defined(USE_SQR_INTEGRAL) && (defined (__ATI__) || defined (__NVIDIA__))
 #define TYPE_IMAGE_SQSUM double
 #else
-#define TYPE_IMAGE_SQSUM ulong
+#define TYPE_IMAGE_SQSUM float
+#endif
+
+#ifndef CN4
+#define CN4 1
+#else
+#define CN4 4
 #endif
 
 //////////////////////////////////////////////////
 // utilities
-#define SQSUMS_PTR(ox, oy) mad24(gidy + oy, img_sqsums_step, gidx + img_sqsums_offset + ox)
+#define SQSUMS_PTR(ox, oy) mad24(gidy + oy, img_sqsums_step, (gidx + img_sqsums_offset + ox) * CN4)
 #define SUMS_PTR(ox, oy) mad24(gidy + oy, img_sums_step, gidx + img_sums_offset + ox)
 // normAcc* are accurate normalization routines which make GPU matchTemplate
 // consistent with CPU one
@@ -95,7 +101,7 @@ float normAcc_SQDIFF(float num, float denum)
 __kernel
 void normalizeKernel_C1_D0
 (
-    __global const TYPE_IMAGE_SQSUM * img_sqsums,
+    __global const float * img_sqsums,
     __global float * res,
     ulong tpl_sqsum,
     int res_rows,
@@ -119,8 +125,8 @@ void normalizeKernel_C1_D0
     if(gidx < res_cols && gidy < res_rows)
     {
         float image_sqsum_ = (float)(
-            (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
+                                 (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
+                                 (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
         res[res_idx] = normAcc(res[res_idx], sqrt(image_sqsum_ * tpl_sqsum));
     }
 }
@@ -152,8 +158,8 @@ void matchTemplate_Prepared_SQDIFF_C1_D0
     if(gidx < res_cols && gidy < res_rows)
     {
         float image_sqsum_ = (float)(
-            (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
+                                 (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
+                                 (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
         res[res_idx] = image_sqsum_ - 2.f * res[res_idx] + tpl_sqsum;
     }
 }
@@ -161,7 +167,7 @@ void matchTemplate_Prepared_SQDIFF_C1_D0
 __kernel
 void matchTemplate_Prepared_SQDIFF_NORMED_C1_D0
 (
-    __global const TYPE_IMAGE_SQSUM * img_sqsums,
+    __global const float * img_sqsums,
     __global float * res,
     ulong tpl_sqsum,
     int res_rows,
@@ -185,10 +191,10 @@ void matchTemplate_Prepared_SQDIFF_NORMED_C1_D0
     if(gidx < res_cols && gidy < res_rows)
     {
         float image_sqsum_ = (float)(
-            (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
+                                 (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
+                                 (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
         res[res_idx] = normAcc_SQDIFF(image_sqsum_ - 2.f * res[res_idx] + tpl_sqsum,
-                                        sqrt(image_sqsum_ * tpl_sqsum));
+                                      sqrt(image_sqsum_ * tpl_sqsum));
     }
 }
 
@@ -441,10 +447,10 @@ void matchTemplate_Naive_CCORR_C1_D0
             __global const uchar * tpl_ptr = tpl + mad24(i, tpl_step, tpl_offset);
             for(j = 0; j < tpl_cols; j ++)
             {
-                sum = mad24(img_ptr[j], tpl_ptr[j], sum);
+                sum = mad24(convert_int(img_ptr[j]), convert_int(tpl_ptr[j]), sum);
             }
         }
-        res[res_idx] = sum;
+        res[res_idx] = (float)sum;
     }
 }
 
@@ -542,7 +548,7 @@ void matchTemplate_Naive_CCORR_C4_D0
                 sum   = mad24(convert_int4(img_ptr[j]), convert_int4(tpl_ptr[j]), sum);
             }
         }
-        res[res_idx] = sum.x + sum.y + sum.z + sum.w;
+        res[res_idx] = (float)(sum.x + sum.y + sum.z + sum.w);
     }
 }
 
@@ -627,9 +633,8 @@ void matchTemplate_Prepared_CCOFF_C1_D0
 
     if(gidx < res_cols && gidy < res_rows)
     {
-        float sum = (float)(
-            (img_sums[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums[SUMS_PTR(0, tpl_rows)] - img_sums[SUMS_PTR(0, 0)]));
+        float sum = (float)((img_sums[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums[SUMS_PTR(tpl_cols, 0)])
+                            -(img_sums[SUMS_PTR(0, tpl_rows)] - img_sums[SUMS_PTR(0, 0)]));
         res[res_idx] -= sum * tpl_sum;
     }
 }
@@ -671,17 +676,17 @@ void matchTemplate_Prepared_CCOFF_C4_D0
     {
         float ccorr = res[res_idx];
         ccorr -= tpl_sum_c0*(float)(
-            (img_sums_c0[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c0[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c0[SUMS_PTR(0, tpl_rows)] - img_sums_c0[SUMS_PTR(0, 0)]));
+                     (img_sums_c0[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c0[SUMS_PTR(tpl_cols, 0)])
+                     - (img_sums_c0[SUMS_PTR(0, tpl_rows)] - img_sums_c0[SUMS_PTR(0, 0)]));
         ccorr -= tpl_sum_c1*(float)(
-            (img_sums_c1[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c1[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c1[SUMS_PTR(0, tpl_rows)] - img_sums_c1[SUMS_PTR(0, 0)]));
+                     (img_sums_c1[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c1[SUMS_PTR(tpl_cols, 0)])
+                     - (img_sums_c1[SUMS_PTR(0, tpl_rows)] - img_sums_c1[SUMS_PTR(0, 0)]));
         ccorr -= tpl_sum_c2*(float)(
-            (img_sums_c2[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c2[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c2[SUMS_PTR(0, tpl_rows)] - img_sums_c2[SUMS_PTR(0, 0)]));
+                     (img_sums_c2[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c2[SUMS_PTR(tpl_cols, 0)])
+                     - (img_sums_c2[SUMS_PTR(0, tpl_rows)] - img_sums_c2[SUMS_PTR(0, 0)]));
         ccorr -= tpl_sum_c3*(float)(
-            (img_sums_c3[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c3[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c3[SUMS_PTR(0, tpl_rows)] - img_sums_c3[SUMS_PTR(0, 0)]));
+                     (img_sums_c3[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c3[SUMS_PTR(tpl_cols, 0)])
+                     - (img_sums_c3[SUMS_PTR(0, tpl_rows)] - img_sums_c3[SUMS_PTR(0, 0)]));
         res[res_idx] = ccorr;
     }
 }
@@ -702,7 +707,7 @@ void matchTemplate_Prepared_CCOFF_NORMED_C1_D0
     __global const uint * img_sums,
     int img_sums_offset,
     int img_sums_step,
-    __global const TYPE_IMAGE_SQSUM * img_sqsums,
+    __global const float * img_sqsums,
     int img_sqsums_offset,
     int img_sqsums_step,
     float tpl_sum,
@@ -725,12 +730,12 @@ void matchTemplate_Prepared_CCOFF_NORMED_C1_D0
     if(gidx < res_cols && gidy < res_rows)
     {
         float image_sum_ =  (float)(
-            (img_sums[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums[SUMS_PTR(0, tpl_rows)] - img_sums[SUMS_PTR(0, 0)]));
+                                (img_sums[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums[SUMS_PTR(tpl_cols, 0)])
+                                - (img_sums[SUMS_PTR(0, tpl_rows)] - img_sums[SUMS_PTR(0, 0)]));
 
         float image_sqsum_ = (float)(
-            (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
+                                 (img_sqsums[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums[SQSUMS_PTR(tpl_cols, 0)]) -
+                                 (img_sqsums[SQSUMS_PTR(0, tpl_rows)] - img_sqsums[SQSUMS_PTR(0, 0)]));
         res[res_idx] = normAcc(res[res_idx] - image_sum_ * tpl_sum,
                                sqrt(tpl_sqsum * (image_sqsum_ - weight * image_sum_ * image_sum_)));
     }
@@ -754,10 +759,10 @@ void matchTemplate_Prepared_CCOFF_NORMED_C4_D0
     __global const uint * img_sums_c3,
     int img_sums_offset,
     int img_sums_step,
-    __global const TYPE_IMAGE_SQSUM * img_sqsums_c0,
-    __global const TYPE_IMAGE_SQSUM * img_sqsums_c1,
-    __global const TYPE_IMAGE_SQSUM * img_sqsums_c2,
-    __global const TYPE_IMAGE_SQSUM * img_sqsums_c3,
+    __global const float * img_sqsums_c0,
+    __global const float * img_sqsums_c1,
+    __global const float * img_sqsums_c2,
+    __global const float * img_sqsums_c3,
     int img_sqsums_offset,
     int img_sqsums_step,
     float tpl_sum_c0,
@@ -782,43 +787,71 @@ void matchTemplate_Prepared_CCOFF_NORMED_C4_D0
     if(gidx < res_cols && gidy < res_rows)
     {
         float image_sum_c0 =  (float)(
-            (img_sums_c0[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c0[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c0[SUMS_PTR(0, tpl_rows)] - img_sums_c0[SUMS_PTR(0, 0)]));
+                                  (img_sums_c0[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c0[SUMS_PTR(tpl_cols, 0)])
+                                  - (img_sums_c0[SUMS_PTR(0, tpl_rows)] - img_sums_c0[SUMS_PTR(0, 0)]));
         float image_sum_c1 =  (float)(
-            (img_sums_c1[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c1[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c1[SUMS_PTR(0, tpl_rows)] - img_sums_c1[SUMS_PTR(0, 0)]));
+                                  (img_sums_c1[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c1[SUMS_PTR(tpl_cols, 0)])
+                                  - (img_sums_c1[SUMS_PTR(0, tpl_rows)] - img_sums_c1[SUMS_PTR(0, 0)]));
         float image_sum_c2 =  (float)(
-            (img_sums_c2[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c2[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c2[SUMS_PTR(0, tpl_rows)] - img_sums_c2[SUMS_PTR(0, 0)]));
+                                  (img_sums_c2[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c2[SUMS_PTR(tpl_cols, 0)])
+                                  - (img_sums_c2[SUMS_PTR(0, tpl_rows)] - img_sums_c2[SUMS_PTR(0, 0)]));
         float image_sum_c3 =  (float)(
-            (img_sums_c3[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c3[SUMS_PTR(tpl_cols, 0)])
-          - (img_sums_c3[SUMS_PTR(0, tpl_rows)] - img_sums_c3[SUMS_PTR(0, 0)]));
+                                  (img_sums_c3[SUMS_PTR(tpl_cols, tpl_rows)] - img_sums_c3[SUMS_PTR(tpl_cols, 0)])
+                                  - (img_sums_c3[SUMS_PTR(0, tpl_rows)] - img_sums_c3[SUMS_PTR(0, 0)]));
 
         float image_sqsum_c0 = (float)(
-            (img_sqsums_c0[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c0[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums_c0[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c0[SQSUMS_PTR(0, 0)]));
+                                   (img_sqsums_c0[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c0[SQSUMS_PTR(tpl_cols, 0)]) -
+                                   (img_sqsums_c0[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c0[SQSUMS_PTR(0, 0)]));
         float image_sqsum_c1 = (float)(
-            (img_sqsums_c1[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c1[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums_c1[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c1[SQSUMS_PTR(0, 0)]));
+                                   (img_sqsums_c1[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c1[SQSUMS_PTR(tpl_cols, 0)]) -
+                                   (img_sqsums_c1[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c1[SQSUMS_PTR(0, 0)]));
         float image_sqsum_c2 = (float)(
-            (img_sqsums_c2[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c2[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums_c2[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c2[SQSUMS_PTR(0, 0)]));
+                                   (img_sqsums_c2[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c2[SQSUMS_PTR(tpl_cols, 0)]) -
+                                   (img_sqsums_c2[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c2[SQSUMS_PTR(0, 0)]));
         float image_sqsum_c3 = (float)(
-            (img_sqsums_c3[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c3[SQSUMS_PTR(tpl_cols, 0)]) -
-            (img_sqsums_c3[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c3[SQSUMS_PTR(0, 0)]));
+                                   (img_sqsums_c3[SQSUMS_PTR(tpl_cols, tpl_rows)] - img_sqsums_c3[SQSUMS_PTR(tpl_cols, 0)]) -
+                                   (img_sqsums_c3[SQSUMS_PTR(0, tpl_rows)] - img_sqsums_c3[SQSUMS_PTR(0, 0)]));
 
         float num = res[res_idx] -
-            image_sum_c0 * tpl_sum_c0 -
-            image_sum_c1 * tpl_sum_c1 -
-            image_sum_c2 * tpl_sum_c2 -
-            image_sum_c3 * tpl_sum_c3;
+                    image_sum_c0 * tpl_sum_c0 -
+                    image_sum_c1 * tpl_sum_c1 -
+                    image_sum_c2 * tpl_sum_c2 -
+                    image_sum_c3 * tpl_sum_c3;
         float denum = sqrt( tpl_sqsum * (
-            image_sqsum_c0 - weight * image_sum_c0 * image_sum_c0 +
-            image_sqsum_c1 - weight * image_sum_c1 * image_sum_c1 +
-            image_sqsum_c2 - weight * image_sum_c2 * image_sum_c2 +
-            image_sqsum_c3 - weight * image_sum_c0 * image_sum_c3)
-            );
+                                image_sqsum_c0 - weight * image_sum_c0 * image_sum_c0 +
+                                image_sqsum_c1 - weight * image_sum_c1 * image_sum_c1 +
+                                image_sqsum_c2 - weight * image_sum_c2 * image_sum_c2 +
+                                image_sqsum_c3 - weight * image_sum_c0 * image_sum_c3)
+                          );
         res[res_idx] = normAcc(num, denum);
     }
 }
 
+//////////////////////////////////////////////////////////////////////
+// extractFirstChannel
+__kernel
+void extractFirstChannel
+(
+    const __global float4* img,
+    __global float* res,
+    int rows,
+    int cols,
+    int img_offset,
+    int res_offset,
+    int img_step,
+    int res_step
+)
+{
+    img_step   /= sizeof(float4);
+    res_step   /= sizeof(float);
+    img_offset /= sizeof(float4);
+    res_offset /= sizeof(float);
+    img += img_offset;
+    res += res_offset;
+    int gidx = get_global_id(0);
+    int gidy = get_global_id(1);
+    if(gidx < cols && gidy < rows)
+    {
+        res[gidx + gidy * res_step] = img[gidx + gidy * img_step].x;
+    }
+}
diff --git a/modules/ocl/src/kernels/meanShift.cl b/modules/ocl/src/opencl/meanShift.cl
similarity index 99%
rename from modules/ocl/src/kernels/meanShift.cl
rename to modules/ocl/src/opencl/meanShift.cl
index 4b5a08b352..a5b110812d 100644
--- a/modules/ocl/src/kernels/meanShift.cl
+++ b/modules/ocl/src/opencl/meanShift.cl
@@ -240,4 +240,3 @@ __kernel void meanshiftproc_kernel( __global uchar4* in, __global uchar4* outr,
 //        outsp[basesp] =(short2)((short)x0,(short)y0);
     }
 }
-
diff --git a/modules/ocl/src/kernels/merge_mat.cl b/modules/ocl/src/opencl/merge_mat.cl
similarity index 100%
rename from modules/ocl/src/kernels/merge_mat.cl
rename to modules/ocl/src/opencl/merge_mat.cl
diff --git a/modules/ocl/src/kernels/moments.cl b/modules/ocl/src/opencl/moments.cl
similarity index 50%
rename from modules/ocl/src/kernels/moments.cl
rename to modules/ocl/src/opencl/moments.cl
index 852dd6af52..5ea94e20c8 100644
--- a/modules/ocl/src/kernels/moments.cl
+++ b/modules/ocl/src/opencl/moments.cl
@@ -1,42 +1,103 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Sen Liu, swjtuls1987@126.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #if defined (DOUBLE_SUPPORT)
+
+#ifdef cl_khr_fp64
 #pragma OPENCL EXTENSION cl_khr_fp64:enable
-#else
-typedef float double;
-typedef float4 double4;
-#define convert_double4 convert_float4
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
 #endif
-//#pragma OPENCL EXTENSION cl_amd_printf:enable
-//#if defined (DOUBLE_SUPPORT)
+typedef double T;
+typedef double F;
+typedef double4 F4;
+#define convert_F4 convert_double4
+
+#else
+typedef float F;
+typedef float4 F4;
+typedef long T;
+#define convert_F4 convert_float4
+#endif
+
+#define DST_ROW_00     0
+#define DST_ROW_10     1
+#define DST_ROW_01     2
+#define DST_ROW_20     3
+#define DST_ROW_11     4
+#define DST_ROW_02     5
+#define DST_ROW_30     6
+#define DST_ROW_21     7
+#define DST_ROW_12     8
+#define DST_ROW_03     9
+
 __kernel void icvContourMoments(int contour_total,
                                 __global float* reader_oclmat_data,
-                                __global double* dst_a00,
-                                __global double* dst_a10,
-                                __global double* dst_a01,
-                                __global double* dst_a20,
-                                __global double* dst_a11,
-                                __global double* dst_a02,
-                                __global double* dst_a30,
-                                __global double* dst_a21,
-                                __global double* dst_a12,
-                                __global double* dst_a03)
+                                __global T* dst_a,
+                                int dst_step)
 {
-    double xi_1, yi_1, xi_12, yi_12, xi, yi, xi2, yi2, dxy, xii_1, yii_1;
+    T xi_1, yi_1, xi_12, yi_12, xi, yi, xi2, yi2, dxy, xii_1, yii_1;
     int idx = get_global_id(0);
 
-    xi_1 = *(reader_oclmat_data + (get_global_id(0) << 1));
-    yi_1 = *(reader_oclmat_data + (get_global_id(0) << 1) + 1);
+    if (idx < 0 || idx >= contour_total)
+        return;
+
+    xi_1 = (T)(*(reader_oclmat_data + (get_global_id(0) << 1)));
+    yi_1 = (T)(*(reader_oclmat_data + (get_global_id(0) << 1) + 1));
     xi_12 = xi_1 * xi_1;
     yi_12 = yi_1 * yi_1;
 
     if(idx == contour_total - 1)
     {
-        xi = *(reader_oclmat_data);
-        yi = *(reader_oclmat_data + 1);
+        xi = (T)(*(reader_oclmat_data));
+        yi = (T)(*(reader_oclmat_data + 1));
     }
     else
     {
-        xi = *(reader_oclmat_data + (idx + 1) * 2);
-        yi = *(reader_oclmat_data + (idx + 1) * 2 + 1);
+        xi = (T)(*(reader_oclmat_data + (idx + 1) * 2));
+        yi = (T)(*(reader_oclmat_data + (idx + 1) * 2 + 1));
     }
 
     xi2 = xi * xi;
@@ -45,36 +106,77 @@ __kernel void icvContourMoments(int contour_total,
     xii_1 = xi_1 + xi;
     yii_1 = yi_1 + yi;
 
-    dst_a00[idx] = dxy;
-    dst_a10[idx] = dxy * xii_1;
-    dst_a01[idx] = dxy * yii_1;
-    dst_a20[idx] = dxy * (xi_1 * xii_1 + xi2);
-    dst_a11[idx] = dxy * (xi_1 * (yii_1 + yi_1) + xi * (yii_1 + yi));
-    dst_a02[idx] = dxy * (yi_1 * yii_1 + yi2);
-    dst_a30[idx] = dxy * xii_1 * (xi_12 + xi2);
-    dst_a03[idx] = dxy * yii_1 * (yi_12 + yi2);
-    dst_a21[idx] =
+    dst_step /= sizeof(T);
+    *( dst_a + DST_ROW_00 * dst_step + idx) = dxy;
+    *( dst_a + DST_ROW_10 * dst_step + idx) = dxy * xii_1;
+    *( dst_a + DST_ROW_01 * dst_step + idx) = dxy * yii_1;
+    *( dst_a + DST_ROW_20 * dst_step + idx) = dxy * (xi_1 * xii_1 + xi2);
+    *( dst_a + DST_ROW_11 * dst_step + idx) = dxy * (xi_1 * (yii_1 + yi_1) + xi * (yii_1 + yi));
+    *( dst_a + DST_ROW_02 * dst_step + idx) = dxy * (yi_1 * yii_1 + yi2);
+    *( dst_a + DST_ROW_30 * dst_step + idx) = dxy * xii_1 * (xi_12 + xi2);
+    *( dst_a + DST_ROW_03 * dst_step + idx) = dxy * yii_1 * (yi_12 + yi2);
+    *( dst_a + DST_ROW_21 * dst_step + idx) =
         dxy * (xi_12 * (3 * yi_1 + yi) + 2 * xi * xi_1 * yii_1 +
                xi2 * (yi_1 + 3 * yi));
-    dst_a12[idx] =
+    *( dst_a + DST_ROW_12 * dst_step + idx) =
         dxy * (yi_12 * (3 * xi_1 + xi) + 2 * yi * yi_1 * xii_1 +
                yi2 * (xi_1 + 3 * xi));
 }
-//#endif
 
-//#if defined (DOUBLE_SUPPORT)
-__kernel void CvMoments_D0(__global uchar16* src_data, int src_rows, int src_cols, int src_step, int tileSize_width, int tileSize_height,
-                           __global double* dst_m00,
-                           __global double* dst_m10,
-                           __global double* dst_m01,
-                           __global double* dst_m20,
-                           __global double* dst_m11,
-                           __global double* dst_m02,
-                           __global double* dst_m30,
-                           __global double* dst_m21,
-                           __global double* dst_m12,
-                           __global double* dst_m03,
-                           int dst_cols, int dst_step, int type, int depth, int cn, int coi, int binary, int TILE_SIZE)
+__kernel void dst_sum(int src_rows, int src_cols, int tile_height, int tile_width, int TILE_SIZE,
+                      __global F* sum, __global F* dst_m, int dst_step)
+{
+    int gidy = get_global_id(0);
+    int gidx = get_global_id(1);
+    int block_y = src_rows/tile_height;
+    int block_x = src_cols/tile_width;
+    int block_num;
+
+    if(src_rows > TILE_SIZE && src_rows % TILE_SIZE != 0)
+        block_y ++;
+    if(src_cols > TILE_SIZE && src_cols % TILE_SIZE != 0)
+        block_x ++;
+    block_num = block_y * block_x;
+    __local F dst_sum[10][128];
+    if(gidy<128-block_num)
+        for(int i=0; i<10; i++)
+            dst_sum[i][gidy+block_num]=0;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    dst_step /= sizeof(F);
+    if(gidy<block_num)
+    {
+        dst_sum[0][gidy] = *(dst_m + mad24(DST_ROW_00 * block_y, dst_step, gidy));
+        dst_sum[1][gidy] = *(dst_m + mad24(DST_ROW_10 * block_y, dst_step, gidy));
+        dst_sum[2][gidy] = *(dst_m + mad24(DST_ROW_01 * block_y, dst_step, gidy));
+        dst_sum[3][gidy] = *(dst_m + mad24(DST_ROW_20 * block_y, dst_step, gidy));
+        dst_sum[4][gidy] = *(dst_m + mad24(DST_ROW_11 * block_y, dst_step, gidy));
+        dst_sum[5][gidy] = *(dst_m + mad24(DST_ROW_02 * block_y, dst_step, gidy));
+        dst_sum[6][gidy] = *(dst_m + mad24(DST_ROW_30 * block_y, dst_step, gidy));
+        dst_sum[7][gidy] = *(dst_m + mad24(DST_ROW_21 * block_y, dst_step, gidy));
+        dst_sum[8][gidy] = *(dst_m + mad24(DST_ROW_12 * block_y, dst_step, gidy));
+        dst_sum[9][gidy] = *(dst_m + mad24(DST_ROW_03 * block_y, dst_step, gidy));
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    for(int lsize=64; lsize>0; lsize>>=1)
+    {
+        if(gidy<lsize)
+        {
+            int lsize2 = gidy + lsize;
+            for(int i=0; i<10; i++)
+                dst_sum[i][gidy] += dst_sum[i][lsize2];
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+    if(gidy==0)
+        for(int i=0; i<10; i++)
+            sum[i] = dst_sum[i][0];
+}
+
+__kernel void CvMoments_D0(__global uchar16* src_data, int src_rows, int src_cols, int src_step,
+                           __global F* dst_m,
+                           int dst_cols, int dst_step, int blocky,
+                           int depth, int cn, int coi, int binary, int TILE_SIZE)
 {
     uchar tmp_coi[16]; // get the coi data
     uchar16 tmp[16];
@@ -90,35 +192,43 @@ __kernel void CvMoments_D0(__global uchar16* src_data, int src_rows, int src_col
     int x = wgidx*TILE_SIZE;  // vector length of uchar
     int kcn = (cn==2)?2:4;
     int rstep = min(src_step, TILE_SIZE);
-    tileSize_height = min(TILE_SIZE, src_rows - y);
-    tileSize_width = min(TILE_SIZE, src_cols - x);
+    int tileSize_height = min(TILE_SIZE, src_rows - y);
+    int tileSize_width = min(TILE_SIZE, src_cols - x);
+
+    if ( y+lidy < src_rows )
+    {
+        if( tileSize_width < TILE_SIZE )
+            for(int i = tileSize_width; i < rstep && (x+i) < src_cols; i++ )
+                *((__global uchar*)src_data+(y+lidy)*src_step+x+i) = 0;
+
+        if( coi > 0 )	//channel of interest
+            for(int i = 0; i < tileSize_width; i += VLEN_C)
+            {
+                for(int j=0; j<VLEN_C; j++)
+                    tmp_coi[j] = *((__global uchar*)src_data+(y+lidy)*src_step+(x+i+j)*kcn+coi-1);
+                tmp[i/VLEN_C] = (uchar16)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7],
+                                          tmp_coi[8],tmp_coi[9],tmp_coi[10],tmp_coi[11],tmp_coi[12],tmp_coi[13],tmp_coi[14],tmp_coi[15]);
+            }
+        else
+            for(int i=0; i < tileSize_width; i+=VLEN_C)
+                tmp[i/VLEN_C] = *(src_data+(y+lidy)*src_step/VLEN_C+(x+i)/VLEN_C);
+    }
 
-    if( tileSize_width < TILE_SIZE )
-        for(int i = tileSize_width; i < rstep; i++ )
-            *((__global uchar*)src_data+(y+lidy)*src_step+x+i) = 0;
-    if( coi > 0 )   //channel of interest
-        for(int i = 0; i < tileSize_width; i += VLEN_C)
-        {
-            for(int j=0; j<VLEN_C; j++)
-                tmp_coi[j] = *((__global uchar*)src_data+(y+lidy)*src_step+(x+i+j)*kcn+coi-1);
-            tmp[i/VLEN_C] = (uchar16)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7],
-                                      tmp_coi[8],tmp_coi[9],tmp_coi[10],tmp_coi[11],tmp_coi[12],tmp_coi[13],tmp_coi[14],tmp_coi[15]);
-        }
-    else
-        for(int i=0; i < tileSize_width; i+=VLEN_C)
-            tmp[i/VLEN_C] = *(src_data+(y+lidy)*src_step/VLEN_C+(x+i)/VLEN_C);
     uchar16 zero = (uchar16)(0);
     uchar16 full = (uchar16)(255);
     if( binary )
         for(int i=0; i < tileSize_width; i+=VLEN_C)
             tmp[i/VLEN_C] = (tmp[i/VLEN_C]!=zero)?full:zero;
-    double mom[10];
+
+    F mom[10];
     __local int m[10][128];
-    if(lidy == 0)
+    if(lidy < 128)
+    {
         for(int i=0; i<10; i++)
-            for(int j=0; j<128; j++)
-                m[i][j]=0;
+            m[i][lidy]=0;
+    }
     barrier(CLK_LOCAL_MEM_FENCE);
+
     int lm[10] = {0};
     int16 x0 = (int16)(0);
     int16 x1 = (int16)(0);
@@ -179,123 +289,58 @@ __kernel void CvMoments_D0(__global uchar16* src_data, int src_rows, int src_col
                 m[i][lidy-j/2] = lm[i];
         barrier(CLK_LOCAL_MEM_FENCE);
     }
+
     if(lidy == 0&&lidx == 0)
     {
         for( int mt = 0; mt < 10; mt++ )
-            mom[mt] = (double)lm[mt];
+            mom[mt] = (F)lm[mt];
         if(binary)
         {
-            double s = 1./255;
+            F s = 1./255;
             for( int mt = 0; mt < 10; mt++ )
                 mom[mt] *= s;
         }
-        double xm = x * mom[0], ym = y * mom[0];
+        F xm = x * mom[0], ym = y * mom[0];
 
         // accumulate moments computed in each tile
+        dst_step /= sizeof(F);
 
         // + m00 ( = m00' )
-        dst_m00[wgidy*dst_cols+wgidx] = mom[0];
+        *(dst_m + mad24(DST_ROW_00 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[0];
 
         // + m10 ( = m10' + x*m00' )
-        dst_m10[wgidy*dst_cols+wgidx]  = mom[1] + xm;
+        *(dst_m + mad24(DST_ROW_10 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[1] + xm;
 
         // + m01 ( = m01' + y*m00' )
-        dst_m01[wgidy*dst_cols+wgidx]  = mom[2] + ym;
+        *(dst_m + mad24(DST_ROW_01 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[2] + ym;
 
         // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
-        dst_m20[wgidy*dst_cols+wgidx]  = mom[3] + x * (mom[1] * 2 + xm);
+        *(dst_m + mad24(DST_ROW_20 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[3] + x * (mom[1] * 2 + xm);
 
         // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
-        dst_m11[wgidy*dst_cols+wgidx]  = mom[4] + x * (mom[2] + ym) + y * mom[1];
+        *(dst_m + mad24(DST_ROW_11 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[4] + x * (mom[2] + ym) + y * mom[1];
 
         // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
-        dst_m02[wgidy*dst_cols+wgidx]  = mom[5] + y * (mom[2] * 2 + ym);
+        *(dst_m + mad24(DST_ROW_02 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[5] + y * (mom[2] * 2 + ym);
 
         // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
-        dst_m30[wgidy*dst_cols+wgidx]  = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
+        *(dst_m + mad24(DST_ROW_30 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
 
         // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
-        dst_m21[wgidy*dst_cols+wgidx]  = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
+        *(dst_m + mad24(DST_ROW_21 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
 
         // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
-        dst_m12[wgidy*dst_cols+wgidx]  = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
+        *(dst_m + mad24(DST_ROW_12 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
 
         // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
-        dst_m03[wgidy*dst_cols+wgidx] = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
+        *(dst_m + mad24(DST_ROW_03 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
     }
 }
-//#endif
-//#if defined (DOUBLE_SUPPORT)
-__kernel void dst_sum(int src_rows, int src_cols, int tile_height, int tile_width, int TILE_SIZE, __global double* sum, __global double* dst_m00,
-                      __global double* dst_m10,
-                      __global double* dst_m01,
-                      __global double* dst_m20,
-                      __global double* dst_m11,
-                      __global double* dst_m02,
-                      __global double* dst_m30,
-                      __global double* dst_m21,
-                      __global double* dst_m12,
-                      __global double* dst_m03)
-{
-    int gidy = get_global_id(0);
-    int gidx = get_global_id(1);
-    int block_y = src_rows/tile_height;
-    int block_x = src_cols/tile_width;
-    int block_num;
 
-    if(src_rows > TILE_SIZE && src_rows % TILE_SIZE != 0)
-        block_y ++;
-    if(src_cols > TILE_SIZE && src_cols % TILE_SIZE != 0)
-        block_x ++;
-    block_num = block_y * block_x;
-    __local double dst_sum[10][128];
-    if(gidy<128-block_num)
-        for(int i=0; i<10; i++)
-            dst_sum[i][gidy+block_num]=0;
-    barrier(CLK_LOCAL_MEM_FENCE);
-    if(gidy<block_num)
-    {
-        dst_sum[0][gidy] = dst_m00[gidy];
-        dst_sum[1][gidy] = dst_m10[gidy];
-        dst_sum[2][gidy] = dst_m01[gidy];
-        dst_sum[3][gidy] = dst_m20[gidy];
-        dst_sum[4][gidy] = dst_m11[gidy];
-        dst_sum[5][gidy] = dst_m02[gidy];
-        dst_sum[6][gidy] = dst_m30[gidy];
-        dst_sum[7][gidy] = dst_m21[gidy];
-        dst_sum[8][gidy] = dst_m12[gidy];
-        dst_sum[9][gidy] = dst_m03[gidy];
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-    for(int lsize=64; lsize>0; lsize>>=1)
-    {
-        if(gidy<lsize)
-        {
-            int lsize2 = gidy + lsize;
-            for(int i=0; i<10; i++)
-                dst_sum[i][gidy] += dst_sum[i][lsize2];
-        }
-        barrier(CLK_LOCAL_MEM_FENCE);
-    }
-    if(gidy==0)
-        for(int i=0; i<10; i++)
-            sum[i] = dst_sum[i][0];
-}
-//#endif
-//#if defined (DOUBLE_SUPPORT)
-__kernel void CvMoments_D2(__global ushort8* src_data, int src_rows, int src_cols, int src_step, int tileSize_width, int tileSize_height,
-                           __global double* dst_m00,
-                           __global double* dst_m10,
-                           __global double* dst_m01,
-                           __global double* dst_m20,
-                           __global double* dst_m11,
-                           __global double* dst_m02,
-                           __global double* dst_m30,
-                           __global double* dst_m21,
-                           __global double* dst_m12,
-                           __global double* dst_m03,
-                           int dst_cols, int dst_step,
-                           int type, int depth, int cn, int coi, int binary, const int TILE_SIZE)
+__kernel void CvMoments_D2(__global ushort8* src_data, int src_rows, int src_cols, int src_step,
+                           __global F* dst_m,
+                           int dst_cols, int dst_step, int blocky,
+                           int depth, int cn, int coi, int binary, const int TILE_SIZE)
 {
     ushort tmp_coi[8]; // get the coi data
     ushort8 tmp[32];
@@ -310,33 +355,38 @@ __kernel void CvMoments_D2(__global ushort8* src_data, int src_rows, int src_col
     int x = wgidx*TILE_SIZE;  // real X index of pixel
     int kcn = (cn==2)?2:4;
     int rstep = min(src_step/2, TILE_SIZE);
-    tileSize_height = min(TILE_SIZE, src_rows - y);
-    tileSize_width = min(TILE_SIZE, src_cols -x);
-    if(src_cols > TILE_SIZE && tileSize_width < TILE_SIZE)
-        for(int i=tileSize_width; i < rstep; i++ )
-            *((__global ushort*)src_data+(y+lidy)*src_step/2+x+i) = 0;
-    if( coi > 0 )
-        for(int i=0; i < tileSize_width; i+=VLEN_US)
-        {
-            for(int j=0; j<VLEN_US; j++)
-                tmp_coi[j] = *((__global ushort*)src_data+(y+lidy)*(int)src_step/2+(x+i+j)*kcn+coi-1);
-            tmp[i/VLEN_US] = (ushort8)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7]);
-        }
-    else
-        for(int i=0; i < tileSize_width; i+=VLEN_US)
-            tmp[i/VLEN_US] = *(src_data+(y+lidy)*src_step/(2*VLEN_US)+(x+i)/VLEN_US);
+    int tileSize_height = min(TILE_SIZE, src_rows - y);
+    int tileSize_width = min(TILE_SIZE, src_cols -x);
+
+    if ( y+lidy < src_rows )
+    {
+        if(src_cols > TILE_SIZE && tileSize_width < TILE_SIZE)
+            for(int i=tileSize_width; i < rstep && (x+i) < src_cols; i++ )
+                *((__global ushort*)src_data+(y+lidy)*src_step/2+x+i) = 0;
+        if( coi > 0 )
+            for(int i=0; i < tileSize_width; i+=VLEN_US)
+            {
+                for(int j=0; j<VLEN_US; j++)
+                    tmp_coi[j] = *((__global ushort*)src_data+(y+lidy)*(int)src_step/2+(x+i+j)*kcn+coi-1);
+                tmp[i/VLEN_US] = (ushort8)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7]);
+            }
+        else
+            for(int i=0; i < tileSize_width; i+=VLEN_US)
+                tmp[i/VLEN_US] = *(src_data+(y+lidy)*src_step/(2*VLEN_US)+(x+i)/VLEN_US);
+    }
+
     ushort8 zero = (ushort8)(0);
     ushort8 full = (ushort8)(255);
     if( binary )
         for(int i=0; i < tileSize_width; i+=VLEN_US)
             tmp[i/VLEN_US] = (tmp[i/VLEN_US]!=zero)?full:zero;
-    double mom[10];
+    F mom[10];
     __local long m[10][128];
-    if(lidy == 0)
+    if(lidy < 128)
         for(int i=0; i<10; i++)
-            for(int j=0; j<128; j++)
-                m[i][j]=0;
+            m[i][lidy]=0;
     barrier(CLK_LOCAL_MEM_FENCE);
+
     long lm[10] = {0};
     int8 x0 = (int8)(0);
     int8 x1 = (int8)(0);
@@ -386,79 +436,75 @@ __kernel void CvMoments_D2(__global ushort8* src_data, int src_rows, int src_col
         lm[0] = x0.s0;             // m00
     }
     barrier(CLK_LOCAL_MEM_FENCE);
+
     for( int j = TILE_SIZE/2; j >= 1; j = j/2 )
     {
         if(lidy < j)
             for( int i = 0; i < 10; i++ )
                 lm[i] = lm[i] + m[i][lidy];
-        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    for( int j = TILE_SIZE/2; j >= 1; j = j/2 )
+    {
         if(lidy >= j/2&&lidy < j)
             for( int i = 0; i < 10; i++ )
                 m[i][lidy-j/2] = lm[i];
-        barrier(CLK_LOCAL_MEM_FENCE);
     }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
     if(lidy == 0&&lidx == 0)
     {
         for(int mt = 0; mt < 10; mt++ )
-            mom[mt] = (double)lm[mt];
+            mom[mt] = (F)lm[mt];
 
         if(binary)
         {
-            double s = 1./255;
+            F s = 1./255;
             for( int mt = 0; mt < 10; mt++ )
                 mom[mt] *= s;
         }
 
-        double xm = x  *mom[0], ym = y * mom[0];
+        F xm = x  *mom[0], ym = y * mom[0];
 
         // accumulate moments computed in each tile
+        dst_step /= sizeof(F);
 
         // + m00 ( = m00' )
-        dst_m00[wgidy*dst_cols+wgidx] = mom[0];
+        *(dst_m + mad24(DST_ROW_00 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[0];
 
         // + m10 ( = m10' + x*m00' )
-        dst_m10[wgidy*dst_cols+wgidx]  = mom[1] + xm;
+        *(dst_m + mad24(DST_ROW_10 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[1] + xm;
 
         // + m01 ( = m01' + y*m00' )
-        dst_m01[wgidy*dst_cols+wgidx]  = mom[2] + ym;
+        *(dst_m + mad24(DST_ROW_01 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[2] + ym;
 
         // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
-        dst_m20[wgidy*dst_cols+wgidx]  = mom[3] + x * (mom[1] * 2 + xm);
+        *(dst_m + mad24(DST_ROW_20 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[3] + x * (mom[1] * 2 + xm);
 
         // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
-        dst_m11[wgidy*dst_cols+wgidx]  = mom[4] + x * (mom[2] + ym) + y * mom[1];
+        *(dst_m + mad24(DST_ROW_11 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[4] + x * (mom[2] + ym) + y * mom[1];
 
         // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
-        dst_m02[wgidy*dst_cols+wgidx]  = mom[5] + y * (mom[2] * 2 + ym);
+        *(dst_m + mad24(DST_ROW_02 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[5] + y * (mom[2] * 2 + ym);
 
         // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
-        dst_m30[wgidy*dst_cols+wgidx]  = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
+        *(dst_m + mad24(DST_ROW_30 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
 
         // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
-        dst_m21[wgidy*dst_cols+wgidx] = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
+        *(dst_m + mad24(DST_ROW_21 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
 
         // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
-        dst_m12[wgidy*dst_cols+wgidx]  = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
+        *(dst_m + mad24(DST_ROW_12 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
 
         // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
-        dst_m03[wgidy*dst_cols+wgidx]  = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
+        *(dst_m + mad24(DST_ROW_03 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
     }
 }
-//#endif
-//#if defined (DOUBLE_SUPPORT)
-__kernel void CvMoments_D3(__global short8* src_data, int src_rows, int src_cols, int src_step, int tileSize_width, int tileSize_height,
-                           __global double* dst_m00,
-                           __global double* dst_m10,
-                           __global double* dst_m01,
-                           __global double* dst_m20,
-                           __global double* dst_m11,
-                           __global double* dst_m02,
-                           __global double* dst_m30,
-                           __global double* dst_m21,
-                           __global double* dst_m12,
-                           __global double* dst_m03,
-                           int dst_cols, int dst_step,
-                           int type, int depth, int cn, int coi, int binary, const int TILE_SIZE)
+
+__kernel void CvMoments_D3(__global short8* src_data, int src_rows, int src_cols, int src_step,
+                           __global F* dst_m,
+                           int dst_cols, int dst_step, int blocky,
+                           int depth, int cn, int coi, int binary, const int TILE_SIZE)
 {
     short tmp_coi[8]; // get the coi data
     short8 tmp[32];
@@ -473,33 +519,37 @@ __kernel void CvMoments_D3(__global short8* src_data, int src_rows, int src_cols
     int x = wgidx*TILE_SIZE;  // real X index of pixel
     int kcn = (cn==2)?2:4;
     int rstep = min(src_step/2, TILE_SIZE);
-    tileSize_height = min(TILE_SIZE, src_rows - y);
-    tileSize_width = min(TILE_SIZE, src_cols -x);
-    if(tileSize_width < TILE_SIZE)
-        for(int i = tileSize_width; i < rstep; i++ )
-            *((__global short*)src_data+(y+lidy)*src_step/2+x+i) = 0;
-    if( coi > 0 )
-        for(int i=0; i < tileSize_width; i+=VLEN_S)
-        {
-            for(int j=0; j<VLEN_S; j++)
-                tmp_coi[j] = *((__global short*)src_data+(y+lidy)*src_step/2+(x+i+j)*kcn+coi-1);
-            tmp[i/VLEN_S] = (short8)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7]);
-        }
-    else
-        for(int i=0; i < tileSize_width; i+=VLEN_S)
-            tmp[i/VLEN_S] = *(src_data+(y+lidy)*src_step/(2*VLEN_S)+(x+i)/VLEN_S);
+    int tileSize_height = min(TILE_SIZE, src_rows - y);
+    int tileSize_width = min(TILE_SIZE, src_cols -x);
+
+    if ( y+lidy < src_rows )
+    {
+        if(tileSize_width < TILE_SIZE)
+            for(int i = tileSize_width; i < rstep && (x+i) < src_cols; i++ )
+                *((__global short*)src_data+(y+lidy)*src_step/2+x+i) = 0;
+        if( coi > 0 )
+            for(int i=0; i < tileSize_width; i+=VLEN_S)
+            {
+                for(int j=0; j<VLEN_S; j++)
+                    tmp_coi[j] = *((__global short*)src_data+(y+lidy)*src_step/2+(x+i+j)*kcn+coi-1);
+                tmp[i/VLEN_S] = (short8)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3],tmp_coi[4],tmp_coi[5],tmp_coi[6],tmp_coi[7]);
+            }
+        else
+            for(int i=0; i < tileSize_width; i+=VLEN_S)
+                tmp[i/VLEN_S] = *(src_data+(y+lidy)*src_step/(2*VLEN_S)+(x+i)/VLEN_S);
+    }
+
     short8 zero = (short8)(0);
     short8 full = (short8)(255);
     if( binary )
         for(int i=0; i < tileSize_width; i+=(VLEN_S))
             tmp[i/VLEN_S] = (tmp[i/VLEN_S]!=zero)?full:zero;
 
-    double mom[10];
+    F mom[10];
     __local long m[10][128];
-    if(lidy == 0)
+    if(lidy < 128)
         for(int i=0; i<10; i++)
-            for(int j=0; j<128; j++)
-                m[i][j]=0;
+            m[i][lidy]=0;
     barrier(CLK_LOCAL_MEM_FENCE);
     long lm[10] = {0};
     int8 x0 = (int8)(0);
@@ -564,65 +614,56 @@ __kernel void CvMoments_D3(__global short8* src_data, int src_rows, int src_cols
     if(lidy ==0 &&lidx ==0)
     {
         for(int mt = 0; mt < 10; mt++ )
-            mom[mt] = (double)lm[mt];
+            mom[mt] = (F)lm[mt];
 
         if(binary)
         {
-            double s = 1./255;
+            F s = 1./255;
             for( int mt = 0; mt < 10; mt++ )
                 mom[mt] *= s;
         }
 
-        double xm = x * mom[0], ym = y*mom[0];
+        F xm = x * mom[0], ym = y*mom[0];
 
         // accumulate moments computed in each tile
+        dst_step /= sizeof(F);
 
         // + m00 ( = m00' )
-        dst_m00[wgidy*dst_cols+wgidx]  = mom[0];
+        *(dst_m + mad24(DST_ROW_00 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[0];
 
         // + m10 ( = m10' + x*m00' )
-        dst_m10[wgidy*dst_cols+wgidx]  = mom[1] + xm;
+        *(dst_m + mad24(DST_ROW_10 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[1] + xm;
 
         // + m01 ( = m01' + y*m00' )
-        dst_m01[wgidy*dst_cols+wgidx]  = mom[2] + ym;
+        *(dst_m + mad24(DST_ROW_01 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[2] + ym;
 
         // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
-        dst_m20[wgidy*dst_cols+wgidx]  = mom[3] + x * (mom[1] * 2 + xm);
+        *(dst_m + mad24(DST_ROW_20 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[3] + x * (mom[1] * 2 + xm);
 
         // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
-        dst_m11[wgidy*dst_cols+wgidx]   = mom[4] + x * (mom[2] + ym) + y * mom[1];
+        *(dst_m + mad24(DST_ROW_11 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[4] + x * (mom[2] + ym) + y * mom[1];
 
         // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
-        dst_m02[wgidy*dst_cols+wgidx]   = mom[5] + y * (mom[2] * 2 + ym);
+        *(dst_m + mad24(DST_ROW_02 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[5] + y * (mom[2] * 2 + ym);
 
         // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
-        dst_m30[wgidy*dst_cols+wgidx]  = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
+        *(dst_m + mad24(DST_ROW_30 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
 
         // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
-        dst_m21[wgidy*dst_cols+wgidx]  = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
+        *(dst_m + mad24(DST_ROW_21 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
 
         // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
-        dst_m12[wgidy*dst_cols+wgidx]  = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
+        *(dst_m + mad24(DST_ROW_12 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
 
         // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
-        dst_m03[wgidy*dst_cols+wgidx]  = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
+        *(dst_m + mad24(DST_ROW_03 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
     }
 }
-//#endif
-//#if defined (DOUBLE_SUPPORT)
-__kernel void CvMoments_D5( __global float* src_data, int src_rows, int src_cols, int src_step, int tileSize_width, int tileSize_height,
-                            __global double* dst_m00,
-                            __global double* dst_m10,
-                            __global double* dst_m01,
-                            __global double* dst_m20,
-                            __global double* dst_m11,
-                            __global double* dst_m02,
-                            __global double* dst_m30,
-                            __global double* dst_m21,
-                            __global double* dst_m12,
-                            __global double* dst_m03,
-                            int dst_cols, int dst_step,
-                            int type, int depth, int cn, int coi, int binary, const int TILE_SIZE)
+
+__kernel void CvMoments_D5( __global float* src_data, int src_rows, int src_cols, int src_step,
+                            __global F* dst_m,
+                            int dst_cols, int dst_step, int blocky,
+                            int depth, int cn, int coi, int binary, const int TILE_SIZE)
 {
     float tmp_coi[4]; // get the coi data
     float4 tmp[64] ;
@@ -637,43 +678,50 @@ __kernel void CvMoments_D5( __global float* src_data, int src_rows, int src_cols
     int x = wgidx*TILE_SIZE;  // real X index of pixel
     int kcn = (cn==2)?2:4;
     int rstep = min(src_step/4, TILE_SIZE);
-    tileSize_height = min(TILE_SIZE, src_rows - y);
-    tileSize_width = min(TILE_SIZE, src_cols -x);
-    if(tileSize_width < TILE_SIZE)
-        for(int i = tileSize_width; i < rstep; i++ )
-            *((__global float*)src_data+(y+lidy)*src_step/4+x+i) = 0;
-    if( coi > 0 )
-        for(int i=0; i < tileSize_width; i+=VLEN_F)
-        {
-            for(int j=0; j<4; j++)
-                tmp_coi[j] = *(src_data+(y+lidy)*src_step/4+(x+i+j)*kcn+coi-1);
-            tmp[i/VLEN_F] = (float4)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3]);
-        }
-    else
-        for(int i=0; i < tileSize_width; i+=VLEN_F)
-            tmp[i/VLEN_F] = (float4)(*(src_data+(y+lidy)*src_step/4+x+i),*(src_data+(y+lidy)*src_step/4+x+i+1),*(src_data+(y+lidy)*src_step/4+x+i+2),*(src_data+(y+lidy)*src_step/4+x+i+3));
+    int tileSize_height = min(TILE_SIZE, src_rows - y);
+    int tileSize_width = min(TILE_SIZE, src_cols -x);
+    int maxIdx = mul24(src_rows, src_cols);
+    int yOff = (y+lidy)*src_step;
+    int index;
+
+    if ( y+lidy < src_rows )
+    {
+        if(tileSize_width < TILE_SIZE)
+            for(int i = tileSize_width; i < rstep && (x+i) < src_cols; i++ )
+                *((__global float*)src_data+(y+lidy)*src_step/4+x+i) = 0;
+        if( coi > 0 )
+            for(int i=0; i < tileSize_width; i+=VLEN_F)
+            {
+                for(int j=0; j<4; j++)
+                    tmp_coi[j] = *(src_data+(y+lidy)*src_step/4+(x+i+j)*kcn+coi-1);
+                tmp[i/VLEN_F] = (float4)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3]);
+            }
+        else
+            for(int i=0; i < tileSize_width; i+=VLEN_F)
+                tmp[i/VLEN_F] = (float4)(*(src_data+(y+lidy)*src_step/4+x+i),*(src_data+(y+lidy)*src_step/4+x+i+1),*(src_data+(y+lidy)*src_step/4+x+i+2),*(src_data+(y+lidy)*src_step/4+x+i+3));
+    }
+
     float4 zero = (float4)(0);
     float4 full = (float4)(255);
     if( binary )
         for(int i=0; i < tileSize_width; i+=4)
             tmp[i/VLEN_F] = (tmp[i/VLEN_F]!=zero)?full:zero;
-    double mom[10];
-    __local double m[10][128];
-    if(lidy == 0)
+    F mom[10];
+    __local F m[10][128];
+    if(lidy < 128)
         for(int i = 0; i < 10; i ++)
-            for(int j = 0; j < 128; j ++)
-                m[i][j] = 0;
+            m[i][lidy] = 0;
     barrier(CLK_LOCAL_MEM_FENCE);
-    double lm[10] = {0};
-    double4 x0 = (double4)(0);
-    double4 x1 = (double4)(0);
-    double4 x2 = (double4)(0);
-    double4 x3 = (double4)(0);
+    F lm[10] = {0};
+    F4 x0 = (F4)(0);
+    F4 x1 = (F4)(0);
+    F4 x2 = (F4)(0);
+    F4 x3 = (F4)(0);
     for( int xt = 0 ; xt < tileSize_width; xt+=VLEN_F )
     {
-        double4 v_xt = (double4)(xt, xt+1, xt+2, xt+3);
-        double4 p = convert_double4(tmp[xt/VLEN_F]);
-        double4 xp = v_xt * p, xxp = xp * v_xt;
+        F4 v_xt = (F4)(xt, xt+1, xt+2, xt+3);
+        F4 p = convert_F4(tmp[xt/VLEN_F]);
+        F4 xp = v_xt * p, xxp = xp * v_xt;
         x0 += p;
         x1 += xp;
         x2 += xxp;
@@ -683,178 +731,14 @@ __kernel void CvMoments_D5( __global float* src_data, int src_rows, int src_cols
     x1.s0 += x1.s1 + x1.s2 + x1.s3;
     x2.s0 += x2.s1 + x2.s2 + x2.s3;
     x3.s0 += x3.s1 + x3.s2 + x3.s3;
-/*
-    double py = lidy * x0.s0, sy = lidy*lidy;
+
+    F py = lidy * x0.s0, sy = lidy*lidy;
     int bheight = min(tileSize_height, TILE_SIZE/2);
     if(bheight >= TILE_SIZE/2&&lidy > bheight-1&&lidy < tileSize_height)
     {
-        m[9][lidy-bheight] = ((double)py) * sy;  // m03
-        m[8][lidy-bheight] = ((double)x1.s0) * sy;  // m12
-        m[7][lidy-bheight] = ((double)x2.s0) * lidy;  // m21
-        m[6][lidy-bheight] = x3.s0;             // m30
-        m[5][lidy-bheight] = x0.s0 * sy;        // m02
-        m[4][lidy-bheight] = x1.s0 * lidy;         // m11
-        m[3][lidy-bheight] = x2.s0;             // m20
-        m[2][lidy-bheight] = py;             // m01
-        m[1][lidy-bheight] = x1.s0;             // m10
-        m[0][lidy-bheight] = x0.s0;             // m00
-    }
-    else if(lidy < bheight)
-    {
-        lm[9] = ((double)py) * sy;  // m03
-        lm[8] = ((double)x1.s0) * sy;  // m12
-        lm[7] = ((double)x2.s0) * lidy;  // m21
-        lm[6] = x3.s0;             // m30
-        lm[5] = x0.s0 * sy;        // m02
-        lm[4] = x1.s0 * lidy;         // m11
-        lm[3] = x2.s0;             // m20
-        lm[2] = py;             // m01
-        lm[1] = x1.s0;             // m10
-        lm[0] = x0.s0;             // m00
-    }
-    barrier(CLK_LOCAL_MEM_FENCE);
-    for( int j = TILE_SIZE/2; j >= 1; j = j/2 )
-    {
-        if(lidy < j)
-            for( int i = 0; i < 10; i++ )
-                lm[i] = lm[i] + m[i][lidy];
-        barrier(CLK_LOCAL_MEM_FENCE);
-        if(lidy >= j/2&&lidy < j)
-            for( int i = 0; i < 10; i++ )
-                m[i][lidy-j/2] = lm[i];
-        barrier(CLK_LOCAL_MEM_FENCE);
-    }
-    if(lidy == 0&&lidx == 0)
-    {
-        for(int mt = 0; mt < 10; mt++ )
-            mom[mt] = (double)lm[mt];
-
-        if(binary)
-        {
-            double s = 1./255;
-            for( int mt = 0; mt < 10; mt++ )
-                mom[mt] *= s;
-        }
-
-        double xm = x * mom[0], ym = y * mom[0];
-
-        // accumulate moments computed in each tile
-
-        // + m00 ( = m00' )
-        dst_m00[wgidy*dst_cols+wgidx]= mom[0];
-
-        // + m10 ( = m10' + x*m00' )
-        dst_m10[wgidy*dst_cols+wgidx] = mom[1] + xm;
-
-        // + m01 ( = m01' + y*m00' )
-        dst_m01[wgidy*dst_cols+wgidx] = mom[2] + ym;
-
-        // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
-        dst_m20[wgidy*dst_cols+wgidx] = mom[3] + x * (mom[1] * 2 + xm);
-
-        // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
-        dst_m11[wgidy*dst_cols+wgidx] = mom[4] + x * (mom[2] + ym) + y * mom[1];
-
-        // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
-        dst_m02[wgidy*dst_cols+wgidx]= mom[5] + y * (mom[2] * 2 + ym);
-
-        // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
-        dst_m30[wgidy*dst_cols+wgidx]= mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
-
-        // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
-        dst_m21[wgidy*dst_cols+wgidx] = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
-
-        // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
-        dst_m12[wgidy*dst_cols+wgidx] = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
-
-        // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
-        dst_m03[wgidy*dst_cols+wgidx]= mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
-    }*/
-}
-//#endif
-//#if defined (DOUBLE_SUPPORT)
-__kernel void CvMoments_D6(__global double* src_data,  int src_rows, int src_cols, int src_step, int tileSize_width, int tileSize_height,
-                           __global double* dst_m00,
-                           __global double* dst_m10,
-                           __global double* dst_m01,
-                           __global double* dst_m20,
-                           __global double* dst_m11,
-                           __global double* dst_m02,
-                           __global double* dst_m30,
-                           __global double* dst_m21,
-                           __global double* dst_m12,
-                           __global double* dst_m03,
-                           int dst_cols, int dst_step,
-                           int type, int depth, int cn, int coi, int binary, const int TILE_SIZE)
-{
-    double tmp_coi[4]; // get the coi data
-    double4 tmp[64];
-    int VLEN_D = 4; // length of vetor
-    int gidy = get_global_id(0);
-    int gidx = get_global_id(1);
-    int wgidy = get_group_id(0);
-    int wgidx = get_group_id(1);
-    int lidy = get_local_id(0);
-    int lidx = get_local_id(1);
-    int y = wgidy*TILE_SIZE;  // real Y index of pixel
-    int x = wgidx*TILE_SIZE;  // real X index of pixel
-    int kcn = (cn==2)?2:4;
-    int rstep = min(src_step/8, TILE_SIZE);
-    tileSize_height = min(TILE_SIZE,  src_rows - y);
-    tileSize_width = min(TILE_SIZE, src_cols - x);
-
-    if(tileSize_width < TILE_SIZE)
-        for(int i = tileSize_width; i < rstep; i++ )
-            *((__global double*)src_data+(y+lidy)*src_step/8+x+i) = 0;
-    if( coi > 0 )
-        for(int i=0; i < tileSize_width; i+=VLEN_D)
-        {
-            for(int j=0; j<4; j++)
-                tmp_coi[j] = *(src_data+(y+lidy)*src_step/8+(x+i+j)*kcn+coi-1);
-            tmp[i/VLEN_D] = (double4)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3]);
-        }
-    else
-        for(int i=0; i < tileSize_width; i+=VLEN_D)
-            tmp[i/VLEN_D] = (double4)(*(src_data+(y+lidy)*src_step/8+x+i),*(src_data+(y+lidy)*src_step/8+x+i+1),*(src_data+(y+lidy)*src_step/8+x+i+2),*(src_data+(y+lidy)*src_step/8+x+i+3));
-    double4 zero = (double4)(0);
-    double4 full = (double4)(255);
-    if( binary )
-        for(int i=0; i < tileSize_width; i+=VLEN_D)
-            tmp[i/VLEN_D] = (tmp[i/VLEN_D]!=zero)?full:zero;
-    double mom[10];
-    __local double m[10][128];
-    if(lidy == 0)
-        for(int i=0; i<10; i++)
-            for(int j=0; j<128; j++)
-                m[i][j]=0;
-    barrier(CLK_LOCAL_MEM_FENCE);
-    double lm[10] = {0};
-    double4 x0 = (double4)(0);
-    double4 x1 = (double4)(0);
-    double4 x2 = (double4)(0);
-    double4 x3 = (double4)(0);
-    for( int xt = 0 ; xt < tileSize_width; xt+=VLEN_D )
-    {
-        double4 v_xt = (double4)(xt, xt+1, xt+2, xt+3);
-        double4 p = tmp[xt/VLEN_D];
-        double4 xp = v_xt * p, xxp = xp * v_xt;
-        x0 += p;
-        x1 += xp;
-        x2 += xxp;
-        x3 += xxp *v_xt;
-    }
-    x0.s0 += x0.s1 + x0.s2 + x0.s3;
-    x1.s0 += x1.s1 + x1.s2 + x1.s3;
-    x2.s0 += x2.s1 + x2.s2 + x2.s3;
-    x3.s0 += x3.s1 + x3.s2 + x3.s3;
-
-    double py = lidy * x0.s0, sy = lidy*lidy;
-    int bheight = min(tileSize_height, TILE_SIZE/2);
-    if(bheight >= TILE_SIZE/2&&lidy > bheight-1&&lidy < tileSize_height)
-    {
-        m[9][lidy-bheight] = ((double)py) * sy;  // m03
-        m[8][lidy-bheight] = ((double)x1.s0) * sy;  // m12
-        m[7][lidy-bheight] = ((double)x2.s0) * lidy;  // m21
+        m[9][lidy-bheight] = ((F)py) * sy;  // m03
+        m[8][lidy-bheight] = ((F)x1.s0) * sy;  // m12
+        m[7][lidy-bheight] = ((F)x2.s0) * lidy;  // m21
         m[6][lidy-bheight] = x3.s0;             // m30
         m[5][lidy-bheight] = x0.s0 * sy;        // m02
         m[4][lidy-bheight] = x1.s0 * lidy;         // m11
@@ -866,9 +750,9 @@ __kernel void CvMoments_D6(__global double* src_data,  int src_rows, int src_col
 
     else if(lidy < bheight)
     {
-        lm[9] = ((double)py) * sy;  // m03
-        lm[8] = ((double)x1.s0) * sy;  // m12
-        lm[7] = ((double)x2.s0) * lidy;  // m21
+        lm[9] = ((F)py) * sy;  // m03
+        lm[8] = ((F)x1.s0) * sy;  // m12
+        lm[7] = ((F)x2.s0) * lidy;  // m21
         lm[6] = x3.s0;             // m30
         lm[5] = x0.s0 * sy;        // m02
         lm[4] = x1.s0 * lidy;         // m11
@@ -892,47 +776,205 @@ __kernel void CvMoments_D6(__global double* src_data,  int src_rows, int src_col
     if(lidy == 0&&lidx == 0)
     {
         for( int mt = 0; mt < 10; mt++ )
-            mom[mt] = (double)lm[mt];
+            mom[mt] = (F)lm[mt];
         if(binary)
         {
-            double s = 1./255;
+            F s = 1./255;
             for( int mt = 0; mt < 10; mt++ )
                 mom[mt] *= s;
         }
 
-        double xm = x * mom[0], ym = y * mom[0];
+        F xm = x * mom[0], ym = y * mom[0];
 
         // accumulate moments computed in each tile
+        dst_step /= sizeof(F);
 
         // + m00 ( = m00' )
-        dst_m00[wgidy*dst_cols+wgidx] = mom[0];
+        *(dst_m + mad24(DST_ROW_00 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[0];
 
         // + m10 ( = m10' + x*m00' )
-        dst_m10[wgidy*dst_cols+wgidx] = mom[1] + xm;
+        *(dst_m + mad24(DST_ROW_10 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[1] + xm;
 
         // + m01 ( = m01' + y*m00' )
-        dst_m01[wgidy*dst_cols+wgidx] = mom[2] + ym;
+        *(dst_m + mad24(DST_ROW_01 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[2] + ym;
 
         // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
-        dst_m20[wgidy*dst_cols+wgidx]  = mom[3] + x * (mom[1] * 2 + xm);
+        *(dst_m + mad24(DST_ROW_20 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[3] + x * (mom[1] * 2 + xm);
 
         // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
-        dst_m11[wgidy*dst_cols+wgidx]  = mom[4] + x * (mom[2] + ym) + y * mom[1];
+        *(dst_m + mad24(DST_ROW_11 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[4] + x * (mom[2] + ym) + y * mom[1];
 
         // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
-        dst_m02[wgidy*dst_cols+wgidx]  = mom[5] + y * (mom[2] * 2 + ym);
+        *(dst_m + mad24(DST_ROW_02 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[5] + y * (mom[2] * 2 + ym);
 
         // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
-        dst_m30[wgidy*dst_cols+wgidx]  = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
+        *(dst_m + mad24(DST_ROW_30 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
 
         // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
-        dst_m21[wgidy*dst_cols+wgidx]  = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
+        *(dst_m + mad24(DST_ROW_21 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
 
         // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
-        dst_m12[wgidy*dst_cols+wgidx]  = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
+        *(dst_m + mad24(DST_ROW_12 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
 
         // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
-        dst_m03[wgidy*dst_cols+wgidx]  = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
+        *(dst_m + mad24(DST_ROW_03 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
+    }
+}
+
+__kernel void CvMoments_D6(__global F* src_data,  int src_rows, int src_cols, int src_step,
+                           __global F* dst_m,
+                           int dst_cols, int dst_step, int blocky,
+                           int depth, int cn, int coi, int binary, const int TILE_SIZE)
+{
+    F tmp_coi[4]; // get the coi data
+    F4 tmp[64];
+    int VLEN_D = 4; // length of vetor
+    int gidy = get_global_id(0);
+    int gidx = get_global_id(1);
+    int wgidy = get_group_id(0);
+    int wgidx = get_group_id(1);
+    int lidy = get_local_id(0);
+    int lidx = get_local_id(1);
+    int y = wgidy*TILE_SIZE;  // real Y index of pixel
+    int x = wgidx*TILE_SIZE;  // real X index of pixel
+    int kcn = (cn==2)?2:4;
+    int rstep = min(src_step/8, TILE_SIZE);
+    int tileSize_height = min(TILE_SIZE,  src_rows - y);
+    int tileSize_width = min(TILE_SIZE, src_cols - x);
+
+    if ( y+lidy < src_rows )
+    {
+        if(tileSize_width < TILE_SIZE)
+            for(int i = tileSize_width; i < rstep && (x+i) < src_cols; i++ )
+                *((__global F*)src_data+(y+lidy)*src_step/8+x+i) = 0;
+        if( coi > 0 )
+            for(int i=0; i < tileSize_width; i+=VLEN_D)
+            {
+                for(int j=0; j<4 && ((x+i+j)*kcn+coi-1)<src_cols; j++)
+                    tmp_coi[j] = *(src_data+(y+lidy)*src_step/8+(x+i+j)*kcn+coi-1);
+                tmp[i/VLEN_D] = (F4)(tmp_coi[0],tmp_coi[1],tmp_coi[2],tmp_coi[3]);
+            }
+        else
+            for(int i=0; i < tileSize_width && (x+i+3) < src_cols; i+=VLEN_D)
+                tmp[i/VLEN_D] = (F4)(*(src_data+(y+lidy)*src_step/8+x+i),*(src_data+(y+lidy)*src_step/8+x+i+1),*(src_data+(y+lidy)*src_step/8+x+i+2),*(src_data+(y+lidy)*src_step/8+x+i+3));
+    }
+
+    F4 zero = (F4)(0);
+    F4 full = (F4)(255);
+    if( binary )
+        for(int i=0; i < tileSize_width; i+=VLEN_D)
+            tmp[i/VLEN_D] = (tmp[i/VLEN_D]!=zero)?full:zero;
+    F mom[10];
+    __local F m[10][128];
+    if(lidy < 128)
+        for(int i=0; i<10; i++)
+            m[i][lidy]=0;
+    barrier(CLK_LOCAL_MEM_FENCE);
+    F lm[10] = {0};
+    F4 x0 = (F4)(0);
+    F4 x1 = (F4)(0);
+    F4 x2 = (F4)(0);
+    F4 x3 = (F4)(0);
+    for( int xt = 0 ; xt < tileSize_width; xt+=VLEN_D )
+    {
+        F4 v_xt = (F4)(xt, xt+1, xt+2, xt+3);
+        F4 p = tmp[xt/VLEN_D];
+        F4 xp = v_xt * p, xxp = xp * v_xt;
+        x0 += p;
+        x1 += xp;
+        x2 += xxp;
+        x3 += xxp *v_xt;
+    }
+    x0.s0 += x0.s1 + x0.s2 + x0.s3;
+    x1.s0 += x1.s1 + x1.s2 + x1.s3;
+    x2.s0 += x2.s1 + x2.s2 + x2.s3;
+    x3.s0 += x3.s1 + x3.s2 + x3.s3;
+
+    F py = lidy * x0.s0, sy = lidy*lidy;
+    int bheight = min(tileSize_height, TILE_SIZE/2);
+    if(bheight >= TILE_SIZE/2&&lidy > bheight-1&&lidy < tileSize_height)
+    {
+        m[9][lidy-bheight] = ((F)py) * sy;  // m03
+        m[8][lidy-bheight] = ((F)x1.s0) * sy;  // m12
+        m[7][lidy-bheight] = ((F)x2.s0) * lidy;  // m21
+        m[6][lidy-bheight] = x3.s0;             // m30
+        m[5][lidy-bheight] = x0.s0 * sy;        // m02
+        m[4][lidy-bheight] = x1.s0 * lidy;         // m11
+        m[3][lidy-bheight] = x2.s0;             // m20
+        m[2][lidy-bheight] = py;             // m01
+        m[1][lidy-bheight] = x1.s0;             // m10
+        m[0][lidy-bheight] = x0.s0;             // m00
+    }
+    else if(lidy < bheight)
+    {
+        lm[9] = ((F)py) * sy;  // m03
+        lm[8] = ((F)x1.s0) * sy;  // m12
+        lm[7] = ((F)x2.s0) * lidy;  // m21
+        lm[6] = x3.s0;             // m30
+        lm[5] = x0.s0 * sy;        // m02
+        lm[4] = x1.s0 * lidy;         // m11
+        lm[3] = x2.s0;             // m20
+        lm[2] = py;             // m01
+        lm[1] = x1.s0;             // m10
+        lm[0] = x0.s0;             // m00
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for( int j = TILE_SIZE/2; j >= 1; j = j/2 )
+    {
+        if(lidy < j)
+            for( int i = 0; i < 10; i++ )
+                lm[i] = lm[i] + m[i][lidy];
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if(lidy >= j/2&&lidy < j)
+            for( int i = 0; i < 10; i++ )
+                m[i][lidy-j/2] = lm[i];
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+    if(lidy == 0&&lidx == 0)
+    {
+        for( int mt = 0; mt < 10; mt++ )
+            mom[mt] = (F)lm[mt];
+        if(binary)
+        {
+            F s = 1./255;
+            for( int mt = 0; mt < 10; mt++ )
+                mom[mt] *= s;
+        }
+
+        F xm = x * mom[0], ym = y * mom[0];
+
+        // accumulate moments computed in each tile
+        dst_step /= sizeof(F);
+
+        // + m00 ( = m00' )
+        *(dst_m + mad24(DST_ROW_00 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[0];
+
+        // + m10 ( = m10' + x*m00' )
+        *(dst_m + mad24(DST_ROW_10 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[1] + xm;
+
+        // + m01 ( = m01' + y*m00' )
+        *(dst_m + mad24(DST_ROW_01 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[2] + ym;
+
+        // + m20 ( = m20' + 2*x*m10' + x*x*m00' )
+        *(dst_m + mad24(DST_ROW_20 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[3] + x * (mom[1] * 2 + xm);
+
+        // + m11 ( = m11' + x*m01' + y*m10' + x*y*m00' )
+        *(dst_m + mad24(DST_ROW_11 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[4] + x * (mom[2] + ym) + y * mom[1];
+
+        // + m02 ( = m02' + 2*y*m01' + y*y*m00' )
+        *(dst_m + mad24(DST_ROW_02 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[5] + y * (mom[2] * 2 + ym);
+
+        // + m30 ( = m30' + 3*x*m20' + 3*x*x*m10' + x*x*x*m00' )
+        *(dst_m + mad24(DST_ROW_30 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[6] + x * (3. * mom[3] + x * (3. * mom[1] + xm));
+
+        // + m21 ( = m21' + x*(2*m11' + 2*y*m10' + x*m01' + x*y*m00') + y*m20')
+        *(dst_m + mad24(DST_ROW_21 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[7] + x * (2 * (mom[4] + y * mom[1]) + x * (mom[2] + ym)) + y * mom[3];
+
+        // + m12 ( = m12' + y*(2*m11' + 2*x*m01' + y*m10' + x*y*m00') + x*m02')
+        *(dst_m + mad24(DST_ROW_12 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[8] + y * (2 * (mom[4] + x * mom[2]) + y * (mom[1] + xm)) + x * mom[5];
+
+        // + m03 ( = m03' + 3*y*m02' + 3*y*y*m01' + y*y*y*m00' )
+        *(dst_m + mad24(DST_ROW_03 * blocky, dst_step, mad24(wgidy, dst_cols, wgidx))) = mom[9] + y * (3. * mom[5] + y * (3. * mom[2] + ym));
     }
 }
-//#endif
diff --git a/modules/ocl/src/opencl/objdetect_hog.cl b/modules/ocl/src/opencl/objdetect_hog.cl
new file mode 100644
index 0000000000..685eccf688
--- /dev/null
+++ b/modules/ocl/src/opencl/objdetect_hog.cl
@@ -0,0 +1,720 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Wenju He, wenju@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#define CELL_WIDTH 8
+#define CELL_HEIGHT 8
+#define CELLS_PER_BLOCK_X 2
+#define CELLS_PER_BLOCK_Y 2
+#define NTHREADS 256
+#define CV_PI_F 3.1415926535897932384626433832795f
+
+//----------------------------------------------------------------------------
+// Histogram computation
+// 12 threads for a cell, 12x4 threads per block
+// Use pre-computed gaussian and interp_weight lookup tables
+__kernel void compute_hists_lut_kernel(
+    const int cblock_stride_x, const int cblock_stride_y,
+    const int cnbins, const int cblock_hist_size, const int img_block_width,
+    const int blocks_in_group, const int blocks_total,
+    const int grad_quadstep, const int qangle_step,
+    __global const float* grad, __global const uchar* qangle,
+    __global const float* gauss_w_lut,
+    __global float* block_hists, __local float* smem)
+{
+    const int lx = get_local_id(0);
+    const int lp = lx / 24; /* local group id */
+    const int gid = get_group_id(0) * blocks_in_group + lp;/* global group id */
+    const int gidY = gid / img_block_width;
+    const int gidX = gid - gidY * img_block_width;
+
+    const int lidX = lx - lp * 24;
+    const int lidY = get_local_id(1);
+
+    const int cell_x = lidX / 12;
+    const int cell_y = lidY;
+    const int cell_thread_x = lidX - cell_x * 12;
+
+    __local float* hists = smem + lp * cnbins * (CELLS_PER_BLOCK_X *
+        CELLS_PER_BLOCK_Y * 12 + CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y);
+    __local float* final_hist = hists + cnbins *
+        (CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y * 12);
+
+    const int offset_x = gidX * cblock_stride_x + (cell_x << 2) + cell_thread_x;
+    const int offset_y = gidY * cblock_stride_y + (cell_y << 2);
+
+    __global const float* grad_ptr = (gid < blocks_total) ?
+        grad + offset_y * grad_quadstep + (offset_x << 1) : grad;
+    __global const uchar* qangle_ptr = (gid < blocks_total) ?
+        qangle + offset_y * qangle_step + (offset_x << 1) : qangle;
+
+    __local float* hist = hists + 12 * (cell_y * CELLS_PER_BLOCK_Y + cell_x) +
+        cell_thread_x;
+    for (int bin_id = 0; bin_id < cnbins; ++bin_id)
+        hist[bin_id * 48] = 0.f;
+
+    const int dist_x = -4 + cell_thread_x - 4 * cell_x;
+    const int dist_center_x = dist_x - 4 * (1 - 2 * cell_x);
+
+    const int dist_y_begin = -4 - 4 * lidY;
+    for (int dist_y = dist_y_begin; dist_y < dist_y_begin + 12; ++dist_y)
+    {
+        float2 vote = (float2) (grad_ptr[0], grad_ptr[1]);
+        uchar2 bin = (uchar2) (qangle_ptr[0], qangle_ptr[1]);
+
+        grad_ptr += grad_quadstep;
+        qangle_ptr += qangle_step;
+
+        int dist_center_y = dist_y - 4 * (1 - 2 * cell_y);
+
+        int idx = (dist_center_y + 8) * 16 + (dist_center_x + 8);
+        float gaussian = gauss_w_lut[idx];
+        idx = (dist_y + 8) * 16 + (dist_x + 8);
+        float interp_weight = gauss_w_lut[256+idx];
+
+        hist[bin.x * 48] += gaussian * interp_weight * vote.x;
+        hist[bin.y * 48] += gaussian * interp_weight * vote.y;
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    volatile __local float* hist_ = hist;
+    for (int bin_id = 0; bin_id < cnbins; ++bin_id, hist_ += 48)
+    {
+        if (cell_thread_x < 6)
+            hist_[0] += hist_[6];
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if (cell_thread_x < 3)
+            hist_[0] += hist_[3];
+#ifdef CPU
+        barrier(CLK_LOCAL_MEM_FENCE);
+#endif
+        if (cell_thread_x == 0)
+            final_hist[(cell_x * 2 + cell_y) * cnbins + bin_id] =
+                hist_[0] + hist_[1] + hist_[2];
+    }
+#ifdef CPU
+    barrier(CLK_LOCAL_MEM_FENCE);
+#endif
+
+    int tid = (cell_y * CELLS_PER_BLOCK_Y + cell_x) * 12 + cell_thread_x;
+    if ((tid < cblock_hist_size) && (gid < blocks_total))
+    {
+        __global float* block_hist = block_hists +
+            (gidY * img_block_width + gidX) * cblock_hist_size;
+        block_hist[tid] = final_hist[tid];
+    }
+}
+
+//-------------------------------------------------------------
+//  Normalization of histograms via L2Hys_norm
+//  optimized for the case of 9 bins
+__kernel void normalize_hists_36_kernel(__global float* block_hists,
+                                        const float threshold, __local float *squares)
+{
+    const int tid = get_local_id(0);
+    const int gid = get_global_id(0);
+    const int bid = tid / 36;      /* block-hist id, (0 - 6) */
+    const int boffset = bid * 36;  /* block-hist offset in the work-group */
+    const int hid = tid - boffset; /* histogram bin id, (0 - 35) */
+
+    float elem = block_hists[gid];
+    squares[tid] = elem * elem;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local float* smem = squares + boffset;
+    float sum = smem[hid];
+    if (hid < 18)
+        smem[hid] = sum = sum + smem[hid + 18];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 9)
+        smem[hid] = sum = sum + smem[hid + 9];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 4)
+        smem[hid] = sum + smem[hid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = smem[0] + smem[1] + smem[2] + smem[3] + smem[8];
+
+    elem = elem / (sqrt(sum) + 3.6f);
+    elem = min(elem, threshold);
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    squares[tid] = elem * elem;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = smem[hid];
+    if (hid < 18)
+      smem[hid] = sum = sum + smem[hid + 18];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 9)
+        smem[hid] = sum = sum + smem[hid + 9];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 4)
+        smem[hid] = sum + smem[hid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = smem[0] + smem[1] + smem[2] + smem[3] + smem[8];
+
+    block_hists[gid] = elem / (sqrt(sum) + 1e-3f);
+}
+
+//-------------------------------------------------------------
+//  Normalization of histograms via L2Hys_norm
+//
+float reduce_smem(volatile __local float* smem, int size)
+{
+    unsigned int tid = get_local_id(0);
+    float sum = smem[tid];
+
+    if (size >= 512) { if (tid < 256) smem[tid] = sum = sum + smem[tid + 256];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 256) { if (tid < 128) smem[tid] = sum = sum + smem[tid + 128];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 128) { if (tid < 64) smem[tid] = sum = sum + smem[tid + 64];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+#ifdef CPU
+    if (size >= 64) { if (tid < 32) smem[tid] = sum = sum + smem[tid + 32];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 32) { if (tid < 16) smem[tid] = sum = sum + smem[tid + 16];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 16) { if (tid < 8) smem[tid] = sum = sum + smem[tid + 8];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 8) { if (tid < 4) smem[tid] = sum = sum + smem[tid + 4];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 4) { if (tid < 2) smem[tid] = sum = sum + smem[tid + 2];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 2) { if (tid < 1) smem[tid] = sum = sum + smem[tid + 1];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+#else
+    if (tid < 32)
+    {
+        if (size >= 64) smem[tid] = sum = sum + smem[tid + 32];
+#if WAVE_SIZE < 32
+    } barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16) {
+#endif
+        if (size >= 32) smem[tid] = sum = sum + smem[tid + 16];
+        if (size >= 16) smem[tid] = sum = sum + smem[tid + 8];
+        if (size >= 8) smem[tid] = sum = sum + smem[tid + 4];
+        if (size >= 4) smem[tid] = sum = sum + smem[tid + 2];
+        if (size >= 2) smem[tid] = sum = sum + smem[tid + 1];
+    }
+#endif
+
+    return sum;
+}
+
+__kernel void normalize_hists_kernel(
+    const int nthreads, const int block_hist_size, const int img_block_width,
+    __global float* block_hists, const float threshold, __local float *squares)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global float* hist = block_hists + (gidY * img_block_width + gidX) *
+        block_hist_size + tid;
+
+    float elem = 0.f;
+    if (tid < block_hist_size)
+        elem = hist[0];
+
+    squares[tid] = elem * elem;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    float sum = reduce_smem(squares, nthreads);
+
+    float scale = 1.0f / (sqrt(sum) + 0.1f * block_hist_size);
+    elem = min(elem * scale, threshold);
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    squares[tid] = elem * elem;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = reduce_smem(squares, nthreads);
+    scale = 1.0f / (sqrt(sum) + 1e-3f);
+
+    if (tid < block_hist_size)
+        hist[0] = elem * scale;
+}
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  48x96 window, 9 bins and default parameters
+//  180 threads, each thread corresponds to a bin in a row
+__kernel void classify_hists_180_kernel(
+    const int cdescr_width, const int cdescr_height, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+
+    for (int i = 0; i < cdescr_height; i++)
+    {
+        product += coefs[i * cdescr_width + tid] *
+            hist[i * img_block_width * cblock_hist_size + tid];
+    }
+
+    __local float products[180];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 90) products[tid] = product = product + products[tid + 90];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 45) products[tid] = product = product + products[tid + 45];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    volatile __local float* smem = products;
+#ifdef CPU
+    if (tid < 13) smem[tid] = product = product + smem[tid + 32];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16) smem[tid] = product = product + smem[tid + 16];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<8) smem[tid] = product = product + smem[tid + 8];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<4) smem[tid] = product = product + smem[tid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<2) smem[tid] = product = product + smem[tid + 2];
+    barrier(CLK_LOCAL_MEM_FENCE);
+#else
+    if (tid < 13)
+    {
+        smem[tid] = product = product + smem[tid + 32];
+    }
+#if WAVE_SIZE < 32
+    barrier(CLK_LOCAL_MEM_FENCE);
+#endif
+    if (tid < 16)
+    {
+        smem[tid] = product = product + smem[tid + 16];
+        smem[tid] = product = product + smem[tid + 8];
+        smem[tid] = product = product + smem[tid + 4];
+        smem[tid] = product = product + smem[tid + 2];
+    }
+#endif
+
+    if (tid == 0){
+        product = product + smem[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  64x128 window, 9 bins and default parameters
+//  256 threads, 252 of them are used
+__kernel void classify_hists_252_kernel(
+    const int cdescr_width, const int cdescr_height, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+    if (tid < cdescr_width)
+    {
+        for (int i = 0; i < cdescr_height; i++)
+            product += coefs[i * cdescr_width + tid] *
+                hist[i * img_block_width * cblock_hist_size + tid];
+    }
+
+    __local float products[NTHREADS];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 128) products[tid] = product = product + products[tid + 128];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 64) products[tid] = product = product + products[tid + 64];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    volatile __local float* smem = products;
+#ifdef CPU
+    if(tid<32) smem[tid] = product = product + smem[tid + 32];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<16) smem[tid] = product = product + smem[tid + 16];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<8) smem[tid] = product = product + smem[tid + 8];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<4) smem[tid] = product = product + smem[tid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<2) smem[tid] = product = product + smem[tid + 2];
+    barrier(CLK_LOCAL_MEM_FENCE);
+#else
+    if (tid < 32)
+    {
+        smem[tid] = product = product + smem[tid + 32];
+#if WAVE_SIZE < 32
+    } barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16) {
+#endif
+        smem[tid] = product = product + smem[tid + 16];
+        smem[tid] = product = product + smem[tid + 8];
+        smem[tid] = product = product + smem[tid + 4];
+        smem[tid] = product = product + smem[tid + 2];
+    }
+#endif
+    if (tid == 0){
+        product = product + smem[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  256 threads
+__kernel void classify_hists_kernel(
+    const int cdescr_size, const int cdescr_width, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int offset_y = i / cdescr_width;
+        int offset_x = i - offset_y * cdescr_width;
+        product += coefs[i] *
+            hist[offset_y * img_block_width * cblock_hist_size + offset_x];
+    }
+
+    __local float products[NTHREADS];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 128) products[tid] = product = product + products[tid + 128];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 64) products[tid] = product = product + products[tid + 64];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    volatile __local float* smem = products;
+#ifdef CPU
+    if(tid<32) smem[tid] = product = product + smem[tid + 32];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<16) smem[tid] = product = product + smem[tid + 16];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<8) smem[tid] = product = product + smem[tid + 8];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<4) smem[tid] = product = product + smem[tid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if(tid<2) smem[tid] = product = product + smem[tid + 2];
+    barrier(CLK_LOCAL_MEM_FENCE);
+#else
+    if (tid < 32)
+    {
+        smem[tid] = product = product + smem[tid + 32];
+#if WAVE_SIZE < 32
+    } barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16) {
+#endif
+        smem[tid] = product = product + smem[tid + 16];
+        smem[tid] = product = product + smem[tid + 8];
+        smem[tid] = product = product + smem[tid + 4];
+        smem[tid] = product = product + smem[tid + 2];
+    }
+#endif
+    if (tid == 0){
+        smem[tid] = product = product + smem[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//----------------------------------------------------------------------------
+// Extract descriptors
+
+__kernel void extract_descrs_by_rows_kernel(
+    const int cblock_hist_size, const int descriptors_quadstep,
+    const int cdescr_size, const int cdescr_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float* block_hists, __global float* descriptors)
+{
+    int tid = get_local_id(0);
+    int gidX = get_group_id(0);
+    int gidY = get_group_id(1);
+
+    // Get left top corner of the window in src
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    // Get left top corner of the window in dst
+    __global float* descriptor = descriptors +
+        (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
+
+    // Copy elements from src to dst
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int offset_y = i / cdescr_width;
+        int offset_x = i - offset_y * cdescr_width;
+        descriptor[i] = hist[offset_y * img_block_width * cblock_hist_size + offset_x];
+    }
+}
+
+__kernel void extract_descrs_by_cols_kernel(
+    const int cblock_hist_size, const int descriptors_quadstep, const int cdescr_size,
+    const int cnblocks_win_x, const int cnblocks_win_y, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float* block_hists, __global float* descriptors)
+{
+    int tid = get_local_id(0);
+    int gidX = get_group_id(0);
+    int gidY = get_group_id(1);
+
+    // Get left top corner of the window in src
+    __global const float* hist = block_hists +  (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    // Get left top corner of the window in dst
+    __global float* descriptor = descriptors +
+        (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
+
+    // Copy elements from src to dst
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int block_idx = i / cblock_hist_size;
+        int idx_in_block = i - block_idx * cblock_hist_size;
+
+        int y = block_idx / cnblocks_win_x;
+        int x = block_idx - y * cnblocks_win_x;
+
+        descriptor[(x * cnblocks_win_y + y) * cblock_hist_size + idx_in_block] =
+            hist[(y * img_block_width  + x) * cblock_hist_size + idx_in_block];
+    }
+}
+
+//----------------------------------------------------------------------------
+// Gradients computation
+
+__kernel void compute_gradients_8UC4_kernel(
+    const int height, const int width,
+    const int img_step, const int grad_quadstep, const int qangle_step,
+    const __global uchar4 * img, __global float * grad, __global uchar * qangle,
+    const float angle_scale, const char correct_gamma, const int cnbins)
+{
+    const int x = get_global_id(0);
+    const int tid = get_local_id(0);
+    const int gSizeX = get_local_size(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const uchar4* row = img + gidY * img_step;
+
+    __local float sh_row[(NTHREADS + 2) * 3];
+
+    uchar4 val;
+    if (x < width)
+        val = row[x];
+    else
+        val = row[width - 2];
+
+    sh_row[tid + 1] = val.x;
+    sh_row[tid + 1 + (NTHREADS + 2)] = val.y;
+    sh_row[tid + 1 + 2 * (NTHREADS + 2)] = val.z;
+
+    if (tid == 0)
+    {
+        val = row[max(x - 1, 1)];
+        sh_row[0] = val.x;
+        sh_row[(NTHREADS + 2)] = val.y;
+        sh_row[2 * (NTHREADS + 2)] = val.z;
+    }
+
+    if (tid == gSizeX - 1)
+    {
+        val = row[min(x + 1, width - 2)];
+        sh_row[gSizeX + 1] = val.x;
+        sh_row[gSizeX + 1 + (NTHREADS + 2)] = val.y;
+        sh_row[gSizeX + 1 + 2 * (NTHREADS + 2)] = val.z;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (x < width)
+    {
+        float3 a = (float3) (sh_row[tid], sh_row[tid + (NTHREADS + 2)],
+            sh_row[tid + 2 * (NTHREADS + 2)]);
+        float3 b = (float3) (sh_row[tid + 2], sh_row[tid + 2 + (NTHREADS + 2)],
+            sh_row[tid + 2 + 2 * (NTHREADS + 2)]);
+
+        float3 dx;
+        if (correct_gamma == 1)
+            dx = sqrt(b) - sqrt(a);
+        else
+            dx = b - a;
+
+        float3 dy = (float3) 0.f;
+
+        if (gidY > 0 && gidY < height - 1)
+        {
+            a = convert_float3(img[(gidY - 1) * img_step + x].xyz);
+            b = convert_float3(img[(gidY + 1) * img_step + x].xyz);
+
+            if (correct_gamma == 1)
+                dy = sqrt(b) - sqrt(a);
+            else
+                dy = b - a;
+        }
+
+        float best_dx = dx.x;
+        float best_dy = dy.x;
+
+        float mag0 = dx.x * dx.x + dy.x * dy.x;
+        float mag1 = dx.y * dx.y + dy.y * dy.y;
+        if (mag0 < mag1)
+        {
+            best_dx = dx.y;
+            best_dy = dy.y;
+            mag0 = mag1;
+        }
+
+        mag1 = dx.z * dx.z + dy.z * dy.z;
+        if (mag0 < mag1)
+        {
+            best_dx = dx.z;
+            best_dy = dy.z;
+            mag0 = mag1;
+        }
+
+        mag0 = sqrt(mag0);
+
+        float ang = (atan2(best_dy, best_dx) + CV_PI_F) * angle_scale - 0.5f;
+        int hidx = (int)floor(ang);
+        ang -= hidx;
+        hidx = (hidx + cnbins) % cnbins;
+
+        qangle[(gidY * qangle_step + x) << 1] = hidx;
+        qangle[((gidY * qangle_step + x) << 1) + 1] = (hidx + 1) % cnbins;
+        grad[(gidY * grad_quadstep + x) << 1] = mag0 * (1.f - ang);
+        grad[((gidY * grad_quadstep + x) << 1) + 1] = mag0 * ang;
+    }
+}
+
+__kernel void compute_gradients_8UC1_kernel(
+    const int height, const int width,
+    const int img_step, const int grad_quadstep, const int qangle_step,
+    __global const uchar * img, __global float * grad, __global uchar * qangle,
+    const float angle_scale, const char correct_gamma, const int cnbins)
+{
+    const int x = get_global_id(0);
+    const int tid = get_local_id(0);
+    const int gSizeX = get_local_size(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const uchar* row = img + gidY * img_step;
+
+    __local float sh_row[NTHREADS + 2];
+
+    if (x < width)
+        sh_row[tid + 1] = row[x];
+    else
+        sh_row[tid + 1] = row[width - 2];
+
+    if (tid == 0)
+        sh_row[0] = row[max(x - 1, 1)];
+
+    if (tid == gSizeX - 1)
+        sh_row[gSizeX + 1] = row[min(x + 1, width - 2)];
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (x < width)
+    {
+        float dx;
+
+        if (correct_gamma == 1)
+            dx = sqrt(sh_row[tid + 2]) - sqrt(sh_row[tid]);
+        else
+            dx = sh_row[tid + 2] - sh_row[tid];
+
+        float dy = 0.f;
+        if (gidY > 0 && gidY < height - 1)
+        {
+            float a = (float) img[ (gidY + 1) * img_step + x ];
+            float b = (float) img[ (gidY - 1) * img_step + x ];
+            if (correct_gamma == 1)
+                dy = sqrt(a) - sqrt(b);
+            else
+                dy = a - b;
+        }
+        float mag = sqrt(dx * dx + dy * dy);
+
+        float ang = (atan2(dy, dx) + CV_PI_F) * angle_scale - 0.5f;
+        int hidx = (int)floor(ang);
+        ang -= hidx;
+        hidx = (hidx + cnbins) % cnbins;
+
+        qangle[ (gidY * qangle_step + x) << 1 ]     = hidx;
+        qangle[ ((gidY * qangle_step + x) << 1) + 1 ] = (hidx + 1) % cnbins;
+        grad[ (gidY * grad_quadstep + x) << 1 ]       = mag * (1.f - ang);
+        grad[ ((gidY * grad_quadstep + x) << 1) + 1 ]   = mag * ang;
+    }
+}
diff --git a/modules/ocl/src/kernels/operator_convertTo.cl b/modules/ocl/src/opencl/operator_convertTo.cl
similarity index 100%
rename from modules/ocl/src/kernels/operator_convertTo.cl
rename to modules/ocl/src/opencl/operator_convertTo.cl
diff --git a/modules/ocl/src/kernels/operator_copyToM.cl b/modules/ocl/src/opencl/operator_copyToM.cl
similarity index 100%
rename from modules/ocl/src/kernels/operator_copyToM.cl
rename to modules/ocl/src/opencl/operator_copyToM.cl
diff --git a/modules/ocl/src/kernels/operator_setTo.cl b/modules/ocl/src/opencl/operator_setTo.cl
similarity index 100%
rename from modules/ocl/src/kernels/operator_setTo.cl
rename to modules/ocl/src/opencl/operator_setTo.cl
diff --git a/modules/ocl/src/kernels/operator_setToM.cl b/modules/ocl/src/opencl/operator_setToM.cl
similarity index 99%
rename from modules/ocl/src/kernels/operator_setToM.cl
rename to modules/ocl/src/opencl/operator_setToM.cl
index 59357fad6d..dde12d86f6 100644
--- a/modules/ocl/src/kernels/operator_setToM.cl
+++ b/modules/ocl/src/opencl/operator_setToM.cl
@@ -57,4 +57,3 @@ __kernel void set_to_with_mask(
         }
 
 }
-
diff --git a/modules/ocl/src/opencl/optical_flow_farneback.cl b/modules/ocl/src/opencl/optical_flow_farneback.cl
new file mode 100644
index 0000000000..7cc564eded
--- /dev/null
+++ b/modules/ocl/src/opencl/optical_flow_farneback.cl
@@ -0,0 +1,450 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Sen Liu, swjtuls1987@126.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#define tx  get_local_id(0)
+#define ty  get_local_id(1)
+#define bx  get_group_id(0)
+#define bdx get_local_size(0)
+
+#define BORDER_SIZE 5
+#define MAX_KSIZE_HALF 100
+
+#ifndef polyN
+#define polyN 5
+#endif
+
+__kernel void polynomialExpansion(__global float * dst,
+                                  __global __const float * src,
+                                  __global __const float * c_g,
+                                  __global __const float * c_xg,
+                                  __global __const float * c_xxg,
+                                  __local float * smem,
+                                  const float4 ig,
+                                  const int height, const int width,
+                                  int dstStep, int srcStep)
+{
+    const int y = get_global_id(1);
+    const int x = bx * (bdx - 2*polyN) + tx - polyN;
+
+    dstStep /= sizeof(*dst);
+    srcStep /= sizeof(*src);
+
+    int xWarped;
+    __local float *row = smem + tx;
+
+    if (y < height && y >= 0)
+    {
+        xWarped = min(max(x, 0), width - 1);
+
+        row[0] = src[mad24(y, srcStep, xWarped)] * c_g[0];
+        row[bdx] = 0.f;
+        row[2*bdx] = 0.f;
+
+#pragma unroll
+        for (int k = 1; k <= polyN; ++k)
+        {
+            float t0 = src[mad24(max(y - k, 0), srcStep, xWarped)];
+            float t1 = src[mad24(min(y + k, height - 1), srcStep, xWarped)];
+
+            row[0] += c_g[k] * (t0 + t1);
+            row[bdx] += c_xg[k] * (t1 - t0);
+            row[2*bdx] += c_xxg[k] * (t0 + t1);
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (y < height && y >= 0 && tx >= polyN && tx + polyN < bdx && x < width)
+    {
+        float b1 = c_g[0] * row[0];
+        float b3 = c_g[0] * row[bdx];
+        float b5 = c_g[0] * row[2*bdx];
+        float b2 = 0, b4 = 0, b6 = 0;
+
+#pragma unroll
+        for (int k = 1; k <= polyN; ++k)
+        {
+            b1 += (row[k] + row[-k]) * c_g[k];
+            b4 += (row[k] + row[-k]) * c_xxg[k];
+            b2 += (row[k] - row[-k]) * c_xg[k];
+            b3 += (row[k + bdx] + row[-k + bdx]) * c_g[k];
+            b6 += (row[k + bdx] - row[-k + bdx]) * c_xg[k];
+            b5 += (row[k + 2*bdx] + row[-k + 2*bdx]) * c_g[k];
+        }
+
+        dst[mad24(y, dstStep, xWarped)] = b3*ig.s0;
+        dst[mad24(height + y, dstStep, xWarped)] = b2*ig.s0;
+        dst[mad24(2*height + y, dstStep, xWarped)] = b1*ig.s1 + b5*ig.s2;
+        dst[mad24(3*height + y, dstStep, xWarped)] = b1*ig.s1 + b4*ig.s2;
+        dst[mad24(4*height + y, dstStep, xWarped)] = b6*ig.s3;
+    }
+}
+
+inline int idx_row_low(const int y, const int last_row)
+{
+    return abs(y) % (last_row + 1);
+}
+
+inline int idx_row_high(const int y, const int last_row)
+{
+    return abs(last_row - abs(last_row - y)) % (last_row + 1);
+}
+
+inline int idx_row(const int y, const int last_row)
+{
+    return idx_row_low(idx_row_high(y, last_row), last_row);
+}
+
+inline int idx_col_low(const int x, const int last_col)
+{
+    return abs(x) % (last_col + 1);
+}
+
+inline int idx_col_high(const int x, const int last_col)
+{
+    return abs(last_col - abs(last_col - x)) % (last_col + 1);
+}
+
+inline int idx_col(const int x, const int last_col)
+{
+    return idx_col_low(idx_col_high(x, last_col), last_col);
+}
+
+__kernel void gaussianBlur(__global float * dst,
+                           __global const float * src,
+                           __global const float * c_gKer,
+                           __local float * smem,
+                           const int height,  const int width,
+                           int dstStep, int srcStep,
+                           const int ksizeHalf)
+{
+    const int y = get_global_id(1);
+    const int x = get_global_id(0);
+
+    dstStep /= sizeof(*dst);
+    srcStep /= sizeof(*src);
+
+    __local float *row = smem + ty * (bdx + 2*ksizeHalf);
+
+    if (y < height)
+    {
+        // Vertical pass
+        for (int i = tx; i < bdx + 2*ksizeHalf; i += bdx)
+        {
+            int xExt = (int)(bx * bdx) + i - ksizeHalf;
+            xExt = idx_col(xExt, width - 1);
+            row[i] = src[mad24(y, srcStep, xExt)] * c_gKer[0];
+            for (int j = 1; j <= ksizeHalf; ++j)
+                row[i] += (src[mad24(idx_row_low(y - j, height - 1), srcStep, xExt)]
+                           + src[mad24(idx_row_high(y + j, height - 1), srcStep, xExt)]) * c_gKer[j];
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (y < height && y >= 0 && x < width && x >= 0)
+    {
+        // Horizontal pass
+        row += tx + ksizeHalf;
+        float res = row[0] * c_gKer[0];
+        for (int i = 1; i <= ksizeHalf; ++i)
+            res += (row[-i] + row[i]) * c_gKer[i];
+
+        dst[mad24(y, dstStep, x)] = res;
+    }
+}
+
+__constant float c_border[BORDER_SIZE + 1] = { 0.14f, 0.14f, 0.4472f, 0.4472f, 0.4472f, 1.f };
+
+__kernel void updateMatrices(__global float * M,
+                             __global const float * flowx, __global const float * flowy,
+                             __global const float * R0, __global const float * R1,
+                             const int height, const int width,
+                             int mStep, int xStep,  int yStep, int R0Step, int R1Step)
+{
+    const int y = get_global_id(1);
+    const int x = get_global_id(0);
+
+    mStep /= sizeof(*M);
+    xStep /= sizeof(*flowx);
+    yStep /= sizeof(*flowy);
+    R0Step /= sizeof(*R0);
+    R1Step /= sizeof(*R1);
+
+    if (y < height && y >= 0 && x < width && x >= 0)
+    {
+        float dx = flowx[mad24(y, xStep, x)];
+        float dy = flowy[mad24(y, yStep, x)];
+        float fx = x + dx;
+        float fy = y + dy;
+
+        int x1 = convert_int(floor(fx));
+        int y1 = convert_int(floor(fy));
+        fx -= x1;
+        fy -= y1;
+
+        float r2, r3, r4, r5, r6;
+
+        if (x1 >= 0 && y1 >= 0 && x1 < width - 1 && y1 < height - 1)
+        {
+            float a00 = (1.f - fx) * (1.f - fy);
+            float a01 = fx * (1.f - fy);
+            float a10 = (1.f - fx) * fy;
+            float a11 = fx * fy;
+
+            r2 = a00 * R1[mad24(y1, R1Step, x1)] +
+                 a01 * R1[mad24(y1, R1Step, x1 + 1)] +
+                 a10 * R1[mad24(y1 + 1, R1Step, x1)] +
+                 a11 * R1[mad24(y1 + 1, R1Step, x1 + 1)];
+
+            r3 = a00 * R1[mad24(height + y1, R1Step, x1)] +
+                 a01 * R1[mad24(height + y1, R1Step, x1 + 1)] +
+                 a10 * R1[mad24(height + y1 + 1, R1Step, x1)] +
+                 a11 * R1[mad24(height + y1 + 1, R1Step, x1 + 1)];
+
+            r4 = a00 * R1[mad24(2*height + y1, R1Step, x1)] +
+                 a01 * R1[mad24(2*height + y1, R1Step, x1 + 1)] +
+                 a10 * R1[mad24(2*height + y1 + 1, R1Step, x1)] +
+                 a11 * R1[mad24(2*height + y1 + 1, R1Step, x1 + 1)];
+
+            r5 = a00 * R1[mad24(3*height + y1, R1Step, x1)] +
+                 a01 * R1[mad24(3*height + y1, R1Step, x1 + 1)] +
+                 a10 * R1[mad24(3*height + y1 + 1, R1Step, x1)] +
+                 a11 * R1[mad24(3*height + y1 + 1, R1Step, x1 + 1)];
+
+            r6 = a00 * R1[mad24(4*height + y1, R1Step, x1)] +
+                 a01 * R1[mad24(4*height + y1, R1Step, x1 + 1)] +
+                 a10 * R1[mad24(4*height + y1 + 1, R1Step, x1)] +
+                 a11 * R1[mad24(4*height + y1 + 1, R1Step, x1 + 1)];
+
+            r4 = (R0[mad24(2*height + y, R0Step, x)] + r4) * 0.5f;
+            r5 = (R0[mad24(3*height + y, R0Step, x)] + r5) * 0.5f;
+            r6 = (R0[mad24(4*height + y, R0Step, x)] + r6) * 0.25f;
+        }
+        else
+        {
+            r2 = r3 = 0.f;
+            r4 = R0[mad24(2*height + y, R0Step, x)];
+            r5 = R0[mad24(3*height + y, R0Step, x)];
+            r6 = R0[mad24(4*height + y, R0Step, x)] * 0.5f;
+        }
+
+        r2 = (R0[mad24(y, R0Step, x)] - r2) * 0.5f;
+        r3 = (R0[mad24(height + y, R0Step, x)] - r3) * 0.5f;
+
+        r2 += r4*dy + r6*dx;
+        r3 += r6*dy + r5*dx;
+
+        float scale =
+            c_border[min(x, BORDER_SIZE)] *
+            c_border[min(y, BORDER_SIZE)] *
+            c_border[min(width - x - 1, BORDER_SIZE)] *
+            c_border[min(height - y - 1, BORDER_SIZE)];
+
+        r2 *= scale;
+        r3 *= scale;
+        r4 *= scale;
+        r5 *= scale;
+        r6 *= scale;
+
+        M[mad24(y, mStep, x)] = r4*r4 + r6*r6;
+        M[mad24(height + y, mStep, x)] = (r4 + r5)*r6;
+        M[mad24(2*height + y, mStep, x)] = r5*r5 + r6*r6;
+        M[mad24(3*height + y, mStep, x)] = r4*r2 + r6*r3;
+        M[mad24(4*height + y, mStep, x)] = r6*r2 + r5*r3;
+    }
+}
+
+__kernel void boxFilter5(__global float * dst,
+                         __global const float * src,
+                         __local float * smem,
+                         const int height,  const int width,
+                         int dstStep, int srcStep,
+                         const int ksizeHalf)
+{
+    const int y = get_global_id(1);
+    const int x = get_global_id(0);
+
+    const float boxAreaInv = 1.f / ((1 + 2*ksizeHalf) * (1 + 2*ksizeHalf));
+    const int smw = bdx + 2*ksizeHalf; // shared memory "width"
+    __local float *row = smem + 5 * ty * smw;
+
+    dstStep /= sizeof(*dst);
+    srcStep /= sizeof(*src);
+
+    if (y < height)
+    {
+        // Vertical pass
+        for (int i = tx; i < bdx + 2*ksizeHalf; i += bdx)
+        {
+            int xExt = (int)(bx * bdx) + i - ksizeHalf;
+            xExt = min(max(xExt, 0), width - 1);
+
+#pragma unroll
+            for (int k = 0; k < 5; ++k)
+                row[k*smw + i] = src[mad24(k*height + y, srcStep, xExt)];
+
+            for (int j = 1; j <= ksizeHalf; ++j)
+#pragma unroll
+                for (int k = 0; k < 5; ++k)
+                    row[k*smw + i] +=
+                        src[mad24(k*height + max(y - j, 0), srcStep, xExt)] +
+                        src[mad24(k*height + min(y + j, height - 1), srcStep, xExt)];
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (y < height && y >= 0 && x < width && x >= 0)
+    {
+        // Horizontal pass
+
+        row += tx + ksizeHalf;
+        float res[5];
+
+#pragma unroll
+        for (int k = 0; k < 5; ++k)
+            res[k] = row[k*smw];
+
+        for (int i = 1; i <= ksizeHalf; ++i)
+#pragma unroll
+            for (int k = 0; k < 5; ++k)
+                res[k] += row[k*smw - i] + row[k*smw + i];
+
+#pragma unroll
+        for (int k = 0; k < 5; ++k)
+            dst[mad24(k*height + y, dstStep, x)] = res[k] * boxAreaInv;
+    }
+}
+
+__kernel void updateFlow(__global float4 * flowx, __global float4 * flowy,
+                         __global const float4 * M,
+                         const int height, const int width,
+                         int xStep, int yStep, int mStep)
+{
+    const int y = get_global_id(1);
+    const int x = get_global_id(0);
+
+    xStep /= sizeof(*flowx);
+    yStep /= sizeof(*flowy);
+    mStep /= sizeof(*M);
+
+    if (y < height && y >= 0 && x < width && x >= 0)
+    {
+        float4 g11 = M[mad24(y, mStep, x)];
+        float4 g12 = M[mad24(height + y, mStep, x)];
+        float4 g22 = M[mad24(2*height + y, mStep, x)];
+        float4 h1 =  M[mad24(3*height + y, mStep, x)];
+        float4 h2 =  M[mad24(4*height + y, mStep, x)];
+
+        float4 detInv = (float4)(1.f) / (g11*g22 - g12*g12 + (float4)(1e-3f));
+
+        flowx[mad24(y, xStep, x)] = (g11*h2 - g12*h1) * detInv;
+        flowy[mad24(y, yStep, x)] = (g22*h1 - g12*h2) * detInv;
+    }
+}
+
+__kernel void gaussianBlur5(__global float * dst,
+                            __global const float * src,
+                            __global const float * c_gKer,
+                            __local float * smem,
+                            const int height,  const int width,
+                            int dstStep, int srcStep,
+                            const int ksizeHalf)
+{
+    const int y = get_global_id(1);
+    const int x = get_global_id(0);
+
+    const int smw = bdx + 2*ksizeHalf; // shared memory "width"
+    __local volatile float *row = smem + 5 * ty * smw;
+
+    dstStep /= sizeof(*dst);
+    srcStep /= sizeof(*src);
+
+    if (y < height)
+    {
+        // Vertical pass
+        for (int i = tx; i < bdx + 2*ksizeHalf; i += bdx)
+        {
+            int xExt = (int)(bx * bdx) + i - ksizeHalf;
+            xExt = idx_col(xExt, width - 1);
+
+#pragma unroll
+            for (int k = 0; k < 5; ++k)
+                row[k*smw + i] = src[mad24(k*height + y, srcStep, xExt)] * c_gKer[0];
+
+            for (int j = 1; j <= ksizeHalf; ++j)
+#pragma unroll
+                for (int k = 0; k < 5; ++k)
+                    row[k*smw + i] +=
+                        (src[mad24(k*height + idx_row_low(y - j, height - 1), srcStep, xExt)] +
+                         src[mad24(k*height + idx_row_high(y + j, height - 1), srcStep, xExt)]) * c_gKer[j];
+        }
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (y < height && y >= 0 && x < width && x >= 0)
+    {
+        // Horizontal pass
+
+        row += tx + ksizeHalf;
+        float res[5];
+
+#pragma unroll
+        for (int k = 0; k < 5; ++k)
+            res[k] = row[k*smw] * c_gKer[0];
+
+        for (int i = 1; i <= ksizeHalf; ++i)
+#pragma unroll
+            for (int k = 0; k < 5; ++k)
+                res[k] += (row[k*smw - i] + row[k*smw + i]) * c_gKer[i];
+
+#pragma unroll
+        for (int k = 0; k < 5; ++k)
+            dst[mad24(k*height + y, dstStep, x)] = res[k];
+    }
+}
diff --git a/modules/ocl/src/kernels/pyr_down.cl b/modules/ocl/src/opencl/pyr_down.cl
similarity index 100%
rename from modules/ocl/src/kernels/pyr_down.cl
rename to modules/ocl/src/opencl/pyr_down.cl
diff --git a/modules/ocl/src/opencl/pyr_up.cl b/modules/ocl/src/opencl/pyr_up.cl
new file mode 100644
index 0000000000..4afa7b7104
--- /dev/null
+++ b/modules/ocl/src/opencl/pyr_up.cl
@@ -0,0 +1,673 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Zhang Chunpeng	chunpeng@multicorewareinc.com
+//    Dachuan Zhao, dachuan@multicorewareinc.com
+//    Yao Wang, yao@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+//#pragma OPENCL EXTENSION cl_amd_printf : enable
+
+uchar get_valid_uchar(float data)
+{
+    return (uchar)(data <= 255 ? data : data > 0 ? 255 : 0);
+}
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_8UC1  //////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C1_D0(__global uchar* src,__global uchar* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+    __local float s_srcPatch[10][10];
+    __local float s_dstPatch[20][16];
+    const int tidx = get_local_id(0);
+    const int tidy = get_local_id(1);
+    const int lsizex = get_local_size(0);
+    const int lsizey = get_local_size(1);
+
+    if( tidx < 10 && tidy < 10 )
+    {
+        int srcx = mad24((int)get_group_id(0), (lsizex>>1), tidx) - 1;
+        int srcy = mad24((int)get_group_id(1), (lsizey>>1), tidy) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[tidy][tidx] = (float)(src[srcx + srcy * srcStep]);
+
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float sum = 0;
+    const int evenFlag = (int)((tidx & 1) == 0);
+    const int oddFlag = (int)((tidx & 1) != 0);
+    const bool  eveny = ((tidy & 1) == 0);
+
+    if(eveny)
+    {
+        sum = (evenFlag * 0.0625f) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 2) >> 1)];
+    }
+
+    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
+
+    if (get_local_id(1) < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = (evenFlag * 0.0625f) * s_srcPatch[lsizey - 16][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 16][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[lsizey - 16][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 16][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[lsizey - 16][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
+    }
+
+    if (get_local_id(1) > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = (evenFlag * 0.0625f) * s_srcPatch[lsizey - 7][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 7][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[lsizey - 7][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 7][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
+        }
+        s_dstPatch[4 + tidy][tidx] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = 0;
+
+    sum = 0.0625f * s_dstPatch[2 + tidy - 2][tidx];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][tidx];
+    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][tidx];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][tidx];
+    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][tidx];
+
+    if ((x < dstCols) && (y < dstRows))
+        dst[x + y * dstStep] = convert_uchar_sat_rte(4.0f * sum);
+
+}
+
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_16UC1  /////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C1_D2(__global ushort* src,__global ushort* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    __local float s_srcPatch[10][10];
+    __local float s_dstPatch[20][16];
+
+    srcStep = srcStep >> 1;
+    dstStep = dstStep >> 1;
+    srcOffset = srcOffset >> 1;
+    dstOffset = dstOffset >> 1;
+
+
+    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
+    {
+        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
+        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[get_local_id(1)][get_local_id(0)] = (float)(src[srcx + srcy * srcStep]);
+
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float sum = 0;
+
+    const int evenFlag = (int)((get_local_id(0) & 1) == 0);
+    const int oddFlag = (int)((get_local_id(0) & 1) != 0);
+    const bool  eveny = ((get_local_id(1) & 1) == 0);
+    const int tidx = get_local_id(0);
+
+    if(eveny)
+    {
+        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)];
+    }
+
+    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
+
+    if (get_local_id(1) < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
+    }
+
+    if (get_local_id(1) > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
+        }
+        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = 0;
+
+    const int tidy = get_local_id(1);
+
+    sum = sum + 0.0625f * s_dstPatch[2 + tidy - 2][get_local_id(0)];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][get_local_id(0)];
+    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][get_local_id(0)];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][get_local_id(0)];
+    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][get_local_id(0)];
+
+    if ((x < dstCols) && (y < dstRows))
+        dst[x + y * dstStep] = convert_short_sat_rte(4.0f * sum);
+
+}
+
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_32FC1  /////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C1_D5(__global float* src,__global float* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+    const int tidx = get_local_id(0);
+    const int tidy = get_local_id(1);
+    const int lsizex = get_local_size(0);
+    const int lsizey = get_local_size(1);
+    __local float s_srcPatch[10][10];
+    __local float s_dstPatch[20][16];
+
+    srcOffset = srcOffset >> 2;
+    dstOffset = dstOffset >> 2;
+    srcStep = srcStep >> 2;
+    dstStep = dstStep >> 2;
+
+
+    if( tidx < 10 && tidy < 10 )
+    {
+        int srcx = mad24((int)get_group_id(0), lsizex>>1, tidx) - 1;
+        int srcy = mad24((int)get_group_id(1), lsizey>>1, tidy) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[tidy][tidx] = (float)(src[srcx + srcy * srcStep]);
+
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float sum = 0;
+    const int evenFlag = (int)((tidx & 1) == 0);
+    const int oddFlag = (int)((tidx & 1) != 0);
+    const bool  eveny = ((tidy & 1) == 0);
+
+
+    if(eveny)
+    {
+        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + (evenFlag * 0.375f ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + (evenFlag * 0.0625f) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 2) >> 1)];
+    }
+
+    s_dstPatch[2 + tidy][tidx] = sum;
+
+    if (tidy < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[lsizey - 16][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 16][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[lsizey - 16][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 16][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[lsizey - 16][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[tidy][tidx] = sum;
+    }
+
+    if (tidy > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[lsizey - 7][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 7][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * 0.375f ) * s_srcPatch[lsizey - 7][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * 0.25f  ) * s_srcPatch[lsizey - 7][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * 0.0625f) * s_srcPatch[lsizey - 7][1 + ((tidx + 2) >> 1)];
+        }
+        s_dstPatch[4 + tidy][tidx] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = 0.0625f * s_dstPatch[2 + tidy - 2][tidx];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy - 1][tidx];
+    sum = sum + 0.375f  * s_dstPatch[2 + tidy    ][tidx];
+    sum = sum + 0.25f   * s_dstPatch[2 + tidy + 1][tidx];
+    sum = sum + 0.0625f * s_dstPatch[2 + tidy + 2][tidx];
+
+    if ((x < dstCols) && (y < dstRows))
+        dst[x + y * dstStep] = (float)(4.0f * sum);
+
+}
+
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_8UC4  //////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C4_D0(__global uchar4* src,__global uchar4* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+    const int tidx = get_local_id(0);
+    const int tidy = get_local_id(1);
+    const int lsizex = get_local_size(0);
+    const int lsizey = get_local_size(1);
+    __local float4 s_srcPatch[10][10];
+    __local float4 s_dstPatch[20][16];
+
+    srcOffset >>= 2;
+    dstOffset >>= 2;
+    srcStep >>= 2;
+    dstStep >>= 2;
+
+
+    if( tidx < 10 && tidy < 10 )
+    {
+        int srcx = mad24((int)get_group_id(0), lsizex>>1, tidx) - 1;
+        int srcy = mad24((int)get_group_id(1), lsizey>>1, tidy) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[tidy][tidx] = convert_float4(src[srcx + srcy * srcStep]);
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float4 sum = (float4)(0,0,0,0);
+
+    const float4 evenFlag = (float4)((tidx & 1) == 0);
+    const float4 oddFlag = (float4)((tidx & 1) != 0);
+    const bool  eveny = ((tidy & 1) == 0);
+
+    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
+    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
+    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
+
+
+    if(eveny)
+    {
+        sum = sum + ( evenFlag * co3) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + ( evenFlag * co1) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + ( evenFlag * co3) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 2) >> 1)];
+
+    }
+
+    s_dstPatch[2 + tidy][tidx] = sum;
+
+    if (tidy < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-16][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-16][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1) * s_srcPatch[lsizey-16][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-16][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-16][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[tidy][tidx] = sum;
+    }
+
+    if (tidy > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-7][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-7][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1) * s_srcPatch[lsizey-7][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-7][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-7][1 + ((tidx + 2) >> 1)];
+
+        }
+        s_dstPatch[4 + tidy][tidx] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = co3 * s_dstPatch[2 + tidy - 2][tidx];
+    sum = sum + co2 * s_dstPatch[2 + tidy - 1][tidx];
+    sum = sum + co1 * s_dstPatch[2 + tidy    ][tidx];
+    sum = sum + co2 * s_dstPatch[2 + tidy + 1][tidx];
+    sum = sum + co3 * s_dstPatch[2 + tidy + 2][tidx];
+
+    if ((x < dstCols) && (y < dstRows))
+    {
+        dst[x + y * dstStep] = convert_uchar4_sat_rte(4.0f * sum);
+    }
+}
+
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_16UC4 //////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C4_D2(__global ushort4* src,__global ushort4* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    __local float4 s_srcPatch[10][10];
+    __local float4 s_dstPatch[20][16];
+
+    srcOffset >>= 3;
+    dstOffset >>= 3;
+    srcStep >>= 3;
+    dstStep >>= 3;
+
+
+    if( get_local_id(0) < 10 && get_local_id(1) < 10 )
+    {
+        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + get_local_id(0)) - 1;
+        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + get_local_id(1)) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[get_local_id(1)][get_local_id(0)] = convert_float4(src[srcx + srcy * srcStep]);
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float4 sum = (float4)(0,0,0,0);
+
+    const float4 evenFlag = (float4)((get_local_id(0) & 1) == 0);
+    const float4 oddFlag = (float4)((get_local_id(0) & 1) != 0);
+    const bool  eveny = ((get_local_id(1) & 1) == 0);
+    const int tidx = get_local_id(0);
+
+    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
+    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
+    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
+
+
+    if(eveny)
+    {
+        sum = sum + ( evenFlag* co3 ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + ( evenFlag* co1 ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + ( evenFlag* co3 ) * s_srcPatch[1 + (get_local_id(1) >> 1)][1 + ((tidx + 2) >> 1)];
+
+    }
+
+    s_dstPatch[2 + get_local_id(1)][get_local_id(0)] = sum;
+
+    if (get_local_id(1) < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3 ) * s_srcPatch[0][1 + ((tidx - 2) >> 1)];
+            sum = sum + (oddFlag * co2  ) * s_srcPatch[0][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1 ) * s_srcPatch[0][1 + ((tidx    ) >> 1)];
+            sum = sum + (oddFlag * co2  ) * s_srcPatch[0][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3 ) * s_srcPatch[0][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[get_local_id(1)][get_local_id(0)] = sum;
+    }
+
+    if (get_local_id(1) > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3) * s_srcPatch[9][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[9][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1) * s_srcPatch[9][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[9][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3) * s_srcPatch[9][1 + ((tidx + 2) >> 1)];
+
+        }
+        s_dstPatch[4 + get_local_id(1)][get_local_id(0)] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = 0;
+
+    const int tidy = get_local_id(1);
+
+    sum = sum + co3 * s_dstPatch[2 + tidy - 2][get_local_id(0)];
+    sum = sum + co2 * s_dstPatch[2 + tidy - 1][get_local_id(0)];
+    sum = sum + co1 * s_dstPatch[2 + tidy    ][get_local_id(0)];
+    sum = sum + co2 * s_dstPatch[2 + tidy + 1][get_local_id(0)];
+    sum = sum + co3 * s_dstPatch[2 + tidy + 2][get_local_id(0)];
+
+    if ((x < dstCols) && (y < dstRows))
+    {
+        dst[x + y * dstStep] = convert_ushort4_sat_rte(4.0f * sum);
+    }
+}
+
+///////////////////////////////////////////////////////////////////////
+//////////////////////////  CV_32FC4 //////////////////////////////////
+///////////////////////////////////////////////////////////////////////
+__kernel void pyrUp_C4_D5(__global float4* src,__global float4* dst,
+                          int srcRows,int dstRows,int srcCols,int dstCols,
+                          int srcOffset,int dstOffset,int srcStep,int dstStep)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+    const int tidx = get_local_id(0);
+    const int tidy = get_local_id(1);
+    const int lsizex = get_local_size(0);
+    const int lsizey = get_local_size(1);
+    __local float4 s_srcPatch[10][10];
+    __local float4 s_dstPatch[20][16];
+
+    srcOffset >>= 4;
+    dstOffset >>= 4;
+    srcStep >>= 4;
+    dstStep >>= 4;
+
+
+    if( tidx < 10 && tidy < 10 )
+    {
+        int srcx = (int)(get_group_id(0) * get_local_size(0) / 2 + tidx) - 1;
+        int srcy = (int)(get_group_id(1) * get_local_size(1) / 2 + tidy) - 1;
+
+        srcx = abs(srcx);
+        srcx = min(srcCols - 1,srcx);
+
+        srcy = abs(srcy);
+        srcy = min(srcRows -1 ,srcy);
+
+        s_srcPatch[tidy][tidx] = (float4)(src[srcx + srcy * srcStep]);
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float4 sum = (float4)(0,0,0,0);
+
+    const float4 evenFlag = (float4)((tidx & 1) == 0);
+    const float4 oddFlag = (float4)((tidx & 1) != 0);
+    const bool  eveny = ((tidy & 1) == 0);
+
+    float4 co1 = (float4)(0.375f, 0.375f, 0.375f, 0.375f);
+    float4 co2 = (float4)(0.25f, 0.25f, 0.25f, 0.25f);
+    float4 co3 = (float4)(0.0625f, 0.0625f, 0.0625f, 0.0625f);
+
+
+    if(eveny)
+    {
+        sum = sum + ( evenFlag* co3 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 2) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx - 1) >> 1)];
+        sum = sum + ( evenFlag* co1 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx    ) >> 1)];
+        sum = sum + ( oddFlag * co2 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 1) >> 1)];
+        sum = sum + ( evenFlag* co3 ) * s_srcPatch[1 + (tidy >> 1)][1 + ((tidx + 2) >> 1)];
+
+    }
+
+    s_dstPatch[2 + tidy][tidx] = sum;
+
+    if (tidy < 2)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3 ) * s_srcPatch[lsizey-16][1 + ((tidx - 2) >> 1)];
+            sum = sum + (oddFlag * co2  ) * s_srcPatch[lsizey-16][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1 ) * s_srcPatch[lsizey-16][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * co2 ) * s_srcPatch[lsizey-16][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3 ) * s_srcPatch[lsizey-16][1 + ((tidx + 2) >> 1)];
+        }
+
+        s_dstPatch[tidy][tidx] = sum;
+    }
+
+    if (tidy > 13)
+    {
+        sum = 0;
+
+        if (eveny)
+        {
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-7][1 + ((tidx - 2) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-7][1 + ((tidx - 1) >> 1)];
+            sum = sum + (evenFlag * co1) * s_srcPatch[lsizey-7][1 + ((tidx    ) >> 1)];
+            sum = sum + ( oddFlag * co2) * s_srcPatch[lsizey-7][1 + ((tidx + 1) >> 1)];
+            sum = sum + (evenFlag * co3) * s_srcPatch[lsizey-7][1 + ((tidx + 2) >> 1)];
+
+        }
+        s_dstPatch[4 + tidy][tidx] = sum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = co3 * s_dstPatch[2 + tidy - 2][tidx];
+    sum = sum + co2 * s_dstPatch[2 + tidy - 1][tidx];
+    sum = sum + co1 * s_dstPatch[2 + tidy    ][tidx];
+    sum = sum + co2 * s_dstPatch[2 + tidy + 1][tidx];
+    sum = sum + co3 * s_dstPatch[2 + tidy + 2][tidx];
+
+    if ((x < dstCols) && (y < dstRows))
+    {
+        dst[x + y * dstStep] = 4.0f * sum;
+    }
+}
diff --git a/modules/ocl/src/opencl/pyrlk.cl b/modules/ocl/src/opencl/pyrlk.cl
new file mode 100644
index 0000000000..02cf3afa44
--- /dev/null
+++ b/modules/ocl/src/opencl/pyrlk.cl
@@ -0,0 +1,1020 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Dachuan Zhao, dachuan@multicorewareinc.com
+//    Yao Wang, bitwangyaoyao@gmail.com
+//    Xiaopeng Fu, fuxiaopeng2222@163.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+//#pragma OPENCL EXTENSION cl_amd_printf : enable
+
+#define	BUFFER	64
+#define	BUFFER2	BUFFER>>1
+#ifndef WAVE_SIZE
+#define WAVE_SIZE 1
+#endif
+#ifdef CPU
+void reduce3(float val1, float val2, float val3,  __local float* smem1,  __local float* smem2,  __local float* smem3, int tid)
+{
+    smem1[tid] = val1;
+    smem2[tid] = val2;
+    smem3[tid] = val3;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for(int i = BUFFER2; i > 0; i >>= 1)
+    {
+        if(tid < i)
+        {
+            smem1[tid] += smem1[tid + i];
+            smem2[tid] += smem2[tid + i];
+            smem3[tid] += smem3[tid + i];
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
+
+void reduce2(float val1, float val2, volatile __local float* smem1, volatile __local float* smem2, int tid)
+{
+    smem1[tid] = val1;
+    smem2[tid] = val2;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for(int i = BUFFER2; i > 0; i >>= 1)
+    {
+        if(tid < i)
+        {
+            smem1[tid] += smem1[tid + i];
+            smem2[tid] += smem2[tid + i];
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
+
+void reduce1(float val1, volatile __local float* smem1, int tid)
+{
+    smem1[tid] = val1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for(int i = BUFFER2; i > 0; i >>= 1)
+    {
+        if(tid < i)
+        {
+            smem1[tid] += smem1[tid + i];
+        }
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+}
+#else
+void reduce3(float val1, float val2, float val3,
+             __local volatile float* smem1, __local volatile float* smem2, __local volatile float* smem3, int tid)
+{
+    smem1[tid] = val1;
+    smem2[tid] = val2;
+    smem3[tid] = val3;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 32)
+    {
+        smem1[tid] += smem1[tid + 32];
+        smem2[tid] += smem2[tid + 32];
+        smem3[tid] += smem3[tid + 32];
+#if WAVE_SIZE < 32
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16)
+    {
+#endif
+        smem1[tid] += smem1[tid + 16];
+        smem2[tid] += smem2[tid + 16];
+        smem3[tid] += smem3[tid + 16];
+#if WAVE_SIZE <16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        smem1[tid] += smem1[tid + 8];
+        smem2[tid] += smem2[tid + 8];
+        smem3[tid] += smem3[tid + 8];
+
+        smem1[tid] += smem1[tid + 4];
+        smem2[tid] += smem2[tid + 4];
+        smem3[tid] += smem3[tid + 4];
+
+        smem1[tid] += smem1[tid + 2];
+        smem2[tid] += smem2[tid + 2];
+        smem3[tid] += smem3[tid + 2];
+
+        smem1[tid] += smem1[tid + 1];
+        smem2[tid] += smem2[tid + 1];
+        smem3[tid] += smem3[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+}
+
+void reduce2(float val1, float val2, __local volatile float* smem1, __local volatile float* smem2, int tid)
+{
+    smem1[tid] = val1;
+    smem2[tid] = val2;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 32)
+    {
+        smem1[tid] += smem1[tid + 32];
+        smem2[tid] += smem2[tid + 32];
+#if WAVE_SIZE < 32
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16)
+    {
+#endif
+        smem1[tid] += smem1[tid + 16];
+        smem2[tid] += smem2[tid + 16];
+#if WAVE_SIZE <16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        smem1[tid] += smem1[tid + 8];
+        smem2[tid] += smem2[tid + 8];
+
+        smem1[tid] += smem1[tid + 4];
+        smem2[tid] += smem2[tid + 4];
+
+        smem1[tid] += smem1[tid + 2];
+        smem2[tid] += smem2[tid + 2];
+
+        smem1[tid] += smem1[tid + 1];
+        smem2[tid] += smem2[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+}
+
+void reduce1(float val1, __local volatile float* smem1, int tid)
+{
+    smem1[tid] = val1;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (tid < 32)
+    {
+        smem1[tid] += smem1[tid + 32];
+#if WAVE_SIZE < 32
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 16)
+    {
+#endif
+        smem1[tid] += smem1[tid + 16];
+#if WAVE_SIZE <16
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (tid < 8)
+    {
+#endif
+        smem1[tid] += smem1[tid + 8];
+        smem1[tid] += smem1[tid + 4];
+        smem1[tid] += smem1[tid + 2];
+        smem1[tid] += smem1[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+}
+#endif
+
+#define SCALE (1.0f / (1 << 20))
+#define	THRESHOLD	0.01f
+
+// Image read mode
+__constant sampler_t sampler    = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_LINEAR;
+
+void SetPatch(image2d_t I, float x, float y,
+              float* Pch, float* Dx, float* Dy,
+              float* A11, float* A12, float* A22)
+{
+    *Pch = read_imagef(I, sampler, (float2)(x, y)).x;
+
+    float dIdx = 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x + 1, y)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)).x -
+                 (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x - 1, y)).x + 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)).x);
+
+    float dIdy = 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x, y + 1)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)).x -
+                 (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)).x + 10.0f * read_imagef(I, sampler, (float2)(x, y - 1)).x + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)).x);
+
+
+    *Dx = dIdx;
+    *Dy = dIdy;
+
+    *A11 += dIdx * dIdx;
+    *A12 += dIdx * dIdy;
+    *A22 += dIdy * dIdy;
+}
+
+void GetPatch(image2d_t J, float x, float y,
+              float* Pch, float* Dx, float* Dy,
+              float* b1, float* b2)
+{
+    float J_val = read_imagef(J, sampler, (float2)(x, y)).x;
+    float diff = (J_val - *Pch) * 32.0f;
+    *b1 += diff**Dx;
+    *b2 += diff**Dy;
+}
+
+void GetError(image2d_t J, const float x, const float y, const float* Pch, float* errval)
+{
+    float diff = read_imagef(J, sampler, (float2)(x,y)).x-*Pch;
+    *errval += fabs(diff);
+}
+
+void SetPatch4(image2d_t I, const float x, const float y,
+               float4* Pch, float4* Dx, float4* Dy,
+               float* A11, float* A12, float* A22)
+{
+    *Pch = read_imagef(I, sampler, (float2)(x, y));
+
+    float4 dIdx = 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x + 1, y)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)) -
+                  (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x - 1, y)) + 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)));
+
+    float4 dIdy = 3.0f * read_imagef(I, sampler, (float2)(x - 1, y + 1)) + 10.0f * read_imagef(I, sampler, (float2)(x, y + 1)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y + 1)) -
+                  (3.0f * read_imagef(I, sampler, (float2)(x - 1, y - 1)) + 10.0f * read_imagef(I, sampler, (float2)(x, y - 1)) + 3.0f * read_imagef(I, sampler, (float2)(x + 1, y - 1)));
+
+
+    *Dx = dIdx;
+    *Dy = dIdy;
+    float4 sqIdx = dIdx * dIdx;
+    *A11 += sqIdx.x + sqIdx.y + sqIdx.z;
+    sqIdx = dIdx * dIdy;
+    *A12 += sqIdx.x + sqIdx.y + sqIdx.z;
+    sqIdx = dIdy * dIdy;
+    *A22 += sqIdx.x + sqIdx.y + sqIdx.z;
+}
+
+void GetPatch4(image2d_t J, const float x, const float y,
+               const float4* Pch, const float4* Dx, const float4* Dy,
+               float* b1, float* b2)
+{
+    float4 J_val = read_imagef(J, sampler, (float2)(x, y));
+    float4 diff = (J_val - *Pch) * 32.0f;
+    float4 xdiff = diff* *Dx;
+    *b1 += xdiff.x + xdiff.y + xdiff.z;
+    xdiff = diff* *Dy;
+    *b2 += xdiff.x + xdiff.y + xdiff.z;
+}
+
+void GetError4(image2d_t J, const float x, const float y, const float4* Pch, float* errval)
+{
+    float4 diff = read_imagef(J, sampler, (float2)(x,y))-*Pch;
+    *errval += fabs(diff.x) + fabs(diff.y) + fabs(diff.z);
+}
+
+#define	GRIDSIZE	3
+__kernel void lkSparse_C1_D5(image2d_t I, image2d_t J,
+                             __global const float2* prevPts, int prevPtsStep, __global float2* nextPts, int nextPtsStep, __global uchar* status, __global float* err,
+                             const int level, const int rows, const int cols, int PATCH_X, int PATCH_Y, int cn, int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
+{
+    __local float smem1[BUFFER];
+    __local float smem2[BUFFER];
+    __local float smem3[BUFFER];
+
+    unsigned int xid=get_local_id(0);
+    unsigned int yid=get_local_id(1);
+    unsigned int gid=get_group_id(0);
+    unsigned int xsize=get_local_size(0);
+    unsigned int ysize=get_local_size(1);
+    int xBase, yBase, i, j, k;
+
+    float2 c_halfWin = (float2)((c_winSize_x - 1)>>1, (c_winSize_y - 1)>>1);
+
+    const int tid = mad24(yid, xsize, xid);
+
+    float2 prevPt = prevPts[gid] / (float2)(1 << level);
+
+    if (prevPt.x < 0 || prevPt.x >= cols || prevPt.y < 0 || prevPt.y >= rows)
+    {
+        if (tid == 0 && level == 0)
+        {
+            status[gid] = 0;
+        }
+
+        return;
+    }
+    prevPt -= c_halfWin;
+
+    // extract the patch from the first image, compute covariation matrix of derivatives
+
+    float A11 = 0;
+    float A12 = 0;
+    float A22 = 0;
+
+    float I_patch[GRIDSIZE][GRIDSIZE];
+    float dIdx_patch[GRIDSIZE][GRIDSIZE];
+    float dIdy_patch[GRIDSIZE][GRIDSIZE];
+
+    yBase=yid;
+    {
+        xBase=xid;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[0][0], &dIdx_patch[0][0], &dIdy_patch[0][0],
+                 &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[0][1], &dIdx_patch[0][1], &dIdy_patch[0][1],
+                 &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[0][2], &dIdx_patch[0][2], &dIdy_patch[0][2],
+                     &A11, &A12, &A22);
+    }
+    yBase+=ysize;
+    {
+        xBase=xid;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[1][0], &dIdx_patch[1][0], &dIdy_patch[1][0],
+                 &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[1][1], &dIdx_patch[1][1], &dIdy_patch[1][1],
+                 &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[1][2], &dIdx_patch[1][2], &dIdy_patch[1][2],
+                     &A11, &A12, &A22);
+    }
+    yBase+=ysize;
+    if(yBase<c_winSize_y)
+    {
+        xBase=xid;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[2][0], &dIdx_patch[2][0], &dIdy_patch[2][0],
+                 &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                 &I_patch[2][1], &dIdx_patch[2][1], &dIdy_patch[2][1],
+                 &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch(I, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[2][2], &dIdx_patch[2][2], &dIdy_patch[2][2],
+                     &A11, &A12, &A22);
+    }
+
+    reduce3(A11, A12, A22, smem1, smem2, smem3, tid);
+
+    A11 = smem1[0];
+    A12 = smem2[0];
+    A22 = smem3[0];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float D = A11 * A22 - A12 * A12;
+
+    if (D < 1.192092896e-07f)
+    {
+        if (tid == 0 && level == 0)
+            status[gid] = 0;
+
+        return;
+    }
+
+    A11 /= D;
+    A12 /= D;
+    A22 /= D;
+
+    prevPt = nextPts[gid] * 2.0f - c_halfWin;
+
+    for (k = 0; k < c_iters; ++k)
+    {
+        if (prevPt.x < -c_halfWin.x || prevPt.x >= cols || prevPt.y < -c_halfWin.y || prevPt.y >= rows)
+        {
+            if (tid == 0 && level == 0)
+                status[gid] = 0;
+            return;
+        }
+
+        float b1 = 0;
+        float b2 = 0;
+
+        yBase=yid;
+        {
+            xBase=xid;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[0][0], &dIdx_patch[0][0], &dIdy_patch[0][0],
+                     &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[0][1], &dIdx_patch[0][1], &dIdy_patch[0][1],
+                     &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[0][2], &dIdx_patch[0][2], &dIdy_patch[0][2],
+                         &b1, &b2);
+        }
+        yBase+=ysize;
+        {
+            xBase=xid;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[1][0], &dIdx_patch[1][0], &dIdy_patch[1][0],
+                     &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[1][1], &dIdx_patch[1][1], &dIdy_patch[1][1],
+                     &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[1][2], &dIdx_patch[1][2], &dIdy_patch[1][2],
+                         &b1, &b2);
+        }
+        yBase+=ysize;
+        if(yBase<c_winSize_y)
+        {
+            xBase=xid;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[2][0], &dIdx_patch[2][0], &dIdy_patch[2][0],
+                     &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[2][1], &dIdx_patch[2][1], &dIdy_patch[2][1],
+                     &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[2][2], &dIdx_patch[2][2], &dIdy_patch[2][2],
+                         &b1, &b2);
+        }
+
+        reduce2(b1, b2, smem1, smem2, tid);
+
+        b1 = smem1[0];
+        b2 = smem2[0];
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        float2 delta;
+        delta.x = A12 * b2 - A22 * b1;
+        delta.y = A12 * b1 - A11 * b2;
+
+        prevPt += delta;
+
+        if (fabs(delta.x) < THRESHOLD && fabs(delta.y) < THRESHOLD)
+            break;
+    }
+
+    D = 0.0f;
+    if (calcErr)
+    {
+        yBase=yid;
+        {
+            xBase=xid;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[0][0], &D);
+
+
+            xBase+=xsize;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[0][1], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[0][2], &D);
+        }
+        yBase+=ysize;
+        {
+            xBase=xid;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[1][0], &D);
+
+
+            xBase+=xsize;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[1][1], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[1][2], &D);
+        }
+        yBase+=ysize;
+        if(yBase<c_winSize_y)
+        {
+            xBase=xid;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[2][0], &D);
+
+
+            xBase+=xsize;
+            GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                     &I_patch[2][1], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError(J, prevPt.x + xBase + 0.5f, prevPt.y + yBase + 0.5f,
+                         &I_patch[2][2], &D);
+        }
+
+        reduce1(D, smem1, tid);
+    }
+
+    if (tid == 0)
+    {
+        prevPt += c_halfWin;
+
+        nextPts[gid] = prevPt;
+
+        if (calcErr)
+            err[gid] = smem1[0] / (float)(c_winSize_x * c_winSize_y);
+    }
+}
+
+
+__kernel void lkSparse_C4_D5(image2d_t I, image2d_t J,
+                             __global const float2* prevPts, int prevPtsStep, __global float2* nextPts, int nextPtsStep, __global uchar* status, __global float* err,
+                             const int level, const int rows, const int cols, int PATCH_X, int PATCH_Y, int cn, int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
+{
+    __local float smem1[BUFFER];
+    __local float smem2[BUFFER];
+    __local float smem3[BUFFER];
+
+    unsigned int xid=get_local_id(0);
+    unsigned int yid=get_local_id(1);
+    unsigned int gid=get_group_id(0);
+    unsigned int xsize=get_local_size(0);
+    unsigned int ysize=get_local_size(1);
+    int xBase, yBase, i, j, k;
+
+    float2 c_halfWin = (float2)((c_winSize_x - 1)>>1, (c_winSize_y - 1)>>1);
+
+    const int tid = mad24(yid, xsize, xid);
+
+    float2 nextPt = prevPts[gid]/(float2)(1<<level);
+
+    if (nextPt.x < 0 || nextPt.x >= cols || nextPt.y < 0 || nextPt.y >= rows)
+    {
+        if (tid == 0 && level == 0)
+        {
+            status[gid] = 0;
+        }
+
+        return;
+    }
+
+    nextPt -= c_halfWin;
+
+    // extract the patch from the first image, compute covariation matrix of derivatives
+
+    float A11 = 0.0f;
+    float A12 = 0.0f;
+    float A22 = 0.0f;
+
+    float4 I_patch[8];
+    float4 dIdx_patch[8];
+    float4 dIdy_patch[8];
+    float4 I_add,Dx_add,Dy_add;
+
+    yBase=yid;
+    {
+        xBase=xid;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[0], &dIdx_patch[0], &dIdy_patch[0],
+                  &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[1], &dIdx_patch[1], &dIdy_patch[1],
+                  &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[2], &dIdx_patch[2], &dIdy_patch[2],
+                      &A11, &A12, &A22);
+
+    }
+    yBase+=ysize;
+    {
+        xBase=xid;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[3], &dIdx_patch[3], &dIdy_patch[3],
+                  &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[4], &dIdx_patch[4], &dIdy_patch[4],
+                  &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[5], &dIdx_patch[5], &dIdy_patch[5],
+                      &A11, &A12, &A22);
+    }
+    yBase+=ysize;
+    if(yBase<c_winSize_y)
+    {
+        xBase=xid;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[6], &dIdx_patch[6], &dIdy_patch[6],
+                  &A11, &A12, &A22);
+
+
+        xBase+=xsize;
+        SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                  &I_patch[7], &dIdx_patch[7], &dIdy_patch[7],
+                  &A11, &A12, &A22);
+
+        xBase+=xsize;
+        if(xBase<c_winSize_x)
+            SetPatch4(I, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_add, &Dx_add, &Dy_add,
+                      &A11, &A12, &A22);
+    }
+
+    reduce3(A11, A12, A22, smem1, smem2, smem3, tid);
+
+    A11 = smem1[0];
+    A12 = smem2[0];
+    A22 = smem3[0];
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float D = A11 * A22 - A12 * A12;
+
+    if (D < 1.192092896e-07f)
+    {
+        if (tid == 0 && level == 0)
+            status[gid] = 0;
+
+        return;
+    }
+
+    A11 /= D;
+    A12 /= D;
+    A22 /= D;
+
+    nextPt = nextPts[gid] * 2.0f - c_halfWin;
+
+    for (k = 0; k < c_iters; ++k)
+    {
+        if (nextPt.x < -c_halfWin.x || nextPt.x >= cols || nextPt.y < -c_halfWin.y || nextPt.y >= rows)
+        {
+            if (tid == 0 && level == 0)
+                status[gid] = 0;
+            return;
+        }
+
+        float b1 = 0;
+        float b2 = 0;
+
+        yBase=yid;
+        {
+            xBase=xid;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[0], &dIdx_patch[0], &dIdy_patch[0],
+                      &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[1], &dIdx_patch[1], &dIdy_patch[1],
+                      &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_patch[2], &dIdx_patch[2], &dIdy_patch[2],
+                          &b1, &b2);
+        }
+        yBase+=ysize;
+        {
+            xBase=xid;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[3], &dIdx_patch[3], &dIdy_patch[3],
+                      &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[4], &dIdx_patch[4], &dIdy_patch[4],
+                      &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_patch[5], &dIdx_patch[5], &dIdy_patch[5],
+                          &b1, &b2);
+        }
+        yBase+=ysize;
+        if(yBase<c_winSize_y)
+        {
+            xBase=xid;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[6], &dIdx_patch[6], &dIdy_patch[6],
+                      &b1, &b2);
+
+
+            xBase+=xsize;
+            GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[7], &dIdx_patch[7], &dIdy_patch[7],
+                      &b1, &b2);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetPatch4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_add, &Dx_add, &Dy_add,
+                          &b1, &b2);
+        }
+
+        reduce2(b1, b2, smem1, smem2, tid);
+
+        b1 = smem1[0];
+        b2 = smem2[0];
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        float2 delta;
+        delta.x = A12 * b2 - A22 * b1;
+        delta.y = A12 * b1 - A11 * b2;
+
+        nextPt +=delta;
+
+        if (fabs(delta.x) < THRESHOLD && fabs(delta.y) < THRESHOLD)
+            break;
+    }
+
+    D = 0.0f;
+    if (calcErr)
+    {
+        yBase=yid;
+        {
+            xBase=xid;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[0], &D);
+
+
+            xBase+=xsize;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[1], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_patch[2], &D);
+        }
+        yBase+=ysize;
+        {
+            xBase=xid;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[3], &D);
+
+
+            xBase+=xsize;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[4], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_patch[5], &D);
+        }
+        yBase+=ysize;
+        if(yBase<c_winSize_y)
+        {
+            xBase=xid;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[6], &D);
+
+
+            xBase+=xsize;
+            GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                      &I_patch[7], &D);
+
+            xBase+=xsize;
+            if(xBase<c_winSize_x)
+                GetError4(J, nextPt.x + xBase + 0.5f, nextPt.y + yBase + 0.5f,
+                          &I_add, &D);
+        }
+
+        reduce1(D, smem1, tid);
+    }
+
+    if (tid == 0)
+    {
+        nextPt += c_halfWin;
+        nextPts[gid] = nextPt;
+
+        if (calcErr)
+            err[gid] = smem1[0] / (float)(3 * c_winSize_x * c_winSize_y);
+    }
+}
+
+__kernel void lkDense_C1_D0(image2d_t I, image2d_t J, __global float* u, int uStep, __global float* v, int vStep, __global const float* prevU, int prevUStep, __global const float* prevV, int prevVStep,
+                            const int rows, const int cols, /*__global float* err, int errStep, int cn,*/ int c_winSize_x, int c_winSize_y, int c_iters, char calcErr)
+{
+    int c_halfWin_x = (c_winSize_x - 1) / 2;
+    int c_halfWin_y = (c_winSize_y - 1) / 2;
+
+    const int patchWidth  = get_local_size(0) + 2 * c_halfWin_x;
+    const int patchHeight = get_local_size(1) + 2 * c_halfWin_y;
+
+    __local int smem[8192];
+
+    __local int* I_patch = smem;
+    __local int* dIdx_patch = I_patch + patchWidth * patchHeight;
+    __local int* dIdy_patch = dIdx_patch + patchWidth * patchHeight;
+
+    const int xBase = get_group_id(0) * get_local_size(0);
+    const int yBase = get_group_id(1) * get_local_size(1);
+
+    sampler_t sampleri    = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_NEAREST;
+
+    for (int i = get_local_id(1); i < patchHeight; i += get_local_size(1))
+    {
+        for (int j = get_local_id(0); j < patchWidth; j += get_local_size(0))
+        {
+            float x = xBase - c_halfWin_x + j + 0.5f;
+            float y = yBase - c_halfWin_y + i + 0.5f;
+
+            I_patch[i * patchWidth + j] = read_imagei(I, sampleri, (float2)(x, y)).x;
+
+            // Sharr Deriv
+
+            dIdx_patch[i * patchWidth + j] = 3 * read_imagei(I, sampleri, (float2)(x+1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x+1, y)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y+1)).x -
+                                             (3 * read_imagei(I, sampleri, (float2)(x-1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x-1, y)).x + 3 * read_imagei(I, sampleri, (float2)(x-1, y+1)).x);
+
+            dIdy_patch[i * patchWidth + j] = 3 * read_imagei(I, sampleri, (float2)(x-1, y+1)).x + 10 * read_imagei(I, sampleri, (float2)(x, y+1)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y+1)).x -
+                                             (3 * read_imagei(I, sampleri, (float2)(x-1, y-1)).x + 10 * read_imagei(I, sampleri, (float2)(x, y-1)).x + 3 * read_imagei(I, sampleri, (float2)(x+1, y-1)).x);
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    // extract the patch from the first image, compute covariation matrix of derivatives
+
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if (x >= cols || y >= rows)
+        return;
+
+    int A11i = 0;
+    int A12i = 0;
+    int A22i = 0;
+
+    for (int i = 0; i < c_winSize_y; ++i)
+    {
+        for (int j = 0; j < c_winSize_x; ++j)
+        {
+            int dIdx = dIdx_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
+            int dIdy = dIdy_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
+
+            A11i += dIdx * dIdx;
+            A12i += dIdx * dIdy;
+            A22i += dIdy * dIdy;
+        }
+    }
+
+    float A11 = A11i;
+    float A12 = A12i;
+    float A22 = A22i;
+
+    float D = A11 * A22 - A12 * A12;
+
+    //if (calcErr && GET_MIN_EIGENVALS)
+    //    (err + y * errStep)[x] = minEig;
+
+    if (D < 1.192092896e-07f)
+    {
+        //if (calcErr)
+        //    err(y, x) = 3.402823466e+38f;
+
+        return;
+    }
+
+    D = 1.f / D;
+
+    A11 *= D;
+    A12 *= D;
+    A22 *= D;
+
+    float2 nextPt;
+    nextPt.x = x + prevU[y/2 * prevUStep / 4 + x/2] * 2.0f;
+    nextPt.y = y + prevV[y/2 * prevVStep / 4 + x/2] * 2.0f;
+
+    for (int k = 0; k < c_iters; ++k)
+    {
+        if (nextPt.x < 0 || nextPt.x >= cols || nextPt.y < 0 || nextPt.y >= rows)
+        {
+            //if (calcErr)
+            //    err(y, x) = 3.402823466e+38f;
+
+            return;
+        }
+
+        int b1 = 0;
+        int b2 = 0;
+
+        for (int i = 0; i < c_winSize_y; ++i)
+        {
+            for (int j = 0; j < c_winSize_x; ++j)
+            {
+                int iI = I_patch[(get_local_id(1) + i) * patchWidth + get_local_id(0) + j];
+                int iJ = read_imagei(J, sampler, (float2)(nextPt.x - c_halfWin_x + j + 0.5f, nextPt.y - c_halfWin_y + i + 0.5f)).x;
+
+                int diff = (iJ - iI) * 32;
+
+                int dIdx = dIdx_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
+                int dIdy = dIdy_patch[(get_local_id(1) + i) * patchWidth + (get_local_id(0) + j)];
+
+                b1 += diff * dIdx;
+                b2 += diff * dIdy;
+            }
+        }
+
+        float2 delta;
+        delta.x = A12 * b2 - A22 * b1;
+        delta.y = A12 * b1 - A11 * b2;
+
+        nextPt.x += delta.x;
+        nextPt.y += delta.y;
+
+        if (fabs(delta.x) < 0.01f && fabs(delta.y) < 0.01f)
+            break;
+    }
+
+    u[y * uStep / 4 + x] = nextPt.x - x;
+    v[y * vStep / 4 + x] = nextPt.y - y;
+
+    if (calcErr)
+    {
+        int errval = 0;
+
+        for (int i = 0; i < c_winSize_y; ++i)
+        {
+            for (int j = 0; j < c_winSize_x; ++j)
+            {
+                int iI = I_patch[(get_local_id(1) + i) * patchWidth + get_local_id(0) + j];
+                int iJ = read_imagei(J, sampler, (float2)(nextPt.x - c_halfWin_x + j + 0.5f, nextPt.y - c_halfWin_y + i + 0.5f)).x;
+
+                errval += abs(iJ - iI);
+            }
+        }
+
+        //err[y * errStep / 4 + x] = static_cast<float>(errval) / (c_winSize_x * c_winSize_y);
+    }
+}
diff --git a/modules/ocl/src/kernels/pyrlk_no_image.cl b/modules/ocl/src/opencl/pyrlk_no_image.cl
similarity index 100%
rename from modules/ocl/src/kernels/pyrlk_no_image.cl
rename to modules/ocl/src/opencl/pyrlk_no_image.cl
diff --git a/modules/ocl/src/kernels/split_mat.cl b/modules/ocl/src/opencl/split_mat.cl
similarity index 87%
rename from modules/ocl/src/kernels/split_mat.cl
rename to modules/ocl/src/opencl/split_mat.cl
index 3c70859264..caee4366de 100644
--- a/modules/ocl/src/kernels/split_mat.cl
+++ b/modules/ocl/src/opencl/split_mat.cl
@@ -51,9 +51,9 @@
 ////////////vector fuction name format: split_vector_C(channels number)_D(data type depth)//////
 ////////////////////////////////////////////////////////////////////////////////////////////////
 __kernel void split_vector_C4_D0 (__global uchar *mat_src,  int src_step,  int src_offset,
-                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global uchar *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global uchar *mat_dst2, int dst2_step, int dst2_offset,
                                   __global uchar *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -61,37 +61,37 @@ __kernel void split_vector_C4_D0 (__global uchar *mat_src,  int src_step,  int s
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
-        int src_idx  = mad24(y, src_step, src_offset + (x << 2)); 
+        int src_idx  = mad24(y, src_step, src_offset + (x << 2));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x) & (int)0xfffffffc;
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x) & (int)0xfffffffc;
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + x) & (int)0xfffffffc;
 
-        int dst3_start = mad24(y, dst3_step, dst3_offset); 
+        int dst3_start = mad24(y, dst3_step, dst3_offset);
         int dst3_end   = mad24(y, dst3_step, dst3_offset + dst_step1);
         int dst3_idx   = mad24(y, dst3_step, dst3_offset + x) & (int)0xfffffffc;
-           
-        uchar4 data_0 = *((global uchar4 *)(mat_src + (src_idx - 12 >= 0 ? src_idx - 12 : src_idx))); 
-        uchar4 data_1 = *((global uchar4 *)(mat_src + (src_idx - 8  >= 0 ? src_idx - 8  : src_idx))); 
-        uchar4 data_2 = *((global uchar4 *)(mat_src + (src_idx - 4  >= 0 ? src_idx - 4  : src_idx))); 
-        uchar4 data_3 = *((global uchar4 *)(mat_src + src_idx + 0 )); 
 
-        int total_bytes = src_offset + rows * src_step; 
-        uchar4 data_4 = *((global uchar4 *)(mat_src + (src_idx + 4  < total_bytes ? src_idx + 4  : src_idx))); 
-        uchar4 data_5 = *((global uchar4 *)(mat_src + (src_idx + 8  < total_bytes ? src_idx + 8  : src_idx))); 
-        uchar4 data_6 = *((global uchar4 *)(mat_src + (src_idx + 12 < total_bytes ? src_idx + 12 : src_idx)));  
+        uchar4 data_0 = *((global uchar4 *)(mat_src + (src_idx - 12 >= 0 ? src_idx - 12 : src_idx)));
+        uchar4 data_1 = *((global uchar4 *)(mat_src + (src_idx - 8  >= 0 ? src_idx - 8  : src_idx)));
+        uchar4 data_2 = *((global uchar4 *)(mat_src + (src_idx - 4  >= 0 ? src_idx - 4  : src_idx)));
+        uchar4 data_3 = *((global uchar4 *)(mat_src + src_idx + 0 ));
+
+        int total_bytes = src_offset + rows * src_step;
+        uchar4 data_4 = *((global uchar4 *)(mat_src + (src_idx + 4  < total_bytes ? src_idx + 4  : src_idx)));
+        uchar4 data_5 = *((global uchar4 *)(mat_src + (src_idx + 8  < total_bytes ? src_idx + 8  : src_idx)));
+        uchar4 data_6 = *((global uchar4 *)(mat_src + (src_idx + 12 < total_bytes ? src_idx + 12 : src_idx)));
 
         uchar4 tmp_data0=1, tmp_data1=2, tmp_data2, tmp_data3;
 
@@ -164,33 +164,33 @@ __kernel void split_vector_C4_D0 (__global uchar *mat_src,  int src_step,  int s
 }
 
 __kernel void split_vector_C3_D0 (__global uchar *mat_src,  int src_step,  int src_offset,
-                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global uchar *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global uchar *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
-        int src_idx  = mad24(y, src_step, src_offset); 
+        int src_idx  = mad24(y, src_step, src_offset);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x  & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + x & (int)0xfffffffc);
-           
+
         uchar4 dst0_data  = *((__global uchar4 *)(mat_dst0 + dst0_idx));
         uchar4 dst1_data  = *((__global uchar4 *)(mat_dst1 + dst1_idx));
         uchar4 dst2_data  = *((__global uchar4 *)(mat_dst2 + dst2_idx));
@@ -227,10 +227,10 @@ __kernel void split_vector_C3_D0 (__global uchar *mat_src,  int src_step,  int s
 
         uchar data[7] = {src_data_0, src_data_3, src_data_6, src_data_9, src_data_12, src_data_15, src_data_18};
         int index = 3 - dst0_offset & 3;
-        tmp_data0 = (uchar4)(data[index], data[index + 1], data[index + 2], data[index + 3]); 
+        tmp_data0 = (uchar4)(data[index], data[index + 1], data[index + 2], data[index + 3]);
 
         uchar4 data0, data1, data2;
-        
+
         data0     = (uchar4)(src_data_1, src_data_4, src_data_7, src_data_10);
         data1     = (dst1_offset & 3) == 2 ? (uchar4)(src_data_4, src_data_7, src_data_10, src_data_13)  : data0;
         data2     = (dst1_offset & 3) == 1 ? (uchar4)(src_data_7, src_data_10, src_data_13, src_data_16) : data1;
@@ -263,33 +263,33 @@ __kernel void split_vector_C3_D0 (__global uchar *mat_src,  int src_step,  int s
 }
 
 __kernel void split_vector_C2_D0 (__global uchar *mat_src,  int src_step,  int src_offset,
-                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global uchar *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global uchar *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
         #define dst0_align ((dst0_offset & 3) << 1)
         #define dst1_align ((dst1_offset & 3) << 1)
-        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 1)); 
-        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 1)); 
+        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 1));
+        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 1));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x & (int)0xfffffffc);
-           
-		int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
-		int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
+
+        int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+        int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
         uchar8 src_data_0 = vload8(0, mat_src + src_idx_0);
         uchar8 src_data_1 = vload8(0, mat_src + src_idx_1);
         if(src_idx_0 == -6)
@@ -326,9 +326,9 @@ __kernel void split_vector_C2_D0 (__global uchar *mat_src,  int src_step,  int s
 }
 
 __kernel void split_vector_C4_D1 (__global char *mat_src,  int src_step,  int src_offset,
-                                  __global char *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global char *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global char *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global char *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global char *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global char *mat_dst2, int dst2_step, int dst2_offset,
                                   __global char *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -336,35 +336,35 @@ __kernel void split_vector_C4_D1 (__global char *mat_src,  int src_step,  int sr
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
-        int src_idx  = mad24(y, src_step, src_offset + (x << 2)); 
+        int src_idx  = mad24(y, src_step, src_offset + (x << 2));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + x & (int)0xfffffffc);
 
-        int dst3_start = mad24(y, dst3_step, dst3_offset); 
+        int dst3_start = mad24(y, dst3_step, dst3_offset);
         int dst3_end   = mad24(y, dst3_step, dst3_offset + dst_step1);
         int dst3_idx   = mad24(y, dst3_step, dst3_offset + x & (int)0xfffffffc);
-           
-        char4 data_0 = *((global char4 *)(mat_src + src_idx - 12)); 
-        char4 data_1 = *((global char4 *)(mat_src + src_idx - 8 )); 
-        char4 data_2 = *((global char4 *)(mat_src + src_idx - 4 )); 
-        char4 data_3 = *((global char4 *)(mat_src + src_idx + 0 )); 
-        char4 data_4 = *((global char4 *)(mat_src + src_idx + 4 )); 
-        char4 data_5 = *((global char4 *)(mat_src + src_idx + 8 )); 
-        char4 data_6 = *((global char4 *)(mat_src + src_idx + 12)); 
+
+        char4 data_0 = *((global char4 *)(mat_src + src_idx - 12));
+        char4 data_1 = *((global char4 *)(mat_src + src_idx - 8 ));
+        char4 data_2 = *((global char4 *)(mat_src + src_idx - 4 ));
+        char4 data_3 = *((global char4 *)(mat_src + src_idx + 0 ));
+        char4 data_4 = *((global char4 *)(mat_src + src_idx + 4 ));
+        char4 data_5 = *((global char4 *)(mat_src + src_idx + 8 ));
+        char4 data_6 = *((global char4 *)(mat_src + src_idx + 12));
 
         char4 tmp_data0=1, tmp_data1=2, tmp_data2, tmp_data3;
 
@@ -437,33 +437,33 @@ __kernel void split_vector_C4_D1 (__global char *mat_src,  int src_step,  int sr
 }
 
 __kernel void split_vector_C3_D1 (__global char *mat_src,  int src_step,  int src_offset,
-                                  __global char *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global char *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global char *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global char *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global char *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global char *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
-        int src_idx  = mad24(y, src_step, src_offset); 
+        int src_idx  = mad24(y, src_step, src_offset);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x  & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + x & (int)0xfffffffc);
-           
+
         char4 dst0_data  = *((__global char4 *)(mat_dst0 + dst0_idx));
         char4 dst1_data  = *((__global char4 *)(mat_dst1 + dst1_idx));
         char4 dst2_data  = *((__global char4 *)(mat_dst2 + dst2_idx));
@@ -500,10 +500,10 @@ __kernel void split_vector_C3_D1 (__global char *mat_src,  int src_step,  int sr
 
         char data[7] = {src_data_0, src_data_3, src_data_6, src_data_9, src_data_12, src_data_15, src_data_18};
         int index = 3 - dst0_offset & 3;
-        tmp_data0 = (char4)(data[index], data[index + 1], data[index + 2], data[index + 3]); 
+        tmp_data0 = (char4)(data[index], data[index + 1], data[index + 2], data[index + 3]);
 
         char4 data0, data1, data2;
-        
+
         data0     = (char4)(src_data_1, src_data_4, src_data_7, src_data_10);
         data1     = (dst1_offset & 3) == 2 ? (char4)(src_data_4, src_data_7, src_data_10, src_data_13)  : data0;
         data2     = (dst1_offset & 3) == 1 ? (char4)(src_data_7, src_data_10, src_data_13, src_data_16) : data1;
@@ -536,32 +536,32 @@ __kernel void split_vector_C3_D1 (__global char *mat_src,  int src_step,  int sr
 }
 
 __kernel void split_vector_C2_D1 (__global char *mat_src,  int src_step,  int src_offset,
-                                  __global char *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global char *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global char *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global char *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 2;
 
         #define dst0_align ((dst0_offset & 3) << 1)
         #define dst1_align ((dst1_offset & 3) << 1)
-        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 1)); 
-        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 1)); 
+        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 1));
+        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 1));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + x & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + x & (int)0xfffffffc);
-   	int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
-		int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
+    int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+        int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
         char8 src_data_0 = vload8(0, mat_src + src_idx_0);
         char8 src_data_1 = vload8(0, mat_src + src_idx_1);
         if(src_idx_0 == -6)
@@ -597,9 +597,9 @@ __kernel void split_vector_C2_D1 (__global char *mat_src,  int src_step,  int sr
 }
 
 __kernel void split_vector_C4_D2 (__global ushort *mat_src,  int src_step,  int src_offset,
-                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global ushort *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global ushort *mat_dst2, int dst2_step, int dst2_offset,
                                   __global ushort *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -607,30 +607,30 @@ __kernel void split_vector_C4_D2 (__global ushort *mat_src,  int src_step,  int
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
-        int src_idx_0  = mad24(y, src_step, src_offset + (x << 3) - 8); 
-        int src_idx_1  = mad24(y, src_step, src_offset + (x << 3) + 8); 
+        int src_idx_0  = mad24(y, src_step, src_offset + (x << 3) - 8);
+        int src_idx_1  = mad24(y, src_step, src_offset + (x << 3) + 8);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst3_start = mad24(y, dst3_step, dst3_offset); 
+        int dst3_start = mad24(y, dst3_step, dst3_offset);
         int dst3_end   = mad24(y, dst3_step, dst3_offset + dst_step1);
         int dst3_idx   = mad24(y, dst3_step, dst3_offset + (x << 1) & (int)0xfffffffc);
-           
-   	int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+
+    int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
         ushort8 src_data0 = vload8(0,(__global ushort *)((__global char *)mat_src + src_idx_0));
              if(src_idx_0 == -6)
             src_data0.s01234567 = src_data0.s67012345;
@@ -672,33 +672,33 @@ __kernel void split_vector_C4_D2 (__global ushort *mat_src,  int src_step,  int
 }
 
 __kernel void split_vector_C3_D2 (__global ushort *mat_src,  int src_step,  int src_offset,
-                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global ushort *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global ushort *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
-        int src_idx  = mad24(y, src_step, src_offset); 
+        int src_idx  = mad24(y, src_step, src_offset);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + (x << 1) & (int)0xfffffffc);
-           
+
         ushort2 dst0_data  = *((__global ushort2 *)((__global char *)mat_dst0 + dst0_idx));
         ushort2 dst1_data  = *((__global ushort2 *)((__global char *)mat_dst1 + dst1_idx));
         ushort2 dst2_data  = *((__global ushort2 *)((__global char *)mat_dst2 + dst2_idx));
@@ -735,48 +735,48 @@ __kernel void split_vector_C3_D2 (__global ushort *mat_src,  int src_step,  int
 }
 
 __kernel void split_vector_C2_D2 (__global ushort *mat_src,  int src_step,  int src_offset,
-                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global ushort *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global ushort *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
         #define dst0_align ((dst0_offset & 3) << 1)
         #define dst1_align ((dst1_offset & 3) << 1)
-        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 2)); 
-        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 2)); 
+        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 2));
+        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 2));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
-           
-		int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
-		int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
+
+        int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+        int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
         ushort4 src_data_0 = vload4(0, (__global ushort *)((__global char *)mat_src + src1_index_fix));
         ushort4 src_data_1 = vload4(0, (__global ushort *)((__global char *)mat_src + src2_index_fix));
-		if(src_idx_0 < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src_idx_0 == -2) ? src_data_0.zwxy : src_data_0.yzwx;
-			src_data_0.xyzw = (src_idx_1 == -1) ? src_data_0.wxyz:tmp.xyzw;
-		}
-		if(src_idx_1 < 0)
-		{
-			ushort4 tmp;
-			tmp.xyzw = (src_idx_1 == -2) ? src_data_1.zwxy : src_data_1.yzwx;
-			src_data_1.xyzw = (src_idx_1 == -1) ? src_data_1.wxyz : tmp.xyzw;
-		}		
-  
+        if(src_idx_0 < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src_idx_0 == -2) ? src_data_0.zwxy : src_data_0.yzwx;
+            src_data_0.xyzw = (src_idx_1 == -1) ? src_data_0.wxyz:tmp.xyzw;
+        }
+        if(src_idx_1 < 0)
+        {
+            ushort4 tmp;
+            tmp.xyzw = (src_idx_1 == -2) ? src_data_1.zwxy : src_data_1.yzwx;
+            src_data_1.xyzw = (src_idx_1 == -1) ? src_data_1.wxyz : tmp.xyzw;
+        }
+
         ushort2 dst0_data  = *((__global ushort2 *)((__global char *)mat_dst0 + dst0_idx));
         ushort2 dst1_data  = *((__global ushort2 *)((__global char *)mat_dst1 + dst1_idx));
 
@@ -793,9 +793,9 @@ __kernel void split_vector_C2_D2 (__global ushort *mat_src,  int src_step,  int
     }
 }
 __kernel void split_vector_C4_D3 (__global short *mat_src,  int src_step,  int src_offset,
-                                  __global short *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global short *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global short *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global short *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global short *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global short *mat_dst2, int dst2_step, int dst2_offset,
                                   __global short *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -803,38 +803,38 @@ __kernel void split_vector_C4_D3 (__global short *mat_src,  int src_step,  int s
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
-        int src_idx_0  = mad24(y, src_step, src_offset + (x << 3) - 8); 
-        int src_idx_1  = mad24(y, src_step, src_offset + (x << 3) + 8); 
+        int src_idx_0  = mad24(y, src_step, src_offset + (x << 3) - 8);
+        int src_idx_1  = mad24(y, src_step, src_offset + (x << 3) + 8);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst3_start = mad24(y, dst3_step, dst3_offset); 
+        int dst3_start = mad24(y, dst3_step, dst3_offset);
         int dst3_end   = mad24(y, dst3_step, dst3_offset + dst_step1);
         int dst3_idx   = mad24(y, dst3_step, dst3_offset + (x << 1) & (int)0xfffffffc);
-     	int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+        int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
         short8 src_data0 = vload8(0,(__global short *)((__global char *)mat_src + src_idx_0));
- 
+
         if(src_idx_0 == -6)
             src_data0.s01234567 = src_data0.s67012345;
         if(src_idx_0 == -4)
             src_data0.s01234567 = src_data0.s45670123;
         if(src_idx_0 == -2)
             src_data0.s01234567 = src_data0.s23456701;
-          
+
         short4 src_data1 = *((__global short4 *)((__global char *)mat_src + src_idx_1));
 
         short2 dst0_data  = *((__global short2 *)((__global char *)mat_dst0 + dst0_idx));
@@ -868,33 +868,33 @@ __kernel void split_vector_C4_D3 (__global short *mat_src,  int src_step,  int s
     }
 }
 __kernel void split_vector_C3_D3 (__global short *mat_src,  int src_step,  int src_offset,
-                                  __global short *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global short *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global short *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global short *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global short *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global short *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
-        int src_idx  = mad24(y, src_step, src_offset); 
+        int src_idx  = mad24(y, src_step, src_offset);
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst2_start = mad24(y, dst2_step, dst2_offset); 
+        int dst2_start = mad24(y, dst2_step, dst2_offset);
         int dst2_end   = mad24(y, dst2_step, dst2_offset + dst_step1);
         int dst2_idx   = mad24(y, dst2_step, dst2_offset + (x << 1) & (int)0xfffffffc);
-           
+
         short2 dst0_data  = *((__global short2 *)((__global char *)mat_dst0 + dst0_idx));
         short2 dst1_data  = *((__global short2 *)((__global char *)mat_dst1 + dst1_idx));
         short2 dst2_data  = *((__global short2 *)((__global char *)mat_dst2 + dst2_idx));
@@ -932,47 +932,47 @@ __kernel void split_vector_C3_D3 (__global short *mat_src,  int src_step,  int s
 
 
 __kernel void split_vector_C2_D3 (__global short *mat_src,  int src_step,  int src_offset,
-                                  __global short *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global short *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global short *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global short *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
         x = x << 1;
 
         #define dst0_align ((dst0_offset & 3) << 1)
         #define dst1_align ((dst1_offset & 3) << 1)
-        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 2)); 
-        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 2)); 
+        int src_idx_0  = mad24(y, src_step, src_offset - dst0_align + (x << 2));
+        int src_idx_1  = mad24(y, src_step, src_offset - dst1_align + (x << 2));
 
-        int dst0_start = mad24(y, dst0_step, dst0_offset); 
+        int dst0_start = mad24(y, dst0_step, dst0_offset);
         int dst0_end   = mad24(y, dst0_step, dst0_offset + dst_step1);
         int dst0_idx   = mad24(y, dst0_step, dst0_offset + (x << 1) & (int)0xfffffffc);
 
-        int dst1_start = mad24(y, dst1_step, dst1_offset); 
+        int dst1_start = mad24(y, dst1_step, dst1_offset);
         int dst1_end   = mad24(y, dst1_step, dst1_offset + dst_step1);
         int dst1_idx   = mad24(y, dst1_step, dst1_offset + (x << 1) & (int)0xfffffffc);
- 		int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
-		int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
+        int src1_index_fix = src_idx_0 < 0 ? 0 : src_idx_0;
+        int src2_index_fix = src_idx_1 < 0 ? 0 : src_idx_1;
         short4 src_data_0 = vload4(0, (__global short *)((__global char *)mat_src + src_idx_0));
         short4 src_data_1 = vload4(0, (__global short *)((__global char *)mat_src + src_idx_1));
-		if(src_idx_0 < 0)
-		{
-			short4 tmp;
-			tmp.xyzw = (src_idx_0 == -2) ? src_data_0.zwxy : src_data_0.yzwx;
-			src_data_0.xyzw = (src_idx_0 == -1) ? src_data_0.wxyz:tmp.xyzw;
-		}
-		if(src_idx_1< 0)
-		{
-			short4 tmp;
-			tmp.xyzw = ( src_idx_1== -2) ? src_data_1.zwxy : src_data_1.yzwx;
-			src_data_1.xyzw = ( src_idx_1== -1) ? src_data_1.wxyz : tmp.xyzw;
-		}		
-             
+        if(src_idx_0 < 0)
+        {
+            short4 tmp;
+            tmp.xyzw = (src_idx_0 == -2) ? src_data_0.zwxy : src_data_0.yzwx;
+            src_data_0.xyzw = (src_idx_0 == -1) ? src_data_0.wxyz:tmp.xyzw;
+        }
+        if(src_idx_1< 0)
+        {
+            short4 tmp;
+            tmp.xyzw = ( src_idx_1== -2) ? src_data_1.zwxy : src_data_1.yzwx;
+            src_data_1.xyzw = ( src_idx_1== -1) ? src_data_1.wxyz : tmp.xyzw;
+        }
+
 
         short2 dst0_data  = *((__global short2 *)((__global char *)mat_dst0 + dst0_idx));
         short2 dst1_data  = *((__global short2 *)((__global char *)mat_dst1 + dst1_idx));
@@ -990,9 +990,9 @@ __kernel void split_vector_C2_D3 (__global short *mat_src,  int src_step,  int s
     }
 }
 __kernel void split_vector_C4_D4 (__global int *mat_src,  int src_step,  int src_offset,
-                                  __global int *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global int *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global int *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global int *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global int *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global int *mat_dst2, int dst2_step, int dst2_offset,
                                   __global int *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -1000,14 +1000,14 @@ __kernel void split_vector_C4_D4 (__global int *mat_src,  int src_step,  int src
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
         int dst3_idx = mad24(y, dst3_step, dst3_offset);
-           
+
         int4 src_data = ((__global int4 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global int *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
@@ -1017,18 +1017,18 @@ __kernel void split_vector_C4_D4 (__global int *mat_src,  int src_step,  int src
     }
 }
 __kernel void split_vector_C3_D4 (__global int *mat_src,  int src_step,  int src_offset,
-                                  __global int *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global int *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global int *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global int *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global int *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global int *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
@@ -1044,20 +1044,20 @@ __kernel void split_vector_C3_D4 (__global int *mat_src,  int src_step,  int src
 }
 
 __kernel void split_vector_C2_D4 (__global int *mat_src,  int src_step,  int src_offset,
-                                  __global int *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global int *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global int *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global int *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
-           
+
         int2 src_data = ((__global int2 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global int *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
@@ -1066,9 +1066,9 @@ __kernel void split_vector_C2_D4 (__global int *mat_src,  int src_step,  int src
 }
 
 __kernel void split_vector_C4_D5 (__global float *mat_src,  int src_step,  int src_offset,
-                                  __global float *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global float *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global float *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global float *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global float *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global float *mat_dst2, int dst2_step, int dst2_offset,
                                   __global float *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -1076,14 +1076,14 @@ __kernel void split_vector_C4_D5 (__global float *mat_src,  int src_step,  int s
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
         int dst3_idx = mad24(y, dst3_step, dst3_offset);
-           
+
         float4 src_data = ((__global float4 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global float *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
@@ -1094,18 +1094,18 @@ __kernel void split_vector_C4_D5 (__global float *mat_src,  int src_step,  int s
 }
 
 __kernel void split_vector_C3_D5 (__global float *mat_src,  int src_step,  int src_offset,
-                                  __global float *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global float *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global float *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global float *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global float *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global float *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
@@ -1121,20 +1121,20 @@ __kernel void split_vector_C3_D5 (__global float *mat_src,  int src_step,  int s
 }
 
 __kernel void split_vector_C2_D5 (__global float *mat_src,  int src_step,  int src_offset,
-                                  __global float *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global float *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global float *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global float *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
-           
+
         float2 src_data = ((__global float2 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global float *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
@@ -1144,9 +1144,9 @@ __kernel void split_vector_C2_D5 (__global float *mat_src,  int src_step,  int s
 
 #if defined (DOUBLE_SUPPORT)
 __kernel void split_vector_C4_D6 (__global double *mat_src,  int src_step,  int src_offset,
-                                  __global double *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global double *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global double *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global double *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global double *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global double *mat_dst2, int dst2_step, int dst2_offset,
                                   __global double *mat_dst3, int dst3_step, int dst3_offset,
                                   int rows, int cols, int dst_step1)
 
@@ -1154,14 +1154,14 @@ __kernel void split_vector_C4_D6 (__global double *mat_src,  int src_step,  int
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
         int dst3_idx = mad24(y, dst3_step, dst3_offset);
-           
+
         double4 src_data = ((__global double4 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global double *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
@@ -1172,18 +1172,18 @@ __kernel void split_vector_C4_D6 (__global double *mat_src,  int src_step,  int
 }
 
 __kernel void split_vector_C3_D6 (__global double *mat_src,  int src_step,  int src_offset,
-                                  __global double *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global double *mat_dst1, int dst1_step, int dst1_offset, 
-	                                __global double *mat_dst2, int dst2_step, int dst2_offset,  
+                                  __global double *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global double *mat_dst1, int dst1_step, int dst1_offset,
+                                    __global double *mat_dst2, int dst2_step, int dst2_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
         int dst2_idx = mad24(y, dst2_step, dst2_offset);
@@ -1199,20 +1199,20 @@ __kernel void split_vector_C3_D6 (__global double *mat_src,  int src_step,  int
 }
 
 __kernel void split_vector_C2_D6 (__global double *mat_src,  int src_step,  int src_offset,
-                                  __global double *mat_dst0, int dst0_step, int dst0_offset,  
-                                  __global double *mat_dst1, int dst1_step, int dst1_offset, 
+                                  __global double *mat_dst0, int dst0_step, int dst0_offset,
+                                  __global double *mat_dst1, int dst1_step, int dst1_offset,
                                   int rows, int cols, int dst_step1)
 
 {
     int x = get_global_id(0);
     int y = get_global_id(1);
 
-    if((x  < cols) && (y < rows)) 
+    if((x  < cols) && (y < rows))
     {
-        int src_idx  = mad24(y, src_step,  src_offset); 
+        int src_idx  = mad24(y, src_step,  src_offset);
         int dst0_idx = mad24(y, dst0_step, dst0_offset);
         int dst1_idx = mad24(y, dst1_step, dst1_offset);
-           
+
         double2 src_data = ((__global double2 *)((__global char *)mat_src + src_idx))[x];
 
         ((__global double *)((__global char *)mat_dst0 + dst0_idx))[x] = src_data.x;
diff --git a/modules/ocl/src/opencl/stereobm.cl b/modules/ocl/src/opencl/stereobm.cl
new file mode 100644
index 0000000000..56f445e429
--- /dev/null
+++ b/modules/ocl/src/opencl/stereobm.cl
@@ -0,0 +1,339 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Sen Liu, swjtuls1987@126.com
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#define ROWSperTHREAD 21     // the number of rows a thread will process
+#define BLOCK_W       128    // the thread block width (464)
+#define N_DISPARITIES 8
+
+#define STEREO_MIND 0                    // The minimum d range to check
+#define STEREO_DISP_STEP N_DISPARITIES   // the d step, must be <= 1 to avoid aliasing
+
+#ifndef radius
+#define radius 64
+#endif
+
+unsigned int CalcSSD(__local unsigned int *col_ssd)
+{
+    unsigned int cache = col_ssd[0];
+
+#pragma unroll
+    for(int i = 1; i <= (radius << 1); i++)
+        cache += col_ssd[i];
+
+    return cache;
+}
+
+uint2 MinSSD(__local unsigned int *col_ssd)
+{
+    unsigned int ssd[N_DISPARITIES];
+    const int win_size = (radius << 1);
+
+    //See above:  #define COL_SSD_SIZE (BLOCK_W + WIN_SIZE)
+    ssd[0] = CalcSSD(col_ssd + 0 * (BLOCK_W + win_size));
+    ssd[1] = CalcSSD(col_ssd + 1 * (BLOCK_W + win_size));
+    ssd[2] = CalcSSD(col_ssd + 2 * (BLOCK_W + win_size));
+    ssd[3] = CalcSSD(col_ssd + 3 * (BLOCK_W + win_size));
+    ssd[4] = CalcSSD(col_ssd + 4 * (BLOCK_W + win_size));
+    ssd[5] = CalcSSD(col_ssd + 5 * (BLOCK_W + win_size));
+    ssd[6] = CalcSSD(col_ssd + 6 * (BLOCK_W + win_size));
+    ssd[7] = CalcSSD(col_ssd + 7 * (BLOCK_W + win_size));
+
+    unsigned int mssd = min(min(min(ssd[0], ssd[1]), min(ssd[4], ssd[5])), min(min(ssd[2], ssd[3]), min(ssd[6], ssd[7])));
+
+    int bestIdx = 0;
+
+    for (int i = 0; i < N_DISPARITIES; i++)
+    {
+        if (mssd == ssd[i])
+            bestIdx = i;
+    }
+
+    return (uint2)(mssd, bestIdx);
+}
+
+void StepDown(int idx1, int idx2, __global unsigned char* imageL,
+              __global unsigned char* imageR, int d,   __local unsigned int *col_ssd)
+{
+    uint8 imgR1 = convert_uint8(vload8(0, imageR + (idx1 - d - 7)));
+    uint8 imgR2 = convert_uint8(vload8(0, imageR + (idx2 - d - 7)));
+    uint8 diff1 = (uint8)(imageL[idx1]) - imgR1;
+    uint8 diff2 = (uint8)(imageL[idx2]) - imgR2;
+    uint8 res = diff2 * diff2 - diff1 * diff1;
+    const int win_size = (radius << 1);
+    col_ssd[0 * (BLOCK_W + win_size)] += res.s7;
+    col_ssd[1 * (BLOCK_W + win_size)] += res.s6;
+    col_ssd[2 * (BLOCK_W + win_size)] += res.s5;
+    col_ssd[3 * (BLOCK_W + win_size)] += res.s4;
+    col_ssd[4 * (BLOCK_W + win_size)] += res.s3;
+    col_ssd[5 * (BLOCK_W + win_size)] += res.s2;
+    col_ssd[6 * (BLOCK_W + win_size)] += res.s1;
+    col_ssd[7 * (BLOCK_W + win_size)] += res.s0;
+}
+
+void InitColSSD(int x_tex, int y_tex, int im_pitch, __global unsigned char* imageL,
+                __global unsigned char* imageR, int d,
+                 __local unsigned int *col_ssd)
+{
+    uint8 leftPixel1;
+    uint8 diffa = 0;
+    int idx = y_tex * im_pitch + x_tex;
+    const int win_size = (radius << 1);
+    for(int i = 0; i < (win_size + 1); i++)
+    {
+        leftPixel1 = (uint8)(imageL[idx]);
+        uint8 imgR = convert_uint8(vload8(0, imageR + (idx - d - 7)));
+        uint8 res = leftPixel1 - imgR;
+        diffa += res * res;
+
+        idx += im_pitch;
+    }
+    //See above:  #define COL_SSD_SIZE (BLOCK_W + WIN_SIZE)
+    col_ssd[0 * (BLOCK_W + win_size)] = diffa.s7;
+    col_ssd[1 * (BLOCK_W + win_size)] = diffa.s6;
+    col_ssd[2 * (BLOCK_W + win_size)] = diffa.s5;
+    col_ssd[3 * (BLOCK_W + win_size)] = diffa.s4;
+    col_ssd[4 * (BLOCK_W + win_size)] = diffa.s3;
+    col_ssd[5 * (BLOCK_W + win_size)] = diffa.s2;
+    col_ssd[6 * (BLOCK_W + win_size)] = diffa.s1;
+    col_ssd[7 * (BLOCK_W + win_size)] = diffa.s0;
+}
+
+__kernel void stereoKernel(__global unsigned char *left, __global unsigned char *right,
+                           __global unsigned int *cminSSDImage, int cminSSD_step,
+                           __global unsigned char *disp, int disp_step,int cwidth, int cheight,
+                           int img_step, int maxdisp,
+                           __local unsigned int *col_ssd_cache)
+{
+    __local unsigned int *col_ssd = col_ssd_cache + get_local_id(0);
+    __local unsigned int *col_ssd_extra = get_local_id(0) < (radius << 1) ? col_ssd + BLOCK_W : 0;
+
+    int X = get_group_id(0) * BLOCK_W + get_local_id(0) + maxdisp + radius;
+
+#define Y (get_group_id(1) * ROWSperTHREAD + radius)
+
+    __global unsigned int* minSSDImage = cminSSDImage + X + Y * cminSSD_step;
+    __global unsigned char* disparImage = disp + X + Y * disp_step;
+
+    int end_row = ROWSperTHREAD < (cheight - Y) ? ROWSperTHREAD:(cheight - Y);
+    int y_tex;
+    int x_tex = X - radius;
+
+    //if (x_tex >= cwidth)
+    //    return;
+
+    for(int d = STEREO_MIND; d < maxdisp; d += STEREO_DISP_STEP)
+    {
+        y_tex = Y - radius;
+
+        InitColSSD(x_tex, y_tex, img_step, left, right, d, col_ssd);
+        if (col_ssd_extra > 0)
+            if (x_tex + BLOCK_W < cwidth)
+                InitColSSD(x_tex + BLOCK_W, y_tex, img_step, left, right, d, col_ssd_extra);
+
+        barrier(CLK_LOCAL_MEM_FENCE); //before MinSSD function
+
+        uint2 minSSD = MinSSD(col_ssd);
+        if (X < cwidth - radius && Y < cheight - radius)
+        {
+            if (minSSD.x < minSSDImage[0])
+            {
+                disparImage[0] = (unsigned char)(d + minSSD.y);
+                minSSDImage[0] = minSSD.x;
+            }
+        }
+
+        for(int row = 1; row < end_row; row++)
+        {
+            int idx1 = y_tex * img_step + x_tex;
+            int idx2 = min(y_tex + ((radius << 1) + 1), cheight - 1) * img_step + x_tex;
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            StepDown(idx1, idx2, left, right, d, col_ssd);
+            if (col_ssd_extra > 0)
+                if (x_tex + BLOCK_W < cwidth)
+                    StepDown(idx1, idx2, left + BLOCK_W, right + BLOCK_W, d, col_ssd_extra);
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            uint2 minSSD = MinSSD(col_ssd);
+            if (X < cwidth - radius && row < cheight - radius - Y)
+            {
+                int idx = row * cminSSD_step;
+                if (minSSD.x < minSSDImage[idx])
+                {
+                    disparImage[disp_step * row] = (unsigned char)(d + minSSD.y);
+                    minSSDImage[idx] = minSSD.x;
+                }
+            }
+
+            y_tex++;
+        } // for row loop
+    } // for d loop
+}
+//////////////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////// Sobel Prefiler (signal channel)//////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////
+
+__kernel void prefilter_xsobel(__global unsigned char *input, __global unsigned char *output,
+                               int rows, int cols, int prefilterCap)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < cols && y < rows)
+    {
+        int cov = input[(y-1) * cols + (x-1)] * (-1) + input[(y-1) * cols + (x+1)] * (1) +
+                  input[(y)   * cols + (x-1)] * (-2) + input[(y)   * cols + (x+1)] * (2) +
+                  input[(y+1) * cols + (x-1)] * (-1) + input[(y+1) * cols + (x+1)] * (1);
+
+        cov = min(min(max(-prefilterCap, cov), prefilterCap) + prefilterCap, 255);
+        output[y * cols + x] = cov & 0xFF;
+    }
+}
+
+
+//////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////// Textureness filtering ////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////
+
+float sobel(__global unsigned char *input, int x, int y, int rows, int cols)
+{
+    float conv = 0;
+    int y1 = y==0? 0 : y-1;
+    int x1 = x==0? 0 : x-1;
+    if(x < cols && y < rows && x > 0 && y > 0)
+    {
+        conv = (float)input[(y1)  * cols + (x1)] * (-1) + (float)input[(y1)  * cols + (x+1)] * (1) +
+               (float)input[(y)   * cols + (x1)] * (-2) + (float)input[(y)   * cols + (x+1)] * (2) +
+               (float)input[(y+1) * cols + (x1)] * (-1) + (float)input[(y+1) * cols + (x+1)] * (1);
+
+    }
+    return fabs(conv);
+}
+
+float CalcSums(__local float *cols, __local float *cols_cache, int winsz)
+{
+    unsigned int cache = cols[0];
+
+#pragma unroll
+    for(int i = 1; i <= winsz; i++)
+        cache += cols[i];
+
+    return cache;
+}
+
+#define RpT (2 * ROWSperTHREAD)  // got experimentally
+__kernel void textureness_kernel(__global unsigned char *disp, int disp_rows, int disp_cols,
+                                 int disp_step, __global unsigned char *input, int input_rows,
+                                 int input_cols,int winsz, float threshold,
+                                 __local float *cols_cache)
+{
+    int winsz2 = winsz/2;
+    int n_dirty_pixels = (winsz2) * 2;
+
+    int local_id_x = get_local_id(0);
+    int group_size_x = get_local_size(0);
+    int group_id_y = get_group_id(1);
+
+    __local float *cols = cols_cache + group_size_x + local_id_x;
+    __local float *cols_extra = local_id_x < n_dirty_pixels ? cols + group_size_x : 0;
+
+    int x = get_global_id(0);
+    int beg_row = group_id_y * RpT;
+    int end_row = min(beg_row + RpT, disp_rows);
+
+
+    int y = beg_row;
+
+    float sum = 0;
+    float sum_extra = 0;
+
+    for(int i = y - winsz2; i <= y + winsz2; ++i)
+    {
+        sum += sobel(input, x - winsz2, i, input_rows, input_cols);
+        if (cols_extra)
+            sum_extra += sobel(input, x + group_size_x - winsz2, i, input_rows, input_cols);
+    }
+    *cols = sum;
+    if (cols_extra)
+        *cols_extra = sum_extra;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    float sum_win = CalcSums(cols, cols_cache + local_id_x, winsz) * 255;
+    if (sum_win < threshold)
+        disp[y * disp_step + x] = 0;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for(int y = beg_row + 1; y < end_row; ++y)
+    {
+        sum = sum - sobel(input, x - winsz2, y - winsz2 - 1, input_rows, input_cols) +
+              sobel(input, x - winsz2, y + winsz2, input_rows, input_cols);
+        *cols = sum;
+
+        if (cols_extra)
+        {
+            sum_extra = sum_extra - sobel(input, x + group_size_x - winsz2, y - winsz2 - 1,input_rows, input_cols)
+                        + sobel(input, x + group_size_x - winsz2, y + winsz2, input_rows, input_cols);
+            *cols_extra = sum_extra;
+        }
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+
+        if (x < disp_cols)
+        {
+            float sum_win = CalcSums(cols, cols_cache + local_id_x, winsz) * 255;
+            if (sum_win < threshold)
+                disp[y * disp_step + x] = 0;
+        }
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+    }
+
+}
diff --git a/modules/ocl/src/opencl/stereobp.cl b/modules/ocl/src/opencl/stereobp.cl
new file mode 100644
index 0000000000..1d523e7885
--- /dev/null
+++ b/modules/ocl/src/opencl/stereobp.cl
@@ -0,0 +1,395 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Peng Xiao,   pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other GpuMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#if defined (DOUBLE_SUPPORT)
+
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64:enable
+#elif defined (cl_amd_fp64)
+#pragma OPENCL EXTENSION cl_amd_fp64:enable
+#endif
+
+#endif
+
+#ifdef T_FLOAT
+#define T float
+#define T4 float4
+#else
+#define T short
+#define T4 short4
+#endif
+
+///////////////////////////////////////////////////////////////
+/////////////////common///////////////////////////////////////
+/////////////////////////////////////////////////////////////
+T saturate_cast(float v){
+#ifdef T_SHORT
+    return convert_short_sat_rte(v);
+#else
+    return v;
+#endif
+}
+
+T4 saturate_cast4(float4 v){
+#ifdef T_SHORT
+    return convert_short4_sat_rte(v);
+#else
+    return v;
+#endif
+}
+
+#define FLOAT_MAX 3.402823466e+38f
+typedef struct
+{
+    int   cndisp;
+    float cmax_data_term;
+    float cdata_weight;
+    float cmax_disc_term;
+    float cdisc_single_jump;
+}con_srtuct_t;
+///////////////////////////////////////////////////////////////
+////////////////////////// comp data //////////////////////////
+///////////////////////////////////////////////////////////////
+
+inline float pix_diff_1(const uchar4 l, __global const uchar *rs)
+{
+    return abs((int)(l.x) - *rs);
+}
+
+float pix_diff_4(const uchar4 l, __global const uchar *rs)
+{
+    uchar4 r;
+    r = *((__global uchar4 *)rs);
+
+    const float tr = 0.299f;
+    const float tg = 0.587f;
+    const float tb = 0.114f;
+
+    float val;
+
+    val  = tb * abs((int)l.x - r.x);
+    val += tg * abs((int)l.y - r.y);
+    val += tr * abs((int)l.z - r.z);
+
+    return val;
+}
+
+inline float pix_diff_3(const uchar4 l, __global const uchar *rs)
+{
+    return pix_diff_4(l, rs);
+}
+
+#ifndef CN
+#define CN 4
+#endif
+
+#ifndef CNDISP
+#define CNDISP 64
+#endif
+
+#define CAT(X,Y) X##Y
+#define CAT2(X,Y) CAT(X,Y)
+
+#define PIX_DIFF CAT2(pix_diff_, CN)
+
+__kernel void comp_data(__global uchar *left,  int left_rows,  int left_cols,  int left_step,
+                        __global uchar *right, int right_step,
+                        __global T *data, int data_step,
+                        __constant con_srtuct_t *con_st)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y > 0 && y < (left_rows - 1) && x > 0 && x < (left_cols - 1))
+    {
+        data_step /= sizeof(T);
+        const __global uchar* ls = left  + y * left_step  + x * CN;
+        const __global uchar* rs = right + y * right_step + x * CN;
+
+        __global T *ds = data + y * data_step + x;
+
+        const unsigned int disp_step = data_step * left_rows;
+        const float weightXterm = con_st -> cdata_weight * con_st -> cmax_data_term;
+        const uchar4 ls_data = vload4(0, ls);
+
+        for (int disp = 0; disp < con_st -> cndisp; disp++)
+        {
+            if (x - disp >= 1)
+            {
+                float val = 0;
+                val = PIX_DIFF(ls_data, rs - disp * CN);
+                ds[disp * disp_step] =  saturate_cast(fmin(con_st -> cdata_weight * val, weightXterm));
+            }
+            else
+            {
+                ds[disp * disp_step] =  saturate_cast(weightXterm);
+            }
+        }
+    }
+}
+
+///////////////////////////////////////////////////////////////
+//////////////////////// data step down ///////////////////////
+///////////////////////////////////////////////////////////////
+__kernel void data_step_down(__global T *src, int src_rows,
+                             __global T *dst, int dst_rows, int dst_cols,
+                             int src_step, int dst_step,
+                             int cndisp)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if (x < dst_cols && y < dst_rows)
+    {
+        src_step /= sizeof(T);
+        dst_step /= sizeof(T);
+        int4 coor_step = (int4)(src_rows * src_step);
+        int4 coor = (int4)(min(2*y+0, src_rows-1) * src_step + 2*x+0,
+                           min(2*y+1, src_rows-1) * src_step + 2*x+0,
+                           min(2*y+0, src_rows-1) * src_step + 2*x+1,
+                           min(2*y+1, src_rows-1) * src_step + 2*x+1);
+
+        for (int d = 0; d < cndisp; ++d)
+        {
+            float dst_reg;
+            dst_reg  = src[coor.x];
+            dst_reg += src[coor.y];
+            dst_reg += src[coor.z];
+            dst_reg += src[coor.w];
+            coor += coor_step;
+
+            dst[(d * dst_rows + y) * dst_step + x] = saturate_cast(dst_reg);
+        }
+    }
+}
+
+///////////////////////////////////////////////////////////////
+/////////////////// level up messages  ////////////////////////
+///////////////////////////////////////////////////////////////
+__kernel void level_up_message(__global T *src, int src_rows, int src_step,
+                               __global T *dst, int dst_rows, int dst_cols, int dst_step,
+                               int cndisp)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if (x < dst_cols && y < dst_rows)
+    {
+        src_step /= sizeof(T);
+        dst_step /= sizeof(T);
+
+        const int dst_disp_step = dst_step * dst_rows;
+        const int src_disp_step = src_step * src_rows;
+
+        __global T       *dstr = dst + y * dst_step + x;
+        __global const T *srcr = src + (y / 2 * src_step) + (x / 2);
+
+        for (int d = 0; d < cndisp; ++d)
+            dstr[d * dst_disp_step] = srcr[d * src_disp_step];
+    }
+}
+
+///////////////////////////////////////////////////////////////
+////////////////////  calc all iterations /////////////////////
+///////////////////////////////////////////////////////////////
+void message(__global T *us_, __global T *ds_, __global T *ls_, __global T *rs_,
+              const __global T *dt,
+              int u_step, int msg_disp_step, int data_disp_step,
+              float4 cmax_disc_term, float4 cdisc_single_jump)
+{
+    __global T *us = us_ + u_step;
+    __global T *ds = ds_ - u_step;
+    __global T *ls = ls_ + 1;
+    __global T *rs = rs_ - 1;
+
+    float4 minimum = (float4)(FLOAT_MAX);
+
+    T4 t_dst[CNDISP];
+    float4 dst_reg;
+    float4 prev;
+    float4 cur;
+
+    T t_us = us[0];
+    T t_ds = ds[0];
+    T t_ls = ls[0];
+    T t_rs = rs[0];
+    T t_dt = dt[0];
+
+    prev = (float4)(t_us + t_ls + t_rs + t_dt,
+                    t_ds + t_ls + t_rs + t_dt,
+                    t_us + t_ds + t_rs + t_dt,
+                    t_us + t_ds + t_ls + t_dt);
+
+    minimum = min(prev, minimum);
+
+    t_dst[0] = saturate_cast4(prev);
+
+    for(int i = 1, idx = msg_disp_step; i < CNDISP; ++i, idx+=msg_disp_step)
+    {
+        t_us = us[idx];
+        t_ds = ds[idx];
+        t_ls = ls[idx];
+        t_rs = rs[idx];
+        t_dt = dt[data_disp_step * i];
+
+        dst_reg = (float4)(t_us + t_ls + t_rs + t_dt,
+                           t_ds + t_ls + t_rs + t_dt,
+                           t_us + t_ds + t_rs + t_dt,
+                           t_us + t_ds + t_ls + t_dt);
+
+        minimum = min(dst_reg, minimum);
+
+        prev += cdisc_single_jump;
+        prev = min(prev, dst_reg);
+
+        t_dst[i] = saturate_cast4(prev);
+    }
+
+    minimum += cmax_disc_term;
+
+    float4 sum = 0;
+    prev = convert_float4(t_dst[CNDISP - 1]);
+    for (int disp = CNDISP - 2; disp >= 0; disp--)
+    {
+        prev += cdisc_single_jump;
+        cur = convert_float4(t_dst[disp]);
+        prev = min(prev, cur);
+        cur = min(prev, minimum);
+        sum += cur;
+
+        t_dst[disp] = saturate_cast4(cur);
+    }
+
+    dst_reg = convert_float4(t_dst[CNDISP - 1]);
+    dst_reg = min(dst_reg, minimum);
+    t_dst[CNDISP - 1] = saturate_cast4(dst_reg);
+    sum += dst_reg;
+
+    sum /= CNDISP;
+#pragma unroll
+    for(int i = 0, idx = 0; i < CNDISP; ++i, idx+=msg_disp_step)
+    {
+        T4 dst = t_dst[i];
+        us_[idx] = dst.x - sum.x;
+        ds_[idx] = dst.y - sum.y;
+        rs_[idx] = dst.z - sum.z;
+        ls_[idx] = dst.w - sum.w;
+    }
+}
+__kernel void one_iteration(__global T *u,    int u_step,
+                            __global T *data, int data_step,
+                            __global T *d,    __global T *l, __global T *r,
+                            int t, int cols, int rows,
+                            float cmax_disc_term, float cdisc_single_jump)
+{
+    const int y = get_global_id(1);
+    const int x = ((get_global_id(0)) << 1) + ((y + t) & 1);
+
+    if ((y > 0) && (y < rows - 1) && (x > 0) && (x < cols - 1))
+    {
+        u_step    /= sizeof(T);
+        data_step /= sizeof(T);
+
+        __global T *us = u + y * u_step + x;
+        __global T *ds = d + y * u_step + x;
+        __global T *ls = l + y * u_step + x;
+        __global T *rs = r + y * u_step + x;
+        const __global  T *dt = data + y * data_step + x;
+
+        int msg_disp_step = u_step * rows;
+        int data_disp_step = data_step * rows;
+
+        message(us, ds, ls, rs, dt,
+                u_step, msg_disp_step, data_disp_step,
+                (float4)(cmax_disc_term), (float4)(cdisc_single_jump));
+    }
+}
+
+///////////////////////////////////////////////////////////////
+/////////////////////////// output ////////////////////////////
+///////////////////////////////////////////////////////////////
+__kernel void output(const __global T *u, int u_step,
+                     const __global T *d, const __global T *l,
+                     const __global T *r, const __global T *data,
+                     __global T *disp, int disp_rows, int disp_cols, int disp_step,
+                     int cndisp)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if (y > 0 && y < disp_rows - 1 && x > 0 && x < disp_cols - 1)
+    {
+        u_step    /= sizeof(T);
+        disp_step /= sizeof(T);
+        const __global T *us = u + (y + 1) * u_step + x;
+        const __global T *ds = d + (y - 1) * u_step + x;
+        const __global T *ls = l + y * u_step + (x + 1);
+        const __global T *rs = r + y * u_step + (x - 1);
+        const __global T *dt = data + y * u_step + x;
+
+        int disp_steps = disp_rows * u_step;
+
+        int best = 0;
+        float best_val = FLOAT_MAX;
+        for (int d = 0; d < cndisp; ++d)
+        {
+            float val;
+            val  = us[d * disp_steps];
+            val += ds[d * disp_steps];
+            val += ls[d * disp_steps];
+            val += rs[d * disp_steps];
+            val += dt[d * disp_steps];
+
+            if (val < best_val)
+            {
+                best_val = val;
+                best = d;
+            }
+        }
+
+        (disp + y * disp_step)[x] = convert_short_sat(best);
+    }
+}
diff --git a/modules/ocl/src/opencl/stereocsbp.cl b/modules/ocl/src/opencl/stereocsbp.cl
new file mode 100644
index 0000000000..89f2bb8dc8
--- /dev/null
+++ b/modules/ocl/src/opencl/stereocsbp.cl
@@ -0,0 +1,1402 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Jin Ma, jin@multicorewareinc.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#ifndef FLT_MAX
+#define FLT_MAX  CL_FLT_MAX
+#endif
+
+#ifndef SHRT_MAX
+#define SHRT_MAX  CL_SHORT_MAX
+#endif
+
+
+///////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////get_first_k_initial_global//////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void get_first_k_initial_global_0(__global short *data_cost_selected_, __global short *selected_disp_pyr,
+    __global short *ctemp, int h, int w, int nr_plane,
+    int cmsg_step1, int cdisp_step1, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        __global short *selected_disparity = selected_disp_pyr      + y * cmsg_step1 + x;
+        __global short *data_cost_selected = data_cost_selected_    + y * cmsg_step1 + x;
+        __global short *data_cost          = ctemp + y * cmsg_step1 + x;
+
+        for(int i = 0; i < nr_plane; i++)
+        {
+            short minimum = SHRT_MAX;
+            int id = 0;
+
+            for(int d = 0; d < cndisp; d++)
+            {
+                short cur = data_cost[d * cdisp_step1];
+                if(cur < minimum)
+                {
+                    minimum = cur;
+                    id = d;
+                }
+            }
+
+            data_cost_selected[i  * cdisp_step1] = minimum;
+            selected_disparity[i  * cdisp_step1] = id;
+            data_cost         [id * cdisp_step1] = SHRT_MAX;
+        }
+    }
+}
+__kernel void get_first_k_initial_global_1(__global  float *data_cost_selected_, __global float *selected_disp_pyr,
+    __global  float *ctemp, int h, int w, int nr_plane,
+    int cmsg_step1, int cdisp_step1, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        __global   float *selected_disparity = selected_disp_pyr      + y * cmsg_step1 + x;
+        __global   float *data_cost_selected = data_cost_selected_    + y * cmsg_step1 + x;
+        __global   float *data_cost          = ctemp + y * cmsg_step1 + x;
+
+        for(int i = 0; i < nr_plane; i++)
+        {
+            float minimum = FLT_MAX;
+            int id = 0;
+
+            for(int d = 0; d < cndisp; d++)
+            {
+                float cur = data_cost[d * cdisp_step1];
+                if(cur < minimum)
+                {
+                    minimum = cur;
+                    id = d;
+                }
+            }
+
+            data_cost_selected[i  * cdisp_step1] = minimum;
+            selected_disparity[i  * cdisp_step1] = id;
+            data_cost         [id * cdisp_step1] = FLT_MAX;
+        }
+    }
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////get_first_k_initial_local////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void get_first_k_initial_local_0(__global  short *data_cost_selected_, __global short *selected_disp_pyr,
+    __global  short *ctemp,int h, int w, int nr_plane,
+    int cmsg_step1, int cdisp_step1, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        __global short *selected_disparity = selected_disp_pyr   + y * cmsg_step1 + x;
+        __global short *data_cost_selected = data_cost_selected_ + y * cmsg_step1 + x;
+        __global short *data_cost = ctemp + y * cmsg_step1 + x;
+
+        int nr_local_minimum = 0;
+
+        short prev = data_cost[0 * cdisp_step1];
+        short cur  = data_cost[1 * cdisp_step1];
+        short next = data_cost[2 * cdisp_step1];
+
+        for (int d = 1; d < cndisp - 1 && nr_local_minimum < nr_plane; d++)
+        {
+
+            if (cur < prev && cur < next)
+            {
+                data_cost_selected[nr_local_minimum * cdisp_step1] = cur;
+                selected_disparity[nr_local_minimum * cdisp_step1] = d;
+                data_cost[d * cdisp_step1] = SHRT_MAX;
+
+                nr_local_minimum++;
+            }
+
+            prev = cur;
+            cur = next;
+            next = data_cost[(d + 1) * cdisp_step1];
+        }
+
+        for (int i = nr_local_minimum; i < nr_plane; i++)
+        {
+            short minimum = SHRT_MAX;
+            int id = 0;
+
+            for (int d = 0; d < cndisp; d++)
+            {
+                cur = data_cost[d * cdisp_step1];
+                if (cur < minimum)
+                {
+                    minimum = cur;
+                    id = d;
+                }
+            }
+
+            data_cost_selected[i * cdisp_step1] = minimum;
+            selected_disparity[i * cdisp_step1] = id;
+            data_cost[id * cdisp_step1] = SHRT_MAX;
+        }
+    }
+}
+
+__kernel void get_first_k_initial_local_1(__global float *data_cost_selected_, __global float *selected_disp_pyr,
+    __global float *ctemp,int h, int w, int nr_plane,
+    int cmsg_step1,  int cdisp_step1, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        __global float *selected_disparity = selected_disp_pyr   + y * cmsg_step1 + x;
+        __global float *data_cost_selected = data_cost_selected_ + y * cmsg_step1 + x;
+        __global float *data_cost = ctemp + y * cmsg_step1 + x;
+
+        int nr_local_minimum = 0;
+
+        float prev = data_cost[0 * cdisp_step1];
+        float cur  = data_cost[1 * cdisp_step1];
+        float next = data_cost[2 * cdisp_step1];
+
+        for (int d = 1; d < cndisp - 1 && nr_local_minimum < nr_plane; d++)
+        {
+            if (cur < prev && cur < next)
+            {
+                data_cost_selected[nr_local_minimum * cdisp_step1] = cur;
+                selected_disparity[nr_local_minimum * cdisp_step1] = d;
+                data_cost[d * cdisp_step1] = FLT_MAX ;
+
+                nr_local_minimum++;
+            }
+
+            prev = cur;
+            cur = next;
+            next = data_cost[(d + 1) * cdisp_step1];
+        }
+
+
+        for (int i = nr_local_minimum; i < nr_plane; i++)
+        {
+            float minimum = FLT_MAX;
+            int id = 0;
+
+            for (int d = 0; d < cndisp; d++)
+            {
+                cur = data_cost[d * cdisp_step1];
+                if (cur < minimum)
+                {
+                    minimum = cur;
+                    id = d;
+                }
+            }
+
+            data_cost_selected[i * cdisp_step1] = minimum;
+            selected_disparity[i * cdisp_step1] = id;
+            data_cost[id * cdisp_step1] = FLT_MAX;
+        }
+    }
+}
+
+///////////////////////////////////////////////////////////////
+/////////////////////// init data cost ////////////////////////
+///////////////////////////////////////////////////////////////
+float compute_3(__global uchar* left, __global uchar* right,
+    float cdata_weight,  float cmax_data_term)
+{
+    float tb = 0.114f * abs((int)left[0] - right[0]);
+    float tg = 0.587f * abs((int)left[1] - right[1]);
+    float tr = 0.299f * abs((int)left[2] - right[2]);
+
+    return fmin(cdata_weight * (tr + tg + tb), cdata_weight * cmax_data_term);
+}
+float compute_1(__global uchar* left, __global uchar* right,
+    float cdata_weight,  float cmax_data_term)
+{
+    return fmin(cdata_weight * abs((int)*left - (int)*right), cdata_weight * cmax_data_term);
+}
+short round_short(float v){
+    return convert_short_sat_rte(v);
+}
+///////////////////////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////init_data_cost///////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void init_data_cost_0(__global short *ctemp, __global uchar *cleft, __global uchar *cright,
+    int h, int w, int level, int channels,
+    int cmsg_step1, float cdata_weight, float cmax_data_term, int cdisp_step1,
+    int cth, int cimg_step, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        int y0 = y << level;
+        int yt = (y + 1) << level;
+
+        int x0 = x << level;
+        int xt = (x + 1) << level;
+
+        __global short *data_cost = ctemp + y * cmsg_step1 + x;
+
+        for(int d = 0; d < cndisp; ++d)
+        {
+            float val = 0.0f;
+            for(int yi = y0; yi < yt; yi++)
+            {
+                for(int xi = x0; xi < xt; xi++)
+                {
+                    int xr = xi - d;
+                    if(d < cth || xr < 0)
+                        val += cdata_weight * cmax_data_term;
+                    else
+                    {
+                        __global uchar *lle = cleft  + yi * cimg_step + xi * channels;
+                        __global uchar *lri = cright + yi * cimg_step + xr * channels;
+
+                        if(channels == 1)
+                            val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                        else
+                            val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+                    }
+                }
+            }
+            data_cost[cdisp_step1 * d] = round_short(val);
+        }
+    }
+}
+__kernel void init_data_cost_1(__global float *ctemp, __global uchar *cleft, __global uchar *cright,
+    int h, int w, int level, int channels,
+    int cmsg_step1, float cdata_weight, float cmax_data_term, int cdisp_step1,
+    int cth, int cimg_step, int cndisp)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        int y0 = y << level;
+        int yt = (y + 1) << level;
+
+        int x0 = x << level;
+        int xt = (x + 1) << level;
+
+        __global float *data_cost = ctemp + y * cmsg_step1 + x;
+
+        for(int d = 0; d < cndisp; ++d)
+        {
+            float val = 0.0f;
+            for(int yi = y0; yi < yt; yi++)
+            {
+                for(int xi = x0; xi < xt; xi++)
+                {
+                    int xr = xi - d;
+                    if(d < cth || xr < 0)
+                        val += cdata_weight * cmax_data_term;
+                    else
+                    {
+                        __global uchar* lle = cleft  + yi * cimg_step + xi * channels;
+                        __global uchar* lri = cright + yi * cimg_step + xr * channels;
+
+                        if(channels == 1)
+                            val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                        else
+                            val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+                    }
+                }
+            }
+            data_cost[cdisp_step1 * d] = val;
+        }
+    }
+}
+////////////////////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////init_data_cost_reduce//////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void init_data_cost_reduce_0(__global short *ctemp, __global uchar *cleft, __global uchar *cright,
+    __local float *smem, int level, int rows, int cols, int h, int winsz, int channels,
+    int cndisp,int cimg_step, float cdata_weight, float cmax_data_term, int cth,
+    int cdisp_step1, int cmsg_step1)
+{
+    int x_out = get_group_id(0);
+    int y_out = get_group_id(1) % h;
+    //int d = (blockIdx.y / h) * blockDim.z + threadIdx.z;
+    int d = (get_group_id(1) / h ) * get_local_size(2) + get_local_id(2);
+
+    int tid = get_local_id(0);
+
+    if (d < cndisp)
+    {
+        int x0 = x_out << level;
+        int y0 = y_out << level;
+
+        int len = min(y0 + winsz, rows) - y0;
+
+        float val = 0.0f;
+        if (x0 + tid < cols)
+        {
+            if (x0 + tid - d < 0 || d < cth)
+                val = cdata_weight * cmax_data_term * len;
+            else
+            {
+                __global uchar* lle =  cleft + y0 * cimg_step + channels * (x0 + tid    );
+                __global uchar* lri = cright + y0 * cimg_step + channels * (x0 + tid - d);
+
+                for(int y = 0; y < len; ++y)
+                {
+                    if(channels == 1)
+                        val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                    else
+                        val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+
+                    lle += cimg_step;
+                    lri += cimg_step;
+                }
+            }
+        }
+
+        __local float* dline = smem + winsz * get_local_id(2);
+
+        dline[tid] = val;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local float* dline = smem + winsz * get_local_id(2);
+        if (winsz >= 256)
+        {
+            if (tid < 128)
+                dline[tid] += dline[tid + 128];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local float* dline = smem + winsz * get_local_id(2);
+        if (winsz >= 128)
+        {
+            if (tid <  64)
+                dline[tid] += dline[tid + 64];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 64)
+            if (tid < 32)
+                vdline[tid] += vdline[tid + 32];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 32)
+            if (tid < 16)
+                vdline[tid] += vdline[tid + 16];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d<cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 16)
+            if (tid <  8)
+                vdline[tid] += vdline[tid + 8];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d<cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 8)
+            if (tid <  4)
+                vdline[tid] += vdline[tid + 4];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d<cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 4)
+            if (tid <  2)
+                vdline[tid] += vdline[tid + 2];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d<cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 2)
+            if (tid <  1)
+                vdline[tid] += vdline[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local float* dline = smem + winsz * get_local_id(2);
+        __global short* data_cost = ctemp + y_out * cmsg_step1 + x_out;
+        if (tid == 0)
+            data_cost[cdisp_step1 * d] = convert_short_sat_rte(dline[0]);
+    }
+}
+
+__kernel void init_data_cost_reduce_1(__global float *ctemp, __global uchar *cleft, __global uchar *cright,
+    __local float *smem, int level, int rows, int cols, int h, int winsz, int channels,
+    int cndisp,int cimg_step, float cdata_weight, float cmax_data_term, int cth,
+    int cdisp_step1, int cmsg_step1)
+{
+    int x_out = get_group_id(0);
+    int y_out = get_group_id(1) % h;
+    int d = (get_group_id(1) / h ) * get_local_size(2) + get_local_id(2);
+
+    int tid = get_local_id(0);
+
+    if (d < cndisp)
+    {
+        int x0 = x_out << level;
+        int y0 = y_out << level;
+
+        int len = min(y0 + winsz, rows) - y0;
+
+        float val = 0.0f;
+        //float val = 528.0f;
+
+        if (x0 + tid < cols)
+        {
+            if (x0 + tid - d < 0 || d < cth)
+                val = cdata_weight * cmax_data_term * len;
+            else
+            {
+                __global uchar* lle =  cleft + y0 * cimg_step + channels * (x0 + tid    );
+                __global uchar* lri = cright + y0 * cimg_step + channels * (x0 + tid - d);
+
+                for(int y = 0; y < len; ++y)
+                {
+                    if(channels == 1)
+                        val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                    else
+                        val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+
+                    lle += cimg_step;
+                    lri += cimg_step;
+                }
+            }
+        }
+
+        __local float* dline = smem + winsz * get_local_id(2);
+
+        dline[tid] = val;
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local float* dline = smem + winsz * get_local_id(2);
+        if (winsz >= 256)
+            if (tid < 128)
+                dline[tid] += dline[tid + 128];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local float* dline = smem + winsz * get_local_id(2);
+        if (winsz >= 128)
+            if (tid < 64)
+                dline[tid] += dline[tid + 64];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 64)
+            if (tid < 32)
+                vdline[tid] += vdline[tid + 32];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 32)
+            if (tid < 16)
+                vdline[tid] += vdline[tid + 16];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 16)
+            if (tid < 8)
+                vdline[tid] += vdline[tid + 8];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 8)
+            if (tid < 4)
+                vdline[tid] += vdline[tid + 4];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 4)
+            if (tid < 2)
+                vdline[tid] += vdline[tid + 2];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 2)
+            if (tid < 1)
+                vdline[tid] += vdline[tid + 1];
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < cndisp)
+    {
+        __global float *data_cost = ctemp + y_out * cmsg_step1 + x_out;
+        __local float* dline = smem + winsz * get_local_id(2);
+        if (tid == 0)
+            data_cost[cdisp_step1 * d] =  dline[0];
+    }
+}
+
+///////////////////////////////////////////////////////////////
+////////////////////// compute data cost //////////////////////
+///////////////////////////////////////////////////////////////
+__kernel void compute_data_cost_0(__global const short *selected_disp_pyr, __global short *data_cost_,
+    __global uchar *cleft, __global uchar *cright,
+    int h, int w, int level, int nr_plane, int channels,
+    int cmsg_step1, int cmsg_step2, int cdisp_step1, int cdisp_step2, float cdata_weight,
+    float cmax_data_term, int cimg_step, int cth)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        int y0 = y << level;
+        int yt = (y + 1) << level;
+
+        int x0 = x << level;
+        int xt = (x + 1) << level;
+
+        __global const short *selected_disparity = selected_disp_pyr + y/2 * cmsg_step2 + x/2;
+        __global       short *data_cost          = data_cost_ + y * cmsg_step1 + x;
+
+        for(int d = 0; d < nr_plane; d++)
+        {
+            float val = 0.0f;
+            for(int yi = y0; yi < yt; yi++)
+            {
+                for(int xi = x0; xi < xt; xi++)
+                {
+                    int sel_disp = selected_disparity[d * cdisp_step2];
+                    int xr = xi - sel_disp;
+
+                    if (xr < 0 || sel_disp < cth)
+                        val += cdata_weight * cmax_data_term;
+
+                    else
+                    {
+                        __global uchar* left_x  = cleft + yi * cimg_step + xi * channels;
+                        __global uchar* right_x = cright + yi * cimg_step + xr * channels;
+
+                        if(channels == 1)
+                            val += compute_1(left_x, right_x, cdata_weight, cmax_data_term);
+                        else
+                            val += compute_3(left_x, right_x, cdata_weight, cmax_data_term);
+                    }
+                }
+            }
+            data_cost[cdisp_step1 * d] = convert_short_sat_rte(val);
+        }
+    }
+}
+__kernel void compute_data_cost_1(__global const float *selected_disp_pyr, __global float *data_cost_,
+    __global uchar *cleft, __global uchar *cright,
+    int h, int w, int level, int nr_plane, int channels,
+    int cmsg_step1, int cmsg_step2, int cdisp_step1, int cdisp_step2, float cdata_weight,
+    float cmax_data_term, int cimg_step, int cth)
+{
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        int y0 = y << level;
+        int yt = (y + 1) << level;
+
+        int x0 = x << level;
+        int xt = (x + 1) << level;
+
+        __global const float *selected_disparity = selected_disp_pyr + y/2 * cmsg_step2 + x/2;
+        __global       float *data_cost          = data_cost_ + y * cmsg_step1 + x;
+
+        for(int d = 0; d < nr_plane; d++)
+        {
+            float val = 0.0f;
+            for(int yi = y0; yi < yt; yi++)
+            {
+                for(int xi = x0; xi < xt; xi++)
+                {
+                    int sel_disp = selected_disparity[d * cdisp_step2];
+                    int xr = xi - sel_disp;
+
+                    if (xr < 0 || sel_disp < cth)
+                        val += cdata_weight * cmax_data_term;
+                    else
+                    {
+                        __global uchar* left_x  = cleft + yi * cimg_step + xi * channels;
+                        __global uchar* right_x = cright + yi * cimg_step + xr * channels;
+
+                        if(channels == 1)
+                            val += compute_1(left_x, right_x, cdata_weight, cmax_data_term);
+                        else
+                            val += compute_3(left_x, right_x, cdata_weight, cmax_data_term);
+                    }
+                }
+            }
+            data_cost[cdisp_step1 * d] = val;
+        }
+    }
+}
+////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////compute_data_cost_reduce//////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+__kernel void compute_data_cost_reduce_0(__global const short* selected_disp_pyr, __global short* data_cost_,
+    __global uchar *cleft, __global uchar *cright,__local float *smem,
+    int level, int rows, int cols, int h, int nr_plane,
+    int channels, int winsz,
+    int cmsg_step1, int cmsg_step2, int cdisp_step1, int cdisp_step2,
+    float cdata_weight,  float cmax_data_term, int cimg_step,int cth)
+
+{
+    int x_out = get_group_id(0);
+    int y_out = get_group_id(1) % h;
+    int d = (get_group_id(1)/ h) * get_local_size(2) + get_local_id(2);
+
+    int tid = get_local_id(0);
+
+    __global const short* selected_disparity = selected_disp_pyr + y_out/2 * cmsg_step2 + x_out/2;
+    __global short* data_cost = data_cost_ + y_out * cmsg_step1 + x_out;
+
+    if (d < nr_plane)
+    {
+        int sel_disp = selected_disparity[d * cdisp_step2];
+
+        int x0 = x_out << level;
+        int y0 = y_out << level;
+
+        int len = min(y0 + winsz, rows) - y0;
+
+        float val = 0.0f;
+        if (x0 + tid < cols)
+        {
+            if (x0 + tid - sel_disp < 0 || sel_disp < cth)
+                val = cdata_weight * cmax_data_term * len;
+            else
+            {
+                __global uchar* lle =  cleft + y0 * cimg_step + channels * (x0 + tid    );
+                __global uchar* lri = cright + y0 * cimg_step + channels * (x0 + tid - sel_disp);
+
+                for(int y = 0; y < len; ++y)
+                {
+                    if(channels == 1)
+                        val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                    else
+                        val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+
+                    lle += cimg_step;
+                    lri += cimg_step;
+                }
+            }
+        }
+
+        __local float* dline = smem + winsz * get_local_id(2);
+
+        dline[tid] = val;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    // if (winsz >= 256) { if (tid < 128) { dline[tid] += dline[tid + 128]; } barrier(CLK_LOCAL_MEM_FENCE); }
+    //if (winsz >= 128) { if (tid <  64) { dline[tid] += dline[tid +  64]; } barrier(CLK_LOCAL_MEM_FENCE); }
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 64)
+        {
+            if (tid < 32)
+                vdline[tid] += vdline[tid + 32];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 32)
+        {
+            if (tid < 16)
+                vdline[tid] += vdline[tid + 16];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 16)
+        {
+            if (tid < 8)
+                vdline[tid] += vdline[tid + 8];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 8)
+        {
+            if (tid < 4)
+                vdline[tid] += vdline[tid + 4];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 4)
+        {
+            if (tid < 2)
+                vdline[tid] += vdline[tid + 2];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 2)
+        {
+            if (tid < 1)
+                vdline[tid] += vdline[tid + 1];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (tid == 0)
+            data_cost[cdisp_step1 * d] = convert_short_sat_rte(vdline[0]);
+    }
+}
+
+__kernel void compute_data_cost_reduce_1(__global const float *selected_disp_pyr, __global float *data_cost_,
+    __global uchar *cleft, __global uchar *cright, __local float *smem,
+    int level, int rows, int cols, int h, int nr_plane,
+    int channels, int winsz,
+    int cmsg_step1, int cmsg_step2, int cdisp_step1,int cdisp_step2, float cdata_weight,
+    float cmax_data_term, int cimg_step, int cth)
+
+{
+    int x_out = get_group_id(0);
+    int y_out = get_group_id(1) % h;
+    int d = (get_group_id(1)/ h) * get_local_size(2) + get_local_id(2);
+
+    int tid = get_local_id(0);
+
+    __global const float *selected_disparity = selected_disp_pyr + y_out/2 * cmsg_step2 + x_out/2;
+    __global float *data_cost = data_cost_ + y_out * cmsg_step1 + x_out;
+
+    if (d < nr_plane)
+    {
+        int sel_disp = selected_disparity[d * cdisp_step2];
+
+        int x0 = x_out << level;
+        int y0 = y_out << level;
+
+        int len = min(y0 + winsz, rows) - y0;
+
+        float val = 0.0f;
+        if (x0 + tid < cols)
+        {
+            if (x0 + tid - sel_disp < 0 || sel_disp < cth)
+                val = cdata_weight * cmax_data_term * len;
+            else
+            {
+                __global uchar* lle =  cleft + y0 * cimg_step + channels * (x0 + tid    );
+                __global uchar* lri = cright + y0 * cimg_step + channels * (x0 + tid - sel_disp);
+
+                for(int y = 0; y < len; ++y)
+                {
+                    if(channels == 1)
+                        val += compute_1(lle, lri, cdata_weight, cmax_data_term);
+                    else
+                        val += compute_3(lle, lri, cdata_weight, cmax_data_term);
+
+                    lle += cimg_step;
+                    lri += cimg_step;
+                }
+            }
+        }
+
+        __local float* dline = smem + winsz * get_local_id(2);
+
+        dline[tid] = val;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 64)
+        {
+            if (tid < 32)
+                vdline[tid] += vdline[tid + 32];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 32)
+        {
+            if (tid < 16)
+                vdline[tid] += vdline[tid + 16];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >= 16)
+        {
+            if (tid <  8)
+                vdline[tid] += vdline[tid + 8];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >=  8)
+        {
+            if (tid <  4)
+                vdline[tid] += vdline[tid + 4];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >=  4)
+        {
+            if (tid <  2)
+                vdline[tid] += vdline[tid + 2];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (winsz >=  2)
+        {
+            if (tid <  1)
+                vdline[tid] += vdline[tid + 1];
+        }
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(d < nr_plane)
+    {
+        __local volatile float* vdline = smem + winsz * get_local_id(2);
+        if (tid == 0)
+            data_cost[cdisp_step1 * d] = vdline[0];
+    }
+}
+
+///////////////////////////////////////////////////////////////
+//////////////////////// init message /////////////////////////
+///////////////////////////////////////////////////////////////
+void get_first_k_element_increase_0(__global short* u_new, __global short *d_new, __global short *l_new,
+    __global short *r_new, __global const short *u_cur, __global const short *d_cur,
+    __global const short *l_cur, __global const short *r_cur,
+    __global short *data_cost_selected, __global short *disparity_selected_new,
+    __global short *data_cost_new, __global const short* data_cost_cur,
+    __global const short *disparity_selected_cur,
+    int nr_plane, int nr_plane2,
+    int cdisp_step1, int cdisp_step2)
+{
+    for(int i = 0; i < nr_plane; i++)
+    {
+        short minimum = SHRT_MAX;
+        int id = 0;
+        for(int j = 0; j < nr_plane2; j++)
+        {
+            short cur = data_cost_new[j * cdisp_step1];
+            if(cur < minimum)
+            {
+                minimum = cur;
+                id = j;
+            }
+        }
+
+        data_cost_selected[i * cdisp_step1] = data_cost_cur[id * cdisp_step1];
+        disparity_selected_new[i * cdisp_step1] = disparity_selected_cur[id * cdisp_step2];
+
+        u_new[i * cdisp_step1] = u_cur[id * cdisp_step2];
+        d_new[i * cdisp_step1] = d_cur[id * cdisp_step2];
+        l_new[i * cdisp_step1] = l_cur[id * cdisp_step2];
+        r_new[i * cdisp_step1] = r_cur[id * cdisp_step2];
+
+        data_cost_new[id * cdisp_step1] = SHRT_MAX;
+    }
+}
+void get_first_k_element_increase_1(__global float *u_new, __global float *d_new, __global float *l_new,
+    __global float *r_new, __global const float *u_cur, __global const float *d_cur,
+    __global const float *l_cur, __global const float *r_cur,
+    __global float *data_cost_selected, __global float *disparity_selected_new,
+    __global float *data_cost_new, __global const float *data_cost_cur,
+    __global const float *disparity_selected_cur,
+    int nr_plane, int nr_plane2,
+    int cdisp_step1, int cdisp_step2)
+{
+    for(int i = 0; i < nr_plane; i++)
+    {
+        float minimum = FLT_MAX;
+        int id = 0;
+
+        for(int j = 0; j < nr_plane2; j++)
+        {
+            float cur = data_cost_new[j * cdisp_step1];
+            if(cur < minimum)
+            {
+                minimum = cur;
+                id = j;
+            }
+        }
+
+        data_cost_selected[i * cdisp_step1] = data_cost_cur[id * cdisp_step1];
+        disparity_selected_new[i * cdisp_step1] = disparity_selected_cur[id * cdisp_step2];
+
+        u_new[i * cdisp_step1] = u_cur[id * cdisp_step2];
+        d_new[i * cdisp_step1] = d_cur[id * cdisp_step2];
+        l_new[i * cdisp_step1] = l_cur[id * cdisp_step2];
+        r_new[i * cdisp_step1] = r_cur[id * cdisp_step2];
+        data_cost_new[id * cdisp_step1] = FLT_MAX;
+
+    }
+}
+__kernel void init_message_0(__global short *u_new_, __global short *d_new_, __global short *l_new_,
+    __global short *r_new_, __global  short *u_cur_, __global const short *d_cur_,
+    __global const short *l_cur_, __global const short *r_cur_, __global short *ctemp,
+    __global short *selected_disp_pyr_new, __global const short *selected_disp_pyr_cur,
+    __global short *data_cost_selected_, __global const short *data_cost_,
+    int h, int w, int nr_plane, int h2, int w2, int nr_plane2,
+    int cdisp_step1, int cdisp_step2, int cmsg_step1, int cmsg_step2)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y < h && x < w)
+    {
+        __global const short *u_cur = u_cur_ + min(h2-1, y/2 + 1) * cmsg_step2 + x/2;
+        __global const short *d_cur = d_cur_ + max(0, y/2 - 1)    * cmsg_step2 + x/2;
+        __global const short *l_cur = l_cur_ + y/2                * cmsg_step2 + min(w2-1, x/2 + 1);
+        __global const short *r_cur = r_cur_ + y/2                * cmsg_step2 + max(0, x/2 - 1);
+
+        __global short *data_cost_new = ctemp + y * cmsg_step1 + x;
+
+        __global const short *disparity_selected_cur = selected_disp_pyr_cur + y/2 * cmsg_step2 + x/2;
+        __global const short *data_cost = data_cost_ + y * cmsg_step1 + x;
+
+        for(int d = 0; d < nr_plane2; d++)
+        {
+            int idx2 = d * cdisp_step2;
+
+            short val  = data_cost[d * cdisp_step1] + u_cur[idx2] + d_cur[idx2] + l_cur[idx2] + r_cur[idx2];
+            data_cost_new[d * cdisp_step1] = val;
+        }
+
+        __global short *data_cost_selected = data_cost_selected_ + y * cmsg_step1 + x;
+        __global short *disparity_selected_new = selected_disp_pyr_new + y * cmsg_step1 + x;
+
+        __global short *u_new = u_new_ + y * cmsg_step1 + x;
+        __global short *d_new = d_new_ + y * cmsg_step1 + x;
+        __global short *l_new = l_new_ + y * cmsg_step1 + x;
+        __global short *r_new = r_new_ + y * cmsg_step1 + x;
+
+        u_cur = u_cur_ + y/2 * cmsg_step2 + x/2;
+        d_cur = d_cur_ + y/2 * cmsg_step2 + x/2;
+        l_cur = l_cur_ + y/2 * cmsg_step2 + x/2;
+        r_cur = r_cur_ + y/2 * cmsg_step2 + x/2;
+
+        get_first_k_element_increase_0(u_new, d_new, l_new, r_new, u_cur, d_cur, l_cur, r_cur,
+            data_cost_selected, disparity_selected_new, data_cost_new,
+            data_cost, disparity_selected_cur, nr_plane, nr_plane2,
+            cdisp_step1, cdisp_step2);
+    }
+}
+__kernel void init_message_1(__global float *u_new_, __global float *d_new_, __global float *l_new_,
+    __global float *r_new_, __global const float *u_cur_, __global const float *d_cur_,
+    __global const float *l_cur_, __global const float *r_cur_, __global float *ctemp,
+    __global float *selected_disp_pyr_new, __global const float *selected_disp_pyr_cur,
+    __global float *data_cost_selected_, __global const float *data_cost_,
+    int h, int w, int nr_plane, int h2, int w2, int nr_plane2,
+    int cdisp_step1, int cdisp_step2, int cmsg_step1, int cmsg_step2)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+
+    __global const float *u_cur = u_cur_ + min(h2-1, y/2 + 1) * cmsg_step2 + x/2;
+    __global const float *d_cur = d_cur_ + max(0, y/2 - 1)    * cmsg_step2 + x/2;
+    __global const float *l_cur = l_cur_ + y/2                * cmsg_step2 + min(w2-1, x/2 + 1);
+    __global const float *r_cur = r_cur_ + y/2                * cmsg_step2 + max(0, x/2 - 1);
+
+    __global float *data_cost_new = ctemp + y * cmsg_step1 + x;
+
+    __global const float *disparity_selected_cur = selected_disp_pyr_cur + y/2 * cmsg_step2 + x/2;
+    __global const float *data_cost = data_cost_ + y * cmsg_step1 + x;
+
+    if (y < h && x < w)
+    {
+        for(int d = 0; d < nr_plane2; d++)
+        {
+            int idx2 = d * cdisp_step2;
+
+            float val  = data_cost[d * cdisp_step1] + u_cur[idx2] + d_cur[idx2] + l_cur[idx2] + r_cur[idx2];
+            data_cost_new[d * cdisp_step1] = val;
+        }
+    }
+
+    __global float *data_cost_selected = data_cost_selected_ + y * cmsg_step1 + x;
+    __global float *disparity_selected_new = selected_disp_pyr_new + y * cmsg_step1 + x;
+
+    __global float *u_new = u_new_ + y * cmsg_step1 + x;
+    __global float *d_new = d_new_ + y * cmsg_step1 + x;
+    __global float *l_new = l_new_ + y * cmsg_step1 + x;
+    __global float *r_new = r_new_ + y * cmsg_step1 + x;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if(y < h && x < w)
+    {
+        u_cur = u_cur_ + y/2 * cmsg_step2 + x/2;
+        d_cur = d_cur_ + y/2 * cmsg_step2 + x/2;
+        l_cur = l_cur_ + y/2 * cmsg_step2 + x/2;
+        r_cur = r_cur_ + y/2 * cmsg_step2 + x/2;
+
+        for(int i = 0; i < nr_plane; i++)
+        {
+            float minimum = FLT_MAX;
+            int id = 0;
+
+            for(int j = 0; j < nr_plane2; j++)
+            {
+                float cur = data_cost_new[j * cdisp_step1];
+                if(cur < minimum)
+                {
+                    minimum = cur;
+                    id = j;
+                }
+            }
+            data_cost_selected[i * cdisp_step1] = data_cost[id * cdisp_step1];
+            disparity_selected_new[i * cdisp_step1] = disparity_selected_cur[id * cdisp_step2];
+            u_new[i * cdisp_step1] = u_cur[id * cdisp_step2];
+            d_new[i * cdisp_step1] = d_cur[id * cdisp_step2];
+            l_new[i * cdisp_step1] = l_cur[id * cdisp_step2];
+            r_new[i * cdisp_step1] = r_cur[id * cdisp_step2];
+            data_cost_new[id * cdisp_step1] = FLT_MAX;
+        }
+    }
+}
+
+///////////////////////////////////////////////////////////////
+////////////////////  calc all iterations /////////////////////
+///////////////////////////////////////////////////////////////
+void message_per_pixel_0(__global const short *data, __global short *msg_dst, __global const short *msg1,
+    __global const short *msg2, __global const short *msg3,
+    __global const short *dst_disp, __global const short *src_disp,
+    int nr_plane, __global short *temp,
+    float cmax_disc_term, int cdisp_step1, float cdisc_single_jump)
+{
+    short minimum = SHRT_MAX;
+    for(int d = 0; d < nr_plane; d++)
+    {
+        int idx = d * cdisp_step1;
+        short val  = data[idx] + msg1[idx] + msg2[idx] + msg3[idx];
+
+        if(val < minimum)
+            minimum = val;
+
+        msg_dst[idx] = val;
+    }
+
+    float sum = 0;
+    for(int d = 0; d < nr_plane; d++)
+    {
+        float cost_min = minimum + cmax_disc_term;
+        short src_disp_reg = src_disp[d * cdisp_step1];
+
+        for(int d2 = 0; d2 < nr_plane; d2++)
+            cost_min = fmin(cost_min, (msg_dst[d2 * cdisp_step1] +
+            cdisc_single_jump * abs(dst_disp[d2 * cdisp_step1] - src_disp_reg)));
+
+        temp[d * cdisp_step1] = convert_short_sat_rte(cost_min);
+        sum += cost_min;
+    }
+    sum /= nr_plane;
+
+    for(int d = 0; d < nr_plane; d++)
+        msg_dst[d * cdisp_step1] = convert_short_sat_rte(temp[d * cdisp_step1] - sum);
+}
+void message_per_pixel_1(__global const float *data, __global float *msg_dst, __global const float *msg1,
+    __global const float *msg2, __global const float *msg3,
+    __global const float *dst_disp, __global const float *src_disp,
+    int nr_plane, __global float *temp,
+    float cmax_disc_term, int cdisp_step1, float cdisc_single_jump)
+{
+    float minimum = FLT_MAX;
+    for(int d = 0; d < nr_plane; d++)
+    {
+        int idx = d * cdisp_step1;
+        float val  = data[idx] + msg1[idx] + msg2[idx] + msg3[idx];
+
+        if(val < minimum)
+            minimum = val;
+
+        msg_dst[idx] = val;
+    }
+
+    float sum = 0;
+    for(int d = 0; d < nr_plane; d++)
+    {
+        float cost_min = minimum + cmax_disc_term;
+        float src_disp_reg = src_disp[d * cdisp_step1];
+
+        for(int d2 = 0; d2 < nr_plane; d2++)
+            cost_min = fmin(cost_min, (msg_dst[d2 * cdisp_step1] +
+            cdisc_single_jump * fabs(dst_disp[d2 * cdisp_step1] - src_disp_reg)));
+
+        temp[d * cdisp_step1] = cost_min;
+        sum += cost_min;
+    }
+    sum /= nr_plane;
+
+    for(int d = 0; d < nr_plane; d++)
+        msg_dst[d * cdisp_step1] = temp[d * cdisp_step1] - sum;
+}
+__kernel void compute_message_0(__global short *u_, __global short *d_, __global short *l_, __global short *r_,
+    __global const short *data_cost_selected, __global const short *selected_disp_pyr_cur,
+    __global short *ctemp, int h, int w, int nr_plane, int i,
+    float cmax_disc_term, int cdisp_step1, int cmsg_step1, float cdisc_single_jump)
+{
+    int y = get_global_id(1);
+    int x = ((get_global_id(0)) << 1) + ((y + i) & 1);
+
+    if (y > 0 && y < h - 1 && x > 0 && x < w - 1)
+    {
+        __global const short *data = data_cost_selected + y * cmsg_step1 + x;
+
+        __global short *u = u_ + y * cmsg_step1 + x;
+        __global short *d = d_ + y * cmsg_step1 + x;
+        __global short *l = l_ + y * cmsg_step1 + x;
+        __global short *r = r_ + y * cmsg_step1 + x;
+
+        __global const short *disp = selected_disp_pyr_cur + y * cmsg_step1 + x;
+
+        __global short *temp = ctemp + y * cmsg_step1 + x;
+
+        message_per_pixel_0(data, u, r - 1, u + cmsg_step1, l + 1, disp, disp - cmsg_step1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_0(data, d, d - cmsg_step1, r - 1, l + 1, disp, disp + cmsg_step1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_0(data, l, u + cmsg_step1, d - cmsg_step1, l + 1, disp, disp - 1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_0(data, r, u + cmsg_step1, d - cmsg_step1, r - 1, disp, disp + 1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+    }
+}
+__kernel void compute_message_1(__global float *u_, __global float *d_, __global float *l_, __global float *r_,
+    __global const float *data_cost_selected, __global const float *selected_disp_pyr_cur,
+    __global float *ctemp, int h, int w, int nr_plane, int i,
+    float cmax_disc_term, int cdisp_step1, int cmsg_step1, float cdisc_single_jump)
+{
+    int y = get_global_id(1);
+    int x = ((get_global_id(0)) << 1) + ((y + i) & 1);
+
+    if (y > 0 && y < h - 1 && x > 0 && x < w - 1)
+    {
+        __global const float *data = data_cost_selected + y * cmsg_step1 + x;
+
+        __global float *u = u_ + y * cmsg_step1 + x;
+        __global float *d = d_ + y * cmsg_step1 + x;
+        __global float *l = l_ + y * cmsg_step1 + x;
+        __global float *r = r_ + y * cmsg_step1 + x;
+
+        __global const float *disp = selected_disp_pyr_cur + y * cmsg_step1 + x;
+        __global float *temp = ctemp + y * cmsg_step1 + x;
+
+        message_per_pixel_1(data, u, r - 1, u + cmsg_step1, l + 1, disp, disp - cmsg_step1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_1(data, d, d - cmsg_step1, r - 1, l + 1, disp, disp + cmsg_step1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_1(data, l, u + cmsg_step1, d - cmsg_step1, l + 1, disp, disp - 1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+        message_per_pixel_1(data, r, u + cmsg_step1, d - cmsg_step1, r - 1, disp, disp + 1, nr_plane, temp,
+            cmax_disc_term, cdisp_step1, cdisc_single_jump);
+    }
+}
+
+///////////////////////////////////////////////////////////////
+/////////////////////////// output ////////////////////////////
+///////////////////////////////////////////////////////////////
+__kernel void compute_disp_0(__global const short *u_, __global const short *d_, __global const short *l_,
+    __global const short *r_, __global const short * data_cost_selected,
+    __global const short *disp_selected_pyr,
+    __global short* disp,
+    int res_step, int cols, int rows, int nr_plane,
+    int cmsg_step1, int cdisp_step1)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y > 0 && y < rows - 1 && x > 0 && x < cols - 1)
+    {
+        __global const short *data = data_cost_selected + y * cmsg_step1 + x;
+        __global const short *disp_selected = disp_selected_pyr + y * cmsg_step1 + x;
+
+        __global const short *u = u_ + (y+1) * cmsg_step1 + (x+0);
+        __global const short *d = d_ + (y-1) * cmsg_step1 + (x+0);
+        __global const short *l = l_ + (y+0) * cmsg_step1 + (x+1);
+        __global const short *r = r_ + (y+0) * cmsg_step1 + (x-1);
+
+        short best = 0;
+        short best_val = SHRT_MAX;
+
+        for (int i = 0; i < nr_plane; ++i)
+        {
+            int idx = i * cdisp_step1;
+            short val = data[idx]+ u[idx] + d[idx] + l[idx] + r[idx];
+
+            if (val < best_val)
+            {
+                best_val = val;
+                best = disp_selected[idx];
+            }
+        }
+        disp[res_step * y + x] = best;
+    }
+}
+__kernel void compute_disp_1(__global const float *u_, __global const float *d_, __global const float *l_,
+    __global const float *r_, __global const float *data_cost_selected,
+    __global const float *disp_selected_pyr,
+    __global short *disp,
+    int res_step, int cols, int rows, int nr_plane,
+    int cmsg_step1, int cdisp_step1)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if (y > 0 && y < rows - 1 && x > 0 && x < cols - 1)
+    {
+        __global const float *data = data_cost_selected + y * cmsg_step1 + x;
+        __global const float *disp_selected = disp_selected_pyr + y * cmsg_step1 + x;
+
+        __global const float *u = u_ + (y+1) * cmsg_step1 + (x+0);
+        __global const float *d = d_ + (y-1) * cmsg_step1 + (x+0);
+        __global const float *l = l_ + (y+0) * cmsg_step1 + (x+1);
+        __global const float *r = r_ + (y+0) * cmsg_step1 + (x-1);
+
+        short best = 0;
+        short best_val = SHRT_MAX;
+        for (int i = 0; i < nr_plane; ++i)
+        {
+            int idx = i * cdisp_step1;
+            float val = data[idx]+ u[idx] + d[idx] + l[idx] + r[idx];
+
+            if (val < best_val)
+            {
+                best_val = val;
+                best = convert_short_sat_rte(disp_selected[idx]);
+            }
+        }
+        disp[res_step * y + x] = best;
+    }
+}
diff --git a/modules/ocl/src/opencl/tvl1flow.cl b/modules/ocl/src/opencl/tvl1flow.cl
new file mode 100644
index 0000000000..095b339f87
--- /dev/null
+++ b/modules/ocl/src/opencl/tvl1flow.cl
@@ -0,0 +1,411 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jin Ma jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+__kernel void centeredGradientKernel(__global const float* src, int src_col, int src_row, int src_step,
+__global float* dx, __global float* dy, int dx_step)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if((x < src_col)&&(y < src_row))
+    {
+        int src_x1 = (x + 1) < (src_col -1)? (x + 1) : (src_col - 1);
+        int src_x2 = (x - 1) > 0 ? (x -1) : 0;
+
+        //if(src[y * src_step + src_x1] == src[y * src_step+ src_x2])
+        //{
+        //    printf("y = %d\n", y);
+        //    printf("src_x1 = %d\n", src_x1);
+        //    printf("src_x2 = %d\n", src_x2);
+        //}
+        dx[y * dx_step+ x] = 0.5f * (src[y * src_step + src_x1] - src[y * src_step+ src_x2]);
+
+        int src_y1 = (y+1) < (src_row - 1) ? (y + 1) : (src_row - 1);
+        int src_y2 = (y - 1) > 0 ? (y - 1) : 0;
+        dy[y * dx_step+ x] = 0.5f * (src[src_y1 * src_step + x] - src[src_y2 * src_step+ x]);
+    }
+
+}
+
+float bicubicCoeff(float x_)
+{
+
+    float x = fabs(x_);
+    if (x <= 1.0f)
+    {
+        return x * x * (1.5f * x - 2.5f) + 1.0f;
+    }
+    else if (x < 2.0f)
+    {
+        return x * (x * (-0.5f * x + 2.5f) - 4.0f) + 2.0f;
+    }
+    else
+    {
+        return 0.0f;
+    }
+
+}
+
+__kernel void warpBackwardKernel(__global const float* I0, int I0_step, int I0_col, int I0_row,
+    image2d_t tex_I1, image2d_t tex_I1x, image2d_t tex_I1y,
+    __global const float* u1, int u1_step,
+    __global const float* u2,
+    __global float* I1w,
+    __global float* I1wx, /*int I1wx_step,*/
+    __global float* I1wy, /*int I1wy_step,*/
+    __global float* grad, /*int grad_step,*/
+    __global float* rho,
+    int I1w_step,
+    int u2_step,
+    int u1_offset_x,
+    int u1_offset_y,
+    int u2_offset_x,
+    int u2_offset_y)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if(x < I0_col&&y < I0_row)
+    {
+        //const float u1Val = u1(y, x);
+        const float u1Val = u1[(y + u1_offset_y) * u1_step + x + u1_offset_x];
+        //const float u2Val = u2(y, x);
+        const float u2Val = u2[(y + u2_offset_y) * u2_step + x + u2_offset_x];
+
+        const float wx = x + u1Val;
+        const float wy = y + u2Val;
+
+        const int xmin = ceil(wx - 2.0f);
+        const int xmax = floor(wx + 2.0f);
+
+        const int ymin = ceil(wy - 2.0f);
+        const int ymax = floor(wy + 2.0f);
+
+        float sum  = 0.0f;
+        float sumx = 0.0f;
+        float sumy = 0.0f;
+        float wsum = 0.0f;
+        sampler_t sampleri = CLK_NORMALIZED_COORDS_FALSE | CLK_ADDRESS_CLAMP_TO_EDGE | CLK_FILTER_NEAREST;
+
+        for (int cy = ymin; cy <= ymax; ++cy)
+        {
+            for (int cx = xmin; cx <= xmax; ++cx)
+            {
+                const float w = bicubicCoeff(wx - cx) * bicubicCoeff(wy - cy);
+
+                //sum  += w * tex2D(tex_I1 , cx, cy);
+                int2 cood = (int2)(cx, cy);
+                sum += w * read_imagef(tex_I1, sampleri, cood).x;
+                //sumx += w * tex2D(tex_I1x, cx, cy);
+                sumx += w * read_imagef(tex_I1x, sampleri, cood).x;
+                //sumy += w * tex2D(tex_I1y, cx, cy);
+                sumy += w * read_imagef(tex_I1y, sampleri, cood).x;
+
+                wsum += w;
+            }
+        }
+
+        const float coeff = 1.0f / wsum;
+
+        const float I1wVal  = sum  * coeff;
+        const float I1wxVal = sumx * coeff;
+        const float I1wyVal = sumy * coeff;
+
+        I1w[y * I1w_step + x]  = I1wVal;
+        I1wx[y * I1w_step + x] = I1wxVal;
+        I1wy[y * I1w_step + x] = I1wyVal;
+
+        const float Ix2 = I1wxVal * I1wxVal;
+        const float Iy2 = I1wyVal * I1wyVal;
+
+        // store the |Grad(I1)|^2
+        grad[y * I1w_step + x] = Ix2 + Iy2;
+
+        // compute the constant part of the rho function
+        const float I0Val = I0[y * I0_step + x];
+        rho[y * I1w_step + x] = I1wVal - I1wxVal * u1Val - I1wyVal * u2Val - I0Val;
+    }
+
+}
+
+float readImage(__global const float *image,  const int x,  const int y,  const int rows,  const int cols, const int elemCntPerRow)
+{
+    int i0 = clamp(x, 0, cols - 1);
+    int j0 = clamp(y, 0, rows - 1);
+    int i1 = clamp(x + 1, 0, cols - 1);
+    int j1 = clamp(y + 1, 0, rows - 1);
+
+    return image[j0 * elemCntPerRow + i0];
+}
+
+__kernel void warpBackwardKernelNoImage2d(__global const float* I0, int I0_step, int I0_col, int I0_row,
+    __global const float* tex_I1, __global const float* tex_I1x, __global const float* tex_I1y,
+    __global const float* u1, int u1_step,
+    __global const float* u2,
+    __global float* I1w,
+    __global float* I1wx, /*int I1wx_step,*/
+    __global float* I1wy, /*int I1wy_step,*/
+    __global float* grad, /*int grad_step,*/
+    __global float* rho,
+    int I1w_step,
+    int u2_step,
+    int I1_step,
+    int I1x_step)
+{
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if(x < I0_col&&y < I0_row)
+    {
+        //const float u1Val = u1(y, x);
+        const float u1Val = u1[y * u1_step + x];
+        //const float u2Val = u2(y, x);
+        const float u2Val = u2[y * u2_step + x];
+
+        const float wx = x + u1Val;
+        const float wy = y + u2Val;
+
+        const int xmin = ceil(wx - 2.0f);
+        const int xmax = floor(wx + 2.0f);
+
+        const int ymin = ceil(wy - 2.0f);
+        const int ymax = floor(wy + 2.0f);
+
+        float sum  = 0.0f;
+        float sumx = 0.0f;
+        float sumy = 0.0f;
+        float wsum = 0.0f;
+
+        for (int cy = ymin; cy <= ymax; ++cy)
+        {
+            for (int cx = xmin; cx <= xmax; ++cx)
+            {
+                const float w = bicubicCoeff(wx - cx) * bicubicCoeff(wy - cy);
+
+                int2 cood = (int2)(cx, cy);
+                sum += w * readImage(tex_I1, cood.x, cood.y, I0_col, I0_row, I1_step);
+                sumx += w * readImage(tex_I1x, cood.x, cood.y, I0_col, I0_row, I1x_step);
+                sumy += w * readImage(tex_I1y, cood.x, cood.y, I0_col, I0_row, I1x_step);
+                wsum += w;
+            }
+        }
+
+        const float coeff = 1.0f / wsum;
+
+        const float I1wVal  = sum  * coeff;
+        const float I1wxVal = sumx * coeff;
+        const float I1wyVal = sumy * coeff;
+
+        I1w[y * I1w_step + x]  = I1wVal;
+        I1wx[y * I1w_step + x] = I1wxVal;
+        I1wy[y * I1w_step + x] = I1wyVal;
+
+        const float Ix2 = I1wxVal * I1wxVal;
+        const float Iy2 = I1wyVal * I1wyVal;
+
+        // store the |Grad(I1)|^2
+        grad[y * I1w_step + x] = Ix2 + Iy2;
+
+        // compute the constant part of the rho function
+        const float I0Val = I0[y * I0_step + x];
+        rho[y * I1w_step + x] = I1wVal - I1wxVal * u1Val - I1wyVal * u2Val - I0Val;
+    }
+
+}
+
+
+__kernel void estimateDualVariablesKernel(__global const float* u1, int u1_col, int u1_row, int u1_step,
+    __global const float* u2,
+    __global float* p11, int p11_step,
+    __global float* p12,
+    __global float* p21,
+    __global float* p22,
+    const float taut,
+    int u2_step,
+    int u1_offset_x,
+    int u1_offset_y,
+    int u2_offset_x,
+    int u2_offset_y)
+{
+
+    //const int x = blockIdx.x * blockDim.x + threadIdx.x;
+    //const int y = blockIdx.y * blockDim.y + threadIdx.y;
+    const int x = get_global_id(0);
+    const int y = get_global_id(1);
+
+    if(x < u1_col && y < u1_row)
+    {
+        int src_x1 = (x + 1) < (u1_col - 1) ? (x + 1) : (u1_col - 1);
+        const float u1x = u1[(y + u1_offset_y) * u1_step + src_x1 + u1_offset_x] - u1[(y + u1_offset_y) * u1_step + x + u1_offset_x];
+
+        int src_y1 = (y + 1) < (u1_row - 1) ? (y + 1) : (u1_row - 1);
+        const float u1y = u1[(src_y1 + u1_offset_y) * u1_step + x + u1_offset_x] - u1[(y + u1_offset_y) * u1_step + x + u1_offset_x];
+
+        int src_x2 = (x + 1) < (u1_col - 1) ? (x + 1) : (u1_col - 1);
+        const float u2x = u2[(y + u2_offset_y) * u2_step + src_x2 + u2_offset_x] - u2[(y + u2_offset_y) * u2_step + x + u2_offset_x];
+
+        int src_y2 = (y + 1) <  (u1_row - 1) ? (y + 1) : (u1_row - 1);
+        const float u2y = u2[(src_y2 + u2_offset_y) * u2_step + x + u2_offset_x] - u2[(y + u2_offset_y) * u2_step + x + u2_offset_x];
+
+        const float g1 = hypot(u1x, u1y);
+        const float g2 = hypot(u2x, u2y);
+
+        const float ng1 = 1.0f + taut * g1;
+        const float ng2 = 1.0f + taut * g2;
+
+        p11[y * p11_step + x] = (p11[y * p11_step + x] + taut * u1x) / ng1;
+        p12[y * p11_step + x] = (p12[y * p11_step + x] + taut * u1y) / ng1;
+        p21[y * p11_step + x] = (p21[y * p11_step + x] + taut * u2x) / ng2;
+        p22[y * p11_step + x] = (p22[y * p11_step + x] + taut * u2y) / ng2;
+    }
+
+}
+
+float divergence(__global const float* v1, __global const float* v2, int y, int x, int v1_step, int v2_step)
+{
+
+    if (x > 0 && y > 0)
+    {
+        const float v1x = v1[y * v1_step + x] - v1[y * v1_step + x - 1];
+        const float v2y = v2[y * v2_step + x] - v2[(y - 1) * v2_step + x];
+        return v1x + v2y;
+    }
+    else
+    {
+        if (y > 0)
+            return v1[y * v1_step + 0] + v2[y * v2_step + 0] - v2[(y - 1) * v2_step + 0];
+        else
+        {
+            if (x > 0)
+                return v1[0 * v1_step + x] - v1[0 * v1_step + x - 1] + v2[0 * v2_step + x];
+            else
+                return v1[0 * v1_step + 0] + v2[0 * v2_step + 0];
+        }
+    }
+
+}
+
+__kernel void estimateUKernel(__global const float* I1wx, int I1wx_col, int I1wx_row, int I1wx_step,
+    __global const float* I1wy, /*int I1wy_step,*/
+    __global const float* grad, /*int grad_step,*/
+    __global const float* rho_c, /*int rho_c_step,*/
+    __global const float* p11, /*int p11_step,*/
+    __global const float* p12, /*int p12_step,*/
+    __global const float* p21, /*int p21_step,*/
+    __global const float* p22, /*int p22_step,*/
+    __global float* u1, int u1_step,
+    __global float* u2,
+    __global float* error, const float l_t, const float theta, int u2_step,
+    int u1_offset_x,
+    int u1_offset_y,
+    int u2_offset_x,
+    int u2_offset_y,
+    char calc_error)
+{
+
+    //const int x = blockIdx.x * blockDim.x + threadIdx.x;
+    //const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+
+    if(x < I1wx_col && y < I1wx_row)
+    {
+        const float I1wxVal = I1wx[y * I1wx_step + x];
+        const float I1wyVal = I1wy[y * I1wx_step + x];
+        const float gradVal = grad[y * I1wx_step + x];
+        const float u1OldVal = u1[(y + u1_offset_y) * u1_step + x + u1_offset_x];
+        const float u2OldVal = u2[(y + u2_offset_y) * u2_step + x + u2_offset_x];
+
+        const float rho = rho_c[y * I1wx_step + x] + (I1wxVal * u1OldVal + I1wyVal * u2OldVal);
+
+        // estimate the values of the variable (v1, v2) (thresholding operator TH)
+
+        float d1 = 0.0f;
+        float d2 = 0.0f;
+
+        if (rho < -l_t * gradVal)
+        {
+            d1 = l_t * I1wxVal;
+            d2 = l_t * I1wyVal;
+        }
+        else if (rho > l_t * gradVal)
+        {
+            d1 = -l_t * I1wxVal;
+            d2 = -l_t * I1wyVal;
+        }
+        else if (gradVal > 1.192092896e-07f)
+        {
+            const float fi = -rho / gradVal;
+            d1 = fi * I1wxVal;
+            d2 = fi * I1wyVal;
+        }
+
+        const float v1 = u1OldVal + d1;
+        const float v2 = u2OldVal + d2;
+
+        // compute the divergence of the dual variable (p1, p2)
+
+        const float div_p1 = divergence(p11, p12, y, x, I1wx_step, I1wx_step);
+        const float div_p2 = divergence(p21, p22, y, x, I1wx_step, I1wx_step);
+
+        // estimate the values of the optical flow (u1, u2)
+
+        const float u1NewVal = v1 + theta * div_p1;
+        const float u2NewVal = v2 + theta * div_p2;
+
+        u1[(y + u1_offset_y) * u1_step + x + u1_offset_x] = u1NewVal;
+        u2[(y + u2_offset_y) * u2_step + x + u2_offset_x] = u2NewVal;
+
+        if(calc_error)
+        {
+            const float n1 = (u1OldVal - u1NewVal) * (u1OldVal - u1NewVal);
+            const float n2 = (u2OldVal - u2NewVal) * (u2OldVal - u2NewVal);
+            error[y * I1wx_step + x] = n1 + n2;
+        }
+    }
+
+}
diff --git a/modules/ocl/src/optical_flow_farneback.cpp b/modules/ocl/src/optical_flow_farneback.cpp
new file mode 100644
index 0000000000..e622446bb8
--- /dev/null
+++ b/modules/ocl/src/optical_flow_farneback.cpp
@@ -0,0 +1,540 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//      Sen Liu, swjtuls1987@126.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#include "precomp.hpp"
+#include "opencv2/video/tracking.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+#define MIN_SIZE 32
+
+namespace cv
+{
+namespace ocl
+{
+///////////////////////////OpenCL kernel strings///////////////////////////
+extern const char *optical_flow_farneback;
+}
+}
+
+namespace cv {
+namespace ocl {
+namespace optflow_farneback
+{
+oclMat g;
+oclMat xg;
+oclMat xxg;
+oclMat gKer;
+
+float ig[4];
+
+inline int divUp(int total, int grain)
+{
+    return (total + grain - 1) / grain;
+}
+
+inline void setGaussianBlurKernel(const float *c_gKer, int ksizeHalf)
+{
+    cv::Mat t_gKer(1, ksizeHalf + 1, CV_32FC1, const_cast<float *>(c_gKer));
+    gKer.upload(t_gKer);
+}
+
+static void gaussianBlurOcl(const oclMat &src, int ksizeHalf, oclMat &dst)
+{
+    string kernelName("gaussianBlur");
+    size_t localThreads[3] = { 256, 1, 1 };
+    size_t globalThreads[3] = { divUp(src.cols, localThreads[0]) * localThreads[0], src.rows, 1 };
+    int smem_size = (localThreads[0] + 2*ksizeHalf) * sizeof(float);
+
+    CV_Assert(dst.size() == src.size());
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&gKer.data));
+    args.push_back(std::make_pair(smem_size, (void *)NULL));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&ksizeHalf));
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1);
+}
+
+static void polynomialExpansionOcl(const oclMat &src, int polyN, oclMat &dst)
+{
+    string kernelName("polynomialExpansion");
+    size_t localThreads[3] = { 256, 1, 1 };
+    size_t globalThreads[3] = { divUp(src.cols, localThreads[0] - 2*polyN) * localThreads[0], src.rows, 1 };
+    int smem_size = 3 * localThreads[0] * sizeof(float);
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&g.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&xg.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&xxg.data));
+    args.push_back(std::make_pair(smem_size, (void *)NULL));
+    args.push_back(std::make_pair(sizeof(cl_float4), (void *)&ig));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+
+    char opt [128];
+    sprintf(opt, "-D polyN=%d", polyN);
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1, opt);
+}
+
+static void updateMatricesOcl(const oclMat &flowx, const oclMat &flowy, const oclMat &R0, const oclMat &R1, oclMat &M)
+{
+    string kernelName("updateMatrices");
+    size_t localThreads[3] = { 32, 8, 1 };
+    size_t globalThreads[3] = { divUp(flowx.cols, localThreads[0]) * localThreads[0],
+                                divUp(flowx.rows, localThreads[1]) * localThreads[1],
+                                1
+                              };
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&M.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&flowx.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&flowy.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&R0.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&R1.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowx.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowx.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&M.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowx.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowy.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&R0.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&R1.step));
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1);
+}
+
+static void boxFilter5Ocl(const oclMat &src, int ksizeHalf, oclMat &dst)
+{
+    string kernelName("boxFilter5");
+    int height = src.rows / 5;
+    size_t localThreads[3] = { 256, 1, 1 };
+    size_t globalThreads[3] = { divUp(src.cols, localThreads[0]) * localThreads[0], height, 1 };
+    int smem_size = (localThreads[0] + 2*ksizeHalf) * 5 * sizeof(float);
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
+    args.push_back(std::make_pair(smem_size, (void *)NULL));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&height));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&ksizeHalf));
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1);
+}
+
+static void updateFlowOcl(const oclMat &M, oclMat &flowx, oclMat &flowy)
+{
+    string kernelName("updateFlow");
+    int cols = divUp(flowx.cols, 4);
+    size_t localThreads[3] = { 32, 8, 1 };
+    size_t globalThreads[3] = { divUp(cols, localThreads[0]) * localThreads[0],
+                                divUp(flowx.rows, localThreads[1]) * localThreads[0],
+                                1
+                              };
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&flowx.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&flowy.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&M.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowx.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowx.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&flowy.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&M.step));
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1);
+}
+
+static void gaussianBlur5Ocl(const oclMat &src, int ksizeHalf, oclMat &dst)
+{
+    string kernelName("gaussianBlur5");
+    int height = src.rows / 5;
+    int width = src.cols;
+    size_t localThreads[3] = { 256, 1, 1 };
+    size_t globalThreads[3] = { divUp(width, localThreads[0]) * localThreads[0], height, 1 };
+    int smem_size = (localThreads[0] + 2*ksizeHalf) * 5 * sizeof(float);
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&dst.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&src.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&gKer.data));
+    args.push_back(std::make_pair(smem_size, (void *)NULL));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&height));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&width));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&dst.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&src.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&ksizeHalf));
+
+    openCLExecuteKernel(Context::getContext(), &optical_flow_farneback, kernelName,
+                        globalThreads, localThreads, args, -1, -1);
+}
+}
+}
+} // namespace cv { namespace ocl { namespace optflow_farneback
+
+static oclMat allocMatFromBuf(int rows, int cols, int type, oclMat &mat)
+{
+    if (!mat.empty() && mat.type() == type && mat.rows >= rows && mat.cols >= cols)
+        return mat(Rect(0, 0, cols, rows));
+    return mat = oclMat(rows, cols, type);
+}
+
+cv::ocl::FarnebackOpticalFlow::FarnebackOpticalFlow()
+{
+    numLevels = 5;
+    pyrScale = 0.5;
+    fastPyramids = false;
+    winSize = 13;
+    numIters = 10;
+    polyN = 5;
+    polySigma = 1.1;
+    flags = 0;
+}
+
+void cv::ocl::FarnebackOpticalFlow::releaseMemory()
+{
+    frames_[0].release();
+    frames_[1].release();
+    pyrLevel_[0].release();
+    pyrLevel_[1].release();
+    M_.release();
+    bufM_.release();
+    R_[0].release();
+    R_[1].release();
+    blurredFrame_[0].release();
+    blurredFrame_[1].release();
+    pyramid0_.clear();
+    pyramid1_.clear();
+}
+
+void cv::ocl::FarnebackOpticalFlow::prepareGaussian(
+    int n, double sigma, float *g, float *xg, float *xxg,
+    double &ig11, double &ig03, double &ig33, double &ig55)
+{
+    double s = 0.;
+    for (int x = -n; x <= n; x++)
+    {
+        g[x] = (float)std::exp(-x*x/(2*sigma*sigma));
+        s += g[x];
+    }
+
+    s = 1./s;
+    for (int x = -n; x <= n; x++)
+    {
+        g[x] = (float)(g[x]*s);
+        xg[x] = (float)(x*g[x]);
+        xxg[x] = (float)(x*x*g[x]);
+    }
+
+    Mat_<double> G(6, 6);
+    G.setTo(0);
+
+    for (int y = -n; y <= n; y++)
+    {
+        for (int x = -n; x <= n; x++)
+        {
+            G(0,0) += g[y]*g[x];
+            G(1,1) += g[y]*g[x]*x*x;
+            G(3,3) += g[y]*g[x]*x*x*x*x;
+            G(5,5) += g[y]*g[x]*x*x*y*y;
+        }
+    }
+
+    //G[0][0] = 1.;
+    G(2,2) = G(0,3) = G(0,4) = G(3,0) = G(4,0) = G(1,1);
+    G(4,4) = G(3,3);
+    G(3,4) = G(4,3) = G(5,5);
+
+    // invG:
+    // [ x        e  e    ]
+    // [    y             ]
+    // [       y          ]
+    // [ e        z       ]
+    // [ e           z    ]
+    // [                u ]
+    Mat_<double> invG = G.inv(DECOMP_CHOLESKY);
+
+    ig11 = invG(1,1);
+    ig03 = invG(0,3);
+    ig33 = invG(3,3);
+    ig55 = invG(5,5);
+}
+
+void cv::ocl::FarnebackOpticalFlow::setPolynomialExpansionConsts(int n, double sigma)
+{
+    vector<float> buf(n*6 + 3);
+    float* g = &buf[0] + n;
+    float* xg = g + n*2 + 1;
+    float* xxg = xg + n*2 + 1;
+
+    if (sigma < FLT_EPSILON)
+        sigma = n*0.3;
+
+    double ig11, ig03, ig33, ig55;
+    prepareGaussian(n, sigma, g, xg, xxg, ig11, ig03, ig33, ig55);
+
+    cv::Mat t_g(1, n + 1, CV_32FC1, g);
+    cv::Mat t_xg(1, n + 1, CV_32FC1, xg);
+    cv::Mat t_xxg(1, n + 1, CV_32FC1, xxg);
+
+    optflow_farneback::g.upload(t_g);
+    optflow_farneback::xg.upload(t_xg);
+    optflow_farneback::xxg.upload(t_xxg);
+
+    optflow_farneback::ig[0] = static_cast<float>(ig11);
+    optflow_farneback::ig[1] = static_cast<float>(ig03);
+    optflow_farneback::ig[2] = static_cast<float>(ig33);
+    optflow_farneback::ig[3] = static_cast<float>(ig55);
+}
+
+void cv::ocl::FarnebackOpticalFlow::updateFlow_boxFilter(
+    const oclMat& R0, const oclMat& R1, oclMat& flowx, oclMat &flowy,
+    oclMat& M, oclMat &bufM, int blockSize, bool updateMatrices)
+{
+    optflow_farneback::boxFilter5Ocl(M, blockSize/2, bufM);
+
+    swap(M, bufM);
+
+    finish();
+
+    optflow_farneback::updateFlowOcl(M, flowx, flowy);
+
+    if (updateMatrices)
+        optflow_farneback::updateMatricesOcl(flowx, flowy, R0, R1, M);
+}
+
+
+void cv::ocl::FarnebackOpticalFlow::updateFlow_gaussianBlur(
+    const oclMat& R0, const oclMat& R1, oclMat& flowx, oclMat& flowy,
+    oclMat& M, oclMat &bufM, int blockSize, bool updateMatrices)
+{
+    optflow_farneback::gaussianBlur5Ocl(M, blockSize/2, bufM);
+
+    swap(M, bufM);
+
+    optflow_farneback::updateFlowOcl(M, flowx, flowy);
+
+    if (updateMatrices)
+        optflow_farneback::updateMatricesOcl(flowx, flowy, R0, R1, M);
+}
+
+
+void cv::ocl::FarnebackOpticalFlow::operator ()(
+    const oclMat &frame0, const oclMat &frame1, oclMat &flowx, oclMat &flowy)
+{
+    CV_Assert(frame0.channels() == 1 && frame1.channels() == 1);
+    CV_Assert(frame0.size() == frame1.size());
+    CV_Assert(polyN == 5 || polyN == 7);
+    CV_Assert(!fastPyramids || std::abs(pyrScale - 0.5) < 1e-6);
+
+    Size size = frame0.size();
+    oclMat prevFlowX, prevFlowY, curFlowX, curFlowY;
+
+    flowx.create(size, CV_32F);
+    flowy.create(size, CV_32F);
+    oclMat flowx0 = flowx;
+    oclMat flowy0 = flowy;
+
+    // Crop unnecessary levels
+    double scale = 1;
+    int numLevelsCropped = 0;
+    for (; numLevelsCropped < numLevels; numLevelsCropped++)
+    {
+        scale *= pyrScale;
+        if (size.width*scale < MIN_SIZE || size.height*scale < MIN_SIZE)
+            break;
+    }
+
+    frame0.convertTo(frames_[0], CV_32F);
+    frame1.convertTo(frames_[1], CV_32F);
+
+    if (fastPyramids)
+    {
+        // Build Gaussian pyramids using pyrDown()
+        pyramid0_.resize(numLevelsCropped + 1);
+        pyramid1_.resize(numLevelsCropped + 1);
+        pyramid0_[0] = frames_[0];
+        pyramid1_[0] = frames_[1];
+        for (int i = 1; i <= numLevelsCropped; ++i)
+        {
+            pyrDown(pyramid0_[i - 1], pyramid0_[i]);
+            pyrDown(pyramid1_[i - 1], pyramid1_[i]);
+        }
+    }
+
+    setPolynomialExpansionConsts(polyN, polySigma);
+
+    for (int k = numLevelsCropped; k >= 0; k--)
+    {
+        scale = 1;
+        for (int i = 0; i < k; i++)
+            scale *= pyrScale;
+
+        double sigma = (1./scale - 1) * 0.5;
+        int smoothSize = cvRound(sigma*5) | 1;
+        smoothSize = std::max(smoothSize, 3);
+
+        int width = cvRound(size.width*scale);
+        int height = cvRound(size.height*scale);
+
+        if (fastPyramids)
+        {
+            width = pyramid0_[k].cols;
+            height = pyramid0_[k].rows;
+        }
+
+        if (k > 0)
+        {
+            curFlowX.create(height, width, CV_32F);
+            curFlowY.create(height, width, CV_32F);
+        }
+        else
+        {
+            curFlowX = flowx0;
+            curFlowY = flowy0;
+        }
+
+        if (!prevFlowX.data)
+        {
+            if (flags & cv::OPTFLOW_USE_INITIAL_FLOW)
+            {
+                resize(flowx0, curFlowX, Size(width, height), 0, 0, INTER_LINEAR);
+                resize(flowy0, curFlowY, Size(width, height), 0, 0, INTER_LINEAR);
+                multiply(scale, curFlowX, curFlowX);
+                multiply(scale, curFlowY, curFlowY);
+            }
+            else
+            {
+                curFlowX.setTo(0);
+                curFlowY.setTo(0);
+            }
+        }
+        else
+        {
+            resize(prevFlowX, curFlowX, Size(width, height), 0, 0, INTER_LINEAR);
+            resize(prevFlowY, curFlowY, Size(width, height), 0, 0, INTER_LINEAR);
+            multiply(1./pyrScale, curFlowX, curFlowX);
+            multiply(1./pyrScale, curFlowY, curFlowY);
+        }
+
+        oclMat M = allocMatFromBuf(5*height, width, CV_32F, M_);
+        oclMat bufM = allocMatFromBuf(5*height, width, CV_32F, bufM_);
+        oclMat R[2] =
+        {
+            allocMatFromBuf(5*height, width, CV_32F, R_[0]),
+            allocMatFromBuf(5*height, width, CV_32F, R_[1])
+        };
+
+        if (fastPyramids)
+        {
+            optflow_farneback::polynomialExpansionOcl(pyramid0_[k], polyN, R[0]);
+            optflow_farneback::polynomialExpansionOcl(pyramid1_[k], polyN, R[1]);
+        }
+        else
+        {
+            oclMat blurredFrame[2] =
+            {
+                allocMatFromBuf(size.height, size.width, CV_32F, blurredFrame_[0]),
+                allocMatFromBuf(size.height, size.width, CV_32F, blurredFrame_[1])
+            };
+            oclMat pyrLevel[2] =
+            {
+                allocMatFromBuf(height, width, CV_32F, pyrLevel_[0]),
+                allocMatFromBuf(height, width, CV_32F, pyrLevel_[1])
+            };
+
+            Mat g = getGaussianKernel(smoothSize, sigma, CV_32F);
+            optflow_farneback::setGaussianBlurKernel(g.ptr<float>(smoothSize/2), smoothSize/2);
+
+            for (int i = 0; i < 2; i++)
+            {
+                optflow_farneback::gaussianBlurOcl(frames_[i], smoothSize/2, blurredFrame[i]);
+                resize(blurredFrame[i], pyrLevel[i], Size(width, height), INTER_LINEAR);
+                optflow_farneback::polynomialExpansionOcl(pyrLevel[i], polyN, R[i]);
+            }
+        }
+
+        optflow_farneback::updateMatricesOcl(curFlowX, curFlowY, R[0], R[1], M);
+
+        if (flags & OPTFLOW_FARNEBACK_GAUSSIAN)
+        {
+            Mat g = getGaussianKernel(winSize, winSize/2*0.3f, CV_32F);
+            optflow_farneback::setGaussianBlurKernel(g.ptr<float>(winSize/2), winSize/2);
+        }
+        for (int i = 0; i < numIters; i++)
+        {
+            if (flags & OPTFLOW_FARNEBACK_GAUSSIAN)
+                updateFlow_gaussianBlur(R[0], R[1], curFlowX, curFlowY, M, bufM, winSize, i < numIters-1);
+            else
+                updateFlow_boxFilter(R[0], R[1], curFlowX, curFlowY, M, bufM, winSize, i < numIters-1);
+        }
+
+        prevFlowX = curFlowX;
+        prevFlowY = curFlowY;
+    }
+
+    flowx = curFlowX;
+    flowy = curFlowY;
+}
diff --git a/modules/ocl/src/precomp.hpp b/modules/ocl/src/precomp.hpp
index 5cbc5364f0..cf8743fccc 100644
--- a/modules/ocl/src/precomp.hpp
+++ b/modules/ocl/src/precomp.hpp
@@ -52,10 +52,6 @@
 #pragma warning( disable: 4267 4324 4244 4251 4710 4711 4514 4996 )
 #endif
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include <map>
 #include <iostream>
 #include <limits>
@@ -65,88 +61,23 @@
 #include <exception>
 #include <stdio.h>
 
-#include "opencv2/ocl.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/core_c.h"
+#undef OPENCV_NOSTL
 
-#include "opencv2/core/internal.hpp"
-//#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
+#include "opencv2/ocl.hpp"
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
 
 #define __ATI__
 
 #if defined (HAVE_OPENCL)
 
-#if defined __APPLE__
-#include <OpenCL/OpenCL.h>
-#else
-#include <CL/opencl.h>
-#endif
-
+#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
+#include "opencv2/ocl/private/util.hpp"
 #include "safe_call.hpp"
 
-namespace cv
-{
-    namespace ocl
-    {
-        ///////////////////////////OpenCL call wrappers////////////////////////////
-        void openCLMallocPitch(Context *clCxt, void **dev_ptr, size_t *pitch,
-                               size_t widthInBytes, size_t height);
-        void openCLMallocPitchEx(Context *clCxt, void **dev_ptr, size_t *pitch,
-                               size_t widthInBytes, size_t height, DevMemRW rw_type, DevMemType mem_type);
-        void openCLMemcpy2D(Context *clCxt, void *dst, size_t dpitch,
-                            const void *src, size_t spitch,
-                            size_t width, size_t height, enum openCLMemcpyKind kind, int channels = -1);
-        void openCLCopyBuffer2D(Context *clCxt, void *dst, size_t dpitch, int dst_offset,
-                                const void *src, size_t spitch,
-                                size_t width, size_t height, int src_offset);
-        void openCLFree(void *devPtr);
-        cl_mem openCLCreateBuffer(Context *clCxt, size_t flag, size_t size);
-        void openCLReadBuffer(Context *clCxt, cl_mem dst_buffer, void *host_buffer, size_t size);
-        cl_kernel openCLGetKernelFromSource(const Context *clCxt,
-                                            const char **source, std::string kernelName);
-        cl_kernel openCLGetKernelFromSource(const Context *clCxt,
-                                            const char **source, std::string kernelName, const char *build_options);
-        void openCLVerifyKernel(const Context *clCxt, cl_kernel kernel, size_t *localThreads);
-        void openCLExecuteKernel(Context *clCxt , const char **source, std::string kernelName, std::vector< std::pair<size_t, const void *> > &args,
-                                 int globalcols , int globalrows, size_t blockSize = 16, int kernel_expand_depth = -1, int kernel_expand_channel = -1);
-        void openCLExecuteKernel_(Context *clCxt , const char **source, std::string kernelName,
-                                  size_t globalThreads[3], size_t localThreads[3],
-                                  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth, const char *build_options);
-        void openCLExecuteKernel(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
-                                 size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels, int depth);
-        void openCLExecuteKernel(Context *clCxt , const char **source, std::string kernelName, size_t globalThreads[3],
-                                 size_t localThreads[3],  std::vector< std::pair<size_t, const void *> > &args, int channels,
-                                 int depth, const char *build_options);
-
-        cl_mem load_constant(cl_context context, cl_command_queue command_queue, const void *value,
-                             const size_t size);
-
-        cl_mem openCLMalloc(cl_context clCxt, size_t size, cl_mem_flags flags, void *host_ptr);
-
-        //void openCLMemcpy2DWithNoPadding(cl_command_queue command_queue, cl_mem buffer, size_t size, size_t offset, void *ptr,
-        //                                 enum openCLMemcpyKind kind, cl_bool blocking_write);
-        int savetofile(const Context *clcxt,  cl_program &program, const char *fileName);
-        struct Context::Impl
-        {
-            //Information of the OpenCL context
-            cl_context clContext;
-            cl_command_queue clCmdQueue;
-            cl_device_id devices;
-            std::string devName;
-            cl_uint maxDimensions;
-            size_t maxWorkGroupSize;
-            size_t maxWorkItemSizes[4];
-            cl_uint maxComputeUnits;
-            int double_support;
-            //extra options to recognize vendor specific fp64 extensions
-            char extra_options[512];
-            std::string Binpath;
-            int unified_memory; //1 means integrated GPU, otherwise this value is 0
-        };
-    }
-}
-
-
 #else /* defined(HAVE_OPENCL) */
 
 static inline void throw_nogpu()
diff --git a/modules/ocl/src/pyrdown.cpp b/modules/ocl/src/pyrdown.cpp
index 6cd7cf5d48..801a29f258 100644
--- a/modules/ocl/src/pyrdown.cpp
+++ b/modules/ocl/src/pyrdown.cpp
@@ -72,7 +72,7 @@ static void pyrdown_run(const oclMat &src, const oclMat &dst)
     //int channels = dst.channels();
     //int depth = dst.depth();
 
-    std::string kernelName = "pyrDown";
+    String kernelName = "pyrDown";
 
     //int vector_lengths[4][7] = {{4, 0, 4, 4, 1, 1, 1},
     //    {4, 0, 4, 4, 1, 1, 1},
@@ -109,4 +109,3 @@ void cv::ocl::pyrDown(const oclMat &src, oclMat &dst)
 
     pyrdown_run(src, dst);
 }
-
diff --git a/modules/ocl/src/pyrlk.cpp b/modules/ocl/src/pyrlk.cpp
index 31009464bc..eae5f85b09 100644
--- a/modules/ocl/src/pyrlk.cpp
+++ b/modules/ocl/src/pyrlk.cpp
@@ -15,8 +15,8 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//		Dachuan Zhao, dachuan@multicorewareinc.com
-//		Yao Wang, yao@multicorewareinc.com
+//      Dachuan Zhao, dachuan@multicorewareinc.com
+//      Yao Wang, yao@multicorewareinc.com
 //      Nathan, liujun@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -47,7 +47,6 @@
 
 
 #include "precomp.hpp"
-#include "mcwutil.hpp"
 using namespace cv;
 using namespace cv::ocl;
 
@@ -55,35 +54,20 @@ namespace cv
 {
 namespace ocl
 {
-///////////////////////////OpenCL kernel strings///////////////////////////
 extern const char *pyrlk;
 extern const char *pyrlk_no_image;
 extern const char *operator_setTo;
 extern const char *operator_convertTo;
 extern const char *operator_copyToM;
-extern const char *arithm_mul;
 extern const char *pyr_down;
 }
 }
-
 struct dim3
 {
     unsigned int x, y, z;
 };
 
-struct float2
-{
-    float x, y;
-};
-
-struct int2
-{
-    int x, y;
-};
-
-namespace
-{
-void calcPatchSize(cv::Size winSize, int cn, dim3 &block, dim3 &patch, bool isDeviceArch11)
+static void calcPatchSize(cv::Size winSize, int cn, dim3 &block, dim3 &patch, bool isDeviceArch11)
 {
     winSize.width *= cn;
 
@@ -103,22 +87,16 @@ void calcPatchSize(cv::Size winSize, int cn, dim3 &block, dim3 &patch, bool isDe
 
     block.z = patch.z = 1;
 }
-}
-
-inline int divUp(int total, int grain)
-{
-    return (total + grain - 1) / grain;
-}
 
 ///////////////////////////////////////////////////////////////////////////
 //////////////////////////////// ConvertTo ////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////
 static void convert_run_cus(const oclMat &src, oclMat &dst, double alpha, double beta)
 {
-    std::string kernelName = "convert_to_S";
+    String kernelName = "convert_to_S";
     std::stringstream idxStr;
     idxStr << src.depth();
-    kernelName += idxStr.str();
+    kernelName = kernelName + idxStr.str().c_str();
     float alpha_f = (float)alpha, beta_f = (float)beta;
     CV_DbgAssert(src.rows == dst.rows && src.cols == dst.cols);
     std::vector<std::pair<size_t , const void *> > args;
@@ -184,7 +162,7 @@ void convertTo( const oclMat &src, oclMat &dst, int rtype, double alpha, double
 //    setTo(s);
 //    return *this;
 //}
-static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar, std::string kernelName)
+static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar, String kernelName)
 {
     std::vector<std::pair<size_t , const void *> > args;
 
@@ -227,7 +205,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 1:
@@ -246,7 +224,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_char4) , (void *)&val.cval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 2:
@@ -265,7 +243,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 3:
@@ -284,7 +262,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_short4) , (void *)&val.shval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 4:
@@ -310,7 +288,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_int4) , (void *)&val.ival ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 5:
@@ -329,7 +307,7 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_float4) , (void *)&val.fval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     case 6:
@@ -348,16 +326,16 @@ static void set_to_withoutmask_run_cus(const oclMat &dst, const Scalar &scalar,
             args.push_back( std::make_pair( sizeof(cl_double4) , (void *)&val.dval ));
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
+            CV_Error(Error::StsUnsupportedFormat, "unsupported channels");
         }
         break;
     default:
-        CV_Error(CV_StsUnsupportedFormat, "unknown depth");
+        CV_Error(Error::StsUnsupportedFormat, "unknown depth");
     }
 #ifdef CL_VERSION_1_2
     if(dst.offset == 0 && dst.cols == dst.wholecols)
     {
-        clEnqueueFillBuffer(dst.clCxt->impl->clCmdQueue, (cl_mem)dst.data, args[0].second, args[0].first, 0, dst.step * dst.rows, 0, NULL, NULL);
+        clEnqueueFillBuffer((cl_command_queue)dst.clCxt->oclCommandQueue(), (cl_mem)dst.data, args[0].second, args[0].first, 0, dst.step * dst.rows, 0, NULL, NULL);
     }
     else
     {
@@ -400,7 +378,7 @@ static oclMat &setTo(oclMat &src, const Scalar &scalar)
 ///////////////////////////////////////////////////////////////////////////
 ////////////////////////////////// CopyTo /////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////
-// static void copy_to_with_mask_cus(const oclMat &src, oclMat &dst, const oclMat &mask, std::string kernelName)
+// static void copy_to_with_mask_cus(const oclMat &src, oclMat &dst, const oclMat &mask, String kernelName)
 // {
 //     CV_DbgAssert( dst.rows == mask.rows && dst.cols == mask.cols &&
 //                   src.rows == dst.rows && src.cols == dst.cols
@@ -408,7 +386,7 @@ static oclMat &setTo(oclMat &src, const Scalar &scalar)
 
 //     std::vector<std::pair<size_t , const void *> > args;
 
-//     std::string string_types[4][7] = {{"uchar", "char", "ushort", "short", "int", "float", "double"},
+//     String string_types[4][7] = {{"uchar", "char", "ushort", "short", "int", "float", "double"},
 //         {"uchar2", "char2", "ushort2", "short2", "int2", "float2", "double2"},
 //         {"uchar3", "char3", "ushort3", "short3", "int3", "float3", "double3"},
 //         {"uchar4", "char4", "ushort4", "short4", "int4", "float4", "double4"}
@@ -449,89 +427,6 @@ static void copyTo(const oclMat &src, oclMat &m )
                        src.data, src.step, src.cols * src.elemSize(), src.rows, src.offset);
 }
 
-// static void copyTo(const oclMat &src, oclMat &mat, const oclMat &mask)
-// {
-//     if (mask.empty())
-//     {
-//         copyTo(src, mat);
-//     }
-//     else
-//     {
-//         mat.create(src.size(), src.type());
-//         copy_to_with_mask_cus(src, mat, mask, "copy_to_with_mask");
-//     }
-// }
-
-static void arithmetic_run(const oclMat &src1, oclMat &dst, std::string kernelName, const char **kernelString, void *_scalar)
-{
-    if(src1.clCxt -> impl -> double_support == 0 && src1.type() == CV_64F)
-    {
-        CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
-        return;
-    }
-
-    //dst.create(src1.size(), src1.type());
-    //CV_Assert(src1.cols == src2.cols && src2.cols == dst.cols &&
-    //          src1.rows == src2.rows && src2.rows == dst.rows);
-    CV_Assert(src1.cols == dst.cols &&
-              src1.rows == dst.rows);
-
-    CV_Assert(src1.type() == dst.type());
-    CV_Assert(src1.depth() != CV_8S);
-
-    Context  *clCxt = src1.clCxt;
-    //int channels = dst.channels();
-    //int depth = dst.depth();
-
-    //int vector_lengths[4][7] = {{4, 0, 4, 4, 1, 1, 1},
-    //    {4, 0, 4, 4, 1, 1, 1},
-    //    {4, 0, 4, 4, 1, 1, 1},
-    //    {4, 0, 4, 4, 1, 1, 1}
-    //};
-
-    //size_t vector_length = vector_lengths[channels-1][depth];
-    //int offset_cols = (dst.offset / dst.elemSize1()) & (vector_length - 1);
-    //int cols = divUp(dst.cols * channels + offset_cols, vector_length);
-
-    size_t localThreads[3]  = { 16, 16, 1 };
-    //size_t globalThreads[3] = { divUp(cols, localThreads[0]) * localThreads[0],
-    //                               divUp(dst.rows, localThreads[1]) * localThreads[1],
-    //                               1
-    //                             };
-    size_t globalThreads[3] = { src1.cols,
-                                src1.rows,
-                                1
-                              };
-
-    int dst_step1 = dst.cols * dst.elemSize();
-    std::vector<std::pair<size_t , const void *> > args;
-    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src1.data ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&src1.step ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&src1.offset ));
-    //args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src2.data ));
-    //args.push_back( std::make_pair( sizeof(cl_int), (void *)&src2.step ));
-    //args.push_back( std::make_pair( sizeof(cl_int), (void *)&src2.offset ));
-    args.push_back( std::make_pair( sizeof(cl_mem), (void *)&dst.data ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst.step ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst.offset ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&src1.rows ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&src1.cols ));
-    args.push_back( std::make_pair( sizeof(cl_int), (void *)&dst_step1 ));
-
-    //if(_scalar != NULL)
-    //{
-    float scalar1 = *((float *)_scalar);
-    args.push_back( std::make_pair( sizeof(float), (float *)&scalar1 ));
-    //}
-
-    openCLExecuteKernel2(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, src1.depth(), CLFLUSH);
-}
-
-static void multiply_cus(const oclMat &src1, oclMat &dst, float scalar)
-{
-    arithmetic_run(src1, dst, "arithm_muls", &arithm_mul, (void *)(&scalar));
-}
-
 static void pyrdown_run_cus(const oclMat &src, const oclMat &dst)
 {
 
@@ -540,7 +435,7 @@ static void pyrdown_run_cus(const oclMat &src, const oclMat &dst)
 
     Context  *clCxt = src.clCxt;
 
-    std::string kernelName = "pyrDown";
+    String kernelName = "pyrDown";
 
     size_t localThreads[3]  = { 256, 1, 1 };
     size_t globalThreads[3] = { src.cols, dst.rows, 1};
@@ -572,20 +467,12 @@ static void lkSparse_run(oclMat &I, oclMat &J,
 {
     Context  *clCxt = I.clCxt;
     int elemCntPerRow = I.step / I.elemSize();
-    std::string kernelName = "lkSparse";
+    String kernelName = "lkSparse";
     bool isImageSupported = support_image2d();
     size_t localThreads[3]  = { 8, isImageSupported ? 8 : 32, 1 };
     size_t globalThreads[3] = { 8 * ptcount, isImageSupported ? 8 : 32, 1};
     int cn = I.oclchannels();
-    char calcErr;
-    if (level == 0)
-    {
-        calcErr = 1;
-    }
-    else
-    {
-        calcErr = 0;
-    }
+    char calcErr = level==0?1:0;
 
     std::vector<std::pair<size_t , const void *> > args;
 
@@ -615,7 +502,16 @@ static void lkSparse_run(oclMat &I, oclMat &J,
 
     if(isImageSupported)
     {
-        openCLExecuteKernel2(clCxt, &pyrlk, kernelName, globalThreads, localThreads, args, I.oclchannels(), I.depth(), CLFLUSH);
+        std::stringstream idxStr;
+        idxStr << kernelName.c_str() << "_C" << I.oclchannels() << "_D" << I.depth();
+        cl_kernel kernel = openCLGetKernelFromSource(clCxt, &pyrlk, idxStr.str().c_str());
+        int wave_size = queryDeviceInfo<WAVEFRONT_SIZE, int>(kernel);
+        openCLSafeCall(clReleaseKernel(kernel));
+
+        static char opt[32] = {0};
+        sprintf(opt, " -D WAVE_SIZE=%d", wave_size);
+
+        openCLExecuteKernel2(clCxt, &pyrlk, kernelName, globalThreads, localThreads, args, I.oclchannels(), I.depth(), opt, CLFLUSH);
         releaseTexture(ITex);
         releaseTexture(JTex);
     }
@@ -657,9 +553,7 @@ void cv::ocl::PyrLKOpticalFlow::sparse(const oclMat &prevImg, const oclMat &next
 
     oclMat temp1 = (useInitialFlow ? nextPts : prevPts).reshape(1);
     oclMat temp2 = nextPts.reshape(1);
-    //oclMat scalar(temp1.rows, temp1.cols, temp1.type(), Scalar(1.0f / (1 << maxLevel) / 2.0f));
-    multiply_cus(temp1, temp2, 1.0f / (1 << maxLevel) / 2.0f);
-    //::multiply(temp1, 1.0f / (1 << maxLevel) / 2.0f, temp2);
+    multiply(1.0f/(1<<maxLevel)/2.0f, temp1, temp2);
 
     ensureSizeIsEnough(1, prevPts.cols, CV_8UC1, status);
     //status.setTo(Scalar::all(1));
@@ -676,7 +570,6 @@ void cv::ocl::PyrLKOpticalFlow::sparse(const oclMat &prevImg, const oclMat &next
     //ensureSizeIsEnough(1, prevPts.cols, CV_32FC1, err);
 
     // build the image pyramids.
-
     prevPyr_.resize(maxLevel + 1);
     nextPyr_.resize(maxLevel + 1);
 
@@ -704,7 +597,6 @@ void cv::ocl::PyrLKOpticalFlow::sparse(const oclMat &prevImg, const oclMat &next
     }
 
     // dI/dx ~ Ix, dI/dy ~ Iy
-
     for (int level = maxLevel; level >= 0; level--)
     {
         lkSparse_run(prevPyr_[level], nextPyr_[level],
@@ -712,7 +604,7 @@ void cv::ocl::PyrLKOpticalFlow::sparse(const oclMat &prevImg, const oclMat &next
                      level, /*block, */patch, winSize, iters);
     }
 
-    clFinish(prevImg.clCxt->impl->clCmdQueue);
+    clFinish((cl_command_queue)prevImg.clCxt->oclCommandQueue());
 
     if(errMat)
         delete err;
@@ -725,7 +617,7 @@ static void lkDense_run(oclMat &I, oclMat &J, oclMat &u, oclMat &v,
     bool isImageSupported = support_image2d();
     int elemCntPerRow = I.step / I.elemSize();
 
-    std::string kernelName = "lkDense";
+    String kernelName = "lkDense";
 
     size_t localThreads[3]  = { 16, 16, 1 };
     size_t globalThreads[3] = { I.cols, I.rows, 1};
@@ -851,5 +743,5 @@ void cv::ocl::PyrLKOpticalFlow::dense(const oclMat &prevImg, const oclMat &nextI
     copyTo(uPyr_[idx], u);
     copyTo(vPyr_[idx], v);
 
-    clFinish(prevImg.clCxt->impl->clCmdQueue);
+    clFinish((cl_command_queue)prevImg.clCxt->oclCommandQueue());
 }
diff --git a/modules/ocl/src/pyrup.cpp b/modules/ocl/src/pyrup.cpp
index 5cec5ba278..1f80f4fb33 100644
--- a/modules/ocl/src/pyrup.cpp
+++ b/modules/ocl/src/pyrup.cpp
@@ -64,7 +64,7 @@ namespace cv
 
             Context *clCxt = src.clCxt;
 
-            const std::string kernelName = "pyrUp";
+            const String kernelName = "pyrUp";
 
             std::vector< std::pair<size_t, const void *> > args;
             args.push_back( std::make_pair( sizeof(cl_mem), (void *)&src.data));
@@ -85,4 +85,4 @@ namespace cv
             openCLExecuteKernel(clCxt, &pyr_up, kernelName, globalThreads, localThreads, args, src.oclchannels(), src.depth());
         }
     }
-}
\ No newline at end of file
+}
diff --git a/modules/ocl/src/safe_call.hpp b/modules/ocl/src/safe_call.hpp
index c8c19f6edb..4cf39b85e8 100644
--- a/modules/ocl/src/safe_call.hpp
+++ b/modules/ocl/src/safe_call.hpp
@@ -47,7 +47,7 @@
 #define __OPENCV_OPENCL_SAFE_CALL_HPP__
 
 #if defined __APPLE__
-#include <OpenCL/OpenCL.h>
+#include <OpenCL/opencl.h>
 #else
 #include <CL/cl.h>
 #endif
@@ -65,19 +65,12 @@ namespace cv
 {
     namespace ocl
     {
-        enum openCLMemcpyKind
-        {
-            clMemcpyHostToDevice = 0,
-            clMemcpyDeviceToHost,
-            clMemcpyDeviceToDevice
-        };
-        void error( const char *error_string, const char *file, const int line, const char *func = "");
         const char *getOpenCLErrorString( int err );
 
         static inline void ___openCLSafeCall(int err, const char *file, const int line, const char *func = "")
         {
             if( CL_SUCCESS != err)
-                cv::ocl::error(getOpenCLErrorString(err), file, line, func);
+                cv::error(Error::OpenCLApiCallError, getOpenCLErrorString(err), func, file, line);
         }
     }
 }
diff --git a/modules/ocl/src/sort_by_key.cpp b/modules/ocl/src/sort_by_key.cpp
new file mode 100644
index 0000000000..0025f0d911
--- /dev/null
+++ b/modules/ocl/src/sort_by_key.cpp
@@ -0,0 +1,454 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include <iomanip>
+#include "precomp.hpp"
+
+namespace cv
+{
+namespace ocl
+{
+
+extern const char * kernel_sort_by_key;
+extern const char * kernel_stablesort_by_key;
+extern const char * kernel_radix_sort_by_key;
+
+void sortByKey(oclMat& keys, oclMat& vals, size_t vecSize, int method, bool isGreaterThan);
+
+//TODO(pengx17): change this value depending on device other than a constant
+const static unsigned int GROUP_SIZE = 256;
+
+const char * depth_strings[] =
+{
+    "uchar",   //CV_8U
+    "char",    //CV_8S
+    "ushort",  //CV_16U
+    "short",   //CV_16S
+    "int",     //CV_32S
+    "float",   //CV_32F
+    "double"   //CV_64F
+};
+
+void static genSortBuildOption(const oclMat& keys, const oclMat& vals, bool isGreaterThan, char * build_opt_buf)
+{
+    sprintf(build_opt_buf, "-D IS_GT=%d -D K_T=%s -D V_T=%s",
+            isGreaterThan?1:0, depth_strings[keys.depth()], depth_strings[vals.depth()]);
+    if(vals.oclchannels() > 1)
+    {
+        sprintf( build_opt_buf + strlen(build_opt_buf), "%d", vals.oclchannels());
+    }
+}
+inline bool isSizePowerOf2(size_t size)
+{
+    return ((size - 1) & (size)) == 0;
+}
+
+namespace bitonic_sort
+{
+static void sortByKey(oclMat& keys, oclMat& vals, size_t vecSize, bool isGreaterThan)
+{
+    CV_Assert(isSizePowerOf2(vecSize));
+
+    Context * cxt = Context::getContext();
+    size_t globalThreads[3] = {vecSize / 2, 1, 1};
+    size_t localThreads[3]  = {GROUP_SIZE, 1, 1};
+
+    // 2^numStages should be equal to vecSize or the output is invalid
+    int numStages = 0;
+    for(int i = vecSize; i > 1; i >>= 1)
+    {
+        ++numStages;
+    }
+    char build_opt_buf [100];
+    genSortBuildOption(keys, vals, isGreaterThan, build_opt_buf);
+    const int argc = 5;
+    std::vector< std::pair<size_t, const void *> > args(argc);
+    String kernelname = "bitonicSort";
+
+    args[0] = std::make_pair(sizeof(cl_mem), (void *)&keys.data);
+    args[1] = std::make_pair(sizeof(cl_mem), (void *)&vals.data);
+    args[2] = std::make_pair(sizeof(cl_int), (void *)&vecSize);
+
+    for(int stage = 0; stage < numStages; ++stage)
+    {
+        args[3] = std::make_pair(sizeof(cl_int), (void *)&stage);
+        for(int passOfStage = 0; passOfStage < stage + 1; ++passOfStage)
+        {
+            args[4] = std::make_pair(sizeof(cl_int), (void *)&passOfStage);
+            openCLExecuteKernel(cxt, &kernel_sort_by_key, kernelname, globalThreads, localThreads, args, -1, -1, build_opt_buf);
+        }
+    }
+}
+}  /* bitonic_sort */
+
+namespace selection_sort
+{
+// FIXME:
+// This function cannot sort arrays with duplicated keys
+static void sortByKey(oclMat& keys, oclMat& vals, size_t vecSize, bool isGreaterThan)
+{
+    CV_Error(-1, "This function is incorrect at the moment.");
+    Context * cxt = Context::getContext();
+
+    size_t globalThreads[3] = {vecSize, 1, 1};
+    size_t localThreads[3]  = {GROUP_SIZE, 1, 1};
+
+    std::vector< std::pair<size_t, const void *> > args;
+    char build_opt_buf [100];
+    genSortBuildOption(keys, vals, isGreaterThan, build_opt_buf);
+
+    //local
+    String kernelname = "selectionSortLocal";
+    int lds_size = GROUP_SIZE * keys.elemSize();
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&keys.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&vals.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&vecSize));
+    args.push_back(std::make_pair(lds_size,       (void*)NULL));
+
+    openCLExecuteKernel(cxt, &kernel_sort_by_key, kernelname, globalThreads, localThreads, args, -1, -1, build_opt_buf);
+
+    //final
+    kernelname = "selectionSortFinal";
+    args.pop_back();
+    openCLExecuteKernel(cxt, &kernel_sort_by_key, kernelname, globalThreads, localThreads, args, -1, -1, build_opt_buf);
+}
+
+}  /* selection_sort */
+
+
+namespace radix_sort
+{
+//FIXME(pengx17):
+// exclusive scan, need to be optimized as this is too naive...
+//void naive_scan_addition(oclMat& input, oclMat& output)
+//{
+//    Context * cxt = Context::getContext();
+//    size_t vecSize = input.cols;
+//    size_t globalThreads[3] = {1, 1, 1};
+//    size_t localThreads[3]  = {1, 1, 1};
+//
+//    String kernelname = "naiveScanAddition";
+//
+//    std::vector< std::pair<size_t, const void *> > args;
+//    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&input.data));
+//    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&output.data));
+//    args.push_back(std::make_pair(sizeof(cl_int), (void *)&vecSize));
+//    openCLExecuteKernel(cxt, &kernel_radix_sort_by_key, kernelname, globalThreads, localThreads, args, -1, -1);
+//}
+
+void static naive_scan_addition_cpu(oclMat& input, oclMat& output)
+{
+    Mat m_input = input, m_output(output.size(), output.type());
+    MatIterator_<int> i_mit = m_input.begin<int>();
+    MatIterator_<int> o_mit = m_output.begin<int>();
+    *o_mit = 0;
+    ++i_mit;
+    ++o_mit;
+    for(; i_mit != m_input.end<int>(); ++i_mit, ++o_mit)
+    {
+        *o_mit = *(o_mit - 1) + *(i_mit - 1);
+    }
+    output = m_output;
+}
+
+
+//radix sort ported from Bolt
+static void sortByKey(oclMat& keys, oclMat& vals, size_t origVecSize, bool isGreaterThan)
+{
+    CV_Assert(keys.depth() == CV_32S || keys.depth() == CV_32F); // we assume keys are 4 bytes
+
+    bool isKeyFloat = keys.type() == CV_32F;
+
+    const int RADIX = 4; //Now you cannot replace this with Radix 8 since there is a
+                         //local array of 16 elements in the histogram kernel.
+    const int RADICES = (1 << RADIX); //Values handeled by each work-item?
+
+    bool  newBuffer = false;
+    size_t vecSize = origVecSize;
+
+    unsigned int groupSize  = RADICES;
+
+    size_t mulFactor = groupSize * RADICES;
+
+    oclMat buffer_keys, buffer_vals;
+
+    if(origVecSize % mulFactor != 0)
+    {
+        vecSize = ((vecSize + mulFactor) / mulFactor) * mulFactor;
+        buffer_keys.create(1, vecSize, keys.type());
+        buffer_vals.create(1, vecSize, vals.type());
+        Scalar padding_value;
+        oclMat roi_buffer_vals = buffer_vals(Rect(0,0,origVecSize,1));
+
+        if(isGreaterThan)
+        {
+            switch(buffer_keys.depth())
+            {
+            case CV_32F:
+                padding_value = Scalar::all(-FLT_MAX);
+                break;
+            case CV_32S:
+                padding_value = Scalar::all(INT_MIN);
+                break;
+            }
+        }
+        else
+        {
+            switch(buffer_keys.depth())
+            {
+            case CV_32F:
+                padding_value = Scalar::all(FLT_MAX);
+                break;
+            case CV_32S:
+                padding_value = Scalar::all(INT_MAX);
+                break;
+            }
+        }
+        ocl::copyMakeBorder(
+            keys(Rect(0,0,origVecSize,1)), buffer_keys,
+            0, 0, 0, vecSize - origVecSize,
+            BORDER_CONSTANT, padding_value);
+        vals(Rect(0,0,origVecSize,1)).copyTo(roi_buffer_vals);
+        newBuffer = true;
+    }
+    else
+    {
+        buffer_keys = keys;
+        buffer_vals = vals;
+        newBuffer = false;
+    }
+    oclMat swap_input_keys(1, vecSize, keys.type());
+    oclMat swap_input_vals(1, vecSize, vals.type());
+    oclMat hist_bin_keys(1, vecSize, CV_32SC1);
+    oclMat hist_bin_dest_keys(1, vecSize, CV_32SC1);
+
+    Context * cxt = Context::getContext();
+
+    size_t globalThreads[3] = {vecSize / RADICES, 1, 1};
+    size_t localThreads[3]  = {groupSize, 1, 1};
+
+    std::vector< std::pair<size_t, const void *> > args;
+    char build_opt_buf [100];
+    genSortBuildOption(keys, vals, isGreaterThan, build_opt_buf);
+
+    //additional build option for radix sort
+    sprintf(build_opt_buf + strlen(build_opt_buf), " -D K_%s", isKeyFloat?"FLT":"INT");
+
+    String kernelnames[2] = {String("histogramRadixN"), String("permuteRadixN")};
+
+    int swap = 0;
+    for(int bits = 0; bits < (static_cast<int>(keys.elemSize()) * 8); bits += RADIX)
+    {
+        args.clear();
+        //Do a histogram pass locally
+        if(swap == 0)
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&buffer_keys.data));
+        }
+        else
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&swap_input_keys.data));
+        }
+        args.push_back(std::make_pair(sizeof(cl_mem), (void *)&hist_bin_keys.data));
+        args.push_back(std::make_pair(sizeof(cl_int), (void *)&bits));
+        openCLExecuteKernel(cxt, &kernel_radix_sort_by_key, kernelnames[0], globalThreads, localThreads,
+            args, -1, -1, build_opt_buf);
+
+        args.clear();
+        //Perform a global scan
+        naive_scan_addition_cpu(hist_bin_keys, hist_bin_dest_keys);
+        // end of scan
+        if(swap == 0)
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&buffer_keys.data));
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&buffer_vals.data));
+        }
+        else
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&swap_input_keys.data));
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&swap_input_vals.data));
+        }
+        args.push_back(std::make_pair(sizeof(cl_mem), (void *)&hist_bin_dest_keys.data));
+        args.push_back(std::make_pair(sizeof(cl_int), (void *)&bits));
+
+        if(swap == 0)
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&swap_input_keys.data));
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&swap_input_vals.data));
+        }
+        else
+        {
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&buffer_keys.data));
+            args.push_back(std::make_pair(sizeof(cl_mem), (void *)&buffer_vals.data));
+        }
+        openCLExecuteKernel(cxt, &kernel_radix_sort_by_key, kernelnames[1], globalThreads, localThreads,
+            args, -1, -1, build_opt_buf);
+        swap = swap ? 0 : 1;
+    }
+    if(newBuffer)
+    {
+        buffer_keys(Rect(0,0,origVecSize,1)).copyTo(keys);
+        buffer_vals(Rect(0,0,origVecSize,1)).copyTo(vals);
+    }
+}
+
+}  /* radix_sort */
+
+namespace merge_sort
+{
+static void sortByKey(oclMat& keys, oclMat& vals, size_t vecSize, bool isGreaterThan)
+{
+    Context * cxt = Context::getContext();
+
+    size_t globalThreads[3] = {vecSize, 1, 1};
+    size_t localThreads[3]  = {GROUP_SIZE, 1, 1};
+
+    std::vector< std::pair<size_t, const void *> > args;
+    char build_opt_buf [100];
+    genSortBuildOption(keys, vals, isGreaterThan, build_opt_buf);
+
+    String kernelname[] = {String("blockInsertionSort"), String("merge")};
+    int keylds_size = GROUP_SIZE * keys.elemSize();
+    int vallds_size = GROUP_SIZE * vals.elemSize();
+    args.push_back(std::make_pair(sizeof(cl_mem),  (void *)&keys.data));
+    args.push_back(std::make_pair(sizeof(cl_mem),  (void *)&vals.data));
+    args.push_back(std::make_pair(sizeof(cl_uint), (void *)&vecSize));
+    args.push_back(std::make_pair(keylds_size,     (void*)NULL));
+    args.push_back(std::make_pair(vallds_size,     (void*)NULL));
+
+    openCLExecuteKernel(cxt, &kernel_stablesort_by_key, kernelname[0], globalThreads, localThreads, args, -1, -1, build_opt_buf);
+
+    //  Early exit for the case of no merge passes, values are already in destination vector
+    if(vecSize <= GROUP_SIZE)
+    {
+        return;
+    }
+
+    //  An odd number of elements requires an extra merge pass to sort
+    size_t numMerges = 0;
+    //  Calculate the log2 of vecSize, taking into acvecSize our block size from kernel 1 is 64
+    //  this is how many merge passes we want
+    size_t log2BlockSize = vecSize >> 6;
+    for( ; log2BlockSize > 1; log2BlockSize >>= 1 )
+    {
+        ++numMerges;
+    }
+    //  Check to see if the input vector size is a power of 2, if not we will need last merge pass
+    numMerges += isSizePowerOf2(vecSize)? 1: 0;
+
+    //  Allocate a flipflop buffer because the merge passes are out of place
+    oclMat tmpKeyBuffer(keys.size(), keys.type());
+    oclMat tmpValBuffer(vals.size(), vals.type());
+    args.resize(8);
+
+    args[4] = std::make_pair(sizeof(cl_uint), (void *)&vecSize);
+    args[6] = std::make_pair(keylds_size,    (void*)NULL);
+    args[7] = std::make_pair(vallds_size,    (void*)NULL);
+
+    for(size_t pass = 1; pass <= numMerges; ++pass )
+    {
+        //  For each pass, flip the input-output buffers
+        if( pass & 0x1 )
+        {
+            args[0] = std::make_pair(sizeof(cl_mem), (void *)&keys.data);
+            args[1] = std::make_pair(sizeof(cl_mem), (void *)&vals.data);
+            args[2] = std::make_pair(sizeof(cl_mem), (void *)&tmpKeyBuffer.data);
+            args[3] = std::make_pair(sizeof(cl_mem), (void *)&tmpValBuffer.data);
+        }
+        else
+        {
+            args[0] = std::make_pair(sizeof(cl_mem), (void *)&tmpKeyBuffer.data);
+            args[1] = std::make_pair(sizeof(cl_mem), (void *)&tmpValBuffer.data);
+            args[2] = std::make_pair(sizeof(cl_mem), (void *)&keys.data);
+            args[3] = std::make_pair(sizeof(cl_mem), (void *)&vals.data);
+        }
+        //  For each pass, the merge window doubles
+        unsigned int srcLogicalBlockSize = static_cast<unsigned int>( localThreads[0] << (pass-1) );
+        args[5] = std::make_pair(sizeof(cl_uint), (void *)&srcLogicalBlockSize);
+        openCLExecuteKernel(cxt, &kernel_stablesort_by_key, kernelname[1], globalThreads, localThreads, args, -1, -1, build_opt_buf);
+    }
+    //  If there are an odd number of merges, then the output data is sitting in the temp buffer.  We need to copy
+    //  the results back into the input array
+    if( numMerges & 1 )
+    {
+        tmpKeyBuffer.copyTo(keys);
+        tmpValBuffer.copyTo(vals);
+    }
+}
+}  /* merge_sort */
+
+}
+} /* namespace cv { namespace ocl */
+
+
+void cv::ocl::sortByKey(oclMat& keys, oclMat& vals, size_t vecSize, int method, bool isGreaterThan)
+{
+    CV_Assert( keys.rows == 1 ); // we only allow one dimensional input
+    CV_Assert( keys.channels() == 1 ); // we only allow one channel keys
+    CV_Assert( vecSize <= static_cast<size_t>(keys.cols) );
+    switch(method)
+    {
+    case SORT_BITONIC:
+        bitonic_sort::sortByKey(keys, vals, vecSize, isGreaterThan);
+        break;
+    case SORT_SELECTION:
+        selection_sort::sortByKey(keys, vals, vecSize, isGreaterThan);
+        break;
+    case SORT_RADIX:
+        radix_sort::sortByKey(keys, vals, vecSize, isGreaterThan);
+        break;
+    case SORT_MERGE:
+        merge_sort::sortByKey(keys, vals, vecSize, isGreaterThan);
+        break;
+    }
+}
+
+void cv::ocl::sortByKey(oclMat& keys, oclMat& vals, int method, bool isGreaterThan)
+{
+    CV_Assert( keys.size() == vals.size() );
+    CV_Assert( keys.rows == 1 ); // we only allow one dimensional input
+    size_t vecSize = static_cast<size_t>(keys.cols);
+    sortByKey(keys, vals, vecSize, method, isGreaterThan);
+}
diff --git a/modules/ocl/src/split_merge.cpp b/modules/ocl/src/split_merge.cpp
index 1916f50dfe..571193c2fc 100644
--- a/modules/ocl/src/split_merge.cpp
+++ b/modules/ocl/src/split_merge.cpp
@@ -84,7 +84,7 @@ namespace cv
             //     int channels = mat_dst.oclchannels();
             //     int depth = mat_dst.depth();
 
-            //     std::string kernelName = "merge_vector";
+            //     String kernelName = "merge_vector";
 
             //     int indexes[4][7] = {{0, 0, 0, 0, 0, 0, 0},
             //         {4, 4, 2, 2, 1, 1, 1},
@@ -125,9 +125,9 @@ namespace cv
 
             static void merge_vector_run(const oclMat *mat_src, size_t n, oclMat &mat_dst)
             {
-                if(mat_dst.clCxt -> impl -> double_support == 0 && mat_dst.type() == CV_64F)
+                if(!mat_dst.clCxt->supportsFeature(Context::CL_DOUBLE) && mat_dst.type() == CV_64F)
                 {
-                    CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+                    CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
                     return;
                 }
 
@@ -135,7 +135,7 @@ namespace cv
                 int channels = mat_dst.oclchannels();
                 int depth = mat_dst.depth();
 
-                std::string kernelName = "merge_vector";
+                String kernelName = "merge_vector";
 
                 int vector_lengths[4][7] = {{0, 0, 0, 0, 0, 0, 0},
                     {2, 2, 1, 1, 1, 1, 1},
@@ -233,7 +233,7 @@ namespace cv
             //     int channels = mat_src.oclchannels();
             //     int depth = mat_src.depth();
 
-            //     std::string kernelName = "split_vector";
+            //     String kernelName = "split_vector";
 
             //     int indexes[4][7] = {{0, 0, 0, 0, 0, 0, 0},
             //         {8, 8, 8, 8, 4, 4, 2},
@@ -274,9 +274,9 @@ namespace cv
             static void split_vector_run(const oclMat &mat_src, oclMat *mat_dst)
             {
 
-                if(mat_src.clCxt -> impl -> double_support == 0 && mat_src.type() == CV_64F)
+                if(!mat_src.clCxt->supportsFeature(Context::CL_DOUBLE) && mat_src.type() == CV_64F)
                 {
-                    CV_Error(CV_GpuNotSupported, "Selected device don't support double\r\n");
+                    CV_Error(Error::GpuNotSupported, "Selected device don't support double\r\n");
                     return;
                 }
 
@@ -284,7 +284,7 @@ namespace cv
                 int channels = mat_src.oclchannels();
                 int depth = mat_src.depth();
 
-                std::string kernelName = "split_vector";
+                String kernelName = "split_vector";
 
                 int vector_lengths[4][7] = {{0, 0, 0, 0, 0, 0, 0},
                     {4, 4, 2, 2, 1, 1, 1},
diff --git a/modules/ocl/src/stereo_csbp.cpp b/modules/ocl/src/stereo_csbp.cpp
new file mode 100644
index 0000000000..f124488b94
--- /dev/null
+++ b/modules/ocl/src/stereo_csbp.cpp
@@ -0,0 +1,756 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Jin Ma, jin@multicorewareinc.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::ocl;
+using namespace std;
+
+#if !defined (HAVE_OPENCL)
+
+namespace cv
+{
+    namespace ocl
+    {
+
+        void cv::ocl::StereoConstantSpaceBP::estimateRecommendedParams(int, int, int &, int &, int &, int &)
+        {
+            throw_nogpu();
+        }
+        cv::ocl::StereoConstantSpaceBP::StereoConstantSpaceBP(int, int, int, int, int)
+        {
+            throw_nogpu();
+        }
+        cv::ocl::StereoConstantSpaceBP::StereoConstantSpaceBP(int, int, int, int, float, float,
+            float, float, int, int)
+        {
+            throw_nogpu();
+        }
+
+        void cv::ocl::StereoConstantSpaceBP::operator()(const oclMat &, const oclMat &, oclMat &)
+        {
+            throw_nogpu();
+        }
+    }
+}
+
+#else /* !defined (HAVE_OPENCL) */
+
+namespace cv
+{
+    namespace ocl
+    {
+
+        ///////////////////////////OpenCL kernel Strings///////////////////////////
+        extern const char *stereocsbp;
+    }
+
+}
+namespace cv
+{
+    namespace ocl
+    {
+        namespace stereoCSBP
+        {
+            //////////////////////////////////////////////////////////////////////////
+            //////////////////////////////common////////////////////////////////////
+            ////////////////////////////////////////////////////////////////////////
+            static inline int divUp(int total, int grain)
+            {
+                return (total + grain - 1) / grain;
+            }
+            static String get_kernel_name(String kernel_name, int data_type)
+            {
+                return kernel_name + (data_type == CV_16S ? "0" : "1");
+            }
+            using cv::ocl::StereoConstantSpaceBP;
+            //////////////////////////////////////////////////////////////////////////////////
+            /////////////////////////////////init_data_cost//////////////////////////////////
+            //////////////////////////////////////////////////////////////////////////////////
+            static void init_data_cost_caller(const oclMat &left, const oclMat &right, oclMat &temp,
+                StereoConstantSpaceBP &rthis,
+                int msg_step, int h, int w, int level)
+            {
+                Context  *clCxt = left.clCxt;
+                int data_type = rthis.msg_type;
+                int channels = left.oclchannels();
+
+                String kernelName = get_kernel_name("init_data_cost_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8 ,1};
+                size_t globalThreads[] = {divUp(w, localThreads[0]) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int cdisp_step1 = msg_step * h;
+                openCLVerifyKernel(clCxt, kernel,  localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0, sizeof(cl_mem),  (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 1, sizeof(cl_mem),  (void *)&left.data));
+                openCLSafeCall(clSetKernelArg(kernel, 2, sizeof(cl_mem),  (void *)&right.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3, sizeof(cl_int),  (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 4, sizeof(cl_int),  (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 5, sizeof(cl_int),  (void *)&level));
+                openCLSafeCall(clSetKernelArg(kernel, 6, sizeof(cl_int),  (void *)&channels));
+                openCLSafeCall(clSetKernelArg(kernel, 7, sizeof(cl_int),  (void *)&msg_step));
+                openCLSafeCall(clSetKernelArg(kernel, 8, sizeof(cl_float), (void *)&rthis.data_weight));
+                openCLSafeCall(clSetKernelArg(kernel, 9, sizeof(cl_float), (void *)&rthis.max_data_term));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int), (void *)&cdisp_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_int), (void *)&rthis.min_disp_th));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_int), (void *)&left.step));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_int), (void *)&rthis.ndisp));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+
+            static void init_data_cost_reduce_caller(const oclMat &left, const oclMat &right, oclMat &temp,
+                StereoConstantSpaceBP &rthis,
+                int msg_step, int h, int w, int level)
+            {
+
+                Context  *clCxt = left.clCxt;
+                int data_type = rthis.msg_type;
+                int channels = left.oclchannels();
+                int win_size = (int)std::pow(2.f, level);
+
+                String kernelName = get_kernel_name("init_data_cost_reduce_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                const int threadsNum = 256;
+                //size_t blockSize = threadsNum;
+                size_t localThreads[3]  = {win_size, 1, threadsNum / win_size};
+                size_t globalThreads[3] = {w *localThreads[0],
+                    h * divUp(rthis.ndisp, localThreads[2]) *localThreads[1], 1 * localThreads[2]
+                };
+
+                int local_mem_size = threadsNum * sizeof(float);
+                int cdisp_step1 = msg_step * h;
+
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem),  (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem),  (void *)&left.data));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem),  (void *)&right.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  local_mem_size,  (void *)NULL));
+                openCLSafeCall(clSetKernelArg(kernel, 4,  sizeof(cl_int),  (void *)&level));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_int),  (void *)&left.rows));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_int),  (void *)&left.cols));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_int),  (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_int),  (void *)&win_size));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_int),  (void *)&channels));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int),  (void *)&rthis.ndisp));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_int),  (void *)&left.step));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_float), (void *)&rthis.data_weight));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_float), (void *)&rthis.max_data_term));
+                openCLSafeCall(clSetKernelArg(kernel, 14, sizeof(cl_int),  (void *)&rthis.min_disp_th));
+                openCLSafeCall(clSetKernelArg(kernel, 15, sizeof(cl_int),  (void *)&cdisp_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 16, sizeof(cl_int),  (void *)&msg_step));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 3, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+
+            static void get_first_initial_local_caller(uchar *data_cost_selected, uchar *disp_selected_pyr,
+                oclMat &temp, StereoConstantSpaceBP &rthis,
+                int h, int w, int nr_plane, int msg_step)
+            {
+                Context  *clCxt = temp.clCxt;
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("get_first_k_initial_local_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8 ,1};
+                size_t globalThreads[] = {divUp(w, localThreads[0]) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int disp_step = msg_step * h;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&data_cost_selected));
+                openCLSafeCall(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3, sizeof(cl_int), (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 4, sizeof(cl_int), (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 5, sizeof(cl_int), (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 6, sizeof(cl_int), (void *)&msg_step));
+                openCLSafeCall(clSetKernelArg(kernel, 7, sizeof(cl_int), (void *)&disp_step));
+                openCLSafeCall(clSetKernelArg(kernel, 8, sizeof(cl_int), (void *)&rthis.ndisp));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+            static void get_first_initial_global_caller(uchar *data_cost_selected, uchar *disp_selected_pyr,
+                oclMat &temp, StereoConstantSpaceBP &rthis,
+                int h, int w, int nr_plane, int msg_step)
+            {
+                Context  *clCxt = temp.clCxt;
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("get_first_k_initial_global_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8, 1};
+                size_t globalThreads[] = {divUp(w, localThreads[0]) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int disp_step = msg_step * h;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&data_cost_selected));
+                openCLSafeCall(clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3, sizeof(cl_int), (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 4, sizeof(cl_int), (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 5, sizeof(cl_int), (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 6, sizeof(cl_int), (void *)&msg_step));
+                openCLSafeCall(clSetKernelArg(kernel, 7, sizeof(cl_int), (void *)&disp_step));
+                openCLSafeCall(clSetKernelArg(kernel, 8, sizeof(cl_int), (void *)&rthis.ndisp));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+
+            static void init_data_cost(const oclMat &left, const oclMat &right, oclMat &temp, StereoConstantSpaceBP &rthis,
+                uchar *disp_selected_pyr, uchar *data_cost_selected,
+                size_t msg_step, int h, int w, int level, int nr_plane)
+            {
+
+                if(level <= 1)
+                    init_data_cost_caller(left, right, temp, rthis, msg_step, h, w, level);
+                else
+                    init_data_cost_reduce_caller(left, right, temp, rthis, msg_step, h, w, level);
+
+                if(rthis.use_local_init_data_cost == true)
+                {
+                    get_first_initial_local_caller(data_cost_selected, disp_selected_pyr, temp, rthis, h, w, nr_plane, msg_step);
+                }
+                else
+                {
+                    get_first_initial_global_caller(data_cost_selected, disp_selected_pyr, temp, rthis, h, w,
+                        nr_plane, msg_step);
+                }
+            }
+
+            ///////////////////////////////////////////////////////////////////////////////////////////////////
+            ///////////////////////////////////compute_data_cost//////////////////////////////////////////////
+            ////////////////////////////////////////////////////////////////////////////////////////////////
+            static void compute_data_cost_caller(uchar *disp_selected_pyr, uchar *data_cost,
+                StereoConstantSpaceBP &rthis, int msg_step1,
+                int msg_step2, const oclMat &left, const oclMat &right, int h,
+                int w, int h2, int level, int nr_plane)
+            {
+                Context  *clCxt = left.clCxt;
+                int channels = left.oclchannels();
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("compute_data_cost_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8, 1};
+                size_t globalThreads[] = {divUp(w, localThreads[0]) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int disp_step1 = msg_step1 * h;
+                int disp_step2 = msg_step2 * h2;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem),  (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem),  (void *)&data_cost));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem),  (void *)&left.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  sizeof(cl_mem),  (void *)&right.data));
+                openCLSafeCall(clSetKernelArg(kernel, 4,  sizeof(cl_int),  (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_int),  (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_int),  (void *)&level));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_int),  (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_int),  (void *)&channels));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_int),  (void *)&msg_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int),  (void *)&msg_step2));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_int),  (void *)&disp_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_int),  (void *)&disp_step2));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_float), (void *)&rthis.data_weight));
+                openCLSafeCall(clSetKernelArg(kernel, 14, sizeof(cl_float), (void *)&rthis.max_data_term));
+                openCLSafeCall(clSetKernelArg(kernel, 15, sizeof(cl_int),  (void *)&left.step));
+                openCLSafeCall(clSetKernelArg(kernel, 16, sizeof(cl_int),  (void *)&rthis.min_disp_th));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+            static void compute_data_cost_reduce_caller(uchar *disp_selected_pyr, uchar *data_cost,
+                StereoConstantSpaceBP &rthis, int msg_step1,
+                int msg_step2, const oclMat &left, const oclMat &right, int h,
+                int w, int h2, int level, int nr_plane)
+            {
+                Context  *clCxt = left.clCxt;
+                int data_type = rthis.msg_type;
+                int channels = left.oclchannels();
+                int win_size = (int)std::pow(2.f, level);
+
+                String kernelName = get_kernel_name("compute_data_cost_reduce_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                const size_t threadsNum = 256;
+                //size_t blockSize = threadsNum;
+                size_t localThreads[3]  = {win_size, 1, threadsNum / win_size};
+                size_t globalThreads[3] = {w *localThreads[0],
+                    h * divUp(nr_plane, localThreads[2]) *localThreads[1], 1 * localThreads[2]
+                };
+
+                int disp_step1 = msg_step1 * h;
+                int disp_step2 = msg_step2 * h2;
+                size_t local_mem_size = threadsNum * sizeof(float);
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem),  (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem),  (void *)&data_cost));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem),  (void *)&left.data));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  sizeof(cl_mem),  (void *)&right.data));
+                openCLSafeCall(clSetKernelArg(kernel, 4, local_mem_size,   (void *)NULL));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_int),  (void *)&level));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_int),  (void *)&left.rows));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_int),  (void *)&left.cols));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_int),  (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_int),  (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int),  (void *)&channels));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_int),  (void *)&win_size));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_int),  (void *)&msg_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_int),  (void *)&msg_step2));
+                openCLSafeCall(clSetKernelArg(kernel, 14, sizeof(cl_int),  (void *)&disp_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 15, sizeof(cl_int),  (void *)&disp_step2));
+                openCLSafeCall(clSetKernelArg(kernel, 16, sizeof(cl_float), (void *)&rthis.data_weight));
+                openCLSafeCall(clSetKernelArg(kernel, 17, sizeof(cl_float), (void *)&rthis.max_data_term));
+                openCLSafeCall(clSetKernelArg(kernel, 18, sizeof(cl_int),  (void *)&left.step));
+                openCLSafeCall(clSetKernelArg(kernel, 19, sizeof(cl_int),  (void *)&rthis.min_disp_th));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 3, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+            static void compute_data_cost(uchar *disp_selected_pyr, uchar *data_cost, StereoConstantSpaceBP &rthis,
+                int msg_step1, int msg_step2, const oclMat &left, const oclMat &right, int h, int w,
+                int h2, int level, int nr_plane)
+            {
+                if(level <= 1)
+                    compute_data_cost_caller(disp_selected_pyr, data_cost, rthis, msg_step1, msg_step2,
+                    left, right, h, w, h2, level, nr_plane);
+                else
+                    compute_data_cost_reduce_caller(disp_selected_pyr, data_cost, rthis,  msg_step1, msg_step2,
+                    left, right, h, w, h2, level, nr_plane);
+            }
+            ////////////////////////////////////////////////////////////////////////////////////////////////
+            //////////////////////////////////////init message//////////////////////////////////////////////
+            ////////////////////////////////////////////////////////////////////////////////////////////////
+            static void init_message(uchar *u_new, uchar *d_new, uchar *l_new, uchar *r_new,
+                uchar *u_cur, uchar *d_cur, uchar *l_cur, uchar *r_cur,
+                uchar *disp_selected_pyr_new, uchar *disp_selected_pyr_cur,
+                uchar *data_cost_selected, uchar *data_cost, oclMat &temp, StereoConstantSpaceBP rthis,
+                size_t msg_step1, size_t msg_step2, int h, int w, int nr_plane,
+                int h2, int w2, int nr_plane2)
+            {
+                Context  *clCxt = temp.clCxt;
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("init_message_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8, 1};
+                size_t globalThreads[] = {divUp(w, localThreads[0]) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int disp_step1 = msg_step1 * h;
+                int disp_step2 = msg_step2 * h2;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem), (void *)&u_new));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem), (void *)&d_new));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem), (void *)&l_new));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  sizeof(cl_mem), (void *)&r_new));
+                openCLSafeCall(clSetKernelArg(kernel, 4,  sizeof(cl_mem), (void *)&u_cur));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_mem), (void *)&d_cur));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_mem), (void *)&l_cur));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_mem), (void *)&r_cur));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_mem), (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_mem), (void *)&disp_selected_pyr_new));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_mem), (void *)&disp_selected_pyr_cur));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_mem), (void *)&data_cost_selected));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_mem), (void *)&data_cost));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_int), (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 14, sizeof(cl_int), (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 15, sizeof(cl_int), (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 16, sizeof(cl_int), (void *)&h2));
+                openCLSafeCall(clSetKernelArg(kernel, 17, sizeof(cl_int), (void *)&w2));
+                openCLSafeCall(clSetKernelArg(kernel, 18, sizeof(cl_int), (void *)&nr_plane2));
+                openCLSafeCall(clSetKernelArg(kernel, 19, sizeof(cl_int), (void *)&disp_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 20, sizeof(cl_int), (void *)&disp_step2));
+                openCLSafeCall(clSetKernelArg(kernel, 21, sizeof(cl_int), (void *)&msg_step1));
+                openCLSafeCall(clSetKernelArg(kernel, 22, sizeof(cl_int), (void *)&msg_step2));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+            ////////////////////////////////////////////////////////////////////////////////////////////////
+            ///////////////////////////calc_all_iterations////////////////////////////////////////////////
+            //////////////////////////////////////////////////////////////////////////////////////////////
+            static void calc_all_iterations_caller(uchar *u, uchar *d, uchar *l, uchar *r, uchar *data_cost_selected,
+                uchar *disp_selected_pyr, oclMat &temp, StereoConstantSpaceBP rthis,
+                int msg_step, int h, int w, int nr_plane, int i)
+            {
+                Context  *clCxt = temp.clCxt;
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("compute_message_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+                size_t localThreads[]  = {32, 8, 1};
+                size_t globalThreads[] = {divUp(w, (localThreads[0]) << 1) *localThreads[0],
+                    divUp(h, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int disp_step = msg_step * h;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem),  (void *)&u));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem),  (void *)&d));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem),  (void *)&l));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  sizeof(cl_mem),  (void *)&r));
+                openCLSafeCall(clSetKernelArg(kernel, 4,  sizeof(cl_mem),  (void *)&data_cost_selected));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_mem),  (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_mem),  (void *)&temp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_int),  (void *)&h));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_int),  (void *)&w));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_int),  (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int),  (void *)&i));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_float), (void *)&rthis.max_disc_term));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_int),  (void *)&disp_step));
+                openCLSafeCall(clSetKernelArg(kernel, 13, sizeof(cl_int),  (void *)&msg_step));
+                openCLSafeCall(clSetKernelArg(kernel, 14, sizeof(cl_float), (void *)&rthis.disc_single_jump));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+            static void calc_all_iterations(uchar *u, uchar *d, uchar *l, uchar *r, uchar *data_cost_selected,
+                uchar *disp_selected_pyr, oclMat &temp, StereoConstantSpaceBP rthis,
+                int msg_step, int h, int w, int nr_plane)
+            {
+                for(int t = 0; t < rthis.iters; t++)
+                    calc_all_iterations_caller(u, d, l, r, data_cost_selected, disp_selected_pyr, temp, rthis,
+                    msg_step, h, w, nr_plane, t & 1);
+            }
+
+            ///////////////////////////////////////////////////////////////////////////////////////////////
+            //////////////////////////compute_disp////////////////////////////////////////////////////////
+            /////////////////////////////////////////////////////////////////////////////////////////////
+            static void compute_disp(uchar *u, uchar *d, uchar *l, uchar *r, uchar *data_cost_selected,
+                uchar *disp_selected_pyr, StereoConstantSpaceBP &rthis, size_t msg_step,
+                oclMat &disp, int nr_plane)
+            {
+                Context  *clCxt = disp.clCxt;
+                int data_type = rthis.msg_type;
+
+                String kernelName = get_kernel_name("compute_disp_", data_type);
+
+                cl_kernel kernel = openCLGetKernelFromSource(clCxt, &stereocsbp, kernelName);
+
+                //size_t blockSize = 256;
+                size_t localThreads[]  = {32, 8, 1};
+                size_t globalThreads[] = {divUp(disp.cols, localThreads[0]) *localThreads[0],
+                    divUp(disp.rows, localThreads[1]) *localThreads[1],
+                    1
+                };
+
+                int step_size = disp.step / disp.elemSize();
+                int disp_step = disp.rows * msg_step;
+                openCLVerifyKernel(clCxt, kernel, localThreads);
+                openCLSafeCall(clSetKernelArg(kernel, 0,  sizeof(cl_mem), (void *)&u));
+                openCLSafeCall(clSetKernelArg(kernel, 1,  sizeof(cl_mem), (void *)&d));
+                openCLSafeCall(clSetKernelArg(kernel, 2,  sizeof(cl_mem), (void *)&l));
+                openCLSafeCall(clSetKernelArg(kernel, 3,  sizeof(cl_mem), (void *)&r));
+                openCLSafeCall(clSetKernelArg(kernel, 4,  sizeof(cl_mem), (void *)&data_cost_selected));
+                openCLSafeCall(clSetKernelArg(kernel, 5,  sizeof(cl_mem), (void *)&disp_selected_pyr));
+                openCLSafeCall(clSetKernelArg(kernel, 6,  sizeof(cl_mem), (void *)&disp.data));
+                openCLSafeCall(clSetKernelArg(kernel, 7,  sizeof(cl_int), (void *)&step_size));
+                openCLSafeCall(clSetKernelArg(kernel, 8,  sizeof(cl_int), (void *)&disp.cols));
+                openCLSafeCall(clSetKernelArg(kernel, 9,  sizeof(cl_int), (void *)&disp.rows));
+                openCLSafeCall(clSetKernelArg(kernel, 10, sizeof(cl_int), (void *)&nr_plane));
+                openCLSafeCall(clSetKernelArg(kernel, 11, sizeof(cl_int), (void *)&msg_step));
+                openCLSafeCall(clSetKernelArg(kernel, 12, sizeof(cl_int), (void *)&disp_step));
+                openCLSafeCall(clEnqueueNDRangeKernel(*(cl_command_queue*)getoclCommandQueue(), kernel, 2, NULL,
+                    globalThreads, localThreads, 0, NULL, NULL));
+
+                clFinish(*(cl_command_queue*)getoclCommandQueue());
+                openCLSafeCall(clReleaseKernel(kernel));
+            }
+        }
+    }
+}
+namespace
+{
+    const float DEFAULT_MAX_DATA_TERM = 30.0f;
+    const float DEFAULT_DATA_WEIGHT = 1.0f;
+    const float DEFAULT_MAX_DISC_TERM = 160.0f;
+    const float DEFAULT_DISC_SINGLE_JUMP = 10.0f;
+}
+
+void cv::ocl::StereoConstantSpaceBP::estimateRecommendedParams(int width, int height, int &ndisp, int &iters, int &levels, int &nr_plane)
+{
+    ndisp = (int) ((float) width / 3.14f);
+    if ((ndisp & 1) != 0)
+        ndisp++;
+
+    int mm = ::max(width, height);
+    iters = mm / 100 + ((mm > 1200) ? - 4 : 4);
+
+    levels = (int)::log(static_cast<double>(mm)) * 2 / 3;
+    if (levels == 0) levels++;
+
+    nr_plane = (int) ((float) ndisp / std::pow(2.0, levels + 1));
+}
+
+cv::ocl::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp_, int iters_, int levels_, int nr_plane_,
+    int msg_type_)
+
+    : ndisp(ndisp_), iters(iters_), levels(levels_), nr_plane(nr_plane_),
+    max_data_term(DEFAULT_MAX_DATA_TERM), data_weight(DEFAULT_DATA_WEIGHT),
+    max_disc_term(DEFAULT_MAX_DISC_TERM), disc_single_jump(DEFAULT_DISC_SINGLE_JUMP), min_disp_th(0),
+    msg_type(msg_type_), use_local_init_data_cost(true)
+{
+    CV_Assert(msg_type_ == CV_32F || msg_type_ == CV_16S);
+}
+
+
+cv::ocl::StereoConstantSpaceBP::StereoConstantSpaceBP(int ndisp_, int iters_, int levels_, int nr_plane_,
+    float max_data_term_, float data_weight_, float max_disc_term_, float disc_single_jump_,
+    int min_disp_th_, int msg_type_)
+    : ndisp(ndisp_), iters(iters_), levels(levels_), nr_plane(nr_plane_),
+    max_data_term(max_data_term_), data_weight(data_weight_),
+    max_disc_term(max_disc_term_), disc_single_jump(disc_single_jump_), min_disp_th(min_disp_th_),
+    msg_type(msg_type_), use_local_init_data_cost(true)
+{
+    CV_Assert(msg_type_ == CV_32F || msg_type_ == CV_16S);
+}
+
+template<class T>
+static void csbp_operator(StereoConstantSpaceBP &rthis, oclMat u[2], oclMat d[2], oclMat l[2], oclMat r[2],
+    oclMat disp_selected_pyr[2], oclMat &data_cost, oclMat &data_cost_selected,
+    oclMat &temp, oclMat &out, const oclMat &left, const oclMat &right, oclMat &disp)
+{
+    CV_DbgAssert(0 < rthis.ndisp && 0 < rthis.iters && 0 < rthis.levels && 0 < rthis.nr_plane
+        && left.rows == right.rows && left.cols == right.cols && left.type() == right.type());
+
+    CV_Assert(rthis.levels <= 8 && (left.type() == CV_8UC1 || left.type() == CV_8UC3));
+
+    const Scalar zero = Scalar::all(0);
+
+    ////////////////////////////////////Init///////////////////////////////////////////////////
+    int rows = left.rows;
+    int cols = left.cols;
+
+    rthis.levels = min(rthis.levels, int(log((double)rthis.ndisp) / log(2.0)));
+    int levels = rthis.levels;
+
+    AutoBuffer<int> buf(levels * 4);
+
+    int *cols_pyr = buf;
+    int *rows_pyr = cols_pyr + levels;
+    int *nr_plane_pyr = rows_pyr + levels;
+    int *step_pyr = nr_plane_pyr + levels;
+
+    cols_pyr[0] = cols;
+    rows_pyr[0] = rows;
+    nr_plane_pyr[0] = rthis.nr_plane;
+
+    const int n = 64;
+    step_pyr[0] = alignSize(cols * sizeof(T), n) / sizeof(T);
+    for (int i = 1; i < levels; i++)
+    {
+        cols_pyr[i] = cols_pyr[i - 1]  / 2;
+        rows_pyr[i] = rows_pyr[i - 1]/ 2;
+
+        nr_plane_pyr[i] = nr_plane_pyr[i - 1] * 2;
+
+        step_pyr[i] = alignSize(cols_pyr[i] * sizeof(T), n) / sizeof(T);
+    }
+
+    Size msg_size(step_pyr[0], rows * nr_plane_pyr[0]);
+    Size data_cost_size(step_pyr[0], rows * nr_plane_pyr[0] * 2);
+
+    u[0].create(msg_size, DataType<T>::type);
+    d[0].create(msg_size, DataType<T>::type);
+    l[0].create(msg_size, DataType<T>::type);
+    r[0].create(msg_size, DataType<T>::type);
+
+    u[1].create(msg_size, DataType<T>::type);
+    d[1].create(msg_size, DataType<T>::type);
+    l[1].create(msg_size, DataType<T>::type);
+    r[1].create(msg_size, DataType<T>::type);
+
+    disp_selected_pyr[0].create(msg_size, DataType<T>::type);
+    disp_selected_pyr[1].create(msg_size, DataType<T>::type);
+
+    data_cost.create(data_cost_size, DataType<T>::type);
+    data_cost_selected.create(msg_size, DataType<T>::type);
+
+    Size temp_size = data_cost_size;
+    if (data_cost_size.width * data_cost_size.height < step_pyr[0] * rows_pyr[levels - 1] * rthis.ndisp)
+        temp_size = Size(step_pyr[0], rows_pyr[levels - 1] * rthis.ndisp);
+
+    temp.create(temp_size, DataType<T>::type);
+    temp = zero;
+
+    ///////////////////////////////// Compute////////////////////////////////////////////////
+
+    //csbp::load_constants(rthis.ndisp, rthis.max_data_term, rthis.data_weight,
+    //   rthis.max_disc_term, rthis.disc_single_jump, rthis.min_disp_th, left, right, temp);
+
+    l[0] = zero;
+    d[0] = zero;
+    r[0] = zero;
+    u[0] = zero;
+    disp_selected_pyr[0] = zero;
+
+    l[1] = zero;
+    d[1] = zero;
+    r[1] = zero;
+    u[1] = zero;
+    disp_selected_pyr[1] = zero;
+
+    data_cost = zero;
+
+    data_cost_selected = zero;
+
+    int cur_idx = 0;
+
+    for (int i = levels - 1; i >= 0; i--)
+    {
+        if (i == levels - 1)
+        {
+            cv::ocl::stereoCSBP::init_data_cost(left, right, temp, rthis, disp_selected_pyr[cur_idx].data,
+                data_cost_selected.data, step_pyr[0], rows_pyr[i], cols_pyr[i],
+                i, nr_plane_pyr[i]);
+        }
+        else
+        {
+            cv::ocl::stereoCSBP::compute_data_cost(
+                disp_selected_pyr[cur_idx].data, data_cost.data, rthis, step_pyr[0],
+                step_pyr[0], left, right, rows_pyr[i], cols_pyr[i], rows_pyr[i + 1], i,
+                nr_plane_pyr[i + 1]);
+
+            int new_idx = (cur_idx + 1) & 1;
+
+            cv::ocl::stereoCSBP::init_message(u[new_idx].data, d[new_idx].data, l[new_idx].data, r[new_idx].data,
+                u[cur_idx].data, d[cur_idx].data, l[cur_idx].data, r[cur_idx].data,
+                disp_selected_pyr[new_idx].data, disp_selected_pyr[cur_idx].data,
+                data_cost_selected.data, data_cost.data, temp, rthis, step_pyr[0],
+                step_pyr[0], rows_pyr[i], cols_pyr[i], nr_plane_pyr[i], rows_pyr[i + 1],
+                cols_pyr[i + 1], nr_plane_pyr[i + 1]);
+            cur_idx = new_idx;
+        }
+        cv::ocl::stereoCSBP::calc_all_iterations(u[cur_idx].data, d[cur_idx].data, l[cur_idx].data, r[cur_idx].data,
+            data_cost_selected.data, disp_selected_pyr[cur_idx].data, temp,
+            rthis, step_pyr[0], rows_pyr[i], cols_pyr[i], nr_plane_pyr[i]);
+    }
+
+    if (disp.empty())
+        disp.create(rows, cols, CV_16S);
+
+    out = ((disp.type() == CV_16S) ? disp : (out.create(rows, cols, CV_16S), out));
+    out = zero;
+
+    stereoCSBP::compute_disp(u[cur_idx].data, d[cur_idx].data, l[cur_idx].data, r[cur_idx].data,
+        data_cost_selected.data, disp_selected_pyr[cur_idx].data, rthis, step_pyr[0],
+        out, nr_plane_pyr[0]);
+    if (disp.type() != CV_16S)
+        out.convertTo(disp, disp.type());
+}
+
+
+typedef void (*csbp_operator_t)(StereoConstantSpaceBP &rthis, oclMat u[2], oclMat d[2], oclMat l[2], oclMat r[2],
+    oclMat disp_selected_pyr[2], oclMat &data_cost, oclMat &data_cost_selected,
+    oclMat &temp, oclMat &out, const oclMat &left, const oclMat &right, oclMat &disp);
+
+const static csbp_operator_t operators[] = {0, 0, 0, csbp_operator<short>, 0, csbp_operator<float>, 0, 0};
+
+void cv::ocl::StereoConstantSpaceBP::operator()(const oclMat &left, const oclMat &right, oclMat &disp)
+{
+
+    CV_Assert(msg_type == CV_32F || msg_type == CV_16S);
+    operators[msg_type](*this, u, d, l, r, disp_selected_pyr, data_cost, data_cost_selected, temp, out,
+        left, right, disp);
+}
+
+#endif /* !defined (HAVE_OPENCL) */
diff --git a/modules/ocl/src/stereobm.cpp b/modules/ocl/src/stereobm.cpp
new file mode 100644
index 0000000000..a5cbe2b9f1
--- /dev/null
+++ b/modules/ocl/src/stereobm.cpp
@@ -0,0 +1,238 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Xiaopeng Fu, xiaopeng@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include <vector>
+
+using namespace cv;
+using namespace cv::ocl;
+
+namespace cv
+{
+namespace ocl
+{
+
+///////////////////////////OpenCL kernel strings///////////////////////////
+extern const char *stereobm;
+
+}
+}
+namespace cv
+{
+namespace ocl
+{
+namespace stereoBM
+{
+/////////////////////////////////////////////////////////////////////////
+//////////////////////////prefilter_xsbel////////////////////////////////
+////////////////////////////////////////////////////////////////////////
+static void prefilter_xsobel(const oclMat &input, oclMat &output, int prefilterCap)
+{
+    String kernelName = "prefilter_xsobel";
+
+    size_t blockSize = 1;
+    size_t globalThreads[3] = { input.cols, input.rows, 1 };
+    size_t localThreads[3]  = { blockSize, blockSize, 1 };
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&input.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&output.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&input.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&input.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&prefilterCap));
+
+    openCLExecuteKernel(Context::getContext(), &stereobm, kernelName,
+        globalThreads, localThreads, args, -1, -1);
+}
+//////////////////////////////////////////////////////////////////////////
+//////////////////////////////common////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
+#define N_DISPARITIES 8
+#define ROWSperTHREAD 21
+#define BLOCK_W 128
+static inline int divUp(int total, int grain)
+{
+    return (total + grain - 1) / grain;
+}
+////////////////////////////////////////////////////////////////////////////
+///////////////////////////////stereoBM_GPU////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+static void stereo_bm(const oclMat &left, const oclMat &right,  oclMat &disp,
+               int maxdisp, int winSize,  oclMat &minSSD_buf)
+{
+    int winsz2 = winSize >> 1;
+
+    String kernelName = "stereoKernel";
+
+    disp.setTo(Scalar_<unsigned char>::all(0));
+    minSSD_buf.setTo(Scalar_<unsigned int>::all(0xFFFFFFFF));
+
+    size_t minssd_step = minSSD_buf.step / minSSD_buf.elemSize();
+    size_t local_mem_size = (N_DISPARITIES * (BLOCK_W + 2 * winsz2)) *
+                            sizeof(cl_uint);
+    //size_t blockSize = 1;
+    size_t localThreads[]  = { BLOCK_W, 1,1};
+    size_t globalThreads[] = { divUp(left.cols - maxdisp - 2 * winsz2, BLOCK_W) *BLOCK_W,
+                               divUp(left.rows - 2 * winsz2, ROWSperTHREAD),
+                               1
+                             };
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&left.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&right.data));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&minSSD_buf.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&minssd_step));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&disp.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&disp.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&left.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&left.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&left.step));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&maxdisp));
+    args.push_back(std::make_pair(local_mem_size, (void *)NULL));
+
+    char opt [128];
+    sprintf(opt, "-D radius=%d", winsz2);
+    openCLExecuteKernel(Context::getContext(), &stereobm, kernelName,
+        globalThreads, localThreads, args, -1, -1, opt);
+}
+////////////////////////////////////////////////////////////////////////////
+///////////////////////////////postfilter_textureness///////////////////////
+////////////////////////////////////////////////////////////////////////////
+static void postfilter_textureness(oclMat &left, int winSize,
+                            float avergeTexThreshold, oclMat &disparity)
+{
+    String kernelName = "textureness_kernel";
+
+    size_t blockSize = 1;
+    size_t localThreads[]  = { BLOCK_W, blockSize ,1};
+    size_t globalThreads[] = { divUp(left.cols, BLOCK_W) *BLOCK_W,
+                               divUp(left.rows, 2 * ROWSperTHREAD),
+                               1
+                             };
+
+    size_t local_mem_size = (localThreads[0] + localThreads[0] + (winSize / 2) * 2) * sizeof(float);
+
+    std::vector< std::pair<size_t, const void *> > args;
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&disparity.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&disparity.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&disparity.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&disparity.step));
+    args.push_back(std::make_pair(sizeof(cl_mem), (void *)&left.data));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&left.rows));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&left.cols));
+    args.push_back(std::make_pair(sizeof(cl_int), (void *)&winSize));
+    args.push_back(std::make_pair(sizeof(cl_float), (void *)&avergeTexThreshold));
+    args.push_back(std::make_pair(local_mem_size, (void*)NULL));
+    openCLExecuteKernel(Context::getContext(), &stereobm, kernelName,
+        globalThreads, localThreads, args, -1, -1);
+}
+//////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////operator/////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////
+static void operator_(oclMat &minSSD, oclMat &leBuf, oclMat &riBuf, int preset, int ndisp,
+               int winSize, float avergeTexThreshold, const oclMat &left,
+               const oclMat &right, oclMat &disparity)
+
+{
+    CV_DbgAssert(left.rows == right.rows && left.cols == right.cols);
+    CV_DbgAssert(left.type() == CV_8UC1);
+    CV_DbgAssert(right.type() == CV_8UC1);
+
+    disparity.create(left.size(), CV_8UC1);
+    minSSD.create(left.size(), CV_32SC1);
+
+    oclMat le_for_bm =  left;
+    oclMat ri_for_bm = right;
+
+    if (preset == cv::ocl::StereoBM_OCL::PREFILTER_XSOBEL)
+    {
+        leBuf.create( left.size(),  left.type());
+        riBuf.create(right.size(), right.type());
+
+        prefilter_xsobel( left, leBuf, 31);
+        prefilter_xsobel(right, riBuf, 31);
+
+        le_for_bm = leBuf;
+        ri_for_bm = riBuf;
+    }
+
+    stereo_bm(le_for_bm, ri_for_bm, disparity, ndisp, winSize, minSSD);
+
+    if (avergeTexThreshold)
+    {
+        postfilter_textureness(le_for_bm, winSize, avergeTexThreshold, disparity);
+    }
+}
+}
+}
+}
+const float defaultAvgTexThreshold = 3;
+
+cv::ocl::StereoBM_OCL::StereoBM_OCL()
+    : preset(BASIC_PRESET), ndisp(DEFAULT_NDISP), winSize(DEFAULT_WINSZ),
+      avergeTexThreshold(defaultAvgTexThreshold)  {}
+
+cv::ocl::StereoBM_OCL::StereoBM_OCL(int preset_, int ndisparities_, int winSize_)
+    : preset(preset_), ndisp(ndisparities_), winSize(winSize_),
+      avergeTexThreshold(defaultAvgTexThreshold)
+{
+    const int max_supported_ndisp = 1 << (sizeof(unsigned char) * 8);
+    CV_Assert(0 < ndisp && ndisp <= max_supported_ndisp);
+    CV_Assert(ndisp % 8 == 0);
+    CV_Assert(winSize % 2 == 1);
+}
+
+bool cv::ocl::StereoBM_OCL::checkIfGpuCallReasonable()
+{
+    return true;
+}
+
+void cv::ocl::StereoBM_OCL::operator() ( const oclMat &left, const oclMat &right,
+        oclMat &disparity)
+{
+    cv::ocl::stereoBM::operator_(minSSD, leBuf, riBuf, preset, ndisp, winSize, avergeTexThreshold, left, right, disparity);
+}
diff --git a/modules/ocl/src/stereobp.cpp b/modules/ocl/src/stereobp.cpp
new file mode 100644
index 0000000000..fa534893cc
--- /dev/null
+++ b/modules/ocl/src/stereobp.cpp
@@ -0,0 +1,517 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Peng Xiao,   pengxiao@outlook.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include <vector>
+#include <cstdio>
+
+using namespace cv;
+using namespace cv::ocl;
+
+////////////////////////////////////////////////////////////////////////
+///////////////// stereoBP /////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
+
+namespace cv
+{
+    namespace ocl
+    {
+
+        ///////////////////////////OpenCL kernel strings///////////////////////////
+        extern const char *stereobp;
+    }
+
+}
+namespace cv
+{
+    namespace ocl
+    {
+        namespace stereoBP
+        {
+            //////////////////////////////////////////////////////////////////////////
+            //////////////////////////////common////////////////////////////////////
+            ////////////////////////////////////////////////////////////////////////
+            typedef struct
+            {
+                int   cndisp;
+                float cmax_data_term;
+                float cdata_weight;
+                float cmax_disc_term;
+                float cdisc_single_jump;
+            } con_struct_t;
+
+            cl_mem cl_con_struct =  NULL;
+            static void load_constants(int ndisp, float max_data_term, float data_weight,
+                                float max_disc_term, float disc_single_jump)
+            {
+                con_struct_t *con_struct = new con_struct_t;
+                con_struct -> cndisp            = ndisp;
+                con_struct -> cmax_data_term    = max_data_term;
+                con_struct -> cdata_weight      = data_weight;
+                con_struct -> cmax_disc_term    = max_disc_term;
+                con_struct -> cdisc_single_jump = disc_single_jump;
+
+                cl_con_struct = load_constant(*((cl_context*)getoclContext()), *((cl_command_queue*)getoclCommandQueue()), (void *)con_struct,
+                                              sizeof(con_struct_t));
+
+                delete con_struct;
+            }
+            static void release_constants()
+            {
+                openCLFree(cl_con_struct);
+            }
+            static inline int divUp(int total, int grain)
+            {
+                return (total + grain - 1) / grain;
+            }
+            /////////////////////////////////////////////////////////////////////////////
+            ///////////////////////////comp data////////////////////////////////////////
+            /////////////////////////////////////////////////////////////////////////
+            static void  comp_data_call(const oclMat &left, const oclMat &right, oclMat &data, int /*disp*/,
+                float /*cmax_data_term*/, float /*cdata_weight*/)
+            {
+                Context  *clCxt = left.clCxt;
+                int channels = left.oclchannels();
+                int data_type = data.type();
+
+                String kernelName = "comp_data";
+
+                std::vector<std::pair<size_t , const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&left.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&left.rows));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&left.cols));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&left.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&right.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&right.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&data.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&data.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&cl_con_struct));
+
+                size_t gt[3] = {left.cols, left.rows, 1}, lt[3] = {16, 16, 1};
+
+                const int OPT_SIZE = 50;
+                char cn_opt [OPT_SIZE] = "";
+                sprintf( cn_opt, "%s -D CN=%d",
+                    (data_type == CV_16S ? "-D T_SHORT":"-D T_FLOAT"),
+                    channels
+                    );
+                openCLExecuteKernel(clCxt, &stereobp, kernelName, gt, lt, args, -1, -1, cn_opt);
+            }
+            ///////////////////////////////////////////////////////////////////////////////////
+            /////////////////////////data set down////////////////////////////////////////////
+            /////////////////////////////////////////////////////////////////////////////////
+            static void data_step_down_call(int dst_cols, int dst_rows, int src_rows,
+                const oclMat &src, oclMat &dst, int disp)
+            {
+                Context  *clCxt = src.clCxt;
+                int data_type = src.type();
+
+                String kernelName = "data_step_down";
+
+                std::vector<std::pair<size_t , const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_rows));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_rows));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_cols));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.step));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.step));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&disp));
+
+                size_t gt[3] = {dst_cols, dst_rows, 1}, lt[3] = {16, 16, 1};
+                const char* t_opt  = data_type == CV_16S ? "-D T_SHORT":"-D T_FLOAT";
+                openCLExecuteKernel(clCxt, &stereobp, kernelName, gt, lt, args, -1, -1, t_opt);
+            }
+            /////////////////////////////////////////////////////////////////////////////////
+            ///////////////////////////live up message////////////////////////////////////////
+            /////////////////////////////////////////////////////////////////////////////////
+            static void level_up_message_call(int dst_cols, int dst_rows, int src_rows,
+                oclMat &src, oclMat &dst, int ndisp)
+            {
+                Context  *clCxt = src.clCxt;
+                int data_type = src.type();
+
+                String kernelName = "level_up_message";
+                std::vector<std::pair<size_t , const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_rows));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_rows));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_cols));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.step));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&ndisp));
+
+                size_t gt[3] = {dst_cols, dst_rows, 1}, lt[3] = {16, 16, 1};
+                const char* t_opt  = data_type == CV_16S ? "-D T_SHORT":"-D T_FLOAT";
+                openCLExecuteKernel(clCxt, &stereobp, kernelName, gt, lt, args, -1, -1, t_opt);
+            }
+            static void level_up_messages_calls(int dst_idx, int dst_cols, int dst_rows, int src_rows,
+                                         oclMat *mus, oclMat *mds, oclMat *mls, oclMat *mrs,
+                                         int ndisp)
+            {
+                int src_idx = (dst_idx + 1) & 1;
+
+                level_up_message_call(dst_cols, dst_rows, src_rows,
+                                      mus[src_idx], mus[dst_idx], ndisp);
+
+                level_up_message_call(dst_cols, dst_rows, src_rows,
+                                      mds[src_idx], mds[dst_idx], ndisp);
+
+                level_up_message_call(dst_cols, dst_rows, src_rows,
+                                      mls[src_idx], mls[dst_idx], ndisp);
+
+                level_up_message_call(dst_cols, dst_rows, src_rows,
+                                      mrs[src_idx], mrs[dst_idx], ndisp);
+            }
+            //////////////////////////////////////////////////////////////////////////////////
+            //////////////////////////////cals_all_iterations_call///////////////////////////
+            /////////////////////////////////////////////////////////////////////////////////
+            static void calc_all_iterations_call(int cols, int rows, oclMat &u, oclMat &d,
+                oclMat &l, oclMat &r, oclMat &data,
+                int t, int cndisp, float cmax_disc_term,
+                float cdisc_single_jump)
+            {
+                Context  *clCxt = l.clCxt;
+                int data_type = u.type();
+
+                String kernelName = "one_iteration";
+
+                std::vector<std::pair<size_t , const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&u.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&u.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&data.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&data.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&d.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&l.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&r.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&t));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&cols));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&rows));
+                args.push_back( std::make_pair( sizeof(cl_float) , (void *)&cmax_disc_term));
+                args.push_back( std::make_pair( sizeof(cl_float) , (void *)&cdisc_single_jump));
+
+                size_t gt[3] = {cols, rows, 1}, lt[3] = {16, 16, 1};
+                char opt[80] = "";
+                sprintf(opt, "-D %s -D CNDISP=%d", data_type == CV_16S ? "T_SHORT":"T_FLOAT", cndisp);
+                openCLExecuteKernel(clCxt, &stereobp, kernelName, gt, lt, args, -1, -1, opt);
+            }
+
+            static void calc_all_iterations_calls(int cols, int rows, int iters, oclMat &u,
+                                           oclMat &d, oclMat &l, oclMat &r,
+                                           oclMat &data, int cndisp, float cmax_disc_term,
+                                           float cdisc_single_jump)
+            {
+                for(int t = 0; t < iters; ++t)
+                    calc_all_iterations_call(cols, rows, u, d, l, r, data, t, cndisp,
+                                             cmax_disc_term, cdisc_single_jump);
+            }
+            ///////////////////////////////////////////////////////////////////////////////
+            ///////////////////////output///////////////////////////////////////////////////
+            ////////////////////////////////////////////////////////////////////////////////
+            static void output_call(const oclMat &u, const oclMat &d, const oclMat l, const oclMat &r,
+                const oclMat &data, oclMat &disp, int ndisp)
+            {
+                Context  *clCxt = u.clCxt;
+                int data_type = u.type();
+
+                String kernelName = "output";
+
+                std::vector<std::pair<size_t , const void *> > args;
+
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&u.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&u.step));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&d.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&l.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&r.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&data.data));
+                args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&disp.data));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&disp.rows));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&disp.cols));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&disp.step));
+                args.push_back( std::make_pair( sizeof(cl_int) , (void *)&ndisp));
+
+                size_t gt[3] = {disp.cols, disp.rows, 1}, lt[3] = {16, 16, 1};
+                const char* t_opt  = data_type == CV_16S ? "-D T_SHORT":"-D T_FLOAT";
+                openCLExecuteKernel(clCxt, &stereobp, kernelName, gt, lt, args, -1, -1, t_opt);
+            }
+        }
+    }
+}
+namespace
+{
+    const float DEFAULT_MAX_DATA_TERM = 10.0f;
+    const float DEFAULT_DATA_WEIGHT = 0.07f;
+    const float DEFAULT_MAX_DISC_TERM = 1.7f;
+    const float DEFAULT_DISC_SINGLE_JUMP = 1.0f;
+}
+
+void cv::ocl::StereoBeliefPropagation::estimateRecommendedParams(int width, int height, int &ndisp, int &iters, int &levels)
+{
+    ndisp = width / 4;
+    if ((ndisp & 1) != 0)
+        ndisp++;
+
+    int mm = ::max(width, height);
+    iters = mm / 100 + 2;
+
+    levels = (int)(::log(static_cast<double>(mm)) + 1) * 4 / 5;
+    if (levels == 0) levels++;
+}
+
+cv::ocl::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp_, int iters_, int levels_, int msg_type_)
+    : ndisp(ndisp_), iters(iters_), levels(levels_),
+      max_data_term(DEFAULT_MAX_DATA_TERM), data_weight(DEFAULT_DATA_WEIGHT),
+      max_disc_term(DEFAULT_MAX_DISC_TERM), disc_single_jump(DEFAULT_DISC_SINGLE_JUMP),
+      msg_type(msg_type_), datas(levels_)
+{
+}
+
+cv::ocl::StereoBeliefPropagation::StereoBeliefPropagation(int ndisp_, int iters_, int levels_, float max_data_term_, float data_weight_, float max_disc_term_, float disc_single_jump_, int msg_type_)
+    : ndisp(ndisp_), iters(iters_), levels(levels_),
+      max_data_term(max_data_term_), data_weight(data_weight_),
+      max_disc_term(max_disc_term_), disc_single_jump(disc_single_jump_),
+      msg_type(msg_type_), datas(levels_)
+{
+}
+
+namespace
+{
+    class StereoBeliefPropagationImpl
+    {
+    public:
+        StereoBeliefPropagationImpl(StereoBeliefPropagation &rthis_,
+                                    oclMat &u_, oclMat &d_, oclMat &l_, oclMat &r_,
+                                    oclMat &u2_, oclMat &d2_, oclMat &l2_, oclMat &r2_,
+                                    std::vector<oclMat> &datas_, oclMat &out_)
+            : rthis(rthis_), u(u_), d(d_), l(l_), r(r_), u2(u2_), d2(d2_), l2(l2_), r2(r2_), datas(datas_), out(out_),
+              zero(Scalar::all(0)), scale(rthis_.msg_type == CV_32F ? 1.0f : 10.0f)
+        {
+            CV_Assert(0 < rthis.ndisp && 0 < rthis.iters && 0 < rthis.levels);
+            CV_Assert(rthis.msg_type == CV_32F || rthis.msg_type == CV_16S);
+            CV_Assert(rthis.msg_type == CV_32F || (1 << (rthis.levels - 1)) * scale * rthis.max_data_term < std::numeric_limits<short>::max());
+        }
+
+        void operator()(const oclMat &left, const oclMat &right, oclMat &disp)
+        {
+            CV_Assert(left.size() == right.size() && left.type() == right.type());
+            CV_Assert(left.type() == CV_8UC1 || left.type() == CV_8UC3 || left.type() == CV_8UC4);
+
+            rows = left.rows;
+            cols = left.cols;
+
+            int divisor = (int)pow(2.f, rthis.levels - 1.0f);
+            int lowest_cols = cols / divisor;
+            int lowest_rows = rows / divisor;
+            const int min_image_dim_size = 2;
+            CV_Assert(min(lowest_cols, lowest_rows) > min_image_dim_size);
+
+            init();
+
+            datas[0].create(rows * rthis.ndisp, cols, rthis.msg_type);
+            datas[0].setTo(Scalar_<short>::all(0));
+
+            cv::ocl::stereoBP::comp_data_call(left, right, datas[0], rthis.ndisp, rthis.max_data_term, scale * rthis.data_weight);
+            calcBP(disp);
+        }
+
+        void operator()(const oclMat &data, oclMat &disp)
+        {
+            CV_Assert((data.type() == rthis.msg_type) && (data.rows % rthis.ndisp == 0));
+
+            rows = data.rows / rthis.ndisp;
+            cols = data.cols;
+
+            int divisor = (int)pow(2.f, rthis.levels - 1.0f);
+            int lowest_cols = cols / divisor;
+            int lowest_rows = rows / divisor;
+            const int min_image_dim_size = 2;
+            CV_Assert(min(lowest_cols, lowest_rows) > min_image_dim_size);
+
+            init();
+
+            datas[0] = data;
+
+            calcBP(disp);
+        }
+    private:
+        void init()
+        {
+            u.create(rows * rthis.ndisp, cols, rthis.msg_type);
+            d.create(rows * rthis.ndisp, cols, rthis.msg_type);
+            l.create(rows * rthis.ndisp, cols, rthis.msg_type);
+            r.create(rows * rthis.ndisp, cols, rthis.msg_type);
+
+            if (rthis.levels & 1)
+            {
+                //can clear less area
+                u = zero;
+                d = zero;
+                l = zero;
+                r = zero;
+            }
+
+            if (rthis.levels > 1)
+            {
+                int less_rows = (rows + 1) / 2;
+                int less_cols = (cols + 1) / 2;
+
+                u2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
+                d2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
+                l2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
+                r2.create(less_rows * rthis.ndisp, less_cols, rthis.msg_type);
+
+                if ((rthis.levels & 1) == 0)
+                {
+                    u2 = zero;
+                    d2 = zero;
+                    l2 = zero;
+                    r2 = zero;
+                }
+            }
+
+            cv::ocl::stereoBP::load_constants(rthis.ndisp, rthis.max_data_term, scale * rthis.data_weight,
+                                              scale * rthis.max_disc_term, scale * rthis.disc_single_jump);
+
+            datas.resize(rthis.levels);
+            cols_all.resize(rthis.levels);
+            rows_all.resize(rthis.levels);
+
+            cols_all[0] = cols;
+            rows_all[0] = rows;
+        }
+
+        void calcBP(oclMat &disp)
+        {
+            using namespace cv::ocl::stereoBP;
+
+            for (int i = 1; i < rthis.levels; ++i)
+            {
+                cols_all[i] = (cols_all[i - 1] + 1) / 2;
+                rows_all[i] = (rows_all[i - 1] + 1) / 2;
+
+                datas[i].create(rows_all[i] * rthis.ndisp, cols_all[i], rthis.msg_type);
+                datas[i].setTo(Scalar_<short>::all(0));
+
+                data_step_down_call(cols_all[i], rows_all[i], rows_all[i - 1],
+                                    datas[i - 1], datas[i], rthis.ndisp);
+            }
+
+            oclMat mus[] = {u, u2};
+            oclMat mds[] = {d, d2};
+            oclMat mrs[] = {r, r2};
+            oclMat mls[] = {l, l2};
+
+            int mem_idx = (rthis.levels & 1) ? 0 : 1;
+
+            for (int i = rthis.levels - 1; i >= 0; --i)
+            {
+                // for lower level we have already computed messages by setting to zero
+                if (i != rthis.levels - 1)
+                    level_up_messages_calls(mem_idx, cols_all[i], rows_all[i], rows_all[i + 1],
+                                            mus, mds, mls, mrs, rthis.ndisp);
+
+                calc_all_iterations_calls(cols_all[i], rows_all[i], rthis.iters, mus[mem_idx],
+                                          mds[mem_idx], mls[mem_idx], mrs[mem_idx], datas[i],
+                                          rthis.ndisp, scale * rthis.max_disc_term,
+                                          scale * rthis.disc_single_jump);
+
+                mem_idx = (mem_idx + 1) & 1;
+            }
+            if (disp.empty())
+                disp.create(rows, cols, CV_16S);
+
+            out = ((disp.type() == CV_16S) ? disp : (out.create(rows, cols, CV_16S), out));
+            out = zero;
+
+            output_call(u, d, l, r, datas.front(), out, rthis.ndisp);
+
+            if (disp.type() != CV_16S)
+                out.convertTo(disp, disp.type());
+
+            release_constants();
+        }
+        StereoBeliefPropagationImpl& operator=(const StereoBeliefPropagationImpl&);
+
+        StereoBeliefPropagation &rthis;
+
+        oclMat &u;
+        oclMat &d;
+        oclMat &l;
+        oclMat &r;
+
+        oclMat &u2;
+        oclMat &d2;
+        oclMat &l2;
+        oclMat &r2;
+
+        std::vector<oclMat> &datas;
+        oclMat &out;
+
+        const Scalar zero;
+        const float scale;
+
+        int rows, cols;
+
+        std::vector<int> cols_all, rows_all;
+    };
+}
+
+void cv::ocl::StereoBeliefPropagation::operator()(const oclMat &left, const oclMat &right, oclMat &disp)
+{
+    ::StereoBeliefPropagationImpl impl(*this, u, d, l, r, u2, d2, l2, r2, datas, out);
+    impl(left, right, disp);
+}
+
+void cv::ocl::StereoBeliefPropagation::operator()(const oclMat &data, oclMat &disp)
+{
+    ::StereoBeliefPropagationImpl impl(*this, u, d, l, r, u2, d2, l2, r2, datas, out);
+    impl(data, disp);
+}
diff --git a/modules/ocl/src/tvl1flow.cpp b/modules/ocl/src/tvl1flow.cpp
new file mode 100644
index 0000000000..daf3a22956
--- /dev/null
+++ b/modules/ocl/src/tvl1flow.cpp
@@ -0,0 +1,491 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//		Jin Ma, jin@multicorewareinc.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#include "precomp.hpp"
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+namespace cv
+{
+    namespace ocl
+    {
+        ///////////////////////////OpenCL kernel strings///////////////////////////
+        extern const char* tvl1flow;
+    }
+}
+
+cv::ocl::OpticalFlowDual_TVL1_OCL::OpticalFlowDual_TVL1_OCL()
+{
+    tau            = 0.25;
+    lambda         = 0.15;
+    theta          = 0.3;
+    nscales        = 5;
+    warps          = 5;
+    epsilon        = 0.01;
+    iterations     = 300;
+    useInitialFlow = false;
+}
+
+void cv::ocl::OpticalFlowDual_TVL1_OCL::operator()(const oclMat& I0, const oclMat& I1, oclMat& flowx, oclMat& flowy)
+{
+    CV_Assert( I0.type() == CV_8UC1 || I0.type() == CV_32FC1 );
+    CV_Assert( I0.size() == I1.size() );
+    CV_Assert( I0.type() == I1.type() );
+    CV_Assert( !useInitialFlow || (flowx.size() == I0.size() && flowx.type() == CV_32FC1 && flowy.size() == flowx.size() && flowy.type() == flowx.type()) );
+    CV_Assert( nscales > 0 );
+
+    // allocate memory for the pyramid structure
+    I0s.resize(nscales);
+    I1s.resize(nscales);
+    u1s.resize(nscales);
+    u2s.resize(nscales);
+    //I0s_step == I1s_step
+    I0.convertTo(I0s[0], CV_32F, I0.depth() == CV_8U ? 1.0 : 255.0);
+    I1.convertTo(I1s[0], CV_32F, I1.depth() == CV_8U ? 1.0 : 255.0);
+
+
+    if (!useInitialFlow)
+    {
+        flowx.create(I0.size(), CV_32FC1);
+        flowy.create(I0.size(), CV_32FC1);
+    }
+    //u1s_step != u2s_step
+    u1s[0] = flowx;
+    u2s[0] = flowy;
+
+    I1x_buf.create(I0.size(), CV_32FC1);
+    I1y_buf.create(I0.size(), CV_32FC1);
+
+    I1w_buf.create(I0.size(), CV_32FC1);
+    I1wx_buf.create(I0.size(), CV_32FC1);
+    I1wy_buf.create(I0.size(), CV_32FC1);
+
+    grad_buf.create(I0.size(), CV_32FC1);
+    rho_c_buf.create(I0.size(), CV_32FC1);
+
+    p11_buf.create(I0.size(), CV_32FC1);
+    p12_buf.create(I0.size(), CV_32FC1);
+    p21_buf.create(I0.size(), CV_32FC1);
+    p22_buf.create(I0.size(), CV_32FC1);
+
+    diff_buf.create(I0.size(), CV_32FC1);
+
+    // create the scales
+    for (int s = 1; s < nscales; ++s)
+    {
+        ocl::pyrDown(I0s[s - 1], I0s[s]);
+        ocl::pyrDown(I1s[s - 1], I1s[s]);
+
+        if (I0s[s].cols < 16 || I0s[s].rows < 16)
+        {
+            nscales = s;
+            break;
+        }
+
+        if (useInitialFlow)
+        {
+            ocl::pyrDown(u1s[s - 1], u1s[s]);
+            ocl::pyrDown(u2s[s - 1], u2s[s]);
+
+            //ocl::multiply(u1s[s], Scalar::all(0.5), u1s[s]);
+            multiply(0.5, u1s[s], u1s[s]);
+            //ocl::multiply(u2s[s], Scalar::all(0.5), u2s[s]);
+            multiply(0.5, u1s[s], u2s[s]);
+        }
+    }
+
+    // pyramidal structure for computing the optical flow
+    for (int s = nscales - 1; s >= 0; --s)
+    {
+        // compute the optical flow at the current scale
+        procOneScale(I0s[s], I1s[s], u1s[s], u2s[s]);
+
+        // if this was the last scale, finish now
+        if (s == 0)
+            break;
+
+        // otherwise, upsample the optical flow
+
+        // zoom the optical flow for the next finer scale
+        ocl::resize(u1s[s], u1s[s - 1], I0s[s - 1].size());
+        ocl::resize(u2s[s], u2s[s - 1], I0s[s - 1].size());
+
+        // scale the optical flow with the appropriate zoom factor
+        multiply(2, u1s[s - 1], u1s[s - 1]);
+        multiply(2, u2s[s - 1], u2s[s - 1]);
+
+    }
+
+}
+
+namespace ocl_tvl1flow
+{
+    void centeredGradient(const oclMat &src, oclMat &dx, oclMat &dy);
+
+    void warpBackward(const oclMat &I0, const oclMat &I1, oclMat &I1x, oclMat &I1y,
+        oclMat &u1, oclMat &u2, oclMat &I1w, oclMat &I1wx, oclMat &I1wy,
+        oclMat &grad, oclMat &rho);
+
+    void estimateU(oclMat &I1wx, oclMat &I1wy, oclMat &grad,
+        oclMat &rho_c, oclMat &p11, oclMat &p12,
+        oclMat &p21, oclMat &p22, oclMat &u1,
+        oclMat &u2, oclMat &error, float l_t, float theta, char calc_error);
+
+    void estimateDualVariables(oclMat &u1, oclMat &u2,
+        oclMat &p11, oclMat &p12, oclMat &p21, oclMat &p22, float taut);
+}
+
+void cv::ocl::OpticalFlowDual_TVL1_OCL::procOneScale(const oclMat &I0, const oclMat &I1, oclMat &u1, oclMat &u2)
+{
+    using namespace ocl_tvl1flow;
+
+    const double scaledEpsilon = epsilon * epsilon * I0.size().area();
+
+    CV_DbgAssert( I1.size() == I0.size() );
+    CV_DbgAssert( I1.type() == I0.type() );
+    CV_DbgAssert( u1.empty() || u1.size() == I0.size() );
+    CV_DbgAssert( u2.size() == u1.size() );
+
+    if (u1.empty())
+    {
+        u1.create(I0.size(), CV_32FC1);
+        u1.setTo(Scalar::all(0));
+
+        u2.create(I0.size(), CV_32FC1);
+        u2.setTo(Scalar::all(0));
+    }
+
+    oclMat I1x = I1x_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat I1y = I1y_buf(Rect(0, 0, I0.cols, I0.rows));
+
+    centeredGradient(I1, I1x, I1y);
+
+    oclMat I1w = I1w_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat I1wx = I1wx_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat I1wy = I1wy_buf(Rect(0, 0, I0.cols, I0.rows));
+
+    oclMat grad = grad_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat rho_c = rho_c_buf(Rect(0, 0, I0.cols, I0.rows));
+
+    oclMat p11 = p11_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat p12 = p12_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat p21 = p21_buf(Rect(0, 0, I0.cols, I0.rows));
+    oclMat p22 = p22_buf(Rect(0, 0, I0.cols, I0.rows));
+    p11.setTo(Scalar::all(0));
+    p12.setTo(Scalar::all(0));
+    p21.setTo(Scalar::all(0));
+    p22.setTo(Scalar::all(0));
+
+    oclMat diff = diff_buf(Rect(0, 0, I0.cols, I0.rows));
+
+    const float l_t = static_cast<float>(lambda * theta);
+    const float taut = static_cast<float>(tau / theta);
+
+    for (int warpings = 0; warpings < warps; ++warpings)
+    {
+        warpBackward(I0, I1, I1x, I1y, u1, u2, I1w, I1wx, I1wy, grad, rho_c);
+
+        double error = numeric_limits<double>::max();
+        double prev_error = 0;
+        for (int n = 0; error > scaledEpsilon && n < iterations; ++n)
+        {
+            // some tweaks to make sum operation less frequently
+            char calc_error = (n & 0x1) && (prev_error < scaledEpsilon);
+            estimateU(I1wx, I1wy, grad, rho_c, p11, p12, p21, p22,
+                      u1, u2, diff, l_t, static_cast<float>(theta), calc_error);
+            if(calc_error)
+            {
+                error = ocl::sum(diff)[0];
+                prev_error = error;
+            }
+            else
+            {
+                error = numeric_limits<double>::max();
+                prev_error -= scaledEpsilon;
+            }
+            estimateDualVariables(u1, u2, p11, p12, p21, p22, taut);
+
+        }
+    }
+
+
+}
+
+void cv::ocl::OpticalFlowDual_TVL1_OCL::collectGarbage()
+{
+    I0s.clear();
+    I1s.clear();
+    u1s.clear();
+    u2s.clear();
+
+    I1x_buf.release();
+    I1y_buf.release();
+
+    I1w_buf.release();
+    I1wx_buf.release();
+    I1wy_buf.release();
+
+    grad_buf.release();
+    rho_c_buf.release();
+
+    p11_buf.release();
+    p12_buf.release();
+    p21_buf.release();
+    p22_buf.release();
+
+    diff_buf.release();
+    norm_buf.release();
+}
+
+void ocl_tvl1flow::centeredGradient(const oclMat &src, oclMat &dx, oclMat &dy)
+{
+    Context  *clCxt = src.clCxt;
+    size_t localThreads[3] = {32, 8, 1};
+    size_t globalThreads[3] = {src.cols, src.rows, 1};
+
+    int srcElementSize = src.elemSize();
+    int src_step = src.step/srcElementSize;
+
+    int dElememntSize = dx.elemSize();
+    int dx_step = dx.step/dElememntSize;
+
+    String kernelName = "centeredGradientKernel";
+    vector< pair<size_t, const void *> > args;
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&src.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&src.cols));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&src.rows));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&src_step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&dx.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&dy.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&dx_step));
+    openCLExecuteKernel(clCxt, &tvl1flow, kernelName, globalThreads, localThreads, args, -1, -1);
+
+}
+
+void ocl_tvl1flow::estimateDualVariables(oclMat &u1, oclMat &u2, oclMat &p11, oclMat &p12, oclMat &p21, oclMat &p22, float taut)
+{
+    Context *clCxt = u1.clCxt;
+
+    size_t localThread[] = {32, 8, 1};
+    size_t globalThread[] =
+    {
+        u1.cols,
+        u1.rows,
+        1
+    };
+
+    int u1_element_size = u1.elemSize();
+    int u1_step = u1.step/u1_element_size;
+
+    int u2_element_size = u2.elemSize();
+    int u2_step = u2.step/u2_element_size;
+
+    int p11_element_size = p11.elemSize();
+    int p11_step = p11.step/p11_element_size;
+
+    int u1_offset_y = u1.offset/u1.step;
+    int u1_offset_x = u1.offset%u1.step;
+    u1_offset_x = u1_offset_x/u1.elemSize();
+
+    int u2_offset_y = u2.offset/u2.step;
+    int u2_offset_x = u2.offset%u2.step;
+    u2_offset_x = u2_offset_x/u2.elemSize();
+
+    String kernelName = "estimateDualVariablesKernel";
+    vector< pair<size_t, const void *> > args;
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u1.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1.cols));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1.rows));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u2.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p11.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&p11_step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p12.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p21.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p22.data));
+    args.push_back( make_pair( sizeof(cl_float), (void*)&taut));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_step));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_y));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_y));
+
+    openCLExecuteKernel(clCxt, &tvl1flow, kernelName, globalThread, localThread, args, -1, -1);
+}
+
+void ocl_tvl1flow::estimateU(oclMat &I1wx, oclMat &I1wy, oclMat &grad,
+    oclMat &rho_c, oclMat &p11, oclMat &p12,
+    oclMat &p21, oclMat &p22, oclMat &u1,
+    oclMat &u2, oclMat &error, float l_t, float theta, char calc_error)
+{
+    Context* clCxt = I1wx.clCxt;
+
+    size_t localThread[] = {32, 8, 1};
+    size_t globalThread[] =
+    {
+        I1wx.cols,
+        I1wx.rows,
+        1
+    };
+
+    int I1wx_element_size = I1wx.elemSize();
+    int I1wx_step = I1wx.step/I1wx_element_size;
+
+    int u1_element_size = u1.elemSize();
+    int u1_step = u1.step/u1_element_size;
+
+    int u2_element_size = u2.elemSize();
+    int u2_step = u2.step/u2_element_size;
+
+    int u1_offset_y = u1.offset/u1.step;
+    int u1_offset_x = u1.offset%u1.step;
+    u1_offset_x = u1_offset_x/u1.elemSize();
+
+    int u2_offset_y = u2.offset/u2.step;
+    int u2_offset_x = u2.offset%u2.step;
+    u2_offset_x = u2_offset_x/u2.elemSize();
+
+    String kernelName = "estimateUKernel";
+    vector< pair<size_t, const void *> > args;
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1wx.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I1wx.cols));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I1wx.rows));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I1wx_step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1wy.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&grad.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&rho_c.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p11.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p12.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p21.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&p22.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u1.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u2.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&error.data));
+    args.push_back( make_pair( sizeof(cl_float), (void*)&l_t));
+    args.push_back( make_pair( sizeof(cl_float), (void*)&theta));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_step));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_y));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_y));
+    args.push_back( make_pair( sizeof(cl_char), (void*)&calc_error));
+
+    openCLExecuteKernel(clCxt, &tvl1flow, kernelName, globalThread, localThread, args, -1, -1);
+}
+
+void ocl_tvl1flow::warpBackward(const oclMat &I0, const oclMat &I1, oclMat &I1x, oclMat &I1y, oclMat &u1, oclMat &u2, oclMat &I1w, oclMat &I1wx, oclMat &I1wy, oclMat &grad, oclMat &rho)
+{
+    Context* clCxt = I0.clCxt;
+    const bool isImgSupported = support_image2d(clCxt);
+
+    CV_Assert(isImgSupported);
+
+    int u1ElementSize = u1.elemSize();
+    int u1Step = u1.step/u1ElementSize;
+
+    int u2ElementSize = u2.elemSize();
+    int u2Step = u2.step/u2ElementSize;
+
+    int I0ElementSize = I0.elemSize();
+    int I0Step = I0.step/I0ElementSize;
+
+    int I1w_element_size = I1w.elemSize();
+    int I1w_step = I1w.step/I1w_element_size;
+
+    int u1_offset_y = u1.offset/u1.step;
+    int u1_offset_x = u1.offset%u1.step;
+    u1_offset_x = u1_offset_x/u1.elemSize();
+
+    int u2_offset_y = u2.offset/u2.step;
+    int u2_offset_x = u2.offset%u2.step;
+    u2_offset_x = u2_offset_x/u2.elemSize();
+
+    size_t localThread[] = {32, 8, 1};
+    size_t globalThread[] =
+    {
+        I0.cols,
+        I0.rows,
+        1
+    };
+
+    cl_mem I1_tex;
+    cl_mem I1x_tex;
+    cl_mem I1y_tex;
+    I1_tex = bindTexture(I1);
+    I1x_tex = bindTexture(I1x);
+    I1y_tex = bindTexture(I1y);
+
+    String kernelName = "warpBackwardKernel";
+    vector< pair<size_t, const void *> > args;
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I0.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I0Step));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I0.cols));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I0.rows));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1_tex));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1x_tex));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1y_tex));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u1.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1Step));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&u2.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1w.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1wx.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&I1wy.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&grad.data));
+    args.push_back( make_pair( sizeof(cl_mem), (void*)&rho.data));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&I1w_step));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2Step));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u1_offset_y));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_x));
+    args.push_back( make_pair( sizeof(cl_int), (void*)&u2_offset_y));
+
+    openCLExecuteKernel(clCxt, &tvl1flow, kernelName, globalThread, localThread, args, -1, -1);
+
+    releaseTexture(I1_tex);
+    releaseTexture(I1x_tex);
+    releaseTexture(I1y_tex);
+}
diff --git a/modules/ocl/test/interpolation.hpp b/modules/ocl/test/interpolation.hpp
deleted file mode 100644
index fb89e701d7..0000000000
--- a/modules/ocl/test/interpolation.hpp
+++ /dev/null
@@ -1,120 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#ifndef __OPENCV_TEST_INTERPOLATION_HPP__
-#define __OPENCV_TEST_INTERPOLATION_HPP__
-
-template <typename T> T readVal(const cv::Mat &src, int y, int x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-{
-    if (border_type == cv::BORDER_CONSTANT)
-        return (y >= 0 && y < src.rows && x >= 0 && x < src.cols) ? src.at<T>(y, x * src.channels() + c) : cv::saturate_cast<T>(borderVal.val[c]);
-
-    return src.at<T>(cv::borderInterpolate(y, src.rows, border_type), cv::borderInterpolate(x, src.cols, border_type) * src.channels() + c);
-}
-
-template <typename T> struct NearestInterpolator
-{
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        return readVal<T>(src, cvFloor(y), cvFloor(x), c, border_type, borderVal);
-    }
-};
-
-template <typename T> struct LinearInterpolator
-{
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        x -= 0.5f;
-        y -= 0.5f;
-
-        int x1 = cvFloor(x);
-        int y1 = cvFloor(y);
-        int x2 = x1 + 1;
-        int y2 = y1 + 1;
-
-        float res = 0;
-
-        res += readVal<T>(src, y1, x1, c, border_type, borderVal) * ((x2 - x) * (y2 - y));
-        res += readVal<T>(src, y1, x2, c, border_type, borderVal) * ((x - x1) * (y2 - y));
-        res += readVal<T>(src, y2, x1, c, border_type, borderVal) * ((x2 - x) * (y - y1));
-        res += readVal<T>(src, y2, x2, c, border_type, borderVal) * ((x - x1) * (y - y1));
-
-        return cv::saturate_cast<T>(res);
-    }
-};
-
-template <typename T> struct CubicInterpolator
-{
-    static float getValue(float p[4], float x)
-    {
-        return p[1] + 0.5 * x * (p[2] - p[0] + x * (2.0 * p[0] - 5.0 * p[1] + 4.0 * p[2] - p[3] + x * (3.0 * (p[1] - p[2]) + p[3] - p[0])));
-    }
-
-    static float getValue(float p[4][4], float x, float y)
-    {
-        float arr[4];
-
-        arr[0] = getValue(p[0], x);
-        arr[1] = getValue(p[1], x);
-        arr[2] = getValue(p[2], x);
-        arr[3] = getValue(p[3], x);
-
-        return getValue(arr, y);
-    }
-
-    static T getValue(const cv::Mat &src, float y, float x, int c, int border_type, cv::Scalar borderVal = cv::Scalar())
-    {
-        int ix = cvRound(x);
-        int iy = cvRound(y);
-
-        float vals[4][4] =
-        {
-            {readVal<T>(src, iy - 2, ix - 2, c, border_type, borderVal), readVal<T>(src, iy - 2, ix - 1, c, border_type, borderVal), readVal<T>(src, iy - 2, ix, c, border_type, borderVal), readVal<T>(src, iy - 2, ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy - 1, ix - 2, c, border_type, borderVal), readVal<T>(src, iy - 1, ix - 1, c, border_type, borderVal), readVal<T>(src, iy - 1, ix, c, border_type, borderVal), readVal<T>(src, iy - 1, ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy    , ix - 2, c, border_type, borderVal), readVal<T>(src, iy    , ix - 1, c, border_type, borderVal), readVal<T>(src, iy    , ix, c, border_type, borderVal), readVal<T>(src, iy    , ix + 1, c, border_type, borderVal)},
-            {readVal<T>(src, iy + 1, ix - 2, c, border_type, borderVal), readVal<T>(src, iy + 1, ix - 1, c, border_type, borderVal), readVal<T>(src, iy + 1, ix, c, border_type, borderVal), readVal<T>(src, iy + 1, ix + 1, c, border_type, borderVal)},
-        };
-
-        return cv::saturate_cast<T>(getValue(vals, (x - ix + 2.0) / 4.0, (y - iy + 2.0) / 4.0));
-    }
-};
-
-#endif // __OPENCV_TEST_INTERPOLATION_HPP__
diff --git a/modules/ocl/test/main.cpp b/modules/ocl/test/main.cpp
index d3273f7fe0..2a7be49005 100644
--- a/modules/ocl/test/main.cpp
+++ b/modules/ocl/test/main.cpp
@@ -39,15 +39,16 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
-using namespace std;
 using namespace cv;
 using namespace cv::ocl;
 using namespace cvtest;
 using namespace testing;
+using std::cout;
+using std::endl;
 
 void print_info()
 {
@@ -73,14 +74,12 @@ void print_info()
 #endif
 
 }
-std::string workdir;
 int main(int argc, char **argv)
 {
-    TS::ptr()->init("ocl");
+    TS::ptr()->init(".");
     InitGoogleTest(&argc, argv);
     const char *keys =
         "{ h | false              | print help message }"
-		"{ w | ../../../samples/c/| set working directory i.e. -w=C:\\}"
         "{ t | gpu                | set device type:i.e. -t=cpu or gpu}"
         "{ p | 0                  | set platform id i.e. -p=0}"
         "{ d | 0                  | set device id i.e. -d=0}";
@@ -88,12 +87,11 @@ int main(int argc, char **argv)
     CommandLineParser cmd(argc, argv, keys);
     if (cmd.get<string>("h")=="true")
     {
-        cout << "Avaible options besides goole test option:" << endl;
+        cout << "Available options besides google test options:" << endl;
         cmd.printMessage();
         return 0;
     }
-    workdir = cmd.get<string>("w");
-    string type = cmd.get<string>("t");
+    string type = cmd.get<String>("t");
     unsigned int pid = cmd.get<unsigned int>("p");
     int device = cmd.get<int>("d");
 
@@ -115,10 +113,11 @@ int main(int argc, char **argv)
         std::cout << "platform invalid\n";
         return -1;
     }
-    if(pid != 0 || device != 0)
-    {
-        setDevice(oclinfo[pid], device);
-    }
+
+    setDevice(oclinfo[pid], device);
+    setBinaryDiskCache(CACHE_UPDATE);
+
+    cout << "Platform name:" << oclinfo[pid].PlatformName << endl;
     cout << "Device type:" << type << endl << "Device name:" << oclinfo[pid].DeviceName[device] << endl;
     return RUN_ALL_TESTS();
 }
diff --git a/modules/ocl/test/precomp.cpp b/modules/ocl/test/precomp.cpp
deleted file mode 100644
index 7d287004ee..0000000000
--- a/modules/ocl/test/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-
diff --git a/modules/ocl/test/test_arithm.cpp b/modules/ocl/test/test_arithm.cpp
index f643864a86..fa9d099990 100644
--- a/modules/ocl/test/test_arithm.cpp
+++ b/modules/ocl/test/test_arithm.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
@@ -21,6 +22,7 @@
 //    Jiang Liyuan,jlyuan001.good@163.com
 //    Rock Li, Rock.Li@amd.com
 //    Zailong Wu, bullet@yeah.net
+//    Yao Wang, bitwangyaoyao@gmail.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -52,7 +54,7 @@
 //#define PRINT_TIME
 
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #include <iomanip>
 
 #ifdef HAVE_OPENCL
@@ -87,14 +89,13 @@ PARAM_TEST_CASE(ArithmTestBase, MatType, bool)
     int maskx;
     int masky;
 
-
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat mat2_roi;
     cv::Mat mask_roi;
     cv::Mat dst_roi;
     cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
     cv::ocl::oclMat gdst1_whole; //bak
@@ -125,10 +126,6 @@ PARAM_TEST_CASE(ArithmTestBase, MatType, bool)
 
         val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -175,14 +172,22 @@ PARAM_TEST_CASE(ArithmTestBase, MatType, bool)
         gmask = mask_roi; //end
     }
 
+    void Near(double threshold = 0.)
+    {
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), threshold);
+    }
+
+    void Near1(double threshold = 0.)
+    {
+        EXPECT_MAT_NEAR(dst1, Mat(gdst1_whole), threshold);
+    }
+
 };
 ////////////////////////////////lut/////////////////////////////////////////////////
-
 struct Lut : ArithmTestBase {};
 #define VARNAME(A) string(#A);
 
 
-
 TEST_P(Lut, Mat)
 {
 
@@ -203,20 +208,12 @@ TEST_P(Lut, Mat)
 
         cv::LUT(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::LUT(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download (cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0, s);
+        Near(0);
     }
 }
 
 
-
-
 ////////////////////////////////exp/////////////////////////////////////////////////
-
 struct Exp : ArithmTestBase {};
 
 TEST_P(Exp, Mat)
@@ -227,20 +224,12 @@ TEST_P(Exp, Mat)
 
         cv::exp(mat1_roi, dst_roi);
         cv::ocl::exp(gmat1, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 2, s);
-
+        Near(2);
     }
 }
 
 
 ////////////////////////////////log/////////////////////////////////////////////////
-
 struct Log : ArithmTestBase {};
 
 TEST_P(Log, Mat)
@@ -249,24 +238,14 @@ TEST_P(Log, Mat)
     {
         random_roi();
 
-
         cv::log(mat1_roi, dst_roi);
         cv::ocl::log(gmat1, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, s);
-
+        Near(1);
     }
 }
 
 
-
-
 ////////////////////////////////add/////////////////////////////////////////////////
-
 struct Add : ArithmTestBase {};
 
 TEST_P(Add, Mat)
@@ -277,12 +256,7 @@ TEST_P(Add, Mat)
 
         cv::add(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::add(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, s);
+        Near(0);
     }
 }
 
@@ -294,14 +268,10 @@ TEST_P(Add, Mat_Mask)
 
         cv::add(mat1_roi, mat2_roi, dst_roi, mask_roi);
         cv::ocl::add(gmat1, gmat2, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, s);
+        Near(0);
     }
 }
+
 TEST_P(Add, Scalar)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
@@ -310,12 +280,7 @@ TEST_P(Add, Scalar)
 
         cv::add(mat1_roi, val, dst_roi);
         cv::ocl::add(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, s);
+        Near(1e-5);
     }
 }
 
@@ -327,12 +292,7 @@ TEST_P(Add, Scalar_Mask)
 
         cv::add(mat1_roi, val, dst_roi, mask_roi);
         cv::ocl::add(gmat1, val, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, s);
+        Near(1e-5);
     }
 }
 
@@ -349,12 +309,7 @@ TEST_P(Sub, Mat)
 
         cv::subtract(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::subtract(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, s);
+        Near(0);
     }
 }
 
@@ -366,14 +321,10 @@ TEST_P(Sub, Mat_Mask)
 
         cv::subtract(mat1_roi, mat2_roi, dst_roi, mask_roi);
         cv::ocl::subtract(gmat1, gmat2, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, s);
+        Near(0);
     }
 }
+
 TEST_P(Sub, Scalar)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
@@ -382,12 +333,7 @@ TEST_P(Sub, Scalar)
 
         cv::subtract(mat1_roi, val, dst_roi);
         cv::ocl::subtract(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, s);
+        Near(1e-5);
     }
 }
 
@@ -399,12 +345,7 @@ TEST_P(Sub, Scalar_Mask)
 
         cv::subtract(mat1_roi, val, dst_roi, mask_roi);
         cv::ocl::subtract(gmat1, val, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, s);
+        Near(1e-5);
     }
 }
 
@@ -421,12 +362,7 @@ TEST_P(Mul, Mat)
 
         cv::multiply(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::multiply(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char s[1024];
-        sprintf(s, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, s);
+        Near(0);
     }
 }
 
@@ -441,12 +377,7 @@ TEST_P(Mul, Mat_Scalar)
 
         cv::multiply(mat1_roi, mat2_roi, dst_roi, s);
         cv::ocl::multiply(gmat1, gmat2, gdst, s);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.001, sss);
+        Near(.001);
     }
 }
 
@@ -462,13 +393,7 @@ TEST_P(Div, Mat)
 
         cv::divide(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::divide(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        Near(1);
     }
 }
 
@@ -483,13 +408,7 @@ TEST_P(Div, Mat_Scalar)
 
         cv::divide(mat1_roi, mat2_roi, dst_roi, s);
         cv::ocl::divide(gmat1, gmat2, gdst, s);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.001, sss);
+        Near(.001);
     }
 }
 
@@ -504,13 +423,7 @@ TEST_P(Absdiff, Mat)
 
         cv::absdiff(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::absdiff(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0, sss);
+        Near(0);
     }
 }
 
@@ -522,13 +435,7 @@ TEST_P(Absdiff, Mat_Scalar)
 
         cv::absdiff(mat1_roi, val, dst_roi);
         cv::ocl::absdiff(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -544,16 +451,8 @@ TEST_P(CartToPolar, angleInDegree)
 
         cv::cartToPolar(mat1_roi, mat2_roi, dst_roi, dst1_roi, 1);
         cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 1);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        cv::Mat cpu_dst1;
-        gdst1_whole.download(cpu_dst1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.5, sss);
-        EXPECT_MAT_NEAR(dst1, cpu_dst1, 0.5, sss);
+        Near(.5);
+        Near1(.5);
     }
 }
 
@@ -565,22 +464,12 @@ TEST_P(CartToPolar, angleInRadians)
 
         cv::cartToPolar(mat1_roi, mat2_roi, dst_roi, dst1_roi, 0);
         cv::ocl::cartToPolar(gmat1, gmat2, gdst, gdst1, 0);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        cv::Mat cpu_dst1;
-        gdst1_whole.download(cpu_dst1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.5, sss);
-        EXPECT_MAT_NEAR(dst1, cpu_dst1, 0.5, sss);
+        Near(.5);
+        Near1(.5);
     }
 }
 
 
-
-
 struct PolarToCart : ArithmTestBase {};
 
 TEST_P(PolarToCart, angleInDegree)
@@ -591,17 +480,8 @@ TEST_P(PolarToCart, angleInDegree)
 
         cv::polarToCart(mat1_roi, mat2_roi, dst_roi, dst1_roi, 1);
         cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 1);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        cv::Mat cpu_dst1;
-        gdst1_whole.download(cpu_dst1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.5, sss);
-        EXPECT_MAT_NEAR(dst1, cpu_dst1, 0.5, sss);
+        Near(.5);
+        Near1(.5);
     }
 }
 
@@ -613,17 +493,8 @@ TEST_P(PolarToCart, angleInRadians)
 
         cv::polarToCart(mat1_roi, mat2_roi, dst_roi, dst1_roi, 0);
         cv::ocl::polarToCart(gmat1, gmat2, gdst, gdst1, 0);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        cv::Mat cpu_dst1;
-        gdst1_whole.download(cpu_dst1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.5, sss);
-        EXPECT_MAT_NEAR(dst1, cpu_dst1, 0.5, sss);
+        Near(.5);
+        Near1(.5);
     }
 }
 
@@ -640,19 +511,11 @@ TEST_P(Magnitude, Mat)
 
         cv::magnitude(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::magnitude(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
 
-
-
 struct Transpose : ArithmTestBase {};
 
 TEST_P(Transpose, Mat)
@@ -663,20 +526,11 @@ TEST_P(Transpose, Mat)
 
         cv::transpose(mat1_roi, dst_roi);
         cv::ocl::transpose(gmat1, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
 
-
-
-
 struct Flip : ArithmTestBase {};
 
 TEST_P(Flip, X)
@@ -687,13 +541,7 @@ TEST_P(Flip, X)
 
         cv::flip(mat1_roi, dst_roi, 0);
         cv::ocl::flip(gmat1, gdst, 0);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -705,13 +553,7 @@ TEST_P(Flip, Y)
 
         cv::flip(mat1_roi, dst_roi, 1);
         cv::ocl::flip(gmat1, gdst, 1);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -723,18 +565,11 @@ TEST_P(Flip, BOTH)
 
         cv::flip(mat1_roi, dst_roi, -1);
         cv::ocl::flip(gmat1, gdst, -1);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
 
-
 struct MinMax : ArithmTestBase {};
 
 TEST_P(MinMax, MAT)
@@ -765,12 +600,8 @@ TEST_P(MinMax, MAT)
         double minVal_, maxVal_;
         cv::ocl::minMax(gmat1, &minVal_, &maxVal_);
 
-        //check results
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_DOUBLE_EQ(minVal_, minVal) << sss;
-        EXPECT_DOUBLE_EQ(maxVal_, maxVal) << sss;
+        EXPECT_DOUBLE_EQ(minVal_, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_, maxVal);
     }
 }
 
@@ -803,12 +634,8 @@ TEST_P(MinMax, MASK)
         double minVal_, maxVal_;
         cv::ocl::minMax(gmat1, &minVal_, &maxVal_, gmask);
 
-        //check results
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_DOUBLE_EQ(minVal_, minVal) << sss;
-        EXPECT_DOUBLE_EQ(maxVal_, maxVal) << sss;
+        EXPECT_DOUBLE_EQ(minVal_, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_, maxVal);
     }
 }
 
@@ -919,17 +746,13 @@ TEST_P(MinMaxLoc, MAT)
             error1 = ::abs(mat1_roi.at<double>(maxLoc_) - mat1_roi.at<double>(maxLoc));
         }
 
-        //check results
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
+        EXPECT_DOUBLE_EQ(minVal_, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_, maxVal);
+        EXPECT_DOUBLE_EQ(minlocVal_, minlocVal);
+        EXPECT_DOUBLE_EQ(maxlocVal_, maxlocVal);
 
-        EXPECT_DOUBLE_EQ(minVal_, minVal) << sss;
-        EXPECT_DOUBLE_EQ(maxVal_, maxVal) << sss;
-        EXPECT_DOUBLE_EQ(minlocVal_, minlocVal) << sss;
-        EXPECT_DOUBLE_EQ(maxlocVal_, maxlocVal) << sss;
-
-        EXPECT_DOUBLE_EQ(error0, 0.0) << sss;
-        EXPECT_DOUBLE_EQ(error1, 0.0) << sss;
+        EXPECT_DOUBLE_EQ(error0, 0.0);
+        EXPECT_DOUBLE_EQ(error1, 0.0);
     }
 }
 
@@ -1040,17 +863,13 @@ TEST_P(MinMaxLoc, MASK)
             error1 = ::abs(mat1_roi.at<double>(maxLoc_) - mat1_roi.at<double>(maxLoc));
         }
 
-        //check results
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
+        EXPECT_DOUBLE_EQ(minVal_, minVal);
+        EXPECT_DOUBLE_EQ(maxVal_, maxVal);
+        EXPECT_DOUBLE_EQ(minlocVal_, minlocVal);
+        EXPECT_DOUBLE_EQ(maxlocVal_, maxlocVal);
 
-        EXPECT_DOUBLE_EQ(minVal_, minVal) << sss;
-        EXPECT_DOUBLE_EQ(maxVal_, maxVal) << sss;
-        EXPECT_DOUBLE_EQ(minlocVal_, minlocVal) << sss;
-        EXPECT_DOUBLE_EQ(maxlocVal_, maxlocVal) << sss;
-
-        EXPECT_DOUBLE_EQ(error0, 0.0) << sss;
-        EXPECT_DOUBLE_EQ(error1, 0.0) << sss;
+        EXPECT_DOUBLE_EQ(error0, 0.0);
+        EXPECT_DOUBLE_EQ(error1, 0.0);
     }
 }
 
@@ -1064,14 +883,12 @@ TEST_P(Sum, MAT)
         random_roi();
         Scalar cpures = cv::sum(mat1_roi);
         Scalar gpures = cv::ocl::sum(gmat1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
 
         //check results
-        EXPECT_NEAR(cpures[0], gpures[0], 0.1) << sss;
-        EXPECT_NEAR(cpures[1], gpures[1], 0.1) << sss;
-        EXPECT_NEAR(cpures[2], gpures[2], 0.1) << sss;
-        EXPECT_NEAR(cpures[3], gpures[3], 0.1) << sss;
+        EXPECT_NEAR(cpures[0], gpures[0], 0.1);
+        EXPECT_NEAR(cpures[1], gpures[1], 0.1);
+        EXPECT_NEAR(cpures[2], gpures[2], 0.1);
+        EXPECT_NEAR(cpures[3], gpures[3], 0.1);
     }
 }
 
@@ -1086,11 +903,7 @@ TEST_P(CountNonZero, MAT)
         int cpures = cv::countNonZero(mat1_roi);
         int gpures = cv::ocl::countNonZero(gmat1);
 
-        //check results
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_DOUBLE_EQ((double)cpures, (double)gpures) << sss;
+        EXPECT_DOUBLE_EQ((double)cpures, (double)gpures);
     }
 }
 
@@ -1112,13 +925,7 @@ TEST_P(Phase, Mat)
             random_roi();
             cv::phase(mat1_roi, mat2_roi, dst_roi, angelInDegrees ? true : false);
             cv::ocl::phase(gmat1, gmat2, gdst, angelInDegrees ? true : false);
-
-            cv::Mat cpu_dst;
-            gdst_whole.download(cpu_dst);
-
-            char sss[1024];
-            sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-            EXPECT_MAT_NEAR(dst, cpu_dst, 1e-2, sss);
+            Near(1e-2);
         }
     }
 }
@@ -1135,13 +942,7 @@ TEST_P(Bitwise_and, Mat)
 
         cv::bitwise_and(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::bitwise_and(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
 
@@ -1153,15 +954,10 @@ TEST_P(Bitwise_and, Mat_Mask)
 
         cv::bitwise_and(mat1_roi, mat2_roi, dst_roi, mask_roi);
         cv::ocl::bitwise_and(gmat1, gmat2, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
+
 TEST_P(Bitwise_and, Scalar)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
@@ -1170,14 +966,7 @@ TEST_P(Bitwise_and, Scalar)
 
         cv::bitwise_and(mat1_roi, val, dst_roi);
         cv::ocl::bitwise_and(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
-
+        Near(1e-5);
     }
 }
 
@@ -1189,14 +978,7 @@ TEST_P(Bitwise_and, Scalar_Mask)
 
         cv::bitwise_and(mat1_roi, val, dst_roi, mask_roi);
         cv::ocl::bitwise_and(gmat1, val, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char *sss = new char[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
-        delete[] sss;
+        Near(1e-5);
     }
 }
 
@@ -1214,13 +996,7 @@ TEST_P(Bitwise_or, Mat)
 
         cv::bitwise_or(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::bitwise_or(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
 
@@ -1232,15 +1008,10 @@ TEST_P(Bitwise_or, Mat_Mask)
 
         cv::bitwise_or(mat1_roi, mat2_roi, dst_roi, mask_roi);
         cv::ocl::bitwise_or(gmat1, gmat2, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
+
 TEST_P(Bitwise_or, Scalar)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
@@ -1249,13 +1020,7 @@ TEST_P(Bitwise_or, Scalar)
 
         cv::bitwise_or(mat1_roi, val, dst_roi);
         cv::ocl::bitwise_or(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -1267,13 +1032,7 @@ TEST_P(Bitwise_or, Scalar_Mask)
 
         cv::bitwise_or(mat1_roi, val, dst_roi, mask_roi);
         cv::ocl::bitwise_or(gmat1, val, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -1291,13 +1050,7 @@ TEST_P(Bitwise_xor, Mat)
 
         cv::bitwise_xor(mat1_roi, mat2_roi, dst_roi);
         cv::ocl::bitwise_xor(gmat1, gmat2, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
 
@@ -1309,15 +1062,10 @@ TEST_P(Bitwise_xor, Mat_Mask)
 
         cv::bitwise_xor(mat1_roi, mat2_roi, dst_roi, mask_roi);
         cv::ocl::bitwise_xor(gmat1, gmat2, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
+
 TEST_P(Bitwise_xor, Scalar)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
@@ -1326,13 +1074,7 @@ TEST_P(Bitwise_xor, Scalar)
 
         cv::bitwise_xor(mat1_roi, val, dst_roi);
         cv::ocl::bitwise_xor(gmat1, val, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -1344,13 +1086,7 @@ TEST_P(Bitwise_xor, Scalar_Mask)
 
         cv::bitwise_xor(mat1_roi, val, dst_roi, mask_roi);
         cv::ocl::bitwise_xor(gmat1, val, gdst, gmask);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
@@ -1367,13 +1103,7 @@ TEST_P(Bitwise_not, Mat)
 
         cv::bitwise_not(mat1_roi, dst_roi);
         cv::ocl::bitwise_not(gmat1, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        Near(0);
     }
 }
 
@@ -1390,7 +1120,7 @@ TEST_P(Compare, Mat)
     }
 
     int cmp_codes[] = {CMP_EQ, CMP_GT, CMP_GE, CMP_LT, CMP_LE, CMP_NE};
-    const char *cmp_str[] = {"CMP_EQ", "CMP_GT", "CMP_GE", "CMP_LT", "CMP_LE", "CMP_NE"};
+    //const char *cmp_str[] = {"CMP_EQ", "CMP_GT", "CMP_GE", "CMP_LT", "CMP_LE", "CMP_NE"};
     int cmp_num = sizeof(cmp_codes) / sizeof(int);
 
     for (int i = 0; i < cmp_num; ++i)
@@ -1402,13 +1132,7 @@ TEST_P(Compare, Mat)
 
             cv::compare(mat1_roi, mat2_roi, dst_roi, cmp_codes[i]);
             cv::ocl::compare(gmat1, gmat2, gdst, cmp_codes[i]);
-
-            cv::Mat cpu_dst;
-            gdst_whole.download(cpu_dst);
-            char sss[1024];
-            sprintf(sss, "cmptype=%s, roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", cmp_str[i], roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-            EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+            Near(0);
         }
     }
 
@@ -1430,14 +1154,7 @@ TEST_P(Pow, Mat)
         double p = 4.5;
         cv::pow(mat1_roi, p, dst_roi);
         cv::ocl::pow(gmat1, p, gdst);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        Near(1);
     }
 }
 
@@ -1448,36 +1165,18 @@ TEST_P(MagnitudeSqr, Mat)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
     {
-        // random_roi();
-        // int64 start, end;
-        // start = cv::getTickCount();
+        random_roi();
         for(int i = 0; i < mat1.rows; ++i)
             for(int j = 0; j < mat1.cols; ++j)
             {
                 float val1 = mat1.at<float>(i, j);
                 float val2 = mat2.at<float>(i, j);
-
                 ((float *)(dst.data))[i * dst.step / 4 + j] = val1 * val1 + val2 * val2;
-
-                //        float val1 =((float *)( mat1.data))[(i*mat1.step/8 +j)*2];
-                //
-                //     float val2 =((float *)( mat1.data))[(i*mat1.step/8 +j)*2+ 1 ];
-
-                //  ((float *)(dst.data))[i*dst.step/4 +j]= val1 * val1 +val2 * val2;
             }
-        // end = cv::getTickCount();
 
-
-
-        cv::ocl::oclMat clmat1(mat1), clmat2(mat2), cldst;
-        cv::ocl::magnitudeSqr(clmat1, clmat2, cldst);
-
-        cv::Mat cpu_dst;
-        cldst.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        cv::ocl::oclMat clmat1(mat1), clmat2(mat2);
+        cv::ocl::magnitudeSqr(clmat1, clmat2, gdst);
+        Near(1);
     }
 }
 
@@ -1498,21 +1197,13 @@ TEST_P(AddWeighted, Mat)
 
         cv::ocl::addWeighted(gmat1, alpha, gmat2, beta, gama, gdst);
 
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
     }
 }
 
 
 
 
-
 //********test****************
 
 INSTANTIATE_TEST_CASE_P(Arithm, Lut, Combine(
@@ -1531,6 +1222,10 @@ INSTANTIATE_TEST_CASE_P(Arithm, Add, Combine(
                             Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4),
                             Values(false)));
 
+INSTANTIATE_TEST_CASE_P(Arithm, Sub, Combine(
+                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4),
+                            Values(false)));
+
 INSTANTIATE_TEST_CASE_P(Arithm, Mul, Combine(
                             Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4),
                             Values(false))); // Values(false) is the reserved parameter
@@ -1586,19 +1281,19 @@ INSTANTIATE_TEST_CASE_P(Arithm, Phase, Combine(Values(CV_32FC1, CV_32FC3, CV_32F
 
 
 INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_and, Combine(
-                            Values(CV_8UC1, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
+                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
 //Values(false) is the reserved parameter
 
 INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_or, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
+                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
 //Values(false) is the reserved parameter
 
 INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_xor, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
+                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
 //Values(false) is the reserved parameter
 
 INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_not, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
+                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false)));
 //Values(false) is the reserved parameter
 
 INSTANTIATE_TEST_CASE_P(Arithm, Compare, Combine(Values(CV_8UC1, CV_32SC1, CV_32FC1), Values(false)));
diff --git a/modules/ocl/test/test_bgfg.cpp b/modules/ocl/test/test_bgfg.cpp
new file mode 100644
index 0000000000..d7492d6007
--- /dev/null
+++ b/modules/ocl/test/test_bgfg.cpp
@@ -0,0 +1,227 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jin Ma, jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_OPENCL
+
+using namespace cv;
+using namespace cv::ocl;
+using namespace cvtest;
+using namespace testing;
+using namespace std;
+
+extern string workdir;
+//////////////////////////////////////////////////////
+// MOG
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(UseGray, bool)
+    IMPLEMENT_PARAM_CLASS(LearningRate, double)
+}
+
+PARAM_TEST_CASE(mog, UseGray, LearningRate, bool)
+{
+    bool useGray;
+    double learningRate;
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        useGray = GET_PARAM(0);
+
+        learningRate = GET_PARAM(1);
+
+        useRoi = GET_PARAM(2);
+    }
+};
+
+TEST_P(mog, Update)
+{
+    std::string inputFile = string(cvtest::TS::ptr()->get_data_path()) + "gpu/video/768x576.avi";
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    cv::ocl::MOG mog;
+    cv::ocl::oclMat foreground = createMat_ocl(frame.size(), CV_8UC1, useRoi);
+
+    Ptr<cv::BackgroundSubtractorMOG> mog_gold = createBackgroundSubtractorMOG();
+    cv::Mat foreground_gold;
+
+    for (int i = 0; i < 10; ++i)
+    {
+        cap >> frame;
+        ASSERT_FALSE(frame.empty());
+
+        if (useGray)
+        {
+            cv::Mat temp;
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+            cv::swap(temp, frame);
+        }
+
+        mog(loadMat_ocl(frame, useRoi), foreground, (float)learningRate);
+
+        mog_gold->apply(frame, foreground_gold, learningRate);
+
+        EXPECT_MAT_NEAR(foreground_gold, foreground, 0.0);
+    }
+}
+INSTANTIATE_TEST_CASE_P(OCL_Video, mog, testing::Combine(
+    testing::Values(UseGray(false), UseGray(true)),
+    testing::Values(LearningRate(0.0), LearningRate(0.01)),
+    Values(true, false)));
+
+//////////////////////////////////////////////////////
+// MOG2
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(DetectShadow, bool)
+}
+
+PARAM_TEST_CASE(mog2, UseGray, DetectShadow, bool)
+{
+    bool useGray;
+    bool detectShadow;
+    bool useRoi;
+    virtual void SetUp()
+    {
+        useGray = GET_PARAM(0);
+        detectShadow = GET_PARAM(1);
+        useRoi = GET_PARAM(2);
+    }
+};
+
+TEST_P(mog2, Update)
+{
+    std::string inputFile = string(cvtest::TS::ptr()->get_data_path()) + "gpu/video/768x576.avi";
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+    cap >> frame;
+    ASSERT_FALSE(frame.empty());
+
+    cv::ocl::MOG2 mog2;
+    mog2.bShadowDetection = detectShadow;
+    cv::ocl::oclMat foreground = createMat_ocl(frame.size(), CV_8UC1, useRoi);
+
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2_gold = createBackgroundSubtractorMOG2();
+    mog2_gold->set("detectShadows", detectShadow);
+    cv::Mat foreground_gold;
+
+    for (int i = 0; i < 10; ++i)
+    {
+        cap >> frame;
+        ASSERT_FALSE(frame.empty());
+
+        if (useGray)
+        {
+            cv::Mat temp;
+            cv::cvtColor(frame, temp, cv::COLOR_BGR2GRAY);
+            cv::swap(temp, frame);
+        }
+
+        mog2(loadMat_ocl(frame, useRoi), foreground);
+
+        mog2_gold->apply(frame, foreground_gold);
+
+        if (detectShadow)
+            EXPECT_MAT_SIMILAR(foreground_gold, foreground, 15e-3)
+        else
+            EXPECT_MAT_NEAR(foreground_gold, foreground, 0)
+    }
+}
+
+TEST_P(mog2, getBackgroundImage)
+{
+    if (useGray)
+        return;
+
+    std::string inputFile = string(cvtest::TS::ptr()->get_data_path()) + "gpu/video/768x576.avi";
+    cv::VideoCapture cap(inputFile);
+    ASSERT_TRUE(cap.isOpened());
+
+    cv::Mat frame;
+
+    cv::ocl::MOG2 mog2;
+    mog2.bShadowDetection = detectShadow;
+    cv::ocl::oclMat foreground;
+
+    cv::Ptr<cv::BackgroundSubtractorMOG2> mog2_gold = createBackgroundSubtractorMOG2();
+    mog2_gold->set("detectShadows", detectShadow);
+    cv::Mat foreground_gold;
+
+    for (int i = 0; i < 10; ++i)
+    {
+        cap >> frame;
+        ASSERT_FALSE(frame.empty());
+
+        mog2(loadMat_ocl(frame, useRoi), foreground);
+
+        mog2_gold->apply(frame, foreground_gold);
+    }
+
+    cv::ocl::oclMat background = createMat_ocl(frame.size(), frame.type(), useRoi);
+    mog2.getBackgroundImage(background);
+
+    cv::Mat background_gold;
+    mog2_gold->getBackgroundImage(background_gold);
+
+    EXPECT_MAT_NEAR(background_gold, background, 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Video, mog2, testing::Combine(
+    testing::Values(UseGray(true), UseGray(false)),
+    testing::Values(DetectShadow(true), DetectShadow(false)),
+    Values(true, false)));
+
+#endif
diff --git a/modules/ocl/test/test_blend.cpp b/modules/ocl/test/test_blend.cpp
index f9c8657d04..ae0cbd41bd 100644
--- a/modules/ocl/test/test_blend.cpp
+++ b/modules/ocl/test/test_blend.cpp
@@ -1,4 +1,48 @@
-#include "precomp.hpp"
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Nathan, liujun@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "test_precomp.hpp"
 #include <iomanip>
 
 using namespace cv;
@@ -33,20 +77,14 @@ void blendLinearGold(const cv::Mat &img1, const cv::Mat &img2, const cv::Mat &we
 
 PARAM_TEST_CASE(Blend, cv::Size, MatType/*, UseRoi*/)
 {
-    //std::vector<cv::ocl::Info> oclinfo;
     cv::Size size;
     int type;
     bool useRoi;
 
     virtual void SetUp()
     {
-        //devInfo = GET_PARAM(0);
         size = GET_PARAM(0);
         type = GET_PARAM(1);
-        /*useRoi = GET_PARAM(3);*/
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
     }
 };
 
@@ -59,12 +97,9 @@ TEST_P(Blend, Accuracy)
     cv::Mat weights1 = randomMat(size, CV_32F, 0, 1);
     cv::Mat weights2 = randomMat(size, CV_32F, 0, 1);
 
-    cv::ocl::oclMat gimg1(size, type), gimg2(size, type), gweights1(size, CV_32F), gweights2(size, CV_32F);
-    cv::ocl::oclMat dst(size, type);
-    gimg1.upload(img1);
-    gimg2.upload(img2);
-    gweights1.upload(weights1);
-    gweights2.upload(weights2);
+    cv::ocl::oclMat gimg1(img1), gimg2(img2), gweights1(weights1), gweights2(weights2);
+    cv::ocl::oclMat dst;
+
     cv::ocl::blendLinear(gimg1, gimg2, gweights1, gweights2, dst);
     cv::Mat result;
     cv::Mat result_gold;
@@ -74,11 +109,11 @@ TEST_P(Blend, Accuracy)
     else
         blendLinearGold<float>(img1, img2, weights1, weights2, result_gold);
 
-    EXPECT_MAT_NEAR(result_gold, result, CV_MAT_DEPTH(type) == CV_8U ? 1.f : 1e-5f, 0);
+    EXPECT_MAT_NEAR(result_gold, result, CV_MAT_DEPTH(type) == CV_8U ? 1.f : 1e-5f);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Blend, Combine(
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, Blend, Combine(
                             DIFFERENT_SIZES,
                             testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_8UC4), MatType(CV_32FC1), MatType(CV_32FC4))
                         ));
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/ocl/test/test_brute_force_matcher.cpp b/modules/ocl/test/test_brute_force_matcher.cpp
index 424781fe0a..4d0b45fb78 100644
--- a/modules/ocl/test/test_brute_force_matcher.cpp
+++ b/modules/ocl/test/test_brute_force_matcher.cpp
@@ -7,12 +7,16 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
-// Copyright (C) 2010-2012, Multicoreware inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
+// @Authors
+//    Nathan, liujun@multicorewareinc.com
+//
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
 //
@@ -21,12 +25,12 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
+//     and/or other oclMaterials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
-// This software is provided by the copyright holders and contributors "as is" and
+// This software is provided by the copyright holders and contributors as is and
 // any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
@@ -39,20 +43,18 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #ifdef HAVE_OPENCL
 namespace
 {
-
     /////////////////////////////////////////////////////////////////////////////////////////////////
     // BruteForceMatcher
-
-    CV_ENUM(DistType, cv::ocl::BruteForceMatcher_OCL_base::L1Dist, cv::ocl::BruteForceMatcher_OCL_base::L2Dist, cv::ocl::BruteForceMatcher_OCL_base::HammingDist)
+    CV_ENUM(DistType, BruteForceMatcher_OCL_base::L1Dist,
+                      BruteForceMatcher_OCL_base::L2Dist,
+                      BruteForceMatcher_OCL_base::HammingDist)
     IMPLEMENT_PARAM_CLASS(DescriptorSize, int)
-
-    PARAM_TEST_CASE(BruteForceMatcher/*, NormCode*/, DistType, DescriptorSize)
+    PARAM_TEST_CASE(BruteForceMatcher, DistType, DescriptorSize)
     {
-        //std::vector<cv::ocl::Info> oclinfo;
         cv::ocl::BruteForceMatcher_OCL_base::DistType distType;
         int normCode;
         int dim;
@@ -64,13 +66,9 @@ namespace
 
         virtual void SetUp()
         {
-            //normCode = GET_PARAM(0);
             distType = (cv::ocl::BruteForceMatcher_OCL_base::DistType)(int)GET_PARAM(0);
             dim = GET_PARAM(1);
 
-            //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-            //CV_Assert(devnums > 0);
-
             queryDescCount = 300; // must be even number because we split train data in some cases in two
             countFactor = 4; // do not change it
 
@@ -164,57 +162,49 @@ namespace
 
     TEST_P(BruteForceMatcher, RadiusMatch_Single)
     {
-        float radius;
-        if(distType == cv::ocl::BruteForceMatcher_OCL_base::L2Dist)
-            radius = 1.f / countFactor / countFactor;
-        else
-            radius = 1.f / countFactor;
+        float radius = 1.f / countFactor;
 
         cv::ocl::BruteForceMatcher_OCL_base matcher(distType);
 
-        // assume support atomic.
-        //if (!supportFeature(devInfo, cv::gpu::GLOBAL_ATOMICS))
-        //{
-        //    try
-        //    {
-        //        std::vector< std::vector<cv::DMatch> > matches;
-        //        matcher.radiusMatch(loadMat(query), loadMat(train), matches, radius);
-        //    }
-        //    catch (const cv::Exception& e)
-        //    {
-        //        ASSERT_EQ(CV_StsNotImplemented, e.code);
-        //    }
-        //}
-        //else
+        std::vector< std::vector<cv::DMatch> > matches;
+        matcher.radiusMatch(cv::ocl::oclMat(query), cv::ocl::oclMat(train), matches, radius);
+
+        ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
+
+        int badCount = 0;
+        for (size_t i = 0; i < matches.size(); i++)
         {
-            std::vector< std::vector<cv::DMatch> > matches;
-            matcher.radiusMatch(cv::ocl::oclMat(query), cv::ocl::oclMat(train), matches, radius);
-
-            ASSERT_EQ(static_cast<size_t>(queryDescCount), matches.size());
-
-            int badCount = 0;
-            for (size_t i = 0; i < matches.size(); i++)
+            if ((int)matches[i].size() != 1)
             {
-                if ((int)matches[i].size() != 1)
-                {
-                    badCount++;
-                }
-                else
-                {
-                    cv::DMatch match = matches[i][0];
-                    if ((match.queryIdx != (int)i) || (match.trainIdx != (int)i * countFactor) || (match.imgIdx != 0))
-                        badCount++;
-                }
+                badCount++;
+            }
+            else
+            {
+                cv::DMatch match = matches[i][0];
+                if ((match.queryIdx != (int)i) || (match.trainIdx != (int)i * countFactor) || (match.imgIdx != 0))
+                    badCount++;
             }
-
-            ASSERT_EQ(0, badCount);
         }
+
+        ASSERT_EQ(0, badCount);
     }
 
-    INSTANTIATE_TEST_CASE_P(GPU_Features2D, BruteForceMatcher, testing::Combine(
-                                //ALL_DEVICES,
-                                testing::Values(DistType(cv::ocl::BruteForceMatcher_OCL_base::L1Dist), DistType(cv::ocl::BruteForceMatcher_OCL_base::L2Dist)),
-                                testing::Values(DescriptorSize(57), DescriptorSize(64), DescriptorSize(83), DescriptorSize(128), DescriptorSize(179), DescriptorSize(256), DescriptorSize(304))));
-
+    INSTANTIATE_TEST_CASE_P(OCL_Features2D, BruteForceMatcher,
+        testing::Combine(
+        testing::Values(
+            DistType(cv::ocl::BruteForceMatcher_OCL_base::L1Dist),
+            DistType(cv::ocl::BruteForceMatcher_OCL_base::L2Dist)/*,
+            DistType(cv::ocl::BruteForceMatcher_OCL_base::HammingDist)*/
+        ),
+        testing::Values(
+            DescriptorSize(57),
+            DescriptorSize(64),
+            DescriptorSize(83),
+            DescriptorSize(128),
+            DescriptorSize(179),
+            DescriptorSize(256),
+            DescriptorSize(304))
+        )
+    );
 } // namespace
 #endif
diff --git a/modules/ocl/test/test_calib3d.cpp b/modules/ocl/test/test_calib3d.cpp
new file mode 100644
index 0000000000..7e5c4a4196
--- /dev/null
+++ b/modules/ocl/test/test_calib3d.cpp
@@ -0,0 +1,196 @@
+///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//     Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+#include <iomanip>
+
+#ifdef HAVE_OPENCL
+
+using namespace cv;
+
+PARAM_TEST_CASE(StereoMatchBM, int, int)
+{
+    int n_disp;
+    int winSize;
+
+    virtual void SetUp()
+    {
+        n_disp  = GET_PARAM(0);
+        winSize = GET_PARAM(1);
+    }
+};
+
+TEST_P(StereoMatchBM, Regression)
+{
+
+    Mat left_image  = readImage("gpu/stereobm/aloe-L.png", IMREAD_GRAYSCALE);
+    Mat right_image = readImage("gpu/stereobm/aloe-R.png", IMREAD_GRAYSCALE);
+    Mat disp_gold   = readImage("gpu/stereobm/aloe-disp.png", IMREAD_GRAYSCALE);
+    ocl::oclMat d_left, d_right;
+    ocl::oclMat d_disp(left_image.size(), CV_8U);
+    Mat  disp;
+
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+    d_left.upload(left_image);
+    d_right.upload(right_image);
+
+    ocl::StereoBM_OCL bm(0, n_disp, winSize);
+
+
+    bm(d_left, d_right, d_disp);
+    d_disp.download(disp);
+
+    EXPECT_MAT_SIMILAR(disp_gold, disp, 1e-3);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Calib3D, StereoMatchBM, testing::Combine(testing::Values(128),
+                                       testing::Values(19)));
+
+PARAM_TEST_CASE(StereoMatchBP, int, int, int, float, float, float, float)
+{
+    int ndisp_;
+    int iters_;
+    int levels_;
+    float max_data_term_;
+    float data_weight_;
+    float max_disc_term_;
+    float disc_single_jump_;
+    virtual void SetUp()
+    {
+        ndisp_          = GET_PARAM(0);
+        iters_          = GET_PARAM(1);
+        levels_         = GET_PARAM(2);
+        max_data_term_  = GET_PARAM(3);
+        data_weight_    = GET_PARAM(4);
+        max_disc_term_     = GET_PARAM(5);
+        disc_single_jump_  = GET_PARAM(6);
+    }
+};
+TEST_P(StereoMatchBP, Regression)
+{
+    Mat left_image  = readImage("gpu/stereobp/aloe-L.png");
+    Mat right_image = readImage("gpu/stereobp/aloe-R.png");
+    Mat disp_gold   = readImage("gpu/stereobp/aloe-disp.png", IMREAD_GRAYSCALE);
+    ocl::oclMat d_left, d_right;
+    ocl::oclMat d_disp;
+    Mat  disp;
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+    d_left.upload(left_image);
+    d_right.upload(right_image);
+    ocl::StereoBeliefPropagation bp(ndisp_, iters_, levels_, max_data_term_, data_weight_,
+        max_disc_term_, disc_single_jump_, CV_16S);
+    bp(d_left, d_right, d_disp);
+    d_disp.download(disp);
+    disp.convertTo(disp, disp_gold.depth());
+    EXPECT_MAT_NEAR(disp_gold, disp, 0.0);
+}
+INSTANTIATE_TEST_CASE_P(OCL_Calib3D, StereoMatchBP, testing::Combine(testing::Values(64),
+    testing::Values(8),testing::Values(2),testing::Values(25.0f),
+    testing::Values(0.1f),testing::Values(15.0f),testing::Values(1.0f)));
+
+//////////////////////////////////////////////////////////////////////////
+//  ConstSpaceBeliefPropagation
+PARAM_TEST_CASE(StereoMatchConstSpaceBP, int, int, int, int, float, float, float, float, int, int)
+{
+    int ndisp_;
+    int iters_;
+    int levels_;
+    int nr_plane_;
+    float max_data_term_;
+    float data_weight_;
+    float max_disc_term_;
+    float disc_single_jump_;
+    int min_disp_th_;
+    int msg_type_;
+
+    virtual void SetUp()
+    {
+        ndisp_          = GET_PARAM(0);
+        iters_          = GET_PARAM(1);
+        levels_         = GET_PARAM(2);
+        nr_plane_ = GET_PARAM(3);
+        max_data_term_  = GET_PARAM(4);
+        data_weight_    = GET_PARAM(5);
+        max_disc_term_     = GET_PARAM(6);
+        disc_single_jump_  = GET_PARAM(7);
+        min_disp_th_ = GET_PARAM(8);
+        msg_type_  = GET_PARAM(9);
+    }
+};
+TEST_P(StereoMatchConstSpaceBP, Regression)
+{
+    Mat left_image  = readImage("gpu/csstereobp/aloe-L.png");
+    Mat right_image = readImage("gpu/csstereobp/aloe-R.png");
+    Mat disp_gold   = readImage("gpu/csstereobp/aloe-disp.png", IMREAD_GRAYSCALE);
+
+    ocl::oclMat d_left, d_right;
+    ocl::oclMat d_disp;
+
+    Mat  disp;
+    ASSERT_FALSE(left_image.empty());
+    ASSERT_FALSE(right_image.empty());
+    ASSERT_FALSE(disp_gold.empty());
+
+    d_left.upload(left_image);
+    d_right.upload(right_image);
+
+    ocl::StereoConstantSpaceBP bp(ndisp_, iters_, levels_, nr_plane_, max_data_term_, data_weight_,
+        max_disc_term_, disc_single_jump_, 0, CV_32F);
+    bp(d_left, d_right, d_disp);
+    d_disp.download(disp);
+    disp.convertTo(disp, disp_gold.depth());
+
+    EXPECT_MAT_SIMILAR(disp_gold, disp, 1e-4);
+    //EXPECT_MAT_NEAR(disp_gold, disp, 1.0, "");
+}
+INSTANTIATE_TEST_CASE_P(OCL_Calib3D, StereoMatchConstSpaceBP, testing::Combine(testing::Values(128),
+    testing::Values(16),testing::Values(4), testing::Values(4), testing::Values(30.0f),
+    testing::Values(1.0f),testing::Values(160.0f),
+    testing::Values(10.0f), testing::Values(0), testing::Values(CV_32F)));
+#endif // HAVE_OPENCL
diff --git a/modules/ocl/test/test_canny.cpp b/modules/ocl/test/test_canny.cpp
index cac6b66f51..5d8d108e70 100644
--- a/modules/ocl/test/test_canny.cpp
+++ b/modules/ocl/test/test_canny.cpp
@@ -43,13 +43,11 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #ifdef HAVE_OPENCL
-#define SHOW_RESULT 0
 
 ////////////////////////////////////////////////////////
 // Canny
-extern std::string workdir;
 IMPLEMENT_PARAM_CLASS(AppertureSize, int);
 IMPLEMENT_PARAM_CLASS(L2gradient, bool);
 
@@ -59,50 +57,33 @@ PARAM_TEST_CASE(Canny, AppertureSize, L2gradient)
     bool useL2gradient;
 
     cv::Mat edges_gold;
-    //std::vector<cv::ocl::Info> oclinfo;
     virtual void SetUp()
     {
         apperture_size = GET_PARAM(0);
         useL2gradient = GET_PARAM(1);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
     }
 };
 
 TEST_P(Canny, Accuracy)
 {
-    cv::Mat img = readImage(workdir + "fruits.jpg", cv::IMREAD_GRAYSCALE);
+    cv::Mat img = readImage("cv/shared/fruits.png", cv::IMREAD_GRAYSCALE);
     ASSERT_FALSE(img.empty());
 
     double low_thresh = 50.0;
     double high_thresh = 100.0;
 
-    cv::resize(img, img, cv::Size(512, 384));
     cv::ocl::oclMat ocl_img = cv::ocl::oclMat(img);
 
     cv::ocl::oclMat edges;
     cv::ocl::Canny(ocl_img, edges, low_thresh, high_thresh, apperture_size, useL2gradient);
 
-    char filename [100];
-    sprintf(filename, "G:/Valve_edges_a%d_L2Grad%d.jpg", apperture_size, (int)useL2gradient);
-
     cv::Mat edges_gold;
     cv::Canny(img, edges_gold, low_thresh, high_thresh, apperture_size, useL2gradient);
 
-#if SHOW_RESULT
-    cv::Mat edges_x2, ocl_edges(edges);
-    edges_x2.create(edges.rows, edges.cols * 2, edges.type());
-    edges_x2.setTo(0);
-    cv::add(edges_gold, cv::Mat(edges_x2, cv::Rect(0, 0, edges_gold.cols, edges_gold.rows)), cv::Mat(edges_x2, cv::Rect(0, 0, edges_gold.cols, edges_gold.rows)));
-    cv::add(ocl_edges, cv::Mat(edges_x2, cv::Rect(edges_gold.cols, 0, edges_gold.cols, edges_gold.rows)), cv::Mat(edges_x2, cv::Rect(edges_gold.cols, 0, edges_gold.cols, edges_gold.rows)));
-    cv::namedWindow("Canny result (left: cpu, right: ocl)");
-    cv::imshow("Canny result (left: cpu, right: ocl)", edges_x2);
-    cv::waitKey();
-#endif //OUTPUT_RESULT
     EXPECT_MAT_SIMILAR(edges_gold, edges, 1e-2);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, Canny, testing::Combine(
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, Canny, testing::Combine(
                             testing::Values(AppertureSize(3), AppertureSize(5)),
                             testing::Values(L2gradient(false), L2gradient(true))));
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/ocl/test/test_color.cpp b/modules/ocl/test/test_color.cpp
index 202967b7a6..b0a223091a 100644
--- a/modules/ocl/test/test_color.cpp
+++ b/modules/ocl/test/test_color.cpp
@@ -43,7 +43,10 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
+
+using namespace cv;
+
 #ifdef HAVE_OPENCL
 
 //#define MAT_DEBUG
@@ -100,7 +103,7 @@ PARAM_TEST_CASE(CvtColor, cv::Size, MatDepth)
     cv::cvtColor(src, dst_gold, CVTCODE(name));\
     cv::Mat dst_mat;\
     dst.download(dst_mat);\
-    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5, "");\
+    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5);\
 }
 
 //add new ones here using macro
@@ -141,7 +144,7 @@ TEST_P(CvtColor_Gray2RGB, Accuracy)
     cv::cvtColor(src, dst_gold, code);
     cv::Mat dst_mat;
     dst.download(dst_mat);
-    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5, "");
+    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5);
 }
 
 
@@ -171,7 +174,7 @@ TEST_P(CvtColor_YUV420, Accuracy)
     cv::Mat dst_mat;
     dst.download(dst_mat);
     MAT_DIFF(dst_mat, dst_gold);
-    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5, "");
+    EXPECT_MAT_NEAR(dst_gold, dst_mat, 1e-5);
 }
 
 INSTANTIATE_TEST_CASE_P(OCL_ImgProc, CvtColor, testing::Combine(
@@ -181,13 +184,13 @@ INSTANTIATE_TEST_CASE_P(OCL_ImgProc, CvtColor, testing::Combine(
 
 INSTANTIATE_TEST_CASE_P(OCL_ImgProc, CvtColor_YUV420, testing::Combine(
                             testing::Values(cv::Size(128, 45), cv::Size(46, 132), cv::Size(1024, 1023)),
-                            testing::Values((int)CV_YUV2RGBA_NV12, (int)CV_YUV2BGRA_NV12, (int)CV_YUV2RGB_NV12, (int)CV_YUV2BGR_NV12)
+                            testing::Values((int)COLOR_YUV2RGBA_NV12, (int)COLOR_YUV2BGRA_NV12, (int)COLOR_YUV2RGB_NV12, (int)COLOR_YUV2BGR_NV12)
                         ));
 
 INSTANTIATE_TEST_CASE_P(OCL_ImgProc, CvtColor_Gray2RGB, testing::Combine(
                             DIFFERENT_SIZES,
                             testing::Values(MatDepth(CV_8U), MatDepth(CV_16U), MatDepth(CV_32F)),
-                            testing::Values((int)CV_GRAY2BGR, (int)CV_GRAY2BGRA, (int)CV_GRAY2RGB, (int)CV_GRAY2RGBA)
+                            testing::Values((int)COLOR_GRAY2BGR, (int)COLOR_GRAY2BGRA, (int)COLOR_GRAY2RGB, (int)COLOR_GRAY2RGBA)
                         ));
 }
 #endif
diff --git a/modules/ocl/test/test_fft.cpp b/modules/ocl/test/test_fft.cpp
index 0fee8b03d4..29a2b815a3 100644
--- a/modules/ocl/test/test_fft.cpp
+++ b/modules/ocl/test/test_fft.cpp
@@ -43,7 +43,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 using namespace std;
 #ifdef HAVE_CLAMDFFT
 ////////////////////////////////////////////////////////////////////////////
@@ -68,7 +68,7 @@ TEST_P(Dft, C2C)
 
     cv::dft(a, b_gold, dft_flags);
     cv::ocl::dft(cv::ocl::oclMat(a), d_b, a.size(), dft_flags);
-    EXPECT_MAT_NEAR(b_gold, cv::Mat(d_b), a.size().area() * 1e-4, "");
+    EXPECT_MAT_NEAR(b_gold, cv::Mat(d_b), a.size().area() * 1e-4);
 }
 
 TEST_P(Dft, R2C)
@@ -81,11 +81,11 @@ TEST_P(Dft, R2C)
     cv::dft(a, b_gold, cv::DFT_COMPLEX_OUTPUT | dft_flags);
 
     b_gold_roi = b_gold(cv::Rect(0, 0, d_b.cols, d_b.rows));
-    EXPECT_MAT_NEAR(b_gold_roi, cv::Mat(d_b), a.size().area() * 1e-4, "");
+    EXPECT_MAT_NEAR(b_gold_roi, cv::Mat(d_b), a.size().area() * 1e-4);
 
     cv::Mat c_gold;
     cv::dft(b_gold, c_gold, cv::DFT_INVERSE | cv::DFT_REAL_OUTPUT | cv::DFT_SCALE);
-    EXPECT_MAT_NEAR(b_gold_roi, cv::Mat(d_b), a.size().area() * 1e-4, "");
+    EXPECT_MAT_NEAR(b_gold_roi, cv::Mat(d_b), a.size().area() * 1e-4);
 }
 
 TEST_P(Dft, R2CthenC2R)
@@ -95,7 +95,7 @@ TEST_P(Dft, R2CthenC2R)
     cv::ocl::oclMat d_b, d_c;
     cv::ocl::dft(cv::ocl::oclMat(a), d_b, a.size(), 0);
     cv::ocl::dft(d_b, d_c, a.size(), cv::DFT_SCALE | cv::DFT_INVERSE | cv::DFT_REAL_OUTPUT);
-    EXPECT_MAT_NEAR(a, d_c, a.size().area() * 1e-4, "");
+    EXPECT_MAT_NEAR(a, d_c, a.size().area() * 1e-4);
 }
 
 
@@ -103,4 +103,138 @@ INSTANTIATE_TEST_CASE_P(OCL_ImgProc, Dft, testing::Combine(
                             testing::Values(cv::Size(2, 3), cv::Size(5, 4), cv::Size(25, 20), cv::Size(512, 1), cv::Size(1024, 768)),
                             testing::Values(0, (int)cv::DFT_ROWS, (int)cv::DFT_SCALE) ));
 
+////////////////////////////////////////////////////////////////////////////
+// MulSpectrums
+
+PARAM_TEST_CASE(MulSpectrums, cv::Size, DftFlags, bool)
+{
+    cv::Size size;
+    int flag;
+    bool ccorr;
+    cv::Mat a, b;
+
+    virtual void SetUp()
+    {
+        size  = GET_PARAM(0);
+        flag  = GET_PARAM(1);
+        ccorr = GET_PARAM(2);
+
+        a = randomMat(size, CV_32FC2);
+        b = randomMat(size, CV_32FC2);
+    }
+};
+
+TEST_P(MulSpectrums, Simple)
+{
+    cv::ocl::oclMat c;
+    cv::ocl::mulSpectrums(cv::ocl::oclMat(a), cv::ocl::oclMat(b), c, flag, 1.0, ccorr);
+
+    cv::Mat c_gold;
+    cv::mulSpectrums(a, b, c_gold, flag, ccorr);
+
+    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
+}
+
+TEST_P(MulSpectrums, Scaled)
+{
+    float scale = 1.f / size.area();
+
+    cv::ocl::oclMat c;
+    cv::ocl::mulSpectrums(cv::ocl::oclMat(a), cv::ocl::oclMat(b), c, flag, scale, ccorr);
+
+    cv::Mat c_gold;
+    cv::mulSpectrums(a, b, c_gold, flag, ccorr);
+    c_gold.convertTo(c_gold, c_gold.type(), scale);
+
+    EXPECT_MAT_NEAR(c_gold, c, 1e-2);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, MulSpectrums, testing::Combine(
+    DIFFERENT_SIZES,
+    testing::Values(DftFlags(0)),
+    testing::Values(false, true)));
+
+
+////////////////////////////////////////////////////////
+// Convolve
+
+void static convolveDFT(const cv::Mat& A, const cv::Mat& B, cv::Mat& C, bool ccorr = false)
+{
+    // reallocate the output array if needed
+    C.create(std::abs(A.rows - B.rows) + 1, std::abs(A.cols - B.cols) + 1, A.type());
+    cv::Size dftSize;
+
+    // compute the size of DFT transform
+    dftSize.width = cv::getOptimalDFTSize(A.cols + B.cols - 1);
+    dftSize.height = cv::getOptimalDFTSize(A.rows + B.rows - 1);
+
+    // allocate temporary buffers and initialize them with 0s
+    cv::Mat tempA(dftSize, A.type(), cv::Scalar::all(0));
+    cv::Mat tempB(dftSize, B.type(), cv::Scalar::all(0));
+
+    // copy A and B to the top-left corners of tempA and tempB, respectively
+    cv::Mat roiA(tempA, cv::Rect(0, 0, A.cols, A.rows));
+    A.copyTo(roiA);
+    cv::Mat roiB(tempB, cv::Rect(0, 0, B.cols, B.rows));
+    B.copyTo(roiB);
+
+    // now transform the padded A & B in-place;
+    // use "nonzeroRows" hint for faster processing
+    cv::dft(tempA, tempA, 0, A.rows);
+    cv::dft(tempB, tempB, 0, B.rows);
+
+    // multiply the spectrums;
+    // the function handles packed spectrum representations well
+    cv::mulSpectrums(tempA, tempB, tempA, 0, ccorr);
+
+    // transform the product back from the frequency domain.
+    // Even though all the result rows will be non-zero,
+    // you need only the first C.rows of them, and thus you
+    // pass nonzeroRows == C.rows
+    cv::dft(tempA, tempA, cv::DFT_INVERSE + cv::DFT_SCALE, C.rows);
+
+    // now copy the result back to C.
+    tempA(cv::Rect(0, 0, C.cols, C.rows)).copyTo(C);
+}
+
+IMPLEMENT_PARAM_CLASS(KSize, int);
+IMPLEMENT_PARAM_CLASS(Ccorr, bool);
+
+PARAM_TEST_CASE(Convolve_DFT, cv::Size, KSize, Ccorr)
+{
+    cv::Size size;
+    int ksize;
+    bool ccorr;
+
+    cv::Mat src;
+    cv::Mat kernel;
+
+    cv::Mat dst_gold;
+
+    virtual void SetUp()
+    {
+        size  = GET_PARAM(0);
+        ksize = GET_PARAM(1);
+        ccorr = GET_PARAM(2);
+    }
+};
+
+TEST_P(Convolve_DFT, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_32FC1, 0.0, 100.0);
+    cv::Mat kernel = randomMat(cv::Size(ksize, ksize), CV_32FC1, 0.0, 1.0);
+
+    cv::ocl::oclMat dst;
+    cv::ocl::convolve(cv::ocl::oclMat(src), cv::ocl::oclMat(kernel), dst, ccorr);
+
+    cv::Mat dst_gold;
+    convolveDFT(src, kernel, dst_gold, ccorr);
+
+    EXPECT_MAT_NEAR(dst, dst_gold, 1e-1);
+}
+#define DIFFERENT_CONVOLVE_SIZES testing::Values(cv::Size(251, 257), cv::Size(113, 113), cv::Size(200, 480), cv::Size(1300, 1300))
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, Convolve_DFT, testing::Combine(
+    DIFFERENT_CONVOLVE_SIZES,
+    testing::Values(KSize(19), KSize(23), KSize(45)),
+    testing::Values(Ccorr(true)/*, Ccorr(false)*/))); // false ccorr cannot pass for some instances
 #endif // HAVE_CLAMDFFT
diff --git a/modules/ocl/test/test_filters.cpp b/modules/ocl/test/test_filters.cpp
index 70d4079445..4a22ec5033 100644
--- a/modules/ocl/test/test_filters.cpp
+++ b/modules/ocl/test/test_filters.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
@@ -19,6 +20,7 @@
 //    Jia Haipeng, jiahaipeng95@gmail.com
 //    Zero Lin, Zero.Lin@amd.com
 //    Zhang Ying, zhangying913@gmail.com
+//    Yao Wang, bitwangyaoyao@gmail.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -46,7 +48,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -55,121 +57,13 @@ using namespace testing;
 using namespace std;
 
 
-PARAM_TEST_CASE(FilterTestBase, MatType, bool)
+PARAM_TEST_CASE(FilterTestBase,
+                MatType,
+                cv::Size, // kernel size
+                cv::Size, // dx,dy
+                int       // border type, or iteration
+                )
 {
-    int type;
-    cv::Scalar val;
-
-    //src mat
-    cv::Mat mat1;
-    cv::Mat mat2;
-    cv::Mat mask;
-    cv::Mat dst;
-    cv::Mat dst1; //bak, for two outputs
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int src2x;
-    int src2y;
-    int dstx;
-    int dsty;
-    int maskx;
-    int masky;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat mat2_roi;
-    cv::Mat mask_roi;
-    cv::Mat dst_roi;
-    cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-    cv::ocl::oclMat gdst1_whole; //bak
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gmat2;
-    cv::ocl::oclMat gdst;
-    cv::ocl::oclMat gdst1;   //bak
-    cv::ocl::oclMat gmask;
-
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        mat2 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        dst1  = randomMat(rng, size, type, 5, 16, false);
-        mask = randomMat(rng, size, CV_8UC1, 0, 2,  false);
-
-        cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
-
-        val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
-    }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(1, mat1.cols);
-        roirows = rng.uniform(1, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        src2x   = rng.uniform(0, mat2.cols - roicols);
-        src2y   = rng.uniform(0, mat2.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-        maskx   = rng.uniform(0, mask.cols - roicols);
-        masky   = rng.uniform(0, mask.rows - roirows);
-#else
-        roicols = mat1.cols;
-        roirows = mat1.rows;
-        src1x = 0;
-        src1y = 0;
-        src2x = 0;
-        src2y = 0;
-        dstx = 0;
-        dsty = 0;
-        maskx = 0;
-        masky = 0;
-#endif
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        mat2_roi = mat2(Rect(src2x, src2y, roicols, roirows));
-        mask_roi = mask(Rect(maskx, masky, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-        dst1_roi = dst1(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gdst1_whole = dst1;
-        gdst1 = gdst1_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gmat1 = mat1_roi;
-        gmat2 = mat2_roi;
-        gmask = mask_roi;
-    }
-
-};
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// blur
-
-PARAM_TEST_CASE(Blur, MatType, cv::Size, int)
-{
-    int type;
-    cv::Size ksize;
-    int bordertype;
-
     //src mat
     cv::Mat mat1;
     cv::Mat dst;
@@ -185,7 +79,7 @@ PARAM_TEST_CASE(Blur, MatType, cv::Size, int)
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -193,23 +87,6 @@ PARAM_TEST_CASE(Blur, MatType, cv::Size, int)
     cv::ocl::oclMat gmat1;
     cv::ocl::oclMat gdst;
 
-    virtual void SetUp()
-    {
-        type = GET_PARAM(0);
-        ksize = GET_PARAM(1);
-        bordertype = GET_PARAM(2);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-    }
-
     void random_roi()
     {
 #ifdef RANDOMROI
@@ -236,10 +113,37 @@ PARAM_TEST_CASE(Blur, MatType, cv::Size, int)
         gdst_whole = dst;
         gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
 
-
         gmat1 = mat1_roi;
     }
 
+    void Init(int mat_type)
+    {
+        cv::Size size(MWIDTH, MHEIGHT);
+        mat1 = randomMat(size, mat_type, 5, 16);
+        dst  = randomMat(size, mat_type, 5, 16);
+    }
+
+    void Near(double threshold)
+    {
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), threshold);
+    }
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// blur
+struct Blur : FilterTestBase
+{
+    int type;
+    cv::Size ksize;
+    int bordertype;
+
+    virtual void SetUp()
+    {
+        type = GET_PARAM(0);
+        ksize = GET_PARAM(1);
+        bordertype = GET_PARAM(3);
+        Init(type);
+    }
 };
 
 TEST_P(Blur, Mat)
@@ -247,116 +151,36 @@ TEST_P(Blur, Mat)
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
-
         cv::blur(mat1_roi, dst_roi, ksize, Point(-1, -1), bordertype);
         cv::ocl::blur(gmat1, gdst, ksize, Point(-1, -1), bordertype);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
+        Near(1.0);
     }
-
 }
 
 
-
 /////////////////////////////////////////////////////////////////////////////////////////////////
 //Laplacian
-
-PARAM_TEST_CASE(LaplacianTestBase, MatType, int)
+struct Laplacian : FilterTestBase
 {
     int type;
-    int ksize;
-
-    //src mat
-    cv::Mat mat;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int srcx;
-    int srcy;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat;
-    cv::ocl::oclMat gdst;
+    cv::Size ksize;
 
     virtual void SetUp()
     {
         type = GET_PARAM(0);
         ksize = GET_PARAM(1);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat  = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-    }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(2, mat.cols);
-        roirows = rng.uniform(2, mat.rows);
-        srcx   = rng.uniform(0, mat.cols - roicols);
-        srcy   = rng.uniform(0, mat.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-#else
-        roicols = mat.cols;
-        roirows = mat.rows;
-        srcx = 0;
-        srcy = 0;
-        dstx = 0;
-        dsty = 0;
-#endif
-
-        mat_roi  = mat(Rect(srcx, srcy, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-        gmat = mat_roi;
+        Init(type);
     }
 };
 
-struct Laplacian : LaplacianTestBase {};
-
 TEST_P(Laplacian, Accuracy)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
-
-        cv::Laplacian(mat_roi, dst_roi, -1, ksize, 1);
-        cv::ocl::Laplacian(gmat, gdst, -1, ksize, 1);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, srcx, srcy, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        cv::Laplacian(mat1_roi, dst_roi, -1, ksize.width, 1);
+        cv::ocl::Laplacian(gmat1, gdst, -1, ksize.width, 1);
+        Near(1e-5);
     }
 }
 
@@ -364,8 +188,7 @@ TEST_P(Laplacian, Accuracy)
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // erode & dilate
-
-PARAM_TEST_CASE(ErodeDilateBase, MatType, int)
+struct ErodeDilate : FilterTestBase
 {
     int type;
     int iterations;
@@ -373,210 +196,54 @@ PARAM_TEST_CASE(ErodeDilateBase, MatType, int)
     //erode or dilate kernel
     cv::Mat kernel;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
     virtual void SetUp()
     {
         type = GET_PARAM(0);
-        iterations = GET_PARAM(1);
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
+        iterations = GET_PARAM(3);
+        Init(type);
         //		rng.fill(kernel, cv::RNG::UNIFORM, cv::Scalar::all(0), cv::Scalar::all(3));
-        kernel = randomMat(rng, Size(3, 3), CV_8UC1, 0, 3, false);
-
-    }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(2, mat1.cols);
-        roirows = rng.uniform(2, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-#else
-        roicols = mat1.cols;
-        roirows = mat1.rows;
-        src1x = 0;
-        src1y = 0;
-        dstx = 0;
-        dsty = 0;
-#endif
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-
-        gmat1 = mat1_roi;
+        kernel = randomMat(Size(3, 3), CV_8UC1, 0, 3);
     }
 
 };
 
-// erode
-
-struct Erode : ErodeDilateBase {};
-
-TEST_P(Erode, Mat)
-{
-    for(int j = 0; j < LOOP_TIMES; j++)
-    {
-        random_roi();
-
-        cv::erode(mat1_roi, dst_roi, kernel, Point(-1, -1), iterations);
-        cv::ocl::erode(gmat1, gdst, kernel, Point(-1, -1), iterations);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
-    }
-
-}
-
-
-
-
-
-// dilate
-
-struct Dilate : ErodeDilateBase {};
-
-TEST_P(Dilate, Mat)
+TEST_P(ErodeDilate, Mat)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
         cv::erode(mat1_roi, dst_roi, kernel, Point(-1, -1), iterations);
         cv::ocl::erode(gmat1, gdst, kernel, Point(-1, -1), iterations);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5, sss);
+        Near(1e-5);
+    }
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        random_roi();
+        cv::dilate(mat1_roi, dst_roi, kernel, Point(-1, -1), iterations);
+        cv::ocl::dilate(gmat1, gdst, kernel, Point(-1, -1), iterations);
+        Near(1e-5);
     }
-
 }
 
 
-
-
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // Sobel
-
-PARAM_TEST_CASE(Sobel, MatType, int, int, int, int)
+struct Sobel : FilterTestBase
 {
     int type;
     int dx, dy, ksize, bordertype;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
     virtual void SetUp()
     {
         type = GET_PARAM(0);
-        dx = GET_PARAM(1);
-        dy = GET_PARAM(2);
-        ksize = GET_PARAM(3);
-        bordertype = GET_PARAM(4);
-
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
+        Size s = GET_PARAM(1);
+        ksize = s.width;
+        s = GET_PARAM(2);
+        dx = s.width;
+        dy = s.height;
+        bordertype = GET_PARAM(3);
+        Init(type);
     }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(2, mat1.cols);
-        roirows = rng.uniform(2, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-#else
-        roicols = mat1.cols;
-        roirows = mat1.rows;
-        src1x = 0;
-        src1y = 0;
-        dstx = 0;
-        dsty = 0;
-#endif
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-
-        gmat1 = mat1_roi;
-    }
-
 };
 
 TEST_P(Sobel, Mat)
@@ -584,103 +251,29 @@ TEST_P(Sobel, Mat)
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
-
         cv::Sobel(mat1_roi, dst_roi, -1, dx, dy, ksize, /*scale*/0.00001,/*delta*/0, bordertype);
         cv::ocl::Sobel(gmat1, gdst, -1, dx, dy, ksize,/*scale*/0.00001,/*delta*/0, bordertype);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        Near(1);
     }
-
 }
 
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // Scharr
-
-PARAM_TEST_CASE(Scharr, MatType, int, int, int)
+struct Scharr : FilterTestBase
 {
     int type;
     int dx, dy, bordertype;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
     virtual void SetUp()
     {
         type = GET_PARAM(0);
-        dx = GET_PARAM(1);
-        dy = GET_PARAM(2);
+        Size s = GET_PARAM(2);
+        dx = s.width;
+        dy = s.height;
         bordertype = GET_PARAM(3);
-        dx = 1;
-        dy = 0;
-
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
+        Init(type);
     }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(2, mat1.cols);
-        roirows = rng.uniform(2, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-#else
-        roicols = mat1.cols;
-        roirows = mat1.rows;
-        src1x = 0;
-        src1y = 0;
-        dstx = 0;
-        dsty = 0;
-#endif
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-
-        gmat1 = mat1_roi;
-    }
-
 };
 
 TEST_P(Scharr, Mat)
@@ -688,16 +281,9 @@ TEST_P(Scharr, Mat)
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
-
         cv::Scharr(mat1_roi, dst_roi, -1, dx, dy, /*scale*/1,/*delta*/0, bordertype);
         cv::ocl::Scharr(gmat1, gdst, -1, dx, dy,/*scale*/1,/*delta*/0, bordertype);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        Near(1);
     }
 
 }
@@ -705,89 +291,23 @@ TEST_P(Scharr, Mat)
 
 /////////////////////////////////////////////////////////////////////////////////////////////////
 // GaussianBlur
-
-PARAM_TEST_CASE(GaussianBlur, MatType, cv::Size, int)
+struct GaussianBlur : FilterTestBase
 {
     int type;
     cv::Size ksize;
     int bordertype;
-
     double sigma1, sigma2;
 
-    //src mat
-    cv::Mat mat1;
-    cv::Mat dst;
-
-    // set up roi
-    int roicols;
-    int roirows;
-    int src1x;
-    int src1y;
-    int dstx;
-    int dsty;
-
-    //src mat with roi
-    cv::Mat mat1_roi;
-    cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
-    //ocl dst mat for testing
-    cv::ocl::oclMat gdst_whole;
-
-    //ocl mat with roi
-    cv::ocl::oclMat gmat1;
-    cv::ocl::oclMat gdst;
-
     virtual void SetUp()
     {
         type = GET_PARAM(0);
         ksize = GET_PARAM(1);
-        bordertype = GET_PARAM(2);
-
+        bordertype = GET_PARAM(3);
+        Init(type);
         cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Size size(MWIDTH, MHEIGHT);
-
         sigma1 = rng.uniform(0.1, 1.0);
         sigma2 = rng.uniform(0.1, 1.0);
-
-        mat1 = randomMat(rng, size, type, 5, 16, false);
-        dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
-
-    void random_roi()
-    {
-#ifdef RANDOMROI
-        //randomize ROI
-        cv::RNG &rng = TS::ptr()->get_rng();
-        roicols = rng.uniform(2, mat1.cols);
-        roirows = rng.uniform(2, mat1.rows);
-        src1x   = rng.uniform(0, mat1.cols - roicols);
-        src1y   = rng.uniform(0, mat1.rows - roirows);
-        dstx    = rng.uniform(0, dst.cols  - roicols);
-        dsty    = rng.uniform(0, dst.rows  - roirows);
-#else
-        roicols = mat1.cols;
-        roirows = mat1.rows;
-        src1x = 0;
-        src1y = 0;
-        dstx = 0;
-        dsty = 0;
-#endif
-
-        mat1_roi = mat1(Rect(src1x, src1y, roicols, roirows));
-        dst_roi  = dst(Rect(dstx, dsty, roicols, roirows));
-
-        gdst_whole = dst;
-        gdst = gdst_whole(Rect(dstx, dsty, roicols, roirows));
-
-
-        gmat1 = mat1_roi;
-    }
-
 };
 
 TEST_P(GaussianBlur, Mat)
@@ -795,54 +315,165 @@ TEST_P(GaussianBlur, Mat)
     for(int j = 0; j < LOOP_TIMES; j++)
     {
         random_roi();
-
         cv::GaussianBlur(mat1_roi, dst_roi, ksize, sigma1, sigma2, bordertype);
         cv::ocl::GaussianBlur(gmat1, gdst, ksize, sigma1, sigma2, bordertype);
-
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
+        Near(1);
     }
 
 }
 
 
 
-INSTANTIATE_TEST_CASE_P(Filter, Blur, Combine(Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4),
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Filter2D
+struct Filter2D : FilterTestBase
+{
+    int type;
+    cv::Size ksize;
+    int bordertype;
+    Point anchor;
+    virtual void SetUp()
+    {
+        type = GET_PARAM(0);
+        ksize = GET_PARAM(1);
+        bordertype = GET_PARAM(3);
+        Init(type);
+        anchor = Point(-1,-1);
+    }
+};
+
+TEST_P(Filter2D, Mat)
+{
+    cv::Mat kernel = randomMat(cv::Size(ksize.width, ksize.height), CV_32FC1, 0.0, 1.0);
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        random_roi();
+        cv::filter2D(mat1_roi, dst_roi, -1, kernel, anchor, 0.0, bordertype);
+        cv::ocl::filter2D(gmat1, gdst, -1, kernel, anchor, bordertype);
+        Near(1);
+    }
+}
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Bilateral
+struct Bilateral : FilterTestBase
+{
+    int type;
+    cv::Size ksize;
+    int bordertype;
+    double sigmacolor, sigmaspace;
+
+    virtual void SetUp()
+    {
+        type = GET_PARAM(0);
+        ksize = GET_PARAM(1);
+        bordertype = GET_PARAM(3);
+        Init(type);
+        cv::RNG &rng = TS::ptr()->get_rng();
+        sigmacolor = rng.uniform(20, 100);
+        sigmaspace = rng.uniform(10, 40);
+    }
+};
+
+TEST_P(Bilateral, Mat)
+{
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        random_roi();
+        cv::bilateralFilter(mat1_roi, dst_roi, ksize.width, sigmacolor, sigmaspace, bordertype);
+        cv::ocl::bilateralFilter(gmat1, gdst, ksize.width, sigmacolor, sigmaspace, bordertype);
+        Near(1);
+    }
+
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// AdaptiveBilateral
+struct AdaptiveBilateral : FilterTestBase
+{
+    int type;
+    cv::Size ksize;
+    int bordertype;
+    Point anchor;
+    virtual void SetUp()
+    {
+        type = GET_PARAM(0);
+        ksize = GET_PARAM(1);
+        bordertype = GET_PARAM(3);
+        Init(type);
+        anchor = Point(-1,-1);
+    }
+};
+
+TEST_P(AdaptiveBilateral, Mat)
+{
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        random_roi();
+        cv::adaptiveBilateralFilter(mat1_roi, dst_roi, ksize, 5, anchor, bordertype);
+        cv::ocl::adaptiveBilateralFilter(gmat1, gdst, ksize, 5, anchor, bordertype);
+        Near(1);
+    }
+
+}
+
+INSTANTIATE_TEST_CASE_P(Filter, Blur, Combine(
+                        Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4),
                         Values(cv::Size(3, 3), cv::Size(5, 5), cv::Size(7, 7)),
+                        Values(Size(0, 0)), //not use
                         Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE, (MatType)cv::BORDER_REFLECT, (MatType)cv::BORDER_REFLECT_101)));
 
 
-INSTANTIATE_TEST_CASE_P(Filters, Laplacian, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
-                            Values(1, 3)));
+INSTANTIATE_TEST_CASE_P(Filter, Laplacian, Combine(
+                        Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
+                        Values(Size(3, 3)),
+                        Values(Size(0, 0)), //not use
+                        Values(0)));        //not use
 
-INSTANTIATE_TEST_CASE_P(Filter, Erode, Combine(Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1)));
-
-//INSTANTIATE_TEST_CASE_P(Filter, Erode, Combine(Values(CV_8UC1, CV_8UC1), Values(false)));
-
-INSTANTIATE_TEST_CASE_P(Filter, Dilate, Combine(Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4), Values(1)));
-
-//INSTANTIATE_TEST_CASE_P(Filter, Dilate, Combine(Values(CV_8UC1, CV_8UC1), Values(false)));
+INSTANTIATE_TEST_CASE_P(Filter, ErodeDilate, Combine(
+                        Values(CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4),
+                        Values(Size(0, 0)), //not use
+                        Values(Size(0, 0)), //not use
+                        Values(1)));
 
 
-INSTANTIATE_TEST_CASE_P(Filter, Sobel, Combine(Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
-                        Values(1, 2), Values(0, 1), Values(3, 5), Values((MatType)cv::BORDER_CONSTANT,
-                                (MatType)cv::BORDER_REPLICATE)));
+INSTANTIATE_TEST_CASE_P(Filter, Sobel, Combine(
+                        Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC3, CV_32FC4),
+                        Values(Size(3, 3), Size(5, 5)),
+                        Values(Size(1, 0), Size(1, 1), Size(2, 0), Size(2, 1)),
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
 
 
 INSTANTIATE_TEST_CASE_P(Filter, Scharr, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4), Values(0, 1), Values(0, 1),
-                            Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
+                        Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4),
+                        Values(Size(0, 0)), //not use
+                        Values(Size(0, 1), Size(1, 0)),
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
 
 INSTANTIATE_TEST_CASE_P(Filter, GaussianBlur, Combine(
-                            Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4),
-                            Values(cv::Size(3, 3), cv::Size(5, 5)),
-                            Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
+                        Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32FC1, CV_32FC4),
+                        Values(Size(3, 3), Size(5, 5)),
+                        Values(Size(0, 0)), //not use
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE)));
 
 
 
+INSTANTIATE_TEST_CASE_P(Filter, Filter2D, testing::Combine(
+                        Values(CV_8UC1, CV_32FC1, CV_32FC4),
+                        Values(Size(3, 3), Size(15, 15), Size(25, 25)),
+                        Values(Size(0, 0)), //not use
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REFLECT101, (MatType)cv::BORDER_REPLICATE, (MatType)cv::BORDER_REFLECT)));
+
+INSTANTIATE_TEST_CASE_P(Filter, Bilateral, Combine(
+                        Values(CV_8UC1, CV_8UC3),
+                        Values(Size(5, 5), Size(9, 9)),
+                        Values(Size(0, 0)), //not use
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE,
+                               (MatType)cv::BORDER_REFLECT, (MatType)cv::BORDER_WRAP, (MatType)cv::BORDER_REFLECT_101)));
+
+INSTANTIATE_TEST_CASE_P(Filter, AdaptiveBilateral, Combine(
+                        Values(CV_8UC1, CV_8UC3),
+                        Values(Size(5, 5), Size(9, 9)),
+                        Values(Size(0, 0)), //not use
+                        Values((MatType)cv::BORDER_CONSTANT, (MatType)cv::BORDER_REPLICATE,
+                               (MatType)cv::BORDER_REFLECT,  (MatType)cv::BORDER_REFLECT_101)));
 #endif // HAVE_OPENCL
diff --git a/modules/ocl/test/test_gemm.cpp b/modules/ocl/test/test_gemm.cpp
index c26a8481f2..00d428770f 100644
--- a/modules/ocl/test/test_gemm.cpp
+++ b/modules/ocl/test/test_gemm.cpp
@@ -43,7 +43,7 @@
 //M*/
 
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 using namespace std;
 #ifdef HAVE_CLAMDBLAS
 ////////////////////////////////////////////////////////////////////////////
@@ -53,13 +53,12 @@ PARAM_TEST_CASE(Gemm, int, cv::Size, int)
     int      type;
     cv::Size mat_size;
     int		 flags;
-    //vector<cv::ocl::Info> info;
+
     virtual void SetUp()
     {
         type     = GET_PARAM(0);
         mat_size = GET_PARAM(1);
         flags    = GET_PARAM(2);
-        //cv::ocl::getDevice(info);
     }
 };
 
@@ -75,7 +74,7 @@ TEST_P(Gemm, Accuracy)
     cv::gemm(a, b, 1.0, c, 1.0, dst, flags);
     cv::ocl::gemm(cv::ocl::oclMat(a), cv::ocl::oclMat(b), 1.0, cv::ocl::oclMat(c), 1.0, ocl_dst, flags);
 
-    EXPECT_MAT_NEAR(dst, ocl_dst, mat_size.area() * 1e-4, "");
+    EXPECT_MAT_NEAR(dst, ocl_dst, mat_size.area() * 1e-4);
 }
 
 INSTANTIATE_TEST_CASE_P(ocl_gemm, Gemm, testing::Combine(
diff --git a/modules/ocl/test/test_haar.cpp b/modules/ocl/test/test_haar.cpp
deleted file mode 100644
index 6d2a912c24..0000000000
--- a/modules/ocl/test/test_haar.cpp
+++ /dev/null
@@ -1,163 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Jia Haipeng, jiahaipeng95@gmail.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "opencv2/objdetect.hpp"
-#include "precomp.hpp"
-
-#ifdef HAVE_OPENCL
-
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-using namespace cv;
-extern string workdir;
-struct getRect
-{
-    Rect operator ()(const CvAvgComp &e) const
-    {
-        return e.rect;
-    }
-};
-
-PARAM_TEST_CASE(HaarTestBase, int, int)
-{
-    //std::vector<cv::ocl::Info> oclinfo;
-    cv::ocl::OclCascadeClassifier cascade, nestedCascade;
-    cv::CascadeClassifier cpucascade, cpunestedCascade;
-    //    Mat img;
-
-    double scale;
-    int index;
-
-    virtual void SetUp()
-    {
-        scale = 1.0;
-        index = 0;
-        string cascadeName = workdir + "../../data/haarcascades/haarcascade_frontalface_alt.xml";
-
-        if( (!cascade.load( cascadeName )) || (!cpucascade.load(cascadeName)))
-        {
-            cout << "ERROR: Could not load classifier cascade" << endl;
-            return;
-        }
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums>0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-        //cv::ocl::setBinpath("E:\\");
-    }
-};
-
-////////////////////////////////faceDetect/////////////////////////////////////////////////
-
-struct Haar : HaarTestBase {};
-
-TEST_F(Haar, FaceDetect)
-{
-    string imgName = workdir + "lena.jpg";
-    Mat img = imread( imgName, 1 );
-
-    if(img.empty())
-    {
-        std::cout << "Couldn't read " << imgName << std::endl;
-        return ;
-    }
-
-    //int i = 0;
-    //double t = 0;
-    vector<Rect> faces, oclfaces;
-
-    // const static Scalar colors[] =  { CV_RGB(0, 0, 255),
-    //                                   CV_RGB(0, 128, 255),
-    //                                   CV_RGB(0, 255, 255),
-    //                                   CV_RGB(0, 255, 0),
-    //                                   CV_RGB(255, 128, 0),
-    //                                   CV_RGB(255, 255, 0),
-    //                                   CV_RGB(255, 0, 0),
-    //                                   CV_RGB(255, 0, 255)
-    //                                 } ;
-
-    Mat gray, smallImg(cvRound (img.rows / scale), cvRound(img.cols / scale), CV_8UC1 );
-    MemStorage storage(cvCreateMemStorage(0));
-    cvtColor( img, gray, CV_BGR2GRAY );
-    resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
-    equalizeHist( smallImg, smallImg );
-
-
-    cv::ocl::oclMat image;
-    CvSeq *_objects;
-    image.upload(smallImg);
-    _objects = cascade.oclHaarDetectObjects( image, storage, 1.1,
-               3, 0
-               | CV_HAAR_SCALE_IMAGE
-               , Size(30, 30), Size(0, 0) );
-    vector<CvAvgComp> vecAvgComp;
-    Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
-    oclfaces.resize(vecAvgComp.size());
-    std::transform(vecAvgComp.begin(), vecAvgComp.end(), oclfaces.begin(), getRect());
-
-    cpucascade.detectMultiScale( smallImg, faces,  1.1,
-                                 3, 0
-                                 | CV_HAAR_SCALE_IMAGE
-                                 , Size(30, 30), Size(0, 0) );
-    EXPECT_EQ(faces.size(), oclfaces.size());
-    /*	for( vector<Rect>::const_iterator r = faces.begin(); r != faces.end(); r++, i++ )
-    {
-    Mat smallImgROI;
-    Point center;
-    Scalar color = colors[i%8];
-    int radius;
-    center.x = cvRound((r->x + r->width*0.5)*scale);
-    center.y = cvRound((r->y + r->height*0.5)*scale);
-    radius = cvRound((r->width + r->height)*0.25*scale);
-    circle( img, center, radius, color, 3, 8, 0 );
-    } */
-    //namedWindow("result");
-    //imshow("result",img);
-    //waitKey(0);
-    //destroyAllWindows();
-
-}
-#endif // HAVE_OPENCL
diff --git a/modules/ocl/test/test_hough.cpp b/modules/ocl/test/test_hough.cpp
index 3a5cec5f95..9e45076cd7 100644
--- a/modules/ocl/test/test_hough.cpp
+++ b/modules/ocl/test/test_hough.cpp
@@ -40,7 +40,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -80,7 +80,7 @@ TEST_P(HoughCircles, Accuracy)
     cv::ocl::oclMat d_src(src);
 
     cv::ocl::oclMat d_circles;
-    cv::ocl::HoughCircles(d_src, d_circles, CV_HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
+    cv::ocl::HoughCircles(d_src, d_circles, cv::HOUGH_GRADIENT, dp, minDist, cannyThreshold, votesThreshold, minRadius, maxRadius);
     ASSERT_TRUE(d_circles.rows > 0);
 
     cv::Mat circles;
diff --git a/modules/ocl/test/test_imgproc.cpp b/modules/ocl/test/test_imgproc.cpp
index 9cc6c1c898..4cad2fabeb 100644
--- a/modules/ocl/test/test_imgproc.cpp
+++ b/modules/ocl/test/test_imgproc.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
@@ -22,6 +23,7 @@
 //    Rock Li, Rock.Li@amd.com
 //    Wu Zailong, bullet@yeah.net
 //    Xu Pang, pangxu010@163.com
+//    Sen Liu, swjtuls1987@126.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -49,7 +51,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -220,10 +222,10 @@ COOR do_meanShift(int x0, int y0, uchar *sptr, uchar *dptr, int sstep, cv::Size
 void meanShiftFiltering_(const Mat &src_roi, Mat &dst_roi, int sp, int sr, cv::TermCriteria crit)
 {
     if( src_roi.empty() )
-        CV_Error( CV_StsBadArg, "The input image is empty" );
+        CV_Error(cv::Error::StsBadArg, "The input image is empty" );
 
     if( src_roi.depth() != CV_8U || src_roi.channels() != 4 )
-        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
+        CV_Error(cv::Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
 
     CV_Assert( (src_roi.cols == dst_roi.cols) && (src_roi.rows == dst_roi.rows) );
     CV_Assert( !(dst_roi.step & 0x3) );
@@ -259,9 +261,9 @@ void meanShiftProc_(const Mat &src_roi, Mat &dst_roi, Mat &dstCoor_roi, int sp,
 {
 
     if( src_roi.empty() )
-        CV_Error( CV_StsBadArg, "The input image is empty" );
+        CV_Error(cv::Error::StsBadArg, "The input image is empty" );
     if( src_roi.depth() != CV_8U || src_roi.channels() != 4 )
-        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
+        CV_Error(cv::Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
     CV_Assert( (src_roi.cols == dst_roi.cols) && (src_roi.rows == dst_roi.rows) &&
                (src_roi.cols == dstCoor_roi.cols) && (src_roi.rows == dstCoor_roi.rows));
     CV_Assert( !(dstCoor_roi.step & 0x3) );
@@ -327,7 +329,7 @@ PARAM_TEST_CASE(ImgprocTestBase, MatType, MatType, MatType, MatType, MatType, bo
     cv::Mat mask_roi;
     cv::Mat dst_roi;
     cv::Mat dst1_roi; //bak
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl mat
     cv::ocl::oclMat clmat1;
     cv::ocl::oclMat clmat2;
@@ -352,10 +354,6 @@ PARAM_TEST_CASE(ImgprocTestBase, MatType, MatType, MatType, MatType, MatType, bo
         cv::RNG &rng = TS::ptr()->get_rng();
         cv::Size size(MWIDTH, MHEIGHT);
         double min = 1, max = 20;
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
 
         if(type1 != nulltype)
         {
@@ -445,6 +443,13 @@ PARAM_TEST_CASE(ImgprocTestBase, MatType, MatType, MatType, MatType, MatType, bo
             clmask_roi = clmask(Rect(maskx, masky, roicols, roirows));
         }
     }
+
+    void Near(double threshold)
+    {
+        cv::Mat cpu_cldst;
+        cldst.download(cpu_cldst);
+        EXPECT_MAT_NEAR(dst, cpu_cldst, threshold);
+    }
 };
 ////////////////////////////////equalizeHist//////////////////////////////////////////
 
@@ -464,84 +469,12 @@ TEST_P(equalizeHist, Mat)
             random_roi();
             cv::equalizeHist(mat1_roi, dst_roi);
             cv::ocl::equalizeHist(clmat1_roi, cldst_roi);
-            cv::Mat cpu_cldst;
-            cldst.download(cpu_cldst);
-            char sss[1024];
-            sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,dst1x=%d,dst1y=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, dst1x, dst1y, maskx, masky, src2x, src2y);
-            EXPECT_MAT_NEAR(dst, cpu_cldst, 1.1, sss);
+            Near(1.1);
         }
     }
 }
 
 
-
-
-
-////////////////////////////////bilateralFilter////////////////////////////////////////////
-
-struct bilateralFilter : ImgprocTestBase {};
-
-TEST_P(bilateralFilter, Mat)
-{
-    double sigmacolor = 50.0;
-    int radius = 9;
-    int d = 2 * radius + 1;
-    double sigmaspace = 20.0;
-    int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE, cv::BORDER_REFLECT, cv::BORDER_WRAP, cv::BORDER_REFLECT_101};
-    const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE", "BORDER_REFLECT", "BORDER_WRAP", "BORDER_REFLECT_101"};
-
-    if (mat1.depth() != CV_8U || mat1.type() != dst.type())
-    {
-        cout << "Unsupported type" << endl;
-        EXPECT_DOUBLE_EQ(0.0, 0.0);
-    }
-    else
-    {
-        for(size_t i = 0; i < sizeof(bordertype) / sizeof(int); i++)
-            for(int j = 0; j < LOOP_TIMES; j++)
-            {
-                random_roi();
-                if(((bordertype[i] != cv::BORDER_CONSTANT) && (bordertype[i] != cv::BORDER_REPLICATE) && (mat1_roi.cols <= radius)) || (mat1_roi.cols <= radius) || (mat1_roi.rows <= radius) || (mat1_roi.rows <= radius))
-                {
-                    continue;
-                }
-                //if((dstx>=radius) && (dsty >= radius) && (dstx+cldst_roi.cols+radius <=cldst_roi.wholecols) && (dsty+cldst_roi.rows+radius <= cldst_roi.wholerows))
-                //{
-                //	dst_roi.adjustROI(radius, radius, radius, radius);
-                //	cldst_roi.adjustROI(radius, radius, radius, radius);
-                //}
-                //else
-                //{
-                //	continue;
-                //}
-
-                cv::bilateralFilter(mat1_roi, dst_roi, d, sigmacolor, sigmaspace, bordertype[i] | cv::BORDER_ISOLATED);
-                cv::ocl::bilateralFilter(clmat1_roi, cldst_roi, d, sigmacolor, sigmaspace, bordertype[i] | cv::BORDER_ISOLATED);
-
-                cv::Mat cpu_cldst;
-                cldst.download(cpu_cldst);
-
-
-                char sss[1024];
-                sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,radius=%d,boredertype=%s", roicols, roirows, src1x, src1y, dstx, dsty, radius, borderstr[i]);
-                //for(int i=0;i<dst.rows;i++)
-                //{
-                //	for(int j=0;j<dst.cols*dst.channels();j++)
-                //	{
-                //		if(dst.at<uchar>(i,j)!=cpu_cldst.at<uchar>(i,j))
-                //		cout<< i <<" "<< j <<" "<< (int)dst.at<uchar>(i,j)<<" "<< (int)cpu_cldst.at<uchar>(i,j)<<"  ";
-                //	}
-                //	cout<<endl;
-                //}
-
-                EXPECT_MAT_NEAR(dst, cpu_cldst, 1.0, sss);
-
-            }
-    }
-}
-
-
-
 ////////////////////////////////copyMakeBorder////////////////////////////////////////////
 
 struct CopyMakeBorder : ImgprocTestBase {};
@@ -549,7 +482,7 @@ struct CopyMakeBorder : ImgprocTestBase {};
 TEST_P(CopyMakeBorder, Mat)
 {
     int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE, cv::BORDER_REFLECT, cv::BORDER_WRAP, cv::BORDER_REFLECT_101};
-    const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE", "BORDER_REFLECT", "BORDER_WRAP", "BORDER_REFLECT_101"};
+    //const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE", "BORDER_REFLECT", "BORDER_WRAP", "BORDER_REFLECT_101"};
     cv::RNG &rng = TS::ptr()->get_rng();
     int top = rng.uniform(0, 10);
     int bottom = rng.uniform(0, 10);
@@ -587,24 +520,12 @@ TEST_P(CopyMakeBorder, Mat)
                 cv::Mat cpu_cldst;
 #ifndef RANDOMROI
                 cldst_roi.download(cpu_cldst);
+                EXPECT_MAT_NEAR(dst_roi, cpu_cldst, 0.0);
 #else
                 cldst.download(cpu_cldst);
+                EXPECT_MAT_NEAR(dst, cpu_cldst, 0.0);
 #endif
-                char sss[1024];
-                sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,dst1x=%d,dst1y=%d,top=%d,bottom=%d,left=%d,right=%d, bordertype=%s", roicols, roirows, src1x, src1y, dstx, dsty, dst1x, dst1y, top, bottom, left, right, borderstr[i]);
-#ifndef RANDOMROI
-                EXPECT_MAT_NEAR(dst_roi, cpu_cldst, 0.0, sss);
-#else
-                //for(int i=0;i<dst.rows;i++)
-                //{
-                //for(int j=0;j<dst.cols;j++)
-                //{
-                //	cout<< (int)dst.at<uchar>(i,j)<<" ";
-                //}
-                //cout<<endl;
-                //}
-                EXPECT_MAT_NEAR(dst, cpu_cldst, 0.0, sss);
-#endif
+
             }
     }
 }
@@ -627,14 +548,7 @@ TEST_P(cornerMinEigenVal, Mat)
         int borderType = cv::BORDER_REFLECT;
         cv::cornerMinEigenVal(mat1_roi, dst_roi, blockSize, apertureSize, borderType);
         cv::ocl::cornerMinEigenVal(clmat1_roi, cldst_roi, blockSize, apertureSize, borderType);
-
-
-        cv::Mat cpu_cldst;
-        cldst.download(cpu_cldst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,dst1x=%d,dst1y=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, dst1x, dst1y, maskx, masky, src2x, src2y);
-        EXPECT_MAT_NEAR(dst, cpu_cldst, 1, sss);
+        Near(1.);
     }
 }
 
@@ -657,13 +571,7 @@ TEST_P(cornerHarris, Mat)
         int borderType = cv::BORDER_REFLECT;
         cv::cornerHarris(mat1_roi, dst_roi, blockSize, apertureSize, k, borderType);
         cv::ocl::cornerHarris(clmat1_roi, cldst_roi, blockSize, apertureSize, k, borderType);
-        cv::Mat cpu_cldst;
-        cldst.download(cpu_cldst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,dst1x=%d,dst1y=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, dst1x, dst1y, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_cldst, 1, sss);
+        Near(1.);
     }
 }
 
@@ -680,15 +588,11 @@ TEST_P(integral, Mat)
 
         cv::ocl::integral(clmat1_roi, cldst_roi, cldst1_roi);
         cv::integral(mat1_roi, dst_roi, dst1_roi);
+        Near(0);
 
-        cv::Mat cpu_cldst, cpu_cldst1;
-        cldst.download(cpu_cldst);
+        cv::Mat cpu_cldst1;
         cldst1.download(cpu_cldst1);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,dst1x=%d,dst1y=%d,maskx=%d,masky=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, dst1x, dst1y, maskx, masky, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_cldst, 0.0, sss);
-        EXPECT_MAT_NEAR(dst1, cpu_cldst1, 0.0, sss);
+        EXPECT_MAT_NEAR(dst1, cpu_cldst1, 0.0);
     }
 }
 
@@ -720,7 +624,7 @@ PARAM_TEST_CASE(WarpTestBase, MatType, int)
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -740,10 +644,6 @@ PARAM_TEST_CASE(WarpTestBase, MatType, int)
         mat1 = randomMat(rng, size, type, 5, 16, false);
         dst  = randomMat(rng, size, type, 5, 16, false);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -805,10 +705,7 @@ TEST_P(WarpAffine, Mat)
 
         cv::Mat cpu_dst;
         gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "src_roicols=%d,src_roirows=%d,dst_roicols=%d,dst_roirows=%d,src1x =%d,src1y=%d,dstx=%d,dsty=%d", src_roicols, src_roirows, dst_roicols, dst_roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
+        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0);
     }
 
 }
@@ -837,10 +734,7 @@ TEST_P(WarpPerspective, Mat)
 
         cv::Mat cpu_dst;
         gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "src_roicols=%d,src_roirows=%d,dst_roicols=%d,dst_roirows=%d,src1x =%d,src1y=%d,dstx=%d,dsty=%d", src_roicols, src_roirows, dst_roicols, dst_roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
+        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0);
     }
 
 }
@@ -905,9 +799,6 @@ PARAM_TEST_CASE(Remap, MatType, MatType, MatType, int, int)
         interpolation = GET_PARAM(3);
         bordertype = GET_PARAM(4);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-
         cv::RNG &rng = TS::ptr()->get_rng();
         cv::Size srcSize = cv::Size(MWIDTH, MHEIGHT);
         cv::Size map1Size = cv::Size(MWIDTH, MHEIGHT);
@@ -1004,7 +895,7 @@ TEST_P(Remap, Mat)
         return;
     }
     int bordertype[] = {cv::BORDER_CONSTANT, cv::BORDER_REPLICATE/*,BORDER_REFLECT,BORDER_WRAP,BORDER_REFLECT_101*/};
-    const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE"/*, "BORDER_REFLECT","BORDER_WRAP","BORDER_REFLECT_101"*/};
+    //const char *borderstr[] = {"BORDER_CONSTANT", "BORDER_REPLICATE"/*, "BORDER_REFLECT","BORDER_WRAP","BORDER_REFLECT_101"*/};
     // for(int i = 0; i < sizeof(bordertype)/sizeof(int); i++)
     for(int j = 0; j < LOOP_TIMES; j++)
     {
@@ -1014,13 +905,9 @@ TEST_P(Remap, Mat)
         cv::Mat cpu_dst;
         gdst.download(cpu_dst);
 
-        char sss[1024];
-        sprintf(sss, "src_roicols=%d,src_roirows=%d,dst_roicols=%d,dst_roirows=%d,src1x =%d,src1y=%d,dstx=%d,dsty=%d bordertype=%s", src_roicols, src_roirows, dst_roicols, dst_roirows, srcx, srcy, dstx, dsty, borderstr[0]);
-
-
         if(interpolation == 0)
-            EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
-        EXPECT_MAT_NEAR(dst, cpu_dst, 2.0, sss);
+            EXPECT_MAT_NEAR(dst, cpu_dst, 1.0);
+        EXPECT_MAT_NEAR(dst, cpu_dst, 2.0);
 
     }
 }
@@ -1051,7 +938,6 @@ PARAM_TEST_CASE(Resize, MatType, cv::Size, double, double, int)
     int dstx;
     int dsty;
 
-    //std::vector<cv::ocl::Info> oclinfo;
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat dst_roi;
@@ -1090,10 +976,6 @@ PARAM_TEST_CASE(Resize, MatType, cv::Size, double, double, int)
         mat1 = randomMat(rng, size, type, 5, 16, false);
         dst  = randomMat(rng, dsize, type, 5, 16, false);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -1149,10 +1031,7 @@ TEST_P(Resize, Mat)
 
         cv::Mat cpu_dst;
         gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "src_roicols=%d,src_roirows=%d,dst_roicols=%d,dst_roirows=%d,src1x =%d,src1y=%d,dstx=%d,dsty=%d", src_roicols, src_roirows, dst_roicols, dst_roirows, src1x, src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0, sss);
+        EXPECT_MAT_NEAR(dst, cpu_dst, 1.0);
     }
 
 }
@@ -1181,7 +1060,7 @@ PARAM_TEST_CASE(Threshold, MatType, ThreshOp)
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -1199,11 +1078,6 @@ PARAM_TEST_CASE(Threshold, MatType, ThreshOp)
 
         mat1 = randomMat(rng, size, type, 5, 16, false);
         dst  = randomMat(rng, size, type, 5, 16, false);
-
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -1251,12 +1125,7 @@ TEST_P(Threshold, Mat)
 
         cv::Mat cpu_dst;
         gdst_whole.download(cpu_dst);
-
-        //EXPECT_MAT_NEAR(dst, cpu_dst, 1e-5)
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x =%d,src1y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x , src1y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1, sss);
+        EXPECT_MAT_NEAR(dst, cpu_dst, 1);
     }
 
 }
@@ -1288,7 +1157,6 @@ PARAM_TEST_CASE(meanShiftTestBase, MatType, MatType, int, int, cv::TermCriteria)
     cv::ocl::oclMat gdst;
     cv::ocl::oclMat gdstCoor;
 
-    //std::vector<cv::ocl::Info> oclinfo;
     //ocl mat with roi
     cv::ocl::oclMat gsrc_roi;
     cv::ocl::oclMat gdst_roi;
@@ -1311,10 +1179,6 @@ PARAM_TEST_CASE(meanShiftTestBase, MatType, MatType, int, int, cv::TermCriteria)
         dst = randomMat(rng, size, type, 5, 16, false);
         dstCoor = randomMat(rng, size, typeCoor, 5, 16, false);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -1367,11 +1231,7 @@ TEST_P(meanShiftFiltering, Mat)
         cv::ocl::meanShiftFiltering(gsrc_roi, gdst_roi, sp, sr, crit);
 
         gdst.download(cpu_gdst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx=%d,srcy=%d,dstx=%d,dsty=%d\n", roicols, roirows, srcx, srcy, dstx, dsty);
-        EXPECT_MAT_NEAR(dst, cpu_gdst, 0.0, sss);
-
+        EXPECT_MAT_NEAR(dst, cpu_gdst, 0.0);
     }
 }
 
@@ -1393,11 +1253,8 @@ TEST_P(meanShiftProc, Mat)
 
         gdst.download(cpu_gdst);
         gdstCoor.download(cpu_gdstCoor);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx=%d,srcy=%d,dstx=%d,dsty=%d\n", roicols, roirows, srcx, srcy, dstx, dsty);
-        EXPECT_MAT_NEAR(dst, cpu_gdst, 0.0, sss);
-        EXPECT_MAT_NEAR(dstCoor, cpu_gdstCoor, 0.0, sss);
+        EXPECT_MAT_NEAR(dst, cpu_gdst, 0.0);
+        EXPECT_MAT_NEAR(dstCoor, cpu_gdstCoor, 0.0);
     }
 }
 
@@ -1436,7 +1293,6 @@ PARAM_TEST_CASE(histTestBase, MatType, MatType)
     cv::ocl::oclMat gdst_hist;
     //ocl mat with roi
     cv::ocl::oclMat gsrc_roi;
-    //    std::vector<cv::ocl::Info> oclinfo;
 
     virtual void SetUp()
     {
@@ -1447,10 +1303,6 @@ PARAM_TEST_CASE(histTestBase, MatType, MatType)
 
         src = randomMat(rng, size, type_src, 0, 256, false);
 
-        //        int devnums = getDevice(oclinfo);
-        //        CV_Assert(devnums > 0);
-        //if you want to use undefault device, set it here
-        //setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -1489,12 +1341,45 @@ TEST_P(calcHist, Mat)
         cv::ocl::calcHist(gsrc_roi, gdst_hist);
 
         gdst_hist.download(cpu_hist);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx=%d,srcy=%d\n", roicols, roirows, srcx, srcy);
-        EXPECT_MAT_NEAR(dst_hist, cpu_hist, 0.0, sss);
+        EXPECT_MAT_NEAR(dst_hist, cpu_hist, 0.0);
     }
 }
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// CLAHE
+
+PARAM_TEST_CASE(CLAHE, cv::Size, double)
+{
+    cv::Size gridSize;
+    double clipLimit;
+
+    cv::Mat src;
+    cv::Mat dst_gold;
+
+    cv::ocl::oclMat g_src;
+    cv::ocl::oclMat g_dst;
+
+    virtual void SetUp()
+    {
+        gridSize = GET_PARAM(0);
+        clipLimit = GET_PARAM(1);
+
+        cv::RNG &rng = TS::ptr()->get_rng();
+        src = randomMat(rng, cv::Size(MWIDTH, MHEIGHT), CV_8UC1, 0, 256, false);
+        g_src.upload(src);
+    }
+};
+
+TEST_P(CLAHE, Accuracy)
+{
+    cv::Ptr<cv::CLAHE> clahe = cv::ocl::createCLAHE(clipLimit, gridSize);
+    clahe->apply(g_src, g_dst);
+    cv::Mat dst(g_dst);
+
+    cv::Ptr<cv::CLAHE> clahe_gold = cv::createCLAHE(clipLimit, gridSize);
+    clahe_gold->apply(src, dst_gold);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1.0);
+}
 
 ///////////////////////////Convolve//////////////////////////////////
 PARAM_TEST_CASE(ConvolveTestBase, MatType, bool)
@@ -1629,15 +1514,52 @@ TEST_P(Convolve, Mat)
 
         cv::Mat cpu_dst;
         gdst_whole.download(cpu_dst);
-
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x=%d,src1y=%d,dstx=%d,dsty=%d,src2x=%d,src2y=%d", roicols, roirows, src1x, src1y, dstx, dsty, src2x, src2y);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 1e-1, sss);
+        EXPECT_MAT_NEAR(dst, cpu_dst, .1);
 
     }
 }
 
+//////////////////////////////// ColumnSum //////////////////////////////////////
+PARAM_TEST_CASE(ColumnSum, cv::Size)
+{
+    cv::Size size;
+    cv::Mat src;
+
+    virtual void SetUp()
+    {
+        size = GET_PARAM(0);
+    }
+};
+
+TEST_P(ColumnSum, Accuracy)
+{
+    cv::Mat src = randomMat(size, CV_32FC1);
+    cv::ocl::oclMat d_dst;
+    cv::ocl::oclMat d_src(src);
+
+    cv::ocl::columnSum(d_src, d_dst);
+
+    cv::Mat dst(d_dst);
+
+    for (int j = 0; j < src.cols; ++j)
+    {
+        float gold = src.at<float>(0, j);
+        float res = dst.at<float>(0, j);
+        ASSERT_NEAR(res, gold, 1e-5);
+    }
+
+    for (int i = 1; i < src.rows; ++i)
+    {
+        for (int j = 0; j < src.cols; ++j)
+        {
+            float gold = src.at<float>(i, j) += src.at<float>(i - 1, j);
+            float res = dst.at<float>(i, j);
+            ASSERT_NEAR(res, gold, 1e-5);
+        }
+    }
+}
+/////////////////////////////////////////////////////////////////////////////////////
+
 INSTANTIATE_TEST_CASE_P(ImgprocTestBase, equalizeHist, Combine(
                             ONE_TYPE(CV_8UC1),
                             NULL_TYPE,
@@ -1646,21 +1568,6 @@ INSTANTIATE_TEST_CASE_P(ImgprocTestBase, equalizeHist, Combine(
                             NULL_TYPE,
                             Values(false))); // Values(false) is the reserved parameter
 
-//INSTANTIATE_TEST_CASE_P(ImgprocTestBase, bilateralFilter, Combine(
-//	ONE_TYPE(CV_8UC1),
-//	NULL_TYPE,
-//	ONE_TYPE(CV_8UC1),
-//	NULL_TYPE,
-//	NULL_TYPE,
-//	Values(false))); // Values(false) is the reserved parameter
-INSTANTIATE_TEST_CASE_P(ImgprocTestBase, bilateralFilter, Combine(
-                            Values(CV_8UC1, CV_8UC3),
-                            NULL_TYPE,
-                            Values(CV_8UC1, CV_8UC3),
-                            NULL_TYPE,
-                            NULL_TYPE,
-                            Values(false))); // Values(false) is the reserved parameter
-
 
 INSTANTIATE_TEST_CASE_P(ImgprocTestBase, CopyMakeBorder, Combine(
                             Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4),
@@ -1749,7 +1656,10 @@ INSTANTIATE_TEST_CASE_P(histTestBase, calcHist, Combine(
                             ONE_TYPE(CV_32SC1) //no use
                         ));
 
-//INSTANTIATE_TEST_CASE_P(ConvolveTestBase, Convolve, Combine(
-//                            Values(CV_32FC1, CV_32FC1),
-//                            Values(false))); // Values(false) is the reserved parameter
+INSTANTIATE_TEST_CASE_P(Imgproc, CLAHE, Combine(
+                        Values(cv::Size(4, 4), cv::Size(32, 8), cv::Size(8, 64)),
+                        Values(0.0, 10.0, 62.0, 300.0)));
+
+INSTANTIATE_TEST_CASE_P(Imgproc, ColumnSum, DIFFERENT_SIZES);
+
 #endif // HAVE_OPENCL
diff --git a/modules/ocl/test/test_kmeans.cpp b/modules/ocl/test/test_kmeans.cpp
new file mode 100644
index 0000000000..008acfaa4f
--- /dev/null
+++ b/modules/ocl/test/test_kmeans.cpp
@@ -0,0 +1,162 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Erping Pang,   pang_er_ping@163.com
+//    Xiaopeng Fu,   fuxiaopeng2222@163.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_OPENCL
+
+using namespace cvtest;
+using namespace testing;
+using namespace std;
+using namespace cv;
+
+#define OCL_KMEANS_USE_INITIAL_LABELS 1
+#define OCL_KMEANS_PP_CENTERS         2
+
+PARAM_TEST_CASE(Kmeans, int, int, int)
+{
+    int type;
+    int K;
+    int flags;
+    cv::Mat src ;
+    ocl::oclMat d_src, d_dists;
+
+    Mat labels, centers;
+    ocl::oclMat d_labels, d_centers;
+    cv::RNG rng ;
+    virtual void SetUp(){
+        K = GET_PARAM(0);
+        type = GET_PARAM(1);
+        flags = GET_PARAM(2);
+        rng = TS::ptr()->get_rng();
+
+        // MWIDTH=256, MHEIGHT=256. defined in utility.hpp
+        cv::Size size = cv::Size(MWIDTH, MHEIGHT);
+        src.create(size, type);
+        int row_idx = 0;
+        const int max_neighbour = MHEIGHT / K - 1;
+        CV_Assert(K <= MWIDTH);
+        for(int i = 0; i < K; i++ )
+        {
+            Mat center_row_header = src.row(row_idx);
+            center_row_header.setTo(0);
+            int nchannel = center_row_header.channels();
+            for(int j = 0; j < nchannel; j++)
+                center_row_header.at<float>(0, i*nchannel+j) = 50000.0;
+
+            for(int j = 0; (j < max_neighbour) ||
+                           (i == K-1 && j < max_neighbour + MHEIGHT%K); j ++)
+            {
+                Mat cur_row_header = src.row(row_idx + 1 + j);
+                center_row_header.copyTo(cur_row_header);
+                Mat tmpmat = randomMat(rng, cur_row_header.size(), cur_row_header.type(), -200, 200, false);
+                cur_row_header += tmpmat;
+            }
+            row_idx += 1 + max_neighbour;
+        }
+    }
+};
+TEST_P(Kmeans, Mat){
+
+    if(flags & KMEANS_USE_INITIAL_LABELS)
+    {
+        // inital a given labels
+        labels.create(src.rows, 1, CV_32S);
+        int *label = labels.ptr<int>();
+        for(int i = 0; i < src.rows; i++)
+            label[i] = rng.uniform(0, K);
+        d_labels.upload(labels);
+    }
+    d_src.upload(src);
+
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        kmeans(src, K, labels,
+            TermCriteria( TermCriteria::EPS + TermCriteria::MAX_ITER, 100, 0),
+            1, flags, centers);
+
+        ocl::kmeans(d_src, K, d_labels,
+            TermCriteria( TermCriteria::EPS + TermCriteria::MAX_ITER, 100, 0),
+            1, flags, d_centers);
+
+        Mat dd_labels(d_labels);
+        Mat dd_centers(d_centers);
+        if(flags & KMEANS_USE_INITIAL_LABELS)
+        {
+            EXPECT_MAT_NEAR(labels, dd_labels, 0);
+            EXPECT_MAT_NEAR(centers, dd_centers, 1e-3);
+        }
+        else
+        {
+            int row_idx = 0;
+            for(int i = 0; i < K; i++)
+            {
+                // verify lables with ground truth resutls
+                int label = labels.at<int>(row_idx);
+                int header_label = dd_labels.at<int>(row_idx);
+                for(int j = 0; (j < MHEIGHT/K)||(i == K-1 && j < MHEIGHT/K+MHEIGHT%K); j++)
+                {
+                    ASSERT_NEAR(labels.at<int>(row_idx+j), label, 0);
+                    ASSERT_NEAR(dd_labels.at<int>(row_idx+j), header_label, 0);
+                }
+
+                // verify centers
+                float *center = centers.ptr<float>(label);
+                float *header_center = dd_centers.ptr<float>(header_label);
+                for(int t = 0; t < centers.cols; t++)
+                    ASSERT_NEAR(center[t], header_center[t], 1e-3);
+
+                row_idx += MHEIGHT/K;
+            }
+        }
+    }
+}
+INSTANTIATE_TEST_CASE_P(OCL_ML, Kmeans, Combine(
+    Values(3, 5, 8),
+    Values(CV_32FC1, CV_32FC2, CV_32FC4),
+    Values(OCL_KMEANS_USE_INITIAL_LABELS/*, OCL_KMEANS_PP_CENTERS*/)));
+
+#endif
diff --git a/modules/ocl/test/test_match_template.cpp b/modules/ocl/test/test_match_template.cpp
index c948e1d533..551c9ff12a 100644
--- a/modules/ocl/test/test_match_template.cpp
+++ b/modules/ocl/test/test_match_template.cpp
@@ -43,7 +43,7 @@
 //M*/
 
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 ////////////////////////////////////////////////////////////////////////////////
@@ -62,7 +62,6 @@ PARAM_TEST_CASE(MatchTemplate8U, cv::Size, TemplateSize, Channels, TemplateMetho
     cv::Size templ_size;
     int cn;
     int method;
-    //std::vector<cv::ocl::Info> oclinfo;
 
     virtual void SetUp()
     {
@@ -70,8 +69,6 @@ PARAM_TEST_CASE(MatchTemplate8U, cv::Size, TemplateSize, Channels, TemplateMetho
         templ_size = GET_PARAM(1);
         cn = GET_PARAM(2);
         method = GET_PARAM(3);
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
     }
 };
 
@@ -92,12 +89,10 @@ TEST_P(MatchTemplate8U, Accuracy)
     cv::Mat dst_gold;
     cv::matchTemplate(image, templ, dst_gold, method);
 
-    char sss [100] = "";
-
     cv::Mat mat_dst;
     dst.download(mat_dst);
 
-    EXPECT_MAT_NEAR(dst_gold, mat_dst, templ_size.area() * 1e-1, sss);
+    EXPECT_MAT_NEAR(dst_gold, mat_dst, templ_size.area() * 1e-1);
 }
 
 PARAM_TEST_CASE(MatchTemplate32F, cv::Size, TemplateSize, Channels, TemplateMethod)
@@ -114,8 +109,6 @@ PARAM_TEST_CASE(MatchTemplate32F, cv::Size, TemplateSize, Channels, TemplateMeth
         templ_size = GET_PARAM(1);
         cn = GET_PARAM(2);
         method = GET_PARAM(3);
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
     }
 };
 
@@ -130,26 +123,24 @@ TEST_P(MatchTemplate32F, Accuracy)
     cv::Mat dst_gold;
     cv::matchTemplate(image, templ, dst_gold, method);
 
-    char sss [100] = "";
-
     cv::Mat mat_dst;
     dst.download(mat_dst);
 
-    EXPECT_MAT_NEAR(dst_gold, mat_dst, templ_size.area() * 1e-1, sss);
+    EXPECT_MAT_NEAR(dst_gold, mat_dst, templ_size.area() * 1e-1);
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate8U,
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, MatchTemplate8U,
                         testing::Combine(
                             MTEMP_SIZES,
-                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16))/*, TemplateSize(cv::Size(30, 30))*/),
+                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
                             testing::Values(Channels(1), Channels(3), Channels(4)),
                             ALL_TEMPLATE_METHODS
                         )
                        );
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, MatchTemplate32F, testing::Combine(
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, MatchTemplate32F, testing::Combine(
                             MTEMP_SIZES,
-                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16))/*, TemplateSize(cv::Size(30, 30))*/),
+                            testing::Values(TemplateSize(cv::Size(5, 5)), TemplateSize(cv::Size(16, 16)), TemplateSize(cv::Size(30, 30))),
                             testing::Values(Channels(1), Channels(3), Channels(4)),
                             testing::Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))));
 #endif
diff --git a/modules/ocl/test/test_matrix_operation.cpp b/modules/ocl/test/test_matrix_operation.cpp
index ef11aaa132..e8b5022324 100644
--- a/modules/ocl/test/test_matrix_operation.cpp
+++ b/modules/ocl/test/test_matrix_operation.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
@@ -43,7 +44,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -72,7 +73,7 @@ PARAM_TEST_CASE(ConvertToTestBase, MatType, MatType)
     //src mat with roi
     cv::Mat mat_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -90,11 +91,6 @@ PARAM_TEST_CASE(ConvertToTestBase, MatType, MatType)
 
         mat = randomMat(rng, size, type, 5, 16, false);
         dst  = randomMat(rng, size, type, 5, 16, false);
-        //std::vector<cv::ocl::Info> oclinfo;
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -139,12 +135,7 @@ TEST_P(ConvertTo, Accuracy)
         mat_roi.convertTo(dst_roi, dst_type);
         gmat.convertTo(gdst, dst_type);
 
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx =%d,srcy=%d,dstx=%d,dsty=%d", roicols, roirows, srcx , srcy, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), 0.0);
     }
 }
 
@@ -175,7 +166,7 @@ PARAM_TEST_CASE(CopyToTestBase, MatType, bool)
     cv::Mat mat_roi;
     cv::Mat mask_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -197,10 +188,6 @@ PARAM_TEST_CASE(CopyToTestBase, MatType, bool)
 
         cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -250,12 +237,7 @@ TEST_P(CopyTo, Without_mask)
         mat_roi.copyTo(dst_roi);
         gmat.copyTo(gdst);
 
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx =%d,srcy=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d", roicols, roirows, srcx , srcy, dstx, dsty, maskx, masky);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), 0.0);
     }
 }
 
@@ -268,12 +250,7 @@ TEST_P(CopyTo, With_mask)
         mat_roi.copyTo(dst_roi, mask_roi);
         gmat.copyTo(gdst, gmask);
 
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx =%d,srcy=%d,dstx=%d,dsty=%d,maskx=%d,masky=%d", roicols, roirows, srcx , srcy, dstx, dsty, maskx, masky);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), 0.0);
     }
 }
 
@@ -301,7 +278,7 @@ PARAM_TEST_CASE(SetToTestBase, MatType, bool)
     //src mat with roi
     cv::Mat mat_roi;
     cv::Mat mask_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gmat_whole;
 
@@ -322,10 +299,6 @@ PARAM_TEST_CASE(SetToTestBase, MatType, bool)
         cv::threshold(mask, mask, 0.5, 255., CV_8UC1);
         val = cv::Scalar(rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0), rng.uniform(-10.0, 10.0));
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -369,12 +342,7 @@ TEST_P(SetTo, Without_mask)
         mat_roi.setTo(val);
         gmat.setTo(val);
 
-        cv::Mat cpu_dst;
-        gmat_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx =%d,srcy=%d,maskx=%d,masky=%d", roicols, roirows, srcx , srcy, maskx, masky);
-
-        EXPECT_MAT_NEAR(mat, cpu_dst, 1., sss);
+        EXPECT_MAT_NEAR(mat, Mat(gmat_whole), 1.);
     }
 }
 
@@ -387,12 +355,7 @@ TEST_P(SetTo, With_mask)
         mat_roi.setTo(val, mask_roi);
         gmat.setTo(val, gmask);
 
-        cv::Mat cpu_dst;
-        gmat_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,srcx =%d,srcy=%d,maskx=%d,masky=%d", roicols, roirows, srcx , srcy, maskx, masky);
-
-        EXPECT_MAT_NEAR(mat, cpu_dst, 1., sss);
+        EXPECT_MAT_NEAR(mat, Mat(gmat_whole), 1.);
     }
 }
 
@@ -417,7 +380,7 @@ PARAM_TEST_CASE(convertC3C4, MatType, cv::Size)
     //src mat with roi
     cv::Mat mat1_roi;
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -430,13 +393,6 @@ PARAM_TEST_CASE(convertC3C4, MatType, cv::Size)
         type = GET_PARAM(0);
         ksize = GET_PARAM(1);
 
-
-
-        //dst  = randomMat(rng, size, type, 5, 16, false);
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[1]);
     }
 
     void random_roi()
@@ -483,11 +439,8 @@ TEST_P(convertC3C4, Accuracy)
 
         mat1 = randomMat(rng, size, type, 0, 40, false);
         gmat1 = mat1;
-        cv::Mat cpu_dst;
-        gmat1.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "cols=%d,rows=%d", mat1.cols, mat1.rows);
-        EXPECT_MAT_NEAR(mat1, cpu_dst, 0.0, sss);
+
+        EXPECT_MAT_NEAR(mat1, Mat(gmat1), 0.0);
     }
 
 }
diff --git a/modules/ocl/test/test_moments.cpp b/modules/ocl/test/test_moments.cpp
index 715ad8963a..50b8379d86 100644
--- a/modules/ocl/test/test_moments.cpp
+++ b/modules/ocl/test/test_moments.cpp
@@ -1,6 +1,5 @@
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #include <iomanip>
-#include "opencv2/imgproc/imgproc_c.h"
 
 #ifdef HAVE_OPENCL
 
@@ -8,9 +7,8 @@ using namespace cv;
 using namespace cv::ocl;
 using namespace cvtest;
 using namespace testing;
-using namespace std;
-extern string workdir;
-PARAM_TEST_CASE(MomentsTestBase, MatType, bool)
+
+PARAM_TEST_CASE(MomentsTest, MatType, bool)
 {
     int type;
     cv::Mat mat1;
@@ -30,13 +28,13 @@ PARAM_TEST_CASE(MomentsTestBase, MatType, bool)
         Mat gpu_dst, cpu_dst;
         HuMoments(cpu, cpu_dst);
         HuMoments(gpu, gpu_dst);
-        EXPECT_MAT_NEAR(gpu_dst,cpu_dst, .5, "");
+        EXPECT_MAT_NEAR(gpu_dst,cpu_dst, .5);
     }
 
 };
-struct ocl_Moments : MomentsTestBase {};
 
-TEST_P(ocl_Moments, Mat)
+
+TEST_P(MomentsTest, Mat)
 {
     bool binaryImage = 0;
     SetUp();
@@ -45,13 +43,13 @@ TEST_P(ocl_Moments, Mat)
     {
         if(test_contours)
         {
-            Mat src = imread( workdir + "../cpp/pic3.png", 1 );
-            Mat src_gray, canny_output;
-            cvtColor( src, src_gray, CV_BGR2GRAY );
+            Mat src = readImage( "cv/shared/pic3.png", IMREAD_GRAYSCALE );
+            ASSERT_FALSE(src.empty());
+            Mat canny_output;
             vector<vector<Point> > contours;
             vector<Vec4i> hierarchy;
-            Canny( src_gray, canny_output, 100, 200, 3 );
-            findContours( canny_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+            Canny( src, canny_output, 100, 200, 3 );
+            findContours( canny_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
             for( size_t i = 0; i < contours.size(); i++ )
             {
                 Moments m = moments( contours[i], false );
@@ -64,9 +62,9 @@ TEST_P(ocl_Moments, Mat)
         cv::Moments oclMom = cv::ocl::ocl_moments(_array, binaryImage);
 
         Compare(CvMom, oclMom);
-
     }
 }
-INSTANTIATE_TEST_CASE_P(Moments, ocl_Moments, Combine(
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, MomentsTest, Combine(
                             Values(CV_8UC1, CV_16UC1, CV_16SC1, CV_64FC1), Values(true,false)));
+
 #endif // HAVE_OPENCL
diff --git a/modules/gpu/test/test_precomp.cpp b/modules/ocl/test/test_norm.cpp
similarity index 77%
rename from modules/gpu/test/test_precomp.cpp
rename to modules/ocl/test/test_norm.cpp
index 8121ae405b..2bd847068e 100644
--- a/modules/gpu/test/test_precomp.cpp
+++ b/modules/ocl/test/test_norm.cpp
@@ -40,3 +40,24 @@
 //M*/
 
 #include "test_precomp.hpp"
+
+typedef ::testing::TestWithParam<cv::Size> normFixture;
+
+TEST_P(normFixture, DISABLED_accuracy)
+{
+    const cv::Size srcSize = GetParam();
+
+    cv::Mat src1(srcSize, CV_8UC1), src2(srcSize, CV_8UC1);
+    cv::randu(src1, 0, 2);
+    cv::randu(src2, 0, 2);
+
+    cv::ocl::oclMat oclSrc1(src1), oclSrc2(src2);
+
+    double value = cv::norm(src1, src2, cv::NORM_INF);
+    double oclValue = cv::ocl::norm(oclSrc1, oclSrc2, cv::NORM_INF);
+
+    ASSERT_EQ(value, oclValue);
+}
+
+INSTANTIATE_TEST_CASE_P(oclNormTest, normFixture,
+                        ::testing::Values(cv::Size(500, 500), cv::Size(1000, 1000)));
diff --git a/modules/ocl/test/test_hog.cpp b/modules/ocl/test/test_objdetect.cpp
similarity index 50%
rename from modules/ocl/test/test_hog.cpp
rename to modules/ocl/test/test_objdetect.cpp
index 9c27c88032..295ae642b0 100644
--- a/modules/ocl/test/test_hog.cpp
+++ b/modules/ocl/test/test_objdetect.cpp
@@ -15,7 +15,7 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//		Wenju He, wenju@multicorewareinc.com
+//		Yao Wang, bitwangyaoyao@gmail.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
 // are permitted provided that the following conditions are met:
@@ -43,53 +43,55 @@
 //
 //M*/
 
-#include "precomp.hpp"
-#include "opencv2/core.hpp"
-using namespace std;
+#include "test_precomp.hpp"
+#include "opencv2/objdetect.hpp"
+
+using namespace cv;
+using namespace testing;
+
 #ifdef HAVE_OPENCL
 
-extern string workdir;
-PARAM_TEST_CASE(HOG, cv::Size, int)
+///////////////////// HOG /////////////////////////////
+PARAM_TEST_CASE(HOG, Size, int)
 {
-    cv::Size winSize;
+    Size winSize;
     int type;
+    Mat img_rgb;
     virtual void SetUp()
     {
         winSize = GET_PARAM(0);
         type = GET_PARAM(1);
+        img_rgb = readImage("gpu/hog/road.png");
+        ASSERT_FALSE(img_rgb.empty());
     }
 };
 
 TEST_P(HOG, GetDescriptors)
 {
-    // Load image
-    cv::Mat img_rgb = readImage(workdir + "lena.jpg");
-    ASSERT_FALSE(img_rgb.empty());
-
     // Convert image
-    cv::Mat img;
+    Mat img;
     switch (type)
     {
     case CV_8UC1:
-        cv::cvtColor(img_rgb, img, CV_BGR2GRAY);
+        cvtColor(img_rgb, img, COLOR_BGR2GRAY);
         break;
     case CV_8UC4:
     default:
-        cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+        cvtColor(img_rgb, img, COLOR_BGR2BGRA);
         break;
     }
-    cv::ocl::oclMat d_img(img);
+    ocl::oclMat d_img(img);
 
     // HOGs
-    cv::ocl::HOGDescriptor ocl_hog;
+    ocl::HOGDescriptor ocl_hog;
     ocl_hog.gamma_correction = true;
-    cv::HOGDescriptor hog;
+    HOGDescriptor hog;
     hog.gammaCorrection = true;
 
     // Compute descriptor
-    cv::ocl::oclMat d_descriptors;
+    ocl::oclMat d_descriptors;
     ocl_hog.getDescriptors(d_img, ocl_hog.win_size, d_descriptors, ocl_hog.DESCR_FORMAT_COL_BY_COL);
-    cv::Mat down_descriptors;
+    Mat down_descriptors;
     d_descriptors.download(down_descriptors);
     down_descriptors = down_descriptors.reshape(0, down_descriptors.cols * down_descriptors.rows);
 
@@ -105,149 +107,123 @@ TEST_P(HOG, GetDescriptors)
         hog.compute(img_rgb, descriptors, ocl_hog.win_size);
         break;
     }
-    cv::Mat cpu_descriptors(descriptors);
+    Mat cpu_descriptors(descriptors);
 
     EXPECT_MAT_SIMILAR(down_descriptors, cpu_descriptors, 1e-2);
 }
 
-
-bool match_rect(cv::Rect r1, cv::Rect r2, int threshold)
-{
-    return ((abs(r1.x - r2.x) < threshold) && (abs(r1.y - r2.y) < threshold) &&
-            (abs(r1.width - r2.width) < threshold) && (abs(r1.height - r2.height) < threshold));
-}
-
 TEST_P(HOG, Detect)
 {
-    // Load image
-    cv::Mat img_rgb = readImage(workdir + "lena.jpg");
-    ASSERT_FALSE(img_rgb.empty());
-
     // Convert image
-    cv::Mat img;
+    Mat img;
     switch (type)
     {
     case CV_8UC1:
-        cv::cvtColor(img_rgb, img, CV_BGR2GRAY);
+        cvtColor(img_rgb, img, COLOR_BGR2GRAY);
         break;
     case CV_8UC4:
     default:
-        cv::cvtColor(img_rgb, img, CV_BGR2BGRA);
+        cvtColor(img_rgb, img, COLOR_BGR2BGRA);
         break;
     }
-    cv::ocl::oclMat d_img(img);
+    ocl::oclMat d_img(img);
 
     // HOGs
-    if ((winSize != cv::Size(48, 96)) && (winSize != cv::Size(64, 128)))
-        winSize = cv::Size(64, 128);
-    cv::ocl::HOGDescriptor ocl_hog(winSize);
+    if ((winSize != Size(48, 96)) && (winSize != Size(64, 128)))
+        winSize = Size(64, 128);
+    ocl::HOGDescriptor ocl_hog(winSize);
     ocl_hog.gamma_correction = true;
 
-    cv::HOGDescriptor hog;
+    HOGDescriptor hog;
     hog.winSize = winSize;
     hog.gammaCorrection = true;
 
     if (winSize.width == 48 && winSize.height == 96)
     {
         // daimler's base
-        ocl_hog.setSVMDetector(ocl_hog.getPeopleDetector48x96());
+        ocl_hog.setSVMDetector(hog.getDaimlerPeopleDetector());
         hog.setSVMDetector(hog.getDaimlerPeopleDetector());
     }
     else if (winSize.width == 64 && winSize.height == 128)
     {
-        ocl_hog.setSVMDetector(ocl_hog.getPeopleDetector64x128());
+        ocl_hog.setSVMDetector(hog.getDefaultPeopleDetector());
         hog.setSVMDetector(hog.getDefaultPeopleDetector());
     }
     else
     {
-        ocl_hog.setSVMDetector(ocl_hog.getDefaultPeopleDetector());
+        ocl_hog.setSVMDetector(hog.getDefaultPeopleDetector());
         hog.setSVMDetector(hog.getDefaultPeopleDetector());
     }
 
     // OpenCL detection
-    std::vector<cv::Rect> d_found;
-    ocl_hog.detectMultiScale(d_img, d_found, 0, cv::Size(8, 8), cv::Size(0, 0), 1.05, 2);
+    std::vector<Rect> d_found;
+    ocl_hog.detectMultiScale(d_img, d_found, 0, Size(8, 8), Size(0, 0), 1.05, 6);
 
     // CPU detection
-    std::vector<cv::Rect> found;
+    std::vector<Rect> found;
     switch (type)
     {
     case CV_8UC1:
-        hog.detectMultiScale(img, found, 0, cv::Size(8, 8), cv::Size(0, 0), 1.05, 2);
+        hog.detectMultiScale(img, found, 0, Size(8, 8), Size(0, 0), 1.05, 6);
         break;
     case CV_8UC4:
     default:
-        hog.detectMultiScale(img_rgb, found, 0, cv::Size(8, 8), cv::Size(0, 0), 1.05, 2);
+        hog.detectMultiScale(img_rgb, found, 0, Size(8, 8), Size(0, 0), 1.05, 6);
         break;
     }
 
-    // Ground-truth rectangular people window
-    cv::Rect win1_64x128(231, 190, 72, 144);
-    cv::Rect win2_64x128(621, 156, 97, 194);
-    cv::Rect win1_48x96(238, 198, 63, 126);
-    cv::Rect win2_48x96(619, 161, 92, 185);
-    cv::Rect win3_48x96(488, 136, 56, 112);
-
-    // Compare whether ground-truth windows are detected and compare the number of windows detected.
-    std::vector<int> d_comp(4);
-    std::vector<int> comp(4);
-    for(int i = 0; i < (int)d_comp.size(); i++)
-    {
-        d_comp[i] = 0;
-        comp[i] = 0;
-    }
-
-    int threshold = 10;
-    int val = 32;
-    d_comp[0] = (int)d_found.size();
-    comp[0] = (int)found.size();
-    if (winSize == cv::Size(48, 96))
-    {
-        for(int i = 0; i < (int)d_found.size(); i++)
-        {
-            if (match_rect(d_found[i], win1_48x96, threshold))
-                d_comp[1] = val;
-            if (match_rect(d_found[i], win2_48x96, threshold))
-                d_comp[2] = val;
-            if (match_rect(d_found[i], win3_48x96, threshold))
-                d_comp[3] = val;
-        }
-        for(int i = 0; i < (int)found.size(); i++)
-        {
-            if (match_rect(found[i], win1_48x96, threshold))
-                comp[1] = val;
-            if (match_rect(found[i], win2_48x96, threshold))
-                comp[2] = val;
-            if (match_rect(found[i], win3_48x96, threshold))
-                comp[3] = val;
-        }
-    }
-    else if (winSize == cv::Size(64, 128))
-    {
-        for(int i = 0; i < (int)d_found.size(); i++)
-        {
-            if (match_rect(d_found[i], win1_64x128, threshold))
-                d_comp[1] = val;
-            if (match_rect(d_found[i], win2_64x128, threshold))
-                d_comp[2] = val;
-        }
-        for(int i = 0; i < (int)found.size(); i++)
-        {
-            if (match_rect(found[i], win1_64x128, threshold))
-                comp[1] = val;
-            if (match_rect(found[i], win2_64x128, threshold))
-                comp[2] = val;
-        }
-    }
-
-    char s[100] = {0};
-    EXPECT_MAT_NEAR(cv::Mat(d_comp), cv::Mat(comp), 3, s);
+    EXPECT_LT(checkRectSimilarity(img.size(), found, d_found), 1.0);
 }
 
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, HOG, testing::Combine(
-                            testing::Values(cv::Size(64, 128), cv::Size(48, 96)),
+INSTANTIATE_TEST_CASE_P(OCL_ObjDetect, HOG, testing::Combine(
+                            testing::Values(Size(64, 128), Size(48, 96)),
                             testing::Values(MatType(CV_8UC1), MatType(CV_8UC4))));
 
 
+///////////////////////////// Haar //////////////////////////////
+IMPLEMENT_PARAM_CLASS(CascadeName, std::string);
+CascadeName cascade_frontalface_alt(std::string("haarcascade_frontalface_alt.xml"));
+CascadeName cascade_frontalface_alt2(std::string("haarcascade_frontalface_alt2.xml"));
+
+PARAM_TEST_CASE(Haar, int, CascadeName)
+{
+    ocl::OclCascadeClassifier cascade, nestedCascade;
+    CascadeClassifier cpucascade, cpunestedCascade;
+
+    int flags;
+    std::string cascadeName;
+    std::vector<Rect> faces, oclfaces;
+    Mat img;
+    ocl::oclMat d_img;
+
+    virtual void SetUp()
+    {
+        flags = GET_PARAM(0);
+        cascadeName = (std::string(cvtest::TS::ptr()->get_data_path()) + "cv/cascadeandhog/cascades/").append(GET_PARAM(1));
+        ASSERT_TRUE(cascade.load( cascadeName ));
+        ASSERT_TRUE(cpucascade.load(cascadeName));
+        img = readImage("cv/shared/lena.png", IMREAD_GRAYSCALE);
+        ASSERT_FALSE(img.empty());
+        equalizeHist(img, img);
+        d_img.upload(img);
+    }
+};
+
+TEST_P(Haar, FaceDetect)
+{
+    cascade.detectMultiScale(d_img, oclfaces, 1.1, 3,
+                             flags, Size(30, 30));
+
+    cpucascade.detectMultiScale(img, faces, 1.1, 3,
+                                flags, Size(30, 30));
+
+    EXPECT_LT(checkRectSimilarity(img.size(), faces, oclfaces), 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_ObjDetect, Haar,
+    Combine(Values((int)CASCADE_SCALE_IMAGE, 0),
+            Values(cascade_frontalface_alt, cascade_frontalface_alt2)));
+
+
 #endif //HAVE_OPENCL
diff --git a/modules/ocl/test/test_optflow.cpp b/modules/ocl/test/test_optflow.cpp
new file mode 100644
index 0000000000..3b6c48050d
--- /dev/null
+++ b/modules/ocl/test/test_optflow.cpp
@@ -0,0 +1,344 @@
+///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+#include <iomanip>
+
+#ifdef HAVE_OPENCL
+
+using namespace cv;
+using namespace cv::ocl;
+using namespace cvtest;
+using namespace testing;
+
+//////////////////////////////////////////////////////
+// GoodFeaturesToTrack
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(MinDistance, double)
+}
+PARAM_TEST_CASE(GoodFeaturesToTrack, MinDistance)
+{
+    double minDistance;
+
+    virtual void SetUp()
+    {
+        minDistance = GET_PARAM(0);
+    }
+};
+
+TEST_P(GoodFeaturesToTrack, Accuracy)
+{
+    cv::Mat frame = readImage("gpu/opticalflow/rubberwhale1.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame.empty());
+
+    int maxCorners = 1000;
+    double qualityLevel = 0.01;
+
+    cv::ocl::GoodFeaturesToTrackDetector_OCL detector(maxCorners, qualityLevel, minDistance);
+
+    cv::ocl::oclMat d_pts;
+    detector(oclMat(frame), d_pts);
+
+    ASSERT_FALSE(d_pts.empty());
+
+    std::vector<cv::Point2f> pts(d_pts.cols);
+
+    detector.downloadPoints(d_pts, pts);
+
+    std::vector<cv::Point2f> pts_gold;
+    cv::goodFeaturesToTrack(frame, pts_gold, maxCorners, qualityLevel, minDistance);
+
+    ASSERT_EQ(pts_gold.size(), pts.size());
+
+    size_t mistmatch = 0;
+    for (size_t i = 0; i < pts.size(); ++i)
+    {
+        cv::Point2i a = pts_gold[i];
+        cv::Point2i b = pts[i];
+
+        bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
+
+        if (!eq)
+            ++mistmatch;
+    }
+
+    double bad_ratio = static_cast<double>(mistmatch) / pts.size();
+
+    ASSERT_LE(bad_ratio, 0.01);
+}
+
+TEST_P(GoodFeaturesToTrack, EmptyCorners)
+{
+    int maxCorners = 1000;
+    double qualityLevel = 0.01;
+
+    cv::ocl::GoodFeaturesToTrackDetector_OCL detector(maxCorners, qualityLevel, minDistance);
+
+    cv::ocl::oclMat src(100, 100, CV_8UC1, cv::Scalar::all(0));
+    cv::ocl::oclMat corners(1, maxCorners, CV_32FC2);
+
+    detector(src, corners);
+
+    ASSERT_TRUE(corners.empty());
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Video, GoodFeaturesToTrack,
+    testing::Values(MinDistance(0.0), MinDistance(3.0)));
+
+//////////////////////////////////////////////////////////////////////////
+PARAM_TEST_CASE(TVL1, bool)
+{
+    bool useRoi;
+
+    virtual void SetUp()
+    {
+        useRoi = GET_PARAM(0);
+    }
+
+};
+
+TEST_P(TVL1, DISABLED_Accuracy) // TODO implementations of TV1 in video module are different in 2.4 and master branches
+{
+    cv::Mat frame0 = readImage("gpu/opticalflow/rubberwhale1.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage("gpu/opticalflow/rubberwhale2.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    cv::ocl::OpticalFlowDual_TVL1_OCL d_alg;
+    cv::RNG &rng = TS::ptr()->get_rng();
+    cv::Mat flowx = randomMat(rng, frame0.size(), CV_32FC1, 0, 0, useRoi);
+    cv::Mat flowy = randomMat(rng, frame0.size(), CV_32FC1, 0, 0, useRoi);
+    cv::ocl::oclMat d_flowx(flowx), d_flowy(flowy);
+    d_alg(oclMat(frame0), oclMat(frame1), d_flowx, d_flowy);
+
+    cv::Ptr<cv::DenseOpticalFlow> alg = cv::createOptFlow_DualTVL1();
+    cv::Mat flow;
+    alg->calc(frame0, frame1, flow);
+    cv::Mat gold[2];
+    cv::split(flow, gold);
+
+    EXPECT_MAT_SIMILAR(gold[0], d_flowx, 3e-3);
+    EXPECT_MAT_SIMILAR(gold[1], d_flowy, 3e-3);
+}
+INSTANTIATE_TEST_CASE_P(OCL_Video, TVL1, Values(false, true));
+
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+// PyrLKOpticalFlow
+
+PARAM_TEST_CASE(Sparse, bool, bool)
+{
+    bool useGray;
+    bool UseSmart;
+
+    virtual void SetUp()
+    {
+        UseSmart = GET_PARAM(0);
+        useGray = GET_PARAM(1);
+    }
+};
+
+TEST_P(Sparse, Mat)
+{
+    cv::Mat frame0 = readImage("gpu/opticalflow/rubberwhale1.png", useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage("gpu/opticalflow/rubberwhale2.png", useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
+    ASSERT_FALSE(frame1.empty());
+
+    cv::Mat gray_frame;
+    if (useGray)
+        gray_frame = frame0;
+    else
+        cv::cvtColor(frame0, gray_frame, cv::COLOR_BGR2GRAY);
+
+    std::vector<cv::Point2f> pts;
+    cv::goodFeaturesToTrack(gray_frame, pts, 1000, 0.01, 0.0);
+
+    cv::ocl::oclMat d_pts;
+    cv::Mat pts_mat(1, (int)pts.size(), CV_32FC2, (void *)&pts[0]);
+    d_pts.upload(pts_mat);
+
+    cv::ocl::PyrLKOpticalFlow pyrLK;
+
+    cv::ocl::oclMat oclFrame0;
+    cv::ocl::oclMat oclFrame1;
+    cv::ocl::oclMat d_nextPts;
+    cv::ocl::oclMat d_status;
+    cv::ocl::oclMat d_err;
+
+    oclFrame0 = frame0;
+    oclFrame1 = frame1;
+
+    pyrLK.sparse(oclFrame0, oclFrame1, d_pts, d_nextPts, d_status, &d_err);
+
+    std::vector<cv::Point2f> nextPts(d_nextPts.cols);
+    cv::Mat nextPts_mat(1, d_nextPts.cols, CV_32FC2, (void *)&nextPts[0]);
+    d_nextPts.download(nextPts_mat);
+
+    std::vector<unsigned char> status(d_status.cols);
+    cv::Mat status_mat(1, d_status.cols, CV_8UC1, (void *)&status[0]);
+    d_status.download(status_mat);
+
+    std::vector<float> err(d_err.cols);
+    cv::Mat err_mat(1, d_err.cols, CV_32FC1, (void*)&err[0]);
+    d_err.download(err_mat);
+
+    std::vector<cv::Point2f> nextPts_gold;
+    std::vector<unsigned char> status_gold;
+    std::vector<float> err_gold;
+    cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts_gold, status_gold, err_gold);
+
+    ASSERT_EQ(nextPts_gold.size(), nextPts.size());
+    ASSERT_EQ(status_gold.size(), status.size());
+
+    size_t mistmatch = 0;
+    for (size_t i = 0; i < nextPts.size(); ++i)
+    {
+        if (status[i] != status_gold[i])
+        {
+            ++mistmatch;
+            continue;
+        }
+
+        if (status[i])
+        {
+            cv::Point2i a = nextPts[i];
+            cv::Point2i b = nextPts_gold[i];
+
+            bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
+            //float errdiff = std::abs(err[i] - err_gold[i]);
+            float errdiff = 0.0f;
+
+            if (!eq || errdiff > 1e-1)
+                ++mistmatch;
+        }
+    }
+
+    double bad_ratio = static_cast<double>(mistmatch) / (nextPts.size());
+
+    ASSERT_LE(bad_ratio, 0.02f);
+
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Video, Sparse, Combine(
+    Values(false, true),
+    Values(false, true)));
+//////////////////////////////////////////////////////
+// FarnebackOpticalFlow
+
+namespace
+{
+    IMPLEMENT_PARAM_CLASS(PyrScale, double)
+        IMPLEMENT_PARAM_CLASS(PolyN, int)
+        CV_FLAGS(FarnebackOptFlowFlags, 0, OPTFLOW_FARNEBACK_GAUSSIAN)
+        IMPLEMENT_PARAM_CLASS(UseInitFlow, bool)
+}
+
+PARAM_TEST_CASE(Farneback, PyrScale, PolyN, FarnebackOptFlowFlags, UseInitFlow)
+{
+    double pyrScale;
+    int polyN;
+    int flags;
+    bool useInitFlow;
+
+    virtual void SetUp()
+    {
+        pyrScale = GET_PARAM(0);
+        polyN = GET_PARAM(1);
+        flags = GET_PARAM(2);
+        useInitFlow = GET_PARAM(3);
+    }
+};
+
+TEST_P(Farneback, Accuracy)
+{
+    cv::Mat frame0 = readImage("gpu/opticalflow/rubberwhale1.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame0.empty());
+
+    cv::Mat frame1 = readImage("gpu/opticalflow/rubberwhale2.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(frame1.empty());
+
+    double polySigma = polyN <= 5 ? 1.1 : 1.5;
+
+    cv::ocl::FarnebackOpticalFlow farn;
+    farn.pyrScale = pyrScale;
+    farn.polyN = polyN;
+    farn.polySigma = polySigma;
+    farn.flags = flags;
+
+    cv::ocl::oclMat d_flowx, d_flowy;
+    farn(oclMat(frame0), oclMat(frame1), d_flowx, d_flowy);
+
+    cv::Mat flow;
+    if (useInitFlow)
+    {
+        cv::Mat flowxy[] = {cv::Mat(d_flowx), cv::Mat(d_flowy)};
+        cv::merge(flowxy, 2, flow);
+
+        farn.flags |= cv::OPTFLOW_USE_INITIAL_FLOW;
+        farn(oclMat(frame0), oclMat(frame1), d_flowx, d_flowy);
+    }
+
+    cv::calcOpticalFlowFarneback(
+        frame0, frame1, flow, farn.pyrScale, farn.numLevels, farn.winSize,
+        farn.numIters, farn.polyN, farn.polySigma, farn.flags);
+
+    std::vector<cv::Mat> flowxy;
+    cv::split(flow, flowxy);
+
+    EXPECT_MAT_SIMILAR(flowxy[0], d_flowx, 0.1);
+    EXPECT_MAT_SIMILAR(flowxy[1], d_flowy, 0.1);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_Video, Farneback, testing::Combine(
+    testing::Values(PyrScale(0.3), PyrScale(0.5), PyrScale(0.8)),
+    testing::Values(PolyN(5), PolyN(7)),
+    testing::Values(FarnebackOptFlowFlags(0), FarnebackOptFlowFlags(cv::OPTFLOW_FARNEBACK_GAUSSIAN)),
+    testing::Values(UseInitFlow(false), UseInitFlow(true))));
+
+#endif // HAVE_OPENCL
diff --git a/modules/ocl/test/precomp.hpp b/modules/ocl/test/test_precomp.hpp
similarity index 93%
rename from modules/ocl/test/precomp.hpp
rename to modules/ocl/test/test_precomp.hpp
index 25c95eb77a..72d5089d6b 100644
--- a/modules/ocl/test/precomp.hpp
+++ b/modules/ocl/test/test_precomp.hpp
@@ -61,20 +61,14 @@
 #include <iterator>
 #include <string>
 #include <cstdarg>
-#include "cvconfig.h"
-#include "opencv2/core.hpp"
+#include "opencv2/ts.hpp"
 #include "opencv2/highgui.hpp"
-//#include "opencv2/calib3d.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/video.hpp"
-#include "opencv2/ts.hpp"
-#include "opencv2/ts/ts_perf.hpp"
 #include "opencv2/ocl.hpp"
-//#include "opencv2/nonfree.hpp"
 
 #include "utility.hpp"
-#include "interpolation.hpp"
-//#include "add_test_info.h"
+
+#include "opencv2/core/private.hpp"
 
 #endif
-
diff --git a/modules/ocl/test/test_pyrdown.cpp b/modules/ocl/test/test_pyramids.cpp
similarity index 72%
rename from modules/ocl/test/test_pyrdown.cpp
rename to modules/ocl/test/test_pyramids.cpp
index c7233cc78a..58179ac185 100644
--- a/modules/ocl/test/test_pyrdown.cpp
+++ b/modules/ocl/test/test_pyramids.cpp
@@ -15,7 +15,6 @@
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
-//    Dachuan Zhao, dachuan@multicorewareinc.com
 //    Yao Wang yao@multicorewareinc.com
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -45,7 +44,7 @@
 //M*/
 
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #include <iomanip>
 
 #ifdef HAVE_OPENCL
@@ -56,54 +55,62 @@ using namespace cvtest;
 using namespace testing;
 using namespace std;
 
-PARAM_TEST_CASE(PyrDown, MatType, int)
+PARAM_TEST_CASE(PyrBase, MatType, int)
 {
     int type;
     int channels;
-
+    Mat dst_cpu;
+    oclMat gdst;
     virtual void SetUp()
     {
         type = GET_PARAM(0);
         channels = GET_PARAM(1);
-
-        //int devnums = getDevice(oclinfo);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
-    }
-
-    void Cleanup()
-    {
     }
 
 };
 
+/////////////////////// PyrDown //////////////////////////
+struct PyrDown : PyrBase {};
 
 TEST_P(PyrDown, Mat)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
     {
-        cv::Size size(MWIDTH, MHEIGHT);
-        cv::RNG &rng = TS::ptr()->get_rng();
-        cv::Mat src = randomMat(rng, size, CV_MAKETYPE(type, channels), 0, 100, false);
+        Size size(MWIDTH, MHEIGHT);
+        Mat src = randomMat(size, CV_MAKETYPE(type, channels));
+        oclMat gsrc(src);
 
-        cv::ocl::oclMat gsrc(src), gdst;
-        cv::Mat dst_cpu;
-        cv::pyrDown(src, dst_cpu);
-        cv::ocl::pyrDown(gsrc, gdst);
+        pyrDown(src, dst_cpu);
+        pyrDown(gsrc, gdst);
 
-        cv::Mat dst;
-        gdst.download(dst);
-        char s[1024] = {0};
-
-        EXPECT_MAT_NEAR(dst, dst_cpu, dst.depth() == CV_32F ? 1e-4f : 1.0f, s);
-
-        Cleanup();
+        EXPECT_MAT_NEAR(dst_cpu, Mat(gdst), type == CV_32F ? 1e-4f : 1.0f);
     }
 }
 
-INSTANTIATE_TEST_CASE_P(GPU_ImgProc, PyrDown, Combine(
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, PyrDown, Combine(
                             Values(CV_8U, CV_32F), Values(1, 3, 4)));
 
+/////////////////////// PyrUp //////////////////////////
 
+struct PyrUp : PyrBase {};
+
+TEST_P(PyrUp, Accuracy)
+{
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        Size size(MWIDTH, MHEIGHT);
+        Mat src = randomMat(size, CV_MAKETYPE(type, channels));
+        oclMat gsrc(src);
+
+        pyrUp(src, dst_cpu);
+        pyrUp(gsrc, gdst);
+
+        EXPECT_MAT_NEAR(dst_cpu, Mat(gdst), (type == CV_32F ? 1e-4f : 1.0));
+    }
+
+}
+
+
+INSTANTIATE_TEST_CASE_P(OCL_ImgProc, PyrUp, testing::Combine(
+                            Values(CV_8U, CV_32F), Values(1, 3, 4)));
 #endif // HAVE_OPENCL
diff --git a/modules/ocl/test/test_pyrlk.cpp b/modules/ocl/test/test_pyrlk.cpp
deleted file mode 100644
index 7c747ee4fc..0000000000
--- a/modules/ocl/test/test_pyrlk.cpp
+++ /dev/null
@@ -1,167 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-#include <iomanip>
-
-#ifdef HAVE_OPENCL
-
-using namespace cv;
-using namespace cv::ocl;
-using namespace cvtest;
-using namespace testing;
-using namespace std;
-
-//#define DUMP
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// BroxOpticalFlow
-extern string workdir;
-#define BROX_OPTICAL_FLOW_DUMP_FILE            "opticalflow/brox_optical_flow.bin"
-#define BROX_OPTICAL_FLOW_DUMP_FILE_CC20       "opticalflow/brox_optical_flow_cc20.bin"
-
-
-/////////////////////////////////////////////////////////////////////////////////////////////////
-// PyrLKOpticalFlow
-
-//IMPLEMENT_PARAM_CLASS(UseGray, bool)
-
-PARAM_TEST_CASE(Sparse, bool, bool)
-{
-    bool useGray;
-    bool UseSmart;
-
-    virtual void SetUp()
-    {
-        UseSmart = GET_PARAM(0);
-        useGray = GET_PARAM(0);
-    }
-};
-
-TEST_P(Sparse, Mat)
-{
-    cv::Mat frame0 = readImage(workdir + "../gpu/rubberwhale1.png", useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    ASSERT_FALSE(frame0.empty());
-
-    cv::Mat frame1 = readImage(workdir + "../gpu/rubberwhale2.png", useGray ? cv::IMREAD_GRAYSCALE : cv::IMREAD_COLOR);
-    ASSERT_FALSE(frame1.empty());
-
-    cv::Mat gray_frame;
-    if (useGray)
-        gray_frame = frame0;
-    else
-        cv::cvtColor(frame0, gray_frame, cv::COLOR_BGR2GRAY);
-
-    std::vector<cv::Point2f> pts;
-    cv::goodFeaturesToTrack(gray_frame, pts, 1000, 0.01, 0.0);
-
-    cv::ocl::oclMat d_pts;
-    cv::Mat pts_mat(1, (int)pts.size(), CV_32FC2, (void *)&pts[0]);
-    d_pts.upload(pts_mat);
-
-    cv::ocl::PyrLKOpticalFlow pyrLK;
-
-    cv::ocl::oclMat oclFrame0;
-    cv::ocl::oclMat oclFrame1;
-    cv::ocl::oclMat d_nextPts;
-    cv::ocl::oclMat d_status;
-    cv::ocl::oclMat d_err;
-
-    oclFrame0 = frame0;
-    oclFrame1 = frame1;
-
-    pyrLK.sparse(oclFrame0, oclFrame1, d_pts, d_nextPts, d_status, &d_err);
-
-    std::vector<cv::Point2f> nextPts(d_nextPts.cols);
-    cv::Mat nextPts_mat(1, d_nextPts.cols, CV_32FC2, (void *)&nextPts[0]);
-    d_nextPts.download(nextPts_mat);
-
-    std::vector<unsigned char> status(d_status.cols);
-    cv::Mat status_mat(1, d_status.cols, CV_8UC1, (void *)&status[0]);
-    d_status.download(status_mat);
-
-    std::vector<float> err(d_err.cols);
-    cv::Mat err_mat(1, d_err.cols, CV_32FC1, (void*)&err[0]);
-    d_err.download(err_mat);
-
-    std::vector<cv::Point2f> nextPts_gold;
-    std::vector<unsigned char> status_gold;
-    std::vector<float> err_gold;
-    cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts_gold, status_gold, err_gold);
-
-    ASSERT_EQ(nextPts_gold.size(), nextPts.size());
-    ASSERT_EQ(status_gold.size(), status.size());
-
-    size_t mistmatch = 0;
-    for (size_t i = 0; i < nextPts.size(); ++i)
-    {
-        if (status[i] != status_gold[i])
-        {
-            ++mistmatch;
-            continue;
-        }
-
-        if (status[i])
-        {
-            cv::Point2i a = nextPts[i];
-            cv::Point2i b = nextPts_gold[i];
-
-            bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
-            //float errdiff = std::abs(err[i] - err_gold[i]);
-            float errdiff = 0.0f;
-
-            if (!eq || errdiff > 1e-1)
-                ++mistmatch;
-        }
-    }
-
-    double bad_ratio = static_cast<double>(mistmatch) / (nextPts.size());
-
-    ASSERT_LE(bad_ratio, 0.02f);
-
-}
-
-INSTANTIATE_TEST_CASE_P(Video, Sparse, Combine(
-                            Values(false, true),
-                            Values(false)));
-
-#endif // HAVE_OPENCL
-
diff --git a/modules/ocl/test/test_sort.cpp b/modules/ocl/test/test_sort.cpp
new file mode 100644
index 0000000000..d303665568
--- /dev/null
+++ b/modules/ocl/test/test_sort.cpp
@@ -0,0 +1,244 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Peng Xiao, pengxiao@outlook.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include <map>
+#include <functional>
+#include "test_precomp.hpp"
+
+using namespace std;
+using namespace cvtest;
+using namespace testing;
+using namespace cv;
+
+
+namespace
+{
+IMPLEMENT_PARAM_CLASS(IsGreaterThan, bool)
+IMPLEMENT_PARAM_CLASS(InputSize, int)
+IMPLEMENT_PARAM_CLASS(SortMethod, int)
+
+
+template<class T>
+struct KV_CVTYPE{ static int toType() {return 0;} };
+
+template<> struct KV_CVTYPE<int>  { static int toType() {return CV_32SC1;} };
+template<> struct KV_CVTYPE<float>{ static int toType() {return CV_32FC1;} };
+template<> struct KV_CVTYPE<Vec2i>{ static int toType() {return CV_32SC2;} };
+template<> struct KV_CVTYPE<Vec2f>{ static int toType() {return CV_32FC2;} };
+
+template<class key_type, class val_type>
+bool kvgreater(pair<key_type, val_type> p1, pair<key_type, val_type> p2)
+{
+    return p1.first > p2.first;
+}
+
+template<class key_type, class val_type>
+bool kvless(pair<key_type, val_type> p1, pair<key_type, val_type> p2)
+{
+    return p1.first < p2.first;
+}
+
+template<class key_type, class val_type>
+void toKVPair(
+    MatConstIterator_<key_type> kit,
+    MatConstIterator_<val_type> vit,
+    int vecSize,
+    vector<pair<key_type, val_type> >& kvres
+    )
+{
+    kvres.clear();
+    for(int i = 0; i < vecSize; i ++)
+    {
+        kvres.push_back(make_pair(*kit, *vit));
+        ++kit;
+        ++vit;
+    }
+}
+
+template<class key_type, class val_type>
+void kvquicksort(Mat& keys, Mat& vals, bool isGreater = false)
+{
+    vector<pair<key_type, val_type> > kvres;
+    toKVPair(keys.begin<key_type>(), vals.begin<val_type>(), keys.cols, kvres);
+
+    if(isGreater)
+    {
+        std::sort(kvres.begin(), kvres.end(), kvgreater<key_type, val_type>);
+    }
+    else
+    {
+        std::sort(kvres.begin(), kvres.end(), kvless<key_type, val_type>);
+    }
+    key_type * kptr = keys.ptr<key_type>();
+    val_type * vptr = vals.ptr<val_type>();
+    for(int i = 0; i < keys.cols; i ++)
+    {
+        kptr[i] = kvres[i].first;
+        vptr[i] = kvres[i].second;
+    }
+}
+
+class SortByKey_STL
+{
+public:
+    static void sort(cv::Mat&, cv::Mat&, bool is_gt);
+private:
+    typedef void (*quick_sorter)(cv::Mat&, cv::Mat&, bool);
+    SortByKey_STL();
+    quick_sorter quick_sorters[CV_64FC4][CV_64FC4];
+    static SortByKey_STL instance;
+};
+
+SortByKey_STL SortByKey_STL::instance = SortByKey_STL();
+
+SortByKey_STL::SortByKey_STL()
+{
+    memset(instance.quick_sorters, 0, sizeof(quick_sorters));
+#define NEW_SORTER(KT, VT) \
+    instance.quick_sorters[KV_CVTYPE<KT>::toType()][KV_CVTYPE<VT>::toType()] = kvquicksort<KT, VT>;
+
+    NEW_SORTER(int, int);
+    NEW_SORTER(int, Vec2i);
+    NEW_SORTER(int, float);
+    NEW_SORTER(int, Vec2f);
+
+    NEW_SORTER(float, int);
+    NEW_SORTER(float, Vec2i);
+    NEW_SORTER(float, float);
+    NEW_SORTER(float, Vec2f);
+#undef NEW_SORTER
+}
+
+void SortByKey_STL::sort(cv::Mat& keys, cv::Mat& vals, bool is_gt)
+{
+    instance.quick_sorters[keys.type()][vals.type()](keys, vals, is_gt);
+}
+
+bool checkUnstableSorterResult(const Mat& gkeys_, const Mat& gvals_,
+                               const Mat& /*dkeys_*/, const Mat& dvals_)
+{
+    int cn_val = gvals_.channels();
+    int count  = gkeys_.cols;
+
+    //for convenience we convert depth to float and channels to 1
+    Mat gkeys, gvals, dkeys, dvals;
+    gkeys_.reshape(1).convertTo(gkeys, CV_32F);
+    gvals_.reshape(1).convertTo(gvals, CV_32F);
+    //dkeys_.reshape(1).convertTo(dkeys, CV_32F);
+    dvals_.reshape(1).convertTo(dvals, CV_32F);
+    float * gkptr = gkeys.ptr<float>();
+    float * gvptr = gvals.ptr<float>();
+    //float * dkptr = dkeys.ptr<float>();
+    float * dvptr = dvals.ptr<float>();
+
+    for(int i = 0; i < count - 1; ++i)
+    {
+        int iden_count = 0;
+        // firstly calculate the number of identical keys
+        while(gkptr[i + iden_count] == gkptr[i + 1 + iden_count])
+        {
+            ++ iden_count;
+        }
+
+        // sort dv and gv
+        int num_of_val = (iden_count + 1) * cn_val;
+        std::sort(gvptr + i * cn_val, gvptr + i * cn_val + num_of_val);
+        std::sort(dvptr + i * cn_val, dvptr + i * cn_val + num_of_val);
+
+        // then check if [i, i + iden_count) is the same
+        for(int j = 0; j < num_of_val; ++j)
+        {
+            if(gvptr[i + j] != dvptr[i + j])
+            {
+                return false;
+            }
+        }
+        i += iden_count;
+    }
+    return true;
+}
+}
+
+#define INPUT_SIZES  Values(InputSize(0x10), InputSize(0x100), InputSize(0x10000)) //2^4, 2^8, 2^16
+#define KEY_TYPES    Values(MatType(CV_32SC1), MatType(CV_32FC1))
+#define VAL_TYPES    Values(MatType(CV_32SC1), MatType(CV_32SC2), MatType(CV_32FC1), MatType(CV_32FC2))
+#define SORT_METHODS Values(SortMethod(cv::ocl::SORT_BITONIC),SortMethod(cv::ocl::SORT_MERGE),SortMethod(cv::ocl::SORT_RADIX)/*,SortMethod(cv::ocl::SORT_SELECTION)*/)
+#define F_OR_T       Values(IsGreaterThan(false), IsGreaterThan(true))
+
+PARAM_TEST_CASE(SortByKey, InputSize, MatType, MatType, SortMethod, IsGreaterThan)
+{
+    InputSize input_size;
+    MatType key_type, val_type;
+    SortMethod method;
+    IsGreaterThan is_gt;
+
+    Mat mat_key, mat_val;
+    virtual void SetUp()
+    {
+        input_size = GET_PARAM(0);
+        key_type   = GET_PARAM(1);
+        val_type   = GET_PARAM(2);
+        method     = GET_PARAM(3);
+        is_gt      = GET_PARAM(4);
+
+        using namespace cv;
+        // fill key and val
+        mat_key = randomMat(Size(input_size, 1), key_type, INT_MIN, INT_MAX);
+        mat_val = randomMat(Size(input_size, 1), val_type, INT_MIN, INT_MAX);
+    }
+};
+
+TEST_P(SortByKey, Accuracy)
+{
+    using namespace cv;
+    ocl::oclMat oclmat_key(mat_key);
+    ocl::oclMat oclmat_val(mat_val);
+
+    ocl::sortByKey(oclmat_key, oclmat_val, method, is_gt);
+    SortByKey_STL::sort(mat_key, mat_val, is_gt);
+
+    EXPECT_MAT_NEAR(mat_key, oclmat_key, 0.0);
+    EXPECT_TRUE(checkUnstableSorterResult(mat_key, mat_val, oclmat_key, oclmat_val));
+}
+INSTANTIATE_TEST_CASE_P(OCL_SORT, SortByKey, Combine(INPUT_SIZES, KEY_TYPES, VAL_TYPES, SORT_METHODS, F_OR_T));
diff --git a/modules/ocl/test/test_split_merge.cpp b/modules/ocl/test/test_split_merge.cpp
index f41d16eeee..9663f5321c 100644
--- a/modules/ocl/test/test_split_merge.cpp
+++ b/modules/ocl/test/test_split_merge.cpp
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // @Authors
@@ -43,7 +44,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 
 #ifdef HAVE_OPENCL
 
@@ -87,7 +88,7 @@ PARAM_TEST_CASE(MergeTestBase, MatType, int)
 
     //dst mat with roi
     cv::Mat dst_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst_whole;
 
@@ -112,10 +113,6 @@ PARAM_TEST_CASE(MergeTestBase, MatType, int)
         mat4 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
         dst  = randomMat(rng, size, CV_MAKETYPE(type, channels), 5, 16, false);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -205,12 +202,7 @@ TEST_P(Merge, Accuracy)
         cv::merge(dev_src, dst_roi);
         cv::ocl::merge(dev_gsrc, gdst);
 
-        cv::Mat cpu_dst;
-        gdst_whole.download(cpu_dst);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,src1x =%d,src1y=%d,src2x =%d,src2y=%d,src3x =%d,src3y=%d,src4x =%d,src4y=%d,dstx=%d,dsty=%d", roicols, roirows, src1x, src1y, src2x , src2y, src3x , src3y, src4x , src4y, dstx, dsty);
-
-        EXPECT_MAT_NEAR(dst, cpu_dst, 0.0, sss);
+        EXPECT_MAT_NEAR(dst, Mat(gdst_whole), 0.0);
     }
 }
 
@@ -252,7 +244,7 @@ PARAM_TEST_CASE(SplitTestBase, MatType, int)
     cv::Mat dst2_roi;
     cv::Mat dst3_roi;
     cv::Mat dst4_roi;
-    //std::vector<cv::ocl::Info> oclinfo;
+
     //ocl dst mat for testing
     cv::ocl::oclMat gdst1_whole;
     cv::ocl::oclMat gdst2_whole;
@@ -280,10 +272,6 @@ PARAM_TEST_CASE(SplitTestBase, MatType, int)
         dst3 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
         dst4 = randomMat(rng, size, CV_MAKETYPE(type, 1), 5, 16, false);
 
-        //int devnums = getDevice(oclinfo, OPENCV_DEFAULT_OPENCL_DEVICE);
-        //CV_Assert(devnums > 0);
-        ////if you want to use undefault device, set it here
-        ////setDevice(oclinfo[0]);
     }
 
     void random_roi()
@@ -356,28 +344,17 @@ TEST_P(Split, Accuracy)
         cv::split(mat_roi, dev_dst);
         cv::ocl::split(gmat, dev_gdst);
 
-        cv::Mat cpu_dst1;
-        cv::Mat cpu_dst2;
-        cv::Mat cpu_dst3;
-        cv::Mat cpu_dst4;
-        gdst1_whole.download(cpu_dst1);
-        gdst2_whole.download(cpu_dst2);
-        gdst3_whole.download(cpu_dst3);
-        gdst4_whole.download(cpu_dst4);
-        char sss[1024];
-        sprintf(sss, "roicols=%d,roirows=%d,dst1x =%d,dsty=%d,dst2x =%d,dst2y=%d,dst3x =%d,dst3y=%d,dst4x =%d,dst4y=%d,srcx=%d,srcy=%d", roicols, roirows, dst1x , dst1y, dst2x , dst2y, dst3x , dst3y, dst4x , dst4y, srcx, srcy);
-
         if(channels >= 1)
-            EXPECT_MAT_NEAR(dst1, cpu_dst1, 0.0, sss);
+            EXPECT_MAT_NEAR(dst1, Mat(gdst1_whole), 0.0);
 
         if(channels >= 2)
-            EXPECT_MAT_NEAR(dst2, cpu_dst2, 0.0, sss);
+            EXPECT_MAT_NEAR(dst2, Mat(gdst2_whole), 0.0);
 
         if(channels >= 3)
-            EXPECT_MAT_NEAR(dst3, cpu_dst3, 0.0, sss);
+            EXPECT_MAT_NEAR(dst3, Mat(gdst3_whole), 0.0);
 
         if(channels >= 4)
-            EXPECT_MAT_NEAR(dst4, cpu_dst4, 0.0, sss);
+            EXPECT_MAT_NEAR(dst4, Mat(gdst4_whole), 0.0);
     }
 }
 
diff --git a/modules/ocl/test/utility.cpp b/modules/ocl/test/utility.cpp
index 4b21081a8b..f8d585cd40 100644
--- a/modules/ocl/test/utility.cpp
+++ b/modules/ocl/test/utility.cpp
@@ -39,7 +39,7 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#include "test_precomp.hpp"
 #define VARNAME(A) #A
 using namespace std;
 using namespace cv;
@@ -100,12 +100,44 @@ Mat randomMat(Size size, int type, double minVal, double maxVal)
     return randomMat(TS::ptr()->get_rng(), size, type, minVal, maxVal, false);
 }
 
+cv::ocl::oclMat createMat_ocl(Size size, int type, bool useRoi)
+{
+    Size size0 = size;
 
+    if (useRoi)
+    {
+        size0.width += randomInt(5, 15);
+        size0.height += randomInt(5, 15);
+    }
 
+    cv::ocl::oclMat d_m(size0, type);
 
+    if (size0 != size)
+        d_m = d_m(Rect((size0.width - size.width) / 2, (size0.height - size.height) / 2, size.width, size.height));
 
+    return d_m;
+}
 
+cv::ocl::oclMat loadMat_ocl(const Mat& m, bool useRoi)
+{
+    CV_Assert(m.type() == CV_8UC1 || m.type() == CV_8UC3);
+    cv::ocl::oclMat d_m;
+    d_m = createMat_ocl(m.size(), m.type(), useRoi);
 
+    Size ls;
+    Point pt;
+
+    d_m.locateROI(ls, pt);
+
+    Rect roi(pt.x, pt.y, d_m.size().width, d_m.size().height);
+
+    cv::ocl::oclMat m_ocl(m);
+
+    cv::ocl::oclMat d_m_roi(d_m, roi);
+
+    m_ocl.copyTo(d_m);
+    return d_m;
+}
 /*
 void showDiff(InputArray gold_, InputArray actual_, double eps)
 {
@@ -137,58 +169,7 @@ void showDiff(InputArray gold_, InputArray actual_, double eps)
 }
 */
 
-/*
-bool supportFeature(const DeviceInfo& info, FeatureSet feature)
-{
-    return TargetArchs::builtWith(feature) && info.supports(feature);
-}
 
-const vector<DeviceInfo>& devices()
-{
-    static vector<DeviceInfo> devs;
-    static bool first = true;
-
-    if (first)
-    {
-        int deviceCount = getCudaEnabledDeviceCount();
-
-        devs.reserve(deviceCount);
-
-        for (int i = 0; i < deviceCount; ++i)
-        {
-            DeviceInfo info(i);
-            if (info.isCompatible())
-                devs.push_back(info);
-        }
-
-        first = false;
-    }
-
-    return devs;
-}
-
-vector<DeviceInfo> devices(FeatureSet feature)
-{
-    const vector<DeviceInfo>& d = devices();
-
-    vector<DeviceInfo> devs_filtered;
-
-    if (TargetArchs::builtWith(feature))
-    {
-        devs_filtered.reserve(d.size());
-
-        for (size_t i = 0, size = d.size(); i < size; ++i)
-        {
-            const DeviceInfo& info = d[i];
-
-            if (info.supports(feature))
-                devs_filtered.push_back(info);
-        }
-    }
-
-    return devs_filtered;
-}
-*/
 
 vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end)
 {
@@ -245,7 +226,7 @@ double checkNorm(const Mat &m1, const Mat &m2)
 double checkSimilarity(const Mat &m1, const Mat &m2)
 {
     Mat diff;
-    matchTemplate(m1, m2, diff, CV_TM_CCORR_NORMED);
+    matchTemplate(m1, m2, diff, TM_CCORR_NORMED);
     return std::abs(diff.at<float>(0, 0) - 1.f);
 }
 
@@ -264,3 +245,47 @@ void PrintTo(const Inverse &inverse, std::ostream *os)
         (*os) << "direct";
 }
 
+double checkRectSimilarity(Size sz, std::vector<Rect>& ob1, std::vector<Rect>& ob2)
+{
+    double final_test_result = 0.0;
+    size_t sz1 = ob1.size();
+    size_t sz2 = ob2.size();
+
+    if(sz1 != sz2)
+    {
+        return sz1 > sz2 ? (double)(sz1 - sz2) : (double)(sz2 - sz1);
+    }
+    else
+    {
+        if(sz1==0 && sz2==0)
+            return 0;
+        cv::Mat cpu_result(sz, CV_8UC1);
+        cpu_result.setTo(0);
+
+        for(vector<Rect>::const_iterator r = ob1.begin(); r != ob1.end(); r++)
+        {
+            cv::Mat cpu_result_roi(cpu_result, *r);
+            cpu_result_roi.setTo(1);
+            cpu_result.copyTo(cpu_result);
+        }
+        int cpu_area = cv::countNonZero(cpu_result > 0);
+
+        cv::Mat gpu_result(sz, CV_8UC1);
+        gpu_result.setTo(0);
+        for(vector<Rect>::const_iterator r2 = ob2.begin(); r2 != ob2.end(); r2++)
+        {
+            cv::Mat gpu_result_roi(gpu_result, *r2);
+            gpu_result_roi.setTo(1);
+            gpu_result.copyTo(gpu_result);
+        }
+
+        cv::Mat result_;
+        multiply(cpu_result, gpu_result, result_);
+        int result = cv::countNonZero(result_ > 0);
+        if(cpu_area!=0 && result!=0)
+            final_test_result = 1.0 - (double)result/(double)cpu_area;
+        else if(cpu_area==0 && result!=0)
+            final_test_result = -1;
+    }
+    return final_test_result;
+}
diff --git a/modules/ocl/test/utility.hpp b/modules/ocl/test/utility.hpp
index 99fd3346d5..ade7620b09 100644
--- a/modules/ocl/test/utility.hpp
+++ b/modules/ocl/test/utility.hpp
@@ -41,6 +41,8 @@
 
 #ifndef __OPENCV_TEST_UTILITY_HPP__
 #define __OPENCV_TEST_UTILITY_HPP__
+#include "opencv2/core.hpp"
+
 #define LOOP_TIMES 1
 #define MWIDTH 256
 #define MHEIGHT 256
@@ -55,13 +57,12 @@ cv::Mat randomMat(cv::Size size, int type, double minVal = 0.0, double maxVal =
 
 void showDiff(cv::InputArray gold, cv::InputArray actual, double eps);
 
-//! return true if device supports specified feature and gpu module was built with support the feature.
-//bool supportFeature(const cv::gpu::DeviceInfo& info, cv::gpu::FeatureSet feature);
+// This function test if gpu_rst matches cpu_rst.
+// If the two vectors are not equal, it will return the difference in vector size
+// Else it will return (total diff of each cpu and gpu rects covered pixels)/(total cpu rects covered pixels)
+// The smaller, the better matched
+double checkRectSimilarity(cv::Size sz, std::vector<cv::Rect>& ob1, std::vector<cv::Rect>& ob2);
 
-//! return all devices compatible with current gpu module build.
-//const std::vector<cv::ocl::DeviceInfo>& devices();
-//! return all devices compatible with current gpu module build which support specified feature.
-//std::vector<cv::ocl::DeviceInfo> devices(cv::gpu::FeatureSet feature);
 
 //! read image from testdata folder.
 cv::Mat readImage(const std::string &fileName, int flags = cv::IMREAD_COLOR);
@@ -71,25 +72,28 @@ double checkNorm(const cv::Mat &m);
 double checkNorm(const cv::Mat &m1, const cv::Mat &m2);
 double checkSimilarity(const cv::Mat &m1, const cv::Mat &m2);
 
+//oclMat create
+cv::ocl::oclMat createMat_ocl(cv::Size size, int type, bool useRoi = false);
+cv::ocl::oclMat loadMat_ocl(const cv::Mat& m, bool useRoi = false);
 #define EXPECT_MAT_NORM(mat, eps) \
 { \
     EXPECT_LE(checkNorm(cv::Mat(mat)), eps) \
 }
 
-/*#define EXPECT_MAT_NEAR(mat1, mat2, eps) \
+#define EXPECT_MAT_NEAR(mat1, mat2, eps) \
 { \
    ASSERT_EQ(mat1.type(), mat2.type()); \
    ASSERT_EQ(mat1.size(), mat2.size()); \
    EXPECT_LE(checkNorm(cv::Mat(mat1), cv::Mat(mat2)), eps); \
-}*/
-
+}
+/*
 #define EXPECT_MAT_NEAR(mat1, mat2, eps,s) \
 { \
     ASSERT_EQ(mat1.type(), mat2.type()); \
     ASSERT_EQ(mat1.size(), mat2.size()); \
     EXPECT_LE(checkNorm(cv::Mat(mat1), cv::Mat(mat2)), eps)<<s; \
 }
-
+*/
 #define EXPECT_MAT_SIMILAR(mat1, mat2, eps) \
 { \
     ASSERT_EQ(mat1.type(), mat2.type()); \
@@ -130,30 +134,21 @@ private:
 
 void PrintTo(const Inverse &useRoi, std::ostream *os);
 
-CV_ENUM(CmpCode, cv::CMP_EQ, cv::CMP_GT, cv::CMP_GE, cv::CMP_LT, cv::CMP_LE, cv::CMP_NE)
-
-CV_ENUM(NormCode, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_TYPE_MASK, cv::NORM_RELATIVE, cv::NORM_MINMAX)
-
 enum {FLIP_BOTH = 0, FLIP_X = 1, FLIP_Y = -1};
 CV_ENUM(FlipCode, FLIP_BOTH, FLIP_X, FLIP_Y)
 
-CV_ENUM(ReduceOp, CV_REDUCE_SUM, CV_REDUCE_AVG, CV_REDUCE_MAX, CV_REDUCE_MIN)
+CV_ENUM(CmpCode, CMP_EQ, CMP_GT, CMP_GE, CMP_LT, CMP_LE, CMP_NE)
+CV_ENUM(NormCode, NORM_INF, NORM_L1, NORM_L2, NORM_TYPE_MASK, NORM_RELATIVE, NORM_MINMAX)
+CV_ENUM(ReduceOp, REDUCE_SUM, REDUCE_AVG, REDUCE_MAX, REDUCE_MIN)
+CV_ENUM(MorphOp, MORPH_OPEN, MORPH_CLOSE, MORPH_GRADIENT, MORPH_TOPHAT, MORPH_BLACKHAT)
+CV_ENUM(ThreshOp, THRESH_BINARY, THRESH_BINARY_INV, THRESH_TRUNC, THRESH_TOZERO, THRESH_TOZERO_INV)
+CV_ENUM(Interpolation, INTER_NEAREST, INTER_LINEAR, INTER_CUBIC)
+CV_ENUM(Border, BORDER_REFLECT101, BORDER_REPLICATE, BORDER_CONSTANT, BORDER_REFLECT, BORDER_WRAP)
+CV_ENUM(TemplateMethod, TM_SQDIFF, TM_SQDIFF_NORMED, TM_CCORR, TM_CCORR_NORMED, TM_CCOEFF, TM_CCOEFF_NORMED)
 
-CV_FLAGS(GemmFlags, cv::GEMM_1_T, cv::GEMM_2_T, cv::GEMM_3_T);
-
-CV_ENUM(MorphOp, cv::MORPH_OPEN, cv::MORPH_CLOSE, cv::MORPH_GRADIENT, cv::MORPH_TOPHAT, cv::MORPH_BLACKHAT)
-
-CV_ENUM(ThreshOp, cv::THRESH_BINARY, cv::THRESH_BINARY_INV, cv::THRESH_TRUNC, cv::THRESH_TOZERO, cv::THRESH_TOZERO_INV)
-
-CV_ENUM(Interpolation, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC)
-
-CV_ENUM(Border, cv::BORDER_REFLECT101, cv::BORDER_REPLICATE, cv::BORDER_CONSTANT, cv::BORDER_REFLECT, cv::BORDER_WRAP)
-
-CV_FLAGS(WarpFlags, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::WARP_INVERSE_MAP)
-
-CV_ENUM(TemplateMethod, cv::TM_SQDIFF, cv::TM_SQDIFF_NORMED, cv::TM_CCORR, cv::TM_CCORR_NORMED, cv::TM_CCOEFF, cv::TM_CCOEFF_NORMED)
-
-CV_FLAGS(DftFlags, cv::DFT_INVERSE, cv::DFT_SCALE, cv::DFT_ROWS, cv::DFT_COMPLEX_OUTPUT, cv::DFT_REAL_OUTPUT)
+CV_FLAGS(GemmFlags, GEMM_1_T, GEMM_2_T, GEMM_3_T);
+CV_FLAGS(WarpFlags, INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, WARP_INVERSE_MAP)
+CV_FLAGS(DftFlags, DFT_INVERSE, DFT_SCALE, DFT_ROWS, DFT_COMPLEX_OUTPUT, DFT_REAL_OUTPUT)
 
 void  run_perf_test();
 
diff --git a/modules/optim/CMakeLists.txt b/modules/optim/CMakeLists.txt
new file mode 100644
index 0000000000..c36c24d9de
--- /dev/null
+++ b/modules/optim/CMakeLists.txt
@@ -0,0 +1,2 @@
+set(the_description "Generic optimization")
+ocv_define_module(optim opencv_core)
diff --git a/modules/optim/doc/downhill_simplex_method.rst b/modules/optim/doc/downhill_simplex_method.rst
new file mode 100644
index 0000000000..94d084c23f
--- /dev/null
+++ b/modules/optim/doc/downhill_simplex_method.rst
@@ -0,0 +1,161 @@
+Downhill Simplex Method
+=======================
+
+.. highlight:: cpp
+
+optim::DownhillSolver
+---------------------------------
+
+.. ocv:class:: optim::DownhillSolver
+
+This class is used to perform the non-linear non-constrained *minimization* of a function, given on an *n*-dimensional Euclidean space,
+using the **Nelder-Mead method**, also known as **downhill simplex method**. The basic idea about the method can be obtained from
+(`http://en.wikipedia.org/wiki/Nelder-Mead\_method <http://en.wikipedia.org/wiki/Nelder-Mead_method>`_). It should be noted, that
+this method, although deterministic, is rather a heuristic and therefore may converge to a local minima, not necessary a global one.
+It is iterative optimization technique, which at each step uses an information about the values of a function evaluated only at
+*n+1* points, arranged as a *simplex* in *n*-dimensional space (hence the second name of the method). At each step new point is
+chosen to evaluate function at, obtained value is compared with previous ones and based on this information simplex changes it's shape
+, slowly moving to the local minimum.
+
+Algorithm stops when the number of function evaluations done exceeds ``termcrit.maxCount``, when the function values at the
+vertices of simplex are within ``termcrit.epsilon`` range or simplex becomes so small that it
+can enclosed in a box with ``termcrit.epsilon`` sides, whatever comes first, for some defined by user
+positive integer ``termcrit.maxCount`` and positive non-integer ``termcrit.epsilon``.
+
+::
+
+    class CV_EXPORTS Solver : public Algorithm
+    {
+    public:
+        class CV_EXPORTS Function
+        {
+        public:
+           virtual ~Function() {}
+           //! ndim - dimensionality
+           virtual double calc(const double* x) const = 0;
+        };
+
+        virtual Ptr<Function> getFunction() const = 0;
+        virtual void setFunction(const Ptr<Function>& f) = 0;
+
+        virtual TermCriteria getTermCriteria() const = 0;
+        virtual void setTermCriteria(const TermCriteria& termcrit) = 0;
+
+        // x contain the initial point before the call and the minima position (if algorithm converged) after. x is assumed to be (something that
+        // after getMat() will return) row-vector or column-vector. *It's size  and should
+        // be consisted with previous dimensionality data given, if any (otherwise, it determines dimensionality)*
+        virtual double minimize(InputOutputArray x) = 0;
+    };
+
+    class CV_EXPORTS DownhillSolver : public Solver
+    {
+    public:
+        //! returns row-vector, even if the column-vector was given
+        virtual void getInitStep(OutputArray step) const=0;
+        //!This should be called at least once before the first call to minimize() and step is assumed to be (something that
+        //! after getMat() will return) row-vector or column-vector. *It's dimensionality determines the dimensionality of a problem.*
+        virtual void setInitStep(InputArray step)=0;
+    };
+
+It should be noted, that ``optim::DownhillSolver`` is a derivative of the abstract interface ``optim::Solver``, which in
+turn is derived from the ``Algorithm`` interface and is used to encapsulate the functionality, common to all non-linear optimization
+algorithms in the ``optim`` module.
+
+optim::DownhillSolver::getFunction
+--------------------------------------------
+
+Getter for the optimized function. The optimized function is represented by ``Solver::Function`` interface, which requires
+derivatives to implement the sole method ``calc(double*)`` to evaluate the function.
+
+.. ocv:function:: Ptr<Solver::Function> optim::DownhillSolver::getFunction()
+
+    :return: Smart-pointer to an object that implements ``Solver::Function`` interface - it represents the function that is being optimized. It can be empty, if no function was given so far.
+
+optim::DownhillSolver::setFunction
+-----------------------------------------------
+
+Setter for the optimized function. *It should be called at least once before the call to* ``DownhillSolver::minimize()``, as
+default value is not usable.
+
+.. ocv:function:: void optim::DownhillSolver::setFunction(const Ptr<Solver::Function>& f)
+
+    :param f: The new function to optimize.
+
+optim::DownhillSolver::getTermCriteria
+----------------------------------------------------
+
+Getter for the previously set terminal criteria for this algorithm.
+
+.. ocv:function:: TermCriteria optim::DownhillSolver::getTermCriteria()
+
+    :return: Deep copy of the terminal criteria used at the moment.
+
+optim::DownhillSolver::setTermCriteria
+------------------------------------------
+
+Set terminal criteria for downhill simplex method. Two things should be noted. First, this method *is not necessary* to be called
+before the first call to ``DownhillSolver::minimize()``, as the default value is sensible. Second, the method will raise an error
+if ``termcrit.type!=(TermCriteria::MAX_ITER+TermCriteria::EPS)``, ``termcrit.epsilon<=0`` or ``termcrit.maxCount<=0``. That is,
+both ``epsilon`` and ``maxCount`` should be set to positive values (non-integer and integer respectively) and they represent
+tolerance and maximal number of function evaluations that is allowed.
+
+Algorithm stops when the number of function evaluations done exceeds ``termcrit.maxCount``, when the function values at the
+vertices of simplex are within ``termcrit.epsilon`` range or simplex becomes so small that it
+can enclosed in a box with ``termcrit.epsilon`` sides, whatever comes first.
+
+.. ocv:function:: void optim::DownhillSolver::setTermCriteria(const TermCriteria& termcrit)
+
+    :param termcrit: Terminal criteria to be used, represented as ``TermCriteria`` structure (defined elsewhere in openCV). Mind you, that it should meet ``(termcrit.type==(TermCriteria::MAX_ITER+TermCriteria::EPS) && termcrit.epsilon>0 && termcrit.maxCount>0)``, otherwise the error will be raised.
+
+optim::DownhillSolver::getInitStep
+-----------------------------------
+
+Returns the initial step that will be used in downhill simplex algorithm. See the description
+of corresponding setter (follows next) for the meaning of this parameter.
+
+.. ocv:function:: void optim::getInitStep(OutputArray step)
+
+    :param step: Initial step that will be used in algorithm. Note, that although corresponding setter accepts column-vectors as well as row-vectors, this method will return a row-vector.
+
+optim::DownhillSolver::setInitStep
+----------------------------------
+
+Sets the initial step that will be used in downhill simplex algorithm. Step, together with initial point (givin in ``DownhillSolver::minimize``)
+are two *n*-dimensional vectors that are used to determine the shape of initial simplex. Roughly said, initial point determines the position
+of a simplex (it will become simplex's centroid), while step determines the spread (size in each dimension) of a simplex. To be more precise,
+if :math:`s,x_0\in\mathbb{R}^n` are the initial step and initial point respectively, the vertices of a simplex will be: :math:`v_0:=x_0-\frac{1}{2}
+s` and :math:`v_i:=x_0+s_i` for :math:`i=1,2,\dots,n` where :math:`s_i` denotes projections of the initial step of *n*-th coordinate (the result
+of projection is treated to be vector given by :math:`s_i:=e_i\cdot\left<e_i\cdot s\right>`, where :math:`e_i` form canonical basis)
+
+.. ocv:function:: void optim::setInitStep(InputArray step)
+
+    :param step: Initial step that will be used in algorithm. Roughly said, it determines the spread (size in each dimension) of an initial simplex.
+
+optim::DownhillSolver::minimize
+-----------------------------------
+
+The main method of the ``DownhillSolver``. It actually runs the algorithm and performs the minimization. The sole input parameter determines the
+centroid of the starting simplex (roughly, it tells where to start), all the others (terminal criteria, initial step, function to be minimized)
+are supposed to be set via the setters before the call to this method or the default values (not always sensible) will be used.
+
+.. ocv:function:: double optim::DownhillSolver::minimize(InputOutputArray x)
+
+    :param x: The initial point, that will become a centroid of an initial simplex. After the algorithm will terminate, it will be setted to the point where the algorithm stops, the point of possible minimum.
+
+    :return: The value of a function at the point found.
+
+optim::createDownhillSolver
+------------------------------------
+
+This function returns the reference to the ready-to-use ``DownhillSolver`` object. All the parameters are optional, so this procedure can be called
+even without parameters at all. In this case, the default values will be used. As default value for terminal criteria are the only sensible ones,
+``DownhillSolver::setFunction()`` and ``DownhillSolver::setInitStep()`` should be called upon the obtained object, if the respective parameters
+were not given to ``createDownhillSolver()``. Otherwise, the two ways (give parameters to ``createDownhillSolver()`` or miss the out and call the
+``DownhillSolver::setFunction()`` and ``DownhillSolver::setInitStep()``) are absolutely equivalent (and will drop the same errors in the same way,
+should invalid input be detected).
+
+.. ocv:function:: Ptr<optim::DownhillSolver> optim::createDownhillSolver(const Ptr<Solver::Function>& f,InputArray initStep, TermCriteria termcrit)
+
+    :param f: Pointer to the function that will be minimized, similarly to the one you submit via ``DownhillSolver::setFunction``.
+    :param step: Initial step, that will be used to construct the initial simplex, similarly to the one you submit via ``DownhillSolver::setInitStep``.
+    :param termcrit: Terminal criteria to the algorithm, similarly to the one you submit via ``DownhillSolver::setTermCriteria``.
diff --git a/modules/optim/doc/linear_programming.rst b/modules/optim/doc/linear_programming.rst
new file mode 100644
index 0000000000..946df9e955
--- /dev/null
+++ b/modules/optim/doc/linear_programming.rst
@@ -0,0 +1,48 @@
+Linear Programming
+==================
+
+.. highlight:: cpp
+
+optim::solveLP
+--------------------
+Solve given (non-integer) linear programming problem using the Simplex Algorithm (Simplex Method).
+What we mean here by "linear programming problem" (or LP problem, for short) can be
+formulated as:
+
+.. math::
+    \mbox{Maximize } c\cdot x\\
+    \mbox{Subject to:}\\
+    Ax\leq b\\
+    x\geq 0
+
+Where :math:`c` is fixed *1*-by-*n* row-vector, :math:`A` is fixed *m*-by-*n* matrix, :math:`b` is fixed *m*-by-*1* column vector and
+:math:`x` is an arbitrary *n*-by-*1* column vector, which satisfies the constraints.
+
+Simplex algorithm is one of many algorithms that are designed to handle this sort of problems efficiently. Although it is not optimal in theoretical
+sense (there exist algorithms that can solve any problem written as above in polynomial type, while simplex method degenerates to exponential time
+for some special cases), it is well-studied, easy to implement and is shown to work well for real-life purposes.
+
+The particular implementation is taken almost verbatim from **Introduction to Algorithms, third edition**
+by T. H. Cormen, C. E. Leiserson, R. L. Rivest and Clifford Stein. In particular, the Bland's rule
+(`http://en.wikipedia.org/wiki/Bland%27s\_rule <http://en.wikipedia.org/wiki/Bland%27s_rule>`_) is used to prevent cycling.
+
+.. ocv:function:: int optim::solveLP(const Mat& Func, const Mat& Constr, Mat& z)
+
+    :param Func: This row-vector corresponds to :math:`c` in the LP problem formulation (see above). It should contain 32- or 64-bit floating point numbers. As a convenience, column-vector may be also submitted, in the latter case it is understood to correspond to :math:`c^T`.
+
+    :param Constr: *m*-by-*n\+1* matrix, whose rightmost column corresponds to :math:`b` in formulation above and the remaining to :math:`A`. It should containt 32- or 64-bit floating point numbers.
+
+    :param z: The solution will be returned here as a column-vector - it corresponds to :math:`c` in the formulation above. It will contain 64-bit floating point numbers.
+
+    :return: One of the return codes:
+
+::
+
+    //!the return codes for solveLP() function
+    enum
+    {
+        SOLVELP_UNBOUNDED    = -2, //problem is unbounded (target function can achieve arbitrary high values)
+        SOLVELP_UNFEASIBLE    = -1, //problem is unfeasible (there are no points that satisfy all the constraints imposed)
+        SOLVELP_SINGLE    = 0, //there is only one maximum for target function
+        SOLVELP_MULTI    = 1 //there are multiple maxima for target function - the arbitrary one is returned
+    };
diff --git a/modules/optim/doc/optim.rst b/modules/optim/doc/optim.rst
new file mode 100644
index 0000000000..b3c7a740b1
--- /dev/null
+++ b/modules/optim/doc/optim.rst
@@ -0,0 +1,12 @@
+**************************************
+optim. Generic numerical optimization
+**************************************
+
+.. highlight:: cpp
+
+.. toctree::
+    :maxdepth: 2
+
+    linear_programming
+    downhill_simplex_method
+    primal_dual_algorithm
diff --git a/modules/optim/doc/primal_dual_algorithm.rst b/modules/optim/doc/primal_dual_algorithm.rst
new file mode 100644
index 0000000000..09d736f242
--- /dev/null
+++ b/modules/optim/doc/primal_dual_algorithm.rst
@@ -0,0 +1,48 @@
+Primal-Dual Algorithm
+=======================
+
+.. highlight:: cpp
+
+optim::denoise_TVL1
+---------------------------------
+
+Primal-dual algorithm is an algorithm for solving special types of variational
+problems (that is, finding a function to minimize some functional)
+. As the image denoising, in particular, may be seen as the variational
+problem, primal-dual algorithm then can be used to perform denoising and this
+is exactly what is implemented.
+
+It should be noted, that this implementation was taken from the July 2013 blog entry [Mordvintsev]_, which also contained
+(slightly more general) ready-to-use
+source code on Python. Subsequently, that code was rewritten on C++ with the usage of openCV by Vadim Pisarevsky
+at the end of July 2013 and finally it was slightly adapted by later authors.
+
+Although the thorough discussion and justification
+of the algorithm involved may be found in [ChambolleEtAl]_, it might make sense to skim over it here, following [Mordvintsev]_. To
+begin with, we consider the 1-byte gray-level images as the functions from the rectangular domain of pixels
+(it may be seen as set :math:`\left\{(x,y)\in\mathbb{N}\times\mathbb{N}\mid 1\leq x\leq n,\;1\leq y\leq m\right\}`
+for some :math:`m,\;n\in\mathbb{N}`) into :math:`\{0,1,\dots,255\}`. We shall denote the noised images as :math:`f_i` and with this
+view, given some image :math:`x` of the same size, we may measure how bad it is by the formula
+
+.. math::
+        \left\|\left\|\nabla x\right\|\right\| + \lambda\sum_i\left\|\left\|x-f_i\right\|\right\|
+
+:math:`\|\|\cdot\|\|` here denotes :math:`L_2`-norm and as you see, the first addend states that we want our image to be smooth
+(ideally, having zero gradient, thus being constant) and the second states that we want our result to be close to the observations we've got.
+If we treat :math:`x` as a function, this is exactly the functional what we seek to minimize and here the Primal-Dual algorithm comes
+into play.
+
+.. ocv:function:: void optim::denoise_TVL1(const std::vector<Mat>& observations,Mat& result, double lambda, int niters)
+
+    :param observations: This array should contain one or more noised versions of the image that is to be restored.
+
+    :param result: Here the denoised image will be stored. There is no need to do pre-allocation of storage space, as it will be automatically allocated, if necessary.
+
+    :param lambda: Corresponds to :math:`\lambda` in the formulas above. As it is enlarged, the smooth (blurred) images are treated more favorably than detailed (but maybe more noised) ones. Roughly speaking, as it becomes smaller, the result will be more blur but more sever outliers will be removed.
+
+    :param niters: Number of iterations that the algorithm will run. Of course, as more iterations as better, but it is hard to quantitatively refine this statement, so just use the default and increase it if the results are poor.
+
+
+.. [ChambolleEtAl] A. Chambolle, V. Caselles, M. Novaga, D. Cremers and T. Pock, An Introduction to Total Variation for Image Analysis, http://hal.archives-ouvertes.fr/docs/00/43/75/81/PDF/preprint.pdf (pdf)
+
+.. [Mordvintsev] Alexander Mordvintsev, ROF and TV-L1 denoising with Primal-Dual algorithm, http://znah.net/rof-and-tv-l1-denoising-with-primal-dual-algorithm.html (blog entry)
diff --git a/modules/optim/include/opencv2/optim.hpp b/modules/optim/include/opencv2/optim.hpp
new file mode 100644
index 0000000000..715372b69a
--- /dev/null
+++ b/modules/optim/include/opencv2/optim.hpp
@@ -0,0 +1,102 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2008-2012, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_OPTIM_HPP__
+#define __OPENCV_OPTIM_HPP__
+
+#include "opencv2/core.hpp"
+
+namespace cv{namespace optim
+{
+class CV_EXPORTS Solver : public Algorithm
+{
+public:
+    class CV_EXPORTS Function
+    {
+    public:
+       virtual ~Function() {}
+       //! ndim - dimensionality
+       virtual double calc(const double* x) const = 0;
+    };
+
+    virtual Ptr<Function> getFunction() const = 0;
+    virtual void setFunction(const Ptr<Function>& f) = 0;
+
+    virtual TermCriteria getTermCriteria() const = 0;
+    virtual void setTermCriteria(const TermCriteria& termcrit) = 0;
+
+    // x contain the initial point before the call and the minima position (if algorithm converged) after. x is assumed to be (something that
+    // after getMat() will return) row-vector or column-vector. *It's size  and should
+    // be consisted with previous dimensionality data given, if any (otherwise, it determines dimensionality)*
+    virtual double minimize(InputOutputArray x) = 0;
+};
+
+//! downhill simplex class
+class CV_EXPORTS DownhillSolver : public Solver
+{
+public:
+    //! returns row-vector, even if the column-vector was given
+    virtual void getInitStep(OutputArray step) const=0;
+    //!This should be called at least once before the first call to minimize() and step is assumed to be (something that
+    //! after getMat() will return) row-vector or column-vector. *It's dimensionality determines the dimensionality of a problem.*
+    virtual void setInitStep(InputArray step)=0;
+};
+
+// both minRange & minError are specified by termcrit.epsilon; In addition, user may specify the number of iterations that the algorithm does.
+CV_EXPORTS_W Ptr<DownhillSolver> createDownhillSolver(const Ptr<Solver::Function>& f=Ptr<Solver::Function>(),
+        InputArray initStep=Mat_<double>(1,1,0.0),
+        TermCriteria termcrit=TermCriteria(TermCriteria::MAX_ITER+TermCriteria::EPS,5000,0.000001));
+
+//!the return codes for solveLP() function
+enum
+{
+    SOLVELP_UNBOUNDED    = -2, //problem is unbounded (target function can achieve arbitrary high values)
+    SOLVELP_UNFEASIBLE    = -1, //problem is unfeasible (there are no points that satisfy all the constraints imposed)
+    SOLVELP_SINGLE    = 0, //there is only one maximum for target function
+    SOLVELP_MULTI    = 1 //there are multiple maxima for target function - the arbitrary one is returned
+};
+
+CV_EXPORTS_W int solveLP(const Mat& Func, const Mat& Constr, Mat& z);
+CV_EXPORTS_W void denoise_TVL1(const std::vector<Mat>& observations,Mat& result, double lambda=1.0, int niters=30);
+}}// cv
+
+#endif
diff --git a/modules/optim/include/opencv2/optim/optim.hpp b/modules/optim/include/opencv2/optim/optim.hpp
new file mode 100644
index 0000000000..b5a9ebf794
--- /dev/null
+++ b/modules/optim/include/opencv2/optim/optim.hpp
@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/optim.hpp"
diff --git a/modules/optim/src/debug.hpp b/modules/optim/src/debug.hpp
new file mode 100644
index 0000000000..fe5d00e87d
--- /dev/null
+++ b/modules/optim/src/debug.hpp
@@ -0,0 +1,18 @@
+namespace cv{namespace optim{
+#ifdef ALEX_DEBUG
+#define dprintf(x) printf x
+static void print_matrix(const Mat& x){
+    printf("\ttype:%d vs %d,\tsize: %d-on-%d\n",x.type(),CV_64FC1,x.rows,x.cols);
+    for(int i=0;i<x.rows;i++){
+        printf("\t[");
+        for(int j=0;j<x.cols;j++){
+            printf("%g, ",x.at<double>(i,j));
+        }
+        printf("]\n");
+    }
+}
+#else
+#define dprintf(x)
+#define print_matrix(x)
+#endif
+}}
diff --git a/modules/optim/src/denoise_tvl1.cpp b/modules/optim/src/denoise_tvl1.cpp
new file mode 100644
index 0000000000..b11ebc0587
--- /dev/null
+++ b/modules/optim/src/denoise_tvl1.cpp
@@ -0,0 +1,113 @@
+#include "precomp.hpp"
+#undef ALEX_DEBUG
+#include "debug.hpp"
+#include <vector>
+#include <algorithm>
+
+#define ABSCLIP(val,threshold) MIN(MAX((val),-(threshold)),(threshold))
+
+namespace cv{namespace optim{
+
+    class AddFloatToCharScaled{
+        public:
+            AddFloatToCharScaled(double scale):_scale(scale){}
+            inline double operator()(double a,uchar b){
+                return a+_scale*((double)b);
+            }
+        private:
+            double _scale;
+    };
+
+    void denoise_TVL1(const std::vector<Mat>& observations,Mat& result, double lambda, int niters){
+
+        CV_Assert(observations.size()>0 && niters>0 && lambda>0);
+
+        const double L2 = 8.0, tau = 0.02, sigma = 1./(L2*tau), theta = 1.0;
+        double clambda = (double)lambda;
+        double s=0;
+        const int workdepth = CV_64F;
+
+        int i, x, y, rows=observations[0].rows, cols=observations[0].cols,count;
+        for(i=1;i<(int)observations.size();i++){
+            CV_Assert(observations[i].rows==rows && observations[i].cols==cols);
+        }
+
+        Mat X, P = Mat::zeros(rows, cols, CV_MAKETYPE(workdepth, 2));
+        observations[0].convertTo(X, workdepth, 1./255);
+        std::vector< Mat_<double> > Rs(observations.size());
+        for(count=0;count<(int)Rs.size();count++){
+            Rs[count]=Mat::zeros(rows,cols,workdepth);
+        }
+
+        for( i = 0; i < niters; i++ )
+        {
+            double currsigma = i == 0 ? 1 + sigma : sigma;
+
+            // P_ = P + sigma*nabla(X)
+            // P(x,y) = P_(x,y)/max(||P(x,y)||,1)
+            for( y = 0; y < rows; y++ )
+            {
+                const double* x_curr = X.ptr<double>(y);
+                const double* x_next = X.ptr<double>(std::min(y+1, rows-1));
+                Point2d* p_curr = P.ptr<Point2d>(y);
+                double dx, dy, m;
+                for( x = 0; x < cols-1; x++ )
+                {
+                    dx = (x_curr[x+1] - x_curr[x])*currsigma + p_curr[x].x;
+                    dy = (x_next[x] - x_curr[x])*currsigma + p_curr[x].y;
+                    m = 1.0/std::max(std::sqrt(dx*dx + dy*dy), 1.0);
+                    p_curr[x].x = dx*m;
+                    p_curr[x].y = dy*m;
+                }
+                dy = (x_next[x] - x_curr[x])*currsigma + p_curr[x].y;
+                m = 1.0/std::max(std::abs(dy), 1.0);
+                p_curr[x].x = 0.0;
+                p_curr[x].y = dy*m;
+            }
+
+
+            //Rs = clip(Rs + sigma*(X-imgs), -clambda, clambda)
+            for(count=0;count<(int)Rs.size();count++){
+                std::transform<MatIterator_<double>,MatConstIterator_<uchar>,MatIterator_<double>,AddFloatToCharScaled>(
+                        Rs[count].begin(),Rs[count].end(),observations[count].begin<uchar>(),
+                        Rs[count].begin(),AddFloatToCharScaled(-sigma/255.0));
+                Rs[count]+=sigma*X;
+                min(Rs[count],clambda,Rs[count]);
+                max(Rs[count],-clambda,Rs[count]);
+            }
+
+            for( y = 0; y < rows; y++ )
+            {
+                double* x_curr = X.ptr<double>(y);
+                const Point2d* p_curr = P.ptr<Point2d>(y);
+                const Point2d* p_prev = P.ptr<Point2d>(std::max(y - 1, 0));
+
+                // X1 = X + tau*(-nablaT(P))
+                x = 0;
+                s=0.0;
+                for(count=0;count<(int)Rs.size();count++){
+                    s=s+Rs[count](y,x);
+                }
+                double x_new = x_curr[x] + tau*(p_curr[x].y - p_prev[x].y)-tau*s;
+                    // X = X2 + theta*(X2 - X)
+                x_curr[x] = x_new + theta*(x_new - x_curr[x]);
+
+
+                for(x = 1; x < cols; x++ )
+                {
+                    s=0.0;
+                    for(count=0;count<(int)Rs.size();count++){
+                        s+=Rs[count](y,x);
+                    }
+                        // X1 = X + tau*(-nablaT(P))
+                    x_new = x_curr[x] + tau*(p_curr[x].x - p_curr[x-1].x + p_curr[x].y - p_prev[x].y)-tau*s;
+                        // X = X2 + theta*(X2 - X)
+                    x_curr[x] = x_new + theta*(x_new - x_curr[x]);
+                }
+            }
+        }
+
+        result.create(X.rows,X.cols,CV_8U);
+        X.convertTo(result, CV_8U, 255);
+    }
+}}
diff --git a/modules/optim/src/lpsolver.cpp b/modules/optim/src/lpsolver.cpp
new file mode 100644
index 0000000000..a046ddae11
--- /dev/null
+++ b/modules/optim/src/lpsolver.cpp
@@ -0,0 +1,316 @@
+#include "precomp.hpp"
+#include <climits>
+#include <algorithm>
+#include <cstdarg>
+#include <debug.hpp>
+
+namespace cv{namespace optim{
+using std::vector;
+
+#ifdef ALEX_DEBUG
+static void print_simplex_state(const Mat& c,const Mat& b,double v,const std::vector<int> N,const std::vector<int> B){
+    printf("\tprint simplex state\n");
+
+    printf("v=%g\n",v);
+
+    printf("here c goes\n");
+    print_matrix(c);
+
+    printf("non-basic: ");
+    print(Mat(N));
+    printf("\n");
+
+    printf("here b goes\n");
+    print_matrix(b);
+    printf("basic: ");
+
+    print(Mat(B));
+    printf("\n");
+}
+#else
+#define print_simplex_state(c,b,v,N,B)
+#endif
+
+/**Due to technical considerations, the format of input b and c is somewhat special:
+ *both b and c should be one column bigger than corresponding b and c of linear problem and the leftmost column will be used internally
+ by this procedure - it should not be cleaned before the call to procedure and may contain mess after
+ it also initializes N and B and does not make any assumptions about their init values
+ * @return SOLVELP_UNFEASIBLE if problem is unfeasible, 0 if feasible.
+*/
+static int initialize_simplex(Mat_<double>& c, Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,vector<unsigned int>& indexToRow);
+static inline void pivot(Mat_<double>& c,Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,int leaving_index,
+        int entering_index,vector<unsigned int>& indexToRow);
+/**@return SOLVELP_UNBOUNDED means the problem is unbdd, SOLVELP_MULTI means multiple solutions, SOLVELP_SINGLE means one solution.
+ */
+static int inner_simplex(Mat_<double>& c, Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,vector<unsigned int>& indexToRow);
+static void swap_columns(Mat_<double>& A,int col1,int col2);
+#define SWAP(type,a,b) {type tmp=(a);(a)=(b);(b)=tmp;}
+
+//return codes:-2 (no_sol - unbdd),-1(no_sol - unfsbl), 0(single_sol), 1(multiple_sol=>least_l2_norm)
+int solveLP(const Mat& Func, const Mat& Constr, Mat& z){
+    dprintf(("call to solveLP\n"));
+
+    //sanity check (size, type, no. of channels)
+    CV_Assert(Func.type()==CV_64FC1 || Func.type()==CV_32FC1);
+    CV_Assert(Constr.type()==CV_64FC1 || Constr.type()==CV_32FC1);
+    CV_Assert((Func.rows==1 && (Constr.cols-Func.cols==1))||
+            (Func.cols==1 && (Constr.cols-Func.rows==1)));
+
+    //copy arguments for we will shall modify them
+    Mat_<double> bigC=Mat_<double>(1,(Func.rows==1?Func.cols:Func.rows)+1),
+        bigB=Mat_<double>(Constr.rows,Constr.cols+1);
+    if(Func.rows==1){
+        Func.convertTo(bigC.colRange(1,bigC.cols),CV_64FC1);
+    }else{
+        Mat FuncT=Func.t();
+        FuncT.convertTo(bigC.colRange(1,bigC.cols),CV_64FC1);
+    }
+    Constr.convertTo(bigB.colRange(1,bigB.cols),CV_64FC1);
+    double v=0;
+    vector<int> N,B;
+    vector<unsigned int> indexToRow;
+
+    if(initialize_simplex(bigC,bigB,v,N,B,indexToRow)==SOLVELP_UNFEASIBLE){
+        return SOLVELP_UNFEASIBLE;
+    }
+    Mat_<double> c=bigC.colRange(1,bigC.cols),
+        b=bigB.colRange(1,bigB.cols);
+
+    int res=0;
+    if((res=inner_simplex(c,b,v,N,B,indexToRow))==SOLVELP_UNBOUNDED){
+        return SOLVELP_UNBOUNDED;
+    }
+
+    //return the optimal solution
+    z.create(c.cols,1,CV_64FC1);
+    MatIterator_<double> it=z.begin<double>();
+    for(int i=1;i<=c.cols;i++,it++){
+        if(indexToRow[i]<N.size()){
+            *it=0;
+        }else{
+            *it=b.at<double>(indexToRow[i]-N.size(),b.cols-1);
+        }
+    }
+
+    return res;
+}
+
+static int initialize_simplex(Mat_<double>& c, Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,vector<unsigned int>& indexToRow){
+    N.resize(c.cols);
+    N[0]=0;
+    for (std::vector<int>::iterator it = N.begin()+1 ; it != N.end(); ++it){
+        *it=it[-1]+1;
+    }
+    B.resize(b.rows);
+    B[0]=N.size();
+    for (std::vector<int>::iterator it = B.begin()+1 ; it != B.end(); ++it){
+        *it=it[-1]+1;
+    }
+    indexToRow.resize(c.cols+b.rows);
+    indexToRow[0]=0;
+    for (std::vector<unsigned int>::iterator it = indexToRow.begin()+1 ; it != indexToRow.end(); ++it){
+        *it=it[-1]+1;
+    }
+    v=0;
+
+    int k=0;
+    {
+        double min=DBL_MAX;
+        for(int i=0;i<b.rows;i++){
+            if(b(i,b.cols-1)<min){
+                min=b(i,b.cols-1);
+                k=i;
+            }
+        }
+    }
+
+    if(b(k,b.cols-1)>=0){
+        N.erase(N.begin());
+        for (std::vector<unsigned int>::iterator it = indexToRow.begin()+1 ; it != indexToRow.end(); ++it){
+            --(*it);
+        }
+        return 0;
+    }
+
+    Mat_<double> old_c=c.clone();
+    c=0;
+    c(0,0)=-1;
+    for(int i=0;i<b.rows;i++){
+        b(i,0)=-1;
+    }
+
+    print_simplex_state(c,b,v,N,B);
+
+    dprintf(("\tWE MAKE PIVOT\n"));
+    pivot(c,b,v,N,B,k,0,indexToRow);
+
+    print_simplex_state(c,b,v,N,B);
+
+    inner_simplex(c,b,v,N,B,indexToRow);
+
+    dprintf(("\tAFTER INNER_SIMPLEX\n"));
+    print_simplex_state(c,b,v,N,B);
+
+    if(indexToRow[0]>=N.size()){
+        int iterator_offset=indexToRow[0]-N.size();
+        if(b(iterator_offset,b.cols-1)>0){
+            return SOLVELP_UNFEASIBLE;
+        }
+        pivot(c,b,v,N,B,iterator_offset,0,indexToRow);
+    }
+
+    vector<int>::iterator iterator;
+    {
+        int iterator_offset=indexToRow[0];
+        iterator=N.begin()+iterator_offset;
+        std::iter_swap(iterator,N.begin());
+        SWAP(int,indexToRow[*iterator],indexToRow[0]);
+        swap_columns(c,iterator_offset,0);
+        swap_columns(b,iterator_offset,0);
+    }
+
+    dprintf(("after swaps\n"));
+    print_simplex_state(c,b,v,N,B);
+
+    //start from 1, because we ignore x_0
+    c=0;
+    v=0;
+    for(int I=1;I<old_c.cols;I++){
+        if(indexToRow[I]<N.size()){
+            dprintf(("I=%d from nonbasic\n",I));
+            int iterator_offset=indexToRow[I];
+            c(0,iterator_offset)+=old_c(0,I);
+            print_matrix(c);
+        }else{
+            dprintf(("I=%d from basic\n",I));
+            int iterator_offset=indexToRow[I]-N.size();
+            c-=old_c(0,I)*b.row(iterator_offset).colRange(0,b.cols-1);
+            v+=old_c(0,I)*b(iterator_offset,b.cols-1);
+            print_matrix(c);
+        }
+    }
+
+    dprintf(("after restore\n"));
+    print_simplex_state(c,b,v,N,B);
+
+    N.erase(N.begin());
+    for (std::vector<unsigned int>::iterator it = indexToRow.begin()+1 ; it != indexToRow.end(); ++it){
+        --(*it);
+    }
+    return 0;
+}
+
+static int inner_simplex(Mat_<double>& c, Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,vector<unsigned int>& indexToRow){
+    int count=0;
+    for(;;){
+        dprintf(("iteration #%d\n",count));
+        count++;
+
+        static MatIterator_<double> pos_ptr;
+        int e=-1,pos_ctr=0,min_var=INT_MAX;
+        bool all_nonzero=true;
+        for(pos_ptr=c.begin();pos_ptr!=c.end();pos_ptr++,pos_ctr++){
+            if(*pos_ptr==0){
+                all_nonzero=false;
+            }
+            if(*pos_ptr>0){
+                if(N[pos_ctr]<min_var){
+                    e=pos_ctr;
+                    min_var=N[pos_ctr];
+                }
+            }
+        }
+        if(e==-1){
+            dprintf(("hello from e==-1\n"));
+            print_matrix(c);
+            if(all_nonzero==true){
+                return SOLVELP_SINGLE;
+            }else{
+                return SOLVELP_MULTI;
+            }
+        }
+
+        int l=-1;
+        min_var=INT_MAX;
+        double min=DBL_MAX;
+        int row_it=0;
+        MatIterator_<double> min_row_ptr=b.begin();
+        for(MatIterator_<double> it=b.begin();it!=b.end();it+=b.cols,row_it++){
+            double myite=0;
+            //check constraints, select the tightest one, reinforcing Bland's rule
+            if((myite=it[e])>0){
+                double val=it[b.cols-1]/myite;
+                if(val<min || (val==min && B[row_it]<min_var)){
+                    min_var=B[row_it];
+                    min_row_ptr=it;
+                    min=val;
+                    l=row_it;
+                }
+            }
+        }
+        if(l==-1){
+            return SOLVELP_UNBOUNDED;
+        }
+        dprintf(("the tightest constraint is in row %d with %g\n",l,min));
+
+        pivot(c,b,v,N,B,l,e,indexToRow);
+
+        dprintf(("objective, v=%g\n",v));
+        print_matrix(c);
+        dprintf(("constraints\n"));
+        print_matrix(b);
+        dprintf(("non-basic: "));
+        print_matrix(Mat(N));
+        dprintf(("basic: "));
+        print_matrix(Mat(B));
+    }
+}
+
+static inline void pivot(Mat_<double>& c,Mat_<double>& b,double& v,vector<int>& N,vector<int>& B,
+        int leaving_index,int entering_index,vector<unsigned int>& indexToRow){
+    double Coef=b(leaving_index,entering_index);
+    for(int i=0;i<b.cols;i++){
+        if(i==entering_index){
+            b(leaving_index,i)=1/Coef;
+        }else{
+            b(leaving_index,i)/=Coef;
+        }
+    }
+
+    for(int i=0;i<b.rows;i++){
+        if(i!=leaving_index){
+            double coef=b(i,entering_index);
+            for(int j=0;j<b.cols;j++){
+                if(j==entering_index){
+                    b(i,j)=-coef*b(leaving_index,j);
+                }else{
+                    b(i,j)-=(coef*b(leaving_index,j));
+                }
+            }
+        }
+    }
+
+    //objective function
+    Coef=c(0,entering_index);
+    for(int i=0;i<(b.cols-1);i++){
+        if(i==entering_index){
+            c(0,i)=-Coef*b(leaving_index,i);
+        }else{
+            c(0,i)-=Coef*b(leaving_index,i);
+        }
+    }
+    dprintf(("v was %g\n",v));
+    v+=Coef*b(leaving_index,b.cols-1);
+
+    SWAP(int,N[entering_index],B[leaving_index]);
+    SWAP(int,indexToRow[N[entering_index]],indexToRow[B[leaving_index]]);
+}
+
+static inline void swap_columns(Mat_<double>& A,int col1,int col2){
+    for(int i=0;i<A.rows;i++){
+        double tmp=A(i,col1);
+        A(i,col1)=A(i,col2);
+        A(i,col2)=tmp;
+    }
+}
+}}
diff --git a/modules/nonfree/src/precomp.cpp b/modules/optim/src/precomp.hpp
similarity index 95%
rename from modules/nonfree/src/precomp.cpp
rename to modules/optim/src/precomp.hpp
index 730edbb63d..7aab572f88 100644
--- a/modules/nonfree/src/precomp.cpp
+++ b/modules/optim/src/precomp.hpp
@@ -40,6 +40,9 @@
 //
 //M*/
 
-#include "precomp.hpp"
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
 
-/* End of file. */
+#include "opencv2/optim.hpp"
+
+#endif
diff --git a/modules/optim/src/simplex.cpp b/modules/optim/src/simplex.cpp
new file mode 100644
index 0000000000..f45d0ce0b9
--- /dev/null
+++ b/modules/optim/src/simplex.cpp
@@ -0,0 +1,273 @@
+#include "precomp.hpp"
+#include "debug.hpp"
+#include "opencv2/core/core_c.h"
+
+namespace cv{namespace optim{
+
+    class DownhillSolverImpl : public DownhillSolver
+    {
+    public:
+        void getInitStep(OutputArray step) const;
+        void setInitStep(InputArray step);
+        Ptr<Function> getFunction() const;
+        void setFunction(const Ptr<Function>& f);
+        TermCriteria getTermCriteria() const;
+        DownhillSolverImpl();
+        void setTermCriteria(const TermCriteria& termcrit);
+        double minimize(InputOutputArray x);
+    protected:
+        Ptr<Solver::Function> _Function;
+        TermCriteria _termcrit;
+        Mat _step;
+    private:
+        inline void createInitialSimplex(Mat_<double>& simplex,Mat& step);
+        inline double innerDownhillSimplex(cv::Mat_<double>& p,double MinRange,double MinError,int& nfunk,
+                const Ptr<Solver::Function>& f,int nmax);
+        inline double tryNewPoint(Mat_<double>& p,Mat_<double>& y,Mat_<double>& coord_sum,const Ptr<Solver::Function>& f,int ihi,
+                double fac,Mat_<double>& ptry);
+    };
+
+    double DownhillSolverImpl::tryNewPoint(
+        Mat_<double>& p,
+        Mat_<double>& y,
+        Mat_<double>&  coord_sum,
+        const Ptr<Solver::Function>& f,
+        int      ihi,
+        double   fac,
+        Mat_<double>& ptry
+        )
+    {
+        int ndim=p.cols;
+        int j;
+        double fac1,fac2,ytry;
+
+        fac1=(1.0-fac)/ndim;
+        fac2=fac1-fac;
+        for (j=0;j<ndim;j++)
+        {
+            ptry(j)=coord_sum(j)*fac1-p(ihi,j)*fac2;
+        }
+        ytry=f->calc((double*)ptry.data);
+        if (ytry < y(ihi))
+        {
+            y(ihi)=ytry;
+            for (j=0;j<ndim;j++)
+            {
+                coord_sum(j) += ptry(j)-p(ihi,j);
+                p(ihi,j)=ptry(j);
+            }
+        }
+
+        return ytry;
+    }
+
+    /*
+    Performs the actual minimization of Solver::Function f (after the initialization was done)
+
+    The matrix p[ndim+1][1..ndim] represents ndim+1 vertices that
+    form a simplex - each row is an ndim vector.
+    On output, nfunk gives the number of function evaluations taken.
+    */
+    double DownhillSolverImpl::innerDownhillSimplex(
+        cv::Mat_<double>&   p,
+        double     MinRange,
+        double     MinError,
+        int&       nfunk,
+        const Ptr<Solver::Function>& f,
+        int nmax
+        )
+    {
+        int ndim=p.cols;
+        double res;
+        int i,ihi,ilo,inhi,j,mpts=ndim+1;
+        double error, range,ysave,ytry;
+        Mat_<double> coord_sum(1,ndim,0.0),buf(1,ndim,0.0),y(1,ndim,0.0);
+
+        nfunk = 0;
+
+        for(i=0;i<ndim+1;++i)
+        {
+            y(i) = f->calc(p[i]);
+        }
+
+        nfunk = ndim+1;
+
+        reduce(p,coord_sum,0,CV_REDUCE_SUM);
+
+        for (;;)
+        {
+            ilo=0;
+            /*  find highest (worst), next-to-worst, and lowest
+                (best) points by going through all of them. */
+            ihi = y(0)>y(1) ? (inhi=1,0) : (inhi=0,1);
+            for (i=0;i<mpts;i++)
+            {
+                if (y(i) <= y(ilo))
+                    ilo=i;
+                if (y(i) > y(ihi))
+                {
+                    inhi=ihi;
+                    ihi=i;
+                }
+                else if (y(i) > y(inhi) && i != ihi)
+                    inhi=i;
+            }
+
+            /* check stop criterion */
+            error=fabs(y(ihi)-y(ilo));
+            range=0;
+            for(i=0;i<ndim;++i)
+            {
+                double min = p(0,i);
+                double max = p(0,i);
+                double d;
+                for(j=1;j<=ndim;++j)
+                {
+                    if( min > p(j,i) ) min = p(j,i);
+                    if( max < p(j,i) ) max = p(j,i);
+                }
+                d = fabs(max-min);
+                if(range < d) range = d;
+            }
+
+            if(range <= MinRange || error <= MinError)
+            { /* Put best point and value in first slot. */
+                std::swap(y(0),y(ilo));
+                for (i=0;i<ndim;i++)
+                {
+                    std::swap(p(0,i),p(ilo,i));
+                }
+                break;
+            }
+
+            if (nfunk >= nmax){
+                dprintf(("nmax exceeded\n"));
+                return y(ilo);
+            }
+            nfunk += 2;
+            /*Begin a new iteration. First, reflect the worst point about the centroid of others */
+            ytry = tryNewPoint(p,y,coord_sum,f,ihi,-1.0,buf);
+            if (ytry <= y(ilo))
+            { /*If that's better than the best point, go twice as far in that direction*/
+                ytry = tryNewPoint(p,y,coord_sum,f,ihi,2.0,buf);
+            }
+            else if (ytry >= y(inhi))
+            {   /* The new point is worse than the second-highest, but better
+                  than the worst so do not go so far in that direction */
+                ysave = y(ihi);
+                ytry = tryNewPoint(p,y,coord_sum,f,ihi,0.5,buf);
+                if (ytry >= ysave)
+                { /* Can't seem to improve things. Contract the simplex to good point
+               in hope to find a simplex landscape. */
+                    for (i=0;i<mpts;i++)
+                    {
+                        if (i != ilo)
+                        {
+                            for (j=0;j<ndim;j++)
+                            {
+                                p(i,j) = coord_sum(j) = 0.5*(p(i,j)+p(ilo,j));
+                            }
+                            y(i)=f->calc((double*)coord_sum.data);
+                        }
+                    }
+                    nfunk += ndim;
+                    reduce(p,coord_sum,0,CV_REDUCE_SUM);
+                }
+            } else --(nfunk); /* correct nfunk */
+            dprintf(("this is simplex on iteration %d\n",nfunk));
+            print_matrix(p);
+        } /* go to next iteration. */
+        res = y(0);
+
+        return res;
+    }
+
+    void DownhillSolverImpl::createInitialSimplex(Mat_<double>& simplex,Mat& step){
+        for(int i=1;i<=step.cols;++i)
+        {
+            simplex.row(0).copyTo(simplex.row(i));
+            simplex(i,i-1)+= 0.5*step.at<double>(0,i-1);
+        }
+        simplex.row(0) -= 0.5*step;
+
+        dprintf(("this is simplex\n"));
+        print_matrix(simplex);
+    }
+
+    double DownhillSolverImpl::minimize(InputOutputArray x){
+        dprintf(("hi from minimize\n"));
+        CV_Assert(_Function.empty()==false);
+        dprintf(("termcrit:\n\ttype: %d\n\tmaxCount: %d\n\tEPS: %g\n",_termcrit.type,_termcrit.maxCount,_termcrit.epsilon));
+        dprintf(("step\n"));
+        print_matrix(_step);
+
+        Mat x_mat=x.getMat();
+        CV_Assert(MIN(x_mat.rows,x_mat.cols)==1);
+        CV_Assert(MAX(x_mat.rows,x_mat.cols)==_step.cols);
+        CV_Assert(x_mat.type()==CV_64FC1);
+
+        Mat_<double> proxy_x;
+
+        if(x_mat.rows>1){
+            proxy_x=x_mat.t();
+        }else{
+            proxy_x=x_mat;
+        }
+
+        int count=0;
+        int ndim=_step.cols;
+        Mat_<double> simplex=Mat_<double>(ndim+1,ndim,0.0);
+
+        simplex.row(0).copyTo(proxy_x);
+        createInitialSimplex(simplex,_step);
+        double res = innerDownhillSimplex(
+                simplex,_termcrit.epsilon, _termcrit.epsilon, count,_Function,_termcrit.maxCount);
+        simplex.row(0).copyTo(proxy_x);
+
+        dprintf(("%d iterations done\n",count));
+
+        if(x_mat.rows>1){
+            Mat(x_mat.rows, 1, CV_64F, (double*)proxy_x.data).copyTo(x);
+        }
+        return res;
+    }
+    DownhillSolverImpl::DownhillSolverImpl(){
+        _Function=Ptr<Function>();
+        _step=Mat_<double>();
+    }
+    Ptr<Solver::Function> DownhillSolverImpl::getFunction()const{
+        return _Function;
+    }
+    void DownhillSolverImpl::setFunction(const Ptr<Function>& f){
+        _Function=f;
+    }
+    TermCriteria DownhillSolverImpl::getTermCriteria()const{
+        return _termcrit;
+    }
+    void DownhillSolverImpl::setTermCriteria(const TermCriteria& termcrit){
+        CV_Assert(termcrit.type==(TermCriteria::MAX_ITER+TermCriteria::EPS) && termcrit.epsilon>0 && termcrit.maxCount>0);
+        _termcrit=termcrit;
+    }
+    // both minRange & minError are specified by termcrit.epsilon; In addition, user may specify the number of iterations that the algorithm does.
+    Ptr<DownhillSolver> createDownhillSolver(const Ptr<Solver::Function>& f, InputArray initStep, TermCriteria termcrit){
+        DownhillSolver *DS=new DownhillSolverImpl();
+        DS->setFunction(f);
+        DS->setInitStep(initStep);
+        DS->setTermCriteria(termcrit);
+        return Ptr<DownhillSolver>(DS);
+    }
+    void DownhillSolverImpl::getInitStep(OutputArray step)const{
+        _step.copyTo(step);
+    }
+    void DownhillSolverImpl::setInitStep(InputArray step){
+        //set dimensionality and make a deep copy of step
+        Mat m=step.getMat();
+        dprintf(("m.cols=%d\nm.rows=%d\n",m.cols,m.rows));
+        CV_Assert(MIN(m.cols,m.rows)==1 && m.type()==CV_64FC1);
+        if(m.rows==1){
+            m.copyTo(_step);
+        }else{
+            transpose(m,_step);
+        }
+    }
+}}
diff --git a/modules/optim/test/test_denoise_tvl1.cpp b/modules/optim/test/test_denoise_tvl1.cpp
new file mode 100644
index 0000000000..2721a7666d
--- /dev/null
+++ b/modules/optim/test/test_denoise_tvl1.cpp
@@ -0,0 +1,80 @@
+#include "test_precomp.hpp"
+#include "opencv2/highgui.hpp"
+
+void make_noisy(const cv::Mat& img, cv::Mat& noisy, double sigma, double pepper_salt_ratio,cv::RNG& rng){
+    noisy.create(img.size(), img.type());
+    cv::Mat noise(img.size(), img.type()), mask(img.size(), CV_8U);
+    rng.fill(noise,cv::RNG::NORMAL,128.0,sigma);
+    cv::addWeighted(img, 1, noise, 1, -128, noisy);
+    cv::randn(noise, cv::Scalar::all(0), cv::Scalar::all(2));
+    noise *= 255;
+    cv::randu(mask, 0, cvRound(1./pepper_salt_ratio));
+    cv::Mat half = mask.colRange(0, img.cols/2);
+    half = cv::Scalar::all(1);
+    noise.setTo(128, mask);
+    cv::addWeighted(noisy, 1, noise, 1, -128, noisy);
+}
+void make_spotty(cv::Mat& img,cv::RNG& rng, int r=3,int n=1000){
+    for(int i=0;i<n;i++){
+        int x=rng(img.cols-r),y=rng(img.rows-r);
+        if(rng(2)==0){
+            img(cv::Range(y,y+r),cv::Range(x,x+r))=(uchar)0;
+        }else{
+            img(cv::Range(y,y+r),cv::Range(x,x+r))=(uchar)255;
+        }
+    }
+}
+
+bool validate_pixel(const cv::Mat& image,int x,int y,uchar val){
+    printf("test: image(%d,%d)=%d vs %d - %s\n",x,y,(int)image.at<uchar>(x,y),val,(val==image.at<uchar>(x,y))?"true":"false");
+    return (image.at<uchar>(x,y)==val);
+}
+
+TEST(Optim_denoise_tvl1, regression_basic){
+    cv::RNG rng(42);
+    cv::Mat img = cv::imread("lena.jpg", 0), noisy,res;
+    if(img.rows!=512 || img.cols!=512){
+        printf("\tplease, put lena.jpg from samples/c in the current folder\n");
+        printf("\tnow, the test will fail...\n");
+        ASSERT_TRUE(false);
+    }
+
+    const int obs_num=5;
+    std::vector<cv::Mat> images(obs_num,cv::Mat());
+    for(int i=0;i<(int)images.size();i++){
+        make_noisy(img,images[i], 20, 0.02,rng);
+        //make_spotty(images[i],rng);
+    }
+
+    //cv::imshow("test", images[0]);
+    cv::optim::denoise_TVL1(images, res);
+    //cv::imshow("denoised", res);
+    //cv::waitKey();
+
+#if 0
+    ASSERT_TRUE(validate_pixel(res,248,334,179));
+    ASSERT_TRUE(validate_pixel(res,489,333,172));
+    ASSERT_TRUE(validate_pixel(res,425,507,104));
+    ASSERT_TRUE(validate_pixel(res,489,486,105));
+    ASSERT_TRUE(validate_pixel(res,223,208,64));
+    ASSERT_TRUE(validate_pixel(res,418,3,78));
+    ASSERT_TRUE(validate_pixel(res,63,76,97));
+    ASSERT_TRUE(validate_pixel(res,29,134,126));
+    ASSERT_TRUE(validate_pixel(res,219,291,174));
+    ASSERT_TRUE(validate_pixel(res,384,124,76));
+#endif
+
+#if 1
+    ASSERT_TRUE(validate_pixel(res,248,334,194));
+    ASSERT_TRUE(validate_pixel(res,489,333,171));
+    ASSERT_TRUE(validate_pixel(res,425,507,103));
+    ASSERT_TRUE(validate_pixel(res,489,486,109));
+    ASSERT_TRUE(validate_pixel(res,223,208,72));
+    ASSERT_TRUE(validate_pixel(res,418,3,58));
+    ASSERT_TRUE(validate_pixel(res,63,76,93));
+    ASSERT_TRUE(validate_pixel(res,29,134,127));
+    ASSERT_TRUE(validate_pixel(res,219,291,180));
+    ASSERT_TRUE(validate_pixel(res,384,124,80));
+#endif
+
+}
diff --git a/modules/optim/test/test_downhill_simplex.cpp b/modules/optim/test/test_downhill_simplex.cpp
new file mode 100644
index 0000000000..95b2c6e9e3
--- /dev/null
+++ b/modules/optim/test/test_downhill_simplex.cpp
@@ -0,0 +1,63 @@
+#include "test_precomp.hpp"
+#include <cstdlib>
+#include <cmath>
+#include <algorithm>
+
+static void mytest(cv::Ptr<cv::optim::DownhillSolver> solver,cv::Ptr<cv::optim::Solver::Function> ptr_F,cv::Mat& x,cv::Mat& step,
+        cv::Mat& etalon_x,double etalon_res){
+    solver->setFunction(ptr_F);
+    int ndim=MAX(step.cols,step.rows);
+    solver->setInitStep(step);
+    cv::Mat settedStep;
+    solver->getInitStep(settedStep);
+    ASSERT_TRUE(settedStep.rows==1 && settedStep.cols==ndim);
+    ASSERT_TRUE(std::equal(step.begin<double>(),step.end<double>(),settedStep.begin<double>()));
+    std::cout<<"step setted:\n\t"<<step<<std::endl;
+    double res=solver->minimize(x);
+    std::cout<<"res:\n\t"<<res<<std::endl;
+    std::cout<<"x:\n\t"<<x<<std::endl;
+    std::cout<<"etalon_res:\n\t"<<etalon_res<<std::endl;
+    std::cout<<"etalon_x:\n\t"<<etalon_x<<std::endl;
+    double tol=solver->getTermCriteria().epsilon;
+    ASSERT_TRUE(std::abs(res-etalon_res)<tol);
+    /*for(cv::Mat_<double>::iterator it1=x.begin<double>(),it2=etalon_x.begin<double>();it1!=x.end<double>();it1++,it2++){
+        ASSERT_TRUE(std::abs((*it1)-(*it2))<tol);
+    }*/
+    std::cout<<"--------------------------\n";
+}
+
+class SphereF:public cv::optim::Solver::Function{
+public:
+    double calc(const double* x)const{
+        return x[0]*x[0]+x[1]*x[1];
+    }
+};
+class RosenbrockF:public cv::optim::Solver::Function{
+    double calc(const double* x)const{
+        return 100*(x[1]-x[0]*x[0])*(x[1]-x[0]*x[0])+(1-x[0])*(1-x[0]);
+    }
+};
+
+TEST(Optim_Downhill, regression_basic){
+    cv::Ptr<cv::optim::DownhillSolver> solver=cv::optim::createDownhillSolver();
+#if 1
+    {
+        cv::Ptr<cv::optim::Solver::Function> ptr_F(new SphereF());
+        cv::Mat x=(cv::Mat_<double>(1,2)<<1.0,1.0),
+            step=(cv::Mat_<double>(2,1)<<-0.5,-0.5),
+            etalon_x=(cv::Mat_<double>(1,2)<<-0.0,0.0);
+        double etalon_res=0.0;
+        mytest(solver,ptr_F,x,step,etalon_x,etalon_res);
+    }
+#endif
+#if 1
+    {
+        cv::Ptr<cv::optim::Solver::Function> ptr_F(new RosenbrockF());
+        cv::Mat x=(cv::Mat_<double>(2,1)<<0.0,0.0),
+            step=(cv::Mat_<double>(2,1)<<0.5,+0.5),
+            etalon_x=(cv::Mat_<double>(2,1)<<1.0,1.0);
+        double etalon_res=0.0;
+        mytest(solver,ptr_F,x,step,etalon_x,etalon_res);
+    }
+#endif
+}
diff --git a/modules/optim/test/test_lpsolver.cpp b/modules/optim/test/test_lpsolver.cpp
new file mode 100644
index 0000000000..f39c7eb371
--- /dev/null
+++ b/modules/optim/test/test_lpsolver.cpp
@@ -0,0 +1,101 @@
+#include "test_precomp.hpp"
+#include <iostream>
+
+TEST(Optim_LpSolver, regression_basic){
+    cv::Mat A,B,z,etalon_z;
+
+#if 1
+    //cormen's example #1
+    A=(cv::Mat_<double>(3,1)<<3,1,2);
+    B=(cv::Mat_<double>(3,4)<<1,1,3,30,2,2,5,24,4,1,2,36);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    cv::optim::solveLP(A,B,z);
+    std::cout<<"here z goes\n"<<z<<"\n";
+    etalon_z=(cv::Mat_<double>(3,1)<<8,4,0);
+    ASSERT_EQ(cv::countNonZero(z!=etalon_z),0);
+#endif
+
+#if 1
+    //cormen's example #2
+    A=(cv::Mat_<double>(1,2)<<18,12.5);
+    B=(cv::Mat_<double>(3,3)<<1,1,20,1,0,20,0,1,16);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    cv::optim::solveLP(A,B,z);
+    std::cout<<"here z goes\n"<<z<<"\n";
+    etalon_z=(cv::Mat_<double>(2,1)<<20,0);
+    ASSERT_EQ(cv::countNonZero(z!=etalon_z),0);
+#endif
+
+#if 1
+    //cormen's example #3
+    A=(cv::Mat_<double>(1,2)<<5,-3);
+    B=(cv::Mat_<double>(2,3)<<1,-1,1,2,1,2);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    cv::optim::solveLP(A,B,z);
+    std::cout<<"here z goes\n"<<z<<"\n";
+    etalon_z=(cv::Mat_<double>(2,1)<<1,0);
+    ASSERT_EQ(cv::countNonZero(z!=etalon_z),0);
+#endif
+}
+
+TEST(Optim_LpSolver, regression_init_unfeasible){
+    cv::Mat A,B,z,etalon_z;
+
+#if 1
+    //cormen's example #4 - unfeasible
+    A=(cv::Mat_<double>(1,3)<<-1,-1,-1);
+    B=(cv::Mat_<double>(2,4)<<-2,-7.5,-3,-10000,-20,-5,-10,-30000);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    cv::optim::solveLP(A,B,z);
+    std::cout<<"here z goes\n"<<z<<"\n";
+    etalon_z=(cv::Mat_<double>(3,1)<<1250,1000,0);
+    ASSERT_EQ(cv::countNonZero(z!=etalon_z),0);
+#endif
+}
+
+TEST(Optim_LpSolver, regression_absolutely_unfeasible){
+    cv::Mat A,B,z,etalon_z;
+
+#if 1
+    //trivial absolutely unfeasible example
+    A=(cv::Mat_<double>(1,1)<<1);
+    B=(cv::Mat_<double>(2,2)<<1,-1);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    int res=cv::optim::solveLP(A,B,z);
+    ASSERT_EQ(res,-1);
+#endif
+}
+
+TEST(Optim_LpSolver, regression_multiple_solutions){
+    cv::Mat A,B,z,etalon_z;
+
+#if 1
+    //trivial example with multiple solutions
+    A=(cv::Mat_<double>(2,1)<<1,1);
+    B=(cv::Mat_<double>(1,3)<<1,1,1);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    int res=cv::optim::solveLP(A,B,z);
+    printf("res=%d\n",res);
+    printf("scalar %g\n",z.dot(A));
+    std::cout<<"here z goes\n"<<z<<"\n";
+    ASSERT_EQ(res,1);
+    ASSERT_EQ(z.dot(A),1);
+#endif
+}
+
+TEST(Optim_LpSolver, regression_cycling){
+    cv::Mat A,B,z,etalon_z;
+
+#if 1
+    //example with cycling from http://people.orie.cornell.edu/miketodd/or630/SimplexCyclingExample.pdf
+    A=(cv::Mat_<double>(4,1)<<10,-57,-9,-24);
+    B=(cv::Mat_<double>(3,5)<<0.5,-5.5,-2.5,9,0,0.5,-1.5,-0.5,1,0,1,0,0,0,1);
+    std::cout<<"here A goes\n"<<A<<"\n";
+    int res=cv::optim::solveLP(A,B,z);
+    printf("res=%d\n",res);
+    printf("scalar %g\n",z.dot(A));
+    std::cout<<"here z goes\n"<<z<<"\n";
+    ASSERT_EQ(z.dot(A),1);
+    //ASSERT_EQ(res,1);
+#endif
+}
diff --git a/modules/calib3d/test/test_precomp.cpp b/modules/optim/test/test_main.cpp
similarity index 58%
rename from modules/calib3d/test/test_precomp.cpp
rename to modules/optim/test/test_main.cpp
index 5956e13e3e..6b24993447 100644
--- a/modules/calib3d/test/test_precomp.cpp
+++ b/modules/optim/test/test_main.cpp
@@ -1 +1,3 @@
 #include "test_precomp.hpp"
+
+CV_TEST_MAIN("cv")
diff --git a/modules/optim/test/test_precomp.hpp b/modules/optim/test/test_precomp.hpp
new file mode 100644
index 0000000000..9a86cab4b5
--- /dev/null
+++ b/modules/optim/test/test_precomp.hpp
@@ -0,0 +1,15 @@
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/optim.hpp"
+
+#endif
diff --git a/modules/photo/CMakeLists.txt b/modules/photo/CMakeLists.txt
index 08a72ea928..5b47d3a1d9 100644
--- a/modules/photo/CMakeLists.txt
+++ b/modules/photo/CMakeLists.txt
@@ -1,2 +1,7 @@
 set(the_description "Computational Photography")
-ocv_define_module(photo opencv_imgproc)
+
+if(HAVE_CUDA)
+  ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wmissing-declarations)
+endif()
+
+ocv_define_module(photo opencv_imgproc OPTIONAL opencv_gpuarithm opencv_gpuimgproc)
diff --git a/modules/photo/doc/denoising.rst b/modules/photo/doc/denoising.rst
index 97625d3b31..4c64b07c29 100644
--- a/modules/photo/doc/denoising.rst
+++ b/modules/photo/doc/denoising.rst
@@ -10,6 +10,8 @@ with several computational optimizations. Noise expected to be a gaussian white
 
 .. ocv:function:: void fastNlMeansDenoising( InputArray src, OutputArray dst, float h=3, int templateWindowSize=7, int searchWindowSize=21 )
 
+.. ocv:pyfunction:: cv2.fastNlMeansDenoising(src[, dst[, h[, templateWindowSize[, searchWindowSize]]]]) -> dst
+
     :param src: Input 8-bit 1-channel, 2-channel or 3-channel image.
 
     :param dst: Output image with the same size and type as  ``src`` .
@@ -30,6 +32,8 @@ Modification of ``fastNlMeansDenoising`` function for colored images
 
 .. ocv:function:: void fastNlMeansDenoisingColored( InputArray src, OutputArray dst, float h=3, float hColor=3, int templateWindowSize=7, int searchWindowSize=21 )
 
+.. ocv:pyfunction:: cv2.fastNlMeansDenoisingColored(src[, dst[, h[, hColor[, templateWindowSize[, searchWindowSize]]]]]) -> dst
+
     :param src: Input 8-bit 3-channel image.
 
     :param dst: Output image with the same size and type as  ``src`` .
@@ -51,6 +55,8 @@ For more details see http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.131
 
 .. ocv:function:: void fastNlMeansDenoisingMulti( InputArrayOfArrays srcImgs, OutputArray dst, int imgToDenoiseIndex, int temporalWindowSize, float h=3, int templateWindowSize=7, int searchWindowSize=21 )
 
+.. ocv:pyfunction:: cv2.fastNlMeansDenoisingMulti(srcImgs, imgToDenoiseIndex, temporalWindowSize[, dst[, h[, templateWindowSize[, searchWindowSize]]]]) -> dst
+
     :param srcImgs: Input 8-bit 1-channel, 2-channel or 3-channel images sequence. All images should have the same type and size.
 
     :param imgToDenoiseIndex: Target image to denoise index in ``srcImgs`` sequence
@@ -71,6 +77,8 @@ Modification of ``fastNlMeansDenoisingMulti`` function for colored images sequen
 
 .. ocv:function:: void fastNlMeansDenoisingColoredMulti( InputArrayOfArrays srcImgs, OutputArray dst, int imgToDenoiseIndex, int temporalWindowSize, float h=3, float hColor=3, int templateWindowSize=7, int searchWindowSize=21 )
 
+.. ocv:pyfunction:: cv2.fastNlMeansDenoisingColoredMulti(srcImgs, imgToDenoiseIndex, temporalWindowSize[, dst[, h[, hColor[, templateWindowSize[, searchWindowSize]]]]]) -> dst
+
     :param srcImgs: Input 8-bit 3-channel images sequence. All images should have the same type and size.
 
     :param imgToDenoiseIndex: Target image to denoise index in ``srcImgs`` sequence
@@ -89,3 +97,101 @@ Modification of ``fastNlMeansDenoisingMulti`` function for colored images sequen
 
 The function converts images to CIELAB colorspace and then separately denoise L and AB components with given h parameters using ``fastNlMeansDenoisingMulti`` function.
 
+
+
+gpu::nonLocalMeans
+-------------------
+Performs pure non local means denoising without any simplification, and thus it is not fast.
+
+.. ocv:function:: void gpu::nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null())
+
+    :param src: Source image. Supports only CV_8UC1, CV_8UC2 and CV_8UC3.
+
+    :param dst: Destination image.
+
+    :param h: Filter sigma regulating filter strength for color.
+
+    :param search_window: Size of search window.
+
+    :param block_size: Size of block used for computing weights.
+
+    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param stream: Stream for the asynchronous version.
+
+.. seealso::
+
+    :ocv:func:`fastNlMeansDenoising`
+
+
+
+gpu::FastNonLocalMeansDenoising
+-------------------------------
+.. ocv:class:: gpu::FastNonLocalMeansDenoising
+
+    ::
+
+        class FastNonLocalMeansDenoising
+        {
+        public:
+            //! Simple method, recommended for grayscale images (though it supports multichannel images)
+            void simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
+            //! Processes luminance and color components separatelly
+            void labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
+        };
+
+The class implements fast approximate Non Local Means Denoising algorithm.
+
+
+
+gpu::FastNonLocalMeansDenoising::simpleMethod()
+-----------------------------------------------
+Perform image denoising using Non-local Means Denoising algorithm http://www.ipol.im/pub/algo/bcm_non_local_means_denoising with several computational optimizations. Noise expected to be a gaussian white noise
+
+.. ocv:function:: void gpu::FastNonLocalMeansDenoising::simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
+
+    :param src: Input 8-bit 1-channel, 2-channel or 3-channel image.
+
+    :param dst: Output image with the same size and type as  ``src`` .
+
+    :param h: Parameter regulating filter strength. Big h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise
+
+    :param search_window: Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater search_window - greater denoising time. Recommended value 21 pixels
+
+    :param block_size: Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels
+
+    :param stream: Stream for the asynchronous invocations.
+
+This function expected to be applied to grayscale images. For colored images look at ``FastNonLocalMeansDenoising::labMethod``.
+
+.. seealso::
+
+    :ocv:func:`fastNlMeansDenoising`
+
+
+
+gpu::FastNonLocalMeansDenoising::labMethod()
+--------------------------------------------
+Modification of ``FastNonLocalMeansDenoising::simpleMethod`` for color images
+
+.. ocv:function:: void gpu::FastNonLocalMeansDenoising::labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null())
+
+    :param src: Input 8-bit 3-channel image.
+
+    :param dst: Output image with the same size and type as  ``src`` .
+
+    :param h_luminance: Parameter regulating filter strength. Big h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise
+
+    :param float: The same as h but for color components. For most images value equals 10 will be enought to remove colored noise and do not distort colors
+
+    :param search_window: Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater search_window - greater denoising time. Recommended value 21 pixels
+
+    :param block_size: Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels
+
+    :param stream: Stream for the asynchronous invocations.
+
+The function converts image to CIELAB colorspace and then separately denoise L and AB components with given h parameters using ``FastNonLocalMeansDenoising::simpleMethod`` function.
+
+.. seealso::
+
+    :ocv:func:`fastNlMeansDenoisingColored`
diff --git a/modules/photo/doc/inpainting.rst b/modules/photo/doc/inpainting.rst
index ef69ebc708..625f70a263 100644
--- a/modules/photo/doc/inpainting.rst
+++ b/modules/photo/doc/inpainting.rst
@@ -12,7 +12,6 @@ Restores the selected region in an image using the region neighborhood.
 .. ocv:pyfunction:: cv2.inpaint(src, inpaintMask, inpaintRadius, flags[, dst]) -> dst
 
 .. ocv:cfunction:: void cvInpaint( const CvArr* src, const CvArr* inpaint_mask, CvArr* dst, double inpaintRange, int flags )
-.. ocv:pyoldfunction:: cv.Inpaint(src, mask, dst, inpaintRadius, flags) -> None
 
     :param src: Input 8-bit 1-channel or 3-channel image.
 
@@ -24,10 +23,21 @@ Restores the selected region in an image using the region neighborhood.
 
     :param flags: Inpainting method that could be one of the following:
 
-            * **INPAINT_NS**     Navier-Stokes based method.
+            * **INPAINT_NS**     Navier-Stokes based method [Navier01]
 
             * **INPAINT_TELEA**     Method by Alexandru Telea  [Telea04]_.
 
 The function reconstructs the selected image area from the pixel near the area boundary. The function may be used to remove dust and scratches from a scanned photo, or to remove undesirable objects from still images or video. See
 http://en.wikipedia.org/wiki/Inpainting
 for more details.
+
+.. note::
+
+   * An example using the inpainting technique can be found at opencv_source_code/samples/cpp/inpaint.cpp
+
+   * (Python) An example using the inpainting technique can be found at opencv_source_code/samples/python2/inpaint.py
+
+
+.. [Telea04] Telea, Alexandru. "An image inpainting technique based on the fast marching method." Journal of graphics tools 9, no. 1 (2004): 23-34.
+
+.. [Navier01] Bertalmio, Marcelo, Andrea L. Bertozzi, and Guillermo Sapiro. "Navier-stokes, fluid dynamics, and image and video inpainting." In Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on, vol. 1, pp. I-355. IEEE, 2001.
diff --git a/modules/photo/include/opencv2/photo.hpp b/modules/photo/include/opencv2/photo.hpp
index 07e7e098a9..185b8dcc90 100644
--- a/modules/photo/include/opencv2/photo.hpp
+++ b/modules/photo/include/opencv2/photo.hpp
@@ -46,10 +46,6 @@
 #include "opencv2/core.hpp"
 #include "opencv2/imgproc.hpp"
 
-#include "opencv2/photo/photo_c.h"
-
-#ifdef __cplusplus
-
 /*! \namespace cv
  Namespace where all the C++ OpenCV functionality resides
  */
@@ -59,8 +55,8 @@ namespace cv
 //! the inpainting algorithm
 enum
 {
-    INPAINT_NS=CV_INPAINT_NS, // Navier-Stokes algorithm
-    INPAINT_TELEA=CV_INPAINT_TELEA // A. Telea algorithm
+    INPAINT_NS    = 0, // Navier-Stokes algorithm
+    INPAINT_TELEA = 1 // A. Telea algorithm
 };
 
 //! restores the damaged image areas using one of the available intpainting algorithms
@@ -84,8 +80,6 @@ CV_EXPORTS_W void fastNlMeansDenoisingColoredMulti( InputArrayOfArrays srcImgs,
                                                     float h = 3, float hColor = 3,
                                                     int templateWindowSize = 7, int searchWindowSize = 21);
 
-}
-
-#endif //__cplusplus
+} // cv
 
 #endif
diff --git a/modules/photo/include/opencv2/photo/gpu.hpp b/modules/photo/include/opencv2/photo/gpu.hpp
new file mode 100644
index 0000000000..a8b3859e75
--- /dev/null
+++ b/modules/photo/include/opencv2/photo/gpu.hpp
@@ -0,0 +1,71 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2008-2012, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PHOTO_GPU_HPP__
+#define __OPENCV_PHOTO_GPU_HPP__
+
+#include "opencv2/core/gpu.hpp"
+
+namespace cv { namespace gpu {
+
+//! Brute force non-local means algorith (slow but universal)
+CV_EXPORTS void nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null());
+
+//! Fast (but approximate)version of non-local means algorith similar to CPU function (running sums technique)
+class CV_EXPORTS FastNonLocalMeansDenoising
+{
+public:
+    //! Simple method, recommended for grayscale images (though it supports multichannel images)
+    void simpleMethod(const GpuMat& src, GpuMat& dst, float h, int search_window = 21, int block_size = 7, Stream& s = Stream::Null());
+
+    //! Processes luminance and color components separatelly
+    void labMethod(const GpuMat& src, GpuMat& dst, float h_luminance, float h_color, int search_window = 21, int block_size = 7, Stream& s = Stream::Null());
+
+private:
+
+    GpuMat buffer, extended_src_buffer;
+    GpuMat lab, l, ab;
+};
+
+}} // namespace cv { namespace gpu {
+
+#endif /* __OPENCV_PHOTO_GPU_HPP__ */
diff --git a/modules/photo/include/opencv2/photo/photo.hpp b/modules/photo/include/opencv2/photo/photo.hpp
index 41aa7ae405..8af5e9f0fa 100644
--- a/modules/photo/include/opencv2/photo/photo.hpp
+++ b/modules/photo/include/opencv2/photo/photo.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/photo.hpp"
\ No newline at end of file
+#include "opencv2/photo.hpp"
diff --git a/modules/gpu/perf/perf_denoising.cpp b/modules/photo/perf/perf_gpu.cpp
similarity index 57%
rename from modules/gpu/perf/perf_denoising.cpp
rename to modules/photo/perf/perf_gpu.cpp
index 6f03994bd9..ec62f7a0cc 100644
--- a/modules/gpu/perf/perf_denoising.cpp
+++ b/modules/photo/perf/perf_gpu.cpp
@@ -1,69 +1,73 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
 #include "perf_precomp.hpp"
 
+#include "opencv2/photo/gpu.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#if defined (HAVE_CUDA) && defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUIMGPROC)
+
 using namespace std;
 using namespace testing;
+using namespace perf;
 
 #define GPU_DENOISING_IMAGE_SIZES testing::Values(perf::szVGA, perf::sz720p)
 
-//////////////////////////////////////////////////////////////////////
-// BilateralFilter
-
-DEF_PARAM_TEST(Sz_Depth_Cn_KernelSz, cv::Size, MatDepth, MatCn, int);
-
-PERF_TEST_P(Sz_Depth_Cn_KernelSz, Denoising_BilateralFilter,
-            Combine(GPU_DENOISING_IMAGE_SIZES,
-                    Values(CV_8U, CV_32F),
-                    GPU_CHANNELS_1_3,
-                    Values(3, 5, 9)))
-{
-    declare.time(60.0);
-
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-    const int kernel_size = GET_PARAM(3);
-
-    const float sigma_color = 7;
-    const float sigma_spatial = 5;
-    const int borderMode = cv::BORDER_REFLECT101;
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-        cv::gpu::GpuMat dst;
-
-        TEST_CYCLE() cv::gpu::bilateralFilter(d_src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        cv::Mat dst;
-
-        TEST_CYCLE() cv::bilateralFilter(src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
-
-        CPU_SANITY_CHECK(dst);
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // nonLocalMeans
 
 DEF_PARAM_TEST(Sz_Depth_Cn_WinSz_BlockSz, cv::Size, MatDepth, MatCn, int, int);
 
-PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, Denoising_NonLocalMeans,
+PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, GPU_NonLocalMeans,
             Combine(GPU_DENOISING_IMAGE_SIZES,
                     Values<MatDepth>(CV_8U),
                     GPU_CHANNELS_1_3,
                     Values(21),
                     Values(5)))
 {
-    declare.time(60.0);
+    declare.time(600.0);
 
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -100,7 +104,7 @@ PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, Denoising_NonLocalMeans,
 
 DEF_PARAM_TEST(Sz_Depth_Cn_WinSz_BlockSz, cv::Size, MatDepth, MatCn, int, int);
 
-PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, Denoising_FastNonLocalMeans,
+PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, GPU_FastNonLocalMeans,
             Combine(GPU_DENOISING_IMAGE_SIZES,
                     Values<MatDepth>(CV_8U),
                     GPU_CHANNELS_1_3,
@@ -146,7 +150,7 @@ PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, Denoising_FastNonLocalMeans,
 
 DEF_PARAM_TEST(Sz_Depth_WinSz_BlockSz, cv::Size, MatDepth, int, int);
 
-PERF_TEST_P(Sz_Depth_WinSz_BlockSz, Denoising_FastNonLocalMeansColored,
+PERF_TEST_P(Sz_Depth_WinSz_BlockSz, GPU_FastNonLocalMeansColored,
             Combine(GPU_DENOISING_IMAGE_SIZES,
                     Values<MatDepth>(CV_8U),
                     Values(21),
@@ -185,3 +189,5 @@ PERF_TEST_P(Sz_Depth_WinSz_BlockSz, Denoising_FastNonLocalMeansColored,
         CPU_SANITY_CHECK(dst);
     }
 }
+
+#endif
diff --git a/modules/photo/perf/perf_inpaint.cpp b/modules/photo/perf/perf_inpaint.cpp
index fdf4ae0692..2debcf5c5e 100644
--- a/modules/photo/perf/perf_inpaint.cpp
+++ b/modules/photo/perf/perf_inpaint.cpp
@@ -14,7 +14,7 @@ typedef perf::TestBaseWithParam<InpaintArea_InpaintingMethod_t> InpaintArea_Inpa
 PERF_TEST_P(InpaintArea_InpaintingMethod, inpaint,
             testing::Combine(
                 testing::Values(::perf::szSmall24, ::perf::szSmall32, ::perf::szSmall64),
-                testing::ValuesIn(InpaintingMethod::all())
+                InpaintingMethod::all()
                 )
             )
 {
diff --git a/modules/photo/perf/perf_main.cpp b/modules/photo/perf/perf_main.cpp
index f5863c1974..fd37812126 100644
--- a/modules/photo/perf/perf_main.cpp
+++ b/modules/photo/perf/perf_main.cpp
@@ -1,3 +1,11 @@
 #include "perf_precomp.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-CV_PERF_TEST_MAIN(photo)
+static const char * impls[] = {
+#ifdef HAVE_CUDA
+    "cuda",
+#endif
+    "plain"
+};
+
+CV_PERF_TEST_MAIN_WITH_IMPLS(photo, impls, perf::printCudaInfo())
diff --git a/modules/photo/perf/perf_precomp.cpp b/modules/photo/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/photo/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/photo/src/arrays.hpp b/modules/photo/src/arrays.hpp
index ae01e9af82..a33018e599 100644
--- a/modules/photo/src/arrays.hpp
+++ b/modules/photo/src/arrays.hpp
@@ -157,5 +157,3 @@ template <class T> struct Array4d {
 };
 
 #endif
-
-
diff --git a/modules/gpu/src/cuda/nlm.cu b/modules/photo/src/cuda/nlm.cu
similarity index 95%
rename from modules/gpu/src/cuda/nlm.cu
rename to modules/photo/src/cuda/nlm.cu
index cd3f0b5c3a..44ed4dc6b9 100644
--- a/modules/gpu/src/cuda/nlm.cu
+++ b/modules/photo/src/cuda/nlm.cu
@@ -12,7 +12,6 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
-// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -29,7 +28,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -41,14 +40,12 @@
 //
 //M*/
 
-#if !defined CUDA_DISABLER
-
-#include "opencv2/gpu/device/common.hpp"
-#include "opencv2/gpu/device/vec_traits.hpp"
-#include "opencv2/gpu/device/vec_math.hpp"
-#include "opencv2/gpu/device/functional.hpp"
-#include "opencv2/gpu/device/reduce.hpp"
-#include "opencv2/gpu/device/border_interpolate.hpp"
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+#include "opencv2/core/cuda/functional.hpp"
+#include "opencv2/core/cuda/reduce.hpp"
+#include "opencv2/core/cuda/border_interpolate.hpp"
 
 using namespace cv::gpu;
 
@@ -58,7 +55,7 @@ typedef unsigned short ushort;
 //////////////////////////////////////////////////////////////////////////////////
 //// Non Local Means Denosing
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -162,11 +159,11 @@ namespace cv { namespace gpu { namespace device
 
             static func_t funcs[] =
             {
-                nlm_caller<T, BrdReflect101>,
-                nlm_caller<T, BrdReplicate>,
                 nlm_caller<T, BrdConstant>,
+                nlm_caller<T, BrdReplicate>,
                 nlm_caller<T, BrdReflect>,
                 nlm_caller<T, BrdWrap>,
+                nlm_caller<T, BrdReflect101>
             };
             funcs[borderMode](src, dst, search_radius, block_radius, h, stream);
         }
@@ -180,7 +177,7 @@ namespace cv { namespace gpu { namespace device
 //////////////////////////////////////////////////////////////////////////////////
 //// Non Local Means Denosing (fast approximate version)
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -191,7 +188,7 @@ namespace cv { namespace gpu { namespace device
             template <int BLOCK_SIZE>
             static __device__ __forceinline__ thrust::tuple<volatile float*, volatile float*> smem_tuple(float* smem)
             {
-                return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE);
+                return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE);
             }
 
             static __device__ __forceinline__ thrust::tuple<float&, float&> tie(float& val1, float& val2)
@@ -210,7 +207,7 @@ namespace cv { namespace gpu { namespace device
             template <int BLOCK_SIZE>
             static __device__ __forceinline__ thrust::tuple<volatile float*, volatile float*, volatile float*> smem_tuple(float* smem)
             {
-                return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE);
+                return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE);
             }
 
             static __device__ __forceinline__ thrust::tuple<float&, float&, float&> tie(float& val1, float2& val2)
@@ -229,7 +226,7 @@ namespace cv { namespace gpu { namespace device
             template <int BLOCK_SIZE>
             static __device__ __forceinline__ thrust::tuple<volatile float*, volatile float*, volatile float*, volatile float*> smem_tuple(float* smem)
             {
-                return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE);
+                return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE);
             }
 
             static __device__ __forceinline__ thrust::tuple<float&, float&, float&, float&> tie(float& val1, float3& val2)
@@ -248,7 +245,7 @@ namespace cv { namespace gpu { namespace device
             template <int BLOCK_SIZE>
             static __device__ __forceinline__ thrust::tuple<volatile float*, volatile float*, volatile float*, volatile float*, volatile float*> smem_tuple(float* smem)
             {
-                return cv::gpu::device::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE, smem + 4 * BLOCK_SIZE);
+                return cv::gpu::cudev::smem_tuple(smem, smem + BLOCK_SIZE, smem + 2 * BLOCK_SIZE, smem + 3 * BLOCK_SIZE, smem + 4 * BLOCK_SIZE);
             }
 
             static __device__ __forceinline__ thrust::tuple<float&, float&, float&, float&, float&> tie(float& val1, float4& val2)
@@ -565,6 +562,3 @@ namespace cv { namespace gpu { namespace device
         }
     }
 }}}
-
-
-#endif /* CUDA_DISABLER */
diff --git a/modules/photo/src/denoising.cpp b/modules/photo/src/denoising.cpp
index 299390f154..b8bae1281f 100644
--- a/modules/photo/src/denoising.cpp
+++ b/modules/photo/src/denoising.cpp
@@ -59,22 +59,22 @@ void cv::fastNlMeansDenoising( InputArray _src, OutputArray _dst, float h,
 
     switch (src.type()) {
         case CV_8U:
-            parallel_for(cv::BlockedRange(0, src.rows),
+            parallel_for_(cv::Range(0, src.rows),
                 FastNlMeansDenoisingInvoker<uchar>(
                     src, dst, templateWindowSize, searchWindowSize, h));
             break;
         case CV_8UC2:
-            parallel_for(cv::BlockedRange(0, src.rows),
+            parallel_for_(cv::Range(0, src.rows),
                 FastNlMeansDenoisingInvoker<cv::Vec2b>(
                     src, dst, templateWindowSize, searchWindowSize, h));
             break;
         case CV_8UC3:
-            parallel_for(cv::BlockedRange(0, src.rows),
+            parallel_for_(cv::Range(0, src.rows),
                 FastNlMeansDenoisingInvoker<cv::Vec3b>(
                     src, dst, templateWindowSize, searchWindowSize, h));
             break;
         default:
-            CV_Error(CV_StsBadArg,
+            CV_Error(Error::StsBadArg,
                 "Unsupported image format! Only CV_8UC1, CV_8UC2 and CV_8UC3 are supported");
     }
 }
@@ -88,12 +88,12 @@ void cv::fastNlMeansDenoisingColored( InputArray _src, OutputArray _dst,
     Mat dst = _dst.getMat();
 
     if (src.type() != CV_8UC3) {
-        CV_Error(CV_StsBadArg, "Type of input image should be CV_8UC3!");
+        CV_Error(Error::StsBadArg, "Type of input image should be CV_8UC3!");
         return;
     }
 
     Mat src_lab;
-    cvtColor(src, src_lab, CV_LBGR2Lab);
+    cvtColor(src, src_lab, COLOR_LBGR2Lab);
 
     Mat l(src.size(), CV_8U);
     Mat ab(src.size(), CV_8UC2);
@@ -108,7 +108,7 @@ void cv::fastNlMeansDenoisingColored( InputArray _src, OutputArray _dst,
     Mat dst_lab(src.size(), src.type());
     mixChannels(l_ab_denoised, 2, &dst_lab, 1, from_to, 3);
 
-    cvtColor(dst_lab, dst, CV_Lab2LBGR);
+    cvtColor(dst_lab, dst, COLOR_Lab2LBGR);
 }
 
 static void fastNlMeansDenoisingMultiCheckPreconditions(
@@ -118,27 +118,27 @@ static void fastNlMeansDenoisingMultiCheckPreconditions(
 {
     int src_imgs_size = (int)srcImgs.size();
     if (src_imgs_size == 0) {
-        CV_Error(CV_StsBadArg, "Input images vector should not be empty!");
+        CV_Error(Error::StsBadArg, "Input images vector should not be empty!");
     }
 
     if (temporalWindowSize % 2 == 0 ||
         searchWindowSize % 2 == 0 ||
         templateWindowSize % 2 == 0) {
-        CV_Error(CV_StsBadArg, "All windows sizes should be odd!");
+        CV_Error(Error::StsBadArg, "All windows sizes should be odd!");
     }
 
     int temporalWindowHalfSize = temporalWindowSize / 2;
     if (imgToDenoiseIndex - temporalWindowHalfSize < 0 ||
         imgToDenoiseIndex + temporalWindowHalfSize >= src_imgs_size)
     {
-        CV_Error(CV_StsBadArg,
+        CV_Error(Error::StsBadArg,
             "imgToDenoiseIndex and temporalWindowSize "
             "should be choosen corresponding srcImgs size!");
     }
 
     for (int i = 1; i < src_imgs_size; i++) {
         if (srcImgs[0].size() != srcImgs[i].size() || srcImgs[0].type() != srcImgs[i].type()) {
-            CV_Error(CV_StsBadArg, "Input images should have the same size and type!");
+            CV_Error(Error::StsBadArg, "Input images should have the same size and type!");
         }
     }
 }
@@ -159,25 +159,25 @@ void cv::fastNlMeansDenoisingMulti( InputArrayOfArrays _srcImgs, OutputArray _ds
 
     switch (srcImgs[0].type()) {
         case CV_8U:
-            parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
+            parallel_for_(cv::Range(0, srcImgs[0].rows),
                 FastNlMeansMultiDenoisingInvoker<uchar>(
                     srcImgs, imgToDenoiseIndex, temporalWindowSize,
                     dst, templateWindowSize, searchWindowSize, h));
             break;
         case CV_8UC2:
-            parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
+            parallel_for_(cv::Range(0, srcImgs[0].rows),
                 FastNlMeansMultiDenoisingInvoker<cv::Vec2b>(
                     srcImgs, imgToDenoiseIndex, temporalWindowSize,
                     dst, templateWindowSize, searchWindowSize, h));
             break;
         case CV_8UC3:
-            parallel_for(cv::BlockedRange(0, srcImgs[0].rows),
+            parallel_for_(cv::Range(0, srcImgs[0].rows),
                 FastNlMeansMultiDenoisingInvoker<cv::Vec3b>(
                     srcImgs, imgToDenoiseIndex, temporalWindowSize,
                     dst, templateWindowSize, searchWindowSize, h));
             break;
         default:
-            CV_Error(CV_StsBadArg,
+            CV_Error(Error::StsBadArg,
                 "Unsupported matrix format! Only uchar, Vec2b, Vec3b are supported");
     }
 }
@@ -201,7 +201,7 @@ void cv::fastNlMeansDenoisingColoredMulti( InputArrayOfArrays _srcImgs, OutputAr
     int src_imgs_size = (int)srcImgs.size();
 
     if (srcImgs[0].type() != CV_8UC3) {
-        CV_Error(CV_StsBadArg, "Type of input images should be CV_8UC3!");
+        CV_Error(Error::StsBadArg, "Type of input images should be CV_8UC3!");
         return;
     }
 
@@ -215,7 +215,7 @@ void cv::fastNlMeansDenoisingColoredMulti( InputArrayOfArrays _srcImgs, OutputAr
         src_lab[i] = Mat::zeros(srcImgs[0].size(), CV_8UC3);
         l[i] = Mat::zeros(srcImgs[0].size(), CV_8UC1);
         ab[i] = Mat::zeros(srcImgs[0].size(), CV_8UC2);
-        cvtColor(srcImgs[i], src_lab[i], CV_LBGR2Lab);
+        cvtColor(srcImgs[i], src_lab[i], COLOR_LBGR2Lab);
 
         Mat l_ab[] = { l[i], ab[i] };
         mixChannels(&src_lab[i], 1, l_ab, 2, from_to, 3);
@@ -236,7 +236,5 @@ void cv::fastNlMeansDenoisingColoredMulti( InputArrayOfArrays _srcImgs, OutputAr
     Mat dst_lab(srcImgs[0].size(), srcImgs[0].type());
     mixChannels(l_ab_denoised, 2, &dst_lab, 1, from_to, 3);
 
-    cvtColor(dst_lab, dst, CV_Lab2LBGR);
+    cvtColor(dst_lab, dst, COLOR_Lab2LBGR);
 }
-
-
diff --git a/modules/gpu/src/denoising.cpp b/modules/photo/src/denoising_gpu.cpp
similarity index 60%
rename from modules/gpu/src/denoising.cpp
rename to modules/photo/src/denoising_gpu.cpp
index 74c1a56686..2c1fabe14a 100644
--- a/modules/gpu/src/denoising.cpp
+++ b/modules/photo/src/denoising_gpu.cpp
@@ -22,13 +22,13 @@
 //
 //   * Redistribution's in binary form must reproduce the above copyright notice,
 //     this list of conditions and the following disclaimer in the documentation
-//     and/or other GpuMaterials provided with the distribution.
+//     and/or other materials provided with the distribution.
 //
 //   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or bpied warranties, including, but not limited to, the bpied
+// any express or implied warranties, including, but not limited to, the implied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -42,74 +42,45 @@
 
 #include "precomp.hpp"
 
+#include "opencv2/photo/gpu.hpp"
+#include "opencv2/core/private.gpu.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUIMGPROC
+#  include "opencv2/gpuimgproc.hpp"
+#endif
+
 using namespace cv;
 using namespace cv::gpu;
 
-#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
-
-void cv::gpu::bilateralFilter(const GpuMat&, GpuMat&, int, float, float, int, Stream&) { throw_nogpu(); }
-void cv::gpu::nonLocalMeans(const GpuMat&, GpuMat&, float, int, int, int, Stream&) { throw_nogpu(); }
-
-void cv::gpu::FastNonLocalMeansDenoising::simpleMethod(const GpuMat&, GpuMat&, float, int, int, Stream&) { throw_nogpu(); }
-void cv::gpu::FastNonLocalMeansDenoising::labMethod( const GpuMat&, GpuMat&, float, float, int, int, Stream&) { throw_nogpu(); }
+#if !defined (HAVE_CUDA) || !defined(HAVE_OPENCV_GPUARITHM) || !defined(HAVE_OPENCV_GPUIMGPROC)
 
+void cv::gpu::nonLocalMeans(const GpuMat&, GpuMat&, float, int, int, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::FastNonLocalMeansDenoising::simpleMethod(const GpuMat&, GpuMat&, float, int, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::FastNonLocalMeansDenoising::labMethod( const GpuMat&, GpuMat&, float, float, int, int, Stream&) { throw_no_cuda(); }
 
 #else
 
 //////////////////////////////////////////////////////////////////////////////////
 //// Non Local Means Denosing (brute force)
 
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
-        template<typename T>
-        void bilateral_filter_gpu(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t stream);
-
         template<typename T>
         void nlm_bruteforce_gpu(const PtrStepSzb& src, PtrStepSzb dst, int search_radius, int block_radius, float h, int borderMode, cudaStream_t stream);
     }
 }}}
 
-void cv::gpu::bilateralFilter(const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode, Stream& s)
-{
-    using cv::gpu::device::imgproc::bilateral_filter_gpu;
-
-    typedef void (*func_t)(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t s);
-
-    static const func_t funcs[6][4] =
-    {
-        {bilateral_filter_gpu<uchar>      , 0 /*bilateral_filter_gpu<uchar2>*/ , bilateral_filter_gpu<uchar3>      , bilateral_filter_gpu<uchar4>      },
-        {0 /*bilateral_filter_gpu<schar>*/, 0 /*bilateral_filter_gpu<schar2>*/ , 0 /*bilateral_filter_gpu<schar3>*/, 0 /*bilateral_filter_gpu<schar4>*/},
-        {bilateral_filter_gpu<ushort>     , 0 /*bilateral_filter_gpu<ushort2>*/, bilateral_filter_gpu<ushort3>     , bilateral_filter_gpu<ushort4>     },
-        {bilateral_filter_gpu<short>      , 0 /*bilateral_filter_gpu<short2>*/ , bilateral_filter_gpu<short3>      , bilateral_filter_gpu<short4>      },
-        {0 /*bilateral_filter_gpu<int>*/  , 0 /*bilateral_filter_gpu<int2>*/   , 0 /*bilateral_filter_gpu<int3>*/  , 0 /*bilateral_filter_gpu<int4>*/  },
-        {bilateral_filter_gpu<float>      , 0 /*bilateral_filter_gpu<float2>*/ , bilateral_filter_gpu<float3>      , bilateral_filter_gpu<float4>      }
-    };
-
-    sigma_color = (sigma_color <= 0 ) ? 1 : sigma_color;
-    sigma_spatial = (sigma_spatial <= 0 ) ? 1 : sigma_spatial;
-
-
-    int radius = (kernel_size <= 0) ? cvRound(sigma_spatial*1.5) : kernel_size/2;
-    kernel_size = std::max(radius, 1)*2 + 1;
-
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
-    const func_t func = funcs[src.depth()][src.channels() - 1];
-    CV_Assert(func != 0);
-
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
-
-    int gpuBorderType;
-    CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
-
-    dst.create(src.size(), src.type());
-    func(src, dst, kernel_size, sigma_spatial, sigma_color, gpuBorderType, StreamAccessor::getStream(s));
-}
-
 void cv::gpu::nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window, int block_window, int borderMode, Stream& s)
 {
-    using cv::gpu::device::imgproc::nlm_bruteforce_gpu;
+    using cv::gpu::cudev::imgproc::nlm_bruteforce_gpu;
     typedef void (*func_t)(const PtrStepSzb& src, PtrStepSzb dst, int search_radius, int block_radius, float h, int borderMode, cudaStream_t stream);
 
     static const func_t funcs[4] = { nlm_bruteforce_gpu<uchar>, nlm_bruteforce_gpu<uchar2>, nlm_bruteforce_gpu<uchar3>, 0/*nlm_bruteforce_gpu<uchar4>,*/ };
@@ -122,19 +93,11 @@ void cv::gpu::nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_
     int b = borderMode;
     CV_Assert(b == BORDER_REFLECT101 || b == BORDER_REPLICATE || b == BORDER_CONSTANT || b == BORDER_REFLECT || b == BORDER_WRAP);
 
-    int gpuBorderType;
-    CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
-
     dst.create(src.size(), src.type());
-    func(src, dst, search_window/2, block_window/2, h, gpuBorderType, StreamAccessor::getStream(s));
+    func(src, dst, search_window/2, block_window/2, h, borderMode, StreamAccessor::getStream(s));
 }
 
-
-//////////////////////////////////////////////////////////////////////////////////
-//// Non Local Means Denosing (fast approxinate)
-
-
-namespace cv { namespace gpu { namespace device
+namespace cv { namespace gpu { namespace cudev
 {
     namespace imgproc
     {
@@ -163,10 +126,10 @@ void cv::gpu::FastNonLocalMeansDenoising::simpleMethod(const GpuMat& src, GpuMat
     GpuMat src_hdr = extended_src(Rect(Point2i(border_size, border_size), src.size()));
 
     int bcols, brows;
-    device::imgproc::nln_fast_get_buffer_size(src_hdr, search_window, block_window, bcols, brows);
+    cudev::imgproc::nln_fast_get_buffer_size(src_hdr, search_window, block_window, bcols, brows);
     buffer.create(brows, bcols, CV_32S);
 
-    using namespace cv::gpu::device::imgproc;
+    using namespace cv::gpu::cudev::imgproc;
     typedef void (*nlm_fast_t)(const PtrStepSzb&, PtrStepSzb, PtrStepi, int, int, float, cudaStream_t);
     static const nlm_fast_t funcs[] = { nlm_fast_gpu<uchar>, nlm_fast_gpu<uchar2>, nlm_fast_gpu<uchar3>, 0};
 
@@ -179,20 +142,17 @@ void cv::gpu::FastNonLocalMeansDenoising::labMethod( const GpuMat& src, GpuMat&
     CV_Assert(src.type() == CV_8UC3);
 
     lab.create(src.size(), src.type());
-    cv::gpu::cvtColor(src, lab, CV_BGR2Lab, 0, s);
+    cv::gpu::cvtColor(src, lab, cv::COLOR_BGR2Lab, 0, s);
 
     l.create(src.size(), CV_8U);
     ab.create(src.size(), CV_8UC2);
-    device::imgproc::fnlm_split_channels(lab, l, ab, StreamAccessor::getStream(s));
+    cudev::imgproc::fnlm_split_channels(lab, l, ab, StreamAccessor::getStream(s));
 
     simpleMethod(l, l, h_luminance, search_window, block_window, s);
     simpleMethod(ab, ab, h_color, search_window, block_window, s);
 
-    device::imgproc::fnlm_merge_channels(l, ab, lab, StreamAccessor::getStream(s));
-    cv::gpu::cvtColor(lab, dst, CV_Lab2BGR, 0, s);
+    cudev::imgproc::fnlm_merge_channels(l, ab, lab, StreamAccessor::getStream(s));
+    cv::gpu::cvtColor(lab, dst, cv::COLOR_Lab2BGR, 0, s);
 }
 
-
 #endif
-
-
diff --git a/modules/photo/src/fast_nlmeans_denoising_invoker.hpp b/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
index c2e240b55f..3e9cc008bc 100644
--- a/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
+++ b/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
@@ -43,9 +43,6 @@
 #define __OPENCV_FAST_NLMEANS_DENOISING_INVOKER_HPP__
 
 #include "precomp.hpp"
-#include <opencv2/core.hpp>
-#include <opencv2/core/internal.hpp>
-#include <opencv2/imgproc.hpp>
 #include <limits>
 
 #include "fast_nlmeans_denoising_invoker_commons.hpp"
@@ -54,12 +51,12 @@
 using namespace cv;
 
 template <typename T>
-struct FastNlMeansDenoisingInvoker {
+struct FastNlMeansDenoisingInvoker : ParallelLoopBody {
     public:
         FastNlMeansDenoisingInvoker(const Mat& src, Mat& dst,
             int template_window_size, int search_window_size, const float h);
 
-        void operator() (const BlockedRange& range) const;
+        void operator() (const Range& range) const;
 
     private:
         void operator= (const FastNlMeansDenoisingInvoker&);
@@ -155,9 +152,9 @@ FastNlMeansDenoisingInvoker<T>::FastNlMeansDenoisingInvoker(
 }
 
 template <class T>
-void FastNlMeansDenoisingInvoker<T>::operator() (const BlockedRange& range) const {
-    int row_from = range.begin();
-    int row_to = range.end() - 1;
+void FastNlMeansDenoisingInvoker<T>::operator() (const Range& range) const {
+    int row_from = range.start;
+    int row_to = range.end - 1;
 
     Array2d<int> dist_sums(search_window_size_, search_window_size_);
 
diff --git a/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp b/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
index fd2b835e32..978f3170c7 100644
--- a/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
+++ b/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
@@ -42,10 +42,6 @@
 #ifndef __OPENCV_FAST_NLMEANS_DENOISING_INVOKER_COMMONS_HPP__
 #define __OPENCV_FAST_NLMEANS_DENOISING_INVOKER_COMMONS_HPP__
 
-#include <opencv2/core.hpp>
-#include <opencv2/core/internal.hpp>
-#include <opencv2/imgproc.hpp>
-
 using namespace cv;
 
 template <typename T> static inline int calcDist(const T a, const T b);
diff --git a/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp b/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
index 037c128a12..e2351a23c0 100644
--- a/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
+++ b/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
@@ -43,9 +43,6 @@
 #define __OPENCV_FAST_NLMEANS_MULTI_DENOISING_INVOKER_HPP__
 
 #include "precomp.hpp"
-#include <opencv2/core.hpp>
-#include <opencv2/core/internal.hpp>
-#include <opencv2/imgproc.hpp>
 #include <limits>
 
 #include "fast_nlmeans_denoising_invoker_commons.hpp"
@@ -54,13 +51,13 @@
 using namespace cv;
 
 template <typename T>
-struct FastNlMeansMultiDenoisingInvoker {
+struct FastNlMeansMultiDenoisingInvoker : ParallelLoopBody {
     public:
         FastNlMeansMultiDenoisingInvoker(
             const std::vector<Mat>& srcImgs, int imgToDenoiseIndex, int temporalWindowSize,
             Mat& dst, int template_window_size, int search_window_size, const float h);
 
-        void operator() (const BlockedRange& range) const;
+        void operator() (const Range& range) const;
 
     private:
         void operator= (const FastNlMeansMultiDenoisingInvoker&);
@@ -174,9 +171,9 @@ FastNlMeansMultiDenoisingInvoker<T>::FastNlMeansMultiDenoisingInvoker(
 }
 
 template <class T>
-void FastNlMeansMultiDenoisingInvoker<T>::operator() (const BlockedRange& range) const {
-    int row_from = range.begin();
-    int row_to = range.end() - 1;
+void FastNlMeansMultiDenoisingInvoker<T>::operator() (const Range& range) const {
+    int row_from = range.start;
+    int row_to = range.end - 1;
 
     Array3d<int> dist_sums(temporal_window_size_, search_window_size_, search_window_size_);
 
diff --git a/modules/photo/src/inpaint.cpp b/modules/photo/src/inpaint.cpp
index 672e6c05ab..f91db5f545 100644
--- a/modules/photo/src/inpaint.cpp
+++ b/modules/photo/src/inpaint.cpp
@@ -47,6 +47,7 @@
 
 #include "precomp.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/photo/photo_c.h"
 
 #undef CV_MAT_ELEM_PTR_FAST
 #define CV_MAT_ELEM_PTR_FAST( mat, row, col, pix_size )  \
@@ -716,6 +717,12 @@ icvNSInpaintFMM(const CvMat *f, CvMat *t, CvMat *out, int range, CvPriorityQueue
       }\
    }
 
+namespace cv {
+template<> void cv::DefaultDeleter<IplConvKernel>::operator ()(IplConvKernel* obj) const
+{
+  cvReleaseStructuringElement(&obj);
+}
+}
 
 void
 cvInpaint( const CvArr* _input_img, const CvArr* _inpaint_mask, CvArr* _output_img,
@@ -752,11 +759,11 @@ cvInpaint( const CvArr* _input_img, const CvArr* _inpaint_mask, CvArr* _output_i
     ecols = input_img->cols + 2;
     erows = input_img->rows + 2;
 
-    f = cvCreateMat(erows, ecols, CV_8UC1);
-    t = cvCreateMat(erows, ecols, CV_32FC1);
-    band = cvCreateMat(erows, ecols, CV_8UC1);
-    mask = cvCreateMat(erows, ecols, CV_8UC1);
-    el_cross = cvCreateStructuringElementEx(3,3,1,1,CV_SHAPE_CROSS,NULL);
+    f.reset(cvCreateMat(erows, ecols, CV_8UC1));
+    t.reset(cvCreateMat(erows, ecols, CV_32FC1));
+    band.reset(cvCreateMat(erows, ecols, CV_8UC1));
+    mask.reset(cvCreateMat(erows, ecols, CV_8UC1));
+    el_cross.reset(cvCreateStructuringElementEx(3,3,1,1,CV_SHAPE_CROSS,NULL));
 
     cvCopy( input_img, output_img );
     cvSet(mask,cvScalar(KNOWN,0,0,0));
@@ -765,7 +772,7 @@ cvInpaint( const CvArr* _input_img, const CvArr* _inpaint_mask, CvArr* _output_i
     cvSet(f,cvScalar(KNOWN,0,0,0));
     cvSet(t,cvScalar(1.0e6f,0,0,0));
     cvDilate(mask,band,el_cross,1);   // image with narrow band
-    Heap=new CvPriorityQueueFloat;
+    Heap=cv::makePtr<CvPriorityQueueFloat>();
     if (!Heap->Init(band))
         return;
     cvSub(band,mask,band,NULL);
@@ -776,14 +783,14 @@ cvInpaint( const CvArr* _input_img, const CvArr* _inpaint_mask, CvArr* _output_i
     cvSet(f,cvScalar(INSIDE,0,0,0),mask);
     cvSet(t,cvScalar(0,0,0,0),band);
 
-    if( flags == CV_INPAINT_TELEA )
+    if( flags == cv::INPAINT_TELEA )
     {
-        out = cvCreateMat(erows, ecols, CV_8UC1);
-        el_range = cvCreateStructuringElementEx(2*range+1,2*range+1,
-            range,range,CV_SHAPE_RECT,NULL);
+        out.reset(cvCreateMat(erows, ecols, CV_8UC1));
+        el_range.reset(cvCreateStructuringElementEx(2*range+1,2*range+1,
+            range,range,CV_SHAPE_RECT,NULL));
         cvDilate(mask,out,el_range,1);
         cvSub(out,mask,out,NULL);
-        Out=new CvPriorityQueueFloat;
+        Out=cv::makePtr<CvPriorityQueueFloat>();
         if (!Out->Init(out))
             return;
         if (!Out->Add(band))
@@ -793,10 +800,10 @@ cvInpaint( const CvArr* _input_img, const CvArr* _inpaint_mask, CvArr* _output_i
         icvCalcFMM(out,t,Out,true);
         icvTeleaInpaintFMM(mask,t,output_img,range,Heap);
     }
-    else if (flags == CV_INPAINT_NS) {
+    else if (flags == cv::INPAINT_NS) {
         icvNSInpaintFMM(mask,t,output_img,range,Heap);
     } else {
-        CV_Error( CV_StsBadArg, "The flags argument must be one of CV_INPAINT_TELEA or CV_INPAINT_NS" );
+        CV_Error( cv::Error::StsBadArg, "The flags argument must be one of CV_INPAINT_TELEA or CV_INPAINT_NS" );
     }
 }
 
diff --git a/modules/photo/src/precomp.cpp b/modules/photo/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/photo/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/photo/src/precomp.hpp b/modules/photo/src/precomp.hpp
index d397dedf4d..38ac3ffcda 100644
--- a/modules/photo/src/precomp.hpp
+++ b/modules/photo/src/precomp.hpp
@@ -43,10 +43,8 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
+#include "opencv2/core/private.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/photo.hpp"
 
 #ifdef HAVE_TEGRA_OPTIMIZATION
diff --git a/modules/photo/test/test_denoising.cpp b/modules/photo/test/test_denoising.cpp
index 57248c7728..ca4f63f222 100644
--- a/modules/photo/test/test_denoising.cpp
+++ b/modules/photo/test/test_denoising.cpp
@@ -62,8 +62,8 @@ TEST(Photo_DenoisingGrayscale, regression)
     string original_path = folder + "lena_noised_gaussian_sigma=10.png";
     string expected_path = folder + "lena_noised_denoised_grayscale_tw=7_sw=21_h=10.png";
 
-    Mat original = imread(original_path, CV_LOAD_IMAGE_GRAYSCALE);
-    Mat expected = imread(expected_path, CV_LOAD_IMAGE_GRAYSCALE);
+    Mat original = imread(original_path, IMREAD_GRAYSCALE);
+    Mat expected = imread(expected_path, IMREAD_GRAYSCALE);
 
     ASSERT_FALSE(original.empty()) << "Could not load input image " << original_path;
     ASSERT_FALSE(expected.empty()) << "Could not load reference image " << expected_path;
@@ -82,8 +82,8 @@ TEST(Photo_DenoisingColored, regression)
     string original_path = folder + "lena_noised_gaussian_sigma=10.png";
     string expected_path = folder + "lena_noised_denoised_lab12_tw=7_sw=21_h=10_h2=10.png";
 
-    Mat original = imread(original_path, CV_LOAD_IMAGE_COLOR);
-    Mat expected = imread(expected_path, CV_LOAD_IMAGE_COLOR);
+    Mat original = imread(original_path, IMREAD_COLOR);
+    Mat expected = imread(expected_path, IMREAD_COLOR);
 
     ASSERT_FALSE(original.empty()) << "Could not load input image " << original_path;
     ASSERT_FALSE(expected.empty()) << "Could not load reference image " << expected_path;
@@ -102,14 +102,14 @@ TEST(Photo_DenoisingGrayscaleMulti, regression)
     string folder = string(cvtest::TS::ptr()->get_data_path()) + "denoising/";
 
     string expected_path = folder + "lena_noised_denoised_multi_tw=7_sw=21_h=15.png";
-    Mat expected = imread(expected_path, CV_LOAD_IMAGE_GRAYSCALE);
+    Mat expected = imread(expected_path, IMREAD_GRAYSCALE);
     ASSERT_FALSE(expected.empty()) << "Could not load reference image " << expected_path;
 
     vector<Mat> original(imgs_count);
     for (int i = 0; i < imgs_count; i++)
     {
         string original_path = format("%slena_noised_gaussian_sigma=20_multi_%d.png", folder.c_str(), i);
-        original[i] = imread(original_path, CV_LOAD_IMAGE_GRAYSCALE);
+        original[i] = imread(original_path, IMREAD_GRAYSCALE);
         ASSERT_FALSE(original[i].empty()) << "Could not load input image " << original_path;
     }
 
@@ -127,14 +127,14 @@ TEST(Photo_DenoisingColoredMulti, regression)
     string folder = string(cvtest::TS::ptr()->get_data_path()) + "denoising/";
 
     string expected_path = folder + "lena_noised_denoised_multi_lab12_tw=7_sw=21_h=10_h2=15.png";
-    Mat expected = imread(expected_path, CV_LOAD_IMAGE_COLOR);
+    Mat expected = imread(expected_path, IMREAD_COLOR);
     ASSERT_FALSE(expected.empty()) << "Could not load reference image " << expected_path;
 
     vector<Mat> original(imgs_count);
     for (int i = 0; i < imgs_count; i++)
     {
         string original_path = format("%slena_noised_gaussian_sigma=20_multi_%d.png", folder.c_str(), i);
-        original[i] = imread(original_path, CV_LOAD_IMAGE_COLOR);
+        original[i] = imread(original_path, IMREAD_COLOR);
         ASSERT_FALSE(original[i].empty()) << "Could not load input image " << original_path;
     }
 
diff --git a/modules/photo/test/test_denoising_gpu.cpp b/modules/photo/test/test_denoising_gpu.cpp
new file mode 100644
index 0000000000..f8de826f2a
--- /dev/null
+++ b/modules/photo/test/test_denoising_gpu.cpp
@@ -0,0 +1,117 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#include "opencv2/photo/gpu.hpp"
+#include "opencv2/ts/gpu_test.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+#include "cvconfig.h"
+
+#if defined (HAVE_CUDA) && defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUIMGPROC)
+
+using namespace cvtest;
+
+////////////////////////////////////////////////////////
+// Brute Force Non local means
+
+TEST(GPU_BruteForceNonLocalMeans, Regression)
+{
+    using cv::gpu::GpuMat;
+
+    cv::Mat bgr  = readImage("../gpu/denoising/lena_noised_gaussian_sigma=20_multi_0.png", cv::IMREAD_COLOR);
+    ASSERT_FALSE(bgr.empty());
+
+    cv::Mat gray;
+    cv::cvtColor(bgr, gray, cv::COLOR_BGR2GRAY);
+
+    GpuMat dbgr, dgray;
+    cv::gpu::nonLocalMeans(GpuMat(bgr),  dbgr, 20);
+    cv::gpu::nonLocalMeans(GpuMat(gray), dgray, 20);
+
+#if 0
+    dumpImage("../gpu/denoising/nlm_denoised_lena_bgr.png", cv::Mat(dbgr));
+    dumpImage("../gpu/denoising/nlm_denoised_lena_gray.png", cv::Mat(dgray));
+#endif
+
+    cv::Mat bgr_gold  = readImage("../gpu/denoising/nlm_denoised_lena_bgr.png", cv::IMREAD_COLOR);
+    cv::Mat gray_gold  = readImage("../gpu/denoising/nlm_denoised_lena_gray.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(bgr_gold.empty() || gray_gold.empty());
+
+    EXPECT_MAT_NEAR(bgr_gold, dbgr, 1e-4);
+    EXPECT_MAT_NEAR(gray_gold, dgray, 1e-4);
+}
+
+////////////////////////////////////////////////////////
+// Fast Force Non local means
+
+TEST(GPU_FastNonLocalMeans, Regression)
+{
+    using cv::gpu::GpuMat;
+
+    cv::Mat bgr  = readImage("../gpu/denoising/lena_noised_gaussian_sigma=20_multi_0.png", cv::IMREAD_COLOR);
+    ASSERT_FALSE(bgr.empty());
+
+    cv::Mat gray;
+    cv::cvtColor(bgr, gray, cv::COLOR_BGR2GRAY);
+
+    GpuMat dbgr, dgray;
+    cv::gpu::FastNonLocalMeansDenoising fnlmd;
+
+    fnlmd.simpleMethod(GpuMat(gray),  dgray, 20);
+    fnlmd.labMethod(GpuMat(bgr),  dbgr, 20, 10);
+
+#if 0
+    dumpImage("../gpu/denoising/fnlm_denoised_lena_bgr.png", cv::Mat(dbgr));
+    dumpImage("../gpu/denoising/fnlm_denoised_lena_gray.png", cv::Mat(dgray));
+#endif
+
+    cv::Mat bgr_gold  = readImage("../gpu/denoising/fnlm_denoised_lena_bgr.png", cv::IMREAD_COLOR);
+    cv::Mat gray_gold  = readImage("../gpu/denoising/fnlm_denoised_lena_gray.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(bgr_gold.empty() || gray_gold.empty());
+
+    EXPECT_MAT_NEAR(bgr_gold, dbgr, 1);
+    EXPECT_MAT_NEAR(gray_gold, dgray, 1);
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/photo/test/test_inpaint.cpp b/modules/photo/test/test_inpaint.cpp
index 2bdb2ddaa5..3c341b27a0 100644
--- a/modules/photo/test/test_inpaint.cpp
+++ b/modules/photo/test/test_inpaint.cpp
@@ -78,14 +78,14 @@ void CV_InpaintTest::run( int )
     mask.convertTo(inv_mask, CV_8UC3, -1.0, 255.0);
 
     Mat mask1ch;
-    cv::cvtColor(mask, mask1ch, CV_BGR2GRAY);
+    cv::cvtColor(mask, mask1ch, COLOR_BGR2GRAY);
 
     Mat test = orig.clone();
     test.setTo(Scalar::all(255), mask1ch);
 
     Mat res1, res2;
-    inpaint( test, mask1ch, res1, 5, CV_INPAINT_NS );
-    inpaint( test, mask1ch, res2, 5, CV_INPAINT_TELEA );
+    inpaint( test, mask1ch, res1, 5, INPAINT_NS );
+    inpaint( test, mask1ch, res2, 5, INPAINT_TELEA );
 
     Mat diff1, diff2;
     absdiff( orig, res1, diff1 );
diff --git a/modules/photo/test/test_precomp.cpp b/modules/photo/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/photo/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/photo/test/test_precomp.hpp b/modules/photo/test/test_precomp.hpp
index 278abb496f..5b22a1c755 100644
--- a/modules/photo/test/test_precomp.hpp
+++ b/modules/photo/test/test_precomp.hpp
@@ -9,9 +9,9 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/photo.hpp"
 #include "opencv2/highgui.hpp"
-#include <iostream>
 
 #endif
diff --git a/modules/python/CMakeLists.txt b/modules/python/CMakeLists.txt
index d39e91921d..d3bed1695b 100644
--- a/modules/python/CMakeLists.txt
+++ b/modules/python/CMakeLists.txt
@@ -5,25 +5,26 @@
 if(WIN32 AND CMAKE_BUILD_TYPE STREQUAL "Debug")
   ocv_module_disable(python)
 endif()
-if(ANDROID OR IOS OR NOT PYTHONLIBS_FOUND OR NOT PYTHON_USE_NUMPY)
+
+if(ANDROID OR IOS OR NOT PYTHONLIBS_FOUND OR NOT PYTHON_NUMPY_INCLUDE_DIRS)
   ocv_module_disable(python)
 endif()
 
 set(the_description "The python bindings")
 ocv_add_module(python BINDINGS opencv_core opencv_flann opencv_imgproc opencv_video opencv_ml opencv_features2d opencv_highgui opencv_calib3d opencv_photo opencv_objdetect opencv_contrib opencv_legacy opencv_softcascade OPTIONAL opencv_nonfree)
 
-add_definitions(-DPYTHON_USE_NUMPY=1)
-
 ocv_module_include_directories(
     "${PYTHON_INCLUDE_PATH}"
-    "${PYTHON_NUMPY_INCLUDE_DIR}"
+    ${PYTHON_NUMPY_INCLUDE_DIRS}
     "${CMAKE_CURRENT_SOURCE_DIR}/src2"
     )
 
-
-
 set(opencv_hdrs
     "${OPENCV_MODULE_opencv_core_LOCATION}/include/opencv2/core.hpp"
+    "${OPENCV_MODULE_opencv_core_LOCATION}/include/opencv2/core/base.hpp"
+    "${OPENCV_MODULE_opencv_core_LOCATION}/include/opencv2/core/types.hpp"
+    "${OPENCV_MODULE_opencv_core_LOCATION}/include/opencv2/core/persistence.hpp"
+    "${OPENCV_MODULE_opencv_core_LOCATION}/include/opencv2/core/utility.hpp"
     "${OPENCV_MODULE_opencv_flann_LOCATION}/include/opencv2/flann/miniflann.hpp"
     "${OPENCV_MODULE_opencv_imgproc_LOCATION}/include/opencv2/imgproc.hpp"
     "${OPENCV_MODULE_opencv_video_LOCATION}/include/opencv2/video/background_segm.hpp"
@@ -49,13 +50,6 @@ set(cv2_generated_hdrs
     "${CMAKE_CURRENT_BINARY_DIR}/pyopencv_generated_type_reg.h"
     "${CMAKE_CURRENT_BINARY_DIR}/pyopencv_generated_const_reg.h")
 
-add_custom_command(
-   OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/generated0.i
-   COMMAND ${PYTHON_EXECUTABLE} "${CMAKE_CURRENT_SOURCE_DIR}/src2/gen.py" "${CMAKE_CURRENT_SOURCE_DIR}/src2"
-   DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/src2/api
-   DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/src2/defs
-   DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/src2/gen.py)
-
 add_custom_command(
    OUTPUT ${cv2_generated_hdrs}
    COMMAND ${PYTHON_EXECUTABLE} "${CMAKE_CURRENT_SOURCE_DIR}/src2/gen2.py" ${CMAKE_CURRENT_BINARY_DIR} ${opencv_hdrs}
@@ -63,7 +57,9 @@ add_custom_command(
    DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/src2/hdr_parser.py
    DEPENDS ${opencv_hdrs})
 
-add_library(${the_module} SHARED src2/cv2.cpp ${CMAKE_CURRENT_BINARY_DIR}/generated0.i ${cv2_generated_hdrs} src2/cv2.cv.hpp)
+add_library(${the_module} SHARED src2/cv2.cpp ${cv2_generated_hdrs})
+set_target_properties(${the_module} PROPERTIES COMPILE_DEFINITIONS OPENCV_NOSTL)
+
 if(PYTHON_DEBUG_LIBRARIES AND NOT PYTHON_LIBRARIES MATCHES "optimized.*debug")
   target_link_libraries(${the_module} debug ${PYTHON_DEBUG_LIBRARIES} optimized ${PYTHON_LIBRARIES})
 else()
@@ -71,7 +67,7 @@ else()
 endif()
 target_link_libraries(${the_module} ${OPENCV_MODULE_${the_module}_DEPS})
 
-execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "import distutils.sysconfig; print distutils.sysconfig.get_config_var('SO')"
+execute_process(COMMAND ${PYTHON_EXECUTABLE} -c "import distutils.sysconfig; print(distutils.sysconfig.get_config_var('SO'))"
                 RESULT_VARIABLE PYTHON_CVPY_PROCESS
                 OUTPUT_VARIABLE CVPY_SUFFIX
                 OUTPUT_STRIP_TRAILING_WHITESPACE)
@@ -85,6 +81,10 @@ if(ENABLE_SOLUTION_FOLDERS)
   set_target_properties(${the_module} PROPERTIES FOLDER "bindings")
 endif()
 
+if(MSVC)
+    add_definitions(-DCVAPI_EXPORTS)
+endif()
+
 if(CMAKE_COMPILER_IS_GNUCXX AND NOT ENABLE_NOISY_WARNINGS)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-function")
 endif()
@@ -112,4 +112,3 @@ install(TARGETS ${the_module}
         LIBRARY DESTINATION ${PYTHON_PACKAGES_PATH} COMPONENT main
         ARCHIVE DESTINATION ${PYTHON_PACKAGES_PATH} COMPONENT main
         )
-install(FILES src2/cv.py ${PYTHON_INSTALL_CONFIGURATIONS} DESTINATION ${PYTHON_PACKAGES_PATH} COMPONENT main)
diff --git a/modules/python/src2/api b/modules/python/src2/api
deleted file mode 100644
index dc02e4ab04..0000000000
--- a/modules/python/src2/api
+++ /dev/null
@@ -1,1805 +0,0 @@
-# Macros
-CV_RGB CvScalar
-  double red
-  double grn
-  double blu
-CV_MAT_CN int
-  int i
-CV_MAT_DEPTH int
-  int i
-Scalar CvScalar
-  double val0
-  double val1 0
-  double val2 0
-  double val3 0
-ScalarAll CvScalar
-  double val0123
-RealScalar CvScalar
-  double val0
-CV_IABS int
-  int a
-CV_CMP int
-  int a
-  int b
-CV_SIGN int
-  int a
-CV_FOURCC int
-  char c1
-  char c2
-  char c3
-  char c4
-CV_MAKETYPE int
-  int depth
-  int cn
-CV_8UC int
-  int n
-CV_8SC int
-  int n
-CV_16UC int
-  int n
-CV_16SC int
-  int n
-CV_32SC int
-  int n
-CV_32FC int
-  int n
-CV_64FC int
-  int n
-
-# Initialization
-CloneImage IplImage*
-  IplImage image
-SetImageCOI
-  IplImage image
-  int coi
-GetImageCOI int
-  IplImage image
-SetImageROI
-  IplImage image
-  CvRect rect
-ResetImageROI
-  IplImage image
-GetImageROI CvRect
-  IplImage image
-CloneMat CvMat*
-  CvMat mat
-CloneMatND CvMatND*
-  CvMatND mat
-
-# Accessing Elements and sub-Arrays
-
-Get1D CvScalar
-  CvArr arr
-  int idx
-Get2D CvScalar
-  CvArr arr
-  int idx0
-  int idx1
-Get3D CvScalar
-  CvArr arr
-  int idx0
-  int idx1
-  int idx2
-GetND CvScalar
-  CvArr arr
-  ints indices
-GetReal1D double
-  CvArr arr
-  int idx0
-GetReal2D double
-  CvArr arr
-  int idx0
-  int idx1
-GetReal3D double
-  CvArr arr
-  int idx0
-  int idx1
-  int idx2
-GetRealND double
-  CvArr arr
-  ints idx
-mGet double
-  CvMat mat
-  int row
-  int col
-Set1D
-  CvArr arr
-  int idx
-  CvScalar value
-Set2D
-  CvArr arr
-  int idx0
-  int idx1
-  CvScalar value
-Set3D
-  CvArr arr
-  int idx0
-  int idx1
-  int idx2
-  CvScalar value
-SetND
-  CvArr arr
-  ints indices
-  CvScalar value
-SetReal1D
-  CvArr arr
-  int idx
-  double value
-SetReal2D
-  CvArr arr
-  int idx0
-  int idx1
-  double value
-SetReal3D
-  CvArr arr
-  int idx0
-  int idx1
-  int idx2
-  double value
-SetRealND
-  CvArr arr
-  ints indices
-  double value
-mSet
-  CvMat mat
-  int row
-  int col
-  double value
-ClearND
-  CvArr arr
-  ints idx
-
-# Sequences
-CV_IS_SEQ_INDEX int
-  CvSeq s
-CV_IS_SEQ_CURVE int
-  CvSeq s
-CV_IS_SEQ_CLOSED int
-  CvSeq s
-CV_IS_SEQ_CONVEX int
-  CvSeq s
-CV_IS_SEQ_HOLE int
-  CvSeq s
-CV_IS_SEQ_SIMPLE int
-  CvSeq s
-
-
-# Curves and Shapes
-Line
-  CvArr img
-  CvPoint pt1
-  CvPoint pt2
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-Rectangle
-  CvArr img
-  CvPoint pt1
-  CvPoint pt2
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-Circle
-  CvArr img
-  CvPoint center
-  int radius
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-Ellipse
-  CvArr img
-  CvPoint center
-  CvSize axes
-  double angle
-  double start_angle
-  double end_angle
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-EllipseBox
-  CvArr img
-  CvBox2D box
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-FillPoly
-  CvArr img
-  pts_npts_contours polys
-  CvScalar color
-  int lineType 8
-  int shift 0
-FillConvexPoly
-  CvArr img
-  CvPoints pn
-  CvScalar color
-  int lineType 8
-  int shift 0
-PolyLine
-  CvArr img
-  pts_npts_contours polys
-  int is_closed
-  CvScalar color
-  int thickness 1
-  int lineType 8
-  int shift 0
-
-#Text
-InitFont font
-  CvFont font /O
-  int fontFace
-  double hscale
-  double vscale
-  double shear 0
-  int thickness 1
-  int lineType 8
-PutText
-  CvArr img
-  char* text
-  CvPoint org
-  CvFont* font
-  CvScalar color
-GetTextSize textSize,baseline
-  char* textString
-  CvFont* font
-  CvSize textSize /O
-  int baseline /O
-
-# Point Sets and Contours
-DrawContours
-  CvArr img
-  CvSeq contour
-  CvScalar external_color
-  CvScalar hole_color
-  int max_level
-  int thickness 1
-  int lineType 8
-  CvPoint offset cvPoint(0,0)
-
-# RTTI and Generic Functions
-Save
-  char* filename
-  generic structPtr
-  char* name NULL
-  char* comment NULL
-Load generic
-  char* filename
-  CvMemStorage storage NULL
-  char* name NULL
-
-# Accessing Elements and sub-Arrays
-GetRow submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  int row
-GetRows submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  int startRow
-  int endRow
-  int deltaRow 1
-GetCol submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  int col
-GetCols submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  int startCol
-  int endCol
-GetDiag submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  int diag 0
-GetSubRect submat
-  CvArr arr
-  CvMat submat /J:arr,O,A
-  CvRect rect
-GetSize CvSize
-  CvArr arr
-GetElemType int
-  CvArr arr
-
-# Copying and Filling
-Copy
-  CvArr src
-  CvArr dst
-  CvArr mask NULL
-Set
-  CvArr arr
-  CvScalar value
-  CvArr mask NULL
-SetZero
-  CvArr arr
-Zero
-  CvArr arr
-SetIdentity
-  CvArr mat
-  CvScalar value cvRealScalar(1)
-Range
-  CvArr mat
-  double start
-  double end
-
-# Transforms and Permutations
-# Reshape, ReshapeND - requires special data refcount code
-Repeat
-  CvArr src
-  CvArr dst
-Flip
-  CvArr src
-  CvArr dst NULL
-  int flipMode 0
-Split
-  CvArr src
-  CvArr dst0
-  CvArr dst1
-  CvArr dst2
-  CvArr dst3
-CvtPixToPlane
-  CvArr src
-  CvArr dst0
-  CvArr dst1
-  CvArr dst2
-  CvArr dst3
-Merge
-  CvArr src0
-  CvArr src1
-  CvArr src2
-  CvArr src3
-  CvArr dst
-MixChannels
-  cvarr_count src /K
-  cvarr_count dst
-  intpair fromTo 
-RandShuffle
-  CvArr mat
-  CvRNG* rng
-  double iter_factor 1.0
-Sort
-  CvArr src
-  CvArr dst
-  CvArr idxmat
-  int flags 0
-
-# Arithmetic, Logic and Comparison
-LUT
-  CvArr src
-  CvArr dst
-  CvArr lut
-ConvertScale
-  CvArr src
-  CvArr dst
-  double scale 1.0
-  double shift 0.0
-CvtScale
-  CvArr src
-  CvArr dst
-  double scale 1.0
-  double shift 0.0
-Scale
-  CvArr src
-  CvArr dst
-  double scale 1.0
-  double shift 0.0
-Convert
-  CvArr src
-  CvArr dst
-ConvertScaleAbs
-  CvArr src
-  CvArr dst
-  double scale 1.0
-  double shift 0.0
-Add
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  CvArr mask NULL
-AddS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-AddWeighted
-  CvArr src1
-  double alpha
-  CvArr src2
-  double beta
-  double gamma
-  CvArr dst
-Sub
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  CvArr mask NULL
-SubS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-SubRS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-Mul
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  double scale 1.0
-Div
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  double scale 1.0
-And
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  CvArr mask NULL
-AndS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-Or
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  CvArr mask NULL
-OrS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-Xor
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  CvArr mask NULL
-XorS
-  CvArr src
-  CvScalar value
-  CvArr dst
-  CvArr mask NULL
-Not
-  CvArr src
-  CvArr dst
-Cmp
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  int cmpOp
-CmpS
-  CvArr src
-  double value
-  CvArr dst
-  int cmpOp
-InRange
-  CvArr src
-  CvArr lower
-  CvArr upper
-  CvArr dst
-InRangeS
-  CvArr src
-  CvScalar lower
-  CvScalar upper
-  CvArr dst
-Max
-  CvArr src1
-  CvArr src2
-  CvArr dst
-MaxS
-  CvArr src
-  double value
-  CvArr dst
-Min
-  CvArr src1
-  CvArr src2
-  CvArr dst
-MinS
-  CvArr src
-  double value
-  CvArr dst
-AbsDiff
-  CvArr src1
-  CvArr src2
-  CvArr dst
-AbsDiffS
-  CvArr src
-  CvArr dst
-  CvScalar value
-Abs
-  CvArr src
-  CvArr dst
-
-# Statistics
-CountNonZero int
-  CvArr arr
-Sum CvScalar
-  CvArr arr
-Avg CvScalar
-  CvArr arr
-  CvArr mask NULL
-AvgSdv mean,stdDev
-  CvArr arr
-  CvScalar mean /O
-  CvScalar stdDev /O
-  CvArr mask NULL
-MinMaxLoc minVal,maxVal,minLoc,maxLoc
-  CvArr arr
-  double minVal /O
-  double maxVal /O
-  CvPoint minLoc /O
-  CvPoint maxLoc /O
-  CvArr mask NULL
-Norm double
-  CvArr arr1
-  CvArr arr2
-  int normType CV_L2
-  CvArr mask NULL
-Reduce
-  CvArr src
-  CvArr dst
-  int dim -1
-  int op CV_REDUCE_SUM
-
-# Linear Algebra
-DotProduct double
-  CvArr src1
-  CvArr src2
-Normalize
-  CvArr src
-  CvArr dst
-  double a 1.0
-  double b 0.0
-  int norm_type CV_L2
-  CvArr mask NULL
-CrossProduct
-  CvArr src1
-  CvArr src2
-  CvArr dst
-ScaleAdd
-  CvArr src1
-  CvScalar scale
-  CvArr src2
-  CvArr dst
-GEMM
-  CvArr src1
-  CvArr src2
-  double alpha
-  CvArr src3
-  double beta
-  CvArr dst
-  int tABC 0
-MatMulAdd
-  CvArr src1
-  CvArr src2
-  CvArr src3
-  CvArr dst
-MatMul
-  CvArr src1
-  CvArr src2
-  CvArr dst
-Transform
-  CvArr src
-  CvArr dst
-  CvMat transmat
-  CvMat shiftvec NULL
-PerspectiveTransform
-  CvArr src
-  CvArr dst
-  CvMat mat
-MulTransposed
-  CvArr src
-  CvArr dst
-  int order
-  CvArr delta NULL
-  double scale 1.0
-Trace CvScalar
-  CvArr mat
-Transpose
-  CvArr src
-  CvArr dst
-Det double
-  CvArr mat
-Invert double
-  CvArr src
-  CvArr dst
-  int method CV_LU
-Solve
-  CvArr A
-  CvArr B
-  CvArr X
-  int method CV_LU
-SVD
-  CvArr A
-  CvArr W
-  CvArr U NULL
-  CvArr V NULL
-  int flags 0
-SVBkSb
-  CvArr W
-  CvArr U
-  CvArr V
-  CvArr B
-  CvArr X
-  int flags
-EigenVV
-  CvArr mat
-  CvArr evects
-  CvArr evals
-  double eps
-  int lowindex 0
-  int highindex 0
-CalcCovarMatrix
-  cvarr_count vects /K
-  CvArr covMat
-  CvArr avg
-  int flags
-Mahalonobis
-  CvArr vec1
-  CvArr vec2
-  CvArr mat
-CalcPCA
-  CvArr data
-  CvArr avg
-  CvArr eigenvalues
-  CvArr eigenvectors
-  int flags
-ProjectPCA
-  CvArr data
-  CvArr avg
-  CvArr eigenvectors
-  CvArr result
-BackProjectPCA
-  CvArr proj
-  CvArr avg
-  CvArr eigenvects
-  CvArr result
-
-# Math Functions
-Round int
-  double value
-Floor int
-  double value
-Ceil int
-  double value
-Sqrt float
-  float value
-InvSqrt float
-  float value
-Cbrt float
-  float value
-FastArctan float
-  float y
-  float x
-IsNaN int
-  double value
-IsInf int
-  double value
-CartToPolar
-  CvArr x
-  CvArr y
-  CvArr magnitude
-  CvArr angle NULL
-  int angleInDegrees 0
-PolarToCart
-  CvArr magnitude
-  CvArr angle
-  CvArr x
-  CvArr y
-  int angleInDegrees 0
-Pow
-  CvArr src
-  CvArr dst
-  double power
-Exp
-  CvArr src
-  CvArr dst
-Log
-  CvArr src
-  CvArr dst
-SolveCubic
-  CvMat coeffs
-  CvMat roots
-SolvePoly
-  CvMat coeffs
-  CvMat roots
-  int maxiter 10
-  int fig 10
-  
-# Random Number Generation
-RNG CvRNG
-  int64 seed -1LL
-RandArr
-  CvRNG* rng
-  CvArr arr
-  int distType
-  CvScalar param1
-  CvScalar param2
-RandInt unsigned
-  CvRNG* rng
-RandReal double
-  CvRNG* rng
-
-# Discrete Transforms
-DFT
-  CvArr src
-  CvArr dst
-  int flags
-  int nonzeroRows 0
-GetOptimalDFTSize int
-  int size0
-MulSpectrums
-  CvArr src1
-  CvArr src2
-  CvArr dst
-  int flags
-DCT
-  CvArr src
-  CvArr dst
-  int flags
-
-# Sequences
-SeqRemove
-  CvSeq seq
-  int index
-ClearSeq
-  CvSeq seq
-CloneSeq
-  CvSeq seq
-  CvMemStorage storage
-SeqRemoveSlice
-  CvSeq seq
-  CvSlice slice
-SeqInvert
-  CvSeq seq
-
-# Miscellaneous Functions
-CheckArr int
-  CvArr arr
-  int flags 0
-  double min_val 0
-  double max_val 0
-KMeans2 double
-  CvArr samples
-  int nclusters
-  CvArr labels
-  CvTermCriteria termcrit
-  int attempts 1
-  int flags 0
-  CvArr centers NULL
-
-# Gradients, Edges, Corners and Features
-Sobel
-  CvArr src
-  CvArr dst
-  int xorder
-  int yorder
-  int apertureSize 3
-Laplace
-  CvArr src
-  CvArr dst
-  int apertureSize 3
-Canny
-  CvArr image
-  CvArr edges
-  double threshold1
-  double threshold2
-  int aperture_size 3
-PreCornerDetect
-  CvArr image
-  CvArr corners
-  int apertureSize 3
-CornerEigenValsAndVecs
-  CvArr image
-  CvArr eigenvv
-  int blockSize
-  int aperture_size 3
-CornerMinEigenVal
-  CvArr image
-  CvArr eigenval
-  int blockSize
-  int aperture_size 3
-CornerHarris
-  CvArr image
-  CvArr harris_dst
-  int blockSize
-  int aperture_size 3
-  double k 0.04
-FindCornerSubPix corners
-  CvArr image
-  CvPoint2D32fs corners
-  CvSize win
-  CvSize zero_zone
-  CvTermCriteria criteria
-GoodFeaturesToTrack cornerCount
-  CvArr image
-  CvArr eigImage
-  CvArr tempImage
-  cvpoint2d32f_count cornerCount
-  double qualityLevel
-  double minDistance
-  CvArr mask NULL
-  int blockSize 3
-  int useHarris 0
-  double k 0.04
-ExtractSURF keypoints,descriptors
-  CvArr image
-  CvArr mask
-  CvSeqOfCvSURFPoint* keypoints /O
-  CvSeqOfCvSURFDescriptor* descriptors /O
-  CvMemStorage storage
-  CvSURFParams params
-GetStarKeypoints CvSeqOfCvStarKeypoint*
-  CvArr image
-  CvMemStorage storage
-  CvStarDetectorParams params cvStarDetectorParams()
-
-# Sampling, Interpolation and Geometrical Transforms
-GetRectSubPix
-  CvArr src
-  CvArr dst
-  CvPoint2D32f center
-GetQuadrangleSubPix
-  CvArr src
-  CvArr dst
-  CvMat mapMatrix
-Resize
-  CvArr src
-  CvArr dst
-  int interpolation CV_INTER_LINEAR
-WarpAffine
-  CvArr src
-  CvArr dst
-  CvMat mapMatrix
-  int flags CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS
-  CvScalar fillval cvScalarAll(0)
-GetAffineTransform
-  CvPoint2D32f* src
-  CvPoint2D32f* dst
-  CvMat mapMatrix
-GetRotationMatrix2D
-  CvPoint2D32f center
-  double angle
-  double scale
-  CvMat mapMatrix
-WarpPerspective
-  CvArr src
-  CvArr dst
-  CvMat mapMatrix
-  int flags CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS
-  CvScalar fillval cvScalarAll(0)
-GetPerspectiveTransform
-  CvPoint2D32f* src
-  CvPoint2D32f* dst
-  CvMat mapMatrix
-Remap
-  CvArr src
-  CvArr dst
-  CvArr mapx
-  CvArr mapy
-  int flags CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS
-  CvScalar fillval cvScalarAll(0)
-ConvertMaps
-  CvArr mapx
-  CvArr mapy
-  CvArr mapxy
-  CvArr mapalpha
-LogPolar
-  CvArr src
-  CvArr dst
-  CvPoint2D32f center
-  double M
-  int flags CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS
-
-# Morphological Operations
-CreateStructuringElementEx IplConvKernel*
-  int cols
-  int rows
-  int anchorX
-  int anchorY
-  int shape
-  ints values {NULL,0}
-Erode
-  CvArr src
-  CvArr dst
-  IplConvKernel* element NULL
-  int iterations 1
-Dilate
-  CvArr src
-  CvArr dst
-  IplConvKernel* element NULL
-  int iterations 1
-MorphologyEx
-  CvArr src
-  CvArr dst
-  CvArr temp
-  IplConvKernel* element
-  int operation
-  int iterations 1
-
-# Filters and Color Conversion
-Smooth
-  CvArr src
-  CvArr dst
-  int smoothtype CV_GAUSSIAN
-  int param1 3
-  int param2 0
-  double param3 0
-  double param4 0
-Filter2D
-  CvArr src
-  CvArr dst
-  CvMat kernel
-  CvPoint anchor cvPoint(-1,-1)
-CopyMakeBorder
-  CvArr src
-  CvArr dst
-  CvPoint offset
-  int bordertype
-  CvScalar value cvScalarAll(0)
-Integral
-  CvArr image
-  CvArr sum
-  CvArr sqsum NULL
-  CvArr tiltedSum NULL
-CvtColor
-  CvArr src
-  CvArr dst
-  int code
-Threshold
-  CvArr src
-  CvArr dst
-  double threshold
-  double maxValue
-  int thresholdType
-AdaptiveThreshold
-  CvArr src
-  CvArr dst
-  double maxValue
-  int adaptive_method CV_ADAPTIVE_THRESH_MEAN_C /ch_adaptive_method
-  int thresholdType CV_THRESH_BINARY /ch_threshold_type
-  int blockSize 3
-  double param1 5
-
-# Pyramids and the Applications
-PyrDown
-  CvArr src
-  CvArr dst
-  int filter CV_GAUSSIAN_5x5
-PyrUp
-  CvArr src
-  CvArr dst
-  int filter CV_GAUSSIAN_5x5
-PyrSegmentation comp
-  IplImage src
-  IplImage dst
-  CvMemStorage storage
-  CvSeq* comp /O
-  int level
-  double threshold1
-  double threshold2
-PyrMeanShiftFiltering
-  CvArr src
-  CvArr dst
-  double sp
-  double sr
-  int max_level 1
-  CvTermCriteria termcrit cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,5,1)
-
-# Image Segmentation, Connected Components and Contour Retrieval
-FloodFill comp
-  CvArr image
-  CvPoint seed_point
-  CvScalar new_val
-  CvScalar lo_diff cvScalarAll(0)
-  CvScalar up_diff cvScalarAll(0)
-  CvConnectedComp comp /O
-  int flags 4
-  CvArr mask NULL
-Watershed
-  CvArr image
-  CvArr markers
-
-# Image and Contour Moments
-Moments moments
-  cvarrseq arr
-  CvMoments moments /O
-  int binary 0
-GetSpatialMoment double
-  CvMoments* moments
-  int x_order
-  int y_order
-GetCentralMoment double
-  CvMoments* moments
-  int x_order
-  int y_order
-GetNormalizedCentralMoment double
-  CvMoments* moments
-  int x_order
-  int y_order
-
-# Special Image Transforms
-HoughLines2 CvSeq*
-  CvArr image
-  CvMemStorage storage
-  int method
-  double rho
-  double theta
-  int threshold
-  double param1 0
-  double param2 0
-HoughCircles
-  CvArr image
-  CvMat circle_storage
-  int method
-  double dp
-  double min_dist
-  double param1 100
-  double param2 100
-  int min_radius 0
-  int max_radius 0
-DistTransform
-  CvArr src
-  CvArr dst
-  int distance_type CV_DIST_L2
-  int mask_size 3
-  floats mask {NULL,0}
-  CvArr labels NULL
-Inpaint
-  CvArr src
-  CvArr mask
-  CvArr dst
-  double inpaintRadius
-  int flags
-
-# Histograms
-ClearHist
-  CvHistogram hist
-CalcArrHist
-  CvArrs image
-  CvHistogram hist
-  int accumulate 0
-  CvArr mask NULL
-CalcHist
-  IplImages image
-  CvHistogram hist
-  int accumulate 0
-  CvArr mask NULL
-NormalizeHist
-  CvHistogram hist
-  double factor
-ThreshHist
-  CvHistogram hist
-  double threshold
-CompareHist double
-  CvHistogram hist1
-  CvHistogram hist2
-  int method
-# CopyHist
-CalcBackProject
-  IplImages image
-  CvArr back_project
-  CvHistogram hist
-CalcArrBackProject
-  CvArrs image
-  CvArr back_project
-  CvHistogram hist
-CalcBackProjectPatch
-  IplImages images
-  CvArr dst
-  CvSize patch_size
-  CvHistogram hist
-  int method
-  float factor
-CalcProbDensity
-  CvHistogram hist1
-  CvHistogram hist2
-  CvHistogram dst_hist
-  double scale 255
-EqualizeHist
-  CvArr src
-  CvArr dst
-QueryHistValue_1D double
-  CvHistogram hist
-  int idx0
-QueryHistValue_2D double
-  CvHistogram hist
-  int idx0
-  int idx1
-QueryHistValue_3D double
-  CvHistogram hist
-  int idx0
-  int idx1
-  int idx2
-QueryHistValue_nD double
-  CvHistogram hist
-  ints idx
-
-# Matching
-MatchTemplate
-  CvArr image
-  CvArr templ
-  CvArr result
-  int method
-MatchShapes double
-  CvSeq object1
-  CvSeq object2
-  int method
-  double parameter 0
-
-# Contour Processing Functions
-ApproxChains CvSeq*
-  CvSeq src_seq
-  CvMemStorage storage
-  int method CV_CHAIN_APPROX_SIMPLE
-  double parameter 0
-  int minimal_perimeter 0
-  int recursive 0
-BoundingRect CvRect
-  cvarrseq points
-  int update 0
-ContourArea double
-  cvarrseq contour
-  CvSlice slice CV_WHOLE_SEQ
-ArcLength double
-  cvarrseq curve
-  CvSlice slice CV_WHOLE_SEQ
-  int isClosed -1
-
-# Computational Geometry
-MaxRect CvRect
-  CvRect* rect1
-  CvRect* rect2
-# TODO PointSeqFromMat
-BoxPoints points
-  CvBox2D box
-  CvPoint2D32f_4 points /O,A
-FitEllipse2 CvBox2D
-  CvArr points
-ConvexHull2 CvSeq*
-  cvarrseq points
-  CvMemStorage storage
-  int orientation CV_CLOCKWISE
-  int return_points 0
-CheckContourConvexity int
-  cvarrseq contour
-ConvexityDefects CvSeqOfCvConvexityDefect*
-  cvarrseq contour
-  CvSeq convexhull
-  CvMemStorage storage
-PointPolygonTest double
-  cvarrseq contour
-  CvPoint2D32f pt
-  int measure_dist
-MinAreaRect2 CvBox2D
-  cvarrseq points
-  CvMemStorage storage NULL
-MinEnclosingCircle int,center,radius
-  cvarrseq points
-  CvPoint2D32f center /O
-  float radius /O
-
-# Planar Subdivisions
-
-Subdiv2DGetEdge CvSubdiv2DEdge
-  CvSubdiv2DEdge edge
-  CvNextEdgeType type
-Subdiv2DNextEdge CvSubdiv2DEdge
-  CvSubdiv2DEdge edge
-Subdiv2DRotateEdge CvSubdiv2DEdge
-  CvSubdiv2DEdge edge
-  int rotate
-Subdiv2DEdgeOrg CvSubdiv2DPoint*
-  CvSubdiv2DEdge edge
-Subdiv2DEdgeDst CvSubdiv2DPoint*
-  CvSubdiv2DEdge edge
-CreateSubdivDelaunay2D CvSubdiv2D*
-  CvRect rect
-  CvMemStorage storage
-SubdivDelaunay2DInsert CvSubdiv2DPoint*
-  CvSubdiv2D* subdiv
-  CvPoint2D32f pt
-CalcSubdivVoronoi2D
-  CvSubdiv2D* subdiv
-ClearSubdivVoronoi2D
-  CvSubdiv2D* subdiv
-FindNearestPoint2D CvSubdiv2DPoint*
-  CvSubdiv2D* subdiv
-  CvPoint2D32f pt
-
-# Object Detection
-HaarDetectObjects CvSeqOfCvAvgComp*
-  CvArr image
-  CvHaarClassifierCascade* cascade
-  CvMemStorage storage
-  double scale_factor 1.1   /ch_doubleAbove1
-  int min_neighbors 3
-  int flags 0
-  CvSize min_size cvSize(0,0)
-
-ComputeCorrespondEpilines
-  CvMat points
-  int whichImage
-  CvMat F
-  CvMat lines
-ConvertPointsHomogeneous
-  CvMat src
-  CvMat dst
-ProjectPoints2
-  CvMat objectPoints
-  CvMat rvec
-  CvMat tvec
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvMat imagePoints
-  CvMat dpdrot NULL
-  CvMat dpdt NULL
-  CvMat dpdf NULL
-  CvMat dpdc NULL
-  CvMat dpddist NULL
-ReprojectImageTo3D
-  CvArr disparity
-  CvArr _3dImage
-  CvMat Q
-  int handleMissingValues 0
-RQDecomp3x3 eulerAngles
-  CvMat M
-  CvMat R
-  CvMat Q
-  CvMat Qx NULL
-  CvMat Qy NULL
-  CvMat Qz NULL
-  CvPoint3D64f eulerAngles /O
-FindHomography
-  CvMat srcPoints
-  CvMat dstPoints
-  CvMat H
-  int method 0
-  double ransacReprojThreshold 3.0
-  CvMat status NULL
-CreateStereoBMState CvStereoBMState*
-  int preset CV_STEREO_BM_BASIC
-  int numberOfDisparities 0
-CreateStereoGCState CvStereoGCState*
-  int numberOfDisparities
-  int maxIters
-FindStereoCorrespondenceBM
-  CvArr left
-  CvArr right
-  CvArr disparity
-  CvStereoBMState* state
-FindStereoCorrespondenceGC
-  CvArr left
-  CvArr right
-  CvArr dispLeft
-  CvArr dispRight
-  CvStereoGCState* state
-  int useDisparityGuess 0
-CalibrateCamera2 
-  CvMat objectPoints
-  CvMat imagePoints
-  CvMat pointCounts
-  CvSize imageSize
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvMat rvecs
-  CvMat tvecs
-  int flags 0
-CalibrationMatrixValues fovx,fovy,focalLength,principalPoint,pixelAspectRatio
-  CvMat calibMatr
-  CvSize image_size
-  double apertureWidth 0
-  double apertureHeight 0
-  double fovx /O
-  double fovy /O
-  double focalLength /O
-  CvPoint2D64f principalPoint /O
-  double pixelAspectRatio /O
-FindExtrinsicCameraParams2
-  CvMat objectPoints
-  CvMat imagePoints
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvMat rvec
-  CvMat tvec
-  int useExtrinsicGuess 0
-FindFundamentalMat int
-  CvMat points1
-  CvMat points2
-  CvMat fundamentalMatrix
-  int method CV_FM_RANSAC
-  double param1 1.
-  double param2 0.99
-  CvMat status NULL
-StereoCalibrate
-  CvMat objectPoints
-  CvMat imagePoints1
-  CvMat imagePoints2
-  CvMat pointCounts
-  CvMat cameraMatrix1
-  CvMat distCoeffs1
-  CvMat cameraMatrix2
-  CvMat distCoeffs2
-  CvSize imageSize
-  CvMat R
-  CvMat T
-  CvMat E NULL
-  CvMat F NULL
-  CvTermCriteria term_crit cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,30,1e-6)
-  int flags CV_CALIB_FIX_INTRINSIC
-GetOptimalNewCameraMatrix
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvSize imageSize
-  double alpha
-  CvMat newCameraMatrix
-  CvSize newImageSize cvSize(0,0)
-  CvRect* validPixROI NULL
-  int centerPrincipalPoint 0
-InitIntrinsicParams2D
-  CvMat objectPoints
-  CvMat imagePoints
-  CvMat npoints
-  CvSize imageSize
-  CvMat cameraMatrix
-  double aspectRatio 1.
-StereoRectify roi1,roi2
-  CvMat cameraMatrix1
-  CvMat cameraMatrix2
-  CvMat distCoeffs1
-  CvMat distCoeffs2
-  CvSize imageSize
-  CvMat R
-  CvMat T
-  CvMat R1
-  CvMat R2
-  CvMat P1
-  CvMat P2
-  CvMat Q NULL
-  int flags CV_CALIB_ZERO_DISPARITY
-  double alpha -1
-  CvSize newImageSize cvSize(0,0)
-  CvRect roi1 /O
-  CvRect roi2 /O
-StereoRectifyUncalibrated
-  CvMat points1
-  CvMat points2
-  CvMat F
-  CvSize imageSize
-  CvMat H1
-  CvMat H2
-  double threshold 5
-Rodrigues2
-  CvMat src
-  CvMat dst
-  CvMat jacobian 0
-Undistort2
-  CvArr src
-  CvArr dst
-  CvMat cameraMatrix
-  CvMat distCoeffs
-InitUndistortMap
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvArr map1
-  CvArr map2
-InitUndistortRectifyMap
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvMat R
-  CvMat newCameraMatrix
-  CvArr map1
-  CvArr map2
-UndistortPoints
-  CvMat src
-  CvMat dst
-  CvMat cameraMatrix
-  CvMat distCoeffs
-  CvMat R NULL
-  CvMat P NULL
-DecomposeProjectionMatrix eulerAngles
-  CvMat projMatrix
-  CvMat cameraMatrix
-  CvMat rotMatrix
-  CvMat transVect
-  CvMat rotMatrX NULL
-  CvMat rotMatrY NULL
-  CvMat rotMatrZ NULL
-  CvPoint3D64f eulerAngles /O
-DrawChessboardCorners
-  CvArr image
-  CvSize patternSize
-  CvPoint2D32fs corners
-  int patternWasFound
-
-CreatePOSITObject CvPOSITObject*
-  CvPoint3D32fs points
-POSIT rotationMatrix,translation_vector
-  CvPOSITObject* posit_object
-  CvPoint2D32f* imagePoints
-  double focal_length
-  CvTermCriteria criteria
-  CvMatr32f_i rotationMatrix /O,A
-  CvVect32f_i translation_vector /O,A
-
-EstimateRigidTransform
-  CvArr A
-  CvArr B
-  CvMat M
-  int full_affine
-
-# Accumulation of Background Statistics
-Acc
-  CvArr image
-  CvArr sum
-  CvArr mask NULL
-SquareAcc
-  CvArr image
-  CvArr sqsum
-  CvArr mask NULL
-MultiplyAcc
-  CvArr image1
-  CvArr image2
-  CvArr acc
-  CvArr mask NULL
-RunningAvg
-  CvArr image
-  CvArr acc
-  double alpha
-  CvArr mask NULL
-
-# Motion Templates
-UpdateMotionHistory
-  CvArr silhouette
-  CvArr mhi
-  double timestamp
-  double duration
-CalcMotionGradient
-  CvArr mhi          /ch_matF
-  CvArr mask
-  CvArr orientation  /ch_matF
-  double delta1
-  double delta2
-  int apertureSize 3 /ch_aperture
-CalcGlobalOrientation double
-  CvArr orientation
-  CvArr mask
-  CvArr mhi
-  double timestamp
-  double duration
-SegmentMotion CvSeq*
-  CvArr mhi
-  CvArr seg_mask
-  CvMemStorage storage
-  double timestamp
-  double seg_thresh
-
-# Object Tracking
-MeanShift comp
-  CvArr prob_image
-  CvRect window
-  CvTermCriteria criteria
-  CvConnectedComp comp /O
-CamShift int,comp,box
-  CvArr prob_image
-  CvRect window
-  CvTermCriteria criteria
-  CvConnectedComp comp /O
-  CvBox2D box /O
-CreateKalman CvKalman*
-  int dynam_params
-  int measure_params
-  int control_params 0
-KalmanCorrect ROCvMat*
-  CvKalman* kalman
-  CvMat measurement
-KalmanPredict ROCvMat*
-  CvKalman* kalman
-  CvMat control NULL
-SnakeImage points
-  IplImage image
-  CvPoints points
-  floats alpha
-  floats beta
-  floats gamma
-  CvSize win
-  CvTermCriteria criteria
-  int calc_gradient 1
-
-# Optical Flow
-CalcOpticalFlowLK
-  CvArr prev
-  CvArr curr
-  CvSize winSize
-  CvArr velx
-  CvArr vely
-CalcOpticalFlowBM
-  CvArr prev    /ch_image8
-  CvArr curr    /ch_image8
-  CvSize blockSize
-  CvSize shiftSize
-  CvSize max_range
-  int usePrevious
-  CvArr velx        /ch_vel
-  CvArr vely        /ch_vel
-CalcOpticalFlowHS
-  CvArr prev    /ch_image8
-  CvArr curr    /ch_image8
-  int usePrevious
-  CvArr velx        /ch_vel_64
-  CvArr vely        /ch_vel_64
-  double lambda
-  CvTermCriteria criteria
-CalcOpticalFlowFarneback
-  CvArr prev /ch_image8
-  CvArr curr /ch_image8
-  CvArr flow
-  double pyr_scale 0.5
-  int levels 3
-  int winsize 15
-  int iterations 3
-  int poly_n 7
-  double poly_sigma 1.5
-  int flags 0
-
-# Highgui
-ConvertImage
-  CvArr src
-  CvArr dst
-  int flags 0
-NamedWindow
-  char* name
-  int flags CV_WINDOW_AUTOSIZE
-DestroyWindow
-  char* name
-DestroyAllWindows
-ResizeWindow
-  char* name
-  int width
-  int height
-MoveWindow
-  char* name
-  int x
-  int y
-ShowImage
-  char* name
-  CvArr image
-GetTrackbarPos int
-  char* trackbarName
-  char* windowName
-SetTrackbarPos
-  char* trackbarName
-  char* windowName
-  int pos
-#WaitKey int
-#  int delay 0
-SaveImage
-  char* filename
-  CvArr image
-CaptureFromFile CvCapture*
-  char* filename
-CreateFileCapture CvCapture*
-  char* filename
-CaptureFromCAM CvCapture*
-  int index
-CreateCameraCapture CvCapture*
-  int index
-GrabFrame int
-  CvCapture* capture
-RetrieveFrame ROIplImage*
-  CvCapture* capture
-QueryFrame ROIplImage*
-  CvCapture* capture
-GetCaptureProperty double
-  CvCapture* capture
-  int property_id
-SetCaptureProperty int
-  CvCapture* capture
-  int property_id
-  double value
-CreateVideoWriter CvVideoWriter*
-  char* filename
-  int fourcc
-  double fps
-  CvSize frame_size
-  int is_color 1
-WriteFrame int
-  CvVideoWriter* writer
-  IplImage image
-EncodeImage CvMat*
-  char* ext
-  CvArr image
-  ints0 params {&zero,1}
-DecodeImage IplImage*
-  CvMat buf
-  int iscolor CV_LOAD_IMAGE_COLOR
-DecodeImageM CvMat*
-  CvMat buf
-  int iscolor CV_LOAD_IMAGE_COLOR
-StartWindowThread
-SetWindowProperty
-  char* name
-  int prop_id
-  double prop_value
-GetWindowProperty double
-  char* name
-  int prop_id
-
-GetTickCount int64
-GetTickFrequency int64
-
-# cvaux stuff
-HOGDetectMultiScale CvSeq*
-  CvArr image
-  CvMemStorage storage
-  CvArr svm_classifier NULL
-  CvSize win_stride cvSize(0,0)
-  double hit_threshold 0
-  double scale 1.05
-  int group_threshold 2
-  CvSize padding cvSize(0,0)
-  CvSize win_size cvSize(64,128)
-  CvSize block_size cvSize(16,16)
-  CvSize block_stride cvSize(8,8)
-  CvSize cell_size cvSize(8,8)
-  int nbins 9
-  int gammaCorrection 1
-
-GrabCut
-  CvArr image
-  CvArr mask
-  CvRect rect
-  CvArr bgdModel
-  CvArr fgdModel
-  int iterCount
-  int mode
-
-# These functions are handwritten in cv.cpp; they appear here as 'doconly' declarations
-# so that their documentation can be auto-generated
-ApproxPoly  /doconly
-  cvarrseq src_seq
-  CvMemStorage storage
-  int method
-  double parameter 0.0
-  int parameter2 0
-CalcEMD2 float /doconly
-  CvArr signature1
-  CvArr signature2
-  int distance_type
-  PyCallableObject* distance_func NULL
-  CvArr cost_matrix NULL
-  CvArr flow NULL
-  float lower_bound 0.0
-  PyObject* userdata NULL
-CalcOpticalFlowPyrLK currFeatures,status,track_error /doconly
-  CvArr prev
-  CvArr curr
-  CvArr prevPyr
-  CvArr currPyr
-  CvPoint2D32f* prevFeatures
-  CvSize winSize
-  int level
-  CvTermCriteria criteria
-  int flags
-  CvPoint2D32f* guesses NULL
-  CvPoint2D32f currFeatures /O
-  char status /O
-  float track_error /O
-ClipLine point1,point2 /doconly
-  CvSize imgSize
-  CvPoint pt1
-  CvPoint pt2
-CreateData /doconly
-  CvArr arr
-CreateHist CvHistogram /doconly
-  ints dims
-  int type
-  ranges ranges None
-  int uniform 1
-CreateImageHeader IplImage* /doconly
-  CvSize size
-  int depth
-  int channels
-CreateImage IplImage* /doconly
-  CvSize size
-  int depth
-  int channels
-CreateMatHeader CvMat /doconly
-  int rows
-  int cols
-  int type
-CreateMat CvMat /doconly
-  int rows
-  int cols
-  int type
-CreateMatNDHeader CvMatND /doconly
-  ints dims
-  int type
-CreateMatND CvMatND /doconly
-  ints dims
-  int type
-CreateMemStorage CvMemStorage /doconly
-  int blockSize 0
-CreateTrackbar /doconly
-  char* trackbarName
-  char* windowName
-  int value
-  int count
-  PyCallableObject* onChange
-FindChessboardCorners corners /doconly
-  CvArr image
-  CvSize patternSize
-  CvPoint2D32fs corners /O
-  int flags CV_CALIB_CB_ADAPTIVE_THRESH
-FindContours CvSeq /doconly
-  CvArr image
-  CvMemStorage storage
-  int mode CV_RETR_LIST
-  int method CV_CHAIN_APPROX_SIMPLE
-  CvPoint offset (0,0)
-FitLine line /doconly
-  CvArr points
-  int dist_type
-  double param
-  double reps
-  double aeps
-  PyObject* line /O
-GetDims dim1,dim2,... /doconly
-  CvArr arr
-GetHuMoments hu /doconly
-  CvMoments moments
-  PyObject* hu /O
-GetImage iplimage /doconly
-  CvMat arr
-GetMat CvMat /doconly
-  IplImage arr
-  int allowND 0
-GetMinMaxHistValue min_value,max_value,min_idx,max_idx /doconly
-  CvHistogram hist
-  CvScalar min_value /O
-  CvScalar max_value /O
-  ints min_idx /O
-  ints max_idx /O
-InitLineIterator line_iterator /doconly
-  CvArr image
-  CvPoint pt1
-  CvPoint pt2
-  iter line_iterator /O
-  int connectivity 8
-  int left_to_right 0
-LoadImageM /doconly
-  char* filename
-  int iscolor CV_LOAD_IMAGE_COLOR
-LoadImage /doconly
-  char* filename
-  int iscolor CV_LOAD_IMAGE_COLOR
-ReshapeMatND CvMat /doconly
-  CvMat arr
-  int newCn
-  ints newDims
-Reshape CvMat /doconly
-  CvArr arr
-  int newCn
-  int newRows 0
-SetData /doconly
-  CvArr arr
-  PyObject* data
-  int step
-SetMouseCallback /doconly
-  char* windowName
-  PyCallableObject* onMouse
-  PyObject* param None
-Subdiv2DLocate loc,where /doconly
-  CvSubdiv2D* subdiv
-  CvPoint2D32f pt
-  int loc /O
-  edgeorpoint where /O
-WaitKey int /doconly
-  int delay 0
diff --git a/modules/python/src2/cv.py b/modules/python/src2/cv.py
deleted file mode 100755
index 4238d05f7a..0000000000
--- a/modules/python/src2/cv.py
+++ /dev/null
@@ -1,3 +0,0 @@
-#/usr/bin/env python
-
-from cv2.cv import *
diff --git a/modules/python/src2/cv2.cpp b/modules/python/src2/cv2.cpp
index bbdd5abbe7..b2bd368a18 100644
--- a/modules/python/src2/cv2.cpp
+++ b/modules/python/src2/cv2.cpp
@@ -1,33 +1,31 @@
 #include <Python.h>
 
-#if !PYTHON_USE_NUMPY
-#error "The module can only be built if NumPy is available"
-#endif
-
 #define MODULESTR "cv2"
-
-#include "numpy/ndarrayobject.h"
+#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
+#include <numpy/ndarrayobject.h>
 
 #include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/contrib.hpp"
 #include "opencv2/flann/miniflann.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/calib3d.hpp"
-#include "opencv2/ml.hpp"
 #include "opencv2/features2d.hpp"
 #include "opencv2/objdetect.hpp"
 #include "opencv2/softcascade.hpp"
-#include "opencv2/video/tracking.hpp"
-#include "opencv2/video/background_segm.hpp"
+#include "opencv2/video.hpp"
 #include "opencv2/photo.hpp"
 #include "opencv2/highgui.hpp"
 
-#include "opencv2/opencv_modules.hpp"
+#include "opencv2/ml.hpp"
 
+#include "opencv2/opencv_modules.hpp"
 #ifdef HAVE_OPENCV_NONFREE
 #  include "opencv2/nonfree.hpp"
 #endif
 
+#include "pycompat.hpp"
+
 using cv::flann::IndexParams;
 using cv::flann::SearchParams;
 
@@ -100,6 +98,7 @@ using namespace cv;
 typedef cv::softcascade::ChannelFeatureBuilder softcascade_ChannelFeatureBuilder;
 
 typedef std::vector<uchar> vector_uchar;
+typedef std::vector<char> vector_char;
 typedef std::vector<int> vector_int;
 typedef std::vector<float> vector_float;
 typedef std::vector<double> vector_double;
@@ -114,7 +113,9 @@ typedef std::vector<Rect> vector_Rect;
 typedef std::vector<KeyPoint> vector_KeyPoint;
 typedef std::vector<Mat> vector_Mat;
 typedef std::vector<DMatch> vector_DMatch;
-typedef std::vector<std::string> vector_string;
+typedef std::vector<String> vector_String;
+
+typedef std::vector<std::vector<char> > vector_vector_char;
 typedef std::vector<std::vector<Point> > vector_vector_Point;
 typedef std::vector<std::vector<Point2f> > vector_vector_Point2f;
 typedef std::vector<std::vector<Point3f> > vector_vector_Point3f;
@@ -125,8 +126,17 @@ typedef Ptr<FeatureDetector> Ptr_FeatureDetector;
 typedef Ptr<DescriptorExtractor> Ptr_DescriptorExtractor;
 typedef Ptr<Feature2D> Ptr_Feature2D;
 typedef Ptr<DescriptorMatcher> Ptr_DescriptorMatcher;
+typedef Ptr<BackgroundSubtractor> Ptr_BackgroundSubtractor;
+typedef Ptr<BackgroundSubtractorMOG> Ptr_BackgroundSubtractorMOG;
+typedef Ptr<BackgroundSubtractorMOG2> Ptr_BackgroundSubtractorMOG2;
+typedef Ptr<BackgroundSubtractorGMG> Ptr_BackgroundSubtractorGMG;
+
+typedef Ptr<StereoMatcher> Ptr_StereoMatcher;
+typedef Ptr<StereoBM> Ptr_StereoBM;
+typedef Ptr<StereoSGBM> Ptr_StereoSGBM;
 
 typedef Ptr<cv::softcascade::ChannelFeatureBuilder> Ptr_ChannelFeatureBuilder;
+typedef Ptr<CLAHE> Ptr_CLAHE;
 
 typedef SimpleBlobDetector::Params SimpleBlobDetector_Params;
 
@@ -183,24 +193,19 @@ public:
                       depth == CV_32S ? NPY_INT : depth == CV_32F ? NPY_FLOAT :
                       depth == CV_64F ? NPY_DOUBLE : f*NPY_ULONGLONG + (f^1)*NPY_UINT;
         int i;
-        npy_intp _sizes[CV_MAX_DIM+1];
+        cv::AutoBuffer<npy_intp> _sizes(dims + 1);
         for( i = 0; i < dims; i++ )
             _sizes[i] = sizes[i];
         if( cn > 1 )
-        {
-            /*if( _sizes[dims-1] == 1 )
-                _sizes[dims-1] = cn;
-            else*/
-                _sizes[dims++] = cn;
-        }
+            _sizes[dims++] = cn;
         PyObject* o = PyArray_SimpleNew(dims, _sizes, typenum);
         if(!o)
-            CV_Error_(CV_StsError, ("The numpy array of typenum=%d, ndims=%d can not be created", typenum, dims));
+            CV_Error_(Error::StsError, ("The numpy array of typenum=%d, ndims=%d can not be created", typenum, dims));
         refcount = refcountFromPyObject(o);
-        npy_intp* _strides = PyArray_STRIDES(o);
+        npy_intp* _strides = PyArray_STRIDES((PyArrayObject*) o);
         for( i = 0; i < dims - (cn > 1); i++ )
             step[i] = (size_t)_strides[i];
-        datastart = data = (uchar*)PyArray_DATA(o);
+        datastart = data = (uchar*)PyArray_DATA((PyArrayObject*) o);
     }
 
     void deallocate(int* refcount, uchar*, uchar*)
@@ -216,11 +221,19 @@ public:
 
 NumpyAllocator g_numpyAllocator;
 
+
+template<typename T> static
+bool pyopencv_to(PyObject* obj, T& p, const char* name = "<unknown>");
+
+template<typename T> static
+PyObject* pyopencv_from(const T& src);
+
 enum { ARG_NONE = 0, ARG_MAT = 1, ARG_SCALAR = 2 };
 
 // special case, when the convertor needs full ArgInfo structure
-static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allowND=true)
+static bool pyopencv_to(PyObject* o, Mat& m, const ArgInfo info)
 {
+    bool allowND = true;
     if(!o || o == Py_None)
     {
         if( !m.data )
@@ -267,8 +280,10 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
         return false;
     }
 
+    PyArrayObject* oarr = (PyArrayObject*) o;
+
     bool needcopy = false, needcast = false;
-    int typenum = PyArray_TYPE(o), new_typenum = typenum;
+    int typenum = PyArray_TYPE(oarr), new_typenum = typenum;
     int type = typenum == NPY_UBYTE ? CV_8U :
                typenum == NPY_BYTE ? CV_8S :
                typenum == NPY_USHORT ? CV_16U :
@@ -293,7 +308,11 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
         }
     }
 
-    int ndims = PyArray_NDIM(o);
+#ifndef CV_MAX_DIM
+    const int CV_MAX_DIM = 32;
+#endif
+
+    int ndims = PyArray_NDIM(oarr);
     if(ndims >= CV_MAX_DIM)
     {
         failmsg("%s dimensionality (=%d) is too high", info.name, ndims);
@@ -301,9 +320,10 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
     }
 
     int size[CV_MAX_DIM+1];
-    size_t step[CV_MAX_DIM+1], elemsize = CV_ELEM_SIZE1(type);
-    const npy_intp* _sizes = PyArray_DIMS(o);
-    const npy_intp* _strides = PyArray_STRIDES(o);
+    size_t step[CV_MAX_DIM+1];
+    size_t elemsize = CV_ELEM_SIZE1(type);
+    const npy_intp* _sizes = PyArray_DIMS(oarr);
+    const npy_intp* _strides = PyArray_STRIDES(oarr);
     bool ismultichannel = ndims == 3 && _sizes[2] <= CV_CN_MAX;
 
     for( int i = ndims-1; i >= 0 && !needcopy; i-- )
@@ -327,11 +347,17 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
             failmsg("Layout of the output array %s is incompatible with cv::Mat (step[ndims-1] != elemsize or step[1] != elemsize*nchannels)", info.name);
             return false;
         }
-        if( needcast )
-            o = (PyObject*)PyArray_Cast((PyArrayObject*)o, new_typenum);
-        else
-            o = (PyObject*)PyArray_GETCONTIGUOUS((PyArrayObject*)o);
-        _strides = PyArray_STRIDES(o);
+
+        if( needcast ) {
+            o = PyArray_Cast(oarr, new_typenum);
+            oarr = (PyArrayObject*) o;
+        }
+        else {
+            oarr = PyArray_GETCONTIGUOUS(oarr);
+            o = (PyObject*) oarr;
+        }
+
+        _strides = PyArray_STRIDES(oarr);
     }
 
     for(int i = 0; i < ndims; i++)
@@ -359,7 +385,7 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
         return false;
     }
 
-    m = Mat(ndims, size, type, PyArray_DATA(o), step);
+    m = Mat(ndims, size, type, PyArray_DATA(oarr), step);
 
     if( m.data )
     {
@@ -375,7 +401,8 @@ static int pyopencv_to(const PyObject* o, Mat& m, const ArgInfo info, bool allow
     return true;
 }
 
-static PyObject* pyopencv_from(const Mat& m)
+template<>
+PyObject* pyopencv_from(const Mat& m)
 {
     if( !m.data )
         Py_RETURN_NONE;
@@ -383,14 +410,15 @@ static PyObject* pyopencv_from(const Mat& m)
     if(!p->refcount || p->allocator != &g_numpyAllocator)
     {
         temp.allocator = &g_numpyAllocator;
-        m.copyTo(temp);
+        ERRWRAP2(m.copyTo(temp));
         p = &temp;
     }
     p->addref();
     return pyObjectFromRefcount(p->refcount);
 }
 
-static bool pyopencv_to(PyObject *o, Scalar& s, const char *name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject *o, Scalar& s, const char *name)
 {
     if(!o || o == Py_None)
         return true;
@@ -424,17 +452,20 @@ static bool pyopencv_to(PyObject *o, Scalar& s, const char *name = "<unknown>")
     return true;
 }
 
-static inline PyObject* pyopencv_from(const Scalar& src)
+template<>
+PyObject* pyopencv_from(const Scalar& src)
 {
     return Py_BuildValue("(dddd)", src[0], src[1], src[2], src[3]);
 }
 
-static PyObject* pyopencv_from(bool value)
+template<>
+PyObject* pyopencv_from(const bool& value)
 {
     return PyBool_FromLong(value);
 }
 
-static bool pyopencv_to(PyObject* obj, bool& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, bool& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -446,12 +477,14 @@ static bool pyopencv_to(PyObject* obj, bool& value, const char* name = "<unknown
     return true;
 }
 
-static PyObject* pyopencv_from(size_t value)
+template<>
+PyObject* pyopencv_from(const size_t& value)
 {
     return PyLong_FromSize_t(value);
 }
 
-static bool pyopencv_to(PyObject* obj, size_t& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, size_t& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -460,22 +493,26 @@ static bool pyopencv_to(PyObject* obj, size_t& value, const char* name = "<unkno
     return value != (size_t)-1 || !PyErr_Occurred();
 }
 
-static PyObject* pyopencv_from(int value)
+template<>
+PyObject* pyopencv_from(const int& value)
 {
     return PyInt_FromLong(value);
 }
 
-static PyObject* pyopencv_from(cvflann_flann_algorithm_t value)
+template<>
+PyObject* pyopencv_from(const cvflann_flann_algorithm_t& value)
 {
     return PyInt_FromLong(int(value));
 }
 
-static PyObject* pyopencv_from(cvflann_flann_distance_t value)
+template<>
+PyObject* pyopencv_from(const cvflann_flann_distance_t& value)
 {
     return PyInt_FromLong(int(value));
 }
 
-static bool pyopencv_to(PyObject* obj, int& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, int& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -489,12 +526,14 @@ static bool pyopencv_to(PyObject* obj, int& value, const char* name = "<unknown>
     return value != -1 || !PyErr_Occurred();
 }
 
-static PyObject* pyopencv_from(uchar value)
+template<>
+PyObject* pyopencv_from(const uchar& value)
 {
     return PyInt_FromLong(value);
 }
 
-static bool pyopencv_to(PyObject* obj, uchar& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, uchar& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -504,12 +543,14 @@ static bool pyopencv_to(PyObject* obj, uchar& value, const char* name = "<unknow
     return ivalue != -1 || !PyErr_Occurred();
 }
 
-static PyObject* pyopencv_from(double value)
+template<>
+PyObject* pyopencv_from(const double& value)
 {
     return PyFloat_FromDouble(value);
 }
 
-static bool pyopencv_to(PyObject* obj, double& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, double& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -521,12 +562,14 @@ static bool pyopencv_to(PyObject* obj, double& value, const char* name = "<unkno
     return !PyErr_Occurred();
 }
 
-static PyObject* pyopencv_from(float value)
+template<>
+PyObject* pyopencv_from(const float& value)
 {
     return PyFloat_FromDouble(value);
 }
 
-static bool pyopencv_to(PyObject* obj, float& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, float& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -538,17 +581,20 @@ static bool pyopencv_to(PyObject* obj, float& value, const char* name = "<unknow
     return !PyErr_Occurred();
 }
 
-static PyObject* pyopencv_from(int64 value)
+template<>
+PyObject* pyopencv_from(const int64& value)
 {
     return PyLong_FromLongLong(value);
 }
 
-static PyObject* pyopencv_from(const std::string& value)
+template<>
+PyObject* pyopencv_from(const String& value)
 {
     return PyString_FromString(value.empty() ? "" : value.c_str());
 }
 
-static bool pyopencv_to(PyObject* obj, std::string& value, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, String& value, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -556,11 +602,12 @@ static bool pyopencv_to(PyObject* obj, std::string& value, const char* name = "<
     char* str = PyString_AsString(obj);
     if(!str)
         return false;
-    value = std::string(str);
+    value = String(str);
     return true;
 }
 
-static inline bool pyopencv_to(PyObject* obj, Size& sz, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Size& sz, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -568,12 +615,14 @@ static inline bool pyopencv_to(PyObject* obj, Size& sz, const char* name = "<unk
     return PyArg_ParseTuple(obj, "ii", &sz.width, &sz.height) > 0;
 }
 
-static inline PyObject* pyopencv_from(const Size& sz)
+template<>
+PyObject* pyopencv_from(const Size& sz)
 {
     return Py_BuildValue("(ii)", sz.width, sz.height);
 }
 
-static inline bool pyopencv_to(PyObject* obj, Rect& r, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Rect& r, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -581,12 +630,14 @@ static inline bool pyopencv_to(PyObject* obj, Rect& r, const char* name = "<unkn
     return PyArg_ParseTuple(obj, "iiii", &r.x, &r.y, &r.width, &r.height) > 0;
 }
 
-static inline PyObject* pyopencv_from(const Rect& r)
+template<>
+PyObject* pyopencv_from(const Rect& r)
 {
     return Py_BuildValue("(iiii)", r.x, r.y, r.width, r.height);
 }
 
-static inline bool pyopencv_to(PyObject* obj, Range& r, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Range& r, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -599,30 +650,14 @@ static inline bool pyopencv_to(PyObject* obj, Range& r, const char* name = "<unk
     return PyArg_ParseTuple(obj, "ii", &r.start, &r.end) > 0;
 }
 
-static inline PyObject* pyopencv_from(const Range& r)
+template<>
+PyObject* pyopencv_from(const Range& r)
 {
     return Py_BuildValue("(ii)", r.start, r.end);
 }
 
-static inline bool pyopencv_to(PyObject* obj, CvSlice& r, const char* name = "<unknown>")
-{
-    (void)name;
-    if(!obj || obj == Py_None)
-        return true;
-    if(PyObject_Size(obj) == 0)
-    {
-        r = CV_WHOLE_SEQ;
-        return true;
-    }
-    return PyArg_ParseTuple(obj, "ii", &r.start_index, &r.end_index) > 0;
-}
-
-static inline PyObject* pyopencv_from(const CvSlice& r)
-{
-    return Py_BuildValue("(ii)", r.start_index, r.end_index);
-}
-
-static inline bool pyopencv_to(PyObject* obj, Point& p, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Point& p, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -637,7 +672,8 @@ static inline bool pyopencv_to(PyObject* obj, Point& p, const char* name = "<unk
     return PyArg_ParseTuple(obj, "ii", &p.x, &p.y) > 0;
 }
 
-static inline bool pyopencv_to(PyObject* obj, Point2f& p, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Point2f& p, const char* name)
 {
     (void)name;
     if(!obj || obj == Py_None)
@@ -652,17 +688,37 @@ static inline bool pyopencv_to(PyObject* obj, Point2f& p, const char* name = "<u
     return PyArg_ParseTuple(obj, "ff", &p.x, &p.y) > 0;
 }
 
-static inline PyObject* pyopencv_from(const Point& p)
+template<>
+bool pyopencv_to(PyObject* obj, Point2d& p, const char* name)
+{
+    (void)name;
+    if(!obj || obj == Py_None)
+        return true;
+    if(!!PyComplex_CheckExact(obj))
+    {
+        Py_complex c = PyComplex_AsCComplex(obj);
+        p.x = saturate_cast<double>(c.real);
+        p.y = saturate_cast<double>(c.imag);
+        return true;
+    }
+    return PyArg_ParseTuple(obj, "dd", &p.x, &p.y) > 0;
+}
+
+
+template<>
+PyObject* pyopencv_from(const Point& p)
 {
     return Py_BuildValue("(ii)", p.x, p.y);
 }
 
-static inline PyObject* pyopencv_from(const Point2f& p)
+template<>
+PyObject* pyopencv_from(const Point2f& p)
 {
     return Py_BuildValue("(dd)", p.x, p.y);
 }
 
-static inline bool pyopencv_to(PyObject* obj, Vec3d& v, const char* name = "<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, Vec3d& v, const char* name)
 {
     (void)name;
     if(!obj)
@@ -670,12 +726,20 @@ static inline bool pyopencv_to(PyObject* obj, Vec3d& v, const char* name = "<unk
     return PyArg_ParseTuple(obj, "ddd", &v[0], &v[1], &v[2]) > 0;
 }
 
-static inline PyObject* pyopencv_from(const Vec3d& v)
+template<>
+PyObject* pyopencv_from(const Vec3d& v)
 {
     return Py_BuildValue("(ddd)", v[0], v[1], v[2]);
 }
 
-static inline PyObject* pyopencv_from(const Point2d& p)
+template<>
+PyObject* pyopencv_from(const Vec2d& v)
+{
+    return Py_BuildValue("(dd)", v[0], v[1]);
+}
+
+template<>
+PyObject* pyopencv_from(const Point2d& p)
 {
     return Py_BuildValue("(dd)", p.x, p.y);
 }
@@ -789,20 +853,18 @@ template<typename _Tp> struct pyopencvVecConverter
     }
 };
 
-
-template<typename _Tp> static inline bool pyopencv_to(PyObject* obj, std::vector<_Tp>& value, const ArgInfo info)
+template<typename _Tp>
+bool pyopencv_to(PyObject* obj, std::vector<_Tp>& value, const ArgInfo info)
 {
     return pyopencvVecConverter<_Tp>::to(obj, value, info);
 }
 
-template<typename _Tp> static inline PyObject* pyopencv_from(const std::vector<_Tp>& value)
+template<typename _Tp>
+PyObject* pyopencv_from(const std::vector<_Tp>& value)
 {
     return pyopencvVecConverter<_Tp>::from(value);
 }
 
-static PyObject* pyopencv_from(const KeyPoint&);
-static PyObject* pyopencv_from(const DMatch&);
-
 template<typename _Tp> static inline bool pyopencv_to_generic_vec(PyObject* obj, std::vector<_Tp>& value, const ArgInfo info)
 {
     if(!obj || obj == Py_None)
@@ -849,9 +911,9 @@ template<typename _Tp> static inline PyObject* pyopencv_from_generic_vec(const s
 
 template<typename _Tp> struct pyopencvVecConverter<std::vector<_Tp> >
 {
-    static bool to(PyObject* obj, std::vector<std::vector<_Tp> >& value, const char* name="<unknown>")
+    static bool to(PyObject* obj, std::vector<std::vector<_Tp> >& value, const ArgInfo info)
     {
-        return pyopencv_to_generic_vec(obj, value, name);
+        return pyopencv_to_generic_vec(obj, value, info);
     }
 
     static PyObject* from(const std::vector<std::vector<_Tp> >& value)
@@ -899,34 +961,21 @@ template<> struct pyopencvVecConverter<DMatch>
     }
 };
 
-template<> struct pyopencvVecConverter<std::string>
+template<> struct pyopencvVecConverter<String>
 {
-    static bool to(PyObject* obj, std::vector<std::string>& value, const ArgInfo info)
+    static bool to(PyObject* obj, std::vector<String>& value, const ArgInfo info)
     {
         return pyopencv_to_generic_vec(obj, value, info);
     }
 
-    static PyObject* from(const std::vector<std::string>& value)
+    static PyObject* from(const std::vector<String>& value)
     {
         return pyopencv_from_generic_vec(value);
     }
 };
 
-
-static inline bool pyopencv_to(PyObject *obj, CvTermCriteria& dst, const char *name="<unknown>")
-{
-    (void)name;
-    if(!obj)
-        return true;
-    return PyArg_ParseTuple(obj, "iid", &dst.type, &dst.max_iter, &dst.epsilon) > 0;
-}
-
-static inline PyObject* pyopencv_from(const CvTermCriteria& src)
-{
-    return Py_BuildValue("(iid)", src.type, src.max_iter, src.epsilon);
-}
-
-static inline bool pyopencv_to(PyObject *obj, TermCriteria& dst, const char *name="<unknown>")
+template<>
+bool pyopencv_to(PyObject *obj, TermCriteria& dst, const char *name)
 {
     (void)name;
     if(!obj)
@@ -934,12 +983,14 @@ static inline bool pyopencv_to(PyObject *obj, TermCriteria& dst, const char *nam
     return PyArg_ParseTuple(obj, "iid", &dst.type, &dst.maxCount, &dst.epsilon) > 0;
 }
 
-static inline PyObject* pyopencv_from(const TermCriteria& src)
+template<>
+PyObject* pyopencv_from(const TermCriteria& src)
 {
     return Py_BuildValue("(iid)", src.type, src.maxCount, src.epsilon);
 }
 
-static inline bool pyopencv_to(PyObject *obj, RotatedRect& dst, const char *name="<unknown>")
+template<>
+bool pyopencv_to(PyObject *obj, RotatedRect& dst, const char *name)
 {
     (void)name;
     if(!obj)
@@ -947,12 +998,14 @@ static inline bool pyopencv_to(PyObject *obj, RotatedRect& dst, const char *name
     return PyArg_ParseTuple(obj, "(ff)(ff)f", &dst.center.x, &dst.center.y, &dst.size.width, &dst.size.height, &dst.angle) > 0;
 }
 
-static inline PyObject* pyopencv_from(const RotatedRect& src)
+template<>
+PyObject* pyopencv_from(const RotatedRect& src)
 {
     return Py_BuildValue("((ff)(ff)f)", src.center.x, src.center.y, src.size.width, src.size.height, src.angle);
 }
 
-static inline PyObject* pyopencv_from(const Moments& m)
+template<>
+PyObject* pyopencv_from(const Moments& m)
 {
     return Py_BuildValue("{s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d,s:d}",
                          "m00", m.m00, "m10", m.m10, "m01", m.m01,
@@ -961,17 +1014,11 @@ static inline PyObject* pyopencv_from(const Moments& m)
                          "mu20", m.mu20, "mu11", m.mu11, "mu02", m.mu02,
                          "mu30", m.mu30, "mu21", m.mu21, "mu12", m.mu12, "mu03", m.mu03,
                          "nu20", m.nu20, "nu11", m.nu11, "nu02", m.nu02,
-                         "nu30", m.nu30, "nu21", m.nu21, "nu12", m.nu12, "mu03", m.nu03);
+                         "nu30", m.nu30, "nu21", m.nu21, "nu12", m.nu12, "nu03", m.nu03);
 }
 
-static inline PyObject* pyopencv_from(const CvDTreeNode* node)
-{
-    double value = node->value;
-    int ivalue = cvRound(value);
-    return value == ivalue ? PyInt_FromLong(ivalue) : PyFloat_FromDouble(value);
-}
-
-static bool pyopencv_to(PyObject *o, cv::flann::IndexParams& p, const char *name="<unknown>")
+template<>
+bool pyopencv_to(PyObject *o, cv::flann::IndexParams& p, const char *name)
 {
     (void)name;
     bool ok = false;
@@ -987,7 +1034,7 @@ static bool pyopencv_to(PyObject *o, cv::flann::IndexParams& p, const char *name
             PyObject* item = PyList_GET_ITEM(values, i);
             if( !PyString_Check(key) )
                 break;
-            std::string k = PyString_AsString(key);
+            String k = PyString_AsString(key);
             if( PyString_Check(item) )
             {
                 const char* value = PyString_AsString(item);
@@ -1019,15 +1066,21 @@ static bool pyopencv_to(PyObject *o, cv::flann::IndexParams& p, const char *name
     return ok;
 }
 
-template <class T>
-static bool pyopencv_to(PyObject *o, Ptr<T>& p, const char *name="<unknown>")
+template<>
+bool pyopencv_to(PyObject* obj, cv::flann::SearchParams & value, const char * name)
 {
-    p = new T();
+    return pyopencv_to<cv::flann::IndexParams>(obj, value, name);
+}
+
+template <typename T>
+bool pyopencv_to(PyObject *o, Ptr<T>& p, const char *name)
+{
+    p = makePtr<T>();
     return pyopencv_to(o, *p, name);
 }
 
-
-static bool pyopencv_to(PyObject *o, cvflann::flann_distance_t& dist, const char *name="<unknown>")
+template<>
+bool pyopencv_to(PyObject *o, cvflann::flann_distance_t& dist, const char *name)
 {
     int d = (int)dist;
     bool ok = pyopencv_to(o, d, name);
@@ -1035,6 +1088,41 @@ static bool pyopencv_to(PyObject *o, cvflann::flann_distance_t& dist, const char
     return ok;
 }
 
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// TODO: REMOVE used only by ml wrapper
+
+template<>
+bool pyopencv_to(PyObject *obj, CvTermCriteria& dst, const char *name)
+{
+    (void)name;
+    if(!obj)
+        return true;
+    return PyArg_ParseTuple(obj, "iid", &dst.type, &dst.max_iter, &dst.epsilon) > 0;
+}
+
+template<>
+bool pyopencv_to(PyObject* obj, CvSlice& r, const char* name)
+{
+    (void)name;
+    if(!obj || obj == Py_None)
+        return true;
+    if(PyObject_Size(obj) == 0)
+    {
+        r = CV_WHOLE_SEQ;
+        return true;
+    }
+    return PyArg_ParseTuple(obj, "ii", &r.start_index, &r.end_index) > 0;
+}
+
+template<>
+PyObject* pyopencv_from(CvDTreeNode* const & node)
+{
+    double value = node->value;
+    int ivalue = cvRound(value);
+    return value == ivalue ? PyInt_FromLong(ivalue) : PyFloat_FromDouble(value);
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 static void OnMouse(int event, int x, int y, int flags, void* param)
@@ -1070,7 +1158,7 @@ static PyObject *pycvSetMouseCallback(PyObject*, PyObject *args, PyObject *kw)
     if (param == NULL) {
         param = Py_None;
     }
-    ERRWRAP2(cvSetMouseCallback(name, OnMouse, Py_BuildValue("OO", on_mouse, param)));
+    ERRWRAP2(setMouseCallback(name, OnMouse, Py_BuildValue("OO", on_mouse, param)));
     Py_RETURN_NONE;
 }
 
@@ -1102,13 +1190,28 @@ static PyObject *pycvCreateTrackbar(PyObject*, PyObject *args)
         PyErr_SetString(PyExc_TypeError, "on_change must be callable");
         return NULL;
     }
-    ERRWRAP2(cvCreateTrackbar2(trackbar_name, window_name, value, count, OnChange, Py_BuildValue("OO", on_change, Py_None)));
+    ERRWRAP2(createTrackbar(trackbar_name, window_name, value, count, OnChange, Py_BuildValue("OO", on_change, Py_None)));
     Py_RETURN_NONE;
 }
 
 ///////////////////////////////////////////////////////////////////////////////////////
 
+static int convert_to_char(PyObject *o, char *dst, const char *name = "no_name")
+{
+  if (PyString_Check(o) && PyString_Size(o) == 1) {
+    *dst = PyString_AsString(o)[0];
+    return 1;
+  } else {
+    (*dst) = 0;
+    return failmsg("Expected single character string for argument '%s'", name);
+  }
+}
+
+#if PY_MAJOR_VERSION >= 3
+#define MKTYPE2(NAME) pyopencv_##NAME##_specials(); if (!to_ok(&pyopencv_##NAME##_Type)) return NULL;
+#else
 #define MKTYPE2(NAME) pyopencv_##NAME##_specials(); if (!to_ok(&pyopencv_##NAME##_Type)) return
+#endif
 
 #ifdef __GNUC__
 #  pragma GCC diagnostic ignored "-Wunused-parameter"
@@ -1122,7 +1225,7 @@ static PyMethodDef methods[] = {
 
 #include "pyopencv_generated_func_tab.h"
   {"createTrackbar", pycvCreateTrackbar, METH_VARARGS, "createTrackbar(trackbarName, windowName, value, count, onChange) -> None"},
-  {"setMouseCallback", (PyCFunction)pycvSetMouseCallback, METH_KEYWORDS, "setMouseCallback(windowName, onMouse [, param]) -> None"},
+  {"setMouseCallback", (PyCFunction)pycvSetMouseCallback, METH_VARARGS | METH_KEYWORDS, "setMouseCallback(windowName, onMouse [, param]) -> None"},
   {NULL, NULL},
 };
 
@@ -1137,25 +1240,35 @@ static int to_ok(PyTypeObject *to)
   return (PyType_Ready(to) == 0);
 }
 
-#include "cv2.cv.hpp"
 
-extern "C"
-#if defined WIN32 || defined _WIN32
-__declspec(dllexport)
-#endif
-void initcv2();
+#if PY_MAJOR_VERSION >= 3
+extern "C" CV_EXPORTS PyObject* PyInit_cv2();
+static struct PyModuleDef cv2_moduledef =
+{
+    PyModuleDef_HEAD_INIT,
+    MODULESTR,
+    "Python wrapper for OpenCV.",
+    -1,     /* size of per-interpreter state of the module,
+               or -1 if the module keeps state in global variables. */
+    methods
+};
+
+PyObject* PyInit_cv2()
+#else
+extern "C" CV_EXPORTS void initcv2();
 
 void initcv2()
+#endif
 {
-#if PYTHON_USE_NUMPY
-    import_array();
-#endif
+  import_array();
 
-#if PYTHON_USE_NUMPY
 #include "pyopencv_generated_type_reg.h"
-#endif
 
+#if PY_MAJOR_VERSION >= 3
+  PyObject* m = PyModule_Create(&cv2_moduledef);
+#else
   PyObject* m = Py_InitModule(MODULESTR, methods);
+#endif
   PyObject* d = PyModule_GetDict(m);
 
   PyDict_SetItemString(d, "__version__", PyString_FromString(CV_VERSION));
@@ -1163,27 +1276,10 @@ void initcv2()
   opencv_error = PyErr_NewException((char*)MODULESTR".error", NULL, NULL);
   PyDict_SetItemString(d, "error", opencv_error);
 
-  PyObject* cv_m = init_cv();
-
-  PyDict_SetItemString(d, "cv", cv_m);
-
 #define PUBLISH(I) PyDict_SetItemString(d, #I, PyInt_FromLong(I))
-#define PUBLISHU(I) PyDict_SetItemString(d, #I, PyLong_FromUnsignedLong(I))
+//#define PUBLISHU(I) PyDict_SetItemString(d, #I, PyLong_FromUnsignedLong(I))
 #define PUBLISH2(I, value) PyDict_SetItemString(d, #I, PyLong_FromLong(value))
 
-  PUBLISHU(IPL_DEPTH_8U);
-  PUBLISHU(IPL_DEPTH_8S);
-  PUBLISHU(IPL_DEPTH_16U);
-  PUBLISHU(IPL_DEPTH_16S);
-  PUBLISHU(IPL_DEPTH_32S);
-  PUBLISHU(IPL_DEPTH_32F);
-  PUBLISHU(IPL_DEPTH_64F);
-
-  PUBLISH(CV_LOAD_IMAGE_COLOR);
-  PUBLISH(CV_LOAD_IMAGE_GRAYSCALE);
-  PUBLISH(CV_LOAD_IMAGE_UNCHANGED);
-  PUBLISH(CV_HIST_ARRAY);
-  PUBLISH(CV_HIST_SPARSE);
   PUBLISH(CV_8U);
   PUBLISH(CV_8UC1);
   PUBLISH(CV_8UC2);
@@ -1219,37 +1315,9 @@ void initcv2()
   PUBLISH(CV_64FC2);
   PUBLISH(CV_64FC3);
   PUBLISH(CV_64FC4);
-  PUBLISH(CV_NEXT_AROUND_ORG);
-  PUBLISH(CV_NEXT_AROUND_DST);
-  PUBLISH(CV_PREV_AROUND_ORG);
-  PUBLISH(CV_PREV_AROUND_DST);
-  PUBLISH(CV_NEXT_AROUND_LEFT);
-  PUBLISH(CV_NEXT_AROUND_RIGHT);
-  PUBLISH(CV_PREV_AROUND_LEFT);
-  PUBLISH(CV_PREV_AROUND_RIGHT);
-
-  PUBLISH(CV_WINDOW_AUTOSIZE);
-
-  PUBLISH(CV_PTLOC_INSIDE);
-  PUBLISH(CV_PTLOC_ON_EDGE);
-  PUBLISH(CV_PTLOC_VERTEX);
-  PUBLISH(CV_PTLOC_OUTSIDE_RECT);
-
-  PUBLISH(GC_BGD);
-  PUBLISH(GC_FGD);
-  PUBLISH(GC_PR_BGD);
-  PUBLISH(GC_PR_FGD);
-  PUBLISH(GC_INIT_WITH_RECT);
-  PUBLISH(GC_INIT_WITH_MASK);
-  PUBLISH(GC_EVAL);
-
-  PUBLISH(CV_ROW_SAMPLE);
-  PUBLISH(CV_VAR_NUMERICAL);
-  PUBLISH(CV_VAR_ORDERED);
-  PUBLISH(CV_VAR_CATEGORICAL);
-
-  PUBLISH(CV_AA);
 
 #include "pyopencv_generated_const_reg.h"
+#if PY_MAJOR_VERSION >= 3
+    return m;
+#endif
 }
-
diff --git a/modules/python/src2/cv2.cv.hpp b/modules/python/src2/cv2.cv.hpp
deleted file mode 100644
index f7a6f73777..0000000000
--- a/modules/python/src2/cv2.cv.hpp
+++ /dev/null
@@ -1,4053 +0,0 @@
-#include "opencv2/legacy.hpp"
-#include "opencv2/legacy/compat.hpp"
-
-#define OLD_MODULESTR "cv2.cv"
-
-struct memtrack_t {
-  PyObject_HEAD
-  int owner;
-  void *ptr;
-  int freeptr;
-  Py_ssize_t size;
-  PyObject *backing;
-  CvArr *backingmat;
-};
-
-struct iplimage_t {
-  PyObject_HEAD
-  IplImage *a;
-  PyObject *data;
-  size_t offset;
-};
-
-struct cvmat_t {
-  PyObject_HEAD
-  CvMat *a;
-  PyObject *data;
-  size_t offset;
-};
-
-struct cvmatnd_t {
-  PyObject_HEAD
-  CvMatND *a;
-  PyObject *data;
-  size_t offset;
-};
-
-struct cvhistogram_t {
-  PyObject_HEAD
-  CvHistogram h;
-  PyObject *bins;
-};
-
-struct cvmemstorage_t {
-  PyObject_HEAD
-  CvMemStorage *a;
-};
-
-struct cvseq_t {
-  PyObject_HEAD
-  CvSeq *a;
-  PyObject *container;  // Containing cvmemstorage_t
-};
-
-struct cvset_t {
-  PyObject_HEAD
-  CvSet *a;
-  PyObject *container;  // Containing cvmemstorage_t
-  int i;
-};
-
-struct cvsubdiv2d_t {
-  PyObject_HEAD
-  CvSubdiv2D *a;
-  PyObject *container;  // Containing cvmemstorage_t
-};
-
-struct cvsubdiv2dpoint_t {
-  PyObject_HEAD
-  CvSubdiv2DPoint *a;
-  PyObject *container;  // Containing cvmemstorage_t
-};
-
-struct cvsubdiv2dedge_t {
-  PyObject_HEAD
-  CvSubdiv2DEdge a;
-  PyObject *container;  // Containing cvmemstorage_t
-};
-
-struct cvlineiterator_t {
-  PyObject_HEAD
-  CvLineIterator iter;
-  int count;
-  int type;
-};
-
-typedef IplImage ROIplImage;
-typedef const CvMat ROCvMat;
-typedef PyObject PyCallableObject;
-
-struct cvfont_t {
-  PyObject_HEAD
-  CvFont a;
-};
-
-struct cvcontourtree_t {
-  PyObject_HEAD
-  CvContourTree *a;
-};
-
-struct cvrng_t {
-  PyObject_HEAD
-  CvRNG a;
-};
-
-static int is_iplimage(PyObject *o);
-static int is_cvmat(PyObject *o);
-static int is_cvmatnd(PyObject *o);
-static int convert_to_CvArr(PyObject *o, CvArr **dst, const char *name = "no_name");
-static int convert_to_IplImage(PyObject *o, IplImage **dst, const char *name = "no_name");
-static int convert_to_CvMat(PyObject *o, CvMat **dst, const char *name = "no_name");
-static int convert_to_CvMatND(PyObject *o, CvMatND **dst, const char *name = "no_name");
-static PyObject *what_data(PyObject *o);
-static PyObject *FROM_CvMat(CvMat *r);
-static PyObject *FROM_ROCvMatPTR(ROCvMat *r);
-static PyObject *shareDataND(PyObject *donor, CvMatND *pdonor, CvMatND *precipient);
-
-#define FROM_double(r)  PyFloat_FromDouble(r)
-#define FROM_float(r)  PyFloat_FromDouble(r)
-#define FROM_int(r)  PyInt_FromLong(r)
-#define FROM_int64(r)  PyLong_FromLongLong(r)
-#define FROM_unsigned(r)  PyLong_FromUnsignedLong(r)
-#define FROM_CvBox2D(r) Py_BuildValue("(ff)(ff)f", r.center.x, r.center.y, r.size.width, r.size.height, r.angle)
-#define FROM_CvScalar(r)  Py_BuildValue("(ffff)", r.val[0], r.val[1], r.val[2], r.val[3])
-#define FROM_CvPoint(r)  Py_BuildValue("(ii)", r.x, r.y)
-#define FROM_CvPoint2D32f(r) Py_BuildValue("(ff)", r.x, r.y)
-#define FROM_CvPoint3D64f(r) Py_BuildValue("(fff)", r.x, r.y, r.z)
-#define FROM_CvSize(r) Py_BuildValue("(ii)", r.width, r.height)
-#define FROM_CvRect(r) Py_BuildValue("(iiii)", r.x, r.y, r.width, r.height)
-#define FROM_CvSeqPTR(r) _FROM_CvSeqPTR(r, pyobj_storage)
-#define FROM_CvSubdiv2DPTR(r) _FROM_CvSubdiv2DPTR(r, pyobj_storage)
-#define FROM_CvPoint2D64f(r) Py_BuildValue("(ff)", r.x, r.y)
-#define FROM_CvConnectedComp(r) Py_BuildValue("(fNN)", (r).area, FROM_CvScalar((r).value), FROM_CvRect((r).rect))
-
-#if PYTHON_USE_NUMPY
-static PyObject *fromarray(PyObject *o, int allowND);
-#endif
-
-static void translate_error_to_exception(void)
-{
-  PyErr_SetString(opencv_error, cvErrorStr(cvGetErrStatus()));
-  cvSetErrStatus(0);
-}
-
-#define ERRCHK do { if (cvGetErrStatus() != 0) { translate_error_to_exception(); return 0; } } while (0)
-#define ERRWRAPN(F, N) \
-    do { \
-        try \
-        { \
-            F; \
-        } \
-        catch (const cv::Exception &e) \
-        { \
-           PyErr_SetString(opencv_error, e.err.c_str()); \
-           return N; \
-        } \
-        ERRCHK; \
-    } while(0)
-#define ERRWRAP(F) ERRWRAPN(F, NULL) // for most functions, exception -> NULL return
-
-/************************************************************************/
-
-/* These get/setters are polymorphic, used in both iplimage and cvmat */
-
-static PyObject *PyObject_FromCvScalar(CvScalar s, int type)
-{
-  int i, spe = CV_MAT_CN(type);
-  PyObject *r;
-  if (spe > 1) {
-    r = PyTuple_New(spe);
-    for (i = 0; i < spe; i++)
-      PyTuple_SET_ITEM(r, i, PyFloat_FromDouble(s.val[i]));
-  } else {
-    r = PyFloat_FromDouble(s.val[0]);
-  }
-  return r;
-}
-
-static PyObject *cvarr_GetItem(PyObject *o, PyObject *key);
-static int cvarr_SetItem(PyObject *o, PyObject *key, PyObject *v);
-
-// o is a Python string or buffer object.  Return its size.
-
-static Py_ssize_t what_size(PyObject *o)
-{
-  void *buffer;
-  Py_ssize_t buffer_len;
-
-  if (PyString_Check(o)) {
-    return PyString_Size(o);
-  } else if (PyObject_AsWriteBuffer(o, &buffer, &buffer_len) == 0) {
-    return buffer_len;
-  } else {
-    assert(0);  // argument must be string or buffer.
-    return 0;
-  }
-}
-
-
-/************************************************************************/
-
-static CvMat *PyCvMat_AsCvMat(PyObject *o)
-{
-  assert(0); // not yet implemented: reference counting for CvMat in Kalman is unclear...
-  return NULL;
-}
-
-#define cvReleaseIplConvKernel(x) cvReleaseStructuringElement(x)
-
-#if defined _MSC_VER && _MSC_VER >= 1200
-    #pragma warning( push )
-    #pragma warning( disable : 4244 )
-#endif
-
-#include "generated3.i"
-
-#if defined _MSC_VER && _MSC_VER >= 1200
-    #pragma warning( pop )
-#endif
-
-/* iplimage */
-
-static void iplimage_dealloc(PyObject *self)
-{
-  iplimage_t *pc = (iplimage_t*)self;
-  cvReleaseImageHeader((IplImage**)&pc->a);
-  Py_DECREF(pc->data);
-  PyObject_Del(self);
-}
-
-static PyObject *iplimage_repr(PyObject *self)
-{
-  iplimage_t *cva = (iplimage_t*)self;
-  IplImage* ipl = (IplImage*)(cva->a);
-  char str[1000];
-  sprintf(str, "<iplimage(");
-  char *d = str + strlen(str);
-  sprintf(d, "nChannels=%d ", ipl->nChannels);
-  d += strlen(d);
-  sprintf(d, "width=%d ", ipl->width);
-  d += strlen(d);
-  sprintf(d, "height=%d ", ipl->height);
-  d += strlen(d);
-  sprintf(d, "widthStep=%d ", ipl->widthStep);
-  d += strlen(d);
-  sprintf(d, ")>");
-  return PyString_FromString(str);
-}
-
-static PyObject *iplimage_tostring(PyObject *self, PyObject *args)
-{
-  IplImage *i=0;
-  if (!convert_to_IplImage(self, &i, "self"))
-    return NULL;
-  if (i == NULL)
-    return NULL;
-  cv::Mat img(i);
-  size_t esz = img.elemSize();
-  int nrows = img.rows, ncols = img.cols;
-
-  if( !img.isContinuous() )
-      img = img.clone();
-  return PyString_FromStringAndSize((char*)img.data, (Py_ssize_t)(esz*nrows*ncols));
-}
-
-static struct PyMethodDef iplimage_methods[] =
-{
-  {"tostring", iplimage_tostring, METH_VARARGS},
-  {NULL,          NULL}
-};
-
-static PyObject *iplimage_getnChannels(iplimage_t *cva)
-{
-  return PyInt_FromLong(((IplImage*)(cva->a))->nChannels);
-}
-static PyObject *iplimage_getwidth(iplimage_t *cva)
-{
-  return PyInt_FromLong(((IplImage*)(cva->a))->width);
-}
-static PyObject *iplimage_getheight(iplimage_t *cva)
-{
-  return PyInt_FromLong(((IplImage*)(cva->a))->height);
-}
-static PyObject *iplimage_getdepth(iplimage_t *cva)
-{
-  return PyLong_FromUnsignedLong((unsigned)((IplImage*)(cva->a))->depth);
-}
-static PyObject *iplimage_getorigin(iplimage_t *cva)
-{
-  return PyInt_FromLong(((IplImage*)(cva->a))->origin);
-}
-static void iplimage_setorigin(iplimage_t *cva, PyObject *v)
-{
-  ((IplImage*)(cva->a))->origin = PyInt_AsLong(v);
-}
-
-static PyGetSetDef iplimage_getseters[] = {
-  {(char*)"nChannels", (getter)iplimage_getnChannels, (setter)NULL, (char*)"nChannels", NULL},
-  {(char*)"channels", (getter)iplimage_getnChannels, (setter)NULL, (char*)"nChannels", NULL},
-  {(char*)"width", (getter)iplimage_getwidth, (setter)NULL, (char*)"width", NULL},
-  {(char*)"height", (getter)iplimage_getheight, (setter)NULL, (char*)"height", NULL},
-  {(char*)"depth", (getter)iplimage_getdepth, (setter)NULL, (char*)"depth", NULL},
-  {(char*)"origin", (getter)iplimage_getorigin, (setter)iplimage_setorigin, (char*)"origin", NULL},
-  {NULL}  /* Sentinel */
-};
-
-static PyMappingMethods iplimage_as_map = {
-  NULL,
-  &cvarr_GetItem,
-  &cvarr_SetItem,
-};
-
-static PyTypeObject iplimage_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".iplimage",                          /*name*/
-  sizeof(iplimage_t),                        /*basicsize*/
-};
-
-static void iplimage_specials(void)
-{
-  iplimage_Type.tp_dealloc = iplimage_dealloc;
-  iplimage_Type.tp_as_mapping = &iplimage_as_map;
-  iplimage_Type.tp_repr = iplimage_repr;
-  iplimage_Type.tp_methods = iplimage_methods;
-  iplimage_Type.tp_getset = iplimage_getseters;
-}
-
-static int is_iplimage(PyObject *o)
-{
-  return PyType_IsSubtype(o->ob_type, &iplimage_Type);
-}
-
-/************************************************************************/
-
-/* cvmat */
-
-static void cvmat_dealloc(PyObject *self)
-{
-  cvmat_t *pc = (cvmat_t*)self;
-  Py_XDECREF(pc->data);
-  //cvDecRefData(pc->a);
-  cvFree(&pc->a);
-  PyObject_Del(self);
-}
-
-static PyObject *cvmat_repr(PyObject *self)
-{
-  CvMat *m = ((cvmat_t*)self)->a;
-  char str[1000];
-  sprintf(str, "<cvmat(");
-  char *d = str + strlen(str);
-  sprintf(d, "type=%08x ", m->type);
-  d += strlen(d);
-  switch (CV_MAT_DEPTH(m->type)) {
-  case CV_8U: strcpy(d, "8U"); break;
-  case CV_8S: strcpy(d, "8S"); break;
-  case CV_16U: strcpy(d, "16U"); break;
-  case CV_16S: strcpy(d, "16S"); break;
-  case CV_32S: strcpy(d, "32S"); break;
-  case CV_32F: strcpy(d, "32F"); break;
-  case CV_64F: strcpy(d, "64F"); break;
-  }
-  d += strlen(d);
-  sprintf(d, "C%d ", CV_MAT_CN(m->type));
-  d += strlen(d);
-  sprintf(d, "rows=%d ", m->rows);
-  d += strlen(d);
-  sprintf(d, "cols=%d ", m->cols);
-  d += strlen(d);
-  sprintf(d, "step=%d ", m->step);
-  d += strlen(d);
-  sprintf(d, ")>");
-  return PyString_FromString(str);
-}
-
-static PyObject *cvmat_tostring(PyObject *self, PyObject *args)
-{
-  CvMat *m;
-  if (!convert_to_CvMat(self, &m, "self"))
-    return NULL;
-
-  int bps;                     // bytes per sample
-
-  switch (CV_MAT_DEPTH(m->type)) {
-  case CV_8U:
-  case CV_8S:
-    bps = CV_MAT_CN(m->type) * 1;
-    break;
-  case CV_16U:
-  case CV_16S:
-    bps = CV_MAT_CN(m->type) * 2;
-    break;
-  case CV_32S:
-  case CV_32F:
-    bps = CV_MAT_CN(m->type) * 4;
-    break;
-  case CV_64F:
-    bps = CV_MAT_CN(m->type) * 8;
-    break;
-  default:
-    failmsg("Unrecognized depth %d", CV_MAT_DEPTH(m->type));
-    return (PyObject*)0;
-  }
-
-  int bpl = m->cols * bps; // bytes per line
-  cvmat_t *pc = (cvmat_t*)self;
-  if (PyString_Check(pc->data) && bpl == m->step && pc->offset == 0 && ((bpl * m->rows) == what_size(pc->data))) {
-    Py_INCREF(pc->data);
-    return pc->data;
-  } else {
-    int l = bpl * m->rows;
-    char *s = new char[l];
-    int y;
-    for (y = 0; y < m->rows; y++) {
-      memcpy(s + y * bpl, m->data.ptr + y * m->step, bpl);
-    }
-    PyObject *r = PyString_FromStringAndSize(s, l);
-    delete[] s;
-    return r;
-  }
-}
-
-static struct PyMethodDef cvmat_methods[] =
-{
-  {"tostring", cvmat_tostring, METH_VARARGS},
-  {NULL,          NULL}
-};
-
-static PyObject *cvmat_gettype(cvmat_t *cva)
-{
-  return PyInt_FromLong(cvGetElemType(cva->a));
-}
-
-static PyObject *cvmat_getstep(cvmat_t *cva)
-{
-  return PyInt_FromLong(cva->a->step);
-}
-
-static PyObject *cvmat_getrows(cvmat_t *cva)
-{
-  return PyInt_FromLong(cva->a->rows);
-}
-
-static PyObject *cvmat_getcols(cvmat_t *cva)
-{
-  return PyInt_FromLong(cva->a->cols);
-}
-
-static PyObject *cvmat_getchannels(cvmat_t *cva)
-{
-  return PyInt_FromLong(CV_MAT_CN(cva->a->type));
-}
-
-#if PYTHON_USE_NUMPY
-#include "numpy/ndarrayobject.h"
-
-// A PyArrayInterface, with an associated python object that should be DECREF'ed on release
-struct arrayTrack {
-  PyArrayInterface s;
-  PyObject *o;
-};
-
-static void arrayTrackDtor(void *p)
-{
-  struct arrayTrack *at = (struct arrayTrack *)p;
-  delete[] at->s.shape;
-  delete[] at->s.strides;
-  if (at->s.descr)
-    Py_DECREF(at->s.descr);
-  Py_DECREF(at->o);
-}
-
-// Fill in fields of PyArrayInterface s using mtype.  This code is common
-// to cvmat and cvmatnd
-
-static void arrayinterface_common(PyArrayInterface *s, int mtype)
-{
-  s->two = 2;
-
-  switch (CV_MAT_DEPTH(mtype)) {
-  case CV_8U:
-    s->typekind = 'u';
-    s->itemsize = 1;
-    break;
-  case CV_8S:
-    s->typekind = 'i';
-    s->itemsize = 1;
-    break;
-  case CV_16U:
-    s->typekind = 'u';
-    s->itemsize = 2;
-    break;
-  case CV_16S:
-    s->typekind = 'i';
-    s->itemsize = 2;
-    break;
-  case CV_32S:
-    s->typekind = 'i';
-    s->itemsize = 4;
-    break;
-  case CV_32F:
-    s->typekind = 'f';
-    s->itemsize = 4;
-    break;
-  case CV_64F:
-    s->typekind = 'f';
-    s->itemsize = 8;
-    break;
-  default:
-    assert(0);
-  }
-
-  s->flags = NPY_WRITEABLE | NPY_NOTSWAPPED;
-}
-
-static PyObject *cvmat_array_struct(cvmat_t *cva)
-{
-  CvMat *m;
-  convert_to_CvMat((PyObject *)cva, &m, "");
-
-  arrayTrack *at = new arrayTrack;
-  PyArrayInterface *s = &at->s;
-
-  at->o = cva->data;
-  Py_INCREF(at->o);
-
-  arrayinterface_common(s, m->type);
-
-  if (CV_MAT_CN(m->type) == 1) {
-    s->nd = 2;
-    s->shape = new npy_intp[2];
-    s->shape[0] = m->rows;
-    s->shape[1] = m->cols;
-    s->strides = new npy_intp[2];
-    s->strides[0] = m->step;
-    s->strides[1] = s->itemsize;
-  } else {
-    s->nd = 3;
-    s->shape = new npy_intp[3];
-    s->shape[0] = m->rows;
-    s->shape[1] = m->cols;
-    s->shape[2] = CV_MAT_CN(m->type);
-    s->strides = new npy_intp[3];
-    s->strides[0] = m->step;
-    s->strides[1] = s->itemsize * CV_MAT_CN(m->type);
-    s->strides[2] = s->itemsize;
-  }
-  s->data = (void*)(m->data.ptr);
-  s->descr = PyList_New(1);
-  char typestr[10];
-  sprintf(typestr, "<%c%d", s->typekind, s->itemsize);
-  PyList_SetItem(s->descr, 0, Py_BuildValue("(ss)", "x", typestr));
-
-  return PyCObject_FromVoidPtr(s, arrayTrackDtor);
-}
-
-static PyObject *cvmatnd_array_struct(cvmatnd_t *cva)
-{
-  CvMatND *m = 0;
-  convert_to_CvMatND((PyObject *)cva, &m, "");
-
-  arrayTrack *at = new arrayTrack;
-  PyArrayInterface *s = &at->s;
-
-  at->o = cva->data;
-  Py_INCREF(at->o);
-
-  arrayinterface_common(s, m->type);
-
-  int i;
-  if (CV_MAT_CN(m->type) == 1) {
-    s->nd = m->dims;
-    s->shape = new npy_intp[s->nd];
-    for (i = 0; i < s->nd; i++)
-      s->shape[i] = m->dim[i].size;
-    s->strides = new npy_intp[s->nd];
-    for (i = 0; i < (s->nd - 1); i++)
-      s->strides[i] = m->dim[i].step;
-    s->strides[s->nd - 1] = s->itemsize;
-  } else {
-    s->nd = m->dims + 1;
-    s->shape = new npy_intp[s->nd];
-    for (i = 0; i < (s->nd - 1); i++)
-      s->shape[i] = m->dim[i].size;
-    s->shape[s->nd - 1] = CV_MAT_CN(m->type);
-
-    s->strides = new npy_intp[s->nd];
-    for (i = 0; i < (s->nd - 2); i++)
-      s->strides[i] = m->dim[i].step;
-    s->strides[s->nd - 2] = s->itemsize * CV_MAT_CN(m->type);
-    s->strides[s->nd - 1] = s->itemsize;
-  }
-  s->data = (void*)(m->data.ptr);
-  s->descr = PyList_New(1);
-  char typestr[10];
-  sprintf(typestr, "<%c%d", s->typekind, s->itemsize);
-  PyList_SetItem(s->descr, 0, Py_BuildValue("(ss)", "x", typestr));
-
-  return PyCObject_FromVoidPtr(s, arrayTrackDtor);
-}
-#endif
-
-static PyGetSetDef cvmat_getseters[] = {
-  {(char*)"type",   (getter)cvmat_gettype, (setter)NULL, (char*)"type",   NULL},
-  {(char*)"step",   (getter)cvmat_getstep, (setter)NULL, (char*)"step",   NULL},
-  {(char*)"rows",   (getter)cvmat_getrows, (setter)NULL, (char*)"rows",   NULL},
-  {(char*)"cols",   (getter)cvmat_getcols, (setter)NULL, (char*)"cols",   NULL},
-  {(char*)"channels",(getter)cvmat_getchannels, (setter)NULL, (char*)"channels",   NULL},
-  {(char*)"width",  (getter)cvmat_getcols, (setter)NULL, (char*)"width",  NULL},
-  {(char*)"height", (getter)cvmat_getrows, (setter)NULL, (char*)"height", NULL},
-#if PYTHON_USE_NUMPY
-  {(char*)"__array_struct__", (getter)cvmat_array_struct, (setter)NULL, (char*)"__array_struct__", NULL},
-#endif
-  {NULL}  /* Sentinel */
-};
-
-static PyMappingMethods cvmat_as_map = {
-  NULL,
-  &cvarr_GetItem,
-  &cvarr_SetItem,
-};
-
-static PyTypeObject cvmat_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvmat",                      /*name*/
-  sizeof(cvmat_t),                        /*basicsize*/
-};
-
-static int illegal_init(PyObject *self, PyObject *args, PyObject *kwds)
-{
-  PyErr_SetString(opencv_error, "Cannot create cvmat directly; use CreateMat() instead");
-  return -1;
-}
-
-static void cvmat_specials(void)
-{
-  cvmat_Type.tp_dealloc = cvmat_dealloc;
-  cvmat_Type.tp_as_mapping = &cvmat_as_map;
-  cvmat_Type.tp_repr = cvmat_repr;
-  cvmat_Type.tp_methods = cvmat_methods;
-  cvmat_Type.tp_getset = cvmat_getseters;
-  cvmat_Type.tp_init = illegal_init;
-}
-
-static int is_cvmat(PyObject *o)
-{
-  return PyType_IsSubtype(o->ob_type, &cvmat_Type);
-}
-
-/************************************************************************/
-
-/* cvmatnd */
-
-static void cvmatnd_dealloc(PyObject *self)
-{
-  cvmatnd_t *pc = (cvmatnd_t*)self;
-  Py_XDECREF(pc->data);
-  cvDecRefData(pc->a);
-  cvFree(&pc->a);
-  PyObject_Del(self);
-}
-
-static PyObject *cvmatnd_repr(PyObject *self)
-{
-  CvMatND *m = ((cvmatnd_t*)self)->a;
-  char str[1000];
-  sprintf(str, "<cvmatnd(");
-  char *d = str + strlen(str);
-  sprintf(d, "type=%08x ", m->type);
-  d += strlen(d);
-  sprintf(d, ")>");
-  return PyString_FromString(str);
-}
-
-static size_t cvmatnd_size(CvMatND *m)
-{
-  int bps = 1;
-  switch (CV_MAT_DEPTH(m->type)) {
-  case CV_8U:
-  case CV_8S:
-    bps = CV_MAT_CN(m->type) * 1;
-    break;
-  case CV_16U:
-  case CV_16S:
-    bps = CV_MAT_CN(m->type) * 2;
-    break;
-  case CV_32S:
-  case CV_32F:
-    bps = CV_MAT_CN(m->type) * 4;
-    break;
-  case CV_64F:
-    bps = CV_MAT_CN(m->type) * 8;
-    break;
-  default:
-    assert(0);
-  }
-  size_t l = bps;
-  for (int d = 0; d < m->dims; d++) {
-    l *= m->dim[d].size;
-  }
-  return l;
-}
-
-static PyObject *cvmatnd_tostring(PyObject *self, PyObject *args)
-{
-  CvMatND *m = 0;
-  if (!convert_to_CvMatND(self, &m, "self"))
-    return NULL;
-
-  int bps;
-  switch (CV_MAT_DEPTH(m->type)) {
-  case CV_8U:
-  case CV_8S:
-    bps = CV_MAT_CN(m->type) * 1;
-    break;
-  case CV_16U:
-  case CV_16S:
-    bps = CV_MAT_CN(m->type) * 2;
-    break;
-  case CV_32S:
-  case CV_32F:
-    bps = CV_MAT_CN(m->type) * 4;
-    break;
-  case CV_64F:
-    bps = CV_MAT_CN(m->type) * 8;
-    break;
-  default:
-          return failmsg("Unrecognised depth %d", CV_MAT_DEPTH(m->type)), (PyObject*)0;
-  }
-
-  int d, l = bps;
-  for (d = 0; d < m->dims; d++) {
-    l *= m->dim[d].size;
-  }
-  int i[CV_MAX_DIM];
-  for (d = 0; d < m->dims; d++) {
-    i[d] = 0;
-  }
-  int rowsize = m->dim[m->dims-1].size * bps;
-  char *s = new char[l];
-  char *ps = s;
-
-  int finished = 0;
-  while (!finished) {
-    memcpy(ps, cvPtrND(m, i), rowsize);
-    ps += rowsize;
-    for (d = m->dims - 2; 0 <= d; d--) {
-      if (++i[d] < cvGetDimSize(m, d)) {
-        break;
-      } else {
-        i[d] = 0;
-      }
-    }
-    if (d < 0)
-      finished = 1;
-  }
-
-  return PyString_FromStringAndSize(s, ps - s);
-}
-
-static struct PyMethodDef cvmatnd_methods[] =
-{
-  {"tostring", cvmatnd_tostring, METH_VARARGS},
-  {NULL,          NULL}
-};
-
-static PyObject *cvmatnd_getchannels(cvmatnd_t *cva)
-{
-  return PyInt_FromLong(CV_MAT_CN(cva->a->type));
-}
-
-static PyGetSetDef cvmatnd_getseters[] = {
-#if PYTHON_USE_NUMPY
-  {(char*)"__array_struct__", (getter)cvmatnd_array_struct, (setter)NULL, (char*)"__array_struct__", NULL},
-#endif
-  {(char*)"channels",(getter)cvmatnd_getchannels, (setter)NULL, (char*)"channels",   NULL},
-  {NULL}  /* Sentinel */
-};
-
-static PyMappingMethods cvmatnd_as_map = {
-  NULL,
-  &cvarr_GetItem,
-  &cvarr_SetItem,
-};
-
-static PyTypeObject cvmatnd_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvmatnd",                          /*name*/
-  sizeof(cvmatnd_t),                        /*basicsize*/
-};
-
-static void cvmatnd_specials(void)
-{
-  cvmatnd_Type.tp_dealloc = cvmatnd_dealloc;
-  cvmatnd_Type.tp_as_mapping = &cvmatnd_as_map;
-  cvmatnd_Type.tp_repr = cvmatnd_repr;
-  cvmatnd_Type.tp_methods = cvmatnd_methods;
-  cvmatnd_Type.tp_getset = cvmatnd_getseters;
-}
-
-static int is_cvmatnd(PyObject *o)
-{
-  return PyType_IsSubtype(o->ob_type, &cvmatnd_Type);
-}
-
-/************************************************************************/
-
-/* cvhistogram */
-
-static void cvhistogram_dealloc(PyObject *self)
-{
-  cvhistogram_t *cvh = (cvhistogram_t*)self;
-  Py_DECREF(cvh->bins);
-  PyObject_Del(self);
-}
-
-static PyTypeObject cvhistogram_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvhistogram",                /*name*/
-  sizeof(cvhistogram_t),                  /*basicsize*/
-};
-
-static PyObject *cvhistogram_getbins(cvhistogram_t *cvh)
-{
-  Py_INCREF(cvh->bins);
-  return cvh->bins;
-}
-
-static PyGetSetDef cvhistogram_getseters[] = {
-  {(char*)"bins", (getter)cvhistogram_getbins, (setter)NULL, (char*)"bins", NULL},
-  {NULL}  /* Sentinel */
-};
-
-static void cvhistogram_specials(void)
-{
-  cvhistogram_Type.tp_dealloc = cvhistogram_dealloc;
-  cvhistogram_Type.tp_getset = cvhistogram_getseters;
-}
-
-/************************************************************************/
-
-/* cvlineiterator */
-
-static PyObject *cvlineiterator_iter(PyObject *o)
-{
-  Py_INCREF(o);
-  return o;
-}
-
-static PyObject *cvlineiterator_next(PyObject *o)
-{
-  cvlineiterator_t *pi = (cvlineiterator_t*)o;
-
-  if (pi->count) {
-      pi->count--;
-
-      CvScalar r;
-      cvRawDataToScalar( (void*)(pi->iter.ptr), pi->type, &r);
-      PyObject *pr = PyObject_FromCvScalar(r, pi->type);
-
-      CV_NEXT_LINE_POINT(pi->iter);
-
-      return pr;
-  } else {
-    return NULL;
-  }
-}
-
-static PyTypeObject cvlineiterator_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvlineiterator",             /*name*/
-  sizeof(cvlineiterator_t),               /*basicsize*/
-};
-
-static void cvlineiterator_specials(void)
-{
-  cvlineiterator_Type.tp_iter = cvlineiterator_iter;
-  cvlineiterator_Type.tp_iternext = cvlineiterator_next;
-}
-
-/************************************************************************/
-
-/* memtrack */
-
-/* Motivation for memtrack is when the storage for a Mat is an array or buffer
-object.  By setting 'data' to be a memtrack, can deallocate the storage at
-object destruction.
-
-For array objects, 'backing' is the actual storage object.  memtrack holds the reference,
-then DECREF's it at dealloc.
-
-For MatND's, we need to cvDecRefData() on release, and this is what field 'backingmat' is for.
-
-If freeptr is true, then a straight cvFree() of ptr happens.
-
-*/
-
-
-static void memtrack_dealloc(PyObject *self)
-{
-  memtrack_t *pi = (memtrack_t*)self;
-  if (pi->backing)
-    Py_DECREF(pi->backing);
-  if (pi->backingmat)
-    cvDecRefData(pi->backingmat);
-  if (pi->freeptr)
-    cvFree(&pi->ptr);
-  PyObject_Del(self);
-}
-
-static PyTypeObject memtrack_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".memtrack",                          /*name*/
-  sizeof(memtrack_t),                        /*basicsize*/
-};
-
-static Py_ssize_t memtrack_getreadbuffer(PyObject *self, Py_ssize_t segment, void **ptrptr)
-{
-  *ptrptr = &((memtrack_t*)self)->ptr;
-  return ((memtrack_t*)self)->size;
-}
-
-static Py_ssize_t memtrack_getwritebuffer(PyObject *self, Py_ssize_t segment, void **ptrptr)
-{
-  *ptrptr = ((memtrack_t*)self)->ptr;
-  return ((memtrack_t*)self)->size;
-}
-
-static Py_ssize_t memtrack_getsegcount(PyObject *self, Py_ssize_t *lenp)
-{
-  return (Py_ssize_t)1;
-}
-
-PyBufferProcs memtrack_as_buffer = {
-  memtrack_getreadbuffer,
-  memtrack_getwritebuffer,
-  memtrack_getsegcount
-};
-
-static void memtrack_specials(void)
-{
-  memtrack_Type.tp_dealloc = memtrack_dealloc;
-  memtrack_Type.tp_as_buffer = &memtrack_as_buffer;
-}
-
-/************************************************************************/
-
-/* cvmemstorage */
-
-static void cvmemstorage_dealloc(PyObject *self)
-{
-  cvmemstorage_t *ps = (cvmemstorage_t*)self;
-  cvReleaseMemStorage(&(ps->a));
-  PyObject_Del(self);
-}
-
-static PyTypeObject cvmemstorage_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvmemstorage",               /*name*/
-  sizeof(cvmemstorage_t),                 /*basicsize*/
-};
-
-static void cvmemstorage_specials(void)
-{
-  cvmemstorage_Type.tp_dealloc = cvmemstorage_dealloc;
-}
-
-/************************************************************************/
-
-/* cvfont */
-
-static PyTypeObject cvfont_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvfont",                     /*name*/
-  sizeof(cvfont_t),                       /*basicsize*/
-};
-
-static void cvfont_specials(void) { }
-
-/************************************************************************/
-
-/* cvrng */
-
-static PyTypeObject cvrng_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvrng",                     /*name*/
-  sizeof(cvrng_t),                       /*basicsize*/
-};
-
-static void cvrng_specials(void)
-{
-}
-
-/************************************************************************/
-
-/* cvcontourtree */
-
-static PyTypeObject cvcontourtree_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvcontourtree",                     /*name*/
-  sizeof(cvcontourtree_t),                       /*basicsize*/
-};
-
-static void cvcontourtree_specials(void) { }
-
-
-/************************************************************************/
-
-/* cvsubdiv2dedge */
-
-static PyTypeObject cvsubdiv2dedge_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvsubdiv2dedge",                     /*name*/
-  sizeof(cvsubdiv2dedge_t),                       /*basicsize*/
-};
-
-static int cvsubdiv2dedge_compare(PyObject *o1, PyObject *o2)
-{
-  cvsubdiv2dedge_t *e1 = (cvsubdiv2dedge_t*)o1;
-  cvsubdiv2dedge_t *e2 = (cvsubdiv2dedge_t*)o2;
-  if (e1->a < e2->a)
-    return -1;
-  else if (e1->a > e2->a)
-    return 1;
-  else
-    return 0;
-}
-
-static PyObject *cvquadedge_repr(PyObject *self)
-{
-  CvSubdiv2DEdge m = ((cvsubdiv2dedge_t*)self)->a;
-  char str[1000];
-  sprintf(str, "<cvsubdiv2dedge(");
-  char *d = str + strlen(str);
-  sprintf(d, "%lux.%d", (unsigned long)(m & ~3), (int)(m & 3));
-  d += strlen(d);
-  sprintf(d, ")>");
-  return PyString_FromString(str);
-}
-
-static void cvsubdiv2dedge_specials(void) {
-  cvsubdiv2dedge_Type.tp_compare = cvsubdiv2dedge_compare;
-  cvsubdiv2dedge_Type.tp_repr = cvquadedge_repr;
-}
-
-/************************************************************************/
-
-/* cvseq */
-
-static void cvseq_dealloc(PyObject *self)
-{
-  cvseq_t *ps = (cvseq_t*)self;
-  Py_DECREF(ps->container);
-  PyObject_Del(self);
-}
-
-static PyObject *cvseq_h_next(PyObject *self, PyObject *args);
-static PyObject *cvseq_h_prev(PyObject *self, PyObject *args);
-static PyObject *cvseq_v_next(PyObject *self, PyObject *args);
-static PyObject *cvseq_v_prev(PyObject *self, PyObject *args);
-
-static struct PyMethodDef cvseq_methods[] =
-{
-  {"h_next", cvseq_h_next, METH_VARARGS},
-  {"h_prev", cvseq_h_prev, METH_VARARGS},
-  {"v_next", cvseq_v_next, METH_VARARGS},
-  {"v_prev", cvseq_v_prev, METH_VARARGS},
-  {NULL,          NULL}
-};
-
-static Py_ssize_t cvseq_seq_length(PyObject *o)
-{
-  cvseq_t *ps = (cvseq_t*)o;
-  if (ps->a == NULL)
-    return (Py_ssize_t)0;
-  else
-    return (Py_ssize_t)(ps->a->total);
-}
-
-static PyObject* cvseq_seq_getitem(PyObject *o, Py_ssize_t i)
-{
-  cvseq_t *ps = (cvseq_t*)o;
-  CvPoint *pt;
-  struct pointpair{
-    CvPoint a, b;
-  } *pp;
-  CvPoint2D32f *pt2;
-  CvPoint3D32f *pt3;
-
-  if (i < cvseq_seq_length(o)) {
-    switch (CV_SEQ_ELTYPE(ps->a)) {
-
-    case CV_SEQ_ELTYPE_POINT:
-      pt = CV_GET_SEQ_ELEM(CvPoint, ps->a, (int)i);
-      return Py_BuildValue("ii", pt->x, pt->y);
-
-    case CV_SEQ_ELTYPE_GENERIC:
-      switch (ps->a->elem_size) {
-      case sizeof(CvQuadEdge2D):
-        {
-          cvsubdiv2dedge_t *r = PyObject_NEW(cvsubdiv2dedge_t, &cvsubdiv2dedge_Type);
-          r->a = (CvSubdiv2DEdge)CV_GET_SEQ_ELEM(CvQuadEdge2D, ps->a, (int)i);
-          r->container = ps->container;
-          Py_INCREF(r->container);
-          return (PyObject*)r;
-        }
-      case sizeof(CvConnectedComp):
-        {
-          CvConnectedComp *cc = CV_GET_SEQ_ELEM(CvConnectedComp, ps->a, (int)i);
-          return FROM_CvConnectedComp(*cc);
-        }
-      default:
-        printf("seq elem size is %d\n", ps->a->elem_size);
-        printf("KIND %d\n", CV_SEQ_KIND(ps->a));
-        assert(0);
-      }
-      return PyInt_FromLong(*CV_GET_SEQ_ELEM(unsigned char, ps->a, (int)i));
-
-    case CV_SEQ_ELTYPE_PTR:
-    case CV_SEQ_ELTYPE_INDEX:
-      return PyInt_FromLong(*CV_GET_SEQ_ELEM(int, ps->a, (int)i));
-
-    case CV_32SC4:
-      pp = CV_GET_SEQ_ELEM(pointpair, ps->a, (int)i);
-      return Py_BuildValue("(ii),(ii)", pp->a.x, pp->a.y, pp->b.x, pp->b.y);
-
-    case CV_32FC2:
-      pt2 = CV_GET_SEQ_ELEM(CvPoint2D32f, ps->a, (int)i);
-      return Py_BuildValue("ff", pt2->x, pt2->y);
-
-    case CV_SEQ_ELTYPE_POINT3D:
-      pt3 = CV_GET_SEQ_ELEM(CvPoint3D32f, ps->a, (int)i);
-      return Py_BuildValue("fff", pt3->x, pt3->y, pt3->z);
-
-    default:
-      printf("Unknown element type %08x\n", CV_SEQ_ELTYPE(ps->a));
-      assert(0);
-      return NULL;
-    }
-  } else
-    return NULL;
-}
-
-static PyObject* cvseq_map_getitem(PyObject *o, PyObject *item)
-{
-  if (PyInt_Check(item)) {
-    int i = (int)PyInt_AS_LONG(item);
-    if (i < 0)
-      i += (int)cvseq_seq_length(o);
-    return cvseq_seq_getitem(o, i);
-  } else if (!!PySlice_Check(item)) {
-    Py_ssize_t start, stop, step, slicelength, cur, i;
-    PyObject* result;
-
-    if (PySlice_GetIndicesEx((PySliceObject*)item, cvseq_seq_length(o),
-         &start, &stop, &step, &slicelength) < 0) {
-      return NULL;
-    }
-
-    if (slicelength <= 0) {
-      return PyList_New(0);
-    } else {
-      result = PyList_New(slicelength);
-      if (!result) return NULL;
-
-      for (cur = start, i = 0; i < slicelength;
-           cur += step, i++) {
-        PyList_SET_ITEM(result, i, cvseq_seq_getitem(o, cur));
-      }
-
-      return result;
-    }
-  } else {
-    PyErr_SetString(PyExc_TypeError, "CvSeq indices must be integers");
-    return NULL;
-  }
-}
-
-static
-PySequenceMethods cvseq_sequence = {
-  cvseq_seq_length,
-  NULL,
-  NULL,
-  cvseq_seq_getitem
-};
-
-static PyMappingMethods cvseq_mapping = {
-  cvseq_seq_length,
-  cvseq_map_getitem,
-  NULL,
-};
-
-static PyTypeObject cvseq_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvseq",                          /*name*/
-  sizeof(cvseq_t),                        /*basicsize*/
-};
-
-static void cvseq_specials(void)
-{
-  cvseq_Type.tp_dealloc = cvseq_dealloc;
-  cvseq_Type.tp_as_sequence = &cvseq_sequence;
-  cvseq_Type.tp_as_mapping = &cvseq_mapping;
-  cvseq_Type.tp_methods = cvseq_methods;
-}
-
-#define MK_ACCESSOR(FIELD) \
-static PyObject *cvseq_##FIELD(PyObject *self, PyObject *args) \
-{ \
-  cvseq_t *ps = (cvseq_t*)self; \
-  CvSeq *s = ps->a; \
-  if (s->FIELD == NULL) { \
-    Py_RETURN_NONE; \
-  } else { \
-    cvseq_t *r = PyObject_NEW(cvseq_t, &cvseq_Type); \
-    r->a = s->FIELD; \
-    r->container = ps->container; \
-    Py_INCREF(r->container); \
-    return (PyObject*)r; \
-  } \
-}
-
-MK_ACCESSOR(h_next)
-MK_ACCESSOR(h_prev)
-MK_ACCESSOR(v_next)
-MK_ACCESSOR(v_prev)
-#undef MK_ACCESSOR
-
-/************************************************************************/
-
-/* cvset */
-
-static void cvset_dealloc(PyObject *self)
-{
-  cvset_t *ps = (cvset_t*)self;
-  Py_DECREF(ps->container);
-  PyObject_Del(self);
-}
-
-static PyTypeObject cvset_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvset",                          /*name*/
-  sizeof(cvset_t),                        /*basicsize*/
-};
-
-static PyObject *cvset_iter(PyObject *o)
-{
-  Py_INCREF(o);
-  cvset_t *ps = (cvset_t*)o;
-  ps->i = 0;
-  return o;
-}
-
-static PyObject *cvset_next(PyObject *o)
-{
-  cvset_t *ps = (cvset_t*)o;
-
-  while (ps->i < ps->a->total) {
-    CvSetElem *e = cvGetSetElem(ps->a, ps->i);
-    int prev_i = ps->i++;
-    if (e != NULL) {
-      return cvseq_seq_getitem(o, prev_i);
-    }
-  }
-  return NULL;
-}
-
-static void cvset_specials(void)
-{
-  cvset_Type.tp_dealloc = cvset_dealloc;
-  cvset_Type.tp_iter = cvset_iter;
-  cvset_Type.tp_iternext = cvset_next;
-}
-
-/************************************************************************/
-
-/* cvsubdiv2d */
-
-static PyTypeObject cvsubdiv2d_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                          /*size*/
-  OLD_MODULESTR".cvsubdiv2d",                     /*name*/
-  sizeof(cvsubdiv2d_t),                       /*basicsize*/
-};
-
-static PyObject *cvsubdiv2d_getattro(PyObject *o, PyObject *name)
-{
-  cvsubdiv2d_t *p = (cvsubdiv2d_t*)o;
-  if (strcmp(PyString_AsString(name), "edges") == 0) {
-    cvset_t *r = PyObject_NEW(cvset_t, &cvset_Type);
-    r->a = p->a->edges;
-    r->container = p->container;
-    Py_INCREF(r->container);
-    return (PyObject*)r;
-  } else {
-    PyErr_SetString(PyExc_TypeError, "cvsubdiv2d has no such attribute");
-    return NULL;
-  }
-}
-
-static void cvsubdiv2d_specials(void)
-{
-  cvsubdiv2d_Type.tp_getattro = cvsubdiv2d_getattro;
-}
-
-/************************************************************************/
-
-/* cvsubdiv2dpoint */
-
-static PyTypeObject cvsubdiv2dpoint_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  OLD_MODULESTR".cvsubdiv2dpoint",                     /*name*/
-  sizeof(cvsubdiv2dpoint_t),                       /*basicsize*/
-};
-
-static PyObject *cvsubdiv2dpoint_getattro(PyObject *o, PyObject *name)
-{
-  cvsubdiv2dpoint_t *p = (cvsubdiv2dpoint_t*)o;
-  if (strcmp(PyString_AsString(name), "first") == 0) {
-    cvsubdiv2dedge_t *r = PyObject_NEW(cvsubdiv2dedge_t, &cvsubdiv2dedge_Type);
-    r->a = p->a->first;
-    r->container = p->container;
-    Py_INCREF(r->container);
-    return (PyObject*)r;
-  } else if (strcmp(PyString_AsString(name), "pt") == 0) {
-    return Py_BuildValue("(ff)", p->a->pt.x, p->a->pt.y);
-  } else {
-    PyErr_SetString(PyExc_TypeError, "cvsubdiv2dpoint has no such attribute");
-    return NULL;
-  }
-}
-
-static void cvsubdiv2dpoint_specials(void)
-{
-  cvsubdiv2dpoint_Type.tp_getattro = cvsubdiv2dpoint_getattro;
-}
-
-/************************************************************************/
-/* convert_to_X: used after PyArg_ParseTuple in the generated code  */
-
-/*static int convert_to_PyObjectPTR(PyObject *o, PyObject **dst, const char *name = "no_name")
-{
-  *dst = o;
-  return 1;
-}
-
-static int convert_to_PyCallableObjectPTR(PyObject *o, PyObject **dst, const char *name = "no_name")
-{
-  *dst = o;
-  return 1;
-}*/
-
-static int convert_to_char(PyObject *o, char *dst, const char *name = "no_name")
-{
-  if (PyString_Check(o) && PyString_Size(o) == 1) {
-    *dst = PyString_AsString(o)[0];
-    return 1;
-  } else {
-    (*dst) = 0;
-    return failmsg("Expected single character string for argument '%s'", name);
-  }
-}
-
-static int convert_to_CvMemStorage(PyObject *o, CvMemStorage **dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvmemstorage_Type)) {
-    (*dst) = (((cvmemstorage_t*)o)->a);
-    return 1;
-  } else {
-    (*dst) = (CvMemStorage*)NULL;
-    return failmsg("Expected CvMemStorage for argument '%s'", name);
-  }
-}
-
-static int convert_to_CvSeq(PyObject *o, CvSeq **dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvseq_Type)) {
-    (*dst) = (((cvseq_t*)o)->a);
-    return 1;
-  } else {
-    (*dst) = (CvSeq*)NULL;
-    return failmsg("Expected CvSeq for argument '%s'", name);
-  }
-}
-
-static int convert_to_CvSize(PyObject *o, CvSize *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "ii", &dst->width, &dst->height))
-    return failmsg("CvSize argument '%s' expects two integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvScalar(PyObject *o, CvScalar *s, const char *name = "no_name")
-{
-  if (PySequence_Check(o)) {
-    PyObject *fi = PySequence_Fast(o, name);
-    if (fi == NULL)
-      return 0;
-    if (4 < PySequence_Fast_GET_SIZE(fi))
-        return failmsg("CvScalar value for argument '%s' is longer than 4", name);
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-      PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-      if (PyFloat_Check(item) || PyInt_Check(item)) {
-        s->val[i] = PyFloat_AsDouble(item);
-      } else {
-        return failmsg("CvScalar value for argument '%s' is not numeric", name);
-      }
-    }
-    Py_DECREF(fi);
-  } else {
-    if (PyFloat_Check(o) || PyInt_Check(o)) {
-      s->val[0] = PyFloat_AsDouble(o);
-    } else {
-      return failmsg("CvScalar value for argument '%s' is not numeric", name);
-    }
-  }
-  return 1;
-}
-
-static int convert_to_CvPointPTR(PyObject *o, CvPoint **p, const char *name = "no_name")
-{
-  if (!PySequence_Check(o))
-    return failmsg("Expected sequence for point list argument '%s'", name);
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  *p = new CvPoint[PySequence_Fast_GET_SIZE(fi)];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!PyTuple_Check(item))
-      return failmsg("Expected tuple for element in point list argument '%s'", name);
-    if (!PyArg_ParseTuple(item, "ii", &((*p)[i].x), &((*p)[i].y))) {
-      return 0;
-    }
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-static int convert_to_CvPoint2D32fPTR(PyObject *o, CvPoint2D32f **p, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  *p = new CvPoint2D32f[PySequence_Fast_GET_SIZE(fi)];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!PyTuple_Check(item))
-      return failmsg("Expected tuple for CvPoint2D32f argument '%s'", name);
-    if (!PyArg_ParseTuple(item, "ff", &((*p)[i].x), &((*p)[i].y))) {
-      return 0;
-    }
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-#if 0 // not used
-static int convert_to_CvPoint3D32fPTR(PyObject *o, CvPoint3D32f **p, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  *p = new CvPoint3D32f[PySequence_Fast_GET_SIZE(fi)];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!PyTuple_Check(item))
-      return failmsg("Expected tuple for CvPoint3D32f argument '%s'", name);
-    if (!PyArg_ParseTuple(item, "fff", &((*p)[i].x), &((*p)[i].y), &((*p)[i].z))) {
-      return 0;
-    }
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-#endif
-
-static int convert_to_CvStarDetectorParams(PyObject *o, CvStarDetectorParams *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o,
-                        "iiiii",
-                        &dst->maxSize,
-                        &dst->responseThreshold,
-                        &dst->lineThresholdProjected,
-                        &dst->lineThresholdBinarized,
-                        &dst->suppressNonmaxSize))
-    return failmsg("CvRect argument '%s' expects four integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvRect(PyObject *o, CvRect *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "iiii", &dst->x, &dst->y, &dst->width, &dst->height))
-    return failmsg("CvRect argument '%s' expects four integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvRectPTR(PyObject *o, CvRect **dst, const char *name = "no_name")
-{
-  *dst = new CvRect;
-  if (!PyArg_ParseTuple(o, "iiii", &(*dst)->x, &(*dst)->y, &(*dst)->width, &(*dst)->height))
-    return failmsg("CvRect argument '%s' expects four integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvSlice(PyObject *o, CvSlice *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "ii", &dst->start_index, &dst->end_index))
-    return failmsg("CvSlice argument '%s' expects two integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvPoint(PyObject *o, CvPoint *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "ii", &dst->x, &dst->y))
-    return failmsg("CvPoint argument '%s' expects two integers", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvPoint2D32f(PyObject *o, CvPoint2D32f *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "ff", &dst->x, &dst->y))
-    return failmsg("CvPoint2D32f argument '%s' expects two floats", name);
-  else
-    return 1;
-}
-
-static int convert_to_CvPoint3D32f(PyObject *o, CvPoint3D32f *dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "fff", &dst->x, &dst->y, &dst->z))
-    return failmsg("CvPoint3D32f argument '%s' expects three floats", name);
-  else
-    return 1;
-}
-
-static int convert_to_IplImage(PyObject *o, IplImage **dst, const char *name)
-{
-  iplimage_t *ipl = (iplimage_t*)o;
-  void *buffer;
-  Py_ssize_t buffer_len;
-
-  if (!is_iplimage(o)) {
-    return failmsg("Argument '%s' must be IplImage", name);
-  } else if (PyString_Check(ipl->data)) {
-    cvSetData(ipl->a, PyString_AsString(ipl->data) + ipl->offset, ipl->a->widthStep);
-    assert(cvGetErrStatus() == 0);
-    *dst = ipl->a;
-    return 1;
-  } else if (ipl->data && PyObject_AsWriteBuffer(ipl->data, &buffer, &buffer_len) == 0) {
-    cvSetData(ipl->a, (void*)((char*)buffer + ipl->offset), ipl->a->widthStep);
-    assert(cvGetErrStatus() == 0);
-    *dst = ipl->a;
-    return 1;
-  } else {
-    return failmsg("IplImage argument '%s' has no data", name);
-  }
-}
-
-static int convert_to_CvMat(PyObject *o, CvMat **dst, const char *name)
-{
-  cvmat_t *m = (cvmat_t*)o;
-  void *buffer;
-  Py_ssize_t buffer_len;
-
-  if (!is_cvmat(o)) {
-//#if !PYTHON_USE_NUMPY
-    return failmsg("Argument '%s' must be CvMat. Use fromarray() to convert numpy arrays to CvMat", name);
-/*#else
-    PyObject *asmat = fromarray(o, 0);
-    if (asmat == NULL)
-      return failmsg("Argument '%s' must be CvMat", name);
-    // now have the array obect as a cvmat, can use regular conversion
-    return convert_to_CvMat(asmat, dst, name);
-#endif*/
-  } else {
-    m->a->refcount = NULL;
-    if (m->data && PyString_Check(m->data)) {
-      assert(cvGetErrStatus() == 0);
-      char *ptr = PyString_AsString(m->data) + m->offset;
-      cvSetData(m->a, ptr, m->a->step);
-      assert(cvGetErrStatus() == 0);
-      *dst = m->a;
-      return 1;
-    } else if (m->data && PyObject_AsWriteBuffer(m->data, &buffer, &buffer_len) == 0) {
-      cvSetData(m->a, (void*)((char*)buffer + m->offset), m->a->step);
-      assert(cvGetErrStatus() == 0);
-      *dst = m->a;
-      return 1;
-    } else if (m->data && m->a->data.ptr){
-      *dst = m->a;
-      return 1;
-    }
-    else {
-      return failmsg("CvMat argument '%s' has no data", name);
-    }
-  }
-}
-
-static int convert_to_CvMatND(PyObject *o, CvMatND **dst, const char *name)
-{
-  cvmatnd_t *m = (cvmatnd_t*)o;
-  void *buffer;
-  Py_ssize_t buffer_len;
-
-  if (!is_cvmatnd(o)) {
-    return failmsg("Argument '%s' must be CvMatND", name);
-  } else if (m->data && PyString_Check(m->data)) {
-    m->a->data.ptr = ((uchar*)PyString_AsString(m->data)) + m->offset;
-    *dst = m->a;
-    return 1;
-  } else if (m->data && PyObject_AsWriteBuffer(m->data, &buffer, &buffer_len) == 0) {
-    m->a->data.ptr = ((uchar*)buffer + m->offset);
-    *dst = m->a;
-    return 1;
-  } else {
-    return failmsg("CvMatND argument '%s' has no data", name);
-  }
-}
-
-static int convert_to_CvArr(PyObject *o, CvArr **dst, const char *name)
-{
-  if (o == Py_None) {
-    *dst = (void*)NULL;
-    return 1;
-  } else if (is_iplimage(o)) {
-    return convert_to_IplImage(o, (IplImage**)dst, name);
-  } else if (is_cvmat(o)) {
-    return convert_to_CvMat(o, (CvMat**)dst, name);
-  } else if (is_cvmatnd(o)) {
-    return convert_to_CvMatND(o, (CvMatND**)dst, name);
-  } else {
-//#if !PYTHON_USE_NUMPY
-    return failmsg("CvArr argument '%s' must be IplImage, CvMat or CvMatND. Use fromarray() to convert numpy arrays to CvMat or cvMatND", name);
-/*#else
-    PyObject *asmat = fromarray(o, 0);
-    if (asmat == NULL)
-      return failmsg("CvArr argument '%s' must be IplImage, CvMat, CvMatND, or support the array interface", name);
-    // now have the array obect as a cvmat, can use regular conversion
-    return convert_to_CvArr(asmat, dst, name);
-#endif*/
-  }
-}
-
-static int convert_to_CvHistogram(PyObject *o, CvHistogram **dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvhistogram_Type)) {
-    cvhistogram_t *ht = (cvhistogram_t*)o;
-    *dst = &ht->h;
-    return convert_to_CvArr(ht->bins, &(ht->h.bins), "bins");
-  } else {
-    *dst = (CvHistogram *)NULL;
-    return failmsg("Expected CvHistogram for argument '%s'", name);
-  }
-}
-
-// Used by FillPoly, FillConvexPoly, PolyLine
-struct pts_npts_contours {
-  CvPoint** pts;
-  int* npts;
-  int contours;
-};
-
-static int convert_to_pts_npts_contours(PyObject *o, pts_npts_contours *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->contours = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->pts = new CvPoint*[dst->contours];
-  dst->npts = new int[dst->contours];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    if (!convert_to_CvPointPTR(PySequence_Fast_GET_ITEM(fi, i), &dst->pts[i], name))
-      return 0;
-    dst->npts[i] = (int)PySequence_Size(PySequence_Fast_GET_ITEM(fi, i)); // safe because convert_ just succeeded
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-class cvarrseq {
-public:
-  union {
-    CvSeq *seq;
-    CvArr *mat;
-  };
-  int freemat;
-  cvarrseq() {
-    freemat = false;
-  }
-  ~cvarrseq() {
-    if (freemat) {
-      cvReleaseMat((CvMat**)&mat);
-    }
-  }
-};
-
-static int is_convertible_to_mat(PyObject *o)
-{
-#if PYTHON_USE_NUMPY
-  if (PyObject_HasAttrString(o, "__array_struct__")) {
-    PyObject *ao = PyObject_GetAttrString(o, "__array_struct__");
-    if (ao != NULL &&
-        PyCObject_Check(ao) &&
-        ((PyArrayInterface*)PyCObject_AsVoidPtr(ao))->two == 2) {
-      return 1;
-    }
-  }
-#endif
-  return is_iplimage(o) && is_cvmat(o) && is_cvmatnd(o);
-}
-
-static int convert_to_cvarrseq(PyObject *o, cvarrseq *dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvseq_Type)) {
-    return convert_to_CvSeq(o, &(dst->seq), name);
-  } else if (is_convertible_to_mat(o)) {
-    int r = convert_to_CvArr(o, &(dst->mat), name);
-    return r;
-  } else if (PySequence_Check(o)) {
-    PyObject *fi = PySequence_Fast(o, name);
-    if (fi == NULL)
-      return 0;
-    Py_ssize_t size = -1;
-    // Make a pass through the sequence, checking that each element is
-    // a sequence and that they are all the same size
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-      PyObject *e = PySequence_Fast_GET_ITEM(fi, i);
-
-      if (!PySequence_Check(e))
-        return failmsg("Sequence '%s' must contain sequences", name);
-      if (i == 0)
-        size = (int)PySequence_Size(e);
-      else if (size != PySequence_Size(e))
-        return failmsg("All elements of sequence '%s' must be same size", name);
-    }
-    assert(size != -1);
-    CvMat *mt = cvCreateMat((int)PySequence_Fast_GET_SIZE(fi), 1, CV_32SC((int)size));
-    dst->freemat = true; // dealloc this mat when done
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-      PyObject *e = PySequence_Fast_GET_ITEM(fi, i);
-      PyObject *fe = PySequence_Fast(e, name);
-      assert(fe != NULL);
-      int *pdst = (int*)cvPtr2D(mt, (int)i, 0);
-      for (Py_ssize_t j = 0; j < size; j++) {
-        PyObject *num = PySequence_Fast_GET_ITEM(fe, j);
-        if (!PyNumber_Check(num)) {
-          return failmsg("Sequence must contain numbers", name);
-        }
-        *pdst++ = PyInt_AsLong(num);
-      }
-      Py_DECREF(fe);
-    }
-    Py_DECREF(fi);
-    dst->mat = mt;
-    return 1;
-  } else {
-    return failmsg("Argument '%s' must be CvSeq, CvArr, or a sequence of numbers");
-  }
-}
-
-struct cvarr_count {
-  CvArr **cvarr;
-  int count;
-};
-
-static int convert_to_cvarr_count(PyObject *o, cvarr_count *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->cvarr = new CvArr*[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    if (!convert_to_CvArr(PySequence_Fast_GET_ITEM(fi, i), &dst->cvarr[i], name))
-      return 0;
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct intpair
-{
-  int *pairs;
-  int count;
-};
-
-static int convert_to_intpair(PyObject *o, intpair *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->pairs = new int[2 * dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!PyArg_ParseTuple(item, "ii", &dst->pairs[2 * i], &dst->pairs[2 * i + 1])) {
-      return 0;
-    }
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct cvpoint2d32f_count {
-  CvPoint2D32f* points;
-  int count;
-};
-
-static int convert_to_cvpoint2d32f_count(PyObject *o, cvpoint2d32f_count *dst, const char *name = "no_name")
-{
-  if (PyInt_Check(o)) {
-    dst->count = PyInt_AsLong(o);
-    dst->points = new CvPoint2D32f[dst->count];
-    return 1;
-  } else {
-    return failmsg("Expected integer for CvPoint2D32f count");
-  }
-}
-
-struct floats {
-  float *f;
-  int count;
-};
-static int convert_to_floats(PyObject *o, floats *dst, const char *name = "no_name")
-{
-  if (PySequence_Check(o)) {
-    PyObject *fi = PySequence_Fast(o, name);
-    if (fi == NULL)
-      return 0;
-    dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-    dst->f = new float[dst->count];
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-      PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-      dst->f[i] = (float)PyFloat_AsDouble(item);
-    }
-    Py_DECREF(fi);
-  } else if (PyNumber_Check(o)) {
-    dst->count = 1;
-    dst->f = new float[1];
-    dst->f[0] = (float)PyFloat_AsDouble(o);
-  } else {
-    return failmsg("Expected list of floats, or float for argument '%s'", name);
-  }
-  return 1;
-}
-
-struct chars {
-  char *f;
-  int count;
-};
-/// convert_to_chars not used
-
-struct CvPoints {
-  CvPoint *p;
-  int count;
-};
-static int convert_to_CvPoints(PyObject *o, CvPoints *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->p = new CvPoint[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    convert_to_CvPoint(item, &dst->p[i], name);
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct CvPoint3D32fs {
-  CvPoint3D32f *p;
-  int count;
-};
-static int convert_to_CvPoint3D32fs(PyObject *o, CvPoint3D32fs *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->p = new CvPoint3D32f[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    convert_to_CvPoint3D32f(item, &dst->p[i], name);
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct CvPoint2D32fs {
-  CvPoint2D32f *p;
-  int count;
-};
-static int convert_to_CvPoint2D32fs(PyObject *o, CvPoint2D32fs *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->p = new CvPoint2D32f[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    convert_to_CvPoint2D32f(item, &dst->p[i], name);
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct ints {
-  int *i;
-  int count;
-};
-static int convert_to_ints(PyObject *o, ints *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->i = new int[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    dst->i[i] = PyInt_AsLong(item);
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct ints0 {
-  int *i;
-  int count;
-};
-static int convert_to_ints0(PyObject *o, ints0 *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->i = new int[dst->count + 1];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    dst->i[i] = PyInt_AsLong(item);
-  }
-  dst->i[dst->count] = 0;
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct dims
-{
-  int count;
-  int i[CV_MAX_DIM];
-  int step[CV_MAX_DIM];
-  int length[CV_MAX_DIM];
-};
-
-static int convert_to_dim(PyObject *item, int i, dims *dst, CvArr *cva, const char *name = "no_name")
-{
-  if (!!PySlice_Check(item)) {
-    Py_ssize_t start, stop, step, slicelength;
-    PySlice_GetIndicesEx((PySliceObject*)item, cvGetDimSize(cva, i), &start, &stop, &step, &slicelength);
-    dst->i[i] = (int)start;
-    dst->step[i] = (int)step;
-    dst->length[i] = (int)slicelength;
-  } else {
-    int index = PyInt_AsLong(item);
-    if (0 <= index)
-      dst->i[i] = index;
-    else
-      dst->i[i] = cvGetDimSize(cva, i) + index;
-    dst->step[i] = 0;
-    dst->length[i] = 1;
-  }
-  return 1;
-}
-
-static int convert_to_dims(PyObject *o, dims *dst, CvArr *cva, const char *name = "no_name")
-{
-  if (!PyTuple_Check(o)) {
-    dst->count = 1;
-    return convert_to_dim(o, 0, dst, cva, name);
-  } else {
-    PyObject *fi = PySequence_Fast(o, name);
-    if (fi == NULL) {
-      PyErr_SetString(PyExc_TypeError, "Expected tuple for index");
-      return 0;
-    }
-    dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-      if (i >= cvGetDims(cva)) {
-        return failmsg("Access specifies %d dimensions, but array only has %d", PySequence_Fast_GET_SIZE(fi), cvGetDims(cva));
-      }
-      PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-      if (!convert_to_dim(item, (int)i, dst, cva, name))
-        return 0;
-    }
-    Py_DECREF(fi);
-    return 1;
-  }
-}
-
-struct IplImages {
-  IplImage **ims;
-  int count;
-};
-static int convert_to_IplImages(PyObject *o, IplImages *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->ims = new IplImage*[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!convert_to_IplImage(item, &dst->ims[i]))
-      return 0;
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-struct CvArrs {
-  CvArr **ims;
-  int count;
-};
-static int convert_to_CvArrs(PyObject *o, CvArrs *dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  dst->count = (int)PySequence_Fast_GET_SIZE(fi);
-  dst->ims = new CvArr*[dst->count];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    if (!convert_to_CvArr(item, &dst->ims[i]))
-      return 0;
-  }
-  Py_DECREF(fi);
-  return 1;
-}
-
-/*static int convert_to_floatPTRPTR(PyObject *o, float*** dst, const char *name = "no_name")
-{
-  PyObject *fi = PySequence_Fast(o, name);
-  if (fi == NULL)
-    return 0;
-  Py_ssize_t sz = (int)PySequence_Fast_GET_SIZE(fi);
-  float **r = new float*[sz];
-  for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++) {
-    PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-    floats ff;
-    if (!convert_to_floats(item, &ff))
-      return 0;
-    r[i] = ff.f;
-  }
-  *dst = r;
-  return 1;
-}*/
-
-static int convert_to_CvFontPTR(PyObject *o, CvFont** dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvfont_Type)) {
-    (*dst) = &(((cvfont_t*)o)->a);
-    return 1;
-  } else {
-    (*dst) = (CvFont*)NULL;
-    return failmsg("Expected CvFont for argument '%s'", name);
-  }
-}
-
-/*static int convert_to_CvContourTreePTR(PyObject *o, CvContourTree** dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvcontourtree_Type)) {
-    (*dst) = ((cvcontourtree_t*)o)->a;
-    return 1;
-  } else {
-    (*dst) = NULL;
-    return failmsg("Expected CvContourTree for argument '%s'", name);
-  }
-}*/
-
-static int convert_to_CvRNGPTR(PyObject *o, CvRNG** dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvrng_Type)) {
-    (*dst) = &(((cvrng_t*)o)->a);
-    return 1;
-  } else {
-    (*dst) = (CvRNG*)NULL;
-    return failmsg("Expected CvRNG for argument '%s'", name);
-  }
-}
-
-typedef void* generic;
-static int convert_to_generic(PyObject *o, generic *dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &iplimage_Type))
-    return convert_to_IplImage(o, (IplImage**)dst, name);
-  else if (PyType_IsSubtype(o->ob_type, &cvmat_Type))
-    return convert_to_CvMat(o, (CvMat**)dst, name);
-  else if (PyType_IsSubtype(o->ob_type, &cvmatnd_Type))
-    return convert_to_CvMatND(o, (CvMatND**)dst, name);
-  else {
-    return failmsg("Cannot identify type of '%s'", name);
-  }
-}
-
-static int convert_to_CvTermCriteria(PyObject *o, CvTermCriteria* dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "iid", &dst->type, &dst->max_iter, &dst->epsilon))
-    return 0;
-  return 1;
-}
-
-static int convert_to_CvBox2D(PyObject *o, CvBox2D* dst, const char *name = "no_name")
-{
-  if (!PyArg_ParseTuple(o, "(ff)(ff)f", &dst->center.x, &dst->center.y, &dst->size.width, &dst->size.height, &dst->angle))
-    return 0;
-  return 1;
-}
-
-static int convert_to_CvSubdiv2DPTR(PyObject *o, CvSubdiv2D** dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvsubdiv2d_Type)) {
-    (*dst) = (((cvsubdiv2d_t*)o)->a);
-    return 1;
-  } else {
-    (*dst) = (CvSubdiv2D*)NULL;
-    return failmsg("Expected CvSubdiv2D for argument '%s'", name);
-  }
-}
-
-static int convert_to_CvNextEdgeType(PyObject *o, CvNextEdgeType *dst, const char *name = "no_name")
-{
-  if (!PyInt_Check(o)) {
-    *dst = (CvNextEdgeType)NULL;
-    return failmsg("Expected number for CvNextEdgeType argument '%s'", name);
-  } else {
-    *dst = (CvNextEdgeType)PyInt_AsLong(o);
-    return 1;
-  }
-}
-
-static int convert_to_CvSubdiv2DEdge(PyObject *o, CvSubdiv2DEdge *dst, const char *name = "no_name")
-{
-  if (PyType_IsSubtype(o->ob_type, &cvsubdiv2dedge_Type)) {
-    (*dst) = (((cvsubdiv2dedge_t*)o)->a);
-    return 1;
-  } else {
-    *dst = 0L;
-    return failmsg("Expected CvSubdiv2DEdge for argument '%s'", name);
-  }
-}
-
-/************************************************************************/
-
-static PyObject *pythonize_CvMat(cvmat_t *m)
-{
-  // Need to make this CvMat look like any other, with a Python
-  // buffer object as its data.
-  CvMat *mat = m->a;
-  assert(mat->step != 0);
-#if 0
-  PyObject *data = PyString_FromStringAndSize((char*)(mat->data.ptr), mat->rows * mat->step);
-#else
-  memtrack_t *o = PyObject_NEW(memtrack_t, &memtrack_Type);
-  size_t gap = mat->data.ptr - (uchar*)mat->refcount;
-  o->ptr = mat->refcount;
-  o->owner = __LINE__;
-  o->freeptr = true;
-  o->size = gap + mat->rows * mat->step;
-  o->backing = NULL;
-  o->backingmat = NULL;
-  PyObject *data = PyBuffer_FromReadWriteObject((PyObject*)o, (size_t)gap, mat->rows * mat->step);
-  if (data == NULL)
-    return NULL;
-#endif
-  m->data = data;
-  m->offset = 0;
-  Py_DECREF(o);
-
-  // Now m has a reference to data, which has a reference to o.
-
-  return (PyObject*)m;
-}
-
-static PyObject *pythonize_IplImage(iplimage_t *cva)
-{
-  // Need to make this iplimage look like any other, with a Python
-  // string as its data.
-  // So copy the image data into a Python string object, then release
-  // it.
-
-  IplImage *ipl = (IplImage*)(cva->a);
-  // PyObject *data = PyString_FromStringAndSize(ipl->imageData, ipl->imageSize);
-
-  memtrack_t *o = PyObject_NEW(memtrack_t, &memtrack_Type);
-  assert(ipl->imageDataOrigin == ipl->imageData);
-  o->ptr = ipl->imageDataOrigin;
-  o->owner = __LINE__;
-  o->freeptr = true;
-  o->size = ipl->height * ipl->widthStep;
-  o->backing = NULL;
-  o->backingmat = NULL;
-  PyObject *data = PyBuffer_FromReadWriteObject((PyObject*)o, (size_t)0, o->size);
-  if (data == NULL)
-    return NULL;
-  Py_DECREF(o);
-  cva->data = data;
-  cva->offset = 0;
-
-  return (PyObject*)cva;
-}
-
-static PyObject *pythonize_CvMatND(cvmatnd_t *m, PyObject *backing = NULL)
-{
-  //
-  // Need to make this CvMatND look like any other, with a Python
-  // buffer object as its data.
-  //
-
-  CvMatND *mat = m->a;
-  assert(mat->dim[0].step != 0);
-#if 0
-  PyObject *data = PyString_FromStringAndSize((char*)(mat->data.ptr), mat->dim[0].size * mat->dim[0].step);
-#else
-  memtrack_t *o = PyObject_NEW(memtrack_t, &memtrack_Type);
-  o->ptr = mat->data.ptr;
-  o->owner = __LINE__;
-  o->freeptr = false;
-  o->size = cvmatnd_size(mat);
-  Py_XINCREF(backing);
-  o->backing = backing;
-  o->backingmat = mat;
-  PyObject *data = PyBuffer_FromReadWriteObject((PyObject*)o, (size_t)0, o->size);
-  Py_DECREF(o); // Now 'data' holds the only reference to 'o'
-  if (data == NULL)
-    return NULL;
-#endif
-  m->data = data;
-  m->offset = 0;
-
-  return (PyObject*)m;
-}
-
-/************************************************************************/
-/* FROM_xxx:   C -> Python converters.
- *
- * Turn various OpenCV types (and some aggregate types above)
- * into Python objects.  Used by the generated code.
- *
- * All these functions and macros return a new reference.
- */
-
-
-static PyObject *_FROM_CvSeqPTR(CvSeq *s, PyObject *storage)
-{
-  cvseq_t *ps = PyObject_NEW(cvseq_t, &cvseq_Type);
-  ps->a = s;
-  ps->container = storage;
-  Py_INCREF(ps->container);
-  return (PyObject*)ps;
-}
-
-static PyObject *_FROM_CvSubdiv2DPTR(CvSubdiv2D *s, PyObject *storage)
-{
-  cvsubdiv2d_t *ps = PyObject_NEW(cvsubdiv2d_t, &cvsubdiv2d_Type);
-  ps->a = s;
-  ps->container = storage;
-  Py_INCREF(ps->container);
-  return (PyObject*)ps;
-}
-
-static PyObject *FROM_floats(floats r)
-{
-  PyObject *pr;
-
-  pr = PyList_New(r.count);
-  for (Py_ssize_t i = 0; i < (Py_ssize_t)r.count; i++) {
-    PyList_SetItem(pr, i, PyFloat_FromDouble(r.f[i]));
-  }
-  return pr;
-}
-
-static PyObject *FROM_chars(chars r)
-{
-  PyObject *pr;
-
-  pr = PyList_New(r.count);
-  for (Py_ssize_t i = 0; i < (Py_ssize_t)r.count; i++) {
-    PyList_SetItem(pr, i, PyInt_FromLong(r.f[i]));
-  }
-  return pr;
-}
-
-static PyObject *FROM_cvpoint2d32f_count(cvpoint2d32f_count r)
-{
-  PyObject *pr;
-
-  pr = PyList_New(r.count);
-  for (Py_ssize_t i = 0; i < (Py_ssize_t)r.count; i++) {
-    PyList_SetItem(pr, i, FROM_CvPoint2D32f(r.points[i]));
-  }
-  return pr;
-}
-
-static PyObject *FROM_CvPoint2D32fs(CvPoint2D32fs r)
-{
-  PyObject *pr;
-
-  pr = PyList_New(r.count);
-  for (Py_ssize_t i = 0; i < (Py_ssize_t)r.count; i++) {
-    PyList_SetItem(pr, i, FROM_CvPoint2D32f(r.p[i]));
-  }
-  return pr;
-}
-
-typedef CvSeq CvSeqOfCvConvexityDefect;
-static PyObject *FROM_CvSeqOfCvConvexityDefectPTR(CvSeqOfCvConvexityDefect *r)
-{
-  PyObject *pr;
-  pr = PyList_New(r->total);
-  for (int i = 0; i < r->total; i++) {
-    CvConvexityDefect *pd = CV_GET_SEQ_ELEM(CvConvexityDefect, r, i);
-    PyList_SetItem(pr, i, Py_BuildValue("(ii)(ii)(ii)f",
-                                        pd->start->x, pd->start->y,
-                                        pd->end->x, pd->end->y,
-                                        pd->depth_point->x, pd->depth_point->y,
-                                        pd->depth));
-  }
-  // This function has copied the CvSeq data into a list.  Hence the
-  // CvSeq is not being returned to the caller.  Hence, no reference
-  // count increase for the storage, unlike _FROM_CvSeqPTR.
-  return pr;
-}
-
-typedef CvSeq CvSeqOfCvAvgComp;
-static PyObject *FROM_CvSeqOfCvAvgCompPTR(CvSeqOfCvAvgComp *r)
-{
-  PyObject *pr;
-  pr = PyList_New(r->total);
-  for (int i = 0; i < r->total; i++) {
-    CvAvgComp *pd = CV_GET_SEQ_ELEM(CvAvgComp, r, i);
-    PyList_SetItem(pr, i, Py_BuildValue("(iiii)i",
-                                        pd->rect.x, pd->rect.y,
-                                        pd->rect.width, pd->rect.height,
-                                        pd->neighbors));
-  }
-  // This function has copied the CvSeq data into a list.  Hence the
-  // CvSeq is not being returned to the caller.  Hence, no reference
-  // count increase for the storage, unlike _FROM_CvSeqPTR.
-  return pr;
-}
-
-typedef CvSeq CvSeqOfCvStarKeypoint;
-static PyObject *FROM_CvSeqOfCvStarKeypointPTR(CvSeqOfCvStarKeypoint *r)
-{
-  PyObject *pr;
-  pr = PyList_New(r->total);
-  for (int i = 0; i < r->total; i++) {
-    CvStarKeypoint *pd = CV_GET_SEQ_ELEM(CvStarKeypoint, r, i);
-    PyList_SetItem(pr, i, Py_BuildValue("(ii)if",
-                                        pd->pt.x, pd->pt.y,
-                                        pd->size,
-                                        pd->response));
-  }
-  // This function has copied the CvSeq data into a list.  Hence the
-  // CvSeq is not being returned to the caller.  Hence, no reference
-  // count increase for the storage, unlike _FROM_CvSeqPTR.
-  return pr;
-}
-
-typedef CvSeq CvSeqOfCvSURFPoint;
-static PyObject *FROM_CvSeqOfCvSURFPointPTR(CvSeqOfCvSURFPoint *r)
-{
-  PyObject *pr;
-  pr = PyList_New(r->total);
-  for (int i = 0; i < r->total; i++) {
-    CvSURFPoint *pd = CV_GET_SEQ_ELEM(CvSURFPoint, r, i);
-    PyList_SetItem(pr, i, Py_BuildValue("(ff)iiff",
-                                        pd->pt.x, pd->pt.y,
-                                        pd->laplacian,
-                                        pd->size,
-                                        pd->dir,
-                                        pd->hessian));
-  }
-  // This function has copied the CvSeq data into a list.  Hence the
-  // CvSeq is not being returned to the caller.  Hence, no reference
-  // count increase for the storage, unlike _FROM_CvSeqPTR.
-  return pr;
-}
-
-typedef CvSeq CvSeqOfCvSURFDescriptor;
-static PyObject *FROM_CvSeqOfCvSURFDescriptorPTR(CvSeqOfCvSURFDescriptor *r)
-{
-  PyObject *pr;
-  pr = PyList_New(r->total);
-  for (int i = 0; i < r->total; i++) {
-    float *pd = (float*)cvGetSeqElem(r, i);
-    int count = r->elem_size / sizeof(float);
-    PyObject *oi = PyList_New(count);
-    for (int j = 0; j < count; j++) {
-      PyList_SetItem(oi, j, PyFloat_FromDouble(pd[j]));
-    }
-    PyList_SetItem(pr, i, oi);
-  }
-  // This function has copied the CvSeq data into a list.  Hence the
-  // CvSeq is not being returned to the caller.  Hence, no reference
-  // count increase for the storage, unlike _FROM_CvSeqPTR.
-  return pr;
-}
-
-typedef CvPoint2D32f CvPoint2D32f_4[4];
-static PyObject *FROM_CvPoint2D32f_4(CvPoint2D32f* r)
-{
-  return Py_BuildValue("(ff)(ff)(ff)(ff)",
-                       r[0].x, r[0].y,
-                       r[1].x, r[1].y,
-                       r[2].x, r[2].y,
-                       r[3].x, r[3].y);
-}
-
-typedef float CvMatr32f_i[9];
-
-static PyObject *FROM_CvMatr32f_i(CvMatr32f_i r)
-{
-  return Py_BuildValue("(fff)(fff)(fff)",
-    r[0], r[1], r[2],
-    r[3], r[4], r[5],
-    r[6], r[7], r[8]);
-}
-
-typedef float CvVect32f_i[3];
-static PyObject *FROM_CvVect32f_i(CvVect32f_i r)
-{
-  return Py_BuildValue("fff",
-    r[0], r[1], r[2]);
-}
-
-static PyObject *FROM_CvFont(CvFont r)
-{
-  cvfont_t *cf = PyObject_NEW(cvfont_t, &cvfont_Type);
-  cf->a = r;
-  return (PyObject*)cf;
-}
-
-static PyObject *FROM_CvSubdiv2DPointPTR(CvSubdiv2DPoint* r)
-{
-  if (r != NULL) {
-    cvsubdiv2dpoint_t *cf = PyObject_NEW(cvsubdiv2dpoint_t, &cvsubdiv2dpoint_Type);
-    cf->a = r;
-    return (PyObject*)cf;
-  } else {
-    Py_INCREF(Py_None);
-    return Py_None;
-  }
-}
-
-static PyObject *FROM_IplImagePTR(IplImage *r)
-{
-  iplimage_t *cva = PyObject_NEW(iplimage_t, &iplimage_Type);
-  cva->a = r;
-  return pythonize_IplImage(cva);
-}
-
-static PyObject *FROM_ROIplImagePTR(ROIplImage *r)
-{
-  if (r != NULL) {
-    iplimage_t *cva = PyObject_NEW(iplimage_t, &iplimage_Type);
-    cva->a = cvCreateImageHeader(cvSize(100,100), 8, 1);
-    *(cva->a) = *r;
-    cva->data = PyBuffer_FromReadWriteMemory(r->imageData, r->height * r->widthStep);
-    cva->offset = 0;
-    return (PyObject*)cva;
-  } else {
-    Py_RETURN_NONE;
-  }
-}
-
-static PyObject *FROM_ROCvMatPTR(ROCvMat *r)
-{
-  if (r != NULL) {
-    cvmat_t *cva = PyObject_NEW(cvmat_t, &cvmat_Type);
-    cva->a = cvCreateMatHeader(100, 100, CV_8U);
-    *(cva->a) = *r;
-    cva->data = PyBuffer_FromReadWriteMemory(r->data.ptr, r->rows * r->step);
-    cva->offset = 0;
-    return (PyObject*)cva;
-  } else {
-    Py_RETURN_NONE;
-  }
-}
-
-static PyObject *FROM_CvMatPTR(CvMat *r)
-{
-  cvmat_t *cvm = PyObject_NEW(cvmat_t, &cvmat_Type);
-  cvm->a = r;
-
-  return pythonize_CvMat(cvm);
-}
-
-static PyObject *FROM_CvMat(CvMat *r)
-{
-  cvmat_t *m = PyObject_NEW(cvmat_t, &cvmat_Type);
-  m->a = r;
-  return pythonize_CvMat(m);
-}
-
-static PyObject *FROM_CvMatNDPTR(CvMatND *r)
-{
-  cvmatnd_t *m = PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-  m->a = r;
-  return pythonize_CvMatND(m);
-}
-
-static PyObject *FROM_CvRNG(CvRNG r)
-{
-  cvrng_t *m = PyObject_NEW(cvrng_t, &cvrng_Type);
-  m->a = r;
-  return (PyObject*)m;
-}
-
-/*static PyObject *FROM_CvContourTreePTR(CvContourTree *r)
-{
-  cvcontourtree_t *m = PyObject_NEW(cvcontourtree_t, &cvcontourtree_Type);
-  m->a = r;
-  return (PyObject*)m;
-}*/
-
-static PyObject *FROM_generic(generic r)
-{
-  if (r == NULL) {
-    failmsg("OpenCV returned NULL");
-    return NULL;
-  }
-  CvTypeInfo* t = cvTypeOf(r);
-  if (strcmp(t->type_name, "opencv-image") == 0)
-    return FROM_IplImagePTR((IplImage*)r);
-  else if (strcmp(t->type_name, "opencv-matrix") == 0)
-    return FROM_CvMat((CvMat*)r);
-  else if (strcmp(t->type_name, "opencv-nd-matrix") == 0)
-    return FROM_CvMatNDPTR((CvMatND*)r);
-  else if (strcmp(t->type_name, "opencv-haar-classifier") == 0)
-    return FROM_CvHaarClassifierCascadePTR((CvHaarClassifierCascade*)r);
-  else {
-    failmsg("Unknown OpenCV type '%s'", t->type_name);
-    return NULL;
-  }
-}
-
-static PyObject *FROM_CvSubdiv2DEdge(CvSubdiv2DEdge r)
-{
-  cvsubdiv2dedge_t *m = PyObject_NEW(cvsubdiv2dedge_t, &cvsubdiv2dedge_Type);
-  m->a = r;
-  m->container = Py_None; // XXX
-  Py_INCREF(m->container);
-  return (PyObject*)m;
-}
-
-static PyObject *FROM_CvPoints(CvPoints src)
-{
-  PyObject *pr;
-  pr = PyList_New(src.count);
-  for (int i = 0; i < src.count; i++) {
-    PyList_SetItem(pr, i, FROM_CvPoint(src.p[i]));
-  }
-  return pr;
-}
-
-/************************************************************************/
-
-/* A few functions are too odd to be generated,
- * so are handwritten here */
-
-static PyObject *pycvWaitKey(PyObject *self, PyObject *args, PyObject *kw)
-{
-  int delay = 0;
-
-  const char *keywords[] = { "delay", NULL };
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "|i", (char**)keywords, &delay))
-    return NULL;
-  int r;
-  Py_BEGIN_ALLOW_THREADS
-  r = cvWaitKey(delay);
-  Py_END_ALLOW_THREADS
-  return FROM_int(r);
-}
-
-static PyObject *pycvLoadImage(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "filename", "iscolor", NULL };
-  char *filename;
-  int iscolor = CV_LOAD_IMAGE_COLOR;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "s|i", (char**)keywords, &filename, &iscolor))
-    return NULL;
-
-  // Inside ALLOW_THREADS, must not reference 'filename' because it might move.
-  // So make a local copy 'filename_copy'.
-  char filename_copy[2048];
-  strncpy(filename_copy, filename, sizeof(filename_copy));
-
-  IplImage *r;
-  Py_BEGIN_ALLOW_THREADS
-  r = cvLoadImage(filename_copy, iscolor);
-  Py_END_ALLOW_THREADS
-
-  if (r == NULL) {
-    PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
-    return NULL;
-  } else {
-    return FROM_IplImagePTR(r);
-  }
-}
-
-static PyObject *pycvLoadImageM(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "filename", "iscolor", NULL };
-  char *filename;
-  int iscolor = CV_LOAD_IMAGE_COLOR;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "s|i", (char**)keywords, &filename, &iscolor))
-    return NULL;
-
-  // Inside ALLOW_THREADS, must not reference 'filename' because it might move.
-  // So make a local copy 'filename_copy'.
-  char filename_copy[2048];
-  strncpy(filename_copy, filename, sizeof(filename_copy));
-
-  CvMat *r;
-  Py_BEGIN_ALLOW_THREADS
-  r = cvLoadImageM(filename_copy, iscolor);
-  Py_END_ALLOW_THREADS
-
-  if (r == NULL) {
-    PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
-    return NULL;
-  } else {
-    return FROM_CvMatPTR(r);
-  }
-}
-
-static PyObject *pycvCreateImageHeader(PyObject *self, PyObject *args)
-{
-  int w, h, depth, channels;
-  if (!PyArg_ParseTuple(args, "(ii)Ii", &w, &h, &depth, &channels))
-    return NULL;
-  iplimage_t *cva = PyObject_NEW(iplimage_t, &iplimage_Type);
-  cva->a = cvCreateImageHeader(cvSize(w, h), depth, channels);
-  if (cva->a == NULL) {
-    PyErr_SetString(PyExc_TypeError, "CreateImage failed");
-    return NULL;
-  } else {
-    cva->data = Py_None;
-    Py_INCREF(cva->data);
-    cva->offset = 0;
-
-    return (PyObject*)cva;
-  }
-}
-
-static PyObject *pycvCreateImage(PyObject *self, PyObject *args)
-{
-  int w, h, depth, channels;
-  if (!PyArg_ParseTuple(args, "(ii)Ii:CreateImage", &w, &h, &depth, &channels))
-    return NULL;
-  iplimage_t *cva = PyObject_NEW(iplimage_t, &iplimage_Type);
-  ERRWRAP(cva->a = cvCreateImage(cvSize(w, h), depth, channels));
-  if (cva->a == NULL) {
-    PyErr_SetString(PyExc_TypeError, "CreateImage failed");
-    return NULL;
-  } else {
-    return pythonize_IplImage(cva);
-  }
-}
-
-static PyObject *pycvCreateMatHeader(PyObject *self, PyObject *args)
-{
-  int rows, cols, type;
-  if (!PyArg_ParseTuple(args, "iii", &rows, &cols, &type))
-    return NULL;
-  cvmat_t *m = PyObject_NEW(cvmat_t, &cvmat_Type);
-  ERRWRAP(m->a = cvCreateMatHeader(rows, cols, type));
-  if (m->a == NULL) {
-    PyErr_SetString(PyExc_TypeError, "CreateMat failed");
-    return NULL;
-  } else {
-    m->data = Py_None;
-    Py_INCREF(m->data);
-    m->offset = 0;
-    return (PyObject*)m;
-  }
-}
-
-static PyObject *pycvCreateMat(PyObject *self, PyObject *args)
-{
-  int rows, cols, type;
-  if (!PyArg_ParseTuple(args, "iii", &rows, &cols, &type))
-    return NULL;
-  cvmat_t *m = PyObject_NEW(cvmat_t, &cvmat_Type);
-  ERRWRAP(m->a = cvCreateMat(rows, cols, type));
-  if (m->a == NULL) {
-    PyErr_SetString(PyExc_TypeError, "CreateMat failed");
-    return NULL;
-  } else {
-    return pythonize_CvMat(m);
-  }
-}
-
-static PyObject *pycvCreateMatNDHeader(PyObject *self, PyObject *args)
-{
-  ints dims;
-  int type;
-
-  if (!PyArg_ParseTuple(args, "O&i", convert_to_ints, (void*)&dims, &type))
-    return NULL;
-  cvmatnd_t *m = PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-  ERRWRAP(m->a = cvCreateMatNDHeader(dims.count, dims.i, type));
-
-  m->data = Py_None;
-  Py_INCREF(m->data);
-  delete [] dims.i;
-  return (PyObject*)m;
-}
-
-
-static PyObject *pycvCreateMatND(PyObject *self, PyObject *args)
-{
-  ints dims;
-  int type;
-
-  if (!PyArg_ParseTuple(args, "O&i", convert_to_ints, (void*)&dims, &type))
-    return NULL;
-  cvmatnd_t *m = PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-  ERRWRAP(m->a = cvCreateMatND(dims.count, dims.i, type));
-  delete [] dims.i;
-  return pythonize_CvMatND(m);
-}
-
-#if PYTHON_USE_NUMPY
-static PyObject *pycvfromarray(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "arr", "allowND", NULL };
-  PyObject *o;
-  int allowND = 0;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "O|i", (char**)keywords, &o, &allowND))
-    return NULL;
-  return fromarray(o, allowND);
-}
-
-static PyObject *fromarray(PyObject *o, int allowND)
-{
-  PyObject *ao = PyObject_GetAttrString(o, "__array_struct__");
-  PyObject *retval;
-
-  if ((ao == NULL) || !PyCObject_Check(ao)) {
-    PyErr_SetString(PyExc_TypeError, "object does not have array interface");
-    return NULL;
-  }
-  PyArrayInterface *pai = (PyArrayInterface*)PyCObject_AsVoidPtr(ao);
-  if (pai->two != 2) {
-    PyErr_SetString(PyExc_TypeError, "object does not have array interface");
-      Py_DECREF(ao);
-      return NULL;
-  }
-
-  int type = -1;
-
-  switch (pai->typekind) {
-  case 'i':
-    if (pai->itemsize == 1)
-      type = CV_8SC1;
-    else if (pai->itemsize == 2)
-      type = CV_16SC1;
-    else if (pai->itemsize == 4)
-      type = CV_32SC1;
-    break;
-
-  case 'u':
-    if (pai->itemsize == 1)
-      type = CV_8UC1;
-    else if (pai->itemsize == 2)
-      type = CV_16UC1;
-    break;
-
-  case 'f':
-    if (pai->itemsize == 4)
-      type = CV_32FC1;
-    else if (pai->itemsize == 8)
-      type = CV_64FC1;
-    break;
-
-  }
-  if (type == -1) {
-     PyErr_SetString(PyExc_TypeError, "the array type is not supported by OpenCV");
-     Py_DECREF(ao);
-     return NULL;
-  }
-
-  if (!allowND) {
-    cvmat_t *m = PyObject_NEW(cvmat_t, &cvmat_Type);
-    if (pai->nd == 2) {
-      if (pai->strides[1] != pai->itemsize) {
-        return failmsg("cv.fromarray array can only accept arrays with contiguous data"), (PyObject*)0;
-      }
-      ERRWRAP(m->a = cvCreateMatHeader((int)pai->shape[0], (int)pai->shape[1], type));
-      m->a->step = (int)pai->strides[0];
-    } else if (pai->nd == 3) {
-      if (pai->shape[2] > CV_CN_MAX) {
-        Py_DECREF(ao);
-        return failmsg("cv.fromarray too many channels, see allowND argument"), (PyObject*)0;
-      }
-      ERRWRAP(m->a = cvCreateMatHeader((int)pai->shape[0], (int)pai->shape[1], type + ((int)(pai->shape[2] - 1) << CV_CN_SHIFT)));
-      m->a->step = (int)pai->strides[0];
-    } else {
-      Py_DECREF(ao);
-      return failmsg("cv.fromarray array can be 2D or 3D only, see allowND argument"), (PyObject*)0;
-    }
-    m->a->data.ptr = (uchar*)pai->data;
-    //retval = pythonize_foreign_CvMat(m);
-    m->data = o;
-    m->offset = 0;
-    retval = (PyObject*)m;
-  } else {
-    int dims[CV_MAX_DIM];
-    int i;
-    for (i = 0; i < pai->nd; i++)
-      dims[i] = (int)pai->shape[i];
-    cvmatnd_t *m = PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-    ERRWRAP(m->a = cvCreateMatNDHeader(pai->nd, dims, type));
-    m->a->data.ptr = (uchar*)pai->data;
-    m->data = o;
-    m->offset = 0;
-    retval = (PyObject*)m;
-    //retval = pythonize_CvMatND(m, ao);
-  }
-  Py_DECREF(ao);
-  Py_INCREF(o);
-  return retval;
-}
-#endif
-
-class ranges {
-public:
-  Py_ssize_t len;
-  float **rr;
-  ranges() {
-    len = 0;
-    rr = NULL;
-  }
-  int fromobj(PyObject *o, const char *name = "no_name") {
-    PyObject *fi = PySequence_Fast(o, name);
-    if (fi == NULL)
-      return 0;
-    len = (int)PySequence_Fast_GET_SIZE(fi);
-    rr = new float*[len];
-    for (Py_ssize_t i = 0; i < len; i++) {
-      PyObject *item = PySequence_Fast_GET_ITEM(fi, i);
-      floats ff; ff.f = 0;
-      if (!convert_to_floats(item, &ff))
-        return 0;
-      rr[i] = ff.f;
-    }
-    Py_DECREF(fi);
-    return 1;
-  }
-  ~ranges() {
-    for (Py_ssize_t i = 0; i < len; i++)
-      delete rr[i];
-    delete[] rr;
-  }
-};
-
-static int ranges_converter(PyObject *o, ranges* dst)
-{
-  return dst->fromobj(o);
-}
-
-static PyObject *pycvCreateHist(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "dims", "type", "ranges", "uniform", NULL };
-  PyObject *dims;
-  int type;
-  int uniform = 1;
-  ranges r;
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "Oi|O&i", (char**)keywords, &dims, &type, ranges_converter, (void*)&r, &uniform)) {
-    return NULL;
-  }
-  cvhistogram_t *h = PyObject_NEW(cvhistogram_t, &cvhistogram_Type);
-  args = Py_BuildValue("Oi", dims, CV_32FC1);
-  memset(&h->h, 0, sizeof(h->h));
-  h->bins = pycvCreateMatND(self, args);
-  Py_DECREF(args);
-  if (h->bins == NULL) {
-    return NULL;
-  }
-  h->h.type = CV_HIST_MAGIC_VAL + CV_HIST_UNIFORM_FLAG;
-  if (!convert_to_CvArr(h->bins, &(h->h.bins), "bins"))
-    return NULL;
-
-  if(r.rr)
-  {
-      ERRWRAP(cvSetHistBinRanges(&(h->h), r.rr, uniform));
-  }
-
-  return (PyObject*)h;
-}
-
-static PyObject *pycvInitLineIterator(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "image", "pt1", "pt2", "connectivity", "left_to_right", NULL };
-  CvArr *image;
-  CvPoint pt1;
-  CvPoint pt2;
-  int connectivity = 8;
-  int left_to_right = 0;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "O&O&O&|ii", (char**)keywords,
-                        convert_to_CvArr, &image,
-                        convert_to_CvPoint, &pt1,
-                        convert_to_CvPoint, &pt2,
-                        &connectivity,
-                        &left_to_right))
-    return NULL;
-
-  cvlineiterator_t *pi = PyObject_NEW(cvlineiterator_t, &cvlineiterator_Type);
-  pi->count = cvInitLineIterator(image, pt1, pt2, &pi->iter, connectivity, left_to_right);
-  ERRWRAP(pi->type = cvGetElemType(image));
-  return (PyObject*)pi;
-}
-
-static PyObject *pycvCreateMemStorage(PyObject *self, PyObject *args)
-{
-  int block_size = 0;
-  if (!PyArg_ParseTuple(args, "|i", &block_size))
-    return NULL;
-  cvmemstorage_t *pm = PyObject_NEW(cvmemstorage_t, &cvmemstorage_Type);
-  pm->a = cvCreateMemStorage(block_size);
-  return (PyObject*)pm;
-}
-
-// single index: return row
-// 2 indices: row, column
-// both row and column can be slices.  column slice must have a step of 1.
-//
-// returns a scalar when all dimensions are specified and all are integers.  Otherwise returns a CvMat.
-//
-static PyObject *cvarr_GetItem(PyObject *o, PyObject *key)
-{
-  dims dd;
-
-  CvArr *cva;
-  if (!convert_to_CvArr(o, &cva, "src"))
-    return NULL;
-
-  if (!convert_to_dims(key, &dd, cva, "key")) {
-    return NULL;
-  }
-
-  // Figure out if all supplied indices have a stride of zero - means they are not slices
-  // and if all indices are positive
-  int all0 = 1;
-  for (int i = 0; i < dd.count; i++) {
-    all0 &= (dd.step[i] == 0) && (0 <= dd.i[i]);
-  }
-
-  // if every dimension supplied, and none are slices, return the scalar
-  if ((cvGetDims(cva) == dd.count) && all0) {
-    CvScalar s;
-    ERRWRAP(s = cvGetND(cva, dd.i));
-    return PyObject_FromCvScalar(s, cvGetElemType(cva));
-  } else {
-    // pad missing dimensions
-    for (int i = dd.count; i < cvGetDims(cva); i++) {
-      dd.i[i] = 0;
-      dd.step[i] = 1;
-      dd.length[i] = cvGetDimSize(cva, i);
-    }
-    dd.count = cvGetDims(cva);
-
-    // negative steps are illegal for OpenCV
-    for (int i = 0; i < dd.count; i++) {
-      if (dd.step[i] < 0)
-        return failmsg("Negative step is illegal"), (PyObject*)0;
-    }
-
-    // zero length illegal for OpenCV
-    for (int i = 0; i < dd.count; i++) {
-      if (dd.length[i] == 0)
-        return failmsg("Zero sized dimension is illegal"), (PyObject*)0;
-    }
-
-    // column step can only be 0 or 1
-    if ((dd.step[dd.count-1] != 0) && (dd.step[dd.count-1] != 1))
-        return failmsg("Column step is illegal"), (PyObject*)0;
-
-    if (is_cvmat(o) || is_iplimage(o)) {
-      cvmat_t *sub = PyObject_NEW(cvmat_t, &cvmat_Type);
-      sub->a = cvCreateMatHeader(dd.length[0], dd.length[1], cvGetElemType(cva));
-      uchar *old0;  // pointer to first element in old mat
-      int oldstep;
-      cvGetRawData(cva, &old0, &oldstep);
-      uchar *new0;  // pointer to first element in new mat
-      ERRWRAP(new0 = cvPtrND(cva, dd.i));
-
-      sub->a->step = oldstep * dd.step[0];
-      sub->data = what_data(o);
-      Py_INCREF(sub->data);
-      sub->offset = new0 - old0;
-      return (PyObject*)sub;
-    } else {
-      cvmatnd_t *sub = PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-      sub->a = cvCreateMatNDHeader(dd.count, dd.length, cvGetElemType(cva));
-      uchar *old0;  // pointer to first element in old mat
-      cvGetRawData(cva, &old0);
-      uchar *new0;  // pointer to first element in new mat
-      ERRWRAP(new0 = cvPtrND(cva, dd.i));
-
-      for (int d = 0; d < dd.count; d++) {
-        int stp = dd.step[d];
-        sub->a->dim[d].step = ((CvMatND*)cva)->dim[d].step * ((stp == 0) ? 1 : stp);
-        sub->a->dim[d].size = dd.length[d];
-      }
-      sub->data = what_data(o);
-      Py_INCREF(sub->data);
-      sub->offset = new0 - old0;
-      return (PyObject*)sub;
-    }
-  }
-}
-
-static int cvarr_SetItem(PyObject *o, PyObject *key, PyObject *v)
-{
-  dims dd;
-
-  CvArr *cva;
-  if (!convert_to_CvArr(o, &cva, "src"))
-    return -1;
-
-  if (!convert_to_dims(key, &dd, cva, "key")) {
-    return -1;
-  }
-
-  if (cvGetDims(cva) != dd.count) {
-    PyErr_SetString(PyExc_TypeError, "key length does not match array dimension");
-    return -1;
-  }
-
-  CvScalar s;
-  if (PySequence_Check(v)) {
-    PyObject *fi = PySequence_Fast(v, "v");
-    if (fi == NULL)
-      return -1;
-    if (PySequence_Fast_GET_SIZE(fi) != CV_MAT_CN(cvGetElemType(cva))) {
-      PyErr_SetString(PyExc_TypeError, "sequence size must be same as channel count");
-      return -1;
-    }
-    for (Py_ssize_t i = 0; i < PySequence_Fast_GET_SIZE(fi); i++)
-      s.val[i] = PyFloat_AsDouble(PySequence_Fast_GET_ITEM(fi, i));
-    Py_DECREF(fi);
-  } else {
-    if (1 != CV_MAT_CN(cvGetElemType(cva))) {
-      PyErr_SetString(PyExc_TypeError, "scalar supplied but channel count does not equal 1");
-      return -1;
-    }
-    s.val[0] = PyFloat_AsDouble(v);
-  }
-  switch (dd.count) {
-  case 1:
-    ERRWRAPN(cvSet1D(cva, dd.i[0], s), -1);
-    break;
-  case 2:
-    ERRWRAPN(cvSet2D(cva, dd.i[0], dd.i[1], s), -1);
-    break;
-  case 3:
-    ERRWRAPN(cvSet3D(cva, dd.i[0], dd.i[1], dd.i[2], s), -1);
-    break;
-  default:
-    ERRWRAPN(cvSetND(cva, dd.i, s), -1);
-    // XXX - OpenCV bug? - seems as if an error in cvSetND does not set error status?
-    break;
-  }
-  if (cvGetErrStatus() != 0) {
-    translate_error_to_exception();
-    return -1;
-  }
-
-  return 0;
-}
-
-
-static PyObject *pycvSetData(PyObject *self, PyObject *args)
-{
-  PyObject *o, *s;
-  int step = CV_AUTO_STEP;
-
-  if (!PyArg_ParseTuple(args, "OO|i", &o, &s, &step))
-    return NULL;
-  if (is_iplimage(o)) {
-    iplimage_t *ipl = (iplimage_t*)o;
-    ipl->a->widthStep = step;
-    Py_DECREF(ipl->data);
-    ipl->data = s;
-    Py_INCREF(ipl->data);
-  } else if (is_cvmat(o)) {
-    cvmat_t *m = (cvmat_t*)o;
-    m->a->step = step;
-    Py_DECREF(m->data);
-    m->data = s;
-    Py_INCREF(m->data);
-  } else if (is_cvmatnd(o)) {
-    cvmatnd_t *m = (cvmatnd_t*)o;
-    Py_DECREF(m->data);
-    m->data = s;
-    Py_INCREF(m->data);
-  } else {
-    PyErr_SetString(PyExc_TypeError, "SetData argument must be either IplImage, CvMat or CvMatND");
-    return NULL;
-  }
-
-  Py_RETURN_NONE;
-}
-
-static PyObject *what_data(PyObject *o)
-{
-  if (is_iplimage(o)) {
-    iplimage_t *ipl = (iplimage_t*)o;
-    return ipl->data;
-  } else if (is_cvmat(o)) {
-    cvmat_t *m = (cvmat_t*)o;
-    return m->data;
-  } else if (is_cvmatnd(o)) {
-    cvmatnd_t *m = (cvmatnd_t*)o;
-    return m->data;
-  } else {
-    assert(0);
-    return NULL;
-  }
-}
-
-static PyObject *pycvCreateData(PyObject *self, PyObject *args)
-{
-  PyObject *o;
-
-  if (!PyArg_ParseTuple(args, "O", &o))
-    return NULL;
-
-  CvArr *a;
-  if (!convert_to_CvArr(o, &a, "arr"))
-    return NULL;
-  ERRWRAP(cvCreateData(a));
-
-  Py_DECREF(what_data(o));
-  if (is_iplimage(o)) {
-    iplimage_t *ipl = (iplimage_t*)o;
-    pythonize_IplImage(ipl);
-  } else if (is_cvmat(o)) {
-    cvmat_t *m = (cvmat_t*)o;
-    pythonize_CvMat(m);
-  } else if (is_cvmatnd(o)) {
-    cvmatnd_t *m = (cvmatnd_t*)o;
-    pythonize_CvMatND(m);
-  } else {
-    PyErr_SetString(PyExc_TypeError, "CreateData argument must be either IplImage, CvMat or CvMatND");
-    return NULL;
-  }
-
-  Py_RETURN_NONE;
-}
-
-static PyObject *pycvGetDims(PyObject *self, PyObject *args)
-{
-  PyObject *o;
-
-  if (!PyArg_ParseTuple(args, "O", &o))
-    return NULL;
-  CvArr *cva;
-  if (!convert_to_CvArr(o, &cva, "src"))
-    return NULL;
-
-  int i, nd;
-  ERRWRAP(nd = cvGetDims(cva));
-  PyObject *r = PyTuple_New(nd);
-  for (i = 0; i < nd; i++)
-    PyTuple_SetItem(r, i, PyInt_FromLong(cvGetDimSize(cva, i)));
-  return r;
-}
-
-static PyObject *pycvGetImage(PyObject *self, PyObject *args)
-{
-  PyObject *o, *r;
-
-  if (!PyArg_ParseTuple(args, "O", &o))
-    return NULL;
-  if (is_iplimage(o)) {
-    r = o;
-    Py_INCREF(o);
-  } else {
-    IplImage *ipl = cvCreateImageHeader(cvSize(100,100), 8, 1); // these args do not matter, because overwritten
-    CvArr *cva;
-    if (!convert_to_CvArr(o, &cva, "src"))
-      return NULL;
-    ERRWRAP(cvGetImage(cva, ipl));
-
-    iplimage_t *oipl = PyObject_NEW(iplimage_t, &iplimage_Type);
-    oipl->a = ipl;
-    oipl->data = what_data(o);
-    Py_INCREF(oipl->data);
-    oipl->offset = 0;
-
-    r = (PyObject*)oipl;
-  }
-  return r;
-}
-
-static PyObject *pycvGetMat(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "arr", "allowND", NULL };
-  PyObject *o, *r;
-  int allowND = 0;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "O|i", (char**)keywords, &o, &allowND))
-    return NULL;
-  if (is_cvmat(o)) {
-    r = o;
-    Py_INCREF(o);
-  } else {
-    CvMat *m = cvCreateMatHeader(100,100, 1); // these args do not matter, because overwritten
-    CvArr *cva;
-    if (!convert_to_CvArr(o, &cva, "src"))
-      return NULL;
-    ERRWRAP(cvGetMat(cva, m, NULL, allowND));
-
-    cvmat_t *om = PyObject_NEW(cvmat_t, &cvmat_Type);
-    om->a = m;
-    om->data = what_data(o);
-    Py_INCREF(om->data);
-    om->offset = 0;
-
-    r = (PyObject*)om;
-  }
-  return r;
-}
-
-static PyObject *pycvReshape(PyObject *self, PyObject *args)
-{
-  PyObject *o;
-  int new_cn;
-  int new_rows = 0;
-
-  if (!PyArg_ParseTuple(args, "Oi|i", &o, &new_cn, &new_rows))
-    return NULL;
-
-  CvMat *m = cvCreateMatHeader(100,100, 1); // these args do not matter, because overwritten
-  CvArr *cva;
-  if (!convert_to_CvArr(o, &cva, "src"))
-    return NULL;
-  ERRWRAP(cvReshape(cva, m, new_cn, new_rows));
-
-  cvmat_t *om = PyObject_NEW(cvmat_t, &cvmat_Type);
-  om->a = m;
-  om->data = what_data(o);
-  Py_INCREF(om->data);
-  om->offset = 0;
-
-  return (PyObject*)om;
-}
-
-static PyObject *pycvReshapeMatND(PyObject *self, PyObject *args)
-{
-  PyObject *o;
-  int new_cn = 0;
-  PyObject *new_dims = NULL;
-
-  if (!PyArg_ParseTuple(args, "OiO", &o, &new_cn, &new_dims))
-    return NULL;
-
-  CvMatND *cva;
-  if (!convert_to_CvMatND(o, &cva, "src"))
-    return NULL;
-  ints dims={0,0};
-  if (new_dims != NULL) {
-    if (!convert_to_ints(new_dims, &dims, "new_dims"))
-      return NULL;
-  }
-
-  if (new_cn == 0)
-    new_cn = CV_MAT_CN(cvGetElemType(cva));
-
-  int i;
-  int count = CV_MAT_CN(cvGetElemType(cva));
-  for (i = 0; i < cva->dims; i++)
-    count *= cva->dim[i].size;
-
-  int newcount = new_cn;
-  for (i = 0; i < dims.count; i++)
-    newcount *= dims.i[i];
-
-  if (count != newcount) {
-    PyErr_SetString(PyExc_TypeError, "Total number of elements must be unchanged");
-    return NULL;
-  }
-
-  CvMatND *pn = cvCreateMatNDHeader(dims.count, dims.i, CV_MAKETYPE(CV_MAT_TYPE(cva->type), new_cn));
-  return shareDataND(o, cva, pn);
-}
-
-static PyObject *pycvFindContours(PyObject *self, PyObject *args, PyObject *kw)
-{
-  CvArr* image;
-  PyObject *pyobj_image = NULL;
-  CvMemStorage* storage;
-  PyObject *pyobj_storage = NULL;
-  CvSeq* first_contour;
-  int header_size = sizeof(CvContour);
-  int mode = CV_RETR_LIST;
-  int method = CV_CHAIN_APPROX_SIMPLE;
-  CvPoint offset = cvPoint(0,0);
-  PyObject *pyobj_offset = NULL;
-
-  const char *keywords[] = { "image", "storage", "mode", "method", "offset", NULL };
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "OO|iiO", (char**)keywords, &pyobj_image, &pyobj_storage, &mode, &method, &pyobj_offset))
-    return NULL;
-  if (!convert_to_CvArr(pyobj_image, &image, "image")) return NULL;
-  if (!convert_to_CvMemStorage(pyobj_storage, &storage, "storage")) return NULL;
-  if ((pyobj_offset != NULL) && !convert_to_CvPoint(pyobj_offset, &offset, "offset")) return NULL;
-  ERRWRAP(cvFindContours(image, storage, &first_contour, header_size, mode, method, offset));
-  cvseq_t *ps = PyObject_NEW(cvseq_t, &cvseq_Type);
-  ps->a = first_contour;
-  ps->container = PyTuple_GetItem(args, 1); // storage
-  Py_INCREF(ps->container);
-  return (PyObject*)ps;
-}
-
-static PyObject *pycvApproxPoly(PyObject *self, PyObject *args, PyObject *kw)
-{
-  cvarrseq src_seq;
-  PyObject *pyobj_src_seq = NULL;
-  int header_size = sizeof(CvContour);
-  CvMemStorage* storage;
-  PyObject *pyobj_storage = NULL;
-  int method;
-  double parameter = 0;
-  int parameter2 = 0;
-
-  const char *keywords[] = { "src_seq", "storage", "method", "parameter", "parameter2", NULL };
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "OOi|di", (char**)keywords, &pyobj_src_seq, &pyobj_storage, &method, &parameter, &parameter2))
-    return NULL;
-  if (!convert_to_cvarrseq(pyobj_src_seq, &src_seq, "src_seq")) return NULL;
-  if (!convert_to_CvMemStorage(pyobj_storage, &storage, "storage")) return NULL;
-  CvSeq* r;
-  ERRWRAP(r = cvApproxPoly(src_seq.mat, header_size, storage, method, parameter, parameter2));
-  return FROM_CvSeqPTR(r);
-}
-
-static float distance_function_glue( const float* a, const float* b, void* user_param )
-{
-  PyObject *o = (PyObject*)user_param;
-  PyObject *args = Py_BuildValue("(ff)(ff)O", a[0], a[1], b[0], b[1], PyTuple_GetItem(o, 1));
-  PyObject *r = PyObject_Call(PyTuple_GetItem(o, 0), args, NULL);
-  Py_DECREF(args);
-  return (float)PyFloat_AsDouble(r);
-}
-
-static PyObject *pycvCalcEMD2(PyObject *self, PyObject *args, PyObject *kw)
-{
-  const char *keywords[] = { "signature1", "signature2", "distance_type", "distance_func", "cost_matrix", "flow", "lower_bound", "userdata", NULL };
-  CvArr* signature1;
-  PyObject *pyobj_signature1;
-  CvArr* signature2;
-  PyObject *pyobj_signature2;
-  int distance_type;
-  PyObject *distance_func = NULL;
-  CvArr* cost_matrix=NULL;
-  PyObject *pyobj_cost_matrix = NULL;
-  CvArr* flow=NULL;
-  PyObject *pyobj_flow = NULL;
-  float lower_bound = FLT_MAX;
-  PyObject *userdata = NULL;
-
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "OOi|OOOfO", (char**)keywords,
-                                   &pyobj_signature1,
-                                   &pyobj_signature2,
-                                   &distance_type,
-                                   &distance_func,
-                                   &pyobj_cost_matrix,
-                                   &pyobj_flow,
-                                   &lower_bound,
-                                   &userdata))
-    return NULL;
-  if (!convert_to_CvArr(pyobj_signature1, &signature1, "signature1")) return NULL;
-  if (!convert_to_CvArr(pyobj_signature2, &signature2, "signature2")) return NULL;
-  if (pyobj_cost_matrix && !convert_to_CvArr(pyobj_cost_matrix, &cost_matrix, "cost_matrix")) return NULL;
-  if (pyobj_flow && !convert_to_CvArr(pyobj_flow, &flow, "flow")) return NULL;
-
-  if (distance_func == NULL) {
-    distance_func = Py_None;
-  }
-  if (userdata == NULL) {
-    userdata = Py_None;
-  }
-
-  PyObject *ud = Py_BuildValue("OO", distance_func, userdata);
-  float r;
-  ERRWRAP(r = cvCalcEMD2(signature1, signature2, distance_type, distance_function_glue, cost_matrix, flow, &lower_bound, (void*)ud));
-  Py_DECREF(ud);
-
-  return PyFloat_FromDouble(r);
-}
-
-static PyObject *pycvSubdiv2DLocate(PyObject *self, PyObject *args)
-{
-  PyObject *pyobj_subdiv;
-  PyObject *pyobj_pt;
-  CvSubdiv2D *subdiv;
-  CvPoint2D32f pt;
-  CvSubdiv2DEdge edge;
-  CvSubdiv2DPoint* vertex;
-
-  if (!PyArg_ParseTuple(args, "OO", &pyobj_subdiv, &pyobj_pt))
-    return NULL;
-  if (!convert_to_CvSubdiv2DPTR(pyobj_subdiv, &subdiv, "subdiv"))
-    return NULL;
-  if (!convert_to_CvPoint2D32f(pyobj_pt, &pt, "pt"))
-    return NULL;
-
-  CvSubdiv2DPointLocation loc = cvSubdiv2DLocate(subdiv, pt, &edge, &vertex);
-  PyObject *r;
-  switch (loc) {
-  case CV_PTLOC_INSIDE:
-  case CV_PTLOC_ON_EDGE:
-    r = FROM_CvSubdiv2DEdge(edge);
-    break;
-  case CV_PTLOC_VERTEX:
-    r = FROM_CvSubdiv2DPointPTR(vertex);
-    break;
-  case CV_PTLOC_OUTSIDE_RECT:
-    r = Py_None;
-    Py_INCREF(Py_None);
-    break;
-  default:
-    return failmsg("Unexpected loc from cvSubdiv2DLocate"), (PyObject*)0;
-  }
-  return Py_BuildValue("iO", (int)loc, r);
-}
-
-static PyObject *pycvCalcOpticalFlowPyrLK(PyObject *self, PyObject *args)
-{
-  CvArr* prev;
-  PyObject *pyobj_prev = NULL;
-  CvArr* curr;
-  PyObject *pyobj_curr = NULL;
-  CvArr* prev_pyr;
-  PyObject *pyobj_prev_pyr = NULL;
-  CvArr* curr_pyr;
-  PyObject *pyobj_curr_pyr = NULL;
-  CvPoint2D32f* prev_features;
-  PyObject *pyobj_prev_features = NULL;
-  PyObject *pyobj_curr_features = NULL;
-  CvPoint2D32f* curr_features;
-  CvSize win_size;
-  int level;
-  CvTermCriteria criteria;
-  int flags;
-
-  if (!PyArg_ParseTuple(args, "OOOOO(ii)i(iif)i|O",
-    &pyobj_prev, &pyobj_curr, &pyobj_prev_pyr, &pyobj_curr_pyr,
-    &pyobj_prev_features,
-    &win_size.width, &win_size.height, &level,
-    &criteria.type, &criteria.max_iter, &criteria.epsilon,
-    &flags,
-    &pyobj_curr_features))
-    return NULL;
-  if (!convert_to_CvArr(pyobj_prev, &prev, "prev")) return NULL;
-  if (!convert_to_CvArr(pyobj_curr, &curr, "curr")) return NULL;
-  if (!convert_to_CvArr(pyobj_prev_pyr, &prev_pyr, "prev_pyr")) return NULL;
-  if (!convert_to_CvArr(pyobj_curr_pyr, &curr_pyr, "curr_pyr")) return NULL;
-  if (!convert_to_CvPoint2D32fPTR(pyobj_prev_features, &prev_features, "prev_features")) return NULL;
-  int count = (int)PySequence_Length(pyobj_prev_features);
-  if (flags & CV_LKFLOW_INITIAL_GUESSES) {
-    failmsg("flag CV_LKFLOW_INITIAL_GUESSES is determined automatically from function arguments - it is not required");
-    return NULL;
-  }
-  if (!pyobj_curr_features) {
-    curr_features = new CvPoint2D32f[count];
-  } else {
-    if (PySequence_Length(pyobj_curr_features) != count) {
-      failmsg("curr_features must have same length as prev_features");
-      return NULL;
-    }
-    if (!convert_to_CvPoint2D32fPTR(pyobj_curr_features, &curr_features, "curr_features")) return NULL;
-    flags |= CV_LKFLOW_INITIAL_GUESSES;
-  }
-  float *track_error = new float[count];
-  char* status = new char[count];
-  ERRWRAP(cvCalcOpticalFlowPyrLK(prev, curr, prev_pyr, curr_pyr, prev_features, curr_features, count, win_size, level, status, track_error, criteria, flags));
-
-  cvpoint2d32f_count r0;
-  r0.points = curr_features;
-  r0.count = count;
-
-  chars r1;
-  r1.f = status;
-  r1.count = count;
-
-  floats r2;
-  r2.f = track_error;
-  r2.count = count;
-
-  return Py_BuildValue("NNN", FROM_cvpoint2d32f_count(r0), FROM_chars(r1), FROM_floats(r2));
-}
-
-// pt1,pt2 are input and output arguments here
-
-static PyObject *pycvClipLine(PyObject *self, PyObject *args)
-{
-  CvSize img_size;
-  PyObject *pyobj_img_size = NULL;
-  CvPoint pt1;
-  PyObject *pyobj_pt1 = NULL;
-  CvPoint pt2;
-  PyObject *pyobj_pt2 = NULL;
-
-  if (!PyArg_ParseTuple(args, "OOO", &pyobj_img_size, &pyobj_pt1, &pyobj_pt2))
-    return NULL;
-  if (!convert_to_CvSize(pyobj_img_size, &img_size, "img_size")) return NULL;
-  if (!convert_to_CvPoint(pyobj_pt1, &pt1, "pt1")) return NULL;
-  if (!convert_to_CvPoint(pyobj_pt2, &pt2, "pt2")) return NULL;
-  int r;
-  ERRWRAP(r = cvClipLine(img_size, &pt1, &pt2));
-  if (r == 0) {
-    Py_RETURN_NONE;
-  } else {
-    return Py_BuildValue("NN", FROM_CvPoint(pt1), FROM_CvPoint(pt2));
-  }
-}
-
-static PyObject *pyfinddatamatrix(PyObject *self, PyObject *args)
-{
-  PyObject *pyim;
-  if (!PyArg_ParseTuple(args, "O", &pyim))
-    return NULL;
-
-  CvMat *image;
-  if (!convert_to_CvMat(pyim, &image, "image")) return NULL;
-
-  std::deque <CvDataMatrixCode> codes;
-  ERRWRAP(codes = cvFindDataMatrix(image));
-
-  PyObject *pycodes = PyList_New(codes.size());
-  for (size_t i = 0; i < codes.size(); i++) {
-    CvDataMatrixCode *pc = &codes[i];
-    PyList_SetItem(pycodes, i, Py_BuildValue("(sOO)", pc->msg, FROM_CvMat(pc->corners), FROM_CvMat(pc->original)));
-  }
-
-  return pycodes;
-}
-
-static PyObject *temp_test(PyObject *self, PyObject *args)
-{
-#if 0
-  CvArr *im = cvLoadImage("../samples/c/lena.jpg", 0);
-  printf("im=%p\n", im);
-  CvMat *m = cvEncodeImage(".jpeg", im);
-#endif
-#if 0
-  CvArr *im = cvLoadImage("lena.jpg", 0);
-  float r0[] = { 0, 255 };
-  float *ranges[] = { r0 };
-  int hist_size[] = { 256 };
-  CvHistogram *hist = cvCreateHist(1, hist_size, CV_HIST_ARRAY, ranges, 1);
-  cvCalcHist(im, hist, 0, 0);
-#endif
-
-#if 0
-  CvMat* mat = cvCreateMat( 3, 3, CV_32F );
-  CvMat row_header, *row;
-  row = cvReshape( mat, &row_header, 0, 1 );
-  printf("%d,%d\n", row_header.rows, row_header.cols);
-  printf("ge %08x\n", cvGetElemType(mat));
-#endif
-
-#if 0
-  CvMat *m = cvCreateMat(1, 10, CV_8UC1);
-  printf("CvMat stride ===> %d\n", m->step);
-#endif
-
-#if 0
-  CvPoint2D32f src[3] = { { 0,0 }, { 1,0 }, { 0,1 } };
-  CvPoint2D32f dst[3] = { { 0,0 }, { 17,0 }, { 0,17 } };
-
-  CvMat* mapping = cvCreateMat(2, 3, CV_32FC1);
-  cvGetAffineTransform(src, dst, mapping);
-  printf("===> %f\n", cvGetReal2D(mapping, 0, 0));
-#endif
-
-#if 0
-  CvArr *im = cvLoadImage("checker77.png");
-  CvPoint2D32f corners[49];
-  int count;
-  cvFindChessboardCorners(im, cvSize(7,7), corners, &count, 0);
-  printf("count=%d\n", count);
-#endif
-
-#if 0
-  CvMat *src = cvCreateMat(512, 512, CV_8UC3);
-  CvMat *dst = cvCreateMat(512, 512, CV_8UC3);
-  cvPyrMeanShiftFiltering(src, dst, 5, 5);
-  return FROM_CvMat(src);
-#endif
-
-  return PyFloat_FromDouble(0.0);
-}
-
-static PyObject *pycvFindChessboardCorners(PyObject *self, PyObject *args, PyObject *kw)
-{
-  CvArr* image;
-  PyObject *pyobj_image = NULL;
-  CvSize pattern_size;
-  PyObject *pyobj_pattern_size = NULL;
-  cvpoint2d32f_count corners;
-  int flags = CV_CALIB_CB_ADAPTIVE_THRESH;
-
-  const char *keywords[] = { "image", "pattern_size", "flags", NULL };
-  if (!PyArg_ParseTupleAndKeywords(args, kw, "OO|i", (char**)keywords, &pyobj_image, &pyobj_pattern_size, &flags))
-    return NULL;
-  if (!convert_to_CvArr(pyobj_image, &image, "image")) return NULL;
-  if (!convert_to_CvSize(pyobj_pattern_size, &pattern_size, "pattern_size")) return NULL;
-  int r;
-  corners.points = new CvPoint2D32f[pattern_size.width * pattern_size.height];
-  ERRWRAP(r = cvFindChessboardCorners(image, pattern_size, corners.points,&corners.count, flags));
-  return Py_BuildValue("NN", FROM_int(r), FROM_cvpoint2d32f_count(corners));
-}
-
-// For functions GetSubRect, GetRow, GetCol.
-// recipient has a view into donor's data, and needs to share it.
-// make recipient use the donor's data, compute the offset,
-// and manage reference counts.
-
-static void preShareData(CvArr *donor, CvMat **recipient)
-{
-  *recipient = cvCreateMatHeader(4, 4, cvGetElemType(donor));
-}
-
-static PyObject *shareData(PyObject *donor, CvArr *pdonor, CvMat *precipient)
-{
-  PyObject *recipient = (PyObject*)PyObject_NEW(cvmat_t, &cvmat_Type);
-  ((cvmat_t*)recipient)->a = precipient;
-  ((cvmat_t*)recipient)->offset = cvPtr1D(precipient, 0) - cvPtr1D(pdonor, 0);
-
-  PyObject *arr_data;
-  if (is_cvmat(donor)) {
-    arr_data = ((cvmat_t*)donor)->data;
-    ((cvmat_t*)recipient)->offset += ((cvmat_t*)donor)->offset;
-  } else if (is_iplimage(donor)) {
-    arr_data = ((iplimage_t*)donor)->data;
-    ((cvmat_t*)recipient)->offset += ((iplimage_t*)donor)->offset;
-  } else {
-    return failmsg("Argument 'mat' must be either IplImage or CvMat"), (PyObject*)0;
-  }
-  ((cvmat_t*)recipient)->data = arr_data;
-  Py_INCREF(arr_data);
-  return recipient;
-}
-
-static PyObject *shareDataND(PyObject *donor, CvMatND *pdonor, CvMatND *precipient)
-{
-  PyObject *recipient = (PyObject*)PyObject_NEW(cvmatnd_t, &cvmatnd_Type);
-  ((cvmatnd_t*)recipient)->a = precipient;
-  ((cvmatnd_t*)recipient)->offset = 0;
-
-  PyObject *arr_data;
-  arr_data = ((cvmatnd_t*)donor)->data;
-  ((cvmatnd_t*)recipient)->data = arr_data;
-  Py_INCREF(arr_data);
-  return recipient;
-}
-
-static PyObject *pycvGetHuMoments(PyObject *self, PyObject *args)
-{
-  CvMoments* moments;
-  PyObject *pyobj_moments = NULL;
-
-  if (!PyArg_ParseTuple(args, "O", &pyobj_moments))
-    return NULL;
-  if (!convert_to_CvMomentsPTR(pyobj_moments, &moments, "moments")) return NULL;
-  CvHuMoments r;
-  ERRWRAP(cvGetHuMoments(moments, &r));
-  return Py_BuildValue("ddddddd", r.hu1, r.hu2, r.hu3, r.hu4, r.hu5, r.hu6, r.hu7);
-}
-
-static PyObject *pycvFitLine(PyObject *self, PyObject *args)
-{
-  cvarrseq points;
-  PyObject *pyobj_points = NULL;
-  int dist_type;
-  float param;
-  float reps;
-  float aeps;
-  float r[6];
-
-  if (!PyArg_ParseTuple(args, "Oifff", &pyobj_points, &dist_type, &param, &reps, &aeps))
-    return NULL;
-  if (!convert_to_cvarrseq(pyobj_points, &points, "points")) return NULL;
-  ERRWRAP(cvFitLine(points.mat, dist_type, param, reps, aeps, r));
-  int dimension;
-  if (strcmp("opencv-matrix", cvTypeOf(points.mat)->type_name) == 0)
-    dimension = CV_MAT_CN(cvGetElemType(points.mat));
-  else {
-    // sequence case... don't think there is a sequence of 3d points,
-    // so assume 2D
-    dimension = 2;
-  }
-  if (dimension == 2)
-    return Py_BuildValue("dddd", r[0], r[1], r[2], r[3]);
-  else
-    return Py_BuildValue("dddddd", r[0], r[1], r[2], r[3], r[4], r[5]);
-}
-
-static PyObject *pycvGetMinMaxHistValue(PyObject *self, PyObject *args)
-{
-  CvHistogram* hist;
-  PyObject *pyobj_hist = NULL;
-  float min_val;
-  float max_val;
-  int min_loc[CV_MAX_DIM];
-  int max_loc[CV_MAX_DIM];
-
-  if (!PyArg_ParseTuple(args, "O", &pyobj_hist))
-    return NULL;
-  if (!convert_to_CvHistogram(pyobj_hist, &hist, "hist")) return NULL;
-  ERRWRAP(cvGetMinMaxHistValue(hist, &min_val, &max_val, min_loc, max_loc));
-  int d = cvGetDims(hist->bins);
-  PyObject *pminloc = PyTuple_New(d), *pmaxloc = PyTuple_New(d);
-  for (int i = 0; i < d; i++) {
-    PyTuple_SetItem(pminloc, i, PyInt_FromLong(min_loc[i]));
-    PyTuple_SetItem(pmaxloc, i, PyInt_FromLong(max_loc[i]));
-  }
-  return Py_BuildValue("ffNN", min_val, max_val, pminloc, pmaxloc);
-}
-
-static CvSeq* cvHOGDetectMultiScale( const CvArr* image, CvMemStorage* storage,
-  const CvArr* svm_classifier=NULL, CvSize win_stride=cvSize(0,0),
-  double hit_threshold=0, double scale=1.05,
-  int group_threshold=2, CvSize padding=cvSize(0,0),
-  CvSize win_size=cvSize(64,128), CvSize block_size=cvSize(16,16),
-  CvSize block_stride=cvSize(8,8), CvSize cell_size=cvSize(8,8),
-  int nbins=9, int gammaCorrection=1 )
-{
-    cv::HOGDescriptor hog(win_size, block_size, block_stride, cell_size, nbins, 1, -1, cv::HOGDescriptor::L2Hys, 0.2, gammaCorrection!=0);
-    if(win_stride.width == 0 && win_stride.height == 0)
-        win_stride = block_stride;
-    cv::Mat img = cv::cvarrToMat(image);
-    std::vector<cv::Rect> found;
-    if(svm_classifier)
-    {
-        CvMat stub, *m = cvGetMat(svm_classifier, &stub);
-        int sz = m->cols*m->rows;
-        CV_Assert(CV_IS_MAT_CONT(m->type) && (m->cols == 1 || m->rows == 1) && CV_MAT_TYPE(m->type) == CV_32FC1);
-        std::vector<float> w(sz);
-        std::copy(m->data.fl, m->data.fl + sz, w.begin());
-        hog.setSVMDetector(w);
-    }
-    else
-        hog.setSVMDetector(cv::HOGDescriptor::getDefaultPeopleDetector());
-    hog.detectMultiScale(img, found, hit_threshold, win_stride, padding, scale, group_threshold);
-    CvSeq* seq = cvCreateSeq(cv::DataType<cv::Rect>::type, sizeof(CvSeq), sizeof(cv::Rect), storage);
-    if(found.size())
-        cvSeqPushMulti(seq, &found[0], (int)found.size());
-    return seq;
-}
-
-static void cvGrabCut(CvArr *image,
-                      CvArr *mask,
-                      CvRect rect,
-                      CvArr *bgdModel,
-                      CvArr *fgdModel,
-                      int iterCount,
-                      int mode)
-{
-  cv::Mat _image = cv::cvarrToMat(image);
-  cv::Mat _mask = cv::cvarrToMat(mask);
-  cv::Mat _bgdModel = cv::cvarrToMat(bgdModel);
-  cv::Mat _fgdModel = cv::cvarrToMat(fgdModel);
-  grabCut(_image, _mask, rect, _bgdModel, _fgdModel, iterCount, mode);
-}
-
-static int zero = 0;
-
-/************************************************************************/
-/* Custom Validators */
-
-#define CVPY_VALIDATE_DrawChessboardCorners() do { \
-  if ((patternSize.width * patternSize.height) != corners.count) \
-    return (PyObject*)0; \
-  } while (0)
-
-#define cvGetRotationMatrix2D cv2DRotationMatrix
-
-/************************************************************************/
-/* Generated functions */
-
-#define constCvMat const CvMat
-#define FROM_constCvMatPTR(x) FROM_CvMatPTR((CvMat*)x)
-
-#define cvSnakeImage(image, points, length, a, b, g, win, criteria, calc_gradient) \
-  do { \
-    int coeff_usage; \
-    if ((alpha.count == 1) && (beta.count == 1) && (gamma.count == 1)) \
-      coeff_usage = CV_VALUE; \
-    else if ((length == alpha.count) && (alpha.count == beta.count) && (beta.count == gamma.count)) \
-      coeff_usage = CV_ARRAY; \
-    else { \
-        failmsg("SnakeImage weights invalid"); \
-        return (PyObject*)0; } \
-    cvSnakeImage(image, points, length, a, b, g, coeff_usage, win, criteria, calc_gradient); \
-  } while (0)
-
-static double cppKMeans(const CvArr* _samples, int cluster_count, CvArr* _labels,
-           CvTermCriteria termcrit, int attempts, int flags, CvArr* _centers)
-{
-    cv::Mat data = cv::cvarrToMat(_samples), labels = cv::cvarrToMat(_labels), centers;
-    if( _centers )
-        centers = cv::cvarrToMat(_centers);
-    CV_Assert( labels.isContinuous() && labels.type() == CV_32S &&
-        (labels.cols == 1 || labels.rows == 1) &&
-        labels.cols + labels.rows - 1 == data.rows );
-    return cv::kmeans(data, cluster_count, labels, termcrit, attempts,
-                        flags, _centers ? cv::_OutputArray(centers) : cv::_OutputArray() );
-}
-
-#define cvKMeans2(samples, nclusters, labels, termcrit, attempts, flags, centers) \
-    cppKMeans(samples, nclusters, labels, termcrit, attempts, flags, centers)
-
-#if defined _MSC_VER && _MSC_VER >= 1200
-    #pragma warning( push )
-    #pragma warning( disable : 4244 )
-#endif
-
-#include "generated0.i"
-
-#if defined _MSC_VER && _MSC_VER >= 1200
-    #pragma warning( pop )
-#endif
-
-static PyMethodDef old_methods[] = {
-
-#if PYTHON_USE_NUMPY
-    {"fromarray", (PyCFunction)pycvfromarray, METH_KEYWORDS, "fromarray(array [, allowND]) -> CvMat"},
-#endif
-
-  {"FindDataMatrix", pyfinddatamatrix, METH_VARARGS},
-  {"temp_test", temp_test, METH_VARARGS},
-
-#include "generated1.i"
-
-  {NULL, NULL},
-};
-
-/************************************************************************/
-/* Module init */
-
-#include "opencv2/opencv_modules.hpp"
-
-#ifdef HAVE_OPENCV_NONFREE
-#  include "opencv2/nonfree.hpp"
-#endif
-
-#ifdef HAVE_OPENCV_FEATURES2D
-#  include "opencv2/features2d.hpp"
-#endif
-
-#ifdef HAVE_OPENCV_VIDEO
-#  include "opencv2/video.hpp"
-#endif
-
-#ifdef HAVE_OPENCV_ML
-#  include "opencv2/ml.hpp"
-#endif
-
-#ifdef HAVE_OPENCV_CONTRIB
-#  include "opencv2/contrib.hpp"
-#endif
-
-static PyObject* init_cv()
-{
-  PyObject *m, *d;
-  cvSetErrMode(CV_ErrModeParent);
-
-    bool init = true;
-#ifdef HAVE_OPENCV_NONFREE
-    init &= cv::initModule_nonfree();
-#endif
-#ifdef HAVE_OPENCV_FEATURES2D
-    init &= cv::initModule_features2d();
-#endif
-#ifdef HAVE_OPENCV_VIDEO
-    init &= cv::initModule_video();
-#endif
-#ifdef HAVE_OPENCV_ML
-    init &= cv::initModule_ml();
-#endif
-    #ifdef HAVE_OPENCV_CONTRIB
-    init &= cv::initModule_contrib();
-#endif
-  if(!init)
-    return 0;
-
-  #define MKTYPE(NAME)  NAME##_specials(); to_ok(&NAME##_Type)
-
-  MKTYPE(cvcontourtree);
-  MKTYPE(cvfont);
-  MKTYPE(cvhistogram);
-  MKTYPE(cvlineiterator);
-  MKTYPE(cvmat);
-  MKTYPE(cvmatnd);
-  MKTYPE(cvmemstorage);
-  MKTYPE(cvsubdiv2dedge);
-  MKTYPE(cvrng);
-  MKTYPE(cvseq);
-  MKTYPE(cvset);
-  MKTYPE(cvsubdiv2d);
-  MKTYPE(cvsubdiv2dpoint);
-  MKTYPE(iplimage);
-  MKTYPE(memtrack);
-
-#include "generated4.i"
-
-  #undef MKTYPE
-
-  m = Py_InitModule(OLD_MODULESTR, old_methods);
-  d = PyModule_GetDict(m);
-
-  PyDict_SetItemString(d, "__version__", PyString_FromString(CV_VERSION));
-  PyDict_SetItemString(d, "error", opencv_error);
-
-  // Couple of warnings about strict aliasing here.  Not clear how to fix.
-  union {
-    PyObject *o;
-    PyTypeObject *to;
-  } convert;
-  convert.to = &iplimage_Type;
-  PyDict_SetItemString(d, "iplimage", convert.o);
-  convert.to = &cvmat_Type;
-  PyDict_SetItemString(d, "cvmat", convert.o);
-
-  // AFAIK the only floating-point constant
-  PyDict_SetItemString(d, "CV_PI", PyFloat_FromDouble(CV_PI));
-
-#define PUBLISH(I) PyDict_SetItemString(d, #I, PyInt_FromLong(I))
-#define PUBLISHU(I) PyDict_SetItemString(d, #I, PyLong_FromUnsignedLong(I))
-#define PUBLISH2(I, value) PyDict_SetItemString(d, #I, PyLong_FromLong(value))
-
-  PUBLISHU(IPL_DEPTH_8U);
-  PUBLISHU(IPL_DEPTH_8S);
-  PUBLISHU(IPL_DEPTH_16U);
-  PUBLISHU(IPL_DEPTH_16S);
-  PUBLISHU(IPL_DEPTH_32S);
-  PUBLISHU(IPL_DEPTH_32F);
-  PUBLISHU(IPL_DEPTH_64F);
-
-  PUBLISH(CV_LOAD_IMAGE_COLOR);
-  PUBLISH(CV_LOAD_IMAGE_GRAYSCALE);
-  PUBLISH(CV_LOAD_IMAGE_UNCHANGED);
-  PUBLISH(CV_HIST_ARRAY);
-  PUBLISH(CV_HIST_SPARSE);
-  PUBLISH(CV_8U);
-  PUBLISH(CV_8UC1);
-  PUBLISH(CV_8UC2);
-  PUBLISH(CV_8UC3);
-  PUBLISH(CV_8UC4);
-  PUBLISH(CV_8S);
-  PUBLISH(CV_8SC1);
-  PUBLISH(CV_8SC2);
-  PUBLISH(CV_8SC3);
-  PUBLISH(CV_8SC4);
-  PUBLISH(CV_16U);
-  PUBLISH(CV_16UC1);
-  PUBLISH(CV_16UC2);
-  PUBLISH(CV_16UC3);
-  PUBLISH(CV_16UC4);
-  PUBLISH(CV_16S);
-  PUBLISH(CV_16SC1);
-  PUBLISH(CV_16SC2);
-  PUBLISH(CV_16SC3);
-  PUBLISH(CV_16SC4);
-  PUBLISH(CV_32S);
-  PUBLISH(CV_32SC1);
-  PUBLISH(CV_32SC2);
-  PUBLISH(CV_32SC3);
-  PUBLISH(CV_32SC4);
-  PUBLISH(CV_32F);
-  PUBLISH(CV_32FC1);
-  PUBLISH(CV_32FC2);
-  PUBLISH(CV_32FC3);
-  PUBLISH(CV_32FC4);
-  PUBLISH(CV_64F);
-  PUBLISH(CV_64FC1);
-  PUBLISH(CV_64FC2);
-  PUBLISH(CV_64FC3);
-  PUBLISH(CV_64FC4);
-  PUBLISH(CV_NEXT_AROUND_ORG);
-  PUBLISH(CV_NEXT_AROUND_DST);
-  PUBLISH(CV_PREV_AROUND_ORG);
-  PUBLISH(CV_PREV_AROUND_DST);
-  PUBLISH(CV_NEXT_AROUND_LEFT);
-  PUBLISH(CV_NEXT_AROUND_RIGHT);
-  PUBLISH(CV_PREV_AROUND_LEFT);
-  PUBLISH(CV_PREV_AROUND_RIGHT);
-
-  PUBLISH(CV_WINDOW_AUTOSIZE);
-
-  PUBLISH(CV_PTLOC_INSIDE);
-  PUBLISH(CV_PTLOC_ON_EDGE);
-  PUBLISH(CV_PTLOC_VERTEX);
-  PUBLISH(CV_PTLOC_OUTSIDE_RECT);
-
-  PUBLISH(GC_BGD);
-  PUBLISH(GC_FGD);
-  PUBLISH(GC_PR_BGD);
-  PUBLISH(GC_PR_FGD);
-  PUBLISH(GC_INIT_WITH_RECT);
-  PUBLISH(GC_INIT_WITH_MASK);
-  PUBLISH(GC_EVAL);
-
-#include "generated2.i"
-
-#undef PUBLISH
-#undef PUBLISHU
-#undef PUBLISH2
-
-  return m;
-}
-
diff --git a/modules/python/src2/defs b/modules/python/src2/defs
deleted file mode 100644
index 0298e5be0a..0000000000
--- a/modules/python/src2/defs
+++ /dev/null
@@ -1,360 +0,0 @@
-#define CV_BLUR_NO_SCALE 0
-#define CV_BLUR  1
-#define CV_GAUSSIAN  2
-#define CV_MEDIAN 3
-#define CV_BILATERAL 4
-#define CV_INPAINT_NS      0
-#define CV_INPAINT_TELEA   1
-#define CV_SCHARR -1
-#define CV_MAX_SOBEL_KSIZE 7
-#define  CV_BGR2BGRA    0
-#define  CV_RGB2RGBA    CV_BGR2BGRA
-#define  CV_BGRA2BGR    1
-#define  CV_RGBA2RGB    CV_BGRA2BGR
-#define  CV_BGR2RGBA    2
-#define  CV_RGB2BGRA    CV_BGR2RGBA
-#define  CV_RGBA2BGR    3
-#define  CV_BGRA2RGB    CV_RGBA2BGR
-#define  CV_BGR2RGB     4
-#define  CV_RGB2BGR     CV_BGR2RGB
-#define  CV_BGRA2RGBA   5
-#define  CV_RGBA2BGRA   CV_BGRA2RGBA
-#define  CV_BGR2GRAY    6
-#define  CV_RGB2GRAY    7
-#define  CV_GRAY2BGR    8
-#define  CV_GRAY2RGB    CV_GRAY2BGR
-#define  CV_GRAY2BGRA   9
-#define  CV_GRAY2RGBA   CV_GRAY2BGRA
-#define  CV_BGRA2GRAY   10
-#define  CV_RGBA2GRAY   11
-#define  CV_BGR2BGR565  12
-#define  CV_RGB2BGR565  13
-#define  CV_BGR5652BGR  14
-#define  CV_BGR5652RGB  15
-#define  CV_BGRA2BGR565 16
-#define  CV_RGBA2BGR565 17
-#define  CV_BGR5652BGRA 18
-#define  CV_BGR5652RGBA 19
-#define  CV_GRAY2BGR565 20
-#define  CV_BGR5652GRAY 21
-#define  CV_BGR2BGR555  22
-#define  CV_RGB2BGR555  23
-#define  CV_BGR5552BGR  24
-#define  CV_BGR5552RGB  25
-#define  CV_BGRA2BGR555 26
-#define  CV_RGBA2BGR555 27
-#define  CV_BGR5552BGRA 28
-#define  CV_BGR5552RGBA 29
-#define  CV_GRAY2BGR555 30
-#define  CV_BGR5552GRAY 31
-#define  CV_BGR2XYZ     32
-#define  CV_RGB2XYZ     33
-#define  CV_XYZ2BGR     34
-#define  CV_XYZ2RGB     35
-#define  CV_BGR2YCrCb   36
-#define  CV_RGB2YCrCb   37
-#define  CV_YCrCb2BGR   38
-#define  CV_YCrCb2RGB   39
-#define  CV_BGR2HSV     40
-#define  CV_RGB2HSV     41
-#define  CV_BGR2Lab     44
-#define  CV_RGB2Lab     45
-#define  CV_BayerBG2BGR 46
-#define  CV_BayerGB2BGR 47
-#define  CV_BayerRG2BGR 48
-#define  CV_BayerGR2BGR 49
-#define  CV_BayerBG2RGB CV_BayerRG2BGR
-#define  CV_BayerGB2RGB CV_BayerGR2BGR
-#define  CV_BayerRG2RGB CV_BayerBG2BGR
-#define  CV_BayerGR2RGB CV_BayerGB2BGR
-#define  CV_BayerBG2BGR_VNG 62
-#define  CV_BayerGB2BGR_VNG 63
-#define  CV_BayerRG2BGR_VNG 64
-#define  CV_BayerGR2BGR_VNG 65
-#define  CV_BGR2Luv     50
-#define  CV_RGB2Luv     51
-#define  CV_BGR2HLS     52
-#define  CV_RGB2HLS     53
-#define  CV_HSV2BGR     54
-#define  CV_HSV2RGB     55
-#define  CV_Lab2BGR     56
-#define  CV_Lab2RGB     57
-#define  CV_Luv2BGR     58
-#define  CV_Luv2RGB     59
-#define  CV_HLS2BGR     60
-#define  CV_HLS2RGB     61
-#define  CV_COLORCVT_MAX  100
-#define  CV_INTER_NN        0
-#define  CV_INTER_LINEAR    1
-#define  CV_INTER_CUBIC     2
-#define  CV_INTER_AREA      3
-#define  CV_WARP_FILL_OUTLIERS 8
-#define  CV_WARP_INVERSE_MAP  16
-#define  CV_SHAPE_RECT      0
-#define  CV_SHAPE_CROSS     1
-#define  CV_SHAPE_ELLIPSE   2
-#define  CV_SHAPE_CUSTOM    100
-#define CV_MOP_OPEN         2
-#define CV_MOP_CLOSE        3
-#define CV_MOP_GRADIENT     4
-#define CV_MOP_TOPHAT       5
-#define CV_MOP_BLACKHAT     6
-#define  CV_TM_SQDIFF        0
-#define  CV_TM_SQDIFF_NORMED 1
-#define  CV_TM_CCORR         2
-#define  CV_TM_CCORR_NORMED  3
-#define  CV_TM_CCOEFF        4
-#define  CV_TM_CCOEFF_NORMED 5
-#define  CV_LKFLOW_PYR_A_READY       1
-#define  CV_LKFLOW_PYR_B_READY       2
-#define  CV_LKFLOW_INITIAL_GUESSES   4
-#define  CV_LKFLOW_GET_MIN_EIGENVALS 8
-#define CV_POLY_APPROX_DP 0
-#define CV_CONTOURS_MATCH_I1  1
-#define CV_CONTOURS_MATCH_I2  2
-#define CV_CONTOURS_MATCH_I3  3
-#define CV_CLOCKWISE         1
-#define CV_COUNTER_CLOCKWISE 2
-#define CV_COMP_CORREL        0
-#define CV_COMP_CHISQR        1
-#define CV_COMP_INTERSECT     2
-#define CV_COMP_BHATTACHARYYA 3
-#define CV_DIST_MASK_3   3
-#define CV_DIST_MASK_5   5
-#define CV_DIST_MASK_PRECISE 0
-#define CV_THRESH_BINARY      0  /* value = value > threshold ? max_value : 0       */
-#define CV_THRESH_BINARY_INV  1  /* value = value > threshold ? 0 : max_value       */
-#define CV_THRESH_TRUNC       2  /* value = value > threshold ? threshold : value   */
-#define CV_THRESH_TOZERO      3  /* value = value > threshold ? value : 0           */
-#define CV_THRESH_TOZERO_INV  4  /* value = value > threshold ? 0 : value           */
-#define CV_THRESH_MASK        7
-#define CV_THRESH_OTSU        8  /* use Otsu algorithm to choose the optimal threshold value;
-#define CV_ADAPTIVE_THRESH_MEAN_C  0
-#define CV_ADAPTIVE_THRESH_GAUSSIAN_C  1
-#define CV_FLOODFILL_FIXED_RANGE (1 << 16)
-#define CV_FLOODFILL_MASK_ONLY   (1 << 17)
-#define CV_CANNY_L2_GRADIENT  (1 << 31)
-#define CV_HOUGH_STANDARD 0
-#define CV_HOUGH_PROBABILISTIC 1
-#define CV_HOUGH_MULTI_SCALE 2
-#define CV_HOUGH_GRADIENT 3
-#define CV_HAAR_DO_CANNY_PRUNING    1
-#define CV_HAAR_SCALE_IMAGE         2
-#define CV_HAAR_FIND_BIGGEST_OBJECT 4 
-#define CV_HAAR_DO_ROUGH_SEARCH     8
-#define CV_LMEDS 4
-#define CV_RANSAC 8
-#define CV_CALIB_CB_ADAPTIVE_THRESH  1
-#define CV_CALIB_CB_NORMALIZE_IMAGE  2
-#define CV_CALIB_CB_FILTER_QUADS     4 
-#define CV_CALIB_USE_INTRINSIC_GUESS  1
-#define CV_CALIB_FIX_ASPECT_RATIO     2
-#define CV_CALIB_FIX_PRINCIPAL_POINT  4
-#define CV_CALIB_ZERO_TANGENT_DIST    8
-#define CV_CALIB_FIX_FOCAL_LENGTH 16
-#define CV_CALIB_FIX_K1  32
-#define CV_CALIB_FIX_K2  64
-#define CV_CALIB_FIX_K3  128
-#define CV_CALIB_FIX_INTRINSIC  256
-#define CV_CALIB_SAME_FOCAL_LENGTH 512
-#define CV_CALIB_ZERO_DISPARITY 1024
-#define CV_FM_7POINT 1
-#define CV_FM_8POINT 2
-#define CV_FM_LMEDS_ONLY  CV_LMEDS
-#define CV_FM_RANSAC_ONLY CV_RANSAC
-#define CV_FM_LMEDS CV_LMEDS
-#define CV_FM_RANSAC CV_RANSAC
-#define CV_STEREO_BM_NORMALIZED_RESPONSE  0
-#define CV_STEREO_BM_BASIC 0
-#define CV_STEREO_BM_FISH_EYE 1
-#define CV_STEREO_BM_NARROW 2
-#define CV_STEREO_GC_OCCLUDED  SHRT_MAX
-#define CV_AUTOSTEP  0x7fffffff
-#define CV_MAX_ARR 10
-#define CV_NO_DEPTH_CHECK     1
-#define CV_NO_CN_CHECK        2
-#define CV_NO_SIZE_CHECK      4
-#define CV_CMP_EQ   0
-#define CV_CMP_GT   1
-#define CV_CMP_GE   2
-#define CV_CMP_LT   3
-#define CV_CMP_LE   4
-#define CV_CMP_NE   5
-#define  CV_CHECK_RANGE    1
-#define  CV_CHECK_QUIET    2
-#define CV_RAND_UNI      0
-#define CV_RAND_NORMAL   1
-#define CV_SORT_EVERY_ROW 0
-#define CV_SORT_EVERY_COLUMN 1
-#define CV_SORT_ASCENDING 0
-#define CV_SORT_DESCENDING 16
-#define CV_GEMM_A_T 1
-#define CV_GEMM_B_T 2
-#define CV_GEMM_C_T 4
-#define CV_SVD_MODIFY_A   1
-#define CV_SVD_U_T        2
-#define CV_SVD_V_T        4
-#define CV_LU  0
-#define CV_SVD 1
-#define CV_SVD_SYM 2
-#define CV_CHOLESKY 3
-#define CV_QR  4
-#define CV_NORMAL 16
-#define CV_COVAR_SCRAMBLED 0
-#define CV_COVAR_NORMAL    1
-#define CV_COVAR_USE_AVG   2
-#define CV_COVAR_SCALE     4
-#define CV_COVAR_ROWS      8
-#define CV_COVAR_COLS     16
-#define CV_PCA_DATA_AS_ROW 0
-#define CV_PCA_DATA_AS_COL 1
-#define CV_PCA_USE_AVG 2
-#define CV_C            1
-#define CV_L1           2
-#define CV_L2           4
-#define CV_NORM_MASK    7
-#define CV_RELATIVE     8
-#define CV_DIFF         16
-#define CV_MINMAX       32
-#define CV_DIFF_C       (CV_DIFF | CV_C)
-#define CV_DIFF_L1      (CV_DIFF | CV_L1)
-#define CV_DIFF_L2      (CV_DIFF | CV_L2)
-#define CV_RELATIVE_C   (CV_RELATIVE | CV_C)
-#define CV_RELATIVE_L1  (CV_RELATIVE | CV_L1)
-#define CV_RELATIVE_L2  (CV_RELATIVE | CV_L2)
-#define CV_REDUCE_SUM 0
-#define CV_REDUCE_AVG 1
-#define CV_REDUCE_MAX 2
-#define CV_REDUCE_MIN 3
-#define CV_DXT_FORWARD  0
-#define CV_DXT_INVERSE  1
-#define CV_DXT_SCALE    2 /* divide result by size of array */
-#define CV_DXT_INV_SCALE (CV_DXT_INVERSE + CV_DXT_SCALE)
-#define CV_DXT_INVERSE_SCALE CV_DXT_INV_SCALE
-#define CV_DXT_ROWS     4 /* transform each row individually */
-#define CV_DXT_MUL_CONJ 8 /* conjugate the second argument of cvMulSpectrums */
-#define CV_FRONT 1
-#define CV_BACK 0
-#define  CV_GRAPH_VERTEX        1
-#define  CV_GRAPH_TREE_EDGE     2
-#define  CV_GRAPH_BACK_EDGE     4
-#define  CV_GRAPH_FORWARD_EDGE  8
-#define  CV_GRAPH_CROSS_EDGE    16
-#define  CV_GRAPH_ANY_EDGE      30
-#define  CV_GRAPH_NEW_TREE      32
-#define  CV_GRAPH_BACKTRACKING  64
-#define  CV_GRAPH_OVER          -1
-#define  CV_GRAPH_ALL_ITEMS    -1
-#define  CV_GRAPH_ITEM_VISITED_FLAG  (1 << 30)
-#define  CV_GRAPH_SEARCH_TREE_NODE_FLAG   (1 << 29)
-#define  CV_GRAPH_FORWARD_EDGE_FLAG       (1 << 28)
-#define CV_FILLED -1
-#define CV_AA 16
-#define CV_FONT_HERSHEY_SIMPLEX         0
-#define CV_FONT_HERSHEY_PLAIN           1
-#define CV_FONT_HERSHEY_DUPLEX          2
-#define CV_FONT_HERSHEY_COMPLEX         3
-#define CV_FONT_HERSHEY_TRIPLEX         4
-#define CV_FONT_HERSHEY_COMPLEX_SMALL   5
-#define CV_FONT_HERSHEY_SCRIPT_SIMPLEX  6
-#define CV_FONT_HERSHEY_SCRIPT_COMPLEX  7
-#define CV_FONT_ITALIC                 16
-#define CV_FONT_VECTOR0    CV_FONT_HERSHEY_SIMPLEX
-#define CV_KMEANS_USE_INITIAL_LABELS    1
-#define CV_ErrModeLeaf     0   /* Print error and exit program */
-#define CV_ErrModeParent   1   /* Print error and continue */
-#define CV_ErrModeSilent   2   /* Don't print and continue */
-#define CV_RETR_EXTERNAL 0
-#define CV_RETR_LIST     1
-#define CV_RETR_CCOMP    2
-#define CV_RETR_TREE     3
-#define CV_CHAIN_CODE               0
-#define CV_CHAIN_APPROX_NONE        1
-#define CV_CHAIN_APPROX_SIMPLE      2
-#define CV_CHAIN_APPROX_TC89_L1     3
-#define CV_CHAIN_APPROX_TC89_KCOS   4
-#define CV_LINK_RUNS                5
-#define CV_SUBDIV2D_VIRTUAL_POINT_FLAG (1 << 30)
-#define CV_DIST_USER    -1  /* User defined distance */
-#define CV_DIST_L1      1   /* distance = |x1-x2| + |y1-y2| */
-#define CV_DIST_L2      2   /* the simple euclidean distance */
-#define CV_DIST_C       3   /* distance = max(|x1-x2|,|y1-y2|) */
-#define CV_DIST_L12     4   /* L1-L2 metric: distance = 2(sqrt(1+x*x/2) - 1)) */
-#define CV_DIST_FAIR    5   /* distance = c^2(|x|/c-log(1+|x|/c)), c = 1.3998 */
-#define CV_DIST_WELSCH  6   /* distance = c^2/2(1-exp(-(x/c)^2)), c = 2.9846 */
-#define CV_DIST_HUBER   7   /* distance = |x|<c ? x^2/2 : c(|x|-c/2), c=1.345 */
-#define CV_HAAR_MAGIC_VAL    0x42500000
-#define CV_HAAR_FEATURE_MAX  3
-#define CV_TERMCRIT_ITER    1
-#define CV_TERMCRIT_NUMBER  CV_TERMCRIT_ITER
-#define CV_TERMCRIT_EPS     2
-#define CV_EVENT_MOUSEMOVE      0
-#define CV_EVENT_LBUTTONDOWN    1
-#define CV_EVENT_RBUTTONDOWN    2
-#define CV_EVENT_MBUTTONDOWN    3
-#define CV_EVENT_LBUTTONUP      4
-#define CV_EVENT_RBUTTONUP      5
-#define CV_EVENT_MBUTTONUP      6
-#define CV_EVENT_LBUTTONDBLCLK  7
-#define CV_EVENT_RBUTTONDBLCLK  8
-#define CV_EVENT_MBUTTONDBLCLK  9
-#define CV_EVENT_FLAG_LBUTTON   1
-#define CV_EVENT_FLAG_RBUTTON   2
-#define CV_EVENT_FLAG_MBUTTON   4
-#define CV_EVENT_FLAG_CTRLKEY   8
-#define CV_EVENT_FLAG_SHIFTKEY  16
-#define CV_EVENT_FLAG_ALTKEY    32
-#define CV_MAX_DIM            32
-#define CV_CAP_PROP_POS_MSEC       0
-#define CV_CAP_PROP_POS_FRAMES     1
-#define CV_CAP_PROP_POS_AVI_RATIO  2
-#define CV_CAP_PROP_FRAME_WIDTH    3
-#define CV_CAP_PROP_FRAME_HEIGHT   4
-#define CV_CAP_PROP_FPS            5
-#define CV_CAP_PROP_FOURCC         6
-#define CV_CAP_PROP_FRAME_COUNT    7
-#define CV_CAP_PROP_FORMAT         8
-#define CV_CAP_PROP_MODE           9
-#define CV_CAP_PROP_BRIGHTNESS    10
-#define CV_CAP_PROP_CONTRAST      11
-#define CV_CAP_PROP_SATURATION    12
-#define CV_CAP_PROP_HUE           13
-#define CV_CAP_PROP_GAIN          14
-#define CV_CAP_PROP_EXPOSURE      15
-#define CV_CAP_PROP_CONVERT_RGB   16
-#define CV_CAP_PROP_RECTIFICATION 18
-#define CV_CAP_OPENNI 900
-#define CV_CAP_OPENNI_DEPTH_GENERATOR 2147483648
-#define CV_CAP_OPENNI_IMAGE_GENERATOR 1073741824
-#define CV_CAP_OPENNI_DEPTH_MAP 0
-#define CV_CAP_OPENNI_POINT_CLOUD_MAP 1
-#define CV_CAP_OPENNI_DISPARITY_MAP 2
-#define CV_CAP_OPENNI_DISPARITY_MAP_32F 3
-#define CV_CAP_OPENNI_VALID_DEPTH_MASK 4
-#define CV_CAP_OPENNI_BGR_IMAGE 5
-#define CV_CAP_OPENNI_GRAY_IMAGE 6
-#define CV_CAP_PROP_OPENNI_OUTPUT_MODE 100
-#define CV_CAP_OPENNI_VGA_30HZ 0
-#define CV_CAP_OPENNI_SXGA_15HZ 1
-#define CV_CAP_PROP_OPENNI_REGISTRATION 104
-#define CV_CAP_PROP_OPENNI_FRAME_MAX_DEPTH 101
-#define CV_CAP_PROP_OPENNI_BASELINE 102
-#define CV_CAP_PROP_OPENNI_FOCAL_LENGTH 103
-#define CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE 1073741924
-#define CV_CAP_OPENNI_DEPTH_GENERATOR_BASELINE 2147483750
-#define CV_CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH 2147483751
-#define CV_CAP_OPENNI_DEPTH_GENERATOR_REGISTRATION 2147483752
-#define CV_CN_SHIFT   3
-#define CV_IMWRITE_JPEG_QUALITY 1
-#define CV_IMWRITE_PNG_COMPRESSION 16
-#define CV_IMWRITE_PXM_BINARY 32
-#define IPL_ORIGIN_TL 0
-#define IPL_ORIGIN_BL 1
-#define CV_GAUSSIAN_5x5
-#define CV_CN_MAX
-#define CV_WINDOW_AUTOSIZE  1
-#define CV_WINDOW_NORMAL	 	 0
-#define CV_WINDOW_FULLSCREEN	 1
-#define HG_AUTOSIZE CV_WINDOW_AUTOSIZE
-#define CV_CVTIMG_FLIP  1
-#define CV_CVTIMG_SWAP_RB 2
diff --git a/modules/python/src2/gen.py b/modules/python/src2/gen.py
deleted file mode 100755
index 65cafc9900..0000000000
--- a/modules/python/src2/gen.py
+++ /dev/null
@@ -1,631 +0,0 @@
-#/usr/bin/env python
-
-import sys
-from string import Template
-
-class argument:
-  def __init__(self, fields):
-    self.ty = fields[0]
-    self.nm = fields[1]
-    self.flags = ""
-    self.init = None
-
-    if len(fields) > 2:
-      if fields[2][0] == '/':
-        self.flags = fields[2][1:].split(",")
-      else:
-        self.init = fields[2]
-
-api = []
-for l in open("%s/api" % sys.argv[1]):
-  if l[0] == '#':
-    continue
-  l = l.rstrip()
-  if (not l.startswith(' ')) and ('/' in l):
-    (l, flags) = l.split('/')
-  else:
-    flags = ""
-  f = l.split()
-  if len(f) != 0:
-    if l[0] != ' ':
-      if len(f) > 1:
-        ty = f[1]
-      else:
-        ty = None
-      api.append((f[0], [], ty, flags))
-    else:
-      api[-1][1].append(argument(f))
-
-# Validation: check that any optional arguments are last
-had_error = False
-for (f, args, ty, flags) in api:
-    has_init = [(a.init != None) for a in args if not 'O' in a.flags]
-    if True in has_init and not all(has_init[has_init.index(True):]):
-        print 'Error in definition for "%s", optional arguments must be last' % f
-        had_error = True
-
-if had_error:
-    sys.exit(1)
-
-def cname(n):
-  if n.startswith("CV"):
-    return '_' + n
-  elif n[0].isdigit():
-    return '_' + n
-  else:
-    return n
-
-# RHS is how the aggregate gets expanded in the C call
-aggregate = {
-  'pts_npts_contours' :  '!.pts,!.npts,!.contours',
-  'cvarr_count' :        '!.cvarr,!.count',
-  'cvarr_plane_count' :  '!.cvarr,!.count',
-  'floats' :             '!.f',
-  'ints' :               '!.i',
-  'ints0' :              '!.i',
-  'CvPoints' :           '!.p,!.count',
-  'CvPoint2D32fs' :      '!.p,!.count',
-  'CvPoint3D32fs' :      '!.p,!.count',
-  'cvarrseq' :           '!.seq',
-  'CvArrs' :             '!.ims',
-  'IplImages' :          '!.ims',
-  'intpair' :            '!.pairs,!.count',
-  'cvpoint2d32f_count' : '!.points,&!.count'
-}
-conversion_types = [
-'char',
-'CvArr',
-'CvArrSeq',
-'CvBox2D', # '((ff)(ff)f)',
-'CvBox2D*',
-'CvCapture*',
-'CvStereoBMState*',
-'CvStereoGCState*',
-'CvKalman*',
-'CvVideoWriter*',
-'CvContourTree*',
-'CvFont',
-'CvFont*',
-'CvHaarClassifierCascade*',
-'CvHistogram',
-'CvMat',
-'CvMatND',
-'CvMemStorage',
-'CvMoments',
-'CvMoments*',
-'CvNextEdgeType',
-'CvPoint',
-'CvPoint*',
-'CvPoint2D32f', # '(ff)',
-'CvPoint2D32f*',
-'CvPoint3D32f*',
-'CvPoint2D64f',
-'CvPOSITObject*',
-'CvRect',
-'CvRect*',
-'CvRNG*',
-'CvScalar',
-'CvSeq',
-'CvSeqOfCvConvexityDefect',
-'CvSize',
-'CvSlice',
-'CvStarDetectorParams',
-'CvSubdiv2D*',
-'CvSubdiv2DEdge',
-'CvTermCriteria',
-'generic',
-'IplConvKernel*',
-'IplImage',
-'PyObject*',
-'PyCallableObject*'
-]
-
-def safename(s):
-  return s.replace('*', 'PTR').replace('[', '_').replace(']', '_')
-
-def has_optional(al):
-    """ return true if any argument is optional """
-    return any([a.init for a in al])
-
-def gen(name, args, ty, flags):
-  yield ""
-  if has_optional(args):
-      yield "static PyObject *pycv%s(PyObject *self, PyObject *args, PyObject *kw)" % cname(name)
-  else:
-      yield "static PyObject *pycv%s(PyObject *self, PyObject *args)" % cname(name)
-  if 'doconly' in flags:
-    yield ";"
-  else:
-    yield "{"
-
-    destinations = []
-    for a in args:
-      remap = {
-       'CvArr' : 'CvArr*',
-       'CvMat' : 'CvMat*',
-       'CvMatND' : 'CvMatND*',
-       'IplImage' : 'IplImage*',
-       'CvMemStorage' : 'CvMemStorage*',
-       'CvHistogram':'CvHistogram*',
-       'CvSeq':'CvSeq*',
-       'CvHaarClassifierCascade' : 'CvHaarClassifierCascade*'
-      }
-      ctype = remap.get(a.ty, a.ty)
-      if a.init:
-        init = " = %s" % a.init
-      else:
-        init = ''
-      yield "  %s %s%s;" % (ctype, a.nm, init)
-      if 'O' in a.flags:
-        continue
-      if a.ty in (conversion_types + aggregate.keys()):
-        yield '  PyObject *pyobj_%s = NULL;' % (a.nm)
-        destinations.append('&pyobj_%s' % (a.nm))
-      elif a.ty in [ 'CvPoint2D32f' ]:
-        destinations.append('&%s.x, &%s.y' % (a.nm, a.nm))
-      elif a.ty in [ 'CvTermCriteria' ]:
-        destinations.append('&%s.type, &%s.max_iter, &%s.epsilon' % ((a.nm,)*3))
-      elif a.ty in [ 'CvSURFParams' ]:
-        destinations.append('&%s.extended, &%s.hessianThreshold, &%s.nOctaves, &%s.nOctaveLayers' % ((a.nm,)*4))
-      elif a.nm in [ 'CvBox2D' ]:
-        s = ", ".join([('&' + a.nm +'.' + fld) for fld in [ 'center.x', 'center.y', 'size.width', 'size.height', 'angle' ] ])
-        destinations.append(s)
-      else:
-        destinations.append('&%s' % a.nm)
-    fmap = {
-      'CvSURFParams' : '(idii)',
-      'double' : 'd',
-      'float' : 'f',
-      'int' : 'i',
-      'int64' : 'L',
-      'char*' : 's',
-    }
-    for k in (conversion_types + aggregate.keys()):
-      fmap[k] = 'O'
-    in_args = [ a for a in args if not 'O' in a.flags ]
-    fmt0 = "".join([ fmap[a.ty] for a in in_args if not a.init])
-    fmt1 = "".join([ fmap[a.ty] for a in in_args if a.init])
-
-    yield ''
-    if len(fmt0 + fmt1) > 0:
-      if len(fmt1) > 0:
-        yield '  const char *keywords[] = { %s };' % (", ".join([ '"%s"' % arg.nm for arg in args if not 'O' in arg.flags ] + ['NULL']))
-        yield '  if (!PyArg_ParseTupleAndKeywords(args, kw, "%s|%s", %s))' % (fmt0, fmt1, ", ".join(['(char**)keywords'] + destinations))
-        if '(' in (fmt0 + fmt1):
-          print "Tuple with kwargs is not allowed, function", name
-          sys.exit(1)
-      else:
-        yield '  if (!PyArg_ParseTuple(args, "%s", %s))' % (fmt0, ", ".join(destinations))
-      yield '    return NULL;'
-
-    # Do the conversions:
-    for a in args:
-      joinwith = [f[2:] for f in a.flags if f.startswith("J:")]
-      if len(joinwith) > 0:
-        yield 'preShareData(%s, &%s);' % (joinwith[0], a.nm)
-      if 'O' in a.flags:
-        continue
-      if a.ty in (conversion_types + aggregate.keys()):
-        if a.init:
-          pred = '(pyobj_%s != NULL) && ' % a.nm
-        else:
-          pred = ''
-        yield '  if (%s!convert_to_%s(pyobj_%s, &%s, "%s")) return NULL;' % (pred, safename(a.ty), a.nm, a.nm, a.nm)
-
-    yield '#ifdef CVPY_VALIDATE_%s' % name
-    yield 'CVPY_VALIDATE_%s();' % name
-    yield '#endif'
-
-    def invokename(a):
-      if 'K' in a.flags:
-        prefix = "(const CvArr **)"
-      elif 'O' in a.flags and not 'A' in a.flags:
-        prefix = "&"
-      else:
-        prefix = ""
-      if a.ty in aggregate:
-        return prefix + aggregate[a.ty].replace('!', a.nm)
-      else:
-        return prefix + a.nm
-
-    def funcname(s):
-      # The name by which the function is called, in C
-      if s.startswith("CV"):
-        return s
-      else:
-        return "cv" + s
-    tocall = '%s(%s)' % (funcname(name), ", ".join(invokename(a) for a in args))
-    if 'stub' in flags:
-      yield '  return stub%s(%s);' % (name, ", ".join(invokename(a) for a in args))
-    elif ty == None:
-      yield '  ERRWRAP(%s);' % tocall
-      yield '  Py_RETURN_NONE;'
-    else:
-      Rtypes = [
-        'int',
-        'int64',
-        'double',
-        'CvCapture*',
-        'CvVideoWriter*',
-        'CvPOSITObject*',
-        'CvScalar',
-        'CvSize',
-        'CvRect',
-        'CvSeq*',
-        'CvBox2D',
-        'CvSeqOfCvAvgComp*',
-        'CvSeqOfCvConvexityDefect*',
-        'CvSeqOfCvStarKeypoint*',
-        'CvSeqOfCvSURFPoint*',
-        'CvSeqOfCvSURFDescriptor*',
-        'CvContourTree*',
-        'IplConvKernel*',
-        'IplImage*',
-        'CvMat*',
-        'constCvMat*',
-        'ROCvMat*',
-        'CvMatND*',
-        'CvPoint2D32f_4',
-        'CvRNG',
-        'CvSubdiv2D*',
-        'CvSubdiv2DPoint*',
-        'CvSubdiv2DEdge',
-        'ROIplImage*',
-        'CvStereoBMState*',
-        'CvStereoGCState*',
-        'CvKalman*',
-        'float',
-        'generic',
-        'unsigned' ]
-
-      if ty in Rtypes:
-        yield '  %s r;' % (ty)
-        yield '  ERRWRAP(r = %s);' % (tocall)
-        yield '  return FROM_%s(r);' % safename(ty)
-      else:
-        all_returns = ty.split(",")
-        return_value_from_call = len(set(Rtypes) & set(all_returns)) != 0
-        if return_value_from_call:
-          yield '  %s r;' % list(set(Rtypes) & set(all_returns))[0]
-          yield '  ERRWRAP(r = %s);' % (tocall)
-        else:
-          yield '  ERRWRAP(%s);' % (tocall)
-        typed = dict([ (a.nm,a.ty) for a in args])
-        for i in range(len(all_returns)):
-          if all_returns[i] in Rtypes:
-            typed['r'] = all_returns[i]
-            all_returns[i] = "r"
-        if len(all_returns) == 1:
-          af = dict([ (a.nm,a.flags) for a in args])
-          joinwith = [f[2:] for f in af.get(all_returns[0], []) if f.startswith("J:")]
-          if len(joinwith) > 0:
-              yield '  return shareData(pyobj_%s, %s, %s);' % (joinwith[0], joinwith[0], all_returns[0])
-          else:
-              yield '  return FROM_%s(%s);' % (safename(typed[all_returns[0]]), all_returns[0])
-        else:
-          yield '  return Py_BuildValue("%s", %s);' % ("N" * len(all_returns), ", ".join(["FROM_%s(%s)" % (safename(typed[n]), n) for n in all_returns]))
-
-    yield '}'
-
-gen_c = [ open("generated%d.i" % i, "w") for i in range(5) ]
-
-print "Generated %d functions" % len(api)
-for nm,args,ty,flags in sorted(api):
-
-  # Figure out docstring into ds_*
-  ds_args = []
-  mandatory = [a.nm for a in args if not ('O' in a.flags) and not a.init]
-  optional = [a.nm for a in args if not ('O' in a.flags) and a.init]
-  ds_args = ", ".join(mandatory)
-  def o2s(o):
-    if o == []:
-        return ""
-    else:
-        return ' [, %s%s]' % (o[0], o2s(o[1:]))
-  ds_args += o2s(optional)
-
-  ds = "%s(%s) -> %s" % (nm, ds_args, str(ty))
-  #print ds
-
-  if has_optional(args):
-      entry = '{"%%s", (PyCFunction)pycv%s, METH_KEYWORDS, "%s"},' % (cname(nm), ds)
-  else:
-      entry = '{"%%s", pycv%s, METH_VARARGS, "%s"},' % (cname(nm), ds)
-  print >>gen_c[1], entry % (nm)
-  if nm.startswith('CV_'):
-    print >>gen_c[1], entry % (nm[3:])
-  for l in gen(nm,args,ty,flags):
-    print >>gen_c[0], l
-
-for l in open("%s/defs" % sys.argv[1]):
-  print >>gen_c[2], "PUBLISH(%s);" % l.split()[1]
-
-########################################################################
-# Generated objects.
-########################################################################
-
-# gen_c[3] is the code, gen_c[4] initializers
-
-gensimple = Template("""
-/*
-  ${cvtype} is the OpenCV C struct
-  ${ourname}_t is the Python object
-*/
-
-struct ${ourname}_t {
-  PyObject_HEAD
-  ${cvtype} v;
-};
-
-static PyObject *${ourname}_repr(PyObject *self)
-{
-  ${ourname}_t *p = (${ourname}_t*)self;
-  char str[1000];
-  sprintf(str, "<${ourname} %p>", p);
-  return PyString_FromString(str);
-}
-
-${getset_funcs}
-
-static PyGetSetDef ${ourname}_getseters[] = {
-
-  ${getset_inits}
-  {NULL}  /* Sentinel */
-};
-
-static PyTypeObject ${ourname}_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  MODULESTR".${ourname}",              /*name*/
-  sizeof(${ourname}_t),                /*basicsize*/
-};
-
-static void ${ourname}_specials(void)
-{
-  ${ourname}_Type.tp_repr = ${ourname}_repr;
-  ${ourname}_Type.tp_getset = ${ourname}_getseters;
-}
-
-static PyObject *FROM_${cvtype}(${cvtype} r)
-{
-  ${ourname}_t *m = PyObject_NEW(${ourname}_t, &${ourname}_Type);
-  m->v = r;
-  return (PyObject*)m;
-}
-
-static int convert_to_${cvtype}PTR(PyObject *o, ${cvtype}** dst, const char *name = "no_name")
-{
-  ${allownull}
-  if (PyType_IsSubtype(o->ob_type, &${ourname}_Type)) {
-    *dst = &(((${ourname}_t*)o)->v);
-    return 1;
-  } else {
-    (*dst) = (${cvtype}*)NULL;
-    return failmsg("Expected ${cvtype} for argument '%s'", name);
-  }
-}
-
-""")
-
-genptr = Template("""
-/*
-  ${cvtype} is the OpenCV C struct
-  ${ourname}_t is the Python object
-*/
-
-struct ${ourname}_t {
-  PyObject_HEAD
-  ${cvtype} *v;
-};
-
-static void ${ourname}_dealloc(PyObject *self)
-{
-  ${ourname}_t *p = (${ourname}_t*)self;
-  cvRelease${ourname}(&p->v);
-  PyObject_Del(self);
-}
-
-static PyObject *${ourname}_repr(PyObject *self)
-{
-  ${ourname}_t *p = (${ourname}_t*)self;
-  char str[1000];
-  sprintf(str, "<${ourname} %p>", p);
-  return PyString_FromString(str);
-}
-
-${getset_funcs}
-
-static PyGetSetDef ${ourname}_getseters[] = {
-
-  ${getset_inits}
-  {NULL}  /* Sentinel */
-};
-
-static PyTypeObject ${ourname}_Type = {
-  PyObject_HEAD_INIT(&PyType_Type)
-  0,                                      /*size*/
-  MODULESTR".${ourname}",              /*name*/
-  sizeof(${ourname}_t),                /*basicsize*/
-};
-
-static void ${ourname}_specials(void)
-{
-  ${ourname}_Type.tp_dealloc = ${ourname}_dealloc;
-  ${ourname}_Type.tp_repr = ${ourname}_repr;
-  ${ourname}_Type.tp_getset = ${ourname}_getseters;
-}
-
-static PyObject *FROM_${cvtype}PTR(${cvtype} *r)
-{
-  ${ourname}_t *m = PyObject_NEW(${ourname}_t, &${ourname}_Type);
-  m->v = r;
-  return (PyObject*)m;
-}
-
-static int convert_to_${cvtype}PTR(PyObject *o, ${cvtype}** dst, const char *name = "no_name")
-{
-  ${allownull}
-  if (PyType_IsSubtype(o->ob_type, &${ourname}_Type)) {
-    *dst = ((${ourname}_t*)o)->v;
-    return 1;
-  } else {
-    (*dst) = (${cvtype}*)NULL;
-    return failmsg("Expected ${cvtype} for argument '%s'", name);
-  }
-}
-
-""")
-
-getset_func_template = Template("""
-static PyObject *${ourname}_get_${member}(${ourname}_t *p, void *closure)
-{
-  return ${rconverter}(p->v${accessor}${member});
-}
-
-static int ${ourname}_set_${member}(${ourname}_t *p, PyObject *value, void *closure)
-{
-  if (value == NULL) {
-    PyErr_SetString(PyExc_TypeError, "Cannot delete the ${member} attribute");
-    return -1;
-  }
-
-  if (! ${checker}(value)) {
-    PyErr_SetString(PyExc_TypeError, "The ${member} attribute value must be a ${typename}");
-    return -1;
-  }
-
-  p->v${accessor}${member} = ${converter}(value);
-  return 0;
-}
-
-""")
-
-getset_init_template = Template("""
-  {(char*)"${member}", (getter)${ourname}_get_${member}, (setter)${ourname}_set_${member}, (char*)"${member}", NULL},
-""")
-
-objects = [
-    ( 'IplConvKernel', ['allownull'], {
-        "nCols" : 'i',
-        "nRows" : 'i',
-        "anchorX" : 'i',
-        "anchorY" : 'i',
-    }),
-    ( 'CvCapture', [], {}),
-    ( 'CvHaarClassifierCascade', [], {}),
-    ( 'CvPOSITObject', [], {}),
-    ( 'CvVideoWriter', [], {}),
-    ( 'CvStereoBMState', [], {
-        "preFilterType" : 'i',
-        "preFilterSize" : 'i',
-        "preFilterCap" : 'i',
-        "SADWindowSize" : 'i',
-        "minDisparity" : 'i',
-        "numberOfDisparities" : 'i',
-        "textureThreshold" : 'i',
-        "uniquenessRatio" : 'i',
-        "speckleWindowSize" : 'i',
-        "speckleRange" : 'i',
-    }),
-    ( 'CvStereoGCState', [], {
-        "Ithreshold" : 'i',
-        "interactionRadius" : 'i',
-        "K" : 'f',
-        "lambda" : 'f',
-        "lambda1" : 'f',
-        "lambda2" : 'f',
-        "occlusionCost" : 'i',
-        "minDisparity" : 'i',
-        "numberOfDisparities" : 'i',
-        "maxIters" : 'i',
-    }),
-    ( 'CvKalman', [], {
-        "MP" : 'i',
-        "DP" : 'i',
-        "CP" : 'i',
-        "state_pre" : 'mr',
-        "state_post" : 'mr',
-        "transition_matrix" : 'mr',
-        "control_matrix" : 'mr',
-        "measurement_matrix" : 'mr',
-        "control_matrix" : 'mr',
-        "process_noise_cov" : 'mr',
-        "measurement_noise_cov" : 'mr',
-        "error_cov_pre" : 'mr',
-        "gain" : 'mr',
-        "error_cov_post" : 'mr',
-    }),
-    ( 'CvMoments', ['copy'], {
-        "m00" : 'f',
-        "m10" : 'f',
-        "m01" : 'f',
-        "m20" : 'f',
-        "m11" : 'f',
-        "m02" : 'f',
-        "m30" : 'f',
-        "m21" : 'f',
-        "m12" : 'f',
-        "m03" : 'f',
-        "mu20" : 'f',
-        "mu11" : 'f',
-        "mu02" : 'f',
-        "mu30" : 'f',
-        "mu21" : 'f',
-        "mu12" : 'f',
-        "mu03" : 'f',
-        "inv_sqrt_m00" : 'f',
-    }),
-]
-
-checkers = {
-    'i' : 'PyNumber_Check',
-    'f' : 'PyNumber_Check',
-    'm' : 'is_cvmat',
-    'mr' : 'is_cvmat'
-}
-# Python -> C
-converters = {
-    'i' : 'PyInt_AsLong',
-    'f' : 'PyFloat_AsDouble',
-    'm' : 'PyCvMat_AsCvMat',
-    'mr' : 'PyCvMat_AsCvMat'
-}
-# C -> Python
-rconverters = {
-    'i' : 'PyInt_FromLong',
-    'f' : 'PyFloat_FromDouble',
-    'm' : 'FROM_CvMat',
-    'mr' : 'FROM_ROCvMatPTR'
-}
-# Human-readable type names
-typenames = {
-    'i' : 'integer',
-    'f' : 'float',
-    'm' : 'list of CvMat',
-    'mr' : 'list of CvMat',
-}
-
-for (t, flags, members) in objects:
-    map = {'cvtype' : t,
-           'ourname' : t.replace('Cv', '')}
-    # gsf is all the generated code for the member accessors
-    if 'copy' in flags:
-        a = '.'
-    else:
-        a = '->'
-    gsf = "".join([getset_func_template.substitute(map, accessor = a, member = m, checker = checkers[t], converter = converters[t], rconverter = rconverters[t], typename = typenames[t]) for (m, t) in members.items()])
-    # gsi is the generated code for the initializer for each accessor
-    gsi = "".join([getset_init_template.substitute(map, member = m) for (m, t) in members.items()])
-    # s is the template that pulls everything together
-    if 'allownull' in flags:
-        nullcode = """if (o == Py_None) { *dst = (%s*)NULL; return 1; }""" % map['cvtype']
-    else:
-        nullcode = ""
-    if 'copy' in flags:
-        print >>gen_c[3], gensimple.substitute(map, getset_funcs = gsf, getset_inits = gsi, allownull = nullcode)
-    else:
-        print >>gen_c[3], genptr.substitute(map, getset_funcs = gsf, getset_inits = gsi, allownull = nullcode)
-    print >>gen_c[4], "MKTYPE(%s);" % map['ourname']
-
-for f in gen_c:
-  f.close()
diff --git a/modules/python/src2/gen2.py b/modules/python/src2/gen2.py
index c7aa72544a..c5df21a399 100755
--- a/modules/python/src2/gen2.py
+++ b/modules/python/src2/gen2.py
@@ -1,24 +1,36 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
-import hdr_parser, sys, re, os, cStringIO
+from __future__ import print_function
+import hdr_parser, sys, re, os
 from string import Template
 
+if sys.version_info[0] >= 3:
+    from io import StringIO
+else:
+    from cStringIO import StringIO
+
+ignored_arg_types = ["RNG*"]
+
 gen_template_check_self = Template("""    if(!PyObject_TypeCheck(self, &pyopencv_${name}_Type))
         return failmsgp("Incorrect type of self (must be '${name}' or its derivative)");
-    $cname* _self_ = ${amp}((pyopencv_${name}_t*)self)->v;
+    $cname* _self_ = ${amp}((pyopencv_${name}_t*)self)->v${get};
 """)
 
 gen_template_check_self_algo = Template("""    if(!PyObject_TypeCheck(self, &pyopencv_${name}_Type))
         return failmsgp("Incorrect type of self (must be '${name}' or its derivative)");
-    $cname* _self_ = dynamic_cast<$cname*>(${amp}((pyopencv_${name}_t*)self)->v.obj);
+    $cname* _self_ = dynamic_cast<$cname*>(${amp}((pyopencv_${name}_t*)self)->v.get());
 """)
 
-gen_template_call_constructor = Template("""self = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
+gen_template_call_constructor_prelude = Template("""self = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
         new (&(self->v)) Ptr<$cname>(); // init Ptr with placement new
-        if(self) ERRWRAP2(self->v = new $cname""")
+        if(self) """)
 
-gen_template_simple_call_constructor = Template("""self = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
-        if(self) ERRWRAP2(self->v = $cname""")
+gen_template_call_constructor = Template("""self->v.reset(new ${cname}${args})""")
+
+gen_template_simple_call_constructor_prelude = Template("""self = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
+        if(self) """)
+
+gen_template_simple_call_constructor = Template("""self->v = ${cname}${args}""")
 
 gen_template_parse_args = Template("""const char* keywords[] = { $kw_list, NULL };
     if( PyArg_ParseTupleAndKeywords(args, kw, "$fmtspec", (char**)keywords, $parse_arglist)$code_cvt )""")
@@ -26,11 +38,18 @@ gen_template_parse_args = Template("""const char* keywords[] = { $kw_list, NULL
 gen_template_func_body = Template("""$code_decl
     $code_parse
     {
-        $code_fcall;
+        ${code_prelude}ERRWRAP2($code_fcall);
         $code_ret;
     }
 """)
 
+py_major_version = sys.version_info[0]
+if py_major_version >= 3:
+    head_init_str = "PyVarObject_HEAD_INIT(&PyType_Type, 0)"
+else:
+    head_init_str = """PyObject_HEAD_INIT(&PyType_Type)
+0,"""
+
 gen_template_simple_type_decl = Template("""
 struct pyopencv_${name}_t
 {
@@ -40,8 +59,7 @@ struct pyopencv_${name}_t
 
 static PyTypeObject pyopencv_${name}_Type =
 {
-    PyObject_HEAD_INIT(&PyType_Type)
-    0,
+    %s
     MODULESTR".$wname",
     sizeof(pyopencv_${name}_t),
 };
@@ -51,26 +69,26 @@ static void pyopencv_${name}_dealloc(PyObject* self)
     PyObject_Del(self);
 }
 
-static PyObject* pyopencv_from(const ${cname}& r)
+template<> PyObject* pyopencv_from(const ${cname}& r)
 {
     pyopencv_${name}_t *m = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
     m->v = r;
     return (PyObject*)m;
 }
 
-static bool pyopencv_to(PyObject* src, ${cname}& dst, const char* name="<unknown>")
+template<> bool pyopencv_to(PyObject* src, ${cname}& dst, const char* name)
 {
     if( src == NULL || src == Py_None )
         return true;
     if(!PyObject_TypeCheck(src, &pyopencv_${name}_Type))
     {
-        failmsg("Expected ${cname} for argument '%s'", name);
+        failmsg("Expected ${cname} for argument '%%s'", name);
         return false;
     }
     dst = ((pyopencv_${name}_t*)src)->v;
     return true;
 }
-""")
+""" % head_init_str)
 
 
 gen_template_type_decl = Template("""
@@ -82,8 +100,7 @@ struct pyopencv_${name}_t
 
 static PyTypeObject pyopencv_${name}_Type =
 {
-    PyObject_HEAD_INIT(&PyType_Type)
-    0,
+    %s
     MODULESTR".$wname",
     sizeof(pyopencv_${name}_t),
 };
@@ -94,7 +111,7 @@ static void pyopencv_${name}_dealloc(PyObject* self)
     PyObject_Del(self);
 }
 
-static PyObject* pyopencv_from(const Ptr<${cname}>& r)
+template<> PyObject* pyopencv_from(const Ptr<${cname}>& r)
 {
     pyopencv_${name}_t *m = PyObject_NEW(pyopencv_${name}_t, &pyopencv_${name}_Type);
     new (&(m->v)) Ptr<$cname1>(); // init Ptr with placement new
@@ -102,23 +119,23 @@ static PyObject* pyopencv_from(const Ptr<${cname}>& r)
     return (PyObject*)m;
 }
 
-static bool pyopencv_to(PyObject* src, Ptr<${cname}>& dst, const char* name="<unknown>")
+template<> bool pyopencv_to(PyObject* src, Ptr<${cname}>& dst, const char* name)
 {
     if( src == NULL || src == Py_None )
         return true;
     if(!PyObject_TypeCheck(src, &pyopencv_${name}_Type))
     {
-        failmsg("Expected ${cname} for argument '%s'", name);
+        failmsg("Expected ${cname} for argument '%%s'", name);
         return false;
     }
-    dst = ((pyopencv_${name}_t*)src)->v;
+    dst = ((pyopencv_${name}_t*)src)->v.dynamicCast<${cname}>();
     return true;
 }
 
-""")
+""" % head_init_str)
 
 gen_template_map_type_cvt = Template("""
-static bool pyopencv_to(PyObject* src, ${cname}& dst, const char* name="<unknown>");
+template<> bool pyopencv_to(PyObject* src, ${cname}& dst, const char* name);
 """)
 
 gen_template_set_prop_from_map = Template("""
@@ -174,7 +191,7 @@ static PyObject* pyopencv_${name}_get_${member}(pyopencv_${name}_t* p, void *clo
 gen_template_get_prop_algo = Template("""
 static PyObject* pyopencv_${name}_get_${member}(pyopencv_${name}_t* p, void *closure)
 {
-    return pyopencv_from(dynamic_cast<$cname*>(p->v.obj)${access}${member});
+    return pyopencv_from(dynamic_cast<$cname*>(p->v.get())${access}${member});
 }
 """)
 
@@ -198,7 +215,7 @@ static int pyopencv_${name}_set_${member}(pyopencv_${name}_t* p, PyObject *value
         PyErr_SetString(PyExc_TypeError, "Cannot delete the ${member} attribute");
         return -1;
     }
-    return pyopencv_to(value, dynamic_cast<$cname*>(p->v.obj)${access}${member}) ? 0 : -1;
+    return pyopencv_to(value, dynamic_cast<$cname*>(p->v.get())${access}${member}) ? 0 : -1;
 }
 """)
 
@@ -214,8 +231,7 @@ simple_argtype_mapping = {
     "int": ("int", "i", "0"),
     "float": ("float", "f", "0.f"),
     "double": ("double", "d", "0"),
-    "c_string": ("char*", "s", '(char*)""'),
-    "string": ("std::string", "s", None)
+    "c_string": ("char*", "s", '(char*)""')
 }
 
 def normalize_class_name(name):
@@ -244,9 +260,9 @@ class ClassInfo(object):
         if decl:
             self.bases = decl[1].split()[1:]
             if len(self.bases) > 1:
-                print "Note: Class %s has more than 1 base class (not supported by Python C extensions)" % (self.name,)
-                print "      Bases: ", " ".join(self.bases)
-                print "      Only the first base class will be used"
+                print("Note: Class %s has more than 1 base class (not supported by Python C extensions)" % (self.name,))
+                print("      Bases: ", " ".join(self.bases))
+                print("      Only the first base class will be used")
                 self.bases = [self.bases[0].strip(",")]
                 #return sys.exit(-1)
             if self.bases and self.bases[0].startswith("cv::"):
@@ -279,8 +295,8 @@ class ClassInfo(object):
         if self.ismap:
             return self.gen_map_code(all_classes)
 
-        getset_code = cStringIO.StringIO()
-        getset_inits = cStringIO.StringIO()
+        getset_code = StringIO()
+        getset_inits = StringIO()
 
         sorted_props = [(p.name, p) for p in self.props]
         sorted_props.sort()
@@ -303,10 +319,10 @@ class ClassInfo(object):
                     getset_code.write(gen_template_set_prop.substitute(name=self.name, member=pname, membertype=p.tp, access=access_op))
                 getset_inits.write(gen_template_rw_prop_init.substitute(name=self.name, member=pname))
 
-        methods_code = cStringIO.StringIO()
-        methods_inits = cStringIO.StringIO()
+        methods_code = StringIO()
+        methods_inits = StringIO()
 
-        sorted_methods = self.methods.items()
+        sorted_methods = list(self.methods.items())
         sorted_methods.sort()
 
         for mname, m in sorted_methods:
@@ -314,7 +330,7 @@ class ClassInfo(object):
             methods_inits.write(m.get_tab_entry())
 
         baseptr = "NULL"
-        if self.bases and all_classes.has_key(self.bases[0]):
+        if self.bases and self.bases[0] in all_classes:
             baseptr = "&pyopencv_" + all_classes[self.bases[0]].name + "_Type"
 
         code = gen_template_type_impl.substitute(name=self.name, wname=self.wname, cname=self.cname,
@@ -426,6 +442,8 @@ class FuncVariant(object):
             argno += 1
             if a.name in self.array_counters:
                 continue
+            if a.tp in ignored_arg_types:
+                continue
             if a.returnarg:
                 outlist.append((a.name, argno))
             if (not a.inputarg) and a.isbig():
@@ -529,7 +547,7 @@ class FuncInfo(object):
             p2 = s.rfind(")")
             docstring_list = [s[:p1+1] + "[" + s[p1+1:p2] + "]" + s[p2:]]
 
-        return Template('    {"$py_funcname", (PyCFunction)$wrap_funcname, METH_KEYWORDS, "$py_docstring"},\n'
+        return Template('    {"$py_funcname", (PyCFunction)$wrap_funcname, METH_VARARGS | METH_KEYWORDS, "$py_docstring"},\n'
                         ).substitute(py_funcname = self.variants[0].wname, wrap_funcname=self.get_wrapper_name(),
                                      py_docstring = "  or  ".join(docstring_list))
 
@@ -545,39 +563,22 @@ class FuncInfo(object):
         if self.classname:
             selfinfo = all_classes[self.classname]
             if not self.isconstructor:
-                amp = ""
-                if selfinfo.issimple:
-                    amp = "&"
+                amp = "&" if selfinfo.issimple else ""
                 if selfinfo.isalgorithm:
                     code += gen_template_check_self_algo.substitute(name=selfinfo.name, cname=selfinfo.cname, amp=amp)
                 else:
-                    code += gen_template_check_self.substitute(name=selfinfo.name, cname=selfinfo.cname, amp=amp)
+                    get = "" if selfinfo.issimple else ".get()"
+                    code += gen_template_check_self.substitute(name=selfinfo.name, cname=selfinfo.cname, amp=amp, get=get)
                 fullname = selfinfo.wname + "." + fullname
 
         all_code_variants = []
         declno = -1
         for v in self.variants:
             code_decl = ""
-            code_fcall = ""
             code_ret = ""
             code_cvt_list = []
 
-            if self.isconstructor:
-                code_decl += "    pyopencv_%s_t* self = 0;\n" % selfinfo.name
-                templ = gen_template_call_constructor
-                if selfinfo.issimple:
-                    templ = gen_template_simple_call_constructor
-                code_fcall = templ.substitute(name=selfinfo.name, cname=selfinfo.cname)
-            else:
-                code_fcall = "ERRWRAP2( "
-                if v.rettype:
-                    code_decl += "    " + simple_argtype_mapping.get(v.rettype, (v.rettype, None, None))[0]  + " retval;\n"
-                    code_fcall += "retval = "
-                if ismethod:
-                    code_fcall += "_self_->" + self.cname
-                else:
-                    code_fcall += self.cname
-            code_fcall += "("
+            code_args = "("
             all_cargs = []
             parse_arglist = []
 
@@ -586,6 +587,16 @@ class FuncInfo(object):
             # form the function/method call,
             # for the list of type mappings
             for a in v.args:
+                if a.tp in ignored_arg_types:
+                    defval = a.defval
+                    if not defval and a.tp.endswith("*"):
+                        defval = 0
+                    assert defval
+                    if not code_args.endswith("("):
+                        code_args += ", "
+                    code_args += defval
+                    all_cargs.append([[None, ""], ""])
+                    continue
                 tp1 = tp = a.tp
                 amp = ""
                 defval0 = ""
@@ -596,7 +607,7 @@ class FuncInfo(object):
                         defval0 = "0"
                         tp1 = tp.replace("*", "_ptr")
                 if tp1.endswith("*"):
-                    print "Error: type with star: a.tp=%s, tp=%s, tp1=%s" % (a.tp, tp, tp1)
+                    print("Error: type with star: a.tp=%s, tp=%s, tp1=%s" % (a.tp, tp, tp1))
                     sys.exit(-1)
 
                 amapping = simple_argtype_mapping.get(tp, (tp, "O", defval0))
@@ -605,7 +616,10 @@ class FuncInfo(object):
                     if amapping[1] == "O":
                         code_decl += "    PyObject* pyobj_%s = NULL;\n" % (a.name,)
                         parse_name = "pyobj_" + a.name
-                        code_cvt_list.append("pyopencv_to(pyobj_%s, %s, %s)" % (a.name, a.name, a.crepr()))
+                        if a.tp == 'char':
+                            code_cvt_list.append("convert_to_char(pyobj_%s, &%s, %s)"% (a.name, a.name, a.crepr()))
+                        else:
+                            code_cvt_list.append("pyopencv_to(pyobj_%s, %s, %s)" % (a.name, a.name, a.crepr()))
 
                 all_cargs.append([amapping, parse_name])
 
@@ -622,11 +636,34 @@ class FuncInfo(object):
                 else:
                     code_decl += "    %s %s;\n" % (amapping[0], a.name)
 
-                if not code_fcall.endswith("("):
-                    code_fcall += ", "
-                code_fcall += amp + a.name
+                if not code_args.endswith("("):
+                    code_args += ", "
+                code_args += amp + a.name
 
-            code_fcall += "))"
+            code_args += ")"
+
+            if self.isconstructor:
+                code_decl += "    pyopencv_%s_t* self = 0;\n" % selfinfo.name
+                if selfinfo.issimple:
+                    templ_prelude = gen_template_simple_call_constructor_prelude
+                    templ = gen_template_simple_call_constructor
+                else:
+                    templ_prelude = gen_template_call_constructor_prelude
+                    templ = gen_template_call_constructor
+
+                code_prelude = templ_prelude.substitute(name=selfinfo.name, cname=selfinfo.cname)
+                code_fcall = templ.substitute(name=selfinfo.name, cname=selfinfo.cname, args=code_args)
+            else:
+                code_prelude = ""
+                code_fcall = ""
+                if v.rettype:
+                    code_decl += "    " + v.rettype + " retval;\n"
+                    code_fcall += "retval = "
+                if ismethod:
+                    code_fcall += "_self_->" + self.cname
+                else:
+                    code_fcall += self.cname
+                code_fcall += code_args
 
             if code_cvt_list:
                 code_cvt_list = [""] + code_cvt_list
@@ -679,7 +716,7 @@ class FuncInfo(object):
                     (fmtspec, ", ".join(["pyopencv_from(" + aname + ")" for aname, argno in v.py_outlist]))
 
             all_code_variants.append(gen_template_func_body.substitute(code_decl=code_decl,
-                code_parse=code_parse, code_fcall=code_fcall, code_ret=code_ret))
+                code_parse=code_parse, code_prelude=code_prelude, code_fcall=code_fcall, code_ret=code_ret))
 
         if len(all_code_variants)==1:
             # if the function/method has only 1 signature, then just put it
@@ -699,11 +736,11 @@ class PythonWrapperGenerator(object):
         self.classes = {}
         self.funcs = {}
         self.consts = {}
-        self.code_types = cStringIO.StringIO()
-        self.code_funcs = cStringIO.StringIO()
-        self.code_func_tab = cStringIO.StringIO()
-        self.code_type_reg = cStringIO.StringIO()
-        self.code_const_reg = cStringIO.StringIO()
+        self.code_types = StringIO()
+        self.code_funcs = StringIO()
+        self.code_func_tab = StringIO()
+        self.code_type_reg = StringIO()
+        self.code_const_reg = StringIO()
         self.class_idx = 0
 
     def add_class(self, stype, name, decl):
@@ -711,9 +748,9 @@ class PythonWrapperGenerator(object):
         classinfo.decl_idx = self.class_idx
         self.class_idx += 1
 
-        if self.classes.has_key(classinfo.name):
-            print "Generator error: class %s (cname=%s) already exists" \
-                % (classinfo.name, classinfo.cname)
+        if classinfo.name in self.classes:
+            print("Generator error: class %s (cname=%s) already exists" \
+                % (classinfo.name, classinfo.cname))
             sys.exit(-1)
         self.classes[classinfo.name] = classinfo
         if classinfo.bases and not classinfo.isalgorithm:
@@ -722,9 +759,9 @@ class PythonWrapperGenerator(object):
     def add_const(self, name, decl):
         constinfo = ConstInfo(name, decl[1])
 
-        if self.consts.has_key(constinfo.name):
-            print "Generator error: constant %s (cname=%s) already exists" \
-                % (constinfo.name, constinfo.cname)
+        if constinfo.name in self.consts:
+            print("Generator error: constant %s (cname=%s) already exists" \
+                % (constinfo.name, constinfo.cname))
             sys.exit(-1)
         self.consts[constinfo.name] = constinfo
 
@@ -763,7 +800,7 @@ class PythonWrapperGenerator(object):
         else:
             classinfo = self.classes.get(classname, ClassInfo(""))
             if not classinfo.name:
-                print "Generator error: the class for method %s is missing" % (name,)
+                print("Generator error: the class for method %s is missing" % (name,))
                 sys.exit(-1)
             func_map = classinfo.methods
 
@@ -803,7 +840,7 @@ class PythonWrapperGenerator(object):
                     self.add_func(decl)
 
         # step 2: generate code for the classes and their methods
-        classlist = self.classes.items()
+        classlist = list(self.classes.items())
         classlist.sort()
         for name, classinfo in classlist:
             if classinfo.ismap:
@@ -828,7 +865,7 @@ class PythonWrapperGenerator(object):
                 self.code_type_reg.write("MKTYPE2(%s);\n" % (classinfo.name,) )
 
         # step 3: generate the code for all the global functions
-        funclist = self.funcs.items()
+        funclist = list(self.funcs.items())
         funclist.sort()
         for name, func in funclist:
             code = func.gen_code(self.classes)
@@ -836,7 +873,7 @@ class PythonWrapperGenerator(object):
             self.code_func_tab.write(func.get_tab_entry())
 
         # step 4: generate the code for constants
-        constlist = self.consts.items()
+        constlist = list(self.consts.items())
         constlist.sort()
         for name, constinfo in constlist:
             self.gen_const_reg(constinfo)
diff --git a/modules/python/src2/hdr_parser.py b/modules/python/src2/hdr_parser.py
index 536e180c2f..b6f21c31eb 100755
--- a/modules/python/src2/hdr_parser.py
+++ b/modules/python/src2/hdr_parser.py
@@ -1,5 +1,6 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
+from __future__ import print_function
 import os, sys, re, string
 
 # the list only for debugging. The real list, used in the real OpenCV build, is specified in CMakeLists.txt
@@ -43,13 +44,13 @@ class CppHeaderParser(object):
     def get_macro_arg(self, arg_str, npos):
         npos2 = npos3 = arg_str.find("(", npos)
         if npos2 < 0:
-            print "Error: no arguments for the macro at %d" % (self.lineno,)
+            print("Error: no arguments for the macro at %d" % (self.lineno,))
             sys.exit(-1)
         balance = 1
         while 1:
             t, npos3 = self.find_next_token(arg_str, ['(', ')'], npos3+1)
             if npos3 < 0:
-                print "Error: no matching ')' in the macro call at %d" % (self.lineno,)
+                print("Error: no matching ')' in the macro call at %d" % (self.lineno,))
                 sys.exit(-1)
             if t == '(':
                 balance += 1
@@ -143,13 +144,13 @@ class CppHeaderParser(object):
                 angle_stack.append(0)
             elif w == "," or w == '>':
                 if not angle_stack:
-                    print "Error at %d: argument contains ',' or '>' not within template arguments" % (self.lineno,)
+                    print("Error at %d: argument contains ',' or '>' not within template arguments" % (self.lineno,))
                     sys.exit(-1)
                 if w == ",":
                     arg_type += "_and_"
                 elif w == ">":
                     if angle_stack[0] == 0:
-                        print "Error at %s:%d: template has no arguments" % (self.hname, self.lineno)
+                        print("Error at %s:%d: template has no arguments" % (self.hname, self.lineno))
                         sys.exit(-1)
                     if angle_stack[0] > 1:
                         arg_type += "_end_"
@@ -173,7 +174,7 @@ class CppHeaderParser(object):
             p1 = arg_name.find("[")
             p2 = arg_name.find("]",p1+1)
             if p2 < 0:
-                print "Error at %d: no closing ]" % (self.lineno,)
+                print("Error at %d: no closing ]" % (self.lineno,))
                 sys.exit(-1)
             counter_str = arg_name[p1+1:p2].strip()
             if counter_str == "":
@@ -335,7 +336,7 @@ class CppHeaderParser(object):
                 if pos >= 0:
                     aname = arg[pos+1:].strip()
                     atype = arg[:pos+1].strip()
-                    if aname.endswith("&") or aname.endswith("*") or (aname in ["int", "string", "Mat"]):
+                    if aname.endswith("&") or aname.endswith("*") or (aname in ["int", "String", "Mat"]):
                         atype = (atype + " " + aname).strip()
                         aname = ""
                 else:
@@ -358,7 +359,7 @@ class CppHeaderParser(object):
         if bool(re.match(r".*\)\s*const(\s*=\s*0)?", decl_str)):
             decl[2].append("/C")
         if "virtual" in decl_str:
-            print decl_str
+            print(decl_str)
         return decl
 
     def parse_func_decl(self, decl_str):
@@ -412,12 +413,12 @@ class CppHeaderParser(object):
         if decl_str.startswith("CVAPI"):
             rtype_end = decl_str.find(")", args_begin+1)
             if rtype_end < 0:
-                print "Error at %d. no terminating ) in CVAPI() macro: %s" % (self.lineno, decl_str)
+                print("Error at %d. no terminating ) in CVAPI() macro: %s" % (self.lineno, decl_str))
                 sys.exit(-1)
             decl_str = decl_str[args_begin+1:rtype_end] + " " + decl_str[rtype_end+1:]
             args_begin = decl_str.find("(")
         if args_begin < 0:
-            print "Error at %d: no args in '%s'" % (self.lineno, decl_str)
+            print("Error at %d: no args in '%s'" % (self.lineno, decl_str))
             sys.exit(-1)
 
         decl_start = decl_str[:args_begin].strip()
@@ -425,7 +426,7 @@ class CppHeaderParser(object):
         if decl_start.endswith("operator"):
             args_begin = decl_str.find("(", args_begin+1)
             if args_begin < 0:
-                print "Error at %d: no args in '%s'" % (self.lineno, decl_str)
+                print("Error at %d: no args in '%s'" % (self.lineno, decl_str))
                 sys.exit(-1)
             decl_start = decl_str[:args_begin].strip()
             # TODO: normalize all type of operators
@@ -455,7 +456,7 @@ class CppHeaderParser(object):
                     return [] # exotic - dynamic 2d array
                 else:
                     #print rettype, funcname, modlist, argno
-                    print "Error at %s:%d the function/method name is missing: '%s'" % (self.hname, self.lineno, decl_start)
+                    print("Error at %s:%d the function/method name is missing: '%s'" % (self.hname, self.lineno, decl_start))
                     sys.exit(-1)
 
         if self.wrap_mode and (("::" in funcname) or funcname.startswith("~")):
@@ -486,9 +487,9 @@ class CppHeaderParser(object):
             npos += 1
             t, npos = self.find_next_token(decl_str, ["(", ")", ",", "<", ">"], npos)
             if not t:
-                print "Error: no closing ')' at %d" % (self.lineno,)
-                print decl_str
-                print decl_str[arg_start:]
+                print("Error: no closing ')' at %d" % (self.lineno,))
+                print(decl_str)
+                print(decl_str[arg_start:])
                 sys.exit(-1)
             if t == "<":
                 angle_balance += 1
@@ -583,7 +584,7 @@ class CppHeaderParser(object):
             if block_type in ["file", "enum"]:
                 continue
             if block_type not in ["struct", "class", "namespace"]:
-                print "Error at %d: there are non-valid entries in the current block stack " % (self.lineno, self.block_stack)
+                print("Error at %d: there are non-valid entries in the current block stack " % (self.lineno, self.block_stack))
                 sys.exit(-1)
             if block_name:
                 n += block_name + "."
@@ -605,7 +606,7 @@ class CppHeaderParser(object):
             stmt_type = "block"
 
         if context == "block":
-            print "Error at %d: should not call parse_stmt inside blocks" % (self.lineno,)
+            print("Error at %d: should not call parse_stmt inside blocks" % (self.lineno,))
             sys.exit(-1)
 
         if context == "class" or context == "struct":
@@ -632,7 +633,7 @@ class CppHeaderParser(object):
                 try:
                     classname, bases, modlist = self.parse_class_decl(stmt[len("typedef "):])
                 except:
-                    print "Error at %s:%d" % (self.hname, self.lineno)
+                    print("Error at %s:%d" % (self.hname, self.lineno))
                     exit(1)
                 if classname.startswith("_Ipl"):
                     classname = classname[1:]
@@ -647,7 +648,7 @@ class CppHeaderParser(object):
                     try:
                         classname, bases, modlist = self.parse_class_decl(stmt)
                     except:
-                        print "Error at %s:%d" % (self.hname, self.lineno)
+                        print("Error at %s:%d" % (self.hname, self.lineno))
                         exit(1)
                     decl = []
                     if ("CV_EXPORTS_W" in stmt) or ("CV_EXPORTS_AS" in stmt) or (not self.wrap_mode):# and ("CV_EXPORTS" in stmt)):
@@ -767,7 +768,7 @@ class CppHeaderParser(object):
                 state = SCAN
 
             if state != SCAN:
-                print "Error at %d: invlid state = %d" % (self.lineno, state)
+                print("Error at %d: invlid state = %d" % (self.lineno, state))
                 sys.exit(-1)
 
             while 1:
@@ -795,7 +796,7 @@ class CppHeaderParser(object):
                     while 1:
                         t2, pos2 = self.find_next_token(l, ["\\", "\""], pos2)
                         if t2 == "":
-                            print "Error at %d: no terminating '\"'" % (self.lineno,)
+                            print("Error at %d: no terminating '\"'" % (self.lineno,))
                             sys.exit(-1)
                         if t2 == "\"":
                             break
@@ -836,7 +837,7 @@ class CppHeaderParser(object):
 
                 if token == "}":
                     if not self.block_stack:
-                        print "Error at %d: the block stack is empty" % (self.lineno,)
+                        print("Error at %d: the block stack is empty" % (self.lineno,))
                     self.block_stack[-1:] = []
                     if pos+1 < len(l) and l[pos+1] == ';':
                         pos += 1
@@ -851,13 +852,13 @@ class CppHeaderParser(object):
         Prints the list of declarations, retrieived by the parse() method
         """
         for d in decls:
-            print d[0], d[1], ";".join(d[2])
+            print(d[0], d[1], ";".join(d[2]))
             for a in d[3]:
-                print "   ", a[0], a[1], a[2],
+                print("   ", a[0], a[1], a[2], end="")
                 if a[3]:
-                    print "; ".join(a[3])
+                    print("; ".join(a[3]))
                 else:
-                    print
+                    print()
 
 if __name__ == '__main__':
     parser = CppHeaderParser()
@@ -867,4 +868,4 @@ if __name__ == '__main__':
     #for hname in sys.argv[1:]:
         #decls += parser.parse(hname, wmode=False)
     parser.print_decls(decls)
-    print len(decls)
+    print(len(decls))
diff --git a/modules/videostab/src/precomp.cpp b/modules/python/src2/pycompat.hpp
similarity index 73%
rename from modules/videostab/src/precomp.cpp
rename to modules/python/src2/pycompat.hpp
index 111385282e..c473fffb20 100644
--- a/modules/videostab/src/precomp.cpp
+++ b/modules/python/src2/pycompat.hpp
@@ -40,4 +40,25 @@
 //
 //M*/
 
-#include "precomp.hpp"
+// Defines for Python 2/3 compatibility.
+#ifndef __PYCOMPAT_HPP__
+#define __PYCOMPAT_HPP__
+
+#if PY_MAJOR_VERSION >= 3
+// Python3 treats all ints as longs, PyInt_X functions have been removed.
+#define PyInt_Check PyLong_Check
+#define PyInt_CheckExact PyLong_CheckExact
+#define PyInt_AsLong PyLong_AsLong
+#define PyInt_AS_LONG PyLong_AS_LONG
+#define PyInt_FromLong PyLong_FromLong
+#define PyNumber_Int PyNumber_Long
+
+// Python3 strings are unicode, these defines mimic the Python2 functionality.
+#define PyString_Check PyUnicode_Check
+#define PyString_FromString PyUnicode_FromString
+#define PyString_AsString PyUnicode_AsUTF8
+#define PyString_FromStringAndSize PyUnicode_FromStringAndSize
+#define PyString_Size PyUnicode_GET_SIZE
+#endif
+
+#endif // END HEADER GUARD
diff --git a/modules/python/test/calchist.py b/modules/python/test/calchist.py
index 0a52258b20..dba6796b3a 100755
--- a/modules/python/test/calchist.py
+++ b/modules/python/test/calchist.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 # Calculating and displaying 2D Hue-Saturation histogram of a color image
 import sys
@@ -51,4 +51,3 @@ if __name__ == '__main__':
     cv.ShowImage("H-S Histogram", hs_histogram(src))
 
     cv.WaitKey(0)
-
diff --git a/modules/python/test/camera_calibration.py b/modules/python/test/camera_calibration.py
index 488dd15c62..8ffc5b1cd9 100755
--- a/modules/python/test/camera_calibration.py
+++ b/modules/python/test/camera_calibration.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys
 import math
diff --git a/modules/python/test/findstereocorrespondence.py b/modules/python/test/findstereocorrespondence.py
index 8f11738cce..40a9603beb 100755
--- a/modules/python/test/findstereocorrespondence.py
+++ b/modules/python/test/findstereocorrespondence.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import sys
 import cv2.cv as cv
diff --git a/modules/python/test/goodfeatures.py b/modules/python/test/goodfeatures.py
index 62907772ab..5ccd5b46c1 100755
--- a/modules/python/test/goodfeatures.py
+++ b/modules/python/test/goodfeatures.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 import unittest
diff --git a/modules/python/test/leak1.py b/modules/python/test/leak1.py
index dde5608951..dbd6040a5a 100755
--- a/modules/python/test/leak1.py
+++ b/modules/python/test/leak1.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 import numpy as np
diff --git a/modules/python/test/leak2.py b/modules/python/test/leak2.py
index af1cb0556c..518226448a 100755
--- a/modules/python/test/leak2.py
+++ b/modules/python/test/leak2.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 import numpy as np
diff --git a/modules/python/test/leak3.py b/modules/python/test/leak3.py
index f72afbbf08..d763c4044d 100755
--- a/modules/python/test/leak3.py
+++ b/modules/python/test/leak3.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 import math
diff --git a/modules/python/test/leak4.py b/modules/python/test/leak4.py
index dcfc5cfdc9..9e5864092b 100755
--- a/modules/python/test/leak4.py
+++ b/modules/python/test/leak4.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 import math
diff --git a/modules/python/test/precornerdetect.py b/modules/python/test/precornerdetect.py
index 29a6ca1ecd..97aa906d4a 100755
--- a/modules/python/test/precornerdetect.py
+++ b/modules/python/test/precornerdetect.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import cv2.cv as cv
 
diff --git a/modules/python/test/test.py b/modules/python/test/test.py
old mode 100755
new mode 100644
index 7c511e4ef7..2da740de9b
--- a/modules/python/test/test.py
+++ b/modules/python/test/test.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import unittest
 import random
@@ -13,2225 +13,120 @@ import os
 import getopt
 import operator
 import functools
+import numpy as np
+import cv2
 
-import cv2.cv as cv
+class NewOpenCVTests(unittest.TestCase):
 
-from test2 import *
-
-class OpenCVTests(unittest.TestCase):
-
-    depths = [ cv.IPL_DEPTH_8U, cv.IPL_DEPTH_8S, cv.IPL_DEPTH_16U, cv.IPL_DEPTH_16S, cv.IPL_DEPTH_32S, cv.IPL_DEPTH_32F, cv.IPL_DEPTH_64F ]
-
-    mat_types = [
-        cv.CV_8UC1,
-        cv.CV_8UC2,
-        cv.CV_8UC3,
-        cv.CV_8UC4,
-        cv.CV_8SC1,
-        cv.CV_8SC2,
-        cv.CV_8SC3,
-        cv.CV_8SC4,
-        cv.CV_16UC1,
-        cv.CV_16UC2,
-        cv.CV_16UC3,
-        cv.CV_16UC4,
-        cv.CV_16SC1,
-        cv.CV_16SC2,
-        cv.CV_16SC3,
-        cv.CV_16SC4,
-        cv.CV_32SC1,
-        cv.CV_32SC2,
-        cv.CV_32SC3,
-        cv.CV_32SC4,
-        cv.CV_32FC1,
-        cv.CV_32FC2,
-        cv.CV_32FC3,
-        cv.CV_32FC4,
-        cv.CV_64FC1,
-        cv.CV_64FC2,
-        cv.CV_64FC3,
-        cv.CV_64FC4,
-    ]
-    mat_types_single = [
-        cv.CV_8UC1,
-        cv.CV_8SC1,
-        cv.CV_16UC1,
-        cv.CV_16SC1,
-        cv.CV_32SC1,
-        cv.CV_32FC1,
-        cv.CV_64FC1,
-    ]
-
-    def depthsize(self, d):
-        return { cv.IPL_DEPTH_8U : 1,
-                 cv.IPL_DEPTH_8S : 1,
-                 cv.IPL_DEPTH_16U : 2,
-                 cv.IPL_DEPTH_16S : 2,
-                 cv.IPL_DEPTH_32S : 4,
-                 cv.IPL_DEPTH_32F : 4,
-                 cv.IPL_DEPTH_64F : 8 }[d]
-
-    def get_sample(self, filename, iscolor = cv.CV_LOAD_IMAGE_COLOR):
+    def get_sample(self, filename, iscolor = cv2.IMREAD_COLOR):
         if not filename in self.image_cache:
             filedata = urllib.urlopen("https://raw.github.com/Itseez/opencv/master/" + filename).read()
-            imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-            cv.SetData(imagefiledata, filedata, len(filedata))
-            self.image_cache[filename] = cv.DecodeImageM(imagefiledata, iscolor)
+            self.image_cache[filename] = cv2.imdecode(np.fromstring(filedata, dtype=np.uint8), iscolor)
         return self.image_cache[filename]
 
     def setUp(self):
         self.image_cache = {}
 
-    def snap(self, img):
-        self.snapL([img])
-
-    def snapL(self, L):
-        for i,img in enumerate(L):
-            cv.NamedWindow("snap-%d" % i, 1)
-            cv.ShowImage("snap-%d" % i, img)
-        cv.WaitKey()
-        cv.DestroyAllWindows()
-
     def hashimg(self, im):
         """ Compute a hash for an image, useful for image comparisons """
         return hashlib.md5(im.tostring()).digest()
 
+    if sys.version_info[:2] == (2, 6):
+        def assertLess(self, a, b, msg=None):
+            if not a < b:
+                self.fail('%s not less than %s' % (repr(a), repr(b)))
+
+        def assertLessEqual(self, a, b, msg=None):
+            if not a <= b:
+                self.fail('%s not less than or equal to %s' % (repr(a), repr(b)))
+
+        def assertGreater(self, a, b, msg=None):
+            if not a > b:
+                self.fail('%s not greater than %s' % (repr(a), repr(b)))
+
 # Tests to run first; check the handful of basic operations that the later tests rely on
 
-class PreliminaryTests(OpenCVTests):
-
-    def test_lena(self):
-        # Check that the lena jpg image has loaded correctly
-        # This test uses a 'golden' MD5 hash of the Lena image
-        # If the JPEG decompressor changes, it is possible that the MD5 hash will change,
-        # so the hash here will need to change.
-
-        im = self.get_sample("samples/c/lena.jpg")
-        # self.snap(im)     # uncomment this line to view the image, when regilding
-        self.assertEqual(hashlib.md5(im.tostring()).hexdigest(), "9dcd9247f9811c6ce86675ba7b0297b6")
-
-    def test_LoadImage(self):
-        self.assertRaises(TypeError, lambda: cv.LoadImage())
-        self.assertRaises(TypeError, lambda: cv.LoadImage(4))
-        self.assertRaises(TypeError, lambda: cv.LoadImage('foo.jpg', 1, 1))
-        self.assertRaises(TypeError, lambda: cv.LoadImage('foo.jpg', xiscolor=cv.CV_LOAD_IMAGE_COLOR))
-
-    def test_types(self):
-        self.assert_(type(cv.CreateImage((7,5), cv.IPL_DEPTH_8U, 1)) == cv.iplimage)
-        self.assert_(type(cv.CreateMat(5, 7, cv.CV_32FC1)) == cv.cvmat)
-        for i,t in enumerate(self.mat_types):
-            basefunc = [
-                cv.CV_8UC,
-                cv.CV_8SC,
-                cv.CV_16UC,
-                cv.CV_16SC,
-                cv.CV_32SC,
-                cv.CV_32FC,
-                cv.CV_64FC,
-            ][i / 4]
-            self.assertEqual(basefunc(1 + (i % 4)), t)
-
-    def test_tostring(self):
-
-        for w in [ 1, 4, 64, 512, 640]:
-            for h in [ 1, 4, 64, 480, 512]:
-                for c in [1, 2, 3, 4]:
-                    for d in self.depths:
-                        a = cv.CreateImage((w,h), d, c);
-                        self.assert_(len(a.tostring()) == w * h * c * self.depthsize(d))
-
-        for w in [ 32, 96, 480 ]:
-            for h in [ 32, 96, 480 ]:
-                depth_size = {
-                    cv.IPL_DEPTH_8U : 1,
-                    cv.IPL_DEPTH_8S : 1,
-                    cv.IPL_DEPTH_16U : 2,
-                    cv.IPL_DEPTH_16S : 2,
-                    cv.IPL_DEPTH_32S : 4,
-                    cv.IPL_DEPTH_32F : 4,
-                    cv.IPL_DEPTH_64F : 8
-                }
-                for f in  self.depths:
-                    for channels in (1,2,3,4):
-                        img = cv.CreateImage((w, h), f, channels)
-                        esize = (w * h * channels * depth_size[f])
-                        self.assert_(len(img.tostring()) == esize)
-                        cv.SetData(img, " " * esize, w * channels * depth_size[f])
-                        self.assert_(len(img.tostring()) == esize)
-
-                mattype_size = {
-                    cv.CV_8UC1 : 1,
-                    cv.CV_8UC2 : 1,
-                    cv.CV_8UC3 : 1,
-                    cv.CV_8UC4 : 1,
-                    cv.CV_8SC1 : 1,
-                    cv.CV_8SC2 : 1,
-                    cv.CV_8SC3 : 1,
-                    cv.CV_8SC4 : 1,
-                    cv.CV_16UC1 : 2,
-                    cv.CV_16UC2 : 2,
-                    cv.CV_16UC3 : 2,
-                    cv.CV_16UC4 : 2,
-                    cv.CV_16SC1 : 2,
-                    cv.CV_16SC2 : 2,
-                    cv.CV_16SC3 : 2,
-                    cv.CV_16SC4 : 2,
-                    cv.CV_32SC1 : 4,
-                    cv.CV_32SC2 : 4,
-                    cv.CV_32SC3 : 4,
-                    cv.CV_32SC4 : 4,
-                    cv.CV_32FC1 : 4,
-                    cv.CV_32FC2 : 4,
-                    cv.CV_32FC3 : 4,
-                    cv.CV_32FC4 : 4,
-                    cv.CV_64FC1 : 8,
-                    cv.CV_64FC2 : 8,
-                    cv.CV_64FC3 : 8,
-                    cv.CV_64FC4 : 8
-                }
-
-                for t in self.mat_types:
-                    for im in [cv.CreateMat(h, w, t), cv.CreateMatND([h, w], t)]:
-                        elemsize = cv.CV_MAT_CN(cv.GetElemType(im)) * mattype_size[cv.GetElemType(im)]
-                        cv.SetData(im, " " * (w * h * elemsize), (w * elemsize))
-                        esize = (w * h * elemsize)
-                        self.assert_(len(im.tostring()) == esize)
-                        cv.SetData(im, " " * esize, w * elemsize)
-                        self.assert_(len(im.tostring()) == esize)
-
-# Tests for specific OpenCV functions
-
-class FunctionTests(OpenCVTests):
-
-    def test_AvgSdv(self):
-        m = cv.CreateMat(1, 8, cv.CV_32FC1)
-        for i,v in enumerate([2, 4, 4, 4, 5, 5, 7, 9]):
-            m[0,i] = (v,)
-        self.assertAlmostEqual(cv.Avg(m)[0], 5.0, 3)
-        avg,sdv = cv.AvgSdv(m)
-        self.assertAlmostEqual(avg[0], 5.0, 3)
-        self.assertAlmostEqual(sdv[0], 2.0, 3)
-
-    def test_CalcEMD2(self):
-        cc = {}
-        for r in [ 5, 10, 37, 38 ]:
-            scratch = cv.CreateImage((100,100), 8, 1)
-            cv.SetZero(scratch)
-            cv.Circle(scratch, (50,50), r, 255, -1)
-            storage = cv.CreateMemStorage()
-            seq = cv.FindContours(scratch, storage, cv.CV_RETR_TREE, cv.CV_CHAIN_APPROX_SIMPLE)
-            arr = cv.CreateMat(len(seq), 3, cv.CV_32FC1)
-            for i,e in enumerate(seq):
-                arr[i,0] = 1
-                arr[i,1] = e[0]
-                arr[i,2] = e[1]
-            cc[r] = arr
-        def myL1(A, B, D):
-            return abs(A[0]-B[0]) + abs(A[1]-B[1])
-        def myL2(A, B, D):
-            return math.sqrt((A[0]-B[0])**2 + (A[1]-B[1])**2)
-        def myC(A, B, D):
-            return max(abs(A[0]-B[0]), abs(A[1]-B[1]))
-        contours = set(cc.values())
-        for c0 in contours:
-            for c1 in contours:
-                self.assert_(abs(cv.CalcEMD2(c0, c1, cv.CV_DIST_L1) - cv.CalcEMD2(c0, c1, cv.CV_DIST_USER, myL1)) < 1e-3)
-                self.assert_(abs(cv.CalcEMD2(c0, c1, cv.CV_DIST_L2) - cv.CalcEMD2(c0, c1, cv.CV_DIST_USER, myL2)) < 1e-3)
-                self.assert_(abs(cv.CalcEMD2(c0, c1, cv.CV_DIST_C) - cv.CalcEMD2(c0, c1, cv.CV_DIST_USER, myC)) < 1e-3)
-
-    def test_CalcOpticalFlowBM(self):
-        a = self.get_sample("samples/c/lena.jpg", 0)
-        b = self.get_sample("samples/c/lena.jpg", 0)
-        (w,h) = cv.GetSize(a)
-        vel_size = (w - 8 + 1, h - 8 + 1)
-        velx = cv.CreateImage(vel_size, cv.IPL_DEPTH_32F, 1)
-        vely = cv.CreateImage(vel_size, cv.IPL_DEPTH_32F, 1)
-        cv.CalcOpticalFlowBM(a, b, (8,8), (1,1), (8,8), 0, velx, vely)
-
-    def test_CalcOpticalFlowPyrLK(self):
-        a = self.get_sample("samples/c/lena.jpg", 0)
-        map = cv.CreateMat(2, 3, cv.CV_32FC1)
-        cv.GetRotationMatrix2D((256, 256), 10, 1.0, map)
-        b = cv.CloneMat(a)
-        cv.WarpAffine(a, b, map)
-
-        eig_image = cv.CreateMat(a.rows, a.cols, cv.CV_32FC1)
-        temp_image = cv.CreateMat(a.rows, a.cols, cv.CV_32FC1)
-
-        prevPyr = cv.CreateMat(a.rows / 3, a.cols + 8, cv.CV_8UC1)
-        currPyr = cv.CreateMat(a.rows / 3, a.cols + 8, cv.CV_8UC1)
-        prevFeatures = cv.GoodFeaturesToTrack(a, eig_image, temp_image, 400, 0.01, 0.01)
-        (currFeatures, status, track_error) = cv.CalcOpticalFlowPyrLK(a,
-                                                                      b,
-                                                                      prevPyr,
-                                                                      currPyr,
-                                                                      prevFeatures,
-                                                                      (10, 10),
-                                                                      3,
-                                                                      (cv.CV_TERMCRIT_ITER|cv.CV_TERMCRIT_EPS,20, 0.03),
-                                                                      0)
-        if 0:  # enable visualization
-            print
-            print sum(status), "Points found in curr image"
-            for prev,this in zip(prevFeatures, currFeatures):
-                iprev = tuple([int(c) for c in prev])
-                ithis = tuple([int(c) for c in this])
-                cv.Circle(a, iprev, 3, 255)
-                cv.Circle(a, ithis, 3, 0)
-                cv.Line(a, iprev, ithis, 128)
-
-            self.snapL([a, b])
-
-    def test_CartToPolar(self):
-        x = cv.CreateMat(5, 5, cv.CV_32F)
-        y = cv.CreateMat(5, 5, cv.CV_32F)
-        mag = cv.CreateMat(5, 5, cv.CV_32F)
-        angle = cv.CreateMat(5, 5, cv.CV_32F)
-        x2 = cv.CreateMat(5, 5, cv.CV_32F)
-        y2 = cv.CreateMat(5, 5, cv.CV_32F)
-
-        for i in range(5):
-            for j in range(5):
-                x[i, j] = i
-                y[i, j] = j
-
-        for in_degrees in [False, True]:
-            cv.CartToPolar(x, y, mag, angle, in_degrees)
-            cv.PolarToCart(mag, angle, x2, y2, in_degrees)
-            for i in range(5):
-                for j in range(5):
-                    self.assertAlmostEqual(x[i, j], x2[i, j], 1)
-                    self.assertAlmostEqual(y[i, j], y2[i, j], 1)
-
-    def test_Circle(self):
-        for w,h in [(2,77), (77,2), (256, 256), (640,480)]:
-            img = cv.CreateImage((w,h), cv.IPL_DEPTH_8U, 1)
-            cv.SetZero(img)
-            tricky = [ -8000, -2, -1, 0, 1, h/2, h-1, h, h+1, w/2, w-1, w, w+1, 8000]
-            for x0 in tricky:
-                for y0 in tricky:
-                    for r in [ 0, 1, 2, 3, 4, 5, w/2, w-1, w, w+1, h/2, h-1, h, h+1, 8000 ]:
-                        for thick in [1, 2, 10]:
-                            for t in [0, 8, 4, cv.CV_AA]:
-                                cv.Circle(img, (x0,y0), r, 255, thick, t)
-        # just check that something was drawn
-        self.assert_(cv.Sum(img)[0] > 0)
-
-    def test_ConvertImage(self):
-        i1 = cv.GetImage(self.get_sample("samples/c/lena.jpg", 1))
-        i2 = cv.CloneImage(i1)
-        i3 = cv.CloneImage(i1)
-        cv.ConvertImage(i1, i2, cv.CV_CVTIMG_FLIP + cv.CV_CVTIMG_SWAP_RB)
-        self.assertNotEqual(self.hashimg(i1), self.hashimg(i2))
-        cv.ConvertImage(i2, i3, cv.CV_CVTIMG_FLIP + cv.CV_CVTIMG_SWAP_RB)
-        self.assertEqual(self.hashimg(i1), self.hashimg(i3))
-
-    def test_ConvexHull2(self):
-        # Draw a series of N-pointed stars, find contours, assert the contour is not convex,
-        # assert the hull has N segments, assert that there are N convexity defects.
-
-        def polar2xy(th, r):
-            return (int(400 + r * math.cos(th)), int(400 + r * math.sin(th)))
-        storage = cv.CreateMemStorage(0)
-        for way in ['CvSeq', 'CvMat', 'list']:
-            for points in range(3,20):
-                scratch = cv.CreateImage((800,800), 8, 1)
-                cv.SetZero(scratch)
-                sides = 2 * points
-                cv.FillPoly(scratch, [ [ polar2xy(i * 2 * math.pi / sides, [100,350][i&1]) for i in range(sides) ] ], 255)
-
-                seq = cv.FindContours(scratch, storage, cv.CV_RETR_TREE, cv.CV_CHAIN_APPROX_SIMPLE)
-
-                if way == 'CvSeq':
-                    # pts is a CvSeq
-                    pts = seq
-                elif way == 'CvMat':
-                    # pts is a CvMat
-                    arr = cv.CreateMat(len(seq), 1, cv.CV_32SC2)
-                    for i,e in enumerate(seq):
-                        arr[i,0] = e
-                    pts = arr
-                elif way == 'list':
-                    # pts is a list of 2-tuples
-                    pts = list(seq)
-                else:
-                    assert False
-
-                self.assert_(cv.CheckContourConvexity(pts) == 0)
-                hull = cv.ConvexHull2(pts, storage, return_points = 1)
-                self.assert_(cv.CheckContourConvexity(hull) == 1)
-                self.assert_(len(hull) == points)
-
-                if way in [ 'CvSeq', 'CvMat' ]:
-                    defects = cv.ConvexityDefects(pts, cv.ConvexHull2(pts, storage), storage)
-                    self.assert_(len([depth for (_,_,_,depth) in defects if (depth > 5)]) == points)
-
-    def test_CreateImage(self):
-        for w in [ 1, 4, 64, 512, 640]:
-            for h in [ 1, 4, 64, 480, 512]:
-                for c in [1, 2, 3, 4]:
-                    for d in self.depths:
-                        a = cv.CreateImage((w,h), d, c);
-                        self.assert_(a.width == w)
-                        self.assert_(a.height == h)
-                        self.assert_(a.nChannels == c)
-                        self.assert_(a.depth == d)
-                        self.assert_(cv.GetSize(a) == (w, h))
-                        # self.assert_(cv.GetElemType(a) == d)
-        self.assertRaises(cv.error, lambda: cv.CreateImage((100, 100), 9, 1))
-
-    def test_CreateMat(self):
-        for rows in [1, 2, 4, 16, 64, 512, 640]:
-            for cols in [1, 2, 4, 16, 64, 512, 640]:
-                for t in self.mat_types:
-                    m = cv.CreateMat(rows, cols, t)
-                    self.assertEqual(cv.GetElemType(m), t)
-                    self.assertEqual(m.type, t)
-        self.assertRaises(cv.error, lambda: cv.CreateMat(-1, 100, cv.CV_8SC4))
-        self.assertRaises(cv.error, lambda: cv.CreateMat(100, -1, cv.CV_8SC4))
-        self.assertRaises(cv.error, lambda: cv.cvmat())
-
-    def test_DrawChessboardCorners(self):
-        im = cv.CreateImage((512,512), cv.IPL_DEPTH_8U, 3)
-        cv.SetZero(im)
-        cv.DrawChessboardCorners(im, (5, 5), [ ((i/5)*100+50,(i%5)*100+50) for i in range(5 * 5) ], 1)
-
-    def test_ExtractSURF(self):
-        img = self.get_sample("samples/c/lena.jpg", 0)
-        w,h = cv.GetSize(img)
-        for hessthresh in [ 300,400,500]:
-            for dsize in [0,1]:
-                for layers in [1,3,10]:
-                    kp,desc = cv.ExtractSURF(img, None, cv.CreateMemStorage(), (dsize, hessthresh, 3, layers))
-                    self.assert_(len(kp) == len(desc))
-                    for d in desc:
-                        self.assert_(len(d) == {0:64, 1:128}[dsize])
-                    for pt,laplacian,size,dir,hessian in kp:
-                        self.assert_((0 <= pt[0]) and (pt[0] <= w))
-                        self.assert_((0 <= pt[1]) and (pt[1] <= h))
-                        self.assert_(laplacian in [-1, 0, 1])
-                        self.assert_((0 <= dir) and (dir <= 360))
-                        self.assert_(hessian >= hessthresh)
-
-    def test_FillPoly(self):
-        scribble = cv.CreateImage((640,480), cv.IPL_DEPTH_8U, 1)
-        random.seed(0)
-        for i in range(50):
-            cv.SetZero(scribble)
-            self.assert_(cv.CountNonZero(scribble) == 0)
-            cv.FillPoly(scribble, [ [ (random.randrange(640), random.randrange(480)) for i in range(100) ] ], (255,))
-            self.assert_(cv.CountNonZero(scribble) != 0)
-
-    def test_FindChessboardCorners(self):
-        im = cv.CreateImage((512,512), cv.IPL_DEPTH_8U, 1)
-        cv.Set(im, 128)
-
-        # Empty image run
-        status,corners = cv.FindChessboardCorners( im, (7,7) )
-
-        # Perfect checkerboard
-        def xf(i,j, o):
-            return ((96 + o) + 40 * i, (96 + o) + 40 * j)
-        for i in range(8):
-            for j in range(8):
-                color = ((i ^ j) & 1) * 255
-                cv.Rectangle(im, xf(i,j, 0), xf(i,j, 39), color, cv.CV_FILLED)
-        status,corners = cv.FindChessboardCorners( im, (7,7) )
-        self.assert_(status)
-        self.assert_(len(corners) == (7 * 7))
-
-        # Exercise corner display
-        im3 = cv.CreateImage(cv.GetSize(im), cv.IPL_DEPTH_8U, 3)
-        cv.Merge(im, im, im, None, im3)
-        cv.DrawChessboardCorners(im3, (7,7), corners, status)
-
-        if 0:
-            self.snap(im3)
-
-        # Run it with too many corners
-        cv.Set(im, 128)
-        for i in range(40):
-            for j in range(40):
-                color = ((i ^ j) & 1) * 255
-                x = 30 + 6 * i
-                y = 30 + 4 * j
-                cv.Rectangle(im, (x, y), (x+4, y+4), color, cv.CV_FILLED)
-        status,corners = cv.FindChessboardCorners( im, (7,7) )
-
-        # XXX - this is very slow
-        if 0:
-            rng = cv.RNG(0)
-            cv.RandArr(rng, im, cv.CV_RAND_UNI, 0, 255.0)
-            self.snap(im)
-            status,corners = cv.FindChessboardCorners( im, (7,7) )
-
-    def test_FindContours(self):
-        random.seed(0)
-
-        storage = cv.CreateMemStorage()
-
-        # First run FindContours on a black image.
-        for mode in [cv.CV_RETR_EXTERNAL, cv.CV_RETR_LIST, cv.CV_RETR_CCOMP, cv.CV_RETR_TREE]:
-            for method in [cv.CV_CHAIN_CODE, cv.CV_CHAIN_APPROX_NONE, cv.CV_CHAIN_APPROX_SIMPLE, cv.CV_CHAIN_APPROX_TC89_L1, cv.CV_CHAIN_APPROX_TC89_KCOS, cv.CV_LINK_RUNS]:
-                scratch = cv.CreateImage((800,800), 8, 1)
-                cv.SetZero(scratch)
-                seq = cv.FindContours(scratch, storage, mode, method)
-                x = len(seq)
-                if seq:
-                    pass
-                for s in seq:
-                    pass
-
-        for trial in range(10):
-            scratch = cv.CreateImage((800,800), 8, 1)
-            cv.SetZero(scratch)
-            def plot(center, radius, mode):
-                cv.Circle(scratch, center, radius, mode, -1)
-                if radius < 20:
-                    return 0
-                else:
-                    newmode = 255 - mode
-                    subs = random.choice([1,2,3])
-                    if subs == 1:
-                        return [ plot(center, radius - 5, newmode) ]
-                    else:
-                        newradius = int({ 2: radius / 2, 3: radius / 2.3 }[subs] - 5)
-                        r = radius / 2
-                        ret = []
-                        for i in range(subs):
-                            th = i * (2 * math.pi) / subs
-                            ret.append(plot((int(center[0] + r * math.cos(th)), int(center[1] + r * math.sin(th))), newradius, newmode))
-                        return sorted(ret)
-
-            actual = plot((400,400), 390, 255 )
-
-            seq = cv.FindContours(scratch, storage, cv.CV_RETR_TREE, cv.CV_CHAIN_APPROX_SIMPLE)
-
-            def traverse(s):
-                if s == None:
-                    return 0
-                else:
-                    self.assert_(abs(cv.ContourArea(s)) > 0.0)
-                    ((x,y),(w,h),th) = cv.MinAreaRect2(s, cv.CreateMemStorage())
-                    self.assert_(((w / h) - 1.0) < 0.01)
-                    self.assert_(abs(cv.ContourArea(s)) > 0.0)
-                    r = []
-                    while s:
-                        r.append(traverse(s.v_next()))
-                        s = s.h_next()
-                    return sorted(r)
-            self.assert_(traverse(seq.v_next()) == actual)
-
-        if 1:
-            original = cv.CreateImage((800,800), 8, 1)
-            cv.SetZero(original)
-            cv.Circle(original, (400, 400), 200, 255, -1)
-            cv.Circle(original, (100, 100), 20, 255, -1)
-        else:
-            original = self.get_sample("samples/c/lena.jpg", 0)
-            cv.Threshold(original, original, 128, 255, cv.CV_THRESH_BINARY);
-
-        contours = cv.FindContours(original, storage, cv.CV_RETR_CCOMP, cv.CV_CHAIN_APPROX_SIMPLE)
-
-
-        def contour_iterator(contour):
-            while contour:
-                yield contour
-                contour = contour.h_next()
-
-        # Should be 2 contours from the two circles above
-        self.assertEqual(len(list(contour_iterator(contours))), 2)
-
-        # Smoke DrawContours
-        sketch = cv.CreateImage(cv.GetSize(original), 8, 3)
-        cv.SetZero(sketch)
-        red = cv.RGB(255, 0, 0)
-        green = cv.RGB(0, 255, 0)
-        for c in contour_iterator(contours):
-            cv.DrawContours(sketch, c, red, green, 0)
-        # self.snap(sketch)
-
-    def test_GetAffineTransform(self):
-        mapping = cv.CreateMat(2, 3, cv.CV_32FC1)
-        cv.GetAffineTransform([ (0,0), (1,0), (0,1) ], [ (0,0), (17,0), (0,17) ], mapping)
-        self.assertAlmostEqual(mapping[0,0], 17, 2)
-        self.assertAlmostEqual(mapping[1,1], 17, 2)
-
-    def test_GetRotationMatrix2D(self):
-        mapping = cv.CreateMat(2, 3, cv.CV_32FC1)
-        for scale in [0.0, 1.0, 2.0]:
-            for angle in [0.0, 360.0]:
-                cv.GetRotationMatrix2D((0,0), angle, scale, mapping)
-                for r in [0, 1]:
-                    for c in [0, 1, 2]:
-                        if r == c:
-                            e = scale
-                        else:
-                            e = 0.0
-                        self.assertAlmostEqual(mapping[r, c], e, 2)
-
-    def test_GetSize(self):
-        self.assert_(cv.GetSize(cv.CreateMat(5, 7, cv.CV_32FC1)) == (7,5))
-        self.assert_(cv.GetSize(cv.CreateImage((7,5), cv.IPL_DEPTH_8U, 1)) == (7,5))
-
-    def test_GetStarKeypoints(self):
-        src = self.get_sample("samples/c/lena.jpg", 0)
-        storage = cv.CreateMemStorage()
-        kp = cv.GetStarKeypoints(src, storage)
-        self.assert_(len(kp) > 0)
-        for (x,y),scale,r in kp:
-            self.assert_(0 <= x)
-            self.assert_(x <= cv.GetSize(src)[0])
-            self.assert_(0 <= y)
-            self.assert_(y <= cv.GetSize(src)[1])
-        return
-        scribble = cv.CreateImage(cv.GetSize(src), 8, 3)
-        cv.CvtColor(src, scribble, cv.CV_GRAY2BGR)
-        for (x,y),scale,r in kp:
-            print x,y,scale,r
-            cv.Circle(scribble, (x,y), scale, cv.RGB(255,0,0))
-        self.snap(scribble)
-
-    def test_GetSubRect(self):
-        src = cv.CreateImage((100,100), 8, 1)
-        data = "z" * (100 * 100)
-
-        cv.SetData(src, data, 100)
-        start_count = sys.getrefcount(data)
-
-        iter = 77
-        subs = []
-        for i in range(iter):
-            sub = cv.GetSubRect(src, (0, 0, 10, 10))
-            subs.append(sub)
-        self.assert_(sys.getrefcount(data) == (start_count + iter))
-
-        src = self.get_sample("samples/c/lena.jpg", 0)
-        made = cv.CreateImage(cv.GetSize(src), 8, 1)
-        sub = cv.CreateMat(32, 32, cv.CV_8UC1)
-        for x in range(0, 512, 32):
-            for y in range(0, 512, 32):
-                sub = cv.GetSubRect(src, (x, y, 32, 32))
-                cv.SetImageROI(made, (x, y, 32, 32))
-                cv.Copy(sub, made)
-        cv.ResetImageROI(made)
-        cv.AbsDiff(made, src, made)
-        self.assert_(cv.CountNonZero(made) == 0)
-
-        for m1 in [cv.CreateMat(1, 10, cv.CV_8UC1), cv.CreateImage((10, 1), 8, 1)]:
-            for i in range(10):
-                m1[0, i] = i
-            def aslist(cvmat): return list(array.array('B', cvmat.tostring()))
-            m2 = cv.GetSubRect(m1, (5, 0, 4, 1))
-            m3 = cv.GetSubRect(m2, (1, 0, 2, 1))
-            self.assertEqual(aslist(m1), range(10))
-            self.assertEqual(aslist(m2), range(5, 9))
-            self.assertEqual(aslist(m3), range(6, 8))
-
-    def xtest_grabCut(self):
-        image = self.get_sample("samples/c/lena.jpg", cv.CV_LOAD_IMAGE_COLOR)
-        tmp1 = cv.CreateMat(1, 13 * 5, cv.CV_32FC1)
-        tmp2 = cv.CreateMat(1, 13 * 5, cv.CV_32FC1)
-        mask = cv.CreateMat(image.rows, image.cols, cv.CV_8UC1)
-        cv.GrabCut(image, mask, (10,10,200,200), tmp1, tmp2, 10, cv.GC_INIT_WITH_RECT)
-
-    def test_HoughLines2_PROBABILISTIC(self):
-        li = cv.HoughLines2(self.yield_line_image(),
-                                                cv.CreateMemStorage(),
-                                                cv.CV_HOUGH_PROBABILISTIC,
-                                                1,
-                                                math.pi/180,
-                                                50,
-                                                50,
-                                                10)
-        self.assert_(len(li) > 0)
-        self.assert_(li[0] != None)
-
-    def test_HoughLines2_STANDARD(self):
-        li = cv.HoughLines2(self.yield_line_image(),
-                                                cv.CreateMemStorage(),
-                                                cv.CV_HOUGH_STANDARD,
-                                                1,
-                                                math.pi/180,
-                                                100,
-                                                0,
-                                                0)
-        self.assert_(len(li) > 0)
-        self.assert_(li[0] != None)
-
-    def test_InPaint(self):
-        src = self.get_sample("samples/cpp/building.jpg")
-        msk = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_8U, 1)
-        damaged = cv.CloneMat(src)
-        repaired = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_8U, 3)
-        difference = cv.CloneImage(repaired)
-        cv.SetZero(msk)
-        for method in [ cv.CV_INPAINT_NS, cv.CV_INPAINT_TELEA ]:
-            for (p0,p1) in [ ((10,10), (400,400)) ]:
-                cv.Line(damaged, p0, p1, cv.RGB(255, 0, 255), 2)
-                cv.Line(msk, p0, p1, 255, 2)
-            cv.Inpaint(damaged, msk, repaired, 10., cv.CV_INPAINT_NS)
-        cv.AbsDiff(src, repaired, difference)
-        #self.snapL([src, damaged, repaired, difference])
-
-    def test_InitLineIterator(self):
-        scribble = cv.CreateImage((640,480), cv.IPL_DEPTH_8U, 1)
-        self.assert_(len(list(cv.InitLineIterator(scribble, (20,10), (30,10)))) == 11)
-
-    def test_InRange(self):
-
-        sz = (256,256)
-        Igray1 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-        Ilow1 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-        Ihi1 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-        Igray2 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-        Ilow2 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-        Ihi2 = cv.CreateImage(sz,cv.IPL_DEPTH_32F,1)
-
-        Imask = cv.CreateImage(sz, cv.IPL_DEPTH_8U,1)
-        Imaskt = cv.CreateImage(sz,cv.IPL_DEPTH_8U,1)
-
-        cv.InRange(Igray1, Ilow1, Ihi1, Imask);
-        cv.InRange(Igray2, Ilow2, Ihi2, Imaskt);
-
-        cv.Or(Imask, Imaskt, Imask);
-
-    def test_Line(self):
-        w,h = 640,480
-        img = cv.CreateImage((w,h), cv.IPL_DEPTH_8U, 1)
-        cv.SetZero(img)
-        tricky = [ -8000, -2, -1, 0, 1, h/2, h-1, h, h+1, w/2, w-1, w, w+1, 8000]
-        for x0 in tricky:
-            for y0 in tricky:
-                for x1 in tricky:
-                    for y1 in tricky:
-                        for thickness in [ 0, 1, 8 ]:
-                            for line_type in [0, 4, 8, cv.CV_AA ]:
-                                cv.Line(img, (x0,y0), (x1,y1), 255, thickness, line_type)
-        # just check that something was drawn
-        self.assert_(cv.Sum(img)[0] > 0)
-
-    def test_MinMaxLoc(self):
-        scribble = cv.CreateImage((640,480), cv.IPL_DEPTH_8U, 1)
-        los = [ (random.randrange(480), random.randrange(640)) for i in range(100) ]
-        his = [ (random.randrange(480), random.randrange(640)) for i in range(100) ]
-        for (lo,hi) in zip(los,his):
-            cv.Set(scribble, 128)
-            scribble[lo] = 0
-            scribble[hi] = 255
-            r = cv.MinMaxLoc(scribble)
-            self.assert_(r == (0, 255, tuple(reversed(lo)), tuple(reversed(hi))))
-
-    def xxx_test_PyrMeanShiftFiltering(self):   # XXX - ticket #306
-        if 0:
-            src = self.get_sample("samples/c/lena.jpg", cv.CV_LOAD_IMAGE_COLOR)
-            dst = cv.CloneMat(src)
-            cv.PyrMeanShiftFiltering(src, dst, 5, 5)
-            print src, dst
-            self.snap(src)
-        else:
-            r = cv.temp_test()
-            print r
-            print len(r.tostring())
-            self.snap(r)
-
-    def test_Reshape(self):
-        # 97 rows
-        # 12 cols
-        rows = 97
-        cols = 12
-        im = cv.CreateMat( rows, cols, cv.CV_32FC1 )
-        elems = rows * cols * 1
-        def crd(im):
-            return cv.GetSize(im) + (cv.CV_MAT_CN(cv.GetElemType(im)),)
-
-        for c in (1, 2, 3, 4):
-            nc,nr,nd = crd(cv.Reshape(im, c))
-            self.assert_(nd == c)
-            self.assert_((nc * nr * nd) == elems)
-
-        nc,nr,nd = crd(cv.Reshape(im, 0, 97*2))
-        self.assert_(nr == 97*2)
-        self.assert_((nc * nr * nd) == elems)
-
-        nc,nr,nd = crd(cv.Reshape(im, 3, 97*2))
-        self.assert_(nr == 97*2)
-        self.assert_(nd == 3)
-        self.assert_((nc * nr * nd) == elems)
-
-        # Now test ReshapeMatND
-        mat = cv.CreateMatND([24], cv.CV_32FC1)
-        cv.Set(mat, 1.0)
-        self.assertEqual(cv.GetDims(cv.ReshapeMatND(mat, 0, [24, 1])), (24, 1))
-        self.assertEqual(cv.GetDims(cv.ReshapeMatND(mat, 0, [6, 4])), (6, 4))
-        self.assertEqual(cv.GetDims(cv.ReshapeMatND(mat, 24, [1])), (1,))
-        self.assertRaises(TypeError, lambda: cv.ReshapeMatND(mat, 12, [1]))
-
-    def test_Save(self):
-        for o in [ cv.CreateImage((128,128), cv.IPL_DEPTH_8U, 1), cv.CreateMat(16, 16, cv.CV_32FC1), cv.CreateMatND([7,9,4], cv.CV_32FC1) ]:
-            cv.Save("test.save", o)
-            loaded = cv.Load("test.save", cv.CreateMemStorage())
-            self.assert_(type(o) == type(loaded))
-
-    def test_SetIdentity(self):
-        for r in range(1,16):
-            for c in range(1, 16):
-                for t in self.mat_types_single:
-                    M = cv.CreateMat(r, c, t)
-                    cv.SetIdentity(M)
-                    for rj in range(r):
-                        for cj in range(c):
-                            if rj == cj:
-                                expected = 1.0
-                            else:
-                                expected = 0.0
-                            self.assertEqual(M[rj,cj], expected)
-
-    def test_SnakeImage(self):
-        src = self.get_sample("samples/c/lena.jpg", 0)
-        pts = [ (512-i,i) for i in range(0, 512, 8) ]
-
-        # Make sure that weight arguments get validated
-        self.assertRaises(TypeError, lambda: cv.SnakeImage(cv.GetImage(src), pts, [1,2], .01, .01, (7,7), (cv.CV_TERMCRIT_ITER, 100, 0.1)))
-
-        # Smoke by making sure that points are changed by call
-        r = cv.SnakeImage(cv.GetImage(src), pts, .01, .01, .01, (7,7), (cv.CV_TERMCRIT_ITER, 100, 0.1))
-        if 0:
-            cv.PolyLine(src, [ r ], 0, 255)
-            self.snap(src)
-        self.assertEqual(len(r), len(pts))
-        self.assertNotEqual(r, pts)
-
-        # Ensure that list of weights is same as scalar weight
-        w = [.01] * len(pts)
-        r2 = cv.SnakeImage(cv.GetImage(src), pts, w, w, w, (7,7), (cv.CV_TERMCRIT_ITER, 100, 0.1))
-        self.assertEqual(r, r2)
-
-    def test_KMeans2(self):
-        size = 500
-        samples = cv.CreateMat(size, 1, cv.CV_32FC3)
-        labels = cv.CreateMat(size, 1, cv.CV_32SC1)
-        centers = cv.CreateMat(2, 3, cv.CV_32FC1)
-
-        cv.Zero(samples)
-        cv.Zero(labels)
-        cv.Zero(centers)
-
-        cv.Set(cv.GetSubRect(samples, (0, 0, 1, size/2)), (255, 255, 255))
-
-        compact = cv.KMeans2(samples, 2, labels, (cv.CV_TERMCRIT_ITER, 100, 0.1), 1, 0, centers)
-
-        self.assertEqual(int(compact), 0)
-
-        random.seed(0)
-        for i in range(50):
-            index = random.randrange(size)
-            if index < size/2:
-                self.assertEqual(samples[index, 0], (255, 255, 255))
-                self.assertEqual(labels[index, 0], 1)
-            else:
-                self.assertEqual(samples[index, 0], (0, 0, 0))
-                self.assertEqual(labels[index, 0], 0)
-
-        for cluster in (0, 1):
-            for channel in (0, 1, 2):
-                self.assertEqual(int(centers[cluster, channel]), cluster*255)
-
-    def test_Sum(self):
-        for r in range(1,11):
-            for c in range(1, 11):
-                for t in self.mat_types_single:
-                    M = cv.CreateMat(r, c, t)
-                    cv.Set(M, 1)
-                    self.assertEqual(cv.Sum(M)[0], r * c)
-
-    def test_Threshold(self):
-    #""" directed test for bug 2790622 """
-        src = self.get_sample("samples/c/lena.jpg", 0)
-        results = set()
-        for i in range(10):
-            dst = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_8U, 1)
-            cv.Threshold(src, dst, 128, 128, cv.CV_THRESH_BINARY)
-            results.add(dst.tostring())
-        # Should have produced the same answer every time, so results set should have size 1
-        self.assert_(len(results) == 1)
-
-        # ticket #71 repro attempt
-        image = self.get_sample("samples/c/lena.jpg", 0)
-        red = cv.CreateImage(cv.GetSize(image), 8, 1)
-        binary = cv.CreateImage(cv.GetSize(image), 8, 1)
-        cv.Split(image, red, None, None, None)
-        cv.Threshold(red, binary, 42, 255, cv.CV_THRESH_BINARY)
-
-    ##############################################################################
-
-    def yield_line_image(self):
-        """ Needed by HoughLines tests """
-        src = self.get_sample("samples/cpp/building.jpg", 0)
-        dst = cv.CreateImage(cv.GetSize(src), 8, 1)
-        cv.Canny(src, dst, 50, 200, 3)
-        return dst
-
-# Tests for functional areas
-
-class AreaTests(OpenCVTests):
-
-    def test_numpy(self):
-        if 'fromarray' in dir(cv):
-            import numpy
-
-            def convert(numpydims):
-                """ Create a numpy array with specified dims, return the OpenCV CvMat """
-                a1 = numpy.array([1] * reduce(operator.__mul__, numpydims)).reshape(*numpydims).astype(numpy.float32)
-                return cv.fromarray(a1)
-            def row_col_chan(m):
-                col = m.cols
-                row = m.rows
-                chan = cv.CV_MAT_CN(cv.GetElemType(m))
-                return (row, col, chan)
-
-            self.assertEqual(row_col_chan(convert((2, 13))), (2, 13, 1))
-            self.assertEqual(row_col_chan(convert((2, 13, 4))), (2, 13, 4))
-            self.assertEqual(row_col_chan(convert((2, 13, cv.CV_CN_MAX))), (2, 13, cv.CV_CN_MAX))
-            self.assertRaises(TypeError, lambda: convert((2,)))
-            self.assertRaises(TypeError, lambda: convert((11, 17, cv.CV_CN_MAX + 1)))
-
-            for t in [cv.CV_16UC1, cv.CV_32SC1, cv.CV_32FC1]:
-                for d in [ (8,), (1,7), (2,3,4), (7,9,2,1,8), (1,2,3,4,5,6,7,8) ]:
-                    total = reduce(operator.__mul__, d)
-                    m = cv.CreateMatND(d, t)
-                    for i in range(total):
-                        cv.Set1D(m, i, i)
-                    na = numpy.asarray(m).reshape((total,))
-                    self.assertEqual(list(na), range(total))
-
-                    # now do numpy -> cvmat, and verify
-                    m2 = cv.fromarray(na, True)
-
-                    # Check that new cvmat m2 contains same counting sequence
-                    for i in range(total):
-                        self.assertEqual(cv.Get1D(m, i)[0], i)
-
-            # Verify round-trip for 2D arrays
-            for rows in [2, 3, 7, 13]:
-                for cols in [2, 3, 7, 13]:
-                    for allowND in [False, True]:
-                        im = cv.CreateMatND([rows, cols], cv.CV_16UC1)
-                        cv.SetZero(im)
-                        a = numpy.asarray(im)
-                        self.assertEqual(a.shape, (rows, cols))
-                        cvmatnd = cv.fromarray(a, allowND)
-                        self.assertEqual(cv.GetDims(cvmatnd), (rows, cols))
-
-                        # im, a and cvmatnd all point to the same data, so...
-                        for i,coord in enumerate([(0,0), (0,1), (1,0), (1,1)]):
-                            v = 5 + i + 7
-                            a[coord] = v
-                            self.assertEqual(im[coord], v)
-                            self.assertEqual(cvmatnd[coord], v)
-
-            # Cv -> Numpy 3 channel check
-            im = cv.CreateMatND([2, 13], cv.CV_16UC3)
-            self.assertEqual(numpy.asarray(im).shape, (2, 13, 3))
-
-            # multi-dimensional NumPy array
-            na = numpy.ones([7,9,2,1,8])
-            cm = cv.fromarray(na, True)
-            self.assertEqual(cv.GetDims(cm), (7,9,2,1,8))
-
-            # Using an array object for a CvArr parameter
-            ones = numpy.ones((640, 480))
-            r = cv.fromarray(numpy.ones((640, 480)))
-            cv.AddS(cv.fromarray(ones), 7, r)
-            self.assert_(numpy.alltrue(r == (8 * ones)))
-
-            # create arrays, use them in OpenCV and replace the the array
-            # looking for leaks
-            def randdim():
-                return [random.randrange(1,6) for i in range(random.randrange(1, 6))]
-            arrays = [numpy.ones(randdim()).astype(numpy.uint8) for i in range(10)]
-            cs = [cv.fromarray(a, True) for a in arrays]
-            for i in range(1000):
-                arrays[random.randrange(10)] = numpy.ones(randdim()).astype(numpy.uint8)
-                cs[random.randrange(10)] = cv.fromarray(arrays[random.randrange(10)], True)
-                for j in range(10):
-                    self.assert_(all([c == chr(1) for c in cs[j].tostring()]))
-
-            #
-            m = numpy.identity(4, dtype = numpy.float32)
-            m = cv.fromarray(m[:3, :3])
-            rvec = cv.CreateMat(3, 1, cv.CV_32FC1)
-            rvec[0,0] = 1
-            rvec[1,0] = 1
-            rvec[2,0] = 1
-            cv.Rodrigues2(rvec, m)
-        #print m
-
-        else:
-            print "SKIPPING test_numpy - numpy support not built"
-
-    def test_boundscatch(self):
-        l2 = cv.CreateMat(256, 1, cv.CV_8U)
-        l2[0,0]     # should be OK
-        self.assertRaises(cv.error, lambda: l2[1,1])
-        l2[0]       # should be OK
-        self.assertRaises(cv.error, lambda: l2[299])
-        for n in range(1, 8):
-            l = cv.CreateMatND([2] * n, cv.CV_8U)
-            l[0] # should be OK
-            self.assertRaises(cv.error, lambda: l[999])
-
-            tup0 = (0,) * n
-            l[tup0] # should be OK
-            tup2 = (2,) * n
-            self.assertRaises(cv.error, lambda: l[tup2])
-
-    def test_stereo(self):
-        bm = cv.CreateStereoBMState()
-        def illegal_delete():
-            bm = cv.CreateStereoBMState()
-            del bm.preFilterType
-        def illegal_assign():
-            bm = cv.CreateStereoBMState()
-            bm.preFilterType = "foo"
-
-        self.assertRaises(TypeError, illegal_delete)
-        self.assertRaises(TypeError, illegal_assign)
-
-        left = self.get_sample("samples/c/lena.jpg", 0)
-        right = self.get_sample("samples/c/lena.jpg", 0)
-        disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-        cv.FindStereoCorrespondenceBM(left, right, disparity, bm)
-
-        gc = cv.CreateStereoGCState(16, 2)
-        left_disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-        right_disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-
-    def test_stereo(self):
-        bm = cv.CreateStereoBMState()
-        def illegal_delete():
-            bm = cv.CreateStereoBMState()
-            del bm.preFilterType
-        def illegal_assign():
-            bm = cv.CreateStereoBMState()
-            bm.preFilterType = "foo"
-
-        self.assertRaises(TypeError, illegal_delete)
-        self.assertRaises(TypeError, illegal_assign)
-
-        left = self.get_sample("samples/c/lena.jpg", 0)
-        right = self.get_sample("samples/c/lena.jpg", 0)
-        disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-        cv.FindStereoCorrespondenceBM(left, right, disparity, bm)
-
-        gc = cv.CreateStereoGCState(16, 2)
-        left_disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-        right_disparity = cv.CreateMat(512, 512, cv.CV_16SC1)
-        cv.FindStereoCorrespondenceGC(left, right, left_disparity, right_disparity, gc)
-
-    def test_kalman(self):
-        k = cv.CreateKalman(2, 1, 0)
-
-    def failing_test_exception(self):
-        a = cv.CreateImage((640, 480), cv.IPL_DEPTH_8U, 1)
-        b = cv.CreateImage((640, 480), cv.IPL_DEPTH_8U, 1)
-        self.assertRaises(cv.error, lambda: cv.Laplace(a, b))
-
-    def test_cvmat_accessors(self):
-        cvm = cv.CreateMat(20, 10, cv.CV_32FC1)
-
-    def test_depths(self):
-    #""" Make sure that the depth enums are unique """
-        self.assert_(len(self.depths) == len(set(self.depths)))
-
-    def test_leak(self):
-    #""" If CreateImage is not releasing image storage, then the loop below should use ~4GB of memory. """
-        for i in range(64000):
-            a = cv.CreateImage((1024,1024), cv.IPL_DEPTH_8U, 1)
-        for i in range(64000):
-            a = cv.CreateMat(1024, 1024, cv.CV_8UC1)
-
-    def test_histograms(self):
-        def split(im):
-            nchans = cv.CV_MAT_CN(cv.GetElemType(im))
-            c = [ cv.CreateImage(cv.GetSize(im), cv.IPL_DEPTH_8U, 1) for i in range(nchans) ] + [None] * (4 - nchans)
-            cv.Split(im, c[0], c[1], c[2], c[3])
-            return c[:nchans]
-        def imh(im):
-            s = split(im)
-            hist = cv.CreateHist([256] * len(s), cv.CV_HIST_ARRAY, [ (0,255) ] * len(s), 1)
-            cv.CalcHist(s, hist, 0)
-            return hist
-
-        dims = [180]
-        ranges = [(0,180)]
-        a = cv.CreateHist(dims, cv.CV_HIST_ARRAY , ranges, 1)
-        src = self.get_sample("samples/c/lena.jpg", 0)
-        h = imh(src)
-        (minv, maxv, minl, maxl) = cv.GetMinMaxHistValue(h)
-        self.assert_(cv.QueryHistValue_nD(h, minl) == minv)
-        self.assert_(cv.QueryHistValue_nD(h, maxl) == maxv)
-        bp = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_8U, 1)
-        cv.CalcBackProject(split(src), bp, h)
-        bp = cv.CreateImage((cv.GetSize(src)[0]-2, cv.GetSize(src)[1]-2), cv.IPL_DEPTH_32F, 1)
-        cv.CalcBackProjectPatch(split(src), bp, (3,3), h, cv.CV_COMP_INTERSECT, 1)
-
-        for meth,expected in [(cv.CV_COMP_CORREL, 1.0), (cv.CV_COMP_CHISQR, 0.0), (cv.CV_COMP_INTERSECT, 1.0), (cv.CV_COMP_BHATTACHARYYA, 0.0)]:
-            self.assertEqual(cv.CompareHist(h, h, meth), expected)
-
-    def test_remap(self):
-        rng = cv.RNG(0)
-        maxError = 1e-6
-        raw = cv.CreateImage((640, 480), cv.IPL_DEPTH_8U, 1)
-        for x in range(0, 640, 20):
-            cv.Line(raw, (x,0), (x,480), 255, 1)
-        for y in range(0, 480, 20):
-            cv.Line(raw, (0,y), (640,y), 255, 1)
-        intrinsic_mat = cv.CreateMat(3, 3, cv.CV_32FC1)
-        distortion_coeffs = cv.CreateMat(1, 4, cv.CV_32FC1)
-
-        cv.SetZero(intrinsic_mat)
-        intrinsic_mat[0,2] = 320.0
-        intrinsic_mat[1,2] = 240.0
-        intrinsic_mat[0,0] = 320.0
-        intrinsic_mat[1,1] = 320.0
-        intrinsic_mat[2,2] = 1.0
-        cv.SetZero(distortion_coeffs)
-        distortion_coeffs[0,0] = 1e-1
-        mapx = cv.CreateImage((640, 480), cv.IPL_DEPTH_32F, 1)
-        mapy = cv.CreateImage((640, 480), cv.IPL_DEPTH_32F, 1)
-        cv.SetZero(mapx)
-        cv.SetZero(mapy)
-        cv.InitUndistortMap(intrinsic_mat, distortion_coeffs, mapx, mapy)
-        rect = cv.CreateImage((640, 480), cv.IPL_DEPTH_8U, 1)
-
-        (w,h) = (640,480)
-        rMapxy = cv.CreateMat(h, w, cv.CV_16SC2)
-        rMapa  = cv.CreateMat(h, w, cv.CV_16UC1)
-        cv.ConvertMaps(mapx,mapy,rMapxy,rMapa)
-
-        cv.Remap(raw, rect, mapx, mapy)
-        cv.Remap(raw, rect, rMapxy, rMapa)
-        cv.Undistort2(raw, rect, intrinsic_mat, distortion_coeffs)
-
-        for w in [1, 4, 4095, 4096, 4097, 4100]:
-            p = cv.CreateImage((w,256), 8, 1)
-            up = cv.CreateImage((w,256), 8, 1)
-            cv.Undistort2(p, up, intrinsic_mat, distortion_coeffs)
-
-        fptypes = [cv.CV_32FC1, cv.CV_64FC1]
-        pointsCount = 7
-        for t0 in fptypes:
-            for t1 in fptypes:
-                for t2 in fptypes:
-                    for t3 in fptypes:
-                        rotation_vector = cv.CreateMat(1, 3, t0)
-                        translation_vector = cv.CreateMat(1, 3, t1)
-                        cv.RandArr(rng, rotation_vector, cv.CV_RAND_UNI, -1.0, 1.0)
-                        cv.RandArr(rng, translation_vector, cv.CV_RAND_UNI, -1.0, 1.0)
-                        object_points = cv.CreateMat(pointsCount, 3, t2)
-                        image_points = cv.CreateMat(pointsCount, 2, t3)
-                        cv.RandArr(rng, object_points, cv.CV_RAND_UNI, -100.0, 100.0)
-                        cv.ProjectPoints2(object_points, rotation_vector, translation_vector, intrinsic_mat, distortion_coeffs, image_points)
-
-                        reshaped_object_points = cv.Reshape(object_points, 1, 3)
-                        reshaped_image_points = cv.CreateMat(2, pointsCount, t3)
-                        cv.ProjectPoints2(object_points, rotation_vector, translation_vector, intrinsic_mat, distortion_coeffs, reshaped_image_points)
-
-                        error = cv.Norm(reshaped_image_points, cv.Reshape(image_points, 1, 2))
-                        self.assert_(error < maxError)
-
-    def test_arithmetic(self):
-        a = cv.CreateMat(4, 4, cv.CV_8UC1)
-        a[0,0] = 50.0
-        b = cv.CreateMat(4, 4, cv.CV_8UC1)
-        b[0,0] = 4.0
-        d = cv.CreateMat(4, 4, cv.CV_8UC1)
-        cv.Add(a, b, d)
-        self.assertEqual(d[0,0], 54.0)
-        cv.Mul(a, b, d)
-        self.assertEqual(d[0,0], 200.0)
-
-
-    def failing_test_cvtcolor(self):
-        src3 = self.get_sample("samples/c/lena.jpg")
-        src1 = self.get_sample("samples/c/lena.jpg", 0)
-        dst8u = dict([(c,cv.CreateImage(cv.GetSize(src1), cv.IPL_DEPTH_8U, c)) for c in (1,2,3,4)])
-        dst16u = dict([(c,cv.CreateImage(cv.GetSize(src1), cv.IPL_DEPTH_16U, c)) for c in (1,2,3,4)])
-        dst32f = dict([(c,cv.CreateImage(cv.GetSize(src1), cv.IPL_DEPTH_32F, c)) for c in (1,2,3,4)])
-
-        for srcf in ["BGR", "RGB"]:
-            for dstf in ["Luv"]:
-                cv.CvtColor(src3, dst8u[3], eval("cv.CV_%s2%s" % (srcf, dstf)))
-                cv.CvtColor(src3, dst32f[3], eval("cv.CV_%s2%s" % (srcf, dstf)))
-                cv.CvtColor(src3, dst8u[3], eval("cv.CV_%s2%s" % (dstf, srcf)))
-
-        for srcf in ["BayerBG", "BayerGB", "BayerGR"]:
-            for dstf in ["RGB", "BGR"]:
-                cv.CvtColor(src1, dst8u[3], eval("cv.CV_%s2%s" % (srcf, dstf)))
-
-    def test_voronoi(self):
-        w,h = 500,500
-
-        storage = cv.CreateMemStorage(0)
-
-        def facet_edges(e0):
-            e = e0
-            while True:
-                e = cv.Subdiv2DGetEdge(e, cv.CV_NEXT_AROUND_LEFT)
-                yield e
-                if e == e0:
-                    break
-
-        def areas(edges):
-            seen = []
-            seensorted = []
-            for edge in edges:
-                pts = [ cv.Subdiv2DEdgeOrg(e) for e in facet_edges(edge) ]
-                if not (None in pts):
-                    l = [p.pt for p in pts]
-                    ls = sorted(l)
-                    if not(ls in seensorted):
-                        seen.append(l)
-                        seensorted.append(ls)
-            return seen
-
-        for npoints in range(1, 200):
-            points = [ (random.randrange(w), random.randrange(h)) for i in range(npoints) ]
-            subdiv = cv.CreateSubdivDelaunay2D( (0,0,w,h), storage )
-            for p in points:
-                cv.SubdivDelaunay2DInsert( subdiv, p)
-            cv.CalcSubdivVoronoi2D(subdiv)
-            ars = areas([ cv.Subdiv2DRotateEdge(e, 1) for e in subdiv.edges ] + [ cv.Subdiv2DRotateEdge(e, 3) for e in subdiv.edges ])
-            self.assert_(len(ars) == len(set(points)))
-
-            if False:
-                img = cv.CreateImage((w,h), cv.IPL_DEPTH_8U, 3)
-                cv.SetZero(img)
-                def T(x): return int(x) # int(300+x/16)
-                for pts in ars:
-                    cv.FillConvexPoly( img, [(T(x),T(y)) for (x,y) in pts], cv.RGB(100+random.randrange(156),random.randrange(256),random.randrange(256)), cv.CV_AA, 0 );
-                for x,y in points:
-                    cv.Circle(img, (T(x), T(y)), 3, cv.RGB(0,0,0), -1)
-
-                cv.ShowImage("snap", img)
-                if cv.WaitKey(10) > 0:
-                    break
-
-    def perf_test_pow(self):
-        mt = cv.CreateMat(1000, 1000, cv.CV_32FC1)
-        dst = cv.CreateMat(1000, 1000, cv.CV_32FC1)
-        rng = cv.RNG(0)
-        cv.RandArr(rng, mt, cv.CV_RAND_UNI, 0, 1000.0)
-        mt[0,0] = 10
-        print
-        for a in [0.5, 2.0, 2.3, 2.4, 3.0, 37.1786] + [2.4]*10:
-            started = time.time()
-            for i in range(10):
-                cv.Pow(mt, dst, a)
-            took = (time.time() - started) / 1e7
-            print "%4.1f took %f ns" % (a, took * 1e9)
-        print dst[0,0], 10 ** 2.4
-
-    def test_access_row_col(self):
-        src = cv.CreateImage((8,3), 8, 1)
-        # Put these words
-        #     Achilles
-        #     Benedict
-        #     Congreve
-        # in an array (3 rows, 8 columns).
-        # Then extract the array in various ways.
-
-        for r,w in enumerate(("Achilles", "Benedict", "Congreve")):
-            for c,v in enumerate(w):
-                src[r,c] = ord(v)
-        self.assertEqual(src.tostring(), "AchillesBenedictCongreve")
-        self.assertEqual(src[:,:].tostring(), "AchillesBenedictCongreve")
-        self.assertEqual(src[:,:4].tostring(), "AchiBeneCong")
-        self.assertEqual(src[:,0].tostring(), "ABC")
-        self.assertEqual(src[:,4:].tostring(), "llesdictreve")
-        self.assertEqual(src[::2,:].tostring(), "AchillesCongreve")
-        self.assertEqual(src[1:,:].tostring(), "BenedictCongreve")
-        self.assertEqual(src[1:2,:].tostring(), "Benedict")
-        self.assertEqual(src[::2,:4].tostring(), "AchiCong")
-        # The mats share the same storage, so updating one should update them all
-        lastword = src[2]
-        self.assertEqual(lastword.tostring(), "Congreve")
-        src[2,0] = ord('K')
-        self.assertEqual(lastword.tostring(), "Kongreve")
-        src[2,0] = ord('C')
-
-        # ABCD
-        # EFGH
-        # IJKL
-        #
-        # MNOP
-        # QRST
-        # UVWX
-
-        mt = cv.CreateMatND([2,3,4], cv.CV_8UC1)
-        for i in range(2):
-            for j in range(3):
-                for k in range(4):
-                    mt[i,j,k] = ord('A') + k + 4 * (j + 3 * i)
-        self.assertEqual(mt[:,:,:1].tostring(), "AEIMQU")
-        self.assertEqual(mt[:,:1,:].tostring(), "ABCDMNOP")
-        self.assertEqual(mt[:1,:,:].tostring(), "ABCDEFGHIJKL")
-        self.assertEqual(mt[1,1].tostring(), "QRST")
-        self.assertEqual(mt[:,::2,:].tostring(), "ABCDIJKLMNOPUVWX")
-
-        # Exercise explicit GetRows
-        self.assertEqual(cv.GetRows(src, 0, 3).tostring(), "AchillesBenedictCongreve")
-        self.assertEqual(cv.GetRows(src, 0, 3, 1).tostring(), "AchillesBenedictCongreve")
-        self.assertEqual(cv.GetRows(src, 0, 3, 2).tostring(), "AchillesCongreve")
-
-        self.assertEqual(cv.GetRow(src, 0).tostring(), "Achilles")
-
-        self.assertEqual(cv.GetCols(src, 0, 4).tostring(), "AchiBeneCong")
-
-        self.assertEqual(cv.GetCol(src, 0).tostring(), "ABC")
-        self.assertEqual(cv.GetCol(src, 1).tostring(), "ceo")
-
-        self.assertEqual(cv.GetDiag(src, 0).tostring(), "Aen")
-
-        # Check that matrix type is preserved by the various operators
-
-        for mt in self.mat_types:
-            m = cv.CreateMat(5, 3, mt)
-            self.assertEqual(mt, cv.GetElemType(cv.GetRows(m, 0, 2)))
-            self.assertEqual(mt, cv.GetElemType(cv.GetRow(m, 0)))
-            self.assertEqual(mt, cv.GetElemType(cv.GetCols(m, 0, 2)))
-            self.assertEqual(mt, cv.GetElemType(cv.GetCol(m, 0)))
-            self.assertEqual(mt, cv.GetElemType(cv.GetDiag(m, 0)))
-            self.assertEqual(mt, cv.GetElemType(m[0]))
-            self.assertEqual(mt, cv.GetElemType(m[::2]))
-            self.assertEqual(mt, cv.GetElemType(m[:,0]))
-            self.assertEqual(mt, cv.GetElemType(m[:,:]))
-            self.assertEqual(mt, cv.GetElemType(m[::2,:]))
-
-    def test_addS_3D(self):
-        for dim in [ [1,1,4], [2,2,3], [7,4,3] ]:
-            for ty,ac in [ (cv.CV_32FC1, 'f'), (cv.CV_64FC1, 'd')]:
-                mat = cv.CreateMatND(dim, ty)
-                mat2 = cv.CreateMatND(dim, ty)
-                for increment in [ 0, 3, -1 ]:
-                    cv.SetData(mat, array.array(ac, range(dim[0] * dim[1] * dim[2])), 0)
-                    cv.AddS(mat, increment, mat2)
-                    for i in range(dim[0]):
-                        for j in range(dim[1]):
-                            for k in range(dim[2]):
-                                self.assert_(mat2[i,j,k] == mat[i,j,k] + increment)
-
-    def test_buffers(self):
-        ar = array.array('f', [7] * (360*640))
-
-        m = cv.CreateMat(360, 640, cv.CV_32FC1)
-        cv.SetData(m, ar, 4 * 640)
-        self.assert_(m[0,0] == 7.0)
-
-        m = cv.CreateMatND((360, 640), cv.CV_32FC1)
-        cv.SetData(m, ar, 4 * 640)
-        self.assert_(m[0,0] == 7.0)
-
-        m = cv.CreateImage((640, 360), cv.IPL_DEPTH_32F, 1)
-        cv.SetData(m, ar, 4 * 640)
-        self.assert_(m[0,0] == 7.0)
-
-    def xxtest_Filters(self):
-        print
-        m = cv.CreateMat(360, 640, cv.CV_32FC1)
-        d = cv.CreateMat(360, 640, cv.CV_32FC1)
-        for k in range(3, 21, 2):
-            started = time.time()
-            for i in range(1000):
-                cv.Smooth(m, m, param1=k)
-            print k, "took", time.time() - started
-
-    def assertSame(self, a, b):
-        w,h = cv.GetSize(a)
-        d = cv.CreateMat(h, w, cv.CV_8UC1)
-        cv.AbsDiff(a, b, d)
-        self.assert_(cv.CountNonZero(d) == 0)
-
-    def test_text(self):
-        img = cv.CreateImage((640,40), cv.IPL_DEPTH_8U, 1)
-        cv.SetZero(img)
-        font = cv.InitFont(cv.CV_FONT_HERSHEY_SIMPLEX, 1, 1)
-        message = "XgfooX"
-        cv.PutText(img, message, (320,30), font, 255)
-        ((w,h),bl) = cv.GetTextSize(message, font)
-
-        # Find nonzero in X and Y
-        Xs = []
-        for x in range(640):
-            cv.SetImageROI(img, (x, 0, 1, 40))
-            Xs.append(cv.Sum(img)[0] > 0)
-        def firstlast(l):
-            return (l.index(True), len(l) - list(reversed(l)).index(True))
-
-        Ys = []
-        for y in range(40):
-            cv.SetImageROI(img, (0, y, 640, 1))
-            Ys.append(cv.Sum(img)[0] > 0)
-
-        x0,x1 = firstlast(Xs)
-        y0,y1 = firstlast(Ys)
-        actual_width = x1 - x0
-        actual_height = y1 - y0
-
-        # actual_width can be up to 8 pixels smaller than GetTextSize says
-        self.assert_(actual_width <= w)
-        self.assert_((w - actual_width) <= 8)
-
-        # actual_height can be up to 4 pixels smaller than GetTextSize says
-        self.assert_(actual_height <= (h + bl))
-        self.assert_(((h + bl) - actual_height) <= 4)
-
-        cv.ResetImageROI(img)
-        self.assert_(w != 0)
-        self.assert_(h != 0)
-
-    def test_sizes(self):
-        sizes = [ 1, 2, 3, 97, 255, 256, 257, 947 ]
-        for w in sizes:
-            for h in sizes:
-                # Create an IplImage
-                im = cv.CreateImage((w,h), cv.IPL_DEPTH_8U, 1)
-                cv.Set(im, 1)
-                self.assert_(cv.Sum(im)[0] == (w * h))
-                del im
-                # Create a CvMat
-                mt = cv.CreateMat(h, w, cv.CV_8UC1)
-                cv.Set(mt, 1)
-                self.assert_(cv.Sum(mt)[0] == (w * h))
-
-        random.seed(7)
-        for dim in range(1, cv.CV_MAX_DIM + 1):
-            for attempt in range(10):
-                dims = [ random.choice([1,1,1,1,2,3]) for i in range(dim) ]
-                mt = cv.CreateMatND(dims, cv.CV_8UC1)
-                cv.SetZero(mt)
-                self.assert_(cv.Sum(mt)[0] == 0)
-                # Set to all-ones, verify the sum
-                cv.Set(mt, 1)
-                expected = 1
-                for d in dims:
-                    expected *= d
-                self.assert_(cv.Sum(mt)[0] == expected)
-
-    def test_random(self):
-        seeds = [ 0, 1, 2**48, 2**48 + 1 ]
-        sequences = set()
-        for s in seeds:
-            rng = cv.RNG(s)
-            sequences.add(str([cv.RandInt(rng) for i in range(10)]))
-        self.assert_(len(seeds) == len(sequences))
-
-        rng = cv.RNG(0)
-        im = cv.CreateImage((1024,1024), cv.IPL_DEPTH_8U, 1)
-        cv.RandArr(rng, im, cv.CV_RAND_UNI, 0, 256)
-        cv.RandArr(rng, im, cv.CV_RAND_NORMAL, 128, 30)
-        if 1:
-            hist = cv.CreateHist([ 256 ], cv.CV_HIST_ARRAY, [ (0,255) ], 1)
-            cv.CalcHist([im], hist)
-
-        rng = cv.RNG()
-        for i in range(1000):
-            v = cv.RandReal(rng)
-            self.assert_(0 <= v)
-            self.assert_(v < 1)
-
-        for mode in [ cv.CV_RAND_UNI, cv.CV_RAND_NORMAL ]:
-            for fmt in self.mat_types:
-                mat = cv.CreateMat(64, 64, fmt)
-                cv.RandArr(cv.RNG(), mat, mode, (0,0,0,0), (1,1,1,1))
-
-    def test_MixChannels(self):
-
-        # First part - test the single case described in the documentation
-        rgba = cv.CreateMat(100, 100, cv.CV_8UC4)
-        bgr = cv.CreateMat(100, 100, cv.CV_8UC3)
-        alpha = cv.CreateMat(100, 100, cv.CV_8UC1)
-        cv.Set(rgba, (1,2,3,4))
-        cv.MixChannels([rgba], [bgr, alpha], [
-           (0, 2),    # rgba[0] -> bgr[2]
-           (1, 1),    # rgba[1] -> bgr[1]
-           (2, 0),    # rgba[2] -> bgr[0]
-           (3, 3)     # rgba[3] -> alpha[0]
-        ])
-        self.assert_(bgr[0,0] == (3,2,1))
-        self.assert_(alpha[0,0] == 4)
-
-        # Second part.  Choose random sets of sources and destinations,
-        # fill them with known values, choose random channel assignments,
-        # run cvMixChannels and check that the result is as expected.
-
-        random.seed(1)
-
-        for rows in [1,2,4,13,64,1000]:
-            for cols in [1,2,4,13,64,1000]:
-                for loop in range(5):
-                    sources = [random.choice([1, 2, 3, 4]) for i in range(8)]
-                    dests = [random.choice([1, 2, 3, 4]) for i in range(8)]
-                    # make sure that fromTo does not have duplicates in dests, otherwise the result is not determined
-                    while 1:
-                        fromTo = [(random.randrange(-1, sum(sources)), random.randrange(sum(dests))) for i in range(random.randrange(1, 30))]
-                        dests_set = list(set([j for (i, j) in fromTo]))
-                        if len(dests_set) == len(dests):
-                            break
-
-                    # print sources
-                    # print dests
-                    # print fromTo
-
-                    def CV_8UC(n):
-                        return [cv.CV_8UC1, cv.CV_8UC2, cv.CV_8UC3, cv.CV_8UC4][n-1]
-                    source_m = [cv.CreateMat(rows, cols, CV_8UC(c)) for c in sources]
-                    dest_m =   [cv.CreateMat(rows, cols, CV_8UC(c)) for c in dests]
-
-                    def m00(m):
-                        # return the contents of the N channel mat m[0,0] as a N-length list
-                        chans = cv.CV_MAT_CN(cv.GetElemType(m))
-                        if chans == 1:
-                            return [m[0,0]]
-                        else:
-                            return list(m[0,0])[:chans]
-
-                    # Sources numbered from 50, destinations numbered from 100
-
-                    for i in range(len(sources)):
-                        s = sum(sources[:i]) + 50
-                        cv.Set(source_m[i], (s, s+1, s+2, s+3))
-                        self.assertEqual(m00(source_m[i]), [s, s+1, s+2, s+3][:sources[i]])
-
-                    for i in range(len(dests)):
-                        s = sum(dests[:i]) + 100
-                        cv.Set(dest_m[i], (s, s+1, s+2, s+3))
-                        self.assertEqual(m00(dest_m[i]), [s, s+1, s+2, s+3][:dests[i]])
-
-                    # now run the sanity check
-
-                    for i in range(len(sources)):
-                        s = sum(sources[:i]) + 50
-                        self.assertEqual(m00(source_m[i]), [s, s+1, s+2, s+3][:sources[i]])
-
-                    for i in range(len(dests)):
-                        s = sum(dests[:i]) + 100
-                        self.assertEqual(m00(dest_m[i]), [s, s+1, s+2, s+3][:dests[i]])
-
-                    cv.MixChannels(source_m, dest_m, fromTo)
-
-                    expected = range(100, 100 + sum(dests))
-                    for (i, j) in fromTo:
-                        if i == -1:
-                            expected[j] = 0.0
-                        else:
-                            expected[j] = 50 + i
-
-                    actual = sum([m00(m) for m in dest_m], [])
-                    self.assertEqual(sum([m00(m) for m in dest_m], []), expected)
-
-    def test_allocs(self):
-        mats = [ 0 for i in range(20) ]
-        for i in range(1000):
-            m = cv.CreateMat(random.randrange(10, 512), random.randrange(10, 512), cv.CV_8UC1)
-            j = random.randrange(len(mats))
-            mats[j] = m
-            cv.SetZero(m)
-
-    def test_access(self):
-        cnames = { 1:cv.CV_32FC1, 2:cv.CV_32FC2, 3:cv.CV_32FC3, 4:cv.CV_32FC4 }
-
-        for w in range(1,11):
-            for h in range(2,11):
-                for c in [1,2]:
-                    for o in [ cv.CreateMat(h, w, cnames[c]), cv.CreateImage((w,h), cv.IPL_DEPTH_32F, c) ][1:]:
-                        pattern = [ (i,j) for i in range(w) for j in range(h) ]
-                        random.shuffle(pattern)
-                        for k,(i,j) in enumerate(pattern):
-                            if c == 1:
-                                o[j,i] = k
-                            else:
-                                o[j,i] = (k,) * c
-                        for k,(i,j) in enumerate(pattern):
-                            if c == 1:
-                                self.assert_(o[j,i] == k)
-                            else:
-                                self.assert_(o[j,i] == (k,)*c)
-
-        test_mat = cv.CreateMat(2, 3, cv.CV_32FC1)
-        cv.SetData(test_mat, array.array('f', range(6)), 12)
-        self.assertEqual(cv.GetDims(test_mat[0]), (1, 3))
-        self.assertEqual(cv.GetDims(test_mat[1]), (1, 3))
-        self.assertEqual(cv.GetDims(test_mat[0:1]), (1, 3))
-        self.assertEqual(cv.GetDims(test_mat[1:2]), (1, 3))
-        self.assertEqual(cv.GetDims(test_mat[-1:]), (1, 3))
-        self.assertEqual(cv.GetDims(test_mat[-1]), (1, 3))
-
-    def xxxtest_corners(self):
-        a = cv.LoadImage("foo-mono.png", 0)
-        cv.AdaptiveThreshold(a, a, 255, param1=5)
-        scribble = cv.CreateImage(cv.GetSize(a), 8, 3)
-        cv.CvtColor(a, scribble, cv.CV_GRAY2BGR)
-        if 0:
-            eig_image = cv.CreateImage(cv.GetSize(a), cv.IPL_DEPTH_32F, 1)
-            temp_image = cv.CreateImage(cv.GetSize(a), cv.IPL_DEPTH_32F, 1)
-            pts = cv.GoodFeaturesToTrack(a, eig_image, temp_image, 100, 0.04, 2, use_harris=1)
-            for p in pts:
-                cv.Circle( scribble, p, 1, cv.RGB(255,0,0), -1 )
-            self.snap(scribble)
-        canny = cv.CreateImage(cv.GetSize(a), 8, 1)
-        cv.SubRS(a, 255, canny)
-        self.snap(canny)
-        li = cv.HoughLines2(canny,
-                                                cv.CreateMemStorage(),
-                                                cv.CV_HOUGH_STANDARD,
-                                                1,
-                                                math.pi/180,
-                                                60,
-                                                0,
-                                                0)
-        for (rho,theta) in li:
-            print rho,theta
-            c = math.cos(theta)
-            s = math.sin(theta)
-            x0 = c*rho
-            y0 = s*rho
-            cv.Line(scribble,
-                            (x0 + 1000*(-s), y0 + 1000*c),
-                            (x0 + -1000*(-s), y0 - 1000*c),
-                            (0,255,0))
-        self.snap(scribble)
-
-    def test_calibration(self):
-
-        def get_corners(mono, refine = False):
-            (ok, corners) = cv.FindChessboardCorners(mono, (num_x_ints, num_y_ints), cv.CV_CALIB_CB_ADAPTIVE_THRESH | cv.CV_CALIB_CB_NORMALIZE_IMAGE)
-            if refine and ok:
-                corners = cv.FindCornerSubPix(mono, corners, (5,5), (-1,-1), ( cv.CV_TERMCRIT_EPS+cv.CV_TERMCRIT_ITER, 30, 0.1 ))
-            return (ok, corners)
-
-        def mk_object_points(nimages, squaresize = 1):
-            opts = cv.CreateMat(nimages * num_pts, 3, cv.CV_32FC1)
-            for i in range(nimages):
-                for j in range(num_pts):
-                    opts[i * num_pts + j, 0] = (j / num_x_ints) * squaresize
-                    opts[i * num_pts + j, 1] = (j % num_x_ints) * squaresize
-                    opts[i * num_pts + j, 2] = 0
-            return opts
-
-        def mk_image_points(goodcorners):
-            ipts = cv.CreateMat(len(goodcorners) * num_pts, 2, cv.CV_32FC1)
-            for (i, co) in enumerate(goodcorners):
-                for j in range(num_pts):
-                    ipts[i * num_pts + j, 0] = co[j][0]
-                    ipts[i * num_pts + j, 1] = co[j][1]
-            return ipts
-
-        def mk_point_counts(nimages):
-            npts = cv.CreateMat(nimages, 1, cv.CV_32SC1)
-            for i in range(nimages):
-                npts[i, 0] = num_pts
-            return npts
-
-        def cvmat_iterator(cvmat):
-            for i in range(cvmat.rows):
-                for j in range(cvmat.cols):
-                    yield cvmat[i,j]
-
-        def image_from_archive(tar, name):
-            member = tar.getmember(name)
-            filedata = tar.extractfile(member).read()
-            imagefiledata = cv.CreateMat(1, len(filedata), cv.CV_8UC1)
-            cv.SetData(imagefiledata, filedata, len(filedata))
-            return cv.DecodeImageM(imagefiledata)
-
-        urllib.urlretrieve("http://opencv.itseez.com/data/camera_calibration.tar.gz", "camera_calibration.tar.gz")
-        tf = tarfile.open("camera_calibration.tar.gz")
-
-        num_x_ints = 8
-        num_y_ints = 6
-        num_pts = num_x_ints * num_y_ints
-
-        leftimages = [image_from_archive(tf, "wide/left%04d.pgm" % i) for i in range(3, 15)]
-        size = cv.GetSize(leftimages[0])
-
-        # Monocular test
-
-        if True:
-            corners = [get_corners(i) for i in leftimages]
-            goodcorners = [co for (im, (ok, co)) in zip(leftimages, corners) if ok]
-
-            ipts = mk_image_points(goodcorners)
-            opts = mk_object_points(len(goodcorners), .1)
-            npts = mk_point_counts(len(goodcorners))
-
-            intrinsics = cv.CreateMat(3, 3, cv.CV_64FC1)
-            distortion = cv.CreateMat(4, 1, cv.CV_64FC1)
-            cv.SetZero(intrinsics)
-            cv.SetZero(distortion)
-            # focal lengths have 1/1 ratio
-            intrinsics[0,0] = 1.0
-            intrinsics[1,1] = 1.0
-            cv.CalibrateCamera2(opts, ipts, npts,
-                       cv.GetSize(leftimages[0]),
-                       intrinsics,
-                       distortion,
-                       cv.CreateMat(len(goodcorners), 3, cv.CV_32FC1),
-                       cv.CreateMat(len(goodcorners), 3, cv.CV_32FC1),
-                       flags = 0) # cv.CV_CALIB_ZERO_TANGENT_DIST)
-            # print "D =", list(cvmat_iterator(distortion))
-            # print "K =", list(cvmat_iterator(intrinsics))
-
-            newK = cv.CreateMat(3, 3, cv.CV_64FC1)
-            cv.GetOptimalNewCameraMatrix(intrinsics, distortion, size, 1.0, newK)
-            # print "newK =", list(cvmat_iterator(newK))
-
-            mapx = cv.CreateImage((640, 480), cv.IPL_DEPTH_32F, 1)
-            mapy = cv.CreateImage((640, 480), cv.IPL_DEPTH_32F, 1)
-            for K in [ intrinsics, newK ]:
-                cv.InitUndistortMap(K, distortion, mapx, mapy)
-                for img in leftimages[:1]:
-                    r = cv.CloneMat(img)
-                    cv.Remap(img, r, mapx, mapy)
-                    # cv.ShowImage("snap", r)
-                    # cv.WaitKey()
-
-        rightimages = [image_from_archive(tf, "wide/right%04d.pgm" % i) for i in range(3, 15)]
-
-        # Stereo test
-
-        if True:
-            lcorners = [get_corners(i) for i in leftimages]
-            rcorners = [get_corners(i) for i in rightimages]
-            good = [(lco, rco) for ((lok, lco), (rok, rco)) in zip(lcorners, rcorners) if (lok and rok)]
-
-            lipts = mk_image_points([l for (l, r) in good])
-            ripts = mk_image_points([r for (l, r) in good])
-            opts = mk_object_points(len(good), .108)
-            npts = mk_point_counts(len(good))
-
-            flags = cv.CV_CALIB_FIX_ASPECT_RATIO | cv.CV_CALIB_FIX_INTRINSIC
-            flags = cv.CV_CALIB_SAME_FOCAL_LENGTH + cv.CV_CALIB_FIX_PRINCIPAL_POINT + cv.CV_CALIB_ZERO_TANGENT_DIST
-            flags = 0
-
-            T = cv.CreateMat(3, 1, cv.CV_64FC1)
-            R = cv.CreateMat(3, 3, cv.CV_64FC1)
-            lintrinsics = cv.CreateMat(3, 3, cv.CV_64FC1)
-            ldistortion = cv.CreateMat(4, 1, cv.CV_64FC1)
-            rintrinsics = cv.CreateMat(3, 3, cv.CV_64FC1)
-            rdistortion = cv.CreateMat(4, 1, cv.CV_64FC1)
-            lR = cv.CreateMat(3, 3, cv.CV_64FC1)
-            rR = cv.CreateMat(3, 3, cv.CV_64FC1)
-            lP = cv.CreateMat(3, 4, cv.CV_64FC1)
-            rP = cv.CreateMat(3, 4, cv.CV_64FC1)
-            lmapx = cv.CreateImage(size, cv.IPL_DEPTH_32F, 1)
-            lmapy = cv.CreateImage(size, cv.IPL_DEPTH_32F, 1)
-            rmapx = cv.CreateImage(size, cv.IPL_DEPTH_32F, 1)
-            rmapy = cv.CreateImage(size, cv.IPL_DEPTH_32F, 1)
-
-            cv.SetIdentity(lintrinsics)
-            cv.SetIdentity(rintrinsics)
-            lintrinsics[0,2] = size[0] * 0.5
-            lintrinsics[1,2] = size[1] * 0.5
-            rintrinsics[0,2] = size[0] * 0.5
-            rintrinsics[1,2] = size[1] * 0.5
-            cv.SetZero(ldistortion)
-            cv.SetZero(rdistortion)
-
-            cv.StereoCalibrate(opts, lipts, ripts, npts,
-                               lintrinsics, ldistortion,
-                               rintrinsics, rdistortion,
-                               size,
-                               R,                                  # R
-                               T,                                  # T
-                               cv.CreateMat(3, 3, cv.CV_32FC1),    # E
-                               cv.CreateMat(3, 3, cv.CV_32FC1),    # F
-                               (cv.CV_TERMCRIT_ITER + cv.CV_TERMCRIT_EPS, 30, 1e-5),
-                               flags)
-
-            for a in [-1, 0, 1]:
-                cv.StereoRectify(lintrinsics,
-                                 rintrinsics,
-                                 ldistortion,
-                                 rdistortion,
-                                 size,
-                                 R,
-                                 T,
-                                 lR, rR, lP, rP,
-                                 alpha = a)
-
-                cv.InitUndistortRectifyMap(lintrinsics, ldistortion, lR, lP, lmapx, lmapy)
-                cv.InitUndistortRectifyMap(rintrinsics, rdistortion, rR, rP, rmapx, rmapy)
-
-                for l,r in zip(leftimages, rightimages)[:1]:
-                    l_ = cv.CloneMat(l)
-                    r_ = cv.CloneMat(r)
-                    cv.Remap(l, l_, lmapx, lmapy)
-                    cv.Remap(r, r_, rmapx, rmapy)
-                    # cv.ShowImage("snap", l_)
-                    # cv.WaitKey()
-
-
-    def xxx_test_Disparity(self):
-        print
-        for t in ["8U", "8S", "16U", "16S", "32S", "32F", "64F" ]:
-          for c in [1,2,3,4]:
-            nm = "%sC%d" % (t, c)
-            print "int32 CV_%s=%d" % (nm, eval("cv.CV_%s" % nm))
-        return
-        integral = cv.CreateImage((641,481), cv.IPL_DEPTH_32S, 1)
-        L = cv.LoadImage("f0-left.png", 0)
-        R = cv.LoadImage("f0-right.png", 0)
-        d = cv.CreateImage(cv.GetSize(L), cv.IPL_DEPTH_8U, 1)
-        Rn = cv.CreateImage(cv.GetSize(L), cv.IPL_DEPTH_8U, 1)
-        started = time.time()
-        for i in range(100):
-            cv.AbsDiff(L, R, d)
-            cv.Integral(d, integral)
-            cv.SetImageROI(R, (1, 1, 639, 479))
-            cv.SetImageROI(Rn, (0, 0, 639, 479))
-            cv.Copy(R, Rn)
-            R = Rn
-            cv.ResetImageROI(R)
-        print 1e3 * (time.time() - started) / 100, "ms"
-        # self.snap(d)
-
-    def local_test_lk(self):
-        seq = [cv.LoadImage("track/%06d.png" % i, 0) for i in range(40)]
-        crit = (cv.CV_TERMCRIT_ITER, 100, 0.1)
-        crit = (cv.CV_TERMCRIT_EPS, 0, 0.001)
-
-        for i in range(1,40):
-            r = cv.CalcOpticalFlowPyrLK(seq[0], seq[i], None, None, [(32,32)], (7,7), 0, crit, 0)
-            pos = r[0][0]
-            #print pos, r[2]
-
-            a = cv.CreateImage((1024,1024), 8, 1)
-            b = cv.CreateImage((1024,1024), 8, 1)
-            cv.Resize(seq[0], a, cv.CV_INTER_NN)
-            cv.Resize(seq[i], b, cv.CV_INTER_NN)
-            cv.Line(a, (0, 512), (1024, 512), 255)
-            cv.Line(a, (512,0), (512,1024), 255)
-            x,y = [int(c) for c in pos]
-            cv.Line(b, (0, y*16), (1024, y*16), 255)
-            cv.Line(b, (x*16,0), (x*16,1024), 255)
-            #self.snapL([a,b])
-
-
-
-    def local_test_Haar(self):
-        import os
-        hcfile = os.environ['OPENCV_ROOT'] + '/share/opencv/haarcascades/haarcascade_frontalface_default.xml'
-        hc = cv.Load(hcfile)
-        img = cv.LoadImage('Stu.jpg', 0)
-        faces = cv.HaarDetectObjects(img, hc, cv.CreateMemStorage())
-        self.assert_(len(faces) > 0)
-        for (x,y,w,h),n in faces:
-            cv.Rectangle(img, (x,y), (x+w,y+h), 255)
-        #self.snap(img)
-
-    def test_create(self):
-    #""" CvCreateImage, CvCreateMat and the header-only form """
-        for (w,h) in [ (320,400), (640,480), (1024, 768) ]:
-            data = "z" * (w * h)
-
-            im = cv.CreateImage((w,h), 8, 1)
-            cv.SetData(im, data, w)
-            im2 = cv.CreateImageHeader((w,h), 8, 1)
-            cv.SetData(im2, data, w)
-            self.assertSame(im, im2)
-
-            m = cv.CreateMat(h, w, cv.CV_8UC1)
-            cv.SetData(m, data, w)
-            m2 = cv.CreateMatHeader(h, w, cv.CV_8UC1)
-            cv.SetData(m2, data, w)
-            self.assertSame(m, m2)
-
-            self.assertSame(im, m)
-            self.assertSame(im2, m2)
-
-
-    def test_casts(self):
-        im = cv.GetImage(self.get_sample("samples/c/lena.jpg", 0))
-        data = im.tostring()
-        cv.SetData(im, data, cv.GetSize(im)[0])
-
-        start_count = sys.getrefcount(data)
-
-        # Conversions should produce same data
-        self.assertSame(im, cv.GetImage(im))
-        m = cv.GetMat(im)
-        self.assertSame(im, m)
-        self.assertSame(m, cv.GetImage(m))
-        im2 = cv.GetImage(m)
-        self.assertSame(im, im2)
-
-        self.assertEqual(sys.getrefcount(data), start_count + 2)
-        del im2
-        self.assertEqual(sys.getrefcount(data), start_count + 1)
-        del m
-        self.assertEqual(sys.getrefcount(data), start_count)
-        del im
-        self.assertEqual(sys.getrefcount(data), start_count - 1)
-
-    def test_morphological(self):
-        im = cv.CreateImage((128, 128), cv.IPL_DEPTH_8U, 1)
-        cv.Resize(cv.GetImage(self.get_sample("samples/c/lena.jpg", 0)), im)
-        dst = cv.CloneImage(im)
-
-        # Check defaults by asserting that all these operations produce the same image
-        funs = [
-            lambda: cv.Dilate(im, dst),
-            lambda: cv.Dilate(im, dst, None),
-            lambda: cv.Dilate(im, dst, iterations = 1),
-            lambda: cv.Dilate(im, dst, element = None),
-            lambda: cv.Dilate(im, dst, iterations = 1, element = None),
-            lambda: cv.Dilate(im, dst, element = None, iterations = 1),
-        ]
-        src_h = self.hashimg(im)
-        hashes = set()
-        for f in funs:
-            f()
-            hashes.add(self.hashimg(dst))
-            self.assertNotEqual(src_h, self.hashimg(dst))
-        # Source image should be untouched
-        self.assertEqual(self.hashimg(im), src_h)
-        # All results should be same
-        self.assertEqual(len(hashes), 1)
-
-        # self.snap(dst)
-        shapes = [eval("cv.CV_SHAPE_%s" % s) for s in ['RECT', 'CROSS', 'ELLIPSE']]
-        elements = [cv.CreateStructuringElementEx(sz, sz, sz / 2 + 1, sz / 2 + 1, shape) for sz in [3, 4, 7, 20] for shape in shapes]
-        elements += [cv.CreateStructuringElementEx(7, 7, 3, 3, cv.CV_SHAPE_CUSTOM, [1] * 49)]
-        for e in elements:
-            for iter in [1, 2]:
-                cv.Dilate(im, dst, e, iter)
-                cv.Erode(im, dst, e, iter)
-                temp = cv.CloneImage(im)
-                for op in ["OPEN", "CLOSE", "GRADIENT", "TOPHAT", "BLACKHAT"]:
-                        cv.MorphologyEx(im, dst, temp, e, eval("cv.CV_MOP_%s" % op), iter)
-
-    def test_getmat_nd(self):
-        # 1D CvMatND should yield (N,1) CvMat
-        matnd = cv.CreateMatND([13], cv.CV_8UC1)
-        self.assertEqual(cv.GetDims(cv.GetMat(matnd, allowND = True)), (13, 1))
-
-        # 2D CvMatND should yield 2D CvMat
-        matnd = cv.CreateMatND([11, 12], cv.CV_8UC1)
-        self.assertEqual(cv.GetDims(cv.GetMat(matnd, allowND = True)), (11, 12))
-
-        if 0: # XXX - ticket #149
-            # 3D CvMatND should yield (N,1) CvMat
-            matnd = cv.CreateMatND([7, 8, 9], cv.CV_8UC1)
-            self.assertEqual(cv.GetDims(cv.GetMat(matnd, allowND = True)), (7 * 8 * 9, 1))
-
-    def test_clipline(self):
-        self.assert_(cv.ClipLine((100,100), (-100,0), (500,0)) == ((0,0), (99,0)))
-        self.assert_(cv.ClipLine((100,100), (-100,0), (-200,0)) == None)
-
-    def test_smoke_image_processing(self):
-        src = self.get_sample("samples/c/lena.jpg", cv.CV_LOAD_IMAGE_GRAYSCALE)
-        #dst = cv.CloneImage(src)
-        for aperture_size in [1, 3, 5, 7]:
-          dst_16s = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_16S, 1)
-          dst_32f = cv.CreateImage(cv.GetSize(src), cv.IPL_DEPTH_32F, 1)
-
-          cv.Sobel(src, dst_16s, 1, 1, aperture_size)
-          cv.Laplace(src, dst_16s, aperture_size)
-          cv.PreCornerDetect(src, dst_32f)
-          eigendst = cv.CreateImage((6*cv.GetSize(src)[0], cv.GetSize(src)[1]), cv.IPL_DEPTH_32F, 1)
-          cv.CornerEigenValsAndVecs(src, eigendst, 8, aperture_size)
-          cv.CornerMinEigenVal(src, dst_32f, 8, aperture_size)
-          cv.CornerHarris(src, dst_32f, 8, aperture_size)
-          cv.CornerHarris(src, dst_32f, 8, aperture_size, 0.1)
-
-        #self.snap(dst)
-
-    def test_fitline(self):
-        cv.FitLine([ (1,1), (10,10) ], cv.CV_DIST_L2, 0, 0.01, 0.01)
-        cv.FitLine([ (1,1,1), (10,10,10) ], cv.CV_DIST_L2, 0, 0.01, 0.01)
-        a = self.get_sample("samples/c/lena.jpg", 0)
-        eig_image = cv.CreateImage(cv.GetSize(a), cv.IPL_DEPTH_32F, 1)
-        temp_image = cv.CreateImage(cv.GetSize(a), cv.IPL_DEPTH_32F, 1)
-        pts = cv.GoodFeaturesToTrack(a, eig_image, temp_image, 100, 0.04, 2, useHarris=1)
-        hull = cv.ConvexHull2(pts, cv.CreateMemStorage(), return_points = 1)
-        cv.FitLine(hull, cv.CV_DIST_L2, 0, 0.01, 0.01)
-
-    def test_moments(self):
-        im = self.get_sample("samples/c/lena.jpg", 0)
-        mo = cv.Moments(im)
-        for fld in ["m00", "m10", "m01", "m20", "m11", "m02", "m30", "m21", "m12", "m03", "mu20", "mu11", "mu02", "mu30", "mu21", "mu12", "mu03", "inv_sqrt_m00"]:
-            self.assert_(isinstance(getattr(mo, fld), float))
-            x = getattr(mo, fld)
-            self.assert_(isinstance(x, float))
-
-        orders = []
-        for x_order in range(4):
-          for y_order in range(4 - x_order):
-            orders.append((x_order, y_order))
-
-        # Just a smoke test for these three functions
-        [ cv.GetSpatialMoment(mo, xo, yo) for (xo,yo) in orders ]
-        [ cv.GetCentralMoment(mo, xo, yo) for (xo,yo) in orders ]
-        [ cv.GetNormalizedCentralMoment(mo, xo, yo) for (xo,yo) in orders ]
-
-        # Hu Moments we can do slightly better.  Check that the first
-        # six are invariant wrt image reflection, and that the 7th
-        # is negated.
-
-        hu0 = cv.GetHuMoments(cv.Moments(im))
-        cv.Flip(im, im, 1)
-        hu1 = cv.GetHuMoments(cv.Moments(im))
-        self.assert_(len(hu0) == 7)
-        self.assert_(len(hu1) == 7)
-        for i in range(5):
-          self.assert_(abs(hu0[i] - hu1[i]) < 1e-6)
-        self.assert_(abs(hu0[i] + hu1[i]) < 1e-6)
-
-    def test_encode(self):
-        im = self.get_sample("samples/c/lena.jpg", 1)
-        jpeg = cv.EncodeImage(".jpeg", im)
-
-        # Smoke jpeg compression at various qualities
-        sizes = dict([(qual, cv.EncodeImage(".jpeg", im, [cv.CV_IMWRITE_JPEG_QUALITY, qual]).cols) for qual in range(5, 100, 5)])
-
-        # Check that the default QUALITY is 95
-        self.assertEqual(cv.EncodeImage(".jpeg", im).cols, sizes[95])
-
-        # Check that the 'round-trip' gives an image of the same size
-        round_trip = cv.DecodeImage(cv.EncodeImage(".jpeg", im, [cv.CV_IMWRITE_JPEG_QUALITY, 10]))
-        self.assert_(cv.GetSize(round_trip) == cv.GetSize(im))
-
-    def test_reduce(self):
-        srcmat = cv.CreateMat(2, 3, cv.CV_32FC1)
-        # 0 1 2
-        # 3 4 5
-        srcmat[0,0] = 0
-        srcmat[0,1] = 1
-        srcmat[0,2] = 2
-        srcmat[1,0] = 3
-        srcmat[1,1] = 4
-        srcmat[1,2] = 5
-        def doreduce(siz, rfunc):
-            dst = cv.CreateMat(siz[0], siz[1], cv.CV_32FC1)
-            rfunc(dst)
-            if siz[0] != 1:
-                return [dst[i,0] for i in range(siz[0])]
-            else:
-                return [dst[0,i] for i in range(siz[1])]
-
-        # exercise dim
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst)), [3, 5, 7])
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, -1)), [3, 5, 7])
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, 0)), [3, 5, 7])
-        self.assertEqual(doreduce((2,1), lambda dst: cv.Reduce(srcmat, dst, 1)), [3, 12])
-
-        # exercise op
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, op = cv.CV_REDUCE_SUM)), [3, 5, 7])
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, op = cv.CV_REDUCE_AVG)), [1.5, 2.5, 3.5])
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, op = cv.CV_REDUCE_MAX)), [3, 4, 5])
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, op = cv.CV_REDUCE_MIN)), [0, 1, 2])
-
-        # exercise both dim and op
-        self.assertEqual(doreduce((1,3), lambda dst: cv.Reduce(srcmat, dst, 0, cv.CV_REDUCE_MAX)), [3, 4, 5])
-        self.assertEqual(doreduce((2,1), lambda dst: cv.Reduce(srcmat, dst, 1, cv.CV_REDUCE_MAX)), [2, 5])
-
-    def test_operations(self):
-        class Im:
-
-            def __init__(self, data = None):
-                self.m = cv.CreateMat(1, 32, cv.CV_32FC1)
-                if data:
-                    cv.SetData(self.m, array.array('f', data), 128)
-
-            def __add__(self, other):
-                r = Im()
-                if isinstance(other, Im):
-                    cv.Add(self.m, other.m, r.m)
-                else:
-                    cv.AddS(self.m, (other,), r.m)
-                return r
-
-            def __sub__(self, other):
-                r = Im()
-                if isinstance(other, Im):
-                    cv.Sub(self.m, other.m, r.m)
-                else:
-                    cv.SubS(self.m, (other,), r.m)
-                return r
-
-            def __rsub__(self, other):
-                r = Im()
-                cv.SubRS(self.m, (other,), r.m)
-                return r
-
-            def __mul__(self, other):
-                r = Im()
-                if isinstance(other, Im):
-                    cv.Mul(self.m, other.m, r.m)
-                else:
-                    cv.ConvertScale(self.m, r.m, other)
-                return r
-
-            def __rmul__(self, other):
-                r = Im()
-                cv.ConvertScale(self.m, r.m, other)
-                return r
-
-            def __div__(self, other):
-                r = Im()
-                if isinstance(other, Im):
-                    cv.Div(self.m, other.m, r.m)
-                else:
-                    cv.ConvertScale(self.m, r.m, 1.0 / other)
-                return r
-
-            def __pow__(self, other):
-                r = Im()
-                cv.Pow(self.m, r.m, other)
-                return r
-
-            def __abs__(self):
-                r = Im()
-                cv.Abs(self.m, r.m)
-                return r
-
-            def __getitem__(self, i):
-                return self.m[0,i]
-
-        def verify(op):
-            r = op(a, b)
-            for i in range(32):
-                expected = op(a[i], b[i])
-                self.assertAlmostEqual(expected, r[i], 4)
-
-        a = Im([random.randrange(1, 256) for i in range(32)])
-        b = Im([random.randrange(1, 256) for i in range(32)])
-
-        # simple operations first
-        verify(lambda x, y: x + y)
-        verify(lambda x, y: x + 3)
-        verify(lambda x, y: x + 0)
-        verify(lambda x, y: x + -8)
-
-        verify(lambda x, y: x - y)
-        verify(lambda x, y: x - 1)
-        verify(lambda x, y: 1 - x)
-
-        verify(lambda x, y: abs(x))
-
-        verify(lambda x, y: x * y)
-        verify(lambda x, y: x * 3)
-
-        verify(lambda x, y: x / y)
-        verify(lambda x, y: x / 2)
-
-        for p in [-2, -1, -0.5, -0.1, 0, 0.1, 0.5, 1, 2 ]:
-            verify(lambda x, y: (x ** p) + (y ** p))
-
-        # Combinations...
-        verify(lambda x, y: x - 4 * abs(y))
-        verify(lambda x, y: abs(y) / x)
-
-        # a polynomial
-        verify(lambda x, y: 2 * x + 3 * (y ** 0.5))
-
-    def temp_test(self):
-        cv.temp_test()
-
-    def failing_test_rand_GetStarKeypoints(self):
-        # GetStarKeypoints [<cvmat(type=4242400d rows=64 cols=64 step=512 )>, <cv.cvmemstorage object at 0xb7cc40d0>, (45, 0.73705234376883488, 0.64282591451367344, 0.1567738743689836, 3)]
-        print cv.CV_MAT_CN(0x4242400d)
-        mat = cv.CreateMat( 64, 64, cv.CV_32FC2)
-        cv.GetStarKeypoints(mat, cv.CreateMemStorage(), (45, 0.73705234376883488, 0.64282591451367344, 0.1567738743689836, 3))
-        print mat
-
-    def test_rand_PutText(self):
-    #""" Test for bug 2829336 """
-        mat = cv.CreateMat( 64, 64, cv.CV_8UC1)
-        font = cv.InitFont(cv.CV_FONT_HERSHEY_SIMPLEX, 1, 1)
-        cv.PutText(mat, chr(127), (20, 20), font, 255)
-
-    def failing_test_rand_FindNearestPoint2D(self):
-        subdiv = cv.CreateSubdivDelaunay2D((0,0,100,100), cv.CreateMemStorage())
-        cv.SubdivDelaunay2DInsert( subdiv, (50, 50))
-        cv.CalcSubdivVoronoi2D(subdiv)
-        print
-        for e in subdiv.edges:
-            print e,
-            print "  ", cv.Subdiv2DEdgeOrg(e)
-            print "  ", cv.Subdiv2DEdgeOrg(cv.Subdiv2DRotateEdge(e, 1)), cv.Subdiv2DEdgeDst(cv.Subdiv2DRotateEdge(e, 1))
-        print "nearest", cv.FindNearestPoint2D(subdiv, (1.0, 1.0))
-
-class DocumentFragmentTests(OpenCVTests):
-    """ Test the fragments of code that are included in the documentation """
-    def setUp(self):
-        OpenCVTests.setUp(self)
-        sys.path.append(".")
-
-    def test_precornerdetect(self):
-        from precornerdetect import precornerdetect
-        im = self.get_sample("samples/cpp/right01.jpg", 0)
-        imf = cv.CreateMat(im.rows, im.cols, cv.CV_32FC1)
-        cv.ConvertScale(im, imf)
-        (r0,r1) = precornerdetect(imf)
-        for r in (r0, r1):
-            self.assertEqual(im.cols, r.cols)
-            self.assertEqual(im.rows, r.rows)
-
-    def test_findstereocorrespondence(self):
-        from findstereocorrespondence import findstereocorrespondence
-        (l,r) = [self.get_sample("samples/cpp/tsukuba_%s.png" % c, cv.CV_LOAD_IMAGE_GRAYSCALE) for c in "lr"]
-
-        (disparity_left, disparity_right) = findstereocorrespondence(l, r)
-
-        disparity_left_visual = cv.CreateMat(l.rows, l.cols, cv.CV_8U)
-        cv.ConvertScale(disparity_left, disparity_left_visual, -16)
-        # self.snap(disparity_left_visual)
-
-    def test_calchist(self):
-        from calchist import hs_histogram
-        i1 = self.get_sample("samples/c/lena.jpg")
-        i2 = self.get_sample("samples/cpp/building.jpg")
-        i3 = cv.CloneMat(i1)
-        cv.Flip(i3, i3, 1)
-        h1 = hs_histogram(i1)
-        h2 = hs_histogram(i2)
-        h3 = hs_histogram(i3)
-        self.assertEqual(self.hashimg(h1), self.hashimg(h3))
-        self.assertNotEqual(self.hashimg(h1), self.hashimg(h2))
+class Hackathon244Tests(NewOpenCVTests):
+
+    def test_int_array(self):
+        a = np.array([-1, 2, -3, 4, -5])
+        absa0 = np.abs(a)
+        self.assert_(cv2.norm(a, cv2.NORM_L1) == 15)
+        absa1 = cv2.absdiff(a, 0)
+        self.assertEqual(cv2.norm(absa1, absa0, cv2.NORM_INF), 0)
+
+    def test_imencode(self):
+        a = np.zeros((480, 640), dtype=np.uint8)
+        flag, ajpg = cv2.imencode("img_q90.jpg", a, [cv2.IMWRITE_JPEG_QUALITY, 90])
+        self.assertEqual(flag, True)
+        self.assertEqual(ajpg.dtype, np.uint8)
+        self.assertGreater(ajpg.shape[0], 1)
+        self.assertEqual(ajpg.shape[1], 1)
+
+    def test_projectPoints(self):
+        objpt = np.float64([[1,2,3]])
+        imgpt0, jac0 = cv2.projectPoints(objpt, np.zeros(3), np.zeros(3), np.eye(3), np.float64([]))
+        imgpt1, jac1 = cv2.projectPoints(objpt, np.zeros(3), np.zeros(3), np.eye(3), None)
+        self.assertEqual(imgpt0.shape, (objpt.shape[0], 1, 2))
+        self.assertEqual(imgpt1.shape, imgpt0.shape)
+        self.assertEqual(jac0.shape, jac1.shape)
+        self.assertEqual(jac0.shape[0], 2*objpt.shape[0])
+
+    def test_estimateAffine3D(self):
+        pattern_size = (11, 8)
+        pattern_points = np.zeros((np.prod(pattern_size), 3), np.float32)
+        pattern_points[:,:2] = np.indices(pattern_size).T.reshape(-1, 2)
+        pattern_points *= 10
+        (retval, out, inliers) = cv2.estimateAffine3D(pattern_points, pattern_points)
+        self.assertEqual(retval, 1)
+        if cv2.norm(out[2,:]) < 1e-3:
+            out[2,2]=1
+        self.assertLess(cv2.norm(out, np.float64([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])), 1e-3)
+        self.assertEqual(cv2.countNonZero(inliers), pattern_size[0]*pattern_size[1])
+
+    def test_fast(self):
+        fd = cv2.FastFeatureDetector(30, True)
+        img = self.get_sample("samples/cpp/right02.jpg", 0)
+        img = cv2.medianBlur(img, 3)
+        imgc = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+        keypoints = fd.detect(img)
+        self.assert_(600 <= len(keypoints) <= 700)
+        for kpt in keypoints:
+            self.assertNotEqual(kpt.response, 0)
+
+    def check_close_angles(self, a, b, angle_delta):
+        self.assert_(abs(a - b) <= angle_delta or
+                     abs(360 - abs(a - b)) <= angle_delta)
+
+    def check_close_pairs(self, a, b, delta):
+        self.assertLessEqual(abs(a[0] - b[0]), delta)
+        self.assertLessEqual(abs(a[1] - b[1]), delta)
+
+    def check_close_boxes(self, a, b, delta, angle_delta):
+        self.check_close_pairs(a[0], b[0], delta)
+        self.check_close_pairs(a[1], b[1], delta)
+        self.check_close_angles(a[2], b[2], angle_delta)
+
+    def test_geometry(self):
+        npt = 100
+        np.random.seed(244)
+        a = np.random.randn(npt,2).astype('float32')*50 + 150
+
+        img = np.zeros((300, 300, 3), dtype='uint8')
+        be = cv2.fitEllipse(a)
+        br = cv2.minAreaRect(a)
+        mc, mr = cv2.minEnclosingCircle(a)
+
+        be0 = ((150.2511749267578, 150.77322387695312), (158.024658203125, 197.57696533203125), 37.57804489135742)
+        br0 = ((161.2974090576172, 154.41793823242188), (199.2301483154297, 207.7177734375), -9.164555549621582)
+        mc0, mr0 = (160.41790771484375, 144.55152893066406), 136.713500977
+
+        self.check_close_boxes(be, be0, 5, 15)
+        self.check_close_boxes(br, br0, 5, 15)
+        self.check_close_pairs(mc, mc0, 5)
+        self.assertLessEqual(abs(mr - mr0), 5)
 
 if __name__ == '__main__':
-    print "testing", cv.__version__
+    print "Testing OpenCV", cv2.__version__
     random.seed(0)
     unittest.main()
-#    optlist, args = getopt.getopt(sys.argv[1:], 'l:rd')
-#    loops = 1
-#    shuffle = 0
-#    doc_frags = False
-#    for o,a in optlist:
-#        if o == '-l':
-#            loops = int(a)
-#        if o == '-r':
-#            shuffle = 1
-#        if o == '-d':
-#            doc_frags = True
-#
-#    cases = [PreliminaryTests, FunctionTests, AreaTests]
-#    if doc_frags:
-#        cases += [DocumentFragmentTests]
-#    everything = [(tc, t) for tc in cases for t in unittest.TestLoader().getTestCaseNames(tc) ]
-#    if len(args) == 0:
-#        # cases = [NewTests]
-#        args = everything
-#    else:
-#        args = [(tc, t) for (tc, t) in everything if t in args]
-#
-#    suite = unittest.TestSuite()
-#    for l in range(loops):
-#        if shuffle:
-#            random.shuffle(args)
-#        for tc,t in args:
-#            suite.addTest(tc(t))
-#	    unittest.TextTestRunner(verbosity=2).run(suite)
diff --git a/modules/python/test/test2.py b/modules/python/test/test2.py
deleted file mode 100644
index a331d69277..0000000000
--- a/modules/python/test/test2.py
+++ /dev/null
@@ -1,120 +0,0 @@
-#/usr/bin/env python
-
-import unittest
-import random
-import time
-import math
-import sys
-import array
-import urllib
-import tarfile
-import hashlib
-import os
-import getopt
-import operator
-import functools
-import numpy as np
-import cv2
-import cv2.cv as cv
-
-class NewOpenCVTests(unittest.TestCase):
-
-    def get_sample(self, filename, iscolor = cv.CV_LOAD_IMAGE_COLOR):
-        if not filename in self.image_cache:
-            filedata = urllib.urlopen("https://raw.github.com/Itseez/opencv/master/" + filename).read()
-            self.image_cache[filename] = cv2.imdecode(np.fromstring(filedata, dtype=np.uint8), iscolor)
-        return self.image_cache[filename]
-
-    def setUp(self):
-        self.image_cache = {}
-
-    def hashimg(self, im):
-        """ Compute a hash for an image, useful for image comparisons """
-        return hashlib.md5(im.tostring()).digest()
-
-# Tests to run first; check the handful of basic operations that the later tests rely on
-
-class Hackathon244Tests(NewOpenCVTests):
-
-    def test_int_array(self):
-        a = np.array([-1, 2, -3, 4, -5])
-        absa0 = np.abs(a)
-        self.assert_(cv2.norm(a, cv2.NORM_L1) == 15)
-        absa1 = cv2.absdiff(a, 0)
-        self.assertEqual(cv2.norm(absa1, absa0, cv2.NORM_INF), 0)
-
-    def test_imencode(self):
-        a = np.zeros((480, 640), dtype=np.uint8)
-        flag, ajpg = cv2.imencode("img_q90.jpg", a, [cv2.IMWRITE_JPEG_QUALITY, 90])
-        self.assertEqual(flag, True)
-        self.assertEqual(ajpg.dtype, np.uint8)
-        self.assertGreater(ajpg.shape[0], 1)
-        self.assertEqual(ajpg.shape[1], 1)
-
-    def test_projectPoints(self):
-        objpt = np.float64([[1,2,3]])
-        imgpt0, jac0 = cv2.projectPoints(objpt, np.zeros(3), np.zeros(3), np.eye(3), np.float64([]))
-        imgpt1, jac1 = cv2.projectPoints(objpt, np.zeros(3), np.zeros(3), np.eye(3), None)
-        self.assertEqual(imgpt0.shape, (objpt.shape[0], 1, 2))
-        self.assertEqual(imgpt1.shape, imgpt0.shape)
-        self.assertEqual(jac0.shape, jac1.shape)
-        self.assertEqual(jac0.shape[0], 2*objpt.shape[0])
-
-    def test_estimateAffine3D(self):
-        pattern_size = (11, 8)
-        pattern_points = np.zeros((np.prod(pattern_size), 3), np.float32)
-        pattern_points[:,:2] = np.indices(pattern_size).T.reshape(-1, 2)
-        pattern_points *= 10
-        (retval, out, inliers) = cv2.estimateAffine3D(pattern_points, pattern_points)
-        self.assertEqual(retval, 1)
-        if cv2.norm(out[2,:]) < 1e-3:
-            out[2,2]=1
-        self.assertLess(cv2.norm(out, np.float64([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0]])), 1e-3)
-        self.assertEqual(cv2.countNonZero(inliers), pattern_size[0]*pattern_size[1])
-
-    def test_fast(self):
-        fd = cv2.FastFeatureDetector(30, True)
-        img = self.get_sample("samples/cpp/right02.jpg", 0)
-        img = cv2.medianBlur(img, 3)
-        imgc = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
-        keypoints = fd.detect(img)
-        self.assert_(600 <= len(keypoints) <= 700)
-        for kpt in keypoints:
-            self.assertNotEqual(kpt.response, 0)
-
-    def check_close_angles(self, a, b, angle_delta):
-        self.assert_(abs(a - b) <= angle_delta or
-                     abs(360 - abs(a - b)) <= angle_delta)
-
-    def check_close_pairs(self, a, b, delta):
-        self.assertLessEqual(abs(a[0] - b[0]), delta)
-        self.assertLessEqual(abs(a[1] - b[1]), delta)
-
-    def check_close_boxes(self, a, b, delta, angle_delta):
-        self.check_close_pairs(a[0], b[0], delta)
-        self.check_close_pairs(a[1], b[1], delta)
-        self.check_close_angles(a[2], b[2], angle_delta)
-
-    def test_geometry(self):
-        npt = 100
-        np.random.seed(244)
-        a = np.random.randn(npt,2).astype('float32')*50 + 150
-
-        img = np.zeros((300, 300, 3), dtype='uint8')
-        be = cv2.fitEllipse(a)
-        br = cv2.minAreaRect(a)
-        mc, mr = cv2.minEnclosingCircle(a)
-
-        be0 = ((150.2511749267578, 150.77322387695312), (158.024658203125, 197.57696533203125), 37.57804489135742)
-        br0 = ((161.2974090576172, 154.41793823242188), (199.2301483154297, 207.7177734375), -9.164555549621582)
-        mc0, mr0 = (160.41790771484375, 144.55152893066406), 136.713500977
-
-        self.check_close_boxes(be, be0, 5, 15)
-        self.check_close_boxes(br, br0, 5, 15)
-        self.check_close_pairs(mc, mc0, 5)
-        self.assertLessEqual(abs(mr - mr0), 5)
-
-if __name__ == '__main__':
-    print "testing", cv2.__version__
-    random.seed(0)
-    unittest.main()
diff --git a/modules/python/test/ticket_6.py b/modules/python/test/ticket_6.py
index 533027f5b9..7249ff2c75 100755
--- a/modules/python/test/ticket_6.py
+++ b/modules/python/test/ticket_6.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import urllib
 import cv2.cv as cv
diff --git a/modules/python/test/tickets.py b/modules/python/test/tickets.py
index 1e756bcd82..de51e7aa16 100755
--- a/modules/python/test/tickets.py
+++ b/modules/python/test/tickets.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 import unittest
 import random
diff --git a/modules/python/test/transformations.py b/modules/python/test/transformations.py
index 1f63bcef22..5dce6b0497 100755
--- a/modules/python/test/transformations.py
+++ b/modules/python/test/transformations.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 # -*- coding: utf-8 -*-
 # transformations.py
diff --git a/modules/softcascade/CMakeLists.txt b/modules/softcascade/CMakeLists.txt
index d558e8d295..84d098d79c 100644
--- a/modules/softcascade/CMakeLists.txt
+++ b/modules/softcascade/CMakeLists.txt
@@ -1,3 +1,3 @@
 set(the_description "Soft Cascade detection and training")
-ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4310 -Wundef -Wsign-promo -Wmissing-declarations -Wmissing-prototypes)
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4310 -Wundef -Wmissing-declarations -Wmissing-prototypes)
 ocv_define_module(softcascade opencv_core opencv_imgproc opencv_ml)
diff --git a/modules/softcascade/doc/softcascade.rst b/modules/softcascade/doc/softcascade.rst
index a8e0417323..4b3a44a36a 100644
--- a/modules/softcascade/doc/softcascade.rst
+++ b/modules/softcascade/doc/softcascade.rst
@@ -9,4 +9,4 @@ softcascade. Soft Cascade object detection and training.
 
     softcascade_detector
     softcascade_training
-    softcascade_cuda
\ No newline at end of file
+    softcascade_cuda
diff --git a/modules/softcascade/doc/softcascade_training.rst b/modules/softcascade/doc/softcascade_training.rst
index 0d19bd6221..cc83b38e43 100644
--- a/modules/softcascade/doc/softcascade_training.rst
+++ b/modules/softcascade/doc/softcascade_training.rst
@@ -31,7 +31,7 @@ Public interface for soft cascade training algorithm. ::
         virtual bool train(const Dataset* dataset, const FeaturePool* pool, int weaks, int treeDepth) = 0;
         virtual void setRejectThresholds(OutputArray thresholds) = 0;
         virtual void write( cv::FileStorage &fs, const FeaturePool* pool, InputArray thresholds) const = 0;
-        virtual void write( CvFileStorage* fs, string name) const = 0;
+        virtual void write( CvFileStorage* fs, String name) const = 0;
 
     };
 
@@ -71,7 +71,7 @@ softcascade::Octave::write
 --------------------------
 
 .. ocv:function:: void softcascade::Octave::train(cv::FileStorage &fs, const FeaturePool* pool, InputArray thresholds) const
-.. ocv:function:: void softcascade::Octave::train( CvFileStorage* fs, string name) const
+.. ocv:function:: void softcascade::Octave::train( CvFileStorage* fs, String name) const
 
     :param fs an output file storage to store trained detector.
 
@@ -140,4 +140,4 @@ Compute feature on integral channel image.
 
     :param si an index of sample.
 
-    :param fs a channel matrix.
\ No newline at end of file
+    :param fs a channel matrix.
diff --git a/modules/softcascade/include/opencv2/softcascade.hpp b/modules/softcascade/include/opencv2/softcascade.hpp
index be418bc64e..60f33aa052 100644
--- a/modules/softcascade/include/opencv2/softcascade.hpp
+++ b/modules/softcascade/include/opencv2/softcascade.hpp
@@ -43,8 +43,10 @@
 #ifndef __OPENCV_SOFTCASCADE_HPP__
 #define __OPENCV_SOFTCASCADE_HPP__
 
+#include <iosfwd>
+
 #include "opencv2/core.hpp"
-#include "opencv2/core/gpumat.hpp"
+#include "opencv2/core/gpu.hpp"
 
 namespace cv { namespace softcascade {
 
@@ -109,7 +111,7 @@ public:
 
     float operator() (const cv::Mat& integrals, const cv::Size& model) const;
 
-    friend void write(cv::FileStorage& fs, const std::string&, const ChannelFeature& f);
+    friend void write(cv::FileStorage& fs, const String&, const ChannelFeature& f);
     friend std::ostream& operator<<(std::ostream& out, const ChannelFeature& f);
 
 private:
@@ -117,7 +119,7 @@ private:
     int channel;
 };
 
-void write(cv::FileStorage& fs, const std::string&, const ChannelFeature& f);
+void write(cv::FileStorage& fs, const String&, const ChannelFeature& f);
 std::ostream& operator<<(std::ostream& out, const ChannelFeature& m);
 
 // ========================================================================== //
@@ -135,7 +137,7 @@ public:
     CV_WRAP virtual int totalChannels() const = 0;
     virtual cv::AlgorithmInfo* info() const = 0;
 
-    CV_WRAP static cv::Ptr<ChannelFeatureBuilder> create(const std::string& featureType);
+    CV_WRAP static cv::Ptr<ChannelFeatureBuilder> create(const String& featureType);
 };
 
 // ========================================================================== //
@@ -211,7 +213,7 @@ public:
     virtual bool train(const Dataset* dataset, const FeaturePool* pool, int weaks, int treeDepth) = 0;
     virtual void setRejectThresholds(OutputArray thresholds) = 0;
     virtual void write( cv::FileStorage &fs, const FeaturePool* pool, InputArray thresholds) const = 0;
-    virtual void write( CvFileStorage* fs, std::string name) const = 0;
+    virtual void write( CvFileStorage* fs, String name) const = 0;
 };
 
 CV_EXPORTS bool initModule_softcascade(void);
@@ -295,4 +297,4 @@ private:
 
 }} // namespace cv { namespace softcascade {
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/softcascade/misc/roc_test.py b/modules/softcascade/misc/roc_test.py
index f91630aab7..30361f2970 100755
--- a/modules/softcascade/misc/roc_test.py
+++ b/modules/softcascade/misc/roc_test.py
@@ -100,4 +100,4 @@ if __name__ == "__main__":
         fppi, miss_rate = sft.computeROC(confidenses, tp, nannotated, nframes, ignored)
         sft.plotLogLog(fppi, miss_rate, plot_colors[idx])
 
-    sft.showPlot(args.output, args.labels)
\ No newline at end of file
+    sft.showPlot(args.output, args.labels)
diff --git a/modules/softcascade/misc/scale_caltech.py b/modules/softcascade/misc/scale_caltech.py
index 29d9707c86..07e7d06bb4 100755
--- a/modules/softcascade/misc/scale_caltech.py
+++ b/modules/softcascade/misc/scale_caltech.py
@@ -139,4 +139,4 @@ if __name__ == "__main__":
         path, ext = os.path.splitext(each)
         path = path + ".seq"
         print path
-        extractPositive(open(each), path, opath, options.octave, options.min_possible)
\ No newline at end of file
+        extractPositive(open(each), path, opath, options.octave, options.min_possible)
diff --git a/modules/softcascade/misc/scale_inria.py b/modules/softcascade/misc/scale_inria.py
index b08a4bc7f2..ebfc6f366e 100755
--- a/modules/softcascade/misc/scale_inria.py
+++ b/modules/softcascade/misc/scale_inria.py
@@ -136,4 +136,4 @@ if __name__ == "__main__":
             cv2.imwrite(os.path.join(cneg_path, out_name), img)
             idx = idx + 1
             print "." ,
-            sys.stdout.flush()
\ No newline at end of file
+            sys.stdout.flush()
diff --git a/modules/softcascade/misc/sft.py b/modules/softcascade/misc/sft.py
index cefbb3700b..6c82d71b6c 100644
--- a/modules/softcascade/misc/sft.py
+++ b/modules/softcascade/misc/sft.py
@@ -278,4 +278,4 @@ def parse_caltech(f):
         for idx, (p, pv, oc) in enumerate(zip(*[pos, posv, occl])):
             annotations[start + idx].append((type, p, oc, pv))
 
-    return annotations
\ No newline at end of file
+    return annotations
diff --git a/modules/softcascade/perf/perf_cuda_softcascade.cpp b/modules/softcascade/perf/perf_cuda_softcascade.cpp
index 1c4061f65b..09f1aad0bf 100644
--- a/modules/softcascade/perf/perf_cuda_softcascade.cpp
+++ b/modules/softcascade/perf/perf_cuda_softcascade.cpp
@@ -233,4 +233,4 @@ RUN_GPU(SCascadeTest, detectStream)
 
 NO_CPU(SCascadeTest, detectStream)
 
-#undef SC_PERF_TEST_P
\ No newline at end of file
+#undef SC_PERF_TEST_P
diff --git a/modules/softcascade/perf/perf_main.cpp b/modules/softcascade/perf/perf_main.cpp
index f045c4b46d..9a3869f5ef 100644
--- a/modules/softcascade/perf/perf_main.cpp
+++ b/modules/softcascade/perf/perf_main.cpp
@@ -41,5 +41,13 @@
 //M*/
 
 #include "perf_precomp.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
 
-CV_PERF_TEST_MAIN(softcascade)
+static const char * impls[] = {
+#ifdef HAVE_CUDA
+    "cuda",
+#endif
+    "plain"
+};
+
+CV_PERF_TEST_MAIN_WITH_IMPLS(softcascade, impls, perf::printCudaInfo())
diff --git a/modules/softcascade/perf/perf_softcascade.cpp b/modules/softcascade/perf/perf_softcascade.cpp
index c8f9fab9ec..8165bd981e 100644
--- a/modules/softcascade/perf/perf_softcascade.cpp
+++ b/modules/softcascade/perf/perf_softcascade.cpp
@@ -44,4 +44,4 @@ PERF_TEST_P(detect, SoftCascadeDetector,
     extractRacts(objectBoxes, rects);
     std::sort(rects.begin(), rects.end(), perf::comparators::RectLess());
     SANITY_CHECK(rects);
-}
\ No newline at end of file
+}
diff --git a/modules/softcascade/src/cuda/channels.cu b/modules/softcascade/src/cuda/channels.cu
index 782e243b5d..78489e1f8b 100644
--- a/modules/softcascade/src/cuda/channels.cu
+++ b/modules/softcascade/src/cuda/channels.cu
@@ -40,28 +40,10 @@
 //
 //M*/
 
-#include "opencv2/core/cuda_devptrs.hpp"
+#include "opencv2/core/gpu_types.hpp"
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace softcascade { namespace internal {
-void error(const char *error_string, const char *file, const int line, const char *func);
-}}}
-#if defined(__GNUC__)
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__)
-#endif
-
-static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
-{
-    if (cudaSuccess != err) cv::softcascade::internal::error(cudaGetErrorString(err), file, line, func);
-}
-
-__host__ __device__ __forceinline__ int divUp(int total, int grain)
-{
-    return (total + grain - 1) / grain;
-}
-
-namespace cv { namespace softcascade { namespace device
+namespace cv { namespace softcascade { namespace cudev
 {
     typedef unsigned int uint;
     typedef unsigned short ushort;
@@ -393,7 +375,7 @@ namespace cv { namespace softcascade { namespace device
 
         {
             const dim3 block(32, 8);
-            const dim3 grid(divUp(integral.cols, block.x), 1);
+            const dim3 grid(cv::gpu::cudev::divUp(integral.cols, block.x), 1);
 
             shfl_integral_vertical<<<grid, block, 0, stream>>>(integral);
             cudaSafeCall( cudaGetLastError() );
@@ -478,7 +460,7 @@ namespace cv { namespace softcascade { namespace device
 
         {
             const dim3 block(32, 8);
-            const dim3 grid(divUp(integral.cols, block.x), 1);
+            const dim3 grid(cv::gpu::cudev::divUp(integral.cols, block.x), 1);
 
             shfl_integral_vertical<<<grid, block, 0, stream>>>((cv::gpu::PtrStepSz<unsigned int>)buffer, integral);
             cudaSafeCall( cudaGetLastError() );
@@ -518,8 +500,8 @@ namespace cv { namespace softcascade { namespace device
     void transform(const cv::gpu::PtrStepSz<uchar3>& bgr, cv::gpu::PtrStepSzb gray)
     {
         const dim3 block(32, 8);
-        const dim3 grid(divUp(bgr.cols, block.x), divUp(bgr.rows, block.y));
+        const dim3 grid(cv::gpu::cudev::divUp(bgr.cols, block.x), cv::gpu::cudev::divUp(bgr.rows, block.y));
         device_transform<<<grid, block>>>(bgr, gray);
         cudaSafeCall(cudaDeviceSynchronize());
     }
-}}}
\ No newline at end of file
+}}}
diff --git a/modules/softcascade/src/cuda/icf-sc.cu b/modules/softcascade/src/cuda/icf-sc.cu
index 8d65aad0e6..b119209dbe 100644
--- a/modules/softcascade/src/cuda/icf-sc.cu
+++ b/modules/softcascade/src/cuda/icf-sc.cu
@@ -43,26 +43,9 @@
 #include <cuda_invoker.hpp>
 #include <float.h>
 #include <stdio.h>
+#include "opencv2/core/cuda/common.hpp"
 
-namespace cv { namespace softcascade { namespace internal {
-void error(const char *error_string, const char *file, const int line, const char *func);
-}}}
-#if defined(__GNUC__)
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__)
-#endif
-
-static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
-{
-    if (cudaSuccess != err) cv::softcascade::internal::error(cudaGetErrorString(err), file, line, func);
-}
-
-#ifndef CV_PI
-    #define CV_PI   3.1415926535897932384626433832795
-#endif
-
-namespace cv { namespace softcascade { namespace device {
+namespace cv { namespace softcascade { namespace cudev {
 
 typedef unsigned char uchar;
 
@@ -82,8 +65,8 @@ typedef unsigned char uchar;
     }
 
     template<int FACTOR>
-    __global__ void shrink(const uchar* __restrict__ hogluv, const int inPitch,
-                                 uchar* __restrict__ shrank, const int outPitch )
+    __global__ void shrink(const uchar* __restrict__ hogluv, const size_t inPitch,
+                                 uchar* __restrict__ shrank, const size_t outPitch )
     {
         const int y = blockIdx.y * blockDim.y + threadIdx.y;
         const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -127,7 +110,7 @@ typedef unsigned char uchar;
         __v = static_cast<uchar>((V + 140.f) * (255.f / (122.f + 140.f )));
     }
 
-    __global__ void bgr2Luv_d(const uchar* rgb, const int rgbPitch, uchar* luvg, const int luvgPitch)
+    __global__ void bgr2Luv_d(const uchar* rgb, const size_t rgbPitch, uchar* luvg, const size_t luvgPitch)
     {
         const int y = blockIdx.y * blockDim.y + threadIdx.y;
         const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -154,10 +137,10 @@ typedef unsigned char uchar;
     template<bool isDefaultNum>
     __device__ __forceinline__ int fast_angle_bin(const float& dx, const float& dy)
     {
-        const float angle_quantum = CV_PI / 6.f;
+        const float angle_quantum = CV_PI_F / 6.f;
         float angle = atan2(dx, dy) + (angle_quantum / 2.f);
 
-        if (angle < 0) angle += CV_PI;
+        if (angle < 0) angle += CV_PI_F;
 
         const float angle_scaling = 1.f / angle_quantum;
         return static_cast<int>(angle * angle_scaling) % 6;
@@ -191,8 +174,8 @@ typedef unsigned char uchar;
         {
             int i = 3;
             float2 bin_vector_i;
-            bin_vector_i.x = ::cos(i * (CV_PI / 6.f));
-            bin_vector_i.y = ::sin(i * (CV_PI / 6.f));
+            bin_vector_i.x = ::cos(i * (CV_PI_F / 6.f));
+            bin_vector_i.y = ::sin(i * (CV_PI_F / 6.f));
 
             const float dot_product = fabs(dx * bin_vector_i.x + dy * bin_vector_i.y);
             if(dot_product > max_dot)
@@ -256,8 +239,8 @@ typedef unsigned char uchar;
 
     // ToDo: use textures or uncached load instruction.
     __global__ void magToHist(const uchar* __restrict__ mag,
-                              const float* __restrict__ angle, const int angPitch,
-                                    uchar* __restrict__ hog,   const int hogPitch, const int fh)
+                              const float* __restrict__ angle, const size_t angPitch,
+                                    uchar* __restrict__ hog,   const size_t hogPitch, const int fh)
     {
         const int y = blockIdx.y * blockDim.y + threadIdx.y;
         const int x = blockIdx.x * blockDim.x + threadIdx.x;
diff --git a/modules/softcascade/src/cuda_invoker.hpp b/modules/softcascade/src/cuda_invoker.hpp
index 9bdd45b883..92f30d01d2 100644
--- a/modules/softcascade/src/cuda_invoker.hpp
+++ b/modules/softcascade/src/cuda_invoker.hpp
@@ -44,8 +44,7 @@
 #ifndef __OPENCV_ICF_HPP__
 #define __OPENCV_ICF_HPP__
 
-// #include <opencv2/gpu/device/common.hpp>
-#include "opencv2/core/cuda_devptrs.hpp"
+#include "opencv2/core/gpu_types.hpp"
 #include "cuda_runtime_api.h"
 
 #if defined __CUDACC__
@@ -55,7 +54,7 @@
 #endif
 
 
-namespace cv { namespace softcascade { namespace device {
+namespace cv { namespace softcascade { namespace cudev {
 
 typedef unsigned char uchar;
 typedef unsigned int uint;
@@ -63,14 +62,14 @@ typedef unsigned short ushort;
 
 struct Octave
 {
+    float scale;
+    ushort2 size;
     ushort index;
     ushort stages;
     ushort shrinkage;
-    ushort2 size;
-    float scale;
 
     Octave(const ushort i, const ushort s, const ushort sh, const ushort2 sz, const float sc)
-    : index(i), stages(s), shrinkage(sh), size(sz), scale(sc) {}
+    : scale(sc), size(sz), index(i), stages(s), shrinkage(sh) {}
 };
 
 struct Level
@@ -111,7 +110,7 @@ struct Detection
 
     Detection(){}
     __device_inline__ Detection(int _x, int _y, uchar _w, uchar _h, float c)
-    : x(_x), y(_y), w(_w), h(_h), confidence(c), kind(0) {};
+    : x(static_cast<ushort>(_x)), y(static_cast<ushort>(_y)), w(_w), h(_h), confidence(c), kind(0) {};
 };
 
 struct GK107PolicyX4
@@ -154,4 +153,4 @@ struct CascadeInvoker
 
 }}}
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/softcascade/src/detector.cpp b/modules/softcascade/src/detector.cpp
index 1a3cae2006..ad74614f41 100644
--- a/modules/softcascade/src/detector.cpp
+++ b/modules/softcascade/src/detector.cpp
@@ -190,7 +190,7 @@ struct ChannelStorage
 
     enum {HOG_BINS = 6, HOG_LUV_BINS = 10};
 
-    ChannelStorage(const cv::Mat& colored, int shr, std::string featureTypeStr) : shrinkage(shr)
+    ChannelStorage(const cv::Mat& colored, int shr, cv::String featureTypeStr) : shrinkage(shr)
     {
         model_height = cvRound(colored.rows / (float)shrinkage);
         if (featureTypeStr == "ICF") featureTypeStr = "HOG6MagLuv";
@@ -240,7 +240,7 @@ struct cv::softcascade::Detector::Fields
     typedef std::vector<SOctave>::iterator  octIt_t;
     typedef std::vector<Detection> dvector;
 
-    std::string featureTypeStr;
+    String featureTypeStr;
 
     void detectAt(const int dx, const int dy, const Level& level, const ChannelStorage& storage, dvector& detections) const
     {
@@ -364,14 +364,14 @@ struct cv::softcascade::Detector::Fields
         static const char *const FEATURE_FORMAT      = "featureFormat";
 
         // only Ada Boost supported
-        std::string stageTypeStr = (std::string)root[SC_STAGE_TYPE];
+        String stageTypeStr = (String)root[SC_STAGE_TYPE];
         CV_Assert(stageTypeStr == SC_BOOST);
 
-        std::string fformat = (std::string)root[FEATURE_FORMAT];
+        String fformat = (String)root[FEATURE_FORMAT];
         bool useBoxes = (fformat == "BOX");
 
         // only HOG-like integral channel features supported
-        featureTypeStr = (std::string)root[SC_FEATURE_TYPE];
+        featureTypeStr = (String)root[SC_FEATURE_TYPE];
         CV_Assert(featureTypeStr == SC_ICF || featureTypeStr == SC_HOG6_MAG_LUV);
 
         origObjWidth  = (int)root[SC_ORIG_W];
@@ -593,4 +593,4 @@ void cv::softcascade::Detector::detect(InputArray _image, InputArray _rois,  Out
         rectPtr[i] = (*it).bb();
         confPtr[i] = (*it).confidence;
     }
-}
\ No newline at end of file
+}
diff --git a/modules/softcascade/src/detector_cuda.cpp b/modules/softcascade/src/detector_cuda.cpp
index bbadc9c54f..ef19fad058 100644
--- a/modules/softcascade/src/detector_cuda.cpp
+++ b/modules/softcascade/src/detector_cuda.cpp
@@ -43,45 +43,32 @@
 #include "precomp.hpp"
 
 #if !defined (HAVE_CUDA)
-#define throw_nogpu() CV_Error(CV_GpuNotSupported, "The library is compiled without CUDA support")
-cv::softcascade::SCascade::SCascade(const double, const double, const int, const int) { throw_nogpu(); }
 
-cv::softcascade::SCascade::~SCascade() { throw_nogpu(); }
+cv::softcascade::SCascade::SCascade(const double, const double, const int, const int) { throw_no_cuda(); }
 
-bool cv::softcascade::SCascade::load(const FileNode&) { throw_nogpu(); return false;}
+cv::softcascade::SCascade::~SCascade() { throw_no_cuda(); }
 
-void cv::softcascade::SCascade::detect(InputArray, InputArray, OutputArray, cv::gpu::Stream&) const { throw_nogpu(); }
+bool cv::softcascade::SCascade::load(const FileNode&) { throw_no_cuda(); return false;}
+
+void cv::softcascade::SCascade::detect(InputArray, InputArray, OutputArray, cv::gpu::Stream&) const { throw_no_cuda(); }
 
 void cv::softcascade::SCascade::read(const FileNode& fn) { Algorithm::read(fn); }
 
-cv::softcascade::ChannelsProcessor::ChannelsProcessor() { throw_nogpu(); }
- cv::softcascade::ChannelsProcessor::~ChannelsProcessor() { throw_nogpu(); }
+cv::softcascade::ChannelsProcessor::ChannelsProcessor() { throw_no_cuda(); }
+ cv::softcascade::ChannelsProcessor::~ChannelsProcessor() { throw_no_cuda(); }
 
 cv::Ptr<cv::softcascade::ChannelsProcessor> cv::softcascade::ChannelsProcessor::create(const int, const int, const int)
-{ throw_nogpu(); return cv::Ptr<cv::softcascade::ChannelsProcessor>(0); }
+{ throw_no_cuda(); return cv::Ptr<cv::softcascade::ChannelsProcessor>(); }
 
 #else
+
 # include "cuda_invoker.hpp"
-# include "opencv2/core/stream_accessor.hpp"
-namespace
-{
-#if defined(__GNUC__)
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__, __func__)
-#else /* defined(__CUDACC__) || defined(__MSVC__) */
-    #define cudaSafeCall(expr)  ___cudaSafeCall(expr, __FILE__, __LINE__)
-#endif
 
-    inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
-    {
-        if (cudaSuccess != err) cv::gpu::error(cudaGetErrorString(err), file, line, func);
-    }
-}
-
-cv::softcascade::device::Level::Level(int idx, const Octave& oct, const float scale, const int w, const int h)
+cv::softcascade::cudev::Level::Level(int idx, const Octave& oct, const float scale, const int w, const int h)
 :  octave(idx), step(oct.stages), relScale(scale / oct.scale)
 {
-    workRect.x = cvRound(w / (float)oct.shrinkage);
-    workRect.y = cvRound(h / (float)oct.shrinkage);
+    workRect.x = (unsigned char)cvRound(w / (float)oct.shrinkage);
+    workRect.y = (unsigned char)cvRound(h / (float)oct.shrinkage);
 
     objSize.x  = cv::saturate_cast<uchar>(oct.size.x * relScale);
     objSize.y  = cv::saturate_cast<uchar>(oct.size.y * relScale);
@@ -96,7 +83,7 @@ cv::softcascade::device::Level::Level(int idx, const Octave& oct, const float sc
     }
 }
 
-namespace cv { namespace softcascade { namespace device {
+namespace cv { namespace softcascade { namespace cudev {
 
     void fillBins(cv::gpu::PtrStepSzb hogluv, const cv::gpu::PtrStepSzf& nangle,
         const int fw, const int fh, const int bins, cudaStream_t stream);
@@ -136,26 +123,26 @@ struct cv::softcascade::SCascade::Fields
         static const char *const SC_F_RECT              = "rect";
 
         // only Ada Boost supported
-        std::string stageTypeStr = (std::string)root[SC_STAGE_TYPE];
+        String stageTypeStr = (String)root[SC_STAGE_TYPE];
         CV_Assert(stageTypeStr == SC_BOOST);
 
         // only HOG-like integral channel features supported
-        std::string featureTypeStr = (std::string)root[SC_FEATURE_TYPE];
+        String featureTypeStr = (String)root[SC_FEATURE_TYPE];
         CV_Assert(featureTypeStr == SC_ICF);
 
         int origWidth  = (int)root[SC_ORIG_W];
         int origHeight = (int)root[SC_ORIG_H];
 
-        std::string fformat = (std::string)root[SC_FEATURE_FORMAT];
+        String fformat = (String)root[SC_FEATURE_FORMAT];
         bool useBoxes = (fformat == "BOX");
         ushort shrinkage = cv::saturate_cast<ushort>((int)root[SC_SHRINKAGE]);
 
         FileNode fn = root[SC_OCTAVES];
         if (fn.empty()) return 0;
 
-        std::vector<device::Octave>  voctaves;
+        std::vector<cudev::Octave>  voctaves;
         std::vector<float>   vstages;
-        std::vector<device::Node>    vnodes;
+        std::vector<cudev::Node>    vnodes;
         std::vector<float>   vleaves;
 
         FileNodeIterator it = fn.begin(), it_end = fn.end();
@@ -168,10 +155,10 @@ struct cv::softcascade::SCascade::Fields
             ushort nweaks = saturate_cast<ushort>((int)fns[SC_OCT_WEAKS]);
 
             ushort2 size;
-            size.x = cvRound(origWidth * scale);
-            size.y = cvRound(origHeight * scale);
+            size.x = (unsigned short)cvRound(origWidth * scale);
+            size.y = (unsigned short)cvRound(origHeight * scale);
 
-            device::Octave octave(octIndex, nweaks, shrinkage, size, scale);
+            cudev::Octave octave(octIndex, nweaks, shrinkage, size, scale);
             CV_Assert(octave.stages > 0);
             voctaves.push_back(octave);
 
@@ -213,7 +200,7 @@ struct cv::softcascade::SCascade::Fields
             }
 
             fns = fns[SC_TREES];
-            if (fn.empty()) return false;
+            if (fn.empty()) return 0;
 
             // for each stage (~ decision tree with H = 2)
             FileNodeIterator st = fns.begin(), st_end = fns.end();
@@ -240,7 +227,7 @@ struct cv::softcascade::SCascade::Fields
                     rect.w = saturate_cast<uchar>(r.height);
 
                     unsigned int channel = saturate_cast<unsigned int>(feature_channels[featureIdx]);
-                    vnodes.push_back(device::Node(rect, channel, th));
+                    vnodes.push_back(cudev::Node(rect, channel, th));
                 }
 
                 intfns = octfn[SC_LEAF];
@@ -252,13 +239,13 @@ struct cv::softcascade::SCascade::Fields
             }
         }
 
-        cv::Mat hoctaves(1, (int) (voctaves.size() * sizeof(device::Octave)), CV_8UC1, (uchar*)&(voctaves[0]));
+        cv::Mat hoctaves(1, (int) (voctaves.size() * sizeof(cudev::Octave)), CV_8UC1, (uchar*)&(voctaves[0]));
         CV_Assert(!hoctaves.empty());
 
         cv::Mat hstages(cv::Mat(vstages).reshape(1,1));
         CV_Assert(!hstages.empty());
 
-        cv::Mat hnodes(1, (int) (vnodes.size() * sizeof(device::Node)), CV_8UC1, (uchar*)&(vnodes[0]) );
+        cv::Mat hnodes(1, (int) (vnodes.size() * sizeof(cudev::Node)), CV_8UC1, (uchar*)&(vnodes[0]) );
         CV_Assert(!hnodes.empty());
 
         cv::Mat hleaves(cv::Mat(vleaves).reshape(1,1));
@@ -274,7 +261,7 @@ struct cv::softcascade::SCascade::Fields
 
     bool check(float mins,float  maxs, int scales)
     {
-        bool updated = (minScale == mins) || (maxScale == maxs) || (totals = scales);
+        bool updated = ((minScale == mins) || (maxScale == maxs) || (totals == scales));
 
         minScale = mins;
         maxScale = maxScale;
@@ -285,7 +272,7 @@ struct cv::softcascade::SCascade::Fields
 
     int createLevels(const int fh, const int fw)
     {
-        std::vector<device::Level> vlevels;
+        std::vector<cudev::Level> vlevels;
         float logFactor = (::log(maxScale) - ::log(minScale)) / (totals -1);
 
         float scale = minScale;
@@ -298,7 +285,7 @@ struct cv::softcascade::SCascade::Fields
             float logScale = ::log(scale);
             int fit = fitOctave(voctaves, logScale);
 
-            device::Level level(fit, voctaves[fit], scale, width, height);
+            cudev::Level level(fit, voctaves[fit], scale, width, height);
 
             if (!width || !height)
                 break;
@@ -312,7 +299,7 @@ struct cv::softcascade::SCascade::Fields
             scale = ::std::min(maxScale, ::expf(::log(scale) + logFactor));
         }
 
-        cv::Mat hlevels = cv::Mat(1, (int) (vlevels.size() * sizeof(device::Level)), CV_8UC1, (uchar*)&(vlevels[0]) );
+        cv::Mat hlevels = cv::Mat(1, (int) (vlevels.size() * sizeof(cudev::Level)), CV_8UC1, (uchar*)&(vlevels[0]) );
         CV_Assert(!hlevels.empty());
         levels.upload(hlevels);
         downscales = dcs;
@@ -348,15 +335,12 @@ struct cv::softcascade::SCascade::Fields
 
     void detect(cv::gpu::GpuMat& objects, cv::gpu::Stream& s) const
     {
-        if (s)
-            s.enqueueMemSet(objects, 0);
-        else
-            cudaMemset(objects.data, 0, sizeof(Detection));
+        objects.setTo(Scalar::all(0), s);
 
         cudaSafeCall( cudaGetLastError());
 
-        device::CascadeInvoker<device::GK107PolicyX4> invoker
-        = device::CascadeInvoker<device::GK107PolicyX4>(levels, stages, nodes, leaves);
+        cudev::CascadeInvoker<cudev::GK107PolicyX4> invoker
+        = cudev::CascadeInvoker<cudev::GK107PolicyX4>(levels, stages, nodes, leaves);
 
         cudaStream_t stream = cv::gpu::StreamAccessor::getStream(s);
         invoker(mask, hogluv, objects, downscales, stream);
@@ -367,31 +351,23 @@ struct cv::softcascade::SCascade::Fields
         cv::gpu::GpuMat ndetections = cv::gpu::GpuMat(objects, cv::Rect(0, 0, sizeof(Detection), 1));
         ensureSizeIsEnough(objects.rows, objects.cols, CV_8UC1, overlaps);
 
-        if (s)
-        {
-            s.enqueueMemSet(overlaps, 0);
-            s.enqueueMemSet(suppressed, 0);
-        }
-        else
-        {
-            overlaps.setTo(0);
-            suppressed.setTo(0);
-        }
+        overlaps.setTo(0, s);
+        suppressed.setTo(0, s);
 
         cudaStream_t stream = cv::gpu::StreamAccessor::getStream(s);
-        device::suppress(objects, overlaps, ndetections, suppressed, stream);
+        cudev::suppress(objects, overlaps, ndetections, suppressed, stream);
     }
 
 private:
 
-    typedef std::vector<device::Octave>::const_iterator  octIt_t;
-    static int fitOctave(const std::vector<device::Octave>& octs, const float& logFactor)
+    typedef std::vector<cudev::Octave>::const_iterator  octIt_t;
+    static int fitOctave(const std::vector<cudev::Octave>& octs, const float& logFactor)
     {
         float minAbsLog = FLT_MAX;
         int res =  0;
         for (int oct = 0; oct < (int)octs.size(); ++oct)
         {
-            const device::Octave& octave =octs[oct];
+            const cudev::Octave& octave =octs[oct];
             float logOctave = ::log(octave.scale);
             float logAbsScale = ::fabs(logFactor - logOctave);
 
@@ -452,7 +428,7 @@ public:
 //     cv::gpu::GpuMat collected;
 
 
-    std::vector<device::Octave> voctaves;
+    std::vector<cudev::Octave> voctaves;
 
 //     DeviceInfo info;
 
@@ -463,6 +439,9 @@ public:
         DEFAULT_FRAME_HEIGHT       = 480,
         HOG_LUV_BINS               = 10
     };
+
+private:
+    cv::softcascade::SCascade::Fields& operator=( const cv::softcascade::SCascade::Fields & );
 };
 
 cv::softcascade::SCascade::SCascade(const double mins, const double maxs, const int sc, const int fl)
@@ -495,23 +474,17 @@ void integral(const cv::gpu::GpuMat& src, cv::gpu::GpuMat& sum, cv::gpu::GpuMat&
     {
         ensureSizeIsEnough(((src.rows + 7) / 8) * 8, ((src.cols + 63) / 64) * 64, CV_32SC1, buffer);
 
-        cv::softcascade::device::shfl_integral(src, buffer, stream);
+        cv::softcascade::cudev::shfl_integral(src, buffer, stream);
 
         sum.create(src.rows + 1, src.cols + 1, CV_32SC1);
-        if (s)
-            s.enqueueMemSet(sum, cv::Scalar::all(0));
-        else
-            sum.setTo(cv::Scalar::all(0));
+        sum.setTo(cv::Scalar::all(0), s);
 
         cv::gpu::GpuMat inner = sum(cv::Rect(1, 1, src.cols, src.rows));
         cv::gpu::GpuMat res = buffer(cv::Rect(0, 0, src.cols, src.rows));
 
-        if (s)
-            s.enqueueCopy(res, inner);
-        else
-            res.copyTo(inner);
+        res.copyTo(inner, s);
     }
-    else {CV_Error(CV_GpuNotSupported, ": CC 3.x required.");}
+    else {CV_Error(cv::Error::GpuNotSupported, ": CC 3.x required.");}
 }
 
 }
@@ -536,7 +509,7 @@ void cv::softcascade::SCascade::detect(InputArray _image, InputArray _rois, Outp
 
     flds.mask.create( rois.cols / shr, rois.rows / shr, rois.type());
 
-    device::shrink(rois, flds.mask);
+    cudev::shrink(rois, flds.mask);
     //cv::gpu::transpose(flds.genRoiTmp, flds.mask, s);
 
     if (type == CV_8UC3)
@@ -547,14 +520,11 @@ void cv::softcascade::SCascade::detect(InputArray _image, InputArray _rois, Outp
             flds.createLevels(image.rows, image.cols);
 
         flds.preprocessor->apply(image, flds.shrunk);
-        integral(flds.shrunk, flds.hogluv, flds.integralBuffer, s);
+        ::integral(flds.shrunk, flds.hogluv, flds.integralBuffer, s);
     }
     else
     {
-        if (s)
-            s.enqueueCopy(image, flds.hogluv);
-        else
-            image.copyTo(flds.hogluv);
+        image.copyTo(flds.hogluv, s);
     }
 
     flds.detect(objects, s);
@@ -581,10 +551,7 @@ using cv::gpu::GpuMat;
 
 inline void setZero(cv::gpu::GpuMat& m, cv::gpu::Stream& s)
 {
-    if (s)
-        s.enqueueMemSet(m, 0);
-    else
-        m.setTo(0);
+    m.setTo(0, s);
 }
 
 struct SeparablePreprocessor : public cv::softcascade::ChannelsProcessor
@@ -604,12 +571,12 @@ struct SeparablePreprocessor : public cv::softcascade::ChannelsProcessor
         setZero(channels, s);
 
         gray.create(bgr.size(), CV_8UC1);
-        cv::softcascade::device::transform(bgr, gray); //cv::gpu::cvtColor(bgr, gray, CV_BGR2GRAY);
-        cv::softcascade::device::gray2hog(gray, channels(cv::Rect(0, 0, bgr.cols, bgr.rows * (bins + 1))), bins);
+        cv::softcascade::cudev::transform(bgr, gray); //cv::gpu::cvtColor(bgr, gray, CV_BGR2GRAY);
+        cv::softcascade::cudev::gray2hog(gray, channels(cv::Rect(0, 0, bgr.cols, bgr.rows * (bins + 1))), bins);
 
         cv::gpu::GpuMat luv(channels, cv::Rect(0, bgr.rows * (bins + 1), bgr.cols, bgr.rows * 3));
-        cv::softcascade::device::bgr2Luv(bgr, luv);
-        cv::softcascade::device::shrink(channels, shrunk);
+        cv::softcascade::cudev::bgr2Luv(bgr, luv);
+        cv::softcascade::cudev::shrink(channels, shrunk);
     }
 
 private:
@@ -619,6 +586,7 @@ private:
     cv::gpu::GpuMat bgr;
     cv::gpu::GpuMat gray;
     cv::gpu::GpuMat channels;
+    SeparablePreprocessor& operator=( const SeparablePreprocessor& );
 };
 
 }
@@ -626,7 +594,7 @@ private:
 cv::Ptr<cv::softcascade::ChannelsProcessor> cv::softcascade::ChannelsProcessor::create(const int s, const int b, const int m)
 {
     CV_Assert((m && SEPARABLE));
-    return cv::Ptr<cv::softcascade::ChannelsProcessor>(new SeparablePreprocessor(s, b));
+    return makePtr<SeparablePreprocessor>(s, b);
 }
 
 cv::softcascade::ChannelsProcessor::ChannelsProcessor() { }
diff --git a/modules/softcascade/src/integral_channel_builder.cpp b/modules/softcascade/src/integral_channel_builder.cpp
index 36277a1f86..08f155e4c6 100644
--- a/modules/softcascade/src/integral_channel_builder.cpp
+++ b/modules/softcascade/src/integral_channel_builder.cpp
@@ -76,7 +76,7 @@ public:
         channels.create(h * N_CHANNELS, w, CV_8UC1);
         channels.setTo(0);
 
-        cvtColor(frame, gray, CV_BGR2GRAY);
+        cvtColor(frame, gray, cv::COLOR_BGR2GRAY);
 
         cv::Mat df_dx, df_dy, mag, angle;
         cv::Sobel(gray, df_dx, CV_32F, 1, 0);
@@ -102,13 +102,13 @@ public:
         }
 
         cv::Mat luv, shrunk;
-        cv::cvtColor(frame, luv, CV_BGR2Luv);
+        cv::cvtColor(frame, luv, cv::COLOR_BGR2Luv);
 
         std::vector<cv::Mat> splited;
         for (int i = 0; i < 3; ++i)
             splited.push_back(channels(cv::Rect(0, h * (7 + i), w, h)));
         split(luv, splited);
-        cv::resize(channels, shrunk, cv::Size(integrals.cols - 1, integrals.rows - 1), -1 , -1, CV_INTER_AREA);
+        cv::resize(channels, shrunk, cv::Size(integrals.cols - 1, integrals.rows - 1), -1 , -1, cv::INTER_AREA);
         cv::integral(shrunk, integrals, cv::noArray(), CV_32S);
     }
 };
@@ -122,7 +122,7 @@ CV_INIT_ALGORITHM(HOG6MagLuv,  "ChannelFeatureBuilder.HOG6MagLuv", );
 
 ChannelFeatureBuilder::~ChannelFeatureBuilder() {}
 
-cv::Ptr<ChannelFeatureBuilder> ChannelFeatureBuilder::create(const std::string& featureType)
+cv::Ptr<ChannelFeatureBuilder> ChannelFeatureBuilder::create(const cv::String& featureType)
 {
     return Algorithm::create<ChannelFeatureBuilder>("ChannelFeatureBuilder." + featureType);
 }
@@ -158,15 +158,14 @@ float ChannelFeature::operator() (const cv::Mat& integrals, const cv::Size& mode
     return (float)(a - b + c - d);
 }
 
-void cv::softcascade::write(cv::FileStorage& fs, const std::string&, const ChannelFeature& f)
+void cv::softcascade::write(cv::FileStorage& fs, const cv::String&, const ChannelFeature& f)
 {
     fs << "{" << "channel" << f.channel << "rect" << f.bb << "}";
 }
 
 std::ostream& cv::softcascade::operator<<(std::ostream& out, const ChannelFeature& m)
 {
-    out << m.channel << " " << m.bb;
-    return out;
+    return out << m.channel << " " << "[" << m.bb.width << " x " << m.bb.height << " from (" << m.bb.x << ", " << m.bb.y << ")]";
 }
 
 ChannelFeature::~ChannelFeature(){}
diff --git a/modules/softcascade/src/octave.cpp b/modules/softcascade/src/octave.cpp
index 26cd460933..5c5aa2eec0 100644
--- a/modules/softcascade/src/octave.cpp
+++ b/modules/softcascade/src/octave.cpp
@@ -41,8 +41,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/ml.hpp"
 #include <queue>
-#include <string>
 
 using cv::InputArray;
 using cv::OutputArray;
@@ -69,7 +69,7 @@ public:
     virtual bool train(const Dataset* dataset, const FeaturePool* pool, int weaks, int treeDepth);
     virtual void setRejectThresholds(OutputArray thresholds);
     virtual void write( cv::FileStorage &fs, const FeaturePool* pool, InputArray thresholds) const;
-    virtual void write( CvFileStorage* fs, std::string name) const;
+    virtual void write( CvFileStorage* fs, cv::String name) const;
 protected:
     virtual float predict( InputArray _sample, InputArray _votes, bool raw_mode, bool return_sum ) const;
     virtual bool train( const cv::Mat& trainData, const cv::Mat& responses, const cv::Mat& varIdx=cv::Mat(),
@@ -321,6 +321,8 @@ void BoostedSoftCascadeOctave::traverse(const CvBoostTree* tree, cv::FileStorage
 
 
     fs << "}";
+
+    delete [] leafs;
 }
 
 void BoostedSoftCascadeOctave::write( cv::FileStorage &fso, const FeaturePool* pool, InputArray _thresholds) const
@@ -436,7 +438,7 @@ float BoostedSoftCascadeOctave::predict( const Mat& _sample, const cv::Range ran
     return CvBoost::predict(&sample, 0, 0, range, false, true);
 }
 
-void BoostedSoftCascadeOctave::write( CvFileStorage* fs, std::string _name) const
+void BoostedSoftCascadeOctave::write( CvFileStorage* fs, cv::String _name) const
 {
     CvBoost::write(fs, _name.c_str());
 }
diff --git a/modules/softcascade/src/precomp.cpp b/modules/softcascade/src/precomp.cpp
deleted file mode 100644
index 62c236d08a..0000000000
--- a/modules/softcascade/src/precomp.cpp
+++ /dev/null
@@ -1,43 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2008-2013, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and / or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
diff --git a/modules/softcascade/src/precomp.hpp b/modules/softcascade/src/precomp.hpp
index 2b6be26640..83ba65b4b0 100644
--- a/modules/softcascade/src/precomp.hpp
+++ b/modules/softcascade/src/precomp.hpp
@@ -43,16 +43,13 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
+#include <iostream>
 
 #include "opencv2/softcascade.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/core_c.h"
-#include "opencv2/core/internal.hpp"
-#include "opencv2/ml.hpp"
+
+#include "opencv2/core/private.hpp"
+#include "opencv2/core/private.gpu.hpp"
 
 namespace cv { namespace softcascade { namespace internal
 {
diff --git a/modules/softcascade/src/softcascade_init.cpp b/modules/softcascade/src/softcascade_init.cpp
index 9563ac629a..6f3c8b6a5d 100644
--- a/modules/softcascade/src/softcascade_init.cpp
+++ b/modules/softcascade/src/softcascade_init.cpp
@@ -58,27 +58,9 @@ CV_INIT_ALGORITHM(SCascade, "CascadeDetector.SCascade",
 
 bool initModule_softcascade(void)
 {
-    Ptr<Algorithm> sc = createSCascade();
-    Ptr<Algorithm> sc1 = createDetector();
+    Ptr<Algorithm> sc = createSCascade_ptr_hidden();
+    Ptr<Algorithm> sc1 = createDetector_ptr_hidden();
     return (sc1->info() != 0) && (sc->info() != 0);
 }
 
-namespace internal {
-void error(const char *error_string, const char *file, const int line, const char *func)
-{
-    int code = CV_GpuApiCallError;
-
-    if (std::uncaught_exception())
-    {
-        const char* errorStr = cvErrorStr(code);
-        const char* function = func ? func : "unknown function";
-
-        std::cerr << "OpenCV Error: " << errorStr << "(" << error_string << ") in " << function << ", file " << file << ", line " << line;
-        std::cerr.flush();
-    }
-    else
-        cv::error( cv::Exception(code, error_string, func, file, line) );
-}
-}
-
-} }
\ No newline at end of file
+} }
diff --git a/modules/softcascade/test/test_cuda_softcascade.cpp b/modules/softcascade/test/test_cuda_softcascade.cpp
index 7787664e40..5973d25f21 100644
--- a/modules/softcascade/test/test_cuda_softcascade.cpp
+++ b/modules/softcascade/test/test_cuda_softcascade.cpp
@@ -41,7 +41,7 @@
 //M*/
 
 #include "test_precomp.hpp"
-#include "opencv2/core/gpumat.hpp"
+#include "opencv2/core/gpu.hpp"
 
 
 #ifdef HAVE_CUDA
@@ -313,4 +313,4 @@ INSTANTIATE_TEST_CASE_P(cuda_accelerated, SCascadeTestAll, testing::Combine( ALL
                     testing::Values(Fixture("cascades/inria_caltech-17.01.2013.xml", 7),
                                     Fixture("cascades/sc_cvpr_2012_to_opencv_new_format.xml", 1291))));
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/softcascade/test/test_main.cpp b/modules/softcascade/test/test_main.cpp
index ab0e8615bd..d3999d4833 100644
--- a/modules/softcascade/test/test_main.cpp
+++ b/modules/softcascade/test/test_main.cpp
@@ -42,4 +42,4 @@
 
 #include "test_precomp.hpp"
 
-CV_TEST_MAIN("cv")
\ No newline at end of file
+CV_TEST_MAIN("cv")
diff --git a/modules/softcascade/test/test_precomp.cpp b/modules/softcascade/test/test_precomp.cpp
deleted file mode 100644
index 278cdd03cf..0000000000
--- a/modules/softcascade/test/test_precomp.cpp
+++ /dev/null
@@ -1,43 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
-// Copyright (C) 2008-2013, Willow Garage Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and / or other materials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "test_precomp.hpp"
diff --git a/modules/softcascade/test/test_precomp.hpp b/modules/softcascade/test/test_precomp.hpp
index 80bff6536d..4a39744783 100644
--- a/modules/softcascade/test/test_precomp.hpp
+++ b/modules/softcascade/test/test_precomp.hpp
@@ -51,10 +51,11 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
-# include "opencv2/ts.hpp"
-# include "opencv2/softcascade.hpp"
-# include "opencv2/imgproc.hpp"
-# include "opencv2/highgui.hpp"
-# include "utility.hpp"
+#include "opencv2/ts.hpp"
+#include "opencv2/softcascade.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "cvconfig.h"
+#include "utility.hpp"
 
 #endif
diff --git a/modules/softcascade/test/test_softcascade.cpp b/modules/softcascade/test/test_softcascade.cpp
index 9004264fcb..59d5c3c4a4 100644
--- a/modules/softcascade/test/test_softcascade.cpp
+++ b/modules/softcascade/test/test_softcascade.cpp
@@ -138,4 +138,4 @@ TEST(SoftCascadeDetector, detectEmptyRoi)
     cascade.detect(colored, cv::Mat::zeros(colored.size(), CV_8UC1), objects);
 
     ASSERT_EQ(0, (int)objects.size());
-}
\ No newline at end of file
+}
diff --git a/modules/softcascade/test/test_training.cpp b/modules/softcascade/test/test_training.cpp
index ef20323d25..5ade81681e 100644
--- a/modules/softcascade/test/test_training.cpp
+++ b/modules/softcascade/test/test_training.cpp
@@ -53,7 +53,7 @@ namespace {
 
 using namespace cv::softcascade;
 
-typedef vector<string> svector;
+typedef vector<cv::String> svector;
 class ScaledDataset : public Dataset
 {
 public:
@@ -157,4 +157,4 @@ TEST(SoftCascade, training)
     ASSERT_FALSE(fn.empty());
 }
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/softcascade/test/utility.cpp b/modules/softcascade/test/utility.cpp
index cb3b1fbf49..b308f416c5 100644
--- a/modules/softcascade/test/utility.cpp
+++ b/modules/softcascade/test/utility.cpp
@@ -75,7 +75,7 @@ void DeviceManager::load(int i)
     if (i < 0 || i >= getCudaEnabledDeviceCount())
     {
         msg << "Incorrect device number - " << i;
-        CV_Error(CV_StsBadArg, msg.str());
+        CV_Error(cv::Error::StsBadArg, msg.str());
     }
 
     DeviceInfo info(i);
@@ -83,7 +83,7 @@ void DeviceManager::load(int i)
     if (!info.isCompatible())
     {
         msg << "Device " << i << " [" << info.name() << "] is NOT compatible with current GPU module build";
-        CV_Error(CV_StsBadArg, msg.str());
+        CV_Error(cv::Error::StsBadArg, msg.str());
     }
 
     devices_.push_back(info);
diff --git a/modules/softcascade/test/utility.hpp b/modules/softcascade/test/utility.hpp
index 9849b525e2..44b0fabf29 100644
--- a/modules/softcascade/test/utility.hpp
+++ b/modules/softcascade/test/utility.hpp
@@ -43,7 +43,7 @@
 #define __OPENCV_SOFTCASCADE_TEST_UTILITY_HPP__
 
 #include "opencv2/core.hpp"
-#include "opencv2/core/gpumat.hpp"
+#include "opencv2/core/gpu.hpp"
 #include "opencv2/ts.hpp"
 
 //////////////////////////////////////////////////////////////////////
diff --git a/modules/stitching/CMakeLists.txt b/modules/stitching/CMakeLists.txt
index 5d48bd378f..389285a569 100644
--- a/modules/stitching/CMakeLists.txt
+++ b/modules/stitching/CMakeLists.txt
@@ -1,3 +1,3 @@
 set(the_description "Images stitching")
-ocv_define_module(stitching opencv_imgproc opencv_features2d opencv_calib3d opencv_objdetect OPTIONAL opencv_gpu opencv_nonfree)
-
+ocv_define_module(stitching opencv_imgproc opencv_features2d opencv_calib3d opencv_objdetect
+                  OPTIONAL opencv_gpu opencv_gpuarithm opencv_gpufilters opencv_gpufeatures2d opencv_nonfree)
diff --git a/modules/stitching/doc/autocalib.rst b/modules/stitching/doc/autocalib.rst
index a9c0c35fdd..021e7b394d 100644
--- a/modules/stitching/doc/autocalib.rst
+++ b/modules/stitching/doc/autocalib.rst
@@ -30,4 +30,3 @@ Estimates focal lengths for each given camera.
     :param pairwise_matches: Matches between all image pairs.
 
     :param focals: Estimated focal lengths for each camera.
-
diff --git a/modules/stitching/doc/exposure_compensation.rst b/modules/stitching/doc/exposure_compensation.rst
index d27701e25b..eff98c01ca 100644
--- a/modules/stitching/doc/exposure_compensation.rst
+++ b/modules/stitching/doc/exposure_compensation.rst
@@ -108,4 +108,3 @@ Exposure compensator which tries to remove exposure related artifacts by adjusti
     };
 
 .. seealso:: :ocv:class:`detail::ExposureCompensator`
-
diff --git a/modules/stitching/doc/high_level.rst b/modules/stitching/doc/high_level.rst
index dec3362d30..955f8f2481 100644
--- a/modules/stitching/doc/high_level.rst
+++ b/modules/stitching/doc/high_level.rst
@@ -88,6 +88,11 @@ High level image stitcher. It's possible to use this class without being aware o
         /* hidden */
     };
 
+.. note::
+
+   * A basic example on image stitching can be found at opencv_source_code/samples/cpp/stitching.cpp
+   * A detailed example on image stitching can be found at opencv_source_code/samples/cpp/stitching_detailed.cpp
+
 Stitcher::createDefault
 -----------------------
 Creates a stitcher with the default parameters.
@@ -203,4 +208,3 @@ Spherical warper factory class. ::
     };
 
 .. seealso:: :ocv:class:`detail::SphericalWarper`
-
diff --git a/modules/stitching/doc/motion_estimation.rst b/modules/stitching/doc/motion_estimation.rst
index 6cda27522d..5859f3e02f 100644
--- a/modules/stitching/doc/motion_estimation.rst
+++ b/modules/stitching/doc/motion_estimation.rst
@@ -18,12 +18,11 @@ Rotation estimator base class. It takes features of all images, pairwise matches
     public:
         virtual ~Estimator() {}
 
-        void operator ()(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
-                         std::vector<CameraParams> &cameras)
-            { estimate(features, pairwise_matches, cameras); }
+        bool operator ()(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
+                         std::vector<CameraParams> &cameras);
 
     protected:
-        virtual void estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
+        virtual bool estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
                               std::vector<CameraParams> &cameras) = 0;
     };
 
@@ -32,7 +31,7 @@ detail::Estimator::operator()
 
 Estimates camera parameters.
 
-.. ocv:function:: detail::Estimator::operator ()(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches, std::vector<CameraParams> &cameras)
+.. ocv:function:: bool detail::Estimator::operator ()(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches, std::vector<CameraParams> &cameras)
 
     :param features: Features of images
 
@@ -40,12 +39,14 @@ Estimates camera parameters.
 
     :param cameras: Estimated camera parameters
 
+    :return: True in case of success, false otherwise
+
 detail::Estimator::estimate
 ---------------------------
 
 This method must implement camera parameters estimation logic in order to make the wrapper `detail::Estimator::operator()`_ work.
 
-.. ocv:function:: void detail::Estimator::estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches, std::vector<CameraParams> &cameras)
+.. ocv:function:: bool detail::Estimator::estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches, std::vector<CameraParams> &cameras)
 
     :param features: Features of images
 
@@ -53,6 +54,8 @@ This method must implement camera parameters estimation logic in order to make t
 
     :param cameras: Estimated camera parameters
 
+    :return: True in case of success, false otherwise
+
 detail::HomographyBasedEstimator
 --------------------------------
 .. ocv:class:: detail::HomographyBasedEstimator : public detail::Estimator
diff --git a/modules/stitching/doc/stitching.rst b/modules/stitching/doc/stitching.rst
index 983b310e78..f071a23501 100644
--- a/modules/stitching/doc/stitching.rst
+++ b/modules/stitching/doc/stitching.rst
@@ -15,5 +15,3 @@ stitching. Images stitching
     seam_estimation
     exposure_compensation
     blenders
-
-
diff --git a/modules/stitching/doc/warpers.rst b/modules/stitching/doc/warpers.rst
index b6826a01a6..1025ffa0cb 100644
--- a/modules/stitching/doc/warpers.rst
+++ b/modules/stitching/doc/warpers.rst
@@ -260,4 +260,4 @@ Construct an instance of the cylindrical warper class.
 
 .. ocv:function:: void detail::CylindricalWarper::CylindricalWarper(float scale)
 
-    :param scale: Projected image scale multiplier
\ No newline at end of file
+    :param scale: Projected image scale multiplier
diff --git a/modules/stitching/include/opencv2/stitching.hpp b/modules/stitching/include/opencv2/stitching.hpp
index 3ccd2878a9..2c48f2f2aa 100644
--- a/modules/stitching/include/opencv2/stitching.hpp
+++ b/modules/stitching/include/opencv2/stitching.hpp
@@ -59,7 +59,13 @@ class CV_EXPORTS Stitcher
 {
 public:
     enum { ORIG_RESOL = -1 };
-    enum Status { OK, ERR_NEED_MORE_IMGS };
+    enum Status
+    {
+        OK = 0,
+        ERR_NEED_MORE_IMGS = 1,
+        ERR_HOMOGRAPHY_EST_FAIL = 2,
+        ERR_CAMERA_PARAMS_ADJUST_FAIL = 3
+    };
 
     // Creates stitcher with default parameters
     static Stitcher createDefault(bool try_use_gpu = false);
@@ -138,7 +144,7 @@ private:
     Stitcher() {}
 
     Status matchImages();
-    void estimateCameraParams();
+    Status estimateCameraParams();
 
     double registr_resol_;
     double seam_est_resol_;
diff --git a/modules/stitching/include/opencv2/stitching/detail/matchers.hpp b/modules/stitching/include/opencv2/stitching/detail/matchers.hpp
index d6e3b78b1f..0b24e784f5 100644
--- a/modules/stitching/include/opencv2/stitching/detail/matchers.hpp
+++ b/modules/stitching/include/opencv2/stitching/detail/matchers.hpp
@@ -47,8 +47,9 @@
 #include "opencv2/features2d.hpp"
 
 #include "opencv2/opencv_modules.hpp"
-#ifdef HAVE_OPENCV_GPU
-#include "opencv2/gpu.hpp"
+
+#ifdef HAVE_OPENCV_NONFREE
+#  include "opencv2/nonfree/gpu.hpp"
 #endif
 
 namespace cv {
@@ -103,7 +104,7 @@ private:
 };
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_NONFREE
 class CV_EXPORTS SurfFeaturesFinderGpu : public FeaturesFinder
 {
 public:
diff --git a/modules/stitching/include/opencv2/stitching/detail/motion_estimators.hpp b/modules/stitching/include/opencv2/stitching/detail/motion_estimators.hpp
index 0d08b09fc1..c0e446c60e 100644
--- a/modules/stitching/include/opencv2/stitching/detail/motion_estimators.hpp
+++ b/modules/stitching/include/opencv2/stitching/detail/motion_estimators.hpp
@@ -56,12 +56,14 @@ class CV_EXPORTS Estimator
 public:
     virtual ~Estimator() {}
 
-    void operator ()(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
+    bool operator ()(const std::vector<ImageFeatures> &features,
+                     const std::vector<MatchesInfo> &pairwise_matches,
                      std::vector<CameraParams> &cameras)
-        { estimate(features, pairwise_matches, cameras); }
+        { return estimate(features, pairwise_matches, cameras); }
 
 protected:
-    virtual void estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
+    virtual bool estimate(const std::vector<ImageFeatures> &features,
+                          const std::vector<MatchesInfo> &pairwise_matches,
                           std::vector<CameraParams> &cameras) = 0;
 };
 
@@ -73,8 +75,9 @@ public:
         : is_focals_estimated_(is_focals_estimated) {}
 
 private:
-    void estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
-                  std::vector<CameraParams> &cameras);
+    virtual bool estimate(const std::vector<ImageFeatures> &features,
+                          const std::vector<MatchesInfo> &pairwise_matches,
+                          std::vector<CameraParams> &cameras);
 
     bool is_focals_estimated_;
 };
@@ -93,8 +96,8 @@ public:
     double confThresh() const { return conf_thresh_; }
     void setConfThresh(double conf_thresh) { conf_thresh_ = conf_thresh; }
 
-    CvTermCriteria termCriteria() { return term_criteria_; }
-    void setTermCriteria(const CvTermCriteria& term_criteria) { term_criteria_ = term_criteria; }
+    TermCriteria termCriteria() { return term_criteria_; }
+    void setTermCriteria(const TermCriteria& term_criteria) { term_criteria_ = term_criteria; }
 
 protected:
     BundleAdjusterBase(int num_params_per_cam, int num_errs_per_measurement)
@@ -103,11 +106,11 @@ protected:
     {
         setRefinementMask(Mat::ones(3, 3, CV_8U));
         setConfThresh(1.);
-        setTermCriteria(cvTermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 1000, DBL_EPSILON));
+        setTermCriteria(TermCriteria(TermCriteria::EPS + TermCriteria::COUNT, 1000, DBL_EPSILON));
     }
 
     // Runs bundle adjustment
-    virtual void estimate(const std::vector<ImageFeatures> &features,
+    virtual bool estimate(const std::vector<ImageFeatures> &features,
                           const std::vector<MatchesInfo> &pairwise_matches,
                           std::vector<CameraParams> &cameras);
 
@@ -132,7 +135,7 @@ protected:
     double conf_thresh_;
 
     //Levenberg–Marquardt algorithm termination criteria
-    CvTermCriteria term_criteria_;
+    TermCriteria term_criteria_;
 
     // Camera parameters matrix (CV_64F)
     Mat cam_params_;
@@ -190,14 +193,17 @@ void CV_EXPORTS waveCorrect(std::vector<Mat> &rmats, WaveCorrectKind kind);
 // Auxiliary functions
 
 // Returns matches graph representation in DOT language
-std::string CV_EXPORTS matchesGraphAsString(std::vector<std::string> &pathes, std::vector<MatchesInfo> &pairwise_matches,
+String CV_EXPORTS matchesGraphAsString(std::vector<String> &pathes, std::vector<MatchesInfo> &pairwise_matches,
                                             float conf_threshold);
 
-std::vector<int> CV_EXPORTS leaveBiggestComponent(std::vector<ImageFeatures> &features, std::vector<MatchesInfo> &pairwise_matches,
-                                                  float conf_threshold);
+std::vector<int> CV_EXPORTS leaveBiggestComponent(
+        std::vector<ImageFeatures> &features,
+        std::vector<MatchesInfo> &pairwise_matches,
+        float conf_threshold);
 
-void CV_EXPORTS findMaxSpanningTree(int num_images, const std::vector<MatchesInfo> &pairwise_matches,
-                                    Graph &span_tree, std::vector<int> &centers);
+void CV_EXPORTS findMaxSpanningTree(
+        int num_images, const std::vector<MatchesInfo> &pairwise_matches,
+        Graph &span_tree, std::vector<int> &centers);
 
 } // namespace detail
 } // namespace cv
diff --git a/modules/stitching/include/opencv2/stitching/detail/util.hpp b/modules/stitching/include/opencv2/stitching/detail/util.hpp
index 5c00fa343a..6f334b9ad5 100644
--- a/modules/stitching/include/opencv2/stitching/detail/util.hpp
+++ b/modules/stitching/include/opencv2/stitching/detail/util.hpp
@@ -56,7 +56,7 @@
   #include <android/log.h>
   #define LOG_STITCHING_MSG(msg) \
     do { \
-        std::stringstream _os; \
+        Stringstream _os; \
         _os << msg; \
        __android_log_print(ANDROID_LOG_DEBUG, "STITCHING", "%s", _os.str().c_str()); \
     } while(0);
diff --git a/modules/stitching/include/opencv2/stitching/detail/warpers.hpp b/modules/stitching/include/opencv2/stitching/detail/warpers.hpp
index 5a6393a69b..5cd8f5141c 100644
--- a/modules/stitching/include/opencv2/stitching/detail/warpers.hpp
+++ b/modules/stitching/include/opencv2/stitching/detail/warpers.hpp
@@ -44,11 +44,9 @@
 #define __OPENCV_STITCHING_WARPERS_HPP__
 
 #include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/opencv_modules.hpp"
-#ifdef HAVE_OPENCV_GPU
-# include "opencv2/gpu.hpp"
-#endif
 
 namespace cv {
 namespace detail {
@@ -329,7 +327,7 @@ public:
 };
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUWARPING
 class CV_EXPORTS PlaneWarperGpu : public PlaneWarper
 {
 public:
@@ -458,7 +456,7 @@ struct SphericalPortraitProjector : ProjectorBase
 
 // Projects image onto unit sphere with origin at (0, 0, 0).
 // Poles are located NOT at (0, -1, 0) and (0, 1, 0) points, BUT at (1, 0, 0) and (-1, 0, 0) points.
-class SphericalPortraitWarper : public RotationWarperBase<SphericalPortraitProjector>
+class CV_EXPORTS SphericalPortraitWarper : public RotationWarperBase<SphericalPortraitProjector>
 {
 public:
     SphericalPortraitWarper(float scale) { projector_.scale = scale; }
@@ -474,7 +472,7 @@ struct CylindricalPortraitProjector : ProjectorBase
 };
 
 
-class CylindricalPortraitWarper : public RotationWarperBase<CylindricalPortraitProjector>
+class CV_EXPORTS CylindricalPortraitWarper : public RotationWarperBase<CylindricalPortraitProjector>
 {
 public:
     CylindricalPortraitWarper(float scale) { projector_.scale = scale; }
@@ -493,7 +491,7 @@ struct PlanePortraitProjector : ProjectorBase
 };
 
 
-class PlanePortraitWarper : public RotationWarperBase<PlanePortraitProjector>
+class CV_EXPORTS PlanePortraitWarper : public RotationWarperBase<PlanePortraitProjector>
 {
 public:
     PlanePortraitWarper(float scale) { projector_.scale = scale; }
diff --git a/modules/stitching/include/opencv2/stitching/detail/warpers_inl.hpp b/modules/stitching/include/opencv2/stitching/detail/warpers_inl.hpp
index c5f79f1a43..144e9e32d4 100644
--- a/modules/stitching/include/opencv2/stitching/detail/warpers_inl.hpp
+++ b/modules/stitching/include/opencv2/stitching/detail/warpers_inl.hpp
@@ -45,6 +45,7 @@
 
 #include "opencv2/core.hpp"
 #include "warpers.hpp" // Make your IDE see declarations
+#include <limits>
 
 namespace cv {
 namespace detail {
diff --git a/modules/stitching/include/opencv2/stitching/warpers.hpp b/modules/stitching/include/opencv2/stitching/warpers.hpp
index 7475d1304a..37cf84b7c2 100644
--- a/modules/stitching/include/opencv2/stitching/warpers.hpp
+++ b/modules/stitching/include/opencv2/stitching/warpers.hpp
@@ -58,33 +58,33 @@ public:
 class PlaneWarper : public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::PlaneWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::PlaneWarper>(scale); }
 };
 
 
 class CylindricalWarper: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::CylindricalWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::CylindricalWarper>(scale); }
 };
 
 
 class SphericalWarper: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::SphericalWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::SphericalWarper>(scale); }
 };
 
 class FisheyeWarper : public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::FisheyeWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::FisheyeWarper>(scale); }
 };
 
 class StereographicWarper: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::StereographicWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::StereographicWarper>(scale); }
 };
 
 class CompressedRectilinearWarper: public WarperCreator
@@ -95,7 +95,7 @@ public:
     {
         a = A; b = B;
     }
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::CompressedRectilinearWarper(scale, a, b); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::CompressedRectilinearWarper>(scale, a, b); }
 };
 
 class CompressedRectilinearPortraitWarper: public WarperCreator
@@ -106,7 +106,7 @@ public:
     {
         a = A; b = B;
     }
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::CompressedRectilinearPortraitWarper(scale, a, b); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::CompressedRectilinearPortraitWarper>(scale, a, b); }
 };
 
 class PaniniWarper: public WarperCreator
@@ -117,7 +117,7 @@ public:
     {
         a = A; b = B;
     }
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::PaniniWarper(scale, a, b); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::PaniniWarper>(scale, a, b); }
 };
 
 class PaniniPortraitWarper: public WarperCreator
@@ -128,42 +128,42 @@ public:
     {
         a = A; b = B;
     }
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::PaniniPortraitWarper(scale, a, b); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::PaniniPortraitWarper>(scale, a, b); }
 };
 
 class MercatorWarper: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::MercatorWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::MercatorWarper>(scale); }
 };
 
 class TransverseMercatorWarper: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::TransverseMercatorWarper(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::TransverseMercatorWarper>(scale); }
 };
 
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUWARPING
 class PlaneWarperGpu: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::PlaneWarperGpu(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::PlaneWarperGpu>(scale); }
 };
 
 
 class CylindricalWarperGpu: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::CylindricalWarperGpu(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::CylindricalWarperGpu>(scale); }
 };
 
 
 class SphericalWarperGpu: public WarperCreator
 {
 public:
-    Ptr<detail::RotationWarper> create(float scale) const { return new detail::SphericalWarperGpu(scale); }
+    Ptr<detail::RotationWarper> create(float scale) const { return makePtr<detail::SphericalWarperGpu>(scale); }
 };
 #endif
 
diff --git a/modules/stitching/perf/perf_precomp.cpp b/modules/stitching/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/stitching/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/stitching/perf/perf_stich.cpp b/modules/stitching/perf/perf_stich.cpp
index 5afe217b02..8cf89651fe 100644
--- a/modules/stitching/perf/perf_stich.cpp
+++ b/modules/stitching/perf/perf_stich.cpp
@@ -1,6 +1,5 @@
 #include "perf_precomp.hpp"
 #include "opencv2/highgui.hpp"
-#include "opencv2/core/internal.hpp"
 #include "opencv2/flann.hpp"
 #include "opencv2/opencv_modules.hpp"
 
@@ -35,12 +34,12 @@ PERF_TEST_P(stitch, a123, TEST_DETECTORS)
     imgs.push_back( imread( getDataPath("stitching/a3.png") ) );
 
     Ptr<detail::FeaturesFinder> featuresFinder = GetParam() == "orb"
-            ? (detail::FeaturesFinder*)new detail::OrbFeaturesFinder()
-            : (detail::FeaturesFinder*)new detail::SurfFeaturesFinder();
+            ? Ptr<detail::FeaturesFinder>(new detail::OrbFeaturesFinder())
+            : Ptr<detail::FeaturesFinder>(new detail::SurfFeaturesFinder());
 
     Ptr<detail::FeaturesMatcher> featuresMatcher = GetParam() == "orb"
-            ? new detail::BestOf2NearestMatcher(false, ORB_MATCH_CONFIDENCE)
-            : new detail::BestOf2NearestMatcher(false, SURF_MATCH_CONFIDENCE);
+            ? makePtr<detail::BestOf2NearestMatcher>(false, ORB_MATCH_CONFIDENCE)
+            : makePtr<detail::BestOf2NearestMatcher>(false, SURF_MATCH_CONFIDENCE);
 
     declare.time(30 * 20).iterations(20);
 
@@ -49,7 +48,7 @@ PERF_TEST_P(stitch, a123, TEST_DETECTORS)
         Stitcher stitcher = Stitcher::createDefault();
         stitcher.setFeaturesFinder(featuresFinder);
         stitcher.setFeaturesMatcher(featuresMatcher);
-        stitcher.setWarper(new SphericalWarper());
+        stitcher.setWarper(makePtr<SphericalWarper>());
         stitcher.setRegistrationResol(WORK_MEGAPIX);
 
         startTimer();
@@ -73,12 +72,12 @@ PERF_TEST_P(stitch, b12, TEST_DETECTORS)
     imgs.push_back( imread( getDataPath("stitching/b2.png") ) );
 
     Ptr<detail::FeaturesFinder> featuresFinder = GetParam() == "orb"
-            ? (detail::FeaturesFinder*)new detail::OrbFeaturesFinder()
-            : (detail::FeaturesFinder*)new detail::SurfFeaturesFinder();
+            ? Ptr<detail::FeaturesFinder>(new detail::OrbFeaturesFinder())
+            : Ptr<detail::FeaturesFinder>(new detail::SurfFeaturesFinder());
 
     Ptr<detail::FeaturesMatcher> featuresMatcher = GetParam() == "orb"
-            ? new detail::BestOf2NearestMatcher(false, ORB_MATCH_CONFIDENCE)
-            : new detail::BestOf2NearestMatcher(false, SURF_MATCH_CONFIDENCE);
+            ? makePtr<detail::BestOf2NearestMatcher>(false, ORB_MATCH_CONFIDENCE)
+            : makePtr<detail::BestOf2NearestMatcher>(false, SURF_MATCH_CONFIDENCE);
 
     declare.time(30 * 20).iterations(20);
 
@@ -87,7 +86,7 @@ PERF_TEST_P(stitch, b12, TEST_DETECTORS)
         Stitcher stitcher = Stitcher::createDefault();
         stitcher.setFeaturesFinder(featuresFinder);
         stitcher.setFeaturesMatcher(featuresMatcher);
-        stitcher.setWarper(new SphericalWarper());
+        stitcher.setWarper(makePtr<SphericalWarper>());
         stitcher.setRegistrationResol(WORK_MEGAPIX);
 
         startTimer();
@@ -115,13 +114,13 @@ PERF_TEST_P( match, bestOf2Nearest, TEST_DETECTORS)
     Ptr<detail::FeaturesMatcher> matcher;
     if (GetParam() == "surf")
     {
-        finder = new detail::SurfFeaturesFinder();
-        matcher = new detail::BestOf2NearestMatcher(false, SURF_MATCH_CONFIDENCE);
+        finder = makePtr<detail::SurfFeaturesFinder>();
+        matcher = makePtr<detail::BestOf2NearestMatcher>(false, SURF_MATCH_CONFIDENCE);
     }
     else if (GetParam() == "orb")
     {
-        finder = new detail::OrbFeaturesFinder();
-        matcher = new detail::BestOf2NearestMatcher(false, ORB_MATCH_CONFIDENCE);
+        finder = makePtr<detail::OrbFeaturesFinder>();
+        matcher = makePtr<detail::BestOf2NearestMatcher>(false, ORB_MATCH_CONFIDENCE);
     }
     else
     {
@@ -170,13 +169,13 @@ PERF_TEST_P( matchVector, bestOf2NearestVectorFeatures, testing::Combine(
     int featuresVectorSize = get<1>(GetParam());
     if (detectorName == "surf")
     {
-        finder = new detail::SurfFeaturesFinder();
-        matcher = new detail::BestOf2NearestMatcher(false, SURF_MATCH_CONFIDENCE);
+        finder = makePtr<detail::SurfFeaturesFinder>();
+        matcher = makePtr<detail::BestOf2NearestMatcher>(false, SURF_MATCH_CONFIDENCE);
     }
     else if (detectorName == "orb")
     {
-        finder = new detail::OrbFeaturesFinder();
-        matcher = new detail::BestOf2NearestMatcher(false, ORB_MATCH_CONFIDENCE);
+        finder = makePtr<detail::OrbFeaturesFinder>();
+        matcher = makePtr<detail::BestOf2NearestMatcher>(false, ORB_MATCH_CONFIDENCE);
     }
     else
     {
diff --git a/modules/stitching/src/blenders.cpp b/modules/stitching/src/blenders.cpp
index dc7d1da618..254c6a1490 100644
--- a/modules/stitching/src/blenders.cpp
+++ b/modules/stitching/src/blenders.cpp
@@ -50,13 +50,13 @@ static const float WEIGHT_EPS = 1e-5f;
 Ptr<Blender> Blender::createDefault(int type, bool try_gpu)
 {
     if (type == NO)
-        return new Blender();
+        return makePtr<Blender>();
     if (type == FEATHER)
-        return new FeatherBlender();
+        return makePtr<FeatherBlender>();
     if (type == MULTI_BAND)
-        return new MultiBandBlender(try_gpu);
-    CV_Error(CV_StsBadArg, "unsupported blending method");
-    return NULL;
+        return makePtr<MultiBandBlender>(try_gpu);
+    CV_Error(Error::StsBadArg, "unsupported blending method");
+    return Ptr<Blender>();
 }
 
 
@@ -187,12 +187,14 @@ Rect FeatherBlender::createWeightMaps(const std::vector<Mat> &masks, const std::
 MultiBandBlender::MultiBandBlender(int try_gpu, int num_bands, int weight_type)
 {
     setNumBands(num_bands);
-#ifdef HAVE_OPENCV_GPU
+
+#if defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUWARPING)
     can_use_gpu_ = try_gpu && gpu::getCudaEnabledDeviceCount();
 #else
-    (void)try_gpu;
+    (void) try_gpu;
     can_use_gpu_ = false;
 #endif
+
     CV_Assert(weight_type == CV_32F || weight_type == CV_16S);
     weight_type_ = weight_type;
 }
@@ -425,7 +427,7 @@ void normalizeUsingWeightMap(const Mat& weight, Mat& src)
 void createWeightMap(const Mat &mask, float sharpness, Mat &weight)
 {
     CV_Assert(mask.type() == CV_8U);
-    distanceTransform(mask, weight, CV_DIST_L1, 3);
+    distanceTransform(mask, weight, DIST_L1, 3);
     threshold(weight * sharpness, weight, 1.f, 1.f, THRESH_TRUNC);
 }
 
@@ -489,7 +491,7 @@ void createLaplacePyr(const Mat &img, int num_levels, std::vector<Mat> &pyr)
 
 void createLaplacePyrGpu(const Mat &img, int num_levels, std::vector<Mat> &pyr)
 {
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUWARPING)
     pyr.resize(num_levels + 1);
 
     std::vector<gpu::GpuMat> gpu_pyr(num_levels + 1);
@@ -529,7 +531,7 @@ void restoreImageFromLaplacePyr(std::vector<Mat> &pyr)
 
 void restoreImageFromLaplacePyrGpu(std::vector<Mat> &pyr)
 {
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUWARPING)
     if (pyr.empty())
         return;
 
diff --git a/modules/stitching/src/exposure_compensate.cpp b/modules/stitching/src/exposure_compensate.cpp
index d2479b7107..78ce6d3714 100644
--- a/modules/stitching/src/exposure_compensate.cpp
+++ b/modules/stitching/src/exposure_compensate.cpp
@@ -48,13 +48,13 @@ namespace detail {
 Ptr<ExposureCompensator> ExposureCompensator::createDefault(int type)
 {
     if (type == NO)
-        return new NoExposureCompensator();
+        return makePtr<NoExposureCompensator>();
     if (type == GAIN)
-        return new GainCompensator();
+        return makePtr<GainCompensator>();
     if (type == GAIN_BLOCKS)
-        return new BlocksGainCompensator();
-    CV_Error(CV_StsBadArg, "unsupported exposure compensation method");
-    return NULL;
+        return makePtr<BlocksGainCompensator>();
+    CV_Error(Error::StsBadArg, "unsupported exposure compensation method");
+    return Ptr<ExposureCompensator>();
 }
 
 
diff --git a/modules/stitching/src/matchers.cpp b/modules/stitching/src/matchers.cpp
index 4351e82e4f..0ad0c1f1d0 100644
--- a/modules/stitching/src/matchers.cpp
+++ b/modules/stitching/src/matchers.cpp
@@ -44,10 +44,7 @@
 
 using namespace cv;
 using namespace cv::detail;
-
-#ifdef HAVE_OPENCV_GPU
 using namespace cv::gpu;
-#endif
 
 #ifdef HAVE_OPENCV_NONFREE
 #include "opencv2/nonfree.hpp"
@@ -65,21 +62,17 @@ struct DistIdxPair
 };
 
 
-struct MatchPairsBody
+struct MatchPairsBody : ParallelLoopBody
 {
-    MatchPairsBody(const MatchPairsBody& other)
-            : matcher(other.matcher), features(other.features),
-              pairwise_matches(other.pairwise_matches), near_pairs(other.near_pairs) {}
-
     MatchPairsBody(FeaturesMatcher &_matcher, const std::vector<ImageFeatures> &_features,
                    std::vector<MatchesInfo> &_pairwise_matches, std::vector<std::pair<int,int> > &_near_pairs)
             : matcher(_matcher), features(_features),
               pairwise_matches(_pairwise_matches), near_pairs(_near_pairs) {}
 
-    void operator ()(const BlockedRange &r) const
+    void operator ()(const Range &r) const
     {
         const int num_images = static_cast<int>(features.size());
-        for (int i = r.begin(); i < r.end(); ++i)
+        for (int i = r.start; i < r.end; ++i)
         {
             int from = near_pairs[i].first;
             int to = near_pairs[i].second;
@@ -132,7 +125,7 @@ private:
     float match_conf_;
 };
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUFEATURES2D
 class GpuMatcher : public FeaturesMatcher
 {
 public:
@@ -162,8 +155,8 @@ void CpuMatcher::match(const ImageFeatures &features1, const ImageFeatures &feat
 
     matches_info.matches.clear();
 
-    Ptr<flann::IndexParams> indexParams = new flann::KDTreeIndexParams();
-    Ptr<flann::SearchParams> searchParams = new flann::SearchParams();
+    Ptr<flann::IndexParams> indexParams = makePtr<flann::KDTreeIndexParams>();
+    Ptr<flann::SearchParams> searchParams = makePtr<flann::SearchParams>();
 
     if (features2.descriptors.depth() == CV_8U)
     {
@@ -207,7 +200,7 @@ void CpuMatcher::match(const ImageFeatures &features1, const ImageFeatures &feat
     LOG("1->2 & 2->1 matches: " << matches_info.matches.size() << endl);
 }
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUFEATURES2D
 void GpuMatcher::match(const ImageFeatures &features1, const ImageFeatures &features2, MatchesInfo& matches_info)
 {
     matches_info.matches.clear();
@@ -321,8 +314,8 @@ SurfFeaturesFinder::SurfFeaturesFinder(double hess_thresh, int num_octaves, int
     if (num_octaves_descr == num_octaves && num_layers_descr == num_layers)
     {
         surf = Algorithm::create<Feature2D>("Feature2D.SURF");
-        if( surf.empty() )
-            CV_Error( CV_StsNotImplemented, "OpenCV was built without SURF support" );
+        if( !surf )
+            CV_Error( Error::StsNotImplemented, "OpenCV was built without SURF support" );
         surf->set("hessianThreshold", hess_thresh);
         surf->set("nOctaves", num_octaves);
         surf->set("nOctaveLayers", num_layers);
@@ -332,8 +325,8 @@ SurfFeaturesFinder::SurfFeaturesFinder(double hess_thresh, int num_octaves, int
         detector_ = Algorithm::create<FeatureDetector>("Feature2D.SURF");
         extractor_ = Algorithm::create<DescriptorExtractor>("Feature2D.SURF");
 
-        if( detector_.empty() || extractor_.empty() )
-            CV_Error( CV_StsNotImplemented, "OpenCV was built without SURF support" );
+        if( !detector_ || !extractor_ )
+            CV_Error( Error::StsNotImplemented, "OpenCV was built without SURF support" );
 
         detector_->set("hessianThreshold", hess_thresh);
         detector_->set("nOctaves", num_octaves);
@@ -347,9 +340,16 @@ SurfFeaturesFinder::SurfFeaturesFinder(double hess_thresh, int num_octaves, int
 void SurfFeaturesFinder::find(const Mat &image, ImageFeatures &features)
 {
     Mat gray_image;
-    CV_Assert(image.type() == CV_8UC3);
-    cvtColor(image, gray_image, CV_BGR2GRAY);
-    if (surf.empty())
+    CV_Assert((image.type() == CV_8UC3) || (image.type() == CV_8UC1));
+    if(image.type() == CV_8UC3)
+    {
+        cvtColor(image, gray_image, COLOR_BGR2GRAY);
+    }
+    else
+    {
+        gray_image = image;
+    }
+    if (!surf)
     {
         detector_->detect(gray_image, features.keypoints);
         extractor_->compute(gray_image, features.keypoints, features.descriptors);
@@ -365,7 +365,7 @@ void SurfFeaturesFinder::find(const Mat &image, ImageFeatures &features)
 OrbFeaturesFinder::OrbFeaturesFinder(Size _grid_size, int n_features, float scaleFactor, int nlevels)
 {
     grid_size = _grid_size;
-    orb = new ORB(n_features * (99 + grid_size.area())/100/grid_size.area(), scaleFactor, nlevels);
+    orb = makePtr<ORB>(n_features * (99 + grid_size.area())/100/grid_size.area(), scaleFactor, nlevels);
 }
 
 void OrbFeaturesFinder::find(const Mat &image, ImageFeatures &features)
@@ -375,13 +375,13 @@ void OrbFeaturesFinder::find(const Mat &image, ImageFeatures &features)
     CV_Assert((image.type() == CV_8UC3) || (image.type() == CV_8UC4) || (image.type() == CV_8UC1));
 
     if (image.type() == CV_8UC3) {
-        cvtColor(image, gray_image, CV_BGR2GRAY);
+        cvtColor(image, gray_image, COLOR_BGR2GRAY);
     } else if (image.type() == CV_8UC4) {
-        cvtColor(image, gray_image, CV_BGRA2GRAY);
+        cvtColor(image, gray_image, COLOR_BGRA2GRAY);
     } else if (image.type() == CV_8UC1) {
         gray_image=image;
     } else {
-        CV_Error(CV_StsUnsupportedFormat, "");
+        CV_Error(Error::StsUnsupportedFormat, "");
     }
 
     if (grid_size.area() == 1)
@@ -428,7 +428,7 @@ void OrbFeaturesFinder::find(const Mat &image, ImageFeatures &features)
     }
 }
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_NONFREE
 SurfFeaturesFinderGpu::SurfFeaturesFinderGpu(double hess_thresh, int num_octaves, int num_layers,
                                              int num_octaves_descr, int num_layers_descr)
 {
@@ -450,7 +450,7 @@ void SurfFeaturesFinderGpu::find(const Mat &image, ImageFeatures &features)
     image_.upload(image);
 
     ensureSizeIsEnough(image.size(), CV_8UC1, gray_image_);
-    cvtColor(image_, gray_image_, CV_BGR2GRAY);
+    cvtColor(image_, gray_image_, COLOR_BGR2GRAY);
 
     surf_.nOctaves = num_octaves_;
     surf_.nOctaveLayers = num_layers_;
@@ -518,9 +518,9 @@ void FeaturesMatcher::operator ()(const std::vector<ImageFeatures> &features, st
     MatchPairsBody body(*this, features, pairwise_matches, near_pairs);
 
     if (is_thread_safe_)
-        parallel_for(BlockedRange(0, static_cast<int>(near_pairs.size())), body);
+        parallel_for_(Range(0, static_cast<int>(near_pairs.size())), body);
     else
-        body(BlockedRange(0, static_cast<int>(near_pairs.size())));
+        body(Range(0, static_cast<int>(near_pairs.size())));
     LOGLN_CHAT("");
 }
 
@@ -529,14 +529,18 @@ void FeaturesMatcher::operator ()(const std::vector<ImageFeatures> &features, st
 
 BestOf2NearestMatcher::BestOf2NearestMatcher(bool try_use_gpu, float match_conf, int num_matches_thresh1, int num_matches_thresh2)
 {
-#ifdef HAVE_OPENCV_GPU
-    if (try_use_gpu && getCudaEnabledDeviceCount() > 0)
-        impl_ = new GpuMatcher(match_conf);
-    else
-#else
     (void)try_use_gpu;
+
+#ifdef HAVE_OPENCV_GPUFEATURES2D
+    if (try_use_gpu && getCudaEnabledDeviceCount() > 0)
+    {
+        impl_ = makePtr<GpuMatcher>(match_conf);
+    }
+    else
 #endif
-        impl_ = new CpuMatcher(match_conf);
+    {
+        impl_ = makePtr<CpuMatcher>(match_conf);
+    }
 
     is_thread_safe_ = impl_->isThreadSafe();
     num_matches_thresh1_ = num_matches_thresh1;
@@ -572,7 +576,7 @@ void BestOf2NearestMatcher::match(const ImageFeatures &features1, const ImageFea
     }
 
     // Find pair-wise motion
-    matches_info.H = findHomography(src_points, dst_points, matches_info.inliers_mask, CV_RANSAC);
+    matches_info.H = findHomography(src_points, dst_points, matches_info.inliers_mask, RANSAC);
     if (matches_info.H.empty() || std::abs(determinant(matches_info.H)) < std::numeric_limits<double>::epsilon())
         return;
 
@@ -619,7 +623,7 @@ void BestOf2NearestMatcher::match(const ImageFeatures &features1, const ImageFea
     }
 
     // Rerun motion estimation on inliers only
-    matches_info.H = findHomography(src_points, dst_points, CV_RANSAC);
+    matches_info.H = findHomography(src_points, dst_points, RANSAC);
 }
 
 void BestOf2NearestMatcher::collectGarbage()
diff --git a/modules/stitching/src/motion_estimators.cpp b/modules/stitching/src/motion_estimators.cpp
index f00f563a2d..d6fd2275d0 100644
--- a/modules/stitching/src/motion_estimators.cpp
+++ b/modules/stitching/src/motion_estimators.cpp
@@ -41,6 +41,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/calib3d/calib3d_c.h"
+#include "opencv2/core/cvdef.h"
 
 using namespace cv;
 using namespace cv::detail;
@@ -68,13 +70,13 @@ struct CalcRotation
         K_from(0,0) = cameras[edge.from].focal;
         K_from(1,1) = cameras[edge.from].focal * cameras[edge.from].aspect;
         K_from(0,2) = cameras[edge.from].ppx;
-        K_from(0,2) = cameras[edge.from].ppy;
+        K_from(1,2) = cameras[edge.from].ppy;
 
         Mat_<double> K_to = Mat::eye(3, 3, CV_64F);
         K_to(0,0) = cameras[edge.to].focal;
         K_to(1,1) = cameras[edge.to].focal * cameras[edge.to].aspect;
         K_to(0,2) = cameras[edge.to].ppx;
-        K_to(0,2) = cameras[edge.to].ppy;
+        K_to(1,2) = cameras[edge.to].ppy;
 
         Mat R = K_from.inv() * pairwise_matches[pair_idx].H.inv() * K_to;
         cameras[edge.to].R = cameras[edge.from].R * R;
@@ -100,8 +102,10 @@ void calcDeriv(const Mat &err1, const Mat &err2, double h, Mat res)
 namespace cv {
 namespace detail {
 
-void HomographyBasedEstimator::estimate(const std::vector<ImageFeatures> &features, const std::vector<MatchesInfo> &pairwise_matches,
-                                        std::vector<CameraParams> &cameras)
+bool HomographyBasedEstimator::estimate(
+        const std::vector<ImageFeatures> &features,
+        const std::vector<MatchesInfo> &pairwise_matches,
+        std::vector<CameraParams> &cameras)
 {
     LOGLN("Estimating rotations...");
 #if ENABLE_LOG
@@ -163,12 +167,13 @@ void HomographyBasedEstimator::estimate(const std::vector<ImageFeatures> &featur
     }
 
     LOGLN("Estimating rotations, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");
+    return true;
 }
 
 
 //////////////////////////////////////////////////////////////////////////////
 
-void BundleAdjusterBase::estimate(const std::vector<ImageFeatures> &features,
+bool BundleAdjusterBase::estimate(const std::vector<ImageFeatures> &features,
                                   const std::vector<MatchesInfo> &pairwise_matches,
                                   std::vector<CameraParams> &cameras)
 {
@@ -244,6 +249,19 @@ void BundleAdjusterBase::estimate(const std::vector<ImageFeatures> &features,
     LOGLN_CHAT("Bundle adjustment, final RMS error: " << std::sqrt(err.dot(err) / total_num_matches_));
     LOGLN_CHAT("Bundle adjustment, iterations done: " << iter);
 
+    // Check if all camera parameters are valid
+    bool ok = true;
+    for (int i = 0; i < cam_params_.rows; ++i)
+    {
+        if (cvIsNaN(cam_params_.at<double>(i,0)))
+        {
+            ok = false;
+            break;
+        }
+    }
+    if (!ok)
+        return false;
+
     obtainRefinedCameraParams(cameras);
 
     // Normalize motion to center image
@@ -255,6 +273,7 @@ void BundleAdjusterBase::estimate(const std::vector<ImageFeatures> &features,
         cameras[i].R = R_inv * cameras[i].R;
 
     LOGLN_CHAT("Bundle adjustment, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");
+    return true;
 }
 
 
@@ -653,7 +672,7 @@ void waveCorrect(std::vector<Mat> &rmats, WaveCorrectKind kind)
 
 //////////////////////////////////////////////////////////////////////////////
 
-std::string matchesGraphAsString(std::vector<std::string> &pathes, std::vector<MatchesInfo> &pairwise_matches,
+String matchesGraphAsString(std::vector<String> &pathes, std::vector<MatchesInfo> &pairwise_matches,
                                 float conf_threshold)
 {
     std::stringstream str;
@@ -685,18 +704,18 @@ std::string matchesGraphAsString(std::vector<std::string> &pathes, std::vector<M
         std::pair<int,int> edge = *itr;
         if (span_tree_edges.find(edge) != span_tree_edges.end())
         {
-            std::string name_src = pathes[edge.first];
+            String name_src = pathes[edge.first];
             size_t prefix_len = name_src.find_last_of("/\\");
-            if (prefix_len != std::string::npos) prefix_len++; else prefix_len = 0;
+            if (prefix_len != String::npos) prefix_len++; else prefix_len = 0;
             name_src = name_src.substr(prefix_len, name_src.size() - prefix_len);
 
-            std::string name_dst = pathes[edge.second];
+            String name_dst = pathes[edge.second];
             prefix_len = name_dst.find_last_of("/\\");
-            if (prefix_len != std::string::npos) prefix_len++; else prefix_len = 0;
+            if (prefix_len != String::npos) prefix_len++; else prefix_len = 0;
             name_dst = name_dst.substr(prefix_len, name_dst.size() - prefix_len);
 
             int pos = edge.first*num_images + edge.second;
-            str << "\"" << name_src << "\" -- \"" << name_dst << "\""
+            str << "\"" << name_src.c_str() << "\" -- \"" << name_dst.c_str() << "\""
                 << "[label=\"Nm=" << pairwise_matches[pos].matches.size()
                 << ", Ni=" << pairwise_matches[pos].num_inliers
                 << ", C=" << pairwise_matches[pos].confidence << "\"];\n";
@@ -707,16 +726,16 @@ std::string matchesGraphAsString(std::vector<std::string> &pathes, std::vector<M
     {
         if (comps.size[comps.findSetByElem((int)i)] == 1)
         {
-            std::string name = pathes[i];
+            String name = pathes[i];
             size_t prefix_len = name.find_last_of("/\\");
-            if (prefix_len != std::string::npos) prefix_len++; else prefix_len = 0;
+            if (prefix_len != String::npos) prefix_len++; else prefix_len = 0;
             name = name.substr(prefix_len, name.size() - prefix_len);
-            str << "\"" << name << "\";\n";
+            str << "\"" << name.c_str() << "\";\n";
         }
     }
 
     str << "}";
-    return str.str();
+    return str.str().c_str();
 }
 
 std::vector<int> leaveBiggestComponent(std::vector<ImageFeatures> &features,  std::vector<MatchesInfo> &pairwise_matches,
@@ -850,4 +869,3 @@ void findMaxSpanningTree(int num_images, const std::vector<MatchesInfo> &pairwis
 
 } // namespace detail
 } // namespace cv
-
diff --git a/modules/stitching/src/precomp.hpp b/modules/stitching/src/precomp.hpp
index 7d75786d7c..ddd216a276 100644
--- a/modules/stitching/src/precomp.hpp
+++ b/modules/stitching/src/precomp.hpp
@@ -43,9 +43,6 @@
 #ifndef __OPENCV_STITCHING_PRECOMP_H__
 #define __OPENCV_STITCHING_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
 #include "opencv2/opencv_modules.hpp"
 
 #include <vector>
@@ -56,7 +53,7 @@
 #include <sstream>
 #include <cmath>
 #include "opencv2/core.hpp"
-#include "opencv2/core/internal.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/stitching.hpp"
 #include "opencv2/stitching/detail/autocalib.hpp"
 #include "opencv2/stitching/detail/blenders.hpp"
@@ -70,12 +67,31 @@
 #include "opencv2/imgproc.hpp"
 #include "opencv2/features2d.hpp"
 #include "opencv2/calib3d.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUWARPING
+#  include "opencv2/gpuwarping.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUFEATURES2D
+#  include "opencv2/gpufeatures2d.hpp"
+#endif
+
 #ifdef HAVE_OPENCV_GPU
-# include "opencv2/gpu.hpp"
+#  include "opencv2/gpu.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_NONFREE
+#  include "opencv2/nonfree/gpu.hpp"
 #endif
 
 #include "../../imgproc/src/gcgraph.hpp"
 
+#include "opencv2/core/private.hpp"
+
 #ifdef HAVE_TEGRA_OPTIMIZATION
 # include "opencv2/stitching/stitching_tegra.hpp"
 #endif
diff --git a/modules/stitching/src/seam_finders.cpp b/modules/stitching/src/seam_finders.cpp
index d55fc70b78..c453b5feb3 100644
--- a/modules/stitching/src/seam_finders.cpp
+++ b/modules/stitching/src/seam_finders.cpp
@@ -139,8 +139,8 @@ void VoronoiSeamFinder::findInPair(size_t first, size_t second, Rect roi)
     Mat unique2 = submask2.clone(); unique2.setTo(0, collision);
 
     Mat dist1, dist2;
-    distanceTransform(unique1 == 0, dist1, CV_DIST_L1, 3);
-    distanceTransform(unique2 == 0, dist2, CV_DIST_L1, 3);
+    distanceTransform(unique1 == 0, dist1, DIST_L1, 3);
+    distanceTransform(unique2 == 0, dist2, DIST_L1, 3);
 
     Mat seam = dist1 < dist2;
 
@@ -522,17 +522,17 @@ void DpSeamFinder::computeGradients(const Mat &image1, const Mat &image2)
     Mat gray;
 
     if (image1.channels() == 3)
-        cvtColor(image1, gray, CV_BGR2GRAY);
+        cvtColor(image1, gray, COLOR_BGR2GRAY);
     else if (image1.channels() == 4)
-        cvtColor(image1, gray, CV_BGRA2GRAY);
+        cvtColor(image1, gray, COLOR_BGRA2GRAY);
 
     Sobel(gray, gradx1_, CV_32F, 1, 0);
     Sobel(gray, grady1_, CV_32F, 0, 1);
 
     if (image2.channels() == 3)
-        cvtColor(image2, gray, CV_BGR2GRAY);
+        cvtColor(image2, gray, COLOR_BGR2GRAY);
     else if (image2.channels() == 4)
-        cvtColor(image2, gray, CV_BGRA2GRAY);
+        cvtColor(image2, gray, COLOR_BGRA2GRAY);
 
     Sobel(gray, gradx2_, CV_32F, 1, 0);
     Sobel(gray, grady2_, CV_32F, 0, 1);
@@ -716,7 +716,7 @@ void DpSeamFinder::computeCosts(
     else if (image1.type() == CV_8UC4 && image2.type() == CV_8UC4)
         diff = diffL2Square4<uchar>;
     else
-        CV_Error(CV_StsBadArg, "both images must have CV_32FC3(4) or CV_8UC3(4) type");
+        CV_Error(Error::StsBadArg, "both images must have CV_32FC3(4) or CV_8UC3(4) type");
 
     int l = comp+1;
     Rect roi(tls_[comp], brs_[comp]);
@@ -1279,7 +1279,7 @@ void GraphCutSeamFinder::Impl::findInPair(size_t first, size_t second, Rect roi)
                                  submask1, submask2, graph);
         break;
     default:
-        CV_Error(CV_StsBadArg, "unsupported pixel similarity measure");
+        CV_Error(Error::StsBadArg, "unsupported pixel similarity measure");
     }
 
     graph.maxFlow();
@@ -1420,7 +1420,7 @@ void GraphCutSeamFinderGpu::findInPair(size_t first, size_t second, Rect roi)
                                  submask1, submask2, terminals, leftT, rightT, top, bottom);
         break;
     default:
-        CV_Error(CV_StsBadArg, "unsupported pixel similarity measure");
+        CV_Error(Error::StsBadArg, "unsupported pixel similarity measure");
     }
 
     gpu::GpuMat terminals_d(terminals);
@@ -1644,4 +1644,3 @@ void GraphCutSeamFinderGpu::setGraphWeightsColorGrad(
 
 } // namespace detail
 } // namespace cv
-
diff --git a/modules/stitching/src/stitcher.cpp b/modules/stitching/src/stitcher.cpp
index cef9fe4e9a..00e02b07cb 100644
--- a/modules/stitching/src/stitcher.cpp
+++ b/modules/stitching/src/stitcher.cpp
@@ -53,34 +53,34 @@ Stitcher Stitcher::createDefault(bool try_use_gpu)
     stitcher.setPanoConfidenceThresh(1);
     stitcher.setWaveCorrection(true);
     stitcher.setWaveCorrectKind(detail::WAVE_CORRECT_HORIZ);
-    stitcher.setFeaturesMatcher(new detail::BestOf2NearestMatcher(try_use_gpu));
-    stitcher.setBundleAdjuster(new detail::BundleAdjusterRay());
+    stitcher.setFeaturesMatcher(makePtr<detail::BestOf2NearestMatcher>(try_use_gpu));
+    stitcher.setBundleAdjuster(makePtr<detail::BundleAdjusterRay>());
 
 #ifdef HAVE_OPENCV_GPU
     if (try_use_gpu && gpu::getCudaEnabledDeviceCount() > 0)
     {
 #ifdef HAVE_OPENCV_NONFREE
-        stitcher.setFeaturesFinder(new detail::SurfFeaturesFinderGpu());
+        stitcher.setFeaturesFinder(makePtr<detail::SurfFeaturesFinderGpu>());
 #else
-        stitcher.setFeaturesFinder(new detail::OrbFeaturesFinder());
+        stitcher.setFeaturesFinder(makePtr<detail::OrbFeaturesFinder>());
 #endif
-        stitcher.setWarper(new SphericalWarperGpu());
-        stitcher.setSeamFinder(new detail::GraphCutSeamFinderGpu());
+        stitcher.setWarper(makePtr<SphericalWarperGpu>());
+        stitcher.setSeamFinder(makePtr<detail::GraphCutSeamFinderGpu>());
     }
     else
 #endif
     {
 #ifdef HAVE_OPENCV_NONFREE
-        stitcher.setFeaturesFinder(new detail::SurfFeaturesFinder());
+        stitcher.setFeaturesFinder(makePtr<detail::SurfFeaturesFinder>());
 #else
-        stitcher.setFeaturesFinder(new detail::OrbFeaturesFinder());
+        stitcher.setFeaturesFinder(makePtr<detail::OrbFeaturesFinder>());
 #endif
-        stitcher.setWarper(new SphericalWarper());
-        stitcher.setSeamFinder(new detail::GraphCutSeamFinder(detail::GraphCutSeamFinderBase::COST_COLOR));
+        stitcher.setWarper(makePtr<SphericalWarper>());
+        stitcher.setSeamFinder(makePtr<detail::GraphCutSeamFinder>(detail::GraphCutSeamFinderBase::COST_COLOR));
     }
 
-    stitcher.setExposureCompensator(new detail::BlocksGainCompensator());
-    stitcher.setBlender(new detail::MultiBandBlender(try_use_gpu));
+    stitcher.setExposureCompensator(makePtr<detail::BlocksGainCompensator>());
+    stitcher.setBlender(makePtr<detail::MultiBandBlender>(try_use_gpu));
 
     return stitcher;
 }
@@ -102,7 +102,8 @@ Stitcher::Status Stitcher::estimateTransform(InputArray images, const std::vecto
     if ((status = matchImages()) != OK)
         return status;
 
-    estimateCameraParams();
+    if ((status = estimateCameraParams()) != OK)
+        return status;
 
     return OK;
 }
@@ -442,10 +443,11 @@ Stitcher::Status Stitcher::matchImages()
 }
 
 
-void Stitcher::estimateCameraParams()
+Stitcher::Status Stitcher::estimateCameraParams()
 {
     detail::HomographyBasedEstimator estimator;
-    estimator(features_, pairwise_matches_, cameras_);
+    if (!estimator(features_, pairwise_matches_, cameras_))
+        return ERR_HOMOGRAPHY_EST_FAIL;
 
     for (size_t i = 0; i < cameras_.size(); ++i)
     {
@@ -456,7 +458,8 @@ void Stitcher::estimateCameraParams()
     }
 
     bundle_adjuster_->setConfThresh(conf_thresh_);
-    (*bundle_adjuster_)(features_, pairwise_matches_, cameras_);
+    if (!(*bundle_adjuster_)(features_, pairwise_matches_, cameras_))
+        return ERR_CAMERA_PARAMS_ADJUST_FAIL;
 
     // Find median focal length and use it as final image scale
     std::vector<double> focals;
@@ -481,6 +484,8 @@ void Stitcher::estimateCameraParams()
         for (size_t i = 0; i < cameras_.size(); ++i)
             cameras_[i].R = rmats[i];
     }
+
+    return OK;
 }
 
 } // namespace cv
diff --git a/modules/stitching/src/warpers.cpp b/modules/stitching/src/warpers.cpp
index 0df91edc60..1082b42d75 100644
--- a/modules/stitching/src/warpers.cpp
+++ b/modules/stitching/src/warpers.cpp
@@ -210,7 +210,7 @@ void SphericalWarper::detectResultRoi(Size src_size, Point &dst_tl, Point &dst_b
 }
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUWARPING
 Rect PlaneWarperGpu::buildMaps(Size src_size, const Mat &K, const Mat &R, gpu::GpuMat &xmap, gpu::GpuMat &ymap)
 {
     return buildMaps(src_size, K, R, Mat::zeros(3, 1, CV_32F), xmap, ymap);
diff --git a/modules/stitching/test/test_matchers.cpp b/modules/stitching/test/test_matchers.cpp
index baa02ec8d5..c7f068b7fe 100644
--- a/modules/stitching/test/test_matchers.cpp
+++ b/modules/stitching/test/test_matchers.cpp
@@ -49,7 +49,7 @@ using namespace std;
 
 TEST(SurfFeaturesFinder, CanFindInROIs)
 {
-    Ptr<detail::FeaturesFinder> finder = new detail::SurfFeaturesFinder();
+    Ptr<detail::FeaturesFinder> finder = makePtr<detail::SurfFeaturesFinder>();
     Mat img  = imread(string(cvtest::TS::ptr()->get_data_path()) + "cv/shared/lena.png");
 
     vector<Rect> rois;
@@ -75,4 +75,4 @@ TEST(SurfFeaturesFinder, CanFindInROIs)
     ASSERT_EQ(bad_count, 0);
 }
 
-#endif
\ No newline at end of file
+#endif
diff --git a/modules/stitching/test/test_precomp.cpp b/modules/stitching/test/test_precomp.cpp
deleted file mode 100644
index 14a070e817..0000000000
--- a/modules/stitching/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
\ No newline at end of file
diff --git a/modules/superres/CMakeLists.txt b/modules/superres/CMakeLists.txt
new file mode 100644
index 0000000000..45150617c6
--- /dev/null
+++ b/modules/superres/CMakeLists.txt
@@ -0,0 +1,9 @@
+if(ANDROID OR IOS)
+  ocv_module_disable(superres)
+endif()
+
+set(the_description "Super Resolution")
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127 -Wundef)
+ocv_define_module(superres opencv_imgproc opencv_video
+                  OPTIONAL opencv_highgui opencv_ocl
+                  opencv_gpuarithm opencv_gpufilters opencv_gpuwarping opencv_gpuimgproc opencv_gpuoptflow opencv_gpucodec)
diff --git a/modules/superres/doc/super_resolution.rst b/modules/superres/doc/super_resolution.rst
new file mode 100644
index 0000000000..1772b0d2db
--- /dev/null
+++ b/modules/superres/doc/super_resolution.rst
@@ -0,0 +1,84 @@
+Super Resolution
+================
+
+.. highlight:: cpp
+
+The Super Resolution module contains a set of functions and classes that can be used to solve the problem of resolution enhancement. There are a few methods implemented, most of them are descibed in the papers [Farsiu03]_ and [Mitzel09]_.
+
+
+
+superres::SuperResolution
+-------------------------
+Base class for Super Resolution algorithms.
+
+.. ocv:class:: superres::SuperResolution : public Algorithm, public superres::FrameSource
+
+The class is only used to define the common interface for the whole family of Super Resolution algorithms.
+
+
+
+superres::SuperResolution::setInput
+-----------------------------------
+Set input frame source for Super Resolution algorithm.
+
+.. ocv:function:: void superres::SuperResolution::setInput(const Ptr<FrameSource>& frameSource)
+
+    :param frameSource: Input frame source
+
+
+
+superres::SuperResolution::nextFrame
+------------------------------------
+Process next frame from input and return output result.
+
+.. ocv:function:: void superres::SuperResolution::nextFrame(OutputArray frame)
+
+    :param frame: Output result
+
+
+
+superres::SuperResolution::collectGarbage
+-----------------------------------------
+Clear all inner buffers.
+
+.. ocv:function:: void superres::SuperResolution::collectGarbage()
+
+
+
+superres::createSuperResolution_BTVL1
+-------------------------------------
+Create Bilateral TV-L1 Super Resolution.
+
+.. ocv:function:: Ptr<SuperResolution> superres::createSuperResolution_BTVL1()
+
+.. ocv:function:: Ptr<SuperResolution> superres::createSuperResolution_BTVL1_GPU()
+
+This class implements Super Resolution algorithm described in the papers [Farsiu03]_ and [Mitzel09]_ .
+
+Here are important members of the class that control the algorithm, which you can set after constructing the class instance:
+
+    * **int scale** Scale factor.
+
+    * **int iterations** Iteration count.
+
+    * **double tau** Asymptotic value of steepest descent method.
+
+    * **double lambda** Weight parameter to balance data term and smoothness term.
+
+    * **double alpha** Parameter of spacial distribution in Bilateral-TV.
+
+    * **int btvKernelSize** Kernel size of Bilateral-TV filter.
+
+    * **int blurKernelSize** Gaussian blur kernel size.
+
+    * **double blurSigma** Gaussian blur sigma.
+
+    * **int temporalAreaRadius** Radius of the temporal search area.
+
+    * **Ptr<DenseOpticalFlowExt> opticalFlow** Dense optical flow algorithm.
+
+
+
+.. [Farsiu03] S. Farsiu, D. Robinson, M. Elad, P. Milanfar. Fast and robust Super-Resolution. Proc 2003 IEEE Int Conf on Image Process, pp. 291–294, 2003.
+
+.. [Mitzel09] D. Mitzel, T. Pock, T. Schoenemann, D. Cremers. Video super resolution using duality based TV-L1 optical flow. DAGM, 2009.
diff --git a/modules/superres/doc/superres.rst b/modules/superres/doc/superres.rst
new file mode 100644
index 0000000000..6a69fcaeb9
--- /dev/null
+++ b/modules/superres/doc/superres.rst
@@ -0,0 +1,8 @@
+**************************
+superres. Super Resolution
+**************************
+
+.. toctree::
+    :maxdepth: 2
+
+    super_resolution
diff --git a/modules/superres/include/opencv2/superres.hpp b/modules/superres/include/opencv2/superres.hpp
new file mode 100644
index 0000000000..ac12c42beb
--- /dev/null
+++ b/modules/superres/include/opencv2/superres.hpp
@@ -0,0 +1,99 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_SUPERRES_HPP__
+#define __OPENCV_SUPERRES_HPP__
+
+#include "opencv2/core.hpp"
+
+namespace cv
+{
+    namespace superres
+    {
+        CV_EXPORTS bool initModule_superres();
+
+        class CV_EXPORTS FrameSource
+        {
+        public:
+            virtual ~FrameSource();
+
+            virtual void nextFrame(OutputArray frame) = 0;
+            virtual void reset() = 0;
+        };
+
+        CV_EXPORTS Ptr<FrameSource> createFrameSource_Empty();
+
+        CV_EXPORTS Ptr<FrameSource> createFrameSource_Video(const String& fileName);
+        CV_EXPORTS Ptr<FrameSource> createFrameSource_Video_GPU(const String& fileName);
+
+        CV_EXPORTS Ptr<FrameSource> createFrameSource_Camera(int deviceId = 0);
+
+        class CV_EXPORTS SuperResolution : public cv::Algorithm, public FrameSource
+        {
+        public:
+            void setInput(const Ptr<FrameSource>& frameSource);
+
+            void nextFrame(OutputArray frame);
+            void reset();
+
+            virtual void collectGarbage();
+
+        protected:
+            SuperResolution();
+
+            virtual void initImpl(Ptr<FrameSource>& frameSource) = 0;
+            virtual void processImpl(Ptr<FrameSource>& frameSource, OutputArray output) = 0;
+
+        private:
+            Ptr<FrameSource> frameSource_;
+            bool firstCall_;
+        };
+
+        // S. Farsiu , D. Robinson, M. Elad, P. Milanfar. Fast and robust multiframe super resolution.
+        // Dennis Mitzel, Thomas Pock, Thomas Schoenemann, Daniel Cremers. Video Super Resolution using Duality Based TV-L1 Optical Flow.
+        CV_EXPORTS Ptr<SuperResolution> createSuperResolution_BTVL1();
+        CV_EXPORTS Ptr<SuperResolution> createSuperResolution_BTVL1_GPU();
+        CV_EXPORTS Ptr<SuperResolution> createSuperResolution_BTVL1_OCL();
+    }
+}
+
+#endif // __OPENCV_SUPERRES_HPP__
diff --git a/modules/superres/include/opencv2/superres/optical_flow.hpp b/modules/superres/include/opencv2/superres/optical_flow.hpp
new file mode 100644
index 0000000000..c51f7b2170
--- /dev/null
+++ b/modules/superres/include/opencv2/superres/optical_flow.hpp
@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_SUPERRES_OPTICAL_FLOW_HPP__
+#define __OPENCV_SUPERRES_OPTICAL_FLOW_HPP__
+
+#include "opencv2/core.hpp"
+
+namespace cv
+{
+    namespace superres
+    {
+        class CV_EXPORTS DenseOpticalFlowExt : public cv::Algorithm
+        {
+        public:
+            virtual void calc(InputArray frame0, InputArray frame1, OutputArray flow1, OutputArray flow2 = noArray()) = 0;
+            virtual void collectGarbage() = 0;
+        };
+
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_Farneback();
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_Farneback_GPU();
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_Farneback_OCL();
+
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_Simple();
+
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_DualTVL1();
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_DualTVL1_GPU();
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_DualTVL1_OCL();
+
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_Brox_GPU();
+
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_PyrLK_GPU();
+        CV_EXPORTS Ptr<DenseOpticalFlowExt> createOptFlow_PyrLK_OCL();
+    }
+}
+
+#endif // __OPENCV_SUPERRES_OPTICAL_FLOW_HPP__
diff --git a/modules/superres/perf/perf_main.cpp b/modules/superres/perf/perf_main.cpp
new file mode 100644
index 0000000000..0a8ab5deaa
--- /dev/null
+++ b/modules/superres/perf/perf_main.cpp
@@ -0,0 +1,54 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace perf;
+
+static const char * impls[] = {
+#ifdef HAVE_CUDA
+    "cuda",
+#endif
+    "plain"
+};
+
+CV_PERF_TEST_MAIN_WITH_IMPLS(superres, impls, printCudaInfo())
diff --git a/modules/superres/perf/perf_precomp.hpp b/modules/superres/perf/perf_precomp.hpp
new file mode 100644
index 0000000000..72236c4ba0
--- /dev/null
+++ b/modules/superres/perf/perf_precomp.hpp
@@ -0,0 +1,65 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/ts.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+#include "opencv2/superres.hpp"
+#include "opencv2/superres/optical_flow.hpp"
+
+#ifdef GTEST_CREATE_SHARED_LIBRARY
+#error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
+#endif
+
+#endif
diff --git a/modules/superres/perf/perf_superres.cpp b/modules/superres/perf/perf_superres.cpp
new file mode 100644
index 0000000000..9d83c949d4
--- /dev/null
+++ b/modules/superres/perf/perf_superres.cpp
@@ -0,0 +1,183 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace std::tr1;
+using namespace testing;
+using namespace perf;
+using namespace cv;
+using namespace cv::superres;
+using namespace cv::gpu;
+
+namespace
+{
+    class OneFrameSource_CPU : public FrameSource
+    {
+    public:
+        explicit OneFrameSource_CPU(const Mat& frame) : frame_(frame) {}
+
+        void nextFrame(OutputArray frame)
+        {
+            frame.getMatRef() = frame_;
+        }
+
+        void reset()
+        {
+        }
+
+    private:
+        Mat frame_;
+    };
+
+    class OneFrameSource_GPU : public FrameSource
+    {
+    public:
+        explicit OneFrameSource_GPU(const GpuMat& frame) : frame_(frame) {}
+
+        void nextFrame(OutputArray frame)
+        {
+            frame.getGpuMatRef() = frame_;
+        }
+
+        void reset()
+        {
+        }
+
+    private:
+        GpuMat frame_;
+    };
+
+    class ZeroOpticalFlow : public DenseOpticalFlowExt
+    {
+    public:
+        void calc(InputArray frame0, InputArray, OutputArray flow1, OutputArray flow2)
+        {
+            cv::Size size = frame0.size();
+
+            if (!flow2.needed())
+            {
+                flow1.create(size, CV_32FC2);
+
+                if (flow1.kind() == cv::_InputArray::GPU_MAT)
+                    flow1.getGpuMatRef().setTo(cv::Scalar::all(0));
+                else
+                    flow1.getMatRef().setTo(cv::Scalar::all(0));
+            }
+            else
+            {
+                flow1.create(size, CV_32FC1);
+                flow2.create(size, CV_32FC1);
+
+                if (flow1.kind() == cv::_InputArray::GPU_MAT)
+                    flow1.getGpuMatRef().setTo(cv::Scalar::all(0));
+                else
+                    flow1.getMatRef().setTo(cv::Scalar::all(0));
+
+                if (flow2.kind() == cv::_InputArray::GPU_MAT)
+                    flow2.getGpuMatRef().setTo(cv::Scalar::all(0));
+                else
+                    flow2.getMatRef().setTo(cv::Scalar::all(0));
+            }
+        }
+
+        void collectGarbage()
+        {
+        }
+    };
+}
+
+PERF_TEST_P(Size_MatType, SuperResolution_BTVL1,
+            Combine(Values(szSmall64, szSmall128),
+                    Values(MatType(CV_8UC1), MatType(CV_8UC3))))
+{
+    declare.time(5 * 60);
+
+    const Size size = get<0>(GetParam());
+    const int type = get<1>(GetParam());
+
+    Mat frame(size, type);
+    declare.in(frame, WARMUP_RNG);
+
+    const int scale = 2;
+    const int iterations = 50;
+    const int temporalAreaRadius = 1;
+    Ptr<DenseOpticalFlowExt> opticalFlow(new ZeroOpticalFlow);
+
+    if (PERF_RUN_GPU())
+    {
+        Ptr<SuperResolution> superRes = createSuperResolution_BTVL1_GPU();
+
+        superRes->set("scale", scale);
+        superRes->set("iterations", iterations);
+        superRes->set("temporalAreaRadius", temporalAreaRadius);
+        superRes->set("opticalFlow", opticalFlow);
+
+        superRes->setInput(makePtr<OneFrameSource_GPU>(GpuMat(frame)));
+
+        GpuMat dst;
+        superRes->nextFrame(dst);
+
+        TEST_CYCLE_N(10) superRes->nextFrame(dst);
+
+        GPU_SANITY_CHECK(dst, 2);
+    }
+    else
+    {
+        Ptr<SuperResolution> superRes = createSuperResolution_BTVL1();
+
+        superRes->set("scale", scale);
+        superRes->set("iterations", iterations);
+        superRes->set("temporalAreaRadius", temporalAreaRadius);
+        superRes->set("opticalFlow", opticalFlow);
+
+        superRes->setInput(makePtr<OneFrameSource_CPU>(frame));
+
+        Mat dst;
+        superRes->nextFrame(dst);
+
+        TEST_CYCLE_N(10) superRes->nextFrame(dst);
+
+        CPU_SANITY_CHECK(dst);
+    }
+}
diff --git a/modules/superres/perf/perf_superres_ocl.cpp b/modules/superres/perf/perf_superres_ocl.cpp
new file mode 100644
index 0000000000..ce8f5934e0
--- /dev/null
+++ b/modules/superres/perf/perf_superres_ocl.cpp
@@ -0,0 +1,146 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "perf_precomp.hpp"
+
+#ifdef HAVE_OPENCL
+
+#include "opencv2/ocl.hpp"
+using namespace std;
+using namespace testing;
+using namespace perf;
+using namespace cv;
+using namespace cv::superres;
+
+namespace
+{
+    class OneFrameSource_OCL : public FrameSource
+    {
+    public:
+        explicit OneFrameSource_OCL(const ocl::oclMat& frame) : frame_(frame) {}
+
+        void nextFrame(OutputArray frame)
+        {
+            ocl::getOclMatRef(frame) = frame_;
+        }
+        void reset()
+        {
+        }
+
+    private:
+        ocl::oclMat frame_;
+    };
+
+
+    class ZeroOpticalFlowOCL : public DenseOpticalFlowExt
+    {
+    public:
+        void calc(InputArray frame0, InputArray, OutputArray flow1, OutputArray flow2)
+        {
+            ocl::oclMat& frame0_ = ocl::getOclMatRef(frame0);
+            ocl::oclMat& flow1_ = ocl::getOclMatRef(flow1);
+            ocl::oclMat& flow2_ = ocl::getOclMatRef(flow2);
+
+            cv::Size size = frame0_.size();
+
+            if(!flow2.needed())
+            {
+                flow1_.create(size, CV_32FC2);
+                flow1_.setTo(Scalar::all(0));
+            }
+            else
+            {
+                flow1_.create(size, CV_32FC1);
+                flow2_.create(size, CV_32FC1);
+
+                flow1_.setTo(Scalar::all(0));
+                flow2_.setTo(Scalar::all(0));
+            }
+        }
+
+        void collectGarbage()
+        {
+        }
+    };
+}
+
+PERF_TEST_P(Size_MatType, SuperResolution_BTVL1_OCL,
+    Combine(Values(szSmall64, szSmall128),
+    Values(MatType(CV_8UC1), MatType(CV_8UC3))))
+{
+    std::vector<cv::ocl::Info>info;
+    cv::ocl::getDevice(info);
+
+    declare.time(5 * 60);
+
+    const Size size = std::tr1::get<0>(GetParam());
+    const int type = std::tr1::get<1>(GetParam());
+
+    Mat frame(size, type);
+    declare.in(frame, WARMUP_RNG);
+
+    ocl::oclMat frame_ocl;
+    frame_ocl.upload(frame);
+
+
+    const int scale = 2;
+    const int iterations = 50;
+    const int temporalAreaRadius = 1;
+    Ptr<DenseOpticalFlowExt> opticalFlowOcl(new ZeroOpticalFlowOCL);
+
+    Ptr<SuperResolution> superRes_ocl = createSuperResolution_BTVL1_OCL();
+
+    superRes_ocl->set("scale", scale);
+    superRes_ocl->set("iterations", iterations);
+    superRes_ocl->set("temporalAreaRadius", temporalAreaRadius);
+    superRes_ocl->set("opticalFlow", opticalFlowOcl);
+
+    superRes_ocl->setInput(makePtr<OneFrameSource_OCL>(frame_ocl));
+
+    ocl::oclMat dst_ocl;
+    superRes_ocl->nextFrame(dst_ocl);
+
+    TEST_CYCLE_N(10) superRes_ocl->nextFrame(dst_ocl);
+    frame_ocl.release();
+    CPU_SANITY_CHECK(dst_ocl);
+}
+#endif
diff --git a/modules/superres/src/btv_l1.cpp b/modules/superres/src/btv_l1.cpp
new file mode 100644
index 0000000000..e0ee7db63a
--- /dev/null
+++ b/modules/superres/src/btv_l1.cpp
@@ -0,0 +1,618 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+// S. Farsiu , D. Robinson, M. Elad, P. Milanfar. Fast and robust multiframe super resolution.
+// Dennis Mitzel, Thomas Pock, Thomas Schoenemann, Daniel Cremers. Video Super Resolution using Duality Based TV-L1 Optical Flow.
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::superres;
+using namespace cv::superres::detail;
+
+namespace
+{
+    void calcRelativeMotions(const std::vector<Mat>& forwardMotions, const std::vector<Mat>& backwardMotions,
+                             std::vector<Mat>& relForwardMotions, std::vector<Mat>& relBackwardMotions,
+                             int baseIdx, Size size)
+    {
+        const int count = static_cast<int>(forwardMotions.size());
+
+        relForwardMotions.resize(count);
+        relForwardMotions[baseIdx].create(size, CV_32FC2);
+        relForwardMotions[baseIdx].setTo(Scalar::all(0));
+
+        relBackwardMotions.resize(count);
+        relBackwardMotions[baseIdx].create(size, CV_32FC2);
+        relBackwardMotions[baseIdx].setTo(Scalar::all(0));
+
+        for (int i = baseIdx - 1; i >= 0; --i)
+        {
+            add(relForwardMotions[i + 1], forwardMotions[i], relForwardMotions[i]);
+
+            add(relBackwardMotions[i + 1], backwardMotions[i + 1], relBackwardMotions[i]);
+        }
+
+        for (int i = baseIdx + 1; i < count; ++i)
+        {
+            add(relForwardMotions[i - 1], backwardMotions[i], relForwardMotions[i]);
+
+            add(relBackwardMotions[i - 1], forwardMotions[i - 1], relBackwardMotions[i]);
+        }
+    }
+
+    void upscaleMotions(const std::vector<Mat>& lowResMotions, std::vector<Mat>& highResMotions, int scale)
+    {
+        highResMotions.resize(lowResMotions.size());
+
+        for (size_t i = 0; i < lowResMotions.size(); ++i)
+        {
+            resize(lowResMotions[i], highResMotions[i], Size(), scale, scale, INTER_CUBIC);
+            multiply(highResMotions[i], Scalar::all(scale), highResMotions[i]);
+        }
+    }
+
+    void buildMotionMaps(const Mat& forwardMotion, const Mat& backwardMotion, Mat& forwardMap, Mat& backwardMap)
+    {
+        forwardMap.create(forwardMotion.size(), CV_32FC2);
+        backwardMap.create(forwardMotion.size(), CV_32FC2);
+
+        for (int y = 0; y < forwardMotion.rows; ++y)
+        {
+            const Point2f* forwardMotionRow = forwardMotion.ptr<Point2f>(y);
+            const Point2f* backwardMotionRow = backwardMotion.ptr<Point2f>(y);
+            Point2f* forwardMapRow = forwardMap.ptr<Point2f>(y);
+            Point2f* backwardMapRow = backwardMap.ptr<Point2f>(y);
+
+            for (int x = 0; x < forwardMotion.cols; ++x)
+            {
+                Point2f base(static_cast<float>(x), static_cast<float>(y));
+
+                forwardMapRow[x] = base + backwardMotionRow[x];
+                backwardMapRow[x] = base + forwardMotionRow[x];
+            }
+        }
+    }
+
+    template <typename T>
+    void upscaleImpl(const Mat& src, Mat& dst, int scale)
+    {
+        dst.create(src.rows * scale, src.cols * scale, src.type());
+        dst.setTo(Scalar::all(0));
+
+        for (int y = 0, Y = 0; y < src.rows; ++y, Y += scale)
+        {
+            const T* srcRow = src.ptr<T>(y);
+            T* dstRow = dst.ptr<T>(Y);
+
+            for (int x = 0, X = 0; x < src.cols; ++x, X += scale)
+                dstRow[X] = srcRow[x];
+        }
+    }
+
+    void upscale(const Mat& src, Mat& dst, int scale)
+    {
+        typedef void (*func_t)(const Mat& src, Mat& dst, int scale);
+        static const func_t funcs[] =
+        {
+            0, upscaleImpl<float>, 0, upscaleImpl<Point3f>
+        };
+
+        CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+
+        const func_t func = funcs[src.channels()];
+
+        func(src, dst, scale);
+    }
+
+    float diffSign(float a, float b)
+    {
+        return a > b ? 1.0f : a < b ? -1.0f : 0.0f;
+    }
+    Point3f diffSign(Point3f a, Point3f b)
+    {
+        return Point3f(
+            a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f,
+            a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f,
+            a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f
+        );
+    }
+
+    void diffSign(const Mat& src1, const Mat& src2, Mat& dst)
+    {
+        const int count = src1.cols * src1.channels();
+
+        dst.create(src1.size(), src1.type());
+
+        for (int y = 0; y < src1.rows; ++y)
+        {
+            const float* src1Ptr = src1.ptr<float>(y);
+            const float* src2Ptr = src2.ptr<float>(y);
+            float* dstPtr = dst.ptr<float>(y);
+
+            for (int x = 0; x < count; ++x)
+                dstPtr[x] = diffSign(src1Ptr[x], src2Ptr[x]);
+        }
+    }
+
+    void calcBtvWeights(int btvKernelSize, double alpha, std::vector<float>& btvWeights)
+    {
+        const size_t size = btvKernelSize * btvKernelSize;
+
+        btvWeights.resize(size);
+
+        const int ksize = (btvKernelSize - 1) / 2;
+        const float alpha_f = static_cast<float>(alpha);
+
+        for (int m = 0, ind = 0; m <= ksize; ++m)
+        {
+            for (int l = ksize; l + m >= 0; --l, ++ind)
+                btvWeights[ind] = pow(alpha_f, std::abs(m) + std::abs(l));
+        }
+    }
+
+    template <typename T>
+    struct BtvRegularizationBody : ParallelLoopBody
+    {
+        void operator ()(const Range& range) const;
+
+        Mat src;
+        mutable Mat dst;
+        int ksize;
+        const float* btvWeights;
+    };
+
+    template <typename T>
+    void BtvRegularizationBody<T>::operator ()(const Range& range) const
+    {
+        for (int i = range.start; i < range.end; ++i)
+        {
+            const T* srcRow = src.ptr<T>(i);
+            T* dstRow = dst.ptr<T>(i);
+
+            for(int j = ksize; j < src.cols - ksize; ++j)
+            {
+                const T srcVal = srcRow[j];
+
+                for (int m = 0, ind = 0; m <= ksize; ++m)
+                {
+                    const T* srcRow2 = src.ptr<T>(i - m);
+                    const T* srcRow3 = src.ptr<T>(i + m);
+
+                    for (int l = ksize; l + m >= 0; --l, ++ind)
+                    {
+                        dstRow[j] += btvWeights[ind] * (diffSign(srcVal, srcRow3[j + l]) - diffSign(srcRow2[j - l], srcVal));
+                    }
+                }
+            }
+        }
+    }
+
+    template <typename T>
+    void calcBtvRegularizationImpl(const Mat& src, Mat& dst, int btvKernelSize, const std::vector<float>& btvWeights)
+    {
+        dst.create(src.size(), src.type());
+        dst.setTo(Scalar::all(0));
+
+        const int ksize = (btvKernelSize - 1) / 2;
+
+        BtvRegularizationBody<T> body;
+
+        body.src = src;
+        body.dst = dst;
+        body.ksize = ksize;
+        body.btvWeights = &btvWeights[0];
+
+        parallel_for_(Range(ksize, src.rows - ksize), body);
+    }
+
+    void calcBtvRegularization(const Mat& src, Mat& dst, int btvKernelSize, const std::vector<float>& btvWeights)
+    {
+        typedef void (*func_t)(const Mat& src, Mat& dst, int btvKernelSize, const std::vector<float>& btvWeights);
+        static const func_t funcs[] =
+        {
+            0, calcBtvRegularizationImpl<float>, 0, calcBtvRegularizationImpl<Point3f>
+        };
+
+        const func_t func = funcs[src.channels()];
+
+        func(src, dst, btvKernelSize, btvWeights);
+    }
+
+    class BTVL1_Base
+    {
+    public:
+        BTVL1_Base();
+
+        void process(const std::vector<Mat>& src, Mat& dst,
+                     const std::vector<Mat>& forwardMotions, const std::vector<Mat>& backwardMotions,
+                     int baseIdx);
+
+        void collectGarbage();
+
+    protected:
+        int scale_;
+        int iterations_;
+        double tau_;
+        double lambda_;
+        double alpha_;
+        int btvKernelSize_;
+        int blurKernelSize_;
+        double blurSigma_;
+        Ptr<DenseOpticalFlowExt> opticalFlow_;
+
+    private:
+        Ptr<FilterEngine> filter_;
+        int curBlurKernelSize_;
+        double curBlurSigma_;
+        int curSrcType_;
+
+        std::vector<float> btvWeights_;
+        int curBtvKernelSize_;
+        double curAlpha_;
+
+        std::vector<Mat> lowResForwardMotions_;
+        std::vector<Mat> lowResBackwardMotions_;
+
+        std::vector<Mat> highResForwardMotions_;
+        std::vector<Mat> highResBackwardMotions_;
+
+        std::vector<Mat> forwardMaps_;
+        std::vector<Mat> backwardMaps_;
+
+        Mat highRes_;
+
+        Mat diffTerm_, regTerm_;
+        Mat a_, b_, c_;
+    };
+
+    BTVL1_Base::BTVL1_Base()
+    {
+        scale_ = 4;
+        iterations_ = 180;
+        lambda_ = 0.03;
+        tau_ = 1.3;
+        alpha_ = 0.7;
+        btvKernelSize_ = 7;
+        blurKernelSize_ = 5;
+        blurSigma_ = 0.0;
+        opticalFlow_ = createOptFlow_Farneback();
+
+        curBlurKernelSize_ = -1;
+        curBlurSigma_ = -1.0;
+        curSrcType_ = -1;
+
+        curBtvKernelSize_ = -1;
+        curAlpha_ = -1.0;
+    }
+
+    void BTVL1_Base::process(const std::vector<Mat>& src, Mat& dst, const std::vector<Mat>& forwardMotions, const std::vector<Mat>& backwardMotions, int baseIdx)
+    {
+        CV_Assert( scale_ > 1 );
+        CV_Assert( iterations_ > 0 );
+        CV_Assert( tau_ > 0.0 );
+        CV_Assert( alpha_ > 0.0 );
+        CV_Assert( btvKernelSize_ > 0 );
+        CV_Assert( blurKernelSize_ > 0 );
+        CV_Assert( blurSigma_ >= 0.0 );
+
+        // update blur filter and btv weights
+
+        if (!filter_ || blurKernelSize_ != curBlurKernelSize_ || blurSigma_ != curBlurSigma_ || src[0].type() != curSrcType_)
+        {
+            filter_ = createGaussianFilter(src[0].type(), Size(blurKernelSize_, blurKernelSize_), blurSigma_);
+            curBlurKernelSize_ = blurKernelSize_;
+            curBlurSigma_ = blurSigma_;
+            curSrcType_ = src[0].type();
+        }
+
+        if (btvWeights_.empty() || btvKernelSize_ != curBtvKernelSize_ || alpha_ != curAlpha_)
+        {
+            calcBtvWeights(btvKernelSize_, alpha_, btvWeights_);
+            curBtvKernelSize_ = btvKernelSize_;
+            curAlpha_ = alpha_;
+        }
+
+        // calc high res motions
+
+        calcRelativeMotions(forwardMotions, backwardMotions, lowResForwardMotions_, lowResBackwardMotions_, baseIdx, src[0].size());
+
+        upscaleMotions(lowResForwardMotions_, highResForwardMotions_, scale_);
+        upscaleMotions(lowResBackwardMotions_, highResBackwardMotions_, scale_);
+
+        forwardMaps_.resize(highResForwardMotions_.size());
+        backwardMaps_.resize(highResForwardMotions_.size());
+        for (size_t i = 0; i < highResForwardMotions_.size(); ++i)
+            buildMotionMaps(highResForwardMotions_[i], highResBackwardMotions_[i], forwardMaps_[i], backwardMaps_[i]);
+
+        // initial estimation
+
+        const Size lowResSize = src[0].size();
+        const Size highResSize(lowResSize.width * scale_, lowResSize.height * scale_);
+
+        resize(src[baseIdx], highRes_, highResSize, 0, 0, INTER_CUBIC);
+
+        // iterations
+
+        diffTerm_.create(highResSize, highRes_.type());
+        a_.create(highResSize, highRes_.type());
+        b_.create(highResSize, highRes_.type());
+        c_.create(lowResSize, highRes_.type());
+
+        for (int i = 0; i < iterations_; ++i)
+        {
+            diffTerm_.setTo(Scalar::all(0));
+
+            for (size_t k = 0; k < src.size(); ++k)
+            {
+                // a = M * Ih
+                remap(highRes_, a_, backwardMaps_[k], noArray(), INTER_NEAREST);
+                // b = HM * Ih
+                filter_->apply(a_, b_);
+                // c = DHM * Ih
+                resize(b_, c_, lowResSize, 0, 0, INTER_NEAREST);
+
+                diffSign(src[k], c_, c_);
+
+                // a = Dt * diff
+                upscale(c_, a_, scale_);
+                // b = HtDt * diff
+                filter_->apply(a_, b_);
+                // a = MtHtDt * diff
+                remap(b_, a_, forwardMaps_[k], noArray(), INTER_NEAREST);
+
+                add(diffTerm_, a_, diffTerm_);
+            }
+
+            if (lambda_ > 0)
+            {
+                calcBtvRegularization(highRes_, regTerm_, btvKernelSize_, btvWeights_);
+                addWeighted(diffTerm_, 1.0, regTerm_, -lambda_, 0.0, diffTerm_);
+            }
+
+            addWeighted(highRes_, 1.0, diffTerm_, tau_, 0.0, highRes_);
+        }
+
+        Rect inner(btvKernelSize_, btvKernelSize_, highRes_.cols - 2 * btvKernelSize_, highRes_.rows - 2 * btvKernelSize_);
+        highRes_(inner).copyTo(dst);
+    }
+
+    void BTVL1_Base::collectGarbage()
+    {
+        filter_.release();
+
+        lowResForwardMotions_.clear();
+        lowResBackwardMotions_.clear();
+
+        highResForwardMotions_.clear();
+        highResBackwardMotions_.clear();
+
+        forwardMaps_.clear();
+        backwardMaps_.clear();
+
+        highRes_.release();
+
+        diffTerm_.release();
+        regTerm_.release();
+        a_.release();
+        b_.release();
+        c_.release();
+    }
+
+////////////////////////////////////////////////////////////////////
+
+    class BTVL1 : public SuperResolution, private BTVL1_Base
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        BTVL1();
+
+        void collectGarbage();
+
+    protected:
+        void initImpl(Ptr<FrameSource>& frameSource);
+        void processImpl(Ptr<FrameSource>& frameSource, OutputArray output);
+
+    private:
+        int temporalAreaRadius_;
+
+        void readNextFrame(Ptr<FrameSource>& frameSource);
+        void processFrame(int idx);
+
+        Mat curFrame_;
+        Mat prevFrame_;
+
+        std::vector<Mat> frames_;
+        std::vector<Mat> forwardMotions_;
+        std::vector<Mat> backwardMotions_;
+        std::vector<Mat> outputs_;
+
+        int storePos_;
+        int procPos_;
+        int outPos_;
+
+        std::vector<Mat> srcFrames_;
+        std::vector<Mat> srcForwardMotions_;
+        std::vector<Mat> srcBackwardMotions_;
+        Mat finalOutput_;
+    };
+
+    CV_INIT_ALGORITHM(BTVL1, "SuperResolution.BTVL1",
+                      obj.info()->addParam(obj, "scale", obj.scale_, false, 0, 0, "Scale factor.");
+                      obj.info()->addParam(obj, "iterations", obj.iterations_, false, 0, 0, "Iteration count.");
+                      obj.info()->addParam(obj, "tau", obj.tau_, false, 0, 0, "Asymptotic value of steepest descent method.");
+                      obj.info()->addParam(obj, "lambda", obj.lambda_, false, 0, 0, "Weight parameter to balance data term and smoothness term.");
+                      obj.info()->addParam(obj, "alpha", obj.alpha_, false, 0, 0, "Parameter of spacial distribution in Bilateral-TV.");
+                      obj.info()->addParam(obj, "btvKernelSize", obj.btvKernelSize_, false, 0, 0, "Kernel size of Bilateral-TV filter.");
+                      obj.info()->addParam(obj, "blurKernelSize", obj.blurKernelSize_, false, 0, 0, "Gaussian blur kernel size.");
+                      obj.info()->addParam(obj, "blurSigma", obj.blurSigma_, false, 0, 0, "Gaussian blur sigma.");
+                      obj.info()->addParam(obj, "temporalAreaRadius", obj.temporalAreaRadius_, false, 0, 0, "Radius of the temporal search area.");
+                      obj.info()->addParam<DenseOpticalFlowExt>(obj, "opticalFlow", obj.opticalFlow_, false, 0, 0, "Dense optical flow algorithm."));
+
+    BTVL1::BTVL1()
+    {
+        temporalAreaRadius_ = 4;
+    }
+
+    void BTVL1::collectGarbage()
+    {
+        curFrame_.release();
+        prevFrame_.release();
+
+        frames_.clear();
+        forwardMotions_.clear();
+        backwardMotions_.clear();
+        outputs_.clear();
+
+        srcFrames_.clear();
+        srcForwardMotions_.clear();
+        srcBackwardMotions_.clear();
+        finalOutput_.release();
+
+        SuperResolution::collectGarbage();
+        BTVL1_Base::collectGarbage();
+    }
+
+    void BTVL1::initImpl(Ptr<FrameSource>& frameSource)
+    {
+        const int cacheSize = 2 * temporalAreaRadius_ + 1;
+
+        frames_.resize(cacheSize);
+        forwardMotions_.resize(cacheSize);
+        backwardMotions_.resize(cacheSize);
+        outputs_.resize(cacheSize);
+
+        storePos_ = -1;
+
+        for (int t = -temporalAreaRadius_; t <= temporalAreaRadius_; ++t)
+            readNextFrame(frameSource);
+
+        for (int i = 0; i <= temporalAreaRadius_; ++i)
+            processFrame(i);
+
+        procPos_ = temporalAreaRadius_;
+        outPos_ = -1;
+    }
+
+    void BTVL1::processImpl(Ptr<FrameSource>& frameSource, OutputArray _output)
+    {
+        if (outPos_ >= storePos_)
+        {
+            _output.release();
+            return;
+        }
+
+        readNextFrame(frameSource);
+
+        if (procPos_ < storePos_)
+        {
+            ++procPos_;
+            processFrame(procPos_);
+        }
+
+        ++outPos_;
+        const Mat& curOutput = at(outPos_, outputs_);
+
+        if (_output.kind() < _InputArray::OPENGL_BUFFER)
+            curOutput.convertTo(_output, CV_8U);
+        else
+        {
+            curOutput.convertTo(finalOutput_, CV_8U);
+            arrCopy(finalOutput_, _output);
+        }
+    }
+
+    void BTVL1::readNextFrame(Ptr<FrameSource>& frameSource)
+    {
+        frameSource->nextFrame(curFrame_);
+
+        if (curFrame_.empty())
+            return;
+
+        ++storePos_;
+        curFrame_.convertTo(at(storePos_, frames_), CV_32F);
+
+        if (storePos_ > 0)
+        {
+            opticalFlow_->calc(prevFrame_, curFrame_, at(storePos_ - 1, forwardMotions_));
+            opticalFlow_->calc(curFrame_, prevFrame_, at(storePos_, backwardMotions_));
+        }
+
+        curFrame_.copyTo(prevFrame_);
+    }
+
+    void BTVL1::processFrame(int idx)
+    {
+        const int startIdx = std::max(idx - temporalAreaRadius_, 0);
+        const int procIdx = idx;
+        const int endIdx = std::min(startIdx + 2 * temporalAreaRadius_, storePos_);
+
+        const int count = endIdx - startIdx + 1;
+
+        srcFrames_.resize(count);
+        srcForwardMotions_.resize(count);
+        srcBackwardMotions_.resize(count);
+
+        int baseIdx = -1;
+
+        for (int i = startIdx, k = 0; i <= endIdx; ++i, ++k)
+        {
+            if (i == procIdx)
+                baseIdx = k;
+
+            srcFrames_[k] = at(i, frames_);
+
+            if (i < endIdx)
+                srcForwardMotions_[k] = at(i, forwardMotions_);
+            if (i > startIdx)
+                srcBackwardMotions_[k] = at(i, backwardMotions_);
+        }
+
+        process(srcFrames_, at(idx, outputs_), srcForwardMotions_, srcBackwardMotions_, baseIdx);
+    }
+}
+
+Ptr<SuperResolution> cv::superres::createSuperResolution_BTVL1()
+{
+    return makePtr<BTVL1>();
+}
diff --git a/modules/superres/src/btv_l1_gpu.cpp b/modules/superres/src/btv_l1_gpu.cpp
new file mode 100644
index 0000000000..f4268b01bd
--- /dev/null
+++ b/modules/superres/src/btv_l1_gpu.cpp
@@ -0,0 +1,584 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+// S. Farsiu , D. Robinson, M. Elad, P. Milanfar. Fast and robust multiframe super resolution.
+// Dennis Mitzel, Thomas Pock, Thomas Schoenemann, Daniel Cremers. Video Super Resolution using Duality Based TV-L1 Optical Flow.
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::superres;
+using namespace cv::superres::detail;
+
+#if !defined(HAVE_CUDA) || !defined(HAVE_OPENCV_GPUARITHM) || !defined(HAVE_OPENCV_GPUWARPING) || !defined(HAVE_OPENCV_GPUFILTERS)
+
+Ptr<SuperResolution> cv::superres::createSuperResolution_BTVL1_GPU()
+{
+    CV_Error(Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<SuperResolution>();
+}
+
+#else // HAVE_CUDA
+
+namespace btv_l1_cudev
+{
+    void buildMotionMaps(PtrStepSzf forwardMotionX, PtrStepSzf forwardMotionY,
+                         PtrStepSzf backwardMotionX, PtrStepSzf bacwardMotionY,
+                         PtrStepSzf forwardMapX, PtrStepSzf forwardMapY,
+                         PtrStepSzf backwardMapX, PtrStepSzf backwardMapY);
+
+    template <int cn>
+    void upscale(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+
+    void diffSign(PtrStepSzf src1, PtrStepSzf src2, PtrStepSzf dst, cudaStream_t stream);
+
+    void loadBtvWeights(const float* weights, size_t count);
+    template <int cn> void calcBtvRegularization(PtrStepSzb src, PtrStepSzb dst, int ksize);
+}
+
+namespace
+{
+    void calcRelativeMotions(const std::vector<std::pair<GpuMat, GpuMat> >& forwardMotions, const std::vector<std::pair<GpuMat, GpuMat> >& backwardMotions,
+                             std::vector<std::pair<GpuMat, GpuMat> >& relForwardMotions, std::vector<std::pair<GpuMat, GpuMat> >& relBackwardMotions,
+                             int baseIdx, Size size)
+    {
+        const int count = static_cast<int>(forwardMotions.size());
+
+        relForwardMotions.resize(count);
+        relForwardMotions[baseIdx].first.create(size, CV_32FC1);
+        relForwardMotions[baseIdx].first.setTo(Scalar::all(0));
+        relForwardMotions[baseIdx].second.create(size, CV_32FC1);
+        relForwardMotions[baseIdx].second.setTo(Scalar::all(0));
+
+        relBackwardMotions.resize(count);
+        relBackwardMotions[baseIdx].first.create(size, CV_32FC1);
+        relBackwardMotions[baseIdx].first.setTo(Scalar::all(0));
+        relBackwardMotions[baseIdx].second.create(size, CV_32FC1);
+        relBackwardMotions[baseIdx].second.setTo(Scalar::all(0));
+
+        for (int i = baseIdx - 1; i >= 0; --i)
+        {
+            gpu::add(relForwardMotions[i + 1].first, forwardMotions[i].first, relForwardMotions[i].first);
+            gpu::add(relForwardMotions[i + 1].second, forwardMotions[i].second, relForwardMotions[i].second);
+
+            gpu::add(relBackwardMotions[i + 1].first, backwardMotions[i + 1].first, relBackwardMotions[i].first);
+            gpu::add(relBackwardMotions[i + 1].second, backwardMotions[i + 1].second, relBackwardMotions[i].second);
+        }
+
+        for (int i = baseIdx + 1; i < count; ++i)
+        {
+            gpu::add(relForwardMotions[i - 1].first, backwardMotions[i].first, relForwardMotions[i].first);
+            gpu::add(relForwardMotions[i - 1].second, backwardMotions[i].second, relForwardMotions[i].second);
+
+            gpu::add(relBackwardMotions[i - 1].first, forwardMotions[i - 1].first, relBackwardMotions[i].first);
+            gpu::add(relBackwardMotions[i - 1].second, forwardMotions[i - 1].second, relBackwardMotions[i].second);
+        }
+    }
+
+    void upscaleMotions(const std::vector<std::pair<GpuMat, GpuMat> >& lowResMotions, std::vector<std::pair<GpuMat, GpuMat> >& highResMotions, int scale)
+    {
+        highResMotions.resize(lowResMotions.size());
+
+        for (size_t i = 0; i < lowResMotions.size(); ++i)
+        {
+            gpu::resize(lowResMotions[i].first, highResMotions[i].first, Size(), scale, scale, INTER_CUBIC);
+            gpu::resize(lowResMotions[i].second, highResMotions[i].second, Size(), scale, scale, INTER_CUBIC);
+
+            gpu::multiply(highResMotions[i].first, Scalar::all(scale), highResMotions[i].first);
+            gpu::multiply(highResMotions[i].second, Scalar::all(scale), highResMotions[i].second);
+        }
+    }
+
+    void buildMotionMaps(const std::pair<GpuMat, GpuMat>& forwardMotion, const std::pair<GpuMat, GpuMat>& backwardMotion,
+                         std::pair<GpuMat, GpuMat>& forwardMap, std::pair<GpuMat, GpuMat>& backwardMap)
+    {
+        forwardMap.first.create(forwardMotion.first.size(), CV_32FC1);
+        forwardMap.second.create(forwardMotion.first.size(), CV_32FC1);
+
+        backwardMap.first.create(forwardMotion.first.size(), CV_32FC1);
+        backwardMap.second.create(forwardMotion.first.size(), CV_32FC1);
+
+        btv_l1_cudev::buildMotionMaps(forwardMotion.first, forwardMotion.second,
+                                       backwardMotion.first, backwardMotion.second,
+                                       forwardMap.first, forwardMap.second,
+                                       backwardMap.first, backwardMap.second);
+    }
+
+    void upscale(const GpuMat& src, GpuMat& dst, int scale, Stream& stream)
+    {
+        typedef void (*func_t)(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+        static const func_t funcs[] =
+        {
+            0, btv_l1_cudev::upscale<1>, 0, btv_l1_cudev::upscale<3>, btv_l1_cudev::upscale<4>
+        };
+
+        CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+
+        dst.create(src.rows * scale, src.cols * scale, src.type());
+        dst.setTo(Scalar::all(0));
+
+        const func_t func = funcs[src.channels()];
+
+        func(src, dst, scale, StreamAccessor::getStream(stream));
+    }
+
+    void diffSign(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream)
+    {
+        dst.create(src1.size(), src1.type());
+
+        btv_l1_cudev::diffSign(src1.reshape(1), src2.reshape(1), dst.reshape(1), StreamAccessor::getStream(stream));
+    }
+
+    void calcBtvWeights(int btvKernelSize, double alpha, std::vector<float>& btvWeights)
+    {
+        const size_t size = btvKernelSize * btvKernelSize;
+
+        btvWeights.resize(size);
+
+        const int ksize = (btvKernelSize - 1) / 2;
+        const float alpha_f = static_cast<float>(alpha);
+
+        for (int m = 0, ind = 0; m <= ksize; ++m)
+        {
+            for (int l = ksize; l + m >= 0; --l, ++ind)
+                btvWeights[ind] = pow(alpha_f, std::abs(m) + std::abs(l));
+        }
+
+        btv_l1_cudev::loadBtvWeights(&btvWeights[0], size);
+    }
+
+    void calcBtvRegularization(const GpuMat& src, GpuMat& dst, int btvKernelSize)
+    {
+        typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, int ksize);
+        static const func_t funcs[] =
+        {
+            0,
+            btv_l1_cudev::calcBtvRegularization<1>,
+            0,
+            btv_l1_cudev::calcBtvRegularization<3>,
+            btv_l1_cudev::calcBtvRegularization<4>
+        };
+
+        dst.create(src.size(), src.type());
+        dst.setTo(Scalar::all(0));
+
+        const int ksize = (btvKernelSize - 1) / 2;
+
+        funcs[src.channels()](src, dst, ksize);
+    }
+
+    class BTVL1_GPU_Base
+    {
+    public:
+        BTVL1_GPU_Base();
+
+        void process(const std::vector<GpuMat>& src, GpuMat& dst,
+                     const std::vector<std::pair<GpuMat, GpuMat> >& forwardMotions, const std::vector<std::pair<GpuMat, GpuMat> >& backwardMotions,
+                     int baseIdx);
+
+        void collectGarbage();
+
+    protected:
+        int scale_;
+        int iterations_;
+        double lambda_;
+        double tau_;
+        double alpha_;
+        int btvKernelSize_;
+        int blurKernelSize_;
+        double blurSigma_;
+        Ptr<DenseOpticalFlowExt> opticalFlow_;
+
+    private:
+        std::vector<Ptr<gpu::Filter> > filters_;
+        int curBlurKernelSize_;
+        double curBlurSigma_;
+        int curSrcType_;
+
+        std::vector<float> btvWeights_;
+        int curBtvKernelSize_;
+        double curAlpha_;
+
+        std::vector<std::pair<GpuMat, GpuMat> > lowResForwardMotions_;
+        std::vector<std::pair<GpuMat, GpuMat> > lowResBackwardMotions_;
+
+        std::vector<std::pair<GpuMat, GpuMat> > highResForwardMotions_;
+        std::vector<std::pair<GpuMat, GpuMat> > highResBackwardMotions_;
+
+        std::vector<std::pair<GpuMat, GpuMat> > forwardMaps_;
+        std::vector<std::pair<GpuMat, GpuMat> > backwardMaps_;
+
+        GpuMat highRes_;
+
+        std::vector<Stream> streams_;
+        std::vector<GpuMat> diffTerms_;
+        std::vector<GpuMat> a_, b_, c_;
+        GpuMat regTerm_;
+    };
+
+    BTVL1_GPU_Base::BTVL1_GPU_Base()
+    {
+        scale_ = 4;
+        iterations_ = 180;
+        lambda_ = 0.03;
+        tau_ = 1.3;
+        alpha_ = 0.7;
+        btvKernelSize_ = 7;
+        blurKernelSize_ = 5;
+        blurSigma_ = 0.0;
+
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+        opticalFlow_ = createOptFlow_Farneback_GPU();
+#else
+        opticalFlow_ = createOptFlow_Farneback();
+#endif
+
+        curBlurKernelSize_ = -1;
+        curBlurSigma_ = -1.0;
+        curSrcType_ = -1;
+
+        curBtvKernelSize_ = -1;
+        curAlpha_ = -1.0;
+    }
+
+    void BTVL1_GPU_Base::process(const std::vector<GpuMat>& src, GpuMat& dst,
+                                 const std::vector<std::pair<GpuMat, GpuMat> >& forwardMotions, const std::vector<std::pair<GpuMat, GpuMat> >& backwardMotions,
+                                 int baseIdx)
+    {
+        CV_Assert( scale_ > 1 );
+        CV_Assert( iterations_ > 0 );
+        CV_Assert( tau_ > 0.0 );
+        CV_Assert( alpha_ > 0.0 );
+        CV_Assert( btvKernelSize_ > 0 && btvKernelSize_ <= 16 );
+        CV_Assert( blurKernelSize_ > 0 );
+        CV_Assert( blurSigma_ >= 0.0 );
+
+        // update blur filter and btv weights
+
+        if (filters_.size() != src.size() || blurKernelSize_ != curBlurKernelSize_ || blurSigma_ != curBlurSigma_ || src[0].type() != curSrcType_)
+        {
+            filters_.resize(src.size());
+            for (size_t i = 0; i < src.size(); ++i)
+                filters_[i] = gpu::createGaussianFilter(src[0].type(), -1, Size(blurKernelSize_, blurKernelSize_), blurSigma_);
+            curBlurKernelSize_ = blurKernelSize_;
+            curBlurSigma_ = blurSigma_;
+            curSrcType_ = src[0].type();
+        }
+
+        if (btvWeights_.empty() || btvKernelSize_ != curBtvKernelSize_ || alpha_ != curAlpha_)
+        {
+            calcBtvWeights(btvKernelSize_, alpha_, btvWeights_);
+            curBtvKernelSize_ = btvKernelSize_;
+            curAlpha_ = alpha_;
+        }
+
+        // calc motions between input frames
+
+        calcRelativeMotions(forwardMotions, backwardMotions, lowResForwardMotions_, lowResBackwardMotions_, baseIdx, src[0].size());
+
+        upscaleMotions(lowResForwardMotions_, highResForwardMotions_, scale_);
+        upscaleMotions(lowResBackwardMotions_, highResBackwardMotions_, scale_);
+
+        forwardMaps_.resize(highResForwardMotions_.size());
+        backwardMaps_.resize(highResForwardMotions_.size());
+        for (size_t i = 0; i < highResForwardMotions_.size(); ++i)
+            buildMotionMaps(highResForwardMotions_[i], highResBackwardMotions_[i], forwardMaps_[i], backwardMaps_[i]);
+
+        // initial estimation
+
+        const Size lowResSize = src[0].size();
+        const Size highResSize(lowResSize.width * scale_, lowResSize.height * scale_);
+
+        gpu::resize(src[baseIdx], highRes_, highResSize, 0, 0, INTER_CUBIC);
+
+        // iterations
+
+        streams_.resize(src.size());
+        diffTerms_.resize(src.size());
+        a_.resize(src.size());
+        b_.resize(src.size());
+        c_.resize(src.size());
+
+        for (int i = 0; i < iterations_; ++i)
+        {
+            for (size_t k = 0; k < src.size(); ++k)
+            {
+                // a = M * Ih
+                gpu::remap(highRes_, a_[k], backwardMaps_[k].first, backwardMaps_[k].second, INTER_NEAREST, BORDER_REPLICATE, Scalar(), streams_[k]);
+                // b = HM * Ih
+                filters_[k]->apply(a_[k], b_[k], streams_[k]);
+                // c = DHF * Ih
+                gpu::resize(b_[k], c_[k], lowResSize, 0, 0, INTER_NEAREST, streams_[k]);
+
+                diffSign(src[k], c_[k], c_[k], streams_[k]);
+
+                // a = Dt * diff
+                upscale(c_[k], a_[k], scale_, streams_[k]);
+                // b = HtDt * diff
+                filters_[k]->apply(a_[k], b_[k], streams_[k]);
+                // diffTerm = MtHtDt * diff
+                gpu::remap(b_[k], diffTerms_[k], forwardMaps_[k].first, forwardMaps_[k].second, INTER_NEAREST, BORDER_REPLICATE, Scalar(), streams_[k]);
+            }
+
+            if (lambda_ > 0)
+            {
+                calcBtvRegularization(highRes_, regTerm_, btvKernelSize_);
+                gpu::addWeighted(highRes_, 1.0, regTerm_, -tau_ * lambda_, 0.0, highRes_);
+            }
+
+            for (size_t k = 0; k < src.size(); ++k)
+            {
+                streams_[k].waitForCompletion();
+                gpu::addWeighted(highRes_, 1.0, diffTerms_[k], tau_, 0.0, highRes_);
+            }
+        }
+
+        Rect inner(btvKernelSize_, btvKernelSize_, highRes_.cols - 2 * btvKernelSize_, highRes_.rows - 2 * btvKernelSize_);
+        highRes_(inner).copyTo(dst);
+    }
+
+    void BTVL1_GPU_Base::collectGarbage()
+    {
+        filters_.clear();
+
+        lowResForwardMotions_.clear();
+        lowResBackwardMotions_.clear();
+
+        highResForwardMotions_.clear();
+        highResBackwardMotions_.clear();
+
+        forwardMaps_.clear();
+        backwardMaps_.clear();
+
+        highRes_.release();
+
+        diffTerms_.clear();
+        a_.clear();
+        b_.clear();
+        c_.clear();
+        regTerm_.release();
+    }
+
+////////////////////////////////////////////////////////////
+
+    class BTVL1_GPU : public SuperResolution, private BTVL1_GPU_Base
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        BTVL1_GPU();
+
+        void collectGarbage();
+
+    protected:
+        void initImpl(Ptr<FrameSource>& frameSource);
+        void processImpl(Ptr<FrameSource>& frameSource, OutputArray output);
+
+    private:
+        int temporalAreaRadius_;
+
+        void readNextFrame(Ptr<FrameSource>& frameSource);
+        void processFrame(int idx);
+
+        GpuMat curFrame_;
+        GpuMat prevFrame_;
+
+        std::vector<GpuMat> frames_;
+        std::vector<std::pair<GpuMat, GpuMat> > forwardMotions_;
+        std::vector<std::pair<GpuMat, GpuMat> > backwardMotions_;
+        std::vector<GpuMat> outputs_;
+
+        int storePos_;
+        int procPos_;
+        int outPos_;
+
+        std::vector<GpuMat> srcFrames_;
+        std::vector<std::pair<GpuMat, GpuMat> > srcForwardMotions_;
+        std::vector<std::pair<GpuMat, GpuMat> > srcBackwardMotions_;
+        GpuMat finalOutput_;
+    };
+
+    CV_INIT_ALGORITHM(BTVL1_GPU, "SuperResolution.BTVL1_GPU",
+                      obj.info()->addParam(obj, "scale", obj.scale_, false, 0, 0, "Scale factor.");
+                      obj.info()->addParam(obj, "iterations", obj.iterations_, false, 0, 0, "Iteration count.");
+                      obj.info()->addParam(obj, "tau", obj.tau_, false, 0, 0, "Asymptotic value of steepest descent method.");
+                      obj.info()->addParam(obj, "lambda", obj.lambda_, false, 0, 0, "Weight parameter to balance data term and smoothness term.");
+                      obj.info()->addParam(obj, "alpha", obj.alpha_, false, 0, 0, "Parameter of spacial distribution in Bilateral-TV.");
+                      obj.info()->addParam(obj, "btvKernelSize", obj.btvKernelSize_, false, 0, 0, "Kernel size of Bilateral-TV filter.");
+                      obj.info()->addParam(obj, "blurKernelSize", obj.blurKernelSize_, false, 0, 0, "Gaussian blur kernel size.");
+                      obj.info()->addParam(obj, "blurSigma", obj.blurSigma_, false, 0, 0, "Gaussian blur sigma.");
+                      obj.info()->addParam(obj, "temporalAreaRadius", obj.temporalAreaRadius_, false, 0, 0, "Radius of the temporal search area.");
+                      obj.info()->addParam<DenseOpticalFlowExt>(obj, "opticalFlow", obj.opticalFlow_, false, 0, 0, "Dense optical flow algorithm."));
+
+    BTVL1_GPU::BTVL1_GPU()
+    {
+        temporalAreaRadius_ = 4;
+    }
+
+    void BTVL1_GPU::collectGarbage()
+    {
+        curFrame_.release();
+        prevFrame_.release();
+
+        frames_.clear();
+        forwardMotions_.clear();
+        backwardMotions_.clear();
+        outputs_.clear();
+
+        srcFrames_.clear();
+        srcForwardMotions_.clear();
+        srcBackwardMotions_.clear();
+        finalOutput_.release();
+
+        SuperResolution::collectGarbage();
+        BTVL1_GPU_Base::collectGarbage();
+    }
+
+    void BTVL1_GPU::initImpl(Ptr<FrameSource>& frameSource)
+    {
+        const int cacheSize = 2 * temporalAreaRadius_ + 1;
+
+        frames_.resize(cacheSize);
+        forwardMotions_.resize(cacheSize);
+        backwardMotions_.resize(cacheSize);
+        outputs_.resize(cacheSize);
+
+        storePos_ = -1;
+
+        for (int t = -temporalAreaRadius_; t <= temporalAreaRadius_; ++t)
+            readNextFrame(frameSource);
+
+        for (int i = 0; i <= temporalAreaRadius_; ++i)
+            processFrame(i);
+
+        procPos_ = temporalAreaRadius_;
+        outPos_ = -1;
+    }
+
+    void BTVL1_GPU::processImpl(Ptr<FrameSource>& frameSource, OutputArray _output)
+    {
+        if (outPos_ >= storePos_)
+        {
+            _output.release();
+            return;
+        }
+
+        readNextFrame(frameSource);
+
+        if (procPos_ < storePos_)
+        {
+            ++procPos_;
+            processFrame(procPos_);
+        }
+
+        ++outPos_;
+        const GpuMat& curOutput = at(outPos_, outputs_);
+
+        if (_output.kind() == _InputArray::GPU_MAT)
+            curOutput.convertTo(_output.getGpuMatRef(), CV_8U);
+        else
+        {
+            curOutput.convertTo(finalOutput_, CV_8U);
+            arrCopy(finalOutput_, _output);
+        }
+    }
+
+    void BTVL1_GPU::readNextFrame(Ptr<FrameSource>& frameSource)
+    {
+        frameSource->nextFrame(curFrame_);
+
+        if (curFrame_.empty())
+            return;
+
+        ++storePos_;
+        curFrame_.convertTo(at(storePos_, frames_), CV_32F);
+
+        if (storePos_ > 0)
+        {
+            std::pair<GpuMat, GpuMat>& forwardMotion = at(storePos_ - 1, forwardMotions_);
+            std::pair<GpuMat, GpuMat>& backwardMotion = at(storePos_, backwardMotions_);
+
+            opticalFlow_->calc(prevFrame_, curFrame_, forwardMotion.first, forwardMotion.second);
+            opticalFlow_->calc(curFrame_, prevFrame_, backwardMotion.first, backwardMotion.second);
+        }
+
+        curFrame_.copyTo(prevFrame_);
+    }
+
+    void BTVL1_GPU::processFrame(int idx)
+    {
+        const int startIdx = std::max(idx - temporalAreaRadius_, 0);
+        const int procIdx = idx;
+        const int endIdx = std::min(startIdx + 2 * temporalAreaRadius_, storePos_);
+
+        const int count = endIdx - startIdx + 1;
+
+        srcFrames_.resize(count);
+        srcForwardMotions_.resize(count);
+        srcBackwardMotions_.resize(count);
+
+        int baseIdx = -1;
+
+        for (int i = startIdx, k = 0; i <= endIdx; ++i, ++k)
+        {
+            if (i == procIdx)
+                baseIdx = k;
+
+            srcFrames_[k] = at(i, frames_);
+
+            if (i < endIdx)
+                srcForwardMotions_[k] = at(i, forwardMotions_);
+            if (i > startIdx)
+                srcBackwardMotions_[k] = at(i, backwardMotions_);
+        }
+
+        process(srcFrames_, at(idx, outputs_), srcForwardMotions_, srcBackwardMotions_, baseIdx);
+    }
+}
+
+Ptr<SuperResolution> cv::superres::createSuperResolution_BTVL1_GPU()
+{
+    return makePtr<BTVL1_GPU>();
+}
+
+#endif // HAVE_CUDA
diff --git a/modules/superres/src/btv_l1_ocl.cpp b/modules/superres/src/btv_l1_ocl.cpp
new file mode 100644
index 0000000000..9d94e61ceb
--- /dev/null
+++ b/modules/superres/src/btv_l1_ocl.cpp
@@ -0,0 +1,748 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//		Jin Ma, jin@multicorewareinc.com
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+// S. Farsiu , D. Robinson, M. Elad, P. Milanfar. Fast and robust multiframe super resolution.
+// Dennis Mitzel, Thomas Pock, Thomas Schoenemann, Daniel Cremers. Video Super Resolution using Duality Based TV-L1 Optical Flow.
+
+#include "precomp.hpp"
+
+#if !defined(HAVE_OPENCL) || !defined(HAVE_OPENCV_OCL)
+
+cv::Ptr<cv::superres::SuperResolution> cv::superres::createSuperResolution_BTVL1_OCL()
+{
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<SuperResolution>();
+}
+
+#else
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+using namespace cv::superres;
+using namespace cv::superres::detail;
+
+namespace cv
+{
+    namespace ocl
+    {
+        extern const char* superres_btvl1;
+
+        float* btvWeights_ = NULL;
+        size_t btvWeights_size = 0;
+    }
+}
+
+namespace btv_l1_device_ocl
+{
+    void buildMotionMaps(const oclMat& forwardMotionX, const oclMat& forwardMotionY,
+        const oclMat& backwardMotionX, const oclMat& bacwardMotionY,
+        oclMat& forwardMapX, oclMat& forwardMapY,
+        oclMat& backwardMapX, oclMat& backwardMapY);
+
+    void upscale(const oclMat& src, oclMat& dst, int scale);
+
+    float diffSign(float a, float b);
+
+    Point3f diffSign(Point3f a, Point3f b);
+
+    void diffSign(const oclMat& src1, const oclMat& src2, oclMat& dst);
+
+    void calcBtvRegularization(const oclMat& src, oclMat& dst, int ksize);
+}
+
+void btv_l1_device_ocl::buildMotionMaps(const oclMat& forwardMotionX, const oclMat& forwardMotionY,
+    const oclMat& backwardMotionX, const oclMat& backwardMotionY,
+    oclMat& forwardMapX, oclMat& forwardMapY,
+    oclMat& backwardMapX, oclMat& backwardMapY)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {forwardMapX.cols, forwardMapX.rows, 1};
+
+    int forwardMotionX_step = (int)(forwardMotionX.step/forwardMotionX.elemSize());
+    int forwardMotionY_step = (int)(forwardMotionY.step/forwardMotionY.elemSize());
+    int backwardMotionX_step = (int)(backwardMotionX.step/backwardMotionX.elemSize());
+    int backwardMotionY_step = (int)(backwardMotionY.step/backwardMotionY.elemSize());
+    int forwardMapX_step = (int)(forwardMapX.step/forwardMapX.elemSize());
+    int forwardMapY_step = (int)(forwardMapY.step/forwardMapY.elemSize());
+    int backwardMapX_step = (int)(backwardMapX.step/backwardMapX.elemSize());
+    int backwardMapY_step = (int)(backwardMapY.step/backwardMapY.elemSize());
+
+    String kernel_name = "buildMotionMapsKernel";
+    vector< pair<size_t, const void*> > args;
+
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&forwardMotionX.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&forwardMotionY.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&backwardMotionX.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&backwardMotionY.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&forwardMapX.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&forwardMapY.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&backwardMapX.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&backwardMapY.data));
+
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMotionX.rows));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMotionY.cols));
+
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMotionX_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMotionY_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&backwardMotionX_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&backwardMotionY_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMapX_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&forwardMapY_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&backwardMapX_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&backwardMapY_step));
+
+    openCLExecuteKernel(clCxt, &superres_btvl1, kernel_name, global_thread, local_thread, args, -1, -1);
+}
+
+void btv_l1_device_ocl::upscale(const oclMat& src, oclMat& dst, int scale)
+{
+    Context* clCxt = Context::getContext();
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {src.cols, src.rows, 1};
+
+    int src_step = (int)(src.step/src.elemSize());
+    int dst_step = (int)(dst.step/dst.elemSize());
+
+    String kernel_name = "upscaleKernel";
+    vector< pair<size_t, const void*> > args;
+
+    int cn = src.oclchannels();
+
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&src.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&dst.data));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&dst_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src.rows));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src.cols));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&scale));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&cn));
+
+    openCLExecuteKernel(clCxt, &superres_btvl1, kernel_name, global_thread, local_thread, args, -1, -1);
+
+}
+
+float btv_l1_device_ocl::diffSign(float a, float b)
+{
+    return a > b ? 1.0f : a < b ? -1.0f : 0.0f;
+}
+
+Point3f btv_l1_device_ocl::diffSign(Point3f a, Point3f b)
+{
+    return Point3f(
+        a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f,
+        a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f,
+        a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f
+        );
+}
+
+void btv_l1_device_ocl::diffSign(const oclMat& src1, const oclMat& src2, oclMat& dst)
+{
+    Context* clCxt = Context::getContext();
+
+    oclMat src1_ = src1.reshape(1);
+    oclMat src2_ = src2.reshape(1);
+    oclMat dst_ = dst.reshape(1);
+
+    int src1_step = (int)(src1_.step/src1_.elemSize());
+    int src2_step = (int)(src2_.step/src2_.elemSize());
+    int dst_step = (int)(dst_.step/dst_.elemSize());
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {src1_.cols, src1_.rows, 1};
+
+    String kernel_name = "diffSignKernel";
+    vector< pair<size_t, const void*> > args;
+
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&src1_.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&src2_.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&dst_.data));
+
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src1_.rows));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src1_.cols));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&dst_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src1_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src2_step));
+
+    openCLExecuteKernel(clCxt, &superres_btvl1, kernel_name, global_thread, local_thread, args, -1, -1);
+}
+
+void btv_l1_device_ocl::calcBtvRegularization(const oclMat& src, oclMat& dst, int ksize)
+{
+    Context* clCxt = Context::getContext();
+
+    oclMat src_ = src.reshape(1);
+    oclMat dst_ = dst.reshape(1);
+
+    size_t local_thread[] = {32, 8, 1};
+    size_t global_thread[] = {src.cols, src.rows, 1};
+
+    int src_step = (int)(src_.step/src_.elemSize());
+    int dst_step = (int)(dst_.step/dst_.elemSize());
+
+    String kernel_name = "calcBtvRegularizationKernel";
+    vector< pair<size_t, const void*> > args;
+
+    int cn = src.oclchannels();
+
+    cl_mem c_btvRegWeights;
+    size_t count = btvWeights_size * sizeof(float);
+    c_btvRegWeights = openCLCreateBuffer(clCxt, CL_MEM_READ_ONLY, count);
+    int cl_safe_check = clEnqueueWriteBuffer((cl_command_queue)clCxt->oclCommandQueue(), c_btvRegWeights, 1, 0, count, btvWeights_, 0, NULL, NULL);
+    CV_Assert(cl_safe_check == CL_SUCCESS);
+
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&src_.data));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&dst_.data));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&dst_step));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src.rows));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&src.cols));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&ksize));
+    args.push_back(make_pair(sizeof(cl_int), (void*)&cn));
+    args.push_back(make_pair(sizeof(cl_mem), (void*)&c_btvRegWeights));
+
+    openCLExecuteKernel(clCxt, &superres_btvl1, kernel_name, global_thread, local_thread, args, -1, -1);
+    cl_safe_check = clReleaseMemObject(c_btvRegWeights);
+    CV_Assert(cl_safe_check == CL_SUCCESS);
+}
+
+namespace
+{
+    void calcRelativeMotions(const vector<pair<oclMat, oclMat> >& forwardMotions, const vector<pair<oclMat, oclMat> >& backwardMotions,
+        vector<pair<oclMat, oclMat> >& relForwardMotions, vector<pair<oclMat, oclMat> >& relBackwardMotions,
+        int baseIdx, Size size)
+    {
+        const int count = static_cast<int>(forwardMotions.size());
+
+        relForwardMotions.resize(count);
+        relForwardMotions[baseIdx].first.create(size, CV_32FC1);
+        relForwardMotions[baseIdx].first.setTo(Scalar::all(0));
+        relForwardMotions[baseIdx].second.create(size, CV_32FC1);
+        relForwardMotions[baseIdx].second.setTo(Scalar::all(0));
+
+        relBackwardMotions.resize(count);
+        relBackwardMotions[baseIdx].first.create(size, CV_32FC1);
+        relBackwardMotions[baseIdx].first.setTo(Scalar::all(0));
+        relBackwardMotions[baseIdx].second.create(size, CV_32FC1);
+        relBackwardMotions[baseIdx].second.setTo(Scalar::all(0));
+
+        for (int i = baseIdx - 1; i >= 0; --i)
+        {
+            ocl::add(relForwardMotions[i + 1].first, forwardMotions[i].first, relForwardMotions[i].first);
+            ocl::add(relForwardMotions[i + 1].second, forwardMotions[i].second, relForwardMotions[i].second);
+
+            ocl::add(relBackwardMotions[i + 1].first, backwardMotions[i + 1].first, relBackwardMotions[i].first);
+            ocl::add(relBackwardMotions[i + 1].second, backwardMotions[i + 1].second, relBackwardMotions[i].second);
+        }
+
+        for (int i = baseIdx + 1; i < count; ++i)
+        {
+            ocl::add(relForwardMotions[i - 1].first, backwardMotions[i].first, relForwardMotions[i].first);
+            ocl::add(relForwardMotions[i - 1].second, backwardMotions[i].second, relForwardMotions[i].second);
+
+            ocl::add(relBackwardMotions[i - 1].first, forwardMotions[i - 1].first, relBackwardMotions[i].first);
+            ocl::add(relBackwardMotions[i - 1].second, forwardMotions[i - 1].second, relBackwardMotions[i].second);
+        }
+    }
+
+    void upscaleMotions(const vector<pair<oclMat, oclMat> >& lowResMotions, vector<pair<oclMat, oclMat> >& highResMotions, int scale)
+    {
+        highResMotions.resize(lowResMotions.size());
+
+        for (size_t i = 0; i < lowResMotions.size(); ++i)
+        {
+            ocl::resize(lowResMotions[i].first, highResMotions[i].first, Size(), scale, scale, INTER_LINEAR);
+            ocl::resize(lowResMotions[i].second, highResMotions[i].second, Size(), scale, scale, INTER_LINEAR);
+
+            ocl::multiply(scale, highResMotions[i].first, highResMotions[i].first);
+            ocl::multiply(scale, highResMotions[i].second, highResMotions[i].second);
+        }
+    }
+
+    void buildMotionMaps(const pair<oclMat, oclMat>& forwardMotion, const pair<oclMat, oclMat>& backwardMotion,
+        pair<oclMat, oclMat>& forwardMap, pair<oclMat, oclMat>& backwardMap)
+    {
+        forwardMap.first.create(forwardMotion.first.size(), CV_32FC1);
+        forwardMap.second.create(forwardMotion.first.size(), CV_32FC1);
+
+        backwardMap.first.create(forwardMotion.first.size(), CV_32FC1);
+        backwardMap.second.create(forwardMotion.first.size(), CV_32FC1);
+
+        btv_l1_device_ocl::buildMotionMaps(forwardMotion.first, forwardMotion.second,
+            backwardMotion.first, backwardMotion.second,
+            forwardMap.first, forwardMap.second,
+            backwardMap.first, backwardMap.second);
+    }
+
+    void upscale(const oclMat& src, oclMat& dst, int scale)
+    {
+        CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+
+        dst.create(src.rows * scale, src.cols * scale, src.type());
+        dst.setTo(Scalar::all(0));
+
+        btv_l1_device_ocl::upscale(src, dst, scale);
+    }
+
+    void diffSign(const oclMat& src1, const oclMat& src2, oclMat& dst)
+    {
+        dst.create(src1.size(), src1.type());
+
+        btv_l1_device_ocl::diffSign(src1, src2, dst);
+    }
+
+    void calcBtvWeights(int btvKernelSize, double alpha, vector<float>& btvWeights)
+    {
+        const size_t size = btvKernelSize * btvKernelSize;
+
+        btvWeights.resize(size);
+
+        const int ksize = (btvKernelSize - 1) / 2;
+        const float alpha_f = static_cast<float>(alpha);
+
+        for (int m = 0, ind = 0; m <= ksize; ++m)
+        {
+            for (int l = ksize; l + m >= 0; --l, ++ind)
+                btvWeights[ind] = pow(alpha_f, std::abs(m) + std::abs(l));
+        }
+
+        btvWeights_ = &btvWeights[0];
+        btvWeights_size = size;
+    }
+
+    void calcBtvRegularization(const oclMat& src, oclMat& dst, int btvKernelSize)
+    {
+        dst.create(src.size(), src.type());
+        dst.setTo(Scalar::all(0));
+
+        const int ksize = (btvKernelSize - 1) / 2;
+
+        btv_l1_device_ocl::calcBtvRegularization(src, dst, ksize);
+    }
+
+    class BTVL1_OCL_Base
+    {
+    public:
+        BTVL1_OCL_Base();
+
+        void process(const vector<oclMat>& src, oclMat& dst,
+            const vector<pair<oclMat, oclMat> >& forwardMotions, const vector<pair<oclMat, oclMat> >& backwardMotions,
+            int baseIdx);
+
+        void collectGarbage();
+
+    protected:
+        int scale_;
+        int iterations_;
+        double lambda_;
+        double tau_;
+        double alpha_;
+        int btvKernelSize_;
+        int blurKernelSize_;
+        double blurSigma_;
+        Ptr<DenseOpticalFlowExt> opticalFlow_;
+
+    private:
+        vector<Ptr<cv::ocl::FilterEngine_GPU> > filters_;
+        int curBlurKernelSize_;
+        double curBlurSigma_;
+        int curSrcType_;
+
+        vector<float> btvWeights_;
+        int curBtvKernelSize_;
+        double curAlpha_;
+
+        vector<pair<oclMat, oclMat> > lowResForwardMotions_;
+        vector<pair<oclMat, oclMat> > lowResBackwardMotions_;
+
+        vector<pair<oclMat, oclMat> > highResForwardMotions_;
+        vector<pair<oclMat, oclMat> > highResBackwardMotions_;
+
+        vector<pair<oclMat, oclMat> > forwardMaps_;
+        vector<pair<oclMat, oclMat> > backwardMaps_;
+
+        oclMat highRes_;
+
+        vector<oclMat> diffTerms_;
+        vector<oclMat> a_, b_, c_;
+        oclMat regTerm_;
+    };
+
+    BTVL1_OCL_Base::BTVL1_OCL_Base()
+    {
+        scale_ = 4;
+        iterations_ = 180;
+        lambda_ = 0.03;
+        tau_ = 1.3;
+        alpha_ = 0.7;
+        btvKernelSize_ = 7;
+        blurKernelSize_ = 5;
+        blurSigma_ = 0.0;
+        opticalFlow_ = createOptFlow_DualTVL1_OCL();
+
+        curBlurKernelSize_ = -1;
+        curBlurSigma_ = -1.0;
+        curSrcType_ = -1;
+
+        curBtvKernelSize_ = -1;
+        curAlpha_ = -1.0;
+    }
+
+    void BTVL1_OCL_Base::process(const vector<oclMat>& src, oclMat& dst,
+        const vector<pair<oclMat, oclMat> >& forwardMotions, const vector<pair<oclMat, oclMat> >& backwardMotions,
+        int baseIdx)
+    {
+        CV_Assert( scale_ > 1 );
+        CV_Assert( iterations_ > 0 );
+        CV_Assert( tau_ > 0.0 );
+        CV_Assert( alpha_ > 0.0 );
+        CV_Assert( btvKernelSize_ > 0 && btvKernelSize_ <= 16 );
+        CV_Assert( blurKernelSize_ > 0 );
+        CV_Assert( blurSigma_ >= 0.0 );
+
+        // update blur filter and btv weights
+
+        if (filters_.size() != src.size() || blurKernelSize_ != curBlurKernelSize_ || blurSigma_ != curBlurSigma_ || src[0].type() != curSrcType_)
+        {
+            filters_.resize(src.size());
+            for (size_t i = 0; i < src.size(); ++i)
+                filters_[i] = cv::ocl::createGaussianFilter_GPU(src[0].type(), Size(blurKernelSize_, blurKernelSize_), blurSigma_);
+            curBlurKernelSize_ = blurKernelSize_;
+            curBlurSigma_ = blurSigma_;
+            curSrcType_ = src[0].type();
+        }
+
+        if (btvWeights_.empty() || btvKernelSize_ != curBtvKernelSize_ || alpha_ != curAlpha_)
+        {
+            calcBtvWeights(btvKernelSize_, alpha_, btvWeights_);
+            curBtvKernelSize_ = btvKernelSize_;
+            curAlpha_ = alpha_;
+        }
+
+        // calc motions between input frames
+
+        calcRelativeMotions(forwardMotions, backwardMotions,
+            lowResForwardMotions_, lowResBackwardMotions_,
+            baseIdx, src[0].size());
+
+        upscaleMotions(lowResForwardMotions_, highResForwardMotions_, scale_);
+        upscaleMotions(lowResBackwardMotions_, highResBackwardMotions_, scale_);
+
+        forwardMaps_.resize(highResForwardMotions_.size());
+        backwardMaps_.resize(highResForwardMotions_.size());
+        for (size_t i = 0; i < highResForwardMotions_.size(); ++i)
+        {
+            buildMotionMaps(highResForwardMotions_[i], highResBackwardMotions_[i], forwardMaps_[i], backwardMaps_[i]);
+        }
+        // initial estimation
+
+        const Size lowResSize = src[0].size();
+        const Size highResSize(lowResSize.width * scale_, lowResSize.height * scale_);
+
+        ocl::resize(src[baseIdx], highRes_, highResSize, 0, 0, INTER_LINEAR);
+
+        // iterations
+
+        diffTerms_.resize(src.size());
+        a_.resize(src.size());
+        b_.resize(src.size());
+        c_.resize(src.size());
+
+        for (int i = 0; i < iterations_; ++i)
+        {
+            for (size_t k = 0; k < src.size(); ++k)
+            {
+                diffTerms_[k].create(highRes_.size(), highRes_.type());
+                a_[k].create(highRes_.size(), highRes_.type());
+                b_[k].create(highRes_.size(), highRes_.type());
+                c_[k].create(lowResSize, highRes_.type());
+
+                // a = M * Ih
+                ocl::remap(highRes_, a_[k], backwardMaps_[k].first, backwardMaps_[k].second, INTER_NEAREST, BORDER_CONSTANT, Scalar());
+                // b = HM * Ih
+                filters_[k]->apply(a_[k], b_[k], Rect(0,0,-1,-1));
+                // c = DHF * Ih
+                ocl::resize(b_[k], c_[k], lowResSize, 0, 0, INTER_NEAREST);
+
+                diffSign(src[k], c_[k], c_[k]);
+
+                // a = Dt * diff
+                upscale(c_[k], a_[k], scale_);
+                // b = HtDt * diff
+                filters_[k]->apply(a_[k], b_[k], Rect(0,0,-1,-1));
+                // diffTerm = MtHtDt * diff
+                ocl::remap(b_[k], diffTerms_[k], forwardMaps_[k].first, forwardMaps_[k].second, INTER_NEAREST, BORDER_CONSTANT, Scalar());
+            }
+
+            if (lambda_ > 0)
+            {
+                calcBtvRegularization(highRes_, regTerm_, btvKernelSize_);
+                ocl::addWeighted(highRes_, 1.0, regTerm_, -tau_ * lambda_, 0.0, highRes_);
+            }
+
+            for (size_t k = 0; k < src.size(); ++k)
+            {
+                ocl::addWeighted(highRes_, 1.0, diffTerms_[k], tau_, 0.0, highRes_);
+            }
+        }
+
+        Rect inner(btvKernelSize_, btvKernelSize_, highRes_.cols - 2 * btvKernelSize_, highRes_.rows - 2 * btvKernelSize_);
+        highRes_(inner).copyTo(dst);
+    }
+
+    void BTVL1_OCL_Base::collectGarbage()
+    {
+        filters_.clear();
+
+        lowResForwardMotions_.clear();
+        lowResBackwardMotions_.clear();
+
+        highResForwardMotions_.clear();
+        highResBackwardMotions_.clear();
+
+        forwardMaps_.clear();
+        backwardMaps_.clear();
+
+        highRes_.release();
+
+        diffTerms_.clear();
+        a_.clear();
+        b_.clear();
+        c_.clear();
+        regTerm_.release();
+    }
+
+    ////////////////////////////////////////////////////////////
+
+    class BTVL1_OCL : public SuperResolution, private BTVL1_OCL_Base
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        BTVL1_OCL();
+
+        void collectGarbage();
+
+    protected:
+        void initImpl(Ptr<FrameSource>& frameSource);
+        void processImpl(Ptr<FrameSource>& frameSource, OutputArray output);
+
+    private:
+        int temporalAreaRadius_;
+
+        void readNextFrame(Ptr<FrameSource>& frameSource);
+        void processFrame(int idx);
+
+        oclMat curFrame_;
+        oclMat prevFrame_;
+
+        vector<oclMat> frames_;
+        vector<pair<oclMat, oclMat> > forwardMotions_;
+        vector<pair<oclMat, oclMat> > backwardMotions_;
+        vector<oclMat> outputs_;
+
+        int storePos_;
+        int procPos_;
+        int outPos_;
+
+        vector<oclMat> srcFrames_;
+        vector<pair<oclMat, oclMat> > srcForwardMotions_;
+        vector<pair<oclMat, oclMat> > srcBackwardMotions_;
+        oclMat finalOutput_;
+    };
+
+    CV_INIT_ALGORITHM(BTVL1_OCL, "SuperResolution.BTVL1_OCL",
+    obj.info()->addParam(obj, "scale", obj.scale_, false, 0, 0, "Scale factor.");
+    obj.info()->addParam(obj, "iterations", obj.iterations_, false, 0, 0, "Iteration count.");
+    obj.info()->addParam(obj, "tau", obj.tau_, false, 0, 0, "Asymptotic value of steepest descent method.");
+    obj.info()->addParam(obj, "lambda", obj.lambda_, false, 0, 0, "Weight parameter to balance data term and smoothness term.");
+    obj.info()->addParam(obj, "alpha", obj.alpha_, false, 0, 0, "Parameter of spacial distribution in Bilateral-TV.");
+    obj.info()->addParam(obj, "btvKernelSize", obj.btvKernelSize_, false, 0, 0, "Kernel size of Bilateral-TV filter.");
+    obj.info()->addParam(obj, "blurKernelSize", obj.blurKernelSize_, false, 0, 0, "Gaussian blur kernel size.");
+    obj.info()->addParam(obj, "blurSigma", obj.blurSigma_, false, 0, 0, "Gaussian blur sigma.");
+    obj.info()->addParam(obj, "temporalAreaRadius", obj.temporalAreaRadius_, false, 0, 0, "Radius of the temporal search area.");
+    obj.info()->addParam<DenseOpticalFlowExt>(obj, "opticalFlow", obj.opticalFlow_, false, 0, 0, "Dense optical flow algorithm."));
+
+    BTVL1_OCL::BTVL1_OCL()
+    {
+        temporalAreaRadius_ = 4;
+    }
+
+    void BTVL1_OCL::collectGarbage()
+    {
+        curFrame_.release();
+        prevFrame_.release();
+
+        frames_.clear();
+        forwardMotions_.clear();
+        backwardMotions_.clear();
+        outputs_.clear();
+
+        srcFrames_.clear();
+        srcForwardMotions_.clear();
+        srcBackwardMotions_.clear();
+        finalOutput_.release();
+
+        SuperResolution::collectGarbage();
+        BTVL1_OCL_Base::collectGarbage();
+    }
+
+    void BTVL1_OCL::initImpl(Ptr<FrameSource>& frameSource)
+    {
+        const int cacheSize = 2 * temporalAreaRadius_ + 1;
+
+        frames_.resize(cacheSize);
+        forwardMotions_.resize(cacheSize);
+        backwardMotions_.resize(cacheSize);
+        outputs_.resize(cacheSize);
+
+        storePos_ = -1;
+
+        for (int t = -temporalAreaRadius_; t <= temporalAreaRadius_; ++t)
+            readNextFrame(frameSource);
+
+        for (int i = 0; i <= temporalAreaRadius_; ++i)
+            processFrame(i);
+
+        procPos_ = temporalAreaRadius_;
+        outPos_ = -1;
+    }
+
+    void BTVL1_OCL::processImpl(Ptr<FrameSource>& frameSource, OutputArray _output)
+    {
+        if (outPos_ >= storePos_)
+        {
+            if(_output.kind() == _InputArray::OCL_MAT)
+            {
+                getOclMatRef(_output).release();
+            }
+            else
+            {
+                _output.release();
+            }
+            return;
+        }
+
+        readNextFrame(frameSource);
+
+        if (procPos_ < storePos_)
+        {
+            ++procPos_;
+            processFrame(procPos_);
+        }
+
+        ++outPos_;
+        const oclMat& curOutput = at(outPos_, outputs_);
+
+        if (_output.kind() == _InputArray::OCL_MAT)
+            curOutput.convertTo(getOclMatRef(_output), CV_8U);
+        else
+        {
+            curOutput.convertTo(finalOutput_, CV_8U);
+            arrCopy(finalOutput_, _output);
+        }
+    }
+
+    void BTVL1_OCL::readNextFrame(Ptr<FrameSource>& frameSource)
+    {
+        curFrame_.release();
+        frameSource->nextFrame(curFrame_);
+
+        if (curFrame_.empty())
+            return;
+
+        ++storePos_;
+        curFrame_.convertTo(at(storePos_, frames_), CV_32F);
+
+        if (storePos_ > 0)
+        {
+            pair<oclMat, oclMat>& forwardMotion = at(storePos_ - 1, forwardMotions_);
+            pair<oclMat, oclMat>& backwardMotion = at(storePos_, backwardMotions_);
+
+            opticalFlow_->calc(prevFrame_, curFrame_, forwardMotion.first, forwardMotion.second);
+            opticalFlow_->calc(curFrame_, prevFrame_, backwardMotion.first, backwardMotion.second);
+        }
+
+        curFrame_.copyTo(prevFrame_);
+    }
+
+    void BTVL1_OCL::processFrame(int idx)
+    {
+        const int startIdx = max(idx - temporalAreaRadius_, 0);
+        const int procIdx = idx;
+        const int endIdx = min(startIdx + 2 * temporalAreaRadius_, storePos_);
+
+        const int count = endIdx - startIdx + 1;
+
+        srcFrames_.resize(count);
+        srcForwardMotions_.resize(count);
+        srcBackwardMotions_.resize(count);
+
+        int baseIdx = -1;
+
+        for (int i = startIdx, k = 0; i <= endIdx; ++i, ++k)
+        {
+            if (i == procIdx)
+                baseIdx = k;
+
+            srcFrames_[k] = at(i, frames_);
+
+            if (i < endIdx)
+                srcForwardMotions_[k] = at(i, forwardMotions_);
+            if (i > startIdx)
+                srcBackwardMotions_[k] = at(i, backwardMotions_);
+        }
+
+        process(srcFrames_, at(idx, outputs_), srcForwardMotions_, srcBackwardMotions_, baseIdx);
+    }
+}
+
+Ptr<SuperResolution> cv::superres::createSuperResolution_BTVL1_OCL()
+{
+    return makePtr<BTVL1_OCL>();
+}
+#endif
diff --git a/modules/superres/src/cuda/btv_l1_gpu.cu b/modules/superres/src/cuda/btv_l1_gpu.cu
new file mode 100644
index 0000000000..22b3e0abf5
--- /dev/null
+++ b/modules/superres/src/cuda/btv_l1_gpu.cu
@@ -0,0 +1,240 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/opencv_modules.hpp"
+
+#if defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUWARPING) && defined(HAVE_OPENCV_GPUFILTERS)
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/transform.hpp"
+#include "opencv2/core/cuda/vec_traits.hpp"
+#include "opencv2/core/cuda/vec_math.hpp"
+
+using namespace cv::gpu;
+using namespace cv::gpu::cudev;
+
+namespace btv_l1_cudev
+{
+    void buildMotionMaps(PtrStepSzf forwardMotionX, PtrStepSzf forwardMotionY,
+                         PtrStepSzf backwardMotionX, PtrStepSzf bacwardMotionY,
+                         PtrStepSzf forwardMapX, PtrStepSzf forwardMapY,
+                         PtrStepSzf backwardMapX, PtrStepSzf backwardMapY);
+
+    template <int cn>
+    void upscale(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+
+    void diffSign(PtrStepSzf src1, PtrStepSzf src2, PtrStepSzf dst, cudaStream_t stream);
+
+    void loadBtvWeights(const float* weights, size_t count);
+    template <int cn> void calcBtvRegularization(PtrStepSzb src, PtrStepSzb dst, int ksize);
+}
+
+namespace btv_l1_cudev
+{
+    __global__ void buildMotionMapsKernel(const PtrStepSzf forwardMotionX, const PtrStepf forwardMotionY,
+                                          PtrStepf backwardMotionX, PtrStepf backwardMotionY,
+                                          PtrStepf forwardMapX, PtrStepf forwardMapY,
+                                          PtrStepf backwardMapX, PtrStepf backwardMapY)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= forwardMotionX.cols || y >= forwardMotionX.rows)
+            return;
+
+        const float fx = forwardMotionX(y, x);
+        const float fy = forwardMotionY(y, x);
+
+        const float bx = backwardMotionX(y, x);
+        const float by = backwardMotionY(y, x);
+
+        forwardMapX(y, x) = x + bx;
+        forwardMapY(y, x) = y + by;
+
+        backwardMapX(y, x) = x + fx;
+        backwardMapY(y, x) = y + fy;
+    }
+
+    void buildMotionMaps(PtrStepSzf forwardMotionX, PtrStepSzf forwardMotionY,
+                         PtrStepSzf backwardMotionX, PtrStepSzf bacwardMotionY,
+                         PtrStepSzf forwardMapX, PtrStepSzf forwardMapY,
+                         PtrStepSzf backwardMapX, PtrStepSzf backwardMapY)
+    {
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(forwardMapX.cols, block.x), divUp(forwardMapX.rows, block.y));
+
+        buildMotionMapsKernel<<<grid, block>>>(forwardMotionX, forwardMotionY,
+                                               backwardMotionX, bacwardMotionY,
+                                               forwardMapX, forwardMapY,
+                                               backwardMapX, backwardMapY);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template <typename T>
+    __global__ void upscaleKernel(const PtrStepSz<T> src, PtrStep<T> dst, const int scale)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y;
+
+        if (x >= src.cols || y >= src.rows)
+            return;
+
+        dst(y * scale, x * scale) = src(y, x);
+    }
+
+    template <int cn>
+    void upscale(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream)
+    {
+        typedef typename TypeVec<float, cn>::vec_type src_t;
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, block.x), divUp(src.rows, block.y));
+
+        upscaleKernel<src_t><<<grid, block, 0, stream>>>((PtrStepSz<src_t>) src, (PtrStepSz<src_t>) dst, scale);
+        cudaSafeCall( cudaGetLastError() );
+
+        if (stream == 0)
+            cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void upscale<1>(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+    template void upscale<3>(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+    template void upscale<4>(const PtrStepSzb src, PtrStepSzb dst, int scale, cudaStream_t stream);
+
+    __device__ __forceinline__ float diffSign(float a, float b)
+    {
+        return a > b ? 1.0f : a < b ? -1.0f : 0.0f;
+    }
+    __device__ __forceinline__ float3 diffSign(const float3& a, const float3& b)
+    {
+        return make_float3(
+            a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f,
+            a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f,
+            a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f
+        );
+    }
+    __device__ __forceinline__ float4 diffSign(const float4& a, const float4& b)
+    {
+        return make_float4(
+            a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f,
+            a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f,
+            a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f,
+            0.0f
+        );
+    }
+
+    struct DiffSign : binary_function<float, float, float>
+    {
+        __device__ __forceinline__ float operator ()(float a, float b) const
+        {
+            return diffSign(a, b);
+        }
+    };
+}
+
+namespace cv { namespace gpu { namespace cudev
+{
+    template <> struct TransformFunctorTraits<btv_l1_cudev::DiffSign> : DefaultTransformFunctorTraits<btv_l1_cudev::DiffSign>
+    {
+        enum { smart_block_dim_y = 8 };
+        enum { smart_shift = 4 };
+    };
+}}}
+
+namespace btv_l1_cudev
+{
+    void diffSign(PtrStepSzf src1, PtrStepSzf src2, PtrStepSzf dst, cudaStream_t stream)
+    {
+        transform(src1, src2, dst, DiffSign(), WithOutMask(), stream);
+    }
+
+    __constant__ float c_btvRegWeights[16*16];
+
+    template <typename T>
+    __global__ void calcBtvRegularizationKernel(const PtrStepSz<T> src, PtrStep<T> dst, const int ksize)
+    {
+        const int x = blockIdx.x * blockDim.x + threadIdx.x + ksize;
+        const int y = blockIdx.y * blockDim.y + threadIdx.y + ksize;
+
+        if (y >= src.rows - ksize || x >= src.cols - ksize)
+            return;
+
+        const T srcVal = src(y, x);
+
+        T dstVal = VecTraits<T>::all(0);
+
+        for (int m = 0, count = 0; m <= ksize; ++m)
+        {
+            for (int l = ksize; l + m >= 0; --l, ++count)
+                dstVal = dstVal + c_btvRegWeights[count] * (diffSign(srcVal, src(y + m, x + l)) - diffSign(src(y - m, x - l), srcVal));
+        }
+
+        dst(y, x) = dstVal;
+    }
+
+    void loadBtvWeights(const float* weights, size_t count)
+    {
+        cudaSafeCall( cudaMemcpyToSymbol(c_btvRegWeights, weights, count * sizeof(float)) );
+    }
+
+    template <int cn>
+    void calcBtvRegularization(PtrStepSzb src, PtrStepSzb dst, int ksize)
+    {
+        typedef typename TypeVec<float, cn>::vec_type src_t;
+
+        const dim3 block(32, 8);
+        const dim3 grid(divUp(src.cols, block.x), divUp(src.rows, block.y));
+
+        calcBtvRegularizationKernel<src_t><<<grid, block>>>((PtrStepSz<src_t>) src, (PtrStepSz<src_t>) dst, ksize);
+        cudaSafeCall( cudaGetLastError() );
+
+        cudaSafeCall( cudaDeviceSynchronize() );
+    }
+
+    template void calcBtvRegularization<1>(PtrStepSzb src, PtrStepSzb dst, int ksize);
+    template void calcBtvRegularization<3>(PtrStepSzb src, PtrStepSzb dst, int ksize);
+    template void calcBtvRegularization<4>(PtrStepSzb src, PtrStepSzb dst, int ksize);
+}
+
+#endif
diff --git a/modules/superres/src/frame_source.cpp b/modules/superres/src/frame_source.cpp
new file mode 100644
index 0000000000..14576d5dd3
--- /dev/null
+++ b/modules/superres/src/frame_source.cpp
@@ -0,0 +1,263 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::superres;
+using namespace cv::superres::detail;
+
+cv::superres::FrameSource::~FrameSource()
+{
+}
+
+//////////////////////////////////////////////////////
+// EmptyFrameSource
+
+namespace
+{
+    class EmptyFrameSource : public FrameSource
+    {
+    public:
+        void nextFrame(OutputArray frame);
+        void reset();
+    };
+
+    void EmptyFrameSource::nextFrame(OutputArray frame)
+    {
+        frame.release();
+    }
+
+    void EmptyFrameSource::reset()
+    {
+    }
+}
+
+Ptr<FrameSource> cv::superres::createFrameSource_Empty()
+{
+    return makePtr<EmptyFrameSource>();
+}
+
+//////////////////////////////////////////////////////
+// VideoFrameSource & CameraFrameSource
+
+#ifndef HAVE_OPENCV_HIGHGUI
+
+Ptr<FrameSource> cv::superres::createFrameSource_Video(const String& fileName)
+{
+    (void) fileName;
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<FrameSource>();
+}
+
+Ptr<FrameSource> cv::superres::createFrameSource_Camera(int deviceId)
+{
+    (void) deviceId;
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<FrameSource>();
+}
+
+#else // HAVE_OPENCV_HIGHGUI
+
+namespace
+{
+    class CaptureFrameSource : public FrameSource
+    {
+    public:
+        void nextFrame(OutputArray frame);
+
+    protected:
+        VideoCapture vc_;
+
+    private:
+        Mat frame_;
+    };
+
+    void CaptureFrameSource::nextFrame(OutputArray _frame)
+    {
+        if (_frame.kind() == _InputArray::MAT)
+        {
+            vc_ >> _frame.getMatRef();
+        }
+        else if(_frame.kind() == _InputArray::GPU_MAT)
+        {
+            vc_ >> frame_;
+            arrCopy(frame_, _frame);
+        }
+        else if(_frame.kind() == _InputArray::OCL_MAT)
+        {
+            vc_ >> frame_;
+            if(!frame_.empty())
+            {
+                arrCopy(frame_, _frame);
+            }
+        }
+        else
+        {
+            //should never get here
+        }
+    }
+
+    class VideoFrameSource : public CaptureFrameSource
+    {
+    public:
+        VideoFrameSource(const String& fileName);
+
+        void reset();
+
+    private:
+        String fileName_;
+    };
+
+    VideoFrameSource::VideoFrameSource(const String& fileName) : fileName_(fileName)
+    {
+        reset();
+    }
+
+    void VideoFrameSource::reset()
+    {
+        vc_.release();
+        vc_.open(fileName_);
+        CV_Assert( vc_.isOpened() );
+    }
+
+    class CameraFrameSource : public CaptureFrameSource
+    {
+    public:
+        CameraFrameSource(int deviceId);
+
+        void reset();
+
+    private:
+        int deviceId_;
+    };
+
+    CameraFrameSource::CameraFrameSource(int deviceId) : deviceId_(deviceId)
+    {
+        reset();
+    }
+
+    void CameraFrameSource::reset()
+    {
+        vc_.release();
+        vc_.open(deviceId_);
+        CV_Assert( vc_.isOpened() );
+    }
+}
+
+Ptr<FrameSource> cv::superres::createFrameSource_Video(const String& fileName)
+{
+    return makePtr<VideoFrameSource>(fileName);
+}
+
+Ptr<FrameSource> cv::superres::createFrameSource_Camera(int deviceId)
+{
+    return makePtr<CameraFrameSource>(deviceId);
+}
+
+#endif // HAVE_OPENCV_HIGHGUI
+
+//////////////////////////////////////////////////////
+// VideoFrameSource_GPU
+
+#ifndef HAVE_OPENCV_GPUCODEC
+
+Ptr<FrameSource> cv::superres::createFrameSource_Video_GPU(const String& fileName)
+{
+    (void) fileName;
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<FrameSource>();
+}
+
+#else // HAVE_OPENCV_GPUCODEC
+
+namespace
+{
+    class VideoFrameSource_GPU : public FrameSource
+    {
+    public:
+        VideoFrameSource_GPU(const String& fileName);
+
+        void nextFrame(OutputArray frame);
+        void reset();
+
+    private:
+        String fileName_;
+        Ptr<gpucodec::VideoReader> reader_;
+        GpuMat frame_;
+    };
+
+    VideoFrameSource_GPU::VideoFrameSource_GPU(const String& fileName) : fileName_(fileName)
+    {
+        reset();
+    }
+
+    void VideoFrameSource_GPU::nextFrame(OutputArray _frame)
+    {
+        if (_frame.kind() == _InputArray::GPU_MAT)
+        {
+            bool res = reader_->nextFrame(_frame.getGpuMatRef());
+            if (!res)
+                _frame.release();
+        }
+        else
+        {
+            bool res = reader_->nextFrame(frame_);
+            if (!res)
+                _frame.release();
+            else
+                arrCopy(frame_, _frame);
+        }
+    }
+
+    void VideoFrameSource_GPU::reset()
+    {
+        reader_ = gpucodec::createVideoReader(fileName_);
+    }
+}
+
+Ptr<FrameSource> cv::superres::createFrameSource_Video_GPU(const String& fileName)
+{
+    return makePtr<VideoFrameSource>(fileName);
+}
+
+#endif // HAVE_OPENCV_GPUCODEC
diff --git a/modules/superres/src/input_array_utility.cpp b/modules/superres/src/input_array_utility.cpp
new file mode 100644
index 0000000000..8a3cbca55d
--- /dev/null
+++ b/modules/superres/src/input_array_utility.cpp
@@ -0,0 +1,351 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+Mat cv::superres::arrGetMat(InputArray arr, Mat& buf)
+{
+    switch (arr.kind())
+    {
+    case _InputArray::GPU_MAT:
+        arr.getGpuMat().download(buf);
+        return buf;
+
+    case _InputArray::OPENGL_BUFFER:
+        arr.getOGlBuffer().copyTo(buf);
+        return buf;
+
+    default:
+        return arr.getMat();
+    }
+}
+
+GpuMat cv::superres::arrGetGpuMat(InputArray arr, GpuMat& buf)
+{
+    switch (arr.kind())
+    {
+    case _InputArray::GPU_MAT:
+        return arr.getGpuMat();
+
+    case _InputArray::OPENGL_BUFFER:
+        arr.getOGlBuffer().copyTo(buf);
+        return buf;
+
+    default:
+        buf.upload(arr.getMat());
+        return buf;
+    }
+}
+
+namespace
+{
+    void mat2mat(InputArray src, OutputArray dst)
+    {
+        src.getMat().copyTo(dst);
+    }
+    void arr2buf(InputArray src, OutputArray dst)
+    {
+        dst.getOGlBufferRef().copyFrom(src);
+    }
+    void mat2gpu(InputArray src, OutputArray dst)
+    {
+        dst.getGpuMatRef().upload(src.getMat());
+    }
+    void buf2arr(InputArray src, OutputArray dst)
+    {
+        src.getOGlBuffer().copyTo(dst);
+    }
+    void gpu2mat(InputArray src, OutputArray dst)
+    {
+        GpuMat d = src.getGpuMat();
+        dst.create(d.size(), d.type());
+        Mat m = dst.getMat();
+        d.download(m);
+    }
+    void gpu2gpu(InputArray src, OutputArray dst)
+    {
+        src.getGpuMat().copyTo(dst.getGpuMatRef());
+    }
+#ifdef HAVE_OPENCV_OCL
+    void ocl2mat(InputArray src, OutputArray dst)
+    {
+        dst.getMatRef() = (Mat)ocl::getOclMatRef(src);
+    }
+    void mat2ocl(InputArray src, OutputArray dst)
+    {
+        Mat m = src.getMat();
+        ocl::getOclMatRef(dst) = (ocl::oclMat)m;
+    }
+    void ocl2ocl(InputArray src, OutputArray dst)
+    {
+        ocl::getOclMatRef(src).copyTo(ocl::getOclMatRef(dst));
+    }
+#else
+    void ocl2mat(InputArray, OutputArray)
+    {
+        CV_Error(Error::StsNotImplemented, "The called functionality is disabled for current build or platform");;
+    }
+    void mat2ocl(InputArray, OutputArray)
+    {
+        CV_Error(Error::StsNotImplemented, "The called functionality is disabled for current build or platform");;
+    }
+    void ocl2ocl(InputArray, OutputArray)
+    {
+        CV_Error(Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    }
+#endif
+}
+
+void cv::superres::arrCopy(InputArray src, OutputArray dst)
+{
+    typedef void (*func_t)(InputArray src, OutputArray dst);
+    static const func_t funcs[11][11] =
+    {
+        {0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, mat2mat, arr2buf, 0 /*arr2tex*/, mat2gpu, mat2ocl},
+        {0, buf2arr, buf2arr, buf2arr, buf2arr, buf2arr, buf2arr, buf2arr, 0 /*buf2arr*/, buf2arr, 0      },
+        {0, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0 /*tex2arr*/, 0},
+        {0, gpu2mat, gpu2mat, gpu2mat, gpu2mat, gpu2mat, gpu2mat, arr2buf, 0 /*arr2tex*/, gpu2gpu, 0      },
+        {0, ocl2mat, ocl2mat, ocl2mat, ocl2mat, ocl2mat, ocl2mat, 0,       0,             0,       ocl2ocl}
+    };
+
+    const int src_kind = src.kind() >> _InputArray::KIND_SHIFT;
+    const int dst_kind = dst.kind() >> _InputArray::KIND_SHIFT;
+
+    CV_DbgAssert( src_kind >= 0 && src_kind < 11 );
+    CV_DbgAssert( dst_kind >= 0 && dst_kind < 11 );
+
+    const func_t func = funcs[src_kind][dst_kind];
+    CV_DbgAssert( func != 0 );
+
+    func(src, dst);
+}
+
+namespace
+{
+    void convertToCn(InputArray src, OutputArray dst, int cn)
+    {
+        CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( cn == 1 || cn == 3 || cn == 4 );
+
+        static const int codes[5][5] =
+        {
+            {-1, -1, -1, -1, -1},
+            {-1, -1, -1, COLOR_GRAY2BGR, COLOR_GRAY2BGRA},
+            {-1, -1, -1, -1, -1},
+            {-1, COLOR_BGR2GRAY, -1, -1, COLOR_BGR2BGRA},
+            {-1, COLOR_BGRA2GRAY, -1, COLOR_BGRA2BGR, -1},
+        };
+
+        const int code = codes[src.channels()][cn];
+        CV_DbgAssert( code >= 0 );
+
+        switch (src.kind())
+        {
+        case _InputArray::GPU_MAT:
+            #ifdef HAVE_OPENCV_GPUIMGPROC
+                gpu::cvtColor(src.getGpuMat(), dst.getGpuMatRef(), code, cn);
+            #else
+                CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+            #endif
+            break;
+
+        default:
+            cv::cvtColor(src, dst, code, cn);
+            break;
+        }
+    }
+    void convertToDepth(InputArray src, OutputArray dst, int depth)
+    {
+        CV_Assert( src.depth() <= CV_64F );
+        CV_Assert( depth == CV_8U || depth == CV_32F );
+
+        static const double maxVals[] =
+        {
+            std::numeric_limits<uchar>::max(),
+            std::numeric_limits<schar>::max(),
+            std::numeric_limits<ushort>::max(),
+            std::numeric_limits<short>::max(),
+            std::numeric_limits<int>::max(),
+            1.0,
+            1.0,
+        };
+
+        const double scale = maxVals[depth] / maxVals[src.depth()];
+
+        switch (src.kind())
+        {
+        case _InputArray::GPU_MAT:
+            src.getGpuMat().convertTo(dst.getGpuMatRef(), depth, scale);
+            break;
+
+        default:
+            src.getMat().convertTo(dst, depth, scale);
+            break;
+        }
+    }
+}
+
+Mat cv::superres::convertToType(const Mat& src, int type, Mat& buf0, Mat& buf1)
+{
+    if (src.type() == type)
+        return src;
+
+    const int depth = CV_MAT_DEPTH(type);
+    const int cn = CV_MAT_CN(type);
+
+    if (src.depth() == depth)
+    {
+        convertToCn(src, buf0, cn);
+        return buf0;
+    }
+
+    if (src.channels() == cn)
+    {
+        convertToDepth(src, buf1, depth);
+        return buf1;
+    }
+
+    convertToCn(src, buf0, cn);
+    convertToDepth(buf0, buf1, depth);
+    return buf1;
+}
+
+GpuMat cv::superres::convertToType(const GpuMat& src, int type, GpuMat& buf0, GpuMat& buf1)
+{
+    if (src.type() == type)
+        return src;
+
+    const int depth = CV_MAT_DEPTH(type);
+    const int cn = CV_MAT_CN(type);
+
+    if (src.depth() == depth)
+    {
+        convertToCn(src, buf0, cn);
+        return buf0;
+    }
+
+    if (src.channels() == cn)
+    {
+        convertToDepth(src, buf1, depth);
+        return buf1;
+    }
+
+    convertToCn(src, buf0, cn);
+    convertToDepth(buf0, buf1, depth);
+    return buf1;
+}
+#ifdef HAVE_OPENCV_OCL
+namespace
+{
+    // TODO(pengx17): remove these overloaded functions until IntputArray fully supports oclMat
+    void convertToCn(const ocl::oclMat& src, ocl::oclMat& dst, int cn)
+    {
+        CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+        CV_Assert( cn == 1 || cn == 3 || cn == 4 );
+
+        static const int codes[5][5] =
+        {
+            {-1, -1, -1, -1, -1},
+            {-1, -1, -1, COLOR_GRAY2BGR, COLOR_GRAY2BGRA},
+            {-1, -1, -1, -1, -1},
+            {-1, COLOR_BGR2GRAY, -1, -1, COLOR_BGR2BGRA},
+            {-1, COLOR_BGRA2GRAY, -1, COLOR_BGRA2BGR, -1},
+        };
+
+        const int code = codes[src.channels()][cn];
+        CV_DbgAssert( code >= 0 );
+
+        ocl::cvtColor(src, dst, code, cn);
+    }
+    void convertToDepth(const ocl::oclMat& src, ocl::oclMat& dst, int depth)
+    {
+        CV_Assert( src.depth() <= CV_64F );
+        CV_Assert( depth == CV_8U || depth == CV_32F );
+
+        static const double maxVals[] =
+        {
+            std::numeric_limits<uchar>::max(),
+            std::numeric_limits<schar>::max(),
+            std::numeric_limits<ushort>::max(),
+            std::numeric_limits<short>::max(),
+            std::numeric_limits<int>::max(),
+            1.0,
+            1.0,
+        };
+        const double scale = maxVals[depth] / maxVals[src.depth()];
+        src.convertTo(dst, depth, scale);
+    }
+}
+ocl::oclMat cv::superres::convertToType(const ocl::oclMat& src, int type, ocl::oclMat& buf0, ocl::oclMat& buf1)
+{
+    if (src.type() == type)
+        return src;
+
+    const int depth = CV_MAT_DEPTH(type);
+    const int cn = CV_MAT_CN(type);
+
+    if (src.depth() == depth)
+    {
+        convertToCn(src, buf0, cn);
+        return buf0;
+    }
+
+    if (src.channels() == cn)
+    {
+        convertToDepth(src, buf1, depth);
+        return buf1;
+    }
+
+    convertToCn(src, buf0, cn);
+    convertToDepth(buf0, buf1, depth);
+    return buf1;
+}
+#endif
diff --git a/modules/superres/src/input_array_utility.hpp b/modules/superres/src/input_array_utility.hpp
new file mode 100644
index 0000000000..0c86143fc6
--- /dev/null
+++ b/modules/superres/src/input_array_utility.hpp
@@ -0,0 +1,70 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_SUPERRES_INPUT_ARRAY_UTILITY_HPP__
+#define __OPENCV_SUPERRES_INPUT_ARRAY_UTILITY_HPP__
+
+#include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
+#ifdef HAVE_OPENCV_OCL
+#include "opencv2/ocl.hpp"
+#endif
+
+namespace cv
+{
+    namespace superres
+    {
+        CV_EXPORTS Mat arrGetMat(InputArray arr, Mat& buf);
+        CV_EXPORTS gpu::GpuMat arrGetGpuMat(InputArray arr, gpu::GpuMat& buf);
+
+        CV_EXPORTS void arrCopy(InputArray src, OutputArray dst);
+
+        CV_EXPORTS Mat convertToType(const Mat& src, int type, Mat& buf0, Mat& buf1);
+        CV_EXPORTS gpu::GpuMat convertToType(const gpu::GpuMat& src, int type, gpu::GpuMat& buf0, gpu::GpuMat& buf1);
+
+#ifdef HAVE_OPENCV_OCL
+        CV_EXPORTS ocl::oclMat convertToType(const ocl::oclMat& src, int type, ocl::oclMat& buf0, ocl::oclMat& buf1);
+#endif
+    }
+}
+
+#endif // __OPENCV_SUPERRES_INPUT_ARRAY_UTILITY_HPP__
diff --git a/modules/superres/src/opencl/superres_btvl1.cl b/modules/superres/src/opencl/superres_btvl1.cl
new file mode 100644
index 0000000000..4720623234
--- /dev/null
+++ b/modules/superres/src/opencl/superres_btvl1.cl
@@ -0,0 +1,261 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Jin Ma jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other oclMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+__kernel void buildMotionMapsKernel(__global float* forwardMotionX,
+    __global float* forwardMotionY,
+    __global float* backwardMotionX,
+    __global float* backwardMotionY,
+    __global float* forwardMapX,
+    __global float* forwardMapY,
+    __global float* backwardMapX,
+    __global float* backwardMapY,
+    int forwardMotionX_row,
+    int forwardMotionX_col,
+    int forwardMotionX_step,
+    int forwardMotionY_step,
+    int backwardMotionX_step,
+    int backwardMotionY_step,
+    int forwardMapX_step,
+    int forwardMapY_step,
+    int backwardMapX_step,
+    int backwardMapY_step
+    )
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < forwardMotionX_col && y < forwardMotionX_row)
+    {
+        float fx = forwardMotionX[y * forwardMotionX_step + x];
+        float fy = forwardMotionY[y * forwardMotionY_step + x];
+
+        float bx = backwardMotionX[y * backwardMotionX_step + x];
+        float by = backwardMotionY[y * backwardMotionY_step + x];
+
+        forwardMapX[y * forwardMapX_step + x] = x + bx;
+        forwardMapY[y * forwardMapY_step + x] = y + by;
+
+        backwardMapX[y * backwardMapX_step + x] = x + fx;
+        backwardMapY[y * backwardMapY_step + x] = y + fy;
+    }
+}
+
+__kernel void upscaleKernel(__global float* src,
+    __global float* dst,
+    int src_step,
+    int dst_step,
+    int src_row,
+    int src_col,
+    int scale,
+    int channels
+    )
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < src_col && y < src_row)
+    {
+        if(channels == 1)
+        {
+            dst[y * scale * dst_step + x * scale] = src[y * src_step + x];
+        }else if(channels == 3)
+        {
+            dst[y * channels * scale * dst_step + 3 * x * scale + 0] = src[y * channels * src_step + 3 * x + 0];
+            dst[y * channels * scale * dst_step + 3 * x * scale + 1] = src[y * channels * src_step + 3 * x + 1];
+            dst[y * channels * scale * dst_step + 3 * x * scale + 2] = src[y * channels * src_step + 3 * x + 2];
+        }else
+        {
+            dst[y * channels * scale * dst_step + 4 * x * scale + 0] = src[y * channels * src_step + 4 * x + 0];
+            dst[y * channels * scale * dst_step + 4 * x * scale + 1] = src[y * channels * src_step + 4 * x + 1];
+            dst[y * channels * scale * dst_step + 4 * x * scale + 2] = src[y * channels * src_step + 4 * x + 2];
+            dst[y * channels * scale * dst_step + 4 * x * scale + 3] = src[y * channels * src_step + 4 * x + 3];
+        }
+    }
+}
+
+
+float diffSign(float a, float b)
+{
+    return a > b ? 1.0f : a < b ? -1.0f : 0.0f;
+}
+
+float3 diffSign3(float3 a, float3 b)
+{
+    float3 pos;
+    pos.x = a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f;
+    pos.y = a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f;
+    pos.z = a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f;
+    return pos;
+}
+
+float4 diffSign4(float4 a, float4 b)
+{
+    float4 pos;
+    pos.x = a.x > b.x ? 1.0f : a.x < b.x ? -1.0f : 0.0f;
+    pos.y = a.y > b.y ? 1.0f : a.y < b.y ? -1.0f : 0.0f;
+    pos.z = a.z > b.z ? 1.0f : a.z < b.z ? -1.0f : 0.0f;
+    pos.w = 0.0f;
+    return pos;
+}
+
+__kernel void diffSignKernel(__global float* src1,
+    __global float* src2,
+    __global float* dst,
+    int src1_row,
+    int src1_col,
+    int dst_step,
+    int src1_step,
+    int src2_step)
+{
+    int x = get_global_id(0);
+    int y = get_global_id(1);
+
+    if(x < src1_col && y < src1_row)
+    {
+        dst[y * dst_step + x] = diffSign(src1[y * src1_step + x], src2[y * src2_step + x]);
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+}
+
+__kernel void calcBtvRegularizationKernel(__global float* src,
+    __global float* dst,
+    int src_step,
+    int dst_step,
+    int src_row,
+    int src_col,
+    int ksize,
+    int channels,
+    __global float* c_btvRegWeights
+    )
+{
+    int x = get_global_id(0) + ksize;
+    int y = get_global_id(1) + ksize;
+
+    if ((y < src_row - ksize) && (x < src_col - ksize))
+    {
+        if(channels == 1)
+        {
+            const float srcVal = src[y * src_step + x];
+            float dstVal = 0.0f;
+
+            for (int m = 0, count = 0; m <= ksize; ++m)
+            {
+                for (int l = ksize; l + m >= 0; --l, ++count)
+                    dstVal = dstVal + c_btvRegWeights[count] * (diffSign(srcVal, src[(y + m) * src_step + (x + l)]) - diffSign(src[(y - m) * src_step + (x - l)], srcVal));
+            }
+            dst[y * dst_step + x] = dstVal;
+        }else if(channels == 3)
+        {
+            float3 srcVal;
+            srcVal.x = src[y * src_step + 3 * x + 0];
+            srcVal.y = src[y * src_step + 3 * x + 1];
+            srcVal.z = src[y * src_step + 3 * x + 2];
+
+            float3 dstVal;
+            dstVal.x = 0.0f;
+            dstVal.y = 0.0f;
+            dstVal.z = 0.0f;
+
+            for (int m = 0, count = 0; m <= ksize; ++m)
+            {
+                for (int l = ksize; l + m >= 0; --l, ++count)
+                {
+                    float3 src1;
+                    src1.x = src[(y + m) * src_step + 3 * (x + l) + 0];
+                    src1.y = src[(y + m) * src_step + 3 * (x + l) + 1];
+                    src1.z = src[(y + m) * src_step + 3 * (x + l) + 2];
+
+                    float3 src2;
+                    src2.x = src[(y - m) * src_step + 3 * (x - l) + 0];
+                    src2.y = src[(y - m) * src_step + 3 * (x - l) + 1];
+                    src2.z = src[(y - m) * src_step + 3 * (x - l) + 2];
+
+                    dstVal = dstVal + c_btvRegWeights[count] * (diffSign3(srcVal, src1) - diffSign3(src2, srcVal));
+                }
+            }
+            dst[y * dst_step + 3 * x + 0] = dstVal.x;
+            dst[y * dst_step + 3 * x + 1] = dstVal.y;
+            dst[y * dst_step + 3 * x + 2] = dstVal.z;
+        }else
+        {
+            float4 srcVal;
+            srcVal.x = src[y * src_step + 4 * x + 0];//r type =float
+            srcVal.y = src[y * src_step + 4 * x + 1];//g
+            srcVal.z = src[y * src_step + 4 * x + 2];//b
+            srcVal.w = src[y * src_step + 4 * x + 3];//a
+
+            float4 dstVal;
+            dstVal.x = 0.0f;
+            dstVal.y = 0.0f;
+            dstVal.z = 0.0f;
+            dstVal.w = 0.0f;
+
+            for (int m = 0, count = 0; m <= ksize; ++m)
+            {
+                for (int l = ksize; l + m >= 0; --l, ++count)
+                {
+                    float4 src1;
+                    src1.x = src[(y + m) * src_step + 4 * (x + l) + 0];
+                    src1.y = src[(y + m) * src_step + 4 * (x + l) + 1];
+                    src1.z = src[(y + m) * src_step + 4 * (x + l) + 2];
+                    src1.w = src[(y + m) * src_step + 4 * (x + l) + 3];
+
+                    float4 src2;
+                    src2.x = src[(y - m) * src_step + 4 * (x - l) + 0];
+                    src2.y = src[(y - m) * src_step + 4 * (x - l) + 1];
+                    src2.z = src[(y - m) * src_step + 4 * (x - l) + 2];
+                    src2.w = src[(y - m) * src_step + 4 * (x - l) + 3];
+
+                    dstVal = dstVal + c_btvRegWeights[count] * (diffSign4(srcVal, src1) - diffSign4(src2, srcVal));
+
+                }
+            }
+            dst[y * dst_step + 4 * x + 0] = dstVal.x;
+            dst[y * dst_step + 4 * x + 1] = dstVal.y;
+            dst[y * dst_step + 4 * x + 2] = dstVal.z;
+            dst[y * dst_step + 4 * x + 3] = dstVal.w;
+        }
+    }
+}
diff --git a/modules/superres/src/optical_flow.cpp b/modules/superres/src/optical_flow.cpp
new file mode 100644
index 0000000000..85c74115a8
--- /dev/null
+++ b/modules/superres/src/optical_flow.cpp
@@ -0,0 +1,986 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cv::superres;
+using namespace cv::superres::detail;
+
+///////////////////////////////////////////////////////////////////
+// CpuOpticalFlow
+
+namespace
+{
+    class CpuOpticalFlow : public DenseOpticalFlowExt
+    {
+    public:
+        explicit CpuOpticalFlow(int work_type);
+
+        void calc(InputArray frame0, InputArray frame1, OutputArray flow1, OutputArray flow2);
+        void collectGarbage();
+
+    protected:
+        virtual void impl(const Mat& input0, const Mat& input1, OutputArray dst) = 0;
+
+    private:
+        int work_type_;
+        Mat buf_[6];
+        Mat flow_;
+        Mat flows_[2];
+    };
+
+    CpuOpticalFlow::CpuOpticalFlow(int work_type) : work_type_(work_type)
+    {
+    }
+
+    void CpuOpticalFlow::calc(InputArray _frame0, InputArray _frame1, OutputArray _flow1, OutputArray _flow2)
+    {
+        Mat frame0 = arrGetMat(_frame0, buf_[0]);
+        Mat frame1 = arrGetMat(_frame1, buf_[1]);
+
+        CV_Assert( frame1.type() == frame0.type() );
+        CV_Assert( frame1.size() == frame0.size() );
+
+        Mat input0 = convertToType(frame0, work_type_, buf_[2], buf_[3]);
+        Mat input1 = convertToType(frame1, work_type_, buf_[4], buf_[5]);
+
+        if (!_flow2.needed() && _flow1.kind() < _InputArray::OPENGL_BUFFER)
+        {
+            impl(input0, input1, _flow1);
+            return;
+        }
+
+        impl(input0, input1, flow_);
+
+        if (!_flow2.needed())
+        {
+            arrCopy(flow_, _flow1);
+        }
+        else
+        {
+            split(flow_, flows_);
+
+            arrCopy(flows_[0], _flow1);
+            arrCopy(flows_[1], _flow2);
+        }
+    }
+
+    void CpuOpticalFlow::collectGarbage()
+    {
+        for (int i = 0; i < 6; ++i)
+            buf_[i].release();
+        flow_.release();
+        flows_[0].release();
+        flows_[1].release();
+    }
+}
+
+///////////////////////////////////////////////////////////////////
+// Farneback
+
+namespace
+{
+    class Farneback : public CpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        Farneback();
+
+    protected:
+        void impl(const Mat& input0, const Mat& input1, OutputArray dst);
+
+    private:
+        double pyrScale_;
+        int numLevels_;
+        int winSize_;
+        int numIters_;
+        int polyN_;
+        double polySigma_;
+        int flags_;
+    };
+
+    CV_INIT_ALGORITHM(Farneback, "DenseOpticalFlowExt.Farneback",
+                      obj.info()->addParam(obj, "pyrScale", obj.pyrScale_);
+                      obj.info()->addParam(obj, "numLevels", obj.numLevels_);
+                      obj.info()->addParam(obj, "winSize", obj.winSize_);
+                      obj.info()->addParam(obj, "numIters", obj.numIters_);
+                      obj.info()->addParam(obj, "polyN", obj.polyN_);
+                      obj.info()->addParam(obj, "polySigma", obj.polySigma_);
+                      obj.info()->addParam(obj, "flags", obj.flags_));
+
+    Farneback::Farneback() : CpuOpticalFlow(CV_8UC1)
+    {
+        pyrScale_ = 0.5;
+        numLevels_ = 5;
+        winSize_ = 13;
+        numIters_ = 10;
+        polyN_ = 5;
+        polySigma_ = 1.1;
+        flags_ = 0;
+    }
+
+    void Farneback::impl(const Mat& input0, const Mat& input1, OutputArray dst)
+    {
+        calcOpticalFlowFarneback(input0, input1, dst, pyrScale_, numLevels_, winSize_, numIters_, polyN_, polySigma_, flags_);
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Farneback()
+{
+    return makePtr<Farneback>();
+}
+
+///////////////////////////////////////////////////////////////////
+// Simple
+
+namespace
+{
+    class Simple : public CpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        Simple();
+
+    protected:
+        void impl(const Mat& input0, const Mat& input1, OutputArray dst);
+
+    private:
+        int layers_;
+        int averagingBlockSize_;
+        int maxFlow_;
+        double sigmaDist_;
+        double sigmaColor_;
+        int postProcessWindow_;
+        double sigmaDistFix_;
+        double sigmaColorFix_;
+        double occThr_;
+        int upscaleAveragingRadius_;
+        double upscaleSigmaDist_;
+        double upscaleSigmaColor_;
+        double speedUpThr_;
+    };
+
+    CV_INIT_ALGORITHM(Simple, "DenseOpticalFlowExt.Simple",
+                      obj.info()->addParam(obj, "layers", obj.layers_);
+                      obj.info()->addParam(obj, "averagingBlockSize", obj.averagingBlockSize_);
+                      obj.info()->addParam(obj, "maxFlow", obj.maxFlow_);
+                      obj.info()->addParam(obj, "sigmaDist", obj.sigmaDist_);
+                      obj.info()->addParam(obj, "sigmaColor", obj.sigmaColor_);
+                      obj.info()->addParam(obj, "postProcessWindow", obj.postProcessWindow_);
+                      obj.info()->addParam(obj, "sigmaDistFix", obj.sigmaDistFix_);
+                      obj.info()->addParam(obj, "sigmaColorFix", obj.sigmaColorFix_);
+                      obj.info()->addParam(obj, "occThr", obj.occThr_);
+                      obj.info()->addParam(obj, "upscaleAveragingRadius", obj.upscaleAveragingRadius_);
+                      obj.info()->addParam(obj, "upscaleSigmaDist", obj.upscaleSigmaDist_);
+                      obj.info()->addParam(obj, "upscaleSigmaColor", obj.upscaleSigmaColor_);
+                      obj.info()->addParam(obj, "speedUpThr", obj.speedUpThr_));
+
+    Simple::Simple() : CpuOpticalFlow(CV_8UC3)
+    {
+        layers_ = 3;
+        averagingBlockSize_ = 2;
+        maxFlow_ = 4;
+        sigmaDist_ = 4.1;
+        sigmaColor_ = 25.5;
+        postProcessWindow_ = 18;
+        sigmaDistFix_ = 55.0;
+        sigmaColorFix_ = 25.5;
+        occThr_ = 0.35;
+        upscaleAveragingRadius_ = 18;
+        upscaleSigmaDist_ = 55.0;
+        upscaleSigmaColor_ = 25.5;
+        speedUpThr_ = 10;
+    }
+
+    void Simple::impl(const Mat& _input0, const Mat& _input1, OutputArray dst)
+    {
+        Mat input0 = _input0;
+        Mat input1 = _input1;
+        calcOpticalFlowSF(input0, input1, dst.getMatRef(),
+                          layers_,
+                          averagingBlockSize_,
+                          maxFlow_,
+                          sigmaDist_,
+                          sigmaColor_,
+                          postProcessWindow_,
+                          sigmaDistFix_,
+                          sigmaColorFix_,
+                          occThr_,
+                          upscaleAveragingRadius_,
+                          upscaleSigmaDist_,
+                          upscaleSigmaColor_,
+                          speedUpThr_);
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Simple()
+{
+    return makePtr<Simple>();
+}
+
+///////////////////////////////////////////////////////////////////
+// DualTVL1
+
+namespace
+{
+    class DualTVL1 : public CpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        DualTVL1();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const Mat& input0, const Mat& input1, OutputArray dst);
+
+    private:
+        double tau_;
+        double lambda_;
+        double theta_;
+        int nscales_;
+        int warps_;
+        double epsilon_;
+        int iterations_;
+        bool useInitialFlow_;
+
+        Ptr<DenseOpticalFlow> alg_;
+    };
+
+    CV_INIT_ALGORITHM(DualTVL1, "DenseOpticalFlowExt.DualTVL1",
+                      obj.info()->addParam(obj, "tau", obj.tau_);
+                      obj.info()->addParam(obj, "lambda", obj.lambda_);
+                      obj.info()->addParam(obj, "theta", obj.theta_);
+                      obj.info()->addParam(obj, "nscales", obj.nscales_);
+                      obj.info()->addParam(obj, "warps", obj.warps_);
+                      obj.info()->addParam(obj, "epsilon", obj.epsilon_);
+                      obj.info()->addParam(obj, "iterations", obj.iterations_);
+                      obj.info()->addParam(obj, "useInitialFlow", obj.useInitialFlow_));
+
+    DualTVL1::DualTVL1() : CpuOpticalFlow(CV_8UC1)
+    {
+        alg_ = cv::createOptFlow_DualTVL1();
+        tau_ = alg_->getDouble("tau");
+        lambda_ = alg_->getDouble("lambda");
+        theta_ = alg_->getDouble("theta");
+        nscales_ = alg_->getInt("nscales");
+        warps_ = alg_->getInt("warps");
+        epsilon_ = alg_->getDouble("epsilon");
+        iterations_ = alg_->getInt("iterations");
+        useInitialFlow_ = alg_->getBool("useInitialFlow");
+    }
+
+    void DualTVL1::impl(const Mat& input0, const Mat& input1, OutputArray dst)
+    {
+        alg_->set("tau", tau_);
+        alg_->set("lambda", lambda_);
+        alg_->set("theta", theta_);
+        alg_->set("nscales", nscales_);
+        alg_->set("warps", warps_);
+        alg_->set("epsilon", epsilon_);
+        alg_->set("iterations", iterations_);
+        alg_->set("useInitialFlow", useInitialFlow_);
+
+        alg_->calc(input0, input1, dst);
+    }
+
+    void DualTVL1::collectGarbage()
+    {
+        alg_->collectGarbage();
+        CpuOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_DualTVL1()
+{
+    return makePtr<DualTVL1>();
+}
+
+///////////////////////////////////////////////////////////////////
+// GpuOpticalFlow
+
+#ifndef HAVE_OPENCV_GPUOPTFLOW
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Farneback_GPU()
+{
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<DenseOpticalFlowExt>();
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_DualTVL1_GPU()
+{
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<DenseOpticalFlowExt>();
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Brox_GPU()
+{
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<DenseOpticalFlowExt>();
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_PyrLK_GPU()
+{
+    CV_Error(cv::Error::StsNotImplemented, "The called functionality is disabled for current build or platform");
+    return Ptr<DenseOpticalFlowExt>();
+}
+
+#else // HAVE_OPENCV_GPUOPTFLOW
+
+namespace
+{
+    class GpuOpticalFlow : public DenseOpticalFlowExt
+    {
+    public:
+        explicit GpuOpticalFlow(int work_type);
+
+        void calc(InputArray frame0, InputArray frame1, OutputArray flow1, OutputArray flow2);
+        void collectGarbage();
+
+    protected:
+        virtual void impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2) = 0;
+
+    private:
+        int work_type_;
+        GpuMat buf_[6];
+        GpuMat u_, v_, flow_;
+    };
+
+    GpuOpticalFlow::GpuOpticalFlow(int work_type) : work_type_(work_type)
+    {
+    }
+
+    void GpuOpticalFlow::calc(InputArray _frame0, InputArray _frame1, OutputArray _flow1, OutputArray _flow2)
+    {
+        GpuMat frame0 = arrGetGpuMat(_frame0, buf_[0]);
+        GpuMat frame1 = arrGetGpuMat(_frame1, buf_[1]);
+
+        CV_Assert( frame1.type() == frame0.type() );
+        CV_Assert( frame1.size() == frame0.size() );
+
+        GpuMat input0 = convertToType(frame0, work_type_, buf_[2], buf_[3]);
+        GpuMat input1 = convertToType(frame1, work_type_, buf_[4], buf_[5]);
+
+        if (_flow2.needed() && _flow1.kind() == _InputArray::GPU_MAT && _flow2.kind() == _InputArray::GPU_MAT)
+        {
+            impl(input0, input1, _flow1.getGpuMatRef(), _flow2.getGpuMatRef());
+            return;
+        }
+
+        impl(input0, input1, u_, v_);
+
+        if (_flow2.needed())
+        {
+            arrCopy(u_, _flow1);
+            arrCopy(v_, _flow2);
+        }
+        else
+        {
+            GpuMat src[] = {u_, v_};
+            merge(src, 2, flow_);
+            arrCopy(flow_, _flow1);
+        }
+    }
+
+    void GpuOpticalFlow::collectGarbage()
+    {
+        for (int i = 0; i < 6; ++i)
+            buf_[i].release();
+        u_.release();
+        v_.release();
+        flow_.release();
+    }
+}
+
+///////////////////////////////////////////////////////////////////
+// Brox_GPU
+
+namespace
+{
+    class Brox_GPU : public GpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        Brox_GPU();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2);
+
+    private:
+        double alpha_;
+        double gamma_;
+        double scaleFactor_;
+        int innerIterations_;
+        int outerIterations_;
+        int solverIterations_;
+
+        BroxOpticalFlow alg_;
+    };
+
+    CV_INIT_ALGORITHM(Brox_GPU, "DenseOpticalFlowExt.Brox_GPU",
+                      obj.info()->addParam(obj, "alpha", obj.alpha_, false, 0, 0, "Flow smoothness");
+                      obj.info()->addParam(obj, "gamma", obj.gamma_, false, 0, 0, "Gradient constancy importance");
+                      obj.info()->addParam(obj, "scaleFactor", obj.scaleFactor_, false, 0, 0, "Pyramid scale factor");
+                      obj.info()->addParam(obj, "innerIterations", obj.innerIterations_, false, 0, 0, "Number of lagged non-linearity iterations (inner loop)");
+                      obj.info()->addParam(obj, "outerIterations", obj.outerIterations_, false, 0, 0, "Number of warping iterations (number of pyramid levels)");
+                      obj.info()->addParam(obj, "solverIterations", obj.solverIterations_, false, 0, 0, "Number of linear system solver iterations"));
+
+    Brox_GPU::Brox_GPU() : GpuOpticalFlow(CV_32FC1), alg_(0.197f, 50.0f, 0.8f, 10, 77, 10)
+    {
+        alpha_ = alg_.alpha;
+        gamma_ = alg_.gamma;
+        scaleFactor_ = alg_.scale_factor;
+        innerIterations_ = alg_.inner_iterations;
+        outerIterations_ = alg_.outer_iterations;
+        solverIterations_ = alg_.solver_iterations;
+    }
+
+    void Brox_GPU::impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2)
+    {
+        alg_.alpha = static_cast<float>(alpha_);
+        alg_.gamma = static_cast<float>(gamma_);
+        alg_.scale_factor = static_cast<float>(scaleFactor_);
+        alg_.inner_iterations = innerIterations_;
+        alg_.outer_iterations = outerIterations_;
+        alg_.solver_iterations = solverIterations_;
+
+        alg_(input0, input1, dst1, dst2);
+    }
+
+    void Brox_GPU::collectGarbage()
+    {
+        alg_.buf.release();
+        GpuOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Brox_GPU()
+{
+    return makePtr<Brox_GPU>();
+}
+
+///////////////////////////////////////////////////////////////////
+// PyrLK_GPU
+
+namespace
+{
+    class PyrLK_GPU : public GpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        PyrLK_GPU();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2);
+
+    private:
+        int winSize_;
+        int maxLevel_;
+        int iterations_;
+
+        PyrLKOpticalFlow alg_;
+    };
+
+    CV_INIT_ALGORITHM(PyrLK_GPU, "DenseOpticalFlowExt.PyrLK_GPU",
+                      obj.info()->addParam(obj, "winSize", obj.winSize_);
+                      obj.info()->addParam(obj, "maxLevel", obj.maxLevel_);
+                      obj.info()->addParam(obj, "iterations", obj.iterations_));
+
+    PyrLK_GPU::PyrLK_GPU() : GpuOpticalFlow(CV_8UC1)
+    {
+        winSize_ = alg_.winSize.width;
+        maxLevel_ = alg_.maxLevel;
+        iterations_ = alg_.iters;
+    }
+
+    void PyrLK_GPU::impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2)
+    {
+        alg_.winSize.width = winSize_;
+        alg_.winSize.height = winSize_;
+        alg_.maxLevel = maxLevel_;
+        alg_.iters = iterations_;
+
+        alg_.dense(input0, input1, dst1, dst2);
+    }
+
+    void PyrLK_GPU::collectGarbage()
+    {
+        alg_.releaseMemory();
+        GpuOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_PyrLK_GPU()
+{
+    return makePtr<PyrLK_GPU>();
+}
+
+///////////////////////////////////////////////////////////////////
+// Farneback_GPU
+
+namespace
+{
+    class Farneback_GPU : public GpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        Farneback_GPU();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2);
+
+    private:
+        double pyrScale_;
+        int numLevels_;
+        int winSize_;
+        int numIters_;
+        int polyN_;
+        double polySigma_;
+        int flags_;
+
+        FarnebackOpticalFlow alg_;
+    };
+
+    CV_INIT_ALGORITHM(Farneback_GPU, "DenseOpticalFlowExt.Farneback_GPU",
+                      obj.info()->addParam(obj, "pyrScale", obj.pyrScale_);
+                      obj.info()->addParam(obj, "numLevels", obj.numLevels_);
+                      obj.info()->addParam(obj, "winSize", obj.winSize_);
+                      obj.info()->addParam(obj, "numIters", obj.numIters_);
+                      obj.info()->addParam(obj, "polyN", obj.polyN_);
+                      obj.info()->addParam(obj, "polySigma", obj.polySigma_);
+                      obj.info()->addParam(obj, "flags", obj.flags_));
+
+    Farneback_GPU::Farneback_GPU() : GpuOpticalFlow(CV_8UC1)
+    {
+        pyrScale_ = alg_.pyrScale;
+        numLevels_ = alg_.numLevels;
+        winSize_ = alg_.winSize;
+        numIters_ = alg_.numIters;
+        polyN_ = alg_.polyN;
+        polySigma_ = alg_.polySigma;
+        flags_ = alg_.flags;
+    }
+
+    void Farneback_GPU::impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2)
+    {
+        alg_.pyrScale = pyrScale_;
+        alg_.numLevels = numLevels_;
+        alg_.winSize = winSize_;
+        alg_.numIters = numIters_;
+        alg_.polyN = polyN_;
+        alg_.polySigma = polySigma_;
+        alg_.flags = flags_;
+
+        alg_(input0, input1, dst1, dst2);
+    }
+
+    void Farneback_GPU::collectGarbage()
+    {
+        alg_.releaseMemory();
+        GpuOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Farneback_GPU()
+{
+    return makePtr<Farneback_GPU>();
+}
+
+///////////////////////////////////////////////////////////////////
+// DualTVL1_GPU
+
+namespace
+{
+    class DualTVL1_GPU : public GpuOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        DualTVL1_GPU();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2);
+
+    private:
+        double tau_;
+        double lambda_;
+        double theta_;
+        int nscales_;
+        int warps_;
+        double epsilon_;
+        int iterations_;
+        bool useInitialFlow_;
+
+        OpticalFlowDual_TVL1_GPU alg_;
+    };
+
+    CV_INIT_ALGORITHM(DualTVL1_GPU, "DenseOpticalFlowExt.DualTVL1_GPU",
+                      obj.info()->addParam(obj, "tau", obj.tau_);
+                      obj.info()->addParam(obj, "lambda", obj.lambda_);
+                      obj.info()->addParam(obj, "theta", obj.theta_);
+                      obj.info()->addParam(obj, "nscales", obj.nscales_);
+                      obj.info()->addParam(obj, "warps", obj.warps_);
+                      obj.info()->addParam(obj, "epsilon", obj.epsilon_);
+                      obj.info()->addParam(obj, "iterations", obj.iterations_);
+                      obj.info()->addParam(obj, "useInitialFlow", obj.useInitialFlow_));
+
+    DualTVL1_GPU::DualTVL1_GPU() : GpuOpticalFlow(CV_8UC1)
+    {
+        tau_ = alg_.tau;
+        lambda_ = alg_.lambda;
+        theta_ = alg_.theta;
+        nscales_ = alg_.nscales;
+        warps_ = alg_.warps;
+        epsilon_ = alg_.epsilon;
+        iterations_ = alg_.iterations;
+        useInitialFlow_ = alg_.useInitialFlow;
+    }
+
+    void DualTVL1_GPU::impl(const GpuMat& input0, const GpuMat& input1, GpuMat& dst1, GpuMat& dst2)
+    {
+        alg_.tau = tau_;
+        alg_.lambda = lambda_;
+        alg_.theta = theta_;
+        alg_.nscales = nscales_;
+        alg_.warps = warps_;
+        alg_.epsilon = epsilon_;
+        alg_.iterations = iterations_;
+        alg_.useInitialFlow = useInitialFlow_;
+
+        alg_(input0, input1, dst1, dst2);
+    }
+
+    void DualTVL1_GPU::collectGarbage()
+    {
+        alg_.collectGarbage();
+        GpuOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_DualTVL1_GPU()
+{
+    return makePtr<DualTVL1_GPU>();
+}
+
+#endif // HAVE_OPENCV_GPUOPTFLOW
+#ifdef HAVE_OPENCV_OCL
+
+namespace
+{
+    class oclOpticalFlow : public DenseOpticalFlowExt
+    {
+    public:
+        explicit oclOpticalFlow(int work_type);
+
+        void calc(InputArray frame0, InputArray frame1, OutputArray flow1, OutputArray flow2);
+        void collectGarbage();
+
+    protected:
+        virtual void impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2) = 0;
+
+    private:
+        int work_type_;
+        cv::ocl::oclMat buf_[6];
+        cv::ocl::oclMat u_, v_, flow_;
+    };
+
+    oclOpticalFlow::oclOpticalFlow(int work_type) : work_type_(work_type)
+    {
+    }
+
+    void oclOpticalFlow::calc(InputArray frame0, InputArray frame1, OutputArray flow1, OutputArray flow2)
+    {
+        ocl::oclMat& _frame0 = ocl::getOclMatRef(frame0);
+        ocl::oclMat& _frame1 = ocl::getOclMatRef(frame1);
+        ocl::oclMat& _flow1  = ocl::getOclMatRef(flow1);
+        ocl::oclMat& _flow2  = ocl::getOclMatRef(flow2);
+
+        CV_Assert( _frame1.type() == _frame0.type() );
+        CV_Assert( _frame1.size() == _frame0.size() );
+
+        cv::ocl::oclMat input0_ = convertToType(_frame0, work_type_, buf_[2], buf_[3]);
+        cv::ocl::oclMat input1_ = convertToType(_frame1, work_type_, buf_[4], buf_[5]);
+
+        impl(input0_, input1_, u_, v_);//go to tvl1 algorithm
+
+        u_.copyTo(_flow1);
+        v_.copyTo(_flow2);
+    }
+
+    void oclOpticalFlow::collectGarbage()
+    {
+        for (int i = 0; i < 6; ++i)
+            buf_[i].release();
+        u_.release();
+        v_.release();
+        flow_.release();
+    }
+}
+///////////////////////////////////////////////////////////////////
+// PyrLK_OCL
+
+namespace
+{
+    class PyrLK_OCL : public oclOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        PyrLK_OCL();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const ocl::oclMat& input0, const ocl::oclMat& input1, ocl::oclMat& dst1, ocl::oclMat& dst2);
+
+    private:
+        int winSize_;
+        int maxLevel_;
+        int iterations_;
+
+        ocl::PyrLKOpticalFlow alg_;
+    };
+
+    CV_INIT_ALGORITHM(PyrLK_OCL, "DenseOpticalFlowExt.PyrLK_OCL",
+        obj.info()->addParam(obj, "winSize", obj.winSize_);
+    obj.info()->addParam(obj, "maxLevel", obj.maxLevel_);
+    obj.info()->addParam(obj, "iterations", obj.iterations_));
+
+    PyrLK_OCL::PyrLK_OCL() : oclOpticalFlow(CV_8UC1)
+    {
+        winSize_ = alg_.winSize.width;
+        maxLevel_ = alg_.maxLevel;
+        iterations_ = alg_.iters;
+    }
+
+    void PyrLK_OCL::impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2)
+    {
+        alg_.winSize.width = winSize_;
+        alg_.winSize.height = winSize_;
+        alg_.maxLevel = maxLevel_;
+        alg_.iters = iterations_;
+
+        alg_.dense(input0, input1, dst1, dst2);
+    }
+
+    void PyrLK_OCL::collectGarbage()
+    {
+        alg_.releaseMemory();
+        oclOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_PyrLK_OCL()
+{
+    return makePtr<PyrLK_OCL>();
+}
+
+///////////////////////////////////////////////////////////////////
+// DualTVL1_OCL
+
+namespace
+{
+    class DualTVL1_OCL : public oclOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        DualTVL1_OCL();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2);
+
+    private:
+        double tau_;
+        double lambda_;
+        double theta_;
+        int nscales_;
+        int warps_;
+        double epsilon_;
+        int iterations_;
+        bool useInitialFlow_;
+
+        ocl::OpticalFlowDual_TVL1_OCL alg_;
+    };
+
+    CV_INIT_ALGORITHM(DualTVL1_OCL, "DenseOpticalFlowExt.DualTVL1_OCL",
+    obj.info()->addParam(obj, "tau", obj.tau_);
+    obj.info()->addParam(obj, "lambda", obj.lambda_);
+    obj.info()->addParam(obj, "theta", obj.theta_);
+    obj.info()->addParam(obj, "nscales", obj.nscales_);
+    obj.info()->addParam(obj, "warps", obj.warps_);
+    obj.info()->addParam(obj, "epsilon", obj.epsilon_);
+    obj.info()->addParam(obj, "iterations", obj.iterations_);
+    obj.info()->addParam(obj, "useInitialFlow", obj.useInitialFlow_));
+
+    DualTVL1_OCL::DualTVL1_OCL() : oclOpticalFlow(CV_8UC1)
+    {
+        tau_ = alg_.tau;
+        lambda_ = alg_.lambda;
+        theta_ = alg_.theta;
+        nscales_ = alg_.nscales;
+        warps_ = alg_.warps;
+        epsilon_ = alg_.epsilon;
+        iterations_ = alg_.iterations;
+        useInitialFlow_ = alg_.useInitialFlow;
+    }
+
+    void DualTVL1_OCL::impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2)
+    {
+        alg_.tau = tau_;
+        alg_.lambda = lambda_;
+        alg_.theta = theta_;
+        alg_.nscales = nscales_;
+        alg_.warps = warps_;
+        alg_.epsilon = epsilon_;
+        alg_.iterations = iterations_;
+        alg_.useInitialFlow = useInitialFlow_;
+
+        alg_(input0, input1, dst1, dst2);
+
+    }
+
+    void DualTVL1_OCL::collectGarbage()
+    {
+        alg_.collectGarbage();
+        oclOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_DualTVL1_OCL()
+{
+    return makePtr<DualTVL1_OCL>();
+}
+
+///////////////////////////////////////////////////////////////////
+// FarneBack
+
+namespace
+{
+    class FarneBack_OCL : public oclOpticalFlow
+    {
+    public:
+        AlgorithmInfo* info() const;
+
+        FarneBack_OCL();
+
+        void collectGarbage();
+
+    protected:
+        void impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2);
+
+    private:
+        double pyrScale_;
+        int numLevels_;
+        int winSize_;
+        int numIters_;
+        int polyN_;
+        double polySigma_;
+        int flags_;
+
+        ocl::FarnebackOpticalFlow alg_;
+    };
+
+    CV_INIT_ALGORITHM(FarneBack_OCL, "DenseOpticalFlowExt.FarneBack_OCL",
+        obj.info()->addParam(obj, "pyrScale", obj.pyrScale_);
+    obj.info()->addParam(obj, "numLevels", obj.numLevels_);
+    obj.info()->addParam(obj, "winSize", obj.winSize_);
+    obj.info()->addParam(obj, "numIters", obj.numIters_);
+    obj.info()->addParam(obj, "polyN", obj.polyN_);
+    obj.info()->addParam(obj, "polySigma", obj.polySigma_);
+    obj.info()->addParam(obj, "flags", obj.flags_));
+
+    FarneBack_OCL::FarneBack_OCL() : oclOpticalFlow(CV_8UC1)
+    {
+        pyrScale_ = alg_.pyrScale;
+        numLevels_ = alg_.numLevels;
+        winSize_ = alg_.winSize;
+        numIters_ = alg_.numIters;
+        polyN_ = alg_.polyN;
+        polySigma_ = alg_.polySigma;
+        flags_ = alg_.flags;
+    }
+
+    void FarneBack_OCL::impl(const cv::ocl::oclMat& input0, const cv::ocl::oclMat& input1, cv::ocl::oclMat& dst1, cv::ocl::oclMat& dst2)
+    {
+        alg_.pyrScale = pyrScale_;
+        alg_.numLevels = numLevels_;
+        alg_.winSize = winSize_;
+        alg_.numIters = numIters_;
+        alg_.polyN = polyN_;
+        alg_.polySigma = polySigma_;
+        alg_.flags = flags_;
+
+        alg_(input0, input1, dst1, dst2);
+    }
+
+    void FarneBack_OCL::collectGarbage()
+    {
+        alg_.releaseMemory();
+        oclOpticalFlow::collectGarbage();
+    }
+}
+
+Ptr<DenseOpticalFlowExt> cv::superres::createOptFlow_Farneback_OCL()
+{
+    return makePtr<FarneBack_OCL>();
+}
+
+#endif
diff --git a/modules/superres/src/precomp.hpp b/modules/superres/src/precomp.hpp
new file mode 100644
index 0000000000..822d488799
--- /dev/null
+++ b/modules/superres/src/precomp.hpp
@@ -0,0 +1,100 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_PRECOMP_H__
+#define __OPENCV_PRECOMP_H__
+
+#include <vector>
+#include <limits>
+
+#include "opencv2/opencv_modules.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/core/opengl.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/video/tracking.hpp"
+#include "opencv2/core/private.hpp"
+
+#include "opencv2/core/private.gpu.hpp"
+
+#ifdef HAVE_OPENCV_GPUARITHM
+#  include "opencv2/gpuarithm.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUWARPING
+#  include "opencv2/gpuwarping.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUFILTERS
+#  include "opencv2/gpufilters.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUIMGPROC
+#  include "opencv2/gpuimgproc.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+#  include "opencv2/gpuoptflow.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPUCODEC
+#  include "opencv2/gpucodec.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_OCL
+    #include "opencv2/ocl/private/util.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_HIGHGUI
+    #include "opencv2/highgui.hpp"
+#endif
+
+#include "opencv2/superres.hpp"
+#include "opencv2/superres/optical_flow.hpp"
+#include "input_array_utility.hpp"
+
+#include "ring_buffer.hpp"
+
+#include "opencv2/core/private.hpp"
+
+#endif /* __OPENCV_PRECOMP_H__ */
diff --git a/modules/superres/src/ring_buffer.hpp b/modules/superres/src/ring_buffer.hpp
new file mode 100644
index 0000000000..a0c0c04b5e
--- /dev/null
+++ b/modules/superres/src/ring_buffer.hpp
@@ -0,0 +1,79 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __RING_BUFFER_HPP__
+#define __RING_BUFFER_HPP__
+
+#include "precomp.hpp"
+
+namespace cv
+{
+    namespace superres
+    {
+        namespace detail
+        {
+            template <typename T, class A>
+            inline const T& at(int index, const std::vector<T, A>& items)
+            {
+                const int len = static_cast<int>(items.size());
+                if (index < 0)
+                    index -= ((index - len + 1) / len) * len;
+                if (index >= len)
+                    index %= len;
+                return items[index];
+            }
+
+            template <typename T, class A>
+            inline T& at(int index, std::vector<T, A>& items)
+            {
+                const int len = static_cast<int>(items.size());
+                if (index < 0)
+                    index -= ((index - len + 1) / len) * len;
+                if (index >= len)
+                    index %= len;
+                return items[index];
+            }
+        }
+    }
+}
+
+#endif // __RING_BUFFER_HPP__
diff --git a/modules/superres/src/super_resolution.cpp b/modules/superres/src/super_resolution.cpp
new file mode 100644
index 0000000000..031a59b295
--- /dev/null
+++ b/modules/superres/src/super_resolution.cpp
@@ -0,0 +1,84 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::superres;
+
+bool cv::superres::initModule_superres()
+{
+    return !createSuperResolution_BTVL1().empty();
+}
+
+cv::superres::SuperResolution::SuperResolution()
+{
+    frameSource_ = createFrameSource_Empty();
+    firstCall_ = true;
+}
+
+void cv::superres::SuperResolution::setInput(const Ptr<FrameSource>& frameSource)
+{
+    frameSource_ = frameSource;
+    firstCall_ = true;
+}
+
+void cv::superres::SuperResolution::nextFrame(OutputArray frame)
+{
+    if (firstCall_)
+    {
+        initImpl(frameSource_);
+        firstCall_ = false;
+    }
+
+    processImpl(frameSource_, frame);
+}
+
+void cv::superres::SuperResolution::reset()
+{
+    frameSource_->reset();
+    firstCall_ = true;
+}
+
+void cv::superres::SuperResolution::collectGarbage()
+{
+}
diff --git a/modules/superres/test/test_main.cpp b/modules/superres/test/test_main.cpp
new file mode 100644
index 0000000000..b1998b9320
--- /dev/null
+++ b/modules/superres/test/test_main.cpp
@@ -0,0 +1,45 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+CV_TEST_MAIN("superres")
diff --git a/modules/superres/test/test_precomp.hpp b/modules/superres/test/test_precomp.hpp
new file mode 100644
index 0000000000..47334e324a
--- /dev/null
+++ b/modules/superres/test/test_precomp.hpp
@@ -0,0 +1,62 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __GNUC__
+#  pragma GCC diagnostic ignored "-Wmissing-declarations"
+#  if defined __clang__ || defined __APPLE__
+#    pragma GCC diagnostic ignored "-Wmissing-prototypes"
+#    pragma GCC diagnostic ignored "-Wextra"
+#  endif
+#endif
+
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/opencv_modules.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/ts.hpp"
+#include "opencv2/superres.hpp"
+#include "cvconfig.h"
+#include "input_array_utility.hpp"
+
+#endif
diff --git a/modules/superres/test/test_superres.cpp b/modules/superres/test/test_superres.cpp
new file mode 100644
index 0000000000..d33e182738
--- /dev/null
+++ b/modules/superres/test/test_superres.cpp
@@ -0,0 +1,290 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+class AllignedFrameSource : public cv::superres::FrameSource
+{
+public:
+    AllignedFrameSource(const cv::Ptr<cv::superres::FrameSource>& base, int scale);
+
+    void nextFrame(cv::OutputArray frame);
+    void reset();
+
+private:
+    cv::Ptr<cv::superres::FrameSource> base_;
+    cv::Mat origFrame_;
+    int scale_;
+};
+
+AllignedFrameSource::AllignedFrameSource(const cv::Ptr<cv::superres::FrameSource>& base, int scale) :
+    base_(base), scale_(scale)
+{
+    CV_Assert( base_ );
+}
+
+void AllignedFrameSource::nextFrame(cv::OutputArray frame)
+{
+    base_->nextFrame(origFrame_);
+
+    if (origFrame_.rows % scale_ == 0 && origFrame_.cols % scale_ == 0)
+    {
+        cv::superres::arrCopy(origFrame_, frame);
+    }
+    else
+    {
+        cv::Rect ROI(0, 0, (origFrame_.cols / scale_) * scale_, (origFrame_.rows / scale_) * scale_);
+        cv::superres::arrCopy(origFrame_(ROI), frame);
+    }
+}
+
+void AllignedFrameSource::reset()
+{
+    base_->reset();
+}
+
+class DegradeFrameSource : public cv::superres::FrameSource
+{
+public:
+    DegradeFrameSource(const cv::Ptr<cv::superres::FrameSource>& base, int scale);
+
+    void nextFrame(cv::OutputArray frame);
+    void reset();
+
+private:
+    cv::Ptr<cv::superres::FrameSource> base_;
+    cv::Mat origFrame_;
+    cv::Mat blurred_;
+    cv::Mat deg_;
+    double iscale_;
+};
+
+DegradeFrameSource::DegradeFrameSource(const cv::Ptr<cv::superres::FrameSource>& base, int scale) :
+    base_(base), iscale_(1.0 / scale)
+{
+    CV_Assert( base_ );
+}
+
+void addGaussNoise(cv::Mat& image, double sigma)
+{
+    cv::Mat noise(image.size(), CV_32FC(image.channels()));
+    cvtest::TS::ptr()->get_rng().fill(noise, cv::RNG::NORMAL, 0.0, sigma);
+
+    cv::addWeighted(image, 1.0, noise, 1.0, 0.0, image, image.depth());
+}
+
+void addSpikeNoise(cv::Mat& image, int frequency)
+{
+    cv::Mat_<uchar> mask(image.size(), 0);
+
+    for (int y = 0; y < mask.rows; ++y)
+    {
+        for (int x = 0; x < mask.cols; ++x)
+        {
+            if (cvtest::TS::ptr()->get_rng().uniform(0, frequency) < 1)
+                mask(y, x) = 255;
+        }
+    }
+
+    image.setTo(cv::Scalar::all(255), mask);
+}
+
+void DegradeFrameSource::nextFrame(cv::OutputArray frame)
+{
+    base_->nextFrame(origFrame_);
+
+    cv::GaussianBlur(origFrame_, blurred_, cv::Size(5, 5), 0);
+    cv::resize(blurred_, deg_, cv::Size(), iscale_, iscale_, cv::INTER_NEAREST);
+
+    addGaussNoise(deg_, 10.0);
+    addSpikeNoise(deg_, 500);
+
+    cv::superres::arrCopy(deg_, frame);
+}
+
+void DegradeFrameSource::reset()
+{
+    base_->reset();
+}
+
+double MSSIM(const cv::Mat& i1, const cv::Mat& i2)
+{
+    const double C1 = 6.5025;
+    const double C2 = 58.5225;
+
+    const int depth = CV_32F;
+
+    cv::Mat I1, I2;
+    i1.convertTo(I1, depth);
+    i2.convertTo(I2, depth);
+
+    cv::Mat I2_2  = I2.mul(I2); // I2^2
+    cv::Mat I1_2  = I1.mul(I1); // I1^2
+    cv::Mat I1_I2 = I1.mul(I2); // I1 * I2
+
+    cv::Mat mu1, mu2;
+    cv::GaussianBlur(I1, mu1, cv::Size(11, 11), 1.5);
+    cv::GaussianBlur(I2, mu2, cv::Size(11, 11), 1.5);
+
+    cv::Mat mu1_2   = mu1.mul(mu1);
+    cv::Mat mu2_2   = mu2.mul(mu2);
+    cv::Mat mu1_mu2 = mu1.mul(mu2);
+
+    cv::Mat sigma1_2, sigma2_2, sigma12;
+
+    cv::GaussianBlur(I1_2, sigma1_2, cv::Size(11, 11), 1.5);
+    sigma1_2 -= mu1_2;
+
+    cv::GaussianBlur(I2_2, sigma2_2, cv::Size(11, 11), 1.5);
+    sigma2_2 -= mu2_2;
+
+    cv::GaussianBlur(I1_I2, sigma12, cv::Size(11, 11), 1.5);
+    sigma12 -= mu1_mu2;
+
+    cv::Mat t1, t2;
+    cv::Mat numerator;
+    cv::Mat denominator;
+
+    // t3 = ((2*mu1_mu2 + C1).*(2*sigma12 + C2))
+    t1 = 2 * mu1_mu2 + C1;
+    t2 = 2 * sigma12 + C2;
+    numerator = t1.mul(t2);
+
+    // t1 =((mu1_2 + mu2_2 + C1).*(sigma1_2 + sigma2_2 + C2))
+    t1 = mu1_2 + mu2_2 + C1;
+    t2 = sigma1_2 + sigma2_2 + C2;
+    denominator = t1.mul(t2);
+
+    // ssim_map =  numerator./denominator;
+    cv::Mat ssim_map;
+    cv::divide(numerator, denominator, ssim_map);
+
+    // mssim = average of ssim map
+    cv::Scalar mssim = cv::mean(ssim_map);
+
+    if (i1.channels() == 1)
+        return mssim[0];
+
+    return (mssim[0] + mssim[1] + mssim[3]) / 3;
+}
+
+class SuperResolution : public testing::Test
+{
+public:
+    void RunTest(cv::Ptr<cv::superres::SuperResolution> superRes);
+};
+
+void SuperResolution::RunTest(cv::Ptr<cv::superres::SuperResolution> superRes)
+{
+    const std::string inputVideoName = cvtest::TS::ptr()->get_data_path() + "car.avi";
+    const int scale = 2;
+    const int iterations = 100;
+    const int temporalAreaRadius = 2;
+
+    ASSERT_FALSE( superRes.empty() );
+
+    const int btvKernelSize = superRes->getInt("btvKernelSize");
+
+    superRes->set("scale", scale);
+    superRes->set("iterations", iterations);
+    superRes->set("temporalAreaRadius", temporalAreaRadius);
+
+    cv::Ptr<cv::superres::FrameSource> goldSource(new AllignedFrameSource(cv::superres::createFrameSource_Video(inputVideoName), scale));
+    cv::Ptr<cv::superres::FrameSource> lowResSource(new DegradeFrameSource(
+        cv::makePtr<AllignedFrameSource>(cv::superres::createFrameSource_Video(inputVideoName), scale), scale));
+
+    // skip first frame
+    cv::Mat frame;
+
+    lowResSource->nextFrame(frame);
+    goldSource->nextFrame(frame);
+
+    cv::Rect inner(btvKernelSize, btvKernelSize, frame.cols - 2 * btvKernelSize, frame.rows - 2 * btvKernelSize);
+
+    superRes->setInput(lowResSource);
+
+    double srAvgMSSIM = 0.0;
+    const int count = 10;
+
+    cv::Mat goldFrame, superResFrame;
+    for (int i = 0; i < count; ++i)
+    {
+        goldSource->nextFrame(goldFrame);
+        ASSERT_FALSE( goldFrame.empty() );
+
+        superRes->nextFrame(superResFrame);
+        ASSERT_FALSE( superResFrame.empty() );
+
+        const double srMSSIM = MSSIM(goldFrame(inner), superResFrame);
+
+        srAvgMSSIM += srMSSIM;
+    }
+
+    srAvgMSSIM /= count;
+
+    EXPECT_GE( srAvgMSSIM, 0.5 );
+}
+
+TEST_F(SuperResolution, BTVL1)
+{
+    RunTest(cv::superres::createSuperResolution_BTVL1());
+}
+
+#if defined(HAVE_CUDA) && defined(HAVE_OPENCV_GPUARITHM) && defined(HAVE_OPENCV_GPUWARPING) && defined(HAVE_OPENCV_GPUFILTERS)
+
+TEST_F(SuperResolution, BTVL1_GPU)
+{
+    RunTest(cv::superres::createSuperResolution_BTVL1_GPU());
+}
+
+#endif
+
+#if defined(HAVE_OPENCV_OCL) && defined(HAVE_OPENCL)
+
+TEST_F(SuperResolution, BTVL1_OCL)
+{
+    std::vector<cv::ocl::Info> infos;
+    cv::ocl::getDevice(infos);
+    RunTest(cv::superres::createSuperResolution_BTVL1_OCL());
+}
+
+#endif
diff --git a/modules/ts/CMakeLists.txt b/modules/ts/CMakeLists.txt
index 455ef876bd..0f9c3fe778 100644
--- a/modules/ts/CMakeLists.txt
+++ b/modules/ts/CMakeLists.txt
@@ -4,22 +4,19 @@ if(IOS)
   ocv_module_disable(ts)
 endif()
 
-if(MINGW)
-  set(OPENCV_MODULE_TYPE STATIC)
-endif()
-
+set(OPENCV_MODULE_TYPE STATIC)
 set(OPENCV_MODULE_IS_PART_OF_WORLD FALSE)
 
-ocv_add_module(ts opencv_core opencv_features2d)
+if(HAVE_CUDA)
+  ocv_include_directories(${CUDA_INCLUDE_DIRS})
+endif()
+
+ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef)
+
+ocv_add_module(ts opencv_core opencv_imgproc opencv_highgui)
 
 ocv_glob_module_sources()
 ocv_module_include_directories()
 ocv_create_module()
 
-if(BUILD_SHARED_LIBS AND NOT MINGW)
-  add_definitions(-DGTEST_CREATE_SHARED_LIBRARY=1)
-else()
-  add_definitions(-DGTEST_CREATE_SHARED_LIBRARY=0)
-endif()
-
 ocv_add_precompiled_headers(${the_module})
diff --git a/modules/ts/include/opencv2/ts.hpp b/modules/ts/include/opencv2/ts.hpp
index 61e75be381..5d98810145 100644
--- a/modules/ts/include/opencv2/ts.hpp
+++ b/modules/ts/include/opencv2/ts.hpp
@@ -1,17 +1,13 @@
 #ifndef __OPENCV_GTESTCV_HPP__
 #define __OPENCV_GTESTCV_HPP__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-#ifndef GTEST_CREATE_SHARED_LIBRARY
-#ifdef BUILD_SHARED_LIBS
-#define GTEST_LINKED_AS_SHARED_LIBRARY 1
-#endif
-#endif
-
+#include "opencv2/core/cvdef.h"
 #include <stdarg.h> // for va_list
 
+#ifdef HAVE_WINRT
+    #pragma warning(disable:4447) // Disable warning 'main' signature found without threading model
+#endif
+
 #ifdef _MSC_VER
 #pragma warning( disable: 4127 )
 #endif
@@ -36,6 +32,7 @@
 #endif
 
 #include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
 
 namespace cvtest
 {
@@ -93,7 +90,7 @@ CV_EXPORTS void add(const Mat& a, double alpha, const Mat& b, double beta,
 CV_EXPORTS void multiply(const Mat& a, const Mat& b, Mat& c, double alpha=1);
 CV_EXPORTS void divide(const Mat& a, const Mat& b, Mat& c, double alpha=1);
 
-CV_EXPORTS void convert(const Mat& src, Mat& dst, int dtype, double alpha=1, double beta=0);
+CV_EXPORTS void convert(const Mat& src, cv::OutputArray dst, int dtype, double alpha=1, double beta=0);
 CV_EXPORTS void copy(const Mat& src, Mat& dst, const Mat& mask=Mat(), bool invertMask=false);
 CV_EXPORTS void set(Mat& dst, const Scalar& gamma, const Mat& mask=Mat());
 
@@ -110,9 +107,9 @@ CV_EXPORTS void  patchZeros( Mat& mat, double level );
 
 CV_EXPORTS void transpose(const Mat& src, Mat& dst);
 CV_EXPORTS void erode(const Mat& src, Mat& dst, const Mat& _kernel, Point anchor=Point(-1,-1),
-                      int borderType=IPL_BORDER_CONSTANT, const Scalar& borderValue=Scalar());
+                      int borderType=0, const Scalar& borderValue=Scalar());
 CV_EXPORTS void dilate(const Mat& src, Mat& dst, const Mat& _kernel, Point anchor=Point(-1,-1),
-                       int borderType=IPL_BORDER_CONSTANT, const Scalar& borderValue=Scalar());
+                       int borderType=0, const Scalar& borderValue=Scalar());
 CV_EXPORTS void filter2D(const Mat& src, Mat& dst, int ddepth, const Mat& kernel,
                          Point anchor, double delta, int borderType,
                          const Scalar& borderValue=Scalar());
@@ -478,26 +475,10 @@ protected:
     virtual int run_test_case( int expected_code, const string& descr );
     virtual void run_func(void) = 0;
     int test_case_idx;
-    int progress;
-    double t, freq;
 
     template<class F>
     int run_test_case( int expected_code, const string& _descr, F f)
     {
-        double new_t = (double)cv::getTickCount(), dt;
-        if( test_case_idx < 0 )
-        {
-            test_case_idx = 0;
-            progress = 0;
-            dt = 0;
-        }
-        else
-        {
-            dt = (new_t - t)/(freq*1000);
-            t = new_t;
-        }
-        progress = update_progress(progress, test_case_idx, 0, dt);
-
         int errcount = 0;
         bool thrown = false;
         const char* descr = _descr.c_str() ? _descr.c_str() : "";
@@ -554,28 +535,25 @@ namespace cvtest
 CV_EXPORTS void fillGradient(Mat& img, int delta = 5);
 CV_EXPORTS void smoothBorder(Mat& img, const Scalar& color, int delta = 3);
 
+CV_EXPORTS void printVersionInfo(bool useStdOut = true);
 } //namespace cvtest
 
-// fills c with zeros
-CV_EXPORTS void cvTsZero( CvMat* c, const CvMat* mask=0 );
-
-// copies a to b (whole matrix or only the selected region)
-CV_EXPORTS void cvTsCopy( const CvMat* a, CvMat* b, const CvMat* mask=0 );
-
-// converts one array to another
-CV_EXPORTS void  cvTsConvert( const CvMat* src, CvMat* dst );
-
-CV_EXPORTS void cvTsGEMM( const CvMat* a, const CvMat* b, double alpha,
-                         const CvMat* c, double beta, CvMat* d, int flags );
-
 #define CV_TEST_MAIN(resourcesubdir) \
 int main(int argc, char **argv) \
 { \
     cvtest::TS::ptr()->init(resourcesubdir); \
     ::testing::InitGoogleTest(&argc, argv); \
+    cvtest::printVersionInfo();\
     return RUN_ALL_TESTS(); \
 }
 
+// This usually only makes sense in perf tests with several implementations,
+// some of which are not available.
+#define CV_TEST_FAIL_NO_IMPL() do { \
+    ::testing::Test::RecordProperty("custom_status", "noimpl"); \
+    FAIL() << "No equivalent implementation."; \
+} while (0)
+
 #endif
 
 #include "opencv2/ts/ts_perf.hpp"
diff --git a/modules/ts/include/opencv2/ts/gpu_perf.hpp b/modules/ts/include/opencv2/ts/gpu_perf.hpp
new file mode 100644
index 0000000000..8f1ba8abc8
--- /dev/null
+++ b/modules/ts/include/opencv2/ts/gpu_perf.hpp
@@ -0,0 +1,127 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPU_PERF_UTILITY_HPP__
+#define __OPENCV_GPU_PERF_UTILITY_HPP__
+
+#include "opencv2/core.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/ts/ts_perf.hpp"
+#include "cvconfig.h"
+
+namespace perf
+{
+    #define ALL_BORDER_MODES BorderMode::all()
+    #define ALL_INTERPOLATIONS Interpolation::all()
+
+    CV_ENUM(BorderMode, BORDER_REFLECT101, BORDER_REPLICATE, BORDER_CONSTANT, BORDER_REFLECT, BORDER_WRAP)
+    CV_ENUM(Interpolation, INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_AREA)
+    CV_ENUM(NormType, NORM_INF, NORM_L1, NORM_L2, NORM_HAMMING, NORM_MINMAX)
+
+    enum { Gray = 1, TwoChannel = 2, BGR = 3, BGRA = 4 };
+    CV_ENUM(MatCn, Gray, TwoChannel, BGR, BGRA)
+
+    #define GPU_CHANNELS_1_3_4 testing::Values(MatCn(Gray), MatCn(BGR), MatCn(BGRA))
+    #define GPU_CHANNELS_1_3 testing::Values(MatCn(Gray), MatCn(BGR))
+
+    #define GET_PARAM(k) std::tr1::get< k >(GetParam())
+
+    #define DEF_PARAM_TEST(name, ...) typedef ::perf::TestBaseWithParam< std::tr1::tuple< __VA_ARGS__ > > name
+    #define DEF_PARAM_TEST_1(name, param_type) typedef ::perf::TestBaseWithParam< param_type > name
+
+    DEF_PARAM_TEST_1(Sz, cv::Size);
+    typedef perf::Size_MatType Sz_Type;
+    DEF_PARAM_TEST(Sz_Depth, cv::Size, perf::MatDepth);
+    DEF_PARAM_TEST(Sz_Depth_Cn, cv::Size, perf::MatDepth, MatCn);
+
+    #define GPU_TYPICAL_MAT_SIZES testing::Values(perf::sz720p, perf::szSXGA, perf::sz1080p)
+
+    #define FAIL_NO_CPU() FAIL() << "No such CPU implementation analogy"
+
+    #define GPU_SANITY_CHECK(mat, ...) \
+        do{ \
+            cv::Mat gpu_##mat(mat); \
+            SANITY_CHECK(gpu_##mat, ## __VA_ARGS__); \
+        } while(0)
+
+    #define CPU_SANITY_CHECK(mat, ...) \
+        do{ \
+            cv::Mat cpu_##mat(mat); \
+            SANITY_CHECK(cpu_##mat, ## __VA_ARGS__); \
+        } while(0)
+
+    CV_EXPORTS cv::Mat readImage(const std::string& fileName, int flags = cv::IMREAD_COLOR);
+
+    struct CvtColorInfo
+    {
+        int scn;
+        int dcn;
+        int code;
+
+        CvtColorInfo() {}
+        explicit CvtColorInfo(int scn_, int dcn_, int code_) : scn(scn_), dcn(dcn_), code(code_) {}
+    };
+    CV_EXPORTS void PrintTo(const CvtColorInfo& info, std::ostream* os);
+
+    CV_EXPORTS void printCudaInfo();
+
+    CV_EXPORTS void sortKeyPoints(std::vector<cv::KeyPoint>& keypoints, cv::InputOutputArray _descriptors = cv::noArray());
+
+#ifdef HAVE_CUDA
+    #define CV_PERF_TEST_CUDA_MAIN(modulename) \
+        int main(int argc, char **argv)\
+        {\
+            const char * impls[] = { "cuda", "plain" };\
+            CV_PERF_TEST_MAIN_INTERNALS(modulename, impls, perf::printCudaInfo())\
+        }
+#else
+    #define CV_PERF_TEST_CUDA_MAIN(modulename) \
+        int main(int argc, char **argv)\
+        {\
+            const char * plain_only[] = { "plain" };\
+            CV_PERF_TEST_MAIN_INTERNALS(modulename, plain_only)\
+        }
+#endif
+}
+
+#endif // __OPENCV_GPU_PERF_UTILITY_HPP__
diff --git a/modules/ts/include/opencv2/ts/gpu_test.hpp b/modules/ts/include/opencv2/ts/gpu_test.hpp
new file mode 100644
index 0000000000..2ff2ca0bfc
--- /dev/null
+++ b/modules/ts/include/opencv2/ts/gpu_test.hpp
@@ -0,0 +1,413 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_GPU_TEST_UTILITY_HPP__
+#define __OPENCV_GPU_TEST_UTILITY_HPP__
+
+#include <stdexcept>
+#include "cvconfig.h"
+#include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/ts.hpp"
+
+namespace cvtest
+{
+    //////////////////////////////////////////////////////////////////////
+    // random generators
+
+    CV_EXPORTS int randomInt(int minVal, int maxVal);
+    CV_EXPORTS double randomDouble(double minVal, double maxVal);
+    CV_EXPORTS cv::Size randomSize(int minVal, int maxVal);
+    CV_EXPORTS cv::Scalar randomScalar(double minVal, double maxVal);
+    CV_EXPORTS cv::Mat randomMat(cv::Size size, int type, double minVal = 0.0, double maxVal = 255.0);
+
+    //////////////////////////////////////////////////////////////////////
+    // GpuMat create
+
+    CV_EXPORTS cv::gpu::GpuMat createMat(cv::Size size, int type, bool useRoi = false);
+    CV_EXPORTS cv::gpu::GpuMat loadMat(const cv::Mat& m, bool useRoi = false);
+
+    //////////////////////////////////////////////////////////////////////
+    // Image load
+
+    //! read image from testdata folder
+    CV_EXPORTS cv::Mat readImage(const std::string& fileName, int flags = cv::IMREAD_COLOR);
+
+    //! read image from testdata folder and convert it to specified type
+    CV_EXPORTS cv::Mat readImageType(const std::string& fname, int type);
+
+    //////////////////////////////////////////////////////////////////////
+    // Gpu devices
+
+    //! return true if device supports specified feature and gpu module was built with support the feature.
+    CV_EXPORTS bool supportFeature(const cv::gpu::DeviceInfo& info, cv::gpu::FeatureSet feature);
+
+    class CV_EXPORTS DeviceManager
+    {
+    public:
+        static DeviceManager& instance();
+
+        void load(int i);
+        void loadAll();
+
+        const std::vector<cv::gpu::DeviceInfo>& values() const { return devices_; }
+
+    private:
+        std::vector<cv::gpu::DeviceInfo> devices_;
+    };
+
+    #define ALL_DEVICES testing::ValuesIn(cvtest::DeviceManager::instance().values())
+
+    //////////////////////////////////////////////////////////////////////
+    // Additional assertion
+
+    CV_EXPORTS void minMaxLocGold(const cv::Mat& src, double* minVal_, double* maxVal_ = 0, cv::Point* minLoc_ = 0, cv::Point* maxLoc_ = 0, const cv::Mat& mask = cv::Mat());
+
+    CV_EXPORTS cv::Mat getMat(cv::InputArray arr);
+
+    CV_EXPORTS testing::AssertionResult assertMatNear(const char* expr1, const char* expr2, const char* eps_expr, cv::InputArray m1, cv::InputArray m2, double eps);
+
+    #define EXPECT_MAT_NEAR(m1, m2, eps) EXPECT_PRED_FORMAT3(cvtest::assertMatNear, m1, m2, eps)
+    #define ASSERT_MAT_NEAR(m1, m2, eps) ASSERT_PRED_FORMAT3(cvtest::assertMatNear, m1, m2, eps)
+
+    #define EXPECT_SCALAR_NEAR(s1, s2, eps) \
+        { \
+            EXPECT_NEAR(s1[0], s2[0], eps); \
+            EXPECT_NEAR(s1[1], s2[1], eps); \
+            EXPECT_NEAR(s1[2], s2[2], eps); \
+            EXPECT_NEAR(s1[3], s2[3], eps); \
+        }
+    #define ASSERT_SCALAR_NEAR(s1, s2, eps) \
+        { \
+            ASSERT_NEAR(s1[0], s2[0], eps); \
+            ASSERT_NEAR(s1[1], s2[1], eps); \
+            ASSERT_NEAR(s1[2], s2[2], eps); \
+            ASSERT_NEAR(s1[3], s2[3], eps); \
+        }
+
+    #define EXPECT_POINT2_NEAR(p1, p2, eps) \
+        { \
+            EXPECT_NEAR(p1.x, p2.x, eps); \
+            EXPECT_NEAR(p1.y, p2.y, eps); \
+        }
+    #define ASSERT_POINT2_NEAR(p1, p2, eps) \
+        { \
+            ASSERT_NEAR(p1.x, p2.x, eps); \
+            ASSERT_NEAR(p1.y, p2.y, eps); \
+        }
+
+    #define EXPECT_POINT3_NEAR(p1, p2, eps) \
+        { \
+            EXPECT_NEAR(p1.x, p2.x, eps); \
+            EXPECT_NEAR(p1.y, p2.y, eps); \
+            EXPECT_NEAR(p1.z, p2.z, eps); \
+        }
+    #define ASSERT_POINT3_NEAR(p1, p2, eps) \
+        { \
+            ASSERT_NEAR(p1.x, p2.x, eps); \
+            ASSERT_NEAR(p1.y, p2.y, eps); \
+            ASSERT_NEAR(p1.z, p2.z, eps); \
+        }
+
+    CV_EXPORTS double checkSimilarity(cv::InputArray m1, cv::InputArray m2);
+
+    #define EXPECT_MAT_SIMILAR(mat1, mat2, eps) \
+        { \
+            ASSERT_EQ(mat1.type(), mat2.type()); \
+            ASSERT_EQ(mat1.size(), mat2.size()); \
+            EXPECT_LE(checkSimilarity(mat1, mat2), eps); \
+        }
+    #define ASSERT_MAT_SIMILAR(mat1, mat2, eps) \
+        { \
+            ASSERT_EQ(mat1.type(), mat2.type()); \
+            ASSERT_EQ(mat1.size(), mat2.size()); \
+            ASSERT_LE(checkSimilarity(mat1, mat2), eps); \
+        }
+
+    //////////////////////////////////////////////////////////////////////
+    // Helper structs for value-parameterized tests
+
+    #define GPU_TEST_P(test_case_name, test_name) \
+      class GTEST_TEST_CLASS_NAME_(test_case_name, test_name) \
+          : public test_case_name { \
+       public: \
+        GTEST_TEST_CLASS_NAME_(test_case_name, test_name)() {} \
+        virtual void TestBody(); \
+       private: \
+        void UnsafeTestBody(); \
+        static int AddToRegistry() { \
+          ::testing::UnitTest::GetInstance()->parameterized_test_registry(). \
+              GetTestCasePatternHolder<test_case_name>(\
+                  #test_case_name, __FILE__, __LINE__)->AddTestPattern(\
+                      #test_case_name, \
+                      #test_name, \
+                      new ::testing::internal::TestMetaFactory< \
+                          GTEST_TEST_CLASS_NAME_(test_case_name, test_name)>()); \
+          return 0; \
+        } \
+        static int gtest_registering_dummy_; \
+        GTEST_DISALLOW_COPY_AND_ASSIGN_(\
+            GTEST_TEST_CLASS_NAME_(test_case_name, test_name)); \
+      }; \
+      int GTEST_TEST_CLASS_NAME_(test_case_name, \
+                                 test_name)::gtest_registering_dummy_ = \
+          GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::AddToRegistry(); \
+      void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::TestBody() \
+      { \
+        try \
+        { \
+          UnsafeTestBody(); \
+        } \
+        catch (...) \
+        { \
+          cv::gpu::resetDevice(); \
+          throw; \
+        } \
+      } \
+      void GTEST_TEST_CLASS_NAME_(test_case_name, test_name)::UnsafeTestBody()
+
+    #define PARAM_TEST_CASE(name, ...) struct name : testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
+    #define GET_PARAM(k) std::tr1::get< k >(GetParam())
+
+    #define DIFFERENT_SIZES testing::Values(cv::Size(128, 128), cv::Size(113, 113))
+
+    // Depth
+
+    using perf::MatDepth;
+
+    #define ALL_DEPTH testing::Values(MatDepth(CV_8U), MatDepth(CV_8S), MatDepth(CV_16U), MatDepth(CV_16S), MatDepth(CV_32S), MatDepth(CV_32F), MatDepth(CV_64F))
+
+    #define DEPTH_PAIRS testing::Values(std::make_pair(MatDepth(CV_8U), MatDepth(CV_8U)),   \
+                                        std::make_pair(MatDepth(CV_8U), MatDepth(CV_16U)),  \
+                                        std::make_pair(MatDepth(CV_8U), MatDepth(CV_16S)),  \
+                                        std::make_pair(MatDepth(CV_8U), MatDepth(CV_32S)),  \
+                                        std::make_pair(MatDepth(CV_8U), MatDepth(CV_32F)),  \
+                                        std::make_pair(MatDepth(CV_8U), MatDepth(CV_64F)),  \
+                                                                                            \
+                                        std::make_pair(MatDepth(CV_16U), MatDepth(CV_16U)), \
+                                        std::make_pair(MatDepth(CV_16U), MatDepth(CV_32S)), \
+                                        std::make_pair(MatDepth(CV_16U), MatDepth(CV_32F)), \
+                                        std::make_pair(MatDepth(CV_16U), MatDepth(CV_64F)), \
+                                                                                            \
+                                        std::make_pair(MatDepth(CV_16S), MatDepth(CV_16S)), \
+                                        std::make_pair(MatDepth(CV_16S), MatDepth(CV_32S)), \
+                                        std::make_pair(MatDepth(CV_16S), MatDepth(CV_32F)), \
+                                        std::make_pair(MatDepth(CV_16S), MatDepth(CV_64F)), \
+                                                                                            \
+                                        std::make_pair(MatDepth(CV_32S), MatDepth(CV_32S)), \
+                                        std::make_pair(MatDepth(CV_32S), MatDepth(CV_32F)), \
+                                        std::make_pair(MatDepth(CV_32S), MatDepth(CV_64F)), \
+                                                                                            \
+                                        std::make_pair(MatDepth(CV_32F), MatDepth(CV_32F)), \
+                                        std::make_pair(MatDepth(CV_32F), MatDepth(CV_64F)), \
+                                                                                            \
+                                        std::make_pair(MatDepth(CV_64F), MatDepth(CV_64F)))
+
+    // Type
+
+    using perf::MatType;
+
+    //! return vector with types from specified range.
+    CV_EXPORTS std::vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end);
+
+    //! return vector with all types (depth: CV_8U-CV_64F, channels: 1-4).
+    CV_EXPORTS const std::vector<MatType>& all_types();
+
+    #define ALL_TYPES testing::ValuesIn(all_types())
+    #define TYPES(depth_start, depth_end, cn_start, cn_end) testing::ValuesIn(types(depth_start, depth_end, cn_start, cn_end))
+
+    // ROI
+
+    class UseRoi
+    {
+    public:
+        inline UseRoi(bool val = false) : val_(val) {}
+
+        inline operator bool() const { return val_; }
+
+    private:
+        bool val_;
+    };
+
+    CV_EXPORTS void PrintTo(const UseRoi& useRoi, std::ostream* os);
+
+    #define WHOLE_SUBMAT testing::Values(UseRoi(false), UseRoi(true))
+
+    // Direct/Inverse
+
+    class Inverse
+    {
+    public:
+        inline Inverse(bool val = false) : val_(val) {}
+
+        inline operator bool() const { return val_; }
+
+    private:
+        bool val_;
+    };
+
+    CV_EXPORTS void PrintTo(const Inverse& useRoi, std::ostream* os);
+
+    #define DIRECT_INVERSE testing::Values(Inverse(false), Inverse(true))
+
+    // Param class
+
+    #define IMPLEMENT_PARAM_CLASS(name, type) \
+        class name \
+        { \
+        public: \
+            name ( type arg = type ()) : val_(arg) {} \
+            operator type () const {return val_;} \
+        private: \
+            type val_; \
+        }; \
+        inline void PrintTo( name param, std::ostream* os) \
+        { \
+            *os << #name <<  "(" << testing::PrintToString(static_cast< type >(param)) << ")"; \
+        }
+
+    IMPLEMENT_PARAM_CLASS(Channels, int)
+
+    #define ALL_CHANNELS testing::Values(Channels(1), Channels(2), Channels(3), Channels(4))
+    #define IMAGE_CHANNELS testing::Values(Channels(1), Channels(3), Channels(4))
+
+    // Flags and enums
+
+    CV_ENUM(NormCode, NORM_INF, NORM_L1, NORM_L2, NORM_TYPE_MASK, NORM_RELATIVE, NORM_MINMAX)
+
+    CV_ENUM(Interpolation, INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_AREA)
+
+    CV_ENUM(BorderType, BORDER_REFLECT101, BORDER_REPLICATE, BORDER_CONSTANT, BORDER_REFLECT, BORDER_WRAP)
+    #define ALL_BORDER_TYPES testing::Values(BorderType(cv::BORDER_REFLECT101), BorderType(cv::BORDER_REPLICATE), BorderType(cv::BORDER_CONSTANT), BorderType(cv::BORDER_REFLECT), BorderType(cv::BORDER_WRAP))
+
+    CV_FLAGS(WarpFlags, INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, WARP_INVERSE_MAP)
+
+    //////////////////////////////////////////////////////////////////////
+    // Features2D
+
+    CV_EXPORTS testing::AssertionResult assertKeyPointsEquals(const char* gold_expr, const char* actual_expr, std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual);
+
+    #define ASSERT_KEYPOINTS_EQ(gold, actual) EXPECT_PRED_FORMAT2(assertKeyPointsEquals, gold, actual)
+
+    CV_EXPORTS int getMatchedPointsCount(std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual);
+    CV_EXPORTS int getMatchedPointsCount(const std::vector<cv::KeyPoint>& keypoints1, const std::vector<cv::KeyPoint>& keypoints2, const std::vector<cv::DMatch>& matches);
+
+    //////////////////////////////////////////////////////////////////////
+    // Other
+
+    CV_EXPORTS void dumpImage(const std::string& fileName, const cv::Mat& image);
+    CV_EXPORTS void showDiff(cv::InputArray gold, cv::InputArray actual, double eps);
+
+    CV_EXPORTS void printCudaInfo();
+}
+
+namespace cv { namespace gpu
+{
+    CV_EXPORTS void PrintTo(const DeviceInfo& info, std::ostream* os);
+}}
+
+#ifdef HAVE_CUDA
+
+#define CV_GPU_TEST_MAIN(resourcesubdir) \
+    int main(int argc, char* argv[]) \
+    { \
+        try \
+        { \
+            cv::CommandLineParser cmd(argc, argv, \
+                "{ h help ?            |      | Print help}" \
+                "{ i info              |      | Print information about system and exit }" \
+                "{ device              | -1   | Device on which tests will be executed (-1 means all devices) }" \
+            ); \
+            if (cmd.has("help")) \
+            { \
+                cmd.printMessage(); \
+                return 0; \
+            } \
+            cvtest::printCudaInfo(); \
+            if (cmd.has("info")) \
+            { \
+                return 0; \
+            } \
+            int device = cmd.get<int>("device"); \
+            if (device < 0) \
+            { \
+                cvtest::DeviceManager::instance().loadAll(); \
+                std::cout << "Run tests on all supported devices \n" << std::endl; \
+            } \
+            else \
+            { \
+                cvtest::DeviceManager::instance().load(device); \
+                cv::gpu::DeviceInfo info(device); \
+                std::cout << "Run tests on device " << device << " [" << info.name() << "] \n" << std::endl; \
+            } \
+            cvtest::TS::ptr()->init( resourcesubdir ); \
+            testing::InitGoogleTest(&argc, argv); \
+            return RUN_ALL_TESTS(); \
+        } \
+        catch (const std::exception& e) \
+        { \
+            std::cerr << e.what() << std::endl; \
+            return -1; \
+        } \
+        catch (...) \
+        { \
+            std::cerr << "Unknown error" << std::endl; \
+            return -1; \
+        } \
+        return 0; \
+    }
+
+#else // HAVE_CUDA
+
+#define CV_GPU_TEST_MAIN(resourcesubdir) \
+    int main() \
+    { \
+        printf("OpenCV was built without CUDA support\n"); \
+        return 0; \
+    }
+
+#endif // HAVE_CUDA
+
+
+#endif // __OPENCV_GPU_TEST_UTILITY_HPP__
diff --git a/modules/ts/include/opencv2/ts/ts_gtest.h b/modules/ts/include/opencv2/ts/ts_gtest.h
index f26059111f..60f30b8f20 100644
--- a/modules/ts/include/opencv2/ts/ts_gtest.h
+++ b/modules/ts/include/opencv2/ts/ts_gtest.h
@@ -128,6 +128,10 @@
 // Low-level types and utilities for porting Google Test to various
 // platforms.  They are subject to change without notice.  DO NOT USE
 // THEM IN USER CODE.
+//
+// This file is fundamental to Google Test.  All other Google Test source
+// files are expected to #include this.  Therefore, it cannot #include
+// any other Google Test header.
 
 #ifndef GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_H_
 #define GTEST_INCLUDE_GTEST_INTERNAL_GTEST_PORT_H_
@@ -1919,7 +1923,7 @@ using ::std::tuple_size;
 #endif  // _MSC_VER
 
 #ifndef GTEST_API_
-# define GTEST_API_
+# define GTEST_API_ CV_EXPORTS
 #endif
 
 #ifdef __GNUC__
@@ -1942,6 +1946,11 @@ class Message;
 
 namespace internal {
 
+// A secret type that Google Test users don't know about.  It has no
+// definition on purpose.  Therefore it's impossible to create a
+// Secret object, which is what we want.
+class Secret;
+
 // The GTEST_COMPILE_ASSERT_ macro can be used to verify that a compile time
 // expression is true. For example, you could use it to verify the
 // size of a static array:
@@ -1962,8 +1971,8 @@ struct CompileAssert {
 };
 
 #define GTEST_COMPILE_ASSERT_(expr, msg) \
-  typedef ::testing::internal::CompileAssert<(bool(expr))> \
-      msg[bool(expr) ? 1 : -1]
+  typedef ::testing::internal::CompileAssert<(static_cast<bool>(expr))> \
+      msg[static_cast<bool>(expr) ? 1 : -1] GTEST_ATTRIBUTE_UNUSED_
 
 // Implementation details of GTEST_COMPILE_ASSERT_:
 //
@@ -3102,12 +3111,265 @@ const char* StringFromGTestEnv(const char* flag, const char* default_val);
 # include <unistd.h>
 #endif  // GTEST_OS_LINUX
 
+#if GTEST_HAS_EXCEPTIONS
+# include <stdexcept>
+#endif
+
 #include <ctype.h>
 #include <string.h>
 #include <iomanip>
 #include <limits>
 #include <set>
 
+// Copyright 2005, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+//
+// Author: wan@google.com (Zhanyong Wan)
+//
+// The Google C++ Testing Framework (Google Test)
+//
+// This header file defines the Message class.
+//
+// IMPORTANT NOTE: Due to limitation of the C++ language, we have to
+// leave some internal implementation details in this header file.
+// They are clearly marked by comments like this:
+//
+//   // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
+//
+// Such code is NOT meant to be used by a user directly, and is subject
+// to CHANGE WITHOUT NOTICE.  Therefore DO NOT DEPEND ON IT in a user
+// program!
+
+#ifndef GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
+#define GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
+
+#include <limits>
+
+
+// Ensures that there is at least one operator<< in the global namespace.
+// See Message& operator<<(...) below for why.
+void operator<<(const testing::internal::Secret&, int);
+
+namespace testing {
+
+// The Message class works like an ostream repeater.
+//
+// Typical usage:
+//
+//   1. You stream a bunch of values to a Message object.
+//      It will remember the text in a stringstream.
+//   2. Then you stream the Message object to an ostream.
+//      This causes the text in the Message to be streamed
+//      to the ostream.
+//
+// For example;
+//
+//   testing::Message foo;
+//   foo << 1 << " != " << 2;
+//   std::cout << foo;
+//
+// will print "1 != 2".
+//
+// Message is not intended to be inherited from.  In particular, its
+// destructor is not virtual.
+//
+// Note that stringstream behaves differently in gcc and in MSVC.  You
+// can stream a NULL char pointer to it in the former, but not in the
+// latter (it causes an access violation if you do).  The Message
+// class hides this difference by treating a NULL char pointer as
+// "(null)".
+class GTEST_API_ Message {
+ private:
+  // The type of basic IO manipulators (endl, ends, and flush) for
+  // narrow streams.
+  typedef std::ostream& (*BasicNarrowIoManip)(std::ostream&);
+
+ public:
+  // Constructs an empty Message.
+  Message();
+
+  // Copy constructor.
+  Message(const Message& msg) : ss_(new ::std::stringstream) {  // NOLINT
+    *ss_ << msg.GetString();
+  }
+
+  // Constructs a Message from a C-string.
+  explicit Message(const char* str) : ss_(new ::std::stringstream) {
+    *ss_ << str;
+  }
+
+#if GTEST_OS_SYMBIAN
+  // Streams a value (either a pointer or not) to this object.
+  template <typename T>
+  inline Message& operator <<(const T& value) {
+    StreamHelper(typename internal::is_pointer<T>::type(), value);
+    return *this;
+  }
+#else
+  // Streams a non-pointer value to this object.
+  template <typename T>
+  inline Message& operator <<(const T& val) {
+    // Some libraries overload << for STL containers.  These
+    // overloads are defined in the global namespace instead of ::std.
+    //
+    // C++'s symbol lookup rule (i.e. Koenig lookup) says that these
+    // overloads are visible in either the std namespace or the global
+    // namespace, but not other namespaces, including the testing
+    // namespace which Google Test's Message class is in.
+    //
+    // To allow STL containers (and other types that has a << operator
+    // defined in the global namespace) to be used in Google Test
+    // assertions, testing::Message must access the custom << operator
+    // from the global namespace.  With this using declaration,
+    // overloads of << defined in the global namespace and those
+    // visible via Koenig lookup are both exposed in this function.
+    using ::operator <<;
+    *ss_ << val;
+    return *this;
+  }
+
+  // Streams a pointer value to this object.
+  //
+  // This function is an overload of the previous one.  When you
+  // stream a pointer to a Message, this definition will be used as it
+  // is more specialized.  (The C++ Standard, section
+  // [temp.func.order].)  If you stream a non-pointer, then the
+  // previous definition will be used.
+  //
+  // The reason for this overload is that streaming a NULL pointer to
+  // ostream is undefined behavior.  Depending on the compiler, you
+  // may get "0", "(nil)", "(null)", or an access violation.  To
+  // ensure consistent result across compilers, we always treat NULL
+  // as "(null)".
+  template <typename T>
+  inline Message& operator <<(T* const& pointer) {  // NOLINT
+    if (pointer == NULL) {
+      *ss_ << "(null)";
+    } else {
+      *ss_ << pointer;
+    }
+    return *this;
+  }
+#endif  // GTEST_OS_SYMBIAN
+
+  // Since the basic IO manipulators are overloaded for both narrow
+  // and wide streams, we have to provide this specialized definition
+  // of operator <<, even though its body is the same as the
+  // templatized version above.  Without this definition, streaming
+  // endl or other basic IO manipulators to Message will confuse the
+  // compiler.
+  Message& operator <<(BasicNarrowIoManip val) {
+    *ss_ << val;
+    return *this;
+  }
+
+  // Instead of 1/0, we want to see true/false for bool values.
+  Message& operator <<(bool b) {
+    return *this << (b ? "true" : "false");
+  }
+
+  // These two overloads allow streaming a wide C string to a Message
+  // using the UTF-8 encoding.
+  Message& operator <<(const wchar_t* wide_c_str);
+  Message& operator <<(wchar_t* wide_c_str);
+
+#if GTEST_HAS_STD_WSTRING
+  // Converts the given wide string to a narrow string using the UTF-8
+  // encoding, and streams the result to this Message object.
+  Message& operator <<(const ::std::wstring& wstr);
+#endif  // GTEST_HAS_STD_WSTRING
+
+#if GTEST_HAS_GLOBAL_WSTRING
+  // Converts the given wide string to a narrow string using the UTF-8
+  // encoding, and streams the result to this Message object.
+  Message& operator <<(const ::wstring& wstr);
+#endif  // GTEST_HAS_GLOBAL_WSTRING
+
+  // Gets the text streamed to this object so far as an std::string.
+  // Each '\0' character in the buffer is replaced with "\\0".
+  //
+  // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
+  std::string GetString() const;
+
+ private:
+
+#if GTEST_OS_SYMBIAN
+  // These are needed as the Nokia Symbian Compiler cannot decide between
+  // const T& and const T* in a function template. The Nokia compiler _can_
+  // decide between class template specializations for T and T*, so a
+  // tr1::type_traits-like is_pointer works, and we can overload on that.
+  template <typename T>
+  inline void StreamHelper(internal::true_type /*is_pointer*/, T* pointer) {
+    if (pointer == NULL) {
+      *ss_ << "(null)";
+    } else {
+      *ss_ << pointer;
+    }
+  }
+  template <typename T>
+  inline void StreamHelper(internal::false_type /*is_pointer*/,
+                           const T& value) {
+    // See the comments in Message& operator <<(const T&) above for why
+    // we need this using statement.
+    using ::operator <<;
+    *ss_ << value;
+  }
+#endif  // GTEST_OS_SYMBIAN
+
+  // We'll hold the text streamed to this object here.
+  const internal::scoped_ptr< ::std::stringstream> ss_;
+
+  // We declare (but don't implement) this to prevent the compiler
+  // from implementing the assignment operator.
+  void operator=(const Message&);
+};
+
+// Streams a Message to an ostream.
+inline std::ostream& operator <<(std::ostream& os, const Message& sb) {
+  return os << sb.GetString();
+}
+
+namespace internal {
+
+// Converts a streamable value to an std::string.  A NULL pointer is
+// converted to "(null)".  When the input value is a ::string,
+// ::std::string, ::wstring, or ::std::wstring object, each NUL
+// character in it is replaced with "\\0".
+template <typename T>
+std::string StreamableToString(const T& streamable) {
+  return (Message() << streamable).GetString();
+}
+
+}  // namespace internal
+}  // namespace testing
+
+#endif  // GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
 // Copyright 2005, Google Inc.
 // All rights reserved.
 //
@@ -3253,16 +3515,14 @@ class GTEST_API_ String {
   static bool EndsWithCaseInsensitive(
       const std::string& str, const std::string& suffix);
 
-  // Formats a list of arguments to an std::string, using the same format
-  // spec string as for printf.
-  //
-  // We do not use the StringPrintf class as it is not universally
-  // available.
-  //
-  // The result is limited to 4096 characters (including the tailing
-  // 0).  If 4096 characters are not enough to format the input,
-  // "<buffer exceeded>" is returned.
-  static std::string Format(const char* format, ...);
+  // Formats an int value as "%02d".
+  static std::string FormatIntWidth2(int value);  // "%02d" for width == 2
+
+  // Formats an int value as "%X".
+  static std::string FormatHexInt(int value);
+
+  // Formats a byte as "%02X".
+  static std::string FormatByte(unsigned char value);
 
  private:
   String();  // Not meant to be instantiated.
@@ -3272,17 +3532,6 @@ class GTEST_API_ String {
 // character in the buffer is replaced with "\\0".
 GTEST_API_ std::string StringStreamToString(::std::stringstream* stream);
 
-// Converts a streamable value to an std::string.  A NULL pointer is
-// converted to "(null)".  When the input value is a ::string,
-// ::std::string, ::wstring, or ::std::wstring object, each NUL
-// character in it is replaced with "\\0".
-
-// Declared here but defined in gtest.h, so that it has access
-// to the definition of the Message class, required by the ARM
-// compiler.
-template <typename T>
-std::string StreamableToString(const T& streamable);
-
 }  // namespace internal
 }  // namespace testing
 
@@ -6834,36 +7083,6 @@ struct TypeList<Types<T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13,
 #define GTEST_CONCAT_TOKEN_(foo, bar) GTEST_CONCAT_TOKEN_IMPL_(foo, bar)
 #define GTEST_CONCAT_TOKEN_IMPL_(foo, bar) foo ## bar
 
-// Google Test defines the testing::Message class to allow construction of
-// test messages via the << operator.  The idea is that anything
-// streamable to std::ostream can be streamed to a testing::Message.
-// This allows a user to use his own types in Google Test assertions by
-// overloading the << operator.
-//
-// util/gtl/stl_logging.h overloads << for STL containers.  These
-// overloads cannot be defined in the std namespace, as that will be
-// undefined behavior.  Therefore, they are defined in the global
-// namespace instead.
-//
-// C++'s symbol lookup rule (i.e. Koenig lookup) says that these
-// overloads are visible in either the std namespace or the global
-// namespace, but not other namespaces, including the testing
-// namespace which Google Test's Message class is in.
-//
-// To allow STL containers (and other types that has a << operator
-// defined in the global namespace) to be used in Google Test assertions,
-// testing::Message must access the custom << operator from the global
-// namespace.  Hence this helper function.
-//
-// Note: Jeffrey Yasskin suggested an alternative fix by "using
-// ::operator<<;" in the definition of Message's operator<<.  That fix
-// doesn't require a helper function, but unfortunately doesn't
-// compile with MSVC.
-template <typename T>
-inline void GTestStreamToHelper(std::ostream* os, const T& val) {
-  *os << val;
-}
-
 class ProtocolMessage;
 namespace proto2 { class Message; }
 
@@ -6895,11 +7114,6 @@ GTEST_API_ extern int g_init_gtest_count;
 // stack trace.
 GTEST_API_ extern const char kStackTraceMarker[];
 
-// A secret type that Google Test users don't know about.  It has no
-// definition on purpose.  Therefore it's impossible to create a
-// Secret object, which is what we want.
-class Secret;
-
 // Two overloaded helpers for checking at compile time whether an
 // expression is a null pointer literal (i.e. NULL or any 0-valued
 // compile-time integral constant).  Their return values have
@@ -6933,6 +7147,31 @@ char (&IsNullLiteralHelper(...))[2];  // NOLINT
 GTEST_API_ std::string AppendUserMessage(
     const std::string& gtest_msg, const Message& user_msg);
 
+#if GTEST_HAS_EXCEPTIONS
+
+// This exception is thrown by (and only by) a failed Google Test
+// assertion when GTEST_FLAG(throw_on_failure) is true (if exceptions
+// are enabled).  We derive it from std::runtime_error, which is for
+// errors presumably detectable only at run time.  Since
+// std::runtime_error inherits from std::exception, many testing
+// frameworks know how to extract and print the message inside it.
+
+#ifdef _MSC_VER
+# pragma warning(push)          // Saves the current warning state.
+# pragma warning(disable:4275)  // Temporarily disables warning 4275.
+#endif  // _MSC_VER
+
+class GTEST_API_ GoogleTestFailureException : public ::std::runtime_error {
+ public:
+  explicit GoogleTestFailureException(const TestPartResult& failure);
+};
+
+#ifdef _MSC_VER
+# pragma warning(pop)           // Restores the warning state.
+#endif  // _MSC_VER
+
+#endif  // GTEST_HAS_EXCEPTIONS
+
 // A helper class for creating scoped traces in user programs.
 class GTEST_API_ ScopedTrace {
  public:
@@ -6952,16 +7191,6 @@ class GTEST_API_ ScopedTrace {
                             // c'tor and d'tor.  Therefore it doesn't
                             // need to be used otherwise.
 
-// Converts a streamable value to an std::string.  A NULL pointer is
-// converted to "(null)".  When the input value is a ::string,
-// ::std::string, ::wstring, or ::std::wstring object, each NUL
-// character in it is replaced with "\\0".
-// Declared here but defined in gtest.h, so that it has access
-// to the definition of the Message class, required by the ARM
-// compiler.
-template <typename T>
-std::string StreamableToString(const T& streamable);
-
 // Constructs and returns the message for an equality assertion
 // (e.g. ASSERT_EQ, EXPECT_STREQ, etc) failure.
 //
@@ -7059,7 +7288,7 @@ class FloatingPoint {
   // bits.  Therefore, 4 should be enough for ordinary use.
   //
   // See the following article for more details on ULP:
-  // http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm.
+  // http://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
   static const size_t kMaxUlps = 4;
 
   // Constructs a FloatingPoint from a raw floating-point number.
@@ -7260,7 +7489,7 @@ typedef void (*TearDownTestCaseFunc)();
 //   test_case_name:   name of the test case
 //   name:             name of the test
 //   type_param        the name of the test's type parameter, or NULL if
-//                     this is not  a typed or a type-parameterized test.
+//                     this is not a typed or a type-parameterized test.
 //   value_param       text representation of the test's value parameter,
 //                     or NULL if this is not a type-parameterized test.
 //   fixture_class_id: ID of the test fixture class
@@ -7270,7 +7499,8 @@ typedef void (*TearDownTestCaseFunc)();
 //                     The newly created TestInfo instance will assume
 //                     ownership of the factory object.
 GTEST_API_ TestInfo* MakeAndRegisterTestInfo(
-    const char* test_case_name, const char* name,
+    const char* test_case_name,
+    const char* name,
     const char* type_param,
     const char* value_param,
     TypeId fixture_class_id,
@@ -7358,8 +7588,8 @@ class TypeParameterizedTest {
     // First, registers the first type-parameterized test in the type
     // list.
     MakeAndRegisterTestInfo(
-        String::Format("%s%s%s/%d", prefix, prefix[0] == '\0' ? "" : "/",
-                       case_name, index).c_str(),
+        (std::string(prefix) + (prefix[0] == '\0' ? "" : "/") + case_name + "/"
+         + StreamableToString(index)).c_str(),
         GetPrefixUntilComma(test_names).c_str(),
         GetTypeName<Type>().c_str(),
         NULL,  // No value parameter.
@@ -8547,234 +8777,6 @@ class GTEST_API_ KilledBySignal {
 }  // namespace testing
 
 #endif  // GTEST_INCLUDE_GTEST_GTEST_DEATH_TEST_H_
-// Copyright 2005, Google Inc.
-// All rights reserved.
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
-// Author: wan@google.com (Zhanyong Wan)
-//
-// The Google C++ Testing Framework (Google Test)
-//
-// This header file defines the Message class.
-//
-// IMPORTANT NOTE: Due to limitation of the C++ language, we have to
-// leave some internal implementation details in this header file.
-// They are clearly marked by comments like this:
-//
-//   // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-//
-// Such code is NOT meant to be used by a user directly, and is subject
-// to CHANGE WITHOUT NOTICE.  Therefore DO NOT DEPEND ON IT in a user
-// program!
-
-#ifndef GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
-#define GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
-
-#include <limits>
-
-
-namespace testing {
-
-// The Message class works like an ostream repeater.
-//
-// Typical usage:
-//
-//   1. You stream a bunch of values to a Message object.
-//      It will remember the text in a stringstream.
-//   2. Then you stream the Message object to an ostream.
-//      This causes the text in the Message to be streamed
-//      to the ostream.
-//
-// For example;
-//
-//   testing::Message foo;
-//   foo << 1 << " != " << 2;
-//   std::cout << foo;
-//
-// will print "1 != 2".
-//
-// Message is not intended to be inherited from.  In particular, its
-// destructor is not virtual.
-//
-// Note that stringstream behaves differently in gcc and in MSVC.  You
-// can stream a NULL char pointer to it in the former, but not in the
-// latter (it causes an access violation if you do).  The Message
-// class hides this difference by treating a NULL char pointer as
-// "(null)".
-class GTEST_API_ Message {
- private:
-  // The type of basic IO manipulators (endl, ends, and flush) for
-  // narrow streams.
-  typedef std::ostream& (*BasicNarrowIoManip)(std::ostream&);
-
- public:
-  // Constructs an empty Message.
-  // We allocate the stringstream separately because otherwise each use of
-  // ASSERT/EXPECT in a procedure adds over 200 bytes to the procedure's
-  // stack frame leading to huge stack frames in some cases; gcc does not reuse
-  // the stack space.
-  Message() : ss_(new ::std::stringstream) {
-    // By default, we want there to be enough precision when printing
-    // a double to a Message.
-    *ss_ << std::setprecision(std::numeric_limits<double>::digits10 + 2);
-  }
-
-  // Copy constructor.
-  Message(const Message& msg) : ss_(new ::std::stringstream) {  // NOLINT
-    *ss_ << msg.GetString();
-  }
-
-  // Constructs a Message from a C-string.
-  explicit Message(const char* str) : ss_(new ::std::stringstream) {
-    *ss_ << str;
-  }
-
-#if GTEST_OS_SYMBIAN
-  // Streams a value (either a pointer or not) to this object.
-  template <typename T>
-  inline Message& operator <<(const T& value) {
-    StreamHelper(typename internal::is_pointer<T>::type(), value);
-    return *this;
-  }
-#else
-  // Streams a non-pointer value to this object.
-  template <typename T>
-  inline Message& operator <<(const T& val) {
-    ::GTestStreamToHelper(ss_.get(), val);
-    return *this;
-  }
-
-  // Streams a pointer value to this object.
-  //
-  // This function is an overload of the previous one.  When you
-  // stream a pointer to a Message, this definition will be used as it
-  // is more specialized.  (The C++ Standard, section
-  // [temp.func.order].)  If you stream a non-pointer, then the
-  // previous definition will be used.
-  //
-  // The reason for this overload is that streaming a NULL pointer to
-  // ostream is undefined behavior.  Depending on the compiler, you
-  // may get "0", "(nil)", "(null)", or an access violation.  To
-  // ensure consistent result across compilers, we always treat NULL
-  // as "(null)".
-  template <typename T>
-  inline Message& operator <<(T* const& pointer) {  // NOLINT
-    if (pointer == NULL) {
-      *ss_ << "(null)";
-    } else {
-      ::GTestStreamToHelper(ss_.get(), pointer);
-    }
-    return *this;
-  }
-#endif  // GTEST_OS_SYMBIAN
-
-  // Since the basic IO manipulators are overloaded for both narrow
-  // and wide streams, we have to provide this specialized definition
-  // of operator <<, even though its body is the same as the
-  // templatized version above.  Without this definition, streaming
-  // endl or other basic IO manipulators to Message will confuse the
-  // compiler.
-  Message& operator <<(BasicNarrowIoManip val) {
-    *ss_ << val;
-    return *this;
-  }
-
-  // Instead of 1/0, we want to see true/false for bool values.
-  Message& operator <<(bool b) {
-    return *this << (b ? "true" : "false");
-  }
-
-  // These two overloads allow streaming a wide C string to a Message
-  // using the UTF-8 encoding.
-  Message& operator <<(const wchar_t* wide_c_str) {
-    return *this << internal::String::ShowWideCString(wide_c_str);
-  }
-  Message& operator <<(wchar_t* wide_c_str) {
-    return *this << internal::String::ShowWideCString(wide_c_str);
-  }
-
-#if GTEST_HAS_STD_WSTRING
-  // Converts the given wide string to a narrow string using the UTF-8
-  // encoding, and streams the result to this Message object.
-  Message& operator <<(const ::std::wstring& wstr);
-#endif  // GTEST_HAS_STD_WSTRING
-
-#if GTEST_HAS_GLOBAL_WSTRING
-  // Converts the given wide string to a narrow string using the UTF-8
-  // encoding, and streams the result to this Message object.
-  Message& operator <<(const ::wstring& wstr);
-#endif  // GTEST_HAS_GLOBAL_WSTRING
-
-  // Gets the text streamed to this object so far as an std::string.
-  // Each '\0' character in the buffer is replaced with "\\0".
-  //
-  // INTERNAL IMPLEMENTATION - DO NOT USE IN A USER PROGRAM.
-  std::string GetString() const {
-    return internal::StringStreamToString(ss_.get());
-  }
-
- private:
-
-#if GTEST_OS_SYMBIAN
-  // These are needed as the Nokia Symbian Compiler cannot decide between
-  // const T& and const T* in a function template. The Nokia compiler _can_
-  // decide between class template specializations for T and T*, so a
-  // tr1::type_traits-like is_pointer works, and we can overload on that.
-  template <typename T>
-  inline void StreamHelper(internal::true_type /*dummy*/, T* pointer) {
-    if (pointer == NULL) {
-      *ss_ << "(null)";
-    } else {
-      ::GTestStreamToHelper(ss_.get(), pointer);
-    }
-  }
-  template <typename T>
-  inline void StreamHelper(internal::false_type /*dummy*/, const T& value) {
-    ::GTestStreamToHelper(ss_.get(), value);
-  }
-#endif  // GTEST_OS_SYMBIAN
-
-  // We'll hold the text streamed to this object here.
-  const internal::scoped_ptr< ::std::stringstream> ss_;
-
-  // We declare (but don't implement) this to prevent the compiler
-  // from implementing the assignment operator.
-  void operator=(const Message&);
-};
-
-// Streams a Message to an ostream.
-inline std::ostream& operator <<(std::ostream& os, const Message& sb) {
-  return os << sb.GetString();
-}
-
-}  // namespace testing
-
-#endif  // GTEST_INCLUDE_GTEST_GTEST_MESSAGE_H_
 // This file was GENERATED by command:
 //     pump.py gtest-param-test.h.pump
 // DO NOT EDIT BY HAND!!!
@@ -10541,10 +10543,10 @@ class ParameterizedTestCaseInfo : public ParameterizedTestCaseInfoBase {
         const string& instantiation_name = gen_it->first;
         ParamGenerator<ParamType> generator((*gen_it->second)());
 
-        Message test_case_name_stream;
+        string test_case_name;
         if ( !instantiation_name.empty() )
-          test_case_name_stream << instantiation_name << "/";
-        test_case_name_stream << test_info->test_case_base_name;
+          test_case_name = instantiation_name + "/";
+        test_case_name += test_info->test_case_base_name;
 
         int i = 0;
         for (typename ParamGenerator<ParamType>::iterator param_it =
@@ -10553,7 +10555,7 @@ class ParameterizedTestCaseInfo : public ParameterizedTestCaseInfoBase {
           Message test_name_stream;
           test_name_stream << test_info->test_base_name << "/" << i;
           MakeAndRegisterTestInfo(
-              test_case_name_stream.GetString().c_str(),
+              test_case_name.c_str(),
               test_name_stream.GetString().c_str(),
               NULL,  // No type parameter.
               PrintToString(*param_it).c_str(),
@@ -17564,6 +17566,9 @@ GTEST_DECLARE_string_(color);
 // the tests to run. If the filter is not given all tests are executed.
 GTEST_DECLARE_string_(filter);
 
+// OpenCV extension: same as filter, but for the parameters string.
+GTEST_DECLARE_string_(param_filter);
+
 // This flag causes the Google Test to list tests. None of the tests listed
 // are actually run if the flag is provided.
 GTEST_DECLARE_bool_(list_tests);
@@ -17615,26 +17620,16 @@ class ExecDeathTest;
 class NoExecDeathTest;
 class FinalSuccessChecker;
 class GTestFlagSaver;
+class StreamingListenerTest;
 class TestResultAccessor;
 class TestEventListenersAccessor;
 class TestEventRepeater;
+class UnitTestRecordPropertyTestHelper;
 class WindowsDeathTest;
 class UnitTestImpl* GetUnitTestImpl();
 void ReportFailureInUnknownLocation(TestPartResult::Type result_type,
                                     const std::string& message);
 
-// Converts a streamable value to an std::string.  A NULL pointer is
-// converted to "(null)".  When the input value is a ::string,
-// ::std::string, ::wstring, or ::std::wstring object, each NUL
-// character in it is replaced with "\\0".
-// Declared in gtest-internal.h but defined here, so that it has access
-// to the definition of the Message class, required by the ARM
-// compiler.
-template <typename T>
-std::string StreamableToString(const T& streamable) {
-  return (Message() << streamable).GetString();
-}
-
 }  // namespace internal
 
 // The friend relationship of some of these classes is cyclic.
@@ -17853,20 +17848,21 @@ class GTEST_API_ Test {
   // non-fatal) failure.
   static bool HasFailure() { return HasFatalFailure() || HasNonfatalFailure(); }
 
-  // Logs a property for the current test.  Only the last value for a given
-  // key is remembered.
-  // These are public static so they can be called from utility functions
-  // that are not members of the test fixture.
-  // The arguments are const char* instead strings, as Google Test is used
-  // on platforms where string doesn't compile.
-  //
-  // Note that a driving consideration for these RecordProperty methods
-  // was to produce xml output suited to the Greenspan charting utility,
-  // which at present will only chart values that fit in a 32-bit int. It
-  // is the user's responsibility to restrict their values to 32-bit ints
-  // if they intend them to be used with Greenspan.
-  static void RecordProperty(const char* key, const char* value);
-  static void RecordProperty(const char* key, int value);
+  // Logs a property for the current test, test case, or for the entire
+  // invocation of the test program when used outside of the context of a
+  // test case.  Only the last value for a given key is remembered.  These
+  // are public static so they can be called from utility functions that are
+  // not members of the test fixture.  Calls to RecordProperty made during
+  // lifespan of the test (from the moment its constructor starts to the
+  // moment its destructor finishes) will be output in XML as attributes of
+  // the <testcase> element.  Properties recorded from fixture's
+  // SetUpTestCase or TearDownTestCase are logged as attributes of the
+  // corresponding <testsuite> element.  Calls to RecordProperty made in the
+  // global context (before or after invocation of RUN_ALL_TESTS and from
+  // SetUp/TearDown method of Environment objects registered with Google
+  // Test) will be output as attributes of the <testsuites> element.
+  static void RecordProperty(const std::string& key, const std::string& value);
+  static void RecordProperty(const std::string& key, int value);
 
  protected:
   // Creates a Test object.
@@ -17935,7 +17931,7 @@ class TestProperty {
   // C'tor.  TestProperty does NOT have a default constructor.
   // Always use this constructor (with parameters) to create a
   // TestProperty object.
-  TestProperty(const char* a_key, const char* a_value) :
+  TestProperty(const std::string& a_key, const std::string& a_value) :
     key_(a_key), value_(a_value) {
   }
 
@@ -17950,7 +17946,7 @@ class TestProperty {
   }
 
   // Sets a new value, overriding the one supplied in the constructor.
-  void SetValue(const char* new_value) {
+  void SetValue(const std::string& new_value) {
     value_ = new_value;
   }
 
@@ -18009,6 +18005,7 @@ class GTEST_API_ TestResult {
 
  private:
   friend class TestInfo;
+  friend class TestCase;
   friend class UnitTest;
   friend class internal::DefaultGlobalTestPartResultReporter;
   friend class internal::ExecDeathTest;
@@ -18033,13 +18030,16 @@ class GTEST_API_ TestResult {
   // a non-fatal failure if invalid (e.g., if it conflicts with reserved
   // key names). If a property is already recorded for the same key, the
   // value will be updated, rather than storing multiple values for the same
-  // key.
-  void RecordProperty(const TestProperty& test_property);
+  // key.  xml_element specifies the element for which the property is being
+  // recorded and is used for validation.
+  void RecordProperty(const std::string& xml_element,
+                      const TestProperty& test_property);
 
   // Adds a failure if the key is a reserved attribute of Google Test
   // testcase tags.  Returns true if the property is valid.
   // TODO(russr): Validate attribute names are legal and human readable.
-  static bool ValidateTestProperty(const TestProperty& test_property);
+  static bool ValidateTestProperty(const std::string& xml_element,
+                                   const TestProperty& test_property);
 
   // Adds a test part result to the list.
   void AddTestPartResult(const TestPartResult& test_part_result);
@@ -18140,8 +18140,10 @@ class GTEST_API_ TestInfo {
   friend class Test;
   friend class TestCase;
   friend class internal::UnitTestImpl;
+  friend class internal::StreamingListenerTest;
   friend TestInfo* internal::MakeAndRegisterTestInfo(
-      const char* test_case_name, const char* name,
+      const char* test_case_name,
+      const char* name,
       const char* type_param,
       const char* value_param,
       internal::TypeId fixture_class_id,
@@ -18151,9 +18153,10 @@ class GTEST_API_ TestInfo {
 
   // Constructs a TestInfo object. The newly constructed instance assumes
   // ownership of the factory object.
-  TestInfo(const char* test_case_name, const char* name,
-           const char* a_type_param,
-           const char* a_value_param,
+  TestInfo(const std::string& test_case_name,
+           const std::string& name,
+           const char* a_type_param,   // NULL if not a type-parameterized test
+           const char* a_value_param,  // NULL if not a value-parameterized test
            internal::TypeId fixture_class_id,
            internal::TestFactoryBase* factory);
 
@@ -18261,6 +18264,10 @@ class GTEST_API_ TestCase {
   // total_test_count() - 1. If i is not in that range, returns NULL.
   const TestInfo* GetTestInfo(int i) const;
 
+  // Returns the TestResult that holds test properties recorded during
+  // execution of SetUpTestCase and TearDownTestCase.
+  const TestResult& ad_hoc_test_result() const { return ad_hoc_test_result_; }
+
  private:
   friend class Test;
   friend class internal::UnitTestImpl;
@@ -18349,6 +18356,9 @@ class GTEST_API_ TestCase {
   bool should_run_;
   // Elapsed time, in milliseconds.
   TimeInMillis elapsed_time_;
+  // Holds test properties recorded during execution of SetUpTestCase and
+  // TearDownTestCase.
+  TestResult ad_hoc_test_result_;
 
   // We disallow copying TestCases.
   GTEST_DISALLOW_COPY_AND_ASSIGN_(TestCase);
@@ -18634,6 +18644,10 @@ class GTEST_API_ UnitTest {
   // total_test_case_count() - 1. If i is not in that range, returns NULL.
   const TestCase* GetTestCase(int i) const;
 
+  // Returns the TestResult containing information on test failures and
+  // properties logged outside of individual test cases.
+  const TestResult& ad_hoc_test_result() const;
+
   // Returns the list of event listeners that can be used to track events
   // inside Google Test.
   TestEventListeners& listeners();
@@ -18661,9 +18675,12 @@ class GTEST_API_ UnitTest {
                          const std::string& os_stack_trace)
       GTEST_LOCK_EXCLUDED_(mutex_);
 
-  // Adds a TestProperty to the current TestResult object. If the result already
-  // contains a property with the same key, the value will be updated.
-  void RecordPropertyForCurrentTest(const char* key, const char* value);
+  // Adds a TestProperty to the current TestResult object when invoked from
+  // inside a test, to current TestCase's ad_hoc_test_result_ when invoked
+  // from SetUpTestCase or TearDownTestCase, or to the global property set
+  // when invoked elsewhere.  If the result already contains a property with
+  // the same key, the value will be updated.
+  void RecordProperty(const std::string& key, const std::string& value);
 
   // Gets the i-th test case among all the test cases. i can range from 0 to
   // total_test_case_count() - 1. If i is not in that range, returns NULL.
@@ -18678,6 +18695,8 @@ class GTEST_API_ UnitTest {
   friend class Test;
   friend class internal::AssertHelper;
   friend class internal::ScopedTrace;
+  friend class internal::StreamingListenerTest;
+  friend class internal::UnitTestRecordPropertyTestHelper;
   friend Environment* AddGlobalTestEnvironment(Environment* env);
   friend internal::UnitTestImpl* internal::GetUnitTestImpl();
   friend void internal::ReportFailureInUnknownLocation(
@@ -19223,7 +19242,12 @@ class WithParamInterface {
   // references static data, to reduce the opportunity for incorrect uses
   // like writing 'WithParamInterface<bool>::GetParam()' for a test that
   // uses a fixture whose parameter type is int.
-  const ParamType& GetParam() const { return *parameter_; }
+  const ParamType& GetParam() const {
+    GTEST_CHECK_(parameter_ != NULL)
+        << "GetParam() can only be called inside a value-parameterized test "
+        << "-- did you intend to write TEST_P instead of TEST_F?";
+    return *parameter_;
+  }
 
  private:
   // Sets parameter value. The caller is responsible for making sure the value
@@ -20040,15 +20064,20 @@ bool StaticAssertTypeEq() {
   GTEST_TEST_(test_fixture, test_name, test_fixture, \
               ::testing::internal::GetTypeId<test_fixture>())
 
-// Use this macro in main() to run all tests.  It returns 0 if all
+}  // namespace testing
+
+// Use this function in main() to run all tests.  It returns 0 if all
 // tests are successful, or 1 otherwise.
 //
 // RUN_ALL_TESTS() should be invoked after the command line has been
 // parsed by InitGoogleTest().
+//
+// This function was formerly a macro; thus, it is in the global
+// namespace and has an all-caps name.
+int RUN_ALL_TESTS() GTEST_MUST_USE_RESULT_;
 
-#define RUN_ALL_TESTS()\
-  (::testing::UnitTest::GetInstance()->Run())
-
-}  // namespace testing
+inline int RUN_ALL_TESTS() {
+  return ::testing::UnitTest::GetInstance()->Run();
+}
 
 #endif  // GTEST_INCLUDE_GTEST_GTEST_H_
diff --git a/modules/ts/include/opencv2/ts/ts_perf.hpp b/modules/ts/include/opencv2/ts/ts_perf.hpp
index cd8209fc67..74ea47e13b 100644
--- a/modules/ts/include/opencv2/ts/ts_perf.hpp
+++ b/modules/ts/include/opencv2/ts/ts_perf.hpp
@@ -2,13 +2,8 @@
 #define __OPENCV_TS_PERF_HPP__
 
 #include "opencv2/core.hpp"
-#include "opencv2/features2d.hpp"
 #include "ts_gtest.h"
 
-#ifdef HAVE_TBB
-#include "tbb/task_scheduler_init.h"
-#endif
-
 #if !(defined(LOGD) || defined(LOGI) || defined(LOGW) || defined(LOGE))
 # if defined(ANDROID) && defined(USE_ANDROID_LOGGING)
 #  include <android/log.h>
@@ -26,6 +21,9 @@
 # endif
 #endif
 
+// declare major namespaces to avoid errors on unknown namespace
+namespace cv { namespace gpu {} namespace ocl {} }
+
 namespace perf
 {
 class TestBase;
@@ -92,66 +90,64 @@ private:
 *     CV_ENUM and CV_FLAGS - macro to create printable wrappers for defines and enums     *
 \*****************************************************************************************/
 
-#define CV_ENUM(class_name, ...) \
-namespace { class CV_EXPORTS class_name {\
-public:\
-  class_name(int val = 0) : _val(val) {}\
-  operator int() const {return _val;}\
-  void PrintTo(std::ostream* os) const {\
-    const int vals[] = {__VA_ARGS__};\
-    const char* svals = #__VA_ARGS__;\
-    for(int i = 0, pos = 0; i < (int)(sizeof(vals)/sizeof(int)); ++i){\
-      while(isspace(svals[pos]) || svals[pos] == ',') ++pos;\
-      int start = pos;\
-      while(!(isspace(svals[pos]) || svals[pos] == ',' || svals[pos] == 0)) ++pos;\
-      if (_val == vals[i]) {\
-        *os << std::string(svals + start, svals + pos);\
-        return;\
-      }\
-    }\
-    *os << "UNKNOWN";\
-  }\
-  struct Container{\
-    typedef class_name value_type;\
-      Container(class_name* first, size_t len): _begin(first), _end(first+len){}\
-      const class_name* begin() const {return _begin;}\
-      const class_name* end() const {return _end;}\
-    private: class_name *_begin, *_end;\
-  };\
-  static Container all(){\
-    static int vals[] = {__VA_ARGS__};\
-    return Container((class_name*)vals, sizeof(vals)/sizeof(vals[0]));\
-  }\
-private: int _val;\
-};\
-inline void PrintTo(const class_name& t, std::ostream* os) { t.PrintTo(os); } }
+#define CV_ENUM(class_name, ...)                                                        \
+    namespace {                                                                         \
+    struct class_name {                                                                 \
+        class_name(int val = 0) : val_(val) {}                                          \
+        operator int() const { return val_; }                                           \
+        void PrintTo(std::ostream* os) const {                                          \
+            using namespace cv;using namespace cv::gpu; using namespace cv::ocl;        \
+            const int vals[] = { __VA_ARGS__ };                                         \
+            const char* svals = #__VA_ARGS__;                                           \
+            for(int i = 0, pos = 0; i < (int)(sizeof(vals)/sizeof(int)); ++i) {         \
+                while(isspace(svals[pos]) || svals[pos] == ',') ++pos;                  \
+                int start = pos;                                                        \
+                while(!(isspace(svals[pos]) || svals[pos] == ',' || svals[pos] == 0))   \
+                    ++pos;                                                              \
+                if (val_ == vals[i]) {                                                  \
+                    *os << std::string(svals + start, svals + pos);                     \
+                    return;                                                             \
+                }                                                                       \
+            }                                                                           \
+            *os << "UNKNOWN";                                                           \
+        }                                                                               \
+        static ::testing::internal::ParamGenerator<class_name> all() {                  \
+            using namespace cv;using namespace cv::gpu; using namespace cv::ocl;        \
+            static class_name vals[] = { __VA_ARGS__ };                                 \
+            return ::testing::ValuesIn(vals);                                           \
+        }                                                                               \
+    private: int val_;                                                                  \
+    };                                                                                  \
+    inline void PrintTo(const class_name& t, std::ostream* os) { t.PrintTo(os); } }
 
-#define CV_FLAGS(class_name, ...) \
-class CV_EXPORTS class_name {\
-public:\
-  class_name(int val = 0) : _val(val) {}\
-  operator int() const {return _val;}\
-  void PrintTo(std::ostream* os) const {\
-    const int vals[] = {__VA_ARGS__};\
-    const char* svals = #__VA_ARGS__;\
-    int value = _val;\
-    bool first = true;\
-    for(int i = 0, pos = 0; i < (int)(sizeof(vals)/sizeof(int)); ++i){\
-      while(isspace(svals[pos]) || svals[pos] == ',') ++pos;\
-      int start = pos;\
-      while(!(isspace(svals[pos]) || svals[pos] == ',' || svals[pos] == 0)) ++pos;\
-      if ((value & vals[i]) == vals[i]) {\
-        value &= ~vals[i]; \
-        if (first) first = false; else *os << "|"; \
-        *os << std::string(svals + start, svals + pos);\
-        if (!value) return;\
-      }\
-    }\
-    if (first) *os << "UNKNOWN";\
-  }\
-private: int _val;\
-};\
-inline void PrintTo(const class_name& t, std::ostream* os) { t.PrintTo(os); }
+#define CV_FLAGS(class_name, ...)                                                       \
+    namespace {                                                                         \
+    struct class_name {                                                                 \
+        class_name(int val = 0) : val_(val) {}                                          \
+        operator int() const { return val_; }                                           \
+        void PrintTo(std::ostream* os) const {                                          \
+            using namespace cv;using namespace cv::gpu; using namespace cv::ocl;        \
+            const int vals[] = { __VA_ARGS__ };                                         \
+            const char* svals = #__VA_ARGS__;                                           \
+            int value = val_;                                                           \
+            bool first = true;                                                          \
+            for(int i = 0, pos = 0; i < (int)(sizeof(vals)/sizeof(int)); ++i) {         \
+                while(isspace(svals[pos]) || svals[pos] == ',') ++pos;                  \
+                int start = pos;                                                        \
+                while(!(isspace(svals[pos]) || svals[pos] == ',' || svals[pos] == 0))   \
+                    ++pos;                                                              \
+                if ((value & vals[i]) == vals[i]) {                                     \
+                    value &= ~vals[i];                                                  \
+                    if (first) first = false; else *os << "|";                          \
+                    *os << std::string(svals + start, svals + pos);                     \
+                    if (!value) return;                                                 \
+                }                                                                       \
+            }                                                                           \
+            if (first) *os << "UNKNOWN";                                                \
+        }                                                                               \
+    private: int val_;                                                                  \
+    };                                                                                  \
+    inline void PrintTo(const class_name& t, std::ostream* os) { t.PrintTo(os); } }
 
 CV_ENUM(MatDepth, CV_8U, CV_8S, CV_16U, CV_16S, CV_32S, CV_32F, CV_64F, CV_USRTYPE1)
 
@@ -208,18 +204,13 @@ private:
 #define SANITY_CHECK_KEYPOINTS(array, ...) ::perf::Regression::addKeypoints(this, #array, array , ## __VA_ARGS__)
 #define SANITY_CHECK_MATCHES(array, ...) ::perf::Regression::addMatches(this, #array, array , ## __VA_ARGS__)
 
-#ifdef HAVE_CUDA
 class CV_EXPORTS GpuPerf
 {
 public:
   static bool targetDevice();
 };
 
-# define PERF_RUN_GPU()  ::perf::GpuPerf::targetDevice()
-#else
-# define PERF_RUN_GPU()  false
-#endif
-
+#define PERF_RUN_GPU()  ::perf::GpuPerf::targetDevice()
 
 /*****************************************************************************************\
 *                            Container for performance metrics                            *
@@ -261,7 +252,11 @@ public:
     TestBase();
 
     static void Init(int argc, const char* const argv[]);
+    static void Init(const std::vector<std::string> & availableImpls,
+                     int argc, const char* const argv[]);
+    static void RecordRunParameters();
     static std::string getDataPath(const std::string& relativePath);
+    static std::string getSelectedImpl();
 
 protected:
     virtual void PerfTestBody() = 0;
@@ -474,15 +469,37 @@ CV_EXPORTS void PrintTo(const Size& sz, ::std::ostream* os);
     INSTANTIATE_TEST_CASE_P(/*none*/, fixture##_##name, params);\
     void fixture##_##name::PerfTestBody()
 
+#if defined(_MSC_VER) && (_MSC_VER <= 1400)
+#define CV_PERF_TEST_MAIN_INTERNALS_ARGS(...)	\
+    while (++argc >= (--argc,-1)) {__VA_ARGS__; break;} /*this ugly construction is needed for VS 2005*/
+#else
+#define CV_PERF_TEST_MAIN_INTERNALS_ARGS(...)	\
+    __VA_ARGS__;
+#endif
 
-#define CV_PERF_TEST_MAIN(testsuitname, ...) \
+#define CV_PERF_TEST_MAIN_INTERNALS(modulename, impls, ...)	\
+    CV_PERF_TEST_MAIN_INTERNALS_ARGS(__VA_ARGS__) \
+    ::perf::Regression::Init(#modulename);\
+    ::perf::TestBase::Init(std::vector<std::string>(impls, impls + sizeof impls / sizeof *impls),\
+                           argc, argv);\
+    ::testing::InitGoogleTest(&argc, argv);\
+    cvtest::printVersionInfo();\
+    ::testing::Test::RecordProperty("cv_module_name", #modulename);\
+    ::perf::TestBase::RecordRunParameters();\
+    return RUN_ALL_TESTS();
+
+// impls must be an array, not a pointer; "plain" should always be one of the implementations
+#define CV_PERF_TEST_MAIN_WITH_IMPLS(modulename, impls, ...) \
 int main(int argc, char **argv)\
 {\
-    while (++argc >= (--argc,-1)) {__VA_ARGS__; break;} /*this ugly construction is needed for VS 2005*/\
-    ::perf::Regression::Init(#testsuitname);\
-    ::perf::TestBase::Init(argc, argv);\
-    ::testing::InitGoogleTest(&argc, argv);\
-    return RUN_ALL_TESTS();\
+    CV_PERF_TEST_MAIN_INTERNALS(modulename, impls, __VA_ARGS__)\
+}
+
+#define CV_PERF_TEST_MAIN(modulename, ...) \
+int main(int argc, char **argv)\
+{\
+    const char * plain_only[] = { "plain" };\
+    CV_PERF_TEST_MAIN_INTERNALS(modulename, plain_only, __VA_ARGS__)\
 }
 
 #define TEST_CYCLE_N(n) for(declare.iterations(n); startTimer(), next(); stopTimer())
diff --git a/modules/ts/misc/perf_tests_timing.py b/modules/ts/misc/perf_tests_timing.py
index d082f9c265..dd44f1a639 100644
--- a/modules/ts/misc/perf_tests_timing.py
+++ b/modules/ts/misc/perf_tests_timing.py
@@ -156,4 +156,4 @@ if __name__ == "__main__":
             if not options.failedOnly:
                 print '\nOverall time: %.2f min\n' % overall_time
             tbl.consolePrintTable(sys.stdout)
-            print 2 * '\n'
\ No newline at end of file
+            print 2 * '\n'
diff --git a/modules/ts/misc/report.py b/modules/ts/misc/report.py
index 6da89e512b..045fee8d15 100755
--- a/modules/ts/misc/report.py
+++ b/modules/ts/misc/report.py
@@ -100,4 +100,4 @@ if __name__ == "__main__":
             tbl.htmlPrintTable(sys.stdout)
             htmlPrintFooter(sys.stdout)
     else:
-        tbl.consolePrintTable(sys.stdout)
\ No newline at end of file
+        tbl.consolePrintTable(sys.stdout)
diff --git a/modules/ts/misc/run.py b/modules/ts/misc/run.py
index 4351713715..a64127f0d4 100755
--- a/modules/ts/misc/run.py
+++ b/modules/ts/misc/run.py
@@ -288,6 +288,16 @@ class TestSuite(object):
             if self.adb:
                 # construct name for aapt tool
                 self.aapt = [os.path.join(os.path.dirname(self.adb[0]), ("aapt","aapt.exe")[hostos == 'nt'])]
+                if not os.path.isfile(self.aapt[0]):
+                    # it's moved in SDK r22
+                    sdk_dir = os.path.dirname( os.path.dirname(self.adb[0]) )
+                    aapt_fn = ("aapt", "aapt.exe")[hostos == 'nt']
+                    for r, ds, fs in os.walk( os.path.join(sdk_dir, 'build-tools') ):
+                        if aapt_fn in fs:
+                            self.aapt = [ os.path.join(r, aapt_fn) ]
+                            break
+                    else:
+                        self.error = "Can't find '%s' tool!" % aapt_fn
 
         # fix has_perf_tests param
         self.has_perf_tests = self.has_perf_tests == "ON"
diff --git a/modules/ts/misc/testlog_parser.py b/modules/ts/misc/testlog_parser.py
index f61b47bba9..4ab0a3ef2f 100755
--- a/modules/ts/misc/testlog_parser.py
+++ b/modules/ts/misc/testlog_parser.py
@@ -1,6 +1,9 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
-import sys, re, os.path
+import collections
+import re
+import os.path
+import sys
 from xml.dom.minidom import parse
 
 class TestInfo(object):
@@ -10,10 +13,17 @@ class TestInfo(object):
         self.name = xmlnode.getAttribute("name")
         self.value_param = xmlnode.getAttribute("value_param")
         self.type_param = xmlnode.getAttribute("type_param")
-        if xmlnode.getElementsByTagName("failure"):
+
+        custom_status = xmlnode.getAttribute("custom_status")
+        failures = xmlnode.getElementsByTagName("failure")
+
+        if len(custom_status) > 0:
+            self.status = custom_status
+        elif len(failures) > 0:
             self.status = "failed"
         else:
             self.status = xmlnode.getAttribute("status")
+
         if self.name.startswith("DISABLED_"):
             self.status = "disabled"
             self.fixture = self.fixture.replace("DISABLED_", "")
@@ -100,34 +110,39 @@ class TestInfo(object):
     def dump(self, units="ms"):
         print "%s ->\t\033[1;31m%s\033[0m = \t%.2f%s" % (str(self), self.status, self.get("gmean", units), units)
 
-    def shortName(self):
+
+    def getName(self):
         pos = self.name.find("/")
         if pos > 0:
-            name = self.name[:pos]
-        else:
-            name = self.name
-        if self.fixture.endswith(name):
-            fixture = self.fixture[:-len(name)]
+            return self.name[:pos]
+        return self.name
+
+
+    def getFixture(self):
+        if self.fixture.endswith(self.getName()):
+            fixture = self.fixture[:-len(self.getName())]
         else:
             fixture = self.fixture
         if fixture.endswith("_"):
             fixture = fixture[:-1]
+        return fixture
+
+
+    def param(self):
+        return '::'.join(filter(None, [self.type_param, self.value_param]))
+
+    def shortName(self):
+        name = self.getName()
+        fixture = self.getFixture()
         return '::'.join(filter(None, [name, fixture]))
 
+
     def __str__(self):
-        pos = self.name.find("/")
-        if pos > 0:
-            name = self.name[:pos]
-        else:
-            name = self.name
-        if self.fixture.endswith(name):
-            fixture = self.fixture[:-len(name)]
-        else:
-            fixture = self.fixture
-        if fixture.endswith("_"):
-            fixture = fixture[:-1]
+        name = self.getName()
+        fixture = self.getFixture()
         return '::'.join(filter(None, [name, fixture, self.type_param, self.value_param]))
 
+
     def __cmp__(self, other):
         r = cmp(self.fixture, other.fixture);
         if r != 0:
@@ -154,12 +169,31 @@ class TestInfo(object):
                 return 1
         return 0
 
+# This is a Sequence for compatibility with old scripts,
+# which treat parseLogFile's return value as a list.
+class TestRunInfo(collections.Sequence):
+    def __init__(self, properties, tests):
+        self.properties = properties
+        self.tests = tests
+
+    def __len__(self):
+        return len(self.tests)
+
+    def __getitem__(self, key):
+        return self.tests[key]
+
 def parseLogFile(filename):
-    tests = []
     log = parse(filename)
-    for case in log.getElementsByTagName("testcase"):
-        tests.append(TestInfo(case))
-    return tests
+
+    properties = {
+        attr_name[3:]: attr_value
+        for (attr_name, attr_value) in log.documentElement.attributes.items()
+        if attr_name.startswith('cv_')
+    }
+
+    tests = map(TestInfo, log.getElementsByTagName("testcase"))
+
+    return TestRunInfo(properties, tests)
 
 
 if __name__ == "__main__":
@@ -168,8 +202,18 @@ if __name__ == "__main__":
         exit(0)
 
     for arg in sys.argv[1:]:
-        print "Tests found in", arg
-        tests = parseLogFile(arg)
-        for t in sorted(tests):
+        print "Processing {}...".format(arg)
+
+        run = parseLogFile(arg)
+
+        print "Properties:"
+
+        for (prop_name, prop_value) in run.properties.items():
+          print "\t{} = {}".format(prop_name, prop_value)
+
+        print "Tests:"
+
+        for t in sorted(run.tests):
             t.dump()
+
         print
diff --git a/modules/ts/misc/xls-report.py b/modules/ts/misc/xls-report.py
new file mode 100755
index 0000000000..e71a7f66c7
--- /dev/null
+++ b/modules/ts/misc/xls-report.py
@@ -0,0 +1,371 @@
+#!/usr/bin/env python
+
+"""
+    This script can generate XLS reports from OpenCV tests' XML output files.
+
+    To use it, first, create a directory for each machine you ran tests on.
+    Each such directory will become a sheet in the report. Put each XML file
+    into the corresponding directory.
+
+    Then, create your configuration file(s). You can have a global configuration
+    file (specified with the -c option), and per-sheet configuration files, which
+    must be called sheet.conf and placed in the directory corresponding to the sheet.
+    The settings in the per-sheet configuration file will override those in the
+    global configuration file, if both are present.
+
+    A configuration file must consist of a Python dictionary. The following keys
+    will be recognized:
+
+    * 'comparisons': [{'from': string, 'to': string}]
+        List of configurations to compare performance between. For each item,
+        the sheet will have a column showing speedup from configuration named
+        'from' to configuration named "to".
+
+    * 'configuration_matchers': [{'properties': {string: object}, 'name': string}]
+        Instructions for matching test run property sets to configuration names.
+
+        For each found XML file:
+
+        1) All attributes of the root element starting with the prefix 'cv_' are
+           placed in a dictionary, with the cv_ prefix stripped and the cv_module_name
+           element deleted.
+
+        2) The first matcher for which the XML's file property set contains the same
+           keys with equal values as its 'properties' dictionary is searched for.
+           A missing property can be matched by using None as the value.
+
+           Corollary 1: you should place more specific matchers before less specific
+           ones.
+
+           Corollary 2: an empty 'properties' dictionary matches every property set.
+
+        3) If a matching matcher is found, its 'name' string is presumed to be the name
+           of the configuration the XML file corresponds to. A warning is printed if
+           two different property sets match to the same configuration name.
+
+        4) If a such a matcher isn't found, if --include-unmatched was specified, the
+           configuration name is assumed to be the relative path from the sheet's
+           directory to the XML file's containing directory. If the XML file isinstance
+           directly inside the sheet's directory, the configuration name is instead
+           a dump of all its properties. If --include-unmatched wasn't specified,
+           the XML file is ignored and a warning is printed.
+
+    * 'configurations': [string]
+        List of names for compile-time and runtime configurations of OpenCV.
+        Each item will correspond to a column of the sheet.
+
+    * 'module_colors': {string: string}
+        Mapping from module name to color name. In the sheet, cells containing module
+        names from this mapping will be colored with the corresponding color. You can
+        find the list of available colors here:
+        <http://www.simplistix.co.uk/presentations/python-excel.pdf>.
+
+    * 'sheet_name': string
+        Name for the sheet. If this parameter is missing, the name of sheet's directory
+        will be used.
+
+    * 'sheet_properties': [(string, string)]
+        List of arbitrary (key, value) pairs that somehow describe the sheet. Will be
+        dumped into the first row of the sheet in string form.
+
+    Note that all keys are optional, although to get useful results, you'll want to
+    specify at least 'configurations' and 'configuration_matchers'.
+
+    Finally, run the script. Use the --help option for usage information.
+"""
+
+from __future__ import division
+
+import ast
+import errno
+import fnmatch
+import logging
+import numbers
+import os, os.path
+import re
+
+from argparse import ArgumentParser
+from glob import glob
+from itertools import ifilter
+
+import xlwt
+
+from testlog_parser import parseLogFile
+
+re_image_size = re.compile(r'^ \d+ x \d+$', re.VERBOSE)
+re_data_type = re.compile(r'^ (?: 8 | 16 | 32 | 64 ) [USF] C [1234] $', re.VERBOSE)
+
+time_style = xlwt.easyxf(num_format_str='#0.00')
+no_time_style = xlwt.easyxf('pattern: pattern solid, fore_color gray25')
+
+speedup_style = time_style
+good_speedup_style = xlwt.easyxf('font: color green', num_format_str='#0.00')
+bad_speedup_style = xlwt.easyxf('font: color red', num_format_str='#0.00')
+no_speedup_style = no_time_style
+error_speedup_style = xlwt.easyxf('pattern: pattern solid, fore_color orange')
+header_style = xlwt.easyxf('font: bold true; alignment: horizontal centre, vertical top, wrap True')
+subheader_style = xlwt.easyxf('alignment: horizontal centre, vertical top')
+
+class Collector(object):
+    def __init__(self, config_match_func, include_unmatched):
+        self.__config_cache = {}
+        self.config_match_func = config_match_func
+        self.include_unmatched = include_unmatched
+        self.tests = {}
+        self.extra_configurations = set()
+
+    # Format a sorted sequence of pairs as if it was a dictionary.
+    # We can't just use a dictionary instead, since we want to preserve the sorted order of the keys.
+    @staticmethod
+    def __format_config_cache_key(pairs, multiline=False):
+        return (
+          ('{\n' if multiline else '{') +
+          (',\n' if multiline else ', ').join(
+             ('  ' if multiline else '') + repr(k) + ': ' + repr(v) for (k, v) in pairs) +
+          ('\n}\n' if multiline else '}')
+        )
+
+    def collect_from(self, xml_path, default_configuration):
+        run = parseLogFile(xml_path)
+
+        module = run.properties['module_name']
+
+        properties = run.properties.copy()
+        del properties['module_name']
+
+        props_key = tuple(sorted(properties.iteritems())) # dicts can't be keys
+
+        if props_key in self.__config_cache:
+            configuration = self.__config_cache[props_key]
+        else:
+            configuration = self.config_match_func(properties)
+
+            if configuration is None:
+                if self.include_unmatched:
+                    if default_configuration is not None:
+                        configuration = default_configuration
+                    else:
+                        configuration = Collector.__format_config_cache_key(props_key, multiline=True)
+
+                    self.extra_configurations.add(configuration)
+                else:
+                    logging.warning('failed to match properties to a configuration: %s',
+                        Collector.__format_config_cache_key(props_key))
+
+            else:
+                same_config_props = [it[0] for it in self.__config_cache.iteritems() if it[1] == configuration]
+                if len(same_config_props) > 0:
+                    logging.warning('property set %s matches the same configuration %r as property set %s',
+                        Collector.__format_config_cache_key(props_key),
+                        configuration,
+                        Collector.__format_config_cache_key(same_config_props[0]))
+
+            self.__config_cache[props_key] = configuration
+
+        if configuration is None: return
+
+        module_tests = self.tests.setdefault(module, {})
+
+        for test in run.tests:
+            test_results = module_tests.setdefault((test.shortName(), test.param()), {})
+            new_result = test.get("gmean") if test.status == 'run' else test.status
+            test_results[configuration] = min(
+              test_results.get(configuration), new_result,
+              key=lambda r: (1, r) if isinstance(r, numbers.Number) else
+                            (2,) if r is not None else
+                            (3,)
+            ) # prefer lower result; prefer numbers to errors and errors to nothing
+
+def make_match_func(matchers):
+    def match_func(properties):
+        for matcher in matchers:
+            if all(properties.get(name) == value
+                   for (name, value) in matcher['properties'].iteritems()):
+                return matcher['name']
+
+        return None
+
+    return match_func
+
+def main():
+    arg_parser = ArgumentParser(description='Build an XLS performance report.')
+    arg_parser.add_argument('sheet_dirs', nargs='+', metavar='DIR', help='directory containing perf test logs')
+    arg_parser.add_argument('-o', '--output', metavar='XLS', default='report.xls', help='name of output file')
+    arg_parser.add_argument('-c', '--config', metavar='CONF', help='global configuration file')
+    arg_parser.add_argument('--include-unmatched', action='store_true',
+        help='include results from XML files that were not recognized by configuration matchers')
+    arg_parser.add_argument('--show-times-per-pixel', action='store_true',
+        help='for tests that have an image size parameter, show per-pixel time, as well as total time')
+
+    args = arg_parser.parse_args()
+
+    logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.DEBUG)
+
+    if args.config is not None:
+        with open(args.config) as global_conf_file:
+            global_conf = ast.literal_eval(global_conf_file.read())
+    else:
+        global_conf = {}
+
+    wb = xlwt.Workbook()
+
+    for sheet_path in args.sheet_dirs:
+        try:
+            with open(os.path.join(sheet_path, 'sheet.conf')) as sheet_conf_file:
+                sheet_conf = ast.literal_eval(sheet_conf_file.read())
+        except IOError as ioe:
+            if ioe.errno != errno.ENOENT: raise
+            sheet_conf = {}
+            logging.debug('no sheet.conf for %s', sheet_path)
+
+        sheet_conf = dict(global_conf.items() + sheet_conf.items())
+
+        config_names = sheet_conf.get('configurations', [])
+        config_matchers = sheet_conf.get('configuration_matchers', [])
+
+        collector = Collector(make_match_func(config_matchers), args.include_unmatched)
+
+        for root, _, filenames in os.walk(sheet_path):
+            logging.info('looking in %s', root)
+            for filename in fnmatch.filter(filenames, '*.xml'):
+                if os.path.normpath(sheet_path) == os.path.normpath(root):
+                  default_conf = None
+                else:
+                  default_conf = os.path.relpath(root, sheet_path)
+                collector.collect_from(os.path.join(root, filename), default_conf)
+
+        config_names.extend(sorted(collector.extra_configurations - set(config_names)))
+
+        sheet = wb.add_sheet(sheet_conf.get('sheet_name', os.path.basename(os.path.abspath(sheet_path))))
+
+        sheet_properties = sheet_conf.get('sheet_properties', [])
+
+        sheet.write(0, 0, 'Properties:')
+
+        sheet.write(0, 1,
+          'N/A' if len(sheet_properties) == 0 else
+          ' '.join(str(k) + '=' + repr(v) for (k, v) in sheet_properties))
+
+        sheet.row(2).height = 800
+        sheet.panes_frozen = True
+        sheet.remove_splits = True
+
+        sheet_comparisons = sheet_conf.get('comparisons', [])
+
+        row = 2
+
+        col = 0
+
+        for (w, caption) in [
+                (2500, 'Module'),
+                (10000, 'Test'),
+                (2000, 'Image\nwidth'),
+                (2000, 'Image\nheight'),
+                (2000, 'Data\ntype'),
+                (7500, 'Other parameters')]:
+            sheet.col(col).width = w
+            if args.show_times_per_pixel:
+                sheet.write_merge(row, row + 1, col, col, caption, header_style)
+            else:
+                sheet.write(row, col, caption, header_style)
+            col += 1
+
+        for config_name in config_names:
+            if args.show_times_per_pixel:
+                sheet.col(col).width = 3000
+                sheet.col(col + 1).width = 3000
+                sheet.write_merge(row, row, col, col + 1, config_name, header_style)
+                sheet.write(row + 1, col, 'total, ms', subheader_style)
+                sheet.write(row + 1, col + 1, 'per pixel, ns', subheader_style)
+                col += 2
+            else:
+                sheet.col(col).width = 4000
+                sheet.write(row, col, config_name, header_style)
+                col += 1
+
+        col += 1 # blank column between configurations and comparisons
+
+        for comp in sheet_comparisons:
+            sheet.col(col).width = 4000
+            caption = comp['to'] + '\nvs\n' + comp['from']
+            if args.show_times_per_pixel:
+                sheet.write_merge(row, row + 1, col, col, caption, header_style)
+            else:
+                sheet.write(row, col, caption, header_style)
+            col += 1
+
+        row += 2 if args.show_times_per_pixel else 1
+
+        sheet.horz_split_pos = row
+        sheet.horz_split_first_visible = row
+
+        module_colors = sheet_conf.get('module_colors', {})
+        module_styles = {module: xlwt.easyxf('pattern: pattern solid, fore_color {}'.format(color))
+                         for module, color in module_colors.iteritems()}
+
+        for module, tests in sorted(collector.tests.iteritems()):
+            for ((test, param), configs) in sorted(tests.iteritems()):
+                sheet.write(row, 0, module, module_styles.get(module, xlwt.Style.default_style))
+                sheet.write(row, 1, test)
+
+                param_list = param[1:-1].split(', ') if param.startswith('(') and param.endswith(')') else [param]
+
+                image_size = next(ifilter(re_image_size.match, param_list), None)
+                if image_size is not None:
+                    (image_width, image_height) = map(int, image_size.split('x', 1))
+                    sheet.write(row, 2, image_width)
+                    sheet.write(row, 3, image_height)
+                    del param_list[param_list.index(image_size)]
+
+                data_type = next(ifilter(re_data_type.match, param_list), None)
+                if data_type is not None:
+                    sheet.write(row, 4, data_type)
+                    del param_list[param_list.index(data_type)]
+
+                sheet.row(row).write(5, ' | '.join(param_list))
+
+                col = 6
+
+                for c in config_names:
+                    if c in configs:
+                        sheet.write(row, col, configs[c], time_style)
+                    else:
+                        sheet.write(row, col, None, no_time_style)
+                    col += 1
+                    if args.show_times_per_pixel:
+                        sheet.write(row, col,
+                          xlwt.Formula('{0} * 1000000 / ({1} * {2})'.format(
+                              xlwt.Utils.rowcol_to_cell(row, col - 1),
+                              xlwt.Utils.rowcol_to_cell(row, 2),
+                              xlwt.Utils.rowcol_to_cell(row, 3)
+                          )),
+                          time_style
+                        )
+                        col += 1
+
+                col += 1 # blank column
+
+                for comp in sheet_comparisons:
+                    cmp_from = configs.get(comp["from"])
+                    cmp_to = configs.get(comp["to"])
+
+                    if isinstance(cmp_from, numbers.Number) and isinstance(cmp_to, numbers.Number):
+                        try:
+                            speedup = cmp_from / cmp_to
+                            sheet.write(row, col, speedup, good_speedup_style if speedup > 1.1 else
+                                                           bad_speedup_style  if speedup < 0.9 else
+                                                           speedup_style)
+                        except ArithmeticError as e:
+                            sheet.write(row, col, None, error_speedup_style)
+                    else:
+                        sheet.write(row, col, None, no_speedup_style)
+
+                    col += 1
+
+                row += 1
+                if row % 1000 == 0: sheet.flush_row_data()
+
+    wb.save(args.output)
+
+if __name__ == '__main__':
+    main()
diff --git a/modules/ts/src/gpu_perf.cpp b/modules/ts/src/gpu_perf.cpp
new file mode 100644
index 0000000000..2bca535c46
--- /dev/null
+++ b/modules/ts/src/gpu_perf.cpp
@@ -0,0 +1,355 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include "opencv2/ts/gpu_perf.hpp"
+#include "opencv2/core/gpu.hpp"
+
+#ifdef HAVE_CUDA
+    #include <cuda_runtime.h>
+#endif
+
+using namespace cv;
+using namespace std;
+
+namespace perf
+{
+    Mat readImage(const string& fileName, int flags)
+    {
+        return imread(perf::TestBase::getDataPath(fileName), flags);
+    }
+
+    void PrintTo(const CvtColorInfo& info, std::ostream* os)
+    {
+        static const char* str[] =
+        {
+            "BGR2BGRA",
+            "BGRA2BGR",
+            "BGR2RGBA",
+            "RGBA2BGR",
+            "BGR2RGB",
+            "BGRA2RGBA",
+
+            "BGR2GRAY",
+            "RGB2GRAY",
+            "GRAY2BGR",
+            "GRAY2BGRA",
+            "BGRA2GRAY",
+            "RGBA2GRAY",
+
+            "BGR2BGR565",
+            "RGB2BGR565",
+            "BGR5652BGR",
+            "BGR5652RGB",
+            "BGRA2BGR565",
+            "RGBA2BGR565",
+            "BGR5652BGRA",
+            "BGR5652RGBA",
+
+            "GRAY2BGR565",
+            "BGR5652GRAY",
+
+            "BGR2BGR555",
+            "RGB2BGR555",
+            "BGR5552BGR",
+            "BGR5552RGB",
+            "BGRA2BGR555",
+            "RGBA2BGR555",
+            "BGR5552BGRA",
+            "BGR5552RGBA",
+
+            "GRAY2BGR555",
+            "BGR5552GRAY",
+
+            "BGR2XYZ",
+            "RGB2XYZ",
+            "XYZ2BGR",
+            "XYZ2RGB",
+
+            "BGR2YCrCb",
+            "RGB2YCrCb",
+            "YCrCb2BGR",
+            "YCrCb2RGB",
+
+            "BGR2HSV",
+            "RGB2HSV",
+
+            "",
+            "",
+
+            "BGR2Lab",
+            "RGB2Lab",
+
+            "BayerBG2BGR",
+            "BayerGB2BGR",
+            "BayerRG2BGR",
+            "BayerGR2BGR",
+
+            "BGR2Luv",
+            "RGB2Luv",
+
+            "BGR2HLS",
+            "RGB2HLS",
+
+            "HSV2BGR",
+            "HSV2RGB",
+
+            "Lab2BGR",
+            "Lab2RGB",
+            "Luv2BGR",
+            "Luv2RGB",
+
+            "HLS2BGR",
+            "HLS2RGB",
+
+            "BayerBG2BGR_VNG",
+            "BayerGB2BGR_VNG",
+            "BayerRG2BGR_VNG",
+            "BayerGR2BGR_VNG",
+
+            "BGR2HSV_FULL",
+            "RGB2HSV_FULL",
+            "BGR2HLS_FULL",
+            "RGB2HLS_FULL",
+
+            "HSV2BGR_FULL",
+            "HSV2RGB_FULL",
+            "HLS2BGR_FULL",
+            "HLS2RGB_FULL",
+
+            "LBGR2Lab",
+            "LRGB2Lab",
+            "LBGR2Luv",
+            "LRGB2Luv",
+
+            "Lab2LBGR",
+            "Lab2LRGB",
+            "Luv2LBGR",
+            "Luv2LRGB",
+
+            "BGR2YUV",
+            "RGB2YUV",
+            "YUV2BGR",
+            "YUV2RGB",
+
+            "BayerBG2GRAY",
+            "BayerGB2GRAY",
+            "BayerRG2GRAY",
+            "BayerGR2GRAY",
+
+            //YUV 4:2:0 formats family
+            "YUV2RGB_NV12",
+            "YUV2BGR_NV12",
+            "YUV2RGB_NV21",
+            "YUV2BGR_NV21",
+
+            "YUV2RGBA_NV12",
+            "YUV2BGRA_NV12",
+            "YUV2RGBA_NV21",
+            "YUV2BGRA_NV21",
+
+            "YUV2RGB_YV12",
+            "YUV2BGR_YV12",
+            "YUV2RGB_IYUV",
+            "YUV2BGR_IYUV",
+
+            "YUV2RGBA_YV12",
+            "YUV2BGRA_YV12",
+            "YUV2RGBA_IYUV",
+            "YUV2BGRA_IYUV",
+
+            "YUV2GRAY_420",
+
+            //YUV 4:2:2 formats family
+            "YUV2RGB_UYVY",
+            "YUV2BGR_UYVY",
+            "YUV2RGB_VYUY",
+            "YUV2BGR_VYUY",
+
+            "YUV2RGBA_UYVY",
+            "YUV2BGRA_UYVY",
+            "YUV2RGBA_VYUY",
+            "YUV2BGRA_VYUY",
+
+            "YUV2RGB_YUY2",
+            "YUV2BGR_YUY2",
+            "YUV2RGB_YVYU",
+            "YUV2BGR_YVYU",
+
+            "YUV2RGBA_YUY2",
+            "YUV2BGRA_YUY2",
+            "YUV2RGBA_YVYU",
+            "YUV2BGRA_YVYU",
+
+            "YUV2GRAY_UYVY",
+            "YUV2GRAY_YUY2",
+
+            // alpha premultiplication
+            "RGBA2mRGBA",
+            "mRGBA2RGBA",
+
+            "COLORCVT_MAX"
+        };
+
+        *os << str[info.code];
+    }
+
+    static void printOsInfo()
+    {
+    #if defined _WIN32
+    #   if defined _WIN64
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x64.\n[----------]\n"), fflush(stdout);
+    #   else
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Windows x32.\n[----------]\n"), fflush(stdout);
+    #   endif
+    #elif defined linux
+    #   if defined _LP64
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x64.\n[----------]\n"), fflush(stdout);
+    #   else
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Linux x32.\n[----------]\n"), fflush(stdout);
+    #   endif
+    #elif defined __APPLE__
+    #   if defined _LP64
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x64.\n[----------]\n"), fflush(stdout);
+    #   else
+            printf("[----------]\n[ GPU INFO ] \tRun on OS Apple x32.\n[----------]\n"), fflush(stdout);
+    #   endif
+    #endif
+
+    }
+
+    void printCudaInfo()
+    {
+        printOsInfo();
+    #ifndef HAVE_CUDA
+        printf("[----------]\n[ GPU INFO ] \tOpenCV was built without CUDA support.\n[----------]\n"), fflush(stdout);
+    #else
+        int driver;
+        cudaDriverGetVersion(&driver);
+
+        printf("[----------]\n"), fflush(stdout);
+        printf("[ GPU INFO ] \tCUDA Driver  version: %d.\n", driver), fflush(stdout);
+        printf("[ GPU INFO ] \tCUDA Runtime version: %d.\n", CUDART_VERSION), fflush(stdout);
+        printf("[----------]\n"), fflush(stdout);
+
+        printf("[----------]\n"), fflush(stdout);
+        printf("[ GPU INFO ] \tGPU module was compiled for the following GPU archs.\n"), fflush(stdout);
+        printf("[      BIN ] \t%s.\n", CUDA_ARCH_BIN), fflush(stdout);
+        printf("[      PTX ] \t%s.\n", CUDA_ARCH_PTX), fflush(stdout);
+        printf("[----------]\n"), fflush(stdout);
+
+        printf("[----------]\n"), fflush(stdout);
+        int deviceCount = cv::gpu::getCudaEnabledDeviceCount();
+        printf("[ GPU INFO ] \tCUDA device count:: %d.\n", deviceCount), fflush(stdout);
+        printf("[----------]\n"), fflush(stdout);
+
+        for (int i = 0; i < deviceCount; ++i)
+        {
+            cv::gpu::DeviceInfo info(i);
+
+            printf("[----------]\n"), fflush(stdout);
+            printf("[ DEVICE   ] \t# %d %s.\n", i, info.name()), fflush(stdout);
+            printf("[          ] \tCompute capability: %d.%d\n", (int)info.majorVersion(), (int)info.minorVersion()), fflush(stdout);
+            printf("[          ] \tMulti Processor Count:  %d\n", info.multiProcessorCount()), fflush(stdout);
+            printf("[          ] \tTotal memory: %d Mb\n", static_cast<int>(static_cast<int>(info.totalMemory() / 1024.0) / 1024.0)), fflush(stdout);
+            printf("[          ] \tFree  memory: %d Mb\n", static_cast<int>(static_cast<int>(info.freeMemory()  / 1024.0) / 1024.0)), fflush(stdout);
+            if (!info.isCompatible())
+                printf("[ GPU INFO ] \tThis device is NOT compatible with current GPU module build\n");
+            printf("[----------]\n"), fflush(stdout);
+        }
+
+    #endif
+    }
+
+    struct KeypointIdxCompare
+    {
+        std::vector<cv::KeyPoint>* keypoints;
+
+        explicit KeypointIdxCompare(std::vector<cv::KeyPoint>* _keypoints) : keypoints(_keypoints) {}
+
+        bool operator ()(size_t i1, size_t i2) const
+        {
+            cv::KeyPoint kp1 = (*keypoints)[i1];
+            cv::KeyPoint kp2 = (*keypoints)[i2];
+            if (kp1.pt.x != kp2.pt.x)
+                return kp1.pt.x < kp2.pt.x;
+            if (kp1.pt.y != kp2.pt.y)
+                return kp1.pt.y < kp2.pt.y;
+            if (kp1.response != kp2.response)
+                return kp1.response < kp2.response;
+            return kp1.octave < kp2.octave;
+        }
+    };
+
+    void sortKeyPoints(std::vector<cv::KeyPoint>& keypoints, cv::InputOutputArray _descriptors)
+    {
+        std::vector<size_t> indexies(keypoints.size());
+        for (size_t i = 0; i < indexies.size(); ++i)
+            indexies[i] = i;
+
+        std::sort(indexies.begin(), indexies.end(), KeypointIdxCompare(&keypoints));
+
+        std::vector<cv::KeyPoint> new_keypoints;
+        cv::Mat new_descriptors;
+
+        new_keypoints.resize(keypoints.size());
+
+        cv::Mat descriptors;
+        if (_descriptors.needed())
+        {
+            descriptors = _descriptors.getMat();
+            new_descriptors.create(descriptors.size(), descriptors.type());
+        }
+
+        for (size_t i = 0; i < indexies.size(); ++i)
+        {
+            size_t new_idx = indexies[i];
+            new_keypoints[i] = keypoints[new_idx];
+            if (!new_descriptors.empty())
+                descriptors.row((int) new_idx).copyTo(new_descriptors.row((int) i));
+        }
+
+        keypoints.swap(new_keypoints);
+        if (_descriptors.needed())
+            new_descriptors.copyTo(_descriptors);
+    }
+}
diff --git a/modules/ts/src/gpu_test.cpp b/modules/ts/src/gpu_test.cpp
new file mode 100644
index 0000000000..4847dbc383
--- /dev/null
+++ b/modules/ts/src/gpu_test.cpp
@@ -0,0 +1,521 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/ts/gpu_test.hpp"
+#include <stdexcept>
+
+using namespace cv;
+using namespace cv::gpu;
+using namespace cvtest;
+using namespace testing;
+using namespace testing::internal;
+
+namespace perf
+{
+    CV_EXPORTS void printCudaInfo();
+}
+
+namespace cvtest
+{
+    //////////////////////////////////////////////////////////////////////
+    // random generators
+
+    int randomInt(int minVal, int maxVal)
+    {
+        RNG& rng = TS::ptr()->get_rng();
+        return rng.uniform(minVal, maxVal);
+    }
+
+    double randomDouble(double minVal, double maxVal)
+    {
+        RNG& rng = TS::ptr()->get_rng();
+        return rng.uniform(minVal, maxVal);
+    }
+
+    Size randomSize(int minVal, int maxVal)
+    {
+        return Size(randomInt(minVal, maxVal), randomInt(minVal, maxVal));
+    }
+
+    Scalar randomScalar(double minVal, double maxVal)
+    {
+        return Scalar(randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal), randomDouble(minVal, maxVal));
+    }
+
+    Mat randomMat(Size size, int type, double minVal, double maxVal)
+    {
+        return randomMat(TS::ptr()->get_rng(), size, type, minVal, maxVal, false);
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // GpuMat create
+
+    GpuMat createMat(Size size, int type, bool useRoi)
+    {
+        Size size0 = size;
+
+        if (useRoi)
+        {
+            size0.width += randomInt(5, 15);
+            size0.height += randomInt(5, 15);
+        }
+
+        GpuMat d_m(size0, type);
+
+        if (size0 != size)
+            d_m = d_m(Rect((size0.width - size.width) / 2, (size0.height - size.height) / 2, size.width, size.height));
+
+        return d_m;
+    }
+
+    GpuMat loadMat(const Mat& m, bool useRoi)
+    {
+        GpuMat d_m = createMat(m.size(), m.type(), useRoi);
+        d_m.upload(m);
+        return d_m;
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // Image load
+
+    Mat readImage(const std::string& fileName, int flags)
+    {
+        return imread(TS::ptr()->get_data_path() + fileName, flags);
+    }
+
+    Mat readImageType(const std::string& fname, int type)
+    {
+        Mat src = readImage(fname, CV_MAT_CN(type) == 1 ? IMREAD_GRAYSCALE : IMREAD_COLOR);
+        if (CV_MAT_CN(type) == 4)
+        {
+            Mat temp;
+            cvtColor(src, temp, COLOR_BGR2BGRA);
+            swap(src, temp);
+        }
+        src.convertTo(src, CV_MAT_DEPTH(type), CV_MAT_DEPTH(type) == CV_32F ? 1.0 / 255.0 : 1.0);
+        return src;
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // Gpu devices
+
+    bool supportFeature(const DeviceInfo& info, FeatureSet feature)
+    {
+        return TargetArchs::builtWith(feature) && info.supports(feature);
+    }
+
+    DeviceManager& DeviceManager::instance()
+    {
+        static DeviceManager obj;
+        return obj;
+    }
+
+    void DeviceManager::load(int i)
+    {
+        devices_.clear();
+        devices_.reserve(1);
+
+        std::ostringstream msg;
+
+        if (i < 0 || i >= getCudaEnabledDeviceCount())
+        {
+            msg << "Incorrect device number - " << i;
+            throw std::runtime_error(msg.str());
+        }
+
+        DeviceInfo info(i);
+
+        if (!info.isCompatible())
+        {
+            msg << "Device " << i << " [" << info.name() << "] is NOT compatible with current GPU module build";
+            throw std::runtime_error(msg.str());
+        }
+
+        devices_.push_back(info);
+    }
+
+    void DeviceManager::loadAll()
+    {
+        int deviceCount = getCudaEnabledDeviceCount();
+
+        devices_.clear();
+        devices_.reserve(deviceCount);
+
+        for (int i = 0; i < deviceCount; ++i)
+        {
+            DeviceInfo info(i);
+            if (info.isCompatible())
+            {
+                devices_.push_back(info);
+            }
+        }
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // Additional assertion
+
+    namespace
+    {
+        template <typename T, typename OutT> std::string printMatValImpl(const Mat& m, Point p)
+        {
+            const int cn = m.channels();
+
+            std::ostringstream ostr;
+            ostr << "(";
+
+            p.x /= cn;
+
+            ostr << static_cast<OutT>(m.at<T>(p.y, p.x * cn));
+            for (int c = 1; c < m.channels(); ++c)
+            {
+                ostr << ", " << static_cast<OutT>(m.at<T>(p.y, p.x * cn + c));
+            }
+            ostr << ")";
+
+            return ostr.str();
+        }
+
+        std::string printMatVal(const Mat& m, Point p)
+        {
+            typedef std::string (*func_t)(const Mat& m, Point p);
+
+            static const func_t funcs[] =
+            {
+                printMatValImpl<uchar, int>, printMatValImpl<schar, int>, printMatValImpl<ushort, int>, printMatValImpl<short, int>,
+                printMatValImpl<int, int>, printMatValImpl<float, float>, printMatValImpl<double, double>
+            };
+
+            return funcs[m.depth()](m, p);
+        }
+    }
+
+    void minMaxLocGold(const Mat& src, double* minVal_, double* maxVal_, Point* minLoc_, Point* maxLoc_, const Mat& mask)
+    {
+        if (src.depth() != CV_8S)
+        {
+            minMaxLoc(src, minVal_, maxVal_, minLoc_, maxLoc_, mask);
+            return;
+        }
+
+        // OpenCV's minMaxLoc doesn't support CV_8S type
+        double minVal = std::numeric_limits<double>::max();
+        Point minLoc(-1, -1);
+
+        double maxVal = -std::numeric_limits<double>::max();
+        Point maxLoc(-1, -1);
+
+        for (int y = 0; y < src.rows; ++y)
+        {
+            const schar* src_row = src.ptr<schar>(y);
+            const uchar* mask_row = mask.empty() ? 0 : mask.ptr<uchar>(y);
+
+            for (int x = 0; x < src.cols; ++x)
+            {
+                if (!mask_row || mask_row[x])
+                {
+                    schar val = src_row[x];
+
+                    if (val < minVal)
+                    {
+                        minVal = val;
+                        minLoc = cv::Point(x, y);
+                    }
+
+                    if (val > maxVal)
+                    {
+                        maxVal = val;
+                        maxLoc = cv::Point(x, y);
+                    }
+                }
+            }
+        }
+
+        if (minVal_) *minVal_ = minVal;
+        if (maxVal_) *maxVal_ = maxVal;
+
+        if (minLoc_) *minLoc_ = minLoc;
+        if (maxLoc_) *maxLoc_ = maxLoc;
+    }
+
+    Mat getMat(InputArray arr)
+    {
+        if (arr.kind() == _InputArray::GPU_MAT)
+        {
+            Mat m;
+            arr.getGpuMat().download(m);
+            return m;
+        }
+
+        return arr.getMat();
+    }
+
+    AssertionResult assertMatNear(const char* expr1, const char* expr2, const char* eps_expr, InputArray m1_, InputArray m2_, double eps)
+    {
+        Mat m1 = getMat(m1_);
+        Mat m2 = getMat(m2_);
+
+        if (m1.size() != m2.size())
+        {
+            return AssertionFailure() << "Matrices \"" << expr1 << "\" and \"" << expr2 << "\" have different sizes : \""
+                                      << expr1 << "\" [" << PrintToString(m1.size()) << "] vs \""
+                                      << expr2 << "\" [" << PrintToString(m2.size()) << "]";
+        }
+
+        if (m1.type() != m2.type())
+        {
+            return AssertionFailure() << "Matrices \"" << expr1 << "\" and \"" << expr2 << "\" have different types : \""
+                                      << expr1 << "\" [" << PrintToString(MatType(m1.type())) << "] vs \""
+                                      << expr2 << "\" [" << PrintToString(MatType(m2.type())) << "]";
+        }
+
+        Mat diff;
+        absdiff(m1.reshape(1), m2.reshape(1), diff);
+
+        double maxVal = 0.0;
+        Point maxLoc;
+        minMaxLocGold(diff, 0, &maxVal, 0, &maxLoc);
+
+        if (maxVal > eps)
+        {
+            return AssertionFailure() << "The max difference between matrices \"" << expr1 << "\" and \"" << expr2
+                                      << "\" is " << maxVal << " at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ")"
+                                      << ", which exceeds \"" << eps_expr << "\", where \""
+                                      << expr1 << "\" at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ") evaluates to " << printMatVal(m1, maxLoc) << ", \""
+                                      << expr2 << "\" at (" << maxLoc.y << ", " << maxLoc.x / m1.channels() << ") evaluates to " << printMatVal(m2, maxLoc) << ", \""
+                                      << eps_expr << "\" evaluates to " << eps;
+        }
+
+        return AssertionSuccess();
+    }
+
+    double checkSimilarity(InputArray m1, InputArray m2)
+    {
+        Mat diff;
+        matchTemplate(getMat(m1), getMat(m2), diff, TM_CCORR_NORMED);
+        return std::abs(diff.at<float>(0, 0) - 1.f);
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // Helper structs for value-parameterized tests
+
+    vector<MatType> types(int depth_start, int depth_end, int cn_start, int cn_end)
+    {
+        vector<MatType> v;
+
+        v.reserve((depth_end - depth_start + 1) * (cn_end - cn_start + 1));
+
+        for (int depth = depth_start; depth <= depth_end; ++depth)
+        {
+            for (int cn = cn_start; cn <= cn_end; ++cn)
+            {
+                v.push_back(MatType(CV_MAKE_TYPE(depth, cn)));
+            }
+        }
+
+        return v;
+    }
+
+    const vector<MatType>& all_types()
+    {
+        static vector<MatType> v = types(CV_8U, CV_64F, 1, 4);
+
+        return v;
+    }
+
+    void PrintTo(const UseRoi& useRoi, std::ostream* os)
+    {
+        if (useRoi)
+            (*os) << "sub matrix";
+        else
+            (*os) << "whole matrix";
+    }
+
+    void PrintTo(const Inverse& inverse, std::ostream* os)
+    {
+        if (inverse)
+            (*os) << "inverse";
+        else
+            (*os) << "direct";
+    }
+
+    //////////////////////////////////////////////////////////////////////
+    // Other
+
+    void dumpImage(const std::string& fileName, const Mat& image)
+    {
+        imwrite(TS::ptr()->get_data_path() + fileName, image);
+    }
+
+    void showDiff(InputArray gold_, InputArray actual_, double eps)
+    {
+        Mat gold = getMat(gold_);
+        Mat actual = getMat(actual_);
+
+        Mat diff;
+        absdiff(gold, actual, diff);
+        threshold(diff, diff, eps, 255.0, cv::THRESH_BINARY);
+
+        namedWindow("gold", WINDOW_NORMAL);
+        namedWindow("actual", WINDOW_NORMAL);
+        namedWindow("diff", WINDOW_NORMAL);
+
+        imshow("gold", gold);
+        imshow("actual", actual);
+        imshow("diff", diff);
+
+        waitKey();
+    }
+
+    namespace
+    {
+        bool keyPointsEquals(const cv::KeyPoint& p1, const cv::KeyPoint& p2)
+        {
+            const double maxPtDif = 1.0;
+            const double maxSizeDif = 1.0;
+            const double maxAngleDif = 2.0;
+            const double maxResponseDif = 0.1;
+
+            double dist = cv::norm(p1.pt - p2.pt);
+
+            if (dist < maxPtDif &&
+                fabs(p1.size - p2.size) < maxSizeDif &&
+                abs(p1.angle - p2.angle) < maxAngleDif &&
+                abs(p1.response - p2.response) < maxResponseDif &&
+                p1.octave == p2.octave &&
+                p1.class_id == p2.class_id)
+            {
+                return true;
+            }
+
+            return false;
+        }
+
+        struct KeyPointLess : std::binary_function<cv::KeyPoint, cv::KeyPoint, bool>
+        {
+            bool operator()(const cv::KeyPoint& kp1, const cv::KeyPoint& kp2) const
+            {
+                return kp1.pt.y < kp2.pt.y || (kp1.pt.y == kp2.pt.y && kp1.pt.x < kp2.pt.x);
+            }
+        };
+    }
+
+    testing::AssertionResult assertKeyPointsEquals(const char* gold_expr, const char* actual_expr, std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual)
+    {
+        if (gold.size() != actual.size())
+        {
+            return testing::AssertionFailure() << "KeyPoints size mistmach\n"
+                                               << "\"" << gold_expr << "\" : " << gold.size() << "\n"
+                                               << "\"" << actual_expr << "\" : " << actual.size();
+        }
+
+        std::sort(actual.begin(), actual.end(), KeyPointLess());
+        std::sort(gold.begin(), gold.end(), KeyPointLess());
+
+        for (size_t i = 0; i < gold.size(); ++i)
+        {
+            const cv::KeyPoint& p1 = gold[i];
+            const cv::KeyPoint& p2 = actual[i];
+
+            if (!keyPointsEquals(p1, p2))
+            {
+                return testing::AssertionFailure() << "KeyPoints differ at " << i << "\n"
+                                                   << "\"" << gold_expr << "\" vs \"" << actual_expr << "\" : \n"
+                                                   << "pt : " << testing::PrintToString(p1.pt) << " vs " << testing::PrintToString(p2.pt) << "\n"
+                                                   << "size : " << p1.size << " vs " << p2.size << "\n"
+                                                   << "angle : " << p1.angle << " vs " << p2.angle << "\n"
+                                                   << "response : " << p1.response << " vs " << p2.response << "\n"
+                                                   << "octave : " << p1.octave << " vs " << p2.octave << "\n"
+                                                   << "class_id : " << p1.class_id << " vs " << p2.class_id;
+            }
+        }
+
+        return ::testing::AssertionSuccess();
+    }
+
+    int getMatchedPointsCount(std::vector<cv::KeyPoint>& gold, std::vector<cv::KeyPoint>& actual)
+    {
+        std::sort(actual.begin(), actual.end(), KeyPointLess());
+        std::sort(gold.begin(), gold.end(), KeyPointLess());
+
+        int validCount = 0;
+
+        for (size_t i = 0; i < gold.size(); ++i)
+        {
+            const cv::KeyPoint& p1 = gold[i];
+            const cv::KeyPoint& p2 = actual[i];
+
+            if (keyPointsEquals(p1, p2))
+                ++validCount;
+        }
+
+        return validCount;
+    }
+
+    int getMatchedPointsCount(const std::vector<cv::KeyPoint>& keypoints1, const std::vector<cv::KeyPoint>& keypoints2, const std::vector<cv::DMatch>& matches)
+    {
+        int validCount = 0;
+
+        for (size_t i = 0; i < matches.size(); ++i)
+        {
+            const cv::DMatch& m = matches[i];
+
+            const cv::KeyPoint& p1 = keypoints1[m.queryIdx];
+            const cv::KeyPoint& p2 = keypoints2[m.trainIdx];
+
+            if (keyPointsEquals(p1, p2))
+                ++validCount;
+        }
+
+        return validCount;
+    }
+
+    void printCudaInfo()
+    {
+        perf::printCudaInfo();
+    }
+}
+
+
+void cv::gpu::PrintTo(const DeviceInfo& info, std::ostream* os)
+{
+    (*os) << info.name();
+}
diff --git a/modules/ts/src/precomp.cpp b/modules/ts/src/precomp.cpp
deleted file mode 100644
index c149df18f9..0000000000
--- a/modules/ts/src/precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "precomp.hpp"
diff --git a/modules/ts/src/precomp.hpp b/modules/ts/src/precomp.hpp
index 0bc076376c..fbb13ec4c8 100644
--- a/modules/ts/src/precomp.hpp
+++ b/modules/ts/src/precomp.hpp
@@ -1,5 +1,7 @@
-#include "opencv2/core/core_c.h"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
 #include "opencv2/ts.hpp"
+#include "cvconfig.h"
 
 #ifdef GTEST_LINKED_AS_SHARED_LIBRARY
 #error ts module should not have GTEST_LINKED_AS_SHARED_LIBRARY defined
diff --git a/modules/ts/src/ts.cpp b/modules/ts/src/ts.cpp
index 7ce37752db..90ea1f9819 100644
--- a/modules/ts/src/ts.cpp
+++ b/modules/ts/src/ts.cpp
@@ -40,6 +40,8 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/core/core_c.h"
+
 #include <ctype.h>
 #include <stdarg.h>
 #include <stdlib.h>
@@ -321,9 +323,7 @@ int BaseTest::update_progress( int progress, int test_case_idx, int count, doubl
 
 BadArgTest::BadArgTest()
 {
-    progress        = -1;
     test_case_idx   = -1;
-    freq            = cv::getTickFrequency();
     // oldErrorCbk     = 0;
     // oldErrorCbkData = 0;
 }
@@ -334,20 +334,6 @@ BadArgTest::~BadArgTest(void)
 
 int BadArgTest::run_test_case( int expected_code, const string& _descr )
 {
-    double new_t = (double)cv::getTickCount(), dt;
-    if( test_case_idx < 0 )
-    {
-        test_case_idx = 0;
-        progress      = 0;
-        dt            = 0;
-    }
-    else
-    {
-        dt = (new_t - t)/(freq*1000);
-        t  = new_t;
-    }
-    progress = update_progress(progress, test_case_idx, 0, dt);
-
     int errcount = 0;
     bool thrown = false;
     const char* descr = _descr.c_str() ? _descr.c_str() : "";
diff --git a/modules/ts/src/ts_arrtest.cpp b/modules/ts/src/ts_arrtest.cpp
index ec3f18330d..a3a058c518 100644
--- a/modules/ts/src/ts_arrtest.cpp
+++ b/modules/ts/src/ts_arrtest.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/core/core_c.h"
 
 namespace cvtest
 {
@@ -158,7 +159,7 @@ int ArrayTest::prepare_test_case( int test_case_idx )
             unsigned t = randInt(rng);
             bool create_mask = true, use_roi = false;
             CvSize size = sizes[i][j], whole_size = size;
-            CvRect roi = {0,0,0,0};
+            CvRect roi;
 
             is_image = !cvmat_allowed ? true : iplimage_allowed ? (t & 1) != 0 : false;
             create_mask = (t & 6) == 0; // ~ each of 3 tests will use mask
diff --git a/modules/ts/src/ts_func.cpp b/modules/ts/src/ts_func.cpp
index 62e16fee4d..35482f3124 100644
--- a/modules/ts/src/ts_func.cpp
+++ b/modules/ts/src/ts_func.cpp
@@ -2,6 +2,10 @@
 #include <float.h>
 #include <limits.h>
 
+#ifdef HAVE_TEGRA_OPTIMIZATION
+#include "tegra.hpp"
+#endif
+
 using namespace cv;
 
 namespace cvtest
@@ -71,7 +75,7 @@ int randomType(RNG& rng, int typeMask, int minChannels, int maxChannels)
 {
     int channels = rng.uniform(minChannels, maxChannels+1);
     int depth = 0;
-    CV_Assert((typeMask & DEPTH_MASK_ALL) != 0);
+    CV_Assert((typeMask & _OutputArray::DEPTH_MASK_ALL) != 0);
     for(;;)
     {
         depth = rng.uniform(CV_8U, CV_64F+1);
@@ -272,10 +276,13 @@ convertTo(const _Tp* src, void* dst, int dtype, size_t total, double alpha, doub
     }
 }
 
-void convert(const Mat& src, Mat& dst, int dtype, double alpha, double beta)
+void convert(const Mat& src, cv::OutputArray _dst, int dtype, double alpha, double beta)
 {
+    if (dtype < 0) dtype = _dst.depth();
+
     dtype = CV_MAKETYPE(CV_MAT_DEPTH(dtype), src.channels());
-    dst.create(src.dims, &src.size[0], dtype);
+    _dst.create(src.dims, &src.size[0], dtype);
+    Mat dst = _dst.getMat();
     if( alpha == 0 )
     {
         set( dst, Scalar::all(beta) );
@@ -642,7 +649,7 @@ void erode(const Mat& _src, Mat& dst, const Mat& _kernel, Point anchor,
     }
     if( anchor == Point(-1,-1) )
         anchor = Point(kernel.cols/2, kernel.rows/2);
-    if( borderType == IPL_BORDER_CONSTANT )
+    if( borderType == BORDER_CONSTANT )
         borderValue = getMaxVal(src.depth());
     copyMakeBorder(_src, src, anchor.y, kernel.rows - anchor.y - 1,
                    anchor.x, kernel.cols - anchor.x - 1,
@@ -699,7 +706,7 @@ void dilate(const Mat& _src, Mat& dst, const Mat& _kernel, Point anchor,
     }
     if( anchor == Point(-1,-1) )
         anchor = Point(kernel.cols/2, kernel.rows/2);
-    if( borderType == IPL_BORDER_CONSTANT )
+    if( borderType == BORDER_CONSTANT )
         borderValue = getMinVal(src.depth());
     copyMakeBorder(_src, src, anchor.y, kernel.rows - anchor.y - 1,
                    anchor.x, kernel.cols - anchor.x - 1,
@@ -775,7 +782,7 @@ void filter2D(const Mat& _src, Mat& dst, int ddepth, const Mat& kernel,
     CV_Assert( kernel.type() == CV_32F || kernel.type() == CV_64F );
     if( anchor == Point(-1,-1) )
         anchor = Point(kernel.cols/2, kernel.rows/2);
-    if( borderType == IPL_BORDER_CONSTANT )
+    if( borderType == BORDER_CONSTANT )
         borderValue = getMinVal(src.depth());
     copyMakeBorder(_src, src, anchor.y, kernel.rows - anchor.y - 1,
                    anchor.x, kernel.cols - anchor.x - 1,
@@ -827,11 +834,11 @@ static int borderInterpolate( int p, int len, int borderType )
 {
     if( (unsigned)p < (unsigned)len )
         ;
-    else if( borderType == IPL_BORDER_REPLICATE )
+    else if( borderType == BORDER_REPLICATE )
         p = p < 0 ? 0 : len - 1;
-    else if( borderType == IPL_BORDER_REFLECT || borderType == IPL_BORDER_REFLECT_101 )
+    else if( borderType == BORDER_REFLECT || borderType == BORDER_REFLECT_101 )
     {
-        int delta = borderType == IPL_BORDER_REFLECT_101;
+        int delta = borderType == BORDER_REFLECT_101;
         if( len == 1 )
             return 0;
         do
@@ -843,17 +850,17 @@ static int borderInterpolate( int p, int len, int borderType )
         }
         while( (unsigned)p >= (unsigned)len );
     }
-    else if( borderType == IPL_BORDER_WRAP )
+    else if( borderType == BORDER_WRAP )
     {
         if( p < 0 )
             p -= ((p-len+1)/len)*len;
         if( p >= len )
             p %= len;
     }
-    else if( borderType == IPL_BORDER_CONSTANT )
+    else if( borderType == BORDER_CONSTANT )
         p = -1;
     else
-        CV_Error( CV_StsBadArg, "Unknown/unsupported border type" );
+        CV_Error( Error::StsBadArg, "Unknown/unsupported border type" );
     return p;
 }
 
@@ -865,7 +872,7 @@ void copyMakeBorder(const Mat& src, Mat& dst, int top, int bottom, int left, int
     int i, j, k, esz = (int)src.elemSize();
     int width = src.cols*esz, width1 = dst.cols*esz;
 
-    if( borderType == IPL_BORDER_CONSTANT )
+    if( borderType == BORDER_CONSTANT )
     {
         vector<uchar> valvec((src.cols + left + right)*esz);
         uchar* val = &valvec[0];
@@ -1301,7 +1308,7 @@ double norm(const Mat& src, int normType, const Mat& mask)
             result = norm_((const double*)sptr, total, cn, normType, result, mptr);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         };
     }
     if( normType0 == NORM_L2 )
@@ -1379,7 +1386,7 @@ double norm(const Mat& src1, const Mat& src2, int normType, const Mat& mask)
             result = norm_((const double*)sptr1, (const double*)sptr2, total, cn, normType, result, mptr);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         };
     }
     if( normType0 == NORM_L2 )
@@ -1438,7 +1445,7 @@ double crossCorr(const Mat& src1, const Mat& src2)
             result += crossCorr_((const double*)sptr1, (const double*)sptr2, total);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         };
     }
     return result;
@@ -1571,7 +1578,7 @@ compare_(const _Tp* src1, const _Tp* src2, uchar* dst, size_t total, int cmpop)
             dst[i] = src1[i] > src2[i] ? 255 : 0;
         break;
     default:
-        CV_Error(CV_StsBadArg, "Unknown comparison operation");
+        CV_Error(Error::StsBadArg, "Unknown comparison operation");
     }
 }
 
@@ -1607,7 +1614,7 @@ compareS_(const _Tp* src1, _WTp value, uchar* dst, size_t total, int cmpop)
             dst[i] = src1[i] > value ? 255 : 0;
         break;
     default:
-        CV_Error(CV_StsBadArg, "Unknown comparison operation");
+        CV_Error(Error::StsBadArg, "Unknown comparison operation");
     }
 }
 
@@ -1654,7 +1661,7 @@ void compare(const Mat& src1, const Mat& src2, Mat& dst, int cmpop)
             compare_((const double*)sptr1, (const double*)sptr2, dptr, total, cmpop);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -1701,7 +1708,7 @@ void compare(const Mat& src, double value, Mat& dst, int cmpop)
             compareS_((const double*)sptr, value, dptr, total, cmpop);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -1833,7 +1840,7 @@ bool cmpUlps(const Mat& src1, const Mat& src2, int imaxDiff, double* _realmaxdif
             realmaxdiff = cmpUlpsFlt_((const int64*)sptr1, (const int64*)sptr2, total, imaxDiff, startidx, idx);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
 
         if(_realmaxdiff)
@@ -1922,7 +1929,7 @@ int check( const Mat& a, double fmin, double fmax, vector<int>* _idx )
                 checkFlt_((const double*)aptr, total, fmin, fmax, startidx, idx);
                 break;
             default:
-                CV_Error(CV_StsUnsupportedFormat, "");
+                CV_Error(Error::StsUnsupportedFormat, "");
         }
 
         if( idx != 0 )
@@ -2341,7 +2348,7 @@ void transform( const Mat& src, Mat& dst, const Mat& transmat, const Mat& _shift
             transform_((const double*)sptr, (double*)dptr, total, scn, dcn, mat);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -2398,7 +2405,7 @@ static void minmax(const Mat& src1, const Mat& src2, Mat& dst, char op)
             minmax_((const double*)sptr1, (const double*)sptr2, (double*)dptr, total, op);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -2466,7 +2473,7 @@ static void minmax(const Mat& src1, double val, Mat& dst, char op)
             minmax_((const double*)sptr1, saturate_cast<double>(val), (double*)dptr, total, op);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -2538,7 +2545,7 @@ static void muldiv(const Mat& src1, const Mat& src2, Mat& dst, double scale, cha
             muldiv_((const double*)sptr1, (const double*)sptr2, (double*)dptr, total, scale, op);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 }
@@ -2623,7 +2630,7 @@ Scalar mean(const Mat& src, const Mat& mask)
             mean_((const double*)sptr, mptr, total, cn, sum, nz);
             break;
         default:
-            CV_Error(CV_StsUnsupportedFormat, "");
+            CV_Error(Error::StsUnsupportedFormat, "");
         }
     }
 
@@ -2860,7 +2867,7 @@ static void writeElems(std::ostream& out, const void* data, int nelems, int dept
         out.precision(pp);
     }
     else
-        CV_Error(CV_StsUnsupportedFormat, "");
+        CV_Error(Error::StsUnsupportedFormat, "");
 }
 
 
@@ -2934,16 +2941,77 @@ MatComparator::operator()(const char* expr1, const char* expr2,
     << "'" << expr2 << "': " << MatPart(m2part, border > 0 ? &loc : 0) << ".\n";
 }
 
+void printVersionInfo(bool useStdOut)
+{
+    ::testing::Test::RecordProperty("cv_version", CV_VERSION);
+    if(useStdOut) std::cout << "OpenCV version: " << CV_VERSION << std::endl;
+
+    std::string buildInfo( cv::getBuildInformation() );
+
+    size_t pos1 = buildInfo.find("Version control");
+    size_t pos2 = buildInfo.find('\n', pos1);
+    if(pos1 != std::string::npos && pos2 != std::string::npos)
+    {
+        size_t value_start = buildInfo.rfind(' ', pos2) + 1;
+        std::string ver( buildInfo.substr(value_start, pos2 - value_start) );
+        ::testing::Test::RecordProperty("cv_vcs_version", ver);
+        if (useStdOut) std::cout << "OpenCV VCS version: " << ver << std::endl;
+    }
+
+    pos1 = buildInfo.find("inner version");
+    pos2 = buildInfo.find('\n', pos1);
+    if(pos1 != std::string::npos && pos2 != std::string::npos)
+    {
+        size_t value_start = buildInfo.rfind(' ', pos2) + 1;
+        std::string ver( buildInfo.substr(value_start, pos2 - value_start) );
+        ::testing::Test::RecordProperty("cv_inner_vcs_version", ver);
+        if(useStdOut) std::cout << "Inner VCS version: " << ver << std::endl;
+    }
+
+    const char* parallel_framework = currentParallelFramework();
+
+    if (parallel_framework) {
+        ::testing::Test::RecordProperty("cv_parallel_framework", parallel_framework);
+        if (useStdOut) std::cout << "Parallel framework: " << parallel_framework << std::endl;
+    }
+
+    std::string cpu_features;
+
+#if CV_SSE
+    if (checkHardwareSupport(CV_CPU_SSE)) cpu_features += " sse";
+#endif
+#if CV_SSE2
+    if (checkHardwareSupport(CV_CPU_SSE2)) cpu_features += " sse2";
+#endif
+#if CV_SSE3
+    if (checkHardwareSupport(CV_CPU_SSE3)) cpu_features += " sse3";
+#endif
+#if CV_SSSE3
+    if (checkHardwareSupport(CV_CPU_SSSE3)) cpu_features += " ssse3";
+#endif
+#if CV_SSE4_1
+    if (checkHardwareSupport(CV_CPU_SSE4_1)) cpu_features += " sse4.1";
+#endif
+#if CV_SSE4_2
+    if (checkHardwareSupport(CV_CPU_SSE4_2)) cpu_features += " sse4.2";
+#endif
+#if CV_AVX
+    if (checkHardwareSupport(CV_CPU_AVX)) cpu_features += " avx";
+#endif
+#if CV_NEON
+    cpu_features += " neon"; // NEON is currently not checked at runtime
+#endif
+
+    cpu_features.erase(0, 1); // erase initial space
+
+    ::testing::Test::RecordProperty("cv_cpu_features", cpu_features);
+    if (useStdOut) std::cout << "CPU features: " << cpu_features << std::endl;
+
+#ifdef HAVE_TEGRA_OPTIMIZATION
+    const char * tegra_optimization = tegra::isDeviceSupported() ? "enabled" : "disabled";
+    ::testing::Test::RecordProperty("cv_tegra_optimization", tegra_optimization);
+    if (useStdOut) std::cout << "Tegra optimization: " << tegra_optimization << std::endl;
+#endif
 }
 
-void cvTsConvert( const CvMat* src, CvMat* dst )
-{
-    Mat _src = cvarrToMat(src), _dst = cvarrToMat(dst);
-    cvtest::convert(_src, _dst, _dst.depth());
-}
-
-void cvTsZero( CvMat* dst, const CvMat* mask )
-{
-    Mat _dst = cvarrToMat(dst), _mask = mask ? cvarrToMat(mask) : Mat();
-    cvtest::set(_dst, Scalar::all(0), _mask);
 }
diff --git a/modules/ts/src/ts_gtest.cpp b/modules/ts/src/ts_gtest.cpp
index a718fc036b..48870913c3 100644
--- a/modules/ts/src/ts_gtest.cpp
+++ b/modules/ts/src/ts_gtest.cpp
@@ -319,6 +319,8 @@ class GTEST_API_ SingleFailureChecker {
 #include <wctype.h>
 
 #include <algorithm>
+#include <iomanip>
+#include <limits>
 #include <ostream>  // NOLINT
 #include <sstream>
 #include <vector>
@@ -464,6 +466,11 @@ class GTEST_API_ SingleFailureChecker {
 #include <vector>
 
 
+#if GTEST_CAN_STREAM_RESULTS_
+# include <arpa/inet.h>  // NOLINT
+# include <netdb.h>  // NOLINT
+#endif
+
 #if GTEST_OS_WINDOWS
 # include <windows.h>  // NOLINT
 #endif  // GTEST_OS_WINDOWS
@@ -490,6 +497,7 @@ const char kBreakOnFailureFlag[] = "break_on_failure";
 const char kCatchExceptionsFlag[] = "catch_exceptions";
 const char kColorFlag[] = "color";
 const char kFilterFlag[] = "filter";
+const char kParamFilterFlag[] = "param_filter";
 const char kListTestsFlag[] = "list_tests";
 const char kOutputFlag[] = "output";
 const char kPrintTimeFlag[] = "print_time";
@@ -568,6 +576,7 @@ class GTestFlagSaver {
     death_test_style_ = GTEST_FLAG(death_test_style);
     death_test_use_fork_ = GTEST_FLAG(death_test_use_fork);
     filter_ = GTEST_FLAG(filter);
+    param_filter_ = GTEST_FLAG(param_filter);
     internal_run_death_test_ = GTEST_FLAG(internal_run_death_test);
     list_tests_ = GTEST_FLAG(list_tests);
     output_ = GTEST_FLAG(output);
@@ -589,6 +598,7 @@ class GTestFlagSaver {
     GTEST_FLAG(death_test_style) = death_test_style_;
     GTEST_FLAG(death_test_use_fork) = death_test_use_fork_;
     GTEST_FLAG(filter) = filter_;
+    GTEST_FLAG(param_filter) = param_filter_;
     GTEST_FLAG(internal_run_death_test) = internal_run_death_test_;
     GTEST_FLAG(list_tests) = list_tests_;
     GTEST_FLAG(output) = output_;
@@ -610,11 +620,11 @@ class GTestFlagSaver {
   std::string death_test_style_;
   bool death_test_use_fork_;
   std::string filter_;
+  std::string param_filter_;
   std::string internal_run_death_test_;
   bool list_tests_;
   std::string output_;
   bool print_time_;
-  bool pretty_;
   internal::Int32 random_seed_;
   internal::Int32 repeat_;
   bool shuffle_;
@@ -626,12 +636,10 @@ class GTestFlagSaver {
 // Converts a Unicode code point to a narrow string in UTF-8 encoding.
 // code_point parameter is of type UInt32 because wchar_t may not be
 // wide enough to contain a code point.
-// The output buffer str must containt at least 32 characters.
-// The function returns the address of the output buffer.
 // If the code_point is not a valid Unicode code point
-// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be output
-// as '(Invalid Unicode 0xXXXXXXXX)'.
-GTEST_API_ char* CodePointToUtf8(UInt32 code_point, char* str);
+// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be converted
+// to "(Invalid Unicode 0xXXXXXXXX)".
+GTEST_API_ std::string CodePointToUtf8(UInt32 code_point);
 
 // Converts a wide string to a narrow string in UTF-8 encoding.
 // The wide string is assumed to have the following encoding:
@@ -750,8 +758,7 @@ class TestPropertyKeyIs {
   // Constructor.
   //
   // TestPropertyKeyIs has NO default constructor.
-  explicit TestPropertyKeyIs(const char* key)
-      : key_(key) {}
+  explicit TestPropertyKeyIs(const std::string& key) : key_(key) {}
 
   // Returns true iff the test name of test property matches on key_.
   bool operator()(const TestProperty& test_property) const {
@@ -1114,6 +1121,12 @@ class GTEST_API_ UnitTestImpl {
     ad_hoc_test_result_.Clear();
   }
 
+  // Adds a TestProperty to the current TestResult object when invoked in a
+  // context of a test or a test case, or to the global property set. If the
+  // result already contains a property with the same key, the value will be
+  // updated.
+  void RecordProperty(const TestProperty& test_property);
+
   enum ReactionToSharding {
     HONOR_SHARDING_PROTOCOL,
     IGNORE_SHARDING_PROTOCOL
@@ -1441,8 +1454,9 @@ bool ParseNaturalNumber(const ::std::string& str, Integer* number) {
 class TestResultAccessor {
  public:
   static void RecordProperty(TestResult* test_result,
+                             const std::string& xml_element,
                              const TestProperty& property) {
-    test_result->RecordProperty(property);
+    test_result->RecordProperty(xml_element, property);
   }
 
   static void ClearTestPartResults(TestResult* test_result) {
@@ -1455,6 +1469,154 @@ class TestResultAccessor {
   }
 };
 
+#if GTEST_CAN_STREAM_RESULTS_
+
+// Streams test results to the given port on the given host machine.
+class StreamingListener : public EmptyTestEventListener {
+ public:
+  // Abstract base class for writing strings to a socket.
+  class AbstractSocketWriter {
+   public:
+    virtual ~AbstractSocketWriter() {}
+
+    // Sends a string to the socket.
+    virtual void Send(const string& message) = 0;
+
+    // Closes the socket.
+    virtual void CloseConnection() {}
+
+    // Sends a string and a newline to the socket.
+    void SendLn(const string& message) {
+      Send(message + "\n");
+    }
+  };
+
+  // Concrete class for actually writing strings to a socket.
+  class SocketWriter : public AbstractSocketWriter {
+   public:
+    SocketWriter(const string& host, const string& port)
+        : sockfd_(-1), host_name_(host), port_num_(port) {
+      MakeConnection();
+    }
+
+    virtual ~SocketWriter() {
+      if (sockfd_ != -1)
+        CloseConnection();
+    }
+
+    // Sends a string to the socket.
+    virtual void Send(const string& message) {
+      GTEST_CHECK_(sockfd_ != -1)
+          << "Send() can be called only when there is a connection.";
+
+      const int len = static_cast<int>(message.length());
+      if (write(sockfd_, message.c_str(), len) != len) {
+        GTEST_LOG_(WARNING)
+            << "stream_result_to: failed to stream to "
+            << host_name_ << ":" << port_num_;
+      }
+    }
+
+   private:
+    // Creates a client socket and connects to the server.
+    void MakeConnection();
+
+    // Closes the socket.
+    void CloseConnection() {
+      GTEST_CHECK_(sockfd_ != -1)
+          << "CloseConnection() can be called only when there is a connection.";
+
+      close(sockfd_);
+      sockfd_ = -1;
+    }
+
+    int sockfd_;  // socket file descriptor
+    const string host_name_;
+    const string port_num_;
+
+    GTEST_DISALLOW_COPY_AND_ASSIGN_(SocketWriter);
+  };  // class SocketWriter
+
+  // Escapes '=', '&', '%', and '\n' characters in str as "%xx".
+  static string UrlEncode(const char* str);
+
+  StreamingListener(const string& host, const string& port)
+      : socket_writer_(new SocketWriter(host, port)) { Start(); }
+
+  explicit StreamingListener(AbstractSocketWriter* socket_writer)
+      : socket_writer_(socket_writer) { Start(); }
+
+  void OnTestProgramStart(const UnitTest& /* unit_test */) {
+    SendLn("event=TestProgramStart");
+  }
+
+  void OnTestProgramEnd(const UnitTest& unit_test) {
+    // Note that Google Test current only report elapsed time for each
+    // test iteration, not for the entire test program.
+    SendLn("event=TestProgramEnd&passed=" + FormatBool(unit_test.Passed()));
+
+    // Notify the streaming server to stop.
+    socket_writer_->CloseConnection();
+  }
+
+  void OnTestIterationStart(const UnitTest& /* unit_test */, int iteration) {
+    SendLn("event=TestIterationStart&iteration=" +
+           StreamableToString(iteration));
+  }
+
+  void OnTestIterationEnd(const UnitTest& unit_test, int /* iteration */) {
+    SendLn("event=TestIterationEnd&passed=" +
+           FormatBool(unit_test.Passed()) + "&elapsed_time=" +
+           StreamableToString(unit_test.elapsed_time()) + "ms");
+  }
+
+  void OnTestCaseStart(const TestCase& test_case) {
+    SendLn(std::string("event=TestCaseStart&name=") + test_case.name());
+  }
+
+  void OnTestCaseEnd(const TestCase& test_case) {
+    SendLn("event=TestCaseEnd&passed=" + FormatBool(test_case.Passed())
+           + "&elapsed_time=" + StreamableToString(test_case.elapsed_time())
+           + "ms");
+  }
+
+  void OnTestStart(const TestInfo& test_info) {
+    SendLn(std::string("event=TestStart&name=") + test_info.name());
+  }
+
+  void OnTestEnd(const TestInfo& test_info) {
+    SendLn("event=TestEnd&passed=" +
+           FormatBool((test_info.result())->Passed()) +
+           "&elapsed_time=" +
+           StreamableToString((test_info.result())->elapsed_time()) + "ms");
+  }
+
+  void OnTestPartResult(const TestPartResult& test_part_result) {
+    const char* file_name = test_part_result.file_name();
+    if (file_name == NULL)
+      file_name = "";
+    SendLn("event=TestPartResult&file=" + UrlEncode(file_name) +
+           "&line=" + StreamableToString(test_part_result.line_number()) +
+           "&message=" + UrlEncode(test_part_result.message()));
+  }
+
+ private:
+  // Sends the given message and a newline to the socket.
+  void SendLn(const string& message) { socket_writer_->SendLn(message); }
+
+  // Called at the start of streaming to notify the receiver what
+  // protocol we are using.
+  void Start() { SendLn("gtest_streaming_protocol_version=1.0"); }
+
+  string FormatBool(bool value) { return value ? "1" : "0"; }
+
+  const scoped_ptr<AbstractSocketWriter> socket_writer_;
+
+  GTEST_DISALLOW_COPY_AND_ASSIGN_(StreamingListener);
+};  // class StreamingListener
+
+#endif  // GTEST_CAN_STREAM_RESULTS_
+
 }  // namespace internal
 }  // namespace testing
 
@@ -1530,7 +1692,7 @@ GTEST_DEFINE_string_(
     "Whether to use colors in the output.  Valid values: yes, no, "
     "and auto.  'auto' means to use colors if the output is "
     "being sent to a terminal and the TERM environment variable "
-    "is set to xterm, xterm-color, xterm-256color, linux or cygwin.");
+    "is set to a terminal type that supports colors.");
 
 GTEST_DEFINE_string_(
     filter,
@@ -1541,6 +1703,12 @@ GTEST_DEFINE_string_(
     "exclude).  A test is run if it matches one of the positive "
     "patterns and does not match any of the negative patterns.");
 
+GTEST_DEFINE_string_(
+    param_filter,
+    internal::StringFromGTestEnv("param_filter", kUniversalFilter),
+    "Same syntax and semantics as for param, but these patterns "
+    "have to match the test's parameters.");
+
 GTEST_DEFINE_bool_(list_tests, false,
                    "List all tests without running them.");
 
@@ -1731,8 +1899,8 @@ std::string UnitTestOptions::GetAbsolutePathToOutputFile() {
   const char* const colon = strchr(gtest_output_flag, ':');
   if (colon == NULL)
     return internal::FilePath::ConcatPaths(
-               internal::FilePath(
-                   UnitTest::GetInstance()->original_working_dir()),
+        internal::FilePath(
+            UnitTest::GetInstance()->original_working_dir()),
         internal::FilePath(kDefaultOutputFile)).string();
 
   internal::FilePath output_name(colon + 1);
@@ -2214,6 +2382,26 @@ static void StreamWideCharsToMessage(const wchar_t* wstr, size_t length,
 
 }  // namespace internal
 
+// Constructs an empty Message.
+// We allocate the stringstream separately because otherwise each use of
+// ASSERT/EXPECT in a procedure adds over 200 bytes to the procedure's
+// stack frame leading to huge stack frames in some cases; gcc does not reuse
+// the stack space.
+Message::Message() : ss_(new ::std::stringstream) {
+  // By default, we want there to be enough precision when printing
+  // a double to a Message.
+  *ss_ << std::setprecision(std::numeric_limits<double>::digits10 + 2);
+}
+
+// These two overloads allow streaming a wide C string to a Message
+// using the UTF-8 encoding.
+Message& Message::operator <<(const wchar_t* wide_c_str) {
+  return *this << internal::String::ShowWideCString(wide_c_str);
+}
+Message& Message::operator <<(wchar_t* wide_c_str) {
+  return *this << internal::String::ShowWideCString(wide_c_str);
+}
+
 #if GTEST_HAS_STD_WSTRING
 // Converts the given wide string to a narrow string using the UTF-8
 // encoding, and streams the result to this Message object.
@@ -2232,6 +2420,12 @@ Message& Message::operator <<(const ::wstring& wstr) {
 }
 #endif  // GTEST_HAS_GLOBAL_WSTRING
 
+// Gets the text streamed to this object so far as an std::string.
+// Each '\0' character in the buffer is replaced with "\\0".
+std::string Message::GetString() const {
+  return internal::StringStreamToString(ss_.get());
+}
+
 // AssertionResult constructors.
 // Used in EXPECT_TRUE/FALSE(assertion_result).
 AssertionResult::AssertionResult(const AssertionResult& other)
@@ -2307,9 +2501,9 @@ AssertionResult EqFailure(const char* expected_expression,
 // Constructs a failure message for Boolean assertions such as EXPECT_TRUE.
 std::string GetBoolAssertionFailureMessage(
     const AssertionResult& assertion_result,
-                                      const char* expression_text,
-                                      const char* actual_predicate_value,
-                                      const char* expected_predicate_value) {
+    const char* expression_text,
+    const char* actual_predicate_value,
+    const char* expected_predicate_value) {
   const char* actual_message = assertion_result.message();
   Message msg;
   msg << "Value of: " << expression_text
@@ -2637,7 +2831,7 @@ AssertionResult HRESULTFailureHelper(const char* expr,
   // want inserts expanded.
   const DWORD kFlags = FORMAT_MESSAGE_FROM_SYSTEM |
                        FORMAT_MESSAGE_IGNORE_INSERTS;
-  const DWORD kBufSize = 4096;  // String::Format can't exceed this length.
+  const DWORD kBufSize = 4096;
   // Gets the system's human readable message string for this HRESULT.
   char error_text[kBufSize] = { '\0' };
   DWORD message_length = ::FormatMessageA(kFlags,
@@ -2647,7 +2841,7 @@ AssertionResult HRESULTFailureHelper(const char* expr,
                                           error_text,  // output buffer
                                           kBufSize,  // buf size
                                           NULL);  // no arguments for inserts
-  // Trims tailing white space (FormatMessage leaves a trailing cr-lf)
+  // Trims tailing white space (FormatMessage leaves a trailing CR-LF)
   for (; message_length && IsSpace(error_text[message_length - 1]);
           --message_length) {
     error_text[message_length - 1] = '\0';
@@ -2655,10 +2849,10 @@ AssertionResult HRESULTFailureHelper(const char* expr,
 
 # endif  // GTEST_OS_WINDOWS_MOBILE
 
-  const std::string error_hex(String::Format("0x%08X ", hr));
+  const std::string error_hex("0x" + String::FormatHexInt(hr));
   return ::testing::AssertionFailure()
       << "Expected: " << expr << " " << expected << ".\n"
-      << "  Actual: " << error_hex << error_text << "\n";
+      << "  Actual: " << error_hex << " " << error_text << "\n";
 }
 
 }  // namespace
@@ -2715,12 +2909,15 @@ inline UInt32 ChopLowBits(UInt32* bits, int n) {
 // Converts a Unicode code point to a narrow string in UTF-8 encoding.
 // code_point parameter is of type UInt32 because wchar_t may not be
 // wide enough to contain a code point.
-// The output buffer str must containt at least 32 characters.
-// The function returns the address of the output buffer.
 // If the code_point is not a valid Unicode code point
-// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be output
-// as '(Invalid Unicode 0xXXXXXXXX)'.
-char* CodePointToUtf8(UInt32 code_point, char* str) {
+// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be converted
+// to "(Invalid Unicode 0xXXXXXXXX)".
+std::string CodePointToUtf8(UInt32 code_point) {
+  if (code_point > kMaxCodePoint4) {
+    return "(Invalid Unicode 0x" + String::FormatHexInt(code_point) + ")";
+  }
+
+  char str[5];  // Big enough for the largest valid code point.
   if (code_point <= kMaxCodePoint1) {
     str[1] = '\0';
     str[0] = static_cast<char>(code_point);                          // 0xxxxxxx
@@ -2733,22 +2930,12 @@ char* CodePointToUtf8(UInt32 code_point, char* str) {
     str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
     str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
     str[0] = static_cast<char>(0xE0 | code_point);                   // 1110xxxx
-  } else if (code_point <= kMaxCodePoint4) {
+  } else {  // code_point <= kMaxCodePoint4
     str[4] = '\0';
     str[3] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
     str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
     str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6));  // 10xxxxxx
     str[0] = static_cast<char>(0xF0 | code_point);                   // 11110xxx
-  } else {
-    // The longest string String::Format can produce when invoked
-    // with these parameters is 28 character long (not including
-    // the terminating nul character). We are asking for 32 character
-    // buffer just in case. This is also enough for strncpy to
-    // null-terminate the destination string.
-    posix::StrNCpy(
-        str, String::Format("(Invalid Unicode 0x%X)", code_point).c_str(), 32);
-    str[31] = '\0';  // Makes sure no change in the format to strncpy leaves
-                     // the result unterminated.
   }
   return str;
 }
@@ -2807,8 +2994,7 @@ std::string WideStringToUtf8(const wchar_t* str, int num_chars) {
       unicode_code_point = static_cast<UInt32>(str[i]);
     }
 
-    char buffer[32];  // CodePointToUtf8 requires a buffer this big.
-    stream << CodePointToUtf8(unicode_code_point, buffer);
+    stream << CodePointToUtf8(unicode_code_point);
   }
   return StringStreamToString(&stream);
 }
@@ -2925,47 +3111,26 @@ bool String::EndsWithCaseInsensitive(
                                       suffix.c_str());
 }
 
-// Formats a list of arguments to an std::string, using the same format
-// spec string as for printf.
-//
-// We do not use the StringPrintf class as it is not universally
-// available.
-//
-// The result is limited to 4096 characters (including the tailing 0).
-// If 4096 characters are not enough to format the input, or if
-// there's an error, "<formatting error or buffer exceeded>" is
-// returned.
-std::string String::Format(const char * format, ...) {
-  va_list args;
-  va_start(args, format);
+// Formats an int value as "%02d".
+std::string String::FormatIntWidth2(int value) {
+  std::stringstream ss;
+  ss << std::setfill('0') << std::setw(2) << value;
+  return ss.str();
+}
 
-  char buffer[4096];
-  const int kBufferSize = sizeof(buffer)/sizeof(buffer[0]);
+// Formats an int value as "%X".
+std::string String::FormatHexInt(int value) {
+  std::stringstream ss;
+  ss << std::hex << std::uppercase << value;
+  return ss.str();
+}
 
-  // MSVC 8 deprecates vsnprintf(), so we want to suppress warning
-  // 4996 (deprecated function) there.
-#ifdef _MSC_VER  // We are using MSVC.
-# pragma warning(push)          // Saves the current warning state.
-# pragma warning(disable:4996)  // Temporarily disables warning 4996.
-
-  const int size = vsnprintf(buffer, kBufferSize, format, args);
-
-# pragma warning(pop)           // Restores the warning state.
-#else  // We are not using MSVC.
-  const int size = vsnprintf(buffer, kBufferSize, format, args);
-#endif  // _MSC_VER
-  va_end(args);
-
-  // vsnprintf()'s behavior is not portable.  When the buffer is not
-  // big enough, it returns a negative value in MSVC, and returns the
-  // needed buffer size on Linux.  When there is an output error, it
-  // always returns a negative value.  For simplicity, we lump the two
-  // error cases together.
-  if (size < 0 || size >= kBufferSize) {
-    return "<formatting error or buffer exceeded>";
-  } else {
-    return std::string(buffer, size);
-  }
+// Formats a byte as "%02X".
+std::string String::FormatByte(unsigned char value) {
+  std::stringstream ss;
+  ss << std::setfill('0') << std::setw(2) << std::hex << std::uppercase
+     << static_cast<unsigned int>(value);
+  return ss.str();
 }
 
 // Converts the buffer in a stringstream to an std::string, converting NUL
@@ -2990,7 +3155,7 @@ std::string StringStreamToString(::std::stringstream* ss) {
 
 // Appends the user-supplied message to the Google-Test-generated message.
 std::string AppendUserMessage(const std::string& gtest_msg,
-                         const Message& user_msg) {
+                              const Message& user_msg) {
   // Appends the user message if it's non-empty.
   const std::string user_msg_string = user_msg.GetString();
   if (user_msg_string.empty()) {
@@ -3045,8 +3210,9 @@ void TestResult::AddTestPartResult(const TestPartResult& test_part_result) {
 // Adds a test property to the list. If a property with the same key as the
 // supplied property is already represented, the value of this test_property
 // replaces the old value for that key.
-void TestResult::RecordProperty(const TestProperty& test_property) {
-  if (!ValidateTestProperty(test_property)) {
+void TestResult::RecordProperty(const std::string& xml_element,
+                                const TestProperty& test_property) {
+  if (!ValidateTestProperty(xml_element, test_property)) {
     return;
   }
   internal::MutexLock lock(&test_properites_mutex_);
@@ -3060,21 +3226,94 @@ void TestResult::RecordProperty(const TestProperty& test_property) {
   property_with_matching_key->SetValue(test_property.value());
 }
 
-// Adds a failure if the key is a reserved attribute of Google Test
-// testcase tags.  Returns true if the property is valid.
-bool TestResult::ValidateTestProperty(const TestProperty& test_property) {
-  const std::string& key = test_property.key();
-  if (key == "name" || key == "status" || key == "time" || key == "classname") {
-    ADD_FAILURE()
-        << "Reserved key used in RecordProperty(): "
-        << key
-        << " ('name', 'status', 'time', and 'classname' are reserved by "
-        << GTEST_NAME_ << ")";
+// The list of reserved attributes used in the <testsuites> element of XML
+// output.
+static const char* const kReservedTestSuitesAttributes[] = {
+  "disabled",
+  "errors",
+  "failures",
+  "name",
+  "random_seed",
+  "tests",
+  "time",
+  "timestamp"
+};
+
+// The list of reserved attributes used in the <testsuite> element of XML
+// output.
+static const char* const kReservedTestSuiteAttributes[] = {
+  "disabled",
+  "errors",
+  "failures",
+  "name",
+  "tests",
+  "time"
+};
+
+// The list of reserved attributes used in the <testcase> element of XML output.
+static const char* const kReservedTestCaseAttributes[] = {
+  "classname",
+  "name",
+  "status",
+  "time",
+  "type_param",
+  "value_param"
+};
+
+template <int kSize>
+std::vector<std::string> ArrayAsVector(const char* const (&array)[kSize]) {
+  return std::vector<std::string>(array, array + kSize);
+}
+
+static std::vector<std::string> GetReservedAttributesForElement(
+    const std::string& xml_element) {
+  if (xml_element == "testsuites") {
+    return ArrayAsVector(kReservedTestSuitesAttributes);
+  } else if (xml_element == "testsuite") {
+    return ArrayAsVector(kReservedTestSuiteAttributes);
+  } else if (xml_element == "testcase") {
+    return ArrayAsVector(kReservedTestCaseAttributes);
+  } else {
+    GTEST_CHECK_(false) << "Unrecognized xml_element provided: " << xml_element;
+  }
+  // This code is unreachable but some compilers may not realizes that.
+  return std::vector<std::string>();
+}
+
+static std::string FormatWordList(const std::vector<std::string>& words) {
+  Message word_list;
+  for (size_t i = 0; i < words.size(); ++i) {
+    if (i > 0 && words.size() > 2) {
+      word_list << ", ";
+    }
+    if (i == words.size() - 1) {
+      word_list << "and ";
+    }
+    word_list << "'" << words[i] << "'";
+  }
+  return word_list.GetString();
+}
+
+static bool ValidateTestPropertyName(const std::string& property_name,
+                              const std::vector<std::string>& reserved_names) {
+  if (std::find(reserved_names.begin(), reserved_names.end(), property_name) !=
+          reserved_names.end()) {
+    ADD_FAILURE() << "Reserved key used in RecordProperty(): " << property_name
+                  << " (" << FormatWordList(reserved_names)
+                  << " are reserved by " << GTEST_NAME_ << ")";
     return false;
   }
   return true;
 }
 
+// Adds a failure if the key is a reserved attribute of the element named
+// xml_element.  Returns true if the property is valid.
+bool TestResult::ValidateTestProperty(const std::string& xml_element,
+                                      const TestProperty& test_property) {
+  return ValidateTestPropertyName(test_property.key(),
+                                  GetReservedAttributesForElement(xml_element));
+}
+
 // Clears the object.
 void TestResult::Clear() {
   test_part_results_.clear();
@@ -3150,12 +3389,12 @@ void Test::TearDown() {
 }
 
 // Allows user supplied key value pairs to be recorded for later output.
-void Test::RecordProperty(const char* key, const char* value) {
-  UnitTest::GetInstance()->RecordPropertyForCurrentTest(key, value);
+void Test::RecordProperty(const std::string& key, const std::string& value) {
+  UnitTest::GetInstance()->RecordProperty(key, value);
 }
 
 // Allows user supplied key value pairs to be recorded for later output.
-void Test::RecordProperty(const char* key, int value) {
+void Test::RecordProperty(const std::string& key, int value) {
   Message value_message;
   value_message << value;
   RecordProperty(key, value_message.GetString().c_str());
@@ -3250,7 +3489,7 @@ bool Test::HasSameFixtureClass() {
 // prohibits creation of objects with destructors on stack in functions
 // using __try (see error C2712).
 static std::string* FormatSehExceptionMessage(DWORD exception_code,
-                                                   const char* location) {
+                                              const char* location) {
   Message message;
   message << "SEH exception with code 0x" << std::setbase(16) <<
     exception_code << std::setbase(10) << " thrown in " << location << ".";
@@ -3260,11 +3499,13 @@ static std::string* FormatSehExceptionMessage(DWORD exception_code,
 
 #endif  // GTEST_HAS_SEH
 
+namespace internal {
+
 #if GTEST_HAS_EXCEPTIONS
 
 // Adds an "exception thrown" fatal failure to the current test.
 static std::string FormatCxxExceptionMessage(const char* description,
-                                                  const char* location) {
+                                             const char* location) {
   Message message;
   if (description != NULL) {
     message << "C++ exception with description \"" << description << "\"";
@@ -3279,20 +3520,12 @@ static std::string FormatCxxExceptionMessage(const char* description,
 static std::string PrintTestPartResultToString(
     const TestPartResult& test_part_result);
 
-// A failed Google Test assertion will throw an exception of this type when
-// GTEST_FLAG(throw_on_failure) is true (if exceptions are enabled).  We
-// derive it from std::runtime_error, which is for errors presumably
-// detectable only at run time.  Since std::runtime_error inherits from
-// std::exception, many testing frameworks know how to extract and print the
-// message inside it.
-class GoogleTestFailureException : public ::std::runtime_error {
- public:
-  explicit GoogleTestFailureException(const TestPartResult& failure)
+GoogleTestFailureException::GoogleTestFailureException(
+    const TestPartResult& failure)
       : ::std::runtime_error(PrintTestPartResultToString(failure).c_str()) {}
-};
+
 #endif  // GTEST_HAS_EXCEPTIONS
 
-namespace internal {
 // We put these helper functions in the internal namespace as IBM's xlC
 // compiler rejects the code if they were declared static.
 
@@ -3358,9 +3591,10 @@ Result HandleExceptionsInMethodIfSupported(
 #if GTEST_HAS_EXCEPTIONS
     try {
       return HandleSehExceptionsInMethodIfSupported(object, method, location);
-    } catch (const GoogleTestFailureException&) {  // NOLINT
-      // This exception doesn't originate in code under test. It makes no
-      // sense to report it as a test failure.
+    } catch (const internal::GoogleTestFailureException&) {  // NOLINT
+      // This exception type can only be thrown by a failed Google
+      // Test assertion with the intention of letting another testing
+      // framework catch it.  Therefore we just re-throw it.
       throw;
     } catch (const std::exception& e) {  // NOLINT
       internal::ReportFailureInUnknownLocation(
@@ -3419,10 +3653,8 @@ bool Test::HasNonfatalFailure() {
 
 // Constructs a TestInfo object. It assumes ownership of the test factory
 // object.
-// TODO(vladl@google.com): Make a_test_case_name and a_name const string&'s
-// to signify they cannot be NULLs.
-TestInfo::TestInfo(const char* a_test_case_name,
-                   const char* a_name,
+TestInfo::TestInfo(const std::string& a_test_case_name,
+                   const std::string& a_name,
                    const char* a_type_param,
                    const char* a_value_param,
                    internal::TypeId fixture_class_id,
@@ -3461,7 +3693,8 @@ namespace internal {
 //                     The newly created TestInfo instance will assume
 //                     ownership of the factory object.
 TestInfo* MakeAndRegisterTestInfo(
-    const char* test_case_name, const char* name,
+    const char* test_case_name,
+    const char* name,
     const char* type_param,
     const char* value_param,
     TypeId fixture_class_id,
@@ -3690,6 +3923,7 @@ void TestCase::Run() {
 
 // Clears the results of all tests in this test case.
 void TestCase::ClearResult() {
+  ad_hoc_test_result_.Clear();
   ForEach(test_info_list_, TestInfo::ClearTestResult);
 }
 
@@ -3711,10 +3945,10 @@ void TestCase::UnshuffleTests() {
 // FormatCountableNoun(1, "formula", "formuli") returns "1 formula".
 // FormatCountableNoun(5, "book", "books") returns "5 books".
 static std::string FormatCountableNoun(int count,
-                                            const char * singular_form,
-                                            const char * plural_form) {
-  return internal::String::Format("%d %s", count,
-                                  count == 1 ? singular_form : plural_form);
+                                       const char * singular_form,
+                                       const char * plural_form) {
+  return internal::StreamableToString(count) + " " +
+      (count == 1 ? singular_form : plural_form);
 }
 
 // Formats the count of tests.
@@ -3748,6 +3982,8 @@ static const char * TestPartResultTypeToString(TestPartResult::Type type) {
   }
 }
 
+namespace internal {
+
 // Prints a TestPartResult to an std::string.
 static std::string PrintTestPartResultToString(
     const TestPartResult& test_part_result) {
@@ -3779,8 +4015,6 @@ static void PrintTestPartResult(const TestPartResult& test_part_result) {
 
 // class PrettyUnitTestResultPrinter
 
-namespace internal {
-
 enum GTestColor {
   COLOR_DEFAULT,
   COLOR_RED,
@@ -3832,6 +4066,7 @@ bool ShouldUseColor(bool stdout_is_tty) {
         String::CStringEquals(term, "xterm-color") ||
         String::CStringEquals(term, "xterm-256color") ||
         String::CStringEquals(term, "screen") ||
+        String::CStringEquals(term, "screen-256color") ||
         String::CStringEquals(term, "linux") ||
         String::CStringEquals(term, "cygwin");
     return stdout_is_tty && term_supports_color;
@@ -3897,6 +4132,11 @@ static void ColoredPrintf(GTestColor color, const char* fmt, ...) {
   va_end(args);
 }
 
+// Text printed in Google Test's text output and --gunit_list_tests
+// output to label the type parameter and value parameter for a test.
+static const char kTypeParamLabel[] = "TypeParam";
+static const char kValueParamLabel[] = "GetParam()";
+
 static void PrintFullTestCommentIfPresent(const TestInfo& test_info) {
   const char* const type_param = test_info.type_param();
   const char* const value_param = test_info.value_param();
@@ -3904,12 +4144,12 @@ static void PrintFullTestCommentIfPresent(const TestInfo& test_info) {
   if (type_param != NULL || value_param != NULL) {
     printf(", where ");
     if (type_param != NULL) {
-      printf("TypeParam = %s", type_param);
+      printf("%s = %s", kTypeParamLabel, type_param);
       if (value_param != NULL)
         printf(" and ");
     }
     if (value_param != NULL) {
-      printf("GetParam() = %s", value_param);
+      printf("%s = %s", kValueParamLabel, value_param);
     }
   }
 }
@@ -3958,6 +4198,14 @@ void PrettyUnitTestResultPrinter::OnTestIterationStart(
                   "Note: %s filter = %s\n", GTEST_NAME_, filter);
   }
 
+  const char* const param_filter = GTEST_FLAG(param_filter).c_str();
+
+  // Ditto.
+  if (!String::CStringEquals(param_filter, kUniversalFilter)) {
+    ColoredPrintf(COLOR_YELLOW,
+                  "Note: %s parameter filter = %s\n", GTEST_NAME_, param_filter);
+  }
+
   if (internal::ShouldShard(kTestTotalShards, kTestShardIndex, false)) {
     const Int32 shard_index = Int32FromEnvOrDie(kTestShardIndex, -1);
     ColoredPrintf(COLOR_YELLOW,
@@ -3994,7 +4242,7 @@ void PrettyUnitTestResultPrinter::OnTestCaseStart(const TestCase& test_case) {
   if (test_case.type_param() == NULL) {
     printf("\n");
   } else {
-    printf(", where TypeParam = %s\n", test_case.type_param());
+    printf(", where %s = %s\n", kTypeParamLabel, test_case.type_param());
   }
   fflush(stdout);
 }
@@ -4259,13 +4507,13 @@ class XmlUnitTestResultPrinter : public EmptyTestEventListener {
   // is_attribute is true, the text is meant to appear as an attribute
   // value, and normalizable whitespace is preserved by replacing it
   // with character references.
-  static std::string EscapeXml(const char* str, bool is_attribute);
+  static std::string EscapeXml(const std::string& str, bool is_attribute);
 
   // Returns the given string with all characters invalid in XML removed.
-  static string RemoveInvalidXmlCharacters(const string& str);
+  static std::string RemoveInvalidXmlCharacters(const std::string& str);
 
   // Convenience wrapper around EscapeXml when str is an attribute value.
-  static std::string EscapeXmlAttribute(const char* str) {
+  static std::string EscapeXmlAttribute(const std::string& str) {
     return EscapeXml(str, true);
   }
 
@@ -4274,6 +4522,13 @@ class XmlUnitTestResultPrinter : public EmptyTestEventListener {
     return EscapeXml(str, false);
   }
 
+  // Verifies that the given attribute belongs to the given element and
+  // streams the attribute as XML.
+  static void OutputXmlAttribute(std::ostream* stream,
+                                 const std::string& element_name,
+                                 const std::string& name,
+                                 const std::string& value);
+
   // Streams an XML CDATA section, escaping invalid CDATA sequences as needed.
   static void OutputXmlCDataSection(::std::ostream* stream, const char* data);
 
@@ -4283,10 +4538,12 @@ class XmlUnitTestResultPrinter : public EmptyTestEventListener {
                                 const TestInfo& test_info);
 
   // Prints an XML representation of a TestCase object
-  static void PrintXmlTestCase(FILE* out, const TestCase& test_case);
+  static void PrintXmlTestCase(::std::ostream* stream,
+                               const TestCase& test_case);
 
   // Prints an XML summary of unit_test to output stream out.
-  static void PrintXmlUnitTest(FILE* out, const UnitTest& unit_test);
+  static void PrintXmlUnitTest(::std::ostream* stream,
+                               const UnitTest& unit_test);
 
   // Produces a string representing the test properties in a result as space
   // delimited XML attributes based on the property key="value" pairs.
@@ -4337,7 +4594,9 @@ void XmlUnitTestResultPrinter::OnTestIterationEnd(const UnitTest& unit_test,
     fflush(stderr);
     exit(EXIT_FAILURE);
   }
-  PrintXmlUnitTest(xmlout, unit_test);
+  std::stringstream stream;
+  PrintXmlUnitTest(&stream, unit_test);
+  fprintf(xmlout, "%s", StringStreamToString(&stream).c_str());
   fclose(xmlout);
 }
 
@@ -4354,12 +4613,12 @@ void XmlUnitTestResultPrinter::OnTestIterationEnd(const UnitTest& unit_test,
 // TODO(wan): It might be nice to have a minimally invasive, human-readable
 // escaping scheme for invalid characters, rather than dropping them.
 std::string XmlUnitTestResultPrinter::EscapeXml(
-    const char* str, bool is_attribute) {
+    const std::string& str, bool is_attribute) {
   Message m;
 
-  if (str != NULL) {
-    for (const char* src = str; *src; ++src) {
-      switch (*src) {
+  for (size_t i = 0; i < str.size(); ++i) {
+    const char ch = str[i];
+    switch (ch) {
         case '<':
           m << "&lt;";
           break;
@@ -4382,16 +4641,16 @@ std::string XmlUnitTestResultPrinter::EscapeXml(
             m << '"';
           break;
         default:
-          if (IsValidXmlCharacter(*src)) {
-            if (is_attribute && IsNormalizableWhitespace(*src))
-              m << String::Format("&#x%02X;", unsigned(*src));
+        if (IsValidXmlCharacter(ch)) {
+          if (is_attribute && IsNormalizableWhitespace(ch))
+            m << "&#x" << String::FormatByte(static_cast<unsigned char>(ch))
+              << ";";
             else
-              m << *src;
+            m << ch;
           }
           break;
       }
     }
-  }
 
   return m.GetString();
 }
@@ -4399,10 +4658,11 @@ std::string XmlUnitTestResultPrinter::EscapeXml(
 // Returns the given string with all characters invalid in XML removed.
 // Currently invalid characters are dropped from the string. An
 // alternative is to replace them with certain characters such as . or ?.
-string XmlUnitTestResultPrinter::RemoveInvalidXmlCharacters(const string& str) {
-  string output;
+std::string XmlUnitTestResultPrinter::RemoveInvalidXmlCharacters(
+    const std::string& str) {
+  std::string output;
   output.reserve(str.size());
-  for (string::const_iterator it = str.begin(); it != str.end(); ++it)
+  for (std::string::const_iterator it = str.begin(); it != str.end(); ++it)
     if (IsValidXmlCharacter(*it))
       output.push_back(*it);
 
@@ -4449,13 +4709,13 @@ std::string FormatEpochTimeInMillisAsIso8601(TimeInMillis ms) {
   if (time_struct == NULL)
     return "";  // Invalid ms value
 
-  return String::Format("%d-%02d-%02dT%02d:%02d:%02d",  // YYYY-MM-DDThh:mm:ss
-                        time_struct->tm_year + 1900,
-                        time_struct->tm_mon + 1,
-                        time_struct->tm_mday,
-                        time_struct->tm_hour,
-                        time_struct->tm_min,
-                        time_struct->tm_sec);
+  // YYYY-MM-DDThh:mm:ss
+  return StreamableToString(time_struct->tm_year + 1900) + "-" +
+      String::FormatIntWidth2(time_struct->tm_mon + 1) + "-" +
+      String::FormatIntWidth2(time_struct->tm_mday) + "T" +
+      String::FormatIntWidth2(time_struct->tm_hour) + ":" +
+      String::FormatIntWidth2(time_struct->tm_min) + ":" +
+      String::FormatIntWidth2(time_struct->tm_sec);
 }
 
 // Streams an XML CDATA section, escaping invalid CDATA sequences as needed.
@@ -4478,30 +4738,47 @@ void XmlUnitTestResultPrinter::OutputXmlCDataSection(::std::ostream* stream,
   *stream << "]]>";
 }
 
+void XmlUnitTestResultPrinter::OutputXmlAttribute(
+    std::ostream* stream,
+    const std::string& element_name,
+    const std::string& name,
+    const std::string& value) {
+  const std::vector<std::string>& allowed_names =
+      GetReservedAttributesForElement(element_name);
+
+  GTEST_CHECK_(std::find(allowed_names.begin(), allowed_names.end(), name) !=
+                   allowed_names.end())
+      << "Attribute " << name << " is not allowed for element <" << element_name
+      << ">.";
+
+  *stream << " " << name << "=\"" << EscapeXmlAttribute(value) << "\"";
+}
+
 // Prints an XML representation of a TestInfo object.
 // TODO(wan): There is also value in printing properties with the plain printer.
 void XmlUnitTestResultPrinter::OutputXmlTestInfo(::std::ostream* stream,
                                                  const char* test_case_name,
                                                  const TestInfo& test_info) {
   const TestResult& result = *test_info.result();
-  *stream << "    <testcase name=\""
-          << EscapeXmlAttribute(test_info.name()).c_str() << "\"";
+  const std::string kTestcase = "testcase";
+
+  *stream << "    <testcase";
+  OutputXmlAttribute(stream, kTestcase, "name", test_info.name());
 
   if (test_info.value_param() != NULL) {
-    *stream << " value_param=\"" << EscapeXmlAttribute(test_info.value_param())
-            << "\"";
+    OutputXmlAttribute(stream, kTestcase, "value_param",
+                       test_info.value_param());
   }
   if (test_info.type_param() != NULL) {
-    *stream << " type_param=\"" << EscapeXmlAttribute(test_info.type_param())
-            << "\"";
+    OutputXmlAttribute(stream, kTestcase, "type_param", test_info.type_param());
   }
 
-  *stream << " status=\""
-          << (test_info.should_run() ? "run" : "notrun")
-          << "\" time=\""
-          << FormatTimeInMillisAsSeconds(result.elapsed_time())
-          << "\" classname=\"" << EscapeXmlAttribute(test_case_name).c_str()
-          << "\"" << TestPropertiesAsXmlAttributes(result).c_str();
+  OutputXmlAttribute(stream, kTestcase, "status",
+                     test_info.should_run() ? "run" : "notrun");
+  OutputXmlAttribute(stream, kTestcase, "time",
+                     FormatTimeInMillisAsSeconds(result.elapsed_time()));
+  OutputXmlAttribute(stream, kTestcase, "classname", test_case_name);
+  *stream << TestPropertiesAsXmlAttributes(result);
 
   int failures = 0;
   for (int i = 0; i < result.total_part_count(); ++i) {
@@ -4529,45 +4806,64 @@ void XmlUnitTestResultPrinter::OutputXmlTestInfo(::std::ostream* stream,
 }
 
 // Prints an XML representation of a TestCase object
-void XmlUnitTestResultPrinter::PrintXmlTestCase(FILE* out,
+void XmlUnitTestResultPrinter::PrintXmlTestCase(std::ostream* stream,
                                                 const TestCase& test_case) {
-  fprintf(out,
-          "  <testsuite name=\"%s\" tests=\"%d\" failures=\"%d\" "
-          "disabled=\"%d\" ",
-          EscapeXmlAttribute(test_case.name()).c_str(),
-          test_case.total_test_count(),
-          test_case.failed_test_count(),
-          test_case.disabled_test_count());
-  fprintf(out,
-          "errors=\"0\" time=\"%s\">\n",
-          FormatTimeInMillisAsSeconds(test_case.elapsed_time()).c_str());
-  for (int i = 0; i < test_case.total_test_count(); ++i) {
-    ::std::stringstream stream;
-    OutputXmlTestInfo(&stream, test_case.name(), *test_case.GetTestInfo(i));
-    fprintf(out, "%s", StringStreamToString(&stream).c_str());
-  }
-  fprintf(out, "  </testsuite>\n");
+  const std::string kTestsuite = "testsuite";
+  *stream << "  <" << kTestsuite;
+  OutputXmlAttribute(stream, kTestsuite, "name", test_case.name());
+  OutputXmlAttribute(stream, kTestsuite, "tests",
+                     StreamableToString(test_case.total_test_count()));
+  OutputXmlAttribute(stream, kTestsuite, "failures",
+                     StreamableToString(test_case.failed_test_count()));
+  OutputXmlAttribute(stream, kTestsuite, "disabled",
+                     StreamableToString(test_case.disabled_test_count()));
+  OutputXmlAttribute(stream, kTestsuite, "errors", "0");
+  OutputXmlAttribute(stream, kTestsuite, "time",
+                     FormatTimeInMillisAsSeconds(test_case.elapsed_time()));
+  *stream << TestPropertiesAsXmlAttributes(test_case.ad_hoc_test_result())
+          << ">\n";
+
+  for (int i = 0; i < test_case.total_test_count(); ++i)
+    OutputXmlTestInfo(stream, test_case.name(), *test_case.GetTestInfo(i));
+  *stream << "  </" << kTestsuite << ">\n";
 }
 
 // Prints an XML summary of unit_test to output stream out.
-void XmlUnitTestResultPrinter::PrintXmlUnitTest(FILE* out,
+void XmlUnitTestResultPrinter::PrintXmlUnitTest(std::ostream* stream,
                                                 const UnitTest& unit_test) {
-  fprintf(out, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n");
-  fprintf(out,
-          "<testsuites tests=\"%d\" failures=\"%d\" disabled=\"%d\" "
-          "errors=\"0\" timestamp=\"%s\" time=\"%s\" ",
-          unit_test.total_test_count(),
-          unit_test.failed_test_count(),
-          unit_test.disabled_test_count(),
-          FormatEpochTimeInMillisAsIso8601(unit_test.start_timestamp()).c_str(),
-          FormatTimeInMillisAsSeconds(unit_test.elapsed_time()).c_str());
+  const std::string kTestsuites = "testsuites";
+
+  *stream << "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";
+  *stream << "<" << kTestsuites;
+
+  OutputXmlAttribute(stream, kTestsuites, "tests",
+                     StreamableToString(unit_test.total_test_count()));
+  OutputXmlAttribute(stream, kTestsuites, "failures",
+                     StreamableToString(unit_test.failed_test_count()));
+  OutputXmlAttribute(stream, kTestsuites, "disabled",
+                     StreamableToString(unit_test.disabled_test_count()));
+  OutputXmlAttribute(stream, kTestsuites, "errors", "0");
+  OutputXmlAttribute(
+      stream, kTestsuites, "timestamp",
+      FormatEpochTimeInMillisAsIso8601(unit_test.start_timestamp()));
+  OutputXmlAttribute(stream, kTestsuites, "time",
+                     FormatTimeInMillisAsSeconds(unit_test.elapsed_time()));
+
   if (GTEST_FLAG(shuffle)) {
-    fprintf(out, "random_seed=\"%d\" ", unit_test.random_seed());
+    OutputXmlAttribute(stream, kTestsuites, "random_seed",
+                       StreamableToString(unit_test.random_seed()));
   }
-  fprintf(out, "name=\"AllTests\">\n");
-  for (int i = 0; i < unit_test.total_test_case_count(); ++i)
-    PrintXmlTestCase(out, *unit_test.GetTestCase(i));
-  fprintf(out, "</testsuites>\n");
+
+  *stream << TestPropertiesAsXmlAttributes(unit_test.ad_hoc_test_result());
+
+  OutputXmlAttribute(stream, kTestsuites, "name", "AllTests");
+  *stream << ">\n";
+
+
+  for (int i = 0; i < unit_test.total_test_case_count(); ++i) {
+    PrintXmlTestCase(stream, *unit_test.GetTestCase(i));
+  }
+  *stream << "</" << kTestsuites << ">\n";
 }
 
 // Produces a string representing the test properties in a result as space
@@ -4587,112 +4883,6 @@ std::string XmlUnitTestResultPrinter::TestPropertiesAsXmlAttributes(
 
 #if GTEST_CAN_STREAM_RESULTS_
 
-// Streams test results to the given port on the given host machine.
-class StreamingListener : public EmptyTestEventListener {
- public:
-  // Escapes '=', '&', '%', and '\n' characters in str as "%xx".
-  static string UrlEncode(const char* str);
-
-  StreamingListener(const string& host, const string& port)
-      : sockfd_(-1), host_name_(host), port_num_(port) {
-    MakeConnection();
-    Send("gtest_streaming_protocol_version=1.0\n");
-  }
-
-  virtual ~StreamingListener() {
-    if (sockfd_ != -1)
-      CloseConnection();
-  }
-
-  void OnTestProgramStart(const UnitTest& /* unit_test */) {
-    Send("event=TestProgramStart\n");
-  }
-
-  void OnTestProgramEnd(const UnitTest& unit_test) {
-    // Note that Google Test current only report elapsed time for each
-    // test iteration, not for the entire test program.
-    Send(String::Format("event=TestProgramEnd&passed=%d\n",
-                        unit_test.Passed()));
-
-    // Notify the streaming server to stop.
-    CloseConnection();
-  }
-
-  void OnTestIterationStart(const UnitTest& /* unit_test */, int iteration) {
-    Send(String::Format("event=TestIterationStart&iteration=%d\n",
-                        iteration));
-  }
-
-  void OnTestIterationEnd(const UnitTest& unit_test, int /* iteration */) {
-    Send(String::Format("event=TestIterationEnd&passed=%d&elapsed_time=%sms\n",
-                        unit_test.Passed(),
-                        StreamableToString(unit_test.elapsed_time()).c_str()));
-  }
-
-  void OnTestCaseStart(const TestCase& test_case) {
-    Send(String::Format("event=TestCaseStart&name=%s\n", test_case.name()));
-  }
-
-  void OnTestCaseEnd(const TestCase& test_case) {
-    Send(String::Format("event=TestCaseEnd&passed=%d&elapsed_time=%sms\n",
-                        test_case.Passed(),
-                        StreamableToString(test_case.elapsed_time()).c_str()));
-  }
-
-  void OnTestStart(const TestInfo& test_info) {
-    Send(String::Format("event=TestStart&name=%s\n", test_info.name()));
-  }
-
-  void OnTestEnd(const TestInfo& test_info) {
-    Send(String::Format(
-        "event=TestEnd&passed=%d&elapsed_time=%sms\n",
-        (test_info.result())->Passed(),
-        StreamableToString((test_info.result())->elapsed_time()).c_str()));
-  }
-
-  void OnTestPartResult(const TestPartResult& test_part_result) {
-    const char* file_name = test_part_result.file_name();
-    if (file_name == NULL)
-      file_name = "";
-    Send(String::Format("event=TestPartResult&file=%s&line=%d&message=",
-                        UrlEncode(file_name).c_str(),
-                        test_part_result.line_number()));
-    Send(UrlEncode(test_part_result.message()) + "\n");
-  }
-
- private:
-  // Creates a client socket and connects to the server.
-  void MakeConnection();
-
-  // Closes the socket.
-  void CloseConnection() {
-    GTEST_CHECK_(sockfd_ != -1)
-        << "CloseConnection() can be called only when there is a connection.";
-
-    close(sockfd_);
-    sockfd_ = -1;
-  }
-
-  // Sends a string to the socket.
-  void Send(const string& message) {
-    GTEST_CHECK_(sockfd_ != -1)
-        << "Send() can be called only when there is a connection.";
-
-    const int len = static_cast<int>(message.length());
-    if (write(sockfd_, message.c_str(), len) != len) {
-      GTEST_LOG_(WARNING)
-          << "stream_result_to: failed to stream to "
-          << host_name_ << ":" << port_num_;
-    }
-  }
-
-  int sockfd_;   // socket file descriptor
-  const string host_name_;
-  const string port_num_;
-
-  GTEST_DISALLOW_COPY_AND_ASSIGN_(StreamingListener);
-};  // class StreamingListener
-
 // Checks if str contains '=', '&', '%' or '\n' characters. If yes,
 // replaces them by "%xx" where xx is their hexadecimal value. For
 // example, replaces "=" with "%3D".  This algorithm is O(strlen(str))
@@ -4707,7 +4897,7 @@ string StreamingListener::UrlEncode(const char* str) {
       case '=':
       case '&':
       case '\n':
-        result.append(String::Format("%%%02x", static_cast<unsigned char>(ch)));
+        result.append("%" + String::FormatByte(static_cast<unsigned char>(ch)));
         break;
       default:
         result.push_back(ch);
@@ -4717,7 +4907,7 @@ string StreamingListener::UrlEncode(const char* str) {
   return result;
 }
 
-void StreamingListener::MakeConnection() {
+void StreamingListener::SocketWriter::MakeConnection() {
   GTEST_CHECK_(sockfd_ == -1)
       << "MakeConnection() can't be called when there is already a connection.";
 
@@ -4976,6 +5166,12 @@ const TestCase* UnitTest::GetTestCase(int i) const {
   return impl()->GetTestCase(i);
 }
 
+// Returns the TestResult containing information on test failures and
+// properties logged outside of individual test cases.
+const TestResult& UnitTest::ad_hoc_test_result() const {
+  return *impl()->ad_hoc_test_result();
+}
+
 // Gets the i-th test case among all the test cases. i can range from 0 to
 // total_test_case_count() - 1. If i is not in that range, returns NULL.
 TestCase* UnitTest::GetMutableTestCase(int i) {
@@ -5011,13 +5207,13 @@ Environment* UnitTest::AddEnvironment(Environment* env) {
 // assertion macros (e.g. ASSERT_TRUE, EXPECT_EQ, etc) eventually call
 // this to report their results.  The user code should use the
 // assertion macros instead of calling this directly.
+GTEST_LOCK_EXCLUDED_(mutex_)
 void UnitTest::AddTestPartResult(
     TestPartResult::Type result_type,
     const char* file_name,
     int line_number,
     const std::string& message,
-    const std::string& os_stack_trace)
-        GTEST_LOCK_EXCLUDED_(mutex_) {
+    const std::string& os_stack_trace) {
   Message msg;
   msg << message;
 
@@ -5064,7 +5260,7 @@ void UnitTest::AddTestPartResult(
 #endif  // GTEST_OS_WINDOWS
     } else if (GTEST_FLAG(throw_on_failure)) {
 #if GTEST_HAS_EXCEPTIONS
-      throw GoogleTestFailureException(result);
+      throw internal::GoogleTestFailureException(result);
 #else
       // We cannot call abort() as it generates a pop-up in debug mode
       // that cannot be suppressed in VC 7.1 or below.
@@ -5074,12 +5270,14 @@ void UnitTest::AddTestPartResult(
   }
 }
 
-// Creates and adds a property to the current TestResult. If a property matching
-// the supplied value already exists, updates its value instead.
-void UnitTest::RecordPropertyForCurrentTest(const char* key,
-                                            const char* value) {
-  const TestProperty test_property(key, value);
-  impl_->current_test_result()->RecordProperty(test_property);
+// Adds a TestProperty to the current TestResult object when invoked from
+// inside a test, to current TestCase's ad_hoc_test_result_ when invoked
+// from SetUpTestCase or TearDownTestCase, or to the global property set
+// when invoked elsewhere.  If the result already contains a property with
+// the same key, the value will be updated.
+void UnitTest::RecordProperty(const std::string& key,
+                              const std::string& value) {
+  impl_->RecordProperty(TestProperty(key, value));
 }
 
 // Runs all tests in this UnitTest object and prints the result.
@@ -5252,6 +5450,28 @@ UnitTestImpl::~UnitTestImpl() {
   delete os_stack_trace_getter_;
 }
 
+// Adds a TestProperty to the current TestResult object when invoked in a
+// context of a test, to current test case's ad_hoc_test_result when invoke
+// from SetUpTestCase/TearDownTestCase, or to the global property set
+// otherwise.  If the result already contains a property with the same key,
+// the value will be updated.
+void UnitTestImpl::RecordProperty(const TestProperty& test_property) {
+  std::string xml_element;
+  TestResult* test_result;  // TestResult appropriate for property recording.
+
+  if (current_test_info_ != NULL) {
+    xml_element = "testcase";
+    test_result = &(current_test_info_->result_);
+  } else if (current_test_case_ != NULL) {
+    xml_element = "testsuite";
+    test_result = &(current_test_case_->ad_hoc_test_result_);
+  } else {
+    xml_element = "testsuites";
+    test_result = &ad_hoc_test_result_;
+  }
+  test_result->RecordProperty(xml_element, test_property);
+}
+
 #if GTEST_HAS_DEATH_TEST
 // Disables event forwarding if the control is currently in a death test
 // subprocess. Must not be called before InitGoogleTest.
@@ -5671,9 +5891,15 @@ int UnitTestImpl::FilterTests(ReactionToSharding shard_tests) {
                                                    kDisableTestFilter);
       test_info->is_disabled_ = is_disabled;
 
+      const std::string value_param(test_info->value_param() == NULL ?
+                                    "" : test_info->value_param());
+
       const bool matches_filter =
           internal::UnitTestOptions::FilterMatchesTest(test_case_name,
-                                                       test_name);
+                                                       test_name) &&
+          internal::UnitTestOptions::MatchesFilter(value_param,
+                                                   GTEST_FLAG(param_filter).c_str());
+
       test_info->matches_filter_ = matches_filter;
 
       const bool is_runnable =
@@ -5695,8 +5921,33 @@ int UnitTestImpl::FilterTests(ReactionToSharding shard_tests) {
   return num_selected_tests;
 }
 
+// Prints the given C-string on a single line by replacing all '\n'
+// characters with string "\\n".  If the output takes more than
+// max_length characters, only prints the first max_length characters
+// and "...".
+static void PrintOnOneLine(const char* str, int max_length) {
+  if (str != NULL) {
+    for (int i = 0; *str != '\0'; ++str) {
+      if (i >= max_length) {
+        printf("...");
+        break;
+      }
+      if (*str == '\n') {
+        printf("\\n");
+        i += 2;
+      } else {
+        printf("%c", *str);
+        ++i;
+      }
+    }
+  }
+}
+
 // Prints the names of the tests matching the user-specified filter flag.
 void UnitTestImpl::ListTestsMatchingFilter() {
+  // Print at most this many characters for each type/value parameter.
+  const int kMaxParamLength = 250;
+
   for (size_t i = 0; i < test_cases_.size(); i++) {
     const TestCase* const test_case = test_cases_[i];
     bool printed_test_case_name = false;
@@ -5707,9 +5958,23 @@ void UnitTestImpl::ListTestsMatchingFilter() {
       if (test_info->matches_filter_) {
         if (!printed_test_case_name) {
           printed_test_case_name = true;
-          printf("%s.\n", test_case->name());
+          printf("%s.", test_case->name());
+          if (test_case->type_param() != NULL) {
+            printf("  # %s = ", kTypeParamLabel);
+            // We print the type parameter on a single line to make
+            // the output easy to parse by a program.
+            PrintOnOneLine(test_case->type_param(), kMaxParamLength);
         }
-        printf("  %s\n", test_info->name());
+          printf("\n");
+        }
+        printf("  %s", test_info->name());
+        if (test_info->value_param() != NULL) {
+          printf("  # %s = ", kValueParamLabel);
+          // We print the value parameter on a single line to make the
+          // output easy to parse by a program.
+          PrintOnOneLine(test_info->value_param(), kMaxParamLength);
+        }
+        printf("\n");
       }
     }
   }
@@ -5784,7 +6049,7 @@ void UnitTestImpl::UnshuffleTests() {
 // GetCurrentOsStackTraceExceptTop(..., 1), Foo() will be included in
 // the trace but Bar() and GetCurrentOsStackTraceExceptTop() won't.
 std::string GetCurrentOsStackTraceExceptTop(UnitTest* /*unit_test*/,
-                                       int skip_count) {
+                                            int skip_count) {
   // We pass skip_count + 1 to skip this wrapper function in addition
   // to what the user really wants to skip.
   return GetUnitTestImpl()->CurrentOsStackTraceExceptTop(skip_count + 1);
@@ -5982,6 +6247,12 @@ static const char kColorEncodedHelpMessage[] =
 "      Run only the tests whose name matches one of the positive patterns but\n"
 "      none of the negative patterns. '?' matches any single character; '*'\n"
 "      matches any substring; ':' separates two patterns.\n"
+"  @G--" GTEST_FLAG_PREFIX_ "param_filter=@YPOSITIVE_PATTERNS"
+    "[@G-@YNEGATIVE_PATTERNS]@D\n"
+"      Like @G--" GTEST_FLAG_PREFIX_
+                      "filter@D, but applies to the test's parameter. If a\n"
+"      test is not parameterized, its parameter is considered to be the\n"
+"      empty string.\n"
 "  @G--" GTEST_FLAG_PREFIX_ "also_run_disabled_tests@D\n"
 "      Run all disabled tests too.\n"
 "\n"
@@ -6059,6 +6330,7 @@ void ParseGoogleTestFlagsOnlyImpl(int* argc, CharType** argv) {
         ParseBoolFlag(arg, kDeathTestUseFork,
                       &GTEST_FLAG(death_test_use_fork)) ||
         ParseStringFlag(arg, kFilterFlag, &GTEST_FLAG(filter)) ||
+        ParseStringFlag(arg, kParamFilterFlag, &GTEST_FLAG(param_filter)) ||
         ParseStringFlag(arg, kInternalRunDeathTestFlag,
                         &GTEST_FLAG(internal_run_death_test)) ||
         ParseBoolFlag(arg, kListTestsFlag, &GTEST_FLAG(list_tests)) ||
@@ -6435,9 +6707,10 @@ static void DeathTestAbort(const std::string& message) {
 # define GTEST_DEATH_TEST_CHECK_(expression) \
   do { \
     if (!::testing::internal::IsTrue(expression)) { \
-      DeathTestAbort(::testing::internal::String::Format( \
-          "CHECK failed: File %s, line %d: %s", \
-          __FILE__, __LINE__, #expression)); \
+      DeathTestAbort( \
+          ::std::string("CHECK failed: File ") + __FILE__ +  ", line " \
+          + ::testing::internal::StreamableToString(__LINE__) + ": " \
+          + #expression); \
     } \
   } while (::testing::internal::AlwaysFalse())
 
@@ -6455,9 +6728,10 @@ static void DeathTestAbort(const std::string& message) {
       gtest_retval = (expression); \
     } while (gtest_retval == -1 && errno == EINTR); \
     if (gtest_retval == -1) { \
-      DeathTestAbort(::testing::internal::String::Format( \
-          "CHECK failed: File %s, line %d: %s != -1", \
-          __FILE__, __LINE__, #expression)); \
+      DeathTestAbort( \
+          ::std::string("CHECK failed: File ") + __FILE__ + ", line " \
+          + ::testing::internal::StreamableToString(__LINE__) + ": " \
+          + #expression + " != -1"); \
     } \
   } while (::testing::internal::AlwaysFalse())
 
@@ -6879,14 +7153,14 @@ DeathTest::TestRole WindowsDeathTest::AssumeRole() {
       info->test_case_name() + "." + info->name();
   const std::string internal_flag =
       std::string("--") + GTEST_FLAG_PREFIX_ + kInternalRunDeathTestFlag +
-      "=" + file_ + "|" + String::Format("%d|%d|%u|%Iu|%Iu", line_,
-      death_test_index,
-      static_cast<unsigned int>(::GetCurrentProcessId()),
-      // size_t has the same with as pointers on both 32-bit and 64-bit
+      "=" + file_ + "|" + StreamableToString(line_) + "|" +
+      StreamableToString(death_test_index) + "|" +
+      StreamableToString(static_cast<unsigned int>(::GetCurrentProcessId())) +
+      // size_t has the same width as pointers on both 32-bit and 64-bit
       // Windows platforms.
       // See http://msdn.microsoft.com/en-us/library/tcxf1dw6.aspx.
-      reinterpret_cast<size_t>(write_handle),
-      reinterpret_cast<size_t>(event_handle_.Get()));
+      "|" + StreamableToString(reinterpret_cast<size_t>(write_handle)) +
+      "|" + StreamableToString(reinterpret_cast<size_t>(event_handle_.Get()));
 
   char executable_path[_MAX_PATH + 1];  // NOLINT
   GTEST_DEATH_TEST_CHECK_(
@@ -7279,13 +7553,13 @@ DeathTest::TestRole ExecDeathTest::AssumeRole() {
   GTEST_DEATH_TEST_CHECK_(fcntl(pipe_fd[1], F_SETFD, 0) != -1);
 
   const std::string filter_flag =
-      String::Format("--%s%s=%s.%s",
-                     GTEST_FLAG_PREFIX_, kFilterFlag,
-                     info->test_case_name(), info->name());
+      std::string("--") + GTEST_FLAG_PREFIX_ + kFilterFlag + "="
+      + info->test_case_name() + "." + info->name();
   const std::string internal_flag =
-      String::Format("--%s%s=%s|%d|%d|%d",
-                     GTEST_FLAG_PREFIX_, kInternalRunDeathTestFlag,
-                     file_, line_, death_test_index, pipe_fd[1]);
+      std::string("--") + GTEST_FLAG_PREFIX_ + kInternalRunDeathTestFlag + "="
+      + file_ + "|" + StreamableToString(line_) + "|"
+      + StreamableToString(death_test_index) + "|"
+      + StreamableToString(pipe_fd[1]);
   Arguments args;
   args.AddArguments(GetArgvsForDeathTestChildProcess());
   args.AddArgument(filter_flag.c_str());
@@ -7324,9 +7598,10 @@ bool DefaultDeathTestFactory::Create(const char* statement, const RE* regex,
 
   if (flag != NULL) {
     if (death_test_index > flag->index()) {
-      DeathTest::set_last_death_test_message(String::Format(
-          "Death test count (%d) somehow exceeded expected maximum (%d)",
-          death_test_index, flag->index()));
+      DeathTest::set_last_death_test_message(
+          "Death test count (" + StreamableToString(death_test_index)
+          + ") somehow exceeded expected maximum ("
+          + StreamableToString(flag->index()) + ")");
       return false;
     }
 
@@ -7355,9 +7630,9 @@ bool DefaultDeathTestFactory::Create(const char* statement, const RE* regex,
 # endif  // GTEST_OS_WINDOWS
 
   else {  // NOLINT - this is more readable than unbalanced brackets inside #if.
-    DeathTest::set_last_death_test_message(String::Format(
-        "Unknown death test style \"%s\" encountered",
-        GTEST_FLAG(death_test_style).c_str()));
+    DeathTest::set_last_death_test_message(
+        "Unknown death test style \"" + GTEST_FLAG(death_test_style)
+        + "\" encountered");
     return false;
   }
 
@@ -7395,8 +7670,8 @@ int GetStatusFileDescriptor(unsigned int parent_process_id,
                                                    FALSE,  // Non-inheritable.
                                                    parent_process_id));
   if (parent_process_handle.Get() == INVALID_HANDLE_VALUE) {
-    DeathTestAbort(String::Format("Unable to open parent process %u",
-                                  parent_process_id));
+    DeathTestAbort("Unable to open parent process " +
+                   StreamableToString(parent_process_id));
   }
 
   // TODO(vladl@google.com): Replace the following check with a
@@ -7416,9 +7691,10 @@ int GetStatusFileDescriptor(unsigned int parent_process_id,
                                  // DUPLICATE_SAME_ACCESS is used.
                          FALSE,  // Request non-inheritable handler.
                          DUPLICATE_SAME_ACCESS)) {
-    DeathTestAbort(String::Format(
-        "Unable to duplicate the pipe handle %Iu from the parent process %u",
-        write_handle_as_size_t, parent_process_id));
+    DeathTestAbort("Unable to duplicate the pipe handle " +
+                   StreamableToString(write_handle_as_size_t) +
+                   " from the parent process " +
+                   StreamableToString(parent_process_id));
   }
 
   const HANDLE event_handle = reinterpret_cast<HANDLE>(event_handle_as_size_t);
@@ -7429,17 +7705,18 @@ int GetStatusFileDescriptor(unsigned int parent_process_id,
                          0x0,
                          FALSE,
                          DUPLICATE_SAME_ACCESS)) {
-    DeathTestAbort(String::Format(
-        "Unable to duplicate the event handle %Iu from the parent process %u",
-        event_handle_as_size_t, parent_process_id));
+    DeathTestAbort("Unable to duplicate the event handle " +
+                   StreamableToString(event_handle_as_size_t) +
+                   " from the parent process " +
+                   StreamableToString(parent_process_id));
   }
 
   const int write_fd =
       ::_open_osfhandle(reinterpret_cast<intptr_t>(dup_write_handle), O_APPEND);
   if (write_fd == -1) {
-    DeathTestAbort(String::Format(
-        "Unable to convert pipe handle %Iu to a file descriptor",
-        write_handle_as_size_t));
+    DeathTestAbort("Unable to convert pipe handle " +
+                   StreamableToString(write_handle_as_size_t) +
+                   " to a file descriptor");
   }
 
   // Signals the parent that the write end of the pipe has been acquired
@@ -7476,9 +7753,8 @@ InternalRunDeathTestFlag* ParseInternalRunDeathTestFlag() {
       || !ParseNaturalNumber(fields[3], &parent_process_id)
       || !ParseNaturalNumber(fields[4], &write_handle_as_size_t)
       || !ParseNaturalNumber(fields[5], &event_handle_as_size_t)) {
-    DeathTestAbort(String::Format(
-        "Bad --gtest_internal_run_death_test flag: %s",
-        GTEST_FLAG(internal_run_death_test).c_str()));
+    DeathTestAbort("Bad --gtest_internal_run_death_test flag: " +
+                   GTEST_FLAG(internal_run_death_test));
   }
   write_fd = GetStatusFileDescriptor(parent_process_id,
                                      write_handle_as_size_t,
@@ -7489,9 +7765,8 @@ InternalRunDeathTestFlag* ParseInternalRunDeathTestFlag() {
       || !ParseNaturalNumber(fields[1], &line)
       || !ParseNaturalNumber(fields[2], &index)
       || !ParseNaturalNumber(fields[3], &write_fd)) {
-    DeathTestAbort(String::Format(
-        "Bad --gtest_internal_run_death_test flag: %s",
-        GTEST_FLAG(internal_run_death_test).c_str()));
+    DeathTestAbort("Bad --gtest_internal_run_death_test flag: "
+        + GTEST_FLAG(internal_run_death_test));
   }
 
 # endif  // GTEST_OS_WINDOWS
@@ -7685,7 +7960,7 @@ FilePath FilePath::MakeFileName(const FilePath& directory,
   if (number == 0) {
     file = base_name.string() + "." + extension;
   } else {
-    file = base_name.string() + "_" + String::Format("%d", number).c_str()
+    file = base_name.string() + "_" + StreamableToString(number)
         + "." + extension;
   }
   return ConcatPaths(directory, FilePath(file));
@@ -8332,15 +8607,15 @@ const char kUnknownFile[] = "unknown file";
 // Formats a source file path and a line number as they would appear
 // in an error message from the compiler used to compile this code.
 GTEST_API_ ::std::string FormatFileLocation(const char* file, int line) {
-  const char* const file_name = file == NULL ? kUnknownFile : file;
+  const std::string file_name(file == NULL ? kUnknownFile : file);
 
   if (line < 0) {
-    return String::Format("%s:", file_name).c_str();
+    return file_name + ":";
   }
 #ifdef _MSC_VER
-  return String::Format("%s(%d):", file_name, line).c_str();
+  return file_name + "(" + StreamableToString(line) + "):";
 #else
-  return String::Format("%s:%d:", file_name, line).c_str();
+  return file_name + ":" + StreamableToString(line) + ":";
 #endif  // _MSC_VER
 }
 
@@ -8351,12 +8626,12 @@ GTEST_API_ ::std::string FormatFileLocation(const char* file, int line) {
 // to the file location it produces, unlike FormatFileLocation().
 GTEST_API_ ::std::string FormatCompilerIndependentFileLocation(
     const char* file, int line) {
-  const char* const file_name = file == NULL ? kUnknownFile : file;
+  const std::string file_name(file == NULL ? kUnknownFile : file);
 
   if (line < 0)
     return file_name;
   else
-    return String::Format("%s:%d", file_name, line).c_str();
+    return file_name + ":" + StreamableToString(line);
 }
 
 
@@ -8857,7 +9132,7 @@ static CharFormat PrintAsCharLiteralTo(Char c, ostream* os) {
         *os << static_cast<char>(c);
         return kAsIs;
       } else {
-        *os << String::Format("\\x%X", static_cast<UnsignedChar>(c));
+        *os << "\\x" + String::FormatHexInt(static_cast<UnsignedChar>(c));
         return kHexEscape;
       }
   }
@@ -8902,7 +9177,7 @@ void PrintCharAndCodeTo(Char c, ostream* os) {
   // obvious).
   if (c == 0)
     return;
-  *os << " (" << String::Format("%d", c).c_str();
+  *os << " (" << static_cast<int>(c);
 
   // For more convenience, we print c's code again in hexidecimal,
   // unless c was already printed in the form '\x##' or the code is in
@@ -8910,8 +9185,7 @@ void PrintCharAndCodeTo(Char c, ostream* os) {
   if (format == kHexEscape || (1 <= c && c <= 9)) {
     // Do nothing.
   } else {
-    *os << String::Format(", 0x%X",
-                          static_cast<UnsignedChar>(c)).c_str();
+    *os << ", 0x" << String::FormatHexInt(static_cast<UnsignedChar>(c));
   }
   *os << ")";
 }
diff --git a/modules/ts/src/ts_perf.cpp b/modules/ts/src/ts_perf.cpp
index f84953ca1d..e3ae8735d0 100644
--- a/modules/ts/src/ts_perf.cpp
+++ b/modules/ts/src/ts_perf.cpp
@@ -1,7 +1,7 @@
 #include "precomp.hpp"
 
 #ifdef HAVE_CUDA
-#include "opencv2/core/gpumat.hpp"
+#include "opencv2/core/gpu.hpp"
 #endif
 
 #ifdef ANDROID
@@ -14,30 +14,10 @@ int64 TestBase::timeLimitDefault = 0;
 unsigned int TestBase::iterationsLimitDefault = (unsigned int)(-1);
 int64 TestBase::_timeadjustment = 0;
 
-const std::string command_line_keys =
-    "{   perf_max_outliers   |8        |percent of allowed outliers}"
-    "{   perf_min_samples    |10       |minimal required numer of samples}"
-    "{   perf_force_samples  |100      |force set maximum number of samples for all tests}"
-    "{   perf_seed           |809564   |seed for random numbers generator}"
-    "{   perf_threads        |-1       |the number of worker threads, if parallel execution is enabled}"
-    "{   perf_write_sanity   |         |create new records for sanity checks}"
-    "{   perf_verify_sanity  |         |fail tests having no regression data for sanity checks}"
-#ifdef ANDROID
-    "{   perf_time_limit     |6.0      |default time limit for a single test (in seconds)}"
-    "{   perf_affinity_mask  |0        |set affinity mask for the main thread}"
-    "{   perf_log_power_checkpoints  | |additional xml logging for power measurement}"
-#else
-    "{   perf_time_limit     |3.0      |default time limit for a single test (in seconds)}"
-#endif
-    "{   perf_max_deviation  |1.0      |}"
-    "{   help h              |         |print help info}"
-#ifdef HAVE_CUDA
-    "{   perf_run_cpu        |false    |run GPU performance tests for analogical CPU functions}"
-    "{   perf_cuda_device    |0        |run GPU test suite onto specific CUDA capable device}"
-    "{   perf_cuda_info_only |false    |print an information about system and an available CUDA devices and then exit.}"
-#endif
-;
+// Item [0] will be considered the default implementation.
+static std::vector<std::string> available_impls;
 
+static std::string  param_impl;
 static double       param_max_outliers;
 static double       param_max_deviation;
 static unsigned int param_min_samples;
@@ -48,7 +28,6 @@ static int          param_threads;
 static bool         param_write_sanity;
 static bool         param_verify_sanity;
 #ifdef HAVE_CUDA
-static bool         param_run_cpu;
 static int          param_cuda_device;
 #endif
 
@@ -72,10 +51,6 @@ static void setCurrentThreadAffinityMask(int mask)
 }
 #endif
 
-#ifdef HAVE_CUDA
-# include <opencv2/core/gpumat.hpp>
-#endif
-
 namespace {
 
 class PerfEnvironment: public ::testing::Environment
@@ -577,12 +552,6 @@ Regression& Regression::operator() (const std::string& name, cv::InputArray arra
 
     std::string nodename = getCurrentTestNodeName();
 
-#ifdef HAVE_CUDA
-    static const std::string prefix = (param_run_cpu)? "CPU_" : "GPU_";
-    if(suiteName == "gpu")
-        nodename = prefix + nodename;
-#endif
-
     cv::FileNode n = rootIn[nodename];
     if(n.isNone())
     {
@@ -646,6 +615,43 @@ performance_metrics::performance_metrics()
 
 void TestBase::Init(int argc, const char* const argv[])
 {
+    std::vector<std::string> plain_only;
+    plain_only.push_back("plain");
+    TestBase::Init(plain_only, argc, argv);
+}
+
+void TestBase::Init(const std::vector<std::string> & availableImpls,
+                 int argc, const char* const argv[])
+{
+    available_impls = availableImpls;
+
+    const std::string command_line_keys =
+        "{   perf_max_outliers           |8        |percent of allowed outliers}"
+        "{   perf_min_samples            |10       |minimal required numer of samples}"
+        "{   perf_force_samples          |100      |force set maximum number of samples for all tests}"
+        "{   perf_seed                   |809564   |seed for random numbers generator}"
+        "{   perf_threads                |-1       |the number of worker threads, if parallel execution is enabled}"
+        "{   perf_write_sanity           |false    |create new records for sanity checks}"
+        "{   perf_verify_sanity          |false    |fail tests having no regression data for sanity checks}"
+        "{   perf_impl                   |" + available_impls[0] +
+                                                  "|the implementation variant of functions under test}"
+        "{   perf_list_impls             |false    |list available implementation variants and exit}"
+        "{   perf_run_cpu                |false    |deprecated, equivalent to --perf_impl=plain}"
+#ifdef ANDROID
+        "{   perf_time_limit             |6.0      |default time limit for a single test (in seconds)}"
+        "{   perf_affinity_mask          |0        |set affinity mask for the main thread}"
+        "{   perf_log_power_checkpoints  |         |additional xml logging for power measurement}"
+#else
+        "{   perf_time_limit             |3.0      |default time limit for a single test (in seconds)}"
+#endif
+        "{   perf_max_deviation          |1.0      |}"
+        "{   help h                      |false    |print help info}"
+#ifdef HAVE_CUDA
+        "{   perf_cuda_device            |0        |run GPU test suite onto specific CUDA capable device}"
+        "{   perf_cuda_info_only         |false    |print an information about system and an available CUDA devices and then exit.}"
+#endif
+    ;
+
     cv::CommandLineParser args(argc, argv, command_line_keys);
     if (args.has("help"))
     {
@@ -655,6 +661,7 @@ void TestBase::Init(int argc, const char* const argv[])
 
     ::testing::AddGlobalTestEnvironment(new PerfEnvironment);
 
+    param_impl          = args.has("perf_run_cpu") ? "plain" : args.get<std::string>("perf_impl");
     param_max_outliers  = std::min(100., std::max(0., args.get<double>("perf_max_outliers")));
     param_min_samples   = std::max(1u, args.get<unsigned int>("perf_min_samples"));
     param_max_deviation = std::max(0., args.get<double>("perf_max_deviation"));
@@ -669,30 +676,52 @@ void TestBase::Init(int argc, const char* const argv[])
     log_power_checkpoints = args.has("perf_log_power_checkpoints");
 #endif
 
+    bool param_list_impls = args.has("perf_list_impls");
+
+    if (param_list_impls)
+    {
+        fputs("Available implementation variants:", stdout);
+        for (size_t i = 0; i < available_impls.size(); ++i) {
+            putchar(' ');
+            fputs(available_impls[i].c_str(), stdout);
+        }
+        putchar('\n');
+        exit(0);
+    }
+
+    if (std::find(available_impls.begin(), available_impls.end(), param_impl) == available_impls.end())
+    {
+        printf("No such implementation: %s\n", param_impl.c_str());
+        exit(1);
+    }
+
 #ifdef HAVE_CUDA
 
     bool printOnly        = args.has("perf_cuda_info_only");
 
     if (printOnly)
         exit(0);
+#endif
+
+    if (available_impls.size() > 1)
+        printf("[----------]\n[   INFO   ] \tImplementation variant: %s.\n[----------]\n", param_impl.c_str()), fflush(stdout);
+
+#ifdef HAVE_CUDA
 
-    param_run_cpu         = args.has("perf_run_cpu");
     param_cuda_device      = std::max(0, std::min(cv::gpu::getCudaEnabledDeviceCount(), args.get<int>("perf_cuda_device")));
 
-    if (param_run_cpu)
-        printf("[----------]\n[ GPU INFO ] \tRun test suite on CPU.\n[----------]\n"), fflush(stdout);
-    else
+    if (param_impl == "cuda")
     {
         cv::gpu::DeviceInfo info(param_cuda_device);
         if (!info.isCompatible())
         {
-            printf("[----------]\n[ FAILURE  ] \tDevice %s is NOT compatible with current GPU module build.\n[----------]\n", info.name().c_str()), fflush(stdout);
+            printf("[----------]\n[ FAILURE  ] \tDevice %s is NOT compatible with current GPU module build.\n[----------]\n", info.name()), fflush(stdout);
             exit(-1);
         }
 
         cv::gpu::setDevice(param_cuda_device);
 
-        printf("[----------]\n[ GPU INFO ] \tRun test suite on %s GPU.\n[----------]\n", info.name().c_str()), fflush(stdout);
+        printf("[----------]\n[ GPU INFO ] \tRun test suite on %s GPU.\n[----------]\n", info.name()), fflush(stdout);
     }
 #endif
 
@@ -707,6 +736,26 @@ void TestBase::Init(int argc, const char* const argv[])
     _timeadjustment = _calibrate();
 }
 
+void TestBase::RecordRunParameters()
+{
+    ::testing::Test::RecordProperty("cv_implementation", param_impl);
+    ::testing::Test::RecordProperty("cv_num_threads", param_threads);
+
+#ifdef HAVE_CUDA
+    if (param_impl == "cuda")
+    {
+        cv::gpu::DeviceInfo info(param_cuda_device);
+        ::testing::Test::RecordProperty("cv_cuda_gpu", info.name());
+    }
+#endif
+}
+
+std::string TestBase::getSelectedImpl()
+{
+    return param_impl;
+}
+
+
 int64 TestBase::_calibrate()
 {
     class _helper : public ::perf::TestBase
@@ -1153,7 +1202,7 @@ void TestBase::RunPerfTestBody()
     {
         metrics.terminationReason = performance_metrics::TERM_EXCEPTION;
         #ifdef HAVE_CUDA
-            if (e.code == CV_GpuApiCallError)
+            if (e.code == cv::Error::GpuApiCallError)
                 cv::gpu::resetDevice();
         #endif
         FAIL() << "Expected: PerfTestBody() doesn't throw an exception.\n  Actual: it throws cv::Exception:\n  " << e.what();
@@ -1324,12 +1373,10 @@ void perf::sort(std::vector<cv::KeyPoint>& pts, cv::InputOutputArray descriptors
 /*****************************************************************************************\
 *                                  ::perf::GpuPerf
 \*****************************************************************************************/
-#ifdef HAVE_CUDA
 bool perf::GpuPerf::targetDevice()
 {
-    return !param_run_cpu;
+    return param_impl == "cuda";
 }
-#endif
 
 /*****************************************************************************************\
 *                                  ::perf::PrintTo
@@ -1367,4 +1414,3 @@ void PrintTo(const Size& sz, ::std::ostream* os)
 }
 
 }  // namespace cv
-
diff --git a/modules/video/doc/motion_analysis_and_object_tracking.rst b/modules/video/doc/motion_analysis_and_object_tracking.rst
index a193c23a14..b392c68544 100644
--- a/modules/video/doc/motion_analysis_and_object_tracking.rst
+++ b/modules/video/doc/motion_analysis_and_object_tracking.rst
@@ -13,7 +13,6 @@ Calculates an optical flow for a sparse feature set using the iterative Lucas-Ka
 .. ocv:pyfunction:: cv2.calcOpticalFlowPyrLK(prevImg, nextImg, prevPts, nextPts[, status[, err[, winSize[, maxLevel[, criteria[, flags[, minEigThreshold]]]]]]]) -> nextPts, status, err
 
 .. ocv:cfunction:: void cvCalcOpticalFlowPyrLK( const CvArr* prev, const CvArr* curr, CvArr* prev_pyr, CvArr* curr_pyr, const CvPoint2D32f* prev_features, CvPoint2D32f* curr_features, int count, CvSize win_size, int level, char* status, float* track_error, CvTermCriteria criteria, int flags )
-.. ocv:pyoldfunction:: cv.CalcOpticalFlowPyrLK(prev, curr, prevPyr, currPyr, prevFeatures, winSize, level, criteria, flags, guesses=None) -> (currFeatures, status, track_error)
 
     :param prevImg: first 8-bit input image or pyramid constructed by :ocv:func:`buildOpticalFlowPyramid`.
 
@@ -42,6 +41,13 @@ Calculates an optical flow for a sparse feature set using the iterative Lucas-Ka
 
 The function implements a sparse iterative version of the Lucas-Kanade optical flow in pyramids. See [Bouguet00]_. The function is parallelized with the TBB library.
 
+.. note::
+
+   * An example using the Lucas-Kanade optical flow algorithm can be found at opencv_source_code/samples/cpp/lkdemo.cpp
+
+   * (Python) An example using the Lucas-Kanade optical flow algorithm can be found at opencv_source_code/samples/python2/lk_track.py
+   * (Python) An example using the Lucas-Kanade tracker for homography matching can be found at opencv_source_code/samples/python2/lk_homography.py
+
 buildOpticalFlowPyramid
 -----------------------
 Constructs the image pyramid which can be passed to :ocv:func:`calcOpticalFlowPyrLK`.
@@ -109,6 +115,11 @@ The function finds an optical flow for each ``prev`` pixel using the [Farneback2
 
     \texttt{prev} (y,x)  \sim \texttt{next} ( y + \texttt{flow} (y,x)[1],  x + \texttt{flow} (y,x)[0])
 
+.. note::
+
+   * An example using the optical flow algorithm described by Gunnar Farneback can be found at opencv_source_code/samples/cpp/fback.cpp
+
+   * (Python) An example using the optical flow algorithm described by Gunnar Farneback can be found at opencv_source_code/samples/python2/opt_flow.py
 
 estimateRigidTransform
 --------------------------
@@ -158,7 +169,8 @@ findTransformECC
 Finds the geometric transform (warp) between two images in terms of the ECC criterion [EP08]_.
 
 .. ocv:function:: double findTransformECC( InputArray templateImage, InputArray inputImage, InputOutputArray warpMatrix, int motionType=MOTION_AFFINE, TermCriteria criteria=TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 50, 0.001))
-.. ocv:cfunction:: double cvFindTransformECC( const CvArr* templateImage, const CvArr* inputImage, CvMat* warpMatrix, const int motionType, const CvTermCriteria criteria)
+
+.. ocv:pyfunction:: cv2.findTransformECC(templateImage, inputImage, warpMatrix[, motionType[, criteria]]) -> retval, warpMatrix
 
     :param templateImage: single-channel template image; ``CV_8U`` or ``CV_32F`` array.
 
@@ -167,9 +179,13 @@ Finds the geometric transform (warp) between two images in terms of the ECC crit
     :param warpMatrix: floating-point :math:`2\times 3` or :math:`3\times 3` mapping matrix (warp).
 
     :param motionType: parameter, specifying the type of motion:
+
         * **MOTION_TRANSLATION** sets a translational motion model; ``warpMatrix`` is :math:`2\times 3` with the first :math:`2\times 2` part being the unity matrix and the rest two parameters being estimated.
+
         * **MOTION_EUCLIDEAN** sets a Euclidean (rigid) transformation as motion model; three parameters are estimated; ``warpMatrix`` is :math:`2\times 3`.
+
         * **MOTION_AFFINE** sets an affine motion model (DEFAULT); six parameters are estimated; ``warpMatrix`` is :math:`2\times 3`.
+
         * **MOTION_HOMOGRAPHY** sets a homography as a motion model; eight parameters are estimated;``warpMatrix`` is :math:`3\times 3`.
 
     :param criteria: parameter, specifying the termination criteria of the ECC algorithm; ``criteria.epsilon`` defines the threshold of the increment in the correlation coefficient between two iterations (a negative ``criteria.epsilon`` makes ``criteria.maxcount`` the only termination criterion). Default values are shown in the declaration above.
@@ -177,13 +193,13 @@ Finds the geometric transform (warp) between two images in terms of the ECC crit
 
 The function estimates the optimum transformation (``warpMatrix``) with respect to ECC criterion ([EP08]_), that is
 
-..math::
+.. math::
 
     \texttt{warpMatrix} = \texttt{warpMatrix} = \arg\max_{W} \texttt{ECC}(\texttt{templateImage}(x,y),\texttt{inputImage}(x',y'))
 
 where
 
-..math::
+.. math::
 
     \begin{bmatrix} x' \\ y' \end{bmatrix} = W \cdot \begin{bmatrix} x \\ y \\ 1 \end{bmatrix}
 
@@ -207,7 +223,6 @@ Updates the motion history image by a moving silhouette.
 .. ocv:pyfunction:: cv2.updateMotionHistory(silhouette, mhi, timestamp, duration) -> mhi
 
 .. ocv:cfunction:: void cvUpdateMotionHistory( const CvArr* silhouette, CvArr* mhi, double timestamp, double duration )
-.. ocv:pyoldfunction:: cv.UpdateMotionHistory(silhouette, mhi, timestamp, duration)-> None
 
     :param silhouette: Silhouette mask that has non-zero pixels where the motion occurs.
 
@@ -241,7 +256,6 @@ Calculates a gradient orientation of a motion history image.
 .. ocv:pyfunction:: cv2.calcMotionGradient(mhi, delta1, delta2[, mask[, orientation[, apertureSize]]]) -> mask, orientation
 
 .. ocv:cfunction:: void cvCalcMotionGradient( const CvArr* mhi, CvArr* mask, CvArr* orientation, double delta1, double delta2, int aperture_size=3 )
-.. ocv:pyoldfunction:: cv.CalcMotionGradient(mhi, mask, orientation, delta1, delta2, apertureSize=3)-> None
 
     :param mhi: Motion history single-channel floating-point image.
 
@@ -270,7 +284,9 @@ In fact,
 :ocv:func:`fastAtan2` and
 :ocv:func:`phase` are used so that the computed angle is measured in degrees and covers the full range 0..360. Also, the ``mask`` is filled to indicate pixels where the computed angle is valid.
 
+.. note::
 
+   * (Python) An example on how to perform a motion template technique can be found at opencv_source_code/samples/python2/motempl.py
 
 calcGlobalOrientation
 -------------------------
@@ -281,7 +297,6 @@ Calculates a global motion orientation in a selected region.
 .. ocv:pyfunction:: cv2.calcGlobalOrientation(orientation, mask, mhi, timestamp, duration) -> retval
 
 .. ocv:cfunction:: double cvCalcGlobalOrientation( const CvArr* orientation, const CvArr* mask, const CvArr* mhi, double timestamp, double duration )
-.. ocv:pyoldfunction:: cv.CalcGlobalOrientation(orientation, mask, mhi, timestamp, duration)-> float
 
     :param orientation: Motion gradient orientation image calculated by the function  :ocv:func:`calcMotionGradient` .
 
@@ -311,7 +326,6 @@ Splits a motion history image into a few parts corresponding to separate indepen
 .. ocv:pyfunction:: cv2.segmentMotion(mhi, timestamp, segThresh[, segmask]) -> segmask, boundingRects
 
 .. ocv:cfunction:: CvSeq* cvSegmentMotion( const CvArr* mhi, CvArr* seg_mask, CvMemStorage* storage, double timestamp, double seg_thresh )
-.. ocv:pyoldfunction:: cv.SegmentMotion(mhi, seg_mask, storage, timestamp, seg_thresh) -> boundingRects
 
     :param mhi: Motion history image.
 
@@ -339,8 +353,6 @@ Finds an object center, size, and orientation.
 
 .. ocv:cfunction:: int cvCamShift( const CvArr* prob_image, CvRect window, CvTermCriteria criteria, CvConnectedComp* comp, CvBox2D* box=NULL )
 
-.. ocv:pyoldfunction:: cv.CamShift(prob_image, window, criteria) -> (int, comp, box)
-
     :param probImage: Back projection of the object histogram. See  :ocv:func:`calcBackProject` .
 
     :param window: Initial search window.
@@ -356,7 +368,9 @@ First, it finds an object center using
 
 See the OpenCV sample ``camshiftdemo.c`` that tracks colored objects.
 
+.. note::
 
+   * (Python) A sample explaining the camshift tracking algorithm can be found at opencv_source_code/samples/python2/camshift.py
 
 meanShift
 ---------
@@ -367,7 +381,6 @@ Finds an object on a back projection image.
 .. ocv:pyfunction:: cv2.meanShift(probImage, window, criteria) -> retval, window
 
 .. ocv:cfunction:: int cvMeanShift( const CvArr* prob_image, CvRect window, CvTermCriteria criteria, CvConnectedComp* comp )
-.. ocv:pyoldfunction:: cv.MeanShift(prob_image, window, criteria) -> comp
 
     :param probImage: Back projection of the object histogram. See  :ocv:func:`calcBackProject` for details.
 
@@ -385,7 +398,9 @@ The function implements the iterative object search algorithm. It takes the inpu
 :ocv:func:`contourArea` ), and rendering the  remaining contours with
 :ocv:func:`drawContours` .
 
+.. note::
 
+   * A mean-shift tracking sample can be found at opencv_source_code/samples/cpp/camshiftdemo.cpp
 
 KalmanFilter
 ------------
@@ -396,7 +411,9 @@ KalmanFilter
 The class implements a standard Kalman filter
 http://en.wikipedia.org/wiki/Kalman_filter, [Welch95]_. However, you can modify ``transitionMatrix``, ``controlMatrix``, and ``measurementMatrix`` to get an extended Kalman filter functionality. See the OpenCV sample ``kalman.cpp`` .
 
+.. note::
 
+   * An example using the standard Kalman filter can be found at opencv_source_code/samples/cpp/kalman.cpp
 
 
 KalmanFilter::KalmanFilter
@@ -410,7 +427,6 @@ The constructors.
 .. ocv:pyfunction:: cv2.KalmanFilter([dynamParams, measureParams[, controlParams[, type]]]) -> <KalmanFilter object>
 
 .. ocv:cfunction:: CvKalman* cvCreateKalman( int dynam_params, int measure_params, int control_params=0 )
-.. ocv:pyoldfunction:: cv.CreateKalman(dynam_params, measure_params, control_params=0) -> CvKalman
 
     The full constructor.
 
@@ -448,7 +464,6 @@ Computes a predicted state.
 .. ocv:pyfunction:: cv2.KalmanFilter.predict([control]) -> retval
 
 .. ocv:cfunction:: const CvMat* cvKalmanPredict( CvKalman* kalman, const CvMat* control=NULL)
-.. ocv:pyoldfunction:: cv.KalmanPredict(kalman, control=None) -> mat
 
     :param control: The optional input control
 
@@ -463,8 +478,6 @@ Updates the predicted state from the measurement.
 
 .. ocv:cfunction:: const CvMat* cvKalmanCorrect( CvKalman* kalman, const CvMat* measurement )
 
-.. ocv:pyoldfunction:: cv.KalmanCorrect(kalman, measurement) -> mat
-
     :param measurement: The measured system parameters
 
 
@@ -479,7 +492,7 @@ Base class for background/foreground segmentation. ::
     {
     public:
         virtual ~BackgroundSubtractor();
-        virtual void operator()(InputArray image, OutputArray fgmask, double learningRate=0);
+        virtual void apply(InputArray image, OutputArray fgmask, double learningRate=0);
         virtual void getBackgroundImage(OutputArray backgroundImage) const;
     };
 
@@ -487,11 +500,11 @@ Base class for background/foreground segmentation. ::
 The class is only used to define the common interface for the whole family of background/foreground segmentation algorithms.
 
 
-BackgroundSubtractor::operator()
+BackgroundSubtractor::apply
 --------------------------------
 Computes a foreground mask.
 
-.. ocv:function:: void BackgroundSubtractor::operator()(InputArray image, OutputArray fgmask, double learningRate=0)
+.. ocv:function:: void BackgroundSubtractor::apply(InputArray image, OutputArray fgmask, double learningRate=-1)
 
 .. ocv:pyfunction:: cv2.BackgroundSubtractor.apply(image[, fgmask[, learningRate]]) -> fgmask
 
@@ -499,6 +512,7 @@ Computes a foreground mask.
 
     :param fgmask: The output foreground mask as an 8-bit binary image.
 
+    :param learningRate: The value between 0 and 1 that indicates how fast the background model is learnt. Negative parameter value makes the algorithm to use some automatically chosen learning rate. 0 means that the background model is not updated at all, 1 means that the background model is completely reinitialized from the last frame.
 
 BackgroundSubtractor::getBackgroundImage
 ----------------------------------------
@@ -517,20 +531,16 @@ BackgroundSubtractorMOG
 
 Gaussian Mixture-based Background/Foreground Segmentation Algorithm.
 
-The class implements the algorithm described in P. KadewTraKuPong and R. Bowden, *An improved adaptive background mixture model for real-time tracking with shadow detection*, Proc. 2nd European Workshop on Advanced Video-Based Surveillance Systems, 2001: http://personal.ee.surrey.ac.uk/Personal/R.Bowden/publications/avbs01/avbs01.pdf
+The class implements the algorithm described in [KB2001]_.
 
 
-
-
-BackgroundSubtractorMOG::BackgroundSubtractorMOG
+createBackgroundSubtractorMOG
 ------------------------------------------------
-The constructors.
+Creates mixture-of-gaussian background subtractor
 
-.. ocv:function:: BackgroundSubtractorMOG::BackgroundSubtractorMOG()
+.. ocv:function:: Ptr<BackgroundSubtractorMOG> createBackgroundSubtractorMOG(int history=200, int nmixtures=5, double backgroundRatio=0.7, double noiseSigma=0)
 
-.. ocv:function:: BackgroundSubtractorMOG::BackgroundSubtractorMOG(int history, int nmixtures, double backgroundRatio, double noiseSigma=0)
-
-.. ocv:pyfunction:: cv2.BackgroundSubtractorMOG([history, nmixtures, backgroundRatio[, noiseSigma]]) -> <BackgroundSubtractorMOG object>
+.. ocv:pyfunction:: cv2.createBackgroundSubtractorMOG([history[, nmixtures[, backgroundRatio[, noiseSigma]]]]) -> retval
 
     :param history: Length of the history.
 
@@ -538,20 +548,7 @@ The constructors.
 
     :param backgroundRatio: Background ratio.
 
-    :param noiseSigma: Noise strength.
-
-Default constructor sets all parameters to default values.
-
-
-
-
-BackgroundSubtractorMOG::operator()
------------------------------------
-Updates the background model and returns the foreground mask
-
-.. ocv:function:: void BackgroundSubtractorMOG::operator()(InputArray image, OutputArray fgmask, double learningRate=0)
-
-Parameters are the same as in :ocv:funcx:`BackgroundSubtractor::operator()`
+    :param noiseSigma: Noise strength (standard deviation of the brightness or each color channel). 0 means some automatic value.
 
 
 BackgroundSubtractorMOG2
@@ -560,84 +557,307 @@ Gaussian Mixture-based Background/Foreground Segmentation Algorithm.
 
 .. ocv:class:: BackgroundSubtractorMOG2 : public BackgroundSubtractor
 
-    Here are important members of the class that control the algorithm, which you can set after constructing the class instance:
-
-    .. ocv:member:: int nmixtures
-
-        Maximum allowed number of mixture components. Actual number is determined dynamically per pixel.
-
-    .. ocv:member:: float backgroundRatio
-
-        Threshold defining whether the component is significant enough to be included into the background model ( corresponds to ``TB=1-cf`` from the paper??which paper??). ``cf=0.1 => TB=0.9`` is default. For ``alpha=0.001``, it means that the mode should exist for approximately 105 frames before it is considered foreground.
-
-    .. ocv:member:: float varThresholdGen
-
-        Threshold for the squared Mahalanobis distance that helps decide when a sample is close to the existing components (corresponds to ``Tg``). If it is not close to any component, a new component is generated. ``3 sigma => Tg=3*3=9`` is default. A smaller ``Tg`` value generates more components. A higher ``Tg`` value may result in a small number of components but they can grow too large.
-
-    .. ocv:member:: float fVarInit
-
-        Initial variance for the newly generated components. It affects the speed of adaptation. The parameter value is based on your estimate of the typical standard deviation from the images. OpenCV uses 15 as a reasonable value.
-
-    .. ocv:member:: float fVarMin
-
-        Parameter used to further control the variance.
-
-    .. ocv:member:: float fVarMax
-
-        Parameter used to further control the variance.
-
-    .. ocv:member:: float fCT
-
-        Complexity reduction parameter. This parameter defines the number of samples needed to accept to prove the component exists. ``CT=0.05`` is a default value for all the samples. By setting ``CT=0`` you get an algorithm very similar to the standard Stauffer&Grimson algorithm.
-
-    .. ocv:member:: uchar nShadowDetection
-
-        The value for marking shadow pixels in the output foreground mask. Default value is 127.
-
-    .. ocv:member:: float fTau
-
-        Shadow threshold. The shadow is detected if the pixel is a darker version of the background. ``Tau`` is a threshold defining how much darker the shadow can be. ``Tau= 0.5`` means that if a pixel is more than twice darker then it is not shadow. See Prati,Mikic,Trivedi,Cucchiarra, *Detecting Moving Shadows...*, IEEE PAMI,2003.
+The class implements the Gaussian mixture model background subtraction described in [Zivkovic2004]_ and [Zivkovic2006]_ .
 
 
-The class implements the Gaussian mixture model background subtraction described in:
-
-  * Z.Zivkovic, *Improved adaptive Gausian mixture model for background subtraction*, International Conference Pattern Recognition, UK, August, 2004, http://www.zoranz.net/Publications/zivkovic2004ICPR.pdf. The code is very fast and performs also shadow detection. Number of Gausssian components is adapted per pixel.
-
-  * Z.Zivkovic, F. van der Heijden, *Efficient Adaptive Density Estimapion per Image Pixel for the Task of Background Subtraction*, Pattern Recognition Letters, vol. 27, no. 7, pages 773-780, 2006. The algorithm similar to the standard Stauffer&Grimson algorithm with additional selection of the number of the Gaussian components based on: Z.Zivkovic, F.van der Heijden, Recursive unsupervised learning of finite mixture models, IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.26, no.5, pages 651-656, 2004.
-
-
-BackgroundSubtractorMOG2::BackgroundSubtractorMOG2
+createBackgroundSubtractorMOG2
 --------------------------------------------------
-The constructors.
+Creates MOG2 Background Subtractor
 
-.. ocv:function:: BackgroundSubtractorMOG2::BackgroundSubtractorMOG2()
+.. ocv:function:: Ptr<BackgroundSubtractorMOG2> createBackgroundSubtractorMOG2( int history=500, double varThreshold=16, bool detectShadows=true )
 
-.. ocv:function:: BackgroundSubtractorMOG2::BackgroundSubtractorMOG2( int history, float varThreshold, bool bShadowDetection=true )
+  :param history: Length of the history.
 
-    :param history: Length of the history.
+  :param varThreshold: Threshold on the squared Mahalanobis distance between the pixel and the model to decide whether a pixel is well described by the background model. This parameter does not affect the background update.
 
-    :param varThreshold: Threshold on the squared Mahalanobis distance to decide whether it is well described by the background model (see Cthr??). This parameter does not affect the background update. A typical value could be 4 sigma, that is, ``varThreshold=4*4=16;`` (see Tb??).
-
-    :param bShadowDetection: Parameter defining whether shadow detection should be enabled (``true`` or ``false``).
+  :param detectShadows: If true, the algorithm will detect shadows and mark them. It decreases the speed a bit, so if you do not need this feature, set the parameter to false.
 
 
+BackgroundSubtractorMOG2::getHistory
+--------------------------------------
+Returns the number of last frames that affect the background model
 
-BackgroundSubtractorMOG2::operator()
-------------------------------------
-Updates the background model and computes the foreground mask
-
-.. ocv:function:: void BackgroundSubtractorMOG2::operator()(InputArray image, OutputArray fgmask, double learningRate=-1)
-
-    See :ocv:funcx:`BackgroundSubtractor::operator()`.
+.. ocv:function:: int BackgroundSubtractorMOG2::getHistory() const
 
 
-BackgroundSubtractorMOG2::getBackgroundImage
---------------------------------------------
-Returns background image
+BackgroundSubtractorMOG2::setHistory
+--------------------------------------
+Sets the number of last frames that affect the background model
 
-.. ocv:function:: void BackgroundSubtractorMOG2::getBackgroundImage(OutputArray backgroundImage)
+.. ocv:function:: void BackgroundSubtractorMOG2::setHistory(int history)
 
-See :ocv:func:`BackgroundSubtractor::getBackgroundImage`.
+
+BackgroundSubtractorMOG2::getNMixtures
+--------------------------------------
+Returns the number of gaussian components in the background model
+
+.. ocv:function:: int BackgroundSubtractorMOG2::getNMixtures() const
+
+
+BackgroundSubtractorMOG2::setNMixtures
+--------------------------------------
+Sets the number of gaussian components in the background model
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setNMixtures(int nmixtures)
+
+
+BackgroundSubtractorMOG2::getBackgroundRatio
+---------------------------------------------
+Returns the "background ratio" parameter of the algorithm
+
+.. ocv:function:: double BackgroundSubtractorMOG2::getBackgroundRatio() const
+
+If a foreground pixel keeps semi-constant value for about ``backgroundRatio*history`` frames, it's considered background and added to the model as a center of a new component. It corresponds to ``TB`` parameter in the paper.
+
+BackgroundSubtractorMOG2::setBackgroundRatio
+---------------------------------------------
+Sets the "background ratio" parameter of the algorithm
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setBackgroundRatio(double ratio)
+
+BackgroundSubtractorMOG2::getVarThresholdGen
+---------------------------------------------
+Returns the variance scale factor for the pixel-model match
+
+.. ocv:function:: double BackgroundSubtractorMOG2::getVarThresholdGen() const
+
+Threshold for the squared Mahalanobis distance that helps decide when a sample is close to the existing components (corresponds to ``Tg`` in the paper). If a pixel is not close to any component, it is considered foreground or added as a new component. ``3 sigma => Tg=3*3=9`` is default. A smaller ``Tg`` value generates more components. A higher ``Tg`` value may result in a small number of components but they can grow too large.
+
+BackgroundSubtractorMOG2::setVarThresholdGen
+---------------------------------------------
+Sets the variance scale factor for the pixel-model match
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setVarThresholdGen(double varThresholdGen)
+
+BackgroundSubtractorMOG2::getVarInit
+---------------------------------------------
+Returns the initial variance of each gaussian component
+
+.. ocv:function:: double BackgroundSubtractorMOG2::getVarInit() const
+
+BackgroundSubtractorMOG2::setVarInit
+---------------------------------------------
+Sets the initial variance of each gaussian component
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setVarInit(double varInit)
+
+
+BackgroundSubtractorMOG2::getComplexityReductionThreshold
+----------------------------------------------------------
+Returns the complexity reduction threshold
+
+.. ocv:function:: double BackgroundSubtractorMOG2::getComplexityReductionThreshold() const
+
+This parameter defines the number of samples needed to accept to prove the component exists. ``CT=0.05`` is a default value for all the samples. By setting ``CT=0`` you get an algorithm very similar to the standard Stauffer&Grimson algorithm.
+
+BackgroundSubtractorMOG2::setComplexityReductionThreshold
+----------------------------------------------------------
+Sets the complexity reduction threshold
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setComplexityReductionThreshold(double ct)
+
+
+BackgroundSubtractorMOG2::getDetectShadows
+---------------------------------------------
+Returns the shadow detection flag
+
+.. ocv:function:: bool BackgroundSubtractorMOG2::getDetectShadows() const
+
+If true, the algorithm detects shadows and marks them. See createBackgroundSubtractorMOG2 for details.
+
+BackgroundSubtractorMOG2::setDetectShadows
+---------------------------------------------
+Enables or disables shadow detection
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setDetectShadows(bool detectShadows)
+
+BackgroundSubtractorMOG2::getShadowValue
+---------------------------------------------
+Returns the shadow value
+
+.. ocv:function:: int BackgroundSubtractorMOG2::getShadowValue() const
+
+Shadow value is the value used to mark shadows in the foreground mask. Default value is 127. Value 0 in the mask always means background, 255 means foreground.
+
+BackgroundSubtractorMOG2::setShadowValue
+---------------------------------------------
+Sets the shadow value
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setShadowValue(int value)
+
+BackgroundSubtractorMOG2::getShadowThreshold
+---------------------------------------------
+Returns the shadow threshold
+
+.. ocv:function:: double BackgroundSubtractorMOG2::getShadowThreshold() const
+
+A shadow is detected if pixel is a darker version of the background. The shadow threshold (``Tau`` in the paper) is a threshold defining how much darker the shadow can be. ``Tau= 0.5`` means that if a pixel is more than twice darker then it is not shadow. See Prati, Mikic, Trivedi and Cucchiarra, *Detecting Moving Shadows...*, IEEE PAMI,2003.
+
+BackgroundSubtractorMOG2::setShadowThreshold
+---------------------------------------------
+Sets the shadow threshold
+
+.. ocv:function:: void BackgroundSubtractorMOG2::setShadowThreshold(double threshold)
+
+
+BackgroundSubtractorGMG
+------------------------
+Background Subtractor module based on the algorithm given in [Gold2012]_.
+
+.. ocv:class:: BackgroundSubtractorGMG : public BackgroundSubtractor
+
+
+createBackgroundSubtractorGMG
+-----------------------------------
+Creates a GMG Background Subtractor
+
+.. ocv:function:: Ptr<BackgroundSubtractorGMG> createBackgroundSubtractorGMG(int initializationFrames=120, double decisionThreshold=0.8)
+
+.. ocv:pyfunction:: cv2.createBackgroundSubtractorGMG([, initializationFrames[, decisionThreshold]]) -> retval
+
+    :param initializationFrames: number of frames used to initialize the background models.
+
+    :param decisionThreshold: Threshold value, above which it is marked foreground, else background.
+
+
+BackgroundSubtractorGMG::getNumFrames
+---------------------------------------
+Returns the number of frames used to initialize background model.
+
+.. ocv:function:: int BackgroundSubtractorGMG::getNumFrames() const
+
+
+BackgroundSubtractorGMG::setNumFrames
+---------------------------------------
+Sets the number of frames used to initialize background model.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setNumFrames(int nframes)
+
+
+BackgroundSubtractorGMG::getDefaultLearningRate
+--------------------------------------------------
+Returns the learning rate of the algorithm. It lies between 0.0 and 1.0. It determines how quickly features are "forgotten" from histograms.
+
+.. ocv:function:: double BackgroundSubtractorGMG::getDefaultLearningRate() const
+
+
+BackgroundSubtractorGMG::setDefaultLearningRate
+--------------------------------------------------
+Sets the learning rate of the algorithm.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setDefaultLearningRate(double lr)
+
+
+BackgroundSubtractorGMG::getDecisionThreshold
+--------------------------------------------------
+Returns the value of decision threshold. Decision value is the value above which pixel is determined to be FG.
+
+.. ocv:function:: double BackgroundSubtractorGMG::getDecisionThreshold() const
+
+
+BackgroundSubtractorGMG::setDecisionThreshold
+--------------------------------------------------
+Sets the value of decision threshold.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setDecisionThreshold(double thresh)
+
+
+BackgroundSubtractorGMG::getMaxFeatures
+--------------------------------------------------
+Returns total number of distinct colors to maintain in histogram.
+
+.. ocv:function:: int BackgroundSubtractorGMG::getMaxFeatures() const
+
+
+BackgroundSubtractorGMG::setMaxFeatures
+--------------------------------------------------
+Sets total number of distinct colors to maintain in histogram.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setMaxFeatures(int maxFeatures)
+
+
+BackgroundSubtractorGMG::getQuantizationLevels
+--------------------------------------------------
+Returns the parameter used for quantization of color-space. It is the number of discrete levels in each channel to be used in histograms.
+
+.. ocv:function:: int BackgroundSubtractorGMG::getQuantizationLevels() const
+
+
+BackgroundSubtractorGMG::setQuantizationLevels
+--------------------------------------------------
+Sets the parameter used for quantization of color-space
+
+.. ocv:function:: void BackgroundSubtractorGMG::setQuantizationLevels(int nlevels)
+
+
+BackgroundSubtractorGMG::getSmoothingRadius
+--------------------------------------------------
+Returns the kernel radius used for morphological operations
+
+.. ocv:function:: int BackgroundSubtractorGMG::getSmoothingRadius() const
+
+
+BackgroundSubtractorGMG::setSmoothingRadius
+--------------------------------------------------
+Sets the kernel radius used for morphological operations
+
+.. ocv:function:: void BackgroundSubtractorGMG::setSmoothingRadius(int radius)
+
+
+BackgroundSubtractorGMG::getUpdateBackgroundModel
+--------------------------------------------------
+Returns the status of background model update
+
+.. ocv:function:: bool BackgroundSubtractorGMG::getUpdateBackgroundModel() const
+
+
+BackgroundSubtractorGMG::setUpdateBackgroundModel
+--------------------------------------------------
+Sets the status of background model update
+
+.. ocv:function:: void BackgroundSubtractorGMG::setUpdateBackgroundModel(bool update)
+
+
+BackgroundSubtractorGMG::getMinVal
+--------------------------------------------------
+Returns the minimum value taken on by pixels in image sequence. Usually 0.
+
+.. ocv:function:: double BackgroundSubtractorGMG::getMinVal() const
+
+
+BackgroundSubtractorGMG::setMinVal
+--------------------------------------------------
+Sets the minimum value taken on by pixels in image sequence.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setMinVal(double val)
+
+
+BackgroundSubtractorGMG::getMaxVal
+--------------------------------------------------
+Returns the maximum value taken on by pixels in image sequence. e.g. 1.0 or 255.
+
+.. ocv:function:: double BackgroundSubtractorGMG::getMaxVal() const
+
+
+BackgroundSubtractorGMG::setMaxVal
+--------------------------------------------------
+Sets the maximum value taken on by pixels in image sequence.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setMaxVal(double val)
+
+
+BackgroundSubtractorGMG::getBackgroundPrior
+--------------------------------------------------
+Returns the prior probability that each individual pixel is a background pixel.
+
+.. ocv:function:: double BackgroundSubtractorGMG::getBackgroundPrior() const
+
+
+BackgroundSubtractorGMG::setBackgroundPrior
+--------------------------------------------------
+Sets the prior probability that each individual pixel is a background pixel.
+
+.. ocv:function:: void BackgroundSubtractorGMG::setBackgroundPrior(double bgprior)
 
 
 calcOpticalFlowSF
@@ -652,7 +872,7 @@ Calculate an optical flow using "SimpleFlow" algorithm.
 
     :param next: Second 8-bit 3-channel image of the same size as ``prev``
 
-    :param flow: computed flow image that has the same size as ``prev`` and type ``CV_32FC2`` 
+    :param flow: computed flow image that has the same size as ``prev`` and type ``CV_32FC2``
 
     :param layers: Number of layers
 
@@ -682,7 +902,9 @@ Calculate an optical flow using "SimpleFlow" algorithm.
 
 See [Tao2012]_. And site of project - http://graphics.berkeley.edu/papers/Tao-SAN-2012-05/.
 
+.. note::
 
+   * An example using the simpleFlow algorithm can be found at opencv_source_code/samples/cpp/simpleflow_demo.cpp
 
 createOptFlow_DualTVL1
 ----------------------
@@ -756,10 +978,16 @@ Releases all inner buffers.
 
 .. [Davis97] Davis, J.W. and Bobick, A.F. “The Representation and Recognition of Action Using Temporal Templates”, CVPR97, 1997
 
+.. [EP08] Evangelidis, G.D. and Psarakis E.Z. "Parametric Image Alignment using Enhanced Correlation Coefficient Maximization", IEEE Transactions on PAMI, vol. 32, no. 10, 2008
+
 .. [Farneback2003] Gunnar Farneback, Two-frame motion estimation based on polynomial expansion, Lecture Notes in Computer Science, 2003, (2749), , 363-370.
 
 .. [Horn81] Berthold K.P. Horn and Brian G. Schunck. Determining Optical Flow. Artificial Intelligence, 17, pp. 185-203, 1981.
 
+.. [KB2001] P. KadewTraKuPong and R. Bowden. "An improved adaptive background mixture model for real-time tracking with shadow detection", Proc. 2nd European Workshop on Advanced Video-Based Surveillance Systems, 2001: http://personal.ee.surrey.ac.uk/Personal/R.Bowden/publications/avbs01/avbs01.pdf
+
+.. [Javier2012] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
+
 .. [Lucas81] Lucas, B., and Kanade, T. An Iterative Image Registration Technique with an Application to Stereo Vision, Proc. of 7th International Joint Conference on Artificial Intelligence (IJCAI), pp. 674-679.
 
 .. [Welch95] Greg Welch and Gary Bishop “An Introduction to the Kalman Filter”, 1995
@@ -768,6 +996,8 @@ Releases all inner buffers.
 
 .. [Zach2007] C. Zach, T. Pock and H. Bischof. "A Duality Based Approach for Realtime TV-L1 Optical Flow", In Proceedings of Pattern Recognition (DAGM), Heidelberg, Germany, pp. 214-223, 2007
 
-.. [Javier2012] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
+.. [Zivkovic2004] Z. Zivkovic. "Improved adaptive Gausian mixture model for background subtraction", International Conference Pattern Recognition, UK, August, 2004, http://www.zoranz.net/Publications/zivkovic2004ICPR.pdf. The code is very fast and performs also shadow detection. Number of Gausssian components is adapted per pixel.
 
-.. [EP08] Evangelidis, G.D. and Psarakis E.Z. "Parametric Image Alignment using Enhanced Correlation Coefficient Maximization", IEEE Transactions on PAMI, vol. 32, no. 10, 2008
+.. [Zivkovic2006] Z.Zivkovic, F. van der Heijden. "Efficient Adaptive Density Estimation per Image Pixel for the Task of Background Subtraction", Pattern Recognition Letters, vol. 27, no. 7, pages 773-780, 2006.
+
+.. [Gold2012] Andrew B. Godbehere, Akihiro Matsukawa, Ken Goldberg, "Visual Tracking of Human Visitors under Variable-Lighting Conditions for a Responsive Audio Art Installation", American Control Conference, Montreal, June 2012.
diff --git a/modules/video/include/opencv2/video.hpp b/modules/video/include/opencv2/video.hpp
index 1dd96f543e..b429ceff98 100644
--- a/modules/video/include/opencv2/video.hpp
+++ b/modules/video/include/opencv2/video.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -46,13 +47,9 @@
 #include "opencv2/video/tracking.hpp"
 #include "opencv2/video/background_segm.hpp"
 
-#ifdef __cplusplus
 namespace cv
 {
-
 CV_EXPORTS bool initModule_video(void);
-
 }
-#endif
 
 #endif //__OPENCV_VIDEO_HPP__
diff --git a/modules/video/include/opencv2/video/background_segm.hpp b/modules/video/include/opencv2/video/background_segm.hpp
index 91e9900aaf..a46dc820ef 100644
--- a/modules/video/include/opencv2/video/background_segm.hpp
+++ b/modules/video/include/opencv2/video/background_segm.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -44,7 +45,7 @@
 #define __OPENCV_BACKGROUND_SEGM_HPP__
 
 #include "opencv2/core.hpp"
-#include <list>
+
 namespace cv
 {
 
@@ -57,17 +58,15 @@ namespace cv
 class CV_EXPORTS_W BackgroundSubtractor : public Algorithm
 {
 public:
-    //! the virtual destructor
-    virtual ~BackgroundSubtractor();
     //! the update operator that takes the next video frame and returns the current foreground mask as 8-bit binary image.
-    CV_WRAP_AS(apply) virtual void operator()(InputArray image, OutputArray fgmask,
-                                              double learningRate=0);
+    CV_WRAP virtual void apply(InputArray image, OutputArray fgmask, double learningRate=-1) = 0;
 
     //! computes a background image
-    virtual void getBackgroundImage(OutputArray backgroundImage) const;
+    CV_WRAP virtual void getBackgroundImage(OutputArray backgroundImage) const = 0;
 };
 
 
+
 /*!
  Gaussian Mixture-based Backbround/Foreground Segmentation Algorithm
 
@@ -81,32 +80,24 @@ public:
 class CV_EXPORTS_W BackgroundSubtractorMOG : public BackgroundSubtractor
 {
 public:
-    //! the default constructor
-    CV_WRAP BackgroundSubtractorMOG();
-    //! the full constructor that takes the length of the history, the number of gaussian mixtures, the background ratio parameter and the noise strength
-    CV_WRAP BackgroundSubtractorMOG(int history, int nmixtures, double backgroundRatio, double noiseSigma=0);
-    //! the destructor
-    virtual ~BackgroundSubtractorMOG();
-    //! the update operator
-    virtual void operator()(InputArray image, OutputArray fgmask, double learningRate=0);
+    CV_WRAP virtual int getHistory() const = 0;
+    CV_WRAP virtual void setHistory(int nframes) = 0;
 
-    //! re-initiaization method
-    virtual void initialize(Size frameSize, int frameType);
+    CV_WRAP virtual int getNMixtures() const = 0;
+    CV_WRAP virtual void setNMixtures(int nmix) = 0;
 
-    virtual AlgorithmInfo* info() const;
+    CV_WRAP virtual double getBackgroundRatio() const = 0;
+    CV_WRAP virtual void setBackgroundRatio(double backgroundRatio) = 0;
 
-protected:
-    Size frameSize;
-    int frameType;
-    Mat bgmodel;
-    int nframes;
-    int history;
-    int nmixtures;
-    double varThreshold;
-    double backgroundRatio;
-    double noiseSigma;
+    CV_WRAP virtual double getNoiseSigma() const = 0;
+    CV_WRAP virtual void setNoiseSigma(double noiseSigma) = 0;
 };
 
+CV_EXPORTS_W Ptr<BackgroundSubtractorMOG>
+    createBackgroundSubtractorMOG(int history=200, int nmixtures=5,
+                                  double backgroundRatio=0.7, double noiseSigma=0);
+
+
 
 /*!
  The class implements the following algorithm:
@@ -114,82 +105,51 @@ protected:
  Z.Zivkovic
  International Conference Pattern Recognition, UK, August, 2004.
  http://www.zoranz.net/Publications/zivkovic2004ICPR.pdf
-*/
-class CV_EXPORTS BackgroundSubtractorMOG2 : public BackgroundSubtractor
+ */
+class CV_EXPORTS_W BackgroundSubtractorMOG2 : public BackgroundSubtractor
 {
 public:
-    //! the default constructor
-    BackgroundSubtractorMOG2();
-    //! the full constructor that takes the length of the history, the number of gaussian mixtures, the background ratio parameter and the noise strength
-    BackgroundSubtractorMOG2(int history,  float varThreshold, bool bShadowDetection=true);
-    //! the destructor
-    virtual ~BackgroundSubtractorMOG2();
-    //! the update operator
-    virtual void operator()(InputArray image, OutputArray fgmask, double learningRate=-1);
+    CV_WRAP virtual int getHistory() const = 0;
+    CV_WRAP virtual void setHistory(int history) = 0;
 
-    //! computes a background image which are the mean of all background gaussians
-    virtual void getBackgroundImage(OutputArray backgroundImage) const;
+    CV_WRAP virtual int getNMixtures() const = 0;
+    CV_WRAP virtual void setNMixtures(int nmixtures) = 0;
 
-    //! re-initiaization method
-    virtual void initialize(Size frameSize, int frameType);
+    CV_WRAP virtual double getBackgroundRatio() const = 0;
+    CV_WRAP virtual void setBackgroundRatio(double ratio) = 0;
 
-    virtual AlgorithmInfo* info() const;
+    CV_WRAP virtual double getVarThreshold() const = 0;
+    CV_WRAP virtual void setVarThreshold(double varThreshold) = 0;
 
-protected:
-    Size frameSize;
-    int frameType;
-    Mat bgmodel;
-    Mat bgmodelUsedModes;//keep track of number of modes per pixel
-    int nframes;
-    int history;
-    int nmixtures;
-    //! here it is the maximum allowed number of mixture components.
-    //! Actual number is determined dynamically per pixel
-    double varThreshold;
-    // threshold on the squared Mahalanobis distance to decide if it is well described
-    // by the background model or not. Related to Cthr from the paper.
-    // This does not influence the update of the background. A typical value could be 4 sigma
-    // and that is varThreshold=4*4=16; Corresponds to Tb in the paper.
+    CV_WRAP virtual double getVarThresholdGen() const = 0;
+    CV_WRAP virtual void setVarThresholdGen(double varThresholdGen) = 0;
 
-    /////////////////////////
-    // less important parameters - things you might change but be carefull
-    ////////////////////////
-    float backgroundRatio;
-    // corresponds to fTB=1-cf from the paper
-    // TB - threshold when the component becomes significant enough to be included into
-    // the background model. It is the TB=1-cf from the paper. So I use cf=0.1 => TB=0.
-    // For alpha=0.001 it means that the mode should exist for approximately 105 frames before
-    // it is considered foreground
-    // float noiseSigma;
-    float varThresholdGen;
-    //correspondts to Tg - threshold on the squared Mahalan. dist. to decide
-    //when a sample is close to the existing components. If it is not close
-    //to any a new component will be generated. I use 3 sigma => Tg=3*3=9.
-    //Smaller Tg leads to more generated components and higher Tg might make
-    //lead to small number of components but they can grow too large
-    float fVarInit;
-    float fVarMin;
-    float fVarMax;
-    //initial variance  for the newly generated components.
-    //It will will influence the speed of adaptation. A good guess should be made.
-    //A simple way is to estimate the typical standard deviation from the images.
-    //I used here 10 as a reasonable value
-    // min and max can be used to further control the variance
-    float fCT;//CT - complexity reduction prior
-    //this is related to the number of samples needed to accept that a component
-    //actually exists. We use CT=0.05 of all the samples. By setting CT=0 you get
-    //the standard Stauffer&Grimson algorithm (maybe not exact but very similar)
+    CV_WRAP virtual double getVarInit() const = 0;
+    CV_WRAP virtual void setVarInit(double varInit) = 0;
 
-    //shadow detection parameters
-    bool bShadowDetection;//default 1 - do shadow detection
-    unsigned char nShadowDetection;//do shadow detection - insert this value as the detection result - 127 default value
-    float fTau;
-    // Tau - shadow threshold. The shadow is detected if the pixel is darker
-    //version of the background. Tau is a threshold on how much darker the shadow can be.
-    //Tau= 0.5 means that if pixel is more than 2 times darker then it is not shadow
-    //See: Prati,Mikic,Trivedi,Cucchiarra,"Detecting Moving Shadows...",IEEE PAMI,2003.
+    CV_WRAP virtual double getVarMin() const = 0;
+    CV_WRAP virtual void setVarMin(double varMin) = 0;
+
+    CV_WRAP virtual double getVarMax() const = 0;
+    CV_WRAP virtual void setVarMax(double varMax) = 0;
+
+    CV_WRAP virtual double getComplexityReductionThreshold() const = 0;
+    CV_WRAP virtual void setComplexityReductionThreshold(double ct) = 0;
+
+    CV_WRAP virtual bool getDetectShadows() const = 0;
+    CV_WRAP virtual void setDetectShadows(bool detectShadows) = 0;
+
+    CV_WRAP virtual int getShadowValue() const = 0;
+    CV_WRAP virtual void setShadowValue(int value) = 0;
+
+    CV_WRAP virtual double getShadowThreshold() const = 0;
+    CV_WRAP virtual void setShadowThreshold(double threshold) = 0;
 };
 
+CV_EXPORTS_W Ptr<BackgroundSubtractorMOG2>
+    createBackgroundSubtractorMOG2(int history=500, double varThreshold=16,
+                                   bool detectShadows=true);
+
 /**
  * Background Subtractor module. Takes a series of images and returns a sequence of mask (8UC1)
  * images of the same size, where 255 indicates Foreground and 0 represents Background.
@@ -197,66 +157,44 @@ protected:
  * Variable-Lighting Conditions for a Responsive Audio Art Installation," A. Godbehere,
  * A. Matsukawa, K. Goldberg, American Control Conference, Montreal, June 2012.
  */
-class CV_EXPORTS BackgroundSubtractorGMG: public cv::BackgroundSubtractor
+class CV_EXPORTS_W BackgroundSubtractorGMG : public BackgroundSubtractor
 {
 public:
-    BackgroundSubtractorGMG();
-    virtual ~BackgroundSubtractorGMG();
-    virtual AlgorithmInfo* info() const;
+    CV_WRAP virtual int getMaxFeatures() const = 0;
+    CV_WRAP virtual void setMaxFeatures(int maxFeatures) = 0;
 
-    /**
-     * Validate parameters and set up data structures for appropriate image size.
-     * Must call before running on data.
-     * @param frameSize input frame size
-     * @param min       minimum value taken on by pixels in image sequence. Usually 0
-     * @param max       maximum value taken on by pixels in image sequence. e.g. 1.0 or 255
-     */
-    void initialize(cv::Size frameSize, double min, double max);
+    CV_WRAP virtual double getDefaultLearningRate() const = 0;
+    CV_WRAP virtual void setDefaultLearningRate(double lr) = 0;
 
-    /**
-     * Performs single-frame background subtraction and builds up a statistical background image
-     * model.
-     * @param image Input image
-     * @param fgmask Output mask image representing foreground and background pixels
-     */
-    virtual void operator()(InputArray image, OutputArray fgmask, double learningRate=-1.0);
+    CV_WRAP virtual int getNumFrames() const = 0;
+    CV_WRAP virtual void setNumFrames(int nframes) = 0;
 
-    /**
-     * Releases all inner buffers.
-     */
-    void release();
+    CV_WRAP virtual int getQuantizationLevels() const = 0;
+    CV_WRAP virtual void setQuantizationLevels(int nlevels) = 0;
 
-    //! Total number of distinct colors to maintain in histogram.
-    int     maxFeatures;
-    //! Set between 0.0 and 1.0, determines how quickly features are "forgotten" from histograms.
-    double  learningRate;
-    //! Number of frames of video to use to initialize histograms.
-    int     numInitializationFrames;
-    //! Number of discrete levels in each channel to be used in histograms.
-    int     quantizationLevels;
-    //! Prior probability that any given pixel is a background pixel. A sensitivity parameter.
-    double  backgroundPrior;
-    //! Value above which pixel is determined to be FG.
-    double  decisionThreshold;
-    //! Smoothing radius, in pixels, for cleaning up FG image.
-    int     smoothingRadius;
-    //! Perform background model update
-    bool updateBackgroundModel;
+    CV_WRAP virtual double getBackgroundPrior() const = 0;
+    CV_WRAP virtual void setBackgroundPrior(double bgprior) = 0;
 
-private:
-    double maxVal_;
-    double minVal_;
+    CV_WRAP virtual int getSmoothingRadius() const = 0;
+    CV_WRAP virtual void setSmoothingRadius(int radius) = 0;
 
-    cv::Size frameSize_;
-    int frameNum_;
+    CV_WRAP virtual double getDecisionThreshold() const = 0;
+    CV_WRAP virtual void setDecisionThreshold(double thresh) = 0;
 
-    cv::Mat_<int> nfeatures_;
-    cv::Mat_<unsigned int> colors_;
-    cv::Mat_<float> weights_;
+    CV_WRAP virtual bool getUpdateBackgroundModel() const = 0;
+    CV_WRAP virtual void setUpdateBackgroundModel(bool update) = 0;
 
-    cv::Mat buf_;
+    CV_WRAP virtual double getMinVal() const = 0;
+    CV_WRAP virtual void setMinVal(double val) = 0;
+
+    CV_WRAP virtual double getMaxVal() const = 0;
+    CV_WRAP virtual void setMaxVal(double val) = 0;
 };
 
-}
+
+CV_EXPORTS_W Ptr<BackgroundSubtractorGMG> createBackgroundSubtractorGMG(int initializationFrames=120,
+                                                                        double decisionThreshold=0.8);
+
+} // cv
 
 #endif
diff --git a/modules/video/include/opencv2/video/tracking.hpp b/modules/video/include/opencv2/video/tracking.hpp
index ed322ef770..2045b19d98 100644
--- a/modules/video/include/opencv2/video/tracking.hpp
+++ b/modules/video/include/opencv2/video/tracking.hpp
@@ -7,11 +7,12 @@
 //  copy or use the software.
 //
 //
-//                           License Agreement
+//                          License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -46,228 +47,92 @@
 #include "opencv2/core.hpp"
 #include "opencv2/imgproc.hpp"
 
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/****************************************************************************************\
-*                                  Motion Analysis                                       *
-\****************************************************************************************/
-
-/************************************ optical flow ***************************************/
-
-#define  CV_LKFLOW_PYR_A_READY       1
-#define  CV_LKFLOW_PYR_B_READY       2
-#define  CV_LKFLOW_INITIAL_GUESSES   4
-#define  CV_LKFLOW_GET_MIN_EIGENVALS 8
-
-/* It is Lucas & Kanade method, modified to use pyramids.
-   Also it does several iterations to get optical flow for
-   every point at every pyramid level.
-   Calculates optical flow between two images for certain set of points (i.e.
-   it is a "sparse" optical flow, which is opposite to the previous 3 methods) */
-CVAPI(void)  cvCalcOpticalFlowPyrLK( const CvArr*  prev, const CvArr*  curr,
-                                     CvArr*  prev_pyr, CvArr*  curr_pyr,
-                                     const CvPoint2D32f* prev_features,
-                                     CvPoint2D32f* curr_features,
-                                     int       count,
-                                     CvSize    win_size,
-                                     int       level,
-                                     char*     status,
-                                     float*    track_error,
-                                     CvTermCriteria criteria,
-                                     int       flags );
-
-
-/* Modification of a previous sparse optical flow algorithm to calculate
-   affine flow */
-CVAPI(void)  cvCalcAffineFlowPyrLK( const CvArr*  prev, const CvArr*  curr,
-                                    CvArr*  prev_pyr, CvArr*  curr_pyr,
-                                    const CvPoint2D32f* prev_features,
-                                    CvPoint2D32f* curr_features,
-                                    float* matrices, int  count,
-                                    CvSize win_size, int  level,
-                                    char* status, float* track_error,
-                                    CvTermCriteria criteria, int flags );
-
-/* Estimate rigid transformation between 2 images or 2 point sets */
-CVAPI(int)  cvEstimateRigidTransform( const CvArr* A, const CvArr* B,
-                                      CvMat* M, int full_affine );
-
-/* Estimate optical flow for each pixel using the two-frame G. Farneback algorithm */
-CVAPI(void) cvCalcOpticalFlowFarneback( const CvArr* prev, const CvArr* next,
-                                        CvArr* flow, double pyr_scale, int levels,
-                                        int winsize, int iterations, int poly_n,
-                                        double poly_sigma, int flags );
-
-/********************************* motion templates *************************************/
-
-/****************************************************************************************\
-*        All the motion template functions work only with single channel images.         *
-*        Silhouette image must have depth IPL_DEPTH_8U or IPL_DEPTH_8S                   *
-*        Motion history image must have depth IPL_DEPTH_32F,                             *
-*        Gradient mask - IPL_DEPTH_8U or IPL_DEPTH_8S,                                   *
-*        Motion orientation image - IPL_DEPTH_32F                                        *
-*        Segmentation mask - IPL_DEPTH_32F                                               *
-*        All the angles are in degrees, all the times are in milliseconds                *
-\****************************************************************************************/
-
-/* Updates motion history image given motion silhouette */
-CVAPI(void)    cvUpdateMotionHistory( const CvArr* silhouette, CvArr* mhi,
-                                      double timestamp, double duration );
-
-/* Calculates gradient of the motion history image and fills
-   a mask indicating where the gradient is valid */
-CVAPI(void)    cvCalcMotionGradient( const CvArr* mhi, CvArr* mask, CvArr* orientation,
-                                     double delta1, double delta2,
-                                     int aperture_size CV_DEFAULT(3));
-
-/* Calculates average motion direction within a selected motion region
-   (region can be selected by setting ROIs and/or by composing a valid gradient mask
-   with the region mask) */
-CVAPI(double)  cvCalcGlobalOrientation( const CvArr* orientation, const CvArr* mask,
-                                        const CvArr* mhi, double timestamp,
-                                        double duration );
-
-/* Splits a motion history image into a few parts corresponding to separate independent motions
-   (e.g. left hand, right hand) */
-CVAPI(CvSeq*)  cvSegmentMotion( const CvArr* mhi, CvArr* seg_mask,
-                                CvMemStorage* storage,
-                                double timestamp, double seg_thresh );
-
-/****************************************************************************************\
-*                                       Tracking                                         *
-\****************************************************************************************/
-
-/* Implements CAMSHIFT algorithm - determines object position, size and orientation
-   from the object histogram back project (extension of meanshift) */
-CVAPI(int)  cvCamShift( const CvArr* prob_image, CvRect  window,
-                        CvTermCriteria criteria, CvConnectedComp* comp,
-                        CvBox2D* box CV_DEFAULT(NULL) );
-
-/* Implements MeanShift algorithm - determines object position
-   from the object histogram back project */
-CVAPI(int)  cvMeanShift( const CvArr* prob_image, CvRect  window,
-                         CvTermCriteria criteria, CvConnectedComp* comp );
-
-/*
-standard Kalman filter (in G. Welch' and G. Bishop's notation):
-
-  x(k)=A*x(k-1)+B*u(k)+w(k)  p(w)~N(0,Q)
-  z(k)=H*x(k)+v(k),   p(v)~N(0,R)
-*/
-typedef struct CvKalman
-{
-    int MP;                     /* number of measurement vector dimensions */
-    int DP;                     /* number of state vector dimensions */
-    int CP;                     /* number of control vector dimensions */
-
-    /* backward compatibility fields */
-#if 1
-    float* PosterState;         /* =state_pre->data.fl */
-    float* PriorState;          /* =state_post->data.fl */
-    float* DynamMatr;           /* =transition_matrix->data.fl */
-    float* MeasurementMatr;     /* =measurement_matrix->data.fl */
-    float* MNCovariance;        /* =measurement_noise_cov->data.fl */
-    float* PNCovariance;        /* =process_noise_cov->data.fl */
-    float* KalmGainMatr;        /* =gain->data.fl */
-    float* PriorErrorCovariance;/* =error_cov_pre->data.fl */
-    float* PosterErrorCovariance;/* =error_cov_post->data.fl */
-    float* Temp1;               /* temp1->data.fl */
-    float* Temp2;               /* temp2->data.fl */
-#endif
-
-    CvMat* state_pre;           /* predicted state (x'(k)):
-                                    x(k)=A*x(k-1)+B*u(k) */
-    CvMat* state_post;          /* corrected state (x(k)):
-                                    x(k)=x'(k)+K(k)*(z(k)-H*x'(k)) */
-    CvMat* transition_matrix;   /* state transition matrix (A) */
-    CvMat* control_matrix;      /* control matrix (B)
-                                   (it is not used if there is no control)*/
-    CvMat* measurement_matrix;  /* measurement matrix (H) */
-    CvMat* process_noise_cov;   /* process noise covariance matrix (Q) */
-    CvMat* measurement_noise_cov; /* measurement noise covariance matrix (R) */
-    CvMat* error_cov_pre;       /* priori error estimate covariance matrix (P'(k)):
-                                    P'(k)=A*P(k-1)*At + Q)*/
-    CvMat* gain;                /* Kalman gain matrix (K(k)):
-                                    K(k)=P'(k)*Ht*inv(H*P'(k)*Ht+R)*/
-    CvMat* error_cov_post;      /* posteriori error estimate covariance matrix (P(k)):
-                                    P(k)=(I-K(k)*H)*P'(k) */
-    CvMat* temp1;               /* temporary matrices */
-    CvMat* temp2;
-    CvMat* temp3;
-    CvMat* temp4;
-    CvMat* temp5;
-} CvKalman;
-
-/* Creates Kalman filter and sets A, B, Q, R and state to some initial values */
-CVAPI(CvKalman*) cvCreateKalman( int dynam_params, int measure_params,
-                                 int control_params CV_DEFAULT(0));
-
-/* Releases Kalman filter state */
-CVAPI(void)  cvReleaseKalman( CvKalman** kalman);
-
-/* Updates Kalman filter by time (predicts future state of the system) */
-CVAPI(const CvMat*)  cvKalmanPredict( CvKalman* kalman,
-                                      const CvMat* control CV_DEFAULT(NULL));
-
-/* Updates Kalman filter by measurement
-   (corrects state of the system and internal matrices) */
-CVAPI(const CvMat*)  cvKalmanCorrect( CvKalman* kalman, const CvMat* measurement );
-
-#define cvKalmanUpdateByTime  cvKalmanPredict
-#define cvKalmanUpdateByMeasurement cvKalmanCorrect
-
-
-/****************************************************************************************\
-*                                       Image Alignment (ECC algorithm)                  *
-\****************************************************************************************/
-enum
-{
-    MOTION_TRANSLATION,
-    MOTION_EUCLIDEAN,
-    MOTION_AFFINE,
-    MOTION_HOMOGRAPHY
-};
-
-/* Estimate the geometric transformation between 2 images (area-based alignment) */
-CVAPI(double)  cvFindTransformECC (const CvArr* templateImage, const CvArr* inputImage,
-                                   CvMat* warpMatrix,
-                                   const int motionType,
-                                   const CvTermCriteria criteria);
-
-#ifdef __cplusplus
-}
-
 namespace cv
 {
 
+enum { OPTFLOW_USE_INITIAL_FLOW     = 4,
+       OPTFLOW_LK_GET_MIN_EIGENVALS = 8,
+       OPTFLOW_FARNEBACK_GAUSSIAN   = 256
+     };
+
+enum { MOTION_TRANSLATION = 0,
+       MOTION_EUCLIDEAN   = 1,
+       MOTION_AFFINE      = 2,
+       MOTION_HOMOGRAPHY  = 3
+     };
+
 //! updates motion history image using the current silhouette
 CV_EXPORTS_W void updateMotionHistory( InputArray silhouette, InputOutputArray mhi,
                                        double timestamp, double duration );
 
 //! computes the motion gradient orientation image from the motion history image
-CV_EXPORTS_W void calcMotionGradient( InputArray mhi, OutputArray mask,
-                                      OutputArray orientation,
-                                      double delta1, double delta2,
-                                      int apertureSize=3 );
+CV_EXPORTS_W void calcMotionGradient( InputArray mhi, OutputArray mask, OutputArray orientation,
+                                      double delta1, double delta2, int apertureSize = 3 );
 
 //! computes the global orientation of the selected motion history image part
-CV_EXPORTS_W double calcGlobalOrientation( InputArray orientation, InputArray mask,
-                                           InputArray mhi, double timestamp,
-                                           double duration );
+CV_EXPORTS_W double calcGlobalOrientation( InputArray orientation, InputArray mask, InputArray mhi,
+                                           double timestamp, double duration );
 
-CV_EXPORTS_W void segmentMotion(InputArray mhi, OutputArray segmask,
-                                CV_OUT std::vector<Rect>& boundingRects,
-                                double timestamp, double segThresh);
+CV_EXPORTS_W void segmentMotion( InputArray mhi, OutputArray segmask,
+                                 CV_OUT std::vector<Rect>& boundingRects,
+                                 double timestamp, double segThresh );
 
 //! updates the object tracking window using CAMSHIFT algorithm
-CV_EXPORTS_W RotatedRect CamShift( InputArray probImage, CV_OUT CV_IN_OUT Rect& window,
+CV_EXPORTS_W RotatedRect CamShift( InputArray probImage, CV_IN_OUT Rect& window,
                                    TermCriteria criteria );
 
 //! updates the object tracking window using meanshift algorithm
-CV_EXPORTS_W int meanShift( InputArray probImage, CV_OUT CV_IN_OUT Rect& window,
-                            TermCriteria criteria );
+CV_EXPORTS_W int meanShift( InputArray probImage, CV_IN_OUT Rect& window, TermCriteria criteria );
+
+//! constructs a pyramid which can be used as input for calcOpticalFlowPyrLK
+CV_EXPORTS_W int buildOpticalFlowPyramid( InputArray img, OutputArrayOfArrays pyramid,
+                                          Size winSize, int maxLevel, bool withDerivatives = true,
+                                          int pyrBorder = BORDER_REFLECT_101,
+                                          int derivBorder = BORDER_CONSTANT,
+                                          bool tryReuseInputImage = true );
+
+//! computes sparse optical flow using multi-scale Lucas-Kanade algorithm
+CV_EXPORTS_W void calcOpticalFlowPyrLK( InputArray prevImg, InputArray nextImg,
+                                        InputArray prevPts, InputOutputArray nextPts,
+                                        OutputArray status, OutputArray err,
+                                        Size winSize = Size(21,21), int maxLevel = 3,
+                                        TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 0.01),
+                                        int flags = 0, double minEigThreshold = 1e-4 );
+
+//! computes dense optical flow using Farneback algorithm
+CV_EXPORTS_W void calcOpticalFlowFarneback( InputArray prev, InputArray next, InputOutputArray flow,
+                                            double pyr_scale, int levels, int winsize,
+                                            int iterations, int poly_n, double poly_sigma,
+                                            int flags );
+
+//! estimates the best-fit Euqcidean, similarity, affine or perspective transformation
+// that maps one 2D point set to another or one image to another.
+CV_EXPORTS_W Mat estimateRigidTransform( InputArray src, InputArray dst, bool fullAffine );
+
+//! estimates the best-fit Translation, Euclidean, Affine or Perspective Transformation
+// with respect to Enhanced Correlation Coefficient criterion that maps one image to
+// another (area-based alignment)
+//
+// see reference:
+// Evangelidis, G. E., Psarakis, E.Z., Parametric Image Alignment using
+// Enhanced Correlation Coefficient Maximization, PAMI, 30(8), 2008
+CV_EXPORTS_W double findTransformECC( InputArray templateImage, InputArray inputImage,
+                                      InputOutputArray warpMatrix, int motionType = MOTION_AFFINE,
+                                      TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 50, 0.001));
+
+
+//! computes dense optical flow using Simple Flow algorithm
+CV_EXPORTS_W void calcOpticalFlowSF( InputArray from, InputArray to, OutputArray flow,
+                                     int layers, int averaging_block_size, int max_flow);
+
+CV_EXPORTS_W void calcOpticalFlowSF( InputArray from, InputArray to, OutputArray flow, int layers,
+                                     int averaging_block_size, int max_flow,
+                                     double sigma_dist, double sigma_color, int postprocess_window,
+                                     double sigma_dist_fix, double sigma_color_fix, double occ_thr,
+                                     int upscale_averaging_radius, double upscale_sigma_dist,
+                                     double upscale_sigma_color, double speed_up_thr );
+
+
 
 /*!
  Kalman filter.
@@ -282,14 +147,14 @@ public:
     //! the default constructor
     CV_WRAP KalmanFilter();
     //! the full constructor taking the dimensionality of the state, of the measurement and of the control vector
-    CV_WRAP KalmanFilter(int dynamParams, int measureParams, int controlParams=0, int type=CV_32F);
+    CV_WRAP KalmanFilter( int dynamParams, int measureParams, int controlParams = 0, int type = CV_32F );
     //! re-initializes Kalman filter. The previous content is destroyed.
-    void init(int dynamParams, int measureParams, int controlParams=0, int type=CV_32F);
+    void init( int dynamParams, int measureParams, int controlParams = 0, int type = CV_32F );
 
     //! computes predicted state
-    CV_WRAP const Mat& predict(const Mat& control=Mat());
+    CV_WRAP const Mat& predict( const Mat& control = Mat() );
     //! updates the predicted state from the measurement
-    CV_WRAP const Mat& correct(const Mat& measurement);
+    CV_WRAP const Mat& correct( const Mat& measurement );
 
     Mat statePre;           //!< predicted state (x'(k)): x(k)=A*x(k-1)+B*u(k)
     Mat statePost;          //!< corrected state (x(k)): x(k)=x'(k)+K(k)*(z(k)-H*x'(k))
@@ -310,81 +175,12 @@ public:
     Mat temp5;
 };
 
-enum
-{
-    OPTFLOW_USE_INITIAL_FLOW = CV_LKFLOW_INITIAL_GUESSES,
-    OPTFLOW_LK_GET_MIN_EIGENVALS = CV_LKFLOW_GET_MIN_EIGENVALS,
-    OPTFLOW_FARNEBACK_GAUSSIAN = 256
-};
 
-//! constructs a pyramid which can be used as input for calcOpticalFlowPyrLK
-CV_EXPORTS_W int buildOpticalFlowPyramid(InputArray img, OutputArrayOfArrays pyramid,
-                                         Size winSize, int maxLevel, bool withDerivatives = true,
-                                         int pyrBorder = BORDER_REFLECT_101, int derivBorder = BORDER_CONSTANT,
-                                         bool tryReuseInputImage = true);
-
-//! computes sparse optical flow using multi-scale Lucas-Kanade algorithm
-CV_EXPORTS_W void calcOpticalFlowPyrLK( InputArray prevImg, InputArray nextImg,
-                           InputArray prevPts, InputOutputArray nextPts,
-                           OutputArray status, OutputArray err,
-                           Size winSize=Size(21,21), int maxLevel=3,
-                           TermCriteria criteria=TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 0.01),
-                           int flags=0, double minEigThreshold=1e-4);
-
-//! computes dense optical flow using Farneback algorithm
-CV_EXPORTS_W void calcOpticalFlowFarneback( InputArray prev, InputArray next,
-                           InputOutputArray flow, double pyr_scale, int levels, int winsize,
-                           int iterations, int poly_n, double poly_sigma, int flags );
-
-//! estimates the best-fit Euqcidean, similarity, affine or perspective transformation
-// that maps one 2D point set to another or one image to another.
-CV_EXPORTS_W Mat estimateRigidTransform( InputArray src, InputArray dst,
-                                         bool fullAffine);
-
-//! estimates the best-fit Translation, Euclidean, Affine or Perspective Transformation
-// with respect to Enhanced Correlation Coefficient criterion that maps one image to
-// another (area-based alignment)
-//
-// see reference:
-// Evangelidis, G. E., Psarakis, E.Z., Parametric Image Alignment using
-// Enhanced Correlation Coefficient Maximization, PAMI, 30(8), 2008
-
-CV_EXPORTS_W double findTransformECC(InputArray templateImage,
-                                     InputArray inputImage,
-                                     InputOutputArray warpMatrix,
-                                     int motionType=MOTION_AFFINE,
-                                     TermCriteria criteria=TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 50, 0.001));
-
-
-//! computes dense optical flow using Simple Flow algorithm
-CV_EXPORTS_W void calcOpticalFlowSF(InputArray from,
-                                    InputArray to,
-                                    OutputArray flow,
-                                    int layers,
-                                    int averaging_block_size,
-                                    int max_flow);
-
-CV_EXPORTS_W void calcOpticalFlowSF(InputArray from,
-                                    InputArray to,
-                                    OutputArray flow,
-                                    int layers,
-                                    int averaging_block_size,
-                                    int max_flow,
-                                    double sigma_dist,
-                                    double sigma_color,
-                                    int postprocess_window,
-                                    double sigma_dist_fix,
-                                    double sigma_color_fix,
-                                    double occ_thr,
-                                    int upscale_averaging_radius,
-                                    double upscale_sigma_dist,
-                                    double upscale_sigma_color,
-                                    double speed_up_thr);
 
 class CV_EXPORTS DenseOpticalFlow : public Algorithm
 {
 public:
-    virtual void calc(InputArray I0, InputArray I1, InputOutputArray flow) = 0;
+    virtual void calc( InputArray I0, InputArray I1, InputOutputArray flow ) = 0;
     virtual void collectGarbage() = 0;
 };
 
@@ -395,8 +191,6 @@ public:
 //   [2] Javier Sanchez, Enric Meinhardt-Llopis and Gabriele Facciolo. "TV-L1 Optical Flow Estimation".
 CV_EXPORTS Ptr<DenseOpticalFlow> createOptFlow_DualTVL1();
 
-}
-
-#endif
+} // cv
 
 #endif
diff --git a/modules/video/include/opencv2/video/tracking_c.h b/modules/video/include/opencv2/video/tracking_c.h
new file mode 100644
index 0000000000..e05a0b3ada
--- /dev/null
+++ b/modules/video/include/opencv2/video/tracking_c.h
@@ -0,0 +1,227 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef __OPENCV_TRACKING_C_H__
+#define __OPENCV_TRACKING_C_H__
+
+#include "opencv2/imgproc/types_c.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/****************************************************************************************\
+*                                  Motion Analysis                                       *
+\****************************************************************************************/
+
+/************************************ optical flow ***************************************/
+
+#define CV_LKFLOW_PYR_A_READY       1
+#define CV_LKFLOW_PYR_B_READY       2
+#define CV_LKFLOW_INITIAL_GUESSES   4
+#define CV_LKFLOW_GET_MIN_EIGENVALS 8
+
+/* It is Lucas & Kanade method, modified to use pyramids.
+   Also it does several iterations to get optical flow for
+   every point at every pyramid level.
+   Calculates optical flow between two images for certain set of points (i.e.
+   it is a "sparse" optical flow, which is opposite to the previous 3 methods) */
+CVAPI(void)  cvCalcOpticalFlowPyrLK( const CvArr*  prev, const CvArr*  curr,
+                                     CvArr*  prev_pyr, CvArr*  curr_pyr,
+                                     const CvPoint2D32f* prev_features,
+                                     CvPoint2D32f* curr_features,
+                                     int       count,
+                                     CvSize    win_size,
+                                     int       level,
+                                     char*     status,
+                                     float*    track_error,
+                                     CvTermCriteria criteria,
+                                     int       flags );
+
+
+/* Modification of a previous sparse optical flow algorithm to calculate
+   affine flow */
+CVAPI(void)  cvCalcAffineFlowPyrLK( const CvArr*  prev, const CvArr*  curr,
+                                    CvArr*  prev_pyr, CvArr*  curr_pyr,
+                                    const CvPoint2D32f* prev_features,
+                                    CvPoint2D32f* curr_features,
+                                    float* matrices, int  count,
+                                    CvSize win_size, int  level,
+                                    char* status, float* track_error,
+                                    CvTermCriteria criteria, int flags );
+
+/* Estimate rigid transformation between 2 images or 2 point sets */
+CVAPI(int)  cvEstimateRigidTransform( const CvArr* A, const CvArr* B,
+                                      CvMat* M, int full_affine );
+
+/* Estimate optical flow for each pixel using the two-frame G. Farneback algorithm */
+CVAPI(void) cvCalcOpticalFlowFarneback( const CvArr* prev, const CvArr* next,
+                                        CvArr* flow, double pyr_scale, int levels,
+                                        int winsize, int iterations, int poly_n,
+                                        double poly_sigma, int flags );
+
+/********************************* motion templates *************************************/
+
+/****************************************************************************************\
+*        All the motion template functions work only with single channel images.         *
+*        Silhouette image must have depth IPL_DEPTH_8U or IPL_DEPTH_8S                   *
+*        Motion history image must have depth IPL_DEPTH_32F,                             *
+*        Gradient mask - IPL_DEPTH_8U or IPL_DEPTH_8S,                                   *
+*        Motion orientation image - IPL_DEPTH_32F                                        *
+*        Segmentation mask - IPL_DEPTH_32F                                               *
+*        All the angles are in degrees, all the times are in milliseconds                *
+\****************************************************************************************/
+
+/* Updates motion history image given motion silhouette */
+CVAPI(void)    cvUpdateMotionHistory( const CvArr* silhouette, CvArr* mhi,
+                                      double timestamp, double duration );
+
+/* Calculates gradient of the motion history image and fills
+   a mask indicating where the gradient is valid */
+CVAPI(void)    cvCalcMotionGradient( const CvArr* mhi, CvArr* mask, CvArr* orientation,
+                                     double delta1, double delta2,
+                                     int aperture_size CV_DEFAULT(3));
+
+/* Calculates average motion direction within a selected motion region
+   (region can be selected by setting ROIs and/or by composing a valid gradient mask
+   with the region mask) */
+CVAPI(double)  cvCalcGlobalOrientation( const CvArr* orientation, const CvArr* mask,
+                                        const CvArr* mhi, double timestamp,
+                                        double duration );
+
+/* Splits a motion history image into a few parts corresponding to separate independent motions
+   (e.g. left hand, right hand) */
+CVAPI(CvSeq*)  cvSegmentMotion( const CvArr* mhi, CvArr* seg_mask,
+                                CvMemStorage* storage,
+                                double timestamp, double seg_thresh );
+
+/****************************************************************************************\
+*                                       Tracking                                         *
+\****************************************************************************************/
+
+/* Implements CAMSHIFT algorithm - determines object position, size and orientation
+   from the object histogram back project (extension of meanshift) */
+CVAPI(int)  cvCamShift( const CvArr* prob_image, CvRect  window,
+                        CvTermCriteria criteria, CvConnectedComp* comp,
+                        CvBox2D* box CV_DEFAULT(NULL) );
+
+/* Implements MeanShift algorithm - determines object position
+   from the object histogram back project */
+CVAPI(int)  cvMeanShift( const CvArr* prob_image, CvRect  window,
+                         CvTermCriteria criteria, CvConnectedComp* comp );
+
+/*
+standard Kalman filter (in G. Welch' and G. Bishop's notation):
+
+  x(k)=A*x(k-1)+B*u(k)+w(k)  p(w)~N(0,Q)
+  z(k)=H*x(k)+v(k),   p(v)~N(0,R)
+*/
+typedef struct CvKalman
+{
+    int MP;                     /* number of measurement vector dimensions */
+    int DP;                     /* number of state vector dimensions */
+    int CP;                     /* number of control vector dimensions */
+
+    /* backward compatibility fields */
+#if 1
+    float* PosterState;         /* =state_pre->data.fl */
+    float* PriorState;          /* =state_post->data.fl */
+    float* DynamMatr;           /* =transition_matrix->data.fl */
+    float* MeasurementMatr;     /* =measurement_matrix->data.fl */
+    float* MNCovariance;        /* =measurement_noise_cov->data.fl */
+    float* PNCovariance;        /* =process_noise_cov->data.fl */
+    float* KalmGainMatr;        /* =gain->data.fl */
+    float* PriorErrorCovariance;/* =error_cov_pre->data.fl */
+    float* PosterErrorCovariance;/* =error_cov_post->data.fl */
+    float* Temp1;               /* temp1->data.fl */
+    float* Temp2;               /* temp2->data.fl */
+#endif
+
+    CvMat* state_pre;           /* predicted state (x'(k)):
+                                    x(k)=A*x(k-1)+B*u(k) */
+    CvMat* state_post;          /* corrected state (x(k)):
+                                    x(k)=x'(k)+K(k)*(z(k)-H*x'(k)) */
+    CvMat* transition_matrix;   /* state transition matrix (A) */
+    CvMat* control_matrix;      /* control matrix (B)
+                                   (it is not used if there is no control)*/
+    CvMat* measurement_matrix;  /* measurement matrix (H) */
+    CvMat* process_noise_cov;   /* process noise covariance matrix (Q) */
+    CvMat* measurement_noise_cov; /* measurement noise covariance matrix (R) */
+    CvMat* error_cov_pre;       /* priori error estimate covariance matrix (P'(k)):
+                                    P'(k)=A*P(k-1)*At + Q)*/
+    CvMat* gain;                /* Kalman gain matrix (K(k)):
+                                    K(k)=P'(k)*Ht*inv(H*P'(k)*Ht+R)*/
+    CvMat* error_cov_post;      /* posteriori error estimate covariance matrix (P(k)):
+                                    P(k)=(I-K(k)*H)*P'(k) */
+    CvMat* temp1;               /* temporary matrices */
+    CvMat* temp2;
+    CvMat* temp3;
+    CvMat* temp4;
+    CvMat* temp5;
+} CvKalman;
+
+/* Creates Kalman filter and sets A, B, Q, R and state to some initial values */
+CVAPI(CvKalman*) cvCreateKalman( int dynam_params, int measure_params,
+                                 int control_params CV_DEFAULT(0));
+
+/* Releases Kalman filter state */
+CVAPI(void)  cvReleaseKalman( CvKalman** kalman);
+
+/* Updates Kalman filter by time (predicts future state of the system) */
+CVAPI(const CvMat*)  cvKalmanPredict( CvKalman* kalman,
+                                      const CvMat* control CV_DEFAULT(NULL));
+
+/* Updates Kalman filter by measurement
+   (corrects state of the system and internal matrices) */
+CVAPI(const CvMat*)  cvKalmanCorrect( CvKalman* kalman, const CvMat* measurement );
+
+#define cvKalmanUpdateByTime  cvKalmanPredict
+#define cvKalmanUpdateByMeasurement cvKalmanCorrect
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+
+#endif // __OPENCV_TRACKING_C_H__
diff --git a/modules/video/include/opencv2/video/video.hpp b/modules/video/include/opencv2/video/video.hpp
index 8837bff1be..8267b85d59 100644
--- a/modules/video/include/opencv2/video/video.hpp
+++ b/modules/video/include/opencv2/video/video.hpp
@@ -45,4 +45,4 @@
 #error this is a compatibility header which should not be used inside the OpenCV library
 #endif
 
-#include "opencv2/video.hpp"
\ No newline at end of file
+#include "opencv2/video.hpp"
diff --git a/modules/video/perf/perf_ecc.cpp b/modules/video/perf/perf_ecc.cpp
index 66fad48bb9..4f0709b88e 100644
--- a/modules/video/perf/perf_ecc.cpp
+++ b/modules/video/perf/perf_ecc.cpp
@@ -35,7 +35,7 @@ PERF_TEST_P(TransformationType, findTransformECC, /*testing::ValuesIn(MotionType
                 0.f, 1.f, 11.839f);
 
             warpAffine(img, templateImage, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_EUCLIDEAN:
             angle = CV_PI/30;
@@ -43,20 +43,20 @@ PERF_TEST_P(TransformationType, findTransformECC, /*testing::ValuesIn(MotionType
             warpGround = (Mat_<float>(2,3) << (float)cos(angle), (float)-sin(angle), 12.123f,
                 (float)sin(angle), (float)cos(angle), 14.789f);
             warpAffine(img, templateImage, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_AFFINE:
             warpGround = (Mat_<float>(2,3) << 0.98f, 0.03f, 15.523f,
                 -0.02f, 0.95f, 10.456f);
             warpAffine(img, templateImage, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_HOMOGRAPHY:
             warpGround = (Mat_<float>(3,3) << 0.98f, 0.03f, 15.523f,
                 -0.02f, 0.95f, 10.456f,
                 0.0002f, 0.0003f, 1.f);
             warpPerspective(img, templateImage, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
     }
 
diff --git a/modules/video/perf/perf_optflowpyrlk.cpp b/modules/video/perf/perf_optflowpyrlk.cpp
index d34bd72150..339cbd0354 100644
--- a/modules/video/perf/perf_optflowpyrlk.cpp
+++ b/modules/video/perf/perf_optflowpyrlk.cpp
@@ -50,7 +50,7 @@ PERF_TEST_P(Path_Idx_Cn_NPoints_WSize, OpticalFlowPyrLK_full, testing::Combine(
     int winSize = get<4>(GetParam());
 
     int maxLevel = 2;
-    TermCriteria criteria(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS, 7, 0.001);
+    TermCriteria criteria(TermCriteria::COUNT|TermCriteria::EPS, 7, 0.001);
     int flags = 0;
     double minEigThreshold = 1e-4;
 
@@ -124,7 +124,7 @@ PERF_TEST_P(Path_Idx_Cn_NPoints_WSize_Deriv, OpticalFlowPyrLK_self, testing::Com
     bool withDerivatives = get<5>(GetParam());
 
     int maxLevel = 2;
-    TermCriteria criteria(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS, 7, 0.001);
+    TermCriteria criteria(TermCriteria::COUNT|TermCriteria::EPS, 7, 0.001);
     int flags = 0;
     double minEigThreshold = 1e-4;
 
@@ -165,7 +165,8 @@ PERF_TEST_P(Path_Idx_Cn_NPoints_WSize_Deriv, OpticalFlowPyrLK_self, testing::Com
     declare.in(pyramid1, pyramid2, inPoints).out(outPoints);
     declare.time(400);
 
-    TEST_CYCLE()
+    int runs = 3;
+    TEST_CYCLE_MULTIRUN(runs)
     {
         calcOpticalFlowPyrLK(pyramid1, pyramid2, inPoints, outPoints, status, err,
                              Size(winSize, winSize), maxLevel, criteria,
@@ -185,7 +186,7 @@ PERF_TEST_P(Path_Win_Deriv_Border_Reuse, OpticalFlowPyrLK_pyr, testing::Combine(
                 testing::Values<std::string>("cv/optflow/frames/720p_01.png"),
                 testing::Values(7, 11),
                 testing::Bool(),
-                testing::ValuesIn(PyrBorderMode::all()),
+                PyrBorderMode::all(),
                 testing::Bool()
                 )
             )
@@ -217,4 +218,4 @@ PERF_TEST_P(Path_Win_Deriv_Border_Reuse, OpticalFlowPyrLK_pyr, testing::Combine(
     }
 
     SANITY_CHECK(pyramid);
-}
\ No newline at end of file
+}
diff --git a/modules/video/perf/perf_precomp.cpp b/modules/video/perf/perf_precomp.cpp
deleted file mode 100644
index 8552ac3d42..0000000000
--- a/modules/video/perf/perf_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "perf_precomp.hpp"
diff --git a/modules/video/perf/perf_precomp.hpp b/modules/video/perf/perf_precomp.hpp
index 94f8d92234..0ea88f0da5 100644
--- a/modules/video/perf/perf_precomp.hpp
+++ b/modules/video/perf/perf_precomp.hpp
@@ -9,10 +9,10 @@
 #ifndef __OPENCV_VIDEO_PRECOMP_HPP__
 #define __OPENCV_VIDEO_PRECOMP_HPP__
 
+#include "opencv2/ts.hpp"
 #include <opencv2/imgproc.hpp>
 #include <opencv2/video.hpp>
 #include <opencv2/highgui.hpp>
-#include "opencv2/ts.hpp"
 
 #ifdef GTEST_CREATE_SHARED_LIBRARY
 #error no modules except ts should have GTEST_CREATE_SHARED_LIBRARY defined
diff --git a/modules/video/perf/perf_tvl1optflow.cpp b/modules/video/perf/perf_tvl1optflow.cpp
index 36f16d9943..0a0f74a82e 100644
--- a/modules/video/perf/perf_tvl1optflow.cpp
+++ b/modules/video/perf/perf_tvl1optflow.cpp
@@ -26,5 +26,5 @@ PERF_TEST_P(ImagePair, OpticalFlowDual_TVL1, testing::Values(impair("cv/optflow/
 
     TEST_CYCLE_N(10) tvl1->calc(frame1, frame2, flow);
 
-    SANITY_CHECK(flow, 0.5);
+    SANITY_CHECK(flow, 0.8);
 }
diff --git a/modules/video/src/affineflow.cpp b/modules/video/src/affineflow.cpp
new file mode 100644
index 0000000000..bad9da8af7
--- /dev/null
+++ b/modules/video/src/affineflow.cpp
@@ -0,0 +1,850 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/video/tracking_c.h"
+
+// to be moved to legacy
+
+static int icvMinimalPyramidSize( CvSize imgSize )
+{
+    return cvAlign(imgSize.width,8) * imgSize.height / 3;
+}
+
+
+static void
+icvInitPyramidalAlgorithm( const CvMat* imgA, const CvMat* imgB,
+                           CvMat* pyrA, CvMat* pyrB,
+                           int level, CvTermCriteria * criteria,
+                           int max_iters, int flags,
+                           uchar *** imgI, uchar *** imgJ,
+                           int **step, CvSize** size,
+                           double **scale, cv::AutoBuffer<uchar>* buffer )
+{
+    const int ALIGN = 8;
+    int pyrBytes, bufferBytes = 0, elem_size;
+    int level1 = level + 1;
+
+    int i;
+    CvSize imgSize, levelSize;
+
+    *imgI = *imgJ = 0;
+    *step = 0;
+    *scale = 0;
+    *size = 0;
+
+    /* check input arguments */
+    if( ((flags & CV_LKFLOW_PYR_A_READY) != 0 && !pyrA) ||
+        ((flags & CV_LKFLOW_PYR_B_READY) != 0 && !pyrB) )
+        CV_Error( CV_StsNullPtr, "Some of the precomputed pyramids are missing" );
+
+    if( level < 0 )
+        CV_Error( CV_StsOutOfRange, "The number of pyramid levels is negative" );
+
+    switch( criteria->type )
+    {
+    case CV_TERMCRIT_ITER:
+        criteria->epsilon = 0.f;
+        break;
+    case CV_TERMCRIT_EPS:
+        criteria->max_iter = max_iters;
+        break;
+    case CV_TERMCRIT_ITER | CV_TERMCRIT_EPS:
+        break;
+    default:
+        assert( 0 );
+        CV_Error( CV_StsBadArg, "Invalid termination criteria" );
+    }
+
+    /* compare squared values */
+    criteria->epsilon *= criteria->epsilon;
+
+    /* set pointers and step for every level */
+    pyrBytes = 0;
+
+    imgSize = cvGetSize(imgA);
+    elem_size = CV_ELEM_SIZE(imgA->type);
+    levelSize = imgSize;
+
+    for( i = 1; i < level1; i++ )
+    {
+        levelSize.width = (levelSize.width + 1) >> 1;
+        levelSize.height = (levelSize.height + 1) >> 1;
+
+        int tstep = cvAlign(levelSize.width,ALIGN) * elem_size;
+        pyrBytes += tstep * levelSize.height;
+    }
+
+    assert( pyrBytes <= imgSize.width * imgSize.height * elem_size * 4 / 3 );
+
+    /* buffer_size = <size for patches> + <size for pyramids> */
+    bufferBytes = (int)((level1 >= 0) * ((pyrA->data.ptr == 0) +
+        (pyrB->data.ptr == 0)) * pyrBytes +
+        (sizeof(imgI[0][0]) * 2 + sizeof(step[0][0]) +
+         sizeof(size[0][0]) + sizeof(scale[0][0])) * level1);
+
+    buffer->allocate( bufferBytes );
+
+    *imgI = (uchar **) (uchar*)(*buffer);
+    *imgJ = *imgI + level1;
+    *step = (int *) (*imgJ + level1);
+    *scale = (double *) (*step + level1);
+    *size = (CvSize *)(*scale + level1);
+
+    imgI[0][0] = imgA->data.ptr;
+    imgJ[0][0] = imgB->data.ptr;
+    step[0][0] = imgA->step;
+    scale[0][0] = 1;
+    size[0][0] = imgSize;
+
+    if( level > 0 )
+    {
+        uchar *bufPtr = (uchar *) (*size + level1);
+        uchar *ptrA = pyrA->data.ptr;
+        uchar *ptrB = pyrB->data.ptr;
+
+        if( !ptrA )
+        {
+            ptrA = bufPtr;
+            bufPtr += pyrBytes;
+        }
+
+        if( !ptrB )
+            ptrB = bufPtr;
+
+        levelSize = imgSize;
+
+        /* build pyramids for both frames */
+        for( i = 1; i <= level; i++ )
+        {
+            int levelBytes;
+            CvMat prev_level, next_level;
+
+            levelSize.width = (levelSize.width + 1) >> 1;
+            levelSize.height = (levelSize.height + 1) >> 1;
+
+            size[0][i] = levelSize;
+            step[0][i] = cvAlign( levelSize.width, ALIGN ) * elem_size;
+            scale[0][i] = scale[0][i - 1] * 0.5;
+
+            levelBytes = step[0][i] * levelSize.height;
+            imgI[0][i] = (uchar *) ptrA;
+            ptrA += levelBytes;
+
+            if( !(flags & CV_LKFLOW_PYR_A_READY) )
+            {
+                prev_level = cvMat( size[0][i-1].height, size[0][i-1].width, CV_8UC1 );
+                next_level = cvMat( size[0][i].height, size[0][i].width, CV_8UC1 );
+                cvSetData( &prev_level, imgI[0][i-1], step[0][i-1] );
+                cvSetData( &next_level, imgI[0][i], step[0][i] );
+                cvPyrDown( &prev_level, &next_level );
+            }
+
+            imgJ[0][i] = (uchar *) ptrB;
+            ptrB += levelBytes;
+
+            if( !(flags & CV_LKFLOW_PYR_B_READY) )
+            {
+                prev_level = cvMat( size[0][i-1].height, size[0][i-1].width, CV_8UC1 );
+                next_level = cvMat( size[0][i].height, size[0][i].width, CV_8UC1 );
+                cvSetData( &prev_level, imgJ[0][i-1], step[0][i-1] );
+                cvSetData( &next_level, imgJ[0][i], step[0][i] );
+                cvPyrDown( &prev_level, &next_level );
+            }
+        }
+    }
+}
+
+
+/* compute dI/dx and dI/dy */
+static void
+icvCalcIxIy_32f( const float* src, int src_step, float* dstX, float* dstY, int dst_step,
+                 CvSize src_size, const float* smooth_k, float* buffer0 )
+{
+    int src_width = src_size.width, dst_width = src_size.width-2;
+    int x, height = src_size.height - 2;
+    float* buffer1 = buffer0 + src_width;
+
+    src_step /= sizeof(src[0]);
+    dst_step /= sizeof(dstX[0]);
+
+    for( ; height--; src += src_step, dstX += dst_step, dstY += dst_step )
+    {
+        const float* src2 = src + src_step;
+        const float* src3 = src + src_step*2;
+
+        for( x = 0; x < src_width; x++ )
+        {
+            float t0 = (src3[x] + src[x])*smooth_k[0] + src2[x]*smooth_k[1];
+            float t1 = src3[x] - src[x];
+            buffer0[x] = t0; buffer1[x] = t1;
+        }
+
+        for( x = 0; x < dst_width; x++ )
+        {
+            float t0 = buffer0[x+2] - buffer0[x];
+            float t1 = (buffer1[x] + buffer1[x+2])*smooth_k[0] + buffer1[x+1]*smooth_k[1];
+            dstX[x] = t0; dstY[x] = t1;
+        }
+    }
+}
+
+
+#undef CV_8TO32F
+#define CV_8TO32F(a) (a)
+
+static const void*
+icvAdjustRect( const void* srcptr, int src_step, int pix_size,
+              CvSize src_size, CvSize win_size,
+              CvPoint ip, CvRect* pRect )
+{
+    CvRect rect;
+    const char* src = (const char*)srcptr;
+
+    if( ip.x >= 0 )
+    {
+        src += ip.x*pix_size;
+        rect.x = 0;
+    }
+    else
+    {
+        rect.x = -ip.x;
+        if( rect.x > win_size.width )
+            rect.x = win_size.width;
+    }
+
+    if( ip.x + win_size.width < src_size.width )
+        rect.width = win_size.width;
+    else
+    {
+        rect.width = src_size.width - ip.x - 1;
+        if( rect.width < 0 )
+        {
+            src += rect.width*pix_size;
+            rect.width = 0;
+        }
+        assert( rect.width <= win_size.width );
+    }
+
+    if( ip.y >= 0 )
+    {
+        src += ip.y * src_step;
+        rect.y = 0;
+    }
+    else
+        rect.y = -ip.y;
+
+    if( ip.y + win_size.height < src_size.height )
+        rect.height = win_size.height;
+    else
+    {
+        rect.height = src_size.height - ip.y - 1;
+        if( rect.height < 0 )
+        {
+            src += rect.height*src_step;
+            rect.height = 0;
+        }
+    }
+
+    *pRect = rect;
+    return src - rect.x*pix_size;
+}
+
+
+static CvStatus CV_STDCALL icvGetRectSubPix_8u32f_C1R
+( const uchar* src, int src_step, CvSize src_size,
+ float* dst, int dst_step, CvSize win_size, CvPoint2D32f center )
+{
+    CvPoint ip;
+    float  a12, a22, b1, b2;
+    float a, b;
+    double s = 0;
+    int i, j;
+
+    center.x -= (win_size.width-1)*0.5f;
+    center.y -= (win_size.height-1)*0.5f;
+
+    ip.x = cvFloor( center.x );
+    ip.y = cvFloor( center.y );
+
+    if( win_size.width <= 0 || win_size.height <= 0 )
+        return CV_BADRANGE_ERR;
+
+    a = center.x - ip.x;
+    b = center.y - ip.y;
+    a = MAX(a,0.0001f);
+    a12 = a*(1.f-b);
+    a22 = a*b;
+    b1 = 1.f - b;
+    b2 = b;
+    s = (1. - a)/a;
+
+    src_step /= sizeof(src[0]);
+    dst_step /= sizeof(dst[0]);
+
+    if( 0 <= ip.x && ip.x + win_size.width < src_size.width &&
+       0 <= ip.y && ip.y + win_size.height < src_size.height )
+    {
+        // extracted rectangle is totally inside the image
+        src += ip.y * src_step + ip.x;
+
+#if 0
+        if( icvCopySubpix_8u32f_C1R_p &&
+           icvCopySubpix_8u32f_C1R_p( src, src_step, dst,
+                                     dst_step*sizeof(dst[0]), win_size, a, b ) >= 0 )
+            return CV_OK;
+#endif
+
+        for( ; win_size.height--; src += src_step, dst += dst_step )
+        {
+            float prev = (1 - a)*(b1*CV_8TO32F(src[0]) + b2*CV_8TO32F(src[src_step]));
+            for( j = 0; j < win_size.width; j++ )
+            {
+                float t = a12*CV_8TO32F(src[j+1]) + a22*CV_8TO32F(src[j+1+src_step]);
+                dst[j] = prev + t;
+                prev = (float)(t*s);
+            }
+        }
+    }
+    else
+    {
+        CvRect r;
+
+        src = (const uchar*)icvAdjustRect( src, src_step*sizeof(*src),
+                                          sizeof(*src), src_size, win_size,ip, &r);
+
+        for( i = 0; i < win_size.height; i++, dst += dst_step )
+        {
+            const uchar *src2 = src + src_step;
+
+            if( i < r.y || i >= r.height )
+                src2 -= src_step;
+
+            for( j = 0; j < r.x; j++ )
+            {
+                float s0 = CV_8TO32F(src[r.x])*b1 +
+                CV_8TO32F(src2[r.x])*b2;
+
+                dst[j] = (float)(s0);
+            }
+
+            if( j < r.width )
+            {
+                float prev = (1 - a)*(b1*CV_8TO32F(src[j]) + b2*CV_8TO32F(src2[j]));
+
+                for( ; j < r.width; j++ )
+                {
+                    float t = a12*CV_8TO32F(src[j+1]) + a22*CV_8TO32F(src2[j+1]);
+                    dst[j] = prev + t;
+                    prev = (float)(t*s);
+                }
+            }
+
+            for( ; j < win_size.width; j++ )
+            {
+                float s0 = CV_8TO32F(src[r.width])*b1 +
+                CV_8TO32F(src2[r.width])*b2;
+
+                dst[j] = (float)(s0);
+            }
+
+            if( i < r.height )
+                src = src2;
+        }
+    }
+
+    return CV_OK;
+}
+
+
+#define ICV_32F8U(x)  ((uchar)cvRound(x))
+
+#define ICV_DEF_GET_QUADRANGLE_SUB_PIX_FUNC( flavor, srctype, dsttype, worktype, cast_macro, cvt ) \
+static CvStatus CV_STDCALL icvGetQuadrangleSubPix_##flavor##_C1R                    \
+              ( const srctype * src, int src_step, CvSize src_size,                 \
+                dsttype *dst, int dst_step, CvSize win_size, const float *matrix )  \
+{                                                                                   \
+    int x, y;                                                                       \
+    double dx = (win_size.width - 1)*0.5;                                           \
+    double dy = (win_size.height - 1)*0.5;                                          \
+    double A11 = matrix[0], A12 = matrix[1], A13 = matrix[2]-A11*dx-A12*dy;         \
+    double A21 = matrix[3], A22 = matrix[4], A23 = matrix[5]-A21*dx-A22*dy;         \
+                                                                                    \
+    src_step /= sizeof(srctype);                                                    \
+    dst_step /= sizeof(dsttype);                                                    \
+                                                                                    \
+    for( y = 0; y < win_size.height; y++, dst += dst_step )                         \
+    {                                                                               \
+        double xs = A12*y + A13;                                                    \
+        double ys = A22*y + A23;                                                    \
+        double xe = A11*(win_size.width-1) + A12*y + A13;                           \
+        double ye = A21*(win_size.width-1) + A22*y + A23;                           \
+                                                                                    \
+        if( (unsigned)(cvFloor(xs)-1) < (unsigned)(src_size.width - 3) &&           \
+            (unsigned)(cvFloor(ys)-1) < (unsigned)(src_size.height - 3) &&          \
+            (unsigned)(cvFloor(xe)-1) < (unsigned)(src_size.width - 3) &&           \
+            (unsigned)(cvFloor(ye)-1) < (unsigned)(src_size.height - 3))            \
+        {                                                                           \
+            for( x = 0; x < win_size.width; x++ )                                   \
+            {                                                                       \
+                int ixs = cvFloor( xs );                                            \
+                int iys = cvFloor( ys );                                            \
+                const srctype *ptr = src + src_step*iys + ixs;                      \
+                double a = xs - ixs, b = ys - iys, a1 = 1.f - a;                    \
+                worktype p0 = cvt(ptr[0])*a1 + cvt(ptr[1])*a;                       \
+                worktype p1 = cvt(ptr[src_step])*a1 + cvt(ptr[src_step+1])*a;       \
+                xs += A11;                                                          \
+                ys += A21;                                                          \
+                                                                                    \
+                dst[x] = cast_macro(p0 + b * (p1 - p0));                            \
+            }                                                                       \
+        }                                                                           \
+        else                                                                        \
+        {                                                                           \
+            for( x = 0; x < win_size.width; x++ )                                   \
+            {                                                                       \
+                int ixs = cvFloor( xs ), iys = cvFloor( ys );                       \
+                double a = xs - ixs, b = ys - iys, a1 = 1.f - a;                    \
+                const srctype *ptr0, *ptr1;                                         \
+                worktype p0, p1;                                                    \
+                xs += A11; ys += A21;                                               \
+                                                                                    \
+                if( (unsigned)iys < (unsigned)(src_size.height-1) )                 \
+                    ptr0 = src + src_step*iys, ptr1 = ptr0 + src_step;              \
+                else                                                                \
+                    ptr0 = ptr1 = src + (iys < 0 ? 0 : src_size.height-1)*src_step; \
+                                                                                    \
+                if( (unsigned)ixs < (unsigned)(src_size.width-1) )                  \
+                {                                                                   \
+                    p0 = cvt(ptr0[ixs])*a1 + cvt(ptr0[ixs+1])*a;                    \
+                    p1 = cvt(ptr1[ixs])*a1 + cvt(ptr1[ixs+1])*a;                    \
+                }                                                                   \
+                else                                                                \
+                {                                                                   \
+                    ixs = ixs < 0 ? 0 : src_size.width - 1;                         \
+                    p0 = cvt(ptr0[ixs]); p1 = cvt(ptr1[ixs]);                       \
+                }                                                                   \
+                dst[x] = cast_macro(p0 + b * (p1 - p0));                            \
+            }                                                                       \
+        }                                                                           \
+    }                                                                               \
+                                                                                    \
+    return CV_OK;                                                                   \
+}
+
+ICV_DEF_GET_QUADRANGLE_SUB_PIX_FUNC( 8u32f, uchar, float, double, cv::saturate_cast<float>, CV_8TO32F )
+
+/* Affine tracking algorithm */
+
+CV_IMPL void
+cvCalcAffineFlowPyrLK( const void* arrA, const void* arrB,
+                       void* pyrarrA, void* pyrarrB,
+                       const CvPoint2D32f * featuresA,
+                       CvPoint2D32f * featuresB,
+                       float *matrices, int count,
+                       CvSize winSize, int level,
+                       char *status, float *error,
+                       CvTermCriteria criteria, int flags )
+{
+    const int MAX_ITERS = 100;
+
+    cv::AutoBuffer<char> _status;
+    cv::AutoBuffer<uchar> buffer;
+    cv::AutoBuffer<uchar> pyr_buffer;
+
+    CvMat stubA, *imgA = (CvMat*)arrA;
+    CvMat stubB, *imgB = (CvMat*)arrB;
+    CvMat pstubA, *pyrA = (CvMat*)pyrarrA;
+    CvMat pstubB, *pyrB = (CvMat*)pyrarrB;
+
+    static const float smoothKernel[] = { 0.09375, 0.3125, 0.09375 };  /* 3/32, 10/32, 3/32 */
+
+    int bufferBytes = 0;
+
+    uchar **imgI = 0;
+    uchar **imgJ = 0;
+    int *step = 0;
+    double *scale = 0;
+    CvSize* size = 0;
+
+    float *patchI;
+    float *patchJ;
+    float *Ix;
+    float *Iy;
+
+    int i, j, k, l;
+
+    CvSize patchSize = cvSize( winSize.width * 2 + 1, winSize.height * 2 + 1 );
+    int patchLen = patchSize.width * patchSize.height;
+    int patchStep = patchSize.width * sizeof( patchI[0] );
+
+    CvSize srcPatchSize = cvSize( patchSize.width + 2, patchSize.height + 2 );
+    int srcPatchLen = srcPatchSize.width * srcPatchSize.height;
+    int srcPatchStep = srcPatchSize.width * sizeof( patchI[0] );
+    CvSize imgSize;
+    float eps = (float)MIN(winSize.width, winSize.height);
+
+    imgA = cvGetMat( imgA, &stubA );
+    imgB = cvGetMat( imgB, &stubB );
+
+    if( CV_MAT_TYPE( imgA->type ) != CV_8UC1 )
+        CV_Error( CV_StsUnsupportedFormat, "" );
+
+    if( !CV_ARE_TYPES_EQ( imgA, imgB ))
+        CV_Error( CV_StsUnmatchedFormats, "" );
+
+    if( !CV_ARE_SIZES_EQ( imgA, imgB ))
+        CV_Error( CV_StsUnmatchedSizes, "" );
+
+    if( imgA->step != imgB->step )
+        CV_Error( CV_StsUnmatchedSizes, "imgA and imgB must have equal steps" );
+
+    if( !matrices )
+        CV_Error( CV_StsNullPtr, "" );
+
+    imgSize = cv::Size(imgA->cols, imgA->rows);
+
+    if( pyrA )
+    {
+        pyrA = cvGetMat( pyrA, &pstubA );
+
+        if( pyrA->step*pyrA->height < icvMinimalPyramidSize( imgSize ) )
+            CV_Error( CV_StsBadArg, "pyramid A has insufficient size" );
+    }
+    else
+    {
+        pyrA = &pstubA;
+        pyrA->data.ptr = 0;
+    }
+
+    if( pyrB )
+    {
+        pyrB = cvGetMat( pyrB, &pstubB );
+
+        if( pyrB->step*pyrB->height < icvMinimalPyramidSize( imgSize ) )
+            CV_Error( CV_StsBadArg, "pyramid B has insufficient size" );
+    }
+    else
+    {
+        pyrB = &pstubB;
+        pyrB->data.ptr = 0;
+    }
+
+    if( count == 0 )
+        return;
+
+    /* check input arguments */
+    if( !featuresA || !featuresB || !matrices )
+        CV_Error( CV_StsNullPtr, "" );
+
+    if( winSize.width <= 1 || winSize.height <= 1 )
+        CV_Error( CV_StsOutOfRange, "the search window is too small" );
+
+    if( count < 0 )
+        CV_Error( CV_StsOutOfRange, "" );
+
+    icvInitPyramidalAlgorithm( imgA, imgB,
+        pyrA, pyrB, level, &criteria, MAX_ITERS, flags,
+        &imgI, &imgJ, &step, &size, &scale, &pyr_buffer );
+
+    /* buffer_size = <size for patches> + <size for pyramids> */
+    bufferBytes = (srcPatchLen + patchLen*3)*sizeof(patchI[0]) + (36*2 + 6)*sizeof(double);
+
+    buffer.allocate(bufferBytes);
+
+    if( !status )
+    {
+        _status.allocate(count);
+        status = _status;
+    }
+
+    patchI = (float *)(uchar*)buffer;
+    patchJ = patchI + srcPatchLen;
+    Ix = patchJ + patchLen;
+    Iy = Ix + patchLen;
+
+    if( status )
+        memset( status, 1, count );
+
+    if( !(flags & CV_LKFLOW_INITIAL_GUESSES) )
+    {
+        memcpy( featuresB, featuresA, count * sizeof( featuresA[0] ));
+        for( i = 0; i < count * 4; i += 4 )
+        {
+            matrices[i] = matrices[i + 3] = 1.f;
+            matrices[i + 1] = matrices[i + 2] = 0.f;
+        }
+    }
+
+    for( i = 0; i < count; i++ )
+    {
+        featuresB[i].x = (float)(featuresB[i].x * scale[level] * 0.5);
+        featuresB[i].y = (float)(featuresB[i].y * scale[level] * 0.5);
+    }
+
+    /* do processing from top pyramid level (smallest image)
+       to the bottom (original image) */
+    for( l = level; l >= 0; l-- )
+    {
+        CvSize levelSize = size[l];
+        int levelStep = step[l];
+
+        /* find flow for each given point at the particular level */
+        for( i = 0; i < count; i++ )
+        {
+            CvPoint2D32f u;
+            float Av[6];
+            double G[36];
+            double meanI = 0, meanJ = 0;
+            int x, y;
+            int pt_status = status[i];
+            CvMat mat;
+
+            if( !pt_status )
+                continue;
+
+            Av[0] = matrices[i*4];
+            Av[1] = matrices[i*4+1];
+            Av[3] = matrices[i*4+2];
+            Av[4] = matrices[i*4+3];
+
+            Av[2] = featuresB[i].x += featuresB[i].x;
+            Av[5] = featuresB[i].y += featuresB[i].y;
+
+            u.x = (float) (featuresA[i].x * scale[l]);
+            u.y = (float) (featuresA[i].y * scale[l]);
+
+            if( u.x < -eps || u.x >= levelSize.width+eps ||
+                u.y < -eps || u.y >= levelSize.height+eps ||
+                icvGetRectSubPix_8u32f_C1R( imgI[l], levelStep,
+                levelSize, patchI, srcPatchStep, srcPatchSize, u ) < 0 )
+            {
+                /* point is outside the image. take the next */
+                if( l == 0 )
+                    status[i] = 0;
+                continue;
+            }
+
+            icvCalcIxIy_32f( patchI, srcPatchStep, Ix, Iy,
+                (srcPatchSize.width-2)*sizeof(patchI[0]), srcPatchSize,
+                smoothKernel, patchJ );
+
+            /* repack patchI (remove borders) */
+            for( k = 0; k < patchSize.height; k++ )
+                memcpy( patchI + k * patchSize.width,
+                        patchI + (k + 1) * srcPatchSize.width + 1, patchStep );
+
+            memset( G, 0, sizeof( G ));
+
+            /* calculate G matrix */
+            for( y = -winSize.height, k = 0; y <= winSize.height; y++ )
+            {
+                for( x = -winSize.width; x <= winSize.width; x++, k++ )
+                {
+                    double ixix = ((double) Ix[k]) * Ix[k];
+                    double ixiy = ((double) Ix[k]) * Iy[k];
+                    double iyiy = ((double) Iy[k]) * Iy[k];
+
+                    double xx, xy, yy;
+
+                    G[0] += ixix;
+                    G[1] += ixiy;
+                    G[2] += x * ixix;
+                    G[3] += y * ixix;
+                    G[4] += x * ixiy;
+                    G[5] += y * ixiy;
+
+                    // G[6] == G[1]
+                    G[7] += iyiy;
+                    // G[8] == G[4]
+                    // G[9] == G[5]
+                    G[10] += x * iyiy;
+                    G[11] += y * iyiy;
+
+                    xx = x * x;
+                    xy = x * y;
+                    yy = y * y;
+
+                    // G[12] == G[2]
+                    // G[13] == G[8] == G[4]
+                    G[14] += xx * ixix;
+                    G[15] += xy * ixix;
+                    G[16] += xx * ixiy;
+                    G[17] += xy * ixiy;
+
+                    // G[18] == G[3]
+                    // G[19] == G[9]
+                    // G[20] == G[15]
+                    G[21] += yy * ixix;
+                    // G[22] == G[17]
+                    G[23] += yy * ixiy;
+
+                    // G[24] == G[4]
+                    // G[25] == G[10]
+                    // G[26] == G[16]
+                    // G[27] == G[22]
+                    G[28] += xx * iyiy;
+                    G[29] += xy * iyiy;
+
+                    // G[30] == G[5]
+                    // G[31] == G[11]
+                    // G[32] == G[17]
+                    // G[33] == G[23]
+                    // G[34] == G[29]
+                    G[35] += yy * iyiy;
+
+                    meanI += patchI[k];
+                }
+            }
+
+            meanI /= patchSize.width*patchSize.height;
+
+            G[8] = G[4];
+            G[9] = G[5];
+            G[22] = G[17];
+
+            // fill part of G below its diagonal
+            for( y = 1; y < 6; y++ )
+                for( x = 0; x < y; x++ )
+                    G[y * 6 + x] = G[x * 6 + y];
+
+            cvInitMatHeader( &mat, 6, 6, CV_64FC1, G );
+
+            if( cvInvert( &mat, &mat, CV_SVD ) < 1e-4 )
+            {
+                /* bad matrix. take the next point */
+                if( l == 0 )
+                    status[i] = 0;
+                continue;
+            }
+
+            for( j = 0; j < criteria.max_iter; j++ )
+            {
+                double b[6] = {0,0,0,0,0,0}, eta[6];
+                double t0, t1, s = 0;
+
+                if( Av[2] < -eps || Av[2] >= levelSize.width+eps ||
+                    Av[5] < -eps || Av[5] >= levelSize.height+eps ||
+                    icvGetQuadrangleSubPix_8u32f_C1R( imgJ[l], levelStep,
+                    levelSize, patchJ, patchStep, patchSize, Av ) < 0 )
+                {
+                    pt_status = 0;
+                    break;
+                }
+
+                for( y = -winSize.height, k = 0, meanJ = 0; y <= winSize.height; y++ )
+                    for( x = -winSize.width; x <= winSize.width; x++, k++ )
+                        meanJ += patchJ[k];
+
+                meanJ = meanJ / (patchSize.width * patchSize.height) - meanI;
+
+                for( y = -winSize.height, k = 0; y <= winSize.height; y++ )
+                {
+                    for( x = -winSize.width; x <= winSize.width; x++, k++ )
+                    {
+                        double t = patchI[k] - patchJ[k] + meanJ;
+                        double ixt = Ix[k] * t;
+                        double iyt = Iy[k] * t;
+
+                        s += t;
+
+                        b[0] += ixt;
+                        b[1] += iyt;
+                        b[2] += x * ixt;
+                        b[3] += y * ixt;
+                        b[4] += x * iyt;
+                        b[5] += y * iyt;
+                    }
+                }
+
+                for( k = 0; k < 6; k++ )
+                    eta[k] = G[k*6]*b[0] + G[k*6+1]*b[1] + G[k*6+2]*b[2] +
+                        G[k*6+3]*b[3] + G[k*6+4]*b[4] + G[k*6+5]*b[5];
+
+                Av[2] = (float)(Av[2] + Av[0] * eta[0] + Av[1] * eta[1]);
+                Av[5] = (float)(Av[5] + Av[3] * eta[0] + Av[4] * eta[1]);
+
+                t0 = Av[0] * (1 + eta[2]) + Av[1] * eta[4];
+                t1 = Av[0] * eta[3] + Av[1] * (1 + eta[5]);
+                Av[0] = (float)t0;
+                Av[1] = (float)t1;
+
+                t0 = Av[3] * (1 + eta[2]) + Av[4] * eta[4];
+                t1 = Av[3] * eta[3] + Av[4] * (1 + eta[5]);
+                Av[3] = (float)t0;
+                Av[4] = (float)t1;
+
+                if( eta[0] * eta[0] + eta[1] * eta[1] < criteria.epsilon )
+                    break;
+            }
+
+            if( pt_status != 0 || l == 0 )
+            {
+                status[i] = (char)pt_status;
+                featuresB[i].x = Av[2];
+                featuresB[i].y = Av[5];
+
+                matrices[i*4] = Av[0];
+                matrices[i*4+1] = Av[1];
+                matrices[i*4+2] = Av[3];
+                matrices[i*4+3] = Av[4];
+            }
+
+            if( pt_status && l == 0 && error )
+            {
+                /* calc error */
+                double err = 0;
+
+                for( y = 0, k = 0; y < patchSize.height; y++ )
+                {
+                    for( x = 0; x < patchSize.width; x++, k++ )
+                    {
+                        double t = patchI[k] - patchJ[k] + meanJ;
+                        err += t * t;
+                    }
+                }
+                error[i] = (float)std::sqrt(err);
+            }
+        }
+    }
+}
diff --git a/modules/video/src/bgfg_gaussmix.cpp b/modules/video/src/bgfg_gaussmix.cpp
index be3732be58..b5ad039e27 100644
--- a/modules/video/src/bgfg_gaussmix.cpp
+++ b/modules/video/src/bgfg_gaussmix.cpp
@@ -7,9 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                          License Agreement
+//                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -22,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -58,15 +60,6 @@
 namespace cv
 {
 
-BackgroundSubtractor::~BackgroundSubtractor() {}
-void BackgroundSubtractor::operator()(InputArray, OutputArray, double)
-{
-}
-
-void BackgroundSubtractor::getBackgroundImage(OutputArray) const
-{
-}
-
 static const int defaultNMixtures = 5;
 static const int defaultHistory = 200;
 static const double defaultBackgroundRatio = 0.7;
@@ -74,55 +67,108 @@ static const double defaultVarThreshold = 2.5*2.5;
 static const double defaultNoiseSigma = 30*0.5;
 static const double defaultInitialWeight = 0.05;
 
-BackgroundSubtractorMOG::BackgroundSubtractorMOG()
+class BackgroundSubtractorMOGImpl : public BackgroundSubtractorMOG
 {
-    frameSize = Size(0,0);
-    frameType = 0;
+public:
+    //! the default constructor
+    BackgroundSubtractorMOGImpl()
+    {
+        frameSize = Size(0,0);
+        frameType = 0;
 
-    nframes = 0;
-    nmixtures = defaultNMixtures;
-    history = defaultHistory;
-    varThreshold = defaultVarThreshold;
-    backgroundRatio = defaultBackgroundRatio;
-    noiseSigma = defaultNoiseSigma;
-}
+        nframes = 0;
+        nmixtures = defaultNMixtures;
+        history = defaultHistory;
+        varThreshold = defaultVarThreshold;
+        backgroundRatio = defaultBackgroundRatio;
+        noiseSigma = defaultNoiseSigma;
+        name_ = "BackgroundSubtractor.MOG";
+    }
+    // the full constructor that takes the length of the history,
+    // the number of gaussian mixtures, the background ratio parameter and the noise strength
+    BackgroundSubtractorMOGImpl(int _history, int _nmixtures, double _backgroundRatio, double _noiseSigma=0)
+    {
+        frameSize = Size(0,0);
+        frameType = 0;
 
-BackgroundSubtractorMOG::BackgroundSubtractorMOG(int _history, int _nmixtures,
-                                                 double _backgroundRatio,
-                                                 double _noiseSigma)
-{
-    frameSize = Size(0,0);
-    frameType = 0;
+        nframes = 0;
+        nmixtures = std::min(_nmixtures > 0 ? _nmixtures : defaultNMixtures, 8);
+        history = _history > 0 ? _history : defaultHistory;
+        varThreshold = defaultVarThreshold;
+        backgroundRatio = std::min(_backgroundRatio > 0 ? _backgroundRatio : 0.95, 1.);
+        noiseSigma = _noiseSigma <= 0 ? defaultNoiseSigma : _noiseSigma;
+    }
 
-    nframes = 0;
-    nmixtures = std::min(_nmixtures > 0 ? _nmixtures : defaultNMixtures, 8);
-    history = _history > 0 ? _history : defaultHistory;
-    varThreshold = defaultVarThreshold;
-    backgroundRatio = std::min(_backgroundRatio > 0 ? _backgroundRatio : 0.95, 1.);
-    noiseSigma = _noiseSigma <= 0 ? defaultNoiseSigma : _noiseSigma;
-}
+    //! the update operator
+    virtual void apply(InputArray image, OutputArray fgmask, double learningRate=0);
 
-BackgroundSubtractorMOG::~BackgroundSubtractorMOG()
-{
-}
+    //! re-initiaization method
+    virtual void initialize(Size _frameSize, int _frameType)
+    {
+        frameSize = _frameSize;
+        frameType = _frameType;
+        nframes = 0;
 
+        int nchannels = CV_MAT_CN(frameType);
+        CV_Assert( CV_MAT_DEPTH(frameType) == CV_8U );
 
-void BackgroundSubtractorMOG::initialize(Size _frameSize, int _frameType)
-{
-    frameSize = _frameSize;
-    frameType = _frameType;
-    nframes = 0;
+        // for each gaussian mixture of each pixel bg model we store ...
+        // the mixture sort key (w/sum_of_variances), the mixture weight (w),
+        // the mean (nchannels values) and
+        // the diagonal covariance matrix (another nchannels values)
+        bgmodel.create( 1, frameSize.height*frameSize.width*nmixtures*(2 + 2*nchannels), CV_32F );
+        bgmodel = Scalar::all(0);
+    }
 
-    int nchannels = CV_MAT_CN(frameType);
-    CV_Assert( CV_MAT_DEPTH(frameType) == CV_8U );
+    virtual AlgorithmInfo* info() const { return 0; }
 
-    // for each gaussian mixture of each pixel bg model we store ...
-    // the mixture sort key (w/sum_of_variances), the mixture weight (w),
-    // the mean (nchannels values) and
-    // the diagonal covariance matrix (another nchannels values)
-    bgmodel.create( 1, frameSize.height*frameSize.width*nmixtures*(2 + 2*nchannels), CV_32F );
-    bgmodel = Scalar::all(0);
-}
+    virtual void getBackgroundImage(OutputArray) const
+    {
+        CV_Error( Error::StsNotImplemented, "" );
+    }
+
+    virtual int getHistory() const { return history; }
+    virtual void setHistory(int _nframes) { history = _nframes; }
+
+    virtual int getNMixtures() const { return nmixtures; }
+    virtual void setNMixtures(int nmix) { nmixtures = nmix; }
+
+    virtual double getBackgroundRatio() const { return backgroundRatio; }
+    virtual void setBackgroundRatio(double _backgroundRatio) { backgroundRatio = _backgroundRatio; }
+
+    virtual double getNoiseSigma() const { return noiseSigma; }
+    virtual void setNoiseSigma(double _noiseSigma) { noiseSigma = _noiseSigma; }
+
+    virtual void write(FileStorage& fs) const
+    {
+        fs << "name" << name_
+           << "history" << history
+           << "nmixtures" << nmixtures
+           << "backgroundRatio" << backgroundRatio
+           << "noiseSigma" << noiseSigma;
+    }
+
+    virtual void read(const FileNode& fn)
+    {
+        CV_Assert( (String)fn["name"] == name_ );
+        history = (int)fn["history"];
+        nmixtures = (int)fn["nmixtures"];
+        backgroundRatio = (double)fn["backgroundRatio"];
+        noiseSigma = (double)fn["noiseSigma"];
+    }
+
+protected:
+    Size frameSize;
+    int frameType;
+    Mat bgmodel;
+    int nframes;
+    int history;
+    int nmixtures;
+    double varThreshold;
+    double backgroundRatio;
+    double noiseSigma;
+    String name_;
+};
 
 
 template<typename VT> struct MixData
@@ -391,7 +437,7 @@ static void process8uC3( const Mat& image, Mat& fgmask, double learningRate,
     }
 }
 
-void BackgroundSubtractorMOG::operator()(InputArray _image, OutputArray _fgmask, double learningRate)
+void BackgroundSubtractorMOGImpl::apply(InputArray _image, OutputArray _fgmask, double learningRate)
 {
     Mat image = _image.getMat();
     bool needToInitialize = nframes == 0 || learningRate >= 1 || image.size() != frameSize || image.type() != frameType;
@@ -412,10 +458,15 @@ void BackgroundSubtractorMOG::operator()(InputArray _image, OutputArray _fgmask,
     else if( image.type() == CV_8UC3 )
         process8uC3( image, fgmask, learningRate, bgmodel, nmixtures, backgroundRatio, varThreshold, noiseSigma );
     else
-        CV_Error( CV_StsUnsupportedFormat, "Only 1- and 3-channel 8-bit images are supported in BackgroundSubtractorMOG" );
+        CV_Error( Error::StsUnsupportedFormat, "Only 1- and 3-channel 8-bit images are supported in BackgroundSubtractorMOG" );
+}
+
+Ptr<BackgroundSubtractorMOG> createBackgroundSubtractorMOG(int history, int nmixtures,
+                                  double backgroundRatio, double noiseSigma)
+{
+    return makePtr<BackgroundSubtractorMOGImpl>(history, nmixtures, backgroundRatio, noiseSigma);
 }
 
 }
 
 /* End of file. */
-
diff --git a/modules/video/src/bgfg_gaussmix2.cpp b/modules/video/src/bgfg_gaussmix2.cpp
index e764348b4f..485e34d26d 100644
--- a/modules/video/src/bgfg_gaussmix2.cpp
+++ b/modules/video/src/bgfg_gaussmix2.cpp
@@ -7,9 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                          License Agreement
+//                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -22,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -114,6 +116,213 @@ static const float defaultfCT2 = 0.05f; // complexity reduction prior constant 0
 static const unsigned char defaultnShadowDetection2 = (unsigned char)127; // value to use in the segmentation mask for shadows, set 0 not to do shadow detection
 static const float defaultfTau = 0.5f; // Tau - shadow threshold, see the paper for explanation
 
+
+class BackgroundSubtractorMOG2Impl : public BackgroundSubtractorMOG2
+{
+public:
+    //! the default constructor
+    BackgroundSubtractorMOG2Impl()
+    {
+        frameSize = Size(0,0);
+        frameType = 0;
+
+        nframes = 0;
+        history = defaultHistory2;
+        varThreshold = defaultVarThreshold2;
+        bShadowDetection = 1;
+
+        nmixtures = defaultNMixtures2;
+        backgroundRatio = defaultBackgroundRatio2;
+        fVarInit = defaultVarInit2;
+        fVarMax  = defaultVarMax2;
+        fVarMin = defaultVarMin2;
+
+        varThresholdGen = defaultVarThresholdGen2;
+        fCT = defaultfCT2;
+        nShadowDetection =  defaultnShadowDetection2;
+        fTau = defaultfTau;
+    }
+    //! the full constructor that takes the length of the history,
+    // the number of gaussian mixtures, the background ratio parameter and the noise strength
+    BackgroundSubtractorMOG2Impl(int _history,  float _varThreshold, bool _bShadowDetection=true)
+    {
+        frameSize = Size(0,0);
+        frameType = 0;
+
+        nframes = 0;
+        history = _history > 0 ? _history : defaultHistory2;
+        varThreshold = (_varThreshold>0)? _varThreshold : defaultVarThreshold2;
+        bShadowDetection = _bShadowDetection;
+
+        nmixtures = defaultNMixtures2;
+        backgroundRatio = defaultBackgroundRatio2;
+        fVarInit = defaultVarInit2;
+        fVarMax  = defaultVarMax2;
+        fVarMin = defaultVarMin2;
+
+        varThresholdGen = defaultVarThresholdGen2;
+        fCT = defaultfCT2;
+        nShadowDetection =  defaultnShadowDetection2;
+        fTau = defaultfTau;
+        name_ = "BackgroundSubtractor.MOG2";
+    }
+    //! the destructor
+    ~BackgroundSubtractorMOG2Impl() {}
+    //! the update operator
+    void apply(InputArray image, OutputArray fgmask, double learningRate=-1);
+
+    //! computes a background image which are the mean of all background gaussians
+    virtual void getBackgroundImage(OutputArray backgroundImage) const;
+
+    //! re-initiaization method
+    void initialize(Size _frameSize, int _frameType)
+    {
+        frameSize = _frameSize;
+        frameType = _frameType;
+        nframes = 0;
+
+        int nchannels = CV_MAT_CN(frameType);
+        CV_Assert( nchannels <= CV_CN_MAX );
+
+        // for each gaussian mixture of each pixel bg model we store ...
+        // the mixture weight (w),
+        // the mean (nchannels values) and
+        // the covariance
+        bgmodel.create( 1, frameSize.height*frameSize.width*nmixtures*(2 + nchannels), CV_32F );
+        //make the array for keeping track of the used modes per pixel - all zeros at start
+        bgmodelUsedModes.create(frameSize,CV_8U);
+        bgmodelUsedModes = Scalar::all(0);
+    }
+
+    virtual AlgorithmInfo* info() const { return 0; }
+
+    virtual int getHistory() const { return history; }
+    virtual void setHistory(int _nframes) { history = _nframes; }
+
+    virtual int getNMixtures() const { return nmixtures; }
+    virtual void setNMixtures(int nmix) { nmixtures = nmix; }
+
+    virtual double getBackgroundRatio() const { return backgroundRatio; }
+    virtual void setBackgroundRatio(double _backgroundRatio) { backgroundRatio = (float)_backgroundRatio; }
+
+    virtual double getVarThreshold() const { return varThreshold; }
+    virtual void setVarThreshold(double _varThreshold) { varThreshold = _varThreshold; }
+
+    virtual double getVarThresholdGen() const { return varThresholdGen; }
+    virtual void setVarThresholdGen(double _varThresholdGen) { varThresholdGen = (float)_varThresholdGen; }
+
+    virtual double getVarInit() const { return fVarInit; }
+    virtual void setVarInit(double varInit) { fVarInit = (float)varInit; }
+
+    virtual double getVarMin() const { return fVarMin; }
+    virtual void setVarMin(double varMin) { fVarMin = (float)varMin; }
+
+    virtual double getVarMax() const { return fVarMax; }
+    virtual void setVarMax(double varMax) { fVarMax = (float)varMax; }
+
+    virtual double getComplexityReductionThreshold() const { return fCT; }
+    virtual void setComplexityReductionThreshold(double ct) { fCT = (float)ct; }
+
+    virtual bool getDetectShadows() const { return bShadowDetection; }
+    virtual void setDetectShadows(bool detectshadows) { bShadowDetection = detectshadows; }
+
+    virtual int getShadowValue() const { return nShadowDetection; }
+    virtual void setShadowValue(int value) { nShadowDetection = (uchar)value; }
+
+    virtual double getShadowThreshold() const { return fTau; }
+    virtual void setShadowThreshold(double value) { fTau = (float)value; }
+
+    virtual void write(FileStorage& fs) const
+    {
+        fs << "name" << name_
+        << "history" << history
+        << "nmixtures" << nmixtures
+        << "backgroundRatio" << backgroundRatio
+        << "varThreshold" << varThreshold
+        << "varThresholdGen" << varThresholdGen
+        << "varInit" << fVarInit
+        << "varMin" << fVarMin
+        << "varMax" << fVarMax
+        << "complexityReductionThreshold" << fCT
+        << "detectShadows" << (int)bShadowDetection
+        << "shadowValue" << (int)nShadowDetection
+        << "shadowThreshold" << fTau;
+    }
+
+    virtual void read(const FileNode& fn)
+    {
+        CV_Assert( (String)fn["name"] == name_ );
+        history = (int)fn["history"];
+        nmixtures = (int)fn["nmixtures"];
+        backgroundRatio = (float)fn["backgroundRatio"];
+        varThreshold = (double)fn["varThreshold"];
+        varThresholdGen = (float)fn["varThresholdGen"];
+        fVarInit = (float)fn["varInit"];
+        fVarMin = (float)fn["varMin"];
+        fVarMax = (float)fn["varMax"];
+        fCT = (float)fn["complexityReductionThreshold"];
+        bShadowDetection = (int)fn["detectShadows"] != 0;
+        nShadowDetection = saturate_cast<uchar>((int)fn["shadowValue"]);
+        fTau = (float)fn["shadowThreshold"];
+    }
+
+protected:
+    Size frameSize;
+    int frameType;
+    Mat bgmodel;
+    Mat bgmodelUsedModes;//keep track of number of modes per pixel
+    int nframes;
+    int history;
+    int nmixtures;
+    //! here it is the maximum allowed number of mixture components.
+    //! Actual number is determined dynamically per pixel
+    double varThreshold;
+    // threshold on the squared Mahalanobis distance to decide if it is well described
+    // by the background model or not. Related to Cthr from the paper.
+    // This does not influence the update of the background. A typical value could be 4 sigma
+    // and that is varThreshold=4*4=16; Corresponds to Tb in the paper.
+
+    /////////////////////////
+    // less important parameters - things you might change but be carefull
+    ////////////////////////
+    float backgroundRatio;
+    // corresponds to fTB=1-cf from the paper
+    // TB - threshold when the component becomes significant enough to be included into
+    // the background model. It is the TB=1-cf from the paper. So I use cf=0.1 => TB=0.
+    // For alpha=0.001 it means that the mode should exist for approximately 105 frames before
+    // it is considered foreground
+    // float noiseSigma;
+    float varThresholdGen;
+    //correspondts to Tg - threshold on the squared Mahalan. dist. to decide
+    //when a sample is close to the existing components. If it is not close
+    //to any a new component will be generated. I use 3 sigma => Tg=3*3=9.
+    //Smaller Tg leads to more generated components and higher Tg might make
+    //lead to small number of components but they can grow too large
+    float fVarInit;
+    float fVarMin;
+    float fVarMax;
+    //initial variance  for the newly generated components.
+    //It will will influence the speed of adaptation. A good guess should be made.
+    //A simple way is to estimate the typical standard deviation from the images.
+    //I used here 10 as a reasonable value
+    // min and max can be used to further control the variance
+    float fCT;//CT - complexity reduction prior
+    //this is related to the number of samples needed to accept that a component
+    //actually exists. We use CT=0.05 of all the samples. By setting CT=0 you get
+    //the standard Stauffer&Grimson algorithm (maybe not exact but very similar)
+
+    //shadow detection parameters
+    bool bShadowDetection;//default 1 - do shadow detection
+    unsigned char nShadowDetection;//do shadow detection - insert this value as the detection result - 127 default value
+    float fTau;
+    // Tau - shadow threshold. The shadow is detected if the pixel is darker
+    //version of the background. Tau is a threshold on how much darker the shadow can be.
+    //Tau= 0.5 means that if pixel is more than 2 times darker then it is not shadow
+    //See: Prati,Mikic,Trivedi,Cucchiarra,"Detecting Moving Shadows...",IEEE PAMI,2003.
+
+    String name_;
+};
+
 struct GaussBGStatModel2Params
 {
     //image info
@@ -248,8 +457,9 @@ detectShadowGMM(const float* data, int nchannels, int nmodes,
 //IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.26, no.5, pages 651-656, 2004
 //http://www.zoranz.net/Publications/zivkovic2004PAMI.pdf
 
-struct MOG2Invoker
+class MOG2Invoker : public ParallelLoopBody
 {
+public:
     MOG2Invoker(const Mat& _src, Mat& _dst,
                 GMM* _gmm, float* _mean,
                 uchar* _modesUsed,
@@ -276,13 +486,11 @@ struct MOG2Invoker
         tau = _tau;
         detectShadows = _detectShadows;
         shadowVal = _shadowVal;
-
-        cvtfunc = src->depth() != CV_32F ? getConvertFunc(src->depth(), CV_32F) : 0;
     }
 
-    void operator()(const BlockedRange& range) const
+    void operator()(const Range& range) const
     {
-        int y0 = range.begin(), y1 = range.end();
+        int y0 = range.start, y1 = range.end;
         int ncols = src->cols, nchannels = src->channels();
         AutoBuffer<float> buf(src->cols*nchannels);
         float alpha1 = 1.f - alphaT;
@@ -291,8 +499,8 @@ struct MOG2Invoker
         for( int y = y0; y < y1; y++ )
         {
             const float* data = buf;
-            if( cvtfunc )
-                cvtfunc( src->ptr(y), src->step, 0, 0, (uchar*)data, 0, Size(ncols*nchannels, 1), 0);
+            if( src->depth() != CV_32F )
+                src->row(y).convertTo(Mat(1, ncols, CV_32FC(nchannels), (void*)data), CV_32F);
             else
                 data = src->ptr<float>(y);
 
@@ -475,79 +683,9 @@ struct MOG2Invoker
 
     bool detectShadows;
     uchar shadowVal;
-
-    BinaryFunc cvtfunc;
 };
 
-BackgroundSubtractorMOG2::BackgroundSubtractorMOG2()
-{
-    frameSize = Size(0,0);
-    frameType = 0;
-
-    nframes = 0;
-    history = defaultHistory2;
-    varThreshold = defaultVarThreshold2;
-    bShadowDetection = 1;
-
-    nmixtures = defaultNMixtures2;
-    backgroundRatio = defaultBackgroundRatio2;
-    fVarInit = defaultVarInit2;
-    fVarMax  = defaultVarMax2;
-    fVarMin = defaultVarMin2;
-
-    varThresholdGen = defaultVarThresholdGen2;
-    fCT = defaultfCT2;
-    nShadowDetection =  defaultnShadowDetection2;
-    fTau = defaultfTau;
-}
-
-BackgroundSubtractorMOG2::BackgroundSubtractorMOG2(int _history,  float _varThreshold, bool _bShadowDetection)
-{
-    frameSize = Size(0,0);
-    frameType = 0;
-
-    nframes = 0;
-    history = _history > 0 ? _history : defaultHistory2;
-    varThreshold = (_varThreshold>0)? _varThreshold : defaultVarThreshold2;
-    bShadowDetection = _bShadowDetection;
-
-    nmixtures = defaultNMixtures2;
-    backgroundRatio = defaultBackgroundRatio2;
-    fVarInit = defaultVarInit2;
-    fVarMax  = defaultVarMax2;
-    fVarMin = defaultVarMin2;
-
-    varThresholdGen = defaultVarThresholdGen2;
-    fCT = defaultfCT2;
-    nShadowDetection =  defaultnShadowDetection2;
-    fTau = defaultfTau;
-}
-
-BackgroundSubtractorMOG2::~BackgroundSubtractorMOG2()
-{
-}
-
-
-void BackgroundSubtractorMOG2::initialize(Size _frameSize, int _frameType)
-{
-    frameSize = _frameSize;
-    frameType = _frameType;
-    nframes = 0;
-
-    int nchannels = CV_MAT_CN(frameType);
-    CV_Assert( nchannels <= CV_CN_MAX );
-
-    // for each gaussian mixture of each pixel bg model we store ...
-    // the mixture weight (w),
-    // the mean (nchannels values) and
-    // the covariance
-    bgmodel.create( 1, frameSize.height*frameSize.width*nmixtures*(2 + nchannels), CV_32F );
-    //make the array for keeping track of the used modes per pixel - all zeros at start
-    bgmodelUsedModes.create(frameSize,CV_8U);
-    bgmodelUsedModes = Scalar::all(0);
-}
-
-void BackgroundSubtractorMOG2::operator()(InputArray _image, OutputArray _fgmask, double learningRate)
+void BackgroundSubtractorMOG2Impl::apply(InputArray _image, OutputArray _fgmask, double learningRate)
 {
     Mat image = _image.getMat();
     bool needToInitialize = nframes == 0 || learningRate >= 1 || image.size() != frameSize || image.type() != frameType;
@@ -562,18 +700,19 @@ void BackgroundSubtractorMOG2::operator()(InputArray _image, OutputArray _fgmask
     learningRate = learningRate >= 0 && nframes > 1 ? learningRate : 1./std::min( 2*nframes, history );
     CV_Assert(learningRate >= 0);
 
-    parallel_for(BlockedRange(0, image.rows),
-                 MOG2Invoker(image, fgmask,
-                             (GMM*)bgmodel.data,
-                             (float*)(bgmodel.data + sizeof(GMM)*nmixtures*image.rows*image.cols),
-                             bgmodelUsedModes.data, nmixtures, (float)learningRate,
-                             (float)varThreshold,
-                             backgroundRatio, varThresholdGen,
-                             fVarInit, fVarMin, fVarMax, float(-learningRate*fCT), fTau,
-                             bShadowDetection, nShadowDetection));
+    parallel_for_(Range(0, image.rows),
+                  MOG2Invoker(image, fgmask,
+                              (GMM*)bgmodel.data,
+                              (float*)(bgmodel.data + sizeof(GMM)*nmixtures*image.rows*image.cols),
+                              bgmodelUsedModes.data, nmixtures, (float)learningRate,
+                              (float)varThreshold,
+                              backgroundRatio, varThresholdGen,
+                              fVarInit, fVarMin, fVarMax, float(-learningRate*fCT), fTau,
+                              bShadowDetection, nShadowDetection),
+                              image.total()/(double)(1 << 16));
 }
 
-void BackgroundSubtractorMOG2::getBackgroundImage(OutputArray backgroundImage) const
+void BackgroundSubtractorMOG2Impl::getBackgroundImage(OutputArray backgroundImage) const
 {
     int nchannels = CV_MAT_CN(frameType);
     CV_Assert( nchannels == 3 );
@@ -622,10 +761,17 @@ void BackgroundSubtractorMOG2::getBackgroundImage(OutputArray backgroundImage) c
     }
 
     default:
-        CV_Error(CV_StsUnsupportedFormat, "");
+        CV_Error(Error::StsUnsupportedFormat, "");
     }
 }
 
+
+Ptr<BackgroundSubtractorMOG2> createBackgroundSubtractorMOG2(int _history, double _varThreshold,
+                                                             bool _bShadowDetection)
+{
+    return makePtr<BackgroundSubtractorMOG2Impl>(_history, (float)_varThreshold, _bShadowDetection);
+}
+
 }
 
 /* End of file. */
diff --git a/modules/video/src/bgfg_gmg.cpp b/modules/video/src/bgfg_gmg.cpp
index 42d6fb1691..f5b7881a0a 100644
--- a/modules/video/src/bgfg_gmg.cpp
+++ b/modules/video/src/bgfg_gmg.cpp
@@ -8,8 +8,10 @@
 //
 //
 //                          License Agreement
+//                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -22,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -47,37 +49,168 @@
  */
 
 #include "precomp.hpp"
+#include <limits>
 
-cv::BackgroundSubtractorGMG::BackgroundSubtractorGMG()
+namespace cv
 {
-    /*
-     * Default Parameter Values. Override with algorithm "set" method.
+
+class BackgroundSubtractorGMGImpl : public BackgroundSubtractorGMG
+{
+public:
+    BackgroundSubtractorGMGImpl()
+    {
+        /*
+         * Default Parameter Values. Override with algorithm "set" method.
+         */
+        maxFeatures = 64;
+        learningRate = 0.025;
+        numInitializationFrames = 120;
+        quantizationLevels = 16;
+        backgroundPrior = 0.8;
+        decisionThreshold = 0.8;
+        smoothingRadius = 7;
+        updateBackgroundModel = true;
+        minVal_ = maxVal_ = 0;
+        name_ = "BackgroundSubtractor.GMG";
+    }
+
+    ~BackgroundSubtractorGMGImpl()
+    {
+    }
+
+    virtual AlgorithmInfo* info() const { return 0; }
+
+    /**
+     * Validate parameters and set up data structures for appropriate image size.
+     * Must call before running on data.
+     * @param frameSize input frame size
+     * @param min       minimum value taken on by pixels in image sequence. Usually 0
+     * @param max       maximum value taken on by pixels in image sequence. e.g. 1.0 or 255
      */
-    maxFeatures = 64;
-    learningRate = 0.025;
-    numInitializationFrames = 120;
-    quantizationLevels = 16;
-    backgroundPrior = 0.8;
-    decisionThreshold = 0.8;
-    smoothingRadius = 7;
-    updateBackgroundModel = true;
-}
+    void initialize(Size frameSize, double minVal, double maxVal);
 
-cv::BackgroundSubtractorGMG::~BackgroundSubtractorGMG()
-{
-}
+    /**
+     * Performs single-frame background subtraction and builds up a statistical background image
+     * model.
+     * @param image Input image
+     * @param fgmask Output mask image representing foreground and background pixels
+     */
+    virtual void apply(InputArray image, OutputArray fgmask, double learningRate=-1.0);
 
-void cv::BackgroundSubtractorGMG::initialize(cv::Size frameSize, double min, double max)
+    /**
+     * Releases all inner buffers.
+     */
+    void release();
+
+    virtual int getMaxFeatures() const { return maxFeatures; }
+    virtual void setMaxFeatures(int _maxFeatures) { maxFeatures = _maxFeatures; }
+
+    virtual double getDefaultLearningRate() const { return learningRate; }
+    virtual void setDefaultLearningRate(double lr) { learningRate = lr; }
+
+    virtual int getNumFrames() const { return numInitializationFrames; }
+    virtual void setNumFrames(int nframes) { numInitializationFrames = nframes; }
+
+    virtual int getQuantizationLevels() const { return quantizationLevels; }
+    virtual void setQuantizationLevels(int nlevels) { quantizationLevels = nlevels; }
+
+    virtual double getBackgroundPrior() const { return backgroundPrior; }
+    virtual void setBackgroundPrior(double bgprior) { backgroundPrior = bgprior; }
+
+    virtual int getSmoothingRadius() const { return smoothingRadius; }
+    virtual void setSmoothingRadius(int radius) { smoothingRadius = radius; }
+
+    virtual double getDecisionThreshold() const { return decisionThreshold; }
+    virtual void setDecisionThreshold(double thresh) { decisionThreshold = thresh; }
+
+    virtual bool getUpdateBackgroundModel() const { return updateBackgroundModel; }
+    virtual void setUpdateBackgroundModel(bool update) { updateBackgroundModel = update; }
+
+    virtual double getMinVal() const { return minVal_; }
+    virtual void setMinVal(double val) { minVal_ = val; }
+
+    virtual double getMaxVal() const  { return maxVal_; }
+    virtual void setMaxVal(double val)  { maxVal_ = val; }
+
+    virtual void getBackgroundImage(OutputArray) const
+    {
+        CV_Error( Error::StsNotImplemented, "" );
+    }
+
+    virtual void write(FileStorage& fs) const
+    {
+        fs << "name" << name_
+        << "maxFeatures" << maxFeatures
+        << "defaultLearningRate" << learningRate
+        << "numFrames" << numInitializationFrames
+        << "quantizationLevels" << quantizationLevels
+        << "backgroundPrior" << backgroundPrior
+        << "decisionThreshold" << decisionThreshold
+        << "smoothingRadius" << smoothingRadius
+        << "updateBackgroundModel" << (int)updateBackgroundModel;
+        // we do not save minVal_ & maxVal_, since they depend on the image type.
+    }
+
+    virtual void read(const FileNode& fn)
+    {
+        CV_Assert( (String)fn["name"] == name_ );
+        maxFeatures = (int)fn["maxFeatures"];
+        learningRate = (double)fn["defaultLearningRate"];
+        numInitializationFrames = (int)fn["numFrames"];
+        quantizationLevels = (int)fn["quantizationLevels"];
+        backgroundPrior = (double)fn["backgroundPrior"];
+        smoothingRadius = (int)fn["smoothingRadius"];
+        decisionThreshold = (double)fn["decisionThreshold"];
+        updateBackgroundModel = (int)fn["updateBackgroundModel"] != 0;
+        minVal_ = maxVal_ = 0;
+        frameSize_ = Size();
+    }
+
+    //! Total number of distinct colors to maintain in histogram.
+    int     maxFeatures;
+    //! Set between 0.0 and 1.0, determines how quickly features are "forgotten" from histograms.
+    double  learningRate;
+    //! Number of frames of video to use to initialize histograms.
+    int     numInitializationFrames;
+    //! Number of discrete levels in each channel to be used in histograms.
+    int     quantizationLevels;
+    //! Prior probability that any given pixel is a background pixel. A sensitivity parameter.
+    double  backgroundPrior;
+    //! Value above which pixel is determined to be FG.
+    double  decisionThreshold;
+    //! Smoothing radius, in pixels, for cleaning up FG image.
+    int     smoothingRadius;
+    //! Perform background model update
+    bool updateBackgroundModel;
+
+private:
+    double maxVal_;
+    double minVal_;
+
+    Size frameSize_;
+    int frameNum_;
+
+    String name_;
+
+    Mat_<int> nfeatures_;
+    Mat_<unsigned int> colors_;
+    Mat_<float> weights_;
+
+    Mat buf_;
+};
+
+
+void BackgroundSubtractorGMGImpl::initialize(Size frameSize, double minVal, double maxVal)
 {
-    CV_Assert(min < max);
+    CV_Assert(minVal < maxVal);
     CV_Assert(maxFeatures > 0);
     CV_Assert(learningRate >= 0.0 && learningRate <= 1.0);
     CV_Assert(numInitializationFrames >= 1);
     CV_Assert(quantizationLevels >= 1 && quantizationLevels <= 255);
     CV_Assert(backgroundPrior >= 0.0 && backgroundPrior <= 1.0);
 
-    minVal_ = min;
-    maxVal_ = max;
+    minVal_ = minVal;
+    maxVal_ = maxVal;
 
     frameSize_ = frameSize;
     frameNum_ = 0;
@@ -86,7 +219,7 @@ void cv::BackgroundSubtractorGMG::initialize(cv::Size frameSize, double min, dou
     colors_.create(frameSize_.area(), maxFeatures);
     weights_.create(frameSize_.area(), maxFeatures);
 
-    nfeatures_.setTo(cv::Scalar::all(0));
+    nfeatures_.setTo(Scalar::all(0));
 }
 
 namespace
@@ -181,10 +314,10 @@ namespace
         }
     };
 
-    class GMG_LoopBody : public cv::ParallelLoopBody
+    class GMG_LoopBody : public ParallelLoopBody
     {
     public:
-        GMG_LoopBody(const cv::Mat& frame, const cv::Mat& fgmask, const cv::Mat_<int>& nfeatures, const cv::Mat_<unsigned int>& colors, const cv::Mat_<float>& weights,
+        GMG_LoopBody(const Mat& frame, const Mat& fgmask, const Mat_<int>& nfeatures, const Mat_<unsigned int>& colors, const Mat_<float>& weights,
                      int maxFeatures, double learningRate, int numInitializationFrames, int quantizationLevels, double backgroundPrior, double decisionThreshold,
                      double maxVal, double minVal, int frameNum, bool updateBackgroundModel) :
             frame_(frame), fgmask_(fgmask), nfeatures_(nfeatures), colors_(colors), weights_(weights),
@@ -194,16 +327,16 @@ namespace
         {
         }
 
-        void operator() (const cv::Range& range) const;
+        void operator() (const Range& range) const;
 
     private:
-        cv::Mat frame_;
+        Mat frame_;
 
-        mutable cv::Mat_<uchar> fgmask_;
+        mutable Mat_<uchar> fgmask_;
 
-        mutable cv::Mat_<int> nfeatures_;
-        mutable cv::Mat_<unsigned int> colors_;
-        mutable cv::Mat_<float> weights_;
+        mutable Mat_<int> nfeatures_;
+        mutable Mat_<unsigned int> colors_;
+        mutable Mat_<float> weights_;
 
         int     maxFeatures_;
         double  learningRate_;
@@ -218,7 +351,7 @@ namespace
         int frameNum_;
     };
 
-    void GMG_LoopBody::operator() (const cv::Range& range) const
+    void GMG_LoopBody::operator() (const Range& range) const
     {
         typedef unsigned int (*func_t)(const void* src_, int x, int cn, double minVal, double maxVal, int quantizationLevels);
         static const func_t funcs[] =
@@ -296,7 +429,7 @@ namespace
     }
 }
 
-void cv::BackgroundSubtractorGMG::operator ()(InputArray _frame, OutputArray _fgmask, double newLearningRate)
+void BackgroundSubtractorGMGImpl::apply(InputArray _frame, OutputArray _fgmask, double newLearningRate)
 {
     Mat frame = _frame.getMat();
 
@@ -310,7 +443,16 @@ void cv::BackgroundSubtractorGMG::operator ()(InputArray _frame, OutputArray _fg
     }
 
     if (frame.size() != frameSize_)
-        initialize(frame.size(), 0.0, frame.depth() == CV_8U ? 255.0 : frame.depth() == CV_16U ? std::numeric_limits<ushort>::max() : 1.0);
+    {
+        double minval = minVal_;
+        double maxval = maxVal_;
+        if( minVal_ == 0 && maxVal_ == 0 )
+        {
+            minval = 0;
+            maxval = frame.depth() == CV_8U ? 255.0 : frame.depth() == CV_16U ? std::numeric_limits<ushort>::max() : 1.0;
+        }
+        initialize(frame.size(), minval, maxval);
+    }
 
     _fgmask.create(frameSize_, CV_8UC1);
     Mat fgmask = _fgmask.getMat();
@@ -323,19 +465,58 @@ void cv::BackgroundSubtractorGMG::operator ()(InputArray _frame, OutputArray _fg
     if (smoothingRadius > 0)
     {
         medianBlur(fgmask, buf_, smoothingRadius);
-        cv::swap(fgmask, buf_);
+        swap(fgmask, buf_);
     }
 
     // keep track of how many frames we have processed
     ++frameNum_;
 }
 
-void cv::BackgroundSubtractorGMG::release()
+void BackgroundSubtractorGMGImpl::release()
 {
-    frameSize_ = cv::Size();
+    frameSize_ = Size();
 
     nfeatures_.release();
     colors_.release();
     weights_.release();
     buf_.release();
 }
+
+
+Ptr<BackgroundSubtractorGMG> createBackgroundSubtractorGMG(int initializationFrames, double decisionThreshold)
+{
+    Ptr<BackgroundSubtractorGMG> bgfg = makePtr<BackgroundSubtractorGMGImpl>();
+    bgfg->setNumFrames(initializationFrames);
+    bgfg->setDecisionThreshold(decisionThreshold);
+
+    return bgfg;
+}
+
+/*
+ ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+ CV_INIT_ALGORITHM(BackgroundSubtractorGMG, "BackgroundSubtractor.GMG",
+ obj.info()->addParam(obj, "maxFeatures", obj.maxFeatures,false,0,0,
+ "Maximum number of features to store in histogram. Harsh enforcement of sparsity constraint.");
+ obj.info()->addParam(obj, "learningRate", obj.learningRate,false,0,0,
+ "Adaptation rate of histogram. Close to 1, slow adaptation. Close to 0, fast adaptation, features forgotten quickly.");
+ obj.info()->addParam(obj, "initializationFrames", obj.numInitializationFrames,false,0,0,
+ "Number of frames to use to initialize histograms of pixels.");
+ obj.info()->addParam(obj, "quantizationLevels", obj.quantizationLevels,false,0,0,
+ "Number of discrete colors to be used in histograms. Up-front quantization.");
+ obj.info()->addParam(obj, "backgroundPrior", obj.backgroundPrior,false,0,0,
+ "Prior probability that each individual pixel is a background pixel.");
+ obj.info()->addParam(obj, "smoothingRadius", obj.smoothingRadius,false,0,0,
+ "Radius of smoothing kernel to filter noise from FG mask image.");
+ obj.info()->addParam(obj, "decisionThreshold", obj.decisionThreshold,false,0,0,
+ "Threshold for FG decision rule. Pixel is FG if posterior probability exceeds threshold.");
+ obj.info()->addParam(obj, "updateBackgroundModel", obj.updateBackgroundModel,false,0,0,
+ "Perform background model update.");
+ obj.info()->addParam(obj, "minVal", obj.minVal_,false,0,0,
+ "Minimum of the value range (mostly for regression testing)");
+ obj.info()->addParam(obj, "maxVal", obj.maxVal_,false,0,0,
+ "Maximum of the value range (mostly for regression testing)");
+ );
+*/
+
+}
diff --git a/modules/video/src/camshift.cpp b/modules/video/src/camshift.cpp
index eadf166c42..9ba02381d8 100644
--- a/modules/video/src/camshift.cpp
+++ b/modules/video/src/camshift.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -40,285 +41,156 @@
 //M*/
 #include "precomp.hpp"
 
-
-/*F///////////////////////////////////////////////////////////////////////////////////////
-//    Name:    cvMeanShift
-//    Purpose: MeanShift algorithm
-//    Context:
-//    Parameters:
-//      imgProb     - 2D object probability distribution
-//      windowIn    - CvRect of CAMSHIFT Window intial size
-//      numIters    - If CAMSHIFT iterates this many times, stop
-//      windowOut   - Location, height and width of converged CAMSHIFT window
-//      len         - If != NULL, return equivalent len
-//      width       - If != NULL, return equivalent width
-//    Returns:
-//      Number of iterations CAMSHIFT took to converge
-//    Notes:
-//F*/
-CV_IMPL int
-cvMeanShift( const void* imgProb, CvRect windowIn,
-             CvTermCriteria criteria, CvConnectedComp* comp )
+int cv::meanShift( InputArray _probImage, Rect& window, TermCriteria criteria )
 {
-    CvMoments moments;
-    int    i = 0, eps;
-    CvMat  stub, *mat = (CvMat*)imgProb;
-    CvMat  cur_win;
-    CvRect cur_rect = windowIn;
+    Mat mat = _probImage.getMat();
+    Rect cur_rect = window;
 
-    if( comp )
-        comp->rect = windowIn;
+    CV_Assert( mat.channels() == 1 );
 
-    moments.m00 = moments.m10 = moments.m01 = 0;
+    if( window.height <= 0 || window.width <= 0 )
+        CV_Error( Error::StsBadArg, "Input window has non-positive sizes" );
 
-    mat = cvGetMat( mat, &stub );
+    window = window & Rect(0, 0, mat.cols, mat.rows);
 
-    if( CV_MAT_CN( mat->type ) > 1 )
-        CV_Error( CV_BadNumChannels, cvUnsupportedFormat );
+    double eps = (criteria.type & TermCriteria::EPS) ? std::max(criteria.epsilon, 0.) : 1.;
+    eps = cvRound(eps*eps);
+    int i, niters = (criteria.type & TermCriteria::MAX_ITER) ? std::max(criteria.maxCount, 1) : 100;
 
-    if( windowIn.height <= 0 || windowIn.width <= 0 )
-        CV_Error( CV_StsBadArg, "Input window has non-positive sizes" );
-
-    windowIn = cv::Rect(windowIn) & cv::Rect(0, 0, mat->cols, mat->rows);
-
-    criteria = cvCheckTermCriteria( criteria, 1., 100 );
-    eps = cvRound( criteria.epsilon * criteria.epsilon );
-
-    for( i = 0; i < criteria.max_iter; i++ )
+    for( i = 0; i < niters; i++ )
     {
-        int dx, dy, nx, ny;
-        double inv_m00;
-        cur_rect = cv::Rect(cur_rect) & cv::Rect(0, 0, mat->cols, mat->rows);
-        if( cv::Rect(cur_rect) == cv::Rect() )
+        cur_rect = cur_rect & Rect(0, 0, mat.cols, mat.rows);
+        if( cur_rect == Rect() )
         {
-            cur_rect.x = mat->cols/2;
-            cur_rect.y = mat->rows/2;
+            cur_rect.x = mat.cols/2;
+            cur_rect.y = mat.rows/2;
         }
-        cur_rect.width = MAX(cur_rect.width, 1);
-        cur_rect.height = MAX(cur_rect.height, 1);
+        cur_rect.width = std::max(cur_rect.width, 1);
+        cur_rect.height = std::max(cur_rect.height, 1);
 
-        cvGetSubRect( mat, &cur_win, cur_rect );
-        cvMoments( &cur_win, &moments );
+        Moments m = moments(mat(cur_rect));
 
-        /* Calculating center of mass */
-        if( fabs(moments.m00) < DBL_EPSILON )
+        // Calculating center of mass
+        if( fabs(m.m00) < DBL_EPSILON )
             break;
 
-        inv_m00 = moments.inv_sqrt_m00*moments.inv_sqrt_m00;
-        dx = cvRound( moments.m10 * inv_m00 - windowIn.width*0.5 );
-        dy = cvRound( moments.m01 * inv_m00 - windowIn.height*0.5 );
+        int dx = cvRound( m.m10/m.m00 - window.width*0.5 );
+        int dy = cvRound( m.m01/m.m00 - window.height*0.5 );
 
-        nx = cur_rect.x + dx;
-        ny = cur_rect.y + dy;
-
-        if( nx < 0 )
-            nx = 0;
-        else if( nx + cur_rect.width > mat->cols )
-            nx = mat->cols - cur_rect.width;
-
-        if( ny < 0 )
-            ny = 0;
-        else if( ny + cur_rect.height > mat->rows )
-            ny = mat->rows - cur_rect.height;
+        int nx = std::min(std::max(cur_rect.x + dx, 0), mat.cols - cur_rect.width);
+        int ny = std::min(std::max(cur_rect.y + dy, 0), mat.rows - cur_rect.height);
 
         dx = nx - cur_rect.x;
         dy = ny - cur_rect.y;
         cur_rect.x = nx;
         cur_rect.y = ny;
 
-        /* Check for coverage centers mass & window */
+        // Check for coverage centers mass & window
         if( dx*dx + dy*dy < eps )
             break;
     }
 
-    if( comp )
-    {
-        comp->rect = cur_rect;
-        comp->area = (float)moments.m00;
-    }
-
+    window = cur_rect;
     return i;
 }
 
 
-/*F///////////////////////////////////////////////////////////////////////////////////////
-//    Name:    cvCamShift
-//    Purpose: CAMSHIFT algorithm
-//    Context:
-//    Parameters:
-//      imgProb     - 2D object probability distribution
-//      windowIn    - CvRect of CAMSHIFT Window intial size
-//      criteria    - criteria of stop finding window
-//      windowOut   - Location, height and width of converged CAMSHIFT window
-//      orientation - If != NULL, return distribution orientation
-//      len         - If != NULL, return equivalent len
-//      width       - If != NULL, return equivalent width
-//      area        - sum of all elements in result window
-//    Returns:
-//      Number of iterations CAMSHIFT took to converge
-//    Notes:
-//F*/
-CV_IMPL int
-cvCamShift( const void* imgProb, CvRect windowIn,
-            CvTermCriteria criteria,
-            CvConnectedComp* _comp,
-            CvBox2D* box )
-{
-    const int TOLERANCE = 10;
-    CvMoments moments;
-    double m00 = 0, m10, m01, mu20, mu11, mu02, inv_m00;
-    double a, b, c, xc, yc;
-    double rotate_a, rotate_c;
-    double theta = 0, square;
-    double cs, sn;
-    double length = 0, width = 0;
-    int itersUsed = 0;
-    CvConnectedComp comp;
-    CvMat  cur_win, stub, *mat = (CvMat*)imgProb;
-
-    comp.rect = windowIn;
-
-    mat = cvGetMat( mat, &stub );
-
-    itersUsed = cvMeanShift( mat, windowIn, criteria, &comp );
-    windowIn = comp.rect;
-
-    windowIn.x -= TOLERANCE;
-    if( windowIn.x < 0 )
-        windowIn.x = 0;
-
-    windowIn.y -= TOLERANCE;
-    if( windowIn.y < 0 )
-        windowIn.y = 0;
-
-    windowIn.width += 2 * TOLERANCE;
-    if( windowIn.x + windowIn.width > mat->width )
-        windowIn.width = mat->width - windowIn.x;
-
-    windowIn.height += 2 * TOLERANCE;
-    if( windowIn.y + windowIn.height > mat->height )
-        windowIn.height = mat->height - windowIn.y;
-
-    cvGetSubRect( mat, &cur_win, windowIn );
-
-    /* Calculating moments in new center mass */
-    cvMoments( &cur_win, &moments );
-
-    m00 = moments.m00;
-    m10 = moments.m10;
-    m01 = moments.m01;
-    mu11 = moments.mu11;
-    mu20 = moments.mu20;
-    mu02 = moments.mu02;
-
-    if( fabs(m00) < DBL_EPSILON )
-        return -1;
-
-    inv_m00 = 1. / m00;
-    xc = cvRound( m10 * inv_m00 + windowIn.x );
-    yc = cvRound( m01 * inv_m00 + windowIn.y );
-    a = mu20 * inv_m00;
-    b = mu11 * inv_m00;
-    c = mu02 * inv_m00;
-
-    /* Calculating width & height */
-    square = sqrt( 4 * b * b + (a - c) * (a - c) );
-
-    /* Calculating orientation */
-    theta = atan2( 2 * b, a - c + square );
-
-    /* Calculating width & length of figure */
-    cs = cos( theta );
-    sn = sin( theta );
-
-    rotate_a = cs * cs * mu20 + 2 * cs * sn * mu11 + sn * sn * mu02;
-    rotate_c = sn * sn * mu20 - 2 * cs * sn * mu11 + cs * cs * mu02;
-    length = sqrt( rotate_a * inv_m00 ) * 4;
-    width = sqrt( rotate_c * inv_m00 ) * 4;
-
-    /* In case, when tetta is 0 or 1.57... the Length & Width may be exchanged */
-    if( length < width )
-    {
-        double t;
-
-        CV_SWAP( length, width, t );
-        CV_SWAP( cs, sn, t );
-        theta = CV_PI*0.5 - theta;
-    }
-
-    /* Saving results */
-    if( _comp || box )
-    {
-        int t0, t1;
-        int _xc = cvRound( xc );
-        int _yc = cvRound( yc );
-
-        t0 = cvRound( fabs( length * cs ));
-        t1 = cvRound( fabs( width * sn ));
-
-        t0 = MAX( t0, t1 ) + 2;
-        comp.rect.width = MIN( t0, (mat->width - _xc) * 2 );
-
-        t0 = cvRound( fabs( length * sn ));
-        t1 = cvRound( fabs( width * cs ));
-
-        t0 = MAX( t0, t1 ) + 2;
-        comp.rect.height = MIN( t0, (mat->height - _yc) * 2 );
-
-        comp.rect.x = MAX( 0, _xc - comp.rect.width / 2 );
-        comp.rect.y = MAX( 0, _yc - comp.rect.height / 2 );
-
-        comp.rect.width = MIN( mat->width - comp.rect.x, comp.rect.width );
-        comp.rect.height = MIN( mat->height - comp.rect.y, comp.rect.height );
-        comp.area = (float) m00;
-    }
-
-    if( _comp )
-        *_comp = comp;
-
-    if( box )
-    {
-        box->size.height = (float)length;
-        box->size.width = (float)width;
-        box->angle = (float)((CV_PI*0.5+theta)*180./CV_PI);
-        while(box->angle < 0)
-            box->angle += 360;
-        while(box->angle >= 360)
-            box->angle -= 360;
-        if(box->angle >= 180)
-            box->angle -= 180;
-        box->center = cvPoint2D32f( comp.rect.x + comp.rect.width*0.5f,
-                                    comp.rect.y + comp.rect.height*0.5f);
-    }
-
-    return itersUsed;
-}
-
-
 cv::RotatedRect cv::CamShift( InputArray _probImage, Rect& window,
                               TermCriteria criteria )
 {
-    CvConnectedComp comp;
-    CvBox2D box;
+    const int TOLERANCE = 10;
+    Mat mat = _probImage.getMat();
 
-    box.center.x = box.center.y = 0; box.angle = 0; box.size.width = box.size.height = 0;
-    comp.rect.x = comp.rect.y = comp.rect.width = comp.rect.height = 0;
+    meanShift( mat, window, criteria );
 
-    Mat probImage = _probImage.getMat();
-    CvMat c_probImage = probImage;
-    cvCamShift(&c_probImage, window, (CvTermCriteria)criteria, &comp, &box);
-    window = comp.rect;
-    return RotatedRect(Point2f(box.center), Size2f(box.size), box.angle);
-}
+    window.x -= TOLERANCE;
+    if( window.x < 0 )
+        window.x = 0;
 
-int cv::meanShift( InputArray _probImage, Rect& window, TermCriteria criteria )
-{
-    CvConnectedComp comp;
-    Mat probImage = _probImage.getMat();
-    CvMat c_probImage = probImage;
-    int iters = cvMeanShift(&c_probImage, window, (CvTermCriteria)criteria, &comp );
-    window = comp.rect;
-    return iters;
+    window.y -= TOLERANCE;
+    if( window.y < 0 )
+        window.y = 0;
+
+    window.width += 2 * TOLERANCE;
+    if( window.x + window.width > mat.cols )
+        window.width = mat.cols - window.x;
+
+    window.height += 2 * TOLERANCE;
+    if( window.y + window.height > mat.rows )
+        window.height = mat.rows - window.y;
+
+    // Calculating moments in new center mass
+    Moments m = moments( mat(window) );
+
+    double m00 = m.m00, m10 = m.m10, m01 = m.m01;
+    double mu11 = m.mu11, mu20 = m.mu20, mu02 = m.mu02;
+
+    if( fabs(m00) < DBL_EPSILON )
+        return RotatedRect();
+
+    double inv_m00 = 1. / m00;
+    int xc = cvRound( m10 * inv_m00 + window.x );
+    int yc = cvRound( m01 * inv_m00 + window.y );
+    double a = mu20 * inv_m00, b = mu11 * inv_m00, c = mu02 * inv_m00;
+
+    // Calculating width & height
+    double square = std::sqrt( 4 * b * b + (a - c) * (a - c) );
+
+    // Calculating orientation
+    double theta = atan2( 2 * b, a - c + square );
+
+    // Calculating width & length of figure
+    double cs = cos( theta );
+    double sn = sin( theta );
+
+    double rotate_a = cs * cs * mu20 + 2 * cs * sn * mu11 + sn * sn * mu02;
+    double rotate_c = sn * sn * mu20 - 2 * cs * sn * mu11 + cs * cs * mu02;
+    double length = std::sqrt( rotate_a * inv_m00 ) * 4;
+    double width = std::sqrt( rotate_c * inv_m00 ) * 4;
+
+    // In case, when tetta is 0 or 1.57... the Length & Width may be exchanged
+    if( length < width )
+    {
+        std::swap( length, width );
+        std::swap( cs, sn );
+        theta = CV_PI*0.5 - theta;
+    }
+
+    // Saving results
+    int _xc = cvRound( xc );
+    int _yc = cvRound( yc );
+
+    int t0 = cvRound( fabs( length * cs ));
+    int t1 = cvRound( fabs( width * sn ));
+
+    t0 = MAX( t0, t1 ) + 2;
+    window.width = MIN( t0, (mat.cols - _xc) * 2 );
+
+    t0 = cvRound( fabs( length * sn ));
+    t1 = cvRound( fabs( width * cs ));
+
+    t0 = MAX( t0, t1 ) + 2;
+    window.height = MIN( t0, (mat.rows - _yc) * 2 );
+
+    window.x = MAX( 0, _xc - window.width / 2 );
+    window.y = MAX( 0, _yc - window.height / 2 );
+
+    window.width = MIN( mat.cols - window.x, window.width );
+    window.height = MIN( mat.rows - window.y, window.height );
+
+    RotatedRect box;
+    box.size.height = (float)length;
+    box.size.width = (float)width;
+    box.angle = (float)((CV_PI*0.5+theta)*180./CV_PI);
+    while(box.angle < 0)
+        box.angle += 360;
+    while(box.angle >= 360)
+        box.angle -= 360;
+    if(box.angle >= 180)
+        box.angle -= 180;
+    box.center = Point2f( window.x + window.width*0.5f, window.y + window.height*0.5f);
+
+    return box;
 }
 
 /* End of file. */
diff --git a/modules/video/src/compat_video.cpp b/modules/video/src/compat_video.cpp
new file mode 100644
index 0000000000..e6dc960303
--- /dev/null
+++ b/modules/video/src/compat_video.cpp
@@ -0,0 +1,389 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+#include "opencv2/video/tracking_c.h"
+
+
+/////////////////////////// Meanshift & CAMShift ///////////////////////////
+
+CV_IMPL int
+cvMeanShift( const void* imgProb, CvRect windowIn,
+             CvTermCriteria criteria, CvConnectedComp* comp )
+{
+    cv::Mat img = cv::cvarrToMat(imgProb);
+    cv::Rect window = windowIn;
+    int iters = cv::meanShift(img, window, criteria);
+
+    if( comp )
+    {
+        comp->rect = window;
+        comp->area = cvRound(cv::sum(img(window))[0]);
+    }
+
+    return iters;
+}
+
+
+CV_IMPL int
+cvCamShift( const void* imgProb, CvRect windowIn,
+            CvTermCriteria criteria,
+            CvConnectedComp* comp,
+            CvBox2D* box )
+{
+    cv::Mat img = cv::cvarrToMat(imgProb);
+    cv::Rect window = windowIn;
+    cv::RotatedRect rr = cv::CamShift(img, window, criteria);
+
+    if( comp )
+    {
+        comp->rect = window;
+        cv::Rect roi = rr.boundingRect() & cv::Rect(0, 0, img.cols, img.rows);
+        comp->area = cvRound(cv::sum(img(roi))[0]);
+    }
+
+    if( box )
+        *box = rr;
+
+    return rr.size.width*rr.size.height > 0.f ? 1 : -1;
+}
+
+
+///////////////////////// Motion Templates ////////////////////////////
+
+CV_IMPL void
+cvUpdateMotionHistory( const void* silhouette, void* mhimg,
+                       double timestamp, double mhi_duration )
+{
+    cv::Mat silh = cv::cvarrToMat(silhouette), mhi = cv::cvarrToMat(mhimg);
+    cv::updateMotionHistory(silh, mhi, timestamp, mhi_duration);
+}
+
+
+CV_IMPL void
+cvCalcMotionGradient( const CvArr* mhimg, CvArr* maskimg,
+                      CvArr* orientation,
+                      double delta1, double delta2,
+                      int aperture_size )
+{
+    cv::Mat mhi = cv::cvarrToMat(mhimg);
+    const cv::Mat mask = cv::cvarrToMat(maskimg), orient = cv::cvarrToMat(orientation);
+    cv::calcMotionGradient(mhi, mask, orient, delta1, delta2, aperture_size);
+}
+
+
+CV_IMPL double
+cvCalcGlobalOrientation( const void* orientation, const void* maskimg, const void* mhimg,
+                         double curr_mhi_timestamp, double mhi_duration )
+{
+    cv::Mat mhi = cv::cvarrToMat(mhimg);
+    cv::Mat mask = cv::cvarrToMat(maskimg), orient = cv::cvarrToMat(orientation);
+    return cv::calcGlobalOrientation(orient, mask, mhi, curr_mhi_timestamp, mhi_duration);
+}
+
+
+CV_IMPL CvSeq*
+cvSegmentMotion( const CvArr* mhimg, CvArr* segmaskimg, CvMemStorage* storage,
+                 double timestamp, double segThresh )
+{
+    cv::Mat mhi = cv::cvarrToMat(mhimg);
+    const cv::Mat segmask = cv::cvarrToMat(segmaskimg);
+    std::vector<cv::Rect> brs;
+    cv::segmentMotion(mhi, segmask, brs, timestamp, segThresh);
+    CvSeq* seq = cvCreateSeq(0, sizeof(CvSeq), sizeof(CvConnectedComp), storage);
+
+    CvConnectedComp comp;
+    memset(&comp, 0, sizeof(comp));
+    for( size_t i = 0; i < brs.size(); i++ )
+    {
+        cv::Rect roi = brs[i];
+        float compLabel = (float)(i+1);
+        int x, y, area = 0;
+
+        cv::Mat part = segmask(roi);
+        for( y = 0; y < roi.height; y++ )
+        {
+            const float* partptr = part.ptr<float>(y);
+            for( x = 0; x < roi.width; x++ )
+                area += partptr[x] == compLabel;
+        }
+
+        comp.value = cv::Scalar(compLabel);
+        comp.rect = roi;
+        comp.area = area;
+        cvSeqPush(seq, &comp);
+    }
+
+    return seq;
+}
+
+
+///////////////////////////////// Kalman ///////////////////////////////
+
+CV_IMPL CvKalman*
+cvCreateKalman( int DP, int MP, int CP )
+{
+    CvKalman *kalman = 0;
+
+    if( DP <= 0 || MP <= 0 )
+        CV_Error( CV_StsOutOfRange,
+        "state and measurement vectors must have positive number of dimensions" );
+
+    if( CP < 0 )
+        CP = DP;
+
+    /* allocating memory for the structure */
+    kalman = (CvKalman *)cvAlloc( sizeof( CvKalman ));
+    memset( kalman, 0, sizeof(*kalman));
+
+    kalman->DP = DP;
+    kalman->MP = MP;
+    kalman->CP = CP;
+
+    kalman->state_pre = cvCreateMat( DP, 1, CV_32FC1 );
+    cvZero( kalman->state_pre );
+
+    kalman->state_post = cvCreateMat( DP, 1, CV_32FC1 );
+    cvZero( kalman->state_post );
+
+    kalman->transition_matrix = cvCreateMat( DP, DP, CV_32FC1 );
+    cvSetIdentity( kalman->transition_matrix );
+
+    kalman->process_noise_cov = cvCreateMat( DP, DP, CV_32FC1 );
+    cvSetIdentity( kalman->process_noise_cov );
+
+    kalman->measurement_matrix = cvCreateMat( MP, DP, CV_32FC1 );
+    cvZero( kalman->measurement_matrix );
+
+    kalman->measurement_noise_cov = cvCreateMat( MP, MP, CV_32FC1 );
+    cvSetIdentity( kalman->measurement_noise_cov );
+
+    kalman->error_cov_pre = cvCreateMat( DP, DP, CV_32FC1 );
+
+    kalman->error_cov_post = cvCreateMat( DP, DP, CV_32FC1 );
+    cvZero( kalman->error_cov_post );
+
+    kalman->gain = cvCreateMat( DP, MP, CV_32FC1 );
+
+    if( CP > 0 )
+    {
+        kalman->control_matrix = cvCreateMat( DP, CP, CV_32FC1 );
+        cvZero( kalman->control_matrix );
+    }
+
+    kalman->temp1 = cvCreateMat( DP, DP, CV_32FC1 );
+    kalman->temp2 = cvCreateMat( MP, DP, CV_32FC1 );
+    kalman->temp3 = cvCreateMat( MP, MP, CV_32FC1 );
+    kalman->temp4 = cvCreateMat( MP, DP, CV_32FC1 );
+    kalman->temp5 = cvCreateMat( MP, 1, CV_32FC1 );
+
+#if 1
+    kalman->PosterState = kalman->state_pre->data.fl;
+    kalman->PriorState = kalman->state_post->data.fl;
+    kalman->DynamMatr = kalman->transition_matrix->data.fl;
+    kalman->MeasurementMatr = kalman->measurement_matrix->data.fl;
+    kalman->MNCovariance = kalman->measurement_noise_cov->data.fl;
+    kalman->PNCovariance = kalman->process_noise_cov->data.fl;
+    kalman->KalmGainMatr = kalman->gain->data.fl;
+    kalman->PriorErrorCovariance = kalman->error_cov_pre->data.fl;
+    kalman->PosterErrorCovariance = kalman->error_cov_post->data.fl;
+#endif
+
+    return kalman;
+}
+
+
+CV_IMPL void
+cvReleaseKalman( CvKalman** _kalman )
+{
+    CvKalman *kalman;
+
+    if( !_kalman )
+        CV_Error( CV_StsNullPtr, "" );
+
+    kalman = *_kalman;
+    if( !kalman )
+        return;
+
+    /* freeing the memory */
+    cvReleaseMat( &kalman->state_pre );
+    cvReleaseMat( &kalman->state_post );
+    cvReleaseMat( &kalman->transition_matrix );
+    cvReleaseMat( &kalman->control_matrix );
+    cvReleaseMat( &kalman->measurement_matrix );
+    cvReleaseMat( &kalman->process_noise_cov );
+    cvReleaseMat( &kalman->measurement_noise_cov );
+    cvReleaseMat( &kalman->error_cov_pre );
+    cvReleaseMat( &kalman->gain );
+    cvReleaseMat( &kalman->error_cov_post );
+    cvReleaseMat( &kalman->temp1 );
+    cvReleaseMat( &kalman->temp2 );
+    cvReleaseMat( &kalman->temp3 );
+    cvReleaseMat( &kalman->temp4 );
+    cvReleaseMat( &kalman->temp5 );
+
+    memset( kalman, 0, sizeof(*kalman));
+
+    /* deallocating the structure */
+    cvFree( _kalman );
+}
+
+
+CV_IMPL const CvMat*
+cvKalmanPredict( CvKalman* kalman, const CvMat* control )
+{
+    if( !kalman )
+        CV_Error( CV_StsNullPtr, "" );
+
+    /* update the state */
+    /* x'(k) = A*x(k) */
+    cvMatMulAdd( kalman->transition_matrix, kalman->state_post, 0, kalman->state_pre );
+
+    if( control && kalman->CP > 0 )
+        /* x'(k) = x'(k) + B*u(k) */
+        cvMatMulAdd( kalman->control_matrix, control, kalman->state_pre, kalman->state_pre );
+
+    /* update error covariance matrices */
+    /* temp1 = A*P(k) */
+    cvMatMulAdd( kalman->transition_matrix, kalman->error_cov_post, 0, kalman->temp1 );
+
+    /* P'(k) = temp1*At + Q */
+    cvGEMM( kalman->temp1, kalman->transition_matrix, 1, kalman->process_noise_cov, 1,
+                     kalman->error_cov_pre, CV_GEMM_B_T );
+
+    /* handle the case when there will be measurement before the next predict */
+    cvCopy(kalman->state_pre, kalman->state_post);
+
+    return kalman->state_pre;
+}
+
+
+CV_IMPL const CvMat*
+cvKalmanCorrect( CvKalman* kalman, const CvMat* measurement )
+{
+    if( !kalman || !measurement )
+        CV_Error( CV_StsNullPtr, "" );
+
+    /* temp2 = H*P'(k) */
+    cvMatMulAdd( kalman->measurement_matrix, kalman->error_cov_pre, 0, kalman->temp2 );
+    /* temp3 = temp2*Ht + R */
+    cvGEMM( kalman->temp2, kalman->measurement_matrix, 1,
+            kalman->measurement_noise_cov, 1, kalman->temp3, CV_GEMM_B_T );
+
+    /* temp4 = inv(temp3)*temp2 = Kt(k) */
+    cvSolve( kalman->temp3, kalman->temp2, kalman->temp4, CV_SVD );
+
+    /* K(k) */
+    cvTranspose( kalman->temp4, kalman->gain );
+
+    /* temp5 = z(k) - H*x'(k) */
+    cvGEMM( kalman->measurement_matrix, kalman->state_pre, -1, measurement, 1, kalman->temp5 );
+
+    /* x(k) = x'(k) + K(k)*temp5 */
+    cvMatMulAdd( kalman->gain, kalman->temp5, kalman->state_pre, kalman->state_post );
+
+    /* P(k) = P'(k) - K(k)*temp2 */
+    cvGEMM( kalman->gain, kalman->temp2, -1, kalman->error_cov_pre, 1,
+                     kalman->error_cov_post, 0 );
+
+    return kalman->state_post;
+}
+
+///////////////////////////////////// Optical Flow ////////////////////////////////
+
+CV_IMPL void
+cvCalcOpticalFlowPyrLK( const void* arrA, const void* arrB,
+                        void* /*pyrarrA*/, void* /*pyrarrB*/,
+                        const CvPoint2D32f * featuresA,
+                        CvPoint2D32f * featuresB,
+                        int count, CvSize winSize, int level,
+                        char *status, float *error,
+                        CvTermCriteria criteria, int flags )
+{
+    if( count <= 0 )
+        return;
+    CV_Assert( featuresA && featuresB );
+    cv::Mat A = cv::cvarrToMat(arrA), B = cv::cvarrToMat(arrB);
+    cv::Mat ptA(count, 1, CV_32FC2, (void*)featuresA);
+    cv::Mat ptB(count, 1, CV_32FC2, (void*)featuresB);
+    cv::Mat st, err;
+
+    if( status )
+        st = cv::Mat(count, 1, CV_8U, (void*)status);
+    if( error )
+        err = cv::Mat(count, 1, CV_32F, (void*)error);
+    cv::calcOpticalFlowPyrLK( A, B, ptA, ptB, st,
+                              error ? cv::_OutputArray(err) : cv::noArray(),
+                              winSize, level, criteria, flags);
+}
+
+
+CV_IMPL void cvCalcOpticalFlowFarneback(
+            const CvArr* _prev, const CvArr* _next,
+            CvArr* _flow, double pyr_scale, int levels,
+            int winsize, int iterations, int poly_n,
+            double poly_sigma, int flags )
+{
+    cv::Mat prev = cv::cvarrToMat(_prev), next = cv::cvarrToMat(_next);
+    cv::Mat flow = cv::cvarrToMat(_flow);
+    CV_Assert( flow.size() == prev.size() && flow.type() == CV_32FC2 );
+    cv::calcOpticalFlowFarneback( prev, next, flow, pyr_scale, levels,
+        winsize, iterations, poly_n, poly_sigma, flags );
+}
+
+
+CV_IMPL int
+cvEstimateRigidTransform( const CvArr* arrA, const CvArr* arrB, CvMat* arrM, int full_affine )
+{
+    cv::Mat matA = cv::cvarrToMat(arrA), matB = cv::cvarrToMat(arrB);
+    const cv::Mat matM0 = cv::cvarrToMat(arrM);
+
+    cv::Mat matM = cv::estimateRigidTransform(matA, matB, full_affine != 0);
+    if( matM.empty() )
+    {
+        matM = cv::cvarrToMat(arrM);
+        matM.setTo(cv::Scalar::all(0));
+        return 0;
+    }
+    matM.convertTo(matM0, matM0.type());
+    return 1;
+}
diff --git a/modules/video/src/ecc.cpp b/modules/video/src/ecc.cpp
index 6555c44ced..f56e1f24b5 100644
--- a/modules/video/src/ecc.cpp
+++ b/modules/video/src/ecc.cpp
@@ -305,23 +305,8 @@ static void update_warping_matrix_ECC (Mat& map_matrix, const Mat& update, const
         mapPtr[3] = (float) sin(new_theta);
         mapPtr[1] = -mapPtr[3];
     }
-
 }
 
-CV_IMPL double cvFindTransformECC (const CvArr* _image1, const CvArr* _image2,
-                                   CvMat* _map_matrix,
-                                   const int motionType,
-                                   const CvTermCriteria _criteria)
-{
-
-    Mat image1 = cvarrToMat(_image1);
-    Mat image2 = cvarrToMat(_image2);
-    Mat map_matrix = cvarrToMat(_map_matrix);
-    double cc = cv::findTransformECC(image1, image2, map_matrix, motionType,
-        TermCriteria(TermCriteria::EPS+TermCriteria::COUNT, _criteria.max_iter, _criteria.epsilon));
-
-    return cc;
-}
 
 double cv::findTransformECC(InputArray templateImage,
                             InputArray inputImage,
@@ -340,14 +325,14 @@ double cv::findTransformECC(InputArray templateImage,
 
 
     if( ! (src.type()==dst.type()))
-        CV_Error( CV_StsUnmatchedFormats, "Both input images must have the same data type" );
+        CV_Error( Error::StsUnmatchedFormats, "Both input images must have the same data type" );
 
     //accept only 1-channel images
     if( src.type() != CV_8UC1 && src.type()!= CV_32FC1)
-        CV_Error( CV_StsUnsupportedFormat, "Images must have 8uC1 or 32fC1 type");
+        CV_Error( Error::StsUnsupportedFormat, "Images must have 8uC1 or 32fC1 type");
 
     if( map.type() != CV_32FC1)
-        CV_Error( CV_StsUnsupportedFormat, "warpMatrix must be single-channel floating-point matrix");
+        CV_Error( Error::StsUnsupportedFormat, "warpMatrix must be single-channel floating-point matrix");
 
     CV_Assert (map.cols == 3);
     CV_Assert (map.rows == 2 || map.rows ==3);
@@ -443,8 +428,8 @@ double cv::findTransformECC(InputArray templateImage,
     Mat deltaP = Mat(numberOfParameters, 1, CV_32F);//transformation parameter correction
     Mat error = Mat(hs, ws, CV_32F);//error as 2D matrix
 
-    const int imageFlags = CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP;
-    const int maskFlags  = CV_INTER_NN+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP;
+    const int imageFlags = INTER_LINEAR  + WARP_INVERSE_MAP;
+    const int maskFlags  = INTER_NEAREST + WARP_INVERSE_MAP;
 
 
     // iteratively update map_matrix
@@ -519,7 +504,7 @@ double cv::findTransformECC(InputArray templateImage,
         if (lambda_d <= 0.0)
         {
             rho = -1;
-            CV_Error(CV_StsNoConv, "The algorithm stopped before its convergence. The correlation is going to be minimized. Images may be uncorrelated or non-overlapped");
+            CV_Error(Error::StsNoConv, "The algorithm stopped before its convergence. The correlation is going to be minimized. Images may be uncorrelated or non-overlapped");
 
         }
         const double lambda = (lambda_n/lambda_d);
diff --git a/modules/video/src/kalman.cpp b/modules/video/src/kalman.cpp
index a5485751ec..793404a393 100644
--- a/modules/video/src/kalman.cpp
+++ b/modules/video/src/kalman.cpp
@@ -40,176 +40,6 @@
 //M*/
 #include "precomp.hpp"
 
-
-CV_IMPL CvKalman*
-cvCreateKalman( int DP, int MP, int CP )
-{
-    CvKalman *kalman = 0;
-
-    if( DP <= 0 || MP <= 0 )
-        CV_Error( CV_StsOutOfRange,
-        "state and measurement vectors must have positive number of dimensions" );
-
-    if( CP < 0 )
-        CP = DP;
-
-    /* allocating memory for the structure */
-    kalman = (CvKalman *)cvAlloc( sizeof( CvKalman ));
-    memset( kalman, 0, sizeof(*kalman));
-
-    kalman->DP = DP;
-    kalman->MP = MP;
-    kalman->CP = CP;
-
-    kalman->state_pre = cvCreateMat( DP, 1, CV_32FC1 );
-    cvZero( kalman->state_pre );
-
-    kalman->state_post = cvCreateMat( DP, 1, CV_32FC1 );
-    cvZero( kalman->state_post );
-
-    kalman->transition_matrix = cvCreateMat( DP, DP, CV_32FC1 );
-    cvSetIdentity( kalman->transition_matrix );
-
-    kalman->process_noise_cov = cvCreateMat( DP, DP, CV_32FC1 );
-    cvSetIdentity( kalman->process_noise_cov );
-
-    kalman->measurement_matrix = cvCreateMat( MP, DP, CV_32FC1 );
-    cvZero( kalman->measurement_matrix );
-
-    kalman->measurement_noise_cov = cvCreateMat( MP, MP, CV_32FC1 );
-    cvSetIdentity( kalman->measurement_noise_cov );
-
-    kalman->error_cov_pre = cvCreateMat( DP, DP, CV_32FC1 );
-
-    kalman->error_cov_post = cvCreateMat( DP, DP, CV_32FC1 );
-    cvZero( kalman->error_cov_post );
-
-    kalman->gain = cvCreateMat( DP, MP, CV_32FC1 );
-
-    if( CP > 0 )
-    {
-        kalman->control_matrix = cvCreateMat( DP, CP, CV_32FC1 );
-        cvZero( kalman->control_matrix );
-    }
-
-    kalman->temp1 = cvCreateMat( DP, DP, CV_32FC1 );
-    kalman->temp2 = cvCreateMat( MP, DP, CV_32FC1 );
-    kalman->temp3 = cvCreateMat( MP, MP, CV_32FC1 );
-    kalman->temp4 = cvCreateMat( MP, DP, CV_32FC1 );
-    kalman->temp5 = cvCreateMat( MP, 1, CV_32FC1 );
-
-#if 1
-    kalman->PosterState = kalman->state_pre->data.fl;
-    kalman->PriorState = kalman->state_post->data.fl;
-    kalman->DynamMatr = kalman->transition_matrix->data.fl;
-    kalman->MeasurementMatr = kalman->measurement_matrix->data.fl;
-    kalman->MNCovariance = kalman->measurement_noise_cov->data.fl;
-    kalman->PNCovariance = kalman->process_noise_cov->data.fl;
-    kalman->KalmGainMatr = kalman->gain->data.fl;
-    kalman->PriorErrorCovariance = kalman->error_cov_pre->data.fl;
-    kalman->PosterErrorCovariance = kalman->error_cov_post->data.fl;
-#endif
-
-    return kalman;
-}
-
-
-CV_IMPL void
-cvReleaseKalman( CvKalman** _kalman )
-{
-    CvKalman *kalman;
-
-    if( !_kalman )
-        CV_Error( CV_StsNullPtr, "" );
-
-    kalman = *_kalman;
-    if( !kalman )
-        return;
-
-    /* freeing the memory */
-    cvReleaseMat( &kalman->state_pre );
-    cvReleaseMat( &kalman->state_post );
-    cvReleaseMat( &kalman->transition_matrix );
-    cvReleaseMat( &kalman->control_matrix );
-    cvReleaseMat( &kalman->measurement_matrix );
-    cvReleaseMat( &kalman->process_noise_cov );
-    cvReleaseMat( &kalman->measurement_noise_cov );
-    cvReleaseMat( &kalman->error_cov_pre );
-    cvReleaseMat( &kalman->gain );
-    cvReleaseMat( &kalman->error_cov_post );
-    cvReleaseMat( &kalman->temp1 );
-    cvReleaseMat( &kalman->temp2 );
-    cvReleaseMat( &kalman->temp3 );
-    cvReleaseMat( &kalman->temp4 );
-    cvReleaseMat( &kalman->temp5 );
-
-    memset( kalman, 0, sizeof(*kalman));
-
-    /* deallocating the structure */
-    cvFree( _kalman );
-}
-
-
-CV_IMPL const CvMat*
-cvKalmanPredict( CvKalman* kalman, const CvMat* control )
-{
-    if( !kalman )
-        CV_Error( CV_StsNullPtr, "" );
-
-    /* update the state */
-    /* x'(k) = A*x(k) */
-    cvMatMulAdd( kalman->transition_matrix, kalman->state_post, 0, kalman->state_pre );
-
-    if( control && kalman->CP > 0 )
-        /* x'(k) = x'(k) + B*u(k) */
-        cvMatMulAdd( kalman->control_matrix, control, kalman->state_pre, kalman->state_pre );
-
-    /* update error covariance matrices */
-    /* temp1 = A*P(k) */
-    cvMatMulAdd( kalman->transition_matrix, kalman->error_cov_post, 0, kalman->temp1 );
-
-    /* P'(k) = temp1*At + Q */
-    cvGEMM( kalman->temp1, kalman->transition_matrix, 1, kalman->process_noise_cov, 1,
-                     kalman->error_cov_pre, CV_GEMM_B_T );
-
-    /* handle the case when there will be measurement before the next predict */
-    cvCopy(kalman->state_pre, kalman->state_post);
-
-    return kalman->state_pre;
-}
-
-
-CV_IMPL const CvMat*
-cvKalmanCorrect( CvKalman* kalman, const CvMat* measurement )
-{
-    if( !kalman || !measurement )
-        CV_Error( CV_StsNullPtr, "" );
-
-    /* temp2 = H*P'(k) */
-    cvMatMulAdd( kalman->measurement_matrix, kalman->error_cov_pre, 0, kalman->temp2 );
-    /* temp3 = temp2*Ht + R */
-    cvGEMM( kalman->temp2, kalman->measurement_matrix, 1,
-            kalman->measurement_noise_cov, 1, kalman->temp3, CV_GEMM_B_T );
-
-    /* temp4 = inv(temp3)*temp2 = Kt(k) */
-    cvSolve( kalman->temp3, kalman->temp2, kalman->temp4, CV_SVD );
-
-    /* K(k) */
-    cvTranspose( kalman->temp4, kalman->gain );
-
-    /* temp5 = z(k) - H*x'(k) */
-    cvGEMM( kalman->measurement_matrix, kalman->state_pre, -1, measurement, 1, kalman->temp5 );
-
-    /* x(k) = x'(k) + K(k)*temp5 */
-    cvMatMulAdd( kalman->gain, kalman->temp5, kalman->state_pre, kalman->state_post );
-
-    /* P(k) = P'(k) - K(k)*temp2 */
-    cvGEMM( kalman->gain, kalman->temp2, -1, kalman->error_cov_pre, 1,
-                     kalman->error_cov_post, 0 );
-
-    return kalman->state_post;
-}
-
 namespace cv
 {
 
@@ -297,5 +127,3 @@ const Mat& KalmanFilter::correct(const Mat& measurement)
 }
 
 };
-
-
diff --git a/modules/video/src/lkpyramid.cpp b/modules/video/src/lkpyramid.cpp
index 28477100fe..155737ba79 100644
--- a/modules/video/src/lkpyramid.cpp
+++ b/modules/video/src/lkpyramid.cpp
@@ -7,10 +7,11 @@
 //  copy or use the software.
 //
 //
-//                        Intel License Agreement
+//                           License Agreement
 //                For Open Source Computer Vision Library
 //
 // Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -23,7 +24,7 @@
 //     this list of conditions and the following disclaimer in the documentation
 //     and/or other materials provided with the distribution.
 //
-//   * The name of Intel Corporation may not be used to endorse or promote products
+//   * The name of the copyright holders may not be used to endorse or promote products
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
@@ -43,6 +44,8 @@
 #include <stdio.h>
 #include "lkpyramid.hpp"
 
+#define  CV_DESCALE(x,n)     (((x) + (1 << ((n)-1))) >> (n))
+
 namespace
 {
 static void calcSharrDeriv(const cv::Mat& src, cv::Mat& dst)
@@ -156,7 +159,15 @@ cv::detail::LKTrackerInvoker::LKTrackerInvoker(
     minEigThreshold = _minEigThreshold;
 }
 
-void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
+#if defined __arm__ && !CV_NEON
+typedef int64 acctype;
+typedef int itemtype;
+#else
+typedef float acctype;
+typedef float itemtype;
+#endif
+
+void cv::detail::LKTrackerInvoker::operator()(const Range& range) const
 {
     Point2f halfWin((winSize.width-1)*0.5f, (winSize.height-1)*0.5f);
     const Mat& I = *prevImg;
@@ -170,7 +181,7 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
     Mat IWinBuf(winSize, CV_MAKETYPE(derivDepth, cn), (deriv_type*)_buf);
     Mat derivIWinBuf(winSize, CV_MAKETYPE(derivDepth, cn2), (deriv_type*)_buf + winSize.area()*cn);
 
-    for( int ptidx = range.begin(); ptidx < range.end(); ptidx++ )
+    for( int ptidx = range.start; ptidx < range.end; ptidx++ )
     {
         Point2f prevPt = prevPts[ptidx]*(float)(1./(1 << level));
         Point2f nextPt;
@@ -215,7 +226,8 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
         int dstep = (int)(derivI.step/derivI.elemSize1());
         int stepI = (int)(I.step/I.elemSize1());
         int stepJ = (int)(J.step/J.elemSize1());
-        float A11 = 0, A12 = 0, A22 = 0;
+        acctype iA11 = 0, iA12 = 0, iA22 = 0;
+        float A11, A12, A22;
 
 #if CV_SSE2
         __m128i qw0 = _mm_set1_epi32(iw00 + (iw01 << 16));
@@ -292,9 +304,9 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
                 dIptr[0] = (short)ixval;
                 dIptr[1] = (short)iyval;
 
-                A11 += (float)(ixval*ixval);
-                A12 += (float)(ixval*iyval);
-                A22 += (float)(iyval*iyval);
+                iA11 += (itemtype)(ixval*ixval);
+                iA12 += (itemtype)(ixval*iyval);
+                iA22 += (itemtype)(iyval*iyval);
             }
         }
 
@@ -303,20 +315,20 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
         _mm_store_ps(A11buf, qA11);
         _mm_store_ps(A12buf, qA12);
         _mm_store_ps(A22buf, qA22);
-        A11 += A11buf[0] + A11buf[1] + A11buf[2] + A11buf[3];
-        A12 += A12buf[0] + A12buf[1] + A12buf[2] + A12buf[3];
-        A22 += A22buf[0] + A22buf[1] + A22buf[2] + A22buf[3];
+        iA11 += A11buf[0] + A11buf[1] + A11buf[2] + A11buf[3];
+        iA12 += A12buf[0] + A12buf[1] + A12buf[2] + A12buf[3];
+        iA22 += A22buf[0] + A22buf[1] + A22buf[2] + A22buf[3];
 #endif
 
-        A11 *= FLT_SCALE;
-        A12 *= FLT_SCALE;
-        A22 *= FLT_SCALE;
+        A11 = iA11*FLT_SCALE;
+        A12 = iA12*FLT_SCALE;
+        A22 = iA22*FLT_SCALE;
 
         float D = A11*A22 - A12*A12;
         float minEig = (A22 + A11 - std::sqrt((A11-A22)*(A11-A22) +
                         4.f*A12*A12))/(2*winSize.width*winSize.height);
 
-        if( err && (flags & CV_LKFLOW_GET_MIN_EIGENVALS) != 0 )
+        if( err && (flags & OPTFLOW_LK_GET_MIN_EIGENVALS) != 0 )
             err[ptidx] = (float)minEig;
 
         if( minEig < minEigThreshold || D < FLT_EPSILON )
@@ -350,7 +362,8 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
             iw01 = cvRound(a*(1.f - b)*(1 << W_BITS));
             iw10 = cvRound((1.f - a)*b*(1 << W_BITS));
             iw11 = (1 << W_BITS) - iw00 - iw01 - iw10;
-            float b1 = 0, b2 = 0;
+            acctype ib1 = 0, ib2 = 0;
+            float b1, b2;
 #if CV_SSE2
             qw0 = _mm_set1_epi32(iw00 + (iw01 << 16));
             qw1 = _mm_set1_epi32(iw10 + (iw11 << 16));
@@ -405,20 +418,20 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
                     int diff = CV_DESCALE(Jptr[x]*iw00 + Jptr[x+cn]*iw01 +
                                           Jptr[x+stepJ]*iw10 + Jptr[x+stepJ+cn]*iw11,
                                           W_BITS1-5) - Iptr[x];
-                    b1 += (float)(diff*dIptr[0]);
-                    b2 += (float)(diff*dIptr[1]);
+                    ib1 += (itemtype)(diff*dIptr[0]);
+                    ib2 += (itemtype)(diff*dIptr[1]);
                 }
             }
 
 #if CV_SSE2
             float CV_DECL_ALIGNED(16) bbuf[4];
             _mm_store_ps(bbuf, _mm_add_ps(qb0, qb1));
-            b1 += bbuf[0] + bbuf[2];
-            b2 += bbuf[1] + bbuf[3];
+            ib1 += bbuf[0] + bbuf[2];
+            ib2 += bbuf[1] + bbuf[3];
 #endif
 
-            b1 *= FLT_SCALE;
-            b2 *= FLT_SCALE;
+            b1 = ib1*FLT_SCALE;
+            b2 = ib2*FLT_SCALE;
 
             Point2f delta( (float)((A12*b2 - A22*b1) * D),
                           (float)((A12*b1 - A11*b2) * D));
@@ -439,7 +452,7 @@ void cv::detail::LKTrackerInvoker::operator()(const BlockedRange& range) const
             prevDelta = delta;
         }
 
-        if( status[ptidx] && err && level == 0 && (flags & CV_LKFLOW_GET_MIN_EIGENVALS) == 0 )
+        if( status[ptidx] && err && level == 0 && (flags & OPTFLOW_LK_GET_MIN_EIGENVALS) == 0 )
         {
             Point2f nextPoint = nextPts[ptidx] - halfWin;
             Point inextPoint;
@@ -733,878 +746,39 @@ void cv::calcOpticalFlowPyrLK( InputArray _prevImg, InputArray _nextImg,
         typedef cv::detail::LKTrackerInvoker LKTrackerInvoker;
 #endif
 
-        parallel_for(BlockedRange(0, npoints), LKTrackerInvoker(prevPyr[level * lvlStep1], derivI,
-                                                                nextPyr[level * lvlStep2], prevPts, nextPts,
-                                                                status, err,
-                                                                winSize, criteria, level, maxLevel,
-                                                                flags, (float)minEigThreshold));
+        parallel_for_(Range(0, npoints), LKTrackerInvoker(prevPyr[level * lvlStep1], derivI,
+                                                          nextPyr[level * lvlStep2], prevPts, nextPts,
+                                                          status, err,
+                                                          winSize, criteria, level, maxLevel,
+                                                          flags, (float)minEigThreshold));
     }
 }
 
-
-static int icvMinimalPyramidSize( CvSize imgSize )
+namespace cv
 {
-    return cvAlign(imgSize.width,8) * imgSize.height / 3;
-}
-
 
 static void
-icvInitPyramidalAlgorithm( const CvMat* imgA, const CvMat* imgB,
-                           CvMat* pyrA, CvMat* pyrB,
-                           int level, CvTermCriteria * criteria,
-                           int max_iters, int flags,
-                           uchar *** imgI, uchar *** imgJ,
-                           int **step, CvSize** size,
-                           double **scale, cv::AutoBuffer<uchar>* buffer )
+getRTMatrix( const Point2f* a, const Point2f* b,
+             int count, Mat& M, bool fullAffine )
 {
-    const int ALIGN = 8;
-    int pyrBytes, bufferBytes = 0, elem_size;
-    int level1 = level + 1;
+    CV_Assert( M.isContinuous() );
 
-    int i;
-    CvSize imgSize, levelSize;
-
-    *imgI = *imgJ = 0;
-    *step = 0;
-    *scale = 0;
-    *size = 0;
-
-    /* check input arguments */
-    if( ((flags & CV_LKFLOW_PYR_A_READY) != 0 && !pyrA) ||
-        ((flags & CV_LKFLOW_PYR_B_READY) != 0 && !pyrB) )
-        CV_Error( CV_StsNullPtr, "Some of the precomputed pyramids are missing" );
-
-    if( level < 0 )
-        CV_Error( CV_StsOutOfRange, "The number of pyramid levels is negative" );
-
-    switch( criteria->type )
+    if( fullAffine )
     {
-    case CV_TERMCRIT_ITER:
-        criteria->epsilon = 0.f;
-        break;
-    case CV_TERMCRIT_EPS:
-        criteria->max_iter = max_iters;
-        break;
-    case CV_TERMCRIT_ITER | CV_TERMCRIT_EPS:
-        break;
-    default:
-        assert( 0 );
-        CV_Error( CV_StsBadArg, "Invalid termination criteria" );
-    }
+        double sa[6][6]={{0.}}, sb[6]={0.};
+        Mat A( 6, 6, CV_64F, &sa[0][0] ), B( 6, 1, CV_64F, sb );
+        Mat MM = M.reshape(1, 6);
 
-    /* compare squared values */
-    criteria->epsilon *= criteria->epsilon;
-
-    /* set pointers and step for every level */
-    pyrBytes = 0;
-
-    imgSize = cvGetSize(imgA);
-    elem_size = CV_ELEM_SIZE(imgA->type);
-    levelSize = imgSize;
-
-    for( i = 1; i < level1; i++ )
-    {
-        levelSize.width = (levelSize.width + 1) >> 1;
-        levelSize.height = (levelSize.height + 1) >> 1;
-
-        int tstep = cvAlign(levelSize.width,ALIGN) * elem_size;
-        pyrBytes += tstep * levelSize.height;
-    }
-
-    assert( pyrBytes <= imgSize.width * imgSize.height * elem_size * 4 / 3 );
-
-    /* buffer_size = <size for patches> + <size for pyramids> */
-    bufferBytes = (int)((level1 >= 0) * ((pyrA->data.ptr == 0) +
-        (pyrB->data.ptr == 0)) * pyrBytes +
-        (sizeof(imgI[0][0]) * 2 + sizeof(step[0][0]) +
-         sizeof(size[0][0]) + sizeof(scale[0][0])) * level1);
-
-    buffer->allocate( bufferBytes );
-
-    *imgI = (uchar **) (uchar*)(*buffer);
-    *imgJ = *imgI + level1;
-    *step = (int *) (*imgJ + level1);
-    *scale = (double *) (*step + level1);
-    *size = (CvSize *)(*scale + level1);
-
-    imgI[0][0] = imgA->data.ptr;
-    imgJ[0][0] = imgB->data.ptr;
-    step[0][0] = imgA->step;
-    scale[0][0] = 1;
-    size[0][0] = imgSize;
-
-    if( level > 0 )
-    {
-        uchar *bufPtr = (uchar *) (*size + level1);
-        uchar *ptrA = pyrA->data.ptr;
-        uchar *ptrB = pyrB->data.ptr;
-
-        if( !ptrA )
+        for( int i = 0; i < count; i++ )
         {
-            ptrA = bufPtr;
-            bufPtr += pyrBytes;
-        }
+            sa[0][0] += a[i].x*a[i].x;
+            sa[0][1] += a[i].y*a[i].x;
+            sa[0][2] += a[i].x;
 
-        if( !ptrB )
-            ptrB = bufPtr;
+            sa[1][1] += a[i].y*a[i].y;
+            sa[1][2] += a[i].y;
 
-        levelSize = imgSize;
-
-        /* build pyramids for both frames */
-        for( i = 1; i <= level; i++ )
-        {
-            int levelBytes;
-            CvMat prev_level, next_level;
-
-            levelSize.width = (levelSize.width + 1) >> 1;
-            levelSize.height = (levelSize.height + 1) >> 1;
-
-            size[0][i] = levelSize;
-            step[0][i] = cvAlign( levelSize.width, ALIGN ) * elem_size;
-            scale[0][i] = scale[0][i - 1] * 0.5;
-
-            levelBytes = step[0][i] * levelSize.height;
-            imgI[0][i] = (uchar *) ptrA;
-            ptrA += levelBytes;
-
-            if( !(flags & CV_LKFLOW_PYR_A_READY) )
-            {
-                prev_level = cvMat( size[0][i-1].height, size[0][i-1].width, CV_8UC1 );
-                next_level = cvMat( size[0][i].height, size[0][i].width, CV_8UC1 );
-                cvSetData( &prev_level, imgI[0][i-1], step[0][i-1] );
-                cvSetData( &next_level, imgI[0][i], step[0][i] );
-                cvPyrDown( &prev_level, &next_level );
-            }
-
-            imgJ[0][i] = (uchar *) ptrB;
-            ptrB += levelBytes;
-
-            if( !(flags & CV_LKFLOW_PYR_B_READY) )
-            {
-                prev_level = cvMat( size[0][i-1].height, size[0][i-1].width, CV_8UC1 );
-                next_level = cvMat( size[0][i].height, size[0][i].width, CV_8UC1 );
-                cvSetData( &prev_level, imgJ[0][i-1], step[0][i-1] );
-                cvSetData( &next_level, imgJ[0][i], step[0][i] );
-                cvPyrDown( &prev_level, &next_level );
-            }
-        }
-    }
-}
-
-
-/* compute dI/dx and dI/dy */
-static void
-icvCalcIxIy_32f( const float* src, int src_step, float* dstX, float* dstY, int dst_step,
-                 CvSize src_size, const float* smooth_k, float* buffer0 )
-{
-    int src_width = src_size.width, dst_width = src_size.width-2;
-    int x, height = src_size.height - 2;
-    float* buffer1 = buffer0 + src_width;
-
-    src_step /= sizeof(src[0]);
-    dst_step /= sizeof(dstX[0]);
-
-    for( ; height--; src += src_step, dstX += dst_step, dstY += dst_step )
-    {
-        const float* src2 = src + src_step;
-        const float* src3 = src + src_step*2;
-
-        for( x = 0; x < src_width; x++ )
-        {
-            float t0 = (src3[x] + src[x])*smooth_k[0] + src2[x]*smooth_k[1];
-            float t1 = src3[x] - src[x];
-            buffer0[x] = t0; buffer1[x] = t1;
-        }
-
-        for( x = 0; x < dst_width; x++ )
-        {
-            float t0 = buffer0[x+2] - buffer0[x];
-            float t1 = (buffer1[x] + buffer1[x+2])*smooth_k[0] + buffer1[x+1]*smooth_k[1];
-            dstX[x] = t0; dstY[x] = t1;
-        }
-    }
-}
-
-
-#undef CV_8TO32F
-#define CV_8TO32F(a) (a)
-
-static const void*
-icvAdjustRect( const void* srcptr, int src_step, int pix_size,
-              CvSize src_size, CvSize win_size,
-              CvPoint ip, CvRect* pRect )
-{
-    CvRect rect;
-    const char* src = (const char*)srcptr;
-
-    if( ip.x >= 0 )
-    {
-        src += ip.x*pix_size;
-        rect.x = 0;
-    }
-    else
-    {
-        rect.x = -ip.x;
-        if( rect.x > win_size.width )
-            rect.x = win_size.width;
-    }
-
-    if( ip.x + win_size.width < src_size.width )
-        rect.width = win_size.width;
-    else
-    {
-        rect.width = src_size.width - ip.x - 1;
-        if( rect.width < 0 )
-        {
-            src += rect.width*pix_size;
-            rect.width = 0;
-        }
-        assert( rect.width <= win_size.width );
-    }
-
-    if( ip.y >= 0 )
-    {
-        src += ip.y * src_step;
-        rect.y = 0;
-    }
-    else
-        rect.y = -ip.y;
-
-    if( ip.y + win_size.height < src_size.height )
-        rect.height = win_size.height;
-    else
-    {
-        rect.height = src_size.height - ip.y - 1;
-        if( rect.height < 0 )
-        {
-            src += rect.height*src_step;
-            rect.height = 0;
-        }
-    }
-
-    *pRect = rect;
-    return src - rect.x*pix_size;
-}
-
-
-static CvStatus CV_STDCALL icvGetRectSubPix_8u32f_C1R
-( const uchar* src, int src_step, CvSize src_size,
- float* dst, int dst_step, CvSize win_size, CvPoint2D32f center )
-{
-    CvPoint ip;
-    float  a12, a22, b1, b2;
-    float a, b;
-    double s = 0;
-    int i, j;
-
-    center.x -= (win_size.width-1)*0.5f;
-    center.y -= (win_size.height-1)*0.5f;
-
-    ip.x = cvFloor( center.x );
-    ip.y = cvFloor( center.y );
-
-    if( win_size.width <= 0 || win_size.height <= 0 )
-        return CV_BADRANGE_ERR;
-
-    a = center.x - ip.x;
-    b = center.y - ip.y;
-    a = MAX(a,0.0001f);
-    a12 = a*(1.f-b);
-    a22 = a*b;
-    b1 = 1.f - b;
-    b2 = b;
-    s = (1. - a)/a;
-
-    src_step /= sizeof(src[0]);
-    dst_step /= sizeof(dst[0]);
-
-    if( 0 <= ip.x && ip.x + win_size.width < src_size.width &&
-       0 <= ip.y && ip.y + win_size.height < src_size.height )
-    {
-        // extracted rectangle is totally inside the image
-        src += ip.y * src_step + ip.x;
-
-#if 0
-        if( icvCopySubpix_8u32f_C1R_p &&
-           icvCopySubpix_8u32f_C1R_p( src, src_step, dst,
-                                     dst_step*sizeof(dst[0]), win_size, a, b ) >= 0 )
-            return CV_OK;
-#endif
-
-        for( ; win_size.height--; src += src_step, dst += dst_step )
-        {
-            float prev = (1 - a)*(b1*CV_8TO32F(src[0]) + b2*CV_8TO32F(src[src_step]));
-            for( j = 0; j < win_size.width; j++ )
-            {
-                float t = a12*CV_8TO32F(src[j+1]) + a22*CV_8TO32F(src[j+1+src_step]);
-                dst[j] = prev + t;
-                prev = (float)(t*s);
-            }
-        }
-    }
-    else
-    {
-        CvRect r;
-
-        src = (const uchar*)icvAdjustRect( src, src_step*sizeof(*src),
-                                          sizeof(*src), src_size, win_size,ip, &r);
-
-        for( i = 0; i < win_size.height; i++, dst += dst_step )
-        {
-            const uchar *src2 = src + src_step;
-
-            if( i < r.y || i >= r.height )
-                src2 -= src_step;
-
-            for( j = 0; j < r.x; j++ )
-            {
-                float s0 = CV_8TO32F(src[r.x])*b1 +
-                CV_8TO32F(src2[r.x])*b2;
-
-                dst[j] = (float)(s0);
-            }
-
-            if( j < r.width )
-            {
-                float prev = (1 - a)*(b1*CV_8TO32F(src[j]) + b2*CV_8TO32F(src2[j]));
-
-                for( ; j < r.width; j++ )
-                {
-                    float t = a12*CV_8TO32F(src[j+1]) + a22*CV_8TO32F(src2[j+1]);
-                    dst[j] = prev + t;
-                    prev = (float)(t*s);
-                }
-            }
-
-            for( ; j < win_size.width; j++ )
-            {
-                float s0 = CV_8TO32F(src[r.width])*b1 +
-                CV_8TO32F(src2[r.width])*b2;
-
-                dst[j] = (float)(s0);
-            }
-
-            if( i < r.height )
-                src = src2;
-        }
-    }
-
-    return CV_OK;
-}
-
-
-#define ICV_32F8U(x)  ((uchar)cvRound(x))
-
-#define ICV_DEF_GET_QUADRANGLE_SUB_PIX_FUNC( flavor, srctype, dsttype,      \
-worktype, cast_macro, cvt )    \
-static CvStatus CV_STDCALL                                                   \
-icvGetQuadrangleSubPix_##flavor##_C1R                                       \
-( const srctype * src, int src_step, CvSize src_size,                       \
-dsttype *dst, int dst_step, CvSize win_size, const float *matrix )        \
-{                                                                           \
-int x, y;                                                               \
-double dx = (win_size.width - 1)*0.5;                                   \
-double dy = (win_size.height - 1)*0.5;                                  \
-double A11 = matrix[0], A12 = matrix[1], A13 = matrix[2]-A11*dx-A12*dy; \
-double A21 = matrix[3], A22 = matrix[4], A23 = matrix[5]-A21*dx-A22*dy; \
-\
-src_step /= sizeof(srctype);                                            \
-dst_step /= sizeof(dsttype);                                            \
-\
-for( y = 0; y < win_size.height; y++, dst += dst_step )                 \
-{                                                                       \
-double xs = A12*y + A13;                                            \
-double ys = A22*y + A23;                                            \
-double xe = A11*(win_size.width-1) + A12*y + A13;                   \
-double ye = A21*(win_size.width-1) + A22*y + A23;                   \
-\
-if( (unsigned)(cvFloor(xs)-1) < (unsigned)(src_size.width - 3) &&   \
-(unsigned)(cvFloor(ys)-1) < (unsigned)(src_size.height - 3) &&  \
-(unsigned)(cvFloor(xe)-1) < (unsigned)(src_size.width - 3) &&   \
-(unsigned)(cvFloor(ye)-1) < (unsigned)(src_size.height - 3))    \
-{                                                                   \
-for( x = 0; x < win_size.width; x++ )                           \
-{                                                               \
-int ixs = cvFloor( xs );                                    \
-int iys = cvFloor( ys );                                    \
-const srctype *ptr = src + src_step*iys + ixs;              \
-double a = xs - ixs, b = ys - iys, a1 = 1.f - a;            \
-worktype p0 = cvt(ptr[0])*a1 + cvt(ptr[1])*a;               \
-worktype p1 = cvt(ptr[src_step])*a1 + cvt(ptr[src_step+1])*a;\
-xs += A11;                                                  \
-ys += A21;                                                  \
-\
-dst[x] = cast_macro(p0 + b * (p1 - p0));                    \
-}                                                               \
-}                                                                   \
-else                                                                \
-{                                                                   \
-for( x = 0; x < win_size.width; x++ )                           \
-{                                                               \
-int ixs = cvFloor( xs ), iys = cvFloor( ys );               \
-double a = xs - ixs, b = ys - iys, a1 = 1.f - a;            \
-const srctype *ptr0, *ptr1;                                 \
-worktype p0, p1;                                            \
-xs += A11; ys += A21;                                       \
-\
-if( (unsigned)iys < (unsigned)(src_size.height-1) )         \
-ptr0 = src + src_step*iys, ptr1 = ptr0 + src_step;      \
-else                                                        \
-ptr0 = ptr1 = src + (iys < 0 ? 0 : src_size.height-1)*src_step; \
-\
-if( (unsigned)ixs < (unsigned)(src_size.width-1) )          \
-{                                                           \
-p0 = cvt(ptr0[ixs])*a1 + cvt(ptr0[ixs+1])*a;            \
-p1 = cvt(ptr1[ixs])*a1 + cvt(ptr1[ixs+1])*a;            \
-}                                                           \
-else                                                        \
-{                                                           \
-ixs = ixs < 0 ? 0 : src_size.width - 1;                 \
-p0 = cvt(ptr0[ixs]); p1 = cvt(ptr1[ixs]);               \
-}                                                           \
-dst[x] = cast_macro(p0 + b * (p1 - p0));                    \
-}                                                               \
-}                                                                   \
-}                                                                       \
-\
-return CV_OK;                                                           \
-}
-
-ICV_DEF_GET_QUADRANGLE_SUB_PIX_FUNC( 8u32f, uchar, float, double, CV_CAST_32F, CV_8TO32F )
-
-
-CV_IMPL void
-cvCalcOpticalFlowPyrLK( const void* arrA, const void* arrB,
-                        void* /*pyrarrA*/, void* /*pyrarrB*/,
-                        const CvPoint2D32f * featuresA,
-                        CvPoint2D32f * featuresB,
-                        int count, CvSize winSize, int level,
-                        char *status, float *error,
-                        CvTermCriteria criteria, int flags )
-{
-    if( count <= 0 )
-        return;
-    CV_Assert( featuresA && featuresB );
-    cv::Mat A = cv::cvarrToMat(arrA), B = cv::cvarrToMat(arrB);
-    cv::Mat ptA(count, 1, CV_32FC2, (void*)featuresA);
-    cv::Mat ptB(count, 1, CV_32FC2, (void*)featuresB);
-    cv::Mat st, err;
-
-    if( status )
-        st = cv::Mat(count, 1, CV_8U, (void*)status);
-    if( error )
-        err = cv::Mat(count, 1, CV_32F, (void*)error);
-    cv::calcOpticalFlowPyrLK( A, B, ptA, ptB, st,
-                              error ? cv::_OutputArray(err) : cv::noArray(),
-                              winSize, level, criteria, flags);
-}
-
-
-/* Affine tracking algorithm */
-
-CV_IMPL void
-cvCalcAffineFlowPyrLK( const void* arrA, const void* arrB,
-                       void* pyrarrA, void* pyrarrB,
-                       const CvPoint2D32f * featuresA,
-                       CvPoint2D32f * featuresB,
-                       float *matrices, int count,
-                       CvSize winSize, int level,
-                       char *status, float *error,
-                       CvTermCriteria criteria, int flags )
-{
-    const int MAX_ITERS = 100;
-
-    cv::AutoBuffer<char> _status;
-    cv::AutoBuffer<uchar> buffer;
-    cv::AutoBuffer<uchar> pyr_buffer;
-
-    CvMat stubA, *imgA = (CvMat*)arrA;
-    CvMat stubB, *imgB = (CvMat*)arrB;
-    CvMat pstubA, *pyrA = (CvMat*)pyrarrA;
-    CvMat pstubB, *pyrB = (CvMat*)pyrarrB;
-
-    static const float smoothKernel[] = { 0.09375, 0.3125, 0.09375 };  /* 3/32, 10/32, 3/32 */
-
-    int bufferBytes = 0;
-
-    uchar **imgI = 0;
-    uchar **imgJ = 0;
-    int *step = 0;
-    double *scale = 0;
-    CvSize* size = 0;
-
-    float *patchI;
-    float *patchJ;
-    float *Ix;
-    float *Iy;
-
-    int i, j, k, l;
-
-    CvSize patchSize = cvSize( winSize.width * 2 + 1, winSize.height * 2 + 1 );
-    int patchLen = patchSize.width * patchSize.height;
-    int patchStep = patchSize.width * sizeof( patchI[0] );
-
-    CvSize srcPatchSize = cvSize( patchSize.width + 2, patchSize.height + 2 );
-    int srcPatchLen = srcPatchSize.width * srcPatchSize.height;
-    int srcPatchStep = srcPatchSize.width * sizeof( patchI[0] );
-    CvSize imgSize;
-    float eps = (float)MIN(winSize.width, winSize.height);
-
-    imgA = cvGetMat( imgA, &stubA );
-    imgB = cvGetMat( imgB, &stubB );
-
-    if( CV_MAT_TYPE( imgA->type ) != CV_8UC1 )
-        CV_Error( CV_StsUnsupportedFormat, "" );
-
-    if( !CV_ARE_TYPES_EQ( imgA, imgB ))
-        CV_Error( CV_StsUnmatchedFormats, "" );
-
-    if( !CV_ARE_SIZES_EQ( imgA, imgB ))
-        CV_Error( CV_StsUnmatchedSizes, "" );
-
-    if( imgA->step != imgB->step )
-        CV_Error( CV_StsUnmatchedSizes, "imgA and imgB must have equal steps" );
-
-    if( !matrices )
-        CV_Error( CV_StsNullPtr, "" );
-
-    imgSize = cvGetMatSize( imgA );
-
-    if( pyrA )
-    {
-        pyrA = cvGetMat( pyrA, &pstubA );
-
-        if( pyrA->step*pyrA->height < icvMinimalPyramidSize( imgSize ) )
-            CV_Error( CV_StsBadArg, "pyramid A has insufficient size" );
-    }
-    else
-    {
-        pyrA = &pstubA;
-        pyrA->data.ptr = 0;
-    }
-
-    if( pyrB )
-    {
-        pyrB = cvGetMat( pyrB, &pstubB );
-
-        if( pyrB->step*pyrB->height < icvMinimalPyramidSize( imgSize ) )
-            CV_Error( CV_StsBadArg, "pyramid B has insufficient size" );
-    }
-    else
-    {
-        pyrB = &pstubB;
-        pyrB->data.ptr = 0;
-    }
-
-    if( count == 0 )
-        return;
-
-    /* check input arguments */
-    if( !featuresA || !featuresB || !matrices )
-        CV_Error( CV_StsNullPtr, "" );
-
-    if( winSize.width <= 1 || winSize.height <= 1 )
-        CV_Error( CV_StsOutOfRange, "the search window is too small" );
-
-    if( count < 0 )
-        CV_Error( CV_StsOutOfRange, "" );
-
-    icvInitPyramidalAlgorithm( imgA, imgB,
-        pyrA, pyrB, level, &criteria, MAX_ITERS, flags,
-        &imgI, &imgJ, &step, &size, &scale, &pyr_buffer );
-
-    /* buffer_size = <size for patches> + <size for pyramids> */
-    bufferBytes = (srcPatchLen + patchLen*3)*sizeof(patchI[0]) + (36*2 + 6)*sizeof(double);
-
-    buffer.allocate(bufferBytes);
-
-    if( !status )
-    {
-        _status.allocate(count);
-        status = _status;
-    }
-
-    patchI = (float *)(uchar*)buffer;
-    patchJ = patchI + srcPatchLen;
-    Ix = patchJ + patchLen;
-    Iy = Ix + patchLen;
-
-    if( status )
-        memset( status, 1, count );
-
-    if( !(flags & CV_LKFLOW_INITIAL_GUESSES) )
-    {
-        memcpy( featuresB, featuresA, count * sizeof( featuresA[0] ));
-        for( i = 0; i < count * 4; i += 4 )
-        {
-            matrices[i] = matrices[i + 3] = 1.f;
-            matrices[i + 1] = matrices[i + 2] = 0.f;
-        }
-    }
-
-    for( i = 0; i < count; i++ )
-    {
-        featuresB[i].x = (float)(featuresB[i].x * scale[level] * 0.5);
-        featuresB[i].y = (float)(featuresB[i].y * scale[level] * 0.5);
-    }
-
-    /* do processing from top pyramid level (smallest image)
-       to the bottom (original image) */
-    for( l = level; l >= 0; l-- )
-    {
-        CvSize levelSize = size[l];
-        int levelStep = step[l];
-
-        /* find flow for each given point at the particular level */
-        for( i = 0; i < count; i++ )
-        {
-            CvPoint2D32f u;
-            float Av[6];
-            double G[36];
-            double meanI = 0, meanJ = 0;
-            int x, y;
-            int pt_status = status[i];
-            CvMat mat;
-
-            if( !pt_status )
-                continue;
-
-            Av[0] = matrices[i*4];
-            Av[1] = matrices[i*4+1];
-            Av[3] = matrices[i*4+2];
-            Av[4] = matrices[i*4+3];
-
-            Av[2] = featuresB[i].x += featuresB[i].x;
-            Av[5] = featuresB[i].y += featuresB[i].y;
-
-            u.x = (float) (featuresA[i].x * scale[l]);
-            u.y = (float) (featuresA[i].y * scale[l]);
-
-            if( u.x < -eps || u.x >= levelSize.width+eps ||
-                u.y < -eps || u.y >= levelSize.height+eps ||
-                icvGetRectSubPix_8u32f_C1R( imgI[l], levelStep,
-                levelSize, patchI, srcPatchStep, srcPatchSize, u ) < 0 )
-            {
-                /* point is outside the image. take the next */
-                if( l == 0 )
-                    status[i] = 0;
-                continue;
-            }
-
-            icvCalcIxIy_32f( patchI, srcPatchStep, Ix, Iy,
-                (srcPatchSize.width-2)*sizeof(patchI[0]), srcPatchSize,
-                smoothKernel, patchJ );
-
-            /* repack patchI (remove borders) */
-            for( k = 0; k < patchSize.height; k++ )
-                memcpy( patchI + k * patchSize.width,
-                        patchI + (k + 1) * srcPatchSize.width + 1, patchStep );
-
-            memset( G, 0, sizeof( G ));
-
-            /* calculate G matrix */
-            for( y = -winSize.height, k = 0; y <= winSize.height; y++ )
-            {
-                for( x = -winSize.width; x <= winSize.width; x++, k++ )
-                {
-                    double ixix = ((double) Ix[k]) * Ix[k];
-                    double ixiy = ((double) Ix[k]) * Iy[k];
-                    double iyiy = ((double) Iy[k]) * Iy[k];
-
-                    double xx, xy, yy;
-
-                    G[0] += ixix;
-                    G[1] += ixiy;
-                    G[2] += x * ixix;
-                    G[3] += y * ixix;
-                    G[4] += x * ixiy;
-                    G[5] += y * ixiy;
-
-                    // G[6] == G[1]
-                    G[7] += iyiy;
-                    // G[8] == G[4]
-                    // G[9] == G[5]
-                    G[10] += x * iyiy;
-                    G[11] += y * iyiy;
-
-                    xx = x * x;
-                    xy = x * y;
-                    yy = y * y;
-
-                    // G[12] == G[2]
-                    // G[13] == G[8] == G[4]
-                    G[14] += xx * ixix;
-                    G[15] += xy * ixix;
-                    G[16] += xx * ixiy;
-                    G[17] += xy * ixiy;
-
-                    // G[18] == G[3]
-                    // G[19] == G[9]
-                    // G[20] == G[15]
-                    G[21] += yy * ixix;
-                    // G[22] == G[17]
-                    G[23] += yy * ixiy;
-
-                    // G[24] == G[4]
-                    // G[25] == G[10]
-                    // G[26] == G[16]
-                    // G[27] == G[22]
-                    G[28] += xx * iyiy;
-                    G[29] += xy * iyiy;
-
-                    // G[30] == G[5]
-                    // G[31] == G[11]
-                    // G[32] == G[17]
-                    // G[33] == G[23]
-                    // G[34] == G[29]
-                    G[35] += yy * iyiy;
-
-                    meanI += patchI[k];
-                }
-            }
-
-            meanI /= patchSize.width*patchSize.height;
-
-            G[8] = G[4];
-            G[9] = G[5];
-            G[22] = G[17];
-
-            // fill part of G below its diagonal
-            for( y = 1; y < 6; y++ )
-                for( x = 0; x < y; x++ )
-                    G[y * 6 + x] = G[x * 6 + y];
-
-            cvInitMatHeader( &mat, 6, 6, CV_64FC1, G );
-
-            if( cvInvert( &mat, &mat, CV_SVD ) < 1e-4 )
-            {
-                /* bad matrix. take the next point */
-                if( l == 0 )
-                    status[i] = 0;
-                continue;
-            }
-
-            for( j = 0; j < criteria.max_iter; j++ )
-            {
-                double b[6] = {0,0,0,0,0,0}, eta[6];
-                double t0, t1, s = 0;
-
-                if( Av[2] < -eps || Av[2] >= levelSize.width+eps ||
-                    Av[5] < -eps || Av[5] >= levelSize.height+eps ||
-                    icvGetQuadrangleSubPix_8u32f_C1R( imgJ[l], levelStep,
-                    levelSize, patchJ, patchStep, patchSize, Av ) < 0 )
-                {
-                    pt_status = 0;
-                    break;
-                }
-
-                for( y = -winSize.height, k = 0, meanJ = 0; y <= winSize.height; y++ )
-                    for( x = -winSize.width; x <= winSize.width; x++, k++ )
-                        meanJ += patchJ[k];
-
-                meanJ = meanJ / (patchSize.width * patchSize.height) - meanI;
-
-                for( y = -winSize.height, k = 0; y <= winSize.height; y++ )
-                {
-                    for( x = -winSize.width; x <= winSize.width; x++, k++ )
-                    {
-                        double t = patchI[k] - patchJ[k] + meanJ;
-                        double ixt = Ix[k] * t;
-                        double iyt = Iy[k] * t;
-
-                        s += t;
-
-                        b[0] += ixt;
-                        b[1] += iyt;
-                        b[2] += x * ixt;
-                        b[3] += y * ixt;
-                        b[4] += x * iyt;
-                        b[5] += y * iyt;
-                    }
-                }
-
-                for( k = 0; k < 6; k++ )
-                    eta[k] = G[k*6]*b[0] + G[k*6+1]*b[1] + G[k*6+2]*b[2] +
-                        G[k*6+3]*b[3] + G[k*6+4]*b[4] + G[k*6+5]*b[5];
-
-                Av[2] = (float)(Av[2] + Av[0] * eta[0] + Av[1] * eta[1]);
-                Av[5] = (float)(Av[5] + Av[3] * eta[0] + Av[4] * eta[1]);
-
-                t0 = Av[0] * (1 + eta[2]) + Av[1] * eta[4];
-                t1 = Av[0] * eta[3] + Av[1] * (1 + eta[5]);
-                Av[0] = (float)t0;
-                Av[1] = (float)t1;
-
-                t0 = Av[3] * (1 + eta[2]) + Av[4] * eta[4];
-                t1 = Av[3] * eta[3] + Av[4] * (1 + eta[5]);
-                Av[3] = (float)t0;
-                Av[4] = (float)t1;
-
-                if( eta[0] * eta[0] + eta[1] * eta[1] < criteria.epsilon )
-                    break;
-            }
-
-            if( pt_status != 0 || l == 0 )
-            {
-                status[i] = (char)pt_status;
-                featuresB[i].x = Av[2];
-                featuresB[i].y = Av[5];
-
-                matrices[i*4] = Av[0];
-                matrices[i*4+1] = Av[1];
-                matrices[i*4+2] = Av[3];
-                matrices[i*4+3] = Av[4];
-            }
-
-            if( pt_status && l == 0 && error )
-            {
-                /* calc error */
-                double err = 0;
-
-                for( y = 0, k = 0; y < patchSize.height; y++ )
-                {
-                    for( x = 0; x < patchSize.width; x++, k++ )
-                    {
-                        double t = patchI[k] - patchJ[k] + meanJ;
-                        err += t * t;
-                    }
-                }
-                error[i] = (float)std::sqrt(err);
-            }
-        }
-    }
-}
-
-
-
-static void
-icvGetRTMatrix( const CvPoint2D32f* a, const CvPoint2D32f* b,
-                int count, CvMat* M, int full_affine )
-{
-    if( full_affine )
-    {
-        double sa[36], sb[6];
-        CvMat A = cvMat( 6, 6, CV_64F, sa ), B = cvMat( 6, 1, CV_64F, sb );
-        CvMat MM = cvMat( 6, 1, CV_64F, M->data.db );
-
-        int i;
-
-        memset( sa, 0, sizeof(sa) );
-        memset( sb, 0, sizeof(sb) );
-
-        for( i = 0; i < count; i++ )
-        {
-            sa[0] += a[i].x*a[i].x;
-            sa[1] += a[i].y*a[i].x;
-            sa[2] += a[i].x;
-
-            sa[6] += a[i].x*a[i].y;
-            sa[7] += a[i].y*a[i].y;
-            sa[8] += a[i].y;
-
-            sa[12] += a[i].x;
-            sa[13] += a[i].y;
-            sa[14] += 1;
+            sa[2][2] += 1;
 
             sb[0] += a[i].x*b[i].x;
             sb[1] += a[i].y*b[i].x;
@@ -1614,50 +788,35 @@ icvGetRTMatrix( const CvPoint2D32f* a, const CvPoint2D32f* b,
             sb[5] += b[i].y;
         }
 
-        sa[21] = sa[0];
-        sa[22] = sa[1];
-        sa[23] = sa[2];
-        sa[27] = sa[6];
-        sa[28] = sa[7];
-        sa[29] = sa[8];
-        sa[33] = sa[12];
-        sa[34] = sa[13];
-        sa[35] = sa[14];
+        sa[3][4] = sa[4][3] = sa[1][0] = sa[0][1];
+        sa[3][5] = sa[5][3] = sa[2][0] = sa[0][2];
+        sa[4][5] = sa[5][4] = sa[2][1] = sa[1][2];
 
-        cvSolve( &A, &B, &MM, CV_SVD );
+        sa[3][3] = sa[0][0];
+        sa[4][4] = sa[1][1];
+        sa[5][5] = sa[2][2];
+
+        solve( A, B, MM, DECOMP_EIG );
     }
     else
     {
-        double sa[16], sb[4], m[4], *om = M->data.db;
-        CvMat A = cvMat( 4, 4, CV_64F, sa ), B = cvMat( 4, 1, CV_64F, sb );
-        CvMat MM = cvMat( 4, 1, CV_64F, m );
+        double sa[4][4]={{0.}}, sb[4]={0.}, m[4];
+        Mat A( 4, 4, CV_64F, sa ), B( 4, 1, CV_64F, sb );
+        Mat MM( 4, 1, CV_64F, m );
 
-        int i;
-
-        memset( sa, 0, sizeof(sa) );
-        memset( sb, 0, sizeof(sb) );
-
-        for( i = 0; i < count; i++ )
+        for( int i = 0; i < count; i++ )
         {
-            sa[0] += a[i].x*a[i].x + a[i].y*a[i].y;
-            sa[1] += 0;
-            sa[2] += a[i].x;
-            sa[3] += a[i].y;
+            sa[0][0] += a[i].x*a[i].x + a[i].y*a[i].y;
+            sa[0][2] += a[i].x;
+            sa[0][3] += a[i].y;
 
-            sa[4] += 0;
-            sa[5] += a[i].x*a[i].x + a[i].y*a[i].y;
-            sa[6] += -a[i].y;
-            sa[7] += a[i].x;
 
-            sa[8] += a[i].x;
-            sa[9] += -a[i].y;
-            sa[10] += 1;
-            sa[11] += 0;
+            sa[2][1] += -a[i].y;
+            sa[2][2] += 1;
 
-            sa[12] += a[i].y;
-            sa[13] += a[i].x;
-            sa[14] += 0;
-            sa[15] += 1;
+            sa[3][0] += a[i].y;
+            sa[3][1] += a[i].x;
+            sa[3][3] += 1;
 
             sb[0] += a[i].x*b[i].x + a[i].y*b[i].y;
             sb[1] += a[i].x*b[i].y - a[i].y*b[i].x;
@@ -1665,8 +824,15 @@ icvGetRTMatrix( const CvPoint2D32f* a, const CvPoint2D32f* b,
             sb[3] += b[i].y;
         }
 
-        cvSolve( &A, &B, &MM, CV_SVD );
+        sa[1][1] = sa[0][0];
+        sa[2][1] = sa[1][2] = -sa[0][3];
+        sa[3][1] = sa[1][3] = sa[2][0] = sa[0][2];
+        sa[2][2] = sa[3][3] = count;
+        sa[3][0] = sa[0][3];
 
+        solve( A, B, MM, DECOMP_EIG );
+
+        double* om = M.ptr<double>();
         om[0] = om[4] = m[0];
         om[1] = -m[1];
         om[3] = m[1];
@@ -1675,88 +841,84 @@ icvGetRTMatrix( const CvPoint2D32f* a, const CvPoint2D32f* b,
     }
 }
 
+}
 
-CV_IMPL int
-cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int full_affine )
+cv::Mat cv::estimateRigidTransform( InputArray src1, InputArray src2, bool fullAffine )
 {
+    Mat M(2, 3, CV_64F), A = src1.getMat(), B = src2.getMat();
+
     const int COUNT = 15;
     const int WIDTH = 160, HEIGHT = 120;
     const int RANSAC_MAX_ITERS = 500;
     const int RANSAC_SIZE0 = 3;
     const double RANSAC_GOOD_RATIO = 0.5;
 
-    cv::Ptr<CvMat> sA, sB;
-    cv::AutoBuffer<CvPoint2D32f> pA, pB;
-    cv::AutoBuffer<int> good_idx;
-    cv::AutoBuffer<char> status;
-    cv::Ptr<CvMat> gray;
+    std::vector<Point2f> pA, pB;
+    std::vector<int> good_idx;
+    std::vector<uchar> status;
 
-    CvMat stubA, *A = cvGetMat( matA, &stubA );
-    CvMat stubB, *B = cvGetMat( matB, &stubB );
-    CvSize sz0, sz1;
-    int cn, equal_sizes;
+    double scale = 1.;
     int i, j, k, k1;
-    int count_x, count_y, count = 0;
-    double scale = 1;
-    CvRNG rng = cvRNG(-1);
-    double m[6]={0};
-    CvMat M = cvMat( 2, 3, CV_64F, m );
+
+    RNG rng((uint64)-1);
     int good_count = 0;
-    CvRect brect;
 
-    if( !CV_IS_MAT(matM) )
-        CV_Error( matM ? CV_StsBadArg : CV_StsNullPtr, "Output parameter M is not a valid matrix" );
+    if( A.size() != B.size() )
+        CV_Error( Error::StsUnmatchedSizes, "Both input images must have the same size" );
 
-    if( !CV_ARE_SIZES_EQ( A, B ) )
-        CV_Error( CV_StsUnmatchedSizes, "Both input images must have the same size" );
+    if( A.type() != B.type() )
+        CV_Error( Error::StsUnmatchedFormats, "Both input images must have the same data type" );
 
-    if( !CV_ARE_TYPES_EQ( A, B ) )
-        CV_Error( CV_StsUnmatchedFormats, "Both input images must have the same data type" );
+    int count = A.checkVector(2);
 
-    if( CV_MAT_TYPE(A->type) == CV_8UC1 || CV_MAT_TYPE(A->type) == CV_8UC3 )
+    if( count > 0 )
     {
-        cn = CV_MAT_CN(A->type);
-        sz0 = cvGetSize(A);
-        sz1 = cvSize(WIDTH, HEIGHT);
+        A.reshape(2, count).convertTo(pA, CV_32F);
+        B.reshape(2, count).convertTo(pB, CV_32F);
+    }
+    else if( A.depth() == CV_8U )
+    {
+        int cn = A.channels();
+        CV_Assert( cn == 1 || cn == 3 || cn == 4 );
+        Size sz0 = A.size();
+        Size sz1(WIDTH, HEIGHT);
+
+        scale = std::max(1., std::max( (double)sz1.width/sz0.width, (double)sz1.height/sz0.height ));
 
-        scale = MAX( (double)sz1.width/sz0.width, (double)sz1.height/sz0.height );
-        scale = MIN( scale, 1. );
         sz1.width = cvRound( sz0.width * scale );
         sz1.height = cvRound( sz0.height * scale );
 
-        equal_sizes = sz1.width == sz0.width && sz1.height == sz0.height;
+        bool equalSizes = sz1.width == sz0.width && sz1.height == sz0.height;
 
-        if( !equal_sizes || cn != 1 )
+        if( !equalSizes || cn != 1 )
         {
-            sA = cvCreateMat( sz1.height, sz1.width, CV_8UC1 );
-            sB = cvCreateMat( sz1.height, sz1.width, CV_8UC1 );
+            Mat sA, sB;
 
             if( cn != 1 )
             {
-                gray = cvCreateMat( sz0.height, sz0.width, CV_8UC1 );
-                cvCvtColor( A, gray, CV_BGR2GRAY );
-                cvResize( gray, sA, CV_INTER_AREA );
-                cvCvtColor( B, gray, CV_BGR2GRAY );
-                cvResize( gray, sB, CV_INTER_AREA );
-                gray.release();
+                Mat gray;
+                cvtColor(A, gray, COLOR_BGR2GRAY);
+                resize(gray, sA, sz1, 0., 0., INTER_AREA);
+                cvtColor(B, gray, COLOR_BGR2GRAY);
+                resize(gray, sB, sz1, 0., 0., INTER_AREA);
             }
             else
             {
-                cvResize( A, sA, CV_INTER_AREA );
-                cvResize( B, sB, CV_INTER_AREA );
+                resize(A, sA, sz1, 0., 0., INTER_AREA);
+                resize(B, sB, sz1, 0., 0., INTER_AREA);
             }
 
             A = sA;
             B = sB;
         }
 
-        count_y = COUNT;
-        count_x = cvRound((double)COUNT*sz1.width/sz1.height);
+        int count_y = COUNT;
+        int count_x = cvRound((double)COUNT*sz1.width/sz1.height);
         count = count_x * count_y;
 
-        pA.allocate(count);
-        pB.allocate(count);
-        status.allocate(count);
+        pA.resize(count);
+        pB.resize(count);
+        status.resize(count);
 
         for( i = 0, k = 0; i < count_y; i++ )
             for( j = 0; j < count_x; j++, k++ )
@@ -1766,8 +928,8 @@ cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int
             }
 
         // find the corresponding points in B
-        cvCalcOpticalFlowPyrLK( A, B, 0, 0, pA, pB, count, cvSize(10,10), 3,
-                                status, 0, cvTermCriteria(CV_TERMCRIT_ITER,40,0.1), 0 );
+        calcOpticalFlowPyrLK(A, B, pA, pB, status, noArray(), Size(21, 21), 3,
+                             TermCriteria(TermCriteria::MAX_ITER,40,0.1));
 
         // repack the remained points
         for( i = 0, k = 0; i < count; i++ )
@@ -1780,48 +942,34 @@ cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int
                 }
                 k++;
             }
-
         count = k;
-    }
-    else if( CV_MAT_TYPE(A->type) == CV_32FC2 || CV_MAT_TYPE(A->type) == CV_32SC2 )
-    {
-        count = A->cols*A->rows;
-        CvMat _pA, _pB;
-        pA.allocate(count);
-        pB.allocate(count);
-        _pA = cvMat( A->rows, A->cols, CV_32FC2, pA );
-        _pB = cvMat( B->rows, B->cols, CV_32FC2, pB );
-        cvConvert( A, &_pA );
-        cvConvert( B, &_pB );
+        pA.resize(count);
+        pB.resize(count);
     }
     else
-        CV_Error( CV_StsUnsupportedFormat, "Both input images must have either 8uC1 or 8uC3 type" );
+        CV_Error( Error::StsUnsupportedFormat, "Both input images must have either 8uC1 or 8uC3 type" );
 
-    good_idx.allocate(count);
+    good_idx.resize(count);
 
     if( count < RANSAC_SIZE0 )
-        return 0;
+        return Mat();
 
-    CvMat _pB = cvMat(1, count, CV_32FC2, pB);
-    brect = cvBoundingRect(&_pB, 1);
+    Rect brect = boundingRect(pB);
 
     // RANSAC stuff:
     // 1. find the consensus
     for( k = 0; k < RANSAC_MAX_ITERS; k++ )
     {
         int idx[RANSAC_SIZE0];
-        CvPoint2D32f a[3];
-        CvPoint2D32f b[3];
-
-        memset( a, 0, sizeof(a) );
-        memset( b, 0, sizeof(b) );
+        Point2f a[RANSAC_SIZE0];
+        Point2f b[RANSAC_SIZE0];
 
         // choose random 3 non-complanar points from A & B
         for( i = 0; i < RANSAC_SIZE0; i++ )
         {
             for( k1 = 0; k1 < RANSAC_MAX_ITERS; k1++ )
             {
-                idx[i] = cvRandInt(&rng) % count;
+                idx[i] = rng.uniform(0, count);
 
                 for( j = 0; j < i; j++ )
                 {
@@ -1871,12 +1019,13 @@ cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int
             continue;
 
         // estimate the transformation using 3 points
-        icvGetRTMatrix( a, b, 3, &M, full_affine );
+        getRTMatrix( a, b, 3, M, fullAffine );
 
+        const double* m = M.ptr<double>();
         for( i = 0, good_count = 0; i < count; i++ )
         {
-            if( fabs( m[0]*pA[i].x + m[1]*pA[i].y + m[2] - pB[i].x ) +
-                fabs( m[3]*pA[i].x + m[4]*pA[i].y + m[5] - pB[i].y ) < MAX(brect.width,brect.height)*0.05 )
+            if( std::abs( m[0]*pA[i].x + m[1]*pA[i].y + m[2] - pB[i].x ) +
+                std::abs( m[3]*pA[i].x + m[4]*pA[i].y + m[5] - pB[i].y ) < std::max(brect.width,brect.height)*0.05 )
                 good_idx[good_count++] = i;
         }
 
@@ -1885,7 +1034,7 @@ cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int
     }
 
     if( k >= RANSAC_MAX_ITERS )
-        return 0;
+        return Mat();
 
     if( good_count < count )
     {
@@ -1897,25 +1046,11 @@ cvEstimateRigidTransform( const CvArr* matA, const CvArr* matB, CvMat* matM, int
         }
     }
 
-    icvGetRTMatrix( pA, pB, good_count, &M, full_affine );
-    m[2] /= scale;
-    m[5] /= scale;
-    cvConvert( &M, matM );
+    getRTMatrix( &pA[0], &pB[0], good_count, M, fullAffine );
+    M.at<double>(0, 2) /= scale;
+    M.at<double>(1, 2) /= scale;
 
-    return 1;
-}
-
-cv::Mat cv::estimateRigidTransform( InputArray src1,
-                                    InputArray src2,
-                                    bool fullAffine )
-{
-    Mat M(2, 3, CV_64F), A = src1.getMat(), B = src2.getMat();
-    CvMat matA = A, matB = B, matM = M;
-    int err = cvEstimateRigidTransform(&matA, &matB, &matM, fullAffine);
-    if (err == 1)
-        return M;
-    else
-        return Mat();
+    return M;
 }
 
 /* End of file. */
diff --git a/modules/video/src/lkpyramid.hpp b/modules/video/src/lkpyramid.hpp
index 390e46bf99..1b9d50bc64 100644
--- a/modules/video/src/lkpyramid.hpp
+++ b/modules/video/src/lkpyramid.hpp
@@ -7,7 +7,7 @@ namespace detail
 
     typedef short deriv_type;
 
-    struct LKTrackerInvoker
+    struct LKTrackerInvoker : ParallelLoopBody
     {
         LKTrackerInvoker( const Mat& _prevImg, const Mat& _prevDeriv, const Mat& _nextImg,
                           const Point2f* _prevPts, Point2f* _nextPts,
@@ -15,7 +15,7 @@ namespace detail
                           Size _winSize, TermCriteria _criteria,
                           int _level, int _maxLevel, int _flags, float _minEigThreshold );
 
-        void operator()(const BlockedRange& range) const;
+        void operator()(const Range& range) const;
 
         const Mat* prevImg;
         const Mat* nextImg;
@@ -34,4 +34,3 @@ namespace detail
 
 }// namespace detail
 }// namespace cv
-
diff --git a/modules/video/src/motempl.cpp b/modules/video/src/motempl.cpp
index f6e52c49e0..aa6d12d8d0 100644
--- a/modules/video/src/motempl.cpp
+++ b/modules/video/src/motempl.cpp
@@ -41,34 +41,23 @@
 
 #include "precomp.hpp"
 
-
-/* motion templates */
-CV_IMPL void
-cvUpdateMotionHistory( const void* silhouette, void* mhimg,
-                       double timestamp, double mhi_duration )
+void cv::updateMotionHistory( InputArray _silhouette, InputOutputArray _mhi,
+                              double timestamp, double duration )
 {
-    CvMat  silhstub, *silh = cvGetMat(silhouette, &silhstub);
-    CvMat  mhistub, *mhi = cvGetMat(mhimg, &mhistub);
+    Mat silh = _silhouette.getMat(), mhi = _mhi.getMat();
 
-    if( !CV_IS_MASK_ARR( silh ))
-        CV_Error( CV_StsBadMask, "" );
+    CV_Assert( silh.type() == CV_8U && mhi.type() == CV_32F );
+    CV_Assert( silh.size() == mhi.size() );
 
-    if( CV_MAT_TYPE( mhi->type ) != CV_32FC1 )
-        CV_Error( CV_StsUnsupportedFormat, "" );
-
-    if( !CV_ARE_SIZES_EQ( mhi, silh ))
-        CV_Error( CV_StsUnmatchedSizes, "" );
-
-    CvSize size = cvGetMatSize( mhi );
-
-    if( CV_IS_MAT_CONT( mhi->type & silh->type ))
+    Size size = silh.size();
+    if( silh.isContinuous() && mhi.isContinuous() )
     {
         size.width *= size.height;
         size.height = 1;
     }
 
     float ts = (float)timestamp;
-    float delbound = (float)(timestamp - mhi_duration);
+    float delbound = (float)(timestamp - duration);
     int x, y;
 #if CV_SSE2
     volatile bool useSIMD = cv::checkHardwareSupport(CV_CPU_SSE2);
@@ -76,8 +65,8 @@ cvUpdateMotionHistory( const void* silhouette, void* mhimg,
 
     for( y = 0; y < size.height; y++ )
     {
-        const uchar* silhData = silh->data.ptr + silh->step*y;
-        float* mhiData = (float*)(mhi->data.ptr + mhi->step*y);
+        const uchar* silhData = silh.ptr<uchar>(y);
+        float* mhiData = mhi.ptr<float>(y);
         x = 0;
 
 #if CV_SSE2
@@ -117,370 +106,269 @@ cvUpdateMotionHistory( const void* silhouette, void* mhimg,
 }
 
 
-CV_IMPL void
-cvCalcMotionGradient( const CvArr* mhiimg, CvArr* maskimg,
-                      CvArr* orientation,
-                      double delta1, double delta2,
-                      int aperture_size )
-{
-    cv::Ptr<CvMat> dX_min, dY_max;
-
-    CvMat  mhistub, *mhi = cvGetMat(mhiimg, &mhistub);
-    CvMat  maskstub, *mask = cvGetMat(maskimg, &maskstub);
-    CvMat  orientstub, *orient = cvGetMat(orientation, &orientstub);
-    CvMat  dX_min_row, dY_max_row, orient_row, mask_row;
-    CvSize size;
-    int x, y;
-
-    float  gradient_epsilon = 1e-4f * aperture_size * aperture_size;
-    float  min_delta, max_delta;
-
-    if( !CV_IS_MASK_ARR( mask ))
-        CV_Error( CV_StsBadMask, "" );
-
-    if( aperture_size < 3 || aperture_size > 7 || (aperture_size & 1) == 0 )
-        CV_Error( CV_StsOutOfRange, "aperture_size must be 3, 5 or 7" );
-
-    if( delta1 <= 0 || delta2 <= 0 )
-        CV_Error( CV_StsOutOfRange, "both delta's must be positive" );
-
-    if( CV_MAT_TYPE( mhi->type ) != CV_32FC1 || CV_MAT_TYPE( orient->type ) != CV_32FC1 )
-        CV_Error( CV_StsUnsupportedFormat,
-        "MHI and orientation must be single-channel floating-point images" );
-
-    if( !CV_ARE_SIZES_EQ( mhi, mask ) || !CV_ARE_SIZES_EQ( orient, mhi ))
-        CV_Error( CV_StsUnmatchedSizes, "" );
-
-    if( orient->data.ptr == mhi->data.ptr )
-        CV_Error( CV_StsInplaceNotSupported, "orientation image must be different from MHI" );
-
-    if( delta1 > delta2 )
-    {
-        double t;
-        CV_SWAP( delta1, delta2, t );
-    }
-
-    size = cvGetMatSize( mhi );
-    min_delta = (float)delta1;
-    max_delta = (float)delta2;
-    dX_min = cvCreateMat( mhi->rows, mhi->cols, CV_32F );
-    dY_max = cvCreateMat( mhi->rows, mhi->cols, CV_32F );
-
-    // calc Dx and Dy
-    cvSobel( mhi, dX_min, 1, 0, aperture_size );
-    cvSobel( mhi, dY_max, 0, 1, aperture_size );
-    cvGetRow( dX_min, &dX_min_row, 0 );
-    cvGetRow( dY_max, &dY_max_row, 0 );
-    cvGetRow( orient, &orient_row, 0 );
-    cvGetRow( mask, &mask_row, 0 );
-
-    // calc gradient
-    for( y = 0; y < size.height; y++ )
-    {
-        dX_min_row.data.ptr = dX_min->data.ptr + y*dX_min->step;
-        dY_max_row.data.ptr = dY_max->data.ptr + y*dY_max->step;
-        orient_row.data.ptr = orient->data.ptr + y*orient->step;
-        mask_row.data.ptr = mask->data.ptr + y*mask->step;
-        cvCartToPolar( &dX_min_row, &dY_max_row, 0, &orient_row, 1 );
-
-        // make orientation zero where the gradient is very small
-        for( x = 0; x < size.width; x++ )
-        {
-            float dY = dY_max_row.data.fl[x];
-            float dX = dX_min_row.data.fl[x];
-
-            if( fabs(dX) < gradient_epsilon && fabs(dY) < gradient_epsilon )
-            {
-                mask_row.data.ptr[x] = 0;
-                orient_row.data.i[x] = 0;
-            }
-            else
-                mask_row.data.ptr[x] = 1;
-        }
-    }
-
-    cvErode( mhi, dX_min, 0, (aperture_size-1)/2);
-    cvDilate( mhi, dY_max, 0, (aperture_size-1)/2);
-
-    // mask off pixels which have little motion difference in their neighborhood
-    for( y = 0; y < size.height; y++ )
-    {
-        dX_min_row.data.ptr = dX_min->data.ptr + y*dX_min->step;
-        dY_max_row.data.ptr = dY_max->data.ptr + y*dY_max->step;
-        mask_row.data.ptr = mask->data.ptr + y*mask->step;
-        orient_row.data.ptr = orient->data.ptr + y*orient->step;
-
-        for( x = 0; x < size.width; x++ )
-        {
-            float d0 = dY_max_row.data.fl[x] - dX_min_row.data.fl[x];
-
-            if( mask_row.data.ptr[x] == 0 || d0 < min_delta || max_delta < d0 )
-            {
-                mask_row.data.ptr[x] = 0;
-                orient_row.data.i[x] = 0;
-            }
-        }
-    }
-}
-
-
-CV_IMPL double
-cvCalcGlobalOrientation( const void* orientation, const void* maskimg, const void* mhiimg,
-                         double curr_mhi_timestamp, double mhi_duration )
-{
-    int hist_size = 12;
-    cv::Ptr<CvHistogram> hist;
-
-    CvMat  mhistub, *mhi = cvGetMat(mhiimg, &mhistub);
-    CvMat  maskstub, *mask = cvGetMat(maskimg, &maskstub);
-    CvMat  orientstub, *orient = cvGetMat(orientation, &orientstub);
-    void*  _orient;
-    float _ranges[] = { 0, 360 };
-    float* ranges = _ranges;
-    int base_orient;
-    float shift_orient = 0, shift_weight = 0;
-    float a, b, fbase_orient;
-    float delbound;
-    CvMat mhi_row, mask_row, orient_row;
-    int x, y, mhi_rows, mhi_cols;
-
-    if( !CV_IS_MASK_ARR( mask ))
-        CV_Error( CV_StsBadMask, "" );
-
-    if( CV_MAT_TYPE( mhi->type ) != CV_32FC1 || CV_MAT_TYPE( orient->type ) != CV_32FC1 )
-        CV_Error( CV_StsUnsupportedFormat,
-        "MHI and orientation must be single-channel floating-point images" );
-
-    if( !CV_ARE_SIZES_EQ( mhi, mask ) || !CV_ARE_SIZES_EQ( orient, mhi ))
-        CV_Error( CV_StsUnmatchedSizes, "" );
-
-    if( mhi_duration <= 0 )
-        CV_Error( CV_StsOutOfRange, "MHI duration must be positive" );
-
-    if( orient->data.ptr == mhi->data.ptr )
-        CV_Error( CV_StsInplaceNotSupported, "orientation image must be different from MHI" );
-
-    // calculate histogram of different orientation values
-    hist = cvCreateHist( 1, &hist_size, CV_HIST_ARRAY, &ranges );
-    _orient = orient;
-    cvCalcArrHist( &_orient, hist, 0, mask );
-
-    // find the maximum index (the dominant orientation)
-    cvGetMinMaxHistValue( hist, 0, 0, 0, &base_orient );
-    fbase_orient = base_orient*360.f/hist_size;
-
-    // override timestamp with the maximum value in MHI
-    cvMinMaxLoc( mhi, 0, &curr_mhi_timestamp, 0, 0, mask );
-
-    // find the shift relative to the dominant orientation as weighted sum of relative angles
-    a = (float)(254. / 255. / mhi_duration);
-    b = (float)(1. - curr_mhi_timestamp * a);
-    delbound = (float)(curr_mhi_timestamp - mhi_duration);
-    mhi_rows = mhi->rows;
-    mhi_cols = mhi->cols;
-
-    if( CV_IS_MAT_CONT( mhi->type & mask->type & orient->type ))
-    {
-        mhi_cols *= mhi_rows;
-        mhi_rows = 1;
-    }
-
-    cvGetRow( mhi, &mhi_row, 0 );
-    cvGetRow( mask, &mask_row, 0 );
-    cvGetRow( orient, &orient_row, 0 );
-
-    /*
-       a = 254/(255*dt)
-       b = 1 - t*a = 1 - 254*t/(255*dur) =
-       (255*dt - 254*t)/(255*dt) =
-       (dt - (t - dt)*254)/(255*dt);
-       --------------------------------------------------------
-       ax + b = 254*x/(255*dt) + (dt - (t - dt)*254)/(255*dt) =
-       (254*x + dt - (t - dt)*254)/(255*dt) =
-       ((x - (t - dt))*254 + dt)/(255*dt) =
-       (((x - (t - dt))/dt)*254 + 1)/255 = (((x - low_time)/dt)*254 + 1)/255
-     */
-    for( y = 0; y < mhi_rows; y++ )
-    {
-        mhi_row.data.ptr = mhi->data.ptr + mhi->step*y;
-        mask_row.data.ptr = mask->data.ptr + mask->step*y;
-        orient_row.data.ptr = orient->data.ptr + orient->step*y;
-
-        for( x = 0; x < mhi_cols; x++ )
-            if( mask_row.data.ptr[x] != 0 && mhi_row.data.fl[x] > delbound )
-            {
-                /*
-                   orient in 0..360, base_orient in 0..360
-                   -> (rel_angle = orient - base_orient) in -360..360.
-                   rel_angle is translated to -180..180
-                 */
-                float weight = mhi_row.data.fl[x] * a + b;
-                float rel_angle = orient_row.data.fl[x] - fbase_orient;
-
-                rel_angle += (rel_angle < -180 ? 360 : 0);
-                rel_angle += (rel_angle > 180 ? -360 : 0);
-
-                if( fabs(rel_angle) < 45 )
-                {
-                    shift_orient += weight * rel_angle;
-                    shift_weight += weight;
-                }
-            }
-    }
-
-    // add the dominant orientation and the relative shift
-    if( shift_weight == 0 )
-        shift_weight = 0.01f;
-
-    fbase_orient += shift_orient / shift_weight;
-    fbase_orient -= (fbase_orient < 360 ? 0 : 360);
-    fbase_orient += (fbase_orient >= 0 ? 0 : 360);
-
-    return fbase_orient;
-}
-
-
-CV_IMPL CvSeq*
-cvSegmentMotion( const CvArr* mhiimg, CvArr* segmask, CvMemStorage* storage,
-                 double timestamp, double seg_thresh )
-{
-    CvSeq* components = 0;
-    cv::Ptr<CvMat> mask8u;
-
-    CvMat  mhistub, *mhi = cvGetMat(mhiimg, &mhistub);
-    CvMat  maskstub, *mask = cvGetMat(segmask, &maskstub);
-    Cv32suf v, comp_idx;
-    int stub_val, ts;
-    int x, y;
-
-    if( !storage )
-        CV_Error( CV_StsNullPtr, "NULL memory storage" );
-
-    mhi = cvGetMat( mhi, &mhistub );
-    mask = cvGetMat( mask, &maskstub );
-
-    if( CV_MAT_TYPE( mhi->type ) != CV_32FC1 || CV_MAT_TYPE( mask->type ) != CV_32FC1 )
-        CV_Error( CV_BadDepth, "Both MHI and the destination mask" );
-
-    if( !CV_ARE_SIZES_EQ( mhi, mask ))
-        CV_Error( CV_StsUnmatchedSizes, "" );
-
-    mask8u = cvCreateMat( mhi->rows + 2, mhi->cols + 2, CV_8UC1 );
-    cvZero( mask8u );
-    cvZero( mask );
-    components = cvCreateSeq( CV_SEQ_KIND_GENERIC, sizeof(CvSeq),
-                              sizeof(CvConnectedComp), storage );
-
-    v.f = (float)timestamp; ts = v.i;
-    v.f = FLT_MAX*0.1f; stub_val = v.i;
-    comp_idx.f = 1;
-
-    for( y = 0; y < mhi->rows; y++ )
-    {
-        int* mhi_row = (int*)(mhi->data.ptr + y*mhi->step);
-        for( x = 0; x < mhi->cols; x++ )
-        {
-            if( mhi_row[x] == 0 )
-                mhi_row[x] = stub_val;
-        }
-    }
-
-    for( y = 0; y < mhi->rows; y++ )
-    {
-        int* mhi_row = (int*)(mhi->data.ptr + y*mhi->step);
-        uchar* mask8u_row = mask8u->data.ptr + (y+1)*mask8u->step + 1;
-
-        for( x = 0; x < mhi->cols; x++ )
-        {
-            if( mhi_row[x] == ts && mask8u_row[x] == 0 )
-            {
-                CvConnectedComp comp;
-                int x1, y1;
-                CvScalar _seg_thresh = cvRealScalar(seg_thresh);
-                CvPoint seed = cvPoint(x,y);
-
-                cvFloodFill( mhi, seed, cvRealScalar(0), _seg_thresh, _seg_thresh,
-                            &comp, CV_FLOODFILL_MASK_ONLY + 2*256 + 4, mask8u );
-
-                for( y1 = 0; y1 < comp.rect.height; y1++ )
-                {
-                    int* mask_row1 = (int*)(mask->data.ptr +
-                                    (comp.rect.y + y1)*mask->step) + comp.rect.x;
-                    uchar* mask8u_row1 = mask8u->data.ptr +
-                                    (comp.rect.y + y1+1)*mask8u->step + comp.rect.x+1;
-
-                    for( x1 = 0; x1 < comp.rect.width; x1++ )
-                    {
-                        if( mask8u_row1[x1] > 1 )
-                        {
-                            mask8u_row1[x1] = 1;
-                            mask_row1[x1] = comp_idx.i;
-                        }
-                    }
-                }
-                comp_idx.f++;
-                cvSeqPush( components, &comp );
-            }
-        }
-    }
-
-    for( y = 0; y < mhi->rows; y++ )
-    {
-        int* mhi_row = (int*)(mhi->data.ptr + y*mhi->step);
-        for( x = 0; x < mhi->cols; x++ )
-        {
-            if( mhi_row[x] == stub_val )
-                mhi_row[x] = 0;
-        }
-    }
-
-    return components;
-}
-
-
-void cv::updateMotionHistory( InputArray _silhouette, InputOutputArray _mhi,
-                              double timestamp, double duration )
-{
-    Mat silhouette = _silhouette.getMat();
-    CvMat c_silhouette = silhouette, c_mhi = _mhi.getMat();
-    cvUpdateMotionHistory( &c_silhouette, &c_mhi, timestamp, duration );
-}
-
 void cv::calcMotionGradient( InputArray _mhi, OutputArray _mask,
                              OutputArray _orientation,
                              double delta1, double delta2,
                              int aperture_size )
 {
+    static int runcase = 0; runcase++;
+
     Mat mhi = _mhi.getMat();
-    _mask.create(mhi.size(), CV_8U);
-    _orientation.create(mhi.size(), CV_32F);
-    CvMat c_mhi = mhi, c_mask = _mask.getMat(), c_orientation = _orientation.getMat();
-    cvCalcMotionGradient(&c_mhi, &c_mask, &c_orientation, delta1, delta2, aperture_size);
+    Size size = mhi.size();
+
+    _mask.create(size, CV_8U);
+    _orientation.create(size, CV_32F);
+
+    Mat mask = _mask.getMat();
+    Mat orient = _orientation.getMat();
+
+    if( aperture_size < 3 || aperture_size > 7 || (aperture_size & 1) == 0 )
+        CV_Error( Error::StsOutOfRange, "aperture_size must be 3, 5 or 7" );
+
+    if( delta1 <= 0 || delta2 <= 0 )
+        CV_Error( Error::StsOutOfRange, "both delta's must be positive" );
+
+    if( mhi.type() != CV_32FC1 )
+        CV_Error( Error::StsUnsupportedFormat,
+                 "MHI must be single-channel floating-point images" );
+
+    if( orient.data == mhi.data )
+    {
+        _orientation.release();
+        _orientation.create(size, CV_32F);
+        orient = _orientation.getMat();
+    }
+
+    if( delta1 > delta2 )
+        std::swap(delta1, delta2);
+
+    float gradient_epsilon = 1e-4f * aperture_size * aperture_size;
+    float min_delta = (float)delta1;
+    float max_delta = (float)delta2;
+
+    Mat dX_min, dY_max;
+
+    // calc Dx and Dy
+    Sobel( mhi, dX_min, CV_32F, 1, 0, aperture_size, 1, 0, BORDER_REPLICATE );
+    Sobel( mhi, dY_max, CV_32F, 0, 1, aperture_size, 1, 0, BORDER_REPLICATE );
+
+    int x, y;
+
+    if( mhi.isContinuous() && orient.isContinuous() && mask.isContinuous() )
+    {
+        size.width *= size.height;
+        size.height = 1;
+    }
+
+    // calc gradient
+    for( y = 0; y < size.height; y++ )
+    {
+        const float* dX_min_row = dX_min.ptr<float>(y);
+        const float* dY_max_row = dY_max.ptr<float>(y);
+        float* orient_row = orient.ptr<float>(y);
+        uchar* mask_row = mask.ptr<uchar>(y);
+
+        fastAtan2(dY_max_row, dX_min_row, orient_row, size.width, true);
+
+        // make orientation zero where the gradient is very small
+        for( x = 0; x < size.width; x++ )
+        {
+            float dY = dY_max_row[x];
+            float dX = dX_min_row[x];
+
+            if( std::abs(dX) < gradient_epsilon && std::abs(dY) < gradient_epsilon )
+            {
+                mask_row[x] = (uchar)0;
+                orient_row[x] = 0.f;
+            }
+            else
+                mask_row[x] = (uchar)1;
+        }
+    }
+
+    erode( mhi, dX_min, noArray(), Point(-1,-1), (aperture_size-1)/2, BORDER_REPLICATE );
+    dilate( mhi, dY_max, noArray(), Point(-1,-1), (aperture_size-1)/2, BORDER_REPLICATE );
+
+    // mask off pixels which have little motion difference in their neighborhood
+    for( y = 0; y < size.height; y++ )
+    {
+        const float* dX_min_row = dX_min.ptr<float>(y);
+        const float* dY_max_row = dY_max.ptr<float>(y);
+        float* orient_row = orient.ptr<float>(y);
+        uchar* mask_row = mask.ptr<uchar>(y);
+
+        for( x = 0; x < size.width; x++ )
+        {
+            float d0 = dY_max_row[x] - dX_min_row[x];
+
+            if( mask_row[x] == 0 || d0 < min_delta || max_delta < d0 )
+            {
+                mask_row[x] = (uchar)0;
+                orient_row[x] = 0.f;
+            }
+        }
+    }
 }
 
 double cv::calcGlobalOrientation( InputArray _orientation, InputArray _mask,
-                                  InputArray _mhi, double timestamp,
+                                  InputArray _mhi, double /*timestamp*/,
                                   double duration )
 {
-    Mat orientation = _orientation.getMat(), mask = _mask.getMat(), mhi = _mhi.getMat();
-    CvMat c_orientation = orientation, c_mask = mask, c_mhi = mhi;
-    return cvCalcGlobalOrientation(&c_orientation, &c_mask, &c_mhi, timestamp, duration);
+    Mat orient = _orientation.getMat(), mask = _mask.getMat(), mhi = _mhi.getMat();
+    Size size = mhi.size();
+
+    CV_Assert( mask.type() == CV_8U && orient.type() == CV_32F && mhi.type() == CV_32F );
+    CV_Assert( mask.size() == size && orient.size() == size );
+    CV_Assert( duration > 0 );
+
+    int histSize = 12;
+    float _ranges[] = { 0.f, 360.f };
+    const float* ranges = _ranges;
+    Mat hist;
+
+    calcHist(&orient, 1, 0, mask, hist, 1, &histSize, &ranges);
+
+    // find the maximum index (the dominant orientation)
+    Point baseOrientPt;
+    minMaxLoc(hist, 0, 0, 0, &baseOrientPt);
+    float fbaseOrient = (baseOrientPt.x + baseOrientPt.y)*360.f/histSize;
+
+    // override timestamp with the maximum value in MHI
+    double timestamp = 0;
+    minMaxLoc( mhi, 0, &timestamp, 0, 0, mask );
+
+    // find the shift relative to the dominant orientation as weighted sum of relative angles
+    float a = (float)(254. / 255. / duration);
+    float b = (float)(1. - timestamp * a);
+    float delbound = (float)(timestamp - duration);
+
+    if( mhi.isContinuous() && mask.isContinuous() && orient.isContinuous() )
+    {
+        size.width *= size.height;
+        size.height = 1;
+    }
+
+    /*
+     a = 254/(255*dt)
+     b = 1 - t*a = 1 - 254*t/(255*dur) =
+     (255*dt - 254*t)/(255*dt) =
+     (dt - (t - dt)*254)/(255*dt);
+     --------------------------------------------------------
+     ax + b = 254*x/(255*dt) + (dt - (t - dt)*254)/(255*dt) =
+     (254*x + dt - (t - dt)*254)/(255*dt) =
+     ((x - (t - dt))*254 + dt)/(255*dt) =
+     (((x - (t - dt))/dt)*254 + 1)/255 = (((x - low_time)/dt)*254 + 1)/255
+     */
+    float shiftOrient = 0, shiftWeight = 0;
+    for( int y = 0; y < size.height; y++ )
+    {
+        const float* mhiptr = mhi.ptr<float>(y);
+        const float* oriptr = orient.ptr<float>(y);
+        const uchar* maskptr = mask.ptr<uchar>(y);
+
+        for( int x = 0; x < size.width; x++ )
+        {
+            if( maskptr[x] != 0 && mhiptr[x] > delbound )
+            {
+                /*
+                 orient in 0..360, base_orient in 0..360
+                 -> (rel_angle = orient - base_orient) in -360..360.
+                 rel_angle is translated to -180..180
+                 */
+                float weight = mhiptr[x] * a + b;
+                float relAngle = oriptr[x] - fbaseOrient;
+
+                relAngle += (relAngle < -180 ? 360 : 0);
+                relAngle += (relAngle > 180 ? -360 : 0);
+
+                if( fabs(relAngle) < 45 )
+                {
+                    shiftOrient += weight * relAngle;
+                    shiftWeight += weight;
+                }
+            }
+        }
+    }
+
+    // add the dominant orientation and the relative shift
+    if( shiftWeight == 0 )
+        shiftWeight = 0.01f;
+
+    fbaseOrient += shiftOrient / shiftWeight;
+    fbaseOrient -= (fbaseOrient < 360 ? 0 : 360);
+    fbaseOrient += (fbaseOrient >= 0 ? 0 : 360);
+
+    return fbaseOrient;
 }
 
+
 void cv::segmentMotion(InputArray _mhi, OutputArray _segmask,
                        std::vector<Rect>& boundingRects,
                        double timestamp, double segThresh)
 {
     Mat mhi = _mhi.getMat();
+
     _segmask.create(mhi.size(), CV_32F);
-    CvMat c_mhi = mhi, c_segmask = _segmask.getMat();
-    Ptr<CvMemStorage> storage = cvCreateMemStorage();
-    Seq<CvConnectedComp> comps = cvSegmentMotion(&c_mhi, &c_segmask, storage, timestamp, segThresh);
-    Seq<CvConnectedComp>::const_iterator it(comps);
-    size_t i, ncomps = comps.size();
-    boundingRects.resize(ncomps);
-    for( i = 0; i < ncomps; i++, ++it)
-        boundingRects[i] = (*it).rect;
+    Mat segmask = _segmask.getMat();
+    segmask = Scalar::all(0);
+
+    CV_Assert( mhi.type() == CV_32F );
+    CV_Assert( segThresh >= 0 );
+
+    Mat mask = Mat::zeros( mhi.rows + 2, mhi.cols + 2, CV_8UC1 );
+
+    int x, y;
+
+    // protect zero mhi pixels from floodfill.
+    for( y = 0; y < mhi.rows; y++ )
+    {
+        const float* mhiptr = mhi.ptr<float>(y);
+        uchar* maskptr = mask.ptr<uchar>(y+1) + 1;
+
+        for( x = 0; x < mhi.cols; x++ )
+        {
+            if( mhiptr[x] == 0 )
+                maskptr[x] = 1;
+        }
+    }
+
+    float ts = (float)timestamp;
+    float comp_idx = 1.f;
+
+    for( y = 0; y < mhi.rows; y++ )
+    {
+        float* mhiptr = mhi.ptr<float>(y);
+        uchar* maskptr = mask.ptr<uchar>(y+1) + 1;
+
+        for( x = 0; x < mhi.cols; x++ )
+        {
+            if( mhiptr[x] == ts && maskptr[x] == 0 )
+            {
+                Rect cc;
+                floodFill( mhi, mask, Point(x,y), Scalar::all(0),
+                           &cc, Scalar::all(segThresh), Scalar::all(segThresh),
+                           FLOODFILL_MASK_ONLY + 2*256 + 4 );
+
+                for( int y1 = 0; y1 < cc.height; y1++ )
+                {
+                    float* segmaskptr = segmask.ptr<float>(cc.y + y1) + cc.x;
+                    uchar* maskptr1 = mask.ptr<uchar>(cc.y + y1 + 1) + cc.x + 1;
+
+                    for( int x1 = 0; x1 < cc.width; x1++ )
+                    {
+                        if( maskptr1[x1] > 1 )
+                        {
+                            maskptr1[x1] = 1;
+                            segmaskptr[x1] = comp_idx;
+                        }
+                    }
+                }
+                comp_idx += 1.f;
+                boundingRects.push_back(cc);
+            }
+        }
+    }
 }
 
+
 /* End of file. */
diff --git a/modules/video/src/optflowgf.cpp b/modules/video/src/optflowgf.cpp
index 650e24262e..18dd3544e4 100644
--- a/modules/video/src/optflowgf.cpp
+++ b/modules/video/src/optflowgf.cpp
@@ -56,7 +56,7 @@ FarnebackPolyExp( const Mat& src, Mat& dst, int n, double sigma )
 {
     int k, x, y;
 
-    assert( src.type() == CV_32FC1 );
+    CV_Assert( src.type() == CV_32FC1 );
     int width = src.cols;
     int height = src.rows;
     AutoBuffer<float> kbuf(n*6 + 3), _row((width + n*2)*3);
@@ -627,7 +627,7 @@ void cv::calcOpticalFlowFarneback( InputArray _prev0, InputArray _next0,
         {
             img[i]->convertTo(fimg, CV_32F);
             GaussianBlur(fimg, fimg, Size(smooth_sz, smooth_sz), sigma, sigma);
-            resize( fimg, I, Size(width, height), CV_INTER_LINEAR );
+            resize( fimg, I, Size(width, height), INTER_LINEAR );
             FarnebackPolyExp( I, R[i], poly_n, poly_sigma );
         }
 
@@ -644,18 +644,3 @@ void cv::calcOpticalFlowFarneback( InputArray _prev0, InputArray _next0,
         prevFlow = flow;
     }
 }
-
-
-CV_IMPL void cvCalcOpticalFlowFarneback(
-            const CvArr* _prev, const CvArr* _next,
-            CvArr* _flow, double pyr_scale, int levels,
-            int winsize, int iterations, int poly_n,
-            double poly_sigma, int flags )
-{
-    cv::Mat prev = cv::cvarrToMat(_prev), next = cv::cvarrToMat(_next);
-    cv::Mat flow = cv::cvarrToMat(_flow);
-    CV_Assert( flow.size() == prev.size() && flow.type() == CV_32FC2 );
-    cv::calcOpticalFlowFarneback( prev, next, flow, pyr_scale, levels,
-        winsize, iterations, poly_n, poly_sigma, flags );
-}
-
diff --git a/modules/video/src/precomp.cpp b/modules/video/src/precomp.cpp
deleted file mode 100644
index 3e0ec42de9..0000000000
--- a/modules/video/src/precomp.cpp
+++ /dev/null
@@ -1,44 +0,0 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                        Intel License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2000, Intel Corporation, all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other materials provided with the distribution.
-//
-//   * The name of Intel Corporation may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
-#include "precomp.hpp"
-
-/* End of file. */
diff --git a/modules/video/src/precomp.hpp b/modules/video/src/precomp.hpp
index 58fc550862..43ff772cc3 100644
--- a/modules/video/src/precomp.hpp
+++ b/modules/video/src/precomp.hpp
@@ -43,16 +43,9 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
-#include "opencv2/video/tracking.hpp"
-#include "opencv2/video/background_segm.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/internal.hpp"
-
-#include <list>
+#include "opencv2/video.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/core/private.hpp"
 
 #ifdef HAVE_TEGRA_OPTIMIZATION
 #include "opencv2/video/video_tegra.hpp"
diff --git a/modules/video/src/simpleflow.cpp b/modules/video/src/simpleflow.cpp
index 8c5c746242..765a342704 100644
--- a/modules/video/src/simpleflow.cpp
+++ b/modules/video/src/simpleflow.cpp
@@ -41,7 +41,6 @@
 //M*/
 
 #include "precomp.hpp"
-#include "simpleflow.hpp"
 
 //
 // 2D dense optical flow algorithm from the following paper:
@@ -54,6 +53,39 @@
 namespace cv
 {
 
+static const uchar MASK_TRUE_VALUE = (uchar)255;
+
+inline static float dist(const Vec3b& p1, const Vec3b& p2) {
+  return (float)((p1[0] - p2[0]) * (p1[0] - p2[0]) +
+         (p1[1] - p2[1]) * (p1[1] - p2[1]) +
+         (p1[2] - p2[2]) * (p1[2] - p2[2]));
+}
+
+inline static float dist(const Vec2f& p1, const Vec2f& p2) {
+  return (p1[0] - p2[0]) * (p1[0] - p2[0]) +
+         (p1[1] - p2[1]) * (p1[1] - p2[1]);
+}
+
+inline static float dist(const Point2f& p1, const Point2f& p2) {
+  return (p1.x - p2.x) * (p1.x - p2.x) +
+         (p1.y - p2.y) * (p1.y - p2.y);
+}
+
+inline static float dist(float x1, float y1, float x2, float y2) {
+  return (x1 - x2) * (x1 - x2) +
+         (y1 - y2) * (y1 - y2);
+}
+
+inline static int dist(int x1, int y1, int x2, int y2) {
+  return (x1 - x2) * (x1 - x2) +
+         (y1 - y2) * (y1 - y2);
+}
+
+template<class T>
+inline static T min(T t1, T t2, T t3) {
+  return (t1 <= t2 && t1 <= t3) ? t1 : min(t2, t3);
+}
+
 static void removeOcclusions(const Mat& flow,
                              const Mat& flow_inv,
                              float occ_thr,
@@ -287,7 +319,7 @@ static Mat upscaleOpticalFlow(int new_rows,
 static Mat calcIrregularityMat(const Mat& flow, int radius) {
   const int rows = flow.rows;
   const int cols = flow.cols;
-  Mat irregularity(rows, cols, CV_32F);
+  Mat irregularity = Mat::zeros(rows, cols, CV_32F);
   for (int r = 0; r < rows; ++r) {
     const int start_row = std::max(0, r - radius);
     const int end_row = std::min(rows - 1, r + radius);
@@ -409,7 +441,7 @@ static void extrapolateFlow(Mat& flow,
                             const Mat& speed_up) {
   const int rows = flow.rows;
   const int cols = flow.cols;
-  Mat done(rows, cols, CV_8U);
+  Mat done = Mat::zeros(rows, cols, CV_8U);
   for (int r = 0; r < rows; ++r) {
     for (int c = 0; c < cols; ++c) {
       if (!done.at<uchar>(r, c) && speed_up.at<uchar>(r, c) > 1) {
@@ -508,8 +540,8 @@ CV_EXPORTS_W void calcOpticalFlowSF(InputArray _from,
   Mat mask = Mat::ones(curr_from.size(), CV_8U);
   Mat mask_inv = Mat::ones(curr_from.size(), CV_8U);
 
-  Mat flow(curr_from.size(), CV_32FC2);
-  Mat flow_inv(curr_to.size(), CV_32FC2);
+  Mat flow = Mat::zeros(curr_from.size(), CV_32FC2);
+  Mat flow_inv = Mat::zeros(curr_to.size(), CV_32FC2);
 
   Mat confidence;
   Mat confidence_inv;
@@ -653,4 +685,3 @@ CV_EXPORTS_W void calcOpticalFlowSF(InputArray from,
 }
 
 }
-
diff --git a/modules/video/src/tvl1flow.cpp b/modules/video/src/tvl1flow.cpp
index 7aa7debd45..8d59932750 100644
--- a/modules/video/src/tvl1flow.cpp
+++ b/modules/video/src/tvl1flow.cpp
@@ -73,6 +73,7 @@
 */
 
 #include "precomp.hpp"
+#include <limits>
 
 using namespace cv;
 
@@ -95,8 +96,11 @@ protected:
     int nscales;
     int warps;
     double epsilon;
-    int iterations;
+    int innerIterations;
+    int outerIterations;
     bool useInitialFlow;
+    double scaleStep;
+    int medianFiltering;
 
 private:
     void procOneScale(const Mat_<float>& I0, const Mat_<float>& I1, Mat_<float>& u1, Mat_<float>& u2);
@@ -144,8 +148,11 @@ OpticalFlowDual_TVL1::OpticalFlowDual_TVL1()
     nscales        = 5;
     warps          = 5;
     epsilon        = 0.01;
-    iterations     = 300;
+    innerIterations = 30;
+    outerIterations = 10;
     useInitialFlow = false;
+    medianFiltering = 5;
+    scaleStep      = 0.8;
 }
 
 void OpticalFlowDual_TVL1::calc(InputArray _I0, InputArray _I1, InputOutputArray _flow)
@@ -168,13 +175,12 @@ void OpticalFlowDual_TVL1::calc(InputArray _I0, InputArray _I1, InputOutputArray
     I0.convertTo(I0s[0], I0s[0].depth(), I0.depth() == CV_8U ? 1.0 : 255.0);
     I1.convertTo(I1s[0], I1s[0].depth(), I1.depth() == CV_8U ? 1.0 : 255.0);
 
+    u1s[0].create(I0.size());
+    u2s[0].create(I0.size());
+
     if (useInitialFlow)
     {
-        u1s[0].create(I0.size());
-        u2s[0].create(I0.size());
-
         Mat_<float> mv[] = {u1s[0], u2s[0]};
-
         split(_flow.getMat(), mv);
     }
 
@@ -210,8 +216,8 @@ void OpticalFlowDual_TVL1::calc(InputArray _I0, InputArray _I1, InputOutputArray
     // create the scales
     for (int s = 1; s < nscales; ++s)
     {
-        pyrDown(I0s[s - 1], I0s[s]);
-        pyrDown(I1s[s - 1], I1s[s]);
+        resize(I0s[s-1], I0s[s], Size(), scaleStep, scaleStep);
+        resize(I1s[s-1], I1s[s], Size(), scaleStep, scaleStep);
 
         if (I0s[s].cols < 16 || I0s[s].rows < 16)
         {
@@ -221,12 +227,23 @@ void OpticalFlowDual_TVL1::calc(InputArray _I0, InputArray _I1, InputOutputArray
 
         if (useInitialFlow)
         {
-            pyrDown(u1s[s - 1], u1s[s]);
-            pyrDown(u2s[s - 1], u2s[s]);
+            resize(u1s[s-1], u1s[s], Size(), scaleStep, scaleStep);
+            resize(u2s[s-1], u2s[s], Size(), scaleStep, scaleStep);
 
-            multiply(u1s[s], Scalar::all(0.5), u1s[s]);
-            multiply(u2s[s], Scalar::all(0.5), u2s[s]);
+            multiply(u1s[s], Scalar::all(scaleStep), u1s[s]);
+            multiply(u2s[s], Scalar::all(scaleStep), u2s[s]);
         }
+        else
+        {
+            u1s[s].create(I0s[s].size());
+            u2s[s].create(I0s[s].size());
+        }
+    }
+
+    if (!useInitialFlow)
+    {
+        u1s[nscales-1].setTo(Scalar::all(0));
+        u2s[nscales-1].setTo(Scalar::all(0));
     }
 
     // pyramidal structure for computing the optical flow
@@ -246,8 +263,8 @@ void OpticalFlowDual_TVL1::calc(InputArray _I0, InputArray _I1, InputOutputArray
         resize(u2s[s], u2s[s - 1], I0s[s - 1].size());
 
         // scale the optical flow with the appropriate zoom factor
-        multiply(u1s[s - 1], Scalar::all(2), u1s[s - 1]);
-        multiply(u2s[s - 1], Scalar::all(2), u2s[s - 1]);
+        multiply(u1s[s - 1], Scalar::all(1/scaleStep), u1s[s - 1]);
+        multiply(u2s[s - 1], Scalar::all(1/scaleStep), u2s[s - 1]);
     }
 
     Mat uxy[] = {u1s[0], u2s[0]};
@@ -792,18 +809,9 @@ void OpticalFlowDual_TVL1::procOneScale(const Mat_<float>& I0, const Mat_<float>
 
     CV_DbgAssert( I1.size() == I0.size() );
     CV_DbgAssert( I1.type() == I0.type() );
-    CV_DbgAssert( u1.empty() || u1.size() == I0.size() );
+    CV_DbgAssert( u1.size() == I0.size() );
     CV_DbgAssert( u2.size() == u1.size() );
 
-    if (u1.empty())
-    {
-        u1.create(I0.size());
-        u1.setTo(Scalar::all(0));
-
-        u2.create(I0.size());
-        u2.setTo(Scalar::all(0));
-    }
-
     Mat_<float> I1x = I1x_buf(Rect(0, 0, I0.cols, I0.rows));
     Mat_<float> I1y = I1y_buf(Rect(0, 0, I0.cols, I0.rows));
     centeredGradient(I1, I1x, I1y);
@@ -852,24 +860,31 @@ void OpticalFlowDual_TVL1::procOneScale(const Mat_<float>& I0, const Mat_<float>
         calcGradRho(I0, I1w, I1wx, I1wy, u1, u2, grad, rho_c);
 
         float error = std::numeric_limits<float>::max();
-        for (int n = 0; error > scaledEpsilon && n < iterations; ++n)
+        for (int n_outer = 0; error > scaledEpsilon && n_outer < outerIterations; ++n_outer)
         {
-            // estimate the values of the variable (v1, v2) (thresholding operator TH)
-            estimateV(I1wx, I1wy, u1, u2, grad, rho_c, v1, v2, l_t);
+            if (medianFiltering > 1) {
+                cv::medianBlur(u1, u1, medianFiltering);
+                cv::medianBlur(u2, u2, medianFiltering);
+            }
+            for (int n_inner = 0; error > scaledEpsilon && n_inner < innerIterations; ++n_inner)
+            {
+                // estimate the values of the variable (v1, v2) (thresholding operator TH)
+                estimateV(I1wx, I1wy, u1, u2, grad, rho_c, v1, v2, l_t);
 
-            // compute the divergence of the dual variable (p1, p2)
-            divergence(p11, p12, div_p1);
-            divergence(p21, p22, div_p2);
+                // compute the divergence of the dual variable (p1, p2)
+                divergence(p11, p12, div_p1);
+                divergence(p21, p22, div_p2);
 
-            // estimate the values of the optical flow (u1, u2)
-            error = estimateU(v1, v2, div_p1, div_p2, u1, u2, static_cast<float>(theta));
+                // estimate the values of the optical flow (u1, u2)
+                error = estimateU(v1, v2, div_p1, div_p2, u1, u2, static_cast<float>(theta));
 
-            // compute the gradient of the optical flow (Du1, Du2)
-            forwardGradient(u1, u1x, u1y);
-            forwardGradient(u2, u2x, u2y);
+                // compute the gradient of the optical flow (Du1, Du2)
+                forwardGradient(u1, u1x, u1y);
+                forwardGradient(u2, u2x, u2y);
 
-            // estimate the values of the dual variable (p1, p2)
-            estimateDualVariables(u1x, u1y, u2x, u2y, p11, p12, p21, p22, taut);
+                // estimate the values of the dual variable (p1, p2)
+                estimateDualVariables(u1x, u1y, u2x, u2y, p11, p12, p21, p22, taut);
+            }
         }
     }
 }
@@ -922,15 +937,21 @@ CV_INIT_ALGORITHM(OpticalFlowDual_TVL1, "DenseOpticalFlow.DualTVL1",
                                        "Number of scales used to create the pyramid of images");
                   obj.info()->addParam(obj, "warps", obj.warps, false, 0, 0,
                                        "Number of warpings per scale");
+                  obj.info()->addParam(obj, "medianFiltering", obj.medianFiltering, false, 0, 0,
+                                       "Median filter kernel size (1 = no filter) (3 or 5)");
+                  obj.info()->addParam(obj, "scaleStep", obj.scaleStep, false, 0, 0,
+                                       "Step between scales (<1)");
                   obj.info()->addParam(obj, "epsilon", obj.epsilon, false, 0, 0,
                                        "Stopping criterion threshold used in the numerical scheme, which is a trade-off between precision and running time");
-                  obj.info()->addParam(obj, "iterations", obj.iterations, false, 0, 0,
-                                       "Stopping criterion iterations number used in the numerical scheme");
+                  obj.info()->addParam(obj, "innerIterations", obj.innerIterations, false, 0, 0,
+                                       "inner iterations (between outlier filtering) used in the numerical scheme");
+                  obj.info()->addParam(obj, "outerIterations", obj.outerIterations, false, 0, 0,
+                                       "outer iterations (number of inner loops) used in the numerical scheme");
                   obj.info()->addParam(obj, "useInitialFlow", obj.useInitialFlow));
 
 } // namespace
 
 Ptr<DenseOpticalFlow> cv::createOptFlow_DualTVL1()
 {
-    return new OpticalFlowDual_TVL1;
+    return makePtr<OpticalFlowDual_TVL1>();
 }
diff --git a/modules/video/src/video_init.cpp b/modules/video/src/video_init.cpp
index 189f4a36ba..118d28a42e 100644
--- a/modules/video/src/video_init.cpp
+++ b/modules/video/src/video_init.cpp
@@ -46,50 +46,9 @@
 namespace cv
 {
 
-///////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-CV_INIT_ALGORITHM(BackgroundSubtractorMOG, "BackgroundSubtractor.MOG",
-    obj.info()->addParam(obj, "history", obj.history);
-    obj.info()->addParam(obj, "nmixtures", obj.nmixtures);
-    obj.info()->addParam(obj, "backgroundRatio", obj.backgroundRatio);
-    obj.info()->addParam(obj, "noiseSigma", obj.noiseSigma));
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-CV_INIT_ALGORITHM(BackgroundSubtractorMOG2, "BackgroundSubtractor.MOG2",
-    obj.info()->addParam(obj, "history", obj.history);
-    obj.info()->addParam(obj, "nmixtures", obj.nmixtures);
-    obj.info()->addParam(obj, "varThreshold", obj.varThreshold);
-    obj.info()->addParam(obj, "detectShadows", obj.bShadowDetection));
-
-///////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-CV_INIT_ALGORITHM(BackgroundSubtractorGMG, "BackgroundSubtractor.GMG",
-                  obj.info()->addParam(obj, "maxFeatures", obj.maxFeatures,false,0,0,
-                                       "Maximum number of features to store in histogram. Harsh enforcement of sparsity constraint.");
-                  obj.info()->addParam(obj, "learningRate", obj.learningRate,false,0,0,
-                                       "Adaptation rate of histogram. Close to 1, slow adaptation. Close to 0, fast adaptation, features forgotten quickly.");
-                  obj.info()->addParam(obj, "initializationFrames", obj.numInitializationFrames,false,0,0,
-                                       "Number of frames to use to initialize histograms of pixels.");
-                  obj.info()->addParam(obj, "quantizationLevels", obj.quantizationLevels,false,0,0,
-                                       "Number of discrete colors to be used in histograms. Up-front quantization.");
-                  obj.info()->addParam(obj, "backgroundPrior", obj.backgroundPrior,false,0,0,
-                                       "Prior probability that each individual pixel is a background pixel.");
-                  obj.info()->addParam(obj, "smoothingRadius", obj.smoothingRadius,false,0,0,
-                                       "Radius of smoothing kernel to filter noise from FG mask image.");
-                  obj.info()->addParam(obj, "decisionThreshold", obj.decisionThreshold,false,0,0,
-                                       "Threshold for FG decision rule. Pixel is FG if posterior probability exceeds threshold.");
-                  obj.info()->addParam(obj, "updateBackgroundModel", obj.updateBackgroundModel,false,0,0,
-                                       "Perform background model update."));
-
 bool initModule_video(void)
 {
-    bool all = true;
-    all &= !BackgroundSubtractorMOG_info_auto.name().empty();
-    all &= !BackgroundSubtractorMOG2_info_auto.name().empty();
-    all &= !BackgroundSubtractorGMG_info_auto.name().empty();
-
-    return all;
+    return true;
 }
 
 }
diff --git a/modules/video/test/test_accum.cpp b/modules/video/test/test_accum.cpp
index b7c6b719ec..ef6c05f980 100644
--- a/modules/video/test/test_accum.cpp
+++ b/modules/video/test/test_accum.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/imgproc/imgproc_c.h"
 
 using namespace cv;
 using namespace std;
diff --git a/modules/video/test/test_backgroundsubtractor_gbh.cpp b/modules/video/test/test_backgroundsubtractor_gbh.cpp
index 00d6fb05e6..99d53e3d32 100644
--- a/modules/video/test/test_backgroundsubtractor_gbh.cpp
+++ b/modules/video/test/test_backgroundsubtractor_gbh.cpp
@@ -37,29 +37,22 @@ void CV_BackgroundSubtractorTest::run(int)
     int width = 2 + ((unsigned int)rng)%98; //!< Mat will be 2 to 100 in width and height
     int height = 2 + ((unsigned int)rng)%98;
 
-    Ptr<BackgroundSubtractorGMG> fgbg =
-            Algorithm::create<BackgroundSubtractorGMG>("BackgroundSubtractor.GMG");
+    Ptr<BackgroundSubtractorGMG> fgbg = createBackgroundSubtractorGMG();
     Mat fgmask;
 
-    if (fgbg.empty())
-        CV_Error(CV_StsError,"Failed to create Algorithm\n");
+    if (!fgbg)
+        CV_Error(Error::StsError,"Failed to create Algorithm\n");
 
     /**
      * Set a few parameters
      */
-    fgbg->set("smoothingRadius",7);
-    fgbg->set("decisionThreshold",0.7);
-    fgbg->set("initializationFrames",120);
+    fgbg->setSmoothingRadius(7);
+    fgbg->setDecisionThreshold(0.7);
+    fgbg->setNumFrames(120);
 
     /**
      * Generate bounds for the values in the matrix for each type
      */
-    uchar maxuc = 0, minuc = 0;
-    char maxc = 0, minc = 0;
-    unsigned int maxui = 0, minui = 0;
-    int maxi=0, mini = 0;
-    long int maxli = 0, minli = 0;
-    float maxf = 0, minf = 0;
     double maxd = 0, mind = 0;
 
     /**
@@ -69,34 +62,34 @@ void CV_BackgroundSubtractorTest::run(int)
     if (type == CV_8U)
     {
         uchar half = UCHAR_MAX/2;
-        maxuc = (unsigned char)rng.uniform(half+32, UCHAR_MAX);
-        minuc = (unsigned char)rng.uniform(0, half-32);
+        maxd = (unsigned char)rng.uniform(half+32, UCHAR_MAX);
+        mind = (unsigned char)rng.uniform(0, half-32);
     }
     else if (type == CV_8S)
     {
-        maxc = (char)rng.uniform(32, CHAR_MAX);
-        minc = (char)rng.uniform(CHAR_MIN, -32);
+        maxd = (char)rng.uniform(32, CHAR_MAX);
+        mind = (char)rng.uniform(CHAR_MIN, -32);
     }
     else if (type == CV_16U)
     {
         ushort half = USHRT_MAX/2;
-        maxui = (unsigned int)rng.uniform(half+32, USHRT_MAX);
-        minui = (unsigned int)rng.uniform(0, half-32);
+        maxd = (unsigned int)rng.uniform(half+32, USHRT_MAX);
+        mind = (unsigned int)rng.uniform(0, half-32);
     }
     else if (type == CV_16S)
     {
-        maxi = rng.uniform(32, SHRT_MAX);
-        mini = rng.uniform(SHRT_MIN, -32);
+        maxd = rng.uniform(32, SHRT_MAX);
+        mind = rng.uniform(SHRT_MIN, -32);
     }
     else if (type == CV_32S)
     {
-        maxli = rng.uniform(32, INT_MAX);
-        minli = rng.uniform(INT_MIN, -32);
+        maxd = rng.uniform(32, INT_MAX);
+        mind = rng.uniform(INT_MIN, -32);
     }
     else if (type == CV_32F)
     {
-        maxf = rng.uniform(32.0f, FLT_MAX);
-        minf = rng.uniform(-FLT_MAX, -32.0f);
+        maxd = rng.uniform(32.0f, FLT_MAX);
+        mind = rng.uniform(-FLT_MAX, -32.0f);
     }
     else if (type == CV_64F)
     {
@@ -104,60 +97,22 @@ void CV_BackgroundSubtractorTest::run(int)
         mind = rng.uniform(-DBL_MAX, -32.0);
     }
 
+    fgbg->setMinVal(mind);
+    fgbg->setMaxVal(maxd);
+
     Mat simImage = Mat::zeros(height, width, channelsAndType);
-    const unsigned int numLearningFrames = 120;
-    for (unsigned int i = 0; i < numLearningFrames; ++i)
+    int numLearningFrames = 120;
+    for (int i = 0; i < numLearningFrames; ++i)
     {
         /**
          * Genrate simulated "image" for any type. Values always confined to upper half of range.
          */
-        if (type == CV_8U)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(unsigned char)(minuc/2+maxuc/2),maxuc);
-            if (i == 0)
-                fgbg->initialize(simImage.size(),minuc,maxuc);
-        }
-        else if (type == CV_8S)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(char)(minc/2+maxc/2),maxc);
-            if (i==0)
-                fgbg->initialize(simImage.size(),minc,maxc);
-        }
-        else if (type == CV_16U)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(unsigned int)(minui/2+maxui/2),maxui);
-            if (i==0)
-                fgbg->initialize(simImage.size(),minui,maxui);
-        }
-        else if (type == CV_16S)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(int)(mini/2+maxi/2),maxi);
-            if (i==0)
-                fgbg->initialize(simImage.size(),mini,maxi);
-        }
-        else if (type == CV_32F)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(float)(minf/2.0+maxf/2.0),maxf);
-            if (i==0)
-                fgbg->initialize(simImage.size(),minf,maxf);
-        }
-        else if (type == CV_32S)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(long int)(minli/2+maxli/2),maxli);
-            if (i==0)
-                fgbg->initialize(simImage.size(),minli,maxli);
-        }
-        else if (type == CV_64F)
-        {
-            rng.fill(simImage,RNG::UNIFORM,(double)(mind/2.0+maxd/2.0),maxd);
-            if (i==0)
-                fgbg->initialize(simImage.size(),mind,maxd);
-        }
+        rng.fill(simImage, RNG::UNIFORM, (mind + maxd)*0.5, maxd);
 
         /**
          * Feed simulated images into background subtractor
          */
-        (*fgbg)(simImage,fgmask);
+        fgbg->apply(simImage,fgmask);
         Mat fullbg = Mat::zeros(simImage.rows, simImage.cols, CV_8U);
 
         //! fgmask should be entirely background during training
@@ -166,22 +121,9 @@ void CV_BackgroundSubtractorTest::run(int)
             ts->set_failed_test_info( code );
     }
     //! generate last image, distinct from training images
-    if (type == CV_8U)
-        rng.fill(simImage,RNG::UNIFORM,minuc,minuc);
-    else if (type == CV_8S)
-        rng.fill(simImage,RNG::UNIFORM,minc,minc);
-    else if (type == CV_16U)
-        rng.fill(simImage,RNG::UNIFORM,minui,minui);
-    else if (type == CV_16S)
-        rng.fill(simImage,RNG::UNIFORM,mini,mini);
-    else if (type == CV_32F)
-        rng.fill(simImage,RNG::UNIFORM,minf,minf);
-    else if (type == CV_32S)
-        rng.fill(simImage,RNG::UNIFORM,minli,minli);
-    else if (type == CV_64F)
-        rng.fill(simImage,RNG::UNIFORM,mind,mind);
+    rng.fill(simImage, RNG::UNIFORM, mind, maxd);
 
-    (*fgbg)(simImage,fgmask);
+    fgbg->apply(simImage,fgmask);
     //! now fgmask should be entirely foreground
     Mat fullfg = 255*Mat::ones(simImage.rows, simImage.cols, CV_8U);
     code = cvtest::cmpEps2( ts, fgmask, fullfg, 255, false, "The final foreground mask" );
diff --git a/modules/video/test/test_camshift.cpp b/modules/video/test/test_camshift.cpp
index 7dcfb19de2..f003286d4d 100644
--- a/modules/video/test/test_camshift.cpp
+++ b/modules/video/test/test_camshift.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/video/tracking_c.h"
 
 using namespace cv;
 using namespace std;
@@ -345,7 +346,7 @@ _exit_:
 
     if( code < 0 )
     {
-#if defined _DEBUG && defined WIN32
+#if 0 //defined _DEBUG && defined WIN32
         IplImage* dst = cvCreateImage( img_size, 8, 3 );
         cvNamedWindow( "test", 1 );
         cvCmpS( img, 0, img, CV_CMP_GT );
@@ -484,7 +485,7 @@ _exit_:
 
     if( code < 0 )
     {
-#if defined _DEBUG && defined WIN32
+#if 0// defined _DEBUG && defined WIN32
         IplImage* dst = cvCreateImage( img_size, 8, 3 );
         cvNamedWindow( "test", 1 );
         cvCmpS( img, 0, img, CV_CMP_GT );
diff --git a/modules/video/test/test_ecc.cpp b/modules/video/test/test_ecc.cpp
index 58389290bd..4065e6182e 100644
--- a/modules/video/test/test_ecc.cpp
+++ b/modules/video/test/test_ecc.cpp
@@ -138,7 +138,7 @@ bool CV_ECC_Test_Translation::testTranslation(int from)
         Mat warpedImage;
 
         warpAffine(testImg, warpedImage, translationGround,
-            Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+            Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
 
         Mat mapTranslation = (Mat_<float>(2,3) << 1, 0, 0, 0, 1, 0);
 
@@ -213,7 +213,7 @@ bool CV_ECC_Test_Euclidean::testEuclidean(int from)
         Mat warpedImage;
 
         warpAffine(testImg, warpedImage, euclideanGround,
-            Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+            Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
 
         Mat mapEuclidean = (Mat_<float>(2,3) << 1, 0, 0, 0, 1, 0);
 
@@ -288,7 +288,7 @@ bool CV_ECC_Test_Affine::testAffine(int from)
         Mat warpedImage;
 
         warpAffine(testImg, warpedImage, affineGround,
-            Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+            Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
 
         Mat mapAffine = (Mat_<float>(2,3) << 1, 0, 0, 0, 1, 0);
 
@@ -363,7 +363,7 @@ bool CV_ECC_Test_Homography::testHomography(int from)
         Mat warpedImage;
 
         warpPerspective(testImg, warpedImage, homoGround,
-            Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+            Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
 
         Mat mapHomography = Mat::eye(3, 3, CV_32F);
 
diff --git a/modules/video/test/test_estimaterigid.cpp b/modules/video/test/test_estimaterigid.cpp
index ddfd007a18..5259ce7ee2 100644
--- a/modules/video/test/test_estimaterigid.cpp
+++ b/modules/video/test/test_estimaterigid.cpp
@@ -93,18 +93,18 @@ bool CV_RigidTransform_Test::testNPoints(int from)
         progress = update_progress(progress, k, ntests, 0);
 
         Mat aff(2, 3, CV_64F);
-        rng.fill(aff, CV_RAND_UNI, Scalar(-2), Scalar(2));
+        rng.fill(aff, RNG::UNIFORM, Scalar(-2), Scalar(2));
 
         int n = (unsigned)rng % 100 + 10;
 
         Mat fpts(1, n, CV_32FC2);
         Mat tpts(1, n, CV_32FC2);
 
-        rng.fill(fpts, CV_RAND_UNI, Scalar(0,0), Scalar(10,10));
+        rng.fill(fpts, RNG::UNIFORM, Scalar(0,0), Scalar(10,10));
         transform(fpts.ptr<Point2f>(), fpts.ptr<Point2f>() + n, tpts.ptr<Point2f>(), WrapAff2D(aff));
 
         Mat noise(1, n, CV_32FC2);
-        rng.fill(noise, CV_RAND_NORMAL, Scalar::all(0), Scalar::all(0.001*(n<=7 ? 0 : n <= 30 ? 1 : 10)));
+        rng.fill(noise, RNG::NORMAL, Scalar::all(0), Scalar::all(0.001*(n<=7 ? 0 : n <= 30 ? 1 : 10)));
         tpts += noise;
 
         Mat aff_est = estimateRigidTransform(fpts, tpts, true);
@@ -153,7 +153,7 @@ bool CV_RigidTransform_Test::testImage()
 
     Mat aff_est = estimateRigidTransform(img, rotated, true);
 
-    const double thres = 0.03;
+    const double thres = 0.033;
     if (norm(aff_est, aff, NORM_INF) > thres)
     {
         ts->set_failed_test_info(cvtest::TS::FAIL_BAD_ACCURACY);
diff --git a/modules/video/test/test_kalman.cpp b/modules/video/test/test_kalman.cpp
index 17b3ee6a43..2aa74e1c21 100644
--- a/modules/video/test/test_kalman.cpp
+++ b/modules/video/test/test_kalman.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/video/tracking_c.h"
 
 using namespace cv;
 
diff --git a/modules/video/test/test_motiontemplates.cpp b/modules/video/test/test_motiontemplates.cpp
index 082131737b..7779327b10 100644
--- a/modules/video/test/test_motiontemplates.cpp
+++ b/modules/video/test/test_motiontemplates.cpp
@@ -169,9 +169,7 @@ double CV_UpdateMHITest::get_success_error_level( int /*test_case_idx*/, int /*i
 
 void CV_UpdateMHITest::run_func()
 {
-    CvMat m = test_mat[INPUT_OUTPUT][0];
     cv::updateMotionHistory( test_mat[INPUT][0], test_mat[INPUT_OUTPUT][0], timestamp, duration);
-    m = test_mat[INPUT_OUTPUT][0];
 }
 
 
@@ -203,8 +201,7 @@ static void test_MHIGradient( const Mat& mhi, Mat& mask, Mat& orientation,
 
     if( delta1 > delta2 )
     {
-        double t;
-        CV_SWAP( delta1, delta2, t );
+        std::swap( delta1, delta2 );
     }
 
     for( int i = 0; i < mhi.rows; i++ )
@@ -417,7 +414,7 @@ void CV_MHIGlobalOrientTest::get_test_array_types_and_sizes( int test_case_idx,
 {
     RNG& rng = ts->get_rng();
     CV_MHIBaseTest::get_test_array_types_and_sizes( test_case_idx, sizes, types );
-    CvSize size = sizes[INPUT][0];
+    Size size = sizes[INPUT][0];
 
     size.width = MAX( size.width, 16 );
     size.height = MAX( size.height, 16 );
@@ -430,8 +427,7 @@ void CV_MHIGlobalOrientTest::get_test_array_types_and_sizes( int test_case_idx,
     max_angle = cvtest::randReal(rng)*359.9;
     if( min_angle >= max_angle )
     {
-        double t;
-        CV_SWAP( min_angle, max_angle, t );
+        std::swap( min_angle, max_angle);
     }
     max_angle += 0.1;
     duration = exp(cvtest::randReal(rng)*max_log_duration);
@@ -474,7 +470,7 @@ int CV_MHIGlobalOrientTest::validate_test_results( int test_case_idx )
                                                    test_mat[INPUT][0], timestamp, duration );
     double err_level = get_success_error_level( test_case_idx, 0, 0 );
     int code = cvtest::TS::OK;
-    int nz = cvCountNonZero( test_array[INPUT][1] );
+    int nz = countNonZero( test_mat[INPUT][1] );
 
     if( nz > 32 && !(min_angle - err_level <= angle &&
           max_angle + err_level >= angle) &&
diff --git a/modules/video/test/test_optflowpyrlk.cpp b/modules/video/test/test_optflowpyrlk.cpp
index 9a9035b98d..e4aa5e5721 100644
--- a/modules/video/test/test_optflowpyrlk.cpp
+++ b/modules/video/test/test_optflowpyrlk.cpp
@@ -40,6 +40,7 @@
 //M*/
 
 #include "test_precomp.hpp"
+#include "opencv2/video/tracking_c.h"
 
 /* ///////////////////// pyrlk_test ///////////////////////// */
 
@@ -72,8 +73,9 @@ void CV_OptFlowPyrLKTest::run( int )
     CvMat *_u = 0, *_v = 0, *_v2 = 0;
     char* status = 0;
 
-    IplImage* imgI = 0;
-    IplImage* imgJ = 0;
+    IplImage imgI;
+    IplImage imgJ;
+    cv::Mat  imgI2, imgJ2;
 
     int  n = 0, i = 0;
 
@@ -115,9 +117,10 @@ void CV_OptFlowPyrLKTest::run( int )
 
     /* read first image */
     sprintf( filename, "%soptflow/%s", ts->get_data_path().c_str(), "rock_1.bmp" );
-    imgI = cvLoadImage( filename, -1 );
+    imgI2 = cv::imread( filename, cv::IMREAD_UNCHANGED );
+    imgI = imgI2;
 
-    if( !imgI )
+    if( imgI2.empty() )
     {
         ts->printf( cvtest::TS::LOG, "could not read %s\n", filename );
         code = cvtest::TS::FAIL_MISSING_TEST_DATA;
@@ -126,9 +129,10 @@ void CV_OptFlowPyrLKTest::run( int )
 
     /* read second image */
     sprintf( filename, "%soptflow/%s", ts->get_data_path().c_str(), "rock_2.bmp" );
-    imgJ = cvLoadImage( filename, -1 );
+    imgJ2 = cv::imread( filename, cv::IMREAD_UNCHANGED );
+    imgJ = imgJ2;
 
-    if( !imgJ )
+    if( imgJ2.empty() )
     {
         ts->printf( cvtest::TS::LOG, "could not read %s\n", filename );
         code = cvtest::TS::FAIL_MISSING_TEST_DATA;
@@ -139,7 +143,7 @@ void CV_OptFlowPyrLKTest::run( int )
     status = (char*)cvAlloc(n*sizeof(status[0]));
 
     /* calculate flow */
-    cvCalcOpticalFlowPyrLK( imgI, imgJ, 0, 0, u, v2, n, cvSize( 41, 41 ),
+    cvCalcOpticalFlowPyrLK( &imgI, &imgJ, 0, 0, u, v2, n, cvSize( 41, 41 ),
                             4, status, 0, cvTermCriteria( CV_TERMCRIT_ITER|
                             CV_TERMCRIT_EPS, 30, 0.01f ), 0 );
 
@@ -201,9 +205,6 @@ _exit_:
     cvReleaseMat( &_v );
     cvReleaseMat( &_v2 );
 
-    cvReleaseImage( &imgI );
-    cvReleaseImage( &imgJ );
-
     if( code < 0 )
         ts->set_failed_test_info( code );
 }
diff --git a/modules/video/test/test_precomp.cpp b/modules/video/test/test_precomp.cpp
deleted file mode 100644
index 5956e13e3e..0000000000
--- a/modules/video/test/test_precomp.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "test_precomp.hpp"
diff --git a/modules/video/test/test_precomp.hpp b/modules/video/test/test_precomp.hpp
index 7025dff8f7..0aae1031f8 100644
--- a/modules/video/test/test_precomp.hpp
+++ b/modules/video/test/test_precomp.hpp
@@ -9,11 +9,10 @@
 #ifndef __OPENCV_TEST_PRECOMP_HPP__
 #define __OPENCV_TEST_PRECOMP_HPP__
 
+#include <iostream>
 #include "opencv2/ts.hpp"
 #include "opencv2/imgproc.hpp"
-#include "opencv2/imgproc/imgproc_c.h"
 #include "opencv2/video.hpp"
 #include "opencv2/highgui.hpp"
-#include <iostream>
 
 #endif
diff --git a/modules/videostab/CMakeLists.txt b/modules/videostab/CMakeLists.txt
index edde3f86dd..4df203c7dd 100644
--- a/modules/videostab/CMakeLists.txt
+++ b/modules/videostab/CMakeLists.txt
@@ -1,3 +1,3 @@
 set(the_description "Video stabilization")
-ocv_define_module(videostab opencv_imgproc opencv_features2d opencv_video opencv_photo opencv_calib3d OPTIONAL opencv_gpu opencv_highgui)
-
+ocv_define_module(videostab opencv_imgproc opencv_features2d opencv_video opencv_photo opencv_calib3d
+                  OPTIONAL opencv_gpu opencv_gpuwarping opencv_gpuoptflow opencv_highgui)
diff --git a/modules/videostab/include/opencv2/videostab/frame_source.hpp b/modules/videostab/include/opencv2/videostab/frame_source.hpp
index 1198231969..0bcc3fc681 100644
--- a/modules/videostab/include/opencv2/videostab/frame_source.hpp
+++ b/modules/videostab/include/opencv2/videostab/frame_source.hpp
@@ -44,7 +44,6 @@
 #define __OPENCV_VIDEOSTAB_FRAME_SOURCE_HPP__
 
 #include <vector>
-#include <string>
 #include "opencv2/core.hpp"
 
 namespace cv
@@ -70,7 +69,7 @@ public:
 class CV_EXPORTS VideoFileSource : public IFrameSource
 {
 public:
-    VideoFileSource(const std::string &path, bool volatileFrame = false);
+    VideoFileSource(const String &path, bool volatileFrame = false);
 
     virtual void reset();
     virtual Mat nextFrame();
diff --git a/modules/videostab/include/opencv2/videostab/global_motion.hpp b/modules/videostab/include/opencv2/videostab/global_motion.hpp
index 5e57081386..8ccc067a38 100644
--- a/modules/videostab/include/opencv2/videostab/global_motion.hpp
+++ b/modules/videostab/include/opencv2/videostab/global_motion.hpp
@@ -44,7 +44,6 @@
 #define __OPENCV_VIDEOSTAB_GLOBAL_MOTION_HPP__
 
 #include <vector>
-#include <string>
 #include <fstream>
 #include "opencv2/core.hpp"
 #include "opencv2/features2d.hpp"
@@ -53,8 +52,8 @@
 #include "opencv2/videostab/motion_core.hpp"
 #include "opencv2/videostab/outlier_rejection.hpp"
 
-#ifdef HAVE_OPENCV_GPU
-  #include "opencv2/gpu.hpp"
+#ifdef HAVE_OPENCV_GPUIMGPROC
+#  include "opencv2/gpuimgproc.hpp"
 #endif
 
 namespace cv
@@ -146,7 +145,7 @@ private:
 class CV_EXPORTS FromFileMotionReader : public ImageMotionEstimatorBase
 {
 public:
-    FromFileMotionReader(const std::string &path);
+    FromFileMotionReader(const String &path);
 
     virtual Mat estimate(const Mat &frame0, const Mat &frame1, bool *ok = 0);
 
@@ -157,7 +156,7 @@ private:
 class CV_EXPORTS ToFileMotionWriter : public ImageMotionEstimatorBase
 {
 public:
-    ToFileMotionWriter(const std::string &path, Ptr<ImageMotionEstimatorBase> estimator);
+    ToFileMotionWriter(const String &path, Ptr<ImageMotionEstimatorBase> estimator);
 
     virtual void setMotionModel(MotionModel val) { motionEstimator_->setMotionModel(val); }
     virtual MotionModel motionModel() const { return motionEstimator_->motionModel(); }
@@ -200,7 +199,8 @@ private:
     std::vector<Point2f> pointsPrevGood_, pointsGood_;
 };
 
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
+
 class CV_EXPORTS KeypointBasedMotionEstimatorGpu : public ImageMotionEstimatorBase
 {
 public:
@@ -217,7 +217,7 @@ public:
 
 private:
     Ptr<MotionEstimatorBase> motionEstimator_;
-    gpu::GoodFeaturesToTrackDetector_GPU detector_;
+    Ptr<gpu::CornersDetector> detector_;
     SparsePyrLkOptFlowEstimatorGpu optFlowEstimator_;
     Ptr<IOutlierRejector> outlierRejector_;
 
@@ -229,7 +229,8 @@ private:
     std::vector<Point2f> hostPointsPrevTmp_, hostPointsTmp_;
     std::vector<uchar> rejectionStatus_;
 };
-#endif
+
+#endif // defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
 
 CV_EXPORTS Mat getMotion(int from, int to, const std::vector<Mat> &motions);
 
diff --git a/modules/videostab/include/opencv2/videostab/optical_flow.hpp b/modules/videostab/include/opencv2/videostab/optical_flow.hpp
index 61f38fdb62..7509c1207a 100644
--- a/modules/videostab/include/opencv2/videostab/optical_flow.hpp
+++ b/modules/videostab/include/opencv2/videostab/optical_flow.hpp
@@ -46,8 +46,8 @@
 #include "opencv2/core.hpp"
 #include "opencv2/opencv_modules.hpp"
 
-#ifdef HAVE_OPENCV_GPU
-  #include "opencv2/gpu.hpp"
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+  #include "opencv2/gpuoptflow.hpp"
 #endif
 
 namespace cv
@@ -99,7 +99,8 @@ public:
             OutputArray status, OutputArray errors);
 };
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+
 class CV_EXPORTS SparsePyrLkOptFlowEstimatorGpu
         : public PyrLkOptFlowEstimatorBase, public ISparseOptFlowEstimator
 {
@@ -135,6 +136,7 @@ private:
     gpu::PyrLKOpticalFlow optFlowEstimator_;
     gpu::GpuMat frame0_, frame1_, flowX_, flowY_, errors_;
 };
+
 #endif
 
 } // namespace videostab
diff --git a/modules/videostab/include/opencv2/videostab/stabilizer.hpp b/modules/videostab/include/opencv2/videostab/stabilizer.hpp
index c13718f0b6..b021b47987 100644
--- a/modules/videostab/include/opencv2/videostab/stabilizer.hpp
+++ b/modules/videostab/include/opencv2/videostab/stabilizer.hpp
@@ -144,7 +144,7 @@ public:
     virtual void reset();
     virtual Mat nextFrame() { return nextStabilizedFrame(); }
 
-private:
+protected:
     virtual void setUp(const Mat &firstFrame);
     virtual Mat estimateMotion();
     virtual Mat estimateStabilizationMotion();
@@ -170,7 +170,7 @@ public:
     virtual void reset();
     virtual Mat nextFrame();
 
-private:
+protected:
     void runPrePassIfNecessary();
 
     virtual void setUp(const Mat &firstFrame);
diff --git a/modules/videostab/include/opencv2/videostab/wobble_suppression.hpp b/modules/videostab/include/opencv2/videostab/wobble_suppression.hpp
index f38de85049..01163b27c4 100644
--- a/modules/videostab/include/opencv2/videostab/wobble_suppression.hpp
+++ b/modules/videostab/include/opencv2/videostab/wobble_suppression.hpp
@@ -45,13 +45,10 @@
 
 #include <vector>
 #include "opencv2/core.hpp"
+#include "opencv2/core/gpu.hpp"
 #include "opencv2/videostab/global_motion.hpp"
 #include "opencv2/videostab/log.hpp"
 
-#ifdef HAVE_OPENCV_GPU
-  #include "opencv2/gpu.hpp"
-#endif
-
 namespace cv
 {
 namespace videostab
@@ -136,4 +133,3 @@ private:
 } // namespace cv
 
 #endif
-
diff --git a/modules/videostab/src/clp.hpp b/modules/videostab/src/clp.hpp
index 223016fe76..41b875b3c0 100644
--- a/modules/videostab/src/clp.hpp
+++ b/modules/videostab/src/clp.hpp
@@ -44,14 +44,6 @@
 #define __OPENCV_VIDEOSTAB_CLP_HPP__
 
 #ifdef HAVE_CLP
-#  undef PACKAGE
-#  undef PACKAGE_BUGREPORT
-#  undef PACKAGE_NAME
-#  undef PACKAGE_STRING
-#  undef PACKAGE_TARNAME
-#  undef PACKAGE_VERSION
-#  undef VERSION
-
 #  define COIN_BIG_INDEX 0
 #  define DEBUG_COIN 0
 #  define PRESOLVE_DEBUG 0
diff --git a/modules/videostab/src/frame_source.cpp b/modules/videostab/src/frame_source.cpp
index b1dc9723cd..7ca4b73de7 100644
--- a/modules/videostab/src/frame_source.cpp
+++ b/modules/videostab/src/frame_source.cpp
@@ -59,7 +59,7 @@ namespace {
 class VideoFileSourceImpl : public IFrameSource
 {
 public:
-    VideoFileSourceImpl(const std::string &path, bool volatileFrame)
+    VideoFileSourceImpl(const String &path, bool volatileFrame)
         : path_(path), volatileFrame_(volatileFrame) { reset(); }
 
     virtual void reset()
@@ -68,7 +68,7 @@ public:
         vc.release();
         vc.open(path_);
         if (!vc.isOpened())
-            throw std::runtime_error("can't open file: " + path_);
+            CV_Error(0, "can't open file: " + path_);
 #else
         CV_Error(CV_StsNotImplemented, "OpenCV has been compiled without video I/O support");
 #endif
@@ -84,10 +84,10 @@ public:
     }
 
 #ifdef HAVE_OPENCV_HIGHGUI
-    int width() {return static_cast<int>(vc.get(CV_CAP_PROP_FRAME_WIDTH));}
-    int height() {return static_cast<int>(vc.get(CV_CAP_PROP_FRAME_HEIGHT));}
-    int count() {return static_cast<int>(vc.get(CV_CAP_PROP_FRAME_COUNT));}
-    double fps() {return vc.get(CV_CAP_PROP_FPS);}
+    int width() {return static_cast<int>(vc.get(CAP_PROP_FRAME_WIDTH));}
+    int height() {return static_cast<int>(vc.get(CAP_PROP_FRAME_HEIGHT));}
+    int count() {return static_cast<int>(vc.get(CAP_PROP_FRAME_COUNT));}
+    double fps() {return vc.get(CAP_PROP_FPS);}
 #else
     int width() {return 0;}
     int height() {return 0;}
@@ -96,7 +96,7 @@ public:
 #endif
 
 private:
-    std::string path_;
+    String path_;
     bool volatileFrame_;
 #ifdef HAVE_OPENCV_HIGHGUI
     VideoCapture vc;
@@ -105,16 +105,16 @@ private:
 
 }//namespace
 
-VideoFileSource::VideoFileSource(const std::string &path, bool volatileFrame)
+VideoFileSource::VideoFileSource(const String &path, bool volatileFrame)
     : impl(new VideoFileSourceImpl(path, volatileFrame)) {}
 
 void VideoFileSource::reset() { impl->reset(); }
 Mat VideoFileSource::nextFrame() { return impl->nextFrame(); }
 
-int VideoFileSource::width() { return ((VideoFileSourceImpl*)impl.obj)->width(); }
-int VideoFileSource::height() { return ((VideoFileSourceImpl*)impl.obj)->height(); }
-int VideoFileSource::count() { return ((VideoFileSourceImpl*)impl.obj)->count(); }
-double VideoFileSource::fps() { return ((VideoFileSourceImpl*)impl.obj)->fps(); }
+int VideoFileSource::width() { return ((VideoFileSourceImpl*)impl.get())->width(); }
+int VideoFileSource::height() { return ((VideoFileSourceImpl*)impl.get())->height(); }
+int VideoFileSource::count() { return ((VideoFileSourceImpl*)impl.get())->count(); }
+double VideoFileSource::fps() { return ((VideoFileSourceImpl*)impl.get())->fps(); }
 
 } // namespace videostab
 } // namespace cv
diff --git a/modules/videostab/src/global_motion.cpp b/modules/videostab/src/global_motion.cpp
index b8927c856b..c9f15a61cf 100644
--- a/modules/videostab/src/global_motion.cpp
+++ b/modules/videostab/src/global_motion.cpp
@@ -47,6 +47,10 @@
 #include "opencv2/opencv_modules.hpp"
 #include "clp.hpp"
 
+#ifdef HAVE_OPENCV_GPU
+#  include "opencv2/gpu.hpp"
+#endif
+
 namespace cv
 {
 namespace videostab
@@ -356,6 +360,9 @@ Mat estimateGlobalMotionRansac(
     const int npoints = points0.getMat().checkVector(2);
     CV_Assert(points1.getMat().checkVector(2) == npoints);
 
+    if (npoints < params.size)
+        return Mat::eye(3, 3, CV_32F);
+
     const Point2f *points0_ = points0.getMat().ptr<Point2f>();
     const Point2f *points1_ = points1.getMat().ptr<Point2f>();
     const int niters = params.niters();
@@ -471,7 +478,7 @@ Mat MotionEstimatorRansacL2::estimate(InputArray points0, InputArray points1, bo
     else
     {
         std::vector<uchar> mask;
-        M = findHomography(points0, points1, mask, CV_LMEDS);
+        M = findHomography(points0, points1, mask, LMEDS);
         for (int i  = 0; i < npoints; ++i)
             if (mask[i]) ninliers++;
     }
@@ -504,7 +511,7 @@ Mat MotionEstimatorL1::estimate(InputArray points0, InputArray points1, bool *ok
 
 #ifndef HAVE_CLP
 
-    CV_Error(CV_StsError, "The library is built without Clp support");
+    CV_Error(Error::StsError, "The library is built without Clp support");
     if (ok) *ok = false;
     return Mat::eye(3, 3, CV_32F);
 
@@ -621,7 +628,7 @@ Mat MotionEstimatorL1::estimate(InputArray points0, InputArray points1, bool *ok
 }
 
 
-FromFileMotionReader::FromFileMotionReader(const std::string &path)
+FromFileMotionReader::FromFileMotionReader(const String &path)
     : ImageMotionEstimatorBase(MM_UNKNOWN)
 {
     file_.open(path.c_str());
@@ -641,7 +648,7 @@ Mat FromFileMotionReader::estimate(const Mat &/*frame0*/, const Mat &/*frame1*/,
 }
 
 
-ToFileMotionWriter::ToFileMotionWriter(const std::string &path, Ptr<ImageMotionEstimatorBase> estimator)
+ToFileMotionWriter::ToFileMotionWriter(const String &path, Ptr<ImageMotionEstimatorBase> estimator)
     : ImageMotionEstimatorBase(estimator->motionModel()), motionEstimator_(estimator)
 {
     file_.open(path.c_str());
@@ -664,9 +671,9 @@ Mat ToFileMotionWriter::estimate(const Mat &frame0, const Mat &frame1, bool *ok)
 KeypointBasedMotionEstimator::KeypointBasedMotionEstimator(Ptr<MotionEstimatorBase> estimator)
     : ImageMotionEstimatorBase(estimator->motionModel()), motionEstimator_(estimator)
 {
-    setDetector(new GoodFeaturesToTrackDetector());
-    setOpticalFlowEstimator(new SparsePyrLkOptFlowEstimator());
-    setOutlierRejector(new NullOutlierRejector());
+    setDetector(makePtr<GoodFeaturesToTrackDetector>());
+    setOpticalFlowEstimator(makePtr<SparsePyrLkOptFlowEstimator>());
+    setOutlierRejector(makePtr<NullOutlierRejector>());
 }
 
 
@@ -674,6 +681,8 @@ Mat KeypointBasedMotionEstimator::estimate(const Mat &frame0, const Mat &frame1,
 {
     // find keypoints
     detector_->detect(frame0, keypointsPrev_);
+    if (keypointsPrev_.empty())
+        return Mat::eye(3, 3, CV_32F);
 
     // extract points from keypoints
     pointsPrev_.resize(keypointsPrev_.size());
@@ -699,7 +708,7 @@ Mat KeypointBasedMotionEstimator::estimate(const Mat &frame0, const Mat &frame1,
 
     // perform outlier rejection
 
-    IOutlierRejector *outlRejector = static_cast<IOutlierRejector*>(outlierRejector_);
+    IOutlierRejector *outlRejector = outlierRejector_.get();
     if (!dynamic_cast<NullOutlierRejector*>(outlRejector))
     {
         pointsPrev_.swap(pointsPrevGood_);
@@ -728,12 +737,15 @@ Mat KeypointBasedMotionEstimator::estimate(const Mat &frame0, const Mat &frame1,
 }
 
 
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
+
 KeypointBasedMotionEstimatorGpu::KeypointBasedMotionEstimatorGpu(Ptr<MotionEstimatorBase> estimator)
     : ImageMotionEstimatorBase(estimator->motionModel()), motionEstimator_(estimator)
 {
+    detector_ = gpu::createGoodFeaturesToTrackDetector(CV_8UC1);
+
     CV_Assert(gpu::getCudaEnabledDeviceCount() > 0);
-    setOutlierRejector(new NullOutlierRejector());
+    setOutlierRejector(makePtr<NullOutlierRejector>());
 }
 
 
@@ -754,12 +766,12 @@ Mat KeypointBasedMotionEstimatorGpu::estimate(const gpu::GpuMat &frame0, const g
         grayFrame0 = frame0;
     else
     {
-        gpu::cvtColor(frame0, grayFrame0_, CV_BGR2GRAY);
+        gpu::cvtColor(frame0, grayFrame0_, COLOR_BGR2GRAY);
         grayFrame0 = grayFrame0_;
     }
 
     // find keypoints
-    detector_(grayFrame0, pointsPrev_);
+    detector_->detect(grayFrame0, pointsPrev_);
 
     // find correspondences
     optFlowEstimator_.run(frame0, frame1, pointsPrev_, points_, status_);
@@ -772,7 +784,7 @@ Mat KeypointBasedMotionEstimatorGpu::estimate(const gpu::GpuMat &frame0, const g
 
     // perform outlier rejection
 
-    IOutlierRejector *rejector = static_cast<IOutlierRejector*>(outlierRejector_);
+    IOutlierRejector *rejector = outlierRejector_.get();
     if (!dynamic_cast<NullOutlierRejector*>(rejector))
     {
         outlierRejector_->process(frame0.size(), hostPointsPrev_, hostPoints_, rejectionStatus_);
@@ -799,7 +811,8 @@ Mat KeypointBasedMotionEstimatorGpu::estimate(const gpu::GpuMat &frame0, const g
     // estimate motion
     return motionEstimator_->estimate(hostPointsPrev_, hostPoints_, ok);
 }
-#endif // HAVE_OPENCV_GPU
+
+#endif // defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
 
 
 Mat getMotion(int from, int to, const std::vector<Mat> &motions)
@@ -821,5 +834,3 @@ Mat getMotion(int from, int to, const std::vector<Mat> &motions)
 
 } // namespace videostab
 } // namespace cv
-
-
diff --git a/modules/videostab/src/inpainting.cpp b/modules/videostab/src/inpainting.cpp
index e24281b5e5..7f5368cffc 100644
--- a/modules/videostab/src/inpainting.cpp
+++ b/modules/videostab/src/inpainting.cpp
@@ -323,10 +323,10 @@ public:
 
 MotionInpainter::MotionInpainter()
 {
-#ifdef HAVE_OPENCV_GPU
-    setOptFlowEstimator(new DensePyrLkOptFlowEstimatorGpu());
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+    setOptFlowEstimator(makePtr<DensePyrLkOptFlowEstimatorGpu>());
 #else
-    CV_Error(CV_StsNotImplemented, "Current implementation of MotionInpainter requires GPU");
+    CV_Error(Error::StsNotImplemented, "Current implementation of MotionInpainter requires GPU");
 #endif
     setFlowErrorThreshold(1e-4f);
     setDistThreshold(5.f);
@@ -357,7 +357,7 @@ void MotionInpainter::inpaint(int idx, Mat &frame, Mat &mask)
         mask1_.setTo(255);
     }
 
-    cvtColor(frame, grayFrame_, CV_BGR2GRAY);
+    cvtColor(frame, grayFrame_, COLOR_BGR2GRAY);
 
     MotionInpaintBody body;
     body.rad = 2;
@@ -383,7 +383,7 @@ void MotionInpainter::inpaint(int idx, Mat &frame, Mat &mask)
                     frame1_, transformedFrame1_, motion1to0, frame1_.size(), INTER_LINEAR,
                     borderMode_);
 
-        cvtColor(transformedFrame1_, transformedGrayFrame1_, CV_BGR2GRAY);
+        cvtColor(transformedFrame1_, transformedGrayFrame1_, COLOR_BGR2GRAY);
 
         // warp mask
 
diff --git a/modules/videostab/src/motion_stabilizing.cpp b/modules/videostab/src/motion_stabilizing.cpp
index 6e9ef7ea34..65bbd73bd1 100644
--- a/modules/videostab/src/motion_stabilizing.cpp
+++ b/modules/videostab/src/motion_stabilizing.cpp
@@ -134,7 +134,7 @@ LpMotionStabilizer::LpMotionStabilizer(MotionModel model)
 
 void LpMotionStabilizer::stabilize(int, const std::vector<Mat>&, std::pair<int,int>, Mat*)
 {
-    CV_Error(CV_StsError, "The library is built without Clp support");
+    CV_Error(Error::StsError, "The library is built without Clp support");
 }
 
 #else
@@ -532,9 +532,9 @@ void LpMotionStabilizer::stabilize(
     model.scaling(1);
 
     ClpPresolve presolveInfo;
-    Ptr<ClpSimplex> presolvedModel = presolveInfo.presolvedModel(model);
+    Ptr<ClpSimplex> presolvedModel(presolveInfo.presolvedModel(model));
 
-    if (!presolvedModel.empty())
+    if (presolvedModel)
     {
         presolvedModel->dual();
         presolveInfo.postsolve(true);
diff --git a/modules/videostab/src/optical_flow.cpp b/modules/videostab/src/optical_flow.cpp
index f5c8288e7e..cee08823a1 100644
--- a/modules/videostab/src/optical_flow.cpp
+++ b/modules/videostab/src/optical_flow.cpp
@@ -58,7 +58,8 @@ void SparsePyrLkOptFlowEstimator::run(
 }
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUOPTFLOW
+
 SparsePyrLkOptFlowEstimatorGpu::SparsePyrLkOptFlowEstimatorGpu()
 {
     CV_Assert(gpu::getCudaEnabledDeviceCount() > 0);
@@ -133,7 +134,8 @@ void DensePyrLkOptFlowEstimatorGpu::run(
     flowX_.download(flowX.getMatRef());
     flowY_.download(flowY.getMatRef());
 }
-#endif // HAVE_OPENCV_GPU
+
+#endif // HAVE_OPENCV_GPUOPTFLOW
 
 } // namespace videostab
 } // namespace cv
diff --git a/modules/videostab/src/precomp.hpp b/modules/videostab/src/precomp.hpp
index bf2d2114ac..aa6026deee 100644
--- a/modules/videostab/src/precomp.hpp
+++ b/modules/videostab/src/precomp.hpp
@@ -43,10 +43,6 @@
 #ifndef __OPENCV_PRECOMP_HPP__
 #define __OPENCV_PRECOMP_HPP__
 
-#ifdef HAVE_CVCONFIG_H
-  #include "cvconfig.h"
-#endif
-
 #include <stdexcept>
 #include <iostream>
 #include <ctime>
@@ -57,6 +53,8 @@
 #include "opencv2/features2d.hpp"
 #include "opencv2/calib3d.hpp"
 
+#include "opencv2/core/private.hpp"
+
 // some aux. functions
 
 inline float sqr(float x) { return x * x; }
@@ -67,4 +65,3 @@ inline float intensity(const cv::Point3_<uchar> &bgr)
 }
 
 #endif
-
diff --git a/modules/videostab/src/stabilizer.cpp b/modules/videostab/src/stabilizer.cpp
index 50ac05c6f7..f9c09ba36c 100644
--- a/modules/videostab/src/stabilizer.cpp
+++ b/modules/videostab/src/stabilizer.cpp
@@ -54,11 +54,11 @@ namespace videostab
 
 StabilizerBase::StabilizerBase()
 {
-    setLog(new LogToStdout());
-    setFrameSource(new NullFrameSource());
-    setMotionEstimator(new KeypointBasedMotionEstimator(new MotionEstimatorRansacL2()));
-    setDeblurer(new NullDeblurer());
-    setInpainter(new NullInpainter());
+    setLog(makePtr<LogToStdout>());
+    setFrameSource(makePtr<NullFrameSource>());
+    setMotionEstimator(makePtr<KeypointBasedMotionEstimator>(makePtr<MotionEstimatorRansacL2>()));
+    setDeblurer(makePtr<NullDeblurer>());
+    setInpainter(makePtr<NullInpainter>());
     setRadius(15);
     setTrimRatio(0);
     setCorrectionForInclusion(false);
@@ -156,7 +156,7 @@ bool StabilizerBase::doOneIteration()
 
 void StabilizerBase::setUp(const Mat &firstFrame)
 {
-    InpainterBase *inpaint = static_cast<InpainterBase*>(inpainter_);
+    InpainterBase *inpaint = inpainter_.get();
     doInpainting_ = dynamic_cast<NullInpainter*>(inpaint) == 0;
     if (doInpainting_)
     {
@@ -167,7 +167,7 @@ void StabilizerBase::setUp(const Mat &firstFrame)
         inpainter_->setStabilizationMotions(stabilizationMotions_);
     }
 
-    DeblurerBase *deblurer = static_cast<DeblurerBase*>(deblurer_);
+    DeblurerBase *deblurer = deblurer_.get();
     doDeblurring_ = dynamic_cast<NullDeblurer*>(deblurer) == 0;
     if (doDeblurring_)
     {
@@ -252,7 +252,7 @@ void StabilizerBase::logProcessingTime()
 
 OnePassStabilizer::OnePassStabilizer()
 {
-    setMotionFilter(new GaussianMotionFilter());
+    setMotionFilter(makePtr<GaussianMotionFilter>());
     reset();
 }
 
@@ -308,8 +308,8 @@ Mat OnePassStabilizer::postProcessFrame(const Mat &frame)
 
 TwoPassStabilizer::TwoPassStabilizer()
 {
-    setMotionStabilizer(new GaussianMotionFilter());
-    setWobbleSuppressor(new NullWobbleSuppressor());
+    setMotionStabilizer(makePtr<GaussianMotionFilter>());
+    setWobbleSuppressor(makePtr<NullWobbleSuppressor>());
     setEstimateTrimRatio(false);
     reset();
 }
@@ -371,7 +371,7 @@ void TwoPassStabilizer::runPrePassIfNecessary()
     {
         // check if we must do wobble suppression
 
-        WobbleSuppressorBase *wobble = static_cast<WobbleSuppressorBase*>(wobbleSuppressor_);
+        WobbleSuppressorBase *wobble = wobbleSuppressor_.get();
         doWobbleSuppression_ = dynamic_cast<NullWobbleSuppressor*>(wobble) == 0;
 
         // estimate motions
@@ -469,7 +469,7 @@ void TwoPassStabilizer::setUp(const Mat &firstFrame)
     for (int i = -radius_; i <= 0; ++i)
         at(i, frames_) = firstFrame;
 
-    WobbleSuppressorBase *wobble = static_cast<WobbleSuppressorBase*>(wobbleSuppressor_);
+    WobbleSuppressorBase *wobble = wobbleSuppressor_.get();
     doWobbleSuppression_ = dynamic_cast<NullWobbleSuppressor*>(wobble) == 0;
     if (doWobbleSuppression_)
     {
diff --git a/modules/videostab/src/wobble_suppression.cpp b/modules/videostab/src/wobble_suppression.cpp
index 3c48df58fd..0b652b9dbc 100644
--- a/modules/videostab/src/wobble_suppression.cpp
+++ b/modules/videostab/src/wobble_suppression.cpp
@@ -44,6 +44,15 @@
 #include "opencv2/videostab/wobble_suppression.hpp"
 #include "opencv2/videostab/ring_buffer.hpp"
 
+#ifdef HAVE_OPENCV_GPUWARPING
+#  include "opencv2/gpuwarping.hpp"
+#endif
+
+#ifdef HAVE_OPENCV_GPU
+#  include "opencv2/gpu.hpp"
+#endif
+
+
 namespace cv
 {
 namespace videostab
@@ -51,7 +60,7 @@ namespace videostab
 
 WobbleSuppressorBase::WobbleSuppressorBase() : motions_(0), stabilizationMotions_(0)
 {
-    setMotionEstimator(new KeypointBasedMotionEstimator(new MotionEstimatorRansacL2(MM_HOMOGRAPHY)));
+    setMotionEstimator(makePtr<KeypointBasedMotionEstimator>(makePtr<MotionEstimatorRansacL2>(MM_HOMOGRAPHY)));
 }
 
 
@@ -113,7 +122,7 @@ void MoreAccurateMotionWobbleSuppressor::suppress(int idx, const Mat &frame, Mat
 }
 
 
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUWARPING
 void MoreAccurateMotionWobbleSuppressorGpu::suppress(int idx, const gpu::GpuMat &frame, gpu::GpuMat &result)
 {
     CV_Assert(motions_ && stabilizationMotions_);
@@ -151,4 +160,3 @@ void MoreAccurateMotionWobbleSuppressorGpu::suppress(int idx, const Mat &frame,
 
 } // namespace videostab
 } // namespace cv
-
diff --git a/modules/viz/include/opencv2/viz/types.hpp b/modules/viz/include/opencv2/viz/types.hpp
index c753945de8..60200eb660 100644
--- a/modules/viz/include/opencv2/viz/types.hpp
+++ b/modules/viz/include/opencv2/viz/types.hpp
@@ -55,8 +55,6 @@
 
 namespace cv
 {
-    typedef std::string String;
-
     namespace viz
     {
         class CV_EXPORTS Color : public Scalar
@@ -104,7 +102,7 @@ namespace cv
 
             //! Create a keyboard event
             //! - Note that action is true if key is pressed, false if released
-            KeyboardEvent(bool action, const std::string& key_sym, unsigned char key, bool alt, bool ctrl, bool shift);
+            KeyboardEvent(bool action, const String& key_sym, unsigned char key, bool alt, bool ctrl, bool shift);
 
             bool isAltPressed() const;
             bool isCtrlPressed() const;
diff --git a/modules/viz/src/interactor_style.cpp b/modules/viz/src/interactor_style.cpp
index ad446f4c0e..312f686163 100644
--- a/modules/viz/src/interactor_style.cpp
+++ b/modules/viz/src/interactor_style.cpp
@@ -321,10 +321,18 @@ cv::viz::InteractorStyle::OnKeyDown()
         Vec2i win_pos(Interactor->GetRenderWindow()->GetPosition());
         Vec2i win_size(Interactor->GetRenderWindow()->GetSize());
 
-        std::cerr << Mat(clip, false).reshape(1, 1) << "/" << Mat(focal, false).reshape(1, 1) << "/" <<
-                     Mat(pose, false).reshape(1, 1) << "/" << Mat(view, false).reshape(1, 1) << "/" <<
-                     cam->GetViewAngle () / 180.0 * CV_PI  << "/" <<
-                     Mat(win_size, false).reshape(1,1) << "/" << Mat(win_pos, false).reshape(1,1) << endl;
+        cv::print(Mat(clip, false).reshape(1, 1));
+        std::cout << "/";
+        cv::print(Mat(focal, false).reshape(1, 1));
+        std::cout << "/";
+        cv::print(Mat(pose, false).reshape(1, 1));
+        std::cout << "/";
+        cv::print(Mat(view, false).reshape(1, 1));
+        std::cout << "/" << cam->GetViewAngle () / 180.0 * CV_PI;
+        cv::print(Mat(win_size, false).reshape(1, 1));
+        std::cout << "/";
+        cv::print(Mat(win_pos, false).reshape(1, 1));
+        std::cout << std::endl;
         break;
     }
     case '=':
diff --git a/modules/viz/src/precomp.hpp b/modules/viz/src/precomp.hpp
index d12011dfdb..808aa6eb15 100644
--- a/modules/viz/src/precomp.hpp
+++ b/modules/viz/src/precomp.hpp
@@ -125,11 +125,14 @@
 #undef __DEPRECATED_DISABLED__
 #endif
 
+#include <opencv2/ts.hpp>
+#include <opencv2/core.hpp>
+
 namespace cv
 {
     namespace viz
     {
-        typedef std::map<std::string, vtkSmartPointer<vtkProp> > WidgetActorMap;
+        typedef std::map<String, vtkSmartPointer<vtkProp> > WidgetActorMap;
     }
 }
 
@@ -139,12 +142,11 @@ namespace cv
 {
     namespace viz
     {
-        typedef std::map<std::string, Viz3d> VizMap;
-        typedef std::pair<std::string, Viz3d> VizPair;
+        typedef std::map<String, Viz3d> VizMap;
+        typedef std::pair<String, Viz3d> VizPair;
     }
 }
 
-#include <opencv2/core.hpp>
 #include <opencv2/viz.hpp>
 #include <opencv2/viz/types.hpp>
 #include "opencv2/viz/widget_accessor.hpp"
diff --git a/modules/viz/src/types.cpp b/modules/viz/src/types.cpp
index 49f6ccee43..31a4699333 100644
--- a/modules/viz/src/types.cpp
+++ b/modules/viz/src/types.cpp
@@ -71,7 +71,7 @@ cv::viz::Color cv::viz::Color::gray()    { return Color(128, 128, 128); }
 ////////////////////////////////////////////////////////////////////
 /// cv::viz::KeyboardEvent
 
-cv::viz::KeyboardEvent::KeyboardEvent(bool _action, const std::string& _key_sym, unsigned char key, bool alt, bool ctrl, bool shift)
+cv::viz::KeyboardEvent::KeyboardEvent(bool _action, const String& _key_sym, unsigned char key, bool alt, bool ctrl, bool shift)
   : action_(_action), modifiers_(0), key_code_(key), key_sym_(_key_sym)
 {
   if (alt)
@@ -223,11 +223,13 @@ cv::viz::Camera::Camera(const Matx44f &proj, const Size &window_size)
     double bottom = near * (proj(1,2)-1) / proj(1,1);
     double top = 2.0 * near / proj(1,1) + bottom;
     
-    if (fabs(left-right) < std::numeric_limits<double>::epsilon()) principal_point_[0] = static_cast<float>(window_size.width) * 0.5f;
+    double epsilon = 2.2204460492503131e-16;
+    
+    if (fabs(left-right) < epsilon) principal_point_[0] = static_cast<float>(window_size.width) * 0.5f;
     else principal_point_[0] = (left * static_cast<float>(window_size.width)) / (left - right); 
     focal_[0] = -near * principal_point_[0] / left;
     
-    if (fabs(top-bottom) < std::numeric_limits<double>::epsilon()) principal_point_[1] = static_cast<float>(window_size.height) * 0.5f; 
+    if (fabs(top-bottom) < epsilon) principal_point_[1] = static_cast<float>(window_size.height) * 0.5f; 
     else principal_point_[1] = (top * static_cast<float>(window_size.height)) / (top - bottom); 
     focal_[1] = near * principal_point_[1] / top;
     
diff --git a/modules/viz/src/viz3d_impl.cpp b/modules/viz/src/viz3d_impl.cpp
index 5c91ec72ef..456b3d62ca 100644
--- a/modules/viz/src/viz3d_impl.cpp
+++ b/modules/viz/src/viz3d_impl.cpp
@@ -59,7 +59,7 @@ vtkRenderWindowInteractor* vtkRenderWindowInteractorFixNew()
 #endif
 
 /////////////////////////////////////////////////////////////////////////////////////////////
-cv::viz::Viz3d::VizImpl::VizImpl(const std::string &name)
+cv::viz::Viz3d::VizImpl::VizImpl(const String &name)
     :  style_(vtkSmartPointer<cv::viz::InteractorStyle>::New())
     , widget_actor_map_(new WidgetActorMap)
     , s_lastDone_(0.0)
@@ -252,7 +252,7 @@ double cv::viz::Viz3d::VizImpl::getDesiredUpdateRate()
 }
 
 /////////////////////////////////////////////////////////////////////////////////////////////
-void cv::viz::Viz3d::VizImpl::saveScreenshot(const std::string &file) { style_->saveScreenshot(file); }
+void cv::viz::Viz3d::VizImpl::saveScreenshot(const String &file) { style_->saveScreenshot(file.c_str()); }
 
 /////////////////////////////////////////////////////////////////////////////////////////////
 void cv::viz::Viz3d::VizImpl::registerMouseCallback(MouseCallback callback, void* cookie)
diff --git a/modules/world/CMakeLists.txt b/modules/world/CMakeLists.txt
index a75037cfc7..f18635e620 100644
--- a/modules/world/CMakeLists.txt
+++ b/modules/world/CMakeLists.txt
@@ -80,6 +80,35 @@ foreach(m ${OPENCV_MODULE_${the_module}_DEPS})
   endforeach()
 endforeach()
 
+
+macro(ios_include_3party_libs)
+  foreach(l ${ARGN})
+    add_dependencies(${the_module} ${l})
+    string(REGEX REPLACE "<MODULE_NAME>" "${l}" objpath1 "${CMAKE_BINARY_DIR}/3rdparty/${l}/${objpath0}")
+    file(GLOB sources ${CMAKE_SOURCE_DIR}/3rdparty/${l}/*.c)
+    foreach(srcname ${sources})
+      if(IS_ABSOLUTE "${srcname}")
+        file(RELATIVE_PATH srcname "${CMAKE_SOURCE_DIR}/3rdparty/${l}" "${srcname}")
+      endif()
+
+      string(REPLACE ".." "__" srcname "${srcname}")
+      get_filename_component(srcname_we ${srcname} NAME_WE)
+      string(REGEX REPLACE <SRC_NAME_WE> "${srcname_we}" objpath2 "${objpath1}")
+      string(REGEX REPLACE <RELATIVE_SRC_NAME> "${srcname}" objpath3 "${objpath2}")
+
+      list(APPEND objlist "\"${objpath3}\"")
+    endforeach() # (srcname ${sources})
+  endforeach()
+endmacro()
+
+if(IOS AND WITH_PNG)
+  ios_include_3party_libs(zlib libpng)
+endif()
+
+if(IOS AND WITH_JPEG)
+  ios_include_3party_libs(libjpeg)
+endif()
+
 string(REPLACE ";" " " objlist "${objlist}")
 
 if(have_cfg)
diff --git a/modules/world/src/precomp.cpp b/modules/world/src/precomp.cpp
deleted file mode 100644
index e69de29bb2..0000000000
diff --git a/modules/world/src/precomp.hpp b/modules/world/src/precomp.hpp
index 343b875979..1aa056848f 100644
--- a/modules/world/src/precomp.hpp
+++ b/modules/world/src/precomp.hpp
@@ -43,10 +43,6 @@
 #ifndef __OPENCV_PRECOMP_H__
 #define __OPENCV_PRECOMP_H__
 
-#ifdef HAVE_CVCONFIG_H
-#include "cvconfig.h"
-#endif
-
 #include "opencv2/opencv_modules.hpp"
 #ifdef HAVE_OPENCV_VIDEO
 #include "opencv2/video.hpp"
diff --git a/android/README.android b/platforms/android/README.android
similarity index 100%
rename from android/README.android
rename to platforms/android/README.android
diff --git a/android/android.toolchain.cmake b/platforms/android/android.toolchain.cmake
similarity index 81%
rename from android/android.toolchain.cmake
rename to platforms/android/android.toolchain.cmake
index f5daf307f7..d7f09c7888 100644
--- a/android/android.toolchain.cmake
+++ b/platforms/android/android.toolchain.cmake
@@ -1,5 +1,5 @@
 # Copyright (c) 2010-2011, Ethan Rublee
-# Copyright (c) 2011-2012, Andrey Kamaev
+# Copyright (c) 2011-2013, Andrey Kamaev
 # All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
@@ -281,8 +281,17 @@
 #     [+] updated for NDK r8c
 #     [+] added support for clang compiler
 #   - December 2012
+#     [+] suppress warning about unused CMAKE_TOOLCHAIN_FILE variable
+#     [+] adjust API level to closest compatible as NDK does
 #     [~] fixed ccache full path search
 #     [+] updated for NDK r8d
+#     [~] compiler options are aligned with NDK r8d
+#   - March 2013
+#     [+] updated for NDK r8e (x86 version)
+#     [+] support x86_64 version of NDK
+#   - April 2013
+#     [+] support non-release NDK layouts (from Linaro git and Android git)
+#     [~] automatically detect if explicit link to crtbegin_*.o is needed
 # ------------------------------------------------------------------------------
 
 cmake_minimum_required( VERSION 2.6.3 )
@@ -292,6 +301,10 @@ if( DEFINED CMAKE_CROSSCOMPILING )
  return()
 endif()
 
+if( CMAKE_TOOLCHAIN_FILE )
+ # touch toolchain variable only to suppress "unused variable" warning
+endif()
+
 get_property( _CMAKE_IN_TRY_COMPILE GLOBAL PROPERTY IN_TRY_COMPILE )
 if( _CMAKE_IN_TRY_COMPILE )
  include( "${CMAKE_CURRENT_SOURCE_DIR}/../android.toolchain.config.cmake" OPTIONAL )
@@ -305,7 +318,7 @@ set( CMAKE_SYSTEM_VERSION 1 )
 # rpath makes low sence for Android
 set( CMAKE_SKIP_RPATH TRUE CACHE BOOL "If set, runtime paths are not added when using shared libraries." )
 
-set( ANDROID_SUPPORTED_NDK_VERSIONS ${ANDROID_EXTRA_NDK_VERSIONS} -r8d -r8c -r8b -r8 -r7c -r7b -r7 -r6b -r6 -r5c -r5b -r5 "" )
+set( ANDROID_SUPPORTED_NDK_VERSIONS ${ANDROID_EXTRA_NDK_VERSIONS} -r8e -r8d -r8c -r8b -r8 -r7c -r7b -r7 -r6b -r6 -r5c -r5b -r5 "" )
 if(NOT DEFINED ANDROID_NDK_SEARCH_PATHS)
  if( CMAKE_HOST_WIN32 )
   file( TO_CMAKE_PATH "$ENV{PROGRAMFILES}" ANDROID_NDK_SEARCH_PATHS )
@@ -449,19 +462,32 @@ if( ANDROID_FORBID_SYGWIN )
  endif()
 endif()
 
+
 # detect current host platform
+if( NOT DEFINED ANDROID_NDK_HOST_X64 AND CMAKE_HOST_SYSTEM_PROCESSOR MATCHES "amd64|x86_64|AMD64")
+ set( ANDROID_NDK_HOST_X64 1 CACHE BOOL "Try to use 64-bit compiler toolchain" )
+ mark_as_advanced( ANDROID_NDK_HOST_X64 )
+endif()
+
 set( TOOL_OS_SUFFIX "" )
 if( CMAKE_HOST_APPLE )
- set( ANDROID_NDK_HOST_SYSTEM_NAME "darwin-x86" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME "darwin-x86_64" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME2 "darwin-x86" )
 elseif( CMAKE_HOST_WIN32 )
- set( ANDROID_NDK_HOST_SYSTEM_NAME "windows" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME "windows-x86_64" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME2 "windows" )
  set( TOOL_OS_SUFFIX ".exe" )
 elseif( CMAKE_HOST_UNIX )
- set( ANDROID_NDK_HOST_SYSTEM_NAME "linux-x86" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME "linux-x86_64" )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME2 "linux-x86" )
 else()
  message( FATAL_ERROR "Cross-compilation on your platform is not supported by this cmake toolchain" )
 endif()
 
+if( NOT ANDROID_NDK_HOST_X64 )
+ set( ANDROID_NDK_HOST_SYSTEM_NAME ${ANDROID_NDK_HOST_SYSTEM_NAME2} )
+endif()
+
 # see if we have path to Android NDK
 __INIT_VARIABLE( ANDROID_NDK PATH ENV_ANDROID_NDK )
 if( NOT ANDROID_NDK )
@@ -493,23 +519,19 @@ if( NOT ANDROID_NDK )
   endif( ANDROID_NDK )
  endif( NOT ANDROID_STANDALONE_TOOLCHAIN )
 endif( NOT ANDROID_NDK )
+
 # remember found paths
 if( ANDROID_NDK )
  get_filename_component( ANDROID_NDK "${ANDROID_NDK}" ABSOLUTE )
- # try to detect change
- if( CMAKE_AR )
-  string( LENGTH "${ANDROID_NDK}" __length )
-  string( SUBSTRING "${CMAKE_AR}" 0 ${__length} __androidNdkPreviousPath )
-  if( NOT __androidNdkPreviousPath STREQUAL ANDROID_NDK )
-   message( FATAL_ERROR "It is not possible to change the path to the NDK on subsequent CMake run. You must remove all generated files from your build folder first.
-   " )
-  endif()
-  unset( __androidNdkPreviousPath )
-  unset( __length )
- endif()
  set( ANDROID_NDK "${ANDROID_NDK}" CACHE INTERNAL "Path of the Android NDK" FORCE )
  set( BUILD_WITH_ANDROID_NDK True )
- file( STRINGS "${ANDROID_NDK}/RELEASE.TXT" ANDROID_NDK_RELEASE LIMIT_COUNT 1 REGEX r[0-9]+[a-z]? )
+ if( EXISTS "${ANDROID_NDK}/RELEASE.TXT" )
+  file( STRINGS "${ANDROID_NDK}/RELEASE.TXT" ANDROID_NDK_RELEASE_FULL LIMIT_COUNT 1 REGEX r[0-9]+[a-z]? )
+  string( REGEX MATCH r[0-9]+[a-z]? ANDROID_NDK_RELEASE "${ANDROID_NDK_RELEASE_FULL}" )
+ else()
+  set( ANDROID_NDK_RELEASE "r1x" )
+  set( ANDROID_NDK_RELEASE_FULL "unreleased" )
+ endif()
 elseif( ANDROID_STANDALONE_TOOLCHAIN )
  get_filename_component( ANDROID_STANDALONE_TOOLCHAIN "${ANDROID_STANDALONE_TOOLCHAIN}" ABSOLUTE )
  # try to detect change
@@ -536,6 +558,51 @@ else()
       sudo ln -s ~/my-android-toolchain ${ANDROID_STANDALONE_TOOLCHAIN_SEARCH_PATH}" )
 endif()
 
+# android NDK layout
+if( BUILD_WITH_ANDROID_NDK )
+ if( NOT DEFINED ANDROID_NDK_LAYOUT )
+  # try to automatically detect the layout
+  if( EXISTS "${ANDROID_NDK}/RELEASE.TXT")
+   set( ANDROID_NDK_LAYOUT "RELEASE" )
+  elseif( EXISTS "${ANDROID_NDK}/../../linux-x86/toolchain/" )
+   set( ANDROID_NDK_LAYOUT "LINARO" )
+  elseif( EXISTS "${ANDROID_NDK}/../../gcc/" )
+   set( ANDROID_NDK_LAYOUT "ANDROID" )
+  endif()
+ endif()
+ set( ANDROID_NDK_LAYOUT "${ANDROID_NDK_LAYOUT}" CACHE STRING "The inner layout of NDK" )
+ mark_as_advanced( ANDROID_NDK_LAYOUT )
+ if( ANDROID_NDK_LAYOUT STREQUAL "LINARO" )
+  set( ANDROID_NDK_HOST_SYSTEM_NAME ${ANDROID_NDK_HOST_SYSTEM_NAME2} ) # only 32-bit at the moment
+  set( ANDROID_NDK_TOOLCHAINS_PATH "${ANDROID_NDK}/../../${ANDROID_NDK_HOST_SYSTEM_NAME}/toolchain" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH  "" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH2 "" )
+ elseif( ANDROID_NDK_LAYOUT STREQUAL "ANDROID" )
+  set( ANDROID_NDK_HOST_SYSTEM_NAME ${ANDROID_NDK_HOST_SYSTEM_NAME2} ) # only 32-bit at the moment
+  set( ANDROID_NDK_TOOLCHAINS_PATH "${ANDROID_NDK}/../../gcc/${ANDROID_NDK_HOST_SYSTEM_NAME}/arm" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH  "" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH2 "" )
+ else() # ANDROID_NDK_LAYOUT STREQUAL "RELEASE"
+  set( ANDROID_NDK_TOOLCHAINS_PATH "${ANDROID_NDK}/toolchains" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH  "/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME}" )
+  set( ANDROID_NDK_TOOLCHAINS_SUBPATH2 "/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME2}" )
+ endif()
+ get_filename_component( ANDROID_NDK_TOOLCHAINS_PATH "${ANDROID_NDK_TOOLCHAINS_PATH}" ABSOLUTE )
+
+ # try to detect change of NDK
+ if( CMAKE_AR )
+  string( LENGTH "${ANDROID_NDK_TOOLCHAINS_PATH}" __length )
+  string( SUBSTRING "${CMAKE_AR}" 0 ${__length} __androidNdkPreviousPath )
+  if( NOT __androidNdkPreviousPath STREQUAL ANDROID_NDK_TOOLCHAINS_PATH )
+   message( FATAL_ERROR "It is not possible to change the path to the NDK on subsequent CMake run. You must remove all generated files from your build folder first.
+   " )
+  endif()
+  unset( __androidNdkPreviousPath )
+  unset( __length )
+ endif()
+endif()
+
+
 # get all the details about standalone toolchain
 if( BUILD_WITH_STANDALONE_TOOLCHAIN )
  __DETECT_NATIVE_API_LEVEL( ANDROID_SUPPORTED_NATIVE_API_LEVELS "${ANDROID_STANDALONE_TOOLCHAIN}/sysroot/usr/include/android/api-level.h" )
@@ -563,22 +630,27 @@ if( BUILD_WITH_STANDALONE_TOOLCHAIN )
  endif()
 endif()
 
-macro( __GLOB_NDK_TOOLCHAINS __availableToolchainsVar )
- foreach( __toolchain ${${__availableToolchainsVar}} )
-  if( "${__toolchain}" MATCHES "-clang3[.][0-9]$" AND NOT EXISTS "${ANDROID_NDK}/toolchains/${__toolchain}/prebuilt/" )
+macro( __GLOB_NDK_TOOLCHAINS __availableToolchainsVar __availableToolchainsLst __toolchain_subpath )
+ foreach( __toolchain ${${__availableToolchainsLst}} )
+  if( "${__toolchain}" MATCHES "-clang3[.][0-9]$" AND NOT EXISTS "${ANDROID_NDK_TOOLCHAINS_PATH}/${__toolchain}${__toolchain_subpath}" )
    string( REGEX REPLACE "-clang3[.][0-9]$" "-4.6" __gcc_toolchain "${__toolchain}" )
   else()
    set( __gcc_toolchain "${__toolchain}" )
   endif()
-  __DETECT_TOOLCHAIN_MACHINE_NAME( __machine "${ANDROID_NDK}/toolchains/${__gcc_toolchain}/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME}" )
+  __DETECT_TOOLCHAIN_MACHINE_NAME( __machine "${ANDROID_NDK_TOOLCHAINS_PATH}/${__gcc_toolchain}${__toolchain_subpath}" )
   if( __machine )
-   string( REGEX MATCH "[0-9]+[.][0-9]+([.][0-9]+)?$" __version "${__gcc_toolchain}" )
-   string( REGEX MATCH "^[^-]+" __arch "${__gcc_toolchain}" )
+   string( REGEX MATCH "[0-9]+[.][0-9]+([.][0-9x]+)?$" __version "${__gcc_toolchain}" )
+   if( __machine MATCHES i686 )
+    set( __arch "x86" )
+   elseif( __machine MATCHES arm )
+    set( __arch "arm" )
+   elseif( __machine MATCHES mipsel )
+    set( __arch "mipsel" )
+   endif()
    list( APPEND __availableToolchainMachines "${__machine}" )
    list( APPEND __availableToolchainArchs "${__arch}" )
    list( APPEND __availableToolchainCompilerVersions "${__version}" )
-  else()
-   list( REMOVE_ITEM ${__availableToolchainsVar} "${__toolchain}" )
+   list( APPEND ${__availableToolchainsVar} "${__toolchain}" )
   endif()
   unset( __gcc_toolchain )
  endforeach()
@@ -592,19 +664,31 @@ if( BUILD_WITH_ANDROID_NDK )
  set( __availableToolchainMachines "" )
  set( __availableToolchainArchs "" )
  set( __availableToolchainCompilerVersions "" )
- if( ANDROID_TOOLCHAIN_NAME AND EXISTS "${ANDROID_NDK}/toolchains/${ANDROID_TOOLCHAIN_NAME}/" )
+ if( ANDROID_TOOLCHAIN_NAME AND EXISTS "${ANDROID_NDK_TOOLCHAINS_PATH}/${ANDROID_TOOLCHAIN_NAME}/" )
   # do not go through all toolchains if we know the name
-  set( __availableToolchains "${ANDROID_TOOLCHAIN_NAME}" )
-  __GLOB_NDK_TOOLCHAINS( __availableToolchains )
+  set( __availableToolchainsLst "${ANDROID_TOOLCHAIN_NAME}" )
+  __GLOB_NDK_TOOLCHAINS( __availableToolchains __availableToolchainsLst "${ANDROID_NDK_TOOLCHAINS_SUBPATH}" )
+  if( NOT __availableToolchains AND NOT ANDROID_NDK_TOOLCHAINS_SUBPATH STREQUAL ANDROID_NDK_TOOLCHAINS_SUBPATH2 )
+   __GLOB_NDK_TOOLCHAINS( __availableToolchains __availableToolchainsLst "${ANDROID_NDK_TOOLCHAINS_SUBPATH2}" )
+   if( __availableToolchains )
+    set( ANDROID_NDK_TOOLCHAINS_SUBPATH ${ANDROID_NDK_TOOLCHAINS_SUBPATH2} )
+   endif()
+  endif()
  endif()
  if( NOT __availableToolchains )
-  file( GLOB __availableToolchains RELATIVE "${ANDROID_NDK}/toolchains" "${ANDROID_NDK}/toolchains/*" )
+  file( GLOB __availableToolchainsLst RELATIVE "${ANDROID_NDK_TOOLCHAINS_PATH}" "${ANDROID_NDK_TOOLCHAINS_PATH}/*" )
   if( __availableToolchains )
-   list(SORT __availableToolchains) # we need clang to go after gcc
+   list(SORT __availableToolchainsLst) # we need clang to go after gcc
+  endif()
+  __LIST_FILTER( __availableToolchainsLst "^[.]" )
+  __LIST_FILTER( __availableToolchainsLst "llvm" )
+  __GLOB_NDK_TOOLCHAINS( __availableToolchains __availableToolchainsLst "${ANDROID_NDK_TOOLCHAINS_SUBPATH}" )
+  if( NOT __availableToolchains AND NOT ANDROID_NDK_TOOLCHAINS_SUBPATH STREQUAL ANDROID_NDK_TOOLCHAINS_SUBPATH2 )
+   __GLOB_NDK_TOOLCHAINS( __availableToolchains __availableToolchainsLst "${ANDROID_NDK_TOOLCHAINS_SUBPATH2}" )
+   if( __availableToolchains )
+    set( ANDROID_NDK_TOOLCHAINS_SUBPATH ${ANDROID_NDK_TOOLCHAINS_SUBPATH2} )
+   endif()
   endif()
-  __LIST_FILTER( __availableToolchains "^[.]" )
-  __LIST_FILTER( __availableToolchains "llvm" )
-  __GLOB_NDK_TOOLCHAINS( __availableToolchains )
  endif()
  if( NOT __availableToolchains )
   message( FATAL_ERROR "Could not find any working toolchain in the NDK. Probably your Android NDK is broken." )
@@ -617,11 +701,11 @@ set( __uniqToolchainArchNames ${__availableToolchainArchs} )
 list( REMOVE_DUPLICATES __uniqToolchainArchNames )
 list( SORT __uniqToolchainArchNames )
 foreach( __arch ${__uniqToolchainArchNames} )
-list( APPEND ANDROID_SUPPORTED_ABIS ${ANDROID_SUPPORTED_ABIS_${__arch}} )
+ list( APPEND ANDROID_SUPPORTED_ABIS ${ANDROID_SUPPORTED_ABIS_${__arch}} )
 endforeach()
 unset( __uniqToolchainArchNames )
 if( NOT ANDROID_SUPPORTED_ABIS )
-message( FATAL_ERROR "No one of known Android ABIs is supported by this cmake toolchain." )
+ message( FATAL_ERROR "No one of known Android ABIs is supported by this cmake toolchain." )
 endif()
 
 # choose target ABI
@@ -733,6 +817,7 @@ else()
   list( GET __availableToolchainArchs ${__idx} __toolchainArch )
   if( __toolchainArch STREQUAL ANDROID_ARCH_FULLNAME )
    list( GET __availableToolchainCompilerVersions ${__idx} __toolchainVersion )
+   string( REPLACE "x" "99" __toolchainVersion "${__toolchainVersion}")
    if( __toolchainVersion VERSION_GREATER __toolchainMaxVersion )
     set( __toolchainMaxVersion "${__toolchainVersion}" )
     set( __toolchainIdx ${__idx} )
@@ -760,11 +845,22 @@ unset( __availableToolchainCompilerVersions )
 # choose native API level
 __INIT_VARIABLE( ANDROID_NATIVE_API_LEVEL ENV_ANDROID_NATIVE_API_LEVEL ANDROID_API_LEVEL ENV_ANDROID_API_LEVEL ANDROID_STANDALONE_TOOLCHAIN_API_LEVEL ANDROID_DEFAULT_NDK_API_LEVEL_${ANDROID_ARCH_NAME} ANDROID_DEFAULT_NDK_API_LEVEL )
 string( REGEX MATCH "[0-9]+" ANDROID_NATIVE_API_LEVEL "${ANDROID_NATIVE_API_LEVEL}" )
-# TODO: filter out unsupported levels
+# adjust API level
+set( __real_api_level ${ANDROID_DEFAULT_NDK_API_LEVEL_${ANDROID_ARCH_NAME}} )
+foreach( __level ${ANDROID_SUPPORTED_NATIVE_API_LEVELS} )
+ if( NOT __level GREATER ANDROID_NATIVE_API_LEVEL AND NOT __level LESS __real_api_level )
+  set( __real_api_level ${__level} )
+ endif()
+endforeach()
+if( __real_api_level AND NOT ANDROID_NATIVE_API_LEVEL EQUAL __real_api_level )
+ message( STATUS "Adjusting Android API level 'android-${ANDROID_NATIVE_API_LEVEL}' to 'android-${__real_api_level}'")
+ set( ANDROID_NATIVE_API_LEVEL ${__real_api_level} )
+endif()
+unset(__real_api_level)
 # validate
 list( FIND ANDROID_SUPPORTED_NATIVE_API_LEVELS "${ANDROID_NATIVE_API_LEVEL}" __levelIdx )
 if( __levelIdx EQUAL -1 )
- message( SEND_ERROR "Specified Android native API level (${ANDROID_NATIVE_API_LEVEL}) is not supported by your NDK/toolchain." )
+ message( SEND_ERROR "Specified Android native API level 'android-${ANDROID_NATIVE_API_LEVEL}' is not supported by your NDK/toolchain." )
 else()
  if( BUILD_WITH_ANDROID_NDK )
   __DETECT_NATIVE_API_LEVEL( __realApiLevel "${ANDROID_NDK}/platforms/android-${ANDROID_NATIVE_API_LEVEL}/arch-${ANDROID_ARCH_NAME}/usr/include/android/api-level.h" )
@@ -925,11 +1021,11 @@ if( "${ANDROID_TOOLCHAIN_NAME}" STREQUAL "standalone-clang" )
 elseif( "${ANDROID_TOOLCHAIN_NAME}" MATCHES "-clang3[.][0-9]?$" )
  string( REGEX MATCH "3[.][0-9]$" ANDROID_CLANG_VERSION "${ANDROID_TOOLCHAIN_NAME}")
  string( REGEX REPLACE "-clang${ANDROID_CLANG_VERSION}$" "-4.6" ANDROID_GCC_TOOLCHAIN_NAME "${ANDROID_TOOLCHAIN_NAME}" )
- if( NOT EXISTS "${ANDROID_NDK}/toolchains/llvm-${ANDROID_CLANG_VERSION}/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME}/bin/clang${TOOL_OS_SUFFIX}" )
-  message( FATAL_ERROR "Could not find the " )
+ if( NOT EXISTS "${ANDROID_NDK_TOOLCHAINS_PATH}/llvm-${ANDROID_CLANG_VERSION}${ANDROID_NDK_TOOLCHAINS_SUBPATH}/bin/clang${TOOL_OS_SUFFIX}" )
+  message( FATAL_ERROR "Could not find the Clang compiler driver" )
  endif()
  set( ANDROID_COMPILER_IS_CLANG 1 )
- set( ANDROID_CLANG_TOOLCHAIN_ROOT "${ANDROID_NDK}/toolchains/llvm-${ANDROID_CLANG_VERSION}/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME}" )
+ set( ANDROID_CLANG_TOOLCHAIN_ROOT "${ANDROID_NDK_TOOLCHAINS_PATH}/llvm-${ANDROID_CLANG_VERSION}${ANDROID_NDK_TOOLCHAINS_SUBPATH}" )
 else()
  set( ANDROID_GCC_TOOLCHAIN_NAME "${ANDROID_TOOLCHAIN_NAME}" )
  unset( ANDROID_COMPILER_IS_CLANG CACHE )
@@ -943,7 +1039,7 @@ endif()
 
 # setup paths and STL for NDK
 if( BUILD_WITH_ANDROID_NDK )
- set( ANDROID_TOOLCHAIN_ROOT "${ANDROID_NDK}/toolchains/${ANDROID_GCC_TOOLCHAIN_NAME}/prebuilt/${ANDROID_NDK_HOST_SYSTEM_NAME}" )
+ set( ANDROID_TOOLCHAIN_ROOT "${ANDROID_NDK_TOOLCHAINS_PATH}/${ANDROID_GCC_TOOLCHAIN_NAME}${ANDROID_NDK_TOOLCHAINS_SUBPATH}" )
  set( ANDROID_SYSROOT "${ANDROID_NDK}/platforms/android-${ANDROID_NATIVE_API_LEVEL}/arch-${ANDROID_ARCH_NAME}" )
 
  if( ANDROID_STL STREQUAL "none" )
@@ -1002,11 +1098,11 @@ if( BUILD_WITH_ANDROID_NDK )
  endif()
  # find libsupc++.a - rtti & exceptions
  if( ANDROID_STL STREQUAL "system_re" OR ANDROID_STL MATCHES "gnustl" )
-  if( ANDROID_NDK_RELEASE STRGREATER "r8" ) # r8b
-   set( __libsupcxx "${ANDROID_NDK}/sources/cxx-stl/gnu-libstdc++/${ANDROID_COMPILER_VERSION}/libs/${ANDROID_NDK_ABI_NAME}/libsupc++.a" )
-  elseif( NOT ANDROID_NDK_RELEASE STRLESS "r7" AND ANDROID_NDK_RELEASE STRLESS "r8b")
-   set( __libsupcxx "${ANDROID_NDK}/sources/cxx-stl/gnu-libstdc++/libs/${ANDROID_NDK_ABI_NAME}/libsupc++.a" )
-  else( ANDROID_NDK_RELEASE STRLESS "r7" )
+  set( __libsupcxx "${ANDROID_NDK}/sources/cxx-stl/gnu-libstdc++/${ANDROID_COMPILER_VERSION}/libs/${ANDROID_NDK_ABI_NAME}/libsupc++.a" ) # r8b or newer
+  if( NOT EXISTS "${__libsupcxx}" )
+   set( __libsupcxx "${ANDROID_NDK}/sources/cxx-stl/gnu-libstdc++/libs/${ANDROID_NDK_ABI_NAME}/libsupc++.a" ) # r7-r8
+  endif()
+  if( NOT EXISTS "${__libsupcxx}" ) # before r7
    if( ARMEABI_V7A )
     if( ANDROID_FORCE_ARM_BUILD )
      set( __libsupcxx "${ANDROID_TOOLCHAIN_ROOT}/${ANDROID_TOOLCHAIN_MACHINE_NAME}/lib/${CMAKE_SYSTEM_PROCESSOR}/libsupc++.a" )
@@ -1056,7 +1152,7 @@ unset( _ndk_ccache )
 
 # setup the cross-compiler
 if( NOT CMAKE_C_COMPILER )
- if( NDK_CCACHE )
+ if( NDK_CCACHE AND NOT ANDROID_SYSROOT MATCHES "[ ;\"]" )
   set( CMAKE_C_COMPILER   "${NDK_CCACHE}" CACHE PATH "ccache as C compiler" )
   set( CMAKE_CXX_COMPILER "${NDK_CCACHE}" CACHE PATH "ccache as C++ compiler" )
   if( ANDROID_COMPILER_IS_CLANG )
@@ -1128,11 +1224,25 @@ set( CMAKE_ASM_SOURCE_FILE_EXTENSIONS s S asm )
 remove_definitions( -DANDROID )
 add_definitions( -DANDROID )
 
-if(ANDROID_SYSROOT MATCHES "[ ;\"]")
- set( ANDROID_CXX_FLAGS "--sysroot=\"${ANDROID_SYSROOT}\"" )
+if( ANDROID_SYSROOT MATCHES "[ ;\"]" )
+ if( CMAKE_HOST_WIN32 )
+  # try to convert path to 8.3 form
+  file( WRITE "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/cvt83.cmd" "@echo %~s1" )
+  execute_process( COMMAND "$ENV{ComSpec}" /c "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/cvt83.cmd" "${ANDROID_SYSROOT}"
+                   OUTPUT_VARIABLE __path OUTPUT_STRIP_TRAILING_WHITESPACE
+                   RESULT_VARIABLE __result ERROR_QUIET )
+  if( __result EQUAL 0 )
+   file( TO_CMAKE_PATH "${__path}" ANDROID_SYSROOT )
+   set( ANDROID_CXX_FLAGS "--sysroot=${ANDROID_SYSROOT}" )
+  else()
+   set( ANDROID_CXX_FLAGS "--sysroot=\"${ANDROID_SYSROOT}\"" )
+  endif()
+ else()
+  set( ANDROID_CXX_FLAGS "'--sysroot=${ANDROID_SYSROOT}'" )
+ endif()
  if( NOT _CMAKE_IN_TRY_COMPILE )
-  # quotes will break try_compile and compiler identification
-  message(WARNING "Your Android system root has non-alphanumeric symbols. It can break compiler features detection and the whole build.")
+  # quotes can break try_compile and compiler identification
+  message(WARNING "Path to your Android NDK (or toolchain) has non-alphanumeric symbols.\nThe build might be broken.\n")
  endif()
 else()
  set( ANDROID_CXX_FLAGS "--sysroot=${ANDROID_SYSROOT}" )
@@ -1140,38 +1250,52 @@ endif()
 
 # NDK flags
 if( ARMEABI OR ARMEABI_V7A )
- set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fpic -D__ARM_ARCH_5__ -D__ARM_ARCH_5T__ -D__ARM_ARCH_5E__ -D__ARM_ARCH_5TE__" )
+ set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fpic -funwind-tables" )
  if( NOT ANDROID_FORCE_ARM_BUILD AND NOT ARMEABI_V6 )
-  # It is recommended to use the -mthumb compiler flag to force the generation
-  # of 16-bit Thumb-1 instructions (the default being 32-bit ARM ones).
-  set( ANDROID_CXX_FLAGS_RELEASE "-mthumb" )
-  set( ANDROID_CXX_FLAGS_DEBUG   "-marm -finline-limit=64" )
+  set( ANDROID_CXX_FLAGS_RELEASE "-mthumb -fomit-frame-pointer -fno-strict-aliasing" )
+  set( ANDROID_CXX_FLAGS_DEBUG   "-marm -fno-omit-frame-pointer -fno-strict-aliasing" )
+  if( NOT ANDROID_COMPILER_IS_CLANG )
+   set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -finline-limit=64" )
+  endif()
  else()
   # always compile ARMEABI_V6 in arm mode; otherwise there is no difference from ARMEABI
-  # O3 instead of O2/Os in release mode - like cmake sets for desktop gcc
-  set( ANDROID_CXX_FLAGS_RELEASE "-marm" )
-  set( ANDROID_CXX_FLAGS_DEBUG   "-marm -finline-limit=300" )
+  set( ANDROID_CXX_FLAGS_RELEASE "-marm -fomit-frame-pointer -fstrict-aliasing" )
+  set( ANDROID_CXX_FLAGS_DEBUG   "-marm -fno-omit-frame-pointer -fno-strict-aliasing" )
+  if( NOT ANDROID_COMPILER_IS_CLANG )
+   set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -funswitch-loops -finline-limit=300" )
+  endif()
  endif()
 elseif( X86 )
- set( ANDROID_CXX_FLAGS         "${ANDROID_CXX_FLAGS} -funwind-tables" )
- set( ANDROID_CXX_FLAGS_RELEASE "" )
- set( ANDROID_CXX_FLAGS_DEBUG   "-finline-limit=300" )
+ set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -funwind-tables" )
+ if( NOT ANDROID_COMPILER_IS_CLANG )
+  set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -funswitch-loops -finline-limit=300" )
+ else()
+  set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fPIC" )
+ endif()
+ set( ANDROID_CXX_FLAGS_RELEASE "-fomit-frame-pointer -fstrict-aliasing" )
+ set( ANDROID_CXX_FLAGS_DEBUG   "-fno-omit-frame-pointer -fno-strict-aliasing" )
 elseif( MIPS )
- set( ANDROID_CXX_FLAGS         "${ANDROID_CXX_FLAGS} -fpic -funwind-tables -fmessage-length=0 -fno-inline-functions-called-once -frename-registers" )
- set( ANDROID_CXX_FLAGS_RELEASE "-finline-limit=300 -fno-strict-aliasing" )
- set( ANDROID_CXX_FLAGS_DEBUG   "-finline-functions -fgcse-after-reload -frerun-cse-after-loop" )
+ set( ANDROID_CXX_FLAGS         "${ANDROID_CXX_FLAGS} -fpic -fno-strict-aliasing -finline-functions -ffunction-sections -funwind-tables -fmessage-length=0" )
+ set( ANDROID_CXX_FLAGS_RELEASE "-fomit-frame-pointer" )
+ set( ANDROID_CXX_FLAGS_DEBUG   "-fno-omit-frame-pointer" )
+ if( NOT ANDROID_COMPILER_IS_CLANG )
+  set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fno-inline-functions-called-once -fgcse-after-reload -frerun-cse-after-loop -frename-registers" )
+  set( ANDROID_CXX_FLAGS_RELEASE "${ANDROID_CXX_FLAGS_RELEASE} -funswitch-loops -finline-limit=300" )
+ endif()
 elseif()
  set( ANDROID_CXX_FLAGS_RELEASE "" )
  set( ANDROID_CXX_FLAGS_DEBUG   "" )
 endif()
 
-if( NOT X86 )
- set( ANDROID_CXX_FLAGS         "-Wno-psabi ${ANDROID_CXX_FLAGS}" )
+set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fsigned-char" ) # good/necessary when porting desktop libraries
+
+if( NOT X86 AND NOT ANDROID_COMPILER_IS_CLANG )
+ set( ANDROID_CXX_FLAGS "-Wno-psabi ${ANDROID_CXX_FLAGS}" )
 endif()
 
-set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -fsigned-char" ) # good/necessary when porting desktop libraries
-set( ANDROID_CXX_FLAGS_RELEASE "${ANDROID_CXX_FLAGS_RELEASE} -fomit-frame-pointer" )
-set( ANDROID_CXX_FLAGS_DEBUG   "${ANDROID_CXX_FLAGS_DEBUG} -fno-strict-aliasing -fno-omit-frame-pointer" )
+if( NOT ANDROID_COMPILER_VERSION VERSION_LESS "4.6" )
+ set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -no-canonical-prefixes" ) # see https://android-review.googlesource.com/#/c/47564/
+endif()
 
 # ABI-specific flags
 if( ARMEABI_V7A )
@@ -1189,22 +1313,18 @@ elseif( ARMEABI )
  set( ANDROID_CXX_FLAGS "${ANDROID_CXX_FLAGS} -march=armv5te -mtune=xscale -msoft-float" )
 endif()
 
+if( ANDROID_STL MATCHES "gnustl" AND (EXISTS "${__libstl}" OR EXISTS "${__libsupcxx}") )
+ set( CMAKE_CXX_CREATE_SHARED_LIBRARY "<CMAKE_C_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
+ set( CMAKE_CXX_CREATE_SHARED_MODULE  "<CMAKE_C_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
+ set( CMAKE_CXX_LINK_EXECUTABLE       "<CMAKE_C_COMPILER> <FLAGS> <CMAKE_CXX_LINK_FLAGS> <LINK_FLAGS> <OBJECTS> -o <TARGET> <LINK_LIBRARIES>" )
+else()
+ set( CMAKE_CXX_CREATE_SHARED_LIBRARY "<CMAKE_CXX_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
+ set( CMAKE_CXX_CREATE_SHARED_MODULE  "<CMAKE_CXX_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
+ set( CMAKE_CXX_LINK_EXECUTABLE       "<CMAKE_CXX_COMPILER> <FLAGS> <CMAKE_CXX_LINK_FLAGS> <LINK_FLAGS> <OBJECTS> -o <TARGET> <LINK_LIBRARIES>" )
+endif()
+
 # STL
 if( EXISTS "${__libstl}" OR EXISTS "${__libsupcxx}" )
- if( ANDROID_STL MATCHES "gnustl" )
-  set( CMAKE_CXX_CREATE_SHARED_LIBRARY "<CMAKE_C_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
-  set( CMAKE_CXX_CREATE_SHARED_MODULE  "<CMAKE_C_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
-  set( CMAKE_CXX_LINK_EXECUTABLE       "<CMAKE_C_COMPILER> <FLAGS> <CMAKE_CXX_LINK_FLAGS> <LINK_FLAGS> <OBJECTS> -o <TARGET> <LINK_LIBRARIES>" )
- else()
-  set( CMAKE_CXX_CREATE_SHARED_LIBRARY "<CMAKE_CXX_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
-  set( CMAKE_CXX_CREATE_SHARED_MODULE  "<CMAKE_CXX_COMPILER> <CMAKE_SHARED_LIBRARY_CXX_FLAGS> <LANGUAGE_COMPILE_FLAGS> <LINK_FLAGS> <CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS> <CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG><TARGET_SONAME> -o <TARGET> <OBJECTS> <LINK_LIBRARIES>" )
-  set( CMAKE_CXX_LINK_EXECUTABLE       "<CMAKE_CXX_COMPILER> <FLAGS> <CMAKE_CXX_LINK_FLAGS> <LINK_FLAGS> <OBJECTS> -o <TARGET> <LINK_LIBRARIES>" )
- endif()
- if ( X86 AND ANDROID_STL MATCHES "gnustl" AND ANDROID_NDK_RELEASE STREQUAL "r6" )
-  # workaround "undefined reference to `__dso_handle'" problem
-  set( CMAKE_CXX_CREATE_SHARED_LIBRARY "${CMAKE_CXX_CREATE_SHARED_LIBRARY} \"${ANDROID_SYSROOT}/usr/lib/crtbegin_so.o\"" )
-  set( CMAKE_CXX_CREATE_SHARED_MODULE  "${CMAKE_CXX_CREATE_SHARED_MODULE} \"${ANDROID_SYSROOT}/usr/lib/crtbegin_so.o\"" )
- endif()
  if( EXISTS "${__libstl}" )
   set( CMAKE_CXX_CREATE_SHARED_LIBRARY "${CMAKE_CXX_CREATE_SHARED_LIBRARY} \"${__libstl}\"" )
   set( CMAKE_CXX_CREATE_SHARED_MODULE  "${CMAKE_CXX_CREATE_SHARED_MODULE} \"${__libstl}\"" )
@@ -1223,9 +1343,12 @@ if( EXISTS "${__libstl}" OR EXISTS "${__libsupcxx}" )
   set( CMAKE_C_LINK_EXECUTABLE       "${CMAKE_C_LINK_EXECUTABLE} \"${__libsupcxx}\"" )
  endif()
  if( ANDROID_STL MATCHES "gnustl" )
-  set( CMAKE_CXX_CREATE_SHARED_LIBRARY "${CMAKE_CXX_CREATE_SHARED_LIBRARY} -lm" )
-  set( CMAKE_CXX_CREATE_SHARED_MODULE  "${CMAKE_CXX_CREATE_SHARED_MODULE} -lm" )
-  set( CMAKE_CXX_LINK_EXECUTABLE       "${CMAKE_CXX_LINK_EXECUTABLE} -lm" )
+  if( NOT EXISTS "${ANDROID_LIBM_PATH}" )
+   set( ANDROID_LIBM_PATH -lm )
+  endif()
+  set( CMAKE_CXX_CREATE_SHARED_LIBRARY "${CMAKE_CXX_CREATE_SHARED_LIBRARY} ${ANDROID_LIBM_PATH}" )
+  set( CMAKE_CXX_CREATE_SHARED_MODULE  "${CMAKE_CXX_CREATE_SHARED_MODULE} ${ANDROID_LIBM_PATH}" )
+  set( CMAKE_CXX_LINK_EXECUTABLE       "${CMAKE_CXX_LINK_EXECUTABLE} ${ANDROID_LIBM_PATH}" )
  endif()
 endif()
 
@@ -1261,7 +1384,14 @@ if( ARMEABI_V7A )
 endif()
 
 if( ANDROID_NO_UNDEFINED )
- set( ANDROID_LINKER_FLAGS "${ANDROID_LINKER_FLAGS} -Wl,--no-undefined" )
+ if( MIPS )
+  # there is some sysroot-related problem in mips linker...
+  if( NOT ANDROID_SYSROOT MATCHES "[ ;\"]" )
+   set( ANDROID_LINKER_FLAGS "${ANDROID_LINKER_FLAGS} -Wl,--no-undefined -Wl,-rpath-link,${ANDROID_SYSROOT}/usr/lib" )
+  endif()
+ else()
+  set( ANDROID_LINKER_FLAGS "${ANDROID_LINKER_FLAGS} -Wl,--no-undefined" )
+ endif()
 endif()
 
 if( ANDROID_SO_UNDEFINED )
@@ -1308,9 +1438,6 @@ if( ANDROID_COMPILER_IS_CLANG )
   set( ANDROID_CXX_FLAGS "-target ${ANDROID_LLVM_TRIPLE} ${ANDROID_CXX_FLAGS}" )
  endif()
  if( BUILD_WITH_ANDROID_NDK )
-  if(ANDROID_ARCH_NAME STREQUAL "arm" )
-   set( ANDROID_CXX_FLAGS "-isystem ${ANDROID_CLANG_TOOLCHAIN_ROOT}/lib/clang/${ANDROID_CLANG_VERSION}/include ${ANDROID_CXX_FLAGS}" )
-  endif()
   set( ANDROID_CXX_FLAGS "-gcc-toolchain ${ANDROID_TOOLCHAIN_ROOT} ${ANDROID_CXX_FLAGS}" )
  endif()
 endif()
@@ -1326,6 +1453,12 @@ set( CMAKE_SHARED_LINKER_FLAGS ""                        CACHE STRING "shared li
 set( CMAKE_MODULE_LINKER_FLAGS ""                        CACHE STRING "module linker flags" )
 set( CMAKE_EXE_LINKER_FLAGS    "-Wl,-z,nocopyreloc"      CACHE STRING "executable linker flags" )
 
+# put flags to cache (for debug purpose only)
+set( ANDROID_CXX_FLAGS         "${ANDROID_CXX_FLAGS}"         CACHE INTERNAL "Android specific c/c++ flags" )
+set( ANDROID_CXX_FLAGS_RELEASE "${ANDROID_CXX_FLAGS_RELEASE}" CACHE INTERNAL "Android specific c/c++ Release flags" )
+set( ANDROID_CXX_FLAGS_DEBUG   "${ANDROID_CXX_FLAGS_DEBUG}"   CACHE INTERNAL "Android specific c/c++ Debug flags" )
+set( ANDROID_LINKER_FLAGS      "${ANDROID_LINKER_FLAGS}"      CACHE INTERNAL "Android specific c/c++ linker flags" )
+
 # finish flags
 set( CMAKE_CXX_FLAGS           "${ANDROID_CXX_FLAGS} ${CMAKE_CXX_FLAGS}" )
 set( CMAKE_C_FLAGS             "${ANDROID_CXX_FLAGS} ${CMAKE_C_FLAGS}" )
@@ -1338,9 +1471,9 @@ set( CMAKE_MODULE_LINKER_FLAGS "${ANDROID_LINKER_FLAGS} ${CMAKE_MODULE_LINKER_FL
 set( CMAKE_EXE_LINKER_FLAGS    "${ANDROID_LINKER_FLAGS} ${CMAKE_EXE_LINKER_FLAGS}" )
 
 if( MIPS AND BUILD_WITH_ANDROID_NDK AND ANDROID_NDK_RELEASE STREQUAL "r8" )
- set( CMAKE_SHARED_LINKER_FLAGS "-Wl,-T,${ANDROID_NDK}/toolchains/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.xsc ${CMAKE_SHARED_LINKER_FLAGS}" )
- set( CMAKE_MODULE_LINKER_FLAGS "-Wl,-T,${ANDROID_NDK}/toolchains/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.xsc ${CMAKE_MODULE_LINKER_FLAGS}" )
- set( CMAKE_EXE_LINKER_FLAGS    "-Wl,-T,${ANDROID_NDK}/toolchains/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.x ${CMAKE_EXE_LINKER_FLAGS}" )
+ set( CMAKE_SHARED_LINKER_FLAGS "-Wl,-T,${ANDROID_NDK_TOOLCHAINS_PATH}/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.xsc ${CMAKE_SHARED_LINKER_FLAGS}" )
+ set( CMAKE_MODULE_LINKER_FLAGS "-Wl,-T,${ANDROID_NDK_TOOLCHAINS_PATH}/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.xsc ${CMAKE_MODULE_LINKER_FLAGS}" )
+ set( CMAKE_EXE_LINKER_FLAGS    "-Wl,-T,${ANDROID_NDK_TOOLCHAINS_PATH}/${ANDROID_GCC_TOOLCHAIN_NAME}/mipself.x ${CMAKE_EXE_LINKER_FLAGS}" )
 endif()
 
 # configure rtti
@@ -1367,6 +1500,43 @@ endif()
 include_directories( SYSTEM "${ANDROID_SYSROOT}/usr/include" ${ANDROID_STL_INCLUDE_DIRS} )
 link_directories( "${CMAKE_INSTALL_PREFIX}/libs/${ANDROID_NDK_ABI_NAME}" )
 
+# detect if need link crtbegin_so.o explicitly
+if( NOT DEFINED ANDROID_EXPLICIT_CRT_LINK )
+ set( __cmd "${CMAKE_CXX_CREATE_SHARED_LIBRARY}" )
+ string( REPLACE "<CMAKE_CXX_COMPILER>" "${CMAKE_CXX_COMPILER} ${CMAKE_CXX_COMPILER_ARG1}" __cmd "${__cmd}" )
+ string( REPLACE "<CMAKE_C_COMPILER>"   "${CMAKE_C_COMPILER} ${CMAKE_C_COMPILER_ARG1}"   __cmd "${__cmd}" )
+ string( REPLACE "<CMAKE_SHARED_LIBRARY_CXX_FLAGS>" "${CMAKE_CXX_FLAGS}" __cmd "${__cmd}" )
+ string( REPLACE "<LANGUAGE_COMPILE_FLAGS>" "" __cmd "${__cmd}" )
+ string( REPLACE "<LINK_FLAGS>" "${CMAKE_SHARED_LINKER_FLAGS}" __cmd "${__cmd}" )
+ string( REPLACE "<CMAKE_SHARED_LIBRARY_CREATE_CXX_FLAGS>" "-shared" __cmd "${__cmd}" )
+ string( REPLACE "<CMAKE_SHARED_LIBRARY_SONAME_CXX_FLAG>" "" __cmd "${__cmd}" )
+ string( REPLACE "<TARGET_SONAME>" "" __cmd "${__cmd}" )
+ string( REPLACE "<TARGET>" "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/toolchain_crtlink_test.so" __cmd "${__cmd}" )
+ string( REPLACE "<OBJECTS>" "\"${ANDROID_SYSROOT}/usr/lib/crtbegin_so.o\"" __cmd "${__cmd}" )
+ string( REPLACE "<LINK_LIBRARIES>" "" __cmd "${__cmd}" )
+ separate_arguments( __cmd )
+ foreach( __var ANDROID_NDK ANDROID_NDK_TOOLCHAINS_PATH ANDROID_STANDALONE_TOOLCHAIN )
+  if( ${__var} )
+   set( __tmp "${${__var}}" )
+   separate_arguments( __tmp )
+   string( REPLACE "${__tmp}" "${${__var}}" __cmd "${__cmd}")
+  endif()
+ endforeach()
+ string( REPLACE "'" "" __cmd "${__cmd}" )
+ string( REPLACE "\"" "" __cmd "${__cmd}" )
+ execute_process( COMMAND ${__cmd} RESULT_VARIABLE __cmd_result OUTPUT_QUIET ERROR_QUIET )
+ if( __cmd_result EQUAL 0 )
+  set( ANDROID_EXPLICIT_CRT_LINK ON )
+ else()
+  set( ANDROID_EXPLICIT_CRT_LINK OFF )
+ endif()
+endif()
+
+if( ANDROID_EXPLICIT_CRT_LINK )
+ set( CMAKE_CXX_CREATE_SHARED_LIBRARY "${CMAKE_CXX_CREATE_SHARED_LIBRARY} \"${ANDROID_SYSROOT}/usr/lib/crtbegin_so.o\"" )
+ set( CMAKE_CXX_CREATE_SHARED_MODULE  "${CMAKE_CXX_CREATE_SHARED_MODULE} \"${ANDROID_SYSROOT}/usr/lib/crtbegin_so.o\"" )
+endif()
+
 # setup output directories
 set( LIBRARY_OUTPUT_PATH_ROOT ${CMAKE_SOURCE_DIR} CACHE PATH "root for library output, set this to change where android libs are installed to" )
 set( CMAKE_INSTALL_PREFIX "${ANDROID_TOOLCHAIN_ROOT}/user" CACHE STRING "path for installing" )
@@ -1456,7 +1626,9 @@ endmacro()
 if( NOT PROJECT_NAME STREQUAL "CMAKE_TRY_COMPILE" )
  set( __toolchain_config "")
  foreach( __var NDK_CCACHE  LIBRARY_OUTPUT_PATH_ROOT  ANDROID_FORBID_SYGWIN  ANDROID_SET_OBSOLETE_VARIABLES
+                ANDROID_NDK_HOST_X64
                 ANDROID_NDK
+                ANDROID_NDK_LAYOUT
                 ANDROID_STANDALONE_TOOLCHAIN
                 ANDROID_TOOLCHAIN_NAME
                 ANDROID_ABI
@@ -1470,6 +1642,8 @@ if( NOT PROJECT_NAME STREQUAL "CMAKE_TRY_COMPILE" )
                 ANDROID_GOLD_LINKER
                 ANDROID_NOEXECSTACK
                 ANDROID_RELRO
+                ANDROID_LIBM_PATH
+                ANDROID_EXPLICIT_CRT_LINK
                 )
   if( DEFINED ${__var} )
    if( "${__var}" MATCHES " ")
@@ -1512,6 +1686,8 @@ endif()
 #   ANDROID_NDK
 #   ANDROID_STANDALONE_TOOLCHAIN
 #   ANDROID_TOOLCHAIN_NAME : the NDK name of compiler toolchain
+#   ANDROID_NDK_HOST_X64 : try to use x86_64 toolchain (default for x64 host systems)
+#   ANDROID_NDK_LAYOUT : the inner NDK structure (RELEASE, LINARO, ANDROID)
 #   LIBRARY_OUTPUT_PATH_ROOT : <any valid path>
 #   NDK_CCACHE : <path to your ccache executable>
 # Obsolete:
@@ -1536,7 +1712,7 @@ endif()
 #   BUILD_WITH_STANDALONE_TOOLCHAIN : TRUE if standalone toolchain is used
 #   ANDROID_NDK_HOST_SYSTEM_NAME : "windows", "linux-x86" or "darwin-x86" depending on host platform
 #   ANDROID_NDK_ABI_NAME : "armeabi", "armeabi-v7a", "x86" or "mips" depending on ANDROID_ABI
-#   ANDROID_NDK_RELEASE : one of r5, r5b, r5c, r6, r6b, r7, r7b, r7c, r8, r8b, r8c, r8d; set only for NDK
+#   ANDROID_NDK_RELEASE : one of r5, r5b, r5c, r6, r6b, r7, r7b, r7c, r8, r8b, r8c, r8d, r8e; set only for NDK
 #   ANDROID_ARCH_NAME : "arm" or "x86" or "mips" depending on ANDROID_ABI
 #   ANDROID_SYSROOT : path to the compiler sysroot
 #   TOOL_OS_SUFFIX : "" or ".exe" depending on host platform
@@ -1557,6 +1733,7 @@ endif()
 #   ANDROID_EXCEPTIONS : if exceptions are enabled by the runtime
 #   ANDROID_GCC_TOOLCHAIN_NAME : read-only, differs from ANDROID_TOOLCHAIN_NAME only if clang is used
 #   ANDROID_CLANG_VERSION : version of clang compiler if clang is used
+#   ANDROID_LIBM_PATH : path to libm.so (set to something like $(TOP)/out/target/product/<product_name>/obj/lib/libm.so) to workaround unresolved `sincos`
 #
 # Defaults:
 #   ANDROID_DEFAULT_NDK_API_LEVEL
diff --git a/android/java.rst b/platforms/android/java.rst
similarity index 100%
rename from android/java.rst
rename to platforms/android/java.rst
diff --git a/android/libinfo/CMakeLists.txt b/platforms/android/libinfo/CMakeLists.txt
similarity index 100%
rename from android/libinfo/CMakeLists.txt
rename to platforms/android/libinfo/CMakeLists.txt
diff --git a/android/libinfo/info.c b/platforms/android/libinfo/info.c
similarity index 100%
rename from android/libinfo/info.c
rename to platforms/android/libinfo/info.c
diff --git a/android/package/AndroidManifest.xml b/platforms/android/package/AndroidManifest.xml
similarity index 100%
rename from android/package/AndroidManifest.xml
rename to platforms/android/package/AndroidManifest.xml
diff --git a/android/package/CMakeLists.txt b/platforms/android/package/CMakeLists.txt
similarity index 98%
rename from android/package/CMakeLists.txt
rename to platforms/android/package/CMakeLists.txt
index 0bfb3fe899..1382a078cf 100644
--- a/android/package/CMakeLists.txt
+++ b/platforms/android/package/CMakeLists.txt
@@ -91,4 +91,4 @@ add_custom_command(
 
 install(FILES "${APK_NAME}" DESTINATION "apk/" COMPONENT main)
 add_custom_target(android_package ALL SOURCES "${APK_NAME}" )
-add_dependencies(android_package opencv_java)
\ No newline at end of file
+add_dependencies(android_package opencv_java)
diff --git a/android/package/res/drawable/icon.png b/platforms/android/package/res/drawable/icon.png
similarity index 100%
rename from android/package/res/drawable/icon.png
rename to platforms/android/package/res/drawable/icon.png
diff --git a/android/package/res/values/strings.xml b/platforms/android/package/res/values/strings.xml
similarity index 100%
rename from android/package/res/values/strings.xml
rename to platforms/android/package/res/values/strings.xml
diff --git a/android/refman.rst b/platforms/android/refman.rst
similarity index 92%
rename from android/refman.rst
rename to platforms/android/refman.rst
index 12d7ea6ec0..b830bbe09a 100644
--- a/android/refman.rst
+++ b/platforms/android/refman.rst
@@ -6,4 +6,4 @@ OpenCV4Android Reference
    :maxdepth: 2
 
    service/doc/index.rst
-   java.rst
\ No newline at end of file
+   java.rst
diff --git a/android/service/CMakeLists.txt b/platforms/android/service/CMakeLists.txt
similarity index 100%
rename from android/service/CMakeLists.txt
rename to platforms/android/service/CMakeLists.txt
diff --git a/android/service/all.py b/platforms/android/service/all.py
similarity index 97%
rename from android/service/all.py
rename to platforms/android/service/all.py
index 0993b89acf..18ab4570bd 100755
--- a/android/service/all.py
+++ b/platforms/android/service/all.py
@@ -40,4 +40,4 @@ if (__name__ ==  "__main__"):
         os.system(PushCommand)
         print("Testing on device \"%s\"" % Name)
         TestCommand = "%s \"%s\" \"%s\" 2>&1" % (os.path.join(HomeDir, "test_native.py"), Arch, Name)
-        os.system(TestCommand)
\ No newline at end of file
+        os.system(TestCommand)
diff --git a/android/service/device.conf b/platforms/android/service/device.conf
similarity index 72%
rename from android/service/device.conf
rename to platforms/android/service/device.conf
index 2953b54510..f8c6b5a865 100644
--- a/android/service/device.conf
+++ b/platforms/android/service/device.conf
@@ -1,3 +1,3 @@
 armeabi;15c000000581404;
 x86;0123456789ABCDEF;
-mips;Novo7 Basic;
\ No newline at end of file
+mips;Novo7 Basic;
diff --git a/android/service/doc/AndroidAppUsageModel.dia b/platforms/android/service/doc/AndroidAppUsageModel.dia
similarity index 100%
rename from android/service/doc/AndroidAppUsageModel.dia
rename to platforms/android/service/doc/AndroidAppUsageModel.dia
diff --git a/android/service/doc/BaseLoaderCallback.rst b/platforms/android/service/doc/BaseLoaderCallback.rst
similarity index 98%
rename from android/service/doc/BaseLoaderCallback.rst
rename to platforms/android/service/doc/BaseLoaderCallback.rst
index 71915c449d..f8c30a8008 100644
--- a/android/service/doc/BaseLoaderCallback.rst
+++ b/platforms/android/service/doc/BaseLoaderCallback.rst
@@ -48,7 +48,7 @@ See the "15-puzzle" OpenCV sample for details.
         super.onResume();
 
         Log.i(TAG, "Trying to load OpenCV library");
-        if (!OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_4, this, mOpenCVCallBack))
+        if (!OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_6, this, mOpenCVCallBack))
         {
             Log.e(TAG, "Cannot connect to OpenCV Manager");
         }
diff --git a/android/service/doc/InstallCallbackInterface.rst b/platforms/android/service/doc/InstallCallbackInterface.rst
similarity index 100%
rename from android/service/doc/InstallCallbackInterface.rst
rename to platforms/android/service/doc/InstallCallbackInterface.rst
diff --git a/android/service/doc/Intro.rst b/platforms/android/service/doc/Intro.rst
similarity index 98%
rename from android/service/doc/Intro.rst
rename to platforms/android/service/doc/Intro.rst
index d0e9d73470..4b00b8e4a8 100644
--- a/android/service/doc/Intro.rst
+++ b/platforms/android/service/doc/Intro.rst
@@ -41,4 +41,4 @@ Subsequent launches of OpenCV apps\:
 Architecture of OpenCV Manager
 ------------------------------
 
-.. image:: img/Structure.png
\ No newline at end of file
+.. image:: img/Structure.png
diff --git a/android/service/doc/JavaHelper.rst b/platforms/android/service/doc/JavaHelper.rst
similarity index 92%
rename from android/service/doc/JavaHelper.rst
rename to platforms/android/service/doc/JavaHelper.rst
index 34798c267e..9262a7cf73 100644
--- a/android/service/doc/JavaHelper.rst
+++ b/platforms/android/service/doc/JavaHelper.rst
@@ -51,3 +51,11 @@ OpenCV version constants
 .. data:: OPENCV_VERSION_2_4_4
 
     OpenCV Library version 2.4.4
+
+.. data:: OPENCV_VERSION_2_4_5
+
+    OpenCV Library version 2.4.5
+
+.. data:: OPENCV_VERSION_2_4_6
+
+    OpenCV Library version 2.4.6
diff --git a/android/service/doc/LibInstallAproved.dia b/platforms/android/service/doc/LibInstallAproved.dia
similarity index 100%
rename from android/service/doc/LibInstallAproved.dia
rename to platforms/android/service/doc/LibInstallAproved.dia
diff --git a/android/service/doc/LibInstallCanceled.dia b/platforms/android/service/doc/LibInstallCanceled.dia
similarity index 100%
rename from android/service/doc/LibInstallCanceled.dia
rename to platforms/android/service/doc/LibInstallCanceled.dia
diff --git a/android/service/doc/LibInstalled.dia b/platforms/android/service/doc/LibInstalled.dia
similarity index 100%
rename from android/service/doc/LibInstalled.dia
rename to platforms/android/service/doc/LibInstalled.dia
diff --git a/android/service/doc/LoaderCallbackInterface.rst b/platforms/android/service/doc/LoaderCallbackInterface.rst
similarity index 96%
rename from android/service/doc/LoaderCallbackInterface.rst
rename to platforms/android/service/doc/LoaderCallbackInterface.rst
index 440b6b6737..63838a1272 100644
--- a/android/service/doc/LoaderCallbackInterface.rst
+++ b/platforms/android/service/doc/LoaderCallbackInterface.rst
@@ -46,4 +46,4 @@ Initialization status constants
 
 .. data:: INIT_FAILED
 
-    OpenCV library initialization failed
\ No newline at end of file
+    OpenCV library initialization failed
diff --git a/android/service/doc/NoService.dia b/platforms/android/service/doc/NoService.dia
similarity index 100%
rename from android/service/doc/NoService.dia
rename to platforms/android/service/doc/NoService.dia
diff --git a/android/service/doc/Structure.dia b/platforms/android/service/doc/Structure.dia
similarity index 100%
rename from android/service/doc/Structure.dia
rename to platforms/android/service/doc/Structure.dia
diff --git a/android/service/doc/UseCases.rst b/platforms/android/service/doc/UseCases.rst
similarity index 95%
rename from android/service/doc/UseCases.rst
rename to platforms/android/service/doc/UseCases.rst
index bbc7da02c3..50ac050e9c 100644
--- a/android/service/doc/UseCases.rst
+++ b/platforms/android/service/doc/UseCases.rst
@@ -29,4 +29,4 @@ Regular application start
 
 OpenCV Manager and OpenCV library has been already installed.
 
-.. image:: img/LibInstalled.png
\ No newline at end of file
+.. image:: img/LibInstalled.png
diff --git a/android/service/doc/build_uml.py b/platforms/android/service/doc/build_uml.py
similarity index 87%
rename from android/service/doc/build_uml.py
rename to platforms/android/service/doc/build_uml.py
index df9eb7bcd5..94b61d6aed 100755
--- a/android/service/doc/build_uml.py
+++ b/platforms/android/service/doc/build_uml.py
@@ -20,4 +20,4 @@ if (not os.path.exists(TARGET_PATH)):
 
 for filename in os.listdir("."):
     if ("dia" == filename[-3:]):
-        os.system("%s --export %s %s" % (DiaPath, os.path.join(TARGET_PATH, filename[0:len(filename)-4] + ".png"), filename))
\ No newline at end of file
+        os.system("%s --export %s %s" % (DiaPath, os.path.join(TARGET_PATH, filename[0:len(filename)-4] + ".png"), filename))
diff --git a/android/service/doc/img/AndroidAppUsageModel.png b/platforms/android/service/doc/img/AndroidAppUsageModel.png
similarity index 100%
rename from android/service/doc/img/AndroidAppUsageModel.png
rename to platforms/android/service/doc/img/AndroidAppUsageModel.png
diff --git a/android/service/doc/img/LibInstallAproved.png b/platforms/android/service/doc/img/LibInstallAproved.png
similarity index 100%
rename from android/service/doc/img/LibInstallAproved.png
rename to platforms/android/service/doc/img/LibInstallAproved.png
diff --git a/android/service/doc/img/LibInstallCanceled.png b/platforms/android/service/doc/img/LibInstallCanceled.png
similarity index 100%
rename from android/service/doc/img/LibInstallCanceled.png
rename to platforms/android/service/doc/img/LibInstallCanceled.png
diff --git a/android/service/doc/img/LibInstalled.png b/platforms/android/service/doc/img/LibInstalled.png
similarity index 100%
rename from android/service/doc/img/LibInstalled.png
rename to platforms/android/service/doc/img/LibInstalled.png
diff --git a/android/service/doc/img/NoService.png b/platforms/android/service/doc/img/NoService.png
similarity index 100%
rename from android/service/doc/img/NoService.png
rename to platforms/android/service/doc/img/NoService.png
diff --git a/android/service/doc/img/Structure.png b/platforms/android/service/doc/img/Structure.png
similarity index 100%
rename from android/service/doc/img/Structure.png
rename to platforms/android/service/doc/img/Structure.png
diff --git a/android/service/doc/index.rst b/platforms/android/service/doc/index.rst
similarity index 100%
rename from android/service/doc/index.rst
rename to platforms/android/service/doc/index.rst
diff --git a/android/service/engine/.classpath b/platforms/android/service/engine/.classpath
similarity index 100%
rename from android/service/engine/.classpath
rename to platforms/android/service/engine/.classpath
diff --git a/android/service/engine/.project b/platforms/android/service/engine/.project
similarity index 100%
rename from android/service/engine/.project
rename to platforms/android/service/engine/.project
diff --git a/android/service/engine/AndroidManifest.xml b/platforms/android/service/engine/AndroidManifest.xml
similarity index 90%
rename from android/service/engine/AndroidManifest.xml
rename to platforms/android/service/engine/AndroidManifest.xml
index f4f0eb94fa..dc992b3a62 100644
--- a/android/service/engine/AndroidManifest.xml
+++ b/platforms/android/service/engine/AndroidManifest.xml
@@ -1,8 +1,8 @@
 <?xml version="1.0" encoding="utf-8"?>
 <manifest xmlns:android="http://schemas.android.com/apk/res/android"
     package="org.opencv.engine"
-    android:versionCode="26@ANDROID_PLATFORM_VERSION_CODE@"
-    android:versionName="2.6" >
+    android:versionCode="210@ANDROID_PLATFORM_VERSION_CODE@"
+    android:versionName="2.10" >
 
     <uses-sdk android:minSdkVersion="@ANDROID_NATIVE_API_LEVEL@" />
     <uses-feature android:name="android.hardware.touchscreen" android:required="false"/>
@@ -26,6 +26,5 @@
             <category android:name="android.intent.category.LAUNCHER" />
         </intent-filter>
     </activity>
-
-</application>
+    </application>
 </manifest>
\ No newline at end of file
diff --git a/android/service/engine/CMakeLists.txt b/platforms/android/service/engine/CMakeLists.txt
similarity index 74%
rename from android/service/engine/CMakeLists.txt
rename to platforms/android/service/engine/CMakeLists.txt
index 8b88393942..b1cac9383c 100644
--- a/android/service/engine/CMakeLists.txt
+++ b/platforms/android/service/engine/CMakeLists.txt
@@ -24,21 +24,34 @@ else()
   message(WARNING "Can not automatically determine the value for ANDROID_PLATFORM_VERSION_CODE")
 endif()
 
-configure_file("${CMAKE_CURRENT_SOURCE_DIR}/${ANDROID_MANIFEST_FILE}" "${OpenCV_BINARY_DIR}/android/service/engine/.build/${ANDROID_MANIFEST_FILE}"  @ONLY)
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/${ANDROID_MANIFEST_FILE}" "${OpenCV_BINARY_DIR}/platforms/android/service/engine/.build/${ANDROID_MANIFEST_FILE}"  @ONLY)
 
-link_directories("${ANDROID_SOURCE_TREE}/out/target/product/generic/system/lib" "${ANDROID_SOURCE_TREE}/out/target/product/${ANDROID_PRODUCT}/system/lib" "${ANDROID_SOURCE_TREE}/bin/${ANDROID_ARCH_NAME}")
+link_directories(
+  "${ANDROID_SOURCE_TREE}/out/target/product/generic/system/lib"
+  "${ANDROID_SOURCE_TREE}/out/target/product/${ANDROID_PRODUCT}/system/lib"
+  "${ANDROID_SOURCE_TREE}/bin/${ANDROID_ARCH_NAME}")
+
+file(GLOB engine_files "jni/BinderComponent/*.cpp" "jni/BinderComponent/*.h" "jni/include/*.h")
+set(engine_libs "z" "binder" "log" "utils")
+
+if (TEGRA_DETECTOR)
+  if (ANDROID_NATIVE_API_LEVEL GREATER 8)
+    add_definitions(-DUSE_TEGRA_HW_DETECTOR)
+    list(APPEND engine_libs ${TEGRA_DETECTOR} GLESv2 EGL)
+  else()
+    message(FATAL_ERROR "Tegra detector required native api level 9 or above")
+  endif()
+endif()
 
 # -D__SUPPORT_ARMEABI_FEATURES key is also available
 add_definitions(-DPLATFORM_ANDROID -D__SUPPORT_ARMEABI_V7A_FEATURES -D__SUPPORT_TEGRA3 -D__SUPPORT_MIPS)
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-rtti -fno-exceptions")
-
 set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -Wl,-allow-shlib-undefined")
-file(GLOB engine_files "jni/BinderComponent/*.cpp" "jni/BinderComponent/*.h" "jni/include/*.h")
 
-include_directories(jni/BinderComponent jni/include)
+include_directories("jni/BinderComponent" "jni/include")
 include_directories(SYSTEM "${ANDROID_SOURCE_TREE}/frameworks/base/include" "${ANDROID_SOURCE_TREE}/system/core/include")
 add_library(${engine} SHARED ${engine_files})
-target_link_libraries(${engine} z binder log utils)
+target_link_libraries(${engine} ${engine_libs})
 
 set_target_properties(${engine} PROPERTIES
     OUTPUT_NAME ${engine}
@@ -51,7 +64,15 @@ add_custom_command(TARGET ${engine} POST_BUILD COMMAND ${CMAKE_STRIP} --strip-un
 file(GLOB engine_jni_files "jni/JNIWrapper/*.cpp" "jni/JNIWrapper/*.h" "jni/include/*.h")
 list(APPEND engine_jni_files jni/NativeService/CommonPackageManager.cpp jni/NativeService/PackageInfo.cpp)
 
-include_directories(jni/include jni/JNIWrapper jni/NativeService jni/BinderComponent "${ANDROID_SOURCE_TREE}/frameworks/base/include" "${ANDROID_SOURCE_TREE}/system/core/include" "${ANDROID_SOURCE_TREE}/frameworks/base/core/jni")
+include_directories(
+  jni/include jni/JNIWrapper
+  jni/NativeService
+  jni/BinderComponent
+  "${ANDROID_SOURCE_TREE}/frameworks/base/include"
+  "${ANDROID_SOURCE_TREE}/system/core/include"
+  "${ANDROID_SOURCE_TREE}/frameworks/base/core/jni"
+  )
+
 add_library(${engine}_jni SHARED ${engine_jni_files})
 target_link_libraries(${engine}_jni z binder log utils android_runtime ${engine})
 
@@ -72,4 +93,3 @@ file(GLOB engine_test_files "jni/Tests/*.cpp")
 
 add_executable(opencv_test_engine ${engine_test_files} jni/Tests/gtest/gtest-all.cpp)
 target_link_libraries(opencv_test_engine z binder log utils android_runtime ${engine} ${engine}_jni)
-
diff --git a/android/service/engine/build.xml b/platforms/android/service/engine/build.xml
similarity index 100%
rename from android/service/engine/build.xml
rename to platforms/android/service/engine/build.xml
diff --git a/android/service/engine/jni/Android.mk b/platforms/android/service/engine/jni/Android.mk
similarity index 98%
rename from android/service/engine/jni/Android.mk
rename to platforms/android/service/engine/jni/Android.mk
index 71d2c464c9..a5c1881109 100644
--- a/android/service/engine/jni/Android.mk
+++ b/platforms/android/service/engine/jni/Android.mk
@@ -83,4 +83,4 @@ include $(BUILD_SHARED_LIBRARY)
 #        Native test application
 #---------------------------------------------------------------------
 
-#include $(LOCAL_PATH)/Tests/Tests.mk
\ No newline at end of file
+#include $(LOCAL_PATH)/Tests/Tests.mk
diff --git a/android/service/engine/jni/Application.mk b/platforms/android/service/engine/jni/Application.mk
similarity index 88%
rename from android/service/engine/jni/Application.mk
rename to platforms/android/service/engine/jni/Application.mk
index 5647110fa2..e530dc1915 100644
--- a/android/service/engine/jni/Application.mk
+++ b/platforms/android/service/engine/jni/Application.mk
@@ -3,4 +3,4 @@ APP_PLATFORM := android-8
 APP_STL := stlport_static
 APP_CPPFLAGS := -fno-rtti -fno-exceptions
 NDK_TOOLCHAIN_VERSION=4.4.3
-#APP_OPTIM := debug
\ No newline at end of file
+#APP_OPTIM := debug
diff --git a/android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp b/platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp
similarity index 99%
rename from android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp
rename to platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp
index 997b8d6ce0..fccb329e90 100644
--- a/android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp
+++ b/platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.cpp
@@ -69,4 +69,4 @@ status_t BnOpenCVEngine::onTransact(uint32_t code, const Parcel& data, android::
     }
 
     return android::NO_ERROR;
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/BinderComponent/BnOpenCVEngine.h b/platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.h
similarity index 98%
rename from android/service/engine/jni/BinderComponent/BnOpenCVEngine.h
rename to platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.h
index bdb88ae1ab..cda2b03e91 100644
--- a/android/service/engine/jni/BinderComponent/BnOpenCVEngine.h
+++ b/platforms/android/service/engine/jni/BinderComponent/BnOpenCVEngine.h
@@ -18,4 +18,4 @@ public:
 
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp b/platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp
similarity index 96%
rename from android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp
rename to platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp
index b1a88ffb66..a9cbc11a21 100644
--- a/android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp
+++ b/platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.cpp
@@ -68,4 +68,4 @@ bool BpOpenCVEngine::InstallVersion(String16 version)
     return static_cast<bool>(reply.readInt32());
 }
 
-IMPLEMENT_META_INTERFACE(OpenCVEngine, OPECV_ENGINE_CLASSNAME)
\ No newline at end of file
+IMPLEMENT_META_INTERFACE(OpenCVEngine, OPECV_ENGINE_CLASSNAME)
diff --git a/android/service/engine/jni/BinderComponent/BpOpenCVEngine.h b/platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.h
similarity index 98%
rename from android/service/engine/jni/BinderComponent/BpOpenCVEngine.h
rename to platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.h
index 01df00eaf6..cb2af532db 100644
--- a/android/service/engine/jni/BinderComponent/BpOpenCVEngine.h
+++ b/platforms/android/service/engine/jni/BinderComponent/BpOpenCVEngine.h
@@ -17,4 +17,4 @@ public:
     virtual bool InstallVersion(android::String16 version);
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/BinderComponent/HardwareDetector.cpp b/platforms/android/service/engine/jni/BinderComponent/HardwareDetector.cpp
similarity index 93%
rename from android/service/engine/jni/BinderComponent/HardwareDetector.cpp
rename to platforms/android/service/engine/jni/BinderComponent/HardwareDetector.cpp
index b5e0fa6001..d0d8514b74 100644
--- a/android/service/engine/jni/BinderComponent/HardwareDetector.cpp
+++ b/platforms/android/service/engine/jni/BinderComponent/HardwareDetector.cpp
@@ -13,7 +13,7 @@ int GetCpuID()
     map<string, string> cpu_info = GetCpuInfo();
     map<string, string>::const_iterator it;
 
-    #if defined(__i386__)
+#if defined(__i386__)
     LOGD("Using X86 HW detector");
     result |= ARCH_X86;
     it = cpu_info.find("flags");
@@ -161,24 +161,18 @@ int GetProcessorCount()
 
 int DetectKnownPlatforms()
 {
+#if defined(__arm__) && defined(USE_TEGRA_HW_DETECTOR)
     int tegra_status = DetectTegra();
-
-    if (3 == tegra_status)
+#else
+    int tegra_status = NOT_TEGRA;
+#endif
+    // All Tegra platforms since Tegra3
+    if (2 < tegra_status)
     {
-        return PLATFORM_TEGRA3;
+        return PLATFORM_TEGRA + tegra_status - 1;
     }
     else
     {
         return PLATFORM_UNKNOWN;
     }
-
-    // NOTE: Uncomment when all Tegras will be supported
-    /*if (tegra_status > 0)
-     *   {
-     *   return PLATFORM_TEGRA + tegra_status - 1;
-     }
-     else
-     {
-         return PLATFORM_UNKNOWN;
-     }*/
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/BinderComponent/HardwareDetector.h b/platforms/android/service/engine/jni/BinderComponent/HardwareDetector.h
similarity index 96%
rename from android/service/engine/jni/BinderComponent/HardwareDetector.h
rename to platforms/android/service/engine/jni/BinderComponent/HardwareDetector.h
index e049db9c5f..1dda8bd14b 100644
--- a/android/service/engine/jni/BinderComponent/HardwareDetector.h
+++ b/platforms/android/service/engine/jni/BinderComponent/HardwareDetector.h
@@ -27,10 +27,11 @@
 #define PLATFORM_TEGRA 1L
 #define PLATFORM_TEGRA2 2L
 #define PLATFORM_TEGRA3 3L
+#define PLATFORM_TEGRA4 4L
 
 int DetectKnownPlatforms();
 int GetProcessorCount();
 std::string GetPlatformName();
 int GetCpuID();
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/BinderComponent/OpenCVEngine.cpp b/platforms/android/service/engine/jni/BinderComponent/OpenCVEngine.cpp
similarity index 98%
rename from android/service/engine/jni/BinderComponent/OpenCVEngine.cpp
rename to platforms/android/service/engine/jni/BinderComponent/OpenCVEngine.cpp
index c10ab54284..b0b2b5137f 100644
--- a/android/service/engine/jni/BinderComponent/OpenCVEngine.cpp
+++ b/platforms/android/service/engine/jni/BinderComponent/OpenCVEngine.cpp
@@ -15,7 +15,7 @@ using namespace android;
 
 const int OpenCVEngine::Platform = DetectKnownPlatforms();
 const int OpenCVEngine::CpuID = GetCpuID();
-const int OpenCVEngine::KnownVersions[] = {2040000, 2040100, 2040200, 2040300, 2040301, 2040302, 2040400};
+const int OpenCVEngine::KnownVersions[] = {2040000, 2040100, 2040200, 2040300, 2040301, 2040302, 2040400, 2040500, 2040600};
 
 bool OpenCVEngine::ValidateVersion(int version)
 {
diff --git a/android/service/engine/jni/BinderComponent/OpenCVEngine.h b/platforms/android/service/engine/jni/BinderComponent/OpenCVEngine.h
similarity index 100%
rename from android/service/engine/jni/BinderComponent/OpenCVEngine.h
rename to platforms/android/service/engine/jni/BinderComponent/OpenCVEngine.h
diff --git a/android/service/engine/jni/BinderComponent/ProcReader.cpp b/platforms/android/service/engine/jni/BinderComponent/ProcReader.cpp
similarity index 100%
rename from android/service/engine/jni/BinderComponent/ProcReader.cpp
rename to platforms/android/service/engine/jni/BinderComponent/ProcReader.cpp
diff --git a/android/service/engine/jni/BinderComponent/ProcReader.h b/platforms/android/service/engine/jni/BinderComponent/ProcReader.h
similarity index 98%
rename from android/service/engine/jni/BinderComponent/ProcReader.h
rename to platforms/android/service/engine/jni/BinderComponent/ProcReader.h
index 333e7c2937..f703ccf9cd 100644
--- a/android/service/engine/jni/BinderComponent/ProcReader.h
+++ b/platforms/android/service/engine/jni/BinderComponent/ProcReader.h
@@ -26,4 +26,4 @@
 // public part
 std::map<std::string, std::string> GetCpuInfo();
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/BinderComponent/StringUtils.cpp b/platforms/android/service/engine/jni/BinderComponent/StringUtils.cpp
similarity index 99%
rename from android/service/engine/jni/BinderComponent/StringUtils.cpp
rename to platforms/android/service/engine/jni/BinderComponent/StringUtils.cpp
index 3809f5aa65..2e6b35a7b1 100644
--- a/android/service/engine/jni/BinderComponent/StringUtils.cpp
+++ b/platforms/android/service/engine/jni/BinderComponent/StringUtils.cpp
@@ -89,4 +89,4 @@ vector<string> SplitStringVector(const string& src, const char seporator)
     }
 
     return result;
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/BinderComponent/StringUtils.h b/platforms/android/service/engine/jni/BinderComponent/StringUtils.h
similarity index 98%
rename from android/service/engine/jni/BinderComponent/StringUtils.h
rename to platforms/android/service/engine/jni/BinderComponent/StringUtils.h
index d1fe36ce77..e36bfcc7c6 100644
--- a/android/service/engine/jni/BinderComponent/StringUtils.h
+++ b/platforms/android/service/engine/jni/BinderComponent/StringUtils.h
@@ -10,4 +10,4 @@ std::set<std::string> SplitString(const std::string& src, const char seporator);
 bool ParseString(const std::string& src, std::string& key, std::string& value);
 std::vector<std::string> SplitStringVector(const std::string& src, const char seporator);
 
-#endif
\ No newline at end of file
+#endif
diff --git a/platforms/android/service/engine/jni/BinderComponent/TegraDetector.h b/platforms/android/service/engine/jni/BinderComponent/TegraDetector.h
new file mode 100644
index 0000000000..4ca930b754
--- /dev/null
+++ b/platforms/android/service/engine/jni/BinderComponent/TegraDetector.h
@@ -0,0 +1,14 @@
+#ifndef __TEGRA_DETECTOR_H__
+#define __TEGRA_DETECTOR_H__
+
+#define TEGRA_DETECTOR_ERROR -2
+#define NOT_TEGRA -1
+#define TEGRA2  2
+#define TEGRA3  3
+#define TEGRA4i 4
+#define TEGRA4  5
+#define TEGRA5  6
+
+int DetectTegra();
+
+#endif
diff --git a/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp b/platforms/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp
similarity index 99%
rename from android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp
rename to platforms/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp
index be30d877b7..3e490a3160 100644
--- a/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp
+++ b/platforms/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.cpp
@@ -22,4 +22,4 @@ JNIEXPORT jint JNICALL Java_org_opencv_engine_HardwareDetector_GetProcessorCount
 JNIEXPORT jint JNICALL Java_org_opencv_engine_HardwareDetector_DetectKnownPlatforms(JNIEnv* , jclass)
 {
     return DetectKnownPlatforms();
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.h b/platforms/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.h
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/HardwareDetector_jni.h
rename to platforms/android/service/engine/jni/JNIWrapper/HardwareDetector_jni.h
diff --git a/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.cpp b/platforms/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.cpp
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.cpp
rename to platforms/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.cpp
diff --git a/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.h b/platforms/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.h
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.h
rename to platforms/android/service/engine/jni/JNIWrapper/JavaBasedPackageManager.h
diff --git a/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.cpp b/platforms/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.cpp
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.cpp
rename to platforms/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.cpp
diff --git a/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.h b/platforms/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.h
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.h
rename to platforms/android/service/engine/jni/JNIWrapper/OpenCVEngine_jni.h
diff --git a/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.cpp b/platforms/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.cpp
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.cpp
rename to platforms/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.cpp
diff --git a/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.h b/platforms/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.h
similarity index 100%
rename from android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.h
rename to platforms/android/service/engine/jni/JNIWrapper/OpenCVLibraryInfo.h
diff --git a/android/service/engine/jni/NativeClient/ClientMain.cpp b/platforms/android/service/engine/jni/NativeClient/ClientMain.cpp
similarity index 99%
rename from android/service/engine/jni/NativeClient/ClientMain.cpp
rename to platforms/android/service/engine/jni/NativeClient/ClientMain.cpp
index 4e9bba2961..e9075dc9ab 100644
--- a/android/service/engine/jni/NativeClient/ClientMain.cpp
+++ b/platforms/android/service/engine/jni/NativeClient/ClientMain.cpp
@@ -39,4 +39,4 @@ int main(int argc, char *argv[])
     printf("OpenCVEngine version %d started", EngineVersion);
 
     return 0;
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/NativeService/CommonPackageManager.cpp b/platforms/android/service/engine/jni/NativeService/CommonPackageManager.cpp
similarity index 98%
rename from android/service/engine/jni/NativeService/CommonPackageManager.cpp
rename to platforms/android/service/engine/jni/NativeService/CommonPackageManager.cpp
index dbcd8ff689..eaa03d4d82 100644
--- a/android/service/engine/jni/NativeService/CommonPackageManager.cpp
+++ b/platforms/android/service/engine/jni/NativeService/CommonPackageManager.cpp
@@ -197,6 +197,7 @@ std::vector<std::pair<int, int> > CommonPackageManager::InitArmRating()
     result.push_back(std::pair<int, int>(PLATFORM_UNKNOWN, ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON));
     result.push_back(std::pair<int, int>(PLATFORM_UNKNOWN, ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_VFPv3d16 | FEATURES_HAS_NEON));
     result.push_back(std::pair<int, int>(PLATFORM_TEGRA3,  ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON));
+    result.push_back(std::pair<int, int>(PLATFORM_TEGRA4,  ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON));
 
     return result;
 }
@@ -218,8 +219,8 @@ std::vector<std::pair<int, int> > CommonPackageManager::InitMipsRating()
 }
 
 const std::vector<std::pair<int, int> > CommonPackageManager::ArchRatings[] = {
-                                           CommonPackageManager::InitArmRating(), 
-                                           CommonPackageManager::InitIntelRating(), 
+                                           CommonPackageManager::InitArmRating(),
+                                           CommonPackageManager::InitIntelRating(),
                                            CommonPackageManager::InitMipsRating()
                                         };
 
diff --git a/android/service/engine/jni/NativeService/CommonPackageManager.h b/platforms/android/service/engine/jni/NativeService/CommonPackageManager.h
similarity index 100%
rename from android/service/engine/jni/NativeService/CommonPackageManager.h
rename to platforms/android/service/engine/jni/NativeService/CommonPackageManager.h
diff --git a/android/service/engine/jni/NativeService/NativePackageManager.cpp b/platforms/android/service/engine/jni/NativeService/NativePackageManager.cpp
similarity index 99%
rename from android/service/engine/jni/NativeService/NativePackageManager.cpp
rename to platforms/android/service/engine/jni/NativeService/NativePackageManager.cpp
index aba13de1e9..87d0ea0c17 100644
--- a/android/service/engine/jni/NativeService/NativePackageManager.cpp
+++ b/platforms/android/service/engine/jni/NativeService/NativePackageManager.cpp
@@ -16,4 +16,4 @@ vector<PackageInfo> NativePackageManager::GetInstalledPackages()
 
 NativePackageManager::~NativePackageManager()
 {
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/NativeService/NativePackageManager.h b/platforms/android/service/engine/jni/NativeService/NativePackageManager.h
similarity index 98%
rename from android/service/engine/jni/NativeService/NativePackageManager.h
rename to platforms/android/service/engine/jni/NativeService/NativePackageManager.h
index c74a083598..329047f0bf 100644
--- a/android/service/engine/jni/NativeService/NativePackageManager.h
+++ b/platforms/android/service/engine/jni/NativeService/NativePackageManager.h
@@ -13,4 +13,4 @@ protected:
     virtual std::vector<PackageInfo> GetInstalledPackages();
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/NativeService/PackageInfo.cpp b/platforms/android/service/engine/jni/NativeService/PackageInfo.cpp
similarity index 97%
rename from android/service/engine/jni/NativeService/PackageInfo.cpp
rename to platforms/android/service/engine/jni/NativeService/PackageInfo.cpp
index 2eb8230739..2f8dde043a 100644
--- a/android/service/engine/jni/NativeService/PackageInfo.cpp
+++ b/platforms/android/service/engine/jni/NativeService/PackageInfo.cpp
@@ -18,6 +18,7 @@ map<int, string> PackageInfo::InitPlatformNameMap()
     result[PLATFORM_TEGRA] = PLATFORM_TEGRA_NAME;
     result[PLATFORM_TEGRA2] = PLATFORM_TEGRA2_NAME;
     result[PLATFORM_TEGRA3] = PLATFORM_TEGRA3_NAME;
+    result[PLATFORM_TEGRA4] = PLATFORM_TEGRA4_NAME;
 
     return result;
 }
@@ -186,6 +187,10 @@ inline int SplitPlatfrom(const vector<string>& features)
         {
             result = PLATFORM_TEGRA3;
         }
+        else if (PLATFORM_TEGRA4_NAME == tmp)
+        {
+            result = PLATFORM_TEGRA4;
+        }
     }
     else
     {
@@ -425,6 +430,10 @@ InstallPath(install_path)
                 {
                     CpuID = ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON;
                 } break;
+                case PLATFORM_TEGRA4:
+                {
+                    CpuID = ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON;
+                } break;
             }
         }
         else
diff --git a/android/service/engine/jni/NativeService/PackageInfo.h b/platforms/android/service/engine/jni/NativeService/PackageInfo.h
similarity index 97%
rename from android/service/engine/jni/NativeService/PackageInfo.h
rename to platforms/android/service/engine/jni/NativeService/PackageInfo.h
index b86ef7a92e..2ce561e2f3 100644
--- a/android/service/engine/jni/NativeService/PackageInfo.h
+++ b/platforms/android/service/engine/jni/NativeService/PackageInfo.h
@@ -12,7 +12,6 @@
 #define ARCH_ARMv7_NAME "armv7a"
 #define ARCH_ARMv8_NAME "armv8"
 
-
 #define FEATURES_HAS_VFPv3d16_NAME "vfpv3d16"
 #define FEATURES_HAS_VFPv3_NAME "vfpv3"
 #define FEATURES_HAS_NEON_NAME "neon"
@@ -25,7 +24,7 @@
 #define PLATFORM_TEGRA_NAME "tegra"
 #define PLATFORM_TEGRA2_NAME "tegra2"
 #define PLATFORM_TEGRA3_NAME "tegra3"
-
+#define PLATFORM_TEGRA4_NAME "tegra4"
 
 class PackageInfo
 {
diff --git a/android/service/engine/jni/NativeService/ServiceMain.cpp b/platforms/android/service/engine/jni/NativeService/ServiceMain.cpp
similarity index 99%
rename from android/service/engine/jni/NativeService/ServiceMain.cpp
rename to platforms/android/service/engine/jni/NativeService/ServiceMain.cpp
index 66ea995980..20cafa6c4f 100644
--- a/android/service/engine/jni/NativeService/ServiceMain.cpp
+++ b/platforms/android/service/engine/jni/NativeService/ServiceMain.cpp
@@ -29,4 +29,4 @@ int main(int argc, char *argv[])
     delete PackageManager;
 
     return 0;
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/Tests/HardwareDetectionTest.cpp b/platforms/android/service/engine/jni/Tests/HardwareDetectionTest.cpp
similarity index 100%
rename from android/service/engine/jni/Tests/HardwareDetectionTest.cpp
rename to platforms/android/service/engine/jni/Tests/HardwareDetectionTest.cpp
diff --git a/android/service/engine/jni/Tests/OpenCVEngineTest.cpp b/platforms/android/service/engine/jni/Tests/OpenCVEngineTest.cpp
similarity index 93%
rename from android/service/engine/jni/Tests/OpenCVEngineTest.cpp
rename to platforms/android/service/engine/jni/Tests/OpenCVEngineTest.cpp
index 4e390386fe..7473387a04 100644
--- a/android/service/engine/jni/Tests/OpenCVEngineTest.cpp
+++ b/platforms/android/service/engine/jni/Tests/OpenCVEngineTest.cpp
@@ -201,6 +201,24 @@ TEST(OpenCVEngineTest, GetPathForCompatiblePackage2)
 #endif
 }
 
+TEST(OpenCVEngineTest, GetPathForCompatiblePackage3)
+{
+    sp<IOpenCVEngine> Engine = InitConnect();
+    Starter.PackageManager->InstalledPackages.clear();
+    Starter.PackageManager->InstallVersion(2040400, PLATFORM_TEGRA4, ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON);
+    EXPECT_FALSE(NULL == Engine.get());
+    String16 result = Engine->GetLibPathByVersion(String16("2.4"));
+    #ifdef __SUPPORT_TEGRA3
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_tegra4/lib", String8(result).string());
+    #else
+    #ifdef __SUPPORT_ARMEABI_V7A_FEATURES
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_armv7a_neon/lib", String8(result).string());
+    #else
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_armv7a/lib", String8(result).string());
+    #endif
+    #endif
+}
+
 TEST(OpenCVEngineTest, InstallAndGetVersion)
 {
     sp<IOpenCVEngine> Engine = InitConnect();
@@ -276,7 +294,7 @@ TEST(OpenCVEngineTest, GetPathFor2_4_5)
     Starter.PackageManager->InstallVersion(2040500, PLATFORM_UNKNOWN, ARCH_ARMv7);
     EXPECT_FALSE(NULL == Engine.get());
     String16 result = Engine->GetLibPathByVersion(String16("2.4.5"));
-    EXPECT_EQ(0, result.size()); // 2.4.5 is not published yet
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_armv7a/lib", String8(result).string());
 }
 #endif
 
diff --git a/android/service/engine/jni/Tests/PackageInfoTest.cpp b/platforms/android/service/engine/jni/Tests/PackageInfoTest.cpp
similarity index 90%
rename from android/service/engine/jni/Tests/PackageInfoTest.cpp
rename to platforms/android/service/engine/jni/Tests/PackageInfoTest.cpp
index 6bc84856ce..36fdae764f 100644
--- a/android/service/engine/jni/Tests/PackageInfoTest.cpp
+++ b/platforms/android/service/engine/jni/Tests/PackageInfoTest.cpp
@@ -85,6 +85,21 @@ TEST(PackageInfo, FullNameTegra3)
     #endif
 }
 
+TEST(PackageInfo, FullNameTegra4)
+{
+    PackageInfo info(2040400, PLATFORM_TEGRA4, ARCH_ARMv7 | FEATURES_HAS_NEON);
+    string name = info.GetFullName();
+    #ifdef __SUPPORT_TEGRA3
+    EXPECT_STREQ("org.opencv.lib_v24_tegra4", name.c_str());
+    #else
+    #ifdef __SUPPORT_ARMEABI_V7A_FEATURES
+    EXPECT_STREQ("org.opencv.lib_v24_armv7a_neon", name.c_str());
+    #else
+    EXPECT_STREQ("org.opencv.lib_v24_armv7a", name.c_str());
+    #endif
+    #endif
+}
+
 TEST(PackageInfo, FullNameX86SSE2)
 {
     PackageInfo info(2030000, PLATFORM_UNKNOWN, ARCH_X86 | FEATURES_HAS_SSE2);
@@ -148,6 +163,13 @@ TEST(PackageInfo, Tegra3FromFullName)
     EXPECT_EQ(PLATFORM_TEGRA3, info.GetPlatform());
 }
 
+TEST(PackageInfo, Tegra4FromFullName)
+{
+    PackageInfo info("org.opencv.lib_v24_tegra4", "/data/data/org.opencv.lib_v24_tegra4");
+    EXPECT_EQ(2040000, info.GetVersion());
+    EXPECT_EQ(PLATFORM_TEGRA4, info.GetPlatform());
+}
+
 #ifdef __SUPPORT_MIPS
 TEST(PackageInfo, MipsFromFullName)
 {
@@ -200,4 +222,3 @@ TEST(PackageInfo, Comparator3)
     EXPECT_EQ(info1, info2);
 }
 #endif
-
diff --git a/android/service/engine/jni/Tests/PackageManagerStub.cpp b/platforms/android/service/engine/jni/Tests/PackageManagerStub.cpp
similarity index 99%
rename from android/service/engine/jni/Tests/PackageManagerStub.cpp
rename to platforms/android/service/engine/jni/Tests/PackageManagerStub.cpp
index 20be63a9b7..e154658629 100644
--- a/android/service/engine/jni/Tests/PackageManagerStub.cpp
+++ b/platforms/android/service/engine/jni/Tests/PackageManagerStub.cpp
@@ -15,4 +15,4 @@ vector<PackageInfo> PackageManagerStub::GetInstalledPackages()
 
 PackageManagerStub::~PackageManagerStub()
 {
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/Tests/PackageManagerStub.h b/platforms/android/service/engine/jni/Tests/PackageManagerStub.h
similarity index 98%
rename from android/service/engine/jni/Tests/PackageManagerStub.h
rename to platforms/android/service/engine/jni/Tests/PackageManagerStub.h
index a37e6b0559..8d1d3a80b5 100644
--- a/android/service/engine/jni/Tests/PackageManagerStub.h
+++ b/platforms/android/service/engine/jni/Tests/PackageManagerStub.h
@@ -14,4 +14,4 @@ protected:
     virtual std::vector<PackageInfo> GetInstalledPackages();
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/Tests/PackageManagmentTest.cpp b/platforms/android/service/engine/jni/Tests/PackageManagmentTest.cpp
similarity index 87%
rename from android/service/engine/jni/Tests/PackageManagmentTest.cpp
rename to platforms/android/service/engine/jni/Tests/PackageManagmentTest.cpp
index f9ccb74844..61d6e01c24 100644
--- a/android/service/engine/jni/Tests/PackageManagmentTest.cpp
+++ b/platforms/android/service/engine/jni/Tests/PackageManagmentTest.cpp
@@ -102,6 +102,22 @@ TEST(PackageManager, GetPackagePathForTegra3)
 #endif
 }
 
+TEST(PackageManager, GetPackagePathForTegra4)
+{
+    PackageManagerStub pm;
+    EXPECT_TRUE(pm.InstallVersion(2040400, PLATFORM_TEGRA4, ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON));
+    string path = pm.GetPackagePathByVersion(2040400, PLATFORM_TEGRA4, ARCH_ARMv7 | FEATURES_HAS_VFPv3 | FEATURES_HAS_NEON);
+    #ifdef __SUPPORT_TEGRA3
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_tegra4/lib", path.c_str());
+    #else
+    #ifdef __SUPPORT_ARMEABI_V7A_FEATURES
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_armv7a_neon/lib", path.c_str());
+    #else
+    EXPECT_STREQ("/data/data/org.opencv.lib_v24_armv7a/lib", path.c_str());
+    #endif
+    #endif
+}
+
 #ifdef __SUPPORT_MIPS
 TEST(PackageManager, GetPackagePathForMips)
 {
@@ -121,5 +137,3 @@ TEST(PackageManager, GetPackagePathForMips)
 //     string path = pm.GetPackagePathByVersion("240", PLATFORM_TEGRA2, 0);
 //     EXPECT_STREQ("/data/data/org.opencv.lib_v24_tegra2/lib", path.c_str());
 // }
-
-
diff --git a/android/service/engine/jni/Tests/TestMain.cpp b/platforms/android/service/engine/jni/Tests/TestMain.cpp
similarity index 98%
rename from android/service/engine/jni/Tests/TestMain.cpp
rename to platforms/android/service/engine/jni/Tests/TestMain.cpp
index 4c994ef447..96d7200ccd 100644
--- a/android/service/engine/jni/Tests/TestMain.cpp
+++ b/platforms/android/service/engine/jni/Tests/TestMain.cpp
@@ -4,4 +4,4 @@ int main(int argc, char **argv)
 {
     ::testing::InitGoogleTest(&argc, argv);
     return RUN_ALL_TESTS();
-}
\ No newline at end of file
+}
diff --git a/android/service/engine/jni/Tests/Tests.mk b/platforms/android/service/engine/jni/Tests/Tests.mk
similarity index 98%
rename from android/service/engine/jni/Tests/Tests.mk
rename to platforms/android/service/engine/jni/Tests/Tests.mk
index 4cfe1c16a3..5b46bde529 100644
--- a/android/service/engine/jni/Tests/Tests.mk
+++ b/platforms/android/service/engine/jni/Tests/Tests.mk
@@ -44,4 +44,4 @@ LOCAL_LDLIBS += -lz -lbinder -llog -lutils
 
 LOCAL_SHARED_LIBRARIES += libOpenCVEngine
 
-include $(BUILD_EXECUTABLE)
\ No newline at end of file
+include $(BUILD_EXECUTABLE)
diff --git a/android/service/engine/jni/Tests/gtest/gtest-all.cpp b/platforms/android/service/engine/jni/Tests/gtest/gtest-all.cpp
similarity index 100%
rename from android/service/engine/jni/Tests/gtest/gtest-all.cpp
rename to platforms/android/service/engine/jni/Tests/gtest/gtest-all.cpp
diff --git a/android/service/engine/jni/Tests/gtest/gtest.h b/platforms/android/service/engine/jni/Tests/gtest/gtest.h
similarity index 100%
rename from android/service/engine/jni/Tests/gtest/gtest.h
rename to platforms/android/service/engine/jni/Tests/gtest/gtest.h
diff --git a/android/service/engine/jni/include/EngineCommon.h b/platforms/android/service/engine/jni/include/EngineCommon.h
similarity index 98%
rename from android/service/engine/jni/include/EngineCommon.h
rename to platforms/android/service/engine/jni/include/EngineCommon.h
index 2bee139ea3..a03f02c68d 100644
--- a/android/service/engine/jni/include/EngineCommon.h
+++ b/platforms/android/service/engine/jni/include/EngineCommon.h
@@ -19,4 +19,4 @@
 
 typedef const char* (*InfoFunctionType)();
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/include/IOpenCVEngine.h b/platforms/android/service/engine/jni/include/IOpenCVEngine.h
similarity index 99%
rename from android/service/engine/jni/include/IOpenCVEngine.h
rename to platforms/android/service/engine/jni/include/IOpenCVEngine.h
index 4dbb251c56..735d3746bb 100644
--- a/android/service/engine/jni/include/IOpenCVEngine.h
+++ b/platforms/android/service/engine/jni/include/IOpenCVEngine.h
@@ -29,4 +29,4 @@ public:
     virtual bool InstallVersion(android::String16 version) = 0;
 };
 
-#endif
\ No newline at end of file
+#endif
diff --git a/android/service/engine/jni/include/IPackageManager.h b/platforms/android/service/engine/jni/include/IPackageManager.h
similarity index 100%
rename from android/service/engine/jni/include/IPackageManager.h
rename to platforms/android/service/engine/jni/include/IPackageManager.h
diff --git a/android/service/engine/jni/include/OpenCVEngineHelper.h b/platforms/android/service/engine/jni/include/OpenCVEngineHelper.h
similarity index 100%
rename from android/service/engine/jni/include/OpenCVEngineHelper.h
rename to platforms/android/service/engine/jni/include/OpenCVEngineHelper.h
diff --git a/android/service/engine_test/project.properties b/platforms/android/service/engine/project.properties
similarity index 100%
rename from android/service/engine_test/project.properties
rename to platforms/android/service/engine/project.properties
diff --git a/android/service/engine/res/drawable/icon.png b/platforms/android/service/engine/res/drawable/icon.png
similarity index 100%
rename from android/service/engine/res/drawable/icon.png
rename to platforms/android/service/engine/res/drawable/icon.png
diff --git a/android/service/engine/res/layout-small/info.xml b/platforms/android/service/engine/res/layout-small/info.xml
similarity index 100%
rename from android/service/engine/res/layout-small/info.xml
rename to platforms/android/service/engine/res/layout-small/info.xml
diff --git a/android/service/engine/res/layout-small/main.xml b/platforms/android/service/engine/res/layout-small/main.xml
similarity index 100%
rename from android/service/engine/res/layout-small/main.xml
rename to platforms/android/service/engine/res/layout-small/main.xml
diff --git a/android/service/engine/res/layout/info.xml b/platforms/android/service/engine/res/layout/info.xml
similarity index 100%
rename from android/service/engine/res/layout/info.xml
rename to platforms/android/service/engine/res/layout/info.xml
diff --git a/android/service/engine/res/layout/main.xml b/platforms/android/service/engine/res/layout/main.xml
similarity index 100%
rename from android/service/engine/res/layout/main.xml
rename to platforms/android/service/engine/res/layout/main.xml
diff --git a/android/service/engine/res/values/strings.xml b/platforms/android/service/engine/res/values/strings.xml
similarity index 100%
rename from android/service/engine/res/values/strings.xml
rename to platforms/android/service/engine/res/values/strings.xml
diff --git a/platforms/android/service/engine/src/org/opencv/engine/BinderConnector.java b/platforms/android/service/engine/src/org/opencv/engine/BinderConnector.java
new file mode 100644
index 0000000000..bde54d5b96
--- /dev/null
+++ b/platforms/android/service/engine/src/org/opencv/engine/BinderConnector.java
@@ -0,0 +1,46 @@
+package org.opencv.engine;
+
+import android.os.IBinder;
+
+public class BinderConnector
+{
+    public BinderConnector(MarketConnector Market) {
+        mMarket = Market;
+    }
+
+    public boolean Init() {
+        boolean result = false;
+        if (mIsReady)
+            result = Init(mMarket);
+
+        return result;
+    }
+
+    public native IBinder Connect();
+
+    public boolean Disconnect()
+    {
+        if (mIsReady)
+            Final();
+
+        return mIsReady;
+    }
+
+    private native boolean Init(MarketConnector Market);
+    private native void Final();
+    private static boolean mIsReady = false;
+    private MarketConnector mMarket;
+
+    static {
+        try {
+            System.loadLibrary("OpenCVEngine");
+            System.loadLibrary("OpenCVEngine_jni");
+            mIsReady = true;
+        }
+        catch(UnsatisfiedLinkError e) {
+            mIsReady = false;
+            e.printStackTrace();
+        }
+    }
+
+}
diff --git a/android/service/engine/src/org/opencv/engine/HardwareDetector.java b/platforms/android/service/engine/src/org/opencv/engine/HardwareDetector.java
similarity index 70%
rename from android/service/engine/src/org/opencv/engine/HardwareDetector.java
rename to platforms/android/service/engine/src/org/opencv/engine/HardwareDetector.java
index 7e2a6135f9..dc82ec30cc 100644
--- a/android/service/engine/src/org/opencv/engine/HardwareDetector.java
+++ b/platforms/android/service/engine/src/org/opencv/engine/HardwareDetector.java
@@ -30,9 +30,12 @@ public class HardwareDetector
     // GPU Acceleration options
     public static final int FEATURES_HAS_GPU = 0x010000;
 
-    public static final int PLATFORM_TEGRA  = 1;
-    public static final int PLATFORM_TEGRA2 = 2;
-    public static final int PLATFORM_TEGRA3 = 3;
+    public static final int PLATFORM_TEGRA   = 1;
+    public static final int PLATFORM_TEGRA2  = 2;
+    public static final int PLATFORM_TEGRA3  = 3;
+    public static final int PLATFORM_TEGRA4i = 4;
+    public static final int PLATFORM_TEGRA4  = 5;
+    public static final int PLATFORM_TEGRA5  = 6;
 
     public static final int PLATFORM_UNKNOWN = 0;
 
@@ -45,9 +48,17 @@ public class HardwareDetector
 
     public static native int DetectKnownPlatforms();
 
-    static
-    {
-        System.loadLibrary("OpenCVEngine");
-        System.loadLibrary("OpenCVEngine_jni");
+    public static boolean mIsReady = false;
+
+    static {
+        try {
+            System.loadLibrary("OpenCVEngine");
+            System.loadLibrary("OpenCVEngine_jni");
+            mIsReady = true;
+        }
+        catch(UnsatisfiedLinkError e) {
+            mIsReady = false;
+            e.printStackTrace();
+        }
     }
 }
diff --git a/android/service/engine/src/org/opencv/engine/MarketConnector.java b/platforms/android/service/engine/src/org/opencv/engine/MarketConnector.java
similarity index 100%
rename from android/service/engine/src/org/opencv/engine/MarketConnector.java
rename to platforms/android/service/engine/src/org/opencv/engine/MarketConnector.java
diff --git a/android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl b/platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl
similarity index 99%
rename from android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl
rename to platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl
index 91de3940c3..a6cf193e30 100644
--- a/android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl
+++ b/platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineInterface.aidl
@@ -30,4 +30,4 @@ interface OpenCVEngineInterface
     * @return Returns OpenCV libraries names seporated by symbol ";" in loading order
     */
     String getLibraryList(String version);
-}
\ No newline at end of file
+}
diff --git a/platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java b/platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java
new file mode 100644
index 0000000000..b3c4ea0575
--- /dev/null
+++ b/platforms/android/service/engine/src/org/opencv/engine/OpenCVEngineService.java
@@ -0,0 +1,73 @@
+package org.opencv.engine;
+
+import android.app.Service;
+import android.content.Intent;
+import android.os.IBinder;
+import android.os.RemoteException;
+import android.util.Log;
+
+public class OpenCVEngineService extends Service
+{
+    private static final String TAG = "OpenCVEngine/Service";
+    private IBinder mEngineInterface = null;
+    private MarketConnector mMarket;
+    private BinderConnector mNativeBinder;
+
+    public void onCreate() {
+        Log.i(TAG, "Service starting");
+        super.onCreate();
+        Log.i(TAG, "Engine binder component creating");
+        mMarket = new MarketConnector(getBaseContext());
+        mNativeBinder = new BinderConnector(mMarket);
+        if (mNativeBinder.Init()) {
+            mEngineInterface = mNativeBinder.Connect();
+            Log.i(TAG, "Service started successfully");
+        } else {
+            Log.e(TAG, "Cannot initialize native part of OpenCV Manager!");
+            Log.e(TAG, "Using stub instead");
+
+            mEngineInterface = new OpenCVEngineInterface.Stub() {
+
+                @Override
+                public boolean installVersion(String version) throws RemoteException {
+                    // TODO Auto-generated method stub
+                    return false;
+                }
+
+                @Override
+                public String getLibraryList(String version) throws RemoteException {
+                    // TODO Auto-generated method stub
+                    return null;
+                }
+
+                @Override
+                public String getLibPathByVersion(String version) throws RemoteException {
+                    // TODO Auto-generated method stub
+                    return null;
+                }
+
+                @Override
+                public int getEngineVersion() throws RemoteException {
+                    return -1;
+                }
+            };
+        }
+    }
+
+    public IBinder onBind(Intent intent) {
+        Log.i(TAG, "Service onBind called for intent " + intent.toString());
+        return mEngineInterface;
+    }
+
+    public boolean onUnbind(Intent intent)
+    {
+        Log.i(TAG, "Service onUnbind called for intent " + intent.toString());
+        return true;
+    }
+    public void OnDestroy()
+    {
+        Log.i(TAG, "OpenCV Engine service destruction");
+        mNativeBinder.Disconnect();
+    }
+
+}
diff --git a/android/service/engine/src/org/opencv/engine/OpenCVLibraryInfo.java b/platforms/android/service/engine/src/org/opencv/engine/OpenCVLibraryInfo.java
similarity index 100%
rename from android/service/engine/src/org/opencv/engine/OpenCVLibraryInfo.java
rename to platforms/android/service/engine/src/org/opencv/engine/OpenCVLibraryInfo.java
diff --git a/android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java b/platforms/android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java
similarity index 86%
rename from android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java
rename to platforms/android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java
index fad2797727..e22f7b5298 100644
--- a/android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java
+++ b/platforms/android/service/engine/src/org/opencv/engine/manager/ManagerActivity.java
@@ -42,6 +42,26 @@ public class ManagerActivity extends Activity
     @Override
     public void onCreate(Bundle savedInstanceState) {
         super.onCreate(savedInstanceState);
+
+        if (!HardwareDetector.mIsReady) {
+            Log.e(TAG, "Cannot initialize native part of OpenCV Manager!");
+
+            AlertDialog dialog = new AlertDialog.Builder(this).create();
+
+            dialog.setTitle("OpenCV Manager Error");
+            dialog.setMessage("OpenCV Manager is incompatible with this device. Please replace it with an appropriate package.");
+            dialog.setCancelable(false);
+            dialog.setButton("OK", new DialogInterface.OnClickListener() {
+
+                public void onClick(DialogInterface dialog, int which) {
+                    finish();
+                }
+            });
+
+            dialog.show();
+            return;
+        }
+
         setContentView(R.layout.main);
 
         TextView OsVersionView = (TextView)findViewById(R.id.OsVersionValue);
@@ -83,10 +103,18 @@ public class ManagerActivity extends Activity
             {
                 HardwarePlatformView.setText("Tegra 2");
             }
-            else
+            else if (HardwareDetector.PLATFORM_TEGRA3 == Platfrom)
             {
                 HardwarePlatformView.setText("Tegra 3");
             }
+            else if (HardwareDetector.PLATFORM_TEGRA4i == Platfrom)
+            {
+                HardwarePlatformView.setText("Tegra 4i");
+            }
+            else
+            {
+                HardwarePlatformView.setText("Tegra 4");
+            }
         }
         else
         {
@@ -182,6 +210,20 @@ public class ManagerActivity extends Activity
             }
         });
 
+        mPackageChangeReciever = new BroadcastReceiver() {
+
+            @Override
+            public void onReceive(Context context, Intent intent) {
+                Log.d("OpenCVManager/Reciever", "Bradcast message " + intent.getAction() + " reciever");
+                Log.d("OpenCVManager/Reciever", "Filling package list on broadcast message");
+                if (!bindService(new Intent("org.opencv.engine.BIND"), new OpenCVEngineServiceConnection(), Context.BIND_AUTO_CREATE))
+                {
+                    TextView EngineVersionView = (TextView)findViewById(R.id.EngineVersionValue);
+                    EngineVersionView.setText("not avaliable");
+                }
+            }
+        };
+
         IntentFilter filter = new IntentFilter();
         filter.addAction(Intent.ACTION_PACKAGE_ADDED);
         filter.addAction(Intent.ACTION_PACKAGE_CHANGED);
@@ -195,17 +237,23 @@ public class ManagerActivity extends Activity
     @Override
     protected void onDestroy() {
         super.onDestroy();
-        unregisterReceiver(mPackageChangeReciever);
+        if (mPackageChangeReciever != null)
+            unregisterReceiver(mPackageChangeReciever);
     }
 
     @Override
     protected void onResume() {
         super.onResume();
-        Log.d(TAG, "Filling package list on resume");
-        if (!bindService(new Intent("org.opencv.engine.BIND"), new OpenCVEngineServiceConnection(), Context.BIND_AUTO_CREATE))
-        {
-            TextView EngineVersionView = (TextView)findViewById(R.id.EngineVersionValue);
-            EngineVersionView.setText("not avaliable");
+        if (HardwareDetector.mIsReady) {
+            Log.d(TAG, "Filling package list on resume");
+            OpenCVEngineServiceConnection connection = new OpenCVEngineServiceConnection();
+            if (!bindService(new Intent("org.opencv.engine.BIND"), connection, Context.BIND_AUTO_CREATE)) {
+                Log.e(TAG, "Cannot bind to OpenCV Manager service!");
+                TextView EngineVersionView = (TextView)findViewById(R.id.EngineVersionValue);
+                if (EngineVersionView != null)
+                    EngineVersionView.setText("not avaliable");
+                unbindService(connection);
+            }
         }
     }
 
@@ -221,19 +269,7 @@ public class ManagerActivity extends Activity
     protected int ManagerApiLevel = 0;
     protected String ManagerVersion;
 
-    protected BroadcastReceiver mPackageChangeReciever = new BroadcastReceiver() {
-
-        @Override
-        public void onReceive(Context context, Intent intent) {
-            Log.d("OpenCVManager/Reciever", "Bradcast message " + intent.getAction() + " reciever");
-            Log.d("OpenCVManager/Reciever", "Filling package list on broadcast message");
-            if (!bindService(new Intent("org.opencv.engine.BIND"), new OpenCVEngineServiceConnection(), Context.BIND_AUTO_CREATE))
-            {
-                TextView EngineVersionView = (TextView)findViewById(R.id.EngineVersionValue);
-                EngineVersionView.setText("not avaliable");
-            }
-        }
-    };
+    protected BroadcastReceiver mPackageChangeReciever = null;
 
     protected class OpenCVEngineServiceConnection implements ServiceConnection
     {
@@ -242,6 +278,12 @@ public class ManagerActivity extends Activity
 
         public void onServiceConnected(ComponentName name, IBinder service) {
             OpenCVEngineInterface EngineService = OpenCVEngineInterface.Stub.asInterface(service);
+            if (EngineService == null) {
+                Log.e(TAG, "Cannot connect to OpenCV Manager Service!");
+                unbindService(this);
+                return;
+            }
+
             try {
                 ManagerApiLevel = EngineService.getEngineVersion();
             } catch (RemoteException e) {
@@ -367,10 +409,10 @@ public class ManagerActivity extends Activity
                 temp.put("Version", NormalizeVersion(OpenCVersion, VersionName));
                 // HACK: OpenCV Manager for Armv7-a Neon already has Tegra3 optimizations
                 // that is enabled on proper hardware
-                if (HardwareDetector.DetectKnownPlatforms() == HardwareDetector.PLATFORM_TEGRA3 &&
+                if (HardwareDetector.DetectKnownPlatforms() >= HardwareDetector.PLATFORM_TEGRA3 &&
                   HardwareName.equals("armv7a neon ") &&  Build.VERSION.SDK_INT >= Build.VERSION_CODES.GINGERBREAD)
                 {
-                    temp.put("Hardware", "Tegra 3");
+                    temp.put("Hardware", "Tegra");
                     if (Tags == null)
                     {
                         Tags = "optimized";
diff --git a/android/service/engine/src/org/opencv/engine/manager/PackageListAdapter.java b/platforms/android/service/engine/src/org/opencv/engine/manager/PackageListAdapter.java
similarity index 100%
rename from android/service/engine/src/org/opencv/engine/manager/PackageListAdapter.java
rename to platforms/android/service/engine/src/org/opencv/engine/manager/PackageListAdapter.java
diff --git a/android/service/engine_test/.classpath b/platforms/android/service/engine_test/.classpath
similarity index 100%
rename from android/service/engine_test/.classpath
rename to platforms/android/service/engine_test/.classpath
diff --git a/android/service/engine_test/.project b/platforms/android/service/engine_test/.project
similarity index 100%
rename from android/service/engine_test/.project
rename to platforms/android/service/engine_test/.project
diff --git a/android/service/engine_test/AndroidManifest.xml b/platforms/android/service/engine_test/AndroidManifest.xml
similarity index 100%
rename from android/service/engine_test/AndroidManifest.xml
rename to platforms/android/service/engine_test/AndroidManifest.xml
diff --git a/android/service/engine_test/build.xml b/platforms/android/service/engine_test/build.xml
similarity index 100%
rename from android/service/engine_test/build.xml
rename to platforms/android/service/engine_test/build.xml
diff --git a/android/service/engine/project.properties b/platforms/android/service/engine_test/project.properties
similarity index 96%
rename from android/service/engine/project.properties
rename to platforms/android/service/engine_test/project.properties
index c6998b3d10..85aac54016 100644
--- a/android/service/engine/project.properties
+++ b/platforms/android/service/engine_test/project.properties
@@ -11,4 +11,4 @@
 #proguard.config=${sdk.dir}/tools/proguard/proguard-android.txt:proguard-project.txt
 
 # Project target.
-target=android-9
+target=android-8
diff --git a/android/service/engine_test/res/drawable-hdpi/ic_launcher.png b/platforms/android/service/engine_test/res/drawable-hdpi/ic_launcher.png
similarity index 100%
rename from android/service/engine_test/res/drawable-hdpi/ic_launcher.png
rename to platforms/android/service/engine_test/res/drawable-hdpi/ic_launcher.png
diff --git a/android/service/engine_test/res/drawable-ldpi/ic_launcher.png b/platforms/android/service/engine_test/res/drawable-ldpi/ic_launcher.png
similarity index 100%
rename from android/service/engine_test/res/drawable-ldpi/ic_launcher.png
rename to platforms/android/service/engine_test/res/drawable-ldpi/ic_launcher.png
diff --git a/android/service/engine_test/res/drawable-mdpi/ic_launcher.png b/platforms/android/service/engine_test/res/drawable-mdpi/ic_launcher.png
similarity index 100%
rename from android/service/engine_test/res/drawable-mdpi/ic_launcher.png
rename to platforms/android/service/engine_test/res/drawable-mdpi/ic_launcher.png
diff --git a/android/service/engine_test/res/layout/main.xml b/platforms/android/service/engine_test/res/layout/main.xml
similarity index 100%
rename from android/service/engine_test/res/layout/main.xml
rename to platforms/android/service/engine_test/res/layout/main.xml
diff --git a/android/service/engine_test/res/values/strings.xml b/platforms/android/service/engine_test/res/values/strings.xml
similarity index 100%
rename from android/service/engine_test/res/values/strings.xml
rename to platforms/android/service/engine_test/res/values/strings.xml
diff --git a/android/service/engine_test/src/org/opencv/engine/test/EngineInterfaceTest.java b/platforms/android/service/engine_test/src/org/opencv/engine/test/EngineInterfaceTest.java
similarity index 100%
rename from android/service/engine_test/src/org/opencv/engine/test/EngineInterfaceTest.java
rename to platforms/android/service/engine_test/src/org/opencv/engine/test/EngineInterfaceTest.java
diff --git a/android/service/push_native.py b/platforms/android/service/push_native.py
similarity index 95%
rename from android/service/push_native.py
rename to platforms/android/service/push_native.py
index 408fefc285..cc14a2fa10 100755
--- a/android/service/push_native.py
+++ b/platforms/android/service/push_native.py
@@ -24,4 +24,4 @@ if (__name__ ==  "__main__"):
     os.system("adb %s push ./engine/libs/%s/OpenCVEngineNativeService %s" % (DEVICE_STR, DEVICE_ARCH, TARGET_DEVICE_PATH))
     os.system("adb %s push ./engine/libs/%s/OpenCVEngineTest %s" % (DEVICE_STR, DEVICE_ARCH, TARGET_DEVICE_PATH))
     os.system("adb %s push ./engine/libs/%s/OpenCVEngineTestApp %s" % (DEVICE_STR, DEVICE_ARCH, TARGET_DEVICE_PATH))
-    os.system("adb %s push ./engine/libs/%s/libOpenCVEngine_jni.so %s" % (DEVICE_STR, DEVICE_ARCH, TARGET_DEVICE_PATH))
\ No newline at end of file
+    os.system("adb %s push ./engine/libs/%s/libOpenCVEngine_jni.so %s" % (DEVICE_STR, DEVICE_ARCH, TARGET_DEVICE_PATH))
diff --git a/android/service/readme.txt b/platforms/android/service/readme.txt
similarity index 74%
rename from android/service/readme.txt
rename to platforms/android/service/readme.txt
index df17c18245..6255df5e93 100644
--- a/android/service/readme.txt
+++ b/platforms/android/service/readme.txt
@@ -14,20 +14,20 @@ manually using adb tool:
 
 .. code-block:: sh
 
-    adb install OpenCV-2.4.4-android-sdk/apk/OpenCV_2.4.4_Manager_2.6_<platform>.apk
+    adb install OpenCV-2.4.6-android-sdk/apk/OpenCV_2.4.6_Manager_2.9_<platform>.apk
 
 Use the table below to determine proper OpenCV Manager package for your device:
 
 +------------------------------+--------------+---------------------------------------------------+
 | Hardware Platform            | Android ver. | Package name                                      |
 +==============================+==============+===================================================+
-| armeabi-v7a (ARMv7-A + NEON) |    >= 2.3    | OpenCV_2.4.4_Manager_2.6_armv7a-neon.apk          |
+| armeabi-v7a (ARMv7-A + NEON) |    >= 2.3    | OpenCV_2.4.6_Manager_2.9_armv7a-neon.apk          |
 +------------------------------+--------------+---------------------------------------------------+
-| armeabi-v7a (ARMv7-A + NEON) |     = 2.2    | OpenCV_2.4.4_Manager_2.6_armv7a-neon-android8.apk |
+| armeabi-v7a (ARMv7-A + NEON) |     = 2.2    | OpenCV_2.4.6_Manager_2.9_armv7a-neon-android8.apk |
 +------------------------------+--------------+---------------------------------------------------+
-| armeabi (ARMv5, ARMv6)       |    >= 2.3    | OpenCV_2.4.4_Manager_2.6_armeabi.apk              |
+| armeabi (ARMv5, ARMv6)       |    >= 2.3    | OpenCV_2.4.6_Manager_2.9_armeabi.apk              |
 +------------------------------+--------------+---------------------------------------------------+
-| Intel x86                    |    >= 2.3    | OpenCV_2.4.4_Manager_2.6_x86.apk                  |
+| Intel x86                    |    >= 2.3    | OpenCV_2.4.6_Manager_2.9_x86.apk                  |
 +------------------------------+--------------+---------------------------------------------------+
-| MIPS                         |    >= 2.3    | OpenCV_2.4.4_Manager_2.6_mips.apk                 |
+| MIPS                         |    >= 2.3    | OpenCV_2.4.6_Manager_2.9_mips.apk                 |
 +------------------------------+--------------+---------------------------------------------------+
diff --git a/android/service/test_native.py b/platforms/android/service/test_native.py
similarity index 99%
rename from android/service/test_native.py
rename to platforms/android/service/test_native.py
index 9a39032b18..328b9a8a51 100755
--- a/android/service/test_native.py
+++ b/platforms/android/service/test_native.py
@@ -34,4 +34,3 @@ if (__name__ ==  "__main__"):
     os.system("adb %s shell mkdir -p \"%s\"" % (DEVICE_STR, DEVICE_LOG_PATH))
 
     RunTestApp("OpenCVEngineTestApp")
-
diff --git a/ios/Info.plist.in b/platforms/ios/Info.plist.in
similarity index 80%
rename from ios/Info.plist.in
rename to platforms/ios/Info.plist.in
index 89ef38625d..b2a3baf524 100644
--- a/ios/Info.plist.in
+++ b/platforms/ios/Info.plist.in
@@ -5,11 +5,11 @@
     <key>CFBundleName</key>
     <string>OpenCV</string>
     <key>CFBundleIdentifier</key>
-    <string>com.itseez.opencv</string>
+    <string>org.opencv</string>
     <key>CFBundleVersion</key>
-    <string>${VERSION}</string>
+    <string>${OPENCV_LIBVERSION}</string>
     <key>CFBundleShortVersionString</key>
-    <string>${VERSION}</string>
+    <string>${OPENCV_LIBVERSION}</string>
     <key>CFBundleSignature</key>
     <string>????</string>
     <key>CFBundlePackageType</key>
diff --git a/ios/build_framework.py b/platforms/ios/build_framework.py
similarity index 81%
rename from ios/build_framework.py
rename to platforms/ios/build_framework.py
index 3f0e85f725..cb3788f7d5 100755
--- a/ios/build_framework.py
+++ b/platforms/ios/build_framework.py
@@ -38,7 +38,7 @@ def build_opencv(srcroot, buildroot, target, arch):
     # for some reason, if you do not specify CMAKE_BUILD_TYPE, it puts libs to "RELEASE" rather than "Release"
     cmakeargs = ("-GXcode " +
                 "-DCMAKE_BUILD_TYPE=Release " +
-                "-DCMAKE_TOOLCHAIN_FILE=%s/ios/cmake/Toolchains/Toolchain-%s_Xcode.cmake " +
+                "-DCMAKE_TOOLCHAIN_FILE=%s/platforms/ios/cmake/Toolchains/Toolchain-%s_Xcode.cmake " +
                 "-DBUILD_opencv_world=ON " +
                 "-DCMAKE_INSTALL_PREFIX=install") % (srcroot, target)
     # if cmake cache exists, just rerun cmake to update OpenCV.xproj if necessary
@@ -71,19 +71,11 @@ def put_framework_together(srcroot, dstroot):
     os.makedirs(framework_dir)
     os.chdir(framework_dir)
 
-    # determine OpenCV version (without subminor part)
-    tdir0 = "../build/" + targetlist[0]
-    cfg = open(tdir0 + "/cvconfig.h", "rt")
-    for l in cfg.readlines():
-        if l.startswith("#define  VERSION"):
-            opencv_version = l[l.find("\"")+1:l.rfind(".")]
-            break
-    cfg.close()
-
     # form the directory tree
     dstdir = "Versions/A"
     os.makedirs(dstdir + "/Resources")
 
+    tdir0 = "../build/" + targetlist[0]
     # copy headers
     shutil.copytree(tdir0 + "/install/include/opencv2", dstdir + "/Headers")
 
@@ -91,22 +83,14 @@ def put_framework_together(srcroot, dstroot):
     wlist = " ".join(["../build/" + t + "/lib/Release/libopencv_world.a" for t in targetlist])
     os.system("lipo -create " + wlist + " -o " + dstdir + "/opencv2")
 
-    # form Info.plist
-    srcfile = open(srcroot + "/ios/Info.plist.in", "rt")
-    dstfile = open(dstdir + "/Resources/Info.plist", "wt")
-    for l in srcfile.readlines():
-        dstfile.write(l.replace("${VERSION}", opencv_version))
-    srcfile.close()
-    dstfile.close()
-
-    # copy cascades
-    # TODO ...
+    # copy Info.plist
+    shutil.copyfile(tdir0 + "/ios/Info.plist", dstdir + "/Resources/Info.plist")
 
     # make symbolic links
-    os.symlink(dstdir + "/Headers", "Headers")
-    os.symlink(dstdir + "/Resources", "Resources")
-    os.symlink(dstdir + "/opencv2", "opencv2")
     os.symlink("A", "Versions/Current")
+    os.symlink("Versions/Current/Headers", "Headers")
+    os.symlink("Versions/Current/Resources", "Resources")
+    os.symlink("Versions/Current/opencv2", "opencv2")
 
 
 def build_framework(srcroot, dstroot):
@@ -125,4 +109,4 @@ if __name__ == "__main__":
         print "Usage:\n\t./build_framework.py <outputdir>\n\n"
         sys.exit(0)
 
-    build_framework(os.path.abspath(os.path.join(os.path.dirname(sys.argv[0]), "..")), os.path.abspath(sys.argv[1]))
\ No newline at end of file
+    build_framework(os.path.abspath(os.path.join(os.path.dirname(sys.argv[0]), "../..")), os.path.abspath(sys.argv[1]))
diff --git a/ios/cmake/Modules/Platform/iOS.cmake b/platforms/ios/cmake/Modules/Platform/iOS.cmake
similarity index 99%
rename from ios/cmake/Modules/Platform/iOS.cmake
rename to platforms/ios/cmake/Modules/Platform/iOS.cmake
index 4d196e65d9..e021aca55f 100644
--- a/ios/cmake/Modules/Platform/iOS.cmake
+++ b/platforms/ios/cmake/Modules/Platform/iOS.cmake
@@ -160,4 +160,4 @@ list (APPEND CMAKE_SYSTEM_PREFIX_PATH
 
     # Project install destination.
     "${CMAKE_INSTALL_PREFIX}"
-)
\ No newline at end of file
+)
diff --git a/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake b/platforms/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake
similarity index 80%
rename from ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake
rename to platforms/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake
index 67343253bd..14555cd6a9 100644
--- a/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake
+++ b/platforms/ios/cmake/Toolchains/Toolchain-iPhoneOS_Xcode.cmake
@@ -4,12 +4,12 @@ set (IPHONEOS TRUE)
 # Standard settings
 set (CMAKE_SYSTEM_NAME iOS)
 # Include extra modules for the iOS platform files
-set (CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/ios/cmake/Modules")
+set (CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/platforms/ios/cmake/Modules")
 
-# Force the compilers to gcc for iOS
+# Force the compilers to clang for iOS
 include (CMakeForceCompiler)
-#CMAKE_FORCE_C_COMPILER (gcc gcc)
-#CMAKE_FORCE_CXX_COMPILER (g++ g++)
+#CMAKE_FORCE_C_COMPILER (clang GNU)
+#CMAKE_FORCE_CXX_COMPILER (clang++ GNU)
 
 set (CMAKE_C_SIZEOF_DATA_PTR 4)
 set (CMAKE_C_HAS_ISYSROOT 1)
@@ -28,4 +28,4 @@ SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM ONLY)
 SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
 SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
 
-message (STATUS "iPhoneOS toolchain loaded")
\ No newline at end of file
+message (STATUS "iPhoneOS toolchain loaded")
diff --git a/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake b/platforms/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake
similarity index 80%
rename from ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake
rename to platforms/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake
index 7ef8113edb..937079d182 100644
--- a/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake
+++ b/platforms/ios/cmake/Toolchains/Toolchain-iPhoneSimulator_Xcode.cmake
@@ -4,12 +4,12 @@ set (IPHONESIMULATOR TRUE)
 # Standard settings
 set (CMAKE_SYSTEM_NAME iOS)
 # Include extra modules for the iOS platform files
-set (CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/ios/cmake/Modules")
+set (CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/platforms/ios/cmake/Modules")
 
-# Force the compilers to gcc for iOS
+# Force the compilers to clang for iOS
 include (CMakeForceCompiler)
-#CMAKE_FORCE_C_COMPILER (gcc gcc)
-#CMAKE_FORCE_CXX_COMPILER (g++ g++)
+#CMAKE_FORCE_C_COMPILER (clang GNU)
+#CMAKE_FORCE_CXX_COMPILER (clang++ GNU)
 
 set (CMAKE_C_SIZEOF_DATA_PTR 4)
 set (CMAKE_C_HAS_ISYSROOT 1)
@@ -28,4 +28,4 @@ SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM ONLY)
 SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
 SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
 
-message (STATUS "iPhoneSimulator toolchain loaded")
\ No newline at end of file
+message (STATUS "iPhoneSimulator toolchain loaded")
diff --git a/platforms/ios/readme.txt b/platforms/ios/readme.txt
new file mode 100644
index 0000000000..0c39e7213c
--- /dev/null
+++ b/platforms/ios/readme.txt
@@ -0,0 +1,7 @@
+Building OpenCV from Source, using CMake and Command Line
+=========================================================
+
+cd ~/<my_working_directory>
+python opencv/platforms/ios/build_framework.py ios
+
+If everything's fine, a few minutes later you will get ~/<my_working_directory>/ios/opencv2.framework. You can add this framework to your Xcode projects.
diff --git a/platforms/linux/scripts/cmake_arm_gnueabi_hardfp.sh b/platforms/linux/scripts/cmake_arm_gnueabi_hardfp.sh
deleted file mode 100755
index 1d2153a384..0000000000
--- a/platforms/linux/scripts/cmake_arm_gnueabi_hardfp.sh
+++ /dev/null
@@ -1,7 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_hardfp
-cd build_hardfp
-
-cmake -DCMAKE_TOOLCHAIN_FILE=../arm-gnueabi.toolchain.cmake $@ ../../..
diff --git a/platforms/linux/scripts/cmake_arm_gnueabi_softfp.sh b/platforms/linux/scripts/cmake_arm_gnueabi_softfp.sh
deleted file mode 100755
index 5caf5a4e1b..0000000000
--- a/platforms/linux/scripts/cmake_arm_gnueabi_softfp.sh
+++ /dev/null
@@ -1,7 +0,0 @@
-#!/bin/sh
-cd `dirname $0`/..
-
-mkdir -p build_softfp
-cd build_softfp
-
-cmake -DSOFTFP=ON -DCMAKE_TOOLCHAIN_FILE=../arm-gnueabi.toolchain.cmake $@ ../../..
diff --git a/platforms/readme.txt b/platforms/readme.txt
index 7e1c4555c5..b039d3dc8a 100644
--- a/platforms/readme.txt
+++ b/platforms/readme.txt
@@ -1 +1,3 @@
-This folder contains toolchains and additional files that are needed for cross compitation.
\ No newline at end of file
+This folder contains toolchains and additional files that are needed for cross compilation.
+For more information see introduction tutorials for target platform in documentation:
+http://docs.opencv.org/doc/tutorials/introduction/table_of_content_introduction/table_of_content_introduction.html#table-of-content-introduction
diff --git a/android/scripts/ABI_compat_generator.py b/platforms/scripts/ABI_compat_generator.py
similarity index 98%
rename from android/scripts/ABI_compat_generator.py
rename to platforms/scripts/ABI_compat_generator.py
index 39253bbdec..d7cc3728aa 100755
--- a/android/scripts/ABI_compat_generator.py
+++ b/platforms/scripts/ABI_compat_generator.py
@@ -6,9 +6,7 @@ import os
 
 
 architecture = 'armeabi'
-excludedHeaders = set(['hdf5.h', 'cap_ios.h',
-    'eigen.hpp', 'cxeigen.hpp' #TOREMOVE
-    ])
+excludedHeaders = set(['hdf5.h', 'cap_ios.h', 'ios.h', 'eigen.hpp', 'cxeigen.hpp']) #TOREMOVE
 systemIncludes = ['sources/cxx-stl/gnu-libstdc++/4.6/include', \
     '/opt/android-ndk-r8c/platforms/android-8/arch-arm', # TODO: check if this one could be passed as command line arg
     'sources/cxx-stl/gnu-libstdc++/4.6/libs/armeabi-v7a/include']
diff --git a/android/scripts/camera_build.conf b/platforms/scripts/camera_build.conf
similarity index 85%
rename from android/scripts/camera_build.conf
rename to platforms/scripts/camera_build.conf
index cd172b4fde..fb79cacc21 100644
--- a/android/scripts/camera_build.conf
+++ b/platforms/scripts/camera_build.conf
@@ -21,3 +21,7 @@ native_camera_r4.2.0; armeabi-v7a; 14; $ANDROID_STUB_ROOT/4.2.0
 native_camera_r4.2.0; armeabi;     14; $ANDROID_STUB_ROOT/4.2.0
 native_camera_r4.2.0; x86;         14; $ANDROID_STUB_ROOT/4.2.0
 native_camera_r4.2.0; mips;        14; $ANDROID_STUB_ROOT/4.2.0
+native_camera_r4.3.0; armeabi;     14; $ANDROID_STUB_ROOT/4.3.0
+native_camera_r4.3.0; armeabi-v7a; 14; $ANDROID_STUB_ROOT/4.3.0
+native_camera_r4.3.0; x86;         14; $ANDROID_STUB_ROOT/4.3.0
+native_camera_r4.3.0; mips;        14; $ANDROID_STUB_ROOT/4.3.0
diff --git a/android/scripts/cmake_android_all_cameras.py b/platforms/scripts/cmake_android_all_cameras.py
similarity index 90%
rename from android/scripts/cmake_android_all_cameras.py
rename to platforms/scripts/cmake_android_all_cameras.py
index 0ef430a3d4..0739004776 100755
--- a/android/scripts/cmake_android_all_cameras.py
+++ b/platforms/scripts/cmake_android_all_cameras.py
@@ -60,7 +60,7 @@ for s in ConfFile.readlines():
 
     os.chdir(BuildDir)
     BuildLog = os.path.join(BuildDir, "build.log")
-    CmakeCmdLine = "cmake -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake -DANDROID_SOURCE_TREE=\"%s\" -DANDROID_NATIVE_API_LEVEL=\"%s\" -DANDROID_ABI=\"%s\" -DANDROID_STL=stlport_static ../../ > \"%s\" 2>&1" % (AndroidTreeRoot, NativeApiLevel, Arch, BuildLog)
+    CmakeCmdLine = "cmake -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake -DANDROID_SOURCE_TREE=\"%s\" -DANDROID_NATIVE_API_LEVEL=\"%s\" -DANDROID_ABI=\"%s\" -DANDROID_STL=stlport_static ../.. > \"%s\" 2>&1" % (AndroidTreeRoot, NativeApiLevel, Arch, BuildLog)
     MakeCmdLine = "make %s >> \"%s\" 2>&1" % (MakeTarget, BuildLog);
     #print(CmakeCmdLine)
     os.system(CmakeCmdLine)
diff --git a/android/scripts/cmake_android.sh b/platforms/scripts/cmake_android_arm.sh
similarity index 50%
rename from android/scripts/cmake_android.sh
rename to platforms/scripts/cmake_android_arm.sh
index 101ba3cee8..84c88a8159 100755
--- a/android/scripts/cmake_android.sh
+++ b/platforms/scripts/cmake_android_arm.sh
@@ -1,8 +1,7 @@
 #!/bin/sh
 cd `dirname $0`/..
 
-mkdir -p build
-cd build
-
-cmake -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON -DCMAKE_TOOLCHAIN_FILE=../android.toolchain.cmake $@ ../..
+mkdir -p build_android_arm
+cd build_android_arm
 
+cmake -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake $@ ../..
diff --git a/platforms/scripts/cmake_android_mips.sh b/platforms/scripts/cmake_android_mips.sh
new file mode 100755
index 0000000000..6bc7944b6d
--- /dev/null
+++ b/platforms/scripts/cmake_android_mips.sh
@@ -0,0 +1,7 @@
+#!/bin/sh
+cd `dirname $0`/..
+
+mkdir -p build_android_mips
+cd build_android_mips
+
+cmake -DANDROID_ABI=mips -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake $@ ../..
diff --git a/platforms/scripts/cmake_android_service.sh b/platforms/scripts/cmake_android_service.sh
new file mode 100755
index 0000000000..7ba8865b2a
--- /dev/null
+++ b/platforms/scripts/cmake_android_service.sh
@@ -0,0 +1,7 @@
+#!/bin/sh
+cd `dirname $0`/..
+
+mkdir -p build_android_service
+cd build_android_service
+
+cmake -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake -DANDROID_TOOLCHAIN_NAME="arm-linux-androideabi-4.4.3" -DANDROID_STL=stlport_static -DANDROID_STL_FORCE_FEATURES=OFF -DBUILD_ANDROID_SERVICE=ON -DANDROID_SOURCE_TREE=~/Projects/AndroidSource/ServiceStub/ $@ ../..
diff --git a/platforms/scripts/cmake_android_x86.sh b/platforms/scripts/cmake_android_x86.sh
new file mode 100755
index 0000000000..8fb8abda7e
--- /dev/null
+++ b/platforms/scripts/cmake_android_x86.sh
@@ -0,0 +1,8 @@
+#!/bin/sh
+
+cd `dirname $0`/..
+
+mkdir -p build_android_x86
+cd build_android_x86
+
+cmake -DANDROID_ABI=x86 -DCMAKE_TOOLCHAIN_FILE=../android/android.toolchain.cmake $@ ../..
diff --git a/platforms/scripts/cmake_arm_gnueabi_hardfp.sh b/platforms/scripts/cmake_arm_gnueabi_hardfp.sh
new file mode 100755
index 0000000000..1fce4f9dc1
--- /dev/null
+++ b/platforms/scripts/cmake_arm_gnueabi_hardfp.sh
@@ -0,0 +1,7 @@
+#!/bin/sh
+cd `dirname $0`/..
+
+mkdir -p build_linux_arm_hardfp
+cd build_linux_arm_hardfp
+
+cmake -DCMAKE_TOOLCHAIN_FILE=../linux/arm-gnueabi.toolchain.cmake $@ ../..
diff --git a/platforms/scripts/cmake_arm_gnueabi_softfp.sh b/platforms/scripts/cmake_arm_gnueabi_softfp.sh
new file mode 100755
index 0000000000..734348907c
--- /dev/null
+++ b/platforms/scripts/cmake_arm_gnueabi_softfp.sh
@@ -0,0 +1,7 @@
+#!/bin/sh
+cd `dirname $0`/..
+
+mkdir -p build_linux_arm_softfp
+cd build_linux_arm_softfp
+
+cmake -DSOFTFP=ON -DCMAKE_TOOLCHAIN_FILE=../linux/arm-gnueabi.toolchain.cmake $@ ../..
diff --git a/platforms/linux/scripts/cmake_carma.sh b/platforms/scripts/cmake_carma.sh
similarity index 100%
rename from platforms/linux/scripts/cmake_carma.sh
rename to platforms/scripts/cmake_carma.sh
diff --git a/platforms/scripts/cmake_winrt.cmd b/platforms/scripts/cmake_winrt.cmd
new file mode 100644
index 0000000000..392781ceb3
--- /dev/null
+++ b/platforms/scripts/cmake_winrt.cmd
@@ -0,0 +1,9 @@
+mkdir build_winrt_arm
+cd build_winrt_arm
+
+set msvc_path=C:\Program Files\Microsoft Visual Studio 11.0
+
+call "%msvc_path%\Common7\Tools\VsDevCmd.bat"
+call "%msvc_path%\VC\bin\x86_arm\vcvarsx86_arm.bat"
+
+cmake.exe -GNinja -DCMAKE_BUILD_TYPE=Release -DENABLE_WINRT_MODE=ON -DWITH_FFMPEG=OFF -DWITH_MSMF=OFF -DWITH_DSHOW=OFF -DWITH_VFW=OFF -DWITH_OPENEXR=OFF -DWITH_CUDA=OFF -DBUILD_opencv_gpu=OFF -DBUILD_opencv_python=OFF -DBUILD_opencv_java=OFF -DCMAKE_TOOLCHAIN_FILE=..\winrt\arm.winrt.toolchain.cmake  %* ..\..
diff --git a/platforms/winrt/arm.winrt.toolchain.cmake b/platforms/winrt/arm.winrt.toolchain.cmake
new file mode 100644
index 0000000000..a5158d01fa
--- /dev/null
+++ b/platforms/winrt/arm.winrt.toolchain.cmake
@@ -0,0 +1,6 @@
+set(CMAKE_SYSTEM_NAME Windows)
+set(CMAKE_SYSTEM_PROCESSOR "arm-v7a")
+
+set(CMAKE_FIND_ROOT_PATH "${CMAKE_SOURCE_DIR}/platforms/winrt")
+set(CMAKE_REQUIRED_DEFINITIONS -D_ARM_WINAPI_PARTITION_DESKTOP_SDK_AVAILABLE)
+add_definitions(-D_ARM_WINAPI_PARTITION_DESKTOP_SDK_AVAILABLE)
diff --git a/samples/MacOSX/FaceTracker/FaceTracker-Info.plist b/samples/MacOSX/FaceTracker/FaceTracker-Info.plist
index 8388dcd4f3..45d8bce13e 100644
--- a/samples/MacOSX/FaceTracker/FaceTracker-Info.plist
+++ b/samples/MacOSX/FaceTracker/FaceTracker-Info.plist
@@ -2,19 +2,19 @@
 <!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
 <dict>
-	<key>CFBundleDevelopmentRegion</key>
-	<string>English</string>
-	<key>CFBundleExecutable</key>
-	<string>${EXECUTABLE_NAME}</string>
-	<key>CFBundleIdentifier</key>
-	<string>de.rwth-aachen.ient.FaceTracker</string>
-	<key>CFBundleInfoDictionaryVersion</key>
-	<string>6.0</string>
-	<key>CFBundlePackageType</key>
-	<string>APPL</string>
-	<key>CFBundleSignature</key>
-	<string>????</string>
-	<key>CFBundleVersion</key>
-	<string>1.0</string>
+    <key>CFBundleDevelopmentRegion</key>
+    <string>English</string>
+    <key>CFBundleExecutable</key>
+    <string>${EXECUTABLE_NAME}</string>
+    <key>CFBundleIdentifier</key>
+    <string>de.rwth-aachen.ient.FaceTracker</string>
+    <key>CFBundleInfoDictionaryVersion</key>
+    <string>6.0</string>
+    <key>CFBundlePackageType</key>
+    <string>APPL</string>
+    <key>CFBundleSignature</key>
+    <string>????</string>
+    <key>CFBundleVersion</key>
+    <string>1.0</string>
 </dict>
 </plist>
diff --git a/samples/android/.gitignore b/samples/android/.gitignore
index 2d406cbbcc..a21956836e 100644
--- a/samples/android/.gitignore
+++ b/samples/android/.gitignore
@@ -4,4 +4,4 @@ build.xml
 local.properties
 proguard-project.txt
 project.properties
-default.properties
\ No newline at end of file
+default.properties
diff --git a/samples/android/15-puzzle/res/menu/activity_puzzle15.xml b/samples/android/15-puzzle/res/menu/activity_puzzle15.xml
deleted file mode 100644
index 7810d81963..0000000000
--- a/samples/android/15-puzzle/res/menu/activity_puzzle15.xml
+++ /dev/null
@@ -1,6 +0,0 @@
-<menu xmlns:android="http://schemas.android.com/apk/res/android">
-    <item android:id="@+id/menu_start_new_game"
-        android:title="@string/menu_start_new_game"
-        android:orderInCategory="100" />
-    <item android:id="@+id/menu_toggle_tile_numbers" android:title="@string/menu_toggle_tile_numbers"></item>
-</menu>
diff --git a/samples/android/15-puzzle/src/org/opencv/samples/puzzle15/Puzzle15Activity.java b/samples/android/15-puzzle/src/org/opencv/samples/puzzle15/Puzzle15Activity.java
index 466400d873..ebd34fc7e2 100644
--- a/samples/android/15-puzzle/src/org/opencv/samples/puzzle15/Puzzle15Activity.java
+++ b/samples/android/15-puzzle/src/org/opencv/samples/puzzle15/Puzzle15Activity.java
@@ -6,6 +6,7 @@ import org.opencv.android.OpenCVLoader;
 import org.opencv.core.Mat;
 import org.opencv.android.CameraBridgeViewBase;
 import org.opencv.android.CameraBridgeViewBase.CvCameraViewListener;
+import org.opencv.android.JavaCameraView;
 
 import android.os.Bundle;
 import android.app.Activity;
@@ -22,6 +23,9 @@ public class Puzzle15Activity extends Activity implements CvCameraViewListener,
 
     private CameraBridgeViewBase mOpenCvCameraView;
     private Puzzle15Processor    mPuzzle15;
+    private MenuItem             mItemHideNumbers;
+    private MenuItem             mItemStartNewGame;
+
 
     private int                  mGameWidth;
     private int                  mGameHeight;
@@ -52,9 +56,9 @@ public class Puzzle15Activity extends Activity implements CvCameraViewListener,
         super.onCreate(savedInstanceState);
         getWindow().addFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
 
-        setContentView(R.layout.activity_puzzle15);
-
-        mOpenCvCameraView = (CameraBridgeViewBase) findViewById(R.id.puzzle_activity_surface_view);
+        Log.d(TAG, "Creating and seting view");
+        mOpenCvCameraView = (CameraBridgeViewBase) new JavaCameraView(this, -1);
+        setContentView(mOpenCvCameraView);
         mOpenCvCameraView.setCvCameraViewListener(this);
         mPuzzle15 = new Puzzle15Processor();
         mPuzzle15.prepareNewGame();
@@ -83,17 +87,19 @@ public class Puzzle15Activity extends Activity implements CvCameraViewListener,
 
     @Override
     public boolean onCreateOptionsMenu(Menu menu) {
-        getMenuInflater().inflate(R.menu.activity_puzzle15, menu);
+        Log.i(TAG, "called onCreateOptionsMenu");
+        mItemHideNumbers = menu.add("Show/hide tile numbers");
+        mItemStartNewGame = menu.add("Start new game");
         return true;
     }
 
     @Override
     public boolean onOptionsItemSelected(MenuItem item) {
         Log.i(TAG, "Menu Item selected " + item);
-        if (item.getItemId() == R.id.menu_start_new_game) {
+        if (item == mItemStartNewGame) {
             /* We need to start new game */
             mPuzzle15.prepareNewGame();
-        } else if (item.getItemId() == R.id.menu_toggle_tile_numbers) {
+        } else if (item == mItemHideNumbers) {
             /* We need to enable or disable drawing of the tile numbers */
             mPuzzle15.toggleTileNumbers();
         }
diff --git a/samples/android/CMakeLists.txt b/samples/android/CMakeLists.txt
index a0794cb972..0dc4a3cd69 100644
--- a/samples/android/CMakeLists.txt
+++ b/samples/android/CMakeLists.txt
@@ -9,13 +9,15 @@ ocv_warnings_disable(CMAKE_CXX_FLAGS -Wmissing-declarations)
 add_subdirectory(15-puzzle)
 add_subdirectory(face-detection)
 add_subdirectory(image-manipulations)
+add_subdirectory(camera-calibration)
 add_subdirectory(color-blob-detection)
-
 add_subdirectory(tutorial-1-camerapreview)
 add_subdirectory(tutorial-2-mixedprocessing)
 add_subdirectory(tutorial-3-cameracontrol)
 
-#hello-android sample
+add_subdirectory(native-activity)
+
+# hello-android sample
 if(HAVE_opencv_highgui)
   ocv_include_modules_recurse(opencv_highgui opencv_core)
   add_executable(hello-android hello-android/main.cpp)
diff --git a/samples/android/camera-calibration/.classpath b/samples/android/camera-calibration/.classpath
new file mode 100644
index 0000000000..46c3d46963
--- /dev/null
+++ b/samples/android/camera-calibration/.classpath
@@ -0,0 +1,9 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<classpath>
+	<classpathentry exported="true" kind="con" path="com.android.ide.eclipse.adt.LIBRARIES"/>
+	<classpathentry kind="con" path="com.android.ide.eclipse.adt.ANDROID_FRAMEWORK"/>
+	<classpathentry exported="true" kind="con" path="com.android.ide.eclipse.adt.DEPENDENCIES"/>
+	<classpathentry kind="src" path="src"/>
+	<classpathentry kind="src" path="gen"/>
+	<classpathentry kind="output" path="bin/classes"/>
+</classpath>
diff --git a/samples/android/camera-calibration/.project b/samples/android/camera-calibration/.project
new file mode 100644
index 0000000000..eae413e1ae
--- /dev/null
+++ b/samples/android/camera-calibration/.project
@@ -0,0 +1,33 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<projectDescription>
+	<name>OpenCV Sample - camera-calibration</name>
+	<comment></comment>
+	<projects>
+	</projects>
+	<buildSpec>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.ResourceManagerBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.PreCompilerBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>org.eclipse.jdt.core.javabuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.ApkBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+	</buildSpec>
+	<natures>
+		<nature>com.android.ide.eclipse.adt.AndroidNature</nature>
+		<nature>org.eclipse.jdt.core.javanature</nature>
+	</natures>
+</projectDescription>
diff --git a/samples/android/camera-calibration/.settings/org.eclipse.jdt.core.prefs b/samples/android/camera-calibration/.settings/org.eclipse.jdt.core.prefs
new file mode 100644
index 0000000000..48ab4c6b11
--- /dev/null
+++ b/samples/android/camera-calibration/.settings/org.eclipse.jdt.core.prefs
@@ -0,0 +1,4 @@
+eclipse.preferences.version=1
+org.eclipse.jdt.core.compiler.codegen.targetPlatform=1.6
+org.eclipse.jdt.core.compiler.compliance=1.6
+org.eclipse.jdt.core.compiler.source=1.6
diff --git a/samples/android/camera-calibration/AndroidManifest.xml b/samples/android/camera-calibration/AndroidManifest.xml
new file mode 100644
index 0000000000..619c919eec
--- /dev/null
+++ b/samples/android/camera-calibration/AndroidManifest.xml
@@ -0,0 +1,38 @@
+<?xml version="1.0" encoding="utf-8"?>
+<manifest xmlns:android="http://schemas.android.com/apk/res/android"
+          package="org.opencv.samples.cameracalibration"
+          android:versionCode="1"
+          android:versionName="1.0">
+
+    <application
+        android:label="@string/app_name"
+        android:icon="@drawable/icon"
+        android:theme="@android:style/Theme.NoTitleBar.Fullscreen">
+
+        <activity android:name="CameraCalibrationActivity"
+                  android:label="@string/app_name"
+                  android:screenOrientation="landscape"
+                  android:configChanges="keyboardHidden|orientation">
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN" />
+                <category android:name="android.intent.category.LAUNCHER" />
+            </intent-filter>
+        </activity>
+    </application>
+
+    <supports-screens android:resizeable="true"
+                      android:smallScreens="true"
+                      android:normalScreens="true"
+                      android:largeScreens="true"
+                      android:anyDensity="true" />
+
+    <uses-sdk android:minSdkVersion="8" />
+
+    <uses-permission android:name="android.permission.CAMERA"/>
+
+    <uses-feature android:name="android.hardware.camera" android:required="false"/>
+    <uses-feature android:name="android.hardware.camera.autofocus" android:required="false"/>
+    <uses-feature android:name="android.hardware.camera.front" android:required="false"/>
+    <uses-feature android:name="android.hardware.camera.front.autofocus" android:required="false"/>
+
+</manifest>
diff --git a/samples/android/camera-calibration/CMakeLists.txt b/samples/android/camera-calibration/CMakeLists.txt
new file mode 100644
index 0000000000..83b11b3647
--- /dev/null
+++ b/samples/android/camera-calibration/CMakeLists.txt
@@ -0,0 +1,6 @@
+set(sample example-camera-calibration)
+
+add_android_project(${sample} "${CMAKE_CURRENT_SOURCE_DIR}" LIBRARY_DEPS ${OpenCV_BINARY_DIR} SDK_TARGET 11 ${ANDROID_SDK_TARGET})
+if(TARGET ${sample})
+  add_dependencies(opencv_android_examples ${sample})
+endif()
diff --git a/samples/android/camera-calibration/res/drawable/icon.png b/samples/android/camera-calibration/res/drawable/icon.png
new file mode 100644
index 0000000000..79ad948452
Binary files /dev/null and b/samples/android/camera-calibration/res/drawable/icon.png differ
diff --git a/samples/android/15-puzzle/res/layout/activity_puzzle15.xml b/samples/android/camera-calibration/res/layout/camera_calibration_surface_view.xml
similarity index 71%
rename from samples/android/15-puzzle/res/layout/activity_puzzle15.xml
rename to samples/android/camera-calibration/res/layout/camera_calibration_surface_view.xml
index 3257ed801c..0feccde3a1 100644
--- a/samples/android/15-puzzle/res/layout/activity_puzzle15.xml
+++ b/samples/android/camera-calibration/res/layout/camera_calibration_surface_view.xml
@@ -1,11 +1,12 @@
-<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
-    xmlns:tools="http://schemas.android.com/tools"
-    android:layout_width="match_parent"
-    android:layout_height="match_parent" >
-
-    <org.opencv.android.JavaCameraView
-        android:layout_width="fill_parent"
-        android:layout_height="fill_parent"
-        android:id="@+id/puzzle_activity_surface_view" />
-
-</LinearLayout>
+<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
+    xmlns:tools="http://schemas.android.com/tools"
+    xmlns:opencv="http://schemas.android.com/apk/res-auto"
+    android:layout_width="match_parent"
+    android:layout_height="match_parent" >
+
+    <org.opencv.android.JavaCameraView
+        android:layout_width="fill_parent"
+        android:layout_height="fill_parent"
+        android:id="@+id/camera_calibration_java_surface_view" />
+
+</LinearLayout>
diff --git a/samples/android/camera-calibration/res/menu/calibration.xml b/samples/android/camera-calibration/res/menu/calibration.xml
new file mode 100644
index 0000000000..9c90f127c1
--- /dev/null
+++ b/samples/android/camera-calibration/res/menu/calibration.xml
@@ -0,0 +1,22 @@
+<?xml version="1.0" encoding="utf-8"?>
+<menu xmlns:android="http://schemas.android.com/apk/res/android" >
+    <group android:checkableBehavior="single">
+        <item android:id="@+id/calibrate"
+              android:title="@string/action_calibrate"
+              android:showAsAction="ifRoom|withText" />
+        <item android:id="@+id/preview_mode"
+              android:title="@string/preview_mode">
+              <menu>
+                  <group android:checkableBehavior="single">
+                      <item android:id="@+id/calibration"
+                            android:title="@string/calibration"
+                            android:checked="true" />
+                      <item android:id="@+id/undistortion"
+                            android:title="@string/undistortion" />
+                      <item android:id="@+id/comparison"
+                            android:title="@string/comparison" />
+                  </group>
+              </menu>
+        </item>
+    </group>
+</menu>
diff --git a/samples/android/camera-calibration/res/values/strings.xml b/samples/android/camera-calibration/res/values/strings.xml
new file mode 100644
index 0000000000..e1ce932a14
--- /dev/null
+++ b/samples/android/camera-calibration/res/values/strings.xml
@@ -0,0 +1,18 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+
+    <string name="app_name">OCV Camera Calibration</string>
+    <string name="action_calibrate">Calibrate</string>
+    <string name="calibration">Calibration</string>
+    <string name="undistortion">Undistortion</string>
+    <string name="comparison">Comparison</string>
+    <string name="preview_mode">Preview mode</string>
+    <string name="calibration_successful">Successfully calibrated!\nAvg. re-projection error:</string>
+    <string name="calibration_unsuccessful">Unsuccessful calibration.\nTry again</string>
+    <string name="more_samples">Please, capture more samples</string>
+    <string name="calibrating">Calibrating...</string>
+    <string name="please_wait">Please, wait</string>
+    <string name="original">Original</string>
+    <string name="undistorted">Undistorted</string>
+
+</resources>
diff --git a/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CalibrationResult.java b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CalibrationResult.java
new file mode 100644
index 0000000000..4b03d5967a
--- /dev/null
+++ b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CalibrationResult.java
@@ -0,0 +1,69 @@
+package org.opencv.samples.cameracalibration;
+
+import org.opencv.core.Mat;
+
+import android.app.Activity;
+import android.content.Context;
+import android.content.SharedPreferences;
+import android.util.Log;
+
+public abstract class CalibrationResult {
+    private static final String TAG = "OCVSample::CalibrationResult";
+
+    private static final int CAMERA_MATRIX_ROWS = 3;
+    private static final int CAMERA_MATRIX_COLS = 3;
+    private static final int DISTORTION_COEFFICIENTS_SIZE = 5;
+
+    public static void save(Activity activity, Mat cameraMatrix, Mat distortionCoefficients) {
+        SharedPreferences sharedPref = activity.getPreferences(Context.MODE_PRIVATE);
+        SharedPreferences.Editor editor = sharedPref.edit();
+
+        double[] cameraMatrixArray = new double[CAMERA_MATRIX_ROWS * CAMERA_MATRIX_COLS];
+        cameraMatrix.get(0,  0, cameraMatrixArray);
+        for (int i = 0; i < CAMERA_MATRIX_ROWS; i++) {
+            for (int j = 0; j < CAMERA_MATRIX_COLS; j++) {
+                Integer id = i * CAMERA_MATRIX_ROWS + j;
+                editor.putFloat(id.toString(), (float)cameraMatrixArray[id]);
+            }
+        }
+
+        double[] distortionCoefficientsArray = new double[DISTORTION_COEFFICIENTS_SIZE];
+        distortionCoefficients.get(0, 0, distortionCoefficientsArray);
+        int shift = CAMERA_MATRIX_ROWS * CAMERA_MATRIX_COLS;
+        for (Integer i = shift; i < DISTORTION_COEFFICIENTS_SIZE + shift; i++) {
+            editor.putFloat(i.toString(), (float)distortionCoefficientsArray[i-shift]);
+        }
+
+        editor.commit();
+        Log.i(TAG, "Saved camera matrix: " + cameraMatrix.dump());
+        Log.i(TAG, "Saved distortion coefficients: " + distortionCoefficients.dump());
+    }
+
+    public static boolean tryLoad(Activity activity, Mat cameraMatrix, Mat distortionCoefficients) {
+        SharedPreferences sharedPref = activity.getPreferences(Context.MODE_PRIVATE);
+        if (sharedPref.getFloat("0", -1) == -1) {
+            Log.i(TAG, "No previous calibration results found");
+            return false;
+        }
+
+        double[] cameraMatrixArray = new double[CAMERA_MATRIX_ROWS * CAMERA_MATRIX_COLS];
+        for (int i = 0; i < CAMERA_MATRIX_ROWS; i++) {
+            for (int j = 0; j < CAMERA_MATRIX_COLS; j++) {
+                Integer id = i * CAMERA_MATRIX_ROWS + j;
+                cameraMatrixArray[id] = sharedPref.getFloat(id.toString(), -1);
+            }
+        }
+        cameraMatrix.put(0, 0, cameraMatrixArray);
+        Log.i(TAG, "Loaded camera matrix: " + cameraMatrix.dump());
+
+        double[] distortionCoefficientsArray = new double[DISTORTION_COEFFICIENTS_SIZE];
+        int shift = CAMERA_MATRIX_ROWS * CAMERA_MATRIX_COLS;
+        for (Integer i = shift; i < DISTORTION_COEFFICIENTS_SIZE + shift; i++) {
+            distortionCoefficientsArray[i - shift] = sharedPref.getFloat(i.toString(), -1);
+        }
+        distortionCoefficients.put(0, 0, distortionCoefficientsArray);
+        Log.i(TAG, "Loaded distortion coefficients: " + distortionCoefficients.dump());
+
+        return true;
+    }
+}
diff --git a/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrationActivity.java b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrationActivity.java
new file mode 100644
index 0000000000..33c9bbbf46
--- /dev/null
+++ b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrationActivity.java
@@ -0,0 +1,216 @@
+// This sample is based on "Camera calibration With OpenCV" tutorial:
+// http://docs.opencv.org/doc/tutorials/calib3d/camera_calibration/camera_calibration.html
+//
+// It uses standard OpenCV asymmetric circles grid pattern 11x4:
+// https://github.com/Itseez/opencv/blob/2.4/doc/acircles_pattern.png.
+// The results are the camera matrix and 5 distortion coefficients.
+//
+// Tap on highlighted pattern to capture pattern corners for calibration.
+// Move pattern along the whole screen and capture data.
+//
+// When you've captured necessary amount of pattern corners (usually ~20 are enough),
+// press "Calibrate" button for performing camera calibration.
+
+package org.opencv.samples.cameracalibration;
+
+import org.opencv.android.BaseLoaderCallback;
+import org.opencv.android.CameraBridgeViewBase;
+import org.opencv.android.CameraBridgeViewBase.CvCameraViewFrame;
+import org.opencv.android.CameraBridgeViewBase.CvCameraViewListener2;
+import org.opencv.android.LoaderCallbackInterface;
+import org.opencv.android.OpenCVLoader;
+import org.opencv.core.Mat;
+
+import android.app.Activity;
+import android.app.ProgressDialog;
+import android.content.res.Resources;
+import android.os.AsyncTask;
+import android.os.Bundle;
+import android.util.Log;
+import android.view.Menu;
+import android.view.MenuItem;
+import android.view.MotionEvent;
+import android.view.SurfaceView;
+import android.view.View;
+import android.view.View.OnTouchListener;
+import android.view.WindowManager;
+import android.widget.Toast;
+
+public class CameraCalibrationActivity extends Activity implements CvCameraViewListener2, OnTouchListener {
+    private static final String TAG = "OCVSample::Activity";
+
+    private CameraBridgeViewBase mOpenCvCameraView;
+    private CameraCalibrator mCalibrator;
+    private OnCameraFrameRender mOnCameraFrameRender;
+    private int mWidth;
+    private int mHeight;
+
+    private BaseLoaderCallback mLoaderCallback = new BaseLoaderCallback(this) {
+        @Override
+        public void onManagerConnected(int status) {
+            switch (status) {
+            case LoaderCallbackInterface.SUCCESS:
+            {
+                Log.i(TAG, "OpenCV loaded successfully");
+                mOpenCvCameraView.enableView();
+                mOpenCvCameraView.setOnTouchListener(CameraCalibrationActivity.this);
+            } break;
+            default:
+            {
+                super.onManagerConnected(status);
+            } break;
+            }
+        }
+    };
+
+    public CameraCalibrationActivity() {
+        Log.i(TAG, "Instantiated new " + this.getClass());
+    }
+
+    @Override
+    public void onCreate(Bundle savedInstanceState) {
+        Log.i(TAG, "called onCreate");
+        super.onCreate(savedInstanceState);
+        getWindow().addFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
+
+        setContentView(R.layout.camera_calibration_surface_view);
+
+        mOpenCvCameraView = (CameraBridgeViewBase) findViewById(R.id.camera_calibration_java_surface_view);
+        mOpenCvCameraView.setVisibility(SurfaceView.VISIBLE);
+        mOpenCvCameraView.setCvCameraViewListener(this);
+    }
+
+    @Override
+    public void onPause()
+    {
+        super.onPause();
+        if (mOpenCvCameraView != null)
+            mOpenCvCameraView.disableView();
+    }
+
+    @Override
+    public void onResume()
+    {
+        super.onResume();
+        OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_2, this, mLoaderCallback);
+    }
+
+    public void onDestroy() {
+        super.onDestroy();
+        if (mOpenCvCameraView != null)
+            mOpenCvCameraView.disableView();
+    }
+
+    @Override
+    public boolean onCreateOptionsMenu(Menu menu) {
+        super.onCreateOptionsMenu(menu);
+        getMenuInflater().inflate(R.menu.calibration, menu);
+
+        return true;
+    }
+
+    @Override
+    public boolean onPrepareOptionsMenu (Menu menu) {
+        super.onPrepareOptionsMenu(menu);
+        menu.findItem(R.id.preview_mode).setEnabled(true);
+        if (!mCalibrator.isCalibrated())
+            menu.findItem(R.id.preview_mode).setEnabled(false);
+
+        return true;
+    }
+
+    @Override
+    public boolean onOptionsItemSelected(MenuItem item) {
+        switch (item.getItemId()) {
+        case R.id.calibration:
+            mOnCameraFrameRender =
+                new OnCameraFrameRender(new CalibrationFrameRender(mCalibrator));
+            item.setChecked(true);
+            return true;
+        case R.id.undistortion:
+            mOnCameraFrameRender =
+                new OnCameraFrameRender(new UndistortionFrameRender(mCalibrator));
+            item.setChecked(true);
+            return true;
+        case R.id.comparison:
+            mOnCameraFrameRender =
+                new OnCameraFrameRender(new ComparisonFrameRender(mCalibrator, mWidth, mHeight, getResources()));
+            item.setChecked(true);
+            return true;
+        case R.id.calibrate:
+            final Resources res = getResources();
+            if (mCalibrator.getCornersBufferSize() < 2) {
+                (Toast.makeText(this, res.getString(R.string.more_samples), Toast.LENGTH_SHORT)).show();
+                return true;
+            }
+
+            mOnCameraFrameRender = new OnCameraFrameRender(new PreviewFrameRender());
+            new AsyncTask<Void, Void, Void>() {
+                private ProgressDialog calibrationProgress;
+
+                @Override
+                protected void onPreExecute() {
+                    calibrationProgress = new ProgressDialog(CameraCalibrationActivity.this);
+                    calibrationProgress.setTitle(res.getString(R.string.calibrating));
+                    calibrationProgress.setMessage(res.getString(R.string.please_wait));
+                    calibrationProgress.setCancelable(false);
+                    calibrationProgress.setIndeterminate(true);
+                    calibrationProgress.show();
+                }
+
+                @Override
+                protected Void doInBackground(Void... arg0) {
+                    mCalibrator.calibrate();
+                    return null;
+                }
+
+                @Override
+                protected void onPostExecute(Void result) {
+                    calibrationProgress.dismiss();
+                    mCalibrator.clearCorners();
+                    mOnCameraFrameRender = new OnCameraFrameRender(new CalibrationFrameRender(mCalibrator));
+                    String resultMessage = (mCalibrator.isCalibrated()) ?
+                            res.getString(R.string.calibration_successful)  + " " + mCalibrator.getAvgReprojectionError() :
+                            res.getString(R.string.calibration_unsuccessful);
+                    (Toast.makeText(CameraCalibrationActivity.this, resultMessage, Toast.LENGTH_SHORT)).show();
+
+                    if (mCalibrator.isCalibrated()) {
+                        CalibrationResult.save(CameraCalibrationActivity.this,
+                                mCalibrator.getCameraMatrix(), mCalibrator.getDistortionCoefficients());
+                    }
+                }
+            }.execute();
+            return true;
+        default:
+            return super.onOptionsItemSelected(item);
+        }
+    }
+
+    public void onCameraViewStarted(int width, int height) {
+        if (mWidth != width || mHeight != height) {
+            mWidth = width;
+            mHeight = height;
+            mCalibrator = new CameraCalibrator(mWidth, mHeight);
+            if (CalibrationResult.tryLoad(this, mCalibrator.getCameraMatrix(), mCalibrator.getDistortionCoefficients())) {
+                mCalibrator.setCalibrated();
+            }
+
+            mOnCameraFrameRender = new OnCameraFrameRender(new CalibrationFrameRender(mCalibrator));
+        }
+    }
+
+    public void onCameraViewStopped() {
+    }
+
+    public Mat onCameraFrame(CvCameraViewFrame inputFrame) {
+        return mOnCameraFrameRender.render(inputFrame);
+    }
+
+    @Override
+    public boolean onTouch(View v, MotionEvent event) {
+        Log.d(TAG, "onTouch invoked");
+
+        mCalibrator.addCorners();
+        return false;
+    }
+}
diff --git a/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrator.java b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrator.java
new file mode 100644
index 0000000000..2f9df6a3cc
--- /dev/null
+++ b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/CameraCalibrator.java
@@ -0,0 +1,169 @@
+package org.opencv.samples.cameracalibration;
+
+import java.util.ArrayList;
+import java.util.List;
+
+import org.opencv.calib3d.Calib3d;
+import org.opencv.core.Core;
+import org.opencv.core.CvType;
+import org.opencv.core.Mat;
+import org.opencv.core.MatOfDouble;
+import org.opencv.core.MatOfPoint2f;
+import org.opencv.core.MatOfPoint3f;
+import org.opencv.core.Point;
+import org.opencv.core.Scalar;
+import org.opencv.core.Size;
+
+import android.util.Log;
+
+public class CameraCalibrator {
+    private static final String TAG = "OCVSample::CameraCalibrator";
+
+    private final Size mPatternSize = new Size(4, 11);
+    private final int mCornersSize = (int)(mPatternSize.width * mPatternSize.height);
+    private boolean mPatternWasFound = false;
+    private MatOfPoint2f mCorners = new MatOfPoint2f();
+    private List<Mat> mCornersBuffer = new ArrayList<Mat>();
+    private boolean mIsCalibrated = false;
+
+    private Mat mCameraMatrix = new Mat();
+    private Mat mDistortionCoefficients = new Mat();
+    private int mFlags;
+    private double mRms;
+    private double mSquareSize = 0.0181;
+    private Size mImageSize;
+
+    public CameraCalibrator(int width, int height) {
+        mImageSize = new Size(width, height);
+        mFlags = Calib3d.CALIB_FIX_PRINCIPAL_POINT +
+                 Calib3d.CALIB_ZERO_TANGENT_DIST +
+                 Calib3d.CALIB_FIX_ASPECT_RATIO +
+                 Calib3d.CALIB_FIX_K4 +
+                 Calib3d.CALIB_FIX_K5;
+        Mat.eye(3, 3, CvType.CV_64FC1).copyTo(mCameraMatrix);
+        mCameraMatrix.put(0, 0, 1.0);
+        Mat.zeros(5, 1, CvType.CV_64FC1).copyTo(mDistortionCoefficients);
+        Log.i(TAG, "Instantiated new " + this.getClass());
+    }
+
+    public void processFrame(Mat grayFrame, Mat rgbaFrame) {
+        findPattern(grayFrame);
+        renderFrame(rgbaFrame);
+    }
+
+    public void calibrate() {
+        ArrayList<Mat> rvecs = new ArrayList<Mat>();
+        ArrayList<Mat> tvecs = new ArrayList<Mat>();
+        Mat reprojectionErrors = new Mat();
+        ArrayList<Mat> objectPoints = new ArrayList<Mat>();
+        objectPoints.add(Mat.zeros(mCornersSize, 1, CvType.CV_32FC3));
+        calcBoardCornerPositions(objectPoints.get(0));
+        for (int i = 1; i < mCornersBuffer.size(); i++) {
+            objectPoints.add(objectPoints.get(0));
+        }
+
+        Calib3d.calibrateCamera(objectPoints, mCornersBuffer, mImageSize,
+                mCameraMatrix, mDistortionCoefficients, rvecs, tvecs, mFlags);
+
+        mIsCalibrated = Core.checkRange(mCameraMatrix)
+                && Core.checkRange(mDistortionCoefficients);
+
+        mRms = computeReprojectionErrors(objectPoints, rvecs, tvecs, reprojectionErrors);
+        Log.i(TAG, String.format("Average re-projection error: %f", mRms));
+        Log.i(TAG, "Camera matrix: " + mCameraMatrix.dump());
+        Log.i(TAG, "Distortion coefficients: " + mDistortionCoefficients.dump());
+    }
+
+    public void clearCorners() {
+        mCornersBuffer.clear();
+    }
+
+    private void calcBoardCornerPositions(Mat corners) {
+        final int cn = 3;
+        float positions[] = new float[mCornersSize * cn];
+
+        for (int i = 0; i < mPatternSize.height; i++) {
+            for (int j = 0; j < mPatternSize.width * cn; j += cn) {
+                positions[(int) (i * mPatternSize.width * cn + j + 0)] =
+                        (2 * (j / cn) + i % 2) * (float) mSquareSize;
+                positions[(int) (i * mPatternSize.width * cn + j + 1)] =
+                        i * (float) mSquareSize;
+                positions[(int) (i * mPatternSize.width * cn + j + 2)] = 0;
+            }
+        }
+        corners.create(mCornersSize, 1, CvType.CV_32FC3);
+        corners.put(0, 0, positions);
+    }
+
+    private double computeReprojectionErrors(List<Mat> objectPoints,
+            List<Mat> rvecs, List<Mat> tvecs, Mat perViewErrors) {
+        MatOfPoint2f cornersProjected = new MatOfPoint2f();
+        double totalError = 0;
+        double error;
+        float viewErrors[] = new float[objectPoints.size()];
+
+        MatOfDouble distortionCoefficients = new MatOfDouble(mDistortionCoefficients);
+        int totalPoints = 0;
+        for (int i = 0; i < objectPoints.size(); i++) {
+            MatOfPoint3f points = new MatOfPoint3f(objectPoints.get(i));
+            Calib3d.projectPoints(points, rvecs.get(i), tvecs.get(i),
+                    mCameraMatrix, distortionCoefficients, cornersProjected);
+            error = Core.norm(mCornersBuffer.get(i), cornersProjected, Core.NORM_L2);
+
+            int n = objectPoints.get(i).rows();
+            viewErrors[i] = (float) Math.sqrt(error * error / n);
+            totalError  += error * error;
+            totalPoints += n;
+        }
+        perViewErrors.create(objectPoints.size(), 1, CvType.CV_32FC1);
+        perViewErrors.put(0, 0, viewErrors);
+
+        return Math.sqrt(totalError / totalPoints);
+    }
+
+    private void findPattern(Mat grayFrame) {
+        mPatternWasFound = Calib3d.findCirclesGrid(grayFrame, mPatternSize,
+                mCorners, Calib3d.CALIB_CB_ASYMMETRIC_GRID);
+    }
+
+    public void addCorners() {
+        if (mPatternWasFound) {
+            mCornersBuffer.add(mCorners.clone());
+        }
+    }
+
+    private void drawPoints(Mat rgbaFrame) {
+        Calib3d.drawChessboardCorners(rgbaFrame, mPatternSize, mCorners, mPatternWasFound);
+    }
+
+    private void renderFrame(Mat rgbaFrame) {
+        drawPoints(rgbaFrame);
+
+        Core.putText(rgbaFrame, "Captured: " + mCornersBuffer.size(), new Point(rgbaFrame.cols() / 3 * 2, rgbaFrame.rows() * 0.1),
+                Core.FONT_HERSHEY_SIMPLEX, 1.0, new Scalar(255, 255, 0));
+    }
+
+    public Mat getCameraMatrix() {
+        return mCameraMatrix;
+    }
+
+    public Mat getDistortionCoefficients() {
+        return mDistortionCoefficients;
+    }
+
+    public int getCornersBufferSize() {
+        return mCornersBuffer.size();
+    }
+
+    public double getAvgReprojectionError() {
+        return mRms;
+    }
+
+    public boolean isCalibrated() {
+        return mIsCalibrated;
+    }
+
+    public void setCalibrated() {
+        mIsCalibrated = true;
+    }
+}
diff --git a/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/OnCameraFrameRender.java b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/OnCameraFrameRender.java
new file mode 100644
index 0000000000..3f155c2bfc
--- /dev/null
+++ b/samples/android/camera-calibration/src/org/opencv/samples/cameracalibration/OnCameraFrameRender.java
@@ -0,0 +1,102 @@
+package org.opencv.samples.cameracalibration;
+
+import java.util.ArrayList;
+import java.util.List;
+
+import org.opencv.android.CameraBridgeViewBase.CvCameraViewFrame;
+import org.opencv.core.Core;
+import org.opencv.core.Mat;
+import org.opencv.core.MatOfPoint;
+import org.opencv.core.Point;
+import org.opencv.core.Range;
+import org.opencv.core.Scalar;
+import org.opencv.imgproc.Imgproc;
+
+import android.content.res.Resources;
+
+abstract class FrameRender {
+    protected CameraCalibrator mCalibrator;
+
+    public abstract Mat render(CvCameraViewFrame inputFrame);
+}
+
+class PreviewFrameRender extends FrameRender {
+    @Override
+    public Mat render(CvCameraViewFrame inputFrame) {
+        return inputFrame.rgba();
+    }
+}
+
+class CalibrationFrameRender extends FrameRender {
+    public CalibrationFrameRender(CameraCalibrator calibrator) {
+        mCalibrator = calibrator;
+    }
+
+    @Override
+    public Mat render(CvCameraViewFrame inputFrame) {
+        Mat rgbaFrame = inputFrame.rgba();
+        Mat grayFrame = inputFrame.gray();
+        mCalibrator.processFrame(grayFrame, rgbaFrame);
+
+        return rgbaFrame;
+    }
+}
+
+class UndistortionFrameRender extends FrameRender {
+    public UndistortionFrameRender(CameraCalibrator calibrator) {
+        mCalibrator = calibrator;
+    }
+
+    @Override
+    public Mat render(CvCameraViewFrame inputFrame) {
+        Mat renderedFrame = new Mat(inputFrame.rgba().size(), inputFrame.rgba().type());
+        Imgproc.undistort(inputFrame.rgba(), renderedFrame,
+                mCalibrator.getCameraMatrix(), mCalibrator.getDistortionCoefficients());
+
+        return renderedFrame;
+    }
+}
+
+class ComparisonFrameRender extends FrameRender {
+    private int mWidth;
+    private int mHeight;
+    private Resources mResources;
+    public ComparisonFrameRender(CameraCalibrator calibrator, int width, int height, Resources resources) {
+        mCalibrator = calibrator;
+        mWidth = width;
+        mHeight = height;
+        mResources = resources;
+    }
+
+    @Override
+    public Mat render(CvCameraViewFrame inputFrame) {
+        Mat undistortedFrame = new Mat(inputFrame.rgba().size(), inputFrame.rgba().type());
+        Imgproc.undistort(inputFrame.rgba(), undistortedFrame,
+                mCalibrator.getCameraMatrix(), mCalibrator.getDistortionCoefficients());
+
+        Mat comparisonFrame = inputFrame.rgba();
+        undistortedFrame.colRange(new Range(0, mWidth / 2)).copyTo(comparisonFrame.colRange(new Range(mWidth / 2, mWidth)));
+        List<MatOfPoint> border = new ArrayList<MatOfPoint>();
+        final int shift = (int)(mWidth * 0.005);
+        border.add(new MatOfPoint(new Point(mWidth / 2 - shift, 0), new Point(mWidth / 2 + shift, 0),
+                new Point(mWidth / 2 + shift, mHeight), new Point(mWidth / 2 - shift, mHeight)));
+        Core.fillPoly(comparisonFrame, border, new Scalar(255, 255, 255));
+
+        Core.putText(comparisonFrame, mResources.getString(R.string.original), new Point(mWidth * 0.1, mHeight * 0.1),
+                Core.FONT_HERSHEY_SIMPLEX, 1.0, new Scalar(255, 255, 0));
+        Core.putText(comparisonFrame, mResources.getString(R.string.undistorted), new Point(mWidth * 0.6, mHeight * 0.1),
+                Core.FONT_HERSHEY_SIMPLEX, 1.0, new Scalar(255, 255, 0));
+
+        return comparisonFrame;
+    }
+}
+
+class OnCameraFrameRender {
+    private FrameRender mFrameRender;
+    public OnCameraFrameRender(FrameRender frameRender) {
+        mFrameRender = frameRender;
+    }
+    public Mat render(CvCameraViewFrame inputFrame) {
+        return mFrameRender.render(inputFrame);
+    }
+}
diff --git a/samples/android/color-blob-detection/CMakeLists.txt b/samples/android/color-blob-detection/CMakeLists.txt
index 400db29ceb..42051bc3a2 100644
--- a/samples/android/color-blob-detection/CMakeLists.txt
+++ b/samples/android/color-blob-detection/CMakeLists.txt
@@ -4,4 +4,3 @@ add_android_project(${sample} "${CMAKE_CURRENT_SOURCE_DIR}" LIBRARY_DEPS ${OpenC
 if(TARGET ${sample})
   add_dependencies(opencv_android_examples ${sample})
 endif()
-
diff --git a/samples/android/color-blob-detection/src/org/opencv/samples/colorblobdetect/ColorBlobDetectionActivity.java b/samples/android/color-blob-detection/src/org/opencv/samples/colorblobdetect/ColorBlobDetectionActivity.java
index 0799df26de..276b03aeba 100644
--- a/samples/android/color-blob-detection/src/org/opencv/samples/colorblobdetect/ColorBlobDetectionActivity.java
+++ b/samples/android/color-blob-detection/src/org/opencv/samples/colorblobdetect/ColorBlobDetectionActivity.java
@@ -187,4 +187,4 @@ public class ColorBlobDetectionActivity extends Activity implements OnTouchListe
 
         return new Scalar(pointMatRgba.get(0, 0));
     }
-}
\ No newline at end of file
+}
diff --git a/samples/android/face-detection/CMakeLists.txt b/samples/android/face-detection/CMakeLists.txt
index 70658a799b..0f89b2a362 100644
--- a/samples/android/face-detection/CMakeLists.txt
+++ b/samples/android/face-detection/CMakeLists.txt
@@ -10,4 +10,3 @@ add_android_project(${sample} "${CMAKE_CURRENT_SOURCE_DIR}" LIBRARY_DEPS ${OpenC
 if(TARGET ${sample})
   add_dependencies(opencv_android_examples ${sample})
 endif()
-
diff --git a/samples/android/face-detection/jni/Android.mk b/samples/android/face-detection/jni/Android.mk
index e6db280c15..6b6642602d 100644
--- a/samples/android/face-detection/jni/Android.mk
+++ b/samples/android/face-detection/jni/Android.mk
@@ -13,4 +13,4 @@ LOCAL_LDLIBS     += -llog -ldl
 
 LOCAL_MODULE     := detection_based_tracker
 
-include $(BUILD_SHARED_LIBRARY)
\ No newline at end of file
+include $(BUILD_SHARED_LIBRARY)
diff --git a/samples/android/face-detection/jni/DetectionBasedTracker_jni.cpp b/samples/android/face-detection/jni/DetectionBasedTracker_jni.cpp
index e0e53e3701..33a8334c6a 100644
--- a/samples/android/face-detection/jni/DetectionBasedTracker_jni.cpp
+++ b/samples/android/face-detection/jni/DetectionBasedTracker_jni.cpp
@@ -26,7 +26,7 @@ public:
             Detector(detector)
     {
         LOGD("CascadeDetectorAdapter::Detect::Detect");
-        CV_Assert(!detector.empty());
+        CV_Assert(detector);
     }
 
     void detect(const cv::Mat &Image, std::vector<cv::Rect> &objects)
@@ -57,11 +57,11 @@ struct DetectorAgregator
             mainDetector(_mainDetector),
             trackingDetector(_trackingDetector)
     {
-        CV_Assert(!_mainDetector.empty());
-        CV_Assert(!_trackingDetector.empty());
+        CV_Assert(_mainDetector);
+        CV_Assert(_trackingDetector);
 
         DetectionBasedTracker::Parameters DetectorParams;
-        tracker = new DetectionBasedTracker(mainDetector.ptr<DetectionBasedTracker::IDetector>(), trackingDetector.ptr<DetectionBasedTracker::IDetector>(), DetectorParams);
+        tracker = makePtr<DetectionBasedTracker>(mainDetector, trackingDetector, DetectorParams);
     }
 };
 
@@ -77,8 +77,10 @@ JNIEXPORT jlong JNICALL Java_org_opencv_samples_facedetect_DetectionBasedTracker
 
     try
     {
-        cv::Ptr<CascadeDetectorAdapter> mainDetector = new CascadeDetectorAdapter(new CascadeClassifier(stdFileName));
-        cv::Ptr<CascadeDetectorAdapter> trackingDetector = new CascadeDetectorAdapter(new CascadeClassifier(stdFileName));
+        cv::Ptr<CascadeDetectorAdapter> mainDetector = makePtr<CascadeDetectorAdapter>(
+            makePtr<CascadeClassifier>(stdFileName));
+        cv::Ptr<CascadeDetectorAdapter> trackingDetector = makePtr<CascadeDetectorAdapter>(
+            makePtr<CascadeClassifier>(stdFileName));
         result = (jlong)new DetectorAgregator(mainDetector, trackingDetector);
         if (faceSize > 0)
         {
diff --git a/samples/android/face-detection/src/org/opencv/samples/facedetect/FdActivity.java b/samples/android/face-detection/src/org/opencv/samples/facedetect/FdActivity.java
index 23727739e4..b06b2cc1c5 100644
--- a/samples/android/face-detection/src/org/opencv/samples/facedetect/FdActivity.java
+++ b/samples/android/face-detection/src/org/opencv/samples/facedetect/FdActivity.java
@@ -215,9 +215,9 @@ public class FdActivity extends Activity implements CvCameraViewListener2 {
         else if (item == mItemFace20)
             setMinFaceSize(0.2f);
         else if (item == mItemType) {
-            mDetectorType = (mDetectorType + 1) % mDetectorName.length;
-            item.setTitle(mDetectorName[mDetectorType]);
-            setDetectorType(mDetectorType);
+            int tmpDetectorType = (mDetectorType + 1) % mDetectorName.length;
+            item.setTitle(mDetectorName[tmpDetectorType]);
+            setDetectorType(tmpDetectorType);
         }
         return true;
     }
diff --git a/samples/android/hello-android/main.cpp b/samples/android/hello-android/main.cpp
index 2b985c6421..cd02e1742a 100644
--- a/samples/android/hello-android/main.cpp
+++ b/samples/android/hello-android/main.cpp
@@ -11,9 +11,9 @@ int main(int argc, char* argv[])
   printf("%s\n", message);
 
   // put message to simple image
-  Size textsize = getTextSize(message, CV_FONT_HERSHEY_COMPLEX, 3, 5, 0);
+  Size textsize = getTextSize(message, FONT_HERSHEY_COMPLEX, 3, 5, 0);
   Mat img(textsize.height + 20, textsize.width + 20, CV_32FC1, Scalar(230,230,230));
-  putText(img, message, Point(10, img.rows - 10), CV_FONT_HERSHEY_COMPLEX, 3, Scalar(0, 0, 0), 5);
+  putText(img, message, Point(10, img.rows - 10), FONT_HERSHEY_COMPLEX, 3, Scalar(0, 0, 0), 5);
 
   // save\show resulting image
 #if ANDROID
@@ -24,4 +24,3 @@ int main(int argc, char* argv[])
 #endif
   return 0;
 }
-
diff --git a/samples/android/image-manipulations/src/org/opencv/samples/imagemanipulations/ImageManipulationsActivity.java b/samples/android/image-manipulations/src/org/opencv/samples/imagemanipulations/ImageManipulationsActivity.java
index 88ade8430f..38f1d5959b 100644
--- a/samples/android/image-manipulations/src/org/opencv/samples/imagemanipulations/ImageManipulationsActivity.java
+++ b/samples/android/image-manipulations/src/org/opencv/samples/imagemanipulations/ImageManipulationsActivity.java
@@ -48,17 +48,12 @@ public class ImageManipulationsActivity extends Activity implements CvCameraView
     private CameraBridgeViewBase mOpenCvCameraView;
 
     private Size                 mSize0;
-    private Size                 mSizeRgba;
-    private Size                 mSizeRgbaInner;
 
-    private Mat                  mRgba;
-    private Mat                  mGray;
     private Mat                  mIntermediateMat;
-    private Mat                  mHist;
     private Mat                  mMat0;
     private MatOfInt             mChannels[];
     private MatOfInt             mHistSize;
-    private int                  mHistSizeNum;
+    private int                  mHistSizeNum = 25;
     private MatOfFloat           mRanges;
     private Scalar               mColorsRGB[];
     private Scalar               mColorsHue[];
@@ -66,10 +61,6 @@ public class ImageManipulationsActivity extends Activity implements CvCameraView
     private Point                mP1;
     private Point                mP2;
     private float                mBuff[];
-    private Mat                  mRgbaInnerWindow;
-    private Mat                  mGrayInnerWindow;
-    private Mat                  mZoomWindow;
-    private Mat                  mZoomCorner;
     private Mat                  mSepiaKernel;
 
     public static int           viewMode = VIEW_MODE_RGBA;
@@ -166,13 +157,9 @@ public class ImageManipulationsActivity extends Activity implements CvCameraView
     }
 
     public void onCameraViewStarted(int width, int height) {
-        mGray = new Mat();
-        mRgba = new Mat();
         mIntermediateMat = new Mat();
         mSize0 = new Size();
-        mHist = new Mat();
         mChannels = new MatOfInt[] { new MatOfInt(0), new MatOfInt(1), new MatOfInt(2) };
-        mHistSizeNum = 25;
         mBuff = new float[mHistSizeNum];
         mHistSize = new MatOfInt(mHistSizeNum);
         mRanges = new MatOfFloat(0f, 256f);
@@ -197,14 +184,22 @@ public class ImageManipulationsActivity extends Activity implements CvCameraView
         mSepiaKernel.put(3, 0, /* A */0.000f, 0.000f, 0.000f, 1f);
     }
 
-    private void CreateAuxiliaryMats() {
-        if (mRgba.empty())
-            return;
+    public void onCameraViewStopped() {
+        // Explicitly deallocate Mats
+        if (mIntermediateMat != null)
+            mIntermediateMat.release();
 
-        mSizeRgba = mRgba.size();
+        mIntermediateMat = null;
+    }
 
-        int rows = (int) mSizeRgba.height;
-        int cols = (int) mSizeRgba.width;
+    public Mat onCameraFrame(CvCameraViewFrame inputFrame) {
+        Mat rgba = inputFrame.rgba();
+        Size sizeRgba = rgba.size();
+
+        Mat rgbaInnerWindow;
+
+        int rows = (int) sizeRgba.height;
+        int cols = (int) sizeRgba.width;
 
         int left = cols / 8;
         int top = rows / 8;
@@ -212,151 +207,107 @@ public class ImageManipulationsActivity extends Activity implements CvCameraView
         int width = cols * 3 / 4;
         int height = rows * 3 / 4;
 
-        if (mRgbaInnerWindow == null)
-            mRgbaInnerWindow = mRgba.submat(top, top + height, left, left + width);
-        mSizeRgbaInner = mRgbaInnerWindow.size();
-
-        if (mGrayInnerWindow == null && !mGray.empty())
-            mGrayInnerWindow = mGray.submat(top, top + height, left, left + width);
-
-        if (mZoomCorner == null)
-            mZoomCorner = mRgba.submat(0, rows / 2 - rows / 10, 0, cols / 2 - cols / 10);
-
-        if (mZoomWindow == null)
-            mZoomWindow = mRgba.submat(rows / 2 - 9 * rows / 100, rows / 2 + 9 * rows / 100, cols / 2 - 9 * cols / 100, cols / 2 + 9 * cols / 100);
-    }
-
-    public void onCameraViewStopped() {
-        // Explicitly deallocate Mats
-        if (mZoomWindow != null)
-            mZoomWindow.release();
-        if (mZoomCorner != null)
-            mZoomCorner.release();
-        if (mGrayInnerWindow != null)
-            mGrayInnerWindow.release();
-        if (mRgbaInnerWindow != null)
-            mRgbaInnerWindow.release();
-        if (mRgba != null)
-            mRgba.release();
-        if (mGray != null)
-            mGray.release();
-        if (mIntermediateMat != null)
-            mIntermediateMat.release();
-
-        mRgba = null;
-        mGray = null;
-        mIntermediateMat = null;
-        mRgbaInnerWindow = null;
-        mGrayInnerWindow = null;
-        mZoomCorner = null;
-        mZoomWindow = null;
-    }
-
-    public Mat onCameraFrame(CvCameraViewFrame inputFrame) {
-        mRgba = inputFrame.rgba();
-
         switch (ImageManipulationsActivity.viewMode) {
         case ImageManipulationsActivity.VIEW_MODE_RGBA:
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_HIST:
-            if ((mSizeRgba == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-            int thikness = (int) (mSizeRgba.width / (mHistSizeNum + 10) / 5);
+            Mat hist = new Mat();
+            int thikness = (int) (sizeRgba.width / (mHistSizeNum + 10) / 5);
             if(thikness > 5) thikness = 5;
-            int offset = (int) ((mSizeRgba.width - (5*mHistSizeNum + 4*10)*thikness)/2);
+            int offset = (int) ((sizeRgba.width - (5*mHistSizeNum + 4*10)*thikness)/2);
             // RGB
             for(int c=0; c<3; c++) {
-                Imgproc.calcHist(Arrays.asList(mRgba), mChannels[c], mMat0, mHist, mHistSize, mRanges);
-                Core.normalize(mHist, mHist, mSizeRgba.height/2, 0, Core.NORM_INF);
-                mHist.get(0, 0, mBuff);
+                Imgproc.calcHist(Arrays.asList(rgba), mChannels[c], mMat0, hist, mHistSize, mRanges);
+                Core.normalize(hist, hist, sizeRgba.height/2, 0, Core.NORM_INF);
+                hist.get(0, 0, mBuff);
                 for(int h=0; h<mHistSizeNum; h++) {
                     mP1.x = mP2.x = offset + (c * (mHistSizeNum + 10) + h) * thikness;
-                    mP1.y = mSizeRgba.height-1;
+                    mP1.y = sizeRgba.height-1;
                     mP2.y = mP1.y - 2 - (int)mBuff[h];
-                    Core.line(mRgba, mP1, mP2, mColorsRGB[c], thikness);
+                    Core.line(rgba, mP1, mP2, mColorsRGB[c], thikness);
                 }
             }
             // Value and Hue
-            Imgproc.cvtColor(mRgba, mIntermediateMat, Imgproc.COLOR_RGB2HSV_FULL);
+            Imgproc.cvtColor(rgba, mIntermediateMat, Imgproc.COLOR_RGB2HSV_FULL);
             // Value
-            Imgproc.calcHist(Arrays.asList(mIntermediateMat), mChannels[2], mMat0, mHist, mHistSize, mRanges);
-            Core.normalize(mHist, mHist, mSizeRgba.height/2, 0, Core.NORM_INF);
-            mHist.get(0, 0, mBuff);
+            Imgproc.calcHist(Arrays.asList(mIntermediateMat), mChannels[2], mMat0, hist, mHistSize, mRanges);
+            Core.normalize(hist, hist, sizeRgba.height/2, 0, Core.NORM_INF);
+            hist.get(0, 0, mBuff);
             for(int h=0; h<mHistSizeNum; h++) {
                 mP1.x = mP2.x = offset + (3 * (mHistSizeNum + 10) + h) * thikness;
-                mP1.y = mSizeRgba.height-1;
+                mP1.y = sizeRgba.height-1;
                 mP2.y = mP1.y - 2 - (int)mBuff[h];
-                Core.line(mRgba, mP1, mP2, mWhilte, thikness);
+                Core.line(rgba, mP1, mP2, mWhilte, thikness);
             }
             // Hue
-            Imgproc.calcHist(Arrays.asList(mIntermediateMat), mChannels[0], mMat0, mHist, mHistSize, mRanges);
-            Core.normalize(mHist, mHist, mSizeRgba.height/2, 0, Core.NORM_INF);
-            mHist.get(0, 0, mBuff);
+            Imgproc.calcHist(Arrays.asList(mIntermediateMat), mChannels[0], mMat0, hist, mHistSize, mRanges);
+            Core.normalize(hist, hist, sizeRgba.height/2, 0, Core.NORM_INF);
+            hist.get(0, 0, mBuff);
             for(int h=0; h<mHistSizeNum; h++) {
                 mP1.x = mP2.x = offset + (4 * (mHistSizeNum + 10) + h) * thikness;
-                mP1.y = mSizeRgba.height-1;
+                mP1.y = sizeRgba.height-1;
                 mP2.y = mP1.y - 2 - (int)mBuff[h];
-                Core.line(mRgba, mP1, mP2, mColorsHue[h], thikness);
+                Core.line(rgba, mP1, mP2, mColorsHue[h], thikness);
             }
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_CANNY:
-             if ((mRgbaInnerWindow == null) || (mGrayInnerWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-            Imgproc.Canny(mRgbaInnerWindow, mIntermediateMat, 80, 90);
-            Imgproc.cvtColor(mIntermediateMat, mRgbaInnerWindow, Imgproc.COLOR_GRAY2BGRA, 4);
+            rgbaInnerWindow = rgba.submat(top, top + height, left, left + width);
+            Imgproc.Canny(rgbaInnerWindow, mIntermediateMat, 80, 90);
+            Imgproc.cvtColor(mIntermediateMat, rgbaInnerWindow, Imgproc.COLOR_GRAY2BGRA, 4);
+            rgbaInnerWindow.release();
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_SOBEL:
-            mGray = inputFrame.gray();
-
-            if ((mRgbaInnerWindow == null) || (mGrayInnerWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-
-            Imgproc.Sobel(mGrayInnerWindow, mIntermediateMat, CvType.CV_8U, 1, 1);
+            Mat gray = inputFrame.gray();
+            Mat grayInnerWindow = gray.submat(top, top + height, left, left + width);
+            rgbaInnerWindow = rgba.submat(top, top + height, left, left + width);
+            Imgproc.Sobel(grayInnerWindow, mIntermediateMat, CvType.CV_8U, 1, 1);
             Core.convertScaleAbs(mIntermediateMat, mIntermediateMat, 10, 0);
-            Imgproc.cvtColor(mIntermediateMat, mRgbaInnerWindow, Imgproc.COLOR_GRAY2BGRA, 4);
+            Imgproc.cvtColor(mIntermediateMat, rgbaInnerWindow, Imgproc.COLOR_GRAY2BGRA, 4);
+            grayInnerWindow.release();
+            rgbaInnerWindow.release();
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_SEPIA:
-            if ((mRgbaInnerWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-            Core.transform(mRgbaInnerWindow, mRgbaInnerWindow, mSepiaKernel);
+            rgbaInnerWindow = rgba.submat(top, top + height, left, left + width);
+            Core.transform(rgbaInnerWindow, rgbaInnerWindow, mSepiaKernel);
+            rgbaInnerWindow.release();
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_ZOOM:
-            if ((mZoomCorner == null) || (mZoomWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-            Imgproc.resize(mZoomWindow, mZoomCorner, mZoomCorner.size());
-
+            Mat zoomCorner = rgba.submat(0, rows / 2 - rows / 10, 0, cols / 2 - cols / 10);
+            Mat mZoomWindow = rgba.submat(rows / 2 - 9 * rows / 100, rows / 2 + 9 * rows / 100, cols / 2 - 9 * cols / 100, cols / 2 + 9 * cols / 100);
+            Imgproc.resize(mZoomWindow, zoomCorner, zoomCorner.size());
             Size wsize = mZoomWindow.size();
             Core.rectangle(mZoomWindow, new Point(1, 1), new Point(wsize.width - 2, wsize.height - 2), new Scalar(255, 0, 0, 255), 2);
+            zoomCorner.release();
+            mZoomWindow.release();
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_PIXELIZE:
-            if ((mRgbaInnerWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
-            Imgproc.resize(mRgbaInnerWindow, mIntermediateMat, mSize0, 0.1, 0.1, Imgproc.INTER_NEAREST);
-            Imgproc.resize(mIntermediateMat, mRgbaInnerWindow, mSizeRgbaInner, 0., 0., Imgproc.INTER_NEAREST);
+            rgbaInnerWindow = rgba.submat(top, top + height, left, left + width);
+            Imgproc.resize(rgbaInnerWindow, mIntermediateMat, mSize0, 0.1, 0.1, Imgproc.INTER_NEAREST);
+            Imgproc.resize(mIntermediateMat, rgbaInnerWindow, rgbaInnerWindow.size(), 0., 0., Imgproc.INTER_NEAREST);
+            rgbaInnerWindow.release();
             break;
 
         case ImageManipulationsActivity.VIEW_MODE_POSTERIZE:
-            if ((mRgbaInnerWindow == null) || (mRgba.cols() != mSizeRgba.width) || (mRgba.height() != mSizeRgba.height))
-                CreateAuxiliaryMats();
             /*
-            Imgproc.cvtColor(mRgbaInnerWindow, mIntermediateMat, Imgproc.COLOR_RGBA2RGB);
+            Imgproc.cvtColor(rgbaInnerWindow, mIntermediateMat, Imgproc.COLOR_RGBA2RGB);
             Imgproc.pyrMeanShiftFiltering(mIntermediateMat, mIntermediateMat, 5, 50);
-            Imgproc.cvtColor(mIntermediateMat, mRgbaInnerWindow, Imgproc.COLOR_RGB2RGBA);
+            Imgproc.cvtColor(mIntermediateMat, rgbaInnerWindow, Imgproc.COLOR_RGB2RGBA);
             */
-
-            Imgproc.Canny(mRgbaInnerWindow, mIntermediateMat, 80, 90);
-            mRgbaInnerWindow.setTo(new Scalar(0, 0, 0, 255), mIntermediateMat);
-            Core.convertScaleAbs(mRgbaInnerWindow, mIntermediateMat, 1./16, 0);
-            Core.convertScaleAbs(mIntermediateMat, mRgbaInnerWindow, 16, 0);
+            rgbaInnerWindow = rgba.submat(top, top + height, left, left + width);
+            Imgproc.Canny(rgbaInnerWindow, mIntermediateMat, 80, 90);
+            rgbaInnerWindow.setTo(new Scalar(0, 0, 0, 255), mIntermediateMat);
+            Core.convertScaleAbs(rgbaInnerWindow, mIntermediateMat, 1./16, 0);
+            Core.convertScaleAbs(mIntermediateMat, rgbaInnerWindow, 16, 0);
+            rgbaInnerWindow.release();
             break;
         }
 
-        return mRgba;
+        return rgba;
     }
 }
diff --git a/samples/android/native-activity/.classpath b/samples/android/native-activity/.classpath
new file mode 100644
index 0000000000..3f9691c5dd
--- /dev/null
+++ b/samples/android/native-activity/.classpath
@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<classpath>
+	<classpathentry kind="con" path="com.android.ide.eclipse.adt.ANDROID_FRAMEWORK"/>
+	<classpathentry kind="con" path="com.android.ide.eclipse.adt.LIBRARIES"/>
+	<classpathentry kind="src" path="src"/>
+	<classpathentry kind="src" path="gen"/>
+	<classpathentry kind="output" path="bin/classes"/>
+</classpath>
diff --git a/samples/android/native-activity/.cproject b/samples/android/native-activity/.cproject
new file mode 100644
index 0000000000..83ca04b55b
--- /dev/null
+++ b/samples/android/native-activity/.cproject
@@ -0,0 +1,61 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<?fileVersion 4.0.0?>
+
+<cproject storage_type_id="org.eclipse.cdt.core.XmlProjectDescriptionStorage">
+	<storageModule moduleId="org.eclipse.cdt.core.settings">
+		<cconfiguration id="0.882924228">
+			<storageModule buildSystemId="org.eclipse.cdt.managedbuilder.core.configurationDataProvider" id="0.882924228" moduleId="org.eclipse.cdt.core.settings" name="Default">
+				<externalSettings/>
+				<extensions>
+					<extension id="org.eclipse.cdt.core.VCErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
+					<extension id="org.eclipse.cdt.core.GmakeErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
+					<extension id="org.eclipse.cdt.core.CWDLocator" point="org.eclipse.cdt.core.ErrorParser"/>
+					<extension id="org.eclipse.cdt.core.GCCErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
+					<extension id="org.eclipse.cdt.core.GASErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
+					<extension id="org.eclipse.cdt.core.GLDErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
+				</extensions>
+			</storageModule>
+			<storageModule moduleId="cdtBuildSystem" version="4.0.0">
+				<configuration artifactName="${ProjName}" buildProperties="" description="" id="0.882924228" name="Default" parent="org.eclipse.cdt.build.core.prefbase.cfg">
+					<folderInfo id="0.882924228." name="/" resourcePath="">
+						<toolChain id="org.eclipse.cdt.build.core.prefbase.toolchain.1667980868" name="No ToolChain" resourceTypeBasedDiscovery="false" superClass="org.eclipse.cdt.build.core.prefbase.toolchain">
+							<targetPlatform id="org.eclipse.cdt.build.core.prefbase.toolchain.1667980868.2108168132" name=""/>
+							<builder autoBuildTarget="" command="${NDKROOT}/ndk-build.cmd" enableAutoBuild="true" enableCleanBuild="false" id="org.eclipse.cdt.build.core.settings.default.builder.328915772" incrementalBuildTarget="" keepEnvironmentInBuildfile="false" managedBuildOn="false" name="Gnu Make Builder" superClass="org.eclipse.cdt.build.core.settings.default.builder"/>
+							<tool id="org.eclipse.cdt.build.core.settings.holder.libs.630148311" name="holder for library settings" superClass="org.eclipse.cdt.build.core.settings.holder.libs"/>
+							<tool id="org.eclipse.cdt.build.core.settings.holder.525090327" name="Assembly" superClass="org.eclipse.cdt.build.core.settings.holder">
+								<inputType id="org.eclipse.cdt.build.core.settings.holder.inType.1491216279" languageId="org.eclipse.cdt.core.assembly" languageName="Assembly" sourceContentType="org.eclipse.cdt.core.asmSource" superClass="org.eclipse.cdt.build.core.settings.holder.inType"/>
+							</tool>
+							<tool id="org.eclipse.cdt.build.core.settings.holder.1242729366" name="GNU C++" superClass="org.eclipse.cdt.build.core.settings.holder">
+								<option id="org.eclipse.cdt.build.core.settings.holder.incpaths.881377735" name="Include Paths" superClass="org.eclipse.cdt.build.core.settings.holder.incpaths" valueType="includePath">
+									<listOptionValue builtIn="false" value="&quot;${NDKROOT}/platforms/android-9/arch-arm/usr/include&quot;"/>
+									<listOptionValue builtIn="false" value="&quot;${NDKROOT}/sources/cxx-stl/gnu-libstdc++/4.6/include&quot;"/>
+									<listOptionValue builtIn="false" value="&quot;${NDKROOT}/sources/cxx-stl/gnu-libstdc++/4.6/libs/armeabi-v7a/include&quot;"/>
+									<listOptionValue builtIn="false" value="&quot;${ProjDirPath}/../../sdk/native/jni/include&quot;"/>
+									<listOptionValue builtIn="false" value="&quot;${NDKROOT}/sources/android/native_app_glue&quot;"/>
+								</option>
+								<inputType id="org.eclipse.cdt.build.core.settings.holder.inType.273216997" languageId="org.eclipse.cdt.core.g++" languageName="GNU C++" sourceContentType="org.eclipse.cdt.core.cxxSource,org.eclipse.cdt.core.cxxHeader" superClass="org.eclipse.cdt.build.core.settings.holder.inType"/>
+							</tool>
+							<tool id="org.eclipse.cdt.build.core.settings.holder.1779128177" name="GNU C" superClass="org.eclipse.cdt.build.core.settings.holder">
+								<inputType id="org.eclipse.cdt.build.core.settings.holder.inType.1778510041" languageId="org.eclipse.cdt.core.gcc" languageName="GNU C" sourceContentType="org.eclipse.cdt.core.cSource,org.eclipse.cdt.core.cHeader" superClass="org.eclipse.cdt.build.core.settings.holder.inType"/>
+							</tool>
+						</toolChain>
+					</folderInfo>
+				</configuration>
+			</storageModule>
+			<storageModule moduleId="org.eclipse.cdt.core.externalSettings"/>
+		</cconfiguration>
+	</storageModule>
+	<storageModule moduleId="cdtBuildSystem" version="4.0.0">
+		<project id="CvNativeActivity.null.708321898" name="CvNativeActivity"/>
+	</storageModule>
+	<storageModule moduleId="scannerConfiguration">
+		<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId=""/>
+		<scannerConfigBuildInfo instanceId="0.882924228">
+			<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId=""/>
+		</scannerConfigBuildInfo>
+	</storageModule>
+	<storageModule moduleId="refreshScope" versionNumber="1">
+		<resource resourceType="PROJECT" workspacePath="/CvNativeActivity"/>
+	</storageModule>
+	<storageModule moduleId="org.eclipse.cdt.internal.ui.text.commentOwnerProjectMappings"/>
+</cproject>
diff --git a/samples/android/native-activity/.project b/samples/android/native-activity/.project
new file mode 100644
index 0000000000..c20be83f60
--- /dev/null
+++ b/samples/android/native-activity/.project
@@ -0,0 +1,101 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<projectDescription>
+	<name>OpenCV Sample - native-activity</name>
+	<comment></comment>
+	<projects>
+	</projects>
+	<buildSpec>
+		<buildCommand>
+			<name>org.eclipse.cdt.managedbuilder.core.genmakebuilder</name>
+			<triggers>auto,full,incremental,</triggers>
+			<arguments>
+				<dictionary>
+					<key>?name?</key>
+					<value></value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.append_environment</key>
+					<value>true</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.autoBuildTarget</key>
+					<value></value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.buildArguments</key>
+					<value></value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.buildCommand</key>
+					<value>&quot;${NDKROOT}/ndk-build.cmd&quot;</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.cleanBuildTarget</key>
+					<value>clean</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.contents</key>
+					<value>org.eclipse.cdt.make.core.activeConfigSettings</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.enableAutoBuild</key>
+					<value>true</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.enableCleanBuild</key>
+					<value>false</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.enableFullBuild</key>
+					<value>true</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.fullBuildTarget</key>
+					<value></value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.stopOnError</key>
+					<value>true</value>
+				</dictionary>
+				<dictionary>
+					<key>org.eclipse.cdt.make.core.useDefaultBuildCmd</key>
+					<value>false</value>
+				</dictionary>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.ResourceManagerBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.PreCompilerBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>org.eclipse.jdt.core.javabuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>com.android.ide.eclipse.adt.ApkBuilder</name>
+			<arguments>
+			</arguments>
+		</buildCommand>
+		<buildCommand>
+			<name>org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder</name>
+			<triggers>full,incremental,</triggers>
+			<arguments>
+			</arguments>
+		</buildCommand>
+	</buildSpec>
+	<natures>
+		<nature>com.android.ide.eclipse.adt.AndroidNature</nature>
+		<nature>org.eclipse.jdt.core.javanature</nature>
+		<nature>org.eclipse.cdt.core.cnature</nature>
+		<nature>org.eclipse.cdt.core.ccnature</nature>
+		<nature>org.eclipse.cdt.managedbuilder.core.managedBuildNature</nature>
+		<nature>org.eclipse.cdt.managedbuilder.core.ScannerConfigNature</nature>
+	</natures>
+</projectDescription>
diff --git a/samples/android/native-activity/.settings/org.eclipse.jdt.core.prefs b/samples/android/native-activity/.settings/org.eclipse.jdt.core.prefs
new file mode 100644
index 0000000000..b080d2ddc8
--- /dev/null
+++ b/samples/android/native-activity/.settings/org.eclipse.jdt.core.prefs
@@ -0,0 +1,4 @@
+eclipse.preferences.version=1
+org.eclipse.jdt.core.compiler.codegen.targetPlatform=1.6
+org.eclipse.jdt.core.compiler.compliance=1.6
+org.eclipse.jdt.core.compiler.source=1.6
diff --git a/samples/android/native-activity/AndroidManifest.xml b/samples/android/native-activity/AndroidManifest.xml
new file mode 100644
index 0000000000..55b696ba47
--- /dev/null
+++ b/samples/android/native-activity/AndroidManifest.xml
@@ -0,0 +1,36 @@
+<?xml version="1.0" encoding="utf-8"?>
+<manifest xmlns:android="http://schemas.android.com/apk/res/android"
+      package="org.opencv.samples.NativeActivity"
+      android:versionCode="1"
+      android:versionName="1.0">
+    <uses-sdk android:minSdkVersion="9" />
+    <application android:label="@string/app_name"
+                 android:icon="@drawable/icon"
+                 android:debuggable="true">
+
+        <activity android:name="CvNativeActivity"
+                  android:label="@string/app_name"
+                  android:screenOrientation="landscape"
+                  android:configChanges="orientation|keyboardHidden">
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN" />
+                    <category android:name="android.intent.category.LAUNCHER" />
+                </intent-filter>
+            </activity>
+        <activity android:name="android.app.NativeActivity"
+                  android:label="@string/app_name"
+                  android:screenOrientation="landscape"
+                  android:configChanges="keyboardHidden|orientation">
+            <meta-data android:name="android.app.lib_name"
+                    android:value="native_activity" />
+        </activity>
+    </application>
+
+   <uses-permission android:name="android.permission.CAMERA"/>
+
+   <uses-feature android:name="android.hardware.camera" android:required="false"/>
+   <uses-feature android:name="android.hardware.camera.autofocus" android:required="false"/>
+   <uses-feature android:name="android.hardware.camera.front" android:required="false"/>
+   <uses-feature android:name="android.hardware.camera.front.autofocus" android:required="false"/>
+
+</manifest>
diff --git a/samples/android/native-activity/CMakeLists.txt b/samples/android/native-activity/CMakeLists.txt
new file mode 100644
index 0000000000..1f6d97439a
--- /dev/null
+++ b/samples/android/native-activity/CMakeLists.txt
@@ -0,0 +1,12 @@
+set(sample example-native-activity)
+
+if(BUILD_FAT_JAVA_LIB)
+  set(native_deps opencv_java)
+else()
+  set(native_deps opencv_highgui opencv_imgproc)
+endif()
+
+add_android_project(${sample} "${CMAKE_CURRENT_SOURCE_DIR}" LIBRARY_DEPS ${OpenCV_BINARY_DIR} SDK_TARGET 9 ${ANDROID_SDK_TARGET} NATIVE_DEPS ${native_deps})
+if(TARGET ${sample})
+  add_dependencies(opencv_android_examples ${sample})
+endif()
diff --git a/samples/android/native-activity/jni/Android.mk b/samples/android/native-activity/jni/Android.mk
new file mode 100644
index 0000000000..7ae31e24ca
--- /dev/null
+++ b/samples/android/native-activity/jni/Android.mk
@@ -0,0 +1,14 @@
+LOCAL_PATH := $(call my-dir)
+
+include $(CLEAR_VARS)
+
+include ../../sdk/native/jni/OpenCV.mk
+
+LOCAL_MODULE    := native_activity
+LOCAL_SRC_FILES := native.cpp
+LOCAL_LDLIBS    += -lm -llog -landroid
+LOCAL_STATIC_LIBRARIES += android_native_app_glue
+
+include $(BUILD_SHARED_LIBRARY)
+
+$(call import-module,android/native_app_glue)
diff --git a/samples/android/native-activity/jni/Application.mk b/samples/android/native-activity/jni/Application.mk
new file mode 100644
index 0000000000..e9392cfed2
--- /dev/null
+++ b/samples/android/native-activity/jni/Application.mk
@@ -0,0 +1,4 @@
+APP_ABI := armeabi-v7a
+APP_STL := gnustl_static
+APP_CPPFLAGS := -frtti -fexceptions
+APP_PLATFORM := android-9
diff --git a/samples/android/native-activity/jni/native.cpp b/samples/android/native-activity/jni/native.cpp
new file mode 100644
index 0000000000..052b866e62
--- /dev/null
+++ b/samples/android/native-activity/jni/native.cpp
@@ -0,0 +1,233 @@
+#include <android_native_app_glue.h>
+
+#include <errno.h>
+#include <jni.h>
+#include <sys/time.h>
+#include <time.h>
+#include <android/log.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <queue>
+
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/highgui.hpp>
+
+#define  LOG_TAG    "OCV:libnative_activity"
+#define  LOGD(...)  __android_log_print(ANDROID_LOG_DEBUG,LOG_TAG,__VA_ARGS__)
+#define  LOGI(...)  __android_log_print(ANDROID_LOG_INFO,LOG_TAG,__VA_ARGS__)
+#define  LOGW(...)  __android_log_print(ANDROID_LOG_WARN,LOG_TAG,__VA_ARGS__)
+#define  LOGE(...)  __android_log_print(ANDROID_LOG_ERROR,LOG_TAG,__VA_ARGS__)
+
+struct Engine
+{
+    android_app* app;
+    cv::Ptr<cv::VideoCapture> capture;
+};
+
+static cv::Size calc_optimal_camera_resolution(const char* supported, int width, int height)
+{
+    int frame_width = 0;
+    int frame_height = 0;
+
+    size_t prev_idx = 0;
+    size_t idx = 0;
+    float min_diff = FLT_MAX;
+
+    do
+    {
+        int tmp_width;
+        int tmp_height;
+
+        prev_idx = idx;
+        while ((supported[idx] != '\0') && (supported[idx] != ','))
+            idx++;
+
+        sscanf(&supported[prev_idx], "%dx%d", &tmp_width, &tmp_height);
+
+        int w_diff = width - tmp_width;
+        int h_diff = height - tmp_height;
+        if ((h_diff >= 0) && (w_diff >= 0))
+        {
+            if ((h_diff <= min_diff) && (tmp_height <= 720))
+            {
+                frame_width = tmp_width;
+                frame_height = tmp_height;
+                min_diff = h_diff;
+            }
+        }
+
+        idx++; // to skip comma symbol
+
+    } while(supported[idx-1] != '\0');
+
+    return cv::Size(frame_width, frame_height);
+}
+
+static void engine_draw_frame(Engine* engine, const cv::Mat& frame)
+{
+    if (engine->app->window == NULL)
+        return; // No window.
+
+    ANativeWindow_Buffer buffer;
+    if (ANativeWindow_lock(engine->app->window, &buffer, NULL) < 0)
+    {
+        LOGW("Unable to lock window buffer");
+        return;
+    }
+
+    int32_t* pixels = (int32_t*)buffer.bits;
+
+    int left_indent = (buffer.width-frame.cols)/2;
+    int top_indent = (buffer.height-frame.rows)/2;
+
+    if (top_indent > 0)
+    {
+        memset(pixels, 0, top_indent*buffer.stride*sizeof(int32_t));
+        pixels += top_indent*buffer.stride;
+    }
+
+    for (int yy = 0; yy < frame.rows; yy++)
+    {
+        if (left_indent > 0)
+        {
+            memset(pixels, 0, left_indent*sizeof(int32_t));
+            memset(pixels+left_indent+frame.cols, 0, (buffer.stride-frame.cols-left_indent)*sizeof(int32_t));
+        }
+        int32_t* line = pixels + left_indent;
+        size_t line_size = frame.cols*4*sizeof(unsigned char);
+        memcpy(line, frame.ptr<unsigned char>(yy), line_size);
+        // go to next line
+        pixels += buffer.stride;
+    }
+    ANativeWindow_unlockAndPost(engine->app->window);
+}
+
+static void engine_handle_cmd(android_app* app, int32_t cmd)
+{
+    Engine* engine = (Engine*)app->userData;
+    switch (cmd)
+    {
+        case APP_CMD_INIT_WINDOW:
+            if (app->window != NULL)
+            {
+                LOGI("APP_CMD_INIT_WINDOW");
+
+                engine->capture = cv::makePtr<cv::VideoCapture>(0);
+
+                union {double prop; const char* name;} u;
+                u.prop = engine->capture->get(cv::CAP_PROP_ANDROID_PREVIEW_SIZES_STRING);
+
+                int view_width = ANativeWindow_getWidth(app->window);
+                int view_height = ANativeWindow_getHeight(app->window);
+
+                cv::Size camera_resolution;
+                if (u.name)
+                    camera_resolution = calc_optimal_camera_resolution(u.name, 640, 480);
+                else
+                {
+                    LOGE("Cannot get supported camera camera_resolutions");
+                    camera_resolution = cv::Size(ANativeWindow_getWidth(app->window),
+                                          ANativeWindow_getHeight(app->window));
+                }
+
+                if ((camera_resolution.width != 0) && (camera_resolution.height != 0))
+                {
+                    engine->capture->set(cv::CAP_PROP_FRAME_WIDTH, camera_resolution.width);
+                    engine->capture->set(cv::CAP_PROP_FRAME_HEIGHT, camera_resolution.height);
+                }
+
+                float scale = std::min((float)view_width/camera_resolution.width,
+                                       (float)view_height/camera_resolution.height);
+
+                if (ANativeWindow_setBuffersGeometry(app->window, (int)(view_width/scale),
+                    int(view_height/scale), WINDOW_FORMAT_RGBA_8888) < 0)
+                {
+                    LOGE("Cannot set pixel format!");
+                    return;
+                }
+
+                LOGI("Camera initialized at resolution %dx%d", camera_resolution.width, camera_resolution.height);
+            }
+            break;
+        case APP_CMD_TERM_WINDOW:
+            LOGI("APP_CMD_TERM_WINDOW");
+
+            engine->capture->release();
+            break;
+    }
+}
+
+void android_main(android_app* app)
+{
+    Engine engine;
+
+    // Make sure glue isn't stripped.
+    app_dummy();
+
+    size_t engine_size = sizeof(engine); // for Eclipse CDT parser
+    memset((void*)&engine, 0, engine_size);
+    app->userData = &engine;
+    app->onAppCmd = engine_handle_cmd;
+    engine.app = app;
+
+    float fps = 0;
+    cv::Mat drawing_frame;
+    std::queue<int64> time_queue;
+
+    // loop waiting for stuff to do.
+    while (1)
+    {
+        // Read all pending events.
+        int ident;
+        int events;
+        android_poll_source* source;
+
+        // Process system events
+        while ((ident=ALooper_pollAll(0, NULL, &events, (void**)&source)) >= 0)
+        {
+            // Process this event.
+            if (source != NULL)
+            {
+                source->process(app, source);
+            }
+
+            // Check if we are exiting.
+            if (app->destroyRequested != 0)
+            {
+                LOGI("Engine thread destroy requested!");
+                return;
+            }
+        }
+
+        int64 then;
+        int64 now = cv::getTickCount();
+        time_queue.push(now);
+
+        // Capture frame from camera and draw it
+        if (!engine.capture.empty())
+        {
+            if (engine.capture->grab())
+                engine.capture->retrieve(drawing_frame, cv::CAP_ANDROID_COLOR_FRAME_RGBA);
+
+             char buffer[256];
+             sprintf(buffer, "Display performance: %dx%d @ %.3f", drawing_frame.cols, drawing_frame.rows, fps);
+             cv::putText(drawing_frame, std::string(buffer), cv::Point(8,64),
+                         cv::FONT_HERSHEY_COMPLEX_SMALL, 1, cv::Scalar(0,255,0,255));
+             engine_draw_frame(&engine, drawing_frame);
+        }
+
+        if (time_queue.size() >= 2)
+            then = time_queue.front();
+        else
+            then = 0;
+
+        if (time_queue.size() >= 25)
+            time_queue.pop();
+
+        fps = time_queue.size() * (float)cv::getTickFrequency() / (now-then);
+    }
+}
diff --git a/samples/android/native-activity/res/drawable/icon.png b/samples/android/native-activity/res/drawable/icon.png
new file mode 100644
index 0000000000..630454927b
Binary files /dev/null and b/samples/android/native-activity/res/drawable/icon.png differ
diff --git a/samples/android/native-activity/res/values/strings.xml b/samples/android/native-activity/res/values/strings.xml
new file mode 100644
index 0000000000..888534efe0
--- /dev/null
+++ b/samples/android/native-activity/res/values/strings.xml
@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <string name="app_name">OCV Native Activity</string>
+</resources>
diff --git a/samples/android/native-activity/src/org/opencv/samples/NativeActivity/CvNativeActivity.java b/samples/android/native-activity/src/org/opencv/samples/NativeActivity/CvNativeActivity.java
new file mode 100644
index 0000000000..04da9a9496
--- /dev/null
+++ b/samples/android/native-activity/src/org/opencv/samples/NativeActivity/CvNativeActivity.java
@@ -0,0 +1,43 @@
+package org.opencv.samples.NativeActivity;
+
+import org.opencv.android.BaseLoaderCallback;
+import org.opencv.android.LoaderCallbackInterface;
+import org.opencv.android.OpenCVLoader;
+
+import android.app.Activity;
+import android.content.Intent;
+import android.util.Log;
+
+public class CvNativeActivity extends Activity {
+    private static final String TAG = "OCVSample::Activity";
+
+    private BaseLoaderCallback mLoaderCallback = new BaseLoaderCallback(this) {
+        @Override
+        public void onManagerConnected(int status) {
+            switch (status) {
+                case LoaderCallbackInterface.SUCCESS:
+                {
+                    Log.i(TAG, "OpenCV loaded successfully");
+                    System.loadLibrary("native_activity");
+                    Intent intent = new Intent(CvNativeActivity.this, android.app.NativeActivity.class);
+                    CvNativeActivity.this.startActivity(intent);
+                } break;
+                default:
+                {
+                    super.onManagerConnected(status);
+                } break;
+            }
+        }
+    };
+
+    public CvNativeActivity() {
+        Log.i(TAG, "Instantiated new " + this.getClass());
+    }
+
+   @Override
+    public void onResume()
+    {
+        super.onResume();
+        OpenCVLoader.initAsync(OpenCVLoader.OPENCV_VERSION_2_4_3, this, mLoaderCallback);
+    }
+}
diff --git a/samples/android/tutorial-1-camerapreview/res/layout/tutorial1_surface_view.xml b/samples/android/tutorial-1-camerapreview/res/layout/tutorial1_surface_view.xml
index 52b9dfaf6e..40f5f49326 100644
--- a/samples/android/tutorial-1-camerapreview/res/layout/tutorial1_surface_view.xml
+++ b/samples/android/tutorial-1-camerapreview/res/layout/tutorial1_surface_view.xml
@@ -1,4 +1,4 @@
-<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
+<FrameLayout xmlns:android="http://schemas.android.com/apk/res/android"
     xmlns:tools="http://schemas.android.com/tools"
     xmlns:opencv="http://schemas.android.com/apk/res-auto"
     android:layout_width="match_parent"
@@ -20,4 +20,4 @@
         opencv:show_fps="true"
         opencv:camera_id="any" />
 
-</LinearLayout>
+</FrameLayout>
diff --git a/samples/android/tutorial-2-mixedprocessing/jni/jni_part.cpp b/samples/android/tutorial-2-mixedprocessing/jni/jni_part.cpp
index e7ed75d213..f8e3ada726 100644
--- a/samples/android/tutorial-2-mixedprocessing/jni/jni_part.cpp
+++ b/samples/android/tutorial-2-mixedprocessing/jni/jni_part.cpp
@@ -8,7 +8,7 @@ using namespace std;
 using namespace cv;
 
 extern "C" {
-JNIEXPORT void JNICALL Java_org_opencv_samples_tutorial2_Tuturial2Activity_FindFeatures(JNIEnv*, jobject, jlong addrGray, jlong addrRgba);
+JNIEXPORT void JNICALL Java_org_opencv_samples_tutorial2_Tutorial2Activity_FindFeatures(JNIEnv*, jobject, jlong addrGray, jlong addrRgba);
 
 JNIEXPORT void JNICALL Java_org_opencv_samples_tutorial2_Tutorial2Activity_FindFeatures(JNIEnv*, jobject, jlong addrGray, jlong addrRgba)
 {
diff --git a/samples/android/tutorial-3-cameracontrol/res/layout/tutorial3_surface_view.xml b/samples/android/tutorial-3-cameracontrol/res/layout/tutorial3_surface_view.xml
index 79fef4eae4..a574905782 100644
--- a/samples/android/tutorial-3-cameracontrol/res/layout/tutorial3_surface_view.xml
+++ b/samples/android/tutorial-3-cameracontrol/res/layout/tutorial3_surface_view.xml
@@ -4,8 +4,8 @@
     android:layout_height="match_parent" >
 
     <org.opencv.samples.tutorial3.Tutorial3View
-        android:layout_width="fill_parent"
-        android:layout_height="fill_parent"
+        android:layout_width="match_parent"
+        android:layout_height="match_parent"
         android:visibility="gone"
         android:id="@+id/tutorial3_activity_java_surface_view" />
 
diff --git a/samples/android/tutorial-3-cameracontrol/src/org/opencv/samples/tutorial3/Tutorial3View.java b/samples/android/tutorial-3-cameracontrol/src/org/opencv/samples/tutorial3/Tutorial3View.java
index 04c163ff88..d102cbcdef 100644
--- a/samples/android/tutorial-3-cameracontrol/src/org/opencv/samples/tutorial3/Tutorial3View.java
+++ b/samples/android/tutorial-3-cameracontrol/src/org/opencv/samples/tutorial3/Tutorial3View.java
@@ -6,17 +6,16 @@ import java.util.List;
 import org.opencv.android.JavaCameraView;
 
 import android.content.Context;
-import android.graphics.Bitmap;
-import android.graphics.BitmapFactory;
 import android.hardware.Camera;
 import android.hardware.Camera.PictureCallback;
 import android.hardware.Camera.Size;
 import android.util.AttributeSet;
 import android.util.Log;
 
-public class Tutorial3View extends JavaCameraView {
+public class Tutorial3View extends JavaCameraView implements PictureCallback {
 
     private static final String TAG = "Sample::Tutorial3View";
+    private String mPictureFileName;
 
     public Tutorial3View(Context context, AttributeSet attrs) {
         super(context, attrs);
@@ -56,26 +55,33 @@ public class Tutorial3View extends JavaCameraView {
     }
 
     public void takePicture(final String fileName) {
-        Log.i(TAG, "Tacking picture");
-        PictureCallback callback = new PictureCallback() {
+        Log.i(TAG, "Taking picture");
+        this.mPictureFileName = fileName;
+        // Postview and jpeg are sent in the same buffers if the queue is not empty when performing a capture.
+        // Clear up buffers to avoid mCamera.takePicture to be stuck because of a memory issue
+        mCamera.setPreviewCallback(null);
 
-            private String mPictureFileName = fileName;
+        // PictureCallback is implemented by the current class
+        mCamera.takePicture(null, null, this);
+    }
 
-            @Override
-            public void onPictureTaken(byte[] data, Camera camera) {
-                Log.i(TAG, "Saving a bitmap to file");
-                Bitmap picture = BitmapFactory.decodeByteArray(data, 0, data.length);
-                try {
-                    FileOutputStream out = new FileOutputStream(mPictureFileName);
-                    picture.compress(Bitmap.CompressFormat.JPEG, 90, out);
-                    picture.recycle();
-                    mCamera.startPreview();
-                } catch (Exception e) {
-                    e.printStackTrace();
-                }
-            }
-        };
+    @Override
+    public void onPictureTaken(byte[] data, Camera camera) {
+        Log.i(TAG, "Saving a bitmap to file");
+        // The camera preview was automatically stopped. Start it again.
+        mCamera.startPreview();
+        mCamera.setPreviewCallback(this);
+
+        // Write the image in a file (in jpeg format)
+        try {
+            FileOutputStream fos = new FileOutputStream(mPictureFileName);
+
+            fos.write(data);
+            fos.close();
+
+        } catch (java.io.IOException e) {
+            Log.e("PictureDemo", "Exception in photoCallback", e);
+        }
 
-        mCamera.takePicture(null, null, callback);
     }
 }
diff --git a/samples/c/CMakeLists.txt b/samples/c/CMakeLists.txt
index dfab5e00af..7ea20b9b2f 100644
--- a/samples/c/CMakeLists.txt
+++ b/samples/c/CMakeLists.txt
@@ -57,4 +57,3 @@ if (INSTALL_C_EXAMPLES AND NOT WIN32)
           DESTINATION share/OpenCV/samples/c
           PERMISSIONS OWNER_READ GROUP_READ WORLD_READ)
 endif ()
-
diff --git a/samples/c/adaptiveskindetector.cpp b/samples/c/adaptiveskindetector.cpp
index b81bff92d2..f94c7a871c 100644
--- a/samples/c/adaptiveskindetector.cpp
+++ b/samples/c/adaptiveskindetector.cpp
@@ -39,8 +39,8 @@
 #include <cstdio>
 #include <cstring>
 #include <ctime>
-#include "opencv2/contrib/contrib.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/contrib/compat.hpp"
+#include "opencv2/highgui/highgui_c.h"
 
 static void help(char **argv)
 {
@@ -410,4 +410,3 @@ int main(int argc, char** argv )
 
     return 0;
 }
-
diff --git a/samples/c/bgfg_codebook.cpp b/samples/c/bgfg_codebook.cpp
index 3054e9971b..eba0b71bea 100644
--- a/samples/c/bgfg_codebook.cpp
+++ b/samples/c/bgfg_codebook.cpp
@@ -20,11 +20,12 @@
      Or: http://oreilly.com/catalog/9780596516130/
      ISBN-10: 0596516134 or: ISBN-13: 978-0596516130
 ************************************************** */
-#include "opencv2/core/core.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/video/background_segm.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/legacy/legacy.hpp"
+#include "opencv2/highgui/highgui_c.h"
+#include "opencv2/legacy.hpp"
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/samples/c/blobtrack_sample.cpp b/samples/c/blobtrack_sample.cpp
index d8209fa305..55f72eb209 100644
--- a/samples/c/blobtrack_sample.cpp
+++ b/samples/c/blobtrack_sample.cpp
@@ -1,7 +1,7 @@
 #include "opencv2/video/background_segm.hpp"
 #include "opencv2/legacy/blobtrack.hpp"
 #include "opencv2/legacy/legacy.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/highgui/highgui_c.h>
 #include <opencv2/imgproc/imgproc_c.h>
 
 #include <stdio.h>
@@ -751,6 +751,3 @@ int main(int argc, char* argv[])
     return 0;
 
 }   /* main() */
-
-
-
diff --git a/samples/c/build_all.sh b/samples/c/build_all.sh
index 787474aa2f..94a5a055c6 100755
--- a/samples/c/build_all.sh
+++ b/samples/c/build_all.sh
@@ -1,16 +1,16 @@
 #!/bin/sh
 
 if [ $# -gt 0 ] ; then
-	base=`basename $1 .c`
-	echo "compiling $base"
-	gcc -ggdb `pkg-config opencv --cflags --libs` $base.c -o $base 
+    base=`basename $1 .c`
+    echo "compiling $base"
+    gcc -ggdb `pkg-config opencv --cflags --libs` $base.c -o $base
 else
-	for i in *.c; do
-	    echo "compiling $i"
-	    gcc -ggdb `pkg-config --cflags opencv` -o `basename $i .c` $i `pkg-config --libs opencv`;
-	done
-	for i in *.cpp; do
-	    echo "compiling $i"
-	    g++ -ggdb `pkg-config --cflags opencv` -o `basename $i .cpp` $i `pkg-config --libs opencv`;
-	done
+    for i in *.c; do
+        echo "compiling $i"
+        gcc -ggdb `pkg-config --cflags opencv` -o `basename $i .c` $i `pkg-config --libs opencv`;
+    done
+    for i in *.cpp; do
+        echo "compiling $i"
+        g++ -ggdb `pkg-config --cflags opencv` -o `basename $i .cpp` $i `pkg-config --libs opencv`;
+    done
 fi
diff --git a/samples/c/convert_cascade.c b/samples/c/convert_cascade.c
index 16c0fa6b19..4337e35e26 100644
--- a/samples/c/convert_cascade.c
+++ b/samples/c/convert_cascade.c
@@ -1,5 +1,5 @@
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
+#include "opencv2/highgui/highgui_c.h"
 
 #include <ctype.h>
 #include <stdio.h>
diff --git a/samples/c/delaunay.c b/samples/c/delaunay.c
index 21fe340a3a..1f09efef90 100644
--- a/samples/c/delaunay.c
+++ b/samples/c/delaunay.c
@@ -1,6 +1,6 @@
 #include <opencv2/imgproc/imgproc_c.h>
-#include <opencv2/legacy/legacy.hpp>
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/highgui/highgui_c.h>
+#include <opencv2/legacy.hpp>
 
 #include <stdio.h>
 static void help( void )
diff --git a/samples/c/example_cmake/CMakeLists.txt b/samples/c/example_cmake/CMakeLists.txt
index 85129a7528..475677a090 100644
--- a/samples/c/example_cmake/CMakeLists.txt
+++ b/samples/c/example_cmake/CMakeLists.txt
@@ -13,6 +13,3 @@ ADD_EXECUTABLE(opencv_example  minarea.c)
 TARGET_LINK_LIBRARIES(opencv_example ${OpenCV_LIBS})
 
 #MESSAGE(STATUS "OpenCV_LIBS: ${OpenCV_LIBS}")
-
-
-
diff --git a/samples/c/example_cmake/README.txt b/samples/c/example_cmake/README.txt
index 7bf53e7acc..7d573e3732 100644
--- a/samples/c/example_cmake/README.txt
+++ b/samples/c/example_cmake/README.txt
@@ -25,8 +25,3 @@ the CMake gui with:
  $ cmake-gui <OPENCV_SRC_PATH>/samples/c/example_cmake/
 
 And pick the correct value for OpenCV_DIR.
-
-
-
-
-
diff --git a/samples/c/example_cmake/minarea.c b/samples/c/example_cmake/minarea.c
index 916acb0efe..c6e0cf310c 100644
--- a/samples/c/example_cmake/minarea.c
+++ b/samples/c/example_cmake/minarea.c
@@ -114,4 +114,3 @@ int main( int argc, char** argv )
 #ifdef _EiC
 main(1,"convexhull.c");
 #endif
-
diff --git a/samples/c/facedetect.cpp b/samples/c/facedetect.cpp
index a7d7f4adba..4c5250664d 100644
--- a/samples/c/facedetect.cpp
+++ b/samples/c/facedetect.cpp
@@ -1,6 +1,9 @@
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/highgui/highgui_c.h"
 
 #include <cctype>
 #include <iostream>
@@ -124,7 +127,7 @@ int main( int argc, const char** argv )
         for(;;)
         {
             IplImage* iplImg = cvQueryFrame( capture );
-            frame = iplImg;
+            frame = cv::cvarrToMat(iplImg);
             if( frame.empty() )
                 break;
             if( iplImg->origin == IPL_ORIGIN_TL )
@@ -206,16 +209,16 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         CV_RGB(255,0,255)} ;
     Mat gray, smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
 
-    cvtColor( img, gray, CV_BGR2GRAY );
+    cvtColor( img, gray, COLOR_BGR2GRAY );
     resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
     equalizeHist( smallImg, smallImg );
 
     t = (double)cvGetTickCount();
     cascade.detectMultiScale( smallImg, faces,
         1.1, 2, 0
-        //|CV_HAAR_FIND_BIGGEST_OBJECT
-        //|CV_HAAR_DO_ROUGH_SEARCH
-        |CV_HAAR_SCALE_IMAGE
+        //|CASCADE_FIND_BIGGEST_OBJECT
+        //|CASCADE_DO_ROUGH_SEARCH
+        |CASCADE_SCALE_IMAGE
         ,
         Size(30, 30) );
     if( tryflip )
@@ -223,9 +226,9 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         flip(smallImg, smallImg, 1);
         cascade.detectMultiScale( smallImg, faces2,
                                  1.1, 2, 0
-                                 //|CV_HAAR_FIND_BIGGEST_OBJECT
-                                 //|CV_HAAR_DO_ROUGH_SEARCH
-                                 |CV_HAAR_SCALE_IMAGE
+                                 //|CASCADE_FIND_BIGGEST_OBJECT
+                                 //|CASCADE_DO_ROUGH_SEARCH
+                                 |CASCADE_SCALE_IMAGE
                                  ,
                                  Size(30, 30) );
         for( vector<Rect>::const_iterator r = faces2.begin(); r != faces2.end(); r++ )
@@ -260,10 +263,10 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         smallImgROI = smallImg(*r);
         nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
             1.1, 2, 0
-            //|CV_HAAR_FIND_BIGGEST_OBJECT
-            //|CV_HAAR_DO_ROUGH_SEARCH
-            //|CV_HAAR_DO_CANNY_PRUNING
-            |CV_HAAR_SCALE_IMAGE
+            //|CASCADE_FIND_BIGGEST_OBJECT
+            //|CASCADE_DO_ROUGH_SEARCH
+            //|CASCADE_DO_CANNY_PRUNING
+            |CASCADE_SCALE_IMAGE
             ,
             Size(30, 30) );
         for( vector<Rect>::const_iterator nr = nestedObjects.begin(); nr != nestedObjects.end(); nr++ )
diff --git a/samples/c/fback_c.c b/samples/c/fback_c.c
index af613d25e4..e0317a4423 100644
--- a/samples/c/fback_c.c
+++ b/samples/c/fback_c.c
@@ -1,5 +1,5 @@
-#include "opencv2/video/tracking.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/video/tracking_c.h"
+#include "opencv2/highgui/highgui_c.h"
 #include "opencv2/imgproc/imgproc_c.h"
 #include <stdio.h>
 
diff --git a/samples/c/find_obj.cpp b/samples/c/find_obj.cpp
index e0431fc6d4..651f439245 100644
--- a/samples/c/find_obj.cpp
+++ b/samples/c/find_obj.cpp
@@ -6,10 +6,10 @@
  */
 #include "opencv2/objdetect/objdetect.hpp"
 #include "opencv2/features2d/features2d.hpp"
-#include "opencv2/highgui/highgui.hpp"
 #include "opencv2/calib3d/calib3d.hpp"
 #include "opencv2/nonfree/nonfree.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/highgui/highgui_c.h"
 #include "opencv2/legacy/legacy.hpp"
 #include "opencv2/legacy/compat.hpp"
 
@@ -231,17 +231,17 @@ int main(int argc, char** argv)
     cvNamedWindow("Object", 1);
     cvNamedWindow("Object Correspond", 1);
 
-    static CvScalar colors[] =
+    static cv::Scalar colors[] =
     {
-        {{0,0,255}},
-        {{0,128,255}},
-        {{0,255,255}},
-        {{0,255,0}},
-        {{255,128,0}},
-        {{255,255,0}},
-        {{255,0,0}},
-        {{255,0,255}},
-        {{255,255,255}}
+        cv::Scalar(0,0,255),
+        cv::Scalar(0,128,255),
+        cv::Scalar(0,255,255),
+        cv::Scalar(0,255,0),
+        cv::Scalar(255,128,0),
+        cv::Scalar(255,255,0),
+        cv::Scalar(255,0,0),
+        cv::Scalar(255,0,255),
+        cv::Scalar(255,255,255)
     };
 
     IplImage* object_color = cvCreateImage(cvGetSize(object), 8, 3);
@@ -262,7 +262,7 @@ int main(int argc, char** argv)
 
     printf( "Extraction time = %gms\n", tt/(cvGetTickFrequency()*1000.));
 
-    CvPoint src_corners[4] = {{0,0}, {object->width,0}, {object->width, object->height}, {0, object->height}};
+    CvPoint src_corners[4] = {CvPoint(0,0), CvPoint(object->width,0), CvPoint(object->width, object->height), CvPoint(0, object->height)};
     CvPoint dst_corners[4];
     IplImage* correspond = cvCreateImage( cvSize(image->width, object->height+image->height), 8, 1 );
     cvSetImageROI( correspond, cvRect( 0, 0, object->width, object->height ) );
diff --git a/samples/c/find_obj_calonder.cpp b/samples/c/find_obj_calonder.cpp
index 02cd266526..cb6551778f 100644
--- a/samples/c/find_obj_calonder.cpp
+++ b/samples/c/find_obj_calonder.cpp
@@ -61,7 +61,7 @@ static void trainCalonderClassifier( const string& classifierFilename, const str
         string str;
         getline( is, str );
         if (str.empty()) break;
-        Mat img = imread( str, CV_LOAD_IMAGE_GRAYSCALE );
+        Mat img = imread( str, IMREAD_GRAYSCALE );
         if( !img.empty() )
             trainImgs.push_back( img );
     }
@@ -106,7 +106,7 @@ static void trainCalonderClassifier( const string& classifierFilename, const str
  */
 static void testCalonderClassifier( const string& classifierFilename, const string& imgFilename )
 {
-    Mat img1 = imread( imgFilename, CV_LOAD_IMAGE_GRAYSCALE ), img2, H12;
+    Mat img1 = imread( imgFilename, IMREAD_GRAYSCALE ), img2, H12;
     if( img1.empty() )
     {
         cout << "Test image can not be read." << endl;
diff --git a/samples/c/find_obj_ferns.cpp b/samples/c/find_obj_ferns.cpp
index 9d19c198f4..cd48684f97 100644
--- a/samples/c/find_obj_ferns.cpp
+++ b/samples/c/find_obj_ferns.cpp
@@ -1,9 +1,9 @@
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/legacy/legacy.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/legacy.hpp"
 
 #include <algorithm>
 #include <iostream>
@@ -32,8 +32,8 @@ int main(int argc, char** argv)
 
     help();
 
-    Mat object = imread( object_filename, CV_LOAD_IMAGE_GRAYSCALE );
-    Mat scene = imread( scene_filename, CV_LOAD_IMAGE_GRAYSCALE );
+    Mat object = imread( object_filename, IMREAD_GRAYSCALE );
+    Mat scene = imread( scene_filename, IMREAD_GRAYSCALE );
 
     if( !object.data || !scene.data )
     {
@@ -47,9 +47,9 @@ int main(int argc, char** argv)
 
     resize(scene, image, Size(), 1./imgscale, 1./imgscale, INTER_CUBIC);
 
-    cvNamedWindow("Object", 1);
-    cvNamedWindow("Image", 1);
-    cvNamedWindow("Object Correspondence", 1);
+    namedWindow("Object", 1);
+    namedWindow("Image", 1);
+    namedWindow("Object Correspondence", 1);
 
     Size patchSize(32, 32);
     LDetector ldetector(7, 20, 2, 2000, patchSize.width, 2);
diff --git a/samples/c/latentsvmdetect.cpp b/samples/c/latentsvmdetect.cpp
index e74b227a62..83a078e5d0 100644
--- a/samples/c/latentsvmdetect.cpp
+++ b/samples/c/latentsvmdetect.cpp
@@ -1,14 +1,8 @@
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
+#include "opencv2/highgui/highgui_c.h"
+#include "opencv2/core/utility.hpp"
 #include <stdio.h>
 
-#ifdef HAVE_CVCONFIG_H
-#include <cvconfig.h>
-#endif
-#ifdef HAVE_TBB
-#include "tbb/task_scheduler_init.h"
-#endif
-
 using namespace cv;
 
 static void help()
@@ -31,28 +25,17 @@ static void detect_and_draw_objects( IplImage* image, CvLatentSvmDetector* detec
     CvSeq* detections = 0;
     int i = 0;
     int64 start = 0, finish = 0;
-#ifdef HAVE_TBB
-    tbb::task_scheduler_init init(tbb::task_scheduler_init::deferred);
-    if (numThreads > 0)
-    {
-        init.initialize(numThreads);
-        printf("Number of threads %i\n", numThreads);
-    }
-    else
-    {
-        printf("Number of threads is not correct for TBB version");
-        return;
-    }
-#endif
+
+    setNumThreads(numThreads);
+    numThreads = getNumThreads();
+    printf("Number of threads %i\n", numThreads);
 
     start = cvGetTickCount();
     detections = cvLatentSvmDetectObjects(image, detector, storage, 0.5f, numThreads);
     finish = cvGetTickCount();
     printf("detection time = %.3f\n", (float)(finish - start) / (float)(cvGetTickFrequency() * 1000000.0));
+    setNumThreads(-1);
 
-#ifdef HAVE_TBB
-    init.terminate();
-#endif
     for( i = 0; i < detections->total; i++ )
     {
         CvObjectDetection detection = *(CvObjectDetection*)cvGetSeqElem( detections, i );
diff --git a/samples/c/morphology.c b/samples/c/morphology.c
index b71046eea0..fc2e2f195c 100644
--- a/samples/c/morphology.c
+++ b/samples/c/morphology.c
@@ -1,4 +1,4 @@
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/highgui/highgui_c.h"
 #include "opencv2/imgproc/imgproc_c.h"
 
 #include <stdio.h>
diff --git a/samples/c/motempl.c b/samples/c/motempl.c
index df089f590f..76b69ade7b 100644
--- a/samples/c/motempl.c
+++ b/samples/c/motempl.c
@@ -1,6 +1,6 @@
-#include "opencv2/video/tracking.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/video/tracking_c.h"
 #include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/highgui/highgui_c.h"
 #include <time.h>
 #include <stdio.h>
 #include <ctype.h>
diff --git a/samples/c/mser_sample.cpp b/samples/c/mser_sample.cpp
index e0217a2d3a..b6b96c3434 100644
--- a/samples/c/mser_sample.cpp
+++ b/samples/c/mser_sample.cpp
@@ -2,9 +2,10 @@
  * Copyright (C) 2009, Liu Liu All rights reserved.
  */
 
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/highgui.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include <iostream>
 #include <stdio.h>
diff --git a/samples/c/mushroom.cpp b/samples/c/mushroom.cpp
index 5763b12464..60eb9f066c 100644
--- a/samples/c/mushroom.cpp
+++ b/samples/c/mushroom.cpp
@@ -90,7 +90,7 @@ static int mushroom_read_database( const char* filename, CvMat** data, CvMat** m
     }
 
     cvReleaseMemStorage( &storage );
-    delete el_ptr;
+    delete [] el_ptr;
     return 1;
 }
 
diff --git a/samples/c/one_way_sample.cpp b/samples/c/one_way_sample.cpp
index c017976b06..d2ca7f4009 100644
--- a/samples/c/one_way_sample.cpp
+++ b/samples/c/one_way_sample.cpp
@@ -49,8 +49,8 @@ int main(int argc, char** argv)
     std::string img2_name = path_name + "/" + std::string(argv[3]);
 
     printf("Reading the images...\n");
-    Mat img1 = imread(img1_name, CV_LOAD_IMAGE_GRAYSCALE);
-    Mat img2 = imread(img2_name, CV_LOAD_IMAGE_GRAYSCALE);
+    Mat img1 = imread(img1_name, IMREAD_GRAYSCALE);
+    Mat img2 = imread(img2_name, IMREAD_GRAYSCALE);
 
     // extract keypoints from the first image
     SURF surf_extractor(5.0e3);
diff --git a/samples/c/polar_transforms.c b/samples/c/polar_transforms.c
index d6915d488c..8ce4831637 100644
--- a/samples/c/polar_transforms.c
+++ b/samples/c/polar_transforms.c
@@ -1,7 +1,5 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-
 #include "opencv2/imgproc/imgproc_c.h"
+#include "opencv2/highgui/highgui_c.h"
 
 #include <ctype.h>
 #include <stdio.h>
diff --git a/samples/c/pyramid_segmentation.c b/samples/c/pyramid_segmentation.c
index 780dc9f301..c15c7cd0c0 100644
--- a/samples/c/pyramid_segmentation.c
+++ b/samples/c/pyramid_segmentation.c
@@ -1,7 +1,6 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
 #include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/legacy/legacy.hpp"
+#include "opencv2/highgui/highgui_c.h"
+#include "opencv2/legacy.hpp"
 #include <stdio.h>
 
 static void help(void)
diff --git a/samples/c/smiledetect.cpp b/samples/c/smiledetect.cpp
index c54c724398..61f990cd2d 100644
--- a/samples/c/smiledetect.cpp
+++ b/samples/c/smiledetect.cpp
@@ -1,6 +1,9 @@
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/core/utility.hpp"
+
+#include "opencv2/highgui/highgui_c.h"
 
 #include <cctype>
 #include <iostream>
@@ -120,7 +123,7 @@ int main( int argc, const char** argv )
         for(;;)
         {
             IplImage* iplImg = cvQueryFrame( capture );
-            frame = iplImg;
+            frame = cv::cvarrToMat(iplImg);
             if( frame.empty() )
                 break;
             if( iplImg->origin == IPL_ORIGIN_TL )
@@ -166,15 +169,15 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         CV_RGB(255,0,255)} ;
     Mat gray, smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
 
-    cvtColor( img, gray, CV_BGR2GRAY );
+    cvtColor( img, gray, COLOR_BGR2GRAY );
     resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
     equalizeHist( smallImg, smallImg );
 
     cascade.detectMultiScale( smallImg, faces,
         1.1, 2, 0
-        //|CV_HAAR_FIND_BIGGEST_OBJECT
-        //|CV_HAAR_DO_ROUGH_SEARCH
-        |CV_HAAR_SCALE_IMAGE
+        //|CASCADE_FIND_BIGGEST_OBJECT
+        //|CASCADE_DO_ROUGH_SEARCH
+        |CASCADE_SCALE_IMAGE
         ,
         Size(30, 30) );
     if( tryflip )
@@ -182,9 +185,9 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         flip(smallImg, smallImg, 1);
         cascade.detectMultiScale( smallImg, faces2,
                                  1.1, 2, 0
-                                 //|CV_HAAR_FIND_BIGGEST_OBJECT
-                                 //|CV_HAAR_DO_ROUGH_SEARCH
-                                 |CV_HAAR_SCALE_IMAGE
+                                 //|CASCADE_FIND_BIGGEST_OBJECT
+                                 //|CASCADE_DO_ROUGH_SEARCH
+                                 |CASCADE_SCALE_IMAGE
                                  ,
                                  Size(30, 30) );
         for( vector<Rect>::const_iterator r = faces2.begin(); r != faces2.end(); r++ )
@@ -220,10 +223,10 @@ void detectAndDraw( Mat& img, CascadeClassifier& cascade,
         smallImgROI = smallImg(*r);
         nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
             1.1, 0, 0
-            //|CV_HAAR_FIND_BIGGEST_OBJECT
-            //|CV_HAAR_DO_ROUGH_SEARCH
-            //|CV_HAAR_DO_CANNY_PRUNING
-            |CV_HAAR_SCALE_IMAGE
+            //|CASCADE_FIND_BIGGEST_OBJECT
+            //|CASCADE_DO_ROUGH_SEARCH
+            //|CASCADE_DO_CANNY_PRUNING
+            |CASCADE_SCALE_IMAGE
             ,
             Size(30, 30) );
 
diff --git a/samples/c/tree_engine.cpp b/samples/c/tree_engine.cpp
index 74993ae490..2c3046fd73 100644
--- a/samples/c/tree_engine.cpp
+++ b/samples/c/tree_engine.cpp
@@ -1,5 +1,6 @@
 #include "opencv2/ml/ml.hpp"
 #include "opencv2/core/core_c.h"
+#include "opencv2/core/utility.hpp"
 #include <stdio.h>
 #include <map>
 
@@ -21,7 +22,7 @@ static void help()
 
 static int count_classes(CvMLData& data)
 {
-    cv::Mat r(data.get_responses());
+    cv::Mat r = cv::cvarrToMat(data.get_responses());
     std::map<int, int> rmap;
     int i, n = (int)r.total();
     for( i = 0; i < n; i++ )
@@ -42,7 +43,7 @@ static void print_result(float train_err, float test_err, const CvMat* _var_imp)
 
     if (_var_imp)
     {
-        cv::Mat var_imp(_var_imp), sorted_idx;
+        cv::Mat var_imp = cv::cvarrToMat(_var_imp), sorted_idx;
         cv::sortIdx(var_imp, sorted_idx, CV_SORT_EVERY_ROW + CV_SORT_DESCENDING);
 
         printf( "variable importance:\n" );
diff --git a/samples/cpp/3calibration.cpp b/samples/cpp/3calibration.cpp
index 061e977952..a6ee45d853 100644
--- a/samples/cpp/3calibration.cpp
+++ b/samples/cpp/3calibration.cpp
@@ -81,14 +81,14 @@ static bool run3Calibration( vector<vector<Point2f> > imagePoints1,
         objpt.resize(imgpt.size(),objpt[0]);
 
         Mat cameraMatrix = Mat::eye(3, 3, CV_64F);
-        if( flags & CV_CALIB_FIX_ASPECT_RATIO )
+        if( flags & CALIB_FIX_ASPECT_RATIO )
             cameraMatrix.at<double>(0,0) = aspectRatio;
 
         Mat distCoeffs = Mat::zeros(5, 1, CV_64F);
 
         double err = calibrateCamera(objpt, imgpt, imageSize, cameraMatrix,
                         distCoeffs, rvecs, tvecs,
-                        flags|CV_CALIB_FIX_K3/*|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5|CV_CALIB_FIX_K6*/);
+                        flags|CALIB_FIX_K3/*|CALIB_FIX_K4|CALIB_FIX_K5|CALIB_FIX_K6*/);
         bool ok = checkRange(cameraMatrix) && checkRange(distCoeffs);
         if(!ok)
         {
@@ -138,7 +138,7 @@ static bool run3Calibration( vector<vector<Point2f> > imagePoints1,
                                      cameraMatrix, distCoeffs,
                                      imageSize, R, T, E, F,
                                      TermCriteria(TermCriteria::COUNT, 30, 0),
-                                     CV_CALIB_FIX_INTRINSIC);
+                                     CALIB_FIX_INTRINSIC);
         printf("Pair (1,%d) calibration reprojection error = %g\n", c, sqrt(err/(N*2)));
         if( c == 2 )
         {
@@ -212,15 +212,15 @@ int main( int argc, char** argv )
         {
             if( sscanf( argv[++i], "%f", &aspectRatio ) != 1 || aspectRatio <= 0 )
                 return printf("Invalid aspect ratio\n" ), -1;
-            flags |= CV_CALIB_FIX_ASPECT_RATIO;
+            flags |= CALIB_FIX_ASPECT_RATIO;
         }
         else if( strcmp( s, "-zt" ) == 0 )
         {
-            flags |= CV_CALIB_ZERO_TANGENT_DIST;
+            flags |= CALIB_ZERO_TANGENT_DIST;
         }
         else if( strcmp( s, "-p" ) == 0 )
         {
-            flags |= CV_CALIB_FIX_PRINCIPAL_POINT;
+            flags |= CALIB_FIX_PRINCIPAL_POINT;
         }
         else if( strcmp( s, "-o" ) == 0 )
         {
@@ -271,8 +271,8 @@ int main( int argc, char** argv )
             {
                 vector<Point2f> ptvec;
                 imageSize = view.size();
-                cvtColor(view, viewGray, CV_BGR2GRAY);
-                bool found = findChessboardCorners( view, boardSize, ptvec, CV_CALIB_CB_ADAPTIVE_THRESH );
+                cvtColor(view, viewGray, COLOR_BGR2GRAY);
+                bool found = findChessboardCorners( view, boardSize, ptvec, CALIB_CB_ADAPTIVE_THRESH );
 
                 drawChessboardCorners( view, boardSize, Mat(ptvec), found );
                 if( found )
@@ -291,13 +291,13 @@ int main( int argc, char** argv )
     printf("Running calibration ...\n");
 
     run3Calibration(imgpt[0], imgpt[1], imgpt[2], imageSize,
-                    boardSize, squareSize, aspectRatio, flags|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5,
+                    boardSize, squareSize, aspectRatio, flags|CALIB_FIX_K4|CALIB_FIX_K5,
                     cameraMatrix[0], distCoeffs[0],
                     cameraMatrix[1], distCoeffs[1],
                     cameraMatrix[2], distCoeffs[2],
                     R12, T12, R13, T13);
 
-    fs.open(outputFilename, CV_STORAGE_WRITE);
+    fs.open(outputFilename, FileStorage::WRITE);
 
     fs << "cameraMatrix1" << cameraMatrix[0];
     fs << "cameraMatrix2" << cameraMatrix[1];
@@ -323,7 +323,7 @@ int main( int argc, char** argv )
              imgpt[0], imgpt[2],
              imageSize, R12, T12, R13, T13,
              R[0], R[1], R[2], P[0], P[1], P[2], Q, -1.,
-             imageSize, 0, 0, CV_CALIB_ZERO_DISPARITY);
+             imageSize, 0, 0, CALIB_ZERO_DISPARITY);
     Mat map1[3], map2[3];
 
     fs << "R1" << R[0];
@@ -359,7 +359,7 @@ int main( int argc, char** argv )
                 continue;
 
             Mat rview = canvas.colRange(k2*imageSize.width, (k2+1)*imageSize.width);
-            remap(view, rview, map1[k1], map2[k1], CV_INTER_LINEAR);
+            remap(view, rview, map1[k1], map2[k1], INTER_LINEAR);
         }
         printf("%s %s %s\n", imageList[i*3].c_str(), imageList[i*3+1].c_str(), imageList[i*3+2].c_str());
         resize( canvas, small_canvas, Size(1500, 1500/3) );
diff --git a/samples/cpp/CMakeLists.txt b/samples/cpp/CMakeLists.txt
index 05bbfe8740..4115bf9fce 100644
--- a/samples/cpp/CMakeLists.txt
+++ b/samples/cpp/CMakeLists.txt
@@ -5,7 +5,7 @@
 
 SET(OPENCV_CPP_SAMPLES_REQUIRED_DEPS opencv_core opencv_flann opencv_imgproc
     opencv_highgui opencv_ml opencv_video opencv_objdetect opencv_photo opencv_nonfree opencv_softcascade
-    opencv_features2d opencv_calib3d opencv_legacy opencv_contrib opencv_stitching opencv_videostab)
+    opencv_features2d opencv_calib3d opencv_legacy opencv_contrib opencv_stitching opencv_videostab opencv_bioinspired)
 
 ocv_check_dependencies(${OPENCV_CPP_SAMPLES_REQUIRED_DEPS})
 
@@ -16,8 +16,17 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
   ocv_include_directories("${OpenCV_SOURCE_DIR}/include")#for opencv.hpp
   ocv_include_modules(${OPENCV_CPP_SAMPLES_REQUIRED_DEPS})
 
-  if (HAVE_opencv_gpu)
-    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpu/include")
+  if(HAVE_opencv_gpuoptflow)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpuoptflow/include")
+  endif()
+  if(HAVE_opencv_gpuimgproc)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpuimgproc/include")
+  endif()
+  if(HAVE_opencv_gpuarithm)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpuarithm/include")
+  endif()
+  if(HAVE_opencv_gpufilters)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpufilters/include")
   endif()
 
   if(CMAKE_COMPILER_IS_GNUCXX AND NOT ENABLE_NOISY_WARNINGS)
@@ -32,17 +41,19 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
     if("${srcs}" MATCHES "tutorial_code")
       set(sample_kind tutorial)
       set(sample_KIND TUTORIAL)
+      set(sample_folder "samples//tutorials")
     else()
       set(sample_kind example)
       set(sample_KIND EXAMPLE)
+      set(sample_folder "samples//cpp")
     endif()
 
     set(the_target "${sample_kind}_${name}")
     add_executable(${the_target} ${srcs})
     target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_CPP_SAMPLES_REQUIRED_DEPS})
 
-    if (HAVE_opencv_gpu)
-      target_link_libraries(${the_target} opencv_gpu)
+    if("${srcs}" MATCHES "gpu/")
+      target_link_libraries(${the_target} opencv_gpuarithm opencv_gpufilters)
     endif()
 
     set_target_properties(${the_target} PROPERTIES
@@ -50,7 +61,7 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
       PROJECT_LABEL "(${sample_KIND}) ${name}")
 
     if(ENABLE_SOLUTION_FOLDERS)
-      set_target_properties(${the_target} PROPERTIES FOLDER "${sample_kind}s//cpp")
+      set_target_properties(${the_target} PROPERTIES FOLDER "${sample_folder}")
     endif()
 
     if(WIN32)
@@ -58,7 +69,7 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
         set_target_properties(${the_target} PROPERTIES LINK_FLAGS "/NODEFAULTLIB:atlthunk.lib /NODEFAULTLIB:atlsd.lib /DEBUG")
       endif()
       install(TARGETS ${the_target}
-              RUNTIME DESTINATION "${sample_kind}s/cpp" COMPONENT main)
+              RUNTIME DESTINATION "${sample_folder}" COMPONENT main)
     endif()
   ENDMACRO()
 
@@ -68,7 +79,7 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
     ocv_list_filterout(cpp_samples Qt_sample)
   endif()
 
-  if(NOT HAVE_opencv_gpu)
+  if(NOT HAVE_opencv_gpuarithm OR NOT HAVE_opencv_gpufilters)
     ocv_list_filterout(cpp_samples "/gpu/")
   endif()
 
@@ -84,4 +95,3 @@ if (INSTALL_C_EXAMPLES AND NOT WIN32)
           DESTINATION share/OpenCV/samples/cpp
           PERMISSIONS OWNER_READ GROUP_READ WORLD_READ)
 endif()
-
diff --git a/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp b/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp
index bfae1ebce4..539115a990 100644
--- a/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp
+++ b/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping.cpp
@@ -10,7 +10,9 @@
 #include <iostream>
 #include <cstring>
 
-#include "opencv2/opencv.hpp"
+#include "opencv2/bioinspired.hpp" // retina based algorithms
+#include "opencv2/imgproc.hpp" // cvCvtcolor function
+#include "opencv2/highgui.hpp" // display
 
 static void help(std::string errorMessage)
 {
@@ -28,7 +30,7 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
     cv::Mat displayedCurveImage = cv::Mat::ones(200, curve.size().height, CV_8U);
 
     cv::Mat windowNormalizedCurve;
-    normalize(curve, windowNormalizedCurve, 0, 200, CV_MINMAX, CV_32F);
+    normalize(curve, windowNormalizedCurve, 0, 200, cv::NORM_MINMAX, CV_32F);
 
     displayedCurveImage = cv::Scalar::all(255); // set a white background
     int binW = cvRound((double)displayedCurveImage.cols/curve.size().height);
@@ -69,7 +71,7 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
      {
          cv::Mat rgbIntImg;
          outputMat.convertTo(rgbIntImg, CV_8UC3);
-         cvtColor(rgbIntImg, intGrayImage, CV_BGR2GRAY);
+         cv::cvtColor(rgbIntImg, intGrayImage, cv::COLOR_BGR2GRAY);
      }
 
      // get histogram density probability in order to cut values under above edges limits (here 5-95%)... usefull for HDR pixel errors cancellation
@@ -80,8 +82,7 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
      normalize(hist, normalizedHist, 1, 0, cv::NORM_L1, CV_32F); // normalize histogram so that its sum equals 1
 
      double min_val, max_val;
-     CvMat histArr(normalizedHist);
-     cvMinMaxLoc(&histArr, &min_val, &max_val);
+     minMaxLoc(normalizedHist, &min_val, &max_val);
      //std::cout<<"Hist max,min = "<<max_val<<", "<<min_val<<std::endl;
 
      // compute density probability
@@ -127,7 +128,7 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
      normalize(imageInputRescaled, imageInputRescaled, 0.0, 255.0, cv::NORM_MINMAX);
  }
 
- cv::Ptr<cv::Retina> retina;
+ cv::Ptr<cv::bioinspired::Retina> retina;
  int retinaHcellsGain;
  int localAdaptation_photoreceptors, localAdaptation_Gcells;
  static void callBack_updateRetinaParams(int, void*)
@@ -175,6 +176,12 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
      }
 
      bool useLogSampling = !strcmp(argv[argc-1], "log"); // check if user wants retina log sampling processing
+     int chosenMethod=0;
+     if (!strcmp(argv[argc-1], "fast"))
+     {
+         chosenMethod=1;
+         std::cout<<"Using fast method (no spectral whithning), adaptation of Meylan&al 2008 method"<<std::endl;
+     }
 
      std::string inputImageName=argv[1];
 
@@ -210,17 +217,22 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
           * -> if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
           */
          if (useLogSampling)
-                {
-                     retina = new cv::Retina(inputImage.size(),true, cv::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+         {
+             retina = cv::bioinspired::createRetina(inputImage.size(),true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
                  }
          else// -> else allocate "classical" retina :
-             retina = new cv::Retina(inputImage.size());
+             retina = cv::bioinspired::createRetina(inputImage.size());
 
-        // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
-        retina->write("RetinaDefaultParameters.xml");
+         // create a fast retina tone mapper (Meyla&al algorithm)
+         std::cout<<"Allocating fast tone mapper..."<<std::endl;
+         //cv::Ptr<cv::RetinaFastToneMapping> fastToneMapper=createRetinaFastToneMapping(inputImage.size());
+         std::cout<<"Fast tone mapper allocated"<<std::endl;
 
-                 // desactivate Magnocellular pathway processing (motion information extraction) since it is not usefull here
-                 retina->activateMovingContoursProcessing(false);
+         // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
+         retina->write("RetinaDefaultParameters.xml");
+
+         // desactivate Magnocellular pathway processing (motion information extraction) since it is not usefull here
+         retina->activateMovingContoursProcessing(false);
 
          // declare retina output buffers
          cv::Mat retinaOutput_parvo;
@@ -230,20 +242,19 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
          histogramClippingValue=0; // default value... updated with interface slider
          //inputRescaleMat = inputImage;
          //outputRescaleMat = imageInputRescaled;
-         cv::namedWindow("Retina input image (with cut edges histogram for basic pixels error avoidance)",1);
-         cv::createTrackbar("histogram edges clipping limit", "Retina input image (with cut edges histogram for basic pixels error avoidance)",&histogramClippingValue,50,callBack_rescaleGrayLevelMat);
+         cv::namedWindow("Processing configuration",1);
+         cv::createTrackbar("histogram edges clipping limit", "Processing configuration",&histogramClippingValue,50,callBack_rescaleGrayLevelMat);
 
-         cv::namedWindow("Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", 1);
          colorSaturationFactor=3;
-         cv::createTrackbar("Color saturation", "Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", &colorSaturationFactor,5,callback_saturateColors);
+         cv::createTrackbar("Color saturation", "Processing configuration", &colorSaturationFactor,5,callback_saturateColors);
 
          retinaHcellsGain=40;
-         cv::createTrackbar("Hcells gain", "Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping",&retinaHcellsGain,100,callBack_updateRetinaParams);
+         cv::createTrackbar("Hcells gain", "Processing configuration",&retinaHcellsGain,100,callBack_updateRetinaParams);
 
          localAdaptation_photoreceptors=197;
          localAdaptation_Gcells=190;
-         cv::createTrackbar("Ph sensitivity", "Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", &localAdaptation_photoreceptors,199,callBack_updateRetinaParams);
-         cv::createTrackbar("Gcells sensitivity", "Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", &localAdaptation_Gcells,199,callBack_updateRetinaParams);
+         cv::createTrackbar("Ph sensitivity", "Processing configuration", &localAdaptation_photoreceptors,199,callBack_updateRetinaParams);
+         cv::createTrackbar("Gcells sensitivity", "Processing configuration", &localAdaptation_Gcells,199,callBack_updateRetinaParams);
 
 
          /////////////////////////////////////////////
@@ -257,11 +268,28 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
          while(continueProcessing)
          {
              // run retina filter
-             retina->run(imageInputRescaled);
-             // Retrieve and display retina output
-             retina->getParvo(retinaOutput_parvo);
-             cv::imshow("Retina input image (with cut edges histogram for basic pixels error avoidance)", imageInputRescaled/255.0);
-             cv::imshow("Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", retinaOutput_parvo);
+             if (!chosenMethod)
+             {
+                 retina->run(imageInputRescaled);
+                 // Retrieve and display retina output
+                 retina->getParvo(retinaOutput_parvo);
+                 cv::imshow("Retina input image (with cut edges histogram for basic pixels error avoidance)", imageInputRescaled/255.0);
+                 cv::imshow("Retina Parvocellular pathway output : 16bit=>8bit image retina tonemapping", retinaOutput_parvo);
+                 cv::imwrite("HDRinput.jpg",imageInputRescaled/255.0);
+                 cv::imwrite("RetinaToneMapping.jpg",retinaOutput_parvo);
+             }
+             else
+             {
+                 // apply the simplified hdr tone mapping method
+                 cv::Mat fastToneMappingOutput;
+                 retina->applyFastToneMapping(imageInputRescaled, fastToneMappingOutput);
+                 cv::imshow("Retina fast tone mapping output : 16bit=>8bit image retina tonemapping", fastToneMappingOutput);
+             }
+             /*cv::Mat fastToneMappingOutput_specificObject;
+             fastToneMapper->setup(3.f, 1.5f, 1.f);
+             fastToneMapper->applyFastToneMapping(imageInputRescaled, fastToneMappingOutput_specificObject);
+             cv::imshow("### Retina fast tone mapping output : 16bit=>8bit image retina tonemapping", fastToneMappingOutput_specificObject);
+*/
              cv::waitKey(10);
          }
      }catch(cv::Exception e)
@@ -274,5 +302,3 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
 
      return 0;
  }
-
-
diff --git a/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping_video.cpp b/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping_video.cpp
index 7c47fcf66e..646b1b7b76 100644
--- a/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping_video.cpp
+++ b/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping_video.cpp
@@ -14,7 +14,9 @@
 #include <stdio.h>
 #include <cstring>
 
-#include "opencv2/opencv.hpp"
+#include "opencv2/bioinspired.hpp" // retina based algorithms
+#include "opencv2/imgproc.hpp" // cvCvtcolor function
+#include "opencv2/highgui.hpp" // display
 
 static void help(std::string errorMessage)
 {
@@ -38,7 +40,7 @@ static void drawPlot(const cv::Mat curve, const std::string figureTitle, const i
     cv::Mat displayedCurveImage = cv::Mat::ones(200, curve.size().height, CV_8U);
 
     cv::Mat windowNormalizedCurve;
-    normalize(curve, windowNormalizedCurve, 0, 200, CV_MINMAX, CV_32F);
+    normalize(curve, windowNormalizedCurve, 0, 200, cv::NORM_MINMAX, CV_32F);
 
     displayedCurveImage = cv::Scalar::all(255); // set a white background
     int binW = cvRound((double)displayedCurveImage.cols/curve.size().height);
@@ -88,7 +90,7 @@ static void rescaleGrayLevelMat(const cv::Mat &inputMat, cv::Mat &outputMat, con
      {
          cv::Mat rgbIntImg;
          normalisedImage.convertTo(rgbIntImg, CV_8UC3);
-         cvtColor(rgbIntImg, intGrayImage, CV_BGR2GRAY);
+         cv::cvtColor(rgbIntImg, intGrayImage, cv::COLOR_BGR2GRAY);
      }
 
      // get histogram density probability in order to cut values under above edges limits (here 5-95%)... usefull for HDR pixel errors cancellation
@@ -159,7 +161,7 @@ static void rescaleGrayLevelMat(const cv::Mat &inputMat, cv::Mat &outputMat, con
 
  }
 
- cv::Ptr<cv::Retina> retina;
+ cv::Ptr<cv::bioinspired::Retina> retina;
  int retinaHcellsGain;
  int localAdaptation_photoreceptors, localAdaptation_Gcells;
  static void callBack_updateRetinaParams(int, void*)
@@ -279,10 +281,10 @@ static void loadNewFrame(const std::string filenamePrototype, const int currentF
           */
          if (useLogSampling)
                 {
-                     retina = new cv::Retina(inputImage.size(),true, cv::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+                     retina = cv::bioinspired::createRetina(inputImage.size(),true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
                  }
          else// -> else allocate "classical" retina :
-             retina = new cv::Retina(inputImage.size());
+             retina = cv::bioinspired::createRetina(inputImage.size());
 
         // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
         retina->write("RetinaDefaultParameters.xml");
@@ -357,5 +359,3 @@ static void loadNewFrame(const std::string filenamePrototype, const int currentF
 
      return 0;
  }
-
-
diff --git a/samples/cpp/Qt_sample/main.cpp b/samples/cpp/Qt_sample/main.cpp
index f987de471a..16345d56a2 100644
--- a/samples/cpp/Qt_sample/main.cpp
+++ b/samples/cpp/Qt_sample/main.cpp
@@ -4,7 +4,15 @@
 
 #include <iostream>
 #include <vector>
-#include <opencv/highgui.h>
+
+#include <opencv2/core/core_c.h>
+#include <opencv2/imgproc/imgproc_c.h>
+#include <opencv2/legacy/compat.hpp>
+#include <opencv2/calib3d/calib3d_c.h>
+
+#include <opencv2/imgproc.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/calib3d.hpp>
 
 #if defined WIN32 || defined _WIN32 || defined WINCE
     #include <windows.h>
@@ -20,9 +28,6 @@
     #include <GL/gl.h>
 #endif
 
-#include <opencv/cxcore.h>
-#include <opencv/cv.h>
-
 using namespace std;
 using namespace cv;
 
@@ -33,7 +38,7 @@ static void help()
             "It works off of the video: cube4.avi\n"
             "Using OpenCV version %s\n" << CV_VERSION << "\n\n"
 " 1). This demo is mainly based on work from Javier Barandiaran Martirena\n"
-"     See this page http://opencv.willowgarage.com/wiki/Posit.\n"
+"     See this page http://code.opencv.org/projects/opencv/wiki/Posit.\n"
 " 2). This is a demo to illustrate how to use **OpenGL Callback**.\n"
 " 3). You need Qt binding to compile this sample with OpenGL support enabled.\n"
 " 4). The features' detection is very basic and could highly be improved \n"
@@ -118,12 +123,12 @@ static void foundCorners(vector<CvPoint2D32f> *srcImagePoints,IplImage* source,
     cvNormalize(grayImage, grayImage, 0, 255, CV_MINMAX);
     cvThreshold( grayImage, grayImage, 26, 255, CV_THRESH_BINARY_INV);//25
 
-    Mat MgrayImage = grayImage;
+    Mat MgrayImage = cv::cvarrToMat(grayImage);
     //For debug
     //MgrayImage = MgrayImage.clone();//deep copy
     vector<vector<Point> > contours;
     vector<Vec4i> hierarchy;
-    findContours(MgrayImage, contours, hierarchy, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE);
+    findContours(MgrayImage, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_NONE);
 
     Point p;
     vector<CvPoint2D32f> srcImagePoints_temp(4,cvPoint2D32f(0,0));
@@ -184,7 +189,7 @@ static void foundCorners(vector<CvPoint2D32f> *srcImagePoints,IplImage* source,
         }
         srcImagePoints->at(3) = srcImagePoints_temp.at(index);
 
-        Mat Msource = source;
+        Mat Msource = cv::cvarrToMat(source);
         stringstream ss;
         for(size_t i = 0 ; i<srcImagePoints_temp.size(); i++ )
         {
@@ -224,19 +229,20 @@ static void createOpenGLMatrixFrom(float *posePOSIT,const CvMatr32f &rotationMat
 int main(void)
 {
     help();
-    CvCapture* video = cvCaptureFromFile("cube4.avi");
-    CV_Assert(video);
+    VideoCapture video("cube4.avi");
+    CV_Assert(video.isOpened());
 
-    IplImage* source = cvCreateImage(cvGetSize(cvQueryFrame(video)),8,3);
-    IplImage* grayImage = cvCreateImage(cvGetSize(cvQueryFrame(video)),8,1);
+    Mat frame; video >> frame;
 
-    cvNamedWindow("original",CV_WINDOW_AUTOSIZE | CV_WINDOW_FREERATIO);
-    cvNamedWindow("POSIT",CV_WINDOW_AUTOSIZE | CV_WINDOW_FREERATIO);
+    IplImage* grayImage = cvCreateImage(frame.size(),8,1);
+
+    namedWindow("original", WINDOW_AUTOSIZE | WINDOW_FREERATIO);
+    namedWindow("POSIT", WINDOW_AUTOSIZE | WINDOW_FREERATIO);
     displayOverlay("POSIT", "We lost the 4 corners' detection quite often (the red circles disappear). This demo is only to illustrate how to use OpenGL callback.\n -- Press ESC to exit.", 10000);
     //For debug
     //cvNamedWindow("tempGray",CV_WINDOW_AUTOSIZE);
     float OpenGLMatrix[]={0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
-    cvSetOpenGlDrawCallback("POSIT",on_opengl,OpenGLMatrix);
+    setOpenGlDrawCallback("POSIT",on_opengl,OpenGLMatrix);
 
     vector<CvPoint3D32f> modelPoints;
     initPOSIT(&modelPoints);
@@ -251,26 +257,27 @@ int main(void)
     vector<CvPoint2D32f> srcImagePoints(4,cvPoint2D32f(0,0));
 
 
-    while(cvWaitKey(33) != 27)
+    while(waitKey(33) != 27)
     {
-        source=cvQueryFrame(video);
-        cvShowImage("original",source);
+        video >> frame;
+        imshow("original", frame);
 
-        foundCorners(&srcImagePoints,source,grayImage);
+        IplImage source = frame;
+        foundCorners(&srcImagePoints, &source, grayImage);
         cvPOSIT( positObject, &srcImagePoints[0], FOCAL_LENGTH, criteria, rotation_matrix, translation_vector );
         createOpenGLMatrixFrom(OpenGLMatrix,rotation_matrix,translation_vector);
 
-        cvShowImage("POSIT",source);
+        imshow("POSIT", frame);
         //For debug
         //cvShowImage("tempGray",grayImage);
 
-        if (cvGetCaptureProperty(video,CV_CAP_PROP_POS_AVI_RATIO)>0.99)
-            cvSetCaptureProperty(video,CV_CAP_PROP_POS_AVI_RATIO,0);
+        if (video.get(CAP_PROP_POS_AVI_RATIO) > 0.99)
+            video.set(CAP_PROP_POS_AVI_RATIO, 0);
     }
 
-    cvDestroyAllWindows();
+    destroyAllWindows();
     cvReleaseImage(&grayImage);
-    cvReleaseCapture(&video);
+    video.release();
     cvReleasePOSITObject(&positObject);
 
     return 0;
diff --git a/samples/cpp/bagofwords_classification.cpp b/samples/cpp/bagofwords_classification.cpp
index e24a770f8e..4506e5b9de 100644
--- a/samples/cpp/bagofwords_classification.cpp
+++ b/samples/cpp/bagofwords_classification.cpp
@@ -2563,19 +2563,19 @@ int main(int argc, char** argv)
     Ptr<FeatureDetector> featureDetector = FeatureDetector::create( ddmParams.detectorType );
     Ptr<DescriptorExtractor> descExtractor = DescriptorExtractor::create( ddmParams.descriptorType );
     Ptr<BOWImgDescriptorExtractor> bowExtractor;
-    if( featureDetector.empty() || descExtractor.empty() )
+    if( !featureDetector || !descExtractor )
     {
         cout << "featureDetector or descExtractor was not created" << endl;
         return -1;
     }
     {
         Ptr<DescriptorMatcher> descMatcher = DescriptorMatcher::create( ddmParams.matcherType );
-        if( featureDetector.empty() || descExtractor.empty() || descMatcher.empty() )
+        if( !featureDetector || !descExtractor || !descMatcher )
         {
             cout << "descMatcher was not created" << endl;
             return -1;
         }
-        bowExtractor = new BOWImgDescriptorExtractor( descExtractor, descMatcher );
+        bowExtractor = makePtr<BOWImgDescriptorExtractor>( descExtractor, descMatcher );
     }
 
     // Print configuration to screen
diff --git a/samples/cpp/bgfg_gmg.cpp b/samples/cpp/bgfg_gmg.cpp
index 3d2da44ab2..226eea4635 100644
--- a/samples/cpp/bgfg_gmg.cpp
+++ b/samples/cpp/bgfg_gmg.cpp
@@ -5,7 +5,9 @@
  *      Author: Andrew B. Godbehere
  */
 
-#include <opencv2/opencv.hpp>
+#include "opencv2/video.hpp"
+#include "opencv2/highgui.hpp"
+#include <opencv2/core/utility.hpp>
 #include <iostream>
 
 using namespace cv;
@@ -32,16 +34,13 @@ int main(int argc, char** argv)
     setUseOptimized(true);
     setNumThreads(8);
 
-    Ptr<BackgroundSubtractorGMG> fgbg = Algorithm::create<BackgroundSubtractorGMG>("BackgroundSubtractor.GMG");
-    if (fgbg.empty())
+    Ptr<BackgroundSubtractor> fgbg = createBackgroundSubtractorGMG(20, 0.7);
+    if (!fgbg)
     {
         std::cerr << "Failed to create BackgroundSubtractor.GMG Algorithm." << std::endl;
         return -1;
     }
 
-    fgbg->set("initializationFrames", 20);
-    fgbg->set("decisionThreshold", 0.7);
-
     VideoCapture cap;
     if (argc > 1)
         cap.open(argv[1]);
@@ -65,9 +64,9 @@ int main(int argc, char** argv)
         if (frame.empty())
             break;
 
-        (*fgbg)(frame, fgmask);
+        fgbg->apply(frame, fgmask);
 
-        frame.copyTo(segm);
+        frame.convertTo(segm, CV_8U, 0.5);
         add(frame, Scalar(100, 100, 0), segm, fgmask);
 
         imshow("FG Segmentation", segm);
@@ -79,4 +78,3 @@ int main(int argc, char** argv)
 
     return 0;
 }
-
diff --git a/samples/cpp/bgfg_segm.cpp b/samples/cpp/bgfg_segm.cpp
index 399c7ad28a..a3d02009a7 100644
--- a/samples/cpp/bgfg_segm.cpp
+++ b/samples/cpp/bgfg_segm.cpp
@@ -1,7 +1,8 @@
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
 #include "opencv2/video/background_segm.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/highgui.hpp"
 #include <stdio.h>
 
 using namespace std;
@@ -46,12 +47,12 @@ int main(int argc, const char** argv)
         return -1;
     }
 
-    namedWindow("image", CV_WINDOW_NORMAL);
-    namedWindow("foreground mask", CV_WINDOW_NORMAL);
-    namedWindow("foreground image", CV_WINDOW_NORMAL);
-    namedWindow("mean background image", CV_WINDOW_NORMAL);
+    namedWindow("image", WINDOW_NORMAL);
+    namedWindow("foreground mask", WINDOW_NORMAL);
+    namedWindow("foreground image", WINDOW_NORMAL);
+    namedWindow("mean background image", WINDOW_NORMAL);
 
-    BackgroundSubtractorMOG2 bg_model;//(100, 3, 0.3, 5);
+    Ptr<BackgroundSubtractor> bg_model = createBackgroundSubtractorMOG2();
 
     Mat img, fgmask, fgimg;
 
@@ -68,13 +69,13 @@ int main(int argc, const char** argv)
           fgimg.create(img.size(), img.type());
 
         //update the model
-        bg_model(img, fgmask, update_bg_model ? -1 : 0);
+        bg_model->apply(img, fgmask, update_bg_model ? -1 : 0);
 
         fgimg = Scalar::all(0);
         img.copyTo(fgimg, fgmask);
 
         Mat bgimg;
-        bg_model.getBackgroundImage(bgimg);
+        bg_model->getBackgroundImage(bgimg);
 
         imshow("image", img);
         imshow("foreground mask", fgmask);
diff --git a/samples/cpp/brief_match_test.cpp b/samples/cpp/brief_match_test.cpp
index dea48d38ee..1f121327f3 100644
--- a/samples/cpp/brief_match_test.cpp
+++ b/samples/cpp/brief_match_test.cpp
@@ -4,11 +4,12 @@
  *  Created on: Oct 17, 2010
  *      Author: ethan
  */
-#include "opencv2/core/core.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/calib3d.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 #include <vector>
 #include <iostream>
 
@@ -65,8 +66,8 @@ int main(int argc, const char ** argv)
   string im1_name = parser.get<string>(0);
   string im2_name = parser.get<string>(1);
 
-  Mat im1 = imread(im1_name, CV_LOAD_IMAGE_GRAYSCALE);
-  Mat im2 = imread(im2_name, CV_LOAD_IMAGE_GRAYSCALE);
+  Mat im1 = imread(im1_name, IMREAD_GRAYSCALE);
+  Mat im2 = imread(im2_name, IMREAD_GRAYSCALE);
 
   if (im1.empty() || im2.empty())
   {
diff --git a/samples/cpp/build3dmodel.cpp b/samples/cpp/build3dmodel.cpp
index b896a39369..5c4d79b494 100644
--- a/samples/cpp/build3dmodel.cpp
+++ b/samples/cpp/build3dmodel.cpp
@@ -1,8 +1,9 @@
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <map>
 
@@ -359,7 +360,7 @@ static void triangulatePoint_test(void)
         pts[0] = imgpt1[i]; pts[1] = imgpt2[i];
         objptt[i] = triangulatePoint(pts, Rv, tv, cameraMatrix);
     }
-    double err = norm(Mat(objpt), Mat(objptt), CV_C);
+    double err = norm(Mat(objpt), Mat(objptt), NORM_INF);
     CV_Assert(err < 1e-1);
 }
 
@@ -381,6 +382,96 @@ struct EqKeypoints
     const Set2i* pairs;
 };
 
+template<typename _Tp, class _EqPredicate> static
+int partition( const std::vector<_Tp>& _vec, std::vector<int>& labels,
+           _EqPredicate predicate=_EqPredicate())
+{
+    int i, j, N = (int)_vec.size();
+    const _Tp* vec = &_vec[0];
+
+    const int PARENT=0;
+    const int RANK=1;
+
+    std::vector<int> _nodes(N*2);
+    int (*nodes)[2] = (int(*)[2])&_nodes[0];
+
+    // The first O(N) pass: create N single-vertex trees
+    for(i = 0; i < N; i++)
+    {
+        nodes[i][PARENT]=-1;
+        nodes[i][RANK] = 0;
+    }
+
+    // The main O(N^2) pass: merge connected components
+    for( i = 0; i < N; i++ )
+    {
+        int root = i;
+
+        // find root
+        while( nodes[root][PARENT] >= 0 )
+            root = nodes[root][PARENT];
+
+        for( j = 0; j < N; j++ )
+        {
+            if( i == j || !predicate(vec[i], vec[j]))
+                continue;
+            int root2 = j;
+
+            while( nodes[root2][PARENT] >= 0 )
+                root2 = nodes[root2][PARENT];
+
+            if( root2 != root )
+            {
+                // unite both trees
+                int rank = nodes[root][RANK], rank2 = nodes[root2][RANK];
+                if( rank > rank2 )
+                    nodes[root2][PARENT] = root;
+                else
+                {
+                    nodes[root][PARENT] = root2;
+                    nodes[root2][RANK] += rank == rank2;
+                    root = root2;
+                }
+                CV_Assert( nodes[root][PARENT] < 0 );
+
+                int k = j, parent;
+
+                // compress the path from node2 to root
+                while( (parent = nodes[k][PARENT]) >= 0 )
+                {
+                    nodes[k][PARENT] = root;
+                    k = parent;
+                }
+
+                // compress the path from node to root
+                k = i;
+                while( (parent = nodes[k][PARENT]) >= 0 )
+                {
+                    nodes[k][PARENT] = root;
+                    k = parent;
+                }
+            }
+        }
+    }
+
+    // Final O(N) pass: enumerate classes
+    labels.resize(N);
+    int nclasses = 0;
+
+    for( i = 0; i < N; i++ )
+    {
+        int root = i;
+        while( nodes[root][PARENT] >= 0 )
+            root = nodes[root][PARENT];
+        // re-use the rank as the class label
+        if( nodes[root][RANK] >= 0 )
+            nodes[root][RANK] = ~nclasses++;
+        labels[i] = ~nodes[root][RANK];
+    }
+
+    return nclasses;
+}
+
 static void build3dmodel( const Ptr<FeatureDetector>& detector,
                           const Ptr<DescriptorExtractor>& descriptorExtractor,
                           const vector<Point3f>& /*modelBox*/,
@@ -418,7 +509,7 @@ static void build3dmodel( const Ptr<FeatureDetector>& detector,
     for( size_t i = 0; i < nimages; i++ )
     {
         Mat img = imread(imageList[i], 1), gray;
-        cvtColor(img, gray, CV_BGR2GRAY);
+        cvtColor(img, gray, COLOR_BGR2GRAY);
 
         vector<KeyPoint> keypoints;
         detector->detect(gray, keypoints);
@@ -603,7 +694,7 @@ static void build3dmodel( const Ptr<FeatureDetector>& detector,
         projectPoints(Mat(model.points), Rs[i], ts[i], cameraMatrix, Mat(), imagePoints);
 
         for( int k = 0; k < (int)imagePoints.size(); k++ )
-            circle(img, imagePoints[k], 2, Scalar(0,255,0), -1, CV_AA, 0);
+            circle(img, imagePoints[k], 2, Scalar(0,255,0), -1, LINE_AA, 0);
 
         imshow("Test", img);
         int c = waitKey();
diff --git a/samples/cpp/calibration.cpp b/samples/cpp/calibration.cpp
index f4ff81077b..bb7c396b43 100644
--- a/samples/cpp/calibration.cpp
+++ b/samples/cpp/calibration.cpp
@@ -1,7 +1,8 @@
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <cctype>
 #include <stdio.h>
@@ -95,7 +96,7 @@ static double computeReprojectionErrors(
     {
         projectPoints(Mat(objectPoints[i]), rvecs[i], tvecs[i],
                       cameraMatrix, distCoeffs, imagePoints2);
-        err = norm(Mat(imagePoints[i]), Mat(imagePoints2), CV_L2);
+        err = norm(Mat(imagePoints[i]), Mat(imagePoints2), NORM_L2);
         int n = (int)objectPoints[i].size();
         perViewErrors[i] = (float)std::sqrt(err*err/n);
         totalErr += err*err;
@@ -127,7 +128,7 @@ static void calcChessboardCorners(Size boardSize, float squareSize, vector<Point
         break;
 
       default:
-        CV_Error(CV_StsBadArg, "Unknown pattern type\n");
+        CV_Error(Error::StsBadArg, "Unknown pattern type\n");
     }
 }
 
@@ -140,7 +141,7 @@ static bool runCalibration( vector<vector<Point2f> > imagePoints,
                     double& totalAvgErr)
 {
     cameraMatrix = Mat::eye(3, 3, CV_64F);
-    if( flags & CV_CALIB_FIX_ASPECT_RATIO )
+    if( flags & CALIB_FIX_ASPECT_RATIO )
         cameraMatrix.at<double>(0,0) = aspectRatio;
 
     distCoeffs = Mat::zeros(8, 1, CV_64F);
@@ -151,8 +152,8 @@ static bool runCalibration( vector<vector<Point2f> > imagePoints,
     objectPoints.resize(imagePoints.size(),objectPoints[0]);
 
     double rms = calibrateCamera(objectPoints, imagePoints, imageSize, cameraMatrix,
-                    distCoeffs, rvecs, tvecs, flags|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5);
-                    ///*|CV_CALIB_FIX_K3*/|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5);
+                    distCoeffs, rvecs, tvecs, flags|CALIB_FIX_K4|CALIB_FIX_K5);
+                    ///*|CALIB_FIX_K3*/|CALIB_FIX_K4|CALIB_FIX_K5);
     printf("RMS error reported by calibrateCamera: %g\n", rms);
 
     bool ok = checkRange(cameraMatrix) && checkRange(distCoeffs);
@@ -191,17 +192,17 @@ static void saveCameraParams( const string& filename,
     fs << "board_height" << boardSize.height;
     fs << "square_size" << squareSize;
 
-    if( flags & CV_CALIB_FIX_ASPECT_RATIO )
+    if( flags & CALIB_FIX_ASPECT_RATIO )
         fs << "aspectRatio" << aspectRatio;
 
     if( flags != 0 )
     {
         sprintf( buf, "flags: %s%s%s%s",
-            flags & CV_CALIB_USE_INTRINSIC_GUESS ? "+use_intrinsic_guess" : "",
-            flags & CV_CALIB_FIX_ASPECT_RATIO ? "+fix_aspectRatio" : "",
-            flags & CV_CALIB_FIX_PRINCIPAL_POINT ? "+fix_principal_point" : "",
-            flags & CV_CALIB_ZERO_TANGENT_DIST ? "+zero_tangent_dist" : "" );
-        cvWriteComment( *fs, buf, 0 );
+            flags & CALIB_USE_INTRINSIC_GUESS ? "+use_intrinsic_guess" : "",
+            flags & CALIB_FIX_ASPECT_RATIO ? "+fix_aspectRatio" : "",
+            flags & CALIB_FIX_PRINCIPAL_POINT ? "+fix_principal_point" : "",
+            flags & CALIB_ZERO_TANGENT_DIST ? "+zero_tangent_dist" : "" );
+        //cvWriteComment( *fs, buf, 0 );
     }
 
     fs << "flags" << flags;
@@ -228,7 +229,7 @@ static void saveCameraParams( const string& filename,
             r = rvecs[i].t();
             t = tvecs[i].t();
         }
-        cvWriteComment( *fs, "a set of 6-tuples (rotation vector + translation vector) for each view", 0 );
+        //cvWriteComment( *fs, "a set of 6-tuples (rotation vector + translation vector) for each view", 0 );
         fs << "extrinsic_parameters" << bigmat;
     }
 
@@ -360,7 +361,7 @@ int main( int argc, char** argv )
         {
             if( sscanf( argv[++i], "%f", &aspectRatio ) != 1 || aspectRatio <= 0 )
                 return printf("Invalid aspect ratio\n" ), -1;
-            flags |= CV_CALIB_FIX_ASPECT_RATIO;
+            flags |= CALIB_FIX_ASPECT_RATIO;
         }
         else if( strcmp( s, "-d" ) == 0 )
         {
@@ -377,11 +378,11 @@ int main( int argc, char** argv )
         }
         else if( strcmp( s, "-zt" ) == 0 )
         {
-            flags |= CV_CALIB_ZERO_TANGENT_DIST;
+            flags |= CALIB_ZERO_TANGENT_DIST;
         }
         else if( strcmp( s, "-p" ) == 0 )
         {
-            flags |= CV_CALIB_FIX_PRINCIPAL_POINT;
+            flags |= CALIB_FIX_PRINCIPAL_POINT;
         }
         else if( strcmp( s, "-v" ) == 0 )
         {
@@ -461,14 +462,14 @@ int main( int argc, char** argv )
             flip( view, view, 0 );
 
         vector<Point2f> pointbuf;
-        cvtColor(view, viewGray, CV_BGR2GRAY);
+        cvtColor(view, viewGray, COLOR_BGR2GRAY);
 
         bool found;
         switch( pattern )
         {
             case CHESSBOARD:
                 found = findChessboardCorners( view, boardSize, pointbuf,
-                    CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_FAST_CHECK | CV_CALIB_CB_NORMALIZE_IMAGE);
+                    CALIB_CB_ADAPTIVE_THRESH | CALIB_CB_FAST_CHECK | CALIB_CB_NORMALIZE_IMAGE);
                 break;
             case CIRCLES_GRID:
                 found = findCirclesGrid( view, boardSize, pointbuf );
@@ -482,7 +483,7 @@ int main( int argc, char** argv )
 
        // improve the found corners' coordinate accuracy
         if( pattern == CHESSBOARD && found) cornerSubPix( viewGray, pointbuf, Size(11,11),
-            Size(-1,-1), TermCriteria( CV_TERMCRIT_EPS+CV_TERMCRIT_ITER, 30, 0.1 ));
+            Size(-1,-1), TermCriteria( TermCriteria::EPS+TermCriteria::COUNT, 30, 0.1 ));
 
         if( mode == CAPTURING && found &&
            (!capture.isOpened() || clock() - prevTimestamp > delay*1e-3*CLOCKS_PER_SEC) )
diff --git a/samples/cpp/calibration_artificial.cpp b/samples/cpp/calibration_artificial.cpp
index b4496615ce..3023127b73 100644
--- a/samples/cpp/calibration_artificial.cpp
+++ b/samples/cpp/calibration_artificial.cpp
@@ -120,7 +120,7 @@ int main()
         imshow("Current chessboard", boards[i]); waitKey(1000);
     }
     cout << "Done" << endl;
-    cvDestroyAllWindows();
+    destroyAllWindows();
 
     Mat camMat_est;
     Mat distCoeffs_est;
@@ -259,15 +259,15 @@ Mat cv::ChessBoardGenerator::generageChessBoard(const Mat& bg, const Mat& camMat
     if (rendererResolutionMultiplier == 1)
     {
         result = bg.clone();
-        drawContours(result, whole_contour, -1, Scalar::all(255), CV_FILLED, CV_AA);
-        drawContours(result, squares_black, -1, Scalar::all(0), CV_FILLED, CV_AA);
+        drawContours(result, whole_contour, -1, Scalar::all(255), FILLED, LINE_AA);
+        drawContours(result, squares_black, -1, Scalar::all(0), FILLED, LINE_AA);
     }
     else
     {
         Mat tmp;
         resize(bg, tmp, bg.size() * rendererResolutionMultiplier);
-        drawContours(tmp, whole_contour, -1, Scalar::all(255), CV_FILLED, CV_AA);
-        drawContours(tmp, squares_black, -1, Scalar::all(0), CV_FILLED, CV_AA);
+        drawContours(tmp, whole_contour, -1, Scalar::all(255), FILLED, LINE_AA);
+        drawContours(tmp, squares_black, -1, Scalar::all(0), FILLED, LINE_AA);
         resize(tmp, result, bg.size(), 0, 0, INTER_AREA);
     }
     return result;
@@ -332,4 +332,3 @@ Mat cv::ChessBoardGenerator::operator ()(const Mat& bg, const Mat& camMat, const
 
     return generageChessBoard(bg, camMat, distCoeffs, zero, pb1, pb2, sqWidth, sqHeight,  pts3d, corners);
 }
-
diff --git a/samples/cpp/camshiftdemo.cpp b/samples/cpp/camshiftdemo.cpp
index 36316e6360..6439cef78e 100644
--- a/samples/cpp/camshiftdemo.cpp
+++ b/samples/cpp/camshiftdemo.cpp
@@ -1,6 +1,7 @@
+#include <opencv2/core/utility.hpp>
 #include "opencv2/video/tracking.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <iostream>
 #include <ctype.h>
@@ -32,12 +33,12 @@ static void onMouse( int event, int x, int y, int, void* )
 
     switch( event )
     {
-    case CV_EVENT_LBUTTONDOWN:
+    case EVENT_LBUTTONDOWN:
         origin = Point(x,y);
         selection = Rect(x,y,0,0);
         selectObject = true;
         break;
-    case CV_EVENT_LBUTTONUP:
+    case EVENT_LBUTTONUP:
         selectObject = false;
         if( selection.width > 0 && selection.height > 0 )
             trackObject = -1;
@@ -113,7 +114,7 @@ int main( int argc, const char** argv )
 
         if( !paused )
         {
-            cvtColor(image, hsv, CV_BGR2HSV);
+            cvtColor(image, hsv, COLOR_BGR2HSV);
 
             if( trackObject )
             {
@@ -129,7 +130,7 @@ int main( int argc, const char** argv )
                 {
                     Mat roi(hue, selection), maskroi(mask, selection);
                     calcHist(&roi, 1, 0, maskroi, hist, 1, &hsize, &phranges);
-                    normalize(hist, hist, 0, 255, CV_MINMAX);
+                    normalize(hist, hist, 0, 255, NORM_MINMAX);
 
                     trackWindow = selection;
                     trackObject = 1;
@@ -139,7 +140,7 @@ int main( int argc, const char** argv )
                     Mat buf(1, hsize, CV_8UC3);
                     for( int i = 0; i < hsize; i++ )
                         buf.at<Vec3b>(i) = Vec3b(saturate_cast<uchar>(i*180./hsize), 255, 255);
-                    cvtColor(buf, buf, CV_HSV2BGR);
+                    cvtColor(buf, buf, COLOR_HSV2BGR);
 
                     for( int i = 0; i < hsize; i++ )
                     {
@@ -153,7 +154,7 @@ int main( int argc, const char** argv )
                 calcBackProject(&hue, 1, 0, hist, backproj, &phranges);
                 backproj &= mask;
                 RotatedRect trackBox = CamShift(backproj, trackWindow,
-                                    TermCriteria( CV_TERMCRIT_EPS | CV_TERMCRIT_ITER, 10, 1 ));
+                                    TermCriteria( TermCriteria::EPS | TermCriteria::COUNT, 10, 1 ));
                 if( trackWindow.area() <= 1 )
                 {
                     int cols = backproj.cols, rows = backproj.rows, r = (MIN(cols, rows) + 5)/6;
@@ -163,8 +164,8 @@ int main( int argc, const char** argv )
                 }
 
                 if( backprojMode )
-                    cvtColor( backproj, image, CV_GRAY2BGR );
-                ellipse( image, trackBox, Scalar(0,0,255), 3, CV_AA );
+                    cvtColor( backproj, image, COLOR_GRAY2BGR );
+                ellipse( image, trackBox, Scalar(0,0,255), 3, LINE_AA );
             }
         }
         else if( trackObject < 0 )
diff --git a/samples/cpp/chamfer.cpp b/samples/cpp/chamfer.cpp
index 8a51b0bae2..ab211f33a5 100644
--- a/samples/cpp/chamfer.cpp
+++ b/samples/cpp/chamfer.cpp
@@ -1,6 +1,7 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/contrib/contrib.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/contrib.hpp"
 
 #include <iostream>
 
@@ -40,7 +41,7 @@ int main( int argc, const char** argv )
         return -1;
     }
     Mat cimg;
-    cvtColor(img, cimg, CV_GRAY2BGR);
+    cvtColor(img, cimg, COLOR_GRAY2BGR);
 
     // if the image and the template are not edge maps but normal grayscale images,
     // you might want to uncomment the lines below to produce the maps. You can also
diff --git a/samples/cpp/connected_components.cpp b/samples/cpp/connected_components.cpp
index d20991ecd1..50c677d372 100644
--- a/samples/cpp/connected_components.cpp
+++ b/samples/cpp/connected_components.cpp
@@ -1,5 +1,6 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 #include <iostream>
 
 using namespace cv;
diff --git a/samples/cpp/contours2.cpp b/samples/cpp/contours2.cpp
index 7d728ec779..0b488959ef 100644
--- a/samples/cpp/contours2.cpp
+++ b/samples/cpp/contours2.cpp
@@ -28,7 +28,7 @@ static void on_trackbar(int, void*)
     Mat cnt_img = Mat::zeros(w, w, CV_8UC3);
     int _levels = levels - 3;
     drawContours( cnt_img, contours, _levels <= 0 ? 3 : -1, Scalar(128,255,255),
-                  3, CV_AA, hierarchy, std::abs(_levels) );
+                  3, LINE_AA, hierarchy, std::abs(_levels) );
 
     imshow("contours", cnt_img);
 }
diff --git a/samples/cpp/convexhull.cpp b/samples/cpp/convexhull.cpp
index ce1503c9a6..2aef70fe7d 100644
--- a/samples/cpp/convexhull.cpp
+++ b/samples/cpp/convexhull.cpp
@@ -41,7 +41,7 @@ int main( int /*argc*/, char** /*argv*/ )
 
         img = Scalar::all(0);
         for( i = 0; i < count; i++ )
-            circle(img, points[i], 3, Scalar(0, 0, 255), CV_FILLED, CV_AA);
+            circle(img, points[i], 3, Scalar(0, 0, 255), FILLED, LINE_AA);
 
         int hullcount = (int)hull.size();
         Point pt0 = points[hull[hullcount-1]];
@@ -49,7 +49,7 @@ int main( int /*argc*/, char** /*argv*/ )
         for( i = 0; i < hullcount; i++ )
         {
             Point pt = points[hull[i]];
-            line(img, pt0, pt, Scalar(0, 255, 0), 1, CV_AA);
+            line(img, pt0, pt, Scalar(0, 255, 0), 1,LINE_AA);
             pt0 = pt;
         }
 
diff --git a/samples/cpp/cout_mat.cpp b/samples/cpp/cout_mat.cpp
index 936a6cebd4..0ef94db125 100644
--- a/samples/cpp/cout_mat.cpp
+++ b/samples/cpp/cout_mat.cpp
@@ -30,16 +30,16 @@ int main(int,char**)
     help();
     Mat I = Mat::eye(4, 4, CV_64F);
     I.at<double>(1,1) = CV_PI;
-    cout << "I = " << I << ";" << endl;
+    cout << "I = \n" << I << ";" << endl << endl;
 
     Mat r = Mat(10, 3, CV_8UC3);
     randu(r, Scalar::all(0), Scalar::all(255));
 
-    cout << "r (default) = " << r << ";" << endl << endl;
-    cout << "r (python) = " << format(r,"python") << ";" << endl << endl;
-    cout << "r (numpy) = " << format(r,"numpy") << ";" << endl << endl;
-    cout << "r (csv) = " << format(r,"csv") << ";" << endl << endl;
-    cout << "r (c) = " << format(r,"C") << ";" << endl << endl;
+    cout << "r (default) = \n" << r << ";" << endl << endl;
+    cout << "r (python) = \n" << format(r, Formatter::FMT_PYTHON) << ";" << endl << endl;
+    cout << "r (numpy) = \n" << format(r, Formatter::FMT_NUMPY) << ";" << endl << endl;
+    cout << "r (csv) = \n" << format(r, Formatter::FMT_CSV) << ";" << endl << endl;
+    cout << "r (c) = \n" << format(r, Formatter::FMT_C) << ";" << endl << endl;
 
     Point2f p(5, 1);
     cout << "p = " << p << ";" << endl;
diff --git a/samples/cpp/dbt_face_detection.cpp b/samples/cpp/dbt_face_detection.cpp
index 35386a78d0..651eaff73f 100644
--- a/samples/cpp/dbt_face_detection.cpp
+++ b/samples/cpp/dbt_face_detection.cpp
@@ -22,7 +22,7 @@ class CascadeDetectorAdapter: public DetectionBasedTracker::IDetector
             IDetector(),
             Detector(detector)
         {
-            CV_Assert(!detector.empty());
+            CV_Assert(detector);
         }
 
         void detect(const cv::Mat &Image, std::vector<cv::Rect> &objects)
@@ -51,11 +51,11 @@ int main(int , char** )
     }
 
     std::string cascadeFrontalfilename = "../../data/lbpcascades/lbpcascade_frontalface.xml";
-    cv::Ptr<cv::CascadeClassifier> cascade = new cv::CascadeClassifier(cascadeFrontalfilename);
-    cv::Ptr<DetectionBasedTracker::IDetector> MainDetector = new CascadeDetectorAdapter(cascade);
+    cv::Ptr<cv::CascadeClassifier> cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> MainDetector = makePtr<CascadeDetectorAdapter>(cascade);
 
-    cascade = new cv::CascadeClassifier(cascadeFrontalfilename);
-    cv::Ptr<DetectionBasedTracker::IDetector> TrackingDetector = new CascadeDetectorAdapter(cascade);
+    cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> TrackingDetector = makePtr<CascadeDetectorAdapter>(cascade);
 
     DetectionBasedTracker::Parameters params;
     DetectionBasedTracker Detector(MainDetector, TrackingDetector, params);
@@ -79,12 +79,12 @@ int main(int , char** )
 
         for (size_t i = 0; i < Faces.size(); i++)
         {
-            rectangle(ReferenceFrame, Faces[i], CV_RGB(0,255,0));
+            rectangle(ReferenceFrame, Faces[i], Scalar(0,255,0));
         }
 
         imshow(WindowName, ReferenceFrame);
 
-        if (cvWaitKey(30) >= 0) break;
+        if (waitKey(30) >= 0) break;
     }
 
     Detector.stop();
diff --git a/samples/cpp/delaunay2.cpp b/samples/cpp/delaunay2.cpp
index 87e6664b67..a00e31a988 100644
--- a/samples/cpp/delaunay2.cpp
+++ b/samples/cpp/delaunay2.cpp
@@ -18,7 +18,7 @@ static void help()
 
 static void draw_subdiv_point( Mat& img, Point2f fp, Scalar color )
 {
-    circle( img, fp, 3, color, CV_FILLED, 8, 0 );
+    circle( img, fp, 3, color, FILLED, LINE_8, 0 );
 }
 
 static void draw_subdiv( Mat& img, Subdiv2D& subdiv, Scalar delaunay_color )
@@ -34,9 +34,9 @@ static void draw_subdiv( Mat& img, Subdiv2D& subdiv, Scalar delaunay_color )
         pt[0] = Point(cvRound(t[0]), cvRound(t[1]));
         pt[1] = Point(cvRound(t[2]), cvRound(t[3]));
         pt[2] = Point(cvRound(t[4]), cvRound(t[5]));
-        line(img, pt[0], pt[1], delaunay_color, 1, CV_AA, 0);
-        line(img, pt[1], pt[2], delaunay_color, 1, CV_AA, 0);
-        line(img, pt[2], pt[0], delaunay_color, 1, CV_AA, 0);
+        line(img, pt[0], pt[1], delaunay_color, 1, LINE_AA, 0);
+        line(img, pt[1], pt[2], delaunay_color, 1, LINE_AA, 0);
+        line(img, pt[2], pt[0], delaunay_color, 1, LINE_AA, 0);
     }
 #else
     vector<Vec4f> edgeList;
@@ -46,7 +46,7 @@ static void draw_subdiv( Mat& img, Subdiv2D& subdiv, Scalar delaunay_color )
         Vec4f e = edgeList[i];
         Point pt0 = Point(cvRound(e[0]), cvRound(e[1]));
         Point pt1 = Point(cvRound(e[2]), cvRound(e[3]));
-        line(img, pt0, pt1, delaunay_color, 1, CV_AA, 0);
+        line(img, pt0, pt1, delaunay_color, 1, LINE_AA, 0);
     }
 #endif
 }
@@ -64,7 +64,7 @@ static void locate_point( Mat& img, Subdiv2D& subdiv, Point2f fp, Scalar active_
         {
             Point2f org, dst;
             if( subdiv.edgeOrg(e, &org) > 0 && subdiv.edgeDst(e, &dst) > 0 )
-                line( img, org, dst, active_color, 3, CV_AA, 0 );
+                line( img, org, dst, active_color, 3, LINE_AA, 0 );
 
             e = subdiv.getEdge(e, Subdiv2D::NEXT_AROUND_LEFT);
         }
@@ -97,8 +97,8 @@ static void paint_voronoi( Mat& img, Subdiv2D& subdiv )
         fillConvexPoly(img, ifacet, color, 8, 0);
 
         ifacets[0] = ifacet;
-        polylines(img, ifacets, true, Scalar(), 1, CV_AA, 0);
-        circle(img, centers[i], 3, Scalar(), -1, CV_AA, 0);
+        polylines(img, ifacets, true, Scalar(), 1, LINE_AA, 0);
+        circle(img, centers[i], 3, Scalar(), FILLED, LINE_AA, 0);
     }
 }
 
diff --git a/samples/cpp/demhist.cpp b/samples/cpp/demhist.cpp
index 8b9189f32d..ba60115df0 100644
--- a/samples/cpp/demhist.cpp
+++ b/samples/cpp/demhist.cpp
@@ -1,5 +1,6 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <iostream>
 
@@ -43,7 +44,7 @@ static void updateBrightnessContrast( int /*arg*/, void* )
     calcHist(&dst, 1, 0, Mat(), hist, 1, &histSize, 0);
     Mat histImage = Mat::ones(200, 320, CV_8U)*255;
 
-    normalize(hist, hist, 0, histImage.rows, CV_MINMAX, CV_32F);
+    normalize(hist, hist, 0, histImage.rows, NORM_MINMAX, CV_32F);
 
     histImage = Scalar::all(255);
     int binW = cvRound((double)histImage.cols/histSize);
diff --git a/samples/cpp/descriptor_extractor_matcher.cpp b/samples/cpp/descriptor_extractor_matcher.cpp
index 7aa5299103..7df3cd02b0 100644
--- a/samples/cpp/descriptor_extractor_matcher.cpp
+++ b/samples/cpp/descriptor_extractor_matcher.cpp
@@ -45,7 +45,7 @@ static int getMatcherFilterType( const string& str )
         return NONE_FILTER;
     if( str == "CrossCheckFilter" )
         return CROSS_CHECK_FILTER;
-    CV_Error(CV_StsBadArg, "Invalid filter name");
+    CV_Error(Error::StsBadArg, "Invalid filter name");
     return -1;
 }
 
@@ -109,12 +109,12 @@ static void doIteration( const Mat& img1, Mat& img2, bool isWarpPerspective,
                   Ptr<DescriptorMatcher>& descriptorMatcher, int matcherFilter, bool eval,
                   double ransacReprojThreshold, RNG& rng )
 {
-    assert( !img1.empty() );
+    CV_Assert( !img1.empty() );
     Mat H12;
     if( isWarpPerspective )
         warpPerspectiveRand(img1, img2, H12, rng );
     else
-        assert( !img2.empty()/* && img2.cols==img1.cols && img2.rows==img1.rows*/ );
+        CV_Assert( !img2.empty()/* && img2.cols==img1.cols && img2.rows==img1.rows*/ );
 
     cout << endl << "< Extracting keypoints from second image..." << endl;
     vector<KeyPoint> keypoints2;
@@ -153,7 +153,7 @@ static void doIteration( const Mat& img1, Mat& img2, bool isWarpPerspective,
     {
         cout << "< Evaluate descriptor matcher..." << endl;
         vector<Point2f> curve;
-        Ptr<GenericDescriptorMatcher> gdm = new VectorDescriptorMatcher( descriptorExtractor, descriptorMatcher );
+        Ptr<GenericDescriptorMatcher> gdm = makePtr<VectorDescriptorMatcher>( descriptorExtractor, descriptorMatcher );
         evaluateGenericDescriptorMatcher( img1, img2, H12, keypoints1, keypoints2, 0, 0, curve, gdm );
 
         Point2f firstPoint = *curve.begin();
@@ -189,7 +189,7 @@ static void doIteration( const Mat& img1, Mat& img2, bool isWarpPerspective,
         cout << "< Computing homography (RANSAC)..." << endl;
         vector<Point2f> points1; KeyPoint::convert(keypoints1, points1, queryIdxs);
         vector<Point2f> points2; KeyPoint::convert(keypoints2, points2, trainIdxs);
-        H12 = findHomography( Mat(points1), Mat(points2), CV_RANSAC, ransacReprojThreshold );
+        H12 = findHomography( Mat(points1), Mat(points2), RANSAC, ransacReprojThreshold );
         cout << ">" << endl;
     }
 
@@ -208,7 +208,7 @@ static void doIteration( const Mat& img1, Mat& img2, bool isWarpPerspective,
                 matchesMask[i1] = 1;
         }
         // draw inliers
-        drawMatches( img1, keypoints1, img2, keypoints2, filteredMatches, drawImg, CV_RGB(0, 255, 0), CV_RGB(0, 0, 255), matchesMask
+        drawMatches( img1, keypoints1, img2, keypoints2, filteredMatches, drawImg, Scalar(0, 255, 0), Scalar(0, 0, 255), matchesMask
 #if DRAW_RICH_KEYPOINTS_MODE
                      , DrawMatchesFlags::DRAW_RICH_KEYPOINTS
 #endif
@@ -218,7 +218,7 @@ static void doIteration( const Mat& img1, Mat& img2, bool isWarpPerspective,
         // draw outliers
         for( size_t i1 = 0; i1 < matchesMask.size(); i1++ )
             matchesMask[i1] = !matchesMask[i1];
-        drawMatches( img1, keypoints1, img2, keypoints2, filteredMatches, drawImg, CV_RGB(0, 0, 255), CV_RGB(255, 0, 0), matchesMask,
+        drawMatches( img1, keypoints1, img2, keypoints2, filteredMatches, drawImg, Scalar(0, 0, 255), Scalar(255, 0, 0), matchesMask,
                      DrawMatchesFlags::DRAW_OVER_OUTIMG | DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
 #endif
 
@@ -253,7 +253,7 @@ int main(int argc, char** argv)
     int mactherFilterType = getMatcherFilterType( argv[4] );
     bool eval = !isWarpPerspective ? false : (atoi(argv[6]) == 0 ? false : true);
     cout << ">" << endl;
-    if( detector.empty() || descriptorExtractor.empty() || descriptorMatcher.empty()  )
+    if( !detector || !descriptorExtractor || !descriptorMatcher )
     {
         cout << "Can not create detector or descriptor exstractor or descriptor matcher of given types" << endl;
         return -1;
diff --git a/samples/cpp/detection_based_tracker_sample.cpp b/samples/cpp/detection_based_tracker_sample.cpp
index 61106f3a9c..81afa789c3 100644
--- a/samples/cpp/detection_based_tracker_sample.cpp
+++ b/samples/cpp/detection_based_tracker_sample.cpp
@@ -1,10 +1,10 @@
 #if defined(__linux__) || defined(LINUX) || defined(__APPLE__) || defined(ANDROID)
 
-#include <opencv2/core/core.hpp>
-#include <opencv2/core/internal.hpp>
-#include <opencv2/imgproc/imgproc.hpp>
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/objdetect/objdetect.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/objdetect.hpp>
 #include "opencv2/contrib/detection_based_tracker.hpp"
 
 #include <vector>
@@ -67,7 +67,7 @@ class CascadeDetectorAdapter: public DetectionBasedTracker::IDetector
         CascadeDetectorAdapter(cv::Ptr<cv::CascadeClassifier> detector):
             Detector(detector)
         {
-            CV_Assert(!detector.empty());
+            CV_Assert(detector);
         }
 
         void detect(const cv::Mat &Image, std::vector<cv::Rect> &objects)
@@ -117,11 +117,11 @@ static int test_FaceDetector(int argc, char *argv[])
     }
 
     std::string cascadeFrontalfilename=cascadefile;
-    cv::Ptr<cv::CascadeClassifier> cascade = new cv::CascadeClassifier(cascadeFrontalfilename);
-    cv::Ptr<DetectionBasedTracker::IDetector> MainDetector = new CascadeDetectorAdapter(cascade);
+    cv::Ptr<cv::CascadeClassifier> cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> MainDetector = makePtr<CascadeDetectorAdapter>(cascade);
 
-    cascade = new cv::CascadeClassifier(cascadeFrontalfilename);
-    cv::Ptr<DetectionBasedTracker::IDetector> TrackingDetector = new CascadeDetectorAdapter(cascade);
+    cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> TrackingDetector = makePtr<CascadeDetectorAdapter>(cascade);
 
     DetectionBasedTracker::Parameters params;
     DetectionBasedTracker fd(MainDetector, TrackingDetector, params);
@@ -144,7 +144,7 @@ static int test_FaceDetector(int argc, char *argv[])
         LOGD("\n\nSTEP n=%d        from prev step %f ms\n", n, t_ms);
         m=images[n-1];
         CV_Assert(! m.empty());
-        cvtColor(m, gray, CV_BGR2GRAY);
+        cvtColor(m, gray, COLOR_BGR2GRAY);
 
         fd.process(gray);
 
diff --git a/samples/cpp/detector_descriptor_evaluation.cpp b/samples/cpp/detector_descriptor_evaluation.cpp
index 650db65d51..dd3cd800e5 100644
--- a/samples/cpp/detector_descriptor_evaluation.cpp
+++ b/samples/cpp/detector_descriptor_evaluation.cpp
@@ -39,10 +39,10 @@
 //
 //M*/
 
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#include "opencv2/legacy/legacy.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/features2d.hpp"
+#include "opencv2/legacy.hpp"
 
 #include <limits>
 #include <cstdio>
@@ -93,7 +93,7 @@ static void calcKeyPointProjections( const vector<KeyPoint>& src, const Mat_<dou
 {
     if(  !src.empty() )
     {
-        assert( !H.empty() && H.cols == 3 && H.rows == 3);
+        CV_Assert( !H.empty() && H.cols == 3 && H.rows == 3);
         dst.resize(src.size());
         vector<KeyPoint>::const_iterator srcIt = src.begin();
         vector<KeyPoint>::iterator       dstIt = dst.begin();
@@ -109,7 +109,7 @@ static void calcKeyPointProjections( const vector<KeyPoint>& src, const Mat_<dou
             Mat_<double> Aff; linearizeHomographyAt(H, srcIt->pt, Aff);
             Mat_<double> dstM; invert(Aff*invM*Aff.t(), dstM);
             Mat_<double> eval; eigen( dstM, eval );
-            assert( eval(0,0) && eval(1,0) );
+            CV_Assert( eval(0,0) && eval(1,0) );
             float dstSize = (float)pow(1./(eval(0,0)*eval(1,0)), 0.25);
 
             // TODO: check angle projection
@@ -526,7 +526,7 @@ inline void writeKeypoints( FileStorage& fs, const vector<KeyPoint>& keypoints,
 
 inline void readKeypoints( FileStorage& fs, vector<KeyPoint>& keypoints, int imgIdx )
 {
-    assert( fs.isOpened() );
+    CV_Assert( fs.isOpened() );
     stringstream imgName; imgName << "img" << imgIdx;
     read( fs[imgName.str()], keypoints);
 }
@@ -535,7 +535,7 @@ void DetectorQualityEvaluator::readAlgorithm ()
 {
     defaultDetector = FeatureDetector::create( algName );
     specificDetector = FeatureDetector::create( algName );
-    if( defaultDetector.empty() )
+    if( !defaultDetector )
     {
         printf( "Algorithm can not be read\n" );
         exit(-1);
@@ -769,14 +769,14 @@ void DescriptorQualityEvaluator::readAlgorithm( )
     defaultDescMatcher = GenericDescriptorMatcher::create( algName );
     specificDescMatcher = GenericDescriptorMatcher::create( algName );
 
-    if( defaultDescMatcher.empty() )
+    if( !defaultDescMatcher )
     {
         Ptr<DescriptorExtractor> extractor = DescriptorExtractor::create( algName );
         Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create( matcherName );
-        defaultDescMatcher = new VectorDescriptorMatch( extractor, matcher );
-        specificDescMatcher = new VectorDescriptorMatch( extractor, matcher );
+        defaultDescMatcher = makePtr<VectorDescriptorMatch>( extractor, matcher );
+        specificDescMatcher = makePtr<VectorDescriptorMatch>( extractor, matcher );
 
-        if( extractor.empty() || matcher.empty() )
+        if( !extractor || !matcher )
         {
             printf("Algorithm can not be read\n");
             exit(-1);
@@ -881,8 +881,9 @@ public:
     virtual void readAlgorithm( )
     {
         string classifierFile = data_path + "/features2d/calonder_classifier.rtc";
-        defaultDescMatcher = new VectorDescriptorMatch( new CalonderDescriptorExtractor<float>( classifierFile ),
-                                                        new BFMatcher(NORM_L2) );
+        defaultDescMatcher = makePtr<VectorDescriptorMatch>(
+            makePtr<CalonderDescriptorExtractor<float> >( classifierFile ),
+            makePtr<BFMatcher>(int(NORM_L2)));
         specificDescMatcher = defaultDescMatcher;
     }
 };
@@ -922,10 +923,11 @@ void OneWayDescriptorQualityTest::processRunParamsFile ()
 
     readAllDatasetsRunParams();
 
-    OneWayDescriptorBase *base = new OneWayDescriptorBase(patchSize, poseCount, pcaFilename,
-                                               trainPath, trainImagesList);
+    Ptr<OneWayDescriptorBase> base(
+        new OneWayDescriptorBase(patchSize, poseCount, pcaFilename,
+                                 trainPath, trainImagesList));
 
-    OneWayDescriptorMatch *match = new OneWayDescriptorMatch ();
+    Ptr<OneWayDescriptorMatch> match = makePtr<OneWayDescriptorMatch>();
     match->initialize( OneWayDescriptorMatch::Params (), base );
     defaultDescMatcher = match;
     writeAllDatasetsRunParams();
@@ -958,18 +960,18 @@ int main( int argc, char** argv )
 
     Ptr<BaseQualityEvaluator> evals[] =
     {
-        new DetectorQualityEvaluator( "FAST", "quality-detector-fast" ),
-        new DetectorQualityEvaluator( "GFTT", "quality-detector-gftt" ),
-        new DetectorQualityEvaluator( "HARRIS", "quality-detector-harris" ),
-        new DetectorQualityEvaluator( "MSER", "quality-detector-mser" ),
-        new DetectorQualityEvaluator( "STAR", "quality-detector-star" ),
-        new DetectorQualityEvaluator( "SIFT", "quality-detector-sift" ),
-        new DetectorQualityEvaluator( "SURF", "quality-detector-surf" ),
+        makePtr<DetectorQualityEvaluator>( "FAST", "quality-detector-fast" ),
+        makePtr<DetectorQualityEvaluator>( "GFTT", "quality-detector-gftt" ),
+        makePtr<DetectorQualityEvaluator>( "HARRIS", "quality-detector-harris" ),
+        makePtr<DetectorQualityEvaluator>( "MSER", "quality-detector-mser" ),
+        makePtr<DetectorQualityEvaluator>( "STAR", "quality-detector-star" ),
+        makePtr<DetectorQualityEvaluator>( "SIFT", "quality-detector-sift" ),
+        makePtr<DetectorQualityEvaluator>( "SURF", "quality-detector-surf" ),
 
-        new DescriptorQualityEvaluator( "SIFT", "quality-descriptor-sift", "BruteForce" ),
-        new DescriptorQualityEvaluator( "SURF", "quality-descriptor-surf", "BruteForce" ),
-        new DescriptorQualityEvaluator( "FERN", "quality-descriptor-fern"),
-        new CalonderDescriptorQualityEvaluator()
+        makePtr<DescriptorQualityEvaluator>( "SIFT", "quality-descriptor-sift", "BruteForce" ),
+        makePtr<DescriptorQualityEvaluator>( "SURF", "quality-descriptor-surf", "BruteForce" ),
+        makePtr<DescriptorQualityEvaluator>( "FERN", "quality-descriptor-fern"),
+        makePtr<CalonderDescriptorQualityEvaluator>()
     };
 
     for( size_t i = 0; i < sizeof(evals)/sizeof(evals[0]); i++ )
diff --git a/samples/cpp/detector_descriptor_matcher_evaluation.cpp b/samples/cpp/detector_descriptor_matcher_evaluation.cpp
index bf92e7da33..194cddfe35 100644
--- a/samples/cpp/detector_descriptor_matcher_evaluation.cpp
+++ b/samples/cpp/detector_descriptor_matcher_evaluation.cpp
@@ -39,9 +39,10 @@
 //
 //M*/
 
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/features2d/features2d.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/features2d.hpp"
 
 #include <stdlib.h>
 #include <stdio.h>
diff --git a/samples/cpp/dft.cpp b/samples/cpp/dft.cpp
index d538c103e6..4708defabb 100644
--- a/samples/cpp/dft.cpp
+++ b/samples/cpp/dft.cpp
@@ -1,6 +1,7 @@
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <stdio.h>
 
@@ -26,7 +27,7 @@ int main(int argc, const char ** argv)
     CommandLineParser parser(argc, argv, keys);
     string filename = parser.get<string>(0);
 
-    Mat img = imread(filename.c_str(), CV_LOAD_IMAGE_GRAYSCALE);
+    Mat img = imread(filename.c_str(), IMREAD_GRAYSCALE);
     if( img.empty() )
     {
         help();
@@ -73,10 +74,9 @@ int main(int argc, const char ** argv)
     q2.copyTo(q1);
     tmp.copyTo(q2);
 
-    normalize(mag, mag, 0, 1, CV_MINMAX);
+    normalize(mag, mag, 0, 1, NORM_MINMAX);
 
     imshow("spectrum magnitude", mag);
     waitKey();
     return 0;
 }
-
diff --git a/samples/cpp/distrans.cpp b/samples/cpp/distrans.cpp
index 472c168e30..1afc9b37ee 100644
--- a/samples/cpp/distrans.cpp
+++ b/samples/cpp/distrans.cpp
@@ -1,15 +1,16 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <stdio.h>
 
 using namespace std;
 using namespace cv;
 
-int maskSize0 = CV_DIST_MASK_5;
+int maskSize0 = DIST_MASK_5;
 int voronoiType = -1;
 int edgeThresh = 100;
-int distType0 = CV_DIST_L1;
+int distType0 = DIST_L1;
 
 // The output and temporary images
 Mat gray;
@@ -30,8 +31,8 @@ static void onTrackbar( int, void* )
         Scalar(255,0,255)
     };
 
-    int maskSize = voronoiType >= 0 ? CV_DIST_MASK_5 : maskSize0;
-    int distType = voronoiType >= 0 ? CV_DIST_L2 : distType0;
+    int maskSize = voronoiType >= 0 ? DIST_MASK_5 : maskSize0;
+    int distType = voronoiType >= 0 ? DIST_L2 : distType0;
 
     Mat edge = gray >= edgeThresh, dist, labels, dist8u;
 
@@ -129,7 +130,7 @@ int main( int argc, const char** argv )
         // Call to update the view
         onTrackbar(0, 0);
 
-        int c = cvWaitKey(0) & 255;
+        int c = waitKey() & 255;
 
         if( c == 27 )
             break;
@@ -139,17 +140,17 @@ int main( int argc, const char** argv )
             voronoiType = -1;
 
         if( c == 'c' || c == 'C' )
-            distType0 = CV_DIST_C;
+            distType0 = DIST_C;
         else if( c == '1' )
-            distType0 = CV_DIST_L1;
+            distType0 = DIST_L1;
         else if( c == '2' )
-            distType0 = CV_DIST_L2;
+            distType0 = DIST_L2;
         else if( c == '3' )
-            maskSize0 = CV_DIST_MASK_3;
+            maskSize0 = DIST_MASK_3;
         else if( c == '5' )
-            maskSize0 = CV_DIST_MASK_5;
+            maskSize0 = DIST_MASK_5;
         else if( c == '0' )
-            maskSize0 = CV_DIST_MASK_PRECISE;
+            maskSize0 = DIST_MASK_PRECISE;
         else if( c == 'v' )
             voronoiType = 0;
         else if( c == 'p' )
@@ -161,18 +162,18 @@ int main( int argc, const char** argv )
             else if( voronoiType == 1 )
             {
                 voronoiType = -1;
-                maskSize0 = CV_DIST_MASK_3;
-                distType0 = CV_DIST_C;
+                maskSize0 = DIST_MASK_3;
+                distType0 = DIST_C;
             }
-            else if( distType0 == CV_DIST_C )
-                distType0 = CV_DIST_L1;
-            else if( distType0 == CV_DIST_L1 )
-                distType0 = CV_DIST_L2;
-            else if( maskSize0 == CV_DIST_MASK_3 )
-                maskSize0 = CV_DIST_MASK_5;
-            else if( maskSize0 == CV_DIST_MASK_5 )
-                maskSize0 = CV_DIST_MASK_PRECISE;
-            else if( maskSize0 == CV_DIST_MASK_PRECISE )
+            else if( distType0 == DIST_C )
+                distType0 = DIST_L1;
+            else if( distType0 == DIST_L1 )
+                distType0 = DIST_L2;
+            else if( maskSize0 == DIST_MASK_3 )
+                maskSize0 = DIST_MASK_5;
+            else if( maskSize0 == DIST_MASK_5 )
+                maskSize0 = DIST_MASK_PRECISE;
+            else if( maskSize0 == DIST_MASK_PRECISE )
                 voronoiType = 0;
         }
     }
diff --git a/samples/cpp/drawing.cpp b/samples/cpp/drawing.cpp
index 5467463902..b0848168df 100644
--- a/samples/cpp/drawing.cpp
+++ b/samples/cpp/drawing.cpp
@@ -21,7 +21,7 @@ int main()
     char wndname[] = "Drawing Demo";
     const int NUMBER = 100;
     const int DELAY = 5;
-    int lineType = CV_AA; // change it to 8 to see non-antialiased graphics
+    int lineType = LINE_AA; // change it to LINE_8 to see non-antialiased graphics
     int i, width = 1000, height = 700;
     int x1 = -width/2, x2 = width*3/2, y1 = -height/2, y2 = height*3/2;
     RNG rng(0xFFFFFFFF);
@@ -157,14 +157,14 @@ int main()
             return 0;
     }
 
-    Size textsize = getTextSize("OpenCV forever!", CV_FONT_HERSHEY_COMPLEX, 3, 5, 0);
+    Size textsize = getTextSize("OpenCV forever!", FONT_HERSHEY_COMPLEX, 3, 5, 0);
     Point org((width - textsize.width)/2, (height - textsize.height)/2);
 
     Mat image2;
     for( i = 0; i < 255; i += 2 )
     {
         image2 = image - Scalar::all(i);
-        putText(image2, "OpenCV forever!", org, CV_FONT_HERSHEY_COMPLEX, 3,
+        putText(image2, "OpenCV forever!", org, FONT_HERSHEY_COMPLEX, 3,
                 Scalar(i, i, 255), 5, lineType);
 
         imshow(wndname, image2);
diff --git a/samples/cpp/edge.cpp b/samples/cpp/edge.cpp
index c6a8db3b6a..262833c672 100644
--- a/samples/cpp/edge.cpp
+++ b/samples/cpp/edge.cpp
@@ -1,5 +1,6 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <stdio.h>
 
@@ -49,7 +50,7 @@ int main( int argc, const char** argv )
         return -1;
     }
     cedge.create(image.size(), image.type());
-    cvtColor(image, gray, CV_BGR2GRAY);
+    cvtColor(image, gray, COLOR_BGR2GRAY);
 
     // Create a window
     namedWindow("Edge map", 1);
diff --git a/samples/cpp/em.cpp b/samples/cpp/em.cpp
index 4c4664b8f8..3a070c8cf3 100644
--- a/samples/cpp/em.cpp
+++ b/samples/cpp/em.cpp
@@ -1,5 +1,5 @@
-#include "opencv2/legacy/legacy.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/legacy.hpp"
 
 using namespace cv;
 
@@ -45,7 +45,7 @@ int main( int /*argc*/, char** /*argv*/ )
     params.start_step         = CvEM::START_AUTO_STEP;
     params.term_crit.max_iter = 300;
     params.term_crit.epsilon  = 0.1;
-    params.term_crit.type     = CV_TERMCRIT_ITER|CV_TERMCRIT_EPS;
+    params.term_crit.type     = TermCriteria::COUNT|TermCriteria::EPS;
 
     // cluster the data
     em_model.train( samples, Mat(), params, &labels );
@@ -76,7 +76,7 @@ int main( int /*argc*/, char** /*argv*/ )
             int response = cvRound(em_model.predict( sample ));
             Scalar c = colors[response];
 
-            circle( img, Point(j, i), 1, c*0.75, CV_FILLED );
+            circle( img, Point(j, i), 1, c*0.75, FILLED );
         }
     }
 
@@ -84,7 +84,7 @@ int main( int /*argc*/, char** /*argv*/ )
     for( i = 0; i < nsamples; i++ )
     {
         Point pt(cvRound(samples.at<float>(i, 0)), cvRound(samples.at<float>(i, 1)));
-        circle( img, pt, 1, colors[labels.at<int>(i)], CV_FILLED );
+        circle( img, pt, 1, colors[labels.at<int>(i)], FILLED );
     }
 
     imshow( "EM-clustering result", img );
diff --git a/samples/cpp/erfilter.cpp b/samples/cpp/erfilter.cpp
new file mode 100644
index 0000000000..8a2a45f9c4
--- /dev/null
+++ b/samples/cpp/erfilter.cpp
@@ -0,0 +1,120 @@
+
+//--------------------------------------------------------------------------------------------------
+//  A demo program of the Extremal Region Filter algorithm described in
+//  Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012
+//--------------------------------------------------------------------------------------------------
+
+#include  "opencv2/opencv.hpp"
+#include  "opencv2/objdetect.hpp"
+#include  "opencv2/highgui.hpp"
+#include  "opencv2/imgproc.hpp"
+
+#include  <vector>
+#include  <iostream>
+#include  <iomanip>
+
+using  namespace std;
+using  namespace cv;
+
+void  er_draw(Mat &src, Mat &dst, ERStat& er);
+
+void  er_draw(Mat &src, Mat &dst, ERStat& er)
+{
+
+    if (er.parent != NULL) // deprecate the root region
+    {
+        int newMaskVal = 255;
+        int flags = 4 + (newMaskVal << 8) + FLOODFILL_FIXED_RANGE + FLOODFILL_MASK_ONLY;
+        floodFill(src,dst,Point(er.pixel%src.cols,er.pixel/src.cols),Scalar(255),0,Scalar(er.level),Scalar(0),flags);
+    }
+
+}
+
+int  main(int argc, const char * argv[])
+{
+
+
+    vector<ERStat> regions;
+
+    if (argc < 2) {
+        cout << "Demo program of the Extremal Region Filter algorithm described in " << endl;
+        cout << "Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012" << endl << endl;
+        cout << "    Usage: " << argv[0] << " input_image <optional_groundtruth_image>" << endl;
+        cout << "    Default classifier files (trained_classifierNM*.xml) should be in ./" << endl;
+        return -1;
+    }
+
+    Mat original = imread(argv[1]);
+    Mat gt;
+    if (argc > 2)
+    {
+        gt = imread(argv[2]);
+        cvtColor(gt, gt, COLOR_RGB2GRAY);
+        threshold(gt, gt, 254, 255, THRESH_BINARY);
+    }
+    Mat grey(original.size(),CV_8UC1);
+    cvtColor(original,grey,COLOR_RGB2GRAY);
+
+    double t = (double)getTickCount();
+
+    // Build ER tree and filter with the 1st stage default classifier
+    Ptr<ERFilter> er_filter1 = createERFilterNM1();
+
+    er_filter1->run(grey, regions);
+
+    t = (double)getTickCount() - t;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+    cout << "\t FIRST STAGE CLASSIFIER done in " << t * 1000. / getTickFrequency() << " ms." << endl;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+    cout << setw(9) << regions.size()+er_filter1->getNumRejected() << "\t Extremal Regions extracted " << endl;
+    cout << setw(9) << regions.size() << "\t Extremal Regions selected by the first stage of the sequential classifier." << endl;
+    cout << "\t \t (saving into out_second_stage.jpg)" << endl;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+
+    er_filter1.release();
+
+    // draw regions
+    Mat mask = Mat::zeros(grey.rows+2,grey.cols+2,CV_8UC1);
+    for (int r=0; r<(int)regions.size(); r++)
+        er_draw(grey, mask, regions.at(r));
+    mask = 255-mask;
+    imwrite("out_first_stage.jpg", mask);
+
+    if (argc > 2)
+    {
+        Mat tmp_mask = (255-gt) & (255-mask(Rect(Point(1,1),Size(mask.cols-2,mask.rows-2))));
+        cout << "Recall for the 1st stage filter = " << (float)countNonZero(tmp_mask) / countNonZero(255-gt) << endl;
+    }
+
+    t = (double)getTickCount();
+
+    // Default second stage classifier
+    Ptr<ERFilter> er_filter2 = createERFilterNM2();
+    er_filter2->run(grey, regions);
+
+    t = (double)getTickCount() - t;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+    cout << "\t SECOND STAGE CLASSIFIER done in " << t * 1000. / getTickFrequency() << " ms." << endl;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+    cout << setw(9) << regions.size() << "\t Extremal Regions selected by the second stage of the sequential classifier." << endl;
+    cout << "\t \t (saving into out_second_stage.jpg)" << endl;
+    cout << " --------------------------------------------------------------------------------------------------" << endl;
+
+    er_filter2.release();
+
+    // draw regions
+    mask = mask*0;
+    for (int r=0; r<(int)regions.size(); r++)
+        er_draw(grey, mask, regions.at(r));
+    mask = 255-mask;
+    imwrite("out_second_stage.jpg", mask);
+
+    if (argc > 2)
+    {
+        Mat tmp_mask = (255-gt) & (255-mask(Rect(Point(1,1),Size(mask.cols-2,mask.rows-2))));
+        cout << "Recall for the 2nd stage filter = " << (float)countNonZero(tmp_mask) / countNonZero(255-gt) << endl;
+    }
+
+    regions.clear();
+
+}
diff --git a/samples/cpp/fabmap_sample.cpp b/samples/cpp/fabmap_sample.cpp
index bbe5f2b4c5..cd06b55cb8 100644
--- a/samples/cpp/fabmap_sample.cpp
+++ b/samples/cpp/fabmap_sample.cpp
@@ -49,9 +49,11 @@
 //
 //M*/
 
+#include <iostream>
 
-#include "opencv2/opencv.hpp"
-#include "opencv2/nonfree/nonfree.hpp"
+#include "opencv2/contrib.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/nonfree.hpp"
 
 using namespace cv;
 using namespace std;
@@ -129,11 +131,11 @@ int main(int argc, char * argv[]) {
     //generate test data
     cout << "Extracting Test Data from images" << endl <<
         endl;
-    Ptr<FeatureDetector> detector =
+    Ptr<FeatureDetector> detector(
         new DynamicAdaptedFeatureDetector(
-        AdjusterAdapter::create("STAR"), 130, 150, 5);
-    Ptr<DescriptorExtractor> extractor =
-        new SurfDescriptorExtractor(1000, 4, 2, false, true);
+            AdjusterAdapter::create("STAR"), 130, 150, 5));
+    Ptr<DescriptorExtractor> extractor(
+        new SurfDescriptorExtractor(1000, 4, 2, false, true));
     Ptr<DescriptorMatcher> matcher =
         DescriptorMatcher::create("FlannBased");
 
@@ -181,8 +183,8 @@ int main(int argc, char * argv[]) {
         endl;
     Ptr<of2::FabMap> fabmap;
 
-    fabmap = new of2::FabMap2(tree, 0.39, 0, of2::FabMap::SAMPLED |
-        of2::FabMap::CHOW_LIU);
+    fabmap.reset(new of2::FabMap2(tree, 0.39, 0, of2::FabMap::SAMPLED |
+        of2::FabMap::CHOW_LIU));
     fabmap->addTraining(trainData);
 
     vector<of2::IMatch> matches;
@@ -204,7 +206,7 @@ int main(int argc, char * argv[]) {
     }
 
     Mat result_large(100, 100, CV_8UC1);
-    resize(result_small, result_large, Size(500, 500), 0, 0, CV_INTER_NN);
+    resize(result_small, result_large, Size(500, 500), 0, 0, INTER_NEAREST);
 
     cout << endl << "Press any key to exit" << endl;
     imshow("Confusion Matrix", result_large);
diff --git a/samples/cpp/facerec_demo.cpp b/samples/cpp/facerec_demo.cpp
index 6402082e82..3a104b86b3 100644
--- a/samples/cpp/facerec_demo.cpp
+++ b/samples/cpp/facerec_demo.cpp
@@ -16,9 +16,10 @@
  *   See <http://www.opensource.org/licenses/bsd-license>
  */
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/contrib/contrib.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/contrib.hpp"
 
 #include <iostream>
 #include <fstream>
@@ -31,7 +32,7 @@ static Mat toGrayscale(InputArray _src) {
     Mat src = _src.getMat();
     // only allow one channel
     if(src.channels() != 1) {
-        CV_Error(CV_StsBadArg, "Only Matrices with one channel are supported");
+        CV_Error(Error::StsBadArg, "Only Matrices with one channel are supported");
     }
     // create and return normalized image
     Mat dst;
@@ -43,7 +44,7 @@ static void read_csv(const string& filename, vector<Mat>& images, vector<int>& l
     std::ifstream file(filename.c_str(), ifstream::in);
     if (!file) {
         string error_message = "No valid input file was given, please check the given filename.";
-        CV_Error(CV_StsBadArg, error_message);
+        CV_Error(Error::StsBadArg, error_message);
     }
     string line, path, classlabel;
     while (getline(file, line)) {
@@ -81,7 +82,7 @@ int main(int argc, const char *argv[]) {
     // Quit if there are not enough images for this demo.
     if(images.size() <= 1) {
         string error_message = "This demo needs at least 2 images to work. Please add more images to your data set!";
-        CV_Error(CV_StsError, error_message);
+        CV_Error(Error::StsError, error_message);
     }
     // Get the height from the first image. We'll need this
     // later in code to reshape the images to their original
diff --git a/samples/cpp/fback.cpp b/samples/cpp/fback.cpp
index 4923e8abcc..5293fd1f7f 100644
--- a/samples/cpp/fback.cpp
+++ b/samples/cpp/fback.cpp
@@ -42,13 +42,13 @@ int main(int, char**)
     for(;;)
     {
         cap >> frame;
-        cvtColor(frame, gray, CV_BGR2GRAY);
+        cvtColor(frame, gray, COLOR_BGR2GRAY);
 
         if( prevgray.data )
         {
             calcOpticalFlowFarneback(prevgray, gray, flow, 0.5, 3, 15, 3, 5, 1.2, 0);
-            cvtColor(prevgray, cflow, CV_GRAY2BGR);
-            drawOptFlowMap(flow, cflow, 16, 1.5, CV_RGB(0, 255, 0));
+            cvtColor(prevgray, cflow, COLOR_GRAY2BGR);
+            drawOptFlowMap(flow, cflow, 16, 1.5, Scalar(0, 255, 0));
             imshow("flow", cflow);
         }
         if(waitKey(30)>=0)
diff --git a/samples/cpp/ffilldemo.cpp b/samples/cpp/ffilldemo.cpp
index a7fff778fd..1cdce9bb5c 100644
--- a/samples/cpp/ffilldemo.cpp
+++ b/samples/cpp/ffilldemo.cpp
@@ -34,14 +34,14 @@ int newMaskVal = 255;
 
 static void onMouse( int event, int x, int y, int, void* )
 {
-    if( event != CV_EVENT_LBUTTONDOWN )
+    if( event != EVENT_LBUTTONDOWN )
         return;
 
     Point seed = Point(x,y);
     int lo = ffillMode == 0 ? 0 : loDiff;
     int up = ffillMode == 0 ? 0 : upDiff;
     int flags = connectivity + (newMaskVal << 8) +
-                (ffillMode == 1 ? CV_FLOODFILL_FIXED_RANGE : 0);
+                (ffillMode == 1 ? FLOODFILL_FIXED_RANGE : 0);
     int b = (unsigned)theRNG() & 255;
     int g = (unsigned)theRNG() & 255;
     int r = (unsigned)theRNG() & 255;
@@ -53,7 +53,7 @@ static void onMouse( int event, int x, int y, int, void* )
 
     if( useMask )
     {
-        threshold(mask, mask, 1, 128, CV_THRESH_BINARY);
+        threshold(mask, mask, 1, 128, THRESH_BINARY);
         area = floodFill(dst, mask, seed, newVal, &ccomp, Scalar(lo, lo, lo),
                   Scalar(up, up, up), flags);
         imshow( "mask", mask );
@@ -81,7 +81,7 @@ int main( int argc, char** argv )
     }
     help();
     image0.copyTo(image);
-    cvtColor(image0, gray, CV_BGR2GRAY);
+    cvtColor(image0, gray, COLOR_BGR2GRAY);
     mask.create(image0.rows+2, image0.cols+2, CV_8UC1);
 
     namedWindow( "image", 0 );
@@ -106,7 +106,7 @@ int main( int argc, char** argv )
             if( isColor )
             {
                 cout << "Grayscale mode is set\n";
-                cvtColor(image0, gray, CV_BGR2GRAY);
+                cvtColor(image0, gray, COLOR_BGR2GRAY);
                 mask = Scalar::all(0);
                 isColor = false;
             }
@@ -135,7 +135,7 @@ int main( int argc, char** argv )
         case 'r':
             cout << "Original image is restored\n";
             image0.copyTo(image);
-            cvtColor(image, gray, CV_BGR2GRAY);
+            cvtColor(image, gray, COLOR_BGR2GRAY);
             mask = Scalar::all(0);
             break;
         case 's':
diff --git a/samples/cpp/fitellipse.cpp b/samples/cpp/fitellipse.cpp
index 55f9e4ab06..c42f8f3c3e 100644
--- a/samples/cpp/fitellipse.cpp
+++ b/samples/cpp/fitellipse.cpp
@@ -64,7 +64,7 @@ void processImage(int /*h*/, void*)
     vector<vector<Point> > contours;
     Mat bimage = image >= sliderPos;
 
-    findContours(bimage, contours, CV_RETR_LIST, CV_CHAIN_APPROX_NONE);
+    findContours(bimage, contours, RETR_LIST, CHAIN_APPROX_NONE);
 
     Mat cimage = Mat::zeros(bimage.size(), CV_8UC3);
 
@@ -82,12 +82,12 @@ void processImage(int /*h*/, void*)
             continue;
         drawContours(cimage, contours, (int)i, Scalar::all(255), 1, 8);
 
-        ellipse(cimage, box, Scalar(0,0,255), 1, CV_AA);
-        ellipse(cimage, box.center, box.size*0.5f, box.angle, 0, 360, Scalar(0,255,255), 1, CV_AA);
+        ellipse(cimage, box, Scalar(0,0,255), 1, LINE_AA);
+        ellipse(cimage, box.center, box.size*0.5f, box.angle, 0, 360, Scalar(0,255,255), 1, LINE_AA);
         Point2f vtx[4];
         box.points(vtx);
         for( int j = 0; j < 4; j++ )
-            line(cimage, vtx[j], vtx[(j+1)%4], Scalar(0,255,0), 1, CV_AA);
+            line(cimage, vtx[j], vtx[(j+1)%4], Scalar(0,255,0), 1, LINE_AA);
     }
 
     imshow("result", cimage);
diff --git a/samples/cpp/freak_demo.cpp b/samples/cpp/freak_demo.cpp
index 60778fd4e2..fa58d20385 100644
--- a/samples/cpp/freak_demo.cpp
+++ b/samples/cpp/freak_demo.cpp
@@ -38,11 +38,11 @@
 #include <string>
 #include <vector>
 
-#include <opencv2/core/core.hpp>
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/features2d/features2d.hpp>
-#include <opencv2/nonfree/features2d.hpp>
-#include <opencv2/legacy/legacy.hpp>
+#include <opencv2/core.hpp>
+#include "opencv2/core/utility.hpp"
+#include <opencv2/highgui.hpp>
+#include <opencv2/features2d.hpp>
+#include <opencv2/nonfree.hpp>
 
 using namespace cv;
 
@@ -56,7 +56,7 @@ static void help( char** argv )
 }
 
 int main( int argc, char** argv ) {
-    // check http://opencv.itseez.com/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.html
+    // check http://docs.opencv.org/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.html
     // for OpenCV general detection/matching framework details
 
     if( argc != 3 ) {
@@ -65,14 +65,14 @@ int main( int argc, char** argv ) {
     }
 
     // Load images
-    Mat imgA = imread(argv[1], CV_LOAD_IMAGE_GRAYSCALE );
+    Mat imgA = imread(argv[1], IMREAD_GRAYSCALE );
     if( !imgA.data ) {
         std::cout<< " --(!) Error reading image " << argv[1] << std::endl;
         return -1;
     }
 
-    Mat imgB = imread(argv[2], CV_LOAD_IMAGE_GRAYSCALE );
-    if( !imgA.data ) {
+    Mat imgB = imread(argv[2], IMREAD_GRAYSCALE );
+    if( !imgB.data ) {
         std::cout << " --(!) Error reading image " << argv[2] << std::endl;
         return -1;
     }
@@ -94,9 +94,9 @@ int main( int argc, char** argv ) {
 
     // MATCHER
     // The standard Hamming distance can be used such as
-    // BruteForceMatcher<Hamming> matcher;
+    // BFMatcher matcher(NORM_HAMMING);
     // or the proposed cascade of hamming distance using SSSE3
-    BruteForceMatcher<Hamming> matcher;
+    BFMatcher matcher(NORM_HAMMING);
 
     // detect
     double t = (double)getTickCount();
@@ -122,7 +122,7 @@ int main( int argc, char** argv ) {
     Mat imgMatch;
     drawMatches(imgA, keypointsA, imgB, keypointsB, matches, imgMatch);
 
-    namedWindow("matches", CV_WINDOW_KEEPRATIO);
+    namedWindow("matches", WINDOW_KEEPRATIO);
     imshow("matches", imgMatch);
     waitKey(0);
 }
diff --git a/samples/cpp/generic_descriptor_match.cpp b/samples/cpp/generic_descriptor_match.cpp
index 7c69ccdeda..359f3c08e1 100644
--- a/samples/cpp/generic_descriptor_match.cpp
+++ b/samples/cpp/generic_descriptor_match.cpp
@@ -33,15 +33,15 @@ int main(int argc, char** argv)
     std::string params_filename = std::string(argv[4]);
 
     Ptr<GenericDescriptorMatcher> descriptorMatcher = GenericDescriptorMatcher::create(alg_name, params_filename);
-    if( descriptorMatcher.empty() )
+    if( !descriptorMatcher )
     {
         printf ("Cannot create descriptor\n");
         return 0;
     }
 
     //printf("Reading the images...\n");
-    Mat img1 = imread(img1_name, CV_LOAD_IMAGE_GRAYSCALE);
-    Mat img2 = imread(img2_name, CV_LOAD_IMAGE_GRAYSCALE);
+    Mat img1 = imread(img1_name, IMREAD_GRAYSCALE);
+    Mat img2 = imread(img2_name, IMREAD_GRAYSCALE);
 
     // extract keypoints from the first image
     SURF surf_extractor(5.0e3);
@@ -80,7 +80,7 @@ Mat DrawCorrespondences(const Mat& img1, const vector<KeyPoint>& features1, cons
 
     for (size_t i = 0; i < features1.size(); i++)
     {
-        circle(img_corr, features1[i].pt, 3, CV_RGB(255, 0, 0));
+        circle(img_corr, features1[i].pt, 3, Scalar(255, 0, 0));
     }
 
     for (size_t i = 0; i < features2.size(); i++)
diff --git a/samples/cpp/grabcut.cpp b/samples/cpp/grabcut.cpp
index 40280e4570..f276e75666 100644
--- a/samples/cpp/grabcut.cpp
+++ b/samples/cpp/grabcut.cpp
@@ -33,13 +33,13 @@ const Scalar BLUE = Scalar(255,0,0);
 const Scalar LIGHTBLUE = Scalar(255,255,160);
 const Scalar GREEN = Scalar(0,255,0);
 
-const int BGD_KEY = CV_EVENT_FLAG_CTRLKEY;
-const int FGD_KEY = CV_EVENT_FLAG_SHIFTKEY;
+const int BGD_KEY = EVENT_FLAG_CTRLKEY;
+const int FGD_KEY = EVENT_FLAG_SHIFTKEY;
 
 static void getBinMask( const Mat& comMask, Mat& binMask )
 {
     if( comMask.empty() || comMask.type()!=CV_8UC1 )
-        CV_Error( CV_StsBadArg, "comMask is empty or has incorrect type (not CV_8UC1)" );
+        CV_Error( Error::StsBadArg, "comMask is empty or has incorrect type (not CV_8UC1)" );
     if( binMask.empty() || binMask.rows!=comMask.rows || binMask.cols!=comMask.cols )
         binMask.create( comMask.size(), CV_8UC1 );
     binMask = comMask & 1;
@@ -132,7 +132,7 @@ void GCApplication::showImage() const
 
 void GCApplication::setRectInMask()
 {
-    assert( !mask.empty() );
+    CV_Assert( !mask.empty() );
     mask.setTo( GC_BGD );
     rect.x = max(0, rect.x);
     rect.y = max(0, rect.y);
@@ -176,7 +176,7 @@ void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
     // TODO add bad args check
     switch( event )
     {
-    case CV_EVENT_LBUTTONDOWN: // set rect or GC_BGD(GC_FGD) labels
+    case EVENT_LBUTTONDOWN: // set rect or GC_BGD(GC_FGD) labels
         {
             bool isb = (flags & BGD_KEY) != 0,
                  isf = (flags & FGD_KEY) != 0;
@@ -189,7 +189,7 @@ void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
                 lblsState = IN_PROCESS;
         }
         break;
-    case CV_EVENT_RBUTTONDOWN: // set GC_PR_BGD(GC_PR_FGD) labels
+    case EVENT_RBUTTONDOWN: // set GC_PR_BGD(GC_PR_FGD) labels
         {
             bool isb = (flags & BGD_KEY) != 0,
                  isf = (flags & FGD_KEY) != 0;
@@ -197,13 +197,13 @@ void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
                 prLblsState = IN_PROCESS;
         }
         break;
-    case CV_EVENT_LBUTTONUP:
+    case EVENT_LBUTTONUP:
         if( rectState == IN_PROCESS )
         {
             rect = Rect( Point(rect.x, rect.y), Point(x,y) );
             rectState = SET;
             setRectInMask();
-            assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
+            CV_Assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
             showImage();
         }
         if( lblsState == IN_PROCESS )
@@ -213,7 +213,7 @@ void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
             showImage();
         }
         break;
-    case CV_EVENT_RBUTTONUP:
+    case EVENT_RBUTTONUP:
         if( prLblsState == IN_PROCESS )
         {
             setLblsInMask(flags, Point(x,y), true);
@@ -221,11 +221,11 @@ void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
             showImage();
         }
         break;
-    case CV_EVENT_MOUSEMOVE:
+    case EVENT_MOUSEMOVE:
         if( rectState == IN_PROCESS )
         {
             rect = Rect( Point(rect.x, rect.y), Point(x,y) );
-            assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
+            CV_Assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
             showImage();
         }
         else if( lblsState == IN_PROCESS )
@@ -296,15 +296,15 @@ int main( int argc, char** argv )
     help();
 
     const string winName = "image";
-    cvNamedWindow( winName.c_str(), CV_WINDOW_AUTOSIZE );
-    cvSetMouseCallback( winName.c_str(), on_mouse, 0 );
+    namedWindow( winName.c_str(), WINDOW_AUTOSIZE );
+    setMouseCallback( winName.c_str(), on_mouse, 0 );
 
     gcapp.setImageAndWinName( image, winName );
     gcapp.showImage();
 
     for(;;)
     {
-        int c = cvWaitKey(0);
+        int c = waitKey();
         switch( (char) c )
         {
         case '\x1b':
@@ -331,6 +331,6 @@ int main( int argc, char** argv )
     }
 
 exit_main:
-    cvDestroyWindow( winName.c_str() );
+    destroyWindow( winName.c_str() );
     return 0;
 }
diff --git a/samples/cpp/houghcircles.cpp b/samples/cpp/houghcircles.cpp
index bc53fd1249..12f1c57677 100644
--- a/samples/cpp/houghcircles.cpp
+++ b/samples/cpp/houghcircles.cpp
@@ -27,18 +27,18 @@ int main(int argc, char** argv)
 
     Mat cimg;
     medianBlur(img, img, 5);
-    cvtColor(img, cimg, CV_GRAY2BGR);
+    cvtColor(img, cimg, COLOR_GRAY2BGR);
 
     vector<Vec3f> circles;
-    HoughCircles(img, circles, CV_HOUGH_GRADIENT, 1, 10,
+    HoughCircles(img, circles, HOUGH_GRADIENT, 1, 10,
                  100, 30, 1, 30 // change the last two parameters
                                 // (min_radius & max_radius) to detect larger circles
                  );
     for( size_t i = 0; i < circles.size(); i++ )
     {
         Vec3i c = circles[i];
-        circle( cimg, Point(c[0], c[1]), c[2], Scalar(0,0,255), 3, CV_AA);
-        circle( cimg, Point(c[0], c[1]), 2, Scalar(0,255,0), 3, CV_AA);
+        circle( cimg, Point(c[0], c[1]), c[2], Scalar(0,0,255), 3, LINE_AA);
+        circle( cimg, Point(c[0], c[1]), 2, Scalar(0,255,0), 3, LINE_AA);
     }
 
     imshow("detected circles", cimg);
diff --git a/samples/cpp/houghlines.cpp b/samples/cpp/houghlines.cpp
index b79ea2f5ff..2fdabe8397 100644
--- a/samples/cpp/houghlines.cpp
+++ b/samples/cpp/houghlines.cpp
@@ -27,7 +27,7 @@ int main(int argc, char** argv)
 
     Mat dst, cdst;
     Canny(src, dst, 50, 200, 3);
-    cvtColor(dst, cdst, CV_GRAY2BGR);
+    cvtColor(dst, cdst, COLOR_GRAY2BGR);
 
 #if 0
     vector<Vec2f> lines;
@@ -51,7 +51,7 @@ int main(int argc, char** argv)
     for( size_t i = 0; i < lines.size(); i++ )
     {
         Vec4i l = lines[i];
-        line( cdst, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0,0,255), 3, CV_AA);
+        line( cdst, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0,0,255), 3, LINE_AA);
     }
 #endif
     imshow("source", src);
@@ -61,4 +61,3 @@ int main(int argc, char** argv)
 
     return 0;
 }
-
diff --git a/samples/cpp/hybridtrackingsample.cpp b/samples/cpp/hybridtrackingsample.cpp
index 8c925ac9d5..b343dfed63 100644
--- a/samples/cpp/hybridtrackingsample.cpp
+++ b/samples/cpp/hybridtrackingsample.cpp
@@ -11,10 +11,6 @@
  *
  */
 
-//#include <cv.h>
-//#include <ml.h>
-//#include <cvaux.h>
-//#include <highgui.h>
 #include <stdio.h>
 #include <time.h>
 #include <iostream>
@@ -47,12 +43,12 @@ static void onMouse(int event, int x, int y, int, void*) {
     }
 
     switch (event) {
-    case CV_EVENT_LBUTTONDOWN:
+    case EVENT_LBUTTONDOWN:
         origin = Point(x, y);
         selection = Rect(x, y, 0, 0);
         selectObject = true;
         break;
-    case CV_EVENT_LBUTTONUP:
+    case EVENT_LBUTTONUP:
         selectObject = false;
         trackObject = -1;
         break;
@@ -96,8 +92,8 @@ int main(int argc, char** argv)
             return 0;
         }
         cout << "Opened camera" << endl;
-        cap.set(CV_CAP_PROP_FRAME_WIDTH, 640);
-        cap.set(CV_CAP_PROP_FRAME_HEIGHT, 480);
+        cap.set(CAP_PROP_FRAME_WIDTH, 640);
+        cap.set(CAP_PROP_FRAME_HEIGHT, 480);
         cap >> frame;
     }
 
@@ -137,7 +133,7 @@ int main(int argc, char** argv)
             int values_read = fscanf(f, "%d %f %f %f %f\n", &i, &w[0], &w[1], &w[2], &w[3]);
             CV_Assert(values_read == 5);
             sprintf(img_file, "seqG/%04d.png", i);
-            image = imread(img_file, CV_LOAD_IMAGE_COLOR);
+            image = imread(img_file, IMREAD_COLOR);
             if (image.empty())
                 break;
             selection = Rect(cvRound(w[0]*image.cols), cvRound(w[1]*image.rows),
diff --git a/samples/cpp/image.cpp b/samples/cpp/image.cpp
index 5647f1c785..80f80c7aff 100644
--- a/samples/cpp/image.cpp
+++ b/samples/cpp/image.cpp
@@ -1,8 +1,9 @@
 #include <stdio.h>
 #include <iostream>
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/flann/miniflann.hpp"
+#include <opencv2/imgproc/imgproc.hpp>
+#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/flann/miniflann.hpp>
+#include <opencv2/core/utility.hpp>
 
 using namespace cv; // all the new API is put into "cv" namespace. Export its content
 using namespace std;
@@ -21,18 +22,22 @@ static void help()
 // enable/disable use of mixed API in the code below.
 #define DEMO_MIXED_API_USE 1
 
+#ifdef DEMO_MIXED_API_USE
+#  include <opencv2/highgui/highgui_c.h>
+#endif
+
 int main( int argc, char** argv )
 {
     help();
     const char* imagename = argc > 1 ? argv[1] : "lena.jpg";
 #if DEMO_MIXED_API_USE
-    Ptr<IplImage> iplimg = cvLoadImage(imagename); // Ptr<T> is safe ref-conting pointer class
-    if(iplimg.empty())
+    Ptr<IplImage> iplimg(cvLoadImage(imagename)); // Ptr<T> is safe ref-counting pointer class
+    if(!iplimg)
     {
         fprintf(stderr, "Can not load image %s\n", imagename);
         return -1;
     }
-    Mat img(iplimg); // cv::Mat replaces the CvMat and IplImage, but it's easy to convert
+    Mat img = cv::cvarrToMat(iplimg); // cv::Mat replaces the CvMat and IplImage, but it's easy to convert
     // between the old and the new data structures (by default, only the header
     // is converted, while the data is shared)
 #else
@@ -48,7 +53,7 @@ int main( int argc, char** argv )
         return -1;
 
     Mat img_yuv;
-    cvtColor(img, img_yuv, CV_BGR2YCrCb); // convert image to YUV color space. The output image will be created automatically
+    cvtColor(img, img_yuv, COLOR_BGR2YCrCb); // convert image to YUV color space. The output image will be created automatically
 
     vector<Mat> planes; // Vector is template vector class, similar to STL's vector. It can store matrices too.
     split(img_yuv, planes); // split the image into separate color planes
@@ -106,10 +111,10 @@ int main( int argc, char** argv )
     // now merge the results back
     merge(planes, img_yuv);
     // and produce the output RGB image
-    cvtColor(img_yuv, img, CV_YCrCb2BGR);
+    cvtColor(img_yuv, img, COLOR_YCrCb2BGR);
 
     // this is counterpart for cvNamedWindow
-    namedWindow("image with grain", CV_WINDOW_AUTOSIZE);
+    namedWindow("image with grain", WINDOW_AUTOSIZE);
 #if DEMO_MIXED_API_USE
     // this is to demonstrate that img and iplimg really share the data - the result of the above
     // processing is stored in img and thus in iplimg too.
diff --git a/samples/cpp/image_alignment.cpp b/samples/cpp/image_alignment.cpp
index 179a751847..56e8209dd9 100644
--- a/samples/cpp/image_alignment.cpp
+++ b/samples/cpp/image_alignment.cpp
@@ -13,10 +13,10 @@
 * Authors: G. Evangelidis, INRIA, Grenoble, France
 *          M. Asbach, Fraunhofer IAIS, St. Augustin, Germany
 */
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/video/video.hpp>
-#include <opencv2/imgproc/imgproc.hpp>
-#include <opencv2/core/core.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/video.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/core/utility.hpp>
 
 #include <stdio.h>
 #include <string>
@@ -238,7 +238,7 @@ int main (const int argc, const char * argv[])
             warpGround = (Mat_<float>(2,3) << 1, 0, (rng.uniform(10.f, 20.f)),
                 0, 1, (rng.uniform(10.f, 20.f)));
             warpAffine(target_image, template_image, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_EUCLIDEAN:
             angle = CV_PI/30 + CV_PI*rng.uniform((double)-2.f, (double)2.f)/180;
@@ -246,7 +246,7 @@ int main (const int argc, const char * argv[])
             warpGround = (Mat_<float>(2,3) << cos(angle), -sin(angle), (rng.uniform(10.f, 20.f)),
                 sin(angle), cos(angle), (rng.uniform(10.f, 20.f)));
             warpAffine(target_image, template_image, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_AFFINE:
 
@@ -255,7 +255,7 @@ int main (const int argc, const char * argv[])
                 (rng.uniform(-0.03f, 0.03f)), (1-rng.uniform(-0.05f, 0.05f)),
                 (rng.uniform(10.f, 20.f)));
             warpAffine(target_image, template_image, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         case MOTION_HOMOGRAPHY:
             warpGround = (Mat_<float>(3,3) << (1-rng.uniform(-0.05f, 0.05f)),
@@ -263,7 +263,7 @@ int main (const int argc, const char * argv[])
                 (rng.uniform(-0.03f, 0.03f)), (1-rng.uniform(-0.05f, 0.05f)),(rng.uniform(10.f, 20.f)),
                 (rng.uniform(0.0001f, 0.0003f)), (rng.uniform(0.0001f, 0.0003f)), 1.f);
             warpPerspective(target_image, template_image, warpGround,
-                Size(200,200), CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+                Size(200,200), INTER_LINEAR + WARP_INVERSE_MAP);
             break;
         }
     }
@@ -333,10 +333,10 @@ int main (const int argc, const char * argv[])
     Mat warped_image = Mat(template_image.rows, template_image.cols, CV_32FC1);
     if (warp_mode != MOTION_HOMOGRAPHY)
         warpAffine      (target_image, warped_image, warp_matrix, warped_image.size(),
-        CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+        INTER_LINEAR + WARP_INVERSE_MAP);
     else
         warpPerspective (target_image, warped_image, warp_matrix, warped_image.size(),
-        CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP);
+        INTER_LINEAR + WARP_INVERSE_MAP);
 
     //save the warped image
     imwrite(warpedImFile, warped_image);
@@ -347,10 +347,10 @@ int main (const int argc, const char * argv[])
 
         cout << "The warped image has been saved in the file: " << warpedImFile << endl << flush;
 
-        namedWindow ("image",    CV_WINDOW_AUTOSIZE);
-        namedWindow ("template", CV_WINDOW_AUTOSIZE);
-        namedWindow ("warped image",   CV_WINDOW_AUTOSIZE);
-        namedWindow ("error (black: no error)", CV_WINDOW_AUTOSIZE);
+        namedWindow ("image",    WINDOW_AUTOSIZE);
+        namedWindow ("template", WINDOW_AUTOSIZE);
+        namedWindow ("warped image",   WINDOW_AUTOSIZE);
+        namedWindow ("error (black: no error)", WINDOW_AUTOSIZE);
 
         moveWindow  ("template", 350, 350);
         moveWindow  ("warped image",   600, 300);
diff --git a/samples/cpp/image_sequence.cpp b/samples/cpp/image_sequence.cpp
new file mode 100644
index 0000000000..d943621611
--- /dev/null
+++ b/samples/cpp/image_sequence.cpp
@@ -0,0 +1,57 @@
+#include <opencv2/core/core.hpp>
+#include <opencv2/highgui/highgui.hpp>
+
+#include <iostream>
+
+using namespace cv;
+using namespace std;
+
+static void help(char** argv)
+{
+    cout << "\nThis sample shows you how to read a sequence of images using the VideoCapture interface.\n"
+         << "Usage: " << argv[0] << " <image_mask> (example mask: example_%%02d.jpg)\n"
+         << "Image mask defines the name variation for the input images that have to be read as a sequence. \n"
+         << "Using the mask example_%%02d.jpg will read in images labeled as 'example_00.jpg', 'example_01.jpg', etc."
+         << endl;
+}
+
+int main(int argc, char** argv)
+{
+    if(argc != 2)
+    {
+        help(argv);
+        return 1;
+    }
+
+    string first_file = argv[1];
+    VideoCapture sequence(first_file);
+
+    if (!sequence.isOpened())
+    {
+        cerr << "Failed to open the image sequence!\n" << endl;
+        return 1;
+    }
+
+    Mat image;
+    namedWindow("Image sequence | press ESC to close", 1);
+
+    for(;;)
+    {
+        // Read in image from sequence
+        sequence >> image;
+
+        // If no image was retrieved -> end of sequence
+        if(image.empty())
+        {
+            cout << "End of Sequence" << endl;
+            break;
+        }
+
+        imshow("Image sequence | press ESC to close", image);
+
+        if(waitKey(500) == 27)
+            break;
+    }
+
+    return 0;
+}
diff --git a/samples/cpp/inpaint.cpp b/samples/cpp/inpaint.cpp
index 223b2de127..7690595cc2 100644
--- a/samples/cpp/inpaint.cpp
+++ b/samples/cpp/inpaint.cpp
@@ -27,11 +27,11 @@ Point prevPt(-1,-1);
 
 static void onMouse( int event, int x, int y, int flags, void* )
 {
-    if( event == CV_EVENT_LBUTTONUP || !(flags & CV_EVENT_FLAG_LBUTTON) )
+    if( event == EVENT_LBUTTONUP || !(flags & EVENT_FLAG_LBUTTON) )
         prevPt = Point(-1,-1);
-    else if( event == CV_EVENT_LBUTTONDOWN )
+    else if( event == EVENT_LBUTTONDOWN )
         prevPt = Point(x,y);
-    else if( event == CV_EVENT_MOUSEMOVE && (flags & CV_EVENT_FLAG_LBUTTON) )
+    else if( event == EVENT_MOUSEMOVE && (flags & EVENT_FLAG_LBUTTON) )
     {
         Point pt(x,y);
         if( prevPt.x < 0 )
@@ -81,7 +81,7 @@ int main( int argc, char** argv )
         if( c == 'i' || c == ' ' )
         {
             Mat inpainted;
-            inpaint(img, inpaintMask, inpainted, 3, CV_INPAINT_TELEA);
+            inpaint(img, inpaintMask, inpainted, 3, INPAINT_TELEA);
             imshow("inpainted image", inpainted);
         }
     }
diff --git a/samples/cpp/kalman.cpp b/samples/cpp/kalman.cpp
index f54adec922..8f9adc6c27 100644
--- a/samples/cpp/kalman.cpp
+++ b/samples/cpp/kalman.cpp
@@ -12,7 +12,7 @@ static inline Point calcPoint(Point2f center, double R, double angle)
 
 static void help()
 {
-    printf( "\nExamle of c calls to OpenCV's Kalman filter.\n"
+    printf( "\nExample of c calls to OpenCV's Kalman filter.\n"
 "   Tracking of rotating point.\n"
 "   Rotation speed is constant.\n"
 "   Both state and measurements vectors are 1D (a point angle),\n"
@@ -40,7 +40,7 @@ int main(int, char**)
     for(;;)
     {
         randn( state, Scalar::all(0), Scalar::all(0.1) );
-        KF.transitionMatrix = *(Mat_<float>(2, 2) << 1, 1, 0, 1);
+        KF.transitionMatrix = (Mat_<float>(2, 2) << 1, 1, 0, 1);
 
         setIdentity(KF.measurementMatrix);
         setIdentity(KF.processNoiseCov, Scalar::all(1e-5));
@@ -69,18 +69,18 @@ int main(int, char**)
             Point measPt = calcPoint(center, R, measAngle);
 
             // plot points
-            #define drawCross( center, color, d )                                 \
-                line( img, Point( center.x - d, center.y - d ),                \
-                             Point( center.x + d, center.y + d ), color, 1, CV_AA, 0); \
-                line( img, Point( center.x + d, center.y - d ),                \
-                             Point( center.x - d, center.y + d ), color, 1, CV_AA, 0 )
+            #define drawCross( center, color, d )                                        \
+                line( img, Point( center.x - d, center.y - d ),                          \
+                             Point( center.x + d, center.y + d ), color, 1, LINE_AA, 0); \
+                line( img, Point( center.x + d, center.y - d ),                          \
+                             Point( center.x - d, center.y + d ), color, 1, LINE_AA, 0 )
 
             img = Scalar::all(0);
             drawCross( statePt, Scalar(255,255,255), 3 );
             drawCross( measPt, Scalar(0,0,255), 3 );
             drawCross( predictPt, Scalar(0,255,0), 3 );
-            line( img, statePt, measPt, Scalar(0,0,255), 3, CV_AA, 0 );
-            line( img, statePt, predictPt, Scalar(0,255,255), 3, CV_AA, 0 );
+            line( img, statePt, measPt, Scalar(0,0,255), 3, LINE_AA, 0 );
+            line( img, statePt, predictPt, Scalar(0,255,255), 3, LINE_AA, 0 );
 
             if(theRNG().uniform(0,4) != 0)
                 KF.correct(measurement);
diff --git a/samples/cpp/kmeans.cpp b/samples/cpp/kmeans.cpp
index 97de6a0a48..0cc313f756 100644
--- a/samples/cpp/kmeans.cpp
+++ b/samples/cpp/kmeans.cpp
@@ -47,13 +47,13 @@ int main( int /*argc*/, char** /*argv*/ )
             Mat pointChunk = points.rowRange(k*sampleCount/clusterCount,
                                              k == clusterCount - 1 ? sampleCount :
                                              (k+1)*sampleCount/clusterCount);
-            rng.fill(pointChunk, CV_RAND_NORMAL, Scalar(center.x, center.y), Scalar(img.cols*0.05, img.rows*0.05));
+            rng.fill(pointChunk, RNG::NORMAL, Scalar(center.x, center.y), Scalar(img.cols*0.05, img.rows*0.05));
         }
 
         randShuffle(points, 1, &rng);
 
         kmeans(points, clusterCount, labels,
-               TermCriteria( CV_TERMCRIT_EPS+CV_TERMCRIT_ITER, 10, 1.0),
+            TermCriteria( TermCriteria::EPS+TermCriteria::COUNT, 10, 1.0),
                3, KMEANS_PP_CENTERS, centers);
 
         img = Scalar::all(0);
@@ -62,7 +62,7 @@ int main( int /*argc*/, char** /*argv*/ )
         {
             int clusterIdx = labels.at<int>(i);
             Point ipt = points.at<Point2f>(i);
-            circle( img, ipt, 2, colorTab[clusterIdx], CV_FILLED, CV_AA );
+            circle( img, ipt, 2, colorTab[clusterIdx], FILLED, LINE_AA );
         }
 
         imshow("clusters", img);
diff --git a/samples/cpp/laplace.cpp b/samples/cpp/laplace.cpp
index 50cb8467ba..45264e3107 100644
--- a/samples/cpp/laplace.cpp
+++ b/samples/cpp/laplace.cpp
@@ -17,8 +17,10 @@ static void help()
             "./laplace [camera #, default 0]\n" << endl;
 }
 
+enum {GAUSSIAN, BLUR, MEDIAN};
+
 int sigma = 3;
-int smoothType = CV_GAUSSIAN;
+int smoothType = GAUSSIAN;
 
 int main( int argc, char** argv )
 {
@@ -32,15 +34,15 @@ int main( int argc, char** argv )
         cap.open(argv[1]);
         if( cap.isOpened() )
             cout << "Video " << argv[1] <<
-                ": width=" << cap.get(CV_CAP_PROP_FRAME_WIDTH) <<
-                ", height=" << cap.get(CV_CAP_PROP_FRAME_HEIGHT) <<
-                ", nframes=" << cap.get(CV_CAP_PROP_FRAME_COUNT) << endl;
+                ": width=" << cap.get(CAP_PROP_FRAME_WIDTH) <<
+                ", height=" << cap.get(CAP_PROP_FRAME_HEIGHT) <<
+                ", nframes=" << cap.get(CAP_PROP_FRAME_COUNT) << endl;
         if( argc > 2 && isdigit(argv[2][0]) )
         {
             int pos;
             sscanf(argv[2], "%d", &pos);
             cout << "seeking to frame #" << pos << endl;
-            cap.set(CV_CAP_PROP_POS_FRAMES, pos);
+            cap.set(CAP_PROP_POS_FRAMES, pos);
         }
     }
 
@@ -63,9 +65,9 @@ int main( int argc, char** argv )
             break;
 
         int ksize = (sigma*5)|1;
-        if(smoothType == CV_GAUSSIAN)
+        if(smoothType == GAUSSIAN)
             GaussianBlur(frame, smoothed, Size(ksize, ksize), sigma, sigma);
-        else if(smoothType == CV_BLUR)
+        else if(smoothType == BLUR)
             blur(frame, smoothed, Size(ksize, ksize));
         else
             medianBlur(frame, smoothed, ksize);
@@ -76,7 +78,7 @@ int main( int argc, char** argv )
 
         int c = waitKey(30);
         if( c == ' ' )
-            smoothType = smoothType == CV_GAUSSIAN ? CV_BLUR : smoothType == CV_BLUR ? CV_MEDIAN : CV_GAUSSIAN;
+            smoothType = smoothType == GAUSSIAN ? BLUR : smoothType == BLUR ? MEDIAN : GAUSSIAN;
         if( c == 'q' || c == 'Q' || (c & 255) == 27 )
             break;
     }
diff --git a/samples/cpp/latentsvm_multidetect.cpp b/samples/cpp/latentsvm_multidetect.cpp
index d2105122ab..f5a8bc56da 100644
--- a/samples/cpp/latentsvm_multidetect.cpp
+++ b/samples/cpp/latentsvm_multidetect.cpp
@@ -3,20 +3,12 @@
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/contrib/contrib.hpp"
 
-#ifdef WIN32
+#if defined(WIN32) || defined(_WIN32)
 #include <io.h>
 #else
 #include <dirent.h>
 #endif
 
-#ifdef HAVE_CVCONFIG_H
-#include <cvconfig.h>
-#endif
-
-#ifdef HAVE_TBB
-#include "tbb/task_scheduler_init.h"
-#endif
-
 using namespace std;
 using namespace cv;
 
@@ -47,7 +39,7 @@ static void detectAndDrawObjects( Mat& image, LatentSvmDetector& detector, const
 
     cout << "Detection time = " << tm.getTimeSec() << " sec" << endl;
 
-    const vector<string> classNames = detector.getClassNames();
+    const vector<String> classNames = detector.getClassNames();
     CV_Assert( colors.size() == classNames.size() );
 
     for( size_t i = 0; i < detections.size(); i++ )
@@ -63,11 +55,11 @@ static void detectAndDrawObjects( Mat& image, LatentSvmDetector& detector, const
     }
 }
 
-static void readDirectory( const string& directoryName, vector<string>& filenames, bool addDirectoryName=true )
+static void readDirectory( const string& directoryName, vector<String>& filenames, bool addDirectoryName=true )
 {
     filenames.clear();
 
-#ifdef WIN32
+#if defined(WIN32) | defined(_WIN32)
     struct _finddata_t s_file;
     string str = directoryName + "\\*.*";
 
@@ -123,7 +115,7 @@ int main(int argc, char* argv[])
         if( argc > 4 ) numThreads = atoi(argv[4]);
     }
 
-    vector<string> images_filenames, models_filenames;
+    vector<String> images_filenames, models_filenames;
     readDirectory( images_folder, images_filenames );
     readDirectory( models_folder, models_filenames );
 
@@ -134,7 +126,7 @@ int main(int argc, char* argv[])
         exit(-1);
     }
 
-    const vector<string>& classNames = detector.getClassNames();
+    const vector<String>& classNames = detector.getClassNames();
     cout << "Loaded " << classNames.size() << " models:" << endl;
     for( size_t i = 0; i < classNames.size(); i++ )
     {
diff --git a/samples/cpp/letter_recog.cpp b/samples/cpp/letter_recog.cpp
index 49a55fc62b..ddbe676291 100644
--- a/samples/cpp/letter_recog.cpp
+++ b/samples/cpp/letter_recog.cpp
@@ -131,7 +131,7 @@ int build_rtrees_classifier( char* data_filename,
             printf( "Could not read the classifier %s\n", filename_to_load );
             return -1;
         }
-        printf( "The classifier %s is loaded.\n", data_filename );
+        printf( "The classifier %s is loaded.\n", filename_to_load );
     }
     else
     {
@@ -262,7 +262,7 @@ int build_boost_classifier( char* data_filename,
             printf( "Could not read the classifier %s\n", filename_to_load );
             return -1;
         }
-        printf( "The classifier %s is loaded.\n", data_filename );
+        printf( "The classifier %s is loaded.\n", filename_to_load );
     }
     else
     {
@@ -403,7 +403,7 @@ int build_mlp_classifier( char* data_filename,
             printf( "Could not read the classifier %s\n", filename_to_load );
             return -1;
         }
-        printf( "The classifier %s is loaded.\n", data_filename );
+        printf( "The classifier %s is loaded.\n", filename_to_load );
     }
     else
     {
@@ -463,7 +463,7 @@ int build_mlp_classifier( char* data_filename,
         int best_class;
         CvMat sample;
         cvGetRow( data, &sample, i );
-        CvPoint max_loc = {0,0};
+        CvPoint max_loc;
         mlp.predict( &sample, mlp_response );
         cvMinMaxLoc( mlp_response, 0, 0, 0, &max_loc, 0 );
         best_class = max_loc.x + 'A';
@@ -639,10 +639,11 @@ int build_nbayes_classifier( char* data_filename )
 }
 
 static
-int build_svm_classifier( char* data_filename )
+int build_svm_classifier( char* data_filename, const char* filename_to_save, const char* filename_to_load )
 {
     CvMat* data = 0;
     CvMat* responses = 0;
+    CvMat* train_resp = 0;
     CvMat train_data;
     int nsamples_all = 0, ntrain_samples = 0;
     int var_count;
@@ -666,13 +667,29 @@ int build_svm_classifier( char* data_filename )
     ntrain_samples = (int)(nsamples_all*0.1);
     var_count = data->cols;
 
-    // train classifier
-    printf( "Training the classifier (may take a few minutes)...\n");
-    cvGetRows( data, &train_data, 0, ntrain_samples );
-    CvMat* train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1);
-    for (int i = 0; i < ntrain_samples; i++)
-        train_resp->data.fl[i] = responses->data.fl[i];
-    svm.train(&train_data, train_resp, 0, 0, param);
+    // Create or load Random Trees classifier
+    if( filename_to_load )
+    {
+        // load classifier from the specified file
+        svm.load( filename_to_load );
+        ntrain_samples = 0;
+        if( svm.get_var_count() == 0 )
+        {
+            printf( "Could not read the classifier %s\n", filename_to_load );
+            return -1;
+        }
+        printf( "The classifier %s is loaded.\n", filename_to_load );
+    }
+    else
+    {
+        // train classifier
+        printf( "Training the classifier (may take a few minutes)...\n");
+        cvGetRows( data, &train_data, 0, ntrain_samples );
+        train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1);
+        for (int i = 0; i < ntrain_samples; i++)
+            train_resp->data.fl[i] = responses->data.fl[i];
+        svm.train(&train_data, train_resp, 0, 0, param);
+    }
 
     // classification
     std::vector<float> _sample(var_count * (nsamples_all - ntrain_samples));
@@ -691,7 +708,10 @@ int build_svm_classifier( char* data_filename )
     CvMat *result = cvCreateMat(1, nsamples_all - ntrain_samples, CV_32FC1);
 
     printf("Classification (may take a few minutes)...\n");
+    double t = (double)cvGetTickCount();
     svm.predict(&sample, result);
+    t = (double)cvGetTickCount() - t;
+    printf("Prediction type: %gms\n", t/(cvGetTickFrequency()*1000.));
 
     int true_resp = 0;
     for (int i = 0; i < nsamples_all - ntrain_samples; i++)
@@ -702,6 +722,9 @@ int build_svm_classifier( char* data_filename )
 
     printf("true_resp = %f%%\n", (float)true_resp / (nsamples_all - ntrain_samples) * 100);
 
+    if( filename_to_save )
+        svm.save( filename_to_save );
+
     cvReleaseMat( &train_resp );
     cvReleaseMat( &result );
     cvReleaseMat( &data );
@@ -772,7 +795,7 @@ int main( int argc, char *argv[] )
         method == 4 ?
         build_nbayes_classifier( data_filename) :
         method == 5 ?
-        build_svm_classifier( data_filename ):
+        build_svm_classifier( data_filename, filename_to_save, filename_to_load ):
         -1) < 0)
     {
         help();
diff --git a/samples/cpp/linemod.cpp b/samples/cpp/linemod.cpp
index 8f0c71a31b..4d11da36b5 100644
--- a/samples/cpp/linemod.cpp
+++ b/samples/cpp/linemod.cpp
@@ -1,8 +1,9 @@
-#include <opencv2/core/core.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
 #include <opencv2/imgproc/imgproc_c.h> // cvFindContours
-#include <opencv2/imgproc/imgproc.hpp>
-#include <opencv2/objdetect/objdetect.hpp>
-#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/objdetect.hpp>
+#include <opencv2/highgui.hpp>
 #include <iterator>
 #include <set>
 #include <cstdio>
@@ -30,7 +31,7 @@ class Mouse
 public:
   static void start(const std::string& a_img_name)
   {
-    cvSetMouseCallback(a_img_name.c_str(), Mouse::cv_on_mouse, 0);
+      cv::setMouseCallback(a_img_name.c_str(), Mouse::cv_on_mouse, 0);
   }
   static int event(void)
   {
@@ -113,7 +114,7 @@ private:
 // Functions to store detector and templates in single XML/YAML file
 static cv::Ptr<cv::linemod::Detector> readLinemod(const std::string& filename)
 {
-  cv::Ptr<cv::linemod::Detector> detector = new cv::linemod::Detector;
+  cv::Ptr<cv::linemod::Detector> detector = cv::makePtr<cv::linemod::Detector>();
   cv::FileStorage fs(filename, cv::FileStorage::READ);
   detector->read(fs.root());
 
@@ -129,7 +130,7 @@ static void writeLinemod(const cv::Ptr<cv::linemod::Detector>& detector, const s
   cv::FileStorage fs(filename, cv::FileStorage::WRITE);
   detector->write(fs);
 
-  std::vector<std::string> ids = detector->classIds();
+  std::vector<cv::String> ids = detector->classIds();
   fs << "classes" << "[";
   for (int i = 0; i < (int)ids.size(); ++i)
   {
@@ -176,7 +177,7 @@ int main(int argc, char * argv[])
   {
     detector = readLinemod(argv[1]);
 
-    std::vector<std::string> ids = detector->classIds();
+    std::vector<cv::String> ids = detector->classIds();
     num_classes = detector->numClasses();
     printf("Loaded %s with %d classes and %d templates\n",
            argv[1], num_classes, detector->numTemplates());
@@ -189,14 +190,14 @@ int main(int argc, char * argv[])
   int num_modalities = (int)detector->getModalities().size();
 
   // Open Kinect sensor
-  cv::VideoCapture capture( CV_CAP_OPENNI );
+  cv::VideoCapture capture( cv::CAP_OPENNI );
   if (!capture.isOpened())
   {
     printf("Could not open OpenNI-capable sensor\n");
     return -1;
   }
-  capture.set(CV_CAP_PROP_OPENNI_REGISTRATION, 1);
-  double focal_length = capture.get(CV_CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH);
+  capture.set(cv::CAP_PROP_OPENNI_REGISTRATION, 1);
+  double focal_length = capture.get(cv::CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH);
   //printf("Focal length = %f\n", focal_length);
 
   // Main loop
@@ -205,8 +206,8 @@ int main(int argc, char * argv[])
   {
     // Capture next color/depth pair
     capture.grab();
-    capture.retrieve(depth, CV_CAP_OPENNI_DEPTH_MAP);
-    capture.retrieve(color, CV_CAP_OPENNI_BGR_IMAGE);
+    capture.retrieve(depth, cv::CAP_OPENNI_DEPTH_MAP);
+    capture.retrieve(color, cv::CAP_OPENNI_BGR_IMAGE);
 
     std::vector<cv::Mat> sources;
     sources.push_back(color);
@@ -223,7 +224,7 @@ int main(int argc, char * argv[])
       cv::Point pt1 = mouse - roi_offset; // top left
       cv::Point pt2 = mouse + roi_offset; // bottom right
 
-      if (event == CV_EVENT_RBUTTONDOWN)
+      if (event == cv::EVENT_RBUTTONDOWN)
       {
         // Compute object mask by subtracting the plane within the ROI
         std::vector<CvPoint> chain(4);
@@ -259,7 +260,7 @@ int main(int argc, char * argv[])
 
     // Perform matching
     std::vector<cv::linemod::Match> matches;
-    std::vector<std::string> class_ids;
+    std::vector<cv::String> class_ids;
     std::vector<cv::Mat> quantized_images;
     match_timer.start();
     detector->match(sources, (float)matching_threshold, matches, class_ids, quantized_images);
@@ -330,7 +331,7 @@ int main(int argc, char * argv[])
     cv::imshow("normals", quantized_images[1]);
 
     cv::FileStorage fs;
-    char key = (char)cvWaitKey(10);
+    char key = (char)cv::waitKey(10);
     if( key == 'q' )
         break;
 
diff --git a/samples/cpp/lkdemo.cpp b/samples/cpp/lkdemo.cpp
index 9ea395c8a3..c665cfdfe2 100644
--- a/samples/cpp/lkdemo.cpp
+++ b/samples/cpp/lkdemo.cpp
@@ -12,9 +12,8 @@ static void help()
 {
     // print a welcome message, and the OpenCV version
     cout << "\nThis is a demo of Lukas-Kanade optical flow lkdemo(),\n"
-            "Using OpenCV version %s\n" << CV_VERSION << "\n"
-            << endl;
-
+            "Using OpenCV version " << CV_VERSION << endl;
+    cout << "\nIt uses camera by default, but you can provide a path to video as an argument.\n";
     cout << "\nHot keys: \n"
             "\tESC - quit the program\n"
             "\tr - auto-initialize tracking\n"
@@ -28,17 +27,19 @@ bool addRemovePt = false;
 
 static void onMouse( int event, int x, int y, int /*flags*/, void* /*param*/ )
 {
-    if( event == CV_EVENT_LBUTTONDOWN )
+    if( event == EVENT_LBUTTONDOWN )
     {
-        point = Point2f((float)x,(float)y);
+        point = Point2f((float)x, (float)y);
         addRemovePt = true;
     }
 }
 
 int main( int argc, char** argv )
 {
+    help();
+
     VideoCapture cap;
-    TermCriteria termcrit(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS,20,0.03);
+    TermCriteria termcrit(TermCriteria::COUNT|TermCriteria::EPS,20,0.03);
     Size subPixWinSize(10,10), winSize(31,31);
 
     const int MAX_COUNT = 500;
@@ -56,8 +57,6 @@ int main( int argc, char** argv )
         return 0;
     }
 
-    help();
-
     namedWindow( "LK Demo", 1 );
     setMouseCallback( "LK Demo", onMouse, 0 );
 
@@ -72,7 +71,7 @@ int main( int argc, char** argv )
             break;
 
         frame.copyTo(image);
-        cvtColor(image, gray, CV_BGR2GRAY);
+        cvtColor(image, gray, COLOR_BGR2GRAY);
 
         if( nightMode )
             image = Scalar::all(0);
@@ -117,7 +116,7 @@ int main( int argc, char** argv )
         {
             vector<Point2f> tmp;
             tmp.push_back(point);
-            cornerSubPix( gray, tmp, winSize, cvSize(-1,-1), termcrit);
+            cornerSubPix( gray, tmp, winSize, Size(-1,-1), termcrit);
             points[1].push_back(tmp[0]);
             addRemovePt = false;
         }
@@ -134,17 +133,16 @@ int main( int argc, char** argv )
             needToInit = true;
             break;
         case 'c':
+            points[0].clear();
             points[1].clear();
             break;
         case 'n':
             nightMode = !nightMode;
             break;
-        default:
-            ;
         }
 
         std::swap(points[1], points[0]);
-        swap(prevGray, gray);
+        cv::swap(prevGray, gray);
     }
 
     return 0;
diff --git a/samples/cpp/logpolar_bsm.cpp b/samples/cpp/logpolar_bsm.cpp
index bfe99ce29e..fc4a4b34b7 100644
--- a/samples/cpp/logpolar_bsm.cpp
+++ b/samples/cpp/logpolar_bsm.cpp
@@ -6,7 +6,9 @@
 * PSPC-lab - University of Genoa
 */
 
-#include "opencv2/opencv.hpp"
+#include "opencv2/contrib.hpp"
+#include "opencv2/highgui.hpp"
+
 #include <iostream>
 #include <cmath>
 
diff --git a/samples/cpp/lsd_lines.cpp b/samples/cpp/lsd_lines.cpp
new file mode 100644
index 0000000000..62692f70bd
--- /dev/null
+++ b/samples/cpp/lsd_lines.cpp
@@ -0,0 +1,54 @@
+#include <iostream>
+#include <string>
+
+#include "opencv2/core/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/highgui/highgui.hpp"
+
+using namespace std;
+using namespace cv;
+
+int main(int argc, char** argv)
+{
+    std::string in;
+    if (argc != 2)
+    {
+        std::cout << "Usage: lsd_lines [input image]. Now loading building.jpg" << std::endl;
+        in = "building.jpg";
+    }
+    else
+    {
+        in = argv[1];
+    }
+
+    Mat image = imread(in, IMREAD_GRAYSCALE);
+
+#if 0
+    Canny(image, image, 50, 200, 3); // Apply canny edge
+#endif
+
+    // Create and LSD detector with standard or no refinement.
+#if 1
+    Ptr<LineSegmentDetector> ls = createLineSegmentDetectorPtr(LSD_REFINE_STD);
+#else
+    Ptr<LineSegmentDetector> ls = createLineSegmentDetectorPtr(LSD_REFINE_NONE);
+#endif
+
+    double start = double(getTickCount());
+    vector<Vec4i> lines_std;
+
+    // Detect the lines
+    ls->detect(image, lines_std);
+
+    double duration_ms = (double(getTickCount()) - start) * 1000 / getTickFrequency();
+    std::cout << "It took " << duration_ms << " ms." << std::endl;
+
+    // Show found lines
+    Mat drawnLines(image);
+    ls->drawSegments(drawnLines, lines_std);
+    imshow("Standard refinement", drawnLines);
+
+    waitKey();
+    return 0;
+}
diff --git a/samples/cpp/matcher_simple.cpp b/samples/cpp/matcher_simple.cpp
index acdf55f700..c1cd6e4c30 100644
--- a/samples/cpp/matcher_simple.cpp
+++ b/samples/cpp/matcher_simple.cpp
@@ -23,8 +23,8 @@ int main(int argc, char** argv)
         return -1;
     }
 
-    Mat img1 = imread(argv[1], CV_LOAD_IMAGE_GRAYSCALE);
-    Mat img2 = imread(argv[2], CV_LOAD_IMAGE_GRAYSCALE);
+    Mat img1 = imread(argv[1], IMREAD_GRAYSCALE);
+    Mat img2 = imread(argv[2], IMREAD_GRAYSCALE);
     if(img1.empty() || img2.empty())
     {
         printf("Can't read one of the images\n");
diff --git a/samples/cpp/matching_to_many_images.cpp b/samples/cpp/matching_to_many_images.cpp
index f6c31b2c48..152b400645 100644
--- a/samples/cpp/matching_to_many_images.cpp
+++ b/samples/cpp/matching_to_many_images.cpp
@@ -84,7 +84,7 @@ static bool createDetectorDescriptorMatcher( const string& detectorType, const s
     descriptorMatcher = DescriptorMatcher::create( matcherType );
     cout << ">" << endl;
 
-    bool isCreated = !( featureDetector.empty() || descriptorExtractor.empty() || descriptorMatcher.empty() );
+    bool isCreated = featureDetector && descriptorExtractor && descriptorMatcher;
     if( !isCreated )
         cout << "Can not create feature detector or descriptor extractor or descriptor matcher of given types." << endl << ">" << endl;
 
@@ -95,7 +95,7 @@ static bool readImages( const string& queryImageName, const string& trainFilenam
                  Mat& queryImage, vector <Mat>& trainImages, vector<string>& trainImageNames )
 {
     cout << "< Reading the images..." << endl;
-    queryImage = imread( queryImageName, CV_LOAD_IMAGE_GRAYSCALE);
+    queryImage = imread( queryImageName, IMREAD_GRAYSCALE);
     if( queryImage.empty() )
     {
         cout << "Query image can not be read." << endl << ">" << endl;
@@ -112,7 +112,7 @@ static bool readImages( const string& queryImageName, const string& trainFilenam
     for( size_t i = 0; i < trainImageNames.size(); i++ )
     {
         string filename = trainDirName + trainImageNames[i];
-        Mat img = imread( filename, CV_LOAD_IMAGE_GRAYSCALE );
+        Mat img = imread( filename, IMREAD_GRAYSCALE );
         if( img.empty() )
             cout << "Train image " << filename << " can not be read." << endl;
         else
diff --git a/samples/cpp/meanshift_segmentation.cpp b/samples/cpp/meanshift_segmentation.cpp
index ce34f6ea7f..eb71619dad 100644
--- a/samples/cpp/meanshift_segmentation.cpp
+++ b/samples/cpp/meanshift_segmentation.cpp
@@ -65,7 +65,7 @@ int main(int argc, char** argv)
     colorRad = 10;
     maxPyrLevel = 1;
 
-    namedWindow( winName, CV_WINDOW_AUTOSIZE );
+    namedWindow( winName, WINDOW_AUTOSIZE );
 
     createTrackbar( "spatialRad", winName, &spatialRad, 80, meanShiftSegmentation );
     createTrackbar( "colorRad", winName, &colorRad, 60, meanShiftSegmentation );
diff --git a/samples/cpp/minarea.cpp b/samples/cpp/minarea.cpp
index eca5dc1d36..13ac7c6d52 100644
--- a/samples/cpp/minarea.cpp
+++ b/samples/cpp/minarea.cpp
@@ -45,16 +45,16 @@ int main( int /*argc*/, char** /*argv*/ )
 
         img = Scalar::all(0);
         for( i = 0; i < count; i++ )
-            circle( img, points[i], 3, Scalar(0, 0, 255), CV_FILLED, CV_AA );
+            circle( img, points[i], 3, Scalar(0, 0, 255), FILLED, LINE_AA );
 
         for( i = 0; i < 4; i++ )
-            line(img, vtx[i], vtx[(i+1)%4], Scalar(0, 255, 0), 1, CV_AA);
+            line(img, vtx[i], vtx[(i+1)%4], Scalar(0, 255, 0), 1, LINE_AA);
 
-        circle(img, center, cvRound(radius), Scalar(0, 255, 255), 1, CV_AA);
+        circle(img, center, cvRound(radius), Scalar(0, 255, 255), 1, LINE_AA);
 
         imshow( "rect & circle", img );
 
-        char key = (char)cvWaitKey();
+        char key = (char)waitKey();
         if( key == 27 || key == 'q' || key == 'Q' ) // 'ESC'
             break;
     }
diff --git a/samples/cpp/morphology2.cpp b/samples/cpp/morphology2.cpp
index d4ab351545..f20b2c7815 100644
--- a/samples/cpp/morphology2.cpp
+++ b/samples/cpp/morphology2.cpp
@@ -1,5 +1,3 @@
-#define CV_NO_BACKWARD_COMPATIBILITY
-
 #include "opencv2/imgproc/imgproc.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include <stdlib.h>
@@ -37,9 +35,9 @@ static void OpenClose(int, void*)
     int an = n > 0 ? n : -n;
     Mat element = getStructuringElement(element_shape, Size(an*2+1, an*2+1), Point(an, an) );
     if( n < 0 )
-        morphologyEx(src, dst, CV_MOP_OPEN, element);
+        morphologyEx(src, dst, MORPH_OPEN, element);
     else
-        morphologyEx(src, dst, CV_MOP_CLOSE, element);
+        morphologyEx(src, dst, MORPH_CLOSE, element);
     imshow("Open/Close",dst);
 }
 
@@ -79,7 +77,7 @@ int main( int argc, char** argv )
 
         OpenClose(open_close_pos, 0);
         ErodeDilate(erode_dilate_pos, 0);
-        c = cvWaitKey(0);
+        c = waitKey();
 
         if( (char)c == 27 )
             break;
diff --git a/samples/cpp/opencv_version.cpp b/samples/cpp/opencv_version.cpp
index b735f2a6cc..ec9b47d58e 100644
--- a/samples/cpp/opencv_version.cpp
+++ b/samples/cpp/opencv_version.cpp
@@ -1,4 +1,4 @@
-#include "opencv2/core/core.hpp"
+#include <opencv2/core/utility.hpp>
 #include <iostream>
 
 const char* keys =
diff --git a/samples/cpp/openni_capture.cpp b/samples/cpp/openni_capture.cpp
index 549f16e6f0..802b474207 100644
--- a/samples/cpp/openni_capture.cpp
+++ b/samples/cpp/openni_capture.cpp
@@ -79,8 +79,8 @@ static void colorizeDisparity( const Mat& gray, Mat& rgb, double maxDisp=-1.f, f
 static float getMaxDisparity( VideoCapture& capture )
 {
     const int minDistance = 400; // mm
-    float b = (float)capture.get( CV_CAP_OPENNI_DEPTH_GENERATOR_BASELINE ); // mm
-    float F = (float)capture.get( CV_CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH ); // pixels
+    float b = (float)capture.get( CAP_OPENNI_DEPTH_GENERATOR_BASELINE ); // mm
+    float F = (float)capture.get( CAP_OPENNI_DEPTH_GENERATOR_FOCAL_LENGTH ); // pixels
     return b * F / minDistance;
 }
 
@@ -142,7 +142,7 @@ static void parseCommandLine( int argc, char* argv[], bool& isColorizeDisp, bool
             {
                 string mask( argv[++i] );
                 if( mask.size() != 5)
-                    CV_Error( CV_StsBadArg, "Incorrect length of -m argument string" );
+                    CV_Error( Error::StsBadArg, "Incorrect length of -m argument string" );
                 int val = atoi(mask.c_str());
 
                 int l = 100000, r = 10000, sum = 0;
@@ -191,7 +191,7 @@ int main( int argc, char* argv[] )
     if( isVideoReading )
         capture.open( filename );
     else
-        capture.open( CV_CAP_OPENNI );
+        capture.open( CAP_OPENNI );
 
     cout << "done." << endl;
 
@@ -207,23 +207,23 @@ int main( int argc, char* argv[] )
         switch ( imageMode )
         {
             case 0:
-                modeRes = capture.set( CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CV_CAP_OPENNI_VGA_30HZ );
+                modeRes = capture.set( CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CAP_OPENNI_VGA_30HZ );
                 break;
             case 1:
-                modeRes = capture.set( CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CV_CAP_OPENNI_SXGA_15HZ );
+                modeRes = capture.set( CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CAP_OPENNI_SXGA_15HZ );
                 break;
             case 2:
-                modeRes = capture.set( CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CV_CAP_OPENNI_SXGA_30HZ );
+                modeRes = capture.set( CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CAP_OPENNI_SXGA_30HZ );
                 break;
                 //The following modes are only supported by the Xtion Pro Live
             case 3:
-                modeRes = capture.set( CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CV_CAP_OPENNI_QVGA_30HZ );
+                modeRes = capture.set( CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CAP_OPENNI_QVGA_30HZ );
                 break;
             case 4:
-                modeRes = capture.set( CV_CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CV_CAP_OPENNI_QVGA_60HZ );
+                modeRes = capture.set( CAP_OPENNI_IMAGE_GENERATOR_OUTPUT_MODE, CAP_OPENNI_QVGA_60HZ );
                 break;
             default:
-                CV_Error( CV_StsBadArg, "Unsupported image mode property.\n");
+                CV_Error( Error::StsBadArg, "Unsupported image mode property.\n");
         }
         if (!modeRes)
             cout << "\nThis image mode is not supported by the device, the default value (CV_CAP_OPENNI_SXGA_15HZ) will be used.\n" << endl;
@@ -231,18 +231,18 @@ int main( int argc, char* argv[] )
 
     // Print some avalible device settings.
     cout << "\nDepth generator output mode:" << endl <<
-            "FRAME_WIDTH      " << capture.get( CV_CAP_PROP_FRAME_WIDTH ) << endl <<
-            "FRAME_HEIGHT     " << capture.get( CV_CAP_PROP_FRAME_HEIGHT ) << endl <<
-            "FRAME_MAX_DEPTH  " << capture.get( CV_CAP_PROP_OPENNI_FRAME_MAX_DEPTH ) << " mm" << endl <<
-            "FPS              " << capture.get( CV_CAP_PROP_FPS ) << endl <<
-            "REGISTRATION     " << capture.get( CV_CAP_PROP_OPENNI_REGISTRATION ) << endl;
-    if( capture.get( CV_CAP_OPENNI_IMAGE_GENERATOR_PRESENT ) )
+            "FRAME_WIDTH      " << capture.get( CAP_PROP_FRAME_WIDTH ) << endl <<
+            "FRAME_HEIGHT     " << capture.get( CAP_PROP_FRAME_HEIGHT ) << endl <<
+            "FRAME_MAX_DEPTH  " << capture.get( CAP_PROP_OPENNI_FRAME_MAX_DEPTH ) << " mm" << endl <<
+            "FPS              " << capture.get( CAP_PROP_FPS ) << endl <<
+            "REGISTRATION     " << capture.get( CAP_PROP_OPENNI_REGISTRATION ) << endl;
+    if( capture.get( CAP_OPENNI_IMAGE_GENERATOR_PRESENT ) )
     {
         cout <<
             "\nImage generator output mode:" << endl <<
-            "FRAME_WIDTH   " << capture.get( CV_CAP_OPENNI_IMAGE_GENERATOR+CV_CAP_PROP_FRAME_WIDTH ) << endl <<
-            "FRAME_HEIGHT  " << capture.get( CV_CAP_OPENNI_IMAGE_GENERATOR+CV_CAP_PROP_FRAME_HEIGHT ) << endl <<
-            "FPS           " << capture.get( CV_CAP_OPENNI_IMAGE_GENERATOR+CV_CAP_PROP_FPS ) << endl;
+            "FRAME_WIDTH   " << capture.get( CAP_OPENNI_IMAGE_GENERATOR+CAP_PROP_FRAME_WIDTH ) << endl <<
+            "FRAME_HEIGHT  " << capture.get( CAP_OPENNI_IMAGE_GENERATOR+CAP_PROP_FRAME_HEIGHT ) << endl <<
+            "FPS           " << capture.get( CAP_OPENNI_IMAGE_GENERATOR+CAP_PROP_FPS ) << endl;
     }
     else
     {
@@ -266,14 +266,14 @@ int main( int argc, char* argv[] )
         }
         else
         {
-            if( retrievedImageFlags[0] && capture.retrieve( depthMap, CV_CAP_OPENNI_DEPTH_MAP ) )
+            if( retrievedImageFlags[0] && capture.retrieve( depthMap, CAP_OPENNI_DEPTH_MAP ) )
             {
                 const float scaleFactor = 0.05f;
                 Mat show; depthMap.convertTo( show, CV_8UC1, scaleFactor );
                 imshow( "depth map", show );
             }
 
-            if( retrievedImageFlags[1] && capture.retrieve( disparityMap, CV_CAP_OPENNI_DISPARITY_MAP ) )
+            if( retrievedImageFlags[1] && capture.retrieve( disparityMap, CAP_OPENNI_DISPARITY_MAP ) )
             {
                 if( isColorizeDisp )
                 {
@@ -289,13 +289,13 @@ int main( int argc, char* argv[] )
                 }
             }
 
-            if( retrievedImageFlags[2] && capture.retrieve( validDepthMap, CV_CAP_OPENNI_VALID_DEPTH_MASK ) )
+            if( retrievedImageFlags[2] && capture.retrieve( validDepthMap, CAP_OPENNI_VALID_DEPTH_MASK ) )
                 imshow( "valid depth mask", validDepthMap );
 
-            if( retrievedImageFlags[3] && capture.retrieve( bgrImage, CV_CAP_OPENNI_BGR_IMAGE ) )
+            if( retrievedImageFlags[3] && capture.retrieve( bgrImage, CAP_OPENNI_BGR_IMAGE ) )
                 imshow( "rgb image", bgrImage );
 
-            if( retrievedImageFlags[4] && capture.retrieve( grayImage, CV_CAP_OPENNI_GRAY_IMAGE ) )
+            if( retrievedImageFlags[4] && capture.retrieve( grayImage, CAP_OPENNI_GRAY_IMAGE ) )
                 imshow( "gray image", grayImage );
         }
 
diff --git a/samples/cpp/pca.cpp b/samples/cpp/pca.cpp
index ed23c7622b..d4272736c2 100644
--- a/samples/cpp/pca.cpp
+++ b/samples/cpp/pca.cpp
@@ -54,7 +54,7 @@ static void read_imgList(const string& filename, vector<Mat>& images) {
     std::ifstream file(filename.c_str(), ifstream::in);
     if (!file) {
         string error_message = "No valid input file was given, please check the given filename.";
-        CV_Error(CV_StsBadArg, error_message);
+        CV_Error(Error::StsBadArg, error_message);
     }
     string line;
     while (getline(file, line)) {
@@ -78,7 +78,7 @@ static Mat toGrayscale(InputArray _src) {
     Mat src = _src.getMat();
     // only allow one channel
     if(src.channels() != 1) {
-        CV_Error(CV_StsBadArg, "Only Matrices with one channel are supported");
+        CV_Error(Error::StsBadArg, "Only Matrices with one channel are supported");
     }
     // create and return normalized image
     Mat dst;
@@ -104,7 +104,7 @@ static void onTrackbar(int pos, void* ptr)
 
     struct params *p = (struct params *)ptr;
 
-    p->pca = PCA(p->data, cv::Mat(), CV_PCA_DATA_AS_ROW, var);
+    p->pca = PCA(p->data, cv::Mat(), PCA::DATA_AS_ROW, var);
 
     Mat point = p->pca.project(p->data.row(0));
     Mat reconstruction = p->pca.backProject(point);
@@ -142,14 +142,14 @@ int main(int argc, char** argv)
     // Quit if there are not enough images for this demo.
     if(images.size() <= 1) {
         string error_message = "This demo needs at least 2 images to work. Please add more images to your data set!";
-        CV_Error(CV_StsError, error_message);
+        CV_Error(Error::StsError, error_message);
     }
 
     // Reshape and stack images into a rowMatrix
     Mat data = formatImagesForPCA(images);
 
     // perform PCA
-    PCA pca(data, cv::Mat(), CV_PCA_DATA_AS_ROW, 0.95); // trackbar is initially set here, also this is a common value for retainedVariance
+    PCA pca(data, cv::Mat(), PCA::DATA_AS_ROW, 0.95); // trackbar is initially set here, also this is a common value for retainedVariance
 
     // Demonstration of the effect of retainedVariance on the first image
     Mat point = pca.project(data.row(0)); // project into the eigenspace, thus the image becomes a "point"
@@ -159,7 +159,7 @@ int main(int argc, char** argv)
 
     // init highgui window
     string winName = "Reconstruction | press 'q' to quit";
-    namedWindow(winName, CV_WINDOW_NORMAL);
+    namedWindow(winName, WINDOW_NORMAL);
 
     // params struct to pass to the trackbar handler
     params p;
diff --git a/samples/cpp/peopledetect.cpp b/samples/cpp/peopledetect.cpp
index fee17897be..c96f3e37d5 100644
--- a/samples/cpp/peopledetect.cpp
+++ b/samples/cpp/peopledetect.cpp
@@ -1,3 +1,4 @@
+#include <opencv2/core/utility.hpp>
 #include <opencv2/imgproc.hpp>
 #include <opencv2/objdetect.hpp>
 #include <opencv2/highgui.hpp>
@@ -83,13 +84,13 @@ int main(int argc, char** argv)
     }
 
     std::string src = parser.get<std::string>("frame");
-    std::vector<std::string> frames;
+    std::vector<cv::String> frames;
     cv::glob(parser.get<std::string>("frame"), frames);
     std::cout << "collected " << src << " " << frames.size() << " frames." << std::endl;
 
     for (int i = 0; i < (int)frames.size(); ++i)
     {
-        std::string& frame_sourse = frames[i];
+        std::string frame_sourse = frames[i];
         cv::Mat frame = cv::imread(frame_sourse);
 
         if(frame.empty())
@@ -140,7 +141,7 @@ int main(int argc, char** argv)
                     conf << d.confidence;
 
                     cv::rectangle(frame, cv::Rect((int)d.x, (int)d.y, (int)d.w, (int)d.h), cv::Scalar(b, 0, 255 - b, 255), 2);
-                    cv::putText(frame, conf.str() , cv::Point((int)d.x + 10, (int)d.y - 5),1, 1.1, cv::Scalar(25, 133, 255, 0), 1, CV_AA);
+                    cv::putText(frame, conf.str() , cv::Point((int)d.x + 10, (int)d.y - 5),1, 1.1, cv::Scalar(25, 133, 255, 0), 1, cv::LINE_AA);
 
                     if (wf)
                         myfile << d.x << "," <<  d.y << "," << d.w << "," <<  d.h << "," << d.confidence << "\n";
diff --git a/samples/cpp/phase_corr.cpp b/samples/cpp/phase_corr.cpp
index d9a1419a7c..cfee809412 100644
--- a/samples/cpp/phase_corr.cpp
+++ b/samples/cpp/phase_corr.cpp
@@ -13,7 +13,7 @@ int main(int, char* [])
     do
     {
         video >> frame;
-        cvtColor(frame, curr, CV_RGB2GRAY);
+        cvtColor(frame, curr, COLOR_RGB2GRAY);
 
         if(prev.empty())
         {
@@ -31,8 +31,8 @@ int main(int, char* [])
         {
             // draw a circle and line indicating the shift direction...
             Point center(curr.cols >> 1, curr.rows >> 1);
-            circle(frame, center, (int)radius, Scalar(0, 255, 0), 3, CV_AA);
-            line(frame, center, Point(center.x + (int)shift.x, center.y + (int)shift.y), Scalar(0, 255, 0), 3, CV_AA);
+            circle(frame, center, (int)radius, Scalar(0, 255, 0), 3, LINE_AA);
+            line(frame, center, Point(center.x + (int)shift.x, center.y + (int)shift.y), Scalar(0, 255, 0), 3, LINE_AA);
         }
 
         imshow("phase shift", frame);
@@ -43,7 +43,3 @@ int main(int, char* [])
 
     return 0;
 }
-
-
-
-
diff --git a/samples/cpp/points_classifier.cpp b/samples/cpp/points_classifier.cpp
index 8bc35e5541..b197b5be59 100644
--- a/samples/cpp/points_classifier.cpp
+++ b/samples/cpp/points_classifier.cpp
@@ -36,7 +36,7 @@ static void on_mouse( int event, int x, int y, int /*flags*/, void* )
 
     int updateFlag = 0;
 
-    if( event == CV_EVENT_LBUTTONUP )
+    if( event == EVENT_LBUTTONUP )
     {
         if( classColors.empty() )
             return;
@@ -45,7 +45,7 @@ static void on_mouse( int event, int x, int y, int /*flags*/, void* )
         trainedPointsMarkers.push_back( (int)(classColors.size()-1) );
         updateFlag = true;
     }
-    else if( event == CV_EVENT_RBUTTONUP )
+    else if( event == EVENT_RBUTTONUP )
     {
 #if _BT_
         if( classColors.size() < 2 )
@@ -503,7 +503,7 @@ int main()
     imgDst.create( 480, 640, CV_8UC3 );
 
     imshow( "points", img );
-    cvSetMouseCallback( "points", on_mouse );
+    setMouseCallback( "points", on_mouse );
 
     for(;;)
     {
diff --git a/samples/cpp/retinaDemo.cpp b/samples/cpp/retinaDemo.cpp
index 7b8d81a49c..790e753837 100644
--- a/samples/cpp/retinaDemo.cpp
+++ b/samples/cpp/retinaDemo.cpp
@@ -9,7 +9,8 @@
 #include <iostream>
 #include <cstring>
 
-#include "opencv2/opencv.hpp"
+#include "opencv2/bioinspired.hpp"
+#include "opencv2/highgui.hpp"
 
 static void help(std::string errorMessage)
 {
@@ -105,15 +106,15 @@ int main(int argc, char* argv[]) {
     try
     {
         // create a retina instance with default parameters setup, uncomment the initialisation you wanna test
-        cv::Ptr<cv::Retina> myRetina;
+        cv::Ptr<cv::bioinspired::Retina> myRetina;
 
         // if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
         if (useLogSampling)
                 {
-                        myRetina = new cv::Retina(inputFrame.size(), true, cv::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+                        myRetina = cv::bioinspired::createRetina(inputFrame.size(), true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
                 }
         else// -> else allocate "classical" retina :
-            myRetina = new cv::Retina(inputFrame.size());
+            myRetina = cv::bioinspired::createRetina(inputFrame.size());
 
         // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
         myRetina->write("RetinaDefaultParameters.xml");
@@ -142,7 +143,8 @@ int main(int argc, char* argv[]) {
             cv::imshow("retina input", inputFrame);
             cv::imshow("Retina Parvo", retinaOutput_parvo);
             cv::imshow("Retina Magno", retinaOutput_magno);
-            cv::waitKey(10);
+
+            cv::waitKey(5);
         }
     }catch(cv::Exception e)
     {
@@ -154,4 +156,3 @@ int main(int argc, char* argv[]) {
 
     return 0;
 }
-
diff --git a/samples/cpp/rgbdodometry.cpp b/samples/cpp/rgbdodometry.cpp
index 660eb31d02..0db2fab847 100644
--- a/samples/cpp/rgbdodometry.cpp
+++ b/samples/cpp/rgbdodometry.cpp
@@ -125,8 +125,8 @@ int main(int argc, char** argv)
     }
 
     Mat grayImage0, grayImage1, depthFlt0, depthFlt1/*in meters*/;
-    cvtColor( colorImage0, grayImage0, CV_BGR2GRAY );
-    cvtColor( colorImage1, grayImage1, CV_BGR2GRAY );
+    cvtColor( colorImage0, grayImage0, COLOR_BGR2GRAY );
+    cvtColor( colorImage1, grayImage1, COLOR_BGR2GRAY );
     depth0.convertTo( depthFlt0, CV_32FC1, 1./1000 );
     depth1.convertTo( depthFlt1, CV_32FC1, 1./1000 );
 
diff --git a/samples/cpp/scenetext.jpg b/samples/cpp/scenetext.jpg
new file mode 100644
index 0000000000..6a9c6bf5e2
Binary files /dev/null and b/samples/cpp/scenetext.jpg differ
diff --git a/samples/cpp/scenetext_GT.png b/samples/cpp/scenetext_GT.png
new file mode 100644
index 0000000000..a5c1905572
Binary files /dev/null and b/samples/cpp/scenetext_GT.png differ
diff --git a/samples/cpp/segment_objects.cpp b/samples/cpp/segment_objects.cpp
index 891b63cffb..d44b035342 100644
--- a/samples/cpp/segment_objects.cpp
+++ b/samples/cpp/segment_objects.cpp
@@ -30,7 +30,7 @@ static void refineSegments(const Mat& img, Mat& mask, Mat& dst)
     erode(temp, temp, Mat(), Point(-1,-1), niters*2);
     dilate(temp, temp, Mat(), Point(-1,-1), niters);
 
-    findContours( temp, contours, hierarchy, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
+    findContours( temp, contours, hierarchy, RETR_CCOMP, CHAIN_APPROX_SIMPLE );
 
     dst = Mat::zeros(img.size(), CV_8UC3);
 
@@ -53,7 +53,7 @@ static void refineSegments(const Mat& img, Mat& mask, Mat& dst)
         }
     }
     Scalar color( 0, 0, 255 );
-    drawContours( dst, contours, largestComp, color, CV_FILLED, 8, hierarchy );
+    drawContours( dst, contours, largestComp, color, FILLED, LINE_8, hierarchy );
 }
 
 
@@ -87,15 +87,15 @@ int main(int argc, char** argv)
     namedWindow("video", 1);
     namedWindow("segmented", 1);
 
-    BackgroundSubtractorMOG bgsubtractor;
-    bgsubtractor.set("noiseSigma", 10);
+    Ptr<BackgroundSubtractorMOG> bgsubtractor=createBackgroundSubtractorMOG();
+    bgsubtractor->setNoiseSigma(10);
 
     for(;;)
     {
         cap >> tmp_frame;
         if( !tmp_frame.data )
             break;
-        bgsubtractor(tmp_frame, bgmask, update_bg_model ? -1 : 0);
+        bgsubtractor->apply(tmp_frame, bgmask, update_bg_model ? -1 : 0);
         //CvMat _bgmask = bgmask;
         //cvSegmentFGMask(&_bgmask);
         refineSegments(tmp_frame, bgmask, out_frame);
diff --git a/samples/cpp/select3dobj.cpp b/samples/cpp/select3dobj.cpp
index a21ee2d291..32d30552f5 100644
--- a/samples/cpp/select3dobj.cpp
+++ b/samples/cpp/select3dobj.cpp
@@ -8,10 +8,11 @@
  *
  */
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <ctype.h>
 #include <stdio.h>
@@ -147,25 +148,25 @@ static Rect extract3DBox(const Mat& frame, Mat& shownFrame, Mat& selectedObjFram
     if( shownFrame.data )
     {
         if( nobjpt == 1 )
-            circle(shownFrame, imgpt[0], 3, Scalar(0,255,0), -1, CV_AA);
+            circle(shownFrame, imgpt[0], 3, Scalar(0,255,0), -1, LINE_AA);
         else if( nobjpt == 2 )
         {
-            circle(shownFrame, imgpt[0], 3, Scalar(0,255,0), -1, CV_AA);
-            circle(shownFrame, imgpt[1], 3, Scalar(0,255,0), -1, CV_AA);
-            line(shownFrame, imgpt[0], imgpt[1], Scalar(0,255,0), 3, CV_AA);
+            circle(shownFrame, imgpt[0], 3, Scalar(0,255,0), -1, LINE_AA);
+            circle(shownFrame, imgpt[1], 3, Scalar(0,255,0), -1, LINE_AA);
+            line(shownFrame, imgpt[0], imgpt[1], Scalar(0,255,0), 3, LINE_AA);
         }
         else if( nobjpt == 3 )
             for( int i = 0; i < 4; i++ )
             {
-                circle(shownFrame, imgpt[i], 3, Scalar(0,255,0), -1, CV_AA);
-                line(shownFrame, imgpt[i], imgpt[(i+1)%4], Scalar(0,255,0), 3, CV_AA);
+                circle(shownFrame, imgpt[i], 3, Scalar(0,255,0), -1, LINE_AA);
+                line(shownFrame, imgpt[i], imgpt[(i+1)%4], Scalar(0,255,0), 3, LINE_AA);
             }
         else
             for( int i = 0; i < 8; i++ )
             {
-                circle(shownFrame, imgpt[i], 3, Scalar(0,255,0), -1, CV_AA);
-                line(shownFrame, imgpt[i], imgpt[(i+1)%4 + (i/4)*4], Scalar(0,255,0), 3, CV_AA);
-                line(shownFrame, imgpt[i], imgpt[i%4], Scalar(0,255,0), 3, CV_AA);
+                circle(shownFrame, imgpt[i], 3, Scalar(0,255,0), -1, LINE_AA);
+                line(shownFrame, imgpt[i], imgpt[(i+1)%4 + (i/4)*4], Scalar(0,255,0), 3, LINE_AA);
+                line(shownFrame, imgpt[i], imgpt[i%4], Scalar(0,255,0), 3, LINE_AA);
             }
     }
 
@@ -211,11 +212,11 @@ static int select3DBox(const string& windowname, const string& selWinName, const
     for(;;)
     {
         float Z = 0.f;
-        bool dragging = (mouse.buttonState & CV_EVENT_FLAG_LBUTTON) != 0;
+        bool dragging = (mouse.buttonState & EVENT_FLAG_LBUTTON) != 0;
         int npt = nobjpt;
 
-        if( (mouse.event == CV_EVENT_LBUTTONDOWN ||
-             mouse.event == CV_EVENT_LBUTTONUP ||
+        if( (mouse.event == EVENT_LBUTTONDOWN ||
+             mouse.event == EVENT_LBUTTONUP ||
              dragging) && nobjpt < 4 )
         {
             Point2f m = mouse.pt;
@@ -258,9 +259,9 @@ static int select3DBox(const string& windowname, const string& selWinName, const
             }
             box[npt] = image2plane(imgpt[npt], R, tvec, cameraMatrix, npt<3 ? 0 : Z);
 
-            if( (npt == 0 && mouse.event == CV_EVENT_LBUTTONDOWN) ||
+            if( (npt == 0 && mouse.event == EVENT_LBUTTONDOWN) ||
                (npt > 0 && norm(box[npt] - box[npt-1]) > eps &&
-                mouse.event == CV_EVENT_LBUTTONUP) )
+                mouse.event == EVENT_LBUTTONUP) )
             {
                 nobjpt++;
                 if( nobjpt < 4 )
diff --git a/samples/cpp/simpleflow_demo.cpp b/samples/cpp/simpleflow_demo.cpp
index 2562a2bb62..c9cde783da 100644
--- a/samples/cpp/simpleflow_demo.cpp
+++ b/samples/cpp/simpleflow_demo.cpp
@@ -1,6 +1,7 @@
+#include <opencv2/core/utility.hpp>
 #include "opencv2/video/tracking.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <cstdio>
 #include <iostream>
diff --git a/samples/cpp/squares.cpp b/samples/cpp/squares.cpp
index 13a9946673..392075d79f 100644
--- a/samples/cpp/squares.cpp
+++ b/samples/cpp/squares.cpp
@@ -83,7 +83,7 @@ static void findSquares( const Mat& image, vector<vector<Point> >& squares )
             }
 
             // find contours and store them all as a list
-            findContours(gray, contours, CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE);
+            findContours(gray, contours, RETR_LIST, CHAIN_APPROX_SIMPLE);
 
             vector<Point> approx;
 
@@ -132,7 +132,7 @@ static void drawSquares( Mat& image, const vector<vector<Point> >& squares )
     {
         const Point* p = &squares[i][0];
         int n = (int)squares[i].size();
-        polylines(image, &p, &n, 1, true, Scalar(0,255,0), 3, CV_AA);
+        polylines(image, &p, &n, 1, true, Scalar(0,255,0), 3, LINE_AA);
     }
 
     imshow(wndname, image);
diff --git a/samples/cpp/starter_imagelist.cpp b/samples/cpp/starter_imagelist.cpp
index 61585a515f..fe89579301 100644
--- a/samples/cpp/starter_imagelist.cpp
+++ b/samples/cpp/starter_imagelist.cpp
@@ -46,15 +46,15 @@ bool readStringList(const string& filename, vector<string>& l)
 
 int process(vector<string> images)
 {
-  namedWindow("image",CV_WINDOW_KEEPRATIO); //resizable window;
-  for (size_t i = 0; i < images.size(); i++)
-  {
-    Mat image = imread(images[i], CV_LOAD_IMAGE_GRAYSCALE); // do grayscale processing?
-    imshow("image",image);
-    cout << "Press a key to see the next image in the list." << endl;
-    waitKey(); // wait indefinitely for a key to be pressed
-  }
-  return 0;
+    namedWindow("image", WINDOW_KEEPRATIO); //resizable window;
+    for (size_t i = 0; i < images.size(); i++)
+    {
+        Mat image = imread(images[i], IMREAD_GRAYSCALE); // do grayscale processing?
+        imshow("image",image);
+        cout << "Press a key to see the next image in the list." << endl;
+        waitKey(); // wait indefinitely for a key to be pressed
+    }
+    return 0;
 }
 
 }
diff --git a/samples/cpp/starter_video.cpp b/samples/cpp/starter_video.cpp
index b8f4086c45..1a3d5b0c9a 100644
--- a/samples/cpp/starter_video.cpp
+++ b/samples/cpp/starter_video.cpp
@@ -4,31 +4,34 @@
 *  Created on: Nov 23, 2010
 *      Author: Ethan Rublee
 *
-* A starter sample for using opencv, get a video stream and display the images
+*  Modified on: April 17, 2013
+*      Author: Kevin Hughes
+*
+* A starter sample for using OpenCV VideoCapture with capture devices, video files or image sequences
 * easy as CV_PI right?
 */
-#include "opencv2/highgui/highgui.hpp"
+
+#include <opencv2/highgui/highgui.hpp>
+
 #include <iostream>
-#include <vector>
 #include <stdio.h>
 
 using namespace cv;
 using namespace std;
 
-
-
 //hide the local functions in an anon namespace
 namespace {
     void help(char** av) {
-        cout << "\nThis program justs gets you started reading images from video\n"
-            "Usage:\n./" << av[0] << " <video device number>\n"
-            << "q,Q,esc -- quit\n"
-            << "space   -- save frame\n\n"
-            << "\tThis is a starter sample, to get you up and going in a copy pasta fashion\n"
-            << "\tThe program captures frames from a camera connected to your computer.\n"
-            << "\tTo find the video device number, try ls /dev/video* \n"
-            << "\tYou may also pass a video file, like my_vide.avi instead of a device number"
-            << endl;
+        cout << "The program captures frames from a video file, image sequence (01.jpg, 02.jpg ... 10.jpg) or camera connected to your computer." << endl
+             << "Usage:\n" << av[0] << " <video file, image sequence or device number>" << endl
+             << "q,Q,esc -- quit" << endl
+             << "space   -- save frame" << endl << endl
+             << "\tTo capture from a camera pass the device number. To find the device number, try ls /dev/video*" << endl
+             << "\texample: " << av[0] << " 0" << endl
+             << "\tYou may also pass a video file instead of a device number" << endl
+             << "\texample: " << av[0] << " video.avi" << endl
+             << "\tYou can also pass the path to an image sequence and OpenCV will treat the sequence just like a video." << endl
+             << "\texample: " << av[0] << " right%%02d.jpg" << endl;
     }
 
     int process(VideoCapture& capture) {
@@ -36,31 +39,33 @@ namespace {
         char filename[200];
         string window_name = "video | q or esc to quit";
         cout << "press space to save a picture. q or esc to quit" << endl;
-        namedWindow(window_name, CV_WINDOW_KEEPRATIO); //resizable window;
+        namedWindow(window_name, WINDOW_KEEPRATIO); //resizable window;
         Mat frame;
+
         for (;;) {
             capture >> frame;
             if (frame.empty())
                 break;
+
             imshow(window_name, frame);
-            char key = (char)waitKey(5); //delay N millis, usually long enough to display and capture input
+            char key = (char)waitKey(30); //delay N millis, usually long enough to display and capture input
+
             switch (key) {
-        case 'q':
-        case 'Q':
-        case 27: //escape key
-            return 0;
-        case ' ': //Save an image
-            sprintf(filename,"filename%.3d.jpg",n++);
-            imwrite(filename,frame);
-            cout << "Saved " << filename << endl;
-            break;
-        default:
-            break;
+            case 'q':
+            case 'Q':
+            case 27: //escape key
+                return 0;
+            case ' ': //Save an image
+                sprintf(filename,"filename%.3d.jpg",n++);
+                imwrite(filename,frame);
+                cout << "Saved " << filename << endl;
+                break;
+            default:
+                break;
             }
         }
         return 0;
     }
-
 }
 
 int main(int ac, char** av) {
@@ -70,11 +75,11 @@ int main(int ac, char** av) {
         return 1;
     }
     std::string arg = av[1];
-    VideoCapture capture(arg); //try to open string, this will attempt to open it as a video file
+    VideoCapture capture(arg); //try to open string, this will attempt to open it as a video file or image sequence
     if (!capture.isOpened()) //if this fails, try to open as a video camera, through the use of an integer param
         capture.open(atoi(arg.c_str()));
     if (!capture.isOpened()) {
-        cerr << "Failed to open a video device or video file!\n" << endl;
+        cerr << "Failed to open the video device, video file or image sequence!\n" << endl;
         help(av);
         return 1;
     }
diff --git a/samples/cpp/stereo_calib.cpp b/samples/cpp/stereo_calib.cpp
index a6d276f8b4..367df4bfef 100644
--- a/samples/cpp/stereo_calib.cpp
+++ b/samples/cpp/stereo_calib.cpp
@@ -14,15 +14,12 @@
      Or: http://oreilly.com/catalog/9780596516130/
      ISBN-10: 0596516134 or: ISBN-13: 978-0596516130
 
-   OTHER OPENCV SITES:
-   * The source code is on sourceforge at:
-     http://sourceforge.net/projects/opencvlibrary/
-   * The OpenCV wiki page (As of Oct 1, 2008 this is down for changing over servers, but should come back):
-     http://opencvlibrary.sourceforge.net/
-   * An active user group is at:
-     http://tech.groups.yahoo.com/group/OpenCV/
-   * The minutes of weekly OpenCV development meetings are at:
-     http://pr.willowgarage.com/wiki/OpenCV
+   OPENCV WEBSITES:
+     Homepage:      http://opencv.org
+     Online docs:   http://docs.opencv.org
+     Q&A forum:     http://answers.opencv.org
+     Issue tracker: http://code.opencv.org
+     GitHub:        https://github.com/Itseez/opencv/
    ************************************************** */
 
 #include "opencv2/calib3d/calib3d.hpp"
@@ -106,7 +103,7 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
                 else
                     resize(img, timg, Size(), scale, scale);
                 found = findChessboardCorners(timg, boardSize, corners,
-                    CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_NORMALIZE_IMAGE);
+                    CALIB_CB_ADAPTIVE_THRESH | CALIB_CB_NORMALIZE_IMAGE);
                 if( found )
                 {
                     if( scale > 1 )
@@ -121,7 +118,7 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
             {
                 cout << filename << endl;
                 Mat cimg, cimg1;
-                cvtColor(img, cimg, CV_GRAY2BGR);
+                cvtColor(img, cimg, COLOR_GRAY2BGR);
                 drawChessboardCorners(cimg, boardSize, corners, found);
                 double sf = 640./MAX(img.rows, img.cols);
                 resize(cimg, cimg1, Size(), sf, sf);
@@ -135,7 +132,7 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
             if( !found )
                 break;
             cornerSubPix(img, corners, Size(11,11), Size(-1,-1),
-                         TermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,
+                         TermCriteria(TermCriteria::COUNT+TermCriteria::EPS,
                                       30, 0.01));
         }
         if( k == 2 )
@@ -175,12 +172,12 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
                     cameraMatrix[0], distCoeffs[0],
                     cameraMatrix[1], distCoeffs[1],
                     imageSize, R, T, E, F,
-                    TermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 100, 1e-5),
-                    CV_CALIB_FIX_ASPECT_RATIO +
-                    CV_CALIB_ZERO_TANGENT_DIST +
-                    CV_CALIB_SAME_FOCAL_LENGTH +
-                    CV_CALIB_RATIONAL_MODEL +
-                    CV_CALIB_FIX_K3 + CV_CALIB_FIX_K4 + CV_CALIB_FIX_K5);
+                    TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 100, 1e-5),
+                    CALIB_FIX_ASPECT_RATIO +
+                    CALIB_ZERO_TANGENT_DIST +
+                    CALIB_SAME_FOCAL_LENGTH +
+                    CALIB_RATIONAL_MODEL +
+                    CALIB_FIX_K3 + CALIB_FIX_K4 + CALIB_FIX_K5);
     cout << "done with RMS error=" << rms << endl;
 
 // CALIBRATION QUALITY CHECK
@@ -214,7 +211,7 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
     cout << "average reprojection err = " <<  err/npoints << endl;
 
     // save intrinsic parameters
-    FileStorage fs("intrinsics.yml", CV_STORAGE_WRITE);
+    FileStorage fs("intrinsics.yml", FileStorage::WRITE);
     if( fs.isOpened() )
     {
         fs << "M1" << cameraMatrix[0] << "D1" << distCoeffs[0] <<
@@ -232,7 +229,7 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
                   imageSize, R, T, R1, R2, P1, P2, Q,
                   CALIB_ZERO_DISPARITY, 1, imageSize, &validRoi[0], &validRoi[1]);
 
-    fs.open("extrinsics.yml", CV_STORAGE_WRITE);
+    fs.open("extrinsics.yml", FileStorage::WRITE);
     if( fs.isOpened() )
     {
         fs << "R" << R << "T" << T << "R1" << R1 << "R2" << R2 << "P1" << P1 << "P2" << P2 << "Q" << Q;
@@ -304,10 +301,10 @@ StereoCalib(const vector<string>& imagelist, Size boardSize, bool useCalibrated=
         for( k = 0; k < 2; k++ )
         {
             Mat img = imread(goodImageList[i*2+k], 0), rimg, cimg;
-            remap(img, rimg, rmap[k][0], rmap[k][1], CV_INTER_LINEAR);
-            cvtColor(rimg, cimg, CV_GRAY2BGR);
+            remap(img, rimg, rmap[k][0], rmap[k][1], INTER_LINEAR);
+            cvtColor(rimg, cimg, COLOR_GRAY2BGR);
             Mat canvasPart = !isVerticalStereo ? canvas(Rect(w*k, 0, w, h)) : canvas(Rect(0, h*k, w, h));
-            resize(cimg, canvasPart, canvasPart.size(), 0, 0, CV_INTER_AREA);
+            resize(cimg, canvasPart, canvasPart.size(), 0, 0, INTER_AREA);
             if( useCalibrated )
             {
                 Rect vroi(cvRound(validRoi[k].x*sf), cvRound(validRoi[k].y*sf),
@@ -404,4 +401,3 @@ int main(int argc, char** argv)
     StereoCalib(imagelist, boardSize, true, showRectified);
     return 0;
 }
-
diff --git a/samples/cpp/stereo_match.cpp b/samples/cpp/stereo_match.cpp
index 7e1f6cb687..6f0b9f38b6 100644
--- a/samples/cpp/stereo_match.cpp
+++ b/samples/cpp/stereo_match.cpp
@@ -11,6 +11,7 @@
 #include "opencv2/imgproc/imgproc.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/contrib/contrib.hpp"
+#include "opencv2/core/utility.hpp"
 
 #include <stdio.h>
 
@@ -66,8 +67,8 @@ int main(int argc, char** argv)
     bool no_display = false;
     float scale = 1.f;
 
-    Ptr<StereoMatcher> bm = createStereoBM(16,9);
-    Ptr<StereoMatcher> sgbm = createStereoSGBM(0,16,3);
+    Ptr<StereoBM> bm = createStereoBM(16,9);
+    Ptr<StereoSGBM> sgbm = createStereoSGBM(0,16,3);
     StereoVar var;
 
     for( int i = 1; i < argc; i++ )
@@ -177,7 +178,7 @@ int main(int argc, char** argv)
     if( intrinsic_filename )
     {
         // reading intrinsic parameters
-        FileStorage fs(intrinsic_filename, CV_STORAGE_READ);
+        FileStorage fs(intrinsic_filename, FileStorage::READ);
         if(!fs.isOpened())
         {
             printf("Failed to open file %s\n", intrinsic_filename);
@@ -193,7 +194,7 @@ int main(int argc, char** argv)
         M1 *= scale;
         M2 *= scale;
 
-        fs.open(extrinsic_filename, CV_STORAGE_READ);
+        fs.open(extrinsic_filename, FileStorage::READ);
         if(!fs.isOpened())
         {
             printf("Failed to open file %s\n", extrinsic_filename);
@@ -220,33 +221,33 @@ int main(int argc, char** argv)
 
     numberOfDisparities = numberOfDisparities > 0 ? numberOfDisparities : ((img_size.width/8) + 15) & -16;
 
-    //bm->set("roi1", roi1);
-    //bm->set("roi2", roi2);
-    bm->set("preFilterCap", 31);
-    bm->set("SADWindowSize", SADWindowSize > 0 ? SADWindowSize : 9);
-    bm->set("minDisparity", 0);
-    bm->set("numDisparities", numberOfDisparities);
-    bm->set("textureThreshold", 10);
-    bm->set("uniquenessRatio", 15);
-    bm->set("speckleWindowSize", 100);
-    bm->set("speckleRange", 32);
-    bm->set("disp12MaxDiff", 1);
+    bm->setROI1(roi1);
+    bm->setROI2(roi2);
+    bm->setPreFilterCap(31);
+    bm->setBlockSize(SADWindowSize > 0 ? SADWindowSize : 9);
+    bm->setMinDisparity(0);
+    bm->setNumDisparities(numberOfDisparities);
+    bm->setTextureThreshold(10);
+    bm->setUniquenessRatio(15);
+    bm->setSpeckleWindowSize(100);
+    bm->setSpeckleRange(32);
+    bm->setDisp12MaxDiff(1);
 
-    sgbm->set("preFilterCap", 63);
+    sgbm->setPreFilterCap(63);
     int sgbmWinSize = SADWindowSize > 0 ? SADWindowSize : 3;
-    sgbm->set("SADWindowSize", sgbmWinSize);
+    sgbm->setBlockSize(sgbmWinSize);
 
     int cn = img1.channels();
 
-    sgbm->set("P1", 8*cn*sgbmWinSize*sgbmWinSize);
-    sgbm->set("P2", 32*cn*sgbmWinSize*sgbmWinSize);
-    sgbm->set("minDisparity", 0);
-    sgbm->set("numDisparities", numberOfDisparities);
-    sgbm->set("uniquenessRatio", 10);
-    sgbm->set("speckleWindowSize", 100);
-    sgbm->set("speckleRange", 32);
-    sgbm->set("disp12MaxDiff", 1);
-    sgbm->set("fullDP", alg == STEREO_HH);
+    sgbm->setP1(8*cn*sgbmWinSize*sgbmWinSize);
+    sgbm->setP2(32*cn*sgbmWinSize*sgbmWinSize);
+    sgbm->setMinDisparity(0);
+    sgbm->setNumDisparities(numberOfDisparities);
+    sgbm->setUniquenessRatio(10);
+    sgbm->setSpeckleWindowSize(100);
+    sgbm->setSpeckleRange(32);
+    sgbm->setDisp12MaxDiff(1);
+    sgbm->setMode(alg == STEREO_HH ? StereoSGBM::MODE_HH : StereoSGBM::MODE_SGBM);
 
     var.levels = 3;                                 // ignored with USE_AUTO_PARAMS
     var.pyrScale = 0.5;                             // ignored with USE_AUTO_PARAMS
diff --git a/samples/cpp/stitching.cpp b/samples/cpp/stitching.cpp
index 1ee6bdb7ce..602edfe86e 100644
--- a/samples/cpp/stitching.cpp
+++ b/samples/cpp/stitching.cpp
@@ -134,5 +134,3 @@ int parseCmdArgs(int argc, char** argv)
     }
     return 0;
 }
-
-
diff --git a/samples/cpp/stitching_detailed.cpp b/samples/cpp/stitching_detailed.cpp
index 73d139e41b..a576816d69 100644
--- a/samples/cpp/stitching_detailed.cpp
+++ b/samples/cpp/stitching_detailed.cpp
@@ -45,7 +45,8 @@
 #include <fstream>
 #include <string>
 #include "opencv2/opencv_modules.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/highgui.hpp"
 #include "opencv2/stitching/detail/autocalib.hpp"
 #include "opencv2/stitching/detail/blenders.hpp"
 #include "opencv2/stitching/detail/camera.hpp"
@@ -120,7 +121,7 @@ static void printUsage()
 
 
 // Default command line args
-vector<string> img_names;
+vector<String> img_names;
 bool preview = false;
 bool try_gpu = false;
 double work_megapix = 0.6;
@@ -355,16 +356,16 @@ int main(int argc, char* argv[])
     Ptr<FeaturesFinder> finder;
     if (features_type == "surf")
     {
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_NONFREE
         if (try_gpu && gpu::getCudaEnabledDeviceCount() > 0)
-            finder = new SurfFeaturesFinderGpu();
+            finder = makePtr<SurfFeaturesFinderGpu>();
         else
 #endif
-            finder = new SurfFeaturesFinder();
+            finder = makePtr<SurfFeaturesFinder>();
     }
     else if (features_type == "orb")
     {
-        finder = new OrbFeaturesFinder();
+        finder = makePtr<OrbFeaturesFinder>();
     }
     else
     {
@@ -445,7 +446,7 @@ int main(int argc, char* argv[])
     // Leave only images we are sure are from the same panorama
     vector<int> indices = leaveBiggestComponent(features, pairwise_matches, conf_thresh);
     vector<Mat> img_subset;
-    vector<string> img_names_subset;
+    vector<String> img_names_subset;
     vector<Size> full_img_sizes_subset;
     for (size_t i = 0; i < indices.size(); ++i)
     {
@@ -468,7 +469,11 @@ int main(int argc, char* argv[])
 
     HomographyBasedEstimator estimator;
     vector<CameraParams> cameras;
-    estimator(features, pairwise_matches, cameras);
+    if (!estimator(features, pairwise_matches, cameras))
+    {
+        cout << "Homography estimation failed.\n";
+        return -1;
+    }
 
     for (size_t i = 0; i < cameras.size(); ++i)
     {
@@ -479,8 +484,8 @@ int main(int argc, char* argv[])
     }
 
     Ptr<detail::BundleAdjusterBase> adjuster;
-    if (ba_cost_func == "reproj") adjuster = new detail::BundleAdjusterReproj();
-    else if (ba_cost_func == "ray") adjuster = new detail::BundleAdjusterRay();
+    if (ba_cost_func == "reproj") adjuster = makePtr<detail::BundleAdjusterReproj>();
+    else if (ba_cost_func == "ray") adjuster = makePtr<detail::BundleAdjusterRay>();
     else
     {
         cout << "Unknown bundle adjustment cost function: '" << ba_cost_func << "'.\n";
@@ -494,7 +499,11 @@ int main(int argc, char* argv[])
     if (ba_refine_mask[3] == 'x') refine_mask(1,1) = 1;
     if (ba_refine_mask[4] == 'x') refine_mask(1,2) = 1;
     adjuster->setRefinementMask(refine_mask);
-    (*adjuster)(features, pairwise_matches, cameras);
+    if (!(*adjuster)(features, pairwise_matches, cameras))
+    {
+        cout << "Camera parameters adjusting failed.\n";
+        return -1;
+    }
 
     // Find median focal length
 
@@ -543,34 +552,52 @@ int main(int argc, char* argv[])
     // Warp images and their masks
 
     Ptr<WarperCreator> warper_creator;
-#ifdef HAVE_OPENCV_GPU
+#ifdef HAVE_OPENCV_GPUWARPING
     if (try_gpu && gpu::getCudaEnabledDeviceCount() > 0)
     {
-        if (warp_type == "plane") warper_creator = new cv::PlaneWarperGpu();
-        else if (warp_type == "cylindrical") warper_creator = new cv::CylindricalWarperGpu();
-        else if (warp_type == "spherical") warper_creator = new cv::SphericalWarperGpu();
+        if (warp_type == "plane")
+            warper_creator = makePtr<cv::PlaneWarperGpu>();
+        else if (warp_type == "cylindrical")
+            warper_creator = makePtr<cv::CylindricalWarperGpu>();
+        else if (warp_type == "spherical")
+            warper_creator = makePtr<cv::SphericalWarperGpu>();
     }
     else
 #endif
     {
-        if (warp_type == "plane") warper_creator = new cv::PlaneWarper();
-        else if (warp_type == "cylindrical") warper_creator = new cv::CylindricalWarper();
-        else if (warp_type == "spherical") warper_creator = new cv::SphericalWarper();
-        else if (warp_type == "fisheye") warper_creator = new cv::FisheyeWarper();
-        else if (warp_type == "stereographic") warper_creator = new cv::StereographicWarper();
-        else if (warp_type == "compressedPlaneA2B1") warper_creator = new cv::CompressedRectilinearWarper(2, 1);
-        else if (warp_type == "compressedPlaneA1.5B1") warper_creator = new cv::CompressedRectilinearWarper(1.5, 1);
-        else if (warp_type == "compressedPlanePortraitA2B1") warper_creator = new cv::CompressedRectilinearPortraitWarper(2, 1);
-        else if (warp_type == "compressedPlanePortraitA1.5B1") warper_creator = new cv::CompressedRectilinearPortraitWarper(1.5, 1);
-        else if (warp_type == "paniniA2B1") warper_creator = new cv::PaniniWarper(2, 1);
-        else if (warp_type == "paniniA1.5B1") warper_creator = new cv::PaniniWarper(1.5, 1);
-        else if (warp_type == "paniniPortraitA2B1") warper_creator = new cv::PaniniPortraitWarper(2, 1);
-        else if (warp_type == "paniniPortraitA1.5B1") warper_creator = new cv::PaniniPortraitWarper(1.5, 1);
-        else if (warp_type == "mercator") warper_creator = new cv::MercatorWarper();
-        else if (warp_type == "transverseMercator") warper_creator = new cv::TransverseMercatorWarper();
+        if (warp_type == "plane")
+            warper_creator = makePtr<cv::PlaneWarper>();
+        else if (warp_type == "cylindrical")
+            warper_creator = makePtr<cv::CylindricalWarper>();
+        else if (warp_type == "spherical")
+            warper_creator = makePtr<cv::SphericalWarper>();
+        else if (warp_type == "fisheye")
+            warper_creator = makePtr<cv::FisheyeWarper>();
+        else if (warp_type == "stereographic")
+            warper_creator = makePtr<cv::StereographicWarper>();
+        else if (warp_type == "compressedPlaneA2B1")
+            warper_creator = makePtr<cv::CompressedRectilinearWarper>(2.0f, 1.0f);
+        else if (warp_type == "compressedPlaneA1.5B1")
+            warper_creator = makePtr<cv::CompressedRectilinearWarper>(1.5f, 1.0f);
+        else if (warp_type == "compressedPlanePortraitA2B1")
+            warper_creator = makePtr<cv::CompressedRectilinearPortraitWarper>(2.0f, 1.0f);
+        else if (warp_type == "compressedPlanePortraitA1.5B1")
+            warper_creator = makePtr<cv::CompressedRectilinearPortraitWarper>(1.5f, 1.0f);
+        else if (warp_type == "paniniA2B1")
+            warper_creator = makePtr<cv::PaniniWarper>(2.0f, 1.0f);
+        else if (warp_type == "paniniA1.5B1")
+            warper_creator = makePtr<cv::PaniniWarper>(1.5f, 1.0f);
+        else if (warp_type == "paniniPortraitA2B1")
+            warper_creator = makePtr<cv::PaniniPortraitWarper>(2.0f, 1.0f);
+        else if (warp_type == "paniniPortraitA1.5B1")
+            warper_creator = makePtr<cv::PaniniPortraitWarper>(1.5f, 1.0f);
+        else if (warp_type == "mercator")
+            warper_creator = makePtr<cv::MercatorWarper>();
+        else if (warp_type == "transverseMercator")
+            warper_creator = makePtr<cv::TransverseMercatorWarper>();
     }
 
-    if (warper_creator.empty())
+    if (!warper_creator)
     {
         cout << "Can't create the following warper '" << warp_type << "'\n";
         return 1;
@@ -603,32 +630,32 @@ int main(int argc, char* argv[])
 
     Ptr<SeamFinder> seam_finder;
     if (seam_find_type == "no")
-        seam_finder = new detail::NoSeamFinder();
+        seam_finder = makePtr<detail::NoSeamFinder>();
     else if (seam_find_type == "voronoi")
-        seam_finder = new detail::VoronoiSeamFinder();
+        seam_finder = makePtr<detail::VoronoiSeamFinder>();
     else if (seam_find_type == "gc_color")
     {
 #ifdef HAVE_OPENCV_GPU
         if (try_gpu && gpu::getCudaEnabledDeviceCount() > 0)
-            seam_finder = new detail::GraphCutSeamFinderGpu(GraphCutSeamFinderBase::COST_COLOR);
+            seam_finder = makePtr<detail::GraphCutSeamFinderGpu>(GraphCutSeamFinderBase::COST_COLOR);
         else
 #endif
-            seam_finder = new detail::GraphCutSeamFinder(GraphCutSeamFinderBase::COST_COLOR);
+            seam_finder = makePtr<detail::GraphCutSeamFinder>(GraphCutSeamFinderBase::COST_COLOR);
     }
     else if (seam_find_type == "gc_colorgrad")
     {
 #ifdef HAVE_OPENCV_GPU
         if (try_gpu && gpu::getCudaEnabledDeviceCount() > 0)
-            seam_finder = new detail::GraphCutSeamFinderGpu(GraphCutSeamFinderBase::COST_COLOR_GRAD);
+            seam_finder = makePtr<detail::GraphCutSeamFinderGpu>(GraphCutSeamFinderBase::COST_COLOR_GRAD);
         else
 #endif
-            seam_finder = new detail::GraphCutSeamFinder(GraphCutSeamFinderBase::COST_COLOR_GRAD);
+            seam_finder = makePtr<detail::GraphCutSeamFinder>(GraphCutSeamFinderBase::COST_COLOR_GRAD);
     }
     else if (seam_find_type == "dp_color")
-        seam_finder = new detail::DpSeamFinder(DpSeamFinder::COLOR);
+        seam_finder = makePtr<detail::DpSeamFinder>(DpSeamFinder::COLOR);
     else if (seam_find_type == "dp_colorgrad")
-        seam_finder = new detail::DpSeamFinder(DpSeamFinder::COLOR_GRAD);
-    if (seam_finder.empty())
+        seam_finder = makePtr<detail::DpSeamFinder>(DpSeamFinder::COLOR_GRAD);
+    if (!seam_finder)
     {
         cout << "Can't create the following seam finder '" << seam_find_type << "'\n";
         return 1;
@@ -726,7 +753,7 @@ int main(int argc, char* argv[])
         resize(dilated_mask, seam_mask, mask_warped.size());
         mask_warped = seam_mask & mask_warped;
 
-        if (blender.empty())
+        if (!blender)
         {
             blender = Blender::createDefault(blend_type, try_gpu);
             Size dst_sz = resultRoi(corners, sizes).size();
@@ -735,13 +762,13 @@ int main(int argc, char* argv[])
                 blender = Blender::createDefault(Blender::NO, try_gpu);
             else if (blend_type == Blender::MULTI_BAND)
             {
-                MultiBandBlender* mb = dynamic_cast<MultiBandBlender*>(static_cast<Blender*>(blender));
+                MultiBandBlender* mb = dynamic_cast<MultiBandBlender*>(blender.get());
                 mb->setNumBands(static_cast<int>(ceil(log(blend_width)/log(2.)) - 1.));
                 LOGLN("Multi-band blender, number of bands: " << mb->numBands());
             }
             else if (blend_type == Blender::FEATHER)
             {
-                FeatherBlender* fb = dynamic_cast<FeatherBlender*>(static_cast<Blender*>(blender));
+                FeatherBlender* fb = dynamic_cast<FeatherBlender*>(blender.get());
                 fb->setSharpness(1.f/blend_width);
                 LOGLN("Feather blender, sharpness: " << fb->sharpness());
             }
@@ -762,5 +789,3 @@ int main(int argc, char* argv[])
     LOGLN("Finished, total time: " << ((getTickCount() - app_start_time) / getTickFrequency()) << " sec");
     return 0;
 }
-
-
diff --git a/samples/cpp/trained_classifierNM1.xml b/samples/cpp/trained_classifierNM1.xml
new file mode 100644
index 0000000000..c395a67df7
--- /dev/null
+++ b/samples/cpp/trained_classifierNM1.xml
@@ -0,0 +1,4046 @@
+<?xml version="1.0"?>
+<opencv_storage>
+<boost type_id="opencv-ml-boost-tree">
+  <boosting_type>RealAdaboost</boosting_type>
+  <splitting_criteria>Gini</splitting_criteria>
+  <ntrees>100</ntrees>
+  <weight_trimming_rate>1.</weight_trimming_rate>
+  <is_classifier>1</is_classifier>
+  <var_all>5</var_all>
+  <var_count>4</var_count>
+  <ord_var_count>4</ord_var_count>
+  <cat_var_count>0</cat_var_count>
+  <training_params>
+    <use_surrogates>0</use_surrogates>
+    <max_categories>10</max_categories>
+    <max_depth>1</max_depth>
+    <min_sample_count>10</min_sample_count>
+    <cross_validation_folds>0</cross_validation_folds>
+    <priors type_id="opencv-matrix">
+      <rows>1</rows>
+      <cols>2</cols>
+      <dt>d</dt>
+      <data>
+        1. 1.</data></priors></training_params>
+  <var_idx type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>4</cols>
+    <dt>i</dt>
+    <data>
+      1 2 3 4</data></var_idx>
+  <var_type>
+    0 0 0 0</var_type>
+  <cat_count type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>1</cols>
+    <dt>i</dt>
+    <data>
+      2</data></cat_count>
+  <cat_map type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>2</cols>
+    <dt>i</dt>
+    <data>
+      1 2</data></cat_map>
+  <trees>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-8.7333809708193622e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>8.9537560939788818e-01</quality>
+              <le>8.5429996252059937e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>281</sample_count>
+          <value>6.7336438045685210e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1578</sample_count>
+          <value>-1.6797288877447023e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>6.8833827526759232e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>6.0157060623168945e-01</quality>
+              <le>1.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1728</sample_count>
+          <value>-2.3830233314576746e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>131</sample_count>
+          <value>1.3270520927417431e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.1546319456101495e-14</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>6.0729509592056274e-01</quality>
+              <le>1.0963299870491028e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>781</sample_count>
+          <value>2.8488485626451149e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1078</sample_count>
+          <value>-1.0290340538001230e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>8.8817841970011734e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.8338499069213867e-01</quality>
+              <le>9.5209401845932007e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1331</sample_count>
+          <value>-4.6437394337436783e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>528</sample_count>
+          <value>4.0542316486679569e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.5931700403371249e-14</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.5465269088745117e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>856</sample_count>
+          <value>5.0630658667623052e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1003</sample_count>
+          <value>-2.3844591468675169e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-4.0523140398818379e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.5272746086120605e-01</quality>
+              <le>5.0000000000000000e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1046</sample_count>
+          <value>-3.5086409511139033e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>813</sample_count>
+          <value>3.2363636038761090e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.5527136788004883e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.4958724975585938e-01</quality>
+              <le>1.6267855167388916e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1811</sample_count>
+          <value>-1.0723654593445622e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>48</sample_count>
+          <value>1.6463996332936000e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.5543122344752215e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.5254465341567993e-01</quality>
+              <le>7.4758999049663544e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>149</sample_count>
+          <value>1.0273978441887677e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1710</sample_count>
+          <value>-1.3682146780021626e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.1990408665951547e-14</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.4537707567214966e-01</quality>
+              <le>6.6345453262329102e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>455</sample_count>
+          <value>-1.0072489119231773e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1404</sample_count>
+          <value>1.1925375544044788e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-7.1054273576010523e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.2146106958389282e-01</quality>
+              <le>2.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1791</sample_count>
+          <value>-4.7305821386203636e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>68</sample_count>
+          <value>1.5160735804641536e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.2081108093261719e-01</quality>
+              <le>5.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1690</sample_count>
+          <value>9.1796156088740075e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>169</sample_count>
+          <value>-4.9059916762377265e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-7.1054273576010523e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.2672868967056274e-01</quality>
+              <le>1.0963299870491028e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>781</sample_count>
+          <value>1.0385059515501381e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1078</sample_count>
+          <value>-5.7169325870124055e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.1102230246251577e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.2609002590179443e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>856</sample_count>
+          <value>3.2591023278276837e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1003</sample_count>
+          <value>-1.6727738809775036e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.5527136788004883e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.1409310102462769e-01</quality>
+              <le>7.1857005357742310e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>114</sample_count>
+          <value>6.4622038086485267e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1745</sample_count>
+          <value>-4.9566471799398701e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-7.4940054162198634e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.1351374387741089e-01</quality>
+              <le>1.4143149554729462e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1571</sample_count>
+          <value>-4.6424296372866977e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>288</sample_count>
+          <value>6.6639156578772774e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.6645352591003686e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.2261954545974731e-01</quality>
+              <le>1.0177600383758545e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>590</sample_count>
+          <value>1.3091283509473614e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1269</sample_count>
+          <value>-3.6264100845476954e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.2181720733642578e-01</quality>
+              <le>8.9620999991893768e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>352</sample_count>
+          <value>-2.4183425262980360e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1507</sample_count>
+          <value>1.8605210755965146e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.1102230246251554e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0836616754531860e-01</quality>
+              <le>1.0963299870491028e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>781</sample_count>
+          <value>5.1954271677409611e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1078</sample_count>
+          <value>-3.3426585985028628e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0880891084671021e-01</quality>
+              <le>1.1348484754562378e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1664</sample_count>
+          <value>-7.4858358240636261e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>195</sample_count>
+          <value>2.4031047447238377e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.3322676295501896e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.1081919670104980e-01</quality>
+              <le>1.2661495208740234e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1730</sample_count>
+          <value>5.6719022662652843e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>129</sample_count>
+          <value>-4.0229379476194682e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>5.5511151231257519e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0904816389083862e-01</quality>
+              <le>2.3689103126525879e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1838</sample_count>
+          <value>-1.9140961693183144e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>21</sample_count>
+          <value>1.7878855228391193e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-9.2703622556201423e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0664407014846802e-01</quality>
+              <le>2.2119045257568359e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1836</sample_count>
+          <value>1.8191803445960011e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>23</sample_count>
+          <value>-9.2985231739132745e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.1102230246251577e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0936371088027954e-01</quality>
+              <le>1.2460045814514160e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1724</sample_count>
+          <value>-5.7918483977216939e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>135</sample_count>
+          <value>3.3577728337370422e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.4424906541753385e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0752711296081543e-01</quality>
+              <le>9.5505046844482422e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1341</sample_count>
+          <value>1.1992717133073164e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>518</sample_count>
+          <value>-1.2680428752841727e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-4.8849813083507124e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0749230384826660e-01</quality>
+              <le>7.8355848789215088e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>764</sample_count>
+          <value>-2.4384603502165489e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1095</sample_count>
+          <value>6.2746713049518996e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.9984014443252778e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0703763961791992e-01</quality>
+              <le>6.3535496592521667e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>59</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1800</sample_count>
+          <value>-1.4076159604991374e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.4512436177016777e-05</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0712978839874268e-01</quality>
+              <le>6.4906999468803406e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>61</sample_count>
+          <value>-2.8822016897573164e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1798</sample_count>
+          <value>1.4394134992880541e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.1009607315063477e-01</quality>
+              <le>6.9339498877525330e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>87</sample_count>
+          <value>1.6547355644710868e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1772</sample_count>
+          <value>-2.1726950528099288e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.1102230246251577e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.0710922479629517e-01</quality>
+              <le>7.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1823</sample_count>
+          <value>1.7568593940350565e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>36</sample_count>
+          <value>-1.1252698341588880e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.7763568394002536e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0767201185226440e-01</quality>
+              <le>8.0150999128818512e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>206</sample_count>
+          <value>2.7041738749866351e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1653</sample_count>
+          <value>-5.8184179168035623e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.3306690738754586e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0510478019714355e-01</quality>
+              <le>6.2720644474029541e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>374</sample_count>
+          <value>-4.5462988716181019e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1485</sample_count>
+          <value>2.3987712659202257e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-4.2188474935756130e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.0439256429672241e-01</quality>
+              <le>2.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1791</sample_count>
+          <value>-1.5635519241144060e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>68</sample_count>
+          <value>6.3562787613239691e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>6.6613381477508951e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.0438237190246582e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>856</sample_count>
+          <value>1.3081676957882793e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1003</sample_count>
+          <value>-6.7484142743913345e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.1102230246251554e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0724387168884277e-01</quality>
+              <le>1.1533749848604202e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>921</sample_count>
+          <value>4.0250806766256482e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>938</sample_count>
+          <value>-3.7703668263161400e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-5.8286708792821058e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0915646553039551e-01</quality>
+              <le>1.4143149554729462e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1571</sample_count>
+          <value>-3.3928131204983447e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>288</sample_count>
+          <value>6.0410410576585261e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.2188474935755767e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0659310817718506e-01</quality>
+              <le>1.4145749807357788e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1572</sample_count>
+          <value>2.0368450770510211e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>287</sample_count>
+          <value>-7.7093695859120004e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.3877787807814476e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0616967678070068e-01</quality>
+              <le>1.6267855167388916e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1811</sample_count>
+          <value>-2.4173163361906828e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>48</sample_count>
+          <value>5.6348494257358028e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-3.8857805861880494e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0677150487899780e-01</quality>
+              <le>1.2661495208740234e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1730</sample_count>
+          <value>4.0402378477571901e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>129</sample_count>
+          <value>-3.4888339817180891e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0588327646255493e-01</quality>
+              <le>1.0797724723815918e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1608</sample_count>
+          <value>-6.8287991647037555e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>251</sample_count>
+          <value>1.7431967242459881e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.6629367034256354e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0476020574569702e-01</quality>
+              <le>9.2052650451660156e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1253</sample_count>
+          <value>1.0513502662517235e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>606</sample_count>
+          <value>-9.1138849679050823e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0690639019012451e-01</quality>
+              <le>9.0829002857208252e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1219</sample_count>
+          <value>-1.4636537853223561e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>640</sample_count>
+          <value>9.5332675553986543e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.2188474935755767e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0534439086914062e-01</quality>
+              <le>9.2052650451660156e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1253</sample_count>
+          <value>1.1293790129864213e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>606</sample_count>
+          <value>-9.5332675553986113e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.1649348980190600e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0435024499893188e-01</quality>
+              <le>9.3664005398750305e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>411</sample_count>
+          <value>8.4555682036086433e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1448</sample_count>
+          <value>-1.0351006034230040e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0704753398895264e-01</quality>
+              <le>9.9698998034000397e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>544</sample_count>
+          <value>-8.5042709024257673e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1315</sample_count>
+          <value>1.6769533771522532e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0574851036071777e-01</quality>
+              <le>1.0386450588703156e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>639</sample_count>
+          <value>5.3643508899729207e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1220</sample_count>
+          <value>-2.1791341988666471e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0597953796386719e-01</quality>
+              <le>1.4143149554729462e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1571</sample_count>
+          <value>-2.3199011211312900e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>288</sample_count>
+          <value>5.7032019870719552e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0620901584625244e-01</quality>
+              <le>1.1533749848604202e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>921</sample_count>
+          <value>3.0260573388611104e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>938</sample_count>
+          <value>-4.3620446423872927e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0698792934417725e-01</quality>
+              <le>1.0789400339126587e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>742</sample_count>
+          <value>-4.7900146843168623e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1117</sample_count>
+          <value>3.0074316296999687e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0459891557693481e-01</quality>
+              <le>9.0829002857208252e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1219</sample_count>
+          <value>-1.1779072393798255e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>640</sample_count>
+          <value>7.8608870331688729e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.3322676295501861e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0399231910705566e-01</quality>
+              <le>3.3933651447296143e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>59</sample_count>
+          <value>1.3770357317310018e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1800</sample_count>
+          <value>-9.0707243801608061e-03</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.6090241078691246e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0383007526397705e-01</quality>
+              <le>1.0386450588703156e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>639</sample_count>
+          <value>4.0916632352681503e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1220</sample_count>
+          <value>-1.8955755646648090e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0380486249923706e-01</quality>
+              <le>9.9698998034000397e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>544</sample_count>
+          <value>-5.8888670701558621e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1315</sample_count>
+          <value>1.3009980239826119e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.2204460492503083e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0560057163238525e-01</quality>
+              <le>9.4531498849391937e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>431</sample_count>
+          <value>8.6647133637941770e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1428</sample_count>
+          <value>-1.3032934253469194e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.3877787807814476e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0492835044860840e-01</quality>
+              <le>8.9620999991893768e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>352</sample_count>
+          <value>-1.0506312397672390e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1507</sample_count>
+          <value>9.4441739612300818e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-9.4368957093138405e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0368678569793701e-01</quality>
+              <le>9.2722505331039429e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>392</sample_count>
+          <value>7.6022467408684444e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1467</sample_count>
+          <value>-9.7486085972466607e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.1086244689504288e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0421810150146484e-01</quality>
+              <le>9.9444493651390076e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>536</sample_count>
+          <value>-6.3752797298286562e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1323</sample_count>
+          <value>1.3328967944519532e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0323176383972168e-01</quality>
+              <le>1.0177600383758545e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>590</sample_count>
+          <value>4.3831310206729070e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1269</sample_count>
+          <value>-1.4864401441489694e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-8.8817841970012602e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0349098443984985e-01</quality>
+              <le>1.0789400339126587e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>742</sample_count>
+          <value>-3.4075730824220232e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1117</sample_count>
+          <value>2.0791874238488089e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-4.4408920985006281e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0392138957977295e-01</quality>
+              <le>1.0963299870491028e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>781</sample_count>
+          <value>2.6986027157566792e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1078</sample_count>
+          <value>-2.9932407189155552e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0362765789031982e-01</quality>
+              <le>2.3689103126525879e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1838</sample_count>
+          <value>-9.2384036977927863e-03</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>21</sample_count>
+          <value>1.0592165467278325e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.9984014443252857e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0543153285980225e-01</quality>
+              <le>2.2119045257568359e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1836</sample_count>
+          <value>1.5914551618823981e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>23</sample_count>
+          <value>-8.3405482949082688e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.5527136788004883e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0523018836975098e-01</quality>
+              <le>1.6267855167388916e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1811</sample_count>
+          <value>-1.9824411253917307e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>48</sample_count>
+          <value>5.8698394468331438e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-4.2188474935756130e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0369459390640259e-01</quality>
+              <le>1.2661495208740234e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1730</sample_count>
+          <value>2.8318382799977653e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>129</sample_count>
+          <value>-2.6718445379285333e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0479084253311157e-01</quality>
+              <le>1.2460045814514160e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1724</sample_count>
+          <value>-3.7451522137317672e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>135</sample_count>
+          <value>2.6178246916992864e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>6.6613381477509353e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0358188152313232e-01</quality>
+              <le>1.2661495208740234e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1730</sample_count>
+          <value>2.7995062956094632e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>129</sample_count>
+          <value>-2.6178246916992898e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-5.5511151231257827e-17</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.0313872098922729e-01</quality>
+              <le>5.0000000000000000e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1046</sample_count>
+          <value>-8.7344555737785268e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>813</sample_count>
+          <value>7.2176940461025546e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.9984014443252857e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0293076038360596e-01</quality>
+              <le>1.2460045814514160e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1724</sample_count>
+          <value>-2.8268637265518849e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>135</sample_count>
+          <value>2.1045949672703843e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.2188474935755767e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.0332474708557129e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>856</sample_count>
+          <value>1.2230679161917310e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1003</sample_count>
+          <value>-5.4692581535459049e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.9984014443252857e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0446850061416626e-01</quality>
+              <le>1.1533749848604202e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>921</sample_count>
+          <value>2.3320665865473968e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>938</sample_count>
+          <value>-4.0391049572091403e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0324887037277222e-01</quality>
+              <le>8.0150999128818512e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>206</sample_count>
+          <value>1.6178738318585617e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1653</sample_count>
+          <value>-4.0534413064780048e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0377529859542847e-01</quality>
+              <le>1.4143149554729462e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1571</sample_count>
+          <value>-1.6597530926694307e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>288</sample_count>
+          <value>4.9094014377722600e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-3.1086244689504478e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0313144922256470e-01</quality>
+              <le>4.5337548851966858e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>110</sample_count>
+          <value>-8.1199669103862704e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1749</sample_count>
+          <value>9.3381880402274662e-03</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0475740432739258e-01</quality>
+              <le>4.5797997713088989e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>113</sample_count>
+          <value>6.8212891092482997e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1746</sample_count>
+          <value>-1.6049266927207175e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0308030843734741e-01</quality>
+              <le>6.4906999468803406e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>61</sample_count>
+          <value>-1.8287112564925503e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1798</sample_count>
+          <value>6.4869817401182221e-03</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>6.6613381477509353e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0292497873306274e-01</quality>
+              <le>1.4145749807357788e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1572</sample_count>
+          <value>1.1075480383243890e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>287</sample_count>
+          <value>-5.8766713493584966e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0272303819656372e-01</quality>
+              <le>1.0132449865341187e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>577</sample_count>
+          <value>-4.4462051722396978e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1282</sample_count>
+          <value>1.2318755109692310e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0431984663009644e-01</quality>
+              <le>9.4531498849391937e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>431</sample_count>
+          <value>7.5132879390893734e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1428</sample_count>
+          <value>-1.1572265040778912e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0449663400650024e-01</quality>
+              <le>8.3246499300003052e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>254</sample_count>
+          <value>-1.3886315377806810e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1605</sample_count>
+          <value>6.5272198323505989e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0269138813018799e-01</quality>
+              <le>8.0150999128818512e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>206</sample_count>
+          <value>1.4709439757444326e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1653</sample_count>
+          <value>-3.6874150180988774e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.0234586000442505e-01</quality>
+              <le>5.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1819</sample_count>
+          <value>-6.0288987564491008e-03</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>40</sample_count>
+          <value>1.0408276196291879e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.0341212749481201e-01</quality>
+              <le>4.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1810</sample_count>
+          <value>1.0137680632639267e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>49</sample_count>
+          <value>-8.1653792840371642e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.2767564783189316e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0258362293243408e-01</quality>
+              <le>9.2052650451660156e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1253</sample_count>
+          <value>7.8639624293829910e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>606</sample_count>
+          <value>-6.5938476875607277e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.8865798640253987e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0378650426864624e-01</quality>
+              <le>6.9388854503631592e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>533</sample_count>
+          <value>-2.7381819728273915e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1326</sample_count>
+          <value>2.8352394260546955e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.1102230246251577e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.0225412845611572e-01</quality>
+              <le>2.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1791</sample_count>
+          <value>-1.0029134530916375e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>68</sample_count>
+          <value>4.8411186971158215e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503178e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0205397605895996e-01</quality>
+              <le>2.3689103126525879e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1838</sample_count>
+          <value>-6.5406523866491815e-03</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>21</sample_count>
+          <value>7.3822507765214740e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.8865798640253987e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0392109155654907e-01</quality>
+              <le>2.2119045257568359e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1836</sample_count>
+          <value>1.2851614092359220e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>23</sample_count>
+          <value>-7.0930664915331199e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-1.7208456881689956e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0325584411621094e-01</quality>
+              <le>1.6267855167388916e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1811</sample_count>
+          <value>-1.4416639629038558e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>48</sample_count>
+          <value>4.8685109113945280e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0334036350250244e-01</quality>
+              <le>1.4522725343704224e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1776</sample_count>
+          <value>2.1134121503482612e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>83</sample_count>
+          <value>-3.2737254080837347e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0246149301528931e-01</quality>
+              <le>9.0829002857208252e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1219</sample_count>
+          <value>-8.5346673521825803e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>640</sample_count>
+          <value>5.7886290686531831e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0454872846603394e-01</quality>
+              <le>7.8601944446563721e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>774</sample_count>
+          <value>1.8066758954429543e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1085</sample_count>
+          <value>-5.0945701357667621e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>3.1086244689504288e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.0212800502777100e-01</quality>
+              <le>1.5322914123535156e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1796</sample_count>
+          <value>-1.4167158667728857e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>63</sample_count>
+          <value>3.0999551339753734e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-8.8817841970012602e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0266414880752563e-01</quality>
+              <le>1.0847599804401398e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>754</sample_count>
+          <value>2.2656461251734570e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1105</sample_count>
+          <value>-2.3970407856964623e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>8.8817841970012444e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0370788574218750e-01</quality>
+              <le>1.0789400339126587e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>742</sample_count>
+          <value>-3.1940419009583935e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1117</sample_count>
+          <value>2.3657134249111228e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0368273258209229e-01</quality>
+              <le>1.0177600383758545e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>590</sample_count>
+          <value>4.4494384607886871e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1269</sample_count>
+          <value>-1.6718599716109608e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0391924381256104e-01</quality>
+              <le>1.0132449865341187e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>577</sample_count>
+          <value>-5.2562531380568607e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1282</sample_count>
+          <value>1.5038797351826086e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.1649348980190600e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0355648994445801e-01</quality>
+              <le>9.2722505331039429e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>392</sample_count>
+          <value>7.2872834808621612e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1467</sample_count>
+          <value>-9.8094551697467572e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>6.6613381477509353e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0229591131210327e-01</quality>
+              <le>1.4143149554729462e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1571</sample_count>
+          <value>-1.1560090798064190e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>288</sample_count>
+          <value>4.2036013274394679e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0365710258483887e-01</quality>
+              <le>1.0963299870491028e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>781</sample_count>
+          <value>2.1235893234003624e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1078</sample_count>
+          <value>-3.5916730278930570e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>-2.2204460492503178e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0305408239364624e-01</quality>
+              <le>9.1055497527122498e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>376</sample_count>
+          <value>-7.5233897759775223e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1483</sample_count>
+          <value>8.1521687774440590e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>1859</sample_count>
+          <value>2.4424906541753385e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.0231134891510010e-01</quality>
+              <le>8.8118501007556915e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>322</sample_count>
+          <value>7.4755757692778257e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1537</sample_count>
+          <value>-6.2031478956482825e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_></trees></boost>
+</opencv_storage>
diff --git a/samples/cpp/trained_classifierNM2.xml b/samples/cpp/trained_classifierNM2.xml
new file mode 100644
index 0000000000..eae66f01ab
--- /dev/null
+++ b/samples/cpp/trained_classifierNM2.xml
@@ -0,0 +1,4046 @@
+<?xml version="1.0"?>
+<opencv_storage>
+<boost type_id="opencv-ml-boost-tree">
+  <boosting_type>RealAdaboost</boosting_type>
+  <splitting_criteria>Gini</splitting_criteria>
+  <ntrees>100</ntrees>
+  <weight_trimming_rate>1.</weight_trimming_rate>
+  <is_classifier>1</is_classifier>
+  <var_all>8</var_all>
+  <var_count>7</var_count>
+  <ord_var_count>7</ord_var_count>
+  <cat_var_count>0</cat_var_count>
+  <training_params>
+    <use_surrogates>0</use_surrogates>
+    <max_categories>10</max_categories>
+    <max_depth>1</max_depth>
+    <min_sample_count>10</min_sample_count>
+    <cross_validation_folds>0</cross_validation_folds>
+    <priors type_id="opencv-matrix">
+      <rows>1</rows>
+      <cols>2</cols>
+      <dt>d</dt>
+      <data>
+        1. 1.</data></priors></training_params>
+  <var_idx type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>7</cols>
+    <dt>i</dt>
+    <data>
+      1 2 3 4 5 6 7</data></var_idx>
+  <var_type>
+    0 0 0 0 0 0 0</var_type>
+  <cat_count type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>1</cols>
+    <dt>i</dt>
+    <data>
+      2</data></cat_count>
+  <cat_map type_id="opencv-matrix">
+    <rows>1</rows>
+    <cols>2</cols>
+    <dt>i</dt>
+    <data>
+      1 2</data></cat_map>
+  <trees>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.0078929470944956e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>8.7790262699127197e-01</quality>
+              <le>8.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2105</sample_count>
+          <value>-1.1353035616351272e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1110</sample_count>
+          <value>2.0526971992043412e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.0428103653102464e-14</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>6.9283843040466309e-01</quality>
+              <le>4.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1558</sample_count>
+          <value>-1.1814132639953403e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1657</sample_count>
+          <value>5.0475867936025332e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2259971643734884e-14</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>7.1280795335769653e-01</quality>
+              <le>1.2044999748468399e-03</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1475</sample_count>
+          <value>-7.9375129344950335e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1740</sample_count>
+          <value>7.6558743305188104e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.4424906541753385e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.9598809480667114e-01</quality>
+              <le>1.6245440244674683e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1385</sample_count>
+          <value>-5.4645841888834568e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1830</sample_count>
+          <value>4.0662339031566280e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>6.2172489379008380e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.5169683694839478e-01</quality>
+              <le>1.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2960</sample_count>
+          <value>-1.2615933276314473e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>255</sample_count>
+          <value>1.1688135468772842e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-8.4376949871512607e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.4812711477279663e-01</quality>
+              <le>5.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2891</sample_count>
+          <value>1.3088588195700304e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>324</sample_count>
+          <value>-9.4959634248367153e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.8865798640253987e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.9334284067153931e-01</quality>
+              <le>9.2769998311996460e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1964</sample_count>
+          <value>-4.5236837327316332e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1251</sample_count>
+          <value>4.7279968034956410e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-1.3322676295501896e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.6080043315887451e-01</quality>
+              <le>2.2512035369873047e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2464</sample_count>
+          <value>-2.1052010276554461e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>751</sample_count>
+          <value>6.7174941695746326e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-7.3274719625260868e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.3896796703338623e-01</quality>
+              <le>1.6267855167388916e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3041</sample_count>
+          <value>-8.1739798958620380e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>174</sample_count>
+          <value>1.8924547997011532e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>3.3306690738754586e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>5.3723198175430298e-01</quality>
+              <le>2.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>851</sample_count>
+          <value>-9.3661052654285037e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2364</sample_count>
+          <value>1.0184709367515142e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.2656542480726624e-14</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.4598015546798706e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1494</sample_count>
+          <value>3.5820791601148183e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1721</sample_count>
+          <value>-2.7428866405164359e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-1.3322676295501896e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.5166077613830566e-01</quality>
+              <le>2.9999999242136255e-05</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1467</sample_count>
+          <value>-3.3467424181887540e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1748</sample_count>
+          <value>3.3182965736372233e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.3322676295501861e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>5.4394942522048950e-01</quality>
+              <le>6.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1817</sample_count>
+          <value>-2.9822080188111177e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1398</sample_count>
+          <value>3.1329368107827943e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.6645352591003686e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.3358083963394165e-01</quality>
+              <le>1.3833449780941010e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2635</sample_count>
+          <value>8.0265603900023089e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>580</sample_count>
+          <value>-1.2162305066944930e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-6.4392935428259497e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.3098827600479126e-01</quality>
+              <le>1.2702149152755737e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2676</sample_count>
+          <value>1.3513810711631610e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>539</sample_count>
+          <value>-4.9909506904561546e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>5.3376436233520508e-01</quality>
+              <le>9.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2193</sample_count>
+          <value>1.3117433044170448e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1022</sample_count>
+          <value>-5.7326180584233111e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.7763568394002473e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.2698892354965210e-01</quality>
+              <le>2.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3127</sample_count>
+          <value>-5.4553392906591960e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>88</sample_count>
+          <value>2.6688100460888684e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-3.0531133177191900e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>5.2742463350296021e-01</quality>
+              <le>1.2500000000000000e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2620</sample_count>
+          <value>-5.4904310687255702e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>595</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.7368061883308035e-04</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.2537131309509277e-01</quality>
+              <le>3.7997097969055176e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3184</sample_count>
+          <value>-5.4291952000757461e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>31</sample_count>
+          <value>1.6789882514261603e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.9984014443252778e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.3548330068588257e-01</quality>
+              <le>6.6345453262329102e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>777</sample_count>
+          <value>-9.1576813577644156e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2438</sample_count>
+          <value>9.8351123166090529e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.2158576250076294e-01</quality>
+              <le>5.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2891</sample_count>
+          <value>8.1007865418959935e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>324</sample_count>
+          <value>-5.9585687406274601e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-6.6613381477509432e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.2358013391494751e-01</quality>
+              <le>1.0659420490264893e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2457</sample_count>
+          <value>-1.3607641362335321e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>758</sample_count>
+          <value>3.6397269406151606e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.2605414390563965e-01</quality>
+              <le>1.3934500217437744e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>909</sample_count>
+          <value>-4.4217410871376006e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2306</sample_count>
+          <value>1.2609161734369648e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.2353090047836304e-01</quality>
+              <le>1.2044999748468399e-03</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1475</sample_count>
+          <value>-2.0545012432588597e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1740</sample_count>
+          <value>2.3659867388791755e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-1.8318679906315118e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.2399057149887085e-01</quality>
+              <le>4.7306199073791504e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3204</sample_count>
+          <value>-4.8506775757358948e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>11</sample_count>
+          <value>2.6435685747751951e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503083e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>6</var>
+              <quality>5.2452313899993896e-01</quality>
+              <le>6.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1817</sample_count>
+          <value>-2.0830922304142732e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1398</sample_count>
+          <value>2.4354914812770342e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>3.7747582837255180e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>1</var>
+              <quality>5.1760864257812500e-01</quality>
+              <le>7.4752002954483032e-02</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>270</sample_count>
+          <value>7.4056582493755241e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2945</sample_count>
+          <value>-5.6004456612259952e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.5543122344752168e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.1595693826675415e-01</quality>
+              <le>5.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2891</sample_count>
+          <value>6.8929447086892887e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>324</sample_count>
+          <value>-5.0204800200277599e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-8.8262730457700718e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.1432991027832031e-01</quality>
+              <le>2.5954151153564453e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2981</sample_count>
+          <value>3.1556263241584975e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>234</sample_count>
+          <value>-1.5189501708336401e+00</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.1588785648345947e-01</quality>
+              <le>1.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2960</sample_count>
+          <value>-6.0429432019906144e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>255</sample_count>
+          <value>5.8525011139658878e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.4424906541753385e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.2001929283142090e-01</quality>
+              <le>1.3934500217437744e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>909</sample_count>
+          <value>-4.6709016373000739e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2306</sample_count>
+          <value>9.2123924914545782e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.6645352591003686e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.1656109094619751e-01</quality>
+              <le>1.2044999748468399e-03</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1475</sample_count>
+          <value>-1.6749125053548522e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1740</sample_count>
+          <value>2.0231653132989663e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-3.7747582837255464e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>3</var>
+              <quality>5.1900207996368408e-01</quality>
+              <le>3.</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1494</sample_count>
+          <value>2.5948749710951657e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1721</sample_count>
+          <value>-1.5086548637238759e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-1.3322676295501896e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>0</var>
+              <quality>5.1032912731170654e-01</quality>
+              <le>1.1417410373687744e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2628</sample_count>
+          <value>-6.3461321700044010e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>587</sample_count>
+          <value>3.3830370061028520e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.1102230246251554e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.1085340976715088e-01</quality>
+              <le>1.3021064996719360e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>734</sample_count>
+          <value>-3.7065768373801600e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2481</sample_count>
+          <value>6.1297280870936215e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>2</var>
+              <quality>5.1171976327896118e-01</quality>
+              <le>1.5000000000000000e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2960</sample_count>
+          <value>-5.1679907432686661e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>255</sample_count>
+          <value>4.8967202182224279e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-3.3306690738754807e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.1276576519012451e-01</quality>
+              <le>1.4958450198173523e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2755</sample_count>
+          <value>7.7575074151822049e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>460</sample_count>
+          <value>-3.4258234102830631e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.0944852828979492e-01</quality>
+              <le>2.5842750072479248e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2980</sample_count>
+          <value>-2.8188779664253410e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>235</sample_count>
+          <value>8.1175021374635614e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-5.2735593669695212e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.1038962602615356e-01</quality>
+              <le>1.4510664939880371e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1007</sample_count>
+          <value>-3.3150262977164136e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2208</sample_count>
+          <value>6.5052948330937457e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>1.3322676295501861e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.1113504171371460e-01</quality>
+              <le>2.4550000671297312e-04</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1471</sample_count>
+          <value>-1.2860595713867043e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1744</sample_count>
+          <value>1.7591144586528801e-01</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-8.8817841970012602e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>4</var>
+              <quality>5.1135826110839844e-01</quality>
+              <le>1.9797500967979431e-01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>2857</sample_count>
+          <value>4.5695890739118755e-02</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>358</sample_count>
+          <value>-5.4593965595422900e-01</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.8865798640253987e-15</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.0861865282058716e-01</quality>
+              <le>4.4535398483276367e+00</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3204</sample_count>
+          <value>-1.7238960071156845e-02</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>11</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>5.4514379135116734e-05</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.1613724231719971e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>5.4514379101591658e-05</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-4.2188474935756130e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>5.8550304174423218e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-4.2188474935756130e-15</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.1834394931793213e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2149631977081299e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2519973516464233e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3510566949844360e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3096535205841064e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.0966378450393677e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3168644905090332e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3053596019744873e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.1062115430831909e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2805336713790894e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3325423002243042e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.0798364877700806e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2765526771545410e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3239699602127075e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.0819625854492188e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2792611122131348e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3205927610397339e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.0529828071594238e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2831163406372070e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3197416067123413e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.0644149780273438e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2831127643585205e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2990629673004150e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.1229735612869263e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2957435846328735e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2827998399734497e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.1507642269134521e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3906562328338623e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>4.4408920985006242e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-4.4408920985006281e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2384939193725586e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-4.4408920985006281e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3710451126098633e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.3055968284606934e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>0.</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2085735797882080e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>-2.2204460492503136e-16</value>
+          <norm_class_idx>0</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_>
+    <_>
+      <best_tree_idx>-1</best_tree_idx>
+      <nodes>
+        <_>
+          <depth>0</depth>
+          <sample_count>3215</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error>
+          <splits>
+            <_><var>5</var>
+              <quality>6.2932032346725464e-01</quality>
+              <le>1.3783290863037109e+01</le></_></splits></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>3214</sample_count>
+          <value>2.2204460492503126e-16</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_>
+        <_>
+          <depth>1</depth>
+          <sample_count>1</sample_count>
+          <value>5.7564577324623896e+00</value>
+          <norm_class_idx>1</norm_class_idx>
+          <Tn>0</Tn>
+          <complexity>0</complexity>
+          <alpha>0.</alpha>
+          <node_risk>0.</node_risk>
+          <tree_risk>0.</tree_risk>
+          <tree_error>0.</tree_error></_></nodes></_></trees></boost>
+</opencv_storage>
diff --git a/samples/cpp/tutorial_code/HighGUI/video-input-psnr-ssim/video-input-psnr-ssim.cpp b/samples/cpp/tutorial_code/HighGUI/video-input-psnr-ssim/video-input-psnr-ssim.cpp
index 56f6e7dfa1..bee1c5b200 100644
--- a/samples/cpp/tutorial_code/HighGUI/video-input-psnr-ssim/video-input-psnr-ssim.cpp
+++ b/samples/cpp/tutorial_code/HighGUI/video-input-psnr-ssim/video-input-psnr-ssim.cpp
@@ -60,10 +60,10 @@ int main(int argc, char *argv[])
         return -1;
     }
 
-    Size refS = Size((int) captRefrnc.get(CV_CAP_PROP_FRAME_WIDTH),
-                     (int) captRefrnc.get(CV_CAP_PROP_FRAME_HEIGHT)),
-         uTSi = Size((int) captUndTst.get(CV_CAP_PROP_FRAME_WIDTH),
-                     (int) captUndTst.get(CV_CAP_PROP_FRAME_HEIGHT));
+    Size refS = Size((int) captRefrnc.get(CAP_PROP_FRAME_WIDTH),
+                     (int) captRefrnc.get(CAP_PROP_FRAME_HEIGHT)),
+         uTSi = Size((int) captUndTst.get(CAP_PROP_FRAME_WIDTH),
+                     (int) captUndTst.get(CAP_PROP_FRAME_HEIGHT));
 
     if (refS != uTSi)
     {
@@ -75,13 +75,13 @@ int main(int argc, char *argv[])
     const char* WIN_RF = "Reference";
 
     // Windows
-    namedWindow(WIN_RF, CV_WINDOW_AUTOSIZE);
-    namedWindow(WIN_UT, CV_WINDOW_AUTOSIZE);
-    cvMoveWindow(WIN_RF, 400       , 0);         //750,  2 (bernat =0)
-    cvMoveWindow(WIN_UT, refS.width, 0);         //1500, 2
+    namedWindow(WIN_RF, WINDOW_AUTOSIZE);
+    namedWindow(WIN_UT, WINDOW_AUTOSIZE);
+    moveWindow(WIN_RF, 400       , 0);         //750,  2 (bernat =0)
+    moveWindow(WIN_UT, refS.width, 0);         //1500, 2
 
     cout << "Reference frame resolution: Width=" << refS.width << "  Height=" << refS.height
-         << " of nr#: " << captRefrnc.get(CV_CAP_PROP_FRAME_COUNT) << endl;
+         << " of nr#: " << captRefrnc.get(CAP_PROP_FRAME_COUNT) << endl;
 
     cout << "PSNR trigger value " << setiosflags(ios::fixed) << setprecision(3)
          << psnrTriggerValue << endl;
@@ -125,7 +125,7 @@ int main(int argc, char *argv[])
         imshow(WIN_RF, frameReference);
         imshow(WIN_UT, frameUnderTest);
 
-        c = (char)cvWaitKey(delay);
+        c = (char)waitKey(delay);
         if (c == 27) break;
     }
 
diff --git a/samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp b/samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp
index 8ee99a10aa..d3b8e44303 100644
--- a/samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp
+++ b/samples/cpp/tutorial_code/HighGUI/video-write/video-write.cpp
@@ -41,19 +41,19 @@ int main(int argc, char *argv[])
 
     string::size_type pAt = source.find_last_of('.');                  // Find extension point
     const string NAME = source.substr(0, pAt) + argv[2][0] + ".avi";   // Form the new name with container
-    int ex = static_cast<int>(inputVideo.get(CV_CAP_PROP_FOURCC));     // Get Codec Type- Int form
+    int ex = static_cast<int>(inputVideo.get(CAP_PROP_FOURCC));     // Get Codec Type- Int form
 
     // Transform from int to char via Bitwise operators
     char EXT[] = {(char)(ex & 0XFF) , (char)((ex & 0XFF00) >> 8),(char)((ex & 0XFF0000) >> 16),(char)((ex & 0XFF000000) >> 24), 0};
 
-    Size S = Size((int) inputVideo.get(CV_CAP_PROP_FRAME_WIDTH),    // Acquire input size
-                  (int) inputVideo.get(CV_CAP_PROP_FRAME_HEIGHT));
+    Size S = Size((int) inputVideo.get(CAP_PROP_FRAME_WIDTH),    // Acquire input size
+                  (int) inputVideo.get(CAP_PROP_FRAME_HEIGHT));
 
     VideoWriter outputVideo;                                        // Open the output
     if (askOutputType)
-        outputVideo.open(NAME, ex=-1, inputVideo.get(CV_CAP_PROP_FPS), S, true);
+        outputVideo.open(NAME, ex=-1, inputVideo.get(CAP_PROP_FPS), S, true);
     else
-        outputVideo.open(NAME, ex, inputVideo.get(CV_CAP_PROP_FPS), S, true);
+        outputVideo.open(NAME, ex, inputVideo.get(CAP_PROP_FPS), S, true);
 
     if (!outputVideo.isOpened())
     {
@@ -62,7 +62,7 @@ int main(int argc, char *argv[])
     }
 
     cout << "Input frame resolution: Width=" << S.width << "  Height=" << S.height
-         << " of nr#: " << inputVideo.get(CV_CAP_PROP_FRAME_COUNT) << endl;
+         << " of nr#: " << inputVideo.get(CAP_PROP_FRAME_COUNT) << endl;
     cout << "Input codec type: " << EXT << endl;
 
     int channel = 2; // Select the channel to save
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/EqualizeHist_Demo.cpp b/samples/cpp/tutorial_code/Histograms_Matching/EqualizeHist_Demo.cpp
index 49e4be0e54..38a5839bb2 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/EqualizeHist_Demo.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/EqualizeHist_Demo.cpp
@@ -31,14 +31,14 @@ int main( int, char** argv )
     }
 
   /// Convert to grayscale
-  cvtColor( src, src, CV_BGR2GRAY );
+  cvtColor( src, src, COLOR_BGR2GRAY );
 
   /// Apply Histogram Equalization
   equalizeHist( src, dst );
 
   /// Display results
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
-  namedWindow( equalized_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
+  namedWindow( equalized_window, WINDOW_AUTOSIZE );
 
   imshow( source_window, src );
   imshow( equalized_window, dst );
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/MatchTemplate_Demo.cpp b/samples/cpp/tutorial_code/Histograms_Matching/MatchTemplate_Demo.cpp
index e3633576f8..b41cf94be5 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/MatchTemplate_Demo.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/MatchTemplate_Demo.cpp
@@ -33,8 +33,8 @@ int main( int, char** argv )
   templ = imread( argv[2], 1 );
 
   /// Create windows
-  namedWindow( image_window, CV_WINDOW_AUTOSIZE );
-  namedWindow( result_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( image_window, WINDOW_AUTOSIZE );
+  namedWindow( result_window, WINDOW_AUTOSIZE );
 
   /// Create Trackbar
   const char* trackbar_label = "Method: \n 0: SQDIFF \n 1: SQDIFF NORMED \n 2: TM CCORR \n 3: TM CCORR NORMED \n 4: TM COEFF \n 5: TM COEFF NORMED";
@@ -74,7 +74,7 @@ void MatchingMethod( int, void* )
 
 
   /// For SQDIFF and SQDIFF_NORMED, the best matches are lower values. For all the other methods, the higher the better
-  if( match_method  == CV_TM_SQDIFF || match_method == CV_TM_SQDIFF_NORMED )
+  if( match_method  == TM_SQDIFF || match_method == TM_SQDIFF_NORMED )
     { matchLoc = minLoc; }
   else
     { matchLoc = maxLoc; }
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo1.cpp b/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo1.cpp
index f422603749..86d2f2e154 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo1.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo1.cpp
@@ -28,7 +28,7 @@ int main( int, char** argv )
   /// Read the image
   src = imread( argv[1], 1 );
   /// Transform it to HSV
-  cvtColor( src, hsv, CV_BGR2HSV );
+  cvtColor( src, hsv, COLOR_BGR2HSV );
 
   /// Use only the Hue value
   hue.create( hsv.size(), hsv.depth() );
@@ -37,7 +37,7 @@ int main( int, char** argv )
 
   /// Create Trackbar to enter the number of bins
   const char* window_image = "Source image";
-  namedWindow( window_image, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_image, WINDOW_AUTOSIZE );
   createTrackbar("* Hue  bins: ", window_image, &bins, 180, Hist_and_Backproj );
   Hist_and_Backproj(0, 0);
 
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo2.cpp b/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo2.cpp
index 42dd01a671..28181c2d31 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo2.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/calcBackProject_Demo2.cpp
@@ -31,10 +31,10 @@ int main( int, char** argv )
   /// Read the image
   src = imread( argv[1], 1 );
   /// Transform it to HSV
-  cvtColor( src, hsv, CV_BGR2HSV );
+  cvtColor( src, hsv, COLOR_BGR2HSV );
 
   /// Show the image
-  namedWindow( window_image, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_image, WINDOW_AUTOSIZE );
   imshow( window_image, src );
 
   /// Set Trackbars for floodfill thresholds
@@ -52,7 +52,7 @@ int main( int, char** argv )
  */
 void pickPoint (int event, int x, int y, int, void* )
 {
-  if( event != CV_EVENT_LBUTTONDOWN )
+  if( event != EVENT_LBUTTONDOWN )
     { return; }
 
   // Fill and get the mask
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/calcHist_Demo.cpp b/samples/cpp/tutorial_code/Histograms_Matching/calcHist_Demo.cpp
index d3af5e7a96..577c8a8b9c 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/calcHist_Demo.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/calcHist_Demo.cpp
@@ -71,7 +71,7 @@ int main( int, char** argv )
   }
 
   /// Display
-  namedWindow("calcHist Demo", CV_WINDOW_AUTOSIZE );
+  namedWindow("calcHist Demo", WINDOW_AUTOSIZE );
   imshow("calcHist Demo", histImage );
 
   waitKey(0);
diff --git a/samples/cpp/tutorial_code/Histograms_Matching/compareHist_Demo.cpp b/samples/cpp/tutorial_code/Histograms_Matching/compareHist_Demo.cpp
index b18aab2e8b..109f8131a3 100644
--- a/samples/cpp/tutorial_code/Histograms_Matching/compareHist_Demo.cpp
+++ b/samples/cpp/tutorial_code/Histograms_Matching/compareHist_Demo.cpp
@@ -33,9 +33,9 @@ int main( int argc, char** argv )
    src_test2 = imread( argv[3], 1 );
 
    /// Convert to HSV
-   cvtColor( src_base, hsv_base, CV_BGR2HSV );
-   cvtColor( src_test1, hsv_test1, CV_BGR2HSV );
-   cvtColor( src_test2, hsv_test2, CV_BGR2HSV );
+   cvtColor( src_base, hsv_base, COLOR_BGR2HSV );
+   cvtColor( src_test1, hsv_test1, COLOR_BGR2HSV );
+   cvtColor( src_test2, hsv_test2, COLOR_BGR2HSV );
 
    hsv_half_down = hsv_base( Range( hsv_base.rows/2, hsv_base.rows - 1 ), Range( 0, hsv_base.cols - 1 ) );
 
diff --git a/samples/cpp/tutorial_code/ImgProc/Morphology_1.cpp b/samples/cpp/tutorial_code/ImgProc/Morphology_1.cpp
index 47030fa4f4..13a96a1f51 100644
--- a/samples/cpp/tutorial_code/ImgProc/Morphology_1.cpp
+++ b/samples/cpp/tutorial_code/ImgProc/Morphology_1.cpp
@@ -37,9 +37,9 @@ int main( int, char** argv )
     { return -1; }
 
   /// Create windows
-  namedWindow( "Erosion Demo", CV_WINDOW_AUTOSIZE );
-  namedWindow( "Dilation Demo", CV_WINDOW_AUTOSIZE );
-  cvMoveWindow( "Dilation Demo", src.cols, 0 );
+  namedWindow( "Erosion Demo", WINDOW_AUTOSIZE );
+  namedWindow( "Dilation Demo", WINDOW_AUTOSIZE );
+  moveWindow( "Dilation Demo", src.cols, 0 );
 
   /// Create Erosion Trackbar
   createTrackbar( "Element:\n 0: Rect \n 1: Cross \n 2: Ellipse", "Erosion Demo",
diff --git a/samples/cpp/tutorial_code/ImgProc/Morphology_2.cpp b/samples/cpp/tutorial_code/ImgProc/Morphology_2.cpp
index c501bfd328..67147b3a04 100644
--- a/samples/cpp/tutorial_code/ImgProc/Morphology_2.cpp
+++ b/samples/cpp/tutorial_code/ImgProc/Morphology_2.cpp
@@ -39,7 +39,7 @@ int main( int, char** argv )
     { return -1; }
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Create Trackbar to select Morphology operation
   createTrackbar("Operator:\n 0: Opening - 1: Closing  \n 2: Gradient - 3: Top Hat \n 4: Black Hat", window_name, &morph_operator, max_operator, Morphology_Operations );
@@ -76,5 +76,3 @@ void Morphology_Operations( int, void* )
   morphologyEx( src, dst, operation, element );
   imshow( window_name, dst );
 }
-
-
diff --git a/samples/cpp/tutorial_code/ImgProc/Pyramids.cpp b/samples/cpp/tutorial_code/ImgProc/Pyramids.cpp
index d0cae3c4f6..fc98d1c21a 100644
--- a/samples/cpp/tutorial_code/ImgProc/Pyramids.cpp
+++ b/samples/cpp/tutorial_code/ImgProc/Pyramids.cpp
@@ -40,7 +40,7 @@ int main( void )
   dst = tmp;
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
   imshow( window_name, dst );
 
   /// Loop
@@ -66,10 +66,3 @@ int main( void )
 
    return 0;
 }
-
-
-
-
-
-
-
diff --git a/samples/cpp/tutorial_code/ImgProc/Smoothing.cpp b/samples/cpp/tutorial_code/ImgProc/Smoothing.cpp
index 5f51d07d29..8513bcf76a 100644
--- a/samples/cpp/tutorial_code/ImgProc/Smoothing.cpp
+++ b/samples/cpp/tutorial_code/ImgProc/Smoothing.cpp
@@ -31,7 +31,7 @@ int display_dst( int delay );
  */
 int main( void )
 {
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Load the source image
   src = imread( "../images/lena.png", 1 );
@@ -89,7 +89,7 @@ int display_caption( const char* caption )
   dst = Mat::zeros( src.size(), src.type() );
   putText( dst, caption,
            Point( src.cols/4, src.rows/2),
-           CV_FONT_HERSHEY_COMPLEX, 1, Scalar(255, 255, 255) );
+           FONT_HERSHEY_COMPLEX, 1, Scalar(255, 255, 255) );
 
   imshow( window_name, dst );
   int c = waitKey( DELAY_CAPTION );
diff --git a/samples/cpp/tutorial_code/ImgProc/Threshold.cpp b/samples/cpp/tutorial_code/ImgProc/Threshold.cpp
index 7505ec297b..1dff6c86ed 100644
--- a/samples/cpp/tutorial_code/ImgProc/Threshold.cpp
+++ b/samples/cpp/tutorial_code/ImgProc/Threshold.cpp
@@ -37,10 +37,10 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert the image to Gray
-  cvtColor( src, src_gray, CV_RGB2GRAY );
+  cvtColor( src, src_gray, COLOR_RGB2GRAY );
 
   /// Create a window to display results
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Create Trackbar to choose type of Threshold
   createTrackbar( trackbar_type,
diff --git a/samples/cpp/tutorial_code/ImgTrans/CannyDetector_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/CannyDetector_Demo.cpp
index c798f2fb42..7851c9f58b 100644
--- a/samples/cpp/tutorial_code/ImgTrans/CannyDetector_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/CannyDetector_Demo.cpp
@@ -58,10 +58,10 @@ int main( int, char** argv )
   dst.create( src.size(), src.type() );
 
   /// Convert the image to grayscale
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
   /// Create a window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Create a Trackbar for user to enter threshold
   createTrackbar( "Min Threshold:", window_name, &lowThreshold, max_lowThreshold, CannyThreshold );
diff --git a/samples/cpp/tutorial_code/ImgTrans/Geometric_Transforms_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/Geometric_Transforms_Demo.cpp
index 2d18357202..0a2a2f5fa4 100644
--- a/samples/cpp/tutorial_code/ImgTrans/Geometric_Transforms_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/Geometric_Transforms_Demo.cpp
@@ -65,13 +65,13 @@ int main( int, char** argv )
 
 
   /// Show what you got
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
-  namedWindow( warp_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( warp_window, WINDOW_AUTOSIZE );
   imshow( warp_window, warp_dst );
 
-  namedWindow( warp_rotate_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( warp_rotate_window, WINDOW_AUTOSIZE );
   imshow( warp_rotate_window, warp_rotate_dst );
 
   /// Wait until user exits the program
diff --git a/samples/cpp/tutorial_code/ImgTrans/HoughCircle_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/HoughCircle_Demo.cpp
index 5782e92a3c..b4cda82f1b 100644
--- a/samples/cpp/tutorial_code/ImgTrans/HoughCircle_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/HoughCircle_Demo.cpp
@@ -26,7 +26,7 @@ int main(int, char** argv)
      { return -1; }
 
    /// Convert it to gray
-    cvtColor( src, src_gray, CV_BGR2GRAY );
+    cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
    /// Reduce the noise so we avoid false circle detection
     GaussianBlur( src_gray, src_gray, Size(9, 9), 2, 2 );
@@ -34,7 +34,7 @@ int main(int, char** argv)
     vector<Vec3f> circles;
 
    /// Apply the Hough Transform to find the circles
-    HoughCircles( src_gray, circles, CV_HOUGH_GRADIENT, 1, src_gray.rows/8, 200, 100, 0, 0 );
+    HoughCircles( src_gray, circles, HOUGH_GRADIENT, 1, src_gray.rows/8, 200, 100, 0, 0 );
 
    /// Draw the circles detected
     for( size_t i = 0; i < circles.size(); i++ )
@@ -48,7 +48,7 @@ int main(int, char** argv)
     }
 
    /// Show your results
-    namedWindow( "Hough Circle Transform Demo", CV_WINDOW_AUTOSIZE );
+    namedWindow( "Hough Circle Transform Demo", WINDOW_AUTOSIZE );
     imshow( "Hough Circle Transform Demo", src );
 
     waitKey(0);
diff --git a/samples/cpp/tutorial_code/ImgTrans/HoughLines_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/HoughLines_Demo.cpp
index 561948a581..6400c43794 100644
--- a/samples/cpp/tutorial_code/ImgTrans/HoughLines_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/HoughLines_Demo.cpp
@@ -46,7 +46,7 @@ int main( int, char** argv )
      }
 
    /// Pass the image to gray
-   cvtColor( src, src_gray, CV_RGB2GRAY );
+   cvtColor( src, src_gray, COLOR_RGB2GRAY );
 
    /// Apply Canny edge detector
    Canny( src_gray, edges, 50, 200, 3 );
@@ -55,10 +55,10 @@ int main( int, char** argv )
    char thresh_label[50];
    sprintf( thresh_label, "Thres: %d + input", min_threshold );
 
-   namedWindow( standard_name, CV_WINDOW_AUTOSIZE );
+   namedWindow( standard_name, WINDOW_AUTOSIZE );
    createTrackbar( thresh_label, standard_name, &s_trackbar, max_trackbar, Standard_Hough);
 
-   namedWindow( probabilistic_name, CV_WINDOW_AUTOSIZE );
+   namedWindow( probabilistic_name, WINDOW_AUTOSIZE );
    createTrackbar( thresh_label, probabilistic_name, &p_trackbar, max_trackbar, Probabilistic_Hough);
 
    /// Initialize
@@ -85,7 +85,7 @@ void help()
 void Standard_Hough( int, void* )
 {
   vector<Vec2f> s_lines;
-  cvtColor( edges, standard_hough, CV_GRAY2BGR );
+  cvtColor( edges, standard_hough, COLOR_GRAY2BGR );
 
   /// 1. Use Standard Hough Transform
   HoughLines( edges, s_lines, 1, CV_PI/180, min_threshold + s_trackbar, 0, 0 );
@@ -100,7 +100,7 @@ void Standard_Hough( int, void* )
 
        Point pt1( cvRound(x0 + alpha*(-sin_t)), cvRound(y0 + alpha*cos_t) );
        Point pt2( cvRound(x0 - alpha*(-sin_t)), cvRound(y0 - alpha*cos_t) );
-       line( standard_hough, pt1, pt2, Scalar(255,0,0), 3, CV_AA);
+       line( standard_hough, pt1, pt2, Scalar(255,0,0), 3, LINE_AA);
      }
 
    imshow( standard_name, standard_hough );
@@ -112,7 +112,7 @@ void Standard_Hough( int, void* )
 void Probabilistic_Hough( int, void* )
 {
   vector<Vec4i> p_lines;
-  cvtColor( edges, probabilistic_hough, CV_GRAY2BGR );
+  cvtColor( edges, probabilistic_hough, COLOR_GRAY2BGR );
 
   /// 2. Use Probabilistic Hough Transform
   HoughLinesP( edges, p_lines, 1, CV_PI/180, min_threshold + p_trackbar, 30, 10 );
@@ -121,7 +121,7 @@ void Probabilistic_Hough( int, void* )
   for( size_t i = 0; i < p_lines.size(); i++ )
      {
        Vec4i l = p_lines[i];
-       line( probabilistic_hough, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(255,0,0), 3, CV_AA);
+       line( probabilistic_hough, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(255,0,0), 3, LINE_AA);
      }
 
    imshow( probabilistic_name, probabilistic_hough );
diff --git a/samples/cpp/tutorial_code/ImgTrans/Laplace_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/Laplace_Demo.cpp
index 279dc6dcb6..f6dff102d8 100644
--- a/samples/cpp/tutorial_code/ImgTrans/Laplace_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/Laplace_Demo.cpp
@@ -34,10 +34,10 @@ int main( int, char** argv )
   GaussianBlur( src, src, Size(3,3), 0, 0, BORDER_DEFAULT );
 
   /// Convert the image to grayscale
-  cvtColor( src, src_gray, CV_RGB2GRAY );
+  cvtColor( src, src_gray, COLOR_RGB2GRAY );
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Apply Laplace function
   Mat abs_dst;
diff --git a/samples/cpp/tutorial_code/ImgTrans/Remap_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/Remap_Demo.cpp
index 40e5118a6d..86b6a2cb60 100644
--- a/samples/cpp/tutorial_code/ImgTrans/Remap_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/Remap_Demo.cpp
@@ -34,7 +34,7 @@ int main( int, char** argv )
   map_y.create( src.size(), CV_32FC1 );
 
   /// Create window
-  namedWindow( remap_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( remap_window, WINDOW_AUTOSIZE );
 
   /// Loop
   for(;;)
@@ -47,7 +47,7 @@ int main( int, char** argv )
 
     /// Update map_x & map_y. Then apply remap
     update_map();
-    remap( src, dst, map_x, map_y, CV_INTER_LINEAR, BORDER_CONSTANT, Scalar(0, 0, 0) );
+    remap( src, dst, map_x, map_y, INTER_LINEAR, BORDER_CONSTANT, Scalar(0, 0, 0) );
 
     // Display results
     imshow( remap_window, dst );
diff --git a/samples/cpp/tutorial_code/ImgTrans/Sobel_Demo.cpp b/samples/cpp/tutorial_code/ImgTrans/Sobel_Demo.cpp
index 0ac9eb1a30..f8c97c4117 100644
--- a/samples/cpp/tutorial_code/ImgTrans/Sobel_Demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/Sobel_Demo.cpp
@@ -33,10 +33,10 @@ int main( int, char** argv )
   GaussianBlur( src, src, Size(3,3), 0, 0, BORDER_DEFAULT );
 
   /// Convert it to gray
-  cvtColor( src, src_gray, CV_RGB2GRAY );
+  cvtColor( src, src_gray, COLOR_RGB2GRAY );
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Generate grad_x and grad_y
   Mat grad_x, grad_y;
@@ -61,5 +61,3 @@ int main( int, char** argv )
 
   return 0;
 }
-
-
diff --git a/samples/cpp/tutorial_code/ImgTrans/copyMakeBorder_demo.cpp b/samples/cpp/tutorial_code/ImgTrans/copyMakeBorder_demo.cpp
index 4b2783e31c..0a441ccdfc 100644
--- a/samples/cpp/tutorial_code/ImgTrans/copyMakeBorder_demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/copyMakeBorder_demo.cpp
@@ -15,7 +15,6 @@ using namespace cv;
 Mat src, dst;
 int top, bottom, left, right;
 int borderType;
-Scalar value;
 const char* window_name = "copyMakeBorder Demo";
 RNG rng(12345);
 
@@ -44,7 +43,7 @@ int main( int, char** argv )
   printf( " ** Press 'ESC' to exit the program \n");
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Initialize arguments for the filter
   top = (int) (0.05*src.rows); bottom = (int) (0.05*src.rows);
@@ -64,7 +63,7 @@ int main( int, char** argv )
          else if( (char)c == 'r' )
            { borderType = BORDER_REPLICATE; }
 
-         value = Scalar( rng.uniform(0, 255), rng.uniform(0, 255), rng.uniform(0, 255) );
+         Scalar value( rng.uniform(0, 255), rng.uniform(0, 255), rng.uniform(0, 255) );
          copyMakeBorder( src, dst, top, bottom, left, right, borderType, value );
 
          imshow( window_name, dst );
@@ -72,5 +71,3 @@ int main( int, char** argv )
 
   return 0;
 }
-
-
diff --git a/samples/cpp/tutorial_code/ImgTrans/filter2D_demo.cpp b/samples/cpp/tutorial_code/ImgTrans/filter2D_demo.cpp
index 7a8dc768cf..86db4d8ed4 100644
--- a/samples/cpp/tutorial_code/ImgTrans/filter2D_demo.cpp
+++ b/samples/cpp/tutorial_code/ImgTrans/filter2D_demo.cpp
@@ -35,7 +35,7 @@ int main ( int, char** argv )
     { return -1; }
 
   /// Create window
-  namedWindow( window_name, CV_WINDOW_AUTOSIZE );
+  namedWindow( window_name, WINDOW_AUTOSIZE );
 
   /// Initialize arguments for the filter
   anchor = Point( -1, -1 );
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/findContours_demo.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/findContours_demo.cpp
index e301476ee7..5eb2d92a7e 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/findContours_demo.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/findContours_demo.cpp
@@ -30,12 +30,12 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert image to gray and blur it
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
   blur( src_gray, src_gray, Size(3,3) );
 
   /// Create Window
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
   createTrackbar( " Canny thresh:", "Source", &thresh, max_thresh, thresh_callback );
@@ -57,7 +57,7 @@ void thresh_callback(int, void* )
   /// Detect edges using canny
   Canny( src_gray, canny_output, thresh, thresh*2, 3 );
   /// Find contours
-  findContours( canny_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+  findContours( canny_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
 
   /// Draw contours
   Mat drawing = Mat::zeros( canny_output.size(), CV_8UC3 );
@@ -68,6 +68,6 @@ void thresh_callback(int, void* )
      }
 
   /// Show in a window
-  namedWindow( "Contours", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Contours", WINDOW_AUTOSIZE );
   imshow( "Contours", drawing );
 }
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo1.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo1.cpp
index b973cfd973..d481d03898 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo1.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo1.cpp
@@ -30,12 +30,12 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert image to gray and blur it
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
   blur( src_gray, src_gray, Size(3,3) );
 
   /// Create Window
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
   createTrackbar( " Threshold:", "Source", &thresh, max_thresh, thresh_callback );
@@ -57,7 +57,7 @@ void thresh_callback(int, void* )
   /// Detect edges using Threshold
   threshold( src_gray, threshold_output, thresh, 255, THRESH_BINARY );
   /// Find contours
-  findContours( threshold_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+  findContours( threshold_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
 
   /// Approximate contours to polygons + get bounding rects and circles
   vector<vector<Point> > contours_poly( contours.size() );
@@ -83,6 +83,6 @@ void thresh_callback(int, void* )
      }
 
   /// Show in a window
-  namedWindow( "Contours", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Contours", WINDOW_AUTOSIZE );
   imshow( "Contours", drawing );
 }
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo2.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo2.cpp
index 70d8663c9b..e1f44712b7 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo2.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/generalContours_demo2.cpp
@@ -30,12 +30,12 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert image to gray and blur it
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
   blur( src_gray, src_gray, Size(3,3) );
 
   /// Create Window
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
   createTrackbar( " Threshold:", "Source", &thresh, max_thresh, thresh_callback );
@@ -57,7 +57,7 @@ void thresh_callback(int, void* )
   /// Detect edges using Threshold
   threshold( src_gray, threshold_output, thresh, 255, THRESH_BINARY );
   /// Find contours
-  findContours( threshold_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+  findContours( threshold_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
 
   /// Find the rotated rectangles and ellipses for each contour
   vector<RotatedRect> minRect( contours.size() );
@@ -85,6 +85,6 @@ void thresh_callback(int, void* )
      }
 
   /// Show in a window
-  namedWindow( "Contours", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Contours", WINDOW_AUTOSIZE );
   imshow( "Contours", drawing );
 }
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/hull_demo.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/hull_demo.cpp
index 8fe5d5b075..395b4b39f0 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/hull_demo.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/hull_demo.cpp
@@ -30,12 +30,12 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert image to gray and blur it
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
   blur( src_gray, src_gray, Size(3,3) );
 
   /// Create Window
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
   createTrackbar( " Threshold:", "Source", &thresh, max_thresh, thresh_callback );
@@ -59,7 +59,7 @@ void thresh_callback(int, void* )
   threshold( src_gray, threshold_output, thresh, 255, THRESH_BINARY );
 
   /// Find contours
-  findContours( threshold_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+  findContours( threshold_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
 
   /// Find the convex hull object for each contour
  vector<vector<Point> >hull( contours.size() );
@@ -76,6 +76,6 @@ void thresh_callback(int, void* )
      }
 
   /// Show in a window
-  namedWindow( "Hull demo", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Hull demo", WINDOW_AUTOSIZE );
   imshow( "Hull demo", drawing );
 }
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/moments_demo.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/moments_demo.cpp
index efd3c63d9e..9ca230f875 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/moments_demo.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/moments_demo.cpp
@@ -30,12 +30,12 @@ int main( int, char** argv )
   src = imread( argv[1], 1 );
 
   /// Convert image to gray and blur it
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
   blur( src_gray, src_gray, Size(3,3) );
 
   /// Create Window
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
 
   createTrackbar( " Canny thresh:", "Source", &thresh, max_thresh, thresh_callback );
@@ -57,7 +57,7 @@ void thresh_callback(int, void* )
   /// Detect edges using canny
   Canny( src_gray, canny_output, thresh, thresh*2, 3 );
   /// Find contours
-  findContours( canny_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, Point(0, 0) );
+  findContours( canny_output, contours, hierarchy, RETR_TREE, CHAIN_APPROX_SIMPLE, Point(0, 0) );
 
   /// Get the moments
   vector<Moments> mu(contours.size() );
@@ -79,7 +79,7 @@ void thresh_callback(int, void* )
      }
 
   /// Show in a window
-  namedWindow( "Contours", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Contours", WINDOW_AUTOSIZE );
   imshow( "Contours", drawing );
 
   /// Calculate the area with the moments 00 and compare with the result of the OpenCV function
@@ -92,4 +92,3 @@ void thresh_callback(int, void* )
        circle( drawing, mc[i], 4, color, -1, 8, 0 );
      }
 }
-
diff --git a/samples/cpp/tutorial_code/ShapeDescriptors/pointPolygonTest_demo.cpp b/samples/cpp/tutorial_code/ShapeDescriptors/pointPolygonTest_demo.cpp
index c15bc22c6f..f55f8f6879 100644
--- a/samples/cpp/tutorial_code/ShapeDescriptors/pointPolygonTest_demo.cpp
+++ b/samples/cpp/tutorial_code/ShapeDescriptors/pointPolygonTest_demo.cpp
@@ -71,13 +71,11 @@ int main( void )
 
   /// Create Window and show your results
   const char* source_window = "Source";
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, src );
-  namedWindow( "Distance", CV_WINDOW_AUTOSIZE );
+  namedWindow( "Distance", WINDOW_AUTOSIZE );
   imshow( "Distance", drawing );
 
   waitKey(0);
   return(0);
 }
-
-
diff --git a/samples/cpp/tutorial_code/TrackingMotion/cornerDetector_Demo.cpp b/samples/cpp/tutorial_code/TrackingMotion/cornerDetector_Demo.cpp
index b4789e3345..31c7004787 100644
--- a/samples/cpp/tutorial_code/TrackingMotion/cornerDetector_Demo.cpp
+++ b/samples/cpp/tutorial_code/TrackingMotion/cornerDetector_Demo.cpp
@@ -40,7 +40,7 @@ int main( int, char** argv )
 {
   /// Load source image and convert it to gray
   src = imread( argv[1], 1 );
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
   /// Set some parameters
   int blockSize = 3; int apertureSize = 3;
@@ -64,7 +64,7 @@ int main( int, char** argv )
   minMaxLoc( Mc, &myHarris_minVal, &myHarris_maxVal, 0, 0, Mat() );
 
   /* Create Window and Trackbar */
-  namedWindow( myHarris_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( myHarris_window, WINDOW_AUTOSIZE );
   createTrackbar( " Quality Level:", myHarris_window, &myHarris_qualityLevel, max_qualityLevel, myHarris_function );
   myHarris_function( 0, 0 );
 
@@ -75,7 +75,7 @@ int main( int, char** argv )
   minMaxLoc( myShiTomasi_dst, &myShiTomasi_minVal, &myShiTomasi_maxVal, 0, 0, Mat() );
 
   /* Create Window and Trackbar */
-  namedWindow( myShiTomasi_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( myShiTomasi_window, WINDOW_AUTOSIZE );
   createTrackbar( " Quality Level:", myShiTomasi_window, &myShiTomasi_qualityLevel, max_qualityLevel, myShiTomasi_function );
   myShiTomasi_function( 0, 0 );
 
@@ -120,4 +120,3 @@ void myHarris_function( int, void* )
      }
   imshow( myHarris_window, myHarris_copy );
 }
-
diff --git a/samples/cpp/tutorial_code/TrackingMotion/cornerHarris_Demo.cpp b/samples/cpp/tutorial_code/TrackingMotion/cornerHarris_Demo.cpp
index 667ad7e32a..e048f057b2 100644
--- a/samples/cpp/tutorial_code/TrackingMotion/cornerHarris_Demo.cpp
+++ b/samples/cpp/tutorial_code/TrackingMotion/cornerHarris_Demo.cpp
@@ -31,10 +31,10 @@ int main( int, char** argv )
 {
   /// Load source image and convert it to gray
   src = imread( argv[1], 1 );
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
   /// Create a window and a trackbar
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   createTrackbar( "Threshold: ", source_window, &thresh, max_thresh, cornerHarris_demo );
   imshow( source_window, src );
 
@@ -77,6 +77,6 @@ void cornerHarris_demo( int, void* )
           }
      }
   /// Showing the result
-  namedWindow( corners_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( corners_window, WINDOW_AUTOSIZE );
   imshow( corners_window, dst_norm_scaled );
 }
diff --git a/samples/cpp/tutorial_code/TrackingMotion/cornerSubPix_Demo.cpp b/samples/cpp/tutorial_code/TrackingMotion/cornerSubPix_Demo.cpp
index 3b538119e8..5957a95d6d 100644
--- a/samples/cpp/tutorial_code/TrackingMotion/cornerSubPix_Demo.cpp
+++ b/samples/cpp/tutorial_code/TrackingMotion/cornerSubPix_Demo.cpp
@@ -32,10 +32,10 @@ int main( int, char** argv )
 {
   /// Load source image and convert it to gray
   src = imread( argv[1], 1 );
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
   /// Create Window
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
 
   /// Create Trackbar to set the number of corners
   createTrackbar( "Max  corners:", source_window, &maxCorners, maxTrackbar, goodFeaturesToTrack_Demo );
@@ -87,13 +87,13 @@ void goodFeaturesToTrack_Demo( int, void* )
      { circle( copy, corners[i], r, Scalar(rng.uniform(0,255), rng.uniform(0,255), rng.uniform(0,255)), -1, 8, 0 ); }
 
   /// Show what you got
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, copy );
 
   /// Set the neeed parameters to find the refined corners
   Size winSize = Size( 5, 5 );
   Size zeroZone = Size( -1, -1 );
-  TermCriteria criteria = TermCriteria( CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 40, 0.001 );
+  TermCriteria criteria = TermCriteria( TermCriteria::EPS + TermCriteria::COUNT, 40, 0.001 );
 
   /// Calculate the refined corner locations
   cornerSubPix( src_gray, corners, winSize, zeroZone, criteria );
@@ -102,4 +102,3 @@ void goodFeaturesToTrack_Demo( int, void* )
   for( size_t i = 0; i < corners.size(); i++ )
      { cout<<" -- Refined Corner ["<<i<<"]  ("<<corners[i].x<<","<<corners[i].y<<")"<<endl; }
 }
-
diff --git a/samples/cpp/tutorial_code/TrackingMotion/goodFeaturesToTrack_Demo.cpp b/samples/cpp/tutorial_code/TrackingMotion/goodFeaturesToTrack_Demo.cpp
index e399c0c8ee..b45d60a083 100644
--- a/samples/cpp/tutorial_code/TrackingMotion/goodFeaturesToTrack_Demo.cpp
+++ b/samples/cpp/tutorial_code/TrackingMotion/goodFeaturesToTrack_Demo.cpp
@@ -32,10 +32,10 @@ int main( int, char** argv )
 {
   /// Load source image and convert it to gray
   src = imread( argv[1], 1 );
-  cvtColor( src, src_gray, CV_BGR2GRAY );
+  cvtColor( src, src_gray, COLOR_BGR2GRAY );
 
   /// Create Window
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
 
   /// Create Trackbar to set the number of corners
   createTrackbar( "Max  corners:", source_window, &maxCorners, maxTrackbar, goodFeaturesToTrack_Demo );
@@ -87,7 +87,6 @@ void goodFeaturesToTrack_Demo( int, void* )
      { circle( copy, corners[i], r, Scalar(rng.uniform(0,255), rng.uniform(0,255), rng.uniform(0,255)), -1, 8, 0 ); }
 
   /// Show what you got
-  namedWindow( source_window, CV_WINDOW_AUTOSIZE );
+  namedWindow( source_window, WINDOW_AUTOSIZE );
   imshow( source_window, copy );
 }
-
diff --git a/samples/cpp/tutorial_code/contrib/retina_tutorial.cpp b/samples/cpp/tutorial_code/bioinspired/retina_tutorial.cpp
similarity index 92%
rename from samples/cpp/tutorial_code/contrib/retina_tutorial.cpp
rename to samples/cpp/tutorial_code/bioinspired/retina_tutorial.cpp
index 8ebcb7cba1..994d881dec 100644
--- a/samples/cpp/tutorial_code/contrib/retina_tutorial.cpp
+++ b/samples/cpp/tutorial_code/bioinspired/retina_tutorial.cpp
@@ -9,7 +9,8 @@
 #include <iostream>
 #include <cstring>
 
-#include "opencv2/opencv.hpp"
+#include "opencv2/bioinspired.hpp"
+#include "opencv2/highgui.hpp"
 
 static void help(std::string errorMessage)
 {
@@ -38,7 +39,7 @@ int main(int argc, char* argv[]) {
     // welcome message
     std::cout<<"****************************************************"<<std::endl;
     std::cout<<"* Retina demonstration : demonstrates the use of is a wrapper class of the Gipsa/Listic Labs retina model."<<std::endl;
-    std::cout<<"* This demo will try to load the file 'RetinaSpecificParameters.xml' (if exists).\nTo create it, copy the autogenerated template 'RetinaDefaultParameters.xml'.\nThen twaek it with your own retina parameters."<<std::endl;
+    std::cout<<"* This demo will try to load the file 'RetinaSpecificParameters.xml' (if exists).\nTo create it, copy the autogenerated template 'RetinaDefaultParameters.xml'.\nThen tweak it with your own retina parameters."<<std::endl;
     // basic input arguments checking
     if (argc<2)
     {
@@ -94,15 +95,17 @@ int main(int argc, char* argv[]) {
     try
     {
         // create a retina instance with default parameters setup, uncomment the initialisation you wanna test
-        cv::Ptr<cv::Retina> myRetina;
+        cv::Ptr<cv::bioinspired::Retina> myRetina;
 
         // if the last parameter is 'log', then activate log sampling (favour foveal vision and subsamples peripheral vision)
         if (useLogSampling)
         {
-                        myRetina = new cv::Retina(inputFrame.size(), true, cv::RETINA_COLOR_BAYER, true, 2.0, 10.0);
+            myRetina = cv::bioinspired::createRetina(inputFrame.size(), true, cv::bioinspired::RETINA_COLOR_BAYER, true, 2.0, 10.0);
         }
         else// -> else allocate "classical" retina :
-            myRetina = new cv::Retina(inputFrame.size());
+        {
+            myRetina = cv::bioinspired::createRetina(inputFrame.size());
+        }
 
         // save default retina parameters file in order to let you see this and maybe modify it and reload using method "setup"
         myRetina->write("RetinaDefaultParameters.xml");
@@ -136,7 +139,7 @@ int main(int argc, char* argv[]) {
         }
     }catch(cv::Exception e)
     {
-        std::cerr<<"Error using Retina : "<<e.what()<<std::endl;
+        std::cerr<<"Error using Retina or end of video sequence reached : "<<e.what()<<std::endl;
     }
 
     // Program end message
diff --git a/samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp b/samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp
index aa4a0361d9..bf74e21a03 100644
--- a/samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp
+++ b/samples/cpp/tutorial_code/calib3d/camera_calibration/camera_calibration.cpp
@@ -3,10 +3,11 @@
 #include <time.h>
 #include <stdio.h>
 
-#include <opencv2/core/core.hpp>
-#include <opencv2/imgproc/imgproc.hpp>
-#include <opencv2/calib3d/calib3d.hpp>
-#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/calib3d.hpp>
+#include <opencv2/highgui.hpp>
 
 using namespace cv;
 using namespace std;
@@ -23,7 +24,7 @@ class Settings
 public:
     Settings() : goodInput(false) {}
     enum Pattern { NOT_EXISTING, CHESSBOARD, CIRCLES_GRID, ASYMMETRIC_CIRCLES_GRID };
-    enum InputType {INVALID, CAMERA, VIDEO_FILE, IMAGE_LIST};
+    enum InputType { INVALID, CAMERA, VIDEO_FILE, IMAGE_LIST };
 
     void write(FileStorage& fs) const                        //Write serialization for this class
     {
@@ -119,9 +120,9 @@ public:
         }
 
         flag = 0;
-        if(calibFixPrincipalPoint) flag |= CV_CALIB_FIX_PRINCIPAL_POINT;
-        if(calibZeroTangentDist)   flag |= CV_CALIB_ZERO_TANGENT_DIST;
-        if(aspectRatio)            flag |= CV_CALIB_FIX_ASPECT_RATIO;
+        if(calibFixPrincipalPoint) flag |= CALIB_FIX_PRINCIPAL_POINT;
+        if(calibZeroTangentDist)   flag |= CALIB_ZERO_TANGENT_DIST;
+        if(aspectRatio)            flag |= CALIB_FIX_ASPECT_RATIO;
 
 
         calibrationPattern = NOT_EXISTING;
@@ -146,7 +147,7 @@ public:
             view0.copyTo(result);
         }
         else if( atImageList < (int)imageList.size() )
-            result = imread(imageList[atImageList++], CV_LOAD_IMAGE_COLOR);
+            result = imread(imageList[atImageList++], IMREAD_COLOR);
 
         return result;
     }
@@ -271,7 +272,7 @@ int main(int argc, char* argv[])
         {
         case Settings::CHESSBOARD:
             found = findChessboardCorners( view, s.boardSize, pointBuf,
-                CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_FAST_CHECK | CV_CALIB_CB_NORMALIZE_IMAGE);
+                CALIB_CB_ADAPTIVE_THRESH | CALIB_CB_FAST_CHECK | CALIB_CB_NORMALIZE_IMAGE);
             break;
         case Settings::CIRCLES_GRID:
             found = findCirclesGrid( view, s.boardSize, pointBuf );
@@ -290,9 +291,9 @@ int main(int argc, char* argv[])
                 if( s.calibrationPattern == Settings::CHESSBOARD)
                 {
                     Mat viewGray;
-                    cvtColor(view, viewGray, CV_BGR2GRAY);
+                    cvtColor(view, viewGray, COLOR_BGR2GRAY);
                     cornerSubPix( viewGray, pointBuf, Size(11,11),
-                        Size(-1,-1), TermCriteria( CV_TERMCRIT_EPS+CV_TERMCRIT_ITER, 30, 0.1 ));
+                        Size(-1,-1), TermCriteria( TermCriteria::EPS+TermCriteria::COUNT, 30, 0.1 ));
                 }
 
                 if( mode == CAPTURING &&  // For camera only take new samples after delay time
@@ -391,7 +392,7 @@ static double computeReprojectionErrors( const vector<vector<Point3f> >& objectP
     {
         projectPoints( Mat(objectPoints[i]), rvecs[i], tvecs[i], cameraMatrix,
                        distCoeffs, imagePoints2);
-        err = norm(Mat(imagePoints[i]), Mat(imagePoints2), CV_L2);
+        err = norm(Mat(imagePoints[i]), Mat(imagePoints2), NORM_L2);
 
         int n = (int)objectPoints[i].size();
         perViewErrors[i] = (float) std::sqrt(err*err/n);
@@ -432,7 +433,7 @@ static bool runCalibration( Settings& s, Size& imageSize, Mat& cameraMatrix, Mat
 {
 
     cameraMatrix = Mat::eye(3, 3, CV_64F);
-    if( s.flag & CV_CALIB_FIX_ASPECT_RATIO )
+    if( s.flag & CALIB_FIX_ASPECT_RATIO )
         cameraMatrix.at<double>(0,0) = 1.0;
 
     distCoeffs = Mat::zeros(8, 1, CV_64F);
@@ -444,7 +445,7 @@ static bool runCalibration( Settings& s, Size& imageSize, Mat& cameraMatrix, Mat
 
     //Find intrinsic and extrinsic camera parameters
     double rms = calibrateCamera(objectPoints, imagePoints, imageSize, cameraMatrix,
-                                 distCoeffs, rvecs, tvecs, s.flag|CV_CALIB_FIX_K4|CV_CALIB_FIX_K5);
+                                 distCoeffs, rvecs, tvecs, s.flag|CALIB_FIX_K4|CALIB_FIX_K5);
 
     cout << "Re-projection error reported by calibrateCamera: "<< rms << endl;
 
@@ -480,17 +481,17 @@ static void saveCameraParams( Settings& s, Size& imageSize, Mat& cameraMatrix, M
     fs << "board_Height" << s.boardSize.height;
     fs << "square_Size" << s.squareSize;
 
-    if( s.flag & CV_CALIB_FIX_ASPECT_RATIO )
+    if( s.flag & CALIB_FIX_ASPECT_RATIO )
         fs << "FixAspectRatio" << s.aspectRatio;
 
     if( s.flag )
     {
         sprintf( buf, "flags: %s%s%s%s",
-            s.flag & CV_CALIB_USE_INTRINSIC_GUESS ? " +use_intrinsic_guess" : "",
-            s.flag & CV_CALIB_FIX_ASPECT_RATIO ? " +fix_aspectRatio" : "",
-            s.flag & CV_CALIB_FIX_PRINCIPAL_POINT ? " +fix_principal_point" : "",
-            s.flag & CV_CALIB_ZERO_TANGENT_DIST ? " +zero_tangent_dist" : "" );
-        cvWriteComment( *fs, buf, 0 );
+            s.flag & CALIB_USE_INTRINSIC_GUESS ? " +use_intrinsic_guess" : "",
+            s.flag & CALIB_FIX_ASPECT_RATIO ? " +fix_aspectRatio" : "",
+            s.flag & CALIB_FIX_PRINCIPAL_POINT ? " +fix_principal_point" : "",
+            s.flag & CALIB_ZERO_TANGENT_DIST ? " +zero_tangent_dist" : "" );
+        //cvWriteComment( *fs, buf, 0 );
 
     }
 
@@ -518,7 +519,7 @@ static void saveCameraParams( Settings& s, Size& imageSize, Mat& cameraMatrix, M
             r = rvecs[i].t();
             t = tvecs[i].t();
         }
-        cvWriteComment( *fs, "a set of 6-tuples (rotation vector + translation vector) for each view", 0 );
+        //cvWriteComment( *fs, "a set of 6-tuples (rotation vector + translation vector) for each view", 0 );
         fs << "Extrinsic_Parameters" << bigmat;
     }
 
diff --git a/samples/cpp/tutorial_code/calib3d/stereoBM/SBM_Sample.cpp b/samples/cpp/tutorial_code/calib3d/stereoBM/SBM_Sample.cpp
index 13d5aa8708..64c650bb82 100644
--- a/samples/cpp/tutorial_code/calib3d/stereoBM/SBM_Sample.cpp
+++ b/samples/cpp/tutorial_code/calib3d/stereoBM/SBM_Sample.cpp
@@ -26,8 +26,8 @@ int main( int argc, char** argv )
   { readme(); return -1; }
 
   //-- 1. Read the images
-  Mat imgLeft = imread( argv[1], CV_LOAD_IMAGE_GRAYSCALE );
-  Mat imgRight = imread( argv[2], CV_LOAD_IMAGE_GRAYSCALE );
+  Mat imgLeft = imread( argv[1], IMREAD_GRAYSCALE );
+  Mat imgRight = imread( argv[2], IMREAD_GRAYSCALE );
   //-- And create the image in which we will save our disparities
   Mat imgDisparity16S = Mat( imgLeft.rows, imgLeft.cols, CV_16S );
   Mat imgDisparity8U = Mat( imgLeft.rows, imgLeft.cols, CV_8UC1 );
@@ -39,12 +39,10 @@ int main( int argc, char** argv )
   int ndisparities = 16*5;   /**< Range of disparity */
   int SADWindowSize = 21; /**< Size of the block window. Must be odd */
 
-  StereoBM sbm( StereoBM::BASIC_PRESET,
-                                ndisparities,
-                SADWindowSize );
+  Ptr<StereoBM> sbm = createStereoBM( ndisparities, SADWindowSize );
 
   //-- 3. Calculate the disparity image
-  sbm( imgLeft, imgRight, imgDisparity16S, CV_16S );
+  sbm->compute( imgLeft, imgRight, imgDisparity16S );
 
   //-- Check its extreme values
   double minVal; double maxVal;
@@ -56,7 +54,7 @@ int main( int argc, char** argv )
   //-- 4. Display it as a CV_8UC1 image
   imgDisparity16S.convertTo( imgDisparity8U, CV_8UC1, 255/(maxVal - minVal));
 
-  namedWindow( windowDisparity, CV_WINDOW_NORMAL );
+  namedWindow( windowDisparity, WINDOW_NORMAL );
   imshow( windowDisparity, imgDisparity8U );
 
   //-- 5. Save the image
diff --git a/samples/cpp/tutorial_code/core/Matrix/Drawing_1.cpp b/samples/cpp/tutorial_code/core/Matrix/Drawing_1.cpp
index 3e2437fef2..5cad99208c 100644
--- a/samples/cpp/tutorial_code/core/Matrix/Drawing_1.cpp
+++ b/samples/cpp/tutorial_code/core/Matrix/Drawing_1.cpp
@@ -64,9 +64,9 @@ int main( void ){
 
   /// 3. Display your stuff!
   imshow( atom_window, atom_image );
-  cvMoveWindow( atom_window, 0, 200 );
+  moveWindow( atom_window, 0, 200 );
   imshow( rook_window, rook_image );
-  cvMoveWindow( rook_window, w, 200 );
+  moveWindow( rook_window, w, 200 );
 
   waitKey( 0 );
   return(0);
@@ -168,5 +168,3 @@ void MyLine( Mat img, Point start, Point end )
     thickness,
     lineType );
 }
-
-
diff --git a/samples/cpp/tutorial_code/core/Matrix/Drawing_2.cpp b/samples/cpp/tutorial_code/core/Matrix/Drawing_2.cpp
index c84eccb69e..844bcd27de 100644
--- a/samples/cpp/tutorial_code/core/Matrix/Drawing_2.cpp
+++ b/samples/cpp/tutorial_code/core/Matrix/Drawing_2.cpp
@@ -304,7 +304,7 @@ int Displaying_Random_Text( Mat image, char* window_name, RNG rng )
  */
 int Displaying_Big_End( Mat image, char* window_name, RNG )
 {
-  Size textsize = getTextSize("OpenCV forever!", CV_FONT_HERSHEY_COMPLEX, 3, 5, 0);
+  Size textsize = getTextSize("OpenCV forever!", FONT_HERSHEY_COMPLEX, 3, 5, 0);
   Point org((window_width - textsize.width)/2, (window_height - textsize.height)/2);
   int lineType = 8;
 
@@ -313,7 +313,7 @@ int Displaying_Big_End( Mat image, char* window_name, RNG )
   for( int i = 0; i < 255; i += 2 )
   {
     image2 = image - Scalar::all(i);
-    putText( image2, "OpenCV forever!", org, CV_FONT_HERSHEY_COMPLEX, 3,
+    putText( image2, "OpenCV forever!", org, FONT_HERSHEY_COMPLEX, 3,
              Scalar(i, i, 255), 5, lineType );
 
     imshow( window_name, image2 );
diff --git a/samples/cpp/tutorial_code/core/discrete_fourier_transform/discrete_fourier_transform.cpp b/samples/cpp/tutorial_code/core/discrete_fourier_transform/discrete_fourier_transform.cpp
index ba54188aee..de7a32ca6b 100644
--- a/samples/cpp/tutorial_code/core/discrete_fourier_transform/discrete_fourier_transform.cpp
+++ b/samples/cpp/tutorial_code/core/discrete_fourier_transform/discrete_fourier_transform.cpp
@@ -22,7 +22,7 @@ int main(int argc, char ** argv)
 
     const char* filename = argc >=2 ? argv[1] : "lena.jpg";
 
-    Mat I = imread(filename, CV_LOAD_IMAGE_GRAYSCALE);
+    Mat I = imread(filename, IMREAD_GRAYSCALE);
     if( I.empty())
         return -1;
 
@@ -67,7 +67,7 @@ int main(int argc, char ** argv)
     q2.copyTo(q1);
     tmp.copyTo(q2);
 
-    normalize(magI, magI, 0, 1, CV_MINMAX); // Transform the matrix with float values into a
+    normalize(magI, magI, 0, 1, NORM_MINMAX); // Transform the matrix with float values into a
                                             // viewable image form (float between values 0 and 1).
 
     imshow("Input Image"       , I   );    // Show the result
@@ -75,4 +75,4 @@ int main(int argc, char ** argv)
     waitKey();
 
     return 0;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/core/file_input_output/file_input_output.cpp b/samples/cpp/tutorial_code/core/file_input_output/file_input_output.cpp
index 9c527a8aa6..388b7744f1 100644
--- a/samples/cpp/tutorial_code/core/file_input_output/file_input_output.cpp
+++ b/samples/cpp/tutorial_code/core/file_input_output/file_input_output.cpp
@@ -151,4 +151,4 @@ int main(int ac, char** av)
         << "Tip: Open up " << filename << " with a text editor to see the serialized data." << endl;
 
     return 0;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/core/how_to_scan_images/how_to_scan_images.cpp b/samples/cpp/tutorial_code/core/how_to_scan_images/how_to_scan_images.cpp
index 49aca426e1..f257d6ed6e 100644
--- a/samples/cpp/tutorial_code/core/how_to_scan_images/how_to_scan_images.cpp
+++ b/samples/cpp/tutorial_code/core/how_to_scan_images/how_to_scan_images.cpp
@@ -1,5 +1,6 @@
-#include <opencv2/core/core.hpp>
-#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/highgui.hpp>
 #include <iostream>
 #include <sstream>
 
@@ -35,9 +36,9 @@ int main( int argc, char* argv[])
 
     Mat I, J;
     if( argc == 4 && !strcmp(argv[3],"G") )
-        I = imread(argv[1], CV_LOAD_IMAGE_GRAYSCALE);
+        I = imread(argv[1], IMREAD_GRAYSCALE);
     else
-        I = imread(argv[1], CV_LOAD_IMAGE_COLOR);
+        I = imread(argv[1], IMREAD_COLOR);
 
     if (!I.data)
     {
@@ -213,4 +214,4 @@ Mat& ScanImageAndReduceRandomAccess(Mat& I, const uchar* const table)
     }
 
     return I;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.cpp b/samples/cpp/tutorial_code/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.cpp
index aeeaf21100..e13f2b6960 100644
--- a/samples/cpp/tutorial_code/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.cpp
+++ b/samples/cpp/tutorial_code/core/interoperability_with_OpenCV_1/interoperability_with_OpenCV_1.cpp
@@ -4,6 +4,7 @@
 #include <opencv2/core/core.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 #include <opencv2/highgui/highgui.hpp>
+#include <opencv2/core/utility.hpp>
 
 using namespace cv;  // The new C++ interface API is inside this namespace. Import it.
 using namespace std;
@@ -21,19 +22,23 @@ static void help( char* progName)
 // comment out the define to use only the latest C++ API
 #define DEMO_MIXED_API_USE
 
+#ifdef DEMO_MIXED_API_USE
+#  include <opencv2/highgui/highgui_c.h>
+#endif
+
 int main( int argc, char** argv )
 {
     help(argv[0]);
     const char* imagename = argc > 1 ? argv[1] : "lena.jpg";
 
 #ifdef DEMO_MIXED_API_USE
-    Ptr<IplImage> IplI = cvLoadImage(imagename);      // Ptr<T> is safe ref-counting pointer class
-    if(IplI.empty())
+    Ptr<IplImage> IplI(cvLoadImage(imagename));      // Ptr<T> is a safe ref-counting pointer class
+    if(!IplI)
     {
         cerr << "Can not load image " <<  imagename << endl;
         return -1;
     }
-    Mat I(IplI); // Convert to the new style container. Only header created. Image not copied.
+    Mat I = cv::cvarrToMat(IplI); // Convert to the new style container. Only header created. Image not copied.
 #else
     Mat I = imread(imagename);        // the newer cvLoadImage alternative, MATLAB-style function
     if( I.empty() )                   // same as if( !I.data )
@@ -45,7 +50,7 @@ int main( int argc, char** argv )
 
     // convert image to YUV color space. The output image will be created automatically.
     Mat I_YUV;
-    cvtColor(I, I_YUV, CV_BGR2YCrCb);
+    cvtColor(I, I_YUV, COLOR_BGR2YCrCb);
 
     vector<Mat> planes;    // Use the STL's vector structure to store multiple Mat objects
     split(I_YUV, planes);  // split the image into separate color planes (Y U V)
@@ -114,10 +119,10 @@ int main( int argc, char** argv )
 
 
     merge(planes, I_YUV);                // now merge the results back
-    cvtColor(I_YUV, I, CV_YCrCb2BGR);  // and produce the output RGB image
+    cvtColor(I_YUV, I, COLOR_YCrCb2BGR);  // and produce the output RGB image
 
 
-    namedWindow("image with grain", CV_WINDOW_AUTOSIZE);   // use this to create images
+    namedWindow("image with grain", WINDOW_AUTOSIZE);   // use this to create images
 
 #ifdef DEMO_MIXED_API_USE
     // this is to demonstrate that I and IplI really share the data - the result of the above
diff --git a/samples/cpp/tutorial_code/core/mat_mask_operations/mat_mask_operations.cpp b/samples/cpp/tutorial_code/core/mat_mask_operations/mat_mask_operations.cpp
index 94e0a29a61..0310b973a6 100644
--- a/samples/cpp/tutorial_code/core/mat_mask_operations/mat_mask_operations.cpp
+++ b/samples/cpp/tutorial_code/core/mat_mask_operations/mat_mask_operations.cpp
@@ -1,6 +1,7 @@
-#include <opencv2/core/core.hpp>
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/imgproc/imgproc.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/imgproc.hpp>
 #include <iostream>
 
 using namespace std;
@@ -26,12 +27,12 @@ int main( int argc, char* argv[])
     Mat I, J, K;
 
     if (argc >= 3 && !strcmp("G", argv[2]))
-        I = imread( filename, CV_LOAD_IMAGE_GRAYSCALE);
+        I = imread( filename, IMREAD_GRAYSCALE);
     else
-        I = imread( filename, CV_LOAD_IMAGE_COLOR);
+        I = imread( filename, IMREAD_COLOR);
 
-    namedWindow("Input", CV_WINDOW_AUTOSIZE);
-    namedWindow("Output", CV_WINDOW_AUTOSIZE);
+    namedWindow("Input", WINDOW_AUTOSIZE);
+    namedWindow("Output", WINDOW_AUTOSIZE);
 
     imshow("Input", I);
     double t = (double)getTickCount();
@@ -42,7 +43,7 @@ int main( int argc, char* argv[])
     cout << "Hand written function times passed in seconds: " << t << endl;
 
     imshow("Output", J);
-    cvWaitKey(0);
+    waitKey();
 
     Mat kern = (Mat_<char>(3,3) <<  0, -1,  0,
                                    -1,  5, -1,
@@ -54,7 +55,7 @@ int main( int argc, char* argv[])
 
     imshow("Output", K);
 
-    cvWaitKey(0);
+    waitKey();
     return 0;
 }
 void Sharpen(const Mat& myImage,Mat& Result)
@@ -83,4 +84,4 @@ void Sharpen(const Mat& myImage,Mat& Result)
     Result.row(Result.rows-1).setTo(Scalar(0));
     Result.col(0).setTo(Scalar(0));
     Result.col(Result.cols-1).setTo(Scalar(0));
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/core/mat_the_basic_image_container/mat_the_basic_image_container.cpp b/samples/cpp/tutorial_code/core/mat_the_basic_image_container/mat_the_basic_image_container.cpp
index 4f26f050da..91cc64f9b2 100644
--- a/samples/cpp/tutorial_code/core/mat_the_basic_image_container/mat_the_basic_image_container.cpp
+++ b/samples/cpp/tutorial_code/core/mat_the_basic_image_container/mat_the_basic_image_container.cpp
@@ -59,10 +59,10 @@ int main(int,char**)
 
     // Demonstrate the output formating options
     cout << "R (default) = " << endl <<        R           << endl << endl;
-    cout << "R (python)  = " << endl << format(R,"python") << endl << endl;
-    cout << "R (numpy)   = " << endl << format(R,"numpy" ) << endl << endl;
-    cout << "R (csv)     = " << endl << format(R,"csv"   ) << endl << endl;
-    cout << "R (c)       = " << endl << format(R,"C"     ) << endl << endl;
+    cout << "R (python)  = " << endl << format(R, Formatter::FMT_PYTHON) << endl << endl;
+    cout << "R (numpy)   = " << endl << format(R, Formatter::FMT_NUMPY ) << endl << endl;
+    cout << "R (csv)     = " << endl << format(R, Formatter::FMT_CSV   ) << endl << endl;
+    cout << "R (c)       = " << endl << format(R, Formatter::FMT_C     ) << endl << endl;
 
     Point2f P(5, 1);
     cout << "Point (2D) = " << P << endl << endl;
@@ -82,4 +82,4 @@ int main(int,char**)
 
     cout << "A vector of 2D Points = " << vPoints << endl << endl;
     return 0;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/features2D/SURF_FlannMatcher.cpp b/samples/cpp/tutorial_code/features2D/SURF_FlannMatcher.cpp
index bdfd0789c0..402d6456ac 100644
--- a/samples/cpp/tutorial_code/features2D/SURF_FlannMatcher.cpp
+++ b/samples/cpp/tutorial_code/features2D/SURF_FlannMatcher.cpp
@@ -25,8 +25,8 @@ int main( int argc, char** argv )
   if( argc != 3 )
   { readme(); return -1; }
 
-  Mat img_1 = imread( argv[1], CV_LOAD_IMAGE_GRAYSCALE );
-  Mat img_2 = imread( argv[2], CV_LOAD_IMAGE_GRAYSCALE );
+  Mat img_1 = imread( argv[1], IMREAD_GRAYSCALE );
+  Mat img_2 = imread( argv[2], IMREAD_GRAYSCALE );
 
   if( !img_1.data || !img_2.data )
   { std::cout<< " --(!) Error reading images " << std::endl; return -1; }
@@ -71,7 +71,7 @@ int main( int argc, char** argv )
   std::vector< DMatch > good_matches;
 
   for( int i = 0; i < descriptors_1.rows; i++ )
-  { if( matches[i].distance < 2*min_dist )
+  { if( matches[i].distance <= 2*min_dist )
     { good_matches.push_back( matches[i]); }
   }
 
diff --git a/samples/cpp/tutorial_code/features2D/SURF_Homography.cpp b/samples/cpp/tutorial_code/features2D/SURF_Homography.cpp
index cf99d4e6e7..47bc3ecfe3 100644
--- a/samples/cpp/tutorial_code/features2D/SURF_Homography.cpp
+++ b/samples/cpp/tutorial_code/features2D/SURF_Homography.cpp
@@ -26,8 +26,8 @@ int main( int argc, char** argv )
   if( argc != 3 )
   { readme(); return -1; }
 
-  Mat img_object = imread( argv[1], CV_LOAD_IMAGE_GRAYSCALE );
-  Mat img_scene = imread( argv[2], CV_LOAD_IMAGE_GRAYSCALE );
+  Mat img_object = imread( argv[1], IMREAD_GRAYSCALE );
+  Mat img_scene = imread( argv[2], IMREAD_GRAYSCALE );
 
   if( !img_object.data || !img_scene.data )
   { std::cout<< " --(!) Error reading images " << std::endl; return -1; }
@@ -92,12 +92,12 @@ int main( int argc, char** argv )
     scene.push_back( keypoints_scene[ good_matches[i].trainIdx ].pt );
   }
 
-  Mat H = findHomography( obj, scene, CV_RANSAC );
+  Mat H = findHomography( obj, scene, RANSAC );
 
   //-- Get the corners from the image_1 ( the object to be "detected" )
   std::vector<Point2f> obj_corners(4);
-  obj_corners[0] = cvPoint(0,0); obj_corners[1] = cvPoint( img_object.cols, 0 );
-  obj_corners[2] = cvPoint( img_object.cols, img_object.rows ); obj_corners[3] = cvPoint( 0, img_object.rows );
+  obj_corners[0] = Point(0,0); obj_corners[1] = Point( img_object.cols, 0 );
+  obj_corners[2] = Point( img_object.cols, img_object.rows ); obj_corners[3] = Point( 0, img_object.rows );
   std::vector<Point2f> scene_corners(4);
 
   perspectiveTransform( obj_corners, scene_corners, H);
diff --git a/samples/cpp/tutorial_code/features2D/SURF_descriptor.cpp b/samples/cpp/tutorial_code/features2D/SURF_descriptor.cpp
index 527e5dd951..56067c83ce 100644
--- a/samples/cpp/tutorial_code/features2D/SURF_descriptor.cpp
+++ b/samples/cpp/tutorial_code/features2D/SURF_descriptor.cpp
@@ -24,8 +24,8 @@ int main( int argc, char** argv )
   if( argc != 3 )
   { return -1; }
 
-  Mat img_1 = imread( argv[1], CV_LOAD_IMAGE_GRAYSCALE );
-  Mat img_2 = imread( argv[2], CV_LOAD_IMAGE_GRAYSCALE );
+  Mat img_1 = imread( argv[1], IMREAD_GRAYSCALE );
+  Mat img_2 = imread( argv[2], IMREAD_GRAYSCALE );
 
   if( !img_1.data || !img_2.data )
   { return -1; }
diff --git a/samples/cpp/tutorial_code/features2D/SURF_detector.cpp b/samples/cpp/tutorial_code/features2D/SURF_detector.cpp
index 2625f1df3b..a1288a8c85 100644
--- a/samples/cpp/tutorial_code/features2D/SURF_detector.cpp
+++ b/samples/cpp/tutorial_code/features2D/SURF_detector.cpp
@@ -24,8 +24,8 @@ int main( int argc, char** argv )
   if( argc != 3 )
   { readme(); return -1; }
 
-  Mat img_1 = imread( argv[1], CV_LOAD_IMAGE_GRAYSCALE );
-  Mat img_2 = imread( argv[2], CV_LOAD_IMAGE_GRAYSCALE );
+  Mat img_1 = imread( argv[1], IMREAD_GRAYSCALE );
+  Mat img_2 = imread( argv[2], IMREAD_GRAYSCALE );
 
   if( !img_1.data || !img_2.data )
   { std::cout<< " --(!) Error reading images " << std::endl; return -1; }
diff --git a/samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp b/samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
index 0bcb6fafeb..e8e961031a 100644
--- a/samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
+++ b/samples/cpp/tutorial_code/gpu/gpu-basics-similarity/gpu-basics-similarity.cpp
@@ -1,10 +1,14 @@
 #include <iostream>                   // Console I/O
 #include <sstream>                    // String to number conversion
 
-#include <opencv2/core/core.hpp>      // Basic OpenCV structures
-#include <opencv2/imgproc/imgproc.hpp>// Image processing methods for the CPU
-#include <opencv2/highgui/highgui.hpp>// Read images
-#include <opencv2/gpu/gpu.hpp>        // GPU structures and methods
+#include <opencv2/core.hpp>      // Basic OpenCV structures
+#include <opencv2/core/utility.hpp>
+#include <opencv2/imgproc.hpp>// Image processing methods for the CPU
+#include <opencv2/highgui.hpp>// Read images
+
+// GPU structures and methods
+#include <opencv2/gpuarithm.hpp>
+#include <opencv2/gpufilters.hpp>
 
 using namespace std;
 using namespace cv;
@@ -304,6 +308,8 @@ Scalar getMSSIM_GPU( const Mat& i1, const Mat& i2)
     gpu::split(tmp2, vI2);
     Scalar mssim;
 
+    Ptr<gpu::Filter> gauss = gpu::createGaussianFilter(vI2[0].type(), -1, Size(11, 11), 1.5);
+
     for( int i = 0; i < gI1.channels(); ++i )
     {
         gpu::GpuMat I2_2, I1_2, I1_I2;
@@ -314,8 +320,8 @@ Scalar getMSSIM_GPU( const Mat& i1, const Mat& i2)
 
         /*************************** END INITS **********************************/
         gpu::GpuMat mu1, mu2;   // PRELIMINARY COMPUTING
-        gpu::GaussianBlur(vI1[i], mu1, Size(11, 11), 1.5);
-        gpu::GaussianBlur(vI2[i], mu2, Size(11, 11), 1.5);
+        gauss->apply(vI1[i], mu1);
+        gauss->apply(vI2[i], mu2);
 
         gpu::GpuMat mu1_2, mu2_2, mu1_mu2;
         gpu::multiply(mu1, mu1, mu1_2);
@@ -324,13 +330,13 @@ Scalar getMSSIM_GPU( const Mat& i1, const Mat& i2)
 
         gpu::GpuMat sigma1_2, sigma2_2, sigma12;
 
-        gpu::GaussianBlur(I1_2, sigma1_2, Size(11, 11), 1.5);
+        gauss->apply(I1_2, sigma1_2);
         gpu::subtract(sigma1_2, mu1_2, sigma1_2); // sigma1_2 -= mu1_2;
 
-        gpu::GaussianBlur(I2_2, sigma2_2, Size(11, 11), 1.5);
+        gauss->apply(I2_2, sigma2_2);
         gpu::subtract(sigma2_2, mu2_2, sigma2_2); // sigma2_2 -= mu2_2;
 
-        gpu::GaussianBlur(I1_I2, sigma12, Size(11, 11), 1.5);
+        gauss->apply(I1_I2, sigma12);
         gpu::subtract(sigma12, mu1_mu2, sigma12); // sigma12 -= mu1_mu2;
 
         ///////////////////////////////// FORMULA ////////////////////////////////
@@ -364,37 +370,37 @@ Scalar getMSSIM_GPU_optimized( const Mat& i1, const Mat& i2, BufferMSSIM& b)
 
     gpu::Stream stream;
 
-    stream.enqueueConvert(b.gI1, b.t1, CV_32F);
-    stream.enqueueConvert(b.gI2, b.t2, CV_32F);
+    b.gI1.convertTo(b.t1, CV_32F, stream);
+    b.gI2.convertTo(b.t2, CV_32F, stream);
 
     gpu::split(b.t1, b.vI1, stream);
     gpu::split(b.t2, b.vI2, stream);
     Scalar mssim;
 
-    gpu::GpuMat buf;
+    Ptr<gpu::Filter> gauss = gpu::createGaussianFilter(b.vI1[0].type(), -1, Size(11, 11), 1.5);
 
     for( int i = 0; i < b.gI1.channels(); ++i )
     {
-        gpu::multiply(b.vI2[i], b.vI2[i], b.I2_2, stream);        // I2^2
-        gpu::multiply(b.vI1[i], b.vI1[i], b.I1_2, stream);        // I1^2
-        gpu::multiply(b.vI1[i], b.vI2[i], b.I1_I2, stream);       // I1 * I2
+        gpu::multiply(b.vI2[i], b.vI2[i], b.I2_2, 1, -1, stream);        // I2^2
+        gpu::multiply(b.vI1[i], b.vI1[i], b.I1_2, 1, -1, stream);        // I1^2
+        gpu::multiply(b.vI1[i], b.vI2[i], b.I1_I2, 1, -1, stream);       // I1 * I2
 
-        gpu::GaussianBlur(b.vI1[i], b.mu1, Size(11, 11), buf, 1.5, 0, BORDER_DEFAULT, -1, stream);
-        gpu::GaussianBlur(b.vI2[i], b.mu2, Size(11, 11), buf, 1.5, 0, BORDER_DEFAULT, -1, stream);
+        gauss->apply(b.vI1[i], b.mu1, stream);
+        gauss->apply(b.vI2[i], b.mu2, stream);
 
-        gpu::multiply(b.mu1, b.mu1, b.mu1_2, stream);
-        gpu::multiply(b.mu2, b.mu2, b.mu2_2, stream);
-        gpu::multiply(b.mu1, b.mu2, b.mu1_mu2, stream);
+        gpu::multiply(b.mu1, b.mu1, b.mu1_2, 1, -1, stream);
+        gpu::multiply(b.mu2, b.mu2, b.mu2_2, 1, -1, stream);
+        gpu::multiply(b.mu1, b.mu2, b.mu1_mu2, 1, -1, stream);
 
-        gpu::GaussianBlur(b.I1_2, b.sigma1_2, Size(11, 11), buf, 1.5, 0, BORDER_DEFAULT, -1, stream);
+        gauss->apply(b.I1_2, b.sigma1_2, stream);
         gpu::subtract(b.sigma1_2, b.mu1_2, b.sigma1_2, gpu::GpuMat(), -1, stream);
         //b.sigma1_2 -= b.mu1_2;  - This would result in an extra data transfer operation
 
-        gpu::GaussianBlur(b.I2_2, b.sigma2_2, Size(11, 11), buf, 1.5, 0, BORDER_DEFAULT, -1, stream);
+        gauss->apply(b.I2_2, b.sigma2_2, stream);
         gpu::subtract(b.sigma2_2, b.mu2_2, b.sigma2_2, gpu::GpuMat(), -1, stream);
         //b.sigma2_2 -= b.mu2_2;
 
-        gpu::GaussianBlur(b.I1_I2, b.sigma12, Size(11, 11), buf, 1.5, 0, BORDER_DEFAULT, -1, stream);
+        gauss->apply(b.I1_I2, b.sigma12, stream);
         gpu::subtract(b.sigma12, b.mu1_mu2, b.sigma12, gpu::GpuMat(), -1, stream);
         //b.sigma12 -= b.mu1_mu2;
 
@@ -424,4 +430,3 @@ Scalar getMSSIM_GPU_optimized( const Mat& i1, const Mat& i2, BufferMSSIM& b)
     }
     return mssim;
 }
-
diff --git a/samples/cpp/tutorial_code/introduction/display_image/display_image.cpp b/samples/cpp/tutorial_code/introduction/display_image/display_image.cpp
index 9f2e31e753..847cd8f6dc 100644
--- a/samples/cpp/tutorial_code/introduction/display_image/display_image.cpp
+++ b/samples/cpp/tutorial_code/introduction/display_image/display_image.cpp
@@ -14,17 +14,17 @@ int main( int argc, char** argv )
     }
 
     Mat image;
-    image = imread(argv[1], CV_LOAD_IMAGE_COLOR);	// Read the file
+    image = imread(argv[1], IMREAD_COLOR); // Read the file
 
-    if(! image.data )                              // Check for invalid input
+    if(! image.data )                      // Check for invalid input
     {
         cout <<  "Could not open or find the image" << std::endl ;
         return -1;
     }
 
-    namedWindow( "Display window", CV_WINDOW_AUTOSIZE );// Create a window for display.
-    imshow( "Display window", image );                   // Show our image inside it.
+    namedWindow( "Display window", WINDOW_AUTOSIZE ); // Create a window for display.
+    imshow( "Display window", image );                // Show our image inside it.
 
-    waitKey(0);											 // Wait for a keystroke in the window
+    waitKey(0); // Wait for a keystroke in the window
     return 0;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/introduction/windows_visual_studio_Opencv/Test.cpp b/samples/cpp/tutorial_code/introduction/windows_visual_studio_Opencv/Test.cpp
index 7f43f0e779..b5470efa30 100644
--- a/samples/cpp/tutorial_code/introduction/windows_visual_studio_Opencv/Test.cpp
+++ b/samples/cpp/tutorial_code/introduction/windows_visual_studio_Opencv/Test.cpp
@@ -59,10 +59,10 @@ int main(int argc, char *argv[])
         return -1;
     }
 
-    Size refS = Size((int) captRefrnc.get(CV_CAP_PROP_FRAME_WIDTH),
-                     (int) captRefrnc.get(CV_CAP_PROP_FRAME_HEIGHT)),
-         uTSi = Size((int) captUndTst.get(CV_CAP_PROP_FRAME_WIDTH),
-                     (int) captUndTst.get(CV_CAP_PROP_FRAME_HEIGHT));
+    Size refS = Size((int) captRefrnc.get(CAP_PROP_FRAME_WIDTH),
+                     (int) captRefrnc.get(CAP_PROP_FRAME_HEIGHT)),
+         uTSi = Size((int) captUndTst.get(CAP_PROP_FRAME_WIDTH),
+                     (int) captUndTst.get(CAP_PROP_FRAME_HEIGHT));
 
     if (refS != uTSi)
     {
@@ -74,13 +74,13 @@ int main(int argc, char *argv[])
     const char* WIN_RF = "Reference";
 
     // Windows
-            namedWindow(WIN_RF, CV_WINDOW_AUTOSIZE );
-            namedWindow(WIN_UT, CV_WINDOW_AUTOSIZE );
-            cvMoveWindow(WIN_RF, 400       ,            0);		 //750,  2 (bernat =0)
-            cvMoveWindow(WIN_UT, refS.width,            0);		 //1500, 2
+            namedWindow(WIN_RF, WINDOW_AUTOSIZE );
+            namedWindow(WIN_UT, WINDOW_AUTOSIZE );
+            moveWindow(WIN_RF, 400       ,            0); //750,  2 (bernat =0)
+            moveWindow(WIN_UT, refS.width,            0); //1500, 2
 
     cout << "Frame resolution: Width=" << refS.width << "  Height=" << refS.height
-         << " of nr#: " << captRefrnc.get(CV_CAP_PROP_FRAME_COUNT) << endl;
+         << " of nr#: " << captRefrnc.get(CAP_PROP_FRAME_COUNT) << endl;
 
     cout << "PSNR trigger value " <<
           setiosflags(ios::fixed) << setprecision(3) << psnrTriggerValue << endl;
@@ -124,7 +124,7 @@ int main(int argc, char *argv[])
         imshow( WIN_RF, frameReference);
         imshow( WIN_UT, frameUnderTest);
 
-        c = (char)cvWaitKey(delay);
+        c = (char)waitKey(delay);
         if (c == 27) break;
     }
 
@@ -203,4 +203,4 @@ Scalar getMSSIM( const Mat& i1, const Mat& i2)
 
     Scalar mssim = mean( ssim_map ); // mssim = average of ssim map
     return mssim;
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp b/samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp
index a24029a232..4f6daecaa9 100644
--- a/samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp
+++ b/samples/cpp/tutorial_code/ml/non_linear_svms/non_linear_svms.cpp
@@ -127,4 +127,4 @@ int main()
     imwrite("result.png", I);	                   // save the Image
     imshow("SVM for Non-Linear Training Data", I); // show it to the user
     waitKey(0);
-}
\ No newline at end of file
+}
diff --git a/samples/cpp/tutorial_code/objectDetection/objectDetection.cpp b/samples/cpp/tutorial_code/objectDetection/objectDetection.cpp
index 0eb1752ece..e7dc3e98fc 100644
--- a/samples/cpp/tutorial_code/objectDetection/objectDetection.cpp
+++ b/samples/cpp/tutorial_code/objectDetection/objectDetection.cpp
@@ -1,11 +1,6 @@
-/**
- * @file objectDetection.cpp
- * @author A. Huaman ( based in the classic facedetect.cpp in samples/c )
- * @brief A simplified version of facedetect.cpp, show how to load a cascade classifier and how to find objects (Face + eyes) in a video stream
- */
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include <iostream>
 #include <stdio.h>
@@ -17,79 +12,73 @@ using namespace cv;
 void detectAndDisplay( Mat frame );
 
 /** Global variables */
-//-- Note, either copy these two files from opencv/data/haarscascades to your current folder, or change these locations
-string face_cascade_name = "haarcascade_frontalface_alt.xml";
-string eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
+String face_cascade_name = "haarcascade_frontalface_alt.xml";
+String eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
 CascadeClassifier face_cascade;
 CascadeClassifier eyes_cascade;
-string window_name = "Capture - Face detection";
-RNG rng(12345);
+String window_name = "Capture - Face detection";
 
-/**
- * @function main
- */
+/** @function main */
 int main( void )
 {
-  CvCapture* capture;
-  Mat frame;
+    VideoCapture capture;
+    Mat frame;
 
-  //-- 1. Load the cascades
-  if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
-  if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
+    //-- 1. Load the cascades
+    if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading face cascade\n"); return -1; };
+    if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading eyes cascade\n"); return -1; };
 
-  //-- 2. Read the video stream
-  capture = cvCaptureFromCAM( -1 );
-  if( capture )
-  {
-    for(;;)
+    //-- 2. Read the video stream
+    capture.open( -1 );
+    if ( ! capture.isOpened() ) { printf("--(!)Error opening video capture\n"); return -1; }
+
+    while ( capture.read(frame) )
     {
-      frame = cvQueryFrame( capture );
+        if( frame.empty() )
+        {
+            printf(" --(!) No captured frame -- Break!");
+            break;
+        }
 
-      //-- 3. Apply the classifier to the frame
-      if( !frame.empty() )
-       { detectAndDisplay( frame ); }
-      else
-       { printf(" --(!) No captured frame -- Break!"); break; }
-
-      int c = waitKey(10);
-      if( (char)c == 'c' ) { break; }
+        //-- 3. Apply the classifier to the frame
+        detectAndDisplay( frame );
 
+        int c = waitKey(10);
+        if( (char)c == 27 ) { break; } // escape
     }
-  }
-  return 0;
+    return 0;
 }
 
-/**
- * @function detectAndDisplay
- */
+/** @function detectAndDisplay */
 void detectAndDisplay( Mat frame )
 {
-   std::vector<Rect> faces;
-   Mat frame_gray;
+    std::vector<Rect> faces;
+    Mat frame_gray;
 
-   cvtColor( frame, frame_gray, CV_BGR2GRAY );
-   equalizeHist( frame_gray, frame_gray );
-   //-- Detect faces
-   face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
+    cvtColor( frame, frame_gray, COLOR_BGR2GRAY );
+    equalizeHist( frame_gray, frame_gray );
 
-   for( size_t i = 0; i < faces.size(); i++ )
+    //-- Detect faces
+    face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CASCADE_SCALE_IMAGE, Size(30, 30) );
+
+    for ( size_t i = 0; i < faces.size(); i++ )
     {
-      Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
-      ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2), 0, 0, 360, Scalar( 255, 0, 255 ), 2, 8, 0 );
+        Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
+        ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2 ), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
 
-      Mat faceROI = frame_gray( faces[i] );
-      std::vector<Rect> eyes;
+        Mat faceROI = frame_gray( faces[i] );
+        std::vector<Rect> eyes;
 
-      //-- In each face, detect eyes
-      eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CV_HAAR_SCALE_IMAGE, Size(30, 30) );
+        //-- In each face, detect eyes
+        eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CASCADE_SCALE_IMAGE, Size(30, 30) );
 
-      for( size_t j = 0; j < eyes.size(); j++ )
-       {
-         Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
-         int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
-         circle( frame, eye_center, radius, Scalar( 255, 0, 0 ), 3, 8, 0 );
-       }
+        for ( size_t j = 0; j < eyes.size(); j++ )
+        {
+            Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
+            int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
+            circle( frame, eye_center, radius, Scalar( 255, 0, 0 ), 4, 8, 0 );
+        }
     }
-   //-- Show what you got
-   imshow( window_name, frame );
+    //-- Show what you got
+    imshow( window_name, frame );
 }
diff --git a/samples/cpp/tutorial_code/objectDetection/objectDetection2.cpp b/samples/cpp/tutorial_code/objectDetection/objectDetection2.cpp
index 086bbaf3bf..3264858398 100644
--- a/samples/cpp/tutorial_code/objectDetection/objectDetection2.cpp
+++ b/samples/cpp/tutorial_code/objectDetection/objectDetection2.cpp
@@ -3,9 +3,9 @@
  * @author A. Huaman ( based in the classic facedetect.cpp in samples/c )
  * @brief A simplified version of facedetect.cpp, show how to load a cascade classifier and how to find objects (Face + eyes) in a video stream - Using LBP here
  */
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/objdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
 
 #include <iostream>
 #include <stdio.h>
@@ -17,46 +17,43 @@ using namespace cv;
 void detectAndDisplay( Mat frame );
 
 /** Global variables */
-string face_cascade_name = "lbpcascade_frontalface.xml";
-string eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
+String face_cascade_name = "lbpcascade_frontalface.xml";
+String eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
 CascadeClassifier face_cascade;
 CascadeClassifier eyes_cascade;
-string window_name = "Capture - Face detection";
-
-RNG rng(12345);
-
+String window_name = "Capture - Face detection";
 /**
  * @function main
  */
 int main( void )
 {
-  CvCapture* capture;
-  Mat frame;
+    VideoCapture capture;
+    Mat frame;
 
-  //-- 1. Load the cascade
-  if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
-  if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
+    //-- 1. Load the cascade
+    if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading face cascade\n"); return -1; };
+    if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading eyes cascade\n"); return -1; };
 
-  //-- 2. Read the video stream
-  capture = cvCaptureFromCAM( -1 );
-  if( capture )
-  {
-    for(;;)
+    //-- 2. Read the video stream
+    capture.open( -1 );
+    if ( ! capture.isOpened() ) { printf("--(!)Error opening video capture\n"); return -1; }
+
+    while ( capture.read(frame) )
     {
-      frame = cvQueryFrame( capture );
+        if( frame.empty() )
+        {
+            printf(" --(!) No captured frame -- Break!");
+            break;
+        }
 
-      //-- 3. Apply the classifier to the frame
-      if( !frame.empty() )
-       { detectAndDisplay( frame ); }
-      else
-       { printf(" --(!) No captured frame -- Break!"); break; }
-
-      int c = waitKey(10);
-      if( (char)c == 'c' ) { break; }
+        //-- 3. Apply the classifier to the frame
+        detectAndDisplay( frame );
 
+        //-- bail out if escape was pressed
+        int c = waitKey(10);
+        if( (char)c == 27 ) { break; }
     }
-  }
-  return 0;
+    return 0;
 }
 
 /**
@@ -64,37 +61,37 @@ int main( void )
  */
 void detectAndDisplay( Mat frame )
 {
-   std::vector<Rect> faces;
-   Mat frame_gray;
+    std::vector<Rect> faces;
+    Mat frame_gray;
 
-   cvtColor( frame, frame_gray, CV_BGR2GRAY );
-   equalizeHist( frame_gray, frame_gray );
+    cvtColor( frame, frame_gray, COLOR_BGR2GRAY );
+    equalizeHist( frame_gray, frame_gray );
 
-   //-- Detect faces
-   face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0, Size(80, 80) );
+    //-- Detect faces
+    face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0, Size(80, 80) );
 
-   for( size_t i = 0; i < faces.size(); i++ )
+    for( size_t i = 0; i < faces.size(); i++ )
     {
-      Mat faceROI = frame_gray( faces[i] );
-      std::vector<Rect> eyes;
+        Mat faceROI = frame_gray( faces[i] );
+        std::vector<Rect> eyes;
 
-      //-- In each face, detect eyes
-      eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CV_HAAR_SCALE_IMAGE, Size(30, 30) );
-      if( eyes.size() == 2)
-      {
-         //-- Draw the face
-         Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
-         ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2), 0, 0, 360, Scalar( 255, 0, 0 ), 2, 8, 0 );
+        //-- In each face, detect eyes
+        eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CASCADE_SCALE_IMAGE, Size(30, 30) );
+        if( eyes.size() == 2)
+        {
+            //-- Draw the face
+            Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
+            ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2 ), 0, 0, 360, Scalar( 255, 0, 0 ), 2, 8, 0 );
 
-         for( size_t j = 0; j < eyes.size(); j++ )
-          { //-- Draw the eyes
-            Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
-            int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
-            circle( frame, eye_center, radius, Scalar( 255, 0, 255 ), 3, 8, 0 );
-          }
-       }
+            for( size_t j = 0; j < eyes.size(); j++ )
+            { //-- Draw the eyes
+                Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
+                int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
+                circle( frame, eye_center, radius, Scalar( 255, 0, 255 ), 3, 8, 0 );
+            }
+        }
 
     }
-   //-- Show what you got
-   imshow( window_name, frame );
+    //-- Show what you got
+    imshow( window_name, frame );
 }
diff --git a/samples/cpp/tvl1_optical_flow.cpp b/samples/cpp/tvl1_optical_flow.cpp
index 4f9a02b5a9..f5e71863c2 100644
--- a/samples/cpp/tvl1_optical_flow.cpp
+++ b/samples/cpp/tvl1_optical_flow.cpp
@@ -1,8 +1,9 @@
 #include <iostream>
 #include <fstream>
 
-#include "opencv2/video/tracking.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/video.hpp"
+#include "opencv2/highgui.hpp"
 
 using namespace cv;
 using namespace std;
diff --git a/samples/cpp/video_dmtx.cpp b/samples/cpp/video_dmtx.cpp
index 4d36a38139..88dc28a418 100644
--- a/samples/cpp/video_dmtx.cpp
+++ b/samples/cpp/video_dmtx.cpp
@@ -44,7 +44,7 @@ namespace
         char filename[200];
         string window_name = "video | q or esc to quit";
         cout << "press space to save a picture. q or esc to quit" << endl;
-        namedWindow(window_name, CV_WINDOW_KEEPRATIO); //resizable window;
+        namedWindow(window_name, WINDOW_KEEPRATIO); //resizable window;
         Mat frame;
         for (;;)
         {
@@ -52,8 +52,8 @@ namespace
             if (frame.empty())
                 break;
             cv::Mat gray;
-            cv::cvtColor(frame,gray,CV_RGB2GRAY);
-            vector<string> codes;
+            cv::cvtColor(frame,gray, COLOR_RGB2GRAY);
+            vector<String> codes;
             Mat corners;
             findDataMatrix(gray, codes, corners);
             drawDataMatrixCodes(frame, codes, corners);
diff --git a/samples/cpp/video_homography.cpp b/samples/cpp/video_homography.cpp
index 01af565c46..bf2559fbaf 100644
--- a/samples/cpp/video_homography.cpp
+++ b/samples/cpp/video_homography.cpp
@@ -152,7 +152,7 @@ int main(int ac, char ** av)
 
     Mat train_desc, query_desc;
     const int DESIRED_FTRS = 500;
-    GridAdaptedFeatureDetector detector(new FastFeatureDetector(10, true), DESIRED_FTRS, 4, 4);
+    GridAdaptedFeatureDetector detector(makePtr<FastFeatureDetector>(10, true), DESIRED_FTRS, 4, 4);
 
     Mat H_prev = Mat::eye(3, 3, CV_32FC1);
     for (;;)
@@ -161,7 +161,7 @@ int main(int ac, char ** av)
         if (frame.empty())
             break;
 
-        cvtColor(frame, gray, CV_RGB2GRAY);
+        cvtColor(frame, gray, COLOR_RGB2GRAY);
 
         detector.detect(gray, query_kpts); //Find interest points
 
diff --git a/samples/cpp/videostab.cpp b/samples/cpp/videostab.cpp
index effd9e61f4..01cdcf37f7 100644
--- a/samples/cpp/videostab.cpp
+++ b/samples/cpp/videostab.cpp
@@ -4,6 +4,7 @@
 #include <sstream>
 #include <stdexcept>
 #include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
 #include "opencv2/video.hpp"
 #include "opencv2/imgproc.hpp"
 #include "opencv2/highgui.hpp"
@@ -47,7 +48,7 @@ void run()
         if (!outputPath.empty())
         {
             if (!writer.isOpened())
-                writer.open(outputPath, CV_FOURCC('X','V','I','D'),
+                writer.open(outputPath, VideoWriter::fourcc('X','V','I','D'),
                             outputFps, stabilizedFrame.size());
             writer << stabilizedFrame;
         }
@@ -192,7 +193,7 @@ public:
 
     virtual Ptr<ImageMotionEstimatorBase> build()
     {
-        MotionEstimatorRansacL2 *est = new MotionEstimatorRansacL2(motionModel(arg(prefix + "model")));
+        Ptr<MotionEstimatorRansacL2> est = makePtr<MotionEstimatorRansacL2>(motionModel(arg(prefix + "model")));
 
         RansacParams ransac = est->ransacParams();
         if (arg(prefix + "subset") != "auto")
@@ -204,10 +205,10 @@ public:
 
         est->setMinInlierRatio(argf(prefix + "min-inlier-ratio"));
 
-        Ptr<IOutlierRejector> outlierRejector = new NullOutlierRejector();
+        Ptr<IOutlierRejector> outlierRejector = makePtr<NullOutlierRejector>();
         if (arg(prefix + "local-outlier-rejection") == "yes")
         {
-            TranslationBasedLocalOutlierRejector *tblor = new TranslationBasedLocalOutlierRejector();
+            Ptr<TranslationBasedLocalOutlierRejector> tblor = makePtr<TranslationBasedLocalOutlierRejector>();
             RansacParams ransacParams = tblor->ransacParams();
             if (arg(prefix + "thresh") != "auto")
                 ransacParams.thresh = argf(prefix + "thresh");
@@ -215,17 +216,17 @@ public:
             outlierRejector = tblor;
         }
 
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
         if (gpu)
         {
-            KeypointBasedMotionEstimatorGpu *kbest = new KeypointBasedMotionEstimatorGpu(est);
+            Ptr<KeypointBasedMotionEstimatorGpu> kbest = makePtr<KeypointBasedMotionEstimatorGpu>(est);
             kbest->setOutlierRejector(outlierRejector);
             return kbest;
         }
 #endif
 
-        KeypointBasedMotionEstimator *kbest = new KeypointBasedMotionEstimator(est);
-        kbest->setDetector(new GoodFeaturesToTrackDetector(argi(prefix + "nkps")));
+        Ptr<KeypointBasedMotionEstimator> kbest = makePtr<KeypointBasedMotionEstimator>(est);
+        kbest->setDetector(makePtr<GoodFeaturesToTrackDetector>(argi(prefix + "nkps")));
         kbest->setOutlierRejector(outlierRejector);
         return kbest;
     }
@@ -243,12 +244,12 @@ public:
 
     virtual Ptr<ImageMotionEstimatorBase> build()
     {
-        MotionEstimatorL1 *est = new MotionEstimatorL1(motionModel(arg(prefix + "model")));
+        Ptr<MotionEstimatorL1> est = makePtr<MotionEstimatorL1>(motionModel(arg(prefix + "model")));
 
-        Ptr<IOutlierRejector> outlierRejector = new NullOutlierRejector();
+        Ptr<IOutlierRejector> outlierRejector = makePtr<NullOutlierRejector>();
         if (arg(prefix + "local-outlier-rejection") == "yes")
         {
-            TranslationBasedLocalOutlierRejector *tblor = new TranslationBasedLocalOutlierRejector();
+            Ptr<TranslationBasedLocalOutlierRejector> tblor = makePtr<TranslationBasedLocalOutlierRejector>();
             RansacParams ransacParams = tblor->ransacParams();
             if (arg(prefix + "thresh") != "auto")
                 ransacParams.thresh = argf(prefix + "thresh");
@@ -256,17 +257,17 @@ public:
             outlierRejector = tblor;
         }
 
-#ifdef HAVE_OPENCV_GPU
+#if defined(HAVE_OPENCV_GPUIMGPROC) && defined(HAVE_OPENCV_GPU) && defined(HAVE_OPENCV_GPUOPTFLOW)
         if (gpu)
         {
-            KeypointBasedMotionEstimatorGpu *kbest = new KeypointBasedMotionEstimatorGpu(est);
+            Ptr<KeypointBasedMotionEstimatorGpu> kbest = makePtr<KeypointBasedMotionEstimatorGpu>(est);
             kbest->setOutlierRejector(outlierRejector);
             return kbest;
         }
 #endif
 
-        KeypointBasedMotionEstimator *kbest = new KeypointBasedMotionEstimator(est);
-        kbest->setDetector(new GoodFeaturesToTrackDetector(argi(prefix + "nkps")));
+        Ptr<KeypointBasedMotionEstimator> kbest = makePtr<KeypointBasedMotionEstimator>(est);
+        kbest->setDetector(makePtr<GoodFeaturesToTrackDetector>(argi(prefix + "nkps")));
         kbest->setOutlierRejector(outlierRejector);
         return kbest;
     }
@@ -362,7 +363,7 @@ int main(int argc, const char **argv)
 
         // get source video parameters
 
-        VideoFileSource *source = new VideoFileSource(inputPath);
+        Ptr<VideoFileSource> source = makePtr<VideoFileSource>(inputPath);
         cout << "frame count (rough): " << source->count() << endl;
         if (arg("fps") == "auto")
             outputFps = source->fps();
@@ -373,15 +374,15 @@ int main(int argc, const char **argv)
 
         Ptr<IMotionEstimatorBuilder> motionEstBuilder;
         if (arg("lin-prog-motion-est") == "yes")
-            motionEstBuilder = new MotionEstimatorL1Builder(cmd, arg("gpu") == "yes");
+            motionEstBuilder.reset(new MotionEstimatorL1Builder(cmd, arg("gpu") == "yes"));
         else
-            motionEstBuilder = new MotionEstimatorRansacL2Builder(cmd, arg("gpu") == "yes");
+            motionEstBuilder.reset(new MotionEstimatorRansacL2Builder(cmd, arg("gpu") == "yes"));
 
         Ptr<IMotionEstimatorBuilder> wsMotionEstBuilder;
         if (arg("ws-lp") == "yes")
-            wsMotionEstBuilder = new MotionEstimatorL1Builder(cmd, arg("gpu") == "yes", "ws-");
+            wsMotionEstBuilder.reset(new MotionEstimatorL1Builder(cmd, arg("gpu") == "yes", "ws-"));
         else
-            wsMotionEstBuilder = new MotionEstimatorRansacL2Builder(cmd, arg("gpu") == "yes", "ws-");
+            wsMotionEstBuilder.reset(new MotionEstimatorRansacL2Builder(cmd, arg("gpu") == "yes", "ws-"));
 
         // determine whether we must use one pass or two pass stabilizer
         bool isTwoPass =
@@ -399,7 +400,7 @@ int main(int argc, const char **argv)
 
             if (arg("lin-prog-stab") == "yes")
             {
-                LpMotionStabilizer *stab = new LpMotionStabilizer();
+                Ptr<LpMotionStabilizer> stab = makePtr<LpMotionStabilizer>();
                 stab->setFrameSize(Size(source->width(), source->height()));
                 stab->setTrimRatio(arg("lps-trim-ratio") == "auto" ? argf("trim-ratio") : argf("lps-trim-ratio"));
                 stab->setWeight1(argf("lps-w1"));
@@ -409,18 +410,18 @@ int main(int argc, const char **argv)
                 twoPassStabilizer->setMotionStabilizer(stab);
             }
             else if (arg("stdev") == "auto")
-                twoPassStabilizer->setMotionStabilizer(new GaussianMotionFilter(argi("radius")));
+                twoPassStabilizer->setMotionStabilizer(makePtr<GaussianMotionFilter>(argi("radius")));
             else
-                twoPassStabilizer->setMotionStabilizer(new GaussianMotionFilter(argi("radius"), argf("stdev")));
+                twoPassStabilizer->setMotionStabilizer(makePtr<GaussianMotionFilter>(argi("radius"), argf("stdev")));
 
             // init wobble suppressor if necessary
 
             if (arg("wobble-suppress") == "yes")
             {
-                MoreAccurateMotionWobbleSuppressorBase *ws = new MoreAccurateMotionWobbleSuppressor();
+                Ptr<MoreAccurateMotionWobbleSuppressorBase> ws = makePtr<MoreAccurateMotionWobbleSuppressor>();
                 if (arg("gpu") == "yes")
 #ifdef HAVE_OPENCV_GPU
-                    ws = new MoreAccurateMotionWobbleSuppressorGpu();
+                    ws = makePtr<MoreAccurateMotionWobbleSuppressorGpu>();
 #else
                     throw runtime_error("OpenCV is built without GPU support");
 #endif
@@ -432,12 +433,12 @@ int main(int argc, const char **argv)
                 MotionModel model = ws->motionEstimator()->motionModel();
                 if (arg("load-motions2") != "no")
                 {
-                    ws->setMotionEstimator(new FromFileMotionReader(arg("load-motions2")));
+                    ws->setMotionEstimator(makePtr<FromFileMotionReader>(arg("load-motions2")));
                     ws->motionEstimator()->setMotionModel(model);
                 }
                 if (arg("save-motions2") != "no")
                 {
-                    ws->setMotionEstimator(new ToFileMotionWriter(arg("save-motions2"), ws->motionEstimator()));
+                    ws->setMotionEstimator(makePtr<ToFileMotionWriter>(arg("save-motions2"), ws->motionEstimator()));
                     ws->motionEstimator()->setMotionModel(model);
                 }
             }
@@ -449,26 +450,26 @@ int main(int argc, const char **argv)
             OnePassStabilizer *onePassStabilizer = new OnePassStabilizer();
             stabilizer = onePassStabilizer;
             if (arg("stdev") == "auto")
-                onePassStabilizer->setMotionFilter(new GaussianMotionFilter(argi("radius")));
+                onePassStabilizer->setMotionFilter(makePtr<GaussianMotionFilter>(argi("radius")));
             else
-                onePassStabilizer->setMotionFilter(new GaussianMotionFilter(argi("radius"), argf("stdev")));
+                onePassStabilizer->setMotionFilter(makePtr<GaussianMotionFilter>(argi("radius"), argf("stdev")));
         }
 
         stabilizer->setFrameSource(source);
         stabilizer->setMotionEstimator(motionEstBuilder->build());
 
         // cast stabilizer to simple frame source interface to read stabilized frames
-        stabilizedFrames = dynamic_cast<IFrameSource*>(stabilizer);
+        stabilizedFrames.reset(dynamic_cast<IFrameSource*>(stabilizer));
 
         MotionModel model = stabilizer->motionEstimator()->motionModel();
         if (arg("load-motions") != "no")
         {
-            stabilizer->setMotionEstimator(new FromFileMotionReader(arg("load-motions")));
+            stabilizer->setMotionEstimator(makePtr<FromFileMotionReader>(arg("load-motions")));
             stabilizer->motionEstimator()->setMotionModel(model);
         }
         if (arg("save-motions") != "no")
         {
-            stabilizer->setMotionEstimator(new ToFileMotionWriter(arg("save-motions"), stabilizer->motionEstimator()));
+            stabilizer->setMotionEstimator(makePtr<ToFileMotionWriter>(arg("save-motions"), stabilizer->motionEstimator()));
             stabilizer->motionEstimator()->setMotionModel(model);
         }
 
@@ -477,7 +478,7 @@ int main(int argc, const char **argv)
         // init deblurer
         if (arg("deblur") == "yes")
         {
-            WeightingDeblurer *deblurer = new WeightingDeblurer();
+            Ptr<WeightingDeblurer> deblurer = makePtr<WeightingDeblurer>();
             deblurer->setRadius(argi("radius"));
             deblurer->setSensitivity(argf("deblur-sens"));
             stabilizer->setDeblurer(deblurer);
@@ -502,22 +503,22 @@ int main(int argc, const char **argv)
         Ptr<InpainterBase> inpainters_(inpainters);
         if (arg("mosaic") == "yes")
         {
-            ConsistentMosaicInpainter *inp = new ConsistentMosaicInpainter();
+            Ptr<ConsistentMosaicInpainter> inp = makePtr<ConsistentMosaicInpainter>();
             inp->setStdevThresh(argf("mosaic-stdev"));
             inpainters->pushBack(inp);
         }
         if (arg("motion-inpaint") == "yes")
         {
-            MotionInpainter *inp = new MotionInpainter();
+            Ptr<MotionInpainter> inp = makePtr<MotionInpainter>();
             inp->setDistThreshold(argf("mi-dist-thresh"));
             inpainters->pushBack(inp);
         }
         if (arg("color-inpaint") == "average")
-            inpainters->pushBack(new ColorAverageInpainter());
+            inpainters->pushBack(makePtr<ColorAverageInpainter>());
         else if (arg("color-inpaint") == "ns")
-            inpainters->pushBack(new ColorInpainter(INPAINT_NS, argd("ci-radius")));
+            inpainters->pushBack(makePtr<ColorInpainter>(int(INPAINT_NS), argd("ci-radius")));
         else if (arg("color-inpaint") == "telea")
-            inpainters->pushBack(new ColorInpainter(INPAINT_TELEA, argd("ci-radius")));
+            inpainters->pushBack(makePtr<ColorInpainter>(int(INPAINT_TELEA), argd("ci-radius")));
         else if (arg("color-inpaint") != "no")
             throw runtime_error("unknown color inpainting method: " + arg("color-inpaint"));
         if (!inpainters->empty())
diff --git a/samples/cpp/watershed.cpp b/samples/cpp/watershed.cpp
index de7f5256f9..cd47505045 100644
--- a/samples/cpp/watershed.cpp
+++ b/samples/cpp/watershed.cpp
@@ -1,5 +1,6 @@
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
 
 #include <cstdio>
 #include <iostream>
@@ -28,11 +29,11 @@ static void onMouse( int event, int x, int y, int flags, void* )
 {
     if( x < 0 || x >= img.cols || y < 0 || y >= img.rows )
         return;
-    if( event == CV_EVENT_LBUTTONUP || !(flags & CV_EVENT_FLAG_LBUTTON) )
+    if( event == EVENT_LBUTTONUP || !(flags & EVENT_FLAG_LBUTTON) )
         prevPt = Point(-1,-1);
-    else if( event == CV_EVENT_LBUTTONDOWN )
+    else if( event == EVENT_LBUTTONDOWN )
         prevPt = Point(x,y);
-    else if( event == CV_EVENT_MOUSEMOVE && (flags & CV_EVENT_FLAG_LBUTTON) )
+    else if( event == EVENT_MOUSEMOVE && (flags & EVENT_FLAG_LBUTTON) )
     {
         Point pt(x, y);
         if( prevPt.x < 0 )
@@ -58,8 +59,8 @@ int main( int argc, char** argv )
     namedWindow( "image", 1 );
 
     img0.copyTo(img);
-    cvtColor(img, markerMask, CV_BGR2GRAY);
-    cvtColor(markerMask, imgGray, CV_GRAY2BGR);
+    cvtColor(img, markerMask, COLOR_BGR2GRAY);
+    cvtColor(markerMask, imgGray, COLOR_GRAY2BGR);
     markerMask = Scalar::all(0);
     imshow( "image", img );
     setMouseCallback( "image", onMouse, 0 );
@@ -84,7 +85,7 @@ int main( int argc, char** argv )
             vector<vector<Point> > contours;
             vector<Vec4i> hierarchy;
 
-            findContours(markerMask, contours, hierarchy, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE);
+            findContours(markerMask, contours, hierarchy, RETR_CCOMP, CHAIN_APPROX_SIMPLE);
 
             if( contours.empty() )
                 continue;
diff --git a/samples/gpu/CMakeLists.txt b/samples/gpu/CMakeLists.txt
index 6d20fc34d4..b953627b1b 100644
--- a/samples/gpu/CMakeLists.txt
+++ b/samples/gpu/CMakeLists.txt
@@ -1,8 +1,10 @@
 SET(OPENCV_GPU_SAMPLES_REQUIRED_DEPS opencv_core opencv_flann opencv_imgproc opencv_highgui
                                      opencv_ml opencv_video opencv_objdetect opencv_features2d
                                      opencv_calib3d opencv_legacy opencv_contrib opencv_gpu
-                                     opencv_nonfree opencv_softcascade)
-
+                                     opencv_nonfree opencv_softcascade opencv_superres
+                                     opencv_gpuarithm opencv_gpufilters opencv_gpuwarping opencv_gpuimgproc
+                                     opencv_gpufeatures2d opencv_gpuoptflow opencv_gpubgsegm
+                                     opencv_gpustereo opencv_gpulegacy)
 ocv_check_dependencies(${OPENCV_GPU_SAMPLES_REQUIRED_DEPS})
 
 if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
@@ -17,10 +19,22 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
     "${OpenCV_SOURCE_DIR}/modules/gpu/src/nvidia/core"
     )
 
+  if(HAVE_opencv_nonfree)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/nonfree/include")
+  endif()
+
+  if(HAVE_opencv_gpucodec)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/gpucodec/include")
+  endif()
+
   if(HAVE_CUDA)
     ocv_include_directories(${CUDA_INCLUDE_DIRS})
   endif()
 
+  if(HAVE_OPENCL)
+    ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/ocl/include")
+  endif()
+
   if(CMAKE_COMPILER_IS_GNUCXX AND NOT ENABLE_NOISY_WARNINGS)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-unused-function")
   endif()
@@ -34,6 +48,21 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
 
     target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_GPU_SAMPLES_REQUIRED_DEPS})
 
+    if(HAVE_CUDA)
+        target_link_libraries(${the_target} ${CUDA_CUDA_LIBRARY})
+    endif()
+
+    if(HAVE_opencv_nonfree)
+      target_link_libraries(${the_target} opencv_nonfree)
+    endif()
+    if(HAVE_opencv_gpucodec)
+      target_link_libraries(${the_target} opencv_gpucodec)
+    endif()
+
+    if(HAVE_OPENCL)
+      target_link_libraries(${the_target} opencv_ocl)
+    endif()
+
     set_target_properties(${the_target} PROPERTIES
       OUTPUT_NAME "${project}-example-${name}"
       PROJECT_LABEL "(EXAMPLE_${project_upper}) ${name}")
@@ -67,4 +96,3 @@ if (INSTALL_C_EXAMPLES AND NOT WIN32)
           DESTINATION share/OpenCV/samples/${project}
           PERMISSIONS OWNER_READ GROUP_READ WORLD_READ)
 endif()
-
diff --git a/samples/gpu/bgfg_segm.cpp b/samples/gpu/bgfg_segm.cpp
index 9456b15e88..1b8e53271c 100644
--- a/samples/gpu/bgfg_segm.cpp
+++ b/samples/gpu/bgfg_segm.cpp
@@ -1,9 +1,16 @@
 #include <iostream>
 #include <string>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/gpu.hpp"
+#include "opencv2/highgui.hpp"
+
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_NONFREE
+#  include "opencv2/nonfree/gpu.hpp"
+#endif
 
 using namespace std;
 using namespace cv;
@@ -11,11 +18,10 @@ using namespace cv::gpu;
 
 enum Method
 {
-    FGD_STAT,
     MOG,
     MOG2,
-    VIBE,
-    GMG
+    GMG,
+    FGD_STAT
 };
 
 int main(int argc, const char** argv)
@@ -23,7 +29,7 @@ int main(int argc, const char** argv)
     cv::CommandLineParser cmd(argc, argv,
         "{ c camera |             | use camera }"
         "{ f file   | 768x576.avi | input video file }"
-        "{ m method | mog         | method (fgd, mog, mog2, vibe, gmg) }"
+        "{ m method | mog         | method (mog, mog2, gmg, fgd) }"
         "{ h help   |             | print help message }");
 
     if (cmd.has("help") || !cmd.check())
@@ -37,13 +43,19 @@ int main(int argc, const char** argv)
     string file = cmd.get<string>("file");
     string method = cmd.get<string>("method");
 
-    if (method != "fgd" && method != "mog" && method != "mog2" && method != "vibe" && method != "gmg")
+    if (method != "mog"
+        && method != "mog2"
+        && method != "gmg"
+        && method != "fgd")
     {
         cerr << "Incorrect method" << endl;
         return -1;
     }
 
-    Method m = method == "fgd" ? FGD_STAT : method == "mog" ? MOG : method == "mog2" ? MOG2 : method == "vibe" ? VIBE : GMG;
+    Method m = method == "mog" ? MOG :
+               method == "mog2" ? MOG2 :
+               method == "fgd" ? FGD_STAT :
+                                  GMG;
 
     VideoCapture cap;
 
@@ -63,12 +75,10 @@ int main(int argc, const char** argv)
 
     GpuMat d_frame(frame);
 
-    FGDStatModel fgd_stat;
-    MOG_GPU mog;
-    MOG2_GPU mog2;
-    VIBE_GPU vibe;
-    GMG_GPU gmg;
-    gmg.numInitializationFrames = 40;
+    Ptr<BackgroundSubtractor> mog = gpu::createBackgroundSubtractorMOG();
+    Ptr<BackgroundSubtractor> mog2 = gpu::createBackgroundSubtractorMOG2();
+    Ptr<BackgroundSubtractor> gmg = gpu::createBackgroundSubtractorGMG(40);
+    Ptr<BackgroundSubtractor> fgd = gpu::createBackgroundSubtractorFGD();
 
     GpuMat d_fgmask;
     GpuMat d_fgimg;
@@ -80,32 +90,30 @@ int main(int argc, const char** argv)
 
     switch (m)
     {
-    case FGD_STAT:
-        fgd_stat.create(d_frame);
-        break;
-
     case MOG:
-        mog(d_frame, d_fgmask, 0.01f);
+        mog->apply(d_frame, d_fgmask, 0.01);
         break;
 
     case MOG2:
-        mog2(d_frame, d_fgmask);
-        break;
-
-    case VIBE:
-        vibe.initialize(d_frame);
+        mog2->apply(d_frame, d_fgmask);
         break;
 
     case GMG:
-        gmg.initialize(d_frame.size());
+        gmg->apply(d_frame, d_fgmask);
+        break;
+
+    case FGD_STAT:
+        fgd->apply(d_frame, d_fgmask);
         break;
     }
 
     namedWindow("image", WINDOW_NORMAL);
     namedWindow("foreground mask", WINDOW_NORMAL);
     namedWindow("foreground image", WINDOW_NORMAL);
-    if (m != VIBE && m != GMG)
+    if (m != GMG)
+    {
         namedWindow("mean background image", WINDOW_NORMAL);
+    }
 
     for(;;)
     {
@@ -119,28 +127,23 @@ int main(int argc, const char** argv)
         //update the model
         switch (m)
         {
-        case FGD_STAT:
-            fgd_stat.update(d_frame);
-            d_fgmask = fgd_stat.foreground;
-            d_bgimg = fgd_stat.background;
-            break;
-
         case MOG:
-            mog(d_frame, d_fgmask, 0.01f);
-            mog.getBackgroundImage(d_bgimg);
+            mog->apply(d_frame, d_fgmask, 0.01);
+            mog->getBackgroundImage(d_bgimg);
             break;
 
         case MOG2:
-            mog2(d_frame, d_fgmask);
-            mog2.getBackgroundImage(d_bgimg);
-            break;
-
-        case VIBE:
-            vibe(d_frame, d_fgmask);
+            mog2->apply(d_frame, d_fgmask);
+            mog2->getBackgroundImage(d_bgimg);
             break;
 
         case GMG:
-            gmg(d_frame, d_fgmask);
+            gmg->apply(d_frame, d_fgmask);
+            break;
+
+        case FGD_STAT:
+            fgd->apply(d_frame, d_fgmask);
+            fgd->getBackgroundImage(d_bgimg);
             break;
         }
 
diff --git a/samples/gpu/brox_optical_flow.cpp b/samples/gpu/brox_optical_flow.cpp
index 56832c4def..1fb85c9038 100644
--- a/samples/gpu/brox_optical_flow.cpp
+++ b/samples/gpu/brox_optical_flow.cpp
@@ -3,10 +3,10 @@
 #include <string>
 #include <cctype>
 
-#include "cvconfig.h"
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/gpu.hpp"
 
 using namespace std;
 using namespace cv;
@@ -85,8 +85,8 @@ int main(int argc, const char* argv[])
 
         Mat frame0Gray, frame1Gray;
 
-        cvtColor(frame0Color, frame0Gray, COLOR_BGR2GRAY);
-        cvtColor(frame1Color, frame1Gray, COLOR_BGR2GRAY);
+        cv::cvtColor(frame0Color, frame0Gray, COLOR_BGR2GRAY);
+        cv::cvtColor(frame1Color, frame1Gray, COLOR_BGR2GRAY);
 
         GpuMat d_frame0(frame0Gray);
         GpuMat d_frame1(frame1Gray);
diff --git a/samples/gpu/cascadeclassifier.cpp b/samples/gpu/cascadeclassifier.cpp
index 412183f107..e27186c271 100644
--- a/samples/gpu/cascadeclassifier.cpp
+++ b/samples/gpu/cascadeclassifier.cpp
@@ -24,12 +24,11 @@ static void help()
 }
 
 
-template<class T>
-void convertAndResize(const T& src, T& gray, T& resized, double scale)
+static void convertAndResize(const Mat& src, Mat& gray, Mat& resized, double scale)
 {
     if (src.channels() == 3)
     {
-        cvtColor( src, gray, CV_BGR2GRAY );
+        cv::cvtColor( src, gray, COLOR_BGR2GRAY );
     }
     else
     {
@@ -40,7 +39,30 @@ void convertAndResize(const T& src, T& gray, T& resized, double scale)
 
     if (scale != 1)
     {
-        resize(gray, resized, sz);
+        cv::resize(gray, resized, sz);
+    }
+    else
+    {
+        resized = gray;
+    }
+}
+
+static void convertAndResize(const GpuMat& src, GpuMat& gray, GpuMat& resized, double scale)
+{
+    if (src.channels() == 3)
+    {
+        cv::gpu::cvtColor( src, gray, COLOR_BGR2GRAY );
+    }
+    else
+    {
+        gray = src;
+    }
+
+    Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
+
+    if (scale != 1)
+    {
+        cv::gpu::resize(gray, resized, sz);
     }
     else
     {
@@ -59,15 +81,15 @@ static void matPrint(Mat &img, int lineOffsY, Scalar fontColor, const string &ss
     Point org;
     org.x = 1;
     org.y = 3 * fontSize.height * (lineOffsY + 1) / 2;
-    putText(img, ss, org, fontFace, fontScale, CV_RGB(0,0,0), 5*fontThickness/2, 16);
+    putText(img, ss, org, fontFace, fontScale, Scalar(0,0,0), 5*fontThickness/2, 16);
     putText(img, ss, org, fontFace, fontScale, fontColor, fontThickness, 16);
 }
 
 
 static void displayState(Mat &canvas, bool bHelp, bool bGpu, bool bLargestFace, bool bFilter, double fps)
 {
-    Scalar fontColorRed = CV_RGB(255,0,0);
-    Scalar fontColorNV  = CV_RGB(118,185,0);
+    Scalar fontColorRed = Scalar(255,0,0);
+    Scalar fontColorNV  = Scalar(118,185,0);
 
     ostringstream ss;
     ss << "FPS = " << setprecision(1) << fixed << fps;
@@ -216,7 +238,7 @@ int main(int argc, const char *argv[])
 
         if (useGPU)
         {
-            cascade_gpu.visualizeInPlace = true;
+            //cascade_gpu.visualizeInPlace = true;
             cascade_gpu.findLargestObject = findLargestObject;
 
             detections_num = cascade_gpu.detectMultiScale(resized_gpu, facesBuf_gpu, 1.2,
@@ -228,8 +250,8 @@ int main(int argc, const char *argv[])
             Size minSize = cascade_gpu.getClassifierSize();
             cascade_cpu.detectMultiScale(resized_cpu, facesBuf_cpu, 1.2,
                                          (filterRects || findLargestObject) ? 4 : 0,
-                                         (findLargestObject ? CV_HAAR_FIND_BIGGEST_OBJECT : 0)
-                                            | CV_HAAR_SCALE_IMAGE,
+                                         (findLargestObject ? CASCADE_FIND_BIGGEST_OBJECT : 0)
+                                            | CASCADE_SCALE_IMAGE,
                                          minSize);
             detections_num = (int)facesBuf_cpu.size();
         }
@@ -245,6 +267,11 @@ int main(int argc, const char *argv[])
         if (useGPU)
         {
             resized_gpu.download(resized_cpu);
+
+             for (int i = 0; i < detections_num; ++i)
+             {
+                rectangle(resized_cpu, faces_downloaded.ptr<cv::Rect>()[i], Scalar(255));
+             }
         }
 
         tm.stop();
@@ -267,7 +294,7 @@ int main(int argc, const char *argv[])
         }
         cout << endl;
 
-        cvtColor(resized_cpu, frameDisp, CV_GRAY2BGR);
+        cv::cvtColor(resized_cpu, frameDisp, COLOR_GRAY2BGR);
         displayState(frameDisp, helpScreen, useGPU, findLargestObject, filterRects, fps);
         imshow("result", frameDisp);
 
diff --git a/samples/gpu/cascadeclassifier_nvidia_api.cpp b/samples/gpu/cascadeclassifier_nvidia_api.cpp
index 99c95ab977..3b27de58c3 100644
--- a/samples/gpu/cascadeclassifier_nvidia_api.cpp
+++ b/samples/gpu/cascadeclassifier_nvidia_api.cpp
@@ -8,26 +8,37 @@
 #include <cstdio>
 #include "opencv2/gpu/gpu.hpp"
 #include "opencv2/highgui/highgui.hpp"
+#include "opencv2/objdetect/objdetect.hpp"
+#include "opencv2/objdetect/objdetect_c.h"
 
 #ifdef HAVE_CUDA
-#include "NCVHaarObjectDetection.hpp"
+#include "opencv2/gpulegacy.hpp"
 #endif
 
 using namespace std;
 using namespace cv;
 
 
-#if !defined(HAVE_CUDA)
+#if !defined(HAVE_CUDA) || defined(__arm__)
+
 int main( int, const char** )
 {
-    cout << "Please compile the library with CUDA support" << endl;
-    return -1;
+#if !defined(HAVE_CUDA)
+    std::cout << "CUDA support is required (CMake key 'WITH_CUDA' must be true)." << std::endl;
+#endif
+
+#if defined(__arm__)
+    std::cout << "Unsupported for ARM CUDA library." << std::endl;
+#endif
+
+    return 0;
 }
+
 #else
 
 
 const Size2i preferredVideoFrameSize(640, 480);
-const string wndTitle = "NVIDIA Computer Vision :: Haar Classifiers Cascade";
+const cv::String wndTitle = "NVIDIA Computer Vision :: Haar Classifiers Cascade";
 
 
 static void matPrint(Mat &img, int lineOffsY, Scalar fontColor, const string &ss)
@@ -40,15 +51,15 @@ static void matPrint(Mat &img, int lineOffsY, Scalar fontColor, const string &ss
     Point org;
     org.x = 1;
     org.y = 3 * fontSize.height * (lineOffsY + 1) / 2;
-    putText(img, ss, org, fontFace, fontScale, CV_RGB(0,0,0), 5*fontThickness/2, 16);
+    putText(img, ss, org, fontFace, fontScale, Scalar(0,0,0), 5*fontThickness/2, 16);
     putText(img, ss, org, fontFace, fontScale, fontColor, fontThickness, 16);
 }
 
 
 static void displayState(Mat &canvas, bool bHelp, bool bGpu, bool bLargestFace, bool bFilter, double fps)
 {
-    Scalar fontColorRed = CV_RGB(255,0,0);
-    Scalar fontColorNV  = CV_RGB(118,185,0);
+    Scalar fontColorRed(0,0,255);
+    Scalar fontColorNV(0,185,118);
 
     ostringstream ss;
     ss << "FPS = " << setprecision(1) << fixed << fps;
@@ -286,7 +297,7 @@ int main(int argc, const char** argv)
     do
     {
         Mat gray;
-        cvtColor((image.empty() ? frame : image), gray, CV_BGR2GRAY);
+        cvtColor((image.empty() ? frame : image), gray, cv::COLOR_BGR2GRAY);
 
         //
         // process
@@ -334,12 +345,12 @@ int main(int argc, const char** argv)
 
         avgTime = (Ncv32f)ncvEndQueryTimerMs(timer);
 
-        cvtColor(gray, frameDisp, CV_GRAY2BGR);
+        cvtColor(gray, frameDisp, cv::COLOR_GRAY2BGR);
         displayState(frameDisp, bHelpScreen, bUseGPU, bLargestObject, bFilterRects, 1000.0f / avgTime);
         imshow(wndTitle, frameDisp);
 
         //handle input
-        switch (cvWaitKey(3))
+        switch (cv::waitKey(3))
         {
         case ' ':
             bUseGPU = !bUseGPU;
@@ -372,7 +383,7 @@ int main(int argc, const char** argv)
         }
     } while (!bQuit);
 
-    cvDestroyWindow(wndTitle.c_str());
+    cv::destroyWindow(wndTitle);
 
     return 0;
 }
diff --git a/samples/gpu/driver_api_multi.cpp b/samples/gpu/driver_api_multi.cpp
index 2d743f0e9c..e78f7e54fd 100644
--- a/samples/gpu/driver_api_multi.cpp
+++ b/samples/gpu/driver_api_multi.cpp
@@ -11,7 +11,19 @@
 #include "opencv2/core/core.hpp"
 #include "opencv2/gpu/gpu.hpp"
 
-#if !defined(HAVE_CUDA) || !defined(HAVE_TBB)
+#ifdef HAVE_TBB
+#  include "tbb/tbb_stddef.h"
+#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#    include "tbb/tbb.h"
+#    include "tbb/task.h"
+#    undef min
+#    undef max
+#  else
+#    undef HAVE_TBB
+#  endif
+#endif
+
+#if !defined(HAVE_CUDA) || !defined(HAVE_TBB) || defined(__arm__)
 
 int main()
 {
@@ -23,6 +35,10 @@ int main()
     std::cout << "TBB support is required (CMake key 'WITH_TBB' must be true).\n";
 #endif
 
+#if defined(__arm__)
+    std::cout << "Unsupported for ARM CUDA library." << std::endl;
+#endif
+
     return 0;
 }
 
@@ -30,7 +46,6 @@ int main()
 
 #include <cuda.h>
 #include <cuda_runtime.h>
-#include "opencv2/core/internal.hpp" // For TBB wrappers
 
 using namespace std;
 using namespace cv;
@@ -96,7 +111,7 @@ int main()
 
     // Execute calculation in two threads using two GPUs
     int devices[] = {0, 1};
-    parallel_do(devices, devices + 2, Worker());
+    tbb::parallel_do(devices, devices + 2, Worker());
 
     destroyContexts();
     return 0;
@@ -115,15 +130,15 @@ void Worker::operator()(int device_id) const
     rng.fill(src, RNG::UNIFORM, 0, 1);
 
     // CPU works
-    transpose(src, dst);
+    cv::transpose(src, dst);
 
     // GPU works
     GpuMat d_src(src);
     GpuMat d_dst;
-    transpose(d_src, d_dst);
+    gpu::transpose(d_src, d_dst);
 
     // Check results
-    bool passed = norm(dst - Mat(d_dst), NORM_INF) < 1e-3;
+    bool passed = cv::norm(dst - Mat(d_dst), NORM_INF) < 1e-3;
     std::cout << "GPU #" << device_id << " (" << DeviceInfo().name() << "): "
         << (passed ? "passed" : "FAILED") << endl;
 
diff --git a/samples/gpu/driver_api_stereo_multi.cpp b/samples/gpu/driver_api_stereo_multi.cpp
index b8f99e810f..d40c20c1e9 100644
--- a/samples/gpu/driver_api_stereo_multi.cpp
+++ b/samples/gpu/driver_api_stereo_multi.cpp
@@ -13,7 +13,19 @@
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/gpu/gpu.hpp"
 
-#if !defined(HAVE_CUDA) || !defined(HAVE_TBB)
+#ifdef HAVE_TBB
+#  include "tbb/tbb_stddef.h"
+#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#    include "tbb/tbb.h"
+#    include "tbb/task.h"
+#    undef min
+#    undef max
+#  else
+#    undef HAVE_TBB
+#  endif
+#endif
+
+#if !defined(HAVE_CUDA) || !defined(HAVE_TBB) || defined(__arm__)
 
 int main()
 {
@@ -25,6 +37,10 @@ int main()
     std::cout << "TBB support is required (CMake key 'WITH_TBB' must be true).\n";
 #endif
 
+#if defined(__arm__)
+    std::cout << "Unsupported for ARM CUDA library." << std::endl;
+#endif
+
     return 0;
 }
 
@@ -32,7 +48,6 @@ int main()
 
 #include <cuda.h>
 #include <cuda_runtime.h>
-#include "opencv2/core/internal.hpp" // For TBB wrappers
 
 using namespace std;
 using namespace cv;
@@ -70,12 +85,9 @@ void inline contextOff()
 // GPUs data
 GpuMat d_left[2];
 GpuMat d_right[2];
-StereoBM_GPU* bm[2];
+Ptr<gpu::StereoBM> bm[2];
 GpuMat d_result[2];
 
-// CPU result
-Mat result;
-
 static void printHelp()
 {
     std::cout << "Usage: driver_api_stereo_multi_gpu --left <left_image> --right <right_image>\n";
@@ -116,12 +128,12 @@ int main(int argc, char** argv)
     {
         if (string(argv[i]) == "--left")
         {
-            left = imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE);
+            left = imread(argv[++i], cv::IMREAD_GRAYSCALE);
             CV_Assert(!left.empty());
         }
         else if (string(argv[i]) == "--right")
         {
-            right = imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE);
+            right = imread(argv[++i], cv::IMREAD_GRAYSCALE);
             CV_Assert(!right.empty());
         }
         else if (string(argv[i]) == "--help")
@@ -150,19 +162,19 @@ int main(int argc, char** argv)
     contextOn(0);
     d_left[0].upload(left.rowRange(0, left.rows / 2));
     d_right[0].upload(right.rowRange(0, right.rows / 2));
-    bm[0] = new StereoBM_GPU();
+    bm[0] = gpu::createStereoBM();
     contextOff();
 
     // Split source images for processing on the GPU #1
     contextOn(1);
     d_left[1].upload(left.rowRange(left.rows / 2, left.rows));
     d_right[1].upload(right.rowRange(right.rows / 2, right.rows));
-    bm[1] = new StereoBM_GPU();
+    bm[1] = gpu::createStereoBM();
     contextOff();
 
     // Execute calculation in two threads using two GPUs
     int devices[] = {0, 1};
-    parallel_do(devices, devices + 2, Worker());
+    tbb::parallel_do(devices, devices + 2, Worker());
 
     // Release the first GPU resources
     contextOn(0);
@@ -170,7 +182,7 @@ int main(int argc, char** argv)
     d_left[0].release();
     d_right[0].release();
     d_result[0].release();
-    delete bm[0];
+    bm[0].release();
     contextOff();
 
     // Release the second GPU resources
@@ -179,7 +191,7 @@ int main(int argc, char** argv)
     d_left[1].release();
     d_right[1].release();
     d_result[1].release();
-    delete bm[1];
+    bm[1].release();
     contextOff();
 
     waitKey();
@@ -192,8 +204,7 @@ void Worker::operator()(int device_id) const
 {
     contextOn(device_id);
 
-    bm[device_id]->operator()(d_left[device_id], d_right[device_id],
-                              d_result[device_id]);
+    bm[device_id]->compute(d_left[device_id], d_right[device_id], d_result[device_id]);
 
     std::cout << "GPU #" << device_id << " (" << DeviceInfo().name()
         << "): finished\n";
diff --git a/samples/gpu/farneback_optical_flow.cpp b/samples/gpu/farneback_optical_flow.cpp
index 77ee49e1a1..c2a5d411e4 100644
--- a/samples/gpu/farneback_optical_flow.cpp
+++ b/samples/gpu/farneback_optical_flow.cpp
@@ -2,10 +2,11 @@
 #include <vector>
 #include <sstream>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/video/video.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/gpu.hpp"
 
 using namespace std;
 using namespace cv;
@@ -21,9 +22,9 @@ inline T mapVal(T x, T a, T b, T c, T d)
 static void colorizeFlow(const Mat &u, const Mat &v, Mat &dst)
 {
     double uMin, uMax;
-    minMaxLoc(u, &uMin, &uMax, 0, 0);
+    cv::minMaxLoc(u, &uMin, &uMax, 0, 0);
     double vMin, vMax;
-    minMaxLoc(v, &vMin, &vMax, 0, 0);
+    cv::minMaxLoc(v, &vMin, &vMax, 0, 0);
     uMin = ::abs(uMin); uMax = ::abs(uMax);
     vMin = ::abs(vMin); vMax = ::abs(vMax);
     float dMax = static_cast<float>(::max(::max(uMin, uMax), ::max(vMin, vMax)));
diff --git a/samples/gpu/generalized_hough.cpp b/samples/gpu/generalized_hough.cpp
index c41e7903c8..1863085256 100644
--- a/samples/gpu/generalized_hough.cpp
+++ b/samples/gpu/generalized_hough.cpp
@@ -2,15 +2,15 @@
 #include <iostream>
 #include <string>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/contrib/contrib.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/gpuimgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/contrib.hpp"
 
 using namespace std;
 using namespace cv;
-using namespace cv::gpu;
 
 static Mat loadImage(const string& name)
 {
@@ -28,8 +28,7 @@ int main(int argc, const char* argv[])
     CommandLineParser cmd(argc, argv,
         "{ image i        | pic1.png  | input image }"
         "{ template t     | templ.png | template image }"
-        "{ scale s        |           | estimate scale }"
-        "{ rotation r     |           | estimate rotation }"
+        "{ full           |           | estimate scale and rotation }"
         "{ gpu            |           | use gpu version }"
         "{ minDist        | 100       | minimum distance between the centers of the detected objects }"
         "{ levels         | 360       | R-Table levels }"
@@ -44,7 +43,7 @@ int main(int argc, const char* argv[])
         "{ minAngle       | 0         | minimal rotation angle to detect in degrees }"
         "{ maxAngle       | 360       | maximal rotation angle to detect in degrees }"
         "{ angleStep      | 1         | angle step in degrees }"
-        "{ maxSize        | 1000      | maximal size of inner buffers }"
+        "{ maxBufSize     | 1000      | maximal size of inner buffers }"
         "{ help h ?       |           | print help message }"
     );
 
@@ -58,8 +57,7 @@ int main(int argc, const char* argv[])
 
     const string templName = cmd.get<string>("template");
     const string imageName = cmd.get<string>("image");
-    const bool estimateScale = cmd.has("scale");
-    const bool estimateRotation = cmd.has("rotation");
+    const bool full = cmd.has("full");
     const bool useGpu = cmd.has("gpu");
     const double minDist = cmd.get<double>("minDist");
     const int levels = cmd.get<int>("levels");
@@ -74,7 +72,7 @@ int main(int argc, const char* argv[])
     const double minAngle = cmd.get<double>("minAngle");
     const double maxAngle = cmd.get<double>("maxAngle");
     const double angleStep = cmd.get<double>("angleStep");
-    const int maxSize = cmd.get<int>("maxSize");
+    const int maxBufSize = cmd.get<int>("maxBufSize");
 
     if (!cmd.check())
     {
@@ -85,93 +83,69 @@ int main(int argc, const char* argv[])
     Mat templ = loadImage(templName);
     Mat image = loadImage(imageName);
 
-    int method = GHT_POSITION;
-    if (estimateScale)
-        method += GHT_SCALE;
-    if (estimateRotation)
-        method += GHT_ROTATION;
+    Ptr<GeneralizedHough> alg;
+
+    if (!full)
+    {
+        Ptr<GeneralizedHoughBallard> ballard = useGpu ? gpu::createGeneralizedHoughBallard() : createGeneralizedHoughBallard();
+
+        ballard->setMinDist(minDist);
+        ballard->setLevels(levels);
+        ballard->setDp(dp);
+        ballard->setMaxBufferSize(maxBufSize);
+        ballard->setVotesThreshold(votesThreshold);
+
+        alg = ballard;
+    }
+    else
+    {
+        Ptr<GeneralizedHoughGuil> guil = useGpu ? gpu::createGeneralizedHoughGuil() : createGeneralizedHoughGuil();
+
+        guil->setMinDist(minDist);
+        guil->setLevels(levels);
+        guil->setDp(dp);
+        guil->setMaxBufferSize(maxBufSize);
+
+        guil->setMinAngle(minAngle);
+        guil->setMaxAngle(maxAngle);
+        guil->setAngleStep(angleStep);
+        guil->setAngleThresh(angleThresh);
+
+        guil->setMinScale(minScale);
+        guil->setMaxScale(maxScale);
+        guil->setScaleStep(scaleStep);
+        guil->setScaleThresh(scaleThresh);
+
+        guil->setPosThresh(posThresh);
+
+        alg = guil;
+    }
 
     vector<Vec4f> position;
-    cv::TickMeter tm;
+    TickMeter tm;
 
     if (useGpu)
     {
-        GpuMat d_templ(templ);
-        GpuMat d_image(image);
-        GpuMat d_position;
+        gpu::GpuMat d_templ(templ);
+        gpu::GpuMat d_image(image);
+        gpu::GpuMat d_position;
 
-        Ptr<GeneralizedHough_GPU> d_hough = GeneralizedHough_GPU::create(method);
-        d_hough->set("minDist", minDist);
-        d_hough->set("levels", levels);
-        d_hough->set("dp", dp);
-        d_hough->set("maxSize", maxSize);
-        if (estimateScale && estimateRotation)
-        {
-            d_hough->set("angleThresh", angleThresh);
-            d_hough->set("scaleThresh", scaleThresh);
-            d_hough->set("posThresh", posThresh);
-        }
-        else
-        {
-            d_hough->set("votesThreshold", votesThreshold);
-        }
-        if (estimateScale)
-        {
-            d_hough->set("minScale", minScale);
-            d_hough->set("maxScale", maxScale);
-            d_hough->set("scaleStep", scaleStep);
-        }
-        if (estimateRotation)
-        {
-            d_hough->set("minAngle", minAngle);
-            d_hough->set("maxAngle", maxAngle);
-            d_hough->set("angleStep", angleStep);
-        }
-
-        d_hough->setTemplate(d_templ);
+        alg->setTemplate(d_templ);
 
         tm.start();
 
-        d_hough->detect(d_image, d_position);
-        d_hough->download(d_position, position);
+        alg->detect(d_image, d_position);
+        d_position.download(position);
 
         tm.stop();
     }
     else
     {
-        Ptr<GeneralizedHough> hough = GeneralizedHough::create(method);
-        hough->set("minDist", minDist);
-        hough->set("levels", levels);
-        hough->set("dp", dp);
-        if (estimateScale && estimateRotation)
-        {
-            hough->set("angleThresh", angleThresh);
-            hough->set("scaleThresh", scaleThresh);
-            hough->set("posThresh", posThresh);
-            hough->set("maxSize", maxSize);
-        }
-        else
-        {
-            hough->set("votesThreshold", votesThreshold);
-        }
-        if (estimateScale)
-        {
-            hough->set("minScale", minScale);
-            hough->set("maxScale", maxScale);
-            hough->set("scaleStep", scaleStep);
-        }
-        if (estimateRotation)
-        {
-            hough->set("minAngle", minAngle);
-            hough->set("maxAngle", maxAngle);
-            hough->set("angleStep", angleStep);
-        }
-
-        hough->setTemplate(templ);
+        alg->setTemplate(templ);
 
         tm.start();
 
-        hough->detect(image, position);
+        alg->detect(image, position);
 
         tm.stop();
     }
@@ -180,7 +154,7 @@ int main(int argc, const char* argv[])
     cout << "Detection time : " << tm.getTimeMilli() << " ms" << endl;
 
     Mat out;
-    cvtColor(image, out, COLOR_GRAY2BGR);
+    cv::cvtColor(image, out, COLOR_GRAY2BGR);
 
     for (size_t i = 0; i < position.size(); ++i)
     {
diff --git a/samples/gpu/hog.cpp b/samples/gpu/hog.cpp
index 2e0b70bcf9..c90cc23335 100644
--- a/samples/gpu/hog.cpp
+++ b/samples/gpu/hog.cpp
@@ -4,8 +4,10 @@
 #include <sstream>
 #include <iomanip>
 #include <stdexcept>
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/gpu.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/objdetect.hpp"
 
 using namespace std;
 using namespace cv;
@@ -295,8 +297,8 @@ void App::run()
             workBegin();
 
             // Change format of the image
-            if (make_gray) cvtColor(frame, img_aux, CV_BGR2GRAY);
-            else if (use_gpu) cvtColor(frame, img_aux, CV_BGR2BGRA);
+            if (make_gray) cvtColor(frame, img_aux, COLOR_BGR2GRAY);
+            else if (use_gpu) cvtColor(frame, img_aux, COLOR_BGR2BGRA);
             else frame.copyTo(img_aux);
 
             // Resize image
@@ -325,7 +327,7 @@ void App::run()
             for (size_t i = 0; i < found.size(); i++)
             {
                 Rect r = found[i];
-                rectangle(img_to_show, r.tl(), r.br(), CV_RGB(0, 255, 0), 3);
+                rectangle(img_to_show, r.tl(), r.br(), Scalar(0, 255, 0), 3);
             }
 
             if (use_gpu)
@@ -344,14 +346,14 @@ void App::run()
             {
                 if (!video_writer.isOpened())
                 {
-                    video_writer.open(args.dst_video, CV_FOURCC('x','v','i','d'), args.dst_video_fps,
+                    video_writer.open(args.dst_video, VideoWriter::fourcc('x','v','i','d'), args.dst_video_fps,
                                       img_to_show.size(), true);
                     if (!video_writer.isOpened())
                         throw std::runtime_error("can't create video writer");
                 }
 
-                if (make_gray) cvtColor(img_to_show, img, CV_GRAY2BGR);
-                else cvtColor(img_to_show, img, CV_BGRA2BGR);
+                if (make_gray) cvtColor(img_to_show, img, COLOR_GRAY2BGR);
+                else cvtColor(img_to_show, img, COLOR_BGRA2BGR);
 
                 video_writer << img;
             }
@@ -456,4 +458,3 @@ inline string App::workFps() const
     ss << work_fps;
     return ss.str();
 }
-
diff --git a/samples/gpu/houghlines.cpp b/samples/gpu/houghlines.cpp
index e98dcc6e7c..14245e5fd9 100644
--- a/samples/gpu/houghlines.cpp
+++ b/samples/gpu/houghlines.cpp
@@ -1,10 +1,11 @@
 #include <cmath>
 #include <iostream>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/gpu.hpp"
 
 using namespace std;
 using namespace cv;
@@ -30,17 +31,17 @@ int main(int argc, const char* argv[])
     }
 
     Mat mask;
-    Canny(src, mask, 100, 200, 3);
+    cv::Canny(src, mask, 100, 200, 3);
 
     Mat dst_cpu;
-    cvtColor(mask, dst_cpu, CV_GRAY2BGR);
+    cv::cvtColor(mask, dst_cpu, COLOR_GRAY2BGR);
     Mat dst_gpu = dst_cpu.clone();
 
     vector<Vec4i> lines_cpu;
     {
         const int64 start = getTickCount();
 
-        HoughLinesP(mask, lines_cpu, 1, CV_PI / 180, 50, 60, 5);
+        cv::HoughLinesP(mask, lines_cpu, 1, CV_PI / 180, 50, 60, 5);
 
         const double timeSec = (getTickCount() - start) / getTickFrequency();
         cout << "CPU Time : " << timeSec * 1000 << " ms" << endl;
@@ -50,16 +51,17 @@ int main(int argc, const char* argv[])
     for (size_t i = 0; i < lines_cpu.size(); ++i)
     {
         Vec4i l = lines_cpu[i];
-        line(dst_cpu, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0, 0, 255), 3, CV_AA);
+        line(dst_cpu, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0, 0, 255), 3, LINE_AA);
     }
 
     GpuMat d_src(mask);
     GpuMat d_lines;
-    HoughLinesBuf d_buf;
     {
         const int64 start = getTickCount();
 
-        gpu::HoughLinesP(d_src, d_lines, d_buf, 1.0f, (float) (CV_PI / 180.0f), 50, 5);
+        Ptr<gpu::HoughSegmentDetector> hough = gpu::createHoughSegmentDetector(1.0f, (float) (CV_PI / 180.0f), 50, 5);
+
+        hough->detect(d_src, d_lines);
 
         const double timeSec = (getTickCount() - start) / getTickFrequency();
         cout << "GPU Time : " << timeSec * 1000 << " ms" << endl;
@@ -76,7 +78,7 @@ int main(int argc, const char* argv[])
     for (size_t i = 0; i < lines_gpu.size(); ++i)
     {
         Vec4i l = lines_gpu[i];
-        line(dst_gpu, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0, 0, 255), 3, CV_AA);
+        line(dst_gpu, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0, 0, 255), 3, LINE_AA);
     }
 
     imshow("source", src);
@@ -86,4 +88,3 @@ int main(int argc, const char* argv[])
 
     return 0;
 }
-
diff --git a/samples/gpu/morphology.cpp b/samples/gpu/morphology.cpp
index 5863eac10f..abc6c980b0 100644
--- a/samples/gpu/morphology.cpp
+++ b/samples/gpu/morphology.cpp
@@ -1,120 +1,186 @@
+#include <iostream>
 
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/gpu/gpu.hpp"
-#include <stdlib.h>
-#include <stdio.h>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/gpufilters.hpp"
+#include "opencv2/gpuimgproc.hpp"
 
 using namespace std;
 using namespace cv;
-using namespace cv::gpu;
 
-static void help()
+class App
 {
+public:
+    App(int argc, const char* argv[]);
 
-printf("\nShow off image morphology: erosion, dialation, open and close\n"
-    "Call:\n   morphology2 [image]\n"
-    "This program also shows use of rect, elipse and cross kernels\n\n");
-printf( "Hot keys: \n"
-    "\tESC - quit the program\n"
-    "\tr - use rectangle structuring element\n"
-    "\te - use elliptic structuring element\n"
-    "\tc - use cross-shaped structuring element\n"
-    "\tSPACE - loop through all the options\n" );
-}
+    int run();
 
-GpuMat src, dst;
+private:
+    void help();
 
-int element_shape = MORPH_RECT;
+    void OpenClose();
+    void ErodeDilate();
 
-//the address of variable which receives trackbar position update
-int max_iters = 10;
-int open_close_pos = 0;
-int erode_dilate_pos = 0;
+    static void OpenCloseCallback(int, void*);
+    static void ErodeDilateCallback(int, void*);
 
-// callback function for open/close trackbar
-static void OpenClose(int, void*)
+    gpu::GpuMat src, dst;
+
+    int element_shape;
+
+    int max_iters;
+    int open_close_pos;
+    int erode_dilate_pos;
+};
+
+App::App(int argc, const char* argv[])
 {
-    int n = open_close_pos - max_iters;
-    int an = n > 0 ? n : -n;
-    Mat element = getStructuringElement(element_shape, Size(an*2+1, an*2+1), Point(an, an) );
-    if( n < 0 )
-        cv::gpu::morphologyEx(src, dst, CV_MOP_OPEN, element);
-    else
-        cv::gpu::morphologyEx(src, dst, CV_MOP_CLOSE, element);
-    imshow("Open/Close",(Mat)dst);
-}
+    element_shape = MORPH_RECT;
+    open_close_pos = erode_dilate_pos = max_iters = 10;
 
-// callback function for erode/dilate trackbar
-static void ErodeDilate(int, void*)
-{
-    int n = erode_dilate_pos - max_iters;
-    int an = n > 0 ? n : -n;
-    Mat element = getStructuringElement(element_shape, Size(an*2+1, an*2+1), Point(an, an) );
-    if( n < 0 )
-        cv::gpu::erode(src, dst, element);
-    else
-        cv::gpu::dilate(src, dst, element);
-    imshow("Erode/Dilate",(Mat)dst);
-}
-
-
-int main( int argc, char** argv )
-{
-    char* filename = argc == 2 ? argv[1] : (char*)"baboon.jpg";
-    if (string(argv[1]) == "--help")
+    if (argc == 2 && String(argv[1]) == "--help")
     {
         help();
-        return -1;
+        exit(0);
     }
 
-    src.upload(imread(filename, 1));
-    if (src.empty())
+    String filename = argc == 2 ? argv[1] : "baboon.jpg";
+
+    Mat img = imread(filename);
+    if (img.empty())
     {
-        help();
-        return -1;
+        cerr << "Can't open image " << filename.c_str() << endl;
+        exit(-1);
     }
 
-    cv::gpu::printShortCudaDeviceInfo(cv::gpu::getDevice());
-
-    help();
-
-
+    src.upload(img);
     if (src.channels() == 3)
     {
         // gpu support only 4th channel images
-        GpuMat src4ch;
-        cv::gpu::cvtColor(src, src4ch, CV_BGR2BGRA);
+        gpu::GpuMat src4ch;
+        gpu::cvtColor(src, src4ch, COLOR_BGR2BGRA);
         src = src4ch;
     }
 
-    //create windows for output images
-    namedWindow("Open/Close",1);
-    namedWindow("Erode/Dilate",1);
+    help();
 
-    open_close_pos = erode_dilate_pos = max_iters;
-    createTrackbar("iterations", "Open/Close",&open_close_pos,max_iters*2+1,OpenClose);
-    createTrackbar("iterations", "Erode/Dilate",&erode_dilate_pos,max_iters*2+1,ErodeDilate);
+    gpu::printShortCudaDeviceInfo(gpu::getDevice());
+}
+
+int App::run()
+{
+    // create windows for output images
+    namedWindow("Open/Close");
+    namedWindow("Erode/Dilate");
+
+    createTrackbar("iterations", "Open/Close", &open_close_pos, max_iters * 2 + 1, OpenCloseCallback, this);
+    createTrackbar("iterations", "Erode/Dilate", &erode_dilate_pos, max_iters * 2 + 1, ErodeDilateCallback, this);
 
     for(;;)
     {
-        int c;
+        OpenClose();
+        ErodeDilate();
 
-        OpenClose(open_close_pos, 0);
-        ErodeDilate(erode_dilate_pos, 0);
-        c = cvWaitKey(0);
+        char c = (char) waitKey();
 
-        if( (char)c == 27 )
+        switch (c)
+        {
+        case 27:
+            return 0;
             break;
-        if( (char)c == 'e' )
+
+        case 'e':
             element_shape = MORPH_ELLIPSE;
-        else if( (char)c == 'r' )
+            break;
+
+        case 'r':
             element_shape = MORPH_RECT;
-        else if( (char)c == 'c' )
+            break;
+
+        case 'c':
             element_shape = MORPH_CROSS;
-        else if( (char)c == ' ' )
+            break;
+
+        case ' ':
             element_shape = (element_shape + 1) % 3;
+            break;
+        }
+    }
+}
+
+void App::help()
+{
+    cout << "Show off image morphology: erosion, dialation, open and close \n";
+    cout << "Call: \n";
+    cout << "   gpu-example-morphology [image] \n";
+    cout << "This program also shows use of rect, elipse and cross kernels \n" << endl;
+
+    cout << "Hot keys: \n";
+    cout << "\tESC - quit the program \n";
+    cout << "\tr - use rectangle structuring element \n";
+    cout << "\te - use elliptic structuring element \n";
+    cout << "\tc - use cross-shaped structuring element \n";
+    cout << "\tSPACE - loop through all the options \n" << endl;
+}
+
+void App::OpenClose()
+{
+    int n = open_close_pos - max_iters;
+    int an = n > 0 ? n : -n;
+
+    Mat element = getStructuringElement(element_shape, Size(an*2+1, an*2+1), Point(an, an));
+
+    if (n < 0)
+    {
+        Ptr<gpu::Filter> openFilter = gpu::createMorphologyFilter(MORPH_OPEN, src.type(), element);
+        openFilter->apply(src, dst);
+    }
+    else
+    {
+        Ptr<gpu::Filter> closeFilter = gpu::createMorphologyFilter(MORPH_CLOSE, src.type(), element);
+        closeFilter->apply(src, dst);
     }
 
-    return 0;
+    Mat h_dst(dst);
+    imshow("Open/Close", h_dst);
+}
+
+void App::ErodeDilate()
+{
+    int n = erode_dilate_pos - max_iters;
+    int an = n > 0 ? n : -n;
+
+    Mat element = getStructuringElement(element_shape, Size(an*2+1, an*2+1), Point(an, an));
+
+    if (n < 0)
+    {
+        Ptr<gpu::Filter> erodeFilter = gpu::createMorphologyFilter(MORPH_ERODE, src.type(), element);
+        erodeFilter->apply(src, dst);
+    }
+    else
+    {
+        Ptr<gpu::Filter> dilateFilter = gpu::createMorphologyFilter(MORPH_DILATE, src.type(), element);
+        dilateFilter->apply(src, dst);
+    }
+
+    Mat h_dst(dst);
+    imshow("Erode/Dilate", h_dst);
+}
+
+void App::OpenCloseCallback(int, void* data)
+{
+    App* thiz = (App*) data;
+    thiz->OpenClose();
+}
+
+void App::ErodeDilateCallback(int, void* data)
+{
+    App* thiz = (App*) data;
+    thiz->ErodeDilate();
+}
+
+int main(int argc, const char* argv[])
+{
+    App app(argc, argv);
+    return app.run();
 }
diff --git a/samples/gpu/multi.cpp b/samples/gpu/multi.cpp
index c3925fc442..b83fd2ce46 100644
--- a/samples/gpu/multi.cpp
+++ b/samples/gpu/multi.cpp
@@ -11,6 +11,18 @@
 #include "opencv2/core/core.hpp"
 #include "opencv2/gpu/gpu.hpp"
 
+#ifdef HAVE_TBB
+#  include "tbb/tbb_stddef.h"
+#  if TBB_VERSION_MAJOR*100 + TBB_VERSION_MINOR >= 202
+#    include "tbb/tbb.h"
+#    include "tbb/task.h"
+#    undef min
+#    undef max
+#  else
+#    undef HAVE_TBB
+#  endif
+#endif
+
 #if !defined(HAVE_CUDA) || !defined(HAVE_TBB)
 
 int main()
@@ -28,8 +40,6 @@ int main()
 
 #else
 
-#include "opencv2/core/internal.hpp" // For TBB wrappers
-
 using namespace std;
 using namespace cv;
 using namespace cv::gpu;
@@ -60,7 +70,7 @@ int main()
 
     // Execute calculation in two threads using two GPUs
     int devices[] = {0, 1};
-    parallel_do(devices, devices + 2, Worker());
+    tbb::parallel_do(devices, devices + 2, Worker());
 
     return 0;
 }
@@ -77,15 +87,15 @@ void Worker::operator()(int device_id) const
     rng.fill(src, RNG::UNIFORM, 0, 1);
 
     // CPU works
-    transpose(src, dst);
+    cv::transpose(src, dst);
 
     // GPU works
     GpuMat d_src(src);
     GpuMat d_dst;
-    transpose(d_src, d_dst);
+    gpu::transpose(d_src, d_dst);
 
     // Check results
-    bool passed = norm(dst - Mat(d_dst), NORM_INF) < 1e-3;
+    bool passed = cv::norm(dst - Mat(d_dst), NORM_INF) < 1e-3;
     std::cout << "GPU #" << device_id << " (" << DeviceInfo().name() << "): "
         << (passed ? "passed" : "FAILED") << endl;
 
diff --git a/samples/gpu/opengl.cpp b/samples/gpu/opengl.cpp
index 80eb44b539..37960d9a9f 100644
--- a/samples/gpu/opengl.cpp
+++ b/samples/gpu/opengl.cpp
@@ -25,7 +25,7 @@ int main()
 
 #include "opencv2/core/core.hpp"
 #include "opencv2/core/opengl.hpp"
-#include "opencv2/core/gpumat.hpp"
+#include "opencv2/core/gpu.hpp"
 #include "opencv2/highgui/highgui.hpp"
 
 using namespace std;
diff --git a/samples/gpu/optical_flow.cpp b/samples/gpu/optical_flow.cpp
index 3f74d1b676..cae00fcc35 100644
--- a/samples/gpu/optical_flow.cpp
+++ b/samples/gpu/optical_flow.cpp
@@ -1,9 +1,10 @@
 #include <iostream>
 #include <fstream>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/highgui.hpp"
+#include "opencv2/gpu.hpp"
 
 using namespace std;
 using namespace cv;
diff --git a/samples/gpu/opticalflow_nvidia_api.cpp b/samples/gpu/opticalflow_nvidia_api.cpp
index 05a37ef69d..4e0863b46a 100644
--- a/samples/gpu/opticalflow_nvidia_api.cpp
+++ b/samples/gpu/opticalflow_nvidia_api.cpp
@@ -13,10 +13,10 @@
 #include <iomanip>
 #include "opencv2/gpu/gpu.hpp"
 #include "opencv2/highgui/highgui.hpp"
+#include "opencv2/highgui/highgui_c.h"
 
 #ifdef HAVE_CUDA
-#include "NPP_staging/NPP_staging.hpp"
-#include "NCVBroxOpticalFlow.hpp"
+#include "opencv2/gpulegacy.hpp"
 #endif
 
 #if !defined(HAVE_CUDA)
diff --git a/samples/gpu/performance/CMakeLists.txt b/samples/gpu/performance/CMakeLists.txt
index 3b3d6e4415..492b4c7900 100644
--- a/samples/gpu/performance/CMakeLists.txt
+++ b/samples/gpu/performance/CMakeLists.txt
@@ -3,9 +3,17 @@ set(the_target "example_gpu_performance")
 file(GLOB sources "performance/*.cpp")
 file(GLOB headers "performance/*.h")
 
+if(HAVE_opencv_nonfree)
+  ocv_include_directories("${OpenCV_SOURCE_DIR}/modules/nonfree/include")
+endif()
+
 add_executable(${the_target} ${sources} ${headers})
 target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_GPU_SAMPLES_REQUIRED_DEPS})
 
+if(HAVE_opencv_nonfree)
+  target_link_libraries(${the_target} opencv_nonfree)
+endif()
+
 set_target_properties(${the_target} PROPERTIES
     OUTPUT_NAME "performance_gpu"
     PROJECT_LABEL "(EXAMPLE_GPU) performance")
diff --git a/samples/gpu/performance/performance.cpp b/samples/gpu/performance/performance.cpp
index 308b58c017..8af0b3d0d4 100644
--- a/samples/gpu/performance/performance.cpp
+++ b/samples/gpu/performance/performance.cpp
@@ -144,7 +144,7 @@ string abspath(const string& relpath)
 }
 
 
-static int CV_CDECL cvErrorCallback(int /*status*/, const char* /*func_name*/,
+static int cvErrorCallback(int /*status*/, const char* /*func_name*/,
                              const char* err_msg, const char* /*file_name*/,
                              int /*line*/, void* /*userdata*/)
 {
diff --git a/samples/gpu/performance/performance.h b/samples/gpu/performance/performance.h
index 4f58061ace..8ee007f0aa 100644
--- a/samples/gpu/performance/performance.h
+++ b/samples/gpu/performance/performance.h
@@ -6,8 +6,8 @@
 #include <vector>
 #include <numeric>
 #include <string>
-#include "opencv2/core/core.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/gpu.hpp"
 
 #define TAB "    "
 
diff --git a/samples/gpu/performance/tests.cpp b/samples/gpu/performance/tests.cpp
index b35102d861..a3df43c6e2 100644
--- a/samples/gpu/performance/tests.cpp
+++ b/samples/gpu/performance/tests.cpp
@@ -1,50 +1,48 @@
 #include <stdexcept>
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/video/video.hpp"
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/nonfree/nonfree.hpp"
-#include "opencv2/legacy/legacy.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/calib3d.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/gpu.hpp"
+
+#include "opencv2/legacy.hpp"
 #include "performance.h"
 
+#include "opencv2/opencv_modules.hpp"
+#ifdef HAVE_OPENCV_NONFREE
+#include "opencv2/nonfree/gpu.hpp"
+#include "opencv2/nonfree/nonfree.hpp"
+#endif
+
 using namespace std;
 using namespace cv;
 
-static void InitMatchTemplate()
-{
-    Mat src; gen(src, 500, 500, CV_32F, 0, 1);
-    Mat templ; gen(templ, 500, 500, CV_32F, 0, 1);
-    gpu::GpuMat d_src(src), d_templ(templ), d_dst;
-    gpu::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
-}
-
 
 TEST(matchTemplate)
 {
-    InitMatchTemplate();
-
     Mat src, templ, dst;
     gen(src, 3000, 3000, CV_32F, 0, 1);
 
     gpu::GpuMat d_src(src), d_templ, d_dst;
 
+    Ptr<gpu::TemplateMatching> alg = gpu::createTemplateMatching(src.type(), TM_CCORR);
+
     for (int templ_size = 5; templ_size < 200; templ_size *= 5)
     {
         SUBTEST << src.cols << 'x' << src.rows << ", 32FC1" << ", templ " << templ_size << 'x' << templ_size << ", CCORR";
 
         gen(templ, templ_size, templ_size, CV_32F, 0, 1);
-        matchTemplate(src, templ, dst, CV_TM_CCORR);
+        matchTemplate(src, templ, dst, TM_CCORR);
 
         CPU_ON;
-        matchTemplate(src, templ, dst, CV_TM_CCORR);
+        matchTemplate(src, templ, dst, TM_CCORR);
         CPU_OFF;
 
         d_templ.upload(templ);
-        gpu::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
+        alg->match(d_src, d_templ, d_dst);
 
         GPU_ON;
-        gpu::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
+        alg->match(d_src, d_templ, d_dst);
         GPU_OFF;
     }
 }
@@ -170,10 +168,12 @@ TEST(cornerHarris)
 
         d_src.upload(src);
 
-        gpu::cornerHarris(d_src, d_dst, 5, 7, 0.1, BORDER_REFLECT101);
+        Ptr<gpu::CornernessCriteria> harris = gpu::createHarrisCorner(src.type(), 5, 7, 0.1, BORDER_REFLECT101);
+
+        harris->compute(d_src, d_dst);
 
         GPU_ON;
-        gpu::cornerHarris(d_src, d_dst, 5, 7, 0.1, BORDER_REFLECT101);
+        harris->compute(d_src, d_dst);
         GPU_OFF;
     }
 }
@@ -266,10 +266,11 @@ TEST(meanShift)
     }
 }
 
+#ifdef HAVE_OPENCV_NONFREE
 
 TEST(SURF)
 {
-    Mat src = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
+    Mat src = imread(abspath("aloeL.jpg"), IMREAD_GRAYSCALE);
     if (src.empty()) throw runtime_error("can't open aloeL.jpg");
 
     SURF surf;
@@ -294,10 +295,12 @@ TEST(SURF)
     GPU_OFF;
 }
 
+#endif
+
 
 TEST(FAST)
 {
-    Mat src = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
+    Mat src = imread(abspath("aloeL.jpg"), IMREAD_GRAYSCALE);
     if (src.empty()) throw runtime_error("can't open aloeL.jpg");
 
     vector<KeyPoint> keypoints;
@@ -322,7 +325,7 @@ TEST(FAST)
 
 TEST(ORB)
 {
-    Mat src = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
+    Mat src = imread(abspath("aloeL.jpg"), IMREAD_GRAYSCALE);
     if (src.empty()) throw runtime_error("can't open aloeL.jpg");
 
     ORB orb(4000);
@@ -596,120 +599,120 @@ TEST(cvtColor)
     gen(src, 4000, 4000, CV_8UC1, 0, 255);
     d_src.upload(src);
 
-    SUBTEST << "4000x4000, 8UC1, CV_GRAY2BGRA";
+    SUBTEST << "4000x4000, 8UC1, COLOR_GRAY2BGRA";
 
-    cvtColor(src, dst, CV_GRAY2BGRA, 4);
+    cvtColor(src, dst, COLOR_GRAY2BGRA, 4);
 
     CPU_ON;
-    cvtColor(src, dst, CV_GRAY2BGRA, 4);
+    cvtColor(src, dst, COLOR_GRAY2BGRA, 4);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_GRAY2BGRA, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_GRAY2BGRA, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_GRAY2BGRA, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_GRAY2BGRA, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC3 vs 8UC4, CV_BGR2YCrCb";
+    SUBTEST << "4000x4000, 8UC3 vs 8UC4, COLOR_BGR2YCrCb";
 
-    cvtColor(src, dst, CV_BGR2YCrCb);
+    cvtColor(src, dst, COLOR_BGR2YCrCb);
 
     CPU_ON;
-    cvtColor(src, dst, CV_BGR2YCrCb);
+    cvtColor(src, dst, COLOR_BGR2YCrCb);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_BGR2YCrCb, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2YCrCb, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_BGR2YCrCb, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2YCrCb, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC4, CV_YCrCb2BGR";
+    SUBTEST << "4000x4000, 8UC4, COLOR_YCrCb2BGR";
 
-    cvtColor(src, dst, CV_YCrCb2BGR, 4);
+    cvtColor(src, dst, COLOR_YCrCb2BGR, 4);
 
     CPU_ON;
-    cvtColor(src, dst, CV_YCrCb2BGR, 4);
+    cvtColor(src, dst, COLOR_YCrCb2BGR, 4);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_YCrCb2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_YCrCb2BGR, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_YCrCb2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_YCrCb2BGR, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC3 vs 8UC4, CV_BGR2XYZ";
+    SUBTEST << "4000x4000, 8UC3 vs 8UC4, COLOR_BGR2XYZ";
 
-    cvtColor(src, dst, CV_BGR2XYZ);
+    cvtColor(src, dst, COLOR_BGR2XYZ);
 
     CPU_ON;
-    cvtColor(src, dst, CV_BGR2XYZ);
+    cvtColor(src, dst, COLOR_BGR2XYZ);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_BGR2XYZ, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2XYZ, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_BGR2XYZ, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2XYZ, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC4, CV_XYZ2BGR";
+    SUBTEST << "4000x4000, 8UC4, COLOR_XYZ2BGR";
 
-    cvtColor(src, dst, CV_XYZ2BGR, 4);
+    cvtColor(src, dst, COLOR_XYZ2BGR, 4);
 
     CPU_ON;
-    cvtColor(src, dst, CV_XYZ2BGR, 4);
+    cvtColor(src, dst, COLOR_XYZ2BGR, 4);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_XYZ2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_XYZ2BGR, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_XYZ2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_XYZ2BGR, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC3 vs 8UC4, CV_BGR2HSV";
+    SUBTEST << "4000x4000, 8UC3 vs 8UC4, COLOR_BGR2HSV";
 
-    cvtColor(src, dst, CV_BGR2HSV);
+    cvtColor(src, dst, COLOR_BGR2HSV);
 
     CPU_ON;
-    cvtColor(src, dst, CV_BGR2HSV);
+    cvtColor(src, dst, COLOR_BGR2HSV);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_BGR2HSV, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2HSV, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_BGR2HSV, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_BGR2HSV, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
     d_src.swap(d_dst);
 
-    SUBTEST << "4000x4000, 8UC4, CV_HSV2BGR";
+    SUBTEST << "4000x4000, 8UC4, COLOR_HSV2BGR";
 
-    cvtColor(src, dst, CV_HSV2BGR, 4);
+    cvtColor(src, dst, COLOR_HSV2BGR, 4);
 
     CPU_ON;
-    cvtColor(src, dst, CV_HSV2BGR, 4);
+    cvtColor(src, dst, COLOR_HSV2BGR, 4);
     CPU_OFF;
 
-    gpu::cvtColor(d_src, d_dst, CV_HSV2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_HSV2BGR, 4);
 
     GPU_ON;
-    gpu::cvtColor(d_src, d_dst, CV_HSV2BGR, 4);
+    gpu::cvtColor(d_src, d_dst, COLOR_HSV2BGR, 4);
     GPU_OFF;
 
     cv::swap(src, dst);
@@ -737,10 +740,12 @@ TEST(erode)
 
         d_src.upload(src);
 
-        gpu::erode(d_src, d_dst, ker, d_buf);
+        Ptr<gpu::Filter> erode = gpu::createMorphologyFilter(MORPH_ERODE, d_src.type(), ker);
+
+        erode->apply(d_src, d_dst);
 
         GPU_ON;
-        gpu::erode(d_src, d_dst, ker, d_buf);
+        erode->apply(d_src, d_dst);
         GPU_OFF;
     }
 }
@@ -920,10 +925,12 @@ TEST(GaussianBlur)
         gpu::GpuMat d_dst(src.size(), src.type());
         gpu::GpuMat d_buf;
 
-        gpu::GaussianBlur(d_src, d_dst, Size(3, 3), d_buf, 1);
+        cv::Ptr<cv::gpu::Filter> gauss = cv::gpu::createGaussianFilter(d_src.type(), -1, cv::Size(3, 3), 1);
+
+        gauss->apply(d_src, d_dst);
 
         GPU_ON;
-        gpu::GaussianBlur(d_src, d_dst, Size(3, 3), d_buf, 1);
+        gauss->apply(d_src, d_dst);
         GPU_OFF;
     }
 }
@@ -952,10 +959,11 @@ TEST(filter2D)
             gpu::GpuMat d_src(src);
             gpu::GpuMat d_dst;
 
-            gpu::filter2D(d_src, d_dst, -1, kernel);
+            Ptr<gpu::Filter> filter2D = gpu::createLinearFilter(d_src.type(), -1, kernel);
+            filter2D->apply(d_src, d_dst);
 
             GPU_ON;
-            gpu::filter2D(d_src, d_dst, -1, kernel);
+            filter2D->apply(d_src, d_dst);
             GPU_OFF;
         }
     }
@@ -1033,13 +1041,12 @@ TEST(equalizeHist)
 
         gpu::GpuMat d_src(src);
         gpu::GpuMat d_dst;
-        gpu::GpuMat d_hist;
         gpu::GpuMat d_buf;
 
-        gpu::equalizeHist(d_src, d_dst, d_hist, d_buf);
+        gpu::equalizeHist(d_src, d_dst, d_buf);
 
         GPU_ON;
-        gpu::equalizeHist(d_src, d_dst, d_hist, d_buf);
+        gpu::equalizeHist(d_src, d_dst, d_buf);
         GPU_OFF;
     }
 }
@@ -1047,7 +1054,7 @@ TEST(equalizeHist)
 
 TEST(Canny)
 {
-    Mat img = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
+    Mat img = imread(abspath("aloeL.jpg"), IMREAD_GRAYSCALE);
 
     if (img.empty()) throw runtime_error("can't open aloeL.jpg");
 
@@ -1059,12 +1066,13 @@ TEST(Canny)
 
     gpu::GpuMat d_img(img);
     gpu::GpuMat d_edges;
-    gpu::CannyBuf d_buf;
 
-    gpu::Canny(d_img, d_buf, d_edges, 50.0, 100.0);
+    Ptr<gpu::CannyEdgeDetector> canny = gpu::createCannyEdgeDetector(50.0, 100.0);
+
+    canny->detect(d_img, d_edges);
 
     GPU_ON;
-    gpu::Canny(d_img, d_buf, d_edges, 50.0, 100.0);
+    canny->detect(d_img, d_edges);
     GPU_OFF;
 }
 
@@ -1085,30 +1093,30 @@ TEST(reduce)
 
         SUBTEST << size << 'x' << size << ", dim = 0";
 
-        reduce(src, dst0, 0, CV_REDUCE_MIN);
+        reduce(src, dst0, 0, REDUCE_MIN);
 
         CPU_ON;
-        reduce(src, dst0, 0, CV_REDUCE_MIN);
+        reduce(src, dst0, 0, REDUCE_MIN);
         CPU_OFF;
 
-        gpu::reduce(d_src, d_dst0, 0, CV_REDUCE_MIN);
+        gpu::reduce(d_src, d_dst0, 0, REDUCE_MIN);
 
         GPU_ON;
-        gpu::reduce(d_src, d_dst0, 0, CV_REDUCE_MIN);
+        gpu::reduce(d_src, d_dst0, 0, REDUCE_MIN);
         GPU_OFF;
 
         SUBTEST << size << 'x' << size << ", dim = 1";
 
-        reduce(src, dst1, 1, CV_REDUCE_MIN);
+        reduce(src, dst1, 1, REDUCE_MIN);
 
         CPU_ON;
-        reduce(src, dst1, 1, CV_REDUCE_MIN);
+        reduce(src, dst1, 1, REDUCE_MIN);
         CPU_OFF;
 
-        gpu::reduce(d_src, d_dst1, 1, CV_REDUCE_MIN);
+        gpu::reduce(d_src, d_dst1, 1, REDUCE_MIN);
 
         GPU_ON;
-        gpu::reduce(d_src, d_dst1, 1, CV_REDUCE_MIN);
+        gpu::reduce(d_src, d_dst1, 1, REDUCE_MIN);
         GPU_OFF;
     }
 }
@@ -1158,15 +1166,15 @@ TEST(GoodFeaturesToTrack)
     goodFeaturesToTrack(src, pts, 8000, 0.01, 0.0);
     CPU_OFF;
 
-    gpu::GoodFeaturesToTrackDetector_GPU detector(8000, 0.01, 0.0);
+    Ptr<gpu::CornersDetector> detector = gpu::createGoodFeaturesToTrackDetector(src.type(), 8000, 0.01, 0.0);
 
     gpu::GpuMat d_src(src);
     gpu::GpuMat d_pts;
 
-    detector(d_src, d_pts);
+    detector->detect(d_src, d_pts);
 
     GPU_ON;
-    detector(d_src, d_pts);
+    detector->detect(d_src, d_pts);
     GPU_OFF;
 }
 
@@ -1253,7 +1261,7 @@ TEST(FarnebackOpticalFlow)
 
 namespace cv
 {
-    template<> void Ptr<CvBGStatModel>::delete_obj()
+    template<> void DefaultDeleter<CvBGStatModel>::operator ()(CvBGStatModel* obj) const
     {
         cvReleaseBGStatModel(&obj);
     }
@@ -1263,14 +1271,14 @@ TEST(FGDStatModel)
 {
     const std::string inputFile = abspath("768x576.avi");
 
-    cv::VideoCapture cap(inputFile);
+    VideoCapture cap(inputFile);
     if (!cap.isOpened()) throw runtime_error("can't open 768x576.avi");
 
-    cv::Mat frame;
+    Mat frame;
     cap >> frame;
 
     IplImage ipl_frame = frame;
-    cv::Ptr<CvBGStatModel> model(cvCreateFGDStatModel(&ipl_frame));
+    Ptr<CvBGStatModel> model(cvCreateFGDStatModel(&ipl_frame));
 
     while (!TestSystem::instance().stop())
     {
@@ -1289,8 +1297,10 @@ TEST(FGDStatModel)
 
     cap >> frame;
 
-    cv::gpu::GpuMat d_frame(frame);
-    cv::gpu::FGDStatModel d_model(d_frame);
+    gpu::GpuMat d_frame(frame), d_fgmask;
+    Ptr<BackgroundSubtractor> d_fgd = gpu::createBackgroundSubtractorFGD();
+
+    d_fgd->apply(d_frame, d_fgmask);
 
     while (!TestSystem::instance().stop())
     {
@@ -1299,7 +1309,7 @@ TEST(FGDStatModel)
 
         TestSystem::instance().gpuOn();
 
-        d_model.update(d_frame);
+        d_fgd->apply(d_frame, d_fgmask);
 
         TestSystem::instance().gpuOff();
     }
@@ -1316,10 +1326,10 @@ TEST(MOG)
     cv::Mat frame;
     cap >> frame;
 
-    cv::BackgroundSubtractorMOG mog;
+    cv::Ptr<cv::BackgroundSubtractor> mog = cv::createBackgroundSubtractorMOG();
     cv::Mat foreground;
 
-    mog(frame, foreground, 0.01);
+    mog->apply(frame, foreground, 0.01);
 
     while (!TestSystem::instance().stop())
     {
@@ -1327,7 +1337,7 @@ TEST(MOG)
 
         TestSystem::instance().cpuOn();
 
-        mog(frame, foreground, 0.01);
+        mog->apply(frame, foreground, 0.01);
 
         TestSystem::instance().cpuOff();
     }
@@ -1338,10 +1348,10 @@ TEST(MOG)
     cap >> frame;
 
     cv::gpu::GpuMat d_frame(frame);
-    cv::gpu::MOG_GPU d_mog;
+    cv::Ptr<cv::BackgroundSubtractor> d_mog = cv::gpu::createBackgroundSubtractorMOG();
     cv::gpu::GpuMat d_foreground;
 
-    d_mog(d_frame, d_foreground, 0.01f);
+    d_mog->apply(d_frame, d_foreground, 0.01);
 
     while (!TestSystem::instance().stop())
     {
@@ -1350,7 +1360,7 @@ TEST(MOG)
 
         TestSystem::instance().gpuOn();
 
-        d_mog(d_frame, d_foreground, 0.01f);
+        d_mog->apply(d_frame, d_foreground, 0.01);
 
         TestSystem::instance().gpuOff();
     }
@@ -1367,12 +1377,12 @@ TEST(MOG2)
     cv::Mat frame;
     cap >> frame;
 
-    cv::BackgroundSubtractorMOG2 mog2;
+    cv::Ptr<cv::BackgroundSubtractor> mog2 = cv::createBackgroundSubtractorMOG2();
     cv::Mat foreground;
     cv::Mat background;
 
-    mog2(frame, foreground);
-    mog2.getBackgroundImage(background);
+    mog2->apply(frame, foreground);
+    mog2->getBackgroundImage(background);
 
     while (!TestSystem::instance().stop())
     {
@@ -1380,8 +1390,8 @@ TEST(MOG2)
 
         TestSystem::instance().cpuOn();
 
-        mog2(frame, foreground);
-        mog2.getBackgroundImage(background);
+        mog2->apply(frame, foreground);
+        mog2->getBackgroundImage(background);
 
         TestSystem::instance().cpuOff();
     }
@@ -1391,13 +1401,13 @@ TEST(MOG2)
 
     cap >> frame;
 
+    cv::Ptr<cv::BackgroundSubtractor> d_mog2 = cv::gpu::createBackgroundSubtractorMOG2();
     cv::gpu::GpuMat d_frame(frame);
-    cv::gpu::MOG2_GPU d_mog2;
     cv::gpu::GpuMat d_foreground;
     cv::gpu::GpuMat d_background;
 
-    d_mog2(d_frame, d_foreground);
-    d_mog2.getBackgroundImage(d_background);
+    d_mog2->apply(d_frame, d_foreground);
+    d_mog2->getBackgroundImage(d_background);
 
     while (!TestSystem::instance().stop())
     {
@@ -1406,8 +1416,8 @@ TEST(MOG2)
 
         TestSystem::instance().gpuOn();
 
-        d_mog2(d_frame, d_foreground);
-        d_mog2.getBackgroundImage(d_background);
+        d_mog2->apply(d_frame, d_foreground);
+        d_mog2->getBackgroundImage(d_background);
 
         TestSystem::instance().gpuOff();
     }
diff --git a/samples/gpu/pyrlk_optical_flow.cpp b/samples/gpu/pyrlk_optical_flow.cpp
index c2e765da13..08717292cd 100644
--- a/samples/gpu/pyrlk_optical_flow.cpp
+++ b/samples/gpu/pyrlk_optical_flow.cpp
@@ -1,12 +1,12 @@
 #include <iostream>
 #include <vector>
 
-#include "cvconfig.h"
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/video/video.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include "opencv2/core.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/video.hpp"
+#include "opencv2/gpu.hpp"
 
 using namespace std;
 using namespace cv;
@@ -170,18 +170,18 @@ int main(int argc, const char* argv[])
     cout << endl;
 
     Mat frame0Gray;
-    cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+    cv::cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
     Mat frame1Gray;
-    cvtColor(frame1, frame1Gray, COLOR_BGR2GRAY);
+    cv::cvtColor(frame1, frame1Gray, COLOR_BGR2GRAY);
 
     // goodFeaturesToTrack
 
-    GoodFeaturesToTrackDetector_GPU detector(points, 0.01, minDist);
-
     GpuMat d_frame0Gray(frame0Gray);
     GpuMat d_prevPts;
 
-    detector(d_frame0Gray, d_prevPts);
+    Ptr<gpu::CornersDetector> detector = gpu::createGoodFeaturesToTrackDetector(d_frame0Gray.type(), points, 0.01, minDist);
+
+    detector->detect(d_frame0Gray, d_prevPts);
 
     // Sparse
 
diff --git a/samples/gpu/softcascade.cpp b/samples/gpu/softcascade.cpp
index acc43af1b3..980d6ffdf7 100644
--- a/samples/gpu/softcascade.cpp
+++ b/samples/gpu/softcascade.cpp
@@ -1,3 +1,4 @@
+#include <opencv2/core/utility.hpp>
 #include <opencv2/gpu.hpp>
 #include <opencv2/softcascade.hpp>
 #include <opencv2/highgui.hpp>
diff --git a/samples/gpu/stereo_match.cpp b/samples/gpu/stereo_match.cpp
index 62745d749d..e404476406 100644
--- a/samples/gpu/stereo_match.cpp
+++ b/samples/gpu/stereo_match.cpp
@@ -3,8 +3,9 @@
 #include <sstream>
 #include <iomanip>
 #include <stdexcept>
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/gpu.hpp"
+#include "opencv2/highgui.hpp"
 
 using namespace cv;
 using namespace std;
@@ -64,9 +65,9 @@ private:
     Mat left, right;
     gpu::GpuMat d_left, d_right;
 
-    gpu::StereoBM_GPU bm;
-    gpu::StereoBeliefPropagation bp;
-    gpu::StereoConstantSpaceBP csbp;
+    Ptr<gpu::StereoBM> bm;
+    Ptr<gpu::StereoBeliefPropagation> bp;
+    Ptr<gpu::StereoConstantSpaceBP> csbp;
 
     int64 work_begin;
     double work_fps;
@@ -162,8 +163,8 @@ void App::run()
     right_src = imread(p.right);
     if (left_src.empty()) throw runtime_error("can't open file \"" + p.left + "\"");
     if (right_src.empty()) throw runtime_error("can't open file \"" + p.right + "\"");
-    cvtColor(left_src, left, CV_BGR2GRAY);
-    cvtColor(right_src, right, CV_BGR2GRAY);
+    cvtColor(left_src, left, COLOR_BGR2GRAY);
+    cvtColor(right_src, right, COLOR_BGR2GRAY);
     d_left.upload(left);
     d_right.upload(right);
 
@@ -171,9 +172,9 @@ void App::run()
     imshow("right", right);
 
     // Set common parameters
-    bm.ndisp = p.ndisp;
-    bp.ndisp = p.ndisp;
-    csbp.ndisp = p.ndisp;
+    bm = gpu::createStereoBM(p.ndisp);
+    bp = gpu::createStereoBeliefPropagation(p.ndisp);
+    csbp = cv::gpu::createStereoConstantSpaceBP(p.ndisp);
 
     // Prepare disparity map of specified type
     Mat disp(left.size(), CV_8U);
@@ -192,18 +193,18 @@ void App::run()
             if (d_left.channels() > 1 || d_right.channels() > 1)
             {
                 cout << "BM doesn't support color images\n";
-                cvtColor(left_src, left, CV_BGR2GRAY);
-                cvtColor(right_src, right, CV_BGR2GRAY);
+                cvtColor(left_src, left, COLOR_BGR2GRAY);
+                cvtColor(right_src, right, COLOR_BGR2GRAY);
                 cout << "image_channels: " << left.channels() << endl;
                 d_left.upload(left);
                 d_right.upload(right);
                 imshow("left", left);
                 imshow("right", right);
             }
-            bm(d_left, d_right, d_disp);
+            bm->compute(d_left, d_right, d_disp);
             break;
-        case Params::BP: bp(d_left, d_right, d_disp); break;
-        case Params::CSBP: csbp(d_left, d_right, d_disp); break;
+        case Params::BP: bp->compute(d_left, d_right, d_disp); break;
+        case Params::CSBP: csbp->compute(d_left, d_right, d_disp); break;
         }
         workEnd();
 
@@ -227,16 +228,16 @@ void App::printParams() const
     switch (p.method)
     {
     case Params::BM:
-        cout << "win_size: " << bm.winSize << endl;
-        cout << "prefilter_sobel: " << bm.preset << endl;
+        cout << "win_size: " << bm->getBlockSize() << endl;
+        cout << "prefilter_sobel: " << bm->getPreFilterType() << endl;
         break;
     case Params::BP:
-        cout << "iter_count: " << bp.iters << endl;
-        cout << "level_count: " << bp.levels << endl;
+        cout << "iter_count: " << bp->getNumIters() << endl;
+        cout << "level_count: " << bp->getNumLevels() << endl;
         break;
     case Params::CSBP:
-        cout << "iter_count: " << csbp.iters << endl;
-        cout << "level_count: " << csbp.levels << endl;
+        cout << "iter_count: " << csbp->getNumIters() << endl;
+        cout << "level_count: " << csbp->getNumLevels() << endl;
         break;
     }
     cout << endl;
@@ -261,8 +262,8 @@ void App::handleKey(char key)
         }
         else
         {
-            cvtColor(left_src, left, CV_BGR2GRAY);
-            cvtColor(right_src, right, CV_BGR2GRAY);
+            cvtColor(left_src, left, COLOR_BGR2GRAY);
+            cvtColor(right_src, right, COLOR_BGR2GRAY);
         }
         d_left.upload(left);
         d_right.upload(right);
@@ -288,95 +289,93 @@ void App::handleKey(char key)
     case 's': case 'S':
         if (p.method == Params::BM)
         {
-            switch (bm.preset)
+            switch (bm->getPreFilterType())
             {
-            case gpu::StereoBM_GPU::BASIC_PRESET:
-                bm.preset = gpu::StereoBM_GPU::PREFILTER_XSOBEL;
+            case 0:
+                bm->setPreFilterType(cv::StereoBM::PREFILTER_XSOBEL);
                 break;
-            case gpu::StereoBM_GPU::PREFILTER_XSOBEL:
-                bm.preset = gpu::StereoBM_GPU::BASIC_PRESET;
+            case cv::StereoBM::PREFILTER_XSOBEL:
+                bm->setPreFilterType(0);
                 break;
             }
-            cout << "prefilter_sobel: " << bm.preset << endl;
+            cout << "prefilter_sobel: " << bm->getPreFilterType() << endl;
         }
         break;
     case '1':
         p.ndisp = p.ndisp == 1 ? 8 : p.ndisp + 8;
         cout << "ndisp: " << p.ndisp << endl;
-        bm.ndisp = p.ndisp;
-        bp.ndisp = p.ndisp;
-        csbp.ndisp = p.ndisp;
+        bm->setNumDisparities(p.ndisp);
+        bp->setNumDisparities(p.ndisp);
+        csbp->setNumDisparities(p.ndisp);
         break;
     case 'q': case 'Q':
         p.ndisp = max(p.ndisp - 8, 1);
         cout << "ndisp: " << p.ndisp << endl;
-        bm.ndisp = p.ndisp;
-        bp.ndisp = p.ndisp;
-        csbp.ndisp = p.ndisp;
+        bm->setNumDisparities(p.ndisp);
+        bp->setNumDisparities(p.ndisp);
+        csbp->setNumDisparities(p.ndisp);
         break;
     case '2':
         if (p.method == Params::BM)
         {
-            bm.winSize = min(bm.winSize + 1, 51);
-            cout << "win_size: " << bm.winSize << endl;
+            bm->setBlockSize(min(bm->getBlockSize() + 1, 51));
+            cout << "win_size: " << bm->getBlockSize() << endl;
         }
         break;
     case 'w': case 'W':
         if (p.method == Params::BM)
         {
-            bm.winSize = max(bm.winSize - 1, 2);
-            cout << "win_size: " << bm.winSize << endl;
+            bm->setBlockSize(max(bm->getBlockSize() - 1, 2));
+            cout << "win_size: " << bm->getBlockSize() << endl;
         }
         break;
     case '3':
         if (p.method == Params::BP)
         {
-            bp.iters += 1;
-            cout << "iter_count: " << bp.iters << endl;
+            bp->setNumIters(bp->getNumIters() + 1);
+            cout << "iter_count: " << bp->getNumIters() << endl;
         }
         else if (p.method == Params::CSBP)
         {
-            csbp.iters += 1;
-            cout << "iter_count: " << csbp.iters << endl;
+            csbp->setNumIters(csbp->getNumIters() + 1);
+            cout << "iter_count: " << csbp->getNumIters() << endl;
         }
         break;
     case 'e': case 'E':
         if (p.method == Params::BP)
         {
-            bp.iters = max(bp.iters - 1, 1);
-            cout << "iter_count: " << bp.iters << endl;
+            bp->setNumIters(max(bp->getNumIters() - 1, 1));
+            cout << "iter_count: " << bp->getNumIters() << endl;
         }
         else if (p.method == Params::CSBP)
         {
-            csbp.iters = max(csbp.iters - 1, 1);
-            cout << "iter_count: " << csbp.iters << endl;
+            csbp->setNumIters(max(csbp->getNumIters() - 1, 1));
+            cout << "iter_count: " << csbp->getNumIters() << endl;
         }
         break;
     case '4':
         if (p.method == Params::BP)
         {
-            bp.levels += 1;
-            cout << "level_count: " << bp.levels << endl;
+            bp->setNumLevels(bp->getNumLevels() + 1);
+            cout << "level_count: " << bp->getNumLevels() << endl;
         }
         else if (p.method == Params::CSBP)
         {
-            csbp.levels += 1;
-            cout << "level_count: " << csbp.levels << endl;
+            csbp->setNumLevels(csbp->getNumLevels() + 1);
+            cout << "level_count: " << csbp->getNumLevels() << endl;
         }
         break;
     case 'r': case 'R':
         if (p.method == Params::BP)
         {
-            bp.levels = max(bp.levels - 1, 1);
-            cout << "level_count: " << bp.levels << endl;
+            bp->setNumLevels(max(bp->getNumLevels() - 1, 1));
+            cout << "level_count: " << bp->getNumLevels() << endl;
         }
         else if (p.method == Params::CSBP)
         {
-            csbp.levels = max(csbp.levels - 1, 1);
-            cout << "level_count: " << csbp.levels << endl;
+            csbp->setNumLevels(max(csbp->getNumLevels() - 1, 1));
+            cout << "level_count: " << csbp->getNumLevels() << endl;
         }
         break;
     }
 }
-
-
diff --git a/samples/gpu/stereo_multi.cpp b/samples/gpu/stereo_multi.cpp
index d424bf90b4..430e15bd7b 100644
--- a/samples/gpu/stereo_multi.cpp
+++ b/samples/gpu/stereo_multi.cpp
@@ -1,152 +1,499 @@
-/* This sample demonstrates working on one piece of data using two GPUs.
-   It splits input into two parts and processes them separately on different
-   GPUs. */
+// This sample demonstrates working on one piece of data using two GPUs.
+// It splits input into two parts and processes them separately on different GPUs.
 
-// Disable some warnings which are caused with CUDA headers
-#if defined(_MSC_VER)
-#pragma warning(disable: 4201 4408 4100)
+#ifdef WIN32
+    #define NOMINMAX
+    #include <windows.h>
+#else
+    #include <pthread.h>
+    #include <unistd.h>
 #endif
 
 #include <iostream>
-#include "cvconfig.h"
-#include "opencv2/core/core.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/gpu/gpu.hpp"
+#include <iomanip>
 
-#if !defined(HAVE_CUDA) || !defined(HAVE_TBB)
-
-int main()
-{
-#if !defined(HAVE_CUDA)
-    std::cout << "CUDA support is required (CMake key 'WITH_CUDA' must be true).\n";
-#endif
-
-#if !defined(HAVE_TBB)
-    std::cout << "TBB support is required (CMake key 'WITH_TBB' must be true).\n";
-#endif
-
-    return 0;
-}
-
-#else
-
-#include "opencv2/core/internal.hpp" // For TBB wrappers
+#include "opencv2/core.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/contrib.hpp"
+#include "opencv2/gpustereo.hpp"
 
 using namespace std;
 using namespace cv;
 using namespace cv::gpu;
 
-struct Worker { void operator()(int device_id) const; };
+///////////////////////////////////////////////////////////
+// Thread
+// OS-specific wrappers for multi-threading
 
-// GPUs data
-GpuMat d_left[2];
-GpuMat d_right[2];
-StereoBM_GPU* bm[2];
-GpuMat d_result[2];
-
-// CPU result
-Mat result;
-
-static void printHelp()
+#ifdef WIN32
+class Thread
 {
-    std::cout << "Usage: stereo_multi_gpu --left <image> --right <image>\n";
+    struct UserData
+    {
+        void (*func)(void* userData);
+        void* param;
+    };
+
+    static DWORD WINAPI WinThreadFunction(LPVOID lpParam)
+    {
+        UserData* userData = static_cast<UserData*>(lpParam);
+
+        userData->func(userData->param);
+
+        return 0;
+    }
+
+    UserData userData_;
+    HANDLE thread_;
+    DWORD threadId_;
+
+public:
+    Thread(void (*func)(void* userData), void* userData)
+    {
+        userData_.func = func;
+        userData_.param = userData;
+
+        thread_ = CreateThread(
+            NULL,                   // default security attributes
+            0,                      // use default stack size
+            WinThreadFunction,      // thread function name
+            &userData_,             // argument to thread function
+            0,                      // use default creation flags
+            &threadId_);            // returns the thread identifier
+    }
+
+    ~Thread()
+    {
+        CloseHandle(thread_);
+    }
+
+    void wait()
+    {
+        WaitForSingleObject(thread_, INFINITE);
+    }
+};
+#else
+class Thread
+{
+    struct UserData
+    {
+        void (*func)(void* userData);
+        void* param;
+    };
+
+    static void* PThreadFunction(void* lpParam)
+    {
+        UserData* userData = static_cast<UserData*>(lpParam);
+
+        userData->func(userData->param);
+
+        return 0;
+    }
+
+    pthread_t thread_;
+    UserData userData_;
+
+public:
+    Thread(void (*func)(void* userData), void* userData)
+    {
+        userData_.func = func;
+        userData_.param = userData;
+
+        pthread_create(&thread_, NULL, PThreadFunction, &userData_);
+    }
+
+    ~Thread()
+    {
+        pthread_detach(thread_);
+    }
+
+    void wait()
+    {
+        pthread_join(thread_, NULL);
+    }
+};
+#endif
+
+///////////////////////////////////////////////////////////
+// StereoSingleGpu
+// Run Stereo algorithm on single GPU
+
+class StereoSingleGpu
+{
+public:
+    explicit StereoSingleGpu(int deviceId = 0);
+    ~StereoSingleGpu();
+
+    void compute(const Mat& leftFrame, const Mat& rightFrame, Mat& disparity);
+
+private:
+    int deviceId_;
+    GpuMat d_leftFrame;
+    GpuMat d_rightFrame;
+    GpuMat d_disparity;
+    Ptr<gpu::StereoBM> d_alg;
+};
+
+StereoSingleGpu::StereoSingleGpu(int deviceId) : deviceId_(deviceId)
+{
+    gpu::setDevice(deviceId_);
+    d_alg = gpu::createStereoBM(256);
 }
 
+StereoSingleGpu::~StereoSingleGpu()
+{
+    gpu::setDevice(deviceId_);
+    d_leftFrame.release();
+    d_rightFrame.release();
+    d_disparity.release();
+    d_alg.release();
+}
+
+void StereoSingleGpu::compute(const Mat& leftFrame, const Mat& rightFrame, Mat& disparity)
+{
+    gpu::setDevice(deviceId_);
+    d_leftFrame.upload(leftFrame);
+    d_rightFrame.upload(rightFrame);
+    d_alg->compute(d_leftFrame, d_rightFrame, d_disparity);
+    d_disparity.download(disparity);
+}
+
+///////////////////////////////////////////////////////////
+// StereoMultiGpuThread
+// Run Stereo algorithm on two GPUs using different host threads
+
+class StereoMultiGpuThread
+{
+public:
+    StereoMultiGpuThread();
+    ~StereoMultiGpuThread();
+
+    void compute(const Mat& leftFrame, const Mat& rightFrame, Mat& disparity);
+
+private:
+    GpuMat d_leftFrames[2];
+    GpuMat d_rightFrames[2];
+    GpuMat d_disparities[2];
+    Ptr<gpu::StereoBM> d_algs[2];
+
+    struct StereoLaunchData
+    {
+        int deviceId;
+        Mat leftFrame;
+        Mat rightFrame;
+        Mat disparity;
+        GpuMat* d_leftFrame;
+        GpuMat* d_rightFrame;
+        GpuMat* d_disparity;
+        Ptr<gpu::StereoBM> d_alg;
+    };
+
+    static void launchGpuStereoAlg(void* userData);
+};
+
+StereoMultiGpuThread::StereoMultiGpuThread()
+{
+    gpu::setDevice(0);
+    d_algs[0] = gpu::createStereoBM(256);
+
+    gpu::setDevice(1);
+    d_algs[1] = gpu::createStereoBM(256);
+}
+
+StereoMultiGpuThread::~StereoMultiGpuThread()
+{
+    gpu::setDevice(0);
+    d_leftFrames[0].release();
+    d_rightFrames[0].release();
+    d_disparities[0].release();
+    d_algs[0].release();
+
+    gpu::setDevice(1);
+    d_leftFrames[1].release();
+    d_rightFrames[1].release();
+    d_disparities[1].release();
+    d_algs[1].release();
+}
+
+void StereoMultiGpuThread::compute(const Mat& leftFrame, const Mat& rightFrame, Mat& disparity)
+{
+    disparity.create(leftFrame.size(), CV_8UC1);
+
+    // Split input data onto two parts for each GPUs.
+    // We add small border for each part,
+    // because original algorithm doesn't calculate disparity on image borders.
+    // With such padding we will get output in the middle of final result.
+
+    StereoLaunchData launchDatas[2];
+
+    launchDatas[0].deviceId = 0;
+    launchDatas[0].leftFrame = leftFrame.rowRange(0, leftFrame.rows / 2 + 32);
+    launchDatas[0].rightFrame = rightFrame.rowRange(0, rightFrame.rows / 2 + 32);
+    launchDatas[0].disparity = disparity.rowRange(0, leftFrame.rows / 2);
+    launchDatas[0].d_leftFrame = &d_leftFrames[0];
+    launchDatas[0].d_rightFrame = &d_rightFrames[0];
+    launchDatas[0].d_disparity = &d_disparities[0];
+    launchDatas[0].d_alg = d_algs[0];
+
+    launchDatas[1].deviceId = 1;
+    launchDatas[1].leftFrame = leftFrame.rowRange(leftFrame.rows / 2 - 32, leftFrame.rows);
+    launchDatas[1].rightFrame = rightFrame.rowRange(leftFrame.rows / 2 - 32, leftFrame.rows);
+    launchDatas[1].disparity = disparity.rowRange(leftFrame.rows / 2, leftFrame.rows);
+    launchDatas[1].d_leftFrame = &d_leftFrames[1];
+    launchDatas[1].d_rightFrame = &d_rightFrames[1];
+    launchDatas[1].d_disparity = &d_disparities[1];
+    launchDatas[1].d_alg = d_algs[1];
+
+    Thread thread0(launchGpuStereoAlg, &launchDatas[0]);
+    Thread thread1(launchGpuStereoAlg, &launchDatas[1]);
+
+    thread0.wait();
+    thread1.wait();
+}
+
+void StereoMultiGpuThread::launchGpuStereoAlg(void* userData)
+{
+    StereoLaunchData* data = static_cast<StereoLaunchData*>(userData);
+
+    gpu::setDevice(data->deviceId);
+    data->d_leftFrame->upload(data->leftFrame);
+    data->d_rightFrame->upload(data->rightFrame);
+    data->d_alg->compute(*data->d_leftFrame, *data->d_rightFrame, *data->d_disparity);
+
+    if (data->deviceId == 0)
+        data->d_disparity->rowRange(0, data->d_disparity->rows - 32).download(data->disparity);
+    else
+        data->d_disparity->rowRange(32, data->d_disparity->rows).download(data->disparity);
+}
+
+///////////////////////////////////////////////////////////
+// StereoMultiGpuStream
+// Run Stereo algorithm on two GPUs from single host thread using async API
+
+class StereoMultiGpuStream
+{
+public:
+    StereoMultiGpuStream();
+    ~StereoMultiGpuStream();
+
+    void compute(const CudaMem& leftFrame, const CudaMem& rightFrame, CudaMem& disparity);
+
+private:
+    GpuMat d_leftFrames[2];
+    GpuMat d_rightFrames[2];
+    GpuMat d_disparities[2];
+    Ptr<gpu::StereoBM> d_algs[2];
+    Ptr<Stream> streams[2];
+};
+
+StereoMultiGpuStream::StereoMultiGpuStream()
+{
+    gpu::setDevice(0);
+    d_algs[0] = gpu::createStereoBM(256);
+    streams[0] = makePtr<Stream>();
+
+    gpu::setDevice(1);
+    d_algs[1] = gpu::createStereoBM(256);
+    streams[1] = makePtr<Stream>();
+}
+
+StereoMultiGpuStream::~StereoMultiGpuStream()
+{
+    gpu::setDevice(0);
+    d_leftFrames[0].release();
+    d_rightFrames[0].release();
+    d_disparities[0].release();
+    d_algs[0].release();
+    streams[0].release();
+
+    gpu::setDevice(1);
+    d_leftFrames[1].release();
+    d_rightFrames[1].release();
+    d_disparities[1].release();
+    d_algs[1].release();
+    streams[1].release();
+}
+
+void StereoMultiGpuStream::compute(const CudaMem& leftFrame, const CudaMem& rightFrame, CudaMem& disparity)
+{
+    disparity.create(leftFrame.size(), CV_8UC1);
+
+    // Split input data onto two parts for each GPUs.
+    // We add small border for each part,
+    // because original algorithm doesn't calculate disparity on image borders.
+    // With such padding we will get output in the middle of final result.
+
+    Mat leftFrameHdr = leftFrame.createMatHeader();
+    Mat rightFrameHdr = rightFrame.createMatHeader();
+    Mat disparityHdr = disparity.createMatHeader();
+    Mat disparityPart0 = disparityHdr.rowRange(0, leftFrame.rows / 2);
+    Mat disparityPart1 = disparityHdr.rowRange(leftFrame.rows / 2, leftFrame.rows);
+
+    gpu::setDevice(0);
+    d_leftFrames[0].upload(leftFrameHdr.rowRange(0, leftFrame.rows / 2 + 32), *streams[0]);
+    d_rightFrames[0].upload(rightFrameHdr.rowRange(0, leftFrame.rows / 2 + 32), *streams[0]);
+    d_algs[0]->compute(d_leftFrames[0], d_rightFrames[0], d_disparities[0], *streams[0]);
+    d_disparities[0].rowRange(0, leftFrame.rows / 2).download(disparityPart0, *streams[0]);
+
+    gpu::setDevice(1);
+    d_leftFrames[1].upload(leftFrameHdr.rowRange(leftFrame.rows / 2 - 32, leftFrame.rows), *streams[1]);
+    d_rightFrames[1].upload(rightFrameHdr.rowRange(leftFrame.rows / 2 - 32, leftFrame.rows), *streams[1]);
+    d_algs[1]->compute(d_leftFrames[1], d_rightFrames[1], d_disparities[1], *streams[1]);
+    d_disparities[1].rowRange(32, d_disparities[1].rows).download(disparityPart1, *streams[1]);
+
+    gpu::setDevice(0);
+    streams[0]->waitForCompletion();
+
+    gpu::setDevice(1);
+    streams[1]->waitForCompletion();
+}
+
+///////////////////////////////////////////////////////////
+// main
+
 int main(int argc, char** argv)
 {
-    if (argc < 5)
+    if (argc != 3)
     {
-        printHelp();
+        cerr << "Usage: stereo_multi_gpu <left_video> <right_video>" << endl;
         return -1;
     }
 
-    int num_devices = getCudaEnabledDeviceCount();
-    if (num_devices < 2)
+    const int numDevices = getCudaEnabledDeviceCount();
+    if (numDevices != 2)
     {
-        std::cout << "Two or more GPUs are required\n";
+        cerr << "Two GPUs are required" << endl;
         return -1;
     }
-    for (int i = 0; i < num_devices; ++i)
-    {
-        cv::gpu::printShortCudaDeviceInfo(i);
 
-        DeviceInfo dev_info(i);
-        if (!dev_info.isCompatible())
+    for (int i = 0; i < numDevices; ++i)
+    {
+        DeviceInfo devInfo(i);
+        if (!devInfo.isCompatible())
         {
-            std::cout << "GPU module isn't built for GPU #" << i << " ("
-                 << dev_info.name() << ", CC " << dev_info.majorVersion()
-                 << dev_info.minorVersion() << "\n";
+            cerr << "GPU module was't built for GPU #" << i << " ("
+                 << devInfo.name() << ", CC " << devInfo.majorVersion()
+                 << devInfo.minorVersion() << endl;
             return -1;
         }
+
+        printShortCudaDeviceInfo(i);
     }
 
-    // Load input data
-    Mat left, right;
-    for (int i = 1; i < argc; ++i)
+    VideoCapture leftVideo(argv[1]);
+    VideoCapture rightVideo(argv[2]);
+
+    if (!leftVideo.isOpened())
     {
-        if (string(argv[i]) == "--left")
+         cerr << "Can't open " << argv[1] << " video file" << endl;
+         return -1;
+    }
+
+    if (!rightVideo.isOpened())
+    {
+         cerr << "Can't open " << argv[2] << " video file" << endl;
+         return -1;
+    }
+
+    cout << endl;
+    cout << "This sample demonstrates working on one piece of data using two GPUs." << endl;
+    cout << "It splits input into two parts and processes them separately on different GPUs." << endl;
+    cout << endl;
+
+    Mat leftFrame, rightFrame;
+    CudaMem leftGrayFrame, rightGrayFrame;
+
+    StereoSingleGpu gpu0Alg(0);
+    StereoSingleGpu gpu1Alg(1);
+    StereoMultiGpuThread multiThreadAlg;
+    StereoMultiGpuStream multiStreamAlg;
+
+    Mat disparityGpu0;
+    Mat disparityGpu1;
+    Mat disparityMultiThread;
+    CudaMem disparityMultiStream;
+
+    Mat disparityGpu0Show;
+    Mat disparityGpu1Show;
+    Mat disparityMultiThreadShow;
+    Mat disparityMultiStreamShow;
+
+    TickMeter tm;
+
+    cout << "-------------------------------------------------------------------" << endl;
+    cout << "| Frame | GPU 0 ms | GPU 1 ms | Multi Thread ms | Multi Stream ms |" << endl;
+    cout << "-------------------------------------------------------------------" << endl;
+
+    for (int i = 0;; ++i)
+    {
+        leftVideo >> leftFrame;
+        rightVideo >> rightFrame;
+
+        if (leftFrame.empty() || rightFrame.empty())
+            break;
+
+        if (leftFrame.size() != rightFrame.size())
         {
-            left = imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE);
-            CV_Assert(!left.empty());
-        }
-        else if (string(argv[i]) == "--right")
-        {
-            right = imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE);
-            CV_Assert(!right.empty());
-        }
-        else if (string(argv[i]) == "--help")
-        {
-            printHelp();
+            cerr << "Frames have different sizes" << endl;
             return -1;
         }
+
+        leftGrayFrame.create(leftFrame.size(), CV_8UC1);
+        rightGrayFrame.create(leftFrame.size(), CV_8UC1);
+
+        cvtColor(leftFrame, leftGrayFrame.createMatHeader(), COLOR_BGR2GRAY);
+        cvtColor(rightFrame, rightGrayFrame.createMatHeader(), COLOR_BGR2GRAY);
+
+        tm.reset(); tm.start();
+        gpu0Alg.compute(leftGrayFrame.createMatHeader(), rightGrayFrame.createMatHeader(),
+                        disparityGpu0);
+        tm.stop();
+
+        const double gpu0Time = tm.getTimeMilli();
+
+        tm.reset(); tm.start();
+        gpu1Alg.compute(leftGrayFrame.createMatHeader(), rightGrayFrame.createMatHeader(),
+                        disparityGpu1);
+        tm.stop();
+
+        const double gpu1Time = tm.getTimeMilli();
+
+        tm.reset(); tm.start();
+        multiThreadAlg.compute(leftGrayFrame.createMatHeader(), rightGrayFrame.createMatHeader(),
+                               disparityMultiThread);
+        tm.stop();
+
+        const double multiThreadTime = tm.getTimeMilli();
+
+        tm.reset(); tm.start();
+        multiStreamAlg.compute(leftGrayFrame, rightGrayFrame, disparityMultiStream);
+        tm.stop();
+
+        const double multiStreamTime = tm.getTimeMilli();
+
+        cout << "| " << setw(5) << i << " | "
+             << setw(8) << setprecision(1) << fixed << gpu0Time << " | "
+             << setw(8) << setprecision(1) << fixed << gpu1Time << " | "
+             << setw(15) << setprecision(1) << fixed << multiThreadTime << " | "
+             << setw(15) << setprecision(1) << fixed << multiStreamTime << " |" << endl;
+
+        resize(disparityGpu0, disparityGpu0Show, Size(1024, 768), 0, 0, INTER_AREA);
+        resize(disparityGpu1, disparityGpu1Show, Size(1024, 768), 0, 0, INTER_AREA);
+        resize(disparityMultiThread, disparityMultiThreadShow, Size(1024, 768), 0, 0, INTER_AREA);
+        resize(disparityMultiStream.createMatHeader(), disparityMultiStreamShow, Size(1024, 768), 0, 0, INTER_AREA);
+
+        imshow("disparityGpu0", disparityGpu0Show);
+        imshow("disparityGpu1", disparityGpu1Show);
+        imshow("disparityMultiThread", disparityMultiThreadShow);
+        imshow("disparityMultiStream", disparityMultiStreamShow);
+
+        const int key = waitKey(30) & 0xff;
+        if (key == 27)
+            break;
     }
 
-    // Split source images for processing on the GPU #0
-    setDevice(0);
-    d_left[0].upload(left.rowRange(0, left.rows / 2));
-    d_right[0].upload(right.rowRange(0, right.rows / 2));
-    bm[0] = new StereoBM_GPU();
+    cout << "-------------------------------------------------------------------" << endl;
 
-    // Split source images for processing on the GPU #1
-    setDevice(1);
-    d_left[1].upload(left.rowRange(left.rows / 2, left.rows));
-    d_right[1].upload(right.rowRange(right.rows / 2, right.rows));
-    bm[1] = new StereoBM_GPU();
-
-    // Execute calculation in two threads using two GPUs
-    int devices[] = {0, 1};
-    parallel_do(devices, devices + 2, Worker());
-
-    // Release the first GPU resources
-    setDevice(0);
-    imshow("GPU #0 result", Mat(d_result[0]));
-    d_left[0].release();
-    d_right[0].release();
-    d_result[0].release();
-    delete bm[0];
-
-    // Release the second GPU resources
-    setDevice(1);
-    imshow("GPU #1 result", Mat(d_result[1]));
-    d_left[1].release();
-    d_right[1].release();
-    d_result[1].release();
-    delete bm[1];
-
-    waitKey();
     return 0;
 }
-
-
-void Worker::operator()(int device_id) const
-{
-    setDevice(device_id);
-
-    bm[device_id]->operator()(d_left[device_id], d_right[device_id],
-                              d_result[device_id]);
-
-    std::cout << "GPU #" << device_id << " (" << DeviceInfo().name()
-        << "): finished\n";
-}
-
-#endif
diff --git a/samples/gpu/super_resolution.cpp b/samples/gpu/super_resolution.cpp
new file mode 100644
index 0000000000..e124b375e3
--- /dev/null
+++ b/samples/gpu/super_resolution.cpp
@@ -0,0 +1,267 @@
+#include <iostream>
+#include <iomanip>
+#include <string>
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/contrib.hpp"
+#include "opencv2/superres.hpp"
+#include "opencv2/superres/optical_flow.hpp"
+#include "opencv2/opencv_modules.hpp"
+
+#if defined(HAVE_OPENCV_OCL)
+#include "opencv2/ocl/ocl.hpp"
+#endif
+
+using namespace std;
+using namespace cv;
+using namespace cv::superres;
+bool useOclChanged;
+#define MEASURE_TIME(op) \
+    { \
+        TickMeter tm; \
+        tm.start(); \
+        op; \
+        tm.stop(); \
+        cout << tm.getTimeSec() << " sec" << endl; \
+    }
+
+static Ptr<DenseOpticalFlowExt> createOptFlow(const string& name, bool useGpu)
+{
+    if (name == "farneback")
+    {
+        if (useGpu)
+            return createOptFlow_Farneback_GPU();
+        else
+            return createOptFlow_Farneback();
+    }
+    else if (name == "simple")
+        return createOptFlow_Simple();
+    else if (name == "tvl1")
+    {
+        if (useGpu)
+            return createOptFlow_DualTVL1_GPU();
+        else
+            return createOptFlow_DualTVL1();
+    }
+    else if (name == "brox")
+        return createOptFlow_Brox_GPU();
+    else if (name == "pyrlk")
+        return createOptFlow_PyrLK_GPU();
+    else
+    {
+        cerr << "Incorrect Optical Flow algorithm - " << name << endl;
+    }
+    return Ptr<DenseOpticalFlowExt>();
+}
+#if defined(HAVE_OPENCV_OCL)
+static Ptr<DenseOpticalFlowExt> createOptFlow(const string& name)
+{
+    if (name == "farneback")
+    {
+        return createOptFlow_Farneback_OCL();
+    }
+    else if (name == "simple")
+    {
+        useOclChanged = true;
+        std::cout<<"simple on OpenCL has not been implemented. Use CPU instead!\n";
+        return createOptFlow_Simple();
+    }
+    else if (name == "tvl1")
+        return createOptFlow_DualTVL1_OCL();
+    else if (name == "brox")
+    {
+        std::cout<<"brox has not been implemented!\n";
+        return Ptr<DenseOpticalFlowExt>();
+    }
+    else if (name == "pyrlk")
+        return createOptFlow_PyrLK_OCL();
+    else
+    {
+        cerr << "Incorrect Optical Flow algorithm - " << name << endl;
+    }
+    return Ptr<DenseOpticalFlowExt>();
+}
+#endif
+int main(int argc, const char* argv[])
+{
+    useOclChanged = false;
+    CommandLineParser cmd(argc, argv,
+        "{ v video      |           | Input video }"
+        "{ o output     |           | Output video }"
+        "{ s scale      | 4         | Scale factor }"
+        "{ i iterations | 180       | Iteration count }"
+        "{ t temporal   | 4         | Radius of the temporal search area }"
+        "{ f flow       | farneback | Optical flow algorithm (farneback, simple, tvl1, brox, pyrlk) }"
+        "{ g            | false     | CPU as default device, cuda for CUDA and ocl for OpenCL }"
+        "{ h help       | false     | Print help message }"
+    );
+
+    if (cmd.get<bool>("help"))
+    {
+        cout << "This sample demonstrates Super Resolution algorithms for video sequence" << endl;
+        cmd.printMessage();
+        return 0;
+    }
+
+    const string inputVideoName = cmd.get<string>("video");
+    const string outputVideoName = cmd.get<string>("output");
+    const int scale = cmd.get<int>("scale");
+    const int iterations = cmd.get<int>("iterations");
+    const int temporalAreaRadius = cmd.get<int>("temporal");
+    const string optFlow = cmd.get<string>("flow");
+    string gpuOption = cmd.get<string>("gpu");
+
+    std::transform(gpuOption.begin(), gpuOption.end(), gpuOption.begin(), ::tolower);
+
+    bool useCuda = false;
+    bool useOcl = false;
+
+    if(gpuOption.compare("ocl") == 0)
+        useOcl = true;
+    else if(gpuOption.compare("cuda") == 0)
+        useCuda = true;
+
+#ifndef HAVE_OPENCV_OCL
+    if(useOcl)
+    {
+        {
+            cout<<"OPENCL is not compiled\n";
+            return 0;
+        }
+    }
+#endif
+#if defined(HAVE_OPENCV_OCL)
+    std::vector<cv::ocl::Info>info;
+    if(useCuda)
+    {
+        CV_Assert(!useOcl);
+        info.clear();
+    }
+
+    if(useOcl)
+    {
+        CV_Assert(!useCuda);
+        cv::ocl::getDevice(info);
+    }
+#endif
+    Ptr<SuperResolution> superRes;
+
+
+#if defined(HAVE_OPENCV_OCL)
+    if(useOcl)
+    {
+        Ptr<DenseOpticalFlowExt> of = createOptFlow(optFlow);
+        if (of.empty())
+            exit(-1);
+        if(useOclChanged)
+        {
+            superRes = createSuperResolution_BTVL1();
+            useOcl = !useOcl;
+        }else
+            superRes = createSuperResolution_BTVL1_OCL();
+        superRes->set("opticalFlow", of);
+    }
+    else
+#endif
+    {
+        if (useCuda)
+            superRes = createSuperResolution_BTVL1_GPU();
+        else
+            superRes = createSuperResolution_BTVL1();
+
+        Ptr<DenseOpticalFlowExt> of = createOptFlow(optFlow, useCuda);
+
+        if (of.empty())
+            exit(-1);
+        superRes->set("opticalFlow", of);
+    }
+
+    superRes->set("scale", scale);
+    superRes->set("iterations", iterations);
+    superRes->set("temporalAreaRadius", temporalAreaRadius);
+
+    Ptr<FrameSource> frameSource;
+    if (useCuda)
+    {
+        // Try to use gpu Video Decoding
+        try
+        {
+            frameSource = createFrameSource_Video_GPU(inputVideoName);
+            Mat frame;
+            frameSource->nextFrame(frame);
+        }
+        catch (const cv::Exception&)
+        {
+            frameSource.release();
+        }
+    }
+    if (!frameSource)
+        frameSource = createFrameSource_Video(inputVideoName);
+
+    // skip first frame, it is usually corrupted
+    {
+        Mat frame;
+        frameSource->nextFrame(frame);
+        cout << "Input           : " << inputVideoName << " " << frame.size() << endl;
+        cout << "Scale factor    : " << scale << endl;
+        cout << "Iterations      : " << iterations << endl;
+        cout << "Temporal radius : " << temporalAreaRadius << endl;
+        cout << "Optical Flow    : " << optFlow << endl;
+#if defined(HAVE_OPENCV_OCL)
+        cout << "Mode            : " << (useCuda ? "CUDA" : useOcl? "OpenCL" : "CPU") << endl;
+#else
+        cout << "Mode            : " << (useCuda ? "CUDA" : "CPU") << endl;
+#endif
+    }
+
+    superRes->setInput(frameSource);
+
+    VideoWriter writer;
+
+    for (int i = 0;; ++i)
+    {
+        cout << '[' << setw(3) << i << "] : ";
+        Mat result;
+
+#if defined(HAVE_OPENCV_OCL)
+        cv::ocl::oclMat result_;
+
+        if(useOcl)
+        {
+            MEASURE_TIME(superRes->nextFrame(result_));
+        }
+        else
+#endif
+        {
+            MEASURE_TIME(superRes->nextFrame(result));
+        }
+
+#ifdef HAVE_OPENCV_OCL
+        if(useOcl)
+        {
+            if(!result_.empty())
+            {
+                result_.download(result);
+            }
+        }
+#endif
+        if (result.empty())
+            break;
+
+        imshow("Super Resolution", result);
+
+        if (waitKey(1000) > 0)
+            break;
+
+        if (!outputVideoName.empty())
+        {
+            if (!writer.isOpened())
+                writer.open(outputVideoName, VideoWriter::fourcc('X', 'V', 'I', 'D'), 25.0, result.size());
+            writer << result;
+        }
+    }
+
+    return 0;
+}
diff --git a/samples/gpu/surf_keypoint_matcher.cpp b/samples/gpu/surf_keypoint_matcher.cpp
index 617cda52bd..fd3578d9f4 100644
--- a/samples/gpu/surf_keypoint_matcher.cpp
+++ b/samples/gpu/surf_keypoint_matcher.cpp
@@ -1,9 +1,14 @@
 #include <iostream>
 
+#include "opencv2/opencv_modules.hpp"
+
+#ifdef HAVE_OPENCV_NONFREE
+
 #include "opencv2/core/core.hpp"
 #include "opencv2/features2d/features2d.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/gpu/gpu.hpp"
+#include "opencv2/nonfree/gpu.hpp"
 
 using namespace std;
 using namespace cv;
@@ -28,12 +33,12 @@ int main(int argc, char* argv[])
     {
         if (string(argv[i]) == "--left")
         {
-            img1.upload(imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE));
+            img1.upload(imread(argv[++i], IMREAD_GRAYSCALE));
             CV_Assert(!img1.empty());
         }
         else if (string(argv[i]) == "--right")
         {
-            img2.upload(imread(argv[++i], CV_LOAD_IMAGE_GRAYSCALE));
+            img2.upload(imread(argv[++i], IMREAD_GRAYSCALE));
             CV_Assert(!img2.empty());
         }
         else if (string(argv[i]) == "--help")
@@ -81,3 +86,13 @@ int main(int argc, char* argv[])
 
     return 0;
 }
+
+#else
+
+int main()
+{
+    std::cerr << "OpenCV was built without nonfree module" << std::endl;
+    return 0;
+}
+
+#endif
diff --git a/samples/gpu/video_reader.cpp b/samples/gpu/video_reader.cpp
index a4c068516d..42f6f91db4 100644
--- a/samples/gpu/video_reader.cpp
+++ b/samples/gpu/video_reader.cpp
@@ -1,14 +1,19 @@
 #include <iostream>
+
+#include "opencv2/opencv_modules.hpp"
+
+#if defined(HAVE_OPENCV_GPUCODEC)
+
 #include <string>
 #include <vector>
 #include <algorithm>
 #include <numeric>
 
-#include <opencv2/core/core.hpp>
+#include <opencv2/core.hpp>
 #include <opencv2/core/opengl.hpp>
-#include <opencv2/gpu/gpu.hpp>
-#include <opencv2/highgui/highgui.hpp>
-#include <opencv2/contrib/contrib.hpp>
+#include <opencv2/gpucodec.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/contrib.hpp>
 
 int main(int argc, const char* argv[])
 {
@@ -25,8 +30,7 @@ int main(int argc, const char* argv[])
     cv::VideoCapture reader(fname);
 
     cv::gpu::GpuMat d_frame;
-    cv::gpu::VideoReader_GPU d_reader(fname);
-    d_reader.dumpFormat(std::cout);
+    cv::Ptr<cv::gpucodec::VideoReader> d_reader = cv::gpucodec::createVideoReader(fname);
 
     cv::TickMeter tm;
     std::vector<double> cpu_times;
@@ -41,7 +45,7 @@ int main(int argc, const char* argv[])
         cpu_times.push_back(tm.getTimeMilli());
 
         tm.reset(); tm.start();
-        if (!d_reader.read(d_frame))
+        if (!d_reader->nextFrame(d_frame))
             break;
         tm.stop();
         gpu_times.push_back(tm.getTimeMilli());
@@ -69,3 +73,13 @@ int main(int argc, const char* argv[])
 
     return 0;
 }
+
+#else
+
+int main()
+{
+    std::cout << "OpenCV was built without GPU Video decoding support\n" << std::endl;
+    return 0;
+}
+
+#endif
diff --git a/samples/gpu/video_writer.cpp b/samples/gpu/video_writer.cpp
index e5154d53a0..c1bcc5d368 100644
--- a/samples/gpu/video_writer.cpp
+++ b/samples/gpu/video_writer.cpp
@@ -1,11 +1,16 @@
 #include <iostream>
+
+#include "opencv2/opencv_modules.hpp"
+
+#if defined(HAVE_OPENCV_GPUCODEC) && defined(WIN32)
+
 #include <vector>
 #include <numeric>
 
-#include "opencv2/core/core.hpp"
-#include "opencv2/gpu/gpu.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/contrib/contrib.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/gpucodec.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/contrib.hpp"
 
 int main(int argc, const char* argv[])
 {
@@ -28,7 +33,7 @@ int main(int argc, const char* argv[])
     cv::gpu::printShortCudaDeviceInfo(cv::gpu::getDevice());
 
     cv::VideoWriter writer;
-    cv::gpu::VideoWriter_GPU d_writer;
+    cv::Ptr<cv::gpucodec::VideoWriter> d_writer;
 
     cv::Mat frame;
     cv::gpu::GpuMat d_frame;
@@ -55,15 +60,15 @@ int main(int argc, const char* argv[])
 
             std::cout << "Open CPU Writer" << std::endl;
 
-            if (!writer.open("output_cpu.avi", CV_FOURCC('X', 'V', 'I', 'D'), FPS, frame.size()))
+            if (!writer.open("output_cpu.avi", cv::VideoWriter::fourcc('X', 'V', 'I', 'D'), FPS, frame.size()))
                 return -1;
         }
 
-        if (!d_writer.isOpened())
+        if (d_writer.empty())
         {
             std::cout << "Open GPU Writer" << std::endl;
 
-            d_writer.open("output_gpu.avi", frame.size(), FPS);
+            d_writer = cv::gpucodec::createVideoWriter("output_gpu.avi", frame.size(), FPS);
         }
 
         d_frame.upload(frame);
@@ -76,7 +81,7 @@ int main(int argc, const char* argv[])
         cpu_times.push_back(tm.getTimeMilli());
 
         tm.reset(); tm.start();
-        d_writer.write(d_frame);
+        d_writer->write(d_frame);
         tm.stop();
         gpu_times.push_back(tm.getTimeMilli());
     }
@@ -94,3 +99,13 @@ int main(int argc, const char* argv[])
 
     return 0;
 }
+
+#else
+
+int main()
+{
+    std::cout << "OpenCV was built without GPU Video encoding support\n" << std::endl;
+    return 0;
+}
+
+#endif
diff --git a/samples/java/sbt/src/main/java/DetectFaceDemo.java b/samples/java/sbt/src/main/java/DetectFaceDemo.java
index fb08567237..07b4202e6a 100644
--- a/samples/java/sbt/src/main/java/DetectFaceDemo.java
+++ b/samples/java/sbt/src/main/java/DetectFaceDemo.java
@@ -41,4 +41,4 @@ public class DetectFaceDemo {
         System.out.println(String.format("Writing %s", filename));
         Highgui.imwrite(filename, image);
     }
-}
\ No newline at end of file
+}
diff --git a/samples/java/sbt/src/main/scala/ScalaCorrespondenceMatchingDemo.scala b/samples/java/sbt/src/main/scala/ScalaCorrespondenceMatchingDemo.scala
index 30ab0553e3..ea50200c9a 100644
--- a/samples/java/sbt/src/main/scala/ScalaCorrespondenceMatchingDemo.scala
+++ b/samples/java/sbt/src/main/scala/ScalaCorrespondenceMatchingDemo.scala
@@ -66,4 +66,4 @@ object ScalaCorrespondenceMatchingDemo {
     println(s"Writing ${filename}")
     assert(Highgui.imwrite(filename, correspondenceImage))
   }
-}
\ No newline at end of file
+}
diff --git a/samples/java/sbt/src/main/scala/ScalaDetectFaceDemo.scala b/samples/java/sbt/src/main/scala/ScalaDetectFaceDemo.scala
index a35eeb7784..85fdeb0d7f 100644
--- a/samples/java/sbt/src/main/scala/ScalaDetectFaceDemo.scala
+++ b/samples/java/sbt/src/main/scala/ScalaDetectFaceDemo.scala
@@ -40,4 +40,4 @@ object ScalaDetectFaceDemo {
     println(s"Writing ${filename}")
     assert(Highgui.imwrite(filename, image))
   }
-}
\ No newline at end of file
+}
diff --git a/samples/ocl/CMakeLists.txt b/samples/ocl/CMakeLists.txt
index 40fe0e6e36..9b04dc3976 100644
--- a/samples/ocl/CMakeLists.txt
+++ b/samples/ocl/CMakeLists.txt
@@ -17,10 +17,6 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
     ocv_include_directories(${OPENCL_INCLUDE_DIR})
   endif()
 
-  if(CMAKE_COMPILER_IS_GNUCXX AND NOT ENABLE_NOISY_WARNINGS)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-unused-function")
-  endif()
-
   # ---------------------------------------------
   #      Define executable targets
   # ---------------------------------------------
@@ -31,7 +27,7 @@ if(BUILD_EXAMPLES AND OCV_DEPENDENCIES_FOUND)
     target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_OCL_SAMPLES_REQUIRED_DEPS})
 
     set_target_properties(${the_target} PROPERTIES
-      OUTPUT_NAME "${name}_${project}"
+      OUTPUT_NAME "${project}-example-${name}"
       PROJECT_LABEL "(EXAMPLE_${project_upper}) ${name}")
 
     if(ENABLE_SOLUTION_FOLDERS)
@@ -58,7 +54,6 @@ endif()
 if (INSTALL_C_EXAMPLES AND NOT WIN32)
   file(GLOB install_list *.c *.cpp *.jpg *.png *.data makefile.* build_all.sh *.dsp *.cmd )
   install(FILES ${install_list}
-          DESTINATION share/opencv/samples/${project}
+          DESTINATION share/OpenCV/samples/${project}
           PERMISSIONS OWNER_READ GROUP_READ WORLD_READ)
 endif()
-
diff --git a/samples/ocl/adaptive_bilateral_filter.cpp b/samples/ocl/adaptive_bilateral_filter.cpp
new file mode 100644
index 0000000000..d6a2415f9a
--- /dev/null
+++ b/samples/ocl/adaptive_bilateral_filter.cpp
@@ -0,0 +1,52 @@
+// This sample shows the difference of adaptive bilateral filter and bilateral filter.
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/ocl.hpp"
+
+using namespace cv;
+using namespace std;
+
+
+int main( int argc, const char** argv )
+{
+    const char* keys =
+        "{ i input   |          | specify input image }"
+        "{ k ksize   |     5    | specify kernel size }";
+    CommandLineParser cmd(argc, argv, keys);
+    string src_path = cmd.get<string>("i");
+    int ks = cmd.get<int>("k");
+    const char * winName[] = {"input", "adaptive bilateral CPU", "adaptive bilateral OpenCL", "bilateralFilter OpenCL"};
+
+    Mat src = imread(src_path);
+    Mat abFilterCPU;
+    if(src.empty()){
+        //cout << "error read image: " << src_path << endl;
+        return -1;
+    }
+
+    std::vector<ocl::Info> infos;
+    ocl::getDevice(infos);
+
+    ocl::oclMat dsrc(src), dABFilter, dBFilter;
+
+    Size ksize(ks, ks);
+    adaptiveBilateralFilter(src,abFilterCPU, ksize, 10);
+    ocl::adaptiveBilateralFilter(dsrc, dABFilter, ksize, 10);
+    ocl::bilateralFilter(dsrc, dBFilter, ks, 30, 9);
+
+    Mat abFilter = dABFilter;
+    Mat bFilter = dBFilter;
+    imshow(winName[0], src);
+
+    imshow(winName[1], abFilterCPU);
+
+    imshow(winName[2], abFilter);
+
+    imshow(winName[3], bFilter);
+
+    waitKey();
+    return 0;
+
+}
diff --git a/samples/ocl/bgfg_segm.cpp b/samples/ocl/bgfg_segm.cpp
new file mode 100644
index 0000000000..87b3ea4eec
--- /dev/null
+++ b/samples/ocl/bgfg_segm.cpp
@@ -0,0 +1,136 @@
+#include <iostream>
+#include <string>
+
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/ocl.hpp"
+#include "opencv2/highgui.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+#define M_MOG  1
+#define M_MOG2 2
+
+int main(int argc, const char** argv)
+{
+
+    cv::CommandLineParser cmd(argc, argv,
+        "{ c camera | false       | use camera }"
+        "{ f file   | 768x576.avi | input video file }"
+        "{ m method | mog         | method (mog, mog2) }"
+        "{ h help   | false       | print help message }");
+
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Usage : bgfg_segm [options]" << endl;
+        cout << "Avaible options:" << endl;
+        cmd.printMessage();
+        return 0;
+    }
+
+    bool useCamera = cmd.get<bool>("camera");
+    string file = cmd.get<string>("file");
+    string method = cmd.get<string>("method");
+
+    if (method != "mog" && method != "mog2")
+    {
+        cerr << "Incorrect method" << endl;
+        return -1;
+    }
+
+    int m = method == "mog" ? M_MOG : M_MOG2;
+
+    VideoCapture cap;
+
+    if (useCamera)
+        cap.open(0);
+    else
+        cap.open(file);
+
+    if (!cap.isOpened())
+    {
+        cerr << "can not open camera or video file" << endl;
+        return -1;
+    }
+
+    std::vector<cv::ocl::Info>info;
+    cv::ocl::getDevice(info);
+
+    Mat frame;
+    cap >> frame;
+
+    oclMat d_frame(frame);
+
+    cv::ocl::MOG mog;
+    cv::ocl::MOG2 mog2;
+
+    oclMat d_fgmask;
+    oclMat d_fgimg;
+    oclMat d_bgimg;
+
+    d_fgimg.create(d_frame.size(), d_frame.type());
+
+    Mat fgmask;
+    Mat fgimg;
+    Mat bgimg;
+
+    switch (m)
+    {
+    case M_MOG:
+        mog(d_frame, d_fgmask, 0.01f);
+        break;
+
+    case M_MOG2:
+        mog2(d_frame, d_fgmask);
+        break;
+    }
+
+    for(;;)
+    {
+        cap >> frame;
+        if (frame.empty())
+            break;
+        d_frame.upload(frame);
+
+        int64 start = cv::getTickCount();
+
+        //update the model
+        switch (m)
+        {
+        case M_MOG:
+            mog(d_frame, d_fgmask, 0.01f);
+            mog.getBackgroundImage(d_bgimg);
+            break;
+
+        case M_MOG2:
+            mog2(d_frame, d_fgmask);
+            mog2.getBackgroundImage(d_bgimg);
+            break;
+        }
+
+        double fps = cv::getTickFrequency() / (cv::getTickCount() - start);
+        std::cout << "FPS : " << fps << std::endl;
+
+        d_fgimg.setTo(Scalar::all(0));
+        d_frame.copyTo(d_fgimg, d_fgmask);
+
+        d_fgmask.download(fgmask);
+        d_fgimg.download(fgimg);
+        if (!d_bgimg.empty())
+            d_bgimg.download(bgimg);
+
+        imshow("image", frame);
+        imshow("foreground mask", fgmask);
+        imshow("foreground image", fgimg);
+        if (!bgimg.empty())
+            imshow("mean background image", bgimg);
+
+        int key = waitKey(30);
+        if (key == 27)
+            break;
+    }
+
+    return 0;
+}
diff --git a/samples/ocl/clahe.cpp b/samples/ocl/clahe.cpp
new file mode 100644
index 0000000000..1ae2c8655d
--- /dev/null
+++ b/samples/ocl/clahe.cpp
@@ -0,0 +1,113 @@
+#include <iostream>
+#include "opencv2/core/core.hpp"
+#include "opencv2/core/utility.hpp"
+#include "opencv2/imgproc/imgproc.hpp"
+#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/ocl/ocl.hpp"
+using namespace cv;
+using namespace std;
+
+Ptr<CLAHE> pFilter;
+int tilesize;
+int cliplimit;
+string outfile;
+
+static void TSize_Callback(int pos)
+{
+    if(pos==0)
+    {
+        pFilter->setTilesGridSize(Size(1,1));
+    }
+    pFilter->setTilesGridSize(Size(tilesize,tilesize));
+}
+
+static void Clip_Callback(int)
+{
+    pFilter->setClipLimit(cliplimit);
+}
+
+int main(int argc, char** argv)
+{
+    const char* keys =
+        "{ i input   |                    | specify input image }"
+        "{ c camera  |    0               | specify camera id   }"
+        "{ s use_cpu |    false           | use cpu algorithm   }"
+        "{ o output  | clahe_output.jpg   | specify output save path}";
+
+    CommandLineParser cmd(argc, argv, keys);
+    string infile = cmd.get<string>("i");
+    outfile = cmd.get<string>("o");
+    int camid = cmd.get<int>("c");
+    bool use_cpu = cmd.get<bool>("s");
+    VideoCapture capture;
+    bool running = true;
+
+    namedWindow("CLAHE");
+    createTrackbar("Tile Size", "CLAHE", &tilesize, 32, (TrackbarCallback)TSize_Callback);
+    createTrackbar("Clip Limit", "CLAHE", &cliplimit, 20, (TrackbarCallback)Clip_Callback);
+
+    vector<ocl::Info> info;
+    CV_Assert(ocl::getDevice(info));
+
+    Mat frame, outframe;
+    ocl::oclMat d_outframe;
+
+    int cur_clip;
+    Size cur_tilesize;
+    if(use_cpu)
+    {
+        pFilter = createCLAHE();
+    }
+    else
+    {
+        pFilter = ocl::createCLAHE();
+    }
+    cur_clip = (int)pFilter->getClipLimit();
+    cur_tilesize = pFilter->getTilesGridSize();
+    setTrackbarPos("Tile Size", "CLAHE", cur_tilesize.width);
+    setTrackbarPos("Clip Limit", "CLAHE", cur_clip);
+    if(infile != "")
+    {
+        frame = imread(infile);
+        if(frame.empty())
+        {
+            cout << "error read image: " << infile << endl;
+            return -1;
+        }
+    }
+    else
+    {
+        capture.open(camid);
+    }
+    cout << "\nControls:\n"
+         << "\to - save output image\n"
+         << "\tESC - exit\n";
+    while(running)
+    {
+        if(capture.isOpened())
+            capture.read(frame);
+        else
+            frame = imread(infile);
+        if(frame.empty())
+        {
+            continue;
+        }
+        if(use_cpu)
+        {
+            cvtColor(frame, frame, COLOR_BGR2GRAY);
+            pFilter->apply(frame, outframe);
+        }
+        else
+        {
+            ocl::oclMat d_frame(frame);
+            ocl::cvtColor(d_frame, d_outframe, COLOR_BGR2GRAY);
+            pFilter->apply(d_outframe, d_outframe);
+            d_outframe.download(outframe);
+        }
+        imshow("CLAHE", outframe);
+        char key = (char)waitKey(3);
+        if(key == 'o') imwrite(outfile, outframe);
+        else if(key == 27) running = false;
+    }
+    return 0;
+}
diff --git a/samples/ocl/facedetect.cpp b/samples/ocl/facedetect.cpp
index 08f45f8853..eef4c9be9f 100644
--- a/samples/ocl/facedetect.cpp
+++ b/samples/ocl/facedetect.cpp
@@ -1,86 +1,105 @@
-//This sample is inherited from facedetect.cpp in smaple/c
-
 #include "opencv2/objdetect/objdetect.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/imgproc/imgproc.hpp"
 #include "opencv2/ocl/ocl.hpp"
+
+#include "opencv2/highgui/highgui_c.h"
+
 #include <iostream>
 #include <stdio.h>
 
+
 using namespace std;
 using namespace cv;
+#define LOOP_NUM 10
 
-static void help()
+const static Scalar colors[] =  { CV_RGB(0,0,255),
+                                  CV_RGB(0,128,255),
+                                  CV_RGB(0,255,255),
+                                  CV_RGB(0,255,0),
+                                  CV_RGB(255,128,0),
+                                  CV_RGB(255,255,0),
+                                  CV_RGB(255,0,0),
+                                  CV_RGB(255,0,255)
+                                } ;
+
+
+int64 work_begin = 0;
+int64 work_end = 0;
+string outputName;
+
+static void workBegin()
 {
-    cout << "\nThis program demonstrates the cascade recognizer.\n"
-        "This classifier can recognize many ~rigid objects, it's most known use is for faces.\n"
-        "Usage:\n"
-        "./facedetect [--cascade=<cascade_path> this is the primary trained classifier such as frontal face]\n"
-        "   [--scale=<image scale greater or equal to 1, try 1.3 for example>\n"
-        "   [filename|camera_index]\n\n"
-        "see facedetect.cmd for one call:\n"
-        "./facedetect --cascade=\"../../data/haarcascades/haarcascade_frontalface_alt.xml\" --scale=1.3 \n"
-        "Hit any key to quit.\n"
-        "Using OpenCV version " << CV_VERSION << "\n" << endl;
+    work_begin = getTickCount();
+}
+static void workEnd()
+{
+    work_end += (getTickCount() - work_begin);
 }
-struct getRect { Rect operator ()(const CvAvgComp& e) const { return e.rect; } };
-void detectAndDraw( Mat& img,
-    cv::ocl::OclCascadeClassifier& cascade, CascadeClassifier& nestedCascade,
-    double scale);
 
-string cascadeName = "../../../data/haarcascades/haarcascade_frontalface_alt.xml";
+static double getTime()
+{
+    return work_end /((double)cvGetTickFrequency() * 1000.);
+}
+
+void detect( Mat& img, vector<Rect>& faces,
+             ocl::OclCascadeClassifier& cascade,
+             double scale, bool calTime);
+
+
+void detectCPU( Mat& img, vector<Rect>& faces,
+                CascadeClassifier& cascade,
+                double scale, bool calTime);
+
+void Draw(Mat& img, vector<Rect>& faces, double scale);
+
+
+// This function test if gpu_rst matches cpu_rst.
+// If the two vectors are not equal, it will return the difference in vector size
+// Else if will return (total diff of each cpu and gpu rects covered pixels)/(total cpu rects covered pixels)
+double checkRectSimilarity(Size sz, vector<Rect>& cpu_rst, vector<Rect>& gpu_rst);
+
 
 int main( int argc, const char** argv )
 {
+    const char* keys =
+        "{ h  help       | false       | print help message }"
+        "{ i  input      |             | specify input image }"
+        "{ t  template   | haarcascade_frontalface_alt.xml |"
+        " specify template file path }"
+        "{ c  scale      |   1.0       | scale image }"
+        "{ s  use_cpu    | false       | use cpu or gpu to process the image }"
+        "{ o  output     | facedetect_output.jpg  |"
+        " specify output image save path(only works when input is images) }";
+
+    CommandLineParser cmd(argc, argv, keys);
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Avaible options:" << endl;
+        return 0;
+    }
     CvCapture* capture = 0;
     Mat frame, frameCopy, image;
-    const string scaleOpt = "--scale=";
-    size_t scaleOptLen = scaleOpt.length();
-    const string cascadeOpt = "--cascade=";
-    size_t cascadeOptLen = cascadeOpt.length();
-    string inputName;
 
-    help();
-    cv::ocl::OclCascadeClassifier cascade;
-    CascadeClassifier  nestedCascade;
-    double scale = 1;
+    bool useCPU = cmd.get<bool>("s");
+    string inputName = cmd.get<string>("i");
+    outputName = cmd.get<string>("o");
+    string cascadeName = cmd.get<string>("t");
+    double scale = cmd.get<double>("c");
+    ocl::OclCascadeClassifier cascade;
+    CascadeClassifier  cpu_cascade;
 
-    for( int i = 1; i < argc; i++ )
-    {
-        cout << "Processing " << i << " " <<  argv[i] << endl;
-        if( cascadeOpt.compare( 0, cascadeOptLen, argv[i], cascadeOptLen ) == 0 )
-        {
-            cascadeName.assign( argv[i] + cascadeOptLen );
-            cout << "  from which we have cascadeName= " << cascadeName << endl;
-        }
-        else if( scaleOpt.compare( 0, scaleOptLen, argv[i], scaleOptLen ) == 0 )
-        {
-            if( !sscanf( argv[i] + scaleOpt.length(), "%lf", &scale ) || scale < 1 )
-                scale = 1;
-            cout << " from which we read scale = " << scale << endl;
-        }
-        else if( argv[i][0] == '-' )
-        {
-            cerr << "WARNING: Unknown option %s" << argv[i] << endl;
-        }
-        else
-            inputName.assign( argv[i] );
-    }
-
-    if( !cascade.load( cascadeName ) )
+    if( !cascade.load( cascadeName ) || !cpu_cascade.load(cascadeName) )
     {
         cerr << "ERROR: Could not load classifier cascade" << endl;
-        cerr << "Usage: facedetect [--cascade=<cascade_path>]\n"
-            "   [--scale[=<image scale>\n"
-            "   [filename|camera_index]\n" << endl ;
         return -1;
     }
 
-    if( inputName.empty() || (isdigit(inputName.c_str()[0]) && inputName.c_str()[1] == '\0') )
+    if( inputName.empty() )
     {
-        capture = cvCaptureFromCAM( inputName.empty() ? 0 : inputName.c_str()[0] - '0' );
-        int c = inputName.empty() ? 0 : inputName.c_str()[0] - '0' ;
-        if(!capture) cout << "Capture from CAM " <<  c << " didn't work" << endl;
+        capture = cvCaptureFromCAM(0);
+        if(!capture)
+            cout << "Capture from CAM 0 didn't work" << endl;
     }
     else if( inputName.size() )
     {
@@ -88,135 +107,144 @@ int main( int argc, const char** argv )
         if( image.empty() )
         {
             capture = cvCaptureFromAVI( inputName.c_str() );
-            if(!capture) cout << "Capture from AVI didn't work" << endl;
+            if(!capture)
+                cout << "Capture from AVI didn't work" << endl;
+            return -1;
         }
     }
     else
     {
         image = imread( "lena.jpg", 1 );
-        if(image.empty()) cout << "Couldn't read lena.jpg" << endl;
+        if(image.empty())
+            cout << "Couldn't read lena.jpg" << endl;
+        return -1;
     }
 
+
     cvNamedWindow( "result", 1 );
-    std::vector<cv::ocl::Info> oclinfo;
-    int devnums = cv::ocl::getDevice(oclinfo);
-    if(devnums<1)
+    vector<ocl::Info> oclinfo;
+    int devnums = ocl::getDevice(oclinfo);
+    if( devnums < 1 )
     {
         std::cout << "no device found\n";
         return -1;
     }
     //if you want to use undefault device, set it here
     //setDevice(oclinfo[0]);
-    //setBinpath(CLBINPATH);
+    ocl::setBinpath("./");
     if( capture )
     {
         cout << "In capture ..." << endl;
         for(;;)
         {
             IplImage* iplImg = cvQueryFrame( capture );
-            frame = iplImg;
+            frame = cv::cvarrToMat(iplImg);
+            vector<Rect> faces;
             if( frame.empty() )
                 break;
             if( iplImg->origin == IPL_ORIGIN_TL )
                 frame.copyTo( frameCopy );
             else
                 flip( frame, frameCopy, 0 );
-
-            detectAndDraw( frameCopy, cascade, nestedCascade, scale );
-
+            if(useCPU)
+            {
+                detectCPU(frameCopy, faces, cpu_cascade, scale, false);
+            }
+            else
+            {
+                detect(frameCopy, faces, cascade, scale, false);
+            }
+            Draw(frameCopy, faces, scale);
             if( waitKey( 10 ) >= 0 )
                 goto _cleanup_;
         }
 
+
         waitKey(0);
 
+
 _cleanup_:
         cvReleaseCapture( &capture );
     }
     else
     {
         cout << "In image read" << endl;
-        if( !image.empty() )
+        vector<Rect> faces;
+        vector<Rect> ref_rst;
+        double accuracy = 0.;
+        for(int i = 0; i <= LOOP_NUM; i ++)
         {
-            detectAndDraw( image, cascade, nestedCascade, scale );
-            waitKey(0);
-        }
-        else if( !inputName.empty() )
-        {
-            /* assume it is a text file containing the
-            list of the image filenames to be processed - one per line */
-            FILE* f = fopen( inputName.c_str(), "rt" );
-            if( f )
+            cout << "loop" << i << endl;
+            if(useCPU)
             {
-                char buf[1000+1];
-                while( fgets( buf, 1000, f ) )
+                detectCPU(image, faces, cpu_cascade, scale, i==0?false:true);
+            }
+            else
+            {
+                detect(image, faces, cascade, scale, i==0?false:true);
+                if(i == 0)
                 {
-                    int len = (int)strlen(buf), c;
-                    while( len > 0 && isspace(buf[len-1]) )
-                        len--;
-                    buf[len] = '\0';
-                    cout << "file " << buf << endl;
-                    image = imread( buf, 1 );
-                    if( !image.empty() )
-                    {
-                        detectAndDraw( image, cascade, nestedCascade, scale );
-                        c = waitKey(0);
-                        if( c == 27 || c == 'q' || c == 'Q' )
-                            break;
-                    }
-                    else
-                    {
-                        cerr << "Aw snap, couldn't read image " << buf << endl;
-                    }
+                    detectCPU(image, ref_rst, cpu_cascade, scale, false);
+                    accuracy = checkRectSimilarity(image.size(), ref_rst, faces);
                 }
-                fclose(f);
+            }
+            if (i == LOOP_NUM)
+            {
+                if (useCPU)
+                    cout << "average CPU time (noCamera) : ";
+                else
+                    cout << "average GPU time (noCamera) : ";
+                cout << getTime() / LOOP_NUM << " ms" << endl;
+                cout << "accuracy value: " << accuracy <<endl;
             }
         }
+        Draw(image, faces, scale);
+        waitKey(0);
     }
 
     cvDestroyWindow("result");
-
     return 0;
 }
 
-void detectAndDraw( Mat& img,
-    cv::ocl::OclCascadeClassifier& cascade, CascadeClassifier&,
-    double scale)
+void detect( Mat& img, vector<Rect>& faces,
+             ocl::OclCascadeClassifier& cascade,
+             double scale, bool calTime)
+{
+    ocl::oclMat image(img);
+    ocl::oclMat gray, smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
+    if(calTime) workBegin();
+    ocl::cvtColor( image, gray, COLOR_BGR2GRAY );
+    ocl::resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
+    ocl::equalizeHist( smallImg, smallImg );
+
+    cascade.detectMultiScale( smallImg, faces, 1.1,
+                              3, 0
+                              |CASCADE_SCALE_IMAGE
+                              , Size(30,30), Size(0, 0) );
+    if(calTime) workEnd();
+}
+
+void detectCPU( Mat& img, vector<Rect>& faces,
+                CascadeClassifier& cascade,
+                double scale, bool calTime)
+{
+    if(calTime) workBegin();
+    Mat cpu_gray, cpu_smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
+    cvtColor(img, cpu_gray, COLOR_BGR2GRAY);
+    resize(cpu_gray, cpu_smallImg, cpu_smallImg.size(), 0, 0, INTER_LINEAR);
+    equalizeHist(cpu_smallImg, cpu_smallImg);
+    cascade.detectMultiScale(cpu_smallImg, faces, 1.1,
+                             3, 0 | CASCADE_SCALE_IMAGE,
+                             Size(30, 30), Size(0, 0));
+    if(calTime) workEnd();
+}
+
+
+void Draw(Mat& img, vector<Rect>& faces, double scale)
 {
     int i = 0;
-    double t = 0;
-    vector<Rect> faces;
-    const static Scalar colors[] =  { CV_RGB(0,0,255),
-        CV_RGB(0,128,255),
-        CV_RGB(0,255,255),
-        CV_RGB(0,255,0),
-        CV_RGB(255,128,0),
-        CV_RGB(255,255,0),
-        CV_RGB(255,0,0),
-        CV_RGB(255,0,255)} ;
-    cv::ocl::oclMat image(img);
-    cv::ocl::oclMat gray, smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
-
-    cv::ocl::cvtColor( image, gray, CV_BGR2GRAY );
-    cv::ocl::resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
-    cv::ocl::equalizeHist( smallImg, smallImg );
-
-    CvSeq* _objects;
-    MemStorage storage(cvCreateMemStorage(0));
-    t = (double)cvGetTickCount();
-    _objects = cascade.oclHaarDetectObjects( smallImg, storage, 1.1,
-        3, 0
-        |CV_HAAR_SCALE_IMAGE
-        , Size(30,30), Size(0, 0) );
-    vector<CvAvgComp> vecAvgComp;
-    Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
-    faces.resize(vecAvgComp.size());
-    std::transform(vecAvgComp.begin(), vecAvgComp.end(), faces.begin(), getRect());
-    t = (double)cvGetTickCount() - t;
-    printf( "detection time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
     for( vector<Rect>::const_iterator r = faces.begin(); r != faces.end(); r++, i++ )
     {
-        Mat smallImgROI;
         Point center;
         Scalar color = colors[i%8];
         int radius;
@@ -225,5 +253,58 @@ void detectAndDraw( Mat& img,
         radius = cvRound((r->width + r->height)*0.25*scale);
         circle( img, center, radius, color, 3, 8, 0 );
     }
-    cv::imshow( "result", img );
+    imwrite( outputName, img );
+    if(abs(scale-1.0)>.001)
+    {
+        resize(img, img, Size((int)(img.cols/scale), (int)(img.rows/scale)));
+    }
+    imshow( "result", img );
+
+}
+
+
+double checkRectSimilarity(Size sz, vector<Rect>& ob1, vector<Rect>& ob2)
+{
+    double final_test_result = 0.0;
+    size_t sz1 = ob1.size();
+    size_t sz2 = ob2.size();
+
+    if(sz1 != sz2)
+    {
+        return sz1 > sz2 ? (double)(sz1 - sz2) : (double)(sz2 - sz1);
+    }
+    else
+    {
+        if(sz1==0 && sz2==0)
+            return 0;
+        Mat cpu_result(sz, CV_8UC1);
+        cpu_result.setTo(0);
+
+        for(vector<Rect>::const_iterator r = ob1.begin(); r != ob1.end(); r++)
+        {
+            Mat cpu_result_roi(cpu_result, *r);
+            cpu_result_roi.setTo(1);
+            cpu_result.copyTo(cpu_result);
+        }
+        int cpu_area = countNonZero(cpu_result > 0);
+
+
+        Mat gpu_result(sz, CV_8UC1);
+        gpu_result.setTo(0);
+        for(vector<Rect>::const_iterator r2 = ob2.begin(); r2 != ob2.end(); r2++)
+        {
+            cv::Mat gpu_result_roi(gpu_result, *r2);
+            gpu_result_roi.setTo(1);
+            gpu_result.copyTo(gpu_result);
+        }
+
+        Mat result_;
+        multiply(cpu_result, gpu_result, result_);
+        int result = countNonZero(result_ > 0);
+        if(cpu_area!=0 && result!=0)
+            final_test_result = 1.0 - (double)result/(double)cpu_area;
+        else if(cpu_area==0 && result!=0)
+            final_test_result = -1;
+    }
+    return final_test_result;
 }
diff --git a/samples/ocl/hog.cpp b/samples/ocl/hog.cpp
index 76b6d2830e..12280b1f80 100644
--- a/samples/ocl/hog.cpp
+++ b/samples/ocl/hog.cpp
@@ -4,75 +4,46 @@
 #include <sstream>
 #include <iomanip>
 #include <stdexcept>
-#include "opencv2/ocl/ocl.hpp"
-#include "opencv2/highgui/highgui.hpp"
+#include <opencv2/core/utility.hpp>
+#include "opencv2/ocl.hpp"
+#include "opencv2/highgui.hpp"
 
 using namespace std;
 using namespace cv;
 
-bool help_showed = false;
-
-class Args
-{
-public:
-    Args();
-    static Args read(int argc, char** argv);
-
-    string src;
-    bool src_is_video;
-    bool src_is_camera;
-    int camera_id;
-
-    bool write_video;
-    string dst_video;
-    double dst_video_fps;
-
-    bool make_gray;
-
-    bool resize_src;
-    int width, height;
-
-    double scale;
-    int nlevels;
-    int gr_threshold;
-
-    double hit_threshold;
-    bool hit_threshold_auto;
-
-    int win_width;
-    int win_stride_width, win_stride_height;
-
-    bool gamma_corr;
-};
-
-
 class App
 {
 public:
-    App(const Args& s);
+    App(CommandLineParser& cmd);
     void run();
-
     void handleKey(char key);
-
     void hogWorkBegin();
     void hogWorkEnd();
     string hogWorkFps() const;
-
     void workBegin();
     void workEnd();
     string workFps() const;
-
     string message() const;
 
+
+// This function test if gpu_rst matches cpu_rst.
+// If the two vectors are not equal, it will return the difference in vector size
+// Else if will return
+// (total diff of each cpu and gpu rects covered pixels)/(total cpu rects covered pixels)
+    double checkRectSimilarity(Size sz,
+                               std::vector<Rect>& cpu_rst,
+                               std::vector<Rect>& gpu_rst);
 private:
     App operator=(App&);
 
-    Args args;
+    //Args args;
     bool running;
-
     bool use_gpu;
     bool make_gray;
     double scale;
+    double resize_scale;
+    int win_width;
+    int win_stride_width, win_stride_height;
     int gr_threshold;
     int nlevels;
     double hit_threshold;
@@ -80,179 +51,112 @@ private:
 
     int64 hog_work_begin;
     double hog_work_fps;
-
     int64 work_begin;
     double work_fps;
-};
 
-static void printHelp()
-{
-    cout << "Histogram of Oriented Gradients descriptor and detector sample.\n"
-         << "\nUsage: hog_gpu\n"
-         << "  (<image>|--video <vide>|--camera <camera_id>) # frames source\n"
-         << "  [--make_gray <true/false>] # convert image to gray one or not\n"
-         << "  [--resize_src <true/false>] # do resize of the source image or not\n"
-         << "  [--width <int>] # resized image width\n"
-         << "  [--height <int>] # resized image height\n"
-         << "  [--hit_threshold <double>] # classifying plane distance threshold (0.0 usually)\n"
-         << "  [--scale <double>] # HOG window scale factor\n"
-         << "  [--nlevels <int>] # max number of HOG window scales\n"
-         << "  [--win_width <int>] # width of the window (48 or 64)\n"
-         << "  [--win_stride_width <int>] # distance by OX axis between neighbour wins\n"
-         << "  [--win_stride_height <int>] # distance by OY axis between neighbour wins\n"
-         << "  [--gr_threshold <int>] # merging similar rects constant\n"
-         << "  [--gamma_correct <int>] # do gamma correction or not\n"
-         << "  [--write_video <bool>] # write video or not\n"
-         << "  [--dst_video <path>] # output video path\n"
-         << "  [--dst_video_fps <double>] # output video fps\n";
-    help_showed = true;
-}
+    string img_source;
+    string vdo_source;
+    string output;
+    int camera_id;
+    bool write_once;
+};
 
 int main(int argc, char** argv)
 {
+    const char* keys =
+        "{ h |  help    | false          | print help message }"
+        "{ i |  input   |                | specify input image}"
+        "{ c | camera   | -1             | enable camera capturing }"
+        "{ v | video    |                | use video as input }"
+        "{ g |  gray    | false          | convert image to gray one or not}"
+        "{ s |  scale   | 1.0            | resize the image before detect}"
+        "{ l |larger_win| false          | use 64x128 window}"
+        "{ o |  output  |                | specify output path when input is images}";
+    CommandLineParser cmd(argc, argv, keys);
+    App app(cmd);
     try
     {
-        if (argc < 2)
-            printHelp();
-        Args args = Args::read(argc, argv);
-        if (help_showed)
-            return -1;
-        App app(args);
         app.run();
     }
-    catch (const Exception& e) { return cout << "error: "  << e.what() << endl, 1; }
-    catch (const exception& e) { return cout << "error: "  << e.what() << endl, 1; }
-    catch(...) { return cout << "unknown exception" << endl, 1; }
+    catch (const Exception& e)
+    {
+        return cout << "error: "  << e.what() << endl, 1;
+    }
+    catch (const exception& e)
+    {
+        return cout << "error: "  << e.what() << endl, 1;
+    }
+    catch(...)
+    {
+        return cout << "unknown exception" << endl, 1;
+    }
     return 0;
 }
 
-
-Args::Args()
+App::App(CommandLineParser& cmd)
 {
-    src_is_video = false;
-    src_is_camera = false;
-    camera_id = 0;
-
-    write_video = false;
-    dst_video_fps = 24.;
-
-    make_gray = false;
-
-    resize_src = false;
-    width = 640;
-    height = 480;
-
-    scale = 1.05;
-    nlevels = 13;
-    gr_threshold = 8;
-    hit_threshold = 1.4;
-    hit_threshold_auto = true;
-
-    win_width = 48;
-    win_stride_width = 8;
-    win_stride_height = 8;
-
-    gamma_corr = true;
-}
-
-
-Args Args::read(int argc, char** argv)
-{
-    Args args;
-    for (int i = 1; i < argc; i++)
-    {
-        if (string(argv[i]) == "--make_gray") args.make_gray = (string(argv[++i]) == "true");
-        else if (string(argv[i]) == "--resize_src") args.resize_src = (string(argv[++i]) == "true");
-        else if (string(argv[i]) == "--width") args.width = atoi(argv[++i]);
-        else if (string(argv[i]) == "--height") args.height = atoi(argv[++i]);
-        else if (string(argv[i]) == "--hit_threshold")
-        {
-            args.hit_threshold = atof(argv[++i]);
-            args.hit_threshold_auto = false;
-        }
-        else if (string(argv[i]) == "--scale") args.scale = atof(argv[++i]);
-        else if (string(argv[i]) == "--nlevels") args.nlevels = atoi(argv[++i]);
-        else if (string(argv[i]) == "--win_width") args.win_width = atoi(argv[++i]);
-        else if (string(argv[i]) == "--win_stride_width") args.win_stride_width = atoi(argv[++i]);
-        else if (string(argv[i]) == "--win_stride_height") args.win_stride_height = atoi(argv[++i]);
-        else if (string(argv[i]) == "--gr_threshold") args.gr_threshold = atoi(argv[++i]);
-        else if (string(argv[i]) == "--gamma_correct") args.gamma_corr = (string(argv[++i]) == "true");
-        else if (string(argv[i]) == "--write_video") args.write_video = (string(argv[++i]) == "true");
-        else if (string(argv[i]) == "--dst_video") args.dst_video = argv[++i];
-        else if (string(argv[i]) == "--dst_video_fps") args.dst_video_fps = atof(argv[++i]);
-        else if (string(argv[i]) == "--help") printHelp();
-        else if (string(argv[i]) == "--video") { args.src = argv[++i]; args.src_is_video = true; }
-        else if (string(argv[i]) == "--camera") { args.camera_id = atoi(argv[++i]); args.src_is_camera = true; }
-        else if (args.src.empty()) args.src = argv[i];
-        else throw runtime_error((string("unknown key: ") + argv[i]));
-    }
-    return args;
-}
-
-
-App::App(const Args& s)
-{
-    args = s;
     cout << "\nControls:\n"
          << "\tESC - exit\n"
          << "\tm - change mode GPU <-> CPU\n"
          << "\tg - convert image to gray or not\n"
+         << "\to - save output image once, or switch on/off video save\n"
          << "\t1/q - increase/decrease HOG scale\n"
          << "\t2/w - increase/decrease levels count\n"
          << "\t3/e - increase/decrease HOG group threshold\n"
          << "\t4/r - increase/decrease hit threshold\n"
          << endl;
 
+
     use_gpu = true;
-    make_gray = args.make_gray;
-    scale = args.scale;
-    gr_threshold = args.gr_threshold;
-    nlevels = args.nlevels;
+    make_gray = cmd.get<bool>("g");
+    resize_scale = cmd.get<double>("s");
+    win_width = cmd.get<bool>("l") == true ? 64 : 48;
+    vdo_source = cmd.get<string>("v");
+    img_source = cmd.get<string>("i");
+    output = cmd.get<string>("o");
+    camera_id = cmd.get<int>("c");
 
-    if (args.hit_threshold_auto)
-        args.hit_threshold = args.win_width == 48 ? 1.4 : 0.;
-    hit_threshold = args.hit_threshold;
+    win_stride_width = 8;
+    win_stride_height = 8;
+    gr_threshold = 8;
+    nlevels = 13;
+    hit_threshold = win_width == 48 ? 1.4 : 0.;
+    scale = 1.05;
+    gamma_corr = true;
+    write_once = false;
 
-    gamma_corr = args.gamma_corr;
-
-    if (args.win_width != 64 && args.win_width != 48)
-        args.win_width = 64;
-
-    cout << "Scale: " << scale << endl;
-    if (args.resize_src)
-        cout << "Resized source: (" << args.width << ", " << args.height << ")\n";
     cout << "Group threshold: " << gr_threshold << endl;
     cout << "Levels number: " << nlevels << endl;
-    cout << "Win width: " << args.win_width << endl;
-    cout << "Win stride: (" << args.win_stride_width << ", " << args.win_stride_height << ")\n";
+    cout << "Win width: " << win_width << endl;
+    cout << "Win stride: (" << win_stride_width << ", " << win_stride_height << ")\n";
     cout << "Hit threshold: " << hit_threshold << endl;
     cout << "Gamma correction: " << gamma_corr << endl;
     cout << endl;
 }
 
-
 void App::run()
 {
-    std::vector<ocl::Info> oclinfo;
+    vector<ocl::Info> oclinfo;
     ocl::getDevice(oclinfo);
     running = true;
-    cv::VideoWriter video_writer;
+    VideoWriter video_writer;
 
-    Size win_size(args.win_width, args.win_width * 2); //(64, 128) or (48, 96)
-    Size win_stride(args.win_stride_width, args.win_stride_height);
+    Size win_size(win_width, win_width * 2);
+    Size win_stride(win_stride_width, win_stride_height);
 
     // Create HOG descriptors and detectors here
     vector<float> detector;
     if (win_size == Size(64, 128))
-        detector = cv::ocl::HOGDescriptor::getPeopleDetector64x128();
+        detector = ocl::HOGDescriptor::getPeopleDetector64x128();
     else
-        detector = cv::ocl::HOGDescriptor::getPeopleDetector48x96();
+        detector = ocl::HOGDescriptor::getPeopleDetector48x96();
 
-    cv::ocl::HOGDescriptor gpu_hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9,
-                                   cv::ocl::HOGDescriptor::DEFAULT_WIN_SIGMA, 0.2, gamma_corr,
-                                   cv::ocl::HOGDescriptor::DEFAULT_NLEVELS);
-    cv::HOGDescriptor cpu_hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9, 1, -1,
-                              HOGDescriptor::L2Hys, 0.2, gamma_corr, cv::HOGDescriptor::DEFAULT_NLEVELS);
+
+    ocl::HOGDescriptor gpu_hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9,
+                               ocl::HOGDescriptor::DEFAULT_WIN_SIGMA, 0.2, gamma_corr,
+                               ocl::HOGDescriptor::DEFAULT_NLEVELS);
+    HOGDescriptor cpu_hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9, 1, -1,
+                          HOGDescriptor::L2Hys, 0.2, gamma_corr, cv::HOGDescriptor::DEFAULT_NLEVELS);
     gpu_hog.setSVMDetector(detector);
     cpu_hog.setSVMDetector(detector);
 
@@ -261,52 +165,55 @@ void App::run()
         VideoCapture vc;
         Mat frame;
 
-        if (args.src_is_video)
+        if (vdo_source!="")
         {
-            vc.open(args.src.c_str());
+            vc.open(vdo_source.c_str());
             if (!vc.isOpened())
-                throw runtime_error(string("can't open video file: " + args.src));
+                throw runtime_error(string("can't open video file: " + vdo_source));
             vc >> frame;
         }
-        else if (args.src_is_camera)
+        else if (camera_id != -1)
         {
-            vc.open(args.camera_id);
+            vc.open(camera_id);
             if (!vc.isOpened())
             {
                 stringstream msg;
-                msg << "can't open camera: " << args.camera_id;
+                msg << "can't open camera: " << camera_id;
                 throw runtime_error(msg.str());
             }
             vc >> frame;
         }
         else
         {
-            frame = imread(args.src);
+            frame = imread(img_source);
             if (frame.empty())
-                throw runtime_error(string("can't open image file: " + args.src));
+                throw runtime_error(string("can't open image file: " + img_source));
         }
 
         Mat img_aux, img, img_to_show;
         ocl::oclMat gpu_img;
 
         // Iterate over all frames
+        bool verify = false;
         while (running && !frame.empty())
         {
             workBegin();
 
             // Change format of the image
-            if (make_gray) cvtColor(frame, img_aux, CV_BGR2GRAY);
-            else if (use_gpu) cvtColor(frame, img_aux, CV_BGR2BGRA);
+            if (make_gray) cvtColor(frame, img_aux, COLOR_BGR2GRAY);
+            else if (use_gpu) cvtColor(frame, img_aux, COLOR_BGR2BGRA);
             else frame.copyTo(img_aux);
 
             // Resize image
-            if (args.resize_src) resize(img_aux, img, Size(args.width, args.height));
+            if (abs(scale-1.0)>0.001)
+            {
+                Size sz((int)((double)img_aux.cols/resize_scale), (int)((double)img_aux.rows/resize_scale));
+                resize(img_aux, img, sz);
+            }
             else img = img_aux;
             img_to_show = img;
-
             gpu_hog.nlevels = nlevels;
             cpu_hog.nlevels = nlevels;
-
             vector<Rect> found;
 
             // Perform HOG classification
@@ -316,16 +223,28 @@ void App::run()
                 gpu_img.upload(img);
                 gpu_hog.detectMultiScale(gpu_img, found, hit_threshold, win_stride,
                                          Size(0, 0), scale, gr_threshold);
+                if (!verify)
+                {
+                    // verify if GPU output same objects with CPU at 1st run
+                    verify = true;
+                    vector<Rect> ref_rst;
+                    cvtColor(img, img, COLOR_BGRA2BGR);
+                    cpu_hog.detectMultiScale(img, ref_rst, hit_threshold, win_stride,
+                                             Size(0, 0), scale, gr_threshold-2);
+                    double accuracy = checkRectSimilarity(img.size(), ref_rst, found);
+                    cout << "\naccuracy value: " << accuracy << endl;
+                }
             }
             else cpu_hog.detectMultiScale(img, found, hit_threshold, win_stride,
-                                          Size(0, 0), scale, gr_threshold);
+                                              Size(0, 0), scale, gr_threshold);
             hogWorkEnd();
 
+
             // Draw positive classified windows
             for (size_t i = 0; i < found.size(); i++)
             {
                 Rect r = found[i];
-                rectangle(img_to_show, r.tl(), r.br(), CV_RGB(0, 255, 0), 3);
+                rectangle(img_to_show, r.tl(), r.br(), Scalar(0, 255, 0), 3);
             }
 
             if (use_gpu)
@@ -335,25 +254,32 @@ void App::run()
             putText(img_to_show, "FPS (HOG only): " + hogWorkFps(), Point(5, 65), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
             putText(img_to_show, "FPS (total): " + workFps(), Point(5, 105), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
             imshow("opencv_gpu_hog", img_to_show);
-
-            if (args.src_is_video || args.src_is_camera) vc >> frame;
+            if (vdo_source!="" || camera_id!=-1) vc >> frame;
 
             workEnd();
 
-            if (args.write_video)
+            if (output!="" && write_once)
             {
-                if (!video_writer.isOpened())
+                if (img_source!="")     // wirte image
                 {
-                    video_writer.open(args.dst_video, CV_FOURCC('x','v','i','d'), args.dst_video_fps,
-                                      img_to_show.size(), true);
-                    if (!video_writer.isOpened())
-                        throw std::runtime_error("can't create video writer");
+                    write_once = false;
+                    imwrite(output, img_to_show);
                 }
+                else                    //write video
+                {
+                    if (!video_writer.isOpened())
+                    {
+                        video_writer.open(output, VideoWriter::fourcc('x','v','i','d'), 24,
+                                          img_to_show.size(), true);
+                        if (!video_writer.isOpened())
+                            throw std::runtime_error("can't create video writer");
+                    }
 
-                if (make_gray) cvtColor(img_to_show, img, CV_GRAY2BGR);
-                else cvtColor(img_to_show, img, CV_BGRA2BGR);
+                    if (make_gray) cvtColor(img_to_show, img, COLOR_GRAY2BGR);
+                    else cvtColor(img_to_show, img, COLOR_BGRA2BGR);
 
-                video_writer << img;
+                    video_writer << img;
+                }
             }
 
             handleKey((char)waitKey(3));
@@ -361,7 +287,6 @@ void App::run()
     }
 }
 
-
 void App::handleKey(char key)
 {
     switch (key)
@@ -420,11 +345,18 @@ void App::handleKey(char key)
         gamma_corr = !gamma_corr;
         cout << "Gamma correction: " << gamma_corr << endl;
         break;
+    case 'o':
+    case 'O':
+        write_once = !write_once;
+        break;
     }
 }
 
 
-inline void App::hogWorkBegin() { hog_work_begin = getTickCount(); }
+inline void App::hogWorkBegin()
+{
+    hog_work_begin = getTickCount();
+}
 
 inline void App::hogWorkEnd()
 {
@@ -440,8 +372,10 @@ inline string App::hogWorkFps() const
     return ss.str();
 }
 
-
-inline void App::workBegin() { work_begin = getTickCount(); }
+inline void App::workBegin()
+{
+    work_begin = getTickCount();
+}
 
 inline void App::workEnd()
 {
@@ -457,3 +391,52 @@ inline string App::workFps() const
     return ss.str();
 }
 
+
+double App::checkRectSimilarity(Size sz,
+                                std::vector<Rect>& ob1,
+                                std::vector<Rect>& ob2)
+{
+    double final_test_result = 0.0;
+    size_t sz1 = ob1.size();
+    size_t sz2 = ob2.size();
+
+    if(sz1 != sz2)
+    {
+        return sz1 > sz2 ? (double)(sz1 - sz2) : (double)(sz2 - sz1);
+    }
+    else
+    {
+        if(sz1==0 && sz2==0)
+            return 0;
+        cv::Mat cpu_result(sz, CV_8UC1);
+        cpu_result.setTo(0);
+
+
+        for(vector<Rect>::const_iterator r = ob1.begin(); r != ob1.end(); r++)
+        {
+            cv::Mat cpu_result_roi(cpu_result, *r);
+            cpu_result_roi.setTo(1);
+            cpu_result.copyTo(cpu_result);
+        }
+        int cpu_area = cv::countNonZero(cpu_result > 0);
+
+
+        cv::Mat gpu_result(sz, CV_8UC1);
+        gpu_result.setTo(0);
+        for(vector<Rect>::const_iterator r2 = ob2.begin(); r2 != ob2.end(); r2++)
+        {
+            cv::Mat gpu_result_roi(gpu_result, *r2);
+            gpu_result_roi.setTo(1);
+            gpu_result.copyTo(gpu_result);
+        }
+
+        cv::Mat result_;
+        multiply(cpu_result, gpu_result, result_);
+        int result = cv::countNonZero(result_ > 0);
+        if(cpu_area!=0 && result!=0)
+            final_test_result = 1.0 - (double)result/(double)cpu_area;
+        else if(cpu_area==0 && result!=0)
+            final_test_result = -1;
+    }
+    return final_test_result;
+}
diff --git a/samples/ocl/performance.cpp b/samples/ocl/performance.cpp
deleted file mode 100644
index 262e4320b2..0000000000
--- a/samples/ocl/performance.cpp
+++ /dev/null
@@ -1,4396 +0,0 @@
-#include <iomanip>
-#include <stdexcept>
-#include <string>
-#include <iostream>
-#include <cstdio>
-#include <vector>
-#include <numeric>
-#include "opencv2/core/core.hpp"
-#include "opencv2/imgproc/imgproc.hpp"
-#include "opencv2/highgui/highgui.hpp"
-#include "opencv2/calib3d/calib3d.hpp"
-#include "opencv2/video/video.hpp"
-#include "opencv2/nonfree/nonfree.hpp"
-#include "opencv2/objdetect/objdetect.hpp"
-#include "opencv2/features2d/features2d.hpp"
-#define USE_OPENCL
-#ifdef USE_OPENCL
-#include "opencv2/ocl/ocl.hpp"
-#endif
-
-#define TAB "    "
-
-using namespace std;
-using namespace cv;
-
-// This program test most of the functions in ocl module and generate data metrix of x-factor in .csv files
-// All images needed in this test are in samples/gpu folder.
-// For haar template, haarcascade_frontalface_alt.xml shouold be in working directory
-
-void gen(Mat &mat, int rows, int cols, int type, Scalar low, Scalar high);
-string abspath(const string &relpath);
-int CV_CDECL cvErrorCallback(int, const char *, const char *, const char *, int, void *);
-typedef struct
-{
-    short x;
-    short y;
-} COOR;
-COOR do_meanShift(int x0, int y0, uchar *sptr, uchar *dptr, int sstep,
-                  cv::Size size, int sp, int sr, int maxIter, float eps, int *tab);
-void meanShiftProc_(const Mat &src_roi, Mat &dst_roi, Mat &dstCoor_roi,
-                    int sp, int sr, cv::TermCriteria crit);
-
-class Runnable
-{
-public:
-    explicit Runnable(const std::string &runname): name_(runname) {}
-    virtual ~Runnable() {}
-
-    const std::string &name() const
-    {
-        return name_;
-    }
-
-    virtual void run() = 0;
-
-private:
-    std::string name_;
-};
-
-class TestSystem
-{
-public:
-    static TestSystem &instance()
-    {
-        static TestSystem me;
-        return me;
-    }
-
-    void setWorkingDir(const std::string &val)
-    {
-        working_dir_ = val;
-    }
-    const std::string &workingDir() const
-    {
-        return working_dir_;
-    }
-
-    void setTestFilter(const std::string &val)
-    {
-        test_filter_ = val;
-    }
-    const std::string &testFilter() const
-    {
-        return test_filter_;
-    }
-
-    void setNumIters(int num_iters)
-    {
-        num_iters_ = num_iters;
-    }
-    void setGPUWarmupIters(int num_iters)
-    {
-        gpu_warmup_iters_ = num_iters;
-    }
-    void setCPUIters(int num_iters)
-    {
-        cpu_num_iters_ = num_iters;
-    }
-
-    void setTopThreshold(double top)
-    {
-        top_ = top;
-    }
-    void setBottomThreshold(double bottom)
-    {
-        bottom_ = bottom;
-    }
-
-    void addInit(Runnable *init)
-    {
-        inits_.push_back(init);
-    }
-    void addTest(Runnable *test)
-    {
-        tests_.push_back(test);
-    }
-    void run();
-
-    // It's public because OpenCV callback uses it
-    void printError(const std::string &msg);
-
-    std::stringstream &startNewSubtest()
-    {
-        finishCurrentSubtest();
-        return cur_subtest_description_;
-    }
-
-    bool stop() const
-    {
-        return cur_iter_idx_ >= num_iters_;
-    }
-
-    bool cpu_stop() const
-    {
-        return cur_iter_idx_ >= cpu_num_iters_;
-    }
-
-    bool warmupStop()
-    {
-        return cur_warmup_idx_++ >= gpu_warmup_iters_;
-    }
-
-    void warmupComplete()
-    {
-        cur_warmup_idx_ = 0;
-    }
-
-    void cpuOn()
-    {
-        cpu_started_ = cv::getTickCount();
-    }
-    void cpuOff()
-    {
-        int64 delta = cv::getTickCount() - cpu_started_;
-        cpu_times_.push_back(delta);
-        ++cur_iter_idx_;
-    }
-    void cpuComplete()
-    {
-        cpu_elapsed_ += meanTime(cpu_times_);
-        cur_subtest_is_empty_ = false;
-        cur_iter_idx_ = 0;
-    }
-
-    void gpuOn()
-    {
-        gpu_started_ = cv::getTickCount();
-    }
-    void gpuOff()
-    {
-        int64 delta = cv::getTickCount() - gpu_started_;
-        gpu_times_.push_back(delta);
-        ++cur_iter_idx_;
-    }
-    void gpuComplete()
-    {
-        gpu_elapsed_ += meanTime(gpu_times_);
-        cur_subtest_is_empty_ = false;
-        cur_iter_idx_ = 0;
-    }
-
-    void gpufullOn()
-    {
-        gpu_full_started_ = cv::getTickCount();
-    }
-    void gpufullOff()
-    {
-        int64 delta = cv::getTickCount() - gpu_full_started_;
-        gpu_full_times_.push_back(delta);
-        ++cur_iter_idx_;
-    }
-    void gpufullComplete()
-    {
-        gpu_full_elapsed_ += meanTime(gpu_full_times_);
-        cur_subtest_is_empty_ = false;
-        cur_iter_idx_ = 0;
-    }
-
-    bool isListMode() const
-    {
-        return is_list_mode_;
-    }
-    void setListMode(bool value)
-    {
-        is_list_mode_ = value;
-    }
-
-    void setRecordName(const std::string &name)
-    {
-        recordname_ = name;
-    }
-
-    void setCurrentTest(const std::string &name)
-    {
-        itname_ = name;
-        itname_changed_ = true;
-    }
-
-private:
-    TestSystem():
-        cur_subtest_is_empty_(true), cpu_elapsed_(0),
-        gpu_elapsed_(0), gpu_full_elapsed_(0), speedup_total_(0.0),
-        num_subtests_called_(0),
-        speedup_faster_count_(0), speedup_slower_count_(0), speedup_equal_count_(0),
-        speedup_full_faster_count_(0), speedup_full_slower_count_(0), speedup_full_equal_count_(0), is_list_mode_(false),
-        num_iters_(10), cpu_num_iters_(2),
-        gpu_warmup_iters_(1), cur_iter_idx_(0), cur_warmup_idx_(0),
-        record_(0), recordname_("performance"), itname_changed_(true)
-    {
-        cpu_times_.reserve(num_iters_);
-        gpu_times_.reserve(num_iters_);
-        gpu_full_times_.reserve(num_iters_);
-    }
-
-    void finishCurrentSubtest();
-    void resetCurrentSubtest()
-    {
-        cpu_elapsed_ = 0;
-        gpu_elapsed_ = 0;
-        gpu_full_elapsed_ = 0;
-        cur_subtest_description_.str("");
-        cur_subtest_is_empty_ = true;
-        cur_iter_idx_ = 0;
-        cpu_times_.clear();
-        gpu_times_.clear();
-        gpu_full_times_.clear();
-    }
-
-    double meanTime(const std::vector<int64> &samples);
-
-    void printHeading();
-    void printSummary();
-    void printMetrics(double cpu_time, double gpu_time, double gpu_full_time, double speedup, double fullspeedup);
-
-    void writeHeading();
-    void writeSummary();
-    void writeMetrics(double cpu_time, double gpu_time, double gpu_full_time,
-                      double speedup, double fullspeedup,
-                      double gpu_min, double gpu_max, double std_dev);
-
-    std::string working_dir_;
-    std::string test_filter_;
-
-    std::vector<Runnable *> inits_;
-    std::vector<Runnable *> tests_;
-
-    std::stringstream cur_subtest_description_;
-    bool cur_subtest_is_empty_;
-
-    int64 cpu_started_;
-    int64 gpu_started_;
-    int64 gpu_full_started_;
-    double cpu_elapsed_;
-    double gpu_elapsed_;
-    double gpu_full_elapsed_;
-
-    double speedup_total_;
-    double speedup_full_total_;
-    int num_subtests_called_;
-
-    int speedup_faster_count_;
-    int speedup_slower_count_;
-    int speedup_equal_count_;
-
-    int speedup_full_faster_count_;
-    int speedup_full_slower_count_;
-    int speedup_full_equal_count_;
-
-    bool is_list_mode_;
-
-    double top_;
-    double bottom_;
-
-    int num_iters_;
-    int cpu_num_iters_;		//there's no need to set cpu running same times with gpu
-    int gpu_warmup_iters_;	//gpu warm up times, default is 1
-    int cur_iter_idx_;
-    int cur_warmup_idx_;	//current gpu warm up times
-    std::vector<int64> cpu_times_;
-    std::vector<int64> gpu_times_;
-    std::vector<int64> gpu_full_times_;
-
-    FILE *record_;
-    std::string recordname_;
-    std::string itname_;
-    bool itname_changed_;
-};
-
-
-#define GLOBAL_INIT(name) \
-    struct name##_init: Runnable { \
-        name##_init(): Runnable(#name) { \
-            TestSystem::instance().addInit(this); \
-        } \
-        void run(); \
-    } name##_init_instance; \
-    void name##_init::run()
-
-
-#define TEST(name) \
-    struct name##_test: Runnable { \
-        name##_test(): Runnable(#name) { \
-            TestSystem::instance().addTest(this); \
-        } \
-        void run(); \
-    } name##_test_instance; \
-    void name##_test::run()
-
-#define SUBTEST TestSystem::instance().startNewSubtest()
-
-#define CPU_ON \
-    while (!TestSystem::instance().cpu_stop()) { \
-        TestSystem::instance().cpuOn()
-#define CPU_OFF \
-        TestSystem::instance().cpuOff(); \
-    } TestSystem::instance().cpuComplete()
-
-#define GPU_ON \
-    while (!TestSystem::instance().stop()) { \
-        TestSystem::instance().gpuOn()
-#define GPU_OFF \
-        TestSystem::instance().gpuOff(); \
-    } TestSystem::instance().gpuComplete()
-
-#define GPU_FULL_ON \
-    while (!TestSystem::instance().stop()) { \
-        TestSystem::instance().gpufullOn()
-#define GPU_FULL_OFF \
-        TestSystem::instance().gpufullOff(); \
-    } TestSystem::instance().gpufullComplete()
-
-#define WARMUP_ON \
-    while (!TestSystem::instance().warmupStop()) {
-#define WARMUP_OFF \
-    } TestSystem::instance().warmupComplete()
-
-void TestSystem::run()
-{
-    if (is_list_mode_)
-    {
-        for (vector<Runnable *>::iterator it = tests_.begin(); it != tests_.end(); ++it)
-        {
-            cout << (*it)->name() << endl;
-        }
-
-        return;
-    }
-
-    // Run test initializers
-    for (vector<Runnable *>::iterator it = inits_.begin(); it != inits_.end(); ++it)
-    {
-        if ((*it)->name().find(test_filter_, 0) != string::npos)
-        {
-            (*it)->run();
-        }
-    }
-
-    printHeading();
-    writeHeading();
-
-    // Run tests
-    for (vector<Runnable *>::iterator it = tests_.begin(); it != tests_.end(); ++it)
-    {
-        try
-        {
-            if ((*it)->name().find(test_filter_, 0) != string::npos)
-            {
-                cout << endl << (*it)->name() << ":\n";
-
-                setCurrentTest((*it)->name());
-                //fprintf(record_,"%s\n",(*it)->name().c_str());
-
-                (*it)->run();
-                finishCurrentSubtest();
-            }
-        }
-        catch (const Exception &)
-        {
-            // Message is printed via callback
-            resetCurrentSubtest();
-        }
-        catch (const runtime_error &e)
-        {
-            printError(e.what());
-            resetCurrentSubtest();
-        }
-    }
-
-#ifdef USE_OPENCL
-    printSummary();
-    writeSummary();
-#endif
-}
-
-
-void TestSystem::finishCurrentSubtest()
-{
-    if (cur_subtest_is_empty_)
-        // There is no need to print subtest statistics
-    {
-        return;
-    }
-
-    double cpu_time = cpu_elapsed_ / getTickFrequency() * 1000.0;
-    double gpu_time = gpu_elapsed_ / getTickFrequency() * 1000.0;
-    double gpu_full_time = gpu_full_elapsed_ / getTickFrequency() * 1000.0;
-
-    double speedup = static_cast<double>(cpu_elapsed_) / std::max(1.0, gpu_elapsed_);
-    speedup_total_ += speedup;
-
-    double fullspeedup = static_cast<double>(cpu_elapsed_) / std::max(1.0, gpu_full_elapsed_);
-    speedup_full_total_ += fullspeedup;
-
-    if (speedup > top_)
-    {
-        speedup_faster_count_++;
-    }
-    else if (speedup < bottom_)
-    {
-        speedup_slower_count_++;
-    }
-    else
-    {
-        speedup_equal_count_++;
-    }
-
-    if (fullspeedup > top_)
-    {
-        speedup_full_faster_count_++;
-    }
-    else if (fullspeedup < bottom_)
-    {
-        speedup_full_slower_count_++;
-    }
-    else
-    {
-        speedup_full_equal_count_++;
-    }
-
-    // compute min, max and
-    std::sort(gpu_times_.begin(), gpu_times_.end());
-    double gpu_min = gpu_times_.front() / getTickFrequency() * 1000.0;
-    double gpu_max = gpu_times_.back() / getTickFrequency() * 1000.0;
-    double deviation = 0;
-
-    if (gpu_times_.size() > 1)
-    {
-        double sum = 0;
-
-        for (size_t i = 0; i < gpu_times_.size(); i++)
-        {
-            int64 diff = gpu_times_[i] - static_cast<int64>(gpu_elapsed_);
-            double diff_time = diff * 1000 / getTickFrequency();
-            sum += diff_time * diff_time;
-        }
-
-        deviation = std::sqrt(sum / gpu_times_.size());
-    }
-
-    printMetrics(cpu_time, gpu_time, gpu_full_time, speedup, fullspeedup);
-    writeMetrics(cpu_time, gpu_time, gpu_full_time, speedup, fullspeedup, gpu_min, gpu_max, deviation);
-
-    num_subtests_called_++;
-    resetCurrentSubtest();
-}
-
-
-double TestSystem::meanTime(const vector<int64> &samples)
-{
-    double sum = accumulate(samples.begin(), samples.end(), 0.);
-    return sum / samples.size();
-}
-
-
-void TestSystem::printHeading()
-{
-    cout << endl;
-    cout << setiosflags(ios_base::left);
-#ifdef USE_OPENCL
-    cout << TAB << setw(10) << "CPU, ms" << setw(10) << "GPU, ms"
-         << setw(14) << "SPEEDUP" << setw(14) << "GPUTOTAL, ms" << setw(14) << "TOTALSPEEDUP"
-         << "DESCRIPTION\n";
-#else
-    cout << TAB << setw(10) << "CPU, ms\n";
-#endif
-    cout << resetiosflags(ios_base::left);
-}
-
-void TestSystem::writeHeading()
-{
-    if (!record_)
-    {
-#ifdef USE_OPENCL
-        recordname_ += "_OCL.csv";
-#else
-        recordname_ += "_CPU.csv";
-#endif
-        record_ = fopen(recordname_.c_str(), "w");
-    }
-
-#ifdef USE_OPENCL
-    fprintf(record_, "NAME,DESCRIPTION,CPU (ms),GPU (ms),SPEEDUP,GPUTOTAL (ms),TOTALSPEEDUP,GPU Min (ms),GPU Max (ms), Standard deviation (ms)\n");
-#else
-    fprintf(record_, "NAME,DESCRIPTION,CPU (ms)\n");
-#endif
-    fflush(record_);
-}
-
-void TestSystem::printSummary()
-{
-    cout << setiosflags(ios_base::fixed);
-    cout << "\naverage GPU speedup: x"
-         << setprecision(3) << speedup_total_ / std::max(1, num_subtests_called_)
-         << endl;
-    cout << "\nGPU exceeded: "
-         << setprecision(3) << speedup_faster_count_
-         << "\nGPU passed: "
-         << setprecision(3) << speedup_equal_count_
-         << "\nGPU failed: "
-         << setprecision(3) << speedup_slower_count_
-         << endl;
-    cout << "\nGPU exceeded rate: "
-         << setprecision(3) << (float)speedup_faster_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << "\nGPU passed rate: "
-         << setprecision(3) << (float)speedup_equal_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << "\nGPU failed rate: "
-         << setprecision(3) << (float)speedup_slower_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << endl;
-    cout << "\naverage GPUTOTAL speedup: x"
-         << setprecision(3) << speedup_full_total_ / std::max(1, num_subtests_called_)
-         << endl;
-    cout << "\nGPUTOTAL exceeded: "
-         << setprecision(3) << speedup_full_faster_count_
-         << "\nGPUTOTAL passed: "
-         << setprecision(3) << speedup_full_equal_count_
-         << "\nGPUTOTAL failed: "
-         << setprecision(3) << speedup_full_slower_count_
-         << endl;
-    cout << "\nGPUTOTAL exceeded rate: "
-         << setprecision(3) << (float)speedup_full_faster_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << "\nGPUTOTAL passed rate: "
-         << setprecision(3) << (float)speedup_full_equal_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << "\nGPUTOTAL failed rate: "
-         << setprecision(3) << (float)speedup_full_slower_count_ / std::max(1, num_subtests_called_) * 100
-         << "%"
-         << endl;
-    cout << resetiosflags(ios_base::fixed);
-}
-
-
-void TestSystem::printMetrics(double cpu_time, double gpu_time, double gpu_full_time, double speedup, double fullspeedup)
-{
-    cout << TAB << setiosflags(ios_base::left);
-    stringstream stream;
-
-    stream << cpu_time;
-    cout << setw(10) << stream.str();
-#ifdef USE_OPENCL
-    stream.str("");
-    stream << gpu_time;
-    cout << setw(10) << stream.str();
-
-    stream.str("");
-    stream << "x" << setprecision(3) << speedup;
-    cout << setw(14) << stream.str();
-
-    stream.str("");
-    stream << gpu_full_time;
-    cout << setw(14) << stream.str();
-
-    stream.str("");
-    stream << "x" << setprecision(3) << fullspeedup;
-    cout << setw(14) << stream.str();
-#endif
-    cout << cur_subtest_description_.str();
-    cout << resetiosflags(ios_base::left) << endl;
-}
-
-void TestSystem::writeMetrics(double cpu_time, double gpu_time, double gpu_full_time, double speedup, double fullspeedup, double gpu_min, double gpu_max, double std_dev)
-{
-    if (!record_)
-    {
-        recordname_ += ".csv";
-        record_ = fopen(recordname_.c_str(), "w");
-    }
-
-#ifdef USE_OPENCL
-    fprintf(record_, "%s,%s,%.3f,%.3f,%.3f,%.3f,%.3f,%.3f,%.3f,%.3f\n", itname_changed_ ? itname_.c_str() : "",
-            cur_subtest_description_.str().c_str(),
-            cpu_time, gpu_time, speedup, gpu_full_time, fullspeedup,
-            gpu_min, gpu_max, std_dev);
-#else
-    fprintf(record_, "%s,%s,%.3f\n",
-            itname_changed_ ? itname_.c_str() : "", cur_subtest_description_.str().c_str(), cpu_time);
-#endif
-
-    if (itname_changed_)
-    {
-        itname_changed_ = false;
-    }
-
-    fflush(record_);
-}
-
-void TestSystem::writeSummary()
-{
-    if (!record_)
-    {
-        recordname_ += ".csv";
-        record_ = fopen(recordname_.c_str(), "w");
-    }
-
-    fprintf(record_, "\nAverage GPU speedup: %.3f\n"
-            "exceeded: %d (%.3f%%)\n"
-            "passed: %d (%.3f%%)\n"
-            "failed: %d (%.3f%%)\n"
-            "\nAverage GPUTOTAL speedup: %.3f\n"
-            "exceeded: %d (%.3f%%)\n"
-            "passed: %d (%.3f%%)\n"
-            "failed: %d (%.3f%%)\n",
-            speedup_total_ / std::max(1, num_subtests_called_),
-            speedup_faster_count_, (float)speedup_faster_count_ / std::max(1, num_subtests_called_) * 100,
-            speedup_equal_count_, (float)speedup_equal_count_ / std::max(1, num_subtests_called_) * 100,
-            speedup_slower_count_, (float)speedup_slower_count_ / std::max(1, num_subtests_called_) * 100,
-            speedup_full_total_ / std::max(1, num_subtests_called_),
-            speedup_full_faster_count_, (float)speedup_full_faster_count_ / std::max(1, num_subtests_called_) * 100,
-            speedup_full_equal_count_, (float)speedup_full_equal_count_ / std::max(1, num_subtests_called_) * 100,
-            speedup_full_slower_count_, (float)speedup_full_slower_count_ / std::max(1, num_subtests_called_) * 100
-           );
-    fflush(record_);
-}
-
-void TestSystem::printError(const std::string &msg)
-{
-    cout << TAB << "[error: " << msg << "] " << cur_subtest_description_.str() << endl;
-}
-
-void gen(Mat &mat, int rows, int cols, int type, Scalar low, Scalar high)
-{
-    mat.create(rows, cols, type);
-    RNG rng(0);
-    rng.fill(mat, RNG::UNIFORM, low, high);
-}
-
-
-string abspath(const string &relpath)
-{
-    return TestSystem::instance().workingDir() + relpath;
-}
-
-
-int CV_CDECL cvErrorCallback(int /*status*/, const char * /*func_name*/,
-                             const char *err_msg, const char * /*file_name*/,
-                             int /*line*/, void * /*userdata*/)
-{
-    TestSystem::instance().printError(err_msg);
-    return 0;
-}
-
-/////////// matchTemplate ////////////////////////
-//void InitMatchTemplate()
-//{
-//	Mat src; gen(src, 500, 500, CV_32F, 0, 1);
-//	Mat templ; gen(templ, 500, 500, CV_32F, 0, 1);
-//#ifdef USE_OPENCL
-//	ocl::oclMat d_src(src), d_templ(templ), d_dst;
-//	ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
-//#endif
-//}
-TEST(matchTemplate)
-{
-    //InitMatchTemplate();
-
-    Mat src, templ, dst;
-    int templ_size = 5;
-
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        int all_type[] = {CV_32FC1, CV_32FC4};
-        std::string type_name[] = {"CV_32FC1", "CV_32FC4"};
-
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            for(templ_size = 5; templ_size <= 5; templ_size *= 5)
-            {
-                gen(src, size, size, all_type[j], 0, 1);
-
-                SUBTEST << src.cols << 'x' << src.rows << "; " << type_name[j] << "; templ " << templ_size << 'x' << templ_size << "; CCORR";
-
-                gen(templ, templ_size, templ_size, all_type[j], 0, 1);
-
-                matchTemplate(src, templ, dst, CV_TM_CCORR);
-
-                CPU_ON;
-                matchTemplate(src, templ, dst, CV_TM_CCORR);
-                CPU_OFF;
-
-#ifdef USE_OPENCL
-                ocl::oclMat d_src(src), d_templ, d_dst;
-
-                d_templ.upload(templ);
-
-                WARMUP_ON;
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
-                WARMUP_OFF;
-
-                GPU_ON;
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
-                GPU_OFF;
-
-                GPU_FULL_ON;
-                d_src.upload(src);
-                d_templ.upload(templ);
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR);
-                d_dst.download(dst);
-                GPU_FULL_OFF;
-#endif
-            }
-        }
-
-        int all_type_8U[] = {CV_8UC1};
-        std::string type_name_8U[] = {"CV_8UC1"};
-
-        for (size_t j = 0; j < sizeof(all_type_8U) / sizeof(int); j++)
-        {
-            for(templ_size = 5; templ_size <= 5; templ_size *= 5)
-            {
-                SUBTEST << src.cols << 'x' << src.rows << "; " << type_name_8U[j] << "; templ " << templ_size << 'x' << templ_size << "; CCORR_NORMED";
-
-                gen(src, size, size, all_type_8U[j], 0, 255);
-
-                gen(templ, templ_size, templ_size, all_type_8U[j], 0, 255);
-
-                matchTemplate(src, templ, dst, CV_TM_CCORR_NORMED);
-
-                CPU_ON;
-                matchTemplate(src, templ, dst, CV_TM_CCORR_NORMED);
-                CPU_OFF;
-
-#ifdef USE_OPENCL
-                ocl::oclMat d_src(src);
-                ocl::oclMat d_templ(templ), d_dst;
-
-                WARMUP_ON;
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR_NORMED);
-                WARMUP_OFF;
-
-                GPU_ON;
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR_NORMED);
-                GPU_OFF;
-
-                GPU_FULL_ON;
-                d_src.upload(src);
-                d_templ.upload(templ);
-                ocl::matchTemplate(d_src, d_templ, d_dst, CV_TM_CCORR_NORMED);
-                d_dst.download(dst);
-                GPU_FULL_OFF;
-#endif
-            }
-        }
-    }
-}
-
-///////////// PyrLKOpticalFlow ////////////////////////
-TEST(PyrLKOpticalFlow)
-{
-    std::string images1[] = {"rubberwhale1.png", "aloeL.jpg"};
-    std::string images2[] = {"rubberwhale2.png", "aloeR.jpg"};
-
-    for (size_t i = 0; i < sizeof(images1) / sizeof(std::string); i++)
-    {
-        Mat frame0 = imread(abspath(images1[i]), i == 0 ? IMREAD_COLOR : IMREAD_GRAYSCALE);
-
-        if (frame0.empty())
-        {
-            std::string errstr = "can't open " + images1[i];
-            throw runtime_error(errstr);
-        }
-
-        Mat frame1 = imread(abspath(images2[i]), i == 0 ? IMREAD_COLOR : IMREAD_GRAYSCALE);
-
-        if (frame1.empty())
-        {
-            std::string errstr = "can't open " + images2[i];
-            throw runtime_error(errstr);
-        }
-
-        Mat gray_frame;
-
-        if (i == 0)
-        {
-            cvtColor(frame0, gray_frame, COLOR_BGR2GRAY);
-        }
-
-        for (int points = 1000; points <= 4000; points *= 2)
-        {
-            if (i == 0)
-                SUBTEST << frame0.cols << "x" << frame0.rows << "; color; " << points << " points";
-            else
-                SUBTEST << frame0.cols << "x" << frame0.rows << "; gray; " << points << " points";
-            Mat nextPts_cpu;
-            Mat status_cpu;
-
-            vector<Point2f> pts;
-            goodFeaturesToTrack(i == 0 ? gray_frame : frame0, pts, points, 0.01, 0.0);
-
-            vector<Point2f> nextPts;
-            vector<unsigned char> status;
-
-            vector<float> err;
-
-            calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, err);
-
-            CPU_ON;
-            calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, err);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            ocl::PyrLKOpticalFlow d_pyrLK;
-
-            ocl::oclMat d_frame0(frame0);
-            ocl::oclMat d_frame1(frame1);
-
-            ocl::oclMat d_pts;
-            Mat pts_mat(1, (int)pts.size(), CV_32FC2, (void *)&pts[0]);
-            d_pts.upload(pts_mat);
-
-            ocl::oclMat d_nextPts;
-            ocl::oclMat d_status;
-            ocl::oclMat d_err;
-
-            WARMUP_ON;
-            d_pyrLK.sparse(d_frame0, d_frame1, d_pts, d_nextPts, d_status, &d_err);
-            WARMUP_OFF;
-
-            GPU_ON;
-            d_pyrLK.sparse(d_frame0, d_frame1, d_pts, d_nextPts, d_status, &d_err);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_frame0.upload(frame0);
-            d_frame1.upload(frame1);
-            d_pts.upload(pts_mat);
-            d_pyrLK.sparse(d_frame0, d_frame1, d_pts, d_nextPts, d_status, &d_err);
-
-            if (!d_nextPts.empty())
-            {
-                d_nextPts.download(nextPts_cpu);
-            }
-
-            if (!d_status.empty())
-            {
-                d_status.download(status_cpu);
-            }
-
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-
-///////////// pyrDown //////////////////////
-TEST(pyrDown)
-{
-    Mat src, dst;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            pyrDown(src, dst);
-
-            CPU_ON;
-            pyrDown(src, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-            ocl::oclMat d_dst;
-
-            WARMUP_ON;
-            ocl::pyrDown(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::pyrDown(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::pyrDown(d_src, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-    }
-}
-
-///////////// pyrUp ////////////////////////
-TEST(pyrUp)
-{
-    Mat src, dst;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 500; size <= 2000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            pyrUp(src, dst);
-
-            CPU_ON;
-            pyrUp(src, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-            ocl::oclMat d_dst;
-
-            WARMUP_ON;
-            ocl::pyrUp(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::pyrUp(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::pyrUp(d_src, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-    }
-}
-
-///////////// Canny ////////////////////////
-TEST(Canny)
-{
-    Mat img = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
-
-    if (img.empty())
-    {
-        throw runtime_error("can't open aloeL.jpg");
-    }
-
-    SUBTEST << img.cols << 'x' << img.rows << "; aloeL.jpg" << "; edges" << "; CV_8UC1";
-
-    Mat edges(img.size(), CV_8UC1);
-
-    CPU_ON;
-    Canny(img, edges, 50.0, 100.0);
-    CPU_OFF;
-
-#ifdef USE_OPENCL
-    ocl::oclMat d_img(img);
-    ocl::oclMat d_edges;
-    ocl::CannyBuf d_buf;
-
-    WARMUP_ON;
-    ocl::Canny(d_img, d_buf, d_edges, 50.0, 100.0);
-    WARMUP_OFF;
-
-    GPU_ON;
-    ocl::Canny(d_img, d_buf, d_edges, 50.0, 100.0);
-    GPU_OFF;
-
-    GPU_FULL_ON;
-    d_img.upload(img);
-    ocl::Canny(d_img, d_buf, d_edges, 50.0, 100.0);
-    d_edges.download(edges);
-    GPU_FULL_OFF;
-#endif
-}
-
-///////////// Haar ////////////////////////
-#ifdef USE_OPENCL
-namespace cv
-{
-namespace ocl
-{
-
-struct getRect
-{
-    Rect operator()(const CvAvgComp &e) const
-    {
-        return e.rect;
-    }
-};
-
-class CascadeClassifier_GPU : public OclCascadeClassifier
-{
-public:
-    void detectMultiScale(oclMat &image,
-                          std::vector<cv::Rect>& faces,
-                          double scaleFactor = 1.1,
-                          int minNeighbors = 3, int flags = 0,
-                          Size minSize = Size(),
-                          Size maxSize = Size())
-    {
-        (void)maxSize;
-        MemStorage storage(cvCreateMemStorage(0));
-        //CvMat img=image;
-        CvSeq *objs = oclHaarDetectObjects(image, storage, scaleFactor, minNeighbors, flags, minSize);
-        vector<CvAvgComp> vecAvgComp;
-        Seq<CvAvgComp>(objs).copyTo(vecAvgComp);
-        faces.resize(vecAvgComp.size());
-        std::transform(vecAvgComp.begin(), vecAvgComp.end(), faces.begin(), getRect());
-    }
-
-};
-
-}
-}
-#endif
-TEST(Haar)
-{
-    Mat img = imread(abspath("basketball1.png"), CV_LOAD_IMAGE_GRAYSCALE);
-
-    if (img.empty())
-    {
-        throw runtime_error("can't open basketball1.png");
-    }
-
-    CascadeClassifier faceCascadeCPU;
-
-    if (!faceCascadeCPU.load(abspath("haarcascade_frontalface_alt.xml")))
-    {
-        throw runtime_error("can't load haarcascade_frontalface_alt.xml");
-    }
-
-    vector<Rect> faces;
-
-    SUBTEST << img.cols << "x" << img.rows << "; scale image";
-    CPU_ON;
-    faceCascadeCPU.detectMultiScale(img, faces,
-                                    1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
-    CPU_OFF;
-
-#ifdef USE_OPENCL
-    ocl::CascadeClassifier_GPU faceCascade;
-
-    if (!faceCascade.load(abspath("haarcascade_frontalface_alt.xml")))
-    {
-        throw runtime_error("can't load haarcascade_frontalface_alt.xml");
-    }
-
-    ocl::oclMat d_img(img);
-
-    faces.clear();
-
-    WARMUP_ON;
-    faceCascade.detectMultiScale(d_img, faces,
-                                 1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
-    WARMUP_OFF;
-
-    faces.clear();
-
-    GPU_ON;
-    faceCascade.detectMultiScale(d_img, faces,
-                                 1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
-    GPU_OFF;
-
-    GPU_FULL_ON;
-    d_img.upload(img);
-    faceCascade.detectMultiScale(d_img, faces,
-                                 1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(30, 30));
-    GPU_FULL_OFF;
-#endif
-}
-
-///////////// blend ////////////////////////
-template <typename T>
-void blendLinearGold(const cv::Mat &img1, const cv::Mat &img2, const cv::Mat &weights1, const cv::Mat &weights2, cv::Mat &result_gold)
-{
-    result_gold.create(img1.size(), img1.type());
-
-    int cn = img1.channels();
-
-    for (int y = 0; y < img1.rows; ++y)
-    {
-        const float *weights1_row = weights1.ptr<float>(y);
-        const float *weights2_row = weights2.ptr<float>(y);
-        const T *img1_row = img1.ptr<T>(y);
-        const T *img2_row = img2.ptr<T>(y);
-        T *result_gold_row = result_gold.ptr<T>(y);
-
-        for (int x = 0; x < img1.cols * cn; ++x)
-        {
-            float w1 = weights1_row[x / cn];
-            float w2 = weights2_row[x / cn];
-            result_gold_row[x] = static_cast<T>((img1_row[x] * w1 + img2_row[x] * w2) / (w1 + w2 + 1e-5f));
-        }
-    }
-}
-TEST(blend)
-{
-    Mat src1, src2, weights1, weights2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_weights1, d_weights2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " and CV_32FC1";
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(weights1, size, size, CV_32FC1, 0, 1);
-            gen(weights2, size, size, CV_32FC1, 0, 1);
-
-            blendLinearGold<uchar>(src1, src2, weights1, weights2, dst);
-
-            CPU_ON;
-            blendLinearGold<uchar>(src1, src2, weights1, weights2, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            d_weights1.upload(weights1);
-            d_weights2.upload(weights2);
-
-            WARMUP_ON;
-            ocl::blendLinear(d_src1, d_src2, d_weights1, d_weights2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::blendLinear(d_src1, d_src2, d_weights1, d_weights2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            d_weights1.upload(weights1);
-            d_weights2.upload(weights2);
-            ocl::blendLinear(d_src1, d_src2, d_weights1, d_weights2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-    }
-}
-///////////// columnSum////////////////////////
-TEST(columnSum)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; CV_32FC1";
-
-        gen(src, size, size, CV_32FC1, 0, 256);
-
-        CPU_ON;
-        dst.create(src.size(), src.type());
-
-        for (int i = 1; i < src.rows; ++i)
-        {
-            for (int j = 0; j < src.cols; ++j)
-            {
-                dst.at<float>(i, j) = src.at<float>(i, j) += src.at<float>(i - 1, j);
-            }
-        }
-
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        d_src.upload(src);
-        WARMUP_ON;
-        ocl::columnSum(d_src, d_dst);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::columnSum(d_src, d_dst);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::columnSum(d_src, d_dst);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-
-///////////// HOG////////////////////////
-TEST(HOG)
-{
-    Mat src = imread(abspath("road.png"), cv::IMREAD_GRAYSCALE);
-
-    if (src.empty())
-    {
-        throw runtime_error("can't open road.png");
-    }
-
-
-    cv::HOGDescriptor hog;
-    hog.setSVMDetector(hog.getDefaultPeopleDetector());
-    std::vector<cv::Rect> found_locations;
-
-    SUBTEST << 768 << 'x' << 576 << "; road.png";
-
-    hog.detectMultiScale(src, found_locations);
-
-    CPU_ON;
-    hog.detectMultiScale(src, found_locations);
-    CPU_OFF;
-
-#ifdef USE_OPENCL
-    cv::ocl::HOGDescriptor ocl_hog;
-    ocl_hog.setSVMDetector(ocl_hog.getDefaultPeopleDetector());
-    ocl::oclMat d_src;
-    d_src.upload(src);
-
-    WARMUP_ON;
-    ocl_hog.detectMultiScale(d_src, found_locations);
-    WARMUP_OFF;
-
-    GPU_ON;
-    ocl_hog.detectMultiScale(d_src, found_locations);
-    GPU_OFF;
-
-    GPU_FULL_ON;
-    d_src.upload(src);
-    ocl_hog.detectMultiScale(d_src, found_locations);
-    GPU_FULL_OFF;
-#endif
-}
-
-///////////// SURF ////////////////////////
-
-TEST(SURF)
-{
-    Mat keypoints_cpu;
-    Mat descriptors_cpu;
-
-    Mat src = imread(abspath("aloeL.jpg"), CV_LOAD_IMAGE_GRAYSCALE);
-
-    if (src.empty())
-    {
-        throw runtime_error("can't open aloeL.jpg");
-    }
-
-    SUBTEST << src.cols << "x" << src.rows << "; aloeL.jpg";
-    SURF surf;
-    vector<KeyPoint> keypoints;
-    Mat descriptors;
-
-    surf(src, Mat(), keypoints, descriptors);
-
-    CPU_ON;
-    keypoints.clear();
-    surf(src, Mat(), keypoints, descriptors);
-    CPU_OFF;
-
-#ifdef USE_OPENCL
-    ocl::SURF_OCL d_surf;
-    ocl::oclMat d_src(src);
-    ocl::oclMat d_keypoints;
-    ocl::oclMat d_descriptors;
-
-    WARMUP_ON;
-    d_surf(d_src, ocl::oclMat(), d_keypoints, d_descriptors);
-    WARMUP_OFF;
-
-    GPU_ON;
-    d_surf(d_src, ocl::oclMat(), d_keypoints, d_descriptors);
-    GPU_OFF;
-
-    GPU_FULL_ON;
-    d_src.upload(src);
-    d_surf(d_src, ocl::oclMat(), d_keypoints, d_descriptors);
-
-    if (!d_keypoints.empty())
-    {
-        d_keypoints.download(keypoints_cpu);
-    }
-
-    if (!d_descriptors.empty())
-    {
-        d_descriptors.download(descriptors_cpu);
-    }
-
-    GPU_FULL_OFF;
-#endif
-}
-//////////////////// BruteForceMatch /////////////////
-TEST(BruteForceMatcher)
-{
-    Mat trainIdx_cpu;
-    Mat distance_cpu;
-    Mat allDist_cpu;
-    Mat nMatches_cpu;
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        // Init CPU matcher
-        int desc_len = 64;
-
-        BFMatcher matcher(NORM_L2);
-
-        Mat query;
-        gen(query, size, desc_len, CV_32F, 0, 1);
-
-        Mat train;
-        gen(train, size, desc_len, CV_32F, 0, 1);
-        // Output
-        vector< vector<DMatch> > matches(2);
-#ifdef USE_OPENCL
-        // Init GPU matcher
-        ocl::BruteForceMatcher_OCL_base d_matcher(ocl::BruteForceMatcher_OCL_base::L2Dist);
-
-        ocl::oclMat d_query(query);
-        ocl::oclMat d_train(train);
-
-        ocl::oclMat d_trainIdx, d_distance, d_allDist, d_nMatches;
-#endif
-        SUBTEST << size << "; match";
-
-        matcher.match(query, train, matches[0]);
-
-        CPU_ON;
-        matcher.match(query, train, matches[0]);
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        WARMUP_ON;
-        d_matcher.matchSingle(d_query, d_train, d_trainIdx, d_distance);
-        WARMUP_OFF;
-
-        GPU_ON;
-        d_matcher.matchSingle(d_query, d_train, d_trainIdx, d_distance);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_query.upload(query);
-        d_train.upload(train);
-        d_matcher.match(d_query, d_train, matches[0]);
-        GPU_FULL_OFF;
-#endif
-
-        SUBTEST << size << "; knnMatch";
-
-        matcher.knnMatch(query, train, matches, 2);
-
-        CPU_ON;
-        matcher.knnMatch(query, train, matches, 2);
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        WARMUP_ON;
-        d_matcher.knnMatchSingle(d_query, d_train, d_trainIdx, d_distance, d_allDist, 2);
-        WARMUP_OFF;
-
-        GPU_ON;
-        d_matcher.knnMatchSingle(d_query, d_train, d_trainIdx, d_distance, d_allDist, 2);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_query.upload(query);
-        d_train.upload(train);
-        d_matcher.knnMatch(d_query, d_train, matches, 2);
-        GPU_FULL_OFF;
-#endif
-        SUBTEST << size << "; radiusMatch";
-
-        float max_distance = 2.0f;
-
-        matcher.radiusMatch(query, train, matches, max_distance);
-
-        CPU_ON;
-        matcher.radiusMatch(query, train, matches, max_distance);
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        d_trainIdx.release();
-
-        WARMUP_ON;
-        d_matcher.radiusMatchSingle(d_query, d_train, d_trainIdx, d_distance, d_nMatches, max_distance);
-        WARMUP_OFF;
-
-        GPU_ON;
-        d_matcher.radiusMatchSingle(d_query, d_train, d_trainIdx, d_distance, d_nMatches, max_distance);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_query.upload(query);
-        d_train.upload(train);
-        d_matcher.radiusMatch(d_query, d_train, matches, max_distance);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-///////////// Lut ////////////////////////
-TEST(lut)
-{
-    Mat src, lut, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_lut, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC3};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC3"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src, size, size, all_type[j], 0, 256);
-            gen(lut, 1, 256, CV_8UC1, 0, 1);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-            LUT(src, lut, dst);
-
-            CPU_ON;
-            LUT(src, lut, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src.upload(src);
-            d_lut.upload(lut);
-
-            WARMUP_ON;
-            ocl::LUT(d_src, d_lut, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::LUT(d_src, d_lut, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            d_lut.upload(lut);
-            ocl::LUT(d_src, d_lut, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Exp ////////////////////////
-TEST(Exp)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; CV_32FC1";
-
-        gen(src, size, size, CV_32FC1, 0, 256);
-        gen(dst, size, size, CV_32FC1, 0, 256);
-
-        exp(src, dst);
-
-        CPU_ON;
-        exp(src, dst);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::exp(d_src, d_dst);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::exp(d_src, d_dst);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::exp(d_src, d_dst);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-
-///////////// LOG ////////////////////////
-TEST(Log)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; 32F";
-
-        gen(src, size, size, CV_32F, 1, 10);
-
-        log(src, dst);
-
-        CPU_ON;
-        log(src, dst);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::log(d_src, d_dst);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::log(d_src, d_dst);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::log(d_src, d_dst);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-
-///////////// Add ////////////////////////
-
-TEST(Add)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src1, size, size, all_type[j], 0, 1);
-            gen(src2, size, size, all_type[j], 0, 1);
-
-            add(src1, src2, dst);
-
-            CPU_ON;
-            add(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::add(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::add(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::add(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Mul ////////////////////////
-TEST(Mul)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            multiply(src1, src2, dst);
-
-            CPU_ON;
-            multiply(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::multiply(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::multiply(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::multiply(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Div ////////////////////////
-TEST(Div)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            divide(src1, src2, dst);
-
-            CPU_ON;
-            divide(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::divide(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::divide(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::divide(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Absdiff ////////////////////////
-TEST(Absdiff)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            absdiff(src1, src2, dst);
-
-            CPU_ON;
-            absdiff(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::absdiff(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::absdiff(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::absdiff(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// CartToPolar ////////////////////////
-TEST(CartToPolar)
-{
-    Mat src1, src2, dst, dst1;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst, d_dst1;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-            gen(dst1, size, size, all_type[j], 0, 256);
-
-
-            cartToPolar(src1, src2, dst, dst1, 1);
-
-            CPU_ON;
-            cartToPolar(src1, src2, dst, dst1, 1);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::cartToPolar(d_src1, d_src2, d_dst, d_dst1, 1);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::cartToPolar(d_src1, d_src2, d_dst, d_dst1, 1);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::cartToPolar(d_src1, d_src2, d_dst, d_dst1, 1);
-            d_dst.download(dst);
-            d_dst1.download(dst1);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// PolarToCart ////////////////////////
-TEST(PolarToCart)
-{
-    Mat src1, src2, dst, dst1;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst, d_dst1;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-            gen(dst1, size, size, all_type[j], 0, 256);
-
-
-            polarToCart(src1, src2, dst, dst1, 1);
-
-            CPU_ON;
-            polarToCart(src1, src2, dst, dst1, 1);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::polarToCart(d_src1, d_src2, d_dst, d_dst1, 1);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::polarToCart(d_src1, d_src2, d_dst, d_dst1, 1);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::polarToCart(d_src1, d_src2, d_dst, d_dst1, 1);
-            d_dst.download(dst);
-            d_dst1.download(dst1);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Magnitude ////////////////////////
-TEST(magnitude)
-{
-    Mat x, y, mag;
-#ifdef USE_OPENCL
-    ocl::oclMat d_x, d_y, d_mag;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(x, size, size, all_type[j], 0, 1);
-            gen(y, size, size, all_type[j], 0, 1);
-
-            magnitude(x, y, mag);
-
-            CPU_ON;
-            magnitude(x, y, mag);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_x.upload(x);
-            d_y.upload(y);
-
-            WARMUP_ON;
-            ocl::magnitude(d_x, d_y, d_mag);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::magnitude(d_x, d_y, d_mag);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_x.upload(x);
-            d_y.upload(y);
-            ocl::magnitude(d_x, d_y, d_mag);
-            d_mag.download(mag);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Transpose ////////////////////////
-TEST(Transpose)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-            transpose(src, dst);
-
-            CPU_ON;
-            transpose(src, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::transpose(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::transpose(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::transpose(d_src, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Flip ////////////////////////
-TEST(Flip)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " ; FLIP_BOTH";
-
-            gen(src, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-            flip(src, dst, 0);
-
-            CPU_ON;
-            flip(src, dst, 0);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::flip(d_src, d_dst, 0);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::flip(d_src, d_dst, 0);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::flip(d_src, d_dst, 0);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// minMax ////////////////////////
-TEST(minMax)
-{
-    Mat src;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src;
-#endif
-    double min_val, max_val;
-    Point min_loc, max_loc;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            CPU_ON;
-            minMaxLoc(src, &min_val, &max_val, &min_loc, &max_loc);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::minMax(d_src, &min_val, &max_val);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::minMax(d_src, &min_val, &max_val);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::minMax(d_src, &min_val, &max_val);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// minMaxLoc ////////////////////////
-TEST(minMaxLoc)
-{
-    Mat src;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src;
-#endif
-    double min_val, max_val;
-    Point min_loc, max_loc;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 1);
-
-            CPU_ON;
-            minMaxLoc(src, &min_val, &max_val, &min_loc, &max_loc);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::minMaxLoc(d_src, &min_val, &max_val, &min_loc, &max_loc);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::minMaxLoc(d_src, &min_val, &max_val, &min_loc, &max_loc);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::minMaxLoc(d_src, &min_val, &max_val, &min_loc, &max_loc);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Sum ////////////////////////
-TEST(Sum)
-{
-    Mat src;
-    Scalar cpures, gpures;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src;
-#endif
-    int all_type[] = {CV_8UC1, CV_32SC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32SC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            cpures = sum(src);
-
-            CPU_ON;
-            cpures = sum(src);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            gpures = ocl::sum(d_src);
-            WARMUP_OFF;
-
-            GPU_ON;
-            gpures = ocl::sum(d_src);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            gpures = ocl::sum(d_src);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// countNonZero ////////////////////////
-TEST(countNonZero)
-{
-    Mat src;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src;
-#endif
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            countNonZero(src);
-
-            CPU_ON;
-            countNonZero(src);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::countNonZero(d_src);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::countNonZero(d_src);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::countNonZero(d_src);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Phase ////////////////////////
-TEST(Phase)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            phase(src1, src2, dst, 1);
-
-            CPU_ON;
-            phase(src1, src2, dst, 1);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::phase(d_src1, d_src2, d_dst, 1);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::phase(d_src1, d_src2, d_dst, 1);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::phase(d_src1, d_src2, d_dst, 1);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// bitwise_and////////////////////////
-TEST(bitwise_and)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32SC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32SC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            bitwise_and(src1, src2, dst);
-
-            CPU_ON;
-            bitwise_and(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::bitwise_and(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::bitwise_and(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::bitwise_and(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// bitwise_or////////////////////////
-TEST(bitwise_or)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32SC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32SC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            bitwise_or(src1, src2, dst);
-
-            CPU_ON;
-            bitwise_or(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::bitwise_or(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::bitwise_or(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::bitwise_or(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// bitwise_xor////////////////////////
-TEST(bitwise_xor)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32SC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32SC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            bitwise_xor(src1, src2, dst);
-
-            CPU_ON;
-            bitwise_xor(src1, src2, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::bitwise_xor(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::bitwise_xor(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::bitwise_xor(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// bitwise_not////////////////////////
-TEST(bitwise_not)
-{
-    Mat src1, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32SC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32SC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            bitwise_not(src1, dst);
-
-            CPU_ON;
-            bitwise_not(src1, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-
-            WARMUP_ON;
-            ocl::bitwise_not(d_src1, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::bitwise_not(d_src1, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            ocl::bitwise_not(d_src1, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// compare////////////////////////
-TEST(compare)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int CMP_EQ = 0;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            compare(src1, src2, dst, CMP_EQ);
-
-            CPU_ON;
-            compare(src1, src2, dst, CMP_EQ);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::compare(d_src1, d_src2, d_dst, CMP_EQ);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::compare(d_src1, d_src2, d_dst, CMP_EQ);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::compare(d_src1, d_src2, d_dst, CMP_EQ);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// pow ////////////////////////
-TEST(pow)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 100);
-            gen(dst, size, size, all_type[j], 0, 100);
-
-            pow(src, -2.0, dst);
-
-            CPU_ON;
-            pow(src, -2.0, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-            d_dst.upload(dst);
-
-            WARMUP_ON;
-            ocl::pow(d_src, -2.0, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::pow(d_src, -2.0, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::pow(d_src, -2.0, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// MagnitudeSqr////////////////////////
-TEST(MagnitudeSqr)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    int all_type[] = {CV_32FC1};
-    std::string type_name[] = {"CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t t = 0; t < sizeof(all_type) / sizeof(int); t++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[t];
-
-            gen(src1, size, size, all_type[t], 0, 256);
-            gen(src2, size, size, all_type[t], 0, 256);
-            gen(dst, size, size, all_type[t], 0, 256);
-
-
-            for (int i = 0; i < src1.rows; ++i)
-
-                for (int j = 0; j < src1.cols; ++j)
-                {
-                    float val1 = src1.at<float>(i, j);
-                    float val2 = src2.at<float>(i, j);
-
-                    ((float *)(dst.data))[i * dst.step / 4 + j] = val1 * val1 + val2 * val2;
-
-                }
-
-            CPU_ON;
-
-            for (int i = 0; i < src1.rows; ++i)
-                for (int j = 0; j < src1.cols; ++j)
-                {
-                    float val1 = src1.at<float>(i, j);
-                    float val2 = src2.at<float>(i, j);
-
-                    ((float *)(dst.data))[i * dst.step / 4 + j] = val1 * val1 + val2 * val2;
-
-                }
-
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::magnitudeSqr(d_src1, d_src2, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::magnitudeSqr(d_src1, d_src2, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::magnitudeSqr(d_src1, d_src2, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// AddWeighted////////////////////////
-TEST(AddWeighted)
-{
-    Mat src1, src2, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_dst;
-#endif
-    double alpha = 2.0, beta = 1.0, gama = 3.0;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(src2, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            addWeighted(src1, alpha, src2, beta, gama, dst);
-
-            CPU_ON;
-            addWeighted(src1, alpha, src2, beta, gama, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-
-            WARMUP_ON;
-            ocl::addWeighted(d_src1, alpha, d_src2, beta, gama, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::addWeighted(d_src1, alpha, d_src2, beta, gama, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            d_src2.upload(src2);
-            ocl::addWeighted(d_src1, alpha, d_src2, beta, gama, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Blur////////////////////////
-TEST(Blur)
-{
-    Mat src1, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_dst;
-#endif
-    Size ksize = Size(3, 3);
-    int bordertype = BORDER_CONSTANT;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            blur(src1, dst, ksize, Point(-1, -1), bordertype);
-
-            CPU_ON;
-            blur(src1, dst, ksize, Point(-1, -1), bordertype);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-
-            WARMUP_ON;
-            ocl::blur(d_src1, d_dst, ksize, Point(-1, -1), bordertype);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::blur(d_src1, d_dst, ksize, Point(-1, -1), bordertype);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            ocl::blur(d_src1, d_dst, ksize, Point(-1, -1), bordertype);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Laplacian////////////////////////
-TEST(Laplacian)
-{
-    Mat src1, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_dst;
-#endif
-    int ksize = 3;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src1, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-
-
-            Laplacian(src1, dst, -1, ksize, 1);
-
-            CPU_ON;
-            Laplacian(src1, dst, -1, ksize, 1);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src1.upload(src1);
-
-            WARMUP_ON;
-            ocl::Laplacian(d_src1, d_dst, -1, ksize, 1);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::Laplacian(d_src1, d_dst, -1, ksize, 1);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src1.upload(src1);
-            ocl::Laplacian(d_src1, d_dst, -1, ksize, 1);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Erode ////////////////////
-TEST(Erode)
-{
-    Mat src, dst, ker;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4", "CV_32FC1", "CV_32FC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], Scalar::all(0), Scalar::all(256));
-            ker = getStructuringElement(MORPH_RECT, Size(3, 3));
-
-            erode(src, dst, ker);
-
-            CPU_ON;
-            erode(src, dst, ker);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::erode(d_src, d_dst, ker);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::erode(d_src, d_dst, ker);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::erode(d_src, d_dst, ker);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Sobel ////////////////////////
-TEST(Sobel)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int dx = 1;
-    int dy = 1;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            Sobel(src, dst, -1, dx, dy);
-
-            CPU_ON;
-            Sobel(src, dst, -1, dx, dy);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::Sobel(d_src, d_dst, -1, dx, dy);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::Sobel(d_src, d_dst, -1, dx, dy);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::Sobel(d_src, d_dst, -1, dx, dy);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Scharr ////////////////////////
-TEST(Scharr)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int dx = 1;
-    int dy = 0;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            Scharr(src, dst, -1, dx, dy);
-
-            CPU_ON;
-            Scharr(src, dst, -1, dx, dy);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::Scharr(d_src, d_dst, -1, dx, dy);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::Scharr(d_src, d_dst, -1, dx, dy);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::Scharr(d_src, d_dst, -1, dx, dy);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// GaussianBlur ////////////////////////
-TEST(GaussianBlur)
-{
-    Mat src, dst;
-    int all_type[] = {CV_8UC1, CV_8UC4, CV_32FC1, CV_32FC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4", "CV_32FC1", "CV_32FC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            GaussianBlur(src, dst, Size(9, 9), 0);
-
-            CPU_ON;
-            GaussianBlur(src, dst, Size(9, 9), 0);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-            ocl::oclMat d_dst(src.size(), src.type());
-            ocl::oclMat d_buf;
-
-            WARMUP_ON;
-            ocl::GaussianBlur(d_src, d_dst, Size(9, 9), 0);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::GaussianBlur(d_src, d_dst, Size(9, 9), 0);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::GaussianBlur(d_src, d_dst, Size(9, 9), 0);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// equalizeHist ////////////////////////
-TEST(equalizeHist)
-{
-    Mat src, dst;
-    int all_type[] = {CV_8UC1};
-    std::string type_name[] = {"CV_8UC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            equalizeHist(src, dst);
-
-            CPU_ON;
-            equalizeHist(src, dst);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-            ocl::oclMat d_dst;
-            ocl::oclMat d_hist;
-            ocl::oclMat d_buf;
-
-            WARMUP_ON;
-            ocl::equalizeHist(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::equalizeHist(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::equalizeHist(d_src, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-/////////// CopyMakeBorder //////////////////////
-TEST(CopyMakeBorder)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_dst;
-#endif
-    int bordertype = BORDER_CONSTANT;
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            copyMakeBorder(src, dst, 7, 5, 5, 7, bordertype, cv::Scalar(1.0));
-
-            CPU_ON;
-            copyMakeBorder(src, dst, 7, 5, 5, 7, bordertype, cv::Scalar(1.0));
-            CPU_OFF;
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-
-            WARMUP_ON;
-            ocl::copyMakeBorder(d_src, d_dst, 7, 5, 5, 7, bordertype, cv::Scalar(1.0));
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::copyMakeBorder(d_src, d_dst, 7, 5, 5, 7, bordertype, cv::Scalar(1.0));
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::copyMakeBorder(d_src, d_dst, 7, 5, 5, 7, bordertype, cv::Scalar(1.0));
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// cornerMinEigenVal ////////////////////////
-TEST(cornerMinEigenVal)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_dst;
-#endif
-    int blockSize = 7, apertureSize = 1 + 2 * (rand() % 4);
-    int borderType = BORDER_REFLECT;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            cornerMinEigenVal(src, dst, blockSize, apertureSize, borderType);
-
-            CPU_ON;
-            cornerMinEigenVal(src, dst, blockSize, apertureSize, borderType);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(src);
-
-            WARMUP_ON;
-            ocl::cornerMinEigenVal(d_src, d_dst, blockSize, apertureSize, borderType);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::cornerMinEigenVal(d_src, d_dst, blockSize, apertureSize, borderType);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::cornerMinEigenVal(d_src, d_dst, blockSize, apertureSize, borderType);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// cornerHarris ////////////////////////
-TEST(cornerHarris)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " ; BORDER_REFLECT";
-
-            gen(src, size, size, all_type[j], 0, 1);
-
-            cornerHarris(src, dst, 5, 7, 0.1, BORDER_REFLECT);
-
-            CPU_ON;
-            cornerHarris(src, dst, 5, 7, 0.1, BORDER_REFLECT);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::cornerHarris(d_src, d_dst, 5, 7, 0.1, BORDER_REFLECT);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::cornerHarris(d_src, d_dst, 5, 7, 0.1, BORDER_REFLECT);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::cornerHarris(d_src, d_dst, 5, 7, 0.1, BORDER_REFLECT);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-
-    }
-}
-///////////// integral ////////////////////////
-TEST(integral)
-{
-    Mat src, sum;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_sum, d_buf;
-#endif
-    int all_type[] = {CV_8UC1};
-    std::string type_name[] = {"CV_8UC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j]  ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            integral(src, sum);
-
-            CPU_ON;
-            integral(src, sum);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::integral(d_src, d_sum);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::integral(d_src, d_sum);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::integral(d_src, d_sum);
-            d_sum.download(sum);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// WarpAffine ////////////////////////
-TEST(WarpAffine)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    static const double coeffs[2][3] =
-    {
-        {cos(3.14 / 6), -sin(3.14 / 6), 100.0},
-        {sin(3.14 / 6), cos(3.14 / 6), -100.0}
-    };
-    Mat M(2, 3, CV_64F, (void *)coeffs);
-    int interpolation = INTER_NEAREST;
-
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-            Size size1 = Size(size, size);
-
-            warpAffine(src, dst, M, size1, interpolation);
-
-            CPU_ON;
-            warpAffine(src, dst, M, size1, interpolation);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::warpAffine(d_src, d_dst, M, size1, interpolation);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::warpAffine(d_src, d_dst, M, size1, interpolation);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::warpAffine(d_src, d_dst, M, size1, interpolation);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// WarpPerspective ////////////////////////
-TEST(WarpPerspective)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    static const double coeffs[3][3] =
-    {
-        {cos(3.14 / 6), -sin(3.14 / 6), 100.0},
-        {sin(3.14 / 6), cos(3.14 / 6), -100.0},
-        {0.0, 0.0, 1.0}
-    };
-    Mat M(3, 3, CV_64F, (void *)coeffs);
-    int interpolation = INTER_NEAREST;
-
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-            gen(dst, size, size, all_type[j], 0, 256);
-            Size size1 = Size(size, size);
-
-            warpPerspective(src, dst, M, size1, interpolation);
-
-            CPU_ON;
-            warpPerspective(src, dst, M, size1, interpolation);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::warpPerspective(d_src, d_dst, M, size1, interpolation);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::warpPerspective(d_src, d_dst, M, size1, interpolation);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::warpPerspective(d_src, d_dst, M, size1, interpolation);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// resize ////////////////////////
-TEST(resize)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " ; up";
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            resize(src, dst, Size(), 2.0, 2.0);
-
-            CPU_ON;
-            resize(src, dst, Size(), 2.0, 2.0);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::resize(d_src, d_dst, Size(), 2.0, 2.0);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::resize(d_src, d_dst, Size(), 2.0, 2.0);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::resize(d_src, d_dst, Size(), 2.0, 2.0);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " ; down";
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            resize(src, dst, Size(), 0.5, 0.5);
-
-            CPU_ON;
-            resize(src, dst, Size(), 0.5, 0.5);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::resize(d_src, d_dst, Size(), 0.5, 0.5);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::resize(d_src, d_dst, Size(), 0.5, 0.5);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::resize(d_src, d_dst, Size(), 0.5, 0.5);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// threshold////////////////////////
-TEST(threshold)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; 8UC1; THRESH_BINARY";
-
-        gen(src, size, size, CV_8U, 0, 100);
-
-        threshold(src, dst, 50.0, 0.0, THRESH_BINARY);
-
-        CPU_ON;
-        threshold(src, dst, 50.0, 0.0, THRESH_BINARY);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_BINARY);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_BINARY);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_BINARY);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; 32FC1; THRESH_TRUNC [NPP]";
-
-        gen(src, size, size, CV_32FC1, 0, 100);
-
-        threshold(src, dst, 50.0, 0.0, THRESH_TRUNC);
-
-        CPU_ON;
-        threshold(src, dst, 50.0, 0.0, THRESH_TRUNC);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_TRUNC);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_TRUNC);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::threshold(d_src, d_dst, 50.0, 0.0, THRESH_TRUNC);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-///////////// meanShiftFiltering////////////////////////
-TEST(meanShiftFiltering)
-{
-    int sp = 10, sr = 10;
-
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; 8UC3 vs 8UC4";
-
-        gen(src, size, size, CV_8UC3, Scalar::all(0), Scalar::all(256));
-
-        pyrMeanShiftFiltering(src, dst, sp, sr);
-
-        CPU_ON;
-        pyrMeanShiftFiltering(src, dst, sp, sr);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        gen(src, size, size, CV_8UC4, Scalar::all(0), Scalar::all(256));
-
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::meanShiftFiltering(d_src, d_dst, sp, sr);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::meanShiftFiltering(d_src, d_dst, sp, sr);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::meanShiftFiltering(d_src, d_dst, sp, sr);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-///////////// meanShiftProc////////////////////////
-COOR do_meanShift(int x0, int y0, uchar *sptr, uchar *dptr, int sstep, cv::Size size, int sp, int sr, int maxIter, float eps, int *tab)
-{
-
-    int isr2 = sr * sr;
-    int c0, c1, c2, c3;
-    int iter;
-    uchar *ptr = NULL;
-    uchar *pstart = NULL;
-    int revx = 0, revy = 0;
-    c0 = sptr[0];
-    c1 = sptr[1];
-    c2 = sptr[2];
-    c3 = sptr[3];
-
-    // iterate meanshift procedure
-    for (iter = 0; iter < maxIter; iter++)
-    {
-        int count = 0;
-        int s0 = 0, s1 = 0, s2 = 0, sx = 0, sy = 0;
-
-        //mean shift: process pixels in window (p-sigmaSp)x(p+sigmaSp)
-        int minx = x0 - sp;
-        int miny = y0 - sp;
-        int maxx = x0 + sp;
-        int maxy = y0 + sp;
-
-        //deal with the image boundary
-        if (minx < 0)
-        {
-            minx = 0;
-        }
-
-        if (miny < 0)
-        {
-            miny = 0;
-        }
-
-        if (maxx >= size.width)
-        {
-            maxx = size.width - 1;
-        }
-
-        if (maxy >= size.height)
-        {
-            maxy = size.height - 1;
-        }
-
-        if (iter == 0)
-        {
-            pstart = sptr;
-        }
-        else
-        {
-            pstart = pstart + revy * sstep + (revx << 2); //point to the new position
-        }
-
-        ptr = pstart;
-        ptr = ptr + (miny - y0) * sstep + ((minx - x0) << 2); //point to the start in the row
-
-        for (int y = miny; y <= maxy; y++, ptr += sstep - ((maxx - minx + 1) << 2))
-        {
-            int rowCount = 0;
-            int x = minx;
-#if CV_ENABLE_UNROLLED
-
-            for (; x + 4 <= maxx; x += 4, ptr += 16)
-            {
-                int t0, t1, t2;
-                t0 = ptr[0], t1 = ptr[1], t2 = ptr[2];
-
-                if (tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
-                {
-                    s0 += t0;
-                    s1 += t1;
-                    s2 += t2;
-                    sx += x;
-                    rowCount++;
-                }
-
-                t0 = ptr[4], t1 = ptr[5], t2 = ptr[6];
-
-                if (tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
-                {
-                    s0 += t0;
-                    s1 += t1;
-                    s2 += t2;
-                    sx += x + 1;
-                    rowCount++;
-                }
-
-                t0 = ptr[8], t1 = ptr[9], t2 = ptr[10];
-
-                if (tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
-                {
-                    s0 += t0;
-                    s1 += t1;
-                    s2 += t2;
-                    sx += x + 2;
-                    rowCount++;
-                }
-
-                t0 = ptr[12], t1 = ptr[13], t2 = ptr[14];
-
-                if (tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
-                {
-                    s0 += t0;
-                    s1 += t1;
-                    s2 += t2;
-                    sx += x + 3;
-                    rowCount++;
-                }
-            }
-
-#endif
-
-            for (; x <= maxx; x++, ptr += 4)
-            {
-                int t0 = ptr[0], t1 = ptr[1], t2 = ptr[2];
-
-                if (tab[t0 - c0 + 255] + tab[t1 - c1 + 255] + tab[t2 - c2 + 255] <= isr2)
-                {
-                    s0 += t0;
-                    s1 += t1;
-                    s2 += t2;
-                    sx += x;
-                    rowCount++;
-                }
-            }
-
-            if (rowCount == 0)
-            {
-                continue;
-            }
-
-            count += rowCount;
-            sy += y * rowCount;
-        }
-
-        if (count == 0)
-        {
-            break;
-        }
-
-        int x1 = sx / count;
-        int y1 = sy / count;
-        s0 = s0 / count;
-        s1 = s1 / count;
-        s2 = s2 / count;
-
-        bool stopFlag = (x0 == x1 && y0 == y1) || (abs(x1 - x0) + abs(y1 - y0) +
-                        tab[s0 - c0 + 255] + tab[s1 - c1 + 255] + tab[s2 - c2 + 255] <= eps);
-
-        //revise the pointer corresponding to the new (y0,x0)
-        revx = x1 - x0;
-        revy = y1 - y0;
-
-        x0 = x1;
-        y0 = y1;
-        c0 = s0;
-        c1 = s1;
-        c2 = s2;
-
-        if (stopFlag)
-        {
-            break;
-        }
-    } //for iter
-
-    dptr[0] = (uchar)c0;
-    dptr[1] = (uchar)c1;
-    dptr[2] = (uchar)c2;
-    dptr[3] = (uchar)c3;
-
-    COOR coor;
-    coor.x = static_cast<short>(x0);
-    coor.y = static_cast<short>(y0);
-    return coor;
-}
-
-void meanShiftProc_(const Mat &src_roi, Mat &dst_roi, Mat &dstCoor_roi, int sp, int sr, cv::TermCriteria crit)
-{
-
-    if (src_roi.empty())
-    {
-        CV_Error(CV_StsBadArg, "The input image is empty");
-    }
-
-    if (src_roi.depth() != CV_8U || src_roi.channels() != 4)
-    {
-        CV_Error(CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported");
-    }
-
-    CV_Assert((src_roi.cols == dst_roi.cols) && (src_roi.rows == dst_roi.rows) &&
-              (src_roi.cols == dstCoor_roi.cols) && (src_roi.rows == dstCoor_roi.rows));
-    CV_Assert(!(dstCoor_roi.step & 0x3));
-
-    if (!(crit.type & cv::TermCriteria::MAX_ITER))
-    {
-        crit.maxCount = 5;
-    }
-
-    int maxIter = std::min(std::max(crit.maxCount, 1), 100);
-    float eps;
-
-    if (!(crit.type & cv::TermCriteria::EPS))
-    {
-        eps = 1.f;
-    }
-
-    eps = (float)std::max(crit.epsilon, 0.0);
-
-    int tab[512];
-
-    for (int i = 0; i < 512; i++)
-    {
-        tab[i] = (i - 255) * (i - 255);
-    }
-
-    uchar *sptr = src_roi.data;
-    uchar *dptr = dst_roi.data;
-    short *dCoorptr = (short *)dstCoor_roi.data;
-    int sstep = (int)src_roi.step;
-    int dstep = (int)dst_roi.step;
-    int dCoorstep = (int)dstCoor_roi.step >> 1;
-    cv::Size size = src_roi.size();
-
-    for (int i = 0; i < size.height; i++, sptr += sstep - (size.width << 2),
-            dptr += dstep - (size.width << 2), dCoorptr += dCoorstep - (size.width << 1))
-    {
-        for (int j = 0; j < size.width; j++, sptr += 4, dptr += 4, dCoorptr += 2)
-        {
-            *((COOR *)dCoorptr) = do_meanShift(j, i, sptr, dptr, sstep, size, sp, sr, maxIter, eps, tab);
-        }
-    }
-
-}
-TEST(meanShiftProc)
-{
-    Mat src, dst, dstCoor_roi;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst, d_dstCoor_roi;
-#endif
-    TermCriteria crit(TermCriteria::COUNT + TermCriteria::EPS, 5, 1);
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; 8UC4 and CV_16SC2 ";
-
-        gen(src, size, size, CV_8UC4, Scalar::all(0), Scalar::all(256));
-        gen(dst, size, size, CV_8UC4, Scalar::all(0), Scalar::all(256));
-        gen(dstCoor_roi, size, size, CV_16SC2, Scalar::all(0), Scalar::all(256));
-
-        meanShiftProc_(src, dst, dstCoor_roi, 5, 6, crit);
-
-        CPU_ON;
-        meanShiftProc_(src, dst, dstCoor_roi, 5, 6, crit);
-        CPU_OFF;
-#ifdef USE_OPENCL
-        d_src.upload(src);
-
-        WARMUP_ON;
-        ocl::meanShiftProc(d_src, d_dst, d_dstCoor_roi, 5, 6, crit);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::meanShiftProc(d_src, d_dst, d_dstCoor_roi, 5, 6, crit);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::meanShiftProc(d_src, d_dst, d_dstCoor_roi, 5, 6, crit);
-        d_dst.download(dst);
-        d_dstCoor_roi.download(dstCoor_roi);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-///////////// ConvertTo////////////////////////
-TEST(ConvertTo)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " to 32FC1";
-
-            gen(src, size, size, all_type[j], 0, 256);
-            //gen(dst, size, size, all_type[j], 0, 256);
-
-            //d_dst.upload(dst);
-
-            src.convertTo(dst, CV_32FC1);
-
-            CPU_ON;
-            src.convertTo(dst, CV_32FC1);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            d_src.convertTo(d_dst, CV_32FC1);
-            WARMUP_OFF;
-
-            GPU_ON;
-            d_src.convertTo(d_dst, CV_32FC1);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            d_src.convertTo(d_dst, CV_32FC1);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// copyTo////////////////////////
-TEST(copyTo)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-            //gen(dst, size, size, all_type[j], 0, 256);
-
-            //d_dst.upload(dst);
-
-            src.copyTo(dst);
-
-            CPU_ON;
-            src.copyTo(dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            d_src.copyTo(d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            d_src.copyTo(d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            d_src.copyTo(d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// setTo////////////////////////
-TEST(setTo)
-{
-    Mat src, dst;
-    Scalar val(1, 2, 3, 4);
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-
-            gen(src, size, size, all_type[j], 0, 256);
-
-            src.setTo(val);
-
-            CPU_ON;
-            src.setTo(val);
-            CPU_OFF;
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            d_src.setTo(val);
-            WARMUP_OFF;
-
-            GPU_ON;
-            d_src.setTo(val);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            d_src.setTo(val);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// Merge////////////////////////
-TEST(Merge)
-{
-    Mat dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_dst;
-#endif
-    int channels = 4;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] ;
-            Size size1 = Size(size, size);
-            std::vector<Mat> src(channels);
-
-            for (int i = 0; i < channels; ++i)
-            {
-                src[i] = Mat(size1, all_type[j], cv::Scalar::all(i));
-            }
-
-            merge(src, dst);
-
-            CPU_ON;
-            merge(src, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            std::vector<ocl::oclMat> d_src(channels);
-
-            for (int i = 0; i < channels; ++i)
-            {
-                d_src[i] = ocl::oclMat(size1, all_type[j], cv::Scalar::all(i));
-            }
-
-            WARMUP_ON;
-            ocl::merge(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::merge(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-
-            for (int i = 0; i < channels; ++i)
-            {
-                d_src[i] = ocl::oclMat(size1, CV_8U, cv::Scalar::all(i));
-            }
-
-            ocl::merge(d_src, d_dst);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// Split////////////////////////
-TEST(Split)
-{
-    //int channels = 4;
-    int all_type[] = {CV_8UC1, CV_32FC1};
-    std::string type_name[] = {"CV_8UC1", "CV_32FC1"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j];
-            Size size1 = Size(size, size);
-
-            Mat src(size1, CV_MAKE_TYPE(all_type[j], 4), cv::Scalar(1, 2, 3, 4));
-
-            std::vector<cv::Mat> dst;
-
-            split(src, dst);
-
-            CPU_ON;
-            split(src, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            ocl::oclMat d_src(size1, CV_MAKE_TYPE(all_type[j], 4), cv::Scalar(1, 2, 3, 4));
-            std::vector<cv::ocl::oclMat> d_dst;
-
-            WARMUP_ON;
-            ocl::split(d_src, d_dst);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::split(d_src, d_dst);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::split(d_src, d_dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-
-///////////// norm////////////////////////
-TEST(norm)
-{
-    Mat src, buf;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_buf;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size << "; CV_8UC1; NORM_INF";
-
-        gen(src, size, size, CV_8UC1, Scalar::all(0), Scalar::all(1));
-        gen(buf, size, size, CV_8UC1, Scalar::all(0), Scalar::all(1));
-
-        norm(src, NORM_INF);
-
-        CPU_ON;
-        norm(src, NORM_INF);
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        d_src.upload(src);
-        d_buf.upload(buf);
-
-        WARMUP_ON;
-        ocl::norm(d_src, d_buf, NORM_INF);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::norm(d_src, d_buf, NORM_INF);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src.upload(src);
-        ocl::norm(d_src, d_buf, NORM_INF);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-///////////// remap////////////////////////
-TEST(remap)
-{
-    Mat src, dst, xmap, ymap;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst, d_xmap, d_ymap;
-#endif
-    int all_type[] = {CV_8UC1, CV_8UC4};
-    std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-    int interpolation = INTER_LINEAR;
-    int borderMode = BORDER_CONSTANT;
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t t = 0; t < sizeof(all_type) / sizeof(int); t++)
-        {
-            SUBTEST << size << 'x' << size << "; src " << type_name[t] << "; map CV_32FC1";
-
-            gen(src, size, size, all_type[t], 0, 256);
-
-            xmap.create(size, size, CV_32FC1);
-            dst.create(size, size, CV_32FC1);
-            ymap.create(size, size, CV_32FC1);
-
-            for (int i = 0; i < size; ++i)
-            {
-                float *xmap_row = xmap.ptr<float>(i);
-                float *ymap_row = ymap.ptr<float>(i);
-
-                for (int j = 0; j < size; ++j)
-                {
-                    xmap_row[j] = (j - size * 0.5f) * 0.75f + size * 0.5f;
-                    ymap_row[j] = (i - size * 0.5f) * 0.75f + size * 0.5f;
-                }
-            }
-
-
-            remap(src, dst, xmap, ymap, interpolation, borderMode);
-
-            CPU_ON;
-            remap(src, dst, xmap, ymap, interpolation, borderMode);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src.upload(src);
-            d_dst.upload(dst);
-            d_xmap.upload(xmap);
-            d_ymap.upload(ymap);
-
-            WARMUP_ON;
-            ocl::remap(d_src, d_dst, d_xmap, d_ymap, interpolation, borderMode);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::remap(d_src, d_dst, d_xmap, d_ymap, interpolation, borderMode);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::remap(d_src, d_dst, d_xmap, d_ymap, interpolation, borderMode);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-///////////// cvtColor////////////////////////
-TEST(cvtColor)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-    int all_type[] = {CV_8UC4};
-    std::string type_name[] = {"CV_8UC4"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            gen(src, size, size, all_type[j], 0, 256);
-            SUBTEST << size << "x" << size << "; " << type_name[j] << " ; CV_RGBA2GRAY";
-
-            cvtColor(src, dst, CV_RGBA2GRAY, 4);
-
-            CPU_ON;
-            cvtColor(src, dst, CV_RGBA2GRAY, 4);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::cvtColor(d_src, d_dst, CV_RGBA2GRAY, 4);
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::cvtColor(d_src, d_dst, CV_RGBA2GRAY, 4);
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::cvtColor(d_src, d_dst, CV_RGBA2GRAY, 4);
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-
-    }
-
-
-}
-///////////// filter2D////////////////////////
-TEST(filter2D)
-{
-    Mat src;
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        int all_type[] = {CV_8UC1, CV_8UC4};
-        std::string type_name[] = {"CV_8UC1", "CV_8UC4"};
-
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            gen(src, size, size, all_type[j], 0, 256);
-
-            for (int ksize = 3; ksize <= 15; ksize = 2*ksize+1)
-            {
-                SUBTEST << "ksize = " << ksize << "; " << size << 'x' << size << "; " << type_name[j] ;
-
-                Mat kernel;
-                gen(kernel, ksize, ksize, CV_32FC1, 0.0, 1.0);
-
-                Mat dst;
-                cv::filter2D(src, dst, -1, kernel);
-
-                CPU_ON;
-                cv::filter2D(src, dst, -1, kernel);
-                CPU_OFF;
-#ifdef USE_OPENCL
-                ocl::oclMat d_src(src);
-                ocl::oclMat d_dst;
-
-                WARMUP_ON;
-                ocl::filter2D(d_src, d_dst, -1, kernel);
-                WARMUP_OFF;
-
-                GPU_ON;
-                ocl::filter2D(d_src, d_dst, -1, kernel);
-                GPU_OFF;
-
-                GPU_FULL_ON;
-                d_src.upload(src);
-                ocl::filter2D(d_src, d_dst, -1, kernel);
-                d_dst.download(dst);
-                GPU_FULL_OFF;
-#endif
-            }
-
-        }
-
-
-    }
-}
-
-
-///////////// dft ////////////////////////
-TEST(dft)
-{
-    Mat src, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src, d_dst;
-#endif
-
-    int all_type[] = {CV_32FC1, CV_32FC2};
-    std::string type_name[] = {"CV_32FC1", "CV_32FC2"};
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        for (size_t j = 0; j < sizeof(all_type) / sizeof(int); j++)
-        {
-            SUBTEST << size << 'x' << size << "; " << type_name[j] << " ; complex-to-complex";
-
-            gen(src, size, size, all_type[j], Scalar::all(0), Scalar::all(1));
-
-            dft(src, dst);
-
-            CPU_ON;
-            dft(src, dst);
-            CPU_OFF;
-
-#ifdef USE_OPENCL
-            d_src.upload(src);
-
-            WARMUP_ON;
-            ocl::dft(d_src, d_dst, Size(size, size));
-            WARMUP_OFF;
-
-            GPU_ON;
-            ocl::dft(d_src, d_dst, Size(size, size));
-            GPU_OFF;
-
-            GPU_FULL_ON;
-            d_src.upload(src);
-            ocl::dft(d_src, d_dst, Size(size, size));
-            d_dst.download(dst);
-            GPU_FULL_OFF;
-#endif
-        }
-
-    }
-}
-
-///////////// gemm ////////////////////////
-TEST(gemm)
-{
-    Mat src1, src2, src3, dst;
-#ifdef USE_OPENCL
-    ocl::oclMat d_src1, d_src2, d_src3, d_dst;
-#endif
-
-    for (int size = 1000; size <= 4000; size *= 2)
-    {
-        SUBTEST << size << 'x' << size;
-
-        gen(src1, size, size, CV_32FC1, Scalar::all(-10), Scalar::all(10));
-        gen(src2, size, size, CV_32FC1, Scalar::all(-10), Scalar::all(10));
-        gen(src3, size, size, CV_32FC1, Scalar::all(-10), Scalar::all(10));
-
-        gemm(src1, src2, 1.0, src3, 1.0, dst);
-
-        CPU_ON;
-        gemm(src1, src2, 1.0, src3, 1.0, dst);
-        CPU_OFF;
-
-#ifdef USE_OPENCL
-        d_src1.upload(src1);
-        d_src2.upload(src2);
-        d_src3.upload(src3);
-
-        WARMUP_ON;
-        ocl::gemm(d_src1, d_src2, 1.0, d_src3, 1.0, d_dst);
-        WARMUP_OFF;
-
-        GPU_ON;
-        ocl::gemm(d_src1, d_src2, 1.0, d_src3, 1.0, d_dst);
-        GPU_OFF;
-
-        GPU_FULL_ON;
-        d_src1.upload(src1);
-        d_src2.upload(src2);
-        d_src3.upload(src3);
-        ocl::gemm(d_src1, d_src2, 1.0, d_src3, 1.0, d_dst);
-        d_dst.download(dst);
-        GPU_FULL_OFF;
-#endif
-    }
-}
-
-int main(int argc, const char *argv[])
-{
-#ifdef USE_OPENCL
-    vector<ocl::Info> oclinfo;
-    int num_devices = getDevice(oclinfo);
-
-    if (num_devices < 1)
-    {
-        cerr << "no device found\n";
-        return -1;
-    }
-
-    int devidx = 0;
-
-    for (size_t i = 0; i < oclinfo.size(); i++)
-    {
-        for (size_t j = 0; j < oclinfo[i].DeviceName.size(); j++)
-        {
-            printf("device %d: %s\n", devidx++, oclinfo[i].DeviceName[j].c_str());
-        }
-    }
-
-#endif
-    redirectError(cvErrorCallback);
-
-    const char *keys =
-        "{ h help    | false | print help message }"
-        "{ f filter  |       | filter for test }"
-        "{ w workdir |       | set working directory }"
-        "{ l list    | false | show all tests }"
-        "{ d device  | 0     | device id }"
-        "{ i iters   | 10    | iteration count }"
-        "{ m warmup  | 1     | gpu warm up iteration count}"
-        "{ t xtop    | 1.1   | xfactor top boundary}"
-        "{ b xbottom | 0.9   | xfactor bottom boundary}"
-        "{ v verify  | false | only run gpu once to verify if problems occur}";
-
-    CommandLineParser cmd(argc, argv, keys);
-
-    if (cmd.get<bool>("help"))
-    {
-        cout << "Avaible options:" << endl;
-        cmd.printMessage();
-        return 0;
-    }
-
-#ifdef USE_OPENCL
-    int device = cmd.get<int>("device");
-
-    if (device < 0 || device >= num_devices)
-    {
-        cerr << "Invalid device ID" << endl;
-        return -1;
-    }
-
-    if (cmd.get<bool>("verify"))
-    {
-        TestSystem::instance().setNumIters(1);
-        TestSystem::instance().setGPUWarmupIters(0);
-        TestSystem::instance().setCPUIters(0);
-    }
-
-    devidx = 0;
-
-    for (size_t i = 0; i < oclinfo.size(); i++)
-    {
-        for (size_t j = 0; j < oclinfo[i].DeviceName.size(); j++, devidx++)
-        {
-            if (device == devidx)
-            {
-                ocl::setDevice(oclinfo[i], (int)j);
-                TestSystem::instance().setRecordName(oclinfo[i].DeviceName[j]);
-                printf("\nuse %d: %s\n", devidx, oclinfo[i].DeviceName[j].c_str());
-                goto END_DEV;
-            }
-        }
-    }
-
-END_DEV:
-
-#endif
-    string filter = cmd.get<string>("filter");
-    string workdir = cmd.get<string>("workdir");
-    bool list = cmd.get<bool>("list");
-    int iters = cmd.get<int>("iters");
-    int wu_iters = cmd.get<int>("warmup");
-    double x_top = cmd.get<double>("xtop");
-    double x_bottom = cmd.get<double>("xbottom");
-
-    TestSystem::instance().setTopThreshold(x_top);
-    TestSystem::instance().setBottomThreshold(x_bottom);
-
-    if (!filter.empty())
-    {
-        TestSystem::instance().setTestFilter(filter);
-    }
-
-    if (!workdir.empty())
-    {
-        if (workdir[workdir.size() - 1] != '/' && workdir[workdir.size() - 1] != '\\')
-        {
-            workdir += '/';
-        }
-
-        TestSystem::instance().setWorkingDir(workdir);
-    }
-
-    if (list)
-    {
-        TestSystem::instance().setListMode(true);
-    }
-
-    TestSystem::instance().setNumIters(iters);
-    TestSystem::instance().setGPUWarmupIters(wu_iters);
-
-    TestSystem::instance().run();
-
-    return 0;
-}
diff --git a/samples/ocl/pyrlk_optical_flow.cpp b/samples/ocl/pyrlk_optical_flow.cpp
new file mode 100644
index 0000000000..f7d0661712
--- /dev/null
+++ b/samples/ocl/pyrlk_optical_flow.cpp
@@ -0,0 +1,273 @@
+#include <iostream>
+#include <vector>
+#include <iomanip>
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/ocl/ocl.hpp"
+#include "opencv2/video/video.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+typedef unsigned char uchar;
+#define LOOP_NUM 10
+int64 work_begin = 0;
+int64 work_end = 0;
+
+static void workBegin()
+{
+    work_begin = getTickCount();
+}
+static void workEnd()
+{
+    work_end += (getTickCount() - work_begin);
+}
+static double getTime()
+{
+    return work_end * 1000. / getTickFrequency();
+}
+
+static void download(const oclMat& d_mat, vector<Point2f>& vec)
+{
+    vec.clear();
+    vec.resize(d_mat.cols);
+    Mat mat(1, d_mat.cols, CV_32FC2, (void*)&vec[0]);
+    d_mat.download(mat);
+}
+
+static void download(const oclMat& d_mat, vector<uchar>& vec)
+{
+    vec.clear();
+    vec.resize(d_mat.cols);
+    Mat mat(1, d_mat.cols, CV_8UC1, (void*)&vec[0]);
+    d_mat.download(mat);
+}
+
+static void drawArrows(Mat& frame, const vector<Point2f>& prevPts, const vector<Point2f>& nextPts, const vector<uchar>& status, Scalar line_color = Scalar(0, 0, 255))
+{
+    for (size_t i = 0; i < prevPts.size(); ++i)
+    {
+        if (status[i])
+        {
+            int line_thickness = 1;
+
+            Point p = prevPts[i];
+            Point q = nextPts[i];
+
+            double angle = atan2((double) p.y - q.y, (double) p.x - q.x);
+
+            double hypotenuse = sqrt( (double)(p.y - q.y)*(p.y - q.y) + (double)(p.x - q.x)*(p.x - q.x) );
+
+            if (hypotenuse < 1.0)
+                continue;
+
+            // Here we lengthen the arrow by a factor of three.
+            q.x = (int) (p.x - 3 * hypotenuse * cos(angle));
+            q.y = (int) (p.y - 3 * hypotenuse * sin(angle));
+
+            // Now we draw the main line of the arrow.
+            line(frame, p, q, line_color, line_thickness);
+
+            // Now draw the tips of the arrow. I do some scaling so that the
+            // tips look proportional to the main line of the arrow.
+
+            p.x = (int) (q.x + 9 * cos(angle + CV_PI / 4));
+            p.y = (int) (q.y + 9 * sin(angle + CV_PI / 4));
+            line(frame, p, q, line_color, line_thickness);
+
+            p.x = (int) (q.x + 9 * cos(angle - CV_PI / 4));
+            p.y = (int) (q.y + 9 * sin(angle - CV_PI / 4));
+            line(frame, p, q, line_color, line_thickness);
+        }
+    }
+}
+
+
+int main(int argc, const char* argv[])
+{
+    static std::vector<Info> ocl_info;
+    ocl::getDevice(ocl_info);
+    //if you want to use undefault device, set it here
+    setDevice(ocl_info[0]);
+
+    //set this to save kernel compile time from second time you run
+    ocl::setBinpath("./");
+    const char* keys =
+        "{ help h           | false           | print help message }"
+        "{ left l           |                 | specify left image }"
+        "{ right r          |                 | specify right image }"
+        "{ camera c         | 0               | enable camera capturing }"
+        "{ use_cpu s        | false           | use cpu or gpu to process the image }"
+        "{ video v          |                 | use video as input }"
+        "{ output o         | pyrlk_output.jpg| specify output save path when input is images }"
+        "{ points           | 1000            | specify points count [GoodFeatureToTrack] }"
+        "{ min_dist         | 0               | specify minimal distance between points [GoodFeatureToTrack] }";
+
+    CommandLineParser cmd(argc, argv, keys);
+
+    if (cmd.has("help"))
+    {
+        cmd.printMessage();
+        return 0;
+    }
+
+    bool defaultPicturesFail = false;
+    string fname0 = cmd.get<string>("left");
+    string fname1 = cmd.get<string>("right");
+    string vdofile = cmd.get<string>("video");
+    string outfile = cmd.get<string>("output");
+    int points = cmd.get<int>("points");
+    double minDist = cmd.get<double>("min_dist");
+    bool useCPU = cmd.has("s");
+    int inputName = cmd.get<int>("c");
+
+    oclMat d_nextPts, d_status;
+    GoodFeaturesToTrackDetector_OCL d_features(points);
+    Mat frame0 = imread(fname0, cv::IMREAD_GRAYSCALE);
+    Mat frame1 = imread(fname1, cv::IMREAD_GRAYSCALE);
+    PyrLKOpticalFlow d_pyrLK;
+    vector<cv::Point2f> pts(points);
+    vector<cv::Point2f> nextPts(points);
+    vector<unsigned char> status(points);
+    vector<float> err;
+
+    cout << "Points count : " << points << endl << endl;
+
+    if (frame0.empty() || frame1.empty())
+    {
+        VideoCapture capture;
+        Mat frame, frameCopy;
+        Mat frame0Gray, frame1Gray;
+        Mat ptr0, ptr1;
+
+        if(vdofile == "")
+            capture.open( inputName );
+        else
+            capture.open(vdofile.c_str());
+
+        int c = inputName ;
+        if(!capture.isOpened())
+        {
+            if(vdofile == "")
+                cout << "Capture from CAM " << c << " didn't work" << endl;
+            else
+                cout << "Capture from file " << vdofile << " failed" <<endl;
+            if (defaultPicturesFail)
+            {
+                return -1;
+            }
+            goto nocamera;
+        }
+
+        cout << "In capture ..." << endl;
+        for(int i = 0;; i++)
+        {
+            if( !capture.read(frame) )
+                break;
+
+            if (i == 0)
+            {
+                frame.copyTo( frame0 );
+                cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+            }
+            else
+            {
+                if (i%2 == 1)
+                {
+                    frame.copyTo(frame1);
+                    cvtColor(frame1, frame1Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame0Gray;
+                    ptr1 = frame1Gray;
+                }
+                else
+                {
+                    frame.copyTo(frame0);
+                    cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame1Gray;
+                    ptr1 = frame0Gray;
+                }
+
+                if (useCPU)
+                {
+                    pts.clear();
+                    goodFeaturesToTrack(ptr0, pts, points, 0.01, 0.0);
+                    if(pts.size() == 0)
+                        continue;
+                    calcOpticalFlowPyrLK(ptr0, ptr1, pts, nextPts, status, err);
+                }
+                else
+                {
+                    oclMat d_img(ptr0), d_prevPts;
+                    d_features(d_img, d_prevPts);
+                    if(!d_prevPts.rows || !d_prevPts.cols)
+                        continue;
+                    d_pyrLK.sparse(d_img, oclMat(ptr1), d_prevPts, d_nextPts, d_status);
+                    d_features.downloadPoints(d_prevPts,pts);
+                    download(d_nextPts, nextPts);
+                    download(d_status, status);
+                }
+                if (i%2 == 1)
+                    frame1.copyTo(frameCopy);
+                else
+                    frame0.copyTo(frameCopy);
+                drawArrows(frameCopy, pts, nextPts, status, Scalar(255, 0, 0));
+                imshow("PyrLK [Sparse]", frameCopy);
+            }
+
+            if( waitKey( 10 ) >= 0 )
+                goto _cleanup_;
+        }
+
+        waitKey(0);
+
+_cleanup_:
+        capture.release();
+    }
+    else
+    {
+nocamera:
+        for(int i = 0; i <= LOOP_NUM; i ++)
+        {
+            cout << "loop" << i << endl;
+            if (i > 0) workBegin();
+
+            if (useCPU)
+            {
+                goodFeaturesToTrack(frame0, pts, points, 0.01, minDist);
+                calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, err);
+            }
+            else
+            {
+                oclMat d_img(frame0), d_prevPts;
+                d_features(d_img, d_prevPts);
+                d_pyrLK.sparse(d_img, oclMat(frame1), d_prevPts, d_nextPts, d_status);
+                d_features.downloadPoints(d_prevPts, pts);
+                download(d_nextPts, nextPts);
+                download(d_status, status);
+            }
+
+            if (i > 0 && i <= LOOP_NUM)
+                workEnd();
+
+            if (i == LOOP_NUM)
+            {
+                if (useCPU)
+                    cout << "average CPU time (noCamera) : ";
+                else
+                    cout << "average GPU time (noCamera) : ";
+
+                cout << getTime() / LOOP_NUM << " ms" << endl;
+
+                drawArrows(frame0, pts, nextPts, status, Scalar(255, 0, 0));
+                imshow("PyrLK [Sparse]", frame0);
+                imwrite(outfile, frame0);
+            }
+        }
+    }
+
+    waitKey();
+
+    return 0;
+}
diff --git a/samples/ocl/squares.cpp b/samples/ocl/squares.cpp
index 6b184161f7..d31e360073 100644
--- a/samples/ocl/squares.cpp
+++ b/samples/ocl/squares.cpp
@@ -2,11 +2,11 @@
 // It loads several images sequentially and tries to find squares in
 // each image
 
-#include "opencv2/core/core.hpp"
+#include "opencv2/core.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/imgproc/imgproc.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/ocl/ocl.hpp"
-
 #include <iostream>
 #include <math.h>
 #include <string.h>
@@ -14,23 +14,50 @@
 using namespace cv;
 using namespace std;
 
-static void help()
-{
-    cout <<
-        "\nA program using OCL module pyramid scaling, Canny, dilate functions, threshold, split; cpu contours, contour simpification and\n"
-        "memory storage (it's got it all folks) to find\n"
-        "squares in a list of images pic1-6.png\n"
-        "Returns sequence of squares detected on the image.\n"
-        "the sequence is stored in the specified memory storage\n"
-        "Call:\n"
-        "./squares\n"
-        "Using OpenCV version %s\n" << CV_VERSION << "\n" << endl;
-}
+#define ACCURACY_CHECK 1
 
+#if ACCURACY_CHECK
+// check if two vectors of vector of points are near or not
+// prior assumption is that they are in correct order
+static bool checkPoints(
+    vector< vector<Point> > set1,
+    vector< vector<Point> > set2,
+    int maxDiff = 5)
+{
+    if(set1.size() != set2.size())
+    {
+        return false;
+    }
+
+    for(vector< vector<Point> >::iterator it1 = set1.begin(), it2 = set2.begin();
+            it1 < set1.end() && it2 < set2.end(); it1 ++, it2 ++)
+    {
+        vector<Point> pts1 = *it1;
+        vector<Point> pts2 = *it2;
+
+
+        if(pts1.size() != pts2.size())
+        {
+            return false;
+        }
+        for(size_t i = 0; i < pts1.size(); i ++)
+        {
+            Point pt1 = pts1[i], pt2 = pts2[i];
+            if(std::abs(pt1.x - pt2.x) > maxDiff ||
+                    std::abs(pt1.y - pt2.y) > maxDiff)
+            {
+                return false;
+            }
+        }
+    }
+    return true;
+}
+#endif
 
 int thresh = 50, N = 11;
 const char* wndname = "OpenCL Square Detection Demo";
 
+
 // helper function:
 // finds a cosine of angle between vectors
 // from pt0->pt1 and from pt0->pt2
@@ -43,9 +70,92 @@ static double angle( Point pt1, Point pt2, Point pt0 )
     return (dx1*dx2 + dy1*dy2)/sqrt((dx1*dx1 + dy1*dy1)*(dx2*dx2 + dy2*dy2) + 1e-10);
 }
 
+
 // returns sequence of squares detected on the image.
 // the sequence is stored in the specified memory storage
 static void findSquares( const Mat& image, vector<vector<Point> >& squares )
+{
+    squares.clear();
+    Mat pyr, timg, gray0(image.size(), CV_8U), gray;
+
+    // down-scale and upscale the image to filter out the noise
+    pyrDown(image, pyr, Size(image.cols/2, image.rows/2));
+    pyrUp(pyr, timg, image.size());
+    vector<vector<Point> > contours;
+
+    // find squares in every color plane of the image
+    for( int c = 0; c < 3; c++ )
+    {
+        int ch[] = {c, 0};
+        mixChannels(&timg, 1, &gray0, 1, ch, 1);
+
+        // try several threshold levels
+        for( int l = 0; l < N; l++ )
+        {
+            // hack: use Canny instead of zero threshold level.
+            // Canny helps to catch squares with gradient shading
+            if( l == 0 )
+            {
+                // apply Canny. Take the upper threshold from slider
+                // and set the lower to 0 (which forces edges merging)
+                Canny(gray0, gray, 0, thresh, 5);
+                // dilate canny output to remove potential
+                // holes between edge segments
+                dilate(gray, gray, Mat(), Point(-1,-1));
+            }
+            else
+            {
+                // apply threshold if l!=0:
+                //     tgray(x,y) = gray(x,y) < (l+1)*255/N ? 255 : 0
+                cv::threshold(gray0, gray, (l+1)*255/N, 255, THRESH_BINARY);
+            }
+
+            // find contours and store them all as a list
+            findContours(gray, contours, RETR_LIST, CHAIN_APPROX_SIMPLE);
+
+            vector<Point> approx;
+
+            // test each contour
+            for( size_t i = 0; i < contours.size(); i++ )
+            {
+                // approximate contour with accuracy proportional
+                // to the contour perimeter
+                approxPolyDP(Mat(contours[i]), approx, arcLength(Mat(contours[i]), true)*0.02, true);
+
+                // square contours should have 4 vertices after approximation
+                // relatively large area (to filter out noisy contours)
+                // and be convex.
+                // Note: absolute value of an area is used because
+                // area may be positive or negative - in accordance with the
+                // contour orientation
+                if( approx.size() == 4 &&
+                        fabs(contourArea(Mat(approx))) > 1000 &&
+                        isContourConvex(Mat(approx)) )
+                {
+                    double maxCosine = 0;
+
+                    for( int j = 2; j < 5; j++ )
+                    {
+                        // find the maximum cosine of the angle between joint edges
+                        double cosine = fabs(angle(approx[j%4], approx[j-2], approx[j-1]));
+                        maxCosine = MAX(maxCosine, cosine);
+                    }
+
+                    // if cosines of all angles are small
+                    // (all angles are ~90 degree) then write quandrange
+                    // vertices to resultant sequence
+                    if( maxCosine < 0.3 )
+                        squares.push_back(approx);
+                }
+            }
+        }
+    }
+}
+
+
+// returns sequence of squares detected on the image.
+// the sequence is stored in the specified memory storage
+static void findSquares_ocl( const Mat& image, vector<vector<Point> >& squares )
 {
     squares.clear();
 
@@ -88,10 +198,9 @@ static void findSquares( const Mat& image, vector<vector<Point> >& squares )
             }
 
             // find contours and store them all as a list
-            findContours(gray, contours, CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE);
+            findContours(gray, contours, RETR_LIST, CHAIN_APPROX_SIMPLE);
 
             vector<Point> approx;
-
             // test each contour
             for( size_t i = 0; i < contours.size(); i++ )
             {
@@ -106,11 +215,10 @@ static void findSquares( const Mat& image, vector<vector<Point> >& squares )
                 // area may be positive or negative - in accordance with the
                 // contour orientation
                 if( approx.size() == 4 &&
-                    fabs(contourArea(Mat(approx))) > 1000 &&
-                    isContourConvex(Mat(approx)) )
+                        fabs(contourArea(Mat(approx))) > 1000 &&
+                        isContourConvex(Mat(approx)) )
                 {
                     double maxCosine = 0;
-
                     for( int j = 2; j < 5; j++ )
                     {
                         // find the maximum cosine of the angle between joint edges
@@ -137,42 +245,95 @@ static void drawSquares( Mat& image, const vector<vector<Point> >& squares )
     {
         const Point* p = &squares[i][0];
         int n = (int)squares[i].size();
-        polylines(image, &p, &n, 1, true, Scalar(0,255,0), 3, CV_AA);
+        polylines(image, &p, &n, 1, true, Scalar(0,255,0), 3, LINE_AA);
     }
-
-    imshow(wndname, image);
 }
 
 
-int main(int /*argc*/, char** /*argv*/)
+// draw both pure-C++ and ocl square results onto a single image
+static Mat drawSquaresBoth( const Mat& image,
+                            const vector<vector<Point> >& sqsCPP,
+                            const vector<vector<Point> >& sqsOCL
+)
 {
+    Mat imgToShow(Size(image.cols * 2, image.rows), image.type());
+    Mat lImg = imgToShow(Rect(Point(0, 0), image.size()));
+    Mat rImg = imgToShow(Rect(Point(image.cols, 0), image.size()));
+    image.copyTo(lImg);
+    image.copyTo(rImg);
+    drawSquares(lImg, sqsCPP);
+    drawSquares(rImg, sqsOCL);
+    float fontScale = 0.8f;
+    Scalar white = Scalar::all(255), black = Scalar::all(0);
+
+    putText(lImg, "C++", Point(10, 20), FONT_HERSHEY_COMPLEX_SMALL, fontScale, black, 2);
+    putText(rImg, "OCL", Point(10, 20), FONT_HERSHEY_COMPLEX_SMALL, fontScale, black, 2);
+    putText(lImg, "C++", Point(10, 20), FONT_HERSHEY_COMPLEX_SMALL, fontScale, white, 1);
+    putText(rImg, "OCL", Point(10, 20), FONT_HERSHEY_COMPLEX_SMALL, fontScale, white, 1);
+
+    return imgToShow;
+}
+
+
+int main(int argc, char** argv)
+{
+    const char* keys =
+        "{ i | input   |                    | specify input image }"
+        "{ o | output  | squares_output.jpg | specify output save path}";
+    CommandLineParser cmd(argc, argv, keys);
+    string inputName = cmd.get<string>("i");
+    string outfile = cmd.get<string>("o");
+    if(inputName.empty())
+    {
+        cout << "Avaible options:" << endl;
+        cmd.printMessage();
+        return 0;
+    }
 
-    //ocl::setBinpath("F:/kernel_bin");
     vector<ocl::Info> info;
     CV_Assert(ocl::getDevice(info));
-
-    static const char* names[] = { "pic1.png", "pic2.png", "pic3.png",
-        "pic4.png", "pic5.png", "pic6.png", 0 };
-    help();
+    int iterations = 10;
     namedWindow( wndname, 1 );
-    vector<vector<Point> > squares;
+    vector<vector<Point> > squares_cpu, squares_ocl;
 
-    for( int i = 0; names[i] != 0; i++ )
+    Mat image = imread(inputName, 1);
+    if( image.empty() )
     {
-        Mat image = imread(names[i], 1);
-        if( image.empty() )
-        {
-            cout << "Couldn't load " << names[i] << endl;
-            continue;
-        }
-
-        findSquares(image, squares);
-        drawSquares(image, squares);
-
-        int c = waitKey();
-        if( (char)c == 27 )
-            break;
+        cout << "Couldn't load " << inputName << endl;
+        return -1;
     }
+    int j = iterations;
+    int64 t_ocl = 0, t_cpp = 0;
+    //warm-ups
+    cout << "warming up ..." << endl;
+    findSquares(image, squares_cpu);
+    findSquares_ocl(image, squares_ocl);
+
+
+#if ACCURACY_CHECK
+    cout << "Checking ocl accuracy ... " << endl;
+    cout << (checkPoints(squares_cpu, squares_ocl) ? "Pass" : "Failed") << endl;
+#endif
+    do
+    {
+        int64 t_start = cv::getTickCount();
+        findSquares(image, squares_cpu);
+        t_cpp += cv::getTickCount() - t_start;
+
+
+        t_start  = cv::getTickCount();
+        findSquares_ocl(image, squares_ocl);
+        t_ocl += cv::getTickCount() - t_start;
+        cout << "run loop: " << j << endl;
+    }
+    while(--j);
+    cout << "cpp average time: " << 1000.0f * (double)t_cpp / getTickFrequency() / iterations << "ms" << endl;
+    cout << "ocl average time: " << 1000.0f * (double)t_ocl / getTickFrequency() / iterations << "ms" << endl;
+
+    Mat result = drawSquaresBoth(image, squares_cpu, squares_ocl);
+    imshow(wndname, result);
+    imwrite(outfile, result);
+    waitKey(0);
 
     return 0;
 }
diff --git a/samples/ocl/stereo_match.cpp b/samples/ocl/stereo_match.cpp
new file mode 100644
index 0000000000..100e62a5ba
--- /dev/null
+++ b/samples/ocl/stereo_match.cpp
@@ -0,0 +1,390 @@
+#include <iostream>
+#include <string>
+#include <sstream>
+#include <iomanip>
+#include <stdexcept>
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/ocl/ocl.hpp"
+#include "opencv2/highgui/highgui.hpp"
+
+using namespace cv;
+using namespace std;
+using namespace ocl;
+
+
+struct App
+{
+    App(CommandLineParser& cmd);
+    void run();
+    void handleKey(char key);
+    void printParams() const;
+
+    void workBegin()
+    {
+        work_begin = getTickCount();
+    }
+    void workEnd()
+    {
+        int64 d = getTickCount() - work_begin;
+        double f = getTickFrequency();
+        work_fps = f / d;
+    }
+    string method_str() const
+    {
+        switch (method)
+        {
+        case BM:
+            return "BM";
+        case BP:
+            return "BP";
+        case CSBP:
+            return "CSBP";
+        }
+        return "";
+    }
+    string text() const
+    {
+        stringstream ss;
+        ss << "(" << method_str() << ") FPS: " << setiosflags(ios::left)
+           << setprecision(4) << work_fps;
+        return ss.str();
+    }
+private:
+    bool running, write_once;
+
+    Mat left_src, right_src;
+    Mat left, right;
+    oclMat d_left, d_right;
+
+    StereoBM_OCL bm;
+    StereoBeliefPropagation bp;
+    StereoConstantSpaceBP csbp;
+
+    int64 work_begin;
+    double work_fps;
+
+    string l_img, r_img;
+    string out_img;
+    enum {BM, BP, CSBP} method;
+    int ndisp; // Max disparity + 1
+    enum {GPU, CPU} type;
+};
+
+int main(int argc, char** argv)
+{
+    const char* keys =
+        "{ h | help     | false                     | print help message }"
+        "{ l | left     |                           | specify left image }"
+        "{ r | right    |                           | specify right image }"
+        "{ m | method   | BM                        | specify match method(BM/BP/CSBP) }"
+        "{ n | ndisp    | 64                        |  specify number of disparity levels }"
+        "{ s | cpu_ocl  | false                     | use cpu or gpu as ocl device to process the image }"
+        "{ o | output   | stereo_match_output.jpg   | specify output path when input is images}";
+    CommandLineParser cmd(argc, argv, keys);
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Avaible options:" << endl;
+        cmd.printMessage();
+        return 0;
+    }
+    try
+    {
+        App app(cmd);
+        int flag = CVCL_DEVICE_TYPE_GPU;
+        if(cmd.get<bool>("s") == true)
+            flag = CVCL_DEVICE_TYPE_CPU;
+
+        vector<Info> info;
+        if(getDevice(info, flag) == 0)
+        {
+            throw runtime_error("Error: Did not find a valid OpenCL device!");
+        }
+        cout << "Device name:" << info[0].DeviceName[0] << endl;
+
+        app.run();
+    }
+    catch (const exception& e)
+    {
+        cout << "error: " << e.what() << endl;
+    }
+    return 0;
+}
+
+App::App(CommandLineParser& cmd)
+    : running(false),method(BM)
+{
+    cout << "stereo_match_ocl sample\n";
+    cout << "\nControls:\n"
+         << "\tesc - exit\n"
+         << "\to - save output image once\n"
+         << "\tp - print current parameters\n"
+         << "\tg - convert source images into gray\n"
+         << "\tm - change stereo match method\n"
+         << "\ts - change Sobel prefiltering flag (for BM only)\n"
+         << "\t1/q - increase/decrease maximum disparity\n"
+         << "\t2/w - increase/decrease window size (for BM only)\n"
+         << "\t3/e - increase/decrease iteration count (for BP and CSBP only)\n"
+         << "\t4/r - increase/decrease level count (for BP and CSBP only)\n";
+    l_img = cmd.get<string>("l");
+    r_img = cmd.get<string>("r");
+    string mstr = cmd.get<string>("m");
+    if(mstr == "BM") method = BM;
+    else if(mstr == "BP") method = BP;
+    else if(mstr == "CSBP") method = CSBP;
+    else cout << "unknown method!\n";
+    ndisp = cmd.get<int>("n");
+    out_img = cmd.get<string>("o");
+    write_once = false;
+}
+
+
+void App::run()
+{
+    // Load images
+    left_src = imread(l_img);
+    right_src = imread(r_img);
+    if (left_src.empty()) throw runtime_error("can't open file \"" + l_img + "\"");
+    if (right_src.empty()) throw runtime_error("can't open file \"" + r_img + "\"");
+
+    cvtColor(left_src, left, COLOR_BGR2GRAY);
+    cvtColor(right_src, right, COLOR_BGR2GRAY);
+
+    d_left.upload(left);
+    d_right.upload(right);
+
+    imshow("left", left);
+    imshow("right", right);
+
+    // Set common parameters
+    bm.ndisp = ndisp;
+    bp.ndisp = ndisp;
+    csbp.ndisp = ndisp;
+
+    cout << endl;
+    printParams();
+
+    running = true;
+    while (running)
+    {
+        // Prepare disparity map of specified type
+        Mat disp;
+        oclMat d_disp;
+        workBegin();
+        switch (method)
+        {
+        case BM:
+            if (d_left.channels() > 1 || d_right.channels() > 1)
+            {
+                cout << "BM doesn't support color images\n";
+                cvtColor(left_src, left, COLOR_BGR2GRAY);
+                cvtColor(right_src, right, COLOR_BGR2GRAY);
+                cout << "image_channels: " << left.channels() << endl;
+                d_left.upload(left);
+                d_right.upload(right);
+                imshow("left", left);
+                imshow("right", right);
+            }
+            bm(d_left, d_right, d_disp);
+            break;
+        case BP:
+            bp(d_left, d_right, d_disp);
+            break;
+        case CSBP:
+            csbp(d_left, d_right, d_disp);
+            break;
+        }
+
+        // Show results
+        d_disp.download(disp);
+        workEnd();
+
+        if (method != BM)
+        {
+            disp.convertTo(disp, 0);
+        }
+        putText(disp, text(), Point(5, 25), FONT_HERSHEY_SIMPLEX, 1.0, Scalar::all(255));
+        imshow("disparity", disp);
+        if(write_once)
+        {
+            imwrite(out_img, disp);
+            write_once = false;
+        }
+        handleKey((char)waitKey(3));
+    }
+}
+
+
+void App::printParams() const
+{
+    cout << "--- Parameters ---\n";
+    cout << "image_size: (" << left.cols << ", " << left.rows << ")\n";
+    cout << "image_channels: " << left.channels() << endl;
+    cout << "method: " << method_str() << endl
+         << "ndisp: " << ndisp << endl;
+    switch (method)
+    {
+    case BM:
+        cout << "win_size: " << bm.winSize << endl;
+        cout << "prefilter_sobel: " << bm.preset << endl;
+        break;
+    case BP:
+        cout << "iter_count: " << bp.iters << endl;
+        cout << "level_count: " << bp.levels << endl;
+        break;
+    case CSBP:
+        cout << "iter_count: " << csbp.iters << endl;
+        cout << "level_count: " << csbp.levels << endl;
+        break;
+    }
+    cout << endl;
+}
+
+
+void App::handleKey(char key)
+{
+    switch (key)
+    {
+    case 27:
+        running = false;
+        break;
+    case 'p':
+    case 'P':
+        printParams();
+        break;
+    case 'g':
+    case 'G':
+        if (left.channels() == 1 && method != BM)
+        {
+            left = left_src;
+            right = right_src;
+        }
+        else
+        {
+            cvtColor(left_src, left, COLOR_BGR2GRAY);
+            cvtColor(right_src, right, COLOR_BGR2GRAY);
+        }
+        d_left.upload(left);
+        d_right.upload(right);
+        cout << "image_channels: " << left.channels() << endl;
+        imshow("left", left);
+        imshow("right", right);
+        break;
+    case 'm':
+    case 'M':
+        switch (method)
+        {
+        case BM:
+            method = BP;
+            break;
+        case BP:
+            method = CSBP;
+            break;
+        case CSBP:
+            method = BM;
+            break;
+        }
+        cout << "method: " << method_str() << endl;
+        break;
+    case 's':
+    case 'S':
+        if (method == BM)
+        {
+            switch (bm.preset)
+            {
+            case StereoBM_OCL::BASIC_PRESET:
+                bm.preset = StereoBM_OCL::PREFILTER_XSOBEL;
+                break;
+            case StereoBM_OCL::PREFILTER_XSOBEL:
+                bm.preset = StereoBM_OCL::BASIC_PRESET;
+                break;
+            }
+            cout << "prefilter_sobel: " << bm.preset << endl;
+        }
+        break;
+    case '1':
+        ndisp == 1 ? ndisp = 8 : ndisp += 8;
+        cout << "ndisp: " << ndisp << endl;
+        bm.ndisp = ndisp;
+        bp.ndisp = ndisp;
+        csbp.ndisp = ndisp;
+        break;
+    case 'q':
+    case 'Q':
+        ndisp = max(ndisp - 8, 1);
+        cout << "ndisp: " << ndisp << endl;
+        bm.ndisp = ndisp;
+        bp.ndisp = ndisp;
+        csbp.ndisp = ndisp;
+        break;
+    case '2':
+        if (method == BM)
+        {
+            bm.winSize = min(bm.winSize + 1, 51);
+            cout << "win_size: " << bm.winSize << endl;
+        }
+        break;
+    case 'w':
+    case 'W':
+        if (method == BM)
+        {
+            bm.winSize = max(bm.winSize - 1, 2);
+            cout << "win_size: " << bm.winSize << endl;
+        }
+        break;
+    case '3':
+        if (method == BP)
+        {
+            bp.iters += 1;
+            cout << "iter_count: " << bp.iters << endl;
+        }
+        else if (method == CSBP)
+        {
+            csbp.iters += 1;
+            cout << "iter_count: " << csbp.iters << endl;
+        }
+        break;
+    case 'e':
+    case 'E':
+        if (method == BP)
+        {
+            bp.iters = max(bp.iters - 1, 1);
+            cout << "iter_count: " << bp.iters << endl;
+        }
+        else if (method == CSBP)
+        {
+            csbp.iters = max(csbp.iters - 1, 1);
+            cout << "iter_count: " << csbp.iters << endl;
+        }
+        break;
+    case '4':
+        if (method == BP)
+        {
+            bp.levels += 1;
+            cout << "level_count: " << bp.levels << endl;
+        }
+        else if (method == CSBP)
+        {
+            csbp.levels += 1;
+            cout << "level_count: " << csbp.levels << endl;
+        }
+        break;
+    case 'r':
+    case 'R':
+        if (method == BP)
+        {
+            bp.levels = max(bp.levels - 1, 1);
+            cout << "level_count: " << bp.levels << endl;
+        }
+        else if (method == CSBP)
+        {
+            csbp.levels = max(csbp.levels - 1, 1);
+            cout << "level_count: " << csbp.levels << endl;
+        }
+        break;
+    case 'o':
+    case 'O':
+        write_once = true;
+        break;
+    }
+}
diff --git a/samples/ocl/surf_matcher.cpp b/samples/ocl/surf_matcher.cpp
index 8462300edc..29619808ae 100644
--- a/samples/ocl/surf_matcher.cpp
+++ b/samples/ocl/surf_matcher.cpp
@@ -1,200 +1,95 @@
-/*M///////////////////////////////////////////////////////////////////////////////////////
-//
-//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
-//
-//  By downloading, copying, installing or using the software you agree to this license.
-//  If you do not agree to this license, do not download, install,
-//  copy or use the software.
-//
-//
-//                           License Agreement
-//                For Open Source Computer Vision Library
-//
-// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
-// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
-// Third party copyrights are property of their respective owners.
-//
-// @Authors
-//    Peng Xiao, pengxiao@multicorewareinc.com
-//
-// Redistribution and use in source and binary forms, with or without modification,
-// are permitted provided that the following conditions are met:
-//
-//   * Redistribution's of source code must retain the above copyright notice,
-//     this list of conditions and the following disclaimer.
-//
-//   * Redistribution's in binary form must reproduce the above copyright notice,
-//     this list of conditions and the following disclaimer in the documentation
-//     and/or other oclMaterials provided with the distribution.
-//
-//   * The name of the copyright holders may not be used to endorse or promote products
-//     derived from this software without specific prior written permission.
-//
-// This software is provided by the copyright holders and contributors as is and
-// any express or implied warranties, including, but not limited to, the implied
-// warranties of merchantability and fitness for a particular purpose are disclaimed.
-// In no event shall the Intel Corporation or contributors be liable for any direct,
-// indirect, incidental, special, exemplary, or consequential damages
-// (including, but not limited to, procurement of substitute goods or services;
-// loss of use, data, or profits; or business interruption) however caused
-// and on any theory of liability, whether in contract, strict liability,
-// or tort (including negligence or otherwise) arising in any way out of
-// the use of this software, even if advised of the possibility of such damage.
-//
-//M*/
-
 #include <iostream>
 #include <stdio.h>
 #include "opencv2/core/core.hpp"
-#include "opencv2/features2d/features2d.hpp"
+#include "opencv2/core/utility.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/ocl/ocl.hpp"
-#include "opencv2/nonfree/nonfree.hpp"
+#include "opencv2/nonfree/ocl.hpp"
 #include "opencv2/calib3d/calib3d.hpp"
+#include "opencv2/nonfree/nonfree.hpp"
 
-using namespace std;
 using namespace cv;
 using namespace cv::ocl;
 
-//#define USE_CPU_DESCRIPTOR // use cpu descriptor extractor until ocl descriptor extractor is fixed
-//#define USE_CPU_BFMATCHER
-void help();
+const int LOOP_NUM = 10;
+const int GOOD_PTS_MAX = 50;
+const float GOOD_PORTION = 0.15f;
 
-void help()
+namespace
 {
-    cout << "\nThis program demonstrates using SURF_OCL features detector and descriptor extractor" << endl;
-    cout << "\nUsage:\n\tsurf_matcher --left <image1> --right <image2>" << endl;
+
+int64 work_begin = 0;
+int64 work_end = 0;
+
+void workBegin()
+{
+    work_begin = getTickCount();
+}
+void workEnd()
+{
+    work_end = getTickCount() - work_begin;
+}
+double getTime()
+{
+    return work_end /((double)getTickFrequency() * 1000.);
 }
 
-
-////////////////////////////////////////////////////
-// This program demonstrates the usage of SURF_OCL.
-// use cpu findHomography interface to calculate the transformation matrix
-int main(int argc, char* argv[])
+template<class KPDetector>
+struct SURFDetector
 {
-    if (argc != 5 && argc != 1)
+    KPDetector surf;
+    SURFDetector(double hessian = 800.0)
+        :surf(hessian)
     {
-        help();
-        return -1;
     }
-    vector<cv::ocl::Info> info;
-    if(!cv::ocl::getDevice(info))
+    template<class T>
+    void operator()(const T& in, const T& mask, std::vector<cv::KeyPoint>& pts, T& descriptors, bool useProvided = false)
     {
-        cout << "Error: Did not find a valid OpenCL device!" << endl;
-        return -1;
+        surf(in, mask, pts, descriptors, useProvided);
     }
-    Mat cpu_img1, cpu_img2, cpu_img1_grey, cpu_img2_grey;
-    oclMat img1, img2;
-    if(argc != 5)
+};
+
+template<class KPMatcher>
+struct SURFMatcher
+{
+    KPMatcher matcher;
+    template<class T>
+    void match(const T& in1, const T& in2, std::vector<cv::DMatch>& matches)
     {
-        cpu_img1 = imread("o.png");
-        cvtColor(cpu_img1, cpu_img1_grey, CV_BGR2GRAY);
-        img1     = cpu_img1_grey;
-        CV_Assert(!img1.empty());
-
-        cpu_img2 = imread("r2.png");
-        cvtColor(cpu_img2, cpu_img2_grey, CV_BGR2GRAY);
-        img2     = cpu_img2_grey;
-    }
-    else
-    {
-        for (int i = 1; i < argc; ++i)
-        {
-            if (string(argv[i]) == "--left")
-            {
-                cpu_img1 = imread(argv[++i]);
-                cvtColor(cpu_img1, cpu_img1_grey, CV_BGR2GRAY);
-                img1     = cpu_img1_grey;
-                CV_Assert(!img1.empty());
-            }
-            else if (string(argv[i]) == "--right")
-            {
-                cpu_img2 = imread(argv[++i]);
-                cvtColor(cpu_img2, cpu_img2_grey, CV_BGR2GRAY);
-                img2     = cpu_img2_grey;
-            }
-            else if (string(argv[i]) == "--help")
-            {
-                help();
-                return -1;
-            }
-        }
+        matcher.match(in1, in2, matches);
     }
+};
 
-    SURF_OCL surf;
-    //surf.hessianThreshold = 400.f;
-    //surf.extended = false;
-
-    // detecting keypoints & computing descriptors
-    oclMat keypoints1GPU, keypoints2GPU;
-    oclMat descriptors1GPU, descriptors2GPU;
-
-    // downloading results
-    vector<KeyPoint> keypoints1, keypoints2;
-    vector<DMatch> matches;
-
-
-#ifndef USE_CPU_DESCRIPTOR
-    surf(img1, oclMat(), keypoints1GPU, descriptors1GPU);
-    surf(img2, oclMat(), keypoints2GPU, descriptors2GPU);
-
-    surf.downloadKeypoints(keypoints1GPU, keypoints1);
-    surf.downloadKeypoints(keypoints2GPU, keypoints2);
-
-
-#ifdef USE_CPU_BFMATCHER
-    //BFMatcher
-    BFMatcher matcher(cv::NORM_L2);
-    matcher.match(Mat(descriptors1GPU), Mat(descriptors2GPU), matches);
-#else
-    BruteForceMatcher_OCL_base matcher(BruteForceMatcher_OCL_base::L2Dist);
-    matcher.match(descriptors1GPU, descriptors2GPU, matches);
-#endif
-
-#else
-    surf(img1, oclMat(), keypoints1GPU);
-    surf(img2, oclMat(), keypoints2GPU);
-    surf.downloadKeypoints(keypoints1GPU, keypoints1);
-    surf.downloadKeypoints(keypoints2GPU, keypoints2);
-
-    // use SURF_OCL to detect keypoints and use SURF to extract descriptors
-    SURF surf_cpu;
-    Mat descriptors1, descriptors2;
-    surf_cpu(cpu_img1, Mat(), keypoints1, descriptors1, true);
-    surf_cpu(cpu_img2, Mat(), keypoints2, descriptors2, true);
-    matcher.match(descriptors1, descriptors2, matches);
-#endif
-    cout << "OCL: FOUND " << keypoints1GPU.cols << " keypoints on first image" << endl;
-    cout << "OCL: FOUND " << keypoints2GPU.cols << " keypoints on second image" << endl;
-
-    double max_dist = 0; double min_dist = 100;
-    //-- Quick calculation of max and min distances between keypoints
-    for( size_t i = 0; i < keypoints1.size(); i++ )
-    {
-        double dist = matches[i].distance;
-        if( dist < min_dist ) min_dist = dist;
-        if( dist > max_dist ) max_dist = dist;
-    }
-
-    printf("-- Max dist : %f \n", max_dist );
-    printf("-- Min dist : %f \n", min_dist );
-
-    //-- Draw only "good" matches (i.e. whose distance is less than 2.5*min_dist )
+Mat drawGoodMatches(
+    const Mat& cpu_img1,
+    const Mat& cpu_img2,
+    const std::vector<KeyPoint>& keypoints1,
+    const std::vector<KeyPoint>& keypoints2,
+    std::vector<DMatch>& matches,
+    std::vector<Point2f>& scene_corners_
+    )
+{
+    //-- Sort matches and preserve top 10% matches
+    std::sort(matches.begin(), matches.end());
     std::vector< DMatch > good_matches;
+    double minDist = matches.front().distance,
+           maxDist = matches.back().distance;
 
-    for( size_t i = 0; i < keypoints1.size(); i++ )
+    const int ptsPairs = std::min(GOOD_PTS_MAX, (int)(matches.size() * GOOD_PORTION));
+    for( int i = 0; i < ptsPairs; i++ )
     {
-        if( matches[i].distance < 3*min_dist )
-        {
-            good_matches.push_back( matches[i]);
-        }
+        good_matches.push_back( matches[i] );
     }
+    std::cout << "\nMax distance: " << maxDist << std::endl;
+    std::cout << "Min distance: " << minDist << std::endl;
+
+    std::cout << "Calculating homography using " << ptsPairs << " point pairs." << std::endl;
 
     // drawing the results
     Mat img_matches;
     drawMatches( cpu_img1, keypoints1, cpu_img2, keypoints2,
-        good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
-        vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
+                 good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
+                 std::vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS  );
 
     //-- Localize the object
     std::vector<Point2f> obj;
@@ -206,26 +101,243 @@ int main(int argc, char* argv[])
         obj.push_back( keypoints1[ good_matches[i].queryIdx ].pt );
         scene.push_back( keypoints2[ good_matches[i].trainIdx ].pt );
     }
-    Mat H = findHomography( obj, scene, CV_RANSAC );
-
     //-- Get the corners from the image_1 ( the object to be "detected" )
     std::vector<Point2f> obj_corners(4);
-    obj_corners[0] = cvPoint(0,0); obj_corners[1] = cvPoint( cpu_img1.cols, 0 );
-    obj_corners[2] = cvPoint( cpu_img1.cols, cpu_img1.rows ); obj_corners[3] = cvPoint( 0, cpu_img1.rows );
+    obj_corners[0] = Point(0,0);
+    obj_corners[1] = Point( cpu_img1.cols, 0 );
+    obj_corners[2] = Point( cpu_img1.cols, cpu_img1.rows );
+    obj_corners[3] = Point( 0, cpu_img1.rows );
     std::vector<Point2f> scene_corners(4);
 
+    Mat H = findHomography( obj, scene, RANSAC );
     perspectiveTransform( obj_corners, scene_corners, H);
 
+    scene_corners_ = scene_corners;
+
     //-- Draw lines between the corners (the mapped object in the scene - image_2 )
-    line( img_matches, scene_corners[0] + Point2f( (float)cpu_img1.cols, 0), scene_corners[1] + Point2f( (float)cpu_img1.cols, 0), Scalar( 0, 255, 0), 4 );
-    line( img_matches, scene_corners[1] + Point2f( (float)cpu_img1.cols, 0), scene_corners[2] + Point2f( (float)cpu_img1.cols, 0), Scalar( 0, 255, 0), 4 );
-    line( img_matches, scene_corners[2] + Point2f( (float)cpu_img1.cols, 0), scene_corners[3] + Point2f( (float)cpu_img1.cols, 0), Scalar( 0, 255, 0), 4 );
-    line( img_matches, scene_corners[3] + Point2f( (float)cpu_img1.cols, 0), scene_corners[0] + Point2f( (float)cpu_img1.cols, 0), Scalar( 0, 255, 0), 4 );
+    line( img_matches,
+          scene_corners[0] + Point2f( (float)cpu_img1.cols, 0), scene_corners[1] + Point2f( (float)cpu_img1.cols, 0),
+          Scalar( 0, 255, 0), 2, LINE_AA );
+    line( img_matches,
+          scene_corners[1] + Point2f( (float)cpu_img1.cols, 0), scene_corners[2] + Point2f( (float)cpu_img1.cols, 0),
+          Scalar( 0, 255, 0), 2, LINE_AA );
+    line( img_matches,
+          scene_corners[2] + Point2f( (float)cpu_img1.cols, 0), scene_corners[3] + Point2f( (float)cpu_img1.cols, 0),
+          Scalar( 0, 255, 0), 2, LINE_AA );
+    line( img_matches,
+          scene_corners[3] + Point2f( (float)cpu_img1.cols, 0), scene_corners[0] + Point2f( (float)cpu_img1.cols, 0),
+          Scalar( 0, 255, 0), 2, LINE_AA );
+    return img_matches;
+}
+
+}
+////////////////////////////////////////////////////
+// This program demonstrates the usage of SURF_OCL.
+// use cpu findHomography interface to calculate the transformation matrix
+int main(int argc, char* argv[])
+{
+    const char* keys =
+        "{ help h    | false           | print help message  }"
+        "{ left l    |                 | specify left image  }"
+        "{ right r   |                 | specify right image }"
+        "{ output o  | SURF_output.jpg | specify output save path (only works in CPU or GPU only mode) }"
+        "{ use_cpu c | false           | use CPU algorithms  }"
+        "{ use_all a | false           | use both CPU and GPU algorithms}";
+    CommandLineParser cmd(argc, argv, keys);
+    if (cmd.get<bool>("help"))
+    {
+        std::cout << "Avaible options:" << std::endl;
+        cmd.printMessage();
+        return 0;
+    }
+
+    std::vector<cv::ocl::Info> info;
+    if(cv::ocl::getDevice(info) == 0)
+    {
+        std::cout << "Error: Did not find a valid OpenCL device!" << std::endl;
+        return -1;
+    }
+    ocl::setDevice(info[0]);
+
+    Mat cpu_img1, cpu_img2, cpu_img1_grey, cpu_img2_grey;
+    oclMat img1, img2;
+    bool useCPU = cmd.get<bool>("c");
+    bool useGPU = false;
+    bool useALL = cmd.get<bool>("a");
+
+    std::string outpath = cmd.get<std::string>("o");
+
+    cpu_img1 = imread(cmd.get<std::string>("l"));
+    CV_Assert(!cpu_img1.empty());
+    cvtColor(cpu_img1, cpu_img1_grey, COLOR_BGR2GRAY);
+    img1 = cpu_img1_grey;
+
+    cpu_img2 = imread(cmd.get<std::string>("r"));
+    CV_Assert(!cpu_img2.empty());
+    cvtColor(cpu_img2, cpu_img2_grey, COLOR_BGR2GRAY);
+    img2 = cpu_img2_grey;
+
+    if(useALL)
+    {
+        useCPU = false;
+        useGPU = false;
+    }
+    else if(useCPU==false && useALL==false)
+    {
+        useGPU = true;
+    }
+
+    if(!useCPU)
+    {
+        std::cout
+                << "Device name:"
+                << info[0].DeviceName[0]
+                << std::endl;
+    }
+    double surf_time = 0.;
+
+    //declare input/output
+    std::vector<KeyPoint> keypoints1, keypoints2;
+    std::vector<DMatch> matches;
+
+    std::vector<KeyPoint> gpu_keypoints1;
+    std::vector<KeyPoint> gpu_keypoints2;
+    std::vector<DMatch> gpu_matches;
+
+    Mat descriptors1CPU, descriptors2CPU;
+
+    oclMat keypoints1GPU, keypoints2GPU;
+    oclMat descriptors1GPU, descriptors2GPU;
+
+    //instantiate detectors/matchers
+    SURFDetector<SURF>     cpp_surf;
+    SURFDetector<SURF_OCL> ocl_surf;
+
+    SURFMatcher<BFMatcher>      cpp_matcher;
+    SURFMatcher<BFMatcher_OCL>  ocl_matcher;
+
+    //-- start of timing section
+    if (useCPU)
+    {
+        for (int i = 0; i <= LOOP_NUM; i++)
+        {
+            if(i == 1) workBegin();
+            cpp_surf(cpu_img1_grey, Mat(), keypoints1, descriptors1CPU);
+            cpp_surf(cpu_img2_grey, Mat(), keypoints2, descriptors2CPU);
+            cpp_matcher.match(descriptors1CPU, descriptors2CPU, matches);
+        }
+        workEnd();
+        std::cout << "CPP: FOUND " << keypoints1.size() << " keypoints on first image" << std::endl;
+        std::cout << "CPP: FOUND " << keypoints2.size() << " keypoints on second image" << std::endl;
+
+        surf_time = getTime();
+        std::cout << "SURF run time: " << surf_time / LOOP_NUM << " ms" << std::endl<<"\n";
+    }
+    else if(useGPU)
+    {
+        for (int i = 0; i <= LOOP_NUM; i++)
+        {
+            if(i == 1) workBegin();
+            ocl_surf(img1, oclMat(), keypoints1, descriptors1GPU);
+            ocl_surf(img2, oclMat(), keypoints2, descriptors2GPU);
+            ocl_matcher.match(descriptors1GPU, descriptors2GPU, matches);
+        }
+        workEnd();
+        std::cout << "OCL: FOUND " << keypoints1.size() << " keypoints on first image" << std::endl;
+        std::cout << "OCL: FOUND " << keypoints2.size() << " keypoints on second image" << std::endl;
+
+        surf_time = getTime();
+        std::cout << "SURF run time: " << surf_time / LOOP_NUM << " ms" << std::endl<<"\n";
+    }
+    else
+    {
+        //cpu runs
+        for (int i = 0; i <= LOOP_NUM; i++)
+        {
+            if(i == 1) workBegin();
+            cpp_surf(cpu_img1_grey, Mat(), keypoints1, descriptors1CPU);
+            cpp_surf(cpu_img2_grey, Mat(), keypoints2, descriptors2CPU);
+            cpp_matcher.match(descriptors1CPU, descriptors2CPU, matches);
+        }
+        workEnd();
+        std::cout << "\nCPP: FOUND " << keypoints1.size() << " keypoints on first image" << std::endl;
+        std::cout << "CPP: FOUND " << keypoints2.size() << " keypoints on second image" << std::endl;
+
+        surf_time = getTime();
+        std::cout << "(CPP)SURF run time: " << surf_time / LOOP_NUM << " ms" << std::endl;
+
+        //gpu runs
+        for (int i = 0; i <= LOOP_NUM; i++)
+        {
+            if(i == 1) workBegin();
+            ocl_surf(img1, oclMat(), gpu_keypoints1, descriptors1GPU);
+            ocl_surf(img2, oclMat(), gpu_keypoints2, descriptors2GPU);
+            ocl_matcher.match(descriptors1GPU, descriptors2GPU, gpu_matches);
+        }
+        workEnd();
+        std::cout << "\nOCL: FOUND " << keypoints1.size() << " keypoints on first image" << std::endl;
+        std::cout << "OCL: FOUND " << keypoints2.size() << " keypoints on second image" << std::endl;
+
+        surf_time = getTime();
+        std::cout << "(OCL)SURF run time: " << surf_time / LOOP_NUM << " ms" << std::endl<<"\n";
+
+    }
+
+    //--------------------------------------------------------------------------
+    std::vector<Point2f> cpu_corner;
+    Mat img_matches = drawGoodMatches(cpu_img1, cpu_img2, keypoints1, keypoints2, matches, cpu_corner);
+
+    std::vector<Point2f> gpu_corner;
+    Mat ocl_img_matches;
+    if(useALL || (!useCPU&&!useGPU))
+    {
+        ocl_img_matches = drawGoodMatches(cpu_img1, cpu_img2, gpu_keypoints1, gpu_keypoints2, gpu_matches, gpu_corner);
+
+        //check accuracy
+        std::cout<<"\nCheck accuracy:\n";
+
+        if(cpu_corner.size()!=gpu_corner.size())
+            std::cout<<"Failed\n";
+        else
+        {
+            bool result = false;
+            for(size_t i = 0; i < cpu_corner.size(); i++)
+            {
+                if((std::abs(cpu_corner[i].x - gpu_corner[i].x) > 10)
+                        ||(std::abs(cpu_corner[i].y - gpu_corner[i].y) > 10))
+                {
+                    std::cout<<"Failed\n";
+                    result = false;
+                    break;
+                }
+                result = true;
+            }
+            if(result)
+                std::cout<<"Passed\n";
+        }
+    }
 
     //-- Show detected matches
-    namedWindow("ocl surf matches", 0);
-    imshow("ocl surf matches", img_matches);
-    waitKey(0);
+    if (useCPU)
+    {
+        namedWindow("cpu surf matches", 0);
+        imshow("cpu surf matches", img_matches);
+        imwrite(outpath, img_matches);
+    }
+    else if(useGPU)
+    {
+        namedWindow("ocl surf matches", 0);
+        imshow("ocl surf matches", img_matches);
+        imwrite(outpath, img_matches);
+    }
+    else
+    {
+        namedWindow("cpu surf matches", 0);
+        imshow("cpu surf matches", img_matches);
 
+        namedWindow("ocl surf matches", 0);
+        imshow("ocl surf matches", ocl_img_matches);
+    }
+    waitKey(0);
     return 0;
 }
diff --git a/samples/ocl/tvl1_optical_flow.cpp b/samples/ocl/tvl1_optical_flow.cpp
new file mode 100644
index 0000000000..478979d956
--- /dev/null
+++ b/samples/ocl/tvl1_optical_flow.cpp
@@ -0,0 +1,264 @@
+#include <iostream>
+#include <vector>
+#include <iomanip>
+
+#include "opencv2/core/utility.hpp"
+#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/ocl/ocl.hpp"
+#include "opencv2/video/video.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+typedef unsigned char uchar;
+#define LOOP_NUM 10
+int64 work_begin = 0;
+int64 work_end = 0;
+
+static void workBegin()
+{
+    work_begin = getTickCount();
+}
+static void workEnd()
+{
+    work_end += (getTickCount() - work_begin);
+}
+static double getTime()
+{
+    return work_end * 1000. / getTickFrequency();
+}
+
+template <typename T> inline T clamp (T x, T a, T b)
+{
+    return ((x) > (a) ? ((x) < (b) ? (x) : (b)) : (a));
+}
+
+template <typename T> inline T mapValue(T x, T a, T b, T c, T d)
+{
+    x = clamp(x, a, b);
+    return c + (d - c) * (x - a) / (b - a);
+}
+
+static void getFlowField(const Mat& u, const Mat& v, Mat& flowField)
+{
+    float maxDisplacement = 1.0f;
+
+    for (int i = 0; i < u.rows; ++i)
+    {
+        const float* ptr_u = u.ptr<float>(i);
+        const float* ptr_v = v.ptr<float>(i);
+
+        for (int j = 0; j < u.cols; ++j)
+        {
+            float d = max(fabsf(ptr_u[j]), fabsf(ptr_v[j]));
+
+            if (d > maxDisplacement)
+                maxDisplacement = d;
+        }
+    }
+
+    flowField.create(u.size(), CV_8UC4);
+
+    for (int i = 0; i < flowField.rows; ++i)
+    {
+        const float* ptr_u = u.ptr<float>(i);
+        const float* ptr_v = v.ptr<float>(i);
+
+
+        Vec4b* row = flowField.ptr<Vec4b>(i);
+
+        for (int j = 0; j < flowField.cols; ++j)
+        {
+            row[j][0] = 0;
+            row[j][1] = static_cast<unsigned char> (mapValue (-ptr_v[j], -maxDisplacement, maxDisplacement, 0.0f, 255.0f));
+            row[j][2] = static_cast<unsigned char> (mapValue ( ptr_u[j], -maxDisplacement, maxDisplacement, 0.0f, 255.0f));
+            row[j][3] = 255;
+        }
+    }
+}
+
+
+int main(int argc, const char* argv[])
+{
+    static std::vector<Info> ocl_info;
+    ocl::getDevice(ocl_info);
+    //if you want to use undefault device, set it here
+    setDevice(ocl_info[0]);
+
+    //set this to save kernel compile time from second time you run
+    ocl::setBinpath("./");
+    const char* keys =
+        "{ h   | help       | false           | print help message }"
+        "{ l   | left       |                 | specify left image }"
+        "{ r   | right      |                 | specify right image }"
+        "{ o   | output     | tvl1_output.jpg | specify output save path }"
+        "{ c   | camera     | 0               | enable camera capturing }"
+        "{ s   | use_cpu    | false           | use cpu or gpu to process the image }"
+        "{ v   | video      |                 | use video as input }";
+
+    CommandLineParser cmd(argc, argv, keys);
+
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Usage: pyrlk_optical_flow [options]" << endl;
+        cout << "Avaible options:" << endl;
+        cmd.printMessage();
+        return 0;
+    }
+
+    bool defaultPicturesFail = false;
+    string fname0 = cmd.get<string>("l");
+    string fname1 = cmd.get<string>("r");
+    string vdofile = cmd.get<string>("v");
+    string outpath = cmd.get<string>("o");
+    bool useCPU = cmd.get<bool>("s");
+    bool useCamera = cmd.get<bool>("c");
+    int inputName = cmd.get<int>("c");
+
+    Mat frame0 = imread(fname0, cv::IMREAD_GRAYSCALE);
+    Mat frame1 = imread(fname1, cv::IMREAD_GRAYSCALE);
+    cv::Ptr<cv::DenseOpticalFlow> alg = cv::createOptFlow_DualTVL1();
+    cv::ocl::OpticalFlowDual_TVL1_OCL d_alg;
+
+
+    Mat flow, show_flow;
+    Mat flow_vec[2];
+    if (frame0.empty() || frame1.empty())
+    {
+        useCamera = true;
+        defaultPicturesFail = true;
+        VideoCapture capture( inputName );
+        if (!capture.isOpened())
+        {
+            cout << "Can't load input images" << endl;
+            return -1;
+        }
+    }
+
+
+    if (useCamera)
+    {
+        VideoCapture capture;
+        Mat frame, frameCopy;
+        Mat frame0Gray, frame1Gray;
+        Mat ptr0, ptr1;
+
+        if(vdofile == "")
+            capture.open( inputName );
+        else
+            capture.open(vdofile.c_str());
+
+        int c = inputName ;
+        if(!capture.isOpened())
+        {
+            if(vdofile == "")
+                cout << "Capture from CAM " << c << " didn't work" << endl;
+            else
+                cout << "Capture from file " << vdofile << " failed" <<endl;
+            if (defaultPicturesFail)
+            {
+                return -1;
+            }
+            goto nocamera;
+        }
+
+        cout << "In capture ..." << endl;
+        for(int i = 0;; i++)
+        {
+            if( !capture.read(frame) )
+                break;
+
+            if (i == 0)
+            {
+                frame.copyTo( frame0 );
+                cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+            }
+            else
+            {
+                if (i%2 == 1)
+                {
+                    frame.copyTo(frame1);
+                    cvtColor(frame1, frame1Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame0Gray;
+                    ptr1 = frame1Gray;
+                }
+                else
+                {
+                    frame.copyTo(frame0);
+                    cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame1Gray;
+                    ptr1 = frame0Gray;
+                }
+
+                if (useCPU)
+                {
+                    alg->calc(ptr0, ptr1, flow);
+                    split(flow, flow_vec);
+                }
+                else
+                {
+                    oclMat d_flowx, d_flowy;
+                    d_alg(oclMat(ptr0), oclMat(ptr1), d_flowx, d_flowy);
+                    d_flowx.download(flow_vec[0]);
+                    d_flowy.download(flow_vec[1]);
+                }
+                if (i%2 == 1)
+                    frame1.copyTo(frameCopy);
+                else
+                    frame0.copyTo(frameCopy);
+                getFlowField(flow_vec[0], flow_vec[1], show_flow);
+                imshow("PyrLK [Sparse]", show_flow);
+            }
+
+            if( waitKey( 10 ) >= 0 )
+                goto _cleanup_;
+        }
+
+        waitKey(0);
+
+_cleanup_:
+        capture.release();
+    }
+    else
+    {
+nocamera:
+        oclMat d_flowx, d_flowy;
+        for(int i = 0; i <= LOOP_NUM; i ++)
+        {
+            cout << "loop" << i << endl;
+
+            if (i > 0) workBegin();
+            if (useCPU)
+            {
+                alg->calc(frame0, frame1, flow);
+                split(flow, flow_vec);
+            }
+            else
+            {
+                d_alg(oclMat(frame0), oclMat(frame1), d_flowx, d_flowy);
+                d_flowx.download(flow_vec[0]);
+                d_flowy.download(flow_vec[1]);
+            }
+            if (i > 0 && i <= LOOP_NUM)
+                workEnd();
+
+            if (i == LOOP_NUM)
+            {
+                if (useCPU)
+                    cout << "average CPU time (noCamera) : ";
+                else
+                    cout << "average GPU time (noCamera) : ";
+                cout << getTime() / LOOP_NUM << " ms" << endl;
+
+                getFlowField(flow_vec[0], flow_vec[1], show_flow);
+                imshow("PyrLK [Sparse]", show_flow);
+                imwrite(outpath, show_flow);
+            }
+        }
+    }
+
+    waitKey();
+
+    return 0;
+}
diff --git a/samples/python/camera.py b/samples/python/camera.py
deleted file mode 100755
index 5d1c63b047..0000000000
--- a/samples/python/camera.py
+++ /dev/null
@@ -1,15 +0,0 @@
-#!/usr/bin/python
-
-import cv2.cv as cv
-import time
-
-cv.NamedWindow("camera", 1)
-
-capture = cv.CaptureFromCAM(0)
-
-while True:
-    img = cv.QueryFrame(capture)
-    cv.ShowImage("camera", img)
-    if cv.WaitKey(10) == 27:
-        break
-cv.DestroyAllWindows()
diff --git a/samples/python/camshift.py b/samples/python/camshift.py
deleted file mode 100755
index 3a08b04d69..0000000000
--- a/samples/python/camshift.py
+++ /dev/null
@@ -1,116 +0,0 @@
-#!/usr/bin/env python
-
-import cv2.cv as cv
-
-def is_rect_nonzero(r):
-    (_,_,w,h) = r
-    return (w > 0) and (h > 0)
-
-class CamShiftDemo:
-
-    def __init__(self):
-        self.capture = cv.CaptureFromCAM(0)
-        cv.NamedWindow( "CamShiftDemo", 1 )
-        cv.NamedWindow( "Histogram", 1 )
-        cv.SetMouseCallback( "CamShiftDemo", self.on_mouse)
-
-        self.drag_start = None      # Set to (x,y) when mouse starts drag
-        self.track_window = None    # Set to rect when the mouse drag finishes
-
-        print( "Keys:\n"
-            "    ESC - quit the program\n"
-            "    b - switch to/from backprojection view\n"
-            "To initialize tracking, drag across the object with the mouse\n" )
-
-    def hue_histogram_as_image(self, hist):
-        """ Returns a nice representation of a hue histogram """
-
-        histimg_hsv = cv.CreateImage( (320,200), 8, 3)
-
-        mybins = cv.CloneMatND(hist.bins)
-        cv.Log(mybins, mybins)
-        (_, hi, _, _) = cv.MinMaxLoc(mybins)
-        cv.ConvertScale(mybins, mybins, 255. / hi)
-
-        w,h = cv.GetSize(histimg_hsv)
-        hdims = cv.GetDims(mybins)[0]
-        for x in range(w):
-            xh = (180 * x) / (w - 1)  # hue sweeps from 0-180 across the image
-            val = int(mybins[int(hdims * x / w)] * h / 255)
-            cv.Rectangle( histimg_hsv, (x, 0), (x, h-val), (xh,255,64), -1)
-            cv.Rectangle( histimg_hsv, (x, h-val), (x, h), (xh,255,255), -1)
-
-        histimg = cv.CreateImage( (320,200), 8, 3)
-        cv.CvtColor(histimg_hsv, histimg, cv.CV_HSV2BGR)
-        return histimg
-
-    def on_mouse(self, event, x, y, flags, param):
-        if event == cv.CV_EVENT_LBUTTONDOWN:
-            self.drag_start = (x, y)
-        if event == cv.CV_EVENT_LBUTTONUP:
-            self.drag_start = None
-            self.track_window = self.selection
-        if self.drag_start:
-            xmin = min(x, self.drag_start[0])
-            ymin = min(y, self.drag_start[1])
-            xmax = max(x, self.drag_start[0])
-            ymax = max(y, self.drag_start[1])
-            self.selection = (xmin, ymin, xmax - xmin, ymax - ymin)
-
-    def run(self):
-        hist = cv.CreateHist([180], cv.CV_HIST_ARRAY, [(0,180)], 1 )
-        backproject_mode = False
-        while True:
-            frame = cv.QueryFrame( self.capture )
-
-            # Convert to HSV and keep the hue
-            hsv = cv.CreateImage(cv.GetSize(frame), 8, 3)
-            cv.CvtColor(frame, hsv, cv.CV_BGR2HSV)
-            self.hue = cv.CreateImage(cv.GetSize(frame), 8, 1)
-            cv.Split(hsv, self.hue, None, None, None)
-
-            # Compute back projection
-            backproject = cv.CreateImage(cv.GetSize(frame), 8, 1)
-
-            # Run the cam-shift
-            cv.CalcArrBackProject( [self.hue], backproject, hist )
-            if self.track_window and is_rect_nonzero(self.track_window):
-                crit = ( cv.CV_TERMCRIT_EPS | cv.CV_TERMCRIT_ITER, 10, 1)
-                (iters, (area, value, rect), track_box) = cv.CamShift(backproject, self.track_window, crit)
-                self.track_window = rect
-
-            # If mouse is pressed, highlight the current selected rectangle
-            # and recompute the histogram
-
-            if self.drag_start and is_rect_nonzero(self.selection):
-                sub = cv.GetSubRect(frame, self.selection)
-                save = cv.CloneMat(sub)
-                cv.ConvertScale(frame, frame, 0.5)
-                cv.Copy(save, sub)
-                x,y,w,h = self.selection
-                cv.Rectangle(frame, (x,y), (x+w,y+h), (255,255,255))
-
-                sel = cv.GetSubRect(self.hue, self.selection )
-                cv.CalcArrHist( [sel], hist, 0)
-                (_, max_val, _, _) = cv.GetMinMaxHistValue( hist)
-                if max_val != 0:
-                    cv.ConvertScale(hist.bins, hist.bins, 255. / max_val)
-            elif self.track_window and is_rect_nonzero(self.track_window):
-                cv.EllipseBox( frame, track_box, cv.CV_RGB(255,0,0), 3, cv.CV_AA, 0 )
-
-            if not backproject_mode:
-                cv.ShowImage( "CamShiftDemo", frame )
-            else:
-                cv.ShowImage( "CamShiftDemo", backproject)
-            cv.ShowImage( "Histogram", self.hue_histogram_as_image(hist))
-
-            c = cv.WaitKey(7) % 0x100
-            if c == 27:
-                break
-            elif c == ord("b"):
-                backproject_mode = not backproject_mode
-
-if __name__=="__main__":
-    demo = CamShiftDemo()
-    demo.run()
-    cv.DestroyAllWindows()
diff --git a/samples/python/chessboard.py b/samples/python/chessboard.py
deleted file mode 100755
index ce80431db5..0000000000
--- a/samples/python/chessboard.py
+++ /dev/null
@@ -1,34 +0,0 @@
-#!/usr/bin/python
-import cv2.cv as cv
-import sys
-import urllib2
-
-if __name__ == "__main__":
-    cv.NamedWindow("win")
-    if len(sys.argv) > 1:
-        filename = sys.argv[1]
-        im = cv.LoadImage(filename, cv.CV_LOAD_IMAGE_GRAYSCALE)
-        im3 = cv.LoadImage(filename, cv.CV_LOAD_IMAGE_COLOR)
-    else:
-       try: # try opening local copy of image
-            fileName = '../cpp/left01.jpg'
-            im = cv.LoadImageM(fileName, False)
-            im3 = cv.LoadImageM(fileName, True)
-       except: # if local copy cannot be opened, try downloading it
-            url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/cpp/left01.jpg'
-            filedata = urllib2.urlopen(url).read()
-            imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-            cv.SetData(imagefiledata, filedata, len(filedata))
-            im = cv.DecodeImageM(imagefiledata, cv.CV_LOAD_IMAGE_GRAYSCALE)
-            im3 = cv.DecodeImageM(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    chessboard_dim = ( 9, 6 )
-
-    found_all, corners = cv.FindChessboardCorners( im, chessboard_dim )
-    print found_all, len(corners)
-
-    cv.DrawChessboardCorners( im3, chessboard_dim, corners, found_all )
-
-    cv.ShowImage("win", im3);
-    cv.WaitKey()
-    cv.DestroyAllWindows()
diff --git a/samples/python/contours.py b/samples/python/contours.py
deleted file mode 100755
index d6058907e6..0000000000
--- a/samples/python/contours.py
+++ /dev/null
@@ -1,136 +0,0 @@
-#! /usr/bin/env python
-
-print "OpenCV Python version of contours"
-
-# import the necessary things for OpenCV
-import cv2.cv as cv
-
-# some default constants
-_SIZE = 500
-_DEFAULT_LEVEL = 3
-
-# definition of some colors
-_red =  (0, 0, 255, 0);
-_green =  (0, 255, 0, 0);
-_white = cv.RealScalar (255)
-_black = cv.RealScalar (0)
-
-# the callback on the trackbar, to set the level of contours we want
-# to display
-def on_trackbar (position):
-
-    # create the image for putting in it the founded contours
-    contours_image = cv.CreateImage ( (_SIZE, _SIZE), 8, 3)
-
-    # compute the real level of display, given the current position
-    levels = position - 3
-
-    # initialisation
-    _contours = contours
-
-    if levels <= 0:
-        # zero or negative value
-        # => get to the nearest face to make it look more funny
-        _contours = contours.h_next().h_next().h_next()
-
-    # first, clear the image where we will draw contours
-    cv.SetZero (contours_image)
-
-    # draw contours in red and green
-    cv.DrawContours (contours_image, _contours,
-                       _red, _green,
-                       levels, 3, cv.CV_AA,
-                        (0, 0))
-
-    # finally, show the image
-    cv.ShowImage ("contours", contours_image)
-
-if __name__ == '__main__':
-
-    # create the image where we want to display results
-    image = cv.CreateImage ( (_SIZE, _SIZE), 8, 1)
-
-    # start with an empty image
-    cv.SetZero (image)
-
-    # draw the original picture
-    for i in range (6):
-        dx = (i % 2) * 250 - 30
-        dy = (i / 2) * 150
-
-        cv.Ellipse (image,
-                       (dx + 150, dy + 100),
-                       (100, 70),
-                      0, 0, 360, _white, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 115, dy + 70),
-                       (30, 20),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 185, dy + 70),
-                       (30, 20),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 115, dy + 70),
-                       (15, 15),
-                      0, 0, 360, _white, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 185, dy + 70),
-                       (15, 15),
-                      0, 0, 360, _white, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 115, dy + 70),
-                       (5, 5),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 185, dy + 70),
-                       (5, 5),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 150, dy + 100),
-                       (10, 5),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 150, dy + 150),
-                       (40, 10),
-                      0, 0, 360, _black, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 27, dy + 100),
-                       (20, 35),
-                      0, 0, 360, _white, -1, 8, 0)
-        cv.Ellipse (image,
-                       (dx + 273, dy + 100),
-                       (20, 35),
-                      0, 0, 360, _white, -1, 8, 0)
-
-    # create window and display the original picture in it
-    cv.NamedWindow ("image", 1)
-    cv.ShowImage ("image", image)
-
-    # create the storage area
-    storage = cv.CreateMemStorage (0)
-
-    # find the contours
-    contours = cv.FindContours(image,
-                               storage,
-                               cv.CV_RETR_TREE,
-                               cv.CV_CHAIN_APPROX_SIMPLE,
-                               (0,0))
-
-    # comment this out if you do not want approximation
-    contours = cv.ApproxPoly (contours,
-                                storage,
-                                cv.CV_POLY_APPROX_DP, 3, 1)
-
-    # create the window for the contours
-    cv.NamedWindow ("contours", 1)
-
-    # create the trackbar, to enable the change of the displayed level
-    cv.CreateTrackbar ("levels+3", "contours", 3, 7, on_trackbar)
-
-    # call one time the callback, so we will have the 1st display done
-    on_trackbar (_DEFAULT_LEVEL)
-
-    # wait a key pressed to end
-    cv.WaitKey (0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/convexhull.py b/samples/python/convexhull.py
deleted file mode 100755
index 17423e053f..0000000000
--- a/samples/python/convexhull.py
+++ /dev/null
@@ -1,65 +0,0 @@
-#! /usr/bin/env python
-
-print "OpenCV Python version of convexhull"
-
-# import the necessary things for OpenCV
-import cv2.cv as cv
-
-# to generate random values
-import random
-
-# how many points we want at max
-_MAX_POINTS = 100
-
-if __name__ == '__main__':
-
-    # main object to get random values from
-    my_random = random.Random ()
-
-    # create the image where we want to display results
-    image = cv.CreateImage ( (500, 500), 8, 3)
-
-    # create the window to put the image in
-    cv.NamedWindow ('hull', cv.CV_WINDOW_AUTOSIZE)
-
-    while True:
-        # do forever
-
-        # get a random number of points
-        count = my_random.randrange (0, _MAX_POINTS) + 1
-
-        # initialisations
-        points = []
-
-        for i in range (count):
-            # generate a random point
-            points.append ( (
-                my_random.randrange (0, image.width / 2) + image.width / 4,
-                my_random.randrange (0, image.width / 2) + image.width / 4
-                ))
-
-        # compute the convex hull
-        storage = cv.CreateMemStorage(0)
-        hull = cv.ConvexHull2 (points, storage, cv.CV_CLOCKWISE, 1)
-
-        # start with an empty image
-        cv.SetZero (image)
-
-        # draw all the points as circles in red
-        for i in range (count):
-            cv.Circle (image, points [i], 2,
-                          (0, 0, 255, 0),
-                         cv.CV_FILLED, cv.CV_AA, 0)
-
-        # Draw the convex hull as a closed polyline in green
-        cv.PolyLine(image, [hull], 1, cv.RGB(0,255,0), 1, cv.CV_AA)
-
-        # display the final image
-        cv.ShowImage ('hull', image)
-
-        # handle events, and wait a key pressed
-        k = cv.WaitKey (0) % 0x100
-        if k == 27:
-            # user has press the ESC key, so exit
-            break
-    cv.DestroyAllWindows()
diff --git a/samples/python/cv20squares.py b/samples/python/cv20squares.py
deleted file mode 100755
index e472fa1973..0000000000
--- a/samples/python/cv20squares.py
+++ /dev/null
@@ -1,169 +0,0 @@
-#!/usr/bin/python
-
-"""
-Find Squares in image by finding countours and filtering
-"""
-#Results slightly different from C version on same images, but is
-#otherwise ok
-
-import math
-import cv2.cv as cv
-
-def angle(pt1, pt2, pt0):
-    "calculate angle contained by 3 points(x, y)"
-    dx1 = pt1[0] - pt0[0]
-    dy1 = pt1[1] - pt0[1]
-    dx2 = pt2[0] - pt0[0]
-    dy2 = pt2[1] - pt0[1]
-
-    nom = dx1*dx2 + dy1*dy2
-    denom = math.sqrt( (dx1*dx1 + dy1*dy1) * (dx2*dx2 + dy2*dy2) + 1e-10 )
-    ang = nom / denom
-    return ang
-
-def is_square(contour):
-    """
-    Squareness checker
-
-    Square contours should:
-        -have 4 vertices after approximation,
-        -have relatively large area (to filter out noisy contours)
-        -be convex.
-        -have angles between sides close to 90deg (cos(ang) ~0 )
-    Note: absolute value of an area is used because area may be
-    positive or negative - in accordance with the contour orientation
-    """
-
-    area = math.fabs( cv.ContourArea(contour) )
-    isconvex = cv.CheckContourConvexity(contour)
-    s = 0
-    if len(contour) == 4 and area > 1000 and isconvex:
-        for i in range(1, 4):
-            # find minimum angle between joint edges (maximum of cosine)
-            pt1 = contour[i]
-            pt2 = contour[i-1]
-            pt0 = contour[i-2]
-
-            t = math.fabs(angle(pt0, pt1, pt2))
-            if s <= t:s = t
-
-        # if cosines of all angles are small (all angles are ~90 degree)
-        # then its a square
-        if s < 0.3:return True
-
-    return False
-
-def find_squares_from_binary( gray ):
-    """
-    use contour search to find squares in binary image
-    returns list of numpy arrays containing 4 points
-    """
-    squares = []
-    storage = cv.CreateMemStorage(0)
-    contours = cv.FindContours(gray, storage, cv.CV_RETR_TREE, cv.CV_CHAIN_APPROX_SIMPLE, (0,0))
-    storage = cv.CreateMemStorage(0)
-    while contours:
-        #approximate contour with accuracy proportional to the contour perimeter
-        arclength = cv.ArcLength(contours)
-        polygon = cv.ApproxPoly( contours, storage, cv.CV_POLY_APPROX_DP, arclength * 0.02, 0)
-        if is_square(polygon):
-            squares.append(polygon[0:4])
-        contours = contours.h_next()
-
-    return squares
-
-def find_squares4(color_img):
-    """
-    Finds multiple squares in image
-
-    Steps:
-    -Use Canny edge to highlight contours, and dilation to connect
-    the edge segments.
-    -Threshold the result to binary edge tokens
-    -Use cv.FindContours: returns a cv.CvSequence of cv.CvContours
-    -Filter each candidate: use Approx poly, keep only contours with 4 vertices,
-    enough area, and ~90deg angles.
-
-    Return all squares contours in one flat list of arrays, 4 x,y points each.
-    """
-    #select even sizes only
-    width, height = (color_img.width & -2, color_img.height & -2 )
-    timg = cv.CloneImage( color_img ) # make a copy of input image
-    gray = cv.CreateImage( (width,height), 8, 1 )
-
-    # select the maximum ROI in the image
-    cv.SetImageROI( timg, (0, 0, width, height) )
-
-    # down-scale and upscale the image to filter out the noise
-    pyr = cv.CreateImage( (width/2, height/2), 8, 3 )
-    cv.PyrDown( timg, pyr, 7 )
-    cv.PyrUp( pyr, timg, 7 )
-
-    tgray = cv.CreateImage( (width,height), 8, 1 )
-    squares = []
-
-    # Find squares in every color plane of the image
-    # Two methods, we use both:
-    # 1. Canny to catch squares with gradient shading. Use upper threshold
-    # from slider, set the lower to 0 (which forces edges merging). Then
-    # dilate canny output to remove potential holes between edge segments.
-    # 2. Binary thresholding at multiple levels
-    N = 11
-    for c in [0, 1, 2]:
-        #extract the c-th color plane
-        cv.SetImageCOI( timg, c+1 );
-        cv.Copy( timg, tgray, None );
-        cv.Canny( tgray, gray, 0, 50, 5 )
-        cv.Dilate( gray, gray)
-        squares = squares + find_squares_from_binary( gray )
-
-        # Look for more squares at several threshold levels
-        for l in range(1, N):
-            cv.Threshold( tgray, gray, (l+1)*255/N, 255, cv.CV_THRESH_BINARY )
-            squares = squares + find_squares_from_binary( gray )
-
-    return squares
-
-
-RED = (0,0,255)
-GREEN = (0,255,0)
-def draw_squares( color_img, squares ):
-    """
-    Squares is py list containing 4-pt numpy arrays. Step through the list
-    and draw a polygon for each 4-group
-    """
-    color, othercolor = RED, GREEN
-    for square in squares:
-        cv.PolyLine(color_img, [square], True, color, 3, cv.CV_AA, 0)
-        color, othercolor = othercolor, color
-
-    cv.ShowImage(WNDNAME, color_img)
-
-
-WNDNAME = "Squares Demo"
-def main():
-    """Open test color images, create display window, start the search"""
-    cv.NamedWindow(WNDNAME, 1)
-    for name in [ "../c/pic%d.png" % i for i in [1, 2, 3, 4, 5, 6] ]:
-        img0 = cv.LoadImage(name, 1)
-        try:
-            img0
-        except ValueError:
-            print "Couldn't load %s\n" % name
-            continue
-
-        # slider deleted from C version, same here and use fixed Canny param=50
-        img = cv.CloneImage(img0)
-
-        cv.ShowImage(WNDNAME, img)
-
-        # force the image processing
-        draw_squares( img, find_squares4( img ) )
-
-        # wait for key.
-        if cv.WaitKey(-1) % 0x100 == 27:
-            break
-
-if __name__ == "__main__":
-    main()
-    cv.DestroyAllWindows()
diff --git a/samples/python/cvutils.py b/samples/python/cvutils.py
deleted file mode 100755
index a2ef8ae325..0000000000
--- a/samples/python/cvutils.py
+++ /dev/null
@@ -1,20 +0,0 @@
-#!/usr/bin/python
-
-import cv2.cv as cv
-import urllib2
-from sys import argv
-
-def load_sample(name=None):
-    if len(argv) > 1:
-        img0 = cv.LoadImage(argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    elif name is not None:
-        try:
-            img0 = cv.LoadImage(name, cv.CV_LOAD_IMAGE_COLOR)
-        except IOError:
-            urlbase = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/'
-            file = name.split('/')[-1]
-            filedata = urllib2.urlopen(urlbase+file).read()
-            imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-            cv.SetData(imagefiledata, filedata, len(filedata))
-            img0 = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-    return img0
diff --git a/samples/python/delaunay.py b/samples/python/delaunay.py
deleted file mode 100755
index b26d4341c5..0000000000
--- a/samples/python/delaunay.py
+++ /dev/null
@@ -1,139 +0,0 @@
-#!/usr/bin/python
-"""
-the script demostrates iterative construction of
-delaunay triangulation and voronoi tesselation
-
-Original Author (C version): ?
-Converted to Python by: Roman Stanchak
-"""
-import cv2.cv as cv
-import random
-
-def draw_subdiv_point( img, fp, color ):
-    cv.Circle( img, (cv.Round(fp[0]), cv.Round(fp[1])), 3, color, cv.CV_FILLED, 8, 0 );
-
-def draw_subdiv_edge( img, edge, color ):
-    org_pt = cv.Subdiv2DEdgeOrg(edge);
-    dst_pt = cv.Subdiv2DEdgeDst(edge);
-
-    if org_pt and dst_pt :
-
-        org = org_pt.pt;
-        dst = dst_pt.pt;
-
-        iorg = ( cv.Round( org[0] ), cv.Round( org[1] ));
-        idst = ( cv.Round( dst[0] ), cv.Round( dst[1] ));
-
-        cv.Line( img, iorg, idst, color, 1, cv.CV_AA, 0 );
-
-
-def draw_subdiv( img, subdiv, delaunay_color, voronoi_color ):
-
-    for edge in subdiv.edges:
-        edge_rot = cv.Subdiv2DRotateEdge( edge, 1 )
-
-        draw_subdiv_edge( img, edge_rot, voronoi_color );
-        draw_subdiv_edge( img, edge, delaunay_color );
-
-
-def locate_point( subdiv, fp, img, active_color ):
-
-    (res, e0) = cv.Subdiv2DLocate( subdiv, fp );
-
-    if res in [ cv.CV_PTLOC_INSIDE, cv.CV_PTLOC_ON_EDGE ]:
-        e = e0
-        while True:
-            draw_subdiv_edge( img, e, active_color );
-            e = cv.Subdiv2DGetEdge(e, cv.CV_NEXT_AROUND_LEFT);
-            if e == e0:
-                break
-
-    draw_subdiv_point( img, fp, active_color );
-
-
-def draw_subdiv_facet( img, edge ):
-
-    t = edge;
-    count = 0;
-
-    # count number of edges in facet
-    while count == 0 or t != edge:
-        count+=1
-        t = cv.Subdiv2DGetEdge( t, cv.CV_NEXT_AROUND_LEFT );
-
-    buf = []
-
-    # gather points
-    t = edge;
-    for i in range(count):
-        assert t>4
-        pt = cv.Subdiv2DEdgeOrg( t );
-        if not pt:
-            break;
-        buf.append( ( cv.Round(pt.pt[0]), cv.Round(pt.pt[1]) ) );
-        t = cv.Subdiv2DGetEdge( t, cv.CV_NEXT_AROUND_LEFT );
-
-    if( len(buf)==count ):
-        pt = cv.Subdiv2DEdgeDst( cv.Subdiv2DRotateEdge( edge, 1 ));
-        cv.FillConvexPoly( img, buf, cv.RGB(random.randrange(256),random.randrange(256),random.randrange(256)), cv.CV_AA, 0 );
-        cv.PolyLine( img, [buf], 1, cv.RGB(0,0,0), 1, cv.CV_AA, 0);
-        draw_subdiv_point( img, pt.pt, cv.RGB(0,0,0));
-
-def paint_voronoi( subdiv, img ):
-
-    cv.CalcSubdivVoronoi2D( subdiv );
-
-    for edge in subdiv.edges:
-
-        # left
-        draw_subdiv_facet( img, cv.Subdiv2DRotateEdge( edge, 1 ));
-
-        # right
-        draw_subdiv_facet( img, cv.Subdiv2DRotateEdge( edge, 3 ));
-
-if __name__ == '__main__':
-    win = "source";
-    rect = ( 0, 0, 600, 600 );
-
-    active_facet_color = cv.RGB( 255, 0, 0 );
-    delaunay_color  = cv.RGB( 0,0,0);
-    voronoi_color = cv.RGB(0, 180, 0);
-    bkgnd_color = cv.RGB(255,255,255);
-
-    img = cv.CreateImage( (rect[2],rect[3]), 8, 3 );
-    cv.Set( img, bkgnd_color );
-
-    cv.NamedWindow( win, 1 );
-
-    storage = cv.CreateMemStorage(0);
-    subdiv = cv.CreateSubdivDelaunay2D( rect, storage );
-
-    print "Delaunay triangulation will be build now interactively."
-    print "To stop the process, press any key\n";
-
-    for i in range(200):
-        fp = ( random.random()*(rect[2]-10)+5, random.random()*(rect[3]-10)+5 )
-
-        locate_point( subdiv, fp, img, active_facet_color );
-        cv.ShowImage( win, img );
-
-        if( cv.WaitKey( 100 ) >= 0 ):
-            break;
-
-        cv.SubdivDelaunay2DInsert( subdiv, fp );
-        cv.CalcSubdivVoronoi2D( subdiv );
-        cv.Set( img, bkgnd_color );
-        draw_subdiv( img, subdiv, delaunay_color, voronoi_color );
-        cv.ShowImage( win, img );
-
-        if( cv.WaitKey( 100 ) >= 0 ):
-            break;
-
-
-    cv.Set( img, bkgnd_color );
-    paint_voronoi( subdiv, img );
-    cv.ShowImage( win, img );
-
-    cv.WaitKey(0);
-
-    cv.DestroyWindow( win );
diff --git a/samples/python/demhist.py b/samples/python/demhist.py
deleted file mode 100755
index 72565d6d74..0000000000
--- a/samples/python/demhist.py
+++ /dev/null
@@ -1,80 +0,0 @@
-#!/usr/bin/python
-import cv2.cv as cv
-import sys
-import urllib2
-
-hist_size = 64
-range_0 = [0, 256]
-ranges = [ range_0 ]
-
-class DemHist:
-
-    def __init__(self, src_image):
-        self.src_image = src_image
-        self.dst_image = cv.CloneMat(src_image)
-        self.hist_image = cv.CreateImage((320, 200), 8, 1)
-        self.hist = cv.CreateHist([hist_size], cv.CV_HIST_ARRAY, ranges, 1)
-
-        self.brightness = 0
-        self.contrast = 0
-
-        cv.NamedWindow("image", 0)
-        cv.NamedWindow("histogram", 0)
-        cv.CreateTrackbar("brightness", "image", 100, 200, self.update_brightness)
-        cv.CreateTrackbar("contrast", "image", 100, 200, self.update_contrast)
-
-        self.update_brightcont()
-
-    def update_brightness(self, val):
-        self.brightness = val - 100
-        self.update_brightcont()
-
-    def update_contrast(self, val):
-        self.contrast = val - 100
-        self.update_brightcont()
-
-    def update_brightcont(self):
-        # The algorithm is by Werner D. Streidt
-        # (http://visca.com/ffactory/archives/5-99/msg00021.html)
-
-        if self.contrast > 0:
-            delta = 127. * self.contrast / 100
-            a = 255. / (255. - delta * 2)
-            b = a * (self.brightness - delta)
-        else:
-            delta = -128. * self.contrast / 100
-            a = (256. - delta * 2) / 255.
-            b = a * self.brightness + delta
-
-        cv.ConvertScale(self.src_image, self.dst_image, a, b)
-        cv.ShowImage("image", self.dst_image)
-
-        cv.CalcArrHist([self.dst_image], self.hist)
-        (min_value, max_value, _, _) = cv.GetMinMaxHistValue(self.hist)
-        cv.Scale(self.hist.bins, self.hist.bins, float(self.hist_image.height) / max_value, 0)
-
-        cv.Set(self.hist_image, cv.ScalarAll(255))
-        bin_w = round(float(self.hist_image.width) / hist_size)
-
-        for i in range(hist_size):
-            cv.Rectangle(self.hist_image, (int(i * bin_w), self.hist_image.height),
-                         (int((i + 1) * bin_w), self.hist_image.height - cv.Round(self.hist.bins[i])),
-                         cv.ScalarAll(0), -1, 8, 0)
-
-        cv.ShowImage("histogram", self.hist_image)
-
-if __name__ == "__main__":
-    # Load the source image.
-    if len(sys.argv) > 1:
-        src_image = cv.GetMat(cv.LoadImage(sys.argv[1], 0))
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/baboon.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        src_image = cv.DecodeImageM(imagefiledata, 0)
-
-    dh = DemHist(src_image)
-
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/dft.py b/samples/python/dft.py
deleted file mode 100755
index 0ecc50ac81..0000000000
--- a/samples/python/dft.py
+++ /dev/null
@@ -1,114 +0,0 @@
-#!/usr/bin/python
-import cv2.cv as cv
-import sys
-import urllib2
-
-# Rearrange the quadrants of Fourier image so that the origin is at
-# the image center
-# src & dst arrays of equal size & type
-def cvShiftDFT(src_arr, dst_arr ):
-
-    size = cv.GetSize(src_arr)
-    dst_size = cv.GetSize(dst_arr)
-
-    if dst_size != size:
-        cv.Error( cv.CV_StsUnmatchedSizes, "cv.ShiftDFT", "Source and Destination arrays must have equal sizes", __FILE__, __LINE__ )
-
-    if(src_arr is dst_arr):
-        tmp = cv.CreateMat(size[1]/2, size[0]/2, cv.GetElemType(src_arr))
-
-    cx = size[0] / 2
-    cy = size[1] / 2 # image center
-
-    q1 = cv.GetSubRect( src_arr, (0,0,cx, cy) )
-    q2 = cv.GetSubRect( src_arr, (cx,0,cx,cy) )
-    q3 = cv.GetSubRect( src_arr, (cx,cy,cx,cy) )
-    q4 = cv.GetSubRect( src_arr, (0,cy,cx,cy) )
-    d1 = cv.GetSubRect( src_arr, (0,0,cx,cy) )
-    d2 = cv.GetSubRect( src_arr, (cx,0,cx,cy) )
-    d3 = cv.GetSubRect( src_arr, (cx,cy,cx,cy) )
-    d4 = cv.GetSubRect( src_arr, (0,cy,cx,cy) )
-
-    if(src_arr is not dst_arr):
-        if( not cv.CV_ARE_TYPES_EQ( q1, d1 )):
-            cv.Error( cv.CV_StsUnmatchedFormats, "cv.ShiftDFT", "Source and Destination arrays must have the same format", __FILE__, __LINE__ )
-
-        cv.Copy(q3, d1)
-        cv.Copy(q4, d2)
-        cv.Copy(q1, d3)
-        cv.Copy(q2, d4)
-
-    else:
-        cv.Copy(q3, tmp)
-        cv.Copy(q1, q3)
-        cv.Copy(tmp, q1)
-        cv.Copy(q4, tmp)
-        cv.Copy(q2, q4)
-        cv.Copy(tmp, q2)
-
-if __name__ == "__main__":
-
-    if len(sys.argv) > 1:
-        im = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_GRAYSCALE)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/baboon.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        im = cv.DecodeImageM(imagefiledata, cv.CV_LOAD_IMAGE_GRAYSCALE)
-
-    realInput = cv.CreateImage( cv.GetSize(im), cv.IPL_DEPTH_64F, 1)
-    imaginaryInput = cv.CreateImage( cv.GetSize(im), cv.IPL_DEPTH_64F, 1)
-    complexInput = cv.CreateImage( cv.GetSize(im), cv.IPL_DEPTH_64F, 2)
-
-    cv.Scale(im, realInput, 1.0, 0.0)
-    cv.Zero(imaginaryInput)
-    cv.Merge(realInput, imaginaryInput, None, None, complexInput)
-
-    dft_M = cv.GetOptimalDFTSize( im.height - 1 )
-    dft_N = cv.GetOptimalDFTSize( im.width - 1 )
-
-    dft_A = cv.CreateMat( dft_M, dft_N, cv.CV_64FC2 )
-    image_Re = cv.CreateImage( (dft_N, dft_M), cv.IPL_DEPTH_64F, 1)
-    image_Im = cv.CreateImage( (dft_N, dft_M), cv.IPL_DEPTH_64F, 1)
-
-    # copy A to dft_A and pad dft_A with zeros
-    tmp = cv.GetSubRect( dft_A, (0,0, im.width, im.height))
-    cv.Copy( complexInput, tmp, None )
-    if(dft_A.width > im.width):
-        tmp = cv.GetSubRect( dft_A, (im.width,0, dft_N - im.width, im.height))
-        cv.Zero( tmp )
-
-    # no need to pad bottom part of dft_A with zeros because of
-    # use nonzero_rows parameter in cv.FT() call below
-
-    cv.DFT( dft_A, dft_A, cv.CV_DXT_FORWARD, complexInput.height )
-
-    cv.NamedWindow("win", 0)
-    cv.NamedWindow("magnitude", 0)
-    cv.ShowImage("win", im)
-
-    # Split Fourier in real and imaginary parts
-    cv.Split( dft_A, image_Re, image_Im, None, None )
-
-    # Compute the magnitude of the spectrum Mag = sqrt(Re^2 + Im^2)
-    cv.Pow( image_Re, image_Re, 2.0)
-    cv.Pow( image_Im, image_Im, 2.0)
-    cv.Add( image_Re, image_Im, image_Re, None)
-    cv.Pow( image_Re, image_Re, 0.5 )
-
-    # Compute log(1 + Mag)
-    cv.AddS( image_Re, cv.ScalarAll(1.0), image_Re, None ) # 1 + Mag
-    cv.Log( image_Re, image_Re ) # log(1 + Mag)
-
-
-    # Rearrange the quadrants of Fourier image so that the origin is at
-    # the image center
-    cvShiftDFT( image_Re, image_Re )
-
-    min, max, pt1, pt2 = cv.MinMaxLoc(image_Re)
-    cv.Scale(image_Re, image_Re, 1.0/(max-min), 1.0*(-min)/(max-min))
-    cv.ShowImage("magnitude", image_Re)
-
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/distrans.py b/samples/python/distrans.py
deleted file mode 100755
index 38ace4419a..0000000000
--- a/samples/python/distrans.py
+++ /dev/null
@@ -1,72 +0,0 @@
-#!/usr/bin/python
-import sys
-import cv2.cv as cv
-import urllib2
-
-wndname = "Distance transform"
-tbarname = "Threshold"
-
-# The output images
-dist = 0
-dist8u1 = 0
-dist8u2 = 0
-dist8u = 0
-dist32s = 0
-
-gray = 0
-edge = 0
-
-# define a trackbar callback
-def on_trackbar(edge_thresh):
-
-    cv.Threshold(gray, edge, float(edge_thresh), float(edge_thresh), cv.CV_THRESH_BINARY)
-    #Distance transform
-    cv.DistTransform(edge, dist, cv.CV_DIST_L2, cv.CV_DIST_MASK_5)
-
-    cv.ConvertScale(dist, dist, 5000.0, 0)
-    cv.Pow(dist, dist, 0.5)
-
-    cv.ConvertScale(dist, dist32s, 1.0, 0.5)
-    cv.AndS(dist32s, cv.ScalarAll(255), dist32s, None)
-    cv.ConvertScale(dist32s, dist8u1, 1, 0)
-    cv.ConvertScale(dist32s, dist32s, -1, 0)
-    cv.AddS(dist32s, cv.ScalarAll(255), dist32s, None)
-    cv.ConvertScale(dist32s, dist8u2, 1, 0)
-    cv.Merge(dist8u1, dist8u2, dist8u2, None, dist8u)
-    cv.ShowImage(wndname, dist8u)
-
-
-if __name__ == "__main__":
-    edge_thresh = 100
-
-    if len(sys.argv) > 1:
-        gray = cv.LoadImage(sys.argv[1], cv.CV_LOAD_IMAGE_GRAYSCALE)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/stuff.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        gray = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_GRAYSCALE)
-
-    # Create the output image
-    dist = cv.CreateImage((gray.width, gray.height), cv.IPL_DEPTH_32F, 1)
-    dist8u1 = cv.CloneImage(gray)
-    dist8u2 = cv.CloneImage(gray)
-    dist8u = cv.CreateImage((gray.width, gray.height), cv.IPL_DEPTH_8U, 3)
-    dist32s = cv.CreateImage((gray.width, gray.height), cv.IPL_DEPTH_32S, 1)
-
-    # Convert to grayscale
-    edge = cv.CloneImage(gray)
-
-    # Create a window
-    cv.NamedWindow(wndname, 1)
-
-    # create a toolbar
-    cv.CreateTrackbar(tbarname, wndname, edge_thresh, 255, on_trackbar)
-
-    # Show the image
-    on_trackbar(edge_thresh)
-
-    # Wait for a key stroke; the same function arranges events processing
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/dmtx.py b/samples/python/dmtx.py
deleted file mode 100755
index 0ff2b43da5..0000000000
--- a/samples/python/dmtx.py
+++ /dev/null
@@ -1,184 +0,0 @@
-#!/usr/bin/python
-
-import cv2.cv as cv
-import time
-from pydmtx import DataMatrix
-import numpy
-import sys
-import math
-
-'''
-Find 2 D barcode based on up to 3 channel datamatrix
-'''
-
-def absnorm8(im, im8):
-    """ im may be any single-channel image type.  Return an 8-bit version, absolute value, normalized so that max is 255 """
-    (minVal, maxVal, _, _) = cv.MinMaxLoc(im)
-    cv.ConvertScaleAbs(im, im8, 255 / max(abs(minVal), abs(maxVal)), 0)
-    return im8
-
-font = cv.InitFont(cv.CV_FONT_HERSHEY_SIMPLEX, 1.0, 1.0, thickness = 2, lineType = cv.CV_AA)
-if 0:
-    started = time.time()
-    print dm_write.decode(bg.width, bg.height, buffer(bg.tostring()), max_count = 1, min_edge = 12, max_edge = 13, shape = DataMatrix.DmtxSymbol10x10) # , timeout = 10)
-    print "took", time.time() - started
-
-class DmtxFinder:
-    def __init__(self):
-        self.cache = {}
-        self.dm = DataMatrix()
-
-    def Cached(self, name, rows, cols, type):
-        key = (name, rows, cols)
-        if not key in self.cache:
-            self.cache[key] = cv.CreateMat(rows, cols, type)
-        return self.cache[key]
-
-    def find0(self, img):
-        started = time.time()
-        self.dm.decode(img.width,
-                       img.height,
-                       buffer(img.tostring()),
-                       max_count = 4,
-                       #min_edge = 6,
-                       #max_edge = 19      # Units of 2 pixels
-                       )
-        print "brute", time.time() - started
-        found = {}
-        for i in range(self.dm.count()):
-            stats = dm_read.stats(i + 1)
-            print stats
-            found[stats[0]] = stats[1]
-        return found
-
-    def find(self, img):
-        started = time.time()
-        gray = self.Cached('gray', img.height, img.width, cv.CV_8UC1)
-        cv.CvtColor(img, gray, cv.CV_BGR2GRAY)
-
-        sobel = self.Cached('sobel', img.height, img.width, cv.CV_16SC1)
-        sobely = self.Cached('sobely', img.height, img.width, cv.CV_16SC1)
-
-        cv.Sobel(gray, sobel, 1, 0)
-        cv.Sobel(gray, sobely, 0, 1)
-        cv.Add(sobel, sobely, sobel)
-
-        sobel8 = self.Cached('sobel8', sobel.height, sobel.width, cv.CV_8UC1)
-        absnorm8(sobel, sobel8)
-        cv.Threshold(sobel8, sobel8, 128.0, 255.0, cv.CV_THRESH_BINARY)
-
-        sobel_integral = self.Cached('sobel_integral', img.height + 1, img.width + 1, cv.CV_32SC1)
-        cv.Integral(sobel8, sobel_integral)
-
-        d = 16
-        _x1y1 = cv.GetSubRect(sobel_integral, (0, 0, sobel_integral.cols - d, sobel_integral.rows - d))
-        _x1y2 = cv.GetSubRect(sobel_integral, (0, d, sobel_integral.cols - d, sobel_integral.rows - d))
-        _x2y1 = cv.GetSubRect(sobel_integral, (d, 0, sobel_integral.cols - d, sobel_integral.rows - d))
-        _x2y2 = cv.GetSubRect(sobel_integral, (d, d, sobel_integral.cols - d, sobel_integral.rows - d))
-
-        summation = cv.CloneMat(_x2y2)
-        cv.Sub(summation, _x1y2, summation)
-        cv.Sub(summation, _x2y1, summation)
-        cv.Add(summation, _x1y1, summation)
-        sum8 = self.Cached('sum8', summation.height, summation.width, cv.CV_8UC1)
-        absnorm8(summation, sum8)
-        cv.Threshold(sum8, sum8, 32.0, 255.0, cv.CV_THRESH_BINARY)
-
-        cv.ShowImage("sum8", sum8)
-        seq = cv.FindContours(sum8, cv.CreateMemStorage(), cv.CV_RETR_EXTERNAL)
-        subimg = cv.GetSubRect(img, (d / 2, d / 2, sum8.cols, sum8.rows))
-        t_cull = time.time() - started
-
-        seqs = []
-        while seq:
-            seqs.append(seq)
-            seq = seq.h_next()
-
-        started = time.time()
-        found = {}
-        print 'seqs', len(seqs)
-        for seq in seqs:
-            area = cv.ContourArea(seq)
-            if area > 1000:
-                rect = cv.BoundingRect(seq)
-                edge = int((14 / 14.) * math.sqrt(area) / 2 + 0.5)
-                candidate = cv.GetSubRect(subimg, rect)
-                sym = self.dm.decode(candidate.width,
-                                     candidate.height,
-                                     buffer(candidate.tostring()),
-                                     max_count = 1,
-                                     #min_edge = 6,
-                                     #max_edge = int(edge)      # Units of 2 pixels
-                                     )
-                if sym:
-                    onscreen = [(d / 2 + rect[0] + x, d / 2 + rect[1] + y) for (x, y) in self.dm.stats(1)[1]]
-                    found[sym] = onscreen
-                else:
-                    print "FAILED"
-        t_brute = time.time() - started
-        print "cull took", t_cull, "brute", t_brute
-        return found
-
-bg = cv.CreateMat(1024, 1024, cv.CV_8UC3)
-cv.Set(bg, cv.RGB(0, 0, 0))
-df = DmtxFinder()
-
-cv.NamedWindow("camera", 1)
-
-def mkdmtx(msg):
-    dm_write = DataMatrix()
-    dm_write.encode(msg)
-    pi = dm_write.image # .resize((14, 14))
-    cv_im = cv.CreateImageHeader(pi.size, cv.IPL_DEPTH_8U, 3)
-    cv.SetData(cv_im, pi.tostring())
-    return cv_im
-
-# test = [('WIL', (100,100))]: # , ('LOW', (250,100)), ('GAR', (300, 300)), ('AGE', (500, 300))]:
-
-test = []
-y = 10
-for j in range(7):
-    r = 28 + j * 4
-    mr = r * math.sqrt(2)
-    y += mr * 1.8
-    test += [(str(deg) + "abcdefgh"[j], (50 + deg * 11, y), math.pi * deg / 180, r) for deg in range(0, 90, 10)]
-
-for (msg, (x, y), angle, r) in test:
-    map = cv.CreateMat(2, 3, cv.CV_32FC1)
-    corners = [(x + r * math.cos(angle + th), y + r * math.sin(angle + th)) for th in [0, math.pi / 2, math.pi, 3 * math.pi / 4]]
-    src = mkdmtx(msg)
-    (sx, sy) = cv.GetSize(src)
-    cv.GetAffineTransform([(0,0), (sx, 0), (sx, sy)], corners[:3], map)
-    temp = cv.CreateMat(bg.rows, bg.cols, cv.CV_8UC3)
-    cv.Set(temp, cv.RGB(0, 0, 0))
-    cv.WarpAffine(src, temp, map)
-    cv.Or(temp, bg, bg)
-
-
-cv.ShowImage("comp", bg)
-scribble = cv.CloneMat(bg)
-
-if 0:
-    for i in range(10):
-        df.find(bg)
-
-for (sym, coords) in df.find(bg).items():
-    print sym
-    cv.PolyLine(scribble, [coords], 1, cv.CV_RGB(255, 0,0), 1, lineType = cv.CV_AA)
-    Xs = [x for (x, y) in coords]
-    Ys = [y for (x, y) in coords]
-    where = ((min(Xs) + max(Xs)) / 2, max(Ys) - 50)
-    cv.PutText(scribble, sym, where, font, cv.RGB(0,255, 0))
-
-cv.ShowImage("results", scribble)
-cv.WaitKey()
-cv.DestroyAllWindows()
-
-sys.exit(0)
-
-capture = cv.CaptureFromCAM(0)
-while True:
-    img = cv.QueryFrame(capture)
-    cv.ShowImage("capture", img)
-    print df.find(img)
-    cv.WaitKey(6)
diff --git a/samples/python/drawing.py b/samples/python/drawing.py
deleted file mode 100755
index be2c244104..0000000000
--- a/samples/python/drawing.py
+++ /dev/null
@@ -1,162 +0,0 @@
-#! /usr/bin/env python
-from random import Random
-import colorsys
-
-print "OpenCV Python version of drawing"
-
-import cv2.cv as cv
-
-def random_color(random):
-    """
-    Return a random color
-    """
-    icolor = random.randint(0, 0xFFFFFF)
-    return cv.Scalar(icolor & 0xff, (icolor >> 8) & 0xff, (icolor >> 16) & 0xff)
-
-if __name__ == '__main__':
-
-    # some "constants"
-    width = 1000
-    height = 700
-    window_name = "Drawing Demo"
-    number = 100
-    delay = 5
-    line_type = cv.CV_AA  # change it to 8 to see non-antialiased graphics
-
-    # create the source image
-    image = cv.CreateImage( (width, height), 8, 3)
-
-    # create window and display the original picture in it
-    cv.NamedWindow(window_name, 1)
-    cv.SetZero(image)
-    cv.ShowImage(window_name, image)
-
-    # create the random number
-    random = Random()
-
-    # draw some lines
-    for i in range(number):
-        pt1 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        pt2 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        cv.Line(image, pt1, pt2,
-                   random_color(random),
-                   random.randrange(0, 10),
-                   line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # draw some rectangles
-    for i in range(number):
-        pt1 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        pt2 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        cv.Rectangle(image, pt1, pt2,
-                        random_color(random),
-                        random.randrange(-1, 9),
-                        line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # draw some ellipes
-    for i in range(number):
-        pt1 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        sz =  (random.randrange(0, 200),
-                        random.randrange(0, 200))
-        angle = random.randrange(0, 1000) * 0.180
-        cv.Ellipse(image, pt1, sz, angle, angle - 100, angle + 200,
-                        random_color(random),
-                        random.randrange(-1, 9),
-                        line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # init the list of polylines
-    nb_polylines = 2
-    polylines_size = 3
-    pt = [0,] * nb_polylines
-    for a in range(nb_polylines):
-        pt [a] = [0,] * polylines_size
-
-    # draw some polylines
-    for i in range(number):
-        for a in range(nb_polylines):
-            for b in range(polylines_size):
-                pt [a][b] =  (random.randrange(-width, 2 * width),
-                                     random.randrange(-height, 2 * height))
-        cv.PolyLine(image, pt, 1,
-                       random_color(random),
-                       random.randrange(1, 9),
-                       line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # draw some filled polylines
-    for i in range(number):
-        for a in range(nb_polylines):
-            for b in range(polylines_size):
-                pt [a][b] =  (random.randrange(-width, 2 * width),
-                                     random.randrange(-height, 2 * height))
-        cv.FillPoly(image, pt,
-                       random_color(random),
-                       line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # draw some circles
-    for i in range(number):
-        pt1 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        cv.Circle(image, pt1, random.randrange(0, 300),
-                     random_color(random),
-                     random.randrange(-1, 9),
-                     line_type, 0)
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # draw some text
-    for i in range(number):
-        pt1 =  (random.randrange(-width, 2 * width),
-                          random.randrange(-height, 2 * height))
-        font = cv.InitFont(random.randrange(0, 8),
-                              random.randrange(0, 100) * 0.05 + 0.01,
-                              random.randrange(0, 100) * 0.05 + 0.01,
-                              random.randrange(0, 5) * 0.1,
-                              random.randrange(0, 10),
-                              line_type)
-
-        cv.PutText(image, "Testing text rendering!",
-                      pt1, font,
-                      random_color(random))
-
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # prepare a text, and get it's properties
-    font = cv.InitFont(cv.CV_FONT_HERSHEY_COMPLEX,
-                          3, 3, 0.0, 5, line_type)
-    text_size, ymin = cv.GetTextSize("OpenCV forever!", font)
-    pt1 = ((width - text_size[0]) / 2, (height + text_size[1]) / 2)
-    image2 = cv.CloneImage(image)
-
-    # now, draw some OpenCV pub ;-)
-    for i in range(0, 512, 2):
-        cv.SubS(image2, cv.ScalarAll(i), image)
-        (r, g, b) = colorsys.hsv_to_rgb((i % 100) / 100., 1, 1)
-        cv.PutText(image, "OpenCV forever!",
-                      pt1, font, cv.RGB(255 * r, 255 * g, 255 * b))
-        cv.ShowImage(window_name, image)
-        cv.WaitKey(delay)
-
-    # wait some key to end
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/edge.py b/samples/python/edge.py
deleted file mode 100755
index 2d3b8efb63..0000000000
--- a/samples/python/edge.py
+++ /dev/null
@@ -1,60 +0,0 @@
-#! /usr/bin/env python
-
-print "OpenCV Python version of edge"
-
-import sys
-import urllib2
-import cv2.cv as cv
-
-# some definitions
-win_name = "Edge"
-trackbar_name = "Threshold"
-
-# the callback on the trackbar
-def on_trackbar(position):
-
-    cv.Smooth(gray, edge, cv.CV_BLUR, 3, 3, 0)
-    cv.Not(gray, edge)
-
-    # run the edge dector on gray scale
-    cv.Canny(gray, edge, position, position * 3, 3)
-
-    # reset
-    cv.SetZero(col_edge)
-
-    # copy edge points
-    cv.Copy(im, col_edge, edge)
-
-    # show the im
-    cv.ShowImage(win_name, col_edge)
-
-if __name__ == '__main__':
-    if len(sys.argv) > 1:
-        im = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        im = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    # create the output im
-    col_edge = cv.CreateImage((im.width, im.height), 8, 3)
-
-    # convert to grayscale
-    gray = cv.CreateImage((im.width, im.height), 8, 1)
-    edge = cv.CreateImage((im.width, im.height), 8, 1)
-    cv.CvtColor(im, gray, cv.CV_BGR2GRAY)
-
-    # create the window
-    cv.NamedWindow(win_name, cv.CV_WINDOW_AUTOSIZE)
-
-    # create the trackbar
-    cv.CreateTrackbar(trackbar_name, win_name, 1, 100, on_trackbar)
-
-    # show the im
-    on_trackbar(0)
-
-    # wait a key pressed to end
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/facedetect.py b/samples/python/facedetect.py
deleted file mode 100755
index 925b73094e..0000000000
--- a/samples/python/facedetect.py
+++ /dev/null
@@ -1,101 +0,0 @@
-#!/usr/bin/python
-"""
-This program is demonstration for face and object detection using haar-like features.
-The program finds faces in a camera image or video stream and displays a red box around them.
-
-Original C implementation by:  ?
-Python implementation by: Roman Stanchak, James Bowman
-"""
-import sys
-import cv2.cv as cv
-from optparse import OptionParser
-
-# Parameters for haar detection
-# From the API:
-# The default parameters (scale_factor=2, min_neighbors=3, flags=0) are tuned
-# for accurate yet slow object detection. For a faster operation on real video
-# images the settings are:
-# scale_factor=1.2, min_neighbors=2, flags=CV_HAAR_DO_CANNY_PRUNING,
-# min_size=<minimum possible face size
-
-min_size = (20, 20)
-image_scale = 2
-haar_scale = 1.2
-min_neighbors = 2
-haar_flags = 0
-
-def detect_and_draw(img, cascade):
-    # allocate temporary images
-    gray = cv.CreateImage((img.width,img.height), 8, 1)
-    small_img = cv.CreateImage((cv.Round(img.width / image_scale),
-                   cv.Round (img.height / image_scale)), 8, 1)
-
-    # convert color input image to grayscale
-    cv.CvtColor(img, gray, cv.CV_BGR2GRAY)
-
-    # scale input image for faster processing
-    cv.Resize(gray, small_img, cv.CV_INTER_LINEAR)
-
-    cv.EqualizeHist(small_img, small_img)
-
-    if(cascade):
-        t = cv.GetTickCount()
-        faces = cv.HaarDetectObjects(small_img, cascade, cv.CreateMemStorage(0),
-                                     haar_scale, min_neighbors, haar_flags, min_size)
-        t = cv.GetTickCount() - t
-        print "detection time = %gms" % (t/(cv.GetTickFrequency()*1000.))
-        if faces:
-            for ((x, y, w, h), n) in faces:
-                # the input to cv.HaarDetectObjects was resized, so scale the
-                # bounding box of each face and convert it to two CvPoints
-                pt1 = (int(x * image_scale), int(y * image_scale))
-                pt2 = (int((x + w) * image_scale), int((y + h) * image_scale))
-                cv.Rectangle(img, pt1, pt2, cv.RGB(255, 0, 0), 3, 8, 0)
-
-    cv.ShowImage("result", img)
-
-if __name__ == '__main__':
-
-    parser = OptionParser(usage = "usage: %prog [options] [filename|camera_index]")
-    parser.add_option("-c", "--cascade", action="store", dest="cascade", type="str", help="Haar cascade file, default %default", default = "../data/haarcascades/haarcascade_frontalface_alt.xml")
-    (options, args) = parser.parse_args()
-
-    cascade = cv.Load(options.cascade)
-
-    if len(args) != 1:
-        parser.print_help()
-        sys.exit(1)
-
-    input_name = args[0]
-    if input_name.isdigit():
-        capture = cv.CreateCameraCapture(int(input_name))
-    else:
-        capture = None
-
-    cv.NamedWindow("result", 1)
-
-    if capture:
-        frame_copy = None
-        while True:
-            frame = cv.QueryFrame(capture)
-            if not frame:
-                cv.WaitKey(0)
-                break
-            if not frame_copy:
-                frame_copy = cv.CreateImage((frame.width,frame.height),
-                                            cv.IPL_DEPTH_8U, frame.nChannels)
-            if frame.origin == cv.IPL_ORIGIN_TL:
-                cv.Copy(frame, frame_copy)
-            else:
-                cv.Flip(frame, frame_copy, 0)
-
-            detect_and_draw(frame_copy, cascade)
-
-            if cv.WaitKey(10) >= 0:
-                break
-    else:
-        image = cv.LoadImage(input_name, 1)
-        detect_and_draw(image, cascade)
-        cv.WaitKey(0)
-
-    cv.DestroyWindow("result")
diff --git a/samples/python/fback.py b/samples/python/fback.py
deleted file mode 100755
index 6d90925378..0000000000
--- a/samples/python/fback.py
+++ /dev/null
@@ -1,56 +0,0 @@
-#!/usr/bin/env python
-
-from cv import *
-
-class FBackDemo:
-    def __init__(self):
-        self.capture = CaptureFromCAM(0)
-        self.mv_step = 16
-        self.mv_scale = 1.5
-        self.mv_color = (0, 255, 0)
-        self.cflow = None
-        self.flow = None
-
-        NamedWindow( "Optical Flow", 1 )
-
-        print( "Press ESC - quit the program\n" )
-
-    def draw_flow(self, flow, prevgray):
-        """ Returns a nice representation of a hue histogram """
-
-        CvtColor(prevgray, self.cflow, CV_GRAY2BGR)
-        for y in range(0, flow.height, self.mv_step):
-            for x in range(0, flow.width, self.mv_step):
-                fx, fy = flow[y, x]
-                Line(self.cflow, (x,y), (x+fx,y+fy), self.mv_color)
-                Circle(self.cflow, (x,y), 2, self.mv_color, -1)
-        ShowImage("Optical Flow", self.cflow)
-
-    def run(self):
-        first_frame = True
-
-        while True:
-            frame = QueryFrame( self.capture )
-
-            if first_frame:
-                gray = CreateImage(GetSize(frame), 8, 1)
-                prev_gray = CreateImage(GetSize(frame), 8, 1)
-                flow = CreateImage(GetSize(frame), 32, 2)
-                self.cflow = CreateImage(GetSize(frame), 8, 3)
-
-            CvtColor(frame, gray, CV_BGR2GRAY)
-            if not first_frame:
-                CalcOpticalFlowFarneback(prev_gray, gray, flow,
-                    pyr_scale=0.5, levels=3, winsize=15,
-                    iterations=3, poly_n=5, poly_sigma=1.2, flags=0)
-                self.draw_flow(flow, prev_gray)
-                c = WaitKey(7)
-                if c in [27, ord('q'), ord('Q')]:
-                    break
-            prev_gray, gray = gray, prev_gray
-            first_frame = False
-
-if __name__=="__main__":
-    demo = FBackDemo()
-    demo.run()
-    cv.DestroyAllWindows()
diff --git a/samples/python/ffilldemo.py b/samples/python/ffilldemo.py
deleted file mode 100755
index 2ecf5ffa72..0000000000
--- a/samples/python/ffilldemo.py
+++ /dev/null
@@ -1,160 +0,0 @@
-#!/usr/bin/python
-import sys
-import random
-import urllib2
-import cv2.cv as cv
-
-im=None;
-mask=None;
-color_img=None;
-gray_img0 = None;
-gray_img = None;
-ffill_case = 1;
-lo_diff = 20
-up_diff = 20;
-connectivity = 4;
-is_color = 1;
-is_mask = 0;
-new_mask_val = 255;
-
-def update_lo( pos ):
-    lo_diff = pos
-def update_up( pos ):
-    up_diff = pos
-
-def on_mouse( event, x, y, flags, param ):
-
-    if( not color_img ):
-        return;
-
-    if event == cv.CV_EVENT_LBUTTONDOWN:
-            my_mask = None
-            seed = (x,y);
-            if ffill_case==0:
-                lo = up = 0
-                flags = connectivity + (new_mask_val << 8)
-            else:
-                lo = lo_diff;
-                up = up_diff;
-                flags = connectivity + (new_mask_val << 8) + cv.CV_FLOODFILL_FIXED_RANGE
-            b = random.randint(0,255)
-            g = random.randint(0,255)
-            r = random.randint(0,255)
-
-            if( is_mask ):
-                my_mask = mask
-                cv.Threshold( mask, mask, 1, 128, cv.CV_THRESH_BINARY );
-
-            if( is_color ):
-
-                color = cv.CV_RGB( r, g, b );
-                comp = cv.FloodFill( color_img, seed, color, cv.CV_RGB( lo, lo, lo ),
-                             cv.CV_RGB( up, up, up ), flags, my_mask );
-                cv.ShowImage( "image", color_img );
-
-            else:
-
-                brightness = cv.RealScalar((r*2 + g*7 + b + 5)/10);
-                comp = cv.FloodFill( gray_img, seed, brightness, cv.RealScalar(lo),
-                             cv.RealScalar(up), flags, my_mask );
-                cv.ShowImage( "image", gray_img );
-
-
-            print "%g pixels were repainted" % comp[0]
-
-            if( is_mask ):
-                cv.ShowImage( "mask", mask );
-
-
-
-
-if __name__ == "__main__":
-
-    if len(sys.argv) > 1:
-        im = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        im = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    print "Hot keys:"
-    print "\tESC - quit the program"
-    print "\tc - switch color/grayscale mode"
-    print "\tm - switch mask mode"
-    print "\tr - restore the original image"
-    print "\ts - use null-range floodfill"
-    print "\tf - use gradient floodfill with fixed(absolute) range"
-    print "\tg - use gradient floodfill with floating(relative) range"
-    print "\t4 - use 4-connectivity mode"
-    print "\t8 - use 8-connectivity mode"
-
-    color_img = cv.CloneImage( im );
-    gray_img0 = cv.CreateImage( (color_img.width, color_img.height), 8, 1 );
-    cv.CvtColor( color_img, gray_img0, cv.CV_BGR2GRAY );
-    gray_img = cv.CloneImage( gray_img0 );
-    mask = cv.CreateImage( (color_img.width + 2, color_img.height + 2), 8, 1 );
-
-    cv.NamedWindow( "image", 1 );
-    cv.CreateTrackbar( "lo_diff", "image", lo_diff, 255, update_lo);
-    cv.CreateTrackbar( "up_diff", "image", up_diff, 255, update_up);
-
-    cv.SetMouseCallback( "image", on_mouse );
-
-    while True:
-        if( is_color ):
-            cv.ShowImage( "image", color_img );
-        else:
-            cv.ShowImage( "image", gray_img );
-
-        c = cv.WaitKey(0) % 0x100
-        if c == 27:
-            print("Exiting ...");
-            sys.exit(0)
-        elif c == ord('c'):
-            if( is_color ):
-
-                print("Grayscale mode is set");
-                cv.CvtColor( color_img, gray_img, cv.CV_BGR2GRAY );
-                is_color = 0;
-
-            else:
-
-                print("Color mode is set");
-                cv.Copy( im, color_img, None );
-                cv.Zero( mask );
-                is_color = 1;
-
-        elif c == ord('m'):
-            if( is_mask ):
-                cv.DestroyWindow( "mask" );
-                is_mask = 0;
-
-            else:
-                cv.NamedWindow( "mask", 0 );
-                cv.Zero( mask );
-                cv.ShowImage( "mask", mask );
-                is_mask = 1;
-
-        elif c == ord('r'):
-            print("Original image is restored");
-            cv.Copy( im, color_img, None );
-            cv.Copy( gray_img0, gray_img, None );
-            cv.Zero( mask );
-        elif c == ord('s'):
-            print("Simple floodfill mode is set");
-            ffill_case = 0;
-        elif c == ord('f'):
-            print("Fixed Range floodfill mode is set");
-            ffill_case = 1;
-        elif c == ord('g'):
-            print("Gradient (floating range) floodfill mode is set");
-            ffill_case = 2;
-        elif c == ord('4'):
-            print("4-connectivity mode is set");
-            connectivity = 4;
-        elif c == ord('8'):
-            print("8-connectivity mode is set");
-            connectivity = 8;
-    cv.DestroyAllWindows()
diff --git a/samples/python/fitellipse.py b/samples/python/fitellipse.py
deleted file mode 100755
index 88a927d451..0000000000
--- a/samples/python/fitellipse.py
+++ /dev/null
@@ -1,104 +0,0 @@
-#!/usr/bin/python
-"""
-This program is a demonstration of ellipse fitting.
-
-Trackbar controls threshold parameter.
-
-Gray lines are contours.  Colored lines are fit ellipses.
-
-Original C implementation by:  Denis Burenkov.
-Python implementation by: Roman Stanchak, James Bowman
-"""
-
-import sys
-import urllib2
-import random
-import cv2.cv as cv
-
-def contour_iterator(contour):
-    while contour:
-        yield contour
-        contour = contour.h_next()
-
-class FitEllipse:
-
-    def __init__(self, source_image, slider_pos):
-        self.source_image = source_image
-        cv.CreateTrackbar("Threshold", "Result", slider_pos, 255, self.process_image)
-        self.process_image(slider_pos)
-
-    def process_image(self, slider_pos):
-        """
-        This function finds contours, draws them and their approximation by ellipses.
-        """
-        stor = cv.CreateMemStorage()
-
-        # Create the destination images
-        image02 = cv.CloneImage(self.source_image)
-        cv.Zero(image02)
-        image04 = cv.CreateImage(cv.GetSize(self.source_image), cv.IPL_DEPTH_8U, 3)
-        cv.Zero(image04)
-
-        # Threshold the source image. This needful for cv.FindContours().
-        cv.Threshold(self.source_image, image02, slider_pos, 255, cv.CV_THRESH_BINARY)
-
-        # Find all contours.
-        cont = cv.FindContours(image02,
-            stor,
-            cv.CV_RETR_LIST,
-            cv.CV_CHAIN_APPROX_NONE,
-            (0, 0))
-
-        for c in contour_iterator(cont):
-            # Number of points must be more than or equal to 6 for cv.FitEllipse2
-            if len(c) >= 6:
-                # Copy the contour into an array of (x,y)s
-                PointArray2D32f = cv.CreateMat(1, len(c), cv.CV_32FC2)
-                for (i, (x, y)) in enumerate(c):
-                    PointArray2D32f[0, i] = (x, y)
-
-                # Draw the current contour in gray
-                gray = cv.CV_RGB(100, 100, 100)
-                cv.DrawContours(image04, c, gray, gray,0,1,8,(0,0))
-
-                # Fits ellipse to current contour.
-                (center, size, angle) = cv.FitEllipse2(PointArray2D32f)
-
-                # Convert ellipse data from float to integer representation.
-                center = (cv.Round(center[0]), cv.Round(center[1]))
-                size = (cv.Round(size[0] * 0.5), cv.Round(size[1] * 0.5))
-
-                # Draw ellipse in random color
-                color = cv.CV_RGB(random.randrange(256),random.randrange(256),random.randrange(256))
-                cv.Ellipse(image04, center, size,
-                          angle, 0, 360,
-                          color, 2, cv.CV_AA, 0)
-
-        # Show image. HighGUI use.
-        cv.ShowImage( "Result", image04 )
-
-
-if __name__ == '__main__':
-    if len(sys.argv) > 1:
-        source_image = cv.LoadImage(sys.argv[1], cv.CV_LOAD_IMAGE_GRAYSCALE)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/stuff.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        source_image = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_GRAYSCALE)
-
-    # Create windows.
-    cv.NamedWindow("Source", 1)
-    cv.NamedWindow("Result", 1)
-
-    # Show the image.
-    cv.ShowImage("Source", source_image)
-
-    fe = FitEllipse(source_image, 70)
-
-    print "Press any key to exit"
-    cv.WaitKey(0)
-
-    cv.DestroyWindow("Source")
-    cv.DestroyWindow("Result")
diff --git a/samples/python/houghlines.py b/samples/python/houghlines.py
deleted file mode 100755
index a437bfe022..0000000000
--- a/samples/python/houghlines.py
+++ /dev/null
@@ -1,58 +0,0 @@
-#!/usr/bin/python
-# This is a standalone program. Pass an image name as a first parameter of the program.
-
-import sys
-from math import sin, cos, sqrt, pi
-import cv2.cv as cv
-import urllib2
-
-# toggle between CV_HOUGH_STANDARD and CV_HOUGH_PROBILISTIC
-USE_STANDARD = True
-
-if __name__ == "__main__":
-    if len(sys.argv) > 1:
-        filename = sys.argv[1]
-        src = cv.LoadImage(filename, cv.CV_LOAD_IMAGE_GRAYSCALE)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/doc/pics/building.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        src = cv.DecodeImageM(imagefiledata, cv.CV_LOAD_IMAGE_GRAYSCALE)
-
-
-    cv.NamedWindow("Source", 1)
-    cv.NamedWindow("Hough", 1)
-
-    while True:
-        dst = cv.CreateImage(cv.GetSize(src), 8, 1)
-        color_dst = cv.CreateImage(cv.GetSize(src), 8, 3)
-        storage = cv.CreateMemStorage(0)
-        lines = 0
-        cv.Canny(src, dst, 50, 200, 3)
-        cv.CvtColor(dst, color_dst, cv.CV_GRAY2BGR)
-
-        if USE_STANDARD:
-            lines = cv.HoughLines2(dst, storage, cv.CV_HOUGH_STANDARD, 1, pi / 180, 100, 0, 0)
-            for (rho, theta) in lines[:100]:
-                a = cos(theta)
-                b = sin(theta)
-                x0 = a * rho
-                y0 = b * rho
-                pt1 = (cv.Round(x0 + 1000*(-b)), cv.Round(y0 + 1000*(a)))
-                pt2 = (cv.Round(x0 - 1000*(-b)), cv.Round(y0 - 1000*(a)))
-                cv.Line(color_dst, pt1, pt2, cv.RGB(255, 0, 0), 3, 8)
-        else:
-            lines = cv.HoughLines2(dst, storage, cv.CV_HOUGH_PROBABILISTIC, 1, pi / 180, 50, 50, 10)
-            for line in lines:
-                cv.Line(color_dst, line[0], line[1], cv.CV_RGB(255, 0, 0), 3, 8)
-
-        cv.ShowImage("Source", src)
-        cv.ShowImage("Hough", color_dst)
-
-        k = cv.WaitKey(0) % 0x100
-        if k == ord(' '):
-            USE_STANDARD = not USE_STANDARD
-        if k == 27:
-            break
-    cv.DestroyAllWindows()
diff --git a/samples/python/inpaint.py b/samples/python/inpaint.py
deleted file mode 100755
index fa7e9a0925..0000000000
--- a/samples/python/inpaint.py
+++ /dev/null
@@ -1,69 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import sys
-import cv2.cv as cv
-
-class Sketcher:
-    def __init__(self, windowname, dests):
-        self.prev_pt = None
-        self.windowname = windowname
-        self.dests = dests
-        cv.SetMouseCallback(self.windowname, self.on_mouse)
-
-    def on_mouse(self, event, x, y, flags, param):
-        pt = (x, y)
-        if event == cv.CV_EVENT_LBUTTONUP or not (flags & cv.CV_EVENT_FLAG_LBUTTON):
-            self.prev_pt = None
-        elif event == cv.CV_EVENT_LBUTTONDOWN:
-            self.prev_pt = pt
-        elif event == cv.CV_EVENT_MOUSEMOVE and (flags & cv.CV_EVENT_FLAG_LBUTTON) :
-            if self.prev_pt:
-                for dst in self.dests:
-                    cv.Line(dst, self.prev_pt, pt, cv.ScalarAll(255), 5, 8, 0)
-            self.prev_pt = pt
-            cv.ShowImage(self.windowname, img)
-
-if __name__=="__main__":
-    if len(sys.argv) > 1:
-        img0 = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        img0 = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    print "Hot keys:"
-    print "\tESC - quit the program"
-    print "\tr - restore the original image"
-    print "\ti or ENTER - run inpainting algorithm"
-    print "\t\t(before running it, paint something on the image)"
-
-    cv.NamedWindow("image", 1)
-    cv.NamedWindow("inpainted image", 1)
-
-    img = cv.CloneImage(img0)
-    inpainted = cv.CloneImage(img0)
-    inpaint_mask = cv.CreateImage(cv.GetSize(img), 8, 1)
-
-    cv.Zero(inpaint_mask)
-    cv.Zero(inpainted)
-    cv.ShowImage("image", img)
-    cv.ShowImage("inpainted image", inpainted)
-
-    sk = Sketcher("image", [img, inpaint_mask])
-    while True:
-        c = cv.WaitKey(0) % 0x100
-
-        if c == 27 or c == ord('q'):
-            break
-
-        if c == ord('r'):
-            cv.Zero(inpaint_mask)
-            cv.Copy(img0, img)
-            cv.ShowImage("image", img)
-
-        if c == ord('i') or c == ord('\n'):
-            cv.Inpaint(img, inpaint_mask, inpainted, 3, cv.CV_INPAINT_TELEA)
-            cv.ShowImage("inpainted image", inpainted)
-    cv.DestroyAllWindows()
diff --git a/samples/python/kalman.py b/samples/python/kalman.py
deleted file mode 100755
index cf040da5cf..0000000000
--- a/samples/python/kalman.py
+++ /dev/null
@@ -1,98 +0,0 @@
-#!/usr/bin/python
-"""
-   Tracking of rotating point.
-   Rotation speed is constant.
-   Both state and measurements vectors are 1D (a point angle),
-   Measurement is the real point angle + gaussian noise.
-   The real and the estimated points are connected with yellow line segment,
-   the real and the measured points are connected with red line segment.
-   (if Kalman filter works correctly,
-    the yellow segment should be shorter than the red one).
-   Pressing any key (except ESC) will reset the tracking with a different speed.
-   Pressing ESC will stop the program.
-"""
-import urllib2
-import cv2.cv as cv
-from math import cos, sin, sqrt
-import sys
-
-if __name__ == "__main__":
-    A = [ [1, 1], [0, 1] ]
-
-    img = cv.CreateImage((500, 500), 8, 3)
-    kalman = cv.CreateKalman(2, 1, 0)
-    state = cv.CreateMat(2, 1, cv.CV_32FC1)  # (phi, delta_phi)
-    process_noise = cv.CreateMat(2, 1, cv.CV_32FC1)
-    measurement = cv.CreateMat(1, 1, cv.CV_32FC1)
-    rng = cv.RNG(-1)
-    code = -1L
-
-    cv.Zero(measurement)
-    cv.NamedWindow("Kalman", 1)
-
-    while True:
-        cv.RandArr(rng, state, cv.CV_RAND_NORMAL, cv.RealScalar(0), cv.RealScalar(0.1))
-
-        kalman.transition_matrix[0,0] = 1
-        kalman.transition_matrix[0,1] = 1
-        kalman.transition_matrix[1,0] = 0
-        kalman.transition_matrix[1,1] = 1
-
-        cv.SetIdentity(kalman.measurement_matrix, cv.RealScalar(1))
-        cv.SetIdentity(kalman.process_noise_cov, cv.RealScalar(1e-5))
-        cv.SetIdentity(kalman.measurement_noise_cov, cv.RealScalar(1e-1))
-        cv.SetIdentity(kalman.error_cov_post, cv.RealScalar(1))
-        cv.RandArr(rng, kalman.state_post, cv.CV_RAND_NORMAL, cv.RealScalar(0), cv.RealScalar(0.1))
-
-
-        while True:
-            def calc_point(angle):
-                return (cv.Round(img.width/2 + img.width/3*cos(angle)),
-                         cv.Round(img.height/2 - img.width/3*sin(angle)))
-
-            state_angle = state[0,0]
-            state_pt = calc_point(state_angle)
-
-            prediction = cv.KalmanPredict(kalman)
-            predict_angle = prediction[0, 0]
-            predict_pt = calc_point(predict_angle)
-
-            cv.RandArr(rng, measurement, cv.CV_RAND_NORMAL, cv.RealScalar(0),
-                       cv.RealScalar(sqrt(kalman.measurement_noise_cov[0, 0])))
-
-            # generate measurement
-            cv.MatMulAdd(kalman.measurement_matrix, state, measurement, measurement)
-
-            measurement_angle = measurement[0, 0]
-            measurement_pt = calc_point(measurement_angle)
-
-            # plot points
-            def draw_cross(center, color, d):
-                cv.Line(img, (center[0] - d, center[1] - d),
-                             (center[0] + d, center[1] + d), color, 1, cv.CV_AA, 0)
-                cv.Line(img, (center[0] + d, center[1] - d),
-                             (center[0] - d, center[1] + d), color, 1, cv.CV_AA, 0)
-
-            cv.Zero(img)
-            draw_cross(state_pt, cv.CV_RGB(255, 255, 255), 3)
-            draw_cross(measurement_pt, cv.CV_RGB(255, 0,0), 3)
-            draw_cross(predict_pt, cv.CV_RGB(0, 255, 0), 3)
-            cv.Line(img, state_pt, measurement_pt, cv.CV_RGB(255, 0,0), 3, cv. CV_AA, 0)
-            cv.Line(img, state_pt, predict_pt, cv.CV_RGB(255, 255, 0), 3, cv. CV_AA, 0)
-
-            cv.KalmanCorrect(kalman, measurement)
-
-            cv.RandArr(rng, process_noise, cv.CV_RAND_NORMAL, cv.RealScalar(0),
-                       cv.RealScalar(sqrt(kalman.process_noise_cov[0, 0])))
-            cv.MatMulAdd(kalman.transition_matrix, state, process_noise, state)
-
-            cv.ShowImage("Kalman", img)
-
-            code = cv.WaitKey(100) % 0x100
-            if code != -1:
-                break
-
-        if code in [27, ord('q'), ord('Q')]:
-            break
-
-    cv.DestroyWindow("Kalman")
diff --git a/samples/python/kmeans.py b/samples/python/kmeans.py
deleted file mode 100755
index 80fe424b34..0000000000
--- a/samples/python/kmeans.py
+++ /dev/null
@@ -1,60 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import cv2.cv as cv
-from random import randint
-MAX_CLUSTERS = 5
-
-if __name__ == "__main__":
-
-    color_tab = [
-        cv.CV_RGB(255, 0,0),
-        cv.CV_RGB(0, 255, 0),
-        cv.CV_RGB(100, 100, 255),
-        cv.CV_RGB(255, 0,255),
-        cv.CV_RGB(255, 255, 0)]
-    img = cv.CreateImage((500, 500), 8, 3)
-    rng = cv.RNG(-1)
-
-    cv.NamedWindow("clusters", 1)
-
-    while True:
-        cluster_count = randint(2, MAX_CLUSTERS)
-        sample_count = randint(1, 1000)
-        points = cv.CreateMat(sample_count, 1, cv.CV_32FC2)
-        clusters = cv.CreateMat(sample_count, 1, cv.CV_32SC1)
-
-        # generate random sample from multigaussian distribution
-        for k in range(cluster_count):
-            center = (cv.RandInt(rng)%img.width, cv.RandInt(rng)%img.height)
-            first = k*sample_count/cluster_count
-            last = sample_count
-            if k != cluster_count:
-                last = (k+1)*sample_count/cluster_count
-
-            point_chunk = cv.GetRows(points, first, last)
-
-            cv.RandArr(rng, point_chunk, cv.CV_RAND_NORMAL,
-                       cv.Scalar(center[0], center[1], 0, 0),
-                       cv.Scalar(img.width*0.1, img.height*0.1, 0, 0))
-
-
-        # shuffle samples
-        cv.RandShuffle(points, rng)
-
-        cv.KMeans2(points, cluster_count, clusters,
-                   (cv.CV_TERMCRIT_EPS + cv.CV_TERMCRIT_ITER, 10, 1.0))
-
-        cv.Zero(img)
-
-        for i in range(sample_count):
-            cluster_idx = int(clusters[i, 0])
-            pt = (cv.Round(points[i, 0][0]), cv.Round(points[i, 0][1]))
-            cv.Circle(img, pt, 2, color_tab[cluster_idx], cv.CV_FILLED, cv.CV_AA, 0)
-
-        cv.ShowImage("clusters", img)
-
-        key = cv.WaitKey(0) % 0x100
-        if key in [27, ord('q'), ord('Q')]:
-            break
-
-    cv.DestroyWindow("clusters")
diff --git a/samples/python/laplace.py b/samples/python/laplace.py
deleted file mode 100755
index df4f2430b6..0000000000
--- a/samples/python/laplace.py
+++ /dev/null
@@ -1,45 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import cv2.cv as cv
-import sys
-
-if __name__ == "__main__":
-    laplace = None
-    colorlaplace = None
-    planes = [ None, None, None ]
-    capture = None
-
-    if len(sys.argv) == 1:
-        capture = cv.CreateCameraCapture(0)
-    elif len(sys.argv) == 2 and sys.argv[1].isdigit():
-        capture = cv.CreateCameraCapture(int(sys.argv[1]))
-    elif len(sys.argv) == 2:
-        capture = cv.CreateFileCapture(sys.argv[1])
-
-    if not capture:
-        print "Could not initialize capturing..."
-        sys.exit(-1)
-
-    cv.NamedWindow("Laplacian", 1)
-
-    while True:
-        frame = cv.QueryFrame(capture)
-        if frame:
-            if not laplace:
-                planes = [cv.CreateImage((frame.width, frame.height), 8, 1) for i in range(3)]
-                laplace = cv.CreateImage((frame.width, frame.height), cv.IPL_DEPTH_16S, 1)
-                colorlaplace = cv.CreateImage((frame.width, frame.height), 8, 3)
-
-            cv.Split(frame, planes[0], planes[1], planes[2], None)
-            for plane in planes:
-                cv.Laplace(plane, laplace, 3)
-                cv.ConvertScaleAbs(laplace, plane, 1, 0)
-
-            cv.Merge(planes[0], planes[1], planes[2], None, colorlaplace)
-
-            cv.ShowImage("Laplacian", colorlaplace)
-
-        if cv.WaitKey(10) != -1:
-            break
-
-    cv.DestroyWindow("Laplacian")
diff --git a/samples/python/lkdemo.py b/samples/python/lkdemo.py
deleted file mode 100755
index 51485fe0a3..0000000000
--- a/samples/python/lkdemo.py
+++ /dev/null
@@ -1,192 +0,0 @@
-#! /usr/bin/env python
-
-print "OpenCV Python version of lkdemo"
-
-import sys
-
-# import the necessary things for OpenCV
-import cv2.cv as cv
-
-#############################################################################
-# some "constants"
-
-win_size = 10
-MAX_COUNT = 500
-
-#############################################################################
-# some "global" variables
-
-image = None
-pt = None
-add_remove_pt = False
-flags = 0
-night_mode = False
-need_to_init = False
-
-#############################################################################
-# the mouse callback
-
-# the callback on the trackbar
-def on_mouse (event, x, y, flags, param):
-
-    # we will use the global pt and add_remove_pt
-    global pt
-    global add_remove_pt
-
-    if image is None:
-        # not initialized, so skip
-        return
-
-    if image.origin != 0:
-        # different origin
-        y = image.height - y
-
-    if event == cv.CV_EVENT_LBUTTONDOWN:
-        # user has click, so memorize it
-        pt = (x, y)
-        add_remove_pt = True
-
-#############################################################################
-# so, here is the main part of the program
-
-if __name__ == '__main__':
-
-    frames = sys.argv[1:]
-    if frames == []:
-      print "usage lkdemo.py <image files>"
-      sys.exit(1)
-
-    # display a small howto use it
-    print "Hot keys: \n" \
-          "\tESC - quit the program\n" \
-          "\tr - auto-initialize tracking\n" \
-          "\tc - delete all the points\n" \
-          "\tn - switch the \"night\" mode on/off\n" \
-          "\tSPACE - next frame\n" \
-          "To add/remove a feature point click it\n"
-
-    # first, create the necessary windows
-    cv.NamedWindow ('LkDemo', cv.CV_WINDOW_AUTOSIZE)
-
-    # register the mouse callback
-    cv.SetMouseCallback ('LkDemo', on_mouse, None)
-
-    fc = 0
-    while 1:
-        # do forever
-
-        frame = cv.LoadImage(frames[fc])
-
-        if image is None:
-            # create the images we need
-            image = cv.CreateImage (cv.GetSize (frame), 8, 3)
-            image.origin = frame.origin
-            grey = cv.CreateImage (cv.GetSize (frame), 8, 1)
-            prev_grey = cv.CreateImage (cv.GetSize (frame), 8, 1)
-            pyramid = cv.CreateImage (cv.GetSize (frame), 8, 1)
-            prev_pyramid = cv.CreateImage (cv.GetSize (frame), 8, 1)
-            features = []
-
-        # copy the frame, so we can draw on it
-        cv.Copy (frame, image)
-
-        # create a grey version of the image
-        cv.CvtColor (image, grey, cv.CV_BGR2GRAY)
-
-        if night_mode:
-            # night mode: only display the points
-            cv.SetZero (image)
-
-        if need_to_init:
-            # we want to search all the good points
-
-            # create the wanted images
-            eig = cv.CreateImage (cv.GetSize (grey), 32, 1)
-            temp = cv.CreateImage (cv.GetSize (grey), 32, 1)
-
-            # the default parameters
-            quality = 0.01
-            min_distance = 10
-
-            # search the good points
-            features = cv.GoodFeaturesToTrack (
-                grey, eig, temp,
-                MAX_COUNT,
-                quality, min_distance, None, 3, 0, 0.04)
-
-            # refine the corner locations
-            features = cv.FindCornerSubPix (
-                grey,
-                features,
-                (win_size, win_size),  (-1, -1),
-                (cv.CV_TERMCRIT_ITER | cv.CV_TERMCRIT_EPS, 20, 0.03))
-
-        elif features != []:
-            # we have points, so display them
-
-            # calculate the optical flow
-            features, status, track_error = cv.CalcOpticalFlowPyrLK (
-                prev_grey, grey, prev_pyramid, pyramid,
-                features,
-                (win_size, win_size), 3,
-                (cv.CV_TERMCRIT_ITER|cv.CV_TERMCRIT_EPS, 20, 0.03),
-                flags)
-
-            # set back the points we keep
-            features = [ p for (st,p) in zip(status, features) if st]
-
-            if add_remove_pt:
-                # we have a point to add, so see if it is close to
-                # another one. If yes, don't use it
-                def ptptdist(p0, p1):
-                    dx = p0[0] - p1[0]
-                    dy = p0[1] - p1[1]
-                    return dx**2 + dy**2
-                if min([ ptptdist(pt, p) for p in features ]) < 25:
-                    # too close
-                    add_remove_pt = 0
-
-            # draw the points as green circles
-            for the_point in features:
-                cv.Circle (image, (int(the_point[0]), int(the_point[1])), 3, (0, 255, 0, 0), -1, 8, 0)
-
-        if add_remove_pt:
-            # we want to add a point
-            # refine this corner location and append it to 'features'
-
-            features += cv.FindCornerSubPix (
-                grey,
-                [pt],
-                (win_size, win_size),  (-1, -1),
-                (cv.CV_TERMCRIT_ITER | cv.CV_TERMCRIT_EPS,
-                20, 0.03))
-            # we are no longer in "add_remove_pt" mode
-            add_remove_pt = False
-
-        # swapping
-        prev_grey, grey = grey, prev_grey
-        prev_pyramid, pyramid = pyramid, prev_pyramid
-        need_to_init = False
-
-        # we can now display the image
-        cv.ShowImage ('LkDemo', image)
-
-        # handle events
-        c = cv.WaitKey(10) % 0x100
-
-        if c == 27:
-            # user has press the ESC key, so exit
-            break
-
-        # processing depending on the character
-        if 32 <= c and c < 128:
-          cc = chr(c).lower()
-          if cc == 'r':
-              need_to_init = True
-          elif cc == 'c':
-              features = []
-          elif cc == 'n':
-              night_mode = not night_mode
-          elif cc == ' ':
-              fc = (fc + 1) % len(frames)
-    cv.DestroyAllWindows()
diff --git a/samples/python/logpolar.py b/samples/python/logpolar.py
deleted file mode 100755
index 338acc2ce8..0000000000
--- a/samples/python/logpolar.py
+++ /dev/null
@@ -1,45 +0,0 @@
-#!/usr/bin/python
-import sys
-import urllib2
-import cv2.cv as cv
-
-src=None
-dst=None
-src2=None
-
-def on_mouse(event, x, y, flags, param):
-
-    if not src:
-        return
-
-    if event==cv.CV_EVENT_LBUTTONDOWN:
-        cv.LogPolar(src, dst, (x, y), 40, cv.CV_INTER_LINEAR + cv.CV_WARP_FILL_OUTLIERS)
-        cv.LogPolar(dst, src2, (x, y), 40, cv.CV_INTER_LINEAR + cv.CV_WARP_FILL_OUTLIERS + cv.CV_WARP_INVERSE_MAP)
-        cv.ShowImage("log-polar", dst)
-        cv.ShowImage("inverse log-polar", src2)
-
-if __name__ == "__main__":
-
-    if len(sys.argv) > 1:
-        src = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        src = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    cv.NamedWindow("original", 1)
-    cv.NamedWindow("log-polar", 1)
-    cv.NamedWindow("inverse log-polar", 1)
-
-
-    dst = cv.CreateImage((256, 256), 8, 3)
-    src2 = cv.CreateImage(cv.GetSize(src), 8, 3)
-
-    cv.SetMouseCallback("original", on_mouse)
-    on_mouse(cv.CV_EVENT_LBUTTONDOWN, src.width/2, src.height/2, None, None)
-
-    cv.ShowImage("original", src)
-    cv.WaitKey()
-    cv.DestroyAllWindows()
diff --git a/samples/python/minarea.py b/samples/python/minarea.py
deleted file mode 100755
index 5fe03f72b9..0000000000
--- a/samples/python/minarea.py
+++ /dev/null
@@ -1,62 +0,0 @@
-#!/usr/bin/python
-
-import urllib2
-import cv2.cv as cv
-from random import randint
-
-def roundxy(pt):
-    return (cv.Round(pt[0]), cv.Round(pt[1]))
-
-def draw_common(points):
-    success, center, radius = cv.MinEnclosingCircle(points)
-    if success:
-        cv.Circle(img, roundxy(center), cv.Round(radius), cv.CV_RGB(255, 255, 0), 1, cv. CV_AA, 0)
-
-    box = cv.MinAreaRect2(points)
-    box_vtx = [roundxy(p) for p in cv.BoxPoints(box)]
-    cv.PolyLine(img, [box_vtx], 1, cv.CV_RGB(0, 255, 255), 1, cv. CV_AA)
-
-def minarea_array(img, count):
-    pointMat = cv.CreateMat(count, 1, cv.CV_32SC2)
-    for i in range(count):
-        pointMat[i, 0] = (randint(img.width/4, img.width*3/4),
-                               randint(img.height/4, img.height*3/4))
-
-    cv.Zero(img)
-
-    for i in range(count):
-        cv.Circle(img, roundxy(pointMat[i, 0]), 2, cv.CV_RGB(255, 0, 0), cv.CV_FILLED, cv. CV_AA, 0)
-
-    draw_common(pointMat)
-
-def minarea_seq(img, count, storage):
-    points = [(randint(img.width/4, img.width*3/4), randint(img.height/4, img.height*3/4)) for i in range(count)]
-    cv.Zero(img)
-
-    for p in points:
-        cv.Circle(img, roundxy(p), 2, cv.CV_RGB(255, 0, 0), cv.CV_FILLED, cv. CV_AA, 0)
-
-    draw_common(points)
-
-if __name__ == "__main__":
-    img = cv.CreateImage((500, 500), 8, 3)
-    storage = cv.CreateMemStorage()
-
-    cv.NamedWindow("rect & circle", 1)
-
-    use_seq = True
-
-    while True:
-        count = randint(1, 100)
-        if use_seq:
-            minarea_seq(img, count, storage)
-        else:
-            minarea_array(img, count)
-
-        cv.ShowImage("rect & circle", img)
-        key = cv.WaitKey() % 0x100
-        if key in [27, ord('q'), ord('Q')]:
-            break
-
-        use_seq = not use_seq
-    cv.DestroyAllWindows()
diff --git a/samples/python/minidemo.py b/samples/python/minidemo.py
deleted file mode 100755
index cbc754e933..0000000000
--- a/samples/python/minidemo.py
+++ /dev/null
@@ -1,13 +0,0 @@
-#! /usr/bin/env python
-
-import cv2.cv as cv
-
-cap = cv.CreateFileCapture("../c/tree.avi")
-img = cv.QueryFrame(cap)
-print "Got frame of dimensions (", img.width, " x ", img.height, ")"
-
-cv.NamedWindow("win", cv.CV_WINDOW_AUTOSIZE)
-cv.ShowImage("win", img)
-cv.MoveWindow("win", 200, 200)
-cv.WaitKey(0)
-cv.DestroyAllWindows()
diff --git a/samples/python/morphology.py b/samples/python/morphology.py
deleted file mode 100755
index 6440272563..0000000000
--- a/samples/python/morphology.py
+++ /dev/null
@@ -1,52 +0,0 @@
-#!/usr/bin/python
-import sys
-import urllib2
-import cv2.cv as cv
-
-src = 0
-image = 0
-dest = 0
-element_shape = cv.CV_SHAPE_RECT
-
-def Opening(pos):
-    element = cv.CreateStructuringElementEx(pos*2+1, pos*2+1, pos, pos, element_shape)
-    cv.Erode(src, image, element, 1)
-    cv.Dilate(image, dest, element, 1)
-    cv.ShowImage("Opening & Closing", dest)
-def Closing(pos):
-    element = cv.CreateStructuringElementEx(pos*2+1, pos*2+1, pos, pos, element_shape)
-    cv.Dilate(src, image, element, 1)
-    cv.Erode(image, dest, element, 1)
-    cv.ShowImage("Opening & Closing", dest)
-def Erosion(pos):
-    element = cv.CreateStructuringElementEx(pos*2+1, pos*2+1, pos, pos, element_shape)
-    cv.Erode(src, dest, element, 1)
-    cv.ShowImage("Erosion & Dilation", dest)
-def Dilation(pos):
-    element = cv.CreateStructuringElementEx(pos*2+1, pos*2+1, pos, pos, element_shape)
-    cv.Dilate(src, dest, element, 1)
-    cv.ShowImage("Erosion & Dilation", dest)
-
-if __name__ == "__main__":
-    if len(sys.argv) > 1:
-        src = cv.LoadImage(sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        src = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    image = cv.CloneImage(src)
-    dest = cv.CloneImage(src)
-    cv.NamedWindow("Opening & Closing", 1)
-    cv.NamedWindow("Erosion & Dilation", 1)
-    cv.ShowImage("Opening & Closing", src)
-    cv.ShowImage("Erosion & Dilation", src)
-    cv.CreateTrackbar("Open", "Opening & Closing", 0, 10, Opening)
-    cv.CreateTrackbar("Close", "Opening & Closing", 0, 10, Closing)
-    cv.CreateTrackbar("Dilate", "Erosion & Dilation", 0, 10, Dilation)
-    cv.CreateTrackbar("Erode", "Erosion & Dilation", 0, 10, Erosion)
-    cv.WaitKey(0)
-    cv.DestroyWindow("Opening & Closing")
-    cv.DestroyWindow("Erosion & Dilation")
diff --git a/samples/python/motempl.py b/samples/python/motempl.py
deleted file mode 100755
index c2944c5c50..0000000000
--- a/samples/python/motempl.py
+++ /dev/null
@@ -1,110 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import sys
-import time
-from math import cos, sin
-import cv2.cv as cv
-
-CLOCKS_PER_SEC = 1.0
-MHI_DURATION = 1
-MAX_TIME_DELTA = 0.5
-MIN_TIME_DELTA = 0.05
-N = 4
-buf = range(10)
-last = 0
-mhi = None # MHI
-orient = None # orientation
-mask = None # valid orientation mask
-segmask = None # motion segmentation map
-storage = None # temporary storage
-
-def update_mhi(img, dst, diff_threshold):
-    global last
-    global mhi
-    global storage
-    global mask
-    global orient
-    global segmask
-    timestamp = time.clock() / CLOCKS_PER_SEC # get current time in seconds
-    size = cv.GetSize(img) # get current frame size
-    idx1 = last
-    if not mhi or cv.GetSize(mhi) != size:
-        for i in range(N):
-            buf[i] = cv.CreateImage(size, cv.IPL_DEPTH_8U, 1)
-            cv.Zero(buf[i])
-        mhi = cv.CreateImage(size,cv. IPL_DEPTH_32F, 1)
-        cv.Zero(mhi) # clear MHI at the beginning
-        orient = cv.CreateImage(size,cv. IPL_DEPTH_32F, 1)
-        segmask = cv.CreateImage(size,cv. IPL_DEPTH_32F, 1)
-        mask = cv.CreateImage(size,cv. IPL_DEPTH_8U, 1)
-
-    cv.CvtColor(img, buf[last], cv.CV_BGR2GRAY) # convert frame to grayscale
-    idx2 = (last + 1) % N # index of (last - (N-1))th frame
-    last = idx2
-    silh = buf[idx2]
-    cv.AbsDiff(buf[idx1], buf[idx2], silh) # get difference between frames
-    cv.Threshold(silh, silh, diff_threshold, 1, cv.CV_THRESH_BINARY) # and threshold it
-    cv.UpdateMotionHistory(silh, mhi, timestamp, MHI_DURATION) # update MHI
-    cv.CvtScale(mhi, mask, 255./MHI_DURATION,
-                (MHI_DURATION - timestamp)*255./MHI_DURATION)
-    cv.Zero(dst)
-    cv.Merge(mask, None, None, None, dst)
-    cv.CalcMotionGradient(mhi, mask, orient, MAX_TIME_DELTA, MIN_TIME_DELTA, 3)
-    if not storage:
-        storage = cv.CreateMemStorage(0)
-    seq = cv.SegmentMotion(mhi, segmask, storage, timestamp, MAX_TIME_DELTA)
-    for (area, value, comp_rect) in seq:
-        if comp_rect[2] + comp_rect[3] > 100: # reject very small components
-            color = cv.CV_RGB(255, 0,0)
-            silh_roi = cv.GetSubRect(silh, comp_rect)
-            mhi_roi = cv.GetSubRect(mhi, comp_rect)
-            orient_roi = cv.GetSubRect(orient, comp_rect)
-            mask_roi = cv.GetSubRect(mask, comp_rect)
-            angle = 360 - cv.CalcGlobalOrientation(orient_roi, mask_roi, mhi_roi, timestamp, MHI_DURATION)
-
-            count = cv.Norm(silh_roi, None, cv.CV_L1, None) # calculate number of points within silhouette ROI
-            if count < (comp_rect[2] * comp_rect[3] * 0.05):
-                continue
-
-            magnitude = 30.
-            center = ((comp_rect[0] + comp_rect[2] / 2), (comp_rect[1] + comp_rect[3] / 2))
-            cv.Circle(dst, center, cv.Round(magnitude*1.2), color, 3, cv.CV_AA, 0)
-            cv.Line(dst,
-                    center,
-                    (cv.Round(center[0] + magnitude * cos(angle * cv.CV_PI / 180)),
-                     cv.Round(center[1] - magnitude * sin(angle * cv.CV_PI / 180))),
-                    color,
-                    3,
-                    cv.CV_AA,
-                    0)
-
-if __name__ == "__main__":
-    motion = 0
-    capture = 0
-
-    if len(sys.argv)==1:
-        capture = cv.CreateCameraCapture(0)
-    elif len(sys.argv)==2 and sys.argv[1].isdigit():
-        capture = cv.CreateCameraCapture(int(sys.argv[1]))
-    elif len(sys.argv)==2:
-        capture = cv.CreateFileCapture(sys.argv[1])
-
-    if not capture:
-        print "Could not initialize capturing..."
-        sys.exit(-1)
-
-    cv.NamedWindow("Motion", 1)
-    while True:
-        image = cv.QueryFrame(capture)
-        if(image):
-            if(not motion):
-                    motion = cv.CreateImage((image.width, image.height), 8, 3)
-                    cv.Zero(motion)
-                    #motion.origin = image.origin
-            update_mhi(image, motion, 30)
-            cv.ShowImage("Motion", motion)
-            if(cv.WaitKey(10) != -1):
-                break
-        else:
-            break
-    cv.DestroyWindow("Motion")
diff --git a/samples/python/numpy_array.py b/samples/python/numpy_array.py
deleted file mode 100755
index a79eec7a95..0000000000
--- a/samples/python/numpy_array.py
+++ /dev/null
@@ -1,66 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import sys
-import cv2.cv as cv
-import numpy
-
-# SRGB-linear conversions using NumPy - see http://en.wikipedia.org/wiki/SRGB
-
-def srgb2lin(x):
-    a = 0.055
-    return numpy.where(x <= 0.04045,
-                       x * (1.0 / 12.92),
-                       numpy.power((x + a) * (1.0 / (1 + a)), 2.4))
-
-def lin2srgb(x):
-    a = 0.055
-    return numpy.where(x <= 0.0031308,
-                       x * 12.92,
-                       (1 + a) * numpy.power(x, 1 / 2.4) - a)
-
-if __name__ == "__main__":
-    if len(sys.argv) > 1:
-        img0 = cv.LoadImageM( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/lena.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        img0 = cv.DecodeImageM(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    cv.NamedWindow("original", 1)
-    cv.ShowImage("original", img0)
-
-    # Image was originally bytes in range 0-255.  Turn it into an array of floats in range 0.0 - 1.0
-    n = numpy.asarray(img0) / 255.0
-
-    # Use NumPy to do some transformations on the image
-
-    # Negate the image by subtracting it from 1.0
-    cv.NamedWindow("negative")
-    cv.ShowImage("negative", cv.fromarray(1.0 - n))
-
-    # Assume the image was sRGB, and compute the linear version.
-    cv.NamedWindow("linear")
-    cv.ShowImage("linear", cv.fromarray(srgb2lin(n)))
-
-    # Look at a subwindow
-    cv.NamedWindow("subwindow")
-    cv.ShowImage("subwindow", cv.fromarray(n[200:300,200:400]))
-
-    # Compute the grayscale image
-    cv.NamedWindow("monochrome")
-    ln = srgb2lin(n)
-    red = ln[:,:,0]
-    grn = ln[:,:,1]
-    blu = ln[:,:,2]
-    linear_mono = 0.3 * red + 0.59 * grn + 0.11 * blu
-    cv.ShowImage("monochrome", cv.fromarray(lin2srgb(linear_mono)))
-
-    # Apply a blur to the NumPy array using OpenCV
-    cv.NamedWindow("gaussian")
-    cv.Smooth(n, n, cv.CV_GAUSSIAN, 15, 15)
-    cv.ShowImage("gaussian", cv.fromarray(n))
-
-    cv.WaitKey(0)
-    cv.DestroyAllWindows()
diff --git a/samples/python/numpy_warhol.py b/samples/python/numpy_warhol.py
deleted file mode 100755
index f440abc037..0000000000
--- a/samples/python/numpy_warhol.py
+++ /dev/null
@@ -1,48 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import sys
-import cv2.cv as cv
-import numpy
-
-if __name__ == "__main__":
-    cv.NamedWindow("camera", 1)
-
-    capture = cv.CaptureFromCAM(0)
-
-    paste = cv.CreateMat(960, 1280, cv.CV_8UC3)
-    topleft = numpy.asarray(cv.GetSubRect(paste, (0, 0, 640, 480)))
-    topright = numpy.asarray(cv.GetSubRect(paste, (640, 0, 640, 480)))
-    bottomleft = numpy.asarray(cv.GetSubRect(paste, (0, 480, 640, 480)))
-    bottomright = numpy.asarray(cv.GetSubRect(paste, (640, 480, 640, 480)))
-
-    while True:
-        img = cv.GetMat(cv.QueryFrame(capture))
-
-        n = (numpy.asarray(img)).astype(numpy.uint8)
-
-        red = n[:,:,0]
-        grn = n[:,:,1]
-        blu = n[:,:,2]
-
-        topleft[:,:,0] = 255 - grn
-        topleft[:,:,1] = red
-        topleft[:,:,2] = blu
-
-        topright[:,:,0] = blu
-        topright[:,:,1] = 255 - red
-        topright[:,:,2] = grn
-
-        bottomright[:,:,0] = red
-        bottomright[:,:,1] = grn
-        bottomright[:,:,2] = 255 - blu
-
-        fgrn = grn.astype(numpy.float32)
-        fred = red.astype(numpy.float32)
-        bottomleft[:,:,0] = blu
-        bottomleft[:,:,1] = (abs(fgrn - fred)).astype(numpy.uint8)
-        bottomleft[:,:,2] = red
-
-        cv.ShowImage("camera", paste)
-        if cv.WaitKey(6) == 27:
-            break
-    cv.DestroyAllWindows()
diff --git a/samples/python/peopledetect.py b/samples/python/peopledetect.py
deleted file mode 100755
index 2824299791..0000000000
--- a/samples/python/peopledetect.py
+++ /dev/null
@@ -1,56 +0,0 @@
-#!/usr/bin/python
-
-import sys
-from cv import *
-
-def inside(r, q):
-    (rx, ry), (rw, rh) = r
-    (qx, qy), (qw, qh) = q
-    return rx > qx and ry > qy and rx + rw < qx + qw and ry + rh < qy + qh
-
-try:
-    img = LoadImage(sys.argv[1])
-except:
-    try:
-        f = open(sys.argv[1], "rt")
-    except:
-        print "cannot read " + sys.argv[1]
-        sys.exit(-1)
-    imglist = list(f.readlines())
-else:
-    imglist = [sys.argv[1]]
-
-NamedWindow("people detection demo", 1)
-storage = CreateMemStorage(0)
-
-for name in imglist:
-    n = name.strip()
-    print n
-    try:
-        img = LoadImage(n)
-    except:
-        continue
-
-    #ClearMemStorage(storage)
-    found = list(HOGDetectMultiScale(img, storage, win_stride=(8,8),
-        padding=(32,32), scale=1.05, group_threshold=2))
-    found_filtered = []
-    for r in found:
-        insidef = False
-        for q in found:
-            if inside(r, q):
-                insidef = True
-                break
-        if not insidef:
-            found_filtered.append(r)
-    for r in found_filtered:
-        (rx, ry), (rw, rh) = r
-        tl = (rx + int(rw*0.1), ry + int(rh*0.07))
-        br = (rx + int(rw*0.9), ry + int(rh*0.87))
-        Rectangle(img, tl, br, (0, 255, 0), 3)
-
-    ShowImage("people detection demo", img)
-    c = WaitKey(0)
-    if c == ord('q'):
-        break
-cv.DestroyAllWindows()
diff --git a/samples/python/pyramid_segmentation.py b/samples/python/pyramid_segmentation.py
deleted file mode 100755
index d9612529f6..0000000000
--- a/samples/python/pyramid_segmentation.py
+++ /dev/null
@@ -1,41 +0,0 @@
-#!/usr/bin/python
-import cv2.cv as cv
-
-class PyrSegmentation:
-    def __init__(self, img0):
-        self.thresh1 = 255
-        self.thresh2 = 30
-        self.level =4
-        self.storage = cv.CreateMemStorage()
-        cv.NamedWindow("Source", 0)
-        cv.ShowImage("Source", img0)
-        cv.NamedWindow("Segmentation", 0)
-        cv.CreateTrackbar("Thresh1", "Segmentation", self.thresh1, 255, self.set_thresh1)
-        cv.CreateTrackbar("Thresh2", "Segmentation",  self.thresh2, 255, self.set_thresh2)
-        self.image0 = cv.CloneImage(img0)
-        self.image1 = cv.CloneImage(img0)
-        cv.ShowImage("Segmentation", self.image1)
-
-    def set_thresh1(self, val):
-        self.thresh1 = val
-        self.on_segment()
-
-    def set_thresh2(self, val):
-        self.thresh2 = val
-        self.on_segment()
-
-    def on_segment(self):
-        comp = cv.PyrSegmentation(self.image0, self.image1, self.storage, \
-                            self.level, self.thresh1+1, self.thresh2+1)
-        cv.ShowImage("Segmentation", self.image1)
-
-    def run(self):
-        self.on_segment()
-        cv.WaitKey(0)
-
-if __name__ == "__main__":
-    img0 = cv.LoadImage("../c/fruits.jpg", 1)
-
-    # segmentation of the color image
-    PyrSegmentation(img0).run()
-    cv.DestroyAllWindows()
diff --git a/samples/python/squares.py b/samples/python/squares.py
deleted file mode 100755
index d5a2f20445..0000000000
--- a/samples/python/squares.py
+++ /dev/null
@@ -1,153 +0,0 @@
-#!/usr/bin/python
-#
-# The full "Square Detector" program.
-# It loads several images subsequentally and tries to find squares in
-# each image
-#
-
-import urllib2
-from math import sqrt
-import cv2.cv as cv
-
-thresh = 50
-img = None
-img0 = None
-storage = None
-wndname = "Square Detection Demo"
-
-def angle(pt1, pt2, pt0):
-    dx1 = pt1.x - pt0.x
-    dy1 = pt1.y - pt0.y
-    dx2 = pt2.x - pt0.x
-    dy2 = pt2.y - pt0.y
-    return (dx1*dx2 + dy1*dy2)/sqrt((dx1*dx1 + dy1*dy1)*(dx2*dx2 + dy2*dy2) + 1e-10)
-
-def findSquares4(img, storage):
-    N = 11
-    sz = (img.width & -2, img.height & -2)
-    timg = cv.CloneImage(img); # make a copy of input image
-    gray = cv.CreateImage(sz, 8, 1)
-    pyr = cv.CreateImage((sz.width/2, sz.height/2), 8, 3)
-    # create empty sequence that will contain points -
-    # 4 points per square (the square's vertices)
-    squares = cv.CreateSeq(0, sizeof_CvSeq, sizeof_CvPoint, storage)
-    squares = CvSeq_CvPoint.cast(squares)
-
-    # select the maximum ROI in the image
-    # with the width and height divisible by 2
-    subimage = cv.GetSubRect(timg, cv.Rect(0, 0, sz.width, sz.height))
-
-    # down-scale and upscale the image to filter out the noise
-    cv.PyrDown(subimage, pyr, 7)
-    cv.PyrUp(pyr, subimage, 7)
-    tgray = cv.CreateImage(sz, 8, 1)
-    # find squares in every color plane of the image
-    for c in range(3):
-        # extract the c-th color plane
-        channels = [None, None, None]
-        channels[c] = tgray
-        cv.Split(subimage, channels[0], channels[1], channels[2], None)
-        for l in range(N):
-            # hack: use Canny instead of zero threshold level.
-            # Canny helps to catch squares with gradient shading
-            if(l == 0):
-                # apply Canny. Take the upper threshold from slider
-                # and set the lower to 0 (which forces edges merging)
-                cv.Canny(tgray, gray, 0, thresh, 5)
-                # dilate canny output to remove potential
-                # holes between edge segments
-                cv.Dilate(gray, gray, None, 1)
-            else:
-                # apply threshold if l!=0:
-                #     tgray(x, y) = gray(x, y) < (l+1)*255/N ? 255 : 0
-                cv.Threshold(tgray, gray, (l+1)*255/N, 255, cv.CV_THRESH_BINARY)
-
-            # find contours and store them all as a list
-            count, contours = cv.FindContours(gray, storage, sizeof_CvContour,
-                cv.CV_RETR_LIST, cv. CV_CHAIN_APPROX_SIMPLE, (0, 0))
-
-            if not contours:
-                continue
-
-            # test each contour
-            for contour in contours.hrange():
-                # approximate contour with accuracy proportional
-                # to the contour perimeter
-                result = cv.ApproxPoly(contour, sizeof_CvContour, storage,
-                    cv.CV_POLY_APPROX_DP, cv.ContourPerimeter(contours)*0.02, 0)
-                # square contours should have 4 vertices after approximation
-                # relatively large area (to filter out noisy contours)
-                # and be convex.
-                # Note: absolute value of an area is used because
-                # area may be positive or negative - in accordance with the
-                # contour orientation
-                if(result.total == 4 and
-                    abs(cv.ContourArea(result)) > 1000 and
-                    cv.CheckContourConvexity(result)):
-                    s = 0
-                    for i in range(5):
-                        # find minimum angle between joint
-                        # edges (maximum of cosine)
-                        if(i >= 2):
-                            t = abs(angle(result[i], result[i-2], result[i-1]))
-                            if s<t:
-                                s=t
-                    # if cosines of all angles are small
-                    # (all angles are ~90 degree) then write quandrange
-                    # vertices to resultant sequence
-                    if(s < 0.3):
-                        for i in range(4):
-                            squares.append(result[i])
-
-    return squares
-
-# the function draws all the squares in the image
-def drawSquares(img, squares):
-    cpy = cv.CloneImage(img)
-    # read 4 sequence elements at a time (all vertices of a square)
-    i=0
-    while i<squares.total:
-        pt = []
-        # read 4 vertices
-        pt.append(squares[i])
-        pt.append(squares[i+1])
-        pt.append(squares[i+2])
-        pt.append(squares[i+3])
-
-        # draw the square as a closed polyline
-        cv.PolyLine(cpy, [pt], 1, cv.CV_RGB(0, 255, 0), 3, cv. CV_AA, 0)
-        i+=4
-
-    # show the resultant image
-    cv.ShowImage(wndname, cpy)
-
-def on_trackbar(a):
-    if(img):
-        drawSquares(img, findSquares4(img, storage))
-
-names =  ["../c/pic1.png", "../c/pic2.png", "../c/pic3.png",
-          "../c/pic4.png", "../c/pic5.png", "../c/pic6.png" ]
-
-if __name__ == "__main__":
-    # create memory storage that will contain all the dynamic data
-    storage = cv.CreateMemStorage(0)
-    for name in names:
-        img0 = cv.LoadImage(name, 1)
-        if not img0:
-            print "Couldn't load %s" % name
-            continue
-        img = cv.CloneImage(img0)
-        # create window and a trackbar (slider) with parent "image" and set callback
-        # (the slider regulates upper threshold, passed to Canny edge detector)
-        cv.NamedWindow(wndname, 1)
-        cv.CreateTrackbar("canny thresh", wndname, thresh, 1000, on_trackbar)
-        # force the image processing
-        on_trackbar(0)
-        # wait for key.
-        # Also the function cv.WaitKey takes care of event processing
-        c = cv.WaitKey(0) % 0x100
-        # clear memory storage - reset free space position
-        cv.ClearMemStorage(storage)
-        if(c == '\x1b'):
-            break
-    cv.DestroyWindow(wndname)
diff --git a/samples/python/watershed.py b/samples/python/watershed.py
deleted file mode 100755
index c1464227c6..0000000000
--- a/samples/python/watershed.py
+++ /dev/null
@@ -1,109 +0,0 @@
-#!/usr/bin/python
-import urllib2
-import sys
-import cv2.cv as cv
-
-class Sketcher:
-    def __init__(self, windowname, dests):
-        self.prev_pt = None
-        self.windowname = windowname
-        self.dests = dests
-        cv.SetMouseCallback(self.windowname, self.on_mouse)
-
-    def on_mouse(self, event, x, y, flags, param):
-        pt = (x, y)
-        if event == cv.CV_EVENT_LBUTTONUP or not (flags & cv.CV_EVENT_FLAG_LBUTTON):
-            self.prev_pt = None
-        elif event == cv.CV_EVENT_LBUTTONDOWN:
-            self.prev_pt = pt
-        elif event == cv.CV_EVENT_MOUSEMOVE and (flags & cv.CV_EVENT_FLAG_LBUTTON) :
-            if self.prev_pt:
-                for dst in self.dests:
-                    cv.Line(dst, self.prev_pt, pt, cv.ScalarAll(255), 5, 8, 0)
-            self.prev_pt = pt
-            cv.ShowImage(self.windowname, img)
-
-if __name__ == "__main__":
-    if len(sys.argv) > 1:
-        img0 = cv.LoadImage( sys.argv[1], cv.CV_LOAD_IMAGE_COLOR)
-    else:
-        url = 'http://code.opencv.org/projects/opencv/repository/revisions/master/raw/samples/c/fruits.jpg'
-        filedata = urllib2.urlopen(url).read()
-        imagefiledata = cv.CreateMatHeader(1, len(filedata), cv.CV_8UC1)
-        cv.SetData(imagefiledata, filedata, len(filedata))
-        img0 = cv.DecodeImage(imagefiledata, cv.CV_LOAD_IMAGE_COLOR)
-
-    rng = cv.RNG(-1)
-
-    print "Hot keys:"
-    print "\tESC - quit the program"
-    print "\tr - restore the original image"
-    print "\tw - run watershed algorithm"
-    print "\t  (before that, roughly outline several markers on the image)"
-
-    cv.NamedWindow("image", 1)
-    cv.NamedWindow("watershed transform", 1)
-
-    img = cv.CloneImage(img0)
-    img_gray = cv.CloneImage(img0)
-    wshed = cv.CloneImage(img0)
-    marker_mask = cv.CreateImage(cv.GetSize(img), 8, 1)
-    markers = cv.CreateImage(cv.GetSize(img), cv.IPL_DEPTH_32S, 1)
-
-    cv.CvtColor(img, marker_mask, cv.CV_BGR2GRAY)
-    cv.CvtColor(marker_mask, img_gray, cv.CV_GRAY2BGR)
-
-    cv.Zero(marker_mask)
-    cv.Zero(wshed)
-
-    cv.ShowImage("image", img)
-    cv.ShowImage("watershed transform", wshed)
-
-    sk = Sketcher("image", [img, marker_mask])
-
-    while True:
-        c = cv.WaitKey(0) % 0x100
-        if c == 27 or c == ord('q'):
-            break
-        if c == ord('r'):
-            cv.Zero(marker_mask)
-            cv.Copy(img0, img)
-            cv.ShowImage("image", img)
-        if c == ord('w'):
-            storage = cv.CreateMemStorage(0)
-            #cv.SaveImage("wshed_mask.png", marker_mask)
-            #marker_mask = cv.LoadImage("wshed_mask.png", 0)
-            contours = cv.FindContours(marker_mask, storage, cv.CV_RETR_CCOMP, cv.CV_CHAIN_APPROX_SIMPLE)
-            def contour_iterator(contour):
-                while contour:
-                    yield contour
-                    contour = contour.h_next()
-
-            cv.Zero(markers)
-            comp_count = 0
-            for c in contour_iterator(contours):
-                cv.DrawContours(markers,
-                                c,
-                                cv.ScalarAll(comp_count + 1),
-                                cv.ScalarAll(comp_count + 1),
-                                -1,
-                                -1,
-                                8)
-                comp_count += 1
-
-            cv.Watershed(img0, markers)
-
-            cv.Set(wshed, cv.ScalarAll(255))
-
-            # paint the watershed image
-            color_tab = [(cv.RandInt(rng) % 180 + 50, cv.RandInt(rng) % 180 + 50, cv.RandInt(rng) % 180 + 50) for i in range(comp_count)]
-            for j in range(markers.height):
-                for i in range(markers.width):
-                    idx = markers[j, i]
-                    if idx != -1:
-                        wshed[j, i] = color_tab[int(idx - 1)]
-
-            cv.AddWeighted(wshed, 0.5, img_gray, 0.5, 0, wshed)
-            cv.ShowImage("watershed transform", wshed)
-    cv.DestroyAllWindows()
-
diff --git a/samples/python2/_coverage.py b/samples/python2/_coverage.py
index 80edffb997..4b23370b50 100755
--- a/samples/python2/_coverage.py
+++ b/samples/python2/_coverage.py
@@ -24,8 +24,3 @@ if __name__ == '__main__':
 
     r = 1.0 * len(cv2_used) / len(cv2_callable)
     print '\ncv2 api coverage: %d / %d  (%.1f%%)' % ( len(cv2_used), len(cv2_callable), r*100 )
-
-    print '\nold (cv) symbols:'
-    for s in found:
-        if s.startswith('cv.'):
-            print s
diff --git a/samples/python2/asift.py b/samples/python2/asift.py
index bdcd3c989a..ae044d527b 100755
--- a/samples/python2/asift.py
+++ b/samples/python2/asift.py
@@ -119,19 +119,19 @@ if __name__ == '__main__':
     img1 = cv2.imread(fn1, 0)
     img2 = cv2.imread(fn2, 0)
     detector, matcher = init_feature(feature_name)
-    
+
     if img1 is None:
         print 'Failed to load fn1:', fn1
         sys.exit(1)
-        
+
     if img2 is None:
         print 'Failed to load fn2:', fn2
         sys.exit(1)
-    
+
     if detector is None:
         print 'unknown feature:', feature_name
         sys.exit(1)
-        
+
     print 'using', feature_name
 
     pool=ThreadPool(processes = cv2.getNumberOfCPUs())
@@ -158,4 +158,3 @@ if __name__ == '__main__':
     match_and_draw('affine find_obj')
     cv2.waitKey()
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/browse.py b/samples/python2/browse.py
index 81ef33bada..462cd17626 100755
--- a/samples/python2/browse.py
+++ b/samples/python2/browse.py
@@ -37,7 +37,7 @@ if __name__ == '__main__':
         img = np.zeros((sz, sz), np.uint8)
         track = np.cumsum(np.random.rand(500000, 2)-0.5, axis=0)
         track = np.int32(track*10 + (sz/2, sz/2))
-        cv2.polylines(img, [track], 0, 255, 1, cv2.CV_AA)
+        cv2.polylines(img, [track], 0, 255, 1, cv2.LINE_AA)
 
 
     small = img
diff --git a/samples/python2/calibrate.py b/samples/python2/calibrate.py
index e23cc323cb..2c759ff972 100755
--- a/samples/python2/calibrate.py
+++ b/samples/python2/calibrate.py
@@ -70,4 +70,3 @@ if __name__ == '__main__':
     print "camera matrix:\n", camera_matrix
     print "distortion coefficients: ", dist_coefs.ravel()
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/camshift.py b/samples/python2/camshift.py
index 910fd82ad2..6e9402095d 100755
--- a/samples/python2/camshift.py
+++ b/samples/python2/camshift.py
@@ -102,7 +102,7 @@ class App(object):
                     vis[:] = prob[...,np.newaxis]
                 try:
                     cv2.ellipse(vis, track_box, (0, 0, 255), 2)
-                except: 
+                except:
                     print track_box
 
             cv2.imshow('camshift', vis)
@@ -119,8 +119,7 @@ if __name__ == '__main__':
     import sys
     try:
         video_src = sys.argv[1]
-    except: 
+    except:
         video_src = 0
     print __doc__
     App(video_src).run()
-
diff --git a/samples/python2/coherence.py b/samples/python2/coherence.py
index 2db26e63d5..92122e876d 100755
--- a/samples/python2/coherence.py
+++ b/samples/python2/coherence.py
@@ -40,9 +40,9 @@ def coherence_filter(img, sigma = 11, str_sigma = 11, blend = 0.5, iter_n = 4):
 
 if __name__ == '__main__':
     import sys
-    try: 
+    try:
         fn = sys.argv[1]
-    except: 
+    except:
         fn = '../cpp/baboon.jpg'
 
     src = cv2.imread(fn)
diff --git a/samples/python2/common.py b/samples/python2/common.py
index 5996695943..e418b703c8 100755
--- a/samples/python2/common.py
+++ b/samples/python2/common.py
@@ -1,3 +1,5 @@
+#!/usr/bin/env python
+
 '''
 This module contains some common routines used by other samples.
 '''
@@ -69,8 +71,8 @@ def mtx2rvec(R):
     return axis * np.arctan2(s, c)
 
 def draw_str(dst, (x, y), s):
-    cv2.putText(dst, s, (x+1, y+1), cv2.FONT_HERSHEY_PLAIN, 1.0, (0, 0, 0), thickness = 2, lineType=cv2.CV_AA)
-    cv2.putText(dst, s, (x, y), cv2.FONT_HERSHEY_PLAIN, 1.0, (255, 255, 255), lineType=cv2.CV_AA)
+    cv2.putText(dst, s, (x+1, y+1), cv2.FONT_HERSHEY_PLAIN, 1.0, (0, 0, 0), thickness = 2, lineType=cv2.LINE_AA)
+    cv2.putText(dst, s, (x, y), cv2.FONT_HERSHEY_PLAIN, 1.0, (255, 255, 255), lineType=cv2.LINE_AA)
 
 class Sketcher:
     def __init__(self, windowname, dests, colors_func):
@@ -215,4 +217,3 @@ def draw_keypoints(vis, keypoints, color = (0, 255, 255)):
     for kp in keypoints:
             x, y = kp.pt
             cv2.circle(vis, (int(x), int(y)), 2, color)
-
diff --git a/samples/python2/contours.py b/samples/python2/contours.py
index f81f21231c..075401b33b 100755
--- a/samples/python2/contours.py
+++ b/samples/python2/contours.py
@@ -53,7 +53,7 @@ if __name__ == '__main__':
         vis = np.zeros((h, w, 3), np.uint8)
         levels = levels - 3
         cv2.drawContours( vis, contours, (-1, 3)[levels <= 0], (128,255,255),
-            3, cv2.CV_AA, hierarchy, abs(levels) )
+            3, cv2.LINE_AA, hierarchy, abs(levels) )
         cv2.imshow('contours', vis)
     update(3)
     cv2.createTrackbar( "levels+3", "contours", 3, 7, update )
diff --git a/samples/python2/deconvolution.py b/samples/python2/deconvolution.py
index 8650e7b0a0..abc8dc181f 100755
--- a/samples/python2/deconvolution.py
+++ b/samples/python2/deconvolution.py
@@ -57,7 +57,7 @@ def motion_kernel(angle, d, sz=65):
 
 def defocus_kernel(d, sz=65):
     kern = np.zeros((sz, sz), np.uint8)
-    cv2.circle(kern, (sz, sz), d, 255, -1, cv2.CV_AA, shift=1)
+    cv2.circle(kern, (sz, sz), d, 255, -1, cv2.LINE_AA, shift=1)
     kern = np.float32(kern) / 255.0
     return kern
 
@@ -69,7 +69,7 @@ if __name__ == '__main__':
     opts = dict(opts)
     try:
         fn = args[0]
-    except: 
+    except:
         fn = 'data/licenseplate_motion.jpg'
 
     win = 'deconvolution'
@@ -78,7 +78,7 @@ if __name__ == '__main__':
     if img is None:
         print 'Failed to load fn1:', fn1
         sys.exit(1)
-    
+
     img = np.float32(img)/255.0
     cv2.imshow('input', img)
 
@@ -125,4 +125,3 @@ if __name__ == '__main__':
         if ch == ord(' '):
             defocus = not defocus
             update(None)
-
diff --git a/samples/python2/demo.py b/samples/python2/demo.py
index c69de825b7..03d624ddcf 100755
--- a/samples/python2/demo.py
+++ b/samples/python2/demo.py
@@ -141,7 +141,7 @@ class App:
         count = tk.IntVar()
         while True:
             match_index = text.search(pattern, 'matchPos', count=count, regexp=regexp, stopindex='end')
-            if not match_index: 
+            if not match_index:
                 break
             end_index = text.index( "%s+%sc" % (match_index, count.get()) )
             text.mark_set('matchPos', end_index)
@@ -161,4 +161,3 @@ class App:
 
 if __name__ == '__main__':
     App().run()
-
diff --git a/samples/python2/dft.py b/samples/python2/dft.py
new file mode 100755
index 0000000000..792ebe1234
--- /dev/null
+++ b/samples/python2/dft.py
@@ -0,0 +1,100 @@
+#!/usr/bin/env python
+
+import cv2
+import numpy as np
+import sys
+
+
+def shift_dft(src, dst=None):
+    '''
+        Rearrange the quadrants of Fourier image so that the origin is at
+        the image center. Swaps quadrant 1 with 3, and 2 with 4.
+
+        src and dst arrays must be equal size & type
+    '''
+
+    if dst is None:
+        dst = np.empty(src.shape, src.dtype)
+    elif src.shape != dst.shape:
+        raise ValueError("src and dst must have equal sizes")
+    elif src.dtype != dst.dtype:
+        raise TypeError("src and dst must have equal types")
+
+    if src is dst:
+        ret = np.empty(src.shape, src.dtype)
+    else:
+        ret = dst
+
+    h, w = src.shape[:2]
+
+    cx1 = cx2 = w/2
+    cy1 = cy2 = h/2
+
+    # if the size is odd, then adjust the bottom/right quadrants
+    if w % 2 != 0:
+        cx2 += 1
+    if h % 2 != 0:
+        cy2 += 1
+
+    # swap quadrants
+
+    # swap q1 and q3
+    ret[h-cy1:, w-cx1:] = src[0:cy1 , 0:cx1 ]   # q1 -> q3
+    ret[0:cy2 , 0:cx2 ] = src[h-cy2:, w-cx2:]   # q3 -> q1
+
+    # swap q2 and q4
+    ret[0:cy2 , w-cx2:] = src[h-cy2:, 0:cx2 ]   # q2 -> q4
+    ret[h-cy1:, 0:cx1 ] = src[0:cy1 , w-cx1:]   # q4 -> q2
+
+    if src is dst:
+        dst[:,:] = ret
+
+    return dst
+
+if __name__ == "__main__":
+
+    if len(sys.argv)>1:
+        im = cv2.imread(sys.argv[1])
+    else :
+        im = cv2.imread('../c/baboon.jpg')
+        print "usage : python dft.py <image_file>"
+
+    # convert to grayscale
+    im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
+    h, w = im.shape[:2]
+
+    realInput = im.astype(np.float64)
+
+    # perform an optimally sized dft
+    dft_M = cv2.getOptimalDFTSize(w)
+    dft_N = cv2.getOptimalDFTSize(h)
+
+    # copy A to dft_A and pad dft_A with zeros
+    dft_A = np.zeros((dft_N, dft_M, 2), dtype=np.float64)
+    dft_A[:h, :w, 0] = realInput
+
+    # no need to pad bottom part of dft_A with zeros because of
+    # use of nonzeroRows parameter in cv2.dft()
+    cv2.dft(dft_A, dst=dft_A, nonzeroRows=h)
+
+    cv2.imshow("win", im)
+
+    # Split fourier into real and imaginary parts
+    image_Re, image_Im = cv2.split(dft_A)
+
+    # Compute the magnitude of the spectrum Mag = sqrt(Re^2 + Im^2)
+    magnitude = cv2.sqrt(image_Re**2.0 + image_Im**2.0)
+
+    # Compute log(1 + Mag)
+    log_spectrum = cv2.log(1.0 + magnitude)
+
+    # Rearrange the quadrants of Fourier image so that the origin is at
+    # the image center
+    shift_dft(log_spectrum, log_spectrum)
+
+    # normalize and display the results as rgb
+    cv2.normalize(log_spectrum, log_spectrum, 0.0, 1.0, cv2.NORM_MINMAX)
+    cv2.imshow("magnitude", log_spectrum)
+
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
diff --git a/samples/python2/distrans.py b/samples/python2/distrans.py
index 083c8d65b6..dc2a137c51 100755
--- a/samples/python2/distrans.py
+++ b/samples/python2/distrans.py
@@ -14,7 +14,6 @@ Keys:
 
 import numpy as np
 import cv2
-import cv2.cv as cv
 
 from common import make_cmap
 
@@ -30,7 +29,7 @@ if __name__ == '__main__':
     if img is None:
         print 'Failed to load fn:', fn
         sys.exit(1)
-        
+
     cm = make_cmap('jet')
     need_update = True
     voronoi = False
@@ -40,7 +39,7 @@ if __name__ == '__main__':
         need_update = False
         thrs = cv2.getTrackbarPos('threshold', 'distrans')
         mark = cv2.Canny(img, thrs, 3*thrs)
-        dist, labels = cv2.distanceTransformWithLabels(~mark, cv.CV_DIST_L2, 5)
+        dist, labels = cv2.distanceTransformWithLabels(~mark, cv2.DIST_L2, 5)
         if voronoi:
             vis = cm[np.uint8(labels)]
         else:
@@ -68,4 +67,3 @@ if __name__ == '__main__':
         if need_update:
             update()
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/edge.py b/samples/python2/edge.py
index 2cc1631a28..bd0c8bde79 100755
--- a/samples/python2/edge.py
+++ b/samples/python2/edge.py
@@ -49,4 +49,3 @@ if __name__ == '__main__':
         if ch == 27:
             break
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/facedetect.py b/samples/python2/facedetect.py
index 1b6e9f1f79..ab82f9a9f5 100755
--- a/samples/python2/facedetect.py
+++ b/samples/python2/facedetect.py
@@ -2,7 +2,6 @@
 
 import numpy as np
 import cv2
-import cv2.cv as cv
 
 # local modules
 from video import create_capture
@@ -13,7 +12,7 @@ USAGE: facedetect.py [--cascade <cascade_fn>] [--nested-cascade <cascade_fn>] [<
 '''
 
 def detect(img, cascade):
-    rects = cascade.detectMultiScale(img, scaleFactor=1.3, minNeighbors=4, minSize=(30, 30), flags = cv.CV_HAAR_SCALE_IMAGE)
+    rects = cascade.detectMultiScale(img, scaleFactor=1.3, minNeighbors=4, minSize=(30, 30), flags = cv2.CASCADE_SCALE_IMAGE)
     if len(rects) == 0:
         return []
     rects[:,2:] += rects[:,:2]
@@ -63,4 +62,3 @@ if __name__ == '__main__':
         if 0xFF & cv2.waitKey(5) == 27:
             break
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/facerec_demo.py b/samples/python2/facerec_demo.py
index 1b0adcc217..9eeb04e0bd 100755
--- a/samples/python2/facerec_demo.py
+++ b/samples/python2/facerec_demo.py
@@ -31,6 +31,11 @@
 # ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 # POSSIBILITY OF SUCH DAMAGE.
 
+# ------------------------------------------------------------------------------------------------
+# Note:
+# When using the FaceRecognizer interface in combination with Python, please stick to Python 2.
+# Some underlying scripts like create_csv will not work in other versions, like Python 3.
+# ------------------------------------------------------------------------------------------------
 
 import os
 import sys
diff --git a/samples/python2/find_obj.py b/samples/python2/find_obj.py
index ccab392144..908da68feb 100755
--- a/samples/python2/find_obj.py
+++ b/samples/python2/find_obj.py
@@ -143,15 +143,15 @@ if __name__ == '__main__':
     if img1 is None:
         print 'Failed to load fn1:', fn1
         sys.exit(1)
-        
+
     if img2 is None:
         print 'Failed to load fn2:', fn2
         sys.exit(1)
-    
+
     if detector is None:
         print 'unknown feature:', feature_name
         sys.exit(1)
-    
+
     print 'using', feature_name
 
     kp1, desc1 = detector.detectAndCompute(img1, None)
diff --git a/samples/python2/fitline.py b/samples/python2/fitline.py
index 876435c9c3..82b45b0232 100755
--- a/samples/python2/fitline.py
+++ b/samples/python2/fitline.py
@@ -42,7 +42,7 @@ def sample_line(p1, p2, n, noise=0.0):
     t = np.random.rand(n,1)
     return p1 + (p2-p1)*t + np.random.normal(size=(n, 2))*noise
 
-dist_func_names = it.cycle('CV_DIST_L2 CV_DIST_L1 CV_DIST_L12 CV_DIST_FAIR CV_DIST_WELSCH CV_DIST_HUBER'.split())
+dist_func_names = it.cycle('DIST_L2 DIST_L1 DIST_L12 DIST_FAIR DIST_WELSCH DIST_HUBER'.split())
 cur_func_name = dist_func_names.next()
 
 def update(_=None):
@@ -63,7 +63,7 @@ def update(_=None):
             cv2.circle(img, toint(p), 2, (255, 255, 255), -1)
         for p in outliers:
             cv2.circle(img, toint(p), 2, (64, 64, 255), -1)
-        func = getattr(cv2.cv, cur_func_name)
+        func = getattr(cv2, cur_func_name)
         vx, vy, cx, cy = cv2.fitLine(np.float32(points), func, 0, 0.01, 0.01)
         cv2.line(img, (int(cx-vx*w), int(cy-vy*w)), (int(cx+vx*w), int(cy+vy*w)), (0, 0, 255))
 
diff --git a/samples/python2/gabor_threads.py b/samples/python2/gabor_threads.py
index 1f56a9beb5..6d10ffdb4b 100755
--- a/samples/python2/gabor_threads.py
+++ b/samples/python2/gabor_threads.py
@@ -51,14 +51,14 @@ if __name__ == '__main__':
     print __doc__
     try:
         img_fn = sys.argv[1]
-    except: 
+    except:
         img_fn = '../cpp/baboon.jpg'
 
     img = cv2.imread(img_fn)
     if img is None:
         print 'Failed to load image file:', img_fn
         sys.exit(1)
-    
+
     filters = build_filters()
 
     with Timer('running single-threaded'):
diff --git a/samples/python2/gaussian_mix.py b/samples/python2/gaussian_mix.py
index 704d30cf6b..6b1b9c24d1 100755
--- a/samples/python2/gaussian_mix.py
+++ b/samples/python2/gaussian_mix.py
@@ -23,7 +23,7 @@ def draw_gaussain(img, mean, cov, color):
     w, u, vt = cv2.SVDecomp(cov)
     ang = np.arctan2(u[1, 0], u[0, 0])*(180/np.pi)
     s1, s2 = np.sqrt(w)*3.0
-    cv2.ellipse(img, (x, y), (s1, s2), ang, 0, 360, color, 1, cv2.CV_AA)
+    cv2.ellipse(img, (x, y), (s1, s2), ang, 0, 360, color, 1, cv2.LINE_AA)
 
 
 if __name__ == '__main__':
diff --git a/samples/python2/grabcut.py b/samples/python2/grabcut.py
new file mode 100644
index 0000000000..1d5b823ddd
--- /dev/null
+++ b/samples/python2/grabcut.py
@@ -0,0 +1,174 @@
+#!/usr/bin/env python
+'''
+===============================================================================
+Interactive Image Segmentation using GrabCut algorithm.
+
+This sample shows interactive image segmentation using grabcut algorithm.
+
+USAGE :
+    python grabcut.py <filename>
+
+README FIRST:
+    Two windows will show up, one for input and one for output.
+
+    At first, in input window, draw a rectangle around the object using
+mouse right button. Then press 'n' to segment the object (once or a few times)
+For any finer touch-ups, you can press any of the keys below and draw lines on
+the areas you want. Then again press 'n' for updating the output.
+
+Key '0' - To select areas of sure background
+Key '1' - To select areas of sure foreground
+Key '2' - To select areas of probable background
+Key '3' - To select areas of probable foreground
+
+Key 'n' - To update the segmentation
+Key 'r' - To reset the setup
+Key 's' - To save the results
+===============================================================================
+'''
+
+import numpy as np
+import cv2
+import sys
+
+BLUE = [255,0,0]        # rectangle color
+RED = [0,0,255]         # PR BG
+GREEN = [0,255,0]       # PR FG
+BLACK = [0,0,0]         # sure BG
+WHITE = [255,255,255]   # sure FG
+
+DRAW_BG = {'color' : BLACK, 'val' : 0}
+DRAW_FG = {'color' : WHITE, 'val' : 1}
+DRAW_PR_FG = {'color' : GREEN, 'val' : 3}
+DRAW_PR_BG = {'color' : RED, 'val' : 2}
+
+# setting up flags
+rect = (0,0,1,1)
+drawing = False         # flag for drawing curves
+rectangle = False       # flag for drawing rect
+rect_over = False       # flag to check if rect drawn
+rect_or_mask = 100      # flag for selecting rect or mask mode
+value = DRAW_FG         # drawing initialized to FG
+thickness = 3           # brush thickness
+
+def onmouse(event,x,y,flags,param):
+    global img,img2,drawing,value,mask,rectangle,rect,rect_or_mask,ix,iy,rect_over
+
+    # Draw Rectangle
+    if event == cv2.EVENT_RBUTTONDOWN:
+        rectangle = True
+        ix,iy = x,y
+
+    elif event == cv2.EVENT_MOUSEMOVE:
+        if rectangle == True:
+            img = img2.copy()
+            cv2.rectangle(img,(ix,iy),(x,y),BLUE,2)
+            rect = (ix,iy,abs(ix-x),abs(iy-y))
+            rect_or_mask = 0
+
+    elif event == cv2.EVENT_RBUTTONUP:
+        rectangle = False
+        rect_over = True
+        cv2.rectangle(img,(ix,iy),(x,y),BLUE,2)
+        rect = (ix,iy,abs(ix-x),abs(iy-y))
+        rect_or_mask = 0
+        print " Now press the key 'n' a few times until no further change \n"
+
+    # draw touchup curves
+
+    if event == cv2.EVENT_LBUTTONDOWN:
+        if rect_over == False:
+            print "first draw rectangle \n"
+        else:
+            drawing = True
+            cv2.circle(img,(x,y),thickness,value['color'],-1)
+            cv2.circle(mask,(x,y),thickness,value['val'],-1)
+
+    elif event == cv2.EVENT_MOUSEMOVE:
+        if drawing == True:
+            cv2.circle(img,(x,y),thickness,value['color'],-1)
+            cv2.circle(mask,(x,y),thickness,value['val'],-1)
+
+    elif event == cv2.EVENT_LBUTTONUP:
+        if drawing == True:
+            drawing = False
+            cv2.circle(img,(x,y),thickness,value['color'],-1)
+            cv2.circle(mask,(x,y),thickness,value['val'],-1)
+
+# print documentation
+print __doc__
+
+# Loading images
+if len(sys.argv) == 2:
+    filename = sys.argv[1] # for drawing purposes
+else:
+    print "No input image given, so loading default image, lena.jpg \n"
+    print "Correct Usage : python grabcut.py <filename> \n"
+    filename = '../cpp/lena.jpg'
+
+img = cv2.imread(filename)
+img2 = img.copy()                               # a copy of original image
+mask = np.zeros(img.shape[:2],dtype = np.uint8) # mask initialized to PR_BG
+output = np.zeros(img.shape,np.uint8)           # output image to be shown
+
+# input and output windows
+cv2.namedWindow('output')
+cv2.namedWindow('input')
+cv2.setMouseCallback('input',onmouse)
+cv2.moveWindow('input',img.shape[1]+10,90)
+
+print " Instructions : \n"
+print " Draw a rectangle around the object using right mouse button \n"
+
+while(1):
+
+    cv2.imshow('output',output)
+    cv2.imshow('input',img)
+    k = 0xFF & cv2.waitKey(1)
+
+    # key bindings
+    if k == 27:         # esc to exit
+        break
+    elif k == ord('0'): # BG drawing
+        print " mark background regions with left mouse button \n"
+        value = DRAW_BG
+    elif k == ord('1'): # FG drawing
+        print " mark foreground regions with left mouse button \n"
+        value = DRAW_FG
+    elif k == ord('2'): # PR_BG drawing
+        value = DRAW_PR_BG
+    elif k == ord('3'): # PR_FG drawing
+        value = DRAW_PR_FG
+    elif k == ord('s'): # save image
+        bar = np.zeros((img.shape[0],5,3),np.uint8)
+        res = np.hstack((img2,bar,img,bar,output))
+        cv2.imwrite('grabcut_output.png',res)
+        print " Result saved as image \n"
+    elif k == ord('r'): # reset everything
+        print "resetting \n"
+        rect = (0,0,1,1)
+        drawing = False
+        rectangle = False
+        rect_or_mask = 100
+        rect_over = False
+        value = DRAW_FG
+        img = img2.copy()
+        mask = np.zeros(img.shape[:2],dtype = np.uint8) # mask initialized to PR_BG
+        output = np.zeros(img.shape,np.uint8)           # output image to be shown
+    elif k == ord('n'): # segment the image
+        print """ For finer touchups, mark foreground and background after pressing keys 0-3
+        and again press 'n' \n"""
+        if (rect_or_mask == 0):         # grabcut with rect
+            bgdmodel = np.zeros((1,65),np.float64)
+            fgdmodel = np.zeros((1,65),np.float64)
+            cv2.grabCut(img2,mask,rect,bgdmodel,fgdmodel,1,cv2.GC_INIT_WITH_RECT)
+            rect_or_mask = 1
+        elif rect_or_mask == 1:         # grabcut with mask
+            bgdmodel = np.zeros((1,65),np.float64)
+            fgdmodel = np.zeros((1,65),np.float64)
+            cv2.grabCut(img2,mask,rect,bgdmodel,fgdmodel,1,cv2.GC_INIT_WITH_MASK)
+
+    mask2 = np.where((mask==1) + (mask==3),255,0).astype('uint8')
+    output = cv2.bitwise_and(img2,img2,mask=mask2)
+
+cv2.destroyAllWindows()
diff --git a/samples/python2/hist.py b/samples/python2/hist.py
index c5e21c4d71..41eec1d3a3 100755
--- a/samples/python2/hist.py
+++ b/samples/python2/hist.py
@@ -61,7 +61,7 @@ if __name__ == '__main__':
         print "usage : python hist.py <image_file>"
 
     im = cv2.imread(fname)
-    
+
     if im is None:
         print 'Failed to load image file:', fname
         sys.exit(1)
@@ -114,4 +114,3 @@ if __name__ == '__main__':
             cv2.destroyAllWindows()
             break
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/houghcircles.py b/samples/python2/houghcircles.py
new file mode 100755
index 0000000000..38ed6f3032
--- /dev/null
+++ b/samples/python2/houghcircles.py
@@ -0,0 +1,33 @@
+#!/usr/bin/python
+
+'''
+This example illustrates how to use cv2.HoughCircles() function.
+Usage: ./houghcircles.py [<image_name>]
+image argument defaults to ../cpp/board.jpg
+'''
+
+import cv2
+import numpy as np
+import sys
+
+
+print __doc__
+try:
+    fn = sys.argv[1]
+except:
+    fn = "../cpp/board.jpg"
+
+src = cv2.imread(fn, 1)
+img = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY)
+img = cv2.medianBlur(img, 5)
+cimg = src.copy() # numpy function
+
+circles = cv2.HoughCircles(img, cv2.HOUGH_GRADIENT, 1, 10, np.array([]), 100, 30, 1, 30)
+a, b, c = circles.shape
+for i in range(b):
+    cv2.circle(cimg, (circles[0][i][0], circles[0][i][1]), circles[0][i][2], (0, 0, 255), 3, cv2.LINE_AA)
+    cv2.circle(cimg, (circles[0][i][0], circles[0][i][1]), 2, (0, 255, 0), 3, cv2.LINE_AA) # draw center of circle
+
+cv2.imshow("source", src)
+cv2.imshow("detected circles", cimg)
+cv2.waitKey(0)
diff --git a/samples/python2/houghlines.py b/samples/python2/houghlines.py
new file mode 100755
index 0000000000..cf67945b19
--- /dev/null
+++ b/samples/python2/houghlines.py
@@ -0,0 +1,41 @@
+#!/usr/bin/python
+'''
+This example illustrates how to use Hough Transform to find lines
+Usage: ./houghlines.py [<image_name>]
+image argument defaults to ../cpp/pic1.png
+'''
+import cv2
+import numpy as np
+import sys
+import math
+
+try:
+    fn = sys.argv[1]
+except:
+    fn = "../cpp/pic1.png"
+print __doc__
+src = cv2.imread(fn)
+dst = cv2.Canny(src, 50, 200)
+cdst = cv2.cvtColor(dst, cv2.COLOR_GRAY2BGR)
+
+# HoughLines()
+# lines = cv2.HoughLines(dst, 1, math.pi/180.0, 50, np.array([]), 0, 0)
+# a,b,c = lines.shape
+# for i in range(b):
+#   rho = lines[0][i][0]
+#   theta = lines[0][i][1]
+#   a = math.cos(theta)
+#   b = math.sin(theta)
+#   x0, y0 = a*rho, b*rho
+#   pt1 = ( int(x0+1000*(-b)), int(y0+1000*(a)) )
+#   pt2 = ( int(x0-1000*(-b)), int(y0-1000*(a)) )
+#   cv2.line(cdst, pt1, pt2, (0, 0, 255), 3, cv2.LINE_AA)
+
+lines = cv2.HoughLinesP(dst, 1, math.pi/180.0, 50, np.array([]), 50, 10)
+a,b,c = lines.shape
+for i in range(b):
+    cv2.line(cdst, (lines[0][i][0], lines[0][i][1]), (lines[0][i][2], lines[0][i][3]), (0, 0, 255), 3, cv2.LINE_AA)
+
+cv2.imshow("source", src)
+cv2.imshow("detected lines", cdst)
+cv2.waitKey(0)
diff --git a/samples/python2/inpaint.py b/samples/python2/inpaint.py
index 15497d406d..5044afb899 100755
--- a/samples/python2/inpaint.py
+++ b/samples/python2/inpaint.py
@@ -23,16 +23,16 @@ if __name__ == '__main__':
     import sys
     try:
         fn = sys.argv[1]
-    except: 
+    except:
         fn = '../cpp/fruits.jpg'
-        
+
     print __doc__
 
     img = cv2.imread(fn)
     if img is None:
         print 'Failed to load image file:', fn
         sys.exit(1)
-        
+
     img_mark = img.copy()
     mark = np.zeros(img.shape[:2], np.uint8)
     sketch = Sketcher('img', [img_mark, mark], lambda : ((255, 255, 255), 255))
@@ -49,4 +49,3 @@ if __name__ == '__main__':
             mark[:] = 0
             sketch.show()
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/lappyr.py b/samples/python2/lappyr.py
index 929d513da0..0c08484de9 100755
--- a/samples/python2/lappyr.py
+++ b/samples/python2/lappyr.py
@@ -64,5 +64,3 @@ if __name__ == '__main__':
 
         if cv2.waitKey(1) == 27:
             break
-
-
diff --git a/samples/python2/morphology.py b/samples/python2/morphology.py
index cfed9b8b04..d9bdb7f1c2 100755
--- a/samples/python2/morphology.py
+++ b/samples/python2/morphology.py
@@ -27,13 +27,13 @@ if __name__ == '__main__':
         fn = sys.argv[1]
     except:
         fn = '../cpp/baboon.jpg'
-    
+
     img = cv2.imread(fn)
-    
+
     if img is None:
         print 'Failed to load image file:', fn
         sys.exit(1)
-    
+
     cv2.imshow('original', img)
 
     modes = cycle(['erode/dilate', 'open/close', 'blackhat/tophat', 'gradient'])
diff --git a/samples/python2/mouse_and_match.py b/samples/python2/mouse_and_match.py
index 66c3b1af7a..f440033850 100755
--- a/samples/python2/mouse_and_match.py
+++ b/samples/python2/mouse_and_match.py
@@ -9,7 +9,6 @@ Demonstrate using a mouse to interact with an image:
  ESC to exit
 '''
 import numpy as np
-import cv2 as cv
 
 # built-in modules
 import os
@@ -24,27 +23,27 @@ sel = (0,0,0,0)
 
 def onmouse(event, x, y, flags, param):
     global drag_start, sel
-    if event == cv.EVENT_LBUTTONDOWN:
+    if event == cv2.EVENT_LBUTTONDOWN:
         drag_start = x, y
         sel = 0,0,0,0
-    elif event == cv.EVENT_LBUTTONUP:
+    elif event == cv2.EVENT_LBUTTONUP:
         if sel[2] > sel[0] and sel[3] > sel[1]:
             patch = gray[sel[1]:sel[3],sel[0]:sel[2]]
-            result = cv.matchTemplate(gray,patch,cv.TM_CCOEFF_NORMED)
+            result = cv2.matchTemplate(gray,patch,cv2.TM_CCOEFF_NORMED)
             result = np.abs(result)**3
-            val, result = cv.threshold(result, 0.01, 0, cv.THRESH_TOZERO)
-            result8 = cv.normalize(result,None,0,255,cv.NORM_MINMAX,cv.CV_8U)
-            cv.imshow("result", result8)
+            val, result = cv2.threshold(result, 0.01, 0, cv2.THRESH_TOZERO)
+            result8 = cv2.normalize(result,None,0,255,cv2.NORM_MINMAX,cv2.CV_8U)
+            cv2.imshow("result", result8)
         drag_start = None
     elif drag_start:
         #print flags
-        if flags & cv.EVENT_FLAG_LBUTTON:
+        if flags & cv2.EVENT_FLAG_LBUTTON:
             minpos = min(drag_start[0], x), min(drag_start[1], y)
             maxpos = max(drag_start[0], x), max(drag_start[1], y)
             sel = minpos[0], minpos[1], maxpos[0], maxpos[1]
-            img = cv.cvtColor(gray, cv.COLOR_GRAY2BGR)
-            cv.rectangle(img, (sel[0], sel[1]), (sel[2], sel[3]), (0,255,255), 1)
-            cv.imshow("gray", img)
+            img = cv2.cvtColor(gray, cv2.COLOR_GRAY2BGR)
+            cv2.rectangle(img, (sel[0], sel[1]), (sel[2], sel[3]), (0,255,255), 1)
+            cv2.imshow("gray", img)
         else:
             print "selection is complete"
             drag_start = None
@@ -55,21 +54,21 @@ if __name__ == '__main__':
     args = parser.parse_args()
     path = args.input
 
-    cv.namedWindow("gray",1)
-    cv.setMouseCallback("gray", onmouse)
+    cv2.namedWindow("gray",1)
+    cv2.setMouseCallback("gray", onmouse)
     '''Loop through all the images in the directory'''
     for infile in glob.glob( os.path.join(path, '*.*') ):
         ext = os.path.splitext(infile)[1][1:] #get the filename extenstion
         if ext == "png" or ext == "jpg" or ext == "bmp" or ext == "tiff" or ext == "pbm":
             print infile
 
-            img=cv.imread(infile,1)
+            img=cv2.imread(infile,1)
             if img == None:
                 continue
             sel = (0,0,0,0)
             drag_start = None
-            gray=cv.cvtColor(img, cv.COLOR_BGR2GRAY)
-            cv.imshow("gray",gray)
-            if (cv.waitKey() & 255) == 27:
+            gray=cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+            cv2.imshow("gray",gray)
+            if (cv2.waitKey() & 255) == 27:
                 break
-    cv.destroyAllWindows()
+    cv2.destroyAllWindows()
diff --git a/samples/python2/opt_flow.py b/samples/python2/opt_flow.py
index 14efbfa8fa..c8a6086eda 100755
--- a/samples/python2/opt_flow.py
+++ b/samples/python2/opt_flow.py
@@ -85,4 +85,3 @@ if __name__ == '__main__':
                 cur_glitch = img.copy()
             print 'glitch is', ['off', 'on'][show_glitch]
     cv2.destroyAllWindows()
-
diff --git a/samples/python2/texture_flow.py b/samples/python2/texture_flow.py
index 38767fae9b..0e1849ce3f 100755
--- a/samples/python2/texture_flow.py
+++ b/samples/python2/texture_flow.py
@@ -24,7 +24,7 @@ if __name__ == '__main__':
     if img is None:
         print 'Failed to load image file:', fn
         sys.exit(1)
-        
+
     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
     h, w = img.shape[:2]
 
@@ -38,7 +38,7 @@ if __name__ == '__main__':
     points =  np.dstack( np.mgrid[d/2:w:d, d/2:h:d] ).reshape(-1, 2)
     for x, y in points:
        vx, vy = np.int32(flow[y, x]*d)
-       cv2.line(vis, (x-vx, y-vy), (x+vx, y+vy), (0, 0, 0), 1, cv2.CV_AA)
+       cv2.line(vis, (x-vx, y-vy), (x+vx, y+vy), (0, 0, 0), 1, cv2.LINE_AA)
     cv2.imshow('input', img)
     cv2.imshow('flow', vis)
     cv2.waitKey()
diff --git a/samples/python2/turing.py b/samples/python2/turing.py
index e847f014c1..94816e6c31 100755
--- a/samples/python2/turing.py
+++ b/samples/python2/turing.py
@@ -9,7 +9,6 @@ Inspired by http://www.jonathanmccabe.com/Cyclic_Symmetric_Multi-Scale_Turing_Pa
 
 import numpy as np
 import cv2
-import cv2.cv as cv
 from common import draw_str
 import getopt, sys
 from itertools import count
@@ -30,7 +29,7 @@ if __name__ == '__main__':
     out = None
     if '-o' in args:
         fn = args['-o']
-        out = cv2.VideoWriter(args['-o'], cv.CV_FOURCC(*'DIB '), 30.0, (w, h), False)
+        out = cv2.VideoWriter(args['-o'], cv2.VideoWriter_fourcc(*'DIB '), 30.0, (w, h), False)
         print 'writing %s ...' % fn
 
     a = np.zeros((h, w), np.float32)
diff --git a/samples/python2/video.py b/samples/python2/video.py
index 8cbc6f03ba..187bb6d512 100755
--- a/samples/python2/video.py
+++ b/samples/python2/video.py
@@ -105,7 +105,7 @@ class Chess(VideoSynthBase):
         img_quads = cv2.projectPoints(quads.reshape(-1, 3), self.rvec, self.tvec, self.K, self.dist_coef) [0]
         img_quads.shape = quads.shape[:2] + (2,)
         for q in img_quads:
-            cv2.fillConvexPoly(img, np.int32(q*4), color, cv2.CV_AA, shift=2)
+            cv2.fillConvexPoly(img, np.int32(q*4), color, cv2.LINE_AA, shift=2)
 
     def render(self, dst):
         t = self.t
@@ -140,7 +140,7 @@ def create_capture(source = 0, fallback = presets['chess']):
     '''
     source = str(source).strip()
     chunks = source.split(':')
-    # hanlde drive letter ('c:', ...)
+    # handle drive letter ('c:', ...)
     if len(chunks) > 1 and len(chunks[0]) == 1 and chunks[0].isalpha():
         chunks[1] = chunks[0] + ':' + chunks[1]
         del chunks[0]
@@ -159,8 +159,8 @@ def create_capture(source = 0, fallback = presets['chess']):
         cap = cv2.VideoCapture(source)
         if 'size' in params:
             w, h = map(int, params['size'].split('x'))
-            cap.set(cv2.cv.CV_CAP_PROP_FRAME_WIDTH, w)
-            cap.set(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT, h)
+            cap.set(cv2.CAP_PROP_FRAME_WIDTH, w)
+            cap.set(cv2.CAP_PROP_FRAME_HEIGHT, h)
     if cap is None or not cap.isOpened():
         print 'Warning: unable to open video source: ', source
         if fallback is not None:
diff --git a/samples/python2/watershed.py b/samples/python2/watershed.py
index 037af1ec45..c6247da22c 100755
--- a/samples/python2/watershed.py
+++ b/samples/python2/watershed.py
@@ -1,4 +1,4 @@
-#/usr/bin/env python
+#!/usr/bin/env python
 
 '''
 Watershed segmentation
diff --git a/samples/winrt/ImageManipulations/AdvancedCapture.xaml b/samples/winrt/ImageManipulations/AdvancedCapture.xaml
new file mode 100644
index 0000000000..07db96f275
--- /dev/null
+++ b/samples/winrt/ImageManipulations/AdvancedCapture.xaml
@@ -0,0 +1,75 @@
+<!--
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+-->
+
+<common:LayoutAwarePage
+    x:Class="SDKSample.MediaCapture.AdvancedCapture"
+    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
+    xmlns:local="using:$rootsnamespace$"
+    xmlns:common="using:SDKSample.Common"
+    xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
+    xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
+    mc:Ignorable="d">
+
+    <Grid x:Name="LayoutRoot" Background="White" HorizontalAlignment="Left" VerticalAlignment="Top">
+        <Grid.RowDefinitions>
+            <RowDefinition Height="Auto"/>
+            <RowDefinition Height="*"/>
+        </Grid.RowDefinitions>
+        <Grid x:Name="Input" Grid.Row="0">
+            <Grid.RowDefinitions>
+                <RowDefinition Height="Auto"/>
+                <RowDefinition Height="Auto"/>
+                <RowDefinition Height="*"/>
+            </Grid.RowDefinitions>
+            <TextBlock TextWrapping="Wrap" Grid.Row="0"  Text="This scenario shows how to enumerate cameras in the system. Choose a camera from the list to preview, record or take a photo from the chosen camera.  You can add the gray scale effect using the checkbox provided." Style="{StaticResource BasicTextStyle}" HorizontalAlignment="Left"/>
+            <StackPanel Orientation="Horizontal" Grid.Row="1" Margin="0,10,0,0">
+                <ListBox x:Name="EnumedDeviceList2" SelectionChanged="lstEnumedDevices_SelectionChanged" />
+                <Button x:Name="btnStartDevice2" Click="btnStartDevice_Click" IsEnabled="true"  Margin="0,0,10,0" Content="StartDevice"/>
+                <Button x:Name="btnStartPreview2" Click="btnStartPreview_Click" IsEnabled="true"  Margin="0,0,10,0" Content="StartPreview"/>
+                <ComboBox x:Name="EffectTypeCombo" Width="120" SelectedIndex="0">
+                    <ComboBoxItem Content="Preview"/>
+                    <ComboBoxItem Content="Grayscale"/>
+                    <ComboBoxItem Content="Canny"/>
+                    <ComboBoxItem Content="Sobel"/>
+                    <ComboBoxItem Content="Histogram"/>
+                </ComboBox>
+                <Button Content="Apply" HorizontalAlignment="Stretch" VerticalAlignment="Top" Click="Button_Click"/>
+            </StackPanel>
+            <StackPanel x:Name="EffectTypeCombo1" Orientation="Horizontal" Grid.Row="1" Margin="324,5,-324,7"/>
+        </Grid>
+
+        <Grid x:Name="Output" HorizontalAlignment="Left" VerticalAlignment="Top" Grid.Row="1">
+            <StackPanel Orientation="Horizontal" Margin="0,10,0,0">
+                <StackPanel>
+                    <TextBlock Style="{StaticResource BasicTextStyle}"  HorizontalAlignment='Center'  VerticalAlignment='Center'  TextAlignment='Center'	Text='Preview' />
+                    <Canvas x:Name="previewCanvas2" Background="Gray">
+                        <CaptureElement x:Name="previewElement2" />
+                    </Canvas>
+                </StackPanel>
+                <StackPanel/>
+                <StackPanel/>
+            </StackPanel>
+        </Grid>
+
+        <!-- Add Storyboards to the visual states below as necessary for supporting the various layouts -->
+        <VisualStateManager.VisualStateGroups>
+            <VisualStateGroup>
+                <VisualState x:Name="FullScreenLandscape"/>
+                <VisualState x:Name="Filled"/>
+                <VisualState x:Name="FullScreenPortrait"/>
+                <VisualState x:Name="Snapped"/>
+            </VisualStateGroup>
+        </VisualStateManager.VisualStateGroups>
+    </Grid>
+
+</common:LayoutAwarePage>
diff --git a/samples/winrt/ImageManipulations/AdvancedCapture.xaml.cpp b/samples/winrt/ImageManipulations/AdvancedCapture.xaml.cpp
new file mode 100644
index 0000000000..cff0a5a799
--- /dev/null
+++ b/samples/winrt/ImageManipulations/AdvancedCapture.xaml.cpp
@@ -0,0 +1,613 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// AdvancedCapture.xaml.cpp
+// Implementation of the AdvancedCapture class
+//
+
+#include "pch.h"
+#include "AdvancedCapture.xaml.h"
+
+using namespace SDKSample::MediaCapture;
+
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Navigation;
+using namespace Windows::UI::Xaml::Data;
+using namespace Windows::System;
+using namespace Windows::Foundation;
+using namespace Windows::Foundation::Collections;
+using namespace Platform;
+using namespace Windows::UI;
+using namespace Windows::UI::Core;
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Controls;
+using namespace Windows::UI::Xaml::Data;
+using namespace Windows::UI::Xaml::Media;
+using namespace Windows::Storage;
+using namespace Windows::Media::MediaProperties;
+using namespace Windows::Storage::Streams;
+using namespace Windows::System;
+using namespace Windows::UI::Xaml::Media::Imaging;
+using namespace Windows::Devices::Enumeration;
+
+ref class ReencodeState sealed
+{
+public:
+    ReencodeState()
+    {
+    }
+
+    virtual ~ReencodeState()
+    {
+        if (InputStream != nullptr)
+        {
+            delete InputStream;
+        }
+        if (OutputStream != nullptr)
+        {
+            delete OutputStream;
+        }
+    }
+
+internal:
+    Windows::Storage::Streams::IRandomAccessStream ^InputStream;
+    Windows::Storage::Streams::IRandomAccessStream ^OutputStream;
+    Windows::Storage::StorageFile ^PhotoStorage;
+    Windows::Graphics::Imaging::BitmapDecoder ^Decoder;
+    Windows::Graphics::Imaging::BitmapEncoder ^Encoder;
+};
+
+AdvancedCapture::AdvancedCapture()
+{
+    InitializeComponent();
+    ScenarioInit();
+}
+
+/// <summary>
+/// Invoked when this page is about to be displayed in a Frame.
+/// </summary>
+/// <param name="e">Event data that describes how this page was reached.  The Parameter
+/// property is typically used to configure the page.</param>
+void AdvancedCapture::OnNavigatedTo(NavigationEventArgs^ e)
+{
+    // A pointer back to the main page.  This is needed if you want to call methods in MainPage such
+    // as NotifyUser()
+    rootPage = MainPage::Current;
+
+    m_orientationChangedEventToken = Windows::Graphics::Display::DisplayProperties::OrientationChanged += ref new Windows::Graphics::Display::DisplayPropertiesEventHandler(this, &AdvancedCapture::DisplayProperties_OrientationChanged);
+}
+
+void AdvancedCapture::OnNavigatedFrom(NavigationEventArgs^ e)
+{
+    Windows::Media::MediaControl::SoundLevelChanged -= m_eventRegistrationToken;
+    Windows::Graphics::Display::DisplayProperties::OrientationChanged  -= m_orientationChangedEventToken;
+}
+
+void  AdvancedCapture::ScenarioInit()
+{
+    rootPage = MainPage::Current;
+    btnStartDevice2->IsEnabled = true;
+    btnStartPreview2->IsEnabled = false;
+    m_bRecording = false;
+    m_bPreviewing = false;
+    m_bEffectAdded = false;
+    previewElement2->Source = nullptr;
+    ShowStatusMessage("");
+    EffectTypeCombo->IsEnabled = false;
+    previewCanvas2->Visibility = Windows::UI::Xaml::Visibility::Collapsed;
+    EnumerateWebcamsAsync();
+    m_bSuspended = false;
+}
+
+void AdvancedCapture::ScenarioReset()
+{
+    previewCanvas2->Visibility = Windows::UI::Xaml::Visibility::Collapsed;
+    ScenarioInit();
+}
+
+void AdvancedCapture::Failed(Windows::Media::Capture::MediaCapture ^currentCaptureObject, Windows::Media::Capture::MediaCaptureFailedEventArgs^ currentFailure)
+{
+    String ^message = "Fatal error" + currentFailure->Message;
+    create_task(Dispatcher->RunAsync(Windows::UI::Core::CoreDispatcherPriority::High,
+        ref new Windows::UI::Core::DispatchedHandler([this, message]()
+    {
+        ShowStatusMessage(message);
+    })));
+}
+
+void AdvancedCapture::btnStartDevice_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e)
+{
+    try
+    {
+        EnableButton(false, "StartDevice");
+        ShowStatusMessage("Starting device");
+        auto mediaCapture = ref new Windows::Media::Capture::MediaCapture();
+        m_mediaCaptureMgr = mediaCapture;
+        auto settings = ref new Windows::Media::Capture::MediaCaptureInitializationSettings();
+        auto chosenDevInfo = m_devInfoCollection->GetAt(EnumedDeviceList2->SelectedIndex);
+        settings->VideoDeviceId = chosenDevInfo->Id;
+        if (chosenDevInfo->EnclosureLocation != nullptr && chosenDevInfo->EnclosureLocation->Panel == Windows::Devices::Enumeration::Panel::Back)
+        {
+            m_bRotateVideoOnOrientationChange = true;
+            m_bReversePreviewRotation = false;
+        }
+        else if (chosenDevInfo->EnclosureLocation != nullptr && chosenDevInfo->EnclosureLocation->Panel == Windows::Devices::Enumeration::Panel::Front)
+        {
+            m_bRotateVideoOnOrientationChange = true;
+            m_bReversePreviewRotation = true;
+        }
+        else
+        {
+            m_bRotateVideoOnOrientationChange = false;
+        }
+
+        create_task(mediaCapture->InitializeAsync(settings)).then([this](task<void> initTask)
+        {
+            try
+            {
+                initTask.get();
+
+                auto mediaCapture =  m_mediaCaptureMgr.Get();
+
+                DisplayProperties_OrientationChanged(nullptr);
+
+                EnableButton(true, "StartPreview");
+                EnableButton(true, "StartStopRecord");
+                EnableButton(true, "TakePhoto");
+                ShowStatusMessage("Device initialized successful");
+                EffectTypeCombo->IsEnabled = true;
+                mediaCapture->Failed += ref new Windows::Media::Capture::MediaCaptureFailedEventHandler(this, &AdvancedCapture::Failed);
+            }
+            catch (Exception ^ e)
+            {
+                ShowExceptionMessage(e);
+            }
+        });
+    }
+    catch (Platform::Exception^ e)
+    {
+        ShowExceptionMessage(e);
+    }
+}
+
+void AdvancedCapture::btnStartPreview_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e)
+{
+    m_bPreviewing = false;
+    try
+    {
+        ShowStatusMessage("Starting preview");
+        EnableButton(false, "StartPreview");
+
+        auto mediaCapture = m_mediaCaptureMgr.Get();
+        previewCanvas2->Visibility = Windows::UI::Xaml::Visibility::Visible;
+        previewElement2->Source = mediaCapture;
+        create_task(mediaCapture->StartPreviewAsync()).then([this](task<void> previewTask)
+        {
+            try
+            {
+                previewTask.get();
+                m_bPreviewing = true;
+                ShowStatusMessage("Start preview successful");
+            }
+            catch (Exception ^e)
+            {
+                ShowExceptionMessage(e);
+            }
+        });
+    }
+    catch (Platform::Exception^ e)
+    {
+        m_bPreviewing = false;
+        previewElement2->Source = nullptr;
+        EnableButton(true, "StartPreview");
+        ShowExceptionMessage(e);
+    }
+}
+
+void AdvancedCapture::lstEnumedDevices_SelectionChanged(Platform::Object^ sender, Windows::UI::Xaml::Controls::SelectionChangedEventArgs^ e)
+{
+     if ( m_bPreviewing )
+     {
+         create_task(m_mediaCaptureMgr->StopPreviewAsync()).then([this](task<void> previewTask)
+         {
+             try
+             {
+                 previewTask.get();
+                 m_bPreviewing = false;
+             }
+             catch (Exception ^e)
+             {
+                ShowExceptionMessage(e);
+             }
+         });
+    }
+
+    btnStartDevice2->IsEnabled = true;
+    btnStartPreview2->IsEnabled = false;
+    m_bRecording = false;
+    previewElement2->Source = nullptr;
+    EffectTypeCombo->IsEnabled = false;
+    m_bEffectAdded = false;
+    m_bEffectAddedToRecord = false;
+    m_bEffectAddedToPhoto = false;
+    ShowStatusMessage("");
+}
+
+void AdvancedCapture::EnumerateWebcamsAsync()
+{
+    try
+    {
+        ShowStatusMessage("Enumerating Webcams...");
+        m_devInfoCollection = nullptr;
+
+        EnumedDeviceList2->Items->Clear();
+
+        task<DeviceInformationCollection^>(DeviceInformation::FindAllAsync(DeviceClass::VideoCapture)).then([this](task<DeviceInformationCollection^> findTask)
+        {
+            try
+            {
+                m_devInfoCollection = findTask.get();
+                if (m_devInfoCollection == nullptr || m_devInfoCollection->Size == 0)
+                {
+                    ShowStatusMessage("No WebCams found.");
+                }
+                else
+                {
+                    for(unsigned int i = 0; i < m_devInfoCollection->Size; i++)
+                    {
+                        auto devInfo = m_devInfoCollection->GetAt(i);
+                        EnumedDeviceList2->Items->Append(devInfo->Name);
+                    }
+                    EnumedDeviceList2->SelectedIndex = 0;
+                    ShowStatusMessage("Enumerating Webcams completed successfully.");
+                    btnStartDevice2->IsEnabled = true;
+                }
+            }
+            catch (Exception ^e)
+            {
+                ShowExceptionMessage(e);
+            }
+        });
+    }
+    catch (Platform::Exception^ e)
+    {
+        ShowExceptionMessage(e);
+    }
+}
+
+void AdvancedCapture::AddEffectToImageStream()
+{
+    auto mediaCapture = m_mediaCaptureMgr.Get();
+    Windows::Media::Capture::VideoDeviceCharacteristic charecteristic = mediaCapture->MediaCaptureSettings->VideoDeviceCharacteristic;
+
+    if((charecteristic != Windows::Media::Capture::VideoDeviceCharacteristic::AllStreamsIdentical) &&
+        (charecteristic != Windows::Media::Capture::VideoDeviceCharacteristic::PreviewPhotoStreamsIdentical) &&
+        (charecteristic != Windows::Media::Capture::VideoDeviceCharacteristic::RecordPhotoStreamsIdentical))
+    {
+        Windows::Media::MediaProperties::IMediaEncodingProperties ^props = mediaCapture->VideoDeviceController->GetMediaStreamProperties(Windows::Media::Capture::MediaStreamType::Photo);
+        if(props->Type->Equals("Image"))
+        {
+            //Switch to a video media type instead since we cant add an effect to a image media type
+            Windows::Foundation::Collections::IVectorView<Windows::Media::MediaProperties::IMediaEncodingProperties^>^ supportedPropsList = mediaCapture->VideoDeviceController->GetAvailableMediaStreamProperties(Windows::Media::Capture::MediaStreamType::Photo);
+            {
+                unsigned int i = 0;
+                while (i < supportedPropsList->Size)
+                {
+                    Windows::Media::MediaProperties::IMediaEncodingProperties^ props = supportedPropsList->GetAt(i);
+
+                    String^ s = props->Type;
+                    if(props->Type->Equals("Video"))
+                    {
+                        task<void>(mediaCapture->VideoDeviceController->SetMediaStreamPropertiesAsync(Windows::Media::Capture::MediaStreamType::Photo,props)).then([this](task<void> changeTypeTask)
+                        {
+                            try
+                            {
+                                changeTypeTask.get();
+                                ShowStatusMessage("Change type on photo stream successful");
+                                //Now add the effect on the image pin
+                                task<void>(m_mediaCaptureMgr->AddEffectAsync(Windows::Media::Capture::MediaStreamType::Photo,"OcvTransform.OcvImageManipulations", nullptr)).then([this](task<void> effectTask3)
+                                {
+                                    try
+                                    {
+                                        effectTask3.get();
+                                        m_bEffectAddedToPhoto = true;
+                                        ShowStatusMessage("Adding effect to photo stream successful");
+                                        EffectTypeCombo->IsEnabled = true;
+
+                                    }
+                                    catch(Exception ^e)
+                                    {
+                                        ShowExceptionMessage(e);
+                                        EffectTypeCombo->IsEnabled = true;
+                                    }
+                                });
+
+                            }
+                            catch(Exception ^e)
+                            {
+                                ShowExceptionMessage(e);
+                                EffectTypeCombo->IsEnabled = true;
+                            }
+
+                        });
+                        break;
+
+                    }
+                    i++;
+                }
+            }
+        }
+        else
+        {
+            //Add the effect to the image pin if the type is already "Video"
+            task<void>(mediaCapture->AddEffectAsync(Windows::Media::Capture::MediaStreamType::Photo,"OcvTransform.OcvImageManipulations", nullptr)).then([this](task<void> effectTask3)
+            {
+                try
+                {
+                    effectTask3.get();
+                    m_bEffectAddedToPhoto = true;
+                    ShowStatusMessage("Adding effect to photo stream successful");
+                    EffectTypeCombo->IsEnabled = true;
+
+                }
+                catch(Exception ^e)
+                {
+                    ShowExceptionMessage(e);
+                    EffectTypeCombo->IsEnabled = true;
+                }
+            });
+        }
+    }
+}
+
+void AdvancedCapture::ShowStatusMessage(Platform::String^ text)
+{
+    rootPage->NotifyUser(text, NotifyType::StatusMessage);
+}
+
+void AdvancedCapture::ShowExceptionMessage(Platform::Exception^ ex)
+{
+    rootPage->NotifyUser(ex->Message, NotifyType::ErrorMessage);
+}
+
+void AdvancedCapture::EnableButton(bool enabled, String^ name)
+{
+    if (name->Equals("StartDevice"))
+    {
+        btnStartDevice2->IsEnabled = enabled;
+    }
+    else if (name->Equals("StartPreview"))
+    {
+        btnStartPreview2->IsEnabled = enabled;
+    }
+}
+
+task<Windows::Storage::StorageFile^> AdvancedCapture::ReencodePhotoAsync(
+    Windows::Storage::StorageFile ^tempStorageFile,
+    Windows::Storage::FileProperties::PhotoOrientation photoRotation)
+{
+    ReencodeState ^state = ref new ReencodeState();
+
+    return create_task(tempStorageFile->OpenAsync(Windows::Storage::FileAccessMode::Read)).then([state](Windows::Storage::Streams::IRandomAccessStream ^stream)
+    {
+        state->InputStream = stream;
+        return Windows::Graphics::Imaging::BitmapDecoder::CreateAsync(state->InputStream);
+    }).then([state](Windows::Graphics::Imaging::BitmapDecoder ^decoder)
+    {
+        state->Decoder = decoder;
+        return Windows::Storage::KnownFolders::PicturesLibrary->CreateFileAsync(PHOTO_FILE_NAME, Windows::Storage::CreationCollisionOption::GenerateUniqueName);
+    }).then([state](Windows::Storage::StorageFile ^storageFile)
+    {
+        state->PhotoStorage = storageFile;
+        return state->PhotoStorage->OpenAsync(Windows::Storage::FileAccessMode::ReadWrite);
+    }).then([state](Windows::Storage::Streams::IRandomAccessStream ^stream)
+    {
+        state->OutputStream = stream;
+        state->OutputStream->Size = 0;
+        return Windows::Graphics::Imaging::BitmapEncoder::CreateForTranscodingAsync(state->OutputStream, state->Decoder);
+    }).then([state, photoRotation](Windows::Graphics::Imaging::BitmapEncoder ^encoder)
+    {
+        state->Encoder = encoder;
+        auto properties = ref new Windows::Graphics::Imaging::BitmapPropertySet();
+        properties->Insert("System.Photo.Orientation",
+            ref new Windows::Graphics::Imaging::BitmapTypedValue((unsigned short)photoRotation, Windows::Foundation::PropertyType::UInt16));
+        return create_task(state->Encoder->BitmapProperties->SetPropertiesAsync(properties));
+    }).then([state]()
+    {
+        return state->Encoder->FlushAsync();
+    }).then([tempStorageFile, state](task<void> previousTask)
+    {
+        auto result = state->PhotoStorage;
+        delete state;
+
+        tempStorageFile->DeleteAsync(Windows::Storage::StorageDeleteOption::PermanentDelete);
+
+        previousTask.get();
+
+        return result;
+    });
+}
+
+Windows::Storage::FileProperties::PhotoOrientation AdvancedCapture::GetCurrentPhotoRotation()
+{
+    bool counterclockwiseRotation = m_bReversePreviewRotation;
+
+    if (m_bRotateVideoOnOrientationChange)
+    {
+        return PhotoRotationLookup(Windows::Graphics::Display::DisplayProperties::CurrentOrientation, counterclockwiseRotation);
+    }
+    else
+    {
+        return Windows::Storage::FileProperties::PhotoOrientation::Normal;
+    }
+}
+
+void AdvancedCapture::PrepareForVideoRecording()
+{
+    Windows::Media::Capture::MediaCapture ^mediaCapture = m_mediaCaptureMgr.Get();
+    if (mediaCapture == nullptr)
+    {
+        return;
+    }
+
+    bool counterclockwiseRotation = m_bReversePreviewRotation;
+
+    if (m_bRotateVideoOnOrientationChange)
+    {
+        mediaCapture->SetRecordRotation(VideoRotationLookup(Windows::Graphics::Display::DisplayProperties::CurrentOrientation, counterclockwiseRotation));
+    }
+    else
+    {
+        mediaCapture->SetRecordRotation(Windows::Media::Capture::VideoRotation::None);
+    }
+}
+
+void AdvancedCapture::DisplayProperties_OrientationChanged(Platform::Object^ sender)
+{
+    Windows::Media::Capture::MediaCapture ^mediaCapture = m_mediaCaptureMgr.Get();
+    if (mediaCapture == nullptr)
+    {
+        return;
+    }
+
+    bool previewMirroring = mediaCapture->GetPreviewMirroring();
+    bool counterclockwiseRotation = (previewMirroring && !m_bReversePreviewRotation) ||
+        (!previewMirroring && m_bReversePreviewRotation);
+
+    if (m_bRotateVideoOnOrientationChange)
+    {
+        mediaCapture->SetPreviewRotation(VideoRotationLookup(Windows::Graphics::Display::DisplayProperties::CurrentOrientation, counterclockwiseRotation));
+    }
+    else
+    {
+        mediaCapture->SetPreviewRotation(Windows::Media::Capture::VideoRotation::None);
+    }
+}
+
+Windows::Storage::FileProperties::PhotoOrientation AdvancedCapture::PhotoRotationLookup(
+    Windows::Graphics::Display::DisplayOrientations displayOrientation, bool counterclockwise)
+{
+    switch (displayOrientation)
+    {
+    case Windows::Graphics::Display::DisplayOrientations::Landscape:
+        return Windows::Storage::FileProperties::PhotoOrientation::Normal;
+
+    case Windows::Graphics::Display::DisplayOrientations::Portrait:
+        return (counterclockwise) ? Windows::Storage::FileProperties::PhotoOrientation::Rotate270:
+            Windows::Storage::FileProperties::PhotoOrientation::Rotate90;
+
+    case Windows::Graphics::Display::DisplayOrientations::LandscapeFlipped:
+        return Windows::Storage::FileProperties::PhotoOrientation::Rotate180;
+
+    case Windows::Graphics::Display::DisplayOrientations::PortraitFlipped:
+        return (counterclockwise) ? Windows::Storage::FileProperties::PhotoOrientation::Rotate90 :
+            Windows::Storage::FileProperties::PhotoOrientation::Rotate270;
+
+    default:
+        return Windows::Storage::FileProperties::PhotoOrientation::Unspecified;
+    }
+}
+
+Windows::Media::Capture::VideoRotation AdvancedCapture::VideoRotationLookup(
+    Windows::Graphics::Display::DisplayOrientations displayOrientation, bool counterclockwise)
+{
+    switch (displayOrientation)
+    {
+    case Windows::Graphics::Display::DisplayOrientations::Landscape:
+        return Windows::Media::Capture::VideoRotation::None;
+
+    case Windows::Graphics::Display::DisplayOrientations::Portrait:
+        return (counterclockwise) ? Windows::Media::Capture::VideoRotation::Clockwise270Degrees :
+            Windows::Media::Capture::VideoRotation::Clockwise90Degrees;
+
+    case Windows::Graphics::Display::DisplayOrientations::LandscapeFlipped:
+        return Windows::Media::Capture::VideoRotation::Clockwise180Degrees;
+
+    case Windows::Graphics::Display::DisplayOrientations::PortraitFlipped:
+        return (counterclockwise) ? Windows::Media::Capture::VideoRotation::Clockwise90Degrees:
+            Windows::Media::Capture::VideoRotation::Clockwise270Degrees ;
+
+    default:
+        return Windows::Media::Capture::VideoRotation::None;
+    }
+}
+
+void SDKSample::MediaCapture::AdvancedCapture::Button_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e)
+{
+    try
+    {
+        create_task(m_mediaCaptureMgr->ClearEffectsAsync(Windows::Media::Capture::MediaStreamType::VideoPreview)).then([this](task<void> cleanTask)
+        {
+            m_bEffectAdded = true;
+            int index = EffectTypeCombo->SelectedIndex;
+            PropertySet^ props = ref new PropertySet();
+            props->Insert(L"{698649BE-8EAE-4551-A4CB-3EC98FBD3D86}", index);
+            create_task(m_mediaCaptureMgr->AddEffectAsync(Windows::Media::Capture::MediaStreamType::VideoPreview,"OcvTransform.OcvImageManipulations", props)).then([this](task<void> effectTask)
+            {
+                try
+                {
+                    effectTask.get();
+
+                    auto mediaCapture = m_mediaCaptureMgr.Get();
+                    Windows::Media::Capture::VideoDeviceCharacteristic charecteristic = mediaCapture->MediaCaptureSettings->VideoDeviceCharacteristic;
+
+                    ShowStatusMessage("Add effect successful to preview stream successful");
+                    if((charecteristic != Windows::Media::Capture::VideoDeviceCharacteristic::AllStreamsIdentical) &&
+                        (charecteristic != Windows::Media::Capture::VideoDeviceCharacteristic::PreviewRecordStreamsIdentical))
+                    {
+                        Windows::Media::MediaProperties::IMediaEncodingProperties ^props = mediaCapture->VideoDeviceController->GetMediaStreamProperties(Windows::Media::Capture::MediaStreamType::VideoRecord);
+                        Windows::Media::MediaProperties::VideoEncodingProperties ^videoEncodingProperties  = static_cast<Windows::Media::MediaProperties::VideoEncodingProperties ^>(props);
+                        if(!videoEncodingProperties->Subtype->Equals("H264")) //Cant add an effect to an H264 stream
+                        {
+                            task<void>(mediaCapture->AddEffectAsync(Windows::Media::Capture::MediaStreamType::VideoRecord,"OcvTransform.OcvImageManipulations", nullptr)).then([this](task<void> effectTask2)
+                            {
+                                try
+                                {
+                                    effectTask2.get();
+                                    ShowStatusMessage("Add effect successful to record stream successful");
+                                    m_bEffectAddedToRecord = true;
+                                    AddEffectToImageStream();
+                                    EffectTypeCombo->IsEnabled = true;
+                                }
+                                catch(Exception ^e)
+                                {
+                                    ShowExceptionMessage(e);
+                                    EffectTypeCombo->IsEnabled = true;
+                                }
+                            });
+                        }
+                        else
+                        {
+                            AddEffectToImageStream();
+                            EffectTypeCombo->IsEnabled = true;
+                        }
+
+                    }
+                    else
+                    {
+                        AddEffectToImageStream();
+                        EffectTypeCombo->IsEnabled = true;
+                    }
+                }
+                catch (Exception ^e)
+                {
+                    ShowExceptionMessage(e);
+                    EffectTypeCombo->IsEnabled = true;
+                }
+            });
+        });
+    }
+    catch (Platform::Exception ^e)
+    {
+        ShowExceptionMessage(e);
+        EffectTypeCombo->IsEnabled = true;
+    }
+}
diff --git a/samples/winrt/ImageManipulations/AdvancedCapture.xaml.h b/samples/winrt/ImageManipulations/AdvancedCapture.xaml.h
new file mode 100644
index 0000000000..94fa87c615
--- /dev/null
+++ b/samples/winrt/ImageManipulations/AdvancedCapture.xaml.h
@@ -0,0 +1,95 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// AdvancedCapture.xaml.h
+// Declaration of the AdvancedCapture class
+//
+
+#pragma once
+
+#include "pch.h"
+#include "AdvancedCapture.g.h"
+#include "MainPage.xaml.h"
+#include <ppl.h>
+
+#define VIDEO_FILE_NAME "video.mp4"
+#define PHOTO_FILE_NAME "photo.jpg"
+#define TEMP_PHOTO_FILE_NAME "photoTmp.jpg"
+
+using namespace concurrency;
+using namespace Windows::Devices::Enumeration;
+
+namespace SDKSample
+{
+    namespace MediaCapture
+    {
+        /// <summary>
+        /// An empty page that can be used on its own or navigated to within a Frame.
+        /// </summary>
+        [Windows::Foundation::Metadata::WebHostHidden]
+        public ref class AdvancedCapture sealed
+        {
+        public:
+            AdvancedCapture();
+
+        protected:
+            virtual void OnNavigatedTo(Windows::UI::Xaml::Navigation::NavigationEventArgs^ e) override;
+            virtual void OnNavigatedFrom(Windows::UI::Xaml::Navigation::NavigationEventArgs^ e) override;
+
+        private:
+            MainPage^ rootPage;
+            void ScenarioInit();
+            void ScenarioReset();
+
+            void Failed(Windows::Media::Capture::MediaCapture ^ mediaCapture, Windows::Media::Capture::MediaCaptureFailedEventArgs ^ args);
+
+            void btnStartDevice_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+
+            void btnStartPreview_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+
+            void lstEnumedDevices_SelectionChanged(Platform::Object^ sender, Windows::UI::Xaml::Controls::SelectionChangedEventArgs^ e);
+            void EnumerateWebcamsAsync();
+
+            void AddEffectToImageStream();
+
+            void ShowStatusMessage(Platform::String^ text);
+            void ShowExceptionMessage(Platform::Exception^ ex);
+
+            void EnableButton(bool enabled, Platform::String ^name);
+
+            task<Windows::Storage::StorageFile^> ReencodePhotoAsync(
+                Windows::Storage::StorageFile ^tempStorageFile,
+                Windows::Storage::FileProperties::PhotoOrientation photoRotation);
+            Windows::Storage::FileProperties::PhotoOrientation GetCurrentPhotoRotation();
+            void PrepareForVideoRecording();
+            void DisplayProperties_OrientationChanged(Platform::Object^ sender);
+            Windows::Storage::FileProperties::PhotoOrientation PhotoRotationLookup(
+                Windows::Graphics::Display::DisplayOrientations displayOrientation, bool counterclockwise);
+            Windows::Media::Capture::VideoRotation VideoRotationLookup(
+                Windows::Graphics::Display::DisplayOrientations displayOrientation, bool counterclockwise);
+
+            Platform::Agile<Windows::Media::Capture::MediaCapture> m_mediaCaptureMgr;
+            Windows::Storage::StorageFile^ m_recordStorageFile;
+            bool m_bRecording;
+            bool m_bEffectAdded;
+            bool m_bEffectAddedToRecord;
+            bool m_bEffectAddedToPhoto;
+            bool m_bSuspended;
+            bool m_bPreviewing;
+            DeviceInformationCollection^ m_devInfoCollection;
+            Windows::Foundation::EventRegistrationToken m_eventRegistrationToken;
+            bool m_bRotateVideoOnOrientationChange;
+            bool m_bReversePreviewRotation;
+            Windows::Foundation::EventRegistrationToken m_orientationChangedEventToken;
+            void Button_Click(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+        };
+    }
+}
diff --git a/samples/winrt/ImageManipulations/App.xaml b/samples/winrt/ImageManipulations/App.xaml
new file mode 100644
index 0000000000..769f4400d2
--- /dev/null
+++ b/samples/winrt/ImageManipulations/App.xaml
@@ -0,0 +1,30 @@
+<!--
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+-->
+
+<Application xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
+    x:Class="SDKSample.App"
+    RequestedTheme="Light">
+    <Application.Resources>
+
+        <ResourceDictionary>
+            <ResourceDictionary.MergedDictionaries>
+                <!--
+                    Styles that define common aspects of the platform look and feel
+                    Required by Visual Studio project and item templates
+                 -->
+                <ResourceDictionary Source="Common/StandardStyles.xaml"/>
+                <ResourceDictionary Source="Sample-Utils/SampleTemplateStyles.xaml"/>
+            </ResourceDictionary.MergedDictionaries>
+        </ResourceDictionary>
+    </Application.Resources>
+</Application>
diff --git a/samples/winrt/ImageManipulations/App.xaml.cpp b/samples/winrt/ImageManipulations/App.xaml.cpp
new file mode 100644
index 0000000000..a24a4f9c0a
--- /dev/null
+++ b/samples/winrt/ImageManipulations/App.xaml.cpp
@@ -0,0 +1,116 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// App.xaml.cpp
+// Implementation of the App.xaml class.
+//
+
+#include "pch.h"
+#include "MainPage.xaml.h"
+#include "AdvancedCapture.xaml.h"
+#include "Common\SuspensionManager.h"
+
+using namespace SDKSample;
+using namespace SDKSample::Common;
+using namespace SDKSample::MediaCapture;
+
+using namespace Concurrency;
+using namespace Platform;
+using namespace Windows::ApplicationModel;
+using namespace Windows::ApplicationModel::Activation;
+using namespace Windows::Foundation;
+using namespace Windows::Foundation::Collections;
+using namespace Windows::UI::Core;
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Controls;
+using namespace Windows::UI::Xaml::Controls::Primitives;
+using namespace Windows::UI::Xaml::Data;
+using namespace Windows::UI::Xaml::Input;
+using namespace Windows::UI::Xaml::Interop;
+using namespace Windows::UI::Xaml::Media;
+using namespace Windows::UI::Xaml::Navigation;
+
+/// <summary>
+/// Initializes the singleton application object.  This is the first line of authored code
+/// executed, and as such is the logical equivalent of main() or WinMain().
+/// </summary>
+App::App()
+{
+    InitializeComponent();
+    this->Suspending += ref new SuspendingEventHandler(this, &SDKSample::App::OnSuspending);
+}
+
+/// <summary>
+/// Invoked when the application is launched normally by the end user.  Other entry points will
+/// be used when the application is launched to open a specific file, to display search results,
+/// and so forth.
+/// </summary>
+/// <param name="pArgs">Details about the launch request and process.</param>
+void App::OnLaunched(LaunchActivatedEventArgs^ pArgs)
+{
+    this->LaunchArgs = pArgs;
+
+    // Do not repeat app initialization when already running, just ensure that
+    // the window is active
+    if (pArgs->PreviousExecutionState == ApplicationExecutionState::Running)
+    {
+        Window::Current->Activate();
+        return;
+    }
+
+    // Create a Frame to act as the navigation context and associate it with
+    // a SuspensionManager key
+    auto rootFrame = ref new Frame();
+    SuspensionManager::RegisterFrame(rootFrame, "AppFrame");
+
+    auto prerequisite = task<void>([](){});
+    if (pArgs->PreviousExecutionState == ApplicationExecutionState::Terminated)
+    {
+        // Restore the saved session state only when appropriate, scheduling the
+        // final launch steps after the restore is complete
+        prerequisite = SuspensionManager::RestoreAsync();
+    }
+    prerequisite.then([=]()
+    {
+        // When the navigation stack isn't restored navigate to the first page,
+        // configuring the new page by passing required information as a navigation
+        // parameter
+        if (rootFrame->Content == nullptr)
+        {
+            if (!rootFrame->Navigate(TypeName(MainPage::typeid)))
+            {
+                throw ref new FailureException("Failed to create initial page");
+            }
+        }
+
+        // Place the frame in the current Window and ensure that it is active
+        Window::Current->Content = rootFrame;
+        Window::Current->Activate();
+    }, task_continuation_context::use_current());
+}
+
+/// <summary>
+/// Invoked when application execution is being suspended.  Application state is saved
+/// without knowing whether the application will be terminated or resumed with the contents
+/// of memory still intact.
+/// </summary>
+/// <param name="sender">The source of the suspend request.</param>
+/// <param name="e">Details about the suspend request.</param>
+void App::OnSuspending(Object^ sender, SuspendingEventArgs^ e)
+{
+    (void) sender;	// Unused parameter
+
+    auto deferral = e->SuspendingOperation->GetDeferral();
+    SuspensionManager::SaveAsync().then([=]()
+    {
+        deferral->Complete();
+    });
+}
diff --git a/samples/winrt/ImageManipulations/App.xaml.h b/samples/winrt/ImageManipulations/App.xaml.h
new file mode 100644
index 0000000000..a8b6064248
--- /dev/null
+++ b/samples/winrt/ImageManipulations/App.xaml.h
@@ -0,0 +1,35 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// App.xaml.h
+// Declaration of the App.xaml class.
+//
+
+#pragma once
+
+#include "pch.h"
+#include "App.g.h"
+#include "MainPage.g.h"
+
+namespace SDKSample
+{
+    ref class App
+    {
+    internal:
+        App();
+        virtual void OnSuspending(Platform::Object^ sender, Windows::ApplicationModel::SuspendingEventArgs^ pArgs);
+        Windows::ApplicationModel::Activation::LaunchActivatedEventArgs^ LaunchArgs;
+    protected:
+        virtual void OnLaunched(Windows::ApplicationModel::Activation::LaunchActivatedEventArgs^ pArgs) override;
+    private:
+        Windows::UI::Xaml::Controls::Frame^ rootFrame;
+    };
+}
diff --git a/samples/winrt/ImageManipulations/Constants.cpp b/samples/winrt/ImageManipulations/Constants.cpp
new file mode 100644
index 0000000000..a26634272b
--- /dev/null
+++ b/samples/winrt/ImageManipulations/Constants.cpp
@@ -0,0 +1,22 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+#include "pch.h"
+#include "MainPage.xaml.h"
+#include "Constants.h"
+
+using namespace SDKSample;
+
+Platform::Array<Scenario>^ MainPage::scenariosInner = ref new Platform::Array<Scenario>
+{
+    // The format here is the following:
+    //     { "Description for the sample", "Fully quaified name for the class that implements the scenario" }
+    { "Enumerate cameras and add a video effect", "SDKSample.MediaCapture.AdvancedCapture" },
+};
diff --git a/samples/winrt/ImageManipulations/Constants.h b/samples/winrt/ImageManipulations/Constants.h
new file mode 100644
index 0000000000..143f06960f
--- /dev/null
+++ b/samples/winrt/ImageManipulations/Constants.h
@@ -0,0 +1,45 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+#pragma once
+
+#include <collection.h>
+namespace SDKSample
+{
+    public value struct Scenario
+    {
+        Platform::String^ Title;
+        Platform::String^ ClassName;
+    };
+
+    partial ref class MainPage
+    {
+    public:
+        static property Platform::String^ FEATURE_NAME
+        {
+            Platform::String^ get()
+            {
+                return ref new Platform::String(L"OpenCV Image Manipulations sample");
+            }
+        }
+
+        static property Platform::Array<Scenario>^ scenarios
+        {
+            Platform::Array<Scenario>^ get()
+            {
+                return scenariosInner;
+            }
+        }
+    private:
+        static Platform::Array<Scenario>^ scenariosInner;
+    };
+
+
+}
diff --git a/samples/winrt/ImageManipulations/MainPage.xaml b/samples/winrt/ImageManipulations/MainPage.xaml
new file mode 100644
index 0000000000..66ce5715f9
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MainPage.xaml
@@ -0,0 +1,156 @@
+<!--
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+-->
+
+<common:LayoutAwarePage
+    x:Class="SDKSample.MainPage"
+    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
+    xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
+    xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
+    xmlns:common="using:SDKSample.Common"
+    mc:Ignorable="d"
+    x:Name="RootPage">
+
+    <common:LayoutAwarePage.Resources>
+        <Style x:Key="BaseStatusStyle" TargetType="TextBlock">
+            <Setter Property="FontFamily" Value="Segoe UI Semilight"/>
+            <Setter Property="FontSize" Value="14.667"/>
+            <Setter Property="Margin" Value="0,0,0,5"/>
+        </Style>
+        <Style x:Key="StatusStyle" BasedOn="{StaticResource BaseStatusStyle}" TargetType="TextBlock">
+            <Setter Property="Foreground" Value="Green"/>
+        </Style>
+        <Style x:Key="ErrorStyle" BasedOn="{StaticResource BaseStatusStyle}" TargetType="TextBlock">
+            <Setter Property="Foreground" Value="Blue"/>
+        </Style>
+    </common:LayoutAwarePage.Resources>
+
+
+    <Grid x:Name="LayoutRoot" Background="{StaticResource ApplicationPageBackgroundThemeBrush}">
+
+        <Grid x:Name="ContentRoot" Background="{StaticResource ApplicationPageBackgroundThemeBrush}" Margin="100,20,100,20">
+            <Grid.RowDefinitions>
+                <RowDefinition Height="Auto"/>
+                <RowDefinition Height="*"/>
+                <RowDefinition Height="Auto"/>
+            </Grid.RowDefinitions>
+
+            <!-- Header -->
+            <StackPanel Orientation="Horizontal" Grid.Row="0">
+                <Image x:Name="WindowsLogo" Stretch="None" Source="Assets/windows-sdk.png" AutomationProperties.Name="Windows Logo" HorizontalAlignment="Left" Grid.Column="0"/>
+                <TextBlock VerticalAlignment="Bottom" Style="{StaticResource TitleTextStyle}" TextWrapping="Wrap" Grid.Column="1" Text="OpenCV for Windows RT"/>
+            </StackPanel>
+            <ScrollViewer x:Name="MainScrollViewer" Grid.Row="1" ZoomMode="Disabled" IsTabStop="False" VerticalScrollBarVisibility="Auto" HorizontalScrollBarVisibility="Auto" Padding="0,0,0,20" >
+                <Grid>
+                    <Grid.RowDefinitions>
+                        <RowDefinition Height="Auto"/>
+                        <RowDefinition Height="*"/>
+                    </Grid.RowDefinitions>
+                    <TextBlock x:Name="FeatureName" Grid.Row="0"  Text="Add Sample Title Here" Style="{StaticResource HeaderTextStyle}" TextWrapping="Wrap"/>
+
+                    <!-- Content -->
+                    <Grid Grid.Row="1">
+
+                        <!-- All XAML in this section is purely for design time so you can see sample content in the designer. -->
+                        <!-- This will be repaced at runtime by live content.                                                  -->
+                        <Grid>
+                            <Grid.RowDefinitions>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="Auto"/>
+                                <RowDefinition Height="*"/>
+                            </Grid.RowDefinitions>
+                            <Grid.ColumnDefinitions>
+                                <ColumnDefinition Width="Auto"/>
+                                <ColumnDefinition Width="*"/>
+                            </Grid.ColumnDefinitions>
+                            <TextBlock Grid.Row="0" Text="Input" Style="{StaticResource H2Style}"/>
+
+                            <TextBlock x:Name="ScenarioListLabel" Text="Select Scenario:" Grid.Row="1"  Style="{StaticResource SubheaderTextStyle}" Margin="0,5,0,0" />
+                            <ListBox x:Name="Scenarios" Margin="0,0,20,0" Grid.Row="2" AutomationProperties.Name="Scenarios" HorizontalAlignment="Left"
+                                         VerticalAlignment="Top" ScrollViewer.HorizontalScrollBarVisibility="Auto"
+                                         AutomationProperties.LabeledBy="{Binding ElementName=ScenarioListLabel}" MaxHeight="125">
+                                <ListBox.ItemTemplate>
+                                    <DataTemplate>
+                                        <TextBlock Text="{Binding Name}"/>
+                                    </DataTemplate>
+                                </ListBox.ItemTemplate>
+                            </ListBox>
+                            <TextBlock x:Name="DescriptionText" Margin="0,5,0,0" Text="Description:" Style="{StaticResource SubheaderTextStyle}" Grid.Row="1" Grid.Column="1"/>
+                            <!-- Input Scenarios -->
+                            <UserControl x:Name="InputSection" Margin="0,5,0,0" IsTabStop="False" Grid.Row="2" Grid.Column="1" HorizontalAlignment="Left" VerticalAlignment="Top"/>
+
+                            <!-- Output section -->
+                            <TextBlock Text="Output" Grid.Row="5"  Margin="0,25,0,20" Style="{StaticResource H2Style}" Grid.ColumnSpan="2"/>
+                            <TextBlock x:Name="StatusBlock" Grid.Row="6" Margin="0,0,0,5" Grid.ColumnSpan="2"/>
+
+                            <!-- Output Scenarios -->
+                            <UserControl x:Name="OutputSection" Grid.Row="7" Grid.ColumnSpan="2" BorderThickness="0"/>
+                        </Grid>
+                    </Grid>
+                </Grid>
+            </ScrollViewer>
+
+            <!-- Footer -->
+            <Grid x:Name="Footer"  Grid.Row="3" Margin="0,10,0,10" VerticalAlignment="Bottom" >
+                <Grid.RowDefinitions>
+                    <RowDefinition Height="Auto"/>
+                    <RowDefinition Height="Auto"/>
+                    <RowDefinition Height="Auto"/>
+                </Grid.RowDefinitions>
+
+                <Grid.ColumnDefinitions>
+                    <ColumnDefinition Width="Auto"/>
+                    <ColumnDefinition Width="*"/>
+                </Grid.ColumnDefinitions>
+                <StackPanel x:Name="FooterPanel" Orientation="Horizontal" Grid.Row="1" Grid.Column="1" HorizontalAlignment="Right"/>
+            </Grid>
+
+
+        </Grid>
+
+        <VisualStateManager.VisualStateGroups>
+            <!-- Visual states reflect the application's view state -->
+            <VisualStateGroup>
+                <VisualState x:Name="FullScreenLandscape">
+                    <Storyboard>
+                    </Storyboard>
+                </VisualState>
+                <VisualState x:Name="Filled">
+                    <Storyboard>
+                    </Storyboard>
+                </VisualState>
+
+                <VisualState x:Name="FullScreenPortrait">
+                    <Storyboard>
+                    </Storyboard>
+                </VisualState>
+
+                <VisualState x:Name="Snapped">
+                    <Storyboard>
+                        <ObjectAnimationUsingKeyFrames Storyboard.TargetProperty="(FrameworkElement.Margin)" Storyboard.TargetName="ContentRoot">
+                            <DiscreteObjectKeyFrame KeyTime="0">
+                                <DiscreteObjectKeyFrame.Value>
+                                    <Thickness>20,20,20,20</Thickness>
+                                </DiscreteObjectKeyFrame.Value>
+                            </DiscreteObjectKeyFrame>
+                        </ObjectAnimationUsingKeyFrames>
+                    </Storyboard>
+                </VisualState>
+            </VisualStateGroup>
+        </VisualStateManager.VisualStateGroups>
+    </Grid>
+</common:LayoutAwarePage>
diff --git a/samples/winrt/ImageManipulations/MainPage.xaml.cpp b/samples/winrt/ImageManipulations/MainPage.xaml.cpp
new file mode 100644
index 0000000000..bd897fcc0b
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MainPage.xaml.cpp
@@ -0,0 +1,315 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// MainPage.xaml.cpp
+// Implementation of the MainPage.xaml class.
+//
+
+#include "pch.h"
+#include "MainPage.xaml.h"
+#include "App.xaml.h"
+
+#include <collection.h>
+
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Controls;
+using namespace Windows::Foundation;
+using namespace Windows::Foundation::Collections;
+using namespace Platform;
+using namespace SDKSample;
+using namespace Windows::UI::Xaml::Navigation;
+using namespace Windows::UI::Xaml::Interop;
+using namespace Windows::Graphics::Display;
+using namespace Windows::UI::ViewManagement;
+
+MainPage^ MainPage::Current = nullptr;
+
+MainPage::MainPage()
+{
+    InitializeComponent();
+
+    // This frame is hidden, meaning it is never shown.  It is simply used to load
+    // each scenario page and then pluck out the input and output sections and
+    // place them into the UserControls on the main page.
+    HiddenFrame = ref new Windows::UI::Xaml::Controls::Frame();
+    HiddenFrame->Visibility = Windows::UI::Xaml::Visibility::Collapsed;
+    ContentRoot->Children->Append(HiddenFrame);
+
+    FeatureName->Text = FEATURE_NAME;
+
+    this->SizeChanged += ref new SizeChangedEventHandler(this, &MainPage::MainPage_SizeChanged);
+    Scenarios->SelectionChanged += ref new SelectionChangedEventHandler(this, &MainPage::Scenarios_SelectionChanged);
+
+    MainPage::Current = this;
+    autoSizeInputSectionWhenSnapped = true;
+}
+
+/// <summary>
+/// We need to handle SizeChanged so that we can make the sample layout property
+/// in the various layouts.
+/// </summary>
+/// <param name="sender"></param>
+/// <param name="e"></param>
+void MainPage::MainPage_SizeChanged(Object^ sender, SizeChangedEventArgs^ e)
+{
+    InvalidateSize();
+    MainPageSizeChangedEventArgs^ args = ref new MainPageSizeChangedEventArgs();
+    args->ViewState = ApplicationView::Value;
+    MainPageResized(this, args);
+
+}
+
+void MainPage::InvalidateSize()
+{
+    // Get the window width
+    double windowWidth = this->ActualWidth;
+
+    if (windowWidth != 0.0)
+    {
+        // Get the width of the ListBox.
+        double listBoxWidth = Scenarios->ActualWidth;
+
+        // Is the ListBox using any margins that we need to consider?
+        double listBoxMarginLeft = Scenarios->Margin.Left;
+        double listBoxMarginRight = Scenarios->Margin.Right;
+
+        // Figure out how much room is left after considering the list box width
+        double availableWidth = windowWidth - listBoxWidth;
+
+        // Is the top most child using margins?
+        double layoutRootMarginLeft = ContentRoot->Margin.Left;
+        double layoutRootMarginRight = ContentRoot->Margin.Right;
+
+        // We have different widths to use depending on the view state
+        if (ApplicationView::Value != ApplicationViewState::Snapped)
+        {
+            // Make us as big as the the left over space, factoring in the ListBox width, the ListBox margins.
+            // and the LayoutRoot's margins
+            InputSection->Width = ((availableWidth) -
+                (layoutRootMarginLeft + layoutRootMarginRight + listBoxMarginLeft + listBoxMarginRight));
+        }
+        else
+        {
+            // Make us as big as the left over space, factoring in just the LayoutRoot's margins.
+            if (autoSizeInputSectionWhenSnapped)
+            {
+                InputSection->Width = (windowWidth - (layoutRootMarginLeft + layoutRootMarginRight));
+            }
+        }
+    }
+    InvalidateViewState();
+}
+
+void MainPage::InvalidateViewState()
+{
+    // Are we going to snapped mode?
+    if (ApplicationView::Value == ApplicationViewState::Snapped)
+    {
+        Grid::SetRow(DescriptionText, 3);
+        Grid::SetColumn(DescriptionText, 0);
+
+        Grid::SetRow(InputSection, 4);
+        Grid::SetColumn(InputSection, 0);
+
+        Grid::SetRow(FooterPanel, 2);
+        Grid::SetColumn(FooterPanel, 0);
+    }
+    else
+    {
+        Grid::SetRow(DescriptionText, 1);
+        Grid::SetColumn(DescriptionText, 1);
+
+        Grid::SetRow(InputSection, 2);
+        Grid::SetColumn(InputSection, 1);
+
+        Grid::SetRow(FooterPanel, 1);
+        Grid::SetColumn(FooterPanel, 1);
+    }
+
+    //  Since we don't load the scenario page in the traditional manner (we just pluck out the
+    // input and output sections from the page) we need to ensure that any VSM code used
+    // by the scenario's input and output sections is fired.
+    VisualStateManager::GoToState(InputSection, "Input" + LayoutAwarePage::DetermineVisualState(ApplicationView::Value), false);
+    VisualStateManager::GoToState(OutputSection, "Output" + LayoutAwarePage::DetermineVisualState(ApplicationView::Value), false);
+}
+
+void MainPage::PopulateScenarios()
+{
+    ScenarioList = ref new Platform::Collections::Vector<Object^>();
+
+    // Populate the ListBox with the list of scenarios as defined in Constants.cpp.
+    for (unsigned int i = 0; i < scenarios->Length; ++i)
+    {
+        Scenario s = scenarios[i];
+        ListBoxItem^ item = ref new ListBoxItem();
+        item->Name = s.ClassName;
+        item->Content = (i + 1).ToString() + ") " + s.Title;
+        ScenarioList->Append(item);
+    }
+
+    // Bind the ListBox to the scenario list.
+    Scenarios->ItemsSource = ScenarioList;
+    Scenarios->ScrollIntoView(Scenarios->SelectedItem);
+}
+
+/// <summary>
+/// This method is responsible for loading the individual input and output sections for each scenario.  This
+/// is based on navigating a hidden Frame to the ScenarioX.xaml page and then extracting out the input
+/// and output sections into the respective UserControl on the main page.
+/// </summary>
+/// <param name="scenarioName"></param>
+void MainPage::LoadScenario(String^ scenarioName)
+{
+    autoSizeInputSectionWhenSnapped = true;
+
+    // Load the ScenarioX.xaml file into the Frame.
+    TypeName scenarioType = {scenarioName, TypeKind::Custom};
+    HiddenFrame->Navigate(scenarioType, this);
+
+    // Get the top element, the Page, so we can look up the elements
+    // that represent the input and output sections of the ScenarioX file.
+    Page^ hiddenPage = safe_cast<Page^>(HiddenFrame->Content);
+
+    // Get each element.
+    UIElement^ input = safe_cast<UIElement^>(hiddenPage->FindName("Input"));
+    UIElement^ output = safe_cast<UIElement^>(hiddenPage->FindName("Output"));
+
+    if (input == nullptr)
+    {
+        // Malformed input section.
+        NotifyUser("Cannot load scenario input section for " + scenarioName +
+            "  Make sure root of input section markup has x:Name of 'Input'", NotifyType::ErrorMessage);
+        return;
+    }
+
+    if (output == nullptr)
+    {
+        // Malformed output section.
+        NotifyUser("Cannot load scenario output section for " + scenarioName +
+            "  Make sure root of output section markup has x:Name of 'Output'", NotifyType::ErrorMessage);
+        return;
+    }
+
+    // Find the LayoutRoot which parents the input and output sections in the main page.
+    Panel^ panel = safe_cast<Panel^>(hiddenPage->FindName("LayoutRoot"));
+
+    if (panel != nullptr)
+    {
+        unsigned int index = 0;
+        UIElementCollection^ collection = panel->Children;
+
+        // Get rid of the content that is currently in the intput and output sections.
+        collection->IndexOf(input, &index);
+        collection->RemoveAt(index);
+
+        collection->IndexOf(output, &index);
+        collection->RemoveAt(index);
+
+        // Populate the input and output sections with the newly loaded content.
+        InputSection->Content = input;
+        OutputSection->Content = output;
+
+        ScenarioLoaded(this, nullptr);
+    }
+    else
+    {
+        // Malformed Scenario file.
+        NotifyUser("Cannot load scenario: " + scenarioName + ".  Make sure root tag in the '" +
+            scenarioName + "' file has an x:Name of 'LayoutRoot'", NotifyType::ErrorMessage);
+    }
+}
+
+void MainPage::Scenarios_SelectionChanged(Object^ sender, SelectionChangedEventArgs^ e)
+{
+    if (Scenarios->SelectedItem != nullptr)
+    {
+        NotifyUser("", NotifyType::StatusMessage);
+
+        LoadScenario((safe_cast<ListBoxItem^>(Scenarios->SelectedItem))->Name);
+        InvalidateSize();
+    }
+}
+
+void MainPage::NotifyUser(String^ strMessage, NotifyType type)
+{
+    switch (type)
+    {
+    case NotifyType::StatusMessage:
+        // Use the status message style.
+        StatusBlock->Style = safe_cast<Windows::UI::Xaml::Style^>(this->Resources->Lookup("StatusStyle"));
+        break;
+    case NotifyType::ErrorMessage:
+        // Use the error message style.
+        StatusBlock->Style = safe_cast<Windows::UI::Xaml::Style^>(this->Resources->Lookup("ErrorStyle"));
+        break;
+    default:
+        break;
+    }
+    StatusBlock->Text = strMessage;
+
+    // Collapsed the StatusBlock if it has no text to conserve real estate.
+    if (StatusBlock->Text != "")
+    {
+        StatusBlock->Visibility = Windows::UI::Xaml::Visibility::Visible;
+    }
+    else
+    {
+        StatusBlock->Visibility = Windows::UI::Xaml::Visibility::Collapsed;
+    }
+}
+
+void MainPage::Footer_Click(Object^ sender, RoutedEventArgs^ e)
+{
+    auto uri = ref new Uri((String^)((HyperlinkButton^)sender)->Tag);
+    Windows::System::Launcher::LaunchUriAsync(uri);
+}
+
+
+/// <summary>
+/// Populates the page with content passed during navigation.  Any saved state is also
+/// provided when recreating a page from a prior session.
+/// </summary>
+/// <param name="navigationParameter">The parameter value passed to
+/// <see cref="Frame::Navigate(Type, Object)"/> when this page was initially requested.
+/// </param>
+/// <param name="pageState">A map of state preserved by this page during an earlier
+/// session.  This will be null the first time a page is visited.</param>
+void MainPage::LoadState(Object^ navigationParameter, IMap<String^, Object^>^ pageState)
+{
+    (void) navigationParameter;    // Unused parameter
+
+    PopulateScenarios();
+
+    // Starting scenario is the first or based upon a previous state.
+    ListBoxItem^ startingScenario = nullptr;
+    int startingScenarioIndex = -1;
+
+    if (pageState != nullptr && pageState->HasKey("SelectedScenarioIndex"))
+    {
+        startingScenarioIndex = safe_cast<int>(pageState->Lookup("SelectedScenarioIndex"));
+    }
+
+    Scenarios->SelectedIndex = startingScenarioIndex != -1 ? startingScenarioIndex : 0;
+
+    InvalidateViewState();
+}
+
+/// <summary>
+/// Preserves state associated with this page in case the application is suspended or the
+/// page is discarded from the navigation cache.  Values must conform to the serialization
+/// requirements of <see cref="SuspensionManager::SessionState"/>.
+/// </summary>
+/// <param name="pageState">An empty map to be populated with serializable state.</param>
+void MainPage::SaveState(IMap<String^, Object^>^ pageState)
+{
+    int selectedListBoxItemIndex = Scenarios->SelectedIndex;
+    pageState->Insert("SelectedScenarioIndex", selectedListBoxItemIndex);
+}
diff --git a/samples/winrt/ImageManipulations/MainPage.xaml.h b/samples/winrt/ImageManipulations/MainPage.xaml.h
new file mode 100644
index 0000000000..36fb7796a3
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MainPage.xaml.h
@@ -0,0 +1,105 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// MainPage.xaml.h
+// Declaration of the MainPage.xaml class.
+//
+
+#pragma once
+
+#include "pch.h"
+#include "MainPage.g.h"
+#include "Common\LayoutAwarePage.h" // Required by generated header
+#include "Constants.h"
+
+namespace SDKSample
+{
+    public enum class NotifyType
+    {
+        StatusMessage,
+        ErrorMessage
+    };
+
+    public ref class MainPageSizeChangedEventArgs sealed
+    {
+    public:
+        property Windows::UI::ViewManagement::ApplicationViewState ViewState
+        {
+            Windows::UI::ViewManagement::ApplicationViewState get()
+            {
+                return viewState;
+            }
+
+            void set(Windows::UI::ViewManagement::ApplicationViewState value)
+            {
+                viewState = value;
+            }
+        }
+
+    private:
+        Windows::UI::ViewManagement::ApplicationViewState viewState;
+    };
+
+    public ref class MainPage sealed
+    {
+    public:
+        MainPage();
+
+    protected:
+        virtual void LoadState(Platform::Object^ navigationParameter,
+            Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ pageState) override;
+        virtual void SaveState(Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ pageState) override;
+
+    internal:
+        property bool AutoSizeInputSectionWhenSnapped
+        {
+            bool get()
+            {
+                return autoSizeInputSectionWhenSnapped;
+            }
+
+            void set(bool value)
+            {
+                autoSizeInputSectionWhenSnapped = value;
+            }
+        }
+
+        property Windows::ApplicationModel::Activation::LaunchActivatedEventArgs^ LaunchArgs
+       {
+            Windows::ApplicationModel::Activation::LaunchActivatedEventArgs^ get()
+            {
+                return safe_cast<App^>(App::Current)->LaunchArgs;
+            }
+        }
+
+        void NotifyUser(Platform::String^ strMessage, NotifyType type);
+        void LoadScenario(Platform::String^ scenarioName);
+        event Windows::Foundation::EventHandler<Platform::Object^>^ ScenarioLoaded;
+        event Windows::Foundation::EventHandler<MainPageSizeChangedEventArgs^>^ MainPageResized;
+
+    private:
+        void PopulateScenarios();
+        void InvalidateSize();
+        void InvalidateViewState();
+
+        Platform::Collections::Vector<Object^>^ ScenarioList;
+        Windows::UI::Xaml::Controls::Frame^ HiddenFrame;
+        void Footer_Click(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+        bool autoSizeInputSectionWhenSnapped;
+
+        void MainPage_SizeChanged(Object^ sender, Windows::UI::Xaml::SizeChangedEventArgs^ e);
+        void Scenarios_SelectionChanged(Object^ sender, Windows::UI::Xaml::Controls::SelectionChangedEventArgs^ e);
+
+    internal:
+        static MainPage^ Current;
+
+    };
+}
diff --git a/samples/winrt/ImageManipulations/MediaCapture.sln b/samples/winrt/ImageManipulations/MediaCapture.sln
new file mode 100644
index 0000000000..e1f9aa3859
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaCapture.sln
@@ -0,0 +1,52 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio 11 Express for Windows 8
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "MediaCapture", "MediaCapture.vcxproj", "{C5B886A7-8300-46FF-B533-9613DE2AF637}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "GrayscaleTransform", "MediaExtensions\OcvTransform\OcvTransform.vcxproj", "{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|ARM = Debug|ARM
+		Debug|Win32 = Debug|Win32
+		Debug|x64 = Debug|x64
+		Release|ARM = Release|ARM
+		Release|Win32 = Release|Win32
+		Release|x64 = Release|x64
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|ARM.ActiveCfg = Debug|ARM
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|ARM.Build.0 = Debug|ARM
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|Win32.ActiveCfg = Debug|Win32
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|Win32.Build.0 = Debug|Win32
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|x64.ActiveCfg = Debug|x64
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Debug|x64.Build.0 = Debug|x64
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|ARM.ActiveCfg = Release|ARM
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|ARM.Build.0 = Release|ARM
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|Win32.ActiveCfg = Release|Win32
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|Win32.Build.0 = Release|Win32
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|x64.ActiveCfg = Release|x64
+		{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}.Release|x64.Build.0 = Release|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|ARM.ActiveCfg = Debug|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|ARM.Build.0 = Debug|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|ARM.Deploy.0 = Debug|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|Win32.ActiveCfg = Debug|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|Win32.Build.0 = Debug|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|Win32.Deploy.0 = Debug|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|x64.ActiveCfg = Debug|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|x64.Build.0 = Debug|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Debug|x64.Deploy.0 = Debug|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|ARM.ActiveCfg = Release|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|ARM.Build.0 = Release|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|ARM.Deploy.0 = Release|ARM
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|Win32.ActiveCfg = Release|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|Win32.Build.0 = Release|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|Win32.Deploy.0 = Release|Win32
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|x64.ActiveCfg = Release|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|x64.Build.0 = Release|x64
+		{C5B886A7-8300-46FF-B533-9613DE2AF637}.Release|x64.Deploy.0 = Release|x64
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+EndGlobal
diff --git a/samples/winrt/ImageManipulations/MediaCapture.vcxproj b/samples/winrt/ImageManipulations/MediaCapture.vcxproj
new file mode 100644
index 0000000000..a0f8c84f9b
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaCapture.vcxproj
@@ -0,0 +1,310 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|ARM">
+      <Configuration>Debug</Configuration>
+      <Platform>ARM</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|Win32">
+      <Configuration>Debug</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|ARM">
+      <Configuration>Release</Configuration>
+      <Platform>ARM</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|Win32">
+      <Configuration>Release</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{C5B886A7-8300-46FF-B533-9613DE2AF637}</ProjectGuid>
+    <RootNamespace>SDKSample</RootNamespace>
+    <DefaultLanguage>en-US</DefaultLanguage>
+    <VCTargetsPath Condition="'$(VCTargetsPath11)' != '' and '$(VSVersion)' == '' and '$(VisualStudioVersion)' == ''">$(VCTargetsPath11)</VCTargetsPath>
+    <MinimumVisualStudioVersion>11.0</MinimumVisualStudioVersion>
+    <AppContainerApplication>true</AppContainerApplication>
+    <ProjectName>MediaCapture</ProjectName>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <ItemDefinitionGroup>
+    <ClCompile>
+      <PrecompiledHeaderFile>pch.h</PrecompiledHeaderFile>
+    </ClCompile>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <ClInclude Include="AdvancedCapture.xaml.h">
+      <DependentUpon>AdvancedCapture.xaml</DependentUpon>
+      <SubType>Code</SubType>
+    </ClInclude>
+    <ClInclude Include="Constants.h" />
+    <ClInclude Include="MainPage.xaml.h">
+      <DependentUpon>MainPage.xaml</DependentUpon>
+    </ClInclude>
+    <ClInclude Include="pch.h" />
+    <ClInclude Include="Common\LayoutAwarePage.h" />
+    <ClInclude Include="Common\SuspensionManager.h" />
+    <ClInclude Include="App.xaml.h">
+      <DependentUpon>App.xaml</DependentUpon>
+    </ClInclude>
+  </ItemGroup>
+  <ItemGroup>
+    <ApplicationDefinition Include="App.xaml">
+      <SubType>Designer</SubType>
+    </ApplicationDefinition>
+    <Page Include="AdvancedCapture.xaml">
+      <SubType>Designer</SubType>
+    </Page>
+    <Page Include="Common\StandardStyles.xaml">
+      <SubType>Designer</SubType>
+    </Page>
+    <Page Include="MainPage.xaml" />
+    <Page Include="Sample-Utils\SampleTemplateStyles.xaml">
+      <SubType>Designer</SubType>
+    </Page>
+  </ItemGroup>
+  <ItemGroup>
+    <AppxManifest Include="Package.appxmanifest">
+      <SubType>Designer</SubType>
+    </AppxManifest>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="AdvancedCapture.xaml.cpp">
+      <DependentUpon>AdvancedCapture.xaml</DependentUpon>
+      <SubType>Code</SubType>
+    </ClCompile>
+    <ClCompile Include="App.xaml.cpp">
+      <DependentUpon>App.xaml</DependentUpon>
+    </ClCompile>
+    <ClCompile Include="Common\LayoutAwarePage.cpp" />
+    <ClCompile Include="Constants.cpp" />
+    <ClCompile Include="Common\SuspensionManager.cpp" />
+    <ClCompile Include="MainPage.xaml.cpp">
+      <DependentUpon>MainPage.xaml</DependentUpon>
+    </ClCompile>
+    <ClCompile Include="pch.cpp">
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">Create</PrecompiledHeader>
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">Create</PrecompiledHeader>
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">Create</PrecompiledHeader>
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">Create</PrecompiledHeader>
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">Create</PrecompiledHeader>
+      <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|x64'">Create</PrecompiledHeader>
+    </ClCompile>
+  </ItemGroup>
+  <ItemGroup>
+    <Image Include="assets\opencv-logo-150.png" />
+    <Image Include="assets\opencv-logo-30.png" />
+    <Image Include="Assets\splash-sdk.png" />
+    <Image Include="Assets\windows-sdk.png" />
+  </ItemGroup>
+  <ItemGroup>
+    <ProjectReference Include="MediaExtensions\OcvTransform\OcvTransform.vcxproj">
+      <Project>{ba69218f-da5c-4d14-a78d-21a9e4dec669}</Project>
+    </ProjectReference>
+  </ItemGroup>
+  <ItemGroup>
+    <None Include="$(OPENCV_DIR)\bin\opencv_calib3d245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_contrib245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_core245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_features2d245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_flann245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_highgui245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_imgproc245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_legacy245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_ml245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_nonfree245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_objdetect245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_photo245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_stitching245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_superres245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_ts245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_video245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+    <None Include="$(OPENCV_DIR)\bin\opencv_videostab245.dll">
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</DeploymentContent>
+      <DeploymentContent Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</DeploymentContent>
+    </None>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/Common/AsyncCB.h b/samples/winrt/ImageManipulations/MediaExtensions/Common/AsyncCB.h
new file mode 100644
index 0000000000..3321b47c95
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/Common/AsyncCB.h
@@ -0,0 +1,81 @@
+#pragma once
+
+//////////////////////////////////////////////////////////////////////////
+//  AsyncCallback [template]
+//
+//  Description:
+//  Helper class that routes IMFAsyncCallback::Invoke calls to a class
+//  method on the parent class.
+//
+//  Usage:
+//  Add this class as a member variable. In the parent class constructor,
+//  initialize the AsyncCallback class like this:
+//  	m_cb(this, &CYourClass::OnInvoke)
+//  where
+//      m_cb       = AsyncCallback object
+//      CYourClass = parent class
+//      OnInvoke   = Method in the parent class to receive Invoke calls.
+//
+//  The parent's OnInvoke method (you can name it anything you like) must
+//  have a signature that matches the InvokeFn typedef below.
+//////////////////////////////////////////////////////////////////////////
+
+// T: Type of the parent object
+template<class T>
+class AsyncCallback : public IMFAsyncCallback
+{
+public:
+    typedef HRESULT (T::*InvokeFn)(IMFAsyncResult *pAsyncResult);
+
+    AsyncCallback(T *pParent, InvokeFn fn) : m_pParent(pParent), m_pInvokeFn(fn)
+    {
+    }
+
+    // IUnknown
+    STDMETHODIMP_(ULONG) AddRef() {
+        // Delegate to parent class.
+        return m_pParent->AddRef();
+    }
+    STDMETHODIMP_(ULONG) Release() {
+        // Delegate to parent class.
+        return m_pParent->Release();
+    }
+    STDMETHODIMP QueryInterface(REFIID iid, void** ppv)
+    {
+        if (!ppv)
+        {
+            return E_POINTER;
+        }
+        if (iid == __uuidof(IUnknown))
+        {
+            *ppv = static_cast<IUnknown*>(static_cast<IMFAsyncCallback*>(this));
+        }
+        else if (iid == __uuidof(IMFAsyncCallback))
+        {
+            *ppv = static_cast<IMFAsyncCallback*>(this);
+        }
+        else
+        {
+            *ppv = NULL;
+            return E_NOINTERFACE;
+        }
+        AddRef();
+        return S_OK;
+    }
+
+
+    // IMFAsyncCallback methods
+    STDMETHODIMP GetParameters(DWORD*, DWORD*)
+    {
+        // Implementation of this method is optional.
+        return E_NOTIMPL;
+    }
+
+    STDMETHODIMP Invoke(IMFAsyncResult* pAsyncResult)
+    {
+        return (m_pParent->*m_pInvokeFn)(pAsyncResult);
+    }
+
+    T *m_pParent;
+    InvokeFn m_pInvokeFn;
+};
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/Common/BufferLock.h b/samples/winrt/ImageManipulations/MediaExtensions/Common/BufferLock.h
new file mode 100644
index 0000000000..1c47deabd0
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/Common/BufferLock.h
@@ -0,0 +1,101 @@
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved
+
+
+#pragma once
+
+
+//////////////////////////////////////////////////////////////////////////
+//  VideoBufferLock
+//
+//  Description:
+//  Locks a video buffer that might or might not support IMF2DBuffer.
+//
+//////////////////////////////////////////////////////////////////////////
+
+class VideoBufferLock
+{
+public:
+    VideoBufferLock(IMFMediaBuffer *pBuffer) : m_p2DBuffer(NULL)
+    {
+        m_pBuffer = pBuffer;
+        m_pBuffer->AddRef();
+
+        // Query for the 2-D buffer interface. OK if this fails.
+        m_pBuffer->QueryInterface(IID_PPV_ARGS(&m_p2DBuffer));
+    }
+
+    ~VideoBufferLock()
+    {
+        UnlockBuffer();
+        SafeRelease(&m_pBuffer);
+        SafeRelease(&m_p2DBuffer);
+    }
+
+    // LockBuffer:
+    // Locks the buffer. Returns a pointer to scan line 0 and returns the stride.
+
+    // The caller must provide the default stride as an input parameter, in case
+    // the buffer does not expose IMF2DBuffer. You can calculate the default stride
+    // from the media type.
+
+    HRESULT LockBuffer(
+        LONG  lDefaultStride,    // Minimum stride (with no padding).
+        DWORD dwHeightInPixels,  // Height of the image, in pixels.
+        BYTE  **ppbScanLine0,    // Receives a pointer to the start of scan line 0.
+        LONG  *plStride          // Receives the actual stride.
+        )
+    {
+        HRESULT hr = S_OK;
+
+        // Use the 2-D version if available.
+        if (m_p2DBuffer)
+        {
+            hr = m_p2DBuffer->Lock2D(ppbScanLine0, plStride);
+        }
+        else
+        {
+            // Use non-2D version.
+            BYTE *pData = NULL;
+
+            hr = m_pBuffer->Lock(&pData, NULL, NULL);
+            if (SUCCEEDED(hr))
+            {
+                *plStride = lDefaultStride;
+                if (lDefaultStride < 0)
+                {
+                    // Bottom-up orientation. Return a pointer to the start of the
+                    // last row *in memory* which is the top row of the image.
+                    *ppbScanLine0 = pData + abs(lDefaultStride) * (dwHeightInPixels - 1);
+                }
+                else
+                {
+                    // Top-down orientation. Return a pointer to the start of the
+                    // buffer.
+                    *ppbScanLine0 = pData;
+                }
+            }
+        }
+        return hr;
+    }
+
+    HRESULT UnlockBuffer()
+    {
+        if (m_p2DBuffer)
+        {
+            return m_p2DBuffer->Unlock2D();
+        }
+        else
+        {
+            return m_pBuffer->Unlock();
+        }
+    }
+
+private:
+    IMFMediaBuffer  *m_pBuffer;
+    IMF2DBuffer     *m_p2DBuffer;
+};
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/Common/CritSec.h b/samples/winrt/ImageManipulations/MediaExtensions/Common/CritSec.h
new file mode 100644
index 0000000000..88851f9f03
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/Common/CritSec.h
@@ -0,0 +1,62 @@
+#pragma once
+
+//////////////////////////////////////////////////////////////////////////
+//  CritSec
+//  Description: Wraps a critical section.
+//////////////////////////////////////////////////////////////////////////
+
+class CritSec
+{
+public:
+    CRITICAL_SECTION m_criticalSection;
+public:
+    CritSec()
+    {
+        InitializeCriticalSectionEx(&m_criticalSection, 100, 0);
+    }
+
+    ~CritSec()
+    {
+        DeleteCriticalSection(&m_criticalSection);
+    }
+
+    _Acquires_lock_(m_criticalSection)
+    void Lock()
+    {
+        EnterCriticalSection(&m_criticalSection);
+    }
+
+    _Releases_lock_(m_criticalSection)
+    void Unlock()
+    {
+        LeaveCriticalSection(&m_criticalSection);
+    }
+};
+
+
+//////////////////////////////////////////////////////////////////////////
+//  AutoLock
+//  Description: Provides automatic locking and unlocking of a
+//               of a critical section.
+//
+//  Note: The AutoLock object must go out of scope before the CritSec.
+//////////////////////////////////////////////////////////////////////////
+
+class AutoLock
+{
+private:
+    CritSec *m_pCriticalSection;
+public:
+    _Acquires_lock_(m_pCriticalSection)
+    AutoLock(CritSec& crit)
+    {
+        m_pCriticalSection = &crit;
+        m_pCriticalSection->Lock();
+    }
+
+    _Releases_lock_(m_pCriticalSection)
+    ~AutoLock()
+    {
+        m_pCriticalSection->Unlock();
+    }
+};
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/Common/LinkList.h b/samples/winrt/ImageManipulations/MediaExtensions/Common/LinkList.h
new file mode 100644
index 0000000000..3657b3d7c5
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/Common/LinkList.h
@@ -0,0 +1,516 @@
+//-----------------------------------------------------------------------------
+// File: Linklist.h
+// Desc: Linked list class.
+//
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+//  Copyright (C) Microsoft Corporation. All rights reserved.
+//-----------------------------------------------------------------------------
+
+#pragma once
+
+// Notes:
+//
+// The List class template implements a simple double-linked list.
+// It uses STL's copy semantics.
+
+// There are two versions of the Clear() method:
+//  Clear(void) clears the list w/out cleaning up the object.
+//  Clear(FN fn) takes a functor object that releases the objects, if they need cleanup.
+
+// The List class supports enumeration. Example of usage:
+//
+// List<T>::POSIITON pos = list.GetFrontPosition();
+// while (pos != list.GetEndPosition())
+// {
+//     T item;
+//     hr = list.GetItemPos(&item);
+//     pos = list.Next(pos);
+// }
+
+// The ComPtrList class template derives from List<> and implements a list of COM pointers.
+
+template <class T>
+struct NoOp
+{
+    void operator()(T& t)
+    {
+    }
+};
+
+template <class T>
+class List
+{
+protected:
+
+    // Nodes in the linked list
+    struct Node
+    {
+        Node *prev;
+        Node *next;
+        T    item;
+
+        Node() : prev(nullptr), next(nullptr)
+        {
+        }
+
+        Node(T item) : prev(nullptr), next(nullptr)
+        {
+            this->item = item;
+        }
+
+        T Item() const { return item; }
+    };
+
+public:
+
+    // Object for enumerating the list.
+    class POSITION
+    {
+        friend class List<T>;
+
+    public:
+        POSITION() : pNode(nullptr)
+        {
+        }
+
+        bool operator==(const POSITION &p) const
+        {
+            return pNode == p.pNode;
+        }
+
+        bool operator!=(const POSITION &p) const
+        {
+            return pNode != p.pNode;
+        }
+
+    private:
+        const Node *pNode;
+
+        POSITION(Node *p) : pNode(p)
+        {
+        }
+    };
+
+protected:
+    Node    m_anchor;  // Anchor node for the linked list.
+    DWORD   m_count;   // Number of items in the list.
+
+    Node* Front() const
+    {
+        return m_anchor.next;
+    }
+
+    Node* Back() const
+    {
+        return m_anchor.prev;
+    }
+
+    virtual HRESULT InsertAfter(T item, Node *pBefore)
+    {
+        if (pBefore == nullptr)
+        {
+            return E_POINTER;
+        }
+
+        Node *pNode = new Node(item);
+        if (pNode == nullptr)
+        {
+            return E_OUTOFMEMORY;
+        }
+
+        Node *pAfter = pBefore->next;
+
+        pBefore->next = pNode;
+        pAfter->prev = pNode;
+
+        pNode->prev = pBefore;
+        pNode->next = pAfter;
+
+        m_count++;
+
+        return S_OK;
+    }
+
+    virtual HRESULT GetItem(const Node *pNode, T* ppItem)
+    {
+        if (pNode == nullptr || ppItem == nullptr)
+        {
+            return E_POINTER;
+        }
+
+        *ppItem = pNode->item;
+        return S_OK;
+    }
+
+    // RemoveItem:
+    // Removes a node and optionally returns the item.
+    // ppItem can be nullptr.
+    virtual HRESULT RemoveItem(Node *pNode, T *ppItem)
+    {
+        if (pNode == nullptr)
+        {
+            return E_POINTER;
+        }
+
+        assert(pNode != &m_anchor); // We should never try to remove the anchor node.
+        if (pNode == &m_anchor)
+        {
+            return E_INVALIDARG;
+        }
+
+
+        T item;
+
+        // The next node's previous is this node's previous.
+        pNode->next->prev = pNode->prev;
+
+        // The previous node's next is this node's next.
+        pNode->prev->next = pNode->next;
+
+        item = pNode->item;
+        delete pNode;
+
+        m_count--;
+
+        if (ppItem)
+        {
+            *ppItem = item;
+        }
+
+        return S_OK;
+    }
+
+public:
+
+    List()
+    {
+        m_anchor.next = &m_anchor;
+        m_anchor.prev = &m_anchor;
+
+        m_count = 0;
+    }
+
+    virtual ~List()
+    {
+        Clear();
+    }
+
+    // Insertion functions
+    HRESULT InsertBack(T item)
+    {
+        return InsertAfter(item, m_anchor.prev);
+    }
+
+
+    HRESULT InsertFront(T item)
+    {
+        return InsertAfter(item, &m_anchor);
+    }
+
+    HRESULT InsertPos(POSITION pos, T item)
+    {
+        if (pos.pNode == nullptr)
+        {
+            return InsertBack(item);
+        }
+
+        return InsertAfter(item, pos.pNode->prev);
+    }
+
+    // RemoveBack: Removes the tail of the list and returns the value.
+    // ppItem can be nullptr if you don't want the item back. (But the method does not release the item.)
+    HRESULT RemoveBack(T *ppItem)
+    {
+        if (IsEmpty())
+        {
+            return E_FAIL;
+        }
+        else
+        {
+            return RemoveItem(Back(), ppItem);
+        }
+    }
+
+    // RemoveFront: Removes the head of the list and returns the value.
+    // ppItem can be nullptr if you don't want the item back. (But the method does not release the item.)
+    HRESULT RemoveFront(T *ppItem)
+    {
+        if (IsEmpty())
+        {
+            return E_FAIL;
+        }
+        else
+        {
+            return RemoveItem(Front(), ppItem);
+        }
+    }
+
+    // GetBack: Gets the tail item.
+    HRESULT GetBack(T *ppItem)
+    {
+        if (IsEmpty())
+        {
+            return E_FAIL;
+        }
+        else
+        {
+            return GetItem(Back(), ppItem);
+        }
+    }
+
+    // GetFront: Gets the front item.
+    HRESULT GetFront(T *ppItem)
+    {
+        if (IsEmpty())
+        {
+            return E_FAIL;
+        }
+        else
+        {
+            return GetItem(Front(), ppItem);
+        }
+    }
+
+
+    // GetCount: Returns the number of items in the list.
+    DWORD GetCount() const { return m_count; }
+
+    bool IsEmpty() const
+    {
+        return (GetCount() == 0);
+    }
+
+    // Clear: Takes a functor object whose operator()
+    // frees the object on the list.
+    template <class FN>
+    void Clear(FN& clear_fn)
+    {
+        Node *n = m_anchor.next;
+
+        // Delete the nodes
+        while (n != &m_anchor)
+        {
+            clear_fn(n->item);
+
+            Node *tmp = n->next;
+            delete n;
+            n = tmp;
+        }
+
+        // Reset the anchor to point at itself
+        m_anchor.next = &m_anchor;
+        m_anchor.prev = &m_anchor;
+
+        m_count = 0;
+    }
+
+    // Clear: Clears the list. (Does not delete or release the list items.)
+    virtual void Clear()
+    {
+        NoOp<T> clearOp;
+        Clear<>(clearOp);
+    }
+
+
+    // Enumerator functions
+
+    POSITION FrontPosition()
+    {
+        if (IsEmpty())
+        {
+            return POSITION(nullptr);
+        }
+        else
+        {
+            return POSITION(Front());
+        }
+    }
+
+    POSITION EndPosition() const
+    {
+        return POSITION();
+    }
+
+    HRESULT GetItemPos(POSITION pos, T *ppItem)
+    {
+        if (pos.pNode)
+        {
+            return GetItem(pos.pNode, ppItem);
+        }
+        else
+        {
+            return E_FAIL;
+        }
+    }
+
+    POSITION Next(const POSITION pos)
+    {
+        if (pos.pNode && (pos.pNode->next != &m_anchor))
+        {
+            return POSITION(pos.pNode->next);
+        }
+        else
+        {
+            return POSITION(nullptr);
+        }
+    }
+
+    // Remove an item at a position.
+    // The item is returns in ppItem, unless ppItem is nullptr.
+    // NOTE: This method invalidates the POSITION object.
+    HRESULT Remove(POSITION& pos, T *ppItem)
+    {
+        if (pos.pNode)
+        {
+            // Remove const-ness temporarily...
+            Node *pNode = const_cast<Node*>(pos.pNode);
+
+            pos = POSITION();
+
+            return RemoveItem(pNode, ppItem);
+        }
+        else
+        {
+            return E_INVALIDARG;
+        }
+    }
+
+};
+
+
+
+// Typical functors for Clear method.
+
+// ComAutoRelease: Releases COM pointers.
+// MemDelete: Deletes pointers to new'd memory.
+
+class ComAutoRelease
+{
+public:
+    void operator()(IUnknown *p)
+    {
+        if (p)
+        {
+            p->Release();
+        }
+    }
+};
+
+class MemDelete
+{
+public:
+    void operator()(void *p)
+    {
+        if (p)
+        {
+            delete p;
+        }
+    }
+};
+
+
+// ComPtrList class
+// Derived class that makes it safer to store COM pointers in the List<> class.
+// It automatically AddRef's the pointers that are inserted onto the list
+// (unless the insertion method fails).
+//
+// T must be a COM interface type.
+// example: ComPtrList<IUnknown>
+//
+// NULLABLE: If true, client can insert nullptr pointers. This means GetItem can
+// succeed but return a nullptr pointer. By default, the list does not allow nullptr
+// pointers.
+
+template <class T, bool NULLABLE = FALSE>
+class ComPtrList : public List<T*>
+{
+public:
+
+    typedef T* Ptr;
+
+    void Clear()
+    {
+        ComAutoRelease car;
+        List<Ptr>::Clear(car);
+    }
+
+    ~ComPtrList()
+    {
+        Clear();
+    }
+
+protected:
+    HRESULT InsertAfter(Ptr item, Node *pBefore)
+    {
+        // Do not allow nullptr item pointers unless NULLABLE is true.
+        if (item == nullptr && !NULLABLE)
+        {
+            return E_POINTER;
+        }
+
+        if (item)
+        {
+            item->AddRef();
+        }
+
+        HRESULT hr = List<Ptr>::InsertAfter(item, pBefore);
+        if (FAILED(hr) && item != nullptr)
+        {
+            item->Release();
+        }
+        return hr;
+    }
+
+    HRESULT GetItem(const Node *pNode, Ptr* ppItem)
+    {
+        Ptr pItem = nullptr;
+
+        // The base class gives us the pointer without AddRef'ing it.
+        // If we return the pointer to the caller, we must AddRef().
+        HRESULT hr = List<Ptr>::GetItem(pNode, &pItem);
+        if (SUCCEEDED(hr))
+        {
+            assert(pItem || NULLABLE);
+            if (pItem)
+            {
+                *ppItem = pItem;
+                (*ppItem)->AddRef();
+            }
+        }
+        return hr;
+    }
+
+    HRESULT RemoveItem(Node *pNode, Ptr *ppItem)
+    {
+        // ppItem can be nullptr, but we need to get the
+        // item so that we can release it.
+
+        // If ppItem is not nullptr, we will AddRef it on the way out.
+
+        Ptr pItem = nullptr;
+
+        HRESULT hr = List<Ptr>::RemoveItem(pNode, &pItem);
+
+        if (SUCCEEDED(hr))
+        {
+            assert(pItem || NULLABLE);
+            if (ppItem && pItem)
+            {
+                *ppItem = pItem;
+                (*ppItem)->AddRef();
+            }
+
+            if (pItem)
+            {
+                pItem->Release();
+                pItem = nullptr;
+            }
+        }
+
+        return hr;
+    }
+};
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/Common/OpQueue.h b/samples/winrt/ImageManipulations/MediaExtensions/Common/OpQueue.h
new file mode 100644
index 0000000000..73f3042ce3
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/Common/OpQueue.h
@@ -0,0 +1,222 @@
+//////////////////////////////////////////////////////////////////////////
+//
+// OpQueue.h
+// Async operation queue.
+//
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved.
+//
+//////////////////////////////////////////////////////////////////////////
+
+#pragma once
+
+#pragma warning( push )
+#pragma warning( disable : 4355 )  // 'this' used in base member initializer list
+
+/*
+    This header file defines an object to help queue and serialize
+    asynchronous operations.
+
+    Background:
+
+    To perform an operation asynchronously in Media Foundation, an object
+    does one of the following:
+
+        1. Calls MFPutWorkItem(Ex), using either a standard work queue
+           identifier or a caller-allocated work queue. The work-queue
+           thread invokes the object's callback.
+
+        2. Creates an async result object (IMFAsyncResult) and calls
+           MFInvokeCallback to invoke the object's callback.
+
+    Ultimately, either of these cause the object's callback to be invoked
+    from a work-queue thread. The object can then complete the operation
+    inside the callback.
+
+    However, the Media Foundation platform may dispatch async callbacks in
+    parallel on several threads. Putting an item on a work queue does NOT
+    guarantee that one operation will complete before the next one starts,
+    or even that work items will be dispatched in the same order they were
+    called.
+
+    To serialize async operations that should not overlap, an object should
+    use a queue. While one operation is pending, subsequent operations are
+    put on the queue, and only dispatched after the previous operation is
+    complete.
+
+    The granularity of a single "operation" depends on the requirements of
+    that particular object. A single operation might involve several
+    asynchronous calls before the object dispatches the next operation on
+    the queue.
+
+
+*/
+
+
+
+//-------------------------------------------------------------------
+// OpQueue class template
+//
+// Base class for an async operation queue.
+//
+// TOperation: The class used to describe operations. This class must
+//          implement IUnknown.
+//
+// The OpQueue class is an abstract class. The derived class must
+// implement the following pure-virtual methods:
+//
+// - IUnknown methods (AddRef, Release, QI)
+//
+// - DispatchOperation:
+//
+//      Performs the asynchronous operation specified by pOp.
+//
+//      At the end of each operation, the derived class must call
+//      ProcessQueue to process the next operation in the queue.
+//
+//      NOTE: An operation is not required to complete inside the
+//      DispatchOperation method. A single operation might consist
+//      of several asynchronous method calls.
+//
+// - ValidateOperation:
+//
+//      Checks whether the object can perform the operation specified
+//      by pOp at this time.
+//
+//      If the object cannot perform the operation now (e.g., because
+//      another operation is still in progress) the method should
+//      return MF_E_NOTACCEPTING.
+//
+//-------------------------------------------------------------------
+#include "linklist.h"
+#include "AsyncCB.h"
+
+template <class T, class TOperation>
+class OpQueue //: public IUnknown
+{
+public:
+
+    typedef ComPtrList<TOperation>   OpList;
+
+    HRESULT QueueOperation(TOperation *pOp);
+
+protected:
+
+    HRESULT ProcessQueue();
+    HRESULT ProcessQueueAsync(IMFAsyncResult *pResult);
+
+    virtual HRESULT DispatchOperation(TOperation *pOp) = 0;
+    virtual HRESULT ValidateOperation(TOperation *pOp) = 0;
+
+    OpQueue(CRITICAL_SECTION& critsec)
+        : m_OnProcessQueue(static_cast<T *>(this), &OpQueue::ProcessQueueAsync),
+          m_critsec(critsec)
+    {
+    }
+
+    virtual ~OpQueue()
+    {
+    }
+
+protected:
+    OpList                  m_OpQueue;         // Queue of operations.
+    CRITICAL_SECTION&       m_critsec;         // Protects the queue state.
+    AsyncCallback<T>  m_OnProcessQueue;  // ProcessQueueAsync callback.
+};
+
+
+
+//-------------------------------------------------------------------
+// Place an operation on the queue.
+// Public method.
+//-------------------------------------------------------------------
+
+template <class T, class TOperation>
+HRESULT OpQueue<T, TOperation>::QueueOperation(TOperation *pOp)
+{
+    HRESULT hr = S_OK;
+
+    EnterCriticalSection(&m_critsec);
+
+    hr = m_OpQueue.InsertBack(pOp);
+    if (SUCCEEDED(hr))
+    {
+        hr = ProcessQueue();
+    }
+
+    LeaveCriticalSection(&m_critsec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// Process the next operation on the queue.
+// Protected method.
+//
+// Note: This method dispatches the operation to a work queue.
+//-------------------------------------------------------------------
+
+template <class T, class TOperation>
+HRESULT OpQueue<T, TOperation>::ProcessQueue()
+{
+    HRESULT hr = S_OK;
+    if (m_OpQueue.GetCount() > 0)
+    {
+        hr = MFPutWorkItem2(
+            MFASYNC_CALLBACK_QUEUE_STANDARD,    // Use the standard work queue.
+            0,                                  // Default priority
+            &m_OnProcessQueue,                  // Callback method.
+            nullptr                             // State object.
+            );
+    }
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// Process the next operation on the queue.
+// Protected method.
+//
+// Note: This method is called from a work-queue thread.
+//-------------------------------------------------------------------
+
+template <class T, class TOperation>
+HRESULT OpQueue<T, TOperation>::ProcessQueueAsync(IMFAsyncResult *pResult)
+{
+    HRESULT hr = S_OK;
+    TOperation *pOp = nullptr;
+
+    EnterCriticalSection(&m_critsec);
+
+    if (m_OpQueue.GetCount() > 0)
+    {
+        hr = m_OpQueue.GetFront(&pOp);
+
+        if (SUCCEEDED(hr))
+        {
+            hr = ValidateOperation(pOp);
+        }
+        if (SUCCEEDED(hr))
+        {
+            hr = m_OpQueue.RemoveFront(nullptr);
+        }
+        if (SUCCEEDED(hr))
+        {
+            (void)DispatchOperation(pOp);
+        }
+    }
+
+    if (pOp != nullptr)
+    {
+        pOp->Release();
+    }
+
+    LeaveCriticalSection(&m_critsec);
+    return hr;
+}
+
+#pragma warning( pop )
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvImageManipulations.idl b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvImageManipulations.idl
new file mode 100644
index 0000000000..37235a092d
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvImageManipulations.idl
@@ -0,0 +1,11 @@
+import "Windows.Media.idl";
+
+#include <sdkddkver.h>
+
+namespace OcvTransform
+{
+    [version(NTDDI_WIN8)]
+    runtimeclass OcvImageManipulations
+    {
+    }
+}
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.cpp b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.cpp
new file mode 100644
index 0000000000..538f353f5e
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.cpp
@@ -0,0 +1,1485 @@
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved.
+
+#include "OcvTransform.h"
+#include "bufferlock.h"
+
+#include "opencv2\core\core.hpp"
+#include "opencv2\imgproc\imgproc.hpp"
+
+using namespace Microsoft::WRL;
+
+/*
+
+This sample implements a video effect as a Media Foundation transform (MFT).
+
+NOTES ON THE MFT IMPLEMENTATION
+
+1. The MFT has fixed streams: One input stream and one output stream.
+
+2. The MFT supports NV12 format only.
+
+3. If the MFT is holding an input sample, SetInputType and SetOutputType both fail.
+
+4. The input and output types must be identical.
+
+5. If both types are set, no type can be set until the current type is cleared.
+
+6. Preferred input types:
+
+     (a) If the output type is set, that's the preferred type.
+     (b) Otherwise, the preferred types are partial types, constructed from the
+         list of supported subtypes.
+
+7. Preferred output types: As above.
+
+8. Streaming:
+
+    The private BeingStreaming() method is called in response to the
+    MFT_MESSAGE_NOTIFY_BEGIN_STREAMING message.
+
+    If the client does not send MFT_MESSAGE_NOTIFY_BEGIN_STREAMING, the MFT calls
+    BeginStreaming inside the first call to ProcessInput or ProcessOutput.
+
+    This is a good approach for allocating resources that your MFT requires for
+    streaming.
+
+9. The configuration attributes are applied in the BeginStreaming method. If the
+   client changes the attributes during streaming, the change is ignored until
+   streaming is stopped (either by changing the media types or by sending the
+   MFT_MESSAGE_NOTIFY_END_STREAMING message) and then restarted.
+
+*/
+
+
+// Static array of media types (preferred and accepted).
+const GUID g_MediaSubtypes[] =
+{
+    MFVideoFormat_NV12
+};
+
+HRESULT GetDefaultStride(IMFMediaType *pType, LONG *plStride);
+
+template <typename T>
+inline T clamp(const T& val, const T& minVal, const T& maxVal)
+{
+    return (val < minVal ? minVal : (val > maxVal ? maxVal : val));
+}
+
+OcvImageManipulations::OcvImageManipulations() :
+    m_pSample(NULL), m_pInputType(NULL), m_pOutputType(NULL),
+    m_imageWidthInPixels(0), m_imageHeightInPixels(0), m_cbImageSize(0),
+    m_TransformType(Preview), m_bStreamingInitialized(false),
+    m_pAttributes(NULL)
+{
+    InitializeCriticalSectionEx(&m_critSec, 3000, 0);
+}
+
+OcvImageManipulations::~OcvImageManipulations()
+{
+    SafeRelease(&m_pInputType);
+    SafeRelease(&m_pOutputType);
+    SafeRelease(&m_pSample);
+    SafeRelease(&m_pAttributes);
+    DeleteCriticalSection(&m_critSec);
+}
+
+// Initialize the instance.
+STDMETHODIMP OcvImageManipulations::RuntimeClassInitialize()
+{
+    // Create the attribute store.
+    return MFCreateAttributes(&m_pAttributes, 3);
+}
+
+// IMediaExtension methods
+
+//-------------------------------------------------------------------
+// SetProperties
+// Sets the configuration of the effect
+//-------------------------------------------------------------------
+HRESULT OcvImageManipulations::SetProperties(ABI::Windows::Foundation::Collections::IPropertySet *pConfiguration)
+{
+    HRESULT hr = S_OK;
+
+    if (!pConfiguration)
+        return hr;
+
+    HSTRING key;
+    WindowsCreateString(L"{698649BE-8EAE-4551-A4CB-3EC98FBD3D86}", 38, &key);
+    Microsoft::WRL::ComPtr<ABI::Windows::Foundation::Collections::IMap<HSTRING, IInspectable *>> spSetting;
+    pConfiguration->QueryInterface(IID_PPV_ARGS(&spSetting));
+    boolean found;
+    spSetting->HasKey(key, &found);
+
+    if (found)
+    {
+        IInspectable* value;
+        spSetting->Lookup(key, &value);
+
+        Microsoft::WRL::ComPtr<ABI::Windows::Foundation::IReference<int>> ref;
+        hr = value->QueryInterface(IID_PPV_ARGS(&ref));
+        int effect = InvalidEffect;
+        hr = ref->get_Value(&effect);
+        if ((effect >= 0) && (effect < InvalidEffect))
+        {
+            m_TransformType = (ProcessingType)effect;
+        }
+    }
+
+    return hr;
+}
+
+// IMFTransform methods. Refer to the Media Foundation SDK documentation for details.
+
+//-------------------------------------------------------------------
+// GetStreamLimits
+// Returns the minimum and maximum number of streams.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetStreamLimits(
+    DWORD   *pdwInputMinimum,
+    DWORD   *pdwInputMaximum,
+    DWORD   *pdwOutputMinimum,
+    DWORD   *pdwOutputMaximum
+)
+{
+    if ((pdwInputMinimum == NULL) ||
+        (pdwInputMaximum == NULL) ||
+        (pdwOutputMinimum == NULL) ||
+        (pdwOutputMaximum == NULL))
+    {
+        return E_POINTER;
+    }
+
+    // This MFT has a fixed number of streams.
+    *pdwInputMinimum = 1;
+    *pdwInputMaximum = 1;
+    *pdwOutputMinimum = 1;
+    *pdwOutputMaximum = 1;
+    return S_OK;
+}
+
+
+//-------------------------------------------------------------------
+// GetStreamCount
+// Returns the actual number of streams.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetStreamCount(
+    DWORD   *pcInputStreams,
+    DWORD   *pcOutputStreams
+)
+{
+    if ((pcInputStreams == NULL) || (pcOutputStreams == NULL))
+
+    {
+        return E_POINTER;
+    }
+
+    // This MFT has a fixed number of streams.
+    *pcInputStreams = 1;
+    *pcOutputStreams = 1;
+    return S_OK;
+}
+
+
+
+//-------------------------------------------------------------------
+// GetStreamIDs
+// Returns stream IDs for the input and output streams.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetStreamIDs(
+    DWORD   dwInputIDArraySize,
+    DWORD   *pdwInputIDs,
+    DWORD   dwOutputIDArraySize,
+    DWORD   *pdwOutputIDs
+)
+{
+    // It is not required to implement this method if the MFT has a fixed number of
+    // streams AND the stream IDs are numbered sequentially from zero (that is, the
+    // stream IDs match the stream indexes).
+
+    // In that case, it is OK to return E_NOTIMPL.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// GetInputStreamInfo
+// Returns information about an input stream.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetInputStreamInfo(
+    DWORD                     dwInputStreamID,
+    MFT_INPUT_STREAM_INFO *   pStreamInfo
+)
+{
+    if (pStreamInfo == NULL)
+    {
+        return E_POINTER;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    // NOTE: This method should succeed even when there is no media type on the
+    //       stream. If there is no media type, we only need to fill in the dwFlags
+    //       member of MFT_INPUT_STREAM_INFO. The other members depend on having a
+    //       a valid media type.
+
+    pStreamInfo->hnsMaxLatency = 0;
+    pStreamInfo->dwFlags = MFT_INPUT_STREAM_WHOLE_SAMPLES | MFT_INPUT_STREAM_SINGLE_SAMPLE_PER_BUFFER;
+
+    if (m_pInputType == NULL)
+    {
+        pStreamInfo->cbSize = 0;
+    }
+    else
+    {
+        pStreamInfo->cbSize = m_cbImageSize;
+    }
+
+    pStreamInfo->cbMaxLookahead = 0;
+    pStreamInfo->cbAlignment = 0;
+
+    LeaveCriticalSection(&m_critSec);
+    return S_OK;
+}
+
+//-------------------------------------------------------------------
+// GetOutputStreamInfo
+// Returns information about an output stream.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetOutputStreamInfo(
+    DWORD                     dwOutputStreamID,
+    MFT_OUTPUT_STREAM_INFO *  pStreamInfo
+)
+{
+    if (pStreamInfo == NULL)
+    {
+        return E_POINTER;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidOutputStream(dwOutputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    // NOTE: This method should succeed even when there is no media type on the
+    //       stream. If there is no media type, we only need to fill in the dwFlags
+    //       member of MFT_OUTPUT_STREAM_INFO. The other members depend on having a
+    //       a valid media type.
+
+    pStreamInfo->dwFlags =
+        MFT_OUTPUT_STREAM_WHOLE_SAMPLES |
+        MFT_OUTPUT_STREAM_SINGLE_SAMPLE_PER_BUFFER |
+        MFT_OUTPUT_STREAM_FIXED_SAMPLE_SIZE ;
+
+    if (m_pOutputType == NULL)
+    {
+        pStreamInfo->cbSize = 0;
+    }
+    else
+    {
+        pStreamInfo->cbSize = m_cbImageSize;
+    }
+
+    pStreamInfo->cbAlignment = 0;
+
+    LeaveCriticalSection(&m_critSec);
+    return S_OK;
+}
+
+
+//-------------------------------------------------------------------
+// GetAttributes
+// Returns the attributes for the MFT.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetAttributes(IMFAttributes** ppAttributes)
+{
+    if (ppAttributes == NULL)
+    {
+        return E_POINTER;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    *ppAttributes = m_pAttributes;
+    (*ppAttributes)->AddRef();
+
+    LeaveCriticalSection(&m_critSec);
+    return S_OK;
+}
+
+
+//-------------------------------------------------------------------
+// GetInputStreamAttributes
+// Returns stream-level attributes for an input stream.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetInputStreamAttributes(
+    DWORD           dwInputStreamID,
+    IMFAttributes   **ppAttributes
+)
+{
+    // This MFT does not support any stream-level attributes, so the method is not implemented.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// GetOutputStreamAttributes
+// Returns stream-level attributes for an output stream.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetOutputStreamAttributes(
+    DWORD           dwOutputStreamID,
+    IMFAttributes   **ppAttributes
+)
+{
+    // This MFT does not support any stream-level attributes, so the method is not implemented.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// DeleteInputStream
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::DeleteInputStream(DWORD dwStreamID)
+{
+    // This MFT has a fixed number of input streams, so the method is not supported.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// AddInputStreams
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::AddInputStreams(
+    DWORD   cStreams,
+    DWORD   *adwStreamIDs
+)
+{
+    // This MFT has a fixed number of output streams, so the method is not supported.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// GetInputAvailableType
+// Returns a preferred input type.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetInputAvailableType(
+    DWORD           dwInputStreamID,
+    DWORD           dwTypeIndex, // 0-based
+    IMFMediaType    **ppType
+)
+{
+    if (ppType == NULL)
+    {
+        return E_INVALIDARG;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    HRESULT hr = S_OK;
+
+    // If the output type is set, return that type as our preferred input type.
+    if (m_pOutputType == NULL)
+    {
+        // The output type is not set. Create a partial media type.
+        hr = OnGetPartialType(dwTypeIndex, ppType);
+    }
+    else if (dwTypeIndex > 0)
+    {
+        hr = MF_E_NO_MORE_TYPES;
+    }
+    else
+    {
+        *ppType = m_pOutputType;
+        (*ppType)->AddRef();
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+
+//-------------------------------------------------------------------
+// GetOutputAvailableType
+// Returns a preferred output type.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetOutputAvailableType(
+    DWORD           dwOutputStreamID,
+    DWORD           dwTypeIndex, // 0-based
+    IMFMediaType    **ppType
+)
+{
+    if (ppType == NULL)
+    {
+        return E_INVALIDARG;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidOutputStream(dwOutputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    HRESULT hr = S_OK;
+
+    if (m_pInputType == NULL)
+    {
+        // The input type is not set. Create a partial media type.
+        hr = OnGetPartialType(dwTypeIndex, ppType);
+    }
+    else if (dwTypeIndex > 0)
+    {
+        hr = MF_E_NO_MORE_TYPES;
+    }
+    else
+    {
+        *ppType = m_pInputType;
+        (*ppType)->AddRef();
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// SetInputType
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::SetInputType(
+    DWORD           dwInputStreamID,
+    IMFMediaType    *pType, // Can be NULL to clear the input type.
+    DWORD           dwFlags
+)
+{
+    // Validate flags.
+    if (dwFlags & ~MFT_SET_TYPE_TEST_ONLY)
+    {
+        return E_INVALIDARG;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    HRESULT hr = S_OK;
+
+    // Does the caller want us to set the type, or just test it?
+    BOOL bReallySet = ((dwFlags & MFT_SET_TYPE_TEST_ONLY) == 0);
+
+    // If we have an input sample, the client cannot change the type now.
+    if (HasPendingOutput())
+    {
+        hr = MF_E_TRANSFORM_CANNOT_CHANGE_MEDIATYPE_WHILE_PROCESSING;
+        goto done;
+    }
+
+    // Validate the type, if non-NULL.
+    if (pType)
+    {
+        hr = OnCheckInputType(pType);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+    }
+
+    // The type is OK. Set the type, unless the caller was just testing.
+    if (bReallySet)
+    {
+        OnSetInputType(pType);
+
+        // When the type changes, end streaming.
+        hr = EndStreaming();
+    }
+
+done:
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+
+//-------------------------------------------------------------------
+// SetOutputType
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::SetOutputType(
+    DWORD           dwOutputStreamID,
+    IMFMediaType    *pType, // Can be NULL to clear the output type.
+    DWORD           dwFlags
+)
+{
+    // Validate flags.
+    if (dwFlags & ~MFT_SET_TYPE_TEST_ONLY)
+    {
+        return E_INVALIDARG;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidOutputStream(dwOutputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    HRESULT hr = S_OK;
+
+    // Does the caller want us to set the type, or just test it?
+    BOOL bReallySet = ((dwFlags & MFT_SET_TYPE_TEST_ONLY) == 0);
+
+    // If we have an input sample, the client cannot change the type now.
+    if (HasPendingOutput())
+    {
+        hr = MF_E_TRANSFORM_CANNOT_CHANGE_MEDIATYPE_WHILE_PROCESSING;
+        goto done;
+    }
+
+    // Validate the type, if non-NULL.
+    if (pType)
+    {
+        hr = OnCheckOutputType(pType);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+    }
+
+    // The type is OK. Set the type, unless the caller was just testing.
+    if (bReallySet)
+    {
+        OnSetOutputType(pType);
+
+        // When the type changes, end streaming.
+        hr = EndStreaming();
+    }
+
+done:
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// GetInputCurrentType
+// Returns the current input type.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetInputCurrentType(
+    DWORD           dwInputStreamID,
+    IMFMediaType    **ppType
+)
+{
+    if (ppType == NULL)
+    {
+        return E_POINTER;
+    }
+
+    HRESULT hr = S_OK;
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        hr = MF_E_INVALIDSTREAMNUMBER;
+    }
+    else if (!m_pInputType)
+    {
+        hr = MF_E_TRANSFORM_TYPE_NOT_SET;
+    }
+    else
+    {
+        *ppType = m_pInputType;
+        (*ppType)->AddRef();
+    }
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// GetOutputCurrentType
+// Returns the current output type.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetOutputCurrentType(
+    DWORD           dwOutputStreamID,
+    IMFMediaType    **ppType
+)
+{
+    if (ppType == NULL)
+    {
+        return E_POINTER;
+    }
+
+    HRESULT hr = S_OK;
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidOutputStream(dwOutputStreamID))
+    {
+        hr = MF_E_INVALIDSTREAMNUMBER;
+    }
+    else if (!m_pOutputType)
+    {
+        hr = MF_E_TRANSFORM_TYPE_NOT_SET;
+    }
+    else
+    {
+        *ppType = m_pOutputType;
+        (*ppType)->AddRef();
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// GetInputStatus
+// Query if the MFT is accepting more input.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetInputStatus(
+    DWORD           dwInputStreamID,
+    DWORD           *pdwFlags
+)
+{
+    if (pdwFlags == NULL)
+    {
+        return E_POINTER;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        LeaveCriticalSection(&m_critSec);
+        return MF_E_INVALIDSTREAMNUMBER;
+    }
+
+    // If an input sample is already queued, do not accept another sample until the
+    // client calls ProcessOutput or Flush.
+
+    // NOTE: It is possible for an MFT to accept more than one input sample. For
+    // example, this might be required in a video decoder if the frames do not
+    // arrive in temporal order. In the case, the decoder must hold a queue of
+    // samples. For the video effect, each sample is transformed independently, so
+    // there is no reason to queue multiple input samples.
+
+    if (m_pSample == NULL)
+    {
+        *pdwFlags = MFT_INPUT_STATUS_ACCEPT_DATA;
+    }
+    else
+    {
+        *pdwFlags = 0;
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return S_OK;
+}
+
+
+
+//-------------------------------------------------------------------
+// GetOutputStatus
+// Query if the MFT can produce output.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::GetOutputStatus(DWORD *pdwFlags)
+{
+    if (pdwFlags == NULL)
+    {
+        return E_POINTER;
+    }
+
+    EnterCriticalSection(&m_critSec);
+
+    // The MFT can produce an output sample if (and only if) there an input sample.
+    if (m_pSample != NULL)
+    {
+        *pdwFlags = MFT_OUTPUT_STATUS_SAMPLE_READY;
+    }
+    else
+    {
+        *pdwFlags = 0;
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return S_OK;
+}
+
+
+//-------------------------------------------------------------------
+// SetOutputBounds
+// Sets the range of time stamps that the MFT will output.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::SetOutputBounds(
+    LONGLONG        hnsLowerBound,
+    LONGLONG        hnsUpperBound
+)
+{
+    // Implementation of this method is optional.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// ProcessEvent
+// Sends an event to an input stream.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::ProcessEvent(
+    DWORD              dwInputStreamID,
+    IMFMediaEvent      *pEvent
+)
+{
+    // This MFT does not handle any stream events, so the method can
+    // return E_NOTIMPL. This tells the pipeline that it can stop
+    // sending any more events to this MFT.
+    return E_NOTIMPL;
+}
+
+
+//-------------------------------------------------------------------
+// ProcessMessage
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::ProcessMessage(
+    MFT_MESSAGE_TYPE    eMessage,
+    ULONG_PTR           ulParam
+)
+{
+    EnterCriticalSection(&m_critSec);
+
+    HRESULT hr = S_OK;
+
+    switch (eMessage)
+    {
+    case MFT_MESSAGE_COMMAND_FLUSH:
+        // Flush the MFT.
+        hr = OnFlush();
+        break;
+
+    case MFT_MESSAGE_COMMAND_DRAIN:
+        // Drain: Tells the MFT to reject further input until all pending samples are
+        // processed. That is our default behavior already, so there is nothing to do.
+        //
+        // For a decoder that accepts a queue of samples, the MFT might need to drain
+        // the queue in response to this command.
+    break;
+
+    case MFT_MESSAGE_SET_D3D_MANAGER:
+        // Sets a pointer to the IDirect3DDeviceManager9 interface.
+
+        // The pipeline should never send this message unless the MFT sets the MF_SA_D3D_AWARE
+        // attribute set to TRUE. Because this MFT does not set MF_SA_D3D_AWARE, it is an error
+        // to send the MFT_MESSAGE_SET_D3D_MANAGER message to the MFT. Return an error code in
+        // this case.
+
+        // NOTE: If this MFT were D3D-enabled, it would cache the IDirect3DDeviceManager9
+        // pointer for use during streaming.
+
+        hr = E_NOTIMPL;
+        break;
+
+    case MFT_MESSAGE_NOTIFY_BEGIN_STREAMING:
+        hr = BeginStreaming();
+        break;
+
+    case MFT_MESSAGE_NOTIFY_END_STREAMING:
+        hr = EndStreaming();
+        break;
+
+    // The next two messages do not require any action from this MFT.
+
+    case MFT_MESSAGE_NOTIFY_END_OF_STREAM:
+        break;
+
+    case MFT_MESSAGE_NOTIFY_START_OF_STREAM:
+        break;
+    }
+
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// ProcessInput
+// Process an input sample.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::ProcessInput(
+    DWORD               dwInputStreamID,
+    IMFSample           *pSample,
+    DWORD               dwFlags
+)
+{
+    // Check input parameters.
+    if (pSample == NULL)
+    {
+        return E_POINTER;
+    }
+
+    if (dwFlags != 0)
+    {
+        return E_INVALIDARG; // dwFlags is reserved and must be zero.
+    }
+
+    HRESULT hr = S_OK;
+
+    EnterCriticalSection(&m_critSec);
+
+    // Validate the input stream number.
+    if (!IsValidInputStream(dwInputStreamID))
+    {
+        hr = MF_E_INVALIDSTREAMNUMBER;
+        goto done;
+    }
+
+    // Check for valid media types.
+    // The client must set input and output types before calling ProcessInput.
+    if (!m_pInputType || !m_pOutputType)
+    {
+        hr = MF_E_NOTACCEPTING;
+        goto done;
+    }
+
+    // Check if an input sample is already queued.
+    if (m_pSample != NULL)
+    {
+        hr = MF_E_NOTACCEPTING;   // We already have an input sample.
+        goto done;
+    }
+
+    // Initialize streaming.
+    hr = BeginStreaming();
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    // Cache the sample. We do the actual work in ProcessOutput.
+    m_pSample = pSample;
+    pSample->AddRef();  // Hold a reference count on the sample.
+
+done:
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+
+//-------------------------------------------------------------------
+// ProcessOutput
+// Process an output sample.
+//-------------------------------------------------------------------
+
+HRESULT OcvImageManipulations::ProcessOutput(
+    DWORD                   dwFlags,
+    DWORD                   cOutputBufferCount,
+    MFT_OUTPUT_DATA_BUFFER  *pOutputSamples, // one per stream
+    DWORD                   *pdwStatus
+)
+{
+    // Check input parameters...
+
+    // This MFT does not accept any flags for the dwFlags parameter.
+
+    // The only defined flag is MFT_PROCESS_OUTPUT_DISCARD_WHEN_NO_BUFFER. This flag
+    // applies only when the MFT marks an output stream as lazy or optional. But this
+    // MFT has no lazy or optional streams, so the flag is not valid.
+
+    if (dwFlags != 0)
+    {
+        return E_INVALIDARG;
+    }
+
+    if (pOutputSamples == NULL || pdwStatus == NULL)
+    {
+        return E_POINTER;
+    }
+
+    // There must be exactly one output buffer.
+    if (cOutputBufferCount != 1)
+    {
+        return E_INVALIDARG;
+    }
+
+    // It must contain a sample.
+    if (pOutputSamples[0].pSample == NULL)
+    {
+        return E_INVALIDARG;
+    }
+
+    HRESULT hr = S_OK;
+
+    IMFMediaBuffer *pInput = NULL;
+    IMFMediaBuffer *pOutput = NULL;
+
+    EnterCriticalSection(&m_critSec);
+
+    // There must be an input sample available for processing.
+    if (m_pSample == NULL)
+    {
+        hr = MF_E_TRANSFORM_NEED_MORE_INPUT;
+        goto done;
+    }
+
+    // Initialize streaming.
+
+    hr = BeginStreaming();
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    // Get the input buffer.
+    hr = m_pSample->ConvertToContiguousBuffer(&pInput);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    // Get the output buffer.
+    hr = pOutputSamples[0].pSample->ConvertToContiguousBuffer(&pOutput);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    hr = OnProcessOutput(pInput, pOutput);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    // Set status flags.
+    pOutputSamples[0].dwStatus = 0;
+    *pdwStatus = 0;
+
+
+    // Copy the duration and time stamp from the input sample, if present.
+
+    LONGLONG hnsDuration = 0;
+    LONGLONG hnsTime = 0;
+
+    if (SUCCEEDED(m_pSample->GetSampleDuration(&hnsDuration)))
+    {
+        hr = pOutputSamples[0].pSample->SetSampleDuration(hnsDuration);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+    }
+
+    if (SUCCEEDED(m_pSample->GetSampleTime(&hnsTime)))
+    {
+        hr = pOutputSamples[0].pSample->SetSampleTime(hnsTime);
+    }
+
+done:
+    SafeRelease(&m_pSample);   // Release our input sample.
+    SafeRelease(&pInput);
+    SafeRelease(&pOutput);
+    LeaveCriticalSection(&m_critSec);
+    return hr;
+}
+
+// PRIVATE METHODS
+
+// All methods that follow are private to this MFT and are not part of the IMFTransform interface.
+
+// Create a partial media type from our list.
+//
+// dwTypeIndex: Index into the list of peferred media types.
+// ppmt:        Receives a pointer to the media type.
+
+HRESULT OcvImageManipulations::OnGetPartialType(DWORD dwTypeIndex, IMFMediaType **ppmt)
+{
+    if (dwTypeIndex >= ARRAYSIZE(g_MediaSubtypes))
+    {
+        return MF_E_NO_MORE_TYPES;
+    }
+
+    IMFMediaType *pmt = NULL;
+
+    HRESULT hr = MFCreateMediaType(&pmt);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    hr = pmt->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    hr = pmt->SetGUID(MF_MT_SUBTYPE, g_MediaSubtypes[dwTypeIndex]);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    *ppmt = pmt;
+    (*ppmt)->AddRef();
+
+done:
+    SafeRelease(&pmt);
+    return hr;
+}
+
+
+// Validate an input media type.
+
+HRESULT OcvImageManipulations::OnCheckInputType(IMFMediaType *pmt)
+{
+    assert(pmt != NULL);
+
+    HRESULT hr = S_OK;
+
+    // If the output type is set, see if they match.
+    if (m_pOutputType != NULL)
+    {
+        DWORD flags = 0;
+        hr = pmt->IsEqual(m_pOutputType, &flags);
+
+        // IsEqual can return S_FALSE. Treat this as failure.
+        if (hr != S_OK)
+        {
+            hr = MF_E_INVALIDMEDIATYPE;
+        }
+    }
+    else
+    {
+        // Output type is not set. Just check this type.
+        hr = OnCheckMediaType(pmt);
+    }
+    return hr;
+}
+
+
+// Validate an output media type.
+
+HRESULT OcvImageManipulations::OnCheckOutputType(IMFMediaType *pmt)
+{
+    assert(pmt != NULL);
+
+    HRESULT hr = S_OK;
+
+    // If the input type is set, see if they match.
+    if (m_pInputType != NULL)
+    {
+        DWORD flags = 0;
+        hr = pmt->IsEqual(m_pInputType, &flags);
+
+        // IsEqual can return S_FALSE. Treat this as failure.
+        if (hr != S_OK)
+        {
+            hr = MF_E_INVALIDMEDIATYPE;
+        }
+
+    }
+    else
+    {
+        // Input type is not set. Just check this type.
+        hr = OnCheckMediaType(pmt);
+    }
+    return hr;
+}
+
+
+// Validate a media type (input or output)
+
+HRESULT OcvImageManipulations::OnCheckMediaType(IMFMediaType *pmt)
+{
+    BOOL bFoundMatchingSubtype = FALSE;
+
+    // Major type must be video.
+    GUID major_type;
+    HRESULT hr = pmt->GetGUID(MF_MT_MAJOR_TYPE, &major_type);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    if (major_type != MFMediaType_Video)
+    {
+        hr = MF_E_INVALIDMEDIATYPE;
+        goto done;
+    }
+
+    // Subtype must be one of the subtypes in our global list.
+
+    // Get the subtype GUID.
+    GUID subtype;
+    hr = pmt->GetGUID(MF_MT_SUBTYPE, &subtype);
+    if (FAILED(hr))
+    {
+        goto done;
+    }
+
+    // Look for the subtype in our list of accepted types.
+    for (DWORD i = 0; i < ARRAYSIZE(g_MediaSubtypes); i++)
+    {
+        if (subtype == g_MediaSubtypes[i])
+        {
+            bFoundMatchingSubtype = TRUE;
+            break;
+        }
+    }
+
+    if (!bFoundMatchingSubtype)
+    {
+        hr = MF_E_INVALIDMEDIATYPE; // The MFT does not support this subtype.
+        goto done;
+    }
+
+    // Reject single-field media types.
+    UINT32 interlace = MFGetAttributeUINT32(pmt, MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
+    if (interlace == MFVideoInterlace_FieldSingleUpper  || interlace == MFVideoInterlace_FieldSingleLower)
+    {
+        hr = MF_E_INVALIDMEDIATYPE;
+    }
+
+done:
+    return hr;
+}
+
+
+// Set or clear the input media type.
+//
+// Prerequisite: The input type was already validated.
+
+void OcvImageManipulations::OnSetInputType(IMFMediaType *pmt)
+{
+    // if pmt is NULL, clear the type.
+    // if pmt is non-NULL, set the type.
+
+    SafeRelease(&m_pInputType);
+    m_pInputType = pmt;
+    if (m_pInputType)
+    {
+        m_pInputType->AddRef();
+    }
+
+    // Update the format information.
+    UpdateFormatInfo();
+}
+
+
+// Set or clears the output media type.
+//
+// Prerequisite: The output type was already validated.
+
+void OcvImageManipulations::OnSetOutputType(IMFMediaType *pmt)
+{
+    // If pmt is NULL, clear the type. Otherwise, set the type.
+
+    SafeRelease(&m_pOutputType);
+    m_pOutputType = pmt;
+    if (m_pOutputType)
+    {
+        m_pOutputType->AddRef();
+    }
+}
+
+
+// Initialize streaming parameters.
+//
+// This method is called if the client sends the MFT_MESSAGE_NOTIFY_BEGIN_STREAMING
+// message, or when the client processes a sample, whichever happens first.
+
+HRESULT OcvImageManipulations::BeginStreaming()
+{
+    HRESULT hr = S_OK;
+
+    if (!m_bStreamingInitialized)
+    {
+        m_bStreamingInitialized = true;
+        hr = S_OK;
+    }
+
+    return hr;
+}
+
+
+// End streaming.
+
+// This method is called if the client sends an MFT_MESSAGE_NOTIFY_END_STREAMING
+// message, or when the media type changes. In general, it should be called whenever
+// the streaming parameters need to be reset.
+
+HRESULT OcvImageManipulations::EndStreaming()
+{
+    m_bStreamingInitialized = false;
+    return S_OK;
+}
+
+
+
+// Generate output data.
+
+HRESULT OcvImageManipulations::OnProcessOutput(IMFMediaBuffer *pIn, IMFMediaBuffer *pOut)
+{
+    BYTE *pDest = NULL;         // Destination buffer.
+    LONG lDestStride = 0;       // Destination stride.
+
+    BYTE *pSrc = NULL;          // Source buffer.
+    LONG lSrcStride = 0;        // Source stride.
+
+    // Helper objects to lock the buffers.
+    VideoBufferLock inputLock(pIn);
+    VideoBufferLock outputLock(pOut);
+
+    // Stride if the buffer does not support IMF2DBuffer
+    LONG lDefaultStride = 0;
+
+    HRESULT hr = GetDefaultStride(m_pInputType, &lDefaultStride);
+    if (FAILED(hr))
+    {
+        return hr;
+    }
+
+    // Lock the input buffer.
+    hr = inputLock.LockBuffer(lDefaultStride, m_imageHeightInPixels, &pSrc, &lSrcStride);
+    if (FAILED(hr))
+    {
+        return hr;
+    }
+
+    // Lock the output buffer.
+    hr = outputLock.LockBuffer(lDefaultStride, m_imageHeightInPixels, &pDest, &lDestStride);
+    if (FAILED(hr))
+    {
+        return hr;
+    }
+
+    cv::Mat InputFrame(m_imageHeightInPixels + m_imageHeightInPixels/2, m_imageWidthInPixels, CV_8UC1, pSrc, lSrcStride);
+    cv::Mat InputGreyScale(InputFrame, cv::Range(0, m_imageHeightInPixels), cv::Range(0, m_imageWidthInPixels));
+    cv::Mat OutputFrame(m_imageHeightInPixels + m_imageHeightInPixels/2, m_imageWidthInPixels, CV_8UC1, pDest, lDestStride);
+
+    switch (m_TransformType)
+    {
+    case Preview:
+        {
+            InputFrame.copyTo(OutputFrame);
+        } break;
+    case GrayScale:
+        {
+            OutputFrame.setTo(cv::Scalar(128));
+            cv::Mat OutputGreyScale(OutputFrame, cv::Range(0, m_imageHeightInPixels), cv::Range(0, m_imageWidthInPixels));
+            InputGreyScale.copyTo(OutputGreyScale);
+        } break;
+    case Canny:
+        {
+            OutputFrame.setTo(cv::Scalar(128));
+            cv::Mat OutputGreyScale(OutputFrame, cv::Range(0, m_imageHeightInPixels), cv::Range(0, m_imageWidthInPixels));
+            cv::Canny(InputGreyScale, OutputGreyScale, 80, 90);
+
+        } break;
+    case Sobel:
+        {
+            OutputFrame.setTo(cv::Scalar(128));
+            cv::Mat OutputGreyScale(OutputFrame, cv::Range(0, m_imageHeightInPixels), cv::Range(0, m_imageWidthInPixels));
+            cv::Sobel(InputGreyScale, OutputGreyScale, CV_8U, 1, 1);
+        } break;
+    case Histogram:
+        {
+            const int mHistSizeNum = 25;
+            const int channels[3][1] = {{0}, {1}, {2}};
+            const int mHistSize[] = {25};
+            const float baseRabge[] = {0.f,256.f};
+            const float* ranges[] = {baseRabge};
+
+            const cv::Scalar mColorsY[] = { cv::Scalar(76), cv::Scalar(149), cv::Scalar(29) };
+            const cv::Scalar mColorsUV[] = { cv::Scalar(84, 255), cv::Scalar(43, 21), cv::Scalar(255, 107) };
+
+            cv::Mat OutputY(m_imageHeightInPixels, m_imageWidthInPixels, CV_8UC1, pDest, lDestStride);
+            cv::Mat OutputUV(m_imageHeightInPixels/2, m_imageWidthInPixels/2,
+                             CV_8UC2, pDest+m_imageHeightInPixels*lDestStride, lDestStride);
+            cv::Mat BgrFrame;
+
+            InputFrame.copyTo(OutputFrame);
+
+            cv::cvtColor(InputFrame, BgrFrame, cv::COLOR_YUV420sp2BGR);
+            int thikness = (int) (BgrFrame.cols / (mHistSizeNum + 10) / 5);
+            if(thikness > 5) thikness = 5;
+            int offset = (int) ((BgrFrame.cols - (5*mHistSizeNum + 4*10)*thikness)/2);
+
+            // RGB
+            for (int c=0; c<3; c++)
+            {
+                cv::Mat hist;
+                cv::calcHist(&BgrFrame, 1, channels[c], cv::Mat(), hist, 1, mHistSize, ranges);
+                cv::normalize(hist, hist, BgrFrame.rows/2, 0, cv::NORM_INF);
+                for(int h=0; h<mHistSizeNum; h++) {
+                    cv::Point mP1, mP2;
+                    // Draw on Y plane
+                    mP1.x = mP2.x = offset + (c * (mHistSizeNum + 10) + h) * thikness;
+                    mP1.y = BgrFrame.rows-1;
+                    mP2.y = mP1.y - 2 - (int)hist.at<float>(h);
+                    cv::line(OutputY, mP1, mP2, mColorsY[c], thikness);
+
+                    // Draw on UV planes
+                    mP1.x /= 2;
+                    mP1.y /= 2;
+                    mP2.x /= 2;
+                    mP2.y /= 2;
+                    cv::line(OutputUV, mP1, mP2, mColorsUV[c], thikness/2);
+                }
+            }
+        } break;
+    default:
+        break;
+    }
+
+    // Set the data size on the output buffer.
+    hr = pOut->SetCurrentLength(m_cbImageSize);
+
+    return hr;
+}
+
+
+// Flush the MFT.
+
+HRESULT OcvImageManipulations::OnFlush()
+{
+    // For this MFT, flushing just means releasing the input sample.
+    SafeRelease(&m_pSample);
+    return S_OK;
+}
+
+
+// Update the format information. This method is called whenever the
+// input type is set.
+
+HRESULT OcvImageManipulations::UpdateFormatInfo()
+{
+    HRESULT hr = S_OK;
+
+    GUID subtype = GUID_NULL;
+
+    m_imageWidthInPixels = 0;
+    m_imageHeightInPixels = 0;
+    m_cbImageSize = 0;
+
+    if (m_pInputType != NULL)
+    {
+        hr = m_pInputType->GetGUID(MF_MT_SUBTYPE, &subtype);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+        if (subtype != MFVideoFormat_NV12)
+        {
+            hr = E_UNEXPECTED;
+            goto done;
+        }
+
+        hr = MFGetAttributeSize(m_pInputType, MF_MT_FRAME_SIZE, &m_imageWidthInPixels, &m_imageHeightInPixels);
+        if (FAILED(hr))
+        {
+            goto done;
+        }
+
+        // Calculate the image size for YUV NV12 image(not including padding)
+        m_cbImageSize = (m_imageHeightInPixels + m_imageHeightInPixels/2)*m_imageWidthInPixels;
+    }
+
+done:
+    return hr;
+}
+
+
+// Get the default stride for a video format.
+HRESULT GetDefaultStride(IMFMediaType *pType, LONG *plStride)
+{
+    LONG lStride = 0;
+
+    // Try to get the default stride from the media type.
+    HRESULT hr = pType->GetUINT32(MF_MT_DEFAULT_STRIDE, (UINT32*)&lStride);
+    if (FAILED(hr))
+    {
+        // Attribute not set. Try to calculate the default stride.
+        GUID subtype = GUID_NULL;
+
+        UINT32 width = 0;
+        UINT32 height = 0;
+
+        // Get the subtype and the image size.
+        hr = pType->GetGUID(MF_MT_SUBTYPE, &subtype);
+        if (SUCCEEDED(hr))
+        {
+            hr = MFGetAttributeSize(pType, MF_MT_FRAME_SIZE, &width, &height);
+        }
+        if (SUCCEEDED(hr))
+        {
+            if (subtype == MFVideoFormat_NV12)
+            {
+                lStride = width;
+            }
+            else if (subtype == MFVideoFormat_YUY2 || subtype == MFVideoFormat_UYVY)
+            {
+                lStride = ((width * 2) + 3) & ~3;
+            }
+            else
+            {
+                hr = E_INVALIDARG;
+            }
+        }
+
+        // Set the attribute for later reference.
+        if (SUCCEEDED(hr))
+        {
+            (void)pType->SetUINT32(MF_MT_DEFAULT_STRIDE, UINT32(lStride));
+        }
+    }
+    if (SUCCEEDED(hr))
+    {
+        *plStride = lStride;
+    }
+    return hr;
+}
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.def b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.def
new file mode 100644
index 0000000000..453349adf5
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.def
@@ -0,0 +1,4 @@
+EXPORTS
+    DllCanUnloadNow                     PRIVATE
+    DllGetActivationFactory             PRIVATE
+    DllGetClassObject                   PRIVATE
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.h b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.h
new file mode 100644
index 0000000000..9162593fc0
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.h
@@ -0,0 +1,247 @@
+// Defines the transform class.
+//
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved.
+
+#ifndef GRAYSCALE_H
+#define GRAYSCALE_H
+
+#include <new>
+#include <mfapi.h>
+#include <mftransform.h>
+#include <mfidl.h>
+#include <mferror.h>
+#include <strsafe.h>
+#include <assert.h>
+
+#include <wrl\implements.h>
+#include <wrl\module.h>
+#include <windows.media.h>
+
+#include "OcvImageManipulations.h"
+
+// CLSID of the MFT.
+DEFINE_GUID(CLSID_GrayscaleMFT,
+0x2f3dbc05, 0xc011, 0x4a8f, 0xb2, 0x64, 0xe4, 0x2e, 0x35, 0xc6, 0x7b, 0xf4);
+
+//
+// * IMPORTANT: If you implement your own MFT, create a new GUID for the CLSID. *
+//
+
+
+// Configuration attributes
+// {698649BE-8EAE-4551-A4CB-3EC98FBD3D86}
+DEFINE_GUID(OCV_IMAGE_EFFECT,
+0x698649be, 0x8eae, 0x4551, 0xa4, 0xcb, 0x3e, 0xc9, 0x8f, 0xbd, 0x3d, 0x86);
+
+
+enum ProcessingType
+{
+    Preview,
+    GrayScale,
+    Canny,
+    Sobel,
+    Histogram,
+    InvalidEffect
+};
+
+template <class T> void SafeRelease(T **ppT)
+{
+    if (*ppT)
+    {
+        (*ppT)->Release();
+        *ppT = NULL;
+    }
+}
+
+// OcvImageManipulations class:
+// Implements a grayscale video effect.
+
+class OcvImageManipulations
+    : public Microsoft::WRL::RuntimeClass<
+           Microsoft::WRL::RuntimeClassFlags< Microsoft::WRL::RuntimeClassType::WinRtClassicComMix >,
+           ABI::Windows::Media::IMediaExtension,
+           IMFTransform >
+{
+    InspectableClass(RuntimeClass_OcvTransform_OcvImageManipulations, BaseTrust)
+
+public:
+    OcvImageManipulations();
+
+    ~OcvImageManipulations();
+
+    STDMETHOD(RuntimeClassInitialize)();
+
+    // IMediaExtension
+    STDMETHODIMP SetProperties(ABI::Windows::Foundation::Collections::IPropertySet *pConfiguration);
+
+    // IMFTransform
+    STDMETHODIMP GetStreamLimits(
+        DWORD   *pdwInputMinimum,
+        DWORD   *pdwInputMaximum,
+        DWORD   *pdwOutputMinimum,
+        DWORD   *pdwOutputMaximum
+    );
+
+    STDMETHODIMP GetStreamCount(
+        DWORD   *pcInputStreams,
+        DWORD   *pcOutputStreams
+    );
+
+    STDMETHODIMP GetStreamIDs(
+        DWORD   dwInputIDArraySize,
+        DWORD   *pdwInputIDs,
+        DWORD   dwOutputIDArraySize,
+        DWORD   *pdwOutputIDs
+    );
+
+    STDMETHODIMP GetInputStreamInfo(
+        DWORD                     dwInputStreamID,
+        MFT_INPUT_STREAM_INFO *   pStreamInfo
+    );
+
+    STDMETHODIMP GetOutputStreamInfo(
+        DWORD                     dwOutputStreamID,
+        MFT_OUTPUT_STREAM_INFO *  pStreamInfo
+    );
+
+    STDMETHODIMP GetAttributes(IMFAttributes** pAttributes);
+
+    STDMETHODIMP GetInputStreamAttributes(
+        DWORD           dwInputStreamID,
+        IMFAttributes   **ppAttributes
+    );
+
+    STDMETHODIMP GetOutputStreamAttributes(
+        DWORD           dwOutputStreamID,
+        IMFAttributes   **ppAttributes
+    );
+
+    STDMETHODIMP DeleteInputStream(DWORD dwStreamID);
+
+    STDMETHODIMP AddInputStreams(
+        DWORD   cStreams,
+        DWORD   *adwStreamIDs
+    );
+
+    STDMETHODIMP GetInputAvailableType(
+        DWORD           dwInputStreamID,
+        DWORD           dwTypeIndex, // 0-based
+        IMFMediaType    **ppType
+    );
+
+    STDMETHODIMP GetOutputAvailableType(
+        DWORD           dwOutputStreamID,
+        DWORD           dwTypeIndex, // 0-based
+        IMFMediaType    **ppType
+    );
+
+    STDMETHODIMP SetInputType(
+        DWORD           dwInputStreamID,
+        IMFMediaType    *pType,
+        DWORD           dwFlags
+    );
+
+    STDMETHODIMP SetOutputType(
+        DWORD           dwOutputStreamID,
+        IMFMediaType    *pType,
+        DWORD           dwFlags
+    );
+
+    STDMETHODIMP GetInputCurrentType(
+        DWORD           dwInputStreamID,
+        IMFMediaType    **ppType
+    );
+
+    STDMETHODIMP GetOutputCurrentType(
+        DWORD           dwOutputStreamID,
+        IMFMediaType    **ppType
+    );
+
+    STDMETHODIMP GetInputStatus(
+        DWORD           dwInputStreamID,
+        DWORD           *pdwFlags
+    );
+
+    STDMETHODIMP GetOutputStatus(DWORD *pdwFlags);
+
+    STDMETHODIMP SetOutputBounds(
+        LONGLONG        hnsLowerBound,
+        LONGLONG        hnsUpperBound
+    );
+
+    STDMETHODIMP ProcessEvent(
+        DWORD              dwInputStreamID,
+        IMFMediaEvent      *pEvent
+    );
+
+    STDMETHODIMP ProcessMessage(
+        MFT_MESSAGE_TYPE    eMessage,
+        ULONG_PTR           ulParam
+    );
+
+    STDMETHODIMP ProcessInput(
+        DWORD               dwInputStreamID,
+        IMFSample           *pSample,
+        DWORD               dwFlags
+    );
+
+    STDMETHODIMP ProcessOutput(
+        DWORD                   dwFlags,
+        DWORD                   cOutputBufferCount,
+        MFT_OUTPUT_DATA_BUFFER  *pOutputSamples, // one per stream
+        DWORD                   *pdwStatus
+    );
+
+
+private:
+    // HasPendingOutput: Returns TRUE if the MFT is holding an input sample.
+    BOOL HasPendingOutput() const { return m_pSample != NULL; }
+
+    // IsValidInputStream: Returns TRUE if dwInputStreamID is a valid input stream identifier.
+    BOOL IsValidInputStream(DWORD dwInputStreamID) const
+    {
+        return dwInputStreamID == 0;
+    }
+
+    // IsValidOutputStream: Returns TRUE if dwOutputStreamID is a valid output stream identifier.
+    BOOL IsValidOutputStream(DWORD dwOutputStreamID) const
+    {
+        return dwOutputStreamID == 0;
+    }
+
+    HRESULT OnGetPartialType(DWORD dwTypeIndex, IMFMediaType **ppmt);
+    HRESULT OnCheckInputType(IMFMediaType *pmt);
+    HRESULT OnCheckOutputType(IMFMediaType *pmt);
+    HRESULT OnCheckMediaType(IMFMediaType *pmt);
+    void    OnSetInputType(IMFMediaType *pmt);
+    void    OnSetOutputType(IMFMediaType *pmt);
+    HRESULT BeginStreaming();
+    HRESULT EndStreaming();
+    HRESULT OnProcessOutput(IMFMediaBuffer *pIn, IMFMediaBuffer *pOut);
+    HRESULT OnFlush();
+    HRESULT UpdateFormatInfo();
+
+    CRITICAL_SECTION            m_critSec;
+
+    // Transformation parameters
+    ProcessingType              m_TransformType;
+
+    // Streaming
+    bool                        m_bStreamingInitialized;
+    IMFSample                   *m_pSample;                 // Input sample.
+    IMFMediaType                *m_pInputType;              // Input media type.
+    IMFMediaType                *m_pOutputType;             // Output media type.
+
+    // Fomat information
+    UINT32                      m_imageWidthInPixels;
+    UINT32                      m_imageHeightInPixels;
+    DWORD                       m_cbImageSize;              // Image size, in bytes.
+
+    IMFAttributes               *m_pAttributes;
+};
+#endif
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.vcxproj b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.vcxproj
new file mode 100644
index 0000000000..54c0ce40ab
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/OcvTransform.vcxproj
@@ -0,0 +1,319 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|ARM">
+      <Configuration>Debug</Configuration>
+      <Platform>ARM</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|Win32">
+      <Configuration>Debug</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|ARM">
+      <Configuration>Release</Configuration>
+      <Platform>ARM</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|Win32">
+      <Configuration>Release</Configuration>
+      <Platform>Win32</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <VCTargetsPath Condition="'$(VCTargetsPath11)' != '' and '$(VSVersion)' == '' and '$(VisualStudioVersion)' == ''">$(VCTargetsPath11)</VCTargetsPath>
+    <ProjectGuid>{BA69218F-DA5C-4D14-A78D-21A9E4DEC669}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <RootNamespace>OcvTransform</RootNamespace>
+    <ProjectName>OcvTransform</ProjectName>
+    <MinimumVisualStudioVersion>11.0</MinimumVisualStudioVersion>
+    <AppContainerApplication>true</AppContainerApplication>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>DynamicLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <PlatformToolset>v110</PlatformToolset>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <PropertyGroup>
+    <DefaultLanguage>en-US</DefaultLanguage>
+  </PropertyGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <GenerateManifest>false</GenerateManifest>
+    <OutDir>$(Configuration)\$(MSBuildProjectName)\</OutDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">
+    <GenerateManifest>false</GenerateManifest>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <GenerateManifest>false</GenerateManifest>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <GenerateManifest>false</GenerateManifest>
+    <OutDir>$(Configuration)\$(MSBuildProjectName)\</OutDir>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">
+    <GenerateManifest>false</GenerateManifest>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <GenerateManifest>false</GenerateManifest>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;_DEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib;opencv_core245.lib;opencv_imgproc245.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(ProjectDir)$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(ProjectDir)$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;_DEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;_DEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;NDEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(ProjectDir)$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(ProjectDir)$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;NDEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <PreprocessorDefinitions>_WINRT_DLL;NDEBUG;_WINDOWS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <PrecompiledHeaderFile>
+      </PrecompiledHeaderFile>
+      <PrecompiledHeaderOutputFile>
+      </PrecompiledHeaderOutputFile>
+      <AdditionalUsingDirectories>$(WindowsSDK_WindowsMetadata);$(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+      <CompileAsWinRT>false</CompileAsWinRT>
+      <AdditionalIncludeDirectories>$(OPENCV_DIR)\include;$(ProjectDir);$(IntermediateOutputPath);%(AdditionalIncludeDirectories);$(ProjectDir)\..\Common</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Console</SubSystem>
+      <AdditionalDependencies>runtimeobject.lib;%(AdditionalDependencies);mf.lib;mfuuid.lib;mfplat.lib</AdditionalDependencies>
+      <IgnoreAllDefaultLibraries>false</IgnoreAllDefaultLibraries>
+      <ModuleDefinitionFile>OcvTransform.def</ModuleDefinitionFile>
+      <AdditionalLibraryDirectories>$(OPENCV_DIR)\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+    </Link>
+    <CustomBuildStep>
+      <Command>mdmerge -metadata_dir "$(WindowsSDK_MetadataPath)" -o "$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)" -i "$(MSBuildProjectDirectory)" -v -partial</Command>
+      <Outputs>$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd</Outputs>
+    </CustomBuildStep>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <ClInclude Include="OcvTransform.h" />
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="dllmain.cpp" />
+    <ClCompile Include="OcvTransform.cpp" />
+  </ItemGroup>
+  <ItemGroup>
+    <None Include="OcvTransform.def" />
+  </ItemGroup>
+  <ItemGroup>
+    <Midl Include="OcvImageManipulations.idl">
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
+      </MetadataFileName>
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">
+      </MetadataFileName>
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">
+      </MetadataFileName>
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">
+      </MetadataFileName>
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+      </MetadataFileName>
+      <MetadataFileName Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+      </MetadataFileName>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <AdditionalMetadataDirectories Condition="'$(Configuration)|$(Platform)'=='Release|x64'">$(WindowsSDK_MetadataPath)</AdditionalMetadataDirectories>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</EnableWindowsRuntime>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</EnableWindowsRuntime>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</EnableWindowsRuntime>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</EnableWindowsRuntime>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</EnableWindowsRuntime>
+      <EnableWindowsRuntime Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</EnableWindowsRuntime>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">%(Filename).h</HeaderFileName>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">%(Filename).h</HeaderFileName>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">%(Filename).h</HeaderFileName>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">%(Filename).h</HeaderFileName>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">%(Filename).h</HeaderFileName>
+      <HeaderFileName Condition="'$(Configuration)|$(Platform)'=='Release|x64'">%(Filename).h</HeaderFileName>
+    </Midl>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ItemGroup>
+    <_MdMergeOutput Condition="'$(Platform)' == 'Win32'" Include="$(ProjectDir)$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd" />
+    <_MdMergeOutput Condition="'$(Platform)' != 'Win32'" Include="$(SolutionDir)$(Platform)\$(Configuration)\$(MSBuildProjectName)\$(ProjectName).winmd" />
+  </ItemGroup>
+  <Target Name="CopyWinmdArtifactsOutputGroup" Returns="@(CopyWinmdArtifactsOutputGroupOutputs)">
+    <ItemGroup>
+      <CopyWinmdArtifactsOutputGroupOutputs Include="@(_MdMergeOutput)">
+        <TargetPath>$(ProjectName).winmd</TargetPath>
+        <Implementation>$(TargetName)$(TargetExt)</Implementation>
+      </CopyWinmdArtifactsOutputGroupOutputs>
+    </ItemGroup>
+  </Target>
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>
diff --git a/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/dllmain.cpp b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/dllmain.cpp
new file mode 100644
index 0000000000..701e7d416a
--- /dev/null
+++ b/samples/winrt/ImageManipulations/MediaExtensions/OcvTransform/dllmain.cpp
@@ -0,0 +1,58 @@
+//////////////////////////////////////////////////////////////////////////
+//
+// dllmain.cpp
+//
+// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
+// PARTICULAR PURPOSE.
+//
+// Copyright (c) Microsoft Corporation. All rights reserved.
+//
+//////////////////////////////////////////////////////////////////////////
+
+#include <initguid.h>
+#include "OcvTransform.h"
+
+using namespace Microsoft::WRL;
+
+namespace Microsoft { namespace Samples {
+    ActivatableClass(OcvImageManipulations);
+}}
+
+BOOL WINAPI DllMain( _In_ HINSTANCE hInstance, _In_ DWORD dwReason, _In_opt_ LPVOID lpReserved )
+{
+    if( DLL_PROCESS_ATTACH == dwReason )
+    {
+        //
+        //  Don't need per-thread callbacks
+        //
+        DisableThreadLibraryCalls( hInstance );
+
+        Module<InProc>::GetModule().Create();
+    }
+    else if( DLL_PROCESS_DETACH == dwReason )
+    {
+        Module<InProc>::GetModule().Terminate();
+    }
+
+    return TRUE;
+}
+
+HRESULT WINAPI DllGetActivationFactory( _In_ HSTRING activatibleClassId, _Outptr_ IActivationFactory** factory )
+{
+    auto &module = Microsoft::WRL::Module< Microsoft::WRL::InProc >::GetModule();
+    return module.GetActivationFactory( activatibleClassId, factory );
+}
+
+HRESULT WINAPI DllCanUnloadNow()
+{
+    auto &module = Microsoft::WRL::Module<Microsoft::WRL::InProc>::GetModule();
+    return (module.Terminate()) ? S_OK : S_FALSE;
+}
+
+STDAPI DllGetClassObject( _In_ REFCLSID rclsid, _In_ REFIID riid, _Outptr_ LPVOID FAR* ppv )
+{
+    auto &module = Microsoft::WRL::Module<Microsoft::WRL::InProc>::GetModule();
+    return module.GetClassObject( rclsid, riid, ppv );
+}
diff --git a/samples/winrt/ImageManipulations/Package.appxmanifest b/samples/winrt/ImageManipulations/Package.appxmanifest
new file mode 100644
index 0000000000..b66e01465e
--- /dev/null
+++ b/samples/winrt/ImageManipulations/Package.appxmanifest
@@ -0,0 +1,36 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Package xmlns="http://schemas.microsoft.com/appx/2010/manifest">
+  <Identity Name="Microsoft.SDKSamples.MediaCapture.CPP" Publisher="CN=Microsoft Corporation, O=Microsoft Corporation, L=Redmond, S=Washington, C=US" Version="1.0.0.0" />
+  <Properties>
+    <DisplayName>MediaCapture CPP sample</DisplayName>
+    <PublisherDisplayName>Microsoft Corporation</PublisherDisplayName>
+    <Logo>Assets\storeLogo-sdk.png</Logo>
+  </Properties>
+  <Prerequisites>
+    <OSMinVersion>6.2.1</OSMinVersion>
+    <OSMaxVersionTested>6.2.1</OSMaxVersionTested>
+  </Prerequisites>
+  <Resources>
+    <Resource Language="x-generate" />
+  </Resources>
+  <Applications>
+    <Application Id="MediaCapture.App" Executable="$targetnametoken$.exe" EntryPoint="MediaCapture.App">
+      <VisualElements DisplayName="OCV Image Manipulations" Logo="assets\opencv-logo-150.png" SmallLogo="assets\opencv-logo-30.png" Description="OpenCV Image Manipulations sample" ForegroundText="light" BackgroundColor="#00b2f0">
+        <DefaultTile ShortName="Ocv ImageManipulations" ShowName="allLogos" />
+        <SplashScreen Image="Assets\splash-sdk.png" BackgroundColor="#00b2f0" />
+      </VisualElements>
+    </Application>
+  </Applications>
+  <Capabilities>
+    <DeviceCapability Name="webcam" />
+    <DeviceCapability Name="microphone" />
+  </Capabilities>
+  <Extensions>
+    <Extension Category="windows.activatableClass.inProcessServer">
+      <InProcessServer>
+        <Path>OcvTransform.dll</Path>
+        <ActivatableClass ActivatableClassId="OcvTransform.OcvImageManipulations" ThreadingModel="both" />
+      </InProcessServer>
+    </Extension>
+  </Extensions>
+</Package>
diff --git a/samples/winrt/ImageManipulations/assets/opencv-logo-150.png b/samples/winrt/ImageManipulations/assets/opencv-logo-150.png
new file mode 100644
index 0000000000..ea685d651a
Binary files /dev/null and b/samples/winrt/ImageManipulations/assets/opencv-logo-150.png differ
diff --git a/samples/winrt/ImageManipulations/assets/opencv-logo-30.png b/samples/winrt/ImageManipulations/assets/opencv-logo-30.png
new file mode 100644
index 0000000000..efaf5468a1
Binary files /dev/null and b/samples/winrt/ImageManipulations/assets/opencv-logo-30.png differ
diff --git a/samples/winrt/ImageManipulations/assets/splash-sdk.png b/samples/winrt/ImageManipulations/assets/splash-sdk.png
new file mode 100644
index 0000000000..901c3b085a
Binary files /dev/null and b/samples/winrt/ImageManipulations/assets/splash-sdk.png differ
diff --git a/samples/winrt/ImageManipulations/assets/windows-sdk.png b/samples/winrt/ImageManipulations/assets/windows-sdk.png
new file mode 100644
index 0000000000..7f753f7b62
Binary files /dev/null and b/samples/winrt/ImageManipulations/assets/windows-sdk.png differ
diff --git a/samples/winrt/ImageManipulations/common/LayoutAwarePage.cpp b/samples/winrt/ImageManipulations/common/LayoutAwarePage.cpp
new file mode 100644
index 0000000000..07092bb745
--- /dev/null
+++ b/samples/winrt/ImageManipulations/common/LayoutAwarePage.cpp
@@ -0,0 +1,452 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+#include "pch.h"
+#include "LayoutAwarePage.h"
+#include "SuspensionManager.h"
+
+using namespace SDKSample::Common;
+
+using namespace Platform;
+using namespace Platform::Collections;
+using namespace Windows::Foundation;
+using namespace Windows::Foundation::Collections;
+using namespace Windows::System;
+using namespace Windows::UI::Core;
+using namespace Windows::UI::ViewManagement;
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Controls;
+using namespace Windows::UI::Xaml::Interop;
+using namespace Windows::UI::Xaml::Navigation;
+
+/// <summary>
+/// Initializes a new instance of the <see cref="LayoutAwarePage"/> class.
+/// </summary>
+LayoutAwarePage::LayoutAwarePage()
+{
+    if (Windows::ApplicationModel::DesignMode::DesignModeEnabled)
+    {
+        return;
+    }
+
+    // Create an empty default view model
+    DefaultViewModel = ref new Map<String^, Object^>(std::less<String^>());
+
+    // When this page is part of the visual tree make two changes:
+    // 1) Map application view state to visual state for the page
+    // 2) Handle keyboard and mouse navigation requests
+    Loaded += ref new RoutedEventHandler(this, &LayoutAwarePage::OnLoaded);
+
+    // Undo the same changes when the page is no longer visible
+    Unloaded += ref new RoutedEventHandler(this, &LayoutAwarePage::OnUnloaded);
+}
+
+static DependencyProperty^ _defaultViewModelProperty =
+    DependencyProperty::Register("DefaultViewModel",
+    TypeName(IObservableMap<String^, Object^>::typeid), TypeName(LayoutAwarePage::typeid), nullptr);
+
+/// <summary>
+/// Identifies the <see cref="DefaultViewModel"/> dependency property.
+/// </summary>
+DependencyProperty^ LayoutAwarePage::DefaultViewModelProperty::get()
+{
+    return _defaultViewModelProperty;
+}
+
+/// <summary>
+/// Gets an implementation of <see cref="IObservableMap&lt;String, Object&gt;"/> designed to be
+/// used as a trivial view model.
+/// </summary>
+IObservableMap<String^, Object^>^ LayoutAwarePage::DefaultViewModel::get()
+{
+    return safe_cast<IObservableMap<String^, Object^>^>(GetValue(DefaultViewModelProperty));
+}
+
+/// <summary>
+/// Sets an implementation of <see cref="IObservableMap&lt;String, Object&gt;"/> designed to be
+/// used as a trivial view model.
+/// </summary>
+void LayoutAwarePage::DefaultViewModel::set(IObservableMap<String^, Object^>^ value)
+{
+    SetValue(DefaultViewModelProperty, value);
+}
+
+/// <summary>
+/// Invoked when the page is part of the visual tree
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="e">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::OnLoaded(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e)
+{
+    this->StartLayoutUpdates(sender, e);
+
+    // Keyboard and mouse navigation only apply when occupying the entire window
+    if (this->ActualHeight == Window::Current->Bounds.Height &&
+        this->ActualWidth == Window::Current->Bounds.Width)
+    {
+        // Listen to the window directly so focus isn't required
+        _acceleratorKeyEventToken = Window::Current->CoreWindow->Dispatcher->AcceleratorKeyActivated +=
+            ref new TypedEventHandler<CoreDispatcher^, AcceleratorKeyEventArgs^>(this,
+            &LayoutAwarePage::CoreDispatcher_AcceleratorKeyActivated);
+        _pointerPressedEventToken = Window::Current->CoreWindow->PointerPressed +=
+            ref new TypedEventHandler<CoreWindow^, PointerEventArgs^>(this,
+            &LayoutAwarePage::CoreWindow_PointerPressed);
+        _navigationShortcutsRegistered = true;
+    }
+}
+
+/// <summary>
+/// Invoked when the page is removed from visual tree
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="e">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::OnUnloaded(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e)
+{
+    if (_navigationShortcutsRegistered)
+    {
+        Window::Current->CoreWindow->Dispatcher->AcceleratorKeyActivated -= _acceleratorKeyEventToken;
+        Window::Current->CoreWindow->PointerPressed -= _pointerPressedEventToken;
+        _navigationShortcutsRegistered = false;
+    }
+    StopLayoutUpdates(sender, e);
+}
+
+#pragma region Navigation support
+
+/// <summary>
+/// Invoked as an event handler to navigate backward in the page's associated <see cref="Frame"/>
+/// until it reaches the top of the navigation stack.
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="e">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::GoHome(Object^ sender, RoutedEventArgs^ e)
+{
+    (void) sender;	// Unused parameter
+    (void) e;	// Unused parameter
+
+    // Use the navigation frame to return to the topmost page
+    if (Frame != nullptr)
+    {
+        while (Frame->CanGoBack)
+        {
+            Frame->GoBack();
+        }
+    }
+}
+
+/// <summary>
+/// Invoked as an event handler to navigate backward in the navigation stack
+/// associated with this page's <see cref="Frame"/>.
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="e">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::GoBack(Object^ sender, RoutedEventArgs^ e)
+{
+    (void) sender;	// Unused parameter
+    (void) e;	// Unused parameter
+
+    // Use the navigation frame to return to the previous page
+    if (Frame != nullptr && Frame->CanGoBack)
+    {
+        Frame->GoBack();
+    }
+}
+
+/// <summary>
+/// Invoked as an event handler to navigate forward in the navigation stack
+/// associated with this page's <see cref="Frame"/>.
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="e">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::GoForward(Object^ sender, RoutedEventArgs^ e)
+{
+    (void) sender;	// Unused parameter
+    (void) e;	// Unused parameter
+
+    // Use the navigation frame to advance to the next page
+    if (Frame != nullptr && Frame->CanGoForward)
+    {
+        Frame->GoForward();
+    }
+}
+
+/// <summary>
+/// Invoked on every keystroke, including system keys such as Alt key combinations, when
+/// this page is active and occupies the entire window.  Used to detect keyboard navigation
+/// between pages even when the page itself doesn't have focus.
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="args">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::CoreDispatcher_AcceleratorKeyActivated(CoreDispatcher^ sender, AcceleratorKeyEventArgs^ args)
+{
+    auto virtualKey = args->VirtualKey;
+
+    // Only investigate further when Left, Right, or the dedicated Previous or Next keys
+    // are pressed
+    if ((args->EventType == CoreAcceleratorKeyEventType::SystemKeyDown ||
+        args->EventType == CoreAcceleratorKeyEventType::KeyDown) &&
+        (virtualKey == VirtualKey::Left || virtualKey == VirtualKey::Right ||
+        (int)virtualKey == 166 || (int)virtualKey == 167))
+    {
+        auto coreWindow = Window::Current->CoreWindow;
+        auto downState = Windows::UI::Core::CoreVirtualKeyStates::Down;
+        bool menuKey = (coreWindow->GetKeyState(VirtualKey::Menu) & downState) == downState;
+        bool controlKey = (coreWindow->GetKeyState(VirtualKey::Control) & downState) == downState;
+        bool shiftKey = (coreWindow->GetKeyState(VirtualKey::Shift) & downState) == downState;
+        bool noModifiers = !menuKey && !controlKey && !shiftKey;
+        bool onlyAlt = menuKey && !controlKey && !shiftKey;
+
+        if (((int)virtualKey == 166 && noModifiers) ||
+            (virtualKey == VirtualKey::Left && onlyAlt))
+        {
+            // When the previous key or Alt+Left are pressed navigate back
+            args->Handled = true;
+            GoBack(this, ref new RoutedEventArgs());
+        }
+        else if (((int)virtualKey == 167 && noModifiers) ||
+            (virtualKey == VirtualKey::Right && onlyAlt))
+        {
+            // When the next key or Alt+Right are pressed navigate forward
+            args->Handled = true;
+            GoForward(this, ref new RoutedEventArgs());
+        }
+    }
+}
+
+/// <summary>
+/// Invoked on every mouse click, touch screen tap, or equivalent interaction when this
+/// page is active and occupies the entire window.  Used to detect browser-style next and
+/// previous mouse button clicks to navigate between pages.
+/// </summary>
+/// <param name="sender">Instance that triggered the event.</param>
+/// <param name="args">Event data describing the conditions that led to the event.</param>
+void LayoutAwarePage::CoreWindow_PointerPressed(CoreWindow^ sender, PointerEventArgs^ args)
+{
+    auto properties = args->CurrentPoint->Properties;
+
+    // Ignore button chords with the left, right, and middle buttons
+    if (properties->IsLeftButtonPressed || properties->IsRightButtonPressed ||
+        properties->IsMiddleButtonPressed) return;
+
+    // If back or foward are pressed (but not both) navigate appropriately
+    bool backPressed = properties->IsXButton1Pressed;
+    bool forwardPressed = properties->IsXButton2Pressed;
+    if (backPressed ^ forwardPressed)
+    {
+        args->Handled = true;
+        if (backPressed) GoBack(this, ref new RoutedEventArgs());
+        if (forwardPressed) GoForward(this, ref new RoutedEventArgs());
+    }
+}
+
+#pragma endregion
+
+#pragma region Visual state switching
+
+/// <summary>
+/// Invoked as an event handler, typically on the <see cref="Loaded"/> event of a
+/// <see cref="Control"/> within the page, to indicate that the sender should start receiving
+/// visual state management changes that correspond to application view state changes.
+/// </summary>
+/// <param name="sender">Instance of <see cref="Control"/> that supports visual state management
+/// corresponding to view states.</param>
+/// <param name="e">Event data that describes how the request was made.</param>
+/// <remarks>The current view state will immediately be used to set the corresponding visual state
+/// when layout updates are requested.  A corresponding <see cref="Unloaded"/> event handler
+/// connected to <see cref="StopLayoutUpdates"/> is strongly encouraged.  Instances of
+/// <see cref="LayoutAwarePage"/> automatically invoke these handlers in their Loaded and Unloaded
+/// events.</remarks>
+/// <seealso cref="DetermineVisualState"/>
+/// <seealso cref="InvalidateVisualState"/>
+void LayoutAwarePage::StartLayoutUpdates(Object^ sender, RoutedEventArgs^ e)
+{
+    (void) e;	// Unused parameter
+
+    auto control = safe_cast<Control^>(sender);
+    if (_layoutAwareControls == nullptr)
+    {
+        // Start listening to view state changes when there are controls interested in updates
+        _layoutAwareControls = ref new Vector<Control^>();
+        _windowSizeEventToken = Window::Current->SizeChanged += ref new WindowSizeChangedEventHandler(this, &LayoutAwarePage::WindowSizeChanged);
+
+        // Page receives notifications for children. Protect the page until we stopped layout updates for all controls.
+        _this = this;
+    }
+    _layoutAwareControls->Append(control);
+
+    // Set the initial visual state of the control
+    VisualStateManager::GoToState(control, DetermineVisualState(ApplicationView::Value), false);
+}
+
+void LayoutAwarePage::WindowSizeChanged(Object^ sender, Windows::UI::Core::WindowSizeChangedEventArgs^ e)
+{
+    (void) sender;	// Unused parameter
+    (void) e;	// Unused parameter
+
+    InvalidateVisualState();
+}
+
+/// <summary>
+/// Invoked as an event handler, typically on the <see cref="Unloaded"/> event of a
+/// <see cref="Control"/>, to indicate that the sender should start receiving visual state
+/// management changes that correspond to application view state changes.
+/// </summary>
+/// <param name="sender">Instance of <see cref="Control"/> that supports visual state management
+/// corresponding to view states.</param>
+/// <param name="e">Event data that describes how the request was made.</param>
+/// <remarks>The current view state will immediately be used to set the corresponding visual state
+/// when layout updates are requested.</remarks>
+/// <seealso cref="StartLayoutUpdates"/>
+void LayoutAwarePage::StopLayoutUpdates(Object^ sender, RoutedEventArgs^ e)
+{
+    (void) e;	// Unused parameter
+
+    auto control = safe_cast<Control^>(sender);
+    unsigned int index;
+    if (_layoutAwareControls != nullptr && _layoutAwareControls->IndexOf(control, &index))
+    {
+        _layoutAwareControls->RemoveAt(index);
+        if (_layoutAwareControls->Size == 0)
+        {
+            // Stop listening to view state changes when no controls are interested in updates
+            Window::Current->SizeChanged -= _windowSizeEventToken;
+            _layoutAwareControls = nullptr;
+            // Last control has received the Unload notification.
+            _this = nullptr;
+        }
+    }
+}
+
+/// <summary>
+/// Translates <see cref="ApplicationViewState"/> values into strings for visual state management
+/// within the page.  The default implementation uses the names of enum values.  Subclasses may
+/// override this method to control the mapping scheme used.
+/// </summary>
+/// <param name="viewState">View state for which a visual state is desired.</param>
+/// <returns>Visual state name used to drive the <see cref="VisualStateManager"/></returns>
+/// <seealso cref="InvalidateVisualState"/>
+String^ LayoutAwarePage::DetermineVisualState(ApplicationViewState viewState)
+{
+    switch (viewState)
+    {
+    case ApplicationViewState::Filled:
+        return "Filled";
+    case ApplicationViewState::Snapped:
+        return "Snapped";
+    case ApplicationViewState::FullScreenPortrait:
+        return "FullScreenPortrait";
+    case ApplicationViewState::FullScreenLandscape:
+    default:
+        return "FullScreenLandscape";
+    }
+}
+
+/// <summary>
+/// Updates all controls that are listening for visual state changes with the correct visual
+/// state.
+/// </summary>
+/// <remarks>
+/// Typically used in conjunction with overriding <see cref="DetermineVisualState"/> to
+/// signal that a different value may be returned even though the view state has not changed.
+/// </remarks>
+void LayoutAwarePage::InvalidateVisualState()
+{
+    if (_layoutAwareControls != nullptr)
+    {
+        String^ visualState = DetermineVisualState(ApplicationView::Value);
+        auto controlIterator = _layoutAwareControls->First();
+        while (controlIterator->HasCurrent)
+        {
+            auto control = controlIterator->Current;
+            VisualStateManager::GoToState(control, visualState, false);
+            controlIterator->MoveNext();
+        }
+    }
+}
+
+#pragma endregion
+
+#pragma region Process lifetime management
+
+/// <summary>
+/// Invoked when this page is about to be displayed in a Frame.
+/// </summary>
+/// <param name="e">Event data that describes how this page was reached.  The Parameter
+/// property provides the group to be displayed.</param>
+void LayoutAwarePage::OnNavigatedTo(NavigationEventArgs^ e)
+{
+    // Returning to a cached page through navigation shouldn't trigger state loading
+    if (_pageKey != nullptr) return;
+
+    auto frameState = SuspensionManager::SessionStateForFrame(Frame);
+    _pageKey = "Page-" + Frame->BackStackDepth;
+
+    if (e->NavigationMode == NavigationMode::New)
+    {
+        // Clear existing state for forward navigation when adding a new page to the
+        // navigation stack
+        auto nextPageKey = _pageKey;
+        int nextPageIndex = Frame->BackStackDepth;
+        while (frameState->HasKey(nextPageKey))
+        {
+            frameState->Remove(nextPageKey);
+            nextPageIndex++;
+            nextPageKey = "Page-" + nextPageIndex;
+        }
+
+        // Pass the navigation parameter to the new page
+        LoadState(e->Parameter, nullptr);
+    }
+    else
+    {
+        // Pass the navigation parameter and preserved page state to the page, using
+        // the same strategy for loading suspended state and recreating pages discarded
+        // from cache
+        LoadState(e->Parameter, safe_cast<IMap<String^, Object^>^>(frameState->Lookup(_pageKey)));
+    }
+}
+
+/// <summary>
+/// Invoked when this page will no longer be displayed in a Frame.
+/// </summary>
+/// <param name="e">Event data that describes how this page was reached.  The Parameter
+/// property provides the group to be displayed.</param>
+void LayoutAwarePage::OnNavigatedFrom(NavigationEventArgs^ e)
+{
+    auto frameState = SuspensionManager::SessionStateForFrame(Frame);
+    auto pageState = ref new Map<String^, Object^>();
+    SaveState(pageState);
+    frameState->Insert(_pageKey, pageState);
+}
+
+/// <summary>
+/// Populates the page with content passed during navigation.  Any saved state is also
+/// provided when recreating a page from a prior session.
+/// </summary>
+/// <param name="navigationParameter">The parameter value passed to
+/// <see cref="Frame.Navigate(Type, Object)"/> when this page was initially requested.
+/// </param>
+/// <param name="pageState">A map of state preserved by this page during an earlier
+/// session.  This will be null the first time a page is visited.</param>
+void LayoutAwarePage::LoadState(Object^ navigationParameter, IMap<String^, Object^>^ pageState)
+{
+}
+
+/// <summary>
+/// Preserves state associated with this page in case the application is suspended or the
+/// page is discarded from the navigation cache.  Values must conform to the serialization
+/// requirements of <see cref="SuspensionManager.SessionState"/>.
+/// </summary>
+/// <param name="pageState">An empty map to be populated with serializable state.</param>
+void LayoutAwarePage::SaveState(IMap<String^, Object^>^ pageState)
+{
+}
+
+#pragma endregion
diff --git a/samples/winrt/ImageManipulations/common/LayoutAwarePage.h b/samples/winrt/ImageManipulations/common/LayoutAwarePage.h
new file mode 100644
index 0000000000..bd71062fe9
--- /dev/null
+++ b/samples/winrt/ImageManipulations/common/LayoutAwarePage.h
@@ -0,0 +1,88 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+#pragma once
+
+#include <collection.h>
+
+namespace SDKSample
+{
+    namespace Common
+    {
+        /// <summary>
+        /// Typical implementation of Page that provides several important conveniences:
+        /// <list type="bullet">
+        /// <item>
+        /// <description>Application view state to visual state mapping</description>
+        /// </item>
+        /// <item>
+        /// <description>GoBack, GoForward, and GoHome event handlers</description>
+        /// </item>
+        /// <item>
+        /// <description>Mouse and keyboard shortcuts for navigation</description>
+        /// </item>
+        /// <item>
+        /// <description>State management for navigation and process lifetime management</description>
+        /// </item>
+        /// <item>
+        /// <description>A default view model</description>
+        /// </item>
+        /// </list>
+        /// </summary>
+        [Windows::Foundation::Metadata::WebHostHidden]
+        public ref class LayoutAwarePage : Windows::UI::Xaml::Controls::Page
+        {
+        internal:
+            LayoutAwarePage();
+
+        public:
+            void StartLayoutUpdates(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            void StopLayoutUpdates(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            void InvalidateVisualState();
+            static property Windows::UI::Xaml::DependencyProperty^ DefaultViewModelProperty
+            {
+                Windows::UI::Xaml::DependencyProperty^ get();
+            };
+            property Windows::Foundation::Collections::IObservableMap<Platform::String^, Platform::Object^>^ DefaultViewModel
+            {
+                Windows::Foundation::Collections::IObservableMap<Platform::String^, Platform::Object^>^ get();
+                void set(Windows::Foundation::Collections::IObservableMap<Platform::String^, Platform::Object^>^ value);
+            }
+
+        protected:
+            virtual void GoHome(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            virtual void GoBack(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            virtual void GoForward(Platform::Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            virtual Platform::String^ DetermineVisualState(Windows::UI::ViewManagement::ApplicationViewState viewState);
+            virtual void OnNavigatedTo(Windows::UI::Xaml::Navigation::NavigationEventArgs^ e) override;
+            virtual void OnNavigatedFrom(Windows::UI::Xaml::Navigation::NavigationEventArgs^ e) override;
+            virtual void LoadState(Platform::Object^ navigationParameter,
+                Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ pageState);
+            virtual void SaveState(Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ pageState);
+
+        private:
+            Platform::String^ _pageKey;
+            bool _navigationShortcutsRegistered;
+            Platform::Collections::Map<Platform::String^, Platform::Object^>^ _defaultViewModel;
+            Windows::Foundation::EventRegistrationToken _windowSizeEventToken,
+                _acceleratorKeyEventToken, _pointerPressedEventToken;
+            Platform::Collections::Vector<Windows::UI::Xaml::Controls::Control^>^ _layoutAwareControls;
+            void WindowSizeChanged(Platform::Object^ sender, Windows::UI::Core::WindowSizeChangedEventArgs^ e);
+            void OnLoaded(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+            void OnUnloaded(Object^ sender, Windows::UI::Xaml::RoutedEventArgs^ e);
+
+            void CoreDispatcher_AcceleratorKeyActivated(Windows::UI::Core::CoreDispatcher^ sender,
+                Windows::UI::Core::AcceleratorKeyEventArgs^ args);
+            void CoreWindow_PointerPressed(Windows::UI::Core::CoreWindow^ sender,
+                Windows::UI::Core::PointerEventArgs^ args);
+            LayoutAwarePage^ _this; // Strong reference to self, cleaned up in OnUnload
+        };
+    }
+}
diff --git a/samples/winrt/ImageManipulations/common/StandardStyles.xaml b/samples/winrt/ImageManipulations/common/StandardStyles.xaml
new file mode 100644
index 0000000000..b4edc81f44
--- /dev/null
+++ b/samples/winrt/ImageManipulations/common/StandardStyles.xaml
@@ -0,0 +1,978 @@
+<!--
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+-->
+
+<ResourceDictionary
+    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml">
+
+    <!-- Non-brush values that vary across themes -->
+
+    <ResourceDictionary.ThemeDictionaries>
+        <ResourceDictionary x:Key="Default">
+            <x:String x:Key="BackButtonGlyph">&#xE071;</x:String>
+            <x:String x:Key="BackButtonSnappedGlyph">&#xE0BA;</x:String>
+        </ResourceDictionary>
+
+        <ResourceDictionary x:Key="HighContrast">
+            <x:String x:Key="BackButtonGlyph">&#xE0A6;</x:String>
+            <x:String x:Key="BackButtonSnappedGlyph">&#xE0C4;</x:String>
+        </ResourceDictionary>
+    </ResourceDictionary.ThemeDictionaries>
+
+    <!-- RichTextBlock styles -->
+
+    <Style x:Key="BasicRichTextStyle" TargetType="RichTextBlock">
+        <Setter Property="Foreground" Value="{StaticResource ApplicationForegroundThemeBrush}"/>
+        <Setter Property="FontSize" Value="{StaticResource ControlContentThemeFontSize}"/>
+        <Setter Property="FontFamily" Value="{StaticResource ContentControlThemeFontFamily}"/>
+        <Setter Property="TextTrimming" Value="WordEllipsis"/>
+        <Setter Property="TextWrapping" Value="Wrap"/>
+        <Setter Property="Typography.StylisticSet20" Value="True"/>
+        <Setter Property="Typography.DiscretionaryLigatures" Value="True"/>
+    </Style>
+
+    <Style x:Key="BaselineRichTextStyle" TargetType="RichTextBlock" BasedOn="{StaticResource BasicRichTextStyle}">
+        <Setter Property="LineHeight" Value="20"/>
+        <Setter Property="LineStackingStrategy" Value="BlockLineHeight"/>
+        <!-- Properly align text along its baseline -->
+        <Setter Property="RenderTransform">
+            <Setter.Value>
+                <TranslateTransform X="-1" Y="4"/>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <Style x:Key="ItemRichTextStyle" TargetType="RichTextBlock" BasedOn="{StaticResource BaselineRichTextStyle}"/>
+
+    <Style x:Key="BodyRichTextStyle" TargetType="RichTextBlock" BasedOn="{StaticResource BaselineRichTextStyle}">
+        <Setter Property="FontWeight" Value="SemiLight"/>
+    </Style>
+
+    <!-- TextBlock styles -->
+
+    <Style x:Key="BasicTextStyle" TargetType="TextBlock">
+        <Setter Property="Foreground" Value="{StaticResource ApplicationForegroundThemeBrush}"/>
+        <Setter Property="FontSize" Value="{StaticResource ControlContentThemeFontSize}"/>
+        <Setter Property="FontFamily" Value="{StaticResource ContentControlThemeFontFamily}"/>
+        <Setter Property="TextTrimming" Value="WordEllipsis"/>
+        <Setter Property="TextWrapping" Value="Wrap"/>
+        <Setter Property="Typography.StylisticSet20" Value="True"/>
+        <Setter Property="Typography.DiscretionaryLigatures" Value="True"/>
+    </Style>
+
+    <Style x:Key="BaselineTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BasicTextStyle}">
+        <Setter Property="LineHeight" Value="20"/>
+        <Setter Property="LineStackingStrategy" Value="BlockLineHeight"/>
+        <!-- Properly align text along its baseline -->
+        <Setter Property="RenderTransform">
+            <Setter.Value>
+                <TranslateTransform X="-1" Y="4"/>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <Style x:Key="HeaderTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}">
+        <Setter Property="FontSize" Value="56"/>
+        <Setter Property="FontWeight" Value="Light"/>
+        <Setter Property="LineHeight" Value="40"/>
+        <Setter Property="RenderTransform">
+            <Setter.Value>
+                <TranslateTransform X="-2" Y="8"/>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <Style x:Key="SubheaderTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}">
+        <Setter Property="FontSize" Value="26.667"/>
+        <Setter Property="FontWeight" Value="Light"/>
+        <Setter Property="LineHeight" Value="30"/>
+        <Setter Property="RenderTransform">
+            <Setter.Value>
+                <TranslateTransform X="-1" Y="6"/>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <Style x:Key="TitleTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}">
+        <Setter Property="FontWeight" Value="SemiBold"/>
+    </Style>
+
+    <Style x:Key="ItemTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}"/>
+
+    <Style x:Key="BodyTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}">
+        <Setter Property="FontWeight" Value="SemiLight"/>
+    </Style>
+
+    <Style x:Key="CaptionTextStyle" TargetType="TextBlock" BasedOn="{StaticResource BaselineTextStyle}">
+        <Setter Property="FontSize" Value="12"/>
+        <Setter Property="Foreground" Value="{StaticResource ApplicationSecondaryForegroundThemeBrush}"/>
+    </Style>
+
+    <!-- Button styles -->
+
+    <!--
+        TextButtonStyle is used to style a Button using subheader-styled text with no other adornment.  This
+        style is used in the GroupedItemsPage as a group header and in the FileOpenPickerPage for triggering
+        commands.
+    -->
+    <Style x:Key="TextButtonStyle" TargetType="Button">
+        <Setter Property="MinWidth" Value="0"/>
+        <Setter Property="MinHeight" Value="0"/>
+        <Setter Property="Template">
+            <Setter.Value>
+                <ControlTemplate TargetType="Button">
+                    <Grid Background="Transparent">
+                        <TextBlock
+                            x:Name="Text"
+                            Text="{TemplateBinding Content}"
+                            Margin="3,-7,3,10"
+                            TextWrapping="NoWrap"
+                            Style="{StaticResource SubheaderTextStyle}"/>
+                        <Rectangle
+                            x:Name="FocusVisualWhite"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualWhiteStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="1.5"/>
+                        <Rectangle
+                            x:Name="FocusVisualBlack"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualBlackStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="0.5"/>
+
+                        <VisualStateManager.VisualStateGroups>
+                            <VisualStateGroup x:Name="CommonStates">
+                                <VisualState x:Name="Normal"/>
+                                <VisualState x:Name="PointerOver">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ApplicationPointerOverForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Pressed">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ApplicationPressedForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Disabled">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ButtonDisabledForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="FocusStates">
+                                <VisualState x:Name="Focused">
+                                    <Storyboard>
+                                        <DoubleAnimation Duration="0" To="1" Storyboard.TargetName="FocusVisualWhite" Storyboard.TargetProperty="Opacity"/>
+                                        <DoubleAnimation Duration="0" To="1" Storyboard.TargetName="FocusVisualBlack" Storyboard.TargetProperty="Opacity"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Unfocused"/>
+                            </VisualStateGroup>
+                        </VisualStateManager.VisualStateGroups>
+                    </Grid>
+                </ControlTemplate>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <!--
+        TextRadioButtonStyle is used to style a RadioButton using subheader-styled text with no other adornment.
+        This style is used in the SearchResultsPage to allow selection among filters.
+    -->
+    <Style x:Key="TextRadioButtonStyle" TargetType="RadioButton">
+        <Setter Property="MinWidth" Value="0"/>
+        <Setter Property="MinHeight" Value="0"/>
+        <Setter Property="Template">
+            <Setter.Value>
+                <ControlTemplate TargetType="RadioButton">
+                    <Grid Background="Transparent">
+                        <TextBlock
+                            x:Name="Text"
+                            Text="{TemplateBinding Content}"
+                            Margin="3,-7,3,10"
+                            TextWrapping="NoWrap"
+                            Style="{StaticResource SubheaderTextStyle}"/>
+                        <Rectangle
+                            x:Name="FocusVisualWhite"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualWhiteStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="1.5"/>
+                        <Rectangle
+                            x:Name="FocusVisualBlack"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualBlackStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="0.5"/>
+
+                        <VisualStateManager.VisualStateGroups>
+                            <VisualStateGroup x:Name="CommonStates">
+                                <VisualState x:Name="Normal"/>
+                                <VisualState x:Name="PointerOver">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ApplicationPointerOverForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Pressed">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ApplicationPressedForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Disabled">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ButtonDisabledForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="FocusStates">
+                                <VisualState x:Name="Focused">
+                                    <Storyboard>
+                                        <DoubleAnimation Duration="0" To="1" Storyboard.TargetName="FocusVisualWhite" Storyboard.TargetProperty="Opacity"/>
+                                        <DoubleAnimation Duration="0" To="1" Storyboard.TargetName="FocusVisualBlack" Storyboard.TargetProperty="Opacity"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Unfocused"/>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="CheckStates">
+                                <VisualState x:Name="Checked"/>
+                                <VisualState x:Name="Unchecked">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Text" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource ApplicationSecondaryForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Indeterminate"/>
+                            </VisualStateGroup>
+                        </VisualStateManager.VisualStateGroups>
+                    </Grid>
+                </ControlTemplate>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <!--
+        AppBarButtonStyle is used to style a Button for use in an App Bar.  Content will be centered and should fit within
+        the 40-pixel radius glyph provided.  16-point Segoe UI Symbol is used for content text to simplify the use of glyphs
+        from that font.  AutomationProperties.Name is used for the text below the glyph.
+    -->
+    <Style x:Key="AppBarButtonStyle" TargetType="Button">
+        <Setter Property="Foreground" Value="{StaticResource AppBarItemForegroundThemeBrush}"/>
+        <Setter Property="VerticalAlignment" Value="Stretch"/>
+        <Setter Property="FontFamily" Value="Segoe UI Symbol"/>
+        <Setter Property="FontWeight" Value="Normal"/>
+        <Setter Property="FontSize" Value="20"/>
+        <Setter Property="AutomationProperties.ItemType" Value="App Bar Button"/>
+        <Setter Property="Template">
+            <Setter.Value>
+                <ControlTemplate TargetType="Button">
+                    <Grid Width="100" Background="Transparent">
+                        <StackPanel VerticalAlignment="Top" Margin="0,14,0,13">
+                            <Grid Width="40" Height="40" Margin="0,0,0,5" HorizontalAlignment="Center">
+                                <TextBlock x:Name="BackgroundGlyph" Text="&#xE0A8;" FontFamily="Segoe UI Symbol" FontSize="53.333" Margin="-4,-19,0,0" Foreground="{StaticResource AppBarItemBackgroundThemeBrush}"/>
+                                <TextBlock x:Name="OutlineGlyph" Text="&#xE0A7;" FontFamily="Segoe UI Symbol" FontSize="53.333" Margin="-4,-19,0,0"/>
+                                <ContentPresenter x:Name="Content" HorizontalAlignment="Center" Margin="-1,-1,0,0" VerticalAlignment="Center"/>
+                            </Grid>
+                            <TextBlock
+                                x:Name="TextLabel"
+                                Text="{TemplateBinding AutomationProperties.Name}"
+                                Margin="0,0,2,0"
+                                FontSize="12"
+                                TextAlignment="Center"
+                                Width="88"
+                                MaxHeight="32"
+                                TextTrimming="WordEllipsis"
+                                Style="{StaticResource BasicTextStyle}"/>
+                        </StackPanel>
+                        <Rectangle
+                                x:Name="FocusVisualWhite"
+                                IsHitTestVisible="False"
+                                Stroke="{StaticResource FocusVisualWhiteStrokeThemeBrush}"
+                                StrokeEndLineCap="Square"
+                                StrokeDashArray="1,1"
+                                Opacity="0"
+                                StrokeDashOffset="1.5"/>
+                        <Rectangle
+                                x:Name="FocusVisualBlack"
+                                IsHitTestVisible="False"
+                                Stroke="{StaticResource FocusVisualBlackStrokeThemeBrush}"
+                                StrokeEndLineCap="Square"
+                                StrokeDashArray="1,1"
+                                Opacity="0"
+                                StrokeDashOffset="0.5"/>
+
+                        <VisualStateManager.VisualStateGroups>
+                            <VisualStateGroup x:Name="CommonStates">
+                                <VisualState x:Name="Normal"/>
+                                <VisualState x:Name="PointerOver">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemPointerOverBackgroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Content" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemPointerOverForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Pressed">
+                                    <Storyboard>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="OutlineGlyph"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="0"
+                                            Duration="0"/>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Content" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemPressedForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Disabled">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="OutlineGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemDisabledForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="Content" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemDisabledForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="TextLabel" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource AppBarItemDisabledForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="FocusStates">
+                                <VisualState x:Name="Focused">
+                                    <Storyboard>
+                                        <DoubleAnimation
+                                                Storyboard.TargetName="FocusVisualWhite"
+                                                Storyboard.TargetProperty="Opacity"
+                                                To="1"
+                                                Duration="0"/>
+                                        <DoubleAnimation
+                                                Storyboard.TargetName="FocusVisualBlack"
+                                                Storyboard.TargetProperty="Opacity"
+                                                To="1"
+                                                Duration="0"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Unfocused" />
+                                <VisualState x:Name="PointerFocused" />
+                            </VisualStateGroup>
+                        </VisualStateManager.VisualStateGroups>
+                    </Grid>
+                </ControlTemplate>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <!-- Standard App Bar buttons -->
+
+    <Style x:Key="SkipBackAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="SkipBackAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Skip Back"/>
+        <Setter Property="Content" Value="&#xE100;"/>
+    </Style>
+    <Style x:Key="SkipAheadAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="SkipAheadAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Skip Ahead"/>
+        <Setter Property="Content" Value="&#xE101;"/>
+    </Style>
+    <Style x:Key="PlayAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="PlayAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Play"/>
+        <Setter Property="Content" Value="&#xE102;"/>
+    </Style>
+    <Style x:Key="PauseAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="PauseAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Pause"/>
+        <Setter Property="Content" Value="&#xE103;"/>
+    </Style>
+    <Style x:Key="EditAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="EditAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Edit"/>
+        <Setter Property="Content" Value="&#xE104;"/>
+    </Style>
+    <Style x:Key="SaveAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="SaveAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Save"/>
+        <Setter Property="Content" Value="&#xE105;"/>
+    </Style>
+    <Style x:Key="DeleteAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="DeleteAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Delete"/>
+        <Setter Property="Content" Value="&#xE106;"/>
+    </Style>
+    <Style x:Key="DiscardAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="DiscardAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Discard"/>
+        <Setter Property="Content" Value="&#xE107;"/>
+    </Style>
+    <Style x:Key="RemoveAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="RemoveAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Remove"/>
+        <Setter Property="Content" Value="&#xE108;"/>
+    </Style>
+    <Style x:Key="AddAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="AddAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Add"/>
+        <Setter Property="Content" Value="&#xE109;"/>
+    </Style>
+    <Style x:Key="NoAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="NoAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="No"/>
+        <Setter Property="Content" Value="&#xE10A;"/>
+    </Style>
+    <Style x:Key="YesAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="YesAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Yes"/>
+        <Setter Property="Content" Value="&#xE10B;"/>
+    </Style>
+    <Style x:Key="MoreAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="MoreAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="More"/>
+        <Setter Property="Content" Value="&#xE10C;"/>
+    </Style>
+    <Style x:Key="RedoAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="RedoAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Redo"/>
+        <Setter Property="Content" Value="&#xE10D;"/>
+    </Style>
+    <Style x:Key="UndoAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="UndoAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Undo"/>
+        <Setter Property="Content" Value="&#xE10E;"/>
+    </Style>
+    <Style x:Key="HomeAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="HomeAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Home"/>
+        <Setter Property="Content" Value="&#xE10F;"/>
+    </Style>
+    <Style x:Key="OutAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="OutAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Out"/>
+        <Setter Property="Content" Value="&#xE110;"/>
+    </Style>
+    <Style x:Key="NextAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="NextAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Next"/>
+        <Setter Property="Content" Value="&#xE111;"/>
+    </Style>
+    <Style x:Key="PreviousAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="PreviousAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Previous"/>
+        <Setter Property="Content" Value="&#xE112;"/>
+    </Style>
+    <Style x:Key="FavoriteAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="FavoriteAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Favorite"/>
+        <Setter Property="Content" Value="&#xE113;"/>
+    </Style>
+    <Style x:Key="PhotoAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="PhotoAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Photo"/>
+        <Setter Property="Content" Value="&#xE114;"/>
+    </Style>
+    <Style x:Key="SettingsAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="SettingsAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Settings"/>
+        <Setter Property="Content" Value="&#xE115;"/>
+    </Style>
+    <Style x:Key="VideoAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="VideoAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Video"/>
+        <Setter Property="Content" Value="&#xE116;"/>
+    </Style>
+    <Style x:Key="RefreshAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="RefreshAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Refresh"/>
+        <Setter Property="Content" Value="&#xE117;"/>
+    </Style>
+    <Style x:Key="DownloadAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="DownloadAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Download"/>
+        <Setter Property="Content" Value="&#xE118;"/>
+    </Style>
+    <Style x:Key="MailAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="MailAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Mail"/>
+        <Setter Property="Content" Value="&#xE119;"/>
+    </Style>
+    <Style x:Key="SearchAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="SearchAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Search"/>
+        <Setter Property="Content" Value="&#xE11A;"/>
+    </Style>
+    <Style x:Key="HelpAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="HelpAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Help"/>
+        <Setter Property="Content" Value="&#xE11B;"/>
+    </Style>
+    <Style x:Key="UploadAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="UploadAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Upload"/>
+        <Setter Property="Content" Value="&#xE11C;"/>
+    </Style>
+    <Style x:Key="PinAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="PinAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Pin"/>
+        <Setter Property="Content" Value="&#xE141;"/>
+    </Style>
+    <Style x:Key="UnpinAppBarButtonStyle" TargetType="Button" BasedOn="{StaticResource AppBarButtonStyle}">
+        <Setter Property="AutomationProperties.AutomationId" Value="UnpinAppBarButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Unpin"/>
+        <Setter Property="Content" Value="&#xE196;"/>
+    </Style>
+
+    <!-- Title area styles -->
+
+    <Style x:Key="PageHeaderTextStyle" TargetType="TextBlock" BasedOn="{StaticResource HeaderTextStyle}">
+        <Setter Property="TextWrapping" Value="NoWrap"/>
+        <Setter Property="VerticalAlignment" Value="Bottom"/>
+        <Setter Property="Margin" Value="0,0,40,40"/>
+    </Style>
+
+    <Style x:Key="PageSubheaderTextStyle" TargetType="TextBlock" BasedOn="{StaticResource SubheaderTextStyle}">
+        <Setter Property="TextWrapping" Value="NoWrap"/>
+        <Setter Property="VerticalAlignment" Value="Bottom"/>
+        <Setter Property="Margin" Value="0,0,0,40"/>
+    </Style>
+
+    <Style x:Key="SnappedPageHeaderTextStyle" TargetType="TextBlock" BasedOn="{StaticResource PageSubheaderTextStyle}">
+        <Setter Property="Margin" Value="0,0,18,40"/>
+    </Style>
+
+    <!--
+        BackButtonStyle is used to style a Button for use in the title area of a page.  Margins appropriate for
+        the conventional page layout are included as part of the style.
+    -->
+    <Style x:Key="BackButtonStyle" TargetType="Button">
+        <Setter Property="MinWidth" Value="0"/>
+        <Setter Property="Width" Value="48"/>
+        <Setter Property="Height" Value="48"/>
+        <Setter Property="Margin" Value="36,0,36,36"/>
+        <Setter Property="VerticalAlignment" Value="Bottom"/>
+        <Setter Property="FontFamily" Value="Segoe UI Symbol"/>
+        <Setter Property="FontWeight" Value="Normal"/>
+        <Setter Property="FontSize" Value="56"/>
+        <Setter Property="AutomationProperties.AutomationId" Value="BackButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Back"/>
+        <Setter Property="AutomationProperties.ItemType" Value="Navigation Button"/>
+        <Setter Property="Template">
+            <Setter.Value>
+                <ControlTemplate TargetType="Button">
+                    <Grid x:Name="RootGrid">
+                        <Grid Margin="-1,-16,0,0">
+                            <TextBlock x:Name="BackgroundGlyph" Text="&#xE0A8;" Foreground="{StaticResource BackButtonBackgroundThemeBrush}"/>
+                            <TextBlock x:Name="NormalGlyph" Text="{StaticResource BackButtonGlyph}" Foreground="{StaticResource BackButtonForegroundThemeBrush}"/>
+                            <TextBlock x:Name="ArrowGlyph" Text="&#xE0A6;" Foreground="{StaticResource BackButtonPressedForegroundThemeBrush}" Opacity="0"/>
+                        </Grid>
+                        <Rectangle
+                            x:Name="FocusVisualWhite"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualWhiteStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="1.5"/>
+                        <Rectangle
+                            x:Name="FocusVisualBlack"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualBlackStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="0.5"/>
+
+                        <VisualStateManager.VisualStateGroups>
+                            <VisualStateGroup x:Name="CommonStates">
+                                <VisualState x:Name="Normal" />
+                                <VisualState x:Name="PointerOver">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonPointerOverBackgroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="NormalGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonPointerOverForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Pressed">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="ArrowGlyph"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="NormalGlyph"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="0"
+                                            Duration="0"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Disabled">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="RootGrid" Storyboard.TargetProperty="Visibility">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="Collapsed"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="FocusStates">
+                                <VisualState x:Name="Focused">
+                                    <Storyboard>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="FocusVisualWhite"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="FocusVisualBlack"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Unfocused" />
+                                <VisualState x:Name="PointerFocused" />
+                            </VisualStateGroup>
+                        </VisualStateManager.VisualStateGroups>
+                    </Grid>
+                </ControlTemplate>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <!--
+        PortraitBackButtonStyle is used to style a Button for use in the title area of a portrait page.  Margins appropriate
+        for the conventional page layout are included as part of the style.
+    -->
+    <Style x:Key="PortraitBackButtonStyle" TargetType="Button" BasedOn="{StaticResource BackButtonStyle}">
+        <Setter Property="Margin" Value="26,0,26,36"/>
+    </Style>
+
+    <!--
+        SnappedBackButtonStyle is used to style a Button for use in the title area of a snapped page.  Margins appropriate
+        for the conventional page layout are included as part of the style.
+
+        The obvious duplication here is necessary as the glyphs used in snapped are not merely smaller versions of the same
+        glyph but are actually distinct.
+    -->
+    <Style x:Key="SnappedBackButtonStyle" TargetType="Button">
+        <Setter Property="MinWidth" Value="0"/>
+        <Setter Property="Margin" Value="20,0,0,0"/>
+        <Setter Property="VerticalAlignment" Value="Bottom"/>
+        <Setter Property="FontFamily" Value="Segoe UI Symbol"/>
+        <Setter Property="FontWeight" Value="Normal"/>
+        <Setter Property="FontSize" Value="26.66667"/>
+        <Setter Property="AutomationProperties.AutomationId" Value="BackButton"/>
+        <Setter Property="AutomationProperties.Name" Value="Back"/>
+        <Setter Property="AutomationProperties.ItemType" Value="Navigation Button"/>
+        <Setter Property="Template">
+            <Setter.Value>
+                <ControlTemplate TargetType="Button">
+                    <Grid x:Name="RootGrid" Width="36" Height="36" Margin="-3,0,7,33">
+                        <Grid Margin="-1,-1,0,0">
+                            <TextBlock x:Name="BackgroundGlyph" Text="&#xE0D4;" Foreground="{StaticResource BackButtonBackgroundThemeBrush}"/>
+                            <TextBlock x:Name="NormalGlyph" Text="{StaticResource BackButtonSnappedGlyph}" Foreground="{StaticResource BackButtonForegroundThemeBrush}"/>
+                            <TextBlock x:Name="ArrowGlyph" Text="&#xE0C4;" Foreground="{StaticResource BackButtonPressedForegroundThemeBrush}" Opacity="0"/>
+                        </Grid>
+                        <Rectangle
+                            x:Name="FocusVisualWhite"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualWhiteStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="1.5"/>
+                        <Rectangle
+                            x:Name="FocusVisualBlack"
+                            IsHitTestVisible="False"
+                            Stroke="{StaticResource FocusVisualBlackStrokeThemeBrush}"
+                            StrokeEndLineCap="Square"
+                            StrokeDashArray="1,1"
+                            Opacity="0"
+                            StrokeDashOffset="0.5"/>
+
+                        <VisualStateManager.VisualStateGroups>
+                            <VisualStateGroup x:Name="CommonStates">
+                                <VisualState x:Name="Normal" />
+                                <VisualState x:Name="PointerOver">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonPointerOverBackgroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="NormalGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonPointerOverForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Pressed">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="BackgroundGlyph" Storyboard.TargetProperty="Foreground">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="{StaticResource BackButtonForegroundThemeBrush}"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="ArrowGlyph"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="NormalGlyph"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="0"
+                                            Duration="0"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Disabled">
+                                    <Storyboard>
+                                        <ObjectAnimationUsingKeyFrames Storyboard.TargetName="RootGrid" Storyboard.TargetProperty="Visibility">
+                                            <DiscreteObjectKeyFrame KeyTime="0" Value="Collapsed"/>
+                                        </ObjectAnimationUsingKeyFrames>
+                                    </Storyboard>
+                                </VisualState>
+                            </VisualStateGroup>
+                            <VisualStateGroup x:Name="FocusStates">
+                                <VisualState x:Name="Focused">
+                                    <Storyboard>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="FocusVisualWhite"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                        <DoubleAnimation
+                                            Storyboard.TargetName="FocusVisualBlack"
+                                            Storyboard.TargetProperty="Opacity"
+                                            To="1"
+                                            Duration="0"/>
+                                    </Storyboard>
+                                </VisualState>
+                                <VisualState x:Name="Unfocused" />
+                                <VisualState x:Name="PointerFocused" />
+                            </VisualStateGroup>
+                        </VisualStateManager.VisualStateGroups>
+                    </Grid>
+                </ControlTemplate>
+            </Setter.Value>
+        </Setter>
+    </Style>
+
+    <!-- Item templates -->
+
+    <!-- Grid-appropriate 250 pixel square item template as seen in the GroupedItemsPage and ItemsPage -->
+    <DataTemplate x:Key="Standard250x250ItemTemplate">
+        <Grid HorizontalAlignment="Left" Width="250" Height="250">
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel VerticalAlignment="Bottom" Background="{StaticResource ListViewItemOverlayBackgroundThemeBrush}">
+                <TextBlock Text="{Binding Title}" Foreground="{StaticResource ListViewItemOverlayForegroundThemeBrush}" Style="{StaticResource TitleTextStyle}" Height="60" Margin="15,0,15,0"/>
+                <TextBlock Text="{Binding Subtitle}" Foreground="{StaticResource ListViewItemOverlaySecondaryForegroundThemeBrush}" Style="{StaticResource CaptionTextStyle}" TextWrapping="NoWrap" Margin="15,0,15,10"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+    <!-- Grid-appropriate 500 by 130 pixel item template as seen in the GroupDetailPage -->
+    <DataTemplate x:Key="Standard500x130ItemTemplate">
+        <Grid Height="110" Width="480" Margin="10">
+            <Grid.ColumnDefinitions>
+                <ColumnDefinition Width="Auto"/>
+                <ColumnDefinition Width="*"/>
+            </Grid.ColumnDefinitions>
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Width="110" Height="110">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel Grid.Column="1" VerticalAlignment="Top" Margin="10,0,0,0">
+                <TextBlock Text="{Binding Title}" Style="{StaticResource TitleTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Subtitle}" Style="{StaticResource CaptionTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Description}" Style="{StaticResource BodyTextStyle}" MaxHeight="60"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+    <!-- List-appropriate 130 pixel high item template as seen in the SplitPage -->
+    <DataTemplate x:Key="Standard130ItemTemplate">
+        <Grid Height="110" Margin="6">
+            <Grid.ColumnDefinitions>
+                <ColumnDefinition Width="Auto"/>
+                <ColumnDefinition Width="*"/>
+            </Grid.ColumnDefinitions>
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Width="110" Height="110">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel Grid.Column="1" VerticalAlignment="Top" Margin="10,0,0,0">
+                <TextBlock Text="{Binding Title}" Style="{StaticResource TitleTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Subtitle}" Style="{StaticResource CaptionTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Description}" Style="{StaticResource BodyTextStyle}" MaxHeight="60"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+    <!--
+        List-appropriate 80 pixel high item template as seen in the SplitPage when Filled, and
+        the following pages when snapped: GroupedItemsPage, GroupDetailPage, and ItemsPage
+    -->
+    <DataTemplate x:Key="Standard80ItemTemplate">
+        <Grid Margin="6">
+            <Grid.ColumnDefinitions>
+                <ColumnDefinition Width="Auto"/>
+                <ColumnDefinition Width="*"/>
+            </Grid.ColumnDefinitions>
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Width="60" Height="60">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel Grid.Column="1" Margin="10,0,0,0">
+                <TextBlock Text="{Binding Title}" Style="{StaticResource ItemTextStyle}" MaxHeight="40"/>
+                <TextBlock Text="{Binding Subtitle}" Style="{StaticResource CaptionTextStyle}" TextWrapping="NoWrap"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+    <!-- Grid-appropriate 300 by 70 pixel item template as seen in the SearchResultsPage -->
+    <DataTemplate x:Key="StandardSmallIcon300x70ItemTemplate">
+        <Grid Width="300">
+            <Grid.ColumnDefinitions>
+                <ColumnDefinition Width="Auto"/>
+                <ColumnDefinition Width="*"/>
+            </Grid.ColumnDefinitions>
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Margin="10,10,0,20" Width="40" Height="40">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel Grid.Column="1" Margin="10,0,10,10">
+                <TextBlock Text="{Binding Title}" Style="{StaticResource BodyTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Subtitle}" Style="{StaticResource BodyTextStyle}" Foreground="{StaticResource ApplicationSecondaryForegroundThemeBrush}" Height="40"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+    <!-- List-appropriate 70 pixel high item template as seen in the SearchResultsPage when Snapped -->
+    <DataTemplate x:Key="StandardSmallIcon70ItemTemplate">
+        <Grid Margin="6">
+            <Grid.ColumnDefinitions>
+                <ColumnDefinition Width="Auto"/>
+                <ColumnDefinition Width="*"/>
+            </Grid.ColumnDefinitions>
+            <Border Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Margin="0,0,0,10" Width="40" Height="40">
+                <Image Source="{Binding Image}" Stretch="UniformToFill"/>
+            </Border>
+            <StackPanel Grid.Column="1" Margin="10,-10,0,0">
+                <TextBlock Text="{Binding Title}" Style="{StaticResource BodyTextStyle}" TextWrapping="NoWrap"/>
+                <TextBlock Text="{Binding Subtitle}" Style="{StaticResource BodyTextStyle}" Foreground="{StaticResource ApplicationSecondaryForegroundThemeBrush}" Height="40"/>
+            </StackPanel>
+        </Grid>
+    </DataTemplate>
+
+  <!--
+      190x130 pixel item template for displaying file previews as seen in the FileOpenPickerPage
+      Includes an elaborate tooltip to display title and description text
+  -->
+  <DataTemplate x:Key="StandardFileWithTooltip190x130ItemTemplate">
+        <Grid>
+            <Grid Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}">
+                <Image
+                    Source="{Binding Image}"
+                    Width="190"
+                    Height="130"
+                    HorizontalAlignment="Center"
+                    VerticalAlignment="Center"
+                    Stretch="Uniform"/>
+            </Grid>
+            <ToolTipService.Placement>Mouse</ToolTipService.Placement>
+            <ToolTipService.ToolTip>
+                <Grid Background="{StaticResource ApplicationPageBackgroundThemeBrush}">
+                    <Grid.ColumnDefinitions>
+                        <ColumnDefinition Width="Auto"/>
+                        <ColumnDefinition Width="*"/>
+                    </Grid.ColumnDefinitions>
+
+                    <Grid Background="{StaticResource ListViewItemPlaceholderBackgroundThemeBrush}" Margin="20">
+                        <Image
+                            Source="{Binding Image}"
+                            Width="160"
+                            Height="160"
+                            HorizontalAlignment="Center"
+                            VerticalAlignment="Center"
+                            Stretch="Uniform"/>
+                    </Grid>
+                    <StackPanel Width="200" Grid.Column="1" Margin="0,20,20,20">
+                        <TextBlock Text="{Binding Title}" TextWrapping="NoWrap" Style="{StaticResource BodyTextStyle}"/>
+                        <TextBlock Text="{Binding Description}" MaxHeight="140" Foreground="{StaticResource ApplicationSecondaryForegroundThemeBrush}" Style="{StaticResource BodyTextStyle}"/>
+                    </StackPanel>
+                </Grid>
+            </ToolTipService.ToolTip>
+        </Grid>
+    </DataTemplate>
+
+    <!-- Default to 10-pixel spacing between grid items (after accounting for 4-pixel insets for focus) -->
+
+    <Style TargetType="GridViewItem">
+        <Setter Property="Margin" Value="0,0,2,2" />
+    </Style>
+
+    <!-- ScrollViewer styles -->
+
+    <Style x:Key="HorizontalScrollViewerStyle" TargetType="ScrollViewer">
+        <Setter Property="HorizontalScrollBarVisibility" Value="Auto"/>
+        <Setter Property="VerticalScrollBarVisibility" Value="Disabled"/>
+        <Setter Property="ScrollViewer.HorizontalScrollMode" Value="Enabled" />
+        <Setter Property="ScrollViewer.VerticalScrollMode" Value="Disabled" />
+        <Setter Property="ScrollViewer.ZoomMode" Value="Disabled" />
+    </Style>
+
+    <Style x:Key="VerticalScrollViewerStyle" TargetType="ScrollViewer">
+        <Setter Property="HorizontalScrollBarVisibility" Value="Disabled"/>
+        <Setter Property="VerticalScrollBarVisibility" Value="Auto"/>
+        <Setter Property="ScrollViewer.HorizontalScrollMode" Value="Disabled" />
+        <Setter Property="ScrollViewer.VerticalScrollMode" Value="Enabled" />
+        <Setter Property="ScrollViewer.ZoomMode" Value="Disabled" />
+    </Style>
+
+    <!-- Page layout roots typically use entrance animations and a theme-appropriate background color -->
+
+    <Style x:Key="LayoutRootStyle" TargetType="Panel">
+        <Setter Property="Background" Value="{StaticResource ApplicationPageBackgroundThemeBrush}"/>
+        <Setter Property="ChildrenTransitions">
+            <Setter.Value>
+                <TransitionCollection>
+                    <EntranceThemeTransition/>
+                </TransitionCollection>
+            </Setter.Value>
+        </Setter>
+    </Style>
+</ResourceDictionary>
diff --git a/samples/winrt/ImageManipulations/common/suspensionmanager.cpp b/samples/winrt/ImageManipulations/common/suspensionmanager.cpp
new file mode 100644
index 0000000000..c1ecf11cfd
--- /dev/null
+++ b/samples/winrt/ImageManipulations/common/suspensionmanager.cpp
@@ -0,0 +1,481 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// SuspensionManager.cpp
+// Implementation of the SuspensionManager class
+//
+
+#include "pch.h"
+#include "SuspensionManager.h"
+
+#include <collection.h>
+#include <algorithm>
+
+using namespace SDKSample::Common;
+
+using namespace Concurrency;
+using namespace Platform;
+using namespace Platform::Collections;
+using namespace Windows::Foundation;
+using namespace Windows::Foundation::Collections;
+using namespace Windows::Storage;
+using namespace Windows::Storage::FileProperties;
+using namespace Windows::Storage::Streams;
+using namespace Windows::UI::Xaml;
+using namespace Windows::UI::Xaml::Controls;
+using namespace Windows::UI::Xaml::Interop;
+
+namespace
+{
+    Map<String^, Object^>^ _sessionState = ref new Map<String^, Object^>();
+    String^ sessionStateFilename = "_sessionState.dat";
+
+    // Forward declarations for object object read / write support
+    void WriteObject(Windows::Storage::Streams::DataWriter^ writer, Platform::Object^ object);
+    Platform::Object^ ReadObject(Windows::Storage::Streams::DataReader^ reader);
+}
+
+/// <summary>
+/// Provides access to global session state for the current session.  This state is serialized by
+/// <see cref="SaveAsync"/> and restored by <see cref="RestoreAsync"/> which require values to be
+/// one of the following: boxed values including integers, floating-point singles and doubles,
+/// wide characters, boolean, Strings and Guids, or Map<String^, Object^> where map values are
+/// subject to the same constraints.  Session state should be as compact as possible.
+/// </summary>
+IMap<String^, Object^>^ SuspensionManager::SessionState::get(void)
+{
+    return _sessionState;
+}
+
+/// <summary>
+/// Wrap a WeakReference as a reference object for use in a collection.
+/// </summary>
+private ref class WeakFrame sealed
+{
+private:
+    WeakReference _frameReference;
+
+internal:
+    WeakFrame(Frame^ frame) { _frameReference = frame; }
+    property Frame^ ResolvedFrame
+    {
+        Frame^ get(void) { return _frameReference.Resolve<Frame>(); }
+    };
+};
+
+namespace
+{
+    std::vector<WeakFrame^> _registeredFrames;
+    DependencyProperty^ FrameSessionStateKeyProperty =
+        DependencyProperty::RegisterAttached("_FrameSessionStateKeyProperty",
+        TypeName(String::typeid), TypeName(SuspensionManager::typeid), nullptr);
+    DependencyProperty^ FrameSessionStateProperty =
+        DependencyProperty::RegisterAttached("_FrameSessionStateProperty",
+        TypeName(IMap<String^, Object^>::typeid), TypeName(SuspensionManager::typeid), nullptr);
+}
+
+/// <summary>
+/// Registers a <see cref="Frame"/> instance to allow its navigation history to be saved to
+/// and restored from <see cref="SessionState"/>.  Frames should be registered once
+/// immediately after creation if they will participate in session state management.  Upon
+/// registration if state has already been restored for the specified key
+/// the navigation history will immediately be restored.  Subsequent invocations of
+/// <see cref="RestoreAsync(String)"/> will also restore navigation history.
+/// </summary>
+/// <param name="frame">An instance whose navigation history should be managed by
+/// <see cref="SuspensionManager"/></param>
+/// <param name="sessionStateKey">A unique key into <see cref="SessionState"/> used to
+/// store navigation-related information.</param>
+void SuspensionManager::RegisterFrame(Frame^ frame, String^ sessionStateKey)
+{
+    if (frame->GetValue(FrameSessionStateKeyProperty) != nullptr)
+    {
+        throw ref new FailureException("Frames can only be registered to one session state key");
+    }
+
+    if (frame->GetValue(FrameSessionStateProperty) != nullptr)
+    {
+        throw ref new FailureException("Frames must be either be registered before accessing frame session state, or not registered at all");
+    }
+
+    // Use a dependency property to associate the session key with a frame, and keep a list of frames whose
+    // navigation state should be managed
+    frame->SetValue(FrameSessionStateKeyProperty, sessionStateKey);
+    _registeredFrames.insert(_registeredFrames.begin(), ref new WeakFrame(frame));
+
+    // Check to see if navigation state can be restored
+    RestoreFrameNavigationState(frame);
+}
+
+/// <summary>
+/// Disassociates a <see cref="Frame"/> previously registered by <see cref="RegisterFrame"/>
+/// from <see cref="SessionState"/>.  Any navigation state previously captured will be
+/// removed.
+/// </summary>
+/// <param name="frame">An instance whose navigation history should no longer be
+/// managed.</param>
+void SuspensionManager::UnregisterFrame(Frame^ frame)
+{
+    // Remove session state and remove the frame from the list of frames whose navigation
+    // state will be saved (along with any weak references that are no longer reachable)
+    auto key = safe_cast<String^>(frame->GetValue(FrameSessionStateKeyProperty));
+    if (SessionState->HasKey(key)) SessionState->Remove(key);
+    _registeredFrames.erase(
+        std::remove_if(_registeredFrames.begin(), _registeredFrames.end(), [=](WeakFrame^& e)
+        {
+            auto testFrame = e->ResolvedFrame;
+            return testFrame == nullptr || testFrame == frame;
+        }),
+        _registeredFrames.end()
+    );
+}
+
+/// <summary>
+/// Provides storage for session state associated with the specified <see cref="Frame"/>.
+/// Frames that have been previously registered with <see cref="RegisterFrame"/> have
+/// their session state saved and restored automatically as a part of the global
+/// <see cref="SessionState"/>.  Frames that are not registered have transient state
+/// that can still be useful when restoring pages that have been discarded from the
+/// navigation cache.
+/// </summary>
+/// <remarks>Apps may choose to rely on <see cref="LayoutAwarePage"/> to manage
+/// page-specific state instead of working with frame session state directly.</remarks>
+/// <param name="frame">The instance for which session state is desired.</param>
+/// <returns>A collection of state subject to the same serialization mechanism as
+/// <see cref="SessionState"/>.</returns>
+IMap<String^, Object^>^ SuspensionManager::SessionStateForFrame(Frame^ frame)
+{
+    auto frameState = safe_cast<IMap<String^, Object^>^>(frame->GetValue(FrameSessionStateProperty));
+
+    if (frameState == nullptr)
+    {
+        auto frameSessionKey = safe_cast<String^>(frame->GetValue(FrameSessionStateKeyProperty));
+        if (frameSessionKey != nullptr)
+        {
+            // Registered frames reflect the corresponding session state
+            if (!_sessionState->HasKey(frameSessionKey))
+            {
+                _sessionState->Insert(frameSessionKey, ref new Map<String^, Object^>());
+            }
+            frameState = safe_cast<IMap<String^, Object^>^>(_sessionState->Lookup(frameSessionKey));
+        }
+        else
+        {
+            // Frames that aren't registered have transient state
+            frameState = ref new Map<String^, Object^>();
+        }
+        frame->SetValue(FrameSessionStateProperty, frameState);
+    }
+    return frameState;
+}
+
+void SuspensionManager::RestoreFrameNavigationState(Frame^ frame)
+{
+    auto frameState = SessionStateForFrame(frame);
+    if (frameState->HasKey("Navigation"))
+    {
+        frame->SetNavigationState(safe_cast<String^>(frameState->Lookup("Navigation")));
+    }
+}
+
+void SuspensionManager::SaveFrameNavigationState(Frame^ frame)
+{
+    auto frameState = SessionStateForFrame(frame);
+    frameState->Insert("Navigation", frame->GetNavigationState());
+}
+
+/// <summary>
+/// Save the current <see cref="SessionState"/>.  Any <see cref="Frame"/> instances
+/// registered with <see cref="RegisterFrame"/> will also preserve their current
+/// navigation stack, which in turn gives their active <see cref="Page"/> an opportunity
+/// to save its state.
+/// </summary>
+/// <returns>An asynchronous task that reflects when session state has been saved.</returns>
+task<void> SuspensionManager::SaveAsync(void)
+{
+    // Save the navigation state for all registered frames
+    for (auto&& weakFrame : _registeredFrames)
+    {
+        auto frame = weakFrame->ResolvedFrame;
+        if (frame != nullptr) SaveFrameNavigationState(frame);
+    }
+
+    // Serialize the session state synchronously to avoid asynchronous access to shared
+    // state
+    auto sessionData = ref new InMemoryRandomAccessStream();
+    auto sessionDataWriter = ref new DataWriter(sessionData->GetOutputStreamAt(0));
+    WriteObject(sessionDataWriter, _sessionState);
+
+    // Once session state has been captured synchronously, begin the asynchronous process
+    // of writing the result to disk
+    return task<unsigned int>(sessionDataWriter->StoreAsync()).then([=](unsigned int)
+    {
+        return sessionDataWriter->FlushAsync();
+    }).then([=](bool flushSucceeded)
+    {
+        (void)flushSucceeded; // Unused parameter
+        return ApplicationData::Current->LocalFolder->CreateFileAsync(sessionStateFilename,
+            CreationCollisionOption::ReplaceExisting);
+    }).then([=](StorageFile^ createdFile)
+    {
+        return createdFile->OpenAsync(FileAccessMode::ReadWrite);
+    }).then([=](IRandomAccessStream^ newStream)
+    {
+        return RandomAccessStream::CopyAndCloseAsync(
+            sessionData->GetInputStreamAt(0), newStream->GetOutputStreamAt(0));
+    }).then([=](UINT64 copiedBytes)
+    {
+        (void)copiedBytes; // Unused parameter
+        return;
+    });
+}
+
+/// <summary>
+/// Restores previously saved <see cref="SessionState"/>.  Any <see cref="Frame"/> instances
+/// registered with <see cref="RegisterFrame"/> will also restore their prior navigation
+/// state, which in turn gives their active <see cref="Page"/> an opportunity restore its
+/// state.
+/// </summary>
+/// <param name="version">A version identifer compared to the session state to prevent
+/// incompatible versions of session state from reaching app code.  Saved state with a
+/// different version will be ignored, resulting in an empty <see cref="SessionState"/>
+/// dictionary.</param>
+/// <returns>An asynchronous task that reflects when session state has been read.  The
+/// content of <see cref="SessionState"/> should not be relied upon until this task
+/// completes.</returns>
+task<void> SuspensionManager::RestoreAsync(void)
+{
+    _sessionState->Clear();
+
+    task<StorageFile^> getFileTask(ApplicationData::Current->LocalFolder->GetFileAsync(sessionStateFilename));
+    return getFileTask.then([=](StorageFile^ stateFile)
+    {
+        task<BasicProperties^> getBasicPropertiesTask(stateFile->GetBasicPropertiesAsync());
+        return getBasicPropertiesTask.then([=](BasicProperties^ stateFileProperties)
+        {
+            auto size = unsigned int(stateFileProperties->Size);
+            if (size != stateFileProperties->Size) throw ref new FailureException("Session state larger than 4GB");
+            task<IRandomAccessStreamWithContentType^> openReadTask(stateFile->OpenReadAsync());
+            return openReadTask.then([=](IRandomAccessStreamWithContentType^ stateFileStream)
+            {
+                auto stateReader = ref new DataReader(stateFileStream);
+                return task<unsigned int>(stateReader->LoadAsync(size)).then([=](unsigned int bytesRead)
+                {
+                    (void)bytesRead; // Unused parameter
+                    // Deserialize the Session State
+                    Object^ content = ReadObject(stateReader);
+                    _sessionState = (Map<String^, Object^>^)content;
+
+                    // Restore any registered frames to their saved state
+                    for (auto&& weakFrame : _registeredFrames)
+                    {
+                        auto frame = weakFrame->ResolvedFrame;
+                        if (frame != nullptr)
+                        {
+                            frame->ClearValue(FrameSessionStateProperty);
+                            RestoreFrameNavigationState(frame);
+                        }
+                    }
+                }, task_continuation_context::use_current());
+            });
+        });
+    });
+}
+
+#pragma region Object serialization for a known set of types
+
+namespace
+{
+    // Codes used for identifying serialized types
+    enum StreamTypes {
+        NullPtrType = 0,
+
+        // Supported IPropertyValue types
+        UInt8Type, UInt16Type, UInt32Type, UInt64Type, Int16Type, Int32Type, Int64Type,
+        SingleType, DoubleType, BooleanType, Char16Type, GuidType, StringType,
+
+        // Additional supported types
+        StringToObjectMapType,
+
+        // Marker values used to ensure stream integrity
+        MapEndMarker
+    };
+
+    void WriteString(DataWriter^ writer, String^ string)
+    {
+        writer->WriteByte(StringType);
+        writer->WriteUInt32(writer->MeasureString(string));
+        writer->WriteString(string);
+    }
+
+    void WriteProperty(DataWriter^ writer, IPropertyValue^ propertyValue)
+    {
+        switch (propertyValue->Type)
+        {
+        case PropertyType::UInt8:
+            writer->WriteByte(UInt8Type);
+            writer->WriteByte(propertyValue->GetUInt8());
+            return;
+        case PropertyType::UInt16:
+            writer->WriteByte(UInt16Type);
+            writer->WriteUInt16(propertyValue->GetUInt16());
+            return;
+        case PropertyType::UInt32:
+            writer->WriteByte(UInt32Type);
+            writer->WriteUInt32(propertyValue->GetUInt32());
+            return;
+        case PropertyType::UInt64:
+            writer->WriteByte(UInt64Type);
+            writer->WriteUInt64(propertyValue->GetUInt64());
+            return;
+        case PropertyType::Int16:
+            writer->WriteByte(Int16Type);
+            writer->WriteUInt16(propertyValue->GetInt16());
+            return;
+        case PropertyType::Int32:
+            writer->WriteByte(Int32Type);
+            writer->WriteUInt32(propertyValue->GetInt32());
+            return;
+        case PropertyType::Int64:
+            writer->WriteByte(Int64Type);
+            writer->WriteUInt64(propertyValue->GetInt64());
+            return;
+        case PropertyType::Single:
+            writer->WriteByte(SingleType);
+            writer->WriteSingle(propertyValue->GetSingle());
+            return;
+        case PropertyType::Double:
+            writer->WriteByte(DoubleType);
+            writer->WriteDouble(propertyValue->GetDouble());
+            return;
+        case PropertyType::Boolean:
+            writer->WriteByte(BooleanType);
+            writer->WriteBoolean(propertyValue->GetBoolean());
+            return;
+        case PropertyType::Char16:
+            writer->WriteByte(Char16Type);
+            writer->WriteUInt16(propertyValue->GetChar16());
+            return;
+        case PropertyType::Guid:
+            writer->WriteByte(GuidType);
+            writer->WriteGuid(propertyValue->GetGuid());
+            return;
+        case PropertyType::String:
+            WriteString(writer, propertyValue->GetString());
+            return;
+        default:
+            throw ref new InvalidArgumentException("Unsupported property type");
+        }
+    }
+
+    void WriteStringToObjectMap(DataWriter^ writer, IMap<String^, Object^>^ map)
+    {
+        writer->WriteByte(StringToObjectMapType);
+        writer->WriteUInt32(map->Size);
+        for (auto&& pair : map)
+        {
+            WriteObject(writer, pair->Key);
+            WriteObject(writer, pair->Value);
+        }
+        writer->WriteByte(MapEndMarker);
+    }
+
+    void WriteObject(DataWriter^ writer, Object^ object)
+    {
+        if (object == nullptr)
+        {
+            writer->WriteByte(NullPtrType);
+            return;
+        }
+
+        auto propertyObject = dynamic_cast<IPropertyValue^>(object);
+        if (propertyObject != nullptr)
+        {
+            WriteProperty(writer, propertyObject);
+            return;
+        }
+
+        auto mapObject = dynamic_cast<IMap<String^, Object^>^>(object);
+        if (mapObject != nullptr)
+        {
+            WriteStringToObjectMap(writer, mapObject);
+            return;
+        }
+
+        throw ref new InvalidArgumentException("Unsupported data type");
+    }
+
+    String^ ReadString(DataReader^ reader)
+    {
+        int length = reader->ReadUInt32();
+        String^ string = reader->ReadString(length);
+        return string;
+    }
+
+    IMap<String^, Object^>^ ReadStringToObjectMap(DataReader^ reader)
+    {
+        auto map = ref new Map<String^, Object^>();
+        auto size = reader->ReadUInt32();
+        for (unsigned int index = 0; index < size; index++)
+        {
+            auto key = safe_cast<String^>(ReadObject(reader));
+            auto value = ReadObject(reader);
+            map->Insert(key, value);
+        }
+        if (reader->ReadByte() != MapEndMarker)
+        {
+            throw ref new InvalidArgumentException("Invalid stream");
+        }
+        return map;
+    }
+
+    Object^ ReadObject(DataReader^ reader)
+    {
+        auto type = reader->ReadByte();
+        switch (type)
+        {
+        case NullPtrType:
+            return nullptr;
+        case UInt8Type:
+            return reader->ReadByte();
+        case UInt16Type:
+            return reader->ReadUInt16();
+        case UInt32Type:
+            return reader->ReadUInt32();
+        case UInt64Type:
+            return reader->ReadUInt64();
+        case Int16Type:
+            return reader->ReadInt16();
+        case Int32Type:
+            return reader->ReadInt32();
+        case Int64Type:
+            return reader->ReadInt64();
+        case SingleType:
+            return reader->ReadSingle();
+        case DoubleType:
+            return reader->ReadDouble();
+        case BooleanType:
+            return reader->ReadBoolean();
+        case Char16Type:
+            return (char16_t)reader->ReadUInt16();
+        case GuidType:
+            return reader->ReadGuid();
+        case StringType:
+            return ReadString(reader);
+        case StringToObjectMapType:
+            return ReadStringToObjectMap(reader);
+        default:
+            throw ref new InvalidArgumentException("Unsupported property type");
+        }
+    }
+}
+
+#pragma endregion
diff --git a/samples/winrt/ImageManipulations/common/suspensionmanager.h b/samples/winrt/ImageManipulations/common/suspensionmanager.h
new file mode 100644
index 0000000000..65e1180a06
--- /dev/null
+++ b/samples/winrt/ImageManipulations/common/suspensionmanager.h
@@ -0,0 +1,50 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// SuspensionManager.h
+// Declaration of the SuspensionManager class
+//
+
+#pragma once
+
+#include <ppltasks.h>
+
+namespace SDKSample
+{
+    namespace Common
+    {
+        /// <summary>
+        /// SuspensionManager captures global session state to simplify process lifetime management
+        /// for an application.  Note that session state will be automatically cleared under a variety
+        /// of conditions and should only be used to store information that would be convenient to
+        /// carry across sessions, but that should be disacarded when an application crashes or is
+        /// upgraded.
+        /// </summary>
+        ref class SuspensionManager sealed
+        {
+        internal:
+            static void RegisterFrame(Windows::UI::Xaml::Controls::Frame^ frame, Platform::String^ sessionStateKey);
+            static void UnregisterFrame(Windows::UI::Xaml::Controls::Frame^ frame);
+            static Concurrency::task<void> SaveAsync(void);
+            static Concurrency::task<void> RestoreAsync(void);
+            static property Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ SessionState
+            {
+                Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ get(void);
+            };
+            static Windows::Foundation::Collections::IMap<Platform::String^, Platform::Object^>^ SessionStateForFrame(
+                Windows::UI::Xaml::Controls::Frame^ frame);
+
+        private:
+            static void RestoreFrameNavigationState(Windows::UI::Xaml::Controls::Frame^ frame);
+            static void SaveFrameNavigationState(Windows::UI::Xaml::Controls::Frame^ frame);
+        };
+    }
+}
diff --git a/samples/winrt/ImageManipulations/pch.cpp b/samples/winrt/ImageManipulations/pch.cpp
new file mode 100644
index 0000000000..97389d94cc
--- /dev/null
+++ b/samples/winrt/ImageManipulations/pch.cpp
@@ -0,0 +1,16 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// pch.cpp
+// Include the standard header and generate the precompiled header.
+//
+
+#include "pch.h"
diff --git a/samples/winrt/ImageManipulations/pch.h b/samples/winrt/ImageManipulations/pch.h
new file mode 100644
index 0000000000..13f9bc34c7
--- /dev/null
+++ b/samples/winrt/ImageManipulations/pch.h
@@ -0,0 +1,23 @@
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+
+//
+// pch.h
+// Header for standard system include files.
+//
+
+#pragma once
+
+#include <collection.h>
+#include <ppltasks.h>
+#include <agile.h>
+#include "Common\LayoutAwarePage.h"
+#include "Common\SuspensionManager.h"
+#include "App.xaml.h"
diff --git a/samples/winrt/ImageManipulations/sample-utils/SampleTemplateStyles.xaml b/samples/winrt/ImageManipulations/sample-utils/SampleTemplateStyles.xaml
new file mode 100644
index 0000000000..cddb264012
--- /dev/null
+++ b/samples/winrt/ImageManipulations/sample-utils/SampleTemplateStyles.xaml
@@ -0,0 +1,51 @@
+<!--
+//*********************************************************
+//
+// Copyright (c) Microsoft. All rights reserved.
+// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF
+// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY
+// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR
+// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT.
+//
+//*********************************************************
+-->
+<ResourceDictionary
+    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml">
+
+    <Style x:Key="TitleTextStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI Light" />
+        <Setter Property="FontSize" Value="16" />
+    </Style>
+    <Style x:Key="HeaderTextStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI Semilight" />
+        <Setter Property="FontSize" Value="26.667" />
+        <Setter Property="Margin" Value="0,0,0,25" />
+    </Style>
+    <Style x:Key="H2Style" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI" />
+        <Setter Property="FontSize" Value="14.667" />
+        <Setter Property="Margin" Value="0,0,0,0" />
+    </Style>
+    <Style x:Key="SubheaderTextStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI Semilight" />
+        <Setter Property="FontSize" Value="14.667" />
+        <Setter Property="Margin" Value="0,0,0,5" />
+    </Style>
+    <Style x:Key="BasicTextStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI Light" />
+        <Setter Property="FontSize" Value="16" />
+    </Style>
+    <Style x:Key="SeparatorStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI" />
+        <Setter Property="FontSize" Value="9" />
+    </Style>
+    <Style x:Key="FooterStyle" TargetType="TextBlock">
+        <Setter Property="FontFamily" Value="Segoe UI" />
+        <Setter Property="FontSize" Value="12" />
+        <Setter Property="Margin" Value="0,8,0,0" />
+    </Style>
+    <Style x:Key="HyperlinkStyle" TargetType="HyperlinkButton">
+        <Setter Property="Padding" Value="5"/>
+    </Style>
+</ResourceDictionary>